From 8e4a95a93d15a6707a29454cd47e10b08314cda2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 24 May 2007 10:41:34 +0100 Subject: First version of the softpipe rasterizer. This will flesh out to hopefully acheive 3 things: - New software rasterizer for mesa - New driver model for mesa & dri with much smaller drivers - The basis for the cell driver. It's got a long way to go yet, but will at least run trivial/tri.c. --- src/mesa/softpipe/Makefile | 2 + src/mesa/softpipe/generic/Makefile | 3 + src/mesa/softpipe/generic/g_context.c | 135 ++++ src/mesa/softpipe/generic/g_context.h | 149 +++++ src/mesa/softpipe/generic/g_draw.c | 696 +++++++++++++++++++++ src/mesa/softpipe/generic/g_draw.h | 56 ++ src/mesa/softpipe/generic/g_headers.h | 115 ++++ src/mesa/softpipe/generic/g_prim.h | 90 +++ src/mesa/softpipe/generic/g_prim_clip.c | 429 +++++++++++++ src/mesa/softpipe/generic/g_prim_cull.c | 119 ++++ src/mesa/softpipe/generic/g_prim_flatshade.c | 149 +++++ src/mesa/softpipe/generic/g_prim_offset.c | 157 +++++ src/mesa/softpipe/generic/g_prim_setup.c | 568 +++++++++++++++++ src/mesa/softpipe/generic/g_prim_setup.h | 121 ++++ src/mesa/softpipe/generic/g_prim_twoside.c | 149 +++++ src/mesa/softpipe/generic/g_prim_unfilled.c | 165 +++++ src/mesa/softpipe/generic/g_state.h | 61 ++ src/mesa/softpipe/generic/g_state_clip.c | 69 ++ src/mesa/softpipe/generic/g_state_derived.c | 136 ++++ src/mesa/softpipe/generic/g_state_fs.c | 50 ++ src/mesa/softpipe/generic/g_state_setup.c | 122 ++++ src/mesa/softpipe/generic/g_state_surface.c | 53 ++ src/mesa/softpipe/generic/g_surface.c | 153 +++++ src/mesa/softpipe/generic/g_surface.h | 115 ++++ src/mesa/softpipe/generic/g_tile.h | 42 ++ src/mesa/softpipe/generic/g_tile_fs.c | 188 ++++++ src/mesa/softpipe/generic/g_tile_output.c | 92 +++ src/mesa/softpipe/sp_context.h | 72 +++ src/mesa/softpipe/sp_defines.h | 45 ++ src/mesa/softpipe/sp_state.h | 132 ++++ src/mesa/softpipe/state_tracker/Makefile | 2 + src/mesa/softpipe/state_tracker/st_atom.c | 168 +++++ src/mesa/softpipe/state_tracker/st_atom.h | 55 ++ src/mesa/softpipe/state_tracker/st_atom_cbuf.c | 72 +++ src/mesa/softpipe/state_tracker/st_atom_clip.c | 75 +++ src/mesa/softpipe/state_tracker/st_atom_fs.c | 56 ++ src/mesa/softpipe/state_tracker/st_atom_setup.c | 175 ++++++ src/mesa/softpipe/state_tracker/st_atom_viewport.c | 117 ++++ src/mesa/softpipe/state_tracker/st_cb_program.c | 165 +++++ src/mesa/softpipe/state_tracker/st_context.c | 76 +++ src/mesa/softpipe/state_tracker/st_context.h | 109 ++++ src/mesa/softpipe/state_tracker/st_draw.c | 107 ++++ src/mesa/softpipe/state_tracker/st_draw.h | 40 ++ src/mesa/softpipe/state_tracker/st_program.h | 68 ++ src/mesa/softpipe/state_tracker/st_public.h | 43 ++ 45 files changed, 5761 insertions(+) create mode 100644 src/mesa/softpipe/Makefile create mode 100644 src/mesa/softpipe/generic/Makefile create mode 100644 src/mesa/softpipe/generic/g_context.c create mode 100644 src/mesa/softpipe/generic/g_context.h create mode 100644 src/mesa/softpipe/generic/g_draw.c create mode 100644 src/mesa/softpipe/generic/g_draw.h create mode 100644 src/mesa/softpipe/generic/g_headers.h create mode 100644 src/mesa/softpipe/generic/g_prim.h create mode 100644 src/mesa/softpipe/generic/g_prim_clip.c create mode 100644 src/mesa/softpipe/generic/g_prim_cull.c create mode 100644 src/mesa/softpipe/generic/g_prim_flatshade.c create mode 100644 src/mesa/softpipe/generic/g_prim_offset.c create mode 100644 src/mesa/softpipe/generic/g_prim_setup.c create mode 100644 src/mesa/softpipe/generic/g_prim_setup.h create mode 100644 src/mesa/softpipe/generic/g_prim_twoside.c create mode 100644 src/mesa/softpipe/generic/g_prim_unfilled.c create mode 100644 src/mesa/softpipe/generic/g_state.h create mode 100644 src/mesa/softpipe/generic/g_state_clip.c create mode 100644 src/mesa/softpipe/generic/g_state_derived.c create mode 100644 src/mesa/softpipe/generic/g_state_fs.c create mode 100644 src/mesa/softpipe/generic/g_state_setup.c create mode 100644 src/mesa/softpipe/generic/g_state_surface.c create mode 100644 src/mesa/softpipe/generic/g_surface.c create mode 100644 src/mesa/softpipe/generic/g_surface.h create mode 100644 src/mesa/softpipe/generic/g_tile.h create mode 100644 src/mesa/softpipe/generic/g_tile_fs.c create mode 100644 src/mesa/softpipe/generic/g_tile_output.c create mode 100644 src/mesa/softpipe/sp_context.h create mode 100644 src/mesa/softpipe/sp_defines.h create mode 100644 src/mesa/softpipe/sp_state.h create mode 100644 src/mesa/softpipe/state_tracker/Makefile create mode 100644 src/mesa/softpipe/state_tracker/st_atom.c create mode 100644 src/mesa/softpipe/state_tracker/st_atom.h create mode 100644 src/mesa/softpipe/state_tracker/st_atom_cbuf.c create mode 100644 src/mesa/softpipe/state_tracker/st_atom_clip.c create mode 100644 src/mesa/softpipe/state_tracker/st_atom_fs.c create mode 100644 src/mesa/softpipe/state_tracker/st_atom_setup.c create mode 100644 src/mesa/softpipe/state_tracker/st_atom_viewport.c create mode 100644 src/mesa/softpipe/state_tracker/st_cb_program.c create mode 100644 src/mesa/softpipe/state_tracker/st_context.c create mode 100644 src/mesa/softpipe/state_tracker/st_context.h create mode 100644 src/mesa/softpipe/state_tracker/st_draw.c create mode 100644 src/mesa/softpipe/state_tracker/st_draw.h create mode 100644 src/mesa/softpipe/state_tracker/st_program.h create mode 100644 src/mesa/softpipe/state_tracker/st_public.h (limited to 'src') diff --git a/src/mesa/softpipe/Makefile b/src/mesa/softpipe/Makefile new file mode 100644 index 0000000000..451911a354 --- /dev/null +++ b/src/mesa/softpipe/Makefile @@ -0,0 +1,2 @@ +default: + cd .. ; make diff --git a/src/mesa/softpipe/generic/Makefile b/src/mesa/softpipe/generic/Makefile new file mode 100644 index 0000000000..12a8bd0409 --- /dev/null +++ b/src/mesa/softpipe/generic/Makefile @@ -0,0 +1,3 @@ +default: + cd ../.. ; make + diff --git a/src/mesa/softpipe/generic/g_context.c b/src/mesa/softpipe/generic/g_context.c new file mode 100644 index 0000000000..c3329b5c7d --- /dev/null +++ b/src/mesa/softpipe/generic/g_context.c @@ -0,0 +1,135 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "tnl/t_context.h" +#include "vf/vf.h" + +#include "g_context.h" +#include "g_prim.h" +#include "g_state.h" +#include "g_draw.h" + +static void generic_destroy( struct softpipe_context *softpipe ) +{ + struct generic_context *generic = generic_context( softpipe ); + + draw_destroy( generic->draw ); + + FREE( generic ); +} + + +static void generic_draw_vb( struct softpipe_context *softpipe, + struct vertex_buffer *VB ) +{ + struct generic_context *generic = generic_context( softpipe ); + + if (generic->dirty) + generic_update_derived( generic ); + + draw_vb( generic->draw, VB ); +} + +struct softpipe_context *generic_create( void ) +{ + struct generic_context *generic = CALLOC_STRUCT(generic_context); + + generic->softpipe.destroy = generic_destroy; + generic->softpipe.set_clip_state = generic_set_clip_state; + generic->softpipe.set_viewport = generic_set_viewport; + generic->softpipe.set_setup_state = generic_set_setup_state; + generic->softpipe.set_scissor_rect = generic_set_scissor_rect; + generic->softpipe.set_fs_state = generic_set_fs_state; + generic->softpipe.set_polygon_stipple = generic_set_polygon_stipple; + generic->softpipe.set_cbuf_state = generic_set_cbuf_state; + generic->softpipe.draw_vb = generic_draw_vb; + + + + generic->prim.setup = prim_setup( generic ); + generic->prim.unfilled = prim_unfilled( generic ); + generic->prim.twoside = prim_twoside( generic ); + generic->prim.offset = prim_offset( generic ); + generic->prim.clip = prim_clip( generic ); + generic->prim.flatshade = prim_flatshade( generic ); + generic->prim.cull = prim_cull( generic ); + + + generic->draw = draw_create( generic ); + + ASSIGN_4V( generic->plane[0], -1, 0, 0, 1 ); + ASSIGN_4V( generic->plane[1], 1, 0, 0, 1 ); + ASSIGN_4V( generic->plane[2], 0, -1, 0, 1 ); + ASSIGN_4V( generic->plane[3], 0, 1, 0, 1 ); + ASSIGN_4V( generic->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ + ASSIGN_4V( generic->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ + generic->nr_planes = 6; + + return &generic->softpipe; +} + + + + + + +#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) + +void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ) +{ + stage->nr_tmps = nr; + + if (nr) { + GLubyte *store = MALLOC(MAX_VERTEX_SIZE * nr); + GLuint i; + + stage->tmp = MALLOC(sizeof(struct vertex_header *) * nr); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } +} + +void prim_free_tmps( struct prim_stage *stage ) +{ + if (stage->tmp) { + FREE(stage->tmp[0]); + FREE(stage->tmp); + } +} + + + + + diff --git a/src/mesa/softpipe/generic/g_context.h b/src/mesa/softpipe/generic/g_context.h new file mode 100644 index 0000000000..24d467f2c3 --- /dev/null +++ b/src/mesa/softpipe/generic/g_context.h @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_CONTEXT_H +#define G_CONTEXT_H + +#include "glheader.h" + +#include "softpipe/sp_state.h" +#include "softpipe/sp_context.h" + + + +struct generic_surface; +struct draw_context; +struct prim_stage; + + +enum interp_mode { + INTERP_CONSTANT, + INTERP_LINEAR, + INTERP_PERSPECTIVE +}; + + +#define G_NEW_VIEWPORT 0x1 +#define G_NEW_SETUP 0x2 +#define G_NEW_FS 0x4 +#define G_NEW_BLEND 0x8 +#define G_NEW_CBUF 0x10 +#define G_NEW_CLIP 0x20 +#define G_NEW_SCISSOR 0x40 +#define G_NEW_STIPPLE 0x80 + + + +struct generic_context { + struct softpipe_context softpipe; + + + /* The most recent drawing state as set by the driver: + */ + struct softpipe_viewport viewport; + struct softpipe_setup_state setup; + struct softpipe_fs_state fs; + struct softpipe_blend_state blend; + struct softpipe_surface cbuf; + struct softpipe_clip_state clip; + struct softpipe_scissor_rect scissor; + struct softpipe_poly_stipple poly_stipple; + GLuint dirty; + + + /* Cbuf derived state??? + */ + struct generic_surface *cbuf_surface; + + /* Clip derived state: + */ + GLfloat plane[12][4]; + GLuint nr_planes; + + /* Setup derived state. TODO: this should be passed in the program + * tokens as parameters to DECL instructions. + * + * For now we just set colors to CONST on flatshade, textures to + * perspective always and everything else to linear. + */ + enum interp_mode interp[VF_ATTRIB_MAX]; + + + /* FS + setup derived state: + */ + GLuint fp_attr_to_slot[VF_ATTRIB_MAX]; + GLuint vf_attr_to_slot[VF_ATTRIB_MAX]; + GLuint nr_attrs; + GLuint nr_frag_attrs; + GLuint attr_mask; + + GLboolean need_z; + GLboolean need_w; + + /* Stipple derived state: + */ + GLubyte stipple_masks[16][16]; + + + /* The software clipper/setup engine. + */ + struct { + struct prim_stage *setup; + struct prim_stage *unfilled; + struct prim_stage *twoside; + struct prim_stage *clip; + struct prim_stage *flatshade; + struct prim_stage *offset; + struct prim_stage *cull; + + struct prim_stage *first; + + GLenum prim; + GLuint vertex_size; + } prim; + + /* Temp kludge: + */ + struct draw_context *draw; +}; + + + + +static INLINE struct generic_context * +generic_context( struct softpipe_context *softpipe ) +{ + return (struct generic_context *)softpipe; +} + + + + +#endif diff --git a/src/mesa/softpipe/generic/g_draw.c b/src/mesa/softpipe/generic/g_draw.c new file mode 100644 index 0000000000..4e094c21a0 --- /dev/null +++ b/src/mesa/softpipe/generic/g_draw.c @@ -0,0 +1,696 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "imports.h" + +#include "tnl/t_context.h" + +#include "g_context.h" +#include "g_prim.h" +#include "g_headers.h" +#include "g_draw.h" + +/* This file is a temporary set of hooks to allow us to use the tnl/ + * and vf/ modules until we have replacements in softpipe. + */ + + +struct draw_context +{ + struct generic_context *generic; + + struct vf_attr_map attrs[VF_ATTRIB_MAX]; + GLuint nr_attrs; + GLuint vertex_size; + struct vertex_fetch *vf; + + GLubyte *verts; + GLuint nr_vertices; + GLboolean in_vb; + + GLenum prim; + + /* Helper for tnl: + */ + GLvector4f header; +}; + + +static struct vertex_header *get_vertex( struct draw_context *pipe, + GLuint i ) +{ + return (struct vertex_header *)(pipe->verts + i * pipe->vertex_size); +} + + + +static void draw_allocate_vertices( struct draw_context *draw, + GLuint nr_vertices ) +{ + draw->nr_vertices = nr_vertices; + draw->verts = MALLOC( nr_vertices * draw->vertex_size ); + + draw->generic->prim.first->begin( draw->generic->prim.first ); +} + +static void draw_set_prim( struct draw_context *draw, + GLenum prim ) +{ + draw->prim = prim; + + /* Not done yet - need to force edgeflags to 1 in strip/fan + * primitives. + */ +#if 0 + switch (prim) { + case GL_TRIANGLES: + case GL_POLYGON: + case GL_QUADS: + case GL_QUAD_STRIP: /* yes, we need this */ + respect_edgeflags( pipe, GL_TRUE ); + break; + + default: + respect_edgeflags( pipe, GL_FALSE ); + break; + } +#endif +} + + + +static void do_quad( struct prim_stage *first, + struct vertex_header *v0, + struct vertex_header *v1, + struct vertex_header *v2, + struct vertex_header *v3 ) +{ + struct prim_header prim; + + { + GLuint tmp = v1->edgeflag; + v1->edgeflag = 0; + + prim.v[0] = v0; + prim.v[1] = v1; + prim.v[2] = v3; + first->tri( first, &prim ); + + v1->edgeflag = tmp; + } + + { + GLuint tmp = v3->edgeflag; + v3->edgeflag = 0; + + prim.v[0] = v1; + prim.v[1] = v2; + prim.v[2] = v3; + first->tri( first, &prim ); + + v3->edgeflag = tmp; + } +} + + + + +static void draw_indexed_prim( struct draw_context *draw, + const GLuint *elts, + GLuint count ) +{ + struct prim_stage * const first = draw->generic->prim.first; + struct prim_header prim; + GLuint i; + + prim.det = 0; /* valid from cull stage onwards */ + prim.v[0] = 0; + prim.v[1] = 0; + prim.v[2] = 0; + + switch (draw->prim) { + case GL_POINTS: + for (i = 0; i < count; i ++) { + prim.v[0] = get_vertex( draw, elts[i] ); + + first->point( first, &prim ); + } + break; + + case GL_LINES: + for (i = 0; i+1 < count; i += 2) { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + + first->line( first, &prim ); + } + break; + + case GL_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + prim.v[0] = get_vertex( draw, elts[i-1] ); + prim.v[1] = get_vertex( draw, elts[i] ); + first->line( first, &prim ); + } + + prim.v[0] = get_vertex( draw, elts[count-1] ); + prim.v[1] = get_vertex( draw, elts[0] ); + first->line( first, &prim ); + } + break; + + case GL_LINE_STRIP: + /* I'm guessing it will be necessary to have something like a + * render->reset_line_stipple() method to properly support + * splitting strips into primitives like this. Alternately we + * could just scan ahead to find individual clipped lines and + * otherwise leave the strip intact - that might be better, but + * require more complex code here. + */ + if (count >= 2) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, elts[0] ); + + for (i = 1; i < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, elts[i] ); + + first->line( first, &prim ); + } + } + break; + + case GL_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + prim.v[0] = get_vertex( draw, elts[i + 1] ); + prim.v[1] = get_vertex( draw, elts[i + 0] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + } + else { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + } + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_FAN: + if (count >= 3) { + prim.v[0] = get_vertex( draw, elts[0] ); + prim.v[1] = 0; + prim.v[2] = get_vertex( draw, elts[1] ); + + for (i = 0; i+2 < count; i++) { + prim.v[1] = prim.v[2]; + prim.v[2] = get_vertex( draw, elts[i+2] ); + + first->tri( first, &prim ); + } + } + break; + + case GL_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( first, + get_vertex( draw, elts[i + 0] ), + get_vertex( draw, elts[i + 1] ), + get_vertex( draw, elts[i + 2] ), + get_vertex( draw, elts[i + 3] )); + } + break; + + case GL_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( first, + get_vertex( draw, elts[i + 2] ), + get_vertex( draw, elts[i + 0] ), + get_vertex( draw, elts[i + 1] ), + get_vertex( draw, elts[i + 3] )); + } + break; + + + case GL_POLYGON: + if (count >= 3) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, elts[1] ); + prim.v[2] = get_vertex( draw, elts[0] ); + + for (i = 0; i+2 < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, elts[i+2] ); + + first->tri( first, &prim ); + } + } + break; + + default: + assert(0); + break; + } +} + +static void draw_prim( struct draw_context *draw, + GLuint start, + GLuint count ) +{ + struct prim_stage * const first = draw->generic->prim.first; + struct prim_header prim; + GLuint i; + +// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); + + prim.det = 0; /* valid from cull stage onwards */ + prim.v[0] = 0; + prim.v[1] = 0; + prim.v[2] = 0; + + switch (draw->prim) { + case GL_POINTS: + for (i = 0; i < count; i ++) { + prim.v[0] = get_vertex( draw, start + i ); + first->point( first, &prim ); + } + break; + + case GL_LINES: + for (i = 0; i+1 < count; i += 2) { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + + first->line( first, &prim ); + } + break; + + case GL_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + prim.v[0] = get_vertex( draw, start + i - 1 ); + prim.v[1] = get_vertex( draw, start + i ); + first->line( first, &prim ); + } + + prim.v[0] = get_vertex( draw, start + count - 1 ); + prim.v[1] = get_vertex( draw, start + 0 ); + first->line( first, &prim ); + } + break; + + case GL_LINE_STRIP: + if (count >= 2) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, start + 0 ); + + for (i = 1; i < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, start + i ); + + first->line( first, &prim ); + } + } + break; + + case GL_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + prim.v[0] = get_vertex( draw, start + i + 1 ); + prim.v[1] = get_vertex( draw, start + i + 0 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + } + else { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + } + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_FAN: + if (count >= 3) { + prim.v[0] = get_vertex( draw, start + 0 ); + prim.v[1] = 0; + prim.v[2] = get_vertex( draw, start + 1 ); + + for (i = 0; i+2 < count; i++) { + prim.v[1] = prim.v[2]; + prim.v[2] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + } + break; + + + case GL_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( first, + get_vertex( draw, start + i + 0 ), + get_vertex( draw, start + i + 1 ), + get_vertex( draw, start + i + 2 ), + get_vertex( draw, start + i + 3 )); + } + break; + + case GL_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( first, + get_vertex( draw, start + i + 2 ), + get_vertex( draw, start + i + 0 ), + get_vertex( draw, start + i + 1 ), + get_vertex( draw, start + i + 3 )); + } + break; + + case GL_POLYGON: + if (count >= 3) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, start + 1 ); + prim.v[2] = get_vertex( draw, start + 0 ); + + for (i = 0; i+2 < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + } + break; + + default: + assert(0); + break; + } +} + + +static void draw_release_vertices( struct draw_context *draw ) +{ + draw->generic->prim.first->end( draw->generic->prim.first ); + + FREE(draw->verts); + draw->verts = NULL; +} + + +struct header_dword { + GLuint clipmask:12; + GLuint edgeflag:1; + GLuint pad:19; +}; + + +static void +build_vertex_headers( struct draw_context *draw, + struct vertex_buffer *VB ) +{ + if (draw->header.storage == NULL) { + draw->header.stride = sizeof(GLfloat); + draw->header.size = 1; + draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(GLfloat), 32 ); + draw->header.data = draw->header.storage; + draw->header.count = 0; + draw->header.flags = VEC_SIZE_1 | VEC_MALLOC; + } + + /* Build vertex header attribute. + * + */ + + { + GLuint i; + struct header_dword *header = (struct header_dword *)draw->header.storage; + + /* yes its a hack + */ + assert(sizeof(*header) == sizeof(GLfloat)); + + draw->header.count = VB->Count; + + if (VB->EdgeFlag) { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = VB->ClipMask[i]; + header[i].edgeflag = VB->EdgeFlag[i]; + header[i].pad = 0; + } + } + else if (VB->ClipOrMask) { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = VB->ClipMask[i]; + header[i].edgeflag = 0; + header[i].pad = 0; + } + } + else { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = 0; + header[i].edgeflag = 0; + header[i].pad = 0; + } + } + } + + VB->AttribPtr[VF_ATTRIB_VERTEX_HEADER] = &draw->header; +} + + + + + +static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) +{ + switch (mode) { + case GL_POINTS: + *first = 1; + *incr = 1; + return 0; + case GL_LINES: + *first = 2; + *incr = 2; + return 0; + case GL_LINE_STRIP: + *first = 2; + *incr = 1; + return 0; + case GL_LINE_LOOP: + *first = 2; + *incr = 1; + return 1; + case GL_TRIANGLES: + *first = 3; + *incr = 3; + return 0; + case GL_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + return 0; + case GL_TRIANGLE_FAN: + case GL_POLYGON: + *first = 3; + *incr = 1; + return 1; + case GL_QUADS: + *first = 4; + *incr = 4; + return 0; + case GL_QUAD_STRIP: + *first = 4; + *incr = 2; + return 0; + default: + assert(0); + *first = 1; + *incr = 1; + return 0; + } +} + + +static GLuint trim( GLuint count, GLuint first, GLuint incr ) +{ + if (count < first) + return 0; + else + return count - (count - first) % incr; +} + + +/* This is a hack & will all go away. + */ +void draw_vb(struct draw_context *draw, + struct vertex_buffer *VB ) +{ + GLuint i; + + VB->AttribPtr[VF_ATTRIB_POS] = VB->NdcPtr; + VB->AttribPtr[VF_ATTRIB_BFC0] = VB->ColorPtr[1]; + VB->AttribPtr[VF_ATTRIB_BFC1] = VB->SecondaryColorPtr[1]; + VB->AttribPtr[VF_ATTRIB_CLIP_POS] = VB->ClipPtr; + + /* Build vertex headers: + */ + build_vertex_headers( draw, VB ); + + draw->in_vb = 1; + + /* Allocate the vertices: + */ + draw_allocate_vertices( draw, VB->Count ); + + /* Bind the vb outputs: + */ + vf_set_sources( draw->vf, VB->AttribPtr, 0 ); + + /* Build the hardware or prim-pipe vertices: + */ + vf_emit_vertices( draw->vf, VB->Count, draw->verts ); + + + for (i = 0; i < VB->PrimitiveCount; i++) { + + GLenum mode = VB->Primitive[i].mode; + GLuint start = VB->Primitive[i].start; + GLuint length, first, incr; + + /* Trim the primitive down to a legal size. + */ + draw_prim_info( mode, &first, &incr ); + length = trim( VB->Primitive[i].count, first, incr ); + + if (!length) + continue; + + if (draw->prim != mode) + draw_set_prim( draw, mode ); + + if (VB->Elts) { + draw_indexed_prim( draw, + VB->Elts + start, + length ); + } + else { + draw_prim( draw, + start, + length ); + } + } + + draw_release_vertices( draw ); + draw->verts = NULL; + draw->in_vb = 0; +} + +void draw_set_viewport( struct draw_context *draw, + const GLfloat *scale, + const GLfloat *translate ) +{ + assert(!draw->in_vb); + vf_set_vp_scale_translate( draw->vf, scale, translate ); +} + + + +struct draw_context *draw_create( struct generic_context *generic ) +{ + struct draw_context *draw = CALLOC_STRUCT( draw_context ); + draw->generic = generic; + draw->vf = vf_create( GL_TRUE ); + + return draw; +} + + +void draw_destroy( struct draw_context *draw ) +{ + if (draw->header.storage) + ALIGN_FREE( draw->header.storage ); + + vf_destroy( draw->vf ); + + FREE( draw ); +} + +#define EMIT_ATTR( ATTR, STYLE ) \ +do { \ + draw->attrs[draw->nr_attrs].attrib = ATTR; \ + draw->attrs[draw->nr_attrs].format = STYLE; \ + draw->nr_attrs++; \ +} while (0) + + +void draw_set_vertex_attributes( struct draw_context *draw, + const GLuint *attrs, + GLuint nr_attrs ) +{ + GLuint i; + + draw->nr_attrs = 0; + + EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F); + EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F); + + assert(attrs[0] == VF_ATTRIB_POS); + EMIT_ATTR(attrs[0], EMIT_4F_VIEWPORT); + + for (i = 1; i < nr_attrs; i++) + EMIT_ATTR(attrs[i], EMIT_4F); + + draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); +} + + diff --git a/src/mesa/softpipe/generic/g_draw.h b/src/mesa/softpipe/generic/g_draw.h new file mode 100644 index 0000000000..78c93b4df7 --- /dev/null +++ b/src/mesa/softpipe/generic/g_draw.h @@ -0,0 +1,56 @@ + +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_DRAW_H +#define G_DRAW_H + +#include "glheader.h" +#include "softpipe/sp_state.h" + + +struct draw_context; + +struct draw_context *draw_create( struct generic_context *generic ); + +void draw_destroy( struct draw_context *draw ); + +void draw_set_viewport( struct draw_context *draw, + const GLfloat *scale, + const GLfloat *translate ); + +void draw_set_vertex_attributes( struct draw_context *draw, + const GLuint *attrs, + GLuint nr_attrs ); + +void draw_vb(struct draw_context *draw, + struct vertex_buffer *VB ); + +#endif diff --git a/src/mesa/softpipe/generic/g_headers.h b/src/mesa/softpipe/generic/g_headers.h new file mode 100644 index 0000000000..1db325a4ab --- /dev/null +++ b/src/mesa/softpipe/generic/g_headers.h @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#ifndef G_HEADERS_H +#define G_HEADERS_H + +#define PRIM_POINT 1 +#define PRIM_LINE 2 +#define PRIM_TRI 3 + +struct prim_header { + GLfloat det; + struct vertex_header *v[3]; +}; + +/* Carry some useful information around with the vertices in the prim + * pipe. + */ +struct vertex_header { + GLuint clipmask:12; + GLuint edgeflag:1; + GLuint pad:19; + + GLfloat clip[4]; + + GLfloat data[][4]; /* Note variable size */ +}; + + + + + +/* The rasterizer generates 2x2 quads of fragment and feeds them to + * the current fp_machine (see below). + */ +#define QUAD_BOTTOM_LEFT 0 +#define QUAD_BOTTOM_RIGHT 1 +#define QUAD_TOP_LEFT 2 +#define QUAD_TOP_RIGHT 3 +#define QUAD_SIZE (2*2) + +#define MASK_BOTTOM_LEFT 0x1 +#define MASK_BOTTOM_RIGHT 0x2 +#define MASK_TOP_LEFT 0x4 +#define MASK_TOP_RIGHT 0x8 +#define MASK_ALL 0xf + + +#define NUM_CHANNELS 4 /* avoid confusion between 4 pixels and 4 channels */ + + +struct setup_coefficient { + GLfloat a0[NUM_CHANNELS]; /* in an xyzw layout */ + GLfloat dadx[NUM_CHANNELS]; + GLfloat dady[NUM_CHANNELS]; +}; + + + +/* Encodes everything we need to know about a 2x2 pixel block. Uses + * "Channel-Serial" or "SoA" layout. + * + * Will expand to include non-attribute things like AA coverage and + * maybe prefetched depth from the depth buffer. + */ +struct quad_header { + GLint x0; + GLint y0; + GLuint mask; + + struct { + GLfloat color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ + GLfloat depth[QUAD_SIZE]; + } outputs; + + const struct setup_coefficient *coef; + + const enum interp_mode *interp; /* XXX: this information should be + * encoded in fragment program DECL + * statements. */ + + GLuint nr_attrs; +}; + + + + + +#endif diff --git a/src/mesa/softpipe/generic/g_prim.h b/src/mesa/softpipe/generic/g_prim.h new file mode 100644 index 0000000000..e3b2c5ea2f --- /dev/null +++ b/src/mesa/softpipe/generic/g_prim.h @@ -0,0 +1,90 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_PRIM_H +#define G_PRIM_H + +#include "glheader.h" +#include "g_headers.h" + +struct generic_context; + +struct prim_stage *prim_setup( struct generic_context *context ); +struct prim_stage *prim_unfilled( struct generic_context *context ); +struct prim_stage *prim_twoside( struct generic_context *context ); +struct prim_stage *prim_offset( struct generic_context *context ); +struct prim_stage *prim_clip( struct generic_context *context ); +struct prim_stage *prim_flatshade( struct generic_context *context ); +struct prim_stage *prim_cull( struct generic_context *context ); + + +/* Internal structs and helpers for the primitive clip/setup pipeline: + */ +struct prim_stage { + struct generic_context *generic; + + struct prim_stage *next; + + struct vertex_header **tmp; + GLuint nr_tmps; + + void (*begin)( struct prim_stage * ); + + void (*point)( struct prim_stage *, + struct prim_header * ); + + void (*line)( struct prim_stage *, + struct prim_header * ); + + void (*tri)( struct prim_stage *, + struct prim_header * ); + + void (*end)( struct prim_stage * ); +}; + + + +/* Get a writeable copy of a vertex: + */ +static INLINE struct vertex_header * +dup_vert( struct prim_stage *stage, + const struct vertex_header *vert, + GLuint idx ) +{ + struct vertex_header *tmp = stage->tmp[idx]; + memcpy(tmp, vert, stage->generic->prim.vertex_size ); + return tmp; +} + +void prim_free_tmps( struct prim_stage *stage ); +void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); + + +#endif diff --git a/src/mesa/softpipe/generic/g_prim_clip.c b/src/mesa/softpipe/generic/g_prim_clip.c new file mode 100644 index 0000000000..742def9c7d --- /dev/null +++ b/src/mesa/softpipe/generic/g_prim_clip.c @@ -0,0 +1,429 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "g_context.h" +#include "g_prim.h" + +struct clipper { + struct prim_stage stage; + + GLuint active_user_planes; +}; + +/* This is a bit confusing: + */ +static INLINE struct clipper *clipper_stage( struct prim_stage *stage ) +{ + return (struct clipper *)stage; +} + + +#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) + + +/* All attributes are float[4], so this is easy: + */ +static void interp_attr( GLfloat *fdst, + GLfloat t, + const GLfloat *fin, + const GLfloat *fout ) +{ + fdst[0] = LINTERP( t, fout[0], fin[0] ); + fdst[1] = LINTERP( t, fout[1], fin[1] ); + fdst[2] = LINTERP( t, fout[2], fin[2] ); + fdst[3] = LINTERP( t, fout[3], fin[3] ); +} + + + + +/* Interpolate between two vertices to produce a third. + */ +static void interp( struct clipper *clip, + struct vertex_header *dst, + GLfloat t, + const struct vertex_header *out, + const struct vertex_header *in ) +{ + const GLuint nr_attrs = clip->stage.generic->nr_attrs; + GLuint j; + + /* Vertex header. + */ + { + dst->clipmask = 0; + dst->edgeflag = 0; + dst->pad = 0; + } + + /* Clip coordinates: interpolate normally + */ + { + interp_attr(dst->clip, t, in->clip, out->clip); + } + + /* Do the projective divide and insert window coordinates: + */ + { + const GLfloat *pos = dst->clip; + const GLfloat *scale = clip->stage.generic->viewport.scale; + const GLfloat *trans = clip->stage.generic->viewport.translate; + GLfloat oow; + + oow = 1.0 / pos[3]; + + dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; + dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; + dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; + dst->data[0][3] = oow; + } + + + /* Other attributes + */ + for (j = 1; j < nr_attrs-1; j++) { + interp_attr(dst->data[j], t, in->data[j], out->data[j]); + } +} + + +#define CLIP_USER_BIT 0x40 +#define CLIP_CULL_BIT 0x80 + + +static INLINE GLfloat dot4( const GLfloat *a, + const GLfloat *b ) +{ + GLfloat result = (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); + + return result; +} + + +#if 0 +static INLINE void do_tri( struct prim_stage *next, + struct prim_header *header ) +{ + GLuint i; + for (i = 0; i < 3; i++) { + GLfloat *ndc = header->v[i]->data[0]; + _mesa_printf("ndc %f %f %f\n", ndc[0], ndc[1], ndc[2]); + assert(ndc[0] >= -1 && ndc[0] <= 641); + assert(ndc[1] >= 30 && ndc[1] <= 481); + } + _mesa_printf("\n"); + next->tri(next, header); +} +#endif + + +static void emit_poly( struct prim_stage *stage, + struct vertex_header **inlist, + GLuint n ) +{ + struct prim_header header; + GLuint i; + + for (i = 2; i < n; i++) { + header.v[0] = inlist[0]; + header.v[1] = inlist[i-1]; + header.v[2] = inlist[i]; + + { + GLuint tmp0 = header.v[0]->edgeflag; + GLuint tmp2 = header.v[2]->edgeflag; + + if (i != 2) header.v[0]->edgeflag = 0; + if (i != n-1) header.v[2]->edgeflag = 0; + + stage->next->tri( stage->next, &header ); + + header.v[0]->edgeflag = tmp0; + header.v[2]->edgeflag = tmp2; + } + } +} + + +#if 0 +static void emit_poly( struct prim_stage *stage ) +{ + GLuint i; + + for (i = 2; i < n; i++) { + header->v[0] = inlist[0]; + header->v[1] = inlist[i-1]; + header->v[2] = inlist[i]; + + stage->next->tri( stage->next, header ); + } +} +#endif + + +/* Clip a triangle against the viewport and user clip planes. + */ +static void +do_clip_tri( struct prim_stage *stage, + struct prim_header *header, + GLuint clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *a[MAX_CLIPPED_VERTICES]; + struct vertex_header *b[MAX_CLIPPED_VERTICES]; + struct vertex_header **inlist = a; + struct vertex_header **outlist = b; + GLuint tmpnr = 0; + GLuint n = 3; + GLuint i; + + inlist[0] = header->v[0]; + inlist[1] = header->v[1]; + inlist[2] = header->v[2]; + + /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove + * this once we correctly use 16bit masks for userclip planes. + */ + clipmask &= ~CLIP_CULL_BIT; + if (clipmask & CLIP_USER_BIT) { + clipmask &= ~CLIP_USER_BIT; + clipmask |= clipper->active_user_planes; + } + + while (clipmask && n >= 3) { + GLuint plane_idx = ffs(clipmask)-1; + const GLfloat *plane = clipper->stage.generic->plane[plane_idx]; + struct vertex_header *vert_prev = inlist[0]; + GLfloat dp_prev = dot4( vert_prev->clip, plane ); + GLuint outcount = 0; + + clipmask &= ~(1<clip, plane ); + + if (!IS_NEGATIVE(dp_prev)) { + outlist[outcount++] = vert_prev; + } + + if (DIFFERENT_SIGNS(dp, dp_prev)) { + struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; + outlist[outcount++] = new_vert; + + if (IS_NEGATIVE(dp)) { + /* Going out of bounds. Avoid division by zero as we + * know dp != dp_prev from DIFFERENT_SIGNS, above. + */ + GLfloat t = dp / (dp - dp_prev); + interp( clipper, new_vert, t, vert, vert_prev ); + + /* Force edgeflag true in this case: + */ + new_vert->edgeflag = 1; + } else { + /* Coming back in. + */ + GLfloat t = dp_prev / (dp_prev - dp); + interp( clipper, new_vert, t, vert_prev, vert ); + + /* Copy starting vert's edgeflag: + */ + new_vert->edgeflag = vert_prev->edgeflag; + } + } + + vert_prev = vert; + dp_prev = dp; + } + + { + struct vertex_header **tmp = inlist; + inlist = outlist; + outlist = tmp; + n = outcount; + } + } + + /* Emit the polygon as triangles to the setup stage: + */ + if (n >= 3) + emit_poly( stage, inlist, n ); +} + + +/* Clip a line against the viewport and user clip planes. + */ +static void +do_clip_line( struct prim_stage *stage, + struct prim_header *header, + GLuint clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const GLfloat *pos0 = v0->clip; + const GLfloat *pos1 = v1->clip; + GLfloat t0 = 0; + GLfloat t1 = 0; + + /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove + * this once we correctly use 16bit masks for userclip planes. + */ + clipmask &= ~CLIP_CULL_BIT; + if (clipmask & CLIP_USER_BIT) { + clipmask &= ~CLIP_USER_BIT; + clipmask |= clipper->active_user_planes; + } + + while (clipmask) { + GLuint plane_idx = ffs(clipmask)-1; + const GLfloat *plane = clipper->stage.generic->plane[plane_idx]; + + clipmask &= ~(1< t1) t1 = t; + } + + if (dp0 < 0) { + GLfloat t = dp0 / (dp0 - dp1); + if (t > t0) t0 = t; + } + + if (t0 + t1 >= 1.0) + return; /* discard */ + } + + if (v0->clipmask) { + interp( clipper, stage->tmp[0], t0, v0, v1 ); + header->v[0] = stage->tmp[0]; + } + + if (v1->clipmask) { + interp( clipper, stage->tmp[1], t1, v1, v0 ); + header->v[1] = stage->tmp[1]; + } + + stage->next->line( stage->next, header ); +} + + + +static void clip_begin( struct prim_stage *stage ) +{ + struct clipper *clipper = clipper_stage(stage); + GLuint nr = stage->generic->nr_planes; + + /* Hacky bitmask to use when we hit CLIP_USER_BIT: + */ + clipper->active_user_planes = ((1<next->begin( stage->next ); +} + +static void +clip_point( struct prim_stage *stage, + struct prim_header *header ) +{ + if (header->v[0]->clipmask == 0) + stage->next->point( stage->next, header ); +} + + +static void +clip_line( struct prim_stage *stage, + struct prim_header *header ) +{ + GLuint clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask); + + if (clipmask == 0) { + stage->next->line( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask) == 0) { + do_clip_line(stage, header, clipmask); + } +} + + +static void +clip_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + GLuint clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask | + header->v[2]->clipmask); + + if (clipmask == 0) { + stage->next->tri( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask & + header->v[2]->clipmask) == 0) { + do_clip_tri(stage, header, clipmask); + } +} + +static void clip_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + + +struct prim_stage *prim_clip( struct generic_context *generic ) +{ + struct clipper *clipper = CALLOC_STRUCT(clipper); + + prim_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); + + clipper->stage.generic = generic; + clipper->stage.begin = clip_begin; + clipper->stage.point = clip_point; + clipper->stage.line = clip_line; + clipper->stage.tri = clip_tri; + clipper->stage.end = clip_end; + + return &clipper->stage; +} diff --git a/src/mesa/softpipe/generic/g_prim_cull.c b/src/mesa/softpipe/generic/g_prim_cull.c new file mode 100644 index 0000000000..633fbcc492 --- /dev/null +++ b/src/mesa/softpipe/generic/g_prim_cull.c @@ -0,0 +1,119 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_prim.h" + + + +struct cull_stage { + struct prim_stage stage; + GLuint mode; +}; + + + +static INLINE struct cull_stage *cull_stage( struct prim_stage *stage ) +{ + return (struct cull_stage *)stage; +} + + +static void cull_begin( struct prim_stage *stage ) +{ + struct cull_stage *cull = cull_stage(stage); + + cull->mode = stage->generic->setup.cull_mode; + + stage->next->begin( stage->next ); +} + + +static void cull_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + /* Window coords: */ + GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); + GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); + GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); + + GLfloat ex = v0[0] - v2[0]; + GLfloat ey = v0[1] - v2[1]; + GLfloat fx = v1[0] - v2[0]; + GLfloat fy = v1[1] - v2[1]; + + header->det = ex * fy - ey * fx; + + _mesa_printf("%s %f\n", __FUNCTION__, header->det ); + + if (header->det != 0) { + GLuint mode = (header->det < 0) ? WINDING_CW : WINDING_CCW; + + if ((mode & cull_stage(stage)->mode) == 0) + stage->next->tri( stage->next, header ); + } +} + + +static void cull_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void cull_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void cull_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_cull( struct generic_context *generic ) +{ + struct cull_stage *cull = CALLOC_STRUCT(cull_stage); + + prim_alloc_tmps( &cull->stage, 0 ); + + cull->stage.generic = generic; + cull->stage.next = NULL; + cull->stage.begin = cull_begin; + cull->stage.point = cull_point; + cull->stage.line = cull_line; + cull->stage.tri = cull_tri; + cull->stage.end = cull_end; + + return &cull->stage; +} diff --git a/src/mesa/softpipe/generic/g_prim_flatshade.c b/src/mesa/softpipe/generic/g_prim_flatshade.c new file mode 100644 index 0000000000..1974a64fa8 --- /dev/null +++ b/src/mesa/softpipe/generic/g_prim_flatshade.c @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_prim.h" + + +struct flatshade_stage { + struct prim_stage stage; + + const GLuint *lookup; +}; + + + +static INLINE struct flatshade_stage *flatshade_stage( struct prim_stage *stage ) +{ + return (struct flatshade_stage *)stage; +} + + +static void flatshade_begin( struct prim_stage *stage ) +{ + stage->next->begin( stage->next ); +} + + + +static INLINE void copy_attr( GLuint attr, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + if (attr) { + memcpy( dst->data[attr], + src->data[attr], + sizeof(src->data[0]) ); + } +} + +static void copy_colors( struct prim_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + struct flatshade_stage *flatshade = flatshade_stage(stage); + const GLuint *lookup = flatshade->lookup; + + copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); + copy_attr( lookup[VF_ATTRIB_COLOR1], dst, src ); + copy_attr( lookup[VF_ATTRIB_BFC0], dst, src ); + copy_attr( lookup[VF_ATTRIB_BFC1], dst, src ); +} + + + +/* Flatshade tri. Required for clipping and when unfilled tris are + * active, otherwise handled by hardware. + */ +static void flatshade_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = header->v[2]; + + copy_colors(stage, tmp.v[0], tmp.v[2]); + copy_colors(stage, tmp.v[1], tmp.v[2]); + + stage->next->tri( stage->next, &tmp ); +} + + +/* Flatshade line. Required for clipping. + */ +static void flatshade_line( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = header->v[1]; + + copy_colors(stage, tmp.v[0], tmp.v[1]); + + stage->next->line( stage->next, &tmp ); +} + + +static void flatshade_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void flatshade_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_flatshade( struct generic_context *generic ) +{ + struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); + + prim_alloc_tmps( &flatshade->stage, 2 ); + + flatshade->stage.generic = generic; + flatshade->stage.next = NULL; + flatshade->stage.begin = flatshade_begin; + flatshade->stage.point = flatshade_point; + flatshade->stage.line = flatshade_line; + flatshade->stage.tri = flatshade_tri; + flatshade->stage.end = flatshade_end; + + flatshade->lookup = generic->vf_attr_to_slot; + + return &flatshade->stage; +} + + diff --git a/src/mesa/softpipe/generic/g_prim_offset.c b/src/mesa/softpipe/generic/g_prim_offset.c new file mode 100644 index 0000000000..801ce9a441 --- /dev/null +++ b/src/mesa/softpipe/generic/g_prim_offset.c @@ -0,0 +1,157 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" +#include "macros.h" + +#include "g_context.h" +#include "g_prim.h" + + + +struct offset_stage { + struct prim_stage stage; + + GLuint hw_data_offset; + + GLfloat scale; + GLfloat units; +}; + + + +static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) +{ + return (struct offset_stage *)stage; +} + + +static void offset_begin( struct prim_stage *stage ) +{ + struct offset_stage *offset = offset_stage(stage); + + offset->units = stage->generic->setup.offset_units; + offset->scale = stage->generic->setup.offset_scale; + + stage->next->begin( stage->next ); +} + + +/* Offset tri. Some hardware can handle this, but not usually when + * doing unfilled rendering. + */ +static void do_offset_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + GLfloat inv_det = 1.0 / header->det; + + /* Window coords: + */ + GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); + GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); + GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); + + GLfloat ex = v0[0] - v2[2]; + GLfloat fx = v1[0] - v2[2]; + GLfloat ey = v0[1] - v2[2]; + GLfloat fy = v1[1] - v2[2]; + GLfloat ez = v0[2] - v2[2]; + GLfloat fz = v1[2] - v2[2]; + + GLfloat a = ey*fz - ez*fy; + GLfloat b = ez*fx - ex*fz; + + GLfloat ac = a * inv_det; + GLfloat bc = b * inv_det; + GLfloat zoffset; + + if ( ac < 0.0f ) ac = -ac; + if ( bc < 0.0f ) bc = -bc; + + zoffset = offset->units + MAX2( ac, bc ) * offset->scale; + + v0[2] += zoffset; + v1[2] += zoffset; + v2[2] += zoffset; + + stage->next->tri( stage->next, header ); +} + + +static void offset_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = dup_vert(stage, header->v[2], 2); + + do_offset_tri( stage->next, &tmp ); +} + + + +static void offset_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void offset_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void offset_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_offset( struct generic_context *generic ) +{ + struct offset_stage *offset = CALLOC_STRUCT(offset_stage); + + prim_alloc_tmps( &offset->stage, 3 ); + + offset->stage.generic = generic; + offset->stage.next = NULL; + offset->stage.begin = offset_begin; + offset->stage.point = offset_point; + offset->stage.line = offset_line; + offset->stage.tri = offset_tri; + offset->stage.end = offset_end; + + return &offset->stage; +} diff --git a/src/mesa/softpipe/generic/g_prim_setup.c b/src/mesa/softpipe/generic/g_prim_setup.c new file mode 100644 index 0000000000..3e9639f7bc --- /dev/null +++ b/src/mesa/softpipe/generic/g_prim_setup.c @@ -0,0 +1,568 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "g_context.h" +#include "g_prim.h" +#include "g_tile.h" + + +struct edge { + GLfloat dx; /* X(v1) - X(v0), used only during setup */ + GLfloat dy; /* Y(v1) - Y(v0), used only during setup */ + GLfloat dxdy; /* dx/dy */ + GLfloat sx; /* first sample point x coord */ + GLfloat sy; + GLint lines; /* number of lines on this edge */ +}; + + +struct setup_stage { + struct prim_stage stage; + + /* Vertices are just an array of floats making up each attribute in + * turn. Currently fixed at 4 floats, but should change in time. + * Codegen will help cope with this. + */ + const struct vertex_header *vmax; + const struct vertex_header *vmid; + const struct vertex_header *vmin; + const struct vertex_header *vprovoke; + + struct edge ebot; + struct edge etop; + struct edge emaj; + + GLfloat oneoverarea; + + struct setup_coefficient coef[FRAG_ATTRIB_MAX]; + struct quad_header quad; + + struct { + GLint left[2]; + GLint right[2]; + GLint y; + GLuint y_flags; + GLuint mask; + } span; + +}; + + + + +static inline struct setup_stage *setup_stage( struct prim_stage *stage ) +{ + return (struct setup_stage *)stage; +} + + + +static inline GLint _min(GLint x, GLint y) +{ + return x < y ? x : y; +} + +static inline GLint _max(GLint x, GLint y) +{ + return x > y ? x : y; +} + +static inline GLint block( GLint x ) +{ + return x & ~1; +} + + + +static void setup_begin( struct prim_stage *stage ) +{ + setup_stage(stage)->quad.nr_attrs = stage->generic->nr_frag_attrs; +} + + + +static void run_shader_block( struct setup_stage *setup, + GLint x, GLint y, GLuint mask ) +{ + setup->quad.x0 = x; + setup->quad.y0 = y; + setup->quad.mask = mask; + + quad_shade( setup->stage.generic, &setup->quad ); +} + + +/* this is pretty nasty... may need to rework flush_spans again to + * fix it, if possible. + */ +static GLuint calculate_mask( struct setup_stage *setup, + GLint x ) +{ + GLuint mask = 0; + + if (x >= setup->span.left[0] && x < setup->span.right[0]) + mask |= MASK_BOTTOM_LEFT; + + if (x >= setup->span.left[1] && x < setup->span.right[1]) + mask |= MASK_TOP_LEFT; + + if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) + mask |= MASK_BOTTOM_RIGHT; + + if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) + mask |= MASK_TOP_RIGHT; + + return mask; +} + + +static void flush_spans( struct setup_stage *setup ) +{ + GLint minleft, maxright; + GLint x; + + switch (setup->span.y_flags) { + case 3: + minleft = _min(setup->span.left[0], + setup->span.left[1]); + maxright = _max(setup->span.right[0], + setup->span.right[1]); + break; + + case 1: + minleft = setup->span.left[0]; + maxright = setup->span.right[0]; + break; + + case 2: + minleft = setup->span.left[1]; + maxright = setup->span.right[1]; + break; + + default: + return; + } + + + for (x = block(minleft); x <= block(maxright); ) + { + run_shader_block( setup, x, + setup->span.y, + calculate_mask( setup, x ) ); + x += 2; + } + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +} + + + +static void +setup_point( struct prim_stage *stage, + struct prim_header *header ) +{ +} + + +static void +setup_line( struct prim_stage *stage, + struct prim_header *header ) +{ +} + + + + + + +static GLboolean setup_sort_vertices( struct setup_stage *setup, + const struct prim_header *prim ) +{ + + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + const struct vertex_header *v2 = prim->v[2]; + + setup->vprovoke = v2; +// setup->oneoverarea = -1.0 / prim->det; + + /* determine bottom to top order of vertices */ + { + GLfloat y0 = v0->data[0][1]; + GLfloat y1 = v1->data[0][1]; + GLfloat y2 = v2->data[0][1]; + if (y0 <= y1) { + if (y1 <= y2) { + /* y0<=y1<=y2 */ + setup->vmin = v0; + setup->vmid = v1; + setup->vmax = v2; + } + else if (y2 <= y0) { + /* y2<=y0<=y1 */ + setup->vmin = v2; + setup->vmid = v0; + setup->vmax = v1; + } + else { + /* y0<=y2<=y1 */ + setup->vmin = v0; + setup->vmid = v2; + setup->vmax = v1; +// setup->oneoverarea = -setup->oneoverarea; + } + } + else { + if (y0 <= y2) { + /* y1<=y0<=y2 */ + setup->vmin = v1; + setup->vmid = v0; + setup->vmax = v2; +// setup->oneoverarea = -setup->oneoverarea; + } + else if (y2 <= y1) { + /* y2<=y1<=y0 */ + setup->vmin = v2; + setup->vmid = v1; + setup->vmax = v0; +// setup->oneoverarea = -setup->oneoverarea; + } + else { + /* y1<=y2<=y0 */ + setup->vmin = v1; + setup->vmid = v2; + setup->vmax = v0; + } + } + } + + setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; + setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; + setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; + + /* xxx: may need to adjust this sign according to the if-tree + * above: + * + * XXX: this is like 'det', but calculated from screen coords?? + */ + { + const GLfloat area = (setup->emaj.dx * setup->ebot.dy - + setup->ebot.dx * setup->emaj.dy); + + setup->oneoverarea = 1.0 / area; + } + + + _mesa_printf("%s one-over-area %f\n", __FUNCTION__, setup->oneoverarea ); + + + return GL_TRUE; +} + + +static void const_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i ) +{ + setup->coef[slot].dadx[i] = 0; + setup->coef[slot].dady[i] = 0; + + /* need provoking vertex info! + */ + setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; +} + + +static void linear_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i) +{ + GLfloat botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; + GLfloat majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; + GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); + + _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", + slot, "xyzw"[i], + setup->coef[slot].a0[i], + setup->coef[slot].dadx[i], + setup->coef[slot].dady[i]); +} + + +static void persp_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i ) +{ + /* premultiply by 1/w: + */ + GLfloat mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; + GLfloat mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; + GLfloat maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; + + GLfloat botda = mida - mina; + GLfloat majda = maxa - mina; + GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; + GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + setup->coef[slot].a0[i] = (mina - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); +} + + + + +static void setup_coefficients( struct setup_stage *setup ) +{ + const enum interp_mode *interp = setup->stage.generic->interp; + GLuint slot, j; + + /* z and w are done by linear interpolation: + */ + linear_coeff(setup, 0, 2); + linear_coeff(setup, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + switch (interp[slot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + linear_coeff(setup, slot, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + persp_coeff(setup, slot, j); + break; + } + } +} + + + +static void setup_edges( struct setup_stage *setup ) +{ + GLfloat vmin_x = setup->vmin->data[0][0] + 0.5; + GLfloat vmid_x = setup->vmid->data[0][0] + 0.5; + + GLfloat vmin_y = setup->vmin->data[0][1] - 0.5; + GLfloat vmid_y = setup->vmid->data[0][1] - 0.5; + GLfloat vmax_y = setup->vmax->data[0][1] - 0.5; + + setup->emaj.sy = ceilf(vmin_y); + setup->emaj.lines = (GLint) ceilf(vmax_y - setup->emaj.sy); + setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; + setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; + + setup->etop.sy = ceilf(vmid_y); + setup->etop.lines = (GLint) ceilf(vmax_y - setup->etop.sy); + setup->etop.dxdy = setup->etop.dx / setup->etop.dy; + setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; + + setup->ebot.sy = ceilf(vmin_y); + setup->ebot.lines = (GLint) ceilf(vmid_y - setup->ebot.sy); + setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; + setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; +} + + + + + + +static void subtriangle( struct setup_stage *setup, + struct edge *eleft, + struct edge *eright, + GLuint lines ) +{ + GLint y, start_y, finish_y; + GLint sy = (GLint)eleft->sy; + + assert((GLint)eleft->sy == (GLint) eright->sy); + assert((GLint)eleft->sy >= 0); /* catch bug in x64? */ + + /* scissor y: + */ + if (setup->stage.generic->setup.scissor) { + start_y = sy; + finish_y = start_y + lines; + + if (start_y < setup->stage.generic->scissor.miny) + start_y = setup->stage.generic->scissor.miny; + + if (finish_y > setup->stage.generic->scissor.maxy) + finish_y = setup->stage.generic->scissor.maxy; + + start_y -= sy; + finish_y -= sy; + } + else { + start_y = 0; + finish_y = lines; + } + + _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + + for (y = start_y; y < finish_y; y++) { + + /* avoid accumulating adds as floats don't have the precision to + * accurately iterate large triangle edges that way. luckily we + * can just multiply these days. + * + * this is all drowned out by the attribute interpolation anyway. + */ + GLint left = (GLint)(eleft->sx + y * eleft->dxdy); + GLint right = (GLint)(eright->sx + y * eright->dxdy); + + /* scissor x: + */ + if (setup->stage.generic->setup.scissor) { + if (left < setup->stage.generic->scissor.minx) + left = setup->stage.generic->scissor.minx; + + if (right > setup->stage.generic->scissor.maxx) + right = setup->stage.generic->scissor.maxx; + } + + if (left < right) { + GLint _y = sy+y; + if (block(_y) != setup->span.y) { + flush_spans(setup); + setup->span.y = block(_y); + } + + setup->span.left[_y&1] = left; + setup->span.right[_y&1] = right; + setup->span.y_flags |= 1<<(_y&1); + } + } + + + /* save the values so that emaj can be restarted: + */ + eleft->sx += lines * eleft->dxdy; + eright->sx += lines * eright->dxdy; + eleft->sy += lines; + eright->sy += lines; +} + + +static void setup_tri( struct prim_stage *stage, + struct prim_header *prim ) +{ + struct setup_stage *setup = setup_stage( stage ); + + _mesa_printf("%s\n", __FUNCTION__ ); + + setup_sort_vertices( setup, prim ); + setup_coefficients( setup ); + setup_edges( setup ); + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +// setup->span.z_mode = tri_z_mode( setup->ctx ); + +// init_constant_attribs( setup ); + + if (setup->oneoverarea < 0.0) { + /* emaj on left: + */ + subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); + subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); + } + else { + /* emaj on right: + */ + subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); + subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); + } + + flush_spans( setup ); +} + +static void setup_end( struct prim_stage *stage ) +{ +} + + +struct prim_stage *prim_setup( struct generic_context *generic ) +{ + struct setup_stage *setup = CALLOC_STRUCT(setup_stage); + + setup->stage.generic = generic; + setup->stage.begin = setup_begin; + setup->stage.point = setup_point; + setup->stage.line = setup_line; + setup->stage.tri = setup_tri; + setup->stage.end = setup_end; + + setup->quad.coef = setup->coef; + + return &setup->stage; +} diff --git a/src/mesa/softpipe/generic/g_prim_setup.h b/src/mesa/softpipe/generic/g_prim_setup.h new file mode 100644 index 0000000000..40a70c543e --- /dev/null +++ b/src/mesa/softpipe/generic/g_prim_setup.h @@ -0,0 +1,121 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef S_TRI_H +#define S_TRI_H + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "imports.h" +#include "s_tri_public.h" +#include "s_context.h" + + + +struct tri_context; +struct fp_context; +struct be_context; + +/* Note the rasterizer does not take a GLcontext argument. This is + * deliberate. + */ +struct tri_context *tri_create_context( GLcontext *ctx ); + +void tri_destroy_context( struct tri_context *tri ); + +void tri_set_fp_context( struct tri_context *tri, + struct fp_context *fp, + void (*fp_run)( struct fp_context *fp, + const struct fp_inputs *, + struct fp_outputs * )); + + +void tri_set_be_context( struct tri_context *tri, + struct be_context *be, + void (*be_run)( struct be_context *be, + const struct fp_outputs * )); + +void tri_set_attribs( struct tri_context *tri, + const struct attr_info *info, + GLuint nr_attrib ); + +void tri_set_backface( struct tri_context *tri, + GLfloat backface ); + +void tri_set_scissor( struct tri_context *tri, + GLint x, + GLint y, + GLuint width, + GLuint height, + GLboolean enabled ); + +void tri_set_stipple( struct tri_context *tri, + const GLuint *pattern, + GLboolean enabled ); + +/* Unfilled triangles will be handled elsewhere (higher in the + * pipeline), as will things like stipple (lower in the pipeline). + */ + +void tri_triangle( struct tri_context *tri, + const struct vertex *v0, + const struct vertex *v1, + const struct vertex *v2 ); + +/* TODO: rasterize_line, rasterize_point?? + * How will linestipple work? + */ + + +#ifdef SETUP_PRIVATE + + + + + + +GLboolean tri_setup( struct tri_context *tri, + const struct vertex *v0, + const struct vertex *v1, + const struct vertex *v2 ); + +void tri_rasterize( struct tri_context *tri ); +void tri_rasterize_spans( struct tri_context *tri ); + + + + + + +#endif +#endif diff --git a/src/mesa/softpipe/generic/g_prim_twoside.c b/src/mesa/softpipe/generic/g_prim_twoside.c new file mode 100644 index 0000000000..334ae2f537 --- /dev/null +++ b/src/mesa/softpipe/generic/g_prim_twoside.c @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" +#include "g_context.h" +#include "g_prim.h" + + +struct twoside_stage { + struct prim_stage stage; + + GLfloat facing; + const GLuint *lookup; +}; + + +static INLINE struct twoside_stage *twoside_stage( struct prim_stage *stage ) +{ + return (struct twoside_stage *)stage; +} + + +static void twoside_begin( struct prim_stage *stage ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + twoside->facing = (stage->generic->setup.front_winding == WINDING_CW) ? -1 : 1; + + stage->next->begin( stage->next ); +} + + +static INLINE void copy_color( GLuint attr_dst, + GLuint attr_src, + struct vertex_header *v ) +{ + if (attr_dst && attr_src) { + memcpy( v->data[attr_dst], + v->data[attr_src], + sizeof(v->data[0]) ); + } +} + + +static struct vertex_header *copy_bfc( struct twoside_stage *twoside, + const struct vertex_header *v, + GLuint idx ) +{ + struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); + + copy_color( twoside->lookup[VF_ATTRIB_COLOR0], + twoside->lookup[VF_ATTRIB_BFC0], + tmp ); + + copy_color( twoside->lookup[VF_ATTRIB_COLOR1], + twoside->lookup[VF_ATTRIB_BFC1], + tmp ); + + return tmp; +} + + +/* Twoside tri: + */ +static void twoside_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + if (header->det * twoside->facing < 0) { + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = copy_bfc(twoside, header->v[0], 0); + tmp.v[1] = copy_bfc(twoside, header->v[1], 1); + tmp.v[2] = copy_bfc(twoside, header->v[2], 2); + + stage->next->tri( stage->next, &tmp ); + } + else { + stage->next->tri( stage->next, header ); + } +} + + +static void twoside_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void twoside_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void twoside_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + + + +struct prim_stage *prim_twoside( struct generic_context *generic ) +{ + struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); + + prim_alloc_tmps( &twoside->stage, 3 ); + + twoside->stage.generic = generic; + twoside->stage.next = NULL; + twoside->stage.begin = twoside_begin; + twoside->stage.point = twoside_point; + twoside->stage.line = twoside_line; + twoside->stage.tri = twoside_tri; + twoside->stage.end = twoside_end; + + twoside->lookup = generic->vf_attr_to_slot; + + return &twoside->stage; +} diff --git a/src/mesa/softpipe/generic/g_prim_unfilled.c b/src/mesa/softpipe/generic/g_prim_unfilled.c new file mode 100644 index 0000000000..7c6ee79f92 --- /dev/null +++ b/src/mesa/softpipe/generic/g_prim_unfilled.c @@ -0,0 +1,165 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_prim.h" + + + +struct unfilled_stage { + struct prim_stage stage; + + GLuint mode[2]; +}; + + +static INLINE struct unfilled_stage *unfilled_stage( struct prim_stage *stage ) +{ + return (struct unfilled_stage *)stage; +} + + +static void unfilled_begin( struct prim_stage *stage ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + + unfilled->mode[0] = stage->generic->setup.fill_ccw; + unfilled->mode[1] = stage->generic->setup.fill_cw; + + stage->next->begin( stage->next ); +} + +static void point( struct prim_stage *stage, + struct vertex_header *v0 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + stage->next->point( stage->next, &tmp ); +} + +static void line( struct prim_stage *stage, + struct vertex_header *v0, + struct vertex_header *v1 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + tmp.v[1] = v1; + stage->next->line( stage->next, &tmp ); +} + + +static void points( struct prim_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (v0->edgeflag) point( stage, v0 ); + if (v1->edgeflag) point( stage, v1 ); + if (v2->edgeflag) point( stage, v2 ); +} + +static void lines( struct prim_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (v0->edgeflag) line( stage, v0, v1 ); + if (v1->edgeflag) line( stage, v1, v2 ); + if (v2->edgeflag) line( stage, v2, v0 ); +} + + +/* Unfilled tri: + * + * Note edgeflags in the vertex struct is not sufficient as we will + * need to manipulate them when decomposing primitives??? + */ +static void unfilled_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + GLuint mode = unfilled->mode[header->det < 0]; + + switch (mode) { + case FILL_TRI: + stage->next->tri( stage->next, header ); + break; + + case FILL_LINE: + lines( stage, header ); + break; + + case GL_POINT: + points( stage, header ); + break; + } +} + +static void unfilled_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void unfilled_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void unfilled_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_unfilled( struct generic_context *generic ) +{ + struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); + + prim_alloc_tmps( &unfilled->stage, 0 ); + + unfilled->stage.generic = generic; + unfilled->stage.next = NULL; + unfilled->stage.tmp = NULL; + unfilled->stage.begin = unfilled_begin; + unfilled->stage.point = unfilled_point; + unfilled->stage.line = unfilled_line; + unfilled->stage.tri = unfilled_tri; + unfilled->stage.end = unfilled_end; + + return &unfilled->stage; +} diff --git a/src/mesa/softpipe/generic/g_state.h b/src/mesa/softpipe/generic/g_state.h new file mode 100644 index 0000000000..5077d37ab7 --- /dev/null +++ b/src/mesa/softpipe/generic/g_state.h @@ -0,0 +1,61 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef PRIM_H +#define PRIM_H + +#include "glheader.h" +#include "softpipe/sp_state.h" + + +void generic_set_clip_state( struct softpipe_context *, + const struct softpipe_clip_state * ); + +void generic_set_viewport( struct softpipe_context *, + const struct softpipe_viewport * ); + +void generic_set_setup_state( struct softpipe_context *, + const struct softpipe_setup_state * ); + +void generic_set_scissor_rect( struct softpipe_context *, + const struct softpipe_scissor_rect * ); + +void generic_set_fs_state( struct softpipe_context *, + const struct softpipe_fs_state * ); + +void generic_set_polygon_stipple( struct softpipe_context *, + const struct softpipe_poly_stipple * ); + +void generic_set_cbuf_state( struct softpipe_context *, + const struct softpipe_surface * ); + +void generic_update_derived( struct generic_context *generic ); + +#endif diff --git a/src/mesa/softpipe/generic/g_state_clip.c b/src/mesa/softpipe/generic/g_state_clip.c new file mode 100644 index 0000000000..b90034e11e --- /dev/null +++ b/src/mesa/softpipe/generic/g_state_clip.c @@ -0,0 +1,69 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_state.h" +#include "g_draw.h" + + + +void generic_set_clip_state( struct softpipe_context *softpipe, + const struct softpipe_clip_state *clip ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy(&generic->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); + + generic->nr_planes = 6 + clip->nr; + generic->dirty |= G_NEW_CLIP; +} + + + +/* Called when driver state tracker notices changes to the viewport + * matrix: + */ +void generic_set_viewport( struct softpipe_context *softpipe, + const struct softpipe_viewport *viewport ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy(&generic->viewport, viewport, sizeof(viewport)); + + /* Using tnl/ and vf/ modules is temporary while getting started. + * Full softpipe will have vertex shader, vertex fetch of its own. + */ + draw_set_viewport( generic->draw, viewport->scale, viewport->translate ); + generic->dirty |= G_NEW_VIEWPORT; +} + + + diff --git a/src/mesa/softpipe/generic/g_state_derived.c b/src/mesa/softpipe/generic/g_state_derived.c new file mode 100644 index 0000000000..678825afe1 --- /dev/null +++ b/src/mesa/softpipe/generic/g_state_derived.c @@ -0,0 +1,136 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "macros.h" +#include "enums.h" +#include "program.h" + +#include "g_context.h" +#include "g_draw.h" +#include "g_state.h" + +#define EMIT_ATTR( ATTR, FRAG_ATTR, INTERP ) \ +do { \ + slot_to_vf_attr[generic->nr_attrs] = ATTR; \ + generic->vf_attr_to_slot[ATTR] = generic->nr_attrs; \ + generic->fp_attr_to_slot[FRAG_ATTR] = generic->nr_attrs; \ + generic->interp[generic->nr_attrs] = INTERP; \ + generic->nr_attrs++; \ + attr_mask |= (1<fs.fp; + const GLuint inputsRead = fp->Base.InputsRead; + GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; + GLuint attr_mask = 0; + GLuint i; + + generic->nr_attrs = 0; + memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); + + memset(generic->fp_attr_to_slot, 0, sizeof(generic->vf_attr_to_slot)); + memset(generic->vf_attr_to_slot, 0, sizeof(generic->fp_attr_to_slot)); + + /* TODO - Figure out if we need to do perspective divide, etc. + */ + EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); + + /* Pull in the rest of the attributes. They are all in float4 + * format. Future optimizations could be to keep some attributes + * as fixed point or ubyte format. + */ + for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { + if (inputsRead & (i << i)) { + EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); + } + } + + for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { + if (inputsRead & (i << i)) { + EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); + } + } + + generic->nr_frag_attrs = generic->nr_attrs; + + /* Additional attributes required for setup: Just twosided + * lighting. Edgeflag is dealt with specially by setting bits in + * the vertex header. + */ + if (generic->setup.light_twoside) { + if (inputsRead & FRAG_BIT_COL0) { + EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ + } + + if (inputsRead & FRAG_BIT_COL1) { + EMIT_ATTR(VF_ATTRIB_BFC1, FRAG_ATTRIB_MAX, 0); /* XXX: discard after setup */ + } + } + + if (attr_mask != generic->attr_mask) { + generic->attr_mask = attr_mask; + + draw_set_vertex_attributes( generic->draw, + slot_to_vf_attr, + generic->nr_attrs ); + } +} + + +/* Hopefully this will remain quite simple, otherwise need to pull in + * something like the state tracker mechanism. + */ +void generic_update_derived( struct generic_context *generic ) +{ + if (generic->dirty & (G_NEW_SETUP | G_NEW_FS)) + calculate_vertex_layout( generic ); + + generic->dirty = 0; +} diff --git a/src/mesa/softpipe/generic/g_state_fs.c b/src/mesa/softpipe/generic/g_state_fs.c new file mode 100644 index 0000000000..8a430c0c65 --- /dev/null +++ b/src/mesa/softpipe/generic/g_state_fs.c @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "g_context.h" +#include "g_state.h" + + + +void generic_set_fs_state( struct softpipe_context *softpipe, + const struct softpipe_fs_state *fs ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy(&generic->fs, fs, sizeof(*fs)); + + generic->dirty |= G_NEW_FS; +} + + + + + + + + + diff --git a/src/mesa/softpipe/generic/g_state_setup.c b/src/mesa/softpipe/generic/g_state_setup.c new file mode 100644 index 0000000000..2a5176c5a1 --- /dev/null +++ b/src/mesa/softpipe/generic/g_state_setup.c @@ -0,0 +1,122 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "g_context.h" +#include "g_state.h" +#include "g_prim.h" + + + + +static void validate_prim_pipe( struct generic_context *generic ) +{ + struct prim_stage *next = generic->prim.setup; + + /* TODO: make the current primitive part of the state and build + * shorter pipelines for lines & points. + */ + if (generic->setup.fill_cw != FILL_TRI || + generic->setup.fill_ccw != FILL_TRI) { + + generic->prim.unfilled->next = next; + next = generic->prim.unfilled; + } + + if (generic->setup.offset_cw || + generic->setup.offset_ccw) { + generic->prim.offset->next = next; + next = generic->prim.offset; + } + + if (generic->setup.light_twoside) { + generic->prim.twoside->next = next; + next = generic->prim.twoside; + } + + /* Always run the cull stage as we calculate determinant there + * also. Fix this.. + */ + { + generic->prim.cull->next = next; + next = generic->prim.cull; + } + + + /* Clip stage + */ + { + generic->prim.clip->next = next; + next = generic->prim.clip; + } + + /* Do software flatshading prior to clipping. XXX: should only do + * this for clipped primitives, ie it is a part of the clip + * routine. + */ + if (generic->setup.flatshade) { + generic->prim.flatshade->next = next; + next = generic->prim.flatshade; + } + + + generic->prim.first = next; +} + + + + +void generic_set_setup_state( struct softpipe_context *softpipe, + const struct softpipe_setup_state *setup ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy( &generic->setup, setup, sizeof(*setup) ); + + validate_prim_pipe( generic ); + generic->dirty |= G_NEW_SETUP; +} + + + +void generic_set_scissor_rect( struct softpipe_context *softpipe, + const struct softpipe_scissor_rect *scissor ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy( &generic->scissor, scissor, sizeof(*scissor) ); + generic->dirty |= G_NEW_SCISSOR; +} + + +void generic_set_polygon_stipple( struct softpipe_context *softpipe, + const struct softpipe_poly_stipple *stipple ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy( &generic->poly_stipple, stipple, sizeof(*stipple) ); + generic->dirty |= G_NEW_STIPPLE; +} diff --git a/src/mesa/softpipe/generic/g_state_surface.c b/src/mesa/softpipe/generic/g_state_surface.c new file mode 100644 index 0000000000..c791ce4640 --- /dev/null +++ b/src/mesa/softpipe/generic/g_state_surface.c @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_state.h" +#include "g_surface.h" + + +/* This is all a total hack. + */ +void generic_set_cbuf_state( struct softpipe_context *softpipe, + const struct softpipe_surface *surface ) +{ + struct generic_context *generic = generic_context(softpipe); + + if (generic->cbuf_surface == NULL) { + generic->cbuf_surface = CALLOC_STRUCT(generic_surface); + generic->cbuf_surface->type = &gs_rgba8; + } + + generic->cbuf_surface->surface = *surface; + generic->dirty |= G_NEW_CBUF; +} + + diff --git a/src/mesa/softpipe/generic/g_surface.c b/src/mesa/softpipe/generic/g_surface.c new file mode 100644 index 0000000000..2df30cde70 --- /dev/null +++ b/src/mesa/softpipe/generic/g_surface.c @@ -0,0 +1,153 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "g_context.h" +#include "g_state.h" +#include "g_surface.h" +#include "g_headers.h" + +static void rgba8_read_quad_f( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + rgba[k][0] = ptr[0] * (1.0 / 255.0); + rgba[k][1] = ptr[1] * (1.0 / 255.0); + rgba[k][2] = ptr[2] * (1.0 / 255.0); + rgba[k][3] = ptr[3] * (1.0 / 255.0); + } + } +} + +static void rgba8_read_quad_f_swz( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + rrrr[0][k] = ptr[0] * (1.0 / 255.0); + rrrr[1][k] = ptr[1] * (1.0 / 255.0); + rrrr[2][k] = ptr[2] * (1.0 / 255.0); + rrrr[3][k] = ptr[3] * (1.0 / 255.0); + } + } +} + +static void rgba8_write_quad_f( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + ptr[0] = rgba[k][0] * 255.0; + ptr[1] = rgba[k][1] * 255.0; + ptr[2] = rgba[k][2] * 255.0; + ptr[3] = rgba[k][3] * 255.0; + } + } +} + +static void rgba8_write_quad_f_swz( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + ptr[0] = rrrr[0][k] * 255.0; + ptr[1] = rrrr[1][k] * 255.0; + ptr[2] = rrrr[2][k] * 255.0; + ptr[3] = rrrr[3][k] * 255.0; + } + } +} + + + + +static void rgba8_read_quad_ub( struct generic_surface *gs, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + rgba[k][0] = ptr[0]; + rgba[k][1] = ptr[1]; + rgba[k][2] = ptr[2]; + rgba[k][3] = ptr[3]; + } + } +} + + +static void rgba8_write_quad_ub( struct generic_surface *gs, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + ptr[0] = rgba[k][0]; + ptr[1] = rgba[k][1]; + ptr[2] = rgba[k][2]; + ptr[3] = rgba[k][3]; + } + } +} + + + + +struct generic_surface_type gs_rgba8 = { + G_SURFACE_RGBA_8888, + rgba8_read_quad_f, + rgba8_read_quad_f_swz, + rgba8_read_quad_ub, + rgba8_write_quad_f, + rgba8_write_quad_f_swz, + rgba8_write_quad_ub, +}; + + + diff --git a/src/mesa/softpipe/generic/g_surface.h b/src/mesa/softpipe/generic/g_surface.h new file mode 100644 index 0000000000..23c28f0bb4 --- /dev/null +++ b/src/mesa/softpipe/generic/g_surface.h @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_SURFACE_H +#define G_SURFACE_H + +#include "glheader.h" +#include "g_headers.h" + +struct generic_surface; + +#define G_SURFACE_RGBA_8888 0x1 + +/* Internal structs and helpers for the primitive clip/setup pipeline: + */ +struct generic_surface_type { + + GLuint format; + + void (*read_quad_f)( struct generic_surface *, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ); + + void (*read_quad_f_swz)( struct generic_surface *, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ); + + void (*read_quad_ub)( struct generic_surface *, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ); + + + void (*write_quad_f)( struct generic_surface *, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ); + + void (*write_quad_f_swz)( struct generic_surface *, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ); + + + void (*write_quad_ub)( struct generic_surface *, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ); + + +}; + + +struct generic_surface { + struct generic_surface_type *type; + struct softpipe_surface surface; +}; + + +static INLINE void gs_read_quad_f( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + gs->type->read_quad_f(gs, x, y, rgba); +} + +static INLINE void gs_read_quad_f_swz( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + gs->type->read_quad_f_swz(gs, x, y, rrrr); +} + +static INLINE void gs_write_quad_f( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + gs->type->write_quad_f(gs, x, y, rgba); +} + +static INLINE void gs_write_quad_f_swz( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + gs->type->write_quad_f_swz(gs, x, y, rrrr); +} + +/* Like this, or hidden? + */ +struct generic_surface_type gs_rgba8; + +#endif diff --git a/src/mesa/softpipe/generic/g_tile.h b/src/mesa/softpipe/generic/g_tile.h new file mode 100644 index 0000000000..6ae5212356 --- /dev/null +++ b/src/mesa/softpipe/generic/g_tile.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#ifndef G_TILE_H +#define G_TILE_H + +struct generic_context; +struct quad_header; + +void quad_shade( struct generic_context *generic, + struct quad_header *quad ); + +void quad_output( struct generic_context *generic, + struct quad_header *quad ); + +#endif diff --git a/src/mesa/softpipe/generic/g_tile_fs.c b/src/mesa/softpipe/generic/g_tile_fs.c new file mode 100644 index 0000000000..35e1ab63de --- /dev/null +++ b/src/mesa/softpipe/generic/g_tile_fs.c @@ -0,0 +1,188 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "g_context.h" +#include "g_headers.h" +#include "g_tile.h" + +struct exec_machine { + const struct setup_coefficient *coef; + + GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE]; +}; + + +static void INLINE cinterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + exec->attr[attrib][i][j] = exec->coef[attrib].a0[i]; + } +} + + +/* Push into the fp: + * + * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx + * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy + */ +static INLINE void linterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + + exec->coef[attrib].dadx[i] * exec->attr[0][0][j] + + exec->coef[attrib].dady[i] * exec->attr[0][1][j]); + } +} + + +/* Push into the fp: + * + * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx + * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy + * INPUT[attr] = MUL INPUT[attr], INPUT_WPOS.wwww + * + * (Or should that be 1/w ???) + */ +static INLINE void pinterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + + exec->coef[attrib].dadx[i] * exec->attr[0][0][j] + + exec->coef[attrib].dady[i] * exec->attr[0][1][j]) * + exec->attr[0][3][j]); + } +} + + + +/* This should be done by the fragment shader execution unit (code + * generated from the decl instructions). Do it here for now. + */ +void quad_shade( struct generic_context *generic, + struct quad_header *quad ) +{ + struct exec_machine exec; + GLfloat fx = quad->x0; + GLfloat fy = quad->y0; + GLuint i, j; + + exec.coef = quad->coef; + + /* Position: + */ + exec.attr[FRAG_ATTRIB_WPOS][0][0] = fx; + exec.attr[FRAG_ATTRIB_WPOS][0][1] = fx + 1.0; + exec.attr[FRAG_ATTRIB_WPOS][0][2] = fx; + exec.attr[FRAG_ATTRIB_WPOS][0][3] = fx + 1.0; + + exec.attr[FRAG_ATTRIB_WPOS][1][0] = fy; + exec.attr[FRAG_ATTRIB_WPOS][1][1] = fy; + exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0; + exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; + + /* Z and W are done by linear interpolation: + */ + if (generic->need_z) { + linterp(&exec, 0, 2); + } + + if (generic->need_w) { + linterp(&exec, 0, 3); +// invert(&exec, 0, 3); + } + + /* Interpolate all the remaining attributes. This will get pushed + * into the fragment program's responsibilities at some point. + */ + for (i = 1; i < quad->nr_attrs; i++) { +#if 1 + for (j = 0; j < NUM_CHANNELS; j++) + linterp(&exec, i, j); +#else + switch (quad->interp[i]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + cinterp(&exec, i, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + linterp(&exec, i, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + pinterp(&exec, i, j); + break; + } +#endif + } + +#if 0 + generic->run_fs( tri->fp, quad, &tri->outputs ); +#else + { + GLuint attr = generic->fp_attr_to_slot[FRAG_ATTRIB_COL0]; + assert(attr); + + memcpy(quad->outputs.color, + exec.attr[attr], + sizeof(quad->outputs.color)); + } +#endif + + + if (quad->mask) + quad_output( generic, quad ); +} + + + + + + + diff --git a/src/mesa/softpipe/generic/g_tile_output.c b/src/mesa/softpipe/generic/g_tile_output.c new file mode 100644 index 0000000000..058ca30193 --- /dev/null +++ b/src/mesa/softpipe/generic/g_tile_output.c @@ -0,0 +1,92 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "g_context.h" +#include "g_headers.h" +#include "g_surface.h" +#include "g_tile.h" + + +static void mask_copy( GLfloat (*dest)[4], + GLfloat (*src)[4], + GLuint mask ) +{ + GLuint i, j; + + for (i = 0; i < 4; i++) { + if (mask & (1<mask != MASK_ALL) + { + GLfloat tmp[4][QUAD_SIZE]; + + /* Yes, we'll probably have a masked write as well, but this is + * how blend will be done at least. + */ + gs_read_quad_f_swz( generic->cbuf_surface, + quad->x0, + quad->y0, + tmp ); + + mask_copy( tmp, quad->outputs.color, quad->mask ); + + gs_write_quad_f_swz( generic->cbuf_surface, + quad->x0, + quad->y0, + tmp ); + } + else + { + gs_write_quad_f_swz( generic->cbuf_surface, + quad->x0, + quad->y0, + quad->outputs.color ); + } +} diff --git a/src/mesa/softpipe/sp_context.h b/src/mesa/softpipe/sp_context.h new file mode 100644 index 0000000000..a11ad26018 --- /dev/null +++ b/src/mesa/softpipe/sp_context.h @@ -0,0 +1,72 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_CONTEXT_H +#define SP_CONTEXT_H + +#include "mtypes.h" + +struct softpipe_context *generic_create( void ); + +/* Drawing currently kludged up via the existing tnl/ module. + */ +struct vertex_buffer; + +struct softpipe_context { + + void (*destroy)( struct softpipe_context * ); + + void (*set_clip_state)( struct softpipe_context *, + const struct softpipe_clip_state * ); + + void (*set_viewport)( struct softpipe_context *, + const struct softpipe_viewport * ); + + void (*set_setup_state)( struct softpipe_context *, + const struct softpipe_setup_state * ); + + void (*set_scissor_rect)( struct softpipe_context *, + const struct softpipe_scissor_rect * ); + + void (*set_fs_state)( struct softpipe_context *, + const struct softpipe_fs_state * ); + + void (*set_polygon_stipple)( struct softpipe_context *, + const struct softpipe_poly_stipple * ); + + void (*set_cbuf_state)( struct softpipe_context *, + const struct softpipe_surface * ); + + + void (*draw_vb)( struct softpipe_context *softpipe, + struct vertex_buffer *VB ); +}; + + + + +#endif diff --git a/src/mesa/softpipe/sp_defines.h b/src/mesa/softpipe/sp_defines.h new file mode 100644 index 0000000000..d8ae8d4baf --- /dev/null +++ b/src/mesa/softpipe/sp_defines.h @@ -0,0 +1,45 @@ + +#define SP_BLENDFACTOR_ONE 0x1 +#define SP_BLENDFACTOR_SRC_COLOR 0x2 +#define SP_BLENDFACTOR_SRC_ALPHA 0x3 +#define SP_BLENDFACTOR_DST_ALPHA 0x4 +#define SP_BLENDFACTOR_DST_COLOR 0x5 +#define SP_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 +#define SP_BLENDFACTOR_CONST_COLOR 0x7 +#define SP_BLENDFACTOR_CONST_ALPHA 0x8 +#define SP_BLENDFACTOR_SRC1_COLOR 0x9 +#define SP_BLENDFACTOR_SRC1_ALPHA 0x0A +#define SP_BLENDFACTOR_ZERO 0x11 +#define SP_BLENDFACTOR_INV_SRC_COLOR 0x12 +#define SP_BLENDFACTOR_INV_SRC_ALPHA 0x13 +#define SP_BLENDFACTOR_INV_DST_ALPHA 0x14 +#define SP_BLENDFACTOR_INV_DST_COLOR 0x15 +#define SP_BLENDFACTOR_INV_CONST_COLOR 0x17 +#define SP_BLENDFACTOR_INV_CONST_ALPHA 0x18 +#define SP_BLENDFACTOR_INV_SRC1_COLOR 0x19 +#define SP_BLENDFACTOR_INV_SRC1_ALPHA 0x1A + +#define SP_BLEND_ADD 0 +#define SP_BLEND_SUBTRACT 1 +#define SP_BLEND_REVERSE_SUBTRACT 2 +#define SP_BLEND_MIN 3 +#define SP_BLEND_MAX 4 + + +#define SP_LOGICOP_CLEAR 0 +#define SP_LOGICOP_NOR 1 +#define SP_LOGICOP_AND_INVERTED 2 +#define SP_LOGICOP_COPY_INVERTED 3 +#define SP_LOGICOP_AND_REVERSE 4 +#define SP_LOGICOP_INVERT 5 +#define SP_LOGICOP_XOR 6 +#define SP_LOGICOP_NAND 7 +#define SP_LOGICOP_AND 8 +#define SP_LOGICOP_EQUIV 9 +#define SP_LOGICOP_NOOP 10 +#define SP_LOGICOP_OR_INVERTED 11 +#define SP_LOGICOP_COPY 12 +#define SP_LOGICOP_OR_REVERSE 13 +#define SP_LOGICOP_OR 14 +#define SP_LOGICOP_SET 15 + diff --git a/src/mesa/softpipe/sp_state.h b/src/mesa/softpipe/sp_state.h new file mode 100644 index 0000000000..d69f9cd164 --- /dev/null +++ b/src/mesa/softpipe/sp_state.h @@ -0,0 +1,132 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_STATE_H +#define SP_STATE_H + +#include "mtypes.h" +#include "vf/vf.h" + +#define WINDING_NONE 0 +#define WINDING_CW 1 +#define WINDING_CCW 2 +#define WINDING_BOTH (WINDING_CW | WINDING_CCW) + +#define FILL_POINT 1 +#define FILL_LINE 2 +#define FILL_TRI 3 + +struct softpipe_setup_state { + GLuint flatshade:1; + GLuint light_twoside:1; + + GLuint front_winding:2; + + GLuint cull_mode:2; + + GLuint fill_cw:2; + GLuint fill_ccw:2; + + GLuint offset_cw:1; + GLuint offset_ccw:1; + + GLuint scissor:1; + GLuint poly_stipple:1; + + GLuint pad:18; + + GLfloat offset_units; + GLfloat offset_scale; +}; + +struct softpipe_poly_stipple { + GLuint stipple[32]; +}; + + +struct softpipe_viewport { + GLfloat scale[4]; + GLfloat translate[4]; +}; + +struct softpipe_scissor_rect { + GLshort minx; + GLshort miny; + GLshort maxx; + GLshort maxy; +}; + + +struct softpipe_clip_state { + GLfloat ucp[6][4]; + GLuint nr; +}; + +struct softpipe_fs_state { + struct gl_fragment_program *fp; +}; + +#define SP_MAX_CONSTANT 32 + +struct softpipe_constant_buffer { + GLfloat constant[SP_MAX_CONSTANT][4]; + GLuint nr_constants; +}; + + +struct softpipe_blend_state { + GLuint blend_enable:1; + GLuint ia_blend_enable:1; + + GLuint ia_dest_blend_factor:5; + GLuint ia_src_blend_factor:5; + GLuint ia_blend_function:3; + + GLuint dest_blend_factor:5; + GLuint src_blend_factor:5; + GLuint blend_function:3; + + GLuint logicop_enable:1; + GLuint logicop_func:4; +}; + +struct softpipe_blend_color { + GLfloat color[4]; +}; + + +/* This will change for hardware softpipes... + */ +struct softpipe_surface { + GLubyte *ptr; + GLint stride; + GLuint cpp; + GLuint format; +}; + + +#endif diff --git a/src/mesa/softpipe/state_tracker/Makefile b/src/mesa/softpipe/state_tracker/Makefile new file mode 100644 index 0000000000..0ab1dc6e6b --- /dev/null +++ b/src/mesa/softpipe/state_tracker/Makefile @@ -0,0 +1,2 @@ +default: + cd ../.. ; make \ No newline at end of file diff --git a/src/mesa/softpipe/state_tracker/st_atom.c b/src/mesa/softpipe/state_tracker/st_atom.c new file mode 100644 index 0000000000..99fba398bf --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom.c @@ -0,0 +1,168 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "glheader.h" +#include "context.h" + +#include "st_context.h" +#include "st_atom.h" + + + +/* This is used to initialize st->atoms[]. We could use this list + * directly except for a single atom, st_update_constants, which has a + * .dirty value which changes according to the parameters of the + * current fragment and vertex programs, and so cannot be a static + * value. + */ +static const struct st_tracked_state *atoms[] = +{ + &st_update_cbuf, + &st_update_clip, + &st_update_fs, + &st_update_setup, + &st_update_viewport, + /* will be patched out at runtime */ +/* &st_update_constants */ +}; + + +void st_init_atoms( struct st_context *st ) +{ + GLuint i; + + st->atoms = _mesa_malloc(sizeof(atoms)); + st->nr_atoms = sizeof(atoms)/sizeof(*atoms); + _mesa_memcpy(st->atoms, atoms, sizeof(atoms)); + + /* Patch in a pointer to the dynamic state atom: + */ + for (i = 0; i < st->nr_atoms; i++) + if (st->atoms[i] == &st_update_constants) + st->atoms[i] = &st->constants.tracked_state; + + _mesa_memcpy(&st->constants.tracked_state, + &st_update_constants, + sizeof(st_update_constants)); +} + + +void st_destroy_atoms( struct st_context *st ) +{ + if (st->atoms) { + _mesa_free(st->atoms); + st->atoms = NULL; + } +} + + +/*********************************************************************** + */ + +static GLboolean check_state( const struct st_state_flags *a, + const struct st_state_flags *b ) +{ + return ((a->mesa & b->mesa) || + (a->st & b->st)); +} + +static void accumulate_state( struct st_state_flags *a, + const struct st_state_flags *b ) +{ + a->mesa |= b->mesa; + a->st |= b->st; +} + + +static void xor_states( struct st_state_flags *result, + const struct st_state_flags *a, + const struct st_state_flags *b ) +{ + result->mesa = a->mesa ^ b->mesa; + result->st = a->st ^ b->st; +} + + +/*********************************************************************** + * Update all derived state: + */ + +void st_validate_state( struct st_context *st ) +{ + struct st_state_flags *state = &st->dirty; + GLuint i; + + if (state->st == 0) + return; + + if (1) { + /* Debug version which enforces various sanity checks on the + * state flags which are generated and checked to help ensure + * state atoms are ordered correctly in the list. + */ + struct st_state_flags examined, prev; + _mesa_memset(&examined, 0, sizeof(examined)); + prev = *state; + + for (i = 0; i < st->nr_atoms; i++) { + const struct st_tracked_state *atom = st->atoms[i]; + struct st_state_flags generated; + + assert(atom->dirty.mesa || + atom->dirty.st); + assert(atom->update); + + if (check_state(state, &atom->dirty)) { + st->atoms[i]->update( st ); + } + + accumulate_state(&examined, &atom->dirty); + + /* generated = (prev ^ state) + * if (examined & generated) + * fail; + */ + xor_states(&generated, &prev, state); + assert(!check_state(&examined, &generated)); + prev = *state; + } + } + else { + const GLuint nr = st->nr_atoms; + + for (i = 0; i < nr; i++) { + if (check_state(state, &st->atoms[i]->dirty)) + st->atoms[i]->update( st ); + } + } + + memset(state, 0, sizeof(*state)); +} + + + diff --git a/src/mesa/softpipe/state_tracker/st_atom.h b/src/mesa/softpipe/state_tracker/st_atom.h new file mode 100644 index 0000000000..87b72940b4 --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom.h @@ -0,0 +1,55 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef ST_ATOM_H +#define ST_ATOM_H + +struct st_context; +struct st_tracked_state; + +void st_init_atoms( struct st_context *st ); +void st_destroy_atoms( struct st_context *st ); + + +void st_validate_state( struct st_context *st ); + + +const struct st_tracked_state st_update_cbuf; +const struct st_tracked_state st_update_clip; +const struct st_tracked_state st_update_fs; +const struct st_tracked_state st_update_setup; +const struct st_tracked_state st_update_viewport; +const struct st_tracked_state st_update_constants; + + +#endif diff --git a/src/mesa/softpipe/state_tracker/st_atom_cbuf.c b/src/mesa/softpipe/state_tracker/st_atom_cbuf.c new file mode 100644 index 0000000000..cd707ec5ef --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom_cbuf.c @@ -0,0 +1,72 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + +extern GLboolean xmesa_get_cbuf_details( GLcontext *ctx, + void **ptr, + GLuint *cpp, + GLint *stride, + GLuint *format ); + + +/* This is a hack to work with the X11 driver as a test harness + */ +static void update_cbuf_state( struct st_context *st ) +{ + struct softpipe_surface cbuf; + GLboolean ok; + + ok = xmesa_get_cbuf_details( st->ctx, + (void **)&cbuf.ptr, + &cbuf.cpp, + &cbuf.stride, + &cbuf.format ); + + assert(ok); + + if (memcmp(&cbuf, &st->state.cbuf, sizeof(cbuf)) != 0) { + st->state.cbuf = cbuf; + st->softpipe->set_cbuf_state( st->softpipe, &cbuf ); + } +} + +const struct st_tracked_state st_update_cbuf = { + .dirty = { + .mesa = _NEW_BUFFERS, + .st = 0, + }, + .update = update_cbuf_state +}; + diff --git a/src/mesa/softpipe/state_tracker/st_atom_clip.c b/src/mesa/softpipe/state_tracker/st_atom_clip.c new file mode 100644 index 0000000000..1a0d2c4053 --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom_clip.c @@ -0,0 +1,75 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + + +/* Second state atom for user clip planes: + */ +static void update_clip( struct st_context *st ) +{ + struct softpipe_clip_state clip; + GLuint i; + + memset(&clip, 0, sizeof(clip)); + + for (i = 0; i < 6; i++) { + if (st->ctx->Transform.ClipPlanesEnabled & (1 << i)) { + memcpy(clip.ucp[clip.nr], + st->ctx->Transform._ClipUserPlane[i], + sizeof(clip.ucp[0])); + clip.nr++; + } + } + + if (memcmp(&clip, &st->state.clip, sizeof(clip)) != 0) { + st->state.clip = clip; + st->softpipe->set_clip_state(st->softpipe, &clip); + } +} + + +const struct st_tracked_state st_update_clip = { + .dirty = { + .mesa = (_NEW_TRANSFORM), + .st = 0, + }, + .update = update_clip +}; + + + + + diff --git a/src/mesa/softpipe/state_tracker/st_atom_fs.c b/src/mesa/softpipe/state_tracker/st_atom_fs.c new file mode 100644 index 0000000000..ca109d2d34 --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom_fs.c @@ -0,0 +1,56 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + + +static void update_fs( struct st_context *st ) +{ + struct softpipe_fs_state fs; + + fs.fp = st->ctx->FragmentProgram._Current; + + if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0) { + st->state.fs = fs; + st->softpipe->set_fs_state(st->softpipe, &fs); + } +} + + +const struct st_tracked_state st_update_fs = { + .dirty = { + .mesa = 0, + .st = ST_NEW_FRAGMENT_PROGRAM, + }, + .update = update_fs +}; diff --git a/src/mesa/softpipe/state_tracker/st_atom_setup.c b/src/mesa/softpipe/state_tracker/st_atom_setup.c new file mode 100644 index 0000000000..2a582ea36c --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom_setup.c @@ -0,0 +1,175 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + +static GLuint translate_fill( GLenum mode ) +{ + switch (mode) { + case GL_POINT: return FILL_POINT; + case GL_LINE: return FILL_LINE; + case GL_FILL: return FILL_TRI; + default: assert(0); return 0; + } +} + +static GLboolean get_offset_flag( GLuint fill_mode, + const struct gl_polygon_attrib *Polygon ) +{ + switch (fill_mode) { + case FILL_POINT: return Polygon->OffsetPoint; + case FILL_LINE: return Polygon->OffsetLine; + case FILL_TRI: return Polygon->OffsetFill; + default: assert(0); return 0; + } +} + + +static void update_setup_state( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + struct softpipe_setup_state setup; + + memset(&setup, 0, sizeof(setup)); + + /* _NEW_POLYGON, _NEW_BUFFERS + */ + { + setup.front_winding = WINDING_CW; + + if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) + setup.front_winding ^= WINDING_BOTH; + + if (ctx->Polygon.FrontFace != GL_CCW) + setup.front_winding ^= WINDING_BOTH; + } + + /* _NEW_LIGHT + */ + if (ctx->Light.ShadeModel == GL_FLAT) + setup.flatshade = 1; + + /* _NEW_LIGHT + * + * Not sure about the light->enabled requirement - does this still + * apply?? + */ + if (ctx->Light.Enabled && + ctx->Light.Model.TwoSide) + setup.light_twoside = 1; + + + /* _NEW_POLYGON + */ + if (ctx->Polygon.CullFlag) { + if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) { + setup.cull_mode = WINDING_BOTH; + } + else if (ctx->Polygon.CullFaceMode == GL_FRONT) { + setup.cull_mode = setup.front_winding; + } + else { + setup.cull_mode = setup.front_winding ^ WINDING_BOTH; + } + } + + /* _NEW_POLYGON + */ + { + GLuint fill_front = translate_fill( ctx->Polygon.FrontMode ); + GLuint fill_back = translate_fill( ctx->Polygon.BackMode ); + + if (setup.front_winding == WINDING_CW) { + setup.fill_cw = fill_front; + setup.fill_ccw = fill_back; + } + else { + setup.fill_cw = fill_back; + setup.fill_ccw = fill_front; + } + + /* Simplify when culling is active: + */ + if (setup.cull_mode & WINDING_CW) { + setup.fill_cw = setup.fill_ccw; + } + + if (setup.cull_mode & WINDING_CCW) { + setup.fill_ccw = setup.fill_cw; + } + } + + /* Hardware does offset for filled prims, but need to do it in + * software for unfilled. + * + * _NEW_POLYGON + */ + if (setup.fill_cw != FILL_TRI) + setup.offset_cw = get_offset_flag( setup.fill_cw, + &ctx->Polygon ); + + if (setup.fill_ccw != FILL_TRI) + setup.offset_ccw = get_offset_flag( setup.fill_ccw, + &ctx->Polygon ); + + + /* _NEW_BUFFERS, _NEW_POLYGON + */ + if (setup.fill_cw != FILL_TRI || + setup.fill_ccw != FILL_TRI) + { + GLfloat mrd = (ctx->DrawBuffer ? + ctx->DrawBuffer->_MRD : + 1.0); + + setup.offset_units = ctx->Polygon.OffsetFactor * mrd; + setup.offset_scale = (ctx->Polygon.OffsetUnits * mrd * + st->polygon_offset_scale); + } + + + if (memcmp(&setup, &st->state.setup, sizeof(setup)) != 0) { + st->state.setup = setup; + st->softpipe->set_setup_state( st->softpipe, &setup ); + } +} + +const struct st_tracked_state st_update_setup = { + .dirty = { + .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_BUFFERS), + .st = 0, + }, + .update = update_setup_state +}; diff --git a/src/mesa/softpipe/state_tracker/st_atom_viewport.c b/src/mesa/softpipe/state_tracker/st_atom_viewport.c new file mode 100644 index 0000000000..af896e2e31 --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom_viewport.c @@ -0,0 +1,117 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "context.h" +#include "colormac.h" +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + + + + + +/** + * Update the viewport transformation matrix. Depends on: + * - viewport pos/size + * - depthrange + * - window pos/size or FBO size + */ +static void update_viewport( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + const GLframebuffer *DrawBuffer = ctx->DrawBuffer; + GLfloat yScale = 1.0; + GLfloat yBias = 0.0; + + /* _NEW_BUFFERS + */ + if (DrawBuffer) { + +#if 0 + if (DrawBuffer->Name) { + /* User created FBO */ + struct st_renderbuffer *irb + = st_renderbuffer(DrawBuffer->_ColorDrawBuffers[0][0]); + if (irb && !irb->RenderToTexture) { + /* y=0=top */ + yScale = -1.0; + yBias = irb->Base.Height; + } + else { + /* y=0=bottom */ + yScale = 1.0; + yBias = 0.0; + } + } + else + { + /* window buffer, y=0=top */ + yScale = -1.0; + yBias = DrawBuffer->Height; + } +#endif + } + + { + /* _NEW_VIEWPORT + */ + GLfloat x = ctx->Viewport.X; + GLfloat y = ctx->Viewport.Y; + GLfloat z = ctx->Viewport.Near; + GLfloat half_width = ctx->Viewport.Width / 2.0; + GLfloat half_height = ctx->Viewport.Height / 2.0; + GLfloat half_depth = (ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; + + struct softpipe_viewport vp; + + vp.scale[0] = half_width; + vp.scale[1] = half_height * yScale; + vp.scale[2] = half_depth; + vp.scale[3] = 1.0; + + vp.translate[0] = (half_width + x); + vp.translate[1] = (half_height + y) * yScale + yBias; + vp.translate[2] = (half_depth + z); + vp.translate[3] = 0.0; + + if (memcmp(&vp, &st->state.viewport, sizeof(vp)) != 0) { + st->state.viewport = vp; + st->softpipe->set_viewport(st->softpipe, &vp); + } + } +} + + +const struct st_tracked_state st_update_viewport = { + .dirty = { + .mesa = _NEW_BUFFERS | _NEW_VIEWPORT, + .st = 0, + }, + .update = update_viewport +}; diff --git a/src/mesa/softpipe/state_tracker/st_cb_program.c b/src/mesa/softpipe/state_tracker/st_cb_program.c new file mode 100644 index 0000000000..67a589ca8d --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_cb_program.c @@ -0,0 +1,165 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "st_context.h" +#include "st_program.h" + +#include "glheader.h" +#include "macros.h" +#include "enums.h" +#include "prog_instruction.h" +#include "prog_parameter.h" +#include "program.h" +#include "programopt.h" +#include "tnl/tnl.h" + + +static void st_bind_program( GLcontext *ctx, + GLenum target, + struct gl_program *prog ) +{ + struct st_context *st = st_context(ctx); + + switch (target) { + case GL_VERTEX_PROGRAM_ARB: + break; + case GL_FRAGMENT_PROGRAM_ARB: + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; + break; + } +} + +static struct gl_program *st_new_program( GLcontext *ctx, + GLenum target, + GLuint id ) +{ + struct st_context *st = st_context(ctx); + + switch (target) { + case GL_VERTEX_PROGRAM_ARB: + return _mesa_init_vertex_program(ctx, + CALLOC_STRUCT(gl_vertex_program), + target, + id); + + case GL_FRAGMENT_PROGRAM_ARB: { + struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program); + + prog->id = st->program_id++; + + return _mesa_init_fragment_program( ctx, + &prog->Base, + target, + id ); + } + + default: + return _mesa_new_program(ctx, target, id); + } +} + +static void st_delete_program( GLcontext *ctx, + struct gl_program *prog ) +{ + _mesa_delete_program( ctx, prog ); +} + + +static GLboolean st_is_program_native( GLcontext *ctx, + GLenum target, + struct gl_program *prog ) +{ + return GL_TRUE; +} + +static void st_program_string_notify( GLcontext *ctx, + GLenum target, + struct gl_program *prog ) +{ + if (target == GL_FRAGMENT_PROGRAM_ARB) { + struct st_context *st = st_context(ctx); + + if (prog == &st->ctx->FragmentProgram._Current->Base) + { + struct st_fragment_program *p = + (struct st_fragment_program *) prog; + + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; + + p->id = st->program_id++; +#if 0 + p->param_state = p->Base.Base.Parameters->StateFlags; + p->translated = 0; +#endif + + /* Gack! do this in the compiler: + */ + if (p->Base.FogOption) { + /* add extra instructions to do fog, then turn off FogOption field */ + _mesa_append_fog_code(ctx, &p->Base); + p->Base.FogOption = GL_NONE; + } + } + } + else if (target == GL_VERTEX_PROGRAM_ARB) { + + /* Also tell tnl about it: + */ + _tnl_program_string(ctx, target, prog); + } +} + + + +void st_init_cb_program( struct st_context *st ) +{ + struct dd_function_table *functions = &st->ctx->Driver; + + /* Need these flags: + */ + st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; + st->ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; + + + assert(functions->ProgramStringNotify == _tnl_program_string); + functions->BindProgram = st_bind_program; + functions->NewProgram = st_new_program; + functions->DeleteProgram = st_delete_program; + functions->IsProgramNative = st_is_program_native; + functions->ProgramStringNotify = st_program_string_notify; +} + + +void st_destroy_cb_program( struct st_context *st ) +{ +} + diff --git a/src/mesa/softpipe/state_tracker/st_context.c b/src/mesa/softpipe/state_tracker/st_context.c new file mode 100644 index 0000000000..8a06dd88df --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_context.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "imports.h" +#include "st_public.h" +#include "st_context.h" +#include "st_atom.h" +#include "st_draw.h" +#include "st_program.h" +#include "softpipe/sp_context.h" + +void st_invalidate_state(GLcontext * ctx, GLuint new_state) +{ + struct st_context *st = st_context(ctx); + + st->dirty.mesa |= new_state; + st->dirty.st |= ST_NEW_MESA; +} + + +struct st_context *st_create_context( GLcontext *ctx, + struct softpipe_context *softpipe ) +{ + struct st_context *st = CALLOC_STRUCT( st_context ); + + ctx->st = st; + + st->ctx = ctx; + st->softpipe = softpipe; + + st->dirty.mesa = ~0; + st->dirty.st = ~0; + + st_init_atoms( st ); + st_init_draw( st ); + st_init_cb_program( st ); + + return st; +} + + +void st_destroy_context( struct st_context *st ) +{ + st_destroy_atoms( st ); + st_destroy_draw( st ); + st_destroy_cb_program( st ); + st->softpipe->destroy( st->softpipe ); + FREE( st ); +} + + + diff --git a/src/mesa/softpipe/state_tracker/st_context.h b/src/mesa/softpipe/state_tracker/st_context.h new file mode 100644 index 0000000000..08cee11797 --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_context.h @@ -0,0 +1,109 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ST_CONTEXT_H +#define ST_CONTEXT_H + +#include "mtypes.h" +#include "softpipe/sp_state.h" + + +struct st_context; +struct st_region; +struct st_texture_object; +struct st_texture_image; +struct st_fragment_program; + +#define ST_NEW_MESA 0x1 /* Mesa state has changed */ +#define ST_NEW_FRAGMENT_PROGRAM 0x2 + +struct st_state_flags { + GLuint mesa; + GLuint st; +}; + +struct st_tracked_state { + struct st_state_flags dirty; + void (*update)( struct st_context *st ); +}; + + + + +struct st_context +{ + GLcontext *ctx; + + struct softpipe_context *softpipe; + + /* Eventually will use a cache to feed the softpipe with + * create/bind/delete calls to constant state objects. Not yet + * though, we just shove random objects across the interface. + */ + struct { + struct softpipe_viewport viewport; + struct softpipe_setup_state setup; + struct softpipe_fs_state fs; + struct softpipe_blend_state blend; + struct softpipe_surface cbuf; + struct softpipe_clip_state clip; + struct softpipe_scissor_rect scissor; + struct softpipe_poly_stipple poly_stipple; + } state; + + struct { + struct st_tracked_state tracked_state; + } constants; + + struct { + struct gl_fragment_program *fragment_program; + } cb; + + /* State to be validated: + */ + struct st_tracked_state **atoms; + GLuint nr_atoms; + + struct st_state_flags dirty; + + /* Counter to track program string changes: + */ + GLuint program_id; + + GLfloat polygon_offset_scale; /* ?? */ +}; + + +/* Need this so that we can implement Mesa callbacks in this module. + */ +static INLINE struct st_context *st_context(GLcontext *ctx) +{ + return ctx->st; +} + + +#endif diff --git a/src/mesa/softpipe/state_tracker/st_draw.c b/src/mesa/softpipe/state_tracker/st_draw.c new file mode 100644 index 0000000000..12646402ee --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_draw.c @@ -0,0 +1,107 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "imports.h" + +#include "tnl/t_context.h" +#include "tnl/t_pipeline.h" + +#include "st_context.h" +#include "st_atom.h" +#include "st_draw.h" +#include "softpipe/sp_context.h" + +/* + * TNL stage which feedsinto the above. + * + * XXX: this needs to go into each driver using this code, because we + * cannot make the leap from ctx->draw_context in this file. The + * driver needs to customize tnl anyway, so this isn't a big deal. + */ +static GLboolean draw( GLcontext * ctx, struct tnl_pipeline_stage *stage ) +{ + struct st_context *st = st_context(ctx); + struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; + + /* Validate driver and softpipe state: + */ + st_validate_state( st ); + + /* Call into the new draw code to handle the VB: + */ + st->softpipe->draw_vb( st->softpipe, VB ); + + /* Finished + */ + return GL_FALSE; +} + +const struct tnl_pipeline_stage st_draw = { + "check state and draw", + NULL, + NULL, + NULL, + NULL, + draw +}; + +static const struct tnl_pipeline_stage *intel_pipeline[] = { + &_tnl_vertex_transform_stage, + &_tnl_vertex_cull_stage, + &_tnl_normal_transform_stage, + &_tnl_lighting_stage, + &_tnl_fog_coordinate_stage, + &_tnl_texgen_stage, + &_tnl_texture_transform_stage, + &_tnl_point_attenuation_stage, + &_tnl_vertex_program_stage, + &st_draw, /* ADD: escape to softpipe */ + 0, +}; + +/* This is all a hack to keep using tnl until we have vertex programs + * up and running. + */ +void st_init_draw( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + + _tnl_destroy_pipeline( ctx ); + _tnl_install_pipeline( ctx, intel_pipeline ); +} + + +void st_destroy_draw( struct st_context *st ) +{ + /* Nothing to do. + */ +} diff --git a/src/mesa/softpipe/state_tracker/st_draw.h b/src/mesa/softpipe/state_tracker/st_draw.h new file mode 100644 index 0000000000..f51059706a --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_draw.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef ST_DRAW_H +#define ST_DRAW_H + +void st_init_draw( struct st_context *st ); +void st_destroy_draw( struct st_context *st ); + +#endif diff --git a/src/mesa/softpipe/state_tracker/st_program.h b/src/mesa/softpipe/state_tracker/st_program.h new file mode 100644 index 0000000000..a47059d7a6 --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_program.h @@ -0,0 +1,68 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef ST_PROGRAM_H +#define ST_PROGRAM_H + +struct st_fragment_program +{ + struct gl_fragment_program Base; + GLboolean error; /* If program is malformed for any reason. */ + + GLuint id; /* String id, for tracking + * ProgramStringNotify changes. + */ + +#if 0 + GLfloat (*cbuffer)[4]; + GLuint nr_constants; + + /* Translate all the parameters, etc, into a constant buffer which + * we update on state changes. + */ + struct + { + GLuint reg; /* Constant idx */ + const GLfloat *values; /* Pointer to tracked values */ + } *param; + GLuint nr_params; + + GLuint param_state; +#endif +}; + + +void st_init_cb_program( struct st_context *st ); +void st_destroy_cb_program( struct st_context *st ); + +#endif diff --git a/src/mesa/softpipe/state_tracker/st_public.h b/src/mesa/softpipe/state_tracker/st_public.h new file mode 100644 index 0000000000..8241dbd52d --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_public.h @@ -0,0 +1,43 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ST_PUBLIC_H +#define ST_PUBLIC_H + +#include "mtypes.h" + +struct st_context; +struct softpipe_context; + +struct st_context *st_create_context( GLcontext *ctx, + struct softpipe_context *softpipe); + +void st_destroy_context( struct st_context *st ); + +void st_invalidate_state(GLcontext * ctx, GLuint new_state); + +#endif -- cgit v1.2.3 From b939adfa155f2b3ca5c5226e86da85629654d79b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 24 May 2007 10:44:53 +0100 Subject: Use the x11 driver as a test harness for the softpipe/state_tracker code. This has some limitations as we currently require a mapped framebuffer, so it only really works with double-buffered ximage rgba8888 windows. --- src/mesa/drivers/x11/Makefile | 2 ++ src/mesa/drivers/x11/xm_api.c | 9 +++++++++ src/mesa/drivers/x11/xm_dd.c | 5 +++++ src/mesa/drivers/x11/xm_span.c | 13 +++++++++++++ src/mesa/drivers/x11/xm_tri.c | 40 ++++++++++++++++++++++++++++++++++++++++ src/mesa/drivers/x11/xmesaP.h | 7 +++++++ src/mesa/main/mtypes.h | 3 ++- src/mesa/sources | 36 ++++++++++++++++++++++++++++++++++++ 8 files changed, 114 insertions(+), 1 deletion(-) create mode 100644 src/mesa/drivers/x11/Makefile (limited to 'src') diff --git a/src/mesa/drivers/x11/Makefile b/src/mesa/drivers/x11/Makefile new file mode 100644 index 0000000000..0ab1dc6e6b --- /dev/null +++ b/src/mesa/drivers/x11/Makefile @@ -0,0 +1,2 @@ +default: + cd ../.. ; make \ No newline at end of file diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index a07d0a90cf..af1a169a76 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -80,6 +80,9 @@ #include "tnl/t_pipeline.h" #include "drivers/common/driverfuncs.h" +#include "softpipe/state_tracker/st_public.h" +#include "softpipe/generic/g_context.h" + /** * Global X driver lock */ @@ -1556,6 +1559,10 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) xmesa_register_swrast_functions( mesaCtx ); _swsetup_Wakeup(mesaCtx); + + st_create_context( mesaCtx, + generic_create() ); + return c; } @@ -1758,6 +1765,8 @@ PUBLIC GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, XMesaBuffer readBuffer ) { + _mesa_printf("%s %p %p %p\n", __FUNCTION__, drawBuffer, readBuffer ); + if (c) { if (!drawBuffer || !readBuffer) return GL_FALSE; /* must specify buffers! */ diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index c8546236fb..63e7115e80 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -53,6 +53,8 @@ #include "tnl/tnl.h" #include "tnl/t_context.h" +#include "softpipe/generic/g_context.h" +#include "softpipe/state_tracker/st_public.h" /* @@ -828,6 +830,9 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) _vbo_InvalidateState( ctx, new_state ); _swsetup_InvalidateState( ctx, new_state ); + st_invalidate_state( ctx, new_state ); + + if (ctx->DrawBuffer->Name != 0) return; diff --git a/src/mesa/drivers/x11/xm_span.c b/src/mesa/drivers/x11/xm_span.c index a24966b3ae..3776891e2e 100644 --- a/src/mesa/drivers/x11/xm_span.c +++ b/src/mesa/drivers/x11/xm_span.c @@ -1303,6 +1303,17 @@ static void put_row_rgb_TRUEDITHER_ximage( RGB_SPAN_ARGS ) } + +static void *get_pointer_4_ximage( GLcontext *ctx, + struct gl_renderbuffer *rb, + GLint x, GLint y ) +{ + GET_XRB(xrb); + return PIXEL_ADDR4(xrb, x, y); +} + + + /* * Write a span of PF_8A8B8G8R-format pixels to an ximage. */ @@ -4593,6 +4604,7 @@ xmesa_set_renderbuffer_funcs(struct xmesa_renderbuffer *xrb, xrb->Base.PutMonoRow = put_mono_row_8A8B8G8R_ximage; xrb->Base.PutValues = put_values_8A8B8G8R_ximage; xrb->Base.PutMonoValues = put_mono_values_8A8B8G8R_ximage; + xrb->Base.GetPointer = get_pointer_4_ximage; } break; case PF_8A8R8G8B: @@ -4609,6 +4621,7 @@ xmesa_set_renderbuffer_funcs(struct xmesa_renderbuffer *xrb, xrb->Base.PutMonoRow = put_mono_row_8A8R8G8B_ximage; xrb->Base.PutValues = put_values_8A8R8G8B_ximage; xrb->Base.PutMonoValues = put_mono_values_8A8R8G8B_ximage; + xrb->Base.GetPointer = get_pointer_4_ximage; } break; case PF_8R8G8B: diff --git a/src/mesa/drivers/x11/xm_tri.c b/src/mesa/drivers/x11/xm_tri.c index 95c6d7c1d2..6158ef30f9 100644 --- a/src/mesa/drivers/x11/xm_tri.c +++ b/src/mesa/drivers/x11/xm_tri.c @@ -1443,6 +1443,46 @@ do { \ #endif + +GLboolean xmesa_get_cbuf_details( GLcontext *ctx, + void **ptr, + GLuint *cpp, + GLint *stride, + GLuint *format ) +{ + XMesaContext xmesa = XMESA_CONTEXT(ctx); + struct gl_framebuffer *fb = ctx->DrawBuffer; + struct gl_renderbuffer *crb = fb->_ColorDrawBuffers[0][0]; + struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(crb->Wrapped); + + *ptr = crb->GetPointer(ctx, crb, 0, 0); + *stride = ((GLubyte *)crb->GetPointer(ctx, crb, 0, 1) - + (GLubyte *)crb->GetPointer(ctx, crb, 0, 0)); + + if (!ptr) + goto bad; + + switch (xmesa->pixelformat) { + case PF_8A8B8G8R: + case PF_8A8R8G8B: + *format = 1; /* whatever */ + *cpp = 4; + break; + default: + goto bad; + } + + return GL_TRUE; + + bad: + *ptr = NULL; + *stride = 0; + *format = 0; + return GL_FALSE; +} + + + /** * Return pointer to line drawing function, or NULL if we should use a * swrast fallback. diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 0198886747..8f0bd34f47 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -578,4 +578,11 @@ extern void xmesa_register_swrast_functions( GLcontext *ctx ); #define ENABLE_EXT_timer_query 0 /* may not have 64-bit GLuint64EXT */ #endif + +GLboolean xmesa_get_cbuf_details( GLcontext *ctx, + void **ptr, + GLuint *cpp, + GLint *stride, + GLuint *format ); + #endif diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 71215d5470..2a8556388c 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -126,6 +126,7 @@ struct gl_pixelstore_attrib; struct gl_texture_format; struct gl_texture_image; struct gl_texture_object; +struct st_context; typedef struct __GLcontextRec GLcontext; typedef struct __GLcontextModesRec GLvisual; typedef struct gl_framebuffer GLframebuffer; @@ -3069,7 +3070,7 @@ struct __GLcontextRec void *swsetup_context; void *swtnl_context; void *swtnl_im; - void *acache_context; + struct st_context *st; void *aelt_context; /*@}*/ }; diff --git a/src/mesa/sources b/src/mesa/sources index dbfc01d0ed..8138fb627d 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -149,6 +149,40 @@ VBO_SOURCES = \ vbo/vbo_save_draw.c \ vbo/vbo_save_loopback.c +VF_SOURCES = \ + vf/vf.c \ + vf/vf_generic.c \ + vf/vf_sse.c + +SOFTPIPE_SOURCES = \ + softpipe/generic/g_context.c \ + softpipe/generic/g_draw.c \ + softpipe/generic/g_prim_clip.c \ + softpipe/generic/g_prim_cull.c \ + softpipe/generic/g_prim_flatshade.c \ + softpipe/generic/g_prim_offset.c \ + softpipe/generic/g_prim_setup.c \ + softpipe/generic/g_prim_twoside.c \ + softpipe/generic/g_prim_unfilled.c \ + softpipe/generic/g_state_clip.c \ + softpipe/generic/g_state_derived.c \ + softpipe/generic/g_state_fs.c \ + softpipe/generic/g_state_setup.c \ + softpipe/generic/g_state_surface.c \ + softpipe/generic/g_surface.c \ + softpipe/generic/g_tile_fs.c \ + softpipe/generic/g_tile_output.c \ + softpipe/state_tracker/st_atom.c \ + softpipe/state_tracker/st_atom_clip.c \ + softpipe/state_tracker/st_atom_fs.c \ + softpipe/state_tracker/st_atom_setup.c \ + softpipe/state_tracker/st_atom_cbuf.c \ + softpipe/state_tracker/st_atom_viewport.c \ + softpipe/state_tracker/st_cb_program.c \ + softpipe/state_tracker/st_draw.c \ + softpipe/state_tracker/st_context.c + + SHADER_SOURCES = \ shader/arbprogparse.c \ @@ -294,6 +328,8 @@ SOLO_SOURCES = \ $(MAIN_SOURCES) \ $(MATH_SOURCES) \ $(VBO_SOURCES) \ + $(VF_SOURCES) \ + $(SOFTPIPE_SOURCES) \ $(TNL_SOURCES) \ $(SHADER_SOURCES) \ $(SWRAST_SOURCES) \ -- cgit v1.2.3 From b1dc66b7bd9b13498204c32182d0935d3a7d3eec Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 24 May 2007 11:37:08 +0100 Subject: Add the vf module. This is a cleaned up version of the code in tnl/t_vertex*. --- src/mesa/vf/vf.c | 372 ++++++++++++++++++ src/mesa/vf/vf.h | 226 +++++++++++ src/mesa/vf/vf_generic.c | 981 +++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/vf/vf_sse.c | 664 ++++++++++++++++++++++++++++++++ 4 files changed, 2243 insertions(+) create mode 100644 src/mesa/vf/vf.c create mode 100644 src/mesa/vf/vf.h create mode 100644 src/mesa/vf/vf_generic.c create mode 100644 src/mesa/vf/vf_sse.c (limited to 'src') diff --git a/src/mesa/vf/vf.c b/src/mesa/vf/vf.c new file mode 100644 index 0000000000..cb25f2e113 --- /dev/null +++ b/src/mesa/vf/vf.c @@ -0,0 +1,372 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + +#include "glheader.h" +#include "context.h" +#include "colormac.h" + +#include "vf.h" + +#define DBG 0 + + + +static GLboolean match_fastpath( struct vertex_fetch *vf, + const struct vf_fastpath *fp) +{ + GLuint j; + + if (vf->attr_count != fp->attr_count) + return GL_FALSE; + + for (j = 0; j < vf->attr_count; j++) + if (vf->attr[j].format != fp->attr[j].format || + vf->attr[j].inputsize != fp->attr[j].size || + vf->attr[j].vertoffset != fp->attr[j].offset) + return GL_FALSE; + + if (fp->match_strides) { + if (vf->vertex_stride != fp->vertex_stride) + return GL_FALSE; + + for (j = 0; j < vf->attr_count; j++) + if (vf->attr[j].inputstride != fp->attr[j].stride) + return GL_FALSE; + } + + return GL_TRUE; +} + +static GLboolean search_fastpath_emit( struct vertex_fetch *vf ) +{ + struct vf_fastpath *fp = vf->fastpath; + + for ( ; fp ; fp = fp->next) { + if (match_fastpath(vf, fp)) { + vf->emit = fp->func; + return GL_TRUE; + } + } + + return GL_FALSE; +} + +void vf_register_fastpath( struct vertex_fetch *vf, + GLboolean match_strides ) +{ + struct vf_fastpath *fastpath = CALLOC_STRUCT(vf_fastpath); + GLuint i; + + fastpath->vertex_stride = vf->vertex_stride; + fastpath->attr_count = vf->attr_count; + fastpath->match_strides = match_strides; + fastpath->func = vf->emit; + fastpath->attr = (struct vf_attr_type *) + _mesa_malloc(vf->attr_count * sizeof(fastpath->attr[0])); + + for (i = 0; i < vf->attr_count; i++) { + fastpath->attr[i].format = vf->attr[i].format; + fastpath->attr[i].stride = vf->attr[i].inputstride; + fastpath->attr[i].size = vf->attr[i].inputsize; + fastpath->attr[i].offset = vf->attr[i].vertoffset; + } + + fastpath->next = vf->fastpath; + vf->fastpath = fastpath; +} + + + + +/*********************************************************************** + * Build codegen functions or return generic ones: + */ +static void choose_emit_func( struct vertex_fetch *vf, + GLuint count, + GLubyte *dest) +{ + vf->emit = NULL; + + /* Does this match an existing (hardwired, codegen or known-bad) + * fastpath? + */ + if (search_fastpath_emit(vf)) { + /* Use this result. If it is null, then it is already known + * that the current state will fail for codegen and there is no + * point trying again. + */ + } + else if (vf->codegen_emit) { + vf->codegen_emit( vf ); + } + + if (!vf->emit) { + vf_generate_hardwired_emit(vf); + } + + /* Otherwise use the generic version: + */ + if (!vf->emit) + vf->emit = vf_generic_emit; + + vf->emit( vf, count, dest ); +} + + + + + +/*********************************************************************** + * Public entrypoints, mostly dispatch to the above: + */ + + + +GLuint vf_set_vertex_attributes( struct vertex_fetch *vf, + const struct vf_attr_map *map, + GLuint nr, + GLuint vertex_stride ) +{ + GLuint offset = 0; + GLuint i, j; + + assert(nr < VF_ATTRIB_MAX); + + memset(vf->lookup, 0, sizeof(vf->lookup)); + + for (j = 0, i = 0; i < nr; i++) { + const GLuint format = map[i].format; + if (format == EMIT_PAD) { + if (DBG) + _mesa_printf("%d: pad %d, offset %d\n", i, + map[i].offset, offset); + + offset += map[i].offset; + + } + else { + assert(vf->lookup[map[i].attrib] == 0); + vf->lookup[map[i].attrib] = &vf->attr[j]; + + vf->attr[j].attrib = map[i].attrib; + vf->attr[j].format = format; + vf->attr[j].insert = vf_format_info[format].insert; + vf->attr[j].extract = vf_format_info[format].extract; + vf->attr[j].vertattrsize = vf_format_info[format].attrsize; + vf->attr[j].vertoffset = offset; + + if (DBG) + _mesa_printf("%d: %s, offset %d\n", i, + vf_format_info[format].name, + vf->attr[j].vertoffset); + + offset += vf_format_info[format].attrsize; + j++; + } + } + + vf->attr_count = j; + vf->vertex_stride = vertex_stride ? vertex_stride : offset; + vf->emit = choose_emit_func; + + assert(vf->vertex_stride >= offset); + return vf->vertex_stride; +} + + + +void vf_set_vp_matrix( struct vertex_fetch *vf, + const GLfloat *viewport ) +{ + assert(vf->allow_viewport_emits); + + /* scale */ + vf->vp[0] = viewport[MAT_SX]; + vf->vp[1] = viewport[MAT_SY]; + vf->vp[2] = viewport[MAT_SZ]; + vf->vp[3] = 1.0; + + /* translate */ + vf->vp[4] = viewport[MAT_TX]; + vf->vp[5] = viewport[MAT_TY]; + vf->vp[6] = viewport[MAT_TZ]; + vf->vp[7] = 0.0; +} + +void vf_set_vp_scale_translate( struct vertex_fetch *vf, + const GLfloat *scale, + const GLfloat *translate ) +{ + assert(vf->allow_viewport_emits); + + vf->vp[0] = scale[0]; + vf->vp[1] = scale[1]; + vf->vp[2] = scale[2]; + vf->vp[3] = scale[3]; + + vf->vp[4] = translate[0]; + vf->vp[5] = translate[1]; + vf->vp[6] = translate[2]; + vf->vp[7] = translate[3]; +} + + +/* Set attribute pointers, adjusted for start position: + */ +void vf_set_sources( struct vertex_fetch *vf, + GLvector4f * const sources[], + GLuint start ) +{ + struct vf_attr *a = vf->attr; + GLuint j; + + for (j = 0; j < vf->attr_count; j++) { + const GLvector4f *vptr = sources[a[j].attrib]; + + if ((a[j].inputstride != vptr->stride) || + (a[j].inputsize != vptr->size)) + vf->emit = choose_emit_func; + + a[j].inputstride = vptr->stride; + a[j].inputsize = vptr->size; + a[j].do_insert = a[j].insert[vptr->size - 1]; + a[j].inputptr = ((GLubyte *)vptr->data) + start * vptr->stride; + } +} + + + +/* Emit count VB vertices to dest. + */ +void vf_emit_vertices( struct vertex_fetch *vf, + GLuint count, + void *dest ) +{ + vf->emit( vf, count, (GLubyte*) dest ); +} + + +/* Extract a named attribute from a hardware vertex. Will have to + * reverse any viewport transformation, swizzling or other conversions + * which may have been applied. + * + * This is mainly required for on-the-fly vertex translations to + * swrast format. + */ +void vf_get_attr( struct vertex_fetch *vf, + const void *vertex, + GLenum attr, + const GLfloat *dflt, + GLfloat *dest ) +{ + const struct vf_attr *a = vf->attr; + const GLuint attr_count = vf->attr_count; + GLuint j; + + for (j = 0; j < attr_count; j++) { + if (a[j].attrib == attr) { + a[j].extract( &a[j], dest, (GLubyte *)vertex + a[j].vertoffset ); + return; + } + } + + /* Else return the value from ctx->Current. + */ + _mesa_memcpy( dest, dflt, 4*sizeof(GLfloat)); +} + + + + +struct vertex_fetch *vf_create( GLboolean allow_viewport_emits ) +{ + struct vertex_fetch *vf = CALLOC_STRUCT(vertex_fetch); + GLuint i; + + for (i = 0; i < VF_ATTRIB_MAX; i++) + vf->attr[i].vf = vf; + + vf->allow_viewport_emits = allow_viewport_emits; + + switch(CHAN_TYPE) { + case GL_UNSIGNED_BYTE: + vf->chan_scale[0] = 255.0; + vf->chan_scale[1] = 255.0; + vf->chan_scale[2] = 255.0; + vf->chan_scale[3] = 255.0; + break; + case GL_UNSIGNED_SHORT: + vf->chan_scale[0] = 65535.0; + vf->chan_scale[1] = 65535.0; + vf->chan_scale[2] = 65535.0; + vf->chan_scale[3] = 65535.0; + break; + default: + vf->chan_scale[0] = 1.0; + vf->chan_scale[1] = 1.0; + vf->chan_scale[2] = 1.0; + vf->chan_scale[3] = 1.0; + break; + } + + vf->identity[0] = 0.0; + vf->identity[1] = 0.0; + vf->identity[2] = 0.0; + vf->identity[3] = 1.0; + + vf->codegen_emit = NULL; + +#ifdef USE_SSE_ASM + if (!_mesa_getenv("MESA_NO_CODEGEN")) + vf->codegen_emit = vf_generate_sse_emit; +#endif + + return vf; +} + + +void vf_destroy( struct vertex_fetch *vf ) +{ + struct vf_fastpath *fp, *tmp; + + for (fp = vf->fastpath ; fp ; fp = tmp) { + tmp = fp->next; + FREE(fp->attr); + + /* KW: At the moment, fp->func is constrained to be allocated by + * _mesa_exec_alloc(), as the hardwired fastpaths in + * t_vertex_generic.c are handled specially. It would be nice + * to unify them, but this probably won't change until this + * module gets another overhaul. + */ + _mesa_exec_free((void *) fp->func); + FREE(fp); + } + + vf->fastpath = NULL; + FREE(vf); +} diff --git a/src/mesa/vf/vf.h b/src/mesa/vf/vf.h new file mode 100644 index 0000000000..fcbf490ce1 --- /dev/null +++ b/src/mesa/vf/vf.h @@ -0,0 +1,226 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + +#ifndef VF_VERTEX_H +#define VF_VERTEX_H + +#include "mtypes.h" +#include "m_vector.h" + +enum { + VF_ATTRIB_POS = 0, + VF_ATTRIB_WEIGHT = 1, + VF_ATTRIB_NORMAL = 2, + VF_ATTRIB_COLOR0 = 3, + VF_ATTRIB_COLOR1 = 4, + VF_ATTRIB_FOG = 5, + VF_ATTRIB_COLOR_INDEX = 6, + VF_ATTRIB_EDGEFLAG = 7, + VF_ATTRIB_TEX0 = 8, + VF_ATTRIB_TEX1 = 9, + VF_ATTRIB_TEX2 = 10, + VF_ATTRIB_TEX3 = 11, + VF_ATTRIB_TEX4 = 12, + VF_ATTRIB_TEX5 = 13, + VF_ATTRIB_TEX6 = 14, + VF_ATTRIB_TEX7 = 15, + VF_ATTRIB_POINTSIZE = 16, + VF_ATTRIB_BFC0 = 17, + VF_ATTRIB_BFC1 = 18, + VF_ATTRIB_CLIP_POS = 19, + VF_ATTRIB_VERTEX_HEADER = 20, + VF_ATTRIB_MAX = 21 +}; + + +enum vf_attr_format { + EMIT_1F, + EMIT_2F, + EMIT_3F, + EMIT_4F, + EMIT_2F_VIEWPORT, /* do viewport transform and emit */ + EMIT_3F_VIEWPORT, /* do viewport transform and emit */ + EMIT_4F_VIEWPORT, /* do viewport transform and emit */ + EMIT_3F_XYW, /* for projective texture */ + EMIT_1UB_1F, /* for fog coordinate */ + EMIT_3UB_3F_RGB, /* for specular color */ + EMIT_3UB_3F_BGR, /* for specular color */ + EMIT_4UB_4F_RGBA, /* for color */ + EMIT_4UB_4F_BGRA, /* for color */ + EMIT_4UB_4F_ARGB, /* for color */ + EMIT_4UB_4F_ABGR, /* for color */ + EMIT_4CHAN_4F_RGBA, /* for swrast color */ + EMIT_PAD, /* leave a hole of 'offset' bytes */ + EMIT_MAX +}; + +struct vf_attr_map { + GLuint attrib; + enum vf_attr_format format; + GLuint offset; +}; + +struct vertex_fetch; + +void vf_set_vp_matrix( struct vertex_fetch *vf, + const GLfloat *viewport ); + +void vf_set_vp_scale_translate( struct vertex_fetch *vf, + const GLfloat *scale, + const GLfloat *translate ); + +GLuint vf_set_vertex_attributes( struct vertex_fetch *vf, + const struct vf_attr_map *map, + GLuint nr, + GLuint vertex_stride ); + +void vf_set_sources( struct vertex_fetch *vf, + GLvector4f * const attrib[], + GLuint start ); + +void vf_emit_vertices( struct vertex_fetch *vf, + GLuint count, + void *dest ); + +void vf_get_attr( struct vertex_fetch *vf, + const void *vertex, + GLenum attr, + const GLfloat *dflt, + GLfloat *dest ); + +struct vertex_fetch *vf_create( GLboolean allow_viewport_emits ); + +void vf_destroy( struct vertex_fetch *vf ); + + + +/*********************************************************************** + * Internal functions and structs: + */ + +struct vf_attr; + +typedef void (*vf_extract_func)( const struct vf_attr *a, + GLfloat *out, + const GLubyte *v ); + +typedef void (*vf_insert_func)( const struct vf_attr *a, + GLubyte *v, + const GLfloat *in ); + +typedef void (*vf_emit_func)( struct vertex_fetch *vf, + GLuint count, + GLubyte *dest ); + + + +/* Describes how to convert/move a vertex attribute from a vertex + * array to a vertex structure. + */ +struct vf_attr +{ + struct vertex_fetch *vf; + + GLuint format; + GLuint inputsize; + GLuint inputstride; + GLuint vertoffset; /* position of the attrib in the vertex struct */ + + GLuint attrib; /* which vertex attrib (0=position, etc) */ + GLuint vertattrsize; /* size of the attribute in bytes */ + + GLubyte *inputptr; + const vf_insert_func *insert; + vf_insert_func do_insert; + vf_extract_func extract; +}; + +struct vertex_fetch +{ + struct vf_attr attr[VF_ATTRIB_MAX]; + GLuint attr_count; + GLuint vertex_stride; + + struct vf_attr *lookup[VF_ATTRIB_MAX]; + + vf_emit_func emit; + + /* Parameters and constants for codegen: + */ + GLboolean allow_viewport_emits; + GLfloat vp[8]; + GLfloat chan_scale[4]; + GLfloat identity[4]; + + struct vf_fastpath *fastpath; + + void (*codegen_emit)( struct vertex_fetch *vf ); +}; + + +struct vf_attr_type { + GLuint format; + GLuint size; + GLuint stride; + GLuint offset; +}; + +struct vf_fastpath { + GLuint vertex_stride; + GLuint attr_count; + GLboolean match_strides; + + struct vf_attr_type *attr; + + vf_emit_func func; + struct vf_fastpath *next; +}; + + +void vf_register_fastpath( struct vertex_fetch *vtx, + GLboolean match_strides ); + +void vf_generic_emit( struct vertex_fetch *vf, + GLuint count, + GLubyte *v ); + +void vf_generate_hardwired_emit( struct vertex_fetch *vf ); + +void vf_generate_sse_emit( struct vertex_fetch *vf ); + + +struct vf_format_info { + const char *name; + vf_extract_func extract; + vf_insert_func insert[4]; + const GLuint attrsize; +}; + +const struct vf_format_info vf_format_info[EMIT_MAX]; + + +#endif diff --git a/src/mesa/vf/vf_generic.c b/src/mesa/vf/vf_generic.c new file mode 100644 index 0000000000..68d8d0897b --- /dev/null +++ b/src/mesa/vf/vf_generic.c @@ -0,0 +1,981 @@ + +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + +#include "glheader.h" +#include "context.h" +#include "colormac.h" +#include "simple_list.h" + +#include "vf/vf.h" + + +/* + * These functions take the NDC coordinates pointed to by 'in', apply the + * NDC->Viewport mapping and store the results at 'v'. + */ + +static INLINE void insert_4f_viewport_4( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = scale[2] * in[2] + trans[2]; + out[3] = in[3]; +} + +static INLINE void insert_4f_viewport_3( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = scale[2] * in[2] + trans[2]; + out[3] = 1; +} + +static INLINE void insert_4f_viewport_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = trans[2]; + out[3] = 1; +} + +static INLINE void insert_4f_viewport_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = trans[1]; + out[2] = trans[2]; + out[3] = 1; +} + +static INLINE void insert_3f_viewport_3( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = scale[2] * in[2] + trans[2]; +} + +static INLINE void insert_3f_viewport_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = scale[2] * in[2] + trans[2]; +} + +static INLINE void insert_3f_viewport_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = trans[1]; + out[2] = trans[2]; +} + +static INLINE void insert_2f_viewport_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; +} + +static INLINE void insert_2f_viewport_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = trans[1]; +} + + +/* + * These functions do the same as above, except for the viewport mapping. + */ + +static INLINE void insert_4f_4( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = in[3]; +} + +static INLINE void insert_4f_3( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = 1; +} + +static INLINE void insert_4f_2( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; + out[3] = 1; +} + +static INLINE void insert_4f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; + out[3] = 1; +} + +static INLINE void insert_3f_xyw_4( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[3]; +} + +static INLINE void insert_3f_xyw_err( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + (void) a; (void) v; (void) in; + _mesa_exit(1); +} + +static INLINE void insert_3f_3( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; +} + +static INLINE void insert_3f_2( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; +} + +static INLINE void insert_3f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; +} + + +static INLINE void insert_2f_2( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; +} + +static INLINE void insert_2f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; +} + +static INLINE void insert_1f_1( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + GLfloat *out = (GLfloat *)(v); + (void) a; + + out[0] = in[0]; +} + +static INLINE void insert_null( const struct vf_attr *a, GLubyte *v, const GLfloat *in ) +{ + (void) a; (void) v; (void) in; +} + +static INLINE void insert_4chan_4f_rgba_4( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLchan *c = (GLchan *)v; + (void) a; + UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); + UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); + UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); + UNCLAMPED_FLOAT_TO_CHAN(c[3], in[3]); +} + +static INLINE void insert_4chan_4f_rgba_3( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLchan *c = (GLchan *)v; + (void) a; + UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); + UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); + UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); + c[3] = CHAN_MAX; +} + +static INLINE void insert_4chan_4f_rgba_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLchan *c = (GLchan *)v; + (void) a; + UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); + UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); + c[2] = 0; + c[3] = CHAN_MAX; +} + +static INLINE void insert_4chan_4f_rgba_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + GLchan *c = (GLchan *)v; + (void) a; + UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); + c[1] = 0; + c[2] = 0; + c[3] = CHAN_MAX; +} + +static INLINE void insert_4ub_4f_rgba_4( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); +} + +static INLINE void insert_4ub_4f_rgba_3( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_rgba_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[2] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_rgba_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + v[1] = 0; + v[2] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_4( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); +} + +static INLINE void insert_4ub_4f_bgra_3( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[0] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + v[1] = 0; + v[0] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_4( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); +} + +static INLINE void insert_4ub_4f_argb_3( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + v[3] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + v[2] = 0x00; + v[3] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_4( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); +} + +static INLINE void insert_4ub_4f_abgr_3( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + v[1] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + v[2] = 0x00; + v[1] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_3ub_3f_rgb_3( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); +} + +static INLINE void insert_3ub_3f_rgb_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[2] = 0; +} + +static INLINE void insert_3ub_3f_rgb_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + v[1] = 0; + v[2] = 0; +} + +static INLINE void insert_3ub_3f_bgr_3( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); +} + +static INLINE void insert_3ub_3f_bgr_2( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[0] = 0; +} + +static INLINE void insert_3ub_3f_bgr_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + v[1] = 0; + v[0] = 0; +} + + +static INLINE void insert_1ub_1f_1( const struct vf_attr *a, GLubyte *v, + const GLfloat *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); +} + + +/*********************************************************************** + * Functions to perform the reverse operations to the above, for + * swrast translation and clip-interpolation. + * + * Currently always extracts a full 4 floats. + */ + +static void extract_4f_viewport( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + const GLfloat *in = (const GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + /* Although included for completeness, the position coordinate is + * usually handled differently during clipping. + */ + out[0] = (in[0] - trans[0]) / scale[0]; + out[1] = (in[1] - trans[1]) / scale[1]; + out[2] = (in[2] - trans[2]) / scale[2]; + out[3] = in[3]; +} + +static void extract_3f_viewport( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + const GLfloat *in = (const GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = (in[0] - trans[0]) / scale[0]; + out[1] = (in[1] - trans[1]) / scale[1]; + out[2] = (in[2] - trans[2]) / scale[2]; + out[3] = 1; +} + + +static void extract_2f_viewport( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + const GLfloat *in = (const GLfloat *)v; + const GLfloat *scale = a->vf->vp; + const GLfloat *trans = a->vf->vp + 4; + + out[0] = (in[0] - trans[0]) / scale[0]; + out[1] = (in[1] - trans[1]) / scale[1]; + out[2] = 0; + out[3] = 1; +} + + +static void extract_4f( const struct vf_attr *a, GLfloat *out, const GLubyte *v ) +{ + const GLfloat *in = (const GLfloat *)v; + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = in[3]; +} + +static void extract_3f_xyw( const struct vf_attr *a, GLfloat *out, const GLubyte *v ) +{ + const GLfloat *in = (const GLfloat *)v; + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; + out[3] = in[2]; +} + + +static void extract_3f( const struct vf_attr *a, GLfloat *out, const GLubyte *v ) +{ + const GLfloat *in = (const GLfloat *)v; + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = 1; +} + + +static void extract_2f( const struct vf_attr *a, GLfloat *out, const GLubyte *v ) +{ + const GLfloat *in = (const GLfloat *)v; + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; + out[3] = 1; +} + +static void extract_1f( const struct vf_attr *a, GLfloat *out, const GLubyte *v ) +{ + const GLfloat *in = (const GLfloat *)v; + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; + out[3] = 1; +} + +static void extract_4chan_4f_rgba( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + GLchan *c = (GLchan *)v; + (void) a; + + out[0] = CHAN_TO_FLOAT(c[0]); + out[1] = CHAN_TO_FLOAT(c[1]); + out[2] = CHAN_TO_FLOAT(c[2]); + out[3] = CHAN_TO_FLOAT(c[3]); +} + +static void extract_4ub_4f_rgba( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + (void) a; + out[0] = UBYTE_TO_FLOAT(v[0]); + out[1] = UBYTE_TO_FLOAT(v[1]); + out[2] = UBYTE_TO_FLOAT(v[2]); + out[3] = UBYTE_TO_FLOAT(v[3]); +} + +static void extract_4ub_4f_bgra( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + (void) a; + out[2] = UBYTE_TO_FLOAT(v[0]); + out[1] = UBYTE_TO_FLOAT(v[1]); + out[0] = UBYTE_TO_FLOAT(v[2]); + out[3] = UBYTE_TO_FLOAT(v[3]); +} + +static void extract_4ub_4f_argb( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + (void) a; + out[3] = UBYTE_TO_FLOAT(v[0]); + out[0] = UBYTE_TO_FLOAT(v[1]); + out[1] = UBYTE_TO_FLOAT(v[2]); + out[2] = UBYTE_TO_FLOAT(v[3]); +} + +static void extract_4ub_4f_abgr( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + (void) a; + out[3] = UBYTE_TO_FLOAT(v[0]); + out[2] = UBYTE_TO_FLOAT(v[1]); + out[1] = UBYTE_TO_FLOAT(v[2]); + out[0] = UBYTE_TO_FLOAT(v[3]); +} + +static void extract_3ub_3f_rgb( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + (void) a; + out[0] = UBYTE_TO_FLOAT(v[0]); + out[1] = UBYTE_TO_FLOAT(v[1]); + out[2] = UBYTE_TO_FLOAT(v[2]); + out[3] = 1; +} + +static void extract_3ub_3f_bgr( const struct vf_attr *a, GLfloat *out, + const GLubyte *v ) +{ + (void) a; + out[2] = UBYTE_TO_FLOAT(v[0]); + out[1] = UBYTE_TO_FLOAT(v[1]); + out[0] = UBYTE_TO_FLOAT(v[2]); + out[3] = 1; +} + +static void extract_1ub_1f( const struct vf_attr *a, GLfloat *out, const GLubyte *v ) +{ + (void) a; + out[0] = UBYTE_TO_FLOAT(v[0]); + out[1] = 0; + out[2] = 0; + out[3] = 1; +} + + +const struct vf_format_info vf_format_info[EMIT_MAX] = +{ + { "1f", + extract_1f, + { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, + sizeof(GLfloat) }, + + { "2f", + extract_2f, + { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, + 2 * sizeof(GLfloat) }, + + { "3f", + extract_3f, + { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, + 3 * sizeof(GLfloat) }, + + { "4f", + extract_4f, + { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, + 4 * sizeof(GLfloat) }, + + { "2f_viewport", + extract_2f_viewport, + { insert_2f_viewport_1, insert_2f_viewport_2, insert_2f_viewport_2, + insert_2f_viewport_2 }, + 2 * sizeof(GLfloat) }, + + { "3f_viewport", + extract_3f_viewport, + { insert_3f_viewport_1, insert_3f_viewport_2, insert_3f_viewport_3, + insert_3f_viewport_3 }, + 3 * sizeof(GLfloat) }, + + { "4f_viewport", + extract_4f_viewport, + { insert_4f_viewport_1, insert_4f_viewport_2, insert_4f_viewport_3, + insert_4f_viewport_4 }, + 4 * sizeof(GLfloat) }, + + { "3f_xyw", + extract_3f_xyw, + { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, + insert_3f_xyw_4 }, + 3 * sizeof(GLfloat) }, + + { "1ub_1f", + extract_1ub_1f, + { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 }, + sizeof(GLubyte) }, + + { "3ub_3f_rgb", + extract_3ub_3f_rgb, + { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3, + insert_3ub_3f_rgb_3 }, + 3 * sizeof(GLubyte) }, + + { "3ub_3f_bgr", + extract_3ub_3f_bgr, + { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3, + insert_3ub_3f_bgr_3 }, + 3 * sizeof(GLubyte) }, + + { "4ub_4f_rgba", + extract_4ub_4f_rgba, + { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, + insert_4ub_4f_rgba_4 }, + 4 * sizeof(GLubyte) }, + + { "4ub_4f_bgra", + extract_4ub_4f_bgra, + { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3, + insert_4ub_4f_bgra_4 }, + 4 * sizeof(GLubyte) }, + + { "4ub_4f_argb", + extract_4ub_4f_argb, + { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3, + insert_4ub_4f_argb_4 }, + 4 * sizeof(GLubyte) }, + + { "4ub_4f_abgr", + extract_4ub_4f_abgr, + { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3, + insert_4ub_4f_abgr_4 }, + 4 * sizeof(GLubyte) }, + + { "4chan_4f_rgba", + extract_4chan_4f_rgba, + { insert_4chan_4f_rgba_1, insert_4chan_4f_rgba_2, insert_4chan_4f_rgba_3, + insert_4chan_4f_rgba_4 }, + 4 * sizeof(GLchan) }, + + { "pad", + NULL, + { NULL, NULL, NULL, NULL }, + 0 } + +}; + + + + +/*********************************************************************** + * Hardwired fastpaths for emitting whole vertices or groups of + * vertices + */ +#define EMIT5(NR, F0, F1, F2, F3, F4, NAME) \ +static void NAME( struct vertex_fetch *vf, \ + GLuint count, \ + GLubyte *v ) \ +{ \ + struct vf_attr *a = vf->attr; \ + GLuint i; \ + \ + for (i = 0 ; i < count ; i++, v += vf->vertex_stride) { \ + if (NR > 0) { \ + F0( &a[0], v + a[0].vertoffset, (GLfloat *)a[0].inputptr ); \ + a[0].inputptr += a[0].inputstride; \ + } \ + \ + if (NR > 1) { \ + F1( &a[1], v + a[1].vertoffset, (GLfloat *)a[1].inputptr ); \ + a[1].inputptr += a[1].inputstride; \ + } \ + \ + if (NR > 2) { \ + F2( &a[2], v + a[2].vertoffset, (GLfloat *)a[2].inputptr ); \ + a[2].inputptr += a[2].inputstride; \ + } \ + \ + if (NR > 3) { \ + F3( &a[3], v + a[3].vertoffset, (GLfloat *)a[3].inputptr ); \ + a[3].inputptr += a[3].inputstride; \ + } \ + \ + if (NR > 4) { \ + F4( &a[4], v + a[4].vertoffset, (GLfloat *)a[4].inputptr ); \ + a[4].inputptr += a[4].inputstride; \ + } \ + } \ +} + + +#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \ + insert_null, insert_null, NAME) + +#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \ + insert_null, NAME) + +#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \ + insert_null, NAME) + + +EMIT2(insert_3f_viewport_3, insert_4ub_4f_rgba_4, emit_viewport3_rgba4) +EMIT2(insert_3f_viewport_3, insert_4ub_4f_bgra_4, emit_viewport3_bgra4) +EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4) + +EMIT3(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_viewport4_rgba4_st2) +EMIT3(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2, emit_viewport4_bgra4_st2) +EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2) + +EMIT4(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_viewport4_rgba4_st2_st2) +EMIT4(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2, insert_2f_2, emit_viewport4_bgra4_st2_st2) +EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2) + + +/* Use the codegen paths to select one of a number of hardwired + * fastpaths. + */ +void vf_generate_hardwired_emit( struct vertex_fetch *vf ) +{ + vf_emit_func func = NULL; + + /* Does it fit a hardwired fastpath? Help! this is growing out of + * control! + */ + switch (vf->attr_count) { + case 2: + if (vf->attr[0].do_insert == insert_3f_viewport_3) { + if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4) + func = emit_viewport3_bgra4; + else if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) + func = emit_viewport3_rgba4; + } + else if (vf->attr[0].do_insert == insert_3f_3 && + vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + func = emit_xyz3_rgba4; + } + break; + case 3: + if (vf->attr[2].do_insert == insert_2f_2) { + if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + if (vf->attr[0].do_insert == insert_4f_viewport_4) + func = emit_viewport4_rgba4_st2; + else if (vf->attr[0].do_insert == insert_4f_4) + func = emit_xyzw4_rgba4_st2; + } + else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 && + vf->attr[0].do_insert == insert_4f_viewport_4) + func = emit_viewport4_bgra4_st2; + } + break; + case 4: + if (vf->attr[2].do_insert == insert_2f_2 && + vf->attr[3].do_insert == insert_2f_2) { + if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + if (vf->attr[0].do_insert == insert_4f_viewport_4) + func = emit_viewport4_rgba4_st2_st2; + else if (vf->attr[0].do_insert == insert_4f_4) + func = emit_xyzw4_rgba4_st2_st2; + } + else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 && + vf->attr[0].do_insert == insert_4f_viewport_4) + func = emit_viewport4_bgra4_st2_st2; + } + break; + } + + vf->emit = func; +} + +/*********************************************************************** + * Generic (non-codegen) functions for whole vertices or groups of + * vertices + */ + +void vf_generic_emit( struct vertex_fetch *vf, + GLuint count, + GLubyte *v ) +{ + struct vf_attr *a = vf->attr; + const GLuint attr_count = vf->attr_count; + const GLuint stride = vf->vertex_stride; + GLuint i, j; + + for (i = 0 ; i < count ; i++, v += stride) { + for (j = 0; j < attr_count; j++) { + GLfloat *in = (GLfloat *)a[j].inputptr; + a[j].inputptr += a[j].inputstride; + a[j].do_insert( &a[j], v + a[j].vertoffset, in ); + } + } +} + + diff --git a/src/mesa/vf/vf_sse.c b/src/mesa/vf/vf_sse.c new file mode 100644 index 0000000000..a5d143461b --- /dev/null +++ b/src/mesa/vf/vf_sse.c @@ -0,0 +1,664 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + +#include "glheader.h" +#include "colormac.h" +#include "simple_list.h" +#include "enums.h" + +#include "vf/vf.h" + +#if defined(USE_SSE_ASM) + +#include "x86/rtasm/x86sse.h" +#include "x86/common_x86_asm.h" + + +#define X 0 +#define Y 1 +#define Z 2 +#define W 3 + + +struct x86_program { + struct x86_function func; + + struct vertex_fetch *vf; + GLboolean inputs_safe; + GLboolean outputs_safe; + GLboolean have_sse2; + + struct x86_reg identity; + struct x86_reg chan0; +}; + + +static struct x86_reg get_identity( struct x86_program *p ) +{ + return p->identity; +} + +static void emit_load4f_4( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movups(&p->func, dest, arg0); +} + +static void emit_load4f_3( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Have to jump through some hoops: + * + * c 0 0 0 + * c 0 0 1 + * 0 0 c 1 + * a b c 1 + */ + sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); + sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); + sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) ); + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load4f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Initialize from identity, then pull in low two words: + */ + sse_movups(&p->func, dest, get_identity(p)); + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load4f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Pull in low word, then swizzle in identity */ + sse_movss(&p->func, dest, arg0); + sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); +} + + + +static void emit_load3f_3( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Over-reads by 1 dword - potential SEGV if input is a vertex + * array. + */ + if (p->inputs_safe) { + sse_movups(&p->func, dest, arg0); + } + else { + /* c 0 0 0 + * c c c c + * a b c c + */ + sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); + sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X)); + sse_movlps(&p->func, dest, arg0); + } +} + +static void emit_load3f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_2(p, dest, arg0); +} + +static void emit_load3f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_1(p, dest, arg0); +} + +static void emit_load2f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load2f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_1(p, dest, arg0); +} + +static void emit_load1f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movss(&p->func, dest, arg0); +} + +static void (*load[4][4])( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) = { + { emit_load1f_1, + emit_load1f_1, + emit_load1f_1, + emit_load1f_1 }, + + { emit_load2f_1, + emit_load2f_2, + emit_load2f_2, + emit_load2f_2 }, + + { emit_load3f_1, + emit_load3f_2, + emit_load3f_3, + emit_load3f_3 }, + + { emit_load4f_1, + emit_load4f_2, + emit_load4f_3, + emit_load4f_4 } +}; + +static void emit_load( struct x86_program *p, + struct x86_reg dest, + GLuint sz, + struct x86_reg src, + GLuint src_sz) +{ + load[sz-1][src_sz-1](p, dest, src); +} + +static void emit_store4f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movups(&p->func, dest, arg0); +} + +static void emit_store3f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + if (p->outputs_safe) { + /* Emit the extra dword anyway. This may hurt writecombining, + * may cause other problems. + */ + sse_movups(&p->func, dest, arg0); + } + else { + /* Alternate strategy - emit two, shuffle, emit one. + */ + sse_movlps(&p->func, dest, arg0); + sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */ + sse_movss(&p->func, x86_make_disp(dest,8), arg0); + } +} + +static void emit_store2f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movlps(&p->func, dest, arg0); +} + +static void emit_store1f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movss(&p->func, dest, arg0); +} + + +static void (*store[4])( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) = +{ + emit_store1f, + emit_store2f, + emit_store3f, + emit_store4f +}; + +static void emit_store( struct x86_program *p, + struct x86_reg dest, + GLuint sz, + struct x86_reg temp ) + +{ + store[sz-1](p, dest, temp); +} + +static void emit_pack_store_4ub( struct x86_program *p, + struct x86_reg dest, + struct x86_reg temp ) +{ + /* Scale by 255.0 + */ + sse_mulps(&p->func, temp, p->chan0); + + if (p->have_sse2) { + sse2_cvtps2dq(&p->func, temp, temp); + sse2_packssdw(&p->func, temp, temp); + sse2_packuswb(&p->func, temp, temp); + sse_movss(&p->func, dest, temp); + } + else { + struct x86_reg mmx0 = x86_make_reg(file_MMX, 0); + struct x86_reg mmx1 = x86_make_reg(file_MMX, 1); + sse_cvtps2pi(&p->func, mmx0, temp); + sse_movhlps(&p->func, temp, temp); + sse_cvtps2pi(&p->func, mmx1, temp); + mmx_packssdw(&p->func, mmx0, mmx1); + mmx_packuswb(&p->func, mmx0, mmx0); + mmx_movd(&p->func, dest, mmx0); + } +} + +static GLint get_offset( const void *a, const void *b ) +{ + return (const char *)b - (const char *)a; +} + +/* Not much happens here. Eventually use this function to try and + * avoid saving/reloading the source pointers each vertex (if some of + * them can fit in registers). + */ +static void get_src_ptr( struct x86_program *p, + struct x86_reg srcREG, + struct x86_reg vfREG, + struct vf_attr *a ) +{ + struct vertex_fetch *vf = p->vf; + struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); + + /* Load current a[j].inputptr + */ + x86_mov(&p->func, srcREG, ptr_to_src); +} + +static void update_src_ptr( struct x86_program *p, + struct x86_reg srcREG, + struct x86_reg vfREG, + struct vf_attr *a ) +{ + if (a->inputstride) { + struct vertex_fetch *vf = p->vf; + struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); + + /* add a[j].inputstride (hardcoded value - could just as easily + * pull the stride value from memory each time). + */ + x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride)); + + /* save new value of a[j].inputptr + */ + x86_mov(&p->func, ptr_to_src, srcREG); + } +} + + +/* Lots of hardcoding + * + * EAX -- pointer to current output vertex + * ECX -- pointer to current attribute + * + */ +static GLboolean build_vertex_emit( struct x86_program *p ) +{ + struct vertex_fetch *vf = p->vf; + GLuint j = 0; + + struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX); + struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX); + struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); + struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI); + struct x86_reg temp = x86_make_reg(file_XMM, 0); + struct x86_reg vp0 = x86_make_reg(file_XMM, 1); + struct x86_reg vp1 = x86_make_reg(file_XMM, 2); + GLubyte *fixup, *label; + + x86_init_func(&p->func); + + /* Push a few regs? + */ + x86_push(&p->func, countEBP); + x86_push(&p->func, vfESI); + + + /* Get vertex count, compare to zero + */ + x86_xor(&p->func, srcECX, srcECX); + x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2)); + x86_cmp(&p->func, countEBP, srcECX); + fixup = x86_jcc_forward(&p->func, cc_E); + + /* Initialize destination register. + */ + x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3)); + + /* Move argument 1 (vf) into a reg: + */ + x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1)); + + + /* Possibly load vp0, vp1 for viewport calcs: + */ + if (vf->allow_viewport_emits) { + sse_movups(&p->func, vp0, x86_make_disp(vfESI, get_offset(vf, &vf->vp[0]))); + sse_movups(&p->func, vp1, x86_make_disp(vfESI, get_offset(vf, &vf->vp[4]))); + } + + /* always load, needed or not: + */ + sse_movups(&p->func, p->chan0, x86_make_disp(vfESI, get_offset(vf, &vf->chan_scale[0]))); + sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0]))); + + /* Note address for loop jump */ + label = x86_get_label(&p->func); + + /* Emit code for each of the attributes. Currently routes + * everything through SSE registers, even when it might be more + * efficient to stick with regular old x86. No optimization or + * other tricks - enough new ground to cover here just getting + * things working. + */ + while (j < vf->attr_count) { + struct vf_attr *a = &vf->attr[j]; + struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset); + + /* Now, load an XMM reg from src, perhaps transform, then save. + * Could be shortcircuited in specific cases: + */ + switch (a->format) { + case EMIT_1F: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 1, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_2F: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 2, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_3F: + /* Potentially the worst case - hardcode 2+1 copying: + */ + if (0) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + } + else { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 2, temp); + if (a->inputsize > 2) { + emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1); + emit_store(p, x86_make_disp(dest,8), 1, temp); + } + else { + sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p)); + } + update_src_ptr(p, srcECX, vfESI, a); + } + break; + case EMIT_4F: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 4, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_2F_VIEWPORT: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + sse_mulps(&p->func, temp, vp0); + sse_addps(&p->func, temp, vp1); + emit_store(p, dest, 2, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_3F_VIEWPORT: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + sse_mulps(&p->func, temp, vp0); + sse_addps(&p->func, temp, vp1); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4F_VIEWPORT: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_mulps(&p->func, temp, vp0); + sse_addps(&p->func, temp, vp1); + emit_store(p, dest, 4, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_3F_XYW: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z)); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + + case EMIT_1UB_1F: + /* Test for PAD3 + 1UB: + */ + if (j > 0 && + a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3) + { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X)); + emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */ + update_src_ptr(p, srcECX, vfESI, a); + } + else { + _mesa_printf("Can't emit 1ub %x %x %d\n", a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); + return GL_FALSE; + } + break; + case EMIT_3UB_3F_RGB: + case EMIT_3UB_3F_BGR: + /* Test for 3UB + PAD1: + */ + if (j == vf->attr_count - 1 || + a[1].vertoffset >= a->vertoffset + 4) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + if (a->format == EMIT_3UB_3F_BGR) + sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + } + /* Test for 3UB + 1UB: + */ + else if (j < vf->attr_count - 1 && + a[1].format == EMIT_1UB_1F && + a[1].vertoffset == a->vertoffset + 3) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + update_src_ptr(p, srcECX, vfESI, a); + + /* Make room for incoming value: + */ + sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); + + get_src_ptr(p, srcECX, vfESI, &a[1]); + emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize); + update_src_ptr(p, srcECX, vfESI, &a[1]); + + /* Rearrange and possibly do BGR conversion: + */ + if (a->format == EMIT_3UB_3F_BGR) + sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); + else + sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X)); + + emit_pack_store_4ub(p, dest, temp); + j++; /* NOTE: two attrs consumed */ + } + else { + _mesa_printf("Can't emit 3ub\n"); + } + return GL_FALSE; /* add this later */ + break; + + case EMIT_4UB_4F_RGBA: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4UB_4F_BGRA: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4UB_4F_ARGB: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4UB_4F_ABGR: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4CHAN_4F_RGBA: + switch (CHAN_TYPE) { + case GL_UNSIGNED_BYTE: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case GL_FLOAT: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 4, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case GL_UNSIGNED_SHORT: + default: + _mesa_printf("unknown CHAN_TYPE %s\n", _mesa_lookup_enum_by_nr(CHAN_TYPE)); + return GL_FALSE; + } + break; + default: + _mesa_printf("unknown a[%d].format %d\n", j, a->format); + return GL_FALSE; /* catch any new opcodes */ + } + + /* Increment j by at least 1 - may have been incremented above also: + */ + j++; + } + + /* Next vertex: + */ + x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vf->vertex_stride)); + + /* decr count, loop if not zero + */ + x86_dec(&p->func, countEBP); + x86_test(&p->func, countEBP, countEBP); + x86_jcc(&p->func, cc_NZ, label); + + /* Exit mmx state? + */ + if (p->func.need_emms) + mmx_emms(&p->func); + + /* Land forward jump here: + */ + x86_fixup_fwd_jump(&p->func, fixup); + + /* Pop regs and return + */ + x86_pop(&p->func, x86_get_base_reg(vfESI)); + x86_pop(&p->func, countEBP); + x86_ret(&p->func); + + vf->emit = (vf_emit_func)x86_get_func(&p->func); + return GL_TRUE; +} + + + +void vf_generate_sse_emit( struct vertex_fetch *vf ) +{ + struct x86_program p; + + if (!cpu_has_xmm) { + vf->codegen_emit = NULL; + return; + } + + _mesa_memset(&p, 0, sizeof(p)); + + p.vf = vf; + p.inputs_safe = 0; /* for now */ + p.outputs_safe = 1; /* for now */ + p.have_sse2 = cpu_has_xmm2; + p.identity = x86_make_reg(file_XMM, 6); + p.chan0 = x86_make_reg(file_XMM, 7); + + x86_init_func(&p.func); + + if (build_vertex_emit(&p)) { + vf_register_fastpath( vf, GL_TRUE ); + } + else { + /* Note the failure so that we don't keep trying to codegen an + * impossible state: + */ + vf_register_fastpath( vf, GL_FALSE ); + x86_release_func(&p.func); + } +} + +#else + +void vf_generate_sse_emit( struct vertex_fetch *vf ) +{ + /* Dummy version for when USE_SSE_ASM not defined */ +} + +#endif -- cgit v1.2.3 From f889ad21da2009f8d49d53a3fa03795e826fb56a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 May 2007 14:36:02 -0600 Subject: define, use SP_MAX_CLIP_PLANES --- src/mesa/softpipe/sp_state.h | 4 +++- src/mesa/softpipe/state_tracker/st_atom_clip.c | 2 +- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/sp_state.h b/src/mesa/softpipe/sp_state.h index d69f9cd164..f8521beed3 100644 --- a/src/mesa/softpipe/sp_state.h +++ b/src/mesa/softpipe/sp_state.h @@ -81,8 +81,10 @@ struct softpipe_scissor_rect { }; +#define SP_MAX_CLIP_PLANES 6 + struct softpipe_clip_state { - GLfloat ucp[6][4]; + GLfloat ucp[SP_MAX_CLIP_PLANES][4]; GLuint nr; }; diff --git a/src/mesa/softpipe/state_tracker/st_atom_clip.c b/src/mesa/softpipe/state_tracker/st_atom_clip.c index 1a0d2c4053..710d6ffc84 100644 --- a/src/mesa/softpipe/state_tracker/st_atom_clip.c +++ b/src/mesa/softpipe/state_tracker/st_atom_clip.c @@ -45,7 +45,7 @@ static void update_clip( struct st_context *st ) memset(&clip, 0, sizeof(clip)); - for (i = 0; i < 6; i++) { + for (i = 0; i < SP_MAX_CLIP_PLANES; i++) { if (st->ctx->Transform.ClipPlanesEnabled & (1 << i)) { memcpy(clip.ucp[clip.nr], st->ctx->Transform._ClipUserPlane[i], -- cgit v1.2.3 From 711af67fe4afa7d982649c101d68474be9294e07 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 May 2007 14:36:48 -0600 Subject: scissor state --- src/mesa/softpipe/state_tracker/st_atom_scissor.c | 69 +++++++++++++++++++++++ 1 file changed, 69 insertions(+) create mode 100644 src/mesa/softpipe/state_tracker/st_atom_scissor.c (limited to 'src') diff --git a/src/mesa/softpipe/state_tracker/st_atom_scissor.c b/src/mesa/softpipe/state_tracker/st_atom_scissor.c new file mode 100644 index 0000000000..74f5cc1a2b --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom_scissor.c @@ -0,0 +1,69 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + + +/* scissor state + */ +static void update_scissor( struct st_context *st ) +{ + struct softpipe_scissor_rect scissor; + + scissor.minx = st->ctx->Scissor.X; + scissor.miny = st->ctx->Scissor.Y; + scissor.maxx = st->ctx->Scissor.X + st->ctx->Scissor.Width; + scissor.maxy = st->ctx->Scissor.Y + st->ctx->Scissor.Height; + + if (memcmp(&scissor, &st->state.scissor, sizeof(scissor)) != 0) { + /* state has changed */ + st->state.scissor = scissor; /* struct copy */ + st->softpipe->set_scissor_rect(st->softpipe, &scissor); /* set new state */ + } +} + + +const struct st_tracked_state st_update_scissor = { + .dirty = { + .mesa = (_NEW_SCISSOR), + .st = 0, + }, + .update = update_scissor +}; + + + + + -- cgit v1.2.3 From 3bf0fd60905d18cc6fef7845d68e0c5b530f531b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 May 2007 14:37:00 -0600 Subject: add scissor state --- src/mesa/softpipe/state_tracker/st_atom.c | 9 +++++---- src/mesa/softpipe/state_tracker/st_atom.h | 1 + 2 files changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/state_tracker/st_atom.c b/src/mesa/softpipe/state_tracker/st_atom.c index 99fba398bf..1eb5c86e46 100644 --- a/src/mesa/softpipe/state_tracker/st_atom.c +++ b/src/mesa/softpipe/state_tracker/st_atom.c @@ -47,6 +47,7 @@ static const struct st_tracked_state *atoms[] = &st_update_fs, &st_update_setup, &st_update_viewport, + &st_update_scissor, /* will be patched out at runtime */ /* &st_update_constants */ }; @@ -58,7 +59,7 @@ void st_init_atoms( struct st_context *st ) st->atoms = _mesa_malloc(sizeof(atoms)); st->nr_atoms = sizeof(atoms)/sizeof(*atoms); - _mesa_memcpy(st->atoms, atoms, sizeof(atoms)); + memcpy(st->atoms, atoms, sizeof(atoms)); /* Patch in a pointer to the dynamic state atom: */ @@ -66,9 +67,9 @@ void st_init_atoms( struct st_context *st ) if (st->atoms[i] == &st_update_constants) st->atoms[i] = &st->constants.tracked_state; - _mesa_memcpy(&st->constants.tracked_state, - &st_update_constants, - sizeof(st_update_constants)); + memcpy(&st->constants.tracked_state, + &st_update_constants, + sizeof(st_update_constants)); } diff --git a/src/mesa/softpipe/state_tracker/st_atom.h b/src/mesa/softpipe/state_tracker/st_atom.h index 87b72940b4..ee3d36a00c 100644 --- a/src/mesa/softpipe/state_tracker/st_atom.h +++ b/src/mesa/softpipe/state_tracker/st_atom.h @@ -50,6 +50,7 @@ const struct st_tracked_state st_update_fs; const struct st_tracked_state st_update_setup; const struct st_tracked_state st_update_viewport; const struct st_tracked_state st_update_constants; +const struct st_tracked_state st_update_scissor; #endif -- cgit v1.2.3 From 9267341bd01900bc78700928e5836d0f52572e38 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 May 2007 16:48:38 -0600 Subject: scissor depends on glScissor() and framebuffer size --- src/mesa/softpipe/state_tracker/st_atom_scissor.c | 37 +++++++++++++++++------ 1 file changed, 28 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/state_tracker/st_atom_scissor.c b/src/mesa/softpipe/state_tracker/st_atom_scissor.c index 74f5cc1a2b..105c2a6dd9 100644 --- a/src/mesa/softpipe/state_tracker/st_atom_scissor.c +++ b/src/mesa/softpipe/state_tracker/st_atom_scissor.c @@ -36,28 +36,47 @@ #include "st_atom.h" -/* scissor state +/** + * Scissor depends on the scissor box, and the framebuffer dimensions. */ -static void update_scissor( struct st_context *st ) +static void +update_scissor( struct st_context *st ) { struct softpipe_scissor_rect scissor; + const struct gl_framebuffer *fb = st->ctx->DrawBuffer; + + scissor.minx = 0; + scissor.miny = 0; + scissor.maxx = fb->Width; + scissor.maxy = fb->Height; + + if (st->ctx->Scissor.Enabled) { + if (st->ctx->Scissor.X > scissor.minx) + scissor.minx = st->ctx->Scissor.X; + if (st->ctx->Scissor.Y > scissor.miny) + scissor.miny = st->ctx->Scissor.Y; + + if (st->ctx->Scissor.X + st->ctx->Scissor.Width < scissor.maxx) + scissor.maxx = st->ctx->Scissor.X + st->ctx->Scissor.Width; + if (st->ctx->Scissor.Y + st->ctx->Scissor.Height < scissor.maxy) + scissor.maxy = st->ctx->Scissor.Y + st->ctx->Scissor.Height; + + /* check for null space */ + if (scissor.minx >= scissor.maxx || scissor.miny >= scissor.maxy) + scissor.minx = scissor.miny = scissor.maxx = scissor.maxy = 0; + } - scissor.minx = st->ctx->Scissor.X; - scissor.miny = st->ctx->Scissor.Y; - scissor.maxx = st->ctx->Scissor.X + st->ctx->Scissor.Width; - scissor.maxy = st->ctx->Scissor.Y + st->ctx->Scissor.Height; - if (memcmp(&scissor, &st->state.scissor, sizeof(scissor)) != 0) { /* state has changed */ st->state.scissor = scissor; /* struct copy */ - st->softpipe->set_scissor_rect(st->softpipe, &scissor); /* set new state */ + st->softpipe->set_scissor_rect(st->softpipe, &scissor); /* activate */ } } const struct st_tracked_state st_update_scissor = { .dirty = { - .mesa = (_NEW_SCISSOR), + .mesa = (_NEW_SCISSOR | _NEW_BUFFERS), .st = 0, }, .update = update_scissor -- cgit v1.2.3 From 7234550cf759fdf3ab325465090c7f5eb3b5d320 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 May 2007 16:49:27 -0600 Subject: #ifndef SP_DEFINES_H protection, copyright --- src/mesa/softpipe/sp_defines.h | 31 ++++++++++++++++++++++++++++++- 1 file changed, 30 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/softpipe/sp_defines.h b/src/mesa/softpipe/sp_defines.h index d8ae8d4baf..f47d782285 100644 --- a/src/mesa/softpipe/sp_defines.h +++ b/src/mesa/softpipe/sp_defines.h @@ -1,3 +1,32 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_DEFINES_H +#define SP_DEFINES_H #define SP_BLENDFACTOR_ONE 0x1 #define SP_BLENDFACTOR_SRC_COLOR 0x2 @@ -25,7 +54,6 @@ #define SP_BLEND_MIN 3 #define SP_BLEND_MAX 4 - #define SP_LOGICOP_CLEAR 0 #define SP_LOGICOP_NOR 1 #define SP_LOGICOP_AND_INVERTED 2 @@ -43,3 +71,4 @@ #define SP_LOGICOP_OR 14 #define SP_LOGICOP_SET 15 +#endif -- cgit v1.2.3 From 6debc8078a7170c26d826e62e2b6542e85813a6c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 May 2007 16:50:22 -0600 Subject: Implement blend state --- src/mesa/softpipe/sp_context.h | 3 + src/mesa/softpipe/sp_state.h | 15 +- src/mesa/softpipe/state_tracker/st_atom.c | 1 + src/mesa/softpipe/state_tracker/st_atom.h | 1 + src/mesa/softpipe/state_tracker/st_atom_blend.c | 202 ++++++++++++++++++++++++ 5 files changed, 214 insertions(+), 8 deletions(-) create mode 100644 src/mesa/softpipe/state_tracker/st_atom_blend.c (limited to 'src') diff --git a/src/mesa/softpipe/sp_context.h b/src/mesa/softpipe/sp_context.h index a11ad26018..726cb96388 100644 --- a/src/mesa/softpipe/sp_context.h +++ b/src/mesa/softpipe/sp_context.h @@ -52,6 +52,9 @@ struct softpipe_context { void (*set_scissor_rect)( struct softpipe_context *, const struct softpipe_scissor_rect * ); + void (*set_blend_state)( struct softpipe_context *, + const struct softpipe_blend_state * ); + void (*set_fs_state)( struct softpipe_context *, const struct softpipe_fs_state * ); diff --git a/src/mesa/softpipe/sp_state.h b/src/mesa/softpipe/sp_state.h index f8521beed3..29bba916a1 100644 --- a/src/mesa/softpipe/sp_state.h +++ b/src/mesa/softpipe/sp_state.h @@ -102,18 +102,17 @@ struct softpipe_constant_buffer { struct softpipe_blend_state { GLuint blend_enable:1; - GLuint ia_blend_enable:1; - GLuint ia_dest_blend_factor:5; - GLuint ia_src_blend_factor:5; - GLuint ia_blend_function:3; + GLuint rgb_func:3; + GLuint rgb_src_factor:5; + GLuint rgb_dst_factor:5; - GLuint dest_blend_factor:5; - GLuint src_blend_factor:5; - GLuint blend_function:3; + GLuint alpha_func:3; + GLuint alpha_src_factor:5; + GLuint alpha_dst_factor:5; GLuint logicop_enable:1; - GLuint logicop_func:4; + GLuint logicop_func:4; }; struct softpipe_blend_color { diff --git a/src/mesa/softpipe/state_tracker/st_atom.c b/src/mesa/softpipe/state_tracker/st_atom.c index 1eb5c86e46..5f426f040a 100644 --- a/src/mesa/softpipe/state_tracker/st_atom.c +++ b/src/mesa/softpipe/state_tracker/st_atom.c @@ -48,6 +48,7 @@ static const struct st_tracked_state *atoms[] = &st_update_setup, &st_update_viewport, &st_update_scissor, + &st_update_blend, /* will be patched out at runtime */ /* &st_update_constants */ }; diff --git a/src/mesa/softpipe/state_tracker/st_atom.h b/src/mesa/softpipe/state_tracker/st_atom.h index ee3d36a00c..dc8397adce 100644 --- a/src/mesa/softpipe/state_tracker/st_atom.h +++ b/src/mesa/softpipe/state_tracker/st_atom.h @@ -51,6 +51,7 @@ const struct st_tracked_state st_update_setup; const struct st_tracked_state st_update_viewport; const struct st_tracked_state st_update_constants; const struct st_tracked_state st_update_scissor; +const struct st_tracked_state st_update_blend; #endif diff --git a/src/mesa/softpipe/state_tracker/st_atom_blend.c b/src/mesa/softpipe/state_tracker/st_atom_blend.c new file mode 100644 index 0000000000..eabb9172ba --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom_blend.c @@ -0,0 +1,202 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "softpipe/sp_context.h" +#include "softpipe/sp_defines.h" + + +/** + * Convert GLenum blend tokens to softpipe tokens. + * Both blend factors and blend funcs are accepted. + */ +static GLuint +gl_blend_to_sp(GLenum blend) +{ + switch (blend) { + /* blend functions */ + case GL_FUNC_ADD: + return SP_BLEND_ADD; + case GL_FUNC_SUBTRACT: + return SP_BLEND_SUBTRACT; + case GL_FUNC_REVERSE_SUBTRACT: + return SP_BLEND_REVERSE_SUBTRACT; + case GL_MIN: + return SP_BLEND_MIN; + case GL_MAX: + return SP_BLEND_MAX; + + /* blend factors */ + case GL_ONE: + return SP_BLENDFACTOR_ONE; + case GL_SRC_COLOR: + return SP_BLENDFACTOR_SRC_COLOR; + case GL_SRC_ALPHA: + return SP_BLENDFACTOR_SRC_ALPHA; + case GL_DST_ALPHA: + return SP_BLENDFACTOR_DST_ALPHA; + case GL_DST_COLOR: + return SP_BLENDFACTOR_DST_COLOR; + case GL_SRC_ALPHA_SATURATE: + return SP_BLENDFACTOR_SRC_ALPHA_SATURATE; + case GL_CONSTANT_COLOR: + return SP_BLENDFACTOR_CONST_COLOR; + case GL_CONSTANT_ALPHA: + return SP_BLENDFACTOR_CONST_ALPHA; + /* + return SP_BLENDFACTOR_SRC1_COLOR; + return SP_BLENDFACTOR_SRC1_ALPHA; + */ + case GL_ZERO: + return SP_BLENDFACTOR_ZERO; + case GL_ONE_MINUS_SRC_COLOR: + return SP_BLENDFACTOR_INV_SRC_COLOR; + case GL_ONE_MINUS_SRC_ALPHA: + return SP_BLENDFACTOR_INV_SRC_ALPHA; + case GL_ONE_MINUS_DST_COLOR: + return SP_BLENDFACTOR_INV_DST_ALPHA; + case GL_ONE_MINUS_DST_ALPHA: + return SP_BLENDFACTOR_INV_DST_COLOR; + case GL_ONE_MINUS_CONSTANT_COLOR: + return SP_BLENDFACTOR_INV_CONST_COLOR; + case GL_ONE_MINUS_CONSTANT_ALPHA: + return SP_BLENDFACTOR_INV_CONST_ALPHA; + /* + return SP_BLENDFACTOR_INV_SRC1_COLOR; + return SP_BLENDFACTOR_INV_SRC1_ALPHA; + */ + default: + assert("invalid GL token in gl_blend_to_sp()" == NULL); + return 0; + } +} + + +/** + * Convert GLenum logicop tokens to softpipe tokens. + */ +static GLuint +gl_logicop_to_sp(GLenum logicop) +{ + switch (logicop) { + case GL_CLEAR: + return SP_LOGICOP_CLEAR; + case GL_NOR: + return SP_LOGICOP_NOR; + case GL_AND_INVERTED: + return SP_LOGICOP_AND_INVERTED; + case GL_COPY_INVERTED: + return SP_LOGICOP_COPY_INVERTED; + case GL_AND_REVERSE: + return SP_LOGICOP_AND_REVERSE; + case GL_INVERT: + return SP_LOGICOP_INVERT; + case GL_XOR: + return SP_LOGICOP_XOR; + case GL_NAND: + return SP_LOGICOP_NAND; + case GL_AND: + return SP_LOGICOP_AND; + case GL_EQUIV: + return SP_LOGICOP_EQUIV; + case GL_NOOP: + return SP_LOGICOP_NOOP; + case GL_OR_INVERTED: + return SP_LOGICOP_OR_INVERTED; + case GL_COPY: + return SP_LOGICOP_COPY; + case GL_OR_REVERSE: + return SP_LOGICOP_OR_REVERSE; + case GL_OR: + return SP_LOGICOP_OR; + case GL_SET: + return SP_LOGICOP_SET; + default: + assert("invalid GL token in gl_logicop_to_sp()" == NULL); + return 0; + } +} + + +static void +update_blend( struct st_context *st ) +{ + struct softpipe_blend_state blend; + + memset(&blend, 0, sizeof(blend)); + + if (st->ctx->Color.ColorLogicOpEnabled || + (st->ctx->Color.BlendEnabled && + st->ctx->Color.BlendEquationRGB == GL_LOGIC_OP)) { + /* logicop enabled */ + blend.logicop_enable = 1; + blend.logicop_func = gl_logicop_to_sp(st->ctx->Color.LogicOp); + } + else if (st->ctx->Color.BlendEnabled) { + /* blending enabled */ + blend.blend_enable = 1; + + blend.rgb_func = gl_blend_to_sp(st->ctx->Color.BlendEquationRGB); + blend.rgb_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcRGB); + blend.rgb_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstRGB); + + blend.alpha_func = gl_blend_to_sp(st->ctx->Color.BlendEquationA); + blend.alpha_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcA); + blend.alpha_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstA); + } + else { + /* no blending / logicop */ + } + + if (memcmp(&blend, &st->state.blend, sizeof(blend)) != 0) { + /* state has changed */ + st->state.blend = blend; /* struct copy */ + st->softpipe->set_blend_state(st->softpipe, &blend); /* set new state */ + } +} + + +const struct st_tracked_state st_update_blend = { + .dirty = { + .mesa = (_NEW_COLOR), /* XXX _NEW_BLEND someday? */ + .st = 0, + }, + .update = update_blend +}; + + + + + -- cgit v1.2.3 From 2248829b99bf388dfe7ee9c6763f98167b52ba8b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 May 2007 16:50:45 -0600 Subject: added st_atom_blend.c, st_atom_scissor.c --- src/mesa/sources | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 8138fb627d..74c92bb381 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -173,10 +173,12 @@ SOFTPIPE_SOURCES = \ softpipe/generic/g_tile_fs.c \ softpipe/generic/g_tile_output.c \ softpipe/state_tracker/st_atom.c \ + softpipe/state_tracker/st_atom_blend.c \ + softpipe/state_tracker/st_atom_cbuf.c \ softpipe/state_tracker/st_atom_clip.c \ softpipe/state_tracker/st_atom_fs.c \ + softpipe/state_tracker/st_atom_scissor.c \ softpipe/state_tracker/st_atom_setup.c \ - softpipe/state_tracker/st_atom_cbuf.c \ softpipe/state_tracker/st_atom_viewport.c \ softpipe/state_tracker/st_cb_program.c \ softpipe/state_tracker/st_draw.c \ -- cgit v1.2.3 From 008fb50174fb6717f3b71836427bc6be4e44613c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 May 2007 17:37:36 -0600 Subject: Stencil state. --- src/mesa/softpipe/sp_context.h | 3 + src/mesa/softpipe/sp_defines.h | 18 +++ src/mesa/softpipe/sp_state.h | 18 +++ src/mesa/softpipe/state_tracker/st_atom.c | 1 + src/mesa/softpipe/state_tracker/st_atom.h | 1 + src/mesa/softpipe/state_tracker/st_atom_stencil.c | 140 ++++++++++++++++++++++ src/mesa/softpipe/state_tracker/st_context.h | 1 + src/mesa/sources | 1 + 8 files changed, 183 insertions(+) create mode 100644 src/mesa/softpipe/state_tracker/st_atom_stencil.c (limited to 'src') diff --git a/src/mesa/softpipe/sp_context.h b/src/mesa/softpipe/sp_context.h index 726cb96388..e047a9330b 100644 --- a/src/mesa/softpipe/sp_context.h +++ b/src/mesa/softpipe/sp_context.h @@ -52,6 +52,9 @@ struct softpipe_context { void (*set_scissor_rect)( struct softpipe_context *, const struct softpipe_scissor_rect * ); + void (*set_stencil_state)( struct softpipe_context *, + const struct softpipe_stencil_state * ); + void (*set_blend_state)( struct softpipe_context *, const struct softpipe_blend_state * ); diff --git a/src/mesa/softpipe/sp_defines.h b/src/mesa/softpipe/sp_defines.h index f47d782285..bc88e3d71e 100644 --- a/src/mesa/softpipe/sp_defines.h +++ b/src/mesa/softpipe/sp_defines.h @@ -71,4 +71,22 @@ #define SP_LOGICOP_OR 14 #define SP_LOGICOP_SET 15 +#define SP_STENCIL_FUNC_NEVER 0 +#define SP_STENCIL_FUNC_LESS 1 +#define SP_STENCIL_FUNC_EQUAL 2 +#define SP_STENCIL_FUNC_LEQUAL 3 +#define SP_STENCIL_FUNC_GREATER 4 +#define SP_STENCIL_FUNC_NOTEQUAL 5 +#define SP_STENCIL_FUNC_GEQUAL 6 +#define SP_STENCIL_FUNC_ALWAYS 7 + +#define SP_STENCIL_OP_KEEP 0 +#define SP_STENCIL_OP_ZERO 1 +#define SP_STENCIL_OP_REPLACE 2 +#define SP_STENCIL_OP_INCR 3 +#define SP_STENCIL_OP_DECR 4 +#define SP_STENCIL_OP_INCR_WRAP 5 +#define SP_STENCIL_OP_DECR_WRAP 6 +#define SP_STENCIL_OP_INVERT 7 + #endif diff --git a/src/mesa/softpipe/sp_state.h b/src/mesa/softpipe/sp_state.h index 29bba916a1..73ed7a619c 100644 --- a/src/mesa/softpipe/sp_state.h +++ b/src/mesa/softpipe/sp_state.h @@ -120,6 +120,24 @@ struct softpipe_blend_color { }; +struct softpipe_stencil_state { + GLuint front_enabled:1; + GLuint front_func:3; /**< SP_STENCIL_FUNC_x */ + GLuint front_fail_op:3; /**< SP_STENCIL_OP_x */ + GLuint front_zpass_op:3; /**< SP_STENCIL_OP_x */ + GLuint front_zfail_op:3; /**< SP_STENCIL_OP_x */ + GLuint back_enabled:1; + GLuint back_func:3; + GLuint back_fail_op:3; + GLuint back_zpass_op:3; + GLuint back_zfail_op:3; + GLint ref_value[2]; /**< [0] = front, [1] = back */ + GLuint value_mask[2]; + GLuint write_mask[2]; + GLuint clear_value; +}; + + /* This will change for hardware softpipes... */ struct softpipe_surface { diff --git a/src/mesa/softpipe/state_tracker/st_atom.c b/src/mesa/softpipe/state_tracker/st_atom.c index 5f426f040a..fdbfb9021d 100644 --- a/src/mesa/softpipe/state_tracker/st_atom.c +++ b/src/mesa/softpipe/state_tracker/st_atom.c @@ -49,6 +49,7 @@ static const struct st_tracked_state *atoms[] = &st_update_viewport, &st_update_scissor, &st_update_blend, + &st_update_stencil, /* will be patched out at runtime */ /* &st_update_constants */ }; diff --git a/src/mesa/softpipe/state_tracker/st_atom.h b/src/mesa/softpipe/state_tracker/st_atom.h index dc8397adce..8dc7460d11 100644 --- a/src/mesa/softpipe/state_tracker/st_atom.h +++ b/src/mesa/softpipe/state_tracker/st_atom.h @@ -52,6 +52,7 @@ const struct st_tracked_state st_update_viewport; const struct st_tracked_state st_update_constants; const struct st_tracked_state st_update_scissor; const struct st_tracked_state st_update_blend; +const struct st_tracked_state st_update_stencil; #endif diff --git a/src/mesa/softpipe/state_tracker/st_atom_stencil.c b/src/mesa/softpipe/state_tracker/st_atom_stencil.c new file mode 100644 index 0000000000..ac8de03702 --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom_stencil.c @@ -0,0 +1,140 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "softpipe/sp_context.h" +#include "softpipe/sp_defines.h" + + +/** + * Convert GLenum stencil func tokens to softpipe tokens. + */ +static GLuint +gl_stencil_func_to_sp(GLenum func) +{ + /* Same values, just biased */ + assert(SP_STENCIL_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(SP_STENCIL_FUNC_LESS == GL_LESS - GL_NEVER); + assert(SP_STENCIL_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(SP_STENCIL_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(SP_STENCIL_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(SP_STENCIL_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(SP_STENCIL_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(SP_STENCIL_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(func >= GL_NEVER); + assert(func <= GL_ALWAYS); + return func - GL_NEVER; +} + + +/** + * Convert GLenum stencil op tokens to softpipe tokens. + */ +static GLuint +gl_stencil_op_to_sp(GLenum func) +{ + switch (func) { + case GL_KEEP: + return SP_STENCIL_OP_KEEP; + case GL_ZERO: + return SP_STENCIL_OP_ZERO; + case GL_REPLACE: + return SP_STENCIL_OP_REPLACE; + case GL_INCR: + return SP_STENCIL_OP_INCR; + case GL_DECR: + return SP_STENCIL_OP_DECR; + case GL_INCR_WRAP: + return SP_STENCIL_OP_INCR_WRAP; + case GL_DECR_WRAP: + return SP_STENCIL_OP_DECR_WRAP; + case GL_INVERT: + return SP_STENCIL_OP_INVERT; + default: + assert("invalid GL token in gl_stencil_op_to_sp()" == NULL); + return 0; + } +} + + +static void +update_stencil( struct st_context *st ) +{ + struct softpipe_stencil_state stencil; + + memset(&stencil, 0, sizeof(stencil)); + + if (st->ctx->Stencil.Enabled) { + stencil.front_enabled = 1; + stencil.front_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[0]); + stencil.front_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[0]); + stencil.front_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[0]); + stencil.front_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[0]); + stencil.ref_value[0] = st->ctx->Stencil.Ref[0]; + stencil.value_mask[0] = st->ctx->Stencil.ValueMask[0]; + stencil.write_mask[0] = st->ctx->Stencil.WriteMask[0]; + if (st->ctx->Stencil.TestTwoSide) { + stencil.back_enabled = 1; + stencil.back_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[1]); + stencil.back_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[1]); + stencil.back_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[1]); + stencil.back_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[1]); + stencil.ref_value[1] = st->ctx->Stencil.Ref[1]; + stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1]; + stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1]; + } + stencil.clear_value = st->ctx->Stencil.Clear; + } + + if (memcmp(&stencil, &st->state.stencil, sizeof(stencil)) != 0) { + /* state has changed */ + st->state.stencil = stencil; /* struct copy */ + st->softpipe->set_stencil_state(st->softpipe, &stencil); /* set new state */ + } +} + + +const struct st_tracked_state st_update_stencil = { + .dirty = { + .mesa = (_NEW_STENCIL), + .st = 0, + }, + .update = update_stencil +}; + + + + + diff --git a/src/mesa/softpipe/state_tracker/st_context.h b/src/mesa/softpipe/state_tracker/st_context.h index 08cee11797..5ff10c1b1f 100644 --- a/src/mesa/softpipe/state_tracker/st_context.h +++ b/src/mesa/softpipe/state_tracker/st_context.h @@ -73,6 +73,7 @@ struct st_context struct softpipe_clip_state clip; struct softpipe_scissor_rect scissor; struct softpipe_poly_stipple poly_stipple; + struct softpipe_stencil_state stencil; } state; struct { diff --git a/src/mesa/sources b/src/mesa/sources index 74c92bb381..c5ee7d9611 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -178,6 +178,7 @@ SOFTPIPE_SOURCES = \ softpipe/state_tracker/st_atom_clip.c \ softpipe/state_tracker/st_atom_fs.c \ softpipe/state_tracker/st_atom_scissor.c \ + softpipe/state_tracker/st_atom_stencil.c \ softpipe/state_tracker/st_atom_setup.c \ softpipe/state_tracker/st_atom_viewport.c \ softpipe/state_tracker/st_cb_program.c \ -- cgit v1.2.3 From cbe4dbef8e8e9ff29d64ca588e81c2bb079d99f1 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 May 2007 22:16:26 -0600 Subject: use GLubyte for stencil ref/mask/clear --- src/mesa/softpipe/sp_state.h | 8 ++++---- src/mesa/softpipe/state_tracker/st_atom_stencil.c | 14 +++++++------- 2 files changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/sp_state.h b/src/mesa/softpipe/sp_state.h index 73ed7a619c..f7d904bffe 100644 --- a/src/mesa/softpipe/sp_state.h +++ b/src/mesa/softpipe/sp_state.h @@ -131,10 +131,10 @@ struct softpipe_stencil_state { GLuint back_fail_op:3; GLuint back_zpass_op:3; GLuint back_zfail_op:3; - GLint ref_value[2]; /**< [0] = front, [1] = back */ - GLuint value_mask[2]; - GLuint write_mask[2]; - GLuint clear_value; + GLubyte ref_value[2]; /**< [0] = front, [1] = back */ + GLubyte value_mask[2]; + GLubyte write_mask[2]; + GLubyte clear_value; }; diff --git a/src/mesa/softpipe/state_tracker/st_atom_stencil.c b/src/mesa/softpipe/state_tracker/st_atom_stencil.c index ac8de03702..32610c3d24 100644 --- a/src/mesa/softpipe/state_tracker/st_atom_stencil.c +++ b/src/mesa/softpipe/state_tracker/st_atom_stencil.c @@ -102,20 +102,20 @@ update_stencil( struct st_context *st ) stencil.front_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[0]); stencil.front_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[0]); stencil.front_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[0]); - stencil.ref_value[0] = st->ctx->Stencil.Ref[0]; - stencil.value_mask[0] = st->ctx->Stencil.ValueMask[0]; - stencil.write_mask[0] = st->ctx->Stencil.WriteMask[0]; + stencil.ref_value[0] = st->ctx->Stencil.Ref[0] & 0xff; + stencil.value_mask[0] = st->ctx->Stencil.ValueMask[0] & 0xff; + stencil.write_mask[0] = st->ctx->Stencil.WriteMask[0] & 0xff; if (st->ctx->Stencil.TestTwoSide) { stencil.back_enabled = 1; stencil.back_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[1]); stencil.back_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[1]); stencil.back_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[1]); stencil.back_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[1]); - stencil.ref_value[1] = st->ctx->Stencil.Ref[1]; - stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1]; - stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1]; + stencil.ref_value[1] = st->ctx->Stencil.Ref[1] & 0xff; + stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1] & 0xff; + stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1] & 0xff; } - stencil.clear_value = st->ctx->Stencil.Clear; + stencil.clear_value = st->ctx->Stencil.Clear & 0xff; } if (memcmp(&stencil, &st->state.stencil, sizeof(stencil)) != 0) { -- cgit v1.2.3 From 00d5334899010b0e7c0ba53fdc9b112d20f6c34b Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 May 2007 15:45:21 -0600 Subject: Add some comments to explain things, code clarifications. Also, add quad.facing field for front/back facing. See comments for details. --- src/mesa/softpipe/generic/g_headers.h | 1 + src/mesa/softpipe/generic/g_prim_setup.c | 78 ++++++++++++++++++++------------ src/mesa/softpipe/generic/g_tile_fs.c | 29 +++++++++--- 3 files changed, 72 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/generic/g_headers.h b/src/mesa/softpipe/generic/g_headers.h index 1db325a4ab..96ff52a453 100644 --- a/src/mesa/softpipe/generic/g_headers.h +++ b/src/mesa/softpipe/generic/g_headers.h @@ -93,6 +93,7 @@ struct quad_header { GLint x0; GLint y0; GLuint mask; + GLuint facing; /**< Front or back facing? */ struct { GLfloat color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ diff --git a/src/mesa/softpipe/generic/g_prim_setup.c b/src/mesa/softpipe/generic/g_prim_setup.c index 3e9639f7bc..ce93c0a6f9 100644 --- a/src/mesa/softpipe/generic/g_prim_setup.c +++ b/src/mesa/softpipe/generic/g_prim_setup.c @@ -68,35 +68,30 @@ struct setup_stage { struct quad_header quad; struct { - GLint left[2]; + GLint left[2]; /**< [0] = row0, [1] = row1 */ GLint right[2]; GLint y; GLuint y_flags; - GLuint mask; + GLuint mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ } span; }; - +/** + * Basically a cast wrapper. + */ static inline struct setup_stage *setup_stage( struct prim_stage *stage ) { return (struct setup_stage *)stage; } - -static inline GLint _min(GLint x, GLint y) -{ - return x < y ? x : y; -} - -static inline GLint _max(GLint x, GLint y) -{ - return x > y ? x : y; -} - +/** + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. + */ static inline GLint block( GLint x ) { return x & ~1; @@ -110,7 +105,9 @@ static void setup_begin( struct prim_stage *stage ) } - +/** + * Run shader on a quad/block. + */ static void run_shader_block( struct setup_stage *setup, GLint x, GLint y, GLuint mask ) { @@ -122,7 +119,11 @@ static void run_shader_block( struct setup_stage *setup, } -/* this is pretty nasty... may need to rework flush_spans again to +/** + * Compute mask which indicates which pixels in the 2x2 quad are actually inside + * the triangle's bounds. + * + * this is pretty nasty... may need to rework flush_spans again to * fix it, if possible. */ static GLuint calculate_mask( struct setup_stage *setup, @@ -146,6 +147,9 @@ static GLuint calculate_mask( struct setup_stage *setup, } +/** + * Render a horizontal span of quads + */ static void flush_spans( struct setup_stage *setup ) { GLint minleft, maxright; @@ -153,10 +157,8 @@ static void flush_spans( struct setup_stage *setup ) switch (setup->span.y_flags) { case 3: - minleft = _min(setup->span.left[0], - setup->span.left[1]); - maxright = _max(setup->span.right[0], - setup->span.right[1]); + minleft = MIN2(setup->span.left[0], setup->span.left[1]); + maxright = MAX2(setup->span.right[0], setup->span.right[1]); break; case 1: @@ -188,8 +190,10 @@ static void flush_spans( struct setup_stage *setup ) setup->span.right[1] = 0; } - +/** + * Do setup for point rasterization, then render the point. + */ static void setup_point( struct prim_stage *stage, struct prim_header *header ) @@ -197,6 +201,9 @@ setup_point( struct prim_stage *stage, } +/** + * Do setup for line rasterization, then render the line. + */ static void setup_line( struct prim_stage *stage, struct prim_header *header ) @@ -204,10 +211,6 @@ setup_line( struct prim_stage *stage, } - - - - static GLboolean setup_sort_vertices( struct setup_stage *setup, const struct prim_header *prim ) { @@ -288,6 +291,11 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, setup->oneoverarea = 1.0 / area; } + /* XXX need to know if this is a front or back-facing triangle: + * - the GLSL gl_FrontFacing fragment attribute (bool) + * - two-sided stencil test + */ + setup->quad.facing = 0; _mesa_printf("%s one-over-area %f\n", __FUNCTION__, setup->oneoverarea ); @@ -296,6 +304,9 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, } +/** + * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + */ static void const_coeff( struct setup_stage *setup, GLuint slot, GLuint i ) @@ -309,6 +320,9 @@ static void const_coeff( struct setup_stage *setup, } +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient. + */ static void linear_coeff( struct setup_stage *setup, GLuint slot, GLuint i) @@ -345,6 +359,9 @@ static void linear_coeff( struct setup_stage *setup, } +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant. + */ static void persp_coeff( struct setup_stage *setup, GLuint slot, GLuint i ) @@ -430,10 +447,10 @@ static void setup_edges( struct setup_stage *setup ) } - - - - +/** + * Render the upper or lower half of a triangle. + * Scissoring is applied here too. + */ static void subtriangle( struct setup_stage *setup, struct edge *eleft, struct edge *eright, @@ -511,6 +528,9 @@ static void subtriangle( struct setup_stage *setup, } +/** + * Do setup for triangle rasterization, then render the triangle. + */ static void setup_tri( struct prim_stage *stage, struct prim_header *prim ) { diff --git a/src/mesa/softpipe/generic/g_tile_fs.c b/src/mesa/softpipe/generic/g_tile_fs.c index 35e1ab63de..8473a9b5f1 100644 --- a/src/mesa/softpipe/generic/g_tile_fs.c +++ b/src/mesa/softpipe/generic/g_tile_fs.c @@ -44,6 +44,9 @@ struct exec_machine { }; +/** + * Compute quad's attributes values, as constants (GL_FLAT shading). + */ static void INLINE cinterp( struct exec_machine *exec, GLuint attrib, GLuint i ) @@ -56,7 +59,10 @@ static void INLINE cinterp( struct exec_machine *exec, } -/* Push into the fp: +/** + * Compute quad's attribute values by linear interpolation. + * + * Push into the fp: * * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy @@ -68,14 +74,20 @@ static INLINE void linterp( struct exec_machine *exec, GLuint j; for (j = 0; j < QUAD_SIZE; j++) { + const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * exec->attr[0][0][j] + - exec->coef[attrib].dady[i] * exec->attr[0][1][j]); + exec->coef[attrib].dadx[i] * x + + exec->coef[attrib].dady[i] * y); } } -/* Push into the fp: +/** + * Compute quad's attribute values by linear interpolation with + * perspective correction. + * + * Push into the fp: * * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy @@ -90,10 +102,12 @@ static INLINE void pinterp( struct exec_machine *exec, GLuint j; for (j = 0; j < QUAD_SIZE; j++) { + const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; + const GLfloat invW = exec->attr[FRAG_ATTRIB_WPOS][3][j]; exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * exec->attr[0][0][j] + - exec->coef[attrib].dady[i] * exec->attr[0][1][j]) * - exec->attr[0][3][j]); + exec->coef[attrib].dadx[i] * x + + exec->coef[attrib].dady[i] * y) * invW); } } @@ -125,6 +139,7 @@ void quad_shade( struct generic_context *generic, exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; /* Z and W are done by linear interpolation: + * XXX we'll probably have to use integers for Z */ if (generic->need_z) { linterp(&exec, 0, 2); -- cgit v1.2.3 From d46fa8bbefdd46953e86514e43ef51cd234ecc74 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 May 2007 15:53:42 -0600 Subject: INLINE goes before the function's return type --- src/mesa/softpipe/generic/g_tile_fs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/softpipe/generic/g_tile_fs.c b/src/mesa/softpipe/generic/g_tile_fs.c index 8473a9b5f1..4c5e6efaed 100644 --- a/src/mesa/softpipe/generic/g_tile_fs.c +++ b/src/mesa/softpipe/generic/g_tile_fs.c @@ -47,7 +47,7 @@ struct exec_machine { /** * Compute quad's attributes values, as constants (GL_FLAT shading). */ -static void INLINE cinterp( struct exec_machine *exec, +static INLINE void cinterp( struct exec_machine *exec, GLuint attrib, GLuint i ) { -- cgit v1.2.3 From 6fc596193370aba52ddb742c61a4687c7831897c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 May 2007 16:13:17 -0600 Subject: implement point rendering --- src/mesa/softpipe/generic/g_prim_setup.c | 113 +++++++++++++++++++++++++++---- 1 file changed, 101 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/generic/g_prim_setup.c b/src/mesa/softpipe/generic/g_prim_setup.c index ce93c0a6f9..614fc16c7c 100644 --- a/src/mesa/softpipe/generic/g_prim_setup.c +++ b/src/mesa/softpipe/generic/g_prim_setup.c @@ -74,7 +74,6 @@ struct setup_stage { GLuint y_flags; GLuint mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ } span; - }; @@ -190,16 +189,6 @@ static void flush_spans( struct setup_stage *setup ) setup->span.right[1] = 0; } - -/** - * Do setup for point rasterization, then render the point. - */ -static void -setup_point( struct prim_stage *stage, - struct prim_header *header ) -{ -} - /** * Do setup for line rasterization, then render the line. @@ -386,7 +375,10 @@ static void persp_coeff( struct setup_stage *setup, - +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. + */ static void setup_coefficients( struct setup_stage *setup ) { const enum interp_mode *interp = setup->stage.generic->interp; @@ -566,6 +558,103 @@ static void setup_tri( struct prim_stage *stage, flush_spans( setup ); } + +/** + * Do setup for point rasterization, then render the point. + * Round or square points... + * XXX could optimize a lot for 1-pixel points. + */ +static void +setup_point( struct prim_stage *stage, + struct prim_header *prim ) +{ + struct setup_stage *setup = setup_stage( stage ); + GLfloat halfSize = 7.3; /*XXX this is a vertex attrib */ + GLfloat halfSizeSquared = halfSize * halfSize; + const struct vertex_header *v0 = prim->v[0]; + const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; + const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; + const GLint ixmin = block((GLint) (x - halfSize)); + const GLint ixmax = block((GLint) (x + halfSize)); + const GLint iymin = block((GLint) (y - halfSize)); + const GLint iymax = block((GLint) (y + halfSize)); + GLboolean round = GL_TRUE; + GLint ix, iy; + GLuint slot, j; + + /* For points, all interpolants are constant-valued. + * However, for point sprites, we'll need to setup texcoords appropriately. + * XXX: which coefficients are the texcoords??? + * We may do point sprites as textured quads... + */ + setup->vprovoke = prim->v[0]; + const_coeff(setup, 0, 2); + const_coeff(setup, 0, 3); + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + } + + /* XXX need to clip against scissor bounds too */ + + for (iy = iymin; iy <= iymax; iy += 2) { + for (ix = ixmin; ix <= ixmax; ix += 2) { + + if (round) { + /* rounded points */ + /* XXX for GL_SMOOTH, need to compute per-fragment coverage too */ + GLfloat dx, dy; + + setup->quad.mask = 0x0; + + dx = (ix + 0.5) - x; + dy = (iy + 0.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_BOTTOM_LEFT; + + dx = (ix + 1.5) - x; + dy = (iy + 0.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_BOTTOM_RIGHT; + + dx = (ix + 0.5) - x; + dy = (iy + 1.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_TOP_LEFT; + + dx = (ix + 1.5) - x; + dy = (iy + 1.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_TOP_RIGHT; + } + else { + /* square points */ + setup->quad.mask = 0xf; + + if (ix + 0.5 < x - halfSize) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + + if (ix + 1.5 > x + halfSize) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + + if (iy + 0.5 < y - halfSize) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + + if (iy + 1.5 > y + halfSize) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + } + + if (setup->quad.mask) { + setup->quad.x0 = ix; + setup->quad.y0 = iy; + quad_shade( setup->stage.generic, &setup->quad ); + } + } + } +} + + + static void setup_end( struct prim_stage *stage ) { } -- cgit v1.2.3 From 4c9b48dcc510929951516f3cd17b98654ac93a9a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 May 2007 16:07:27 -0600 Subject: depth(Z) state --- src/mesa/softpipe/state_tracker/st_atom_depth.c | 88 +++++++++++++++++++++++++ 1 file changed, 88 insertions(+) create mode 100644 src/mesa/softpipe/state_tracker/st_atom_depth.c (limited to 'src') diff --git a/src/mesa/softpipe/state_tracker/st_atom_depth.c b/src/mesa/softpipe/state_tracker/st_atom_depth.c new file mode 100644 index 0000000000..5532abc8fd --- /dev/null +++ b/src/mesa/softpipe/state_tracker/st_atom_depth.c @@ -0,0 +1,88 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "softpipe/sp_context.h" +#include "softpipe/sp_defines.h" + + +/** + * Convert GLenum depth func tokens to softpipe tokens. + */ +static GLuint +gl_depth_func_to_sp(GLenum func) +{ + /* Same values, just biased */ + assert(SP_DEPTH_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(SP_DEPTH_FUNC_LESS == GL_LESS - GL_NEVER); + assert(SP_DEPTH_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(SP_DEPTH_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(SP_DEPTH_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(SP_DEPTH_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(SP_DEPTH_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(SP_DEPTH_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(func >= GL_NEVER); + assert(func <= GL_ALWAYS); + return func - GL_NEVER; +} + + +static void +update_depth( struct st_context *st ) +{ + struct softpipe_depth_state depth; + + memset(&depth, 0, sizeof(depth)); + + depth.enabled = st->ctx->Depth.Test; + depth.writemask = st->ctx->Depth.Mask; + depth.func = gl_depth_func_to_sp(st->ctx->Depth.Func); + depth.clear = st->ctx->Depth.Clear; + + if (memcmp(&depth, &st->state.depth, sizeof(depth)) != 0) { + /* state has changed */ + st->state.depth = depth; /* struct copy */ + st->softpipe->set_depth_state(st->softpipe, &depth); /* set new state */ + } +} + + +const struct st_tracked_state st_update_depth = { + .dirty = { + .mesa = (_NEW_DEPTH), + .st = 0, + }, + .update = update_depth +}; -- cgit v1.2.3 From 8f2888758129b0a40d71fa4ce10d606b4bcf7631 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 May 2007 16:07:39 -0600 Subject: added depth(Z) state support --- src/mesa/softpipe/sp_context.h | 3 +++ src/mesa/softpipe/sp_defines.h | 10 ++++++++++ src/mesa/softpipe/sp_state.h | 9 +++++++++ src/mesa/softpipe/state_tracker/st_atom.h | 1 + src/mesa/softpipe/state_tracker/st_context.h | 1 + 5 files changed, 24 insertions(+) (limited to 'src') diff --git a/src/mesa/softpipe/sp_context.h b/src/mesa/softpipe/sp_context.h index e047a9330b..e2b678f80f 100644 --- a/src/mesa/softpipe/sp_context.h +++ b/src/mesa/softpipe/sp_context.h @@ -43,6 +43,9 @@ struct softpipe_context { void (*set_clip_state)( struct softpipe_context *, const struct softpipe_clip_state * ); + void (*set_depth_state)( struct softpipe_context *, + const struct softpipe_depth_state * ); + void (*set_viewport)( struct softpipe_context *, const struct softpipe_viewport * ); diff --git a/src/mesa/softpipe/sp_defines.h b/src/mesa/softpipe/sp_defines.h index bc88e3d71e..e1c773a407 100644 --- a/src/mesa/softpipe/sp_defines.h +++ b/src/mesa/softpipe/sp_defines.h @@ -89,4 +89,14 @@ #define SP_STENCIL_OP_DECR_WRAP 6 #define SP_STENCIL_OP_INVERT 7 +#define SP_DEPTH_FUNC_NEVER 0 +#define SP_DEPTH_FUNC_LESS 1 +#define SP_DEPTH_FUNC_EQUAL 2 +#define SP_DEPTH_FUNC_LEQUAL 3 +#define SP_DEPTH_FUNC_GREATER 4 +#define SP_DEPTH_FUNC_NOTEQUAL 5 +#define SP_DEPTH_FUNC_GEQUAL 6 +#define SP_DEPTH_FUNC_ALWAYS 7 + + #endif diff --git a/src/mesa/softpipe/sp_state.h b/src/mesa/softpipe/sp_state.h index f7d904bffe..634b9e9868 100644 --- a/src/mesa/softpipe/sp_state.h +++ b/src/mesa/softpipe/sp_state.h @@ -100,6 +100,15 @@ struct softpipe_constant_buffer { }; +struct softpipe_depth_state +{ + GLuint enabled:1; /**< depth test enabled? */ + GLuint writemask:1; /**< allow depth buffer writes? */ + GLuint func:3; /**< depth test func */ + GLfloat clear; /**< Clear value in [0,1] (XXX correct place?) */ +}; + + struct softpipe_blend_state { GLuint blend_enable:1; diff --git a/src/mesa/softpipe/state_tracker/st_atom.h b/src/mesa/softpipe/state_tracker/st_atom.h index 8dc7460d11..1d8da46336 100644 --- a/src/mesa/softpipe/state_tracker/st_atom.h +++ b/src/mesa/softpipe/state_tracker/st_atom.h @@ -46,6 +46,7 @@ void st_validate_state( struct st_context *st ); const struct st_tracked_state st_update_cbuf; const struct st_tracked_state st_update_clip; +const struct st_tracked_state st_update_depth; const struct st_tracked_state st_update_fs; const struct st_tracked_state st_update_setup; const struct st_tracked_state st_update_viewport; diff --git a/src/mesa/softpipe/state_tracker/st_context.h b/src/mesa/softpipe/state_tracker/st_context.h index 5ff10c1b1f..317d377371 100644 --- a/src/mesa/softpipe/state_tracker/st_context.h +++ b/src/mesa/softpipe/state_tracker/st_context.h @@ -71,6 +71,7 @@ struct st_context struct softpipe_blend_state blend; struct softpipe_surface cbuf; struct softpipe_clip_state clip; + struct softpipe_depth_state depth; struct softpipe_scissor_rect scissor; struct softpipe_poly_stipple poly_stipple; struct softpipe_stencil_state stencil; -- cgit v1.2.3 From 092a9ecbba4d766da9c0becf888818c020ac226d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 May 2007 16:10:31 -0600 Subject: comments, clean-up, alphabetize members --- src/mesa/softpipe/sp_context.h | 48 ++++++++++++++++++++++++------------------ 1 file changed, 28 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/sp_context.h b/src/mesa/softpipe/sp_context.h index e2b678f80f..716ddd27a9 100644 --- a/src/mesa/softpipe/sp_context.h +++ b/src/mesa/softpipe/sp_context.h @@ -30,24 +30,47 @@ #include "mtypes.h" -struct softpipe_context *generic_create( void ); +extern struct softpipe_context *generic_create( void ); /* Drawing currently kludged up via the existing tnl/ module. */ struct vertex_buffer; + +/** + * Software pipeline rendering context. Basically a collection of + * state setting functions, plus VBO drawing entrypoint. + */ struct softpipe_context { void (*destroy)( struct softpipe_context * ); + /* + * Drawing + */ + void (*draw_vb)( struct softpipe_context *softpipe, + struct vertex_buffer *VB ); + + /* + * State functions + */ + void (*set_blend_state)( struct softpipe_context *, + const struct softpipe_blend_state * ); + + void (*set_cbuf_state)( struct softpipe_context *, + const struct softpipe_surface * ); + void (*set_clip_state)( struct softpipe_context *, const struct softpipe_clip_state * ); void (*set_depth_state)( struct softpipe_context *, const struct softpipe_depth_state * ); - void (*set_viewport)( struct softpipe_context *, - const struct softpipe_viewport * ); + void (*set_fs_state)( struct softpipe_context *, + const struct softpipe_fs_state * ); + + void (*set_polygon_stipple)( struct softpipe_context *, + const struct softpipe_poly_stipple * ); void (*set_setup_state)( struct softpipe_context *, const struct softpipe_setup_state * ); @@ -58,24 +81,9 @@ struct softpipe_context { void (*set_stencil_state)( struct softpipe_context *, const struct softpipe_stencil_state * ); - void (*set_blend_state)( struct softpipe_context *, - const struct softpipe_blend_state * ); - - void (*set_fs_state)( struct softpipe_context *, - const struct softpipe_fs_state * ); - - void (*set_polygon_stipple)( struct softpipe_context *, - const struct softpipe_poly_stipple * ); - - void (*set_cbuf_state)( struct softpipe_context *, - const struct softpipe_surface * ); - - - void (*draw_vb)( struct softpipe_context *softpipe, - struct vertex_buffer *VB ); + void (*set_viewport)( struct softpipe_context *, + const struct softpipe_viewport * ); }; - - #endif -- cgit v1.2.3 From 02a47549ad182d28d330063d4f452108e3a4b452 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 May 2007 16:26:55 -0600 Subject: sketch out texture sampler state --- src/mesa/softpipe/sp_defines.h | 19 +++++++++++++++++++ src/mesa/softpipe/sp_state.h | 24 ++++++++++++++++++++++++ 2 files changed, 43 insertions(+) (limited to 'src') diff --git a/src/mesa/softpipe/sp_defines.h b/src/mesa/softpipe/sp_defines.h index e1c773a407..c381865dab 100644 --- a/src/mesa/softpipe/sp_defines.h +++ b/src/mesa/softpipe/sp_defines.h @@ -89,6 +89,7 @@ #define SP_STENCIL_OP_DECR_WRAP 6 #define SP_STENCIL_OP_INVERT 7 +/* Note: same as stencil funcs. Also used for shadow/depth compare */ #define SP_DEPTH_FUNC_NEVER 0 #define SP_DEPTH_FUNC_LESS 1 #define SP_DEPTH_FUNC_EQUAL 2 @@ -98,5 +99,23 @@ #define SP_DEPTH_FUNC_GEQUAL 6 #define SP_DEPTH_FUNC_ALWAYS 7 +#define SP_TEX_WRAP_REPEAT 0 +#define SP_TEX_WRAP_CLAMP 1 +#define SP_TEX_WRAP_CLAMP_TO_EDGE 2 +#define SP_TEX_WRAP_CLAMP_TO_BORDER 3 +#define SP_TEX_WRAP_MIRROR_REPEAT 4 +#define SP_TEX_WRAP_MIRROR_CLAMP 5 +#define SP_TEX_WRAP_MIRROR_CLAMP_TO_EDGE 6 +#define SP_TEX_WRAP_MIRROR_CLAMP_TO_BORDER 7 + +#define SP_TEX_FILTER_NEAREST 0 +#define SP_TEX_FILTER_LINEAR 1 +#define SP_TEX_FILTER_NEAREST_MIPMAP_NEAREST 2 +#define SP_TEX_FILTER_NEAREST_MIPMAP_LINEAR 3 +#define SP_TEX_FILTER_LINEAR_MIPMAP_NEAREST 4 +#define SP_TEX_FILTER_LINEAR_MIPMAP_LINEAR 5 + +#define SP_TEX_COMPARE_NONE 0 +#define SP_TEX_COMPARE_R_TO_TEXTURE 1 #endif diff --git a/src/mesa/softpipe/sp_state.h b/src/mesa/softpipe/sp_state.h index 634b9e9868..150fa9fcad 100644 --- a/src/mesa/softpipe/sp_state.h +++ b/src/mesa/softpipe/sp_state.h @@ -157,4 +157,28 @@ struct softpipe_surface { }; +/** + * Texture sampler state. + */ +struct softpipe_sampler_state +{ + GLuint wrap_s:3; /**< SP_TEX_WRAP_x */ + GLuint wrap_t:3; /**< SP_TEX_WRAP_x */ + GLuint wrap_r:3; /**< SP_TEX_WRAP_x */ + GLuint min_filter:3; /**< SP_TEX_FILTER_x */ + GLuint mag_filter:1; /**< SP_TEX_FILTER_LINEAR or _NEAREST */ + GLfloat min_lod; + GLfloat max_lod; + GLfloat lod_bias; +#if 0 /* need these? */ + GLint BaseLevel; /**< min mipmap level, OpenGL 1.2 */ + GLint MaxLevel; /**< max mipmap level, OpenGL 1.2 */ +#endif + GLfloat max_anisotropy; + GLuint compare:1; /**< shadow/depth compare enabled? */ + GLenum compare_mode:1; /**< SP_TEX_COMPARE_x */ + GLenum compare_func:3; /**< SP_DEPTH_FUNC_x */ + GLfloat shadow_ambient; /**< shadow test fail color/intensity */ +}; + #endif -- cgit v1.2.3 From c9527eae8eb461c82632f89eb7c5244a92e885bc Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 May 2007 16:27:17 -0600 Subject: added st_atom_depth.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index c5ee7d9611..eda3ef7beb 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -176,6 +176,7 @@ SOFTPIPE_SOURCES = \ softpipe/state_tracker/st_atom_blend.c \ softpipe/state_tracker/st_atom_cbuf.c \ softpipe/state_tracker/st_atom_clip.c \ + softpipe/state_tracker/st_atom_depth.c \ softpipe/state_tracker/st_atom_fs.c \ softpipe/state_tracker/st_atom_scissor.c \ softpipe/state_tracker/st_atom_stencil.c \ -- cgit v1.2.3 From 6378dc3ae302e8a69f482fd99a63a23d1a1730a1 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Jun 2007 08:47:49 -0600 Subject: simple line drawing code --- src/mesa/softpipe/generic/g_prim_setup.c | 272 +++++++++++++++++++++++++++---- 1 file changed, 240 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/generic/g_prim_setup.c b/src/mesa/softpipe/generic/g_prim_setup.c index 614fc16c7c..125d361b92 100644 --- a/src/mesa/softpipe/generic/g_prim_setup.c +++ b/src/mesa/softpipe/generic/g_prim_setup.c @@ -36,6 +36,9 @@ #include "g_tile.h" +/** + * Triangle edge info + */ struct edge { GLfloat dx; /* X(v1) - X(v0), used only during setup */ GLfloat dy; /* Y(v1) - Y(v0), used only during setup */ @@ -46,9 +49,13 @@ struct edge { }; +/** + * Triangle setup info (derived from prim_stage). + * Also used for line drawing (taking some liberties). + */ struct setup_stage { - struct prim_stage stage; - + struct prim_stage stage; /**< This must be first */ + /* Vertices are just an array of floats making up each attribute in * turn. Currently fixed at 4 floats, but should change in time. * Codegen will help cope with this. @@ -57,7 +64,7 @@ struct setup_stage { const struct vertex_header *vmid; const struct vertex_header *vmin; const struct vertex_header *vprovoke; - + struct edge ebot; struct edge etop; struct edge emaj; @@ -190,20 +197,9 @@ static void flush_spans( struct setup_stage *setup ) } -/** - * Do setup for line rasterization, then render the line. - */ -static void -setup_line( struct prim_stage *stage, - struct prim_header *header ) -{ -} - - static GLboolean setup_sort_vertices( struct setup_stage *setup, const struct prim_header *prim ) { - const struct vertex_header *v0 = prim->v[0]; const struct vertex_header *v1 = prim->v[1]; const struct vertex_header *v2 = prim->v[2]; @@ -310,11 +306,12 @@ static void const_coeff( struct setup_stage *setup, /** - * Compute a0, dadx and dady for a linearly interpolated coefficient. + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a triangle. */ -static void linear_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i) +static void tri_linear_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i) { GLfloat botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; GLfloat majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; @@ -349,11 +346,12 @@ static void linear_coeff( struct setup_stage *setup, /** - * Compute a0, dadx and dady for a perspective-corrected interpolant. + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a triangle. */ -static void persp_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i ) +static void tri_persp_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i ) { /* premultiply by 1/w: */ @@ -379,15 +377,15 @@ static void persp_coeff( struct setup_stage *setup, * Compute the setup->coef[] array dadx, dady, a0 values. * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. */ -static void setup_coefficients( struct setup_stage *setup ) +static void setup_tri_coefficients( struct setup_stage *setup ) { const enum interp_mode *interp = setup->stage.generic->interp; GLuint slot, j; /* z and w are done by linear interpolation: */ - linear_coeff(setup, 0, 2); - linear_coeff(setup, 0, 3); + tri_linear_coeff(setup, 0, 2); + tri_linear_coeff(setup, 0, 3); /* setup interpolation for all the remaining attributes: */ @@ -400,12 +398,12 @@ static void setup_coefficients( struct setup_stage *setup ) case INTERP_LINEAR: for (j = 0; j < NUM_CHANNELS; j++) - linear_coeff(setup, slot, j); + tri_linear_coeff(setup, slot, j); break; case INTERP_PERSPECTIVE: for (j = 0; j < NUM_CHANNELS; j++) - persp_coeff(setup, slot, j); + tri_persp_coeff(setup, slot, j); break; } } @@ -413,7 +411,7 @@ static void setup_coefficients( struct setup_stage *setup ) -static void setup_edges( struct setup_stage *setup ) +static void setup_tri_edges( struct setup_stage *setup ) { GLfloat vmin_x = setup->vmin->data[0][0] + 0.5; GLfloat vmid_x = setup->vmid->data[0][0] + 0.5; @@ -531,8 +529,8 @@ static void setup_tri( struct prim_stage *stage, _mesa_printf("%s\n", __FUNCTION__ ); setup_sort_vertices( setup, prim ); - setup_coefficients( setup ); - setup_edges( setup ); + setup_tri_coefficients( setup ); + setup_tri_edges( setup ); setup->span.y = 0; setup->span.y_flags = 0; @@ -559,14 +557,224 @@ static void setup_tri( struct prim_stage *stage, } + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a line. + */ +static void +line_linear_coeff(struct setup_stage *setup, GLuint slot, GLuint i) +{ + const GLfloat dz = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + const GLfloat dadx = dz * setup->emaj.dx * setup->oneoverarea; + const GLfloat dady = dz * setup->emaj.dy * setup->oneoverarea; + setup->coef[slot].dadx[i] = dadx; + setup->coef[slot].dady[i] = dady; + setup->coef[slot].a0[i] + = (setup->vmin->data[slot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5) + + dady * (setup->vmin->data[0][1] - 0.5))); +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a line. + */ +static void +line_persp_coeff(struct setup_stage *setup, GLuint slot, GLuint i) +{ + /* XXX to do */ + line_linear_coeff(setup, slot, i); /* XXX temporary */ +} + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmax are initialized. + */ +static INLINE void +setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) +{ + const enum interp_mode *interp = setup->stage.generic->interp; + GLuint slot, j; + + /* use setup->vmin, vmax to point to vertices */ + setup->vprovoke = prim->v[1]; + setup->vmin = prim->v[0]; + setup->vmax = prim->v[1]; + + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + /* NOTE: this is not really 1/area */ + setup->oneoverarea = 1.0 / (setup->emaj.dx * setup->emaj.dx + + setup->emaj.dy * setup->emaj.dy); + + /* z and w are done by linear interpolation: + */ + line_linear_coeff(setup, 0, 2); + line_linear_coeff(setup, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + switch (interp[slot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + line_linear_coeff(setup, slot, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + line_persp_coeff(setup, slot, j); + break; + } + } +} + + +/** + * Plot a pixel in a line segment. + */ +static INLINE void +plot(struct setup_stage *setup, GLint x, GLint y) +{ + const GLint quadX = block(x); + const GLint quadY = block(y); + + if ((quadX != setup->quad.x0 || quadY != setup->quad.y0) + && setup->quad.x0 != -1) { + /* flush prev quad, start new quad */ + quad_shade(setup->stage.generic, &setup->quad); + setup->quad.mask = 0x0; + } + setup->quad.x0 = quadX; + setup->quad.y0 = quadY; + + if (x & 1) { + if (y & 1) + setup->quad.mask |= MASK_TOP_RIGHT; + else + setup->quad.mask |= MASK_BOTTOM_RIGHT; + } + else { + if (y & 1) + setup->quad.mask |= MASK_TOP_LEFT; + else + setup->quad.mask |= MASK_BOTTOM_LEFT; + } +} + + + +/** + * Do setup for line rasterization, then render the line. + * XXX single-pixel width, no stipple, etc + * XXX no scissoring yet. + */ +static void +setup_line(struct prim_stage *stage, struct prim_header *prim) +{ + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + struct setup_stage *setup = setup_stage( stage ); + + GLint x0 = (GLint) v0->data[0][0]; + GLint x1 = (GLint) v1->data[0][0]; + GLint y0 = (GLint) v0->data[0][1]; + GLint y1 = (GLint) v1->data[0][1]; + GLint dx = x1 - x0; + GLint dy = y1 - y0; + GLint xstep, ystep; + + if (dx == 0 && dy == 0) + return; + + setup_line_coefficients(setup, prim); + + if (dx < 0) { + dx = -dx; /* make positive */ + xstep = -1; + } + else { + xstep = 1; + } + + if (dy < 0) { + dy = -dy; /* make positive */ + ystep = -1; + } + else { + ystep = 1; + } + + assert(dx >= 0); + assert(dy >= 0); + + setup->quad.x0 = setup->quad.y0 = -1; + setup->quad.mask = 0x0; + + if (dx > dy) { + /*** X-major line ***/ + GLint i; + const GLint errorInc = dy + dy; + GLint error = errorInc - dx; + const GLint errorDec = error - dx; + + for (i = 0; i < dx; i++) { + plot(setup, x0, y0); + + x0 += xstep; + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + y0 += ystep; + } + } + } + else { + /*** Y-major line ***/ + GLint i; + const GLint errorInc = dx + dx; + GLint error = errorInc - dy; + const GLint errorDec = error - dy; + + for (i = 0; i < dy; i++) { + plot(setup, x0, y0); + + y0 += ystep; + + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + x0 += xstep; + } + } + } + + /* draw final quad */ + if (setup->quad.mask) { + quad_shade(setup->stage.generic, &setup->quad); + } +} + + /** * Do setup for point rasterization, then render the point. * Round or square points... * XXX could optimize a lot for 1-pixel points. */ static void -setup_point( struct prim_stage *stage, - struct prim_header *prim ) +setup_point(struct prim_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); GLfloat halfSize = 7.3; /*XXX this is a vertex attrib */ -- cgit v1.2.3 From 790ec6ad3face4fb2e1c4b69d9d42bdf332841d6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 13 Jun 2007 15:37:52 +0100 Subject: Remove some dead code. --- src/mesa/softpipe/generic/g_prim_setup.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/generic/g_prim_setup.c b/src/mesa/softpipe/generic/g_prim_setup.c index 125d361b92..0e1bc03b4f 100644 --- a/src/mesa/softpipe/generic/g_prim_setup.c +++ b/src/mesa/softpipe/generic/g_prim_setup.c @@ -205,7 +205,6 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, const struct vertex_header *v2 = prim->v[2]; setup->vprovoke = v2; -// setup->oneoverarea = -1.0 / prim->det; /* determine bottom to top order of vertices */ { @@ -230,7 +229,6 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, setup->vmin = v0; setup->vmid = v2; setup->vmax = v1; -// setup->oneoverarea = -setup->oneoverarea; } } else { @@ -239,14 +237,12 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, setup->vmin = v1; setup->vmid = v0; setup->vmax = v2; -// setup->oneoverarea = -setup->oneoverarea; } else if (y2 <= y1) { /* y2<=y1<=y0 */ setup->vmin = v2; setup->vmid = v1; setup->vmax = v0; -// setup->oneoverarea = -setup->oneoverarea; } else { /* y1<=y2<=y0 */ -- cgit v1.2.3 From 7050a4a7bc779b2c30fbd8244d4e77e6bd3dc30f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 14 Jun 2007 15:18:27 +0100 Subject: Small optimization for line drawing. And some comments on point sprites. --- src/mesa/softpipe/generic/g_prim_setup.c | 49 ++++++++++++++++++-------------- 1 file changed, 28 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/softpipe/generic/g_prim_setup.c b/src/mesa/softpipe/generic/g_prim_setup.c index 0e1bc03b4f..41e1381a60 100644 --- a/src/mesa/softpipe/generic/g_prim_setup.c +++ b/src/mesa/softpipe/generic/g_prim_setup.c @@ -640,30 +640,26 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) static INLINE void plot(struct setup_stage *setup, GLint x, GLint y) { - const GLint quadX = block(x); - const GLint quadY = block(y); - - if ((quadX != setup->quad.x0 || quadY != setup->quad.y0) - && setup->quad.x0 != -1) { + const GLint iy = y & 1; + const GLint ix = x & 1; + const GLint quadX = x - ix; + const GLint quadY = y - iy; + const GLint mask = (1 << ix) << (2 * iy); + + if (quadX != setup->quad.x0 || + quadY != setup->quad.y0) + { /* flush prev quad, start new quad */ - quad_shade(setup->stage.generic, &setup->quad); + + if (setup->quad.x0 != -1) + quad_shade(setup->stage.generic, &setup->quad); + + setup->quad.x0 = quadX; + setup->quad.y0 = quadY; setup->quad.mask = 0x0; } - setup->quad.x0 = quadX; - setup->quad.y0 = quadY; - - if (x & 1) { - if (y & 1) - setup->quad.mask |= MASK_TOP_RIGHT; - else - setup->quad.mask |= MASK_BOTTOM_RIGHT; - } - else { - if (y & 1) - setup->quad.mask |= MASK_TOP_LEFT; - else - setup->quad.mask |= MASK_BOTTOM_LEFT; - } + + setup->quad.mask |= mask; } @@ -790,6 +786,17 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) * However, for point sprites, we'll need to setup texcoords appropriately. * XXX: which coefficients are the texcoords??? * We may do point sprites as textured quads... + * + * KW: We don't know which coefficients are texcoords - ultimately + * the choice of what interpolation mode to use for each attribute + * should be determined by the fragment program, using + * per-attribute declaration statements that include interpolation + * mode as a parameter. So either the fragment program will have + * to be adjusted for pointsprite vs normal point behaviour, or + * otherwise a special interpolation mode will have to be defined + * which matches the required behaviour for point sprites. But - + * the latter is not a feature of normal hardware, and as such + * probably should be ruled out on that basis. */ setup->vprovoke = prim->v[0]; const_coeff(setup, 0, 2); -- cgit v1.2.3 From 6393cda6766b707ef01e925d378239a66d143ae0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 14 Jun 2007 18:11:48 +0100 Subject: Renamed softpipe directories and files to something less confusing. softpipe/state_tracker --> state_tracker/ softpipe/ --> pipe/ softpipe/generic --> pipe/softpipe/ I don't think pipe is a great name, but I disliked all the others too. Luckily it's fairly easy to rename with git, so this can be revisited later. --- src/mesa/pipe/Makefile | 2 + src/mesa/pipe/p_context.h | 89 +++ src/mesa/pipe/p_defines.h | 121 +++ src/mesa/pipe/p_state.h | 184 +++++ src/mesa/pipe/softpipe/Makefile | 3 + src/mesa/pipe/softpipe/g_context.c | 135 ++++ src/mesa/pipe/softpipe/g_context.h | 149 ++++ src/mesa/pipe/softpipe/g_draw.c | 696 ++++++++++++++++ src/mesa/pipe/softpipe/g_draw.h | 56 ++ src/mesa/pipe/softpipe/g_headers.h | 116 +++ src/mesa/pipe/softpipe/g_prim.h | 90 +++ src/mesa/pipe/softpipe/g_prim_clip.c | 429 ++++++++++ src/mesa/pipe/softpipe/g_prim_cull.c | 119 +++ src/mesa/pipe/softpipe/g_prim_flatshade.c | 149 ++++ src/mesa/pipe/softpipe/g_prim_offset.c | 157 ++++ src/mesa/pipe/softpipe/g_prim_setup.c | 888 +++++++++++++++++++++ src/mesa/pipe/softpipe/g_prim_setup.h | 121 +++ src/mesa/pipe/softpipe/g_prim_twoside.c | 149 ++++ src/mesa/pipe/softpipe/g_prim_unfilled.c | 165 ++++ src/mesa/pipe/softpipe/g_state.h | 61 ++ src/mesa/pipe/softpipe/g_state_clip.c | 69 ++ src/mesa/pipe/softpipe/g_state_derived.c | 136 ++++ src/mesa/pipe/softpipe/g_state_fs.c | 50 ++ src/mesa/pipe/softpipe/g_state_setup.c | 122 +++ src/mesa/pipe/softpipe/g_state_surface.c | 53 ++ src/mesa/pipe/softpipe/g_surface.c | 153 ++++ src/mesa/pipe/softpipe/g_surface.h | 115 +++ src/mesa/pipe/softpipe/g_tile.h | 42 + src/mesa/pipe/softpipe/g_tile_fs.c | 203 +++++ src/mesa/pipe/softpipe/g_tile_output.c | 92 +++ src/mesa/pipe/softpipe/sp_context.c | 135 ++++ src/mesa/pipe/softpipe/sp_context.h | 149 ++++ src/mesa/pipe/softpipe/sp_draw.c | 696 ++++++++++++++++ src/mesa/pipe/softpipe/sp_draw.h | 56 ++ src/mesa/pipe/softpipe/sp_headers.h | 116 +++ src/mesa/pipe/softpipe/sp_prim.h | 90 +++ src/mesa/pipe/softpipe/sp_prim_clip.c | 429 ++++++++++ src/mesa/pipe/softpipe/sp_prim_cull.c | 119 +++ src/mesa/pipe/softpipe/sp_prim_flatshade.c | 149 ++++ src/mesa/pipe/softpipe/sp_prim_offset.c | 157 ++++ src/mesa/pipe/softpipe/sp_prim_setup.c | 888 +++++++++++++++++++++ src/mesa/pipe/softpipe/sp_prim_setup.h | 121 +++ src/mesa/pipe/softpipe/sp_prim_twoside.c | 149 ++++ src/mesa/pipe/softpipe/sp_prim_unfilled.c | 165 ++++ src/mesa/pipe/softpipe/sp_state.h | 61 ++ src/mesa/pipe/softpipe/sp_state_clip.c | 69 ++ src/mesa/pipe/softpipe/sp_state_derived.c | 136 ++++ src/mesa/pipe/softpipe/sp_state_fs.c | 50 ++ src/mesa/pipe/softpipe/sp_state_setup.c | 122 +++ src/mesa/pipe/softpipe/sp_state_surface.c | 53 ++ src/mesa/pipe/softpipe/sp_surface.c | 153 ++++ src/mesa/pipe/softpipe/sp_surface.h | 115 +++ src/mesa/pipe/softpipe/sp_tile.h | 42 + src/mesa/pipe/softpipe/sp_tile_fs.c | 203 +++++ src/mesa/pipe/softpipe/sp_tile_output.c | 92 +++ src/mesa/softpipe/Makefile | 2 - src/mesa/softpipe/generic/Makefile | 3 - src/mesa/softpipe/generic/g_context.c | 135 ---- src/mesa/softpipe/generic/g_context.h | 149 ---- src/mesa/softpipe/generic/g_draw.c | 696 ---------------- src/mesa/softpipe/generic/g_draw.h | 56 -- src/mesa/softpipe/generic/g_headers.h | 116 --- src/mesa/softpipe/generic/g_prim.h | 90 --- src/mesa/softpipe/generic/g_prim_clip.c | 429 ---------- src/mesa/softpipe/generic/g_prim_cull.c | 119 --- src/mesa/softpipe/generic/g_prim_flatshade.c | 149 ---- src/mesa/softpipe/generic/g_prim_offset.c | 157 ---- src/mesa/softpipe/generic/g_prim_setup.c | 888 --------------------- src/mesa/softpipe/generic/g_prim_setup.h | 121 --- src/mesa/softpipe/generic/g_prim_twoside.c | 149 ---- src/mesa/softpipe/generic/g_prim_unfilled.c | 165 ---- src/mesa/softpipe/generic/g_state.h | 61 -- src/mesa/softpipe/generic/g_state_clip.c | 69 -- src/mesa/softpipe/generic/g_state_derived.c | 136 ---- src/mesa/softpipe/generic/g_state_fs.c | 50 -- src/mesa/softpipe/generic/g_state_setup.c | 122 --- src/mesa/softpipe/generic/g_state_surface.c | 53 -- src/mesa/softpipe/generic/g_surface.c | 153 ---- src/mesa/softpipe/generic/g_surface.h | 115 --- src/mesa/softpipe/generic/g_tile.h | 42 - src/mesa/softpipe/generic/g_tile_fs.c | 203 ----- src/mesa/softpipe/generic/g_tile_output.c | 92 --- src/mesa/softpipe/sp_context.h | 89 --- src/mesa/softpipe/sp_defines.h | 121 --- src/mesa/softpipe/sp_state.h | 184 ----- src/mesa/softpipe/state_tracker/Makefile | 2 - src/mesa/softpipe/state_tracker/st_atom.c | 171 ---- src/mesa/softpipe/state_tracker/st_atom.h | 59 -- src/mesa/softpipe/state_tracker/st_atom_blend.c | 202 ----- src/mesa/softpipe/state_tracker/st_atom_cbuf.c | 72 -- src/mesa/softpipe/state_tracker/st_atom_clip.c | 75 -- src/mesa/softpipe/state_tracker/st_atom_depth.c | 88 -- src/mesa/softpipe/state_tracker/st_atom_fs.c | 56 -- src/mesa/softpipe/state_tracker/st_atom_scissor.c | 88 -- src/mesa/softpipe/state_tracker/st_atom_setup.c | 175 ---- src/mesa/softpipe/state_tracker/st_atom_stencil.c | 140 ---- src/mesa/softpipe/state_tracker/st_atom_viewport.c | 117 --- src/mesa/softpipe/state_tracker/st_cb_program.c | 165 ---- src/mesa/softpipe/state_tracker/st_context.c | 76 -- src/mesa/softpipe/state_tracker/st_context.h | 111 --- src/mesa/softpipe/state_tracker/st_draw.c | 107 --- src/mesa/softpipe/state_tracker/st_draw.h | 40 - src/mesa/softpipe/state_tracker/st_program.h | 68 -- src/mesa/softpipe/state_tracker/st_public.h | 43 - src/mesa/state_tracker/Makefile | 2 + src/mesa/state_tracker/st_atom.c | 171 ++++ src/mesa/state_tracker/st_atom.h | 59 ++ src/mesa/state_tracker/st_atom_blend.c | 202 +++++ src/mesa/state_tracker/st_atom_cbuf.c | 72 ++ src/mesa/state_tracker/st_atom_clip.c | 75 ++ src/mesa/state_tracker/st_atom_depth.c | 88 ++ src/mesa/state_tracker/st_atom_fs.c | 56 ++ src/mesa/state_tracker/st_atom_scissor.c | 88 ++ src/mesa/state_tracker/st_atom_setup.c | 175 ++++ src/mesa/state_tracker/st_atom_stencil.c | 140 ++++ src/mesa/state_tracker/st_atom_viewport.c | 117 +++ src/mesa/state_tracker/st_cb_program.c | 165 ++++ src/mesa/state_tracker/st_context.c | 76 ++ src/mesa/state_tracker/st_context.h | 111 +++ src/mesa/state_tracker/st_draw.c | 107 +++ src/mesa/state_tracker/st_draw.h | 40 + src/mesa/state_tracker/st_program.h | 68 ++ src/mesa/state_tracker/st_public.h | 43 + 123 files changed, 11284 insertions(+), 6769 deletions(-) create mode 100644 src/mesa/pipe/Makefile create mode 100644 src/mesa/pipe/p_context.h create mode 100644 src/mesa/pipe/p_defines.h create mode 100644 src/mesa/pipe/p_state.h create mode 100644 src/mesa/pipe/softpipe/Makefile create mode 100644 src/mesa/pipe/softpipe/g_context.c create mode 100644 src/mesa/pipe/softpipe/g_context.h create mode 100644 src/mesa/pipe/softpipe/g_draw.c create mode 100644 src/mesa/pipe/softpipe/g_draw.h create mode 100644 src/mesa/pipe/softpipe/g_headers.h create mode 100644 src/mesa/pipe/softpipe/g_prim.h create mode 100644 src/mesa/pipe/softpipe/g_prim_clip.c create mode 100644 src/mesa/pipe/softpipe/g_prim_cull.c create mode 100644 src/mesa/pipe/softpipe/g_prim_flatshade.c create mode 100644 src/mesa/pipe/softpipe/g_prim_offset.c create mode 100644 src/mesa/pipe/softpipe/g_prim_setup.c create mode 100644 src/mesa/pipe/softpipe/g_prim_setup.h create mode 100644 src/mesa/pipe/softpipe/g_prim_twoside.c create mode 100644 src/mesa/pipe/softpipe/g_prim_unfilled.c create mode 100644 src/mesa/pipe/softpipe/g_state.h create mode 100644 src/mesa/pipe/softpipe/g_state_clip.c create mode 100644 src/mesa/pipe/softpipe/g_state_derived.c create mode 100644 src/mesa/pipe/softpipe/g_state_fs.c create mode 100644 src/mesa/pipe/softpipe/g_state_setup.c create mode 100644 src/mesa/pipe/softpipe/g_state_surface.c create mode 100644 src/mesa/pipe/softpipe/g_surface.c create mode 100644 src/mesa/pipe/softpipe/g_surface.h create mode 100644 src/mesa/pipe/softpipe/g_tile.h create mode 100644 src/mesa/pipe/softpipe/g_tile_fs.c create mode 100644 src/mesa/pipe/softpipe/g_tile_output.c create mode 100644 src/mesa/pipe/softpipe/sp_context.c create mode 100644 src/mesa/pipe/softpipe/sp_context.h create mode 100644 src/mesa/pipe/softpipe/sp_draw.c create mode 100644 src/mesa/pipe/softpipe/sp_draw.h create mode 100644 src/mesa/pipe/softpipe/sp_headers.h create mode 100644 src/mesa/pipe/softpipe/sp_prim.h create mode 100644 src/mesa/pipe/softpipe/sp_prim_clip.c create mode 100644 src/mesa/pipe/softpipe/sp_prim_cull.c create mode 100644 src/mesa/pipe/softpipe/sp_prim_flatshade.c create mode 100644 src/mesa/pipe/softpipe/sp_prim_offset.c create mode 100644 src/mesa/pipe/softpipe/sp_prim_setup.c create mode 100644 src/mesa/pipe/softpipe/sp_prim_setup.h create mode 100644 src/mesa/pipe/softpipe/sp_prim_twoside.c create mode 100644 src/mesa/pipe/softpipe/sp_prim_unfilled.c create mode 100644 src/mesa/pipe/softpipe/sp_state.h create mode 100644 src/mesa/pipe/softpipe/sp_state_clip.c create mode 100644 src/mesa/pipe/softpipe/sp_state_derived.c create mode 100644 src/mesa/pipe/softpipe/sp_state_fs.c create mode 100644 src/mesa/pipe/softpipe/sp_state_setup.c create mode 100644 src/mesa/pipe/softpipe/sp_state_surface.c create mode 100644 src/mesa/pipe/softpipe/sp_surface.c create mode 100644 src/mesa/pipe/softpipe/sp_surface.h create mode 100644 src/mesa/pipe/softpipe/sp_tile.h create mode 100644 src/mesa/pipe/softpipe/sp_tile_fs.c create mode 100644 src/mesa/pipe/softpipe/sp_tile_output.c delete mode 100644 src/mesa/softpipe/Makefile delete mode 100644 src/mesa/softpipe/generic/Makefile delete mode 100644 src/mesa/softpipe/generic/g_context.c delete mode 100644 src/mesa/softpipe/generic/g_context.h delete mode 100644 src/mesa/softpipe/generic/g_draw.c delete mode 100644 src/mesa/softpipe/generic/g_draw.h delete mode 100644 src/mesa/softpipe/generic/g_headers.h delete mode 100644 src/mesa/softpipe/generic/g_prim.h delete mode 100644 src/mesa/softpipe/generic/g_prim_clip.c delete mode 100644 src/mesa/softpipe/generic/g_prim_cull.c delete mode 100644 src/mesa/softpipe/generic/g_prim_flatshade.c delete mode 100644 src/mesa/softpipe/generic/g_prim_offset.c delete mode 100644 src/mesa/softpipe/generic/g_prim_setup.c delete mode 100644 src/mesa/softpipe/generic/g_prim_setup.h delete mode 100644 src/mesa/softpipe/generic/g_prim_twoside.c delete mode 100644 src/mesa/softpipe/generic/g_prim_unfilled.c delete mode 100644 src/mesa/softpipe/generic/g_state.h delete mode 100644 src/mesa/softpipe/generic/g_state_clip.c delete mode 100644 src/mesa/softpipe/generic/g_state_derived.c delete mode 100644 src/mesa/softpipe/generic/g_state_fs.c delete mode 100644 src/mesa/softpipe/generic/g_state_setup.c delete mode 100644 src/mesa/softpipe/generic/g_state_surface.c delete mode 100644 src/mesa/softpipe/generic/g_surface.c delete mode 100644 src/mesa/softpipe/generic/g_surface.h delete mode 100644 src/mesa/softpipe/generic/g_tile.h delete mode 100644 src/mesa/softpipe/generic/g_tile_fs.c delete mode 100644 src/mesa/softpipe/generic/g_tile_output.c delete mode 100644 src/mesa/softpipe/sp_context.h delete mode 100644 src/mesa/softpipe/sp_defines.h delete mode 100644 src/mesa/softpipe/sp_state.h delete mode 100644 src/mesa/softpipe/state_tracker/Makefile delete mode 100644 src/mesa/softpipe/state_tracker/st_atom.c delete mode 100644 src/mesa/softpipe/state_tracker/st_atom.h delete mode 100644 src/mesa/softpipe/state_tracker/st_atom_blend.c delete mode 100644 src/mesa/softpipe/state_tracker/st_atom_cbuf.c delete mode 100644 src/mesa/softpipe/state_tracker/st_atom_clip.c delete mode 100644 src/mesa/softpipe/state_tracker/st_atom_depth.c delete mode 100644 src/mesa/softpipe/state_tracker/st_atom_fs.c delete mode 100644 src/mesa/softpipe/state_tracker/st_atom_scissor.c delete mode 100644 src/mesa/softpipe/state_tracker/st_atom_setup.c delete mode 100644 src/mesa/softpipe/state_tracker/st_atom_stencil.c delete mode 100644 src/mesa/softpipe/state_tracker/st_atom_viewport.c delete mode 100644 src/mesa/softpipe/state_tracker/st_cb_program.c delete mode 100644 src/mesa/softpipe/state_tracker/st_context.c delete mode 100644 src/mesa/softpipe/state_tracker/st_context.h delete mode 100644 src/mesa/softpipe/state_tracker/st_draw.c delete mode 100644 src/mesa/softpipe/state_tracker/st_draw.h delete mode 100644 src/mesa/softpipe/state_tracker/st_program.h delete mode 100644 src/mesa/softpipe/state_tracker/st_public.h create mode 100644 src/mesa/state_tracker/Makefile create mode 100644 src/mesa/state_tracker/st_atom.c create mode 100644 src/mesa/state_tracker/st_atom.h create mode 100644 src/mesa/state_tracker/st_atom_blend.c create mode 100644 src/mesa/state_tracker/st_atom_cbuf.c create mode 100644 src/mesa/state_tracker/st_atom_clip.c create mode 100644 src/mesa/state_tracker/st_atom_depth.c create mode 100644 src/mesa/state_tracker/st_atom_fs.c create mode 100644 src/mesa/state_tracker/st_atom_scissor.c create mode 100644 src/mesa/state_tracker/st_atom_setup.c create mode 100644 src/mesa/state_tracker/st_atom_stencil.c create mode 100644 src/mesa/state_tracker/st_atom_viewport.c create mode 100644 src/mesa/state_tracker/st_cb_program.c create mode 100644 src/mesa/state_tracker/st_context.c create mode 100644 src/mesa/state_tracker/st_context.h create mode 100644 src/mesa/state_tracker/st_draw.c create mode 100644 src/mesa/state_tracker/st_draw.h create mode 100644 src/mesa/state_tracker/st_program.h create mode 100644 src/mesa/state_tracker/st_public.h (limited to 'src') diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile new file mode 100644 index 0000000000..451911a354 --- /dev/null +++ b/src/mesa/pipe/Makefile @@ -0,0 +1,2 @@ +default: + cd .. ; make diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h new file mode 100644 index 0000000000..716ddd27a9 --- /dev/null +++ b/src/mesa/pipe/p_context.h @@ -0,0 +1,89 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_CONTEXT_H +#define SP_CONTEXT_H + +#include "mtypes.h" + +extern struct softpipe_context *generic_create( void ); + +/* Drawing currently kludged up via the existing tnl/ module. + */ +struct vertex_buffer; + + +/** + * Software pipeline rendering context. Basically a collection of + * state setting functions, plus VBO drawing entrypoint. + */ +struct softpipe_context { + + void (*destroy)( struct softpipe_context * ); + + /* + * Drawing + */ + void (*draw_vb)( struct softpipe_context *softpipe, + struct vertex_buffer *VB ); + + /* + * State functions + */ + void (*set_blend_state)( struct softpipe_context *, + const struct softpipe_blend_state * ); + + void (*set_cbuf_state)( struct softpipe_context *, + const struct softpipe_surface * ); + + void (*set_clip_state)( struct softpipe_context *, + const struct softpipe_clip_state * ); + + void (*set_depth_state)( struct softpipe_context *, + const struct softpipe_depth_state * ); + + void (*set_fs_state)( struct softpipe_context *, + const struct softpipe_fs_state * ); + + void (*set_polygon_stipple)( struct softpipe_context *, + const struct softpipe_poly_stipple * ); + + void (*set_setup_state)( struct softpipe_context *, + const struct softpipe_setup_state * ); + + void (*set_scissor_rect)( struct softpipe_context *, + const struct softpipe_scissor_rect * ); + + void (*set_stencil_state)( struct softpipe_context *, + const struct softpipe_stencil_state * ); + + void (*set_viewport)( struct softpipe_context *, + const struct softpipe_viewport * ); +}; + + +#endif diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h new file mode 100644 index 0000000000..c381865dab --- /dev/null +++ b/src/mesa/pipe/p_defines.h @@ -0,0 +1,121 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_DEFINES_H +#define SP_DEFINES_H + +#define SP_BLENDFACTOR_ONE 0x1 +#define SP_BLENDFACTOR_SRC_COLOR 0x2 +#define SP_BLENDFACTOR_SRC_ALPHA 0x3 +#define SP_BLENDFACTOR_DST_ALPHA 0x4 +#define SP_BLENDFACTOR_DST_COLOR 0x5 +#define SP_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 +#define SP_BLENDFACTOR_CONST_COLOR 0x7 +#define SP_BLENDFACTOR_CONST_ALPHA 0x8 +#define SP_BLENDFACTOR_SRC1_COLOR 0x9 +#define SP_BLENDFACTOR_SRC1_ALPHA 0x0A +#define SP_BLENDFACTOR_ZERO 0x11 +#define SP_BLENDFACTOR_INV_SRC_COLOR 0x12 +#define SP_BLENDFACTOR_INV_SRC_ALPHA 0x13 +#define SP_BLENDFACTOR_INV_DST_ALPHA 0x14 +#define SP_BLENDFACTOR_INV_DST_COLOR 0x15 +#define SP_BLENDFACTOR_INV_CONST_COLOR 0x17 +#define SP_BLENDFACTOR_INV_CONST_ALPHA 0x18 +#define SP_BLENDFACTOR_INV_SRC1_COLOR 0x19 +#define SP_BLENDFACTOR_INV_SRC1_ALPHA 0x1A + +#define SP_BLEND_ADD 0 +#define SP_BLEND_SUBTRACT 1 +#define SP_BLEND_REVERSE_SUBTRACT 2 +#define SP_BLEND_MIN 3 +#define SP_BLEND_MAX 4 + +#define SP_LOGICOP_CLEAR 0 +#define SP_LOGICOP_NOR 1 +#define SP_LOGICOP_AND_INVERTED 2 +#define SP_LOGICOP_COPY_INVERTED 3 +#define SP_LOGICOP_AND_REVERSE 4 +#define SP_LOGICOP_INVERT 5 +#define SP_LOGICOP_XOR 6 +#define SP_LOGICOP_NAND 7 +#define SP_LOGICOP_AND 8 +#define SP_LOGICOP_EQUIV 9 +#define SP_LOGICOP_NOOP 10 +#define SP_LOGICOP_OR_INVERTED 11 +#define SP_LOGICOP_COPY 12 +#define SP_LOGICOP_OR_REVERSE 13 +#define SP_LOGICOP_OR 14 +#define SP_LOGICOP_SET 15 + +#define SP_STENCIL_FUNC_NEVER 0 +#define SP_STENCIL_FUNC_LESS 1 +#define SP_STENCIL_FUNC_EQUAL 2 +#define SP_STENCIL_FUNC_LEQUAL 3 +#define SP_STENCIL_FUNC_GREATER 4 +#define SP_STENCIL_FUNC_NOTEQUAL 5 +#define SP_STENCIL_FUNC_GEQUAL 6 +#define SP_STENCIL_FUNC_ALWAYS 7 + +#define SP_STENCIL_OP_KEEP 0 +#define SP_STENCIL_OP_ZERO 1 +#define SP_STENCIL_OP_REPLACE 2 +#define SP_STENCIL_OP_INCR 3 +#define SP_STENCIL_OP_DECR 4 +#define SP_STENCIL_OP_INCR_WRAP 5 +#define SP_STENCIL_OP_DECR_WRAP 6 +#define SP_STENCIL_OP_INVERT 7 + +/* Note: same as stencil funcs. Also used for shadow/depth compare */ +#define SP_DEPTH_FUNC_NEVER 0 +#define SP_DEPTH_FUNC_LESS 1 +#define SP_DEPTH_FUNC_EQUAL 2 +#define SP_DEPTH_FUNC_LEQUAL 3 +#define SP_DEPTH_FUNC_GREATER 4 +#define SP_DEPTH_FUNC_NOTEQUAL 5 +#define SP_DEPTH_FUNC_GEQUAL 6 +#define SP_DEPTH_FUNC_ALWAYS 7 + +#define SP_TEX_WRAP_REPEAT 0 +#define SP_TEX_WRAP_CLAMP 1 +#define SP_TEX_WRAP_CLAMP_TO_EDGE 2 +#define SP_TEX_WRAP_CLAMP_TO_BORDER 3 +#define SP_TEX_WRAP_MIRROR_REPEAT 4 +#define SP_TEX_WRAP_MIRROR_CLAMP 5 +#define SP_TEX_WRAP_MIRROR_CLAMP_TO_EDGE 6 +#define SP_TEX_WRAP_MIRROR_CLAMP_TO_BORDER 7 + +#define SP_TEX_FILTER_NEAREST 0 +#define SP_TEX_FILTER_LINEAR 1 +#define SP_TEX_FILTER_NEAREST_MIPMAP_NEAREST 2 +#define SP_TEX_FILTER_NEAREST_MIPMAP_LINEAR 3 +#define SP_TEX_FILTER_LINEAR_MIPMAP_NEAREST 4 +#define SP_TEX_FILTER_LINEAR_MIPMAP_LINEAR 5 + +#define SP_TEX_COMPARE_NONE 0 +#define SP_TEX_COMPARE_R_TO_TEXTURE 1 + +#endif diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h new file mode 100644 index 0000000000..150fa9fcad --- /dev/null +++ b/src/mesa/pipe/p_state.h @@ -0,0 +1,184 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_STATE_H +#define SP_STATE_H + +#include "mtypes.h" +#include "vf/vf.h" + +#define WINDING_NONE 0 +#define WINDING_CW 1 +#define WINDING_CCW 2 +#define WINDING_BOTH (WINDING_CW | WINDING_CCW) + +#define FILL_POINT 1 +#define FILL_LINE 2 +#define FILL_TRI 3 + +struct softpipe_setup_state { + GLuint flatshade:1; + GLuint light_twoside:1; + + GLuint front_winding:2; + + GLuint cull_mode:2; + + GLuint fill_cw:2; + GLuint fill_ccw:2; + + GLuint offset_cw:1; + GLuint offset_ccw:1; + + GLuint scissor:1; + GLuint poly_stipple:1; + + GLuint pad:18; + + GLfloat offset_units; + GLfloat offset_scale; +}; + +struct softpipe_poly_stipple { + GLuint stipple[32]; +}; + + +struct softpipe_viewport { + GLfloat scale[4]; + GLfloat translate[4]; +}; + +struct softpipe_scissor_rect { + GLshort minx; + GLshort miny; + GLshort maxx; + GLshort maxy; +}; + + +#define SP_MAX_CLIP_PLANES 6 + +struct softpipe_clip_state { + GLfloat ucp[SP_MAX_CLIP_PLANES][4]; + GLuint nr; +}; + +struct softpipe_fs_state { + struct gl_fragment_program *fp; +}; + +#define SP_MAX_CONSTANT 32 + +struct softpipe_constant_buffer { + GLfloat constant[SP_MAX_CONSTANT][4]; + GLuint nr_constants; +}; + + +struct softpipe_depth_state +{ + GLuint enabled:1; /**< depth test enabled? */ + GLuint writemask:1; /**< allow depth buffer writes? */ + GLuint func:3; /**< depth test func */ + GLfloat clear; /**< Clear value in [0,1] (XXX correct place?) */ +}; + + +struct softpipe_blend_state { + GLuint blend_enable:1; + + GLuint rgb_func:3; + GLuint rgb_src_factor:5; + GLuint rgb_dst_factor:5; + + GLuint alpha_func:3; + GLuint alpha_src_factor:5; + GLuint alpha_dst_factor:5; + + GLuint logicop_enable:1; + GLuint logicop_func:4; +}; + +struct softpipe_blend_color { + GLfloat color[4]; +}; + + +struct softpipe_stencil_state { + GLuint front_enabled:1; + GLuint front_func:3; /**< SP_STENCIL_FUNC_x */ + GLuint front_fail_op:3; /**< SP_STENCIL_OP_x */ + GLuint front_zpass_op:3; /**< SP_STENCIL_OP_x */ + GLuint front_zfail_op:3; /**< SP_STENCIL_OP_x */ + GLuint back_enabled:1; + GLuint back_func:3; + GLuint back_fail_op:3; + GLuint back_zpass_op:3; + GLuint back_zfail_op:3; + GLubyte ref_value[2]; /**< [0] = front, [1] = back */ + GLubyte value_mask[2]; + GLubyte write_mask[2]; + GLubyte clear_value; +}; + + +/* This will change for hardware softpipes... + */ +struct softpipe_surface { + GLubyte *ptr; + GLint stride; + GLuint cpp; + GLuint format; +}; + + +/** + * Texture sampler state. + */ +struct softpipe_sampler_state +{ + GLuint wrap_s:3; /**< SP_TEX_WRAP_x */ + GLuint wrap_t:3; /**< SP_TEX_WRAP_x */ + GLuint wrap_r:3; /**< SP_TEX_WRAP_x */ + GLuint min_filter:3; /**< SP_TEX_FILTER_x */ + GLuint mag_filter:1; /**< SP_TEX_FILTER_LINEAR or _NEAREST */ + GLfloat min_lod; + GLfloat max_lod; + GLfloat lod_bias; +#if 0 /* need these? */ + GLint BaseLevel; /**< min mipmap level, OpenGL 1.2 */ + GLint MaxLevel; /**< max mipmap level, OpenGL 1.2 */ +#endif + GLfloat max_anisotropy; + GLuint compare:1; /**< shadow/depth compare enabled? */ + GLenum compare_mode:1; /**< SP_TEX_COMPARE_x */ + GLenum compare_func:3; /**< SP_DEPTH_FUNC_x */ + GLfloat shadow_ambient; /**< shadow test fail color/intensity */ +}; + +#endif diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile new file mode 100644 index 0000000000..12a8bd0409 --- /dev/null +++ b/src/mesa/pipe/softpipe/Makefile @@ -0,0 +1,3 @@ +default: + cd ../.. ; make + diff --git a/src/mesa/pipe/softpipe/g_context.c b/src/mesa/pipe/softpipe/g_context.c new file mode 100644 index 0000000000..c3329b5c7d --- /dev/null +++ b/src/mesa/pipe/softpipe/g_context.c @@ -0,0 +1,135 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "tnl/t_context.h" +#include "vf/vf.h" + +#include "g_context.h" +#include "g_prim.h" +#include "g_state.h" +#include "g_draw.h" + +static void generic_destroy( struct softpipe_context *softpipe ) +{ + struct generic_context *generic = generic_context( softpipe ); + + draw_destroy( generic->draw ); + + FREE( generic ); +} + + +static void generic_draw_vb( struct softpipe_context *softpipe, + struct vertex_buffer *VB ) +{ + struct generic_context *generic = generic_context( softpipe ); + + if (generic->dirty) + generic_update_derived( generic ); + + draw_vb( generic->draw, VB ); +} + +struct softpipe_context *generic_create( void ) +{ + struct generic_context *generic = CALLOC_STRUCT(generic_context); + + generic->softpipe.destroy = generic_destroy; + generic->softpipe.set_clip_state = generic_set_clip_state; + generic->softpipe.set_viewport = generic_set_viewport; + generic->softpipe.set_setup_state = generic_set_setup_state; + generic->softpipe.set_scissor_rect = generic_set_scissor_rect; + generic->softpipe.set_fs_state = generic_set_fs_state; + generic->softpipe.set_polygon_stipple = generic_set_polygon_stipple; + generic->softpipe.set_cbuf_state = generic_set_cbuf_state; + generic->softpipe.draw_vb = generic_draw_vb; + + + + generic->prim.setup = prim_setup( generic ); + generic->prim.unfilled = prim_unfilled( generic ); + generic->prim.twoside = prim_twoside( generic ); + generic->prim.offset = prim_offset( generic ); + generic->prim.clip = prim_clip( generic ); + generic->prim.flatshade = prim_flatshade( generic ); + generic->prim.cull = prim_cull( generic ); + + + generic->draw = draw_create( generic ); + + ASSIGN_4V( generic->plane[0], -1, 0, 0, 1 ); + ASSIGN_4V( generic->plane[1], 1, 0, 0, 1 ); + ASSIGN_4V( generic->plane[2], 0, -1, 0, 1 ); + ASSIGN_4V( generic->plane[3], 0, 1, 0, 1 ); + ASSIGN_4V( generic->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ + ASSIGN_4V( generic->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ + generic->nr_planes = 6; + + return &generic->softpipe; +} + + + + + + +#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) + +void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ) +{ + stage->nr_tmps = nr; + + if (nr) { + GLubyte *store = MALLOC(MAX_VERTEX_SIZE * nr); + GLuint i; + + stage->tmp = MALLOC(sizeof(struct vertex_header *) * nr); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } +} + +void prim_free_tmps( struct prim_stage *stage ) +{ + if (stage->tmp) { + FREE(stage->tmp[0]); + FREE(stage->tmp); + } +} + + + + + diff --git a/src/mesa/pipe/softpipe/g_context.h b/src/mesa/pipe/softpipe/g_context.h new file mode 100644 index 0000000000..24d467f2c3 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_context.h @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_CONTEXT_H +#define G_CONTEXT_H + +#include "glheader.h" + +#include "softpipe/sp_state.h" +#include "softpipe/sp_context.h" + + + +struct generic_surface; +struct draw_context; +struct prim_stage; + + +enum interp_mode { + INTERP_CONSTANT, + INTERP_LINEAR, + INTERP_PERSPECTIVE +}; + + +#define G_NEW_VIEWPORT 0x1 +#define G_NEW_SETUP 0x2 +#define G_NEW_FS 0x4 +#define G_NEW_BLEND 0x8 +#define G_NEW_CBUF 0x10 +#define G_NEW_CLIP 0x20 +#define G_NEW_SCISSOR 0x40 +#define G_NEW_STIPPLE 0x80 + + + +struct generic_context { + struct softpipe_context softpipe; + + + /* The most recent drawing state as set by the driver: + */ + struct softpipe_viewport viewport; + struct softpipe_setup_state setup; + struct softpipe_fs_state fs; + struct softpipe_blend_state blend; + struct softpipe_surface cbuf; + struct softpipe_clip_state clip; + struct softpipe_scissor_rect scissor; + struct softpipe_poly_stipple poly_stipple; + GLuint dirty; + + + /* Cbuf derived state??? + */ + struct generic_surface *cbuf_surface; + + /* Clip derived state: + */ + GLfloat plane[12][4]; + GLuint nr_planes; + + /* Setup derived state. TODO: this should be passed in the program + * tokens as parameters to DECL instructions. + * + * For now we just set colors to CONST on flatshade, textures to + * perspective always and everything else to linear. + */ + enum interp_mode interp[VF_ATTRIB_MAX]; + + + /* FS + setup derived state: + */ + GLuint fp_attr_to_slot[VF_ATTRIB_MAX]; + GLuint vf_attr_to_slot[VF_ATTRIB_MAX]; + GLuint nr_attrs; + GLuint nr_frag_attrs; + GLuint attr_mask; + + GLboolean need_z; + GLboolean need_w; + + /* Stipple derived state: + */ + GLubyte stipple_masks[16][16]; + + + /* The software clipper/setup engine. + */ + struct { + struct prim_stage *setup; + struct prim_stage *unfilled; + struct prim_stage *twoside; + struct prim_stage *clip; + struct prim_stage *flatshade; + struct prim_stage *offset; + struct prim_stage *cull; + + struct prim_stage *first; + + GLenum prim; + GLuint vertex_size; + } prim; + + /* Temp kludge: + */ + struct draw_context *draw; +}; + + + + +static INLINE struct generic_context * +generic_context( struct softpipe_context *softpipe ) +{ + return (struct generic_context *)softpipe; +} + + + + +#endif diff --git a/src/mesa/pipe/softpipe/g_draw.c b/src/mesa/pipe/softpipe/g_draw.c new file mode 100644 index 0000000000..4e094c21a0 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_draw.c @@ -0,0 +1,696 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "imports.h" + +#include "tnl/t_context.h" + +#include "g_context.h" +#include "g_prim.h" +#include "g_headers.h" +#include "g_draw.h" + +/* This file is a temporary set of hooks to allow us to use the tnl/ + * and vf/ modules until we have replacements in softpipe. + */ + + +struct draw_context +{ + struct generic_context *generic; + + struct vf_attr_map attrs[VF_ATTRIB_MAX]; + GLuint nr_attrs; + GLuint vertex_size; + struct vertex_fetch *vf; + + GLubyte *verts; + GLuint nr_vertices; + GLboolean in_vb; + + GLenum prim; + + /* Helper for tnl: + */ + GLvector4f header; +}; + + +static struct vertex_header *get_vertex( struct draw_context *pipe, + GLuint i ) +{ + return (struct vertex_header *)(pipe->verts + i * pipe->vertex_size); +} + + + +static void draw_allocate_vertices( struct draw_context *draw, + GLuint nr_vertices ) +{ + draw->nr_vertices = nr_vertices; + draw->verts = MALLOC( nr_vertices * draw->vertex_size ); + + draw->generic->prim.first->begin( draw->generic->prim.first ); +} + +static void draw_set_prim( struct draw_context *draw, + GLenum prim ) +{ + draw->prim = prim; + + /* Not done yet - need to force edgeflags to 1 in strip/fan + * primitives. + */ +#if 0 + switch (prim) { + case GL_TRIANGLES: + case GL_POLYGON: + case GL_QUADS: + case GL_QUAD_STRIP: /* yes, we need this */ + respect_edgeflags( pipe, GL_TRUE ); + break; + + default: + respect_edgeflags( pipe, GL_FALSE ); + break; + } +#endif +} + + + +static void do_quad( struct prim_stage *first, + struct vertex_header *v0, + struct vertex_header *v1, + struct vertex_header *v2, + struct vertex_header *v3 ) +{ + struct prim_header prim; + + { + GLuint tmp = v1->edgeflag; + v1->edgeflag = 0; + + prim.v[0] = v0; + prim.v[1] = v1; + prim.v[2] = v3; + first->tri( first, &prim ); + + v1->edgeflag = tmp; + } + + { + GLuint tmp = v3->edgeflag; + v3->edgeflag = 0; + + prim.v[0] = v1; + prim.v[1] = v2; + prim.v[2] = v3; + first->tri( first, &prim ); + + v3->edgeflag = tmp; + } +} + + + + +static void draw_indexed_prim( struct draw_context *draw, + const GLuint *elts, + GLuint count ) +{ + struct prim_stage * const first = draw->generic->prim.first; + struct prim_header prim; + GLuint i; + + prim.det = 0; /* valid from cull stage onwards */ + prim.v[0] = 0; + prim.v[1] = 0; + prim.v[2] = 0; + + switch (draw->prim) { + case GL_POINTS: + for (i = 0; i < count; i ++) { + prim.v[0] = get_vertex( draw, elts[i] ); + + first->point( first, &prim ); + } + break; + + case GL_LINES: + for (i = 0; i+1 < count; i += 2) { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + + first->line( first, &prim ); + } + break; + + case GL_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + prim.v[0] = get_vertex( draw, elts[i-1] ); + prim.v[1] = get_vertex( draw, elts[i] ); + first->line( first, &prim ); + } + + prim.v[0] = get_vertex( draw, elts[count-1] ); + prim.v[1] = get_vertex( draw, elts[0] ); + first->line( first, &prim ); + } + break; + + case GL_LINE_STRIP: + /* I'm guessing it will be necessary to have something like a + * render->reset_line_stipple() method to properly support + * splitting strips into primitives like this. Alternately we + * could just scan ahead to find individual clipped lines and + * otherwise leave the strip intact - that might be better, but + * require more complex code here. + */ + if (count >= 2) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, elts[0] ); + + for (i = 1; i < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, elts[i] ); + + first->line( first, &prim ); + } + } + break; + + case GL_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + prim.v[0] = get_vertex( draw, elts[i + 1] ); + prim.v[1] = get_vertex( draw, elts[i + 0] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + } + else { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + } + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_FAN: + if (count >= 3) { + prim.v[0] = get_vertex( draw, elts[0] ); + prim.v[1] = 0; + prim.v[2] = get_vertex( draw, elts[1] ); + + for (i = 0; i+2 < count; i++) { + prim.v[1] = prim.v[2]; + prim.v[2] = get_vertex( draw, elts[i+2] ); + + first->tri( first, &prim ); + } + } + break; + + case GL_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( first, + get_vertex( draw, elts[i + 0] ), + get_vertex( draw, elts[i + 1] ), + get_vertex( draw, elts[i + 2] ), + get_vertex( draw, elts[i + 3] )); + } + break; + + case GL_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( first, + get_vertex( draw, elts[i + 2] ), + get_vertex( draw, elts[i + 0] ), + get_vertex( draw, elts[i + 1] ), + get_vertex( draw, elts[i + 3] )); + } + break; + + + case GL_POLYGON: + if (count >= 3) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, elts[1] ); + prim.v[2] = get_vertex( draw, elts[0] ); + + for (i = 0; i+2 < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, elts[i+2] ); + + first->tri( first, &prim ); + } + } + break; + + default: + assert(0); + break; + } +} + +static void draw_prim( struct draw_context *draw, + GLuint start, + GLuint count ) +{ + struct prim_stage * const first = draw->generic->prim.first; + struct prim_header prim; + GLuint i; + +// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); + + prim.det = 0; /* valid from cull stage onwards */ + prim.v[0] = 0; + prim.v[1] = 0; + prim.v[2] = 0; + + switch (draw->prim) { + case GL_POINTS: + for (i = 0; i < count; i ++) { + prim.v[0] = get_vertex( draw, start + i ); + first->point( first, &prim ); + } + break; + + case GL_LINES: + for (i = 0; i+1 < count; i += 2) { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + + first->line( first, &prim ); + } + break; + + case GL_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + prim.v[0] = get_vertex( draw, start + i - 1 ); + prim.v[1] = get_vertex( draw, start + i ); + first->line( first, &prim ); + } + + prim.v[0] = get_vertex( draw, start + count - 1 ); + prim.v[1] = get_vertex( draw, start + 0 ); + first->line( first, &prim ); + } + break; + + case GL_LINE_STRIP: + if (count >= 2) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, start + 0 ); + + for (i = 1; i < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, start + i ); + + first->line( first, &prim ); + } + } + break; + + case GL_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + prim.v[0] = get_vertex( draw, start + i + 1 ); + prim.v[1] = get_vertex( draw, start + i + 0 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + } + else { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + } + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_FAN: + if (count >= 3) { + prim.v[0] = get_vertex( draw, start + 0 ); + prim.v[1] = 0; + prim.v[2] = get_vertex( draw, start + 1 ); + + for (i = 0; i+2 < count; i++) { + prim.v[1] = prim.v[2]; + prim.v[2] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + } + break; + + + case GL_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( first, + get_vertex( draw, start + i + 0 ), + get_vertex( draw, start + i + 1 ), + get_vertex( draw, start + i + 2 ), + get_vertex( draw, start + i + 3 )); + } + break; + + case GL_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( first, + get_vertex( draw, start + i + 2 ), + get_vertex( draw, start + i + 0 ), + get_vertex( draw, start + i + 1 ), + get_vertex( draw, start + i + 3 )); + } + break; + + case GL_POLYGON: + if (count >= 3) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, start + 1 ); + prim.v[2] = get_vertex( draw, start + 0 ); + + for (i = 0; i+2 < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + } + break; + + default: + assert(0); + break; + } +} + + +static void draw_release_vertices( struct draw_context *draw ) +{ + draw->generic->prim.first->end( draw->generic->prim.first ); + + FREE(draw->verts); + draw->verts = NULL; +} + + +struct header_dword { + GLuint clipmask:12; + GLuint edgeflag:1; + GLuint pad:19; +}; + + +static void +build_vertex_headers( struct draw_context *draw, + struct vertex_buffer *VB ) +{ + if (draw->header.storage == NULL) { + draw->header.stride = sizeof(GLfloat); + draw->header.size = 1; + draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(GLfloat), 32 ); + draw->header.data = draw->header.storage; + draw->header.count = 0; + draw->header.flags = VEC_SIZE_1 | VEC_MALLOC; + } + + /* Build vertex header attribute. + * + */ + + { + GLuint i; + struct header_dword *header = (struct header_dword *)draw->header.storage; + + /* yes its a hack + */ + assert(sizeof(*header) == sizeof(GLfloat)); + + draw->header.count = VB->Count; + + if (VB->EdgeFlag) { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = VB->ClipMask[i]; + header[i].edgeflag = VB->EdgeFlag[i]; + header[i].pad = 0; + } + } + else if (VB->ClipOrMask) { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = VB->ClipMask[i]; + header[i].edgeflag = 0; + header[i].pad = 0; + } + } + else { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = 0; + header[i].edgeflag = 0; + header[i].pad = 0; + } + } + } + + VB->AttribPtr[VF_ATTRIB_VERTEX_HEADER] = &draw->header; +} + + + + + +static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) +{ + switch (mode) { + case GL_POINTS: + *first = 1; + *incr = 1; + return 0; + case GL_LINES: + *first = 2; + *incr = 2; + return 0; + case GL_LINE_STRIP: + *first = 2; + *incr = 1; + return 0; + case GL_LINE_LOOP: + *first = 2; + *incr = 1; + return 1; + case GL_TRIANGLES: + *first = 3; + *incr = 3; + return 0; + case GL_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + return 0; + case GL_TRIANGLE_FAN: + case GL_POLYGON: + *first = 3; + *incr = 1; + return 1; + case GL_QUADS: + *first = 4; + *incr = 4; + return 0; + case GL_QUAD_STRIP: + *first = 4; + *incr = 2; + return 0; + default: + assert(0); + *first = 1; + *incr = 1; + return 0; + } +} + + +static GLuint trim( GLuint count, GLuint first, GLuint incr ) +{ + if (count < first) + return 0; + else + return count - (count - first) % incr; +} + + +/* This is a hack & will all go away. + */ +void draw_vb(struct draw_context *draw, + struct vertex_buffer *VB ) +{ + GLuint i; + + VB->AttribPtr[VF_ATTRIB_POS] = VB->NdcPtr; + VB->AttribPtr[VF_ATTRIB_BFC0] = VB->ColorPtr[1]; + VB->AttribPtr[VF_ATTRIB_BFC1] = VB->SecondaryColorPtr[1]; + VB->AttribPtr[VF_ATTRIB_CLIP_POS] = VB->ClipPtr; + + /* Build vertex headers: + */ + build_vertex_headers( draw, VB ); + + draw->in_vb = 1; + + /* Allocate the vertices: + */ + draw_allocate_vertices( draw, VB->Count ); + + /* Bind the vb outputs: + */ + vf_set_sources( draw->vf, VB->AttribPtr, 0 ); + + /* Build the hardware or prim-pipe vertices: + */ + vf_emit_vertices( draw->vf, VB->Count, draw->verts ); + + + for (i = 0; i < VB->PrimitiveCount; i++) { + + GLenum mode = VB->Primitive[i].mode; + GLuint start = VB->Primitive[i].start; + GLuint length, first, incr; + + /* Trim the primitive down to a legal size. + */ + draw_prim_info( mode, &first, &incr ); + length = trim( VB->Primitive[i].count, first, incr ); + + if (!length) + continue; + + if (draw->prim != mode) + draw_set_prim( draw, mode ); + + if (VB->Elts) { + draw_indexed_prim( draw, + VB->Elts + start, + length ); + } + else { + draw_prim( draw, + start, + length ); + } + } + + draw_release_vertices( draw ); + draw->verts = NULL; + draw->in_vb = 0; +} + +void draw_set_viewport( struct draw_context *draw, + const GLfloat *scale, + const GLfloat *translate ) +{ + assert(!draw->in_vb); + vf_set_vp_scale_translate( draw->vf, scale, translate ); +} + + + +struct draw_context *draw_create( struct generic_context *generic ) +{ + struct draw_context *draw = CALLOC_STRUCT( draw_context ); + draw->generic = generic; + draw->vf = vf_create( GL_TRUE ); + + return draw; +} + + +void draw_destroy( struct draw_context *draw ) +{ + if (draw->header.storage) + ALIGN_FREE( draw->header.storage ); + + vf_destroy( draw->vf ); + + FREE( draw ); +} + +#define EMIT_ATTR( ATTR, STYLE ) \ +do { \ + draw->attrs[draw->nr_attrs].attrib = ATTR; \ + draw->attrs[draw->nr_attrs].format = STYLE; \ + draw->nr_attrs++; \ +} while (0) + + +void draw_set_vertex_attributes( struct draw_context *draw, + const GLuint *attrs, + GLuint nr_attrs ) +{ + GLuint i; + + draw->nr_attrs = 0; + + EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F); + EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F); + + assert(attrs[0] == VF_ATTRIB_POS); + EMIT_ATTR(attrs[0], EMIT_4F_VIEWPORT); + + for (i = 1; i < nr_attrs; i++) + EMIT_ATTR(attrs[i], EMIT_4F); + + draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); +} + + diff --git a/src/mesa/pipe/softpipe/g_draw.h b/src/mesa/pipe/softpipe/g_draw.h new file mode 100644 index 0000000000..78c93b4df7 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_draw.h @@ -0,0 +1,56 @@ + +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_DRAW_H +#define G_DRAW_H + +#include "glheader.h" +#include "softpipe/sp_state.h" + + +struct draw_context; + +struct draw_context *draw_create( struct generic_context *generic ); + +void draw_destroy( struct draw_context *draw ); + +void draw_set_viewport( struct draw_context *draw, + const GLfloat *scale, + const GLfloat *translate ); + +void draw_set_vertex_attributes( struct draw_context *draw, + const GLuint *attrs, + GLuint nr_attrs ); + +void draw_vb(struct draw_context *draw, + struct vertex_buffer *VB ); + +#endif diff --git a/src/mesa/pipe/softpipe/g_headers.h b/src/mesa/pipe/softpipe/g_headers.h new file mode 100644 index 0000000000..96ff52a453 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_headers.h @@ -0,0 +1,116 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#ifndef G_HEADERS_H +#define G_HEADERS_H + +#define PRIM_POINT 1 +#define PRIM_LINE 2 +#define PRIM_TRI 3 + +struct prim_header { + GLfloat det; + struct vertex_header *v[3]; +}; + +/* Carry some useful information around with the vertices in the prim + * pipe. + */ +struct vertex_header { + GLuint clipmask:12; + GLuint edgeflag:1; + GLuint pad:19; + + GLfloat clip[4]; + + GLfloat data[][4]; /* Note variable size */ +}; + + + + + +/* The rasterizer generates 2x2 quads of fragment and feeds them to + * the current fp_machine (see below). + */ +#define QUAD_BOTTOM_LEFT 0 +#define QUAD_BOTTOM_RIGHT 1 +#define QUAD_TOP_LEFT 2 +#define QUAD_TOP_RIGHT 3 +#define QUAD_SIZE (2*2) + +#define MASK_BOTTOM_LEFT 0x1 +#define MASK_BOTTOM_RIGHT 0x2 +#define MASK_TOP_LEFT 0x4 +#define MASK_TOP_RIGHT 0x8 +#define MASK_ALL 0xf + + +#define NUM_CHANNELS 4 /* avoid confusion between 4 pixels and 4 channels */ + + +struct setup_coefficient { + GLfloat a0[NUM_CHANNELS]; /* in an xyzw layout */ + GLfloat dadx[NUM_CHANNELS]; + GLfloat dady[NUM_CHANNELS]; +}; + + + +/* Encodes everything we need to know about a 2x2 pixel block. Uses + * "Channel-Serial" or "SoA" layout. + * + * Will expand to include non-attribute things like AA coverage and + * maybe prefetched depth from the depth buffer. + */ +struct quad_header { + GLint x0; + GLint y0; + GLuint mask; + GLuint facing; /**< Front or back facing? */ + + struct { + GLfloat color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ + GLfloat depth[QUAD_SIZE]; + } outputs; + + const struct setup_coefficient *coef; + + const enum interp_mode *interp; /* XXX: this information should be + * encoded in fragment program DECL + * statements. */ + + GLuint nr_attrs; +}; + + + + + +#endif diff --git a/src/mesa/pipe/softpipe/g_prim.h b/src/mesa/pipe/softpipe/g_prim.h new file mode 100644 index 0000000000..e3b2c5ea2f --- /dev/null +++ b/src/mesa/pipe/softpipe/g_prim.h @@ -0,0 +1,90 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_PRIM_H +#define G_PRIM_H + +#include "glheader.h" +#include "g_headers.h" + +struct generic_context; + +struct prim_stage *prim_setup( struct generic_context *context ); +struct prim_stage *prim_unfilled( struct generic_context *context ); +struct prim_stage *prim_twoside( struct generic_context *context ); +struct prim_stage *prim_offset( struct generic_context *context ); +struct prim_stage *prim_clip( struct generic_context *context ); +struct prim_stage *prim_flatshade( struct generic_context *context ); +struct prim_stage *prim_cull( struct generic_context *context ); + + +/* Internal structs and helpers for the primitive clip/setup pipeline: + */ +struct prim_stage { + struct generic_context *generic; + + struct prim_stage *next; + + struct vertex_header **tmp; + GLuint nr_tmps; + + void (*begin)( struct prim_stage * ); + + void (*point)( struct prim_stage *, + struct prim_header * ); + + void (*line)( struct prim_stage *, + struct prim_header * ); + + void (*tri)( struct prim_stage *, + struct prim_header * ); + + void (*end)( struct prim_stage * ); +}; + + + +/* Get a writeable copy of a vertex: + */ +static INLINE struct vertex_header * +dup_vert( struct prim_stage *stage, + const struct vertex_header *vert, + GLuint idx ) +{ + struct vertex_header *tmp = stage->tmp[idx]; + memcpy(tmp, vert, stage->generic->prim.vertex_size ); + return tmp; +} + +void prim_free_tmps( struct prim_stage *stage ); +void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); + + +#endif diff --git a/src/mesa/pipe/softpipe/g_prim_clip.c b/src/mesa/pipe/softpipe/g_prim_clip.c new file mode 100644 index 0000000000..742def9c7d --- /dev/null +++ b/src/mesa/pipe/softpipe/g_prim_clip.c @@ -0,0 +1,429 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "g_context.h" +#include "g_prim.h" + +struct clipper { + struct prim_stage stage; + + GLuint active_user_planes; +}; + +/* This is a bit confusing: + */ +static INLINE struct clipper *clipper_stage( struct prim_stage *stage ) +{ + return (struct clipper *)stage; +} + + +#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) + + +/* All attributes are float[4], so this is easy: + */ +static void interp_attr( GLfloat *fdst, + GLfloat t, + const GLfloat *fin, + const GLfloat *fout ) +{ + fdst[0] = LINTERP( t, fout[0], fin[0] ); + fdst[1] = LINTERP( t, fout[1], fin[1] ); + fdst[2] = LINTERP( t, fout[2], fin[2] ); + fdst[3] = LINTERP( t, fout[3], fin[3] ); +} + + + + +/* Interpolate between two vertices to produce a third. + */ +static void interp( struct clipper *clip, + struct vertex_header *dst, + GLfloat t, + const struct vertex_header *out, + const struct vertex_header *in ) +{ + const GLuint nr_attrs = clip->stage.generic->nr_attrs; + GLuint j; + + /* Vertex header. + */ + { + dst->clipmask = 0; + dst->edgeflag = 0; + dst->pad = 0; + } + + /* Clip coordinates: interpolate normally + */ + { + interp_attr(dst->clip, t, in->clip, out->clip); + } + + /* Do the projective divide and insert window coordinates: + */ + { + const GLfloat *pos = dst->clip; + const GLfloat *scale = clip->stage.generic->viewport.scale; + const GLfloat *trans = clip->stage.generic->viewport.translate; + GLfloat oow; + + oow = 1.0 / pos[3]; + + dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; + dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; + dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; + dst->data[0][3] = oow; + } + + + /* Other attributes + */ + for (j = 1; j < nr_attrs-1; j++) { + interp_attr(dst->data[j], t, in->data[j], out->data[j]); + } +} + + +#define CLIP_USER_BIT 0x40 +#define CLIP_CULL_BIT 0x80 + + +static INLINE GLfloat dot4( const GLfloat *a, + const GLfloat *b ) +{ + GLfloat result = (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); + + return result; +} + + +#if 0 +static INLINE void do_tri( struct prim_stage *next, + struct prim_header *header ) +{ + GLuint i; + for (i = 0; i < 3; i++) { + GLfloat *ndc = header->v[i]->data[0]; + _mesa_printf("ndc %f %f %f\n", ndc[0], ndc[1], ndc[2]); + assert(ndc[0] >= -1 && ndc[0] <= 641); + assert(ndc[1] >= 30 && ndc[1] <= 481); + } + _mesa_printf("\n"); + next->tri(next, header); +} +#endif + + +static void emit_poly( struct prim_stage *stage, + struct vertex_header **inlist, + GLuint n ) +{ + struct prim_header header; + GLuint i; + + for (i = 2; i < n; i++) { + header.v[0] = inlist[0]; + header.v[1] = inlist[i-1]; + header.v[2] = inlist[i]; + + { + GLuint tmp0 = header.v[0]->edgeflag; + GLuint tmp2 = header.v[2]->edgeflag; + + if (i != 2) header.v[0]->edgeflag = 0; + if (i != n-1) header.v[2]->edgeflag = 0; + + stage->next->tri( stage->next, &header ); + + header.v[0]->edgeflag = tmp0; + header.v[2]->edgeflag = tmp2; + } + } +} + + +#if 0 +static void emit_poly( struct prim_stage *stage ) +{ + GLuint i; + + for (i = 2; i < n; i++) { + header->v[0] = inlist[0]; + header->v[1] = inlist[i-1]; + header->v[2] = inlist[i]; + + stage->next->tri( stage->next, header ); + } +} +#endif + + +/* Clip a triangle against the viewport and user clip planes. + */ +static void +do_clip_tri( struct prim_stage *stage, + struct prim_header *header, + GLuint clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *a[MAX_CLIPPED_VERTICES]; + struct vertex_header *b[MAX_CLIPPED_VERTICES]; + struct vertex_header **inlist = a; + struct vertex_header **outlist = b; + GLuint tmpnr = 0; + GLuint n = 3; + GLuint i; + + inlist[0] = header->v[0]; + inlist[1] = header->v[1]; + inlist[2] = header->v[2]; + + /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove + * this once we correctly use 16bit masks for userclip planes. + */ + clipmask &= ~CLIP_CULL_BIT; + if (clipmask & CLIP_USER_BIT) { + clipmask &= ~CLIP_USER_BIT; + clipmask |= clipper->active_user_planes; + } + + while (clipmask && n >= 3) { + GLuint plane_idx = ffs(clipmask)-1; + const GLfloat *plane = clipper->stage.generic->plane[plane_idx]; + struct vertex_header *vert_prev = inlist[0]; + GLfloat dp_prev = dot4( vert_prev->clip, plane ); + GLuint outcount = 0; + + clipmask &= ~(1<clip, plane ); + + if (!IS_NEGATIVE(dp_prev)) { + outlist[outcount++] = vert_prev; + } + + if (DIFFERENT_SIGNS(dp, dp_prev)) { + struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; + outlist[outcount++] = new_vert; + + if (IS_NEGATIVE(dp)) { + /* Going out of bounds. Avoid division by zero as we + * know dp != dp_prev from DIFFERENT_SIGNS, above. + */ + GLfloat t = dp / (dp - dp_prev); + interp( clipper, new_vert, t, vert, vert_prev ); + + /* Force edgeflag true in this case: + */ + new_vert->edgeflag = 1; + } else { + /* Coming back in. + */ + GLfloat t = dp_prev / (dp_prev - dp); + interp( clipper, new_vert, t, vert_prev, vert ); + + /* Copy starting vert's edgeflag: + */ + new_vert->edgeflag = vert_prev->edgeflag; + } + } + + vert_prev = vert; + dp_prev = dp; + } + + { + struct vertex_header **tmp = inlist; + inlist = outlist; + outlist = tmp; + n = outcount; + } + } + + /* Emit the polygon as triangles to the setup stage: + */ + if (n >= 3) + emit_poly( stage, inlist, n ); +} + + +/* Clip a line against the viewport and user clip planes. + */ +static void +do_clip_line( struct prim_stage *stage, + struct prim_header *header, + GLuint clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const GLfloat *pos0 = v0->clip; + const GLfloat *pos1 = v1->clip; + GLfloat t0 = 0; + GLfloat t1 = 0; + + /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove + * this once we correctly use 16bit masks for userclip planes. + */ + clipmask &= ~CLIP_CULL_BIT; + if (clipmask & CLIP_USER_BIT) { + clipmask &= ~CLIP_USER_BIT; + clipmask |= clipper->active_user_planes; + } + + while (clipmask) { + GLuint plane_idx = ffs(clipmask)-1; + const GLfloat *plane = clipper->stage.generic->plane[plane_idx]; + + clipmask &= ~(1< t1) t1 = t; + } + + if (dp0 < 0) { + GLfloat t = dp0 / (dp0 - dp1); + if (t > t0) t0 = t; + } + + if (t0 + t1 >= 1.0) + return; /* discard */ + } + + if (v0->clipmask) { + interp( clipper, stage->tmp[0], t0, v0, v1 ); + header->v[0] = stage->tmp[0]; + } + + if (v1->clipmask) { + interp( clipper, stage->tmp[1], t1, v1, v0 ); + header->v[1] = stage->tmp[1]; + } + + stage->next->line( stage->next, header ); +} + + + +static void clip_begin( struct prim_stage *stage ) +{ + struct clipper *clipper = clipper_stage(stage); + GLuint nr = stage->generic->nr_planes; + + /* Hacky bitmask to use when we hit CLIP_USER_BIT: + */ + clipper->active_user_planes = ((1<next->begin( stage->next ); +} + +static void +clip_point( struct prim_stage *stage, + struct prim_header *header ) +{ + if (header->v[0]->clipmask == 0) + stage->next->point( stage->next, header ); +} + + +static void +clip_line( struct prim_stage *stage, + struct prim_header *header ) +{ + GLuint clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask); + + if (clipmask == 0) { + stage->next->line( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask) == 0) { + do_clip_line(stage, header, clipmask); + } +} + + +static void +clip_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + GLuint clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask | + header->v[2]->clipmask); + + if (clipmask == 0) { + stage->next->tri( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask & + header->v[2]->clipmask) == 0) { + do_clip_tri(stage, header, clipmask); + } +} + +static void clip_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + + +struct prim_stage *prim_clip( struct generic_context *generic ) +{ + struct clipper *clipper = CALLOC_STRUCT(clipper); + + prim_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); + + clipper->stage.generic = generic; + clipper->stage.begin = clip_begin; + clipper->stage.point = clip_point; + clipper->stage.line = clip_line; + clipper->stage.tri = clip_tri; + clipper->stage.end = clip_end; + + return &clipper->stage; +} diff --git a/src/mesa/pipe/softpipe/g_prim_cull.c b/src/mesa/pipe/softpipe/g_prim_cull.c new file mode 100644 index 0000000000..633fbcc492 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_prim_cull.c @@ -0,0 +1,119 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_prim.h" + + + +struct cull_stage { + struct prim_stage stage; + GLuint mode; +}; + + + +static INLINE struct cull_stage *cull_stage( struct prim_stage *stage ) +{ + return (struct cull_stage *)stage; +} + + +static void cull_begin( struct prim_stage *stage ) +{ + struct cull_stage *cull = cull_stage(stage); + + cull->mode = stage->generic->setup.cull_mode; + + stage->next->begin( stage->next ); +} + + +static void cull_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + /* Window coords: */ + GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); + GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); + GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); + + GLfloat ex = v0[0] - v2[0]; + GLfloat ey = v0[1] - v2[1]; + GLfloat fx = v1[0] - v2[0]; + GLfloat fy = v1[1] - v2[1]; + + header->det = ex * fy - ey * fx; + + _mesa_printf("%s %f\n", __FUNCTION__, header->det ); + + if (header->det != 0) { + GLuint mode = (header->det < 0) ? WINDING_CW : WINDING_CCW; + + if ((mode & cull_stage(stage)->mode) == 0) + stage->next->tri( stage->next, header ); + } +} + + +static void cull_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void cull_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void cull_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_cull( struct generic_context *generic ) +{ + struct cull_stage *cull = CALLOC_STRUCT(cull_stage); + + prim_alloc_tmps( &cull->stage, 0 ); + + cull->stage.generic = generic; + cull->stage.next = NULL; + cull->stage.begin = cull_begin; + cull->stage.point = cull_point; + cull->stage.line = cull_line; + cull->stage.tri = cull_tri; + cull->stage.end = cull_end; + + return &cull->stage; +} diff --git a/src/mesa/pipe/softpipe/g_prim_flatshade.c b/src/mesa/pipe/softpipe/g_prim_flatshade.c new file mode 100644 index 0000000000..1974a64fa8 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_prim_flatshade.c @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_prim.h" + + +struct flatshade_stage { + struct prim_stage stage; + + const GLuint *lookup; +}; + + + +static INLINE struct flatshade_stage *flatshade_stage( struct prim_stage *stage ) +{ + return (struct flatshade_stage *)stage; +} + + +static void flatshade_begin( struct prim_stage *stage ) +{ + stage->next->begin( stage->next ); +} + + + +static INLINE void copy_attr( GLuint attr, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + if (attr) { + memcpy( dst->data[attr], + src->data[attr], + sizeof(src->data[0]) ); + } +} + +static void copy_colors( struct prim_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + struct flatshade_stage *flatshade = flatshade_stage(stage); + const GLuint *lookup = flatshade->lookup; + + copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); + copy_attr( lookup[VF_ATTRIB_COLOR1], dst, src ); + copy_attr( lookup[VF_ATTRIB_BFC0], dst, src ); + copy_attr( lookup[VF_ATTRIB_BFC1], dst, src ); +} + + + +/* Flatshade tri. Required for clipping and when unfilled tris are + * active, otherwise handled by hardware. + */ +static void flatshade_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = header->v[2]; + + copy_colors(stage, tmp.v[0], tmp.v[2]); + copy_colors(stage, tmp.v[1], tmp.v[2]); + + stage->next->tri( stage->next, &tmp ); +} + + +/* Flatshade line. Required for clipping. + */ +static void flatshade_line( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = header->v[1]; + + copy_colors(stage, tmp.v[0], tmp.v[1]); + + stage->next->line( stage->next, &tmp ); +} + + +static void flatshade_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void flatshade_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_flatshade( struct generic_context *generic ) +{ + struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); + + prim_alloc_tmps( &flatshade->stage, 2 ); + + flatshade->stage.generic = generic; + flatshade->stage.next = NULL; + flatshade->stage.begin = flatshade_begin; + flatshade->stage.point = flatshade_point; + flatshade->stage.line = flatshade_line; + flatshade->stage.tri = flatshade_tri; + flatshade->stage.end = flatshade_end; + + flatshade->lookup = generic->vf_attr_to_slot; + + return &flatshade->stage; +} + + diff --git a/src/mesa/pipe/softpipe/g_prim_offset.c b/src/mesa/pipe/softpipe/g_prim_offset.c new file mode 100644 index 0000000000..801ce9a441 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_prim_offset.c @@ -0,0 +1,157 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" +#include "macros.h" + +#include "g_context.h" +#include "g_prim.h" + + + +struct offset_stage { + struct prim_stage stage; + + GLuint hw_data_offset; + + GLfloat scale; + GLfloat units; +}; + + + +static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) +{ + return (struct offset_stage *)stage; +} + + +static void offset_begin( struct prim_stage *stage ) +{ + struct offset_stage *offset = offset_stage(stage); + + offset->units = stage->generic->setup.offset_units; + offset->scale = stage->generic->setup.offset_scale; + + stage->next->begin( stage->next ); +} + + +/* Offset tri. Some hardware can handle this, but not usually when + * doing unfilled rendering. + */ +static void do_offset_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + GLfloat inv_det = 1.0 / header->det; + + /* Window coords: + */ + GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); + GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); + GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); + + GLfloat ex = v0[0] - v2[2]; + GLfloat fx = v1[0] - v2[2]; + GLfloat ey = v0[1] - v2[2]; + GLfloat fy = v1[1] - v2[2]; + GLfloat ez = v0[2] - v2[2]; + GLfloat fz = v1[2] - v2[2]; + + GLfloat a = ey*fz - ez*fy; + GLfloat b = ez*fx - ex*fz; + + GLfloat ac = a * inv_det; + GLfloat bc = b * inv_det; + GLfloat zoffset; + + if ( ac < 0.0f ) ac = -ac; + if ( bc < 0.0f ) bc = -bc; + + zoffset = offset->units + MAX2( ac, bc ) * offset->scale; + + v0[2] += zoffset; + v1[2] += zoffset; + v2[2] += zoffset; + + stage->next->tri( stage->next, header ); +} + + +static void offset_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = dup_vert(stage, header->v[2], 2); + + do_offset_tri( stage->next, &tmp ); +} + + + +static void offset_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void offset_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void offset_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_offset( struct generic_context *generic ) +{ + struct offset_stage *offset = CALLOC_STRUCT(offset_stage); + + prim_alloc_tmps( &offset->stage, 3 ); + + offset->stage.generic = generic; + offset->stage.next = NULL; + offset->stage.begin = offset_begin; + offset->stage.point = offset_point; + offset->stage.line = offset_line; + offset->stage.tri = offset_tri; + offset->stage.end = offset_end; + + return &offset->stage; +} diff --git a/src/mesa/pipe/softpipe/g_prim_setup.c b/src/mesa/pipe/softpipe/g_prim_setup.c new file mode 100644 index 0000000000..41e1381a60 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_prim_setup.c @@ -0,0 +1,888 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "g_context.h" +#include "g_prim.h" +#include "g_tile.h" + + +/** + * Triangle edge info + */ +struct edge { + GLfloat dx; /* X(v1) - X(v0), used only during setup */ + GLfloat dy; /* Y(v1) - Y(v0), used only during setup */ + GLfloat dxdy; /* dx/dy */ + GLfloat sx; /* first sample point x coord */ + GLfloat sy; + GLint lines; /* number of lines on this edge */ +}; + + +/** + * Triangle setup info (derived from prim_stage). + * Also used for line drawing (taking some liberties). + */ +struct setup_stage { + struct prim_stage stage; /**< This must be first */ + + /* Vertices are just an array of floats making up each attribute in + * turn. Currently fixed at 4 floats, but should change in time. + * Codegen will help cope with this. + */ + const struct vertex_header *vmax; + const struct vertex_header *vmid; + const struct vertex_header *vmin; + const struct vertex_header *vprovoke; + + struct edge ebot; + struct edge etop; + struct edge emaj; + + GLfloat oneoverarea; + + struct setup_coefficient coef[FRAG_ATTRIB_MAX]; + struct quad_header quad; + + struct { + GLint left[2]; /**< [0] = row0, [1] = row1 */ + GLint right[2]; + GLint y; + GLuint y_flags; + GLuint mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ + } span; +}; + + + +/** + * Basically a cast wrapper. + */ +static inline struct setup_stage *setup_stage( struct prim_stage *stage ) +{ + return (struct setup_stage *)stage; +} + + +/** + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. + */ +static inline GLint block( GLint x ) +{ + return x & ~1; +} + + + +static void setup_begin( struct prim_stage *stage ) +{ + setup_stage(stage)->quad.nr_attrs = stage->generic->nr_frag_attrs; +} + + +/** + * Run shader on a quad/block. + */ +static void run_shader_block( struct setup_stage *setup, + GLint x, GLint y, GLuint mask ) +{ + setup->quad.x0 = x; + setup->quad.y0 = y; + setup->quad.mask = mask; + + quad_shade( setup->stage.generic, &setup->quad ); +} + + +/** + * Compute mask which indicates which pixels in the 2x2 quad are actually inside + * the triangle's bounds. + * + * this is pretty nasty... may need to rework flush_spans again to + * fix it, if possible. + */ +static GLuint calculate_mask( struct setup_stage *setup, + GLint x ) +{ + GLuint mask = 0; + + if (x >= setup->span.left[0] && x < setup->span.right[0]) + mask |= MASK_BOTTOM_LEFT; + + if (x >= setup->span.left[1] && x < setup->span.right[1]) + mask |= MASK_TOP_LEFT; + + if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) + mask |= MASK_BOTTOM_RIGHT; + + if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) + mask |= MASK_TOP_RIGHT; + + return mask; +} + + +/** + * Render a horizontal span of quads + */ +static void flush_spans( struct setup_stage *setup ) +{ + GLint minleft, maxright; + GLint x; + + switch (setup->span.y_flags) { + case 3: + minleft = MIN2(setup->span.left[0], setup->span.left[1]); + maxright = MAX2(setup->span.right[0], setup->span.right[1]); + break; + + case 1: + minleft = setup->span.left[0]; + maxright = setup->span.right[0]; + break; + + case 2: + minleft = setup->span.left[1]; + maxright = setup->span.right[1]; + break; + + default: + return; + } + + + for (x = block(minleft); x <= block(maxright); ) + { + run_shader_block( setup, x, + setup->span.y, + calculate_mask( setup, x ) ); + x += 2; + } + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +} + + +static GLboolean setup_sort_vertices( struct setup_stage *setup, + const struct prim_header *prim ) +{ + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + const struct vertex_header *v2 = prim->v[2]; + + setup->vprovoke = v2; + + /* determine bottom to top order of vertices */ + { + GLfloat y0 = v0->data[0][1]; + GLfloat y1 = v1->data[0][1]; + GLfloat y2 = v2->data[0][1]; + if (y0 <= y1) { + if (y1 <= y2) { + /* y0<=y1<=y2 */ + setup->vmin = v0; + setup->vmid = v1; + setup->vmax = v2; + } + else if (y2 <= y0) { + /* y2<=y0<=y1 */ + setup->vmin = v2; + setup->vmid = v0; + setup->vmax = v1; + } + else { + /* y0<=y2<=y1 */ + setup->vmin = v0; + setup->vmid = v2; + setup->vmax = v1; + } + } + else { + if (y0 <= y2) { + /* y1<=y0<=y2 */ + setup->vmin = v1; + setup->vmid = v0; + setup->vmax = v2; + } + else if (y2 <= y1) { + /* y2<=y1<=y0 */ + setup->vmin = v2; + setup->vmid = v1; + setup->vmax = v0; + } + else { + /* y1<=y2<=y0 */ + setup->vmin = v1; + setup->vmid = v2; + setup->vmax = v0; + } + } + } + + setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; + setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; + setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; + + /* xxx: may need to adjust this sign according to the if-tree + * above: + * + * XXX: this is like 'det', but calculated from screen coords?? + */ + { + const GLfloat area = (setup->emaj.dx * setup->ebot.dy - + setup->ebot.dx * setup->emaj.dy); + + setup->oneoverarea = 1.0 / area; + } + + /* XXX need to know if this is a front or back-facing triangle: + * - the GLSL gl_FrontFacing fragment attribute (bool) + * - two-sided stencil test + */ + setup->quad.facing = 0; + + _mesa_printf("%s one-over-area %f\n", __FUNCTION__, setup->oneoverarea ); + + + return GL_TRUE; +} + + +/** + * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + */ +static void const_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i ) +{ + setup->coef[slot].dadx[i] = 0; + setup->coef[slot].dady[i] = 0; + + /* need provoking vertex info! + */ + setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; +} + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a triangle. + */ +static void tri_linear_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i) +{ + GLfloat botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; + GLfloat majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; + GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); + + _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", + slot, "xyzw"[i], + setup->coef[slot].a0[i], + setup->coef[slot].dadx[i], + setup->coef[slot].dady[i]); +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a triangle. + */ +static void tri_persp_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i ) +{ + /* premultiply by 1/w: + */ + GLfloat mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; + GLfloat mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; + GLfloat maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; + + GLfloat botda = mida - mina; + GLfloat majda = maxa - mina; + GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; + GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + setup->coef[slot].a0[i] = (mina - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); +} + + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. + */ +static void setup_tri_coefficients( struct setup_stage *setup ) +{ + const enum interp_mode *interp = setup->stage.generic->interp; + GLuint slot, j; + + /* z and w are done by linear interpolation: + */ + tri_linear_coeff(setup, 0, 2); + tri_linear_coeff(setup, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + switch (interp[slot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + tri_linear_coeff(setup, slot, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + tri_persp_coeff(setup, slot, j); + break; + } + } +} + + + +static void setup_tri_edges( struct setup_stage *setup ) +{ + GLfloat vmin_x = setup->vmin->data[0][0] + 0.5; + GLfloat vmid_x = setup->vmid->data[0][0] + 0.5; + + GLfloat vmin_y = setup->vmin->data[0][1] - 0.5; + GLfloat vmid_y = setup->vmid->data[0][1] - 0.5; + GLfloat vmax_y = setup->vmax->data[0][1] - 0.5; + + setup->emaj.sy = ceilf(vmin_y); + setup->emaj.lines = (GLint) ceilf(vmax_y - setup->emaj.sy); + setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; + setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; + + setup->etop.sy = ceilf(vmid_y); + setup->etop.lines = (GLint) ceilf(vmax_y - setup->etop.sy); + setup->etop.dxdy = setup->etop.dx / setup->etop.dy; + setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; + + setup->ebot.sy = ceilf(vmin_y); + setup->ebot.lines = (GLint) ceilf(vmid_y - setup->ebot.sy); + setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; + setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; +} + + +/** + * Render the upper or lower half of a triangle. + * Scissoring is applied here too. + */ +static void subtriangle( struct setup_stage *setup, + struct edge *eleft, + struct edge *eright, + GLuint lines ) +{ + GLint y, start_y, finish_y; + GLint sy = (GLint)eleft->sy; + + assert((GLint)eleft->sy == (GLint) eright->sy); + assert((GLint)eleft->sy >= 0); /* catch bug in x64? */ + + /* scissor y: + */ + if (setup->stage.generic->setup.scissor) { + start_y = sy; + finish_y = start_y + lines; + + if (start_y < setup->stage.generic->scissor.miny) + start_y = setup->stage.generic->scissor.miny; + + if (finish_y > setup->stage.generic->scissor.maxy) + finish_y = setup->stage.generic->scissor.maxy; + + start_y -= sy; + finish_y -= sy; + } + else { + start_y = 0; + finish_y = lines; + } + + _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + + for (y = start_y; y < finish_y; y++) { + + /* avoid accumulating adds as floats don't have the precision to + * accurately iterate large triangle edges that way. luckily we + * can just multiply these days. + * + * this is all drowned out by the attribute interpolation anyway. + */ + GLint left = (GLint)(eleft->sx + y * eleft->dxdy); + GLint right = (GLint)(eright->sx + y * eright->dxdy); + + /* scissor x: + */ + if (setup->stage.generic->setup.scissor) { + if (left < setup->stage.generic->scissor.minx) + left = setup->stage.generic->scissor.minx; + + if (right > setup->stage.generic->scissor.maxx) + right = setup->stage.generic->scissor.maxx; + } + + if (left < right) { + GLint _y = sy+y; + if (block(_y) != setup->span.y) { + flush_spans(setup); + setup->span.y = block(_y); + } + + setup->span.left[_y&1] = left; + setup->span.right[_y&1] = right; + setup->span.y_flags |= 1<<(_y&1); + } + } + + + /* save the values so that emaj can be restarted: + */ + eleft->sx += lines * eleft->dxdy; + eright->sx += lines * eright->dxdy; + eleft->sy += lines; + eright->sy += lines; +} + + +/** + * Do setup for triangle rasterization, then render the triangle. + */ +static void setup_tri( struct prim_stage *stage, + struct prim_header *prim ) +{ + struct setup_stage *setup = setup_stage( stage ); + + _mesa_printf("%s\n", __FUNCTION__ ); + + setup_sort_vertices( setup, prim ); + setup_tri_coefficients( setup ); + setup_tri_edges( setup ); + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +// setup->span.z_mode = tri_z_mode( setup->ctx ); + +// init_constant_attribs( setup ); + + if (setup->oneoverarea < 0.0) { + /* emaj on left: + */ + subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); + subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); + } + else { + /* emaj on right: + */ + subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); + subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); + } + + flush_spans( setup ); +} + + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a line. + */ +static void +line_linear_coeff(struct setup_stage *setup, GLuint slot, GLuint i) +{ + const GLfloat dz = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + const GLfloat dadx = dz * setup->emaj.dx * setup->oneoverarea; + const GLfloat dady = dz * setup->emaj.dy * setup->oneoverarea; + setup->coef[slot].dadx[i] = dadx; + setup->coef[slot].dady[i] = dady; + setup->coef[slot].a0[i] + = (setup->vmin->data[slot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5) + + dady * (setup->vmin->data[0][1] - 0.5))); +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a line. + */ +static void +line_persp_coeff(struct setup_stage *setup, GLuint slot, GLuint i) +{ + /* XXX to do */ + line_linear_coeff(setup, slot, i); /* XXX temporary */ +} + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmax are initialized. + */ +static INLINE void +setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) +{ + const enum interp_mode *interp = setup->stage.generic->interp; + GLuint slot, j; + + /* use setup->vmin, vmax to point to vertices */ + setup->vprovoke = prim->v[1]; + setup->vmin = prim->v[0]; + setup->vmax = prim->v[1]; + + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + /* NOTE: this is not really 1/area */ + setup->oneoverarea = 1.0 / (setup->emaj.dx * setup->emaj.dx + + setup->emaj.dy * setup->emaj.dy); + + /* z and w are done by linear interpolation: + */ + line_linear_coeff(setup, 0, 2); + line_linear_coeff(setup, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + switch (interp[slot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + line_linear_coeff(setup, slot, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + line_persp_coeff(setup, slot, j); + break; + } + } +} + + +/** + * Plot a pixel in a line segment. + */ +static INLINE void +plot(struct setup_stage *setup, GLint x, GLint y) +{ + const GLint iy = y & 1; + const GLint ix = x & 1; + const GLint quadX = x - ix; + const GLint quadY = y - iy; + const GLint mask = (1 << ix) << (2 * iy); + + if (quadX != setup->quad.x0 || + quadY != setup->quad.y0) + { + /* flush prev quad, start new quad */ + + if (setup->quad.x0 != -1) + quad_shade(setup->stage.generic, &setup->quad); + + setup->quad.x0 = quadX; + setup->quad.y0 = quadY; + setup->quad.mask = 0x0; + } + + setup->quad.mask |= mask; +} + + + +/** + * Do setup for line rasterization, then render the line. + * XXX single-pixel width, no stipple, etc + * XXX no scissoring yet. + */ +static void +setup_line(struct prim_stage *stage, struct prim_header *prim) +{ + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + struct setup_stage *setup = setup_stage( stage ); + + GLint x0 = (GLint) v0->data[0][0]; + GLint x1 = (GLint) v1->data[0][0]; + GLint y0 = (GLint) v0->data[0][1]; + GLint y1 = (GLint) v1->data[0][1]; + GLint dx = x1 - x0; + GLint dy = y1 - y0; + GLint xstep, ystep; + + if (dx == 0 && dy == 0) + return; + + setup_line_coefficients(setup, prim); + + if (dx < 0) { + dx = -dx; /* make positive */ + xstep = -1; + } + else { + xstep = 1; + } + + if (dy < 0) { + dy = -dy; /* make positive */ + ystep = -1; + } + else { + ystep = 1; + } + + assert(dx >= 0); + assert(dy >= 0); + + setup->quad.x0 = setup->quad.y0 = -1; + setup->quad.mask = 0x0; + + if (dx > dy) { + /*** X-major line ***/ + GLint i; + const GLint errorInc = dy + dy; + GLint error = errorInc - dx; + const GLint errorDec = error - dx; + + for (i = 0; i < dx; i++) { + plot(setup, x0, y0); + + x0 += xstep; + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + y0 += ystep; + } + } + } + else { + /*** Y-major line ***/ + GLint i; + const GLint errorInc = dx + dx; + GLint error = errorInc - dy; + const GLint errorDec = error - dy; + + for (i = 0; i < dy; i++) { + plot(setup, x0, y0); + + y0 += ystep; + + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + x0 += xstep; + } + } + } + + /* draw final quad */ + if (setup->quad.mask) { + quad_shade(setup->stage.generic, &setup->quad); + } +} + + +/** + * Do setup for point rasterization, then render the point. + * Round or square points... + * XXX could optimize a lot for 1-pixel points. + */ +static void +setup_point(struct prim_stage *stage, struct prim_header *prim) +{ + struct setup_stage *setup = setup_stage( stage ); + GLfloat halfSize = 7.3; /*XXX this is a vertex attrib */ + GLfloat halfSizeSquared = halfSize * halfSize; + const struct vertex_header *v0 = prim->v[0]; + const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; + const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; + const GLint ixmin = block((GLint) (x - halfSize)); + const GLint ixmax = block((GLint) (x + halfSize)); + const GLint iymin = block((GLint) (y - halfSize)); + const GLint iymax = block((GLint) (y + halfSize)); + GLboolean round = GL_TRUE; + GLint ix, iy; + GLuint slot, j; + + /* For points, all interpolants are constant-valued. + * However, for point sprites, we'll need to setup texcoords appropriately. + * XXX: which coefficients are the texcoords??? + * We may do point sprites as textured quads... + * + * KW: We don't know which coefficients are texcoords - ultimately + * the choice of what interpolation mode to use for each attribute + * should be determined by the fragment program, using + * per-attribute declaration statements that include interpolation + * mode as a parameter. So either the fragment program will have + * to be adjusted for pointsprite vs normal point behaviour, or + * otherwise a special interpolation mode will have to be defined + * which matches the required behaviour for point sprites. But - + * the latter is not a feature of normal hardware, and as such + * probably should be ruled out on that basis. + */ + setup->vprovoke = prim->v[0]; + const_coeff(setup, 0, 2); + const_coeff(setup, 0, 3); + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + } + + /* XXX need to clip against scissor bounds too */ + + for (iy = iymin; iy <= iymax; iy += 2) { + for (ix = ixmin; ix <= ixmax; ix += 2) { + + if (round) { + /* rounded points */ + /* XXX for GL_SMOOTH, need to compute per-fragment coverage too */ + GLfloat dx, dy; + + setup->quad.mask = 0x0; + + dx = (ix + 0.5) - x; + dy = (iy + 0.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_BOTTOM_LEFT; + + dx = (ix + 1.5) - x; + dy = (iy + 0.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_BOTTOM_RIGHT; + + dx = (ix + 0.5) - x; + dy = (iy + 1.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_TOP_LEFT; + + dx = (ix + 1.5) - x; + dy = (iy + 1.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_TOP_RIGHT; + } + else { + /* square points */ + setup->quad.mask = 0xf; + + if (ix + 0.5 < x - halfSize) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + + if (ix + 1.5 > x + halfSize) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + + if (iy + 0.5 < y - halfSize) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + + if (iy + 1.5 > y + halfSize) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + } + + if (setup->quad.mask) { + setup->quad.x0 = ix; + setup->quad.y0 = iy; + quad_shade( setup->stage.generic, &setup->quad ); + } + } + } +} + + + +static void setup_end( struct prim_stage *stage ) +{ +} + + +struct prim_stage *prim_setup( struct generic_context *generic ) +{ + struct setup_stage *setup = CALLOC_STRUCT(setup_stage); + + setup->stage.generic = generic; + setup->stage.begin = setup_begin; + setup->stage.point = setup_point; + setup->stage.line = setup_line; + setup->stage.tri = setup_tri; + setup->stage.end = setup_end; + + setup->quad.coef = setup->coef; + + return &setup->stage; +} diff --git a/src/mesa/pipe/softpipe/g_prim_setup.h b/src/mesa/pipe/softpipe/g_prim_setup.h new file mode 100644 index 0000000000..40a70c543e --- /dev/null +++ b/src/mesa/pipe/softpipe/g_prim_setup.h @@ -0,0 +1,121 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef S_TRI_H +#define S_TRI_H + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "imports.h" +#include "s_tri_public.h" +#include "s_context.h" + + + +struct tri_context; +struct fp_context; +struct be_context; + +/* Note the rasterizer does not take a GLcontext argument. This is + * deliberate. + */ +struct tri_context *tri_create_context( GLcontext *ctx ); + +void tri_destroy_context( struct tri_context *tri ); + +void tri_set_fp_context( struct tri_context *tri, + struct fp_context *fp, + void (*fp_run)( struct fp_context *fp, + const struct fp_inputs *, + struct fp_outputs * )); + + +void tri_set_be_context( struct tri_context *tri, + struct be_context *be, + void (*be_run)( struct be_context *be, + const struct fp_outputs * )); + +void tri_set_attribs( struct tri_context *tri, + const struct attr_info *info, + GLuint nr_attrib ); + +void tri_set_backface( struct tri_context *tri, + GLfloat backface ); + +void tri_set_scissor( struct tri_context *tri, + GLint x, + GLint y, + GLuint width, + GLuint height, + GLboolean enabled ); + +void tri_set_stipple( struct tri_context *tri, + const GLuint *pattern, + GLboolean enabled ); + +/* Unfilled triangles will be handled elsewhere (higher in the + * pipeline), as will things like stipple (lower in the pipeline). + */ + +void tri_triangle( struct tri_context *tri, + const struct vertex *v0, + const struct vertex *v1, + const struct vertex *v2 ); + +/* TODO: rasterize_line, rasterize_point?? + * How will linestipple work? + */ + + +#ifdef SETUP_PRIVATE + + + + + + +GLboolean tri_setup( struct tri_context *tri, + const struct vertex *v0, + const struct vertex *v1, + const struct vertex *v2 ); + +void tri_rasterize( struct tri_context *tri ); +void tri_rasterize_spans( struct tri_context *tri ); + + + + + + +#endif +#endif diff --git a/src/mesa/pipe/softpipe/g_prim_twoside.c b/src/mesa/pipe/softpipe/g_prim_twoside.c new file mode 100644 index 0000000000..334ae2f537 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_prim_twoside.c @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" +#include "g_context.h" +#include "g_prim.h" + + +struct twoside_stage { + struct prim_stage stage; + + GLfloat facing; + const GLuint *lookup; +}; + + +static INLINE struct twoside_stage *twoside_stage( struct prim_stage *stage ) +{ + return (struct twoside_stage *)stage; +} + + +static void twoside_begin( struct prim_stage *stage ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + twoside->facing = (stage->generic->setup.front_winding == WINDING_CW) ? -1 : 1; + + stage->next->begin( stage->next ); +} + + +static INLINE void copy_color( GLuint attr_dst, + GLuint attr_src, + struct vertex_header *v ) +{ + if (attr_dst && attr_src) { + memcpy( v->data[attr_dst], + v->data[attr_src], + sizeof(v->data[0]) ); + } +} + + +static struct vertex_header *copy_bfc( struct twoside_stage *twoside, + const struct vertex_header *v, + GLuint idx ) +{ + struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); + + copy_color( twoside->lookup[VF_ATTRIB_COLOR0], + twoside->lookup[VF_ATTRIB_BFC0], + tmp ); + + copy_color( twoside->lookup[VF_ATTRIB_COLOR1], + twoside->lookup[VF_ATTRIB_BFC1], + tmp ); + + return tmp; +} + + +/* Twoside tri: + */ +static void twoside_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + if (header->det * twoside->facing < 0) { + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = copy_bfc(twoside, header->v[0], 0); + tmp.v[1] = copy_bfc(twoside, header->v[1], 1); + tmp.v[2] = copy_bfc(twoside, header->v[2], 2); + + stage->next->tri( stage->next, &tmp ); + } + else { + stage->next->tri( stage->next, header ); + } +} + + +static void twoside_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void twoside_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void twoside_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + + + +struct prim_stage *prim_twoside( struct generic_context *generic ) +{ + struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); + + prim_alloc_tmps( &twoside->stage, 3 ); + + twoside->stage.generic = generic; + twoside->stage.next = NULL; + twoside->stage.begin = twoside_begin; + twoside->stage.point = twoside_point; + twoside->stage.line = twoside_line; + twoside->stage.tri = twoside_tri; + twoside->stage.end = twoside_end; + + twoside->lookup = generic->vf_attr_to_slot; + + return &twoside->stage; +} diff --git a/src/mesa/pipe/softpipe/g_prim_unfilled.c b/src/mesa/pipe/softpipe/g_prim_unfilled.c new file mode 100644 index 0000000000..7c6ee79f92 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_prim_unfilled.c @@ -0,0 +1,165 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_prim.h" + + + +struct unfilled_stage { + struct prim_stage stage; + + GLuint mode[2]; +}; + + +static INLINE struct unfilled_stage *unfilled_stage( struct prim_stage *stage ) +{ + return (struct unfilled_stage *)stage; +} + + +static void unfilled_begin( struct prim_stage *stage ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + + unfilled->mode[0] = stage->generic->setup.fill_ccw; + unfilled->mode[1] = stage->generic->setup.fill_cw; + + stage->next->begin( stage->next ); +} + +static void point( struct prim_stage *stage, + struct vertex_header *v0 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + stage->next->point( stage->next, &tmp ); +} + +static void line( struct prim_stage *stage, + struct vertex_header *v0, + struct vertex_header *v1 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + tmp.v[1] = v1; + stage->next->line( stage->next, &tmp ); +} + + +static void points( struct prim_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (v0->edgeflag) point( stage, v0 ); + if (v1->edgeflag) point( stage, v1 ); + if (v2->edgeflag) point( stage, v2 ); +} + +static void lines( struct prim_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (v0->edgeflag) line( stage, v0, v1 ); + if (v1->edgeflag) line( stage, v1, v2 ); + if (v2->edgeflag) line( stage, v2, v0 ); +} + + +/* Unfilled tri: + * + * Note edgeflags in the vertex struct is not sufficient as we will + * need to manipulate them when decomposing primitives??? + */ +static void unfilled_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + GLuint mode = unfilled->mode[header->det < 0]; + + switch (mode) { + case FILL_TRI: + stage->next->tri( stage->next, header ); + break; + + case FILL_LINE: + lines( stage, header ); + break; + + case GL_POINT: + points( stage, header ); + break; + } +} + +static void unfilled_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void unfilled_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void unfilled_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_unfilled( struct generic_context *generic ) +{ + struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); + + prim_alloc_tmps( &unfilled->stage, 0 ); + + unfilled->stage.generic = generic; + unfilled->stage.next = NULL; + unfilled->stage.tmp = NULL; + unfilled->stage.begin = unfilled_begin; + unfilled->stage.point = unfilled_point; + unfilled->stage.line = unfilled_line; + unfilled->stage.tri = unfilled_tri; + unfilled->stage.end = unfilled_end; + + return &unfilled->stage; +} diff --git a/src/mesa/pipe/softpipe/g_state.h b/src/mesa/pipe/softpipe/g_state.h new file mode 100644 index 0000000000..5077d37ab7 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_state.h @@ -0,0 +1,61 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef PRIM_H +#define PRIM_H + +#include "glheader.h" +#include "softpipe/sp_state.h" + + +void generic_set_clip_state( struct softpipe_context *, + const struct softpipe_clip_state * ); + +void generic_set_viewport( struct softpipe_context *, + const struct softpipe_viewport * ); + +void generic_set_setup_state( struct softpipe_context *, + const struct softpipe_setup_state * ); + +void generic_set_scissor_rect( struct softpipe_context *, + const struct softpipe_scissor_rect * ); + +void generic_set_fs_state( struct softpipe_context *, + const struct softpipe_fs_state * ); + +void generic_set_polygon_stipple( struct softpipe_context *, + const struct softpipe_poly_stipple * ); + +void generic_set_cbuf_state( struct softpipe_context *, + const struct softpipe_surface * ); + +void generic_update_derived( struct generic_context *generic ); + +#endif diff --git a/src/mesa/pipe/softpipe/g_state_clip.c b/src/mesa/pipe/softpipe/g_state_clip.c new file mode 100644 index 0000000000..b90034e11e --- /dev/null +++ b/src/mesa/pipe/softpipe/g_state_clip.c @@ -0,0 +1,69 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_state.h" +#include "g_draw.h" + + + +void generic_set_clip_state( struct softpipe_context *softpipe, + const struct softpipe_clip_state *clip ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy(&generic->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); + + generic->nr_planes = 6 + clip->nr; + generic->dirty |= G_NEW_CLIP; +} + + + +/* Called when driver state tracker notices changes to the viewport + * matrix: + */ +void generic_set_viewport( struct softpipe_context *softpipe, + const struct softpipe_viewport *viewport ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy(&generic->viewport, viewport, sizeof(viewport)); + + /* Using tnl/ and vf/ modules is temporary while getting started. + * Full softpipe will have vertex shader, vertex fetch of its own. + */ + draw_set_viewport( generic->draw, viewport->scale, viewport->translate ); + generic->dirty |= G_NEW_VIEWPORT; +} + + + diff --git a/src/mesa/pipe/softpipe/g_state_derived.c b/src/mesa/pipe/softpipe/g_state_derived.c new file mode 100644 index 0000000000..678825afe1 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_state_derived.c @@ -0,0 +1,136 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "macros.h" +#include "enums.h" +#include "program.h" + +#include "g_context.h" +#include "g_draw.h" +#include "g_state.h" + +#define EMIT_ATTR( ATTR, FRAG_ATTR, INTERP ) \ +do { \ + slot_to_vf_attr[generic->nr_attrs] = ATTR; \ + generic->vf_attr_to_slot[ATTR] = generic->nr_attrs; \ + generic->fp_attr_to_slot[FRAG_ATTR] = generic->nr_attrs; \ + generic->interp[generic->nr_attrs] = INTERP; \ + generic->nr_attrs++; \ + attr_mask |= (1<fs.fp; + const GLuint inputsRead = fp->Base.InputsRead; + GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; + GLuint attr_mask = 0; + GLuint i; + + generic->nr_attrs = 0; + memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); + + memset(generic->fp_attr_to_slot, 0, sizeof(generic->vf_attr_to_slot)); + memset(generic->vf_attr_to_slot, 0, sizeof(generic->fp_attr_to_slot)); + + /* TODO - Figure out if we need to do perspective divide, etc. + */ + EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); + + /* Pull in the rest of the attributes. They are all in float4 + * format. Future optimizations could be to keep some attributes + * as fixed point or ubyte format. + */ + for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { + if (inputsRead & (i << i)) { + EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); + } + } + + for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { + if (inputsRead & (i << i)) { + EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); + } + } + + generic->nr_frag_attrs = generic->nr_attrs; + + /* Additional attributes required for setup: Just twosided + * lighting. Edgeflag is dealt with specially by setting bits in + * the vertex header. + */ + if (generic->setup.light_twoside) { + if (inputsRead & FRAG_BIT_COL0) { + EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ + } + + if (inputsRead & FRAG_BIT_COL1) { + EMIT_ATTR(VF_ATTRIB_BFC1, FRAG_ATTRIB_MAX, 0); /* XXX: discard after setup */ + } + } + + if (attr_mask != generic->attr_mask) { + generic->attr_mask = attr_mask; + + draw_set_vertex_attributes( generic->draw, + slot_to_vf_attr, + generic->nr_attrs ); + } +} + + +/* Hopefully this will remain quite simple, otherwise need to pull in + * something like the state tracker mechanism. + */ +void generic_update_derived( struct generic_context *generic ) +{ + if (generic->dirty & (G_NEW_SETUP | G_NEW_FS)) + calculate_vertex_layout( generic ); + + generic->dirty = 0; +} diff --git a/src/mesa/pipe/softpipe/g_state_fs.c b/src/mesa/pipe/softpipe/g_state_fs.c new file mode 100644 index 0000000000..8a430c0c65 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_state_fs.c @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "g_context.h" +#include "g_state.h" + + + +void generic_set_fs_state( struct softpipe_context *softpipe, + const struct softpipe_fs_state *fs ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy(&generic->fs, fs, sizeof(*fs)); + + generic->dirty |= G_NEW_FS; +} + + + + + + + + + diff --git a/src/mesa/pipe/softpipe/g_state_setup.c b/src/mesa/pipe/softpipe/g_state_setup.c new file mode 100644 index 0000000000..2a5176c5a1 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_state_setup.c @@ -0,0 +1,122 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "g_context.h" +#include "g_state.h" +#include "g_prim.h" + + + + +static void validate_prim_pipe( struct generic_context *generic ) +{ + struct prim_stage *next = generic->prim.setup; + + /* TODO: make the current primitive part of the state and build + * shorter pipelines for lines & points. + */ + if (generic->setup.fill_cw != FILL_TRI || + generic->setup.fill_ccw != FILL_TRI) { + + generic->prim.unfilled->next = next; + next = generic->prim.unfilled; + } + + if (generic->setup.offset_cw || + generic->setup.offset_ccw) { + generic->prim.offset->next = next; + next = generic->prim.offset; + } + + if (generic->setup.light_twoside) { + generic->prim.twoside->next = next; + next = generic->prim.twoside; + } + + /* Always run the cull stage as we calculate determinant there + * also. Fix this.. + */ + { + generic->prim.cull->next = next; + next = generic->prim.cull; + } + + + /* Clip stage + */ + { + generic->prim.clip->next = next; + next = generic->prim.clip; + } + + /* Do software flatshading prior to clipping. XXX: should only do + * this for clipped primitives, ie it is a part of the clip + * routine. + */ + if (generic->setup.flatshade) { + generic->prim.flatshade->next = next; + next = generic->prim.flatshade; + } + + + generic->prim.first = next; +} + + + + +void generic_set_setup_state( struct softpipe_context *softpipe, + const struct softpipe_setup_state *setup ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy( &generic->setup, setup, sizeof(*setup) ); + + validate_prim_pipe( generic ); + generic->dirty |= G_NEW_SETUP; +} + + + +void generic_set_scissor_rect( struct softpipe_context *softpipe, + const struct softpipe_scissor_rect *scissor ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy( &generic->scissor, scissor, sizeof(*scissor) ); + generic->dirty |= G_NEW_SCISSOR; +} + + +void generic_set_polygon_stipple( struct softpipe_context *softpipe, + const struct softpipe_poly_stipple *stipple ) +{ + struct generic_context *generic = generic_context(softpipe); + + memcpy( &generic->poly_stipple, stipple, sizeof(*stipple) ); + generic->dirty |= G_NEW_STIPPLE; +} diff --git a/src/mesa/pipe/softpipe/g_state_surface.c b/src/mesa/pipe/softpipe/g_state_surface.c new file mode 100644 index 0000000000..c791ce4640 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_state_surface.c @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "g_context.h" +#include "g_state.h" +#include "g_surface.h" + + +/* This is all a total hack. + */ +void generic_set_cbuf_state( struct softpipe_context *softpipe, + const struct softpipe_surface *surface ) +{ + struct generic_context *generic = generic_context(softpipe); + + if (generic->cbuf_surface == NULL) { + generic->cbuf_surface = CALLOC_STRUCT(generic_surface); + generic->cbuf_surface->type = &gs_rgba8; + } + + generic->cbuf_surface->surface = *surface; + generic->dirty |= G_NEW_CBUF; +} + + diff --git a/src/mesa/pipe/softpipe/g_surface.c b/src/mesa/pipe/softpipe/g_surface.c new file mode 100644 index 0000000000..2df30cde70 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_surface.c @@ -0,0 +1,153 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "g_context.h" +#include "g_state.h" +#include "g_surface.h" +#include "g_headers.h" + +static void rgba8_read_quad_f( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + rgba[k][0] = ptr[0] * (1.0 / 255.0); + rgba[k][1] = ptr[1] * (1.0 / 255.0); + rgba[k][2] = ptr[2] * (1.0 / 255.0); + rgba[k][3] = ptr[3] * (1.0 / 255.0); + } + } +} + +static void rgba8_read_quad_f_swz( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + rrrr[0][k] = ptr[0] * (1.0 / 255.0); + rrrr[1][k] = ptr[1] * (1.0 / 255.0); + rrrr[2][k] = ptr[2] * (1.0 / 255.0); + rrrr[3][k] = ptr[3] * (1.0 / 255.0); + } + } +} + +static void rgba8_write_quad_f( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + ptr[0] = rgba[k][0] * 255.0; + ptr[1] = rgba[k][1] * 255.0; + ptr[2] = rgba[k][2] * 255.0; + ptr[3] = rgba[k][3] * 255.0; + } + } +} + +static void rgba8_write_quad_f_swz( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + ptr[0] = rrrr[0][k] * 255.0; + ptr[1] = rrrr[1][k] * 255.0; + ptr[2] = rrrr[2][k] * 255.0; + ptr[3] = rrrr[3][k] * 255.0; + } + } +} + + + + +static void rgba8_read_quad_ub( struct generic_surface *gs, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + rgba[k][0] = ptr[0]; + rgba[k][1] = ptr[1]; + rgba[k][2] = ptr[2]; + rgba[k][3] = ptr[3]; + } + } +} + + +static void rgba8_write_quad_ub( struct generic_surface *gs, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + ptr[0] = rgba[k][0]; + ptr[1] = rgba[k][1]; + ptr[2] = rgba[k][2]; + ptr[3] = rgba[k][3]; + } + } +} + + + + +struct generic_surface_type gs_rgba8 = { + G_SURFACE_RGBA_8888, + rgba8_read_quad_f, + rgba8_read_quad_f_swz, + rgba8_read_quad_ub, + rgba8_write_quad_f, + rgba8_write_quad_f_swz, + rgba8_write_quad_ub, +}; + + + diff --git a/src/mesa/pipe/softpipe/g_surface.h b/src/mesa/pipe/softpipe/g_surface.h new file mode 100644 index 0000000000..23c28f0bb4 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_surface.h @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_SURFACE_H +#define G_SURFACE_H + +#include "glheader.h" +#include "g_headers.h" + +struct generic_surface; + +#define G_SURFACE_RGBA_8888 0x1 + +/* Internal structs and helpers for the primitive clip/setup pipeline: + */ +struct generic_surface_type { + + GLuint format; + + void (*read_quad_f)( struct generic_surface *, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ); + + void (*read_quad_f_swz)( struct generic_surface *, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ); + + void (*read_quad_ub)( struct generic_surface *, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ); + + + void (*write_quad_f)( struct generic_surface *, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ); + + void (*write_quad_f_swz)( struct generic_surface *, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ); + + + void (*write_quad_ub)( struct generic_surface *, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ); + + +}; + + +struct generic_surface { + struct generic_surface_type *type; + struct softpipe_surface surface; +}; + + +static INLINE void gs_read_quad_f( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + gs->type->read_quad_f(gs, x, y, rgba); +} + +static INLINE void gs_read_quad_f_swz( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + gs->type->read_quad_f_swz(gs, x, y, rrrr); +} + +static INLINE void gs_write_quad_f( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + gs->type->write_quad_f(gs, x, y, rgba); +} + +static INLINE void gs_write_quad_f_swz( struct generic_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + gs->type->write_quad_f_swz(gs, x, y, rrrr); +} + +/* Like this, or hidden? + */ +struct generic_surface_type gs_rgba8; + +#endif diff --git a/src/mesa/pipe/softpipe/g_tile.h b/src/mesa/pipe/softpipe/g_tile.h new file mode 100644 index 0000000000..6ae5212356 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_tile.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#ifndef G_TILE_H +#define G_TILE_H + +struct generic_context; +struct quad_header; + +void quad_shade( struct generic_context *generic, + struct quad_header *quad ); + +void quad_output( struct generic_context *generic, + struct quad_header *quad ); + +#endif diff --git a/src/mesa/pipe/softpipe/g_tile_fs.c b/src/mesa/pipe/softpipe/g_tile_fs.c new file mode 100644 index 0000000000..4c5e6efaed --- /dev/null +++ b/src/mesa/pipe/softpipe/g_tile_fs.c @@ -0,0 +1,203 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "g_context.h" +#include "g_headers.h" +#include "g_tile.h" + +struct exec_machine { + const struct setup_coefficient *coef; + + GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE]; +}; + + +/** + * Compute quad's attributes values, as constants (GL_FLAT shading). + */ +static INLINE void cinterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + exec->attr[attrib][i][j] = exec->coef[attrib].a0[i]; + } +} + + +/** + * Compute quad's attribute values by linear interpolation. + * + * Push into the fp: + * + * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx + * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy + */ +static INLINE void linterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; + exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + + exec->coef[attrib].dadx[i] * x + + exec->coef[attrib].dady[i] * y); + } +} + + +/** + * Compute quad's attribute values by linear interpolation with + * perspective correction. + * + * Push into the fp: + * + * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx + * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy + * INPUT[attr] = MUL INPUT[attr], INPUT_WPOS.wwww + * + * (Or should that be 1/w ???) + */ +static INLINE void pinterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; + const GLfloat invW = exec->attr[FRAG_ATTRIB_WPOS][3][j]; + exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + + exec->coef[attrib].dadx[i] * x + + exec->coef[attrib].dady[i] * y) * invW); + } +} + + + +/* This should be done by the fragment shader execution unit (code + * generated from the decl instructions). Do it here for now. + */ +void quad_shade( struct generic_context *generic, + struct quad_header *quad ) +{ + struct exec_machine exec; + GLfloat fx = quad->x0; + GLfloat fy = quad->y0; + GLuint i, j; + + exec.coef = quad->coef; + + /* Position: + */ + exec.attr[FRAG_ATTRIB_WPOS][0][0] = fx; + exec.attr[FRAG_ATTRIB_WPOS][0][1] = fx + 1.0; + exec.attr[FRAG_ATTRIB_WPOS][0][2] = fx; + exec.attr[FRAG_ATTRIB_WPOS][0][3] = fx + 1.0; + + exec.attr[FRAG_ATTRIB_WPOS][1][0] = fy; + exec.attr[FRAG_ATTRIB_WPOS][1][1] = fy; + exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0; + exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; + + /* Z and W are done by linear interpolation: + * XXX we'll probably have to use integers for Z + */ + if (generic->need_z) { + linterp(&exec, 0, 2); + } + + if (generic->need_w) { + linterp(&exec, 0, 3); +// invert(&exec, 0, 3); + } + + /* Interpolate all the remaining attributes. This will get pushed + * into the fragment program's responsibilities at some point. + */ + for (i = 1; i < quad->nr_attrs; i++) { +#if 1 + for (j = 0; j < NUM_CHANNELS; j++) + linterp(&exec, i, j); +#else + switch (quad->interp[i]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + cinterp(&exec, i, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + linterp(&exec, i, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + pinterp(&exec, i, j); + break; + } +#endif + } + +#if 0 + generic->run_fs( tri->fp, quad, &tri->outputs ); +#else + { + GLuint attr = generic->fp_attr_to_slot[FRAG_ATTRIB_COL0]; + assert(attr); + + memcpy(quad->outputs.color, + exec.attr[attr], + sizeof(quad->outputs.color)); + } +#endif + + + if (quad->mask) + quad_output( generic, quad ); +} + + + + + + + diff --git a/src/mesa/pipe/softpipe/g_tile_output.c b/src/mesa/pipe/softpipe/g_tile_output.c new file mode 100644 index 0000000000..058ca30193 --- /dev/null +++ b/src/mesa/pipe/softpipe/g_tile_output.c @@ -0,0 +1,92 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "g_context.h" +#include "g_headers.h" +#include "g_surface.h" +#include "g_tile.h" + + +static void mask_copy( GLfloat (*dest)[4], + GLfloat (*src)[4], + GLuint mask ) +{ + GLuint i, j; + + for (i = 0; i < 4; i++) { + if (mask & (1<mask != MASK_ALL) + { + GLfloat tmp[4][QUAD_SIZE]; + + /* Yes, we'll probably have a masked write as well, but this is + * how blend will be done at least. + */ + gs_read_quad_f_swz( generic->cbuf_surface, + quad->x0, + quad->y0, + tmp ); + + mask_copy( tmp, quad->outputs.color, quad->mask ); + + gs_write_quad_f_swz( generic->cbuf_surface, + quad->x0, + quad->y0, + tmp ); + } + else + { + gs_write_quad_f_swz( generic->cbuf_surface, + quad->x0, + quad->y0, + quad->outputs.color ); + } +} diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c new file mode 100644 index 0000000000..8e547cbdc3 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -0,0 +1,135 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "tnl/t_context.h" +#include "vf/vf.h" + +#include "sp_context.h" +#include "sp_prim.h" +#include "sp_state.h" +#include "sp_draw.h" + +static void softpipe_destroy( struct pipe_context *pipe ) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + + draw_destroy( softpipe->draw ); + + FREE( softpipe ); +} + + +static void softpipe_draw_vb( struct pipe_context *pipe, + struct vertex_buffer *VB ) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + + if (softpipe->dirty) + softpipe_update_derived( softpipe ); + + draw_vb( softpipe->draw, VB ); +} + +struct pipe_context *softpipe_create( void ) +{ + struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); + + softpipe->pipe.destroy = softpipe_destroy; + softpipe->pipe.set_clip_state = softpipe_set_clip_state; + softpipe->pipe.set_viewport = softpipe_set_viewport; + softpipe->pipe.set_setup_state = softpipe_set_setup_state; + softpipe->pipe.set_scissor_rect = softpipe_set_scissor_rect; + softpipe->pipe.set_fs_state = softpipe_set_fs_state; + softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; + softpipe->pipe.set_cbuf_state = softpipe_set_cbuf_state; + softpipe->pipe.draw_vb = softpipe_draw_vb; + + + + softpipe->prim.setup = prim_setup( softpipe ); + softpipe->prim.unfilled = prim_unfilled( softpipe ); + softpipe->prim.twoside = prim_twoside( softpipe ); + softpipe->prim.offset = prim_offset( softpipe ); + softpipe->prim.clip = prim_clip( softpipe ); + softpipe->prim.flatshade = prim_flatshade( softpipe ); + softpipe->prim.cull = prim_cull( softpipe ); + + + softpipe->draw = draw_create( softpipe ); + + ASSIGN_4V( softpipe->plane[0], -1, 0, 0, 1 ); + ASSIGN_4V( softpipe->plane[1], 1, 0, 0, 1 ); + ASSIGN_4V( softpipe->plane[2], 0, -1, 0, 1 ); + ASSIGN_4V( softpipe->plane[3], 0, 1, 0, 1 ); + ASSIGN_4V( softpipe->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ + ASSIGN_4V( softpipe->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ + softpipe->nr_planes = 6; + + return &softpipe->pipe; +} + + + + + + +#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) + +void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ) +{ + stage->nr_tmps = nr; + + if (nr) { + GLubyte *store = MALLOC(MAX_VERTEX_SIZE * nr); + GLuint i; + + stage->tmp = MALLOC(sizeof(struct vertex_header *) * nr); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } +} + +void prim_free_tmps( struct prim_stage *stage ) +{ + if (stage->tmp) { + FREE(stage->tmp[0]); + FREE(stage->tmp); + } +} + + + + + diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h new file mode 100644 index 0000000000..da6c17aaf6 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_CONTEXT_H +#define G_CONTEXT_H + +#include "glheader.h" + +#include "pipe/p_state.h" +#include "pipe/p_context.h" + + + +struct softpipe_surface; +struct draw_context; +struct prim_stage; + + +enum interp_mode { + INTERP_CONSTANT, + INTERP_LINEAR, + INTERP_PERSPECTIVE +}; + + +#define G_NEW_VIEWPORT 0x1 +#define G_NEW_SETUP 0x2 +#define G_NEW_FS 0x4 +#define G_NEW_BLEND 0x8 +#define G_NEW_CBUF 0x10 +#define G_NEW_CLIP 0x20 +#define G_NEW_SCISSOR 0x40 +#define G_NEW_STIPPLE 0x80 + + + +struct softpipe_context { + struct pipe_context pipe; + + + /* The most recent drawing state as set by the driver: + */ + struct pipe_viewport viewport; + struct pipe_setup_state setup; + struct pipe_fs_state fs; + struct pipe_blend_state blend; + struct pipe_surface cbuf; + struct pipe_clip_state clip; + struct pipe_scissor_rect scissor; + struct pipe_poly_stipple poly_stipple; + GLuint dirty; + + + /* Cbuf derived state??? + */ + struct softpipe_surface *cbuf_surface; + + /* Clip derived state: + */ + GLfloat plane[12][4]; + GLuint nr_planes; + + /* Setup derived state. TODO: this should be passed in the program + * tokens as parameters to DECL instructions. + * + * For now we just set colors to CONST on flatshade, textures to + * perspective always and everything else to linear. + */ + enum interp_mode interp[VF_ATTRIB_MAX]; + + + /* FS + setup derived state: + */ + GLuint fp_attr_to_slot[VF_ATTRIB_MAX]; + GLuint vf_attr_to_slot[VF_ATTRIB_MAX]; + GLuint nr_attrs; + GLuint nr_frag_attrs; + GLuint attr_mask; + + GLboolean need_z; + GLboolean need_w; + + /* Stipple derived state: + */ + GLubyte stipple_masks[16][16]; + + + /* The software clipper/setup engine. + */ + struct { + struct prim_stage *setup; + struct prim_stage *unfilled; + struct prim_stage *twoside; + struct prim_stage *clip; + struct prim_stage *flatshade; + struct prim_stage *offset; + struct prim_stage *cull; + + struct prim_stage *first; + + GLenum prim; + GLuint vertex_size; + } prim; + + /* Temp kludge: + */ + struct draw_context *draw; +}; + + + + +static INLINE struct softpipe_context * +softpipe_context( struct pipe_context *pipe ) +{ + return (struct softpipe_context *)pipe; +} + + + + +#endif diff --git a/src/mesa/pipe/softpipe/sp_draw.c b/src/mesa/pipe/softpipe/sp_draw.c new file mode 100644 index 0000000000..5813e37776 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_draw.c @@ -0,0 +1,696 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "imports.h" + +#include "tnl/t_context.h" + +#include "sp_context.h" +#include "sp_prim.h" +#include "sp_headers.h" +#include "sp_draw.h" + +/* This file is a temporary set of hooks to allow us to use the tnl/ + * and vf/ modules until we have replacements in pipe. + */ + + +struct draw_context +{ + struct softpipe_context *softpipe; + + struct vf_attr_map attrs[VF_ATTRIB_MAX]; + GLuint nr_attrs; + GLuint vertex_size; + struct vertex_fetch *vf; + + GLubyte *verts; + GLuint nr_vertices; + GLboolean in_vb; + + GLenum prim; + + /* Helper for tnl: + */ + GLvector4f header; +}; + + +static struct vertex_header *get_vertex( struct draw_context *pipe, + GLuint i ) +{ + return (struct vertex_header *)(pipe->verts + i * pipe->vertex_size); +} + + + +static void draw_allocate_vertices( struct draw_context *draw, + GLuint nr_vertices ) +{ + draw->nr_vertices = nr_vertices; + draw->verts = MALLOC( nr_vertices * draw->vertex_size ); + + draw->softpipe->prim.first->begin( draw->softpipe->prim.first ); +} + +static void draw_set_prim( struct draw_context *draw, + GLenum prim ) +{ + draw->prim = prim; + + /* Not done yet - need to force edgeflags to 1 in strip/fan + * primitives. + */ +#if 0 + switch (prim) { + case GL_TRIANGLES: + case GL_POLYGON: + case GL_QUADS: + case GL_QUAD_STRIP: /* yes, we need this */ + respect_edgeflags( pipe, GL_TRUE ); + break; + + default: + respect_edgeflags( pipe, GL_FALSE ); + break; + } +#endif +} + + + +static void do_quad( struct prim_stage *first, + struct vertex_header *v0, + struct vertex_header *v1, + struct vertex_header *v2, + struct vertex_header *v3 ) +{ + struct prim_header prim; + + { + GLuint tmp = v1->edgeflag; + v1->edgeflag = 0; + + prim.v[0] = v0; + prim.v[1] = v1; + prim.v[2] = v3; + first->tri( first, &prim ); + + v1->edgeflag = tmp; + } + + { + GLuint tmp = v3->edgeflag; + v3->edgeflag = 0; + + prim.v[0] = v1; + prim.v[1] = v2; + prim.v[2] = v3; + first->tri( first, &prim ); + + v3->edgeflag = tmp; + } +} + + + + +static void draw_indexed_prim( struct draw_context *draw, + const GLuint *elts, + GLuint count ) +{ + struct prim_stage * const first = draw->softpipe->prim.first; + struct prim_header prim; + GLuint i; + + prim.det = 0; /* valid from cull stage onwards */ + prim.v[0] = 0; + prim.v[1] = 0; + prim.v[2] = 0; + + switch (draw->prim) { + case GL_POINTS: + for (i = 0; i < count; i ++) { + prim.v[0] = get_vertex( draw, elts[i] ); + + first->point( first, &prim ); + } + break; + + case GL_LINES: + for (i = 0; i+1 < count; i += 2) { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + + first->line( first, &prim ); + } + break; + + case GL_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + prim.v[0] = get_vertex( draw, elts[i-1] ); + prim.v[1] = get_vertex( draw, elts[i] ); + first->line( first, &prim ); + } + + prim.v[0] = get_vertex( draw, elts[count-1] ); + prim.v[1] = get_vertex( draw, elts[0] ); + first->line( first, &prim ); + } + break; + + case GL_LINE_STRIP: + /* I'm guessing it will be necessary to have something like a + * render->reset_line_stipple() method to properly support + * splitting strips into primitives like this. Alternately we + * could just scan ahead to find individual clipped lines and + * otherwise leave the strip intact - that might be better, but + * require more complex code here. + */ + if (count >= 2) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, elts[0] ); + + for (i = 1; i < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, elts[i] ); + + first->line( first, &prim ); + } + } + break; + + case GL_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + prim.v[0] = get_vertex( draw, elts[i + 1] ); + prim.v[1] = get_vertex( draw, elts[i + 0] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + } + else { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + } + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_FAN: + if (count >= 3) { + prim.v[0] = get_vertex( draw, elts[0] ); + prim.v[1] = 0; + prim.v[2] = get_vertex( draw, elts[1] ); + + for (i = 0; i+2 < count; i++) { + prim.v[1] = prim.v[2]; + prim.v[2] = get_vertex( draw, elts[i+2] ); + + first->tri( first, &prim ); + } + } + break; + + case GL_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( first, + get_vertex( draw, elts[i + 0] ), + get_vertex( draw, elts[i + 1] ), + get_vertex( draw, elts[i + 2] ), + get_vertex( draw, elts[i + 3] )); + } + break; + + case GL_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( first, + get_vertex( draw, elts[i + 2] ), + get_vertex( draw, elts[i + 0] ), + get_vertex( draw, elts[i + 1] ), + get_vertex( draw, elts[i + 3] )); + } + break; + + + case GL_POLYGON: + if (count >= 3) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, elts[1] ); + prim.v[2] = get_vertex( draw, elts[0] ); + + for (i = 0; i+2 < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, elts[i+2] ); + + first->tri( first, &prim ); + } + } + break; + + default: + assert(0); + break; + } +} + +static void draw_prim( struct draw_context *draw, + GLuint start, + GLuint count ) +{ + struct prim_stage * const first = draw->softpipe->prim.first; + struct prim_header prim; + GLuint i; + +// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); + + prim.det = 0; /* valid from cull stage onwards */ + prim.v[0] = 0; + prim.v[1] = 0; + prim.v[2] = 0; + + switch (draw->prim) { + case GL_POINTS: + for (i = 0; i < count; i ++) { + prim.v[0] = get_vertex( draw, start + i ); + first->point( first, &prim ); + } + break; + + case GL_LINES: + for (i = 0; i+1 < count; i += 2) { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + + first->line( first, &prim ); + } + break; + + case GL_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + prim.v[0] = get_vertex( draw, start + i - 1 ); + prim.v[1] = get_vertex( draw, start + i ); + first->line( first, &prim ); + } + + prim.v[0] = get_vertex( draw, start + count - 1 ); + prim.v[1] = get_vertex( draw, start + 0 ); + first->line( first, &prim ); + } + break; + + case GL_LINE_STRIP: + if (count >= 2) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, start + 0 ); + + for (i = 1; i < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, start + i ); + + first->line( first, &prim ); + } + } + break; + + case GL_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + prim.v[0] = get_vertex( draw, start + i + 1 ); + prim.v[1] = get_vertex( draw, start + i + 0 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + } + else { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + } + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_FAN: + if (count >= 3) { + prim.v[0] = get_vertex( draw, start + 0 ); + prim.v[1] = 0; + prim.v[2] = get_vertex( draw, start + 1 ); + + for (i = 0; i+2 < count; i++) { + prim.v[1] = prim.v[2]; + prim.v[2] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + } + break; + + + case GL_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( first, + get_vertex( draw, start + i + 0 ), + get_vertex( draw, start + i + 1 ), + get_vertex( draw, start + i + 2 ), + get_vertex( draw, start + i + 3 )); + } + break; + + case GL_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( first, + get_vertex( draw, start + i + 2 ), + get_vertex( draw, start + i + 0 ), + get_vertex( draw, start + i + 1 ), + get_vertex( draw, start + i + 3 )); + } + break; + + case GL_POLYGON: + if (count >= 3) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, start + 1 ); + prim.v[2] = get_vertex( draw, start + 0 ); + + for (i = 0; i+2 < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + } + break; + + default: + assert(0); + break; + } +} + + +static void draw_release_vertices( struct draw_context *draw ) +{ + draw->softpipe->prim.first->end( draw->softpipe->prim.first ); + + FREE(draw->verts); + draw->verts = NULL; +} + + +struct header_dword { + GLuint clipmask:12; + GLuint edgeflag:1; + GLuint pad:19; +}; + + +static void +build_vertex_headers( struct draw_context *draw, + struct vertex_buffer *VB ) +{ + if (draw->header.storage == NULL) { + draw->header.stride = sizeof(GLfloat); + draw->header.size = 1; + draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(GLfloat), 32 ); + draw->header.data = draw->header.storage; + draw->header.count = 0; + draw->header.flags = VEC_SIZE_1 | VEC_MALLOC; + } + + /* Build vertex header attribute. + * + */ + + { + GLuint i; + struct header_dword *header = (struct header_dword *)draw->header.storage; + + /* yes its a hack + */ + assert(sizeof(*header) == sizeof(GLfloat)); + + draw->header.count = VB->Count; + + if (VB->EdgeFlag) { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = VB->ClipMask[i]; + header[i].edgeflag = VB->EdgeFlag[i]; + header[i].pad = 0; + } + } + else if (VB->ClipOrMask) { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = VB->ClipMask[i]; + header[i].edgeflag = 0; + header[i].pad = 0; + } + } + else { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = 0; + header[i].edgeflag = 0; + header[i].pad = 0; + } + } + } + + VB->AttribPtr[VF_ATTRIB_VERTEX_HEADER] = &draw->header; +} + + + + + +static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) +{ + switch (mode) { + case GL_POINTS: + *first = 1; + *incr = 1; + return 0; + case GL_LINES: + *first = 2; + *incr = 2; + return 0; + case GL_LINE_STRIP: + *first = 2; + *incr = 1; + return 0; + case GL_LINE_LOOP: + *first = 2; + *incr = 1; + return 1; + case GL_TRIANGLES: + *first = 3; + *incr = 3; + return 0; + case GL_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + return 0; + case GL_TRIANGLE_FAN: + case GL_POLYGON: + *first = 3; + *incr = 1; + return 1; + case GL_QUADS: + *first = 4; + *incr = 4; + return 0; + case GL_QUAD_STRIP: + *first = 4; + *incr = 2; + return 0; + default: + assert(0); + *first = 1; + *incr = 1; + return 0; + } +} + + +static GLuint trim( GLuint count, GLuint first, GLuint incr ) +{ + if (count < first) + return 0; + else + return count - (count - first) % incr; +} + + +/* This is a hack & will all go away. + */ +void draw_vb(struct draw_context *draw, + struct vertex_buffer *VB ) +{ + GLuint i; + + VB->AttribPtr[VF_ATTRIB_POS] = VB->NdcPtr; + VB->AttribPtr[VF_ATTRIB_BFC0] = VB->ColorPtr[1]; + VB->AttribPtr[VF_ATTRIB_BFC1] = VB->SecondaryColorPtr[1]; + VB->AttribPtr[VF_ATTRIB_CLIP_POS] = VB->ClipPtr; + + /* Build vertex headers: + */ + build_vertex_headers( draw, VB ); + + draw->in_vb = 1; + + /* Allocate the vertices: + */ + draw_allocate_vertices( draw, VB->Count ); + + /* Bind the vb outputs: + */ + vf_set_sources( draw->vf, VB->AttribPtr, 0 ); + + /* Build the hardware or prim-pipe vertices: + */ + vf_emit_vertices( draw->vf, VB->Count, draw->verts ); + + + for (i = 0; i < VB->PrimitiveCount; i++) { + + GLenum mode = VB->Primitive[i].mode; + GLuint start = VB->Primitive[i].start; + GLuint length, first, incr; + + /* Trim the primitive down to a legal size. + */ + draw_prim_info( mode, &first, &incr ); + length = trim( VB->Primitive[i].count, first, incr ); + + if (!length) + continue; + + if (draw->prim != mode) + draw_set_prim( draw, mode ); + + if (VB->Elts) { + draw_indexed_prim( draw, + VB->Elts + start, + length ); + } + else { + draw_prim( draw, + start, + length ); + } + } + + draw_release_vertices( draw ); + draw->verts = NULL; + draw->in_vb = 0; +} + +void draw_set_viewport( struct draw_context *draw, + const GLfloat *scale, + const GLfloat *translate ) +{ + assert(!draw->in_vb); + vf_set_vp_scale_translate( draw->vf, scale, translate ); +} + + + +struct draw_context *draw_create( struct softpipe_context *softpipe ) +{ + struct draw_context *draw = CALLOC_STRUCT( draw_context ); + draw->softpipe = softpipe; + draw->vf = vf_create( GL_TRUE ); + + return draw; +} + + +void draw_destroy( struct draw_context *draw ) +{ + if (draw->header.storage) + ALIGN_FREE( draw->header.storage ); + + vf_destroy( draw->vf ); + + FREE( draw ); +} + +#define EMIT_ATTR( ATTR, STYLE ) \ +do { \ + draw->attrs[draw->nr_attrs].attrib = ATTR; \ + draw->attrs[draw->nr_attrs].format = STYLE; \ + draw->nr_attrs++; \ +} while (0) + + +void draw_set_vertex_attributes( struct draw_context *draw, + const GLuint *attrs, + GLuint nr_attrs ) +{ + GLuint i; + + draw->nr_attrs = 0; + + EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F); + EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F); + + assert(attrs[0] == VF_ATTRIB_POS); + EMIT_ATTR(attrs[0], EMIT_4F_VIEWPORT); + + for (i = 1; i < nr_attrs; i++) + EMIT_ATTR(attrs[i], EMIT_4F); + + draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); +} + + diff --git a/src/mesa/pipe/softpipe/sp_draw.h b/src/mesa/pipe/softpipe/sp_draw.h new file mode 100644 index 0000000000..a138f812fa --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_draw.h @@ -0,0 +1,56 @@ + +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_DRAW_H +#define G_DRAW_H + +#include "glheader.h" +#include "pipe/p_state.h" + + +struct draw_context; + +struct draw_context *draw_create( struct softpipe_context *softpipe ); + +void draw_destroy( struct draw_context *draw ); + +void draw_set_viewport( struct draw_context *draw, + const GLfloat *scale, + const GLfloat *translate ); + +void draw_set_vertex_attributes( struct draw_context *draw, + const GLuint *attrs, + GLuint nr_attrs ); + +void draw_vb(struct draw_context *draw, + struct vertex_buffer *VB ); + +#endif diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h new file mode 100644 index 0000000000..96ff52a453 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -0,0 +1,116 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#ifndef G_HEADERS_H +#define G_HEADERS_H + +#define PRIM_POINT 1 +#define PRIM_LINE 2 +#define PRIM_TRI 3 + +struct prim_header { + GLfloat det; + struct vertex_header *v[3]; +}; + +/* Carry some useful information around with the vertices in the prim + * pipe. + */ +struct vertex_header { + GLuint clipmask:12; + GLuint edgeflag:1; + GLuint pad:19; + + GLfloat clip[4]; + + GLfloat data[][4]; /* Note variable size */ +}; + + + + + +/* The rasterizer generates 2x2 quads of fragment and feeds them to + * the current fp_machine (see below). + */ +#define QUAD_BOTTOM_LEFT 0 +#define QUAD_BOTTOM_RIGHT 1 +#define QUAD_TOP_LEFT 2 +#define QUAD_TOP_RIGHT 3 +#define QUAD_SIZE (2*2) + +#define MASK_BOTTOM_LEFT 0x1 +#define MASK_BOTTOM_RIGHT 0x2 +#define MASK_TOP_LEFT 0x4 +#define MASK_TOP_RIGHT 0x8 +#define MASK_ALL 0xf + + +#define NUM_CHANNELS 4 /* avoid confusion between 4 pixels and 4 channels */ + + +struct setup_coefficient { + GLfloat a0[NUM_CHANNELS]; /* in an xyzw layout */ + GLfloat dadx[NUM_CHANNELS]; + GLfloat dady[NUM_CHANNELS]; +}; + + + +/* Encodes everything we need to know about a 2x2 pixel block. Uses + * "Channel-Serial" or "SoA" layout. + * + * Will expand to include non-attribute things like AA coverage and + * maybe prefetched depth from the depth buffer. + */ +struct quad_header { + GLint x0; + GLint y0; + GLuint mask; + GLuint facing; /**< Front or back facing? */ + + struct { + GLfloat color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ + GLfloat depth[QUAD_SIZE]; + } outputs; + + const struct setup_coefficient *coef; + + const enum interp_mode *interp; /* XXX: this information should be + * encoded in fragment program DECL + * statements. */ + + GLuint nr_attrs; +}; + + + + + +#endif diff --git a/src/mesa/pipe/softpipe/sp_prim.h b/src/mesa/pipe/softpipe/sp_prim.h new file mode 100644 index 0000000000..b6cbaae085 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim.h @@ -0,0 +1,90 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_PRIM_H +#define G_PRIM_H + +#include "glheader.h" +#include "sp_headers.h" + +struct softpipe_context; + +struct prim_stage *prim_setup( struct softpipe_context *context ); +struct prim_stage *prim_unfilled( struct softpipe_context *context ); +struct prim_stage *prim_twoside( struct softpipe_context *context ); +struct prim_stage *prim_offset( struct softpipe_context *context ); +struct prim_stage *prim_clip( struct softpipe_context *context ); +struct prim_stage *prim_flatshade( struct softpipe_context *context ); +struct prim_stage *prim_cull( struct softpipe_context *context ); + + +/* Internal structs and helpers for the primitive clip/setup pipeline: + */ +struct prim_stage { + struct softpipe_context *softpipe; + + struct prim_stage *next; + + struct vertex_header **tmp; + GLuint nr_tmps; + + void (*begin)( struct prim_stage * ); + + void (*point)( struct prim_stage *, + struct prim_header * ); + + void (*line)( struct prim_stage *, + struct prim_header * ); + + void (*tri)( struct prim_stage *, + struct prim_header * ); + + void (*end)( struct prim_stage * ); +}; + + + +/* Get a writeable copy of a vertex: + */ +static INLINE struct vertex_header * +dup_vert( struct prim_stage *stage, + const struct vertex_header *vert, + GLuint idx ) +{ + struct vertex_header *tmp = stage->tmp[idx]; + memcpy(tmp, vert, stage->softpipe->prim.vertex_size ); + return tmp; +} + +void prim_free_tmps( struct prim_stage *stage ); +void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); + + +#endif diff --git a/src/mesa/pipe/softpipe/sp_prim_clip.c b/src/mesa/pipe/softpipe/sp_prim_clip.c new file mode 100644 index 0000000000..304c43c3f4 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_clip.c @@ -0,0 +1,429 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "sp_context.h" +#include "sp_prim.h" + +struct clipper { + struct prim_stage stage; + + GLuint active_user_planes; +}; + +/* This is a bit confusing: + */ +static INLINE struct clipper *clipper_stage( struct prim_stage *stage ) +{ + return (struct clipper *)stage; +} + + +#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) + + +/* All attributes are float[4], so this is easy: + */ +static void interp_attr( GLfloat *fdst, + GLfloat t, + const GLfloat *fin, + const GLfloat *fout ) +{ + fdst[0] = LINTERP( t, fout[0], fin[0] ); + fdst[1] = LINTERP( t, fout[1], fin[1] ); + fdst[2] = LINTERP( t, fout[2], fin[2] ); + fdst[3] = LINTERP( t, fout[3], fin[3] ); +} + + + + +/* Interpolate between two vertices to produce a third. + */ +static void interp( struct clipper *clip, + struct vertex_header *dst, + GLfloat t, + const struct vertex_header *out, + const struct vertex_header *in ) +{ + const GLuint nr_attrs = clip->stage.softpipe->nr_attrs; + GLuint j; + + /* Vertex header. + */ + { + dst->clipmask = 0; + dst->edgeflag = 0; + dst->pad = 0; + } + + /* Clip coordinates: interpolate normally + */ + { + interp_attr(dst->clip, t, in->clip, out->clip); + } + + /* Do the projective divide and insert window coordinates: + */ + { + const GLfloat *pos = dst->clip; + const GLfloat *scale = clip->stage.softpipe->viewport.scale; + const GLfloat *trans = clip->stage.softpipe->viewport.translate; + GLfloat oow; + + oow = 1.0 / pos[3]; + + dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; + dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; + dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; + dst->data[0][3] = oow; + } + + + /* Other attributes + */ + for (j = 1; j < nr_attrs-1; j++) { + interp_attr(dst->data[j], t, in->data[j], out->data[j]); + } +} + + +#define CLIP_USER_BIT 0x40 +#define CLIP_CULL_BIT 0x80 + + +static INLINE GLfloat dot4( const GLfloat *a, + const GLfloat *b ) +{ + GLfloat result = (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); + + return result; +} + + +#if 0 +static INLINE void do_tri( struct prim_stage *next, + struct prim_header *header ) +{ + GLuint i; + for (i = 0; i < 3; i++) { + GLfloat *ndc = header->v[i]->data[0]; + _mesa_printf("ndc %f %f %f\n", ndc[0], ndc[1], ndc[2]); + assert(ndc[0] >= -1 && ndc[0] <= 641); + assert(ndc[1] >= 30 && ndc[1] <= 481); + } + _mesa_printf("\n"); + next->tri(next, header); +} +#endif + + +static void emit_poly( struct prim_stage *stage, + struct vertex_header **inlist, + GLuint n ) +{ + struct prim_header header; + GLuint i; + + for (i = 2; i < n; i++) { + header.v[0] = inlist[0]; + header.v[1] = inlist[i-1]; + header.v[2] = inlist[i]; + + { + GLuint tmp0 = header.v[0]->edgeflag; + GLuint tmp2 = header.v[2]->edgeflag; + + if (i != 2) header.v[0]->edgeflag = 0; + if (i != n-1) header.v[2]->edgeflag = 0; + + stage->next->tri( stage->next, &header ); + + header.v[0]->edgeflag = tmp0; + header.v[2]->edgeflag = tmp2; + } + } +} + + +#if 0 +static void emit_poly( struct prim_stage *stage ) +{ + GLuint i; + + for (i = 2; i < n; i++) { + header->v[0] = inlist[0]; + header->v[1] = inlist[i-1]; + header->v[2] = inlist[i]; + + stage->next->tri( stage->next, header ); + } +} +#endif + + +/* Clip a triangle against the viewport and user clip planes. + */ +static void +do_clip_tri( struct prim_stage *stage, + struct prim_header *header, + GLuint clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *a[MAX_CLIPPED_VERTICES]; + struct vertex_header *b[MAX_CLIPPED_VERTICES]; + struct vertex_header **inlist = a; + struct vertex_header **outlist = b; + GLuint tmpnr = 0; + GLuint n = 3; + GLuint i; + + inlist[0] = header->v[0]; + inlist[1] = header->v[1]; + inlist[2] = header->v[2]; + + /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove + * this once we correctly use 16bit masks for userclip planes. + */ + clipmask &= ~CLIP_CULL_BIT; + if (clipmask & CLIP_USER_BIT) { + clipmask &= ~CLIP_USER_BIT; + clipmask |= clipper->active_user_planes; + } + + while (clipmask && n >= 3) { + GLuint plane_idx = ffs(clipmask)-1; + const GLfloat *plane = clipper->stage.softpipe->plane[plane_idx]; + struct vertex_header *vert_prev = inlist[0]; + GLfloat dp_prev = dot4( vert_prev->clip, plane ); + GLuint outcount = 0; + + clipmask &= ~(1<clip, plane ); + + if (!IS_NEGATIVE(dp_prev)) { + outlist[outcount++] = vert_prev; + } + + if (DIFFERENT_SIGNS(dp, dp_prev)) { + struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; + outlist[outcount++] = new_vert; + + if (IS_NEGATIVE(dp)) { + /* Going out of bounds. Avoid division by zero as we + * know dp != dp_prev from DIFFERENT_SIGNS, above. + */ + GLfloat t = dp / (dp - dp_prev); + interp( clipper, new_vert, t, vert, vert_prev ); + + /* Force edgeflag true in this case: + */ + new_vert->edgeflag = 1; + } else { + /* Coming back in. + */ + GLfloat t = dp_prev / (dp_prev - dp); + interp( clipper, new_vert, t, vert_prev, vert ); + + /* Copy starting vert's edgeflag: + */ + new_vert->edgeflag = vert_prev->edgeflag; + } + } + + vert_prev = vert; + dp_prev = dp; + } + + { + struct vertex_header **tmp = inlist; + inlist = outlist; + outlist = tmp; + n = outcount; + } + } + + /* Emit the polygon as triangles to the setup stage: + */ + if (n >= 3) + emit_poly( stage, inlist, n ); +} + + +/* Clip a line against the viewport and user clip planes. + */ +static void +do_clip_line( struct prim_stage *stage, + struct prim_header *header, + GLuint clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const GLfloat *pos0 = v0->clip; + const GLfloat *pos1 = v1->clip; + GLfloat t0 = 0; + GLfloat t1 = 0; + + /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove + * this once we correctly use 16bit masks for userclip planes. + */ + clipmask &= ~CLIP_CULL_BIT; + if (clipmask & CLIP_USER_BIT) { + clipmask &= ~CLIP_USER_BIT; + clipmask |= clipper->active_user_planes; + } + + while (clipmask) { + GLuint plane_idx = ffs(clipmask)-1; + const GLfloat *plane = clipper->stage.softpipe->plane[plane_idx]; + + clipmask &= ~(1< t1) t1 = t; + } + + if (dp0 < 0) { + GLfloat t = dp0 / (dp0 - dp1); + if (t > t0) t0 = t; + } + + if (t0 + t1 >= 1.0) + return; /* discard */ + } + + if (v0->clipmask) { + interp( clipper, stage->tmp[0], t0, v0, v1 ); + header->v[0] = stage->tmp[0]; + } + + if (v1->clipmask) { + interp( clipper, stage->tmp[1], t1, v1, v0 ); + header->v[1] = stage->tmp[1]; + } + + stage->next->line( stage->next, header ); +} + + + +static void clip_begin( struct prim_stage *stage ) +{ + struct clipper *clipper = clipper_stage(stage); + GLuint nr = stage->softpipe->nr_planes; + + /* Hacky bitmask to use when we hit CLIP_USER_BIT: + */ + clipper->active_user_planes = ((1<next->begin( stage->next ); +} + +static void +clip_point( struct prim_stage *stage, + struct prim_header *header ) +{ + if (header->v[0]->clipmask == 0) + stage->next->point( stage->next, header ); +} + + +static void +clip_line( struct prim_stage *stage, + struct prim_header *header ) +{ + GLuint clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask); + + if (clipmask == 0) { + stage->next->line( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask) == 0) { + do_clip_line(stage, header, clipmask); + } +} + + +static void +clip_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + GLuint clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask | + header->v[2]->clipmask); + + if (clipmask == 0) { + stage->next->tri( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask & + header->v[2]->clipmask) == 0) { + do_clip_tri(stage, header, clipmask); + } +} + +static void clip_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + + +struct prim_stage *prim_clip( struct softpipe_context *softpipe ) +{ + struct clipper *clipper = CALLOC_STRUCT(clipper); + + prim_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); + + clipper->stage.softpipe = softpipe; + clipper->stage.begin = clip_begin; + clipper->stage.point = clip_point; + clipper->stage.line = clip_line; + clipper->stage.tri = clip_tri; + clipper->stage.end = clip_end; + + return &clipper->stage; +} diff --git a/src/mesa/pipe/softpipe/sp_prim_cull.c b/src/mesa/pipe/softpipe/sp_prim_cull.c new file mode 100644 index 0000000000..16920d478a --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_cull.c @@ -0,0 +1,119 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_prim.h" + + + +struct cull_stage { + struct prim_stage stage; + GLuint mode; +}; + + + +static INLINE struct cull_stage *cull_stage( struct prim_stage *stage ) +{ + return (struct cull_stage *)stage; +} + + +static void cull_begin( struct prim_stage *stage ) +{ + struct cull_stage *cull = cull_stage(stage); + + cull->mode = stage->softpipe->setup.cull_mode; + + stage->next->begin( stage->next ); +} + + +static void cull_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + /* Window coords: */ + GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); + GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); + GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); + + GLfloat ex = v0[0] - v2[0]; + GLfloat ey = v0[1] - v2[1]; + GLfloat fx = v1[0] - v2[0]; + GLfloat fy = v1[1] - v2[1]; + + header->det = ex * fy - ey * fx; + + _mesa_printf("%s %f\n", __FUNCTION__, header->det ); + + if (header->det != 0) { + GLuint mode = (header->det < 0) ? WINDING_CW : WINDING_CCW; + + if ((mode & cull_stage(stage)->mode) == 0) + stage->next->tri( stage->next, header ); + } +} + + +static void cull_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void cull_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void cull_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_cull( struct softpipe_context *softpipe ) +{ + struct cull_stage *cull = CALLOC_STRUCT(cull_stage); + + prim_alloc_tmps( &cull->stage, 0 ); + + cull->stage.softpipe = softpipe; + cull->stage.next = NULL; + cull->stage.begin = cull_begin; + cull->stage.point = cull_point; + cull->stage.line = cull_line; + cull->stage.tri = cull_tri; + cull->stage.end = cull_end; + + return &cull->stage; +} diff --git a/src/mesa/pipe/softpipe/sp_prim_flatshade.c b/src/mesa/pipe/softpipe/sp_prim_flatshade.c new file mode 100644 index 0000000000..0e9de5faeb --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_flatshade.c @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_prim.h" + + +struct flatshade_stage { + struct prim_stage stage; + + const GLuint *lookup; +}; + + + +static INLINE struct flatshade_stage *flatshade_stage( struct prim_stage *stage ) +{ + return (struct flatshade_stage *)stage; +} + + +static void flatshade_begin( struct prim_stage *stage ) +{ + stage->next->begin( stage->next ); +} + + + +static INLINE void copy_attr( GLuint attr, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + if (attr) { + memcpy( dst->data[attr], + src->data[attr], + sizeof(src->data[0]) ); + } +} + +static void copy_colors( struct prim_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + struct flatshade_stage *flatshade = flatshade_stage(stage); + const GLuint *lookup = flatshade->lookup; + + copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); + copy_attr( lookup[VF_ATTRIB_COLOR1], dst, src ); + copy_attr( lookup[VF_ATTRIB_BFC0], dst, src ); + copy_attr( lookup[VF_ATTRIB_BFC1], dst, src ); +} + + + +/* Flatshade tri. Required for clipping and when unfilled tris are + * active, otherwise handled by hardware. + */ +static void flatshade_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = header->v[2]; + + copy_colors(stage, tmp.v[0], tmp.v[2]); + copy_colors(stage, tmp.v[1], tmp.v[2]); + + stage->next->tri( stage->next, &tmp ); +} + + +/* Flatshade line. Required for clipping. + */ +static void flatshade_line( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = header->v[1]; + + copy_colors(stage, tmp.v[0], tmp.v[1]); + + stage->next->line( stage->next, &tmp ); +} + + +static void flatshade_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void flatshade_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_flatshade( struct softpipe_context *softpipe ) +{ + struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); + + prim_alloc_tmps( &flatshade->stage, 2 ); + + flatshade->stage.softpipe = softpipe; + flatshade->stage.next = NULL; + flatshade->stage.begin = flatshade_begin; + flatshade->stage.point = flatshade_point; + flatshade->stage.line = flatshade_line; + flatshade->stage.tri = flatshade_tri; + flatshade->stage.end = flatshade_end; + + flatshade->lookup = softpipe->vf_attr_to_slot; + + return &flatshade->stage; +} + + diff --git a/src/mesa/pipe/softpipe/sp_prim_offset.c b/src/mesa/pipe/softpipe/sp_prim_offset.c new file mode 100644 index 0000000000..5fd6ac911a --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_offset.c @@ -0,0 +1,157 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" +#include "macros.h" + +#include "sp_context.h" +#include "sp_prim.h" + + + +struct offset_stage { + struct prim_stage stage; + + GLuint hw_data_offset; + + GLfloat scale; + GLfloat units; +}; + + + +static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) +{ + return (struct offset_stage *)stage; +} + + +static void offset_begin( struct prim_stage *stage ) +{ + struct offset_stage *offset = offset_stage(stage); + + offset->units = stage->softpipe->setup.offset_units; + offset->scale = stage->softpipe->setup.offset_scale; + + stage->next->begin( stage->next ); +} + + +/* Offset tri. Some hardware can handle this, but not usually when + * doing unfilled rendering. + */ +static void do_offset_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + GLfloat inv_det = 1.0 / header->det; + + /* Window coords: + */ + GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); + GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); + GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); + + GLfloat ex = v0[0] - v2[2]; + GLfloat fx = v1[0] - v2[2]; + GLfloat ey = v0[1] - v2[2]; + GLfloat fy = v1[1] - v2[2]; + GLfloat ez = v0[2] - v2[2]; + GLfloat fz = v1[2] - v2[2]; + + GLfloat a = ey*fz - ez*fy; + GLfloat b = ez*fx - ex*fz; + + GLfloat ac = a * inv_det; + GLfloat bc = b * inv_det; + GLfloat zoffset; + + if ( ac < 0.0f ) ac = -ac; + if ( bc < 0.0f ) bc = -bc; + + zoffset = offset->units + MAX2( ac, bc ) * offset->scale; + + v0[2] += zoffset; + v1[2] += zoffset; + v2[2] += zoffset; + + stage->next->tri( stage->next, header ); +} + + +static void offset_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = dup_vert(stage, header->v[2], 2); + + do_offset_tri( stage->next, &tmp ); +} + + + +static void offset_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void offset_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void offset_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_offset( struct softpipe_context *softpipe ) +{ + struct offset_stage *offset = CALLOC_STRUCT(offset_stage); + + prim_alloc_tmps( &offset->stage, 3 ); + + offset->stage.softpipe = softpipe; + offset->stage.next = NULL; + offset->stage.begin = offset_begin; + offset->stage.point = offset_point; + offset->stage.line = offset_line; + offset->stage.tri = offset_tri; + offset->stage.end = offset_end; + + return &offset->stage; +} diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c new file mode 100644 index 0000000000..a1addc4ee1 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -0,0 +1,888 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "sp_context.h" +#include "sp_prim.h" +#include "sp_tile.h" + + +/** + * Triangle edge info + */ +struct edge { + GLfloat dx; /* X(v1) - X(v0), used only during setup */ + GLfloat dy; /* Y(v1) - Y(v0), used only during setup */ + GLfloat dxdy; /* dx/dy */ + GLfloat sx; /* first sample point x coord */ + GLfloat sy; + GLint lines; /* number of lines on this edge */ +}; + + +/** + * Triangle setup info (derived from prim_stage). + * Also used for line drawing (taking some liberties). + */ +struct setup_stage { + struct prim_stage stage; /**< This must be first */ + + /* Vertices are just an array of floats making up each attribute in + * turn. Currently fixed at 4 floats, but should change in time. + * Codegen will help cope with this. + */ + const struct vertex_header *vmax; + const struct vertex_header *vmid; + const struct vertex_header *vmin; + const struct vertex_header *vprovoke; + + struct edge ebot; + struct edge etop; + struct edge emaj; + + GLfloat oneoverarea; + + struct setup_coefficient coef[FRAG_ATTRIB_MAX]; + struct quad_header quad; + + struct { + GLint left[2]; /**< [0] = row0, [1] = row1 */ + GLint right[2]; + GLint y; + GLuint y_flags; + GLuint mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ + } span; +}; + + + +/** + * Basically a cast wrapper. + */ +static inline struct setup_stage *setup_stage( struct prim_stage *stage ) +{ + return (struct setup_stage *)stage; +} + + +/** + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. + */ +static inline GLint block( GLint x ) +{ + return x & ~1; +} + + + +static void setup_begin( struct prim_stage *stage ) +{ + setup_stage(stage)->quad.nr_attrs = stage->softpipe->nr_frag_attrs; +} + + +/** + * Run shader on a quad/block. + */ +static void run_shader_block( struct setup_stage *setup, + GLint x, GLint y, GLuint mask ) +{ + setup->quad.x0 = x; + setup->quad.y0 = y; + setup->quad.mask = mask; + + quad_shade( setup->stage.softpipe, &setup->quad ); +} + + +/** + * Compute mask which indicates which pixels in the 2x2 quad are actually inside + * the triangle's bounds. + * + * this is pretty nasty... may need to rework flush_spans again to + * fix it, if possible. + */ +static GLuint calculate_mask( struct setup_stage *setup, + GLint x ) +{ + GLuint mask = 0; + + if (x >= setup->span.left[0] && x < setup->span.right[0]) + mask |= MASK_BOTTOM_LEFT; + + if (x >= setup->span.left[1] && x < setup->span.right[1]) + mask |= MASK_TOP_LEFT; + + if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) + mask |= MASK_BOTTOM_RIGHT; + + if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) + mask |= MASK_TOP_RIGHT; + + return mask; +} + + +/** + * Render a horizontal span of quads + */ +static void flush_spans( struct setup_stage *setup ) +{ + GLint minleft, maxright; + GLint x; + + switch (setup->span.y_flags) { + case 3: + minleft = MIN2(setup->span.left[0], setup->span.left[1]); + maxright = MAX2(setup->span.right[0], setup->span.right[1]); + break; + + case 1: + minleft = setup->span.left[0]; + maxright = setup->span.right[0]; + break; + + case 2: + minleft = setup->span.left[1]; + maxright = setup->span.right[1]; + break; + + default: + return; + } + + + for (x = block(minleft); x <= block(maxright); ) + { + run_shader_block( setup, x, + setup->span.y, + calculate_mask( setup, x ) ); + x += 2; + } + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +} + + +static GLboolean setup_sort_vertices( struct setup_stage *setup, + const struct prim_header *prim ) +{ + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + const struct vertex_header *v2 = prim->v[2]; + + setup->vprovoke = v2; + + /* determine bottom to top order of vertices */ + { + GLfloat y0 = v0->data[0][1]; + GLfloat y1 = v1->data[0][1]; + GLfloat y2 = v2->data[0][1]; + if (y0 <= y1) { + if (y1 <= y2) { + /* y0<=y1<=y2 */ + setup->vmin = v0; + setup->vmid = v1; + setup->vmax = v2; + } + else if (y2 <= y0) { + /* y2<=y0<=y1 */ + setup->vmin = v2; + setup->vmid = v0; + setup->vmax = v1; + } + else { + /* y0<=y2<=y1 */ + setup->vmin = v0; + setup->vmid = v2; + setup->vmax = v1; + } + } + else { + if (y0 <= y2) { + /* y1<=y0<=y2 */ + setup->vmin = v1; + setup->vmid = v0; + setup->vmax = v2; + } + else if (y2 <= y1) { + /* y2<=y1<=y0 */ + setup->vmin = v2; + setup->vmid = v1; + setup->vmax = v0; + } + else { + /* y1<=y2<=y0 */ + setup->vmin = v1; + setup->vmid = v2; + setup->vmax = v0; + } + } + } + + setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; + setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; + setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; + + /* xxx: may need to adjust this sign according to the if-tree + * above: + * + * XXX: this is like 'det', but calculated from screen coords?? + */ + { + const GLfloat area = (setup->emaj.dx * setup->ebot.dy - + setup->ebot.dx * setup->emaj.dy); + + setup->oneoverarea = 1.0 / area; + } + + /* XXX need to know if this is a front or back-facing triangle: + * - the GLSL gl_FrontFacing fragment attribute (bool) + * - two-sided stencil test + */ + setup->quad.facing = 0; + + _mesa_printf("%s one-over-area %f\n", __FUNCTION__, setup->oneoverarea ); + + + return GL_TRUE; +} + + +/** + * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + */ +static void const_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i ) +{ + setup->coef[slot].dadx[i] = 0; + setup->coef[slot].dady[i] = 0; + + /* need provoking vertex info! + */ + setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; +} + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a triangle. + */ +static void tri_linear_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i) +{ + GLfloat botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; + GLfloat majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; + GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); + + _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", + slot, "xyzw"[i], + setup->coef[slot].a0[i], + setup->coef[slot].dadx[i], + setup->coef[slot].dady[i]); +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a triangle. + */ +static void tri_persp_coeff( struct setup_stage *setup, + GLuint slot, + GLuint i ) +{ + /* premultiply by 1/w: + */ + GLfloat mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; + GLfloat mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; + GLfloat maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; + + GLfloat botda = mida - mina; + GLfloat majda = maxa - mina; + GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; + GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + setup->coef[slot].a0[i] = (mina - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); +} + + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. + */ +static void setup_tri_coefficients( struct setup_stage *setup ) +{ + const enum interp_mode *interp = setup->stage.softpipe->interp; + GLuint slot, j; + + /* z and w are done by linear interpolation: + */ + tri_linear_coeff(setup, 0, 2); + tri_linear_coeff(setup, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + switch (interp[slot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + tri_linear_coeff(setup, slot, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + tri_persp_coeff(setup, slot, j); + break; + } + } +} + + + +static void setup_tri_edges( struct setup_stage *setup ) +{ + GLfloat vmin_x = setup->vmin->data[0][0] + 0.5; + GLfloat vmid_x = setup->vmid->data[0][0] + 0.5; + + GLfloat vmin_y = setup->vmin->data[0][1] - 0.5; + GLfloat vmid_y = setup->vmid->data[0][1] - 0.5; + GLfloat vmax_y = setup->vmax->data[0][1] - 0.5; + + setup->emaj.sy = ceilf(vmin_y); + setup->emaj.lines = (GLint) ceilf(vmax_y - setup->emaj.sy); + setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; + setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; + + setup->etop.sy = ceilf(vmid_y); + setup->etop.lines = (GLint) ceilf(vmax_y - setup->etop.sy); + setup->etop.dxdy = setup->etop.dx / setup->etop.dy; + setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; + + setup->ebot.sy = ceilf(vmin_y); + setup->ebot.lines = (GLint) ceilf(vmid_y - setup->ebot.sy); + setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; + setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; +} + + +/** + * Render the upper or lower half of a triangle. + * Scissoring is applied here too. + */ +static void subtriangle( struct setup_stage *setup, + struct edge *eleft, + struct edge *eright, + GLuint lines ) +{ + GLint y, start_y, finish_y; + GLint sy = (GLint)eleft->sy; + + assert((GLint)eleft->sy == (GLint) eright->sy); + assert((GLint)eleft->sy >= 0); /* catch bug in x64? */ + + /* scissor y: + */ + if (setup->stage.softpipe->setup.scissor) { + start_y = sy; + finish_y = start_y + lines; + + if (start_y < setup->stage.softpipe->scissor.miny) + start_y = setup->stage.softpipe->scissor.miny; + + if (finish_y > setup->stage.softpipe->scissor.maxy) + finish_y = setup->stage.softpipe->scissor.maxy; + + start_y -= sy; + finish_y -= sy; + } + else { + start_y = 0; + finish_y = lines; + } + + _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + + for (y = start_y; y < finish_y; y++) { + + /* avoid accumulating adds as floats don't have the precision to + * accurately iterate large triangle edges that way. luckily we + * can just multiply these days. + * + * this is all drowned out by the attribute interpolation anyway. + */ + GLint left = (GLint)(eleft->sx + y * eleft->dxdy); + GLint right = (GLint)(eright->sx + y * eright->dxdy); + + /* scissor x: + */ + if (setup->stage.softpipe->setup.scissor) { + if (left < setup->stage.softpipe->scissor.minx) + left = setup->stage.softpipe->scissor.minx; + + if (right > setup->stage.softpipe->scissor.maxx) + right = setup->stage.softpipe->scissor.maxx; + } + + if (left < right) { + GLint _y = sy+y; + if (block(_y) != setup->span.y) { + flush_spans(setup); + setup->span.y = block(_y); + } + + setup->span.left[_y&1] = left; + setup->span.right[_y&1] = right; + setup->span.y_flags |= 1<<(_y&1); + } + } + + + /* save the values so that emaj can be restarted: + */ + eleft->sx += lines * eleft->dxdy; + eright->sx += lines * eright->dxdy; + eleft->sy += lines; + eright->sy += lines; +} + + +/** + * Do setup for triangle rasterization, then render the triangle. + */ +static void setup_tri( struct prim_stage *stage, + struct prim_header *prim ) +{ + struct setup_stage *setup = setup_stage( stage ); + + _mesa_printf("%s\n", __FUNCTION__ ); + + setup_sort_vertices( setup, prim ); + setup_tri_coefficients( setup ); + setup_tri_edges( setup ); + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +// setup->span.z_mode = tri_z_mode( setup->ctx ); + +// init_constant_attribs( setup ); + + if (setup->oneoverarea < 0.0) { + /* emaj on left: + */ + subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); + subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); + } + else { + /* emaj on right: + */ + subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); + subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); + } + + flush_spans( setup ); +} + + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a line. + */ +static void +line_linear_coeff(struct setup_stage *setup, GLuint slot, GLuint i) +{ + const GLfloat dz = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + const GLfloat dadx = dz * setup->emaj.dx * setup->oneoverarea; + const GLfloat dady = dz * setup->emaj.dy * setup->oneoverarea; + setup->coef[slot].dadx[i] = dadx; + setup->coef[slot].dady[i] = dady; + setup->coef[slot].a0[i] + = (setup->vmin->data[slot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5) + + dady * (setup->vmin->data[0][1] - 0.5))); +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a line. + */ +static void +line_persp_coeff(struct setup_stage *setup, GLuint slot, GLuint i) +{ + /* XXX to do */ + line_linear_coeff(setup, slot, i); /* XXX temporary */ +} + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmax are initialized. + */ +static INLINE void +setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) +{ + const enum interp_mode *interp = setup->stage.softpipe->interp; + GLuint slot, j; + + /* use setup->vmin, vmax to point to vertices */ + setup->vprovoke = prim->v[1]; + setup->vmin = prim->v[0]; + setup->vmax = prim->v[1]; + + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + /* NOTE: this is not really 1/area */ + setup->oneoverarea = 1.0 / (setup->emaj.dx * setup->emaj.dx + + setup->emaj.dy * setup->emaj.dy); + + /* z and w are done by linear interpolation: + */ + line_linear_coeff(setup, 0, 2); + line_linear_coeff(setup, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + switch (interp[slot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + line_linear_coeff(setup, slot, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + line_persp_coeff(setup, slot, j); + break; + } + } +} + + +/** + * Plot a pixel in a line segment. + */ +static INLINE void +plot(struct setup_stage *setup, GLint x, GLint y) +{ + const GLint iy = y & 1; + const GLint ix = x & 1; + const GLint quadX = x - ix; + const GLint quadY = y - iy; + const GLint mask = (1 << ix) << (2 * iy); + + if (quadX != setup->quad.x0 || + quadY != setup->quad.y0) + { + /* flush prev quad, start new quad */ + + if (setup->quad.x0 != -1) + quad_shade(setup->stage.softpipe, &setup->quad); + + setup->quad.x0 = quadX; + setup->quad.y0 = quadY; + setup->quad.mask = 0x0; + } + + setup->quad.mask |= mask; +} + + + +/** + * Do setup for line rasterization, then render the line. + * XXX single-pixel width, no stipple, etc + * XXX no scissoring yet. + */ +static void +setup_line(struct prim_stage *stage, struct prim_header *prim) +{ + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + struct setup_stage *setup = setup_stage( stage ); + + GLint x0 = (GLint) v0->data[0][0]; + GLint x1 = (GLint) v1->data[0][0]; + GLint y0 = (GLint) v0->data[0][1]; + GLint y1 = (GLint) v1->data[0][1]; + GLint dx = x1 - x0; + GLint dy = y1 - y0; + GLint xstep, ystep; + + if (dx == 0 && dy == 0) + return; + + setup_line_coefficients(setup, prim); + + if (dx < 0) { + dx = -dx; /* make positive */ + xstep = -1; + } + else { + xstep = 1; + } + + if (dy < 0) { + dy = -dy; /* make positive */ + ystep = -1; + } + else { + ystep = 1; + } + + assert(dx >= 0); + assert(dy >= 0); + + setup->quad.x0 = setup->quad.y0 = -1; + setup->quad.mask = 0x0; + + if (dx > dy) { + /*** X-major line ***/ + GLint i; + const GLint errorInc = dy + dy; + GLint error = errorInc - dx; + const GLint errorDec = error - dx; + + for (i = 0; i < dx; i++) { + plot(setup, x0, y0); + + x0 += xstep; + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + y0 += ystep; + } + } + } + else { + /*** Y-major line ***/ + GLint i; + const GLint errorInc = dx + dx; + GLint error = errorInc - dy; + const GLint errorDec = error - dy; + + for (i = 0; i < dy; i++) { + plot(setup, x0, y0); + + y0 += ystep; + + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + x0 += xstep; + } + } + } + + /* draw final quad */ + if (setup->quad.mask) { + quad_shade(setup->stage.softpipe, &setup->quad); + } +} + + +/** + * Do setup for point rasterization, then render the point. + * Round or square points... + * XXX could optimize a lot for 1-pixel points. + */ +static void +setup_point(struct prim_stage *stage, struct prim_header *prim) +{ + struct setup_stage *setup = setup_stage( stage ); + GLfloat halfSize = 7.3; /*XXX this is a vertex attrib */ + GLfloat halfSizeSquared = halfSize * halfSize; + const struct vertex_header *v0 = prim->v[0]; + const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; + const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; + const GLint ixmin = block((GLint) (x - halfSize)); + const GLint ixmax = block((GLint) (x + halfSize)); + const GLint iymin = block((GLint) (y - halfSize)); + const GLint iymax = block((GLint) (y + halfSize)); + GLboolean round = GL_TRUE; + GLint ix, iy; + GLuint slot, j; + + /* For points, all interpolants are constant-valued. + * However, for point sprites, we'll need to setup texcoords appropriately. + * XXX: which coefficients are the texcoords??? + * We may do point sprites as textured quads... + * + * KW: We don't know which coefficients are texcoords - ultimately + * the choice of what interpolation mode to use for each attribute + * should be determined by the fragment program, using + * per-attribute declaration statements that include interpolation + * mode as a parameter. So either the fragment program will have + * to be adjusted for pointsprite vs normal point behaviour, or + * otherwise a special interpolation mode will have to be defined + * which matches the required behaviour for point sprites. But - + * the latter is not a feature of normal hardware, and as such + * probably should be ruled out on that basis. + */ + setup->vprovoke = prim->v[0]; + const_coeff(setup, 0, 2); + const_coeff(setup, 0, 3); + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + } + + /* XXX need to clip against scissor bounds too */ + + for (iy = iymin; iy <= iymax; iy += 2) { + for (ix = ixmin; ix <= ixmax; ix += 2) { + + if (round) { + /* rounded points */ + /* XXX for GL_SMOOTH, need to compute per-fragment coverage too */ + GLfloat dx, dy; + + setup->quad.mask = 0x0; + + dx = (ix + 0.5) - x; + dy = (iy + 0.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_BOTTOM_LEFT; + + dx = (ix + 1.5) - x; + dy = (iy + 0.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_BOTTOM_RIGHT; + + dx = (ix + 0.5) - x; + dy = (iy + 1.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_TOP_LEFT; + + dx = (ix + 1.5) - x; + dy = (iy + 1.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_TOP_RIGHT; + } + else { + /* square points */ + setup->quad.mask = 0xf; + + if (ix + 0.5 < x - halfSize) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + + if (ix + 1.5 > x + halfSize) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + + if (iy + 0.5 < y - halfSize) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + + if (iy + 1.5 > y + halfSize) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + } + + if (setup->quad.mask) { + setup->quad.x0 = ix; + setup->quad.y0 = iy; + quad_shade( setup->stage.softpipe, &setup->quad ); + } + } + } +} + + + +static void setup_end( struct prim_stage *stage ) +{ +} + + +struct prim_stage *prim_setup( struct softpipe_context *softpipe ) +{ + struct setup_stage *setup = CALLOC_STRUCT(setup_stage); + + setup->stage.softpipe = softpipe; + setup->stage.begin = setup_begin; + setup->stage.point = setup_point; + setup->stage.line = setup_line; + setup->stage.tri = setup_tri; + setup->stage.end = setup_end; + + setup->quad.coef = setup->coef; + + return &setup->stage; +} diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h new file mode 100644 index 0000000000..40a70c543e --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_setup.h @@ -0,0 +1,121 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef S_TRI_H +#define S_TRI_H + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "imports.h" +#include "s_tri_public.h" +#include "s_context.h" + + + +struct tri_context; +struct fp_context; +struct be_context; + +/* Note the rasterizer does not take a GLcontext argument. This is + * deliberate. + */ +struct tri_context *tri_create_context( GLcontext *ctx ); + +void tri_destroy_context( struct tri_context *tri ); + +void tri_set_fp_context( struct tri_context *tri, + struct fp_context *fp, + void (*fp_run)( struct fp_context *fp, + const struct fp_inputs *, + struct fp_outputs * )); + + +void tri_set_be_context( struct tri_context *tri, + struct be_context *be, + void (*be_run)( struct be_context *be, + const struct fp_outputs * )); + +void tri_set_attribs( struct tri_context *tri, + const struct attr_info *info, + GLuint nr_attrib ); + +void tri_set_backface( struct tri_context *tri, + GLfloat backface ); + +void tri_set_scissor( struct tri_context *tri, + GLint x, + GLint y, + GLuint width, + GLuint height, + GLboolean enabled ); + +void tri_set_stipple( struct tri_context *tri, + const GLuint *pattern, + GLboolean enabled ); + +/* Unfilled triangles will be handled elsewhere (higher in the + * pipeline), as will things like stipple (lower in the pipeline). + */ + +void tri_triangle( struct tri_context *tri, + const struct vertex *v0, + const struct vertex *v1, + const struct vertex *v2 ); + +/* TODO: rasterize_line, rasterize_point?? + * How will linestipple work? + */ + + +#ifdef SETUP_PRIVATE + + + + + + +GLboolean tri_setup( struct tri_context *tri, + const struct vertex *v0, + const struct vertex *v1, + const struct vertex *v2 ); + +void tri_rasterize( struct tri_context *tri ); +void tri_rasterize_spans( struct tri_context *tri ); + + + + + + +#endif +#endif diff --git a/src/mesa/pipe/softpipe/sp_prim_twoside.c b/src/mesa/pipe/softpipe/sp_prim_twoside.c new file mode 100644 index 0000000000..6eee1de588 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_twoside.c @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" +#include "sp_context.h" +#include "sp_prim.h" + + +struct twoside_stage { + struct prim_stage stage; + + GLfloat facing; + const GLuint *lookup; +}; + + +static INLINE struct twoside_stage *twoside_stage( struct prim_stage *stage ) +{ + return (struct twoside_stage *)stage; +} + + +static void twoside_begin( struct prim_stage *stage ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + twoside->facing = (stage->softpipe->setup.front_winding == WINDING_CW) ? -1 : 1; + + stage->next->begin( stage->next ); +} + + +static INLINE void copy_color( GLuint attr_dst, + GLuint attr_src, + struct vertex_header *v ) +{ + if (attr_dst && attr_src) { + memcpy( v->data[attr_dst], + v->data[attr_src], + sizeof(v->data[0]) ); + } +} + + +static struct vertex_header *copy_bfc( struct twoside_stage *twoside, + const struct vertex_header *v, + GLuint idx ) +{ + struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); + + copy_color( twoside->lookup[VF_ATTRIB_COLOR0], + twoside->lookup[VF_ATTRIB_BFC0], + tmp ); + + copy_color( twoside->lookup[VF_ATTRIB_COLOR1], + twoside->lookup[VF_ATTRIB_BFC1], + tmp ); + + return tmp; +} + + +/* Twoside tri: + */ +static void twoside_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + if (header->det * twoside->facing < 0) { + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = copy_bfc(twoside, header->v[0], 0); + tmp.v[1] = copy_bfc(twoside, header->v[1], 1); + tmp.v[2] = copy_bfc(twoside, header->v[2], 2); + + stage->next->tri( stage->next, &tmp ); + } + else { + stage->next->tri( stage->next, header ); + } +} + + +static void twoside_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void twoside_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void twoside_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + + + +struct prim_stage *prim_twoside( struct softpipe_context *softpipe ) +{ + struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); + + prim_alloc_tmps( &twoside->stage, 3 ); + + twoside->stage.softpipe = softpipe; + twoside->stage.next = NULL; + twoside->stage.begin = twoside_begin; + twoside->stage.point = twoside_point; + twoside->stage.line = twoside_line; + twoside->stage.tri = twoside_tri; + twoside->stage.end = twoside_end; + + twoside->lookup = softpipe->vf_attr_to_slot; + + return &twoside->stage; +} diff --git a/src/mesa/pipe/softpipe/sp_prim_unfilled.c b/src/mesa/pipe/softpipe/sp_prim_unfilled.c new file mode 100644 index 0000000000..73147337fd --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_unfilled.c @@ -0,0 +1,165 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_prim.h" + + + +struct unfilled_stage { + struct prim_stage stage; + + GLuint mode[2]; +}; + + +static INLINE struct unfilled_stage *unfilled_stage( struct prim_stage *stage ) +{ + return (struct unfilled_stage *)stage; +} + + +static void unfilled_begin( struct prim_stage *stage ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + + unfilled->mode[0] = stage->softpipe->setup.fill_ccw; + unfilled->mode[1] = stage->softpipe->setup.fill_cw; + + stage->next->begin( stage->next ); +} + +static void point( struct prim_stage *stage, + struct vertex_header *v0 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + stage->next->point( stage->next, &tmp ); +} + +static void line( struct prim_stage *stage, + struct vertex_header *v0, + struct vertex_header *v1 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + tmp.v[1] = v1; + stage->next->line( stage->next, &tmp ); +} + + +static void points( struct prim_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (v0->edgeflag) point( stage, v0 ); + if (v1->edgeflag) point( stage, v1 ); + if (v2->edgeflag) point( stage, v2 ); +} + +static void lines( struct prim_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (v0->edgeflag) line( stage, v0, v1 ); + if (v1->edgeflag) line( stage, v1, v2 ); + if (v2->edgeflag) line( stage, v2, v0 ); +} + + +/* Unfilled tri: + * + * Note edgeflags in the vertex struct is not sufficient as we will + * need to manipulate them when decomposing primitives??? + */ +static void unfilled_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + GLuint mode = unfilled->mode[header->det < 0]; + + switch (mode) { + case FILL_TRI: + stage->next->tri( stage->next, header ); + break; + + case FILL_LINE: + lines( stage, header ); + break; + + case GL_POINT: + points( stage, header ); + break; + } +} + +static void unfilled_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void unfilled_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void unfilled_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_unfilled( struct softpipe_context *softpipe ) +{ + struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); + + prim_alloc_tmps( &unfilled->stage, 0 ); + + unfilled->stage.softpipe = softpipe; + unfilled->stage.next = NULL; + unfilled->stage.tmp = NULL; + unfilled->stage.begin = unfilled_begin; + unfilled->stage.point = unfilled_point; + unfilled->stage.line = unfilled_line; + unfilled->stage.tri = unfilled_tri; + unfilled->stage.end = unfilled_end; + + return &unfilled->stage; +} diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h new file mode 100644 index 0000000000..2abefc6037 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -0,0 +1,61 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef PRIM_H +#define PRIM_H + +#include "glheader.h" +#include "pipe/p_state.h" + + +void softpipe_set_clip_state( struct pipe_context *, + const struct pipe_clip_state * ); + +void softpipe_set_viewport( struct pipe_context *, + const struct pipe_viewport * ); + +void softpipe_set_setup_state( struct pipe_context *, + const struct pipe_setup_state * ); + +void softpipe_set_scissor_rect( struct pipe_context *, + const struct pipe_scissor_rect * ); + +void softpipe_set_fs_state( struct pipe_context *, + const struct pipe_fs_state * ); + +void softpipe_set_polygon_stipple( struct pipe_context *, + const struct pipe_poly_stipple * ); + +void softpipe_set_cbuf_state( struct pipe_context *, + const struct pipe_surface * ); + +void softpipe_update_derived( struct softpipe_context *softpipe ); + +#endif diff --git a/src/mesa/pipe/softpipe/sp_state_clip.c b/src/mesa/pipe/softpipe/sp_state_clip.c new file mode 100644 index 0000000000..6e83b055fb --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_clip.c @@ -0,0 +1,69 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_draw.h" + + + +void softpipe_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + memcpy(&softpipe->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); + + softpipe->nr_planes = 6 + clip->nr; + softpipe->dirty |= G_NEW_CLIP; +} + + + +/* Called when driver state tracker notices changes to the viewport + * matrix: + */ +void softpipe_set_viewport( struct pipe_context *pipe, + const struct pipe_viewport *viewport ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + memcpy(&softpipe->viewport, viewport, sizeof(viewport)); + + /* Using tnl/ and vf/ modules is temporary while getting started. + * Full pipe will have vertex shader, vertex fetch of its own. + */ + draw_set_viewport( softpipe->draw, viewport->scale, viewport->translate ); + softpipe->dirty |= G_NEW_VIEWPORT; +} + + + diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c new file mode 100644 index 0000000000..1f7329600a --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -0,0 +1,136 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "macros.h" +#include "enums.h" +#include "program.h" + +#include "sp_context.h" +#include "sp_draw.h" +#include "sp_state.h" + +#define EMIT_ATTR( ATTR, FRAG_ATTR, INTERP ) \ +do { \ + slot_to_vf_attr[softpipe->nr_attrs] = ATTR; \ + softpipe->vf_attr_to_slot[ATTR] = softpipe->nr_attrs; \ + softpipe->fp_attr_to_slot[FRAG_ATTR] = softpipe->nr_attrs; \ + softpipe->interp[softpipe->nr_attrs] = INTERP; \ + softpipe->nr_attrs++; \ + attr_mask |= (1<fs.fp; + const GLuint inputsRead = fp->Base.InputsRead; + GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; + GLuint attr_mask = 0; + GLuint i; + + softpipe->nr_attrs = 0; + memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); + + memset(softpipe->fp_attr_to_slot, 0, sizeof(softpipe->vf_attr_to_slot)); + memset(softpipe->vf_attr_to_slot, 0, sizeof(softpipe->fp_attr_to_slot)); + + /* TODO - Figure out if we need to do perspective divide, etc. + */ + EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); + + /* Pull in the rest of the attributes. They are all in float4 + * format. Future optimizations could be to keep some attributes + * as fixed point or ubyte format. + */ + for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { + if (inputsRead & (i << i)) { + EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); + } + } + + for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { + if (inputsRead & (i << i)) { + EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); + } + } + + softpipe->nr_frag_attrs = softpipe->nr_attrs; + + /* Additional attributes required for setup: Just twosided + * lighting. Edgeflag is dealt with specially by setting bits in + * the vertex header. + */ + if (softpipe->setup.light_twoside) { + if (inputsRead & FRAG_BIT_COL0) { + EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ + } + + if (inputsRead & FRAG_BIT_COL1) { + EMIT_ATTR(VF_ATTRIB_BFC1, FRAG_ATTRIB_MAX, 0); /* XXX: discard after setup */ + } + } + + if (attr_mask != softpipe->attr_mask) { + softpipe->attr_mask = attr_mask; + + draw_set_vertex_attributes( softpipe->draw, + slot_to_vf_attr, + softpipe->nr_attrs ); + } +} + + +/* Hopefully this will remain quite simple, otherwise need to pull in + * something like the state tracker mechanism. + */ +void softpipe_update_derived( struct softpipe_context *softpipe ) +{ + if (softpipe->dirty & (G_NEW_SETUP | G_NEW_FS)) + calculate_vertex_layout( softpipe ); + + softpipe->dirty = 0; +} diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c new file mode 100644 index 0000000000..29f965e50e --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "sp_context.h" +#include "sp_state.h" + + + +void softpipe_set_fs_state( struct pipe_context *pipe, + const struct pipe_fs_state *fs ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + memcpy(&softpipe->fs, fs, sizeof(*fs)); + + softpipe->dirty |= G_NEW_FS; +} + + + + + + + + + diff --git a/src/mesa/pipe/softpipe/sp_state_setup.c b/src/mesa/pipe/softpipe/sp_state_setup.c new file mode 100644 index 0000000000..7be7d7fe40 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_setup.c @@ -0,0 +1,122 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_prim.h" + + + + +static void validate_prim_pipe( struct softpipe_context *softpipe ) +{ + struct prim_stage *next = softpipe->prim.setup; + + /* TODO: make the current primitive part of the state and build + * shorter pipelines for lines & points. + */ + if (softpipe->setup.fill_cw != FILL_TRI || + softpipe->setup.fill_ccw != FILL_TRI) { + + softpipe->prim.unfilled->next = next; + next = softpipe->prim.unfilled; + } + + if (softpipe->setup.offset_cw || + softpipe->setup.offset_ccw) { + softpipe->prim.offset->next = next; + next = softpipe->prim.offset; + } + + if (softpipe->setup.light_twoside) { + softpipe->prim.twoside->next = next; + next = softpipe->prim.twoside; + } + + /* Always run the cull stage as we calculate determinant there + * also. Fix this.. + */ + { + softpipe->prim.cull->next = next; + next = softpipe->prim.cull; + } + + + /* Clip stage + */ + { + softpipe->prim.clip->next = next; + next = softpipe->prim.clip; + } + + /* Do software flatshading prior to clipping. XXX: should only do + * this for clipped primitives, ie it is a part of the clip + * routine. + */ + if (softpipe->setup.flatshade) { + softpipe->prim.flatshade->next = next; + next = softpipe->prim.flatshade; + } + + + softpipe->prim.first = next; +} + + + + +void softpipe_set_setup_state( struct pipe_context *pipe, + const struct pipe_setup_state *setup ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + memcpy( &softpipe->setup, setup, sizeof(*setup) ); + + validate_prim_pipe( softpipe ); + softpipe->dirty |= G_NEW_SETUP; +} + + + +void softpipe_set_scissor_rect( struct pipe_context *pipe, + const struct pipe_scissor_rect *scissor ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + memcpy( &softpipe->scissor, scissor, sizeof(*scissor) ); + softpipe->dirty |= G_NEW_SCISSOR; +} + + +void softpipe_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + memcpy( &softpipe->poly_stipple, stipple, sizeof(*stipple) ); + softpipe->dirty |= G_NEW_STIPPLE; +} diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c new file mode 100644 index 0000000000..989f2e29d8 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_surface.h" + + +/* This is all a total hack. + */ +void softpipe_set_cbuf_state( struct pipe_context *pipe, + const struct pipe_surface *surface ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + if (softpipe->cbuf_surface == NULL) { + softpipe->cbuf_surface = CALLOC_STRUCT(softpipe_surface); + softpipe->cbuf_surface->type = &gs_rgba8; + } + + softpipe->cbuf_surface->surface = *surface; + softpipe->dirty |= G_NEW_CBUF; +} + + diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c new file mode 100644 index 0000000000..16bbacb12b --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -0,0 +1,153 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_surface.h" +#include "sp_headers.h" + +static void rgba8_read_quad_f( struct softpipe_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + rgba[k][0] = ptr[0] * (1.0 / 255.0); + rgba[k][1] = ptr[1] * (1.0 / 255.0); + rgba[k][2] = ptr[2] * (1.0 / 255.0); + rgba[k][3] = ptr[3] * (1.0 / 255.0); + } + } +} + +static void rgba8_read_quad_f_swz( struct softpipe_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + rrrr[0][k] = ptr[0] * (1.0 / 255.0); + rrrr[1][k] = ptr[1] * (1.0 / 255.0); + rrrr[2][k] = ptr[2] * (1.0 / 255.0); + rrrr[3][k] = ptr[3] * (1.0 / 255.0); + } + } +} + +static void rgba8_write_quad_f( struct softpipe_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + ptr[0] = rgba[k][0] * 255.0; + ptr[1] = rgba[k][1] * 255.0; + ptr[2] = rgba[k][2] * 255.0; + ptr[3] = rgba[k][3] * 255.0; + } + } +} + +static void rgba8_write_quad_f_swz( struct softpipe_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + ptr[0] = rrrr[0][k] * 255.0; + ptr[1] = rrrr[1][k] * 255.0; + ptr[2] = rrrr[2][k] * 255.0; + ptr[3] = rrrr[3][k] * 255.0; + } + } +} + + + + +static void rgba8_read_quad_ub( struct softpipe_surface *gs, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + rgba[k][0] = ptr[0]; + rgba[k][1] = ptr[1]; + rgba[k][2] = ptr[2]; + rgba[k][3] = ptr[3]; + } + } +} + + +static void rgba8_write_quad_ub( struct softpipe_surface *gs, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ) +{ + GLuint i, j, k = 0; + + for (i = 0; i < 2; i++) { + for (j = 0; j < 2; j++, k++) { + GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; + ptr[0] = rgba[k][0]; + ptr[1] = rgba[k][1]; + ptr[2] = rgba[k][2]; + ptr[3] = rgba[k][3]; + } + } +} + + + + +struct softpipe_surface_type gs_rgba8 = { + G_SURFACE_RGBA_8888, + rgba8_read_quad_f, + rgba8_read_quad_f_swz, + rgba8_read_quad_ub, + rgba8_write_quad_f, + rgba8_write_quad_f_swz, + rgba8_write_quad_ub, +}; + + + diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h new file mode 100644 index 0000000000..08b6889257 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_SURFACE_H +#define G_SURFACE_H + +#include "glheader.h" +#include "sp_headers.h" + +struct softpipe_surface; + +#define G_SURFACE_RGBA_8888 0x1 + +/* Internal structs and helpers for the primitive clip/setup pipeline: + */ +struct softpipe_surface_type { + + GLuint format; + + void (*read_quad_f)( struct softpipe_surface *, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ); + + void (*read_quad_f_swz)( struct softpipe_surface *, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ); + + void (*read_quad_ub)( struct softpipe_surface *, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ); + + + void (*write_quad_f)( struct softpipe_surface *, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ); + + void (*write_quad_f_swz)( struct softpipe_surface *, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ); + + + void (*write_quad_ub)( struct softpipe_surface *, + GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS] ); + + +}; + + +struct softpipe_surface { + struct softpipe_surface_type *type; + struct pipe_surface surface; +}; + + +static INLINE void gs_read_quad_f( struct softpipe_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + gs->type->read_quad_f(gs, x, y, rgba); +} + +static INLINE void gs_read_quad_f_swz( struct softpipe_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + gs->type->read_quad_f_swz(gs, x, y, rrrr); +} + +static INLINE void gs_write_quad_f( struct softpipe_surface *gs, + GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS] ) +{ + gs->type->write_quad_f(gs, x, y, rgba); +} + +static INLINE void gs_write_quad_f_swz( struct softpipe_surface *gs, + GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE] ) +{ + gs->type->write_quad_f_swz(gs, x, y, rrrr); +} + +/* Like this, or hidden? + */ +struct softpipe_surface_type gs_rgba8; + +#endif diff --git a/src/mesa/pipe/softpipe/sp_tile.h b/src/mesa/pipe/softpipe/sp_tile.h new file mode 100644 index 0000000000..f4808b99d2 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tile.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#ifndef G_TILE_H +#define G_TILE_H + +struct softpipe_context; +struct quad_header; + +void quad_shade( struct softpipe_context *softpipe, + struct quad_header *quad ); + +void quad_output( struct softpipe_context *softpipe, + struct quad_header *quad ); + +#endif diff --git a/src/mesa/pipe/softpipe/sp_tile_fs.c b/src/mesa/pipe/softpipe/sp_tile_fs.c new file mode 100644 index 0000000000..c145fcf089 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tile_fs.c @@ -0,0 +1,203 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_tile.h" + +struct exec_machine { + const struct setup_coefficient *coef; + + GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE]; +}; + + +/** + * Compute quad's attributes values, as constants (GL_FLAT shading). + */ +static INLINE void cinterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + exec->attr[attrib][i][j] = exec->coef[attrib].a0[i]; + } +} + + +/** + * Compute quad's attribute values by linear interpolation. + * + * Push into the fp: + * + * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx + * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy + */ +static INLINE void linterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; + exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + + exec->coef[attrib].dadx[i] * x + + exec->coef[attrib].dady[i] * y); + } +} + + +/** + * Compute quad's attribute values by linear interpolation with + * perspective correction. + * + * Push into the fp: + * + * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx + * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy + * INPUT[attr] = MUL INPUT[attr], INPUT_WPOS.wwww + * + * (Or should that be 1/w ???) + */ +static INLINE void pinterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; + const GLfloat invW = exec->attr[FRAG_ATTRIB_WPOS][3][j]; + exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + + exec->coef[attrib].dadx[i] * x + + exec->coef[attrib].dady[i] * y) * invW); + } +} + + + +/* This should be done by the fragment shader execution unit (code + * generated from the decl instructions). Do it here for now. + */ +void quad_shade( struct softpipe_context *softpipe, + struct quad_header *quad ) +{ + struct exec_machine exec; + GLfloat fx = quad->x0; + GLfloat fy = quad->y0; + GLuint i, j; + + exec.coef = quad->coef; + + /* Position: + */ + exec.attr[FRAG_ATTRIB_WPOS][0][0] = fx; + exec.attr[FRAG_ATTRIB_WPOS][0][1] = fx + 1.0; + exec.attr[FRAG_ATTRIB_WPOS][0][2] = fx; + exec.attr[FRAG_ATTRIB_WPOS][0][3] = fx + 1.0; + + exec.attr[FRAG_ATTRIB_WPOS][1][0] = fy; + exec.attr[FRAG_ATTRIB_WPOS][1][1] = fy; + exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0; + exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; + + /* Z and W are done by linear interpolation: + * XXX we'll probably have to use integers for Z + */ + if (softpipe->need_z) { + linterp(&exec, 0, 2); + } + + if (softpipe->need_w) { + linterp(&exec, 0, 3); +// invert(&exec, 0, 3); + } + + /* Interpolate all the remaining attributes. This will get pushed + * into the fragment program's responsibilities at some point. + */ + for (i = 1; i < quad->nr_attrs; i++) { +#if 1 + for (j = 0; j < NUM_CHANNELS; j++) + linterp(&exec, i, j); +#else + switch (quad->interp[i]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + cinterp(&exec, i, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + linterp(&exec, i, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + pinterp(&exec, i, j); + break; + } +#endif + } + +#if 0 + softpipe->run_fs( tri->fp, quad, &tri->outputs ); +#else + { + GLuint attr = softpipe->fp_attr_to_slot[FRAG_ATTRIB_COL0]; + assert(attr); + + memcpy(quad->outputs.color, + exec.attr[attr], + sizeof(quad->outputs.color)); + } +#endif + + + if (quad->mask) + quad_output( softpipe, quad ); +} + + + + + + + diff --git a/src/mesa/pipe/softpipe/sp_tile_output.c b/src/mesa/pipe/softpipe/sp_tile_output.c new file mode 100644 index 0000000000..b1eb9e8c9f --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tile_output.c @@ -0,0 +1,92 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_tile.h" + + +static void mask_copy( GLfloat (*dest)[4], + GLfloat (*src)[4], + GLuint mask ) +{ + GLuint i, j; + + for (i = 0; i < 4; i++) { + if (mask & (1<mask != MASK_ALL) + { + GLfloat tmp[4][QUAD_SIZE]; + + /* Yes, we'll probably have a masked write as well, but this is + * how blend will be done at least. + */ + gs_read_quad_f_swz( softpipe->cbuf_surface, + quad->x0, + quad->y0, + tmp ); + + mask_copy( tmp, quad->outputs.color, quad->mask ); + + gs_write_quad_f_swz( softpipe->cbuf_surface, + quad->x0, + quad->y0, + tmp ); + } + else + { + gs_write_quad_f_swz( softpipe->cbuf_surface, + quad->x0, + quad->y0, + quad->outputs.color ); + } +} diff --git a/src/mesa/softpipe/Makefile b/src/mesa/softpipe/Makefile deleted file mode 100644 index 451911a354..0000000000 --- a/src/mesa/softpipe/Makefile +++ /dev/null @@ -1,2 +0,0 @@ -default: - cd .. ; make diff --git a/src/mesa/softpipe/generic/Makefile b/src/mesa/softpipe/generic/Makefile deleted file mode 100644 index 12a8bd0409..0000000000 --- a/src/mesa/softpipe/generic/Makefile +++ /dev/null @@ -1,3 +0,0 @@ -default: - cd ../.. ; make - diff --git a/src/mesa/softpipe/generic/g_context.c b/src/mesa/softpipe/generic/g_context.c deleted file mode 100644 index c3329b5c7d..0000000000 --- a/src/mesa/softpipe/generic/g_context.c +++ /dev/null @@ -1,135 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - -#include "imports.h" -#include "macros.h" - -#include "tnl/t_context.h" -#include "vf/vf.h" - -#include "g_context.h" -#include "g_prim.h" -#include "g_state.h" -#include "g_draw.h" - -static void generic_destroy( struct softpipe_context *softpipe ) -{ - struct generic_context *generic = generic_context( softpipe ); - - draw_destroy( generic->draw ); - - FREE( generic ); -} - - -static void generic_draw_vb( struct softpipe_context *softpipe, - struct vertex_buffer *VB ) -{ - struct generic_context *generic = generic_context( softpipe ); - - if (generic->dirty) - generic_update_derived( generic ); - - draw_vb( generic->draw, VB ); -} - -struct softpipe_context *generic_create( void ) -{ - struct generic_context *generic = CALLOC_STRUCT(generic_context); - - generic->softpipe.destroy = generic_destroy; - generic->softpipe.set_clip_state = generic_set_clip_state; - generic->softpipe.set_viewport = generic_set_viewport; - generic->softpipe.set_setup_state = generic_set_setup_state; - generic->softpipe.set_scissor_rect = generic_set_scissor_rect; - generic->softpipe.set_fs_state = generic_set_fs_state; - generic->softpipe.set_polygon_stipple = generic_set_polygon_stipple; - generic->softpipe.set_cbuf_state = generic_set_cbuf_state; - generic->softpipe.draw_vb = generic_draw_vb; - - - - generic->prim.setup = prim_setup( generic ); - generic->prim.unfilled = prim_unfilled( generic ); - generic->prim.twoside = prim_twoside( generic ); - generic->prim.offset = prim_offset( generic ); - generic->prim.clip = prim_clip( generic ); - generic->prim.flatshade = prim_flatshade( generic ); - generic->prim.cull = prim_cull( generic ); - - - generic->draw = draw_create( generic ); - - ASSIGN_4V( generic->plane[0], -1, 0, 0, 1 ); - ASSIGN_4V( generic->plane[1], 1, 0, 0, 1 ); - ASSIGN_4V( generic->plane[2], 0, -1, 0, 1 ); - ASSIGN_4V( generic->plane[3], 0, 1, 0, 1 ); - ASSIGN_4V( generic->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ - ASSIGN_4V( generic->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ - generic->nr_planes = 6; - - return &generic->softpipe; -} - - - - - - -#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) - -void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ) -{ - stage->nr_tmps = nr; - - if (nr) { - GLubyte *store = MALLOC(MAX_VERTEX_SIZE * nr); - GLuint i; - - stage->tmp = MALLOC(sizeof(struct vertex_header *) * nr); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); - } -} - -void prim_free_tmps( struct prim_stage *stage ) -{ - if (stage->tmp) { - FREE(stage->tmp[0]); - FREE(stage->tmp); - } -} - - - - - diff --git a/src/mesa/softpipe/generic/g_context.h b/src/mesa/softpipe/generic/g_context.h deleted file mode 100644 index 24d467f2c3..0000000000 --- a/src/mesa/softpipe/generic/g_context.h +++ /dev/null @@ -1,149 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_CONTEXT_H -#define G_CONTEXT_H - -#include "glheader.h" - -#include "softpipe/sp_state.h" -#include "softpipe/sp_context.h" - - - -struct generic_surface; -struct draw_context; -struct prim_stage; - - -enum interp_mode { - INTERP_CONSTANT, - INTERP_LINEAR, - INTERP_PERSPECTIVE -}; - - -#define G_NEW_VIEWPORT 0x1 -#define G_NEW_SETUP 0x2 -#define G_NEW_FS 0x4 -#define G_NEW_BLEND 0x8 -#define G_NEW_CBUF 0x10 -#define G_NEW_CLIP 0x20 -#define G_NEW_SCISSOR 0x40 -#define G_NEW_STIPPLE 0x80 - - - -struct generic_context { - struct softpipe_context softpipe; - - - /* The most recent drawing state as set by the driver: - */ - struct softpipe_viewport viewport; - struct softpipe_setup_state setup; - struct softpipe_fs_state fs; - struct softpipe_blend_state blend; - struct softpipe_surface cbuf; - struct softpipe_clip_state clip; - struct softpipe_scissor_rect scissor; - struct softpipe_poly_stipple poly_stipple; - GLuint dirty; - - - /* Cbuf derived state??? - */ - struct generic_surface *cbuf_surface; - - /* Clip derived state: - */ - GLfloat plane[12][4]; - GLuint nr_planes; - - /* Setup derived state. TODO: this should be passed in the program - * tokens as parameters to DECL instructions. - * - * For now we just set colors to CONST on flatshade, textures to - * perspective always and everything else to linear. - */ - enum interp_mode interp[VF_ATTRIB_MAX]; - - - /* FS + setup derived state: - */ - GLuint fp_attr_to_slot[VF_ATTRIB_MAX]; - GLuint vf_attr_to_slot[VF_ATTRIB_MAX]; - GLuint nr_attrs; - GLuint nr_frag_attrs; - GLuint attr_mask; - - GLboolean need_z; - GLboolean need_w; - - /* Stipple derived state: - */ - GLubyte stipple_masks[16][16]; - - - /* The software clipper/setup engine. - */ - struct { - struct prim_stage *setup; - struct prim_stage *unfilled; - struct prim_stage *twoside; - struct prim_stage *clip; - struct prim_stage *flatshade; - struct prim_stage *offset; - struct prim_stage *cull; - - struct prim_stage *first; - - GLenum prim; - GLuint vertex_size; - } prim; - - /* Temp kludge: - */ - struct draw_context *draw; -}; - - - - -static INLINE struct generic_context * -generic_context( struct softpipe_context *softpipe ) -{ - return (struct generic_context *)softpipe; -} - - - - -#endif diff --git a/src/mesa/softpipe/generic/g_draw.c b/src/mesa/softpipe/generic/g_draw.c deleted file mode 100644 index 4e094c21a0..0000000000 --- a/src/mesa/softpipe/generic/g_draw.c +++ /dev/null @@ -1,696 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "imports.h" - -#include "tnl/t_context.h" - -#include "g_context.h" -#include "g_prim.h" -#include "g_headers.h" -#include "g_draw.h" - -/* This file is a temporary set of hooks to allow us to use the tnl/ - * and vf/ modules until we have replacements in softpipe. - */ - - -struct draw_context -{ - struct generic_context *generic; - - struct vf_attr_map attrs[VF_ATTRIB_MAX]; - GLuint nr_attrs; - GLuint vertex_size; - struct vertex_fetch *vf; - - GLubyte *verts; - GLuint nr_vertices; - GLboolean in_vb; - - GLenum prim; - - /* Helper for tnl: - */ - GLvector4f header; -}; - - -static struct vertex_header *get_vertex( struct draw_context *pipe, - GLuint i ) -{ - return (struct vertex_header *)(pipe->verts + i * pipe->vertex_size); -} - - - -static void draw_allocate_vertices( struct draw_context *draw, - GLuint nr_vertices ) -{ - draw->nr_vertices = nr_vertices; - draw->verts = MALLOC( nr_vertices * draw->vertex_size ); - - draw->generic->prim.first->begin( draw->generic->prim.first ); -} - -static void draw_set_prim( struct draw_context *draw, - GLenum prim ) -{ - draw->prim = prim; - - /* Not done yet - need to force edgeflags to 1 in strip/fan - * primitives. - */ -#if 0 - switch (prim) { - case GL_TRIANGLES: - case GL_POLYGON: - case GL_QUADS: - case GL_QUAD_STRIP: /* yes, we need this */ - respect_edgeflags( pipe, GL_TRUE ); - break; - - default: - respect_edgeflags( pipe, GL_FALSE ); - break; - } -#endif -} - - - -static void do_quad( struct prim_stage *first, - struct vertex_header *v0, - struct vertex_header *v1, - struct vertex_header *v2, - struct vertex_header *v3 ) -{ - struct prim_header prim; - - { - GLuint tmp = v1->edgeflag; - v1->edgeflag = 0; - - prim.v[0] = v0; - prim.v[1] = v1; - prim.v[2] = v3; - first->tri( first, &prim ); - - v1->edgeflag = tmp; - } - - { - GLuint tmp = v3->edgeflag; - v3->edgeflag = 0; - - prim.v[0] = v1; - prim.v[1] = v2; - prim.v[2] = v3; - first->tri( first, &prim ); - - v3->edgeflag = tmp; - } -} - - - - -static void draw_indexed_prim( struct draw_context *draw, - const GLuint *elts, - GLuint count ) -{ - struct prim_stage * const first = draw->generic->prim.first; - struct prim_header prim; - GLuint i; - - prim.det = 0; /* valid from cull stage onwards */ - prim.v[0] = 0; - prim.v[1] = 0; - prim.v[2] = 0; - - switch (draw->prim) { - case GL_POINTS: - for (i = 0; i < count; i ++) { - prim.v[0] = get_vertex( draw, elts[i] ); - - first->point( first, &prim ); - } - break; - - case GL_LINES: - for (i = 0; i+1 < count; i += 2) { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - - first->line( first, &prim ); - } - break; - - case GL_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - prim.v[0] = get_vertex( draw, elts[i-1] ); - prim.v[1] = get_vertex( draw, elts[i] ); - first->line( first, &prim ); - } - - prim.v[0] = get_vertex( draw, elts[count-1] ); - prim.v[1] = get_vertex( draw, elts[0] ); - first->line( first, &prim ); - } - break; - - case GL_LINE_STRIP: - /* I'm guessing it will be necessary to have something like a - * render->reset_line_stipple() method to properly support - * splitting strips into primitives like this. Alternately we - * could just scan ahead to find individual clipped lines and - * otherwise leave the strip intact - that might be better, but - * require more complex code here. - */ - if (count >= 2) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, elts[0] ); - - for (i = 1; i < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, elts[i] ); - - first->line( first, &prim ); - } - } - break; - - case GL_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - prim.v[0] = get_vertex( draw, elts[i + 1] ); - prim.v[1] = get_vertex( draw, elts[i + 0] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - } - else { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - } - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_FAN: - if (count >= 3) { - prim.v[0] = get_vertex( draw, elts[0] ); - prim.v[1] = 0; - prim.v[2] = get_vertex( draw, elts[1] ); - - for (i = 0; i+2 < count; i++) { - prim.v[1] = prim.v[2]; - prim.v[2] = get_vertex( draw, elts[i+2] ); - - first->tri( first, &prim ); - } - } - break; - - case GL_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( first, - get_vertex( draw, elts[i + 0] ), - get_vertex( draw, elts[i + 1] ), - get_vertex( draw, elts[i + 2] ), - get_vertex( draw, elts[i + 3] )); - } - break; - - case GL_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( first, - get_vertex( draw, elts[i + 2] ), - get_vertex( draw, elts[i + 0] ), - get_vertex( draw, elts[i + 1] ), - get_vertex( draw, elts[i + 3] )); - } - break; - - - case GL_POLYGON: - if (count >= 3) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, elts[1] ); - prim.v[2] = get_vertex( draw, elts[0] ); - - for (i = 0; i+2 < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, elts[i+2] ); - - first->tri( first, &prim ); - } - } - break; - - default: - assert(0); - break; - } -} - -static void draw_prim( struct draw_context *draw, - GLuint start, - GLuint count ) -{ - struct prim_stage * const first = draw->generic->prim.first; - struct prim_header prim; - GLuint i; - -// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - - prim.det = 0; /* valid from cull stage onwards */ - prim.v[0] = 0; - prim.v[1] = 0; - prim.v[2] = 0; - - switch (draw->prim) { - case GL_POINTS: - for (i = 0; i < count; i ++) { - prim.v[0] = get_vertex( draw, start + i ); - first->point( first, &prim ); - } - break; - - case GL_LINES: - for (i = 0; i+1 < count; i += 2) { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - - first->line( first, &prim ); - } - break; - - case GL_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - prim.v[0] = get_vertex( draw, start + i - 1 ); - prim.v[1] = get_vertex( draw, start + i ); - first->line( first, &prim ); - } - - prim.v[0] = get_vertex( draw, start + count - 1 ); - prim.v[1] = get_vertex( draw, start + 0 ); - first->line( first, &prim ); - } - break; - - case GL_LINE_STRIP: - if (count >= 2) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, start + 0 ); - - for (i = 1; i < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, start + i ); - - first->line( first, &prim ); - } - } - break; - - case GL_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - prim.v[0] = get_vertex( draw, start + i + 1 ); - prim.v[1] = get_vertex( draw, start + i + 0 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - } - else { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - } - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_FAN: - if (count >= 3) { - prim.v[0] = get_vertex( draw, start + 0 ); - prim.v[1] = 0; - prim.v[2] = get_vertex( draw, start + 1 ); - - for (i = 0; i+2 < count; i++) { - prim.v[1] = prim.v[2]; - prim.v[2] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); - } - } - break; - - - case GL_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( first, - get_vertex( draw, start + i + 0 ), - get_vertex( draw, start + i + 1 ), - get_vertex( draw, start + i + 2 ), - get_vertex( draw, start + i + 3 )); - } - break; - - case GL_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( first, - get_vertex( draw, start + i + 2 ), - get_vertex( draw, start + i + 0 ), - get_vertex( draw, start + i + 1 ), - get_vertex( draw, start + i + 3 )); - } - break; - - case GL_POLYGON: - if (count >= 3) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, start + 1 ); - prim.v[2] = get_vertex( draw, start + 0 ); - - for (i = 0; i+2 < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); - } - } - break; - - default: - assert(0); - break; - } -} - - -static void draw_release_vertices( struct draw_context *draw ) -{ - draw->generic->prim.first->end( draw->generic->prim.first ); - - FREE(draw->verts); - draw->verts = NULL; -} - - -struct header_dword { - GLuint clipmask:12; - GLuint edgeflag:1; - GLuint pad:19; -}; - - -static void -build_vertex_headers( struct draw_context *draw, - struct vertex_buffer *VB ) -{ - if (draw->header.storage == NULL) { - draw->header.stride = sizeof(GLfloat); - draw->header.size = 1; - draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(GLfloat), 32 ); - draw->header.data = draw->header.storage; - draw->header.count = 0; - draw->header.flags = VEC_SIZE_1 | VEC_MALLOC; - } - - /* Build vertex header attribute. - * - */ - - { - GLuint i; - struct header_dword *header = (struct header_dword *)draw->header.storage; - - /* yes its a hack - */ - assert(sizeof(*header) == sizeof(GLfloat)); - - draw->header.count = VB->Count; - - if (VB->EdgeFlag) { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = VB->ClipMask[i]; - header[i].edgeflag = VB->EdgeFlag[i]; - header[i].pad = 0; - } - } - else if (VB->ClipOrMask) { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = VB->ClipMask[i]; - header[i].edgeflag = 0; - header[i].pad = 0; - } - } - else { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = 0; - header[i].edgeflag = 0; - header[i].pad = 0; - } - } - } - - VB->AttribPtr[VF_ATTRIB_VERTEX_HEADER] = &draw->header; -} - - - - - -static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) -{ - switch (mode) { - case GL_POINTS: - *first = 1; - *incr = 1; - return 0; - case GL_LINES: - *first = 2; - *incr = 2; - return 0; - case GL_LINE_STRIP: - *first = 2; - *incr = 1; - return 0; - case GL_LINE_LOOP: - *first = 2; - *incr = 1; - return 1; - case GL_TRIANGLES: - *first = 3; - *incr = 3; - return 0; - case GL_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - return 0; - case GL_TRIANGLE_FAN: - case GL_POLYGON: - *first = 3; - *incr = 1; - return 1; - case GL_QUADS: - *first = 4; - *incr = 4; - return 0; - case GL_QUAD_STRIP: - *first = 4; - *incr = 2; - return 0; - default: - assert(0); - *first = 1; - *incr = 1; - return 0; - } -} - - -static GLuint trim( GLuint count, GLuint first, GLuint incr ) -{ - if (count < first) - return 0; - else - return count - (count - first) % incr; -} - - -/* This is a hack & will all go away. - */ -void draw_vb(struct draw_context *draw, - struct vertex_buffer *VB ) -{ - GLuint i; - - VB->AttribPtr[VF_ATTRIB_POS] = VB->NdcPtr; - VB->AttribPtr[VF_ATTRIB_BFC0] = VB->ColorPtr[1]; - VB->AttribPtr[VF_ATTRIB_BFC1] = VB->SecondaryColorPtr[1]; - VB->AttribPtr[VF_ATTRIB_CLIP_POS] = VB->ClipPtr; - - /* Build vertex headers: - */ - build_vertex_headers( draw, VB ); - - draw->in_vb = 1; - - /* Allocate the vertices: - */ - draw_allocate_vertices( draw, VB->Count ); - - /* Bind the vb outputs: - */ - vf_set_sources( draw->vf, VB->AttribPtr, 0 ); - - /* Build the hardware or prim-pipe vertices: - */ - vf_emit_vertices( draw->vf, VB->Count, draw->verts ); - - - for (i = 0; i < VB->PrimitiveCount; i++) { - - GLenum mode = VB->Primitive[i].mode; - GLuint start = VB->Primitive[i].start; - GLuint length, first, incr; - - /* Trim the primitive down to a legal size. - */ - draw_prim_info( mode, &first, &incr ); - length = trim( VB->Primitive[i].count, first, incr ); - - if (!length) - continue; - - if (draw->prim != mode) - draw_set_prim( draw, mode ); - - if (VB->Elts) { - draw_indexed_prim( draw, - VB->Elts + start, - length ); - } - else { - draw_prim( draw, - start, - length ); - } - } - - draw_release_vertices( draw ); - draw->verts = NULL; - draw->in_vb = 0; -} - -void draw_set_viewport( struct draw_context *draw, - const GLfloat *scale, - const GLfloat *translate ) -{ - assert(!draw->in_vb); - vf_set_vp_scale_translate( draw->vf, scale, translate ); -} - - - -struct draw_context *draw_create( struct generic_context *generic ) -{ - struct draw_context *draw = CALLOC_STRUCT( draw_context ); - draw->generic = generic; - draw->vf = vf_create( GL_TRUE ); - - return draw; -} - - -void draw_destroy( struct draw_context *draw ) -{ - if (draw->header.storage) - ALIGN_FREE( draw->header.storage ); - - vf_destroy( draw->vf ); - - FREE( draw ); -} - -#define EMIT_ATTR( ATTR, STYLE ) \ -do { \ - draw->attrs[draw->nr_attrs].attrib = ATTR; \ - draw->attrs[draw->nr_attrs].format = STYLE; \ - draw->nr_attrs++; \ -} while (0) - - -void draw_set_vertex_attributes( struct draw_context *draw, - const GLuint *attrs, - GLuint nr_attrs ) -{ - GLuint i; - - draw->nr_attrs = 0; - - EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F); - EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F); - - assert(attrs[0] == VF_ATTRIB_POS); - EMIT_ATTR(attrs[0], EMIT_4F_VIEWPORT); - - for (i = 1; i < nr_attrs; i++) - EMIT_ATTR(attrs[i], EMIT_4F); - - draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); -} - - diff --git a/src/mesa/softpipe/generic/g_draw.h b/src/mesa/softpipe/generic/g_draw.h deleted file mode 100644 index 78c93b4df7..0000000000 --- a/src/mesa/softpipe/generic/g_draw.h +++ /dev/null @@ -1,56 +0,0 @@ - -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_DRAW_H -#define G_DRAW_H - -#include "glheader.h" -#include "softpipe/sp_state.h" - - -struct draw_context; - -struct draw_context *draw_create( struct generic_context *generic ); - -void draw_destroy( struct draw_context *draw ); - -void draw_set_viewport( struct draw_context *draw, - const GLfloat *scale, - const GLfloat *translate ); - -void draw_set_vertex_attributes( struct draw_context *draw, - const GLuint *attrs, - GLuint nr_attrs ); - -void draw_vb(struct draw_context *draw, - struct vertex_buffer *VB ); - -#endif diff --git a/src/mesa/softpipe/generic/g_headers.h b/src/mesa/softpipe/generic/g_headers.h deleted file mode 100644 index 96ff52a453..0000000000 --- a/src/mesa/softpipe/generic/g_headers.h +++ /dev/null @@ -1,116 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#ifndef G_HEADERS_H -#define G_HEADERS_H - -#define PRIM_POINT 1 -#define PRIM_LINE 2 -#define PRIM_TRI 3 - -struct prim_header { - GLfloat det; - struct vertex_header *v[3]; -}; - -/* Carry some useful information around with the vertices in the prim - * pipe. - */ -struct vertex_header { - GLuint clipmask:12; - GLuint edgeflag:1; - GLuint pad:19; - - GLfloat clip[4]; - - GLfloat data[][4]; /* Note variable size */ -}; - - - - - -/* The rasterizer generates 2x2 quads of fragment and feeds them to - * the current fp_machine (see below). - */ -#define QUAD_BOTTOM_LEFT 0 -#define QUAD_BOTTOM_RIGHT 1 -#define QUAD_TOP_LEFT 2 -#define QUAD_TOP_RIGHT 3 -#define QUAD_SIZE (2*2) - -#define MASK_BOTTOM_LEFT 0x1 -#define MASK_BOTTOM_RIGHT 0x2 -#define MASK_TOP_LEFT 0x4 -#define MASK_TOP_RIGHT 0x8 -#define MASK_ALL 0xf - - -#define NUM_CHANNELS 4 /* avoid confusion between 4 pixels and 4 channels */ - - -struct setup_coefficient { - GLfloat a0[NUM_CHANNELS]; /* in an xyzw layout */ - GLfloat dadx[NUM_CHANNELS]; - GLfloat dady[NUM_CHANNELS]; -}; - - - -/* Encodes everything we need to know about a 2x2 pixel block. Uses - * "Channel-Serial" or "SoA" layout. - * - * Will expand to include non-attribute things like AA coverage and - * maybe prefetched depth from the depth buffer. - */ -struct quad_header { - GLint x0; - GLint y0; - GLuint mask; - GLuint facing; /**< Front or back facing? */ - - struct { - GLfloat color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ - GLfloat depth[QUAD_SIZE]; - } outputs; - - const struct setup_coefficient *coef; - - const enum interp_mode *interp; /* XXX: this information should be - * encoded in fragment program DECL - * statements. */ - - GLuint nr_attrs; -}; - - - - - -#endif diff --git a/src/mesa/softpipe/generic/g_prim.h b/src/mesa/softpipe/generic/g_prim.h deleted file mode 100644 index e3b2c5ea2f..0000000000 --- a/src/mesa/softpipe/generic/g_prim.h +++ /dev/null @@ -1,90 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_PRIM_H -#define G_PRIM_H - -#include "glheader.h" -#include "g_headers.h" - -struct generic_context; - -struct prim_stage *prim_setup( struct generic_context *context ); -struct prim_stage *prim_unfilled( struct generic_context *context ); -struct prim_stage *prim_twoside( struct generic_context *context ); -struct prim_stage *prim_offset( struct generic_context *context ); -struct prim_stage *prim_clip( struct generic_context *context ); -struct prim_stage *prim_flatshade( struct generic_context *context ); -struct prim_stage *prim_cull( struct generic_context *context ); - - -/* Internal structs and helpers for the primitive clip/setup pipeline: - */ -struct prim_stage { - struct generic_context *generic; - - struct prim_stage *next; - - struct vertex_header **tmp; - GLuint nr_tmps; - - void (*begin)( struct prim_stage * ); - - void (*point)( struct prim_stage *, - struct prim_header * ); - - void (*line)( struct prim_stage *, - struct prim_header * ); - - void (*tri)( struct prim_stage *, - struct prim_header * ); - - void (*end)( struct prim_stage * ); -}; - - - -/* Get a writeable copy of a vertex: - */ -static INLINE struct vertex_header * -dup_vert( struct prim_stage *stage, - const struct vertex_header *vert, - GLuint idx ) -{ - struct vertex_header *tmp = stage->tmp[idx]; - memcpy(tmp, vert, stage->generic->prim.vertex_size ); - return tmp; -} - -void prim_free_tmps( struct prim_stage *stage ); -void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); - - -#endif diff --git a/src/mesa/softpipe/generic/g_prim_clip.c b/src/mesa/softpipe/generic/g_prim_clip.c deleted file mode 100644 index 742def9c7d..0000000000 --- a/src/mesa/softpipe/generic/g_prim_clip.c +++ /dev/null @@ -1,429 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "imports.h" -#include "macros.h" - -#include "g_context.h" -#include "g_prim.h" - -struct clipper { - struct prim_stage stage; - - GLuint active_user_planes; -}; - -/* This is a bit confusing: - */ -static INLINE struct clipper *clipper_stage( struct prim_stage *stage ) -{ - return (struct clipper *)stage; -} - - -#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) - - -/* All attributes are float[4], so this is easy: - */ -static void interp_attr( GLfloat *fdst, - GLfloat t, - const GLfloat *fin, - const GLfloat *fout ) -{ - fdst[0] = LINTERP( t, fout[0], fin[0] ); - fdst[1] = LINTERP( t, fout[1], fin[1] ); - fdst[2] = LINTERP( t, fout[2], fin[2] ); - fdst[3] = LINTERP( t, fout[3], fin[3] ); -} - - - - -/* Interpolate between two vertices to produce a third. - */ -static void interp( struct clipper *clip, - struct vertex_header *dst, - GLfloat t, - const struct vertex_header *out, - const struct vertex_header *in ) -{ - const GLuint nr_attrs = clip->stage.generic->nr_attrs; - GLuint j; - - /* Vertex header. - */ - { - dst->clipmask = 0; - dst->edgeflag = 0; - dst->pad = 0; - } - - /* Clip coordinates: interpolate normally - */ - { - interp_attr(dst->clip, t, in->clip, out->clip); - } - - /* Do the projective divide and insert window coordinates: - */ - { - const GLfloat *pos = dst->clip; - const GLfloat *scale = clip->stage.generic->viewport.scale; - const GLfloat *trans = clip->stage.generic->viewport.translate; - GLfloat oow; - - oow = 1.0 / pos[3]; - - dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; - dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; - dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; - dst->data[0][3] = oow; - } - - - /* Other attributes - */ - for (j = 1; j < nr_attrs-1; j++) { - interp_attr(dst->data[j], t, in->data[j], out->data[j]); - } -} - - -#define CLIP_USER_BIT 0x40 -#define CLIP_CULL_BIT 0x80 - - -static INLINE GLfloat dot4( const GLfloat *a, - const GLfloat *b ) -{ - GLfloat result = (a[0]*b[0] + - a[1]*b[1] + - a[2]*b[2] + - a[3]*b[3]); - - return result; -} - - -#if 0 -static INLINE void do_tri( struct prim_stage *next, - struct prim_header *header ) -{ - GLuint i; - for (i = 0; i < 3; i++) { - GLfloat *ndc = header->v[i]->data[0]; - _mesa_printf("ndc %f %f %f\n", ndc[0], ndc[1], ndc[2]); - assert(ndc[0] >= -1 && ndc[0] <= 641); - assert(ndc[1] >= 30 && ndc[1] <= 481); - } - _mesa_printf("\n"); - next->tri(next, header); -} -#endif - - -static void emit_poly( struct prim_stage *stage, - struct vertex_header **inlist, - GLuint n ) -{ - struct prim_header header; - GLuint i; - - for (i = 2; i < n; i++) { - header.v[0] = inlist[0]; - header.v[1] = inlist[i-1]; - header.v[2] = inlist[i]; - - { - GLuint tmp0 = header.v[0]->edgeflag; - GLuint tmp2 = header.v[2]->edgeflag; - - if (i != 2) header.v[0]->edgeflag = 0; - if (i != n-1) header.v[2]->edgeflag = 0; - - stage->next->tri( stage->next, &header ); - - header.v[0]->edgeflag = tmp0; - header.v[2]->edgeflag = tmp2; - } - } -} - - -#if 0 -static void emit_poly( struct prim_stage *stage ) -{ - GLuint i; - - for (i = 2; i < n; i++) { - header->v[0] = inlist[0]; - header->v[1] = inlist[i-1]; - header->v[2] = inlist[i]; - - stage->next->tri( stage->next, header ); - } -} -#endif - - -/* Clip a triangle against the viewport and user clip planes. - */ -static void -do_clip_tri( struct prim_stage *stage, - struct prim_header *header, - GLuint clipmask ) -{ - struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *a[MAX_CLIPPED_VERTICES]; - struct vertex_header *b[MAX_CLIPPED_VERTICES]; - struct vertex_header **inlist = a; - struct vertex_header **outlist = b; - GLuint tmpnr = 0; - GLuint n = 3; - GLuint i; - - inlist[0] = header->v[0]; - inlist[1] = header->v[1]; - inlist[2] = header->v[2]; - - /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove - * this once we correctly use 16bit masks for userclip planes. - */ - clipmask &= ~CLIP_CULL_BIT; - if (clipmask & CLIP_USER_BIT) { - clipmask &= ~CLIP_USER_BIT; - clipmask |= clipper->active_user_planes; - } - - while (clipmask && n >= 3) { - GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->stage.generic->plane[plane_idx]; - struct vertex_header *vert_prev = inlist[0]; - GLfloat dp_prev = dot4( vert_prev->clip, plane ); - GLuint outcount = 0; - - clipmask &= ~(1<clip, plane ); - - if (!IS_NEGATIVE(dp_prev)) { - outlist[outcount++] = vert_prev; - } - - if (DIFFERENT_SIGNS(dp, dp_prev)) { - struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; - outlist[outcount++] = new_vert; - - if (IS_NEGATIVE(dp)) { - /* Going out of bounds. Avoid division by zero as we - * know dp != dp_prev from DIFFERENT_SIGNS, above. - */ - GLfloat t = dp / (dp - dp_prev); - interp( clipper, new_vert, t, vert, vert_prev ); - - /* Force edgeflag true in this case: - */ - new_vert->edgeflag = 1; - } else { - /* Coming back in. - */ - GLfloat t = dp_prev / (dp_prev - dp); - interp( clipper, new_vert, t, vert_prev, vert ); - - /* Copy starting vert's edgeflag: - */ - new_vert->edgeflag = vert_prev->edgeflag; - } - } - - vert_prev = vert; - dp_prev = dp; - } - - { - struct vertex_header **tmp = inlist; - inlist = outlist; - outlist = tmp; - n = outcount; - } - } - - /* Emit the polygon as triangles to the setup stage: - */ - if (n >= 3) - emit_poly( stage, inlist, n ); -} - - -/* Clip a line against the viewport and user clip planes. - */ -static void -do_clip_line( struct prim_stage *stage, - struct prim_header *header, - GLuint clipmask ) -{ - struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - const GLfloat *pos0 = v0->clip; - const GLfloat *pos1 = v1->clip; - GLfloat t0 = 0; - GLfloat t1 = 0; - - /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove - * this once we correctly use 16bit masks for userclip planes. - */ - clipmask &= ~CLIP_CULL_BIT; - if (clipmask & CLIP_USER_BIT) { - clipmask &= ~CLIP_USER_BIT; - clipmask |= clipper->active_user_planes; - } - - while (clipmask) { - GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->stage.generic->plane[plane_idx]; - - clipmask &= ~(1< t1) t1 = t; - } - - if (dp0 < 0) { - GLfloat t = dp0 / (dp0 - dp1); - if (t > t0) t0 = t; - } - - if (t0 + t1 >= 1.0) - return; /* discard */ - } - - if (v0->clipmask) { - interp( clipper, stage->tmp[0], t0, v0, v1 ); - header->v[0] = stage->tmp[0]; - } - - if (v1->clipmask) { - interp( clipper, stage->tmp[1], t1, v1, v0 ); - header->v[1] = stage->tmp[1]; - } - - stage->next->line( stage->next, header ); -} - - - -static void clip_begin( struct prim_stage *stage ) -{ - struct clipper *clipper = clipper_stage(stage); - GLuint nr = stage->generic->nr_planes; - - /* Hacky bitmask to use when we hit CLIP_USER_BIT: - */ - clipper->active_user_planes = ((1<next->begin( stage->next ); -} - -static void -clip_point( struct prim_stage *stage, - struct prim_header *header ) -{ - if (header->v[0]->clipmask == 0) - stage->next->point( stage->next, header ); -} - - -static void -clip_line( struct prim_stage *stage, - struct prim_header *header ) -{ - GLuint clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask); - - if (clipmask == 0) { - stage->next->line( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask) == 0) { - do_clip_line(stage, header, clipmask); - } -} - - -static void -clip_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - GLuint clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask | - header->v[2]->clipmask); - - if (clipmask == 0) { - stage->next->tri( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask & - header->v[2]->clipmask) == 0) { - do_clip_tri(stage, header, clipmask); - } -} - -static void clip_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - - -struct prim_stage *prim_clip( struct generic_context *generic ) -{ - struct clipper *clipper = CALLOC_STRUCT(clipper); - - prim_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); - - clipper->stage.generic = generic; - clipper->stage.begin = clip_begin; - clipper->stage.point = clip_point; - clipper->stage.line = clip_line; - clipper->stage.tri = clip_tri; - clipper->stage.end = clip_end; - - return &clipper->stage; -} diff --git a/src/mesa/softpipe/generic/g_prim_cull.c b/src/mesa/softpipe/generic/g_prim_cull.c deleted file mode 100644 index 633fbcc492..0000000000 --- a/src/mesa/softpipe/generic/g_prim_cull.c +++ /dev/null @@ -1,119 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_prim.h" - - - -struct cull_stage { - struct prim_stage stage; - GLuint mode; -}; - - - -static INLINE struct cull_stage *cull_stage( struct prim_stage *stage ) -{ - return (struct cull_stage *)stage; -} - - -static void cull_begin( struct prim_stage *stage ) -{ - struct cull_stage *cull = cull_stage(stage); - - cull->mode = stage->generic->setup.cull_mode; - - stage->next->begin( stage->next ); -} - - -static void cull_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - /* Window coords: */ - GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); - GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); - GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); - - GLfloat ex = v0[0] - v2[0]; - GLfloat ey = v0[1] - v2[1]; - GLfloat fx = v1[0] - v2[0]; - GLfloat fy = v1[1] - v2[1]; - - header->det = ex * fy - ey * fx; - - _mesa_printf("%s %f\n", __FUNCTION__, header->det ); - - if (header->det != 0) { - GLuint mode = (header->det < 0) ? WINDING_CW : WINDING_CCW; - - if ((mode & cull_stage(stage)->mode) == 0) - stage->next->tri( stage->next, header ); - } -} - - -static void cull_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void cull_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void cull_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_cull( struct generic_context *generic ) -{ - struct cull_stage *cull = CALLOC_STRUCT(cull_stage); - - prim_alloc_tmps( &cull->stage, 0 ); - - cull->stage.generic = generic; - cull->stage.next = NULL; - cull->stage.begin = cull_begin; - cull->stage.point = cull_point; - cull->stage.line = cull_line; - cull->stage.tri = cull_tri; - cull->stage.end = cull_end; - - return &cull->stage; -} diff --git a/src/mesa/softpipe/generic/g_prim_flatshade.c b/src/mesa/softpipe/generic/g_prim_flatshade.c deleted file mode 100644 index 1974a64fa8..0000000000 --- a/src/mesa/softpipe/generic/g_prim_flatshade.c +++ /dev/null @@ -1,149 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_prim.h" - - -struct flatshade_stage { - struct prim_stage stage; - - const GLuint *lookup; -}; - - - -static INLINE struct flatshade_stage *flatshade_stage( struct prim_stage *stage ) -{ - return (struct flatshade_stage *)stage; -} - - -static void flatshade_begin( struct prim_stage *stage ) -{ - stage->next->begin( stage->next ); -} - - - -static INLINE void copy_attr( GLuint attr, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - if (attr) { - memcpy( dst->data[attr], - src->data[attr], - sizeof(src->data[0]) ); - } -} - -static void copy_colors( struct prim_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - struct flatshade_stage *flatshade = flatshade_stage(stage); - const GLuint *lookup = flatshade->lookup; - - copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); - copy_attr( lookup[VF_ATTRIB_COLOR1], dst, src ); - copy_attr( lookup[VF_ATTRIB_BFC0], dst, src ); - copy_attr( lookup[VF_ATTRIB_BFC1], dst, src ); -} - - - -/* Flatshade tri. Required for clipping and when unfilled tris are - * active, otherwise handled by hardware. - */ -static void flatshade_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = header->v[2]; - - copy_colors(stage, tmp.v[0], tmp.v[2]); - copy_colors(stage, tmp.v[1], tmp.v[2]); - - stage->next->tri( stage->next, &tmp ); -} - - -/* Flatshade line. Required for clipping. - */ -static void flatshade_line( struct prim_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = header->v[1]; - - copy_colors(stage, tmp.v[0], tmp.v[1]); - - stage->next->line( stage->next, &tmp ); -} - - -static void flatshade_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void flatshade_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_flatshade( struct generic_context *generic ) -{ - struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); - - prim_alloc_tmps( &flatshade->stage, 2 ); - - flatshade->stage.generic = generic; - flatshade->stage.next = NULL; - flatshade->stage.begin = flatshade_begin; - flatshade->stage.point = flatshade_point; - flatshade->stage.line = flatshade_line; - flatshade->stage.tri = flatshade_tri; - flatshade->stage.end = flatshade_end; - - flatshade->lookup = generic->vf_attr_to_slot; - - return &flatshade->stage; -} - - diff --git a/src/mesa/softpipe/generic/g_prim_offset.c b/src/mesa/softpipe/generic/g_prim_offset.c deleted file mode 100644 index 801ce9a441..0000000000 --- a/src/mesa/softpipe/generic/g_prim_offset.c +++ /dev/null @@ -1,157 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" -#include "macros.h" - -#include "g_context.h" -#include "g_prim.h" - - - -struct offset_stage { - struct prim_stage stage; - - GLuint hw_data_offset; - - GLfloat scale; - GLfloat units; -}; - - - -static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) -{ - return (struct offset_stage *)stage; -} - - -static void offset_begin( struct prim_stage *stage ) -{ - struct offset_stage *offset = offset_stage(stage); - - offset->units = stage->generic->setup.offset_units; - offset->scale = stage->generic->setup.offset_scale; - - stage->next->begin( stage->next ); -} - - -/* Offset tri. Some hardware can handle this, but not usually when - * doing unfilled rendering. - */ -static void do_offset_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct offset_stage *offset = offset_stage(stage); - GLfloat inv_det = 1.0 / header->det; - - /* Window coords: - */ - GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); - GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); - GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); - - GLfloat ex = v0[0] - v2[2]; - GLfloat fx = v1[0] - v2[2]; - GLfloat ey = v0[1] - v2[2]; - GLfloat fy = v1[1] - v2[2]; - GLfloat ez = v0[2] - v2[2]; - GLfloat fz = v1[2] - v2[2]; - - GLfloat a = ey*fz - ez*fy; - GLfloat b = ez*fx - ex*fz; - - GLfloat ac = a * inv_det; - GLfloat bc = b * inv_det; - GLfloat zoffset; - - if ( ac < 0.0f ) ac = -ac; - if ( bc < 0.0f ) bc = -bc; - - zoffset = offset->units + MAX2( ac, bc ) * offset->scale; - - v0[2] += zoffset; - v1[2] += zoffset; - v2[2] += zoffset; - - stage->next->tri( stage->next, header ); -} - - -static void offset_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = dup_vert(stage, header->v[2], 2); - - do_offset_tri( stage->next, &tmp ); -} - - - -static void offset_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void offset_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void offset_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_offset( struct generic_context *generic ) -{ - struct offset_stage *offset = CALLOC_STRUCT(offset_stage); - - prim_alloc_tmps( &offset->stage, 3 ); - - offset->stage.generic = generic; - offset->stage.next = NULL; - offset->stage.begin = offset_begin; - offset->stage.point = offset_point; - offset->stage.line = offset_line; - offset->stage.tri = offset_tri; - offset->stage.end = offset_end; - - return &offset->stage; -} diff --git a/src/mesa/softpipe/generic/g_prim_setup.c b/src/mesa/softpipe/generic/g_prim_setup.c deleted file mode 100644 index 41e1381a60..0000000000 --- a/src/mesa/softpipe/generic/g_prim_setup.c +++ /dev/null @@ -1,888 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "imports.h" -#include "macros.h" - -#include "g_context.h" -#include "g_prim.h" -#include "g_tile.h" - - -/** - * Triangle edge info - */ -struct edge { - GLfloat dx; /* X(v1) - X(v0), used only during setup */ - GLfloat dy; /* Y(v1) - Y(v0), used only during setup */ - GLfloat dxdy; /* dx/dy */ - GLfloat sx; /* first sample point x coord */ - GLfloat sy; - GLint lines; /* number of lines on this edge */ -}; - - -/** - * Triangle setup info (derived from prim_stage). - * Also used for line drawing (taking some liberties). - */ -struct setup_stage { - struct prim_stage stage; /**< This must be first */ - - /* Vertices are just an array of floats making up each attribute in - * turn. Currently fixed at 4 floats, but should change in time. - * Codegen will help cope with this. - */ - const struct vertex_header *vmax; - const struct vertex_header *vmid; - const struct vertex_header *vmin; - const struct vertex_header *vprovoke; - - struct edge ebot; - struct edge etop; - struct edge emaj; - - GLfloat oneoverarea; - - struct setup_coefficient coef[FRAG_ATTRIB_MAX]; - struct quad_header quad; - - struct { - GLint left[2]; /**< [0] = row0, [1] = row1 */ - GLint right[2]; - GLint y; - GLuint y_flags; - GLuint mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ - } span; -}; - - - -/** - * Basically a cast wrapper. - */ -static inline struct setup_stage *setup_stage( struct prim_stage *stage ) -{ - return (struct setup_stage *)stage; -} - - -/** - * Given an X or Y coordinate, return the block/quad coordinate that it - * belongs to. - */ -static inline GLint block( GLint x ) -{ - return x & ~1; -} - - - -static void setup_begin( struct prim_stage *stage ) -{ - setup_stage(stage)->quad.nr_attrs = stage->generic->nr_frag_attrs; -} - - -/** - * Run shader on a quad/block. - */ -static void run_shader_block( struct setup_stage *setup, - GLint x, GLint y, GLuint mask ) -{ - setup->quad.x0 = x; - setup->quad.y0 = y; - setup->quad.mask = mask; - - quad_shade( setup->stage.generic, &setup->quad ); -} - - -/** - * Compute mask which indicates which pixels in the 2x2 quad are actually inside - * the triangle's bounds. - * - * this is pretty nasty... may need to rework flush_spans again to - * fix it, if possible. - */ -static GLuint calculate_mask( struct setup_stage *setup, - GLint x ) -{ - GLuint mask = 0; - - if (x >= setup->span.left[0] && x < setup->span.right[0]) - mask |= MASK_BOTTOM_LEFT; - - if (x >= setup->span.left[1] && x < setup->span.right[1]) - mask |= MASK_TOP_LEFT; - - if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) - mask |= MASK_BOTTOM_RIGHT; - - if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) - mask |= MASK_TOP_RIGHT; - - return mask; -} - - -/** - * Render a horizontal span of quads - */ -static void flush_spans( struct setup_stage *setup ) -{ - GLint minleft, maxright; - GLint x; - - switch (setup->span.y_flags) { - case 3: - minleft = MIN2(setup->span.left[0], setup->span.left[1]); - maxright = MAX2(setup->span.right[0], setup->span.right[1]); - break; - - case 1: - minleft = setup->span.left[0]; - maxright = setup->span.right[0]; - break; - - case 2: - minleft = setup->span.left[1]; - maxright = setup->span.right[1]; - break; - - default: - return; - } - - - for (x = block(minleft); x <= block(maxright); ) - { - run_shader_block( setup, x, - setup->span.y, - calculate_mask( setup, x ) ); - x += 2; - } - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; -} - - -static GLboolean setup_sort_vertices( struct setup_stage *setup, - const struct prim_header *prim ) -{ - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - const struct vertex_header *v2 = prim->v[2]; - - setup->vprovoke = v2; - - /* determine bottom to top order of vertices */ - { - GLfloat y0 = v0->data[0][1]; - GLfloat y1 = v1->data[0][1]; - GLfloat y2 = v2->data[0][1]; - if (y0 <= y1) { - if (y1 <= y2) { - /* y0<=y1<=y2 */ - setup->vmin = v0; - setup->vmid = v1; - setup->vmax = v2; - } - else if (y2 <= y0) { - /* y2<=y0<=y1 */ - setup->vmin = v2; - setup->vmid = v0; - setup->vmax = v1; - } - else { - /* y0<=y2<=y1 */ - setup->vmin = v0; - setup->vmid = v2; - setup->vmax = v1; - } - } - else { - if (y0 <= y2) { - /* y1<=y0<=y2 */ - setup->vmin = v1; - setup->vmid = v0; - setup->vmax = v2; - } - else if (y2 <= y1) { - /* y2<=y1<=y0 */ - setup->vmin = v2; - setup->vmid = v1; - setup->vmax = v0; - } - else { - /* y1<=y2<=y0 */ - setup->vmin = v1; - setup->vmid = v2; - setup->vmax = v0; - } - } - } - - setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; - setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; - setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; - - /* xxx: may need to adjust this sign according to the if-tree - * above: - * - * XXX: this is like 'det', but calculated from screen coords?? - */ - { - const GLfloat area = (setup->emaj.dx * setup->ebot.dy - - setup->ebot.dx * setup->emaj.dy); - - setup->oneoverarea = 1.0 / area; - } - - /* XXX need to know if this is a front or back-facing triangle: - * - the GLSL gl_FrontFacing fragment attribute (bool) - * - two-sided stencil test - */ - setup->quad.facing = 0; - - _mesa_printf("%s one-over-area %f\n", __FUNCTION__, setup->oneoverarea ); - - - return GL_TRUE; -} - - -/** - * Compute a0 for a constant-valued coefficient (GL_FLAT shading). - */ -static void const_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i ) -{ - setup->coef[slot].dadx[i] = 0; - setup->coef[slot].dady[i] = 0; - - /* need provoking vertex info! - */ - setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; -} - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a triangle. - */ -static void tri_linear_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i) -{ - GLfloat botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; - GLfloat majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; - GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; - - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; - - /* calculate a0 as the value which would be sampled for the - * fragment at (0,0), taking into account that we want to sample at - * pixel centers, in other words (0.5, 0.5). - * - * this is neat but unfortunately not a good way to do things for - * triangles with very large values of dadx or dady as it will - * result in the subtraction and re-addition from a0 of a very - * large number, which means we'll end up loosing a lot of the - * fractional bits and precision from a0. the way to fix this is - * to define a0 as the sample at a pixel center somewhere near vmin - * instead - i'll switch to this later. - */ - setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); - - _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", - slot, "xyzw"[i], - setup->coef[slot].a0[i], - setup->coef[slot].dadx[i], - setup->coef[slot].dady[i]); -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a triangle. - */ -static void tri_persp_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i ) -{ - /* premultiply by 1/w: - */ - GLfloat mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; - GLfloat mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; - GLfloat maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; - - GLfloat botda = mida - mina; - GLfloat majda = maxa - mina; - GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; - GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; - - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; - setup->coef[slot].a0[i] = (mina - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); -} - - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. - */ -static void setup_tri_coefficients( struct setup_stage *setup ) -{ - const enum interp_mode *interp = setup->stage.generic->interp; - GLuint slot, j; - - /* z and w are done by linear interpolation: - */ - tri_linear_coeff(setup, 0, 2); - tri_linear_coeff(setup, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - switch (interp[slot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - tri_linear_coeff(setup, slot, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - tri_persp_coeff(setup, slot, j); - break; - } - } -} - - - -static void setup_tri_edges( struct setup_stage *setup ) -{ - GLfloat vmin_x = setup->vmin->data[0][0] + 0.5; - GLfloat vmid_x = setup->vmid->data[0][0] + 0.5; - - GLfloat vmin_y = setup->vmin->data[0][1] - 0.5; - GLfloat vmid_y = setup->vmid->data[0][1] - 0.5; - GLfloat vmax_y = setup->vmax->data[0][1] - 0.5; - - setup->emaj.sy = ceilf(vmin_y); - setup->emaj.lines = (GLint) ceilf(vmax_y - setup->emaj.sy); - setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; - setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; - - setup->etop.sy = ceilf(vmid_y); - setup->etop.lines = (GLint) ceilf(vmax_y - setup->etop.sy); - setup->etop.dxdy = setup->etop.dx / setup->etop.dy; - setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; - - setup->ebot.sy = ceilf(vmin_y); - setup->ebot.lines = (GLint) ceilf(vmid_y - setup->ebot.sy); - setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; - setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; -} - - -/** - * Render the upper or lower half of a triangle. - * Scissoring is applied here too. - */ -static void subtriangle( struct setup_stage *setup, - struct edge *eleft, - struct edge *eright, - GLuint lines ) -{ - GLint y, start_y, finish_y; - GLint sy = (GLint)eleft->sy; - - assert((GLint)eleft->sy == (GLint) eright->sy); - assert((GLint)eleft->sy >= 0); /* catch bug in x64? */ - - /* scissor y: - */ - if (setup->stage.generic->setup.scissor) { - start_y = sy; - finish_y = start_y + lines; - - if (start_y < setup->stage.generic->scissor.miny) - start_y = setup->stage.generic->scissor.miny; - - if (finish_y > setup->stage.generic->scissor.maxy) - finish_y = setup->stage.generic->scissor.maxy; - - start_y -= sy; - finish_y -= sy; - } - else { - start_y = 0; - finish_y = lines; - } - - _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); - - for (y = start_y; y < finish_y; y++) { - - /* avoid accumulating adds as floats don't have the precision to - * accurately iterate large triangle edges that way. luckily we - * can just multiply these days. - * - * this is all drowned out by the attribute interpolation anyway. - */ - GLint left = (GLint)(eleft->sx + y * eleft->dxdy); - GLint right = (GLint)(eright->sx + y * eright->dxdy); - - /* scissor x: - */ - if (setup->stage.generic->setup.scissor) { - if (left < setup->stage.generic->scissor.minx) - left = setup->stage.generic->scissor.minx; - - if (right > setup->stage.generic->scissor.maxx) - right = setup->stage.generic->scissor.maxx; - } - - if (left < right) { - GLint _y = sy+y; - if (block(_y) != setup->span.y) { - flush_spans(setup); - setup->span.y = block(_y); - } - - setup->span.left[_y&1] = left; - setup->span.right[_y&1] = right; - setup->span.y_flags |= 1<<(_y&1); - } - } - - - /* save the values so that emaj can be restarted: - */ - eleft->sx += lines * eleft->dxdy; - eright->sx += lines * eright->dxdy; - eleft->sy += lines; - eright->sy += lines; -} - - -/** - * Do setup for triangle rasterization, then render the triangle. - */ -static void setup_tri( struct prim_stage *stage, - struct prim_header *prim ) -{ - struct setup_stage *setup = setup_stage( stage ); - - _mesa_printf("%s\n", __FUNCTION__ ); - - setup_sort_vertices( setup, prim ); - setup_tri_coefficients( setup ); - setup_tri_edges( setup ); - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; -// setup->span.z_mode = tri_z_mode( setup->ctx ); - -// init_constant_attribs( setup ); - - if (setup->oneoverarea < 0.0) { - /* emaj on left: - */ - subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); - subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); - } - else { - /* emaj on right: - */ - subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); - subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); - } - - flush_spans( setup ); -} - - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a line. - */ -static void -line_linear_coeff(struct setup_stage *setup, GLuint slot, GLuint i) -{ - const GLfloat dz = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - const GLfloat dadx = dz * setup->emaj.dx * setup->oneoverarea; - const GLfloat dady = dz * setup->emaj.dy * setup->oneoverarea; - setup->coef[slot].dadx[i] = dadx; - setup->coef[slot].dady[i] = dady; - setup->coef[slot].a0[i] - = (setup->vmin->data[slot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5) + - dady * (setup->vmin->data[0][1] - 0.5))); -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a line. - */ -static void -line_persp_coeff(struct setup_stage *setup, GLuint slot, GLuint i) -{ - /* XXX to do */ - line_linear_coeff(setup, slot, i); /* XXX temporary */ -} - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmax are initialized. - */ -static INLINE void -setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) -{ - const enum interp_mode *interp = setup->stage.generic->interp; - GLuint slot, j; - - /* use setup->vmin, vmax to point to vertices */ - setup->vprovoke = prim->v[1]; - setup->vmin = prim->v[0]; - setup->vmax = prim->v[1]; - - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - /* NOTE: this is not really 1/area */ - setup->oneoverarea = 1.0 / (setup->emaj.dx * setup->emaj.dx + - setup->emaj.dy * setup->emaj.dy); - - /* z and w are done by linear interpolation: - */ - line_linear_coeff(setup, 0, 2); - line_linear_coeff(setup, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - switch (interp[slot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - line_linear_coeff(setup, slot, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - line_persp_coeff(setup, slot, j); - break; - } - } -} - - -/** - * Plot a pixel in a line segment. - */ -static INLINE void -plot(struct setup_stage *setup, GLint x, GLint y) -{ - const GLint iy = y & 1; - const GLint ix = x & 1; - const GLint quadX = x - ix; - const GLint quadY = y - iy; - const GLint mask = (1 << ix) << (2 * iy); - - if (quadX != setup->quad.x0 || - quadY != setup->quad.y0) - { - /* flush prev quad, start new quad */ - - if (setup->quad.x0 != -1) - quad_shade(setup->stage.generic, &setup->quad); - - setup->quad.x0 = quadX; - setup->quad.y0 = quadY; - setup->quad.mask = 0x0; - } - - setup->quad.mask |= mask; -} - - - -/** - * Do setup for line rasterization, then render the line. - * XXX single-pixel width, no stipple, etc - * XXX no scissoring yet. - */ -static void -setup_line(struct prim_stage *stage, struct prim_header *prim) -{ - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - struct setup_stage *setup = setup_stage( stage ); - - GLint x0 = (GLint) v0->data[0][0]; - GLint x1 = (GLint) v1->data[0][0]; - GLint y0 = (GLint) v0->data[0][1]; - GLint y1 = (GLint) v1->data[0][1]; - GLint dx = x1 - x0; - GLint dy = y1 - y0; - GLint xstep, ystep; - - if (dx == 0 && dy == 0) - return; - - setup_line_coefficients(setup, prim); - - if (dx < 0) { - dx = -dx; /* make positive */ - xstep = -1; - } - else { - xstep = 1; - } - - if (dy < 0) { - dy = -dy; /* make positive */ - ystep = -1; - } - else { - ystep = 1; - } - - assert(dx >= 0); - assert(dy >= 0); - - setup->quad.x0 = setup->quad.y0 = -1; - setup->quad.mask = 0x0; - - if (dx > dy) { - /*** X-major line ***/ - GLint i; - const GLint errorInc = dy + dy; - GLint error = errorInc - dx; - const GLint errorDec = error - dx; - - for (i = 0; i < dx; i++) { - plot(setup, x0, y0); - - x0 += xstep; - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - y0 += ystep; - } - } - } - else { - /*** Y-major line ***/ - GLint i; - const GLint errorInc = dx + dx; - GLint error = errorInc - dy; - const GLint errorDec = error - dy; - - for (i = 0; i < dy; i++) { - plot(setup, x0, y0); - - y0 += ystep; - - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - x0 += xstep; - } - } - } - - /* draw final quad */ - if (setup->quad.mask) { - quad_shade(setup->stage.generic, &setup->quad); - } -} - - -/** - * Do setup for point rasterization, then render the point. - * Round or square points... - * XXX could optimize a lot for 1-pixel points. - */ -static void -setup_point(struct prim_stage *stage, struct prim_header *prim) -{ - struct setup_stage *setup = setup_stage( stage ); - GLfloat halfSize = 7.3; /*XXX this is a vertex attrib */ - GLfloat halfSizeSquared = halfSize * halfSize; - const struct vertex_header *v0 = prim->v[0]; - const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; - const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; - const GLint ixmin = block((GLint) (x - halfSize)); - const GLint ixmax = block((GLint) (x + halfSize)); - const GLint iymin = block((GLint) (y - halfSize)); - const GLint iymax = block((GLint) (y + halfSize)); - GLboolean round = GL_TRUE; - GLint ix, iy; - GLuint slot, j; - - /* For points, all interpolants are constant-valued. - * However, for point sprites, we'll need to setup texcoords appropriately. - * XXX: which coefficients are the texcoords??? - * We may do point sprites as textured quads... - * - * KW: We don't know which coefficients are texcoords - ultimately - * the choice of what interpolation mode to use for each attribute - * should be determined by the fragment program, using - * per-attribute declaration statements that include interpolation - * mode as a parameter. So either the fragment program will have - * to be adjusted for pointsprite vs normal point behaviour, or - * otherwise a special interpolation mode will have to be defined - * which matches the required behaviour for point sprites. But - - * the latter is not a feature of normal hardware, and as such - * probably should be ruled out on that basis. - */ - setup->vprovoke = prim->v[0]; - const_coeff(setup, 0, 2); - const_coeff(setup, 0, 3); - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - } - - /* XXX need to clip against scissor bounds too */ - - for (iy = iymin; iy <= iymax; iy += 2) { - for (ix = ixmin; ix <= ixmax; ix += 2) { - - if (round) { - /* rounded points */ - /* XXX for GL_SMOOTH, need to compute per-fragment coverage too */ - GLfloat dx, dy; - - setup->quad.mask = 0x0; - - dx = (ix + 0.5) - x; - dy = (iy + 0.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_BOTTOM_LEFT; - - dx = (ix + 1.5) - x; - dy = (iy + 0.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_BOTTOM_RIGHT; - - dx = (ix + 0.5) - x; - dy = (iy + 1.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_TOP_LEFT; - - dx = (ix + 1.5) - x; - dy = (iy + 1.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_TOP_RIGHT; - } - else { - /* square points */ - setup->quad.mask = 0xf; - - if (ix + 0.5 < x - halfSize) - setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - - if (ix + 1.5 > x + halfSize) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - - if (iy + 0.5 < y - halfSize) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); - - if (iy + 1.5 > y + halfSize) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - } - - if (setup->quad.mask) { - setup->quad.x0 = ix; - setup->quad.y0 = iy; - quad_shade( setup->stage.generic, &setup->quad ); - } - } - } -} - - - -static void setup_end( struct prim_stage *stage ) -{ -} - - -struct prim_stage *prim_setup( struct generic_context *generic ) -{ - struct setup_stage *setup = CALLOC_STRUCT(setup_stage); - - setup->stage.generic = generic; - setup->stage.begin = setup_begin; - setup->stage.point = setup_point; - setup->stage.line = setup_line; - setup->stage.tri = setup_tri; - setup->stage.end = setup_end; - - setup->quad.coef = setup->coef; - - return &setup->stage; -} diff --git a/src/mesa/softpipe/generic/g_prim_setup.h b/src/mesa/softpipe/generic/g_prim_setup.h deleted file mode 100644 index 40a70c543e..0000000000 --- a/src/mesa/softpipe/generic/g_prim_setup.h +++ /dev/null @@ -1,121 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef S_TRI_H -#define S_TRI_H - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "glheader.h" -#include "imports.h" -#include "s_tri_public.h" -#include "s_context.h" - - - -struct tri_context; -struct fp_context; -struct be_context; - -/* Note the rasterizer does not take a GLcontext argument. This is - * deliberate. - */ -struct tri_context *tri_create_context( GLcontext *ctx ); - -void tri_destroy_context( struct tri_context *tri ); - -void tri_set_fp_context( struct tri_context *tri, - struct fp_context *fp, - void (*fp_run)( struct fp_context *fp, - const struct fp_inputs *, - struct fp_outputs * )); - - -void tri_set_be_context( struct tri_context *tri, - struct be_context *be, - void (*be_run)( struct be_context *be, - const struct fp_outputs * )); - -void tri_set_attribs( struct tri_context *tri, - const struct attr_info *info, - GLuint nr_attrib ); - -void tri_set_backface( struct tri_context *tri, - GLfloat backface ); - -void tri_set_scissor( struct tri_context *tri, - GLint x, - GLint y, - GLuint width, - GLuint height, - GLboolean enabled ); - -void tri_set_stipple( struct tri_context *tri, - const GLuint *pattern, - GLboolean enabled ); - -/* Unfilled triangles will be handled elsewhere (higher in the - * pipeline), as will things like stipple (lower in the pipeline). - */ - -void tri_triangle( struct tri_context *tri, - const struct vertex *v0, - const struct vertex *v1, - const struct vertex *v2 ); - -/* TODO: rasterize_line, rasterize_point?? - * How will linestipple work? - */ - - -#ifdef SETUP_PRIVATE - - - - - - -GLboolean tri_setup( struct tri_context *tri, - const struct vertex *v0, - const struct vertex *v1, - const struct vertex *v2 ); - -void tri_rasterize( struct tri_context *tri ); -void tri_rasterize_spans( struct tri_context *tri ); - - - - - - -#endif -#endif diff --git a/src/mesa/softpipe/generic/g_prim_twoside.c b/src/mesa/softpipe/generic/g_prim_twoside.c deleted file mode 100644 index 334ae2f537..0000000000 --- a/src/mesa/softpipe/generic/g_prim_twoside.c +++ /dev/null @@ -1,149 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" -#include "g_context.h" -#include "g_prim.h" - - -struct twoside_stage { - struct prim_stage stage; - - GLfloat facing; - const GLuint *lookup; -}; - - -static INLINE struct twoside_stage *twoside_stage( struct prim_stage *stage ) -{ - return (struct twoside_stage *)stage; -} - - -static void twoside_begin( struct prim_stage *stage ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - - twoside->facing = (stage->generic->setup.front_winding == WINDING_CW) ? -1 : 1; - - stage->next->begin( stage->next ); -} - - -static INLINE void copy_color( GLuint attr_dst, - GLuint attr_src, - struct vertex_header *v ) -{ - if (attr_dst && attr_src) { - memcpy( v->data[attr_dst], - v->data[attr_src], - sizeof(v->data[0]) ); - } -} - - -static struct vertex_header *copy_bfc( struct twoside_stage *twoside, - const struct vertex_header *v, - GLuint idx ) -{ - struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); - - copy_color( twoside->lookup[VF_ATTRIB_COLOR0], - twoside->lookup[VF_ATTRIB_BFC0], - tmp ); - - copy_color( twoside->lookup[VF_ATTRIB_COLOR1], - twoside->lookup[VF_ATTRIB_BFC1], - tmp ); - - return tmp; -} - - -/* Twoside tri: - */ -static void twoside_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - - if (header->det * twoside->facing < 0) { - struct prim_header tmp; - - tmp.det = header->det; - tmp.v[0] = copy_bfc(twoside, header->v[0], 0); - tmp.v[1] = copy_bfc(twoside, header->v[1], 1); - tmp.v[2] = copy_bfc(twoside, header->v[2], 2); - - stage->next->tri( stage->next, &tmp ); - } - else { - stage->next->tri( stage->next, header ); - } -} - - -static void twoside_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void twoside_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void twoside_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - - - -struct prim_stage *prim_twoside( struct generic_context *generic ) -{ - struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); - - prim_alloc_tmps( &twoside->stage, 3 ); - - twoside->stage.generic = generic; - twoside->stage.next = NULL; - twoside->stage.begin = twoside_begin; - twoside->stage.point = twoside_point; - twoside->stage.line = twoside_line; - twoside->stage.tri = twoside_tri; - twoside->stage.end = twoside_end; - - twoside->lookup = generic->vf_attr_to_slot; - - return &twoside->stage; -} diff --git a/src/mesa/softpipe/generic/g_prim_unfilled.c b/src/mesa/softpipe/generic/g_prim_unfilled.c deleted file mode 100644 index 7c6ee79f92..0000000000 --- a/src/mesa/softpipe/generic/g_prim_unfilled.c +++ /dev/null @@ -1,165 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_prim.h" - - - -struct unfilled_stage { - struct prim_stage stage; - - GLuint mode[2]; -}; - - -static INLINE struct unfilled_stage *unfilled_stage( struct prim_stage *stage ) -{ - return (struct unfilled_stage *)stage; -} - - -static void unfilled_begin( struct prim_stage *stage ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - - unfilled->mode[0] = stage->generic->setup.fill_ccw; - unfilled->mode[1] = stage->generic->setup.fill_cw; - - stage->next->begin( stage->next ); -} - -static void point( struct prim_stage *stage, - struct vertex_header *v0 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - stage->next->point( stage->next, &tmp ); -} - -static void line( struct prim_stage *stage, - struct vertex_header *v0, - struct vertex_header *v1 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - tmp.v[1] = v1; - stage->next->line( stage->next, &tmp ); -} - - -static void points( struct prim_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - - if (v0->edgeflag) point( stage, v0 ); - if (v1->edgeflag) point( stage, v1 ); - if (v2->edgeflag) point( stage, v2 ); -} - -static void lines( struct prim_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - - if (v0->edgeflag) line( stage, v0, v1 ); - if (v1->edgeflag) line( stage, v1, v2 ); - if (v2->edgeflag) line( stage, v2, v0 ); -} - - -/* Unfilled tri: - * - * Note edgeflags in the vertex struct is not sufficient as we will - * need to manipulate them when decomposing primitives??? - */ -static void unfilled_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - GLuint mode = unfilled->mode[header->det < 0]; - - switch (mode) { - case FILL_TRI: - stage->next->tri( stage->next, header ); - break; - - case FILL_LINE: - lines( stage, header ); - break; - - case GL_POINT: - points( stage, header ); - break; - } -} - -static void unfilled_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void unfilled_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void unfilled_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_unfilled( struct generic_context *generic ) -{ - struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); - - prim_alloc_tmps( &unfilled->stage, 0 ); - - unfilled->stage.generic = generic; - unfilled->stage.next = NULL; - unfilled->stage.tmp = NULL; - unfilled->stage.begin = unfilled_begin; - unfilled->stage.point = unfilled_point; - unfilled->stage.line = unfilled_line; - unfilled->stage.tri = unfilled_tri; - unfilled->stage.end = unfilled_end; - - return &unfilled->stage; -} diff --git a/src/mesa/softpipe/generic/g_state.h b/src/mesa/softpipe/generic/g_state.h deleted file mode 100644 index 5077d37ab7..0000000000 --- a/src/mesa/softpipe/generic/g_state.h +++ /dev/null @@ -1,61 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef PRIM_H -#define PRIM_H - -#include "glheader.h" -#include "softpipe/sp_state.h" - - -void generic_set_clip_state( struct softpipe_context *, - const struct softpipe_clip_state * ); - -void generic_set_viewport( struct softpipe_context *, - const struct softpipe_viewport * ); - -void generic_set_setup_state( struct softpipe_context *, - const struct softpipe_setup_state * ); - -void generic_set_scissor_rect( struct softpipe_context *, - const struct softpipe_scissor_rect * ); - -void generic_set_fs_state( struct softpipe_context *, - const struct softpipe_fs_state * ); - -void generic_set_polygon_stipple( struct softpipe_context *, - const struct softpipe_poly_stipple * ); - -void generic_set_cbuf_state( struct softpipe_context *, - const struct softpipe_surface * ); - -void generic_update_derived( struct generic_context *generic ); - -#endif diff --git a/src/mesa/softpipe/generic/g_state_clip.c b/src/mesa/softpipe/generic/g_state_clip.c deleted file mode 100644 index b90034e11e..0000000000 --- a/src/mesa/softpipe/generic/g_state_clip.c +++ /dev/null @@ -1,69 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_state.h" -#include "g_draw.h" - - - -void generic_set_clip_state( struct softpipe_context *softpipe, - const struct softpipe_clip_state *clip ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy(&generic->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); - - generic->nr_planes = 6 + clip->nr; - generic->dirty |= G_NEW_CLIP; -} - - - -/* Called when driver state tracker notices changes to the viewport - * matrix: - */ -void generic_set_viewport( struct softpipe_context *softpipe, - const struct softpipe_viewport *viewport ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy(&generic->viewport, viewport, sizeof(viewport)); - - /* Using tnl/ and vf/ modules is temporary while getting started. - * Full softpipe will have vertex shader, vertex fetch of its own. - */ - draw_set_viewport( generic->draw, viewport->scale, viewport->translate ); - generic->dirty |= G_NEW_VIEWPORT; -} - - - diff --git a/src/mesa/softpipe/generic/g_state_derived.c b/src/mesa/softpipe/generic/g_state_derived.c deleted file mode 100644 index 678825afe1..0000000000 --- a/src/mesa/softpipe/generic/g_state_derived.c +++ /dev/null @@ -1,136 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "macros.h" -#include "enums.h" -#include "program.h" - -#include "g_context.h" -#include "g_draw.h" -#include "g_state.h" - -#define EMIT_ATTR( ATTR, FRAG_ATTR, INTERP ) \ -do { \ - slot_to_vf_attr[generic->nr_attrs] = ATTR; \ - generic->vf_attr_to_slot[ATTR] = generic->nr_attrs; \ - generic->fp_attr_to_slot[FRAG_ATTR] = generic->nr_attrs; \ - generic->interp[generic->nr_attrs] = INTERP; \ - generic->nr_attrs++; \ - attr_mask |= (1<fs.fp; - const GLuint inputsRead = fp->Base.InputsRead; - GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; - GLuint attr_mask = 0; - GLuint i; - - generic->nr_attrs = 0; - memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); - - memset(generic->fp_attr_to_slot, 0, sizeof(generic->vf_attr_to_slot)); - memset(generic->vf_attr_to_slot, 0, sizeof(generic->fp_attr_to_slot)); - - /* TODO - Figure out if we need to do perspective divide, etc. - */ - EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); - - /* Pull in the rest of the attributes. They are all in float4 - * format. Future optimizations could be to keep some attributes - * as fixed point or ubyte format. - */ - for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { - if (inputsRead & (i << i)) { - EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); - } - } - - for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { - if (inputsRead & (i << i)) { - EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); - } - } - - generic->nr_frag_attrs = generic->nr_attrs; - - /* Additional attributes required for setup: Just twosided - * lighting. Edgeflag is dealt with specially by setting bits in - * the vertex header. - */ - if (generic->setup.light_twoside) { - if (inputsRead & FRAG_BIT_COL0) { - EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ - } - - if (inputsRead & FRAG_BIT_COL1) { - EMIT_ATTR(VF_ATTRIB_BFC1, FRAG_ATTRIB_MAX, 0); /* XXX: discard after setup */ - } - } - - if (attr_mask != generic->attr_mask) { - generic->attr_mask = attr_mask; - - draw_set_vertex_attributes( generic->draw, - slot_to_vf_attr, - generic->nr_attrs ); - } -} - - -/* Hopefully this will remain quite simple, otherwise need to pull in - * something like the state tracker mechanism. - */ -void generic_update_derived( struct generic_context *generic ) -{ - if (generic->dirty & (G_NEW_SETUP | G_NEW_FS)) - calculate_vertex_layout( generic ); - - generic->dirty = 0; -} diff --git a/src/mesa/softpipe/generic/g_state_fs.c b/src/mesa/softpipe/generic/g_state_fs.c deleted file mode 100644 index 8a430c0c65..0000000000 --- a/src/mesa/softpipe/generic/g_state_fs.c +++ /dev/null @@ -1,50 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "g_context.h" -#include "g_state.h" - - - -void generic_set_fs_state( struct softpipe_context *softpipe, - const struct softpipe_fs_state *fs ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy(&generic->fs, fs, sizeof(*fs)); - - generic->dirty |= G_NEW_FS; -} - - - - - - - - - diff --git a/src/mesa/softpipe/generic/g_state_setup.c b/src/mesa/softpipe/generic/g_state_setup.c deleted file mode 100644 index 2a5176c5a1..0000000000 --- a/src/mesa/softpipe/generic/g_state_setup.c +++ /dev/null @@ -1,122 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "g_context.h" -#include "g_state.h" -#include "g_prim.h" - - - - -static void validate_prim_pipe( struct generic_context *generic ) -{ - struct prim_stage *next = generic->prim.setup; - - /* TODO: make the current primitive part of the state and build - * shorter pipelines for lines & points. - */ - if (generic->setup.fill_cw != FILL_TRI || - generic->setup.fill_ccw != FILL_TRI) { - - generic->prim.unfilled->next = next; - next = generic->prim.unfilled; - } - - if (generic->setup.offset_cw || - generic->setup.offset_ccw) { - generic->prim.offset->next = next; - next = generic->prim.offset; - } - - if (generic->setup.light_twoside) { - generic->prim.twoside->next = next; - next = generic->prim.twoside; - } - - /* Always run the cull stage as we calculate determinant there - * also. Fix this.. - */ - { - generic->prim.cull->next = next; - next = generic->prim.cull; - } - - - /* Clip stage - */ - { - generic->prim.clip->next = next; - next = generic->prim.clip; - } - - /* Do software flatshading prior to clipping. XXX: should only do - * this for clipped primitives, ie it is a part of the clip - * routine. - */ - if (generic->setup.flatshade) { - generic->prim.flatshade->next = next; - next = generic->prim.flatshade; - } - - - generic->prim.first = next; -} - - - - -void generic_set_setup_state( struct softpipe_context *softpipe, - const struct softpipe_setup_state *setup ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy( &generic->setup, setup, sizeof(*setup) ); - - validate_prim_pipe( generic ); - generic->dirty |= G_NEW_SETUP; -} - - - -void generic_set_scissor_rect( struct softpipe_context *softpipe, - const struct softpipe_scissor_rect *scissor ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy( &generic->scissor, scissor, sizeof(*scissor) ); - generic->dirty |= G_NEW_SCISSOR; -} - - -void generic_set_polygon_stipple( struct softpipe_context *softpipe, - const struct softpipe_poly_stipple *stipple ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy( &generic->poly_stipple, stipple, sizeof(*stipple) ); - generic->dirty |= G_NEW_STIPPLE; -} diff --git a/src/mesa/softpipe/generic/g_state_surface.c b/src/mesa/softpipe/generic/g_state_surface.c deleted file mode 100644 index c791ce4640..0000000000 --- a/src/mesa/softpipe/generic/g_state_surface.c +++ /dev/null @@ -1,53 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_state.h" -#include "g_surface.h" - - -/* This is all a total hack. - */ -void generic_set_cbuf_state( struct softpipe_context *softpipe, - const struct softpipe_surface *surface ) -{ - struct generic_context *generic = generic_context(softpipe); - - if (generic->cbuf_surface == NULL) { - generic->cbuf_surface = CALLOC_STRUCT(generic_surface); - generic->cbuf_surface->type = &gs_rgba8; - } - - generic->cbuf_surface->surface = *surface; - generic->dirty |= G_NEW_CBUF; -} - - diff --git a/src/mesa/softpipe/generic/g_surface.c b/src/mesa/softpipe/generic/g_surface.c deleted file mode 100644 index 2df30cde70..0000000000 --- a/src/mesa/softpipe/generic/g_surface.c +++ /dev/null @@ -1,153 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "g_context.h" -#include "g_state.h" -#include "g_surface.h" -#include "g_headers.h" - -static void rgba8_read_quad_f( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - rgba[k][0] = ptr[0] * (1.0 / 255.0); - rgba[k][1] = ptr[1] * (1.0 / 255.0); - rgba[k][2] = ptr[2] * (1.0 / 255.0); - rgba[k][3] = ptr[3] * (1.0 / 255.0); - } - } -} - -static void rgba8_read_quad_f_swz( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - rrrr[0][k] = ptr[0] * (1.0 / 255.0); - rrrr[1][k] = ptr[1] * (1.0 / 255.0); - rrrr[2][k] = ptr[2] * (1.0 / 255.0); - rrrr[3][k] = ptr[3] * (1.0 / 255.0); - } - } -} - -static void rgba8_write_quad_f( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - ptr[0] = rgba[k][0] * 255.0; - ptr[1] = rgba[k][1] * 255.0; - ptr[2] = rgba[k][2] * 255.0; - ptr[3] = rgba[k][3] * 255.0; - } - } -} - -static void rgba8_write_quad_f_swz( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - ptr[0] = rrrr[0][k] * 255.0; - ptr[1] = rrrr[1][k] * 255.0; - ptr[2] = rrrr[2][k] * 255.0; - ptr[3] = rrrr[3][k] * 255.0; - } - } -} - - - - -static void rgba8_read_quad_ub( struct generic_surface *gs, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - rgba[k][0] = ptr[0]; - rgba[k][1] = ptr[1]; - rgba[k][2] = ptr[2]; - rgba[k][3] = ptr[3]; - } - } -} - - -static void rgba8_write_quad_ub( struct generic_surface *gs, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - ptr[0] = rgba[k][0]; - ptr[1] = rgba[k][1]; - ptr[2] = rgba[k][2]; - ptr[3] = rgba[k][3]; - } - } -} - - - - -struct generic_surface_type gs_rgba8 = { - G_SURFACE_RGBA_8888, - rgba8_read_quad_f, - rgba8_read_quad_f_swz, - rgba8_read_quad_ub, - rgba8_write_quad_f, - rgba8_write_quad_f_swz, - rgba8_write_quad_ub, -}; - - - diff --git a/src/mesa/softpipe/generic/g_surface.h b/src/mesa/softpipe/generic/g_surface.h deleted file mode 100644 index 23c28f0bb4..0000000000 --- a/src/mesa/softpipe/generic/g_surface.h +++ /dev/null @@ -1,115 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_SURFACE_H -#define G_SURFACE_H - -#include "glheader.h" -#include "g_headers.h" - -struct generic_surface; - -#define G_SURFACE_RGBA_8888 0x1 - -/* Internal structs and helpers for the primitive clip/setup pipeline: - */ -struct generic_surface_type { - - GLuint format; - - void (*read_quad_f)( struct generic_surface *, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ); - - void (*read_quad_f_swz)( struct generic_surface *, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ); - - void (*read_quad_ub)( struct generic_surface *, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ); - - - void (*write_quad_f)( struct generic_surface *, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ); - - void (*write_quad_f_swz)( struct generic_surface *, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ); - - - void (*write_quad_ub)( struct generic_surface *, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ); - - -}; - - -struct generic_surface { - struct generic_surface_type *type; - struct softpipe_surface surface; -}; - - -static INLINE void gs_read_quad_f( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - gs->type->read_quad_f(gs, x, y, rgba); -} - -static INLINE void gs_read_quad_f_swz( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - gs->type->read_quad_f_swz(gs, x, y, rrrr); -} - -static INLINE void gs_write_quad_f( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - gs->type->write_quad_f(gs, x, y, rgba); -} - -static INLINE void gs_write_quad_f_swz( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - gs->type->write_quad_f_swz(gs, x, y, rrrr); -} - -/* Like this, or hidden? - */ -struct generic_surface_type gs_rgba8; - -#endif diff --git a/src/mesa/softpipe/generic/g_tile.h b/src/mesa/softpipe/generic/g_tile.h deleted file mode 100644 index 6ae5212356..0000000000 --- a/src/mesa/softpipe/generic/g_tile.h +++ /dev/null @@ -1,42 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#ifndef G_TILE_H -#define G_TILE_H - -struct generic_context; -struct quad_header; - -void quad_shade( struct generic_context *generic, - struct quad_header *quad ); - -void quad_output( struct generic_context *generic, - struct quad_header *quad ); - -#endif diff --git a/src/mesa/softpipe/generic/g_tile_fs.c b/src/mesa/softpipe/generic/g_tile_fs.c deleted file mode 100644 index 4c5e6efaed..0000000000 --- a/src/mesa/softpipe/generic/g_tile_fs.c +++ /dev/null @@ -1,203 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "glheader.h" -#include "g_context.h" -#include "g_headers.h" -#include "g_tile.h" - -struct exec_machine { - const struct setup_coefficient *coef; - - GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE]; -}; - - -/** - * Compute quad's attributes values, as constants (GL_FLAT shading). - */ -static INLINE void cinterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) -{ - GLuint j; - - for (j = 0; j < QUAD_SIZE; j++) { - exec->attr[attrib][i][j] = exec->coef[attrib].a0[i]; - } -} - - -/** - * Compute quad's attribute values by linear interpolation. - * - * Push into the fp: - * - * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx - * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy - */ -static INLINE void linterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) -{ - GLuint j; - - for (j = 0; j < QUAD_SIZE; j++) { - const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; - exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * x + - exec->coef[attrib].dady[i] * y); - } -} - - -/** - * Compute quad's attribute values by linear interpolation with - * perspective correction. - * - * Push into the fp: - * - * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx - * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy - * INPUT[attr] = MUL INPUT[attr], INPUT_WPOS.wwww - * - * (Or should that be 1/w ???) - */ -static INLINE void pinterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) -{ - GLuint j; - - for (j = 0; j < QUAD_SIZE; j++) { - const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; - const GLfloat invW = exec->attr[FRAG_ATTRIB_WPOS][3][j]; - exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * x + - exec->coef[attrib].dady[i] * y) * invW); - } -} - - - -/* This should be done by the fragment shader execution unit (code - * generated from the decl instructions). Do it here for now. - */ -void quad_shade( struct generic_context *generic, - struct quad_header *quad ) -{ - struct exec_machine exec; - GLfloat fx = quad->x0; - GLfloat fy = quad->y0; - GLuint i, j; - - exec.coef = quad->coef; - - /* Position: - */ - exec.attr[FRAG_ATTRIB_WPOS][0][0] = fx; - exec.attr[FRAG_ATTRIB_WPOS][0][1] = fx + 1.0; - exec.attr[FRAG_ATTRIB_WPOS][0][2] = fx; - exec.attr[FRAG_ATTRIB_WPOS][0][3] = fx + 1.0; - - exec.attr[FRAG_ATTRIB_WPOS][1][0] = fy; - exec.attr[FRAG_ATTRIB_WPOS][1][1] = fy; - exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0; - exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; - - /* Z and W are done by linear interpolation: - * XXX we'll probably have to use integers for Z - */ - if (generic->need_z) { - linterp(&exec, 0, 2); - } - - if (generic->need_w) { - linterp(&exec, 0, 3); -// invert(&exec, 0, 3); - } - - /* Interpolate all the remaining attributes. This will get pushed - * into the fragment program's responsibilities at some point. - */ - for (i = 1; i < quad->nr_attrs; i++) { -#if 1 - for (j = 0; j < NUM_CHANNELS; j++) - linterp(&exec, i, j); -#else - switch (quad->interp[i]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - cinterp(&exec, i, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - linterp(&exec, i, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - pinterp(&exec, i, j); - break; - } -#endif - } - -#if 0 - generic->run_fs( tri->fp, quad, &tri->outputs ); -#else - { - GLuint attr = generic->fp_attr_to_slot[FRAG_ATTRIB_COL0]; - assert(attr); - - memcpy(quad->outputs.color, - exec.attr[attr], - sizeof(quad->outputs.color)); - } -#endif - - - if (quad->mask) - quad_output( generic, quad ); -} - - - - - - - diff --git a/src/mesa/softpipe/generic/g_tile_output.c b/src/mesa/softpipe/generic/g_tile_output.c deleted file mode 100644 index 058ca30193..0000000000 --- a/src/mesa/softpipe/generic/g_tile_output.c +++ /dev/null @@ -1,92 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "glheader.h" -#include "g_context.h" -#include "g_headers.h" -#include "g_surface.h" -#include "g_tile.h" - - -static void mask_copy( GLfloat (*dest)[4], - GLfloat (*src)[4], - GLuint mask ) -{ - GLuint i, j; - - for (i = 0; i < 4; i++) { - if (mask & (1<mask != MASK_ALL) - { - GLfloat tmp[4][QUAD_SIZE]; - - /* Yes, we'll probably have a masked write as well, but this is - * how blend will be done at least. - */ - gs_read_quad_f_swz( generic->cbuf_surface, - quad->x0, - quad->y0, - tmp ); - - mask_copy( tmp, quad->outputs.color, quad->mask ); - - gs_write_quad_f_swz( generic->cbuf_surface, - quad->x0, - quad->y0, - tmp ); - } - else - { - gs_write_quad_f_swz( generic->cbuf_surface, - quad->x0, - quad->y0, - quad->outputs.color ); - } -} diff --git a/src/mesa/softpipe/sp_context.h b/src/mesa/softpipe/sp_context.h deleted file mode 100644 index 716ddd27a9..0000000000 --- a/src/mesa/softpipe/sp_context.h +++ /dev/null @@ -1,89 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SP_CONTEXT_H -#define SP_CONTEXT_H - -#include "mtypes.h" - -extern struct softpipe_context *generic_create( void ); - -/* Drawing currently kludged up via the existing tnl/ module. - */ -struct vertex_buffer; - - -/** - * Software pipeline rendering context. Basically a collection of - * state setting functions, plus VBO drawing entrypoint. - */ -struct softpipe_context { - - void (*destroy)( struct softpipe_context * ); - - /* - * Drawing - */ - void (*draw_vb)( struct softpipe_context *softpipe, - struct vertex_buffer *VB ); - - /* - * State functions - */ - void (*set_blend_state)( struct softpipe_context *, - const struct softpipe_blend_state * ); - - void (*set_cbuf_state)( struct softpipe_context *, - const struct softpipe_surface * ); - - void (*set_clip_state)( struct softpipe_context *, - const struct softpipe_clip_state * ); - - void (*set_depth_state)( struct softpipe_context *, - const struct softpipe_depth_state * ); - - void (*set_fs_state)( struct softpipe_context *, - const struct softpipe_fs_state * ); - - void (*set_polygon_stipple)( struct softpipe_context *, - const struct softpipe_poly_stipple * ); - - void (*set_setup_state)( struct softpipe_context *, - const struct softpipe_setup_state * ); - - void (*set_scissor_rect)( struct softpipe_context *, - const struct softpipe_scissor_rect * ); - - void (*set_stencil_state)( struct softpipe_context *, - const struct softpipe_stencil_state * ); - - void (*set_viewport)( struct softpipe_context *, - const struct softpipe_viewport * ); -}; - - -#endif diff --git a/src/mesa/softpipe/sp_defines.h b/src/mesa/softpipe/sp_defines.h deleted file mode 100644 index c381865dab..0000000000 --- a/src/mesa/softpipe/sp_defines.h +++ /dev/null @@ -1,121 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SP_DEFINES_H -#define SP_DEFINES_H - -#define SP_BLENDFACTOR_ONE 0x1 -#define SP_BLENDFACTOR_SRC_COLOR 0x2 -#define SP_BLENDFACTOR_SRC_ALPHA 0x3 -#define SP_BLENDFACTOR_DST_ALPHA 0x4 -#define SP_BLENDFACTOR_DST_COLOR 0x5 -#define SP_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 -#define SP_BLENDFACTOR_CONST_COLOR 0x7 -#define SP_BLENDFACTOR_CONST_ALPHA 0x8 -#define SP_BLENDFACTOR_SRC1_COLOR 0x9 -#define SP_BLENDFACTOR_SRC1_ALPHA 0x0A -#define SP_BLENDFACTOR_ZERO 0x11 -#define SP_BLENDFACTOR_INV_SRC_COLOR 0x12 -#define SP_BLENDFACTOR_INV_SRC_ALPHA 0x13 -#define SP_BLENDFACTOR_INV_DST_ALPHA 0x14 -#define SP_BLENDFACTOR_INV_DST_COLOR 0x15 -#define SP_BLENDFACTOR_INV_CONST_COLOR 0x17 -#define SP_BLENDFACTOR_INV_CONST_ALPHA 0x18 -#define SP_BLENDFACTOR_INV_SRC1_COLOR 0x19 -#define SP_BLENDFACTOR_INV_SRC1_ALPHA 0x1A - -#define SP_BLEND_ADD 0 -#define SP_BLEND_SUBTRACT 1 -#define SP_BLEND_REVERSE_SUBTRACT 2 -#define SP_BLEND_MIN 3 -#define SP_BLEND_MAX 4 - -#define SP_LOGICOP_CLEAR 0 -#define SP_LOGICOP_NOR 1 -#define SP_LOGICOP_AND_INVERTED 2 -#define SP_LOGICOP_COPY_INVERTED 3 -#define SP_LOGICOP_AND_REVERSE 4 -#define SP_LOGICOP_INVERT 5 -#define SP_LOGICOP_XOR 6 -#define SP_LOGICOP_NAND 7 -#define SP_LOGICOP_AND 8 -#define SP_LOGICOP_EQUIV 9 -#define SP_LOGICOP_NOOP 10 -#define SP_LOGICOP_OR_INVERTED 11 -#define SP_LOGICOP_COPY 12 -#define SP_LOGICOP_OR_REVERSE 13 -#define SP_LOGICOP_OR 14 -#define SP_LOGICOP_SET 15 - -#define SP_STENCIL_FUNC_NEVER 0 -#define SP_STENCIL_FUNC_LESS 1 -#define SP_STENCIL_FUNC_EQUAL 2 -#define SP_STENCIL_FUNC_LEQUAL 3 -#define SP_STENCIL_FUNC_GREATER 4 -#define SP_STENCIL_FUNC_NOTEQUAL 5 -#define SP_STENCIL_FUNC_GEQUAL 6 -#define SP_STENCIL_FUNC_ALWAYS 7 - -#define SP_STENCIL_OP_KEEP 0 -#define SP_STENCIL_OP_ZERO 1 -#define SP_STENCIL_OP_REPLACE 2 -#define SP_STENCIL_OP_INCR 3 -#define SP_STENCIL_OP_DECR 4 -#define SP_STENCIL_OP_INCR_WRAP 5 -#define SP_STENCIL_OP_DECR_WRAP 6 -#define SP_STENCIL_OP_INVERT 7 - -/* Note: same as stencil funcs. Also used for shadow/depth compare */ -#define SP_DEPTH_FUNC_NEVER 0 -#define SP_DEPTH_FUNC_LESS 1 -#define SP_DEPTH_FUNC_EQUAL 2 -#define SP_DEPTH_FUNC_LEQUAL 3 -#define SP_DEPTH_FUNC_GREATER 4 -#define SP_DEPTH_FUNC_NOTEQUAL 5 -#define SP_DEPTH_FUNC_GEQUAL 6 -#define SP_DEPTH_FUNC_ALWAYS 7 - -#define SP_TEX_WRAP_REPEAT 0 -#define SP_TEX_WRAP_CLAMP 1 -#define SP_TEX_WRAP_CLAMP_TO_EDGE 2 -#define SP_TEX_WRAP_CLAMP_TO_BORDER 3 -#define SP_TEX_WRAP_MIRROR_REPEAT 4 -#define SP_TEX_WRAP_MIRROR_CLAMP 5 -#define SP_TEX_WRAP_MIRROR_CLAMP_TO_EDGE 6 -#define SP_TEX_WRAP_MIRROR_CLAMP_TO_BORDER 7 - -#define SP_TEX_FILTER_NEAREST 0 -#define SP_TEX_FILTER_LINEAR 1 -#define SP_TEX_FILTER_NEAREST_MIPMAP_NEAREST 2 -#define SP_TEX_FILTER_NEAREST_MIPMAP_LINEAR 3 -#define SP_TEX_FILTER_LINEAR_MIPMAP_NEAREST 4 -#define SP_TEX_FILTER_LINEAR_MIPMAP_LINEAR 5 - -#define SP_TEX_COMPARE_NONE 0 -#define SP_TEX_COMPARE_R_TO_TEXTURE 1 - -#endif diff --git a/src/mesa/softpipe/sp_state.h b/src/mesa/softpipe/sp_state.h deleted file mode 100644 index 150fa9fcad..0000000000 --- a/src/mesa/softpipe/sp_state.h +++ /dev/null @@ -1,184 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SP_STATE_H -#define SP_STATE_H - -#include "mtypes.h" -#include "vf/vf.h" - -#define WINDING_NONE 0 -#define WINDING_CW 1 -#define WINDING_CCW 2 -#define WINDING_BOTH (WINDING_CW | WINDING_CCW) - -#define FILL_POINT 1 -#define FILL_LINE 2 -#define FILL_TRI 3 - -struct softpipe_setup_state { - GLuint flatshade:1; - GLuint light_twoside:1; - - GLuint front_winding:2; - - GLuint cull_mode:2; - - GLuint fill_cw:2; - GLuint fill_ccw:2; - - GLuint offset_cw:1; - GLuint offset_ccw:1; - - GLuint scissor:1; - GLuint poly_stipple:1; - - GLuint pad:18; - - GLfloat offset_units; - GLfloat offset_scale; -}; - -struct softpipe_poly_stipple { - GLuint stipple[32]; -}; - - -struct softpipe_viewport { - GLfloat scale[4]; - GLfloat translate[4]; -}; - -struct softpipe_scissor_rect { - GLshort minx; - GLshort miny; - GLshort maxx; - GLshort maxy; -}; - - -#define SP_MAX_CLIP_PLANES 6 - -struct softpipe_clip_state { - GLfloat ucp[SP_MAX_CLIP_PLANES][4]; - GLuint nr; -}; - -struct softpipe_fs_state { - struct gl_fragment_program *fp; -}; - -#define SP_MAX_CONSTANT 32 - -struct softpipe_constant_buffer { - GLfloat constant[SP_MAX_CONSTANT][4]; - GLuint nr_constants; -}; - - -struct softpipe_depth_state -{ - GLuint enabled:1; /**< depth test enabled? */ - GLuint writemask:1; /**< allow depth buffer writes? */ - GLuint func:3; /**< depth test func */ - GLfloat clear; /**< Clear value in [0,1] (XXX correct place?) */ -}; - - -struct softpipe_blend_state { - GLuint blend_enable:1; - - GLuint rgb_func:3; - GLuint rgb_src_factor:5; - GLuint rgb_dst_factor:5; - - GLuint alpha_func:3; - GLuint alpha_src_factor:5; - GLuint alpha_dst_factor:5; - - GLuint logicop_enable:1; - GLuint logicop_func:4; -}; - -struct softpipe_blend_color { - GLfloat color[4]; -}; - - -struct softpipe_stencil_state { - GLuint front_enabled:1; - GLuint front_func:3; /**< SP_STENCIL_FUNC_x */ - GLuint front_fail_op:3; /**< SP_STENCIL_OP_x */ - GLuint front_zpass_op:3; /**< SP_STENCIL_OP_x */ - GLuint front_zfail_op:3; /**< SP_STENCIL_OP_x */ - GLuint back_enabled:1; - GLuint back_func:3; - GLuint back_fail_op:3; - GLuint back_zpass_op:3; - GLuint back_zfail_op:3; - GLubyte ref_value[2]; /**< [0] = front, [1] = back */ - GLubyte value_mask[2]; - GLubyte write_mask[2]; - GLubyte clear_value; -}; - - -/* This will change for hardware softpipes... - */ -struct softpipe_surface { - GLubyte *ptr; - GLint stride; - GLuint cpp; - GLuint format; -}; - - -/** - * Texture sampler state. - */ -struct softpipe_sampler_state -{ - GLuint wrap_s:3; /**< SP_TEX_WRAP_x */ - GLuint wrap_t:3; /**< SP_TEX_WRAP_x */ - GLuint wrap_r:3; /**< SP_TEX_WRAP_x */ - GLuint min_filter:3; /**< SP_TEX_FILTER_x */ - GLuint mag_filter:1; /**< SP_TEX_FILTER_LINEAR or _NEAREST */ - GLfloat min_lod; - GLfloat max_lod; - GLfloat lod_bias; -#if 0 /* need these? */ - GLint BaseLevel; /**< min mipmap level, OpenGL 1.2 */ - GLint MaxLevel; /**< max mipmap level, OpenGL 1.2 */ -#endif - GLfloat max_anisotropy; - GLuint compare:1; /**< shadow/depth compare enabled? */ - GLenum compare_mode:1; /**< SP_TEX_COMPARE_x */ - GLenum compare_func:3; /**< SP_DEPTH_FUNC_x */ - GLfloat shadow_ambient; /**< shadow test fail color/intensity */ -}; - -#endif diff --git a/src/mesa/softpipe/state_tracker/Makefile b/src/mesa/softpipe/state_tracker/Makefile deleted file mode 100644 index 0ab1dc6e6b..0000000000 --- a/src/mesa/softpipe/state_tracker/Makefile +++ /dev/null @@ -1,2 +0,0 @@ -default: - cd ../.. ; make \ No newline at end of file diff --git a/src/mesa/softpipe/state_tracker/st_atom.c b/src/mesa/softpipe/state_tracker/st_atom.c deleted file mode 100644 index fdbfb9021d..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom.c +++ /dev/null @@ -1,171 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "glheader.h" -#include "context.h" - -#include "st_context.h" -#include "st_atom.h" - - - -/* This is used to initialize st->atoms[]. We could use this list - * directly except for a single atom, st_update_constants, which has a - * .dirty value which changes according to the parameters of the - * current fragment and vertex programs, and so cannot be a static - * value. - */ -static const struct st_tracked_state *atoms[] = -{ - &st_update_cbuf, - &st_update_clip, - &st_update_fs, - &st_update_setup, - &st_update_viewport, - &st_update_scissor, - &st_update_blend, - &st_update_stencil, - /* will be patched out at runtime */ -/* &st_update_constants */ -}; - - -void st_init_atoms( struct st_context *st ) -{ - GLuint i; - - st->atoms = _mesa_malloc(sizeof(atoms)); - st->nr_atoms = sizeof(atoms)/sizeof(*atoms); - memcpy(st->atoms, atoms, sizeof(atoms)); - - /* Patch in a pointer to the dynamic state atom: - */ - for (i = 0; i < st->nr_atoms; i++) - if (st->atoms[i] == &st_update_constants) - st->atoms[i] = &st->constants.tracked_state; - - memcpy(&st->constants.tracked_state, - &st_update_constants, - sizeof(st_update_constants)); -} - - -void st_destroy_atoms( struct st_context *st ) -{ - if (st->atoms) { - _mesa_free(st->atoms); - st->atoms = NULL; - } -} - - -/*********************************************************************** - */ - -static GLboolean check_state( const struct st_state_flags *a, - const struct st_state_flags *b ) -{ - return ((a->mesa & b->mesa) || - (a->st & b->st)); -} - -static void accumulate_state( struct st_state_flags *a, - const struct st_state_flags *b ) -{ - a->mesa |= b->mesa; - a->st |= b->st; -} - - -static void xor_states( struct st_state_flags *result, - const struct st_state_flags *a, - const struct st_state_flags *b ) -{ - result->mesa = a->mesa ^ b->mesa; - result->st = a->st ^ b->st; -} - - -/*********************************************************************** - * Update all derived state: - */ - -void st_validate_state( struct st_context *st ) -{ - struct st_state_flags *state = &st->dirty; - GLuint i; - - if (state->st == 0) - return; - - if (1) { - /* Debug version which enforces various sanity checks on the - * state flags which are generated and checked to help ensure - * state atoms are ordered correctly in the list. - */ - struct st_state_flags examined, prev; - _mesa_memset(&examined, 0, sizeof(examined)); - prev = *state; - - for (i = 0; i < st->nr_atoms; i++) { - const struct st_tracked_state *atom = st->atoms[i]; - struct st_state_flags generated; - - assert(atom->dirty.mesa || - atom->dirty.st); - assert(atom->update); - - if (check_state(state, &atom->dirty)) { - st->atoms[i]->update( st ); - } - - accumulate_state(&examined, &atom->dirty); - - /* generated = (prev ^ state) - * if (examined & generated) - * fail; - */ - xor_states(&generated, &prev, state); - assert(!check_state(&examined, &generated)); - prev = *state; - } - } - else { - const GLuint nr = st->nr_atoms; - - for (i = 0; i < nr; i++) { - if (check_state(state, &st->atoms[i]->dirty)) - st->atoms[i]->update( st ); - } - } - - memset(state, 0, sizeof(*state)); -} - - - diff --git a/src/mesa/softpipe/state_tracker/st_atom.h b/src/mesa/softpipe/state_tracker/st_atom.h deleted file mode 100644 index 1d8da46336..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom.h +++ /dev/null @@ -1,59 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef ST_ATOM_H -#define ST_ATOM_H - -struct st_context; -struct st_tracked_state; - -void st_init_atoms( struct st_context *st ); -void st_destroy_atoms( struct st_context *st ); - - -void st_validate_state( struct st_context *st ); - - -const struct st_tracked_state st_update_cbuf; -const struct st_tracked_state st_update_clip; -const struct st_tracked_state st_update_depth; -const struct st_tracked_state st_update_fs; -const struct st_tracked_state st_update_setup; -const struct st_tracked_state st_update_viewport; -const struct st_tracked_state st_update_constants; -const struct st_tracked_state st_update_scissor; -const struct st_tracked_state st_update_blend; -const struct st_tracked_state st_update_stencil; - - -#endif diff --git a/src/mesa/softpipe/state_tracker/st_atom_blend.c b/src/mesa/softpipe/state_tracker/st_atom_blend.c deleted file mode 100644 index eabb9172ba..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom_blend.c +++ /dev/null @@ -1,202 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#include "st_context.h" -#include "st_atom.h" -#include "softpipe/sp_context.h" -#include "softpipe/sp_defines.h" - - -/** - * Convert GLenum blend tokens to softpipe tokens. - * Both blend factors and blend funcs are accepted. - */ -static GLuint -gl_blend_to_sp(GLenum blend) -{ - switch (blend) { - /* blend functions */ - case GL_FUNC_ADD: - return SP_BLEND_ADD; - case GL_FUNC_SUBTRACT: - return SP_BLEND_SUBTRACT; - case GL_FUNC_REVERSE_SUBTRACT: - return SP_BLEND_REVERSE_SUBTRACT; - case GL_MIN: - return SP_BLEND_MIN; - case GL_MAX: - return SP_BLEND_MAX; - - /* blend factors */ - case GL_ONE: - return SP_BLENDFACTOR_ONE; - case GL_SRC_COLOR: - return SP_BLENDFACTOR_SRC_COLOR; - case GL_SRC_ALPHA: - return SP_BLENDFACTOR_SRC_ALPHA; - case GL_DST_ALPHA: - return SP_BLENDFACTOR_DST_ALPHA; - case GL_DST_COLOR: - return SP_BLENDFACTOR_DST_COLOR; - case GL_SRC_ALPHA_SATURATE: - return SP_BLENDFACTOR_SRC_ALPHA_SATURATE; - case GL_CONSTANT_COLOR: - return SP_BLENDFACTOR_CONST_COLOR; - case GL_CONSTANT_ALPHA: - return SP_BLENDFACTOR_CONST_ALPHA; - /* - return SP_BLENDFACTOR_SRC1_COLOR; - return SP_BLENDFACTOR_SRC1_ALPHA; - */ - case GL_ZERO: - return SP_BLENDFACTOR_ZERO; - case GL_ONE_MINUS_SRC_COLOR: - return SP_BLENDFACTOR_INV_SRC_COLOR; - case GL_ONE_MINUS_SRC_ALPHA: - return SP_BLENDFACTOR_INV_SRC_ALPHA; - case GL_ONE_MINUS_DST_COLOR: - return SP_BLENDFACTOR_INV_DST_ALPHA; - case GL_ONE_MINUS_DST_ALPHA: - return SP_BLENDFACTOR_INV_DST_COLOR; - case GL_ONE_MINUS_CONSTANT_COLOR: - return SP_BLENDFACTOR_INV_CONST_COLOR; - case GL_ONE_MINUS_CONSTANT_ALPHA: - return SP_BLENDFACTOR_INV_CONST_ALPHA; - /* - return SP_BLENDFACTOR_INV_SRC1_COLOR; - return SP_BLENDFACTOR_INV_SRC1_ALPHA; - */ - default: - assert("invalid GL token in gl_blend_to_sp()" == NULL); - return 0; - } -} - - -/** - * Convert GLenum logicop tokens to softpipe tokens. - */ -static GLuint -gl_logicop_to_sp(GLenum logicop) -{ - switch (logicop) { - case GL_CLEAR: - return SP_LOGICOP_CLEAR; - case GL_NOR: - return SP_LOGICOP_NOR; - case GL_AND_INVERTED: - return SP_LOGICOP_AND_INVERTED; - case GL_COPY_INVERTED: - return SP_LOGICOP_COPY_INVERTED; - case GL_AND_REVERSE: - return SP_LOGICOP_AND_REVERSE; - case GL_INVERT: - return SP_LOGICOP_INVERT; - case GL_XOR: - return SP_LOGICOP_XOR; - case GL_NAND: - return SP_LOGICOP_NAND; - case GL_AND: - return SP_LOGICOP_AND; - case GL_EQUIV: - return SP_LOGICOP_EQUIV; - case GL_NOOP: - return SP_LOGICOP_NOOP; - case GL_OR_INVERTED: - return SP_LOGICOP_OR_INVERTED; - case GL_COPY: - return SP_LOGICOP_COPY; - case GL_OR_REVERSE: - return SP_LOGICOP_OR_REVERSE; - case GL_OR: - return SP_LOGICOP_OR; - case GL_SET: - return SP_LOGICOP_SET; - default: - assert("invalid GL token in gl_logicop_to_sp()" == NULL); - return 0; - } -} - - -static void -update_blend( struct st_context *st ) -{ - struct softpipe_blend_state blend; - - memset(&blend, 0, sizeof(blend)); - - if (st->ctx->Color.ColorLogicOpEnabled || - (st->ctx->Color.BlendEnabled && - st->ctx->Color.BlendEquationRGB == GL_LOGIC_OP)) { - /* logicop enabled */ - blend.logicop_enable = 1; - blend.logicop_func = gl_logicop_to_sp(st->ctx->Color.LogicOp); - } - else if (st->ctx->Color.BlendEnabled) { - /* blending enabled */ - blend.blend_enable = 1; - - blend.rgb_func = gl_blend_to_sp(st->ctx->Color.BlendEquationRGB); - blend.rgb_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcRGB); - blend.rgb_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstRGB); - - blend.alpha_func = gl_blend_to_sp(st->ctx->Color.BlendEquationA); - blend.alpha_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcA); - blend.alpha_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstA); - } - else { - /* no blending / logicop */ - } - - if (memcmp(&blend, &st->state.blend, sizeof(blend)) != 0) { - /* state has changed */ - st->state.blend = blend; /* struct copy */ - st->softpipe->set_blend_state(st->softpipe, &blend); /* set new state */ - } -} - - -const struct st_tracked_state st_update_blend = { - .dirty = { - .mesa = (_NEW_COLOR), /* XXX _NEW_BLEND someday? */ - .st = 0, - }, - .update = update_blend -}; - - - - - diff --git a/src/mesa/softpipe/state_tracker/st_atom_cbuf.c b/src/mesa/softpipe/state_tracker/st_atom_cbuf.c deleted file mode 100644 index cd707ec5ef..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom_cbuf.c +++ /dev/null @@ -1,72 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "st_context.h" -#include "softpipe/sp_context.h" -#include "st_atom.h" - -extern GLboolean xmesa_get_cbuf_details( GLcontext *ctx, - void **ptr, - GLuint *cpp, - GLint *stride, - GLuint *format ); - - -/* This is a hack to work with the X11 driver as a test harness - */ -static void update_cbuf_state( struct st_context *st ) -{ - struct softpipe_surface cbuf; - GLboolean ok; - - ok = xmesa_get_cbuf_details( st->ctx, - (void **)&cbuf.ptr, - &cbuf.cpp, - &cbuf.stride, - &cbuf.format ); - - assert(ok); - - if (memcmp(&cbuf, &st->state.cbuf, sizeof(cbuf)) != 0) { - st->state.cbuf = cbuf; - st->softpipe->set_cbuf_state( st->softpipe, &cbuf ); - } -} - -const struct st_tracked_state st_update_cbuf = { - .dirty = { - .mesa = _NEW_BUFFERS, - .st = 0, - }, - .update = update_cbuf_state -}; - diff --git a/src/mesa/softpipe/state_tracker/st_atom_clip.c b/src/mesa/softpipe/state_tracker/st_atom_clip.c deleted file mode 100644 index 710d6ffc84..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom_clip.c +++ /dev/null @@ -1,75 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#include "st_context.h" -#include "softpipe/sp_context.h" -#include "st_atom.h" - - -/* Second state atom for user clip planes: - */ -static void update_clip( struct st_context *st ) -{ - struct softpipe_clip_state clip; - GLuint i; - - memset(&clip, 0, sizeof(clip)); - - for (i = 0; i < SP_MAX_CLIP_PLANES; i++) { - if (st->ctx->Transform.ClipPlanesEnabled & (1 << i)) { - memcpy(clip.ucp[clip.nr], - st->ctx->Transform._ClipUserPlane[i], - sizeof(clip.ucp[0])); - clip.nr++; - } - } - - if (memcmp(&clip, &st->state.clip, sizeof(clip)) != 0) { - st->state.clip = clip; - st->softpipe->set_clip_state(st->softpipe, &clip); - } -} - - -const struct st_tracked_state st_update_clip = { - .dirty = { - .mesa = (_NEW_TRANSFORM), - .st = 0, - }, - .update = update_clip -}; - - - - - diff --git a/src/mesa/softpipe/state_tracker/st_atom_depth.c b/src/mesa/softpipe/state_tracker/st_atom_depth.c deleted file mode 100644 index 5532abc8fd..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom_depth.c +++ /dev/null @@ -1,88 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#include "st_context.h" -#include "st_atom.h" -#include "softpipe/sp_context.h" -#include "softpipe/sp_defines.h" - - -/** - * Convert GLenum depth func tokens to softpipe tokens. - */ -static GLuint -gl_depth_func_to_sp(GLenum func) -{ - /* Same values, just biased */ - assert(SP_DEPTH_FUNC_NEVER == GL_NEVER - GL_NEVER); - assert(SP_DEPTH_FUNC_LESS == GL_LESS - GL_NEVER); - assert(SP_DEPTH_FUNC_EQUAL == GL_EQUAL - GL_NEVER); - assert(SP_DEPTH_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); - assert(SP_DEPTH_FUNC_GREATER == GL_GREATER - GL_NEVER); - assert(SP_DEPTH_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); - assert(SP_DEPTH_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); - assert(SP_DEPTH_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); - assert(func >= GL_NEVER); - assert(func <= GL_ALWAYS); - return func - GL_NEVER; -} - - -static void -update_depth( struct st_context *st ) -{ - struct softpipe_depth_state depth; - - memset(&depth, 0, sizeof(depth)); - - depth.enabled = st->ctx->Depth.Test; - depth.writemask = st->ctx->Depth.Mask; - depth.func = gl_depth_func_to_sp(st->ctx->Depth.Func); - depth.clear = st->ctx->Depth.Clear; - - if (memcmp(&depth, &st->state.depth, sizeof(depth)) != 0) { - /* state has changed */ - st->state.depth = depth; /* struct copy */ - st->softpipe->set_depth_state(st->softpipe, &depth); /* set new state */ - } -} - - -const struct st_tracked_state st_update_depth = { - .dirty = { - .mesa = (_NEW_DEPTH), - .st = 0, - }, - .update = update_depth -}; diff --git a/src/mesa/softpipe/state_tracker/st_atom_fs.c b/src/mesa/softpipe/state_tracker/st_atom_fs.c deleted file mode 100644 index ca109d2d34..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom_fs.c +++ /dev/null @@ -1,56 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "st_context.h" -#include "softpipe/sp_context.h" -#include "st_atom.h" - - -static void update_fs( struct st_context *st ) -{ - struct softpipe_fs_state fs; - - fs.fp = st->ctx->FragmentProgram._Current; - - if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0) { - st->state.fs = fs; - st->softpipe->set_fs_state(st->softpipe, &fs); - } -} - - -const struct st_tracked_state st_update_fs = { - .dirty = { - .mesa = 0, - .st = ST_NEW_FRAGMENT_PROGRAM, - }, - .update = update_fs -}; diff --git a/src/mesa/softpipe/state_tracker/st_atom_scissor.c b/src/mesa/softpipe/state_tracker/st_atom_scissor.c deleted file mode 100644 index 105c2a6dd9..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom_scissor.c +++ /dev/null @@ -1,88 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#include "st_context.h" -#include "softpipe/sp_context.h" -#include "st_atom.h" - - -/** - * Scissor depends on the scissor box, and the framebuffer dimensions. - */ -static void -update_scissor( struct st_context *st ) -{ - struct softpipe_scissor_rect scissor; - const struct gl_framebuffer *fb = st->ctx->DrawBuffer; - - scissor.minx = 0; - scissor.miny = 0; - scissor.maxx = fb->Width; - scissor.maxy = fb->Height; - - if (st->ctx->Scissor.Enabled) { - if (st->ctx->Scissor.X > scissor.minx) - scissor.minx = st->ctx->Scissor.X; - if (st->ctx->Scissor.Y > scissor.miny) - scissor.miny = st->ctx->Scissor.Y; - - if (st->ctx->Scissor.X + st->ctx->Scissor.Width < scissor.maxx) - scissor.maxx = st->ctx->Scissor.X + st->ctx->Scissor.Width; - if (st->ctx->Scissor.Y + st->ctx->Scissor.Height < scissor.maxy) - scissor.maxy = st->ctx->Scissor.Y + st->ctx->Scissor.Height; - - /* check for null space */ - if (scissor.minx >= scissor.maxx || scissor.miny >= scissor.maxy) - scissor.minx = scissor.miny = scissor.maxx = scissor.maxy = 0; - } - - if (memcmp(&scissor, &st->state.scissor, sizeof(scissor)) != 0) { - /* state has changed */ - st->state.scissor = scissor; /* struct copy */ - st->softpipe->set_scissor_rect(st->softpipe, &scissor); /* activate */ - } -} - - -const struct st_tracked_state st_update_scissor = { - .dirty = { - .mesa = (_NEW_SCISSOR | _NEW_BUFFERS), - .st = 0, - }, - .update = update_scissor -}; - - - - - diff --git a/src/mesa/softpipe/state_tracker/st_atom_setup.c b/src/mesa/softpipe/state_tracker/st_atom_setup.c deleted file mode 100644 index 2a582ea36c..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom_setup.c +++ /dev/null @@ -1,175 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#include "st_context.h" -#include "softpipe/sp_context.h" -#include "st_atom.h" - -static GLuint translate_fill( GLenum mode ) -{ - switch (mode) { - case GL_POINT: return FILL_POINT; - case GL_LINE: return FILL_LINE; - case GL_FILL: return FILL_TRI; - default: assert(0); return 0; - } -} - -static GLboolean get_offset_flag( GLuint fill_mode, - const struct gl_polygon_attrib *Polygon ) -{ - switch (fill_mode) { - case FILL_POINT: return Polygon->OffsetPoint; - case FILL_LINE: return Polygon->OffsetLine; - case FILL_TRI: return Polygon->OffsetFill; - default: assert(0); return 0; - } -} - - -static void update_setup_state( struct st_context *st ) -{ - GLcontext *ctx = st->ctx; - struct softpipe_setup_state setup; - - memset(&setup, 0, sizeof(setup)); - - /* _NEW_POLYGON, _NEW_BUFFERS - */ - { - setup.front_winding = WINDING_CW; - - if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) - setup.front_winding ^= WINDING_BOTH; - - if (ctx->Polygon.FrontFace != GL_CCW) - setup.front_winding ^= WINDING_BOTH; - } - - /* _NEW_LIGHT - */ - if (ctx->Light.ShadeModel == GL_FLAT) - setup.flatshade = 1; - - /* _NEW_LIGHT - * - * Not sure about the light->enabled requirement - does this still - * apply?? - */ - if (ctx->Light.Enabled && - ctx->Light.Model.TwoSide) - setup.light_twoside = 1; - - - /* _NEW_POLYGON - */ - if (ctx->Polygon.CullFlag) { - if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) { - setup.cull_mode = WINDING_BOTH; - } - else if (ctx->Polygon.CullFaceMode == GL_FRONT) { - setup.cull_mode = setup.front_winding; - } - else { - setup.cull_mode = setup.front_winding ^ WINDING_BOTH; - } - } - - /* _NEW_POLYGON - */ - { - GLuint fill_front = translate_fill( ctx->Polygon.FrontMode ); - GLuint fill_back = translate_fill( ctx->Polygon.BackMode ); - - if (setup.front_winding == WINDING_CW) { - setup.fill_cw = fill_front; - setup.fill_ccw = fill_back; - } - else { - setup.fill_cw = fill_back; - setup.fill_ccw = fill_front; - } - - /* Simplify when culling is active: - */ - if (setup.cull_mode & WINDING_CW) { - setup.fill_cw = setup.fill_ccw; - } - - if (setup.cull_mode & WINDING_CCW) { - setup.fill_ccw = setup.fill_cw; - } - } - - /* Hardware does offset for filled prims, but need to do it in - * software for unfilled. - * - * _NEW_POLYGON - */ - if (setup.fill_cw != FILL_TRI) - setup.offset_cw = get_offset_flag( setup.fill_cw, - &ctx->Polygon ); - - if (setup.fill_ccw != FILL_TRI) - setup.offset_ccw = get_offset_flag( setup.fill_ccw, - &ctx->Polygon ); - - - /* _NEW_BUFFERS, _NEW_POLYGON - */ - if (setup.fill_cw != FILL_TRI || - setup.fill_ccw != FILL_TRI) - { - GLfloat mrd = (ctx->DrawBuffer ? - ctx->DrawBuffer->_MRD : - 1.0); - - setup.offset_units = ctx->Polygon.OffsetFactor * mrd; - setup.offset_scale = (ctx->Polygon.OffsetUnits * mrd * - st->polygon_offset_scale); - } - - - if (memcmp(&setup, &st->state.setup, sizeof(setup)) != 0) { - st->state.setup = setup; - st->softpipe->set_setup_state( st->softpipe, &setup ); - } -} - -const struct st_tracked_state st_update_setup = { - .dirty = { - .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_BUFFERS), - .st = 0, - }, - .update = update_setup_state -}; diff --git a/src/mesa/softpipe/state_tracker/st_atom_stencil.c b/src/mesa/softpipe/state_tracker/st_atom_stencil.c deleted file mode 100644 index 32610c3d24..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom_stencil.c +++ /dev/null @@ -1,140 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#include "st_context.h" -#include "st_atom.h" -#include "softpipe/sp_context.h" -#include "softpipe/sp_defines.h" - - -/** - * Convert GLenum stencil func tokens to softpipe tokens. - */ -static GLuint -gl_stencil_func_to_sp(GLenum func) -{ - /* Same values, just biased */ - assert(SP_STENCIL_FUNC_NEVER == GL_NEVER - GL_NEVER); - assert(SP_STENCIL_FUNC_LESS == GL_LESS - GL_NEVER); - assert(SP_STENCIL_FUNC_EQUAL == GL_EQUAL - GL_NEVER); - assert(SP_STENCIL_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); - assert(SP_STENCIL_FUNC_GREATER == GL_GREATER - GL_NEVER); - assert(SP_STENCIL_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); - assert(SP_STENCIL_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); - assert(SP_STENCIL_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); - assert(func >= GL_NEVER); - assert(func <= GL_ALWAYS); - return func - GL_NEVER; -} - - -/** - * Convert GLenum stencil op tokens to softpipe tokens. - */ -static GLuint -gl_stencil_op_to_sp(GLenum func) -{ - switch (func) { - case GL_KEEP: - return SP_STENCIL_OP_KEEP; - case GL_ZERO: - return SP_STENCIL_OP_ZERO; - case GL_REPLACE: - return SP_STENCIL_OP_REPLACE; - case GL_INCR: - return SP_STENCIL_OP_INCR; - case GL_DECR: - return SP_STENCIL_OP_DECR; - case GL_INCR_WRAP: - return SP_STENCIL_OP_INCR_WRAP; - case GL_DECR_WRAP: - return SP_STENCIL_OP_DECR_WRAP; - case GL_INVERT: - return SP_STENCIL_OP_INVERT; - default: - assert("invalid GL token in gl_stencil_op_to_sp()" == NULL); - return 0; - } -} - - -static void -update_stencil( struct st_context *st ) -{ - struct softpipe_stencil_state stencil; - - memset(&stencil, 0, sizeof(stencil)); - - if (st->ctx->Stencil.Enabled) { - stencil.front_enabled = 1; - stencil.front_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[0]); - stencil.front_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[0]); - stencil.front_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[0]); - stencil.front_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[0]); - stencil.ref_value[0] = st->ctx->Stencil.Ref[0] & 0xff; - stencil.value_mask[0] = st->ctx->Stencil.ValueMask[0] & 0xff; - stencil.write_mask[0] = st->ctx->Stencil.WriteMask[0] & 0xff; - if (st->ctx->Stencil.TestTwoSide) { - stencil.back_enabled = 1; - stencil.back_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[1]); - stencil.back_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[1]); - stencil.back_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[1]); - stencil.back_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[1]); - stencil.ref_value[1] = st->ctx->Stencil.Ref[1] & 0xff; - stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1] & 0xff; - stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1] & 0xff; - } - stencil.clear_value = st->ctx->Stencil.Clear & 0xff; - } - - if (memcmp(&stencil, &st->state.stencil, sizeof(stencil)) != 0) { - /* state has changed */ - st->state.stencil = stencil; /* struct copy */ - st->softpipe->set_stencil_state(st->softpipe, &stencil); /* set new state */ - } -} - - -const struct st_tracked_state st_update_stencil = { - .dirty = { - .mesa = (_NEW_STENCIL), - .st = 0, - }, - .update = update_stencil -}; - - - - - diff --git a/src/mesa/softpipe/state_tracker/st_atom_viewport.c b/src/mesa/softpipe/state_tracker/st_atom_viewport.c deleted file mode 100644 index af896e2e31..0000000000 --- a/src/mesa/softpipe/state_tracker/st_atom_viewport.c +++ /dev/null @@ -1,117 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "context.h" -#include "colormac.h" -#include "st_context.h" -#include "softpipe/sp_context.h" -#include "st_atom.h" - - - - - -/** - * Update the viewport transformation matrix. Depends on: - * - viewport pos/size - * - depthrange - * - window pos/size or FBO size - */ -static void update_viewport( struct st_context *st ) -{ - GLcontext *ctx = st->ctx; - const GLframebuffer *DrawBuffer = ctx->DrawBuffer; - GLfloat yScale = 1.0; - GLfloat yBias = 0.0; - - /* _NEW_BUFFERS - */ - if (DrawBuffer) { - -#if 0 - if (DrawBuffer->Name) { - /* User created FBO */ - struct st_renderbuffer *irb - = st_renderbuffer(DrawBuffer->_ColorDrawBuffers[0][0]); - if (irb && !irb->RenderToTexture) { - /* y=0=top */ - yScale = -1.0; - yBias = irb->Base.Height; - } - else { - /* y=0=bottom */ - yScale = 1.0; - yBias = 0.0; - } - } - else - { - /* window buffer, y=0=top */ - yScale = -1.0; - yBias = DrawBuffer->Height; - } -#endif - } - - { - /* _NEW_VIEWPORT - */ - GLfloat x = ctx->Viewport.X; - GLfloat y = ctx->Viewport.Y; - GLfloat z = ctx->Viewport.Near; - GLfloat half_width = ctx->Viewport.Width / 2.0; - GLfloat half_height = ctx->Viewport.Height / 2.0; - GLfloat half_depth = (ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; - - struct softpipe_viewport vp; - - vp.scale[0] = half_width; - vp.scale[1] = half_height * yScale; - vp.scale[2] = half_depth; - vp.scale[3] = 1.0; - - vp.translate[0] = (half_width + x); - vp.translate[1] = (half_height + y) * yScale + yBias; - vp.translate[2] = (half_depth + z); - vp.translate[3] = 0.0; - - if (memcmp(&vp, &st->state.viewport, sizeof(vp)) != 0) { - st->state.viewport = vp; - st->softpipe->set_viewport(st->softpipe, &vp); - } - } -} - - -const struct st_tracked_state st_update_viewport = { - .dirty = { - .mesa = _NEW_BUFFERS | _NEW_VIEWPORT, - .st = 0, - }, - .update = update_viewport -}; diff --git a/src/mesa/softpipe/state_tracker/st_cb_program.c b/src/mesa/softpipe/state_tracker/st_cb_program.c deleted file mode 100644 index 67a589ca8d..0000000000 --- a/src/mesa/softpipe/state_tracker/st_cb_program.c +++ /dev/null @@ -1,165 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "st_context.h" -#include "st_program.h" - -#include "glheader.h" -#include "macros.h" -#include "enums.h" -#include "prog_instruction.h" -#include "prog_parameter.h" -#include "program.h" -#include "programopt.h" -#include "tnl/tnl.h" - - -static void st_bind_program( GLcontext *ctx, - GLenum target, - struct gl_program *prog ) -{ - struct st_context *st = st_context(ctx); - - switch (target) { - case GL_VERTEX_PROGRAM_ARB: - break; - case GL_FRAGMENT_PROGRAM_ARB: - st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - break; - } -} - -static struct gl_program *st_new_program( GLcontext *ctx, - GLenum target, - GLuint id ) -{ - struct st_context *st = st_context(ctx); - - switch (target) { - case GL_VERTEX_PROGRAM_ARB: - return _mesa_init_vertex_program(ctx, - CALLOC_STRUCT(gl_vertex_program), - target, - id); - - case GL_FRAGMENT_PROGRAM_ARB: { - struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program); - - prog->id = st->program_id++; - - return _mesa_init_fragment_program( ctx, - &prog->Base, - target, - id ); - } - - default: - return _mesa_new_program(ctx, target, id); - } -} - -static void st_delete_program( GLcontext *ctx, - struct gl_program *prog ) -{ - _mesa_delete_program( ctx, prog ); -} - - -static GLboolean st_is_program_native( GLcontext *ctx, - GLenum target, - struct gl_program *prog ) -{ - return GL_TRUE; -} - -static void st_program_string_notify( GLcontext *ctx, - GLenum target, - struct gl_program *prog ) -{ - if (target == GL_FRAGMENT_PROGRAM_ARB) { - struct st_context *st = st_context(ctx); - - if (prog == &st->ctx->FragmentProgram._Current->Base) - { - struct st_fragment_program *p = - (struct st_fragment_program *) prog; - - st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - - p->id = st->program_id++; -#if 0 - p->param_state = p->Base.Base.Parameters->StateFlags; - p->translated = 0; -#endif - - /* Gack! do this in the compiler: - */ - if (p->Base.FogOption) { - /* add extra instructions to do fog, then turn off FogOption field */ - _mesa_append_fog_code(ctx, &p->Base); - p->Base.FogOption = GL_NONE; - } - } - } - else if (target == GL_VERTEX_PROGRAM_ARB) { - - /* Also tell tnl about it: - */ - _tnl_program_string(ctx, target, prog); - } -} - - - -void st_init_cb_program( struct st_context *st ) -{ - struct dd_function_table *functions = &st->ctx->Driver; - - /* Need these flags: - */ - st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; - st->ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; - - - assert(functions->ProgramStringNotify == _tnl_program_string); - functions->BindProgram = st_bind_program; - functions->NewProgram = st_new_program; - functions->DeleteProgram = st_delete_program; - functions->IsProgramNative = st_is_program_native; - functions->ProgramStringNotify = st_program_string_notify; -} - - -void st_destroy_cb_program( struct st_context *st ) -{ -} - diff --git a/src/mesa/softpipe/state_tracker/st_context.c b/src/mesa/softpipe/state_tracker/st_context.c deleted file mode 100644 index 8a06dd88df..0000000000 --- a/src/mesa/softpipe/state_tracker/st_context.c +++ /dev/null @@ -1,76 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "imports.h" -#include "st_public.h" -#include "st_context.h" -#include "st_atom.h" -#include "st_draw.h" -#include "st_program.h" -#include "softpipe/sp_context.h" - -void st_invalidate_state(GLcontext * ctx, GLuint new_state) -{ - struct st_context *st = st_context(ctx); - - st->dirty.mesa |= new_state; - st->dirty.st |= ST_NEW_MESA; -} - - -struct st_context *st_create_context( GLcontext *ctx, - struct softpipe_context *softpipe ) -{ - struct st_context *st = CALLOC_STRUCT( st_context ); - - ctx->st = st; - - st->ctx = ctx; - st->softpipe = softpipe; - - st->dirty.mesa = ~0; - st->dirty.st = ~0; - - st_init_atoms( st ); - st_init_draw( st ); - st_init_cb_program( st ); - - return st; -} - - -void st_destroy_context( struct st_context *st ) -{ - st_destroy_atoms( st ); - st_destroy_draw( st ); - st_destroy_cb_program( st ); - st->softpipe->destroy( st->softpipe ); - FREE( st ); -} - - - diff --git a/src/mesa/softpipe/state_tracker/st_context.h b/src/mesa/softpipe/state_tracker/st_context.h deleted file mode 100644 index 317d377371..0000000000 --- a/src/mesa/softpipe/state_tracker/st_context.h +++ /dev/null @@ -1,111 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef ST_CONTEXT_H -#define ST_CONTEXT_H - -#include "mtypes.h" -#include "softpipe/sp_state.h" - - -struct st_context; -struct st_region; -struct st_texture_object; -struct st_texture_image; -struct st_fragment_program; - -#define ST_NEW_MESA 0x1 /* Mesa state has changed */ -#define ST_NEW_FRAGMENT_PROGRAM 0x2 - -struct st_state_flags { - GLuint mesa; - GLuint st; -}; - -struct st_tracked_state { - struct st_state_flags dirty; - void (*update)( struct st_context *st ); -}; - - - - -struct st_context -{ - GLcontext *ctx; - - struct softpipe_context *softpipe; - - /* Eventually will use a cache to feed the softpipe with - * create/bind/delete calls to constant state objects. Not yet - * though, we just shove random objects across the interface. - */ - struct { - struct softpipe_viewport viewport; - struct softpipe_setup_state setup; - struct softpipe_fs_state fs; - struct softpipe_blend_state blend; - struct softpipe_surface cbuf; - struct softpipe_clip_state clip; - struct softpipe_depth_state depth; - struct softpipe_scissor_rect scissor; - struct softpipe_poly_stipple poly_stipple; - struct softpipe_stencil_state stencil; - } state; - - struct { - struct st_tracked_state tracked_state; - } constants; - - struct { - struct gl_fragment_program *fragment_program; - } cb; - - /* State to be validated: - */ - struct st_tracked_state **atoms; - GLuint nr_atoms; - - struct st_state_flags dirty; - - /* Counter to track program string changes: - */ - GLuint program_id; - - GLfloat polygon_offset_scale; /* ?? */ -}; - - -/* Need this so that we can implement Mesa callbacks in this module. - */ -static INLINE struct st_context *st_context(GLcontext *ctx) -{ - return ctx->st; -} - - -#endif diff --git a/src/mesa/softpipe/state_tracker/st_draw.c b/src/mesa/softpipe/state_tracker/st_draw.c deleted file mode 100644 index 12646402ee..0000000000 --- a/src/mesa/softpipe/state_tracker/st_draw.c +++ /dev/null @@ -1,107 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "imports.h" - -#include "tnl/t_context.h" -#include "tnl/t_pipeline.h" - -#include "st_context.h" -#include "st_atom.h" -#include "st_draw.h" -#include "softpipe/sp_context.h" - -/* - * TNL stage which feedsinto the above. - * - * XXX: this needs to go into each driver using this code, because we - * cannot make the leap from ctx->draw_context in this file. The - * driver needs to customize tnl anyway, so this isn't a big deal. - */ -static GLboolean draw( GLcontext * ctx, struct tnl_pipeline_stage *stage ) -{ - struct st_context *st = st_context(ctx); - struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - - /* Validate driver and softpipe state: - */ - st_validate_state( st ); - - /* Call into the new draw code to handle the VB: - */ - st->softpipe->draw_vb( st->softpipe, VB ); - - /* Finished - */ - return GL_FALSE; -} - -const struct tnl_pipeline_stage st_draw = { - "check state and draw", - NULL, - NULL, - NULL, - NULL, - draw -}; - -static const struct tnl_pipeline_stage *intel_pipeline[] = { - &_tnl_vertex_transform_stage, - &_tnl_vertex_cull_stage, - &_tnl_normal_transform_stage, - &_tnl_lighting_stage, - &_tnl_fog_coordinate_stage, - &_tnl_texgen_stage, - &_tnl_texture_transform_stage, - &_tnl_point_attenuation_stage, - &_tnl_vertex_program_stage, - &st_draw, /* ADD: escape to softpipe */ - 0, -}; - -/* This is all a hack to keep using tnl until we have vertex programs - * up and running. - */ -void st_init_draw( struct st_context *st ) -{ - GLcontext *ctx = st->ctx; - - _tnl_destroy_pipeline( ctx ); - _tnl_install_pipeline( ctx, intel_pipeline ); -} - - -void st_destroy_draw( struct st_context *st ) -{ - /* Nothing to do. - */ -} diff --git a/src/mesa/softpipe/state_tracker/st_draw.h b/src/mesa/softpipe/state_tracker/st_draw.h deleted file mode 100644 index f51059706a..0000000000 --- a/src/mesa/softpipe/state_tracker/st_draw.h +++ /dev/null @@ -1,40 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef ST_DRAW_H -#define ST_DRAW_H - -void st_init_draw( struct st_context *st ); -void st_destroy_draw( struct st_context *st ); - -#endif diff --git a/src/mesa/softpipe/state_tracker/st_program.h b/src/mesa/softpipe/state_tracker/st_program.h deleted file mode 100644 index a47059d7a6..0000000000 --- a/src/mesa/softpipe/state_tracker/st_program.h +++ /dev/null @@ -1,68 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef ST_PROGRAM_H -#define ST_PROGRAM_H - -struct st_fragment_program -{ - struct gl_fragment_program Base; - GLboolean error; /* If program is malformed for any reason. */ - - GLuint id; /* String id, for tracking - * ProgramStringNotify changes. - */ - -#if 0 - GLfloat (*cbuffer)[4]; - GLuint nr_constants; - - /* Translate all the parameters, etc, into a constant buffer which - * we update on state changes. - */ - struct - { - GLuint reg; /* Constant idx */ - const GLfloat *values; /* Pointer to tracked values */ - } *param; - GLuint nr_params; - - GLuint param_state; -#endif -}; - - -void st_init_cb_program( struct st_context *st ); -void st_destroy_cb_program( struct st_context *st ); - -#endif diff --git a/src/mesa/softpipe/state_tracker/st_public.h b/src/mesa/softpipe/state_tracker/st_public.h deleted file mode 100644 index 8241dbd52d..0000000000 --- a/src/mesa/softpipe/state_tracker/st_public.h +++ /dev/null @@ -1,43 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef ST_PUBLIC_H -#define ST_PUBLIC_H - -#include "mtypes.h" - -struct st_context; -struct softpipe_context; - -struct st_context *st_create_context( GLcontext *ctx, - struct softpipe_context *softpipe); - -void st_destroy_context( struct st_context *st ); - -void st_invalidate_state(GLcontext * ctx, GLuint new_state); - -#endif diff --git a/src/mesa/state_tracker/Makefile b/src/mesa/state_tracker/Makefile new file mode 100644 index 0000000000..0ab1dc6e6b --- /dev/null +++ b/src/mesa/state_tracker/Makefile @@ -0,0 +1,2 @@ +default: + cd ../.. ; make \ No newline at end of file diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c new file mode 100644 index 0000000000..fdbfb9021d --- /dev/null +++ b/src/mesa/state_tracker/st_atom.c @@ -0,0 +1,171 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "glheader.h" +#include "context.h" + +#include "st_context.h" +#include "st_atom.h" + + + +/* This is used to initialize st->atoms[]. We could use this list + * directly except for a single atom, st_update_constants, which has a + * .dirty value which changes according to the parameters of the + * current fragment and vertex programs, and so cannot be a static + * value. + */ +static const struct st_tracked_state *atoms[] = +{ + &st_update_cbuf, + &st_update_clip, + &st_update_fs, + &st_update_setup, + &st_update_viewport, + &st_update_scissor, + &st_update_blend, + &st_update_stencil, + /* will be patched out at runtime */ +/* &st_update_constants */ +}; + + +void st_init_atoms( struct st_context *st ) +{ + GLuint i; + + st->atoms = _mesa_malloc(sizeof(atoms)); + st->nr_atoms = sizeof(atoms)/sizeof(*atoms); + memcpy(st->atoms, atoms, sizeof(atoms)); + + /* Patch in a pointer to the dynamic state atom: + */ + for (i = 0; i < st->nr_atoms; i++) + if (st->atoms[i] == &st_update_constants) + st->atoms[i] = &st->constants.tracked_state; + + memcpy(&st->constants.tracked_state, + &st_update_constants, + sizeof(st_update_constants)); +} + + +void st_destroy_atoms( struct st_context *st ) +{ + if (st->atoms) { + _mesa_free(st->atoms); + st->atoms = NULL; + } +} + + +/*********************************************************************** + */ + +static GLboolean check_state( const struct st_state_flags *a, + const struct st_state_flags *b ) +{ + return ((a->mesa & b->mesa) || + (a->st & b->st)); +} + +static void accumulate_state( struct st_state_flags *a, + const struct st_state_flags *b ) +{ + a->mesa |= b->mesa; + a->st |= b->st; +} + + +static void xor_states( struct st_state_flags *result, + const struct st_state_flags *a, + const struct st_state_flags *b ) +{ + result->mesa = a->mesa ^ b->mesa; + result->st = a->st ^ b->st; +} + + +/*********************************************************************** + * Update all derived state: + */ + +void st_validate_state( struct st_context *st ) +{ + struct st_state_flags *state = &st->dirty; + GLuint i; + + if (state->st == 0) + return; + + if (1) { + /* Debug version which enforces various sanity checks on the + * state flags which are generated and checked to help ensure + * state atoms are ordered correctly in the list. + */ + struct st_state_flags examined, prev; + _mesa_memset(&examined, 0, sizeof(examined)); + prev = *state; + + for (i = 0; i < st->nr_atoms; i++) { + const struct st_tracked_state *atom = st->atoms[i]; + struct st_state_flags generated; + + assert(atom->dirty.mesa || + atom->dirty.st); + assert(atom->update); + + if (check_state(state, &atom->dirty)) { + st->atoms[i]->update( st ); + } + + accumulate_state(&examined, &atom->dirty); + + /* generated = (prev ^ state) + * if (examined & generated) + * fail; + */ + xor_states(&generated, &prev, state); + assert(!check_state(&examined, &generated)); + prev = *state; + } + } + else { + const GLuint nr = st->nr_atoms; + + for (i = 0; i < nr; i++) { + if (check_state(state, &st->atoms[i]->dirty)) + st->atoms[i]->update( st ); + } + } + + memset(state, 0, sizeof(*state)); +} + + + diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h new file mode 100644 index 0000000000..1d8da46336 --- /dev/null +++ b/src/mesa/state_tracker/st_atom.h @@ -0,0 +1,59 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef ST_ATOM_H +#define ST_ATOM_H + +struct st_context; +struct st_tracked_state; + +void st_init_atoms( struct st_context *st ); +void st_destroy_atoms( struct st_context *st ); + + +void st_validate_state( struct st_context *st ); + + +const struct st_tracked_state st_update_cbuf; +const struct st_tracked_state st_update_clip; +const struct st_tracked_state st_update_depth; +const struct st_tracked_state st_update_fs; +const struct st_tracked_state st_update_setup; +const struct st_tracked_state st_update_viewport; +const struct st_tracked_state st_update_constants; +const struct st_tracked_state st_update_scissor; +const struct st_tracked_state st_update_blend; +const struct st_tracked_state st_update_stencil; + + +#endif diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c new file mode 100644 index 0000000000..eabb9172ba --- /dev/null +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -0,0 +1,202 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "softpipe/sp_context.h" +#include "softpipe/sp_defines.h" + + +/** + * Convert GLenum blend tokens to softpipe tokens. + * Both blend factors and blend funcs are accepted. + */ +static GLuint +gl_blend_to_sp(GLenum blend) +{ + switch (blend) { + /* blend functions */ + case GL_FUNC_ADD: + return SP_BLEND_ADD; + case GL_FUNC_SUBTRACT: + return SP_BLEND_SUBTRACT; + case GL_FUNC_REVERSE_SUBTRACT: + return SP_BLEND_REVERSE_SUBTRACT; + case GL_MIN: + return SP_BLEND_MIN; + case GL_MAX: + return SP_BLEND_MAX; + + /* blend factors */ + case GL_ONE: + return SP_BLENDFACTOR_ONE; + case GL_SRC_COLOR: + return SP_BLENDFACTOR_SRC_COLOR; + case GL_SRC_ALPHA: + return SP_BLENDFACTOR_SRC_ALPHA; + case GL_DST_ALPHA: + return SP_BLENDFACTOR_DST_ALPHA; + case GL_DST_COLOR: + return SP_BLENDFACTOR_DST_COLOR; + case GL_SRC_ALPHA_SATURATE: + return SP_BLENDFACTOR_SRC_ALPHA_SATURATE; + case GL_CONSTANT_COLOR: + return SP_BLENDFACTOR_CONST_COLOR; + case GL_CONSTANT_ALPHA: + return SP_BLENDFACTOR_CONST_ALPHA; + /* + return SP_BLENDFACTOR_SRC1_COLOR; + return SP_BLENDFACTOR_SRC1_ALPHA; + */ + case GL_ZERO: + return SP_BLENDFACTOR_ZERO; + case GL_ONE_MINUS_SRC_COLOR: + return SP_BLENDFACTOR_INV_SRC_COLOR; + case GL_ONE_MINUS_SRC_ALPHA: + return SP_BLENDFACTOR_INV_SRC_ALPHA; + case GL_ONE_MINUS_DST_COLOR: + return SP_BLENDFACTOR_INV_DST_ALPHA; + case GL_ONE_MINUS_DST_ALPHA: + return SP_BLENDFACTOR_INV_DST_COLOR; + case GL_ONE_MINUS_CONSTANT_COLOR: + return SP_BLENDFACTOR_INV_CONST_COLOR; + case GL_ONE_MINUS_CONSTANT_ALPHA: + return SP_BLENDFACTOR_INV_CONST_ALPHA; + /* + return SP_BLENDFACTOR_INV_SRC1_COLOR; + return SP_BLENDFACTOR_INV_SRC1_ALPHA; + */ + default: + assert("invalid GL token in gl_blend_to_sp()" == NULL); + return 0; + } +} + + +/** + * Convert GLenum logicop tokens to softpipe tokens. + */ +static GLuint +gl_logicop_to_sp(GLenum logicop) +{ + switch (logicop) { + case GL_CLEAR: + return SP_LOGICOP_CLEAR; + case GL_NOR: + return SP_LOGICOP_NOR; + case GL_AND_INVERTED: + return SP_LOGICOP_AND_INVERTED; + case GL_COPY_INVERTED: + return SP_LOGICOP_COPY_INVERTED; + case GL_AND_REVERSE: + return SP_LOGICOP_AND_REVERSE; + case GL_INVERT: + return SP_LOGICOP_INVERT; + case GL_XOR: + return SP_LOGICOP_XOR; + case GL_NAND: + return SP_LOGICOP_NAND; + case GL_AND: + return SP_LOGICOP_AND; + case GL_EQUIV: + return SP_LOGICOP_EQUIV; + case GL_NOOP: + return SP_LOGICOP_NOOP; + case GL_OR_INVERTED: + return SP_LOGICOP_OR_INVERTED; + case GL_COPY: + return SP_LOGICOP_COPY; + case GL_OR_REVERSE: + return SP_LOGICOP_OR_REVERSE; + case GL_OR: + return SP_LOGICOP_OR; + case GL_SET: + return SP_LOGICOP_SET; + default: + assert("invalid GL token in gl_logicop_to_sp()" == NULL); + return 0; + } +} + + +static void +update_blend( struct st_context *st ) +{ + struct softpipe_blend_state blend; + + memset(&blend, 0, sizeof(blend)); + + if (st->ctx->Color.ColorLogicOpEnabled || + (st->ctx->Color.BlendEnabled && + st->ctx->Color.BlendEquationRGB == GL_LOGIC_OP)) { + /* logicop enabled */ + blend.logicop_enable = 1; + blend.logicop_func = gl_logicop_to_sp(st->ctx->Color.LogicOp); + } + else if (st->ctx->Color.BlendEnabled) { + /* blending enabled */ + blend.blend_enable = 1; + + blend.rgb_func = gl_blend_to_sp(st->ctx->Color.BlendEquationRGB); + blend.rgb_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcRGB); + blend.rgb_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstRGB); + + blend.alpha_func = gl_blend_to_sp(st->ctx->Color.BlendEquationA); + blend.alpha_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcA); + blend.alpha_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstA); + } + else { + /* no blending / logicop */ + } + + if (memcmp(&blend, &st->state.blend, sizeof(blend)) != 0) { + /* state has changed */ + st->state.blend = blend; /* struct copy */ + st->softpipe->set_blend_state(st->softpipe, &blend); /* set new state */ + } +} + + +const struct st_tracked_state st_update_blend = { + .dirty = { + .mesa = (_NEW_COLOR), /* XXX _NEW_BLEND someday? */ + .st = 0, + }, + .update = update_blend +}; + + + + + diff --git a/src/mesa/state_tracker/st_atom_cbuf.c b/src/mesa/state_tracker/st_atom_cbuf.c new file mode 100644 index 0000000000..cd707ec5ef --- /dev/null +++ b/src/mesa/state_tracker/st_atom_cbuf.c @@ -0,0 +1,72 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + +extern GLboolean xmesa_get_cbuf_details( GLcontext *ctx, + void **ptr, + GLuint *cpp, + GLint *stride, + GLuint *format ); + + +/* This is a hack to work with the X11 driver as a test harness + */ +static void update_cbuf_state( struct st_context *st ) +{ + struct softpipe_surface cbuf; + GLboolean ok; + + ok = xmesa_get_cbuf_details( st->ctx, + (void **)&cbuf.ptr, + &cbuf.cpp, + &cbuf.stride, + &cbuf.format ); + + assert(ok); + + if (memcmp(&cbuf, &st->state.cbuf, sizeof(cbuf)) != 0) { + st->state.cbuf = cbuf; + st->softpipe->set_cbuf_state( st->softpipe, &cbuf ); + } +} + +const struct st_tracked_state st_update_cbuf = { + .dirty = { + .mesa = _NEW_BUFFERS, + .st = 0, + }, + .update = update_cbuf_state +}; + diff --git a/src/mesa/state_tracker/st_atom_clip.c b/src/mesa/state_tracker/st_atom_clip.c new file mode 100644 index 0000000000..710d6ffc84 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_clip.c @@ -0,0 +1,75 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + + +/* Second state atom for user clip planes: + */ +static void update_clip( struct st_context *st ) +{ + struct softpipe_clip_state clip; + GLuint i; + + memset(&clip, 0, sizeof(clip)); + + for (i = 0; i < SP_MAX_CLIP_PLANES; i++) { + if (st->ctx->Transform.ClipPlanesEnabled & (1 << i)) { + memcpy(clip.ucp[clip.nr], + st->ctx->Transform._ClipUserPlane[i], + sizeof(clip.ucp[0])); + clip.nr++; + } + } + + if (memcmp(&clip, &st->state.clip, sizeof(clip)) != 0) { + st->state.clip = clip; + st->softpipe->set_clip_state(st->softpipe, &clip); + } +} + + +const struct st_tracked_state st_update_clip = { + .dirty = { + .mesa = (_NEW_TRANSFORM), + .st = 0, + }, + .update = update_clip +}; + + + + + diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c new file mode 100644 index 0000000000..5532abc8fd --- /dev/null +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -0,0 +1,88 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "softpipe/sp_context.h" +#include "softpipe/sp_defines.h" + + +/** + * Convert GLenum depth func tokens to softpipe tokens. + */ +static GLuint +gl_depth_func_to_sp(GLenum func) +{ + /* Same values, just biased */ + assert(SP_DEPTH_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(SP_DEPTH_FUNC_LESS == GL_LESS - GL_NEVER); + assert(SP_DEPTH_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(SP_DEPTH_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(SP_DEPTH_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(SP_DEPTH_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(SP_DEPTH_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(SP_DEPTH_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(func >= GL_NEVER); + assert(func <= GL_ALWAYS); + return func - GL_NEVER; +} + + +static void +update_depth( struct st_context *st ) +{ + struct softpipe_depth_state depth; + + memset(&depth, 0, sizeof(depth)); + + depth.enabled = st->ctx->Depth.Test; + depth.writemask = st->ctx->Depth.Mask; + depth.func = gl_depth_func_to_sp(st->ctx->Depth.Func); + depth.clear = st->ctx->Depth.Clear; + + if (memcmp(&depth, &st->state.depth, sizeof(depth)) != 0) { + /* state has changed */ + st->state.depth = depth; /* struct copy */ + st->softpipe->set_depth_state(st->softpipe, &depth); /* set new state */ + } +} + + +const struct st_tracked_state st_update_depth = { + .dirty = { + .mesa = (_NEW_DEPTH), + .st = 0, + }, + .update = update_depth +}; diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c new file mode 100644 index 0000000000..ca109d2d34 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -0,0 +1,56 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + + +static void update_fs( struct st_context *st ) +{ + struct softpipe_fs_state fs; + + fs.fp = st->ctx->FragmentProgram._Current; + + if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0) { + st->state.fs = fs; + st->softpipe->set_fs_state(st->softpipe, &fs); + } +} + + +const struct st_tracked_state st_update_fs = { + .dirty = { + .mesa = 0, + .st = ST_NEW_FRAGMENT_PROGRAM, + }, + .update = update_fs +}; diff --git a/src/mesa/state_tracker/st_atom_scissor.c b/src/mesa/state_tracker/st_atom_scissor.c new file mode 100644 index 0000000000..105c2a6dd9 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_scissor.c @@ -0,0 +1,88 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + + +/** + * Scissor depends on the scissor box, and the framebuffer dimensions. + */ +static void +update_scissor( struct st_context *st ) +{ + struct softpipe_scissor_rect scissor; + const struct gl_framebuffer *fb = st->ctx->DrawBuffer; + + scissor.minx = 0; + scissor.miny = 0; + scissor.maxx = fb->Width; + scissor.maxy = fb->Height; + + if (st->ctx->Scissor.Enabled) { + if (st->ctx->Scissor.X > scissor.minx) + scissor.minx = st->ctx->Scissor.X; + if (st->ctx->Scissor.Y > scissor.miny) + scissor.miny = st->ctx->Scissor.Y; + + if (st->ctx->Scissor.X + st->ctx->Scissor.Width < scissor.maxx) + scissor.maxx = st->ctx->Scissor.X + st->ctx->Scissor.Width; + if (st->ctx->Scissor.Y + st->ctx->Scissor.Height < scissor.maxy) + scissor.maxy = st->ctx->Scissor.Y + st->ctx->Scissor.Height; + + /* check for null space */ + if (scissor.minx >= scissor.maxx || scissor.miny >= scissor.maxy) + scissor.minx = scissor.miny = scissor.maxx = scissor.maxy = 0; + } + + if (memcmp(&scissor, &st->state.scissor, sizeof(scissor)) != 0) { + /* state has changed */ + st->state.scissor = scissor; /* struct copy */ + st->softpipe->set_scissor_rect(st->softpipe, &scissor); /* activate */ + } +} + + +const struct st_tracked_state st_update_scissor = { + .dirty = { + .mesa = (_NEW_SCISSOR | _NEW_BUFFERS), + .st = 0, + }, + .update = update_scissor +}; + + + + + diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c new file mode 100644 index 0000000000..2a582ea36c --- /dev/null +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -0,0 +1,175 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + +static GLuint translate_fill( GLenum mode ) +{ + switch (mode) { + case GL_POINT: return FILL_POINT; + case GL_LINE: return FILL_LINE; + case GL_FILL: return FILL_TRI; + default: assert(0); return 0; + } +} + +static GLboolean get_offset_flag( GLuint fill_mode, + const struct gl_polygon_attrib *Polygon ) +{ + switch (fill_mode) { + case FILL_POINT: return Polygon->OffsetPoint; + case FILL_LINE: return Polygon->OffsetLine; + case FILL_TRI: return Polygon->OffsetFill; + default: assert(0); return 0; + } +} + + +static void update_setup_state( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + struct softpipe_setup_state setup; + + memset(&setup, 0, sizeof(setup)); + + /* _NEW_POLYGON, _NEW_BUFFERS + */ + { + setup.front_winding = WINDING_CW; + + if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) + setup.front_winding ^= WINDING_BOTH; + + if (ctx->Polygon.FrontFace != GL_CCW) + setup.front_winding ^= WINDING_BOTH; + } + + /* _NEW_LIGHT + */ + if (ctx->Light.ShadeModel == GL_FLAT) + setup.flatshade = 1; + + /* _NEW_LIGHT + * + * Not sure about the light->enabled requirement - does this still + * apply?? + */ + if (ctx->Light.Enabled && + ctx->Light.Model.TwoSide) + setup.light_twoside = 1; + + + /* _NEW_POLYGON + */ + if (ctx->Polygon.CullFlag) { + if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) { + setup.cull_mode = WINDING_BOTH; + } + else if (ctx->Polygon.CullFaceMode == GL_FRONT) { + setup.cull_mode = setup.front_winding; + } + else { + setup.cull_mode = setup.front_winding ^ WINDING_BOTH; + } + } + + /* _NEW_POLYGON + */ + { + GLuint fill_front = translate_fill( ctx->Polygon.FrontMode ); + GLuint fill_back = translate_fill( ctx->Polygon.BackMode ); + + if (setup.front_winding == WINDING_CW) { + setup.fill_cw = fill_front; + setup.fill_ccw = fill_back; + } + else { + setup.fill_cw = fill_back; + setup.fill_ccw = fill_front; + } + + /* Simplify when culling is active: + */ + if (setup.cull_mode & WINDING_CW) { + setup.fill_cw = setup.fill_ccw; + } + + if (setup.cull_mode & WINDING_CCW) { + setup.fill_ccw = setup.fill_cw; + } + } + + /* Hardware does offset for filled prims, but need to do it in + * software for unfilled. + * + * _NEW_POLYGON + */ + if (setup.fill_cw != FILL_TRI) + setup.offset_cw = get_offset_flag( setup.fill_cw, + &ctx->Polygon ); + + if (setup.fill_ccw != FILL_TRI) + setup.offset_ccw = get_offset_flag( setup.fill_ccw, + &ctx->Polygon ); + + + /* _NEW_BUFFERS, _NEW_POLYGON + */ + if (setup.fill_cw != FILL_TRI || + setup.fill_ccw != FILL_TRI) + { + GLfloat mrd = (ctx->DrawBuffer ? + ctx->DrawBuffer->_MRD : + 1.0); + + setup.offset_units = ctx->Polygon.OffsetFactor * mrd; + setup.offset_scale = (ctx->Polygon.OffsetUnits * mrd * + st->polygon_offset_scale); + } + + + if (memcmp(&setup, &st->state.setup, sizeof(setup)) != 0) { + st->state.setup = setup; + st->softpipe->set_setup_state( st->softpipe, &setup ); + } +} + +const struct st_tracked_state st_update_setup = { + .dirty = { + .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_BUFFERS), + .st = 0, + }, + .update = update_setup_state +}; diff --git a/src/mesa/state_tracker/st_atom_stencil.c b/src/mesa/state_tracker/st_atom_stencil.c new file mode 100644 index 0000000000..32610c3d24 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_stencil.c @@ -0,0 +1,140 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "softpipe/sp_context.h" +#include "softpipe/sp_defines.h" + + +/** + * Convert GLenum stencil func tokens to softpipe tokens. + */ +static GLuint +gl_stencil_func_to_sp(GLenum func) +{ + /* Same values, just biased */ + assert(SP_STENCIL_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(SP_STENCIL_FUNC_LESS == GL_LESS - GL_NEVER); + assert(SP_STENCIL_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(SP_STENCIL_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(SP_STENCIL_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(SP_STENCIL_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(SP_STENCIL_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(SP_STENCIL_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(func >= GL_NEVER); + assert(func <= GL_ALWAYS); + return func - GL_NEVER; +} + + +/** + * Convert GLenum stencil op tokens to softpipe tokens. + */ +static GLuint +gl_stencil_op_to_sp(GLenum func) +{ + switch (func) { + case GL_KEEP: + return SP_STENCIL_OP_KEEP; + case GL_ZERO: + return SP_STENCIL_OP_ZERO; + case GL_REPLACE: + return SP_STENCIL_OP_REPLACE; + case GL_INCR: + return SP_STENCIL_OP_INCR; + case GL_DECR: + return SP_STENCIL_OP_DECR; + case GL_INCR_WRAP: + return SP_STENCIL_OP_INCR_WRAP; + case GL_DECR_WRAP: + return SP_STENCIL_OP_DECR_WRAP; + case GL_INVERT: + return SP_STENCIL_OP_INVERT; + default: + assert("invalid GL token in gl_stencil_op_to_sp()" == NULL); + return 0; + } +} + + +static void +update_stencil( struct st_context *st ) +{ + struct softpipe_stencil_state stencil; + + memset(&stencil, 0, sizeof(stencil)); + + if (st->ctx->Stencil.Enabled) { + stencil.front_enabled = 1; + stencil.front_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[0]); + stencil.front_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[0]); + stencil.front_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[0]); + stencil.front_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[0]); + stencil.ref_value[0] = st->ctx->Stencil.Ref[0] & 0xff; + stencil.value_mask[0] = st->ctx->Stencil.ValueMask[0] & 0xff; + stencil.write_mask[0] = st->ctx->Stencil.WriteMask[0] & 0xff; + if (st->ctx->Stencil.TestTwoSide) { + stencil.back_enabled = 1; + stencil.back_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[1]); + stencil.back_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[1]); + stencil.back_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[1]); + stencil.back_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[1]); + stencil.ref_value[1] = st->ctx->Stencil.Ref[1] & 0xff; + stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1] & 0xff; + stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1] & 0xff; + } + stencil.clear_value = st->ctx->Stencil.Clear & 0xff; + } + + if (memcmp(&stencil, &st->state.stencil, sizeof(stencil)) != 0) { + /* state has changed */ + st->state.stencil = stencil; /* struct copy */ + st->softpipe->set_stencil_state(st->softpipe, &stencil); /* set new state */ + } +} + + +const struct st_tracked_state st_update_stencil = { + .dirty = { + .mesa = (_NEW_STENCIL), + .st = 0, + }, + .update = update_stencil +}; + + + + + diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c new file mode 100644 index 0000000000..af896e2e31 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -0,0 +1,117 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "context.h" +#include "colormac.h" +#include "st_context.h" +#include "softpipe/sp_context.h" +#include "st_atom.h" + + + + + +/** + * Update the viewport transformation matrix. Depends on: + * - viewport pos/size + * - depthrange + * - window pos/size or FBO size + */ +static void update_viewport( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + const GLframebuffer *DrawBuffer = ctx->DrawBuffer; + GLfloat yScale = 1.0; + GLfloat yBias = 0.0; + + /* _NEW_BUFFERS + */ + if (DrawBuffer) { + +#if 0 + if (DrawBuffer->Name) { + /* User created FBO */ + struct st_renderbuffer *irb + = st_renderbuffer(DrawBuffer->_ColorDrawBuffers[0][0]); + if (irb && !irb->RenderToTexture) { + /* y=0=top */ + yScale = -1.0; + yBias = irb->Base.Height; + } + else { + /* y=0=bottom */ + yScale = 1.0; + yBias = 0.0; + } + } + else + { + /* window buffer, y=0=top */ + yScale = -1.0; + yBias = DrawBuffer->Height; + } +#endif + } + + { + /* _NEW_VIEWPORT + */ + GLfloat x = ctx->Viewport.X; + GLfloat y = ctx->Viewport.Y; + GLfloat z = ctx->Viewport.Near; + GLfloat half_width = ctx->Viewport.Width / 2.0; + GLfloat half_height = ctx->Viewport.Height / 2.0; + GLfloat half_depth = (ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; + + struct softpipe_viewport vp; + + vp.scale[0] = half_width; + vp.scale[1] = half_height * yScale; + vp.scale[2] = half_depth; + vp.scale[3] = 1.0; + + vp.translate[0] = (half_width + x); + vp.translate[1] = (half_height + y) * yScale + yBias; + vp.translate[2] = (half_depth + z); + vp.translate[3] = 0.0; + + if (memcmp(&vp, &st->state.viewport, sizeof(vp)) != 0) { + st->state.viewport = vp; + st->softpipe->set_viewport(st->softpipe, &vp); + } + } +} + + +const struct st_tracked_state st_update_viewport = { + .dirty = { + .mesa = _NEW_BUFFERS | _NEW_VIEWPORT, + .st = 0, + }, + .update = update_viewport +}; diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c new file mode 100644 index 0000000000..67a589ca8d --- /dev/null +++ b/src/mesa/state_tracker/st_cb_program.c @@ -0,0 +1,165 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "st_context.h" +#include "st_program.h" + +#include "glheader.h" +#include "macros.h" +#include "enums.h" +#include "prog_instruction.h" +#include "prog_parameter.h" +#include "program.h" +#include "programopt.h" +#include "tnl/tnl.h" + + +static void st_bind_program( GLcontext *ctx, + GLenum target, + struct gl_program *prog ) +{ + struct st_context *st = st_context(ctx); + + switch (target) { + case GL_VERTEX_PROGRAM_ARB: + break; + case GL_FRAGMENT_PROGRAM_ARB: + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; + break; + } +} + +static struct gl_program *st_new_program( GLcontext *ctx, + GLenum target, + GLuint id ) +{ + struct st_context *st = st_context(ctx); + + switch (target) { + case GL_VERTEX_PROGRAM_ARB: + return _mesa_init_vertex_program(ctx, + CALLOC_STRUCT(gl_vertex_program), + target, + id); + + case GL_FRAGMENT_PROGRAM_ARB: { + struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program); + + prog->id = st->program_id++; + + return _mesa_init_fragment_program( ctx, + &prog->Base, + target, + id ); + } + + default: + return _mesa_new_program(ctx, target, id); + } +} + +static void st_delete_program( GLcontext *ctx, + struct gl_program *prog ) +{ + _mesa_delete_program( ctx, prog ); +} + + +static GLboolean st_is_program_native( GLcontext *ctx, + GLenum target, + struct gl_program *prog ) +{ + return GL_TRUE; +} + +static void st_program_string_notify( GLcontext *ctx, + GLenum target, + struct gl_program *prog ) +{ + if (target == GL_FRAGMENT_PROGRAM_ARB) { + struct st_context *st = st_context(ctx); + + if (prog == &st->ctx->FragmentProgram._Current->Base) + { + struct st_fragment_program *p = + (struct st_fragment_program *) prog; + + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; + + p->id = st->program_id++; +#if 0 + p->param_state = p->Base.Base.Parameters->StateFlags; + p->translated = 0; +#endif + + /* Gack! do this in the compiler: + */ + if (p->Base.FogOption) { + /* add extra instructions to do fog, then turn off FogOption field */ + _mesa_append_fog_code(ctx, &p->Base); + p->Base.FogOption = GL_NONE; + } + } + } + else if (target == GL_VERTEX_PROGRAM_ARB) { + + /* Also tell tnl about it: + */ + _tnl_program_string(ctx, target, prog); + } +} + + + +void st_init_cb_program( struct st_context *st ) +{ + struct dd_function_table *functions = &st->ctx->Driver; + + /* Need these flags: + */ + st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; + st->ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; + + + assert(functions->ProgramStringNotify == _tnl_program_string); + functions->BindProgram = st_bind_program; + functions->NewProgram = st_new_program; + functions->DeleteProgram = st_delete_program; + functions->IsProgramNative = st_is_program_native; + functions->ProgramStringNotify = st_program_string_notify; +} + + +void st_destroy_cb_program( struct st_context *st ) +{ +} + diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c new file mode 100644 index 0000000000..8a06dd88df --- /dev/null +++ b/src/mesa/state_tracker/st_context.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "imports.h" +#include "st_public.h" +#include "st_context.h" +#include "st_atom.h" +#include "st_draw.h" +#include "st_program.h" +#include "softpipe/sp_context.h" + +void st_invalidate_state(GLcontext * ctx, GLuint new_state) +{ + struct st_context *st = st_context(ctx); + + st->dirty.mesa |= new_state; + st->dirty.st |= ST_NEW_MESA; +} + + +struct st_context *st_create_context( GLcontext *ctx, + struct softpipe_context *softpipe ) +{ + struct st_context *st = CALLOC_STRUCT( st_context ); + + ctx->st = st; + + st->ctx = ctx; + st->softpipe = softpipe; + + st->dirty.mesa = ~0; + st->dirty.st = ~0; + + st_init_atoms( st ); + st_init_draw( st ); + st_init_cb_program( st ); + + return st; +} + + +void st_destroy_context( struct st_context *st ) +{ + st_destroy_atoms( st ); + st_destroy_draw( st ); + st_destroy_cb_program( st ); + st->softpipe->destroy( st->softpipe ); + FREE( st ); +} + + + diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h new file mode 100644 index 0000000000..317d377371 --- /dev/null +++ b/src/mesa/state_tracker/st_context.h @@ -0,0 +1,111 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ST_CONTEXT_H +#define ST_CONTEXT_H + +#include "mtypes.h" +#include "softpipe/sp_state.h" + + +struct st_context; +struct st_region; +struct st_texture_object; +struct st_texture_image; +struct st_fragment_program; + +#define ST_NEW_MESA 0x1 /* Mesa state has changed */ +#define ST_NEW_FRAGMENT_PROGRAM 0x2 + +struct st_state_flags { + GLuint mesa; + GLuint st; +}; + +struct st_tracked_state { + struct st_state_flags dirty; + void (*update)( struct st_context *st ); +}; + + + + +struct st_context +{ + GLcontext *ctx; + + struct softpipe_context *softpipe; + + /* Eventually will use a cache to feed the softpipe with + * create/bind/delete calls to constant state objects. Not yet + * though, we just shove random objects across the interface. + */ + struct { + struct softpipe_viewport viewport; + struct softpipe_setup_state setup; + struct softpipe_fs_state fs; + struct softpipe_blend_state blend; + struct softpipe_surface cbuf; + struct softpipe_clip_state clip; + struct softpipe_depth_state depth; + struct softpipe_scissor_rect scissor; + struct softpipe_poly_stipple poly_stipple; + struct softpipe_stencil_state stencil; + } state; + + struct { + struct st_tracked_state tracked_state; + } constants; + + struct { + struct gl_fragment_program *fragment_program; + } cb; + + /* State to be validated: + */ + struct st_tracked_state **atoms; + GLuint nr_atoms; + + struct st_state_flags dirty; + + /* Counter to track program string changes: + */ + GLuint program_id; + + GLfloat polygon_offset_scale; /* ?? */ +}; + + +/* Need this so that we can implement Mesa callbacks in this module. + */ +static INLINE struct st_context *st_context(GLcontext *ctx) +{ + return ctx->st; +} + + +#endif diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c new file mode 100644 index 0000000000..12646402ee --- /dev/null +++ b/src/mesa/state_tracker/st_draw.c @@ -0,0 +1,107 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "imports.h" + +#include "tnl/t_context.h" +#include "tnl/t_pipeline.h" + +#include "st_context.h" +#include "st_atom.h" +#include "st_draw.h" +#include "softpipe/sp_context.h" + +/* + * TNL stage which feedsinto the above. + * + * XXX: this needs to go into each driver using this code, because we + * cannot make the leap from ctx->draw_context in this file. The + * driver needs to customize tnl anyway, so this isn't a big deal. + */ +static GLboolean draw( GLcontext * ctx, struct tnl_pipeline_stage *stage ) +{ + struct st_context *st = st_context(ctx); + struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; + + /* Validate driver and softpipe state: + */ + st_validate_state( st ); + + /* Call into the new draw code to handle the VB: + */ + st->softpipe->draw_vb( st->softpipe, VB ); + + /* Finished + */ + return GL_FALSE; +} + +const struct tnl_pipeline_stage st_draw = { + "check state and draw", + NULL, + NULL, + NULL, + NULL, + draw +}; + +static const struct tnl_pipeline_stage *intel_pipeline[] = { + &_tnl_vertex_transform_stage, + &_tnl_vertex_cull_stage, + &_tnl_normal_transform_stage, + &_tnl_lighting_stage, + &_tnl_fog_coordinate_stage, + &_tnl_texgen_stage, + &_tnl_texture_transform_stage, + &_tnl_point_attenuation_stage, + &_tnl_vertex_program_stage, + &st_draw, /* ADD: escape to softpipe */ + 0, +}; + +/* This is all a hack to keep using tnl until we have vertex programs + * up and running. + */ +void st_init_draw( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + + _tnl_destroy_pipeline( ctx ); + _tnl_install_pipeline( ctx, intel_pipeline ); +} + + +void st_destroy_draw( struct st_context *st ) +{ + /* Nothing to do. + */ +} diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h new file mode 100644 index 0000000000..f51059706a --- /dev/null +++ b/src/mesa/state_tracker/st_draw.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef ST_DRAW_H +#define ST_DRAW_H + +void st_init_draw( struct st_context *st ); +void st_destroy_draw( struct st_context *st ); + +#endif diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h new file mode 100644 index 0000000000..a47059d7a6 --- /dev/null +++ b/src/mesa/state_tracker/st_program.h @@ -0,0 +1,68 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef ST_PROGRAM_H +#define ST_PROGRAM_H + +struct st_fragment_program +{ + struct gl_fragment_program Base; + GLboolean error; /* If program is malformed for any reason. */ + + GLuint id; /* String id, for tracking + * ProgramStringNotify changes. + */ + +#if 0 + GLfloat (*cbuffer)[4]; + GLuint nr_constants; + + /* Translate all the parameters, etc, into a constant buffer which + * we update on state changes. + */ + struct + { + GLuint reg; /* Constant idx */ + const GLfloat *values; /* Pointer to tracked values */ + } *param; + GLuint nr_params; + + GLuint param_state; +#endif +}; + + +void st_init_cb_program( struct st_context *st ); +void st_destroy_cb_program( struct st_context *st ); + +#endif diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h new file mode 100644 index 0000000000..8241dbd52d --- /dev/null +++ b/src/mesa/state_tracker/st_public.h @@ -0,0 +1,43 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ST_PUBLIC_H +#define ST_PUBLIC_H + +#include "mtypes.h" + +struct st_context; +struct softpipe_context; + +struct st_context *st_create_context( GLcontext *ctx, + struct softpipe_context *softpipe); + +void st_destroy_context( struct st_context *st ); + +void st_invalidate_state(GLcontext * ctx, GLuint new_state); + +#endif -- cgit v1.2.3 From 943964a1e5bad86bdceb0a06d60fb3b302ebce6a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 14 Jun 2007 18:23:43 +0100 Subject: Rename directories again?! Some git wierdness going on. --- src/mesa/drivers/x11/xm_api.c | 6 +- src/mesa/drivers/x11/xm_dd.c | 4 +- src/mesa/pipe/p_context.h | 54 +- src/mesa/pipe/p_defines.h | 164 +++--- src/mesa/pipe/p_state.h | 62 +-- src/mesa/pipe/softpipe/g_context.c | 135 ----- src/mesa/pipe/softpipe/g_context.h | 149 ----- src/mesa/pipe/softpipe/g_draw.c | 696 ----------------------- src/mesa/pipe/softpipe/g_draw.h | 56 -- src/mesa/pipe/softpipe/g_headers.h | 116 ---- src/mesa/pipe/softpipe/g_prim.h | 90 --- src/mesa/pipe/softpipe/g_prim_clip.c | 429 --------------- src/mesa/pipe/softpipe/g_prim_cull.c | 119 ---- src/mesa/pipe/softpipe/g_prim_flatshade.c | 149 ----- src/mesa/pipe/softpipe/g_prim_offset.c | 157 ------ src/mesa/pipe/softpipe/g_prim_setup.c | 888 ------------------------------ src/mesa/pipe/softpipe/g_prim_setup.h | 121 ---- src/mesa/pipe/softpipe/g_prim_twoside.c | 149 ----- src/mesa/pipe/softpipe/g_prim_unfilled.c | 165 ------ src/mesa/pipe/softpipe/g_state.h | 61 -- src/mesa/pipe/softpipe/g_state_clip.c | 69 --- src/mesa/pipe/softpipe/g_state_derived.c | 136 ----- src/mesa/pipe/softpipe/g_state_fs.c | 50 -- src/mesa/pipe/softpipe/g_state_setup.c | 122 ---- src/mesa/pipe/softpipe/g_state_surface.c | 53 -- src/mesa/pipe/softpipe/g_surface.c | 153 ----- src/mesa/pipe/softpipe/g_surface.h | 115 ---- src/mesa/pipe/softpipe/g_tile.h | 42 -- src/mesa/pipe/softpipe/g_tile_fs.c | 203 ------- src/mesa/pipe/softpipe/g_tile_output.c | 92 ---- src/mesa/sources | 63 ++- src/mesa/state_tracker/st_atom_blend.c | 92 ++-- src/mesa/state_tracker/st_atom_cbuf.c | 6 +- src/mesa/state_tracker/st_atom_clip.c | 8 +- src/mesa/state_tracker/st_atom_depth.c | 26 +- src/mesa/state_tracker/st_atom_fs.c | 6 +- src/mesa/state_tracker/st_atom_scissor.c | 6 +- src/mesa/state_tracker/st_atom_setup.c | 6 +- src/mesa/state_tracker/st_atom_stencil.c | 44 +- src/mesa/state_tracker/st_atom_viewport.c | 6 +- src/mesa/state_tracker/st_context.c | 8 +- src/mesa/state_tracker/st_context.h | 26 +- src/mesa/state_tracker/st_draw.c | 8 +- src/mesa/state_tracker/st_public.h | 4 +- 44 files changed, 302 insertions(+), 4812 deletions(-) delete mode 100644 src/mesa/pipe/softpipe/g_context.c delete mode 100644 src/mesa/pipe/softpipe/g_context.h delete mode 100644 src/mesa/pipe/softpipe/g_draw.c delete mode 100644 src/mesa/pipe/softpipe/g_draw.h delete mode 100644 src/mesa/pipe/softpipe/g_headers.h delete mode 100644 src/mesa/pipe/softpipe/g_prim.h delete mode 100644 src/mesa/pipe/softpipe/g_prim_clip.c delete mode 100644 src/mesa/pipe/softpipe/g_prim_cull.c delete mode 100644 src/mesa/pipe/softpipe/g_prim_flatshade.c delete mode 100644 src/mesa/pipe/softpipe/g_prim_offset.c delete mode 100644 src/mesa/pipe/softpipe/g_prim_setup.c delete mode 100644 src/mesa/pipe/softpipe/g_prim_setup.h delete mode 100644 src/mesa/pipe/softpipe/g_prim_twoside.c delete mode 100644 src/mesa/pipe/softpipe/g_prim_unfilled.c delete mode 100644 src/mesa/pipe/softpipe/g_state.h delete mode 100644 src/mesa/pipe/softpipe/g_state_clip.c delete mode 100644 src/mesa/pipe/softpipe/g_state_derived.c delete mode 100644 src/mesa/pipe/softpipe/g_state_fs.c delete mode 100644 src/mesa/pipe/softpipe/g_state_setup.c delete mode 100644 src/mesa/pipe/softpipe/g_state_surface.c delete mode 100644 src/mesa/pipe/softpipe/g_surface.c delete mode 100644 src/mesa/pipe/softpipe/g_surface.h delete mode 100644 src/mesa/pipe/softpipe/g_tile.h delete mode 100644 src/mesa/pipe/softpipe/g_tile_fs.c delete mode 100644 src/mesa/pipe/softpipe/g_tile_output.c (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index af1a169a76..396a0923e9 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -80,8 +80,8 @@ #include "tnl/t_pipeline.h" #include "drivers/common/driverfuncs.h" -#include "softpipe/state_tracker/st_public.h" -#include "softpipe/generic/g_context.h" +#include "state_tracker/st_public.h" +#include "pipe/softpipe/sp_context.h" /** * Global X driver lock @@ -1561,7 +1561,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) st_create_context( mesaCtx, - generic_create() ); + softpipe_create() ); return c; } diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index 63e7115e80..eb59d1d05e 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -53,8 +53,8 @@ #include "tnl/tnl.h" #include "tnl/t_context.h" -#include "softpipe/generic/g_context.h" -#include "softpipe/state_tracker/st_public.h" +#include "pipe/softpipe/sp_context.h" +#include "state_tracker/st_public.h" /* diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 716ddd27a9..dc7a80744a 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -25,12 +25,14 @@ * **************************************************************************/ -#ifndef SP_CONTEXT_H -#define SP_CONTEXT_H +#ifndef PIPE_CONTEXT_H +#define PIPE_CONTEXT_H #include "mtypes.h" -extern struct softpipe_context *generic_create( void ); +/* Kludge: + */ +extern struct pipe_context *softpipe_create( void ); /* Drawing currently kludged up via the existing tnl/ module. */ @@ -41,48 +43,48 @@ struct vertex_buffer; * Software pipeline rendering context. Basically a collection of * state setting functions, plus VBO drawing entrypoint. */ -struct softpipe_context { +struct pipe_context { - void (*destroy)( struct softpipe_context * ); + void (*destroy)( struct pipe_context * ); /* * Drawing */ - void (*draw_vb)( struct softpipe_context *softpipe, + void (*draw_vb)( struct pipe_context *pipe, struct vertex_buffer *VB ); /* * State functions */ - void (*set_blend_state)( struct softpipe_context *, - const struct softpipe_blend_state * ); + void (*set_blend_state)( struct pipe_context *, + const struct pipe_blend_state * ); - void (*set_cbuf_state)( struct softpipe_context *, - const struct softpipe_surface * ); + void (*set_cbuf_state)( struct pipe_context *, + const struct pipe_surface * ); - void (*set_clip_state)( struct softpipe_context *, - const struct softpipe_clip_state * ); + void (*set_clip_state)( struct pipe_context *, + const struct pipe_clip_state * ); - void (*set_depth_state)( struct softpipe_context *, - const struct softpipe_depth_state * ); + void (*set_depth_state)( struct pipe_context *, + const struct pipe_depth_state * ); - void (*set_fs_state)( struct softpipe_context *, - const struct softpipe_fs_state * ); + void (*set_fs_state)( struct pipe_context *, + const struct pipe_fs_state * ); - void (*set_polygon_stipple)( struct softpipe_context *, - const struct softpipe_poly_stipple * ); + void (*set_polygon_stipple)( struct pipe_context *, + const struct pipe_poly_stipple * ); - void (*set_setup_state)( struct softpipe_context *, - const struct softpipe_setup_state * ); + void (*set_setup_state)( struct pipe_context *, + const struct pipe_setup_state * ); - void (*set_scissor_rect)( struct softpipe_context *, - const struct softpipe_scissor_rect * ); + void (*set_scissor_rect)( struct pipe_context *, + const struct pipe_scissor_rect * ); - void (*set_stencil_state)( struct softpipe_context *, - const struct softpipe_stencil_state * ); + void (*set_stencil_state)( struct pipe_context *, + const struct pipe_stencil_state * ); - void (*set_viewport)( struct softpipe_context *, - const struct softpipe_viewport * ); + void (*set_viewport)( struct pipe_context *, + const struct pipe_viewport * ); }; diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index c381865dab..1ca717e7ec 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -25,97 +25,97 @@ * **************************************************************************/ -#ifndef SP_DEFINES_H -#define SP_DEFINES_H +#ifndef PIPE_DEFINES_H +#define PIPE_DEFINES_H -#define SP_BLENDFACTOR_ONE 0x1 -#define SP_BLENDFACTOR_SRC_COLOR 0x2 -#define SP_BLENDFACTOR_SRC_ALPHA 0x3 -#define SP_BLENDFACTOR_DST_ALPHA 0x4 -#define SP_BLENDFACTOR_DST_COLOR 0x5 -#define SP_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 -#define SP_BLENDFACTOR_CONST_COLOR 0x7 -#define SP_BLENDFACTOR_CONST_ALPHA 0x8 -#define SP_BLENDFACTOR_SRC1_COLOR 0x9 -#define SP_BLENDFACTOR_SRC1_ALPHA 0x0A -#define SP_BLENDFACTOR_ZERO 0x11 -#define SP_BLENDFACTOR_INV_SRC_COLOR 0x12 -#define SP_BLENDFACTOR_INV_SRC_ALPHA 0x13 -#define SP_BLENDFACTOR_INV_DST_ALPHA 0x14 -#define SP_BLENDFACTOR_INV_DST_COLOR 0x15 -#define SP_BLENDFACTOR_INV_CONST_COLOR 0x17 -#define SP_BLENDFACTOR_INV_CONST_ALPHA 0x18 -#define SP_BLENDFACTOR_INV_SRC1_COLOR 0x19 -#define SP_BLENDFACTOR_INV_SRC1_ALPHA 0x1A +#define PIPE_BLENDFACTOR_ONE 0x1 +#define PIPE_BLENDFACTOR_SRC_COLOR 0x2 +#define PIPE_BLENDFACTOR_SRC_ALPHA 0x3 +#define PIPE_BLENDFACTOR_DST_ALPHA 0x4 +#define PIPE_BLENDFACTOR_DST_COLOR 0x5 +#define PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 +#define PIPE_BLENDFACTOR_CONST_COLOR 0x7 +#define PIPE_BLENDFACTOR_CONST_ALPHA 0x8 +#define PIPE_BLENDFACTOR_SRC1_COLOR 0x9 +#define PIPE_BLENDFACTOR_SRC1_ALPHA 0x0A +#define PIPE_BLENDFACTOR_ZERO 0x11 +#define PIPE_BLENDFACTOR_INV_SRC_COLOR 0x12 +#define PIPE_BLENDFACTOR_INV_SRC_ALPHA 0x13 +#define PIPE_BLENDFACTOR_INV_DST_ALPHA 0x14 +#define PIPE_BLENDFACTOR_INV_DST_COLOR 0x15 +#define PIPE_BLENDFACTOR_INV_CONST_COLOR 0x17 +#define PIPE_BLENDFACTOR_INV_CONST_ALPHA 0x18 +#define PIPE_BLENDFACTOR_INV_SRC1_COLOR 0x19 +#define PIPE_BLENDFACTOR_INV_SRC1_ALPHA 0x1A -#define SP_BLEND_ADD 0 -#define SP_BLEND_SUBTRACT 1 -#define SP_BLEND_REVERSE_SUBTRACT 2 -#define SP_BLEND_MIN 3 -#define SP_BLEND_MAX 4 +#define PIPE_BLEND_ADD 0 +#define PIPE_BLEND_SUBTRACT 1 +#define PIPE_BLEND_REVERSE_SUBTRACT 2 +#define PIPE_BLEND_MIN 3 +#define PIPE_BLEND_MAX 4 -#define SP_LOGICOP_CLEAR 0 -#define SP_LOGICOP_NOR 1 -#define SP_LOGICOP_AND_INVERTED 2 -#define SP_LOGICOP_COPY_INVERTED 3 -#define SP_LOGICOP_AND_REVERSE 4 -#define SP_LOGICOP_INVERT 5 -#define SP_LOGICOP_XOR 6 -#define SP_LOGICOP_NAND 7 -#define SP_LOGICOP_AND 8 -#define SP_LOGICOP_EQUIV 9 -#define SP_LOGICOP_NOOP 10 -#define SP_LOGICOP_OR_INVERTED 11 -#define SP_LOGICOP_COPY 12 -#define SP_LOGICOP_OR_REVERSE 13 -#define SP_LOGICOP_OR 14 -#define SP_LOGICOP_SET 15 +#define PIPE_LOGICOP_CLEAR 0 +#define PIPE_LOGICOP_NOR 1 +#define PIPE_LOGICOP_AND_INVERTED 2 +#define PIPE_LOGICOP_COPY_INVERTED 3 +#define PIPE_LOGICOP_AND_REVERSE 4 +#define PIPE_LOGICOP_INVERT 5 +#define PIPE_LOGICOP_XOR 6 +#define PIPE_LOGICOP_NAND 7 +#define PIPE_LOGICOP_AND 8 +#define PIPE_LOGICOP_EQUIV 9 +#define PIPE_LOGICOP_NOOP 10 +#define PIPE_LOGICOP_OR_INVERTED 11 +#define PIPE_LOGICOP_COPY 12 +#define PIPE_LOGICOP_OR_REVERSE 13 +#define PIPE_LOGICOP_OR 14 +#define PIPE_LOGICOP_SET 15 -#define SP_STENCIL_FUNC_NEVER 0 -#define SP_STENCIL_FUNC_LESS 1 -#define SP_STENCIL_FUNC_EQUAL 2 -#define SP_STENCIL_FUNC_LEQUAL 3 -#define SP_STENCIL_FUNC_GREATER 4 -#define SP_STENCIL_FUNC_NOTEQUAL 5 -#define SP_STENCIL_FUNC_GEQUAL 6 -#define SP_STENCIL_FUNC_ALWAYS 7 +#define PIPE_STENCIL_FUNC_NEVER 0 +#define PIPE_STENCIL_FUNC_LESS 1 +#define PIPE_STENCIL_FUNC_EQUAL 2 +#define PIPE_STENCIL_FUNC_LEQUAL 3 +#define PIPE_STENCIL_FUNC_GREATER 4 +#define PIPE_STENCIL_FUNC_NOTEQUAL 5 +#define PIPE_STENCIL_FUNC_GEQUAL 6 +#define PIPE_STENCIL_FUNC_ALWAYS 7 -#define SP_STENCIL_OP_KEEP 0 -#define SP_STENCIL_OP_ZERO 1 -#define SP_STENCIL_OP_REPLACE 2 -#define SP_STENCIL_OP_INCR 3 -#define SP_STENCIL_OP_DECR 4 -#define SP_STENCIL_OP_INCR_WRAP 5 -#define SP_STENCIL_OP_DECR_WRAP 6 -#define SP_STENCIL_OP_INVERT 7 +#define PIPE_STENCIL_OP_KEEP 0 +#define PIPE_STENCIL_OP_ZERO 1 +#define PIPE_STENCIL_OP_REPLACE 2 +#define PIPE_STENCIL_OP_INCR 3 +#define PIPE_STENCIL_OP_DECR 4 +#define PIPE_STENCIL_OP_INCR_WRAP 5 +#define PIPE_STENCIL_OP_DECR_WRAP 6 +#define PIPE_STENCIL_OP_INVERT 7 /* Note: same as stencil funcs. Also used for shadow/depth compare */ -#define SP_DEPTH_FUNC_NEVER 0 -#define SP_DEPTH_FUNC_LESS 1 -#define SP_DEPTH_FUNC_EQUAL 2 -#define SP_DEPTH_FUNC_LEQUAL 3 -#define SP_DEPTH_FUNC_GREATER 4 -#define SP_DEPTH_FUNC_NOTEQUAL 5 -#define SP_DEPTH_FUNC_GEQUAL 6 -#define SP_DEPTH_FUNC_ALWAYS 7 +#define PIPE_DEPTH_FUNC_NEVER 0 +#define PIPE_DEPTH_FUNC_LESS 1 +#define PIPE_DEPTH_FUNC_EQUAL 2 +#define PIPE_DEPTH_FUNC_LEQUAL 3 +#define PIPE_DEPTH_FUNC_GREATER 4 +#define PIPE_DEPTH_FUNC_NOTEQUAL 5 +#define PIPE_DEPTH_FUNC_GEQUAL 6 +#define PIPE_DEPTH_FUNC_ALWAYS 7 -#define SP_TEX_WRAP_REPEAT 0 -#define SP_TEX_WRAP_CLAMP 1 -#define SP_TEX_WRAP_CLAMP_TO_EDGE 2 -#define SP_TEX_WRAP_CLAMP_TO_BORDER 3 -#define SP_TEX_WRAP_MIRROR_REPEAT 4 -#define SP_TEX_WRAP_MIRROR_CLAMP 5 -#define SP_TEX_WRAP_MIRROR_CLAMP_TO_EDGE 6 -#define SP_TEX_WRAP_MIRROR_CLAMP_TO_BORDER 7 +#define PIPE_TEX_WRAP_REPEAT 0 +#define PIPE_TEX_WRAP_CLAMP 1 +#define PIPE_TEX_WRAP_CLAMP_TO_EDGE 2 +#define PIPE_TEX_WRAP_CLAMP_TO_BORDER 3 +#define PIPE_TEX_WRAP_MIRROR_REPEAT 4 +#define PIPE_TEX_WRAP_MIRROR_CLAMP 5 +#define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE 6 +#define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER 7 -#define SP_TEX_FILTER_NEAREST 0 -#define SP_TEX_FILTER_LINEAR 1 -#define SP_TEX_FILTER_NEAREST_MIPMAP_NEAREST 2 -#define SP_TEX_FILTER_NEAREST_MIPMAP_LINEAR 3 -#define SP_TEX_FILTER_LINEAR_MIPMAP_NEAREST 4 -#define SP_TEX_FILTER_LINEAR_MIPMAP_LINEAR 5 +#define PIPE_TEX_FILTER_NEAREST 0 +#define PIPE_TEX_FILTER_LINEAR 1 +#define PIPE_TEX_FILTER_NEAREST_MIPMAP_NEAREST 2 +#define PIPE_TEX_FILTER_NEAREST_MIPMAP_LINEAR 3 +#define PIPE_TEX_FILTER_LINEAR_MIPMAP_NEAREST 4 +#define PIPE_TEX_FILTER_LINEAR_MIPMAP_LINEAR 5 -#define SP_TEX_COMPARE_NONE 0 -#define SP_TEX_COMPARE_R_TO_TEXTURE 1 +#define PIPE_TEX_COMPARE_NONE 0 +#define PIPE_TEX_COMPARE_R_TO_TEXTURE 1 #endif diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 150fa9fcad..7264f253d9 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -25,8 +25,8 @@ * **************************************************************************/ -#ifndef SP_STATE_H -#define SP_STATE_H +#ifndef PIPE_STATE_H +#define PIPE_STATE_H #include "mtypes.h" #include "vf/vf.h" @@ -40,7 +40,7 @@ #define FILL_LINE 2 #define FILL_TRI 3 -struct softpipe_setup_state { +struct pipe_setup_state { GLuint flatshade:1; GLuint light_twoside:1; @@ -63,17 +63,17 @@ struct softpipe_setup_state { GLfloat offset_scale; }; -struct softpipe_poly_stipple { +struct pipe_poly_stipple { GLuint stipple[32]; }; -struct softpipe_viewport { +struct pipe_viewport { GLfloat scale[4]; GLfloat translate[4]; }; -struct softpipe_scissor_rect { +struct pipe_scissor_rect { GLshort minx; GLshort miny; GLshort maxx; @@ -81,26 +81,26 @@ struct softpipe_scissor_rect { }; -#define SP_MAX_CLIP_PLANES 6 +#define PIPE_MAX_CLIP_PLANES 6 -struct softpipe_clip_state { - GLfloat ucp[SP_MAX_CLIP_PLANES][4]; +struct pipe_clip_state { + GLfloat ucp[PIPE_MAX_CLIP_PLANES][4]; GLuint nr; }; -struct softpipe_fs_state { +struct pipe_fs_state { struct gl_fragment_program *fp; }; -#define SP_MAX_CONSTANT 32 +#define PIPE_MAX_CONSTANT 32 -struct softpipe_constant_buffer { - GLfloat constant[SP_MAX_CONSTANT][4]; +struct pipe_constant_buffer { + GLfloat constant[PIPE_MAX_CONSTANT][4]; GLuint nr_constants; }; -struct softpipe_depth_state +struct pipe_depth_state { GLuint enabled:1; /**< depth test enabled? */ GLuint writemask:1; /**< allow depth buffer writes? */ @@ -109,7 +109,7 @@ struct softpipe_depth_state }; -struct softpipe_blend_state { +struct pipe_blend_state { GLuint blend_enable:1; GLuint rgb_func:3; @@ -124,17 +124,17 @@ struct softpipe_blend_state { GLuint logicop_func:4; }; -struct softpipe_blend_color { +struct pipe_blend_color { GLfloat color[4]; }; -struct softpipe_stencil_state { +struct pipe_stencil_state { GLuint front_enabled:1; - GLuint front_func:3; /**< SP_STENCIL_FUNC_x */ - GLuint front_fail_op:3; /**< SP_STENCIL_OP_x */ - GLuint front_zpass_op:3; /**< SP_STENCIL_OP_x */ - GLuint front_zfail_op:3; /**< SP_STENCIL_OP_x */ + GLuint front_func:3; /**< PIPE_STENCIL_FUNC_x */ + GLuint front_fail_op:3; /**< PIPE_STENCIL_OP_x */ + GLuint front_zpass_op:3; /**< PIPE_STENCIL_OP_x */ + GLuint front_zfail_op:3; /**< PIPE_STENCIL_OP_x */ GLuint back_enabled:1; GLuint back_func:3; GLuint back_fail_op:3; @@ -147,9 +147,9 @@ struct softpipe_stencil_state { }; -/* This will change for hardware softpipes... +/* This will change for hardware pipes... */ -struct softpipe_surface { +struct pipe_surface { GLubyte *ptr; GLint stride; GLuint cpp; @@ -160,13 +160,13 @@ struct softpipe_surface { /** * Texture sampler state. */ -struct softpipe_sampler_state +struct pipe_sampler_state { - GLuint wrap_s:3; /**< SP_TEX_WRAP_x */ - GLuint wrap_t:3; /**< SP_TEX_WRAP_x */ - GLuint wrap_r:3; /**< SP_TEX_WRAP_x */ - GLuint min_filter:3; /**< SP_TEX_FILTER_x */ - GLuint mag_filter:1; /**< SP_TEX_FILTER_LINEAR or _NEAREST */ + GLuint wrap_s:3; /**< PIPE_TEX_WRAP_x */ + GLuint wrap_t:3; /**< PIPE_TEX_WRAP_x */ + GLuint wrap_r:3; /**< PIPE_TEX_WRAP_x */ + GLuint min_filter:3; /**< PIPE_TEX_FILTER_x */ + GLuint mag_filter:1; /**< PIPE_TEX_FILTER_LINEAR or _NEAREST */ GLfloat min_lod; GLfloat max_lod; GLfloat lod_bias; @@ -176,8 +176,8 @@ struct softpipe_sampler_state #endif GLfloat max_anisotropy; GLuint compare:1; /**< shadow/depth compare enabled? */ - GLenum compare_mode:1; /**< SP_TEX_COMPARE_x */ - GLenum compare_func:3; /**< SP_DEPTH_FUNC_x */ + GLenum compare_mode:1; /**< PIPE_TEX_COMPARE_x */ + GLenum compare_func:3; /**< PIPE_DEPTH_FUNC_x */ GLfloat shadow_ambient; /**< shadow test fail color/intensity */ }; diff --git a/src/mesa/pipe/softpipe/g_context.c b/src/mesa/pipe/softpipe/g_context.c deleted file mode 100644 index c3329b5c7d..0000000000 --- a/src/mesa/pipe/softpipe/g_context.c +++ /dev/null @@ -1,135 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - -#include "imports.h" -#include "macros.h" - -#include "tnl/t_context.h" -#include "vf/vf.h" - -#include "g_context.h" -#include "g_prim.h" -#include "g_state.h" -#include "g_draw.h" - -static void generic_destroy( struct softpipe_context *softpipe ) -{ - struct generic_context *generic = generic_context( softpipe ); - - draw_destroy( generic->draw ); - - FREE( generic ); -} - - -static void generic_draw_vb( struct softpipe_context *softpipe, - struct vertex_buffer *VB ) -{ - struct generic_context *generic = generic_context( softpipe ); - - if (generic->dirty) - generic_update_derived( generic ); - - draw_vb( generic->draw, VB ); -} - -struct softpipe_context *generic_create( void ) -{ - struct generic_context *generic = CALLOC_STRUCT(generic_context); - - generic->softpipe.destroy = generic_destroy; - generic->softpipe.set_clip_state = generic_set_clip_state; - generic->softpipe.set_viewport = generic_set_viewport; - generic->softpipe.set_setup_state = generic_set_setup_state; - generic->softpipe.set_scissor_rect = generic_set_scissor_rect; - generic->softpipe.set_fs_state = generic_set_fs_state; - generic->softpipe.set_polygon_stipple = generic_set_polygon_stipple; - generic->softpipe.set_cbuf_state = generic_set_cbuf_state; - generic->softpipe.draw_vb = generic_draw_vb; - - - - generic->prim.setup = prim_setup( generic ); - generic->prim.unfilled = prim_unfilled( generic ); - generic->prim.twoside = prim_twoside( generic ); - generic->prim.offset = prim_offset( generic ); - generic->prim.clip = prim_clip( generic ); - generic->prim.flatshade = prim_flatshade( generic ); - generic->prim.cull = prim_cull( generic ); - - - generic->draw = draw_create( generic ); - - ASSIGN_4V( generic->plane[0], -1, 0, 0, 1 ); - ASSIGN_4V( generic->plane[1], 1, 0, 0, 1 ); - ASSIGN_4V( generic->plane[2], 0, -1, 0, 1 ); - ASSIGN_4V( generic->plane[3], 0, 1, 0, 1 ); - ASSIGN_4V( generic->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ - ASSIGN_4V( generic->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ - generic->nr_planes = 6; - - return &generic->softpipe; -} - - - - - - -#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) - -void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ) -{ - stage->nr_tmps = nr; - - if (nr) { - GLubyte *store = MALLOC(MAX_VERTEX_SIZE * nr); - GLuint i; - - stage->tmp = MALLOC(sizeof(struct vertex_header *) * nr); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); - } -} - -void prim_free_tmps( struct prim_stage *stage ) -{ - if (stage->tmp) { - FREE(stage->tmp[0]); - FREE(stage->tmp); - } -} - - - - - diff --git a/src/mesa/pipe/softpipe/g_context.h b/src/mesa/pipe/softpipe/g_context.h deleted file mode 100644 index 24d467f2c3..0000000000 --- a/src/mesa/pipe/softpipe/g_context.h +++ /dev/null @@ -1,149 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_CONTEXT_H -#define G_CONTEXT_H - -#include "glheader.h" - -#include "softpipe/sp_state.h" -#include "softpipe/sp_context.h" - - - -struct generic_surface; -struct draw_context; -struct prim_stage; - - -enum interp_mode { - INTERP_CONSTANT, - INTERP_LINEAR, - INTERP_PERSPECTIVE -}; - - -#define G_NEW_VIEWPORT 0x1 -#define G_NEW_SETUP 0x2 -#define G_NEW_FS 0x4 -#define G_NEW_BLEND 0x8 -#define G_NEW_CBUF 0x10 -#define G_NEW_CLIP 0x20 -#define G_NEW_SCISSOR 0x40 -#define G_NEW_STIPPLE 0x80 - - - -struct generic_context { - struct softpipe_context softpipe; - - - /* The most recent drawing state as set by the driver: - */ - struct softpipe_viewport viewport; - struct softpipe_setup_state setup; - struct softpipe_fs_state fs; - struct softpipe_blend_state blend; - struct softpipe_surface cbuf; - struct softpipe_clip_state clip; - struct softpipe_scissor_rect scissor; - struct softpipe_poly_stipple poly_stipple; - GLuint dirty; - - - /* Cbuf derived state??? - */ - struct generic_surface *cbuf_surface; - - /* Clip derived state: - */ - GLfloat plane[12][4]; - GLuint nr_planes; - - /* Setup derived state. TODO: this should be passed in the program - * tokens as parameters to DECL instructions. - * - * For now we just set colors to CONST on flatshade, textures to - * perspective always and everything else to linear. - */ - enum interp_mode interp[VF_ATTRIB_MAX]; - - - /* FS + setup derived state: - */ - GLuint fp_attr_to_slot[VF_ATTRIB_MAX]; - GLuint vf_attr_to_slot[VF_ATTRIB_MAX]; - GLuint nr_attrs; - GLuint nr_frag_attrs; - GLuint attr_mask; - - GLboolean need_z; - GLboolean need_w; - - /* Stipple derived state: - */ - GLubyte stipple_masks[16][16]; - - - /* The software clipper/setup engine. - */ - struct { - struct prim_stage *setup; - struct prim_stage *unfilled; - struct prim_stage *twoside; - struct prim_stage *clip; - struct prim_stage *flatshade; - struct prim_stage *offset; - struct prim_stage *cull; - - struct prim_stage *first; - - GLenum prim; - GLuint vertex_size; - } prim; - - /* Temp kludge: - */ - struct draw_context *draw; -}; - - - - -static INLINE struct generic_context * -generic_context( struct softpipe_context *softpipe ) -{ - return (struct generic_context *)softpipe; -} - - - - -#endif diff --git a/src/mesa/pipe/softpipe/g_draw.c b/src/mesa/pipe/softpipe/g_draw.c deleted file mode 100644 index 4e094c21a0..0000000000 --- a/src/mesa/pipe/softpipe/g_draw.c +++ /dev/null @@ -1,696 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "imports.h" - -#include "tnl/t_context.h" - -#include "g_context.h" -#include "g_prim.h" -#include "g_headers.h" -#include "g_draw.h" - -/* This file is a temporary set of hooks to allow us to use the tnl/ - * and vf/ modules until we have replacements in softpipe. - */ - - -struct draw_context -{ - struct generic_context *generic; - - struct vf_attr_map attrs[VF_ATTRIB_MAX]; - GLuint nr_attrs; - GLuint vertex_size; - struct vertex_fetch *vf; - - GLubyte *verts; - GLuint nr_vertices; - GLboolean in_vb; - - GLenum prim; - - /* Helper for tnl: - */ - GLvector4f header; -}; - - -static struct vertex_header *get_vertex( struct draw_context *pipe, - GLuint i ) -{ - return (struct vertex_header *)(pipe->verts + i * pipe->vertex_size); -} - - - -static void draw_allocate_vertices( struct draw_context *draw, - GLuint nr_vertices ) -{ - draw->nr_vertices = nr_vertices; - draw->verts = MALLOC( nr_vertices * draw->vertex_size ); - - draw->generic->prim.first->begin( draw->generic->prim.first ); -} - -static void draw_set_prim( struct draw_context *draw, - GLenum prim ) -{ - draw->prim = prim; - - /* Not done yet - need to force edgeflags to 1 in strip/fan - * primitives. - */ -#if 0 - switch (prim) { - case GL_TRIANGLES: - case GL_POLYGON: - case GL_QUADS: - case GL_QUAD_STRIP: /* yes, we need this */ - respect_edgeflags( pipe, GL_TRUE ); - break; - - default: - respect_edgeflags( pipe, GL_FALSE ); - break; - } -#endif -} - - - -static void do_quad( struct prim_stage *first, - struct vertex_header *v0, - struct vertex_header *v1, - struct vertex_header *v2, - struct vertex_header *v3 ) -{ - struct prim_header prim; - - { - GLuint tmp = v1->edgeflag; - v1->edgeflag = 0; - - prim.v[0] = v0; - prim.v[1] = v1; - prim.v[2] = v3; - first->tri( first, &prim ); - - v1->edgeflag = tmp; - } - - { - GLuint tmp = v3->edgeflag; - v3->edgeflag = 0; - - prim.v[0] = v1; - prim.v[1] = v2; - prim.v[2] = v3; - first->tri( first, &prim ); - - v3->edgeflag = tmp; - } -} - - - - -static void draw_indexed_prim( struct draw_context *draw, - const GLuint *elts, - GLuint count ) -{ - struct prim_stage * const first = draw->generic->prim.first; - struct prim_header prim; - GLuint i; - - prim.det = 0; /* valid from cull stage onwards */ - prim.v[0] = 0; - prim.v[1] = 0; - prim.v[2] = 0; - - switch (draw->prim) { - case GL_POINTS: - for (i = 0; i < count; i ++) { - prim.v[0] = get_vertex( draw, elts[i] ); - - first->point( first, &prim ); - } - break; - - case GL_LINES: - for (i = 0; i+1 < count; i += 2) { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - - first->line( first, &prim ); - } - break; - - case GL_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - prim.v[0] = get_vertex( draw, elts[i-1] ); - prim.v[1] = get_vertex( draw, elts[i] ); - first->line( first, &prim ); - } - - prim.v[0] = get_vertex( draw, elts[count-1] ); - prim.v[1] = get_vertex( draw, elts[0] ); - first->line( first, &prim ); - } - break; - - case GL_LINE_STRIP: - /* I'm guessing it will be necessary to have something like a - * render->reset_line_stipple() method to properly support - * splitting strips into primitives like this. Alternately we - * could just scan ahead to find individual clipped lines and - * otherwise leave the strip intact - that might be better, but - * require more complex code here. - */ - if (count >= 2) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, elts[0] ); - - for (i = 1; i < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, elts[i] ); - - first->line( first, &prim ); - } - } - break; - - case GL_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - prim.v[0] = get_vertex( draw, elts[i + 1] ); - prim.v[1] = get_vertex( draw, elts[i + 0] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - } - else { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - } - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_FAN: - if (count >= 3) { - prim.v[0] = get_vertex( draw, elts[0] ); - prim.v[1] = 0; - prim.v[2] = get_vertex( draw, elts[1] ); - - for (i = 0; i+2 < count; i++) { - prim.v[1] = prim.v[2]; - prim.v[2] = get_vertex( draw, elts[i+2] ); - - first->tri( first, &prim ); - } - } - break; - - case GL_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( first, - get_vertex( draw, elts[i + 0] ), - get_vertex( draw, elts[i + 1] ), - get_vertex( draw, elts[i + 2] ), - get_vertex( draw, elts[i + 3] )); - } - break; - - case GL_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( first, - get_vertex( draw, elts[i + 2] ), - get_vertex( draw, elts[i + 0] ), - get_vertex( draw, elts[i + 1] ), - get_vertex( draw, elts[i + 3] )); - } - break; - - - case GL_POLYGON: - if (count >= 3) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, elts[1] ); - prim.v[2] = get_vertex( draw, elts[0] ); - - for (i = 0; i+2 < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, elts[i+2] ); - - first->tri( first, &prim ); - } - } - break; - - default: - assert(0); - break; - } -} - -static void draw_prim( struct draw_context *draw, - GLuint start, - GLuint count ) -{ - struct prim_stage * const first = draw->generic->prim.first; - struct prim_header prim; - GLuint i; - -// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - - prim.det = 0; /* valid from cull stage onwards */ - prim.v[0] = 0; - prim.v[1] = 0; - prim.v[2] = 0; - - switch (draw->prim) { - case GL_POINTS: - for (i = 0; i < count; i ++) { - prim.v[0] = get_vertex( draw, start + i ); - first->point( first, &prim ); - } - break; - - case GL_LINES: - for (i = 0; i+1 < count; i += 2) { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - - first->line( first, &prim ); - } - break; - - case GL_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - prim.v[0] = get_vertex( draw, start + i - 1 ); - prim.v[1] = get_vertex( draw, start + i ); - first->line( first, &prim ); - } - - prim.v[0] = get_vertex( draw, start + count - 1 ); - prim.v[1] = get_vertex( draw, start + 0 ); - first->line( first, &prim ); - } - break; - - case GL_LINE_STRIP: - if (count >= 2) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, start + 0 ); - - for (i = 1; i < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, start + i ); - - first->line( first, &prim ); - } - } - break; - - case GL_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - prim.v[0] = get_vertex( draw, start + i + 1 ); - prim.v[1] = get_vertex( draw, start + i + 0 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - } - else { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - } - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_FAN: - if (count >= 3) { - prim.v[0] = get_vertex( draw, start + 0 ); - prim.v[1] = 0; - prim.v[2] = get_vertex( draw, start + 1 ); - - for (i = 0; i+2 < count; i++) { - prim.v[1] = prim.v[2]; - prim.v[2] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); - } - } - break; - - - case GL_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( first, - get_vertex( draw, start + i + 0 ), - get_vertex( draw, start + i + 1 ), - get_vertex( draw, start + i + 2 ), - get_vertex( draw, start + i + 3 )); - } - break; - - case GL_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( first, - get_vertex( draw, start + i + 2 ), - get_vertex( draw, start + i + 0 ), - get_vertex( draw, start + i + 1 ), - get_vertex( draw, start + i + 3 )); - } - break; - - case GL_POLYGON: - if (count >= 3) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, start + 1 ); - prim.v[2] = get_vertex( draw, start + 0 ); - - for (i = 0; i+2 < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); - } - } - break; - - default: - assert(0); - break; - } -} - - -static void draw_release_vertices( struct draw_context *draw ) -{ - draw->generic->prim.first->end( draw->generic->prim.first ); - - FREE(draw->verts); - draw->verts = NULL; -} - - -struct header_dword { - GLuint clipmask:12; - GLuint edgeflag:1; - GLuint pad:19; -}; - - -static void -build_vertex_headers( struct draw_context *draw, - struct vertex_buffer *VB ) -{ - if (draw->header.storage == NULL) { - draw->header.stride = sizeof(GLfloat); - draw->header.size = 1; - draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(GLfloat), 32 ); - draw->header.data = draw->header.storage; - draw->header.count = 0; - draw->header.flags = VEC_SIZE_1 | VEC_MALLOC; - } - - /* Build vertex header attribute. - * - */ - - { - GLuint i; - struct header_dword *header = (struct header_dword *)draw->header.storage; - - /* yes its a hack - */ - assert(sizeof(*header) == sizeof(GLfloat)); - - draw->header.count = VB->Count; - - if (VB->EdgeFlag) { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = VB->ClipMask[i]; - header[i].edgeflag = VB->EdgeFlag[i]; - header[i].pad = 0; - } - } - else if (VB->ClipOrMask) { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = VB->ClipMask[i]; - header[i].edgeflag = 0; - header[i].pad = 0; - } - } - else { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = 0; - header[i].edgeflag = 0; - header[i].pad = 0; - } - } - } - - VB->AttribPtr[VF_ATTRIB_VERTEX_HEADER] = &draw->header; -} - - - - - -static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) -{ - switch (mode) { - case GL_POINTS: - *first = 1; - *incr = 1; - return 0; - case GL_LINES: - *first = 2; - *incr = 2; - return 0; - case GL_LINE_STRIP: - *first = 2; - *incr = 1; - return 0; - case GL_LINE_LOOP: - *first = 2; - *incr = 1; - return 1; - case GL_TRIANGLES: - *first = 3; - *incr = 3; - return 0; - case GL_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - return 0; - case GL_TRIANGLE_FAN: - case GL_POLYGON: - *first = 3; - *incr = 1; - return 1; - case GL_QUADS: - *first = 4; - *incr = 4; - return 0; - case GL_QUAD_STRIP: - *first = 4; - *incr = 2; - return 0; - default: - assert(0); - *first = 1; - *incr = 1; - return 0; - } -} - - -static GLuint trim( GLuint count, GLuint first, GLuint incr ) -{ - if (count < first) - return 0; - else - return count - (count - first) % incr; -} - - -/* This is a hack & will all go away. - */ -void draw_vb(struct draw_context *draw, - struct vertex_buffer *VB ) -{ - GLuint i; - - VB->AttribPtr[VF_ATTRIB_POS] = VB->NdcPtr; - VB->AttribPtr[VF_ATTRIB_BFC0] = VB->ColorPtr[1]; - VB->AttribPtr[VF_ATTRIB_BFC1] = VB->SecondaryColorPtr[1]; - VB->AttribPtr[VF_ATTRIB_CLIP_POS] = VB->ClipPtr; - - /* Build vertex headers: - */ - build_vertex_headers( draw, VB ); - - draw->in_vb = 1; - - /* Allocate the vertices: - */ - draw_allocate_vertices( draw, VB->Count ); - - /* Bind the vb outputs: - */ - vf_set_sources( draw->vf, VB->AttribPtr, 0 ); - - /* Build the hardware or prim-pipe vertices: - */ - vf_emit_vertices( draw->vf, VB->Count, draw->verts ); - - - for (i = 0; i < VB->PrimitiveCount; i++) { - - GLenum mode = VB->Primitive[i].mode; - GLuint start = VB->Primitive[i].start; - GLuint length, first, incr; - - /* Trim the primitive down to a legal size. - */ - draw_prim_info( mode, &first, &incr ); - length = trim( VB->Primitive[i].count, first, incr ); - - if (!length) - continue; - - if (draw->prim != mode) - draw_set_prim( draw, mode ); - - if (VB->Elts) { - draw_indexed_prim( draw, - VB->Elts + start, - length ); - } - else { - draw_prim( draw, - start, - length ); - } - } - - draw_release_vertices( draw ); - draw->verts = NULL; - draw->in_vb = 0; -} - -void draw_set_viewport( struct draw_context *draw, - const GLfloat *scale, - const GLfloat *translate ) -{ - assert(!draw->in_vb); - vf_set_vp_scale_translate( draw->vf, scale, translate ); -} - - - -struct draw_context *draw_create( struct generic_context *generic ) -{ - struct draw_context *draw = CALLOC_STRUCT( draw_context ); - draw->generic = generic; - draw->vf = vf_create( GL_TRUE ); - - return draw; -} - - -void draw_destroy( struct draw_context *draw ) -{ - if (draw->header.storage) - ALIGN_FREE( draw->header.storage ); - - vf_destroy( draw->vf ); - - FREE( draw ); -} - -#define EMIT_ATTR( ATTR, STYLE ) \ -do { \ - draw->attrs[draw->nr_attrs].attrib = ATTR; \ - draw->attrs[draw->nr_attrs].format = STYLE; \ - draw->nr_attrs++; \ -} while (0) - - -void draw_set_vertex_attributes( struct draw_context *draw, - const GLuint *attrs, - GLuint nr_attrs ) -{ - GLuint i; - - draw->nr_attrs = 0; - - EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F); - EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F); - - assert(attrs[0] == VF_ATTRIB_POS); - EMIT_ATTR(attrs[0], EMIT_4F_VIEWPORT); - - for (i = 1; i < nr_attrs; i++) - EMIT_ATTR(attrs[i], EMIT_4F); - - draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); -} - - diff --git a/src/mesa/pipe/softpipe/g_draw.h b/src/mesa/pipe/softpipe/g_draw.h deleted file mode 100644 index 78c93b4df7..0000000000 --- a/src/mesa/pipe/softpipe/g_draw.h +++ /dev/null @@ -1,56 +0,0 @@ - -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_DRAW_H -#define G_DRAW_H - -#include "glheader.h" -#include "softpipe/sp_state.h" - - -struct draw_context; - -struct draw_context *draw_create( struct generic_context *generic ); - -void draw_destroy( struct draw_context *draw ); - -void draw_set_viewport( struct draw_context *draw, - const GLfloat *scale, - const GLfloat *translate ); - -void draw_set_vertex_attributes( struct draw_context *draw, - const GLuint *attrs, - GLuint nr_attrs ); - -void draw_vb(struct draw_context *draw, - struct vertex_buffer *VB ); - -#endif diff --git a/src/mesa/pipe/softpipe/g_headers.h b/src/mesa/pipe/softpipe/g_headers.h deleted file mode 100644 index 96ff52a453..0000000000 --- a/src/mesa/pipe/softpipe/g_headers.h +++ /dev/null @@ -1,116 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#ifndef G_HEADERS_H -#define G_HEADERS_H - -#define PRIM_POINT 1 -#define PRIM_LINE 2 -#define PRIM_TRI 3 - -struct prim_header { - GLfloat det; - struct vertex_header *v[3]; -}; - -/* Carry some useful information around with the vertices in the prim - * pipe. - */ -struct vertex_header { - GLuint clipmask:12; - GLuint edgeflag:1; - GLuint pad:19; - - GLfloat clip[4]; - - GLfloat data[][4]; /* Note variable size */ -}; - - - - - -/* The rasterizer generates 2x2 quads of fragment and feeds them to - * the current fp_machine (see below). - */ -#define QUAD_BOTTOM_LEFT 0 -#define QUAD_BOTTOM_RIGHT 1 -#define QUAD_TOP_LEFT 2 -#define QUAD_TOP_RIGHT 3 -#define QUAD_SIZE (2*2) - -#define MASK_BOTTOM_LEFT 0x1 -#define MASK_BOTTOM_RIGHT 0x2 -#define MASK_TOP_LEFT 0x4 -#define MASK_TOP_RIGHT 0x8 -#define MASK_ALL 0xf - - -#define NUM_CHANNELS 4 /* avoid confusion between 4 pixels and 4 channels */ - - -struct setup_coefficient { - GLfloat a0[NUM_CHANNELS]; /* in an xyzw layout */ - GLfloat dadx[NUM_CHANNELS]; - GLfloat dady[NUM_CHANNELS]; -}; - - - -/* Encodes everything we need to know about a 2x2 pixel block. Uses - * "Channel-Serial" or "SoA" layout. - * - * Will expand to include non-attribute things like AA coverage and - * maybe prefetched depth from the depth buffer. - */ -struct quad_header { - GLint x0; - GLint y0; - GLuint mask; - GLuint facing; /**< Front or back facing? */ - - struct { - GLfloat color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ - GLfloat depth[QUAD_SIZE]; - } outputs; - - const struct setup_coefficient *coef; - - const enum interp_mode *interp; /* XXX: this information should be - * encoded in fragment program DECL - * statements. */ - - GLuint nr_attrs; -}; - - - - - -#endif diff --git a/src/mesa/pipe/softpipe/g_prim.h b/src/mesa/pipe/softpipe/g_prim.h deleted file mode 100644 index e3b2c5ea2f..0000000000 --- a/src/mesa/pipe/softpipe/g_prim.h +++ /dev/null @@ -1,90 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_PRIM_H -#define G_PRIM_H - -#include "glheader.h" -#include "g_headers.h" - -struct generic_context; - -struct prim_stage *prim_setup( struct generic_context *context ); -struct prim_stage *prim_unfilled( struct generic_context *context ); -struct prim_stage *prim_twoside( struct generic_context *context ); -struct prim_stage *prim_offset( struct generic_context *context ); -struct prim_stage *prim_clip( struct generic_context *context ); -struct prim_stage *prim_flatshade( struct generic_context *context ); -struct prim_stage *prim_cull( struct generic_context *context ); - - -/* Internal structs and helpers for the primitive clip/setup pipeline: - */ -struct prim_stage { - struct generic_context *generic; - - struct prim_stage *next; - - struct vertex_header **tmp; - GLuint nr_tmps; - - void (*begin)( struct prim_stage * ); - - void (*point)( struct prim_stage *, - struct prim_header * ); - - void (*line)( struct prim_stage *, - struct prim_header * ); - - void (*tri)( struct prim_stage *, - struct prim_header * ); - - void (*end)( struct prim_stage * ); -}; - - - -/* Get a writeable copy of a vertex: - */ -static INLINE struct vertex_header * -dup_vert( struct prim_stage *stage, - const struct vertex_header *vert, - GLuint idx ) -{ - struct vertex_header *tmp = stage->tmp[idx]; - memcpy(tmp, vert, stage->generic->prim.vertex_size ); - return tmp; -} - -void prim_free_tmps( struct prim_stage *stage ); -void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); - - -#endif diff --git a/src/mesa/pipe/softpipe/g_prim_clip.c b/src/mesa/pipe/softpipe/g_prim_clip.c deleted file mode 100644 index 742def9c7d..0000000000 --- a/src/mesa/pipe/softpipe/g_prim_clip.c +++ /dev/null @@ -1,429 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "imports.h" -#include "macros.h" - -#include "g_context.h" -#include "g_prim.h" - -struct clipper { - struct prim_stage stage; - - GLuint active_user_planes; -}; - -/* This is a bit confusing: - */ -static INLINE struct clipper *clipper_stage( struct prim_stage *stage ) -{ - return (struct clipper *)stage; -} - - -#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) - - -/* All attributes are float[4], so this is easy: - */ -static void interp_attr( GLfloat *fdst, - GLfloat t, - const GLfloat *fin, - const GLfloat *fout ) -{ - fdst[0] = LINTERP( t, fout[0], fin[0] ); - fdst[1] = LINTERP( t, fout[1], fin[1] ); - fdst[2] = LINTERP( t, fout[2], fin[2] ); - fdst[3] = LINTERP( t, fout[3], fin[3] ); -} - - - - -/* Interpolate between two vertices to produce a third. - */ -static void interp( struct clipper *clip, - struct vertex_header *dst, - GLfloat t, - const struct vertex_header *out, - const struct vertex_header *in ) -{ - const GLuint nr_attrs = clip->stage.generic->nr_attrs; - GLuint j; - - /* Vertex header. - */ - { - dst->clipmask = 0; - dst->edgeflag = 0; - dst->pad = 0; - } - - /* Clip coordinates: interpolate normally - */ - { - interp_attr(dst->clip, t, in->clip, out->clip); - } - - /* Do the projective divide and insert window coordinates: - */ - { - const GLfloat *pos = dst->clip; - const GLfloat *scale = clip->stage.generic->viewport.scale; - const GLfloat *trans = clip->stage.generic->viewport.translate; - GLfloat oow; - - oow = 1.0 / pos[3]; - - dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; - dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; - dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; - dst->data[0][3] = oow; - } - - - /* Other attributes - */ - for (j = 1; j < nr_attrs-1; j++) { - interp_attr(dst->data[j], t, in->data[j], out->data[j]); - } -} - - -#define CLIP_USER_BIT 0x40 -#define CLIP_CULL_BIT 0x80 - - -static INLINE GLfloat dot4( const GLfloat *a, - const GLfloat *b ) -{ - GLfloat result = (a[0]*b[0] + - a[1]*b[1] + - a[2]*b[2] + - a[3]*b[3]); - - return result; -} - - -#if 0 -static INLINE void do_tri( struct prim_stage *next, - struct prim_header *header ) -{ - GLuint i; - for (i = 0; i < 3; i++) { - GLfloat *ndc = header->v[i]->data[0]; - _mesa_printf("ndc %f %f %f\n", ndc[0], ndc[1], ndc[2]); - assert(ndc[0] >= -1 && ndc[0] <= 641); - assert(ndc[1] >= 30 && ndc[1] <= 481); - } - _mesa_printf("\n"); - next->tri(next, header); -} -#endif - - -static void emit_poly( struct prim_stage *stage, - struct vertex_header **inlist, - GLuint n ) -{ - struct prim_header header; - GLuint i; - - for (i = 2; i < n; i++) { - header.v[0] = inlist[0]; - header.v[1] = inlist[i-1]; - header.v[2] = inlist[i]; - - { - GLuint tmp0 = header.v[0]->edgeflag; - GLuint tmp2 = header.v[2]->edgeflag; - - if (i != 2) header.v[0]->edgeflag = 0; - if (i != n-1) header.v[2]->edgeflag = 0; - - stage->next->tri( stage->next, &header ); - - header.v[0]->edgeflag = tmp0; - header.v[2]->edgeflag = tmp2; - } - } -} - - -#if 0 -static void emit_poly( struct prim_stage *stage ) -{ - GLuint i; - - for (i = 2; i < n; i++) { - header->v[0] = inlist[0]; - header->v[1] = inlist[i-1]; - header->v[2] = inlist[i]; - - stage->next->tri( stage->next, header ); - } -} -#endif - - -/* Clip a triangle against the viewport and user clip planes. - */ -static void -do_clip_tri( struct prim_stage *stage, - struct prim_header *header, - GLuint clipmask ) -{ - struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *a[MAX_CLIPPED_VERTICES]; - struct vertex_header *b[MAX_CLIPPED_VERTICES]; - struct vertex_header **inlist = a; - struct vertex_header **outlist = b; - GLuint tmpnr = 0; - GLuint n = 3; - GLuint i; - - inlist[0] = header->v[0]; - inlist[1] = header->v[1]; - inlist[2] = header->v[2]; - - /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove - * this once we correctly use 16bit masks for userclip planes. - */ - clipmask &= ~CLIP_CULL_BIT; - if (clipmask & CLIP_USER_BIT) { - clipmask &= ~CLIP_USER_BIT; - clipmask |= clipper->active_user_planes; - } - - while (clipmask && n >= 3) { - GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->stage.generic->plane[plane_idx]; - struct vertex_header *vert_prev = inlist[0]; - GLfloat dp_prev = dot4( vert_prev->clip, plane ); - GLuint outcount = 0; - - clipmask &= ~(1<clip, plane ); - - if (!IS_NEGATIVE(dp_prev)) { - outlist[outcount++] = vert_prev; - } - - if (DIFFERENT_SIGNS(dp, dp_prev)) { - struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; - outlist[outcount++] = new_vert; - - if (IS_NEGATIVE(dp)) { - /* Going out of bounds. Avoid division by zero as we - * know dp != dp_prev from DIFFERENT_SIGNS, above. - */ - GLfloat t = dp / (dp - dp_prev); - interp( clipper, new_vert, t, vert, vert_prev ); - - /* Force edgeflag true in this case: - */ - new_vert->edgeflag = 1; - } else { - /* Coming back in. - */ - GLfloat t = dp_prev / (dp_prev - dp); - interp( clipper, new_vert, t, vert_prev, vert ); - - /* Copy starting vert's edgeflag: - */ - new_vert->edgeflag = vert_prev->edgeflag; - } - } - - vert_prev = vert; - dp_prev = dp; - } - - { - struct vertex_header **tmp = inlist; - inlist = outlist; - outlist = tmp; - n = outcount; - } - } - - /* Emit the polygon as triangles to the setup stage: - */ - if (n >= 3) - emit_poly( stage, inlist, n ); -} - - -/* Clip a line against the viewport and user clip planes. - */ -static void -do_clip_line( struct prim_stage *stage, - struct prim_header *header, - GLuint clipmask ) -{ - struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - const GLfloat *pos0 = v0->clip; - const GLfloat *pos1 = v1->clip; - GLfloat t0 = 0; - GLfloat t1 = 0; - - /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove - * this once we correctly use 16bit masks for userclip planes. - */ - clipmask &= ~CLIP_CULL_BIT; - if (clipmask & CLIP_USER_BIT) { - clipmask &= ~CLIP_USER_BIT; - clipmask |= clipper->active_user_planes; - } - - while (clipmask) { - GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->stage.generic->plane[plane_idx]; - - clipmask &= ~(1< t1) t1 = t; - } - - if (dp0 < 0) { - GLfloat t = dp0 / (dp0 - dp1); - if (t > t0) t0 = t; - } - - if (t0 + t1 >= 1.0) - return; /* discard */ - } - - if (v0->clipmask) { - interp( clipper, stage->tmp[0], t0, v0, v1 ); - header->v[0] = stage->tmp[0]; - } - - if (v1->clipmask) { - interp( clipper, stage->tmp[1], t1, v1, v0 ); - header->v[1] = stage->tmp[1]; - } - - stage->next->line( stage->next, header ); -} - - - -static void clip_begin( struct prim_stage *stage ) -{ - struct clipper *clipper = clipper_stage(stage); - GLuint nr = stage->generic->nr_planes; - - /* Hacky bitmask to use when we hit CLIP_USER_BIT: - */ - clipper->active_user_planes = ((1<next->begin( stage->next ); -} - -static void -clip_point( struct prim_stage *stage, - struct prim_header *header ) -{ - if (header->v[0]->clipmask == 0) - stage->next->point( stage->next, header ); -} - - -static void -clip_line( struct prim_stage *stage, - struct prim_header *header ) -{ - GLuint clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask); - - if (clipmask == 0) { - stage->next->line( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask) == 0) { - do_clip_line(stage, header, clipmask); - } -} - - -static void -clip_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - GLuint clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask | - header->v[2]->clipmask); - - if (clipmask == 0) { - stage->next->tri( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask & - header->v[2]->clipmask) == 0) { - do_clip_tri(stage, header, clipmask); - } -} - -static void clip_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - - -struct prim_stage *prim_clip( struct generic_context *generic ) -{ - struct clipper *clipper = CALLOC_STRUCT(clipper); - - prim_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); - - clipper->stage.generic = generic; - clipper->stage.begin = clip_begin; - clipper->stage.point = clip_point; - clipper->stage.line = clip_line; - clipper->stage.tri = clip_tri; - clipper->stage.end = clip_end; - - return &clipper->stage; -} diff --git a/src/mesa/pipe/softpipe/g_prim_cull.c b/src/mesa/pipe/softpipe/g_prim_cull.c deleted file mode 100644 index 633fbcc492..0000000000 --- a/src/mesa/pipe/softpipe/g_prim_cull.c +++ /dev/null @@ -1,119 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_prim.h" - - - -struct cull_stage { - struct prim_stage stage; - GLuint mode; -}; - - - -static INLINE struct cull_stage *cull_stage( struct prim_stage *stage ) -{ - return (struct cull_stage *)stage; -} - - -static void cull_begin( struct prim_stage *stage ) -{ - struct cull_stage *cull = cull_stage(stage); - - cull->mode = stage->generic->setup.cull_mode; - - stage->next->begin( stage->next ); -} - - -static void cull_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - /* Window coords: */ - GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); - GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); - GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); - - GLfloat ex = v0[0] - v2[0]; - GLfloat ey = v0[1] - v2[1]; - GLfloat fx = v1[0] - v2[0]; - GLfloat fy = v1[1] - v2[1]; - - header->det = ex * fy - ey * fx; - - _mesa_printf("%s %f\n", __FUNCTION__, header->det ); - - if (header->det != 0) { - GLuint mode = (header->det < 0) ? WINDING_CW : WINDING_CCW; - - if ((mode & cull_stage(stage)->mode) == 0) - stage->next->tri( stage->next, header ); - } -} - - -static void cull_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void cull_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void cull_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_cull( struct generic_context *generic ) -{ - struct cull_stage *cull = CALLOC_STRUCT(cull_stage); - - prim_alloc_tmps( &cull->stage, 0 ); - - cull->stage.generic = generic; - cull->stage.next = NULL; - cull->stage.begin = cull_begin; - cull->stage.point = cull_point; - cull->stage.line = cull_line; - cull->stage.tri = cull_tri; - cull->stage.end = cull_end; - - return &cull->stage; -} diff --git a/src/mesa/pipe/softpipe/g_prim_flatshade.c b/src/mesa/pipe/softpipe/g_prim_flatshade.c deleted file mode 100644 index 1974a64fa8..0000000000 --- a/src/mesa/pipe/softpipe/g_prim_flatshade.c +++ /dev/null @@ -1,149 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_prim.h" - - -struct flatshade_stage { - struct prim_stage stage; - - const GLuint *lookup; -}; - - - -static INLINE struct flatshade_stage *flatshade_stage( struct prim_stage *stage ) -{ - return (struct flatshade_stage *)stage; -} - - -static void flatshade_begin( struct prim_stage *stage ) -{ - stage->next->begin( stage->next ); -} - - - -static INLINE void copy_attr( GLuint attr, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - if (attr) { - memcpy( dst->data[attr], - src->data[attr], - sizeof(src->data[0]) ); - } -} - -static void copy_colors( struct prim_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - struct flatshade_stage *flatshade = flatshade_stage(stage); - const GLuint *lookup = flatshade->lookup; - - copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); - copy_attr( lookup[VF_ATTRIB_COLOR1], dst, src ); - copy_attr( lookup[VF_ATTRIB_BFC0], dst, src ); - copy_attr( lookup[VF_ATTRIB_BFC1], dst, src ); -} - - - -/* Flatshade tri. Required for clipping and when unfilled tris are - * active, otherwise handled by hardware. - */ -static void flatshade_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = header->v[2]; - - copy_colors(stage, tmp.v[0], tmp.v[2]); - copy_colors(stage, tmp.v[1], tmp.v[2]); - - stage->next->tri( stage->next, &tmp ); -} - - -/* Flatshade line. Required for clipping. - */ -static void flatshade_line( struct prim_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = header->v[1]; - - copy_colors(stage, tmp.v[0], tmp.v[1]); - - stage->next->line( stage->next, &tmp ); -} - - -static void flatshade_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void flatshade_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_flatshade( struct generic_context *generic ) -{ - struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); - - prim_alloc_tmps( &flatshade->stage, 2 ); - - flatshade->stage.generic = generic; - flatshade->stage.next = NULL; - flatshade->stage.begin = flatshade_begin; - flatshade->stage.point = flatshade_point; - flatshade->stage.line = flatshade_line; - flatshade->stage.tri = flatshade_tri; - flatshade->stage.end = flatshade_end; - - flatshade->lookup = generic->vf_attr_to_slot; - - return &flatshade->stage; -} - - diff --git a/src/mesa/pipe/softpipe/g_prim_offset.c b/src/mesa/pipe/softpipe/g_prim_offset.c deleted file mode 100644 index 801ce9a441..0000000000 --- a/src/mesa/pipe/softpipe/g_prim_offset.c +++ /dev/null @@ -1,157 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" -#include "macros.h" - -#include "g_context.h" -#include "g_prim.h" - - - -struct offset_stage { - struct prim_stage stage; - - GLuint hw_data_offset; - - GLfloat scale; - GLfloat units; -}; - - - -static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) -{ - return (struct offset_stage *)stage; -} - - -static void offset_begin( struct prim_stage *stage ) -{ - struct offset_stage *offset = offset_stage(stage); - - offset->units = stage->generic->setup.offset_units; - offset->scale = stage->generic->setup.offset_scale; - - stage->next->begin( stage->next ); -} - - -/* Offset tri. Some hardware can handle this, but not usually when - * doing unfilled rendering. - */ -static void do_offset_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct offset_stage *offset = offset_stage(stage); - GLfloat inv_det = 1.0 / header->det; - - /* Window coords: - */ - GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); - GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); - GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); - - GLfloat ex = v0[0] - v2[2]; - GLfloat fx = v1[0] - v2[2]; - GLfloat ey = v0[1] - v2[2]; - GLfloat fy = v1[1] - v2[2]; - GLfloat ez = v0[2] - v2[2]; - GLfloat fz = v1[2] - v2[2]; - - GLfloat a = ey*fz - ez*fy; - GLfloat b = ez*fx - ex*fz; - - GLfloat ac = a * inv_det; - GLfloat bc = b * inv_det; - GLfloat zoffset; - - if ( ac < 0.0f ) ac = -ac; - if ( bc < 0.0f ) bc = -bc; - - zoffset = offset->units + MAX2( ac, bc ) * offset->scale; - - v0[2] += zoffset; - v1[2] += zoffset; - v2[2] += zoffset; - - stage->next->tri( stage->next, header ); -} - - -static void offset_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = dup_vert(stage, header->v[2], 2); - - do_offset_tri( stage->next, &tmp ); -} - - - -static void offset_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void offset_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void offset_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_offset( struct generic_context *generic ) -{ - struct offset_stage *offset = CALLOC_STRUCT(offset_stage); - - prim_alloc_tmps( &offset->stage, 3 ); - - offset->stage.generic = generic; - offset->stage.next = NULL; - offset->stage.begin = offset_begin; - offset->stage.point = offset_point; - offset->stage.line = offset_line; - offset->stage.tri = offset_tri; - offset->stage.end = offset_end; - - return &offset->stage; -} diff --git a/src/mesa/pipe/softpipe/g_prim_setup.c b/src/mesa/pipe/softpipe/g_prim_setup.c deleted file mode 100644 index 41e1381a60..0000000000 --- a/src/mesa/pipe/softpipe/g_prim_setup.c +++ /dev/null @@ -1,888 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "imports.h" -#include "macros.h" - -#include "g_context.h" -#include "g_prim.h" -#include "g_tile.h" - - -/** - * Triangle edge info - */ -struct edge { - GLfloat dx; /* X(v1) - X(v0), used only during setup */ - GLfloat dy; /* Y(v1) - Y(v0), used only during setup */ - GLfloat dxdy; /* dx/dy */ - GLfloat sx; /* first sample point x coord */ - GLfloat sy; - GLint lines; /* number of lines on this edge */ -}; - - -/** - * Triangle setup info (derived from prim_stage). - * Also used for line drawing (taking some liberties). - */ -struct setup_stage { - struct prim_stage stage; /**< This must be first */ - - /* Vertices are just an array of floats making up each attribute in - * turn. Currently fixed at 4 floats, but should change in time. - * Codegen will help cope with this. - */ - const struct vertex_header *vmax; - const struct vertex_header *vmid; - const struct vertex_header *vmin; - const struct vertex_header *vprovoke; - - struct edge ebot; - struct edge etop; - struct edge emaj; - - GLfloat oneoverarea; - - struct setup_coefficient coef[FRAG_ATTRIB_MAX]; - struct quad_header quad; - - struct { - GLint left[2]; /**< [0] = row0, [1] = row1 */ - GLint right[2]; - GLint y; - GLuint y_flags; - GLuint mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ - } span; -}; - - - -/** - * Basically a cast wrapper. - */ -static inline struct setup_stage *setup_stage( struct prim_stage *stage ) -{ - return (struct setup_stage *)stage; -} - - -/** - * Given an X or Y coordinate, return the block/quad coordinate that it - * belongs to. - */ -static inline GLint block( GLint x ) -{ - return x & ~1; -} - - - -static void setup_begin( struct prim_stage *stage ) -{ - setup_stage(stage)->quad.nr_attrs = stage->generic->nr_frag_attrs; -} - - -/** - * Run shader on a quad/block. - */ -static void run_shader_block( struct setup_stage *setup, - GLint x, GLint y, GLuint mask ) -{ - setup->quad.x0 = x; - setup->quad.y0 = y; - setup->quad.mask = mask; - - quad_shade( setup->stage.generic, &setup->quad ); -} - - -/** - * Compute mask which indicates which pixels in the 2x2 quad are actually inside - * the triangle's bounds. - * - * this is pretty nasty... may need to rework flush_spans again to - * fix it, if possible. - */ -static GLuint calculate_mask( struct setup_stage *setup, - GLint x ) -{ - GLuint mask = 0; - - if (x >= setup->span.left[0] && x < setup->span.right[0]) - mask |= MASK_BOTTOM_LEFT; - - if (x >= setup->span.left[1] && x < setup->span.right[1]) - mask |= MASK_TOP_LEFT; - - if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) - mask |= MASK_BOTTOM_RIGHT; - - if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) - mask |= MASK_TOP_RIGHT; - - return mask; -} - - -/** - * Render a horizontal span of quads - */ -static void flush_spans( struct setup_stage *setup ) -{ - GLint minleft, maxright; - GLint x; - - switch (setup->span.y_flags) { - case 3: - minleft = MIN2(setup->span.left[0], setup->span.left[1]); - maxright = MAX2(setup->span.right[0], setup->span.right[1]); - break; - - case 1: - minleft = setup->span.left[0]; - maxright = setup->span.right[0]; - break; - - case 2: - minleft = setup->span.left[1]; - maxright = setup->span.right[1]; - break; - - default: - return; - } - - - for (x = block(minleft); x <= block(maxright); ) - { - run_shader_block( setup, x, - setup->span.y, - calculate_mask( setup, x ) ); - x += 2; - } - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; -} - - -static GLboolean setup_sort_vertices( struct setup_stage *setup, - const struct prim_header *prim ) -{ - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - const struct vertex_header *v2 = prim->v[2]; - - setup->vprovoke = v2; - - /* determine bottom to top order of vertices */ - { - GLfloat y0 = v0->data[0][1]; - GLfloat y1 = v1->data[0][1]; - GLfloat y2 = v2->data[0][1]; - if (y0 <= y1) { - if (y1 <= y2) { - /* y0<=y1<=y2 */ - setup->vmin = v0; - setup->vmid = v1; - setup->vmax = v2; - } - else if (y2 <= y0) { - /* y2<=y0<=y1 */ - setup->vmin = v2; - setup->vmid = v0; - setup->vmax = v1; - } - else { - /* y0<=y2<=y1 */ - setup->vmin = v0; - setup->vmid = v2; - setup->vmax = v1; - } - } - else { - if (y0 <= y2) { - /* y1<=y0<=y2 */ - setup->vmin = v1; - setup->vmid = v0; - setup->vmax = v2; - } - else if (y2 <= y1) { - /* y2<=y1<=y0 */ - setup->vmin = v2; - setup->vmid = v1; - setup->vmax = v0; - } - else { - /* y1<=y2<=y0 */ - setup->vmin = v1; - setup->vmid = v2; - setup->vmax = v0; - } - } - } - - setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; - setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; - setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; - - /* xxx: may need to adjust this sign according to the if-tree - * above: - * - * XXX: this is like 'det', but calculated from screen coords?? - */ - { - const GLfloat area = (setup->emaj.dx * setup->ebot.dy - - setup->ebot.dx * setup->emaj.dy); - - setup->oneoverarea = 1.0 / area; - } - - /* XXX need to know if this is a front or back-facing triangle: - * - the GLSL gl_FrontFacing fragment attribute (bool) - * - two-sided stencil test - */ - setup->quad.facing = 0; - - _mesa_printf("%s one-over-area %f\n", __FUNCTION__, setup->oneoverarea ); - - - return GL_TRUE; -} - - -/** - * Compute a0 for a constant-valued coefficient (GL_FLAT shading). - */ -static void const_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i ) -{ - setup->coef[slot].dadx[i] = 0; - setup->coef[slot].dady[i] = 0; - - /* need provoking vertex info! - */ - setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; -} - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a triangle. - */ -static void tri_linear_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i) -{ - GLfloat botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; - GLfloat majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; - GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; - - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; - - /* calculate a0 as the value which would be sampled for the - * fragment at (0,0), taking into account that we want to sample at - * pixel centers, in other words (0.5, 0.5). - * - * this is neat but unfortunately not a good way to do things for - * triangles with very large values of dadx or dady as it will - * result in the subtraction and re-addition from a0 of a very - * large number, which means we'll end up loosing a lot of the - * fractional bits and precision from a0. the way to fix this is - * to define a0 as the sample at a pixel center somewhere near vmin - * instead - i'll switch to this later. - */ - setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); - - _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", - slot, "xyzw"[i], - setup->coef[slot].a0[i], - setup->coef[slot].dadx[i], - setup->coef[slot].dady[i]); -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a triangle. - */ -static void tri_persp_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i ) -{ - /* premultiply by 1/w: - */ - GLfloat mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; - GLfloat mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; - GLfloat maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; - - GLfloat botda = mida - mina; - GLfloat majda = maxa - mina; - GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; - GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; - - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; - setup->coef[slot].a0[i] = (mina - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); -} - - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. - */ -static void setup_tri_coefficients( struct setup_stage *setup ) -{ - const enum interp_mode *interp = setup->stage.generic->interp; - GLuint slot, j; - - /* z and w are done by linear interpolation: - */ - tri_linear_coeff(setup, 0, 2); - tri_linear_coeff(setup, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - switch (interp[slot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - tri_linear_coeff(setup, slot, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - tri_persp_coeff(setup, slot, j); - break; - } - } -} - - - -static void setup_tri_edges( struct setup_stage *setup ) -{ - GLfloat vmin_x = setup->vmin->data[0][0] + 0.5; - GLfloat vmid_x = setup->vmid->data[0][0] + 0.5; - - GLfloat vmin_y = setup->vmin->data[0][1] - 0.5; - GLfloat vmid_y = setup->vmid->data[0][1] - 0.5; - GLfloat vmax_y = setup->vmax->data[0][1] - 0.5; - - setup->emaj.sy = ceilf(vmin_y); - setup->emaj.lines = (GLint) ceilf(vmax_y - setup->emaj.sy); - setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; - setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; - - setup->etop.sy = ceilf(vmid_y); - setup->etop.lines = (GLint) ceilf(vmax_y - setup->etop.sy); - setup->etop.dxdy = setup->etop.dx / setup->etop.dy; - setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; - - setup->ebot.sy = ceilf(vmin_y); - setup->ebot.lines = (GLint) ceilf(vmid_y - setup->ebot.sy); - setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; - setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; -} - - -/** - * Render the upper or lower half of a triangle. - * Scissoring is applied here too. - */ -static void subtriangle( struct setup_stage *setup, - struct edge *eleft, - struct edge *eright, - GLuint lines ) -{ - GLint y, start_y, finish_y; - GLint sy = (GLint)eleft->sy; - - assert((GLint)eleft->sy == (GLint) eright->sy); - assert((GLint)eleft->sy >= 0); /* catch bug in x64? */ - - /* scissor y: - */ - if (setup->stage.generic->setup.scissor) { - start_y = sy; - finish_y = start_y + lines; - - if (start_y < setup->stage.generic->scissor.miny) - start_y = setup->stage.generic->scissor.miny; - - if (finish_y > setup->stage.generic->scissor.maxy) - finish_y = setup->stage.generic->scissor.maxy; - - start_y -= sy; - finish_y -= sy; - } - else { - start_y = 0; - finish_y = lines; - } - - _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); - - for (y = start_y; y < finish_y; y++) { - - /* avoid accumulating adds as floats don't have the precision to - * accurately iterate large triangle edges that way. luckily we - * can just multiply these days. - * - * this is all drowned out by the attribute interpolation anyway. - */ - GLint left = (GLint)(eleft->sx + y * eleft->dxdy); - GLint right = (GLint)(eright->sx + y * eright->dxdy); - - /* scissor x: - */ - if (setup->stage.generic->setup.scissor) { - if (left < setup->stage.generic->scissor.minx) - left = setup->stage.generic->scissor.minx; - - if (right > setup->stage.generic->scissor.maxx) - right = setup->stage.generic->scissor.maxx; - } - - if (left < right) { - GLint _y = sy+y; - if (block(_y) != setup->span.y) { - flush_spans(setup); - setup->span.y = block(_y); - } - - setup->span.left[_y&1] = left; - setup->span.right[_y&1] = right; - setup->span.y_flags |= 1<<(_y&1); - } - } - - - /* save the values so that emaj can be restarted: - */ - eleft->sx += lines * eleft->dxdy; - eright->sx += lines * eright->dxdy; - eleft->sy += lines; - eright->sy += lines; -} - - -/** - * Do setup for triangle rasterization, then render the triangle. - */ -static void setup_tri( struct prim_stage *stage, - struct prim_header *prim ) -{ - struct setup_stage *setup = setup_stage( stage ); - - _mesa_printf("%s\n", __FUNCTION__ ); - - setup_sort_vertices( setup, prim ); - setup_tri_coefficients( setup ); - setup_tri_edges( setup ); - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; -// setup->span.z_mode = tri_z_mode( setup->ctx ); - -// init_constant_attribs( setup ); - - if (setup->oneoverarea < 0.0) { - /* emaj on left: - */ - subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); - subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); - } - else { - /* emaj on right: - */ - subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); - subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); - } - - flush_spans( setup ); -} - - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a line. - */ -static void -line_linear_coeff(struct setup_stage *setup, GLuint slot, GLuint i) -{ - const GLfloat dz = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - const GLfloat dadx = dz * setup->emaj.dx * setup->oneoverarea; - const GLfloat dady = dz * setup->emaj.dy * setup->oneoverarea; - setup->coef[slot].dadx[i] = dadx; - setup->coef[slot].dady[i] = dady; - setup->coef[slot].a0[i] - = (setup->vmin->data[slot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5) + - dady * (setup->vmin->data[0][1] - 0.5))); -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a line. - */ -static void -line_persp_coeff(struct setup_stage *setup, GLuint slot, GLuint i) -{ - /* XXX to do */ - line_linear_coeff(setup, slot, i); /* XXX temporary */ -} - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmax are initialized. - */ -static INLINE void -setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) -{ - const enum interp_mode *interp = setup->stage.generic->interp; - GLuint slot, j; - - /* use setup->vmin, vmax to point to vertices */ - setup->vprovoke = prim->v[1]; - setup->vmin = prim->v[0]; - setup->vmax = prim->v[1]; - - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - /* NOTE: this is not really 1/area */ - setup->oneoverarea = 1.0 / (setup->emaj.dx * setup->emaj.dx + - setup->emaj.dy * setup->emaj.dy); - - /* z and w are done by linear interpolation: - */ - line_linear_coeff(setup, 0, 2); - line_linear_coeff(setup, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - switch (interp[slot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - line_linear_coeff(setup, slot, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - line_persp_coeff(setup, slot, j); - break; - } - } -} - - -/** - * Plot a pixel in a line segment. - */ -static INLINE void -plot(struct setup_stage *setup, GLint x, GLint y) -{ - const GLint iy = y & 1; - const GLint ix = x & 1; - const GLint quadX = x - ix; - const GLint quadY = y - iy; - const GLint mask = (1 << ix) << (2 * iy); - - if (quadX != setup->quad.x0 || - quadY != setup->quad.y0) - { - /* flush prev quad, start new quad */ - - if (setup->quad.x0 != -1) - quad_shade(setup->stage.generic, &setup->quad); - - setup->quad.x0 = quadX; - setup->quad.y0 = quadY; - setup->quad.mask = 0x0; - } - - setup->quad.mask |= mask; -} - - - -/** - * Do setup for line rasterization, then render the line. - * XXX single-pixel width, no stipple, etc - * XXX no scissoring yet. - */ -static void -setup_line(struct prim_stage *stage, struct prim_header *prim) -{ - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - struct setup_stage *setup = setup_stage( stage ); - - GLint x0 = (GLint) v0->data[0][0]; - GLint x1 = (GLint) v1->data[0][0]; - GLint y0 = (GLint) v0->data[0][1]; - GLint y1 = (GLint) v1->data[0][1]; - GLint dx = x1 - x0; - GLint dy = y1 - y0; - GLint xstep, ystep; - - if (dx == 0 && dy == 0) - return; - - setup_line_coefficients(setup, prim); - - if (dx < 0) { - dx = -dx; /* make positive */ - xstep = -1; - } - else { - xstep = 1; - } - - if (dy < 0) { - dy = -dy; /* make positive */ - ystep = -1; - } - else { - ystep = 1; - } - - assert(dx >= 0); - assert(dy >= 0); - - setup->quad.x0 = setup->quad.y0 = -1; - setup->quad.mask = 0x0; - - if (dx > dy) { - /*** X-major line ***/ - GLint i; - const GLint errorInc = dy + dy; - GLint error = errorInc - dx; - const GLint errorDec = error - dx; - - for (i = 0; i < dx; i++) { - plot(setup, x0, y0); - - x0 += xstep; - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - y0 += ystep; - } - } - } - else { - /*** Y-major line ***/ - GLint i; - const GLint errorInc = dx + dx; - GLint error = errorInc - dy; - const GLint errorDec = error - dy; - - for (i = 0; i < dy; i++) { - plot(setup, x0, y0); - - y0 += ystep; - - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - x0 += xstep; - } - } - } - - /* draw final quad */ - if (setup->quad.mask) { - quad_shade(setup->stage.generic, &setup->quad); - } -} - - -/** - * Do setup for point rasterization, then render the point. - * Round or square points... - * XXX could optimize a lot for 1-pixel points. - */ -static void -setup_point(struct prim_stage *stage, struct prim_header *prim) -{ - struct setup_stage *setup = setup_stage( stage ); - GLfloat halfSize = 7.3; /*XXX this is a vertex attrib */ - GLfloat halfSizeSquared = halfSize * halfSize; - const struct vertex_header *v0 = prim->v[0]; - const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; - const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; - const GLint ixmin = block((GLint) (x - halfSize)); - const GLint ixmax = block((GLint) (x + halfSize)); - const GLint iymin = block((GLint) (y - halfSize)); - const GLint iymax = block((GLint) (y + halfSize)); - GLboolean round = GL_TRUE; - GLint ix, iy; - GLuint slot, j; - - /* For points, all interpolants are constant-valued. - * However, for point sprites, we'll need to setup texcoords appropriately. - * XXX: which coefficients are the texcoords??? - * We may do point sprites as textured quads... - * - * KW: We don't know which coefficients are texcoords - ultimately - * the choice of what interpolation mode to use for each attribute - * should be determined by the fragment program, using - * per-attribute declaration statements that include interpolation - * mode as a parameter. So either the fragment program will have - * to be adjusted for pointsprite vs normal point behaviour, or - * otherwise a special interpolation mode will have to be defined - * which matches the required behaviour for point sprites. But - - * the latter is not a feature of normal hardware, and as such - * probably should be ruled out on that basis. - */ - setup->vprovoke = prim->v[0]; - const_coeff(setup, 0, 2); - const_coeff(setup, 0, 3); - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - } - - /* XXX need to clip against scissor bounds too */ - - for (iy = iymin; iy <= iymax; iy += 2) { - for (ix = ixmin; ix <= ixmax; ix += 2) { - - if (round) { - /* rounded points */ - /* XXX for GL_SMOOTH, need to compute per-fragment coverage too */ - GLfloat dx, dy; - - setup->quad.mask = 0x0; - - dx = (ix + 0.5) - x; - dy = (iy + 0.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_BOTTOM_LEFT; - - dx = (ix + 1.5) - x; - dy = (iy + 0.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_BOTTOM_RIGHT; - - dx = (ix + 0.5) - x; - dy = (iy + 1.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_TOP_LEFT; - - dx = (ix + 1.5) - x; - dy = (iy + 1.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_TOP_RIGHT; - } - else { - /* square points */ - setup->quad.mask = 0xf; - - if (ix + 0.5 < x - halfSize) - setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - - if (ix + 1.5 > x + halfSize) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - - if (iy + 0.5 < y - halfSize) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); - - if (iy + 1.5 > y + halfSize) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - } - - if (setup->quad.mask) { - setup->quad.x0 = ix; - setup->quad.y0 = iy; - quad_shade( setup->stage.generic, &setup->quad ); - } - } - } -} - - - -static void setup_end( struct prim_stage *stage ) -{ -} - - -struct prim_stage *prim_setup( struct generic_context *generic ) -{ - struct setup_stage *setup = CALLOC_STRUCT(setup_stage); - - setup->stage.generic = generic; - setup->stage.begin = setup_begin; - setup->stage.point = setup_point; - setup->stage.line = setup_line; - setup->stage.tri = setup_tri; - setup->stage.end = setup_end; - - setup->quad.coef = setup->coef; - - return &setup->stage; -} diff --git a/src/mesa/pipe/softpipe/g_prim_setup.h b/src/mesa/pipe/softpipe/g_prim_setup.h deleted file mode 100644 index 40a70c543e..0000000000 --- a/src/mesa/pipe/softpipe/g_prim_setup.h +++ /dev/null @@ -1,121 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#ifndef S_TRI_H -#define S_TRI_H - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "glheader.h" -#include "imports.h" -#include "s_tri_public.h" -#include "s_context.h" - - - -struct tri_context; -struct fp_context; -struct be_context; - -/* Note the rasterizer does not take a GLcontext argument. This is - * deliberate. - */ -struct tri_context *tri_create_context( GLcontext *ctx ); - -void tri_destroy_context( struct tri_context *tri ); - -void tri_set_fp_context( struct tri_context *tri, - struct fp_context *fp, - void (*fp_run)( struct fp_context *fp, - const struct fp_inputs *, - struct fp_outputs * )); - - -void tri_set_be_context( struct tri_context *tri, - struct be_context *be, - void (*be_run)( struct be_context *be, - const struct fp_outputs * )); - -void tri_set_attribs( struct tri_context *tri, - const struct attr_info *info, - GLuint nr_attrib ); - -void tri_set_backface( struct tri_context *tri, - GLfloat backface ); - -void tri_set_scissor( struct tri_context *tri, - GLint x, - GLint y, - GLuint width, - GLuint height, - GLboolean enabled ); - -void tri_set_stipple( struct tri_context *tri, - const GLuint *pattern, - GLboolean enabled ); - -/* Unfilled triangles will be handled elsewhere (higher in the - * pipeline), as will things like stipple (lower in the pipeline). - */ - -void tri_triangle( struct tri_context *tri, - const struct vertex *v0, - const struct vertex *v1, - const struct vertex *v2 ); - -/* TODO: rasterize_line, rasterize_point?? - * How will linestipple work? - */ - - -#ifdef SETUP_PRIVATE - - - - - - -GLboolean tri_setup( struct tri_context *tri, - const struct vertex *v0, - const struct vertex *v1, - const struct vertex *v2 ); - -void tri_rasterize( struct tri_context *tri ); -void tri_rasterize_spans( struct tri_context *tri ); - - - - - - -#endif -#endif diff --git a/src/mesa/pipe/softpipe/g_prim_twoside.c b/src/mesa/pipe/softpipe/g_prim_twoside.c deleted file mode 100644 index 334ae2f537..0000000000 --- a/src/mesa/pipe/softpipe/g_prim_twoside.c +++ /dev/null @@ -1,149 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" -#include "g_context.h" -#include "g_prim.h" - - -struct twoside_stage { - struct prim_stage stage; - - GLfloat facing; - const GLuint *lookup; -}; - - -static INLINE struct twoside_stage *twoside_stage( struct prim_stage *stage ) -{ - return (struct twoside_stage *)stage; -} - - -static void twoside_begin( struct prim_stage *stage ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - - twoside->facing = (stage->generic->setup.front_winding == WINDING_CW) ? -1 : 1; - - stage->next->begin( stage->next ); -} - - -static INLINE void copy_color( GLuint attr_dst, - GLuint attr_src, - struct vertex_header *v ) -{ - if (attr_dst && attr_src) { - memcpy( v->data[attr_dst], - v->data[attr_src], - sizeof(v->data[0]) ); - } -} - - -static struct vertex_header *copy_bfc( struct twoside_stage *twoside, - const struct vertex_header *v, - GLuint idx ) -{ - struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); - - copy_color( twoside->lookup[VF_ATTRIB_COLOR0], - twoside->lookup[VF_ATTRIB_BFC0], - tmp ); - - copy_color( twoside->lookup[VF_ATTRIB_COLOR1], - twoside->lookup[VF_ATTRIB_BFC1], - tmp ); - - return tmp; -} - - -/* Twoside tri: - */ -static void twoside_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - - if (header->det * twoside->facing < 0) { - struct prim_header tmp; - - tmp.det = header->det; - tmp.v[0] = copy_bfc(twoside, header->v[0], 0); - tmp.v[1] = copy_bfc(twoside, header->v[1], 1); - tmp.v[2] = copy_bfc(twoside, header->v[2], 2); - - stage->next->tri( stage->next, &tmp ); - } - else { - stage->next->tri( stage->next, header ); - } -} - - -static void twoside_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void twoside_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void twoside_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - - - -struct prim_stage *prim_twoside( struct generic_context *generic ) -{ - struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); - - prim_alloc_tmps( &twoside->stage, 3 ); - - twoside->stage.generic = generic; - twoside->stage.next = NULL; - twoside->stage.begin = twoside_begin; - twoside->stage.point = twoside_point; - twoside->stage.line = twoside_line; - twoside->stage.tri = twoside_tri; - twoside->stage.end = twoside_end; - - twoside->lookup = generic->vf_attr_to_slot; - - return &twoside->stage; -} diff --git a/src/mesa/pipe/softpipe/g_prim_unfilled.c b/src/mesa/pipe/softpipe/g_prim_unfilled.c deleted file mode 100644 index 7c6ee79f92..0000000000 --- a/src/mesa/pipe/softpipe/g_prim_unfilled.c +++ /dev/null @@ -1,165 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_prim.h" - - - -struct unfilled_stage { - struct prim_stage stage; - - GLuint mode[2]; -}; - - -static INLINE struct unfilled_stage *unfilled_stage( struct prim_stage *stage ) -{ - return (struct unfilled_stage *)stage; -} - - -static void unfilled_begin( struct prim_stage *stage ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - - unfilled->mode[0] = stage->generic->setup.fill_ccw; - unfilled->mode[1] = stage->generic->setup.fill_cw; - - stage->next->begin( stage->next ); -} - -static void point( struct prim_stage *stage, - struct vertex_header *v0 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - stage->next->point( stage->next, &tmp ); -} - -static void line( struct prim_stage *stage, - struct vertex_header *v0, - struct vertex_header *v1 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - tmp.v[1] = v1; - stage->next->line( stage->next, &tmp ); -} - - -static void points( struct prim_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - - if (v0->edgeflag) point( stage, v0 ); - if (v1->edgeflag) point( stage, v1 ); - if (v2->edgeflag) point( stage, v2 ); -} - -static void lines( struct prim_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - - if (v0->edgeflag) line( stage, v0, v1 ); - if (v1->edgeflag) line( stage, v1, v2 ); - if (v2->edgeflag) line( stage, v2, v0 ); -} - - -/* Unfilled tri: - * - * Note edgeflags in the vertex struct is not sufficient as we will - * need to manipulate them when decomposing primitives??? - */ -static void unfilled_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - GLuint mode = unfilled->mode[header->det < 0]; - - switch (mode) { - case FILL_TRI: - stage->next->tri( stage->next, header ); - break; - - case FILL_LINE: - lines( stage, header ); - break; - - case GL_POINT: - points( stage, header ); - break; - } -} - -static void unfilled_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void unfilled_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void unfilled_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_unfilled( struct generic_context *generic ) -{ - struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); - - prim_alloc_tmps( &unfilled->stage, 0 ); - - unfilled->stage.generic = generic; - unfilled->stage.next = NULL; - unfilled->stage.tmp = NULL; - unfilled->stage.begin = unfilled_begin; - unfilled->stage.point = unfilled_point; - unfilled->stage.line = unfilled_line; - unfilled->stage.tri = unfilled_tri; - unfilled->stage.end = unfilled_end; - - return &unfilled->stage; -} diff --git a/src/mesa/pipe/softpipe/g_state.h b/src/mesa/pipe/softpipe/g_state.h deleted file mode 100644 index 5077d37ab7..0000000000 --- a/src/mesa/pipe/softpipe/g_state.h +++ /dev/null @@ -1,61 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef PRIM_H -#define PRIM_H - -#include "glheader.h" -#include "softpipe/sp_state.h" - - -void generic_set_clip_state( struct softpipe_context *, - const struct softpipe_clip_state * ); - -void generic_set_viewport( struct softpipe_context *, - const struct softpipe_viewport * ); - -void generic_set_setup_state( struct softpipe_context *, - const struct softpipe_setup_state * ); - -void generic_set_scissor_rect( struct softpipe_context *, - const struct softpipe_scissor_rect * ); - -void generic_set_fs_state( struct softpipe_context *, - const struct softpipe_fs_state * ); - -void generic_set_polygon_stipple( struct softpipe_context *, - const struct softpipe_poly_stipple * ); - -void generic_set_cbuf_state( struct softpipe_context *, - const struct softpipe_surface * ); - -void generic_update_derived( struct generic_context *generic ); - -#endif diff --git a/src/mesa/pipe/softpipe/g_state_clip.c b/src/mesa/pipe/softpipe/g_state_clip.c deleted file mode 100644 index b90034e11e..0000000000 --- a/src/mesa/pipe/softpipe/g_state_clip.c +++ /dev/null @@ -1,69 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_state.h" -#include "g_draw.h" - - - -void generic_set_clip_state( struct softpipe_context *softpipe, - const struct softpipe_clip_state *clip ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy(&generic->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); - - generic->nr_planes = 6 + clip->nr; - generic->dirty |= G_NEW_CLIP; -} - - - -/* Called when driver state tracker notices changes to the viewport - * matrix: - */ -void generic_set_viewport( struct softpipe_context *softpipe, - const struct softpipe_viewport *viewport ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy(&generic->viewport, viewport, sizeof(viewport)); - - /* Using tnl/ and vf/ modules is temporary while getting started. - * Full softpipe will have vertex shader, vertex fetch of its own. - */ - draw_set_viewport( generic->draw, viewport->scale, viewport->translate ); - generic->dirty |= G_NEW_VIEWPORT; -} - - - diff --git a/src/mesa/pipe/softpipe/g_state_derived.c b/src/mesa/pipe/softpipe/g_state_derived.c deleted file mode 100644 index 678825afe1..0000000000 --- a/src/mesa/pipe/softpipe/g_state_derived.c +++ /dev/null @@ -1,136 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "macros.h" -#include "enums.h" -#include "program.h" - -#include "g_context.h" -#include "g_draw.h" -#include "g_state.h" - -#define EMIT_ATTR( ATTR, FRAG_ATTR, INTERP ) \ -do { \ - slot_to_vf_attr[generic->nr_attrs] = ATTR; \ - generic->vf_attr_to_slot[ATTR] = generic->nr_attrs; \ - generic->fp_attr_to_slot[FRAG_ATTR] = generic->nr_attrs; \ - generic->interp[generic->nr_attrs] = INTERP; \ - generic->nr_attrs++; \ - attr_mask |= (1<fs.fp; - const GLuint inputsRead = fp->Base.InputsRead; - GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; - GLuint attr_mask = 0; - GLuint i; - - generic->nr_attrs = 0; - memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); - - memset(generic->fp_attr_to_slot, 0, sizeof(generic->vf_attr_to_slot)); - memset(generic->vf_attr_to_slot, 0, sizeof(generic->fp_attr_to_slot)); - - /* TODO - Figure out if we need to do perspective divide, etc. - */ - EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); - - /* Pull in the rest of the attributes. They are all in float4 - * format. Future optimizations could be to keep some attributes - * as fixed point or ubyte format. - */ - for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { - if (inputsRead & (i << i)) { - EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); - } - } - - for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { - if (inputsRead & (i << i)) { - EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); - } - } - - generic->nr_frag_attrs = generic->nr_attrs; - - /* Additional attributes required for setup: Just twosided - * lighting. Edgeflag is dealt with specially by setting bits in - * the vertex header. - */ - if (generic->setup.light_twoside) { - if (inputsRead & FRAG_BIT_COL0) { - EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ - } - - if (inputsRead & FRAG_BIT_COL1) { - EMIT_ATTR(VF_ATTRIB_BFC1, FRAG_ATTRIB_MAX, 0); /* XXX: discard after setup */ - } - } - - if (attr_mask != generic->attr_mask) { - generic->attr_mask = attr_mask; - - draw_set_vertex_attributes( generic->draw, - slot_to_vf_attr, - generic->nr_attrs ); - } -} - - -/* Hopefully this will remain quite simple, otherwise need to pull in - * something like the state tracker mechanism. - */ -void generic_update_derived( struct generic_context *generic ) -{ - if (generic->dirty & (G_NEW_SETUP | G_NEW_FS)) - calculate_vertex_layout( generic ); - - generic->dirty = 0; -} diff --git a/src/mesa/pipe/softpipe/g_state_fs.c b/src/mesa/pipe/softpipe/g_state_fs.c deleted file mode 100644 index 8a430c0c65..0000000000 --- a/src/mesa/pipe/softpipe/g_state_fs.c +++ /dev/null @@ -1,50 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "g_context.h" -#include "g_state.h" - - - -void generic_set_fs_state( struct softpipe_context *softpipe, - const struct softpipe_fs_state *fs ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy(&generic->fs, fs, sizeof(*fs)); - - generic->dirty |= G_NEW_FS; -} - - - - - - - - - diff --git a/src/mesa/pipe/softpipe/g_state_setup.c b/src/mesa/pipe/softpipe/g_state_setup.c deleted file mode 100644 index 2a5176c5a1..0000000000 --- a/src/mesa/pipe/softpipe/g_state_setup.c +++ /dev/null @@ -1,122 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "g_context.h" -#include "g_state.h" -#include "g_prim.h" - - - - -static void validate_prim_pipe( struct generic_context *generic ) -{ - struct prim_stage *next = generic->prim.setup; - - /* TODO: make the current primitive part of the state and build - * shorter pipelines for lines & points. - */ - if (generic->setup.fill_cw != FILL_TRI || - generic->setup.fill_ccw != FILL_TRI) { - - generic->prim.unfilled->next = next; - next = generic->prim.unfilled; - } - - if (generic->setup.offset_cw || - generic->setup.offset_ccw) { - generic->prim.offset->next = next; - next = generic->prim.offset; - } - - if (generic->setup.light_twoside) { - generic->prim.twoside->next = next; - next = generic->prim.twoside; - } - - /* Always run the cull stage as we calculate determinant there - * also. Fix this.. - */ - { - generic->prim.cull->next = next; - next = generic->prim.cull; - } - - - /* Clip stage - */ - { - generic->prim.clip->next = next; - next = generic->prim.clip; - } - - /* Do software flatshading prior to clipping. XXX: should only do - * this for clipped primitives, ie it is a part of the clip - * routine. - */ - if (generic->setup.flatshade) { - generic->prim.flatshade->next = next; - next = generic->prim.flatshade; - } - - - generic->prim.first = next; -} - - - - -void generic_set_setup_state( struct softpipe_context *softpipe, - const struct softpipe_setup_state *setup ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy( &generic->setup, setup, sizeof(*setup) ); - - validate_prim_pipe( generic ); - generic->dirty |= G_NEW_SETUP; -} - - - -void generic_set_scissor_rect( struct softpipe_context *softpipe, - const struct softpipe_scissor_rect *scissor ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy( &generic->scissor, scissor, sizeof(*scissor) ); - generic->dirty |= G_NEW_SCISSOR; -} - - -void generic_set_polygon_stipple( struct softpipe_context *softpipe, - const struct softpipe_poly_stipple *stipple ) -{ - struct generic_context *generic = generic_context(softpipe); - - memcpy( &generic->poly_stipple, stipple, sizeof(*stipple) ); - generic->dirty |= G_NEW_STIPPLE; -} diff --git a/src/mesa/pipe/softpipe/g_state_surface.c b/src/mesa/pipe/softpipe/g_state_surface.c deleted file mode 100644 index c791ce4640..0000000000 --- a/src/mesa/pipe/softpipe/g_state_surface.c +++ /dev/null @@ -1,53 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "g_context.h" -#include "g_state.h" -#include "g_surface.h" - - -/* This is all a total hack. - */ -void generic_set_cbuf_state( struct softpipe_context *softpipe, - const struct softpipe_surface *surface ) -{ - struct generic_context *generic = generic_context(softpipe); - - if (generic->cbuf_surface == NULL) { - generic->cbuf_surface = CALLOC_STRUCT(generic_surface); - generic->cbuf_surface->type = &gs_rgba8; - } - - generic->cbuf_surface->surface = *surface; - generic->dirty |= G_NEW_CBUF; -} - - diff --git a/src/mesa/pipe/softpipe/g_surface.c b/src/mesa/pipe/softpipe/g_surface.c deleted file mode 100644 index 2df30cde70..0000000000 --- a/src/mesa/pipe/softpipe/g_surface.c +++ /dev/null @@ -1,153 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "g_context.h" -#include "g_state.h" -#include "g_surface.h" -#include "g_headers.h" - -static void rgba8_read_quad_f( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - rgba[k][0] = ptr[0] * (1.0 / 255.0); - rgba[k][1] = ptr[1] * (1.0 / 255.0); - rgba[k][2] = ptr[2] * (1.0 / 255.0); - rgba[k][3] = ptr[3] * (1.0 / 255.0); - } - } -} - -static void rgba8_read_quad_f_swz( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - rrrr[0][k] = ptr[0] * (1.0 / 255.0); - rrrr[1][k] = ptr[1] * (1.0 / 255.0); - rrrr[2][k] = ptr[2] * (1.0 / 255.0); - rrrr[3][k] = ptr[3] * (1.0 / 255.0); - } - } -} - -static void rgba8_write_quad_f( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - ptr[0] = rgba[k][0] * 255.0; - ptr[1] = rgba[k][1] * 255.0; - ptr[2] = rgba[k][2] * 255.0; - ptr[3] = rgba[k][3] * 255.0; - } - } -} - -static void rgba8_write_quad_f_swz( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - ptr[0] = rrrr[0][k] * 255.0; - ptr[1] = rrrr[1][k] * 255.0; - ptr[2] = rrrr[2][k] * 255.0; - ptr[3] = rrrr[3][k] * 255.0; - } - } -} - - - - -static void rgba8_read_quad_ub( struct generic_surface *gs, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - rgba[k][0] = ptr[0]; - rgba[k][1] = ptr[1]; - rgba[k][2] = ptr[2]; - rgba[k][3] = ptr[3]; - } - } -} - - -static void rgba8_write_quad_ub( struct generic_surface *gs, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - ptr[0] = rgba[k][0]; - ptr[1] = rgba[k][1]; - ptr[2] = rgba[k][2]; - ptr[3] = rgba[k][3]; - } - } -} - - - - -struct generic_surface_type gs_rgba8 = { - G_SURFACE_RGBA_8888, - rgba8_read_quad_f, - rgba8_read_quad_f_swz, - rgba8_read_quad_ub, - rgba8_write_quad_f, - rgba8_write_quad_f_swz, - rgba8_write_quad_ub, -}; - - - diff --git a/src/mesa/pipe/softpipe/g_surface.h b/src/mesa/pipe/softpipe/g_surface.h deleted file mode 100644 index 23c28f0bb4..0000000000 --- a/src/mesa/pipe/softpipe/g_surface.h +++ /dev/null @@ -1,115 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_SURFACE_H -#define G_SURFACE_H - -#include "glheader.h" -#include "g_headers.h" - -struct generic_surface; - -#define G_SURFACE_RGBA_8888 0x1 - -/* Internal structs and helpers for the primitive clip/setup pipeline: - */ -struct generic_surface_type { - - GLuint format; - - void (*read_quad_f)( struct generic_surface *, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ); - - void (*read_quad_f_swz)( struct generic_surface *, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ); - - void (*read_quad_ub)( struct generic_surface *, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ); - - - void (*write_quad_f)( struct generic_surface *, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ); - - void (*write_quad_f_swz)( struct generic_surface *, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ); - - - void (*write_quad_ub)( struct generic_surface *, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ); - - -}; - - -struct generic_surface { - struct generic_surface_type *type; - struct softpipe_surface surface; -}; - - -static INLINE void gs_read_quad_f( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - gs->type->read_quad_f(gs, x, y, rgba); -} - -static INLINE void gs_read_quad_f_swz( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - gs->type->read_quad_f_swz(gs, x, y, rrrr); -} - -static INLINE void gs_write_quad_f( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - gs->type->write_quad_f(gs, x, y, rgba); -} - -static INLINE void gs_write_quad_f_swz( struct generic_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - gs->type->write_quad_f_swz(gs, x, y, rrrr); -} - -/* Like this, or hidden? - */ -struct generic_surface_type gs_rgba8; - -#endif diff --git a/src/mesa/pipe/softpipe/g_tile.h b/src/mesa/pipe/softpipe/g_tile.h deleted file mode 100644 index 6ae5212356..0000000000 --- a/src/mesa/pipe/softpipe/g_tile.h +++ /dev/null @@ -1,42 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#ifndef G_TILE_H -#define G_TILE_H - -struct generic_context; -struct quad_header; - -void quad_shade( struct generic_context *generic, - struct quad_header *quad ); - -void quad_output( struct generic_context *generic, - struct quad_header *quad ); - -#endif diff --git a/src/mesa/pipe/softpipe/g_tile_fs.c b/src/mesa/pipe/softpipe/g_tile_fs.c deleted file mode 100644 index 4c5e6efaed..0000000000 --- a/src/mesa/pipe/softpipe/g_tile_fs.c +++ /dev/null @@ -1,203 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "glheader.h" -#include "g_context.h" -#include "g_headers.h" -#include "g_tile.h" - -struct exec_machine { - const struct setup_coefficient *coef; - - GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE]; -}; - - -/** - * Compute quad's attributes values, as constants (GL_FLAT shading). - */ -static INLINE void cinterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) -{ - GLuint j; - - for (j = 0; j < QUAD_SIZE; j++) { - exec->attr[attrib][i][j] = exec->coef[attrib].a0[i]; - } -} - - -/** - * Compute quad's attribute values by linear interpolation. - * - * Push into the fp: - * - * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx - * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy - */ -static INLINE void linterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) -{ - GLuint j; - - for (j = 0; j < QUAD_SIZE; j++) { - const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; - exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * x + - exec->coef[attrib].dady[i] * y); - } -} - - -/** - * Compute quad's attribute values by linear interpolation with - * perspective correction. - * - * Push into the fp: - * - * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx - * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy - * INPUT[attr] = MUL INPUT[attr], INPUT_WPOS.wwww - * - * (Or should that be 1/w ???) - */ -static INLINE void pinterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) -{ - GLuint j; - - for (j = 0; j < QUAD_SIZE; j++) { - const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; - const GLfloat invW = exec->attr[FRAG_ATTRIB_WPOS][3][j]; - exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * x + - exec->coef[attrib].dady[i] * y) * invW); - } -} - - - -/* This should be done by the fragment shader execution unit (code - * generated from the decl instructions). Do it here for now. - */ -void quad_shade( struct generic_context *generic, - struct quad_header *quad ) -{ - struct exec_machine exec; - GLfloat fx = quad->x0; - GLfloat fy = quad->y0; - GLuint i, j; - - exec.coef = quad->coef; - - /* Position: - */ - exec.attr[FRAG_ATTRIB_WPOS][0][0] = fx; - exec.attr[FRAG_ATTRIB_WPOS][0][1] = fx + 1.0; - exec.attr[FRAG_ATTRIB_WPOS][0][2] = fx; - exec.attr[FRAG_ATTRIB_WPOS][0][3] = fx + 1.0; - - exec.attr[FRAG_ATTRIB_WPOS][1][0] = fy; - exec.attr[FRAG_ATTRIB_WPOS][1][1] = fy; - exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0; - exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; - - /* Z and W are done by linear interpolation: - * XXX we'll probably have to use integers for Z - */ - if (generic->need_z) { - linterp(&exec, 0, 2); - } - - if (generic->need_w) { - linterp(&exec, 0, 3); -// invert(&exec, 0, 3); - } - - /* Interpolate all the remaining attributes. This will get pushed - * into the fragment program's responsibilities at some point. - */ - for (i = 1; i < quad->nr_attrs; i++) { -#if 1 - for (j = 0; j < NUM_CHANNELS; j++) - linterp(&exec, i, j); -#else - switch (quad->interp[i]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - cinterp(&exec, i, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - linterp(&exec, i, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - pinterp(&exec, i, j); - break; - } -#endif - } - -#if 0 - generic->run_fs( tri->fp, quad, &tri->outputs ); -#else - { - GLuint attr = generic->fp_attr_to_slot[FRAG_ATTRIB_COL0]; - assert(attr); - - memcpy(quad->outputs.color, - exec.attr[attr], - sizeof(quad->outputs.color)); - } -#endif - - - if (quad->mask) - quad_output( generic, quad ); -} - - - - - - - diff --git a/src/mesa/pipe/softpipe/g_tile_output.c b/src/mesa/pipe/softpipe/g_tile_output.c deleted file mode 100644 index 058ca30193..0000000000 --- a/src/mesa/pipe/softpipe/g_tile_output.c +++ /dev/null @@ -1,92 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "glheader.h" -#include "g_context.h" -#include "g_headers.h" -#include "g_surface.h" -#include "g_tile.h" - - -static void mask_copy( GLfloat (*dest)[4], - GLfloat (*src)[4], - GLuint mask ) -{ - GLuint i, j; - - for (i = 0; i < 4; i++) { - if (mask & (1<mask != MASK_ALL) - { - GLfloat tmp[4][QUAD_SIZE]; - - /* Yes, we'll probably have a masked write as well, but this is - * how blend will be done at least. - */ - gs_read_quad_f_swz( generic->cbuf_surface, - quad->x0, - quad->y0, - tmp ); - - mask_copy( tmp, quad->outputs.color, quad->mask ); - - gs_write_quad_f_swz( generic->cbuf_surface, - quad->x0, - quad->y0, - tmp ); - } - else - { - gs_write_quad_f_swz( generic->cbuf_surface, - quad->x0, - quad->y0, - quad->outputs.color ); - } -} diff --git a/src/mesa/sources b/src/mesa/sources index eda3ef7beb..af8d678878 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -155,36 +155,38 @@ VF_SOURCES = \ vf/vf_sse.c SOFTPIPE_SOURCES = \ - softpipe/generic/g_context.c \ - softpipe/generic/g_draw.c \ - softpipe/generic/g_prim_clip.c \ - softpipe/generic/g_prim_cull.c \ - softpipe/generic/g_prim_flatshade.c \ - softpipe/generic/g_prim_offset.c \ - softpipe/generic/g_prim_setup.c \ - softpipe/generic/g_prim_twoside.c \ - softpipe/generic/g_prim_unfilled.c \ - softpipe/generic/g_state_clip.c \ - softpipe/generic/g_state_derived.c \ - softpipe/generic/g_state_fs.c \ - softpipe/generic/g_state_setup.c \ - softpipe/generic/g_state_surface.c \ - softpipe/generic/g_surface.c \ - softpipe/generic/g_tile_fs.c \ - softpipe/generic/g_tile_output.c \ - softpipe/state_tracker/st_atom.c \ - softpipe/state_tracker/st_atom_blend.c \ - softpipe/state_tracker/st_atom_cbuf.c \ - softpipe/state_tracker/st_atom_clip.c \ - softpipe/state_tracker/st_atom_depth.c \ - softpipe/state_tracker/st_atom_fs.c \ - softpipe/state_tracker/st_atom_scissor.c \ - softpipe/state_tracker/st_atom_stencil.c \ - softpipe/state_tracker/st_atom_setup.c \ - softpipe/state_tracker/st_atom_viewport.c \ - softpipe/state_tracker/st_cb_program.c \ - softpipe/state_tracker/st_draw.c \ - softpipe/state_tracker/st_context.c + pipe/softpipe/sp_context.c \ + pipe/softpipe/sp_draw.c \ + pipe/softpipe/sp_prim_clip.c \ + pipe/softpipe/sp_prim_cull.c \ + pipe/softpipe/sp_prim_flatshade.c \ + pipe/softpipe/sp_prim_offset.c \ + pipe/softpipe/sp_prim_setup.c \ + pipe/softpipe/sp_prim_twoside.c \ + pipe/softpipe/sp_prim_unfilled.c \ + pipe/softpipe/sp_state_clip.c \ + pipe/softpipe/sp_state_derived.c \ + pipe/softpipe/sp_state_fs.c \ + pipe/softpipe/sp_state_setup.c \ + pipe/softpipe/sp_state_surface.c \ + pipe/softpipe/sp_surface.c \ + pipe/softpipe/sp_tile_fs.c \ + pipe/softpipe/sp_tile_output.c + +STATETRACKER_SOURCES = \ + state_tracker/st_atom.c \ + state_tracker/st_atom_blend.c \ + state_tracker/st_atom_cbuf.c \ + state_tracker/st_atom_clip.c \ + state_tracker/st_atom_depth.c \ + state_tracker/st_atom_fs.c \ + state_tracker/st_atom_scissor.c \ + state_tracker/st_atom_stencil.c \ + state_tracker/st_atom_setup.c \ + state_tracker/st_atom_viewport.c \ + state_tracker/st_cb_program.c \ + state_tracker/st_draw.c \ + state_tracker/st_context.c @@ -334,6 +336,7 @@ SOLO_SOURCES = \ $(VBO_SOURCES) \ $(VF_SOURCES) \ $(SOFTPIPE_SOURCES) \ + $(STATETRACKER_SOURCES) \ $(TNL_SOURCES) \ $(SHADER_SOURCES) \ $(SWRAST_SOURCES) \ diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index eabb9172ba..03fbd290e0 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -34,12 +34,12 @@ #include "st_context.h" #include "st_atom.h" -#include "softpipe/sp_context.h" -#include "softpipe/sp_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" /** - * Convert GLenum blend tokens to softpipe tokens. + * Convert GLenum blend tokens to pipe tokens. * Both blend factors and blend funcs are accepted. */ static GLuint @@ -48,54 +48,54 @@ gl_blend_to_sp(GLenum blend) switch (blend) { /* blend functions */ case GL_FUNC_ADD: - return SP_BLEND_ADD; + return PIPE_BLEND_ADD; case GL_FUNC_SUBTRACT: - return SP_BLEND_SUBTRACT; + return PIPE_BLEND_SUBTRACT; case GL_FUNC_REVERSE_SUBTRACT: - return SP_BLEND_REVERSE_SUBTRACT; + return PIPE_BLEND_REVERSE_SUBTRACT; case GL_MIN: - return SP_BLEND_MIN; + return PIPE_BLEND_MIN; case GL_MAX: - return SP_BLEND_MAX; + return PIPE_BLEND_MAX; /* blend factors */ case GL_ONE: - return SP_BLENDFACTOR_ONE; + return PIPE_BLENDFACTOR_ONE; case GL_SRC_COLOR: - return SP_BLENDFACTOR_SRC_COLOR; + return PIPE_BLENDFACTOR_SRC_COLOR; case GL_SRC_ALPHA: - return SP_BLENDFACTOR_SRC_ALPHA; + return PIPE_BLENDFACTOR_SRC_ALPHA; case GL_DST_ALPHA: - return SP_BLENDFACTOR_DST_ALPHA; + return PIPE_BLENDFACTOR_DST_ALPHA; case GL_DST_COLOR: - return SP_BLENDFACTOR_DST_COLOR; + return PIPE_BLENDFACTOR_DST_COLOR; case GL_SRC_ALPHA_SATURATE: - return SP_BLENDFACTOR_SRC_ALPHA_SATURATE; + return PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE; case GL_CONSTANT_COLOR: - return SP_BLENDFACTOR_CONST_COLOR; + return PIPE_BLENDFACTOR_CONST_COLOR; case GL_CONSTANT_ALPHA: - return SP_BLENDFACTOR_CONST_ALPHA; + return PIPE_BLENDFACTOR_CONST_ALPHA; /* - return SP_BLENDFACTOR_SRC1_COLOR; - return SP_BLENDFACTOR_SRC1_ALPHA; + return PIPE_BLENDFACTOR_SRC1_COLOR; + return PIPE_BLENDFACTOR_SRC1_ALPHA; */ case GL_ZERO: - return SP_BLENDFACTOR_ZERO; + return PIPE_BLENDFACTOR_ZERO; case GL_ONE_MINUS_SRC_COLOR: - return SP_BLENDFACTOR_INV_SRC_COLOR; + return PIPE_BLENDFACTOR_INV_SRC_COLOR; case GL_ONE_MINUS_SRC_ALPHA: - return SP_BLENDFACTOR_INV_SRC_ALPHA; + return PIPE_BLENDFACTOR_INV_SRC_ALPHA; case GL_ONE_MINUS_DST_COLOR: - return SP_BLENDFACTOR_INV_DST_ALPHA; + return PIPE_BLENDFACTOR_INV_DST_ALPHA; case GL_ONE_MINUS_DST_ALPHA: - return SP_BLENDFACTOR_INV_DST_COLOR; + return PIPE_BLENDFACTOR_INV_DST_COLOR; case GL_ONE_MINUS_CONSTANT_COLOR: - return SP_BLENDFACTOR_INV_CONST_COLOR; + return PIPE_BLENDFACTOR_INV_CONST_COLOR; case GL_ONE_MINUS_CONSTANT_ALPHA: - return SP_BLENDFACTOR_INV_CONST_ALPHA; + return PIPE_BLENDFACTOR_INV_CONST_ALPHA; /* - return SP_BLENDFACTOR_INV_SRC1_COLOR; - return SP_BLENDFACTOR_INV_SRC1_ALPHA; + return PIPE_BLENDFACTOR_INV_SRC1_COLOR; + return PIPE_BLENDFACTOR_INV_SRC1_ALPHA; */ default: assert("invalid GL token in gl_blend_to_sp()" == NULL); @@ -105,44 +105,44 @@ gl_blend_to_sp(GLenum blend) /** - * Convert GLenum logicop tokens to softpipe tokens. + * Convert GLenum logicop tokens to pipe tokens. */ static GLuint gl_logicop_to_sp(GLenum logicop) { switch (logicop) { case GL_CLEAR: - return SP_LOGICOP_CLEAR; + return PIPE_LOGICOP_CLEAR; case GL_NOR: - return SP_LOGICOP_NOR; + return PIPE_LOGICOP_NOR; case GL_AND_INVERTED: - return SP_LOGICOP_AND_INVERTED; + return PIPE_LOGICOP_AND_INVERTED; case GL_COPY_INVERTED: - return SP_LOGICOP_COPY_INVERTED; + return PIPE_LOGICOP_COPY_INVERTED; case GL_AND_REVERSE: - return SP_LOGICOP_AND_REVERSE; + return PIPE_LOGICOP_AND_REVERSE; case GL_INVERT: - return SP_LOGICOP_INVERT; + return PIPE_LOGICOP_INVERT; case GL_XOR: - return SP_LOGICOP_XOR; + return PIPE_LOGICOP_XOR; case GL_NAND: - return SP_LOGICOP_NAND; + return PIPE_LOGICOP_NAND; case GL_AND: - return SP_LOGICOP_AND; + return PIPE_LOGICOP_AND; case GL_EQUIV: - return SP_LOGICOP_EQUIV; + return PIPE_LOGICOP_EQUIV; case GL_NOOP: - return SP_LOGICOP_NOOP; + return PIPE_LOGICOP_NOOP; case GL_OR_INVERTED: - return SP_LOGICOP_OR_INVERTED; + return PIPE_LOGICOP_OR_INVERTED; case GL_COPY: - return SP_LOGICOP_COPY; + return PIPE_LOGICOP_COPY; case GL_OR_REVERSE: - return SP_LOGICOP_OR_REVERSE; + return PIPE_LOGICOP_OR_REVERSE; case GL_OR: - return SP_LOGICOP_OR; + return PIPE_LOGICOP_OR; case GL_SET: - return SP_LOGICOP_SET; + return PIPE_LOGICOP_SET; default: assert("invalid GL token in gl_logicop_to_sp()" == NULL); return 0; @@ -153,7 +153,7 @@ gl_logicop_to_sp(GLenum logicop) static void update_blend( struct st_context *st ) { - struct softpipe_blend_state blend; + struct pipe_blend_state blend; memset(&blend, 0, sizeof(blend)); @@ -183,7 +183,7 @@ update_blend( struct st_context *st ) if (memcmp(&blend, &st->state.blend, sizeof(blend)) != 0) { /* state has changed */ st->state.blend = blend; /* struct copy */ - st->softpipe->set_blend_state(st->softpipe, &blend); /* set new state */ + st->pipe->set_blend_state(st->pipe, &blend); /* set new state */ } } diff --git a/src/mesa/state_tracker/st_atom_cbuf.c b/src/mesa/state_tracker/st_atom_cbuf.c index cd707ec5ef..0f90aa7646 100644 --- a/src/mesa/state_tracker/st_atom_cbuf.c +++ b/src/mesa/state_tracker/st_atom_cbuf.c @@ -31,7 +31,7 @@ */ #include "st_context.h" -#include "softpipe/sp_context.h" +#include "pipe/p_context.h" #include "st_atom.h" extern GLboolean xmesa_get_cbuf_details( GLcontext *ctx, @@ -45,7 +45,7 @@ extern GLboolean xmesa_get_cbuf_details( GLcontext *ctx, */ static void update_cbuf_state( struct st_context *st ) { - struct softpipe_surface cbuf; + struct pipe_surface cbuf; GLboolean ok; ok = xmesa_get_cbuf_details( st->ctx, @@ -58,7 +58,7 @@ static void update_cbuf_state( struct st_context *st ) if (memcmp(&cbuf, &st->state.cbuf, sizeof(cbuf)) != 0) { st->state.cbuf = cbuf; - st->softpipe->set_cbuf_state( st->softpipe, &cbuf ); + st->pipe->set_cbuf_state( st->pipe, &cbuf ); } } diff --git a/src/mesa/state_tracker/st_atom_clip.c b/src/mesa/state_tracker/st_atom_clip.c index 710d6ffc84..8ccad637d5 100644 --- a/src/mesa/state_tracker/st_atom_clip.c +++ b/src/mesa/state_tracker/st_atom_clip.c @@ -32,7 +32,7 @@ #include "st_context.h" -#include "softpipe/sp_context.h" +#include "pipe/p_context.h" #include "st_atom.h" @@ -40,12 +40,12 @@ */ static void update_clip( struct st_context *st ) { - struct softpipe_clip_state clip; + struct pipe_clip_state clip; GLuint i; memset(&clip, 0, sizeof(clip)); - for (i = 0; i < SP_MAX_CLIP_PLANES; i++) { + for (i = 0; i < PIPE_MAX_CLIP_PLANES; i++) { if (st->ctx->Transform.ClipPlanesEnabled & (1 << i)) { memcpy(clip.ucp[clip.nr], st->ctx->Transform._ClipUserPlane[i], @@ -56,7 +56,7 @@ static void update_clip( struct st_context *st ) if (memcmp(&clip, &st->state.clip, sizeof(clip)) != 0) { st->state.clip = clip; - st->softpipe->set_clip_state(st->softpipe, &clip); + st->pipe->set_clip_state(st->pipe, &clip); } } diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index 5532abc8fd..4c891e6e54 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -34,25 +34,25 @@ #include "st_context.h" #include "st_atom.h" -#include "softpipe/sp_context.h" -#include "softpipe/sp_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" /** - * Convert GLenum depth func tokens to softpipe tokens. + * Convert GLenum depth func tokens to pipe tokens. */ static GLuint gl_depth_func_to_sp(GLenum func) { /* Same values, just biased */ - assert(SP_DEPTH_FUNC_NEVER == GL_NEVER - GL_NEVER); - assert(SP_DEPTH_FUNC_LESS == GL_LESS - GL_NEVER); - assert(SP_DEPTH_FUNC_EQUAL == GL_EQUAL - GL_NEVER); - assert(SP_DEPTH_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); - assert(SP_DEPTH_FUNC_GREATER == GL_GREATER - GL_NEVER); - assert(SP_DEPTH_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); - assert(SP_DEPTH_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); - assert(SP_DEPTH_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(PIPE_DEPTH_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(PIPE_DEPTH_FUNC_LESS == GL_LESS - GL_NEVER); + assert(PIPE_DEPTH_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(PIPE_DEPTH_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(PIPE_DEPTH_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(PIPE_DEPTH_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(PIPE_DEPTH_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(PIPE_DEPTH_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); assert(func >= GL_NEVER); assert(func <= GL_ALWAYS); return func - GL_NEVER; @@ -62,7 +62,7 @@ gl_depth_func_to_sp(GLenum func) static void update_depth( struct st_context *st ) { - struct softpipe_depth_state depth; + struct pipe_depth_state depth; memset(&depth, 0, sizeof(depth)); @@ -74,7 +74,7 @@ update_depth( struct st_context *st ) if (memcmp(&depth, &st->state.depth, sizeof(depth)) != 0) { /* state has changed */ st->state.depth = depth; /* struct copy */ - st->softpipe->set_depth_state(st->softpipe, &depth); /* set new state */ + st->pipe->set_depth_state(st->pipe, &depth); /* set new state */ } } diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index ca109d2d34..9c6bc1ce2a 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -30,19 +30,19 @@ */ #include "st_context.h" -#include "softpipe/sp_context.h" +#include "pipe/p_context.h" #include "st_atom.h" static void update_fs( struct st_context *st ) { - struct softpipe_fs_state fs; + struct pipe_fs_state fs; fs.fp = st->ctx->FragmentProgram._Current; if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0) { st->state.fs = fs; - st->softpipe->set_fs_state(st->softpipe, &fs); + st->pipe->set_fs_state(st->pipe, &fs); } } diff --git a/src/mesa/state_tracker/st_atom_scissor.c b/src/mesa/state_tracker/st_atom_scissor.c index 105c2a6dd9..75bead388f 100644 --- a/src/mesa/state_tracker/st_atom_scissor.c +++ b/src/mesa/state_tracker/st_atom_scissor.c @@ -32,7 +32,7 @@ #include "st_context.h" -#include "softpipe/sp_context.h" +#include "pipe/p_context.h" #include "st_atom.h" @@ -42,7 +42,7 @@ static void update_scissor( struct st_context *st ) { - struct softpipe_scissor_rect scissor; + struct pipe_scissor_rect scissor; const struct gl_framebuffer *fb = st->ctx->DrawBuffer; scissor.minx = 0; @@ -69,7 +69,7 @@ update_scissor( struct st_context *st ) if (memcmp(&scissor, &st->state.scissor, sizeof(scissor)) != 0) { /* state has changed */ st->state.scissor = scissor; /* struct copy */ - st->softpipe->set_scissor_rect(st->softpipe, &scissor); /* activate */ + st->pipe->set_scissor_rect(st->pipe, &scissor); /* activate */ } } diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index 2a582ea36c..d88d403c1d 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -32,7 +32,7 @@ #include "st_context.h" -#include "softpipe/sp_context.h" +#include "pipe/p_context.h" #include "st_atom.h" static GLuint translate_fill( GLenum mode ) @@ -60,7 +60,7 @@ static GLboolean get_offset_flag( GLuint fill_mode, static void update_setup_state( struct st_context *st ) { GLcontext *ctx = st->ctx; - struct softpipe_setup_state setup; + struct pipe_setup_state setup; memset(&setup, 0, sizeof(setup)); @@ -162,7 +162,7 @@ static void update_setup_state( struct st_context *st ) if (memcmp(&setup, &st->state.setup, sizeof(setup)) != 0) { st->state.setup = setup; - st->softpipe->set_setup_state( st->softpipe, &setup ); + st->pipe->set_setup_state( st->pipe, &setup ); } } diff --git a/src/mesa/state_tracker/st_atom_stencil.c b/src/mesa/state_tracker/st_atom_stencil.c index 32610c3d24..b9f24f49e6 100644 --- a/src/mesa/state_tracker/st_atom_stencil.c +++ b/src/mesa/state_tracker/st_atom_stencil.c @@ -34,25 +34,25 @@ #include "st_context.h" #include "st_atom.h" -#include "softpipe/sp_context.h" -#include "softpipe/sp_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" /** - * Convert GLenum stencil func tokens to softpipe tokens. + * Convert GLenum stencil func tokens to pipe tokens. */ static GLuint gl_stencil_func_to_sp(GLenum func) { /* Same values, just biased */ - assert(SP_STENCIL_FUNC_NEVER == GL_NEVER - GL_NEVER); - assert(SP_STENCIL_FUNC_LESS == GL_LESS - GL_NEVER); - assert(SP_STENCIL_FUNC_EQUAL == GL_EQUAL - GL_NEVER); - assert(SP_STENCIL_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); - assert(SP_STENCIL_FUNC_GREATER == GL_GREATER - GL_NEVER); - assert(SP_STENCIL_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); - assert(SP_STENCIL_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); - assert(SP_STENCIL_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(PIPE_STENCIL_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(PIPE_STENCIL_FUNC_LESS == GL_LESS - GL_NEVER); + assert(PIPE_STENCIL_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(PIPE_STENCIL_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(PIPE_STENCIL_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(PIPE_STENCIL_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(PIPE_STENCIL_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(PIPE_STENCIL_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); assert(func >= GL_NEVER); assert(func <= GL_ALWAYS); return func - GL_NEVER; @@ -60,28 +60,28 @@ gl_stencil_func_to_sp(GLenum func) /** - * Convert GLenum stencil op tokens to softpipe tokens. + * Convert GLenum stencil op tokens to pipe tokens. */ static GLuint gl_stencil_op_to_sp(GLenum func) { switch (func) { case GL_KEEP: - return SP_STENCIL_OP_KEEP; + return PIPE_STENCIL_OP_KEEP; case GL_ZERO: - return SP_STENCIL_OP_ZERO; + return PIPE_STENCIL_OP_ZERO; case GL_REPLACE: - return SP_STENCIL_OP_REPLACE; + return PIPE_STENCIL_OP_REPLACE; case GL_INCR: - return SP_STENCIL_OP_INCR; + return PIPE_STENCIL_OP_INCR; case GL_DECR: - return SP_STENCIL_OP_DECR; + return PIPE_STENCIL_OP_DECR; case GL_INCR_WRAP: - return SP_STENCIL_OP_INCR_WRAP; + return PIPE_STENCIL_OP_INCR_WRAP; case GL_DECR_WRAP: - return SP_STENCIL_OP_DECR_WRAP; + return PIPE_STENCIL_OP_DECR_WRAP; case GL_INVERT: - return SP_STENCIL_OP_INVERT; + return PIPE_STENCIL_OP_INVERT; default: assert("invalid GL token in gl_stencil_op_to_sp()" == NULL); return 0; @@ -92,7 +92,7 @@ gl_stencil_op_to_sp(GLenum func) static void update_stencil( struct st_context *st ) { - struct softpipe_stencil_state stencil; + struct pipe_stencil_state stencil; memset(&stencil, 0, sizeof(stencil)); @@ -121,7 +121,7 @@ update_stencil( struct st_context *st ) if (memcmp(&stencil, &st->state.stencil, sizeof(stencil)) != 0) { /* state has changed */ st->state.stencil = stencil; /* struct copy */ - st->softpipe->set_stencil_state(st->softpipe, &stencil); /* set new state */ + st->pipe->set_stencil_state(st->pipe, &stencil); /* set new state */ } } diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index af896e2e31..765bcf3647 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -29,7 +29,7 @@ #include "context.h" #include "colormac.h" #include "st_context.h" -#include "softpipe/sp_context.h" +#include "pipe/p_context.h" #include "st_atom.h" @@ -88,7 +88,7 @@ static void update_viewport( struct st_context *st ) GLfloat half_height = ctx->Viewport.Height / 2.0; GLfloat half_depth = (ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; - struct softpipe_viewport vp; + struct pipe_viewport vp; vp.scale[0] = half_width; vp.scale[1] = half_height * yScale; @@ -102,7 +102,7 @@ static void update_viewport( struct st_context *st ) if (memcmp(&vp, &st->state.viewport, sizeof(vp)) != 0) { st->state.viewport = vp; - st->softpipe->set_viewport(st->softpipe, &vp); + st->pipe->set_viewport(st->pipe, &vp); } } } diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 8a06dd88df..6308e81a61 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -31,7 +31,7 @@ #include "st_atom.h" #include "st_draw.h" #include "st_program.h" -#include "softpipe/sp_context.h" +#include "pipe/p_context.h" void st_invalidate_state(GLcontext * ctx, GLuint new_state) { @@ -43,14 +43,14 @@ void st_invalidate_state(GLcontext * ctx, GLuint new_state) struct st_context *st_create_context( GLcontext *ctx, - struct softpipe_context *softpipe ) + struct pipe_context *pipe ) { struct st_context *st = CALLOC_STRUCT( st_context ); ctx->st = st; st->ctx = ctx; - st->softpipe = softpipe; + st->pipe = pipe; st->dirty.mesa = ~0; st->dirty.st = ~0; @@ -68,7 +68,7 @@ void st_destroy_context( struct st_context *st ) st_destroy_atoms( st ); st_destroy_draw( st ); st_destroy_cb_program( st ); - st->softpipe->destroy( st->softpipe ); + st->pipe->destroy( st->pipe ); FREE( st ); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 317d377371..6d9460e2f0 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -29,7 +29,7 @@ #define ST_CONTEXT_H #include "mtypes.h" -#include "softpipe/sp_state.h" +#include "pipe/p_state.h" struct st_context; @@ -58,23 +58,23 @@ struct st_context { GLcontext *ctx; - struct softpipe_context *softpipe; + struct pipe_context *pipe; - /* Eventually will use a cache to feed the softpipe with + /* Eventually will use a cache to feed the pipe with * create/bind/delete calls to constant state objects. Not yet * though, we just shove random objects across the interface. */ struct { - struct softpipe_viewport viewport; - struct softpipe_setup_state setup; - struct softpipe_fs_state fs; - struct softpipe_blend_state blend; - struct softpipe_surface cbuf; - struct softpipe_clip_state clip; - struct softpipe_depth_state depth; - struct softpipe_scissor_rect scissor; - struct softpipe_poly_stipple poly_stipple; - struct softpipe_stencil_state stencil; + struct pipe_viewport viewport; + struct pipe_setup_state setup; + struct pipe_fs_state fs; + struct pipe_blend_state blend; + struct pipe_surface cbuf; + struct pipe_clip_state clip; + struct pipe_depth_state depth; + struct pipe_scissor_rect scissor; + struct pipe_poly_stipple poly_stipple; + struct pipe_stencil_state stencil; } state; struct { diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 12646402ee..4db8189988 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -38,7 +38,7 @@ #include "st_context.h" #include "st_atom.h" #include "st_draw.h" -#include "softpipe/sp_context.h" +#include "pipe/p_context.h" /* * TNL stage which feedsinto the above. @@ -52,13 +52,13 @@ static GLboolean draw( GLcontext * ctx, struct tnl_pipeline_stage *stage ) struct st_context *st = st_context(ctx); struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - /* Validate driver and softpipe state: + /* Validate driver and pipe state: */ st_validate_state( st ); /* Call into the new draw code to handle the VB: */ - st->softpipe->draw_vb( st->softpipe, VB ); + st->pipe->draw_vb( st->pipe, VB ); /* Finished */ @@ -84,7 +84,7 @@ static const struct tnl_pipeline_stage *intel_pipeline[] = { &_tnl_texture_transform_stage, &_tnl_point_attenuation_stage, &_tnl_vertex_program_stage, - &st_draw, /* ADD: escape to softpipe */ + &st_draw, /* ADD: escape to pipe */ 0, }; diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 8241dbd52d..3191549a2f 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -31,10 +31,10 @@ #include "mtypes.h" struct st_context; -struct softpipe_context; +struct pipe_context; struct st_context *st_create_context( GLcontext *ctx, - struct softpipe_context *softpipe); + struct pipe_context *pipe); void st_destroy_context( struct st_context *st ); -- cgit v1.2.3 From fe1f94f26c5cb43934513a38f60bf00aa542e64c Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 15 Jun 2007 09:51:27 -0600 Subject: framebuffer state --- src/mesa/state_tracker/st_atom_framebuffer.c | 78 ++++++++++++++++++++++++++++ 1 file changed, 78 insertions(+) create mode 100644 src/mesa/state_tracker/st_atom_framebuffer.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c new file mode 100644 index 0000000000..849616f81e --- /dev/null +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -0,0 +1,78 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "st_context.h" +#include "st_atom.h" +#include "pipe/p_context.h" + + +static struct pipe_surface * +xmesa_get_color_surface(GLcontext *ctx, GLuint i) +{ + return NULL; +} + + +/** + * Update framebuffer state (color, depth, stencil, etc. buffers) + * XXX someday: separate draw/read buffers. + */ +static void +update_framebuffer_state( struct st_context *st ) +{ + struct pipe_framebuffer_state framebuffer; + GLuint i; + + /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state + * to determine which surfaces to draw to + */ + framebuffer.num_cbufs = st->ctx->DrawBuffer->_NumColorDrawBuffers[0]; + for (i = 0; i < framebuffer.num_cbufs; i++) { + framebuffer.cbufs[i] = xmesa_get_color_surface(st->ctx, i); + } + + if (memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0) { + st->state.framebuffer = framebuffer; + st->pipe->set_framebuffer_state( st->pipe, &framebuffer ); + } +} + + +const struct st_tracked_state st_update_framebuffer = { + .dirty = { + .mesa = _NEW_BUFFERS, + .st = 0, + }, + .update = update_framebuffer_state +}; + -- cgit v1.2.3 From 8f6b6eed35832733674372dea7bd5ce69a965c4e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 15 Jun 2007 09:51:53 -0600 Subject: initial framebuffer state --- src/mesa/pipe/p_context.h | 3 +++ src/mesa/pipe/p_state.h | 15 ++++++++++++++- src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/sources | 1 + src/mesa/state_tracker/st_context.h | 1 + 5 files changed, 20 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index dc7a80744a..5c9b9b14fe 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -59,6 +59,9 @@ struct pipe_context { void (*set_blend_state)( struct pipe_context *, const struct pipe_blend_state * ); + void (*set_framebuffer_state)( struct pipe_context *, + const struct pipe_framebuffer_state * ); + void (*set_cbuf_state)( struct pipe_context *, const struct pipe_surface * ); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 7264f253d9..dba1e11715 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -149,7 +149,9 @@ struct pipe_stencil_state { /* This will change for hardware pipes... */ -struct pipe_surface { +struct pipe_surface +{ + GLuint width, height; GLubyte *ptr; GLint stride; GLuint cpp; @@ -157,6 +159,17 @@ struct pipe_surface { }; +struct pipe_framebuffer_state +{ + GLuint num_cbufs; /**< Number of color bufs to draw to */ + struct pipe_surface *cbufs[4]; /**< OpenGL can write to as many as + 4 color buffers at once */ + struct pipe_surface *zbuf; /**< Z buffer */ + struct pipe_surface *sbuf; /**< Stencil buffer */ + struct pipe_surface *abuf; /**< Accum buffer */ +}; + + /** * Texture sampler state. */ diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index da6c17aaf6..9050b59f77 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -67,6 +67,7 @@ struct softpipe_context { /* The most recent drawing state as set by the driver: */ + struct pipe_framebuffer_state framebuffer; struct pipe_viewport viewport; struct pipe_setup_state setup; struct pipe_fs_state fs; diff --git a/src/mesa/sources b/src/mesa/sources index af8d678878..1aef2f76eb 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -180,6 +180,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_clip.c \ state_tracker/st_atom_depth.c \ state_tracker/st_atom_fs.c \ + state_tracker/st_atom_framebuffer.c \ state_tracker/st_atom_scissor.c \ state_tracker/st_atom_stencil.c \ state_tracker/st_atom_setup.c \ diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 6d9460e2f0..b43fc0925b 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -65,6 +65,7 @@ struct st_context * though, we just shove random objects across the interface. */ struct { + struct pipe_framebuffer_state framebuffer; struct pipe_viewport viewport; struct pipe_setup_state setup; struct pipe_fs_state fs; -- cgit v1.2.3 From fde755aefb3559731a734110c7172badee9f14f9 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 15 Jun 2007 10:50:48 -0600 Subject: In softpipe_set_viewport() use ordinary (struct) assignment to update softpipe->viewport. The previous memcpy() was incorrect since it only copied 4 bytes instead of 32. With struct assignment we avoid data size errors. --- src/mesa/pipe/softpipe/sp_state_clip.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_clip.c b/src/mesa/pipe/softpipe/sp_state_clip.c index 6e83b055fb..4b83f2b07e 100644 --- a/src/mesa/pipe/softpipe/sp_state_clip.c +++ b/src/mesa/pipe/softpipe/sp_state_clip.c @@ -56,7 +56,7 @@ void softpipe_set_viewport( struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); - memcpy(&softpipe->viewport, viewport, sizeof(viewport)); + softpipe->viewport = *viewport; /* struct copy */ /* Using tnl/ and vf/ modules is temporary while getting started. * Full pipe will have vertex shader, vertex fetch of its own. -- cgit v1.2.3 From 1119852d0272c5ae43b09024b61f6e53c356ee65 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 15 Jun 2007 10:54:08 -0600 Subject: framebuffer state --- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_context.h | 2 +- src/mesa/pipe/softpipe/sp_state.h | 3 +++ src/mesa/pipe/softpipe/sp_state_surface.c | 14 ++++++++++++++ 4 files changed, 19 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 8e547cbdc3..9a054265fb 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -66,6 +66,7 @@ struct pipe_context *softpipe_create( void ) struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); softpipe->pipe.destroy = softpipe_destroy; + softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_viewport = softpipe_set_viewport; softpipe->pipe.set_setup_state = softpipe_set_setup_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 9050b59f77..f2ebc4272a 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -58,7 +58,7 @@ enum interp_mode { #define G_NEW_CLIP 0x20 #define G_NEW_SCISSOR 0x40 #define G_NEW_STIPPLE 0x80 - +#define G_NEW_FRAMEBUFFER 0x100 struct softpipe_context { diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 2abefc6037..14afce157e 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -35,6 +35,9 @@ #include "pipe/p_state.h" +void softpipe_set_framebuffer_state( struct pipe_context *, + const struct pipe_framebuffer_state * ); + void softpipe_set_clip_state( struct pipe_context *, const struct pipe_clip_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index 989f2e29d8..d089fe0632 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -51,3 +51,17 @@ void softpipe_set_cbuf_state( struct pipe_context *pipe, } +/* + * XXX this might get moved someday + */ +void softpipe_set_framebuffer_state( struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->framebuffer = *fb; /* struct copy */ + + softpipe->dirty |= G_NEW_FRAMEBUFFER; +} + + -- cgit v1.2.3 From 3cdd8bfe8e5afa57001c605b42bfac1f3fbc4eb8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 15 Jun 2007 10:55:12 -0600 Subject: Added st_update_framebuffer struct/object. --- src/mesa/state_tracker/st_atom.c | 1 + src/mesa/state_tracker/st_atom.h | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index fdbfb9021d..5b3d32cce3 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -42,6 +42,7 @@ */ static const struct st_tracked_state *atoms[] = { + &st_update_framebuffer, &st_update_cbuf, &st_update_clip, &st_update_fs, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 1d8da46336..8575bfea1a 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -44,6 +44,7 @@ void st_destroy_atoms( struct st_context *st ); void st_validate_state( struct st_context *st ); +const struct st_tracked_state st_update_framebuffer; const struct st_tracked_state st_update_cbuf; const struct st_tracked_state st_update_clip; const struct st_tracked_state st_update_depth; -- cgit v1.2.3 From 97fc7685742bfd60d36001ab832f14a003152bae Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 15 Jun 2007 10:57:44 -0600 Subject: s/PRIM_H/SP_STATE_H/ --- src/mesa/pipe/softpipe/sp_state.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 14afce157e..bdf10112f6 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -28,8 +28,8 @@ /* Authors: Keith Whitwell */ -#ifndef PRIM_H -#define PRIM_H +#ifndef SP_STATE_H +#define SP_STATE_H #include "glheader.h" #include "pipe/p_state.h" -- cgit v1.2.3 From efe6c50089e945902917c8091c92ce3a6155caac Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 18 Jun 2007 17:53:09 -0600 Subject: Define just one set of PIPE_FUNC_LESS, GREATER, EQUAL, etc tokens. Used for depth test, alpha test, stencil test, shadow test, etc. --- src/mesa/pipe/p_defines.h | 30 ++++++++++++------------------ src/mesa/state_tracker/st_atom_depth.c | 16 ++++++++-------- src/mesa/state_tracker/st_atom_stencil.c | 16 ++++++++-------- 3 files changed, 28 insertions(+), 34 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 1ca717e7ec..11b70175b0 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -71,14 +71,18 @@ #define PIPE_LOGICOP_OR 14 #define PIPE_LOGICOP_SET 15 -#define PIPE_STENCIL_FUNC_NEVER 0 -#define PIPE_STENCIL_FUNC_LESS 1 -#define PIPE_STENCIL_FUNC_EQUAL 2 -#define PIPE_STENCIL_FUNC_LEQUAL 3 -#define PIPE_STENCIL_FUNC_GREATER 4 -#define PIPE_STENCIL_FUNC_NOTEQUAL 5 -#define PIPE_STENCIL_FUNC_GEQUAL 6 -#define PIPE_STENCIL_FUNC_ALWAYS 7 +/** + * Inequality functions. Used for depth test, stencil compare, alpha + * test, shadow compare, etc. + */ +#define PIPE_FUNC_NEVER 0 +#define PIPE_FUNC_LESS 1 +#define PIPE_FUNC_EQUAL 2 +#define PIPE_FUNC_LEQUAL 3 +#define PIPE_FUNC_GREATER 4 +#define PIPE_FUNC_NOTEQUAL 5 +#define PIPE_FUNC_GEQUAL 6 +#define PIPE_FUNC_ALWAYS 7 #define PIPE_STENCIL_OP_KEEP 0 #define PIPE_STENCIL_OP_ZERO 1 @@ -89,16 +93,6 @@ #define PIPE_STENCIL_OP_DECR_WRAP 6 #define PIPE_STENCIL_OP_INVERT 7 -/* Note: same as stencil funcs. Also used for shadow/depth compare */ -#define PIPE_DEPTH_FUNC_NEVER 0 -#define PIPE_DEPTH_FUNC_LESS 1 -#define PIPE_DEPTH_FUNC_EQUAL 2 -#define PIPE_DEPTH_FUNC_LEQUAL 3 -#define PIPE_DEPTH_FUNC_GREATER 4 -#define PIPE_DEPTH_FUNC_NOTEQUAL 5 -#define PIPE_DEPTH_FUNC_GEQUAL 6 -#define PIPE_DEPTH_FUNC_ALWAYS 7 - #define PIPE_TEX_WRAP_REPEAT 0 #define PIPE_TEX_WRAP_CLAMP 1 #define PIPE_TEX_WRAP_CLAMP_TO_EDGE 2 diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index 4c891e6e54..a1523e06c2 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -45,14 +45,14 @@ static GLuint gl_depth_func_to_sp(GLenum func) { /* Same values, just biased */ - assert(PIPE_DEPTH_FUNC_NEVER == GL_NEVER - GL_NEVER); - assert(PIPE_DEPTH_FUNC_LESS == GL_LESS - GL_NEVER); - assert(PIPE_DEPTH_FUNC_EQUAL == GL_EQUAL - GL_NEVER); - assert(PIPE_DEPTH_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); - assert(PIPE_DEPTH_FUNC_GREATER == GL_GREATER - GL_NEVER); - assert(PIPE_DEPTH_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); - assert(PIPE_DEPTH_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); - assert(PIPE_DEPTH_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(PIPE_FUNC_LESS == GL_LESS - GL_NEVER); + assert(PIPE_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(PIPE_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(PIPE_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(PIPE_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(PIPE_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(PIPE_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); assert(func >= GL_NEVER); assert(func <= GL_ALWAYS); return func - GL_NEVER; diff --git a/src/mesa/state_tracker/st_atom_stencil.c b/src/mesa/state_tracker/st_atom_stencil.c index b9f24f49e6..d037335e9e 100644 --- a/src/mesa/state_tracker/st_atom_stencil.c +++ b/src/mesa/state_tracker/st_atom_stencil.c @@ -45,14 +45,14 @@ static GLuint gl_stencil_func_to_sp(GLenum func) { /* Same values, just biased */ - assert(PIPE_STENCIL_FUNC_NEVER == GL_NEVER - GL_NEVER); - assert(PIPE_STENCIL_FUNC_LESS == GL_LESS - GL_NEVER); - assert(PIPE_STENCIL_FUNC_EQUAL == GL_EQUAL - GL_NEVER); - assert(PIPE_STENCIL_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); - assert(PIPE_STENCIL_FUNC_GREATER == GL_GREATER - GL_NEVER); - assert(PIPE_STENCIL_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); - assert(PIPE_STENCIL_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); - assert(PIPE_STENCIL_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(PIPE_FUNC_LESS == GL_LESS - GL_NEVER); + assert(PIPE_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(PIPE_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(PIPE_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(PIPE_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(PIPE_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(PIPE_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); assert(func >= GL_NEVER); assert(func <= GL_ALWAYS); return func - GL_NEVER; -- cgit v1.2.3 From 493f7b5f3ae2eb64eaa65e24e79a7a071c359b81 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 18 Jun 2007 17:54:38 -0600 Subject: Added alpha test state. --- src/mesa/pipe/p_context.h | 9 ++- src/mesa/pipe/p_state.h | 25 ++++---- src/mesa/pipe/softpipe/sp_context.h | 2 + src/mesa/sources | 2 + src/mesa/state_tracker/st_atom_alphatest.c | 94 ++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_context.h | 1 + 6 files changed, 120 insertions(+), 13 deletions(-) create mode 100644 src/mesa/state_tracker/st_atom_alphatest.c (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 5c9b9b14fe..f7374a8380 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -56,12 +56,12 @@ struct pipe_context { /* * State functions */ + void (*set_alpha_test_state)( struct pipe_context *, + const struct pipe_alpha_test_state * ); + void (*set_blend_state)( struct pipe_context *, const struct pipe_blend_state * ); - void (*set_framebuffer_state)( struct pipe_context *, - const struct pipe_framebuffer_state * ); - void (*set_cbuf_state)( struct pipe_context *, const struct pipe_surface * ); @@ -71,6 +71,9 @@ struct pipe_context { void (*set_depth_state)( struct pipe_context *, const struct pipe_depth_state * ); + void (*set_framebuffer_state)( struct pipe_context *, + const struct pipe_framebuffer_state * ); + void (*set_fs_state)( struct pipe_context *, const struct pipe_fs_state * ); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index dba1e11715..71f1e99964 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -104,34 +104,39 @@ struct pipe_depth_state { GLuint enabled:1; /**< depth test enabled? */ GLuint writemask:1; /**< allow depth buffer writes? */ - GLuint func:3; /**< depth test func */ + GLuint func:3; /**< depth test func (PIPE_FUNC_x) */ GLfloat clear; /**< Clear value in [0,1] (XXX correct place?) */ }; +struct pipe_alpha_test_state { + GLuint enable:1; + GLuint func:3; /**< PIPE_FUNC_x */ + GLfloat ref; /**< reference value */ +}; -struct pipe_blend_state { - GLuint blend_enable:1; +struct pipe_blend_state { + GLuint blend_enable:1; - GLuint rgb_func:3; - GLuint rgb_src_factor:5; - GLuint rgb_dst_factor:5; + GLuint rgb_func:3; + GLuint rgb_src_factor:5; + GLuint rgb_dst_factor:5; GLuint alpha_func:3; GLuint alpha_src_factor:5; GLuint alpha_dst_factor:5; - GLuint logicop_enable:1; + GLuint logicop_enable:1; GLuint logicop_func:4; }; struct pipe_blend_color { - GLfloat color[4]; + GLfloat color[4]; }; struct pipe_stencil_state { GLuint front_enabled:1; - GLuint front_func:3; /**< PIPE_STENCIL_FUNC_x */ + GLuint front_func:3; /**< PIPE_FUNC_x */ GLuint front_fail_op:3; /**< PIPE_STENCIL_OP_x */ GLuint front_zpass_op:3; /**< PIPE_STENCIL_OP_x */ GLuint front_zfail_op:3; /**< PIPE_STENCIL_OP_x */ @@ -190,7 +195,7 @@ struct pipe_sampler_state GLfloat max_anisotropy; GLuint compare:1; /**< shadow/depth compare enabled? */ GLenum compare_mode:1; /**< PIPE_TEX_COMPARE_x */ - GLenum compare_func:3; /**< PIPE_DEPTH_FUNC_x */ + GLenum compare_func:3; /**< PIPE_FUNC_x */ GLfloat shadow_ambient; /**< shadow test fail color/intensity */ }; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index f2ebc4272a..efe453ea2c 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -59,6 +59,7 @@ enum interp_mode { #define G_NEW_SCISSOR 0x40 #define G_NEW_STIPPLE 0x80 #define G_NEW_FRAMEBUFFER 0x100 +#define G_NEW_ALPHA_TEST 0x200 struct softpipe_context { @@ -72,6 +73,7 @@ struct softpipe_context { struct pipe_setup_state setup; struct pipe_fs_state fs; struct pipe_blend_state blend; + struct pipe_alpha_test_state alpha_test; struct pipe_surface cbuf; struct pipe_clip_state clip; struct pipe_scissor_rect scissor; diff --git a/src/mesa/sources b/src/mesa/sources index 1aef2f76eb..d07d4dabe8 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -175,6 +175,7 @@ SOFTPIPE_SOURCES = \ STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ + state_tracker/st_atom_alphatest.c \ state_tracker/st_atom_blend.c \ state_tracker/st_atom_cbuf.c \ state_tracker/st_atom_clip.c \ @@ -291,6 +292,7 @@ X11_DRIVER_SOURCES = \ drivers/x11/xm_glide.c \ drivers/x11/xm_line.c \ drivers/x11/xm_span.c \ + drivers/x11/xm_surface.c \ drivers/x11/xm_tri.c OSMESA_DRIVER_SOURCES = \ diff --git a/src/mesa/state_tracker/st_atom_alphatest.c b/src/mesa/state_tracker/st_atom_alphatest.c new file mode 100644 index 0000000000..b7cc0d70a9 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_alphatest.c @@ -0,0 +1,94 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + + +/** + * Convert GLenum stencil func tokens to pipe tokens. + */ +static GLuint +gl_alpha_func_to_sp(GLenum func) +{ + /* Same values, just biased */ + assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(PIPE_FUNC_LESS == GL_LESS - GL_NEVER); + assert(PIPE_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(PIPE_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(PIPE_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(PIPE_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(PIPE_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(PIPE_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(func >= GL_NEVER); + assert(func <= GL_ALWAYS); + return func - GL_NEVER; +} + + +static void +update_alpha_test( struct st_context *st ) +{ + struct pipe_alpha_test_state alpha; + + memset(&alpha, 0, sizeof(alpha)); + + if (st->ctx->Color.AlphaEnabled) { + alpha.enable = 1; + alpha.func = gl_alpha_func_to_sp(st->ctx->Color.AlphaFunc); + alpha.ref = st->ctx->Color.AlphaRef; + } + + if (memcmp(&alpha, &st->state.alpha_test, sizeof(alpha)) != 0) { + /* state has changed */ + st->state.alpha_test = alpha; /* struct copy */ + st->pipe->set_alpha_test_state(st->pipe, &alpha); /* set new state */ + } +} + + +const struct st_tracked_state st_update_alpha_test = { + .dirty = { + .mesa = (_NEW_COLOR), + .st = 0, + }, + .update = update_alpha_test +}; + + + + + diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index b43fc0925b..d98dd3aada 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -69,6 +69,7 @@ struct st_context struct pipe_viewport viewport; struct pipe_setup_state setup; struct pipe_fs_state fs; + struct pipe_alpha_test_state alpha_test; struct pipe_blend_state blend; struct pipe_surface cbuf; struct pipe_clip_state clip; -- cgit v1.2.3 From 2137e3053ec87e4463fdbd952537d1bbd1a3dbcd Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Jun 2007 08:43:05 -0600 Subject: add point/line/polygon state --- src/mesa/pipe/p_defines.h | 15 +++++++++++++++ src/mesa/pipe/p_state.h | 26 ++++++++++++++++++++++++++ 2 files changed, 41 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 11b70175b0..550240e647 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -84,6 +84,21 @@ #define PIPE_FUNC_GEQUAL 6 #define PIPE_FUNC_ALWAYS 7 +/** Polygon fill mode */ +#define PIPE_POLYGON_MODE_FILL 0 +#define PIPE_POLYGON_MODE_LINE 1 +#define PIPE_POLYGON_MODE_POINT 2 + +/** Polygon cull mode */ +#define PIPE_POLYGON_CULL_NONE 0 +#define PIPE_POLYGON_CULL_FRONT 1 +#define PIPE_POLYGON_CULL_BACK 2 +#define PIPE_POLYGON_CULL_BOTH 3 + +/** Polygon front winding order */ +#define PIPE_POLYGON_FRONT_CW 0 +#define PIPE_POLYGON_FRONT_CCW 1 + #define PIPE_STENCIL_OP_KEEP 0 #define PIPE_STENCIL_OP_ZERO 1 #define PIPE_STENCIL_OP_REPLACE 2 diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 71f1e99964..c88c0de6ad 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -133,6 +133,32 @@ struct pipe_blend_color { GLfloat color[4]; }; +struct pipe_line_state +{ + GLuint smooth:1; + GLuint stipple:1; + GLushort stipple_pattern; + GLint stipple_factor; + GLfloat width; +}; + +struct pipe_point_state +{ + GLuint smooth:1; + GLfloat size; + GLfloat min_size, max_size; + GLfloat attenuation[3]; +}; + +struct pipe_polygon_state { + GLuint cull_mode:2; /**< PIPE_POLYGON_CULL_x */ + GLuint front_winding:1; /**< PIPE_POLYGON_FRONT_CCW,CW */ + GLuint front_mode:2; /**< PIPE_POLYGON_MODE_x */ + GLuint back_mode:2; /**< PIPE_POLYGON_MODE_x */ + GLuint stipple:1; /**< enable */ + GLuint smooth:1; /**< enable */ + /* XXX Polygon offset? */ +}; struct pipe_stencil_state { GLuint front_enabled:1; -- cgit v1.2.3 From 0e9838263c88d2b88abbedcd338491d08d61a6ac Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Jun 2007 08:45:34 -0600 Subject: just use regular malloc(), free(), memcpy() --- src/mesa/state_tracker/st_atom.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 5b3d32cce3..02720804d4 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -60,7 +60,7 @@ void st_init_atoms( struct st_context *st ) { GLuint i; - st->atoms = _mesa_malloc(sizeof(atoms)); + st->atoms = malloc(sizeof(atoms)); st->nr_atoms = sizeof(atoms)/sizeof(*atoms); memcpy(st->atoms, atoms, sizeof(atoms)); @@ -79,7 +79,7 @@ void st_init_atoms( struct st_context *st ) void st_destroy_atoms( struct st_context *st ) { if (st->atoms) { - _mesa_free(st->atoms); + free(st->atoms); st->atoms = NULL; } } @@ -130,7 +130,7 @@ void st_validate_state( struct st_context *st ) * state atoms are ordered correctly in the list. */ struct st_state_flags examined, prev; - _mesa_memset(&examined, 0, sizeof(examined)); + memset(&examined, 0, sizeof(examined)); prev = *state; for (i = 0; i < st->nr_atoms; i++) { -- cgit v1.2.3 From af9b5ca0359b5712509d7815a7fbc81a3255f4af Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Jun 2007 19:19:31 -0600 Subject: Re-org of surface/framebuffer state. We should be able to render to any depth/format of X window now. --- src/mesa/drivers/x11/xm_surface.c | 231 +++++++++++++++++++++++++++ src/mesa/drivers/x11/xmesaP.h | 7 + src/mesa/pipe/softpipe/sp_context.c | 1 - src/mesa/pipe/softpipe/sp_context.h | 7 - src/mesa/pipe/softpipe/sp_state.h | 3 - src/mesa/pipe/softpipe/sp_state_surface.c | 17 -- src/mesa/pipe/softpipe/sp_surface.h | 53 +----- src/mesa/pipe/softpipe/sp_tile_output.c | 45 +++--- src/mesa/sources | 4 - src/mesa/state_tracker/st_atom.c | 1 - src/mesa/state_tracker/st_atom.h | 1 - src/mesa/state_tracker/st_atom_framebuffer.c | 7 +- 12 files changed, 269 insertions(+), 108 deletions(-) create mode 100644 src/mesa/drivers/x11/xm_surface.c (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c new file mode 100644 index 0000000000..c0c56c093a --- /dev/null +++ b/src/mesa/drivers/x11/xm_surface.c @@ -0,0 +1,231 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * \file xm_surface.h + * Code to allow the softpipe code to write to X windows/buffers. + * This is a bit of a hack for now. We've basically got two different + * abstractions for color buffers: gl_renderbuffer and softpipe_surface. + * They'll need to get merged someday... + * For now, they're separate things that point to each other. + */ + + +#include "glxheader.h" +#include "GL/xmesa.h" +#include "xmesaP.h" +#include "context.h" +#include "imports.h" +#include "macros.h" +#include "framebuffer.h" +#include "renderbuffer.h" + +#include "pipe/p_state.h" +#include "pipe/softpipe/sp_context.h" +#include "pipe/softpipe/sp_surface.h" + + +/** + * An xm_surface is derived from a softpipe_surface + */ +struct xmesa_surface +{ + struct softpipe_surface sps; + struct xmesa_renderbuffer *xrb; /** ptr back to matching xmesa_renderbuffer */ +}; + + +/** + * Cast wrapper + */ +static INLINE struct xmesa_surface * +xmesa_surface(struct softpipe_surface *sps) +{ + return (struct xmesa_surface *) sps; +} + + +/** + * quad reading/writing + * These functions are just wrappers around the existing renderbuffer + * functions. + */ + +static void +read_quad_f(struct softpipe_surface *gs, GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(gs); + struct xmesa_renderbuffer *xrb = xmsurf->xrb; + GLubyte temp[16]; + GLfloat *dst = (GLfloat *) rgba; + GLuint i; + GET_CURRENT_CONTEXT(ctx); + xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, temp); + xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, temp + 8); + for (i = 0; i < 16; i++) { + dst[i] = UBYTE_TO_FLOAT(temp[i]); + } +} + +static void +read_quad_f_swz(struct softpipe_surface *gs, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(gs); + struct xmesa_renderbuffer *xrb = xmsurf->xrb; + GLubyte temp[16]; + GLfloat *dst = (GLfloat *) rrrr; + GLuint i, j; + GET_CURRENT_CONTEXT(ctx); + xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, temp); + xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, temp + 8); + for (i = 0; i < 4; i++) { + for (j = 0; j < 4; j++) { + dst[j * 4 + i] = UBYTE_TO_FLOAT(temp[i * 4 + j]); + } + } +} + +static void +write_quad_f(struct softpipe_surface *gs, GLint x, GLint y, + GLfloat (*rgba)[NUM_CHANNELS]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(gs); + struct xmesa_renderbuffer *xrb = xmsurf->xrb; + GLubyte temp[16]; + const GLfloat *src = (const GLfloat *) rgba; + GLuint i; + GET_CURRENT_CONTEXT(ctx); + for (i = 0; i < 16; i++) { + temp[i] = FLOAT_TO_UBYTE(src[i]); + } + xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y, temp, NULL); + xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y + 1, temp + 8, NULL); +} + +static void +write_quad_f_swz(struct softpipe_surface *gs, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(gs); + struct xmesa_renderbuffer *xrb = xmsurf->xrb; + GLubyte temp[16]; + const GLfloat *src = (const GLfloat *) rrrr; + GLuint i, j; + GET_CURRENT_CONTEXT(ctx); + for (i = 0; i < 4; i++) { + for (j = 0; j < 4; j++) { + temp[j * 4 + i] = FLOAT_TO_UBYTE(src[i * 4 + j]); + } + } + xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y, temp, NULL); + xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y + 1, temp + 8, NULL); +} + +static void +read_quad_ub(struct softpipe_surface *gs, GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(gs); + struct xmesa_renderbuffer *xrb = xmsurf->xrb; + GET_CURRENT_CONTEXT(ctx); + xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, rgba); + xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, rgba + 2); +} + +static void +write_quad_ub(struct softpipe_surface *gs, GLint x, GLint y, + GLubyte (*rgba)[NUM_CHANNELS]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(gs); + struct xmesa_renderbuffer *xrb = xmsurf->xrb; + GET_CURRENT_CONTEXT(ctx); + xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, rgba); + xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, rgba + 2); +} + + +static struct xmesa_surface * +create_surface(XMesaContext xmctx, struct xmesa_renderbuffer *xrb) +{ + struct xmesa_surface *xmsurf; + + xmsurf = CALLOC_STRUCT(xmesa_surface); + if (xmsurf) { + xmsurf->xrb = xrb; + xmsurf->sps.surface.width = xrb->Base.Width; + xmsurf->sps.surface.height = xrb->Base.Height; + + xmsurf->sps.read_quad_f = read_quad_f; + xmsurf->sps.read_quad_f_swz = read_quad_f_swz; + xmsurf->sps.read_quad_ub = read_quad_ub; + xmsurf->sps.write_quad_f = write_quad_f; + xmsurf->sps.write_quad_f_swz = write_quad_f_swz; + xmsurf->sps.write_quad_ub = write_quad_ub; +#if 0 + if (xrb->ximage) { + xmsurf->sps.surface.ptr = (GLubyte *) xrb->ximage->data; + xmsurf->sps.surface.stride = xrb->ximage->bytes_per_line; + xmsurf->sps.surface.cpp = xrb->ximage->depth; + + } +#endif + } + return xmsurf; +} + + +static void +free_surface(struct softpipe_surface *sps) +{ + /* XXX may need to do more in the future */ + free(sps); +} + + +/** + * Return generic surface pointer corresponding to the current color buffer. + */ +struct pipe_surface * +xmesa_get_color_surface(GLcontext *ctx, GLuint buf) +{ + XMesaContext xmctx = XMESA_CONTEXT(ctx); + struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[0][buf]; + struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); + struct softpipe_surface *sps = (struct softpipe_surface *) xrb->pSurface; + + if (!sps) { + xrb->pSurface = create_surface(xmctx, xrb); + } + else if (sps->surface.width != rb->Width || + sps->surface.height != rb->Height) { + free_surface(sps); + xrb->pSurface = create_surface(xmctx, xrb); + } + + return (struct pipe_surface *) xrb->pSurface; +} + diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 8f0bd34f47..c36b0966d9 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -196,6 +196,8 @@ struct xmesa_renderbuffer GLint bottom; /* used for FLIP macro, equals height - 1 */ ClearFunc clearFunc; + + void *pSurface; /** pipe surface */ }; @@ -585,4 +587,9 @@ GLboolean xmesa_get_cbuf_details( GLcontext *ctx, GLint *stride, GLuint *format ); + +struct pipe_surface; +struct pipe_surface * +xmesa_get_color_surface(GLcontext *ctx, GLuint buf); + #endif diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 9a054265fb..018f67302d 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -73,7 +73,6 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.set_scissor_rect = softpipe_set_scissor_rect; softpipe->pipe.set_fs_state = softpipe_set_fs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; - softpipe->pipe.set_cbuf_state = softpipe_set_cbuf_state; softpipe->pipe.draw_vb = softpipe_draw_vb; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index efe453ea2c..a873301368 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -54,7 +54,6 @@ enum interp_mode { #define G_NEW_SETUP 0x2 #define G_NEW_FS 0x4 #define G_NEW_BLEND 0x8 -#define G_NEW_CBUF 0x10 #define G_NEW_CLIP 0x20 #define G_NEW_SCISSOR 0x40 #define G_NEW_STIPPLE 0x80 @@ -74,17 +73,11 @@ struct softpipe_context { struct pipe_fs_state fs; struct pipe_blend_state blend; struct pipe_alpha_test_state alpha_test; - struct pipe_surface cbuf; struct pipe_clip_state clip; struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; GLuint dirty; - - /* Cbuf derived state??? - */ - struct softpipe_surface *cbuf_surface; - /* Clip derived state: */ GLfloat plane[12][4]; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index bdf10112f6..5002ce88fc 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -56,9 +56,6 @@ void softpipe_set_fs_state( struct pipe_context *, void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); -void softpipe_set_cbuf_state( struct pipe_context *, - const struct pipe_surface * ); - void softpipe_update_derived( struct softpipe_context *softpipe ); #endif diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index d089fe0632..99332fdd52 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -34,23 +34,6 @@ #include "sp_surface.h" -/* This is all a total hack. - */ -void softpipe_set_cbuf_state( struct pipe_context *pipe, - const struct pipe_surface *surface ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - if (softpipe->cbuf_surface == NULL) { - softpipe->cbuf_surface = CALLOC_STRUCT(softpipe_surface); - softpipe->cbuf_surface->type = &gs_rgba8; - } - - softpipe->cbuf_surface->surface = *surface; - softpipe->dirty |= G_NEW_CBUF; -} - - /* * XXX this might get moved someday */ diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 08b6889257..dde6b90448 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -38,12 +38,16 @@ struct softpipe_surface; #define G_SURFACE_RGBA_8888 0x1 -/* Internal structs and helpers for the primitive clip/setup pipeline: + +/** + * Softpipe surface is derived from pipe_surface. */ -struct softpipe_surface_type { - - GLuint format; +struct softpipe_surface { + struct pipe_surface surface; + /** + * Functions for read/writing surface data + */ void (*read_quad_f)( struct softpipe_surface *, GLint x, GLint y, GLfloat (*rgba)[NUM_CHANNELS] ); @@ -69,47 +73,6 @@ struct softpipe_surface_type { void (*write_quad_ub)( struct softpipe_surface *, GLint x, GLint y, GLubyte (*rgba)[NUM_CHANNELS] ); - - }; - -struct softpipe_surface { - struct softpipe_surface_type *type; - struct pipe_surface surface; -}; - - -static INLINE void gs_read_quad_f( struct softpipe_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - gs->type->read_quad_f(gs, x, y, rgba); -} - -static INLINE void gs_read_quad_f_swz( struct softpipe_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - gs->type->read_quad_f_swz(gs, x, y, rrrr); -} - -static INLINE void gs_write_quad_f( struct softpipe_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - gs->type->write_quad_f(gs, x, y, rgba); -} - -static INLINE void gs_write_quad_f_swz( struct softpipe_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - gs->type->write_quad_f_swz(gs, x, y, rrrr); -} - -/* Like this, or hidden? - */ -struct softpipe_surface_type gs_rgba8; - #endif diff --git a/src/mesa/pipe/softpipe/sp_tile_output.c b/src/mesa/pipe/softpipe/sp_tile_output.c index b1eb9e8c9f..d4add4b162 100644 --- a/src/mesa/pipe/softpipe/sp_tile_output.c +++ b/src/mesa/pipe/softpipe/sp_tile_output.c @@ -55,38 +55,35 @@ static void mask_copy( GLfloat (*dest)[4], } -/* Write to the output, taking mask into account. +/** + * Write quad to framebuffer, taking mask into account. * * Note that surfaces support only full quad reads and writes. */ void quad_output( struct softpipe_context *softpipe, struct quad_header *quad ) { + GLuint i; - if (quad->mask != MASK_ALL) - { - GLfloat tmp[4][QUAD_SIZE]; + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + struct softpipe_surface *sps + = (struct softpipe_surface *) softpipe->framebuffer.cbufs[i]; - /* Yes, we'll probably have a masked write as well, but this is - * how blend will be done at least. - */ - gs_read_quad_f_swz( softpipe->cbuf_surface, - quad->x0, - quad->y0, - tmp ); - - mask_copy( tmp, quad->outputs.color, quad->mask ); + if (quad->mask != MASK_ALL) { + GLfloat tmp[4][QUAD_SIZE]; - gs_write_quad_f_swz( softpipe->cbuf_surface, - quad->x0, - quad->y0, - tmp ); - } - else - { - gs_write_quad_f_swz( softpipe->cbuf_surface, - quad->x0, - quad->y0, - quad->outputs.color ); + /* Yes, we'll probably have a masked write as well, but this is + * how blend will be done at least. + */ + + sps->read_quad_f_swz(sps, quad->x0, quad->y0, tmp); + + mask_copy( tmp, quad->outputs.color, quad->mask ); + + sps->write_quad_f_swz(sps, quad->x0, quad->y0, tmp); + } + else { + sps->write_quad_f_swz(sps, quad->x0, quad->y0, quad->outputs.color); + } } } diff --git a/src/mesa/sources b/src/mesa/sources index d07d4dabe8..eb67dd9f7b 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -169,7 +169,6 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_state_fs.c \ pipe/softpipe/sp_state_setup.c \ pipe/softpipe/sp_state_surface.c \ - pipe/softpipe/sp_surface.c \ pipe/softpipe/sp_tile_fs.c \ pipe/softpipe/sp_tile_output.c @@ -177,7 +176,6 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ state_tracker/st_atom_alphatest.c \ state_tracker/st_atom_blend.c \ - state_tracker/st_atom_cbuf.c \ state_tracker/st_atom_clip.c \ state_tracker/st_atom_depth.c \ state_tracker/st_atom_fs.c \ @@ -190,8 +188,6 @@ STATETRACKER_SOURCES = \ state_tracker/st_draw.c \ state_tracker/st_context.c - - SHADER_SOURCES = \ shader/arbprogparse.c \ shader/arbprogram.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 02720804d4..228e7889b7 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -43,7 +43,6 @@ static const struct st_tracked_state *atoms[] = { &st_update_framebuffer, - &st_update_cbuf, &st_update_clip, &st_update_fs, &st_update_setup, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 8575bfea1a..8a75c9c6d5 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -45,7 +45,6 @@ void st_validate_state( struct st_context *st ); const struct st_tracked_state st_update_framebuffer; -const struct st_tracked_state st_update_cbuf; const struct st_tracked_state st_update_clip; const struct st_tracked_state st_update_depth; const struct st_tracked_state st_update_fs; diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 849616f81e..f203e1df60 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -36,11 +36,8 @@ #include "pipe/p_context.h" -static struct pipe_surface * -xmesa_get_color_surface(GLcontext *ctx, GLuint i) -{ - return NULL; -} +extern struct pipe_surface * +xmesa_get_color_surface(GLcontext *ctx, GLuint i); /** -- cgit v1.2.3 From 6cb2d0cb71d2019bd2c941a8c042e56275b22c1c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Jun 2007 19:52:25 -0600 Subject: hook up point state --- src/mesa/pipe/p_context.h | 3 + src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_context.h | 3 + src/mesa/pipe/softpipe/sp_prim_setup.c | 131 ++++++++++++++++++-------------- src/mesa/pipe/softpipe/sp_state.h | 3 + src/mesa/pipe/softpipe/sp_state_point.c | 47 ++++++++++++ src/mesa/sources | 2 + src/mesa/state_tracker/st_atom.c | 1 + src/mesa/state_tracker/st_atom.h | 1 + src/mesa/state_tracker/st_atom_point.c | 70 +++++++++++++++++ src/mesa/state_tracker/st_context.h | 2 +- 11 files changed, 204 insertions(+), 60 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_state_point.c create mode 100644 src/mesa/state_tracker/st_atom_point.c (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index f7374a8380..3e0c61eda2 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -71,6 +71,9 @@ struct pipe_context { void (*set_depth_state)( struct pipe_context *, const struct pipe_depth_state * ); + void (*set_point_state)( struct pipe_context *, + const struct pipe_point_state * ); + void (*set_framebuffer_state)( struct pipe_context *, const struct pipe_framebuffer_state * ); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 018f67302d..e0acf4177b 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -68,6 +68,7 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.destroy = softpipe_destroy; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_clip_state = softpipe_set_clip_state; + softpipe->pipe.set_point_state = softpipe_set_point_state; softpipe->pipe.set_viewport = softpipe_set_viewport; softpipe->pipe.set_setup_state = softpipe_set_setup_state; softpipe->pipe.set_scissor_rect = softpipe_set_scissor_rect; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index a873301368..43cb6f3f0b 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -54,11 +54,13 @@ enum interp_mode { #define G_NEW_SETUP 0x2 #define G_NEW_FS 0x4 #define G_NEW_BLEND 0x8 +#define G_NEW_POINT 0x10 #define G_NEW_CLIP 0x20 #define G_NEW_SCISSOR 0x40 #define G_NEW_STIPPLE 0x80 #define G_NEW_FRAMEBUFFER 0x100 #define G_NEW_ALPHA_TEST 0x200 +#define G_NEW_DEPTH_TEST 0x400 struct softpipe_context { @@ -74,6 +76,7 @@ struct softpipe_context { struct pipe_blend_state blend; struct pipe_alpha_test_state alpha_test; struct pipe_clip_state clip; + struct pipe_point_state point; struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; GLuint dirty; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index a1addc4ee1..8ef0fcbf9c 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -769,17 +769,12 @@ static void setup_point(struct prim_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - GLfloat halfSize = 7.3; /*XXX this is a vertex attrib */ - GLfloat halfSizeSquared = halfSize * halfSize; + /*XXX this should be a vertex attrib! */ + GLfloat halfSize = 0.5 * setup->stage.softpipe->point.size; + GLboolean round = setup->stage.softpipe->point.smooth; const struct vertex_header *v0 = prim->v[0]; const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; - const GLint ixmin = block((GLint) (x - halfSize)); - const GLint ixmax = block((GLint) (x + halfSize)); - const GLint iymin = block((GLint) (y - halfSize)); - const GLint iymax = block((GLint) (y + halfSize)); - GLboolean round = GL_TRUE; - GLint ix, iy; GLuint slot, j; /* For points, all interpolants are constant-valued. @@ -808,57 +803,75 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) /* XXX need to clip against scissor bounds too */ - for (iy = iymin; iy <= iymax; iy += 2) { - for (ix = ixmin; ix <= ixmax; ix += 2) { - - if (round) { - /* rounded points */ - /* XXX for GL_SMOOTH, need to compute per-fragment coverage too */ - GLfloat dx, dy; - - setup->quad.mask = 0x0; - - dx = (ix + 0.5) - x; - dy = (iy + 0.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_BOTTOM_LEFT; - - dx = (ix + 1.5) - x; - dy = (iy + 0.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_BOTTOM_RIGHT; - - dx = (ix + 0.5) - x; - dy = (iy + 1.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_TOP_LEFT; - - dx = (ix + 1.5) - x; - dy = (iy + 1.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) - setup->quad.mask |= MASK_TOP_RIGHT; - } - else { - /* square points */ - setup->quad.mask = 0xf; - - if (ix + 0.5 < x - halfSize) - setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - - if (ix + 1.5 > x + halfSize) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - - if (iy + 0.5 < y - halfSize) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); - - if (iy + 1.5 > y + halfSize) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - } - - if (setup->quad.mask) { - setup->quad.x0 = ix; - setup->quad.y0 = iy; - quad_shade( setup->stage.softpipe, &setup->quad ); + if (halfSize <= 0.5 && !round) { + /* special case for 1-pixel points */ + const GLint ix = ((GLint) x) & 1; + const GLint iy = ((GLint) y) & 1; + setup->quad.x0 = x - ix; + setup->quad.y0 = y - iy; + setup->quad.mask = (1 << ix) << (2 * iy); + quad_shade(setup->stage.softpipe, &setup->quad); + } + else { + const GLint ixmin = block((GLint) (x - halfSize)); + const GLint ixmax = block((GLint) (x + halfSize)); + const GLint iymin = block((GLint) (y - halfSize)); + const GLint iymax = block((GLint) (y + halfSize)); + GLfloat halfSizeSquared = halfSize * halfSize; + GLint ix, iy; + + for (iy = iymin; iy <= iymax; iy += 2) { + for (ix = ixmin; ix <= ixmax; ix += 2) { + + if (round) { + /* rounded points */ + /* XXX for GL_SMOOTH, need to compute per-fragment coverage too */ + GLfloat dx, dy; + + setup->quad.mask = 0x0; + + dx = (ix + 0.5) - x; + dy = (iy + 0.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_BOTTOM_LEFT; + + dx = (ix + 1.5) - x; + dy = (iy + 0.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_BOTTOM_RIGHT; + + dx = (ix + 0.5) - x; + dy = (iy + 1.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_TOP_LEFT; + + dx = (ix + 1.5) - x; + dy = (iy + 1.5) - y; + if (dx * dx + dy * dy <= halfSizeSquared) + setup->quad.mask |= MASK_TOP_RIGHT; + } + else { + /* square points */ + setup->quad.mask = 0xf; + + if (ix + 0.5 < x - halfSize) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + + if (ix + 1.5 > x + halfSize) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + + if (iy + 0.5 < y - halfSize) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + + if (iy + 1.5 > y + halfSize) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + } + + if (setup->quad.mask) { + setup->quad.x0 = ix; + setup->quad.y0 = iy; + quad_shade( setup->stage.softpipe, &setup->quad ); + } } } } diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 5002ce88fc..4d18e80b38 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -53,6 +53,9 @@ void softpipe_set_scissor_rect( struct pipe_context *, void softpipe_set_fs_state( struct pipe_context *, const struct pipe_fs_state * ); +void softpipe_set_point_state( struct pipe_context *, + const struct pipe_point_state * ); + void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/pipe/softpipe/sp_state_point.c b/src/mesa/pipe/softpipe/sp_state_point.c new file mode 100644 index 0000000000..566b33e696 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_point.c @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_draw.h" + + + +void softpipe_set_point_state( struct pipe_context *pipe, + const struct pipe_point_state *point ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->point = *point; + + softpipe->dirty |= G_NEW_POINT; +} + diff --git a/src/mesa/sources b/src/mesa/sources index eb67dd9f7b..0bf7dbceb3 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -167,6 +167,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_state_clip.c \ pipe/softpipe/sp_state_derived.c \ pipe/softpipe/sp_state_fs.c \ + pipe/softpipe/sp_state_point.c \ pipe/softpipe/sp_state_setup.c \ pipe/softpipe/sp_state_surface.c \ pipe/softpipe/sp_tile_fs.c \ @@ -180,6 +181,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_depth.c \ state_tracker/st_atom_fs.c \ state_tracker/st_atom_framebuffer.c \ + state_tracker/st_atom_point.c \ state_tracker/st_atom_scissor.c \ state_tracker/st_atom_stencil.c \ state_tracker/st_atom_setup.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 228e7889b7..8ac435041d 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -45,6 +45,7 @@ static const struct st_tracked_state *atoms[] = &st_update_framebuffer, &st_update_clip, &st_update_fs, + &st_update_point, &st_update_setup, &st_update_viewport, &st_update_scissor, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 8a75c9c6d5..56c33f3076 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -48,6 +48,7 @@ const struct st_tracked_state st_update_framebuffer; const struct st_tracked_state st_update_clip; const struct st_tracked_state st_update_depth; const struct st_tracked_state st_update_fs; +const struct st_tracked_state st_update_point; const struct st_tracked_state st_update_setup; const struct st_tracked_state st_update_viewport; const struct st_tracked_state st_update_constants; diff --git a/src/mesa/state_tracker/st_atom_point.c b/src/mesa/state_tracker/st_atom_point.c new file mode 100644 index 0000000000..7299142932 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_point.c @@ -0,0 +1,70 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + + +static void +update_point( struct st_context *st ) +{ + struct pipe_point_state point; + + memset(&point, 0, sizeof(point)); + + point.smooth = st->ctx->Point.SmoothFlag; + point.size = st->ctx->Point.Size; + point.min_size = st->ctx->Point.MinSize; + point.max_size = st->ctx->Point.MaxSize; + point.attenuation[0] = st->ctx->Point.Params[0]; + point.attenuation[1] = st->ctx->Point.Params[1]; + point.attenuation[2] = st->ctx->Point.Params[2]; + + if (memcmp(&point, &st->state.point, sizeof(point)) != 0) { + /* state has changed */ + st->state.point = point; /* struct copy */ + st->pipe->set_point_state(st->pipe, &point); /* set new state */ + } +} + + +const struct st_tracked_state st_update_point = { + .dirty = { + .mesa = (_NEW_POINT), + .st = 0, + }, + .update = update_point +}; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index d98dd3aada..94388f1aee 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -71,9 +71,9 @@ struct st_context struct pipe_fs_state fs; struct pipe_alpha_test_state alpha_test; struct pipe_blend_state blend; - struct pipe_surface cbuf; struct pipe_clip_state clip; struct pipe_depth_state depth; + struct pipe_point_state point; struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; struct pipe_stencil_state stencil; -- cgit v1.2.3 From 73f96c51052bf5233191d852ef463462306bf1d5 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 08:47:09 -0600 Subject: Initial work for glClear(), clear color state. --- src/mesa/pipe/p_context.h | 10 +++-- src/mesa/pipe/p_state.h | 5 +++ src/mesa/pipe/softpipe/sp_clear.c | 47 +++++++++++++++++++++ src/mesa/pipe/softpipe/sp_clear.h | 43 +++++++++++++++++++ src/mesa/pipe/softpipe/sp_context.c | 4 +- src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_state.h | 3 ++ src/mesa/pipe/softpipe/sp_state_surface.c | 15 ++++++- src/mesa/sources | 2 + src/mesa/state_tracker/st_atom.c | 1 + src/mesa/state_tracker/st_atom.h | 1 + src/mesa/state_tracker/st_atom_clear_color.c | 62 ++++++++++++++++++++++++++++ src/mesa/state_tracker/st_context.h | 1 + 13 files changed, 189 insertions(+), 6 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_clear.c create mode 100644 src/mesa/pipe/softpipe/sp_clear.h create mode 100644 src/mesa/state_tracker/st_atom_clear_color.c (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 3e0c61eda2..32444343be 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -53,6 +53,10 @@ struct pipe_context { void (*draw_vb)( struct pipe_context *pipe, struct vertex_buffer *VB ); + /** Clear framebuffer */ + void (*clear)(struct pipe_context *pipe, GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum); + /* * State functions */ @@ -62,12 +66,12 @@ struct pipe_context { void (*set_blend_state)( struct pipe_context *, const struct pipe_blend_state * ); - void (*set_cbuf_state)( struct pipe_context *, - const struct pipe_surface * ); - void (*set_clip_state)( struct pipe_context *, const struct pipe_clip_state * ); + void (*set_clear_color_state)( struct pipe_context *, + const struct pipe_clear_color_state * ); + void (*set_depth_state)( struct pipe_context *, const struct pipe_depth_state * ); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index c88c0de6ad..c4bf0d2195 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -133,6 +133,11 @@ struct pipe_blend_color { GLfloat color[4]; }; +struct pipe_clear_color_state +{ + GLfloat color[4]; +}; + struct pipe_line_state { GLuint smooth:1; diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c new file mode 100644 index 0000000000..c7fbca4229 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + */ + + +#include "sp_clear.h" + + +void +softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum) +{ + /* validate state (scissor)? */ + + if (color) { + } + if (depth) { + } + +} diff --git a/src/mesa/pipe/softpipe/sp_clear.h b/src/mesa/pipe/softpipe/sp_clear.h new file mode 100644 index 0000000000..f9db99dd32 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_clear.h @@ -0,0 +1,43 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + */ + +#ifndef SP_CLEAR_H +#define SP_CLEAR_H + +#include "pipe/p_state.h" +struct pipe_context; + +extern void +softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum); + + +#endif /* SP_CLEAR_H */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index e0acf4177b..7ab65162bd 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -36,6 +36,7 @@ #include "vf/vf.h" #include "sp_context.h" +#include "sp_clear.h" #include "sp_prim.h" #include "sp_state.h" #include "sp_draw.h" @@ -68,6 +69,7 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.destroy = softpipe_destroy; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_clip_state = softpipe_set_clip_state; + softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_point_state = softpipe_set_point_state; softpipe->pipe.set_viewport = softpipe_set_viewport; softpipe->pipe.set_setup_state = softpipe_set_setup_state; @@ -75,7 +77,7 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.set_fs_state = softpipe_set_fs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.draw_vb = softpipe_draw_vb; - + softpipe->pipe.clear = softpipe_clear; softpipe->prim.setup = prim_setup( softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 43cb6f3f0b..0a183ea385 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -76,6 +76,7 @@ struct softpipe_context { struct pipe_blend_state blend; struct pipe_alpha_test_state alpha_test; struct pipe_clip_state clip; + struct pipe_clear_color_state clear_color; struct pipe_point_state point; struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 4d18e80b38..bc5a3512ae 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -38,6 +38,9 @@ void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); +void softpipe_set_clear_color_state( struct pipe_context *, + const struct pipe_clear_color_state * ); + void softpipe_set_clip_state( struct pipe_context *, const struct pipe_clip_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index 99332fdd52..4d27814e1b 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -37,8 +37,9 @@ /* * XXX this might get moved someday */ -void softpipe_set_framebuffer_state( struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb ) +void +softpipe_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -48,3 +49,13 @@ void softpipe_set_framebuffer_state( struct pipe_context *pipe, } + + +void +softpipe_set_clear_color_state(struct pipe_context *pipe, + const struct pipe_clear_color_state *clear) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->clear_color = *clear; /* struct copy */ +} diff --git a/src/mesa/sources b/src/mesa/sources index 0bf7dbceb3..97801648c4 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -155,6 +155,7 @@ VF_SOURCES = \ vf/vf_sse.c SOFTPIPE_SOURCES = \ + pipe/softpipe/sp_clear.c \ pipe/softpipe/sp_context.c \ pipe/softpipe/sp_draw.c \ pipe/softpipe/sp_prim_clip.c \ @@ -177,6 +178,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ state_tracker/st_atom_alphatest.c \ state_tracker/st_atom_blend.c \ + state_tracker/st_atom_clear_color.c \ state_tracker/st_atom_clip.c \ state_tracker/st_atom_depth.c \ state_tracker/st_atom_fs.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 8ac435041d..e1d187f2db 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -43,6 +43,7 @@ static const struct st_tracked_state *atoms[] = { &st_update_framebuffer, + &st_update_clear_color, &st_update_clip, &st_update_fs, &st_update_point, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 56c33f3076..8c1219fefd 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -46,6 +46,7 @@ void st_validate_state( struct st_context *st ); const struct st_tracked_state st_update_framebuffer; const struct st_tracked_state st_update_clip; +const struct st_tracked_state st_update_clear_color; const struct st_tracked_state st_update_depth; const struct st_tracked_state st_update_fs; const struct st_tracked_state st_update_point; diff --git a/src/mesa/state_tracker/st_atom_clear_color.c b/src/mesa/state_tracker/st_atom_clear_color.c new file mode 100644 index 0000000000..adf730cd8c --- /dev/null +++ b/src/mesa/state_tracker/st_atom_clear_color.c @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "st_context.h" +#include "st_atom.h" +#include "pipe/p_context.h" + + +static void +update_clear_color_state( struct st_context *st ) +{ + struct pipe_clear_color_state clear; + + clear.color[0] = st->ctx->Color.ClearColor[0]; + clear.color[1] = st->ctx->Color.ClearColor[1]; + clear.color[2] = st->ctx->Color.ClearColor[2]; + clear.color[3] = st->ctx->Color.ClearColor[3]; + + if (memcmp(&clear, &st->state.clear_color, sizeof(clear)) != 0) { + st->state.clear_color = clear; + st->pipe->set_clear_color_state( st->pipe, &clear ); + } +} + + +const struct st_tracked_state st_update_clear_color = { + .dirty = { + .mesa = _NEW_COLOR, + .st = 0, + }, + .update = update_clear_color_state +}; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 94388f1aee..0f1fa2327b 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -71,6 +71,7 @@ struct st_context struct pipe_fs_state fs; struct pipe_alpha_test_state alpha_test; struct pipe_blend_state blend; + struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_depth_state depth; struct pipe_point_state point; -- cgit v1.2.3 From 837f2c18c477f0695150b69b69a3a921b08943ec Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 09:41:41 -0600 Subject: implement softpipe clearing (untested) --- src/mesa/pipe/softpipe/sp_clear.c | 26 +++++++++++++++++++++++++- src/mesa/pipe/softpipe/sp_surface.h | 13 +++++++++++++ src/mesa/pipe/softpipe/sp_tile_output.c | 2 +- 3 files changed, 39 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index c7fbca4229..536f0d3924 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -31,16 +31,40 @@ #include "sp_clear.h" +#include "sp_context.h" +#include "sp_surface.h" +#include "colormac.h" void softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) { - /* validate state (scissor)? */ + struct softpipe_context *softpipe = softpipe_context(pipe); if (color) { + GLuint i; + const GLint x = softpipe->scissor.minx; + const GLint y = softpipe->scissor.miny; + const GLint w = softpipe->scissor.maxx - x; + const GLint h = softpipe->scissor.maxy - y; + GLubyte clr[4]; + + UNCLAMPED_FLOAT_TO_UBYTE(clr[0], softpipe->clear_color.color[0]); + UNCLAMPED_FLOAT_TO_UBYTE(clr[1], softpipe->clear_color.color[1]); + UNCLAMPED_FLOAT_TO_UBYTE(clr[2], softpipe->clear_color.color[2]); + UNCLAMPED_FLOAT_TO_UBYTE(clr[3], softpipe->clear_color.color[3]); + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + struct pipe_surface *ps = softpipe->framebuffer.cbufs[i]; + struct softpipe_surface *sps = softpipe_surface(ps); + GLint j; + for (j = 0; j < h; j++) { + sps->write_mono_row_ub(sps, w, x, y + j, clr); + } + } } + if (depth) { } diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index dde6b90448..fc9557dee3 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -73,6 +73,19 @@ struct softpipe_surface { void (*write_quad_ub)( struct softpipe_surface *, GLint x, GLint y, GLubyte (*rgba)[NUM_CHANNELS] ); + + void (*write_mono_row_ub)( struct softpipe_surface *, + GLuint count, GLint x, GLint y, + GLubyte rgba[NUM_CHANNELS] ); }; + +/** Cast wrapper */ +static INLINE struct softpipe_surface * +softpipe_surface(struct pipe_surface *ps) +{ + return (struct softpipe_surface *) ps; +} + + #endif diff --git a/src/mesa/pipe/softpipe/sp_tile_output.c b/src/mesa/pipe/softpipe/sp_tile_output.c index d4add4b162..23086b7020 100644 --- a/src/mesa/pipe/softpipe/sp_tile_output.c +++ b/src/mesa/pipe/softpipe/sp_tile_output.c @@ -67,7 +67,7 @@ void quad_output( struct softpipe_context *softpipe, for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { struct softpipe_surface *sps - = (struct softpipe_surface *) softpipe->framebuffer.cbufs[i]; + = softpipe_surface(softpipe->framebuffer.cbufs[i]); if (quad->mask != MASK_ALL) { GLfloat tmp[4][QUAD_SIZE]; -- cgit v1.2.3 From 83b6c46b50ab138551f3af114afbcc2adbee81d2 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 12:23:47 -0600 Subject: plug in write_mono_row_ub --- src/mesa/drivers/x11/xm_surface.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index c0c56c093a..ee06b77377 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -167,6 +167,16 @@ write_quad_ub(struct softpipe_surface *gs, GLint x, GLint y, xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, rgba + 2); } +static void +write_mono_row_ub(struct softpipe_surface *gs, GLuint count, GLint x, GLint y, + GLubyte rgba[NUM_CHANNELS]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(gs); + struct xmesa_renderbuffer *xrb = xmsurf->xrb; + GET_CURRENT_CONTEXT(ctx); + xrb->Base.PutMonoRow(ctx, &xrb->Base, count, x, y, rgba, NULL); +} + static struct xmesa_surface * create_surface(XMesaContext xmctx, struct xmesa_renderbuffer *xrb) @@ -185,6 +195,8 @@ create_surface(XMesaContext xmctx, struct xmesa_renderbuffer *xrb) xmsurf->sps.write_quad_f = write_quad_f; xmsurf->sps.write_quad_f_swz = write_quad_f_swz; xmsurf->sps.write_quad_ub = write_quad_ub; + xmsurf->sps.write_mono_row_ub = write_mono_row_ub; + #if 0 if (xrb->ximage) { xmsurf->sps.surface.ptr = (GLubyte *) xrb->ximage->data; -- cgit v1.2.3 From abf45c2a3db39fc1690e282e7f7603bc1d81f647 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 12:50:16 -0600 Subject: fix comment --- src/mesa/state_tracker/st_draw.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 4db8189988..1f6c261bc2 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -41,7 +41,7 @@ #include "pipe/p_context.h" /* - * TNL stage which feedsinto the above. + * TNL stage which feeds into the above. * * XXX: this needs to go into each driver using this code, because we * cannot make the leap from ctx->draw_context in this file. The -- cgit v1.2.3 From b7c646d1bcf4b6fa285996d1b9b660ce478190f6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 13:10:48 -0600 Subject: actually use new glClear code --- src/mesa/drivers/x11/xm_dd.c | 12 ++++++++++++ src/mesa/pipe/softpipe/sp_clear.c | 10 +++++----- src/mesa/state_tracker/st_draw.c | 14 ++++++++++++++ src/mesa/state_tracker/st_draw.h | 4 ++++ 4 files changed, 35 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index eb59d1d05e..5725414856 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -55,6 +55,8 @@ #include "pipe/softpipe/sp_context.h" #include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_draw.h" /* @@ -393,6 +395,7 @@ clear_buffers(GLcontext *ctx, GLbitfield buffers) /* we can't handle color or index masking */ if (*colorMask == 0xffffffff && ctx->Color.IndexMask == 0xffffffff) { +#if 0 if (buffers & BUFFER_BIT_FRONT_LEFT) { /* clear front color buffer */ struct gl_renderbuffer *frontRb @@ -416,6 +419,15 @@ clear_buffers(GLcontext *ctx, GLbitfield buffers) buffers &= ~BUFFER_BIT_BACK_LEFT; } } +#else + /* Clear with state-tracker/pipe interface */ + struct st_context *st = st_context(ctx); + GLboolean color = (buffers & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)) ? 1: 0; + GLboolean depth = (buffers & BUFFER_BIT_DEPTH) ? 1 : 0; + GLboolean stencil = (buffers & BUFFER_BIT_STENCIL) ? 1 : 0; + GLboolean accum = (buffers & BUFFER_BIT_ACCUM) ? 1 : 0; + st_clear(st, color, depth, stencil, accum); +#endif } } if (buffers) diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 536f0d3924..e83bc053ef 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -40,14 +40,14 @@ void softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) { - struct softpipe_context *softpipe = softpipe_context(pipe); + const struct softpipe_context *softpipe = softpipe_context(pipe); + const GLint x = softpipe->scissor.minx; + const GLint y = softpipe->scissor.miny; + const GLint w = softpipe->scissor.maxx - x; + const GLint h = softpipe->scissor.maxy - y; if (color) { GLuint i; - const GLint x = softpipe->scissor.minx; - const GLint y = softpipe->scissor.miny; - const GLint w = softpipe->scissor.maxx - x; - const GLint h = softpipe->scissor.maxy - y; GLubyte clr[4]; UNCLAMPED_FLOAT_TO_UBYTE(clr[0], softpipe->clear_color.color[0]); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 1f6c261bc2..55b98629db 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -105,3 +105,17 @@ void st_destroy_draw( struct st_context *st ) /* Nothing to do. */ } + + +/** XXX temporary here */ +void +st_clear(struct st_context *st, GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum) +{ + /* Validate driver and pipe state: + */ + st_validate_state( st ); + + st->pipe->clear(st->pipe, color, depth, stencil, accum); +} + diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h index f51059706a..7a3ba52130 100644 --- a/src/mesa/state_tracker/st_draw.h +++ b/src/mesa/state_tracker/st_draw.h @@ -37,4 +37,8 @@ void st_init_draw( struct st_context *st ); void st_destroy_draw( struct st_context *st ); +/** XXX temporary here */ +void st_clear(struct st_context *st, GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum); + #endif -- cgit v1.2.3 From 1edb5aafadc16ac0d7c604a3cd4a9c2e91d9b705 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 14:27:42 -0600 Subject: better comment --- src/mesa/state_tracker/st_draw.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 55b98629db..24abee132f 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -112,8 +112,7 @@ void st_clear(struct st_context *st, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) { - /* Validate driver and pipe state: - */ + /* This makes sure the softpipe has the latest scissor, etc values */ st_validate_state( st ); st->pipe->clear(st->pipe, color, depth, stencil, accum); -- cgit v1.2.3 From fb5cdbd078d4d44fb43d417843debe41148f3714 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 14:29:14 -0600 Subject: Initial implementation of a software pipeline for quad rasterization (fragment ops). This is very much like the clipper/setup pipeline for primitives. --- src/mesa/pipe/softpipe/sp_context.c | 6 +-- src/mesa/pipe/softpipe/sp_context.h | 15 +++++- src/mesa/pipe/softpipe/sp_prim_setup.c | 22 +++++++-- src/mesa/pipe/softpipe/sp_state_derived.c | 3 ++ src/mesa/pipe/softpipe/sp_tile.c | 24 ++++++++++ src/mesa/pipe/softpipe/sp_tile.h | 22 +++++++-- src/mesa/pipe/softpipe/sp_tile_blend.c | 70 +++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_tile_depth_test.c | 73 +++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_tile_fs.c | 19 +++++--- src/mesa/pipe/softpipe/sp_tile_output.c | 18 ++++++- src/mesa/sources | 3 ++ 11 files changed, 252 insertions(+), 23 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_tile.c create mode 100644 src/mesa/pipe/softpipe/sp_tile_blend.c create mode 100644 src/mesa/pipe/softpipe/sp_tile_depth_test.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 7ab65162bd..d1023fb782 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -32,9 +32,6 @@ #include "imports.h" #include "macros.h" -#include "tnl/t_context.h" -#include "vf/vf.h" - #include "sp_context.h" #include "sp_clear.h" #include "sp_prim.h" @@ -88,6 +85,9 @@ struct pipe_context *softpipe_create( void ) softpipe->prim.flatshade = prim_flatshade( softpipe ); softpipe->prim.cull = prim_cull( softpipe ); + softpipe->quad.blend = sp_quad_blend_stage(softpipe); + softpipe->quad.shade = sp_quad_shade_stage(softpipe); + softpipe->quad.output = sp_quad_output_stage(softpipe); softpipe->draw = draw_create( softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 0a183ea385..4e6168fe33 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -36,6 +36,7 @@ #include "pipe/p_state.h" #include "pipe/p_context.h" +#include "sp_tile.h" struct softpipe_surface; @@ -129,6 +130,18 @@ struct softpipe_context { GLuint vertex_size; } prim; + /* + * Software quad rendering pipeline + */ + struct { + struct quad_stage *shade; + struct quad_stage *depth_test; + struct quad_stage *blend; + struct quad_stage *output; + + struct quad_stage *first; /**< points to one of the above stages */ + } quad; + /* Temp kludge: */ struct draw_context *draw; @@ -144,6 +157,4 @@ softpipe_context( struct pipe_context *pipe ) } - - #endif diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 8ef0fcbf9c..29bb936540 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -36,6 +36,18 @@ #include "sp_tile.h" + +/** + * Emit/render a quad. + * This passes the quad to the first stage of per-fragment operations. + */ +static INLINE void +quad_emit(struct softpipe_context *sp, struct quad_header *quad) +{ + sp->quad.first->run(sp->quad.first, quad); +} + + /** * Triangle edge info */ @@ -121,7 +133,7 @@ static void run_shader_block( struct setup_stage *setup, setup->quad.y0 = y; setup->quad.mask = mask; - quad_shade( setup->stage.softpipe, &setup->quad ); + quad_emit(setup->stage.softpipe, &setup->quad); } @@ -652,7 +664,7 @@ plot(struct setup_stage *setup, GLint x, GLint y) /* flush prev quad, start new quad */ if (setup->quad.x0 != -1) - quad_shade(setup->stage.softpipe, &setup->quad); + quad_emit(setup->stage.softpipe, &setup->quad); setup->quad.x0 = quadX; setup->quad.y0 = quadY; @@ -755,7 +767,7 @@ setup_line(struct prim_stage *stage, struct prim_header *prim) /* draw final quad */ if (setup->quad.mask) { - quad_shade(setup->stage.softpipe, &setup->quad); + quad_emit(setup->stage.softpipe, &setup->quad); } } @@ -810,7 +822,7 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) setup->quad.x0 = x - ix; setup->quad.y0 = y - iy; setup->quad.mask = (1 << ix) << (2 * iy); - quad_shade(setup->stage.softpipe, &setup->quad); + quad_emit(setup->stage.softpipe, &setup->quad); } else { const GLint ixmin = block((GLint) (x - halfSize)); @@ -870,7 +882,7 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) if (setup->quad.mask) { setup->quad.x0 = ix; setup->quad.y0 = iy; - quad_shade( setup->stage.softpipe, &setup->quad ); + quad_emit( setup->stage.softpipe, &setup->quad ); } } } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 1f7329600a..b53c842366 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -132,5 +132,8 @@ void softpipe_update_derived( struct softpipe_context *softpipe ) if (softpipe->dirty & (G_NEW_SETUP | G_NEW_FS)) calculate_vertex_layout( softpipe ); + if (softpipe->dirty & (G_NEW_BLEND | G_NEW_FS)) + sp_build_quad_pipeline(softpipe); + softpipe->dirty = 0; } diff --git a/src/mesa/pipe/softpipe/sp_tile.c b/src/mesa/pipe/softpipe/sp_tile.c new file mode 100644 index 0000000000..168872c64d --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tile.c @@ -0,0 +1,24 @@ + + +#include "sp_context.h" + + + +void +sp_build_quad_pipeline(struct softpipe_context *sp) +{ + + sp->quad.first = sp->quad.output; + + if (sp->blend.blend_enable) { + sp->quad.blend->next = sp->quad.first; + sp->quad.first = sp->quad.blend; + } + + /* XXX always enable shader? */ + if (1) { + sp->quad.shade->next = sp->quad.first; + sp->quad.first = sp->quad.shade; + } + +} diff --git a/src/mesa/pipe/softpipe/sp_tile.h b/src/mesa/pipe/softpipe/sp_tile.h index f4808b99d2..86c17180bf 100644 --- a/src/mesa/pipe/softpipe/sp_tile.h +++ b/src/mesa/pipe/softpipe/sp_tile.h @@ -33,10 +33,24 @@ struct softpipe_context; struct quad_header; -void quad_shade( struct softpipe_context *softpipe, - struct quad_header *quad ); -void quad_output( struct softpipe_context *softpipe, - struct quad_header *quad ); +struct quad_stage { + struct softpipe_context *softpipe; + + struct quad_stage *next; + + /** the stage action */ + void (*run)(struct quad_stage *qs, struct quad_header *quad); +}; + + + +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); + +void +sp_build_quad_pipeline(struct softpipe_context *sp); #endif diff --git a/src/mesa/pipe/softpipe/sp_tile_blend.c b/src/mesa/pipe/softpipe/sp_tile_blend.c new file mode 100644 index 0000000000..5681528d06 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tile_blend.c @@ -0,0 +1,70 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * quad blending + */ + +#include "glheader.h" +#include "imports.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_tile.h" + + + +static void +blend_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + GLfloat dest[4][QUAD_SIZE], result[4][QUAD_SIZE]; + GLuint i; + + /* XXX we're also looping in output_quad() !?! */ + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + struct softpipe_surface *sps + = softpipe_surface(softpipe->framebuffer.cbufs[i]); + + sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + + /* XXX do blend here */ + + qs->next->run(qs->next, quad); + } +} + + + + +struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = blend_quad; + + return stage; +} diff --git a/src/mesa/pipe/softpipe/sp_tile_depth_test.c b/src/mesa/pipe/softpipe/sp_tile_depth_test.c new file mode 100644 index 0000000000..5bc2648531 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tile_depth_test.c @@ -0,0 +1,73 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * quad blending + */ + +#include "glheader.h" +#include "imports.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_tile.h" + + + +static void +depth_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ +#if 0 + struct softpipe_context *softpipe = qs->softpipe; + GLfloat dest[4][QUAD_SIZE], result[4][QUAD_SIZE]; + GLuint i; + + /* XXX we're also looping in output_quad() !?! */ + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + struct softpipe_surface *sps + = softpipe_surface(softpipe->framebuffer.cbufs[i]); + + sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + + /* XXX do blend here */ + + } +#endif + + qs->next->run(qs->next, quad); +} + + + + +struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = depth_test_quad; + + return stage; +} diff --git a/src/mesa/pipe/softpipe/sp_tile_fs.c b/src/mesa/pipe/softpipe/sp_tile_fs.c index c145fcf089..912831fefa 100644 --- a/src/mesa/pipe/softpipe/sp_tile_fs.c +++ b/src/mesa/pipe/softpipe/sp_tile_fs.c @@ -33,6 +33,7 @@ */ #include "glheader.h" +#include "imports.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_tile.h" @@ -116,9 +117,10 @@ static INLINE void pinterp( struct exec_machine *exec, /* This should be done by the fragment shader execution unit (code * generated from the decl instructions). Do it here for now. */ -void quad_shade( struct softpipe_context *softpipe, - struct quad_header *quad ) +static void +shade_quad( struct quad_stage *qs, struct quad_header *quad ) { + struct softpipe_context *softpipe = qs->softpipe; struct exec_machine exec; GLfloat fx = quad->x0; GLfloat fy = quad->y0; @@ -190,14 +192,17 @@ void quad_shade( struct softpipe_context *softpipe, } #endif - - if (quad->mask) - quad_output( softpipe, quad ); + qs->next->run(qs->next, quad); } +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + stage->softpipe = softpipe; + stage->run = shade_quad; - - + return stage; +} diff --git a/src/mesa/pipe/softpipe/sp_tile_output.c b/src/mesa/pipe/softpipe/sp_tile_output.c index 23086b7020..4f4421891b 100644 --- a/src/mesa/pipe/softpipe/sp_tile_output.c +++ b/src/mesa/pipe/softpipe/sp_tile_output.c @@ -33,6 +33,7 @@ */ #include "glheader.h" +#include "imports.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" @@ -60,9 +61,10 @@ static void mask_copy( GLfloat (*dest)[4], * * Note that surfaces support only full quad reads and writes. */ -void quad_output( struct softpipe_context *softpipe, - struct quad_header *quad ) +static void +output_quad(struct quad_stage *qs, struct quad_header *quad) { + struct softpipe_context *softpipe = qs->softpipe; GLuint i; for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { @@ -87,3 +89,15 @@ void quad_output( struct softpipe_context *softpipe, } } } + + +struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = output_quad; + + return stage; +} + diff --git a/src/mesa/sources b/src/mesa/sources index 97801648c4..d07bd3fbe0 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -171,6 +171,9 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_state_point.c \ pipe/softpipe/sp_state_setup.c \ pipe/softpipe/sp_state_surface.c \ + pipe/softpipe/sp_tile.c \ + pipe/softpipe/sp_tile_blend.c \ + pipe/softpipe/sp_tile_depth_test.c \ pipe/softpipe/sp_tile_fs.c \ pipe/softpipe/sp_tile_output.c -- cgit v1.2.3 From 7658a0208aa31f1443be6eaa720a383b34e2bbb3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 15:14:47 -0600 Subject: quad blending works now, but many blend terms need to be added in blend_quad(). --- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_state.h | 3 + src/mesa/pipe/softpipe/sp_state_blend.c | 47 ++++++++++ src/mesa/pipe/softpipe/sp_tile_blend.c | 157 +++++++++++++++++++++++++++----- src/mesa/sources | 1 + 5 files changed, 185 insertions(+), 24 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_state_blend.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index d1023fb782..593be0e132 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -65,6 +65,7 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.destroy = softpipe_destroy; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; + softpipe->pipe.set_blend_state = softpipe_set_blend_state; softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_point_state = softpipe_set_point_state; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index bc5a3512ae..5a657c4a8f 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -38,6 +38,9 @@ void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); +void softpipe_set_blend_state( struct pipe_context *, + const struct pipe_blend_state * ); + void softpipe_set_clear_color_state( struct pipe_context *, const struct pipe_clear_color_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c new file mode 100644 index 0000000000..1fd7a44105 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_draw.h" + + + +void softpipe_set_blend_state( struct pipe_context *pipe, + const struct pipe_blend_state *blend ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->blend = *blend; + + softpipe->dirty |= G_NEW_BLEND; +} + diff --git a/src/mesa/pipe/softpipe/sp_tile_blend.c b/src/mesa/pipe/softpipe/sp_tile_blend.c index 5681528d06..40af57149d 100644 --- a/src/mesa/pipe/softpipe/sp_tile_blend.c +++ b/src/mesa/pipe/softpipe/sp_tile_blend.c @@ -1,33 +1,39 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ /** * quad blending + * \author Brian Paul */ #include "glheader.h" #include "imports.h" +#include "macros.h" +#include "pipe/p_defines.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" @@ -39,19 +45,122 @@ static void blend_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - GLfloat dest[4][QUAD_SIZE], result[4][QUAD_SIZE]; - GLuint i; + GLfloat source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; + GLuint i, j; /* XXX we're also looping in output_quad() !?! */ + /* copy quad's colors since we'll modify them in the loop */ + memcpy(source, quad->outputs.color, 4 * 4 * sizeof(GLfloat)); + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + GLfloat srcTerm[4], dstTerm[4]; struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.cbufs[i]); sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); - /* XXX do blend here */ + /* XXX this loop could be factored out - we'd compute the src/dstTerm + * for all four pixels in the quad at once. + */ + for (j = 0; j < QUAD_SIZE; j++) { + switch (softpipe->blend.rgb_src_factor) { + case PIPE_BLENDFACTOR_ONE: + srcTerm[0] = srcTerm[1] = srcTerm[2] = 1.0; + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + srcTerm[0] = srcTerm[1] = srcTerm[2] = quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + srcTerm[0] = srcTerm[1] = srcTerm[2] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.alpha_src_factor) { + case PIPE_BLENDFACTOR_ONE: + srcTerm[3] = 1.0; + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + srcTerm[3] = quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + srcTerm[3] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.rgb_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0; + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0 - quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + dstTerm[0] = dstTerm[1] = dstTerm[2] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.alpha_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + dstTerm[3] = 1.0; + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + dstTerm[3] = 1.0 - quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + dstTerm[3] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.rgb_func) { + case PIPE_BLEND_ADD: + quad->outputs.color[0][j] = source[0][j] * srcTerm[0] + dest[0][j] * dstTerm[0]; + quad->outputs.color[1][j] = source[1][j] * srcTerm[1] + dest[1][j] * dstTerm[1]; + quad->outputs.color[2][j] = source[2][j] * srcTerm[2] + dest[2][j] * dstTerm[2]; + quad->outputs.color[3][j] = source[3][j] * srcTerm[3] + dest[3][j] * dstTerm[3]; + break; + case PIPE_BLEND_SUBTRACT: + quad->outputs.color[0][j] = source[0][j] * srcTerm[0] - dest[0][j] * dstTerm[0]; + quad->outputs.color[1][j] = source[1][j] * srcTerm[1] - dest[1][j] * dstTerm[1]; + quad->outputs.color[2][j] = source[2][j] * srcTerm[2] - dest[2][j] * dstTerm[2]; + quad->outputs.color[3][j] = source[3][j] * srcTerm[3] - dest[3][j] * dstTerm[3]; + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + quad->outputs.color[0][j] = dest[0][j] * dstTerm[0] - source[0][j] * srcTerm[0]; + quad->outputs.color[1][j] = dest[1][j] * dstTerm[1] - source[1][j] * srcTerm[1]; + quad->outputs.color[2][j] = dest[2][j] * dstTerm[2] - source[2][j] * srcTerm[2]; + quad->outputs.color[3][j] = dest[3][j] * dstTerm[3] - source[3][j] * srcTerm[3]; + break; + case PIPE_BLEND_MIN: + quad->outputs.color[0][j] = MIN2(dest[0][j], source[0][j]); + quad->outputs.color[1][j] = MIN2(dest[1][j], source[1][j]); + quad->outputs.color[2][j] = MIN2(dest[2][j], source[2][j]); + quad->outputs.color[3][j] = MIN2(dest[3][j], source[3][j]); + break; + case PIPE_BLEND_MAX: + quad->outputs.color[0][j] = MAX2(dest[0][j], source[0][j]); + quad->outputs.color[1][j] = MAX2(dest[1][j], source[1][j]); + quad->outputs.color[2][j] = MAX2(dest[2][j], source[2][j]); + quad->outputs.color[3][j] = MAX2(dest[3][j], source[3][j]); + break; + default: + abort(); + } + } /* loop over quads */ + /* pass blended quad to next stage */ qs->next->run(qs->next, quad); } } diff --git a/src/mesa/sources b/src/mesa/sources index d07bd3fbe0..cc675eab39 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -165,6 +165,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_prim_setup.c \ pipe/softpipe/sp_prim_twoside.c \ pipe/softpipe/sp_prim_unfilled.c \ + pipe/softpipe/sp_state_blend.c \ pipe/softpipe/sp_state_clip.c \ pipe/softpipe/sp_state_derived.c \ pipe/softpipe/sp_state_fs.c \ -- cgit v1.2.3 From acef7c750113e34a92fd00c74308443d8fb3f974 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 15:22:39 -0600 Subject: only blend active pixels in the quad --- src/mesa/pipe/softpipe/sp_tile_blend.c | 191 +++++++++++++++++---------------- 1 file changed, 98 insertions(+), 93 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_blend.c b/src/mesa/pipe/softpipe/sp_tile_blend.c index 40af57149d..9ac0ccbfa5 100644 --- a/src/mesa/pipe/softpipe/sp_tile_blend.c +++ b/src/mesa/pipe/softpipe/sp_tile_blend.c @@ -64,101 +64,106 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) * for all four pixels in the quad at once. */ for (j = 0; j < QUAD_SIZE; j++) { - switch (softpipe->blend.rgb_src_factor) { - case PIPE_BLENDFACTOR_ONE: - srcTerm[0] = srcTerm[1] = srcTerm[2] = 1.0; - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - srcTerm[0] = srcTerm[1] = srcTerm[2] = quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - srcTerm[0] = srcTerm[1] = srcTerm[2] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.alpha_src_factor) { - case PIPE_BLENDFACTOR_ONE: - srcTerm[3] = 1.0; - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - srcTerm[3] = quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - srcTerm[3] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.rgb_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0; - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0 - quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - dstTerm[0] = dstTerm[1] = dstTerm[2] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.alpha_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - dstTerm[3] = 1.0; - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - dstTerm[3] = 1.0 - quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - dstTerm[3] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - switch (softpipe->blend.rgb_func) { - case PIPE_BLEND_ADD: - quad->outputs.color[0][j] = source[0][j] * srcTerm[0] + dest[0][j] * dstTerm[0]; - quad->outputs.color[1][j] = source[1][j] * srcTerm[1] + dest[1][j] * dstTerm[1]; - quad->outputs.color[2][j] = source[2][j] * srcTerm[2] + dest[2][j] * dstTerm[2]; - quad->outputs.color[3][j] = source[3][j] * srcTerm[3] + dest[3][j] * dstTerm[3]; - break; - case PIPE_BLEND_SUBTRACT: - quad->outputs.color[0][j] = source[0][j] * srcTerm[0] - dest[0][j] * dstTerm[0]; - quad->outputs.color[1][j] = source[1][j] * srcTerm[1] - dest[1][j] * dstTerm[1]; - quad->outputs.color[2][j] = source[2][j] * srcTerm[2] - dest[2][j] * dstTerm[2]; - quad->outputs.color[3][j] = source[3][j] * srcTerm[3] - dest[3][j] * dstTerm[3]; - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - quad->outputs.color[0][j] = dest[0][j] * dstTerm[0] - source[0][j] * srcTerm[0]; - quad->outputs.color[1][j] = dest[1][j] * dstTerm[1] - source[1][j] * srcTerm[1]; - quad->outputs.color[2][j] = dest[2][j] * dstTerm[2] - source[2][j] * srcTerm[2]; - quad->outputs.color[3][j] = dest[3][j] * dstTerm[3] - source[3][j] * srcTerm[3]; - break; - case PIPE_BLEND_MIN: - quad->outputs.color[0][j] = MIN2(dest[0][j], source[0][j]); - quad->outputs.color[1][j] = MIN2(dest[1][j], source[1][j]); - quad->outputs.color[2][j] = MIN2(dest[2][j], source[2][j]); - quad->outputs.color[3][j] = MIN2(dest[3][j], source[3][j]); - break; - case PIPE_BLEND_MAX: - quad->outputs.color[0][j] = MAX2(dest[0][j], source[0][j]); - quad->outputs.color[1][j] = MAX2(dest[1][j], source[1][j]); - quad->outputs.color[2][j] = MAX2(dest[2][j], source[2][j]); - quad->outputs.color[3][j] = MAX2(dest[3][j], source[3][j]); - break; - default: - abort(); + /* if this pixel in the quad is alive */ + if (quad->mask & (1 << j)) { + + switch (softpipe->blend.rgb_src_factor) { + case PIPE_BLENDFACTOR_ONE: + srcTerm[0] = srcTerm[1] = srcTerm[2] = 1.0; + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + srcTerm[0] = srcTerm[1] = srcTerm[2] = quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + srcTerm[0] = srcTerm[1] = srcTerm[2] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.alpha_src_factor) { + case PIPE_BLENDFACTOR_ONE: + srcTerm[3] = 1.0; + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + srcTerm[3] = quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + srcTerm[3] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.rgb_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0; + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0 - quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + dstTerm[0] = dstTerm[1] = dstTerm[2] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.alpha_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + dstTerm[3] = 1.0; + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + dstTerm[3] = 1.0 - quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + dstTerm[3] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.rgb_func) { + case PIPE_BLEND_ADD: + quad->outputs.color[0][j] = source[0][j] * srcTerm[0] + dest[0][j] * dstTerm[0]; + quad->outputs.color[1][j] = source[1][j] * srcTerm[1] + dest[1][j] * dstTerm[1]; + quad->outputs.color[2][j] = source[2][j] * srcTerm[2] + dest[2][j] * dstTerm[2]; + quad->outputs.color[3][j] = source[3][j] * srcTerm[3] + dest[3][j] * dstTerm[3]; + break; + case PIPE_BLEND_SUBTRACT: + quad->outputs.color[0][j] = source[0][j] * srcTerm[0] - dest[0][j] * dstTerm[0]; + quad->outputs.color[1][j] = source[1][j] * srcTerm[1] - dest[1][j] * dstTerm[1]; + quad->outputs.color[2][j] = source[2][j] * srcTerm[2] - dest[2][j] * dstTerm[2]; + quad->outputs.color[3][j] = source[3][j] * srcTerm[3] - dest[3][j] * dstTerm[3]; + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + quad->outputs.color[0][j] = dest[0][j] * dstTerm[0] - source[0][j] * srcTerm[0]; + quad->outputs.color[1][j] = dest[1][j] * dstTerm[1] - source[1][j] * srcTerm[1]; + quad->outputs.color[2][j] = dest[2][j] * dstTerm[2] - source[2][j] * srcTerm[2]; + quad->outputs.color[3][j] = dest[3][j] * dstTerm[3] - source[3][j] * srcTerm[3]; + break; + case PIPE_BLEND_MIN: + quad->outputs.color[0][j] = MIN2(dest[0][j], source[0][j]); + quad->outputs.color[1][j] = MIN2(dest[1][j], source[1][j]); + quad->outputs.color[2][j] = MIN2(dest[2][j], source[2][j]); + quad->outputs.color[3][j] = MIN2(dest[3][j], source[3][j]); + break; + case PIPE_BLEND_MAX: + quad->outputs.color[0][j] = MAX2(dest[0][j], source[0][j]); + quad->outputs.color[1][j] = MAX2(dest[1][j], source[1][j]); + quad->outputs.color[2][j] = MAX2(dest[2][j], source[2][j]); + quad->outputs.color[3][j] = MAX2(dest[3][j], source[3][j]); + break; + default: + abort(); + } } - } /* loop over quads */ + } /* loop over quad's pixels*/ /* pass blended quad to next stage */ qs->next->run(qs->next, quad); -- cgit v1.2.3 From e0398f43a2d7b454592135e1ed33a8a6c303ecc9 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 16:12:26 -0600 Subject: Rename "tile" files to "quad". --- src/mesa/pipe/softpipe/sp_quad.c | 24 ++++ src/mesa/pipe/softpipe/sp_quad.h | 56 ++++++++ src/mesa/pipe/softpipe/sp_quad_blend.c | 184 ++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_quad_depth_test.c | 73 ++++++++++ src/mesa/pipe/softpipe/sp_quad_fs.c | 208 ++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_quad_output.c | 103 ++++++++++++++ src/mesa/pipe/softpipe/sp_tile.c | 24 ---- src/mesa/pipe/softpipe/sp_tile.h | 56 -------- src/mesa/pipe/softpipe/sp_tile_blend.c | 184 ------------------------ src/mesa/pipe/softpipe/sp_tile_depth_test.c | 73 ---------- src/mesa/pipe/softpipe/sp_tile_fs.c | 208 ---------------------------- src/mesa/pipe/softpipe/sp_tile_output.c | 103 -------------- 12 files changed, 648 insertions(+), 648 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_quad.c create mode 100644 src/mesa/pipe/softpipe/sp_quad.h create mode 100644 src/mesa/pipe/softpipe/sp_quad_blend.c create mode 100644 src/mesa/pipe/softpipe/sp_quad_depth_test.c create mode 100644 src/mesa/pipe/softpipe/sp_quad_fs.c create mode 100644 src/mesa/pipe/softpipe/sp_quad_output.c delete mode 100644 src/mesa/pipe/softpipe/sp_tile.c delete mode 100644 src/mesa/pipe/softpipe/sp_tile.h delete mode 100644 src/mesa/pipe/softpipe/sp_tile_blend.c delete mode 100644 src/mesa/pipe/softpipe/sp_tile_depth_test.c delete mode 100644 src/mesa/pipe/softpipe/sp_tile_fs.c delete mode 100644 src/mesa/pipe/softpipe/sp_tile_output.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c new file mode 100644 index 0000000000..168872c64d --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -0,0 +1,24 @@ + + +#include "sp_context.h" + + + +void +sp_build_quad_pipeline(struct softpipe_context *sp) +{ + + sp->quad.first = sp->quad.output; + + if (sp->blend.blend_enable) { + sp->quad.blend->next = sp->quad.first; + sp->quad.first = sp->quad.blend; + } + + /* XXX always enable shader? */ + if (1) { + sp->quad.shade->next = sp->quad.first; + sp->quad.first = sp->quad.shade; + } + +} diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h new file mode 100644 index 0000000000..86c17180bf --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -0,0 +1,56 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#ifndef G_TILE_H +#define G_TILE_H + +struct softpipe_context; +struct quad_header; + + +struct quad_stage { + struct softpipe_context *softpipe; + + struct quad_stage *next; + + /** the stage action */ + void (*run)(struct quad_stage *qs, struct quad_header *quad); +}; + + + +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); + +void +sp_build_quad_pipeline(struct softpipe_context *sp); + +#endif diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c new file mode 100644 index 0000000000..9ac0ccbfa5 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -0,0 +1,184 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * quad blending + * \author Brian Paul + */ + +#include "glheader.h" +#include "imports.h" +#include "macros.h" +#include "pipe/p_defines.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_tile.h" + + + +static void +blend_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + GLfloat source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; + GLuint i, j; + + /* XXX we're also looping in output_quad() !?! */ + + /* copy quad's colors since we'll modify them in the loop */ + memcpy(source, quad->outputs.color, 4 * 4 * sizeof(GLfloat)); + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + GLfloat srcTerm[4], dstTerm[4]; + struct softpipe_surface *sps + = softpipe_surface(softpipe->framebuffer.cbufs[i]); + + sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + + /* XXX this loop could be factored out - we'd compute the src/dstTerm + * for all four pixels in the quad at once. + */ + for (j = 0; j < QUAD_SIZE; j++) { + + /* if this pixel in the quad is alive */ + if (quad->mask & (1 << j)) { + + switch (softpipe->blend.rgb_src_factor) { + case PIPE_BLENDFACTOR_ONE: + srcTerm[0] = srcTerm[1] = srcTerm[2] = 1.0; + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + srcTerm[0] = srcTerm[1] = srcTerm[2] = quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + srcTerm[0] = srcTerm[1] = srcTerm[2] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.alpha_src_factor) { + case PIPE_BLENDFACTOR_ONE: + srcTerm[3] = 1.0; + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + srcTerm[3] = quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + srcTerm[3] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.rgb_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0; + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0 - quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + dstTerm[0] = dstTerm[1] = dstTerm[2] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.alpha_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + dstTerm[3] = 1.0; + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + dstTerm[3] = 1.0 - quad->outputs.color[3][j]; + break; + case PIPE_BLENDFACTOR_ZERO: + dstTerm[3] = 0.0; + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + switch (softpipe->blend.rgb_func) { + case PIPE_BLEND_ADD: + quad->outputs.color[0][j] = source[0][j] * srcTerm[0] + dest[0][j] * dstTerm[0]; + quad->outputs.color[1][j] = source[1][j] * srcTerm[1] + dest[1][j] * dstTerm[1]; + quad->outputs.color[2][j] = source[2][j] * srcTerm[2] + dest[2][j] * dstTerm[2]; + quad->outputs.color[3][j] = source[3][j] * srcTerm[3] + dest[3][j] * dstTerm[3]; + break; + case PIPE_BLEND_SUBTRACT: + quad->outputs.color[0][j] = source[0][j] * srcTerm[0] - dest[0][j] * dstTerm[0]; + quad->outputs.color[1][j] = source[1][j] * srcTerm[1] - dest[1][j] * dstTerm[1]; + quad->outputs.color[2][j] = source[2][j] * srcTerm[2] - dest[2][j] * dstTerm[2]; + quad->outputs.color[3][j] = source[3][j] * srcTerm[3] - dest[3][j] * dstTerm[3]; + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + quad->outputs.color[0][j] = dest[0][j] * dstTerm[0] - source[0][j] * srcTerm[0]; + quad->outputs.color[1][j] = dest[1][j] * dstTerm[1] - source[1][j] * srcTerm[1]; + quad->outputs.color[2][j] = dest[2][j] * dstTerm[2] - source[2][j] * srcTerm[2]; + quad->outputs.color[3][j] = dest[3][j] * dstTerm[3] - source[3][j] * srcTerm[3]; + break; + case PIPE_BLEND_MIN: + quad->outputs.color[0][j] = MIN2(dest[0][j], source[0][j]); + quad->outputs.color[1][j] = MIN2(dest[1][j], source[1][j]); + quad->outputs.color[2][j] = MIN2(dest[2][j], source[2][j]); + quad->outputs.color[3][j] = MIN2(dest[3][j], source[3][j]); + break; + case PIPE_BLEND_MAX: + quad->outputs.color[0][j] = MAX2(dest[0][j], source[0][j]); + quad->outputs.color[1][j] = MAX2(dest[1][j], source[1][j]); + quad->outputs.color[2][j] = MAX2(dest[2][j], source[2][j]); + quad->outputs.color[3][j] = MAX2(dest[3][j], source[3][j]); + break; + default: + abort(); + } + } + } /* loop over quad's pixels*/ + + /* pass blended quad to next stage */ + qs->next->run(qs->next, quad); + } +} + + + + +struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = blend_quad; + + return stage; +} diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c new file mode 100644 index 0000000000..5bc2648531 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -0,0 +1,73 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * quad blending + */ + +#include "glheader.h" +#include "imports.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_tile.h" + + + +static void +depth_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ +#if 0 + struct softpipe_context *softpipe = qs->softpipe; + GLfloat dest[4][QUAD_SIZE], result[4][QUAD_SIZE]; + GLuint i; + + /* XXX we're also looping in output_quad() !?! */ + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + struct softpipe_surface *sps + = softpipe_surface(softpipe->framebuffer.cbufs[i]); + + sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + + /* XXX do blend here */ + + } +#endif + + qs->next->run(qs->next, quad); +} + + + + +struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = depth_test_quad; + + return stage; +} diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c new file mode 100644 index 0000000000..912831fefa --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -0,0 +1,208 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "imports.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_tile.h" + +struct exec_machine { + const struct setup_coefficient *coef; + + GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE]; +}; + + +/** + * Compute quad's attributes values, as constants (GL_FLAT shading). + */ +static INLINE void cinterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + exec->attr[attrib][i][j] = exec->coef[attrib].a0[i]; + } +} + + +/** + * Compute quad's attribute values by linear interpolation. + * + * Push into the fp: + * + * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx + * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy + */ +static INLINE void linterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; + exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + + exec->coef[attrib].dadx[i] * x + + exec->coef[attrib].dady[i] * y); + } +} + + +/** + * Compute quad's attribute values by linear interpolation with + * perspective correction. + * + * Push into the fp: + * + * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx + * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy + * INPUT[attr] = MUL INPUT[attr], INPUT_WPOS.wwww + * + * (Or should that be 1/w ???) + */ +static INLINE void pinterp( struct exec_machine *exec, + GLuint attrib, + GLuint i ) +{ + GLuint j; + + for (j = 0; j < QUAD_SIZE; j++) { + const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; + const GLfloat invW = exec->attr[FRAG_ATTRIB_WPOS][3][j]; + exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + + exec->coef[attrib].dadx[i] * x + + exec->coef[attrib].dady[i] * y) * invW); + } +} + + + +/* This should be done by the fragment shader execution unit (code + * generated from the decl instructions). Do it here for now. + */ +static void +shade_quad( struct quad_stage *qs, struct quad_header *quad ) +{ + struct softpipe_context *softpipe = qs->softpipe; + struct exec_machine exec; + GLfloat fx = quad->x0; + GLfloat fy = quad->y0; + GLuint i, j; + + exec.coef = quad->coef; + + /* Position: + */ + exec.attr[FRAG_ATTRIB_WPOS][0][0] = fx; + exec.attr[FRAG_ATTRIB_WPOS][0][1] = fx + 1.0; + exec.attr[FRAG_ATTRIB_WPOS][0][2] = fx; + exec.attr[FRAG_ATTRIB_WPOS][0][3] = fx + 1.0; + + exec.attr[FRAG_ATTRIB_WPOS][1][0] = fy; + exec.attr[FRAG_ATTRIB_WPOS][1][1] = fy; + exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0; + exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; + + /* Z and W are done by linear interpolation: + * XXX we'll probably have to use integers for Z + */ + if (softpipe->need_z) { + linterp(&exec, 0, 2); + } + + if (softpipe->need_w) { + linterp(&exec, 0, 3); +// invert(&exec, 0, 3); + } + + /* Interpolate all the remaining attributes. This will get pushed + * into the fragment program's responsibilities at some point. + */ + for (i = 1; i < quad->nr_attrs; i++) { +#if 1 + for (j = 0; j < NUM_CHANNELS; j++) + linterp(&exec, i, j); +#else + switch (quad->interp[i]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + cinterp(&exec, i, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + linterp(&exec, i, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + pinterp(&exec, i, j); + break; + } +#endif + } + +#if 0 + softpipe->run_fs( tri->fp, quad, &tri->outputs ); +#else + { + GLuint attr = softpipe->fp_attr_to_slot[FRAG_ATTRIB_COL0]; + assert(attr); + + memcpy(quad->outputs.color, + exec.attr[attr], + sizeof(quad->outputs.color)); + } +#endif + + qs->next->run(qs->next, quad); +} + + + +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = shade_quad; + + return stage; +} diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c new file mode 100644 index 0000000000..4f4421891b --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -0,0 +1,103 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "glheader.h" +#include "imports.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_tile.h" + + +static void mask_copy( GLfloat (*dest)[4], + GLfloat (*src)[4], + GLuint mask ) +{ + GLuint i, j; + + for (i = 0; i < 4; i++) { + if (mask & (1<softpipe; + GLuint i; + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + struct softpipe_surface *sps + = softpipe_surface(softpipe->framebuffer.cbufs[i]); + + if (quad->mask != MASK_ALL) { + GLfloat tmp[4][QUAD_SIZE]; + + /* Yes, we'll probably have a masked write as well, but this is + * how blend will be done at least. + */ + + sps->read_quad_f_swz(sps, quad->x0, quad->y0, tmp); + + mask_copy( tmp, quad->outputs.color, quad->mask ); + + sps->write_quad_f_swz(sps, quad->x0, quad->y0, tmp); + } + else { + sps->write_quad_f_swz(sps, quad->x0, quad->y0, quad->outputs.color); + } + } +} + + +struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = output_quad; + + return stage; +} + diff --git a/src/mesa/pipe/softpipe/sp_tile.c b/src/mesa/pipe/softpipe/sp_tile.c deleted file mode 100644 index 168872c64d..0000000000 --- a/src/mesa/pipe/softpipe/sp_tile.c +++ /dev/null @@ -1,24 +0,0 @@ - - -#include "sp_context.h" - - - -void -sp_build_quad_pipeline(struct softpipe_context *sp) -{ - - sp->quad.first = sp->quad.output; - - if (sp->blend.blend_enable) { - sp->quad.blend->next = sp->quad.first; - sp->quad.first = sp->quad.blend; - } - - /* XXX always enable shader? */ - if (1) { - sp->quad.shade->next = sp->quad.first; - sp->quad.first = sp->quad.shade; - } - -} diff --git a/src/mesa/pipe/softpipe/sp_tile.h b/src/mesa/pipe/softpipe/sp_tile.h deleted file mode 100644 index 86c17180bf..0000000000 --- a/src/mesa/pipe/softpipe/sp_tile.h +++ /dev/null @@ -1,56 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#ifndef G_TILE_H -#define G_TILE_H - -struct softpipe_context; -struct quad_header; - - -struct quad_stage { - struct softpipe_context *softpipe; - - struct quad_stage *next; - - /** the stage action */ - void (*run)(struct quad_stage *qs, struct quad_header *quad); -}; - - - -struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); - -void -sp_build_quad_pipeline(struct softpipe_context *sp); - -#endif diff --git a/src/mesa/pipe/softpipe/sp_tile_blend.c b/src/mesa/pipe/softpipe/sp_tile_blend.c deleted file mode 100644 index 9ac0ccbfa5..0000000000 --- a/src/mesa/pipe/softpipe/sp_tile_blend.c +++ /dev/null @@ -1,184 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * quad blending - * \author Brian Paul - */ - -#include "glheader.h" -#include "imports.h" -#include "macros.h" -#include "pipe/p_defines.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_tile.h" - - - -static void -blend_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - GLfloat source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; - GLuint i, j; - - /* XXX we're also looping in output_quad() !?! */ - - /* copy quad's colors since we'll modify them in the loop */ - memcpy(source, quad->outputs.color, 4 * 4 * sizeof(GLfloat)); - - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - GLfloat srcTerm[4], dstTerm[4]; - struct softpipe_surface *sps - = softpipe_surface(softpipe->framebuffer.cbufs[i]); - - sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); - - /* XXX this loop could be factored out - we'd compute the src/dstTerm - * for all four pixels in the quad at once. - */ - for (j = 0; j < QUAD_SIZE; j++) { - - /* if this pixel in the quad is alive */ - if (quad->mask & (1 << j)) { - - switch (softpipe->blend.rgb_src_factor) { - case PIPE_BLENDFACTOR_ONE: - srcTerm[0] = srcTerm[1] = srcTerm[2] = 1.0; - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - srcTerm[0] = srcTerm[1] = srcTerm[2] = quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - srcTerm[0] = srcTerm[1] = srcTerm[2] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.alpha_src_factor) { - case PIPE_BLENDFACTOR_ONE: - srcTerm[3] = 1.0; - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - srcTerm[3] = quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - srcTerm[3] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.rgb_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0; - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0 - quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - dstTerm[0] = dstTerm[1] = dstTerm[2] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.alpha_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - dstTerm[3] = 1.0; - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - dstTerm[3] = 1.0 - quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - dstTerm[3] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.rgb_func) { - case PIPE_BLEND_ADD: - quad->outputs.color[0][j] = source[0][j] * srcTerm[0] + dest[0][j] * dstTerm[0]; - quad->outputs.color[1][j] = source[1][j] * srcTerm[1] + dest[1][j] * dstTerm[1]; - quad->outputs.color[2][j] = source[2][j] * srcTerm[2] + dest[2][j] * dstTerm[2]; - quad->outputs.color[3][j] = source[3][j] * srcTerm[3] + dest[3][j] * dstTerm[3]; - break; - case PIPE_BLEND_SUBTRACT: - quad->outputs.color[0][j] = source[0][j] * srcTerm[0] - dest[0][j] * dstTerm[0]; - quad->outputs.color[1][j] = source[1][j] * srcTerm[1] - dest[1][j] * dstTerm[1]; - quad->outputs.color[2][j] = source[2][j] * srcTerm[2] - dest[2][j] * dstTerm[2]; - quad->outputs.color[3][j] = source[3][j] * srcTerm[3] - dest[3][j] * dstTerm[3]; - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - quad->outputs.color[0][j] = dest[0][j] * dstTerm[0] - source[0][j] * srcTerm[0]; - quad->outputs.color[1][j] = dest[1][j] * dstTerm[1] - source[1][j] * srcTerm[1]; - quad->outputs.color[2][j] = dest[2][j] * dstTerm[2] - source[2][j] * srcTerm[2]; - quad->outputs.color[3][j] = dest[3][j] * dstTerm[3] - source[3][j] * srcTerm[3]; - break; - case PIPE_BLEND_MIN: - quad->outputs.color[0][j] = MIN2(dest[0][j], source[0][j]); - quad->outputs.color[1][j] = MIN2(dest[1][j], source[1][j]); - quad->outputs.color[2][j] = MIN2(dest[2][j], source[2][j]); - quad->outputs.color[3][j] = MIN2(dest[3][j], source[3][j]); - break; - case PIPE_BLEND_MAX: - quad->outputs.color[0][j] = MAX2(dest[0][j], source[0][j]); - quad->outputs.color[1][j] = MAX2(dest[1][j], source[1][j]); - quad->outputs.color[2][j] = MAX2(dest[2][j], source[2][j]); - quad->outputs.color[3][j] = MAX2(dest[3][j], source[3][j]); - break; - default: - abort(); - } - } - } /* loop over quad's pixels*/ - - /* pass blended quad to next stage */ - qs->next->run(qs->next, quad); - } -} - - - - -struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->run = blend_quad; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_tile_depth_test.c b/src/mesa/pipe/softpipe/sp_tile_depth_test.c deleted file mode 100644 index 5bc2648531..0000000000 --- a/src/mesa/pipe/softpipe/sp_tile_depth_test.c +++ /dev/null @@ -1,73 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * quad blending - */ - -#include "glheader.h" -#include "imports.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_tile.h" - - - -static void -depth_test_quad(struct quad_stage *qs, struct quad_header *quad) -{ -#if 0 - struct softpipe_context *softpipe = qs->softpipe; - GLfloat dest[4][QUAD_SIZE], result[4][QUAD_SIZE]; - GLuint i; - - /* XXX we're also looping in output_quad() !?! */ - - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps - = softpipe_surface(softpipe->framebuffer.cbufs[i]); - - sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); - - /* XXX do blend here */ - - } -#endif - - qs->next->run(qs->next, quad); -} - - - - -struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->run = depth_test_quad; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_tile_fs.c b/src/mesa/pipe/softpipe/sp_tile_fs.c deleted file mode 100644 index 912831fefa..0000000000 --- a/src/mesa/pipe/softpipe/sp_tile_fs.c +++ /dev/null @@ -1,208 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "glheader.h" -#include "imports.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_tile.h" - -struct exec_machine { - const struct setup_coefficient *coef; - - GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE]; -}; - - -/** - * Compute quad's attributes values, as constants (GL_FLAT shading). - */ -static INLINE void cinterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) -{ - GLuint j; - - for (j = 0; j < QUAD_SIZE; j++) { - exec->attr[attrib][i][j] = exec->coef[attrib].a0[i]; - } -} - - -/** - * Compute quad's attribute values by linear interpolation. - * - * Push into the fp: - * - * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx - * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy - */ -static INLINE void linterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) -{ - GLuint j; - - for (j = 0; j < QUAD_SIZE; j++) { - const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; - exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * x + - exec->coef[attrib].dady[i] * y); - } -} - - -/** - * Compute quad's attribute values by linear interpolation with - * perspective correction. - * - * Push into the fp: - * - * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx - * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy - * INPUT[attr] = MUL INPUT[attr], INPUT_WPOS.wwww - * - * (Or should that be 1/w ???) - */ -static INLINE void pinterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) -{ - GLuint j; - - for (j = 0; j < QUAD_SIZE; j++) { - const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; - const GLfloat invW = exec->attr[FRAG_ATTRIB_WPOS][3][j]; - exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * x + - exec->coef[attrib].dady[i] * y) * invW); - } -} - - - -/* This should be done by the fragment shader execution unit (code - * generated from the decl instructions). Do it here for now. - */ -static void -shade_quad( struct quad_stage *qs, struct quad_header *quad ) -{ - struct softpipe_context *softpipe = qs->softpipe; - struct exec_machine exec; - GLfloat fx = quad->x0; - GLfloat fy = quad->y0; - GLuint i, j; - - exec.coef = quad->coef; - - /* Position: - */ - exec.attr[FRAG_ATTRIB_WPOS][0][0] = fx; - exec.attr[FRAG_ATTRIB_WPOS][0][1] = fx + 1.0; - exec.attr[FRAG_ATTRIB_WPOS][0][2] = fx; - exec.attr[FRAG_ATTRIB_WPOS][0][3] = fx + 1.0; - - exec.attr[FRAG_ATTRIB_WPOS][1][0] = fy; - exec.attr[FRAG_ATTRIB_WPOS][1][1] = fy; - exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0; - exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; - - /* Z and W are done by linear interpolation: - * XXX we'll probably have to use integers for Z - */ - if (softpipe->need_z) { - linterp(&exec, 0, 2); - } - - if (softpipe->need_w) { - linterp(&exec, 0, 3); -// invert(&exec, 0, 3); - } - - /* Interpolate all the remaining attributes. This will get pushed - * into the fragment program's responsibilities at some point. - */ - for (i = 1; i < quad->nr_attrs; i++) { -#if 1 - for (j = 0; j < NUM_CHANNELS; j++) - linterp(&exec, i, j); -#else - switch (quad->interp[i]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - cinterp(&exec, i, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - linterp(&exec, i, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - pinterp(&exec, i, j); - break; - } -#endif - } - -#if 0 - softpipe->run_fs( tri->fp, quad, &tri->outputs ); -#else - { - GLuint attr = softpipe->fp_attr_to_slot[FRAG_ATTRIB_COL0]; - assert(attr); - - memcpy(quad->outputs.color, - exec.attr[attr], - sizeof(quad->outputs.color)); - } -#endif - - qs->next->run(qs->next, quad); -} - - - -struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->run = shade_quad; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_tile_output.c b/src/mesa/pipe/softpipe/sp_tile_output.c deleted file mode 100644 index 4f4421891b..0000000000 --- a/src/mesa/pipe/softpipe/sp_tile_output.c +++ /dev/null @@ -1,103 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "glheader.h" -#include "imports.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_tile.h" - - -static void mask_copy( GLfloat (*dest)[4], - GLfloat (*src)[4], - GLuint mask ) -{ - GLuint i, j; - - for (i = 0; i < 4; i++) { - if (mask & (1<softpipe; - GLuint i; - - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps - = softpipe_surface(softpipe->framebuffer.cbufs[i]); - - if (quad->mask != MASK_ALL) { - GLfloat tmp[4][QUAD_SIZE]; - - /* Yes, we'll probably have a masked write as well, but this is - * how blend will be done at least. - */ - - sps->read_quad_f_swz(sps, quad->x0, quad->y0, tmp); - - mask_copy( tmp, quad->outputs.color, quad->mask ); - - sps->write_quad_f_swz(sps, quad->x0, quad->y0, tmp); - } - else { - sps->write_quad_f_swz(sps, quad->x0, quad->y0, quad->outputs.color); - } - } -} - - -struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->run = output_quad; - - return stage; -} - -- cgit v1.2.3 From 47e970b2cc0f375a7432a6c28f6d9218db9807ce Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 16:14:55 -0600 Subject: s/tile/quad/ --- src/mesa/sources | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index cc675eab39..36317f84ba 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -165,18 +165,18 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_prim_setup.c \ pipe/softpipe/sp_prim_twoside.c \ pipe/softpipe/sp_prim_unfilled.c \ + pipe/softpipe/sp_quad.c \ + pipe/softpipe/sp_quad_blend.c \ + pipe/softpipe/sp_quad_depth_test.c \ + pipe/softpipe/sp_quad_fs.c \ + pipe/softpipe/sp_quad_output.c \ pipe/softpipe/sp_state_blend.c \ pipe/softpipe/sp_state_clip.c \ pipe/softpipe/sp_state_derived.c \ pipe/softpipe/sp_state_fs.c \ pipe/softpipe/sp_state_point.c \ pipe/softpipe/sp_state_setup.c \ - pipe/softpipe/sp_state_surface.c \ - pipe/softpipe/sp_tile.c \ - pipe/softpipe/sp_tile_blend.c \ - pipe/softpipe/sp_tile_depth_test.c \ - pipe/softpipe/sp_tile_fs.c \ - pipe/softpipe/sp_tile_output.c + pipe/softpipe/sp_state_surface.c STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ -- cgit v1.2.3 From 2fd7b30bfef0ca29a02f22eb98ce3d3b34bedb27 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 16:15:09 -0600 Subject: update includes (s/tile/quad/) --- src/mesa/pipe/softpipe/sp_context.h | 2 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 2 +- src/mesa/pipe/softpipe/sp_quad_blend.c | 2 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/pipe/softpipe/sp_quad_output.c | 2 +- 6 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 4e6168fe33..81f6ae3292 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -36,7 +36,7 @@ #include "pipe/p_state.h" #include "pipe/p_context.h" -#include "sp_tile.h" +#include "sp_quad.h" struct softpipe_surface; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 29bb936540..a559b31a52 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -33,7 +33,7 @@ #include "sp_context.h" #include "sp_prim.h" -#include "sp_tile.h" +#include "sp_quad.h" diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 9ac0ccbfa5..86f9fb633c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -37,7 +37,7 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" -#include "sp_tile.h" +#include "sp_quad.h" diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 5bc2648531..756141db17 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -31,7 +31,7 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" -#include "sp_tile.h" +#include "sp_quad.h" diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 912831fefa..02fcc05467 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -36,7 +36,7 @@ #include "imports.h" #include "sp_context.h" #include "sp_headers.h" -#include "sp_tile.h" +#include "sp_quad.h" struct exec_machine { const struct setup_coefficient *coef; diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index 4f4421891b..e898757a06 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -37,7 +37,7 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" -#include "sp_tile.h" +#include "sp_quad.h" static void mask_copy( GLfloat (*dest)[4], -- cgit v1.2.3 From b8657e44dc397625f447c83504f545ef13d9851d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 20 Jun 2007 23:03:26 +0100 Subject: Remove dependency on vf.h in public headers --- src/mesa/pipe/p_state.h | 1 - src/mesa/pipe/softpipe/sp_context.h | 8 +++++--- src/mesa/pipe/softpipe/sp_draw.c | 1 + src/mesa/pipe/softpipe/sp_prim_flatshade.c | 2 ++ src/mesa/pipe/softpipe/sp_prim_twoside.c | 2 ++ src/mesa/pipe/softpipe/sp_state_derived.c | 2 ++ 6 files changed, 12 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index c4bf0d2195..b60e266806 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -29,7 +29,6 @@ #define PIPE_STATE_H #include "mtypes.h" -#include "vf/vf.h" #define WINDING_NONE 0 #define WINDING_CW 1 diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 81f6ae3292..ebe39fa8bf 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -64,6 +64,8 @@ enum interp_mode { #define G_NEW_DEPTH_TEST 0x400 +#define PIPE_ATTRIB_MAX 32 + struct softpipe_context { struct pipe_context pipe; @@ -94,13 +96,13 @@ struct softpipe_context { * For now we just set colors to CONST on flatshade, textures to * perspective always and everything else to linear. */ - enum interp_mode interp[VF_ATTRIB_MAX]; + enum interp_mode interp[PIPE_ATTRIB_MAX]; /* FS + setup derived state: */ - GLuint fp_attr_to_slot[VF_ATTRIB_MAX]; - GLuint vf_attr_to_slot[VF_ATTRIB_MAX]; + GLuint fp_attr_to_slot[PIPE_ATTRIB_MAX]; + GLuint vf_attr_to_slot[PIPE_ATTRIB_MAX]; GLuint nr_attrs; GLuint nr_frag_attrs; GLuint attr_mask; diff --git a/src/mesa/pipe/softpipe/sp_draw.c b/src/mesa/pipe/softpipe/sp_draw.c index 5813e37776..3fc30dd203 100644 --- a/src/mesa/pipe/softpipe/sp_draw.c +++ b/src/mesa/pipe/softpipe/sp_draw.c @@ -33,6 +33,7 @@ #include "imports.h" #include "tnl/t_context.h" +#include "vf/vf.h" #include "sp_context.h" #include "sp_prim.h" diff --git a/src/mesa/pipe/softpipe/sp_prim_flatshade.c b/src/mesa/pipe/softpipe/sp_prim_flatshade.c index 0e9de5faeb..3a7d9de466 100644 --- a/src/mesa/pipe/softpipe/sp_prim_flatshade.c +++ b/src/mesa/pipe/softpipe/sp_prim_flatshade.c @@ -29,6 +29,8 @@ */ #include "imports.h" +#include "vf/vf.h" + #include "sp_context.h" #include "sp_prim.h" diff --git a/src/mesa/pipe/softpipe/sp_prim_twoside.c b/src/mesa/pipe/softpipe/sp_prim_twoside.c index 6eee1de588..8ac0b5c130 100644 --- a/src/mesa/pipe/softpipe/sp_prim_twoside.c +++ b/src/mesa/pipe/softpipe/sp_prim_twoside.c @@ -28,6 +28,8 @@ /* Authors: Keith Whitwell */ #include "imports.h" +#include "vf/vf.h" + #include "sp_context.h" #include "sp_prim.h" diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index b53c842366..8ab325b72d 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -30,6 +30,8 @@ #include "enums.h" #include "program.h" +#include "vf/vf.h" + #include "sp_context.h" #include "sp_draw.h" #include "sp_state.h" -- cgit v1.2.3 From 78cc48f70afee7edc5291e195af8e6b6ebbf2de9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 20 Jun 2007 23:04:00 +0100 Subject: Rename static variable in temporary tnl code --- src/mesa/state_tracker/st_draw.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 24abee132f..a424d1dd05 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -74,7 +74,7 @@ const struct tnl_pipeline_stage st_draw = { draw }; -static const struct tnl_pipeline_stage *intel_pipeline[] = { +static const struct tnl_pipeline_stage *st_pipeline[] = { &_tnl_vertex_transform_stage, &_tnl_vertex_cull_stage, &_tnl_normal_transform_stage, @@ -96,7 +96,7 @@ void st_init_draw( struct st_context *st ) GLcontext *ctx = st->ctx; _tnl_destroy_pipeline( ctx ); - _tnl_install_pipeline( ctx, intel_pipeline ); + _tnl_install_pipeline( ctx, st_pipeline ); } -- cgit v1.2.3 From 24a989c5d1d3a9ba4536d546d20b91c0c3ffa03e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 20 Jun 2007 23:04:56 +0100 Subject: Add state tracker create/destroy calls to i915 driver. --- src/mesa/drivers/dri/Makefile.template | 2 +- src/mesa/drivers/dri/i915tex/intel_context.c | 12 ++++++++++++ 2 files changed, 13 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template index 6f2314ee8c..43c0e912bf 100644 --- a/src/mesa/drivers/dri/Makefile.template +++ b/src/mesa/drivers/dri/Makefile.template @@ -85,7 +85,7 @@ $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) touch depend $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ - $(ASM_SOURCES) 2>&1 /dev/null + $(ASM_SOURCES) 2> /dev/null # Emacs tags diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 20b2b41ef2..c37092337c 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -60,6 +60,10 @@ #include "intel_buffer_objects.h" #include "intel_fbo.h" +#include "pipe/softpipe/sp_context.h" +#include "state_tracker/st_public.h" + + #include "drirenderbuffer.h" #include "vblank.h" #include "utils.h" @@ -244,6 +248,9 @@ intelInvalidateState(GLcontext * ctx, GLuint new_state) _vbo_InvalidateState(ctx, new_state); _tnl_InvalidateState(ctx, new_state); _tnl_invalidate_vertex_state(ctx, new_state); + + st_invalidate_state( ctx, new_state ); + intel_context(ctx)->NewGLState |= new_state; } @@ -493,6 +500,11 @@ intelInitContext(struct intel_context *intel, FALLBACK(intel, INTEL_FALLBACK_USER, 1); } + + st_create_context( &intel->ctx, + softpipe_create() ); + + return GL_TRUE; } -- cgit v1.2.3 From ecfa794037e8be351ecfec0229d1e3b1677ae369 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Jun 2007 17:20:02 -0600 Subject: checkpoint: implement z/depth testing --- src/mesa/drivers/x11/xm_surface.c | 14 +++++ src/mesa/drivers/x11/xm_tri.c | 4 +- src/mesa/drivers/x11/xmesaP.h | 11 +++- src/mesa/pipe/softpipe/sp_context.c | 3 ++ src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_quad.c | 6 +++ src/mesa/pipe/softpipe/sp_quad_depth_test.c | 79 +++++++++++++++++++++++----- src/mesa/pipe/softpipe/sp_state.h | 6 +++ src/mesa/pipe/softpipe/sp_state_blend.c | 26 +++++++++ src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- src/mesa/pipe/softpipe/sp_surface.h | 5 ++ src/mesa/state_tracker/st_atom.c | 1 + src/mesa/state_tracker/st_atom_framebuffer.c | 14 +++++ 13 files changed, 153 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index ee06b77377..a937df3ade 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -241,3 +241,17 @@ xmesa_get_color_surface(GLcontext *ctx, GLuint buf) return (struct pipe_surface *) xrb->pSurface; } + +struct pipe_surface * +xmesa_get_z_surface(GLcontext *ctx, GLuint i) +{ + return NULL; +} + + +struct pipe_surface * +xmesa_get_stencil_surface(GLcontext *ctx, GLuint i) +{ + return NULL; +} + diff --git a/src/mesa/drivers/x11/xm_tri.c b/src/mesa/drivers/x11/xm_tri.c index 6158ef30f9..9f17083f90 100644 --- a/src/mesa/drivers/x11/xm_tri.c +++ b/src/mesa/drivers/x11/xm_tri.c @@ -1443,7 +1443,7 @@ do { \ #endif - +#if 0 GLboolean xmesa_get_cbuf_details( GLcontext *ctx, void **ptr, GLuint *cpp, @@ -1480,7 +1480,7 @@ GLboolean xmesa_get_cbuf_details( GLcontext *ctx, *format = 0; return GL_FALSE; } - +#endif /** diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index c36b0966d9..1d5df3d935 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -580,16 +580,23 @@ extern void xmesa_register_swrast_functions( GLcontext *ctx ); #define ENABLE_EXT_timer_query 0 /* may not have 64-bit GLuint64EXT */ #endif - +#if 0 GLboolean xmesa_get_cbuf_details( GLcontext *ctx, void **ptr, GLuint *cpp, GLint *stride, GLuint *format ); - +#endif struct pipe_surface; struct pipe_surface * xmesa_get_color_surface(GLcontext *ctx, GLuint buf); +struct pipe_surface * +xmesa_get_z_surface(GLcontext *ctx, GLuint i); + +struct pipe_surface * +xmesa_get_stencil_surface(GLcontext *ctx, GLuint i); + + #endif diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 593be0e132..f27d2dd8bc 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -65,9 +65,11 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.destroy = softpipe_destroy; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; + softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_state = softpipe_set_blend_state; softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; + softpipe->pipe.set_depth_state = softpipe_set_depth_test_state; softpipe->pipe.set_point_state = softpipe_set_point_state; softpipe->pipe.set_viewport = softpipe_set_viewport; softpipe->pipe.set_setup_state = softpipe_set_setup_state; @@ -87,6 +89,7 @@ struct pipe_context *softpipe_create( void ) softpipe->prim.cull = prim_cull( softpipe ); softpipe->quad.blend = sp_quad_blend_stage(softpipe); + softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); softpipe->quad.shade = sp_quad_shade_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index ebe39fa8bf..d01fc38b81 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -80,6 +80,7 @@ struct softpipe_context { struct pipe_alpha_test_state alpha_test; struct pipe_clip_state clip; struct pipe_clear_color_state clear_color; + struct pipe_depth_state depth_test; struct pipe_point_state point; struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 168872c64d..32085ab8c4 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -7,6 +7,7 @@ void sp_build_quad_pipeline(struct softpipe_context *sp) { + /* build up the pipeline in reverse order... */ sp->quad.first = sp->quad.output; @@ -15,6 +16,11 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.blend; } + if (sp->depth_test.enabled) { + sp->quad.depth_test->next = sp->quad.first; + sp->quad.first = sp->quad.depth_test; + } + /* XXX always enable shader? */ if (1) { sp->quad.shade->next = sp->quad.first; diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 756141db17..76a6ee1bb6 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -32,29 +32,80 @@ #include "sp_headers.h" #include "sp_surface.h" #include "sp_quad.h" - +#include "pipe/p_defines.h" static void depth_test_quad(struct quad_stage *qs, struct quad_header *quad) { -#if 0 struct softpipe_context *softpipe = qs->softpipe; - GLfloat dest[4][QUAD_SIZE], result[4][QUAD_SIZE]; - GLuint i; - - /* XXX we're also looping in output_quad() !?! */ - - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps - = softpipe_surface(softpipe->framebuffer.cbufs[i]); - - sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + GLuint j; + struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); + GLfloat zzzz[QUAD_SIZE]; /**< Z for four pixels in quad */ - /* XXX do blend here */ +#if 0 + assert(sps); /* shouldn't get here if there's no zbuffer */ +#else + if (!sps) + return; +#endif + /* XXX get zquad from zbuffer */ + sps->read_quad_z(sps, quad->x0, quad->y0, zzzz); + + switch (softpipe->depth_test.func) { + case PIPE_FUNC_NEVER: + quad->mask = 0x0; + break; + case PIPE_FUNC_LESS: + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + if (quad->outputs.depth[j] >= zzzz[j]) { + /* fail */ + quad->mask &= (1 << j); + } + else if (softpipe->depth_test.writemask) { + /* pass, and update Z buffer */ + zzzz[j] = quad->outputs.depth[j]; + } + } + } + break; + case PIPE_FUNC_EQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + if (quad->outputs.depth[j] != zzzz[j]) { + /* fail */ + quad->mask &= (1 << j); + } + else if (softpipe->depth_test.writemask) { + /* pass, and update Z buffer */ + zzzz[j] = quad->outputs.depth[j]; + } + } + } + break; + case PIPE_FUNC_LEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + if (quad->outputs.depth[j] > zzzz[j]) { + /* fail */ + quad->mask &= (1 << j); + } + else if (softpipe->depth_test.writemask) { + /* pass, and update Z buffer */ + zzzz[j] = quad->outputs.depth[j]; + } + } + } + break; + /* XXX fill in remaining cases */ + default: + abort(); } -#endif + + /* XXX write updated zquad to zbuffer */ + sps->write_quad_z(sps, quad->x0, quad->y0, zzzz); qs->next->run(qs->next, quad); } diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 5a657c4a8f..6253b9c9e5 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -38,6 +38,9 @@ void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); +void softpipe_set_alpha_test_state( struct pipe_context *, + const struct pipe_alpha_test_state * ); + void softpipe_set_blend_state( struct pipe_context *, const struct pipe_blend_state * ); @@ -47,6 +50,9 @@ void softpipe_set_clear_color_state( struct pipe_context *, void softpipe_set_clip_state( struct pipe_context *, const struct pipe_clip_state * ); +void softpipe_set_depth_test_state( struct pipe_context *, + const struct pipe_depth_state * ); + void softpipe_set_viewport( struct pipe_context *, const struct pipe_viewport * ); diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 1fd7a44105..c364d8a319 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -45,3 +45,29 @@ void softpipe_set_blend_state( struct pipe_context *pipe, softpipe->dirty |= G_NEW_BLEND; } + +/** XXX move someday? Or consolidate all these simple state setters + * into one file. + */ +void +softpipe_set_depth_test_state(struct pipe_context *pipe, + const struct pipe_depth_state *depth) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->depth_test = *depth; + + softpipe->dirty |= G_NEW_DEPTH_TEST; +} + +void +softpipe_set_alpha_test_state(struct pipe_context *pipe, + const struct pipe_alpha_test_state *alpha) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->alpha_test = *alpha; + + softpipe->dirty |= G_NEW_ALPHA_TEST; +} + diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 8ab325b72d..5aee4be6b8 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -134,7 +134,7 @@ void softpipe_update_derived( struct softpipe_context *softpipe ) if (softpipe->dirty & (G_NEW_SETUP | G_NEW_FS)) calculate_vertex_layout( softpipe ); - if (softpipe->dirty & (G_NEW_BLEND | G_NEW_FS)) + if (softpipe->dirty & (G_NEW_BLEND | G_NEW_DEPTH_TEST | G_NEW_ALPHA_TEST | G_NEW_FS)) sp_build_quad_pipeline(softpipe); softpipe->dirty = 0; diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index fc9557dee3..05b125d17b 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -77,6 +77,11 @@ struct softpipe_surface { void (*write_mono_row_ub)( struct softpipe_surface *, GLuint count, GLint x, GLint y, GLubyte rgba[NUM_CHANNELS] ); + + void (*read_quad_z)(struct softpipe_surface *, + GLint x, GLint y, GLfloat zzzz[QUAD_SIZE]); + void (*write_quad_z)(struct softpipe_surface *, + GLint x, GLint y, const GLfloat zzzz[QUAD_SIZE]); }; diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index e1d187f2db..5fcd9d7af5 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -44,6 +44,7 @@ static const struct st_tracked_state *atoms[] = { &st_update_framebuffer, &st_update_clear_color, + &st_update_depth, &st_update_clip, &st_update_fs, &st_update_point, diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index f203e1df60..8e98cbc2df 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -39,6 +39,12 @@ extern struct pipe_surface * xmesa_get_color_surface(GLcontext *ctx, GLuint i); +extern struct pipe_surface * +xmesa_get_z_surface(GLcontext *ctx, GLuint i); + +extern struct pipe_surface * +xmesa_get_stencil_surface(GLcontext *ctx, GLuint i); + /** * Update framebuffer state (color, depth, stencil, etc. buffers) @@ -58,6 +64,14 @@ update_framebuffer_state( struct st_context *st ) framebuffer.cbufs[i] = xmesa_get_color_surface(st->ctx, i); } + if (st->ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer) { + framebuffer.zbuf = xmesa_get_z_surface(st->ctx, i); + } + + if (st->ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer) { + framebuffer.sbuf = xmesa_get_stencil_surface(st->ctx, i); + } + if (memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0) { st->state.framebuffer = framebuffer; st->pipe->set_framebuffer_state( st->pipe, &framebuffer ); -- cgit v1.2.3 From 493ed9fc11a2bf272a2c1e9e5a072e4f02b46554 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 21 Jun 2007 20:08:10 -0600 Subject: alpha test and misc changes --- src/mesa/pipe/p_state.h | 2 +- src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_quad.c | 5 ++ src/mesa/pipe/softpipe/sp_quad.h | 1 + src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 81 +++++++++++++++++++++++++++++ src/mesa/sources | 1 + src/mesa/state_tracker/st_atom_alphatest.c | 2 +- 7 files changed, 91 insertions(+), 2 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_quad_alpha_test.c (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index b60e266806..8e68fac1bd 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -108,7 +108,7 @@ struct pipe_depth_state }; struct pipe_alpha_test_state { - GLuint enable:1; + GLuint enabled:1; GLuint func:3; /**< PIPE_FUNC_x */ GLfloat ref; /**< reference value */ }; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index d01fc38b81..329544eb74 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -138,6 +138,7 @@ struct softpipe_context { */ struct { struct quad_stage *shade; + struct quad_stage *alpha_test; struct quad_stage *depth_test; struct quad_stage *blend; struct quad_stage *output; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 32085ab8c4..63126a4da3 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -21,6 +21,11 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.depth_test; } + if (sp->alpha_test.enabled) { + sp->quad.alpha_test->next = sp->quad.first; + sp->quad.first = sp->quad.alpha_test; + } + /* XXX always enable shader? */ if (1) { sp->quad.shade->next = sp->quad.first; diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 86c17180bf..351cd4d323 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -46,6 +46,7 @@ struct quad_stage { struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c new file mode 100644 index 0000000000..f7cebca6ea --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -0,0 +1,81 @@ + +/** + * quad alpha test + */ + +#include "glheader.h" +#include "imports.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "pipe/p_defines.h" + + +static void +alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + GLuint j; + struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); + const GLfloat ref = softpipe->alpha_test.ref; + +#if 0 + assert(sps); /* shouldn't get here if there's no zbuffer */ +#else + if (!sps) + return; +#endif + + switch (softpipe->alpha_test.func) { + case PIPE_FUNC_NEVER: + quad->mask = 0x0; + break; + case PIPE_FUNC_LESS: + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + if (quad->outputs.color[3][j] >= ref) { + /* fail */ + quad->mask &= (1 << j); + } + } + } + break; + case PIPE_FUNC_EQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + if (quad->outputs.color[3][j] != ref) { + /* fail */ + quad->mask &= (1 << j); + } + } + } + break; + case PIPE_FUNC_LEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + if (quad->outputs.color[3][j] > ref) { + /* fail */ + quad->mask &= (1 << j); + } + } + } + break; + /* XXX fill in remaining cases */ + default: + abort(); + } + + qs->next->run(qs->next, quad); +} + + +struct quad_stage * +sp_quad_alpha_test_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = alpha_test_quad; + + return stage; +} diff --git a/src/mesa/sources b/src/mesa/sources index 36317f84ba..fd84ed5826 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -166,6 +166,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_prim_twoside.c \ pipe/softpipe/sp_prim_unfilled.c \ pipe/softpipe/sp_quad.c \ + pipe/softpipe/sp_quad_alpha_test.c \ pipe/softpipe/sp_quad_blend.c \ pipe/softpipe/sp_quad_depth_test.c \ pipe/softpipe/sp_quad_fs.c \ diff --git a/src/mesa/state_tracker/st_atom_alphatest.c b/src/mesa/state_tracker/st_atom_alphatest.c index b7cc0d70a9..1e2e449795 100644 --- a/src/mesa/state_tracker/st_atom_alphatest.c +++ b/src/mesa/state_tracker/st_atom_alphatest.c @@ -67,7 +67,7 @@ update_alpha_test( struct st_context *st ) memset(&alpha, 0, sizeof(alpha)); if (st->ctx->Color.AlphaEnabled) { - alpha.enable = 1; + alpha.enabled = 1; alpha.func = gl_alpha_func_to_sp(st->ctx->Color.AlphaFunc); alpha.ref = st->ctx->Color.AlphaRef; } -- cgit v1.2.3 From f79c225d9e5adee6287a9bba35f014c3fe00d3f9 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Jun 2007 12:47:04 -0600 Subject: Assorted token renaming/removal, minor state changes, etc. --- src/mesa/pipe/p_defines.h | 15 +++---- src/mesa/pipe/p_state.h | 55 ++++++++++++------------- src/mesa/pipe/softpipe/sp_prim_cull.c | 5 ++- src/mesa/pipe/softpipe/sp_prim_twoside.c | 3 +- src/mesa/pipe/softpipe/sp_prim_unfilled.c | 12 +++--- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 8 ---- src/mesa/pipe/softpipe/sp_state_setup.c | 5 ++- src/mesa/state_tracker/st_atom_setup.c | 62 ++++++++++++++++++----------- 8 files changed, 84 insertions(+), 81 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 550240e647..75573107ff 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -89,16 +89,13 @@ #define PIPE_POLYGON_MODE_LINE 1 #define PIPE_POLYGON_MODE_POINT 2 -/** Polygon cull mode */ -#define PIPE_POLYGON_CULL_NONE 0 -#define PIPE_POLYGON_CULL_FRONT 1 -#define PIPE_POLYGON_CULL_BACK 2 -#define PIPE_POLYGON_CULL_BOTH 3 - -/** Polygon front winding order */ -#define PIPE_POLYGON_FRONT_CW 0 -#define PIPE_POLYGON_FRONT_CCW 1 +/** Polygon front/back window, also for culling */ +#define PIPE_WINDING_NONE 0 +#define PIPE_WINDING_CW 1 +#define PIPE_WINDING_CCW 2 +#define PIPE_WINDING_BOTH (PIPE_WINDING_CW | PIPE_WINDING_CCW) +/** Stencil ops */ #define PIPE_STENCIL_OP_KEEP 0 #define PIPE_STENCIL_OP_ZERO 1 #define PIPE_STENCIL_OP_REPLACE 2 diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 8e68fac1bd..b841181cf0 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -25,38 +25,43 @@ * **************************************************************************/ + +/** + * Abstract graphics pipe state objects. + * + * Basic notes: + * 1. Want compact representations, so we use bitfields. + * 2. Put bitfields before other (GLfloat) fields. + */ + + #ifndef PIPE_STATE_H #define PIPE_STATE_H #include "mtypes.h" -#define WINDING_NONE 0 -#define WINDING_CW 1 -#define WINDING_CCW 2 -#define WINDING_BOTH (WINDING_CW | WINDING_CCW) -#define FILL_POINT 1 -#define FILL_LINE 2 -#define FILL_TRI 3 - -struct pipe_setup_state { +/** + * Primitive (point/line/tri) setup info + */ +struct pipe_setup_state +{ GLuint flatshade:1; GLuint light_twoside:1; - GLuint front_winding:2; + GLuint front_winding:2; /**< PIPE_WINDING_x */ - GLuint cull_mode:2; + GLuint cull_mode:2; /**< PIPE_WINDING_x */ - GLuint fill_cw:2; - GLuint fill_ccw:2; + GLuint fill_cw:2; /**< PIPE_POLYGON_MODE_x */ + GLuint fill_ccw:2; /**< PIPE_POLYGON_MODE_x */ GLuint offset_cw:1; GLuint offset_ccw:1; GLuint scissor:1; GLuint poly_stipple:1; - - GLuint pad:18; + GLuint poly_smooth:1; GLfloat offset_units; GLfloat offset_scale; @@ -137,6 +142,7 @@ struct pipe_clear_color_state GLfloat color[4]; }; +/** XXXX probably merge into pipe_setup_state */ struct pipe_line_state { GLuint smooth:1; @@ -146,6 +152,7 @@ struct pipe_line_state GLfloat width; }; +/** XXXX probably merge into pipe_setup_state */ struct pipe_point_state { GLuint smooth:1; @@ -154,16 +161,6 @@ struct pipe_point_state GLfloat attenuation[3]; }; -struct pipe_polygon_state { - GLuint cull_mode:2; /**< PIPE_POLYGON_CULL_x */ - GLuint front_winding:1; /**< PIPE_POLYGON_FRONT_CCW,CW */ - GLuint front_mode:2; /**< PIPE_POLYGON_MODE_x */ - GLuint back_mode:2; /**< PIPE_POLYGON_MODE_x */ - GLuint stipple:1; /**< enable */ - GLuint smooth:1; /**< enable */ - /* XXX Polygon offset? */ -}; - struct pipe_stencil_state { GLuint front_enabled:1; GLuint front_func:3; /**< PIPE_FUNC_x */ @@ -215,6 +212,10 @@ struct pipe_sampler_state GLuint wrap_r:3; /**< PIPE_TEX_WRAP_x */ GLuint min_filter:3; /**< PIPE_TEX_FILTER_x */ GLuint mag_filter:1; /**< PIPE_TEX_FILTER_LINEAR or _NEAREST */ + GLuint compare:1; /**< shadow/depth compare enabled? */ + GLenum compare_mode:1; /**< PIPE_TEX_COMPARE_x */ + GLenum compare_func:3; /**< PIPE_FUNC_x */ + GLfloat shadow_ambient; /**< shadow test fail color/intensity */ GLfloat min_lod; GLfloat max_lod; GLfloat lod_bias; @@ -223,10 +224,6 @@ struct pipe_sampler_state GLint MaxLevel; /**< max mipmap level, OpenGL 1.2 */ #endif GLfloat max_anisotropy; - GLuint compare:1; /**< shadow/depth compare enabled? */ - GLenum compare_mode:1; /**< PIPE_TEX_COMPARE_x */ - GLenum compare_func:3; /**< PIPE_FUNC_x */ - GLfloat shadow_ambient; /**< shadow test fail color/intensity */ }; #endif diff --git a/src/mesa/pipe/softpipe/sp_prim_cull.c b/src/mesa/pipe/softpipe/sp_prim_cull.c index 16920d478a..63099fbee0 100644 --- a/src/mesa/pipe/softpipe/sp_prim_cull.c +++ b/src/mesa/pipe/softpipe/sp_prim_cull.c @@ -29,6 +29,7 @@ */ #include "imports.h" +#include "pipe/p_defines.h" #include "sp_context.h" #include "sp_prim.h" @@ -75,8 +76,8 @@ static void cull_tri( struct prim_stage *stage, _mesa_printf("%s %f\n", __FUNCTION__, header->det ); if (header->det != 0) { - GLuint mode = (header->det < 0) ? WINDING_CW : WINDING_CCW; - + GLuint mode = (header->det < 0) ? PIPE_WINDING_CW : PIPE_WINDING_CCW; + if ((mode & cull_stage(stage)->mode) == 0) stage->next->tri( stage->next, header ); } diff --git a/src/mesa/pipe/softpipe/sp_prim_twoside.c b/src/mesa/pipe/softpipe/sp_prim_twoside.c index 8ac0b5c130..5e9f218d1e 100644 --- a/src/mesa/pipe/softpipe/sp_prim_twoside.c +++ b/src/mesa/pipe/softpipe/sp_prim_twoside.c @@ -30,6 +30,7 @@ #include "imports.h" #include "vf/vf.h" +#include "pipe/p_defines.h" #include "sp_context.h" #include "sp_prim.h" @@ -52,7 +53,7 @@ static void twoside_begin( struct prim_stage *stage ) { struct twoside_stage *twoside = twoside_stage(stage); - twoside->facing = (stage->softpipe->setup.front_winding == WINDING_CW) ? -1 : 1; + twoside->facing = (stage->softpipe->setup.front_winding == PIPE_WINDING_CW) ? -1 : 1; stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/softpipe/sp_prim_unfilled.c b/src/mesa/pipe/softpipe/sp_prim_unfilled.c index 73147337fd..ab0dab09d4 100644 --- a/src/mesa/pipe/softpipe/sp_prim_unfilled.c +++ b/src/mesa/pipe/softpipe/sp_prim_unfilled.c @@ -31,7 +31,7 @@ #include "sp_context.h" #include "sp_prim.h" - +#include "pipe/p_defines.h" struct unfilled_stage { @@ -113,17 +113,17 @@ static void unfilled_tri( struct prim_stage *stage, GLuint mode = unfilled->mode[header->det < 0]; switch (mode) { - case FILL_TRI: + case PIPE_POLYGON_MODE_FILL: stage->next->tri( stage->next, header ); break; - - case FILL_LINE: + case PIPE_POLYGON_MODE_LINE: lines( stage, header ); break; - - case GL_POINT: + case PIPE_POLYGON_MODE_POINT: points( stage, header ); break; + default: + abort(); } } diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index f7cebca6ea..0059b3abf4 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -16,16 +16,8 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; GLuint j; - struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); const GLfloat ref = softpipe->alpha_test.ref; -#if 0 - assert(sps); /* shouldn't get here if there's no zbuffer */ -#else - if (!sps) - return; -#endif - switch (softpipe->alpha_test.func) { case PIPE_FUNC_NEVER: quad->mask = 0x0; diff --git a/src/mesa/pipe/softpipe/sp_state_setup.c b/src/mesa/pipe/softpipe/sp_state_setup.c index 7be7d7fe40..a249ceb992 100644 --- a/src/mesa/pipe/softpipe/sp_state_setup.c +++ b/src/mesa/pipe/softpipe/sp_state_setup.c @@ -25,6 +25,7 @@ * **************************************************************************/ +#include "pipe/p_defines.h" #include "sp_context.h" #include "sp_state.h" #include "sp_prim.h" @@ -39,8 +40,8 @@ static void validate_prim_pipe( struct softpipe_context *softpipe ) /* TODO: make the current primitive part of the state and build * shorter pipelines for lines & points. */ - if (softpipe->setup.fill_cw != FILL_TRI || - softpipe->setup.fill_ccw != FILL_TRI) { + if (softpipe->setup.fill_cw != PIPE_POLYGON_MODE_FILL || + softpipe->setup.fill_ccw != PIPE_POLYGON_MODE_FILL) { softpipe->prim.unfilled->next = next; next = softpipe->prim.unfilled; diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index d88d403c1d..fcda2b3a41 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -33,15 +33,21 @@ #include "st_context.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" #include "st_atom.h" static GLuint translate_fill( GLenum mode ) { switch (mode) { - case GL_POINT: return FILL_POINT; - case GL_LINE: return FILL_LINE; - case GL_FILL: return FILL_TRI; - default: assert(0); return 0; + case GL_POINT: + return PIPE_POLYGON_MODE_POINT; + case GL_LINE: + return PIPE_POLYGON_MODE_LINE; + case GL_FILL: + return PIPE_POLYGON_MODE_FILL; + default: + assert(0); + return 0; } } @@ -49,10 +55,15 @@ static GLboolean get_offset_flag( GLuint fill_mode, const struct gl_polygon_attrib *Polygon ) { switch (fill_mode) { - case FILL_POINT: return Polygon->OffsetPoint; - case FILL_LINE: return Polygon->OffsetLine; - case FILL_TRI: return Polygon->OffsetFill; - default: assert(0); return 0; + case PIPE_POLYGON_MODE_POINT: + return Polygon->OffsetPoint; + case PIPE_POLYGON_MODE_LINE: + return Polygon->OffsetLine; + case PIPE_POLYGON_MODE_FILL: + return Polygon->OffsetFill; + default: + assert(0); + return 0; } } @@ -67,13 +78,13 @@ static void update_setup_state( struct st_context *st ) /* _NEW_POLYGON, _NEW_BUFFERS */ { - setup.front_winding = WINDING_CW; + setup.front_winding = PIPE_WINDING_CW; if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) - setup.front_winding ^= WINDING_BOTH; + setup.front_winding ^= PIPE_WINDING_BOTH; if (ctx->Polygon.FrontFace != GL_CCW) - setup.front_winding ^= WINDING_BOTH; + setup.front_winding ^= PIPE_WINDING_BOTH; } /* _NEW_LIGHT @@ -90,18 +101,23 @@ static void update_setup_state( struct st_context *st ) ctx->Light.Model.TwoSide) setup.light_twoside = 1; + if (ctx->Polygon.SmoothFlag) + setup.poly_smooth = 1; + + if (ctx->Polygon.StippleFlag) + setup.poly_stipple = 1; /* _NEW_POLYGON */ if (ctx->Polygon.CullFlag) { if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) { - setup.cull_mode = WINDING_BOTH; + setup.cull_mode = PIPE_WINDING_BOTH; } else if (ctx->Polygon.CullFaceMode == GL_FRONT) { setup.cull_mode = setup.front_winding; } else { - setup.cull_mode = setup.front_winding ^ WINDING_BOTH; + setup.cull_mode = setup.front_winding ^ PIPE_WINDING_BOTH; } } @@ -111,7 +127,7 @@ static void update_setup_state( struct st_context *st ) GLuint fill_front = translate_fill( ctx->Polygon.FrontMode ); GLuint fill_back = translate_fill( ctx->Polygon.BackMode ); - if (setup.front_winding == WINDING_CW) { + if (setup.front_winding == PIPE_WINDING_CW) { setup.fill_cw = fill_front; setup.fill_ccw = fill_back; } @@ -122,11 +138,11 @@ static void update_setup_state( struct st_context *st ) /* Simplify when culling is active: */ - if (setup.cull_mode & WINDING_CW) { + if (setup.cull_mode & PIPE_WINDING_CW) { setup.fill_cw = setup.fill_ccw; } - if (setup.cull_mode & WINDING_CCW) { + if (setup.cull_mode & PIPE_WINDING_CCW) { setup.fill_ccw = setup.fill_cw; } } @@ -136,19 +152,17 @@ static void update_setup_state( struct st_context *st ) * * _NEW_POLYGON */ - if (setup.fill_cw != FILL_TRI) - setup.offset_cw = get_offset_flag( setup.fill_cw, - &ctx->Polygon ); + if (setup.fill_cw != PIPE_POLYGON_MODE_FILL) + setup.offset_cw = get_offset_flag( setup.fill_cw, &ctx->Polygon ); - if (setup.fill_ccw != FILL_TRI) - setup.offset_ccw = get_offset_flag( setup.fill_ccw, - &ctx->Polygon ); + if (setup.fill_ccw != PIPE_POLYGON_MODE_FILL) + setup.offset_ccw = get_offset_flag( setup.fill_ccw, &ctx->Polygon ); /* _NEW_BUFFERS, _NEW_POLYGON */ - if (setup.fill_cw != FILL_TRI || - setup.fill_ccw != FILL_TRI) + if (setup.fill_cw != PIPE_POLYGON_MODE_FILL || + setup.fill_ccw != PIPE_POLYGON_MODE_FILL) { GLfloat mrd = (ctx->DrawBuffer ? ctx->DrawBuffer->_MRD : -- cgit v1.2.3 From 8754b59667f8196eb884113ef1427b0792de47fd Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Jun 2007 12:49:23 -0600 Subject: added blend comments --- src/mesa/pipe/p_state.h | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index b841181cf0..d1e387ce43 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -121,16 +121,16 @@ struct pipe_alpha_test_state { struct pipe_blend_state { GLuint blend_enable:1; - GLuint rgb_func:3; - GLuint rgb_src_factor:5; - GLuint rgb_dst_factor:5; + GLuint rgb_func:3; /**< PIPE_BLEND_x */ + GLuint rgb_src_factor:5; /**< PIPE_BLENDFACTOR_x */ + GLuint rgb_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ - GLuint alpha_func:3; - GLuint alpha_src_factor:5; - GLuint alpha_dst_factor:5; + GLuint alpha_func:3; /**< PIPE_BLEND_x */ + GLuint alpha_src_factor:5; /**< PIPE_BLENDFACTOR_x */ + GLuint alpha_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ GLuint logicop_enable:1; - GLuint logicop_func:4; + GLuint logicop_func:4; /**< PIPE_LOGICOP_x */ }; struct pipe_blend_color { -- cgit v1.2.3 From 13682d959ddacde1ce65843aa8c5b43dc9017b32 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Jun 2007 13:08:49 -0600 Subject: more texture sampler work --- src/mesa/pipe/p_context.h | 4 + src/mesa/pipe/p_state.h | 15 ++-- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_context.h | 2 + src/mesa/pipe/softpipe/sp_state.h | 4 + src/mesa/pipe/softpipe/sp_state_sampler.c | 49 +++++++++++ src/mesa/sources | 2 + src/mesa/state_tracker/st_atom_sampler.c | 135 ++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_context.h | 1 + 9 files changed, 208 insertions(+), 5 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_state_sampler.c create mode 100644 src/mesa/state_tracker/st_atom_sampler.c (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 32444343be..7e68699d3e 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -96,6 +96,10 @@ struct pipe_context { void (*set_stencil_state)( struct pipe_context *, const struct pipe_stencil_state * ); + void (*set_sampler_state)( struct pipe_context *, + GLuint unit, + const struct pipe_sampler_state * ); + void (*set_viewport)( struct pipe_context *, const struct pipe_viewport * ); }; diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index d1e387ce43..f3723eb87e 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -41,6 +41,16 @@ #include "mtypes.h" + +/** + * Implementation limits + */ +#define PIPE_MAX_SAMPLERS 8 +#define PIPE_MAX_CLIP_PLANES 6 +#define PIPE_MAX_CONSTANT 32 + + + /** * Primitive (point/line/tri) setup info */ @@ -84,9 +94,6 @@ struct pipe_scissor_rect { GLshort maxy; }; - -#define PIPE_MAX_CLIP_PLANES 6 - struct pipe_clip_state { GLfloat ucp[PIPE_MAX_CLIP_PLANES][4]; GLuint nr; @@ -96,8 +103,6 @@ struct pipe_fs_state { struct gl_fragment_program *fp; }; -#define PIPE_MAX_CONSTANT 32 - struct pipe_constant_buffer { GLfloat constant[PIPE_MAX_CONSTANT][4]; GLuint nr_constants; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index f27d2dd8bc..07b529fc6e 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -76,6 +76,7 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.set_scissor_rect = softpipe_set_scissor_rect; softpipe->pipe.set_fs_state = softpipe_set_fs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; + softpipe->pipe.set_sampler_state = softpipe_set_sampler_state; softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.clear = softpipe_clear; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 329544eb74..d8c0de0ad1 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -62,6 +62,7 @@ enum interp_mode { #define G_NEW_FRAMEBUFFER 0x100 #define G_NEW_ALPHA_TEST 0x200 #define G_NEW_DEPTH_TEST 0x400 +#define G_NEW_SAMPLER 0x800 #define PIPE_ATTRIB_MAX 32 @@ -84,6 +85,7 @@ struct softpipe_context { struct pipe_point_state point; struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; + struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; GLuint dirty; /* Clip derived state: diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 6253b9c9e5..faaa043a56 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -59,6 +59,10 @@ void softpipe_set_viewport( struct pipe_context *, void softpipe_set_setup_state( struct pipe_context *, const struct pipe_setup_state * ); +void softpipe_set_sampler_state( struct pipe_context *, + GLuint unit, + const struct pipe_sampler_state * ); + void softpipe_set_scissor_rect( struct pipe_context *, const struct pipe_scissor_rect * ); diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c new file mode 100644 index 0000000000..6b422acfb6 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -0,0 +1,49 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: + * Brian Paul + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_draw.h" + + + +void +softpipe_set_sampler_state(struct pipe_context *pipe, + GLuint unit, + const struct pipe_sampler_state *sampler) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->sampler[unit] = *sampler; + + softpipe->dirty |= G_NEW_SAMPLER; +} diff --git a/src/mesa/sources b/src/mesa/sources index fd84ed5826..3887e9f3c2 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -177,6 +177,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_state_fs.c \ pipe/softpipe/sp_state_point.c \ pipe/softpipe/sp_state_setup.c \ + pipe/softpipe/sp_state_sampler.c \ pipe/softpipe/sp_state_surface.c STATETRACKER_SOURCES = \ @@ -189,6 +190,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_fs.c \ state_tracker/st_atom_framebuffer.c \ state_tracker/st_atom_point.c \ + state_tracker/st_atom_sampler.c \ state_tracker/st_atom_scissor.c \ state_tracker/st_atom_stencil.c \ state_tracker/st_atom_setup.c \ diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c new file mode 100644 index 0000000000..1aa9da8484 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -0,0 +1,135 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + + +/** + * Convert GLenum texcoord wrap tokens to pipe tokens. + */ +static GLuint +gl_wrap_to_sp(GLenum wrap) +{ + switch (wrap) { + case GL_REPEAT: + return PIPE_TEX_WRAP_REPEAT; + case GL_CLAMP: + return PIPE_TEX_WRAP_CLAMP; + case GL_CLAMP_TO_EDGE: + return PIPE_TEX_WRAP_CLAMP_TO_EDGE; + case GL_CLAMP_TO_BORDER: + return PIPE_TEX_WRAP_CLAMP_TO_BORDER; + case GL_MIRRORED_REPEAT: + return PIPE_TEX_WRAP_MIRROR_REPEAT; + case GL_MIRROR_CLAMP_EXT: + return PIPE_TEX_WRAP_MIRROR_CLAMP; + case GL_MIRROR_CLAMP_TO_EDGE_EXT: + return PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE; + case GL_MIRROR_CLAMP_TO_BORDER_EXT: + return PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER; + default: + abort(); + return 0; + } +} + + +static GLuint +gl_filter_to_sp(GLenum filter) +{ + switch (filter) { + case GL_NEAREST: + return PIPE_TEX_FILTER_NEAREST; + case GL_LINEAR: + return PIPE_TEX_FILTER_LINEAR; + case GL_NEAREST_MIPMAP_NEAREST: + return PIPE_TEX_FILTER_NEAREST_MIPMAP_NEAREST; + case GL_NEAREST_MIPMAP_LINEAR: + return PIPE_TEX_FILTER_NEAREST_MIPMAP_LINEAR; + case GL_LINEAR_MIPMAP_NEAREST: + return PIPE_TEX_FILTER_LINEAR_MIPMAP_NEAREST; + case GL_LINEAR_MIPMAP_LINEAR: + return PIPE_TEX_FILTER_LINEAR_MIPMAP_LINEAR; + default: + abort(); + return 0; + } +} + + +static void +update_samplers(struct st_context *st) +{ + GLuint u; + + for (u = 0; u < st->ctx->Const.MaxTextureImageUnits; u++) { + const struct gl_texture_object *texobj + = st->ctx->Texture.Unit[u]._Current; + struct pipe_sampler_state sampler; + + memset(&sampler, 0, sizeof(sampler)); + + sampler.wrap_s = gl_wrap_to_sp(texobj->WrapS); + sampler.wrap_t = gl_wrap_to_sp(texobj->WrapT); + sampler.wrap_r = gl_wrap_to_sp(texobj->WrapR); + + sampler.min_filter = gl_filter_to_sp(texobj->MinFilter); + sampler.mag_filter = gl_filter_to_sp(texobj->MagFilter); + + /* XXX more sampler state here */ + + if (memcmp(&sampler, &st->state.sampler[u], sizeof(sampler)) != 0) { + /* state has changed */ + st->state.sampler[u] = sampler; + st->pipe->set_sampler_state(st->pipe, u, &sampler); + } + } +} + + +const struct st_tracked_state st_update_sampler = { + .dirty = { + .mesa = _NEW_TEXTURE, + .st = 0, + }, + .update = update_samplers +}; + + + + + diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 0f1fa2327b..582d586a73 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -75,6 +75,7 @@ struct st_context struct pipe_clip_state clip; struct pipe_depth_state depth; struct pipe_point_state point; + struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; struct pipe_stencil_state stencil; -- cgit v1.2.3 From 5d69aeb0028f44d06093faede5c545908b0df89a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Jun 2007 13:37:47 -0600 Subject: initial texture object, texture format code --- src/mesa/pipe/p_context.h | 4 ++++ src/mesa/pipe/p_defines.h | 16 ++++++++++++++++ src/mesa/pipe/p_state.h | 14 ++++++++++++++ src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_context.h | 2 ++ src/mesa/pipe/softpipe/sp_state.h | 4 ++++ src/mesa/pipe/softpipe/sp_state_sampler.c | 15 +++++++++++++++ 7 files changed, 56 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 7e68699d3e..d9546a24d2 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -100,6 +100,10 @@ struct pipe_context { GLuint unit, const struct pipe_sampler_state * ); + void (*set_texture_state)( struct pipe_context *, + GLuint unit, + struct pipe_texture_object * ); + void (*set_viewport)( struct pipe_context *, const struct pipe_viewport * ); }; diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 75573107ff..a0c4d72a3c 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -124,4 +124,20 @@ #define PIPE_TEX_COMPARE_NONE 0 #define PIPE_TEX_COMPARE_R_TO_TEXTURE 1 +/** + * Texture/surface image formats + */ +#define PIPE_FORMAT_U_R8_G8_B8_A8 0 /**< ubyte[4] RGBA */ +#define PIPE_FORMAT_U_A8_R8_G8_B8 1 /**< ubyte[4] ARGB */ +#define PIPE_FORMAT_U_R5_G6_B5 2 /**< 5/6/5 RGB */ +#define PIPE_FORMAT_U_L8 3 /**< ubyte luminance */ +#define PIPE_FORMAT_U_A8 4 /**< ubyte alpha */ +#define PIPE_FORMAT_U_I8 5 /**< ubyte intensity */ +#define PIPE_FORMAT_U_L8_A8 6 /**< ubyte luminance, alpha */ + +#define PIPE_FORMAT_U_Z16 7 /**< ushort Z/depth */ +#define PIPE_FORMAT_F_Z32 8 /**< float Z/depth */ +#define PIPE_FORMAT_YCBCR 9 +#define PIPE_FORMAT_YCBCR_REV 10 + #endif diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index f3723eb87e..3e3da7cce6 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -231,4 +231,18 @@ struct pipe_sampler_state GLfloat max_anisotropy; }; + +/** + * XXX rough approximation... + */ +struct pipe_texture_object +{ + GLuint format:5; /**< PIPE_FORMAT_x */ + GLuint width:13; /**< 13 bits = 8K max size */ + GLuint height:13; + GLuint depth:13; + GLubyte *data; /**< only valid while buffer mapped? */ +}; + + #endif diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 07b529fc6e..0d00c7eb6c 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -77,6 +77,7 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.set_fs_state = softpipe_set_fs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.set_sampler_state = softpipe_set_sampler_state; + softpipe->pipe.set_texture_state = softpipe_set_texture_state; softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.clear = softpipe_clear; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index d8c0de0ad1..7c816dbc1a 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -63,6 +63,7 @@ enum interp_mode { #define G_NEW_ALPHA_TEST 0x200 #define G_NEW_DEPTH_TEST 0x400 #define G_NEW_SAMPLER 0x800 +#define G_NEW_TEXTURE 0x1000 #define PIPE_ATTRIB_MAX 32 @@ -86,6 +87,7 @@ struct softpipe_context { struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; + struct pipe_texture_object *texture[PIPE_MAX_SAMPLERS]; GLuint dirty; /* Clip derived state: diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index faaa043a56..735d039748 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -63,6 +63,10 @@ void softpipe_set_sampler_state( struct pipe_context *, GLuint unit, const struct pipe_sampler_state * ); +void softpipe_set_texture_state( struct pipe_context *, + GLuint unit, + struct pipe_texture_object * ); + void softpipe_set_scissor_rect( struct pipe_context *, const struct pipe_scissor_rect * ); diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 6b422acfb6..060e2c8c98 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -43,7 +43,22 @@ softpipe_set_sampler_state(struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); + assert(unit < PIPE_MAX_SAMPLERS); softpipe->sampler[unit] = *sampler; softpipe->dirty |= G_NEW_SAMPLER; } + + +void +softpipe_set_texture_state(struct pipe_context *pipe, + GLuint unit, + struct pipe_texture_object *texture) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + assert(unit < PIPE_MAX_SAMPLERS); + softpipe->texture[unit] = texture; /* ptr, not struct */ + + softpipe->dirty |= G_NEW_TEXTURE; +} -- cgit v1.2.3 From c2577bb1a0a2381789759420576af7b4c8acdfeb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 24 Jun 2007 10:32:48 +0100 Subject: Restructure z test code slightly. Make the logic slightly closer to an eventual SSE or SPE implementation. --- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 57 ++++++++++++----------------- 1 file changed, 24 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 76a6ee1bb6..9104b943ec 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -42,6 +42,7 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) GLuint j; struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); GLfloat zzzz[QUAD_SIZE]; /**< Z for four pixels in quad */ + GLuint zmask = 0; #if 0 assert(sps); /* shouldn't get here if there's no zbuffer */ @@ -55,48 +56,25 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) switch (softpipe->depth_test.func) { case PIPE_FUNC_NEVER: - quad->mask = 0x0; break; case PIPE_FUNC_LESS: + /* Note this is pretty much a single sse or cell instruction. + */ for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - if (quad->outputs.depth[j] >= zzzz[j]) { - /* fail */ - quad->mask &= (1 << j); - } - else if (softpipe->depth_test.writemask) { - /* pass, and update Z buffer */ - zzzz[j] = quad->outputs.depth[j]; - } - } + if (quad->outputs.depth[j] < zzzz[j]) + zmask |= 1 << j; } break; case PIPE_FUNC_EQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - if (quad->outputs.depth[j] != zzzz[j]) { - /* fail */ - quad->mask &= (1 << j); - } - else if (softpipe->depth_test.writemask) { - /* pass, and update Z buffer */ - zzzz[j] = quad->outputs.depth[j]; - } - } + if (quad->outputs.depth[j] == zzzz[j]) + zmask |= 1 << j; } break; case PIPE_FUNC_LEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - if (quad->outputs.depth[j] > zzzz[j]) { - /* fail */ - quad->mask &= (1 << j); - } - else if (softpipe->depth_test.writemask) { - /* pass, and update Z buffer */ - zzzz[j] = quad->outputs.depth[j]; - } - } + if (quad->outputs.depth[j] <= zzzz[j]) + zmask |= (1 << j); } break; /* XXX fill in remaining cases */ @@ -104,8 +82,21 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) abort(); } - /* XXX write updated zquad to zbuffer */ - sps->write_quad_z(sps, quad->x0, quad->y0, zzzz); + quad->mask &= zmask; + + if (softpipe->depth_test.writemask) { + + /* This is also efficient with sse / spe instructions: + */ + for (j = 0; j < QUAD_SIZE; j++) { + if (zmask & (1 << j)) { + zzzz[j] = quad->outputs.depth[j]; + } + } + + /* XXX write updated zquad to zbuffer */ + sps->write_quad_z(sps, quad->x0, quad->y0, zzzz); + } qs->next->run(qs->next, quad); } -- cgit v1.2.3 From 8f6d06d037524f012416da5b56889e74bcf09f8f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Jun 2007 09:43:02 -0600 Subject: code for functional Z buffer surface --- src/mesa/drivers/x11/xm_surface.c | 77 ++++++++++++++++++++++++++-- src/mesa/drivers/x11/xmesaP.h | 4 +- src/mesa/state_tracker/st_atom_framebuffer.c | 10 ++-- 3 files changed, 81 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index a937df3ade..30c9049cbf 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -54,6 +54,7 @@ struct xmesa_surface { struct softpipe_surface sps; struct xmesa_renderbuffer *xrb; /** ptr back to matching xmesa_renderbuffer */ + struct gl_renderbuffer *rb; /* ptr to matching gl_renderbuffer */ }; @@ -242,15 +243,85 @@ xmesa_get_color_surface(GLcontext *ctx, GLuint buf) } +static void +read_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, GLfloat zzzz[QUAD_SIZE]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(sps); + struct gl_renderbuffer *rb = xmsurf->rb; + GLushort temp[4]; + GLuint i; + GET_CURRENT_CONTEXT(ctx); + rb->GetRow(ctx, rb, 2, x, y, temp); + rb->GetRow(ctx, rb, 2, x, y + 1, temp + 2); + for (i = 0; i < 4; i++) { + zzzz[i] = USHORT_TO_FLOAT(temp[i]); + } +} + +static void +write_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, const GLfloat zzzz[QUAD_SIZE]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(sps); + struct gl_renderbuffer *rb = xmsurf->rb; + GLushort temp[4]; + GLuint i; + GET_CURRENT_CONTEXT(ctx); + for (i = 0; i < 4; i++) { + CLAMPED_FLOAT_TO_USHORT(temp[i], zzzz[i]); + } + rb->PutRow(ctx, rb, 2, x, y, temp, NULL); + rb->PutRow(ctx, rb, 2, x, y + 1, temp + 2, NULL); +} + + +static struct xmesa_surface * +create_z_surface(XMesaContext xmctx, struct gl_renderbuffer *rb) +{ + struct xmesa_surface *xmsurf; + + xmsurf = CALLOC_STRUCT(xmesa_surface); + if (xmsurf) { + xmsurf->sps.surface.width = rb->Width; + xmsurf->sps.surface.height = rb->Height; + xmsurf->sps.read_quad_z = read_quad_z; + xmsurf->sps.write_quad_z = write_quad_z; + xmsurf->rb = rb; + } + return xmsurf; +} + +/** + * Return a pipe_surface that wraps the current Z/depth buffer. + * XXX this is pretty much a total hack until gl_renderbuffers and + * pipe_surfaces are merged... + */ struct pipe_surface * -xmesa_get_z_surface(GLcontext *ctx, GLuint i) +xmesa_get_z_surface(GLcontext *ctx) { - return NULL; + XMesaContext xmctx = XMESA_CONTEXT(ctx); + struct gl_renderbuffer *rb = ctx->DrawBuffer->_DepthBuffer; + static struct xmesa_surface *xms = NULL; + + if (!rb) + return NULL; + + if (!xms) { + xms = create_z_surface(xmctx, rb); + } + else if (xms->sps.surface.width != rb->Width || + xms->sps.surface.height != rb->Height) { + free_surface(&xms->sps); + xms = create_z_surface(xmctx, rb); + } + + return (struct pipe_surface *) &xms->sps.surface; } struct pipe_surface * -xmesa_get_stencil_surface(GLcontext *ctx, GLuint i) +xmesa_get_stencil_surface(GLcontext *ctx) { return NULL; } diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 1d5df3d935..c0eed4c2a3 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -593,10 +593,10 @@ struct pipe_surface * xmesa_get_color_surface(GLcontext *ctx, GLuint buf); struct pipe_surface * -xmesa_get_z_surface(GLcontext *ctx, GLuint i); +xmesa_get_z_surface(GLcontext *ctx); struct pipe_surface * -xmesa_get_stencil_surface(GLcontext *ctx, GLuint i); +xmesa_get_stencil_surface(GLcontext *ctx); #endif diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 8e98cbc2df..595f390b28 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -40,10 +40,10 @@ extern struct pipe_surface * xmesa_get_color_surface(GLcontext *ctx, GLuint i); extern struct pipe_surface * -xmesa_get_z_surface(GLcontext *ctx, GLuint i); +xmesa_get_z_surface(GLcontext *ctx); extern struct pipe_surface * -xmesa_get_stencil_surface(GLcontext *ctx, GLuint i); +xmesa_get_stencil_surface(GLcontext *ctx); /** @@ -64,12 +64,12 @@ update_framebuffer_state( struct st_context *st ) framebuffer.cbufs[i] = xmesa_get_color_surface(st->ctx, i); } - if (st->ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer) { - framebuffer.zbuf = xmesa_get_z_surface(st->ctx, i); + if (st->ctx->DrawBuffer->_DepthBuffer/*Attachment[BUFFER_DEPTH].Renderbuffer*/) { + framebuffer.zbuf = xmesa_get_z_surface(st->ctx); } if (st->ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer) { - framebuffer.sbuf = xmesa_get_stencil_surface(st->ctx, i); + framebuffer.sbuf = xmesa_get_stencil_surface(st->ctx); } if (memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0) { -- cgit v1.2.3 From 0717daf4f9861647ae486e0231a56df529630c76 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Jun 2007 09:44:28 -0600 Subject: emit quad->outputs.depth[], hack need_z --- src/mesa/pipe/softpipe/sp_quad_fs.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 02fcc05467..b4470ac94a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -125,6 +125,7 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) GLfloat fx = quad->x0; GLfloat fy = quad->y0; GLuint i, j; + GLboolean need_z = softpipe->depth_test.enabled; /* XXX hack */ exec.coef = quad->coef; @@ -143,12 +144,12 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) /* Z and W are done by linear interpolation: * XXX we'll probably have to use integers for Z */ - if (softpipe->need_z) { - linterp(&exec, 0, 2); + if (/*softpipe->*/need_z) { + linterp(&exec, 0, 2); /* attr[0].z */ } if (softpipe->need_w) { - linterp(&exec, 0, 3); + linterp(&exec, 0, 3); /* attr[0].w */ // invert(&exec, 0, 3); } @@ -189,6 +190,13 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) memcpy(quad->outputs.color, exec.attr[attr], sizeof(quad->outputs.color)); + + if (need_z) { + quad->outputs.depth[0] = exec.attr[0][2][0]; + quad->outputs.depth[1] = exec.attr[0][2][1]; + quad->outputs.depth[2] = exec.attr[0][2][2]; + quad->outputs.depth[3] = exec.attr[0][2][3]; + } } #endif -- cgit v1.2.3 From 363e6020429f31d6b923fcd05407d7f3e36bc9ba Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Jun 2007 09:45:10 -0600 Subject: clean-ups --- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 9104b943ec..5229b31016 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -44,14 +44,9 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) GLfloat zzzz[QUAD_SIZE]; /**< Z for four pixels in quad */ GLuint zmask = 0; -#if 0 assert(sps); /* shouldn't get here if there's no zbuffer */ -#else - if (!sps) - return; -#endif - /* XXX get zquad from zbuffer */ + /* get zquad from zbuffer */ sps->read_quad_z(sps, quad->x0, quad->y0, zzzz); switch (softpipe->depth_test.func) { @@ -94,7 +89,7 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } } - /* XXX write updated zquad to zbuffer */ + /* write updated zquad to zbuffer */ sps->write_quad_z(sps, quad->x0, quad->y0, zzzz); } -- cgit v1.2.3 From 192f9ab81b3e2c88812f540c7a5c820e3783ff8a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Jun 2007 09:46:04 -0600 Subject: fix logic error in updating of quad's Z values. Basic depth testing works now --- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 5229b31016..b72bbac0d4 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -84,7 +84,7 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) /* This is also efficient with sse / spe instructions: */ for (j = 0; j < QUAD_SIZE; j++) { - if (zmask & (1 << j)) { + if (quad->mask & (1 << j)) { zzzz[j] = quad->outputs.depth[j]; } } -- cgit v1.2.3 From 1be17dc446aa6b0770d76a3eccf79d0faf6608c0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Jun 2007 17:35:24 -0600 Subject: consolidate point/line state into pipe_setup_state --- src/mesa/pipe/p_context.h | 3 --- src/mesa/pipe/p_state.h | 31 +++++++++++-------------------- src/mesa/pipe/softpipe/sp_context.c | 1 - src/mesa/pipe/softpipe/sp_context.h | 1 - src/mesa/pipe/softpipe/sp_prim_setup.c | 4 ++-- src/mesa/pipe/softpipe/sp_state.h | 3 --- src/mesa/sources | 6 ++---- src/mesa/state_tracker/st_atom.c | 1 - src/mesa/state_tracker/st_atom.h | 1 - src/mesa/state_tracker/st_atom_setup.c | 28 +++++++++++++++++++++------- src/mesa/state_tracker/st_context.h | 1 - 11 files changed, 36 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index d9546a24d2..93dbbe6ab7 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -75,9 +75,6 @@ struct pipe_context { void (*set_depth_state)( struct pipe_context *, const struct pipe_depth_state * ); - void (*set_point_state)( struct pipe_context *, - const struct pipe_point_state * ); - void (*set_framebuffer_state)( struct pipe_context *, const struct pipe_framebuffer_state * ); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 3e3da7cce6..e0fa4d54c3 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -70,9 +70,19 @@ struct pipe_setup_state GLuint offset_ccw:1; GLuint scissor:1; - GLuint poly_stipple:1; + GLuint poly_smooth:1; + GLuint poly_stipple_enable:1; + + GLuint line_smooth:1; + GLuint line_stipple_enable:1; + + GLuint point_smooth:1; + GLubyte line_stipple_factor; /**< [1..255] only */ + GLushort line_stipple_pattern; + GLfloat line_width; + GLfloat point_size; /**< used when no per-vertex size */ GLfloat offset_units; GLfloat offset_scale; }; @@ -147,25 +157,6 @@ struct pipe_clear_color_state GLfloat color[4]; }; -/** XXXX probably merge into pipe_setup_state */ -struct pipe_line_state -{ - GLuint smooth:1; - GLuint stipple:1; - GLushort stipple_pattern; - GLint stipple_factor; - GLfloat width; -}; - -/** XXXX probably merge into pipe_setup_state */ -struct pipe_point_state -{ - GLuint smooth:1; - GLfloat size; - GLfloat min_size, max_size; - GLfloat attenuation[3]; -}; - struct pipe_stencil_state { GLuint front_enabled:1; GLuint front_func:3; /**< PIPE_FUNC_x */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 0d00c7eb6c..7a0aad0973 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -70,7 +70,6 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_depth_state = softpipe_set_depth_test_state; - softpipe->pipe.set_point_state = softpipe_set_point_state; softpipe->pipe.set_viewport = softpipe_set_viewport; softpipe->pipe.set_setup_state = softpipe_set_setup_state; softpipe->pipe.set_scissor_rect = softpipe_set_scissor_rect; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 7c816dbc1a..c38102e2fd 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -83,7 +83,6 @@ struct softpipe_context { struct pipe_clip_state clip; struct pipe_clear_color_state clear_color; struct pipe_depth_state depth_test; - struct pipe_point_state point; struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index a559b31a52..3f4602feb0 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -782,8 +782,8 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); /*XXX this should be a vertex attrib! */ - GLfloat halfSize = 0.5 * setup->stage.softpipe->point.size; - GLboolean round = setup->stage.softpipe->point.smooth; + GLfloat halfSize = 0.5 * setup->stage.softpipe->setup.point_size; + GLboolean round = setup->stage.softpipe->setup.point_smooth; const struct vertex_header *v0 = prim->v[0]; const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 735d039748..4086c16bcc 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -73,9 +73,6 @@ void softpipe_set_scissor_rect( struct pipe_context *, void softpipe_set_fs_state( struct pipe_context *, const struct pipe_fs_state * ); -void softpipe_set_point_state( struct pipe_context *, - const struct pipe_point_state * ); - void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/sources b/src/mesa/sources index 3887e9f3c2..c0c4b35355 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -175,9 +175,8 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_state_clip.c \ pipe/softpipe/sp_state_derived.c \ pipe/softpipe/sp_state_fs.c \ - pipe/softpipe/sp_state_point.c \ - pipe/softpipe/sp_state_setup.c \ pipe/softpipe/sp_state_sampler.c \ + pipe/softpipe/sp_state_setup.c \ pipe/softpipe/sp_state_surface.c STATETRACKER_SOURCES = \ @@ -189,11 +188,10 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_depth.c \ state_tracker/st_atom_fs.c \ state_tracker/st_atom_framebuffer.c \ - state_tracker/st_atom_point.c \ state_tracker/st_atom_sampler.c \ state_tracker/st_atom_scissor.c \ - state_tracker/st_atom_stencil.c \ state_tracker/st_atom_setup.c \ + state_tracker/st_atom_stencil.c \ state_tracker/st_atom_viewport.c \ state_tracker/st_cb_program.c \ state_tracker/st_draw.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 5fcd9d7af5..88c6c776a3 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -47,7 +47,6 @@ static const struct st_tracked_state *atoms[] = &st_update_depth, &st_update_clip, &st_update_fs, - &st_update_point, &st_update_setup, &st_update_viewport, &st_update_scissor, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 8c1219fefd..7ea3301ea5 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -49,7 +49,6 @@ const struct st_tracked_state st_update_clip; const struct st_tracked_state st_update_clear_color; const struct st_tracked_state st_update_depth; const struct st_tracked_state st_update_fs; -const struct st_tracked_state st_update_point; const struct st_tracked_state st_update_setup; const struct st_tracked_state st_update_viewport; const struct st_tracked_state st_update_constants; diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index fcda2b3a41..23adf0b7c8 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -101,12 +101,6 @@ static void update_setup_state( struct st_context *st ) ctx->Light.Model.TwoSide) setup.light_twoside = 1; - if (ctx->Polygon.SmoothFlag) - setup.poly_smooth = 1; - - if (ctx->Polygon.StippleFlag) - setup.poly_stipple = 1; - /* _NEW_POLYGON */ if (ctx->Polygon.CullFlag) { @@ -158,6 +152,12 @@ static void update_setup_state( struct st_context *st ) if (setup.fill_ccw != PIPE_POLYGON_MODE_FILL) setup.offset_ccw = get_offset_flag( setup.fill_ccw, &ctx->Polygon ); + if (ctx->Polygon.SmoothFlag) + setup.poly_smooth = 1; + + if (ctx->Polygon.StippleFlag) + setup.poly_stipple_enable = 1; + /* _NEW_BUFFERS, _NEW_POLYGON */ @@ -173,6 +173,19 @@ static void update_setup_state( struct st_context *st ) st->polygon_offset_scale); } + /* _NEW_POINT + */ + setup.point_size = ctx->Point.Size; + setup.point_smooth = ctx->Point.SmoothFlag; + + /* _NEW_LINE + */ + setup.line_width = ctx->Line.Width; + setup.line_smooth = ctx->Line.SmoothFlag; + setup.line_stipple_enable = ctx->Line.StippleFlag; + setup.line_stipple_pattern = ctx->Line.StipplePattern; + setup.line_stipple_factor = ctx->Line.StippleFactor; + if (memcmp(&setup, &st->state.setup, sizeof(setup)) != 0) { st->state.setup = setup; @@ -182,7 +195,8 @@ static void update_setup_state( struct st_context *st ) const struct st_tracked_state st_update_setup = { .dirty = { - .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_BUFFERS), + .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | + _NEW_POINT | _NEW_BUFFERS), .st = 0, }, .update = update_setup_state diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 582d586a73..48ea7f12ee 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -74,7 +74,6 @@ struct st_context struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_depth_state depth; - struct pipe_point_state point; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_scissor_rect scissor; struct pipe_poly_stipple poly_stipple; -- cgit v1.2.3 From 16fdae16b6c34dd93867e98320152fbaac71139e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 29 Jun 2007 12:16:37 -0600 Subject: s/G_CONTEXT_H/SP_CONTEXT_H/ --- src/mesa/pipe/softpipe/sp_context.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index c38102e2fd..7ba78b8090 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -28,8 +28,8 @@ /* Authors: Keith Whitwell */ -#ifndef G_CONTEXT_H -#define G_CONTEXT_H +#ifndef SP_CONTEXT_H +#define SP_CONTEXT_H #include "glheader.h" @@ -164,4 +164,4 @@ softpipe_context( struct pipe_context *pipe ) } -#endif +#endif /* SP_CONTEXT_H */ -- cgit v1.2.3 From 34cbcf92af902ac96553a0578badac5140417daa Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 29 Jun 2007 12:59:02 -0600 Subject: texture object construction --- src/mesa/sources | 3 +- src/mesa/state_tracker/st_texobj.c | 101 +++++++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_texobj.h | 41 +++++++++++++++ 3 files changed, 144 insertions(+), 1 deletion(-) create mode 100644 src/mesa/state_tracker/st_texobj.c create mode 100644 src/mesa/state_tracker/st_texobj.h (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index c0c4b35355..643e841b15 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -195,7 +195,8 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_viewport.c \ state_tracker/st_cb_program.c \ state_tracker/st_draw.c \ - state_tracker/st_context.c + state_tracker/st_context.c \ + state_tracker/st_texobj.c SHADER_SOURCES = \ shader/arbprogparse.c \ diff --git a/src/mesa/state_tracker/st_texobj.c b/src/mesa/state_tracker/st_texobj.c new file mode 100644 index 0000000000..455e89b6bd --- /dev/null +++ b/src/mesa/state_tracker/st_texobj.c @@ -0,0 +1,101 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * Authors: + * Brian Paul + */ + +#include "imports.h" +#include "texformat.h" + +#include "st_context.h" +#include "st_texobj.h" +#include "pipe/p_defines.h" + + +/** + * Create a pipe_texture_object from a Mesa texture object. + * Eventually, gl_texture_object may be derived from this... + */ +struct pipe_texture_object * +create_texture_object(struct gl_texture_object *texObj) +{ + struct pipe_texture_object *pto; + const struct gl_texture_image *texImage; + + pto = calloc(1, sizeof(*pto)); + if (!pto) + return NULL; + + assert(texObj->Complete); + + switch (texObj->Target) { + case GL_TEXTURE_1D: + pto->type = PIPE_TEXTURE_1D; + break; + case GL_TEXTURE_2D: + pto->type = PIPE_TEXTURE_2D; + break; + case GL_TEXTURE_3D: + pto->type = PIPE_TEXTURE_3D; + break; + case GL_TEXTURE_CUBE_MAP: + pto->type = PIPE_TEXTURE_CUBE; + break; + default: + assert(0); + return NULL; + } + + texImage = texObj->Image[0][texObj->BaseLevel]; + assert(texImage); + + switch (texImage->TexFormat->MesaFormat) { + case MESA_FORMAT_RGBA8888: + pto->format = PIPE_FORMAT_U_R8_G8_B8_A8; + break; + case MESA_FORMAT_RGB565: + pto->format = PIPE_FORMAT_U_R5_G6_B5; + break; + + /* XXX fill in more formats */ + + default: + assert(0); + return NULL; + } + + pto->width = texImage->Width; + pto->height = texImage->Height; + pto->depth = texImage->Depth; + + /* XXX verify this */ + pto->mipmapped = texObj->Image[0][texObj->BaseLevel + 1] != NULL; + + return pto; +} diff --git a/src/mesa/state_tracker/st_texobj.h b/src/mesa/state_tracker/st_texobj.h new file mode 100644 index 0000000000..3c66031018 --- /dev/null +++ b/src/mesa/state_tracker/st_texobj.h @@ -0,0 +1,41 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * Authors: + * Brian Paul + */ + +#ifndef ST_TEXOBJ_H +#define ST_TEXOBJ_H 1 + + +extern struct pipe_texture_object * +create_texture_object(struct gl_texture_object *texObj); + + +#endif /* ST_TEXOBJ_H */ -- cgit v1.2.3 From 48076dc784c910dfee1ebaa6970375495209b839 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 29 Jun 2007 12:59:58 -0600 Subject: added texture types and mapping types --- src/mesa/pipe/p_defines.h | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index a0c4d72a3c..7212040b36 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -134,10 +134,24 @@ #define PIPE_FORMAT_U_A8 4 /**< ubyte alpha */ #define PIPE_FORMAT_U_I8 5 /**< ubyte intensity */ #define PIPE_FORMAT_U_L8_A8 6 /**< ubyte luminance, alpha */ - #define PIPE_FORMAT_U_Z16 7 /**< ushort Z/depth */ #define PIPE_FORMAT_F_Z32 8 /**< float Z/depth */ #define PIPE_FORMAT_YCBCR 9 #define PIPE_FORMAT_YCBCR_REV 10 +/** + * Texture typess + */ +#define PIPE_TEXTURE_1D 0 +#define PIPE_TEXTURE_2D 1 +#define PIPE_TEXTURE_3D 2 +#define PIPE_TEXTURE_CUBE 3 + +/** + * Buffer mapping access modes + */ +#define PIPE_MAP_READ 1 +#define PIPE_MAP_WRITE 2 +#define PIPE_MAP_READ_WRITE 3 + #endif -- cgit v1.2.3 From e62cf5c69ab605b78324639f5455ac9921e68dba Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 29 Jun 2007 13:03:01 -0600 Subject: more work on tex objects, surfaces, mappable buffers --- src/mesa/pipe/p_state.h | 84 ++++++++++++++++++++++++++++++++++++++----------- 1 file changed, 65 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index e0fa4d54c3..6a9a86fa97 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -50,6 +50,14 @@ #define PIPE_MAX_CONSTANT 32 +/* fwd decl */ +struct pipe_surface; + + +/*** + *** State objects + ***/ + /** * Primitive (point/line/tri) setup info @@ -164,29 +172,17 @@ struct pipe_stencil_state { GLuint front_zpass_op:3; /**< PIPE_STENCIL_OP_x */ GLuint front_zfail_op:3; /**< PIPE_STENCIL_OP_x */ GLuint back_enabled:1; - GLuint back_func:3; - GLuint back_fail_op:3; - GLuint back_zpass_op:3; - GLuint back_zfail_op:3; - GLubyte ref_value[2]; /**< [0] = front, [1] = back */ + GLuint back_func:3; /**< PIPE_FUNC_x */ + GLuint back_fail_op:3; /**< PIPE_STENCIL_OP_x */ + GLuint back_zpass_op:3; /**< PIPE_STENCIL_OP_x */ + GLuint back_zfail_op:3; /**< PIPE_STENCIL_OP_x */ + GLubyte ref_value[2]; /**< [0] = front, [1] = back */ GLubyte value_mask[2]; GLubyte write_mask[2]; GLubyte clear_value; }; -/* This will change for hardware pipes... - */ -struct pipe_surface -{ - GLuint width, height; - GLubyte *ptr; - GLint stride; - GLuint cpp; - GLuint format; -}; - - struct pipe_framebuffer_state { GLuint num_cbufs; /**< Number of color bufs to draw to */ @@ -223,16 +219,66 @@ struct pipe_sampler_state }; +/*** + *** Non-state Objects + ***/ + + +/** + * A mappable buffer (vertex data, pixel data, etc) + */ +struct pipe_buffer +{ + void *(*map)(struct pipe_buffer *pb, GLuint access_mode); + void (*unmap)(struct pipe_buffer *pb); + void *ptr; /**< address, only valid while mapped */ + GLuint mode; /**< PIPE_MAP_x, only valid while mapped */ +}; + + +/** + * 2D surface. + * May be a renderbuffer, texture mipmap level, etc. + */ +struct pipe_surface +{ + struct pipe_buffer buffer; /**< surfaces can be mapped */ + GLuint format:5; /**< PIPE_FORMAT_x */ + GLuint width, height; +#if 0 + GLubyte *ptr; + GLint stride; + GLuint cpp; + GLuint format; +#endif +}; + + /** - * XXX rough approximation... + * Texture object. + * Mipmap levels, cube faces, 3D slices can be accessed as surfaces. */ struct pipe_texture_object { + GLuint type:2; /**< PIPE_TEXTURE_x */ GLuint format:5; /**< PIPE_FORMAT_x */ GLuint width:13; /**< 13 bits = 8K max size */ GLuint height:13; GLuint depth:13; - GLubyte *data; /**< only valid while buffer mapped? */ + GLuint mipmapped:1; + + /** to access a 1D or 2D texture object as a surface */ + struct pipe_surface *(*get_2d_surface)(struct pipe_texture_object *pto, + GLuint level); + /** to access a 3D texture object as a surface */ + struct pipe_surface *(*get_3d_surface)(struct pipe_texture_object *pto, + GLuint level, GLuint slice); + /** to access a cube texture object as a surface */ + struct pipe_surface *(*get_cube_surface)(struct pipe_texture_object *pto, + GLuint face, GLuint level); + /** when finished with surface: */ + void (*release_surface)(struct pipe_texture_object *pto, + struct pipe_surface *ps); }; -- cgit v1.2.3 From d523a04853fe9924d8dbf86da08019d17d4f58ef Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 29 Jun 2007 13:10:47 -0600 Subject: added buffer_data(), buffer_sub_data() to pipe_buffer --- src/mesa/pipe/p_state.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 6a9a86fa97..66a53483a3 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -229,6 +229,9 @@ struct pipe_sampler_state */ struct pipe_buffer { + void (*buffer_data)(struct pipe_buffer *pb, GLuint size, const void *src); + void (*buffer_sub_data)(struct pipe_buffer *pb, GLuint offset, GLuint size, + const void *src); void *(*map)(struct pipe_buffer *pb, GLuint access_mode); void (*unmap)(struct pipe_buffer *pb); void *ptr; /**< address, only valid while mapped */ -- cgit v1.2.3 From ea92566ed9cabf5eb5d0993b39c4372d5bfcf3f1 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 2 Jul 2007 15:32:12 -0600 Subject: rename a few structs (use _state suffix consistantly), reorder/sort fields in some structs --- src/mesa/pipe/p_context.h | 8 ++++---- src/mesa/pipe/p_state.h | 4 ++-- src/mesa/pipe/softpipe/sp_context.c | 4 ++-- src/mesa/pipe/softpipe/sp_context.h | 16 ++++++++-------- src/mesa/pipe/softpipe/sp_state.h | 8 ++++---- src/mesa/pipe/softpipe/sp_state_clip.c | 4 ++-- src/mesa/pipe/softpipe/sp_state_setup.c | 4 ++-- src/mesa/state_tracker/st_atom_scissor.c | 4 ++-- src/mesa/state_tracker/st_atom_viewport.c | 4 ++-- src/mesa/state_tracker/st_context.h | 16 ++++++++-------- 10 files changed, 36 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 93dbbe6ab7..b87c42844c 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -87,8 +87,8 @@ struct pipe_context { void (*set_setup_state)( struct pipe_context *, const struct pipe_setup_state * ); - void (*set_scissor_rect)( struct pipe_context *, - const struct pipe_scissor_rect * ); + void (*set_scissor_state)( struct pipe_context *, + const struct pipe_scissor_state * ); void (*set_stencil_state)( struct pipe_context *, const struct pipe_stencil_state * ); @@ -101,8 +101,8 @@ struct pipe_context { GLuint unit, struct pipe_texture_object * ); - void (*set_viewport)( struct pipe_context *, - const struct pipe_viewport * ); + void (*set_viewport_state)( struct pipe_context *, + const struct pipe_viewport_state * ); }; diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 66a53483a3..7f9a6d12de 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -100,12 +100,12 @@ struct pipe_poly_stipple { }; -struct pipe_viewport { +struct pipe_viewport_state { GLfloat scale[4]; GLfloat translate[4]; }; -struct pipe_scissor_rect { +struct pipe_scissor_state { GLshort minx; GLshort miny; GLshort maxx; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 7a0aad0973..5bd652f3a3 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -70,9 +70,9 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_depth_state = softpipe_set_depth_test_state; - softpipe->pipe.set_viewport = softpipe_set_viewport; + softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; softpipe->pipe.set_setup_state = softpipe_set_setup_state; - softpipe->pipe.set_scissor_rect = softpipe_set_scissor_rect; + softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; softpipe->pipe.set_fs_state = softpipe_set_fs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.set_sampler_state = softpipe_set_sampler_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 7ba78b8090..a8c952591e 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -74,19 +74,19 @@ struct softpipe_context { /* The most recent drawing state as set by the driver: */ - struct pipe_framebuffer_state framebuffer; - struct pipe_viewport viewport; - struct pipe_setup_state setup; - struct pipe_fs_state fs; - struct pipe_blend_state blend; struct pipe_alpha_test_state alpha_test; - struct pipe_clip_state clip; - struct pipe_clear_color_state clear_color; + struct pipe_blend_state blend; + struct pipe_clear_color_state clear_color; + struct pipe_clip_state clip; struct pipe_depth_state depth_test; - struct pipe_scissor_rect scissor; + struct pipe_framebuffer_state framebuffer; + struct pipe_fs_state fs; struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; + struct pipe_setup_state setup; struct pipe_texture_object *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; GLuint dirty; /* Clip derived state: diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 4086c16bcc..60c24885e4 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -53,8 +53,8 @@ void softpipe_set_clip_state( struct pipe_context *, void softpipe_set_depth_test_state( struct pipe_context *, const struct pipe_depth_state * ); -void softpipe_set_viewport( struct pipe_context *, - const struct pipe_viewport * ); +void softpipe_set_viewport_state( struct pipe_context *, + const struct pipe_viewport_state * ); void softpipe_set_setup_state( struct pipe_context *, const struct pipe_setup_state * ); @@ -67,8 +67,8 @@ void softpipe_set_texture_state( struct pipe_context *, GLuint unit, struct pipe_texture_object * ); -void softpipe_set_scissor_rect( struct pipe_context *, - const struct pipe_scissor_rect * ); +void softpipe_set_scissor_state( struct pipe_context *, + const struct pipe_scissor_state * ); void softpipe_set_fs_state( struct pipe_context *, const struct pipe_fs_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_clip.c b/src/mesa/pipe/softpipe/sp_state_clip.c index 4b83f2b07e..c907550019 100644 --- a/src/mesa/pipe/softpipe/sp_state_clip.c +++ b/src/mesa/pipe/softpipe/sp_state_clip.c @@ -51,8 +51,8 @@ void softpipe_set_clip_state( struct pipe_context *pipe, /* Called when driver state tracker notices changes to the viewport * matrix: */ -void softpipe_set_viewport( struct pipe_context *pipe, - const struct pipe_viewport *viewport ) +void softpipe_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) { struct softpipe_context *softpipe = softpipe_context(pipe); diff --git a/src/mesa/pipe/softpipe/sp_state_setup.c b/src/mesa/pipe/softpipe/sp_state_setup.c index a249ceb992..55803aeac5 100644 --- a/src/mesa/pipe/softpipe/sp_state_setup.c +++ b/src/mesa/pipe/softpipe/sp_state_setup.c @@ -103,8 +103,8 @@ void softpipe_set_setup_state( struct pipe_context *pipe, -void softpipe_set_scissor_rect( struct pipe_context *pipe, - const struct pipe_scissor_rect *scissor ) +void softpipe_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) { struct softpipe_context *softpipe = softpipe_context(pipe); diff --git a/src/mesa/state_tracker/st_atom_scissor.c b/src/mesa/state_tracker/st_atom_scissor.c index 75bead388f..05a9f3eed1 100644 --- a/src/mesa/state_tracker/st_atom_scissor.c +++ b/src/mesa/state_tracker/st_atom_scissor.c @@ -42,7 +42,7 @@ static void update_scissor( struct st_context *st ) { - struct pipe_scissor_rect scissor; + struct pipe_scissor_state scissor; const struct gl_framebuffer *fb = st->ctx->DrawBuffer; scissor.minx = 0; @@ -69,7 +69,7 @@ update_scissor( struct st_context *st ) if (memcmp(&scissor, &st->state.scissor, sizeof(scissor)) != 0) { /* state has changed */ st->state.scissor = scissor; /* struct copy */ - st->pipe->set_scissor_rect(st->pipe, &scissor); /* activate */ + st->pipe->set_scissor_state(st->pipe, &scissor); /* activate */ } } diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index 765bcf3647..ac91f628aa 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -88,7 +88,7 @@ static void update_viewport( struct st_context *st ) GLfloat half_height = ctx->Viewport.Height / 2.0; GLfloat half_depth = (ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; - struct pipe_viewport vp; + struct pipe_viewport_state vp; vp.scale[0] = half_width; vp.scale[1] = half_height * yScale; @@ -102,7 +102,7 @@ static void update_viewport( struct st_context *st ) if (memcmp(&vp, &st->state.viewport, sizeof(vp)) != 0) { st->state.viewport = vp; - st->pipe->set_viewport(st->pipe, &vp); + st->pipe->set_viewport_state(st->pipe, &vp); } } } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 48ea7f12ee..3b47785a9a 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -65,19 +65,19 @@ struct st_context * though, we just shove random objects across the interface. */ struct { - struct pipe_framebuffer_state framebuffer; - struct pipe_viewport viewport; - struct pipe_setup_state setup; - struct pipe_fs_state fs; struct pipe_alpha_test_state alpha_test; struct pipe_blend_state blend; struct pipe_clear_color_state clear_color; - struct pipe_clip_state clip; - struct pipe_depth_state depth; - struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; - struct pipe_scissor_rect scissor; + struct pipe_clip_state clip; + struct pipe_depth_state depth; + struct pipe_framebuffer_state framebuffer; + struct pipe_fs_state fs; struct pipe_poly_stipple poly_stipple; + struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; + struct pipe_scissor_state scissor; + struct pipe_setup_state setup; struct pipe_stencil_state stencil; + struct pipe_viewport_state viewport; } state; struct { -- cgit v1.2.3 From a813f654872380547faf67fba75fa30792f87ea6 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 3 Jul 2007 10:58:03 -0600 Subject: clarify that line_stipple_factor is in [0,255] corresponding to GL's range of [1,256]. --- src/mesa/pipe/p_state.h | 2 +- src/mesa/state_tracker/st_atom_setup.c | 3 ++- 2 files changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 7f9a6d12de..9b4b336654 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -87,7 +87,7 @@ struct pipe_setup_state GLuint point_smooth:1; - GLubyte line_stipple_factor; /**< [1..255] only */ + GLubyte line_stipple_factor; /**< [1..256] actually */ GLushort line_stipple_pattern; GLfloat line_width; GLfloat point_size; /**< used when no per-vertex size */ diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index 23adf0b7c8..457bfaa9ba 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -184,7 +184,8 @@ static void update_setup_state( struct st_context *st ) setup.line_smooth = ctx->Line.SmoothFlag; setup.line_stipple_enable = ctx->Line.StippleFlag; setup.line_stipple_pattern = ctx->Line.StipplePattern; - setup.line_stipple_factor = ctx->Line.StippleFactor; + /* GL stipple factor is in [1,256], remap to [0, 255] here */ + setup.line_stipple_factor = ctx->Line.StippleFactor - 1; if (memcmp(&setup, &st->state.setup, sizeof(setup)) != 0) { -- cgit v1.2.3 From f6d4f5e1e9d626356311e1868b35b79d355af739 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 3 Jul 2007 17:12:11 -0600 Subject: Rewrite blending in terms of SIMD operations. --- src/mesa/pipe/softpipe/sp_quad_blend.c | 431 +++++++++++++++++++++++++-------- 1 file changed, 324 insertions(+), 107 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 86f9fb633c..2599e7a2f9 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -40,130 +40,347 @@ #include "sp_quad.h" +#define VEC4_COPY(DST, SRC) \ +do { \ + DST[0] = SRC[0]; \ + DST[1] = SRC[1]; \ + DST[2] = SRC[2]; \ + DST[3] = SRC[3]; \ +} while(0) + +#define VEC4_SCALAR(DST, SRC) \ +do { \ + DST[0] = SRC; \ + DST[1] = SRC; \ + DST[2] = SRC; \ + DST[3] = SRC; \ +} while(0) + +#define VEC4_ADD(SUM, A, B) \ +do { \ + SUM[0] = A[0] + B[0]; \ + SUM[1] = A[1] + B[1]; \ + SUM[2] = A[2] + B[2]; \ + SUM[3] = A[3] + B[3]; \ +} while (0) + +#define VEC4_SUB(SUM, A, B) \ +do { \ + SUM[0] = A[0] - B[0]; \ + SUM[1] = A[1] - B[1]; \ + SUM[2] = A[2] - B[2]; \ + SUM[3] = A[3] - B[3]; \ +} while (0) + +#define VEC4_MUL(SUM, A, B) \ +do { \ + SUM[0] = A[0] * B[0]; \ + SUM[1] = A[1] * B[1]; \ + SUM[2] = A[2] * B[2]; \ + SUM[3] = A[3] * B[3]; \ +} while (0) + +#define VEC4_MIN(SUM, A, B) \ +do { \ + SUM[0] = (A[0] < B[0]) ? A[0] : B[0]; \ + SUM[1] = (A[1] < B[1]) ? A[1] : B[1]; \ + SUM[2] = (A[2] < B[2]) ? A[2] : B[2]; \ + SUM[3] = (A[3] < B[3]) ? A[3] : B[3]; \ +} while (0) + +#define VEC4_MAX(SUM, A, B) \ +do { \ + SUM[0] = (A[0] > B[0]) ? A[0] : B[0]; \ + SUM[1] = (A[1] > B[1]) ? A[1] : B[1]; \ + SUM[2] = (A[2] > B[2]) ? A[2] : B[2]; \ + SUM[3] = (A[3] > B[3]) ? A[3] : B[3]; \ +} while (0) + + static void blend_quad(struct quad_stage *qs, struct quad_header *quad) { + static const GLfloat zero[4] = { 0, 0, 0, 0 }; + static const GLfloat one[4] = { 1, 1, 1, 1 }; struct softpipe_context *softpipe = qs->softpipe; GLfloat source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; - GLuint i, j; + GLuint i; /* XXX we're also looping in output_quad() !?! */ - /* copy quad's colors since we'll modify them in the loop */ - memcpy(source, quad->outputs.color, 4 * 4 * sizeof(GLfloat)); - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - GLfloat srcTerm[4], dstTerm[4]; struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.cbufs[i]); sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); - /* XXX this loop could be factored out - we'd compute the src/dstTerm - * for all four pixels in the quad at once. + /* + * Compute src/first term RGB + */ + switch (softpipe->blend.rgb_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[0], quad->outputs.color[0]); /* R */ + VEC4_COPY(source[1], quad->outputs.color[1]); /* G */ + VEC4_COPY(source[2], quad->outputs.color[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + VEC4_MUL(source[0], quad->outputs.color[0], quad->outputs.color[0]); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], quad->outputs.color[1]); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], quad->outputs.color[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const GLfloat *alpha = quad->outputs.color[3]; + VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_DST_COLOR: + VEC4_MUL(source[0], quad->outputs.color[0], dest[0]); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], dest[1]); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], dest[2]); /* B */ + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + { + const GLfloat *alpha = dest[3]; + VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + { + GLfloat comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ + VEC4_MUL(source[0], quad->outputs.color[0], comp); /* R */ + VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ + VEC4_MUL(source[1], quad->outputs.color[1], comp); /* G */ + VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ + VEC4_MUL(source[2], quad->outputs.color[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + GLfloat alpha[4]; + VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_SRC1_ALPHA: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[0], zero); /* R */ + VEC4_COPY(source[1], zero); /* G */ + VEC4_COPY(source[2], zero); /* B */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + { + GLfloat inv_comp[4]; + VEC4_SUB(inv_comp, one, quad->outputs.color[0]); /* R */ + VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, quad->outputs.color[1]); /* G */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, quad->outputs.color[2]); /* B */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + GLfloat inv_alpha[4]; + VEC4_SUB(inv_alpha, one, quad->outputs.color[3]); + VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + GLfloat inv_alpha[4]; + VEC4_SUB(inv_alpha, one, dest[3]); + VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + { + GLfloat inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[0]); /* R */ + VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, dest[1]); /* G */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, dest[2]); /* B */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + { + GLfloat inv_comp[4]; + /* R */ + VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[0]); + VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); + /* G */ + VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[1]); + VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); + /* B */ + VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[2]); + VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + GLfloat alpha[4], inv_alpha[4]; + VEC4_SCALAR(alpha, 1.0 - softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + assert(0); /* to do */ + break; + default: + abort(); + } + + /* + * Compute src/first term A + */ + switch (softpipe->blend.alpha_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[3], quad->outputs.color[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const GLfloat *alpha = quad->outputs.color[3]; + VEC4_MUL(source[3], quad->outputs.color[3], alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[3], zero); /* A */ + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + + /* + * Compute dest/second term RGB + */ + switch (softpipe->blend.rgb_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + GLfloat one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); + VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ + VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ + VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[0], zero); /* R */ + VEC4_COPY(dest[1], zero); /* G */ + VEC4_COPY(dest[2], zero); /* B */ + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + /* + * Compute dest/second term A */ - for (j = 0; j < QUAD_SIZE; j++) { - - /* if this pixel in the quad is alive */ - if (quad->mask & (1 << j)) { - - switch (softpipe->blend.rgb_src_factor) { - case PIPE_BLENDFACTOR_ONE: - srcTerm[0] = srcTerm[1] = srcTerm[2] = 1.0; - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - srcTerm[0] = srcTerm[1] = srcTerm[2] = quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - srcTerm[0] = srcTerm[1] = srcTerm[2] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.alpha_src_factor) { - case PIPE_BLENDFACTOR_ONE: - srcTerm[3] = 1.0; - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - srcTerm[3] = quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - srcTerm[3] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.rgb_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0; - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - dstTerm[0] = dstTerm[1] = dstTerm[2] = 1.0 - quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - dstTerm[0] = dstTerm[1] = dstTerm[2] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.alpha_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - dstTerm[3] = 1.0; - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - dstTerm[3] = 1.0 - quad->outputs.color[3][j]; - break; - case PIPE_BLENDFACTOR_ZERO: - dstTerm[3] = 0.0; - break; - /* XXX fill in remaining terms */ - default: - abort(); - } - - switch (softpipe->blend.rgb_func) { - case PIPE_BLEND_ADD: - quad->outputs.color[0][j] = source[0][j] * srcTerm[0] + dest[0][j] * dstTerm[0]; - quad->outputs.color[1][j] = source[1][j] * srcTerm[1] + dest[1][j] * dstTerm[1]; - quad->outputs.color[2][j] = source[2][j] * srcTerm[2] + dest[2][j] * dstTerm[2]; - quad->outputs.color[3][j] = source[3][j] * srcTerm[3] + dest[3][j] * dstTerm[3]; - break; - case PIPE_BLEND_SUBTRACT: - quad->outputs.color[0][j] = source[0][j] * srcTerm[0] - dest[0][j] * dstTerm[0]; - quad->outputs.color[1][j] = source[1][j] * srcTerm[1] - dest[1][j] * dstTerm[1]; - quad->outputs.color[2][j] = source[2][j] * srcTerm[2] - dest[2][j] * dstTerm[2]; - quad->outputs.color[3][j] = source[3][j] * srcTerm[3] - dest[3][j] * dstTerm[3]; - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - quad->outputs.color[0][j] = dest[0][j] * dstTerm[0] - source[0][j] * srcTerm[0]; - quad->outputs.color[1][j] = dest[1][j] * dstTerm[1] - source[1][j] * srcTerm[1]; - quad->outputs.color[2][j] = dest[2][j] * dstTerm[2] - source[2][j] * srcTerm[2]; - quad->outputs.color[3][j] = dest[3][j] * dstTerm[3] - source[3][j] * srcTerm[3]; - break; - case PIPE_BLEND_MIN: - quad->outputs.color[0][j] = MIN2(dest[0][j], source[0][j]); - quad->outputs.color[1][j] = MIN2(dest[1][j], source[1][j]); - quad->outputs.color[2][j] = MIN2(dest[2][j], source[2][j]); - quad->outputs.color[3][j] = MIN2(dest[3][j], source[3][j]); - break; - case PIPE_BLEND_MAX: - quad->outputs.color[0][j] = MAX2(dest[0][j], source[0][j]); - quad->outputs.color[1][j] = MAX2(dest[1][j], source[1][j]); - quad->outputs.color[2][j] = MAX2(dest[2][j], source[2][j]); - quad->outputs.color[3][j] = MAX2(dest[3][j], source[3][j]); - break; - default: - abort(); - } + switch (softpipe->blend.alpha_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + GLfloat one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); + VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ } - } /* loop over quad's pixels*/ + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[3], zero); /* A */ + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + /* + * Combine RGB terms + */ + switch (softpipe->blend.rgb_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quad->outputs.color[0], source[0], dest[0]); /* R */ + VEC4_ADD(quad->outputs.color[1], source[1], dest[1]); /* G */ + VEC4_ADD(quad->outputs.color[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quad->outputs.color[0], source[0], dest[0]); /* R */ + VEC4_SUB(quad->outputs.color[1], source[1], dest[1]); /* G */ + VEC4_SUB(quad->outputs.color[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quad->outputs.color[0], dest[0], source[0]); /* R */ + VEC4_SUB(quad->outputs.color[1], dest[1], source[1]); /* G */ + VEC4_SUB(quad->outputs.color[2], dest[2], source[2]); /* B */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quad->outputs.color[0], source[0], dest[0]); /* R */ + VEC4_MIN(quad->outputs.color[1], source[1], dest[1]); /* G */ + VEC4_MIN(quad->outputs.color[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quad->outputs.color[0], source[0], dest[0]); /* R */ + VEC4_MAX(quad->outputs.color[1], source[1], dest[1]); /* G */ + VEC4_MAX(quad->outputs.color[2], source[2], dest[2]); /* B */ + break; + default: + abort(); + } + + /* + * Combine A terms + */ + switch (softpipe->blend.alpha_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quad->outputs.color[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quad->outputs.color[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quad->outputs.color[3], dest[3], source[3]); /* A */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quad->outputs.color[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quad->outputs.color[3], source[3], dest[3]); /* A */ + default: + abort(); + } /* pass blended quad to next stage */ qs->next->run(qs->next, quad); -- cgit v1.2.3 From 111880798a7b67b79af3f7986e3297268c4b778a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 3 Jul 2007 17:13:04 -0600 Subject: Blend MIN/MAX modes are special: the src/dest terms are always one. --- src/mesa/state_tracker/st_atom_blend.c | 24 ++++++++++++++++++++---- 1 file changed, 20 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index 03fbd290e0..e0215c9eef 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -169,12 +169,28 @@ update_blend( struct st_context *st ) blend.blend_enable = 1; blend.rgb_func = gl_blend_to_sp(st->ctx->Color.BlendEquationRGB); - blend.rgb_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcRGB); - blend.rgb_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstRGB); + if (st->ctx->Color.BlendEquationRGB == GL_MIN || + st->ctx->Color.BlendEquationRGB == GL_MAX) { + /* Min/max are special */ + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ONE; + } + else { + blend.rgb_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcRGB); + blend.rgb_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstRGB); + } blend.alpha_func = gl_blend_to_sp(st->ctx->Color.BlendEquationA); - blend.alpha_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcA); - blend.alpha_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstA); + if (st->ctx->Color.BlendEquationA == GL_MIN || + st->ctx->Color.BlendEquationA == GL_MAX) { + /* Min/max are special */ + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ONE; + } + else { + blend.alpha_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcA); + blend.alpha_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstA); + } } else { /* no blending / logicop */ -- cgit v1.2.3 From d378a2c0b87dfc0bee0e8310312fec566148cb63 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 3 Jul 2007 17:13:45 -0600 Subject: added blend_color --- src/mesa/pipe/softpipe/sp_context.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index a8c952591e..40688208e0 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -76,6 +76,7 @@ struct softpipe_context { */ struct pipe_alpha_test_state alpha_test; struct pipe_blend_state blend; + struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_depth_state depth_test; -- cgit v1.2.3 From 284efcfc27aaeb447e0898ae3342e9fee20c584a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 3 Jul 2007 17:19:30 -0600 Subject: hook in state tracking for blend color --- src/mesa/pipe/p_context.h | 3 +++ src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_state.h | 3 +++ src/mesa/pipe/softpipe/sp_state_blend.c | 11 +++++++++++ src/mesa/state_tracker/st_atom_blend.c | 9 +++++++++ src/mesa/state_tracker/st_context.h | 1 + 6 files changed, 28 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index b87c42844c..e4115226e7 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -66,6 +66,9 @@ struct pipe_context { void (*set_blend_state)( struct pipe_context *, const struct pipe_blend_state * ); + void (*set_blend_color)( struct pipe_context *, + const struct pipe_blend_color * ); + void (*set_clip_state)( struct pipe_context *, const struct pipe_clip_state * ); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 5bd652f3a3..6bd1d9f16e 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -67,6 +67,7 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_state = softpipe_set_blend_state; + softpipe->pipe.set_blend_color = softpipe_set_blend_color; softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_depth_state = softpipe_set_depth_test_state; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 60c24885e4..38a212b79a 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -44,6 +44,9 @@ void softpipe_set_alpha_test_state( struct pipe_context *, void softpipe_set_blend_state( struct pipe_context *, const struct pipe_blend_state * ); +void softpipe_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ); + void softpipe_set_clear_color_state( struct pipe_context *, const struct pipe_clear_color_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index c364d8a319..6ca42e8a01 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -46,6 +46,17 @@ void softpipe_set_blend_state( struct pipe_context *pipe, } +void softpipe_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->blend_color = *blend_color; + + softpipe->dirty |= G_NEW_BLEND; +} + + /** XXX move someday? Or consolidate all these simple state setters * into one file. */ diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index e0215c9eef..3e5410cfab 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -201,6 +201,15 @@ update_blend( struct st_context *st ) st->state.blend = blend; /* struct copy */ st->pipe->set_blend_state(st->pipe, &blend); /* set new state */ } + + if (memcmp(st->ctx->Color.BlendColor, &st->state.blend_color, 4 * sizeof(GLfloat)) != 0) { + /* state has changed */ + st->state.blend_color.color[0] = st->ctx->Color.BlendColor[0]; + st->state.blend_color.color[1] = st->ctx->Color.BlendColor[1]; + st->state.blend_color.color[2] = st->ctx->Color.BlendColor[2]; + st->state.blend_color.color[3] = st->ctx->Color.BlendColor[3]; + st->pipe->set_blend_color(st->pipe, (struct pipe_blend_color *) st->ctx->Color.BlendColor); + } } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 3b47785a9a..c1d868604c 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -67,6 +67,7 @@ struct st_context struct { struct pipe_alpha_test_state alpha_test; struct pipe_blend_state blend; + struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_depth_state depth; -- cgit v1.2.3 From ac5523a9c31efbb89e58ef99da47b0d81a21409c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 3 Jul 2007 17:31:59 -0600 Subject: SIMD comments --- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 4 ++++ src/mesa/pipe/softpipe/sp_quad_depth_test.c | 1 + 2 files changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index 0059b3abf4..e0f7225d74 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -23,6 +23,10 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) quad->mask = 0x0; break; case PIPE_FUNC_LESS: + /* + * If quad->mask were an array [4] we could do this SIMD-style: + * quad->mask &= (quad->outputs.color[3] <= vec4(ref)); + */ for (j = 0; j < QUAD_SIZE; j++) { if (quad->mask & (1 << j)) { if (quad->outputs.color[3][j] >= ref) { diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index b72bbac0d4..fcd6a22e1d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -54,6 +54,7 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_FUNC_LESS: /* Note this is pretty much a single sse or cell instruction. + * Like this: quad->mask &= (quad->outputs.depth < zzzz); */ for (j = 0; j < QUAD_SIZE; j++) { if (quad->outputs.depth[j] < zzzz[j]) -- cgit v1.2.3 From ffe58739da9eee2e99682747cc8f26e412c87430 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 5 Jul 2007 19:55:38 +0200 Subject: Softpipe: import TGSI tree. Not hooked-up yet. --- src/mesa/pipe/tgsi/Makefile | 3 + src/mesa/pipe/tgsi/core/Makefile | 3 + src/mesa/pipe/tgsi/core/tgsi_build.c | 1315 +++++++++++++++++++ src/mesa/pipe/tgsi/core/tgsi_build.h | 309 +++++ src/mesa/pipe/tgsi/core/tgsi_core.h | 12 + src/mesa/pipe/tgsi/core/tgsi_dump.c | 871 +++++++++++++ src/mesa/pipe/tgsi/core/tgsi_dump.h | 22 + src/mesa/pipe/tgsi/core/tgsi_exec.c | 2193 ++++++++++++++++++++++++++++++++ src/mesa/pipe/tgsi/core/tgsi_exec.h | 137 ++ src/mesa/pipe/tgsi/core/tgsi_parse.c | 284 +++++ src/mesa/pipe/tgsi/core/tgsi_parse.h | 120 ++ src/mesa/pipe/tgsi/core/tgsi_token.h | 1045 +++++++++++++++ src/mesa/pipe/tgsi/core/tgsi_util.c | 263 ++++ src/mesa/pipe/tgsi/core/tgsi_util.h | 70 + src/mesa/pipe/tgsi/mesa/Makefile | 3 + src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 499 ++++++++ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 25 + src/mesa/pipe/tgsi/mesa/tgsi_mesa.h | 8 + src/mesa/pipe/tgsi/tgsi_platform.h | 18 + src/mesa/sources | 19 +- src/mesa/state_tracker/st_cb_program.c | 14 +- src/mesa/state_tracker/st_texobj.c | 3 +- 22 files changed, 7229 insertions(+), 7 deletions(-) create mode 100644 src/mesa/pipe/tgsi/Makefile create mode 100644 src/mesa/pipe/tgsi/core/Makefile create mode 100644 src/mesa/pipe/tgsi/core/tgsi_build.c create mode 100644 src/mesa/pipe/tgsi/core/tgsi_build.h create mode 100644 src/mesa/pipe/tgsi/core/tgsi_core.h create mode 100644 src/mesa/pipe/tgsi/core/tgsi_dump.c create mode 100644 src/mesa/pipe/tgsi/core/tgsi_dump.h create mode 100644 src/mesa/pipe/tgsi/core/tgsi_exec.c create mode 100644 src/mesa/pipe/tgsi/core/tgsi_exec.h create mode 100644 src/mesa/pipe/tgsi/core/tgsi_parse.c create mode 100644 src/mesa/pipe/tgsi/core/tgsi_parse.h create mode 100644 src/mesa/pipe/tgsi/core/tgsi_token.h create mode 100644 src/mesa/pipe/tgsi/core/tgsi_util.c create mode 100644 src/mesa/pipe/tgsi/core/tgsi_util.h create mode 100644 src/mesa/pipe/tgsi/mesa/Makefile create mode 100644 src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c create mode 100644 src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h create mode 100644 src/mesa/pipe/tgsi/mesa/tgsi_mesa.h create mode 100644 src/mesa/pipe/tgsi/tgsi_platform.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/Makefile b/src/mesa/pipe/tgsi/Makefile new file mode 100644 index 0000000000..12a8bd0409 --- /dev/null +++ b/src/mesa/pipe/tgsi/Makefile @@ -0,0 +1,3 @@ +default: + cd ../.. ; make + diff --git a/src/mesa/pipe/tgsi/core/Makefile b/src/mesa/pipe/tgsi/core/Makefile new file mode 100644 index 0000000000..eb8b14e0e8 --- /dev/null +++ b/src/mesa/pipe/tgsi/core/Makefile @@ -0,0 +1,3 @@ +default: + cd ../../.. ; make + diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.c b/src/mesa/pipe/tgsi/core/tgsi_build.c new file mode 100644 index 0000000000..2a482a78dc --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_build.c @@ -0,0 +1,1315 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ) +{ + struct tgsi_version version; + + version.MajorVersion = 1; + version.MinorVersion = 1; + version.Padding = 0; + + return version; +} + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ) +{ + struct tgsi_header header; + + header.HeaderSize = 1; + header.BodySize = 0; + + return header; +} + +static void +header_headersize_grow( struct tgsi_header *header ) +{ + assert (header->HeaderSize < 0xFF); + assert (header->BodySize == 0); + + header->HeaderSize++; +} + +static void +header_bodysize_grow( struct tgsi_header *header ) +{ + assert (header->BodySize < 0xFFFFFF); + + header->BodySize++; +} + +struct tgsi_processor +tgsi_default_processor( void ) +{ + struct tgsi_processor processor; + + processor.Processor = TGSI_PROCESSOR_FRAGMENT; + processor.Padding = 0; + + return processor; +} + +struct tgsi_processor +tgsi_build_processor( + GLuint type, + struct tgsi_header *header ) +{ + struct tgsi_processor processor; + + processor = tgsi_default_processor(); + processor.Processor = type; + + header_headersize_grow( header ); + + return processor; +} + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ) +{ + struct tgsi_declaration declaration; + + declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; + declaration.Size = 1; + declaration.File = TGSI_FILE_NULL; + declaration.Declare = TGSI_DECLARE_RANGE; + declaration.Interpolate = 0; + declaration.Padding = 0; + declaration.Extended = 0; + + return declaration; +} + +struct tgsi_declaration +tgsi_build_declaration( + GLuint file, + GLuint declare, + GLuint interpolate, + struct tgsi_header *header ) +{ + struct tgsi_declaration declaration; + + assert (file <= TGSI_FILE_IMMEDIATE); + assert (declare <= TGSI_DECLARE_MASK); + + declaration = tgsi_default_declaration(); + declaration.File = file; + declaration.Declare = declare; + declaration.Interpolate = interpolate; + + header_bodysize_grow( header ); + + return declaration; +} + +static void +declaration_grow( + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + assert (declaration->Size < 0xFF); + + declaration->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ) +{ + struct tgsi_full_declaration full_declaration; + + full_declaration.Declaration = tgsi_default_declaration(); + full_declaration.Interpolation = tgsi_default_declaration_interpolation(); + + return full_declaration; +} + +GLuint +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + GLuint maxsize ) +{ + GLuint size = 0; + struct tgsi_declaration *declaration; + + if( maxsize <= size ) + return 0; + declaration = (struct tgsi_declaration *) &tokens[size]; + size++; + + *declaration = tgsi_build_declaration( + full_decl->Declaration.File, + full_decl->Declaration.Declare, + full_decl->Declaration.Interpolate, + header ); + + switch( full_decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + { + struct tgsi_declaration_range *dr; + + if( maxsize <= size ) + return 0; + dr = (struct tgsi_declaration_range *) &tokens[size]; + size++; + + *dr = tgsi_build_declaration_range( + full_decl->u.DeclarationRange.First, + full_decl->u.DeclarationRange.Last, + declaration, + header ); + break; + } + + case TGSI_DECLARE_MASK: + { + struct tgsi_declaration_mask *dm; + + if( maxsize <= size ) + return 0; + dm = (struct tgsi_declaration_mask *) &tokens[size]; + size++; + + *dm = tgsi_build_declaration_mask( + full_decl->u.DeclarationMask.Mask, + declaration, + header ); + break; + } + + default: + assert( 0 ); + } + + if( full_decl->Declaration.Interpolate ) { + struct tgsi_declaration_interpolation *di; + + if( maxsize <= size ) + return 0; + di = (struct tgsi_declaration_interpolation *) &tokens[size]; + size++; + + *di = tgsi_build_declaration_interpolation( + full_decl->Interpolation.Interpolate, + declaration, + header ); + } + + return size; +} + +struct tgsi_declaration_range +tgsi_build_declaration_range( + GLuint first, + GLuint last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_range declaration_range; + + assert (last >= first); + assert (last <= 0xFFFF); + + declaration_range.First = first; + declaration_range.Last = last; + + declaration_grow( declaration, header ); + + return declaration_range; +} + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + GLuint mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_mask declaration_mask; + + declaration_mask.Mask = mask; + + declaration_grow( declaration, header ); + + return declaration_mask; +} + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ) +{ + struct tgsi_declaration_interpolation di; + + di.Interpolate = TGSI_INTERPOLATE_CONSTANT; + di.Padding = 0; + + return di; +} + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + GLuint interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_interpolation di; + + assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); + + di = tgsi_default_declaration_interpolation(); + di.Interpolate = interpolate; + + declaration_grow( declaration, header ); + + return di; +} + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ) +{ + struct tgsi_immediate immediate; + + immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + immediate.Size = 1; + immediate.DataType = TGSI_IMM_FLOAT32; + immediate.Padding = 0; + immediate.Extended = 0; + + return immediate; +} + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ) +{ + struct tgsi_immediate immediate; + + immediate = tgsi_default_immediate(); + + header_bodysize_grow( header ); + + return immediate; +} + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ) +{ + struct tgsi_full_immediate fullimm; + + fullimm.Immediate = tgsi_default_immediate(); + fullimm.u.Pointer = (void *) 0; + + return fullimm; +} + +static void +immediate_grow( + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + assert( immediate->Size < 0xFF ); + + immediate->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + GLfloat value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + struct tgsi_immediate_float32 immediate_float32; + + immediate_float32.Float = value; + + immediate_grow( immediate, header ); + + return immediate_float32; +} + +GLuint +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + GLuint maxsize ) +{ + GLuint size = 0, i; + struct tgsi_immediate *immediate; + + if( maxsize <= size ) + return 0; + immediate = (struct tgsi_immediate *) &tokens[size]; + size++; + + *immediate = tgsi_build_immediate( header ); + + for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { + struct tgsi_immediate_float32 *if32; + + if( maxsize <= size ) + return 0; + if32 = (struct tgsi_immediate_float32 *) &tokens[size]; + size++; + + *if32 = tgsi_build_immediate_float32( + full_imm->u.ImmediateFloat32[i].Float, + immediate, + header ); + } + + return size; +} + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ) +{ + struct tgsi_instruction instruction; + + instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; + instruction.Size = 1; + instruction.Opcode = TGSI_OPCODE_MOV; + instruction.Saturate = TGSI_SAT_NONE; + instruction.NumDstRegs = 1; + instruction.NumSrcRegs = 1; + instruction.Padding = 0; + instruction.Extended = 0; + + return instruction; +} + +struct tgsi_instruction +tgsi_build_instruction( + GLuint opcode, + GLuint saturate, + GLuint num_dst_regs, + GLuint num_src_regs, + struct tgsi_header *header ) +{ + struct tgsi_instruction instruction; + + assert (opcode <= TGSI_OPCODE_LAST); + assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); + assert (num_dst_regs <= 3); + assert (num_src_regs <= 15); + + instruction = tgsi_default_instruction(); + instruction.Opcode = opcode; + instruction.Saturate = saturate; + instruction.NumDstRegs = num_dst_regs; + instruction.NumSrcRegs = num_src_regs; + + header_bodysize_grow( header ); + + return instruction; +} + +static void +instruction_grow( + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + assert (instruction->Size < 0xFF); + + instruction->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ) +{ + struct tgsi_full_instruction full_instruction; + GLuint i; + + full_instruction.Instruction = tgsi_default_instruction(); + full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); + full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); + full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); + for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { + full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); + } + for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { + full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); + } + + return full_instruction; +} + +GLuint +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + GLuint maxsize ) +{ + GLuint size = 0; + GLuint i; + struct tgsi_instruction *instruction; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + instruction = (struct tgsi_instruction *) &tokens[size]; + size++; + + *instruction = tgsi_build_instruction( + full_inst->Instruction.Opcode, + full_inst->Instruction.Saturate, + full_inst->Instruction.NumDstRegs, + full_inst->Instruction.NumSrcRegs, + header ); + prev_token = (struct tgsi_token *) instruction; + + if( tgsi_compare_instruction_ext_nv( + full_inst->InstructionExtNv, + tgsi_default_instruction_ext_nv() ) ) { + struct tgsi_instruction_ext_nv *instruction_ext_nv; + + if( maxsize <= size ) + return 0; + instruction_ext_nv = + (struct tgsi_instruction_ext_nv *) &tokens[size]; + size++; + + *instruction_ext_nv = tgsi_build_instruction_ext_nv( + full_inst->InstructionExtNv.Precision, + full_inst->InstructionExtNv.CondDstIndex, + full_inst->InstructionExtNv.CondFlowIndex, + full_inst->InstructionExtNv.CondMask, + full_inst->InstructionExtNv.CondSwizzleX, + full_inst->InstructionExtNv.CondSwizzleY, + full_inst->InstructionExtNv.CondSwizzleZ, + full_inst->InstructionExtNv.CondSwizzleW, + full_inst->InstructionExtNv.CondDstUpdate, + full_inst->InstructionExtNv.CondFlowEnable, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_nv; + } + + if( tgsi_compare_instruction_ext_label( + full_inst->InstructionExtLabel, + tgsi_default_instruction_ext_label() ) ) { + struct tgsi_instruction_ext_label *instruction_ext_label; + + if( maxsize <= size ) + return 0; + instruction_ext_label = + (struct tgsi_instruction_ext_label *) &tokens[size]; + size++; + + *instruction_ext_label = tgsi_build_instruction_ext_label( + full_inst->InstructionExtLabel.Label, + full_inst->InstructionExtLabel.Target, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_label; + } + + if( tgsi_compare_instruction_ext_texture( + full_inst->InstructionExtTexture, + tgsi_default_instruction_ext_texture() ) ) { + struct tgsi_instruction_ext_texture *instruction_ext_texture; + + if( maxsize <= size ) + return 0; + instruction_ext_texture = + (struct tgsi_instruction_ext_texture *) &tokens[size]; + size++; + + *instruction_ext_texture = tgsi_build_instruction_ext_texture( + full_inst->InstructionExtTexture.Texture, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_texture; + } + + for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { + const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; + struct tgsi_dst_register *dst_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + dst_register = (struct tgsi_dst_register *) &tokens[size]; + size++; + + *dst_register = tgsi_build_dst_register( + reg->DstRegister.File, + reg->DstRegister.WriteMask, + reg->DstRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register; + + if( tgsi_compare_dst_register_ext_concode( + reg->DstRegisterExtConcode, + tgsi_default_dst_register_ext_concode() ) ) { + struct tgsi_dst_register_ext_concode *dst_register_ext_concode; + + if( maxsize <= size ) + return 0; + dst_register_ext_concode = + (struct tgsi_dst_register_ext_concode *) &tokens[size]; + size++; + + *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( + reg->DstRegisterExtConcode.CondMask, + reg->DstRegisterExtConcode.CondSwizzleX, + reg->DstRegisterExtConcode.CondSwizzleY, + reg->DstRegisterExtConcode.CondSwizzleZ, + reg->DstRegisterExtConcode.CondSwizzleW, + reg->DstRegisterExtConcode.CondSrcIndex, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_concode; + } + + if( tgsi_compare_dst_register_ext_modulate( + reg->DstRegisterExtModulate, + tgsi_default_dst_register_ext_modulate() ) ) { + struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; + + if( maxsize <= size ) + return 0; + dst_register_ext_modulate = + (struct tgsi_dst_register_ext_modulate *) &tokens[size]; + size++; + + *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( + reg->DstRegisterExtModulate.Modulate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_modulate; + } + } + + for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { + const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; + struct tgsi_src_register *src_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + src_register = (struct tgsi_src_register *) &tokens[size]; + size++; + + *src_register = tgsi_build_src_register( + reg->SrcRegister.File, + reg->SrcRegister.SwizzleX, + reg->SrcRegister.SwizzleY, + reg->SrcRegister.SwizzleZ, + reg->SrcRegister.SwizzleW, + reg->SrcRegister.Negate, + reg->SrcRegister.Indirect, + reg->SrcRegister.Dimension, + reg->SrcRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register; + + if( tgsi_compare_src_register_ext_swz( + reg->SrcRegisterExtSwz, + tgsi_default_src_register_ext_swz() ) ) { + struct tgsi_src_register_ext_swz *src_register_ext_swz; + + if( maxsize <= size ) + return 0; + src_register_ext_swz = + (struct tgsi_src_register_ext_swz *) &tokens[size]; + size++; + + *src_register_ext_swz = tgsi_build_src_register_ext_swz( + reg->SrcRegisterExtSwz.ExtSwizzleX, + reg->SrcRegisterExtSwz.ExtSwizzleY, + reg->SrcRegisterExtSwz.ExtSwizzleZ, + reg->SrcRegisterExtSwz.ExtSwizzleW, + reg->SrcRegisterExtSwz.NegateX, + reg->SrcRegisterExtSwz.NegateY, + reg->SrcRegisterExtSwz.NegateZ, + reg->SrcRegisterExtSwz.NegateW, + reg->SrcRegisterExtSwz.ExtDivide, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_swz; + } + + if( tgsi_compare_src_register_ext_mod( + reg->SrcRegisterExtMod, + tgsi_default_src_register_ext_mod() ) ) { + struct tgsi_src_register_ext_mod *src_register_ext_mod; + + if( maxsize <= size ) + return 0; + src_register_ext_mod = + (struct tgsi_src_register_ext_mod *) &tokens[size]; + size++; + + *src_register_ext_mod = tgsi_build_src_register_ext_mod( + reg->SrcRegisterExtMod.Complement, + reg->SrcRegisterExtMod.Bias, + reg->SrcRegisterExtMod.Scale2X, + reg->SrcRegisterExtMod.Absolute, + reg->SrcRegisterExtMod.Negate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_mod; + } + + if( reg->SrcRegister.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterInd.File, + reg->SrcRegisterInd.SwizzleX, + reg->SrcRegisterInd.SwizzleY, + reg->SrcRegisterInd.SwizzleZ, + reg->SrcRegisterInd.SwizzleW, + reg->SrcRegisterInd.Negate, + reg->SrcRegisterInd.Indirect, + reg->SrcRegisterInd.Dimension, + reg->SrcRegisterInd.Index, + instruction, + header ); + } + + if( reg->SrcRegister.Dimension ) { + struct tgsi_dimension *dim; + + assert( !reg->SrcRegisterDim.Dimension ); + + if( maxsize <= size ) + return 0; + dim = (struct tgsi_dimension *) &tokens[size]; + size++; + + *dim = tgsi_build_dimension( + reg->SrcRegisterDim.Indirect, + reg->SrcRegisterDim.Index, + instruction, + header ); + + if( reg->SrcRegisterDim.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterDimInd.File, + reg->SrcRegisterDimInd.SwizzleX, + reg->SrcRegisterDimInd.SwizzleY, + reg->SrcRegisterDimInd.SwizzleZ, + reg->SrcRegisterDimInd.SwizzleW, + reg->SrcRegisterDimInd.Negate, + reg->SrcRegisterDimInd.Indirect, + reg->SrcRegisterDimInd.Dimension, + reg->SrcRegisterDimInd.Index, + instruction, + header ); + } + } + } + + return size; +} + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; + instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; + instruction_ext_nv.CondDstIndex = 0; + instruction_ext_nv.CondFlowIndex = 0; + instruction_ext_nv.CondMask = TGSI_CC_TR; + instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; + instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; + instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; + instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; + instruction_ext_nv.CondDstUpdate = 0; + instruction_ext_nv.CondFlowEnable = 0; + instruction_ext_nv.Padding = 0; + instruction_ext_nv.Extended = 0; + + return instruction_ext_nv; +} + +union token_u32 +{ + GLuint u32; +}; + +GLuint +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + GLuint precision, + GLuint cond_dst_index, + GLuint cond_flow_index, + GLuint cond_mask, + GLuint cond_swizzle_x, + GLuint cond_swizzle_y, + GLuint cond_swizzle_z, + GLuint cond_swizzle_w, + GLuint cond_dst_update, + GLuint cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv = tgsi_default_instruction_ext_nv(); + instruction_ext_nv.Precision = precision; + instruction_ext_nv.CondDstIndex = cond_dst_index; + instruction_ext_nv.CondFlowIndex = cond_flow_index; + instruction_ext_nv.CondMask = cond_mask; + instruction_ext_nv.CondSwizzleX = cond_swizzle_x; + instruction_ext_nv.CondSwizzleY = cond_swizzle_y; + instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; + instruction_ext_nv.CondSwizzleW = cond_swizzle_w; + instruction_ext_nv.CondDstUpdate = cond_dst_update; + instruction_ext_nv.CondFlowEnable = cond_flow_update; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_nv; +} + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; + instruction_ext_label.Label = 0; + instruction_ext_label.Target = 0; + instruction_ext_label.Padding = 0; + instruction_ext_label.Extended = 0; + + return instruction_ext_label; +} + +GLuint +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + GLuint label, + GLuint target, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label = tgsi_default_instruction_ext_label(); + instruction_ext_label.Label = label; + instruction_ext_label.Target = target; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_label; +} + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; + instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; + instruction_ext_texture.Padding = 0; + instruction_ext_texture.Extended = 0; + + return instruction_ext_texture; +} + +GLuint +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + GLuint texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture = tgsi_default_instruction_ext_texture(); + instruction_ext_texture.Texture = texture; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_texture; +} + +struct tgsi_src_register +tgsi_default_src_register( void ) +{ + struct tgsi_src_register src_register; + + src_register.File = TGSI_FILE_NULL; + src_register.SwizzleX = TGSI_SWIZZLE_X; + src_register.SwizzleY = TGSI_SWIZZLE_Y; + src_register.SwizzleZ = TGSI_SWIZZLE_Z; + src_register.SwizzleW = TGSI_SWIZZLE_W; + src_register.Negate = 0; + src_register.Indirect = 0; + src_register.Dimension = 0; + src_register.Index = 0; + src_register.Extended = 0; + + return src_register; +} + +struct tgsi_src_register +tgsi_build_src_register( + GLuint file, + GLuint swizzle_x, + GLuint swizzle_y, + GLuint swizzle_z, + GLuint swizzle_w, + GLuint negate, + GLuint indirect, + GLuint dimension, + GLint index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register src_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( negate <= 1 ); + assert( index >= -0x8000 && index <= 0x7FFF ); + + src_register = tgsi_default_src_register(); + src_register.File = file; + src_register.SwizzleX = swizzle_x; + src_register.SwizzleY = swizzle_y; + src_register.SwizzleZ = swizzle_z; + src_register.SwizzleW = swizzle_w; + src_register.Negate = negate; + src_register.Indirect = indirect; + src_register.Dimension = dimension; + src_register.Index = index; + + instruction_grow( instruction, header ); + + return src_register; +} + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ) +{ + struct tgsi_full_src_register full_src_register; + + full_src_register.SrcRegister = tgsi_default_src_register(); + full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); + full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); + full_src_register.SrcRegisterInd = tgsi_default_src_register(); + full_src_register.SrcRegisterDim = tgsi_default_dimension(); + full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); + + return full_src_register; +} + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; + src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; + src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; + src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; + src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; + src_register_ext_swz.NegateX = 0; + src_register_ext_swz.NegateY = 0; + src_register_ext_swz.NegateZ = 0; + src_register_ext_swz.NegateW = 0; + src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; + src_register_ext_swz.Padding = 0; + src_register_ext_swz.Extended = 0; + + return src_register_ext_swz; +} + +GLuint +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + GLuint ext_swizzle_x, + GLuint ext_swizzle_y, + GLuint ext_swizzle_z, + GLuint ext_swizzle_w, + GLuint negate_x, + GLuint negate_y, + GLuint negate_z, + GLuint negate_w, + GLuint ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + assert (ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE); + assert (ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE); + assert (ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE); + assert (ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE); + assert (negate_x <= 1); + assert (negate_y <= 1); + assert (negate_z <= 1); + assert (negate_w <= 1); + assert (ext_divide <= TGSI_EXTSWIZZLE_ONE); + + src_register_ext_swz = tgsi_default_src_register_ext_swz(); + src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; + src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; + src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; + src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; + src_register_ext_swz.NegateX = negate_x; + src_register_ext_swz.NegateY = negate_y; + src_register_ext_swz.NegateZ = negate_z; + src_register_ext_swz.NegateW = negate_w; + src_register_ext_swz.ExtDivide = ext_divide; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_swz; +} + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; + src_register_ext_mod.Complement = 0; + src_register_ext_mod.Bias = 0; + src_register_ext_mod.Scale2X = 0; + src_register_ext_mod.Absolute = 0; + src_register_ext_mod.Negate = 0; + src_register_ext_mod.Padding = 0; + src_register_ext_mod.Extended = 0; + + return src_register_ext_mod; +} + +GLuint +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + GLuint complement, + GLuint bias, + GLuint scale_2x, + GLuint absolute, + GLuint negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + assert (complement <= 1); + assert (bias <= 1); + assert (scale_2x <= 1); + assert (absolute <= 1); + assert (negate <= 1); + + src_register_ext_mod = tgsi_default_src_register_ext_mod(); + src_register_ext_mod.Complement = complement; + src_register_ext_mod.Bias = bias; + src_register_ext_mod.Scale2X = scale_2x; + src_register_ext_mod.Absolute = absolute; + src_register_ext_mod.Negate = negate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_mod; +} + +struct tgsi_dimension +tgsi_default_dimension( void ) +{ + struct tgsi_dimension dimension; + + dimension.Indirect = 0; + dimension.Dimension = 0; + dimension.Padding = 0; + dimension.Index = 0; + dimension.Extended = 0; + + return dimension; +} + +struct tgsi_dimension +tgsi_build_dimension( + GLuint indirect, + GLuint index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dimension dimension; + + dimension = tgsi_default_dimension(); + dimension.Indirect = indirect; + dimension.Index = index; + + instruction_grow( instruction, header ); + + return dimension; +} + +struct tgsi_dst_register +tgsi_default_dst_register( void ) +{ + struct tgsi_dst_register dst_register; + + dst_register.File = TGSI_FILE_NULL; + dst_register.WriteMask = TGSI_WRITEMASK_XYZW; + dst_register.Indirect = 0; + dst_register.Dimension = 0; + dst_register.Index = 0; + dst_register.Padding = 0; + dst_register.Extended = 0; + + return dst_register; +} + +struct tgsi_dst_register +tgsi_build_dst_register( + GLuint file, + GLuint mask, + GLint index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register dst_register; + + assert (file <= TGSI_FILE_IMMEDIATE); + assert (mask <= TGSI_WRITEMASK_XYZW); + assert (index >= -32768 && index <= 32767); + + dst_register = tgsi_default_dst_register(); + dst_register.File = file; + dst_register.WriteMask = mask; + dst_register.Index = index; + + instruction_grow( instruction, header ); + + return dst_register; +} + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ) +{ + struct tgsi_full_dst_register full_dst_register; + + full_dst_register.DstRegister = tgsi_default_dst_register(); + full_dst_register.DstRegisterExtConcode = + tgsi_default_dst_register_ext_concode(); + full_dst_register.DstRegisterExtModulate = + tgsi_default_dst_register_ext_modulate(); + + return full_dst_register; +} + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; + dst_register_ext_concode.CondMask = TGSI_CC_TR; + dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; + dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; + dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; + dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; + dst_register_ext_concode.CondSrcIndex = 0; + dst_register_ext_concode.Padding = 0; + dst_register_ext_concode.Extended = 0; + + return dst_register_ext_concode; +} + +GLuint +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + GLuint cc, + GLuint swizzle_x, + GLuint swizzle_y, + GLuint swizzle_z, + GLuint swizzle_w, + GLint index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + assert (cc <= TGSI_CC_FL); + assert (swizzle_x <= TGSI_SWIZZLE_W); + assert (swizzle_y <= TGSI_SWIZZLE_W); + assert (swizzle_z <= TGSI_SWIZZLE_W); + assert (swizzle_w <= TGSI_SWIZZLE_W); + assert (index >= -32768 && index <= 32767); + + dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); + dst_register_ext_concode.CondMask = cc; + dst_register_ext_concode.CondSwizzleX = swizzle_x; + dst_register_ext_concode.CondSwizzleY = swizzle_y; + dst_register_ext_concode.CondSwizzleZ = swizzle_z; + dst_register_ext_concode.CondSwizzleW = swizzle_w; + dst_register_ext_concode.CondSrcIndex = index; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_concode; +} + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; + dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; + dst_register_ext_modulate.Padding = 0; + dst_register_ext_modulate.Extended = 0; + + return dst_register_ext_modulate; +} + +GLuint +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + GLuint modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + assert (modulate <= TGSI_MODULATE_EIGHTH); + + dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); + dst_register_ext_modulate.Modulate = modulate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_modulate; +} + diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.h b/src/mesa/pipe/tgsi/core/tgsi_build.h new file mode 100644 index 0000000000..db25956519 --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_build.h @@ -0,0 +1,309 @@ +#if !defined TGSI_BUILD_H +#define TGSI_BUILD_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ); + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ); + +struct tgsi_processor +tgsi_default_processor( void ); + +struct tgsi_processor +tgsi_build_processor( + GLuint processor, + struct tgsi_header *header ); + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ); + +struct tgsi_declaration +tgsi_build_declaration( + GLuint file, + GLuint declare, + GLuint interpolate, + struct tgsi_header *header ); + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ); + +GLuint +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + GLuint maxsize ); + +struct tgsi_declaration_range +tgsi_build_declaration_range( + GLuint first, + GLuint last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + GLuint mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ); + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + GLuint interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ); + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ); + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ); + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + GLfloat value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ); + +GLuint +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + GLuint maxsize ); + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ); + +struct tgsi_instruction +tgsi_build_instruction( + GLuint opcode, + GLuint saturate, + GLuint num_dst_regs, + GLuint num_src_regs, + struct tgsi_header *header ); + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ); + +GLuint +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + GLuint maxsize ); + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ); + +GLuint +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ); + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + GLuint precision, + GLuint cond_dst_index, + GLuint cond_flow_index, + GLuint cond_mask, + GLuint cond_swizzle_x, + GLuint cond_swizzle_y, + GLuint cond_swizzle_z, + GLuint cond_swizzle_w, + GLuint cond_dst_update, + GLuint cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ); + +GLuint +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ); + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + GLuint label, + GLuint target, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ); + +GLuint +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ); + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + GLuint texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register +tgsi_default_src_register( void ); + +struct tgsi_src_register +tgsi_build_src_register( + GLuint file, + GLuint swizzle_x, + GLuint swizzle_y, + GLuint swizzle_z, + GLuint swizzle_w, + GLuint negate, + GLuint indirect, + GLuint dimension, + GLint index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ); + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ); + +GLuint +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ); + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + GLuint ext_swizzle_x, + GLuint ext_swizzle_y, + GLuint ext_swizzle_z, + GLuint ext_swizzle_w, + GLuint negate_x, + GLuint negate_y, + GLuint negate_z, + GLuint negate_w, + GLuint ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ); + +GLuint +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ); + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + GLuint complement, + GLuint bias, + GLuint scale_2x, + GLuint absolute, + GLuint negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dimension +tgsi_default_dimension( void ); + +struct tgsi_dimension +tgsi_build_dimension( + GLuint indirect, + GLuint index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register +tgsi_default_dst_register( void ); + +struct tgsi_dst_register +tgsi_build_dst_register( + GLuint file, + GLuint mask, + GLint index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ); + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ); + +GLuint +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ); + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + GLuint cc, + GLuint swizzle_x, + GLuint swizzle_y, + GLuint swizzle_z, + GLuint swizzle_w, + GLint index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ); + +GLuint +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ); + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + GLuint modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_BUILD_H + diff --git a/src/mesa/pipe/tgsi/core/tgsi_core.h b/src/mesa/pipe/tgsi/core/tgsi_core.h new file mode 100644 index 0000000000..1f5f00a38e --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_core.h @@ -0,0 +1,12 @@ +#if !defined TGSI_CORE_H +#define TGSI_CORE_H + +#include "tgsi_token.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" +#include "tgsi_exec.h" +#include "tgsi_dump.h" +#include "tgsi_util.h" + +#endif // !defined TGSI_CORE_H + diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c new file mode 100644 index 0000000000..fecb246ab1 --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -0,0 +1,871 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +struct text_dump +{ + FILE *file; + GLuint tabs; +}; + +static void +text_dump_write( + struct text_dump *dump, + const void *buffer, + GLuint size ) +{ + fwrite( buffer, size, 1, dump->file ); +} + +static void +text_dump_str( + struct text_dump *dump, + const char *str ) +{ + GLuint i; + GLuint len = strlen( str ); + + for( i = 0; i < len; i++ ) { + text_dump_write( dump, &str[i], 1 ); + + if( str[i] == '\n' ) { + GLuint i; + + for( i = 0; i < dump->tabs; i++ ) { + text_dump_write( dump, " ", 4 ); + } + } + } +} + +static void +text_dump_chr( + struct text_dump *dump, + const char chr ) +{ + char str[2]; + + str[0] = chr; + str[1] = '\0'; + text_dump_str( dump, str ); +} + +static void +text_dump_uix( + struct text_dump *dump, + const GLuint ui) +{ + char str[36]; + + sprintf( str, "0x%x", ui ); + text_dump_str( dump, str ); +} + +static void +text_dump_uid( + struct text_dump *dump, + const GLuint ui ) +{ + char str[16]; + + sprintf( str, "%u", ui ); + text_dump_str( dump, str ); +} + +static void +text_dump_sid( + struct text_dump *dump, + const GLint si ) +{ + char str[16]; + + sprintf( str, "%d", si ); + text_dump_str( dump, str ); +} + +static void +text_dump_flt( + struct text_dump *dump, + const GLfloat f ) +{ + char str[48]; + + sprintf( str, "%40.6f", f ); + text_dump_str( dump, str ); +} + +static void +text_dump_enum( + struct text_dump *dump, + const GLuint e, + const char **enums, + const GLuint enums_count ) +{ + if( e >= enums_count ) { + text_dump_uid( dump, e ); + } + else { + text_dump_str( dump, enums[e] ); + } +} + +static void +text_dump_tab( + struct text_dump *dump ) +{ + dump->tabs++; +} + +static void +text_dump_untab( + struct text_dump *dump ) +{ + assert( dump->tabs > 0 ); + + --dump->tabs; +} + +#define TXT(S) text_dump_str( &dump, S ) +#define CHR(C) text_dump_chr( &dump, C ) +#define UIX(I) text_dump_uix( &dump, I ) +#define UID(I) text_dump_uid( &dump, I ) +#define SID(I) text_dump_sid( &dump, I ) +#define FLT(F) text_dump_flt( &dump, F ) +#define TAB() text_dump_tab( &dump ) +#define UNT() text_dump_untab( &dump ) +#define ENM(E,ENUMS) text_dump_enum( &dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) + +static const char *TGSI_PROCESSOR_TYPES[] = +{ + "PROCESSOR_FRAGMENT", + "PROCESSOR_VERTEX" +}; + +static const char *TGSI_TOKEN_TYPES[] = +{ + "TOKEN_TYPE_DECLARATION", + "TOKEN_TYPE_IMMEDIATE", + "TOKEN_TYPE_INSTRUCTION" +}; + +static const char *TGSI_FILES[] = +{ + "FILE_NULL", + "FILE_CONSTANT", + "FILE_INPUT", + "FILE_OUTPUT", + "FILE_TEMPORARY", + "FILE_SAMPLER", + "FILE_ADDRESS", + "FILE_IMMEDIATE" +}; + +static const char *TGSI_DECLARES[] = +{ + "DECLARE_RANGE", + "DECLARE_MASK" +}; + +static const char *TGSI_INTERPOLATES[] = +{ + "INTERPOLATE_CONSTANT", + "INTERPOLATE_LINEAR", + "INTERPOLATE_PERSPECTIVE" +}; + +static const char *TGSI_IMMS[] = +{ + "IMM_FLOAT32" +}; + +static const char *TGSI_OPCODES[] = +{ + "OPCODE_MOV", + "OPCODE_LIT", + "OPCODE_RCP", + "OPCODE_RSQ", + "OPCODE_EXP", + "OPCODE_LOG", + "OPCODE_MUL", + "OPCODE_ADD", + "OPCODE_DP3", + "OPCODE_DP4", + "OPCODE_DST", + "OPCODE_MIN", + "OPCODE_MAX", + "OPCODE_SLT", + "OPCODE_SGE", + "OPCODE_MAD", + "OPCODE_SUB", + "OPCODE_LERP", + "OPCODE_CND", + "OPCODE_CND0", + "OPCODE_DOT2ADD", + "OPCODE_INDEX", + "OPCODE_NEGATE", + "OPCODE_FRAC", + "OPCODE_CLAMP", + "OPCODE_FLOOR", + "OPCODE_ROUND", + "OPCODE_EXPBASE2", + "OPCODE_LOGBASE2", + "OPCODE_POWER", + "OPCODE_CROSSPRODUCT", + "OPCODE_MULTIPLYMATRIX", + "OPCODE_ABS", + "OPCODE_RCC", + "OPCODE_DPH", + "OPCODE_COS", + "OPCODE_DDX", + "OPCODE_DDY", + "OPCODE_KIL", + "OPCODE_PK2H", + "OPCODE_PK2US", + "OPCODE_PK4B", + "OPCODE_PK4UB", + "OPCODE_RFL", + "OPCODE_SEQ", + "OPCODE_SFL", + "OPCODE_SGT", + "OPCODE_SIN", + "OPCODE_SLE", + "OPCODE_SNE", + "OPCODE_STR", + "OPCODE_TEX", + "OPCODE_TXD", + "OPCODE_UP2H", + "OPCODE_UP2US", + "OPCODE_UP4B", + "OPCODE_UP4UB", + "OPCODE_X2D", + "OPCODE_ARA", + "OPCODE_ARR", + "OPCODE_BRA", + "OPCODE_CAL", + "OPCODE_RET", + "OPCODE_SSG", + "OPCODE_CMP", + "OPCODE_SCS", + "OPCODE_TXB", + "OPCODE_NRM", + "OPCODE_DIV", + "OPCODE_DP2", + "OPCODE_TXL", + "OPCODE_BRK", + "OPCODE_IF", + "OPCODE_LOOP", + "OPCODE_REP", + "OPCODE_ELSE", + "OPCODE_ENDIF", + "OPCODE_ENDLOOP", + "OPCODE_ENDREP", + "OPCODE_PUSHA", + "OPCODE_POPA", + "OPCODE_CEIL", + "OPCODE_I2F", + "OPCODE_NOT", + "OPCODE_TRUNC", + "OPCODE_SHL", + "OPCODE_SHR", + "OPCODE_AND", + "OPCODE_OR", + "OPCODE_MOD", + "OPCODE_XOR", + "OPCODE_SAD", + "OPCODE_TXF", + "OPCODE_TXQ", + "OPCODE_CONT", + "OPCODE_EMIT", + "OPCODE_ENDPRIM" +}; + +static const char *TGSI_SATS[] = +{ + "SAT_NONE", + "SAT_ZERO_ONE", + "SAT_MINUS_PLUS_ONE" +}; + +static const char *TGSI_INSTRUCTION_EXTS[] = +{ + "INSTRUCTION_EXT_TYPE_NV", + "INSTRUCTION_EXT_TYPE_LABEL", + "INSTRUCTION_EXT_TYPE_TEXTURE" +}; + +static const char *TGSI_PRECISIONS[] = +{ + "PRECISION_DEFAULT", + "TGSI_PRECISION_FLOAT32", + "TGSI_PRECISION_FLOAT16", + "TGSI_PRECISION_FIXED12" +}; + +static const char *TGSI_CCS[] = +{ + "CC_GT", + "CC_EQ", + "CC_LT", + "CC_UN", + "CC_GE", + "CC_LE", + "CC_NE", + "CC_TR", + "CC_FL" +}; + +static const char *TGSI_SWIZZLES[] = +{ + "SWIZZLE_X", + "SWIZZLE_Y", + "SWIZZLE_Z", + "SWIZZLE_W" +}; + +static const char *TGSI_TEXTURES[] = +{ + "TEXTURE_UNKNOWN", + "TEXTURE_1D", + "TEXTURE_2D", + "TEXTURE_3D", + "TEXTURE_CUBE", + "TEXTURE_RECT", + "TEXTURE_SHADOW1D", + "TEXTURE_SHADOW2D", + "TEXTURE_SHADOWRECT" +}; + +static const char *TGSI_SRC_REGISTER_EXTS[] = +{ + "SRC_REGISTER_EXT_TYPE_SWZ", + "SRC_REGISTER_EXT_TYPE_MOD" +}; + +static const char *TGSI_EXTSWIZZLES[] = +{ + "EXTSWIZZLE_X", + "EXTSWIZZLE_Y", + "EXTSWIZZLE_Z", + "EXTSWIZZLE_W", + "EXTSWIZZLE_ZERO", + "EXTSWIZZLE_ONE" +}; + +static const char *TGSI_WRITEMASKS[] = +{ + "0", + "WRITEMASK_X", + "WRITEMASK_Y", + "WRITEMASK_XY", + "WRITEMASK_Z", + "WRITEMASK_XZ", + "WRITEMASK_YZ", + "WRITEMASK_XYZ", + "WRITEMASK_W", + "WRITEMASK_XW", + "WRITEMASK_YW", + "WRITEMASK_XYW", + "WRITEMASK_ZW", + "WRITEMASK_XZW", + "WRITEMASK_YZW", + "WRITEMASK_XYZW" +}; + +static const char *TGSI_DST_REGISTER_EXTS[] = +{ + "DST_REGISTER_EXT_TYPE_CONDCODE", + "DST_REGISTER_EXT_TYPE_MODULATE" +}; + +static const char *TGSI_MODULATES[] = +{ + "MODULATE_1X", + "MODULATE_2X", + "MODULATE_4X", + "MODULATE_8X", + "MODULATE_HALF", + "MODULATE_QUARTER", + "MODULATE_EIGHTH" +}; + +void +tgsi_dump( + const struct tgsi_token *tokens, + GLuint flags ) +{ + struct text_dump dump; + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + GLuint ignored = !(flags & TGSI_DUMP_NO_IGNORED); + GLuint deflt = !(flags & TGSI_DUMP_NO_DEFAULT); + + { + static GLuint counter = 0; + char buffer[64]; + + sprintf( buffer, "sbir-dump-%.4u.txt", counter++ ); + dump.file = fopen( buffer, "wt" ); + dump.tabs = 0; + } + + tgsi_parse_init( &parse, tokens ); + + TXT( "sbir-dump" ); + + CHR( '\n' ); + TXT( "\nMajorVersion: " ); + UID( parse.FullVersion.Version.MajorVersion ); + TXT( "\nMinorVersion: " ); + UID( parse.FullVersion.Version.MinorVersion ); + + CHR( '\n' ); + TXT( "\nHeaderSize: " ); + UID( parse.FullHeader.Header.HeaderSize ); + TXT( "\nBodySize : " ); + UID( parse.FullHeader.Header.BodySize ); + TXT( "\nProcessor : " ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + GLuint i; + + tgsi_parse_token( &parse ); + + CHR( '\n' ); + TXT( "\nType : " ); + ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); + if( ignored ) { + TXT( "\nSize : " ); + UID( parse.FullToken.Token.Size ); + if( deflt || parse.FullToken.Token.Extended ) { + TXT( "\nExtended : " ); + UID( parse.FullToken.Token.Extended ); + } + } + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; + + TXT( "\nFile : " ); + ENM( decl->Declaration.File, TGSI_FILES ); + TXT( "\nDeclare : " ); + ENM( decl->Declaration.Declare, TGSI_DECLARES ); + if( deflt || fd.Declaration.Interpolate != decl->Declaration.Interpolate ) { + TXT( "\nInterpolate: " ); + UID( decl->Declaration.Interpolate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Declaration.Padding ); + } + + CHR( '\n' ); + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + TXT( "\nFirst: " ); + UID( decl->u.DeclarationRange.First ); + TXT( "\nLast : " ); + UID( decl->u.DeclarationRange.Last ); + break; + + case TGSI_DECLARE_MASK: + TXT( "\nMask: " ); + UIX( decl->u.DeclarationMask.Mask ); + break; + + default: + assert( 0 ); + } + + if( decl->Declaration.Interpolate ) { + CHR( '\n' ); + TXT( "\nInterpolate: " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Interpolation.Padding ); + } + } + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + TXT( "\nDataType : " ); + ENM( parse.FullToken.FullImmediate.Immediate.DataType, TGSI_IMMS ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( parse.FullToken.FullImmediate.Immediate.Padding ); + } + + for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size - 1; i++ ) { + CHR( '\n' ); + switch( parse.FullToken.FullImmediate.Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + TXT( "\nFloat: " ); + FLT( parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + } + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction; + + TXT( "\nOpcode : " ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES ); + if( deflt || fi.Instruction.Saturate != inst->Instruction.Saturate ) { + TXT( "\nSaturate : " ); + ENM( inst->Instruction.Saturate, TGSI_SATS ); + } + if( deflt || fi.Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { + TXT( "\nNumDstRegs : " ); + UID( inst->Instruction.NumDstRegs ); + } + if( deflt || fi.Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { + TXT( "\nNumSrcRegs : " ); + UID( inst->Instruction.NumSrcRegs ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->Instruction.Padding ); + } + + if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi.InstructionExtNv)) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi.InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { + TXT( "\nPrecision : " ); + ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); + } + if( deflt || fi.InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { + TXT( "\nCondDstIndex : " ); + UID( inst->InstructionExtNv.CondDstIndex ); + } + if( deflt || fi.InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { + TXT( "\nCondFlowIndex : " ); + UID( inst->InstructionExtNv.CondFlowIndex ); + } + if( deflt || fi.InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { + TXT( "\nCondMask : " ); + ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); + } + if( deflt || fi.InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { + TXT( "\nCondSwizzleX : " ); + ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fi.InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { + TXT( "\nCondSwizzleY : " ); + ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fi.InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ : " ); + ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fi.InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { + TXT( "\nCondSwizzleW : " ); + ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fi.InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { + TXT( "\nCondDstUpdate : " ); + UID( inst->InstructionExtNv.CondDstUpdate ); + } + if( deflt || fi.InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { + TXT( "\nCondFlowEnable: " ); + UID( inst->InstructionExtNv.CondFlowEnable ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtNv.Padding ); + if( deflt || fi.InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { + TXT( "\nExtended : " ); + UID( inst->InstructionExtNv.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi.InstructionExtLabel ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi.InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { + TXT( "\nLabel : " ); + UID( inst->InstructionExtLabel.Label ); + } + if( deflt || fi.InstructionExtLabel.Target != inst->InstructionExtLabel.Target ) { + TXT( "\nTarget : " ); + UID( inst->InstructionExtLabel.Target ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtLabel.Padding ); + if( deflt || fi.InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtLabel.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi.InstructionExtTexture ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi.InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { + TXT( "\nTexture : " ); + ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtTexture.Padding ); + if( deflt || fi.InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtTexture.Extended ); + } + } + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + struct tgsi_full_dst_register *fd = &fi.FullDstRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : " ); + ENM( dst->DstRegister.File, TGSI_FILES ); + if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { + TXT( "\nWriteMask: " ); + ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); + } + if( ignored ) { + if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( dst->DstRegister.Indirect ); + } + if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( dst->DstRegister.Dimension ); + } + } + if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { + TXT( "\nIndex : " ); + SID( dst->DstRegister.Index ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegister.Padding ); + if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegister.Extended ); + } + } + + if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { + TXT( "\nCondMask : " ); + ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { + TXT( "\nCondSwizzleX: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { + TXT( "\nCondSwizzleY: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { + TXT( "\nCondSwizzleW: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { + TXT( "\nCondSrcIndex: " ); + UID( dst->DstRegisterExtConcode.CondSrcIndex ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtConcode.Padding ); + if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegisterExtConcode.Extended ); + } + } + } + + if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { + TXT( "\nModulate: " ); + ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtModulate.Padding ); + if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { + TXT( "\nExtended: " ); + UID( dst->DstRegisterExtModulate.Extended ); + } + } + } + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + struct tgsi_full_src_register *fs = &fi.FullSrcRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : "); + ENM( src->SrcRegister.File, TGSI_FILES ); + if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { + TXT( "\nSwizzleX : " ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { + TXT( "\nSwizzleY : " ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { + TXT( "\nSwizzleZ : " ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { + TXT( "\nSwizzleW : " ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegister.Negate ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( src->SrcRegister.Indirect ); + } + if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( src->SrcRegister.Dimension ); + } + } + if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { + TXT( "\nIndex : " ); + SID( src->SrcRegister.Index ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegister.Extended ); + } + } + + if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { + TXT( "\nExtSwizzleX: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { + TXT( "\nExtSwizzleY: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { + TXT( "\nExtSwizzleZ: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { + TXT( "\nExtSwizzleW: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { + TXT( "\nNegateX : " ); + UID( src->SrcRegisterExtSwz.NegateX ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { + TXT( "\nNegateY : " ); + UID( src->SrcRegisterExtSwz.NegateY ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { + TXT( "\nNegateZ : " ); + UID( src->SrcRegisterExtSwz.NegateZ ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { + TXT( "\nNegateW : " ); + UID( src->SrcRegisterExtSwz.NegateW ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { + TXT( "\nExtDivide : " ); + ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtSwz.Padding ); + if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtSwz.Extended ); + } + } + } + + if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { + TXT( "\nComplement: " ); + UID( src->SrcRegisterExtMod.Complement ); + } + if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { + TXT( "\nBias : " ); + UID( src->SrcRegisterExtMod.Bias ); + } + if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { + TXT( "\nScale2X : " ); + UID( src->SrcRegisterExtMod.Scale2X ); + } + if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { + TXT( "\nAbsolute : " ); + UID( src->SrcRegisterExtMod.Absolute ); + } + if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegisterExtMod.Negate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtMod.Padding ); + if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtMod.Extended ); + } + } + } + } + } + break; + + default: + assert( 0 ); + } + } + + tgsi_parse_free( &parse ); + + fclose( dump.file ); +} + diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.h b/src/mesa/pipe/tgsi/core/tgsi_dump.h new file mode 100644 index 0000000000..dc34a840de --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.h @@ -0,0 +1,22 @@ +#if !defined TGSI_DUMP_H +#define TGSI_DUMP_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +#define TGSI_DUMP_VERBOSE 0 +#define TGSI_DUMP_NO_IGNORED 1 +#define TGSI_DUMP_NO_DEFAULT 2 + +void +tgsi_dump( + const struct tgsi_token *tokens, + GLuint flags ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_DUMP_H + diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c new file mode 100644 index 0000000000..189a41167a --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -0,0 +1,2193 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +#define TILE_BOTTOM_LEFT 0 +#define TILE_BOTTOM_RIGHT 1 +#define TILE_TOP_LEFT 2 +#define TILE_TOP_RIGHT 3 + +#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I +#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C +#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I +#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C +#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I +#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C +#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I +#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C +#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I +#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C +#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I +#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C +#define TEMP_128_I TGSI_EXEC_TEMP_128_I +#define TEMP_128_C TGSI_EXEC_TEMP_128_C +#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I +#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C +#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I +#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C +#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I +#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C +#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I +#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +#define FOR_EACH_CHANNEL(CHAN)\ + for (CHAN = 0; CHAN < 4; CHAN++) + +#define IS_CHANNEL_ENABLED(INST, CHAN)\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IS_CHANNEL_ENABLED2(INST, CHAN)\ + ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) + +#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED2( INST, CHAN )) + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach, + struct tgsi_token *tokens ) +{ + GLuint i; + struct tgsi_parse_context parse; + + mach->Tokens = tokens; + + tgsi_parse_init (&parse, mach->Tokens); + mach->Processor = parse.FullHeader.Processor.Processor; + tgsi_parse_free (&parse); + + mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); + mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; + +#if XXX_SSE + tgsi_emit_sse (tokens, + &mach->Function); +#endif + + /* Setup constants. */ + for( i = 0; i < 4; i++ ) { + mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; + mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; + mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; + mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; + mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; + mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; + mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; + mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; + } +} + +void +tgsi_exec_prepare( + struct tgsi_exec_machine *mach, + struct tgsi_exec_labels *labels ) +{ + struct tgsi_parse_context parse; + + mach->ImmLimit = 0; + labels->count = 0; + + tgsi_parse_init( &parse, mach->Tokens ); + while( !tgsi_parse_end_of_tokens( &parse ) ) { + GLuint pointer = parse.Position; + GLuint i; + tgsi_parse_token( &parse ); + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + assert( (parse.FullToken.FullImmediate.Immediate.Size - 1) % 4 == 0 ); + assert( mach->ImmLimit + (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4 <= 256 ); + for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size - 1; i++ ) { + mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } + mach->ImmLimit += (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4; + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + if( parse.FullToken.FullInstruction.InstructionExtLabel.Label && + parse.FullToken.FullInstruction.InstructionExtLabel.Target ) { + assert( labels->count < 128 ); + labels->labels[labels->count][0] = parse.FullToken.FullInstruction.InstructionExtLabel.Label; + labels->labels[labels->count][1] = pointer; + labels->count++; + } + break; + default: + assert( 0 ); + } + } + tgsi_parse_free (&parse); +} + +void +tgsi_exec_machine_run( + struct tgsi_exec_machine *mach ) +{ + struct tgsi_exec_labels labels; + + tgsi_exec_prepare( mach, &labels ); + tgsi_exec_machine_run2( mach, &labels ); +} + +static void +micro_abs( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) fabs( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) fabs( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) fabs( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) fabs( (GLdouble) src->f[3] ); +} + +static void +micro_add( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] + src1->f[0]; + dst->f[1] = src0->f[1] + src1->f[1]; + dst->f[2] = src0->f[2] + src1->f[2]; + dst->f[3] = src0->f[3] + src1->f[3]; +} + +static void +micro_iadd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] + src1->i[0]; + dst->i[1] = src0->i[1] + src1->i[1]; + dst->i[2] = src0->i[2] + src1->i[2]; + dst->i[3] = src0->i[3] + src1->i[3]; +} + +static void +micro_and( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] & src1->u[0]; + dst->u[1] = src0->u[1] & src1->u[1]; + dst->u[2] = src0->u[2] & src1->u[2]; + dst->u[3] = src0->u[3] & src1->u[3]; +} + +static void +micro_ceil( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) ceil( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) ceil( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) ceil( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) ceil( (GLdouble) src->f[3] ); +} + +static void +micro_cos( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) cos( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) cos( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) cos( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) cos( (GLdouble) src->f[3] ); +} + +static void +micro_div( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] / src1->f[0]; + dst->f[1] = src0->f[1] / src1->f[1]; + dst->f[2] = src0->f[2] / src1->f[2]; + dst->f[3] = src0->f[3] / src1->f[3]; +} + +static void +micro_udiv( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] / src1->u[0]; + dst->u[1] = src0->u[1] / src1->u[1]; + dst->u[2] = src0->u[2] / src1->u[2]; + dst->u[3] = src0->u[3] / src1->u[3]; +} + +static void +micro_eq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ieq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_exp2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) pow( 2.0, (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) pow( 2.0, (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) pow( 2.0, (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) pow( 2.0, (GLdouble) src->f[3] ); +} + +static void +micro_f2it( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = (GLint) src->f[0]; + dst->i[1] = (GLint) src->f[1]; + dst->i[2] = (GLint) src->f[2]; + dst->i[3] = (GLint) src->f[3]; +} + +static void +micro_f2ut( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = (GLuint) src->f[0]; + dst->u[1] = (GLuint) src->f[1]; + dst->u[2] = (GLuint) src->f[2]; + dst->u[3] = (GLuint) src->f[3]; +} + +static void +micro_flr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) floor( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) floor( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) floor( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) floor( (GLdouble) src->f[3] ); +} + +static void +micro_frc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = src->f[0] - (GLfloat) floor( (GLdouble) src->f[0] ); + dst->f[1] = src->f[1] - (GLfloat) floor( (GLdouble) src->f[1] ); + dst->f[2] = src->f[2] - (GLfloat) floor( (GLdouble) src->f[2] ); + dst->f[3] = src->f[3] - (GLfloat) floor( (GLdouble) src->f[3] ); +} + +static void +micro_i2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) src->i[0]; + dst->f[1] = (GLfloat) src->i[1]; + dst->f[2] = (GLfloat) src->i[2]; + dst->f[3] = (GLfloat) src->i[3]; +} + +static void +micro_lg2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) log( (GLdouble) src->f[0] ) * 1.442695f; + dst->f[1] = (GLfloat) log( (GLdouble) src->f[1] ) * 1.442695f; + dst->f[2] = (GLfloat) log( (GLdouble) src->f[2] ) * 1.442695f; + dst->f[3] = (GLfloat) log( (GLdouble) src->f[3] ) * 1.442695f; +} + +static void +micro_lt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ilt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_ult( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; +} + +static void +micro_max( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_min( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_umod( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] % src1->u[0]; + dst->u[1] = src0->u[1] % src1->u[1]; + dst->u[2] = src0->u[2] % src1->u[2]; + dst->u[3] = src0->u[3] % src1->u[3]; +} + +static void +micro_mul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] * src1->f[0]; + dst->f[1] = src0->f[1] * src1->f[1]; + dst->f[2] = src0->f[2] * src1->f[2]; + dst->f[3] = src0->f[3] * src1->f[3]; +} + +static void +micro_imul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] * src1->i[0]; + dst->i[1] = src0->i[1] * src1->i[1]; + dst->i[2] = src0->i[2] * src1->i[2]; + dst->i[3] = src0->i[3] * src1->i[3]; +} + +static void +micro_imul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->i[0] = src0->i[0] * src1->i[0]; + dst1->i[1] = src0->i[1] * src1->i[1]; + dst1->i[2] = src0->i[2] * src1->i[2]; + dst1->i[3] = src0->i[3] * src1->i[3]; + dst0->i[0] = 0; + dst0->i[1] = 0; + dst0->i[2] = 0; + dst0->i[3] = 0; +} + +static void +micro_umul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->u[0] = src0->u[0] * src1->u[0]; + dst1->u[1] = src0->u[1] * src1->u[1]; + dst1->u[2] = src0->u[2] * src1->u[2]; + dst1->u[3] = src0->u[3] * src1->u[3]; + dst0->u[0] = 0; + dst0->u[1] = 0; + dst0->u[2] = 0; + dst0->u[3] = 0; +} + +static void +micro_movc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2 ) +{ + dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; + dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; + dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; + dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; +} + +static void +micro_neg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = -src->f[0]; + dst->f[1] = -src->f[1]; + dst->f[2] = -src->f[2]; + dst->f[3] = -src->f[3]; +} + +static void +micro_ineg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = -src->i[0]; + dst->i[1] = -src->i[1]; + dst->i[2] = -src->i[2]; + dst->i[3] = -src->i[3]; +} + +static void +micro_not( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = ~src->u[0]; + dst->u[1] = ~src->u[1]; + dst->u[2] = ~src->u[2]; + dst->u[3] = ~src->u[3]; +} + +static void +micro_or( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] | src1->u[0]; + dst->u[1] = src0->u[1] | src1->u[1]; + dst->u[2] = src0->u[2] | src1->u[2]; + dst->u[3] = src0->u[3] | src1->u[3]; +} + +static void +micro_pow( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = (GLfloat) pow( (GLdouble) src0->f[0], (GLdouble) src1->f[0] ); + dst->f[1] = (GLfloat) pow( (GLdouble) src0->f[1], (GLdouble) src1->f[1] ); + dst->f[2] = (GLfloat) pow( (GLdouble) src0->f[2], (GLdouble) src1->f[2] ); + dst->f[3] = (GLfloat) pow( (GLdouble) src0->f[3], (GLdouble) src1->f[3] ); +} + +static void +micro_rnd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) floor( (GLdouble) (src->f[0] + 0.5f) ); + dst->f[1] = (GLfloat) floor( (GLdouble) (src->f[1] + 0.5f) ); + dst->f[2] = (GLfloat) floor( (GLdouble) (src->f[2] + 0.5f) ); + dst->f[3] = (GLfloat) floor( (GLdouble) (src->f[3] + 0.5f) ); +} + +static void +micro_shl( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] << src1->i[0]; + dst->i[1] = src0->i[1] << src1->i[1]; + dst->i[2] = src0->i[2] << src1->i[2]; + dst->i[3] = src0->i[3] << src1->i[3]; +} + +static void +micro_ishr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] >> src1->i[0]; + dst->i[1] = src0->i[1] >> src1->i[1]; + dst->i[2] = src0->i[2] >> src1->i[2]; + dst->i[3] = src0->i[3] >> src1->i[3]; +} + +static void +micro_ushr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] >> src1->u[0]; + dst->u[1] = src0->u[1] >> src1->u[1]; + dst->u[2] = src0->u[2] >> src1->u[2]; + dst->u[3] = src0->u[3] >> src1->u[3]; +} + +static void +micro_sin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) sin( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) sin( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) sin( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) sin( (GLdouble) src->f[3] ); +} + +static void +micro_sqrt( union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) sqrt( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) sqrt( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) sqrt( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) sqrt( (GLdouble) src->f[3] ); +} + +static void +micro_sub( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] - src1->f[0]; + dst->f[1] = src0->f[1] - src1->f[1]; + dst->f[2] = src0->f[2] - src1->f[2]; + dst->f[3] = src0->f[3] - src1->f[3]; +} + +static void +micro_u2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) src->u[0]; + dst->f[1] = (GLfloat) src->u[1]; + dst->f[2] = (GLfloat) src->u[2]; + dst->f[3] = (GLfloat) src->u[3]; +} + +static void +micro_xor( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] ^ src1->u[0]; + dst->u[1] = src0->u[1] ^ src1->u[1]; + dst->u[2] = src0->u[2] ^ src1->u[2]; + dst->u[3] = src0->u[3] ^ src1->u[3]; +} + +static void +fetch_src_file_channel( + const struct tgsi_exec_machine *mach, + const GLuint file, + const GLuint swizzle, + const union tgsi_exec_channel *index, + union tgsi_exec_channel *chan ) +{ + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( file ) { + case TGSI_FILE_CONSTANT: + chan->f[0] = mach->Consts[index->i[0]][swizzle]; + chan->f[1] = mach->Consts[index->i[1]][swizzle]; + chan->f[2] = mach->Consts[index->i[2]][swizzle]; + chan->f[3] = mach->Consts[index->i[3]][swizzle]; + break; + + case TGSI_FILE_INPUT: + chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_TEMPORARY: + chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_IMMEDIATE: + assert( index->i[0] < (GLint) mach->ImmLimit ); + chan->f[0] = mach->Imms[index->i[0]][swizzle]; + assert( index->i[1] < (GLint) mach->ImmLimit ); + chan->f[1] = mach->Imms[index->i[1]][swizzle]; + assert( index->i[2] < (GLint) mach->ImmLimit ); + chan->f[2] = mach->Imms[index->i[2]][swizzle]; + assert( index->i[3] < (GLint) mach->ImmLimit ); + chan->f[3] = mach->Imms[index->i[3]][swizzle]; + break; + + case TGSI_FILE_ADDRESS: + chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; + break; + + case TGSI_EXTSWIZZLE_ONE: + *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; + break; + + default: + assert( 0 ); + } +} + +static void +fetch_source( + const struct tgsi_exec_machine *mach, + union tgsi_exec_channel *chan, + const struct tgsi_full_src_register *reg, + const GLuint chan_index ) +{ + union tgsi_exec_channel index; + GLuint swizzle; + + index.i[0] = + index.i[1] = + index.i[2] = + index.i[3] = reg->SrcRegister.Index; + + if (reg->SrcRegister.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + + if( reg->SrcRegister.Dimension ) { + switch( reg->SrcRegister.File ) { + case TGSI_FILE_INPUT: + index.i[0] *= 17; + index.i[1] *= 17; + index.i[2] *= 17; + index.i[3] *= 17; + break; + case TGSI_FILE_CONSTANT: + index.i[0] *= 4096; + index.i[1] *= 4096; + index.i[2] *= 4096; + index.i[3] *= 4096; + break; + default: + assert( 0 ); + } + + index.i[0] += reg->SrcRegisterDim.Index; + index.i[1] += reg->SrcRegisterDim.Index; + index.i[2] += reg->SrcRegisterDim.Index; + index.i[3] += reg->SrcRegisterDim.Index; + + if (reg->SrcRegisterDim.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterDimInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterDimInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + } + + swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + fetch_src_file_channel( + mach, + reg->SrcRegister.File, + swizzle, + &index, + chan ); + + switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { + case TGSI_UTIL_SIGN_CLEAR: + micro_abs( chan, chan ); + break; + + case TGSI_UTIL_SIGN_SET: + micro_abs( chan, chan ); + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } +} + +static void +store_dest( + struct tgsi_exec_machine *mach, + const union tgsi_exec_channel *chan, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + GLuint chan_index ) +{ + union tgsi_exec_channel *dst; + + switch( reg->DstRegister.File ) { + case TGSI_FILE_NULL: + return; + + case TGSI_FILE_OUTPUT: + dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_TEMPORARY: + dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_ADDRESS: + dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; + break; + + default: + assert( 0 ); + } + + switch (inst->Instruction.Saturate) + { + case TGSI_SAT_NONE: + *dst = *chan; + break; + + case TGSI_SAT_ZERO_ONE: + micro_lt( dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], chan ); + micro_lt( dst, chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + + default: + assert( 0 ); + } +} + +#define FETCH(VAL,INDEX,CHAN)\ + fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) + +#define STORE(VAL,INDEX,CHAN)\ + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN) + +static void +exec_kil (struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + GLuint uniquemask; + GLuint chan_index; + GLuint kilmask = 0; + union tgsi_exec_channel r[1]; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + GLuint swizzle; + GLuint i; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle ( + &inst->FullSrcRegisters[0], + chan_index); + + /* check if the component has not been already tested */ + if (uniquemask & (1 << swizzle)) + continue; + uniquemask |= 1 << swizzle; + + FETCH(&r[0], 0, chan_index); + for (i = 0; i < 4; i++) + if (r[0].f[i] < 0.0f) + kilmask |= 1 << (i * 4); + } + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; +} + +#if MESA +/* + * Fetch a texel using S texture coordinate. + */ +static void +fetch_texel_1d( GLcontext *ctx, + struct tgsi_sampler_state *sampler, + const union tgsi_exec_channel *s, + GLuint unit, + union tgsi_exec_channel *r, + union tgsi_exec_channel *g, + union tgsi_exec_channel *b, + union tgsi_exec_channel *a ) +{ + SWcontext *swrast = SWRAST_CONTEXT(ctx); + GLuint fragment_index; + GLfloat stpq[4][4]; + GLfloat lambdas[4]; + GLchan rgba[4][4]; + + for (fragment_index = 0; fragment_index < 4; fragment_index++) + { + stpq[fragment_index][0] = s->f[fragment_index]; + } + + if (sampler->NeedLambda) + { + GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; + GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; + + GLfloat rho, lambda; + + dsdx = FABSF(dsdx); + dsdy = FABSF(dsdy); + + rho = MAX2(dsdx, dsdy) * sampler->ImageWidth; + + lambda = LOG2(rho); + + if (sampler->NeedLodBias) + lambda += sampler->LodBias; + + if (sampler->NeedLambdaClamp) + lambda = CLAMP(lambda, sampler->MinLod, sampler->MaxLod); + + /* XXX: Use the same lambda value throughout the tile. Could + * end up with four unique values by recalculating partial + * derivs in the other row and column, and calculating lambda + * using the dx and dy values appropriate for each fragment in + * the tile. + */ + lambdas[0] = + lambdas[1] = + lambdas[2] = + lambdas[3] = lambda; + } + + /* XXX use a float-valued TextureSample routine here!!! */ + swrast->TextureSample[unit] (ctx, + ctx->Texture.Unit[unit]._Current, + 4, + (const GLfloat (*)[4])stpq, + lambdas, + rgba); + + for (fragment_index = 0; fragment_index < 4; fragment_index++) + { + r->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][0]); + g->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][1]); + b->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][2]); + a->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][3]); + } +} + +/* + * Fetch a texel using ST texture coordinates. + */ +static void +fetch_texel_2d( GLcontext *ctx, + struct tgsi_sampler_state *sampler, + const union tgsi_exec_channel *s, + const union tgsi_exec_channel *t, + GLuint unit, + union tgsi_exec_channel *r, + union tgsi_exec_channel *g, + union tgsi_exec_channel *b, + union tgsi_exec_channel *a ) +{ + SWcontext *swrast = SWRAST_CONTEXT( ctx ); + GLuint fragment_index; + GLfloat stpq[4][4]; + GLfloat lambdas[4]; + GLchan rgba[4][4]; + + for (fragment_index = 0; fragment_index < 4; fragment_index++) { + stpq[fragment_index][0] = s->f[fragment_index]; + stpq[fragment_index][1] = t->f[fragment_index]; + } + + if (sampler->NeedLambda) { + GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; + GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; + + GLfloat dtdx = t->f[TILE_BOTTOM_RIGHT] - t->f[TILE_BOTTOM_LEFT]; + GLfloat dtdy = t->f[TILE_TOP_LEFT] - t->f[TILE_BOTTOM_LEFT]; + + GLfloat maxU, maxV, rho, lambda; + + dsdx = FABSF( dsdx ); + dsdy = FABSF( dsdy ); + dtdx = FABSF( dtdx ); + dtdy = FABSF( dtdy ); + + maxU = MAX2( dsdx, dsdy ) * sampler->ImageWidth; + maxV = MAX2( dtdx, dtdy ) * sampler->ImageHeight; + + rho = MAX2( maxU, maxV ); + + lambda = LOG2( rho ); + + if (sampler->NeedLodBias) + lambda += sampler->LodBias; + + if (sampler->NeedLambdaClamp) + lambda = CLAMP( + lambda, + sampler->MinLod, + sampler->MaxLod ); + + /* XXX: Use the same lambda value throughout the tile. Could + * end up with four unique values by recalculating partial + * derivs in the other row and column, and calculating lambda + * using the dx and dy values appropriate for each fragment in + * the tile. + */ + lambdas[0] = + lambdas[1] = + lambdas[2] = + lambdas[3] = lambda; + } + + /* XXX use a float-valued TextureSample routine here!!! */ + swrast->TextureSample[unit]( + ctx, + ctx->Texture.Unit[unit]._Current, + 4, + (const GLfloat (*)[4]) stpq, + lambdas, + rgba ); + + for (fragment_index = 0; fragment_index < 4; fragment_index++) { + r->f[fragment_index] = CHAN_TO_FLOAT( rgba[fragment_index][0] ); + g->f[fragment_index] = CHAN_TO_FLOAT( rgba[fragment_index][1] ); + b->f[fragment_index] = CHAN_TO_FLOAT( rgba[fragment_index][2] ); + a->f[fragment_index] = CHAN_TO_FLOAT( rgba[fragment_index][3] ); + } +} + +/* + * Fetch a texel using STR texture coordinates. + */ +static void +fetch_texel_3d( GLcontext *ctx, + struct tgsi_sampler_state *sampler, + const union tgsi_exec_channel *s, + const union tgsi_exec_channel *t, + const union tgsi_exec_channel *p, + GLuint unit, + union tgsi_exec_channel *r, + union tgsi_exec_channel *g, + union tgsi_exec_channel *b, + union tgsi_exec_channel *a ) +{ + SWcontext *swrast = SWRAST_CONTEXT(ctx); + GLuint fragment_index; + GLfloat stpq[4][4]; + GLfloat lambdas[4]; + GLchan rgba[4][4]; + + for (fragment_index = 0; fragment_index < 4; fragment_index++) + { + stpq[fragment_index][0] = s->f[fragment_index]; + stpq[fragment_index][1] = t->f[fragment_index]; + stpq[fragment_index][2] = p->f[fragment_index]; + } + + if (sampler->NeedLambda) + { + GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; + GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; + + GLfloat dtdx = t->f[TILE_BOTTOM_RIGHT] - t->f[TILE_BOTTOM_LEFT]; + GLfloat dtdy = t->f[TILE_TOP_LEFT] - t->f[TILE_BOTTOM_LEFT]; + + GLfloat dpdx = p->f[TILE_BOTTOM_RIGHT] - p->f[TILE_BOTTOM_LEFT]; + GLfloat dpdy = p->f[TILE_TOP_LEFT] - p->f[TILE_BOTTOM_LEFT]; + + GLfloat maxU, maxV, maxW, rho, lambda; + + dsdx = FABSF(dsdx); + dsdy = FABSF(dsdy); + dtdx = FABSF(dtdx); + dtdy = FABSF(dtdy); + dpdx = FABSF(dpdx); + dpdy = FABSF(dpdy); + + maxU = MAX2(dsdx, dsdy) * sampler->ImageWidth; + maxV = MAX2(dtdx, dtdy) * sampler->ImageHeight; + maxW = MAX2(dpdx, dpdy) * sampler->ImageDepth; + + rho = MAX2(maxU, MAX2(maxV, maxW)); + + lambda = LOG2(rho); + + if (sampler->NeedLodBias) + lambda += sampler->LodBias; + + if (sampler->NeedLambdaClamp) + lambda = CLAMP(lambda, sampler->MinLod, sampler->MaxLod); + + /* XXX: Use the same lambda value throughout the tile. Could + * end up with four unique values by recalculating partial + * derivs in the other row and column, and calculating lambda + * using the dx and dy values appropriate for each fragment in + * the tile. + */ + lambdas[0] = + lambdas[1] = + lambdas[2] = + lambdas[3] = lambda; + } + + /* XXX use a float-valued TextureSample routine here!!! */ + swrast->TextureSample[unit] (ctx, + ctx->Texture.Unit[unit]._Current, + 4, + (const GLfloat (*)[4])stpq, + lambdas, + rgba); + + for (fragment_index = 0; fragment_index < 4; fragment_index++) + { + r->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][0]); + g->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][1]); + b->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][2]); + a->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][3]); + } +} +#endif + +static GLuint +map_label( + GLuint label, + struct tgsi_exec_labels *labels ) +{ + GLuint i; + + for( i = 0; i < labels->count; i++ ) { + if( labels->labels[i][0] == label ) { + return labels->labels[i][1]; + } + } + assert( 0 ); + return 0; +} + +static void +exec_instruction( + struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + struct tgsi_exec_labels *labels, + GLuint *programCounter ) +{ +#if MESA + GET_CURRENT_CONTEXT(ctx); +#endif + GLuint chan_index; + union tgsi_exec_channel r[8]; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_f2it( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_X ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[1], 0, CHAN_Y ); + micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + + FETCH( &r[2], 0, CHAN_W ); + micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); + micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); + micro_pow( &r[1], &r[1], &r[2] ); + micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Z ); + } + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( &r[0], 0, CHAN_X ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( &r[0], 0, CHAN_X ); + micro_sqrt( &r[0], &r[0] ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + assert (0); + break; + + case TGSI_OPCODE_LOG: + assert (0); + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_mul( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Z ); + FETCH( &r[2], 1, CHAN_Z ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_W); + FETCH(&r[2], 1, CHAN_W); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + FETCH( &r[0], 0, CHAN_Y ); + FETCH( &r[1], 1, CHAN_Y); + micro_mul( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_Z ); + STORE( &r[0], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + FETCH( &r[0], 1, CHAN_W ); + STORE( &r[0], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_mul( &r[0], &r[0], &r[1] ); + FETCH( &r[1], 2, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_sub( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_sub( &r[1], &r[1], &r[2] ); + micro_mul( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_CND: + assert (0); + break; + + case TGSI_OPCODE_CND0: + assert (0); + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert (0); + break; + + case TGSI_OPCODE_INDEX: + assert (0); + break; + + case TGSI_OPCODE_NEGATE: + assert (0); + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_frc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + assert (0); + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_flr( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_rnd( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH(&r[0], 0, CHAN_X); + + micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( &r[0], 0, CHAN_X ); + micro_lg2( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_pow( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + FETCH(&r[0], 0, CHAN_Y); + FETCH(&r[1], 1, CHAN_Z); + + micro_mul( &r[2], &r[0], &r[1] ); + + FETCH(&r[3], 0, CHAN_Z); + FETCH(&r[4], 1, CHAN_Y); + + micro_mul( &r[5], &r[3], &r[4] ); + micro_sub( &r[2], &r[2], &r[5] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &r[2], 0, CHAN_X ); + } + + FETCH(&r[2], 1, CHAN_X); + + micro_mul( &r[3], &r[3], &r[2] ); + + FETCH(&r[5], 0, CHAN_X); + + micro_mul( &r[1], &r[1], &r[5] ); + micro_sub( &r[3], &r[3], &r[1] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + STORE( &r[3], 0, CHAN_Y ); + } + + micro_mul( &r[5], &r[5], &r[4] ); + micro_mul( &r[0], &r[0], &r[2] ); + micro_sub( &r[5], &r[5], &r[0] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( &r[5], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + assert (0); + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + + micro_abs( &r[0], &r[0] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_RCC: + assert (0); + break; + + case TGSI_OPCODE_DPH: + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 1, CHAN_W); + + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH(&r[0], 0, CHAN_X); + + micro_cos( &r[0], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + assert (0); + break; + + case TGSI_OPCODE_DDY: + assert (0); + break; + + case TGSI_OPCODE_KIL: + exec_kil (mach, inst); + break; + + case TGSI_OPCODE_PK2H: + assert (0); + break; + + case TGSI_OPCODE_PK2US: + assert (0); + break; + + case TGSI_OPCODE_PK4B: + assert (0); + break; + + case TGSI_OPCODE_PK4UB: + assert (0); + break; + + case TGSI_OPCODE_RFL: + assert (0); + break; + + case TGSI_OPCODE_SEQ: + assert (0); + break; + + case TGSI_OPCODE_SFL: + assert (0); + break; + + case TGSI_OPCODE_SGT: + assert (0); + break; + + case TGSI_OPCODE_SIN: + FETCH(&r[0], 0, CHAN_X); + + micro_sin( &r[0], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + assert (0); + break; + + case TGSI_OPCODE_SNE: + assert (0); + break; + + case TGSI_OPCODE_STR: + assert (0); + break; + + case TGSI_OPCODE_TEX: + + switch (inst->InstructionExtTexture.Texture) + { + case TGSI_TEXTURE_1D: + + FETCH(&r[0], 0, CHAN_X); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) + { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[1], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[1] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } +#if MESA + fetch_texel_1d (ctx, + &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], + &r[0], + inst->FullSrcRegisters[1].SrcRegister.Index, + &r[0], &r[1], &r[2], &r[3]); +#endif + break; + + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) + { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[2], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[2] ); + micro_div( &r[1], &r[1], &r[2] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + +#if MESA + fetch_texel_2d (ctx, + &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], + &r[0], &r[1], + inst->FullSrcRegisters[1].SrcRegister.Index, + &r[0], &r[1], &r[2], &r[3]); +#endif + break; + + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) + { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + +#if MESA + fetch_texel_3d (ctx, + &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], + &r[0], &r[1], &r[2], + inst->FullSrcRegisters[1].SrcRegister.Index, + &r[0], &r[1], &r[2], &r[3]); +#endif + break; + + default: + assert (0); + } + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TXD: + assert (0); + break; + + case TGSI_OPCODE_UP2H: + assert (0); + break; + + case TGSI_OPCODE_UP2US: + assert (0); + break; + + case TGSI_OPCODE_UP4B: + assert (0); + break; + + case TGSI_OPCODE_UP4UB: + assert (0); + break; + + case TGSI_OPCODE_X2D: + assert (0); + break; + + case TGSI_OPCODE_ARA: + assert (0); + break; + + case TGSI_OPCODE_ARR: + assert (0); + break; + + case TGSI_OPCODE_BRA: + assert (0); + break; + + case TGSI_OPCODE_CAL: + assert (0); + break; + + case TGSI_OPCODE_RET: + /* XXX: end of shader! */ + /*assert (0);*/ + break; + + case TGSI_OPCODE_SSG: + assert (0); + break; + + case TGSI_OPCODE_CMP: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_SCS: + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( &r[0], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + micro_cos( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + micro_sin( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_Y ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_TXB: + assert (0); + break; + + case TGSI_OPCODE_NRM: + assert (0); + break; + + case TGSI_OPCODE_DIV: + assert( 0 ); + break; + + case TGSI_OPCODE_DP2: + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TXL: + assert (0); + break; + + case TGSI_OPCODE_BRK: + assert (0); + break; + + case TGSI_OPCODE_IF: + assert (0); + break; + + case TGSI_OPCODE_LOOP: + assert (0); + break; + + case TGSI_OPCODE_REP: + assert (0); + break; + + case TGSI_OPCODE_ELSE: + assert (0); + break; + + case TGSI_OPCODE_ENDIF: + assert (0); + break; + + case TGSI_OPCODE_ENDLOOP: + assert (0); + break; + + case TGSI_OPCODE_ENDREP: + assert (0); + break; + + case TGSI_OPCODE_PUSHA: + assert (0); + break; + + case TGSI_OPCODE_POPA: + assert (0); + break; + + case TGSI_OPCODE_CEIL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ceil( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_I2F: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_i2f( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_NOT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_not( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TRUNC: + assert (0); + break; + + case TGSI_OPCODE_SHL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_shl( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ishr( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_AND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_and( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_OR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_or( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOD: + assert (0); + break; + + case TGSI_OPCODE_XOR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_xor( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SAD: + assert (0); + break; + + case TGSI_OPCODE_TXF: + assert (0); + break; + + case TGSI_OPCODE_TXQ: + assert (0); + break; + + case TGSI_OPCODE_CONT: + assert (0); + break; + + case TGSI_OPCODE_EMIT: + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; + break; + + case TGSI_OPCODE_ENDPRIM: + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; + break; + + default: + assert( 0 ); + } +} + + +#if !defined(XSTDCALL) +#if defined(WIN32) +#define XSTDCALL __stdcall +#else +#define XSTDCALL +#endif +#endif + +typedef void (XSTDCALL *fp_function) (const struct tgsi_exec_vector *input, + struct tgsi_exec_vector *output, + GLfloat (*constant)[4], + struct tgsi_exec_vector *temporary); + +void +tgsi_exec_machine_run2( + struct tgsi_exec_machine *mach, + struct tgsi_exec_labels *labels ) +{ +#if MESA + GET_CURRENT_CONTEXT(ctx); + GLuint i; +#endif + +#if XXX_SSE + fp_function function; + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + + function = (fp_function) x86_get_func (&mach->Function); + + function (mach->Inputs, + mach->Outputs, + mach->Consts, + mach->Temps); +#else + struct tgsi_parse_context parse; + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; + + if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; + mach->Primitives[0] = 0; + } + + tgsi_parse_init( &parse, mach->Tokens ); + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + exec_instruction( mach, &parse.FullToken.FullInstruction, labels, &parse.Position ); + break; + default: + assert( 0 ); + } + } + tgsi_parse_free (&parse); +#endif + +#if MESA + if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { + /* + * Scale back depth component. + */ + for (i = 0; i < 4; i++) + mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; + } +#endif +} + diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h new file mode 100644 index 0000000000..abce6ee924 --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -0,0 +1,137 @@ +#if !defined TGSI_EXEC_H +#define TGSI_EXEC_H + +#if 0 +#include "x86/rtasm/x86sse.h" +#endif + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +union tgsi_exec_channel +{ + GLfloat f[4]; + GLint i[4]; + GLuint u[4]; +}; + +struct tgsi_exec_vector +{ + union tgsi_exec_channel xyzw[4]; +}; + +struct tgsi_sampler_state +{ + GLboolean NeedLambda; + GLboolean NeedLodBias; /* if NeedLambda */ + GLboolean NeedLambdaClamp; /* if NeedLambda */ + GLfloat LodBias; /* if NeedLodBias */ + GLfloat MinLod; /* if NeedLambdaClamp */ + GLfloat MaxLod; /* if NeedLambdaClamp */ + GLfloat ImageWidth; + GLfloat ImageHeight; + GLfloat ImageDepth; +}; + +struct tgsi_exec_labels +{ + GLuint labels[128][2]; + GLuint count; +}; + +#define TGSI_EXEC_TEMP_00000000_I 32 +#define TGSI_EXEC_TEMP_00000000_C 0 + +#define TGSI_EXEC_TEMP_7FFFFFFF_I 32 +#define TGSI_EXEC_TEMP_7FFFFFFF_C 1 + +#define TGSI_EXEC_TEMP_80000000_I 32 +#define TGSI_EXEC_TEMP_80000000_C 2 + +#define TGSI_EXEC_TEMP_FFFFFFFF_I 32 +#define TGSI_EXEC_TEMP_FFFFFFFF_C 3 + +#define TGSI_EXEC_TEMP_ONE_I 33 +#define TGSI_EXEC_TEMP_ONE_C 0 + +#define TGSI_EXEC_TEMP_TWO_I 33 +#define TGSI_EXEC_TEMP_TWO_C 1 + +#define TGSI_EXEC_TEMP_128_I 33 +#define TGSI_EXEC_TEMP_128_C 2 + +#define TGSI_EXEC_TEMP_MINUS_128_I 33 +#define TGSI_EXEC_TEMP_MINUS_128_C 3 + +#define TGSI_EXEC_TEMP_KILMASK_I 34 +#define TGSI_EXEC_TEMP_KILMASK_C 0 + +#define TGSI_EXEC_TEMP_OUTPUT_I 34 +#define TGSI_EXEC_TEMP_OUTPUT_C 1 + +#define TGSI_EXEC_TEMP_PRIMITIVE_I 34 +#define TGSI_EXEC_TEMP_PRIMITIVE_C 2 + +#define TGSI_EXEC_TEMP_R0 35 + +#define TGSI_EXEC_NUM_TEMPS (32 + 4) +#define TGSI_EXEC_NUM_ADDRS 1 + +struct tgsi_exec_machine +{ + /* + * 32 program temporaries + * 4 internal temporaries + * 1 address + * 1 temporary of padding to align to 16 bytes + */ + struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS + 1]; + + /* + * This will point to _Temps after aligning to 16B boundary. + */ + struct tgsi_exec_vector *Temps; + struct tgsi_exec_vector *Addrs; + + struct tgsi_sampler_state Samplers[16]; + + GLfloat Imms[256][4]; + GLuint ImmLimit; + GLfloat (*Consts)[4]; + const struct tgsi_exec_vector *Inputs; + struct tgsi_exec_vector *Outputs; + struct tgsi_token *Tokens; + GLuint Processor; + + GLuint *Primitives; +#if XXX_SSE + struct x86_function Function; +#endif +}; + +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach, + struct tgsi_token *tokens ); + +void +tgsi_exec_prepare( + struct tgsi_exec_machine *mach, + struct tgsi_exec_labels *labels ); + +void +tgsi_exec_machine_run( + struct tgsi_exec_machine *mach ); + +void +tgsi_exec_machine_run2( + struct tgsi_exec_machine *mach, + struct tgsi_exec_labels *labels ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_EXEC_H + diff --git a/src/mesa/pipe/tgsi/core/tgsi_parse.c b/src/mesa/pipe/tgsi/core/tgsi_parse.c new file mode 100644 index 0000000000..df34fa7d7e --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_parse.c @@ -0,0 +1,284 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ) +{ + full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; +} + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ) +{ + if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) + free( full_token->FullImmediate.u.Pointer ); +} + +GLuint +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ) +{ + ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; + if( ctx->FullVersion.Version.MajorVersion > 1 ) { + return TGSI_PARSE_ERROR; + } + + ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; + if( ctx->FullHeader.Header.HeaderSize >= 2 ) { + ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; + } + else { + ctx->FullHeader.Processor = tgsi_default_processor(); + } + + ctx->Tokens = tokens; + ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; + + tgsi_full_token_init( &ctx->FullToken ); + + return TGSI_PARSE_OK; +} + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ) +{ + tgsi_full_token_free( &ctx->FullToken ); +} + +GLuint +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ) +{ + return ctx->Position >= + 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; +} + +static void +next_token( + struct tgsi_parse_context *ctx, + void *token ) +{ + assert( !tgsi_parse_end_of_tokens( ctx ) ); + + *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; +} + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ) +{ + struct tgsi_token token; + GLuint i; + + tgsi_full_token_free( &ctx->FullToken ); + tgsi_full_token_init( &ctx->FullToken ); + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; + + *decl = tgsi_default_full_declaration(); + decl->Declaration = *(struct tgsi_declaration *) &token; + + switch( decl->Declaration.Type ) { + case TGSI_DECLARE_RANGE: + next_token( ctx, &decl->u.DeclarationRange ); + break; + + case TGSI_DECLARE_MASK: + next_token( ctx, &decl->u.DeclarationMask ); + break; + + default: + assert (0); + } + + if( decl->Declaration.Interpolate ) { + next_token( ctx, &decl->Interpolation ); + } + + break; + } + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; + + *imm = tgsi_default_full_immediate(); + imm->Immediate = *(struct tgsi_immediate *) &token; + + assert( !imm->Immediate.Extended ); + + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + imm->u.Pointer = malloc( + sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + next_token( ctx, &imm->u.ImmediateFloat32[i] ); + } + break; + + default: + assert( 0 ); + } + + break; + } + + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; + GLuint extended; + + *inst = tgsi_default_full_instruction(); + inst->Instruction = *(struct tgsi_instruction *) &token; + + extended = inst->Instruction.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_INSTRUCTION_EXT_TYPE_NV: + inst->InstructionExtNv = + *(struct tgsi_instruction_ext_nv *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_LABEL: + inst->InstructionExtLabel = + *(struct tgsi_instruction_ext_label *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: + inst->InstructionExtTexture = + *(struct tgsi_instruction_ext_texture *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + GLuint extended; + + next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); + assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); + + extended = inst->FullDstRegisters[i].DstRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: + inst->FullDstRegisters[i].DstRegisterExtConcode = + *(struct tgsi_dst_register_ext_concode *) &token; + break; + + case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: + inst->FullDstRegisters[i].DstRegisterExtModulate = + *(struct tgsi_dst_register_ext_modulate *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + } + + assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + GLuint extended; + + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); + + extended = inst->FullSrcRegisters[i].SrcRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: + inst->FullSrcRegisters[i].SrcRegisterExtSwz = + *(struct tgsi_src_register_ext_swz *) &token; + break; + + case TGSI_SRC_REGISTER_EXT_TYPE_MOD: + inst->FullSrcRegisters[i].SrcRegisterExtMod = + *(struct tgsi_src_register_ext_mod *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + + if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); + + /* + * No support for multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); + + if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + } + } + + break; + } + + default: + assert( 0 ); + } +} + diff --git a/src/mesa/pipe/tgsi/core/tgsi_parse.h b/src/mesa/pipe/tgsi/core/tgsi_parse.h new file mode 100644 index 0000000000..61ad0669b1 --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_parse.h @@ -0,0 +1,120 @@ +#if !defined TGSI_PARSE_H +#define TGSI_PARSE_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_full_version +{ + struct tgsi_version Version; +}; + +struct tgsi_full_header +{ + struct tgsi_header Header; + struct tgsi_processor Processor; +}; + +struct tgsi_full_dst_register +{ + struct tgsi_dst_register DstRegister; + struct tgsi_dst_register_ext_concode DstRegisterExtConcode; + struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; +}; + +struct tgsi_full_src_register +{ + struct tgsi_src_register SrcRegister; + struct tgsi_src_register_ext_swz SrcRegisterExtSwz; + struct tgsi_src_register_ext_mod SrcRegisterExtMod; + struct tgsi_src_register SrcRegisterInd; + struct tgsi_dimension SrcRegisterDim; + struct tgsi_src_register SrcRegisterDimInd; +}; + +struct tgsi_full_declaration +{ + struct tgsi_declaration Declaration; + union + { + struct tgsi_declaration_range DeclarationRange; + struct tgsi_declaration_mask DeclarationMask; + } u; + struct tgsi_declaration_interpolation Interpolation; +}; + +struct tgsi_full_immediate +{ + struct tgsi_immediate Immediate; + union + { + void *Pointer; + struct tgsi_immediate_float32 *ImmediateFloat32; + } u; +}; + +#define TGSI_FULL_MAX_DST_REGISTERS 2 +#define TGSI_FULL_MAX_SRC_REGISTERS 3 + +struct tgsi_full_instruction +{ + struct tgsi_instruction Instruction; + struct tgsi_instruction_ext_nv InstructionExtNv; + struct tgsi_instruction_ext_label InstructionExtLabel; + struct tgsi_instruction_ext_texture InstructionExtTexture; + struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; + struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; +}; + +union tgsi_full_token +{ + struct tgsi_token Token; + struct tgsi_full_declaration FullDeclaration; + struct tgsi_full_immediate FullImmediate; + struct tgsi_full_instruction FullInstruction; +}; + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ); + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ); + +struct tgsi_parse_context +{ + const struct tgsi_token *Tokens; + GLuint Position; + struct tgsi_full_version FullVersion; + struct tgsi_full_header FullHeader; + union tgsi_full_token FullToken; +}; + +#define TGSI_PARSE_OK 0 +#define TGSI_PARSE_ERROR 1 + +GLuint +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ); + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ); + +GLuint +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ); + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_PARSE_H + diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h new file mode 100644 index 0000000000..becdd480fb --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -0,0 +1,1045 @@ +#if !defined TGSI_TOKEN_H +#define TGSI_TOKEN_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_version +{ + GLuint MajorVersion : 8; + GLuint MinorVersion : 8; + GLuint Padding : 16; +}; + +struct tgsi_header +{ + GLuint HeaderSize : 8; + GLuint BodySize : 24; +}; + +#define TGSI_PROCESSOR_FRAGMENT 0 +#define TGSI_PROCESSOR_VERTEX 1 +#define TGSI_PROCESSOR_GEOMETRY 2 + +struct tgsi_processor +{ + GLuint Processor : 4; /* TGSI_PROCESSOR_ */ + GLuint Padding : 28; +}; + +#define TGSI_TOKEN_TYPE_DECLARATION 0 +#define TGSI_TOKEN_TYPE_IMMEDIATE 1 +#define TGSI_TOKEN_TYPE_INSTRUCTION 2 + +struct tgsi_token +{ + GLuint Type : 4; /* TGSI_TOKEN_TYPE_ */ + GLuint Size : 8; /* UINT */ + GLuint Padding : 19; + GLuint Extended : 1; /* BOOL */ +}; + +#define TGSI_FILE_NULL 0 +#define TGSI_FILE_CONSTANT 1 +#define TGSI_FILE_INPUT 2 +#define TGSI_FILE_OUTPUT 3 +#define TGSI_FILE_TEMPORARY 4 +#define TGSI_FILE_SAMPLER 5 +#define TGSI_FILE_ADDRESS 6 +#define TGSI_FILE_IMMEDIATE 7 + +#define TGSI_DECLARE_RANGE 0 +#define TGSI_DECLARE_MASK 1 + +struct tgsi_declaration +{ + GLuint Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ + GLuint Size : 8; /* UINT */ + GLuint File : 4; /* TGSI_FILE_ */ + GLuint Declare : 4; /* TGSI_DECLARE_ */ + GLuint Interpolate : 1; /* BOOL */ + GLuint Padding : 10; + GLuint Extended : 1; /* BOOL */ +}; + +struct tgsi_declaration_range +{ + GLuint First : 16; /* UINT */ + GLuint Last : 16; /* UINT */ +}; + +struct tgsi_declaration_mask +{ + GLuint Mask : 32; /* UINT */ +}; + +#define TGSI_INTERPOLATE_CONSTANT 0 +#define TGSI_INTERPOLATE_LINEAR 1 +#define TGSI_INTERPOLATE_PERSPECTIVE 2 + +struct tgsi_declaration_interpolation +{ + GLuint Interpolate : 4; /* TGSI_INTERPOLATE_ */ + GLuint Padding : 28; +}; + +#define TGSI_IMM_FLOAT32 0 + +struct tgsi_immediate +{ + GLuint Type : 4; /* TGSI_TOKEN_TYPE_IMMEDIATE */ + GLuint Size : 8; /* UINT */ + GLuint DataType : 4; /* TGSI_IMM_ */ + GLuint Padding : 15; + GLuint Extended : 1; /* BOOL */ +}; + +struct tgsi_immediate_float32 +{ + GLfloat Float; +}; + +/* + * GL_NV_vertex_program + */ +#define TGSI_OPCODE_ARL 0 +#define TGSI_OPCODE_MOV 1 +#define TGSI_OPCODE_LIT 2 +#define TGSI_OPCODE_RCP 3 +#define TGSI_OPCODE_RSQ 4 +#define TGSI_OPCODE_EXP 5 +#define TGSI_OPCODE_LOG 6 +#define TGSI_OPCODE_MUL 7 +#define TGSI_OPCODE_ADD 8 +#define TGSI_OPCODE_DP3 9 +#define TGSI_OPCODE_DP4 10 +#define TGSI_OPCODE_DST 11 +#define TGSI_OPCODE_MIN 12 +#define TGSI_OPCODE_MAX 13 +#define TGSI_OPCODE_SLT 14 +#define TGSI_OPCODE_SGE 15 +#define TGSI_OPCODE_MAD 16 + +/* + * GL_ATI_fragment_shader + */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_MUL */ +#define TGSI_OPCODE_SUB 17 +#define TGSI_OPCODE_DOT3 TGSI_OPCODE_DP3 +#define TGSI_OPCODE_DOT4 TGSI_OPCODE_DP4 +/* TGSI_OPCODE_MAD */ +#define TGSI_OPCODE_LERP 18 +#define TGSI_OPCODE_CND 19 +#define TGSI_OPCODE_CND0 20 +#define TGSI_OPCODE_DOT2ADD 21 + +/* + * GL_EXT_vertex_shader + */ +#define TGSI_OPCODE_INDEX 22 +#define TGSI_OPCODE_NEGATE 23 +/* TGSI_OPCODE_DOT3 */ +/* TGSI_OPCODE_DOT4 */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_MADD TGSI_OPCODE_MAD +#define TGSI_OPCODE_FRAC 24 +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +#define TGSI_OPCODE_SETGE TGSI_OPCODE_SGE +#define TGSI_OPCODE_SETLT TGSI_OPCODE_SLT +#define TGSI_OPCODE_CLAMP 25 +#define TGSI_OPCODE_FLOOR 26 +#define TGSI_OPCODE_ROUND 27 +#define TGSI_OPCODE_EXPBASE2 28 +#define TGSI_OPCODE_LOGBASE2 29 +#define TGSI_OPCODE_POWER 30 +#define TGSI_OPCODE_RECIP TGSI_OPCODE_RCP +#define TGSI_OPCODE_RECIPSQRT TGSI_OPCODE_RSQ +/* TGSI_OPCODE_SUB */ +#define TGSI_OPCODE_CROSSPRODUCT 31 +#define TGSI_OPCODE_MULTIPLYMATRIX 32 +/* TGSI_OPCODE_MOV */ + +/* + * GL_NV_vertex_program1_1 + */ +/* TGSI_OPCODE_ARL */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_LIT */ +#define TGSI_OPCODE_ABS 33 +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_LOG */ +#define TGSI_OPCODE_RCC 34 +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +#define TGSI_OPCODE_DPH 35 +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ + +/* + * GL_NV_fragment_program + */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_COS 36 +#define TGSI_OPCODE_DDX 37 +#define TGSI_OPCODE_DDY 38 +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DST */ +#define TGSI_OPCODE_EX2 TGSI_OPCODE_EXPBASE2 +#define TGSI_OPCODE_FLR TGSI_OPCODE_FLOOR +#define TGSI_OPCODE_FRC TGSI_OPCODE_FRAC +#define TGSI_OPCODE_KIL 39 +#define TGSI_OPCODE_LG2 TGSI_OPCODE_LOGBASE2 +/* TGSI_OPCODE_LIT */ +#define TGSI_OPCODE_LRP TGSI_OPCODE_LERP +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +#define TGSI_OPCODE_PK2H 40 +#define TGSI_OPCODE_PK2US 41 +#define TGSI_OPCODE_PK4B 42 +#define TGSI_OPCODE_PK4UB 43 +#define TGSI_OPCODE_POW TGSI_OPCODE_POWER +/* TGSI_OPCODE_RCP */ +#define TGSI_OPCODE_RFL 44 +/* TGSI_OPCODE_RSQ */ +#define TGSI_OPCODE_SEQ 45 +#define TGSI_OPCODE_SFL 46 +/* TGSI_OPCODE_SGE */ +#define TGSI_OPCODE_SGT 47 +#define TGSI_OPCODE_SIN 48 +#define TGSI_OPCODE_SLE 49 +/* TGSI_OPCODE_SLT */ +#define TGSI_OPCODE_SNE 50 +#define TGSI_OPCODE_STR 51 +/* TGSI_OPCODE_SUB */ +#define TGSI_OPCODE_TEX 52 +#define TGSI_OPCODE_TXD 53 +/* TGSI_OPCODE_TXP - use TGSI_OPCODE_TEX */ +#define TGSI_OPCODE_UP2H 54 +#define TGSI_OPCODE_UP2US 55 +#define TGSI_OPCODE_UP4B 56 +#define TGSI_OPCODE_UP4UB 57 +#define TGSI_OPCODE_X2D 58 + +/* + * GL_NV_vertex_program2 + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_ARA 59 +/* TGSI_OPCODE_ARL */ +#define TGSI_OPCODE_ARR 60 +#define TGSI_OPCODE_BRA 61 +#define TGSI_OPCODE_CAL 62 +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCC */ +/* TGSI_OPCODE_RCP */ +#define TGSI_OPCODE_RET 63 +/* TGSI_OPCODE_RSQNV */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SNE */ +#define TGSI_OPCODE_SSG 64 +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_SUB */ + +/* + * GL_ARB_vertex_program + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_ARL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +#define TGSI_OPCODE_SWZ TGSI_OPCODE_MOV +#define TGSI_OPCODE_XPD TGSI_OPCODE_CROSSPRODUCT + +/* + * GL_ARB_fragment_program + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_CMP 65 +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +#define TGSI_OPCODE_SCS 66 +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXP */ +#define TGSI_OPCODE_TXB 67 +/* TGSI_OPCODE_KIL */ + +/* + * GL_NV_fragment_program_option + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_DDX */ +/* TGSI_OPCODE_DDY */ +/* TGSI_OPCODE_PK2H */ +/* TGSI_OPCODE_PK2US */ +/* TGSI_OPCODE_PK4B */ +/* TGSI_OPCODE_PK4UB */ +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SCS */ +/* TGSI_OPCODE_UP2H */ +/* TGSI_OPCODE_UP2US */ +/* TGSI_OPCODE_UP4B */ +/* TGSI_OPCODE_UP4UB */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_RFL */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_CMP */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_X2D */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXP */ +/* TGSI_OPCODE_TXB */ +/* TGSI_OPCODE_KIL */ +/* TGSI_OPCODE_TXD */ + +/* + * GL_NV_fragment_program2 + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_DDX */ +/* TGSI_OPCODE_DDY */ +/* TGSI_OPCODE_PK2H */ +/* TGSI_OPCODE_PK2US */ +/* TGSI_OPCODE_PK4B */ +/* TGSI_OPCODE_PK4UB */ +#define TGSI_OPCODE_NRM 68 +#define TGSI_OPCODE_DIV 69 +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SCS */ +/* TGSI_OPCODE_UP2H */ +/* TGSI_OPCODE_UP2US */ +/* TGSI_OPCODE_UP4B */ +/* TGSI_OPCODE_UP4UB */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_RFL */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +#define TGSI_OPCODE_DP2 70 +/* TGSI_OPCODE_CMP */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_X2D */ +#define TGSI_OPCODE_DP2A TGSI_OPCODE_DOT2ADD +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXP */ +/* TGSI_OPCODE_TXB */ +#define TGSI_OPCODE_TXL 71 +/* TGSI_OPCODE_KIL */ +/* TGSI_OPCODE_TXD */ +/* TGSI_OPCODE_CAL */ +/* TGSI_OPCODE_RET */ +#define TGSI_OPCODE_BRK 72 +#define TGSI_OPCODE_IF 73 +#define TGSI_OPCODE_LOOP 74 +#define TGSI_OPCODE_REP 75 +#define TGSI_OPCODE_ELSE 76 +#define TGSI_OPCODE_ENDIF 77 +#define TGSI_OPCODE_ENDLOOP 78 +#define TGSI_OPCODE_ENDREP 79 + +/* + * GL_NV_vertex_program2_option + */ +/* TGSI_OPCODE_ARL */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_SSG */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_RCC */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_ARR */ +/* TGSI_OPCODE_ARA */ +/* TGSI_OPCODE_BRA */ +/* TGSI_OPCODE_CAL */ +/* TGSI_OPCODE_RET */ + +/* + * GL_NV_vertex_program3 + */ +/* TGSI_OPCODE_ARL */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_SSG */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_RCC */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_ARR */ +/* TGSI_OPCODE_ARA */ +/* TGSI_OPCODE_BRA */ +/* TGSI_OPCODE_CAL */ +/* TGSI_OPCODE_RET */ +#define TGSI_OPCODE_PUSHA 80 +#define TGSI_OPCODE_POPA 81 +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXP */ +/* TGSI_OPCODE_TXB */ +/* TGSI_OPCODE_TXL */ + +/* + * GL_NV_gpu_program4 + */ +/* TGSI_OPCODE_ABS */ +#define TGSI_OPCODE_CEIL 82 +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +#define TGSI_OPCODE_I2F 83 +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +#define TGSI_OPCODE_NOT 84 +/* TGSI_OPCODE_NRM */ +/* TGSI_OPCODE_PK2H */ +/* TGSI_OPCODE_PK2US */ +/* TGSI_OPCODE_PK4B */ +/* TGSI_OPCODE_PK4UB */ +/* TGSI_OPCODE_ROUND */ +/* TGSI_OPCODE_SSG */ +#define TGSI_OPCODE_TRUNC 85 +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_RCC */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SCS */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_UP2H */ +/* TGSI_OPCODE_UP2US */ +/* TGSI_OPCODE_UP4B */ +/* TGSI_OPCODE_UP4UB */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_DIV */ +#define TGSI_OPCODE_SHL 86 +#define TGSI_OPCODE_SHR 87 +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_AND 88 +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +#define TGSI_OPCODE_OR 89 +/* TGSI_OPCODE_RFL */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_DP2 */ +#define TGSI_OPCODE_MOD 90 +#define TGSI_OPCODE_XOR 91 +/* TGSI_OPCODE_CMP */ +/* TGSI_OPCODE_DP2A */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_MAD */ +#define TGSI_OPCODE_SAD 92 +/* TGSI_OPCODE_X2D */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXB */ +#define TGSI_OPCODE_TXF 93 +/* TGSI_OPCODE_TXL */ +/* TGSI_OPCODE_TXP */ +#define TGSI_OPCODE_TXQ 94 +/* TGSI_OPCODE_TXD */ +/* TGSI_OPCODE_CAL */ +/* TGSI_OPCODE_RET */ +/* TGSI_OPCODE_BRK */ +#define TGSI_OPCODE_CONT 95 +/* TGSI_OPCODE_IF */ +/* TGSI_OPCODE_REP */ +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +/* TGSI_OPCODE_ENDREP */ + +/* + * GL_NV_vertex_program4 + */ +/* Same as GL_NV_gpu_program4 */ + +/* + * GL_NV_fragment_program4 + */ +/* Same as GL_NV_gpu_program4 */ +/* TGSI_OPCODE_KIL */ +/* TGSI_OPCODE_DDX */ +/* TGSI_OPCODE_DDY */ + +/* + * GL_NV_geometry_program4 + */ +/* Same as GL_NV_gpu_program4 */ +#define TGSI_OPCODE_EMIT 96 +#define TGSI_OPCODE_ENDPRIM 97 + +#define TGSI_OPCODE_LAST 98 + +#define TGSI_SAT_NONE 0 /* do not saturate */ +#define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ +#define TGSI_SAT_MINUS_PLUS_ONE 2 /* clamp to [-1,1] */ + +/* + * Opcode is the operation code to execute. A given operation defines the + * semantics how the source registers (if any) are interpreted and what is + * written to the destination registers (if any) as a result of execution. + * + * NumDstRegs and NumSrcRegs is the number of destination and source registers, + * respectively. For a given operation code, those numbers are fixed and are + * present here only for convenience. + * + * If Extended is TRUE, it is now executed. + * + * Saturate controls how are final results in destination registers modified. + */ + +struct tgsi_instruction +{ + GLuint Type : 4; /* TGSI_TOKEN_TYPE_INSTRUCTION */ + GLuint Size : 8; /* UINT */ + GLuint Opcode : 8; /* TGSI_OPCODE_ */ + GLuint Saturate : 2; /* TGSI_SAT_ */ + GLuint NumDstRegs : 2; /* UINT */ + GLuint NumSrcRegs : 4; /* UINT */ + GLuint Padding : 3; + GLuint Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows. + * + * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow. + * + * Then, tgsi_instruction::NumSrcRegs of tgsi_src_register follow. + * + * tgsi_instruction::Size contains the total number of words that make the + * instruction, including the instruction word. + */ + +#define TGSI_INSTRUCTION_EXT_TYPE_NV 0 +#define TGSI_INSTRUCTION_EXT_TYPE_LABEL 1 +#define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE 2 + +struct tgsi_instruction_ext +{ + GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_ */ + GLuint Padding : 27; + GLuint Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_NV, it should + * be cast to tgsi_instruction_ext_nv. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_LABEL, it + * should be cast to tgsi_instruction_ext_label. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it + * should be cast to tgsi_instruction_ext_texture. + * + * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext + * follows. + */ + +#define TGSI_PRECISION_DEFAULT 0 +#define TGSI_PRECISION_FLOAT32 1 +#define TGSI_PRECISION_FLOAT16 2 +#define TGSI_PRECISION_FIXED12 3 + +#define TGSI_CC_GT 0 +#define TGSI_CC_EQ 1 +#define TGSI_CC_LT 2 +#define TGSI_CC_UN 3 +#define TGSI_CC_GE 4 +#define TGSI_CC_LE 5 +#define TGSI_CC_NE 6 +#define TGSI_CC_TR 7 +#define TGSI_CC_FL 8 + +#define TGSI_SWIZZLE_X 0 +#define TGSI_SWIZZLE_Y 1 +#define TGSI_SWIZZLE_Z 2 +#define TGSI_SWIZZLE_W 3 + +/* + * Precision controls the precision at which the operation should be executed. + * + * CondDstUpdate enables condition code register writes. When this field is + * TRUE, CondDstIndex specifies the index of the condition code register to + * update. + * + * CondFlowEnable enables conditional execution of the operation. When this + * field is TRUE, CondFlowIndex specifies the index of the condition code + * register to test against CondMask with component swizzle controled by + * CondSwizzleX, CondSwizzleY, CondSwizzleZ and CondSwizzleW. If the test fails, + * the operation is not executed. + */ + +struct tgsi_instruction_ext_nv +{ + GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_NV */ + GLuint Precision : 4; /* TGSI_PRECISION_ */ + GLuint CondDstIndex : 4; /* UINT */ + GLuint CondFlowIndex : 4; /* UINT */ + GLuint CondMask : 4; /* TGSI_CC_ */ + GLuint CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + GLuint CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + GLuint CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + GLuint CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + GLuint CondDstUpdate : 1; /* BOOL */ + GLuint CondFlowEnable : 1; /* BOOL */ + GLuint Padding : 1; + GLuint Extended : 1; /* BOOL */ +}; + +struct tgsi_instruction_ext_label +{ + GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ + GLuint Label : 24; /* UINT */ + GLuint Target : 1; /* BOOL */ + GLuint Padding : 2; + GLuint Extended : 1; /* BOOL */ +}; + +#define TGSI_TEXTURE_UNKNOWN 0 +#define TGSI_TEXTURE_1D 1 +#define TGSI_TEXTURE_2D 2 +#define TGSI_TEXTURE_3D 3 +#define TGSI_TEXTURE_CUBE 4 +#define TGSI_TEXTURE_RECT 5 +#define TGSI_TEXTURE_SHADOW1D 6 +#define TGSI_TEXTURE_SHADOW2D 7 +#define TGSI_TEXTURE_SHADOWRECT 8 + +struct tgsi_instruction_ext_texture +{ + GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */ + GLuint Texture : 8; /* TGSI_TEXTURE_ */ + GLuint Padding : 19; + GLuint Extended : 1; /* BOOL */ +}; + +/* + * File specifies the register array to access. + * + * Index specifies the element number of a register in the register file. + * + * If Indirect is TRUE, Index should be offset by the X component of a source + * register that follows. The register can be now fetched into local storage + * for further processing. + * + * If Negate is TRUE, all components of the fetched register are negated. + * + * The fetched register components are swizzled according to SwizzleX, SwizzleY, + * SwizzleZ and SwizzleW. + * + * If Extended is TRUE, any further modifications to the source register are + * made to this temporary storage. + */ + +struct tgsi_src_register +{ + GLuint File : 4; /* TGSI_FILE_ */ + GLuint SwizzleX : 2; /* TGSI_SWIZZLE_ */ + GLuint SwizzleY : 2; /* TGSI_SWIZZLE_ */ + GLuint SwizzleZ : 2; /* TGSI_SWIZZLE_ */ + GLuint SwizzleW : 2; /* TGSI_SWIZZLE_ */ + GLuint Negate : 1; /* BOOL */ + GLuint Indirect : 1; /* BOOL */ + GLuint Dimension : 1; /* BOOL */ + GLint Index : 16; /* SINT */ + GLuint Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_src_register::Extended is TRUE, tgsi_src_register_ext follows. + * + * Then, if tgsi_src_register::Indirect is TRUE, another tgsi_src_register + * follows. + * + * Then, if tgsi_src_register::Dimension is TRUE, tgsi_dimension follows. + */ + +#define TGSI_SRC_REGISTER_EXT_TYPE_SWZ 0 +#define TGSI_SRC_REGISTER_EXT_TYPE_MOD 1 + +struct tgsi_src_register_ext +{ + GLuint Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_ */ + GLuint Padding : 27; + GLuint Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_SWZ, + * it should be cast to tgsi_src_register_ext_extswz. + * + * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_MOD, + * it should be cast to tgsi_src_register_ext_mod. + * + * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext + * follows. + */ + +#define TGSI_EXTSWIZZLE_X TGSI_SWIZZLE_X +#define TGSI_EXTSWIZZLE_Y TGSI_SWIZZLE_Y +#define TGSI_EXTSWIZZLE_Z TGSI_SWIZZLE_Z +#define TGSI_EXTSWIZZLE_W TGSI_SWIZZLE_W +#define TGSI_EXTSWIZZLE_ZERO 4 +#define TGSI_EXTSWIZZLE_ONE 5 + +/* + * ExtSwizzleX, ExtSwizzleY, ExtSwizzleZ and ExtSwizzleW swizzle the source + * register in an extended manner. + * + * NegateX, NegateY, NegateZ and NegateW negate individual components of the + * source register. + * + * ExtDivide specifies which component is used to divide all components of the + * source register. + */ + +struct tgsi_src_register_ext_swz +{ + GLuint Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */ + GLuint ExtSwizzleX : 4; /* TGSI_EXTSWIZZLE_ */ + GLuint ExtSwizzleY : 4; /* TGSI_EXTSWIZZLE_ */ + GLuint ExtSwizzleZ : 4; /* TGSI_EXTSWIZZLE_ */ + GLuint ExtSwizzleW : 4; /* TGSI_EXTSWIZZLE_ */ + GLuint NegateX : 1; /* BOOL */ + GLuint NegateY : 1; /* BOOL */ + GLuint NegateZ : 1; /* BOOL */ + GLuint NegateW : 1; /* BOOL */ + GLuint ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ + GLuint Padding : 3; + GLuint Extended : 1; /* BOOL */ +}; + +/* + * If Complement is TRUE, the source register is modified by subtracting it + * from 1.0. + * + * If Bias is TRUE, the source register is modified by subtracting 0.5 from it. + * + * If Scale2X is TRUE, the source register is modified by multiplying it by 2.0. + * + * If Absolute is TRUE, the source register is modified by removing the sign. + * + * If Negate is TRUE, the source register is modified by negating it. + */ + +struct tgsi_src_register_ext_mod +{ + GLuint Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */ + GLuint Complement : 1; /* BOOL */ + GLuint Bias : 1; /* BOOL */ + GLuint Scale2X : 1; /* BOOL */ + GLuint Absolute : 1; /* BOOL */ + GLuint Negate : 1; /* BOOL */ + GLuint Padding : 22; + GLuint Extended : 1; /* BOOL */ +}; + +struct tgsi_dimension +{ + GLuint Indirect : 1; /* BOOL */ + GLuint Dimension : 1; /* BOOL */ + GLuint Padding : 13; + GLint Index : 16; /* SINT */ + GLuint Extended : 1; /* BOOL */ +}; + +#define TGSI_WRITEMASK_NONE 0x00 +#define TGSI_WRITEMASK_X 0x01 +#define TGSI_WRITEMASK_Y 0x02 +#define TGSI_WRITEMASK_XY 0x03 +#define TGSI_WRITEMASK_Z 0x04 +#define TGSI_WRITEMASK_XZ 0x05 +#define TGSI_WRITEMASK_YZ 0x06 +#define TGSI_WRITEMASK_XYZ 0x07 +#define TGSI_WRITEMASK_W 0x08 +#define TGSI_WRITEMASK_XW 0x09 +#define TGSI_WRITEMASK_YW 0x0A +#define TGSI_WRITEMASK_XYW 0x0B +#define TGSI_WRITEMASK_ZW 0x0C +#define TGSI_WRITEMASK_XZW 0x0D +#define TGSI_WRITEMASK_YZW 0x0E +#define TGSI_WRITEMASK_XYZW 0x0F + +struct tgsi_dst_register +{ + GLuint File : 4; /* TGSI_FILE_ */ + GLuint WriteMask : 4; /* TGSI_WRITEMASK_ */ + GLuint Indirect : 1; /* BOOL */ + GLuint Dimension : 1; /* BOOL */ + GLint Index : 16; /* SINT */ + GLuint Padding : 5; + GLuint Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_dst_register::Extended is TRUE, tgsi_dst_register_ext follows. + * + * Then, if tgsi_dst_register::Indirect is TRUE, tgsi_src_register follows. + */ + +#define TGSI_DST_REGISTER_EXT_TYPE_CONDCODE 0 +#define TGSI_DST_REGISTER_EXT_TYPE_MODULATE 1 + +struct tgsi_dst_register_ext +{ + GLuint Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_ */ + GLuint Padding : 27; + GLuint Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_CONDCODE, + * it should be cast to tgsi_dst_register_ext_condcode. + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_MODULATE, + * it should be cast to tgsi_dst_register_ext_modulate. + * + * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext + * follows. + */ + +struct tgsi_dst_register_ext_concode +{ + GLuint Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ + GLuint CondMask : 4; /* TGSI_CC_ */ + GLuint CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + GLuint CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + GLuint CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + GLuint CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + GLuint CondSrcIndex : 4; /* UINT */ + GLuint Padding : 11; + GLuint Extended : 1; /* BOOL */ +}; + +#define TGSI_MODULATE_1X 0 +#define TGSI_MODULATE_2X 1 +#define TGSI_MODULATE_4X 2 +#define TGSI_MODULATE_8X 3 +#define TGSI_MODULATE_HALF 4 +#define TGSI_MODULATE_QUARTER 5 +#define TGSI_MODULATE_EIGHTH 6 + +struct tgsi_dst_register_ext_modulate +{ + GLuint Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */ + GLuint Modulate : 4; /* TGSI_MODULATE_ */ + GLuint Padding : 23; + GLuint Extended : 1; /* BOOL */ +}; + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_TOKEN_H + diff --git a/src/mesa/pipe/tgsi/core/tgsi_util.c b/src/mesa/pipe/tgsi/core/tgsi_util.c new file mode 100644 index 0000000000..2331affdfd --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_util.c @@ -0,0 +1,263 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +void * +tgsi_align_128bit( + void *unaligned ) +{ + GLuint *ptr, addr; + + ptr = (GLuint *) unaligned; + addr = (*(GLuint *) &ptr + 15) & ~15; + return *(void **) &addr; +} + +GLuint +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + GLuint component ) +{ + switch( component ) { + case 0: + return reg->SwizzleX; + case 1: + return reg->SwizzleY; + case 2: + return reg->SwizzleZ; + case 3: + return reg->SwizzleW; + default: + assert( 0 ); + } + return 0; +} + +GLuint +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + GLuint component ) +{ + switch( component ) { + case 0: + return reg->ExtSwizzleX; + case 1: + return reg->ExtSwizzleY; + case 2: + return reg->ExtSwizzleZ; + case 3: + return reg->ExtSwizzleW; + default: + assert( 0 ); + } + return 0; +} + +GLuint +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + GLuint component ) +{ + GLuint swizzle; + + /* + * First, calculate the extended swizzle for a given channel. This will give + * us either a channel index into the simple swizzle or a constant 1 or 0. + */ + swizzle = tgsi_util_get_src_register_extswizzle( + ®->SrcRegisterExtSwz, + component ); + + assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); + assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); + assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); + assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); + + /* + * Second, calculate the simple swizzle for the unswizzled channel index. + * Leave the constants intact, they are not affected by the simple swizzle. + */ + if( swizzle <= TGSI_SWIZZLE_W ) { + swizzle = tgsi_util_get_src_register_swizzle( + ®->SrcRegister, + swizzle ); + } + + return swizzle; +} + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + GLuint swizzle, + GLuint component ) +{ + switch( component ) { + case 0: + reg->SwizzleX = swizzle; + break; + case 1: + reg->SwizzleY = swizzle; + break; + case 2: + reg->SwizzleZ = swizzle; + break; + case 3: + reg->SwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + GLuint swizzle, + GLuint component ) +{ + switch( component ) { + case 0: + reg->ExtSwizzleX = swizzle; + break; + case 1: + reg->ExtSwizzleY = swizzle; + break; + case 2: + reg->ExtSwizzleZ = swizzle; + break; + case 3: + reg->ExtSwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +GLuint +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + GLuint component ) +{ + switch( component ) { + case 0: + return reg->NegateX; + case 1: + return reg->NegateY; + case 2: + return reg->NegateZ; + case 3: + return reg->NegateW; + default: + assert( 0 ); + } + return 0; +} + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + GLuint negate, + GLuint component ) +{ + switch( component ) { + case 0: + reg->NegateX = negate; + break; + case 1: + reg->NegateY = negate; + break; + case 2: + reg->NegateZ = negate; + break; + case 3: + reg->NegateW = negate; + break; + default: + assert( 0 ); + } +} + +GLuint +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + GLuint component ) +{ + GLuint sign_mode; + + if( reg->SrcRegisterExtMod.Absolute ) { + /* Consider only the post-abs negation. */ + + if( reg->SrcRegisterExtMod.Negate ) { + sign_mode = TGSI_UTIL_SIGN_SET; + } + else { + sign_mode = TGSI_UTIL_SIGN_CLEAR; + } + } + else { + /* Accumulate the three negations. */ + + GLuint negate; + + negate = reg->SrcRegister.Negate; + if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { + negate = !negate; + } + if( reg->SrcRegisterExtMod.Negate ) { + negate = !negate; + } + + if( negate ) { + sign_mode = TGSI_UTIL_SIGN_TOGGLE; + } + else { + sign_mode = TGSI_UTIL_SIGN_KEEP; + } + } + + return sign_mode; +} + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + GLuint sign_mode ) +{ + reg->SrcRegisterExtSwz.NegateX = 0; + reg->SrcRegisterExtSwz.NegateY = 0; + reg->SrcRegisterExtSwz.NegateZ = 0; + reg->SrcRegisterExtSwz.NegateW = 0; + + switch (sign_mode) + { + case TGSI_UTIL_SIGN_CLEAR: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_SET: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 1; + break; + + case TGSI_UTIL_SIGN_TOGGLE: + reg->SrcRegister.Negate = 1; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_KEEP: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + default: + assert( 0 ); + } +} + diff --git a/src/mesa/pipe/tgsi/core/tgsi_util.h b/src/mesa/pipe/tgsi/core/tgsi_util.h new file mode 100644 index 0000000000..70c48690c5 --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_util.h @@ -0,0 +1,70 @@ +#if !defined TGSI_UTIL_H +#define TGSI_UTIL_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +void * +tgsi_align_128bit( + void *unaligned ); + +GLuint +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + GLuint component ); + +GLuint +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + GLuint component); + +GLuint +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + GLuint component ); + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + GLuint swizzle, + GLuint component ); + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + GLuint swizzle, + GLuint component ); + +GLuint +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + GLuint component ); + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + GLuint negate, + GLuint component ); + +#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ +#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ +#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ +#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ + +GLuint +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + GLuint component ); + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + GLuint sign_mode ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_UTIL_H + diff --git a/src/mesa/pipe/tgsi/mesa/Makefile b/src/mesa/pipe/tgsi/mesa/Makefile new file mode 100644 index 0000000000..eb8b14e0e8 --- /dev/null +++ b/src/mesa/pipe/tgsi/mesa/Makefile @@ -0,0 +1,3 @@ +default: + cd ../../.. ; make + diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c new file mode 100644 index 0000000000..e37778fe41 --- /dev/null +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -0,0 +1,499 @@ +#include "tgsi_platform.h" +#include "tgsi_mesa.h" + +/* + * Map mesa register file to SBIR register file. + */ +static GLuint +map_register_file( enum register_file file ) +{ + switch (file) { + case PROGRAM_UNDEFINED: + return TGSI_FILE_NULL; + case PROGRAM_TEMPORARY: + return TGSI_FILE_TEMPORARY; + //case PROGRAM_LOCAL_PARAM: + //case PROGRAM_ENV_PARAM: + case PROGRAM_STATE_VAR: + //case PROGRAM_NAMED_PARAM: + case PROGRAM_CONSTANT: + return TGSI_FILE_CONSTANT; + case PROGRAM_INPUT: + return TGSI_FILE_INPUT; + case PROGRAM_OUTPUT: + return TGSI_FILE_OUTPUT; + case PROGRAM_ADDRESS: + return TGSI_FILE_ADDRESS; + default: + assert (0); + return TGSI_FILE_NULL; + } +} + +/* + * Map mesa register file index to SBIR index. + * Take special care when processing input and output indices. + */ +static GLuint +map_register_file_index( GLuint processor, + GLuint file, + GLuint index, + GLuint usage_bitmask ) +{ + GLuint mapped_index; + GLuint i; + + switch (file) + { + case TGSI_FILE_INPUT: + assert (index < 32); + assert (usage_bitmask & (1 << index)); + mapped_index = 0; + for (i = 0; i < index; i++) { + if (usage_bitmask & (1 << i)) + mapped_index++; + } + break; + + case TGSI_FILE_OUTPUT: + assert (usage_bitmask == 0); + if (processor == TGSI_PROCESSOR_FRAGMENT) { + if (index == FRAG_RESULT_DEPR) { + mapped_index = 0; + } else { + assert (index == FRAG_RESULT_COLR); + mapped_index = index + 1; + } + } else { + mapped_index = index; + } + break; + + default: + mapped_index = index; + } + + return mapped_index; +} + +/* + * Map mesa texture target to SBIR texture target. + */ +static GLuint +map_texture_target( GLuint textarget ) +{ + switch (textarget) { + case TEXTURE_1D_INDEX: + return TGSI_TEXTURE_1D; + case TEXTURE_2D_INDEX: + return TGSI_TEXTURE_2D; + case TEXTURE_3D_INDEX: + return TGSI_TEXTURE_3D; + case TEXTURE_CUBE_INDEX: + return TGSI_TEXTURE_CUBE; + case TEXTURE_RECT_INDEX: + return TGSI_TEXTURE_RECT; + default: + assert (0); + } + return TGSI_TEXTURE_1D; +} + +static GLuint +convert_sat( GLuint sat ) +{ + switch (sat) { + case SATURATE_OFF: + return TGSI_SAT_NONE; + case SATURATE_ZERO_ONE: + return TGSI_SAT_ZERO_ONE; + case SATURATE_PLUS_MINUS_ONE: + return TGSI_SAT_MINUS_PLUS_ONE; + default: + assert (0); + return TGSI_SAT_NONE; + } +} + +static GLuint +convert_writemask( GLuint writemask ) +{ + assert (WRITEMASK_X == TGSI_WRITEMASK_X); + assert (WRITEMASK_Y == TGSI_WRITEMASK_Y); + assert (WRITEMASK_Z == TGSI_WRITEMASK_Z); + assert (WRITEMASK_W == TGSI_WRITEMASK_W); + assert ((writemask & ~TGSI_WRITEMASK_XYZW) == 0); + + return writemask; +} + +static GLboolean +compile_instruction( struct prog_instruction *inst, + struct tgsi_full_instruction *fullinst, + GLuint inputs_read, + GLuint processor ) +{ + GLuint i; + struct tgsi_full_dst_register *fulldst; + struct tgsi_full_src_register *fullsrc; + + *fullinst = tgsi_default_full_instruction (); + + fullinst->Instruction.Saturate = convert_sat (inst->SaturateMode); + fullinst->Instruction.NumDstRegs = 1; + fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs (inst->Opcode); + + fulldst = &fullinst->FullDstRegisters[0]; + fulldst->DstRegister.File = + map_register_file (inst->DstReg.File); + fulldst->DstRegister.Index = + map_register_file_index (processor, + fulldst->DstRegister.File, + inst->DstReg.Index, + 0); + fulldst->DstRegister.WriteMask = + convert_writemask (inst->DstReg.WriteMask); + + for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) { + GLuint j; + + fullsrc = &fullinst->FullSrcRegisters[i]; + fullsrc->SrcRegister.File = + map_register_file (inst->SrcReg[i].File); + fullsrc->SrcRegister.Index = + map_register_file_index (processor, + fullsrc->SrcRegister.File, + inst->SrcReg[i].Index, + inputs_read); + + for (j = 0; j < 4; j++) { + GLuint swz; + + swz = GET_SWZ(inst->SrcReg[i].Swizzle, j); + if (swz > SWIZZLE_W) { + tgsi_util_set_src_register_extswizzle ( + &fullsrc->SrcRegisterExtSwz, + swz, + j); + } else { + tgsi_util_set_src_register_swizzle ( + &fullsrc->SrcRegister, + swz, + j); + } + } + + if (inst->SrcReg[i].NegateBase == NEGATE_XYZW) { + fullsrc->SrcRegister.Negate = 1; + } else if (inst->SrcReg[i].NegateBase != NEGATE_NONE) { + if (inst->SrcReg[i].NegateBase & NEGATE_X) + fullsrc->SrcRegisterExtSwz.NegateX = 1; + if (inst->SrcReg[i].NegateBase & NEGATE_Y) + fullsrc->SrcRegisterExtSwz.NegateY = 1; + if (inst->SrcReg[i].NegateBase & NEGATE_Z) + fullsrc->SrcRegisterExtSwz.NegateZ = 1; + if (inst->SrcReg[i].NegateBase & NEGATE_W) + fullsrc->SrcRegisterExtSwz.NegateW = 1; + } + + if (inst->SrcReg[i].Abs) + fullsrc->SrcRegisterExtMod.Absolute = 1; + + if (inst->SrcReg[i].NegateAbs) + fullsrc->SrcRegisterExtMod.Negate = 1; + + if (inst->SrcReg[i].RelAddr) { + fullsrc->SrcRegister.Indirect = 1; + + fullsrc->SrcRegisterInd.File = TGSI_FILE_ADDRESS; + fullsrc->SrcRegisterInd.Index = 0; + } + } + + switch (inst->Opcode) { + case OPCODE_ARL: + fullinst->Instruction.Opcode = TGSI_OPCODE_ARL; + break; + case OPCODE_ABS: + fullinst->Instruction.Opcode = TGSI_OPCODE_ABS; + break; + case OPCODE_ADD: + fullinst->Instruction.Opcode = TGSI_OPCODE_ADD; + break; + case OPCODE_CMP: + fullinst->Instruction.Opcode = TGSI_OPCODE_CMP; + break; + case OPCODE_COS: + fullinst->Instruction.Opcode = TGSI_OPCODE_COS; + break; + case OPCODE_DP3: + fullinst->Instruction.Opcode = TGSI_OPCODE_DP3; + break; + case OPCODE_DP4: + fullinst->Instruction.Opcode = TGSI_OPCODE_DP4; + break; + case OPCODE_DPH: + fullinst->Instruction.Opcode = TGSI_OPCODE_DPH; + break; + case OPCODE_DST: + fullinst->Instruction.Opcode = TGSI_OPCODE_DST; + break; + case OPCODE_EX2: + fullinst->Instruction.Opcode = TGSI_OPCODE_EX2; + break; + case OPCODE_FLR: + fullinst->Instruction.Opcode = TGSI_OPCODE_FLR; + break; + case OPCODE_FRC: + fullinst->Instruction.Opcode = TGSI_OPCODE_FRC; + break; + case OPCODE_KIL: + fullinst->Instruction.Opcode = TGSI_OPCODE_KIL; + break; + case OPCODE_LG2: + fullinst->Instruction.Opcode = TGSI_OPCODE_LG2; + break; + case OPCODE_LIT: + fullinst->Instruction.Opcode = TGSI_OPCODE_LIT; + break; + case OPCODE_LRP: + fullinst->Instruction.Opcode = TGSI_OPCODE_LRP; + break; + case OPCODE_MAD: + fullinst->Instruction.Opcode = TGSI_OPCODE_MAD; + break; + case OPCODE_MAX: + fullinst->Instruction.Opcode = TGSI_OPCODE_MAX; + break; + case OPCODE_MIN: + fullinst->Instruction.Opcode = TGSI_OPCODE_MIN; + break; + case OPCODE_MOV: + fullinst->Instruction.Opcode = TGSI_OPCODE_MOV; + break; + case OPCODE_MUL: + fullinst->Instruction.Opcode = TGSI_OPCODE_MUL; + break; + case OPCODE_POW: + fullinst->Instruction.Opcode = TGSI_OPCODE_POW; + break; + case OPCODE_RCP: + fullinst->Instruction.Opcode = TGSI_OPCODE_RCP; + break; + case OPCODE_RSQ: + fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ; + tgsi_util_set_full_src_register_sign_mode (&fullinst->FullSrcRegisters[0], + TGSI_UTIL_SIGN_CLEAR); + break; + case OPCODE_SCS: + fullinst->Instruction.Opcode = TGSI_OPCODE_SCS; + fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XY; + break; + case OPCODE_SGE: + fullinst->Instruction.Opcode = TGSI_OPCODE_SGE; + break; + case OPCODE_SIN: + fullinst->Instruction.Opcode = TGSI_OPCODE_SIN; + break; + case OPCODE_SLT: + fullinst->Instruction.Opcode = TGSI_OPCODE_SLT; + break; + case OPCODE_SUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_SUB; + break; + case OPCODE_SWZ: + fullinst->Instruction.Opcode = TGSI_OPCODE_SWZ; + break; + case OPCODE_TEX: + fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target (inst->TexSrcTarget); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXB: + fullinst->Instruction.Opcode = TGSI_OPCODE_TXB; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target (inst->TexSrcTarget); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXP: + fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target (inst->TexSrcTarget); + fullinst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide = TGSI_EXTSWIZZLE_W; + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_XPD: + fullinst->Instruction.Opcode = TGSI_OPCODE_XPD; + fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; + break; + case OPCODE_END: + return GL_TRUE; + default: + assert (0); + } + + return GL_FALSE; +} + +GLboolean +tgsi_compile_fp_program( const struct gl_fragment_program *program, + struct tgsi_token *tokens, + GLuint max_token_count, + GLuint *token_count ) +{ + GLuint i, ti; + struct tgsi_header *header; + struct tgsi_full_declaration fulldecl; + struct tgsi_full_instruction fullinst; + struct tgsi_full_dst_register *fulldst; + struct tgsi_full_src_register *fullsrc; + GLuint inputs_read; + + *(struct tgsi_version *) &tokens[0] = tgsi_build_version (); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header (); + + ti = 2; + + /* + * Input 0 is always read, at least implicitly by the instruction generated + * above, so mark it as used. + */ + inputs_read = program->Base.InputsRead | 1; + + /* + * Declare input attributes. + */ + fulldecl = tgsi_default_full_declaration(); + + fulldecl.Declaration.File = TGSI_FILE_INPUT; + fulldecl.Declaration.Declare = TGSI_DECLARE_RANGE; + fulldecl.Declaration.Interpolate = 1; + + /* + * Do not interpolate fragment position. + */ + fulldecl.u.DeclarationRange.First = 0; + fulldecl.u.DeclarationRange.Last = 0; + + fulldecl.Interpolation.Interpolate = TGSI_INTERPOLATE_CONSTANT; + + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + max_token_count - ti ); + + /* + * Interpolate generic attributes. + */ + fulldecl.u.DeclarationRange.First = 1; + fulldecl.u.DeclarationRange.Last = 1; + for( i = 1; i < 32; i++ ) { + if( inputs_read & (1 << i) ) { + fulldecl.u.DeclarationRange.Last++; + } + } + + fulldecl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR; + + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + max_token_count - ti ); + + /* + * Copy input fragment xyz to output xyz. + * If the shader writes depth, do not copy the z component. + */ + + fullinst = tgsi_default_full_instruction (); + + fullinst.Instruction.Opcode = TGSI_OPCODE_MOV; + fullinst.Instruction.NumDstRegs = 1; + fullinst.Instruction.NumSrcRegs = 1; + + fulldst = &fullinst.FullDstRegisters[0]; + fulldst->DstRegister.File = TGSI_FILE_OUTPUT; + fulldst->DstRegister.Index = 0; + if (program->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR)) { + fulldst->DstRegister.WriteMask = TGSI_WRITEMASK_XY; + } else { + fulldst->DstRegister.WriteMask = TGSI_WRITEMASK_XYZ; + } + + fullsrc = &fullinst.FullSrcRegisters[0]; + fullsrc->SrcRegister.File = TGSI_FILE_INPUT; + fullsrc->SrcRegister.Index = 0; + + ti += tgsi_build_full_instruction (&fullinst, + &tokens[ti], + header, + max_token_count - ti); + + for( i = 0; i < program->Base.NumInstructions; i++ ) { + if (compile_instruction (&program->Base.Instructions[i], + &fullinst, + inputs_read, + TGSI_PROCESSOR_FRAGMENT)) { + assert (i == program->Base.NumInstructions - 1); + tgsi_dump (tokens, TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT); + break; + } + + ti += tgsi_build_full_instruction (&fullinst, + &tokens[ti], + header, + max_token_count - ti); + } + + return GL_TRUE; +} + +GLboolean +tgsi_compile_vp_program( const struct gl_vertex_program *program, + struct tgsi_token *tokens, + GLuint max_token_count, + GLuint *token_count ) +{ + GLuint ii, ti; + struct tgsi_header *header; + struct tgsi_processor *processor; + struct tgsi_full_instruction fullinst; + GLuint inputs_read = ~0; + + *(struct tgsi_version *) &tokens[0] = tgsi_build_version (); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header (); + + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor (TGSI_PROCESSOR_VERTEX, header); + + ti = 3; + + for (ii = 0; ii < program->Base.NumInstructions; ii++) { + if (compile_instruction (&program->Base.Instructions[ii], + &fullinst, + inputs_read, + TGSI_PROCESSOR_VERTEX)) { + assert (ii == program->Base.NumInstructions - 1); + tgsi_dump (tokens, TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT); + break; + } + + ti += tgsi_build_full_instruction (&fullinst, + &tokens[ti], + header, + max_token_count - ti); + } + + return GL_TRUE; +} + diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h new file mode 100644 index 0000000000..4c1141e579 --- /dev/null +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -0,0 +1,25 @@ +#if !defined MESA_TO_TGSI_H +#define MESA_TO_TGSI_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +GLboolean +tgsi_mesa_compile_fp_program( + const struct gl_fragment_program *program, + struct tgsi_token *tokens, + GLuint maxTokens ); + +GLboolean +tgsi_mesa_compile_vp_program( + const struct gl_vertex_program *program, + struct tgsi_token *tokens, + GLuint maxTokens ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined MESA_TO_TGSI_H + diff --git a/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h b/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h new file mode 100644 index 0000000000..0053748fde --- /dev/null +++ b/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h @@ -0,0 +1,8 @@ +#if !defined TGSI_MESA_H +#define TGSI_MESA_H + +#include "../core/tgsi_core.h" +#include "mesa_to_tgsi.h" + +#endif // !defined TGSI_MESA_H + diff --git a/src/mesa/pipe/tgsi/tgsi_platform.h b/src/mesa/pipe/tgsi/tgsi_platform.h new file mode 100644 index 0000000000..553f0b26a8 --- /dev/null +++ b/src/mesa/pipe/tgsi/tgsi_platform.h @@ -0,0 +1,18 @@ +#if !defined TGSI_PLATFORM_H +#define TGSI_PLATFORM_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +#include "imports.h" +#include "mtypes.h" +#include "prog_instruction.h" +#include "program.h" + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_PLATFORM_H + diff --git a/src/mesa/sources b/src/mesa/sources index 643e841b15..b1ff3798a6 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -179,6 +179,16 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_state_setup.c \ pipe/softpipe/sp_state_surface.c +TGSICORE_SOURCES = \ + pipe/tgsi/core/tgsi_build.c \ + pipe/tgsi/core/tgsi_dump.c \ + pipe/tgsi/core/tgsi_exec.c \ + pipe/tgsi/core/tgsi_parse.c \ + pipe/tgsi/core/tgsi_util.c + +TGSIMESA_SOURCES = \ + pipe/tgsi/mesa/mesa_to_tgsi.c + STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ state_tracker/st_atom_alphatest.c \ @@ -344,8 +354,10 @@ SOLO_SOURCES = \ $(MATH_SOURCES) \ $(VBO_SOURCES) \ $(VF_SOURCES) \ - $(SOFTPIPE_SOURCES) \ - $(STATETRACKER_SOURCES) \ + $(SOFTPIPE_SOURCES) \ + $(TGSICORE_SOURCES) \ + $(TGSIMESA_SOURCES) \ + $(STATETRACKER_SOURCES) \ $(TNL_SOURCES) \ $(SHADER_SOURCES) \ $(SWRAST_SOURCES) \ @@ -391,4 +403,5 @@ INCLUDE_DIRS = \ -I$(TOP)/src/mesa/shader/grammar \ -I$(TOP)/src/mesa/shader/slang \ -I$(TOP)/src/mesa/swrast \ - -I$(TOP)/src/mesa/swrast_setup + -I$(TOP)/src/mesa/swrast_setup \ + -I$(TOP)/src/mesa/pipe/tgsi diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 67a589ca8d..327b627722 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -41,6 +41,7 @@ #include "program.h" #include "programopt.h" #include "tnl/tnl.h" +#include "pipe/tgsi/mesa/tgsi_mesa.h" static void st_bind_program( GLcontext *ctx, @@ -89,7 +90,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, static void st_delete_program( GLcontext *ctx, struct gl_program *prog ) -{ +{ _mesa_delete_program( ctx, prog ); } @@ -115,7 +116,7 @@ static void st_program_string_notify( GLcontext *ctx, st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - p->id = st->program_id++; + p->id = st->program_id++; #if 0 p->param_state = p->Base.Base.Parameters->StateFlags; p->translated = 0; @@ -128,6 +129,14 @@ static void st_program_string_notify( GLcontext *ctx, _mesa_append_fog_code(ctx, &p->Base); p->Base.FogOption = GL_NONE; } + + /* XXX: Not hooked-up yet. */ + { + struct tgsi_token tokens[1024]; + + tgsi_mesa_compile_fp_program( prog, tokens, 1024 ); + tgsi_dump( tokens, TGSI_DUMP_VERBOSE ); + } } } else if (target == GL_VERTEX_PROGRAM_ARB) { @@ -148,7 +157,6 @@ void st_init_cb_program( struct st_context *st ) */ st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; st->ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; - assert(functions->ProgramStringNotify == _tnl_program_string); functions->BindProgram = st_bind_program; diff --git a/src/mesa/state_tracker/st_texobj.c b/src/mesa/state_tracker/st_texobj.c index 455e89b6bd..eb5bdb2d08 100644 --- a/src/mesa/state_tracker/st_texobj.c +++ b/src/mesa/state_tracker/st_texobj.c @@ -52,7 +52,8 @@ create_texture_object(struct gl_texture_object *texObj) if (!pto) return NULL; - assert(texObj->Complete); + /* XXX: Member not defined. Comment-out to get it compile. */ + /*assert(texObj->Complete);*/ switch (texObj->Target) { case GL_TEXTURE_1D: -- cgit v1.2.3 From 6d8fb1a54301e846c5ff5185ec0aa3b5d4cba823 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 6 Jul 2007 18:53:55 +0200 Subject: Indent. Fix function prototype. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 319 ++++++++++++++++++--------------- 1 file changed, 171 insertions(+), 148 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index e37778fe41..e6298f579f 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -5,9 +5,10 @@ * Map mesa register file to SBIR register file. */ static GLuint -map_register_file( enum register_file file ) +map_register_file( + enum register_file file ) { - switch (file) { + switch( file ) { case PROGRAM_UNDEFINED: return TGSI_FILE_NULL; case PROGRAM_TEMPORARY: @@ -25,7 +26,7 @@ map_register_file( enum register_file file ) case PROGRAM_ADDRESS: return TGSI_FILE_ADDRESS; default: - assert (0); + assert( 0 ); return TGSI_FILE_NULL; } } @@ -35,37 +36,40 @@ map_register_file( enum register_file file ) * Take special care when processing input and output indices. */ static GLuint -map_register_file_index( GLuint processor, - GLuint file, - GLuint index, - GLuint usage_bitmask ) +map_register_file_index( + GLuint processor, + GLuint file, + GLuint index, + GLuint usage_bitmask ) { GLuint mapped_index; GLuint i; - switch (file) - { + switch( file ) { case TGSI_FILE_INPUT: - assert (index < 32); - assert (usage_bitmask & (1 << index)); + assert( index < 32 ); + assert( usage_bitmask & (1 << index) ); mapped_index = 0; - for (i = 0; i < index; i++) { - if (usage_bitmask & (1 << i)) - mapped_index++; + for( i = 0; i < index; i++ ) { + if( usage_bitmask & (1 << i) ) { + mapped_index++; + } } break; case TGSI_FILE_OUTPUT: - assert (usage_bitmask == 0); - if (processor == TGSI_PROCESSOR_FRAGMENT) { - if (index == FRAG_RESULT_DEPR) { - mapped_index = 0; - } else { - assert (index == FRAG_RESULT_COLR); - mapped_index = index + 1; - } - } else { - mapped_index = index; + assert( usage_bitmask == 0 ); + if( processor == TGSI_PROCESSOR_FRAGMENT ) { + if( index == FRAG_RESULT_DEPR ) { + mapped_index = 0; + } + else { + assert( index == FRAG_RESULT_COLR ); + mapped_index = index + 1; + } + } + else { + mapped_index = index; } break; @@ -80,9 +84,10 @@ map_register_file_index( GLuint processor, * Map mesa texture target to SBIR texture target. */ static GLuint -map_texture_target( GLuint textarget ) +map_texture_target( + GLuint textarget ) { - switch (textarget) { + switch( textarget ) { case TEXTURE_1D_INDEX: return TGSI_TEXTURE_1D; case TEXTURE_2D_INDEX: @@ -94,15 +99,17 @@ map_texture_target( GLuint textarget ) case TEXTURE_RECT_INDEX: return TGSI_TEXTURE_RECT; default: - assert (0); + assert( 0 ); } + return TGSI_TEXTURE_1D; } static GLuint -convert_sat( GLuint sat ) +convert_sat( + GLuint sat ) { - switch (sat) { + switch( sat ) { case SATURATE_OFF: return TGSI_SAT_NONE; case SATURATE_ZERO_ONE: @@ -110,107 +117,114 @@ convert_sat( GLuint sat ) case SATURATE_PLUS_MINUS_ONE: return TGSI_SAT_MINUS_PLUS_ONE; default: - assert (0); + assert( 0 ); return TGSI_SAT_NONE; } } static GLuint -convert_writemask( GLuint writemask ) +convert_writemask( + GLuint writemask ) { - assert (WRITEMASK_X == TGSI_WRITEMASK_X); - assert (WRITEMASK_Y == TGSI_WRITEMASK_Y); - assert (WRITEMASK_Z == TGSI_WRITEMASK_Z); - assert (WRITEMASK_W == TGSI_WRITEMASK_W); - assert ((writemask & ~TGSI_WRITEMASK_XYZW) == 0); + assert( WRITEMASK_X == TGSI_WRITEMASK_X ); + assert( WRITEMASK_Y == TGSI_WRITEMASK_Y ); + assert( WRITEMASK_Z == TGSI_WRITEMASK_Z ); + assert( WRITEMASK_W == TGSI_WRITEMASK_W ); + assert( (writemask & ~TGSI_WRITEMASK_XYZW) == 0 ); return writemask; } static GLboolean -compile_instruction( struct prog_instruction *inst, - struct tgsi_full_instruction *fullinst, - GLuint inputs_read, - GLuint processor ) +compile_instruction( + struct prog_instruction *inst, + struct tgsi_full_instruction *fullinst, + GLuint inputs_read, + GLuint processor ) { GLuint i; struct tgsi_full_dst_register *fulldst; struct tgsi_full_src_register *fullsrc; - *fullinst = tgsi_default_full_instruction (); + *fullinst = tgsi_default_full_instruction(); - fullinst->Instruction.Saturate = convert_sat (inst->SaturateMode); + fullinst->Instruction.Saturate = convert_sat( inst->SaturateMode ); fullinst->Instruction.NumDstRegs = 1; - fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs (inst->Opcode); + fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode ); fulldst = &fullinst->FullDstRegisters[0]; - fulldst->DstRegister.File = - map_register_file (inst->DstReg.File); - fulldst->DstRegister.Index = - map_register_file_index (processor, - fulldst->DstRegister.File, - inst->DstReg.Index, - 0); - fulldst->DstRegister.WriteMask = - convert_writemask (inst->DstReg.WriteMask); - - for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) { + fulldst->DstRegister.File = map_register_file( inst->DstReg.File ); + fulldst->DstRegister.Index = map_register_file_index( + processor, + fulldst->DstRegister.File, + inst->DstReg.Index, + 0 ); + fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); + + for( i = 0; i < fullinst->Instruction.NumSrcRegs; i++ ) { GLuint j; fullsrc = &fullinst->FullSrcRegisters[i]; - fullsrc->SrcRegister.File = - map_register_file (inst->SrcReg[i].File); - fullsrc->SrcRegister.Index = - map_register_file_index (processor, - fullsrc->SrcRegister.File, - inst->SrcReg[i].Index, - inputs_read); - - for (j = 0; j < 4; j++) { - GLuint swz; - - swz = GET_SWZ(inst->SrcReg[i].Swizzle, j); - if (swz > SWIZZLE_W) { - tgsi_util_set_src_register_extswizzle ( - &fullsrc->SrcRegisterExtSwz, - swz, - j); - } else { - tgsi_util_set_src_register_swizzle ( - &fullsrc->SrcRegister, - swz, - j); - } + fullsrc->SrcRegister.File = map_register_file( inst->SrcReg[i].File ); + fullsrc->SrcRegister.Index = map_register_file_index( + processor, + fullsrc->SrcRegister.File, + inst->SrcReg[i].Index, + inputs_read ); + + for( j = 0; j < 4; j++ ) { + GLuint swz; + + swz = GET_SWZ( inst->SrcReg[i].Swizzle, j ); + if( swz > SWIZZLE_W ) { + tgsi_util_set_src_register_extswizzle( + &fullsrc->SrcRegisterExtSwz, + swz, + j ); + } + else { + tgsi_util_set_src_register_swizzle( + &fullsrc->SrcRegister, + swz, + j ); + } } - if (inst->SrcReg[i].NegateBase == NEGATE_XYZW) { - fullsrc->SrcRegister.Negate = 1; - } else if (inst->SrcReg[i].NegateBase != NEGATE_NONE) { - if (inst->SrcReg[i].NegateBase & NEGATE_X) - fullsrc->SrcRegisterExtSwz.NegateX = 1; - if (inst->SrcReg[i].NegateBase & NEGATE_Y) - fullsrc->SrcRegisterExtSwz.NegateY = 1; - if (inst->SrcReg[i].NegateBase & NEGATE_Z) - fullsrc->SrcRegisterExtSwz.NegateZ = 1; - if (inst->SrcReg[i].NegateBase & NEGATE_W) - fullsrc->SrcRegisterExtSwz.NegateW = 1; + if( inst->SrcReg[i].NegateBase == NEGATE_XYZW ) { + fullsrc->SrcRegister.Negate = 1; + } + else if( inst->SrcReg[i].NegateBase != NEGATE_NONE ) { + if( inst->SrcReg[i].NegateBase & NEGATE_X ) { + fullsrc->SrcRegisterExtSwz.NegateX = 1; + } + if( inst->SrcReg[i].NegateBase & NEGATE_Y ) { + fullsrc->SrcRegisterExtSwz.NegateY = 1; + } + if( inst->SrcReg[i].NegateBase & NEGATE_Z ) { + fullsrc->SrcRegisterExtSwz.NegateZ = 1; + } + if( inst->SrcReg[i].NegateBase & NEGATE_W ) { + fullsrc->SrcRegisterExtSwz.NegateW = 1; + } } - if (inst->SrcReg[i].Abs) - fullsrc->SrcRegisterExtMod.Absolute = 1; + if( inst->SrcReg[i].Abs ) { + fullsrc->SrcRegisterExtMod.Absolute = 1; + } - if (inst->SrcReg[i].NegateAbs) - fullsrc->SrcRegisterExtMod.Negate = 1; + if( inst->SrcReg[i].NegateAbs ) { + fullsrc->SrcRegisterExtMod.Negate = 1; + } - if (inst->SrcReg[i].RelAddr) { - fullsrc->SrcRegister.Indirect = 1; + if( inst->SrcReg[i].RelAddr ) { + fullsrc->SrcRegister.Indirect = 1; - fullsrc->SrcRegisterInd.File = TGSI_FILE_ADDRESS; - fullsrc->SrcRegisterInd.Index = 0; + fullsrc->SrcRegisterInd.File = TGSI_FILE_ADDRESS; + fullsrc->SrcRegisterInd.Index = 0; } } - switch (inst->Opcode) { + switch( inst->Opcode ) { case OPCODE_ARL: fullinst->Instruction.Opcode = TGSI_OPCODE_ARL; break; @@ -282,8 +296,9 @@ compile_instruction( struct prog_instruction *inst, break; case OPCODE_RSQ: fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ; - tgsi_util_set_full_src_register_sign_mode (&fullinst->FullSrcRegisters[0], - TGSI_UTIL_SIGN_CLEAR); + tgsi_util_set_full_src_register_sign_mode( + &fullinst->FullSrcRegisters[0], + TGSI_UTIL_SIGN_CLEAR ); break; case OPCODE_SCS: fullinst->Instruction.Opcode = TGSI_OPCODE_SCS; @@ -307,21 +322,21 @@ compile_instruction( struct prog_instruction *inst, case OPCODE_TEX: fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target (inst->TexSrcTarget); + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; break; case OPCODE_TXB: fullinst->Instruction.Opcode = TGSI_OPCODE_TXB; fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target (inst->TexSrcTarget); + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; break; case OPCODE_TXP: fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target (inst->TexSrcTarget); + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); fullinst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide = TGSI_EXTSWIZZLE_W; fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; @@ -333,17 +348,17 @@ compile_instruction( struct prog_instruction *inst, case OPCODE_END: return GL_TRUE; default: - assert (0); + assert( 0 ); } return GL_FALSE; } GLboolean -tgsi_compile_fp_program( const struct gl_fragment_program *program, - struct tgsi_token *tokens, - GLuint max_token_count, - GLuint *token_count ) +tgsi_compile_fp_program( + const struct gl_fragment_program *program, + struct tgsi_token *tokens, + GLuint maxTokens ) { GLuint i, ti; struct tgsi_header *header; @@ -353,10 +368,10 @@ tgsi_compile_fp_program( const struct gl_fragment_program *program, struct tgsi_full_src_register *fullsrc; GLuint inputs_read; - *(struct tgsi_version *) &tokens[0] = tgsi_build_version (); + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header (); + *header = tgsi_build_header(); ti = 2; @@ -387,7 +402,7 @@ tgsi_compile_fp_program( const struct gl_fragment_program *program, &fulldecl, &tokens[ti], header, - max_token_count - ti ); + maxTokens - ti ); /* * Interpolate generic attributes. @@ -406,14 +421,14 @@ tgsi_compile_fp_program( const struct gl_fragment_program *program, &fulldecl, &tokens[ti], header, - max_token_count - ti ); + maxTokens - ti ); /* * Copy input fragment xyz to output xyz. * If the shader writes depth, do not copy the z component. */ - fullinst = tgsi_default_full_instruction (); + fullinst = tgsi_default_full_instruction(); fullinst.Instruction.Opcode = TGSI_OPCODE_MOV; fullinst.Instruction.NumDstRegs = 1; @@ -422,9 +437,10 @@ tgsi_compile_fp_program( const struct gl_fragment_program *program, fulldst = &fullinst.FullDstRegisters[0]; fulldst->DstRegister.File = TGSI_FILE_OUTPUT; fulldst->DstRegister.Index = 0; - if (program->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR)) { + if( program->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR) ) { fulldst->DstRegister.WriteMask = TGSI_WRITEMASK_XY; - } else { + } + else { fulldst->DstRegister.WriteMask = TGSI_WRITEMASK_XYZ; } @@ -432,35 +448,40 @@ tgsi_compile_fp_program( const struct gl_fragment_program *program, fullsrc->SrcRegister.File = TGSI_FILE_INPUT; fullsrc->SrcRegister.Index = 0; - ti += tgsi_build_full_instruction (&fullinst, - &tokens[ti], - header, - max_token_count - ti); + ti += tgsi_build_full_instruction( + &fullinst, + &tokens[ti], + header, + maxTokens - ti ); for( i = 0; i < program->Base.NumInstructions; i++ ) { - if (compile_instruction (&program->Base.Instructions[i], - &fullinst, - inputs_read, - TGSI_PROCESSOR_FRAGMENT)) { - assert (i == program->Base.NumInstructions - 1); - tgsi_dump (tokens, TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT); - break; + if( compile_instruction( + &program->Base.Instructions[i], + &fullinst, + inputs_read, + TGSI_PROCESSOR_FRAGMENT ) ) { + assert( i == program->Base.NumInstructions - 1 ); + tgsi_dump( + tokens, + TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT ); + break; } - ti += tgsi_build_full_instruction (&fullinst, - &tokens[ti], - header, - max_token_count - ti); + ti += tgsi_build_full_instruction( + &fullinst, + &tokens[ti], + header, + maxTokens - ti ); } return GL_TRUE; } GLboolean -tgsi_compile_vp_program( const struct gl_vertex_program *program, - struct tgsi_token *tokens, - GLuint max_token_count, - GLuint *token_count ) +tgsi_compile_vp_program( + const struct gl_vertex_program *program, + struct tgsi_token *tokens, + GLuint maxTokens ) { GLuint ii, ti; struct tgsi_header *header; @@ -468,30 +489,32 @@ tgsi_compile_vp_program( const struct gl_vertex_program *program, struct tgsi_full_instruction fullinst; GLuint inputs_read = ~0; - *(struct tgsi_version *) &tokens[0] = tgsi_build_version (); + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header (); + *header = tgsi_build_header(); processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor (TGSI_PROCESSOR_VERTEX, header); + *processor = tgsi_build_processor( TGSI_PROCESSOR_VERTEX, header ); ti = 3; - for (ii = 0; ii < program->Base.NumInstructions; ii++) { - if (compile_instruction (&program->Base.Instructions[ii], - &fullinst, - inputs_read, - TGSI_PROCESSOR_VERTEX)) { - assert (ii == program->Base.NumInstructions - 1); - tgsi_dump (tokens, TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT); - break; + for( ii = 0; ii < program->Base.NumInstructions; ii++ ) { + if( compile_instruction( + &program->Base.Instructions[ii], + &fullinst, + inputs_read, + TGSI_PROCESSOR_VERTEX ) ) { + assert( ii == program->Base.NumInstructions - 1 ); + tgsi_dump( tokens, TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT ); + break; } - ti += tgsi_build_full_instruction (&fullinst, - &tokens[ti], - header, - max_token_count - ti); + ti += tgsi_build_full_instruction( + &fullinst, + &tokens[ti], + header, + maxTokens - ti ); } return GL_TRUE; -- cgit v1.2.3 From 9b32c33e9d87e4571e85efdbb14c6a98a090cae6 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 6 Jul 2007 18:57:47 +0200 Subject: Fix function names. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index e6298f579f..291e62d262 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -355,7 +355,7 @@ compile_instruction( } GLboolean -tgsi_compile_fp_program( +tgsi_mesa_compile_fp_program( const struct gl_fragment_program *program, struct tgsi_token *tokens, GLuint maxTokens ) @@ -478,7 +478,7 @@ tgsi_compile_fp_program( } GLboolean -tgsi_compile_vp_program( +tgsi_mesa_compile_vp_program( const struct gl_vertex_program *program, struct tgsi_token *tokens, GLuint maxTokens ) -- cgit v1.2.3 From 10f4b37bd70c12c8caaa85ff5b94aec532788abd Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 9 Jul 2007 16:04:08 -0600 Subject: Rename/move some files to modularize the primitive/draw code. --- src/mesa/pipe/draw/draw_clip.c | 429 ++++++++++++++++++ src/mesa/pipe/draw/draw_context.h | 56 +++ src/mesa/pipe/draw/draw_cull.c | 120 +++++ src/mesa/pipe/draw/draw_flatshade.c | 151 +++++++ src/mesa/pipe/draw/draw_offset.c | 157 +++++++ src/mesa/pipe/draw/draw_private.h | 90 ++++ src/mesa/pipe/draw/draw_twoside.c | 152 +++++++ src/mesa/pipe/draw/draw_unfilled.c | 165 +++++++ src/mesa/pipe/draw/draw_vb.c | 697 +++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_draw.c | 697 ----------------------------- src/mesa/pipe/softpipe/sp_draw.h | 56 --- src/mesa/pipe/softpipe/sp_prim.h | 90 ---- src/mesa/pipe/softpipe/sp_prim_clip.c | 429 ------------------ src/mesa/pipe/softpipe/sp_prim_cull.c | 120 ----- src/mesa/pipe/softpipe/sp_prim_flatshade.c | 151 ------- src/mesa/pipe/softpipe/sp_prim_offset.c | 157 ------- src/mesa/pipe/softpipe/sp_prim_twoside.c | 152 ------- src/mesa/pipe/softpipe/sp_prim_unfilled.c | 165 ------- 18 files changed, 2017 insertions(+), 2017 deletions(-) create mode 100644 src/mesa/pipe/draw/draw_clip.c create mode 100644 src/mesa/pipe/draw/draw_context.h create mode 100644 src/mesa/pipe/draw/draw_cull.c create mode 100644 src/mesa/pipe/draw/draw_flatshade.c create mode 100644 src/mesa/pipe/draw/draw_offset.c create mode 100644 src/mesa/pipe/draw/draw_private.h create mode 100644 src/mesa/pipe/draw/draw_twoside.c create mode 100644 src/mesa/pipe/draw/draw_unfilled.c create mode 100644 src/mesa/pipe/draw/draw_vb.c delete mode 100644 src/mesa/pipe/softpipe/sp_draw.c delete mode 100644 src/mesa/pipe/softpipe/sp_draw.h delete mode 100644 src/mesa/pipe/softpipe/sp_prim.h delete mode 100644 src/mesa/pipe/softpipe/sp_prim_clip.c delete mode 100644 src/mesa/pipe/softpipe/sp_prim_cull.c delete mode 100644 src/mesa/pipe/softpipe/sp_prim_flatshade.c delete mode 100644 src/mesa/pipe/softpipe/sp_prim_offset.c delete mode 100644 src/mesa/pipe/softpipe/sp_prim_twoside.c delete mode 100644 src/mesa/pipe/softpipe/sp_prim_unfilled.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c new file mode 100644 index 0000000000..304c43c3f4 --- /dev/null +++ b/src/mesa/pipe/draw/draw_clip.c @@ -0,0 +1,429 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "sp_context.h" +#include "sp_prim.h" + +struct clipper { + struct prim_stage stage; + + GLuint active_user_planes; +}; + +/* This is a bit confusing: + */ +static INLINE struct clipper *clipper_stage( struct prim_stage *stage ) +{ + return (struct clipper *)stage; +} + + +#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) + + +/* All attributes are float[4], so this is easy: + */ +static void interp_attr( GLfloat *fdst, + GLfloat t, + const GLfloat *fin, + const GLfloat *fout ) +{ + fdst[0] = LINTERP( t, fout[0], fin[0] ); + fdst[1] = LINTERP( t, fout[1], fin[1] ); + fdst[2] = LINTERP( t, fout[2], fin[2] ); + fdst[3] = LINTERP( t, fout[3], fin[3] ); +} + + + + +/* Interpolate between two vertices to produce a third. + */ +static void interp( struct clipper *clip, + struct vertex_header *dst, + GLfloat t, + const struct vertex_header *out, + const struct vertex_header *in ) +{ + const GLuint nr_attrs = clip->stage.softpipe->nr_attrs; + GLuint j; + + /* Vertex header. + */ + { + dst->clipmask = 0; + dst->edgeflag = 0; + dst->pad = 0; + } + + /* Clip coordinates: interpolate normally + */ + { + interp_attr(dst->clip, t, in->clip, out->clip); + } + + /* Do the projective divide and insert window coordinates: + */ + { + const GLfloat *pos = dst->clip; + const GLfloat *scale = clip->stage.softpipe->viewport.scale; + const GLfloat *trans = clip->stage.softpipe->viewport.translate; + GLfloat oow; + + oow = 1.0 / pos[3]; + + dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; + dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; + dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; + dst->data[0][3] = oow; + } + + + /* Other attributes + */ + for (j = 1; j < nr_attrs-1; j++) { + interp_attr(dst->data[j], t, in->data[j], out->data[j]); + } +} + + +#define CLIP_USER_BIT 0x40 +#define CLIP_CULL_BIT 0x80 + + +static INLINE GLfloat dot4( const GLfloat *a, + const GLfloat *b ) +{ + GLfloat result = (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); + + return result; +} + + +#if 0 +static INLINE void do_tri( struct prim_stage *next, + struct prim_header *header ) +{ + GLuint i; + for (i = 0; i < 3; i++) { + GLfloat *ndc = header->v[i]->data[0]; + _mesa_printf("ndc %f %f %f\n", ndc[0], ndc[1], ndc[2]); + assert(ndc[0] >= -1 && ndc[0] <= 641); + assert(ndc[1] >= 30 && ndc[1] <= 481); + } + _mesa_printf("\n"); + next->tri(next, header); +} +#endif + + +static void emit_poly( struct prim_stage *stage, + struct vertex_header **inlist, + GLuint n ) +{ + struct prim_header header; + GLuint i; + + for (i = 2; i < n; i++) { + header.v[0] = inlist[0]; + header.v[1] = inlist[i-1]; + header.v[2] = inlist[i]; + + { + GLuint tmp0 = header.v[0]->edgeflag; + GLuint tmp2 = header.v[2]->edgeflag; + + if (i != 2) header.v[0]->edgeflag = 0; + if (i != n-1) header.v[2]->edgeflag = 0; + + stage->next->tri( stage->next, &header ); + + header.v[0]->edgeflag = tmp0; + header.v[2]->edgeflag = tmp2; + } + } +} + + +#if 0 +static void emit_poly( struct prim_stage *stage ) +{ + GLuint i; + + for (i = 2; i < n; i++) { + header->v[0] = inlist[0]; + header->v[1] = inlist[i-1]; + header->v[2] = inlist[i]; + + stage->next->tri( stage->next, header ); + } +} +#endif + + +/* Clip a triangle against the viewport and user clip planes. + */ +static void +do_clip_tri( struct prim_stage *stage, + struct prim_header *header, + GLuint clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *a[MAX_CLIPPED_VERTICES]; + struct vertex_header *b[MAX_CLIPPED_VERTICES]; + struct vertex_header **inlist = a; + struct vertex_header **outlist = b; + GLuint tmpnr = 0; + GLuint n = 3; + GLuint i; + + inlist[0] = header->v[0]; + inlist[1] = header->v[1]; + inlist[2] = header->v[2]; + + /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove + * this once we correctly use 16bit masks for userclip planes. + */ + clipmask &= ~CLIP_CULL_BIT; + if (clipmask & CLIP_USER_BIT) { + clipmask &= ~CLIP_USER_BIT; + clipmask |= clipper->active_user_planes; + } + + while (clipmask && n >= 3) { + GLuint plane_idx = ffs(clipmask)-1; + const GLfloat *plane = clipper->stage.softpipe->plane[plane_idx]; + struct vertex_header *vert_prev = inlist[0]; + GLfloat dp_prev = dot4( vert_prev->clip, plane ); + GLuint outcount = 0; + + clipmask &= ~(1<clip, plane ); + + if (!IS_NEGATIVE(dp_prev)) { + outlist[outcount++] = vert_prev; + } + + if (DIFFERENT_SIGNS(dp, dp_prev)) { + struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; + outlist[outcount++] = new_vert; + + if (IS_NEGATIVE(dp)) { + /* Going out of bounds. Avoid division by zero as we + * know dp != dp_prev from DIFFERENT_SIGNS, above. + */ + GLfloat t = dp / (dp - dp_prev); + interp( clipper, new_vert, t, vert, vert_prev ); + + /* Force edgeflag true in this case: + */ + new_vert->edgeflag = 1; + } else { + /* Coming back in. + */ + GLfloat t = dp_prev / (dp_prev - dp); + interp( clipper, new_vert, t, vert_prev, vert ); + + /* Copy starting vert's edgeflag: + */ + new_vert->edgeflag = vert_prev->edgeflag; + } + } + + vert_prev = vert; + dp_prev = dp; + } + + { + struct vertex_header **tmp = inlist; + inlist = outlist; + outlist = tmp; + n = outcount; + } + } + + /* Emit the polygon as triangles to the setup stage: + */ + if (n >= 3) + emit_poly( stage, inlist, n ); +} + + +/* Clip a line against the viewport and user clip planes. + */ +static void +do_clip_line( struct prim_stage *stage, + struct prim_header *header, + GLuint clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const GLfloat *pos0 = v0->clip; + const GLfloat *pos1 = v1->clip; + GLfloat t0 = 0; + GLfloat t1 = 0; + + /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove + * this once we correctly use 16bit masks for userclip planes. + */ + clipmask &= ~CLIP_CULL_BIT; + if (clipmask & CLIP_USER_BIT) { + clipmask &= ~CLIP_USER_BIT; + clipmask |= clipper->active_user_planes; + } + + while (clipmask) { + GLuint plane_idx = ffs(clipmask)-1; + const GLfloat *plane = clipper->stage.softpipe->plane[plane_idx]; + + clipmask &= ~(1< t1) t1 = t; + } + + if (dp0 < 0) { + GLfloat t = dp0 / (dp0 - dp1); + if (t > t0) t0 = t; + } + + if (t0 + t1 >= 1.0) + return; /* discard */ + } + + if (v0->clipmask) { + interp( clipper, stage->tmp[0], t0, v0, v1 ); + header->v[0] = stage->tmp[0]; + } + + if (v1->clipmask) { + interp( clipper, stage->tmp[1], t1, v1, v0 ); + header->v[1] = stage->tmp[1]; + } + + stage->next->line( stage->next, header ); +} + + + +static void clip_begin( struct prim_stage *stage ) +{ + struct clipper *clipper = clipper_stage(stage); + GLuint nr = stage->softpipe->nr_planes; + + /* Hacky bitmask to use when we hit CLIP_USER_BIT: + */ + clipper->active_user_planes = ((1<next->begin( stage->next ); +} + +static void +clip_point( struct prim_stage *stage, + struct prim_header *header ) +{ + if (header->v[0]->clipmask == 0) + stage->next->point( stage->next, header ); +} + + +static void +clip_line( struct prim_stage *stage, + struct prim_header *header ) +{ + GLuint clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask); + + if (clipmask == 0) { + stage->next->line( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask) == 0) { + do_clip_line(stage, header, clipmask); + } +} + + +static void +clip_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + GLuint clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask | + header->v[2]->clipmask); + + if (clipmask == 0) { + stage->next->tri( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask & + header->v[2]->clipmask) == 0) { + do_clip_tri(stage, header, clipmask); + } +} + +static void clip_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + + +struct prim_stage *prim_clip( struct softpipe_context *softpipe ) +{ + struct clipper *clipper = CALLOC_STRUCT(clipper); + + prim_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); + + clipper->stage.softpipe = softpipe; + clipper->stage.begin = clip_begin; + clipper->stage.point = clip_point; + clipper->stage.line = clip_line; + clipper->stage.tri = clip_tri; + clipper->stage.end = clip_end; + + return &clipper->stage; +} diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h new file mode 100644 index 0000000000..a138f812fa --- /dev/null +++ b/src/mesa/pipe/draw/draw_context.h @@ -0,0 +1,56 @@ + +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_DRAW_H +#define G_DRAW_H + +#include "glheader.h" +#include "pipe/p_state.h" + + +struct draw_context; + +struct draw_context *draw_create( struct softpipe_context *softpipe ); + +void draw_destroy( struct draw_context *draw ); + +void draw_set_viewport( struct draw_context *draw, + const GLfloat *scale, + const GLfloat *translate ); + +void draw_set_vertex_attributes( struct draw_context *draw, + const GLuint *attrs, + GLuint nr_attrs ); + +void draw_vb(struct draw_context *draw, + struct vertex_buffer *VB ); + +#endif diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c new file mode 100644 index 0000000000..63099fbee0 --- /dev/null +++ b/src/mesa/pipe/draw/draw_cull.c @@ -0,0 +1,120 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "pipe/p_defines.h" +#include "sp_context.h" +#include "sp_prim.h" + + + +struct cull_stage { + struct prim_stage stage; + GLuint mode; +}; + + + +static INLINE struct cull_stage *cull_stage( struct prim_stage *stage ) +{ + return (struct cull_stage *)stage; +} + + +static void cull_begin( struct prim_stage *stage ) +{ + struct cull_stage *cull = cull_stage(stage); + + cull->mode = stage->softpipe->setup.cull_mode; + + stage->next->begin( stage->next ); +} + + +static void cull_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + /* Window coords: */ + GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); + GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); + GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); + + GLfloat ex = v0[0] - v2[0]; + GLfloat ey = v0[1] - v2[1]; + GLfloat fx = v1[0] - v2[0]; + GLfloat fy = v1[1] - v2[1]; + + header->det = ex * fy - ey * fx; + + _mesa_printf("%s %f\n", __FUNCTION__, header->det ); + + if (header->det != 0) { + GLuint mode = (header->det < 0) ? PIPE_WINDING_CW : PIPE_WINDING_CCW; + + if ((mode & cull_stage(stage)->mode) == 0) + stage->next->tri( stage->next, header ); + } +} + + +static void cull_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void cull_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void cull_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_cull( struct softpipe_context *softpipe ) +{ + struct cull_stage *cull = CALLOC_STRUCT(cull_stage); + + prim_alloc_tmps( &cull->stage, 0 ); + + cull->stage.softpipe = softpipe; + cull->stage.next = NULL; + cull->stage.begin = cull_begin; + cull->stage.point = cull_point; + cull->stage.line = cull_line; + cull->stage.tri = cull_tri; + cull->stage.end = cull_end; + + return &cull->stage; +} diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c new file mode 100644 index 0000000000..3a7d9de466 --- /dev/null +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -0,0 +1,151 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "vf/vf.h" + +#include "sp_context.h" +#include "sp_prim.h" + + +struct flatshade_stage { + struct prim_stage stage; + + const GLuint *lookup; +}; + + + +static INLINE struct flatshade_stage *flatshade_stage( struct prim_stage *stage ) +{ + return (struct flatshade_stage *)stage; +} + + +static void flatshade_begin( struct prim_stage *stage ) +{ + stage->next->begin( stage->next ); +} + + + +static INLINE void copy_attr( GLuint attr, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + if (attr) { + memcpy( dst->data[attr], + src->data[attr], + sizeof(src->data[0]) ); + } +} + +static void copy_colors( struct prim_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + struct flatshade_stage *flatshade = flatshade_stage(stage); + const GLuint *lookup = flatshade->lookup; + + copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); + copy_attr( lookup[VF_ATTRIB_COLOR1], dst, src ); + copy_attr( lookup[VF_ATTRIB_BFC0], dst, src ); + copy_attr( lookup[VF_ATTRIB_BFC1], dst, src ); +} + + + +/* Flatshade tri. Required for clipping and when unfilled tris are + * active, otherwise handled by hardware. + */ +static void flatshade_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = header->v[2]; + + copy_colors(stage, tmp.v[0], tmp.v[2]); + copy_colors(stage, tmp.v[1], tmp.v[2]); + + stage->next->tri( stage->next, &tmp ); +} + + +/* Flatshade line. Required for clipping. + */ +static void flatshade_line( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = header->v[1]; + + copy_colors(stage, tmp.v[0], tmp.v[1]); + + stage->next->line( stage->next, &tmp ); +} + + +static void flatshade_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void flatshade_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_flatshade( struct softpipe_context *softpipe ) +{ + struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); + + prim_alloc_tmps( &flatshade->stage, 2 ); + + flatshade->stage.softpipe = softpipe; + flatshade->stage.next = NULL; + flatshade->stage.begin = flatshade_begin; + flatshade->stage.point = flatshade_point; + flatshade->stage.line = flatshade_line; + flatshade->stage.tri = flatshade_tri; + flatshade->stage.end = flatshade_end; + + flatshade->lookup = softpipe->vf_attr_to_slot; + + return &flatshade->stage; +} + + diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c new file mode 100644 index 0000000000..5fd6ac911a --- /dev/null +++ b/src/mesa/pipe/draw/draw_offset.c @@ -0,0 +1,157 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" +#include "macros.h" + +#include "sp_context.h" +#include "sp_prim.h" + + + +struct offset_stage { + struct prim_stage stage; + + GLuint hw_data_offset; + + GLfloat scale; + GLfloat units; +}; + + + +static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) +{ + return (struct offset_stage *)stage; +} + + +static void offset_begin( struct prim_stage *stage ) +{ + struct offset_stage *offset = offset_stage(stage); + + offset->units = stage->softpipe->setup.offset_units; + offset->scale = stage->softpipe->setup.offset_scale; + + stage->next->begin( stage->next ); +} + + +/* Offset tri. Some hardware can handle this, but not usually when + * doing unfilled rendering. + */ +static void do_offset_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + GLfloat inv_det = 1.0 / header->det; + + /* Window coords: + */ + GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); + GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); + GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); + + GLfloat ex = v0[0] - v2[2]; + GLfloat fx = v1[0] - v2[2]; + GLfloat ey = v0[1] - v2[2]; + GLfloat fy = v1[1] - v2[2]; + GLfloat ez = v0[2] - v2[2]; + GLfloat fz = v1[2] - v2[2]; + + GLfloat a = ey*fz - ez*fy; + GLfloat b = ez*fx - ex*fz; + + GLfloat ac = a * inv_det; + GLfloat bc = b * inv_det; + GLfloat zoffset; + + if ( ac < 0.0f ) ac = -ac; + if ( bc < 0.0f ) bc = -bc; + + zoffset = offset->units + MAX2( ac, bc ) * offset->scale; + + v0[2] += zoffset; + v1[2] += zoffset; + v2[2] += zoffset; + + stage->next->tri( stage->next, header ); +} + + +static void offset_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = dup_vert(stage, header->v[2], 2); + + do_offset_tri( stage->next, &tmp ); +} + + + +static void offset_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void offset_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void offset_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_offset( struct softpipe_context *softpipe ) +{ + struct offset_stage *offset = CALLOC_STRUCT(offset_stage); + + prim_alloc_tmps( &offset->stage, 3 ); + + offset->stage.softpipe = softpipe; + offset->stage.next = NULL; + offset->stage.begin = offset_begin; + offset->stage.point = offset_point; + offset->stage.line = offset_line; + offset->stage.tri = offset_tri; + offset->stage.end = offset_end; + + return &offset->stage; +} diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h new file mode 100644 index 0000000000..b6cbaae085 --- /dev/null +++ b/src/mesa/pipe/draw/draw_private.h @@ -0,0 +1,90 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef G_PRIM_H +#define G_PRIM_H + +#include "glheader.h" +#include "sp_headers.h" + +struct softpipe_context; + +struct prim_stage *prim_setup( struct softpipe_context *context ); +struct prim_stage *prim_unfilled( struct softpipe_context *context ); +struct prim_stage *prim_twoside( struct softpipe_context *context ); +struct prim_stage *prim_offset( struct softpipe_context *context ); +struct prim_stage *prim_clip( struct softpipe_context *context ); +struct prim_stage *prim_flatshade( struct softpipe_context *context ); +struct prim_stage *prim_cull( struct softpipe_context *context ); + + +/* Internal structs and helpers for the primitive clip/setup pipeline: + */ +struct prim_stage { + struct softpipe_context *softpipe; + + struct prim_stage *next; + + struct vertex_header **tmp; + GLuint nr_tmps; + + void (*begin)( struct prim_stage * ); + + void (*point)( struct prim_stage *, + struct prim_header * ); + + void (*line)( struct prim_stage *, + struct prim_header * ); + + void (*tri)( struct prim_stage *, + struct prim_header * ); + + void (*end)( struct prim_stage * ); +}; + + + +/* Get a writeable copy of a vertex: + */ +static INLINE struct vertex_header * +dup_vert( struct prim_stage *stage, + const struct vertex_header *vert, + GLuint idx ) +{ + struct vertex_header *tmp = stage->tmp[idx]; + memcpy(tmp, vert, stage->softpipe->prim.vertex_size ); + return tmp; +} + +void prim_free_tmps( struct prim_stage *stage ); +void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); + + +#endif diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c new file mode 100644 index 0000000000..5e9f218d1e --- /dev/null +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -0,0 +1,152 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" +#include "vf/vf.h" + +#include "pipe/p_defines.h" +#include "sp_context.h" +#include "sp_prim.h" + + +struct twoside_stage { + struct prim_stage stage; + + GLfloat facing; + const GLuint *lookup; +}; + + +static INLINE struct twoside_stage *twoside_stage( struct prim_stage *stage ) +{ + return (struct twoside_stage *)stage; +} + + +static void twoside_begin( struct prim_stage *stage ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + twoside->facing = (stage->softpipe->setup.front_winding == PIPE_WINDING_CW) ? -1 : 1; + + stage->next->begin( stage->next ); +} + + +static INLINE void copy_color( GLuint attr_dst, + GLuint attr_src, + struct vertex_header *v ) +{ + if (attr_dst && attr_src) { + memcpy( v->data[attr_dst], + v->data[attr_src], + sizeof(v->data[0]) ); + } +} + + +static struct vertex_header *copy_bfc( struct twoside_stage *twoside, + const struct vertex_header *v, + GLuint idx ) +{ + struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); + + copy_color( twoside->lookup[VF_ATTRIB_COLOR0], + twoside->lookup[VF_ATTRIB_BFC0], + tmp ); + + copy_color( twoside->lookup[VF_ATTRIB_COLOR1], + twoside->lookup[VF_ATTRIB_BFC1], + tmp ); + + return tmp; +} + + +/* Twoside tri: + */ +static void twoside_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + if (header->det * twoside->facing < 0) { + struct prim_header tmp; + + tmp.det = header->det; + tmp.v[0] = copy_bfc(twoside, header->v[0], 0); + tmp.v[1] = copy_bfc(twoside, header->v[1], 1); + tmp.v[2] = copy_bfc(twoside, header->v[2], 2); + + stage->next->tri( stage->next, &tmp ); + } + else { + stage->next->tri( stage->next, header ); + } +} + + +static void twoside_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void twoside_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void twoside_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + + + +struct prim_stage *prim_twoside( struct softpipe_context *softpipe ) +{ + struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); + + prim_alloc_tmps( &twoside->stage, 3 ); + + twoside->stage.softpipe = softpipe; + twoside->stage.next = NULL; + twoside->stage.begin = twoside_begin; + twoside->stage.point = twoside_point; + twoside->stage.line = twoside_line; + twoside->stage.tri = twoside_tri; + twoside->stage.end = twoside_end; + + twoside->lookup = softpipe->vf_attr_to_slot; + + return &twoside->stage; +} diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c new file mode 100644 index 0000000000..ab0dab09d4 --- /dev/null +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -0,0 +1,165 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_prim.h" +#include "pipe/p_defines.h" + + +struct unfilled_stage { + struct prim_stage stage; + + GLuint mode[2]; +}; + + +static INLINE struct unfilled_stage *unfilled_stage( struct prim_stage *stage ) +{ + return (struct unfilled_stage *)stage; +} + + +static void unfilled_begin( struct prim_stage *stage ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + + unfilled->mode[0] = stage->softpipe->setup.fill_ccw; + unfilled->mode[1] = stage->softpipe->setup.fill_cw; + + stage->next->begin( stage->next ); +} + +static void point( struct prim_stage *stage, + struct vertex_header *v0 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + stage->next->point( stage->next, &tmp ); +} + +static void line( struct prim_stage *stage, + struct vertex_header *v0, + struct vertex_header *v1 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + tmp.v[1] = v1; + stage->next->line( stage->next, &tmp ); +} + + +static void points( struct prim_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (v0->edgeflag) point( stage, v0 ); + if (v1->edgeflag) point( stage, v1 ); + if (v2->edgeflag) point( stage, v2 ); +} + +static void lines( struct prim_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (v0->edgeflag) line( stage, v0, v1 ); + if (v1->edgeflag) line( stage, v1, v2 ); + if (v2->edgeflag) line( stage, v2, v0 ); +} + + +/* Unfilled tri: + * + * Note edgeflags in the vertex struct is not sufficient as we will + * need to manipulate them when decomposing primitives??? + */ +static void unfilled_tri( struct prim_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + GLuint mode = unfilled->mode[header->det < 0]; + + switch (mode) { + case PIPE_POLYGON_MODE_FILL: + stage->next->tri( stage->next, header ); + break; + case PIPE_POLYGON_MODE_LINE: + lines( stage, header ); + break; + case PIPE_POLYGON_MODE_POINT: + points( stage, header ); + break; + default: + abort(); + } +} + +static void unfilled_line( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void unfilled_point( struct prim_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void unfilled_end( struct prim_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct prim_stage *prim_unfilled( struct softpipe_context *softpipe ) +{ + struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); + + prim_alloc_tmps( &unfilled->stage, 0 ); + + unfilled->stage.softpipe = softpipe; + unfilled->stage.next = NULL; + unfilled->stage.tmp = NULL; + unfilled->stage.begin = unfilled_begin; + unfilled->stage.point = unfilled_point; + unfilled->stage.line = unfilled_line; + unfilled->stage.tri = unfilled_tri; + unfilled->stage.end = unfilled_end; + + return &unfilled->stage; +} diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c new file mode 100644 index 0000000000..3fc30dd203 --- /dev/null +++ b/src/mesa/pipe/draw/draw_vb.c @@ -0,0 +1,697 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "imports.h" + +#include "tnl/t_context.h" +#include "vf/vf.h" + +#include "sp_context.h" +#include "sp_prim.h" +#include "sp_headers.h" +#include "sp_draw.h" + +/* This file is a temporary set of hooks to allow us to use the tnl/ + * and vf/ modules until we have replacements in pipe. + */ + + +struct draw_context +{ + struct softpipe_context *softpipe; + + struct vf_attr_map attrs[VF_ATTRIB_MAX]; + GLuint nr_attrs; + GLuint vertex_size; + struct vertex_fetch *vf; + + GLubyte *verts; + GLuint nr_vertices; + GLboolean in_vb; + + GLenum prim; + + /* Helper for tnl: + */ + GLvector4f header; +}; + + +static struct vertex_header *get_vertex( struct draw_context *pipe, + GLuint i ) +{ + return (struct vertex_header *)(pipe->verts + i * pipe->vertex_size); +} + + + +static void draw_allocate_vertices( struct draw_context *draw, + GLuint nr_vertices ) +{ + draw->nr_vertices = nr_vertices; + draw->verts = MALLOC( nr_vertices * draw->vertex_size ); + + draw->softpipe->prim.first->begin( draw->softpipe->prim.first ); +} + +static void draw_set_prim( struct draw_context *draw, + GLenum prim ) +{ + draw->prim = prim; + + /* Not done yet - need to force edgeflags to 1 in strip/fan + * primitives. + */ +#if 0 + switch (prim) { + case GL_TRIANGLES: + case GL_POLYGON: + case GL_QUADS: + case GL_QUAD_STRIP: /* yes, we need this */ + respect_edgeflags( pipe, GL_TRUE ); + break; + + default: + respect_edgeflags( pipe, GL_FALSE ); + break; + } +#endif +} + + + +static void do_quad( struct prim_stage *first, + struct vertex_header *v0, + struct vertex_header *v1, + struct vertex_header *v2, + struct vertex_header *v3 ) +{ + struct prim_header prim; + + { + GLuint tmp = v1->edgeflag; + v1->edgeflag = 0; + + prim.v[0] = v0; + prim.v[1] = v1; + prim.v[2] = v3; + first->tri( first, &prim ); + + v1->edgeflag = tmp; + } + + { + GLuint tmp = v3->edgeflag; + v3->edgeflag = 0; + + prim.v[0] = v1; + prim.v[1] = v2; + prim.v[2] = v3; + first->tri( first, &prim ); + + v3->edgeflag = tmp; + } +} + + + + +static void draw_indexed_prim( struct draw_context *draw, + const GLuint *elts, + GLuint count ) +{ + struct prim_stage * const first = draw->softpipe->prim.first; + struct prim_header prim; + GLuint i; + + prim.det = 0; /* valid from cull stage onwards */ + prim.v[0] = 0; + prim.v[1] = 0; + prim.v[2] = 0; + + switch (draw->prim) { + case GL_POINTS: + for (i = 0; i < count; i ++) { + prim.v[0] = get_vertex( draw, elts[i] ); + + first->point( first, &prim ); + } + break; + + case GL_LINES: + for (i = 0; i+1 < count; i += 2) { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + + first->line( first, &prim ); + } + break; + + case GL_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + prim.v[0] = get_vertex( draw, elts[i-1] ); + prim.v[1] = get_vertex( draw, elts[i] ); + first->line( first, &prim ); + } + + prim.v[0] = get_vertex( draw, elts[count-1] ); + prim.v[1] = get_vertex( draw, elts[0] ); + first->line( first, &prim ); + } + break; + + case GL_LINE_STRIP: + /* I'm guessing it will be necessary to have something like a + * render->reset_line_stipple() method to properly support + * splitting strips into primitives like this. Alternately we + * could just scan ahead to find individual clipped lines and + * otherwise leave the strip intact - that might be better, but + * require more complex code here. + */ + if (count >= 2) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, elts[0] ); + + for (i = 1; i < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, elts[i] ); + + first->line( first, &prim ); + } + } + break; + + case GL_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + prim.v[0] = get_vertex( draw, elts[i + 1] ); + prim.v[1] = get_vertex( draw, elts[i + 0] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + } + else { + prim.v[0] = get_vertex( draw, elts[i + 0] ); + prim.v[1] = get_vertex( draw, elts[i + 1] ); + prim.v[2] = get_vertex( draw, elts[i + 2] ); + } + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_FAN: + if (count >= 3) { + prim.v[0] = get_vertex( draw, elts[0] ); + prim.v[1] = 0; + prim.v[2] = get_vertex( draw, elts[1] ); + + for (i = 0; i+2 < count; i++) { + prim.v[1] = prim.v[2]; + prim.v[2] = get_vertex( draw, elts[i+2] ); + + first->tri( first, &prim ); + } + } + break; + + case GL_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( first, + get_vertex( draw, elts[i + 0] ), + get_vertex( draw, elts[i + 1] ), + get_vertex( draw, elts[i + 2] ), + get_vertex( draw, elts[i + 3] )); + } + break; + + case GL_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( first, + get_vertex( draw, elts[i + 2] ), + get_vertex( draw, elts[i + 0] ), + get_vertex( draw, elts[i + 1] ), + get_vertex( draw, elts[i + 3] )); + } + break; + + + case GL_POLYGON: + if (count >= 3) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, elts[1] ); + prim.v[2] = get_vertex( draw, elts[0] ); + + for (i = 0; i+2 < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, elts[i+2] ); + + first->tri( first, &prim ); + } + } + break; + + default: + assert(0); + break; + } +} + +static void draw_prim( struct draw_context *draw, + GLuint start, + GLuint count ) +{ + struct prim_stage * const first = draw->softpipe->prim.first; + struct prim_header prim; + GLuint i; + +// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); + + prim.det = 0; /* valid from cull stage onwards */ + prim.v[0] = 0; + prim.v[1] = 0; + prim.v[2] = 0; + + switch (draw->prim) { + case GL_POINTS: + for (i = 0; i < count; i ++) { + prim.v[0] = get_vertex( draw, start + i ); + first->point( first, &prim ); + } + break; + + case GL_LINES: + for (i = 0; i+1 < count; i += 2) { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + + first->line( first, &prim ); + } + break; + + case GL_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + prim.v[0] = get_vertex( draw, start + i - 1 ); + prim.v[1] = get_vertex( draw, start + i ); + first->line( first, &prim ); + } + + prim.v[0] = get_vertex( draw, start + count - 1 ); + prim.v[1] = get_vertex( draw, start + 0 ); + first->line( first, &prim ); + } + break; + + case GL_LINE_STRIP: + if (count >= 2) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, start + 0 ); + + for (i = 1; i < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, start + i ); + + first->line( first, &prim ); + } + } + break; + + case GL_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + prim.v[0] = get_vertex( draw, start + i + 1 ); + prim.v[1] = get_vertex( draw, start + i + 0 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + } + else { + prim.v[0] = get_vertex( draw, start + i + 0 ); + prim.v[1] = get_vertex( draw, start + i + 1 ); + prim.v[2] = get_vertex( draw, start + i + 2 ); + } + + first->tri( first, &prim ); + } + break; + + case GL_TRIANGLE_FAN: + if (count >= 3) { + prim.v[0] = get_vertex( draw, start + 0 ); + prim.v[1] = 0; + prim.v[2] = get_vertex( draw, start + 1 ); + + for (i = 0; i+2 < count; i++) { + prim.v[1] = prim.v[2]; + prim.v[2] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + } + break; + + + case GL_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( first, + get_vertex( draw, start + i + 0 ), + get_vertex( draw, start + i + 1 ), + get_vertex( draw, start + i + 2 ), + get_vertex( draw, start + i + 3 )); + } + break; + + case GL_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( first, + get_vertex( draw, start + i + 2 ), + get_vertex( draw, start + i + 0 ), + get_vertex( draw, start + i + 1 ), + get_vertex( draw, start + i + 3 )); + } + break; + + case GL_POLYGON: + if (count >= 3) { + prim.v[0] = 0; + prim.v[1] = get_vertex( draw, start + 1 ); + prim.v[2] = get_vertex( draw, start + 0 ); + + for (i = 0; i+2 < count; i++) { + prim.v[0] = prim.v[1]; + prim.v[1] = get_vertex( draw, start + i + 2 ); + + first->tri( first, &prim ); + } + } + break; + + default: + assert(0); + break; + } +} + + +static void draw_release_vertices( struct draw_context *draw ) +{ + draw->softpipe->prim.first->end( draw->softpipe->prim.first ); + + FREE(draw->verts); + draw->verts = NULL; +} + + +struct header_dword { + GLuint clipmask:12; + GLuint edgeflag:1; + GLuint pad:19; +}; + + +static void +build_vertex_headers( struct draw_context *draw, + struct vertex_buffer *VB ) +{ + if (draw->header.storage == NULL) { + draw->header.stride = sizeof(GLfloat); + draw->header.size = 1; + draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(GLfloat), 32 ); + draw->header.data = draw->header.storage; + draw->header.count = 0; + draw->header.flags = VEC_SIZE_1 | VEC_MALLOC; + } + + /* Build vertex header attribute. + * + */ + + { + GLuint i; + struct header_dword *header = (struct header_dword *)draw->header.storage; + + /* yes its a hack + */ + assert(sizeof(*header) == sizeof(GLfloat)); + + draw->header.count = VB->Count; + + if (VB->EdgeFlag) { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = VB->ClipMask[i]; + header[i].edgeflag = VB->EdgeFlag[i]; + header[i].pad = 0; + } + } + else if (VB->ClipOrMask) { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = VB->ClipMask[i]; + header[i].edgeflag = 0; + header[i].pad = 0; + } + } + else { + for (i = 0; i < VB->Count; i++) { + header[i].clipmask = 0; + header[i].edgeflag = 0; + header[i].pad = 0; + } + } + } + + VB->AttribPtr[VF_ATTRIB_VERTEX_HEADER] = &draw->header; +} + + + + + +static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) +{ + switch (mode) { + case GL_POINTS: + *first = 1; + *incr = 1; + return 0; + case GL_LINES: + *first = 2; + *incr = 2; + return 0; + case GL_LINE_STRIP: + *first = 2; + *incr = 1; + return 0; + case GL_LINE_LOOP: + *first = 2; + *incr = 1; + return 1; + case GL_TRIANGLES: + *first = 3; + *incr = 3; + return 0; + case GL_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + return 0; + case GL_TRIANGLE_FAN: + case GL_POLYGON: + *first = 3; + *incr = 1; + return 1; + case GL_QUADS: + *first = 4; + *incr = 4; + return 0; + case GL_QUAD_STRIP: + *first = 4; + *incr = 2; + return 0; + default: + assert(0); + *first = 1; + *incr = 1; + return 0; + } +} + + +static GLuint trim( GLuint count, GLuint first, GLuint incr ) +{ + if (count < first) + return 0; + else + return count - (count - first) % incr; +} + + +/* This is a hack & will all go away. + */ +void draw_vb(struct draw_context *draw, + struct vertex_buffer *VB ) +{ + GLuint i; + + VB->AttribPtr[VF_ATTRIB_POS] = VB->NdcPtr; + VB->AttribPtr[VF_ATTRIB_BFC0] = VB->ColorPtr[1]; + VB->AttribPtr[VF_ATTRIB_BFC1] = VB->SecondaryColorPtr[1]; + VB->AttribPtr[VF_ATTRIB_CLIP_POS] = VB->ClipPtr; + + /* Build vertex headers: + */ + build_vertex_headers( draw, VB ); + + draw->in_vb = 1; + + /* Allocate the vertices: + */ + draw_allocate_vertices( draw, VB->Count ); + + /* Bind the vb outputs: + */ + vf_set_sources( draw->vf, VB->AttribPtr, 0 ); + + /* Build the hardware or prim-pipe vertices: + */ + vf_emit_vertices( draw->vf, VB->Count, draw->verts ); + + + for (i = 0; i < VB->PrimitiveCount; i++) { + + GLenum mode = VB->Primitive[i].mode; + GLuint start = VB->Primitive[i].start; + GLuint length, first, incr; + + /* Trim the primitive down to a legal size. + */ + draw_prim_info( mode, &first, &incr ); + length = trim( VB->Primitive[i].count, first, incr ); + + if (!length) + continue; + + if (draw->prim != mode) + draw_set_prim( draw, mode ); + + if (VB->Elts) { + draw_indexed_prim( draw, + VB->Elts + start, + length ); + } + else { + draw_prim( draw, + start, + length ); + } + } + + draw_release_vertices( draw ); + draw->verts = NULL; + draw->in_vb = 0; +} + +void draw_set_viewport( struct draw_context *draw, + const GLfloat *scale, + const GLfloat *translate ) +{ + assert(!draw->in_vb); + vf_set_vp_scale_translate( draw->vf, scale, translate ); +} + + + +struct draw_context *draw_create( struct softpipe_context *softpipe ) +{ + struct draw_context *draw = CALLOC_STRUCT( draw_context ); + draw->softpipe = softpipe; + draw->vf = vf_create( GL_TRUE ); + + return draw; +} + + +void draw_destroy( struct draw_context *draw ) +{ + if (draw->header.storage) + ALIGN_FREE( draw->header.storage ); + + vf_destroy( draw->vf ); + + FREE( draw ); +} + +#define EMIT_ATTR( ATTR, STYLE ) \ +do { \ + draw->attrs[draw->nr_attrs].attrib = ATTR; \ + draw->attrs[draw->nr_attrs].format = STYLE; \ + draw->nr_attrs++; \ +} while (0) + + +void draw_set_vertex_attributes( struct draw_context *draw, + const GLuint *attrs, + GLuint nr_attrs ) +{ + GLuint i; + + draw->nr_attrs = 0; + + EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F); + EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F); + + assert(attrs[0] == VF_ATTRIB_POS); + EMIT_ATTR(attrs[0], EMIT_4F_VIEWPORT); + + for (i = 1; i < nr_attrs; i++) + EMIT_ATTR(attrs[i], EMIT_4F); + + draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); +} + + diff --git a/src/mesa/pipe/softpipe/sp_draw.c b/src/mesa/pipe/softpipe/sp_draw.c deleted file mode 100644 index 3fc30dd203..0000000000 --- a/src/mesa/pipe/softpipe/sp_draw.c +++ /dev/null @@ -1,697 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "imports.h" - -#include "tnl/t_context.h" -#include "vf/vf.h" - -#include "sp_context.h" -#include "sp_prim.h" -#include "sp_headers.h" -#include "sp_draw.h" - -/* This file is a temporary set of hooks to allow us to use the tnl/ - * and vf/ modules until we have replacements in pipe. - */ - - -struct draw_context -{ - struct softpipe_context *softpipe; - - struct vf_attr_map attrs[VF_ATTRIB_MAX]; - GLuint nr_attrs; - GLuint vertex_size; - struct vertex_fetch *vf; - - GLubyte *verts; - GLuint nr_vertices; - GLboolean in_vb; - - GLenum prim; - - /* Helper for tnl: - */ - GLvector4f header; -}; - - -static struct vertex_header *get_vertex( struct draw_context *pipe, - GLuint i ) -{ - return (struct vertex_header *)(pipe->verts + i * pipe->vertex_size); -} - - - -static void draw_allocate_vertices( struct draw_context *draw, - GLuint nr_vertices ) -{ - draw->nr_vertices = nr_vertices; - draw->verts = MALLOC( nr_vertices * draw->vertex_size ); - - draw->softpipe->prim.first->begin( draw->softpipe->prim.first ); -} - -static void draw_set_prim( struct draw_context *draw, - GLenum prim ) -{ - draw->prim = prim; - - /* Not done yet - need to force edgeflags to 1 in strip/fan - * primitives. - */ -#if 0 - switch (prim) { - case GL_TRIANGLES: - case GL_POLYGON: - case GL_QUADS: - case GL_QUAD_STRIP: /* yes, we need this */ - respect_edgeflags( pipe, GL_TRUE ); - break; - - default: - respect_edgeflags( pipe, GL_FALSE ); - break; - } -#endif -} - - - -static void do_quad( struct prim_stage *first, - struct vertex_header *v0, - struct vertex_header *v1, - struct vertex_header *v2, - struct vertex_header *v3 ) -{ - struct prim_header prim; - - { - GLuint tmp = v1->edgeflag; - v1->edgeflag = 0; - - prim.v[0] = v0; - prim.v[1] = v1; - prim.v[2] = v3; - first->tri( first, &prim ); - - v1->edgeflag = tmp; - } - - { - GLuint tmp = v3->edgeflag; - v3->edgeflag = 0; - - prim.v[0] = v1; - prim.v[1] = v2; - prim.v[2] = v3; - first->tri( first, &prim ); - - v3->edgeflag = tmp; - } -} - - - - -static void draw_indexed_prim( struct draw_context *draw, - const GLuint *elts, - GLuint count ) -{ - struct prim_stage * const first = draw->softpipe->prim.first; - struct prim_header prim; - GLuint i; - - prim.det = 0; /* valid from cull stage onwards */ - prim.v[0] = 0; - prim.v[1] = 0; - prim.v[2] = 0; - - switch (draw->prim) { - case GL_POINTS: - for (i = 0; i < count; i ++) { - prim.v[0] = get_vertex( draw, elts[i] ); - - first->point( first, &prim ); - } - break; - - case GL_LINES: - for (i = 0; i+1 < count; i += 2) { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - - first->line( first, &prim ); - } - break; - - case GL_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - prim.v[0] = get_vertex( draw, elts[i-1] ); - prim.v[1] = get_vertex( draw, elts[i] ); - first->line( first, &prim ); - } - - prim.v[0] = get_vertex( draw, elts[count-1] ); - prim.v[1] = get_vertex( draw, elts[0] ); - first->line( first, &prim ); - } - break; - - case GL_LINE_STRIP: - /* I'm guessing it will be necessary to have something like a - * render->reset_line_stipple() method to properly support - * splitting strips into primitives like this. Alternately we - * could just scan ahead to find individual clipped lines and - * otherwise leave the strip intact - that might be better, but - * require more complex code here. - */ - if (count >= 2) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, elts[0] ); - - for (i = 1; i < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, elts[i] ); - - first->line( first, &prim ); - } - } - break; - - case GL_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - prim.v[0] = get_vertex( draw, elts[i + 1] ); - prim.v[1] = get_vertex( draw, elts[i + 0] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - } - else { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - } - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_FAN: - if (count >= 3) { - prim.v[0] = get_vertex( draw, elts[0] ); - prim.v[1] = 0; - prim.v[2] = get_vertex( draw, elts[1] ); - - for (i = 0; i+2 < count; i++) { - prim.v[1] = prim.v[2]; - prim.v[2] = get_vertex( draw, elts[i+2] ); - - first->tri( first, &prim ); - } - } - break; - - case GL_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( first, - get_vertex( draw, elts[i + 0] ), - get_vertex( draw, elts[i + 1] ), - get_vertex( draw, elts[i + 2] ), - get_vertex( draw, elts[i + 3] )); - } - break; - - case GL_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( first, - get_vertex( draw, elts[i + 2] ), - get_vertex( draw, elts[i + 0] ), - get_vertex( draw, elts[i + 1] ), - get_vertex( draw, elts[i + 3] )); - } - break; - - - case GL_POLYGON: - if (count >= 3) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, elts[1] ); - prim.v[2] = get_vertex( draw, elts[0] ); - - for (i = 0; i+2 < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, elts[i+2] ); - - first->tri( first, &prim ); - } - } - break; - - default: - assert(0); - break; - } -} - -static void draw_prim( struct draw_context *draw, - GLuint start, - GLuint count ) -{ - struct prim_stage * const first = draw->softpipe->prim.first; - struct prim_header prim; - GLuint i; - -// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - - prim.det = 0; /* valid from cull stage onwards */ - prim.v[0] = 0; - prim.v[1] = 0; - prim.v[2] = 0; - - switch (draw->prim) { - case GL_POINTS: - for (i = 0; i < count; i ++) { - prim.v[0] = get_vertex( draw, start + i ); - first->point( first, &prim ); - } - break; - - case GL_LINES: - for (i = 0; i+1 < count; i += 2) { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - - first->line( first, &prim ); - } - break; - - case GL_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - prim.v[0] = get_vertex( draw, start + i - 1 ); - prim.v[1] = get_vertex( draw, start + i ); - first->line( first, &prim ); - } - - prim.v[0] = get_vertex( draw, start + count - 1 ); - prim.v[1] = get_vertex( draw, start + 0 ); - first->line( first, &prim ); - } - break; - - case GL_LINE_STRIP: - if (count >= 2) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, start + 0 ); - - for (i = 1; i < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, start + i ); - - first->line( first, &prim ); - } - } - break; - - case GL_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - prim.v[0] = get_vertex( draw, start + i + 1 ); - prim.v[1] = get_vertex( draw, start + i + 0 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - } - else { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - } - - first->tri( first, &prim ); - } - break; - - case GL_TRIANGLE_FAN: - if (count >= 3) { - prim.v[0] = get_vertex( draw, start + 0 ); - prim.v[1] = 0; - prim.v[2] = get_vertex( draw, start + 1 ); - - for (i = 0; i+2 < count; i++) { - prim.v[1] = prim.v[2]; - prim.v[2] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); - } - } - break; - - - case GL_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( first, - get_vertex( draw, start + i + 0 ), - get_vertex( draw, start + i + 1 ), - get_vertex( draw, start + i + 2 ), - get_vertex( draw, start + i + 3 )); - } - break; - - case GL_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( first, - get_vertex( draw, start + i + 2 ), - get_vertex( draw, start + i + 0 ), - get_vertex( draw, start + i + 1 ), - get_vertex( draw, start + i + 3 )); - } - break; - - case GL_POLYGON: - if (count >= 3) { - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, start + 1 ); - prim.v[2] = get_vertex( draw, start + 0 ); - - for (i = 0; i+2 < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); - } - } - break; - - default: - assert(0); - break; - } -} - - -static void draw_release_vertices( struct draw_context *draw ) -{ - draw->softpipe->prim.first->end( draw->softpipe->prim.first ); - - FREE(draw->verts); - draw->verts = NULL; -} - - -struct header_dword { - GLuint clipmask:12; - GLuint edgeflag:1; - GLuint pad:19; -}; - - -static void -build_vertex_headers( struct draw_context *draw, - struct vertex_buffer *VB ) -{ - if (draw->header.storage == NULL) { - draw->header.stride = sizeof(GLfloat); - draw->header.size = 1; - draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(GLfloat), 32 ); - draw->header.data = draw->header.storage; - draw->header.count = 0; - draw->header.flags = VEC_SIZE_1 | VEC_MALLOC; - } - - /* Build vertex header attribute. - * - */ - - { - GLuint i; - struct header_dword *header = (struct header_dword *)draw->header.storage; - - /* yes its a hack - */ - assert(sizeof(*header) == sizeof(GLfloat)); - - draw->header.count = VB->Count; - - if (VB->EdgeFlag) { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = VB->ClipMask[i]; - header[i].edgeflag = VB->EdgeFlag[i]; - header[i].pad = 0; - } - } - else if (VB->ClipOrMask) { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = VB->ClipMask[i]; - header[i].edgeflag = 0; - header[i].pad = 0; - } - } - else { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = 0; - header[i].edgeflag = 0; - header[i].pad = 0; - } - } - } - - VB->AttribPtr[VF_ATTRIB_VERTEX_HEADER] = &draw->header; -} - - - - - -static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) -{ - switch (mode) { - case GL_POINTS: - *first = 1; - *incr = 1; - return 0; - case GL_LINES: - *first = 2; - *incr = 2; - return 0; - case GL_LINE_STRIP: - *first = 2; - *incr = 1; - return 0; - case GL_LINE_LOOP: - *first = 2; - *incr = 1; - return 1; - case GL_TRIANGLES: - *first = 3; - *incr = 3; - return 0; - case GL_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - return 0; - case GL_TRIANGLE_FAN: - case GL_POLYGON: - *first = 3; - *incr = 1; - return 1; - case GL_QUADS: - *first = 4; - *incr = 4; - return 0; - case GL_QUAD_STRIP: - *first = 4; - *incr = 2; - return 0; - default: - assert(0); - *first = 1; - *incr = 1; - return 0; - } -} - - -static GLuint trim( GLuint count, GLuint first, GLuint incr ) -{ - if (count < first) - return 0; - else - return count - (count - first) % incr; -} - - -/* This is a hack & will all go away. - */ -void draw_vb(struct draw_context *draw, - struct vertex_buffer *VB ) -{ - GLuint i; - - VB->AttribPtr[VF_ATTRIB_POS] = VB->NdcPtr; - VB->AttribPtr[VF_ATTRIB_BFC0] = VB->ColorPtr[1]; - VB->AttribPtr[VF_ATTRIB_BFC1] = VB->SecondaryColorPtr[1]; - VB->AttribPtr[VF_ATTRIB_CLIP_POS] = VB->ClipPtr; - - /* Build vertex headers: - */ - build_vertex_headers( draw, VB ); - - draw->in_vb = 1; - - /* Allocate the vertices: - */ - draw_allocate_vertices( draw, VB->Count ); - - /* Bind the vb outputs: - */ - vf_set_sources( draw->vf, VB->AttribPtr, 0 ); - - /* Build the hardware or prim-pipe vertices: - */ - vf_emit_vertices( draw->vf, VB->Count, draw->verts ); - - - for (i = 0; i < VB->PrimitiveCount; i++) { - - GLenum mode = VB->Primitive[i].mode; - GLuint start = VB->Primitive[i].start; - GLuint length, first, incr; - - /* Trim the primitive down to a legal size. - */ - draw_prim_info( mode, &first, &incr ); - length = trim( VB->Primitive[i].count, first, incr ); - - if (!length) - continue; - - if (draw->prim != mode) - draw_set_prim( draw, mode ); - - if (VB->Elts) { - draw_indexed_prim( draw, - VB->Elts + start, - length ); - } - else { - draw_prim( draw, - start, - length ); - } - } - - draw_release_vertices( draw ); - draw->verts = NULL; - draw->in_vb = 0; -} - -void draw_set_viewport( struct draw_context *draw, - const GLfloat *scale, - const GLfloat *translate ) -{ - assert(!draw->in_vb); - vf_set_vp_scale_translate( draw->vf, scale, translate ); -} - - - -struct draw_context *draw_create( struct softpipe_context *softpipe ) -{ - struct draw_context *draw = CALLOC_STRUCT( draw_context ); - draw->softpipe = softpipe; - draw->vf = vf_create( GL_TRUE ); - - return draw; -} - - -void draw_destroy( struct draw_context *draw ) -{ - if (draw->header.storage) - ALIGN_FREE( draw->header.storage ); - - vf_destroy( draw->vf ); - - FREE( draw ); -} - -#define EMIT_ATTR( ATTR, STYLE ) \ -do { \ - draw->attrs[draw->nr_attrs].attrib = ATTR; \ - draw->attrs[draw->nr_attrs].format = STYLE; \ - draw->nr_attrs++; \ -} while (0) - - -void draw_set_vertex_attributes( struct draw_context *draw, - const GLuint *attrs, - GLuint nr_attrs ) -{ - GLuint i; - - draw->nr_attrs = 0; - - EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F); - EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F); - - assert(attrs[0] == VF_ATTRIB_POS); - EMIT_ATTR(attrs[0], EMIT_4F_VIEWPORT); - - for (i = 1; i < nr_attrs; i++) - EMIT_ATTR(attrs[i], EMIT_4F); - - draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); -} - - diff --git a/src/mesa/pipe/softpipe/sp_draw.h b/src/mesa/pipe/softpipe/sp_draw.h deleted file mode 100644 index a138f812fa..0000000000 --- a/src/mesa/pipe/softpipe/sp_draw.h +++ /dev/null @@ -1,56 +0,0 @@ - -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_DRAW_H -#define G_DRAW_H - -#include "glheader.h" -#include "pipe/p_state.h" - - -struct draw_context; - -struct draw_context *draw_create( struct softpipe_context *softpipe ); - -void draw_destroy( struct draw_context *draw ); - -void draw_set_viewport( struct draw_context *draw, - const GLfloat *scale, - const GLfloat *translate ); - -void draw_set_vertex_attributes( struct draw_context *draw, - const GLuint *attrs, - GLuint nr_attrs ); - -void draw_vb(struct draw_context *draw, - struct vertex_buffer *VB ); - -#endif diff --git a/src/mesa/pipe/softpipe/sp_prim.h b/src/mesa/pipe/softpipe/sp_prim.h deleted file mode 100644 index b6cbaae085..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim.h +++ /dev/null @@ -1,90 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef G_PRIM_H -#define G_PRIM_H - -#include "glheader.h" -#include "sp_headers.h" - -struct softpipe_context; - -struct prim_stage *prim_setup( struct softpipe_context *context ); -struct prim_stage *prim_unfilled( struct softpipe_context *context ); -struct prim_stage *prim_twoside( struct softpipe_context *context ); -struct prim_stage *prim_offset( struct softpipe_context *context ); -struct prim_stage *prim_clip( struct softpipe_context *context ); -struct prim_stage *prim_flatshade( struct softpipe_context *context ); -struct prim_stage *prim_cull( struct softpipe_context *context ); - - -/* Internal structs and helpers for the primitive clip/setup pipeline: - */ -struct prim_stage { - struct softpipe_context *softpipe; - - struct prim_stage *next; - - struct vertex_header **tmp; - GLuint nr_tmps; - - void (*begin)( struct prim_stage * ); - - void (*point)( struct prim_stage *, - struct prim_header * ); - - void (*line)( struct prim_stage *, - struct prim_header * ); - - void (*tri)( struct prim_stage *, - struct prim_header * ); - - void (*end)( struct prim_stage * ); -}; - - - -/* Get a writeable copy of a vertex: - */ -static INLINE struct vertex_header * -dup_vert( struct prim_stage *stage, - const struct vertex_header *vert, - GLuint idx ) -{ - struct vertex_header *tmp = stage->tmp[idx]; - memcpy(tmp, vert, stage->softpipe->prim.vertex_size ); - return tmp; -} - -void prim_free_tmps( struct prim_stage *stage ); -void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); - - -#endif diff --git a/src/mesa/pipe/softpipe/sp_prim_clip.c b/src/mesa/pipe/softpipe/sp_prim_clip.c deleted file mode 100644 index 304c43c3f4..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_clip.c +++ /dev/null @@ -1,429 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "imports.h" -#include "macros.h" - -#include "sp_context.h" -#include "sp_prim.h" - -struct clipper { - struct prim_stage stage; - - GLuint active_user_planes; -}; - -/* This is a bit confusing: - */ -static INLINE struct clipper *clipper_stage( struct prim_stage *stage ) -{ - return (struct clipper *)stage; -} - - -#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) - - -/* All attributes are float[4], so this is easy: - */ -static void interp_attr( GLfloat *fdst, - GLfloat t, - const GLfloat *fin, - const GLfloat *fout ) -{ - fdst[0] = LINTERP( t, fout[0], fin[0] ); - fdst[1] = LINTERP( t, fout[1], fin[1] ); - fdst[2] = LINTERP( t, fout[2], fin[2] ); - fdst[3] = LINTERP( t, fout[3], fin[3] ); -} - - - - -/* Interpolate between two vertices to produce a third. - */ -static void interp( struct clipper *clip, - struct vertex_header *dst, - GLfloat t, - const struct vertex_header *out, - const struct vertex_header *in ) -{ - const GLuint nr_attrs = clip->stage.softpipe->nr_attrs; - GLuint j; - - /* Vertex header. - */ - { - dst->clipmask = 0; - dst->edgeflag = 0; - dst->pad = 0; - } - - /* Clip coordinates: interpolate normally - */ - { - interp_attr(dst->clip, t, in->clip, out->clip); - } - - /* Do the projective divide and insert window coordinates: - */ - { - const GLfloat *pos = dst->clip; - const GLfloat *scale = clip->stage.softpipe->viewport.scale; - const GLfloat *trans = clip->stage.softpipe->viewport.translate; - GLfloat oow; - - oow = 1.0 / pos[3]; - - dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; - dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; - dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; - dst->data[0][3] = oow; - } - - - /* Other attributes - */ - for (j = 1; j < nr_attrs-1; j++) { - interp_attr(dst->data[j], t, in->data[j], out->data[j]); - } -} - - -#define CLIP_USER_BIT 0x40 -#define CLIP_CULL_BIT 0x80 - - -static INLINE GLfloat dot4( const GLfloat *a, - const GLfloat *b ) -{ - GLfloat result = (a[0]*b[0] + - a[1]*b[1] + - a[2]*b[2] + - a[3]*b[3]); - - return result; -} - - -#if 0 -static INLINE void do_tri( struct prim_stage *next, - struct prim_header *header ) -{ - GLuint i; - for (i = 0; i < 3; i++) { - GLfloat *ndc = header->v[i]->data[0]; - _mesa_printf("ndc %f %f %f\n", ndc[0], ndc[1], ndc[2]); - assert(ndc[0] >= -1 && ndc[0] <= 641); - assert(ndc[1] >= 30 && ndc[1] <= 481); - } - _mesa_printf("\n"); - next->tri(next, header); -} -#endif - - -static void emit_poly( struct prim_stage *stage, - struct vertex_header **inlist, - GLuint n ) -{ - struct prim_header header; - GLuint i; - - for (i = 2; i < n; i++) { - header.v[0] = inlist[0]; - header.v[1] = inlist[i-1]; - header.v[2] = inlist[i]; - - { - GLuint tmp0 = header.v[0]->edgeflag; - GLuint tmp2 = header.v[2]->edgeflag; - - if (i != 2) header.v[0]->edgeflag = 0; - if (i != n-1) header.v[2]->edgeflag = 0; - - stage->next->tri( stage->next, &header ); - - header.v[0]->edgeflag = tmp0; - header.v[2]->edgeflag = tmp2; - } - } -} - - -#if 0 -static void emit_poly( struct prim_stage *stage ) -{ - GLuint i; - - for (i = 2; i < n; i++) { - header->v[0] = inlist[0]; - header->v[1] = inlist[i-1]; - header->v[2] = inlist[i]; - - stage->next->tri( stage->next, header ); - } -} -#endif - - -/* Clip a triangle against the viewport and user clip planes. - */ -static void -do_clip_tri( struct prim_stage *stage, - struct prim_header *header, - GLuint clipmask ) -{ - struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *a[MAX_CLIPPED_VERTICES]; - struct vertex_header *b[MAX_CLIPPED_VERTICES]; - struct vertex_header **inlist = a; - struct vertex_header **outlist = b; - GLuint tmpnr = 0; - GLuint n = 3; - GLuint i; - - inlist[0] = header->v[0]; - inlist[1] = header->v[1]; - inlist[2] = header->v[2]; - - /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove - * this once we correctly use 16bit masks for userclip planes. - */ - clipmask &= ~CLIP_CULL_BIT; - if (clipmask & CLIP_USER_BIT) { - clipmask &= ~CLIP_USER_BIT; - clipmask |= clipper->active_user_planes; - } - - while (clipmask && n >= 3) { - GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->stage.softpipe->plane[plane_idx]; - struct vertex_header *vert_prev = inlist[0]; - GLfloat dp_prev = dot4( vert_prev->clip, plane ); - GLuint outcount = 0; - - clipmask &= ~(1<clip, plane ); - - if (!IS_NEGATIVE(dp_prev)) { - outlist[outcount++] = vert_prev; - } - - if (DIFFERENT_SIGNS(dp, dp_prev)) { - struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; - outlist[outcount++] = new_vert; - - if (IS_NEGATIVE(dp)) { - /* Going out of bounds. Avoid division by zero as we - * know dp != dp_prev from DIFFERENT_SIGNS, above. - */ - GLfloat t = dp / (dp - dp_prev); - interp( clipper, new_vert, t, vert, vert_prev ); - - /* Force edgeflag true in this case: - */ - new_vert->edgeflag = 1; - } else { - /* Coming back in. - */ - GLfloat t = dp_prev / (dp_prev - dp); - interp( clipper, new_vert, t, vert_prev, vert ); - - /* Copy starting vert's edgeflag: - */ - new_vert->edgeflag = vert_prev->edgeflag; - } - } - - vert_prev = vert; - dp_prev = dp; - } - - { - struct vertex_header **tmp = inlist; - inlist = outlist; - outlist = tmp; - n = outcount; - } - } - - /* Emit the polygon as triangles to the setup stage: - */ - if (n >= 3) - emit_poly( stage, inlist, n ); -} - - -/* Clip a line against the viewport and user clip planes. - */ -static void -do_clip_line( struct prim_stage *stage, - struct prim_header *header, - GLuint clipmask ) -{ - struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - const GLfloat *pos0 = v0->clip; - const GLfloat *pos1 = v1->clip; - GLfloat t0 = 0; - GLfloat t1 = 0; - - /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove - * this once we correctly use 16bit masks for userclip planes. - */ - clipmask &= ~CLIP_CULL_BIT; - if (clipmask & CLIP_USER_BIT) { - clipmask &= ~CLIP_USER_BIT; - clipmask |= clipper->active_user_planes; - } - - while (clipmask) { - GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->stage.softpipe->plane[plane_idx]; - - clipmask &= ~(1< t1) t1 = t; - } - - if (dp0 < 0) { - GLfloat t = dp0 / (dp0 - dp1); - if (t > t0) t0 = t; - } - - if (t0 + t1 >= 1.0) - return; /* discard */ - } - - if (v0->clipmask) { - interp( clipper, stage->tmp[0], t0, v0, v1 ); - header->v[0] = stage->tmp[0]; - } - - if (v1->clipmask) { - interp( clipper, stage->tmp[1], t1, v1, v0 ); - header->v[1] = stage->tmp[1]; - } - - stage->next->line( stage->next, header ); -} - - - -static void clip_begin( struct prim_stage *stage ) -{ - struct clipper *clipper = clipper_stage(stage); - GLuint nr = stage->softpipe->nr_planes; - - /* Hacky bitmask to use when we hit CLIP_USER_BIT: - */ - clipper->active_user_planes = ((1<next->begin( stage->next ); -} - -static void -clip_point( struct prim_stage *stage, - struct prim_header *header ) -{ - if (header->v[0]->clipmask == 0) - stage->next->point( stage->next, header ); -} - - -static void -clip_line( struct prim_stage *stage, - struct prim_header *header ) -{ - GLuint clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask); - - if (clipmask == 0) { - stage->next->line( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask) == 0) { - do_clip_line(stage, header, clipmask); - } -} - - -static void -clip_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - GLuint clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask | - header->v[2]->clipmask); - - if (clipmask == 0) { - stage->next->tri( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask & - header->v[2]->clipmask) == 0) { - do_clip_tri(stage, header, clipmask); - } -} - -static void clip_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - - -struct prim_stage *prim_clip( struct softpipe_context *softpipe ) -{ - struct clipper *clipper = CALLOC_STRUCT(clipper); - - prim_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); - - clipper->stage.softpipe = softpipe; - clipper->stage.begin = clip_begin; - clipper->stage.point = clip_point; - clipper->stage.line = clip_line; - clipper->stage.tri = clip_tri; - clipper->stage.end = clip_end; - - return &clipper->stage; -} diff --git a/src/mesa/pipe/softpipe/sp_prim_cull.c b/src/mesa/pipe/softpipe/sp_prim_cull.c deleted file mode 100644 index 63099fbee0..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_cull.c +++ /dev/null @@ -1,120 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "pipe/p_defines.h" -#include "sp_context.h" -#include "sp_prim.h" - - - -struct cull_stage { - struct prim_stage stage; - GLuint mode; -}; - - - -static INLINE struct cull_stage *cull_stage( struct prim_stage *stage ) -{ - return (struct cull_stage *)stage; -} - - -static void cull_begin( struct prim_stage *stage ) -{ - struct cull_stage *cull = cull_stage(stage); - - cull->mode = stage->softpipe->setup.cull_mode; - - stage->next->begin( stage->next ); -} - - -static void cull_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - /* Window coords: */ - GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); - GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); - GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); - - GLfloat ex = v0[0] - v2[0]; - GLfloat ey = v0[1] - v2[1]; - GLfloat fx = v1[0] - v2[0]; - GLfloat fy = v1[1] - v2[1]; - - header->det = ex * fy - ey * fx; - - _mesa_printf("%s %f\n", __FUNCTION__, header->det ); - - if (header->det != 0) { - GLuint mode = (header->det < 0) ? PIPE_WINDING_CW : PIPE_WINDING_CCW; - - if ((mode & cull_stage(stage)->mode) == 0) - stage->next->tri( stage->next, header ); - } -} - - -static void cull_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void cull_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void cull_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_cull( struct softpipe_context *softpipe ) -{ - struct cull_stage *cull = CALLOC_STRUCT(cull_stage); - - prim_alloc_tmps( &cull->stage, 0 ); - - cull->stage.softpipe = softpipe; - cull->stage.next = NULL; - cull->stage.begin = cull_begin; - cull->stage.point = cull_point; - cull->stage.line = cull_line; - cull->stage.tri = cull_tri; - cull->stage.end = cull_end; - - return &cull->stage; -} diff --git a/src/mesa/pipe/softpipe/sp_prim_flatshade.c b/src/mesa/pipe/softpipe/sp_prim_flatshade.c deleted file mode 100644 index 3a7d9de466..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_flatshade.c +++ /dev/null @@ -1,151 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "vf/vf.h" - -#include "sp_context.h" -#include "sp_prim.h" - - -struct flatshade_stage { - struct prim_stage stage; - - const GLuint *lookup; -}; - - - -static INLINE struct flatshade_stage *flatshade_stage( struct prim_stage *stage ) -{ - return (struct flatshade_stage *)stage; -} - - -static void flatshade_begin( struct prim_stage *stage ) -{ - stage->next->begin( stage->next ); -} - - - -static INLINE void copy_attr( GLuint attr, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - if (attr) { - memcpy( dst->data[attr], - src->data[attr], - sizeof(src->data[0]) ); - } -} - -static void copy_colors( struct prim_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - struct flatshade_stage *flatshade = flatshade_stage(stage); - const GLuint *lookup = flatshade->lookup; - - copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); - copy_attr( lookup[VF_ATTRIB_COLOR1], dst, src ); - copy_attr( lookup[VF_ATTRIB_BFC0], dst, src ); - copy_attr( lookup[VF_ATTRIB_BFC1], dst, src ); -} - - - -/* Flatshade tri. Required for clipping and when unfilled tris are - * active, otherwise handled by hardware. - */ -static void flatshade_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = header->v[2]; - - copy_colors(stage, tmp.v[0], tmp.v[2]); - copy_colors(stage, tmp.v[1], tmp.v[2]); - - stage->next->tri( stage->next, &tmp ); -} - - -/* Flatshade line. Required for clipping. - */ -static void flatshade_line( struct prim_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = header->v[1]; - - copy_colors(stage, tmp.v[0], tmp.v[1]); - - stage->next->line( stage->next, &tmp ); -} - - -static void flatshade_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void flatshade_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_flatshade( struct softpipe_context *softpipe ) -{ - struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); - - prim_alloc_tmps( &flatshade->stage, 2 ); - - flatshade->stage.softpipe = softpipe; - flatshade->stage.next = NULL; - flatshade->stage.begin = flatshade_begin; - flatshade->stage.point = flatshade_point; - flatshade->stage.line = flatshade_line; - flatshade->stage.tri = flatshade_tri; - flatshade->stage.end = flatshade_end; - - flatshade->lookup = softpipe->vf_attr_to_slot; - - return &flatshade->stage; -} - - diff --git a/src/mesa/pipe/softpipe/sp_prim_offset.c b/src/mesa/pipe/softpipe/sp_prim_offset.c deleted file mode 100644 index 5fd6ac911a..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_offset.c +++ /dev/null @@ -1,157 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" -#include "macros.h" - -#include "sp_context.h" -#include "sp_prim.h" - - - -struct offset_stage { - struct prim_stage stage; - - GLuint hw_data_offset; - - GLfloat scale; - GLfloat units; -}; - - - -static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) -{ - return (struct offset_stage *)stage; -} - - -static void offset_begin( struct prim_stage *stage ) -{ - struct offset_stage *offset = offset_stage(stage); - - offset->units = stage->softpipe->setup.offset_units; - offset->scale = stage->softpipe->setup.offset_scale; - - stage->next->begin( stage->next ); -} - - -/* Offset tri. Some hardware can handle this, but not usually when - * doing unfilled rendering. - */ -static void do_offset_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct offset_stage *offset = offset_stage(stage); - GLfloat inv_det = 1.0 / header->det; - - /* Window coords: - */ - GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); - GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); - GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); - - GLfloat ex = v0[0] - v2[2]; - GLfloat fx = v1[0] - v2[2]; - GLfloat ey = v0[1] - v2[2]; - GLfloat fy = v1[1] - v2[2]; - GLfloat ez = v0[2] - v2[2]; - GLfloat fz = v1[2] - v2[2]; - - GLfloat a = ey*fz - ez*fy; - GLfloat b = ez*fx - ex*fz; - - GLfloat ac = a * inv_det; - GLfloat bc = b * inv_det; - GLfloat zoffset; - - if ( ac < 0.0f ) ac = -ac; - if ( bc < 0.0f ) bc = -bc; - - zoffset = offset->units + MAX2( ac, bc ) * offset->scale; - - v0[2] += zoffset; - v1[2] += zoffset; - v2[2] += zoffset; - - stage->next->tri( stage->next, header ); -} - - -static void offset_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = dup_vert(stage, header->v[2], 2); - - do_offset_tri( stage->next, &tmp ); -} - - - -static void offset_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void offset_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void offset_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_offset( struct softpipe_context *softpipe ) -{ - struct offset_stage *offset = CALLOC_STRUCT(offset_stage); - - prim_alloc_tmps( &offset->stage, 3 ); - - offset->stage.softpipe = softpipe; - offset->stage.next = NULL; - offset->stage.begin = offset_begin; - offset->stage.point = offset_point; - offset->stage.line = offset_line; - offset->stage.tri = offset_tri; - offset->stage.end = offset_end; - - return &offset->stage; -} diff --git a/src/mesa/pipe/softpipe/sp_prim_twoside.c b/src/mesa/pipe/softpipe/sp_prim_twoside.c deleted file mode 100644 index 5e9f218d1e..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_twoside.c +++ /dev/null @@ -1,152 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" -#include "vf/vf.h" - -#include "pipe/p_defines.h" -#include "sp_context.h" -#include "sp_prim.h" - - -struct twoside_stage { - struct prim_stage stage; - - GLfloat facing; - const GLuint *lookup; -}; - - -static INLINE struct twoside_stage *twoside_stage( struct prim_stage *stage ) -{ - return (struct twoside_stage *)stage; -} - - -static void twoside_begin( struct prim_stage *stage ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - - twoside->facing = (stage->softpipe->setup.front_winding == PIPE_WINDING_CW) ? -1 : 1; - - stage->next->begin( stage->next ); -} - - -static INLINE void copy_color( GLuint attr_dst, - GLuint attr_src, - struct vertex_header *v ) -{ - if (attr_dst && attr_src) { - memcpy( v->data[attr_dst], - v->data[attr_src], - sizeof(v->data[0]) ); - } -} - - -static struct vertex_header *copy_bfc( struct twoside_stage *twoside, - const struct vertex_header *v, - GLuint idx ) -{ - struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); - - copy_color( twoside->lookup[VF_ATTRIB_COLOR0], - twoside->lookup[VF_ATTRIB_BFC0], - tmp ); - - copy_color( twoside->lookup[VF_ATTRIB_COLOR1], - twoside->lookup[VF_ATTRIB_BFC1], - tmp ); - - return tmp; -} - - -/* Twoside tri: - */ -static void twoside_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - - if (header->det * twoside->facing < 0) { - struct prim_header tmp; - - tmp.det = header->det; - tmp.v[0] = copy_bfc(twoside, header->v[0], 0); - tmp.v[1] = copy_bfc(twoside, header->v[1], 1); - tmp.v[2] = copy_bfc(twoside, header->v[2], 2); - - stage->next->tri( stage->next, &tmp ); - } - else { - stage->next->tri( stage->next, header ); - } -} - - -static void twoside_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void twoside_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void twoside_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - - - -struct prim_stage *prim_twoside( struct softpipe_context *softpipe ) -{ - struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); - - prim_alloc_tmps( &twoside->stage, 3 ); - - twoside->stage.softpipe = softpipe; - twoside->stage.next = NULL; - twoside->stage.begin = twoside_begin; - twoside->stage.point = twoside_point; - twoside->stage.line = twoside_line; - twoside->stage.tri = twoside_tri; - twoside->stage.end = twoside_end; - - twoside->lookup = softpipe->vf_attr_to_slot; - - return &twoside->stage; -} diff --git a/src/mesa/pipe/softpipe/sp_prim_unfilled.c b/src/mesa/pipe/softpipe/sp_prim_unfilled.c deleted file mode 100644 index ab0dab09d4..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_unfilled.c +++ /dev/null @@ -1,165 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "sp_context.h" -#include "sp_prim.h" -#include "pipe/p_defines.h" - - -struct unfilled_stage { - struct prim_stage stage; - - GLuint mode[2]; -}; - - -static INLINE struct unfilled_stage *unfilled_stage( struct prim_stage *stage ) -{ - return (struct unfilled_stage *)stage; -} - - -static void unfilled_begin( struct prim_stage *stage ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - - unfilled->mode[0] = stage->softpipe->setup.fill_ccw; - unfilled->mode[1] = stage->softpipe->setup.fill_cw; - - stage->next->begin( stage->next ); -} - -static void point( struct prim_stage *stage, - struct vertex_header *v0 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - stage->next->point( stage->next, &tmp ); -} - -static void line( struct prim_stage *stage, - struct vertex_header *v0, - struct vertex_header *v1 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - tmp.v[1] = v1; - stage->next->line( stage->next, &tmp ); -} - - -static void points( struct prim_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - - if (v0->edgeflag) point( stage, v0 ); - if (v1->edgeflag) point( stage, v1 ); - if (v2->edgeflag) point( stage, v2 ); -} - -static void lines( struct prim_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - - if (v0->edgeflag) line( stage, v0, v1 ); - if (v1->edgeflag) line( stage, v1, v2 ); - if (v2->edgeflag) line( stage, v2, v0 ); -} - - -/* Unfilled tri: - * - * Note edgeflags in the vertex struct is not sufficient as we will - * need to manipulate them when decomposing primitives??? - */ -static void unfilled_tri( struct prim_stage *stage, - struct prim_header *header ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - GLuint mode = unfilled->mode[header->det < 0]; - - switch (mode) { - case PIPE_POLYGON_MODE_FILL: - stage->next->tri( stage->next, header ); - break; - case PIPE_POLYGON_MODE_LINE: - lines( stage, header ); - break; - case PIPE_POLYGON_MODE_POINT: - points( stage, header ); - break; - default: - abort(); - } -} - -static void unfilled_line( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void unfilled_point( struct prim_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void unfilled_end( struct prim_stage *stage ) -{ - stage->next->end( stage->next ); -} - -struct prim_stage *prim_unfilled( struct softpipe_context *softpipe ) -{ - struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); - - prim_alloc_tmps( &unfilled->stage, 0 ); - - unfilled->stage.softpipe = softpipe; - unfilled->stage.next = NULL; - unfilled->stage.tmp = NULL; - unfilled->stage.begin = unfilled_begin; - unfilled->stage.point = unfilled_point; - unfilled->stage.line = unfilled_line; - unfilled->stage.tri = unfilled_tri; - unfilled->stage.end = unfilled_end; - - return &unfilled->stage; -} -- cgit v1.2.3 From 9fbdf500788e78d63247a17226a75f7a079ae315 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 9 Jul 2007 16:12:13 -0600 Subject: New 'draw' module for handling the various stages of primitive drawing (clipping, culling, etc). --- src/mesa/pipe/draw/draw_context.c | 182 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 182 insertions(+) create mode 100644 src/mesa/pipe/draw/draw_context.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c new file mode 100644 index 0000000000..c91be23eef --- /dev/null +++ b/src/mesa/pipe/draw/draw_context.c @@ -0,0 +1,182 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "imports.h" +#include "macros.h" + +#include "draw_context.h" +#include "draw_private.h" + + +struct draw_context *draw_create( void ) +{ + struct draw_context *draw = CALLOC_STRUCT( draw_context ); + + /* create pipeline stages */ + draw->pipeline.unfilled = prim_unfilled( draw ); + draw->pipeline.twoside = prim_twoside( draw ); + draw->pipeline.offset = prim_offset( draw ); + draw->pipeline.clip = prim_clip( draw ); + draw->pipeline.flatshade = prim_flatshade( draw ); + draw->pipeline.cull = prim_cull( draw ); + + ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); + ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); + ASSIGN_4V( draw->plane[2], 0, -1, 0, 1 ); + ASSIGN_4V( draw->plane[3], 0, 1, 0, 1 ); + ASSIGN_4V( draw->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ + ASSIGN_4V( draw->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ + draw->nr_planes = 6; + + draw->vf = vf_create( GL_TRUE ); + + return draw; +} + + +void draw_destroy( struct draw_context *draw ) +{ + if (draw->header.storage) + ALIGN_FREE( draw->header.storage ); + + vf_destroy( draw->vf ); + + FREE( draw ); +} + + +/** + * Rebuild the rendering pipeline. + */ +static void validate_pipeline( struct draw_context *draw ) +{ + struct prim_stage *next = draw->pipeline.setup; + + /* TODO: make the current primitive part of the state and build + * shorter pipelines for lines & points. + */ + if (draw->setup.fill_cw != PIPE_POLYGON_MODE_FILL || + draw->setup.fill_ccw != PIPE_POLYGON_MODE_FILL) { + + draw->pipeline.unfilled->next = next; + next = draw->pipeline.unfilled; + } + + if (draw->setup.offset_cw || + draw->setup.offset_ccw) { + draw->pipeline.offset->next = next; + next = draw->pipeline.offset; + } + + if (draw->setup.light_twoside) { + draw->pipeline.twoside->next = next; + next = draw->pipeline.twoside; + } + + /* Always run the cull stage as we calculate determinant there + * also. Fix this.. + */ + { + draw->pipeline.cull->next = next; + next = draw->pipeline.cull; + } + + + /* Clip stage + */ + { + draw->pipeline.clip->next = next; + next = draw->pipeline.clip; + } + + /* Do software flatshading prior to clipping. XXX: should only do + * this for clipped primitives, ie it is a part of the clip + * routine. + */ + if (draw->setup.flatshade) { + draw->pipeline.flatshade->next = next; + next = draw->pipeline.flatshade; + } + + + draw->pipeline.first = next; +} + + +/** + * Register new primitive setup/rendering state. + * This causes the drawing pipeline to be rebuilt. + */ +void draw_set_setup_state( struct draw_context *draw, + const struct pipe_setup_state *setup ) +{ + memcpy( &draw->setup, setup, sizeof(*setup) ); + validate_pipeline( draw ); +} + + +/** + * Plug in the primitive rendering/rasterization stage. + * This is provided by the device driver. + */ +void draw_set_setup_stage( struct draw_context *draw, + struct prim_stage *stage ) +{ + draw->pipeline.setup = stage; +} + + +/** + * Set the draw module's clipping state. + */ +void draw_set_clip_state( struct draw_context *draw, + const struct pipe_clip_state *clip ) +{ + memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); + draw->nr_planes = 6 + clip->nr; +} + + +/** + * Set the draw module's viewport state. + */ +void draw_set_viewport_state( struct draw_context *draw, + const struct pipe_viewport_state *viewport ) +{ + draw->viewport = *viewport; /* struct copy */ + + vf_set_vp_scale_translate( draw->vf, viewport->scale, viewport->translate ); + + /* Using tnl/ and vf/ modules is temporary while getting started. + * Full pipe will have vertex shader, vertex fetch of its own. + */ +} -- cgit v1.2.3 From 279ffe3f163fd6a5e7bfa108db14c81acbb06ece Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 9 Jul 2007 16:14:26 -0600 Subject: New 'draw' module for primitive drawing (clipping, culling, etc). --- src/mesa/pipe/draw/draw_clip.c | 35 +++++--- src/mesa/pipe/draw/draw_context.h | 32 +++++-- src/mesa/pipe/draw/draw_cull.c | 21 +++-- src/mesa/pipe/draw/draw_flatshade.c | 31 +++---- src/mesa/pipe/draw/draw_offset.c | 24 +++--- src/mesa/pipe/draw/draw_private.h | 138 +++++++++++++++++++++++++----- src/mesa/pipe/draw/draw_twoside.c | 24 ++++-- src/mesa/pipe/draw/draw_unfilled.c | 17 ++-- src/mesa/pipe/draw/draw_vb.c | 92 ++++++++------------ src/mesa/pipe/p_state.h | 2 + src/mesa/pipe/softpipe/sp_context.c | 74 +++------------- src/mesa/pipe/softpipe/sp_context.h | 27 ------ src/mesa/pipe/softpipe/sp_headers.h | 21 ----- src/mesa/pipe/softpipe/sp_prim_setup.c | 54 +++++++----- src/mesa/pipe/softpipe/sp_prim_setup.h | 6 ++ src/mesa/pipe/softpipe/sp_state_blend.c | 2 - src/mesa/pipe/softpipe/sp_state_clip.c | 32 +++++-- src/mesa/pipe/softpipe/sp_state_derived.c | 16 ++-- src/mesa/pipe/softpipe/sp_state_sampler.c | 2 +- src/mesa/pipe/softpipe/sp_state_setup.c | 30 ++----- src/mesa/sources | 22 +++-- 21 files changed, 368 insertions(+), 334 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 304c43c3f4..10a6c1b823 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -28,18 +28,18 @@ /* Authors: Keith Whitwell */ -#include "imports.h" -#include "macros.h" +#include "main/macros.h" +#include "draw_private.h" -#include "sp_context.h" -#include "sp_prim.h" struct clipper { - struct prim_stage stage; + struct prim_stage stage; /**< base class */ GLuint active_user_planes; + GLfloat (*plane)[4]; }; + /* This is a bit confusing: */ static INLINE struct clipper *clipper_stage( struct prim_stage *stage ) @@ -75,7 +75,7 @@ static void interp( struct clipper *clip, const struct vertex_header *out, const struct vertex_header *in ) { - const GLuint nr_attrs = clip->stage.softpipe->nr_attrs; + const GLuint nr_attrs = clip->stage.draw->nr_attrs; GLuint j; /* Vertex header. @@ -96,8 +96,8 @@ static void interp( struct clipper *clip, */ { const GLfloat *pos = dst->clip; - const GLfloat *scale = clip->stage.softpipe->viewport.scale; - const GLfloat *trans = clip->stage.softpipe->viewport.translate; + const GLfloat *scale = clip->stage.draw->viewport.scale; + const GLfloat *trans = clip->stage.draw->viewport.translate; GLfloat oow; oow = 1.0 / pos[3]; @@ -225,7 +225,7 @@ do_clip_tri( struct prim_stage *stage, while (clipmask && n >= 3) { GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->stage.softpipe->plane[plane_idx]; + const GLfloat *plane = clipper->plane[plane_idx]; struct vertex_header *vert_prev = inlist[0]; GLfloat dp_prev = dot4( vert_prev->clip, plane ); GLuint outcount = 0; @@ -314,7 +314,7 @@ do_clip_line( struct prim_stage *stage, while (clipmask) { GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->stage.softpipe->plane[plane_idx]; + const GLfloat *plane = clipper->plane[plane_idx]; clipmask &= ~(1<softpipe->nr_planes; + GLuint nr = stage->draw->nr_planes; /* Hacky bitmask to use when we hit CLIP_USER_BIT: */ @@ -379,6 +379,7 @@ clip_line( struct prim_stage *stage, header->v[1]->clipmask); if (clipmask == 0) { + /* no clipping needed */ stage->next->line( stage->next, header ); } else if ((header->v[0]->clipmask & @@ -397,6 +398,7 @@ clip_tri( struct prim_stage *stage, header->v[2]->clipmask); if (clipmask == 0) { + /* no clipping needed */ stage->next->tri( stage->next, header ); } else if ((header->v[0]->clipmask & @@ -406,24 +408,31 @@ clip_tri( struct prim_stage *stage, } } + static void clip_end( struct prim_stage *stage ) { stage->next->end( stage->next ); } -struct prim_stage *prim_clip( struct softpipe_context *softpipe ) +/** + * Allocate a new clipper stage. + * \return pointer to new stage object + */ +struct prim_stage *prim_clip( struct draw_context *draw ) { struct clipper *clipper = CALLOC_STRUCT(clipper); prim_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); - clipper->stage.softpipe = softpipe; + clipper->stage.draw = draw; clipper->stage.begin = clip_begin; clipper->stage.point = clip_point; clipper->stage.line = clip_line; clipper->stage.tri = clip_tri; clipper->stage.end = clip_end; + clipper->plane = draw->plane; + return &clipper->stage; } diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index a138f812fa..85f2ace75f 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -26,25 +26,42 @@ * **************************************************************************/ +/** + * \brief Public interface into the drawing module. + */ + /* Authors: Keith Whitwell */ -#ifndef G_DRAW_H -#define G_DRAW_H + +#ifndef DRAW_CONTEXT_H +#define DRAW_CONTEXT_H + #include "glheader.h" #include "pipe/p_state.h" +struct vertex_buffer; struct draw_context; +struct prim_stage; + -struct draw_context *draw_create( struct softpipe_context *softpipe ); +struct draw_context *draw_create( void ); void draw_destroy( struct draw_context *draw ); -void draw_set_viewport( struct draw_context *draw, - const GLfloat *scale, - const GLfloat *translate ); +void draw_set_viewport_state( struct draw_context *draw, + const struct pipe_viewport_state *viewport ); + +void draw_set_clip_state( struct draw_context *pipe, + const struct pipe_clip_state *clip ); + +void draw_set_setup_state( struct draw_context *draw, + const struct pipe_setup_state *setup ); + +void draw_set_setup_stage( struct draw_context *draw, + struct prim_stage *stage ); void draw_set_vertex_attributes( struct draw_context *draw, const GLuint *attrs, @@ -53,4 +70,5 @@ void draw_set_vertex_attributes( struct draw_context *draw, void draw_vb(struct draw_context *draw, struct vertex_buffer *VB ); -#endif + +#endif /* DRAW_CONTEXT_H */ diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index 63099fbee0..d27d33a40d 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -27,11 +27,10 @@ /* Authors: Keith Whitwell */ -#include "imports.h" +#include "main/imports.h" #include "pipe/p_defines.h" -#include "sp_context.h" -#include "sp_prim.h" +#include "draw_private.h" @@ -52,7 +51,7 @@ static void cull_begin( struct prim_stage *stage ) { struct cull_stage *cull = cull_stage(stage); - cull->mode = stage->softpipe->setup.cull_mode; + cull->mode = stage->draw->setup.cull_mode; stage->next->begin( stage->next ); } @@ -76,10 +75,13 @@ static void cull_tri( struct prim_stage *stage, _mesa_printf("%s %f\n", __FUNCTION__, header->det ); if (header->det != 0) { + /* non-zero area */ GLuint mode = (header->det < 0) ? PIPE_WINDING_CW : PIPE_WINDING_CCW; - if ((mode & cull_stage(stage)->mode) == 0) + if ((mode & cull_stage(stage)->mode) == 0) { + /* triangle is not culled, pass to next stage */ stage->next->tri( stage->next, header ); + } } } @@ -97,18 +99,23 @@ static void cull_point( struct prim_stage *stage, stage->next->point( stage->next, header ); } + static void cull_end( struct prim_stage *stage ) { stage->next->end( stage->next ); } -struct prim_stage *prim_cull( struct softpipe_context *softpipe ) + +/** + * Create a new polygon culling stage. + */ +struct prim_stage *prim_cull( struct draw_context *draw ) { struct cull_stage *cull = CALLOC_STRUCT(cull_stage); prim_alloc_tmps( &cull->stage, 0 ); - cull->stage.softpipe = softpipe; + cull->stage.draw = draw; cull->stage.next = NULL; cull->stage.begin = cull_begin; cull->stage.point = cull_point; diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 3a7d9de466..004b5bdc96 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -27,12 +27,9 @@ /* Authors: Keith Whitwell */ -#include "imports.h" -#include "vf/vf.h" - -#include "sp_context.h" -#include "sp_prim.h" +#include "main/imports.h" +#include "draw_private.h" struct flatshade_stage { @@ -67,11 +64,12 @@ static INLINE void copy_attr( GLuint attr, } } -static void copy_colors( struct prim_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) + +static INLINE void copy_colors( struct prim_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) { - struct flatshade_stage *flatshade = flatshade_stage(stage); + const struct flatshade_stage *flatshade = flatshade_stage(stage); const GLuint *lookup = flatshade->lookup; copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); @@ -81,8 +79,8 @@ static void copy_colors( struct prim_stage *stage, } - -/* Flatshade tri. Required for clipping and when unfilled tris are +/** + * Flatshade tri. Required for clipping and when unfilled tris are * active, otherwise handled by hardware. */ static void flatshade_tri( struct prim_stage *stage, @@ -102,7 +100,8 @@ static void flatshade_tri( struct prim_stage *stage, } -/* Flatshade line. Required for clipping. +/** + * Flatshade line. Required for clipping. */ static void flatshade_line( struct prim_stage *stage, struct prim_header *header ) @@ -124,18 +123,20 @@ static void flatshade_point( struct prim_stage *stage, stage->next->point( stage->next, header ); } + static void flatshade_end( struct prim_stage *stage ) { stage->next->end( stage->next ); } -struct prim_stage *prim_flatshade( struct softpipe_context *softpipe ) + +struct prim_stage *prim_flatshade( struct draw_context *draw ) { struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); prim_alloc_tmps( &flatshade->stage, 2 ); - flatshade->stage.softpipe = softpipe; + flatshade->stage.draw = draw; flatshade->stage.next = NULL; flatshade->stage.begin = flatshade_begin; flatshade->stage.point = flatshade_point; @@ -143,7 +144,7 @@ struct prim_stage *prim_flatshade( struct softpipe_context *softpipe ) flatshade->stage.tri = flatshade_tri; flatshade->stage.end = flatshade_end; - flatshade->lookup = softpipe->vf_attr_to_slot; + flatshade->lookup = draw->vf_attr_to_slot; return &flatshade->stage; } diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 5fd6ac911a..0fa8cf29d3 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -27,11 +27,10 @@ /* Authors: Keith Whitwell */ -#include "imports.h" -#include "macros.h" -#include "sp_context.h" -#include "sp_prim.h" +#include "main/imports.h" +#include "main/macros.h" +#include "draw_private.h" @@ -56,14 +55,15 @@ static void offset_begin( struct prim_stage *stage ) { struct offset_stage *offset = offset_stage(stage); - offset->units = stage->softpipe->setup.offset_units; - offset->scale = stage->softpipe->setup.offset_scale; + offset->units = stage->draw->setup.offset_units; + offset->scale = stage->draw->setup.offset_scale; stage->next->begin( stage->next ); } -/* Offset tri. Some hardware can handle this, but not usually when +/** + * Offset tri Z. Some hardware can handle this, but not usually when * doing unfilled rendering. */ static void do_offset_tri( struct prim_stage *stage, @@ -92,8 +92,8 @@ static void do_offset_tri( struct prim_stage *stage, GLfloat bc = b * inv_det; GLfloat zoffset; - if ( ac < 0.0f ) ac = -ac; - if ( bc < 0.0f ) bc = -bc; + ac = FABSF(ac); + bc = FABSF(bc); zoffset = offset->units + MAX2( ac, bc ) * offset->scale; @@ -115,7 +115,7 @@ static void offset_tri( struct prim_stage *stage, tmp.v[1] = dup_vert(stage, header->v[1], 1); tmp.v[2] = dup_vert(stage, header->v[2], 2); - do_offset_tri( stage->next, &tmp ); + do_offset_tri( stage, &tmp ); } @@ -139,13 +139,13 @@ static void offset_end( struct prim_stage *stage ) stage->next->end( stage->next ); } -struct prim_stage *prim_offset( struct softpipe_context *softpipe ) +struct prim_stage *prim_offset( struct draw_context *draw ) { struct offset_stage *offset = CALLOC_STRUCT(offset_stage); prim_alloc_tmps( &offset->stage, 3 ); - offset->stage.softpipe = softpipe; + offset->stage.draw = draw; offset->stage.next = NULL; offset->stage.begin = offset_begin; offset->stage.point = offset_point; diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index b6cbaae085..ac25628a08 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -25,32 +25,63 @@ * **************************************************************************/ -/* Authors: Keith Whitwell +/** + * Private data structures, etc for the draw module. */ -#ifndef G_PRIM_H -#define G_PRIM_H -#include "glheader.h" -#include "sp_headers.h" +/** + * Authors: + * Keith Whitwell + * Brian Paul + */ + -struct softpipe_context; +#ifndef DRAW_PRIVATE_H +#define DRAW_PRIVATE_H -struct prim_stage *prim_setup( struct softpipe_context *context ); -struct prim_stage *prim_unfilled( struct softpipe_context *context ); -struct prim_stage *prim_twoside( struct softpipe_context *context ); -struct prim_stage *prim_offset( struct softpipe_context *context ); -struct prim_stage *prim_clip( struct softpipe_context *context ); -struct prim_stage *prim_flatshade( struct softpipe_context *context ); -struct prim_stage *prim_cull( struct softpipe_context *context ); +#include "main/glheader.h" +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "vf/vf.h" -/* Internal structs and helpers for the primitive clip/setup pipeline: + +/** + * Basic vertex info. + * Carry some useful information around with the vertices in the prim pipe. */ -struct prim_stage { - struct softpipe_context *softpipe; +struct vertex_header { + GLuint clipmask:12; + GLuint edgeflag:1; + GLuint pad:19; - struct prim_stage *next; + GLfloat clip[4]; + + GLfloat data[][4]; /* Note variable size */ +}; + + +/** + * Basic info for a point/line/triangle primitive. + */ +struct prim_header { + GLfloat det; /**< front/back face determinant */ + struct vertex_header *v[3]; /**< 1 to 3 vertex pointers */ +}; + + + +struct draw_context; + +/** + * Base class for all primitive drawing stages. + */ +struct prim_stage +{ + struct draw_context *draw; /**< parent context */ + + struct prim_stage *next; /**< next stage in pipeline */ struct vertex_header **tmp; GLuint nr_tmps; @@ -70,8 +101,72 @@ struct prim_stage { }; +/** + * Private context for the drawing module. + */ +struct draw_context +{ + struct { + struct prim_stage *first; /**< one of the following */ + + /* stages (in logical order) */ + struct prim_stage *flatshade; + struct prim_stage *clip; + struct prim_stage *cull; + struct prim_stage *twoside; + struct prim_stage *offset; + struct prim_stage *unfilled; + struct prim_stage *setup; /* aka render/rasterize */ + } pipeline; + + /* pipe state that we need: */ + struct pipe_setup_state setup; + struct pipe_viewport_state viewport; + + /* Clip derived state: + */ + GLfloat plane[12][4]; + GLuint nr_planes; + + GLuint vf_attr_to_slot[PIPE_ATTRIB_MAX]; + + struct vf_attr_map attrs[VF_ATTRIB_MAX]; + GLuint nr_attrs; + GLuint vertex_size; /**< in bytes */ + struct vertex_fetch *vf; + + GLubyte *verts; + GLuint nr_vertices; + GLboolean in_vb; + + GLenum prim; /**< GL_POINTS, GL_LINE_STRIP, GL_QUADS, etc */ + + /* Helper for tnl: + */ + GLvector4f header; +}; + + -/* Get a writeable copy of a vertex: +extern struct prim_stage *prim_unfilled( struct draw_context *context ); +extern struct prim_stage *prim_twoside( struct draw_context *context ); +extern struct prim_stage *prim_offset( struct draw_context *context ); +extern struct prim_stage *prim_clip( struct draw_context *context ); +extern struct prim_stage *prim_flatshade( struct draw_context *context ); +extern struct prim_stage *prim_cull( struct draw_context *context ); + + +extern void prim_free_tmps( struct prim_stage *stage ); +extern void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); + + + +/** + * Get a writeable copy of a vertex. + * \param stage drawing stage info + * \param vert the vertex to copy (source) + * \param idx index into stage's tmp[] array to put the copy (dest) + * \return pointer to the copied vertex */ static INLINE struct vertex_header * dup_vert( struct prim_stage *stage, @@ -79,12 +174,9 @@ dup_vert( struct prim_stage *stage, GLuint idx ) { struct vertex_header *tmp = stage->tmp[idx]; - memcpy(tmp, vert, stage->softpipe->prim.vertex_size ); + memcpy(tmp, vert, stage->draw->vertex_size ); return tmp; } -void prim_free_tmps( struct prim_stage *stage ); -void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); - -#endif +#endif /* DRAW_PRIVATE_H */ diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 5e9f218d1e..88e164ec5e 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -27,12 +27,10 @@ /* Authors: Keith Whitwell */ -#include "imports.h" -#include "vf/vf.h" +#include "main/imports.h" #include "pipe/p_defines.h" -#include "sp_context.h" -#include "sp_prim.h" +#include "draw_private.h" struct twoside_stage { @@ -53,7 +51,7 @@ static void twoside_begin( struct prim_stage *stage ) { struct twoside_stage *twoside = twoside_stage(stage); - twoside->facing = (stage->softpipe->setup.front_winding == PIPE_WINDING_CW) ? -1 : 1; + twoside->facing = (stage->draw->setup.front_winding == PIPE_WINDING_CW) ? -1 : 1; stage->next->begin( stage->next ); } @@ -97,9 +95,11 @@ static void twoside_tri( struct prim_stage *stage, struct twoside_stage *twoside = twoside_stage(stage); if (header->det * twoside->facing < 0) { + /* this is a back-facing triangle */ struct prim_header tmp; tmp.det = header->det; + /* copy back colors to front color slots */ tmp.v[0] = copy_bfc(twoside, header->v[0], 0); tmp.v[1] = copy_bfc(twoside, header->v[1], 1); tmp.v[2] = copy_bfc(twoside, header->v[2], 2); @@ -115,6 +115,7 @@ static void twoside_tri( struct prim_stage *stage, static void twoside_line( struct prim_stage *stage, struct prim_header *header ) { + /* pass-through */ stage->next->line( stage->next, header ); } @@ -122,23 +123,28 @@ static void twoside_line( struct prim_stage *stage, static void twoside_point( struct prim_stage *stage, struct prim_header *header ) { + /* pass-through */ stage->next->point( stage->next, header ); } + static void twoside_end( struct prim_stage *stage ) { + /* pass-through */ stage->next->end( stage->next ); } - -struct prim_stage *prim_twoside( struct softpipe_context *softpipe ) +/** + * Create twoside pipeline stage. + */ +struct prim_stage *prim_twoside( struct draw_context *draw ) { struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); prim_alloc_tmps( &twoside->stage, 3 ); - twoside->stage.softpipe = softpipe; + twoside->stage.draw = draw; twoside->stage.next = NULL; twoside->stage.begin = twoside_begin; twoside->stage.point = twoside_point; @@ -146,7 +152,7 @@ struct prim_stage *prim_twoside( struct softpipe_context *softpipe ) twoside->stage.tri = twoside_tri; twoside->stage.end = twoside_end; - twoside->lookup = softpipe->vf_attr_to_slot; + twoside->lookup = draw->vf_attr_to_slot; return &twoside->stage; } diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index ab0dab09d4..a1d9d14352 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -27,11 +27,10 @@ /* Authors: Keith Whitwell */ -#include "imports.h" -#include "sp_context.h" -#include "sp_prim.h" +#include "main/imports.h" #include "pipe/p_defines.h" +#include "draw_private.h" struct unfilled_stage { @@ -51,8 +50,8 @@ static void unfilled_begin( struct prim_stage *stage ) { struct unfilled_stage *unfilled = unfilled_stage(stage); - unfilled->mode[0] = stage->softpipe->setup.fill_ccw; - unfilled->mode[1] = stage->softpipe->setup.fill_cw; + unfilled->mode[0] = stage->draw->setup.fill_ccw; + unfilled->mode[1] = stage->draw->setup.fill_cw; stage->next->begin( stage->next ); } @@ -128,14 +127,14 @@ static void unfilled_tri( struct prim_stage *stage, } static void unfilled_line( struct prim_stage *stage, - struct prim_header *header ) + struct prim_header *header ) { stage->next->line( stage->next, header ); } static void unfilled_point( struct prim_stage *stage, - struct prim_header *header ) + struct prim_header *header ) { stage->next->point( stage->next, header ); } @@ -146,13 +145,13 @@ static void unfilled_end( struct prim_stage *stage ) stage->next->end( stage->next ); } -struct prim_stage *prim_unfilled( struct softpipe_context *softpipe ) +struct prim_stage *prim_unfilled( struct draw_context *draw ) { struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); prim_alloc_tmps( &unfilled->stage, 0 ); - unfilled->stage.softpipe = softpipe; + unfilled->stage.draw = draw; unfilled->stage.next = NULL; unfilled->stage.tmp = NULL; unfilled->stage.begin = unfilled_begin; diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index 3fc30dd203..66573a93fe 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -31,41 +31,22 @@ */ #include "imports.h" +#include "macros.h" #include "tnl/t_context.h" #include "vf/vf.h" -#include "sp_context.h" -#include "sp_prim.h" -#include "sp_headers.h" -#include "sp_draw.h" +#include "pipe/softpipe/sp_context.h" +#include "pipe/softpipe/sp_headers.h" +#include "draw_private.h" +#include "draw_context.h" + /* This file is a temporary set of hooks to allow us to use the tnl/ * and vf/ modules until we have replacements in pipe. */ -struct draw_context -{ - struct softpipe_context *softpipe; - - struct vf_attr_map attrs[VF_ATTRIB_MAX]; - GLuint nr_attrs; - GLuint vertex_size; - struct vertex_fetch *vf; - - GLubyte *verts; - GLuint nr_vertices; - GLboolean in_vb; - - GLenum prim; - - /* Helper for tnl: - */ - GLvector4f header; -}; - - static struct vertex_header *get_vertex( struct draw_context *pipe, GLuint i ) { @@ -80,7 +61,7 @@ static void draw_allocate_vertices( struct draw_context *draw, draw->nr_vertices = nr_vertices; draw->verts = MALLOC( nr_vertices * draw->vertex_size ); - draw->softpipe->prim.first->begin( draw->softpipe->prim.first ); + draw->pipeline.first->begin( draw->pipeline.first ); } static void draw_set_prim( struct draw_context *draw, @@ -149,7 +130,7 @@ static void draw_indexed_prim( struct draw_context *draw, const GLuint *elts, GLuint count ) { - struct prim_stage * const first = draw->softpipe->prim.first; + struct prim_stage * const first = draw->pipeline.first; struct prim_header prim; GLuint i; @@ -299,7 +280,7 @@ static void draw_prim( struct draw_context *draw, GLuint start, GLuint count ) { - struct prim_stage * const first = draw->softpipe->prim.first; + struct prim_stage * const first = draw->pipeline.first; struct prim_header prim; GLuint i; @@ -442,7 +423,7 @@ static void draw_prim( struct draw_context *draw, static void draw_release_vertices( struct draw_context *draw ) { - draw->softpipe->prim.first->end( draw->softpipe->prim.first ); + draw->pipeline.first->end( draw->pipeline.first ); FREE(draw->verts); draw->verts = NULL; @@ -636,35 +617,6 @@ void draw_vb(struct draw_context *draw, draw->in_vb = 0; } -void draw_set_viewport( struct draw_context *draw, - const GLfloat *scale, - const GLfloat *translate ) -{ - assert(!draw->in_vb); - vf_set_vp_scale_translate( draw->vf, scale, translate ); -} - - - -struct draw_context *draw_create( struct softpipe_context *softpipe ) -{ - struct draw_context *draw = CALLOC_STRUCT( draw_context ); - draw->softpipe = softpipe; - draw->vf = vf_create( GL_TRUE ); - - return draw; -} - - -void draw_destroy( struct draw_context *draw ) -{ - if (draw->header.storage) - ALIGN_FREE( draw->header.storage ); - - vf_destroy( draw->vf ); - - FREE( draw ); -} #define EMIT_ATTR( ATTR, STYLE ) \ do { \ @@ -695,3 +647,27 @@ void draw_set_vertex_attributes( struct draw_context *draw, } +#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) + +void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ) +{ + stage->nr_tmps = nr; + + if (nr) { + GLubyte *store = MALLOC(MAX_VERTEX_SIZE * nr); + GLuint i; + + stage->tmp = MALLOC(sizeof(struct vertex_header *) * nr); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } +} + +void prim_free_tmps( struct prim_stage *stage ) +{ + if (stage->tmp) { + FREE(stage->tmp[0]); + FREE(stage->tmp); + } +} diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 9b4b336654..581ea5ddd8 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -48,6 +48,8 @@ #define PIPE_MAX_SAMPLERS 8 #define PIPE_MAX_CLIP_PLANES 6 #define PIPE_MAX_CONSTANT 32 +#define PIPE_ATTRIB_MAX 32 + /* fwd decl */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 6bd1d9f16e..9becab4918 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -29,14 +29,14 @@ * Keith Whitwell */ -#include "imports.h" -#include "macros.h" - +#include "main/imports.h" +#include "main/macros.h" +#include "pipe/draw/draw_context.h" #include "sp_context.h" #include "sp_clear.h" -#include "sp_prim.h" #include "sp_state.h" -#include "sp_draw.h" +#include "sp_prim_setup.h" + static void softpipe_destroy( struct pipe_context *pipe ) { @@ -44,7 +44,7 @@ static void softpipe_destroy( struct pipe_context *pipe ) draw_destroy( softpipe->draw ); - FREE( softpipe ); + free( softpipe ); } @@ -59,6 +59,7 @@ static void softpipe_draw_vb( struct pipe_context *pipe, draw_vb( softpipe->draw, VB ); } + struct pipe_context *softpipe_create( void ) { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); @@ -81,64 +82,17 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.clear = softpipe_clear; - - softpipe->prim.setup = prim_setup( softpipe ); - softpipe->prim.unfilled = prim_unfilled( softpipe ); - softpipe->prim.twoside = prim_twoside( softpipe ); - softpipe->prim.offset = prim_offset( softpipe ); - softpipe->prim.clip = prim_clip( softpipe ); - softpipe->prim.flatshade = prim_flatshade( softpipe ); - softpipe->prim.cull = prim_cull( softpipe ); - + softpipe->quad.shade = sp_quad_shade_stage(softpipe); + softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); softpipe->quad.blend = sp_quad_blend_stage(softpipe); softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); - softpipe->quad.shade = sp_quad_shade_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); - softpipe->draw = draw_create( softpipe ); - - ASSIGN_4V( softpipe->plane[0], -1, 0, 0, 1 ); - ASSIGN_4V( softpipe->plane[1], 1, 0, 0, 1 ); - ASSIGN_4V( softpipe->plane[2], 0, -1, 0, 1 ); - ASSIGN_4V( softpipe->plane[3], 0, 1, 0, 1 ); - ASSIGN_4V( softpipe->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ - ASSIGN_4V( softpipe->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ - softpipe->nr_planes = 6; + /* + * Create drawing context and plug our render/setup stage into it. + */ + softpipe->draw = draw_create(); + draw_set_setup_stage(softpipe->draw, prim_setup(softpipe)); return &softpipe->pipe; } - - - - - - -#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) - -void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ) -{ - stage->nr_tmps = nr; - - if (nr) { - GLubyte *store = MALLOC(MAX_VERTEX_SIZE * nr); - GLuint i; - - stage->tmp = MALLOC(sizeof(struct vertex_header *) * nr); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); - } -} - -void prim_free_tmps( struct prim_stage *stage ) -{ - if (stage->tmp) { - FREE(stage->tmp[0]); - FREE(stage->tmp); - } -} - - - - - diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 40688208e0..0c082978bd 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -55,7 +55,6 @@ enum interp_mode { #define G_NEW_SETUP 0x2 #define G_NEW_FS 0x4 #define G_NEW_BLEND 0x8 -#define G_NEW_POINT 0x10 #define G_NEW_CLIP 0x20 #define G_NEW_SCISSOR 0x40 #define G_NEW_STIPPLE 0x80 @@ -66,12 +65,9 @@ enum interp_mode { #define G_NEW_TEXTURE 0x1000 -#define PIPE_ATTRIB_MAX 32 - struct softpipe_context { struct pipe_context pipe; - /* The most recent drawing state as set by the driver: */ struct pipe_alpha_test_state alpha_test; @@ -90,11 +86,6 @@ struct softpipe_context { struct pipe_viewport_state viewport; GLuint dirty; - /* Clip derived state: - */ - GLfloat plane[12][4]; - GLuint nr_planes; - /* Setup derived state. TODO: this should be passed in the program * tokens as parameters to DECL instructions. * @@ -119,24 +110,6 @@ struct softpipe_context { */ GLubyte stipple_masks[16][16]; - - /* The software clipper/setup engine. - */ - struct { - struct prim_stage *setup; - struct prim_stage *unfilled; - struct prim_stage *twoside; - struct prim_stage *clip; - struct prim_stage *flatshade; - struct prim_stage *offset; - struct prim_stage *cull; - - struct prim_stage *first; - - GLenum prim; - GLuint vertex_size; - } prim; - /* * Software quad rendering pipeline */ diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index 96ff52a453..561597d9e2 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -34,27 +34,6 @@ #define PRIM_LINE 2 #define PRIM_TRI 3 -struct prim_header { - GLfloat det; - struct vertex_header *v[3]; -}; - -/* Carry some useful information around with the vertices in the prim - * pipe. - */ -struct vertex_header { - GLuint clipmask:12; - GLuint edgeflag:1; - GLuint pad:19; - - GLfloat clip[4]; - - GLfloat data[][4]; /* Note variable size */ -}; - - - - /* The rasterizer generates 2x2 quads of fragment and feeds them to * the current fp_machine (see below). diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 3f4602feb0..ea10ef53a0 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -32,8 +32,10 @@ #include "macros.h" #include "sp_context.h" -#include "sp_prim.h" +#include "sp_headers.h" +#include "pipe/draw/draw_private.h" #include "sp_quad.h" +#include "sp_prim_setup.h" @@ -66,7 +68,10 @@ struct edge { * Also used for line drawing (taking some liberties). */ struct setup_stage { - struct prim_stage stage; /**< This must be first */ + struct prim_stage stage; /**< This must be first (base class) */ + + /*XXX NEW */ + struct softpipe_context *softpipe; /* Vertices are just an array of floats making up each attribute in * turn. Currently fixed at 4 floats, but should change in time. @@ -119,7 +124,9 @@ static inline GLint block( GLint x ) static void setup_begin( struct prim_stage *stage ) { - setup_stage(stage)->quad.nr_attrs = stage->softpipe->nr_frag_attrs; + struct setup_stage *setup = setup_stage(stage); + + setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; } @@ -133,7 +140,7 @@ static void run_shader_block( struct setup_stage *setup, setup->quad.y0 = y; setup->quad.mask = mask; - quad_emit(setup->stage.softpipe, &setup->quad); + quad_emit(setup->/*stage.*/softpipe, &setup->quad); } @@ -387,7 +394,7 @@ static void tri_persp_coeff( struct setup_stage *setup, */ static void setup_tri_coefficients( struct setup_stage *setup ) { - const enum interp_mode *interp = setup->stage.softpipe->interp; + const enum interp_mode *interp = setup->/*stage.*/softpipe->interp; GLuint slot, j; /* z and w are done by linear interpolation: @@ -462,15 +469,15 @@ static void subtriangle( struct setup_stage *setup, /* scissor y: */ - if (setup->stage.softpipe->setup.scissor) { + if (setup->/*stage.*/softpipe->setup.scissor) { start_y = sy; finish_y = start_y + lines; - if (start_y < setup->stage.softpipe->scissor.miny) - start_y = setup->stage.softpipe->scissor.miny; + if (start_y < setup->/*stage.*/softpipe->scissor.miny) + start_y = setup->/*stage.*/softpipe->scissor.miny; - if (finish_y > setup->stage.softpipe->scissor.maxy) - finish_y = setup->stage.softpipe->scissor.maxy; + if (finish_y > setup->/*stage.*/softpipe->scissor.maxy) + finish_y = setup->/*stage.*/softpipe->scissor.maxy; start_y -= sy; finish_y -= sy; @@ -495,12 +502,12 @@ static void subtriangle( struct setup_stage *setup, /* scissor x: */ - if (setup->stage.softpipe->setup.scissor) { - if (left < setup->stage.softpipe->scissor.minx) - left = setup->stage.softpipe->scissor.minx; + if (setup->/*stage.*/softpipe->setup.scissor) { + if (left < setup->/*stage.*/softpipe->scissor.minx) + left = setup->/*stage.*/softpipe->scissor.minx; - if (right > setup->stage.softpipe->scissor.maxx) - right = setup->stage.softpipe->scissor.maxx; + if (right > setup->/*stage.*/softpipe->scissor.maxx) + right = setup->/*stage.*/softpipe->scissor.maxx; } if (left < right) { @@ -604,7 +611,7 @@ line_persp_coeff(struct setup_stage *setup, GLuint slot, GLuint i) static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { - const enum interp_mode *interp = setup->stage.softpipe->interp; + const enum interp_mode *interp = setup->/*stage.*/softpipe->interp; GLuint slot, j; /* use setup->vmin, vmax to point to vertices */ @@ -664,7 +671,7 @@ plot(struct setup_stage *setup, GLint x, GLint y) /* flush prev quad, start new quad */ if (setup->quad.x0 != -1) - quad_emit(setup->stage.softpipe, &setup->quad); + quad_emit(setup->/*stage.*/softpipe, &setup->quad); setup->quad.x0 = quadX; setup->quad.y0 = quadY; @@ -767,7 +774,7 @@ setup_line(struct prim_stage *stage, struct prim_header *prim) /* draw final quad */ if (setup->quad.mask) { - quad_emit(setup->stage.softpipe, &setup->quad); + quad_emit(setup->/*stage.*/softpipe, &setup->quad); } } @@ -782,8 +789,8 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); /*XXX this should be a vertex attrib! */ - GLfloat halfSize = 0.5 * setup->stage.softpipe->setup.point_size; - GLboolean round = setup->stage.softpipe->setup.point_smooth; + GLfloat halfSize = 0.5 * setup->/*stage.*/softpipe->setup.point_size; + GLboolean round = setup->/*stage.*/softpipe->setup.point_smooth; const struct vertex_header *v0 = prim->v[0]; const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; @@ -822,7 +829,7 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) setup->quad.x0 = x - ix; setup->quad.y0 = y - iy; setup->quad.mask = (1 << ix) << (2 * iy); - quad_emit(setup->stage.softpipe, &setup->quad); + quad_emit(setup->/*stage.*/softpipe, &setup->quad); } else { const GLint ixmin = block((GLint) (x - halfSize)); @@ -882,7 +889,7 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) if (setup->quad.mask) { setup->quad.x0 = ix; setup->quad.y0 = iy; - quad_emit( setup->stage.softpipe, &setup->quad ); + quad_emit( setup->/*stage.*/softpipe, &setup->quad ); } } } @@ -900,7 +907,8 @@ struct prim_stage *prim_setup( struct softpipe_context *softpipe ) { struct setup_stage *setup = CALLOC_STRUCT(setup_stage); - setup->stage.softpipe = softpipe; + setup->softpipe = softpipe; + setup->stage.draw = softpipe->draw; setup->stage.begin = setup_begin; setup->stage.point = setup_point; setup->stage.line = setup_line; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h index 40a70c543e..4d26c05458 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.h +++ b/src/mesa/pipe/softpipe/sp_prim_setup.h @@ -37,11 +37,16 @@ #include "glheader.h" #include "imports.h" +#if 0 #include "s_tri_public.h" +#endif #include "s_context.h" +extern struct prim_stage *prim_setup( struct softpipe_context *softpipe ); + +#if 0 /* UNUSED? */ struct tri_context; struct fp_context; struct be_context; @@ -117,5 +122,6 @@ void tri_rasterize_spans( struct tri_context *tri ); +#endif #endif #endif diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 6ca42e8a01..e0ff549d70 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -31,8 +31,6 @@ #include "sp_context.h" #include "sp_state.h" -#include "sp_draw.h" - void softpipe_set_blend_state( struct pipe_context *pipe, diff --git a/src/mesa/pipe/softpipe/sp_state_clip.c b/src/mesa/pipe/softpipe/sp_state_clip.c index c907550019..3b69d3e7a7 100644 --- a/src/mesa/pipe/softpipe/sp_state_clip.c +++ b/src/mesa/pipe/softpipe/sp_state_clip.c @@ -31,8 +31,7 @@ #include "sp_context.h" #include "sp_state.h" -#include "sp_draw.h" - +#include "pipe/draw/draw_context.h" void softpipe_set_clip_state( struct pipe_context *pipe, @@ -40,10 +39,8 @@ void softpipe_set_clip_state( struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); - memcpy(&softpipe->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); - - softpipe->nr_planes = 6 + clip->nr; - softpipe->dirty |= G_NEW_CLIP; + /* pass the clip state to the draw module */ + draw_set_clip_state(softpipe->draw, clip); } @@ -57,13 +54,32 @@ void softpipe_set_viewport_state( struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); softpipe->viewport = *viewport; /* struct copy */ + softpipe->dirty |= G_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + draw_set_viewport_state(softpipe->draw, viewport); /* Using tnl/ and vf/ modules is temporary while getting started. * Full pipe will have vertex shader, vertex fetch of its own. */ - draw_set_viewport( softpipe->draw, viewport->scale, viewport->translate ); - softpipe->dirty |= G_NEW_VIEWPORT; } +void softpipe_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + memcpy( &softpipe->scissor, scissor, sizeof(*scissor) ); + softpipe->dirty |= G_NEW_SCISSOR; +} + + +void softpipe_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + memcpy( &softpipe->poly_stipple, stipple, sizeof(*stipple) ); + softpipe->dirty |= G_NEW_STIPPLE; +} diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 5aee4be6b8..4454dd9ba9 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -25,17 +25,17 @@ * **************************************************************************/ -#include "glheader.h" -#include "macros.h" -#include "enums.h" -#include "program.h" +#include "main/glheader.h" +#include "main/macros.h" +#include "main/enums.h" +#include "shader/program.h" #include "vf/vf.h" - +#include "pipe/draw/draw_context.h" #include "sp_context.h" -#include "sp_draw.h" #include "sp_state.h" + #define EMIT_ATTR( ATTR, FRAG_ATTR, INTERP ) \ do { \ slot_to_vf_attr[softpipe->nr_attrs] = ATTR; \ @@ -77,8 +77,8 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) softpipe->nr_attrs = 0; memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); - memset(softpipe->fp_attr_to_slot, 0, sizeof(softpipe->vf_attr_to_slot)); - memset(softpipe->vf_attr_to_slot, 0, sizeof(softpipe->fp_attr_to_slot)); + memset(softpipe->fp_attr_to_slot, 0, sizeof(softpipe->fp_attr_to_slot)); + memset(softpipe->vf_attr_to_slot, 0, sizeof(softpipe->vf_attr_to_slot)); /* TODO - Figure out if we need to do perspective divide, etc. */ diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 060e2c8c98..62ff5c50b8 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -28,11 +28,11 @@ /* Authors: * Brian Paul */ + #include "imports.h" #include "sp_context.h" #include "sp_state.h" -#include "sp_draw.h" diff --git a/src/mesa/pipe/softpipe/sp_state_setup.c b/src/mesa/pipe/softpipe/sp_state_setup.c index 55803aeac5..2dea49c86d 100644 --- a/src/mesa/pipe/softpipe/sp_state_setup.c +++ b/src/mesa/pipe/softpipe/sp_state_setup.c @@ -28,11 +28,10 @@ #include "pipe/p_defines.h" #include "sp_context.h" #include "sp_state.h" -#include "sp_prim.h" - - +#include "pipe/draw/draw_context.h" +#if 0 static void validate_prim_pipe( struct softpipe_context *softpipe ) { struct prim_stage *next = softpipe->prim.setup; @@ -87,6 +86,7 @@ static void validate_prim_pipe( struct softpipe_context *softpipe ) softpipe->prim.first = next; } +#endif @@ -95,29 +95,15 @@ void softpipe_set_setup_state( struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); + /* pass-through to draw module */ + draw_set_setup_state(softpipe->draw, setup); + memcpy( &softpipe->setup, setup, sizeof(*setup) ); +#if 0 validate_prim_pipe( softpipe ); +#endif softpipe->dirty |= G_NEW_SETUP; } - -void softpipe_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - memcpy( &softpipe->scissor, scissor, sizeof(*scissor) ); - softpipe->dirty |= G_NEW_SCISSOR; -} - - -void softpipe_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - memcpy( &softpipe->poly_stipple, stipple, sizeof(*stipple) ); - softpipe->dirty |= G_NEW_STIPPLE; -} diff --git a/src/mesa/sources b/src/mesa/sources index b1ff3798a6..5bef7606d4 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -157,14 +157,6 @@ VF_SOURCES = \ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_clear.c \ pipe/softpipe/sp_context.c \ - pipe/softpipe/sp_draw.c \ - pipe/softpipe/sp_prim_clip.c \ - pipe/softpipe/sp_prim_cull.c \ - pipe/softpipe/sp_prim_flatshade.c \ - pipe/softpipe/sp_prim_offset.c \ - pipe/softpipe/sp_prim_setup.c \ - pipe/softpipe/sp_prim_twoside.c \ - pipe/softpipe/sp_prim_unfilled.c \ pipe/softpipe/sp_quad.c \ pipe/softpipe/sp_quad_alpha_test.c \ pipe/softpipe/sp_quad_blend.c \ @@ -177,7 +169,18 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_state_fs.c \ pipe/softpipe/sp_state_sampler.c \ pipe/softpipe/sp_state_setup.c \ - pipe/softpipe/sp_state_surface.c + pipe/softpipe/sp_state_surface.c \ + pipe/softpipe/sp_prim_setup.c + +DRAW_SOURCES = \ + pipe/draw/draw_clip.c \ + pipe/draw/draw_context.c\ + pipe/draw/draw_cull.c \ + pipe/draw/draw_flatshade.c \ + pipe/draw/draw_offset.c \ + pipe/draw/draw_twoside.c \ + pipe/draw/draw_unfilled.c \ + pipe/draw/draw_vb.c TGSICORE_SOURCES = \ pipe/tgsi/core/tgsi_build.c \ @@ -355,6 +358,7 @@ SOLO_SOURCES = \ $(VBO_SOURCES) \ $(VF_SOURCES) \ $(SOFTPIPE_SOURCES) \ + $(DRAW_SOURCES) \ $(TGSICORE_SOURCES) \ $(TGSIMESA_SOURCES) \ $(STATETRACKER_SOURCES) \ -- cgit v1.2.3 From 546eff6ba81e0fc8f15038de6db2a2a5fe228e01 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 9 Jul 2007 16:15:20 -0600 Subject: Obsolete file. --- src/mesa/pipe/softpipe/sp_state_point.c | 47 --------------------------------- 1 file changed, 47 deletions(-) delete mode 100644 src/mesa/pipe/softpipe/sp_state_point.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_point.c b/src/mesa/pipe/softpipe/sp_state_point.c deleted file mode 100644 index 566b33e696..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_point.c +++ /dev/null @@ -1,47 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_draw.h" - - - -void softpipe_set_point_state( struct pipe_context *pipe, - const struct pipe_point_state *point ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->point = *point; - - softpipe->dirty |= G_NEW_POINT; -} - -- cgit v1.2.3 From a380f98b6cc1da78d0845746ab86c3490a800126 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 9 Jul 2007 16:17:50 -0600 Subject: Rename G_NEW_* tokens to SP_NEW_* --- src/mesa/pipe/softpipe/sp_context.h | 24 ++++++++++++------------ src/mesa/pipe/softpipe/sp_state_blend.c | 8 ++++---- src/mesa/pipe/softpipe/sp_state_clip.c | 6 +++--- src/mesa/pipe/softpipe/sp_state_derived.c | 4 ++-- src/mesa/pipe/softpipe/sp_state_fs.c | 2 +- src/mesa/pipe/softpipe/sp_state_sampler.c | 4 ++-- src/mesa/pipe/softpipe/sp_state_setup.c | 2 +- src/mesa/pipe/softpipe/sp_state_surface.c | 2 +- 8 files changed, 26 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 0c082978bd..21d17c474f 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -51,18 +51,18 @@ enum interp_mode { }; -#define G_NEW_VIEWPORT 0x1 -#define G_NEW_SETUP 0x2 -#define G_NEW_FS 0x4 -#define G_NEW_BLEND 0x8 -#define G_NEW_CLIP 0x20 -#define G_NEW_SCISSOR 0x40 -#define G_NEW_STIPPLE 0x80 -#define G_NEW_FRAMEBUFFER 0x100 -#define G_NEW_ALPHA_TEST 0x200 -#define G_NEW_DEPTH_TEST 0x400 -#define G_NEW_SAMPLER 0x800 -#define G_NEW_TEXTURE 0x1000 +#define SP_NEW_VIEWPORT 0x1 +#define SP_NEW_SETUP 0x2 +#define SP_NEW_FS 0x4 +#define SP_NEW_BLEND 0x8 +#define SP_NEW_CLIP 0x10 +#define SP_NEW_SCISSOR 0x20 +#define SP_NEW_STIPPLE 0x40 +#define SP_NEW_FRAMEBUFFER 0x80 +#define SP_NEW_ALPHA_TEST 0x100 +#define SP_NEW_DEPTH_TEST 0x200 +#define SP_NEW_SAMPLER 0x400 +#define SP_NEW_TEXTURE 0x800 struct softpipe_context { diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index e0ff549d70..ae8950e386 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -40,7 +40,7 @@ void softpipe_set_blend_state( struct pipe_context *pipe, softpipe->blend = *blend; - softpipe->dirty |= G_NEW_BLEND; + softpipe->dirty |= SP_NEW_BLEND; } @@ -51,7 +51,7 @@ void softpipe_set_blend_color( struct pipe_context *pipe, softpipe->blend_color = *blend_color; - softpipe->dirty |= G_NEW_BLEND; + softpipe->dirty |= SP_NEW_BLEND; } @@ -66,7 +66,7 @@ softpipe_set_depth_test_state(struct pipe_context *pipe, softpipe->depth_test = *depth; - softpipe->dirty |= G_NEW_DEPTH_TEST; + softpipe->dirty |= SP_NEW_DEPTH_TEST; } void @@ -77,6 +77,6 @@ softpipe_set_alpha_test_state(struct pipe_context *pipe, softpipe->alpha_test = *alpha; - softpipe->dirty |= G_NEW_ALPHA_TEST; + softpipe->dirty |= SP_NEW_ALPHA_TEST; } diff --git a/src/mesa/pipe/softpipe/sp_state_clip.c b/src/mesa/pipe/softpipe/sp_state_clip.c index 3b69d3e7a7..8cf4383d3f 100644 --- a/src/mesa/pipe/softpipe/sp_state_clip.c +++ b/src/mesa/pipe/softpipe/sp_state_clip.c @@ -54,7 +54,7 @@ void softpipe_set_viewport_state( struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); softpipe->viewport = *viewport; /* struct copy */ - softpipe->dirty |= G_NEW_VIEWPORT; + softpipe->dirty |= SP_NEW_VIEWPORT; /* pass the viewport info to the draw module */ draw_set_viewport_state(softpipe->draw, viewport); @@ -71,7 +71,7 @@ void softpipe_set_scissor_state( struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); memcpy( &softpipe->scissor, scissor, sizeof(*scissor) ); - softpipe->dirty |= G_NEW_SCISSOR; + softpipe->dirty |= SP_NEW_SCISSOR; } @@ -81,5 +81,5 @@ void softpipe_set_polygon_stipple( struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); memcpy( &softpipe->poly_stipple, stipple, sizeof(*stipple) ); - softpipe->dirty |= G_NEW_STIPPLE; + softpipe->dirty |= SP_NEW_STIPPLE; } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 4454dd9ba9..dd8a490f3b 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -131,10 +131,10 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ void softpipe_update_derived( struct softpipe_context *softpipe ) { - if (softpipe->dirty & (G_NEW_SETUP | G_NEW_FS)) + if (softpipe->dirty & (SP_NEW_SETUP | SP_NEW_FS)) calculate_vertex_layout( softpipe ); - if (softpipe->dirty & (G_NEW_BLEND | G_NEW_DEPTH_TEST | G_NEW_ALPHA_TEST | G_NEW_FS)) + if (softpipe->dirty & (SP_NEW_BLEND | SP_NEW_DEPTH_TEST | SP_NEW_ALPHA_TEST | SP_NEW_FS)) sp_build_quad_pipeline(softpipe); softpipe->dirty = 0; diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 29f965e50e..c7ef1f1cfc 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -37,7 +37,7 @@ void softpipe_set_fs_state( struct pipe_context *pipe, memcpy(&softpipe->fs, fs, sizeof(*fs)); - softpipe->dirty |= G_NEW_FS; + softpipe->dirty |= SP_NEW_FS; } diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 62ff5c50b8..9ef71f73cb 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -46,7 +46,7 @@ softpipe_set_sampler_state(struct pipe_context *pipe, assert(unit < PIPE_MAX_SAMPLERS); softpipe->sampler[unit] = *sampler; - softpipe->dirty |= G_NEW_SAMPLER; + softpipe->dirty |= SP_NEW_SAMPLER; } @@ -60,5 +60,5 @@ softpipe_set_texture_state(struct pipe_context *pipe, assert(unit < PIPE_MAX_SAMPLERS); softpipe->texture[unit] = texture; /* ptr, not struct */ - softpipe->dirty |= G_NEW_TEXTURE; + softpipe->dirty |= SP_NEW_TEXTURE; } diff --git a/src/mesa/pipe/softpipe/sp_state_setup.c b/src/mesa/pipe/softpipe/sp_state_setup.c index 2dea49c86d..6dbb87ae79 100644 --- a/src/mesa/pipe/softpipe/sp_state_setup.c +++ b/src/mesa/pipe/softpipe/sp_state_setup.c @@ -103,7 +103,7 @@ void softpipe_set_setup_state( struct pipe_context *pipe, #if 0 validate_prim_pipe( softpipe ); #endif - softpipe->dirty |= G_NEW_SETUP; + softpipe->dirty |= SP_NEW_SETUP; } diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index 4d27814e1b..8ce81eb2b0 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -45,7 +45,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, softpipe->framebuffer = *fb; /* struct copy */ - softpipe->dirty |= G_NEW_FRAMEBUFFER; + softpipe->dirty |= SP_NEW_FRAMEBUFFER; } -- cgit v1.2.3 From ebafc193fac22c4d1d8d4f1c5a129d5dbe1679a2 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 9 Jul 2007 16:20:16 -0600 Subject: s/G_/SP_/ --- src/mesa/pipe/softpipe/sp_headers.h | 11 +++++------ src/mesa/pipe/softpipe/sp_quad.h | 13 +++++++------ src/mesa/pipe/softpipe/sp_surface.h | 6 +++--- 3 files changed, 15 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index 561597d9e2..82f03ec078 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -27,8 +27,10 @@ /* Authors: Keith Whitwell */ -#ifndef G_HEADERS_H -#define G_HEADERS_H + +#ifndef SP_HEADERS_H +#define SP_HEADERS_H + #define PRIM_POINT 1 #define PRIM_LINE 2 @@ -89,7 +91,4 @@ struct quad_header { }; - - - -#endif +#endif /* SP_HEADERS_H */ diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 351cd4d323..79f52721a3 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -27,8 +27,10 @@ /* Authors: Keith Whitwell */ -#ifndef G_TILE_H -#define G_TILE_H + +#ifndef SP_TILE_H +#define SP_TILE_H + struct softpipe_context; struct quad_header; @@ -44,14 +46,13 @@ struct quad_stage { }; - struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); -void -sp_build_quad_pipeline(struct softpipe_context *sp); +void sp_build_quad_pipeline(struct softpipe_context *sp); + -#endif +#endif /* SP_TILE_H */ diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 05b125d17b..ac66ffe891 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -28,8 +28,8 @@ /* Authors: Keith Whitwell */ -#ifndef G_SURFACE_H -#define G_SURFACE_H +#ifndef SP_SURFACE_H +#define SP_SURFACE_H #include "glheader.h" #include "sp_headers.h" @@ -93,4 +93,4 @@ softpipe_surface(struct pipe_surface *ps) } -#endif +#endif /* SP_SURFACE_H */ -- cgit v1.2.3 From f4053ce858739167fe57a36aebe1da7110d1a04e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 9 Jul 2007 16:21:18 -0600 Subject: Remove softpipe includes --- src/mesa/pipe/draw/draw_vb.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index 66573a93fe..74e7d7f243 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -36,8 +36,6 @@ #include "tnl/t_context.h" #include "vf/vf.h" -#include "pipe/softpipe/sp_context.h" -#include "pipe/softpipe/sp_headers.h" #include "draw_private.h" #include "draw_context.h" -- cgit v1.2.3 From d348a48ac353d7c646a0c4f976f23a764990e779 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 10 Jul 2007 18:07:38 +0200 Subject: First attempt at using private back/z buffers. Allocate and use private back and z/stencil buffers. This is still very broken. --- src/mesa/drivers/dri/i915tex/intel_blit.c | 8 +++-- src/mesa/drivers/dri/i915tex/intel_context.c | 11 ++++++ src/mesa/drivers/dri/i915tex/intel_context.h | 2 +- src/mesa/drivers/dri/i915tex/intel_fbo.c | 25 +++++++++++++- src/mesa/drivers/dri/i915tex/intel_screen.c | 51 +++++++++++++++++++++++++--- src/mesa/drivers/dri/i915tex/intel_screen.h | 1 + 6 files changed, 89 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c index dbe4ba2ac5..5dcb0ddebb 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ b/src/mesa/drivers/dri/i915tex/intel_blit.c @@ -88,6 +88,7 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, const int nbox = dPriv->numClipRects; const drm_clip_rect_t *pbox = dPriv->pClipRects; const int pitch = frontRegion->pitch; + const int srcpitch = backRegion->pitch; const int cpp = frontRegion->cpp; int BR13, CMD; int i; @@ -96,9 +97,12 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ ASSERT(frontRegion); ASSERT(backRegion); - ASSERT(frontRegion->pitch == backRegion->pitch); +// ASSERT(frontRegion->pitch == backRegion->pitch); ASSERT(frontRegion->cpp == backRegion->cpp); + DBG("copy buffer, front pitch %d back pitch %d\n", + frontRegion->pitch, backRegion->pitch); + if (cpp == 2) { BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); CMD = XY_SRC_COPY_BLT_CMD; @@ -142,7 +146,7 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, OUT_RELOC(frontRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); OUT_BATCH((pbox->y1 << 16) | pbox->x1); - OUT_BATCH(BR13 & 0xffff); + OUT_BATCH((srcpitch * cpp) & 0xffff); OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index c927dca8e5..028e57e5ab 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -569,6 +569,13 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driReadPriv) { +#if 0 + if (driDrawPriv) { + fprintf(stderr, "x %d, y %d, width %d, height %d\n", + driDrawPriv->x, driDrawPriv->y, driDrawPriv->w, driDrawPriv->h); + } +#endif + if (driContextPriv) { struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; @@ -576,6 +583,9 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, (struct intel_framebuffer *) driDrawPriv->driverPrivate; GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; + /* this is a hack so we have a valid context when the region allocation + is done. Need a per-screen context? */ + intel->intelScreen->dummyctxptr = intel; /* XXX FBO temporary fix-ups! */ /* if the renderbuffers don't have regions, init them from the context */ @@ -613,6 +623,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, } _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); + intel->intelScreen->dummyctxptr = &intel->ctx; /* The drawbuffer won't always be updated by _mesa_make_current: */ diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index 9d060eb866..7a9af1b5fd 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -353,7 +353,7 @@ __memcpy(void *to, const void *from, size_t n) /* ================================================================ * Debugging: */ -#define DO_DEBUG 0 +#define DO_DEBUG 1 #if DO_DEBUG extern int INTEL_DEBUG; #else diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c index 349912ffec..d3f14bcd25 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c @@ -257,7 +257,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, break; default: _mesa_problem(ctx, - "Unexpected format in intel_alloc_renderbuffer_storage"); + "Unexpected format (%x) in intel_alloc_renderbuffer_storage", internalFormat); return GL_FALSE; } @@ -453,6 +453,29 @@ intel_create_renderbuffer(GLenum intFormat, GLsizei width, GLsizei height, return irb; } +struct gl_renderbuffer * +intel_new_renderbuffer_fb(GLcontext * ctx, GLuint intFormat) +{ + struct intel_renderbuffer *irb; + + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, 0); + irb->Base.ClassID = INTEL_RB_CLASS; + irb->Base.InternalFormat = intFormat; + + /* intel-specific methods */ + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; + irb->Base.GetPointer = intel_get_pointer; + /* span routines set in alloc_storage function */ + + return &irb->Base; +} /** * Create a new renderbuffer object. diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c index 2acdead63d..a5a9c79625 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ b/src/mesa/drivers/dri/i915tex/intel_screen.c @@ -156,7 +156,7 @@ intel_recreate_static(intelScreenPrivate *intelScreen, } return region; } - + /* Create intel_region structs to describe the static front,back,depth * buffers created by the xserver. @@ -192,7 +192,7 @@ intel_recreate_static_regions(intelScreenPrivate *intelScreen) intelScreen->rotated.pitch / intelScreen->cpp, intelScreen->height); - +#if 0 intelScreen->back_region = intel_recreate_static(intelScreen, intelScreen->back_region, @@ -226,6 +226,7 @@ intel_recreate_static_regions(intelScreenPrivate *intelScreen) intelScreen->cpp, intelScreen->depth.pitch / intelScreen->cpp, intelScreen->height); +#endif } /** @@ -382,7 +383,7 @@ intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, intelScreen->rotatedWidth = sarea->virtualX; intelScreen->rotatedHeight = sarea->virtualY; - if (0) + if (1) intelPrintSAREA(sarea); } @@ -614,6 +615,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, &intel_fb->color_rb[0]->Base); } +#if 0 if (mesaVis->doubleBufferMode) { intel_fb->color_rb[1] = intel_create_renderbuffer(rgbFormat, @@ -640,7 +642,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base); } } - if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { /* combined depth/stencil buffer */ struct intel_renderbuffer *depthStencilRb @@ -670,6 +671,39 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); } +#else + if (mesaVis->doubleBufferMode) { + intel_fb->color_rb[1] + = intel_new_renderbuffer_fb(NULL, rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, + &intel_fb->color_rb[1]->Base); + if (screen->third.handle) { + struct gl_renderbuffer *tmp_rb = NULL; + + intel_fb->color_rb[2] + = intel_new_renderbuffer_fb(NULL, rgbFormat); + _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base); + } + } + if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { + /* combined depth/stencil buffer */ + struct intel_renderbuffer *depthStencilRb + = intel_new_renderbuffer_fb(NULL, GL_DEPTH24_STENCIL8_EXT); + /* note: bind RB to two attachment points */ + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, + &depthStencilRb->Base); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL, + &depthStencilRb->Base); + } + else if (mesaVis->depthBits == 16) { + /* just 16-bit depth buffer, no hw stencil */ + struct intel_renderbuffer *depthRb + = intel_new_renderbuffer_fb(NULL, GL_DEPTH_COMPONENT16); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); + } + +#endif + /* now add any/all software-based renderbuffers we may need */ _mesa_add_soft_renderbuffers(&intel_fb->Base, GL_FALSE, /* never sw color */ @@ -939,11 +973,18 @@ struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) * context at screen creation. For now just use the current context. */ - GET_CURRENT_CONTEXT(ctx); +/* GET_CURRENT_CONTEXT(ctx); if (ctx == NULL) { _mesa_problem(NULL, "No current context in intelScreenContext\n"); return NULL; } return intel_context(ctx); +*/ + if (intelScreen->dummyctxptr == NULL) { + _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; + } + return intelScreen->dummyctxptr; + } diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.h b/src/mesa/drivers/dri/i915tex/intel_screen.h index bac43aaddd..783ee35524 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.h +++ b/src/mesa/drivers/dri/i915tex/intel_screen.h @@ -96,6 +96,7 @@ typedef struct struct _DriBufferPool *staticPool; unsigned int maxBatchSize; GLboolean havePools; + struct intel_context *dummyctxptr; } intelScreenPrivate; -- cgit v1.2.3 From d68ea4e23fd0cf1fedabb667a8067bfdc3fb308c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 10:24:28 -0600 Subject: remove some dead/disabled stuff --- src/mesa/pipe/softpipe/sp_prim_setup.c | 38 +++++++++++++++++----------------- 1 file changed, 19 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index ea10ef53a0..34e3bedb06 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -140,7 +140,7 @@ static void run_shader_block( struct setup_stage *setup, setup->quad.y0 = y; setup->quad.mask = mask; - quad_emit(setup->/*stage.*/softpipe, &setup->quad); + quad_emit(setup->softpipe, &setup->quad); } @@ -394,7 +394,7 @@ static void tri_persp_coeff( struct setup_stage *setup, */ static void setup_tri_coefficients( struct setup_stage *setup ) { - const enum interp_mode *interp = setup->/*stage.*/softpipe->interp; + const enum interp_mode *interp = setup->softpipe->interp; GLuint slot, j; /* z and w are done by linear interpolation: @@ -469,15 +469,15 @@ static void subtriangle( struct setup_stage *setup, /* scissor y: */ - if (setup->/*stage.*/softpipe->setup.scissor) { + if (setup->softpipe->setup.scissor) { start_y = sy; finish_y = start_y + lines; - if (start_y < setup->/*stage.*/softpipe->scissor.miny) - start_y = setup->/*stage.*/softpipe->scissor.miny; + if (start_y < setup->softpipe->scissor.miny) + start_y = setup->softpipe->scissor.miny; - if (finish_y > setup->/*stage.*/softpipe->scissor.maxy) - finish_y = setup->/*stage.*/softpipe->scissor.maxy; + if (finish_y > setup->softpipe->scissor.maxy) + finish_y = setup->softpipe->scissor.maxy; start_y -= sy; finish_y -= sy; @@ -502,12 +502,12 @@ static void subtriangle( struct setup_stage *setup, /* scissor x: */ - if (setup->/*stage.*/softpipe->setup.scissor) { - if (left < setup->/*stage.*/softpipe->scissor.minx) - left = setup->/*stage.*/softpipe->scissor.minx; + if (setup->softpipe->setup.scissor) { + if (left < setup->softpipe->scissor.minx) + left = setup->softpipe->scissor.minx; - if (right > setup->/*stage.*/softpipe->scissor.maxx) - right = setup->/*stage.*/softpipe->scissor.maxx; + if (right > setup->softpipe->scissor.maxx) + right = setup->softpipe->scissor.maxx; } if (left < right) { @@ -611,7 +611,7 @@ line_persp_coeff(struct setup_stage *setup, GLuint slot, GLuint i) static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { - const enum interp_mode *interp = setup->/*stage.*/softpipe->interp; + const enum interp_mode *interp = setup->softpipe->interp; GLuint slot, j; /* use setup->vmin, vmax to point to vertices */ @@ -671,7 +671,7 @@ plot(struct setup_stage *setup, GLint x, GLint y) /* flush prev quad, start new quad */ if (setup->quad.x0 != -1) - quad_emit(setup->/*stage.*/softpipe, &setup->quad); + quad_emit(setup->softpipe, &setup->quad); setup->quad.x0 = quadX; setup->quad.y0 = quadY; @@ -774,7 +774,7 @@ setup_line(struct prim_stage *stage, struct prim_header *prim) /* draw final quad */ if (setup->quad.mask) { - quad_emit(setup->/*stage.*/softpipe, &setup->quad); + quad_emit(setup->softpipe, &setup->quad); } } @@ -789,8 +789,8 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); /*XXX this should be a vertex attrib! */ - GLfloat halfSize = 0.5 * setup->/*stage.*/softpipe->setup.point_size; - GLboolean round = setup->/*stage.*/softpipe->setup.point_smooth; + GLfloat halfSize = 0.5 * setup->softpipe->setup.point_size; + GLboolean round = setup->softpipe->setup.point_smooth; const struct vertex_header *v0 = prim->v[0]; const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; @@ -829,7 +829,7 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) setup->quad.x0 = x - ix; setup->quad.y0 = y - iy; setup->quad.mask = (1 << ix) << (2 * iy); - quad_emit(setup->/*stage.*/softpipe, &setup->quad); + quad_emit(setup->softpipe, &setup->quad); } else { const GLint ixmin = block((GLint) (x - halfSize)); @@ -889,7 +889,7 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) if (setup->quad.mask) { setup->quad.x0 = ix; setup->quad.y0 = iy; - quad_emit( setup->/*stage.*/softpipe, &setup->quad ); + quad_emit( setup->softpipe, &setup->quad ); } } } -- cgit v1.2.3 From 8c1fa904edf991159a53b0f4bba04e2f5e326437 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 10:53:29 -0600 Subject: Fix EMIT_ATTR() to populate the draw->vf_attr_to_slot[] array. Note that attribute index has to be biased by two, since vertex->data[] starts after the header and clipcoord fields. See comments for details. Added a bunch of comments/docs. --- src/mesa/pipe/draw/draw_vb.c | 47 ++++++++++++++++++++++++++++++++++---------- 1 file changed, 37 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index 74e7d7f243..905e465447 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -616,32 +616,59 @@ void draw_vb(struct draw_context *draw, } -#define EMIT_ATTR( ATTR, STYLE ) \ -do { \ - draw->attrs[draw->nr_attrs].attrib = ATTR; \ - draw->attrs[draw->nr_attrs].format = STYLE; \ - draw->nr_attrs++; \ +/** + * Accumulate another attribute's info. + * Note the "- 2" factor here. We need this because the vertex->data[] + * array does not include the first two attributes we emit (VERTEX_HEADER + * and CLIP_POS). So, the 3rd attribute actually winds up in the 1st + * position of the data[] array. + */ +#define EMIT_ATTR( VF_ATTR, STYLE ) \ +do { \ + if (draw->nr_attrs >= 2) \ + draw->vf_attr_to_slot[VF_ATTR] = draw->nr_attrs - 2; \ + draw->attrs[draw->nr_attrs].attrib = VF_ATTR; \ + draw->attrs[draw->nr_attrs].format = STYLE; \ + draw->nr_attrs++; \ } while (0) +/** + * Tell the draw module about the layout of attributes in the vertex. + * We need this in order to know which vertex slot has color0, etc. + * + * \param slot_to_vf_attr an array which maps slot indexes to vertex + * format tokens (VF_*). + * \param nr_attrs the size of the slot_to_vf_attr array + * (and number of attributes) + */ void draw_set_vertex_attributes( struct draw_context *draw, - const GLuint *attrs, + const GLuint *slot_to_vf_attr, GLuint nr_attrs ) { GLuint i; + memset(draw->vf_attr_to_slot, 0, sizeof(draw->vf_attr_to_slot)); draw->nr_attrs = 0; + /* + * First three attribs are always the same: header, clip pos, winpos + */ EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F); EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F); - assert(attrs[0] == VF_ATTRIB_POS); - EMIT_ATTR(attrs[0], EMIT_4F_VIEWPORT); + assert(slot_to_vf_attr[0] == VF_ATTRIB_POS); + EMIT_ATTR(slot_to_vf_attr[0], EMIT_4F_VIEWPORT); + /* + * Remaining attribs (color, texcoords, etc) + */ for (i = 1; i < nr_attrs; i++) - EMIT_ATTR(attrs[i], EMIT_4F); + EMIT_ATTR(slot_to_vf_attr[i], EMIT_4F); - draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); + /* tell the vertex format module how to construct vertices for us */ + draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, + draw->nr_attrs, 0 ); } -- cgit v1.2.3 From 57a53856522b749dbe9f88f5887933c253524212 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 10:53:57 -0600 Subject: Added comments, assertions. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 34e3bedb06..8e43f5d04d 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -306,11 +306,18 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, /** * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + * The value value comes from vertex->data[slot][i]. + * The result will be put into setup->coef[slot].a0[i]. + * \param slot which attribute slot + * \param i which component of the slot (0..3) */ static void const_coeff( struct setup_stage *setup, GLuint slot, GLuint i ) { + assert(slot < FRAG_ATTRIB_MAX); + assert(i <= 3); + setup->coef[slot].dadx[i] = 0; setup->coef[slot].dady[i] = 0; @@ -333,6 +340,9 @@ static void tri_linear_coeff( struct setup_stage *setup, GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + assert(slot < FRAG_ATTRIB_MAX); + assert(i <= 3); + setup->coef[slot].dadx[i] = a * setup->oneoverarea; setup->coef[slot].dady[i] = b * setup->oneoverarea; @@ -379,6 +389,9 @@ static void tri_persp_coeff( struct setup_stage *setup, GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + assert(slot < FRAG_ATTRIB_MAX); + assert(i <= 3); + setup->coef[slot].dadx[i] = a * setup->oneoverarea; setup->coef[slot].dady[i] = b * setup->oneoverarea; setup->coef[slot].a0[i] = (mina - -- cgit v1.2.3 From e7a208514812506eeb9862b91258392da646b618 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 11:00:24 -0600 Subject: comments --- src/mesa/pipe/softpipe/sp_context.h | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 21d17c474f..c6046624a1 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -66,7 +66,7 @@ enum interp_mode { struct softpipe_context { - struct pipe_context pipe; + struct pipe_context pipe; /**< base class */ /* The most recent drawing state as set by the driver: */ @@ -97,22 +97,23 @@ struct softpipe_context { /* FS + setup derived state: */ + + /** Map fragment program attribute to quad/coef array slot */ GLuint fp_attr_to_slot[PIPE_ATTRIB_MAX]; + /** Map vertex format attribute to a vertex attribute slot */ GLuint vf_attr_to_slot[PIPE_ATTRIB_MAX]; GLuint nr_attrs; - GLuint nr_frag_attrs; - GLuint attr_mask; + GLuint nr_frag_attrs; /**< number of active fragment attribs */ + GLbitfield attr_mask; /**< bitfield of VF_ATTRIB_ indexes/bits */ - GLboolean need_z; - GLboolean need_w; + GLboolean need_z; /**< produce quad/fragment Z values? */ + GLboolean need_w; /**< produce quad/fragment W values? */ /* Stipple derived state: */ GLubyte stipple_masks[16][16]; - /* - * Software quad rendering pipeline - */ + /** Software quad rendering pipeline */ struct { struct quad_stage *shade; struct quad_stage *alpha_test; @@ -123,8 +124,7 @@ struct softpipe_context { struct quad_stage *first; /**< points to one of the above stages */ } quad; - /* Temp kludge: - */ + /** The primitive drawing context */ struct draw_context *draw; }; -- cgit v1.2.3 From 17b86cfa4e9403b2d887d832556dc4f1a1271ce0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 11:03:51 -0600 Subject: Comments, minor clean-up. --- src/mesa/pipe/softpipe/sp_state_derived.c | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index dd8a490f3b..429ae55397 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -36,18 +36,18 @@ #include "sp_state.h" -#define EMIT_ATTR( ATTR, FRAG_ATTR, INTERP ) \ -do { \ - slot_to_vf_attr[softpipe->nr_attrs] = ATTR; \ - softpipe->vf_attr_to_slot[ATTR] = softpipe->nr_attrs; \ +#define EMIT_ATTR( VF_ATTR, FRAG_ATTR, INTERP ) \ +do { \ + slot_to_vf_attr[softpipe->nr_attrs] = VF_ATTR; \ + softpipe->vf_attr_to_slot[VF_ATTR] = softpipe->nr_attrs; \ softpipe->fp_attr_to_slot[FRAG_ATTR] = softpipe->nr_attrs; \ - softpipe->interp[softpipe->nr_attrs] = INTERP; \ - softpipe->nr_attrs++; \ - attr_mask |= (1<interp[softpipe->nr_attrs] = INTERP; \ + softpipe->nr_attrs++; \ + attr_mask |= (1 << (VF_ATTR)); \ } while (0) -static GLuint frag_to_vf[FRAG_ATTRIB_MAX] = +static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = { VF_ATTRIB_POS, VF_ATTRIB_COLOR0, @@ -71,7 +71,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) struct gl_fragment_program *fp = softpipe->fs.fp; const GLuint inputsRead = fp->Base.InputsRead; GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; - GLuint attr_mask = 0; + GLbitfield attr_mask = 0x0; GLuint i; softpipe->nr_attrs = 0; @@ -116,6 +116,9 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) } } + /* If the attributes have changed, tell the draw module (which in turn + * tells the vf module) about the new vertex layout. + */ if (attr_mask != softpipe->attr_mask) { softpipe->attr_mask = attr_mask; -- cgit v1.2.3 From 1fe4cf83cd929a0ecc1119d3bb541e80618d3987 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 11:54:03 -0600 Subject: Fix front/back mix-up. added comments. --- src/mesa/pipe/draw/draw_unfilled.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index a1d9d14352..8ee23d7493 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -25,6 +25,11 @@ * **************************************************************************/ +/** + * \brief Drawing stage for handling glPolygonMode(line/point). + * Convert triangles to points or lines as needed. + */ + /* Authors: Keith Whitwell */ @@ -36,6 +41,10 @@ struct unfilled_stage { struct prim_stage stage; + /** [0] = front face, [1] = back face. + * legal values: PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE, + * and PIPE_POLYGON_MODE_POINT, + */ GLuint mode[2]; }; @@ -50,8 +59,8 @@ static void unfilled_begin( struct prim_stage *stage ) { struct unfilled_stage *unfilled = unfilled_stage(stage); - unfilled->mode[0] = stage->draw->setup.fill_ccw; - unfilled->mode[1] = stage->draw->setup.fill_cw; + unfilled->mode[0] = stage->draw->setup.fill_ccw; /* front */ + unfilled->mode[1] = stage->draw->setup.fill_cw; /* back */ stage->next->begin( stage->next ); } @@ -87,6 +96,7 @@ static void points( struct prim_stage *stage, if (v2->edgeflag) point( stage, v2 ); } + static void lines( struct prim_stage *stage, struct prim_header *header ) { @@ -109,7 +119,7 @@ static void unfilled_tri( struct prim_stage *stage, struct prim_header *header ) { struct unfilled_stage *unfilled = unfilled_stage(stage); - GLuint mode = unfilled->mode[header->det < 0]; + GLuint mode = unfilled->mode[header->det > 0.0]; switch (mode) { case PIPE_POLYGON_MODE_FILL: -- cgit v1.2.3 From 720c0eb71db9908c5ecef15263c2ae638d61d07b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 11:54:16 -0600 Subject: added comments --- src/mesa/pipe/draw/draw_cull.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index d27d33a40d..e7343cfd76 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -25,22 +25,25 @@ * **************************************************************************/ +/** + * \brief Drawing stage for polygon culling + */ + /* Authors: Keith Whitwell */ + #include "main/imports.h" #include "pipe/p_defines.h" #include "draw_private.h" - struct cull_stage { struct prim_stage stage; - GLuint mode; + GLuint mode; /**< one of PIPE_WINDING_x */ }; - static INLINE struct cull_stage *cull_stage( struct prim_stage *stage ) { return (struct cull_stage *)stage; -- cgit v1.2.3 From 1278514ff48b262ee0a4f2ac698c6df648b326a0 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 10 Jul 2007 19:56:44 +0200 Subject: copybuffer/cliprect fixups --- src/mesa/drivers/dri/i915tex/intel_blit.c | 6 ++++- src/mesa/drivers/dri/i915tex/intel_buffers.c | 40 ++++++++++++++++++++++++++-- src/mesa/drivers/dri/i915tex/intel_context.h | 1 + 3 files changed, 44 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c index 5dcb0ddebb..b4fc1691e6 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ b/src/mesa/drivers/dri/i915tex/intel_blit.c @@ -115,6 +115,7 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, for (i = 0; i < nbox; i++, pbox++) { drm_clip_rect_t box; + drm_clip_rect_t sbox; if (pbox->x1 > pbox->x2 || pbox->y1 > pbox->y2 || @@ -137,6 +138,9 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, continue; } + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); OUT_BATCH(CMD); OUT_BATCH(BR13); @@ -145,7 +149,7 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, OUT_RELOC(frontRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((pbox->y1 << 16) | pbox->x1); + OUT_BATCH((sbox.y1 << 16) | sbox.x1); OUT_BATCH((srcpitch * cpp) & 0xffff); OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 62ff54b007..e8dd7cd524 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -146,6 +146,27 @@ intelSetRenderbufferClipRects(struct intel_context *intel) intel->drawY = 0; } +/** + * As above, but for rendering private front/back buffer of a window. + * \sa intelSetPrivbufClipRects + */ + +static void +intelSetPrivbufClipRects(struct intel_context *intel) +{ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + if (!dPriv) + return; + + intel->fakeClipRect.x1 = 0; + intel->fakeClipRect.y1 = 0; + intel->fakeClipRect.x2 = dPriv->w; + intel->fakeClipRect.y2 = dPriv->h; + intel->pClipRects = &intel->fakeClipRect; + intel->numClipRects = 1; + intel->drawX = 0; + intel->drawY = 0; +} /** * As above, but for rendering to front buffer of a window. @@ -212,7 +233,7 @@ intelWindowMoved(struct intel_context *intel) /* when would this happen? -BP */ intelSetFrontClipRects(intel); } - else if (intel->ctx.DrawBuffer->Name != 0) { + else if (1 || intel->ctx.DrawBuffer->Name != 0) { /* drawing to user-created FBO - do nothing */ /* Cliprects would be set from intelDrawBuffer() */ } @@ -953,10 +974,17 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) if (ctx->Visual.doubleBufferMode) { drm_clip_rect_t rect; +#if 1 rect.x1 = x + dPriv->x; rect.y1 = (dPriv->h - y - h) + dPriv->y; rect.x2 = rect.x1 + w; rect.y2 = rect.y1 + h; +#else + rect.x1 = x; + rect.y1 = dPriv->h - y; + rect.x2 = rect.x1 + w; + rect.y2 = rect.y1 + h; +#endif _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ intelCopyBuffer(dPriv, &rect); } @@ -991,7 +1019,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) return; } - /* Do this here, note core Mesa, since this function is called from + /* Do this here, not core Mesa, since this function is called from * many places within the driver. */ if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { @@ -1042,11 +1070,19 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) if (fb->Name == 0) { /* drawing to window system buffer */ if (front) { +#if 0 intelSetFrontClipRects(intel); +#else + intelSetPrivbufClipRects(intel); +#endif colorRegion = intel_get_rb_region(fb, BUFFER_FRONT_LEFT); } else { +#if 0 intelSetBackClipRects(intel); +#else + intelSetPrivbufClipRects(intel); +#endif colorRegion = intel_get_rb_region(fb, BUFFER_BACK_LEFT); } } diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index 7a9af1b5fd..d05b20abbd 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -254,6 +254,7 @@ struct intel_context GLuint numClipRects; /**< cliprects for drawing */ drm_clip_rect_t *pClipRects; drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ + drm_clip_rect_t fakeClipRect; /**< cliprect for priv back/fake front buffers rendering */ int perf_boxes; -- cgit v1.2.3 From 5e4379b5a430bcaf7dea4814341830fce0820cb1 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 13:33:45 -0600 Subject: comments, clean-up --- src/mesa/pipe/draw/draw_context.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index c91be23eef..d4654791bb 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -81,12 +81,15 @@ static void validate_pipeline( struct draw_context *draw ) { struct prim_stage *next = draw->pipeline.setup; - /* TODO: make the current primitive part of the state and build + /* + * NOTE: we build up the pipeline in end-to-start order. + * + * TODO: make the current primitive part of the state and build * shorter pipelines for lines & points. */ + if (draw->setup.fill_cw != PIPE_POLYGON_MODE_FILL || draw->setup.fill_ccw != PIPE_POLYGON_MODE_FILL) { - draw->pipeline.unfilled->next = next; next = draw->pipeline.unfilled; } @@ -110,7 +113,6 @@ static void validate_pipeline( struct draw_context *draw ) next = draw->pipeline.cull; } - /* Clip stage */ { @@ -127,7 +129,6 @@ static void validate_pipeline( struct draw_context *draw ) next = draw->pipeline.flatshade; } - draw->pipeline.first = next; } -- cgit v1.2.3 From f6dffd6ee70473dcbf65420a9c635049199f7a4e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 13:34:23 -0600 Subject: fix problems with polygon offset --- src/mesa/state_tracker/st_atom_setup.c | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index 457bfaa9ba..842a87c8a1 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -52,15 +52,15 @@ static GLuint translate_fill( GLenum mode ) } static GLboolean get_offset_flag( GLuint fill_mode, - const struct gl_polygon_attrib *Polygon ) + const struct gl_polygon_attrib *p ) { switch (fill_mode) { case PIPE_POLYGON_MODE_POINT: - return Polygon->OffsetPoint; + return p->OffsetPoint; case PIPE_POLYGON_MODE_LINE: - return Polygon->OffsetLine; + return p->OffsetLine; case PIPE_POLYGON_MODE_FILL: - return Polygon->OffsetFill; + return p->OffsetFill; default: assert(0); return 0; @@ -141,16 +141,15 @@ static void update_setup_state( struct st_context *st ) } } - /* Hardware does offset for filled prims, but need to do it in - * software for unfilled. - * - * _NEW_POLYGON + /* _NEW_POLYGON */ - if (setup.fill_cw != PIPE_POLYGON_MODE_FILL) + if (ctx->Polygon.OffsetUnits != 0.0 || + ctx->Polygon.OffsetFactor != 0.0) { setup.offset_cw = get_offset_flag( setup.fill_cw, &ctx->Polygon ); - - if (setup.fill_ccw != PIPE_POLYGON_MODE_FILL) setup.offset_ccw = get_offset_flag( setup.fill_ccw, &ctx->Polygon ); + setup.offset_units = ctx->Polygon.OffsetUnits; + setup.offset_scale = ctx->Polygon.OffsetFactor; + } if (ctx->Polygon.SmoothFlag) setup.poly_smooth = 1; -- cgit v1.2.3 From 1deafdb1dec24c2920ab92098f1433273b2ccbdd Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 13:53:21 -0600 Subject: Do depth testing with integer values. Using floats (and float->ushort->float conversion) introduces errors. Only GLushort depth buffers work for now... --- src/mesa/drivers/x11/xm_surface.c | 12 ++++++--- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 39 +++++++++++++++++++++++------ src/mesa/pipe/softpipe/sp_surface.h | 4 +-- 3 files changed, 41 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 30c9049cbf..5158e42d9a 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -43,6 +43,7 @@ #include "renderbuffer.h" #include "pipe/p_state.h" +#include "pipe/p_defines.h" #include "pipe/softpipe/sp_context.h" #include "pipe/softpipe/sp_surface.h" @@ -245,7 +246,7 @@ xmesa_get_color_surface(GLcontext *ctx, GLuint buf) static void read_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, GLfloat zzzz[QUAD_SIZE]) + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) { struct xmesa_surface *xmsurf = xmesa_surface(sps); struct gl_renderbuffer *rb = xmsurf->rb; @@ -254,22 +255,24 @@ read_quad_z(struct softpipe_surface *sps, GET_CURRENT_CONTEXT(ctx); rb->GetRow(ctx, rb, 2, x, y, temp); rb->GetRow(ctx, rb, 2, x, y + 1, temp + 2); + /* convert from GLushort to GLuint */ for (i = 0; i < 4; i++) { - zzzz[i] = USHORT_TO_FLOAT(temp[i]); + zzzz[i] = temp[i]; } } static void write_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, const GLfloat zzzz[QUAD_SIZE]) + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) { struct xmesa_surface *xmsurf = xmesa_surface(sps); struct gl_renderbuffer *rb = xmsurf->rb; GLushort temp[4]; GLuint i; GET_CURRENT_CONTEXT(ctx); + /* convert from GLuint to GLushort */ for (i = 0; i < 4; i++) { - CLAMPED_FLOAT_TO_USHORT(temp[i], zzzz[i]); + temp[i] = zzzz[i]; } rb->PutRow(ctx, rb, 2, x, y, temp, NULL); rb->PutRow(ctx, rb, 2, x, y + 1, temp + 2, NULL); @@ -283,6 +286,7 @@ create_z_surface(XMesaContext xmctx, struct gl_renderbuffer *rb) xmsurf = CALLOC_STRUCT(xmesa_surface); if (xmsurf) { + xmsurf->sps.surface.format = PIPE_FORMAT_U_Z16; xmsurf->sps.surface.width = rb->Width; xmsurf->sps.surface.height = rb->Height; xmsurf->sps.read_quad_z = read_quad_z; diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index fcd6a22e1d..0b5d909b2d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -39,15 +39,38 @@ static void depth_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - GLuint j; struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); - GLfloat zzzz[QUAD_SIZE]; /**< Z for four pixels in quad */ + GLuint bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ + GLuint qzzzz[QUAD_SIZE]; /**< Z values from the quad */ GLuint zmask = 0; + GLuint j; + GLfloat scale; assert(sps); /* shouldn't get here if there's no zbuffer */ + /* + * To increase efficiency, we should probably have multiple versions + * of this function that are specifically for Z16, Z32 and FP Z buffers. + * Try to effectively do that with codegen... + */ + if (sps->surface.format == PIPE_FORMAT_U_Z16) + scale = 65535.0; + else + assert(0); /* XXX fix this someday */ + + /* + * Convert quad's float depth values to int depth values. + * If the Z buffer stores integer values, we _have_ to do the depth + * compares with integers (not floats). Otherwise, the float->int->float + * conversion of Z values (which isn't an identity function) will cause + * Z-fighting errors. + */ + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (GLuint) (quad->outputs.depth[j] * scale); + } + /* get zquad from zbuffer */ - sps->read_quad_z(sps, quad->x0, quad->y0, zzzz); + sps->read_quad_z(sps, quad->x0, quad->y0, bzzzz); switch (softpipe->depth_test.func) { case PIPE_FUNC_NEVER: @@ -57,19 +80,19 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) * Like this: quad->mask &= (quad->outputs.depth < zzzz); */ for (j = 0; j < QUAD_SIZE; j++) { - if (quad->outputs.depth[j] < zzzz[j]) + if (qzzzz[j] < bzzzz[j]) zmask |= 1 << j; } break; case PIPE_FUNC_EQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->outputs.depth[j] == zzzz[j]) + if (qzzzz[j] == bzzzz[j]) zmask |= 1 << j; } break; case PIPE_FUNC_LEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->outputs.depth[j] <= zzzz[j]) + if (qzzzz[j] <= bzzzz[j]) zmask |= (1 << j); } break; @@ -86,12 +109,12 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) */ for (j = 0; j < QUAD_SIZE; j++) { if (quad->mask & (1 << j)) { - zzzz[j] = quad->outputs.depth[j]; + bzzzz[j] = qzzzz[j]; } } /* write updated zquad to zbuffer */ - sps->write_quad_z(sps, quad->x0, quad->y0, zzzz); + sps->write_quad_z(sps, quad->x0, quad->y0, bzzzz); } qs->next->run(qs->next, quad); diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index ac66ffe891..450542abdd 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -79,9 +79,9 @@ struct softpipe_surface { GLubyte rgba[NUM_CHANNELS] ); void (*read_quad_z)(struct softpipe_surface *, - GLint x, GLint y, GLfloat zzzz[QUAD_SIZE]); + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]); void (*write_quad_z)(struct softpipe_surface *, - GLint x, GLint y, const GLfloat zzzz[QUAD_SIZE]); + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]); }; -- cgit v1.2.3 From 6aa9c8ebc19ffa4425669862caeb99b22c079090 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 14:36:10 -0600 Subject: remove unneeded casts, added comments --- src/mesa/pipe/draw/draw_cull.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index e7343cfd76..b6c578b3c9 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -64,19 +64,19 @@ static void cull_tri( struct prim_stage *stage, struct prim_header *header ) { /* Window coords: */ - GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); - GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); - GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); + const GLfloat *v0 = header->v[0]->data[0]; + const GLfloat *v1 = header->v[1]->data[0]; + const GLfloat *v2 = header->v[2]->data[0]; + /* edge vectors e = v0 - v2, f = v1 - v2 */ GLfloat ex = v0[0] - v2[0]; GLfloat ey = v0[1] - v2[1]; GLfloat fx = v1[0] - v2[0]; GLfloat fy = v1[1] - v2[1]; + /* det = cross(e,f).z */ header->det = ex * fy - ey * fx; - _mesa_printf("%s %f\n", __FUNCTION__, header->det ); - if (header->det != 0) { /* non-zero area */ GLuint mode = (header->det < 0) ? PIPE_WINDING_CW : PIPE_WINDING_CCW; -- cgit v1.2.3 From 13aa51de41a93fff1aa5050cebec2a6e0985b45a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 14:38:10 -0600 Subject: clamp after offsetting, new comments --- src/mesa/pipe/draw/draw_offset.c | 34 ++++++++++++++++++---------------- 1 file changed, 18 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 0fa8cf29d3..62ab2e4643 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -74,32 +74,34 @@ static void do_offset_tri( struct prim_stage *stage, /* Window coords: */ - GLfloat *v0 = (GLfloat *)&(header->v[0]->data[0]); - GLfloat *v1 = (GLfloat *)&(header->v[1]->data[0]); - GLfloat *v2 = (GLfloat *)&(header->v[2]->data[0]); - + GLfloat *v0 = header->v[0]->data[0]; + GLfloat *v1 = header->v[1]->data[0]; + GLfloat *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ GLfloat ex = v0[0] - v2[2]; - GLfloat fx = v1[0] - v2[2]; GLfloat ey = v0[1] - v2[2]; - GLfloat fy = v1[1] - v2[2]; GLfloat ez = v0[2] - v2[2]; + GLfloat fx = v1[0] - v2[2]; + GLfloat fy = v1[1] - v2[2]; GLfloat fz = v1[2] - v2[2]; + /* (a,b) = cross(e,f).xy */ GLfloat a = ey*fz - ez*fy; GLfloat b = ez*fx - ex*fz; - GLfloat ac = a * inv_det; - GLfloat bc = b * inv_det; - GLfloat zoffset; - - ac = FABSF(ac); - bc = FABSF(bc); + GLfloat dzdx = FABSF(a * inv_det); + GLfloat dzdy = FABSF(b * inv_det); - zoffset = offset->units + MAX2( ac, bc ) * offset->scale; + GLfloat zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; - v0[2] += zoffset; - v1[2] += zoffset; - v2[2] += zoffset; + /* + * Note: we're applying the offset and clamping per-vertex. + * Ideally, the offset is applied per-fragment prior to fragment shading. + */ + v0[2] = CLAMP(v0[2] + zoffset, 0.0, 1.0); + v1[2] = CLAMP(v1[2] + zoffset, 0.0, 1.0); + v2[2] = CLAMP(v2[2] + zoffset, 0.0, 1.0); stage->next->tri( stage->next, header ); } -- cgit v1.2.3 From a49a23efc5e320d8c9aa6f4f39be855632aa5cb8 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 15:24:31 -0600 Subject: Fix line clipping bug. New comments. Need to use a new prim_header for the post-clipped line. Otherwise, we were changing the header passed to us from the vb code. That messed up the vertex pointers for the next line primitive. --- src/mesa/pipe/draw/draw_clip.c | 39 +++++++++++++++++++++++---------------- 1 file changed, 23 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 10a6c1b823..a1c4cf01f0 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -69,7 +69,7 @@ static void interp_attr( GLfloat *fdst, /* Interpolate between two vertices to produce a third. */ -static void interp( struct clipper *clip, +static void interp( const struct clipper *clip, struct vertex_header *dst, GLfloat t, const struct vertex_header *out, @@ -98,9 +98,7 @@ static void interp( struct clipper *clip, const GLfloat *pos = dst->clip; const GLfloat *scale = clip->stage.draw->viewport.scale; const GLfloat *trans = clip->stage.draw->viewport.translate; - GLfloat oow; - - oow = 1.0 / pos[3]; + const GLfloat oow = 1.0 / pos[3]; dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; @@ -110,6 +108,9 @@ static void interp( struct clipper *clip, /* Other attributes + * Note: start at 1 to skip winpos (data[0]) and subtract one + * since there's two vertex attrib slots we want to ignore (the header + * and the clippos. */ for (j = 1; j < nr_attrs-1; j++) { interp_attr(dst->data[j], t, in->data[j], out->data[j]); @@ -295,13 +296,14 @@ do_clip_line( struct prim_stage *stage, struct prim_header *header, GLuint clipmask ) { - struct clipper *clipper = clipper_stage( stage ); + const struct clipper *clipper = clipper_stage( stage ); struct vertex_header *v0 = header->v[0]; struct vertex_header *v1 = header->v[1]; const GLfloat *pos0 = v0->clip; const GLfloat *pos1 = v1->clip; GLfloat t0 = 0; GLfloat t1 = 0; + struct prim_header newprim; /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove * this once we correctly use 16bit masks for userclip planes. @@ -313,43 +315,47 @@ do_clip_line( struct prim_stage *stage, } while (clipmask) { - GLuint plane_idx = ffs(clipmask)-1; + const GLuint plane_idx = ffs(clipmask)-1; const GLfloat *plane = clipper->plane[plane_idx]; - - clipmask &= ~(1< t1) t1 = t; + t1 = MIN2(t1, t); } if (dp0 < 0) { GLfloat t = dp0 / (dp0 - dp1); - if (t > t0) t0 = t; + t0 = MIN2(t0, t); } if (t0 + t1 >= 1.0) return; /* discard */ + + clipmask &= ~(1 << plane_idx); /* turn off this plane's bit */ } if (v0->clipmask) { interp( clipper, stage->tmp[0], t0, v0, v1 ); - header->v[0] = stage->tmp[0]; + newprim.v[0] = stage->tmp[0]; + } + else { + newprim.v[0] = v0; } if (v1->clipmask) { interp( clipper, stage->tmp[1], t1, v1, v0 ); - header->v[1] = stage->tmp[1]; + newprim.v[1] = stage->tmp[1]; + } + else { + newprim.v[1] = v1; } - stage->next->line( stage->next, header ); + stage->next->line( stage->next, &newprim ); } - static void clip_begin( struct prim_stage *stage ) { struct clipper *clipper = clipper_stage(stage); @@ -361,7 +367,8 @@ static void clip_begin( struct prim_stage *stage ) stage->next->begin( stage->next ); } - + + static void clip_point( struct prim_stage *stage, struct prim_header *header ) -- cgit v1.2.3 From 093d1b42d0d867dae1bcb59d36f3f309994badff Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 15:29:17 -0600 Subject: remove debug printf --- src/mesa/drivers/x11/xm_api.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index d1a635ee24..f20e8104fb 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -1836,8 +1836,6 @@ PUBLIC GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, XMesaBuffer readBuffer ) { - _mesa_printf("%s %p %p %p\n", __FUNCTION__, drawBuffer, readBuffer ); - if (c) { if (!drawBuffer || !readBuffer) return GL_FALSE; /* must specify buffers! */ -- cgit v1.2.3 From d015d2e0f45ad8e79ccb256b612597ef8ed51d4a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 16:25:43 -0600 Subject: Fill in remaining switch cases. Only call next stage if quad->mask != 0. --- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 57 +++++++++++++++++++---------- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 26 ++++++++++++- 2 files changed, 61 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index e0f7225d74..8c28a824be 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -15,8 +15,8 @@ static void alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - GLuint j; const GLfloat ref = softpipe->alpha_test.ref; + GLuint passMask = 0x0, j; switch (softpipe->alpha_test.func) { case PIPE_FUNC_NEVER: @@ -24,44 +24,61 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_FUNC_LESS: /* - * If quad->mask were an array [4] we could do this SIMD-style: - * quad->mask &= (quad->outputs.color[3] <= vec4(ref)); + * If mask were an array [4] we could do this SIMD-style: + * passMask = (quad->outputs.color[3] <= vec4(ref)); */ for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - if (quad->outputs.color[3][j] >= ref) { - /* fail */ - quad->mask &= (1 << j); - } + if (quad->outputs.color[3][j] < ref) { + passMask |= (1 << j); } } break; case PIPE_FUNC_EQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - if (quad->outputs.color[3][j] != ref) { - /* fail */ - quad->mask &= (1 << j); - } + if (quad->outputs.color[3][j] == ref) { + passMask |= (1 << j); } } break; case PIPE_FUNC_LEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - if (quad->outputs.color[3][j] > ref) { - /* fail */ - quad->mask &= (1 << j); - } + if (quad->outputs.color[3][j] <= ref) { + passMask |= (1 << j); } } break; - /* XXX fill in remaining cases */ + case PIPE_FUNC_GREATER: + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->outputs.color[3][j] > ref) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_NOTEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->outputs.color[3][j] != ref) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_GEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->outputs.color[3][j] >= ref) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_ALWAYS: + passMask = MASK_ALL; + break; default: abort(); } - qs->next->run(qs->next, quad); + quad->mask &= passMask; + + if (quad->mask) + qs->next->run(qs->next, quad); } diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 0b5d909b2d..268a1f947e 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -74,6 +74,7 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) switch (softpipe->depth_test.func) { case PIPE_FUNC_NEVER: + /* zmask = 0 */ break; case PIPE_FUNC_LESS: /* Note this is pretty much a single sse or cell instruction. @@ -96,7 +97,27 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) zmask |= (1 << j); } break; - /* XXX fill in remaining cases */ + case PIPE_FUNC_GREATER: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] > bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_NOTEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] != bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_GEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] >= bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_ALWAYS: + zmask = MASK_ALL; + break; default: abort(); } @@ -117,7 +138,8 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) sps->write_quad_z(sps, quad->x0, quad->y0, bzzzz); } - qs->next->run(qs->next, quad); + if (quad->mask) + qs->next->run(qs->next, quad); } -- cgit v1.2.3 From 227c95737bcf373f934bb96aa0777c45c8575074 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 16:37:03 -0600 Subject: clean-ups --- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 268a1f947e..f7dc5c877b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -23,16 +23,16 @@ */ /** - * quad blending + * \brief Quad depth testing */ -#include "glheader.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/imports.h" +#include "pipe/p_defines.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" #include "sp_quad.h" -#include "pipe/p_defines.h" static void -- cgit v1.2.3 From e6eca5c37e13fd0f9100de127075b1bbed0821c0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 16:37:18 -0600 Subject: Begin hooking up stenciling. --- src/mesa/pipe/softpipe/sp_context.c | 11 +-- src/mesa/pipe/softpipe/sp_context.h | 3 + src/mesa/pipe/softpipe/sp_quad.c | 7 +- src/mesa/pipe/softpipe/sp_quad.h | 1 + src/mesa/pipe/softpipe/sp_quad_stencil.c | 134 +++++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_state.h | 23 +++--- src/mesa/pipe/softpipe/sp_state_blend.c | 11 +++ src/mesa/sources | 1 + 8 files changed, 175 insertions(+), 16 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_quad_stencil.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 9becab4918..6afc0c1320 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -65,20 +65,21 @@ struct pipe_context *softpipe_create( void ) struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); softpipe->pipe.destroy = softpipe_destroy; - softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; - softpipe->pipe.set_blend_state = softpipe_set_blend_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; + softpipe->pipe.set_blend_state = softpipe_set_blend_state; softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_depth_state = softpipe_set_depth_test_state; - softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; - softpipe->pipe.set_setup_state = softpipe_set_setup_state; - softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; + softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_fs_state = softpipe_set_fs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.set_sampler_state = softpipe_set_sampler_state; + softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; + softpipe->pipe.set_setup_state = softpipe_set_setup_state; + softpipe->pipe.set_stencil_state = softpipe_set_stencil_state; softpipe->pipe.set_texture_state = softpipe_set_texture_state; + softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.clear = softpipe_clear; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index c6046624a1..2379a99b81 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -63,6 +63,7 @@ enum interp_mode { #define SP_NEW_DEPTH_TEST 0x200 #define SP_NEW_SAMPLER 0x400 #define SP_NEW_TEXTURE 0x800 +#define SP_NEW_STENCIL 0x1000 struct softpipe_context { @@ -82,6 +83,7 @@ struct softpipe_context { struct pipe_scissor_state scissor; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_setup_state setup; + struct pipe_stencil_state stencil; struct pipe_texture_object *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; GLuint dirty; @@ -117,6 +119,7 @@ struct softpipe_context { struct { struct quad_stage *shade; struct quad_stage *alpha_test; + struct quad_stage *stencil_test; struct quad_stage *depth_test; struct quad_stage *blend; struct quad_stage *output; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 63126a4da3..31278f529d 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -16,7 +16,12 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.blend; } - if (sp->depth_test.enabled) { + if ( sp->stencil.front_enabled + || sp->stencil.front_enabled) { + sp->quad.stencil_test->next = sp->quad.first; + sp->quad.first = sp->quad.stencil_test; + } + else if (sp->depth_test.enabled) { sp->quad.depth_test->next = sp->quad.first; sp->quad.first = sp->quad.depth_test; } diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 79f52721a3..c09905c249 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -48,6 +48,7 @@ struct quad_stage { struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c new file mode 100644 index 0000000000..9f59d09906 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -0,0 +1,134 @@ + +/** + * \brief Quad stencil testing + */ + + +#include "main/glheader.h" +#include "main/imports.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" +#include "pipe/p_defines.h" + + +static void +stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); + GLuint bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ + GLuint qzzzz[QUAD_SIZE]; /**< Z values from the quad */ + GLuint zmask = 0; + GLuint j; + GLfloat scale; + + assert(sps); /* shouldn't get here if there's no zbuffer */ + + /* + * To increase efficiency, we should probably have multiple versions + * of this function that are specifically for Z16, Z32 and FP Z buffers. + * Try to effectively do that with codegen... + */ + if (sps->surface.format == PIPE_FORMAT_U_Z16) + scale = 65535.0; + else + assert(0); /* XXX fix this someday */ + + /* + * Convert quad's float depth values to int depth values. + * If the Z buffer stores integer values, we _have_ to do the depth + * compares with integers (not floats). Otherwise, the float->int->float + * conversion of Z values (which isn't an identity function) will cause + * Z-fighting errors. + */ + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (GLuint) (quad->outputs.depth[j] * scale); + } + + /* get zquad from zbuffer */ + sps->read_quad_z(sps, quad->x0, quad->y0, bzzzz); + + switch (softpipe->depth_test.func) { + case PIPE_FUNC_NEVER: + /* zmask = 0 */ + break; + case PIPE_FUNC_LESS: + /* Note this is pretty much a single sse or cell instruction. + * Like this: quad->mask &= (quad->outputs.depth < zzzz); + */ + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] < bzzzz[j]) + zmask |= 1 << j; + } + break; + case PIPE_FUNC_EQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] == bzzzz[j]) + zmask |= 1 << j; + } + break; + case PIPE_FUNC_LEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] <= bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_GREATER: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] > bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_NOTEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] != bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_GEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] >= bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_ALWAYS: + zmask = MASK_ALL; + break; + default: + abort(); + } + + quad->mask &= zmask; + + if (softpipe->depth_test.writemask) { + + /* This is also efficient with sse / spe instructions: + */ + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + bzzzz[j] = qzzzz[j]; + } + } + + /* write updated zquad to zbuffer */ + sps->write_quad_z(sps, quad->x0, quad->y0, bzzzz); + } + + if (quad->mask) + qs->next->run(qs->next, quad); +} + + + + +struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = stencil_test_quad; + + return stage; +} diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 38a212b79a..71c1a2d9ba 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -56,8 +56,14 @@ void softpipe_set_clip_state( struct pipe_context *, void softpipe_set_depth_test_state( struct pipe_context *, const struct pipe_depth_state * ); -void softpipe_set_viewport_state( struct pipe_context *, - const struct pipe_viewport_state * ); +void softpipe_set_fs_state( struct pipe_context *, + const struct pipe_fs_state * ); + +void softpipe_set_polygon_stipple( struct pipe_context *, + const struct pipe_poly_stipple * ); + +void softpipe_set_scissor_state( struct pipe_context *, + const struct pipe_scissor_state * ); void softpipe_set_setup_state( struct pipe_context *, const struct pipe_setup_state * ); @@ -66,18 +72,15 @@ void softpipe_set_sampler_state( struct pipe_context *, GLuint unit, const struct pipe_sampler_state * ); +void softpipe_set_stencil_state( struct pipe_context *, + const struct pipe_stencil_state * ); + void softpipe_set_texture_state( struct pipe_context *, GLuint unit, struct pipe_texture_object * ); -void softpipe_set_scissor_state( struct pipe_context *, - const struct pipe_scissor_state * ); - -void softpipe_set_fs_state( struct pipe_context *, - const struct pipe_fs_state * ); - -void softpipe_set_polygon_stipple( struct pipe_context *, - const struct pipe_poly_stipple * ); +void softpipe_set_viewport_state( struct pipe_context *, + const struct pipe_viewport_state * ); void softpipe_update_derived( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index ae8950e386..8bc22b0efc 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -80,3 +80,14 @@ softpipe_set_alpha_test_state(struct pipe_context *pipe, softpipe->dirty |= SP_NEW_ALPHA_TEST; } +void +softpipe_set_stencil_state(struct pipe_context *pipe, + const struct pipe_stencil_state *stencil) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->stencil = *stencil; + + softpipe->dirty |= SP_NEW_STENCIL; +} + diff --git a/src/mesa/sources b/src/mesa/sources index 5bef7606d4..1ffac8df0a 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -163,6 +163,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_quad_depth_test.c \ pipe/softpipe/sp_quad_fs.c \ pipe/softpipe/sp_quad_output.c \ + pipe/softpipe/sp_quad_stencil.c \ pipe/softpipe/sp_state_blend.c \ pipe/softpipe/sp_state_clip.c \ pipe/softpipe/sp_state_derived.c \ -- cgit v1.2.3 From 73daa688541ec88119804ad190ce5b429e50ea44 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 10 Jul 2007 18:59:17 -0600 Subject: Checkpoint: stencil roughly working, some bugs to fix... --- src/mesa/drivers/x11/xm_surface.c | 65 ++++++- src/mesa/pipe/p_defines.h | 5 +- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_quad.h | 1 + src/mesa/pipe/softpipe/sp_quad_depth_test.c | 15 +- src/mesa/pipe/softpipe/sp_quad_stencil.c | 277 +++++++++++++++++++++------- src/mesa/pipe/softpipe/sp_surface.h | 5 + 7 files changed, 303 insertions(+), 66 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 5158e42d9a..2bef5e6b9e 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -244,6 +244,8 @@ xmesa_get_color_surface(GLcontext *ctx, GLuint buf) } + + static void read_quad_z(struct softpipe_surface *sps, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) @@ -296,6 +298,51 @@ create_z_surface(XMesaContext xmctx, struct gl_renderbuffer *rb) return xmsurf; } + + + +static void +read_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(sps); + struct gl_renderbuffer *rb = xmsurf->rb; + GET_CURRENT_CONTEXT(ctx); + rb->GetRow(ctx, rb, 2, x, y, ssss); + rb->GetRow(ctx, rb, 2, x, y + 1, ssss + 2); +} + +static void +write_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) +{ + struct xmesa_surface *xmsurf = xmesa_surface(sps); + struct gl_renderbuffer *rb = xmsurf->rb; + GET_CURRENT_CONTEXT(ctx); + rb->PutRow(ctx, rb, 2, x, y, ssss, NULL); + rb->PutRow(ctx, rb, 2, x, y + 1, ssss + 2, NULL); +} + +static struct xmesa_surface * +create_stencil_surface(XMesaContext xmctx, struct gl_renderbuffer *rb) +{ + struct xmesa_surface *xmsurf; + + xmsurf = CALLOC_STRUCT(xmesa_surface); + if (xmsurf) { + xmsurf->sps.surface.format = PIPE_FORMAT_S8; + xmsurf->sps.surface.width = rb->Width; + xmsurf->sps.surface.height = rb->Height; + xmsurf->sps.read_quad_stencil = read_quad_stencil; + xmsurf->sps.write_quad_stencil = write_quad_stencil; + xmsurf->rb = rb; + } + return xmsurf; +} + + + + /** * Return a pipe_surface that wraps the current Z/depth buffer. * XXX this is pretty much a total hack until gl_renderbuffers and @@ -327,6 +374,22 @@ xmesa_get_z_surface(GLcontext *ctx) struct pipe_surface * xmesa_get_stencil_surface(GLcontext *ctx) { - return NULL; + XMesaContext xmctx = XMESA_CONTEXT(ctx); + struct gl_renderbuffer *rb = ctx->DrawBuffer->_StencilBuffer; + static struct xmesa_surface *xms = NULL; + + if (!rb) + return NULL; + + if (!xms) { + xms = create_stencil_surface(xmctx, rb); + } + else if (xms->sps.surface.width != rb->Width || + xms->sps.surface.height != rb->Height) { + free_surface(&xms->sps); + xms = create_stencil_surface(xmctx, rb); + } + + return (struct pipe_surface *) &xms->sps.surface; } diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 7212040b36..fbdd015619 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -125,7 +125,7 @@ #define PIPE_TEX_COMPARE_R_TO_TEXTURE 1 /** - * Texture/surface image formats + * Texture/surface image formats (preliminary) */ #define PIPE_FORMAT_U_R8_G8_B8_A8 0 /**< ubyte[4] RGBA */ #define PIPE_FORMAT_U_A8_R8_G8_B8 1 /**< ubyte[4] ARGB */ @@ -138,6 +138,9 @@ #define PIPE_FORMAT_F_Z32 8 /**< float Z/depth */ #define PIPE_FORMAT_YCBCR 9 #define PIPE_FORMAT_YCBCR_REV 10 +#define PIPE_FORMAT_S8 11 /**< 8-bit stencil */ +#define PIPE_FORMAT_Z24_S8 12 /**< 24-bit Z + 8-bit stencil */ + /** * Texture typess diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 6afc0c1320..50434600c3 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -87,6 +87,7 @@ struct pipe_context *softpipe_create( void ) softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); softpipe->quad.blend = sp_quad_blend_stage(softpipe); softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); + softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); /* diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index c09905c249..df416dfa7f 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -55,5 +55,6 @@ struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); void sp_build_quad_pipeline(struct softpipe_context *sp); +void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad); #endif /* SP_TILE_H */ diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index f7dc5c877b..d47c4c42b8 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -35,8 +35,12 @@ #include "sp_quad.h" -static void -depth_test_quad(struct quad_stage *qs, struct quad_header *quad) +/** + * Do depth testing for a quad. + * Not static since it's used by the stencil code. + */ +void +sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); @@ -137,6 +141,13 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) /* write updated zquad to zbuffer */ sps->write_quad_z(sps, quad->x0, quad->y0, bzzzz); } +} + + +static void +depth_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ + sp_depth_test_quad(qs, quad); if (quad->mask) qs->next->run(qs->next, quad); diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 9f59d09906..5b59addce2 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -13,108 +13,261 @@ #include "pipe/p_defines.h" -static void -stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); - GLuint bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ - GLuint qzzzz[QUAD_SIZE]; /**< Z values from the quad */ - GLuint zmask = 0; - GLuint j; - GLfloat scale; +/** Only 8-bit stencil supported */ +#define STENCIL_MAX 0xff - assert(sps); /* shouldn't get here if there's no zbuffer */ - /* - * To increase efficiency, we should probably have multiple versions - * of this function that are specifically for Z16, Z32 and FP Z buffers. - * Try to effectively do that with codegen... - */ - if (sps->surface.format == PIPE_FORMAT_U_Z16) - scale = 65535.0; - else - assert(0); /* XXX fix this someday */ - - /* - * Convert quad's float depth values to int depth values. - * If the Z buffer stores integer values, we _have_ to do the depth - * compares with integers (not floats). Otherwise, the float->int->float - * conversion of Z values (which isn't an identity function) will cause - * Z-fighting errors. - */ - for (j = 0; j < QUAD_SIZE; j++) { - qzzzz[j] = (GLuint) (quad->outputs.depth[j] * scale); - } +/** + * Do the basic stencil test (compare stencil buffer values against the + * reference value. + * + * \param stencilVals the stencil values from the stencil buffer + * \param func the stencil func (PIPE_FUNC_x) + * \param ref the stencil reference value + * \param valMask the stencil value mask indicating which bits of the stencil + * values and ref value are to be used. + * \return mask indicating which pixels passed the stencil test + */ +static GLbitfield +do_stencil_test(const GLubyte stencilVals[QUAD_SIZE], GLuint func, + GLbitfield ref, GLbitfield valMask) +{ + GLbitfield passMask = 0x0; + GLuint j; - /* get zquad from zbuffer */ - sps->read_quad_z(sps, quad->x0, quad->y0, bzzzz); + ref &= valMask; - switch (softpipe->depth_test.func) { + switch (func) { case PIPE_FUNC_NEVER: - /* zmask = 0 */ + /* passMask = 0x0 */ break; case PIPE_FUNC_LESS: - /* Note this is pretty much a single sse or cell instruction. - * Like this: quad->mask &= (quad->outputs.depth < zzzz); - */ for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] < bzzzz[j]) - zmask |= 1 << j; + if ((stencilVals[j] & valMask) < ref) { + passMask |= (1 << j); + } } break; case PIPE_FUNC_EQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] == bzzzz[j]) - zmask |= 1 << j; + if ((stencilVals[j] & valMask) == ref) { + passMask |= (1 << j); + } } break; case PIPE_FUNC_LEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] <= bzzzz[j]) - zmask |= (1 << j); + if ((stencilVals[j] & valMask) <= ref) { + passMask |= (1 << j); + } } break; case PIPE_FUNC_GREATER: for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] > bzzzz[j]) - zmask |= (1 << j); + if ((stencilVals[j] & valMask) > ref) { + passMask |= (1 << j); + } } break; case PIPE_FUNC_NOTEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] != bzzzz[j]) - zmask |= (1 << j); + if ((stencilVals[j] & valMask) != ref) { + passMask |= (1 << j); + } } break; case PIPE_FUNC_GEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] >= bzzzz[j]) - zmask |= (1 << j); + if ((stencilVals[j] & valMask) >= ref) { + passMask |= (1 << j); + } } break; case PIPE_FUNC_ALWAYS: - zmask = MASK_ALL; + passMask = MASK_ALL; break; default: - abort(); + assert(0); } - quad->mask &= zmask; + return passMask; +} + + +/** + * Apply the stencil operator to stencil values. + * + * \param stencilVals the stencil buffer values (read and written) + * \param mask indicates which pixels to update + * \param op the stencil operator (PIPE_STENCIL_OP_x) + * \param ref the stencil reference value + * \param wrtMask writemask controlling which bits are changed in the + * stencil values + */ +static void +apply_stencil_op(GLubyte stencilVals[QUAD_SIZE], + GLbitfield mask, GLuint op, GLubyte ref, GLubyte wrtMask) +{ + GLuint j; + GLubyte newstencil[QUAD_SIZE]; + + for (j = 0; j < QUAD_SIZE; j++) { + newstencil[j] = stencilVals[j]; + } - if (softpipe->depth_test.writemask) { - - /* This is also efficient with sse / spe instructions: - */ + switch (op) { + case PIPE_STENCIL_OP_KEEP: + /* no-op */ + break; + case PIPE_STENCIL_OP_ZERO: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - bzzzz[j] = qzzzz[j]; - } + if (mask & (1 << j)) { + newstencil[j] = 0; + } } + break; + case PIPE_STENCIL_OP_REPLACE: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + newstencil[j] = ref; + } + } + break; + case PIPE_STENCIL_OP_INCR: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + if (stencilVals[j] < STENCIL_MAX) { + newstencil[j] = stencilVals[j] + 1; + } + } + } + break; + case PIPE_STENCIL_OP_DECR: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + if (stencilVals[j] > 0) { + newstencil[j] = stencilVals[j] - 1; + } + } + } + break; + case PIPE_STENCIL_OP_INCR_WRAP: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + newstencil[j] = stencilVals[j] + 1; + } + } + break; + case PIPE_STENCIL_OP_DECR_WRAP: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + newstencil[j] = stencilVals[j] - 1; + } + } + break; + case PIPE_STENCIL_OP_INVERT: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + newstencil[j] = ~stencilVals[j]; + } + } + break; + default: + assert(0); + } - /* write updated zquad to zbuffer */ - sps->write_quad_z(sps, quad->x0, quad->y0, bzzzz); + /* + * update the stencil values + */ + if (wrtMask != STENCIL_MAX) { + /* apply bit-wise stencil buffer writemask */ + for (j = 0; j < QUAD_SIZE; j++) { + stencilVals[j] = (wrtMask & newstencil[j]) | (~wrtMask & stencilVals[j]); + } } + else { + for (j = 0; j < QUAD_SIZE; j++) { + stencilVals[j] = newstencil[j]; + } + } +} + + +/** + * Do stencil (and depth) testing. Stenciling depends on the outcome of + * depth testing. + */ +static void +stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + struct softpipe_surface *s_surf = softpipe_surface(softpipe->framebuffer.sbuf); + GLuint func, zFailOp, zPassOp, failOp; + GLuint face = 0; + GLubyte ref, wrtMask, valMask; + GLubyte stencilVals[QUAD_SIZE]; + + /* choose front or back face function, operator, etc */ + if (softpipe->stencil.back_enabled && face == 1) { + func = softpipe->stencil.back_func; + failOp = softpipe->stencil.back_fail_op; + zFailOp = softpipe->stencil.back_zfail_op; + zPassOp = softpipe->stencil.back_zpass_op; + ref = softpipe->stencil.ref_value[1]; + wrtMask = softpipe->stencil.write_mask[1]; + valMask = softpipe->stencil.value_mask[1]; + } + else { + func = softpipe->stencil.front_func; + failOp = softpipe->stencil.front_fail_op; + zFailOp = softpipe->stencil.front_zfail_op; + zPassOp = softpipe->stencil.front_zpass_op; + ref = softpipe->stencil.ref_value[0]; + wrtMask = softpipe->stencil.write_mask[0]; + valMask = softpipe->stencil.value_mask[0]; + } + + assert(s_surf); /* shouldn't get here if there's no stencil buffer */ + s_surf->read_quad_stencil(s_surf, quad->x0, quad->y0, stencilVals); + + /* do the stencil test first */ + { + GLbitfield passMask, failMask; + passMask = do_stencil_test(stencilVals, func, ref, valMask); + failMask = quad->mask & ~passMask; + quad->mask &= passMask; + + if (failOp != PIPE_STENCIL_OP_KEEP) { + apply_stencil_op(stencilVals, failMask, failOp, ref, wrtMask); + } + } + + if (!quad->mask) + return; + + /* now the pixels that passed the stencil test are depth tested */ + if (softpipe->depth_test.enabled) { + const GLbitfield origMask = quad->mask; + + sp_depth_test_quad(qs, quad); /* quad->mask is updated */ + + /* update stencil buffer values according to z pass/fail result */ + if (zFailOp != PIPE_STENCIL_OP_KEEP) { + const GLbitfield failMask = origMask & ~quad->mask; + apply_stencil_op(stencilVals, failMask, zFailOp, ref, wrtMask); + } + + if (zPassOp != PIPE_STENCIL_OP_KEEP) { + const GLbitfield passMask = origMask & quad->mask; + apply_stencil_op(stencilVals, passMask, zPassOp, ref, wrtMask); + } + } + else { + /* no depth test, apply Zpass operator to stencil buffer values */ + apply_stencil_op(stencilVals, quad->mask, zPassOp, ref, wrtMask); + } + + s_surf->write_quad_stencil(s_surf, quad->x0, quad->y0, stencilVals); if (quad->mask) qs->next->run(qs->next, quad); diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 450542abdd..3ba732cebe 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -82,6 +82,11 @@ struct softpipe_surface { GLint x, GLint y, GLuint zzzz[QUAD_SIZE]); void (*write_quad_z)(struct softpipe_surface *, GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]); + + void (*read_quad_stencil)(struct softpipe_surface *, + GLint x, GLint y, GLubyte ssss[QUAD_SIZE]); + void (*write_quad_stencil)(struct softpipe_surface *, + GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]); }; -- cgit v1.2.3 From 43e24ff50773575d28763d899a4f25bb430418b5 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 11 Jul 2007 14:20:39 +0200 Subject: some fixes, fake frontbuffer still doesn't work quite right (resize). Fake frontbuffer doesn't copy in real frontbuffer. Don't even think about doing rotation/page flip/triple buffering for now... More cleanups needed (fake cliprects etc.) --- src/mesa/drivers/dri/i915tex/i915_vtbl.c | 2 ++ src/mesa/drivers/dri/i915tex/intel_blit.c | 18 ++++++++++++------ src/mesa/drivers/dri/i915tex/intel_buffers.c | 23 ++++++++++++----------- src/mesa/drivers/dri/i915tex/intel_context.c | 18 ++++++++++++------ src/mesa/drivers/dri/i915tex/intel_screen.c | 14 +++++++++++++- src/mesa/drivers/dri/i915tex/intel_state.c | 24 ++++++++---------------- 6 files changed, 59 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/i915_vtbl.c b/src/mesa/drivers/dri/i915tex/i915_vtbl.c index f80e8d6327..51aff68840 100644 --- a/src/mesa/drivers/dri/i915tex/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915tex/i915_vtbl.c @@ -463,6 +463,7 @@ i915_state_draw_region(struct intel_context *intel, * Set stride/cpp values */ if (color_region) { +// fprintf(stderr, "color pitch %d\n", color_region->pitch); state->Buffer[I915_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; state->Buffer[I915_DESTREG_CBUFADDR1] = (BUF_3D_ID_COLOR_BACK | @@ -471,6 +472,7 @@ i915_state_draw_region(struct intel_context *intel, } if (depth_region) { +// fprintf(stderr, "depth pitch %d\n", depth_region->pitch); state->Buffer[I915_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD; state->Buffer[I915_DESTREG_DBUFADDR1] = (BUF_3D_ID_DEPTH | diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c index b4fc1691e6..d6651f62be 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ b/src/mesa/drivers/dri/i915tex/intel_blit.c @@ -82,9 +82,11 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, if (dPriv && dPriv->numClipRects) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; const struct intel_region *frontRegion - = intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT); + = intelScreen->front_region; const struct intel_region *backRegion - = intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); + = intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? + intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : + intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); const int nbox = dPriv->numClipRects; const drm_clip_rect_t *pbox = dPriv->pClipRects; const int pitch = frontRegion->pitch; @@ -100,7 +102,7 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, // ASSERT(frontRegion->pitch == backRegion->pitch); ASSERT(frontRegion->cpp == backRegion->cpp); - DBG("copy buffer, front pitch %d back pitch %d\n", + DBG("front pitch %d back pitch %d\n", frontRegion->pitch, backRegion->pitch); if (cpp == 2) { @@ -138,6 +140,11 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, continue; } + DBG("box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + /* XXX should make sure only the minimum area based on + old draw buffer and new front clip rects is copied */ sbox.x1 = box.x1 - dPriv->x; sbox.y1 = box.y1 - dPriv->y; @@ -417,9 +424,8 @@ intelClearWithBlit(GLcontext * ctx, GLbitfield mask) b = *box; } - if (0) - _mesa_printf("clear %d,%d..%d,%d, mask %x\n", - b.x1, b.y1, b.x2, b.y2, mask); + DBG("clear %d,%d..%d,%d, mask %x\n", + b.x1, b.y1, b.x2, b.y2, mask); /* Loop over all renderbuffers */ for (buf = 0; buf < BUFFER_COUNT && clearMask; buf++) { diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index e8dd7cd524..fbaa4ee4f3 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -930,7 +930,7 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) GLboolean missed_target; struct intel_framebuffer *intel_fb = dPriv->driverPrivate; int64_t ust; - + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ if (screen->current_rotation != 0 || @@ -956,6 +956,16 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) intel_fb->swap_ust = ust; } + if (dPriv && intel->lastStamp != dPriv->lastStamp) { + /* XXX this doesn't appear to work quite right. + And in any case, it will never get called with single buffered + rendering here... + And if it's only a window move (not resize), don't need to do anything. */ + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("doing defered drawable update\n"); + intelWindowMoved(intel); + intel->lastStamp = dPriv->lastStamp; + } } else { /* XXX this shouldn't be an error but we can't handle it for now */ @@ -1068,21 +1078,12 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) * And set up cliprects. */ if (fb->Name == 0) { + intelSetPrivbufClipRects(intel); /* drawing to window system buffer */ if (front) { -#if 0 - intelSetFrontClipRects(intel); -#else - intelSetPrivbufClipRects(intel); -#endif colorRegion = intel_get_rb_region(fb, BUFFER_FRONT_LEFT); } else { -#if 0 - intelSetBackClipRects(intel); -#else - intelSetPrivbufClipRects(intel); -#endif colorRegion = intel_get_rb_region(fb, BUFFER_BACK_LEFT); } } diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 028e57e5ab..76426e3c88 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -282,18 +282,22 @@ intelFlush(GLcontext * ctx) * Check if we need to rotate/warp the front color buffer to the * rotated screen. We generally need to do this when we get a glFlush * or glFinish after drawing to the front color buffer. + * If no rotation, just copy the private fake front buffer to the real one. */ static void -intelCheckFrontRotate(GLcontext * ctx) +intelCheckFrontUpdate(GLcontext * ctx) { struct intel_context *intel = intel_context(ctx); if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT) { intelScreenPrivate *screen = intel->intelScreen; + __DRIdrawablePrivate *dPriv = intel->driDrawable; if (screen->current_rotation != 0) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT); } + else { + intelCopyBuffer(dPriv, NULL); + } } } @@ -305,7 +309,7 @@ static void intelglFlush(GLcontext * ctx) { intelFlush(ctx); - intelCheckFrontRotate(ctx); + intelCheckFrontUpdate(ctx); } void @@ -319,7 +323,7 @@ intelFinish(GLcontext * ctx) driFenceUnReference(intel->batch->last_fence); intel->batch->last_fence = NULL; } - intelCheckFrontRotate(ctx); + intelCheckFrontUpdate(ctx); } @@ -724,8 +728,10 @@ intelContendedLock(struct intel_context *intel, GLuint flags) /* Drawable changed? */ if (dPriv && intel->lastStamp != dPriv->lastStamp) { - intelWindowMoved(intel); - intel->lastStamp = dPriv->lastStamp; + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("drawable change detected but defering update\n"); +/* intelWindowMoved(intel); + intel->lastStamp = dPriv->lastStamp;*/ } } diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c index a5a9c79625..82b65ea842 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ b/src/mesa/drivers/dri/i915tex/intel_screen.c @@ -172,6 +172,7 @@ intel_recreate_static(intelScreenPrivate *intelScreen, static void intel_recreate_static_regions(intelScreenPrivate *intelScreen) { +/* this is the real front buffer which is only used for blitting to */ intelScreen->front_region = intel_recreate_static(intelScreen, intelScreen->front_region, @@ -601,6 +602,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); +#if 0 /* setup the hardware-based renderbuffers */ { intel_fb->color_rb[0] @@ -615,7 +617,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, &intel_fb->color_rb[0]->Base); } -#if 0 if (mesaVis->doubleBufferMode) { intel_fb->color_rb[1] = intel_create_renderbuffer(rgbFormat, @@ -672,11 +673,22 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, } #else + { + /* fake frontbuffer */ + /* XXX allocation should only happen in the unusual case + it's actually needed */ + intel_fb->color_rb[0] + = intel_new_renderbuffer_fb(NULL, rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, + &intel_fb->color_rb[0]->Base); + } + if (mesaVis->doubleBufferMode) { intel_fb->color_rb[1] = intel_new_renderbuffer_fb(NULL, rgbFormat); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, &intel_fb->color_rb[1]->Base); + if (screen->third.handle) { struct gl_renderbuffer *tmp_rb = NULL; diff --git a/src/mesa/drivers/dri/i915tex/intel_state.c b/src/mesa/drivers/dri/i915tex/intel_state.c index 271511037e..b2773990e4 100644 --- a/src/mesa/drivers/dri/i915tex/intel_state.c +++ b/src/mesa/drivers/dri/i915tex/intel_state.c @@ -205,25 +205,17 @@ intelCalcViewport(GLcontext * ctx) GLfloat *m = intel->ViewportMatrix.m; GLfloat yScale, yBias; - if (ctx->DrawBuffer->Name) { - /* User created FBO */ - struct intel_renderbuffer *irb + struct intel_renderbuffer *irb = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); - if (irb && !irb->RenderToTexture) { - /* y=0=top */ - yScale = -1.0; - yBias = irb->Base.Height; - } - else { - /* y=0=bottom */ - yScale = 1.0; - yBias = 0.0; - } + if (irb && !irb->RenderToTexture) { + /* y=0=top */ + yScale = -1.0; + yBias = irb->Base.Height; } else { - /* window buffer, y=0=top */ - yScale = -1.0; - yBias = (intel->driDrawable) ? intel->driDrawable->h : 0.0F; + /* y=0=bottom */ + yScale = 1.0; + yBias = 0.0; } m[MAT_SX] = v[MAT_SX]; -- cgit v1.2.3 From eb2a896f68e5402764955a30b949d8408d435e73 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 11 Jul 2007 16:08:14 +0200 Subject: drop some more code no longer needed --- src/mesa/drivers/dri/i915tex/intel_buffers.c | 1 + src/mesa/drivers/dri/i915tex/intel_context.c | 2 ++ src/mesa/drivers/dri/i915tex/intel_screen.c | 2 ++ 3 files changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index fbaa4ee4f3..4e4f0d14c6 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -500,6 +500,7 @@ void intelRotateWindow(struct intel_context *intel, __DRIdrawablePrivate * dPriv, GLuint srcBuf) { + intelScreenPrivate *screen = intel->intelScreen; drm_clip_rect_t fullRect; struct intel_framebuffer *intel_fb; diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 76426e3c88..4b96cd7b7e 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -682,6 +682,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) intelUpdateScreenRotation(sPriv, sarea); } +#if 0 if (sarea->width != intel->width || sarea->height != intel->height || sarea->rotation != intel->current_rotation) { @@ -724,6 +725,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) intel->height = sarea->height; intel->current_rotation = sarea->rotation; } +#endif /* Drawable changed? */ diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c index 82b65ea842..08057a6600 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ b/src/mesa/drivers/dri/i915tex/intel_screen.c @@ -88,6 +88,7 @@ intelMapScreenRegions(__DRIscreenPrivate * sPriv) _mesa_warning(NULL, "no front buffer handle in intelMapScreenRegions!"); } +#if 0 if (0) _mesa_printf("Back 0x%08x ", intelScreen->back.handle); if (drmMap(sPriv->fd, @@ -119,6 +120,7 @@ intelMapScreenRegions(__DRIscreenPrivate * sPriv) intelUnmapScreenRegions(intelScreen); return GL_FALSE; } +#endif #if 0 _mesa_printf("TEX 0x%08x ", intelScreen->tex.handle); -- cgit v1.2.3 From fb67b1609e211ed16828e10ffc1c36e54f655112 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 11 Jul 2007 17:30:01 +0200 Subject: remove bogus assertion, change drawable updates don't update driDrawable information until later at all (blit needs fixing). --- src/mesa/drivers/dri/i915tex/intel_buffers.c | 17 ++++++++++++----- src/mesa/drivers/dri/i915tex/intel_context.c | 5 ++++- src/mesa/drivers/dri/i915tex/intel_context.h | 1 + src/mesa/drivers/dri/i915tex/intel_fbo.c | 4 +--- 4 files changed, 18 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 4e4f0d14c6..67f992e4a2 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -957,15 +957,22 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) intel_fb->swap_ust = ust; } - if (dPriv && intel->lastStamp != dPriv->lastStamp) { + if (intel->revalidateDrawable) { + __DRIscreenPrivate *sPriv = intel->driScreen; + LOCK_HARDWARE(intel); + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + intel->revalidateDrawable = GL_FALSE; + UNLOCK_HARDWARE(intel); + if (dPriv && intel->lastStamp != dPriv->lastStamp) { /* XXX this doesn't appear to work quite right. And in any case, it will never get called with single buffered rendering here... And if it's only a window move (not resize), don't need to do anything. */ - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("doing defered drawable update\n"); - intelWindowMoved(intel); - intel->lastStamp = dPriv->lastStamp; + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("doing defered drawable update\n"); + intelWindowMoved(intel); + intel->lastStamp = dPriv->lastStamp; + } } } else { diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 4b96cd7b7e..aa2078e762 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -673,7 +673,8 @@ intelContendedLock(struct intel_context *intel, GLuint flags) * checking must be done *after* this call: */ if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + intel->revalidateDrawable = GL_TRUE; +// DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); if (sarea->width != intelScreen->width || sarea->height != intelScreen->height || @@ -727,6 +728,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) } #endif +#if 0 /* Drawable changed? */ if (dPriv && intel->lastStamp != dPriv->lastStamp) { @@ -735,6 +737,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) /* intelWindowMoved(intel); intel->lastStamp = dPriv->lastStamp;*/ } +#endif } diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index d05b20abbd..7b97c564a3 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -273,6 +273,7 @@ struct intel_context drmI830Sarea *sarea; GLuint lastStamp; + GLuint revalidateDrawable; /** * Configuration cache diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c index d3f14bcd25..2dd3617f93 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c @@ -176,7 +176,7 @@ intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb, /** * Called via glRenderbufferStorageEXT() to set the format and allocate - * storage for a user-created renderbuffer. + * storage for a user-created (or priv buffer) renderbuffer. */ static GLboolean intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, @@ -188,8 +188,6 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, GLboolean softwareBuffer = GL_FALSE; int cpp; - ASSERT(rb->Name != 0); - switch (internalFormat) { case GL_R3_G3_B2: case GL_RGB4: -- cgit v1.2.3 From 0ac0fb91bd711ec80a058ab23bfe8011baa0a487 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 11 Jul 2007 09:52:00 -0600 Subject: Compute quad.facing from prim->det and polygon winding. Updated comments/questions about area vs. prim->det. --- src/mesa/pipe/softpipe/sp_headers.h | 2 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 22 ++++++++++++++-------- 2 files changed, 15 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index 82f03ec078..cd2dc984c5 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -74,7 +74,7 @@ struct quad_header { GLint x0; GLint y0; GLuint mask; - GLuint facing; /**< Front or back facing? */ + GLuint facing; /**< Front (0) or back (1) facing? */ struct { GLfloat color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 8e43f5d04d..332113979c 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -279,26 +279,32 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; - /* xxx: may need to adjust this sign according to the if-tree - * above: + /* + * Compute triangle's area. Use 1/area to compute partial + * derivatives of attributes later. * - * XXX: this is like 'det', but calculated from screen coords?? + * The area will be the same as prim->det, but the sign may be + * different depending on how the vertices get sorted above. + * + * To determine whether the primitive is front or back facing we + * use the prim->det value because its sign is correct. */ { const GLfloat area = (setup->emaj.dx * setup->ebot.dy - setup->ebot.dx * setup->emaj.dy); setup->oneoverarea = 1.0 / area; + /* + _mesa_printf("%s one-over-area %f area %f det %f\n", + __FUNCTION__, setup->oneoverarea, area, prim->det ); + */ } - /* XXX need to know if this is a front or back-facing triangle: + /* We need to know if this is a front or back-facing triangle for: * - the GLSL gl_FrontFacing fragment attribute (bool) * - two-sided stencil test */ - setup->quad.facing = 0; - - _mesa_printf("%s one-over-area %f\n", __FUNCTION__, setup->oneoverarea ); - + setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->setup.front_winding == PIPE_WINDING_CW); return GL_TRUE; } -- cgit v1.2.3 From d6aa360eddec0463f3563eff07b06c5d60037826 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 11 Jul 2007 09:52:32 -0600 Subject: fix stencil logic error, use proper front/back face value --- src/mesa/pipe/softpipe/sp_quad_stencil.c | 41 ++++++++++++++++---------------- 1 file changed, 21 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 5b59addce2..0b37474c1a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -203,12 +203,12 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *s_surf = softpipe_surface(softpipe->framebuffer.sbuf); GLuint func, zFailOp, zPassOp, failOp; - GLuint face = 0; GLubyte ref, wrtMask, valMask; GLubyte stencilVals[QUAD_SIZE]; /* choose front or back face function, operator, etc */ - if (softpipe->stencil.back_enabled && face == 1) { + /* XXX we could do these initializations once per primitive */ + if (softpipe->stencil.back_enabled && quad->facing) { func = softpipe->stencil.back_func; failOp = softpipe->stencil.back_fail_op; zFailOp = softpipe->stencil.back_zfail_op; @@ -242,29 +242,30 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) } } - if (!quad->mask) - return; + if (quad->mask) { - /* now the pixels that passed the stencil test are depth tested */ - if (softpipe->depth_test.enabled) { - const GLbitfield origMask = quad->mask; + /* now the pixels that passed the stencil test are depth tested */ + if (softpipe->depth_test.enabled) { + const GLbitfield origMask = quad->mask; - sp_depth_test_quad(qs, quad); /* quad->mask is updated */ + sp_depth_test_quad(qs, quad); /* quad->mask is updated */ - /* update stencil buffer values according to z pass/fail result */ - if (zFailOp != PIPE_STENCIL_OP_KEEP) { - const GLbitfield failMask = origMask & ~quad->mask; - apply_stencil_op(stencilVals, failMask, zFailOp, ref, wrtMask); - } + /* update stencil buffer values according to z pass/fail result */ + if (zFailOp != PIPE_STENCIL_OP_KEEP) { + const GLbitfield failMask = origMask & ~quad->mask; + apply_stencil_op(stencilVals, failMask, zFailOp, ref, wrtMask); + } - if (zPassOp != PIPE_STENCIL_OP_KEEP) { - const GLbitfield passMask = origMask & quad->mask; - apply_stencil_op(stencilVals, passMask, zPassOp, ref, wrtMask); + if (zPassOp != PIPE_STENCIL_OP_KEEP) { + const GLbitfield passMask = origMask & quad->mask; + apply_stencil_op(stencilVals, passMask, zPassOp, ref, wrtMask); + } } - } - else { - /* no depth test, apply Zpass operator to stencil buffer values */ - apply_stencil_op(stencilVals, quad->mask, zPassOp, ref, wrtMask); + else { + /* no depth test, apply Zpass operator to stencil buffer values */ + apply_stencil_op(stencilVals, quad->mask, zPassOp, ref, wrtMask); + } + } s_surf->write_quad_stencil(s_surf, quad->x0, quad->y0, stencilVals); -- cgit v1.2.3 From 300e97081e7e752c0ff9133149d15935baac7a46 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 11 Jul 2007 09:56:46 -0600 Subject: need to clamp floats to [0,1] in write functions --- src/mesa/drivers/x11/xm_surface.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 2bef5e6b9e..8ae5c248f4 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -122,7 +122,7 @@ write_quad_f(struct softpipe_surface *gs, GLint x, GLint y, GLuint i; GET_CURRENT_CONTEXT(ctx); for (i = 0; i < 16; i++) { - temp[i] = FLOAT_TO_UBYTE(src[i]); + UNCLAMPED_FLOAT_TO_UBYTE(temp[i], src[i]); } xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y, temp, NULL); xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y + 1, temp + 8, NULL); @@ -140,7 +140,7 @@ write_quad_f_swz(struct softpipe_surface *gs, GLint x, GLint y, GET_CURRENT_CONTEXT(ctx); for (i = 0; i < 4; i++) { for (j = 0; j < 4; j++) { - temp[j * 4 + i] = FLOAT_TO_UBYTE(src[i * 4 + j]); + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + i], src[i * 4 + j]); } } xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y, temp, NULL); -- cgit v1.2.3 From 20cf13e03b1703937b4228aba8355b34d664aafb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 11 Jul 2007 17:01:30 +0100 Subject: Fix screen corruption on resize. Move buffer resize check to immediately after swapbuffers. Update cliprects inside the locked region of swapbuffers. --- src/mesa/drivers/dri/i915tex/intel_blit.c | 13 ++++ src/mesa/drivers/dri/i915tex/intel_buffers.c | 95 ++-------------------------- src/mesa/drivers/dri/i915tex/intel_context.c | 11 ---- 3 files changed, 17 insertions(+), 102 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c index d6651f62be..c755eac6b2 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ b/src/mesa/drivers/dri/i915tex/intel_blit.c @@ -79,6 +79,13 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, */ LOCK_HARDWARE(intel); + if (intel->revalidateDrawable) { + __DRIscreenPrivate *sPriv = intel->driScreen; + if (dPriv) { + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + } + } + if (dPriv && dPriv->numClipRects) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; const struct intel_region *frontRegion @@ -171,6 +178,12 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, } UNLOCK_HARDWARE(intel); + + if (intel->revalidateDrawable) { + intel->revalidateDrawable = GL_FALSE; + intelWindowMoved(intel); + } + } diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 67f992e4a2..5aed3ad0ec 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -162,60 +162,12 @@ intelSetPrivbufClipRects(struct intel_context *intel) intel->fakeClipRect.y1 = 0; intel->fakeClipRect.x2 = dPriv->w; intel->fakeClipRect.y2 = dPriv->h; - intel->pClipRects = &intel->fakeClipRect; intel->numClipRects = 1; + intel->pClipRects = &intel->fakeClipRect; intel->drawX = 0; intel->drawY = 0; } -/** - * As above, but for rendering to front buffer of a window. - * \sa intelSetRenderbufferClipRects - */ -static void -intelSetFrontClipRects(struct intel_context *intel) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - if (!dPriv) - return; - - intel->numClipRects = dPriv->numClipRects; - intel->pClipRects = dPriv->pClipRects; - intel->drawX = dPriv->x; - intel->drawY = dPriv->y; -} - - -/** - * As above, but for rendering to back buffer of a window. - */ -static void -intelSetBackClipRects(struct intel_context *intel) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - struct intel_framebuffer *intel_fb; - - if (!dPriv) - return; - - intel_fb = dPriv->driverPrivate; - - if (intel_fb->pf_active || dPriv->numBackClipRects == 0) { - /* use the front clip rects */ - intel->numClipRects = dPriv->numClipRects; - intel->pClipRects = dPriv->pClipRects; - intel->drawX = dPriv->x; - intel->drawY = dPriv->y; - } - else { - /* use the back clip rects */ - intel->numClipRects = dPriv->numBackClipRects; - intel->pClipRects = dPriv->pBackClipRects; - intel->drawX = dPriv->backX; - intel->drawY = dPriv->backY; - } -} /** @@ -231,27 +183,10 @@ intelWindowMoved(struct intel_context *intel) if (!intel->ctx.DrawBuffer) { /* when would this happen? -BP */ - intelSetFrontClipRects(intel); - } - else if (1 || intel->ctx.DrawBuffer->Name != 0) { - /* drawing to user-created FBO - do nothing */ - /* Cliprects would be set from intelDrawBuffer() */ - } - else { - /* drawing to a window */ - switch (intel_fb->Base._ColorDrawBufferMask[0]) { - case BUFFER_BIT_FRONT_LEFT: - intelSetFrontClipRects(intel); - break; - case BUFFER_BIT_BACK_LEFT: - intelSetBackClipRects(intel); - break; - default: - /* glDrawBuffer(GL_NONE or GL_FRONT_AND_BACK): software fallback */ - intelSetFrontClipRects(intel); - } + intel->numClipRects = 0; } + if (intel->intelScreen->driScrnPriv->ddxMinor >= 7) { drmI830Sarea *sarea = intel->sarea; drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, @@ -957,23 +892,6 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) intel_fb->swap_ust = ust; } - if (intel->revalidateDrawable) { - __DRIscreenPrivate *sPriv = intel->driScreen; - LOCK_HARDWARE(intel); - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - intel->revalidateDrawable = GL_FALSE; - UNLOCK_HARDWARE(intel); - if (dPriv && intel->lastStamp != dPriv->lastStamp) { - /* XXX this doesn't appear to work quite right. - And in any case, it will never get called with single buffered - rendering here... - And if it's only a window move (not resize), don't need to do anything. */ - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("doing defered drawable update\n"); - intelWindowMoved(intel); - intel->lastStamp = dPriv->lastStamp; - } - } } else { /* XXX this shouldn't be an error but we can't handle it for now */ @@ -1061,12 +979,7 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) /* * How many color buffers are we drawing into? */ - if (fb->_NumColorDrawBuffers[0] != 1 -#if 0 - /* XXX FBO temporary - always use software rendering */ - || 1 -#endif - ) { + if (fb->_NumColorDrawBuffers[0] != 1) { /* writing to 0 or 2 or 4 color buffers */ /*_mesa_debug(ctx, "Software rendering\n");*/ FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index aa2078e762..531b4adde8 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -727,17 +727,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags) intel->current_rotation = sarea->rotation; } #endif - -#if 0 - /* Drawable changed? - */ - if (dPriv && intel->lastStamp != dPriv->lastStamp) { - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("drawable change detected but defering update\n"); -/* intelWindowMoved(intel); - intel->lastStamp = dPriv->lastStamp;*/ - } -#endif } -- cgit v1.2.3 From e89bd0fbc56ecfb96f3aff926c5891c45221dd37 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 11 Jul 2007 11:34:19 -0600 Subject: Implement polygon stipple state tracking, application. --- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_prim_setup.c | 3 ++ src/mesa/pipe/softpipe/sp_quad.c | 4 ++ src/mesa/pipe/softpipe/sp_quad.h | 1 + src/mesa/pipe/softpipe/sp_quad_stipple.c | 56 ++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_state_derived.c | 7 +++- src/mesa/sources | 2 + src/mesa/state_tracker/st_atom.c | 1 + src/mesa/state_tracker/st_atom.h | 1 + src/mesa/state_tracker/st_atom_stipple.c | 62 +++++++++++++++++++++++++++++++ 11 files changed, 138 insertions(+), 1 deletion(-) create mode 100644 src/mesa/pipe/softpipe/sp_quad_stipple.c create mode 100644 src/mesa/state_tracker/st_atom_stipple.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 50434600c3..a5bd61b784 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -83,6 +83,7 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.clear = softpipe_clear; + softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); softpipe->quad.shade = sp_quad_shade_stage(softpipe); softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); softpipe->quad.blend = sp_quad_blend_stage(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 2379a99b81..21d6108ac6 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -117,6 +117,7 @@ struct softpipe_context { /** Software quad rendering pipeline */ struct { + struct quad_stage *polygon_stipple; struct quad_stage *shade; struct quad_stage *alpha_test; struct quad_stage *stencil_test; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 332113979c..0148a26259 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -922,6 +922,9 @@ static void setup_end( struct prim_stage *stage ) } +/** + * Create a new primitive setup/render stage. + */ struct prim_stage *prim_setup( struct softpipe_context *softpipe ) { struct setup_stage *setup = CALLOC_STRUCT(setup_stage); diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 31278f529d..aba5ab280e 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -37,4 +37,8 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.shade; } + if (sp->setup.poly_stipple_enable) { + sp->quad.polygon_stipple->next = sp->quad.first; + sp->quad.first = sp->quad.polygon_stipple; + } } diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index df416dfa7f..8b8e12261b 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -46,6 +46,7 @@ struct quad_stage { }; +struct quad_stage *sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c new file mode 100644 index 0000000000..f9a3c0ba26 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -0,0 +1,56 @@ + +/** + * quad polygon stipple stage + */ + +#include "glheader.h" +#include "imports.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "pipe/p_defines.h" + + +/** + * Apply polygon stipple to quads produced by triangle rasterization + */ +static void +stipple_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + const GLint col0 = quad->x0 % 32; + const GLint row0 = quad->y0 % 32; + const GLuint stipple0 = softpipe->poly_stipple.stipple[row0]; + const GLuint stipple1 = softpipe->poly_stipple.stipple[row0 + 1]; + GLbitfield mask = 0x0; + + /* XXX this could be optimize a bit to use just two conditionals */ + if ((1 << col0) & stipple0) + mask |= MASK_BOTTOM_LEFT; + + if ((2 << col0) & stipple0) + mask |= MASK_BOTTOM_RIGHT; + + if ((1 << col0) & stipple1) + mask |= MASK_TOP_LEFT; + + if ((2 << col0) & stipple1) + mask |= MASK_TOP_RIGHT; + + quad->mask &= mask; + + if (quad->mask) + qs->next->run(qs->next, quad); +} + + +struct quad_stage * +sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = stipple_quad; + + return stage; +} diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 429ae55397..34c893e396 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -137,7 +137,12 @@ void softpipe_update_derived( struct softpipe_context *softpipe ) if (softpipe->dirty & (SP_NEW_SETUP | SP_NEW_FS)) calculate_vertex_layout( softpipe ); - if (softpipe->dirty & (SP_NEW_BLEND | SP_NEW_DEPTH_TEST | SP_NEW_ALPHA_TEST | SP_NEW_FS)) + if (softpipe->dirty & (SP_NEW_BLEND | + SP_NEW_DEPTH_TEST | + SP_NEW_ALPHA_TEST | + SP_NEW_STENCIL | + SP_NEW_SETUP | + SP_NEW_FS)) sp_build_quad_pipeline(softpipe); softpipe->dirty = 0; diff --git a/src/mesa/sources b/src/mesa/sources index 1ffac8df0a..a7d132f4fd 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -163,6 +163,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_quad_depth_test.c \ pipe/softpipe/sp_quad_fs.c \ pipe/softpipe/sp_quad_output.c \ + pipe/softpipe/sp_quad_stipple.c \ pipe/softpipe/sp_quad_stencil.c \ pipe/softpipe/sp_state_blend.c \ pipe/softpipe/sp_state_clip.c \ @@ -206,6 +207,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_scissor.c \ state_tracker/st_atom_setup.c \ state_tracker/st_atom_stencil.c \ + state_tracker/st_atom_stipple.c \ state_tracker/st_atom_viewport.c \ state_tracker/st_cb_program.c \ state_tracker/st_draw.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 88c6c776a3..dfebfb4768 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -48,6 +48,7 @@ static const struct st_tracked_state *atoms[] = &st_update_clip, &st_update_fs, &st_update_setup, + &st_update_polygon_stipple, &st_update_viewport, &st_update_scissor, &st_update_blend, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 7ea3301ea5..a56483ac39 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -50,6 +50,7 @@ const struct st_tracked_state st_update_clear_color; const struct st_tracked_state st_update_depth; const struct st_tracked_state st_update_fs; const struct st_tracked_state st_update_setup; +const struct st_tracked_state st_update_polygon_stipple; const struct st_tracked_state st_update_viewport; const struct st_tracked_state st_update_constants; const struct st_tracked_state st_update_scissor; diff --git a/src/mesa/state_tracker/st_atom_stipple.c b/src/mesa/state_tracker/st_atom_stipple.c new file mode 100644 index 0000000000..dd04d2158c --- /dev/null +++ b/src/mesa/state_tracker/st_atom_stipple.c @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * \brief polygon stipple state + * + * Authors: + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + + +static void +update_stipple( struct st_context *st ) +{ + const GLuint sz = sizeof(st->state.poly_stipple.stipple); + assert(sz == sizeof(st->ctx->PolygonStipple)); + + if (memcmp(&st->state.poly_stipple.stipple, st->ctx->PolygonStipple, sz)) { + /* state has changed */ + memcpy(st->state.poly_stipple.stipple, st->ctx->PolygonStipple, sz); + st->pipe->set_polygon_stipple(st->pipe, &st->state.poly_stipple); + } +} + + +const struct st_tracked_state st_update_polygon_stipple = { + .dirty = { + .mesa = (_NEW_POLYGONSTIPPLE), + .st = 0, + }, + .update = update_stipple +}; -- cgit v1.2.3 From bc4dd63f12059813f6de832e335521eb766f4afb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 11 Jul 2007 11:36:22 -0600 Subject: Obsolete file. --- src/mesa/state_tracker/st_atom_point.c | 70 ---------------------------------- 1 file changed, 70 deletions(-) delete mode 100644 src/mesa/state_tracker/st_atom_point.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_point.c b/src/mesa/state_tracker/st_atom_point.c deleted file mode 100644 index 7299142932..0000000000 --- a/src/mesa/state_tracker/st_atom_point.c +++ /dev/null @@ -1,70 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#include "st_context.h" -#include "st_atom.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" - - -static void -update_point( struct st_context *st ) -{ - struct pipe_point_state point; - - memset(&point, 0, sizeof(point)); - - point.smooth = st->ctx->Point.SmoothFlag; - point.size = st->ctx->Point.Size; - point.min_size = st->ctx->Point.MinSize; - point.max_size = st->ctx->Point.MaxSize; - point.attenuation[0] = st->ctx->Point.Params[0]; - point.attenuation[1] = st->ctx->Point.Params[1]; - point.attenuation[2] = st->ctx->Point.Params[2]; - - if (memcmp(&point, &st->state.point, sizeof(point)) != 0) { - /* state has changed */ - st->state.point = point; /* struct copy */ - st->pipe->set_point_state(st->pipe, &point); /* set new state */ - } -} - - -const struct st_tracked_state st_update_point = { - .dirty = { - .mesa = (_NEW_POINT), - .st = 0, - }, - .update = update_point -}; -- cgit v1.2.3 From 9870f7bccc55f9b1728fa8b9ed78a72ab15d7f4f Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 11 Jul 2007 20:18:51 +0200 Subject: restrict blitted area to size of rendered area --- src/mesa/drivers/dri/i915tex/intel_blit.c | 38 ++++++++++++++++++---------- src/mesa/drivers/dri/i915tex/intel_blit.h | 2 +- src/mesa/drivers/dri/i915tex/intel_buffers.c | 14 +++------- 3 files changed, 30 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c index c755eac6b2..3e42d79980 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ b/src/mesa/drivers/dri/i915tex/intel_blit.c @@ -49,7 +49,7 @@ * Used for SwapBuffers(). */ void -intelCopyBuffer(const __DRIdrawablePrivate * dPriv, +intelCopyBuffer(__DRIdrawablePrivate * dPriv, const drm_clip_rect_t * rect) { @@ -94,6 +94,8 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, = intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); + const int backWidth = intel_fb->Base.Width; + const int backHeight = intel_fb->Base.Height; const int nbox = dPriv->numClipRects; const drm_clip_rect_t *pbox = dPriv->pClipRects; const int pitch = frontRegion->pitch; @@ -134,32 +136,42 @@ intelCopyBuffer(const __DRIdrawablePrivate * dPriv, box = *pbox; if (rect) { - if (rect->x1 > box.x1) - box.x1 = rect->x1; - if (rect->y1 > box.y1) - box.y1 = rect->y1; - if (rect->x2 < box.x2) - box.x2 = rect->x2; - if (rect->y2 < box.y2) - box.y2 = rect->y2; + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; if (box.x1 > box.x2 || box.y1 > box.y2) continue; } + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > backWidth) + box.x2 = backWidth + box.x1; + if (box.y2 - box.y1 > backHeight) + box.y2 = backHeight + box.y1; + DBG("box x1 x2 y1 y2 %d %d %d %d\n", box.x1, box.x2, box.y1, box.y2); - /* XXX should make sure only the minimum area based on - old draw buffer and new front clip rects is copied */ sbox.x1 = box.x1 - dPriv->x; sbox.y1 = box.y1 - dPriv->y; BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); OUT_BATCH(CMD); OUT_BATCH(BR13); - OUT_BATCH((pbox->y1 << 16) | pbox->x1); - OUT_BATCH((pbox->y2 << 16) | pbox->x2); + OUT_BATCH((box.y1 << 16) | box.x1); + OUT_BATCH((box.y2 << 16) | box.x2); OUT_RELOC(frontRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.h b/src/mesa/drivers/dri/i915tex/intel_blit.h index e7bc280f58..77686444fa 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.h +++ b/src/mesa/drivers/dri/i915tex/intel_blit.h @@ -32,7 +32,7 @@ #include "intel_ioctl.h" #include "dri_bufmgr.h" -extern void intelCopyBuffer(const __DRIdrawablePrivate * dpriv, +extern void intelCopyBuffer(__DRIdrawablePrivate * dpriv, const drm_clip_rect_t * rect); extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask); diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 5aed3ad0ec..20e9521105 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -910,17 +910,11 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) if (ctx->Visual.doubleBufferMode) { drm_clip_rect_t rect; -#if 1 - rect.x1 = x + dPriv->x; - rect.y1 = (dPriv->h - y - h) + dPriv->y; - rect.x2 = rect.x1 + w; - rect.y2 = rect.y1 + h; -#else + /* fixup cliprect (driDrawable may have changed?) later */ rect.x1 = x; - rect.y1 = dPriv->h - y; - rect.x2 = rect.x1 + w; - rect.y2 = rect.y1 + h; -#endif + rect.y1 = y; + rect.x2 = w; + rect.y2 = h; _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ intelCopyBuffer(dPriv, &rect); } -- cgit v1.2.3 From aac212966314852936352b024bfd897485aa5e83 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 11 Jul 2007 13:02:03 -0600 Subject: Fix broken depth offset. The edge vectors were totally wrong. Need to multiply bias units by the min resolvable depth delta. --- src/mesa/pipe/draw/draw_offset.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 62ab2e4643..1a41265617 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -54,8 +54,9 @@ static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) static void offset_begin( struct prim_stage *stage ) { struct offset_stage *offset = offset_stage(stage); + GLfloat mrd = 1.0 / 65535.0; /* XXX this depends on depthbuffer bits! */ - offset->units = stage->draw->setup.offset_units; + offset->units = stage->draw->setup.offset_units * mrd; offset->scale = stage->draw->setup.offset_scale; stage->next->begin( stage->next ); @@ -79,11 +80,11 @@ static void do_offset_tri( struct prim_stage *stage, GLfloat *v2 = header->v[2]->data[0]; /* edge vectors e = v0 - v2, f = v1 - v2 */ - GLfloat ex = v0[0] - v2[2]; - GLfloat ey = v0[1] - v2[2]; + GLfloat ex = v0[0] - v2[0]; + GLfloat ey = v0[1] - v2[1]; GLfloat ez = v0[2] - v2[2]; - GLfloat fx = v1[0] - v2[2]; - GLfloat fy = v1[1] - v2[2]; + GLfloat fx = v1[0] - v2[0]; + GLfloat fy = v1[1] - v2[1]; GLfloat fz = v1[2] - v2[2]; /* (a,b) = cross(e,f).xy */ -- cgit v1.2.3 From 46d75518fa4d0cb59dd73a9cd22bfd128ecb2c79 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 11 Jul 2007 13:13:00 -0600 Subject: disable debug printfs --- src/mesa/pipe/softpipe/sp_prim_setup.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 0148a26259..b194f14c03 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -368,11 +368,13 @@ static void tri_linear_coeff( struct setup_stage *setup, (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); + /* _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", slot, "xyzw"[i], setup->coef[slot].a0[i], setup->coef[slot].dadx[i], setup->coef[slot].dady[i]); + */ } @@ -506,7 +508,9 @@ static void subtriangle( struct setup_stage *setup, finish_y = lines; } + /* _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + */ for (y = start_y; y < finish_y; y++) { @@ -560,7 +564,9 @@ static void setup_tri( struct prim_stage *stage, { struct setup_stage *setup = setup_stage( stage ); + /* _mesa_printf("%s\n", __FUNCTION__ ); + */ setup_sort_vertices( setup, prim ); setup_tri_coefficients( setup ); -- cgit v1.2.3 From 2371ed36f7761c4fa452e51a41947ed34f2b661b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 11 Jul 2007 16:17:23 -0600 Subject: remove unused hw_data_offset --- src/mesa/pipe/draw/draw_offset.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 1a41265617..a2705d5cd6 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -25,7 +25,11 @@ * **************************************************************************/ -/* Authors: Keith Whitwell +/** + * \brief polygon offset state + * + * \author Keith Whitwell + * \author Brian Paul */ #include "main/imports.h" @@ -37,8 +41,6 @@ struct offset_stage { struct prim_stage stage; - GLuint hw_data_offset; - GLfloat scale; GLfloat units; }; @@ -47,7 +49,7 @@ struct offset_stage { static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) { - return (struct offset_stage *)stage; + return (struct offset_stage *) stage; } -- cgit v1.2.3 From 5b06424a1507dadad8832d557e79f68a3b68b9c2 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 10:43:33 -0600 Subject: fix MIN/MAX mix-up --- src/mesa/pipe/draw/draw_clip.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index a1c4cf01f0..619ad7d6d6 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -25,9 +25,13 @@ * **************************************************************************/ -/* Authors: Keith Whitwell +/** + * \brief Clipping stage + * + * \author Keith Whitwell */ + #include "main/macros.h" #include "draw_private.h" @@ -322,12 +326,12 @@ do_clip_line( struct prim_stage *stage, if (dp1 < 0) { GLfloat t = dp1 / (dp1 - dp0); - t1 = MIN2(t1, t); + t1 = MAX2(t1, t); } if (dp0 < 0) { GLfloat t = dp0 / (dp0 - dp1); - t0 = MIN2(t0, t); + t0 = MAX2(t0, t); } if (t0 + t1 >= 1.0) -- cgit v1.2.3 From 62f5f18b3e5b80b6d9592442dc3c994d590c782d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 11:50:52 -0600 Subject: handle edge flags for GL_POLYGONs --- src/mesa/pipe/draw/draw_vb.c | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index 905e465447..67e196618a 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -255,16 +255,28 @@ static void draw_indexed_prim( struct draw_context *draw, case GL_POLYGON: if (count >= 3) { + int e1save, e2save; prim.v[0] = 0; prim.v[1] = get_vertex( draw, elts[1] ); prim.v[2] = get_vertex( draw, elts[0] ); + e2save = prim.v[2]->edgeflag; for (i = 0; i+2 < count; i++) { prim.v[0] = prim.v[1]; prim.v[1] = get_vertex( draw, elts[i+2] ); + /* save v1 edge flag, and clear if not last triangle */ + e1save = prim.v[1]->edgeflag; + if (i + 3 < count) + prim.v[1]->edgeflag = 0; + + /* draw */ first->tri( first, &prim ); + + prim.v[1]->edgeflag = e1save; /* restore */ + prim.v[2]->edgeflag = 0; /* disable edge after 1st tri */ } + prim.v[2]->edgeflag = e2save; } break; @@ -399,16 +411,28 @@ static void draw_prim( struct draw_context *draw, case GL_POLYGON: if (count >= 3) { + int e1save, e2save; prim.v[0] = 0; prim.v[1] = get_vertex( draw, start + 1 ); prim.v[2] = get_vertex( draw, start + 0 ); - + e2save = prim.v[2]->edgeflag; + for (i = 0; i+2 < count; i++) { prim.v[0] = prim.v[1]; prim.v[1] = get_vertex( draw, start + i + 2 ); - + + /* save v1 edge flag, and clear if not last triangle */ + e1save = prim.v[1]->edgeflag; + if (i + 3 < count) + prim.v[1]->edgeflag = 0; + + /* draw */ first->tri( first, &prim ); + + prim.v[1]->edgeflag = e1save; /* restore */ + prim.v[2]->edgeflag = 0; /* disable edge after 1st tri */ } + prim.v[2]->edgeflag = e2save; } break; -- cgit v1.2.3 From 9e94e9b7980b08038d7d4c6b76b8f184d70ba163 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 11:59:15 -0600 Subject: culling was inverted --- src/mesa/pipe/draw/draw_cull.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index b6c578b3c9..1498874fcf 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -79,7 +79,7 @@ static void cull_tri( struct prim_stage *stage, if (header->det != 0) { /* non-zero area */ - GLuint mode = (header->det < 0) ? PIPE_WINDING_CW : PIPE_WINDING_CCW; + GLuint mode = (header->det > 0) ? PIPE_WINDING_CW : PIPE_WINDING_CCW; if ((mode & cull_stage(stage)->mode) == 0) { /* triangle is not culled, pass to next stage */ -- cgit v1.2.3 From bd5533bb7c90d8da06f6a1c38bdc84bb607c5b85 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 11:59:43 -0600 Subject: front/back-face determination was wrong --- src/mesa/pipe/draw/draw_twoside.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 88e164ec5e..e8221639bb 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -51,7 +51,7 @@ static void twoside_begin( struct prim_stage *stage ) { struct twoside_stage *twoside = twoside_stage(stage); - twoside->facing = (stage->draw->setup.front_winding == PIPE_WINDING_CW) ? -1 : 1; + twoside->facing = (stage->draw->setup.front_winding == PIPE_WINDING_CW) ? 1 : -1; stage->next->begin( stage->next ); } -- cgit v1.2.3 From 86352ff70d8c9a31fe0ebb4d02ce4bb4644fe54a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 12:20:14 -0600 Subject: Added colormask, dither, multisample state. Implement colormasking stage. --- src/mesa/pipe/p_defines.h | 5 ++ src/mesa/pipe/p_state.h | 5 ++ src/mesa/pipe/softpipe/sp_context.c | 3 +- src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_quad.c | 5 ++ src/mesa/pipe/softpipe/sp_quad.h | 1 + src/mesa/pipe/softpipe/sp_quad_colormask.c | 88 ++++++++++++++++++++++++++++++ src/mesa/sources | 1 + src/mesa/state_tracker/st_atom_blend.c | 13 +++++ src/mesa/state_tracker/st_atom_setup.c | 5 +- 10 files changed, 125 insertions(+), 2 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_quad_colormask.c (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index fbdd015619..821521ac75 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -71,6 +71,11 @@ #define PIPE_LOGICOP_OR 14 #define PIPE_LOGICOP_SET 15 +#define PIPE_MASK_R 0x1 +#define PIPE_MASK_G 0x2 +#define PIPE_MASK_B 0x4 +#define PIPE_MASK_A 0x8 + /** * Inequality functions. Used for depth test, stencil compare, alpha * test, shadow compare, etc. diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 581ea5ddd8..3dfe584af8 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -89,6 +89,8 @@ struct pipe_setup_state GLuint point_smooth:1; + GLuint multisample:1; /* XXX maybe more ms state in future */ + GLubyte line_stipple_factor; /**< [1..256] actually */ GLushort line_stipple_pattern; GLfloat line_width; @@ -156,6 +158,9 @@ struct pipe_blend_state { GLuint logicop_enable:1; GLuint logicop_func:4; /**< PIPE_LOGICOP_x */ + + GLuint colormask:4; /**< bitmask of PIPE_MASK_R/G/B/A */ + GLuint dither:1; }; struct pipe_blend_color { diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index a5bd61b784..671ef2764b 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -86,9 +86,10 @@ struct pipe_context *softpipe_create( void ) softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); softpipe->quad.shade = sp_quad_shade_stage(softpipe); softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); - softpipe->quad.blend = sp_quad_blend_stage(softpipe); softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); + softpipe->quad.blend = sp_quad_blend_stage(softpipe); + softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); /* diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 21d6108ac6..47ce2f0ad8 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -123,6 +123,7 @@ struct softpipe_context { struct quad_stage *stencil_test; struct quad_stage *depth_test; struct quad_stage *blend; + struct quad_stage *colormask; struct quad_stage *output; struct quad_stage *first; /**< points to one of the above stages */ diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index aba5ab280e..419a720de9 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -11,6 +11,11 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.output; + if (sp->blend.colormask != 0xf) { + sp->quad.colormask->next = sp->quad.first; + sp->quad.first = sp->quad.colormask; + } + if (sp->blend.blend_enable) { sp->quad.blend->next = sp->quad.first; sp->quad.first = sp->quad.blend; diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 8b8e12261b..966f72d3b5 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -52,6 +52,7 @@ struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ) struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); void sp_build_quad_pipeline(struct softpipe_context *sp); diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c new file mode 100644 index 0000000000..6fb228fdda --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -0,0 +1,88 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief quad colormask stage + * \author Brian Paul + */ + +#include "main/glheader.h" +#include "main/imports.h" +#include "main/macros.h" +#include "pipe/p_defines.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" + + + +static void +colormask_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + GLfloat dest[4][QUAD_SIZE]; + + /* XXX buffer looping */ + + struct softpipe_surface *sps + = softpipe_surface(softpipe->framebuffer.cbufs[0]); + + sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + + /* R */ + if (!(softpipe->blend.colormask & PIPE_MASK_R)) + COPY_4FV(quad->outputs.color[0], dest[0]); + + /* G */ + if (!(softpipe->blend.colormask & PIPE_MASK_G)) + COPY_4FV(quad->outputs.color[1], dest[1]); + + /* B */ + if (!(softpipe->blend.colormask & PIPE_MASK_B)) + COPY_4FV(quad->outputs.color[2], dest[2]); + + /* A */ + if (!(softpipe->blend.colormask & PIPE_MASK_A)) + COPY_4FV(quad->outputs.color[3], dest[3]); + + /* pass quad to next stage */ + qs->next->run(qs->next, quad); +} + + + + +struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = colormask_quad; + + return stage; +} diff --git a/src/mesa/sources b/src/mesa/sources index a7d132f4fd..78a030b767 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -160,6 +160,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_quad.c \ pipe/softpipe/sp_quad_alpha_test.c \ pipe/softpipe/sp_quad_blend.c \ + pipe/softpipe/sp_quad_colormask.c \ pipe/softpipe/sp_quad_depth_test.c \ pipe/softpipe/sp_quad_fs.c \ pipe/softpipe/sp_quad_output.c \ diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index 3e5410cfab..256f13471a 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -196,6 +196,19 @@ update_blend( struct st_context *st ) /* no blending / logicop */ } + /* Colormask - maybe reverse these bits? */ + if (st->ctx->Color.ColorMask[0]) + blend.colormask |= PIPE_MASK_R; + if (st->ctx->Color.ColorMask[1]) + blend.colormask |= PIPE_MASK_G; + if (st->ctx->Color.ColorMask[2]) + blend.colormask |= PIPE_MASK_B; + if (st->ctx->Color.ColorMask[3]) + blend.colormask |= PIPE_MASK_A; + + if (st->ctx->Color.DitherFlag) + blend.dither = 1; + if (memcmp(&blend, &st->state.blend, sizeof(blend)) != 0) { /* state has changed */ st->state.blend = blend; /* struct copy */ diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index 842a87c8a1..8b95ea958e 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -186,6 +186,9 @@ static void update_setup_state( struct st_context *st ) /* GL stipple factor is in [1,256], remap to [0, 255] here */ setup.line_stipple_factor = ctx->Line.StippleFactor - 1; + /* _NEW_MULTISAMPLE */ + if (ctx->Multisample.Enabled) + setup.multisample = 1; if (memcmp(&setup, &st->state.setup, sizeof(setup)) != 0) { st->state.setup = setup; @@ -196,7 +199,7 @@ static void update_setup_state( struct st_context *st ) const struct st_tracked_state st_update_setup = { .dirty = { .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | - _NEW_POINT | _NEW_BUFFERS), + _NEW_POINT | _NEW_BUFFERS | _NEW_MULTISAMPLE), .st = 0, }, .update = update_setup_state -- cgit v1.2.3 From ea470eec86715cd2bc9aa86d36e6ea803d0d4017 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 13:32:31 -0600 Subject: Rename prim_stage -> draw_stage --- src/mesa/pipe/draw/draw_clip.c | 28 +++++++++---------- src/mesa/pipe/draw/draw_context.c | 16 +++++------ src/mesa/pipe/draw/draw_context.h | 4 +-- src/mesa/pipe/draw/draw_cull.c | 18 ++++++------- src/mesa/pipe/draw/draw_flatshade.c | 25 +++++++++-------- src/mesa/pipe/draw/draw_offset.c | 24 ++++++++++------- src/mesa/pipe/draw/draw_private.h | 48 ++++++++++++++++----------------- src/mesa/pipe/draw/draw_twoside.c | 18 ++++++------- src/mesa/pipe/draw/draw_unfilled.c | 30 ++++++++++++--------- src/mesa/pipe/draw/draw_vb.c | 10 +++---- src/mesa/pipe/softpipe/sp_context.c | 5 ++++ src/mesa/pipe/softpipe/sp_context.h | 2 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 29 +++++++++++--------- src/mesa/pipe/softpipe/sp_prim_setup.h | 2 +- src/mesa/pipe/softpipe/sp_state_setup.c | 2 +- 15 files changed, 141 insertions(+), 120 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 619ad7d6d6..7ede51fc6c 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -37,7 +37,7 @@ struct clipper { - struct prim_stage stage; /**< base class */ + struct draw_stage stage; /**< base class */ GLuint active_user_planes; GLfloat (*plane)[4]; @@ -46,7 +46,7 @@ struct clipper { /* This is a bit confusing: */ -static INLINE struct clipper *clipper_stage( struct prim_stage *stage ) +static INLINE struct clipper *clipper_stage( struct draw_stage *stage ) { return (struct clipper *)stage; } @@ -139,7 +139,7 @@ static INLINE GLfloat dot4( const GLfloat *a, #if 0 -static INLINE void do_tri( struct prim_stage *next, +static INLINE void do_tri( struct draw_stage *next, struct prim_header *header ) { GLuint i; @@ -155,7 +155,7 @@ static INLINE void do_tri( struct prim_stage *next, #endif -static void emit_poly( struct prim_stage *stage, +static void emit_poly( struct draw_stage *stage, struct vertex_header **inlist, GLuint n ) { @@ -184,7 +184,7 @@ static void emit_poly( struct prim_stage *stage, #if 0 -static void emit_poly( struct prim_stage *stage ) +static void emit_poly( struct draw_stage *stage ) { GLuint i; @@ -202,7 +202,7 @@ static void emit_poly( struct prim_stage *stage ) /* Clip a triangle against the viewport and user clip planes. */ static void -do_clip_tri( struct prim_stage *stage, +do_clip_tri( struct draw_stage *stage, struct prim_header *header, GLuint clipmask ) { @@ -296,7 +296,7 @@ do_clip_tri( struct prim_stage *stage, /* Clip a line against the viewport and user clip planes. */ static void -do_clip_line( struct prim_stage *stage, +do_clip_line( struct draw_stage *stage, struct prim_header *header, GLuint clipmask ) { @@ -360,7 +360,7 @@ do_clip_line( struct prim_stage *stage, } -static void clip_begin( struct prim_stage *stage ) +static void clip_begin( struct draw_stage *stage ) { struct clipper *clipper = clipper_stage(stage); GLuint nr = stage->draw->nr_planes; @@ -374,7 +374,7 @@ static void clip_begin( struct prim_stage *stage ) static void -clip_point( struct prim_stage *stage, +clip_point( struct draw_stage *stage, struct prim_header *header ) { if (header->v[0]->clipmask == 0) @@ -383,7 +383,7 @@ clip_point( struct prim_stage *stage, static void -clip_line( struct prim_stage *stage, +clip_line( struct draw_stage *stage, struct prim_header *header ) { GLuint clipmask = (header->v[0]->clipmask | @@ -401,7 +401,7 @@ clip_line( struct prim_stage *stage, static void -clip_tri( struct prim_stage *stage, +clip_tri( struct draw_stage *stage, struct prim_header *header ) { GLuint clipmask = (header->v[0]->clipmask | @@ -420,7 +420,7 @@ clip_tri( struct prim_stage *stage, } -static void clip_end( struct prim_stage *stage ) +static void clip_end( struct draw_stage *stage ) { stage->next->end( stage->next ); } @@ -430,11 +430,11 @@ static void clip_end( struct prim_stage *stage ) * Allocate a new clipper stage. * \return pointer to new stage object */ -struct prim_stage *prim_clip( struct draw_context *draw ) +struct draw_stage *draw_clip_stage( struct draw_context *draw ) { struct clipper *clipper = CALLOC_STRUCT(clipper); - prim_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); + draw_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); clipper->stage.draw = draw; clipper->stage.begin = clip_begin; diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index d4654791bb..a97f488387 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -42,12 +42,12 @@ struct draw_context *draw_create( void ) struct draw_context *draw = CALLOC_STRUCT( draw_context ); /* create pipeline stages */ - draw->pipeline.unfilled = prim_unfilled( draw ); - draw->pipeline.twoside = prim_twoside( draw ); - draw->pipeline.offset = prim_offset( draw ); - draw->pipeline.clip = prim_clip( draw ); - draw->pipeline.flatshade = prim_flatshade( draw ); - draw->pipeline.cull = prim_cull( draw ); + draw->pipeline.unfilled = draw_unfilled_stage( draw ); + draw->pipeline.twoside = draw_twoside_stage( draw ); + draw->pipeline.offset = draw_offset_stage( draw ); + draw->pipeline.clip = draw_clip_stage( draw ); + draw->pipeline.flatshade = draw_flatshade_stage( draw ); + draw->pipeline.cull = draw_cull_stage( draw ); ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); @@ -79,7 +79,7 @@ void draw_destroy( struct draw_context *draw ) */ static void validate_pipeline( struct draw_context *draw ) { - struct prim_stage *next = draw->pipeline.setup; + struct draw_stage *next = draw->pipeline.setup; /* * NOTE: we build up the pipeline in end-to-start order. @@ -150,7 +150,7 @@ void draw_set_setup_state( struct draw_context *draw, * This is provided by the device driver. */ void draw_set_setup_stage( struct draw_context *draw, - struct prim_stage *stage ) + struct draw_stage *stage ) { draw->pipeline.setup = stage; } diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 85f2ace75f..c298d4f46d 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -44,7 +44,7 @@ struct vertex_buffer; struct draw_context; -struct prim_stage; +struct draw_stage; struct draw_context *draw_create( void ); @@ -61,7 +61,7 @@ void draw_set_setup_state( struct draw_context *draw, const struct pipe_setup_state *setup ); void draw_set_setup_stage( struct draw_context *draw, - struct prim_stage *stage ); + struct draw_stage *stage ); void draw_set_vertex_attributes( struct draw_context *draw, const GLuint *attrs, diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index 1498874fcf..863686f150 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -39,18 +39,18 @@ struct cull_stage { - struct prim_stage stage; + struct draw_stage stage; GLuint mode; /**< one of PIPE_WINDING_x */ }; -static INLINE struct cull_stage *cull_stage( struct prim_stage *stage ) +static INLINE struct cull_stage *cull_stage( struct draw_stage *stage ) { return (struct cull_stage *)stage; } -static void cull_begin( struct prim_stage *stage ) +static void cull_begin( struct draw_stage *stage ) { struct cull_stage *cull = cull_stage(stage); @@ -60,7 +60,7 @@ static void cull_begin( struct prim_stage *stage ) } -static void cull_tri( struct prim_stage *stage, +static void cull_tri( struct draw_stage *stage, struct prim_header *header ) { /* Window coords: */ @@ -89,21 +89,21 @@ static void cull_tri( struct prim_stage *stage, } -static void cull_line( struct prim_stage *stage, +static void cull_line( struct draw_stage *stage, struct prim_header *header ) { stage->next->line( stage->next, header ); } -static void cull_point( struct prim_stage *stage, +static void cull_point( struct draw_stage *stage, struct prim_header *header ) { stage->next->point( stage->next, header ); } -static void cull_end( struct prim_stage *stage ) +static void cull_end( struct draw_stage *stage ) { stage->next->end( stage->next ); } @@ -112,11 +112,11 @@ static void cull_end( struct prim_stage *stage ) /** * Create a new polygon culling stage. */ -struct prim_stage *prim_cull( struct draw_context *draw ) +struct draw_stage *draw_cull_stage( struct draw_context *draw ) { struct cull_stage *cull = CALLOC_STRUCT(cull_stage); - prim_alloc_tmps( &cull->stage, 0 ); + draw_alloc_tmps( &cull->stage, 0 ); cull->stage.draw = draw; cull->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 004b5bdc96..09547865fa 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -33,20 +33,20 @@ struct flatshade_stage { - struct prim_stage stage; + struct draw_stage stage; const GLuint *lookup; }; -static INLINE struct flatshade_stage *flatshade_stage( struct prim_stage *stage ) +static INLINE struct flatshade_stage *flatshade_stage( struct draw_stage *stage ) { return (struct flatshade_stage *)stage; } -static void flatshade_begin( struct prim_stage *stage ) +static void flatshade_begin( struct draw_stage *stage ) { stage->next->begin( stage->next ); } @@ -65,7 +65,7 @@ static INLINE void copy_attr( GLuint attr, } -static INLINE void copy_colors( struct prim_stage *stage, +static INLINE void copy_colors( struct draw_stage *stage, struct vertex_header *dst, const struct vertex_header *src ) { @@ -83,7 +83,7 @@ static INLINE void copy_colors( struct prim_stage *stage, * Flatshade tri. Required for clipping and when unfilled tris are * active, otherwise handled by hardware. */ -static void flatshade_tri( struct prim_stage *stage, +static void flatshade_tri( struct draw_stage *stage, struct prim_header *header ) { struct prim_header tmp; @@ -103,7 +103,7 @@ static void flatshade_tri( struct prim_stage *stage, /** * Flatshade line. Required for clipping. */ -static void flatshade_line( struct prim_stage *stage, +static void flatshade_line( struct draw_stage *stage, struct prim_header *header ) { struct prim_header tmp; @@ -117,24 +117,27 @@ static void flatshade_line( struct prim_stage *stage, } -static void flatshade_point( struct prim_stage *stage, - struct prim_header *header ) +static void flatshade_point( struct draw_stage *stage, + struct prim_header *header ) { stage->next->point( stage->next, header ); } -static void flatshade_end( struct prim_stage *stage ) +static void flatshade_end( struct draw_stage *stage ) { stage->next->end( stage->next ); } -struct prim_stage *prim_flatshade( struct draw_context *draw ) +/** + * Create flatshading drawing stage. + */ +struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) { struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); - prim_alloc_tmps( &flatshade->stage, 2 ); + draw_alloc_tmps( &flatshade->stage, 2 ); flatshade->stage.draw = draw; flatshade->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index a2705d5cd6..1a7f56d0c5 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -39,7 +39,7 @@ struct offset_stage { - struct prim_stage stage; + struct draw_stage stage; GLfloat scale; GLfloat units; @@ -47,13 +47,13 @@ struct offset_stage { -static INLINE struct offset_stage *offset_stage( struct prim_stage *stage ) +static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) { return (struct offset_stage *) stage; } -static void offset_begin( struct prim_stage *stage ) +static void offset_begin( struct draw_stage *stage ) { struct offset_stage *offset = offset_stage(stage); GLfloat mrd = 1.0 / 65535.0; /* XXX this depends on depthbuffer bits! */ @@ -69,7 +69,7 @@ static void offset_begin( struct prim_stage *stage ) * Offset tri Z. Some hardware can handle this, but not usually when * doing unfilled rendering. */ -static void do_offset_tri( struct prim_stage *stage, +static void do_offset_tri( struct draw_stage *stage, struct prim_header *header ) { struct offset_stage *offset = offset_stage(stage); @@ -110,7 +110,7 @@ static void do_offset_tri( struct prim_stage *stage, } -static void offset_tri( struct prim_stage *stage, +static void offset_tri( struct draw_stage *stage, struct prim_header *header ) { struct prim_header tmp; @@ -125,30 +125,34 @@ static void offset_tri( struct prim_stage *stage, -static void offset_line( struct prim_stage *stage, +static void offset_line( struct draw_stage *stage, struct prim_header *header ) { stage->next->line( stage->next, header ); } -static void offset_point( struct prim_stage *stage, +static void offset_point( struct draw_stage *stage, struct prim_header *header ) { stage->next->point( stage->next, header ); } -static void offset_end( struct prim_stage *stage ) +static void offset_end( struct draw_stage *stage ) { stage->next->end( stage->next ); } -struct prim_stage *prim_offset( struct draw_context *draw ) + +/** + * Create polygon offset drawing stage. + */ +struct draw_stage *draw_offset_stage( struct draw_context *draw ) { struct offset_stage *offset = CALLOC_STRUCT(offset_stage); - prim_alloc_tmps( &offset->stage, 3 ); + draw_alloc_tmps( &offset->stage, 3 ); offset->stage.draw = draw; offset->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index ac25628a08..573b5bf588 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -77,27 +77,27 @@ struct draw_context; /** * Base class for all primitive drawing stages. */ -struct prim_stage +struct draw_stage { struct draw_context *draw; /**< parent context */ - struct prim_stage *next; /**< next stage in pipeline */ + struct draw_stage *next; /**< next stage in pipeline */ struct vertex_header **tmp; GLuint nr_tmps; - void (*begin)( struct prim_stage * ); + void (*begin)( struct draw_stage * ); - void (*point)( struct prim_stage *, + void (*point)( struct draw_stage *, struct prim_header * ); - void (*line)( struct prim_stage *, + void (*line)( struct draw_stage *, struct prim_header * ); - void (*tri)( struct prim_stage *, + void (*tri)( struct draw_stage *, struct prim_header * ); - void (*end)( struct prim_stage * ); + void (*end)( struct draw_stage * ); }; @@ -107,16 +107,16 @@ struct prim_stage struct draw_context { struct { - struct prim_stage *first; /**< one of the following */ + struct draw_stage *first; /**< one of the following */ /* stages (in logical order) */ - struct prim_stage *flatshade; - struct prim_stage *clip; - struct prim_stage *cull; - struct prim_stage *twoside; - struct prim_stage *offset; - struct prim_stage *unfilled; - struct prim_stage *setup; /* aka render/rasterize */ + struct draw_stage *flatshade; + struct draw_stage *clip; + struct draw_stage *cull; + struct draw_stage *twoside; + struct draw_stage *offset; + struct draw_stage *unfilled; + struct draw_stage *setup; /* aka render/rasterize */ } pipeline; /* pipe state that we need: */ @@ -148,16 +148,16 @@ struct draw_context -extern struct prim_stage *prim_unfilled( struct draw_context *context ); -extern struct prim_stage *prim_twoside( struct draw_context *context ); -extern struct prim_stage *prim_offset( struct draw_context *context ); -extern struct prim_stage *prim_clip( struct draw_context *context ); -extern struct prim_stage *prim_flatshade( struct draw_context *context ); -extern struct prim_stage *prim_cull( struct draw_context *context ); +extern struct draw_stage *draw_unfilled_stage( struct draw_context *context ); +extern struct draw_stage *draw_twoside_stage( struct draw_context *context ); +extern struct draw_stage *draw_offset_stage( struct draw_context *context ); +extern struct draw_stage *draw_clip_stage( struct draw_context *context ); +extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); +extern struct draw_stage *draw_cull_stage( struct draw_context *context ); -extern void prim_free_tmps( struct prim_stage *stage ); -extern void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); +extern void draw_free_tmps( struct draw_stage *stage ); +extern void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ); @@ -169,7 +169,7 @@ extern void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ); * \return pointer to the copied vertex */ static INLINE struct vertex_header * -dup_vert( struct prim_stage *stage, +dup_vert( struct draw_stage *stage, const struct vertex_header *vert, GLuint idx ) { diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index e8221639bb..e86123e84d 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -34,20 +34,20 @@ struct twoside_stage { - struct prim_stage stage; + struct draw_stage stage; GLfloat facing; const GLuint *lookup; }; -static INLINE struct twoside_stage *twoside_stage( struct prim_stage *stage ) +static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage ) { return (struct twoside_stage *)stage; } -static void twoside_begin( struct prim_stage *stage ) +static void twoside_begin( struct draw_stage *stage ) { struct twoside_stage *twoside = twoside_stage(stage); @@ -89,7 +89,7 @@ static struct vertex_header *copy_bfc( struct twoside_stage *twoside, /* Twoside tri: */ -static void twoside_tri( struct prim_stage *stage, +static void twoside_tri( struct draw_stage *stage, struct prim_header *header ) { struct twoside_stage *twoside = twoside_stage(stage); @@ -112,7 +112,7 @@ static void twoside_tri( struct prim_stage *stage, } -static void twoside_line( struct prim_stage *stage, +static void twoside_line( struct draw_stage *stage, struct prim_header *header ) { /* pass-through */ @@ -120,7 +120,7 @@ static void twoside_line( struct prim_stage *stage, } -static void twoside_point( struct prim_stage *stage, +static void twoside_point( struct draw_stage *stage, struct prim_header *header ) { /* pass-through */ @@ -128,7 +128,7 @@ static void twoside_point( struct prim_stage *stage, } -static void twoside_end( struct prim_stage *stage ) +static void twoside_end( struct draw_stage *stage ) { /* pass-through */ stage->next->end( stage->next ); @@ -138,11 +138,11 @@ static void twoside_end( struct prim_stage *stage ) /** * Create twoside pipeline stage. */ -struct prim_stage *prim_twoside( struct draw_context *draw ) +struct draw_stage *draw_twoside_stage( struct draw_context *draw ) { struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); - prim_alloc_tmps( &twoside->stage, 3 ); + draw_alloc_tmps( &twoside->stage, 3 ); twoside->stage.draw = draw; twoside->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index 8ee23d7493..f0aee58333 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -39,7 +39,7 @@ struct unfilled_stage { - struct prim_stage stage; + struct draw_stage stage; /** [0] = front face, [1] = back face. * legal values: PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE, @@ -49,13 +49,13 @@ struct unfilled_stage { }; -static INLINE struct unfilled_stage *unfilled_stage( struct prim_stage *stage ) +static INLINE struct unfilled_stage *unfilled_stage( struct draw_stage *stage ) { return (struct unfilled_stage *)stage; } -static void unfilled_begin( struct prim_stage *stage ) +static void unfilled_begin( struct draw_stage *stage ) { struct unfilled_stage *unfilled = unfilled_stage(stage); @@ -65,7 +65,7 @@ static void unfilled_begin( struct prim_stage *stage ) stage->next->begin( stage->next ); } -static void point( struct prim_stage *stage, +static void point( struct draw_stage *stage, struct vertex_header *v0 ) { struct prim_header tmp; @@ -73,7 +73,7 @@ static void point( struct prim_stage *stage, stage->next->point( stage->next, &tmp ); } -static void line( struct prim_stage *stage, +static void line( struct draw_stage *stage, struct vertex_header *v0, struct vertex_header *v1 ) { @@ -84,7 +84,7 @@ static void line( struct prim_stage *stage, } -static void points( struct prim_stage *stage, +static void points( struct draw_stage *stage, struct prim_header *header ) { struct vertex_header *v0 = header->v[0]; @@ -97,7 +97,7 @@ static void points( struct prim_stage *stage, } -static void lines( struct prim_stage *stage, +static void lines( struct draw_stage *stage, struct prim_header *header ) { struct vertex_header *v0 = header->v[0]; @@ -115,7 +115,7 @@ static void lines( struct prim_stage *stage, * Note edgeflags in the vertex struct is not sufficient as we will * need to manipulate them when decomposing primitives??? */ -static void unfilled_tri( struct prim_stage *stage, +static void unfilled_tri( struct draw_stage *stage, struct prim_header *header ) { struct unfilled_stage *unfilled = unfilled_stage(stage); @@ -136,30 +136,34 @@ static void unfilled_tri( struct prim_stage *stage, } } -static void unfilled_line( struct prim_stage *stage, +static void unfilled_line( struct draw_stage *stage, struct prim_header *header ) { stage->next->line( stage->next, header ); } -static void unfilled_point( struct prim_stage *stage, +static void unfilled_point( struct draw_stage *stage, struct prim_header *header ) { stage->next->point( stage->next, header ); } -static void unfilled_end( struct prim_stage *stage ) +static void unfilled_end( struct draw_stage *stage ) { stage->next->end( stage->next ); } -struct prim_stage *prim_unfilled( struct draw_context *draw ) + +/** + * Create unfilled triangle stage. + */ +struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) { struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); - prim_alloc_tmps( &unfilled->stage, 0 ); + draw_alloc_tmps( &unfilled->stage, 0 ); unfilled->stage.draw = draw; unfilled->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index 67e196618a..ac721c589d 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -88,7 +88,7 @@ static void draw_set_prim( struct draw_context *draw, -static void do_quad( struct prim_stage *first, +static void do_quad( struct draw_stage *first, struct vertex_header *v0, struct vertex_header *v1, struct vertex_header *v2, @@ -128,7 +128,7 @@ static void draw_indexed_prim( struct draw_context *draw, const GLuint *elts, GLuint count ) { - struct prim_stage * const first = draw->pipeline.first; + struct draw_stage * const first = draw->pipeline.first; struct prim_header prim; GLuint i; @@ -290,7 +290,7 @@ static void draw_prim( struct draw_context *draw, GLuint start, GLuint count ) { - struct prim_stage * const first = draw->pipeline.first; + struct draw_stage * const first = draw->pipeline.first; struct prim_header prim; GLuint i; @@ -698,7 +698,7 @@ void draw_set_vertex_attributes( struct draw_context *draw, #define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) -void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ) +void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ) { stage->nr_tmps = nr; @@ -713,7 +713,7 @@ void prim_alloc_tmps( struct prim_stage *stage, GLuint nr ) } } -void prim_free_tmps( struct prim_stage *stage ) +void draw_free_tmps( struct draw_stage *stage ) { if (stage->tmp) { FREE(stage->tmp[0]); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 671ef2764b..d3aea3698a 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -98,5 +98,10 @@ struct pipe_context *softpipe_create( void ) softpipe->draw = draw_create(); draw_set_setup_stage(softpipe->draw, prim_setup(softpipe)); + /* + * XXX we could plug GL selection/feedback into the drawing pipeline + * by specifying a different setup/render stage. + */ + return &softpipe->pipe; } diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 47ce2f0ad8..3ae1de71cf 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -41,7 +41,7 @@ struct softpipe_surface; struct draw_context; -struct prim_stage; +struct draw_stage; enum interp_mode { diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index b194f14c03..5db05ecccb 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -25,9 +25,14 @@ * **************************************************************************/ -/* Authors: Keith Whitwell +/** + * \brief Primitive rasterization/rendering (points, lines, triangles) + * + * \author Keith Whitwell + * \author Brian Paul */ + #include "imports.h" #include "macros.h" @@ -64,11 +69,11 @@ struct edge { /** - * Triangle setup info (derived from prim_stage). + * Triangle setup info (derived from draw_stage). * Also used for line drawing (taking some liberties). */ struct setup_stage { - struct prim_stage stage; /**< This must be first (base class) */ + struct draw_stage stage; /**< This must be first (base class) */ /*XXX NEW */ struct softpipe_context *softpipe; @@ -105,7 +110,7 @@ struct setup_stage { /** * Basically a cast wrapper. */ -static inline struct setup_stage *setup_stage( struct prim_stage *stage ) +static inline struct setup_stage *setup_stage( struct draw_stage *stage ) { return (struct setup_stage *)stage; } @@ -122,7 +127,7 @@ static inline GLint block( GLint x ) -static void setup_begin( struct prim_stage *stage ) +static void setup_begin( struct draw_stage *stage ) { struct setup_stage *setup = setup_stage(stage); @@ -559,7 +564,7 @@ static void subtriangle( struct setup_stage *setup, /** * Do setup for triangle rasterization, then render the triangle. */ -static void setup_tri( struct prim_stage *stage, +static void setup_tri( struct draw_stage *stage, struct prim_header *prim ) { struct setup_stage *setup = setup_stage( stage ); @@ -576,9 +581,9 @@ static void setup_tri( struct prim_stage *stage, setup->span.y_flags = 0; setup->span.right[0] = 0; setup->span.right[1] = 0; -// setup->span.z_mode = tri_z_mode( setup->ctx ); + /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ -// init_constant_attribs( setup ); + /* init_constant_attribs( setup ); */ if (setup->oneoverarea < 0.0) { /* emaj on left: @@ -714,7 +719,7 @@ plot(struct setup_stage *setup, GLint x, GLint y) * XXX no scissoring yet. */ static void -setup_line(struct prim_stage *stage, struct prim_header *prim) +setup_line(struct draw_stage *stage, struct prim_header *prim) { const struct vertex_header *v0 = prim->v[0]; const struct vertex_header *v1 = prim->v[1]; @@ -810,7 +815,7 @@ setup_line(struct prim_stage *stage, struct prim_header *prim) * XXX could optimize a lot for 1-pixel points. */ static void -setup_point(struct prim_stage *stage, struct prim_header *prim) +setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); /*XXX this should be a vertex attrib! */ @@ -923,7 +928,7 @@ setup_point(struct prim_stage *stage, struct prim_header *prim) -static void setup_end( struct prim_stage *stage ) +static void setup_end( struct draw_stage *stage ) { } @@ -931,7 +936,7 @@ static void setup_end( struct prim_stage *stage ) /** * Create a new primitive setup/render stage. */ -struct prim_stage *prim_setup( struct softpipe_context *softpipe ) +struct draw_stage *prim_setup( struct softpipe_context *softpipe ) { struct setup_stage *setup = CALLOC_STRUCT(setup_stage); diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h index 4d26c05458..be00b0182b 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.h +++ b/src/mesa/pipe/softpipe/sp_prim_setup.h @@ -43,7 +43,7 @@ #include "s_context.h" -extern struct prim_stage *prim_setup( struct softpipe_context *softpipe ); +extern struct draw_stage *prim_setup( struct softpipe_context *softpipe ); #if 0 /* UNUSED? */ diff --git a/src/mesa/pipe/softpipe/sp_state_setup.c b/src/mesa/pipe/softpipe/sp_state_setup.c index 6dbb87ae79..cef70d42d9 100644 --- a/src/mesa/pipe/softpipe/sp_state_setup.c +++ b/src/mesa/pipe/softpipe/sp_state_setup.c @@ -34,7 +34,7 @@ #if 0 static void validate_prim_pipe( struct softpipe_context *softpipe ) { - struct prim_stage *next = softpipe->prim.setup; + struct draw_stage *next = softpipe->prim.setup; /* TODO: make the current primitive part of the state and build * shorter pipelines for lines & points. -- cgit v1.2.3 From 4f218958cc0bc57196a748cb005f94aeeace037c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 13:33:33 -0600 Subject: Remove dead code --- src/mesa/pipe/softpipe/sp_state_setup.c | 62 --------------------------------- 1 file changed, 62 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_setup.c b/src/mesa/pipe/softpipe/sp_state_setup.c index cef70d42d9..4715a26f55 100644 --- a/src/mesa/pipe/softpipe/sp_state_setup.c +++ b/src/mesa/pipe/softpipe/sp_state_setup.c @@ -31,65 +31,6 @@ #include "pipe/draw/draw_context.h" -#if 0 -static void validate_prim_pipe( struct softpipe_context *softpipe ) -{ - struct draw_stage *next = softpipe->prim.setup; - - /* TODO: make the current primitive part of the state and build - * shorter pipelines for lines & points. - */ - if (softpipe->setup.fill_cw != PIPE_POLYGON_MODE_FILL || - softpipe->setup.fill_ccw != PIPE_POLYGON_MODE_FILL) { - - softpipe->prim.unfilled->next = next; - next = softpipe->prim.unfilled; - } - - if (softpipe->setup.offset_cw || - softpipe->setup.offset_ccw) { - softpipe->prim.offset->next = next; - next = softpipe->prim.offset; - } - - if (softpipe->setup.light_twoside) { - softpipe->prim.twoside->next = next; - next = softpipe->prim.twoside; - } - - /* Always run the cull stage as we calculate determinant there - * also. Fix this.. - */ - { - softpipe->prim.cull->next = next; - next = softpipe->prim.cull; - } - - - /* Clip stage - */ - { - softpipe->prim.clip->next = next; - next = softpipe->prim.clip; - } - - /* Do software flatshading prior to clipping. XXX: should only do - * this for clipped primitives, ie it is a part of the clip - * routine. - */ - if (softpipe->setup.flatshade) { - softpipe->prim.flatshade->next = next; - next = softpipe->prim.flatshade; - } - - - softpipe->prim.first = next; -} - -#endif - - - void softpipe_set_setup_state( struct pipe_context *pipe, const struct pipe_setup_state *setup ) { @@ -100,9 +41,6 @@ void softpipe_set_setup_state( struct pipe_context *pipe, memcpy( &softpipe->setup, setup, sizeof(*setup) ); -#if 0 - validate_prim_pipe( softpipe ); -#endif softpipe->dirty |= SP_NEW_SETUP; } -- cgit v1.2.3 From c78f372b1720cf17b0112d7c3296020044bef97b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 13:43:33 -0600 Subject: s/prim_setup/sp_draw_render_stage/, new comments --- src/mesa/pipe/softpipe/sp_context.c | 4 ++-- src/mesa/pipe/softpipe/sp_prim_setup.c | 27 ++++++++++++++++++--------- src/mesa/pipe/softpipe/sp_prim_setup.h | 21 ++++++++------------- 3 files changed, 28 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index d3aea3698a..9511254a96 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -93,10 +93,10 @@ struct pipe_context *softpipe_create( void ) softpipe->quad.output = sp_quad_output_stage(softpipe); /* - * Create drawing context and plug our render/setup stage into it. + * Create drawing context and plug our rendering stage into it. */ softpipe->draw = draw_create(); - draw_set_setup_stage(softpipe->draw, prim_setup(softpipe)); + draw_set_setup_stage(softpipe->draw, sp_draw_render_stage(softpipe)); /* * XXX we could plug GL selection/feedback into the drawing pipeline diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 5db05ecccb..e94ca139aa 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -127,14 +127,6 @@ static inline GLint block( GLint x ) -static void setup_begin( struct draw_stage *stage ) -{ - struct setup_stage *setup = setup_stage(stage); - - setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; -} - - /** * Run shader on a quad/block. */ @@ -928,15 +920,32 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) +static void setup_begin( struct draw_stage *stage ) +{ + struct setup_stage *setup = setup_stage(stage); + + setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; + + /* + * XXX this is where we might map() the renderbuffers to begin + * s/w rendering. + */ +} + + static void setup_end( struct draw_stage *stage ) { + /* + * XXX this is where we might unmap() the renderbuffers after + * s/w rendering. + */ } /** * Create a new primitive setup/render stage. */ -struct draw_stage *prim_setup( struct softpipe_context *softpipe ) +struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) { struct setup_stage *setup = CALLOC_STRUCT(setup_stage); diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h index be00b0182b..e5abefcd49 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.h +++ b/src/mesa/pipe/softpipe/sp_prim_setup.h @@ -22,8 +22,10 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#ifndef S_TRI_H -#define S_TRI_H + +#ifndef SP_PRIM_SETUP_H +#define SP_PRIM_SETUP_H + /* Vertices are just an array of floats, with all the attributes * packed. We currently assume a layout like: @@ -43,7 +45,7 @@ #include "s_context.h" -extern struct draw_stage *prim_setup( struct softpipe_context *softpipe ); +extern struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ); #if 0 /* UNUSED? */ @@ -104,11 +106,6 @@ void tri_triangle( struct tri_context *tri, #ifdef SETUP_PRIVATE - - - - - GLboolean tri_setup( struct tri_context *tri, const struct vertex *v0, const struct vertex *v1, @@ -117,11 +114,9 @@ GLboolean tri_setup( struct tri_context *tri, void tri_rasterize( struct tri_context *tri ); void tri_rasterize_spans( struct tri_context *tri ); - - - +#endif #endif -#endif -#endif + +#endif /* SP_PRIM_SETUP_H */ -- cgit v1.2.3 From e8ceb5a2eb174f5444eaf5f52925fa161ea7d0a9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 13:44:43 -0600 Subject: XXX comment about skipping stipple for points/lines --- src/mesa/pipe/softpipe/sp_quad_stipple.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index f9a3c0ba26..e77bbb81ee 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -13,6 +13,7 @@ /** * Apply polygon stipple to quads produced by triangle rasterization + * XXX we need to skip this for lines and points!!! */ static void stipple_quad(struct quad_stage *qs, struct quad_header *quad) -- cgit v1.2.3 From 8b6517abfc20dc508c2a4cb1f6f22f34a8a70805 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Thu, 12 Jul 2007 21:52:02 +0200 Subject: don't use driDrawable information directly, don't resize at makecurrent driDrawable information now isn't used in most parts of the driver. makecurrent only updates fb information when the context is new (driDrawable is still updated by the dri common code). --- src/mesa/drivers/dri/i915tex/intel_blit.c | 9 +++---- src/mesa/drivers/dri/i915tex/intel_buffers.c | 22 ++++----------- src/mesa/drivers/dri/i915tex/intel_context.c | 18 ++++++++----- src/mesa/drivers/dri/i915tex/intel_context.h | 1 - src/mesa/drivers/dri/i915tex/intel_fbo.c | 2 ++ src/mesa/drivers/dri/i915tex/intel_ioctl.c | 3 +-- src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c | 18 ++++++------- src/mesa/drivers/dri/i915tex/intel_pixel_copy.c | 33 +++++++++-------------- src/mesa/drivers/dri/i915tex/intel_pixel_draw.c | 27 ++++++++++--------- src/mesa/drivers/dri/i915tex/intel_pixel_read.c | 26 +++++++----------- src/mesa/drivers/dri/i915tex/intel_span.c | 15 +++++------ src/mesa/drivers/dri/i915tex/intel_tex_copy.c | 17 +++--------- 12 files changed, 77 insertions(+), 114 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c index 3e42d79980..2a0a639315 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ b/src/mesa/drivers/dri/i915tex/intel_blit.c @@ -91,7 +91,7 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, const struct intel_region *frontRegion = intelScreen->front_region; const struct intel_region *backRegion - = intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? + = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); const int backWidth = intel_fb->Base.Width; @@ -108,7 +108,6 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ ASSERT(frontRegion); ASSERT(backRegion); -// ASSERT(frontRegion->pitch == backRegion->pitch); ASSERT(frontRegion->cpp == backRegion->cpp); DBG("front pitch %d back pitch %d\n", @@ -405,6 +404,7 @@ intelClearWithBlit(GLcontext * ctx, GLbitfield mask) LOCK_HARDWARE(intel); if (intel->numClipRects) { + assert(intel->numClipRects == 1); GLint cx, cy, cw, ch; drm_clip_rect_t clear; int i; @@ -417,10 +417,9 @@ intelClearWithBlit(GLcontext * ctx, GLbitfield mask) if (fb->Name == 0) { /* clearing a window */ - /* flip top to bottom */ - clear.x1 = cx + intel->drawX; - clear.y1 = intel->driDrawable->y + intel->driDrawable->h - cy - ch; + clear.x1 = cx; + clear.y1 = intel->pClipRects->y2 - cy - ch; clear.x2 = clear.x1 + cw; clear.y2 = clear.y1 + ch; } diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 20e9521105..59b291a709 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -128,8 +128,6 @@ intel_readbuf_region(struct intel_context *intel) * Update the following fields for rendering to a user-created FBO: * intel->numClipRects * intel->pClipRects - * intel->drawX - * intel->drawY */ static void intelSetRenderbufferClipRects(struct intel_context *intel) @@ -142,8 +140,6 @@ intelSetRenderbufferClipRects(struct intel_context *intel) intel->fboRect.y2 = intel->ctx.DrawBuffer->Height; intel->numClipRects = 1; intel->pClipRects = &intel->fboRect; - intel->drawX = 0; - intel->drawY = 0; } /** @@ -154,18 +150,17 @@ intelSetRenderbufferClipRects(struct intel_context *intel) static void intelSetPrivbufClipRects(struct intel_context *intel) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - if (!dPriv) + if (!intel->ctx.DrawBuffer) { + fprintf(stderr, "%s: DrawBuffer not set!\n", __FUNCTION__); return; + } intel->fakeClipRect.x1 = 0; intel->fakeClipRect.y1 = 0; - intel->fakeClipRect.x2 = dPriv->w; - intel->fakeClipRect.y2 = dPriv->h; + intel->fakeClipRect.x2 = intel->ctx.DrawBuffer->Width; + intel->fakeClipRect.y2 = intel->ctx.DrawBuffer->Height; intel->numClipRects = 1; intel->pClipRects = &intel->fakeClipRect; - intel->drawX = 0; - intel->drawY = 0; } @@ -445,7 +440,6 @@ intelRotateWindow(struct intel_context *intel, int i; GLenum format, type; - int xOrig, yOrig; int origNumClipRects; drm_clip_rect_t *origRects; @@ -469,8 +463,6 @@ intelRotateWindow(struct intel_context *intel, /* save current drawing origin and cliprects (restored at end) */ - xOrig = intel->drawX; - yOrig = intel->drawY; origNumClipRects = intel->numClipRects; origRects = intel->pClipRects; @@ -481,8 +473,6 @@ intelRotateWindow(struct intel_context *intel, fullRect.y1 = 0; fullRect.x2 = screen->rotatedWidth; fullRect.y2 = screen->rotatedHeight; - intel->drawX = 0; - intel->drawY = 0; intel->numClipRects = 1; intel->pClipRects = &fullRect; @@ -565,8 +555,6 @@ intelRotateWindow(struct intel_context *intel, intel_batchbuffer_flush(intel->batch); /* restore original drawing origin and cliprects */ - intel->drawX = xOrig; - intel->drawY = yOrig; intel->numClipRects = origNumClipRects; intel->pClipRects = origRects; diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 531b4adde8..451c7fd1ba 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -370,7 +370,7 @@ intelInitContext(struct intel_context *intel, if (!havePools) return GL_FALSE; - + if (!_mesa_initialize_context(&intel->ctx, mesaVis, shareCtx, functions, (void *) intel)) @@ -572,6 +572,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv) { + GLuint updatebufsize = GL_FALSE; #if 0 if (driDrawPriv) { @@ -619,19 +620,22 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, } } - /* set GLframebuffer size to match window, if needed */ - driUpdateFramebufferSize(&intel->ctx, driDrawPriv); + /* only update GLframebuffer size to match window + if here for the first time */ + if (intel->ctx.FirstTimeCurrent) { + updatebufsize = GL_TRUE; + driUpdateFramebufferSize(&intel->ctx, driDrawPriv); - if (driReadPriv != driDrawPriv) { - driUpdateFramebufferSize(&intel->ctx, driReadPriv); + if (driReadPriv != driDrawPriv) { + driUpdateFramebufferSize(&intel->ctx, driReadPriv); + } } _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); - intel->intelScreen->dummyctxptr = &intel->ctx; /* The drawbuffer won't always be updated by _mesa_make_current: */ - if (intel->ctx.DrawBuffer == &intel_fb->Base) { + if (updatebufsize && intel->ctx.DrawBuffer == &intel_fb->Base) { if (intel->driDrawable != driDrawPriv) { intel_fb->vblank_flags = (intel->intelScreen->irq_active != 0) diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index 7b97c564a3..f85b3a6362 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -250,7 +250,6 @@ struct intel_context /* These refer to the current drawing buffer: */ - int drawX, drawY; /**< origin of drawing area within region */ GLuint numClipRects; /**< cliprects for drawing */ drm_clip_rect_t *pClipRects; drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c index 2dd3617f93..aeb909cce1 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c @@ -146,6 +146,8 @@ intel_delete_renderbuffer(struct gl_renderbuffer *rb) ASSERT(irb); + DBG("freeing renderbuffer\n"); + if (irb->PairedStencil || irb->PairedDepth) { intel_unpair_depth_stencil(ctx, irb); } diff --git a/src/mesa/drivers/dri/i915tex/intel_ioctl.c b/src/mesa/drivers/dri/i915tex/intel_ioctl.c index 3250c6b3a9..54721c78ab 100644 --- a/src/mesa/drivers/dri/i915tex/intel_ioctl.c +++ b/src/mesa/drivers/dri/i915tex/intel_ioctl.c @@ -116,8 +116,7 @@ intel_batch_ioctl(struct intel_context *intel, batch.cliprects = intel->pClipRects; batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; batch.DR1 = 0; - batch.DR4 = ((((GLuint) intel->drawX) & 0xffff) | - (((GLuint) intel->drawY) << 16)); + batch.DR4 = 0; /* still need this ? */ DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n", __FUNCTION__, diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c b/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c index 65bf338589..96d3cbef5f 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c @@ -195,18 +195,18 @@ do_blit_bitmap( GLcontext *ctx, LOCK_HARDWARE(intel); - if (intel->driDrawable->numClipRects) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - drm_clip_rect_t *box = dPriv->pClipRects; + if (intel->numClipRects) { + assert(intel->numClipRects == 1); + drm_clip_rect_t *box = intel->pClipRects; drm_clip_rect_t dest_rect; - GLint nbox = dPriv->numClipRects; + GLint nbox = intel->numClipRects; GLint srcx = 0, srcy = 0; GLint orig_screen_x1, orig_screen_y2; GLuint i; - orig_screen_x1 = dPriv->x + dstx; - orig_screen_y2 = dPriv->y + (dPriv->h - dsty); + orig_screen_x1 = dstx; + orig_screen_y2 = box->y2 - dsty; /* Do scissoring in GL coordinates: */ @@ -223,8 +223,8 @@ x if (ctx->Scissor.Enabled) /* Convert from GL to hardware coordinates: */ - dsty = dPriv->y + (dPriv->h - dsty - height); - dstx = dPriv->x + dstx; + dsty = box->y2 - dsty - height; + dstx = dstx; dest_rect.x1 = dstx; dest_rect.y1 = dsty; @@ -235,7 +235,7 @@ x if (ctx->Scissor.Enabled) drm_clip_rect_t rect; int box_w, box_h; GLint px, py; - GLuint stipple[32]; + GLuint stipple[32]; if (!intel_intersect_cliprects(&rect, &dest_rect, &box[i])) continue; diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c index 9d478283e4..03c7353ee1 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c @@ -191,14 +191,10 @@ do_texture_copypixels(GLcontext * ctx, LOCK_HARDWARE(intel); - if (intel->driDrawable->numClipRects) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - - srcy = dPriv->h - srcy - height; /* convert from gl to hardware coords */ - - srcx += dPriv->x; - srcy += dPriv->y; + if (intel->numClipRects) { + int bufHeight = intel->pClipRects->y2; + assert (intel->numClipRects == 1); + srcy = bufHeight - srcy - height; /* convert from gl to hardware coords */ /* Clip against the source region. This is the only source * clipping we do. XXX: Just set the texcord wrap mode to clamp @@ -223,8 +219,8 @@ do_texture_copypixels(GLcontext * ctx, intel_meta_draw_quad(intel, dstx, dstx + width * ctx->Pixel.ZoomX, - dPriv->h - (dsty + height * ctx->Pixel.ZoomY), - dPriv->h - (dsty), 0, /* XXX: what z value? */ + bufHeight - (dsty + height * ctx->Pixel.ZoomY), + bufHeight - (dsty), 0, /* XXX: what z value? */ 0x00ff00ff, srcx, srcx + width, srcy, srcy + height); @@ -271,11 +267,11 @@ do_blit_copypixels(GLcontext * ctx, LOCK_HARDWARE(intel); - if (intel->driDrawable->numClipRects) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - drm_clip_rect_t *box = dPriv->pClipRects; + if (intel->numClipRects) { + assert(intel->numClipRects == 1); + drm_clip_rect_t *box = intel->pClipRects; drm_clip_rect_t dest_rect; - GLint nbox = dPriv->numClipRects; + GLint nbox = intel->numClipRects; GLint delta_x = 0; GLint delta_y = 0; GLuint i; @@ -299,13 +295,10 @@ do_blit_copypixels(GLcontext * ctx, } /* Convert from GL to hardware coordinates: + * cliprect should be size of buffer */ - dsty = dPriv->h - dsty - height; - srcy = dPriv->h - srcy - height; - dstx += dPriv->x; - dsty += dPriv->y; - srcx += dPriv->x; - srcy += dPriv->y; + dsty = box->y2 - dsty - height; + srcy = box->y2 - srcy - height; /* Clip against the source region. This is the only source * clipping we do. Dst is clipped with cliprects below. diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c index 77c67c821e..2d46c4b8c0 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c @@ -138,13 +138,15 @@ do_texture_drawpixels(GLcontext * ctx, LOCK_HARDWARE(intel); - if (intel->driDrawable->numClipRects) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; + if (intel->numClipRects) { + assert(intel->numClipRects == 1); + int bufHeight = intel->pClipRects->y2; + GLint srcx, srcy; GLint dstx, dsty; dstx = x; - dsty = dPriv->h - (y + height); + dsty = bufHeight - (y + height); srcx = 0; /* skiprows/pixels already done */ srcy = 0; @@ -172,8 +174,8 @@ do_texture_drawpixels(GLcontext * ctx, */ intel_meta_draw_quad(intel, dstx, dstx + width * ctx->Pixel.ZoomX, - dPriv->h - (y + height * ctx->Pixel.ZoomY), - dPriv->h - (y), + bufHeight - (y + height * ctx->Pixel.ZoomY), + bufHeight - (y), -ctx->Current.RasterPos[2] * .5, 0x00ff00ff, srcx, srcx + width, srcy + height, srcy); @@ -289,21 +291,21 @@ do_blit_drawpixels(GLcontext * ctx, src_offset = (GLuint) _mesa_image_address(2, unpack, pixels, width, height, format, type, 0, 0, 0); + /* don't need a lock as cliprects shouldn't change */ intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); - if (intel->driDrawable->numClipRects) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - int nbox = dPriv->numClipRects; - drm_clip_rect_t *box = dPriv->pClipRects; + if (intel->numClipRects) { + assert(intel->numClipRects == 1); + int nbox = intel->numClipRects; + drm_clip_rect_t *box = intel->pClipRects; drm_clip_rect_t rect; drm_clip_rect_t dest_rect; struct _DriBufferObject *src_buffer = intel_bufferobj_buffer(intel, src, INTEL_READ); int i; - dest_rect.x1 = dPriv->x + x; - dest_rect.y1 = dPriv->y + dPriv->h - (y + height); + dest_rect.x1 = x; + dest_rect.y1 = box->y2 - (y + height); dest_rect.x2 = dest_rect.x1 + width; dest_rect.y2 = dest_rect.y1 + height; @@ -327,7 +329,6 @@ do_blit_drawpixels(GLcontext * ctx, fence = intel_batchbuffer_flush(intel->batch); driFenceReference(fence); } - UNLOCK_HARDWARE(intel); if (fence) { driFenceFinish(fence, DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c b/src/mesa/drivers/dri/i915tex/intel_pixel_read.c index 24e49ae066..9481cb4800 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_read.c @@ -77,7 +77,6 @@ do_texture_readpixels(GLcontext * ctx, struct intel_context *intel = intel_context(ctx); intelScreenPrivate *screen = intel->intelScreen; GLint pitch = pack->RowLength ? pack->RowLength : width; - __DRIdrawablePrivate *dPriv = intel->driDrawable; int textureFormat; GLenum glTextureFormat; int destFormat, depthFormat, destPitch; @@ -107,7 +106,7 @@ do_texture_readpixels(GLcontext * ctx, LOCK_HARDWARE(intel); - if (intel->driDrawable->numClipRects) { + if (intel->numClipRects) { intel->vtbl.install_meta_state(intel); intel->vtbl.meta_no_depth_write(intel); intel->vtbl.meta_no_stencil_write(intel); @@ -120,9 +119,7 @@ do_texture_readpixels(GLcontext * ctx, return GL_TRUE; } - y = dPriv->h - y - height; - x += dPriv->x; - y += dPriv->y; + y = intel->pClipRects->y2 - y - height; /* Set the frontbuffer up as a large rectangular texture. @@ -230,29 +227,25 @@ do_blit_readpixels(GLcontext * ctx, dst_offset = (GLuint) _mesa_image_address(2, pack, pixels, width, height, format, type, 0, 0, 0); - - /* Although the blits go on the command buffer, need to do this and - * fire with lock held to guarentee cliprects are correct. - */ + /* reading from priv buffer, cliprects should not change */ intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); - if (intel->driDrawable->numClipRects) { + if (intel->numClipRects) { + assert (intel->numClipRects == 1); GLboolean all = (width * height * src->cpp == dst->Base.Size && x == 0 && dst_offset == 0); struct _DriBufferObject *dst_buffer = intel_bufferobj_buffer(intel, dst, all ? INTEL_WRITE_FULL : INTEL_WRITE_PART); - __DRIdrawablePrivate *dPriv = intel->driDrawable; - int nbox = dPriv->numClipRects; - drm_clip_rect_t *box = dPriv->pClipRects; + int nbox = intel->numClipRects; + drm_clip_rect_t *box = intel->pClipRects; drm_clip_rect_t rect; drm_clip_rect_t src_rect; int i; - src_rect.x1 = dPriv->x + x; - src_rect.y1 = dPriv->y + dPriv->h - (y + height); + src_rect.x1 = x; + src_rect.y1 = box->y2 - (y + height); src_rect.x2 = src_rect.x1 + width; src_rect.y2 = src_rect.y1 + height; @@ -279,7 +272,6 @@ do_blit_readpixels(GLcontext * ctx, driFenceReference(fence); } - UNLOCK_HARDWARE(intel); if (fence) { driFenceFinish(fence, DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, diff --git a/src/mesa/drivers/dri/i915tex/intel_span.c b/src/mesa/drivers/dri/i915tex/intel_span.c index ab0874e4fd..368e0ec3c0 100644 --- a/src/mesa/drivers/dri/i915tex/intel_span.c +++ b/src/mesa/drivers/dri/i915tex/intel_span.c @@ -51,8 +51,7 @@ struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ const GLint yScale = irb->RenderToTexture ? 1 : -1; \ const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ - GLubyte *buf = (GLubyte *) irb->pfMap \ - + (intel->drawY * irb->pfPitch + intel->drawX) * irb->region->cpp;\ + GLubyte *buf = (GLubyte *) irb->pfMap; \ GLuint p; \ assert(irb->pfMap);\ (void) p; @@ -65,10 +64,10 @@ do { \ int _nc = intel->numClipRects; \ while ( _nc-- ) { \ - int minx = intel->pClipRects[_nc].x1 - intel->drawX; \ - int miny = intel->pClipRects[_nc].y1 - intel->drawY; \ - int maxx = intel->pClipRects[_nc].x2 - intel->drawX; \ - int maxy = intel->pClipRects[_nc].y2 - intel->drawY; + int minx = intel->pClipRects[_nc].x1; \ + int miny = intel->pClipRects[_nc].y1; \ + int maxx = intel->pClipRects[_nc].x2; \ + int maxy = intel->pClipRects[_nc].y2; @@ -106,9 +105,7 @@ const GLuint pitch = irb->pfPitch/***XXX region->pitch*/; /* in pixels */ \ const GLint yScale = irb->RenderToTexture ? 1 : -1; \ const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ - char *buf = (char *) irb->pfMap/*XXX use region->map*/ + \ - (intel->drawY * pitch + intel->drawX) * irb->region->cpp; - + char *buf = (char *) irb->pfMap/*XXX use region->map*/ ; #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_copy.c b/src/mesa/drivers/dri/i915tex/intel_tex_copy.c index b85a25642a..90c2b9ebb5 100644 --- a/src/mesa/drivers/dri/i915tex/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915tex/intel_tex_copy.c @@ -101,6 +101,7 @@ do_copy_texsubimage(struct intel_context *intel, } intelFlush(ctx); + /* XXX still need the lock ? */ LOCK_HARDWARE(intel); { GLuint image_offset = intel_miptree_image_offset(intelImage->mt, @@ -117,20 +118,8 @@ do_copy_texsubimage(struct intel_context *intel, dstx += x - orig_x; dsty += y - orig_y; - if (ctx->ReadBuffer->Name == 0) { - /* reading from a window, adjust x, y */ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - GLuint window_y; - /* window_y = position of window on screen if y=0=bottom */ - window_y = intel->intelScreen->height - (dPriv->y + dPriv->h); - y = window_y + y; - x += dPriv->x; - } - else { - /* reading from a FBO */ - /* invert Y */ - y = ctx->ReadBuffer->Height - y - 1; - } + /* invert Y */ + y = ctx->ReadBuffer->Height - y - 1; /* A bit of fiddling to get the blitter to work with -ve -- cgit v1.2.3 From a48d767cf2ec82d0c0f893a22d22e8593901b206 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 14:13:48 -0600 Subject: Add a quad 'bufloop' stage to handle glDrawBuffer(GL_FRONT_AND_BACK). This removes the notion of multiple color buffers from all other stages. Will need a bit more work when shaders with multiple render targets arrive. --- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_context.h | 3 + src/mesa/pipe/softpipe/sp_quad.c | 10 + src/mesa/pipe/softpipe/sp_quad.h | 1 + src/mesa/pipe/softpipe/sp_quad_blend.c | 536 ++++++++++++++--------------- src/mesa/pipe/softpipe/sp_quad_bufloop.c | 59 ++++ src/mesa/pipe/softpipe/sp_quad_colormask.c | 6 +- src/mesa/pipe/softpipe/sp_quad_output.c | 28 +- src/mesa/sources | 1 + 9 files changed, 351 insertions(+), 294 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_quad_bufloop.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 9511254a96..cc4b8d5914 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -88,6 +88,7 @@ struct pipe_context *softpipe_create( void ) softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); + softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); softpipe->quad.blend = sp_quad_blend_stage(softpipe); softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 3ae1de71cf..81d2b58cf9 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -122,6 +122,7 @@ struct softpipe_context { struct quad_stage *alpha_test; struct quad_stage *stencil_test; struct quad_stage *depth_test; + struct quad_stage *bufloop; struct quad_stage *blend; struct quad_stage *colormask; struct quad_stage *output; @@ -131,6 +132,8 @@ struct softpipe_context { /** The primitive drawing context */ struct draw_context *draw; + + struct pipe_surface *cbuf; /**< current color buffer (one of cbufs) */ }; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 419a720de9..5f34f5f1b3 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -21,6 +21,16 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.blend; } + if (sp->framebuffer.num_cbufs == 1) { + /* the usual case: write to exactly one colorbuf */ + sp->cbuf = sp->framebuffer.cbufs[0]; + } + else { + /* insert bufloop stage */ + sp->quad.bufloop->next = sp->quad.first; + sp->quad.first = sp->quad.bufloop; + } + if ( sp->stencil.front_enabled || sp->stencil.front_enabled) { sp->quad.stencil_test->next = sp->quad.first; diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 966f72d3b5..719d4774cc 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -51,6 +51,7 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 2599e7a2f9..3d097aea65 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -104,287 +104,281 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) static const GLfloat zero[4] = { 0, 0, 0, 0 }; static const GLfloat one[4] = { 1, 1, 1, 1 }; struct softpipe_context *softpipe = qs->softpipe; + struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); GLfloat source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; - GLuint i; - - /* XXX we're also looping in output_quad() !?! */ - - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps - = softpipe_surface(softpipe->framebuffer.cbufs[i]); - sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + /* get colors from framebuffer */ + sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); - /* - * Compute src/first term RGB - */ - switch (softpipe->blend.rgb_src_factor) { - case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[0], quad->outputs.color[0]); /* R */ - VEC4_COPY(source[1], quad->outputs.color[1]); /* G */ - VEC4_COPY(source[2], quad->outputs.color[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - VEC4_MUL(source[0], quad->outputs.color[0], quad->outputs.color[0]); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], quad->outputs.color[1]); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], quad->outputs.color[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - { - const GLfloat *alpha = quad->outputs.color[3]; - VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_DST_COLOR: - VEC4_MUL(source[0], quad->outputs.color[0], dest[0]); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], dest[1]); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], dest[2]); /* B */ - break; - case PIPE_BLENDFACTOR_DST_ALPHA: - { - const GLfloat *alpha = dest[3]; - VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - { - GLfloat comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ - VEC4_MUL(source[0], quad->outputs.color[0], comp); /* R */ - VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ - VEC4_MUL(source[1], quad->outputs.color[1], comp); /* G */ - VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ - VEC4_MUL(source[2], quad->outputs.color[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - GLfloat alpha[4]; - VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_SRC1_COLOR: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_SRC1_ALPHA: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(source[0], zero); /* R */ - VEC4_COPY(source[1], zero); /* G */ - VEC4_COPY(source[2], zero); /* B */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - { - GLfloat inv_comp[4]; - VEC4_SUB(inv_comp, one, quad->outputs.color[0]); /* R */ - VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, quad->outputs.color[1]); /* G */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, quad->outputs.color[2]); /* B */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - GLfloat inv_alpha[4]; - VEC4_SUB(inv_alpha, one, quad->outputs.color[3]); - VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - GLfloat inv_alpha[4]; - VEC4_SUB(inv_alpha, one, dest[3]); - VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - { - GLfloat inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[0]); /* R */ - VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, dest[1]); /* G */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, dest[2]); /* B */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - { - GLfloat inv_comp[4]; - /* R */ - VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[0]); - VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); - /* G */ - VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[1]); - VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); - /* B */ - VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[2]); - VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); - } - break; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - GLfloat alpha[4], inv_alpha[4]; - VEC4_SCALAR(alpha, 1.0 - softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: - assert(0); /* to do */ - break; - default: - abort(); + /* + * Compute src/first term RGB + */ + switch (softpipe->blend.rgb_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[0], quad->outputs.color[0]); /* R */ + VEC4_COPY(source[1], quad->outputs.color[1]); /* G */ + VEC4_COPY(source[2], quad->outputs.color[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + VEC4_MUL(source[0], quad->outputs.color[0], quad->outputs.color[0]); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], quad->outputs.color[1]); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], quad->outputs.color[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const GLfloat *alpha = quad->outputs.color[3]; + VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ } - - /* - * Compute src/first term A - */ - switch (softpipe->blend.alpha_src_factor) { - case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[3], quad->outputs.color[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - { - const GLfloat *alpha = quad->outputs.color[3]; - VEC4_MUL(source[3], quad->outputs.color[3], alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(source[3], zero); /* A */ - break; - /* XXX fill in remaining terms */ - default: - abort(); + break; + case PIPE_BLENDFACTOR_DST_COLOR: + VEC4_MUL(source[0], quad->outputs.color[0], dest[0]); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], dest[1]); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], dest[2]); /* B */ + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + { + const GLfloat *alpha = dest[3]; + VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ } - - - /* - * Compute dest/second term RGB - */ - switch (softpipe->blend.rgb_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - /* dest = dest * 1 NO-OP, leave dest as-is */ - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - GLfloat one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); - VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ - VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ - VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[0], zero); /* R */ - VEC4_COPY(dest[1], zero); /* G */ - VEC4_COPY(dest[2], zero); /* B */ - break; - /* XXX fill in remaining terms */ - default: - abort(); + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + { + GLfloat comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ + VEC4_MUL(source[0], quad->outputs.color[0], comp); /* R */ + VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ + VEC4_MUL(source[1], quad->outputs.color[1], comp); /* G */ + VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ + VEC4_MUL(source[2], quad->outputs.color[2], comp); /* B */ } - - /* - * Compute dest/second term A - */ - switch (softpipe->blend.alpha_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - /* dest = dest * 1 NO-OP, leave dest as-is */ - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - GLfloat one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); - VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[3], zero); /* A */ - break; - /* XXX fill in remaining terms */ - default: - abort(); + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + GLfloat alpha[4]; + VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ } - - /* - * Combine RGB terms - */ - switch (softpipe->blend.rgb_func) { - case PIPE_BLEND_ADD: - VEC4_ADD(quad->outputs.color[0], source[0], dest[0]); /* R */ - VEC4_ADD(quad->outputs.color[1], source[1], dest[1]); /* G */ - VEC4_ADD(quad->outputs.color[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quad->outputs.color[0], source[0], dest[0]); /* R */ - VEC4_SUB(quad->outputs.color[1], source[1], dest[1]); /* G */ - VEC4_SUB(quad->outputs.color[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quad->outputs.color[0], dest[0], source[0]); /* R */ - VEC4_SUB(quad->outputs.color[1], dest[1], source[1]); /* G */ - VEC4_SUB(quad->outputs.color[2], dest[2], source[2]); /* B */ - break; - case PIPE_BLEND_MIN: - VEC4_MIN(quad->outputs.color[0], source[0], dest[0]); /* R */ - VEC4_MIN(quad->outputs.color[1], source[1], dest[1]); /* G */ - VEC4_MIN(quad->outputs.color[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_MAX: - VEC4_MAX(quad->outputs.color[0], source[0], dest[0]); /* R */ - VEC4_MAX(quad->outputs.color[1], source[1], dest[1]); /* G */ - VEC4_MAX(quad->outputs.color[2], source[2], dest[2]); /* B */ - break; - default: - abort(); + break; + case PIPE_BLENDFACTOR_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_SRC1_ALPHA: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[0], zero); /* R */ + VEC4_COPY(source[1], zero); /* G */ + VEC4_COPY(source[2], zero); /* B */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + { + GLfloat inv_comp[4]; + VEC4_SUB(inv_comp, one, quad->outputs.color[0]); /* R */ + VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, quad->outputs.color[1]); /* G */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, quad->outputs.color[2]); /* B */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); /* B */ } - - /* - * Combine A terms - */ - switch (softpipe->blend.alpha_func) { - case PIPE_BLEND_ADD: - VEC4_ADD(quad->outputs.color[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quad->outputs.color[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quad->outputs.color[3], dest[3], source[3]); /* A */ - break; - case PIPE_BLEND_MIN: - VEC4_MIN(quad->outputs.color[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_MAX: - VEC4_MAX(quad->outputs.color[3], source[3], dest[3]); /* A */ - default: - abort(); + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + GLfloat inv_alpha[4]; + VEC4_SUB(inv_alpha, one, quad->outputs.color[3]); + VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ } - - /* pass blended quad to next stage */ - qs->next->run(qs->next, quad); + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + GLfloat inv_alpha[4]; + VEC4_SUB(inv_alpha, one, dest[3]); + VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + { + GLfloat inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[0]); /* R */ + VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, dest[1]); /* G */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, dest[2]); /* B */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + { + GLfloat inv_comp[4]; + /* R */ + VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[0]); + VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); + /* G */ + VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[1]); + VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); + /* B */ + VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[2]); + VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + GLfloat alpha[4], inv_alpha[4]; + VEC4_SCALAR(alpha, 1.0 - softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + assert(0); /* to do */ + break; + default: + abort(); + } + + /* + * Compute src/first term A + */ + switch (softpipe->blend.alpha_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[3], quad->outputs.color[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const GLfloat *alpha = quad->outputs.color[3]; + VEC4_MUL(source[3], quad->outputs.color[3], alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[3], zero); /* A */ + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + + /* + * Compute dest/second term RGB + */ + switch (softpipe->blend.rgb_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + GLfloat one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); + VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ + VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ + VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[0], zero); /* R */ + VEC4_COPY(dest[1], zero); /* G */ + VEC4_COPY(dest[2], zero); /* B */ + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + /* + * Compute dest/second term A + */ + switch (softpipe->blend.alpha_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + GLfloat one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); + VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[3], zero); /* A */ + break; + /* XXX fill in remaining terms */ + default: + abort(); + } + + /* + * Combine RGB terms + */ + switch (softpipe->blend.rgb_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quad->outputs.color[0], source[0], dest[0]); /* R */ + VEC4_ADD(quad->outputs.color[1], source[1], dest[1]); /* G */ + VEC4_ADD(quad->outputs.color[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quad->outputs.color[0], source[0], dest[0]); /* R */ + VEC4_SUB(quad->outputs.color[1], source[1], dest[1]); /* G */ + VEC4_SUB(quad->outputs.color[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quad->outputs.color[0], dest[0], source[0]); /* R */ + VEC4_SUB(quad->outputs.color[1], dest[1], source[1]); /* G */ + VEC4_SUB(quad->outputs.color[2], dest[2], source[2]); /* B */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quad->outputs.color[0], source[0], dest[0]); /* R */ + VEC4_MIN(quad->outputs.color[1], source[1], dest[1]); /* G */ + VEC4_MIN(quad->outputs.color[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quad->outputs.color[0], source[0], dest[0]); /* R */ + VEC4_MAX(quad->outputs.color[1], source[1], dest[1]); /* G */ + VEC4_MAX(quad->outputs.color[2], source[2], dest[2]); /* B */ + break; + default: + abort(); } + + /* + * Combine A terms + */ + switch (softpipe->blend.alpha_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quad->outputs.color[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quad->outputs.color[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quad->outputs.color[3], dest[3], source[3]); /* A */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quad->outputs.color[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quad->outputs.color[3], source[3], dest[3]); /* A */ + default: + abort(); + } + + /* pass blended quad to next stage */ + qs->next->run(qs->next, quad); } diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c new file mode 100644 index 0000000000..977cc74c26 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -0,0 +1,59 @@ + + +#include "main/glheader.h" +#include "main/imports.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" + + +/** + * Loop over colorbuffers, passing quad to next stage each time. + */ +static void +cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + const GLuint sz = sizeof(quad->outputs.color); + GLfloat tmp[4][QUAD_SIZE]; + GLuint i; + + assert(sz == sizeof(tmp)); + + /* make copy of original colors since they can get modified + * by blending and masking. + */ + memcpy(tmp, quad->outputs.color, sz); + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + /* set current cbuffer */ + softpipe->cbuf = softpipe->framebuffer.cbufs[i]; + + /* pass blended quad to next stage */ + qs->next->run(qs->next, quad); + + if (i + 1 < softpipe->framebuffer.num_cbufs) { + /* restore quad's colors for next buffer */ + memcpy(quad->outputs.color, tmp, sz); + } + } + + softpipe->cbuf = NULL; /* prevent accidental use */ +} + + +/** + * Create the colorbuffer loop stage. + * This is used to implement GL_FRONT_AND_BACK rendering. + */ +struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = cbuf_loop_quad; + + return stage; +} + diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index 6fb228fdda..fff6efa8f6 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -45,12 +45,8 @@ static void colormask_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; + struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); GLfloat dest[4][QUAD_SIZE]; - - /* XXX buffer looping */ - - struct softpipe_surface *sps - = softpipe_surface(softpipe->framebuffer.cbufs[0]); sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index e898757a06..12ab1eca1c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -65,28 +65,21 @@ static void output_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - GLuint i; + struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps - = softpipe_surface(softpipe->framebuffer.cbufs[i]); + if (quad->mask != MASK_ALL) { + GLfloat tmp[4][QUAD_SIZE]; - if (quad->mask != MASK_ALL) { - GLfloat tmp[4][QUAD_SIZE]; + /* XXX probably add a masked-write function someday */ - /* Yes, we'll probably have a masked write as well, but this is - * how blend will be done at least. - */ + sps->read_quad_f_swz(sps, quad->x0, quad->y0, tmp); - sps->read_quad_f_swz(sps, quad->x0, quad->y0, tmp); + mask_copy( tmp, quad->outputs.color, quad->mask ); - mask_copy( tmp, quad->outputs.color, quad->mask ); - - sps->write_quad_f_swz(sps, quad->x0, quad->y0, tmp); - } - else { - sps->write_quad_f_swz(sps, quad->x0, quad->y0, quad->outputs.color); - } + sps->write_quad_f_swz(sps, quad->x0, quad->y0, tmp); + } + else if (quad->mask) { + sps->write_quad_f_swz(sps, quad->x0, quad->y0, quad->outputs.color); } } @@ -100,4 +93,3 @@ struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ) return stage; } - diff --git a/src/mesa/sources b/src/mesa/sources index 78a030b767..4856f7b023 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -160,6 +160,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_quad.c \ pipe/softpipe/sp_quad_alpha_test.c \ pipe/softpipe/sp_quad_blend.c \ + pipe/softpipe/sp_quad_bufloop.c \ pipe/softpipe/sp_quad_colormask.c \ pipe/softpipe/sp_quad_depth_test.c \ pipe/softpipe/sp_quad_fs.c \ -- cgit v1.2.3 From f0f9a22609ccf2b8edc5760480f1a7a78cb504d7 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 12 Jul 2007 16:17:14 -0600 Subject: minor changes, comments about multiple color buffers, MRT. --- src/mesa/pipe/p_state.h | 10 ++++++---- src/mesa/pipe/softpipe/sp_quad_bufloop.c | 9 ++++++++- 2 files changed, 14 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 3dfe584af8..2b0d659993 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -49,7 +49,7 @@ #define PIPE_MAX_CLIP_PLANES 6 #define PIPE_MAX_CONSTANT 32 #define PIPE_ATTRIB_MAX 32 - +#define PIPE_MAX_COLOR_BUFS 8 /* fwd decl */ @@ -192,9 +192,10 @@ struct pipe_stencil_state { struct pipe_framebuffer_state { - GLuint num_cbufs; /**< Number of color bufs to draw to */ - struct pipe_surface *cbufs[4]; /**< OpenGL can write to as many as - 4 color buffers at once */ + /** multiple colorbuffers for multiple render targets */ + GLuint num_cbufs; + struct pipe_surface *cbufs[PIPE_MAX_COLOR_BUFS]; + struct pipe_surface *zbuf; /**< Z buffer */ struct pipe_surface *sbuf; /**< Stencil buffer */ struct pipe_surface *abuf; /**< Accum buffer */ @@ -221,6 +222,7 @@ struct pipe_sampler_state #if 0 /* need these? */ GLint BaseLevel; /**< min mipmap level, OpenGL 1.2 */ GLint MaxLevel; /**< max mipmap level, OpenGL 1.2 */ + GLfloat border_color[4]; #endif GLfloat max_anisotropy; }; diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c index 977cc74c26..76e7fd5889 100644 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -23,9 +23,15 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) /* make copy of original colors since they can get modified * by blending and masking. + * XXX we won't have to do this if the fragment program actually emits + * N separate colors and we're drawing to N color buffers (MRT). + * But if we emitted one color and glDrawBuffer(GL_FRONT_AND_BACK) is + * in effect, we need to save/restore colors like this. */ memcpy(tmp, quad->outputs.color, sz); + assert(softpipe->framebuffer.num_cbufs <= PIPE_MAX_COLOR_BUFS); + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { /* set current cbuffer */ softpipe->cbuf = softpipe->framebuffer.cbufs[i]; @@ -45,7 +51,8 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) /** * Create the colorbuffer loop stage. - * This is used to implement GL_FRONT_AND_BACK rendering. + * This is used to implement multiple render targets and GL_FRONT_AND_BACK + * rendering. */ struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ) { -- cgit v1.2.3 From 60a21f6c54788c0f22fd1cf4a4a13e1af389f3e6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 13 Jul 2007 11:27:57 +0100 Subject: Simplify slightly. Don't bother trying to save a small memcpy. --- src/mesa/pipe/softpipe/sp_quad_bufloop.c | 14 +++++--------- 1 file changed, 5 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c index 76e7fd5889..be32d02a46 100644 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -15,11 +15,11 @@ static void cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - const GLuint sz = sizeof(quad->outputs.color); GLfloat tmp[4][QUAD_SIZE]; GLuint i; - assert(sz == sizeof(tmp)); + assert(sizeof(quad->outputs.color) == sizeof(tmp)); + assert(softpipe->framebuffer.num_cbufs <= PIPE_MAX_COLOR_BUFS); /* make copy of original colors since they can get modified * by blending and masking. @@ -28,9 +28,7 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) * But if we emitted one color and glDrawBuffer(GL_FRONT_AND_BACK) is * in effect, we need to save/restore colors like this. */ - memcpy(tmp, quad->outputs.color, sz); - - assert(softpipe->framebuffer.num_cbufs <= PIPE_MAX_COLOR_BUFS); + memcpy(tmp, quad->outputs.color, sizeof(tmp)); for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { /* set current cbuffer */ @@ -39,10 +37,8 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) /* pass blended quad to next stage */ qs->next->run(qs->next, quad); - if (i + 1 < softpipe->framebuffer.num_cbufs) { - /* restore quad's colors for next buffer */ - memcpy(quad->outputs.color, tmp, sz); - } + /* restore quad's colors for next buffer */ + memcpy(quad->outputs.color, tmp, sizeof(tmp)); } softpipe->cbuf = NULL; /* prevent accidental use */ -- cgit v1.2.3 From aa0f415c8d5d79d0763fca49d91b1963bf0975f6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 13 Jul 2007 11:28:42 +0100 Subject: Add conditional-free implementation (disabled, not tested) --- src/mesa/pipe/softpipe/sp_quad_stipple.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index e77bbb81ee..928771e454 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -23,13 +23,15 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) const GLint row0 = quad->y0 % 32; const GLuint stipple0 = softpipe->poly_stipple.stipple[row0]; const GLuint stipple1 = softpipe->poly_stipple.stipple[row0 + 1]; + + /* XXX this should be acheivable without conditionals */ +#if 1 GLbitfield mask = 0x0; - /* XXX this could be optimize a bit to use just two conditionals */ if ((1 << col0) & stipple0) mask |= MASK_BOTTOM_LEFT; - if ((2 << col0) & stipple0) + if ((2 << col0) & stipple0) /* note: col0 <= 30 */ mask |= MASK_BOTTOM_RIGHT; if ((1 << col0) & stipple1) @@ -39,6 +41,13 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) mask |= MASK_TOP_RIGHT; quad->mask &= mask; +#else + /* XXX there may be a better way to lay out the stored stipple + * values to further simplify this computation. + */ + quad->mask &= (((stipple0 >> col0) & 0x3) | + (((stipple1 >> col0) & 0x3) << 2)); +#endif if (quad->mask) qs->next->run(qs->next, quad); -- cgit v1.2.3 From 676fcf0de270651311a6b812e98f2e71d7dae925 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 09:23:41 -0600 Subject: minor rename --- src/mesa/drivers/x11/xm_surface.c | 2 +- src/mesa/pipe/p_defines.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 8ae5c248f4..58081a36a0 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -330,7 +330,7 @@ create_stencil_surface(XMesaContext xmctx, struct gl_renderbuffer *rb) xmsurf = CALLOC_STRUCT(xmesa_surface); if (xmsurf) { - xmsurf->sps.surface.format = PIPE_FORMAT_S8; + xmsurf->sps.surface.format = PIPE_FORMAT_U_S8; xmsurf->sps.surface.width = rb->Width; xmsurf->sps.surface.height = rb->Height; xmsurf->sps.read_quad_stencil = read_quad_stencil; diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 821521ac75..1b799f1451 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -143,7 +143,7 @@ #define PIPE_FORMAT_F_Z32 8 /**< float Z/depth */ #define PIPE_FORMAT_YCBCR 9 #define PIPE_FORMAT_YCBCR_REV 10 -#define PIPE_FORMAT_S8 11 /**< 8-bit stencil */ +#define PIPE_FORMAT_U_S8 11 /**< 8-bit stencil */ #define PIPE_FORMAT_Z24_S8 12 /**< 24-bit Z + 8-bit stencil */ -- cgit v1.2.3 From 563479552e2f491fb94e7fac5772f3c72cee962a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 09:25:57 -0600 Subject: Added basic occlusion counting --- src/mesa/main/queryobj.c | 15 +++++++ src/mesa/pipe/p_context.h | 10 ++++- src/mesa/pipe/p_state.h | 1 + src/mesa/pipe/softpipe/sp_context.c | 17 ++++++++ src/mesa/pipe/softpipe/sp_context.h | 3 ++ src/mesa/pipe/softpipe/sp_quad.c | 5 +++ src/mesa/pipe/softpipe/sp_quad.h | 1 + src/mesa/pipe/softpipe/sp_quad_occlusion.c | 67 ++++++++++++++++++++++++++++++ src/mesa/sources | 1 + src/mesa/state_tracker/st_atom_depth.c | 4 ++ 10 files changed, 123 insertions(+), 1 deletion(-) create mode 100644 src/mesa/pipe/softpipe/sp_quad_occlusion.c (limited to 'src') diff --git a/src/mesa/main/queryobj.c b/src/mesa/main/queryobj.c index 0e59ba615a..fc04dde3f4 100644 --- a/src/mesa/main/queryobj.c +++ b/src/mesa/main/queryobj.c @@ -30,6 +30,11 @@ #include "queryobj.h" #include "mtypes.h" +#if 1 /*PIPE*/ +#include "pipe/p_context.h" +#include "state_tracker/st_context.h" +#endif + /** * Allocate a new query object. This is a fallback routine called via @@ -220,6 +225,10 @@ _mesa_BeginQueryARB(GLenum target, GLuint id) q->Result = 0; q->Ready = GL_FALSE; +#if 1 /*PIPE*/ + ctx->st->pipe->reset_occlusion_counter(ctx->st->pipe); +#endif + if (target == GL_SAMPLES_PASSED_ARB) { ctx->Query.CurrentOcclusionObject = q; } @@ -282,6 +291,12 @@ _mesa_EndQueryARB(GLenum target) /* if we're using software rendering/querying */ q->Ready = GL_TRUE; } + +#if 1 /*PIPE*/ + if (target == GL_SAMPLES_PASSED_ARB) { + q->Result = ctx->st->pipe->get_occlusion_counter(ctx->st->pipe); + } +#endif } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index e4115226e7..05a175c8dc 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -28,7 +28,9 @@ #ifndef PIPE_CONTEXT_H #define PIPE_CONTEXT_H -#include "mtypes.h" +#include "main/mtypes.h" +#include "p_state.h" + /* Kludge: */ @@ -57,6 +59,12 @@ struct pipe_context { void (*clear)(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum); + /** occlusion counting (XXX this may be temporary - we should probably + * have generic query objects with begin/end methods) + */ + void (*reset_occlusion_counter)(struct pipe_context *pipe); + GLuint (*get_occlusion_counter)(struct pipe_context *pipe); + /* * State functions */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 2b0d659993..fd5e7ad3af 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -136,6 +136,7 @@ struct pipe_depth_state GLuint enabled:1; /**< depth test enabled? */ GLuint writemask:1; /**< allow depth buffer writes? */ GLuint func:3; /**< depth test func (PIPE_FUNC_x) */ + GLuint occlusion_count:1; /**< XXX move this elsewhere? */ GLfloat clear; /**< Clear value in [0,1] (XXX correct place?) */ }; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index cc4b8d5914..685e55be84 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -60,6 +60,20 @@ static void softpipe_draw_vb( struct pipe_context *pipe, } +static void softpipe_reset_occlusion_counter(struct pipe_context *pipe) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + softpipe->occlusion_counter = 0; +} + +/* XXX pipe param should be const */ +static GLuint softpipe_get_occlusion_counter(struct pipe_context *pipe) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + return softpipe->occlusion_counter; +} + + struct pipe_context *softpipe_create( void ) { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); @@ -82,12 +96,15 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.clear = softpipe_clear; + softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; + softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); softpipe->quad.shade = sp_quad_shade_stage(softpipe); softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); + softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); softpipe->quad.blend = sp_quad_blend_stage(softpipe); softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 81d2b58cf9..38f2977bd8 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -115,6 +115,8 @@ struct softpipe_context { */ GLubyte stipple_masks[16][16]; + GLuint occlusion_counter; + /** Software quad rendering pipeline */ struct { struct quad_stage *polygon_stipple; @@ -122,6 +124,7 @@ struct softpipe_context { struct quad_stage *alpha_test; struct quad_stage *stencil_test; struct quad_stage *depth_test; + struct quad_stage *occlusion; struct quad_stage *bufloop; struct quad_stage *blend; struct quad_stage *colormask; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 5f34f5f1b3..c27f14f32e 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -31,6 +31,11 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.bufloop; } + if (sp->depth_test.occlusion_count) { + sp->quad.occlusion->next = sp->quad.first; + sp->quad.first = sp->quad.occlusion; + } + if ( sp->stencil.front_enabled || sp->stencil.front_enabled) { sp->quad.stencil_test->next = sp->quad.first; diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 719d4774cc..072b51f9bf 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -51,6 +51,7 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c new file mode 100644 index 0000000000..843c462d48 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_occlusion.c @@ -0,0 +1,67 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * \brief Quad occlusion counter stage + * \author Brian Paul + */ + + +#include "main/glheader.h" +#include "main/imports.h" +#include "pipe/p_defines.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" + + +static void +occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + + softpipe->occlusion_counter += (quad->mask ) & 1; + softpipe->occlusion_counter += (quad->mask >> 1) & 1; + softpipe->occlusion_counter += (quad->mask >> 2) & 1; + softpipe->occlusion_counter += (quad->mask >> 3) & 1; + + if (quad->mask) + qs->next->run(qs->next, quad); +} + + +struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = occlusion_count_quad; + + return stage; +} diff --git a/src/mesa/sources b/src/mesa/sources index 4856f7b023..9bb05e35f7 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -164,6 +164,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_quad_colormask.c \ pipe/softpipe/sp_quad_depth_test.c \ pipe/softpipe/sp_quad_fs.c \ + pipe/softpipe/sp_quad_occlusion.c \ pipe/softpipe/sp_quad_output.c \ pipe/softpipe/sp_quad_stipple.c \ pipe/softpipe/sp_quad_stencil.c \ diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index a1523e06c2..7fc51953dc 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -71,6 +71,10 @@ update_depth( struct st_context *st ) depth.func = gl_depth_func_to_sp(st->ctx->Depth.Func); depth.clear = st->ctx->Depth.Clear; + if (st->ctx->Query.CurrentOcclusionObject && + st->ctx->Query.CurrentOcclusionObject->Active) + depth.occlusion_count = 1; + if (memcmp(&depth, &st->state.depth, sizeof(depth)) != 0) { /* state has changed */ st->state.depth = depth; /* struct copy */ -- cgit v1.2.3 From 46bba80a54afbcabc0f064433cc2194473661c30 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 10:33:48 -0600 Subject: Implement AA points and AA coverage application in quad pipeline. --- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_headers.h | 2 + src/mesa/pipe/softpipe/sp_prim_setup.c | 97 +++++++++++++++++++++---------- src/mesa/pipe/softpipe/sp_quad.c | 7 +++ src/mesa/pipe/softpipe/sp_quad.h | 1 + src/mesa/pipe/softpipe/sp_quad_coverage.c | 79 +++++++++++++++++++++++++ src/mesa/sources | 1 + 8 files changed, 159 insertions(+), 30 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_quad_coverage.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 685e55be84..3bb04a6835 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -105,6 +105,7 @@ struct pipe_context *softpipe_create( void ) softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); + softpipe->quad.coverage = sp_quad_coverage_stage(softpipe); softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); softpipe->quad.blend = sp_quad_blend_stage(softpipe); softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 38f2977bd8..ef1a06ccaf 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -125,6 +125,7 @@ struct softpipe_context { struct quad_stage *stencil_test; struct quad_stage *depth_test; struct quad_stage *occlusion; + struct quad_stage *coverage; struct quad_stage *bufloop; struct quad_stage *blend; struct quad_stage *colormask; diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index cd2dc984c5..bcc0f99add 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -81,6 +81,8 @@ struct quad_header { GLfloat depth[QUAD_SIZE]; } outputs; + GLfloat coverage[QUAD_SIZE]; /** fragment coverage for antialiasing */ + const struct setup_coefficient *coef; const enum interp_mode *interp; /* XXX: this information should be diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index e94ca139aa..3d3f2b74fc 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -858,60 +858,97 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const GLint ixmax = block((GLint) (x + halfSize)); const GLint iymin = block((GLint) (y - halfSize)); const GLint iymax = block((GLint) (y + halfSize)); - GLfloat halfSizeSquared = halfSize * halfSize; GLint ix, iy; - for (iy = iymin; iy <= iymax; iy += 2) { - for (ix = ixmin; ix <= ixmax; ix += 2) { + if (round) { + /* rounded points */ + const GLfloat rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ + const GLfloat rmax = halfSize + 0.7071F; + const GLfloat rmin2 = MAX2(0.0F, rmin * rmin); + const GLfloat rmax2 = rmax * rmax; + const GLfloat cscale = 1.0F / (rmax2 - rmin2); - if (round) { - /* rounded points */ - /* XXX for GL_SMOOTH, need to compute per-fragment coverage too */ - GLfloat dx, dy; + for (iy = iymin; iy <= iymax; iy += 2) { + for (ix = ixmin; ix <= ixmax; ix += 2) { + GLfloat dx, dy, dist2, cover; setup->quad.mask = 0x0; dx = (ix + 0.5) - x; dy = (iy + 0.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0); setup->quad.mask |= MASK_BOTTOM_LEFT; + } dx = (ix + 1.5) - x; dy = (iy + 0.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0); setup->quad.mask |= MASK_BOTTOM_RIGHT; + } dx = (ix + 0.5) - x; dy = (iy + 1.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0); setup->quad.mask |= MASK_TOP_LEFT; + } dx = (ix + 1.5) - x; dy = (iy + 1.5) - y; - if (dx * dx + dy * dy <= halfSizeSquared) + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0); setup->quad.mask |= MASK_TOP_RIGHT; - } - else { - /* square points */ - setup->quad.mask = 0xf; - - if (ix + 0.5 < x - halfSize) - setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - - if (ix + 1.5 > x + halfSize) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - - if (iy + 0.5 < y - halfSize) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + } - if (iy + 1.5 > y + halfSize) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + if (setup->quad.mask) { + setup->quad.x0 = ix; + setup->quad.y0 = iy; + quad_emit( setup->softpipe, &setup->quad ); + } } + } + } + else { + /* square points */ + for (iy = iymin; iy <= iymax; iy += 2) { + for (ix = ixmin; ix <= ixmax; ix += 2) { + setup->quad.mask = 0xf; - if (setup->quad.mask) { - setup->quad.x0 = ix; - setup->quad.y0 = iy; - quad_emit( setup->softpipe, &setup->quad ); + if (ix + 0.5 < x - halfSize) { + /* fragment is past left edge of point, turn off left bits */ + setup->quad.mask &= ~(MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + } + + if (ix + 1.5 > x + halfSize) { + /* past the right edge */ + setup->quad.mask &= ~(MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + } + + if (iy + 0.5 < y - halfSize) { + /* below the bottom edge */ + setup->quad.mask &= ~(MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + } + + if (iy + 1.5 > y + halfSize) { + /* above the top edge */ + setup->quad.mask &= ~(MASK_TOP_LEFT | MASK_TOP_RIGHT); + } + + if (setup->quad.mask) { + setup->quad.x0 = ix; + setup->quad.y0 = iy; + quad_emit( setup->softpipe, &setup->quad ); + } } } } diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index c27f14f32e..0053b16e68 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -36,6 +36,13 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.occlusion; } + if (sp->setup.poly_smooth || + sp->setup.line_smooth || + sp->setup.point_smooth) { + sp->quad.coverage->next = sp->quad.first; + sp->quad.first = sp->quad.coverage; + } + if ( sp->stencil.front_enabled || sp->stencil.front_enabled) { sp->quad.stencil_test->next = sp->quad.first; diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 072b51f9bf..2ee53bdecd 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -52,6 +52,7 @@ struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ) struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c new file mode 100644 index 0000000000..05cd03bce0 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -0,0 +1,79 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * \brief Apply AA coverage to quad alpha valus + * \author Brian Paul + */ + + +#include "main/glheader.h" +#include "main/macros.h" +#include "pipe/p_defines.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" + + +/** + * Multiply quad's alpha values by the fragment coverage. + */ +static void +apply_aa_coverage_quad(struct quad_stage *qs, struct quad_header *quad) +{ +#if 0 + struct softpipe_context *softpipe = qs->softpipe; + + /* XXX need to know if this quad is from a point, line or polygon! */ + if ((softpipe->setup.poly_smooth && prim == POLYGON) || + (softpipe->setup.line_smooth && prim == LINE) || + (softpipe->setup.point_smooth && prim == POINT)) { +#endif + GLuint j; + for (j = 0; j < QUAD_SIZE; j++) { + assert(quad->coverage[j] >= 0.0); + assert(quad->coverage[j] <= 1.0); + quad->outputs.color[3][j] *= quad->coverage[j]; + } +#if 0 + } +#endif + + qs->next->run(qs->next, quad); +} + + +struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->run = apply_aa_coverage_quad; + + return stage; +} diff --git a/src/mesa/sources b/src/mesa/sources index 9bb05e35f7..a76e41bdda 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -162,6 +162,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_quad_blend.c \ pipe/softpipe/sp_quad_bufloop.c \ pipe/softpipe/sp_quad_colormask.c \ + pipe/softpipe/sp_quad_coverage.c \ pipe/softpipe/sp_quad_depth_test.c \ pipe/softpipe/sp_quad_fs.c \ pipe/softpipe/sp_quad_occlusion.c \ -- cgit v1.2.3 From 80362a90d8ad1fca14d7276169fc962f953d936d Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 11:15:10 -0600 Subject: Add 'prim' field to quad so that stipple and aa coverage stages can do the right thing. --- src/mesa/pipe/softpipe/sp_headers.h | 11 ++++--- src/mesa/pipe/softpipe/sp_prim_setup.c | 5 ++++ src/mesa/pipe/softpipe/sp_quad_coverage.c | 15 ++++------ src/mesa/pipe/softpipe/sp_quad_stipple.c | 49 ++++++++++++++++--------------- 4 files changed, 40 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index bcc0f99add..68a84621f2 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -64,17 +64,16 @@ struct setup_coefficient { -/* Encodes everything we need to know about a 2x2 pixel block. Uses +/** + * Encodes everything we need to know about a 2x2 pixel block. Uses * "Channel-Serial" or "SoA" layout. - * - * Will expand to include non-attribute things like AA coverage and - * maybe prefetched depth from the depth buffer. */ struct quad_header { GLint x0; GLint y0; - GLuint mask; - GLuint facing; /**< Front (0) or back (1) facing? */ + GLuint mask:4; + GLuint facing:1; /**< Front (0) or back (1) facing? */ + GLuint prim:2; /**< PRIM_POINT, LINE, TRI */ struct { GLfloat color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 3d3f2b74fc..6420fc8809 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -569,6 +569,8 @@ static void setup_tri( struct draw_stage *stage, setup_tri_coefficients( setup ); setup_tri_edges( setup ); + setup->quad.prim = PRIM_TRI; + setup->span.y = 0; setup->span.y_flags = 0; setup->span.right[0] = 0; @@ -751,6 +753,7 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) setup->quad.x0 = setup->quad.y0 = -1; setup->quad.mask = 0x0; + setup->quad.prim = PRIM_LINE; if (dx > dy) { /*** X-major line ***/ @@ -842,6 +845,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const_coeff(setup, slot, j); } + setup->quad.prim = PRIM_POINT; + /* XXX need to clip against scissor bounds too */ if (halfSize <= 0.5 && !round) { diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c index 05cd03bce0..cdd8890c7f 100644 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -44,25 +44,20 @@ * Multiply quad's alpha values by the fragment coverage. */ static void -apply_aa_coverage_quad(struct quad_stage *qs, struct quad_header *quad) +coverage_quad(struct quad_stage *qs, struct quad_header *quad) { -#if 0 struct softpipe_context *softpipe = qs->softpipe; - /* XXX need to know if this quad is from a point, line or polygon! */ - if ((softpipe->setup.poly_smooth && prim == POLYGON) || - (softpipe->setup.line_smooth && prim == LINE) || - (softpipe->setup.point_smooth && prim == POINT)) { -#endif + if ((softpipe->setup.poly_smooth && quad->prim == PRIM_TRI) || + (softpipe->setup.line_smooth && quad->prim == PRIM_LINE) || + (softpipe->setup.point_smooth && quad->prim == PRIM_POINT)) { GLuint j; for (j = 0; j < QUAD_SIZE; j++) { assert(quad->coverage[j] >= 0.0); assert(quad->coverage[j] <= 1.0); quad->outputs.color[3][j] *= quad->coverage[j]; } -#if 0 } -#endif qs->next->run(qs->next, quad); } @@ -73,7 +68,7 @@ struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ) struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; - stage->run = apply_aa_coverage_quad; + stage->run = coverage_quad; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index 928771e454..532ffc61b0 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -13,44 +13,45 @@ /** * Apply polygon stipple to quads produced by triangle rasterization - * XXX we need to skip this for lines and points!!! */ static void stipple_quad(struct quad_stage *qs, struct quad_header *quad) { - struct softpipe_context *softpipe = qs->softpipe; - const GLint col0 = quad->x0 % 32; - const GLint row0 = quad->y0 % 32; - const GLuint stipple0 = softpipe->poly_stipple.stipple[row0]; - const GLuint stipple1 = softpipe->poly_stipple.stipple[row0 + 1]; + if (quad->prim == PRIM_TRI) { + struct softpipe_context *softpipe = qs->softpipe; + const GLint col0 = quad->x0 % 32; + const GLint row0 = quad->y0 % 32; + const GLuint stipple0 = softpipe->poly_stipple.stipple[row0]; + const GLuint stipple1 = softpipe->poly_stipple.stipple[row0 + 1]; - /* XXX this should be acheivable without conditionals */ + /* XXX this should be acheivable without conditionals */ #if 1 - GLbitfield mask = 0x0; + GLbitfield mask = 0x0; - if ((1 << col0) & stipple0) - mask |= MASK_BOTTOM_LEFT; + if ((1 << col0) & stipple0) + mask |= MASK_BOTTOM_LEFT; - if ((2 << col0) & stipple0) /* note: col0 <= 30 */ - mask |= MASK_BOTTOM_RIGHT; + if ((2 << col0) & stipple0) /* note: col0 <= 30 */ + mask |= MASK_BOTTOM_RIGHT; - if ((1 << col0) & stipple1) - mask |= MASK_TOP_LEFT; + if ((1 << col0) & stipple1) + mask |= MASK_TOP_LEFT; - if ((2 << col0) & stipple1) - mask |= MASK_TOP_RIGHT; + if ((2 << col0) & stipple1) + mask |= MASK_TOP_RIGHT; - quad->mask &= mask; + quad->mask &= mask; #else - /* XXX there may be a better way to lay out the stored stipple - * values to further simplify this computation. - */ - quad->mask &= (((stipple0 >> col0) & 0x3) | - (((stipple1 >> col0) & 0x3) << 2)); + /* XXX there may be a better way to lay out the stored stipple + * values to further simplify this computation. + */ + quad->mask &= (((stipple0 >> col0) & 0x3) | + (((stipple1 >> col0) & 0x3) << 2)); #endif - if (quad->mask) - qs->next->run(qs->next, quad); + if (quad->mask) + qs->next->run(qs->next, quad); + } } -- cgit v1.2.3 From a6eccf9e4f4d4815e004db9c8c61c5cdac63bbfd Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 11:16:13 -0600 Subject: Enable Keith's branch-less stipple code - it works. --- src/mesa/pipe/softpipe/sp_quad_stipple.c | 19 ------------------- 1 file changed, 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index 532ffc61b0..cad1a1400c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -24,30 +24,11 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) const GLuint stipple0 = softpipe->poly_stipple.stipple[row0]; const GLuint stipple1 = softpipe->poly_stipple.stipple[row0 + 1]; - /* XXX this should be acheivable without conditionals */ -#if 1 - GLbitfield mask = 0x0; - - if ((1 << col0) & stipple0) - mask |= MASK_BOTTOM_LEFT; - - if ((2 << col0) & stipple0) /* note: col0 <= 30 */ - mask |= MASK_BOTTOM_RIGHT; - - if ((1 << col0) & stipple1) - mask |= MASK_TOP_LEFT; - - if ((2 << col0) & stipple1) - mask |= MASK_TOP_RIGHT; - - quad->mask &= mask; -#else /* XXX there may be a better way to lay out the stored stipple * values to further simplify this computation. */ quad->mask &= (((stipple0 >> col0) & 0x3) | (((stipple1 >> col0) & 0x3) << 2)); -#endif if (quad->mask) qs->next->run(qs->next, quad); -- cgit v1.2.3 From a47b5764c0d801fb9fc23ee80c698ff042006cb6 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 12:23:34 -0600 Subject: comments --- src/mesa/pipe/softpipe/sp_quad_fs.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index b4470ac94a..7b1c90cba6 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -39,7 +39,7 @@ #include "sp_quad.h" struct exec_machine { - const struct setup_coefficient *coef; + const struct setup_coefficient *coef; /**< will point to quad->coef */ GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE]; }; @@ -200,7 +200,9 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) } #endif - qs->next->run(qs->next, quad); + /* shader may cull fragments */ + if (quad->mask) + qs->next->run(qs->next, quad); } -- cgit v1.2.3 From 5810b40d6e9aae6b184879a99c67f83107fc6637 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 12:26:09 -0600 Subject: set INTERP_CONSTANT when appropriate Not really signficant now since draw_flat stage copies the provoking vertex color to all prim verts. --- src/mesa/pipe/softpipe/sp_state_derived.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 34c893e396..18dfb50e38 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -90,7 +90,11 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { if (inputsRead & (i << i)) { - EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); + if (softpipe->setup.flatshade + && (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1)) + EMIT_ATTR(frag_to_vf[i], i, INTERP_CONSTANT); + else + EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); } } -- cgit v1.2.3 From 2bf4a500de24347476ce96cdd48d68ddeecbb019 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 12:28:42 -0600 Subject: Fix for-loop in interp() so we don't go out of bounds. Improved comments for that loop. Added some sanity check assertions regarding vertex layout. --- src/mesa/pipe/draw/draw_clip.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 7ede51fc6c..b220cc3643 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -110,13 +110,15 @@ static void interp( const struct clipper *clip, dst->data[0][3] = oow; } - /* Other attributes - * Note: start at 1 to skip winpos (data[0]) and subtract one - * since there's two vertex attrib slots we want to ignore (the header - * and the clippos. + * Note: start at 1 to skip winpos (data[0]) since we just computed + * it above. + * Subtract two from nr_attrs since the first two attribs (always + * VF_ATTRIB_VERTEX_HEADER and VF_ATTRIB_CLIP_POS, see + * draw_set_vertex_attributes()) are in the vertex_header struct, + * not in the data[] array. */ - for (j = 1; j < nr_attrs-1; j++) { + for (j = 1; j < nr_attrs - 2; j++) { interp_attr(dst->data[j], t, in->data[j], out->data[j]); } } @@ -365,6 +367,11 @@ static void clip_begin( struct draw_stage *stage ) struct clipper *clipper = clipper_stage(stage); GLuint nr = stage->draw->nr_planes; + /* sanity checks. If these fail, review the clip/interp code! */ + assert(stage->draw->nr_attrs >= 3); + assert(stage->draw->attrs[0].attrib == VF_ATTRIB_VERTEX_HEADER); + assert(stage->draw->attrs[1].attrib == VF_ATTRIB_CLIP_POS); + /* Hacky bitmask to use when we hit CLIP_USER_BIT: */ clipper->active_user_planes = ((1< Date: Fri, 13 Jul 2007 20:47:17 +0200 Subject: bugfixes (wrt fb updates), cleanups move drawable validation back to where it was now that the driDrawable information isn't used directly anymore. Fix bogus fb updates (the context we get for SwapBuffer processing may not have the drawable being processed attached!) glthreads behaves a bit better but still not correct. --- src/mesa/drivers/dri/i915tex/intel_blit.c | 18 ++++++++--------- src/mesa/drivers/dri/i915tex/intel_context.c | 24 ++++++++++------------- src/mesa/drivers/dri/i915tex/intel_context.h | 1 - src/mesa/drivers/dri/i915tex/intel_fbo.c | 14 ++++++++------ src/mesa/drivers/dri/i915tex/intel_fbo.h | 4 +++- src/mesa/drivers/dri/i915tex/intel_screen.c | 29 ++++++++++++++-------------- 6 files changed, 44 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c index 2a0a639315..db72f59a2c 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ b/src/mesa/drivers/dri/i915tex/intel_blit.c @@ -79,13 +79,6 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, */ LOCK_HARDWARE(intel); - if (intel->revalidateDrawable) { - __DRIscreenPrivate *sPriv = intel->driScreen; - if (dPriv) { - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - } - } - if (dPriv && dPriv->numClipRects) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; const struct intel_region *frontRegion @@ -190,9 +183,14 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, UNLOCK_HARDWARE(intel); - if (intel->revalidateDrawable) { - intel->revalidateDrawable = GL_FALSE; - intelWindowMoved(intel); + /* XXX this is bogus. The context here may not even be bound to this drawable! */ + if (intel->lastStamp != dPriv->lastStamp) { + GET_CURRENT_CONTEXT(currctx); + struct intel_context *intelcurrent = intel_context(currctx); + if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { + intelWindowMoved(intel); + intel->lastStamp = dPriv->lastStamp; + } } } diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 451c7fd1ba..41eaafece2 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -572,7 +572,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv) { - GLuint updatebufsize = GL_FALSE; #if 0 if (driDrawPriv) { @@ -620,22 +619,18 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, } } - /* only update GLframebuffer size to match window - if here for the first time */ - if (intel->ctx.FirstTimeCurrent) { - updatebufsize = GL_TRUE; - driUpdateFramebufferSize(&intel->ctx, driDrawPriv); + /* update GLframebuffer size to match window if needed */ + driUpdateFramebufferSize(&intel->ctx, driDrawPriv); - if (driReadPriv != driDrawPriv) { - driUpdateFramebufferSize(&intel->ctx, driReadPriv); - } + if (driReadPriv != driDrawPriv) { + driUpdateFramebufferSize(&intel->ctx, driReadPriv); } _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); /* The drawbuffer won't always be updated by _mesa_make_current: */ - if (updatebufsize && intel->ctx.DrawBuffer == &intel_fb->Base) { + if (intel->ctx.DrawBuffer == &intel_fb->Base) { if (intel->driDrawable != driDrawPriv) { intel_fb->vblank_flags = (intel->intelScreen->irq_active != 0) @@ -646,9 +641,11 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, &intel_fb->vbl_seq); intel->driDrawable = driDrawPriv; intelWindowMoved(intel); + intel->lastStamp = driDrawPriv->lastStamp; + } + else if (intel->lastStamp != driDrawPriv->lastStamp) { + intel_draw_buffer(&intel->ctx, &intel_fb->Base); } - - intel_draw_buffer(&intel->ctx, &intel_fb->Base); } } else { @@ -677,8 +674,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) * checking must be done *after* this call: */ if (dPriv) - intel->revalidateDrawable = GL_TRUE; -// DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); if (sarea->width != intelScreen->width || sarea->height != intelScreen->height || diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index f85b3a6362..b5defdc47e 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -272,7 +272,6 @@ struct intel_context drmI830Sarea *sarea; GLuint lastStamp; - GLuint revalidateDrawable; /** * Configuration cache diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c index aeb909cce1..afee07c689 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c @@ -299,7 +299,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } - +#if 0 /** * Called for each hardware renderbuffer when a _window_ is resized. * Just update fields. @@ -316,6 +316,7 @@ intel_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb, return GL_TRUE; } +#endif static void intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, @@ -352,7 +353,7 @@ intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } - +#if 0 /** * Create a new intel_renderbuffer which corresponds to an on-screen window, * not a user-created renderbuffer. @@ -452,15 +453,16 @@ intel_create_renderbuffer(GLenum intFormat, GLsizei width, GLsizei height, return irb; } +#endif -struct gl_renderbuffer * -intel_new_renderbuffer_fb(GLcontext * ctx, GLuint intFormat) +struct intel_renderbuffer * +intel_new_renderbuffer_fb(GLuint intFormat) { struct intel_renderbuffer *irb; irb = CALLOC_STRUCT(intel_renderbuffer); if (!irb) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); + _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer"); return NULL; } @@ -474,7 +476,7 @@ intel_new_renderbuffer_fb(GLcontext * ctx, GLuint intFormat) irb->Base.GetPointer = intel_get_pointer; /* span routines set in alloc_storage function */ - return &irb->Base; + return irb; } /** diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.h b/src/mesa/drivers/dri/i915tex/intel_fbo.h index 963f5e706f..762aac594e 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.h +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.h @@ -84,7 +84,7 @@ struct intel_renderbuffer GLuint vbl_pending; /**< vblank sequence number of pending flip */ }; - +#if 0 extern struct intel_renderbuffer *intel_create_renderbuffer(GLenum intFormat, GLsizei width, GLsizei height, @@ -92,7 +92,9 @@ extern struct intel_renderbuffer *intel_create_renderbuffer(GLenum intFormat, int pitch, int cpp, void *map); +#endif +extern struct intel_renderbuffer *intel_new_renderbuffer_fb(GLuint intFormat); extern void intel_fbo_init(struct intel_context *intel); diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c index 08057a6600..dcc2992383 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ b/src/mesa/drivers/dri/i915tex/intel_screen.c @@ -680,14 +680,14 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, /* XXX allocation should only happen in the unusual case it's actually needed */ intel_fb->color_rb[0] - = intel_new_renderbuffer_fb(NULL, rgbFormat); + = intel_new_renderbuffer_fb(rgbFormat); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, &intel_fb->color_rb[0]->Base); } if (mesaVis->doubleBufferMode) { intel_fb->color_rb[1] - = intel_new_renderbuffer_fb(NULL, rgbFormat); + = intel_new_renderbuffer_fb(rgbFormat); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, &intel_fb->color_rb[1]->Base); @@ -695,14 +695,14 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, struct gl_renderbuffer *tmp_rb = NULL; intel_fb->color_rb[2] - = intel_new_renderbuffer_fb(NULL, rgbFormat); + = intel_new_renderbuffer_fb(rgbFormat); _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base); } } if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { /* combined depth/stencil buffer */ struct intel_renderbuffer *depthStencilRb - = intel_new_renderbuffer_fb(NULL, GL_DEPTH24_STENCIL8_EXT); + = intel_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); /* note: bind RB to two attachment points */ _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthStencilRb->Base); @@ -712,7 +712,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, else if (mesaVis->depthBits == 16) { /* just 16-bit depth buffer, no hw stencil */ struct intel_renderbuffer *depthRb - = intel_new_renderbuffer_fb(NULL, GL_DEPTH_COMPONENT16); + = intel_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); } @@ -987,18 +987,19 @@ struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) * context at screen creation. For now just use the current context. */ -/* GET_CURRENT_CONTEXT(ctx); + GET_CURRENT_CONTEXT(ctx); if (ctx == NULL) { - _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; +/* _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; */ + /* need a context for the first time makecurrent is called (for hw lock + when allocating priv buffers) */ + if (intelScreen->dummyctxptr == NULL) { + _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; + } + return intelScreen->dummyctxptr; } return intel_context(ctx); -*/ - if (intelScreen->dummyctxptr == NULL) { - _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; - } - return intelScreen->dummyctxptr; } -- cgit v1.2.3 From dc313b578386dc07f4916fba98da061af3ab18e5 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 14:22:46 -0600 Subject: Fix more polygon winding, culling confusion. If the determinant of the triangle is positive, its winding is CCW (right-handed coord system). --- src/mesa/pipe/draw/draw_cull.c | 12 +++++++----- src/mesa/pipe/draw/draw_twoside.c | 12 ++++++++---- src/mesa/state_tracker/st_atom_setup.c | 18 ++++++++++++------ 3 files changed, 27 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index 863686f150..e563f9f45f 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -40,7 +40,7 @@ struct cull_stage { struct draw_stage stage; - GLuint mode; /**< one of PIPE_WINDING_x */ + GLuint winding; /**< which winding(s) to cull (one of PIPE_WINDING_x) */ }; @@ -54,7 +54,7 @@ static void cull_begin( struct draw_stage *stage ) { struct cull_stage *cull = cull_stage(stage); - cull->mode = stage->draw->setup.cull_mode; + cull->winding = stage->draw->setup.cull_mode; stage->next->begin( stage->next ); } @@ -78,10 +78,12 @@ static void cull_tri( struct draw_stage *stage, header->det = ex * fy - ey * fx; if (header->det != 0) { - /* non-zero area */ - GLuint mode = (header->det > 0) ? PIPE_WINDING_CW : PIPE_WINDING_CCW; + /* if (det > 0 then Z points toward camera and triangle is + * counter-clockwise winding. + */ + GLuint winding = (header->det > 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; - if ((mode & cull_stage(stage)->mode) == 0) { + if ((winding & cull_stage(stage)->winding) == 0) { /* triangle is not culled, pass to next stage */ stage->next->tri( stage->next, header ); } diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index e86123e84d..9f26335fb6 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -35,8 +35,7 @@ struct twoside_stage { struct draw_stage stage; - - GLfloat facing; + GLfloat sign; /**< +1 or -1 */ const GLuint *lookup; }; @@ -51,7 +50,12 @@ static void twoside_begin( struct draw_stage *stage ) { struct twoside_stage *twoside = twoside_stage(stage); - twoside->facing = (stage->draw->setup.front_winding == PIPE_WINDING_CW) ? 1 : -1; + /* + * We'll multiply the primitive's determinant by this sign to determine + * if the triangle is back-facing (negative). + * sign = 1 for CCW, -1 for CW + */ + twoside->sign = (stage->draw->setup.front_winding == PIPE_WINDING_CCW) ? 1 : -1; stage->next->begin( stage->next ); } @@ -94,7 +98,7 @@ static void twoside_tri( struct draw_stage *stage, { struct twoside_stage *twoside = twoside_stage(stage); - if (header->det * twoside->facing < 0) { + if (header->det * twoside->sign < 0.0) { /* this is a back-facing triangle */ struct prim_header tmp; diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index 8b95ea958e..08f7a8a732 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -78,13 +78,19 @@ static void update_setup_state( struct st_context *st ) /* _NEW_POLYGON, _NEW_BUFFERS */ { - setup.front_winding = PIPE_WINDING_CW; - + if (ctx->Polygon.FrontFace == GL_CCW) + setup.front_winding = PIPE_WINDING_CCW; + else + setup.front_winding = PIPE_WINDING_CW; + + /* XXX + * I think the intention here is that user-created framebuffer objects + * use Y=0=TOP layout instead of OpenGL's normal Y=0=bottom layout. + * Flipping Y changes CW to CCW and vice-versa. + * But this is an implementation/driver-specific artifact - remove... + */ if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) - setup.front_winding ^= PIPE_WINDING_BOTH; - - if (ctx->Polygon.FrontFace != GL_CCW) - setup.front_winding ^= PIPE_WINDING_BOTH; + setup.front_winding ^= PIPE_WINDING_BOTH; } /* _NEW_LIGHT -- cgit v1.2.3 From f0636d9653d142bdf71475d2d2066f5c7c436978 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 13 Jul 2007 14:28:15 -0600 Subject: Fix setup.light_twoside determination. When a vertex program is used, use value of GL_VERTEX_PROGRAM_TWO_SIDE. Otherwise, if FF lighting, use value of GL_LIGHT_MODEL_TWO_SIDE. --- src/mesa/state_tracker/st_atom_setup.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index 08f7a8a732..6ee96f6178 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -98,14 +98,18 @@ static void update_setup_state( struct st_context *st ) if (ctx->Light.ShadeModel == GL_FLAT) setup.flatshade = 1; - /* _NEW_LIGHT + /* _NEW_LIGHT | _NEW_PROGRAM * - * Not sure about the light->enabled requirement - does this still - * apply?? + * Back-face colors can come from traditional lighting (when + * GL_LIGHT_MODEL_TWO_SIDE is set) or from vertex programs (when + * GL_VERTEX_PROGRAM_TWO_SIDE is set). Note the logic here. */ - if (ctx->Light.Enabled && - ctx->Light.Model.TwoSide) + if (ctx->VertexProgram._Enabled) { + setup.light_twoside = ctx->VertexProgram.TwoSideEnabled; + } + else if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { setup.light_twoside = 1; + } /* _NEW_POLYGON */ -- cgit v1.2.3 From a04056d313a8346dc9e6e1b062762a6f115d1b46 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Sat, 14 Jul 2007 14:24:20 +0200 Subject: fix assertion, z/stencil swarst readback, (still looks suspicious) copyTexSubimage --- src/mesa/drivers/dri/common/drirenderbuffer.c | 4 ++++ src/mesa/drivers/dri/i915tex/intel_pixel_read.c | 2 +- src/mesa/drivers/dri/i915tex/intel_span.c | 4 ++-- src/mesa/drivers/dri/i915tex/intel_tex_copy.c | 8 +++++--- 4 files changed, 12 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/common/drirenderbuffer.c b/src/mesa/drivers/dri/common/drirenderbuffer.c index d36af3e5be..d4f7f2527b 100644 --- a/src/mesa/drivers/dri/common/drirenderbuffer.c +++ b/src/mesa/drivers/dri/common/drirenderbuffer.c @@ -209,7 +209,11 @@ driUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv) struct gl_framebuffer *fb = (struct gl_framebuffer *) dPriv->driverPrivate; if (fb && (dPriv->w != fb->Width || dPriv->h != fb->Height)) { ctx->Driver.ResizeBuffers(ctx, fb, dPriv->w, dPriv->h); +#if 0 + /* if the driver needs the hw lock for ResizeBuffers, the drawable + might have changed again by now */ assert(fb->Width == dPriv->w); assert(fb->Height == dPriv->h); +#endif } } diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c b/src/mesa/drivers/dri/i915tex/intel_pixel_read.c index 9481cb4800..c7bdb6bc85 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_read.c @@ -183,7 +183,7 @@ do_blit_readpixels(GLcontext * ctx, */ if (!_mesa_validate_pbo_access(2, pack, width, height, 1, format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels"); + _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels"); return GL_TRUE; } } diff --git a/src/mesa/drivers/dri/i915tex/intel_span.c b/src/mesa/drivers/dri/i915tex/intel_span.c index 368e0ec3c0..74f118121c 100644 --- a/src/mesa/drivers/dri/i915tex/intel_span.c +++ b/src/mesa/drivers/dri/i915tex/intel_span.c @@ -264,7 +264,7 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) /* depth buffer (Note wrapper!) */ if (ctx->DrawBuffer->_DepthBuffer) { irb = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); - if (irb && irb->region && irb->Base.Name != 0) { + if (irb && irb->region) { if (map) { intel_region_map(intel->intelScreen, irb->region); irb->pfMap = irb->region->map; @@ -281,7 +281,7 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) /* stencil buffer (Note wrapper!) */ if (ctx->DrawBuffer->_StencilBuffer) { irb = intel_renderbuffer(ctx->DrawBuffer->_StencilBuffer->Wrapped); - if (irb && irb->region && irb->Base.Name != 0) { + if (irb && irb->region) { if (map) { intel_region_map(intel->intelScreen, irb->region); irb->pfMap = irb->region->map; diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_copy.c b/src/mesa/drivers/dri/i915tex/intel_tex_copy.c index 90c2b9ebb5..8d7f8f9d91 100644 --- a/src/mesa/drivers/dri/i915tex/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915tex/intel_tex_copy.c @@ -118,9 +118,11 @@ do_copy_texsubimage(struct intel_context *intel, dstx += x - orig_x; dsty += y - orig_y; - /* invert Y */ - y = ctx->ReadBuffer->Height - y - 1; - + if (!(ctx->ReadBuffer->Name == 0)) { + /* XXX this looks bogus ? */ + /* FBO: invert Y */ + y = ctx->ReadBuffer->Height - y - 1; + } /* A bit of fiddling to get the blitter to work with -ve * pitches. But we get a nice inverted blit this way, so it's -- cgit v1.2.3 From cd3afa2a952a9ec05fb9fa959dfe90d620bd2d88 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Sun, 15 Jul 2007 18:17:36 +0200 Subject: begin getting rid of fake clip rects (copy_pixel) --- src/mesa/drivers/dri/i915tex/intel_context.c | 1 + src/mesa/drivers/dri/i915tex/intel_pixel_copy.c | 133 ++++++++++++++---------- 2 files changed, 80 insertions(+), 54 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 41eaafece2..146426fd37 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -192,6 +192,7 @@ const struct dri_extension card_extensions[] = { #if 1 /* XXX FBO temporary? */ {"GL_EXT_packed_depth_stencil", NULL}, #endif + {"GL_EXT_pixel_buffer_object", NULL}, {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, {"GL_EXT_stencil_wrap", NULL}, {"GL_EXT_texture_edge_clamp", NULL}, diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c index 03c7353ee1..77d0e24f00 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c @@ -42,21 +42,25 @@ #include "intel_regions.h" #include "intel_tris.h" #include "intel_pixel.h" +#include "intel_fbo.h" #define FILE_DEBUG_FLAG DEBUG_PIXEL static struct intel_region * copypix_src_region(struct intel_context *intel, GLenum type) { + struct intel_renderbuffer *irb; + switch (type) { case GL_COLOR: return intel_readbuf_region(intel); case GL_DEPTH: /* Don't think this is really possible execpt at 16bpp, when we have no stencil. */ - if (intel->intelScreen->depth_region && - intel->intelScreen->depth_region->cpp == 2) - return intel->intelScreen->depth_region; + irb = intel_renderbuffer(intel->ctx.ReadBuffer->_DepthBuffer->Wrapped); + if (irb && irb->region && irb->region->cpp == 2) + return irb->region; + break; case GL_STENCIL: /* Don't think this is really possible. */ @@ -64,7 +68,10 @@ copypix_src_region(struct intel_context *intel, GLenum type) case GL_DEPTH_STENCIL_EXT: /* Does it matter whether it is stencil/depth or depth/stencil? */ - return intel->intelScreen->depth_region; + irb = intel_renderbuffer(intel->ctx.ReadBuffer->_DepthBuffer->Wrapped); + if (irb && irb->region) + return irb->region; + break; default: break; } @@ -164,7 +171,8 @@ do_texture_copypixels(GLcontext * ctx, /* Set the 3d engine to draw into the destination region: */ - intel->vtbl.meta_draw_region(intel, dst, intel->intelScreen->depth_region); + intel->vtbl.meta_draw_region(intel, dst, + (intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped))->region); intel->vtbl.meta_import_pixel_state(intel); @@ -191,9 +199,8 @@ do_texture_copypixels(GLcontext * ctx, LOCK_HARDWARE(intel); - if (intel->numClipRects) { - int bufHeight = intel->pClipRects->y2; - assert (intel->numClipRects == 1); + { + int bufHeight = ctx->DrawBuffer->Height; srcy = bufHeight - srcy - height; /* convert from gl to hardware coords */ /* Clip against the source region. This is the only source @@ -205,7 +212,8 @@ do_texture_copypixels(GLcontext * ctx, GLint orig_x = srcx; GLint orig_y = srcy; - if (!_mesa_clip_to_region(0, 0, src->pitch, src->height, + if (!_mesa_clip_to_region(0, 0, ctx->ReadBuffer->Width - 1, + ctx->ReadBuffer->Height - 1, &srcx, &srcy, &width, &height)) goto out; @@ -248,8 +256,10 @@ do_blit_copypixels(GLcontext * ctx, GLint dstx, GLint dsty, GLenum type) { struct intel_context *intel = intel_context(ctx); - struct intel_region *dst = intel_drawbuf_region(intel); - struct intel_region *src = copypix_src_region(intel, type); + struct intel_renderbuffer *irbread; + struct intel_renderbuffer *irbdraw; + struct intel_region *dst; + struct intel_region *src; /* Copypixels can be more than a straight copy. Ensure all the * extra operations are disabled: @@ -258,23 +268,43 @@ do_blit_copypixels(GLcontext * ctx, ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F) return GL_FALSE; - if (!src || !dst) - return GL_FALSE; - - - intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); + if (type == GL_COLOR) { + irbread = intel_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + irbdraw = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); + if (!irbread || !irbread->region || !irbdraw || !irbdraw->region) + return GL_FALSE; + } + else if (type == GL_DEPTH) { + /* Don't think this is really possible execpt at 16bpp, when we have no stencil. + */ + irbread = intel_renderbuffer(ctx->ReadBuffer->_DepthBuffer->Wrapped); + irbdraw = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + if (!irbread || !irbread->region || !irbdraw || !irbdraw->region + || !(irbread->region->cpp == 2)) + return GL_FALSE; + } + else if (GL_DEPTH_STENCIL_EXT) { + /* Does it matter whether it is stencil/depth or depth/stencil? + */ + irbread = intel_renderbuffer(ctx->ReadBuffer->_DepthBuffer->Wrapped); + irbdraw = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + if (!irbread || !irbread->region || !irbdraw || !irbdraw->region) + return GL_FALSE; + } + else if (GL_STENCIL) { + /* Don't think this is really possible. + */ + return GL_FALSE; + } + + src = irbread->region; + dst = irbdraw->region; - if (intel->numClipRects) { - assert(intel->numClipRects == 1); - drm_clip_rect_t *box = intel->pClipRects; - drm_clip_rect_t dest_rect; - GLint nbox = intel->numClipRects; + { GLint delta_x = 0; GLint delta_y = 0; - GLuint i; /* Do scissoring in GL coordinates: */ @@ -289,58 +319,54 @@ do_blit_copypixels(GLcontext * ctx, if (!_mesa_clip_to_region(x, y, x+w-1, y+h-1, &dstx, &dsty, &width, &height)) goto out; - + srcx = dstx - dx; srcy = dsty - dy; } /* Convert from GL to hardware coordinates: - * cliprect should be size of buffer */ - dsty = box->y2 - dsty - height; - srcy = box->y2 - srcy - height; + dsty = irbdraw->Base.Height - dsty - height; + srcy = irbread->Base.Height - srcy - height; - /* Clip against the source region. This is the only source - * clipping we do. Dst is clipped with cliprects below. + /* Clip against the source region: */ { delta_x = srcx - dstx; delta_y = srcy - dsty; - if (!_mesa_clip_to_region(0, 0, src->pitch, src->height, - &srcx, &srcy, &width, &height)) + if (!_mesa_clip_to_region(0, 0, irbread->Base.Width - 1, + irbread->Base.Height - 1, + &srcx, &srcy, &width, &height)) goto out; dstx = srcx - delta_x; dsty = srcy - delta_y; } - dest_rect.x1 = dstx; - dest_rect.y1 = dsty; - dest_rect.x2 = dstx + width; - dest_rect.y2 = dsty + height; - - /* Could do slightly more clipping: Eg, take the intersection of - * the existing set of cliprects and those cliprects translated - * by delta_x, delta_y: - * - * This code will not overwrite other windows, but will - * introduce garbage when copying from obscured window regions. + /* Clip against the dest region: */ - for (i = 0; i < nbox; i++) { - drm_clip_rect_t rect; + { + delta_x = dstx - srcx; + delta_y = dsty - srcy; - if (!intel_intersect_cliprects(&rect, &dest_rect, &box[i])) - continue; + if (!_mesa_clip_to_region(0, 0, irbdraw->Base.Width - 1, + irbdraw->Base.Height - 1, + &dstx, &dsty, &width, &height)) + goto out; + srcx = dstx - delta_x; + srcy = dsty - delta_y; + } - intelEmitCopyBlit(intel, dst->cpp, - src->pitch, src->buffer, 0, - dst->pitch, dst->buffer, 0, - rect.x1 + delta_x, - rect.y1 + delta_y, /* srcx, srcy */ - rect.x1, rect.y1, /* dstx, dsty */ - rect.x2 - rect.x1, rect.y2 - rect.y1, + { + + intelEmitCopyBlit(intel, dst->cpp, + src->pitch, src->buffer, 0, + dst->pitch, dst->buffer, 0, + srcx, srcy, + dstx, dsty, + width, height, ctx->Color.ColorLogicOpEnabled ? ctx->Color.LogicOp : GL_COPY); } @@ -348,7 +374,6 @@ do_blit_copypixels(GLcontext * ctx, out: intel_batchbuffer_flush(intel->batch); } - UNLOCK_HARDWARE(intel); DBG("%s: success\n", __FUNCTION__); return GL_TRUE; -- cgit v1.2.3 From 8172f50419b370c4608d1bbc6cac4d77e3e45804 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Sun, 15 Jul 2007 20:15:19 +0200 Subject: fix the texture copypixels path --- src/mesa/drivers/dri/i915tex/intel_pixel_copy.c | 23 ++++++++++++++--------- 1 file changed, 14 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c index 77d0e24f00..57e746ae22 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c @@ -119,6 +119,7 @@ do_texture_copypixels(GLcontext * ctx, struct intel_context *intel = intel_context(ctx); struct intel_region *dst = intel_drawbuf_region(intel); struct intel_region *src = copypix_src_region(intel, type); + struct intel_region *depthreg = NULL; GLenum src_format; GLenum src_type; @@ -171,8 +172,11 @@ do_texture_copypixels(GLcontext * ctx, /* Set the 3d engine to draw into the destination region: */ - intel->vtbl.meta_draw_region(intel, dst, - (intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped))->region); + if (ctx->DrawBuffer->_DepthBuffer && + ctx->DrawBuffer->_DepthBuffer->Wrapped) + depthreg = (intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped))->region; + + intel->vtbl.meta_draw_region(intel, dst, depthreg); intel->vtbl.meta_import_pixel_state(intel); @@ -200,8 +204,9 @@ do_texture_copypixels(GLcontext * ctx, LOCK_HARDWARE(intel); { - int bufHeight = ctx->DrawBuffer->Height; - srcy = bufHeight - srcy - height; /* convert from gl to hardware coords */ + int dstbufHeight = ctx->DrawBuffer->Height; + /* convert from gl to hardware coords */ + srcy = ctx->ReadBuffer->Height - srcy - height; /* Clip against the source region. This is the only source * clipping we do. XXX: Just set the texcord wrap mode to clamp @@ -224,11 +229,11 @@ do_texture_copypixels(GLcontext * ctx, /* Just use the regular cliprect mechanism... Does this need to * even hold the lock??? */ - intel_meta_draw_quad(intel, - dstx, - dstx + width * ctx->Pixel.ZoomX, - bufHeight - (dsty + height * ctx->Pixel.ZoomY), - bufHeight - (dsty), 0, /* XXX: what z value? */ + intel_meta_draw_quad(intel, + dstx, + dstx + width * ctx->Pixel.ZoomX, + dstbufHeight - (dsty + height * ctx->Pixel.ZoomY), + dstbufHeight - (dsty), 0, /* XXX: what z value? */ 0x00ff00ff, srcx, srcx + width, srcy, srcy + height); -- cgit v1.2.3 From a1ec23a30f3ff9583b32428c2c357b9ef66f3a50 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Sun, 15 Jul 2007 22:47:42 +0200 Subject: fix bogus fb/drawable information (fixes xdemos/wincopy) the framebuffer objects attached to drawables can have invalidate state associated with them, since for the window framebuffer this is per-context state and not per-fbo state. Since drivers may rely on that information (otherwise would need to check if currently the window-framebuffer is bound in a lot of places) fix it up in _mesa_make_current (ugly). --- src/mesa/drivers/dri/i915tex/intel_context.c | 5 +++-- src/mesa/main/context.c | 11 +++++++++++ 2 files changed, 14 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 146426fd37..cb3ec4ccfa 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -289,8 +289,9 @@ static void intelCheckFrontUpdate(GLcontext * ctx) { struct intel_context *intel = intel_context(ctx); - if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT) { + /* can't use _ColorDrawBufferMask as its value + might change if a different drawable is bound! */ + if (ctx->Color.DrawBuffer[0] == GL_FRONT_LEFT) { intelScreenPrivate *screen = intel->intelScreen; __DRIdrawablePrivate *dPriv = intel->driDrawable; if (screen->current_rotation != 0) { diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 255023c0fa..183a552d40 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1495,9 +1495,20 @@ _mesa_make_current( GLcontext *newCtx, GLframebuffer *drawBuffer, */ if (!newCtx->DrawBuffer || newCtx->DrawBuffer->Name == 0) { _mesa_reference_framebuffer(&newCtx->DrawBuffer, drawBuffer); + /* fix up the fb fields - these will end up wrong otherwise + if the DRIdrawable changes, and someone may rely on them. + */ + /* What a mess!?! */ + int i; + GLenum buffers[MAX_DRAW_BUFFERS]; + for(i = 0; i < newCtx->Const.MaxDrawBuffers; i++) { + buffers[i] = newCtx->Color.DrawBuffer[i]; + } + _mesa_drawbuffers(newCtx, newCtx->Const.MaxDrawBuffers, buffers, NULL); } if (!newCtx->ReadBuffer || newCtx->ReadBuffer->Name == 0) { _mesa_reference_framebuffer(&newCtx->ReadBuffer, readBuffer); + _mesa_ReadBuffer(newCtx->Pixel.ReadBuffer); } newCtx->NewState |= _NEW_BUFFERS; -- cgit v1.2.3 From 99c7e38d8363eafe2fc079ed6e7bad289d65e965 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Sun, 15 Jul 2007 22:53:59 +0200 Subject: change back driver change from previous commit --- src/mesa/drivers/dri/i915tex/intel_context.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index cb3ec4ccfa..ec4a7a746d 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -289,9 +289,12 @@ static void intelCheckFrontUpdate(GLcontext * ctx) { struct intel_context *intel = intel_context(ctx); - /* can't use _ColorDrawBufferMask as its value - might change if a different drawable is bound! */ - if (ctx->Color.DrawBuffer[0] == GL_FRONT_LEFT) { + /* rely on _ColorDrawBufferMask being kept up to date by mesa + even for window-fbos. */ + /* not sure. Might need that for all masks including + BUFFER_BIT_FRONT_LEFT maybe? */ + if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT) { intelScreenPrivate *screen = intel->intelScreen; __DRIdrawablePrivate *dPriv = intel->driDrawable; if (screen->current_rotation != 0) { -- cgit v1.2.3 From d69cae95a072d3641bd8340a5216662e0365ed6f Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Mon, 16 Jul 2007 11:51:08 +0200 Subject: allow blit path for read pixels if pack->alignment isn't 1. pack->alignment doesn't matter usually if the image is sufficiently aligned anyway. Speeds up some tests/readrate cases by a factor of 100 (since the pack->alignment default which noone ever bothers to change is 4). --- src/mesa/drivers/dri/i915tex/intel_pixel_read.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c b/src/mesa/drivers/dri/i915tex/intel_pixel_read.c index c7bdb6bc85..c41ca33f8f 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_read.c @@ -203,17 +203,18 @@ do_blit_readpixels(GLcontext * ctx, return GL_FALSE; } - if (pack->Alignment != 1 || pack->SwapBytes || pack->LsbFirst) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s: bad packing params\n", __FUNCTION__); - return GL_FALSE; - } - if (pack->RowLength > 0) rowLength = pack->RowLength; else rowLength = width; + if (((rowLength * src->cpp) % pack->Alignment) || + pack->SwapBytes || pack->LsbFirst) { + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s: bad packing params\n", __FUNCTION__); + return GL_FALSE; + } + if (pack->Invert) { if (INTEL_DEBUG & DEBUG_PIXEL) _mesa_printf("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__); -- cgit v1.2.3 From 8b93d9238e4dd51acaf95c4744d5f6a795d9d35c Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Mon, 16 Jul 2007 13:57:49 +0200 Subject: get rid of more cliprects (readpixel)... --- src/mesa/drivers/dri/i915tex/intel_pixel_read.c | 167 +++++++++++++----------- 1 file changed, 91 insertions(+), 76 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c b/src/mesa/drivers/dri/i915tex/intel_pixel_read.c index c41ca33f8f..76a6959625 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_read.c @@ -42,6 +42,8 @@ #include "intel_regions.h" #include "intel_pixel.h" #include "intel_buffer_objects.h" +#include "intel_fbo.h" +#include "intel_tris.h" /* For many applications, the new ability to pull the source buffers * back out of the GTT and then do the packing/conversion operations @@ -54,6 +56,8 @@ * an argument for blit/texture readpixels, or for blitting to a * temporary and then pulling that back. * + * XXX this is not true for fake frontbuffers... + * When the destination is a pbo, however, it's not clear if it is * ever going to be pulled to main memory (though the access param * will be a good hint). So it sounds like we do want to be able to @@ -68,88 +72,105 @@ static GLboolean do_texture_readpixels(GLcontext * ctx, - GLint x, GLint y, GLsizei width, GLsizei height, + GLint srcx, GLint srcy, GLsizei width, GLsizei height, GLenum format, GLenum type, const struct gl_pixelstore_attrib *pack, struct intel_region *dest_region) { #if 0 struct intel_context *intel = intel_context(ctx); - intelScreenPrivate *screen = intel->intelScreen; - GLint pitch = pack->RowLength ? pack->RowLength : width; - int textureFormat; - GLenum glTextureFormat; - int destFormat, depthFormat, destPitch; - drm_clip_rect_t tmp; + struct intel_region *depthreg = NULL; + struct intel_region *src = intel_readbuf_region(intel); +// struct intel_region *src = copypix_src_region(intel, type); + GLenum src_format; + GLenum src_type; if (INTEL_DEBUG & DEBUG_PIXEL) fprintf(stderr, "%s\n", __FUNCTION__); - - if (ctx->_ImageTransferState || - pack->SwapBytes || pack->LsbFirst || !pack->Invert) { - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: check_color failed\n", __FUNCTION__); - return GL_FALSE; - } - - intel->vtbl.meta_texrect_source(intel, intel_readbuf_region(intel)); - - if (!intel->vtbl.meta_render_dest(intel, dest_region, type, format)) { - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: couldn't set dest %s/%s\n", - __FUNCTION__, - _mesa_lookup_enum_by_nr(type), - _mesa_lookup_enum_by_nr(format)); + if (!src || type != GL_COLOR) return GL_FALSE; - } - LOCK_HARDWARE(intel); + intelFlush(&intel->ctx); - if (intel->numClipRects) { - intel->vtbl.install_meta_state(intel); - intel->vtbl.meta_no_depth_write(intel); - intel->vtbl.meta_no_stencil_write(intel); + intel->vtbl.install_meta_state(intel); - if (!driClipRectToFramebuffer(ctx->ReadBuffer, &x, &y, &width, &height)) { - UNLOCK_HARDWARE(intel); - SET_STATE(i830, state); - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: cliprect failed\n", __FUNCTION__); - return GL_TRUE; - } + /* Is this true? Also will need to turn depth testing on according + * to state: + */ + intel->vtbl.meta_no_stencil_write(intel); + intel->vtbl.meta_no_depth_write(intel); - y = intel->pClipRects->y2 - y - height; + /* Set the 3d engine to draw into the destination region: + */ + if (ctx->DrawBuffer->_DepthBuffer && + ctx->DrawBuffer->_DepthBuffer->Wrapped) + depthreg = (intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped))->region; + intel->vtbl.meta_draw_region(intel, dest_region, depthreg); - /* Set the frontbuffer up as a large rectangular texture. - */ - intel->vtbl.meta_tex_rect_source(intel, src_region, textureFormat); + intel->vtbl.meta_import_pixel_state(intel); + if (src->cpp == 2) { + src_format = GL_RGB; + src_type = GL_UNSIGNED_SHORT_5_6_5; + } + else { + src_format = GL_BGRA; + src_type = GL_UNSIGNED_BYTE; + } - intel->vtbl.meta_texture_blend_replace(i830, glTextureFormat); + /* Set the frontbuffer up as a large rectangular texture. + */ + if (!intel->vtbl.meta_tex_rect_source(intel, src->buffer, 0, + src->pitch, + src->height, src_format, src_type)) { + intel->vtbl.leave_meta_state(intel); + return GL_FALSE; + } - /* Set the 3d engine to draw into the destination region: - */ + intel->vtbl.meta_texture_blend_replace(intel); - intel->vtbl.meta_draw_region(intel, dest_region); - intel->vtbl.meta_draw_format(intel, destFormat, depthFormat); /* ?? */ + LOCK_HARDWARE(intel); + { + int dstbufHeight = height * ctx->Pixel.ZoomY; /* ? */ + /* convert from gl to hardware coords */ + srcy = ctx->ReadBuffer->Height - srcy - height; - /* Draw a single quad, no cliprects: + /* Clip against the source region. This is the only source + * clipping we do. XXX: Just set the texcord wrap mode to clamp + * or similar. + * */ - intel->vtbl.meta_disable_cliprects(intel); + if (0) { + if (!_mesa_clip_to_region(0, 0, ctx->ReadBuffer->Width - 1, + ctx->ReadBuffer->Height - 1, + &srcx, &srcy, &width, &height)) + goto out; - intel->vtbl.draw_quad(intel, - 0, width, 0, height, - 0x00ff00ff, x, x + width, y, y + height); + } + /* Just use the regular cliprect mechanism... Does this need to + * even hold the lock??? + */ + intel_meta_draw_quad(intel, + 0, + width * ctx->Pixel.ZoomX, + dstbufHeight - (height * ctx->Pixel.ZoomY), + dstbufHeight, 0, /* XXX: what z value? */ + 0x00ff00ff, + srcx, srcx + width, srcy, srcy + height); + + out: intel->vtbl.leave_meta_state(intel); + intel_batchbuffer_flush(intel->batch); } UNLOCK_HARDWARE(intel); - intel_region_wait_fence(ctx, dest_region); /* required by GL */ + if (INTEL_DEBUG & DEBUG_PIXEL) + _mesa_printf("%s: success\n", __FUNCTION__); return GL_TRUE; #endif @@ -166,7 +187,8 @@ do_blit_readpixels(GLcontext * ctx, const struct gl_pixelstore_attrib *pack, GLvoid * pixels) { struct intel_context *intel = intel_context(ctx); - struct intel_region *src = intel_readbuf_region(intel); + struct intel_renderbuffer *irbread; + struct intel_region *src; struct intel_buffer_object *dst = intel_buffer_object(pack->BufferObj); GLuint dst_offset; GLuint rowLength; @@ -175,8 +197,10 @@ do_blit_readpixels(GLcontext * ctx, if (INTEL_DEBUG & DEBUG_PIXEL) _mesa_printf("%s\n", __FUNCTION__); - if (!src) + irbread = intel_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + if (!irbread || !irbread->region) return GL_FALSE; + src = irbread->region; if (dst) { /* XXX This validation should be done by core mesa: @@ -228,44 +252,35 @@ do_blit_readpixels(GLcontext * ctx, dst_offset = (GLuint) _mesa_image_address(2, pack, pixels, width, height, format, type, 0, 0, 0); - /* reading from priv buffer, cliprects should not change */ intelFlush(&intel->ctx); - if (intel->numClipRects) { - assert (intel->numClipRects == 1); + { + GLint srcx = x; + GLint srcy = irbread->Base.Height - (y + height); + GLint srcwidth = width; + GLint srcheight = height; + GLboolean all = (width * height * src->cpp == dst->Base.Size && x == 0 && dst_offset == 0); struct _DriBufferObject *dst_buffer = intel_bufferobj_buffer(intel, dst, all ? INTEL_WRITE_FULL : INTEL_WRITE_PART); - int nbox = intel->numClipRects; - drm_clip_rect_t *box = intel->pClipRects; - drm_clip_rect_t rect; - drm_clip_rect_t src_rect; - int i; - - src_rect.x1 = x; - src_rect.y1 = box->y2 - (y + height); - src_rect.x2 = src_rect.x1 + width; - src_rect.y2 = src_rect.y1 + height; - - - for (i = 0; i < nbox; i++) { - if (!intel_intersect_cliprects(&rect, &src_rect, &box[i])) - continue; + /* clip to src region */ + if (_mesa_clip_to_region(0, 0, irbread->Base.Width - 1, + irbread->Base.Height - 1, + &srcx, &srcy, &srcwidth, &srcheight)); + { intelEmitCopyBlit(intel, src->cpp, src->pitch, src->buffer, 0, rowLength, dst_buffer, dst_offset, - rect.x1, - rect.y1, - rect.x1 - src_rect.x1, - rect.y2 - src_rect.y2, - rect.x2 - rect.x1, rect.y2 - rect.y1, + srcx, srcy, + 0, 0, + width, height, GL_COPY); } -- cgit v1.2.3 From 6104044f6bd9464b5972234f95801ca6ba5db510 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Mon, 16 Jul 2007 15:33:09 +0200 Subject: get rid of more fake cliprects (drawpixel), and enable blit draw with scissor --- src/mesa/drivers/dri/i915tex/intel_pixel.c | 3 +- src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c | 15 ++-- src/mesa/drivers/dri/i915tex/intel_pixel_copy.c | 29 +----- src/mesa/drivers/dri/i915tex/intel_pixel_draw.c | 104 ++++++++++++++++------ 4 files changed, 87 insertions(+), 64 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel.c b/src/mesa/drivers/dri/i915tex/intel_pixel.c index 9018e3daef..8e61cc8184 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel.c @@ -44,13 +44,12 @@ intel_check_blit_fragment_ops(GLcontext * ctx) if (ctx->NewState) _mesa_update_state(ctx); - /* XXX Note: Scissor could be done with the blitter: + /* XXX Note: Scissor done with blitter: */ return !(ctx->_ImageTransferState || ctx->Color.AlphaEnabled || ctx->Depth.Test || ctx->Fog.Enabled || - ctx->Scissor.Enabled || ctx->Stencil.Enabled || !ctx->Color.ColorMask[0] || !ctx->Color.ColorMask[1] || diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c b/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c index 96d3cbef5f..c205c6a715 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c @@ -195,16 +195,17 @@ do_blit_bitmap( GLcontext *ctx, LOCK_HARDWARE(intel); - if (intel->numClipRects) { - assert(intel->numClipRects == 1); - drm_clip_rect_t *box = intel->pClipRects; + { + drm_clip_rect_t box; drm_clip_rect_t dest_rect; - GLint nbox = intel->numClipRects; GLint srcx = 0, srcy = 0; GLint orig_screen_x1, orig_screen_y2; GLuint i; - + box.x1 = 0; + box.y1 = 0; + box.x2 = ctx->DrawBuffer->Width; + box.y2 = ctx->DrawBuffer->Height; orig_screen_x1 = dstx; orig_screen_y2 = box->y2 - dsty; @@ -231,13 +232,13 @@ x if (ctx->Scissor.Enabled) dest_rect.x2 = dstx + width; dest_rect.y2 = dsty + height; - for (i = 0; i < nbox; i++) { + for (i = 0; i < 1; i++) { drm_clip_rect_t rect; int box_w, box_h; GLint px, py; GLuint stipple[32]; - if (!intel_intersect_cliprects(&rect, &dest_rect, &box[i])) + if (!intel_intersect_cliprects(&rect, &dest_rect, &box[0])) continue; /* Now go back to GL coordinates to figure out what subset of diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c index 57e746ae22..3148a2b0d4 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c @@ -80,33 +80,6 @@ copypix_src_region(struct intel_context *intel, GLenum type) } -/** - * Check if any fragment operations are in effect which might effect - * glCopyPixels. Differs from intel_check_blit_fragment_ops in that - * we allow Scissor. - */ -static GLboolean -intel_check_copypixel_blit_fragment_ops(GLcontext * ctx) -{ - if (ctx->NewState) - _mesa_update_state(ctx); - - /* Could do logicop with the blitter: - */ - return !(ctx->_ImageTransferState || - ctx->Color.AlphaEnabled || - ctx->Depth.Test || - ctx->Fog.Enabled || - ctx->Stencil.Enabled || - !ctx->Color.ColorMask[0] || - !ctx->Color.ColorMask[1] || - !ctx->Color.ColorMask[2] || - !ctx->Color.ColorMask[3] || - ctx->Texture._EnabledUnits || - ctx->FragmentProgram._Enabled || - ctx->Color.BlendEnabled); -} - /* Doesn't work for overlapping regions. Could do a double copy or * just fallback. */ @@ -269,7 +242,7 @@ do_blit_copypixels(GLcontext * ctx, /* Copypixels can be more than a straight copy. Ensure all the * extra operations are disabled: */ - if (!intel_check_copypixel_blit_fragment_ops(ctx) || + if (!intel_check_blit_fragment_ops(ctx) || ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F) return GL_FALSE; diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c index 2d46c4b8c0..b1ecf2b0ce 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c @@ -43,7 +43,7 @@ #include "intel_pixel.h" #include "intel_buffer_objects.h" #include "intel_tris.h" - +#include "intel_fbo.h" static GLboolean @@ -138,9 +138,8 @@ do_texture_drawpixels(GLcontext * ctx, LOCK_HARDWARE(intel); - if (intel->numClipRects) { - assert(intel->numClipRects == 1); - int bufHeight = intel->pClipRects->y2; + { + int bufHeight = ctx->DrawBuffer->Height; GLint srcx, srcy; GLint dstx, dsty; @@ -208,22 +207,50 @@ do_texture_drawpixels(GLcontext * ctx, */ static GLboolean do_blit_drawpixels(GLcontext * ctx, - GLint x, GLint y, + GLint dstx, GLint dsty, GLsizei width, GLsizei height, GLenum format, GLenum type, const struct gl_pixelstore_attrib *unpack, const GLvoid * pixels) { struct intel_context *intel = intel_context(ctx); - struct intel_region *dest = intel_drawbuf_region(intel); + struct intel_renderbuffer *irbdraw; + struct intel_region *dest; struct intel_buffer_object *src = intel_buffer_object(unpack->BufferObj); GLuint src_offset; GLuint rowLength; + GLuint height_orig = height; struct _DriFenceObject *fence = NULL; if (INTEL_DEBUG & DEBUG_PIXEL) _mesa_printf("%s\n", __FUNCTION__); + if (type == GL_COLOR) { + irbdraw = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); + if (!irbdraw || !irbdraw->region) + return GL_FALSE; + } + else if (type == GL_DEPTH) { + /* Don't think this is really possible execpt at 16bpp, when we have no stencil. + */ + irbdraw = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + if (!irbdraw || !irbdraw->region || !(irbdraw->region->cpp == 2)) + return GL_FALSE; + } + else if (GL_DEPTH_STENCIL_EXT) { + /* Does it matter whether it is stencil/depth or depth/stencil? + */ + irbdraw = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + if (!irbdraw || !irbdraw->region) + return GL_FALSE; + } + else if (GL_STENCIL) { + /* Don't think this is really possible. + */ + return GL_FALSE; + } + + dest = irbdraw->region; if (!dest) { if (INTEL_DEBUG & DEBUG_PIXEL) @@ -277,7 +304,7 @@ do_blit_drawpixels(GLcontext * ctx, if (INTEL_DEBUG & DEBUG_PIXEL) _mesa_printf("%s - bad PixelZoomY for blit\n", __FUNCTION__); return GL_FALSE; /* later */ - y -= height; + dsty -= height; } else if (ctx->Pixel.ZoomY == 1.0F) { rowLength = -rowLength; @@ -291,38 +318,59 @@ do_blit_drawpixels(GLcontext * ctx, src_offset = (GLuint) _mesa_image_address(2, unpack, pixels, width, height, format, type, 0, 0, 0); - /* don't need a lock as cliprects shouldn't change */ + /* don't need a lock as we have no cliprects ? */ intelFlush(&intel->ctx); - if (intel->numClipRects) { - assert(intel->numClipRects == 1); - int nbox = intel->numClipRects; - drm_clip_rect_t *box = intel->pClipRects; - drm_clip_rect_t rect; - drm_clip_rect_t dest_rect; + { + GLuint srcx = 0; + GLuint srcy = 0; + GLint dx = dstx; + GLint dy = dsty; + + /* Do scissoring in GL coordinates: + */ + if (ctx->Scissor.Enabled) + { + GLint x = ctx->Scissor.X; + GLint y = ctx->Scissor.Y; + GLuint w = ctx->Scissor.Width; + GLuint h = ctx->Scissor.Height; + height_orig = height; + + + if (!_mesa_clip_to_region(x, y, x+w-1, y+h-1, &dstx, &dsty, &width, &height)) + goto out; + + } + + /* no need to clip against pbo src region, but clip against dest */ + { + if (!_mesa_clip_to_region(0, 0, irbdraw->Base.Width - 1, + irbdraw->Base.Height - 1, + &dstx, &dsty, &width, &height)) + goto out; + + srcx = dstx - dx; + srcy = dsty - dy; + } + struct _DriBufferObject *src_buffer = intel_bufferobj_buffer(intel, src, INTEL_READ); - int i; - dest_rect.x1 = x; - dest_rect.y1 = box->y2 - (y + height); - dest_rect.x2 = dest_rect.x1 + width; - dest_rect.y2 = dest_rect.y1 + height; - - for (i = 0; i < nbox; i++) { - if (!intel_intersect_cliprects(&rect, &dest_rect, &box[i])) - continue; + /* Convert from GL to hardware coordinates: + */ + dsty = irbdraw->Base.Height - dsty - height; + srcy = height_orig - srcy - height; + { intelEmitCopyBlit(intel, dest->cpp, rowLength, src_buffer, src_offset, dest->pitch, dest->buffer, 0, - rect.x1 - dest_rect.x1, - rect.y2 - dest_rect.y2, - rect.x1, - rect.y1, rect.x2 - rect.x1, rect.y2 - rect.y1, + srcx, srcy, + dstx, dsty, width, height, ctx->Color.ColorLogicOpEnabled ? ctx->Color.LogicOp : GL_COPY); } @@ -335,6 +383,8 @@ do_blit_drawpixels(GLcontext * ctx, driFenceUnReference(fence); } + out: + if (INTEL_DEBUG & DEBUG_PIXEL) _mesa_printf("%s - DONE\n", __FUNCTION__); -- cgit v1.2.3 From 56cbdfd7d541e5d99b7e7c4fb7b03ac6ac1536a1 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Mon, 16 Jul 2007 15:53:23 +0200 Subject: more cliprect elimination (swrast span) --- src/mesa/drivers/dri/i915tex/intel_span.c | 19 +++++++------------ 1 file changed, 7 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_span.c b/src/mesa/drivers/dri/i915tex/intel_span.c index 74f118121c..5a978d9ce2 100644 --- a/src/mesa/drivers/dri/i915tex/intel_span.c +++ b/src/mesa/drivers/dri/i915tex/intel_span.c @@ -47,7 +47,6 @@ #define DBG 0 #define LOCAL_VARS \ - struct intel_context *intel = intel_context(ctx); \ struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ const GLint yScale = irb->RenderToTexture ? 1 : -1; \ const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ @@ -56,20 +55,17 @@ assert(irb->pfMap);\ (void) p; -/* XXX FBO: this is identical to the macro in spantmp2.h except we get - * the cliprect info from the context, not the driDrawable. - * Move this into spantmp2.h someday. +/* There is just a single cliploop! */ #define HW_CLIPLOOP() \ do { \ - int _nc = intel->numClipRects; \ - while ( _nc-- ) { \ - int minx = intel->pClipRects[_nc].x1; \ - int miny = intel->pClipRects[_nc].y1; \ - int maxx = intel->pClipRects[_nc].x2; \ - int maxy = intel->pClipRects[_nc].y2; - + int minx = 0; \ + int miny = 0; \ + int maxx = irb->Base.Width - 1; \ + int maxy = irb->Base.Height - 1; +#define HW_ENDCLIPLOOP() \ + } while (0) #define Y_FLIP(_y) ((_y) * yScale + yBias) @@ -100,7 +96,6 @@ #define LOCAL_DEPTH_VARS \ - struct intel_context *intel = intel_context(ctx); \ struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ const GLuint pitch = irb->pfPitch/***XXX region->pitch*/; /* in pixels */ \ const GLint yScale = irb->RenderToTexture ? 1 : -1; \ -- cgit v1.2.3 From 894557d625b6813dfb23602b96fde6a59d3db54e Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Mon, 16 Jul 2007 17:41:22 +0200 Subject: get rid of more cliprects... --- src/mesa/drivers/dri/i915tex/intel_blit.c | 33 ++++++++----------------- src/mesa/drivers/dri/i915tex/intel_pixel_draw.c | 7 ++++++ 2 files changed, 17 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c index db72f59a2c..3ecd2e8e48 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ b/src/mesa/drivers/dri/i915tex/intel_blit.c @@ -402,10 +402,8 @@ intelClearWithBlit(GLcontext * ctx, GLbitfield mask) LOCK_HARDWARE(intel); if (intel->numClipRects) { - assert(intel->numClipRects == 1); GLint cx, cy, cw, ch; - drm_clip_rect_t clear; - int i; + drm_clip_rect_t b; /* Get clear bounds after locking */ cx = fb->_Xmin; @@ -416,36 +414,25 @@ intelClearWithBlit(GLcontext * ctx, GLbitfield mask) if (fb->Name == 0) { /* clearing a window */ /* flip top to bottom */ - clear.x1 = cx; - clear.y1 = intel->pClipRects->y2 - cy - ch; - clear.x2 = clear.x1 + cw; - clear.y2 = clear.y1 + ch; + b.x1 = cx; + b.y1 = fb->Height - cy - ch; + b.x2 = b.x1 + cw; + b.y2 = b.y1 + ch; } else { /* clearing FBO */ - assert(intel->numClipRects == 1); - assert(intel->pClipRects == &intel->fboRect); - clear.x1 = cx; - clear.y1 = cy; - clear.x2 = clear.x1 + cw; - clear.y2 = clear.y1 + ch; + b.x1 = cx; + b.y1 = cy; + b.x2 = b.x1 + cw; + b.y2 = b.y1 + ch; /* no change to mask */ } - for (i = 0; i < intel->numClipRects; i++) { - const drm_clip_rect_t *box = &intel->pClipRects[i]; - drm_clip_rect_t b; + { GLuint buf; GLuint clearMask = mask; /* use copy, since we modify it below */ GLboolean all = (cw == fb->Width && ch == fb->Height); - if (!all) { - intel_intersect_cliprects(&b, &clear, box); - } - else { - b = *box; - } - DBG("clear %d,%d..%d,%d, mask %x\n", b.x1, b.y1, b.x2, b.y2, mask); diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c index b1ecf2b0ce..1c83028d8a 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c @@ -421,10 +421,17 @@ intelDrawPixels(GLcontext * ctx, * wise happily run the fragment program on each pixel in the image). */ struct gl_fragment_program *fpSave = ctx->FragmentProgram._Current; + /* can't just set current frag prog to 0 here as on buffer resize + we'll get new state checks which will segfault (actually don't get them + with current priv buffers). Remains a hack. */ ctx->FragmentProgram._Current = NULL; + ctx->FragmentProgram._UseTexEnvProgram = GL_FALSE; + ctx->FragmentProgram._Active = GL_FALSE; _swrast_DrawPixels( ctx, x, y, width, height, format, type, unpack, pixels ); ctx->FragmentProgram._Current = fpSave; + ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; + ctx->FragmentProgram._Active = GL_TRUE; } else { _swrast_DrawPixels( ctx, x, y, width, height, format, type, -- cgit v1.2.3 From 5fc7973d5493039e4b3037eae10e9a005d4cf717 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Mon, 16 Jul 2007 19:01:47 +0200 Subject: small cleanups --- src/mesa/drivers/dri/i915tex/intel_buffers.c | 44 ++-------------------------- src/mesa/drivers/dri/i915tex/intel_context.h | 3 +- 2 files changed, 4 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 59b291a709..d3d97f8df0 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -125,13 +125,14 @@ intel_readbuf_region(struct intel_context *intel) /** - * Update the following fields for rendering to a user-created FBO: + * Update the following fields for rendering: * intel->numClipRects * intel->pClipRects */ static void intelSetRenderbufferClipRects(struct intel_context *intel) { + /* zero-sized buffers might be legal? */ assert(intel->ctx.DrawBuffer->Width > 0); assert(intel->ctx.DrawBuffer->Height > 0); intel->fboRect.x1 = 0; @@ -142,28 +143,6 @@ intelSetRenderbufferClipRects(struct intel_context *intel) intel->pClipRects = &intel->fboRect; } -/** - * As above, but for rendering private front/back buffer of a window. - * \sa intelSetPrivbufClipRects - */ - -static void -intelSetPrivbufClipRects(struct intel_context *intel) -{ - if (!intel->ctx.DrawBuffer) { - fprintf(stderr, "%s: DrawBuffer not set!\n", __FUNCTION__); - return; - } - - intel->fakeClipRect.x1 = 0; - intel->fakeClipRect.y1 = 0; - intel->fakeClipRect.x2 = intel->ctx.DrawBuffer->Width; - intel->fakeClipRect.y2 = intel->ctx.DrawBuffer->Height; - intel->numClipRects = 1; - intel->pClipRects = &intel->fakeClipRect; -} - - /** * This will be called whenever the currently bound window is moved/resized. @@ -327,7 +306,6 @@ intelClearWithTris(struct intel_context *intel, GLbitfield mask) LOCK_HARDWARE(intel); - /* XXX FBO: was: intel->driDrawable->numClipRects */ if (intel->numClipRects) { GLint cx, cy, cw, ch; GLuint buf; @@ -930,7 +908,6 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) struct intel_context *intel = intel_context(ctx); struct intel_region *colorRegion, *depthRegion = NULL; struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; - int front = 0; /* drawing to front color buffer? */ if (!fb) { /* this can happen during the initial context initialization */ @@ -965,33 +942,18 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) /* writing to 0 or 2 or 4 color buffers */ /*_mesa_debug(ctx, "Software rendering\n");*/ FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); - front = 1; /* might not have back color buffer */ } else { /* draw to exactly one color buffer */ /*_mesa_debug(ctx, "Hardware rendering\n");*/ FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); - if (fb->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT) { - front = 1; - } } /* * Get the intel_renderbuffer for the colorbuffer we're drawing into. * And set up cliprects. */ - if (fb->Name == 0) { - intelSetPrivbufClipRects(intel); - /* drawing to window system buffer */ - if (front) { - colorRegion = intel_get_rb_region(fb, BUFFER_FRONT_LEFT); - } - else { - colorRegion = intel_get_rb_region(fb, BUFFER_BACK_LEFT); - } - } - else { - /* drawing to user-created FBO */ + { struct intel_renderbuffer *irb; intelSetRenderbufferClipRects(intel); irb = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index b5defdc47e..e2008c4e18 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -252,8 +252,7 @@ struct intel_context */ GLuint numClipRects; /**< cliprects for drawing */ drm_clip_rect_t *pClipRects; - drm_clip_rect_t fboRect; /**< cliprect for FBO rendering */ - drm_clip_rect_t fakeClipRect; /**< cliprect for priv back/fake front buffers rendering */ + drm_clip_rect_t fboRect; /**< cliprect for rendering */ int perf_boxes; -- cgit v1.2.3 From cc9cf90b090b80e51ac03af6fe113287e083accd Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Mon, 16 Jul 2007 20:53:33 +0200 Subject: fix resize bugs (fb size updated too late), fix typos, cleanups... --- src/mesa/drivers/dri/i915tex/i915_state.c | 2 ++ src/mesa/drivers/dri/i915tex/intel_buffers.c | 10 ++++++---- src/mesa/drivers/dri/i915tex/intel_pixel_copy.c | 4 ++-- src/mesa/drivers/dri/i915tex/intel_pixel_draw.c | 4 ++-- src/mesa/drivers/dri/i915tex/intel_screen.c | 2 ++ 5 files changed, 14 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/i915_state.c b/src/mesa/drivers/dri/i915tex/i915_state.c index e5d8d27993..f523206bef 100644 --- a/src/mesa/drivers/dri/i915tex/i915_state.c +++ b/src/mesa/drivers/dri/i915tex/i915_state.c @@ -904,6 +904,7 @@ i915_init_packets(struct i915_context *i915) { I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); +#if 0 /* color buffer offset/stride */ i915->state.Buffer[I915_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; /* XXX FBO: remove this? Also get set in i915_set_draw_region() */ @@ -914,6 +915,7 @@ i915_init_packets(struct i915_context *i915) /* XXX FBO: remove this? Also get set in i915_set_draw_region() */ i915->state.Buffer[I915_DESTREG_DBUFADDR1] = (BUF_3D_ID_DEPTH | BUF_3D_PITCH(screen->depth.pitch) | /* pitch in bytes */ BUF_3D_USE_FENCE); +#endif i915->state.Buffer[I915_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD; diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index d3d97f8df0..74e2bfd769 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -160,6 +160,9 @@ intelWindowMoved(struct intel_context *intel) intel->numClipRects = 0; } + /* Update Mesa's notion of window size */ + driUpdateFramebufferSize(ctx, dPriv); + intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ if (intel->intelScreen->driScrnPriv->ddxMinor >= 7) { drmI830Sarea *sarea = intel->sarea; @@ -277,10 +280,6 @@ intelWindowMoved(struct intel_context *intel) intel_fb->vblank_flags &= ~VBLANK_FLAG_SECONDARY; } - /* Update Mesa's notion of window size */ - driUpdateFramebufferSize(ctx, dPriv); - intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ - /* Update hardware scissor */ ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, ctx->Scissor.Width, ctx->Scissor.Height); @@ -707,6 +706,9 @@ intelPageFlip(const __DRIdrawablePrivate * dPriv) intel_flip_renderbuffers(intel_fb); intel_draw_buffer(&intel->ctx, &intel_fb->Base); + if (INTEL_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s: success\n", __FUNCTION__); + return GL_TRUE; } diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c index 3148a2b0d4..ad111d4d72 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c @@ -263,7 +263,7 @@ do_blit_copypixels(GLcontext * ctx, || !(irbread->region->cpp == 2)) return GL_FALSE; } - else if (GL_DEPTH_STENCIL_EXT) { + else if (type == GL_DEPTH_STENCIL_EXT) { /* Does it matter whether it is stencil/depth or depth/stencil? */ irbread = intel_renderbuffer(ctx->ReadBuffer->_DepthBuffer->Wrapped); @@ -271,7 +271,7 @@ do_blit_copypixels(GLcontext * ctx, if (!irbread || !irbread->region || !irbdraw || !irbdraw->region) return GL_FALSE; } - else if (GL_STENCIL) { + else if (type == GL_STENCIL) { /* Don't think this is really possible. */ return GL_FALSE; diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c index 1c83028d8a..eebd63ec01 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c @@ -237,14 +237,14 @@ do_blit_drawpixels(GLcontext * ctx, if (!irbdraw || !irbdraw->region || !(irbdraw->region->cpp == 2)) return GL_FALSE; } - else if (GL_DEPTH_STENCIL_EXT) { + else if (type == GL_DEPTH_STENCIL_EXT) { /* Does it matter whether it is stencil/depth or depth/stencil? */ irbdraw = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); if (!irbdraw || !irbdraw->region) return GL_FALSE; } - else if (GL_STENCIL) { + else if (type == GL_STENCIL) { /* Don't think this is really possible. */ return GL_FALSE; diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c index dcc2992383..f0d1439dd7 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ b/src/mesa/drivers/dri/i915tex/intel_screen.c @@ -355,6 +355,7 @@ intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, intelScreen->front.handle = sarea->front_handle; intelScreen->front.size = sarea->front_size; +#if 0 intelScreen->back.offset = sarea->back_offset; intelScreen->back.pitch = sarea->pitch * intelScreen->cpp; intelScreen->back.handle = sarea->back_handle; @@ -376,6 +377,7 @@ intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, intelScreen->logTextureGranularity = sarea->log_tex_granularity; intelScreen->tex.handle = sarea->tex_handle; intelScreen->tex.size = sarea->tex_size; +#endif intelScreen->rotated.offset = sarea->rotated_offset; intelScreen->rotated.pitch = sarea->rotated_pitch * intelScreen->cpp; -- cgit v1.2.3 From 4dd8e2648d7ec8634d1005aae2da562b57b0267f Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 17 Jul 2007 12:56:30 +0200 Subject: increase MAX_RELOCS so never run out before batch buffer is full (fixes xdemos/shape) --- src/mesa/drivers/dri/i915tex/intel_batchbuffer.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h index 59261f7274..212f130101 100644 --- a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h @@ -9,7 +9,7 @@ struct intel_context; #define BATCH_SZ 16384 #define BATCH_RESERVED 16 -#define MAX_RELOCS 400 +#define MAX_RELOCS 4096 #define INTEL_BATCH_NO_CLIPRECTS 0x1 #define INTEL_BATCH_CLIPRECTS 0x2 -- cgit v1.2.3 From 04d49ff24d0f9c7c3a99c45b50f62aa32b0a28ed Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 17 Jul 2007 15:56:43 +0200 Subject: remove old code, remaining bits of static handles, disable (no longer functional anyway) rotation code --- src/mesa/drivers/dri/i915tex/i830_state.c | 10 - src/mesa/drivers/dri/i915tex/intel_buffers.c | 12 +- src/mesa/drivers/dri/i915tex/intel_buffers.h | 2 + src/mesa/drivers/dri/i915tex/intel_context.c | 34 +-- src/mesa/drivers/dri/i915tex/intel_pixel_copy.c | 2 +- src/mesa/drivers/dri/i915tex/intel_pixel_draw.c | 9 +- src/mesa/drivers/dri/i915tex/intel_screen.c | 266 +----------------------- src/mesa/drivers/dri/i915tex/intel_screen.h | 9 - src/mesa/main/framebuffer.c | 2 + 9 files changed, 32 insertions(+), 314 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/i830_state.c b/src/mesa/drivers/dri/i915tex/i830_state.c index 3c149e6905..490d681429 100644 --- a/src/mesa/drivers/dri/i915tex/i830_state.c +++ b/src/mesa/drivers/dri/i915tex/i830_state.c @@ -879,7 +879,6 @@ i830Enable(GLcontext * ctx, GLenum cap, GLboolean state) static void i830_init_packets(struct i830_context *i830) { - intelScreenPrivate *screen = i830->intel.intelScreen; /* Zero all state */ memset(&i830->state, 0, sizeof(i830->state)); @@ -1033,15 +1032,6 @@ i830_init_packets(struct i830_context *i830) i830->state.Stipple[I830_STPREG_ST0] = _3DSTATE_STIPPLE; - i830->state.Buffer[I830_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - i830->state.Buffer[I830_DESTREG_CBUFADDR1] = (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(screen->front.pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - - - i830->state.Buffer[I830_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - i830->state.Buffer[I830_DESTREG_DBUFADDR1] = (BUF_3D_ID_DEPTH | BUF_3D_PITCH(screen->depth.pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - i830->state.Buffer[I830_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD; #if 0 diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 74e2bfd769..d06fc0eaea 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -193,7 +193,7 @@ intelWindowMoved(struct intel_context *intel) intel_fb->pf_current_page = (intel->sarea->pf_current_page >> (intel_fb->pf_pipes & 0x2)) & 0x3; - intel_fb->pf_num_pages = intel->intelScreen->third.handle ? 3 : 2; + intel_fb->pf_num_pages = 2 /*intel->intelScreen->third.handle ? 3 : 2*/; pf_active = pf_pipes && (pf_pipes & intel->sarea->pf_active) == pf_pipes; @@ -396,8 +396,9 @@ intelClearWithTris(struct intel_context *intel, GLbitfield mask) } - - +#if 0 +/* Xorg seems to handle that nowadays, and this code here no longer works with priv + buffers */ /** * Copy the window contents named by dPriv to the rotated (or reflected) * color buffer. @@ -537,7 +538,7 @@ intelRotateWindow(struct intel_context *intel, UNLOCK_HARDWARE(intel); } - +#endif /** * Called by ctx->Driver.Clear. @@ -845,10 +846,11 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) if (screen->current_rotation != 0 || !intelPageFlip(dPriv)) { intelCopyBuffer(dPriv, NULL); } - +#if 0 if (screen->current_rotation != 0) { intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT); } +#endif } intel_fb->swap_count++; diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.h b/src/mesa/drivers/dri/i915tex/intel_buffers.h index 3b686cb5c1..3c7a9cf139 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.h +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.h @@ -52,8 +52,10 @@ extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); extern void intelInitBufferFuncs(struct dd_function_table *functions); +#if 0 extern void intelRotateWindow(struct intel_context *intel, __DRIdrawablePrivate * dPriv, GLuint srcBuf); +#endif #endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index ec4a7a746d..ea2b7e5277 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -295,14 +295,18 @@ intelCheckFrontUpdate(GLcontext * ctx) BUFFER_BIT_FRONT_LEFT maybe? */ if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT) { - intelScreenPrivate *screen = intel->intelScreen; __DRIdrawablePrivate *dPriv = intel->driDrawable; +#if 0 + intelScreenPrivate *screen = intel->intelScreen; if (screen->current_rotation != 0) { intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT); } else { intelCopyBuffer(dPriv, NULL); } +#else + intelCopyBuffer(dPriv, NULL); +#endif } } @@ -596,34 +600,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, is done. Need a per-screen context? */ intel->intelScreen->dummyctxptr = intel; - /* XXX FBO temporary fix-ups! */ - /* if the renderbuffers don't have regions, init them from the context */ - { - struct intel_renderbuffer *irbDepth - = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH); - struct intel_renderbuffer *irbStencil - = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL); - - if (intel_fb->color_rb[0] && !intel_fb->color_rb[0]->region) { - intel_region_reference(&intel_fb->color_rb[0]->region, - intel->intelScreen->front_region); - } - if (intel_fb->color_rb[1] && !intel_fb->color_rb[1]->region) { - intel_region_reference(&intel_fb->color_rb[1]->region, - intel->intelScreen->back_region); - } - if (intel_fb->color_rb[2] && !intel_fb->color_rb[2]->region) { - intel_region_reference(&intel_fb->color_rb[2]->region, - intel->intelScreen->third_region); - } - if (irbDepth && !irbDepth->region) { - intel_region_reference(&irbDepth->region, intel->intelScreen->depth_region); - } - if (irbStencil && !irbStencil->region) { - intel_region_reference(&irbStencil->region, intel->intelScreen->depth_region); - } - } - /* update GLframebuffer size to match window if needed */ driUpdateFramebufferSize(&intel->ctx, driDrawPriv); diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c index ad111d4d72..58c415215e 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c @@ -271,7 +271,7 @@ do_blit_copypixels(GLcontext * ctx, if (!irbread || !irbread->region || !irbdraw || !irbdraw->region) return GL_FALSE; } - else if (type == GL_STENCIL) { + else /* GL_STENCIL */ { /* Don't think this is really possible. */ return GL_FALSE; diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c index eebd63ec01..4d6b4af589 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c @@ -57,6 +57,7 @@ do_texture_drawpixels(GLcontext * ctx, struct intel_context *intel = intel_context(ctx); struct intel_region *dst = intel_drawbuf_region(intel); struct intel_buffer_object *src = intel_buffer_object(unpack->BufferObj); + struct intel_region *depthreg = NULL; GLuint rowLength = unpack->RowLength ? unpack->RowLength : width; GLuint src_offset; @@ -112,7 +113,11 @@ do_texture_drawpixels(GLcontext * ctx, /* Set the 3d engine to draw into the destination region: */ - intel->vtbl.meta_draw_region(intel, dst, intel->intelScreen->depth_region); + if (ctx->DrawBuffer->_DepthBuffer && + ctx->DrawBuffer->_DepthBuffer->Wrapped) + depthreg = (intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped))->region; + + intel->vtbl.meta_draw_region(intel, dst, depthreg); intel->vtbl.meta_import_pixel_state(intel); @@ -244,7 +249,7 @@ do_blit_drawpixels(GLcontext * ctx, if (!irbdraw || !irbdraw->region) return GL_FALSE; } - else if (type == GL_STENCIL) { + else /* GL_STENCIL */ { /* Don't think this is really possible. */ return GL_FALSE; diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c index f0d1439dd7..4cabc22225 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ b/src/mesa/drivers/dri/i915tex/intel_screen.c @@ -88,55 +88,8 @@ intelMapScreenRegions(__DRIscreenPrivate * sPriv) _mesa_warning(NULL, "no front buffer handle in intelMapScreenRegions!"); } -#if 0 - if (0) - _mesa_printf("Back 0x%08x ", intelScreen->back.handle); - if (drmMap(sPriv->fd, - intelScreen->back.handle, - intelScreen->back.size, - (drmAddress *) & intelScreen->back.map) != 0) { - intelUnmapScreenRegions(intelScreen); - return GL_FALSE; - } - - if (intelScreen->third.handle) { - if (0) - _mesa_printf("Third 0x%08x ", intelScreen->third.handle); - if (drmMap(sPriv->fd, - intelScreen->third.handle, - intelScreen->third.size, - (drmAddress *) & intelScreen->third.map) != 0) { - intelUnmapScreenRegions(intelScreen); - return GL_FALSE; - } - } - - if (0) - _mesa_printf("Depth 0x%08x ", intelScreen->depth.handle); - if (drmMap(sPriv->fd, - intelScreen->depth.handle, - intelScreen->depth.size, - (drmAddress *) & intelScreen->depth.map) != 0) { - intelUnmapScreenRegions(intelScreen); - return GL_FALSE; - } -#endif - -#if 0 - _mesa_printf("TEX 0x%08x ", intelScreen->tex.handle); - if (drmMap(sPriv->fd, - intelScreen->tex.handle, - intelScreen->tex.size, - (drmAddress *) & intelScreen->tex.map) != 0) { - intelUnmapScreenRegions(intelScreen); - return GL_FALSE; - } -#endif if (0) - printf("Mappings: front: %p back: %p third: %p depth: %p tex: %p\n", - intelScreen->front.map, - intelScreen->back.map, intelScreen->third.map, - intelScreen->depth.map, intelScreen->tex.map); + printf("Mappings: front: %p\n", intelScreen->front.map); return GL_TRUE; } @@ -161,12 +114,8 @@ intel_recreate_static(intelScreenPrivate *intelScreen, /* Create intel_region structs to describe the static front,back,depth - * buffers created by the xserver. - * - * Although FBO's mean we now no longer use these as render targets in - * all circumstances, they won't go away until the back and depth - * buffers become private, and the front and rotated buffers will - * remain even then. + * buffers created by the xserver. + * Only used for real front buffer now. * * Note that these don't allocate video memory, just describe * allocations alread made by the X server. @@ -185,51 +134,6 @@ intel_recreate_static_regions(intelScreenPrivate *intelScreen) intelScreen->front.pitch / intelScreen->cpp, intelScreen->height); - intelScreen->rotated_region = - intel_recreate_static(intelScreen, - intelScreen->rotated_region, - DRM_BO_FLAG_MEM_TT, - intelScreen->rotated.offset, - intelScreen->rotated.map, - intelScreen->cpp, - intelScreen->rotated.pitch / - intelScreen->cpp, intelScreen->height); - -#if 0 - intelScreen->back_region = - intel_recreate_static(intelScreen, - intelScreen->back_region, - DRM_BO_FLAG_MEM_TT, - intelScreen->back.offset, - intelScreen->back.map, - intelScreen->cpp, - intelScreen->back.pitch / intelScreen->cpp, - intelScreen->height); - - if (intelScreen->third.handle) { - intelScreen->third_region = - intel_recreate_static(intelScreen, - intelScreen->third_region, - DRM_BO_FLAG_MEM_TT, - intelScreen->third.offset, - intelScreen->third.map, - intelScreen->cpp, - intelScreen->third.pitch / intelScreen->cpp, - intelScreen->height); - } - - /* Still assuming front.cpp == depth.cpp - */ - intelScreen->depth_region = - intel_recreate_static(intelScreen, - intelScreen->depth_region, - DRM_BO_FLAG_MEM_TT, - intelScreen->depth.offset, - intelScreen->depth.map, - intelScreen->cpp, - intelScreen->depth.pitch / intelScreen->cpp, - intelScreen->height); -#endif } /** @@ -261,32 +165,6 @@ intelUnmapScreenRegions(intelScreenPrivate * intelScreen) #endif intelScreen->front.map = NULL; } - if (intelScreen->back.map) { -#if REALLY_UNMAP - if (drmUnmap(intelScreen->back.map, intelScreen->back.size) != 0) - printf("drmUnmap back failed!\n"); -#endif - intelScreen->back.map = NULL; - } - if (intelScreen->third.map) { -#if REALLY_UNMAP - if (drmUnmap(intelScreen->third.map, intelScreen->third.size) != 0) - printf("drmUnmap third failed!\n"); -#endif - intelScreen->third.map = NULL; - } - if (intelScreen->depth.map) { -#if REALLY_UNMAP - drmUnmap(intelScreen->depth.map, intelScreen->depth.size); - intelScreen->depth.map = NULL; -#endif - } - if (intelScreen->tex.map) { -#if REALLY_UNMAP - drmUnmap(intelScreen->tex.map, intelScreen->tex.size); - intelScreen->tex.map = NULL; -#endif - } } @@ -297,17 +175,6 @@ intelPrintDRIInfo(intelScreenPrivate * intelScreen, fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", intelScreen->front.size, intelScreen->front.offset, intelScreen->front.pitch); - fprintf(stderr, "*** Back size: 0x%x offset: 0x%x pitch: %d\n", - intelScreen->back.size, intelScreen->back.offset, - intelScreen->back.pitch); - fprintf(stderr, "*** Depth size: 0x%x offset: 0x%x pitch: %d\n", - intelScreen->depth.size, intelScreen->depth.offset, - intelScreen->depth.pitch); - fprintf(stderr, "*** Rotated size: 0x%x offset: 0x%x pitch: %d\n", - intelScreen->rotated.size, intelScreen->rotated.offset, - intelScreen->rotated.pitch); - fprintf(stderr, "*** Texture size: 0x%x offset: 0x%x\n", - intelScreen->tex.size, intelScreen->tex.offset); fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); } @@ -355,39 +222,6 @@ intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, intelScreen->front.handle = sarea->front_handle; intelScreen->front.size = sarea->front_size; -#if 0 - intelScreen->back.offset = sarea->back_offset; - intelScreen->back.pitch = sarea->pitch * intelScreen->cpp; - intelScreen->back.handle = sarea->back_handle; - intelScreen->back.size = sarea->back_size; - - if (intelScreen->driScrnPriv->ddxMinor >= 8) { - intelScreen->third.offset = sarea->third_offset; - intelScreen->third.pitch = sarea->pitch * intelScreen->cpp; - intelScreen->third.handle = sarea->third_handle; - intelScreen->third.size = sarea->third_size; - } - - intelScreen->depth.offset = sarea->depth_offset; - intelScreen->depth.pitch = sarea->pitch * intelScreen->cpp; - intelScreen->depth.handle = sarea->depth_handle; - intelScreen->depth.size = sarea->depth_size; - - intelScreen->tex.offset = sarea->tex_offset; - intelScreen->logTextureGranularity = sarea->log_tex_granularity; - intelScreen->tex.handle = sarea->tex_handle; - intelScreen->tex.size = sarea->tex_size; -#endif - - intelScreen->rotated.offset = sarea->rotated_offset; - intelScreen->rotated.pitch = sarea->rotated_pitch * intelScreen->cpp; - intelScreen->rotated.size = sarea->rotated_size; - intelScreen->current_rotation = sarea->rotation; - matrix23Rotate(&intelScreen->rotMatrix, - sarea->width, sarea->height, sarea->rotation); - intelScreen->rotatedWidth = sarea->virtualX; - intelScreen->rotatedHeight = sarea->virtualY; - if (1) intelPrintSAREA(sarea); } @@ -495,23 +329,6 @@ intelInitDriver(__DRIscreenPrivate * sPriv) return GL_FALSE; } -#if 0 - - /* - * FIXME: Remove this code and its references. - */ - - intelScreen->tex.offset = gDRIPriv->textureOffset; - intelScreen->logTextureGranularity = gDRIPriv->logTextureGranularity; - intelScreen->tex.handle = gDRIPriv->textures; - intelScreen->tex.size = gDRIPriv->textureSize; - -#else - intelScreen->tex.offset = 0; - intelScreen->logTextureGranularity = 0; - intelScreen->tex.handle = 0; - intelScreen->tex.size = 0; -#endif intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset; @@ -589,7 +406,9 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, __DRIdrawablePrivate * driDrawPriv, const __GLcontextModes * mesaVis, GLboolean isPixmap) { +#if 0 intelScreenPrivate *screen = (intelScreenPrivate *) driScrnPriv->private; +#endif if (isPixmap) { return GL_FALSE; /* not implemented */ @@ -606,77 +425,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); -#if 0 - /* setup the hardware-based renderbuffers */ - { - intel_fb->color_rb[0] - = intel_create_renderbuffer(rgbFormat, - screen->width, screen->height, - screen->front.offset, - screen->front.pitch, - screen->cpp, - screen->front.map); - intel_set_span_functions(&intel_fb->color_rb[0]->Base); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, - &intel_fb->color_rb[0]->Base); - } - - if (mesaVis->doubleBufferMode) { - intel_fb->color_rb[1] - = intel_create_renderbuffer(rgbFormat, - screen->width, screen->height, - screen->back.offset, - screen->back.pitch, - screen->cpp, - screen->back.map); - intel_set_span_functions(&intel_fb->color_rb[1]->Base); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, - &intel_fb->color_rb[1]->Base); - - if (screen->third.handle) { - struct gl_renderbuffer *tmp_rb = NULL; - - intel_fb->color_rb[2] - = intel_create_renderbuffer(rgbFormat, - screen->width, screen->height, - screen->third.offset, - screen->third.pitch, - screen->cpp, - screen->third.map); - intel_set_span_functions(&intel_fb->color_rb[2]->Base); - _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base); - } - } - if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { - /* combined depth/stencil buffer */ - struct intel_renderbuffer *depthStencilRb - = intel_create_renderbuffer(GL_DEPTH24_STENCIL8_EXT, - screen->width, screen->height, - screen->depth.offset, - screen->depth.pitch, - screen->cpp, /* 4! */ - screen->depth.map); - intel_set_span_functions(&depthStencilRb->Base); - /* note: bind RB to two attachment points */ - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, - &depthStencilRb->Base); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL, - &depthStencilRb->Base); - } - else if (mesaVis->depthBits == 16) { - /* just 16-bit depth buffer, no hw stencil */ - struct intel_renderbuffer *depthRb - = intel_create_renderbuffer(GL_DEPTH_COMPONENT16, - screen->width, screen->height, - screen->depth.offset, - screen->depth.pitch, - screen->cpp, /* 2! */ - screen->depth.map); - intel_set_span_functions(&depthRb->Base); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); - } - -#else { /* fake frontbuffer */ /* XXX allocation should only happen in the unusual case @@ -693,6 +441,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, &intel_fb->color_rb[1]->Base); +#if 0 if (screen->third.handle) { struct gl_renderbuffer *tmp_rb = NULL; @@ -700,7 +449,9 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, = intel_new_renderbuffer_fb(rgbFormat); _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base); } +#endif } + if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { /* combined depth/stencil buffer */ struct intel_renderbuffer *depthStencilRb @@ -718,7 +469,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); } -#endif /* now add any/all software-based renderbuffers we may need */ _mesa_add_soft_renderbuffers(&intel_fb->Base, diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.h b/src/mesa/drivers/dri/i915tex/intel_screen.h index 783ee35524..89671e12c9 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.h +++ b/src/mesa/drivers/dri/i915tex/intel_screen.h @@ -50,17 +50,8 @@ typedef struct typedef struct { intelRegion front; - intelRegion back; - intelRegion third; - intelRegion rotated; - intelRegion depth; - intelRegion tex; struct intel_region *front_region; - struct intel_region *back_region; - struct intel_region *third_region; - struct intel_region *depth_region; - struct intel_region *rotated_region; int deviceID; int width; diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c index 1fd31a5321..a95e44c513 100644 --- a/src/mesa/main/framebuffer.c +++ b/src/mesa/main/framebuffer.c @@ -668,7 +668,9 @@ update_color_read_buffer(GLcontext *ctx, struct gl_framebuffer *fb) void _mesa_update_framebuffer(GLcontext *ctx) { + /* XXX might not be quite correct for different draw/read buffers ? */ struct gl_framebuffer *fb = ctx->DrawBuffer; +// struct gl_framebuffer *fbread = ctx->ReadBuffer; /* Completeness only matters for user-created framebuffers */ if (fb->Name != 0) { -- cgit v1.2.3 From d9499a151df85fef6bb6f31b58657e44cadba50a Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 17 Jul 2007 17:29:55 +0200 Subject: fix mesa fb binding Make sure that we bind the right buffer (draw or read) when rebinding the window framebuffer (the api doesn't allow binding different draw and read buffers at the same time, but the default window framebuffer is basically 2 fb objects, one for read, one for write, which can be different). Pass both of these two down the driver api (no driver uses this right now). --- src/mesa/drivers/dri/i915tex/intel_fbo.c | 2 +- src/mesa/drivers/dri/nouveau/nouveau_buffers.c | 7 +++++-- src/mesa/main/dd.h | 2 +- src/mesa/main/fbobject.c | 8 +++++--- 4 files changed, 12 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c index afee07c689..b0c243ebe0 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c @@ -513,7 +513,7 @@ intel_new_renderbuffer(GLcontext * ctx, GLuint name) */ static void intel_bind_framebuffer(GLcontext * ctx, GLenum target, - struct gl_framebuffer *fb) + struct gl_framebuffer *fb, struct gl_framebuffer *fbread) { if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) { intel_draw_buffer(ctx, fb); diff --git a/src/mesa/drivers/dri/nouveau/nouveau_buffers.c b/src/mesa/drivers/dri/nouveau/nouveau_buffers.c index 857cd30584..6a2c2013e3 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_buffers.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_buffers.c @@ -395,9 +395,12 @@ nouveauNewRenderbuffer(GLcontext *ctx, GLuint name) } static void -nouveauBindFramebuffer(GLcontext *ctx, GLenum target, struct gl_framebuffer *fb) +nouveauBindFramebuffer(GLcontext *ctx, GLenum target, + struct gl_framebuffer *fb, struct gl_framebuffer *fbread) { - nouveau_build_framebuffer(ctx, fb); + if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) { + nouveau_build_framebuffer(ctx, fb); + } } static void diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index 88f33943b3..caa50dd682 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -782,7 +782,7 @@ struct dd_function_table { struct gl_framebuffer * (*NewFramebuffer)(GLcontext *ctx, GLuint name); struct gl_renderbuffer * (*NewRenderbuffer)(GLcontext *ctx, GLuint name); void (*BindFramebuffer)(GLcontext *ctx, GLenum target, - struct gl_framebuffer *fb); + struct gl_framebuffer *fb, struct gl_framebuffer *fbread); void (*FramebufferRenderbuffer)(GLcontext *ctx, struct gl_framebuffer *fb, GLenum attachment, diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index e3bada5ae8..5345310ca1 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -924,7 +924,7 @@ check_end_texture_render(GLcontext *ctx, struct gl_framebuffer *fb) void GLAPIENTRY _mesa_BindFramebufferEXT(GLenum target, GLuint framebuffer) { - struct gl_framebuffer *newFb; + struct gl_framebuffer *newFb, *newFbread; GLboolean bindReadBuf, bindDrawBuf; GET_CURRENT_CONTEXT(ctx); @@ -984,12 +984,14 @@ _mesa_BindFramebufferEXT(GLenum target, GLuint framebuffer) } _mesa_HashInsert(ctx->Shared->FrameBuffers, framebuffer, newFb); } + newFbread = newFb; } else { /* Binding the window system framebuffer (which was originally set * with MakeCurrent). */ newFb = ctx->WinSysDrawBuffer; + newFbread = ctx->WinSysReadBuffer; } ASSERT(newFb); @@ -1000,7 +1002,7 @@ _mesa_BindFramebufferEXT(GLenum target, GLuint framebuffer) */ if (bindReadBuf) { - _mesa_reference_framebuffer(&ctx->ReadBuffer, newFb); + _mesa_reference_framebuffer(&ctx->ReadBuffer, newFbread); } if (bindDrawBuf) { @@ -1015,7 +1017,7 @@ _mesa_BindFramebufferEXT(GLenum target, GLuint framebuffer) } if (ctx->Driver.BindFramebuffer) { - ctx->Driver.BindFramebuffer(ctx, target, newFb); + ctx->Driver.BindFramebuffer(ctx, target, newFb, newFbread); } } -- cgit v1.2.3 From a8bf93e00e28031d7e2444f79df2686f262b4420 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 17 Jul 2007 19:22:23 +0200 Subject: fix up mesa's probably bogus framebuffer updates with different read/write fbs --- src/mesa/main/framebuffer.c | 40 ++++++++++++++++++++++++++-------------- 1 file changed, 26 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c index a95e44c513..f8716f30d3 100644 --- a/src/mesa/main/framebuffer.c +++ b/src/mesa/main/framebuffer.c @@ -652,6 +652,27 @@ update_color_read_buffer(GLcontext *ctx, struct gl_framebuffer *fb) } +static void +update_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb) +{ + /* Completeness only matters for user-created framebuffers */ + if (fb->Name != 0) { + /* XXX: EXT_framebuffer_blit: + framebuffer must still be complete wrt read/draw? */ + _mesa_test_framebuffer_completeness(ctx, fb); + _mesa_update_framebuffer_visual(fb); + } + + /* update_color_draw/read_buffers not needed for + read/draw only fb, but shouldn't hurt ??? */ + update_color_draw_buffers(ctx, fb); + update_color_read_buffer(ctx, fb); + _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH); + _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL); + + compute_depth_max(fb); +} + /** * Update state related to the current draw/read framebuffers. * Specifically, update these framebuffer fields: @@ -668,22 +689,13 @@ update_color_read_buffer(GLcontext *ctx, struct gl_framebuffer *fb) void _mesa_update_framebuffer(GLcontext *ctx) { - /* XXX might not be quite correct for different draw/read buffers ? */ struct gl_framebuffer *fb = ctx->DrawBuffer; -// struct gl_framebuffer *fbread = ctx->ReadBuffer; - - /* Completeness only matters for user-created framebuffers */ - if (fb->Name != 0) { - _mesa_test_framebuffer_completeness(ctx, fb); - _mesa_update_framebuffer_visual(fb); - } + struct gl_framebuffer *fbread = ctx->ReadBuffer; - update_color_draw_buffers(ctx, fb); - update_color_read_buffer(ctx, fb); - _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH); - _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL); - - compute_depth_max(fb); + update_framebuffer(ctx, fb); + if (fbread != fb && fbread != NULL /* can happen at make_current - + core/driver circular dependencies, should be fixed up */) + update_framebuffer(ctx, fbread); } -- cgit v1.2.3 From 881d71495b893516d685e3c68b60c0aa2742d2a7 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 17 Jul 2007 19:37:05 +0200 Subject: get rid of more pageflip/rotation code. --- src/mesa/drivers/dri/i915tex/intel_buffers.c | 48 +++------------------------- src/mesa/drivers/dri/i915tex/intel_context.c | 7 +--- src/mesa/drivers/dri/i915tex/intel_context.h | 15 --------- src/mesa/drivers/dri/i915tex/intel_screen.h | 2 -- 4 files changed, 6 insertions(+), 66 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index d06fc0eaea..5ec8ad0835 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -650,7 +650,7 @@ intel_wait_flips(struct intel_context *intel, GLuint batch_flags) } } - +#if 0 /* Flip the front & back buffers */ static GLboolean @@ -712,35 +712,8 @@ intelPageFlip(const __DRIdrawablePrivate * dPriv) return GL_TRUE; } +#endif -#if 0 -void -intelSwapBuffers(__DRIdrawablePrivate * dPriv) -{ - if (dPriv->driverPrivate) { - const struct gl_framebuffer *fb - = (struct gl_framebuffer *) dPriv->driverPrivate; - if (fb->Visual.doubleBufferMode) { - GET_CURRENT_CONTEXT(ctx); - if (ctx && ctx->DrawBuffer == fb) { - _mesa_notifySwapBuffers(ctx); /* flush pending rendering */ - } - if (intel->doPageFlip) { - intelPageFlip(dPriv); - } - else { - intelCopyBuffer(dPriv); - } - } - } - else { - _mesa_problem(NULL, - "dPriv has no gl_framebuffer pointer in intelSwapBuffers"); - } -} -#else -/* Trunk version: - */ static GLboolean intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) @@ -755,8 +728,7 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) GLboolean ret; if ((intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || - intelScreen->current_rotation != 0 || - intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) + intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) return GL_FALSE; swap.seqtype = DRM_VBLANK_ABSOLUTE; @@ -831,26 +803,17 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) intel = intel_context(ctx); if (ctx->Visual.doubleBufferMode) { - intelScreenPrivate *screen = intel->intelScreen; GLboolean missed_target; struct intel_framebuffer *intel_fb = dPriv->driverPrivate; int64_t ust; _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - if (screen->current_rotation != 0 || - !intelScheduleSwap(dPriv, &missed_target)) { + if (!intelScheduleSwap(dPriv, &missed_target)) { driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, &missed_target); - if (screen->current_rotation != 0 || !intelPageFlip(dPriv)) { - intelCopyBuffer(dPriv, NULL); - } -#if 0 - if (screen->current_rotation != 0) { - intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT); - } -#endif + intelCopyBuffer(dPriv, NULL); } intel_fb->swap_count++; @@ -868,7 +831,6 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); } } -#endif void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index ea2b7e5277..527b33f4af 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -390,10 +390,6 @@ intelInitContext(struct intel_context *intel, intel->driScreen = sPriv; intel->sarea = saPriv; - intel->width = intelScreen->width; - intel->height = intelScreen->height; - intel->current_rotation = intelScreen->current_rotation; - if (!lockMutexInit) { lockMutexInit = GL_TRUE; _glthread_INIT_MUTEX(lockMutex); @@ -658,8 +654,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); if (sarea->width != intelScreen->width || - sarea->height != intelScreen->height || - sarea->rotation != intelScreen->current_rotation) { + sarea->height != intelScreen->height) { intelUpdateScreenRotation(sPriv, sarea); } diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index e2008c4e18..5c8e277799 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -234,13 +234,6 @@ struct intel_context GLuint vertex_size; GLubyte *verts; /* points to tnl->clipspace.vertex_buf */ -#if 0 - struct intel_region *front_region; /* XXX FBO: obsolete */ - struct intel_region *rotated_region; /* XXX FBO: obsolete */ - struct intel_region *back_region; /* XXX FBO: obsolete */ - struct intel_region *draw_region; /* XXX FBO: rename to color_region */ - struct intel_region *depth_region; /**< currently bound depth/Z region */ -#endif /* Fallback rasterization functions */ @@ -276,14 +269,6 @@ struct intel_context * Configuration cache */ driOptionCache optionCache; - - /* Rotation. Need to match that of the - * current screen. - */ - - int width; - int height; - int current_rotation; }; /* These are functions now: diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.h b/src/mesa/drivers/dri/i915tex/intel_screen.h index 89671e12c9..9a5e3ac76e 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.h +++ b/src/mesa/drivers/dri/i915tex/intel_screen.h @@ -74,8 +74,6 @@ typedef struct struct matrix23 rotMatrix; - int current_rotation; /* 0, 90, 180 or 270 */ - int rotatedWidth, rotatedHeight; /** * Configuration cache with default values for all contexts -- cgit v1.2.3 From 7664105fc2a27ae0862edff66a0a5a33e4111fe1 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 17 Jul 2007 19:49:25 +0200 Subject: remove some more old stuff --- src/mesa/drivers/dri/i915tex/intel_screen.c | 2 -- src/mesa/drivers/dri/i915tex/intel_screen.h | 4 ---- 2 files changed, 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c index 4cabc22225..0b9ba7eda7 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ b/src/mesa/drivers/dri/i915tex/intel_screen.c @@ -310,10 +310,8 @@ intelInitDriver(__DRIscreenPrivate * sPriv) switch (gDRIPriv->bitsPerPixel) { case 16: - intelScreen->fbFormat = DV_PF_565; break; case 32: - intelScreen->fbFormat = DV_PF_8888; break; default: exit(1); diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.h b/src/mesa/drivers/dri/i915tex/intel_screen.h index 9a5e3ac76e..8e4e0b9c12 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.h +++ b/src/mesa/drivers/dri/i915tex/intel_screen.h @@ -59,10 +59,6 @@ typedef struct int mem; /* unused */ int cpp; /* for front and back buffers */ -/* int bitsPerPixel; */ - int fbFormat; /* XXX FBO: this is obsolete - remove after i830 updates */ - - int logTextureGranularity; __DRIscreenPrivate *driScrnPriv; unsigned int sarea_priv_offset; -- cgit v1.2.3 From c5a8bc712542d24a035aa8fe58cf3c28c7c6fbd1 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 17 Jul 2007 20:06:44 +0200 Subject: get rid of checks for old ddx --- src/mesa/drivers/dri/i915tex/intel_buffers.c | 5 ++--- src/mesa/drivers/dri/i915tex/intel_screen.c | 2 +- 2 files changed, 3 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 5ec8ad0835..bcd99106b9 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -157,6 +157,7 @@ intelWindowMoved(struct intel_context *intel) if (!intel->ctx.DrawBuffer) { /* when would this happen? -BP */ + assert(0); intel->numClipRects = 0; } @@ -164,7 +165,7 @@ intelWindowMoved(struct intel_context *intel) driUpdateFramebufferSize(ctx, dPriv); intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ - if (intel->intelScreen->driScrnPriv->ddxMinor >= 7) { + { drmI830Sarea *sarea = intel->sarea; drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; @@ -276,8 +277,6 @@ intelWindowMoved(struct intel_context *intel) intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited; } } - } else { - intel_fb->vblank_flags &= ~VBLANK_FLAG_SECONDARY; } /* Update hardware scissor */ diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c index 0b9ba7eda7..5ad0932e2e 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ b/src/mesa/drivers/dri/i915tex/intel_screen.c @@ -692,7 +692,7 @@ __driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, __GLcontextModes ** driver_modes) { __DRIscreenPrivate *psp; - static const __DRIversion ddx_expected = { 1, 5, 0 }; + static const __DRIversion ddx_expected = { 1, 7, 0 }; static const __DRIversion dri_expected = { 4, 0, 0 }; static const __DRIversion drm_expected = { 1, 7, 0 }; -- cgit v1.2.3 From ccb713cdb8831617880ac71ac357361b586d05c6 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 18 Jul 2007 15:18:43 +0200 Subject: swapbuffers with non-current contexts, cleanups manywin got broken (when intelPageflip got no longer called). Make sure that intelFlush is getting called when a context gets unbound, to handle later swapbuffer calls on that context's drawable better. Related, fix non-current cliprects getting used on unbound drawables. --- src/mesa/drivers/dri/i915tex/intel_blit.c | 6 ++++++ src/mesa/drivers/dri/i915tex/intel_buffers.c | 13 ++----------- src/mesa/drivers/dri/i915tex/intel_context.c | 13 +++++++++---- src/mesa/drivers/dri/i915tex/intel_context.h | 2 -- 4 files changed, 17 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c index 3ecd2e8e48..28441f4b83 100644 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ b/src/mesa/drivers/dri/i915tex/intel_blit.c @@ -78,6 +78,12 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, * should work regardless. */ LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + current context (which is what intelScreenContext should return) might + not get a contended lock and thus cliprects not updated (tests/manywin) */ + if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + if (dPriv && dPriv->numClipRects) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index bcd99106b9..0ecf3aa1d4 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -356,7 +356,8 @@ intelClearWithTris(struct intel_context *intel, GLbitfield mask) /* XXX: Using INTEL_BATCH_NO_CLIPRECTS here is dangerous as the * drawing origin may not be correctly emitted. */ - intel_meta_draw_quad(intel, clear.x1, clear.x2, clear.y1, clear.y2, intel->ctx.Depth.Clear, clearColor, 0, 0, 0, 0); /* texcoords */ + intel_meta_draw_quad(intel, clear.x1, clear.x2, clear.y1, clear.y2, + intel->ctx.Depth.Clear, clearColor, 0, 0, 0, 0); /* texcoords */ mask &= ~(BUFFER_BIT_BACK_LEFT | BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH); @@ -991,16 +992,6 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) /** ** Release old regions, reference new regions **/ -#if 0 /* XXX FBO: this seems to be redundant with i915_state_draw_region() */ - if (intel->draw_region != colorRegion) { - intel_region_release(&intel->draw_region); - intel_region_reference(&intel->draw_region, colorRegion); - } - if (intel->intelScreen->depth_region != depthRegion) { - intel_region_release(&intel->intelScreen->depth_region); - intel_region_reference(&intel->intelScreen->depth_region, depthRegion); - } -#endif intel->vtbl.set_draw_region(intel, colorRegion, depthRegion); diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 527b33f4af..5acfd923b9 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -569,6 +569,9 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv) { + struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; + FLUSH_VERTICES((&intel->ctx), 0); + intelFlush(&intel->ctx); return GL_TRUE; } @@ -616,14 +619,16 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, (*dri_interface->getUST) (&intel_fb->swap_ust); driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); + } + } + + if ((intel->driDrawable != driDrawPriv) || + (intel->lastStamp != driDrawPriv->lastStamp)) { intel->driDrawable = driDrawPriv; intelWindowMoved(intel); intel->lastStamp = driDrawPriv->lastStamp; - } - else if (intel->lastStamp != driDrawPriv->lastStamp) { - intel_draw_buffer(&intel->ctx, &intel_fb->Base); - } } + } else { _mesa_make_current(NULL, NULL, NULL); diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index 5c8e277799..628a15f8de 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -247,8 +247,6 @@ struct intel_context drm_clip_rect_t *pClipRects; drm_clip_rect_t fboRect; /**< cliprect for rendering */ - int perf_boxes; - GLuint do_usleeps; int do_irqs; GLuint irqsEmitted; -- cgit v1.2.3 From 3f318c2eb98cf7216ae7033806fa764ac40192bf Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 18 Jul 2007 18:07:12 +0200 Subject: more fixes for mesa's fbo handling (fixes tests/fbotest1/2) --- src/mesa/main/buffers.c | 5 ++--- src/mesa/main/context.c | 5 +++++ src/mesa/main/fbobject.c | 7 +++++++ 3 files changed, 14 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index c280f89e1d..68a0575d93 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -557,9 +557,8 @@ _mesa_ReadBuffer(GLenum buffer) } } - if (fb->Name == 0) { - ctx->Pixel.ReadBuffer = buffer; - } + ctx->Pixel.ReadBuffer = buffer; + fb->ColorReadBuffer = buffer; fb->_ColorReadBufferIndex = srcBuffer; diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 183a552d40..5239e73956 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1499,6 +1499,11 @@ _mesa_make_current( GLcontext *newCtx, GLframebuffer *drawBuffer, if the DRIdrawable changes, and someone may rely on them. */ /* What a mess!?! */ + /* XXX this is still not quite correct. Imagine a user-created fbo + bound on a context. Now rebind with a completely new drawable. + Upon rebinding to the window-framebuffer, we have no idea what + the read and write buffers should be (front, back, ...) - that + information was only available in the previously used drawable... */ int i; GLenum buffers[MAX_DRAW_BUFFERS]; for(i = 0; i < newCtx->Const.MaxDrawBuffers; i++) { diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 5345310ca1..f300e481ce 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -1003,13 +1003,20 @@ _mesa_BindFramebufferEXT(GLenum target, GLuint framebuffer) if (bindReadBuf) { _mesa_reference_framebuffer(&ctx->ReadBuffer, newFbread); + /* set context value */ + ctx->Pixel.ReadBuffer = newFbread->ColorReadBuffer; } if (bindDrawBuf) { + GLuint i; /* check if old FB had any texture attachments */ check_end_texture_render(ctx, ctx->DrawBuffer); /* check if time to delete this framebuffer */ _mesa_reference_framebuffer(&ctx->DrawBuffer, newFb); + /* set context value */ + for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { + ctx->Color.DrawBuffer[i] = newFb->ColorDrawBuffer[i]; + } if (newFb->Name != 0) { /* check if newly bound framebuffer has any texture attachments */ check_begin_texture_render(ctx, newFb); -- cgit v1.2.3 From 131a03631208afb1fce90292eb27115511f58257 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 18 Jul 2007 20:17:14 +0200 Subject: fix mesa's handling of fbo's / window fb (again) Make sure the relevant fields in window fbs get updated at appropriate time (those are NOT the same as fbos!!!), and fix up related code accordingly. This is a bit ugly, but there's a reason the issues section in EXT_fbo is a couple hundred pages long... Hopefully correct now. --- src/mesa/main/attrib.c | 4 +++ src/mesa/main/buffers.c | 72 ++++++++++++++++++++++++++++++--------------- src/mesa/main/buffers.h | 3 ++ src/mesa/main/context.c | 12 ++------ src/mesa/main/fbobject.c | 24 ++++++++++----- src/mesa/main/framebuffer.c | 3 +- 6 files changed, 77 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index 1aa0a02fc7..b422198f92 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -98,9 +98,13 @@ _mesa_PushAttrib(GLbitfield mask) } if (mask & GL_COLOR_BUFFER_BIT) { + GLuint i; struct gl_colorbuffer_attrib *attr; attr = MALLOC_STRUCT( gl_colorbuffer_attrib ); MEMCPY( attr, &ctx->Color, sizeof(struct gl_colorbuffer_attrib) ); + /* push the Draw FBO's DrawBuffer[] state, not ctx->Color.DrawBuffer[] */ + for (i = 0; i < ctx->Const.MaxDrawBuffers; i ++) + attr->DrawBuffer[i] = ctx->DrawBuffer->ColorDrawBuffer[i]; newnode = new_attrib_node( GL_COLOR_BUFFER_BIT ); newnode->data = attr; newnode->next = head; diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index 68a0575d93..7764a5d3b2 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -370,6 +370,14 @@ _mesa_DrawBuffer(GLenum buffer) /* if we get here, there's no error so set new state */ _mesa_drawbuffers(ctx, 1, &buffer, &destMask); + + /* + * Call device driver function. + */ + if (ctx->Driver.DrawBuffers) + ctx->Driver.DrawBuffers(ctx, 1, &buffer); + else if (ctx->Driver.DrawBuffer) + ctx->Driver.DrawBuffer(ctx, buffer); } @@ -435,6 +443,14 @@ _mesa_DrawBuffersARB(GLsizei n, const GLenum *buffers) /* OK, if we get here, there were no errors so set the new state */ _mesa_drawbuffers(ctx, n, buffers, destMask); + + /* + * Call device driver function. + */ + if (ctx->Driver.DrawBuffers) + ctx->Driver.DrawBuffers(ctx, n, buffers); + else if (ctx->Driver.DrawBuffer) + ctx->Driver.DrawBuffer(ctx, buffers[0]); } @@ -463,14 +479,15 @@ set_color_output(GLcontext *ctx, GLuint output, GLenum buffer, /* not really needed, will be set later */ fb->_NumColorDrawBuffers[output] = 0; + if (fb->Name == 0) /* Set traditional state var */ - ctx->Color.DrawBuffer[output] = buffer; + ctx->Color.DrawBuffer[output] = buffer; } /** * Helper routine used by _mesa_DrawBuffer, _mesa_DrawBuffersARB and - * _mesa_PopAttrib to set drawbuffer state. + * other places (window fbo fixup) to set fbo (and the old ctx) fields. * All error checking will have been done prior to calling this function * so nothing should go wrong at this point. * \param ctx current context @@ -479,6 +496,7 @@ set_color_output(GLcontext *ctx, GLuint output, GLenum buffer, * \param destMask array[n] of BUFFER_* bitmasks which correspond to the * colorbuffer names. (i.e. GL_FRONT_AND_BACK => * BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT). + * \param callDriver call driver or not (bad idea sometimes this is called) */ void _mesa_drawbuffers(GLcontext *ctx, GLuint n, const GLenum *buffers, @@ -509,30 +527,15 @@ _mesa_drawbuffers(GLcontext *ctx, GLuint n, const GLenum *buffers, } ctx->NewState |= _NEW_COLOR; - - /* - * Call device driver function. - */ - if (ctx->Driver.DrawBuffers) - ctx->Driver.DrawBuffers(ctx, n, buffers); - else if (ctx->Driver.DrawBuffer) - ctx->Driver.DrawBuffer(ctx, buffers[0]); } - -/** - * Called by glReadBuffer to set the source renderbuffer for reading pixels. - * \param mode color buffer such as GL_FRONT, GL_BACK, etc. - */ -void GLAPIENTRY -_mesa_ReadBuffer(GLenum buffer) +GLboolean +_mesa_readbuffer_update_fields(GLcontext *ctx, GLenum buffer) { struct gl_framebuffer *fb; GLbitfield supportedMask; GLint srcBuffer; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); fb = ctx->ReadBuffer; @@ -548,20 +551,43 @@ _mesa_ReadBuffer(GLenum buffer) srcBuffer = read_buffer_enum_to_index(buffer); if (srcBuffer == -1) { _mesa_error(ctx, GL_INVALID_ENUM, "glReadBuffer(buffer=0x%x)", buffer); - return; + return GL_FALSE; } supportedMask = supported_buffer_bitmask(ctx, fb); if (((1 << srcBuffer) & supportedMask) == 0) { _mesa_error(ctx, GL_INVALID_OPERATION, "glReadBuffer(buffer=0x%x)", buffer); - return; + return GL_FALSE; } } - ctx->Pixel.ReadBuffer = buffer; - + if (fb->Name == 0) { + ctx->Pixel.ReadBuffer = buffer; + } fb->ColorReadBuffer = buffer; fb->_ColorReadBufferIndex = srcBuffer; + return GL_TRUE; +} + + + +/** + * Called by glReadBuffer to set the source renderbuffer for reading pixels. + * \param mode color buffer such as GL_FRONT, GL_BACK, etc. + * \param callDriver call driver or not (bad idea sometimes this is called) + */ +void GLAPIENTRY +_mesa_ReadBuffer(GLenum buffer) +{ + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + + if (MESA_VERBOSE & VERBOSE_API) + _mesa_debug(ctx, "glReadBuffer %s\n", _mesa_lookup_enum_by_nr(buffer)); + + if (!_mesa_readbuffer_update_fields(ctx, buffer)) + return; + ctx->NewState |= _NEW_PIXEL; /* diff --git a/src/mesa/main/buffers.h b/src/mesa/main/buffers.h index fcc2152342..208e7af2b9 100644 --- a/src/mesa/main/buffers.h +++ b/src/mesa/main/buffers.h @@ -56,6 +56,9 @@ extern void _mesa_drawbuffers(GLcontext *ctx, GLuint n, const GLenum *buffers, const GLbitfield *destMask); +extern GLboolean +_mesa_readbuffer_update_fields(GLcontext *ctx, GLenum buffer); + extern void GLAPIENTRY _mesa_ReadBuffer( GLenum mode ); diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 5239e73956..92d8ceda29 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1496,14 +1496,8 @@ _mesa_make_current( GLcontext *newCtx, GLframebuffer *drawBuffer, if (!newCtx->DrawBuffer || newCtx->DrawBuffer->Name == 0) { _mesa_reference_framebuffer(&newCtx->DrawBuffer, drawBuffer); /* fix up the fb fields - these will end up wrong otherwise - if the DRIdrawable changes, and someone may rely on them. - */ - /* What a mess!?! */ - /* XXX this is still not quite correct. Imagine a user-created fbo - bound on a context. Now rebind with a completely new drawable. - Upon rebinding to the window-framebuffer, we have no idea what - the read and write buffers should be (front, back, ...) - that - information was only available in the previously used drawable... */ + if the DRIdrawable changes, and everything relies on them. + This is a bit messy (same as needed in _mesa_BindFramebufferEXT) */ int i; GLenum buffers[MAX_DRAW_BUFFERS]; for(i = 0; i < newCtx->Const.MaxDrawBuffers; i++) { @@ -1513,7 +1507,7 @@ _mesa_make_current( GLcontext *newCtx, GLframebuffer *drawBuffer, } if (!newCtx->ReadBuffer || newCtx->ReadBuffer->Name == 0) { _mesa_reference_framebuffer(&newCtx->ReadBuffer, readBuffer); - _mesa_ReadBuffer(newCtx->Pixel.ReadBuffer); + _mesa_readbuffer_update_fields(newCtx, newCtx->Pixel.ReadBuffer); } newCtx->NewState |= _NEW_BUFFERS; diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index f300e481ce..6f7effcce7 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -29,6 +29,7 @@ */ +#include "buffers.h" #include "context.h" #include "fbobject.h" #include "framebuffer.h" @@ -1001,23 +1002,32 @@ _mesa_BindFramebufferEXT(GLenum target, GLuint framebuffer) * XXX check if re-binding same buffer and skip some of this code. */ + /* for window-framebuffers, re-initialize the fbo values, as they + could be wrong (makecurrent with a new drawable while still a fbo + was bound will lead to default init fbo values). + note that therefore the context ReadBuffer/DrawBuffer values are not + valid while fbo's are bound!!! */ if (bindReadBuf) { _mesa_reference_framebuffer(&ctx->ReadBuffer, newFbread); - /* set context value */ - ctx->Pixel.ReadBuffer = newFbread->ColorReadBuffer; + if (!newFbread->Name) { + _mesa_readbuffer_update_fields(ctx, ctx->Pixel.ReadBuffer); + } } if (bindDrawBuf) { - GLuint i; /* check if old FB had any texture attachments */ check_end_texture_render(ctx, ctx->DrawBuffer); /* check if time to delete this framebuffer */ _mesa_reference_framebuffer(&ctx->DrawBuffer, newFb); - /* set context value */ - for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { - ctx->Color.DrawBuffer[i] = newFb->ColorDrawBuffer[i]; + if (!newFb->Name) { + GLuint i; + GLenum buffers[MAX_DRAW_BUFFERS]; + for(i = 0; i < ctx->Const.MaxDrawBuffers; i++) { + buffers[i] = ctx->Color.DrawBuffer[i]; + } + _mesa_drawbuffers(ctx, ctx->Const.MaxDrawBuffers, buffers, NULL); } - if (newFb->Name != 0) { + else { /* check if newly bound framebuffer has any texture attachments */ check_begin_texture_render(ctx, newFb); } diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c index f8716f30d3..f737c74f35 100644 --- a/src/mesa/main/framebuffer.c +++ b/src/mesa/main/framebuffer.c @@ -693,8 +693,7 @@ _mesa_update_framebuffer(GLcontext *ctx) struct gl_framebuffer *fbread = ctx->ReadBuffer; update_framebuffer(ctx, fb); - if (fbread != fb && fbread != NULL /* can happen at make_current - - core/driver circular dependencies, should be fixed up */) + if (fbread != fb) update_framebuffer(ctx, fbread); } -- cgit v1.2.3 From da21eff86be088852457cdcb725aed5337c4519d Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Thu, 19 Jul 2007 16:41:14 +0200 Subject: fix drawables not getting freed if context is made current with new drawables --- src/mesa/drivers/dri/common/dri_util.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c index dd52f7e915..d785382f39 100644 --- a/src/mesa/drivers/dri/common/dri_util.c +++ b/src/mesa/drivers/dri/common/dri_util.c @@ -120,7 +120,7 @@ static __DRIdrawable *__driFindDrawable(void *drawHash, __DRIid draw) * Find drawables in the local hash that have been destroyed on the * server. * - * \param drawHash Hash-table containing all know drawables. + * \param drawHash Hash-table containing all known drawables. */ static void __driGarbageCollectDrawables(void *drawHash) { @@ -235,6 +235,12 @@ static GLboolean driUnbindContext(__DRInativeDisplay *dpy, int scrn, prp->refcount--; } + /* destroy the drawables if they no longer exist on the server */ + if ((pdp->refcount == 0) || (prp->refcount == 0)) { + /* probably shouldn't need the collector here, + as we know the affected drawables (or could there be others?) */ + __driGarbageCollectDrawables(pdp->driScreenPriv->drawHash); + } /* XXX this is disabled so that if we call SwapBuffers on an unbound * window we can determine the last context bound to the window and -- cgit v1.2.3 From 79ddb7f9a27421089efea090de6c41d2d630bb8f Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Thu, 19 Jul 2007 17:42:14 +0200 Subject: more cleanups (looks pretty reasonable now) remove some already ifdefed out, no longer functional and used code. Don't do our own scissor clipping in the pixeldraw/copy paths, as meas already does that for us... --- src/mesa/drivers/dri/i915tex/intel_buffers.c | 144 ------------------------ src/mesa/drivers/dri/i915tex/intel_buffers.h | 6 - src/mesa/drivers/dri/i915tex/intel_context.c | 12 +- src/mesa/drivers/dri/i915tex/intel_fbo.c | 120 -------------------- src/mesa/drivers/dri/i915tex/intel_pixel_copy.c | 77 +++++-------- src/mesa/drivers/dri/i915tex/intel_pixel_draw.c | 44 +++----- 6 files changed, 45 insertions(+), 358 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 0ecf3aa1d4..ee3266e09b 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -396,150 +396,6 @@ intelClearWithTris(struct intel_context *intel, GLbitfield mask) } -#if 0 -/* Xorg seems to handle that nowadays, and this code here no longer works with priv - buffers */ -/** - * Copy the window contents named by dPriv to the rotated (or reflected) - * color buffer. - * srcBuf is BUFFER_BIT_FRONT_LEFT or BUFFER_BIT_BACK_LEFT to indicate the source. - */ -void -intelRotateWindow(struct intel_context *intel, - __DRIdrawablePrivate * dPriv, GLuint srcBuf) -{ - - intelScreenPrivate *screen = intel->intelScreen; - drm_clip_rect_t fullRect; - struct intel_framebuffer *intel_fb; - struct intel_region *src; - const drm_clip_rect_t *clipRects; - int numClipRects; - int i; - GLenum format, type; - - int origNumClipRects; - drm_clip_rect_t *origRects; - - /* - * set up hardware state - */ - intelFlush(&intel->ctx); - - LOCK_HARDWARE(intel); - - if (!intel->numClipRects) { - UNLOCK_HARDWARE(intel); - return; - } - - intel->vtbl.install_meta_state(intel); - - intel->vtbl.meta_no_depth_write(intel); - intel->vtbl.meta_no_stencil_write(intel); - intel->vtbl.meta_color_mask(intel, GL_FALSE); - - - /* save current drawing origin and cliprects (restored at end) */ - origNumClipRects = intel->numClipRects; - origRects = intel->pClipRects; - - /* - * set drawing origin, cliprects for full-screen access to rotated screen - */ - fullRect.x1 = 0; - fullRect.y1 = 0; - fullRect.x2 = screen->rotatedWidth; - fullRect.y2 = screen->rotatedHeight; - intel->numClipRects = 1; - intel->pClipRects = &fullRect; - - intel->vtbl.meta_draw_region(intel, screen->rotated_region, NULL); /* ? */ - - intel_fb = dPriv->driverPrivate; - - if ((srcBuf == BUFFER_BIT_BACK_LEFT && !intel_fb->pf_active)) { - src = intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); - clipRects = dPriv->pBackClipRects; - numClipRects = dPriv->numBackClipRects; - } - else { - src = intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT); - clipRects = dPriv->pClipRects; - numClipRects = dPriv->numClipRects; - } - - if (src->cpp == 4) { - format = GL_BGRA; - type = GL_UNSIGNED_BYTE; - } - else { - format = GL_BGR; - type = GL_UNSIGNED_SHORT_5_6_5_REV; - } - - /* set the whole screen up as a texture to avoid alignment issues */ - intel->vtbl.meta_tex_rect_source(intel, - src->buffer, - screen->width, - screen->height, src->pitch, format, type); - - intel->vtbl.meta_texture_blend_replace(intel); - - /* - * loop over the source window's cliprects - */ - for (i = 0; i < numClipRects; i++) { - int srcX0 = clipRects[i].x1; - int srcY0 = clipRects[i].y1; - int srcX1 = clipRects[i].x2; - int srcY1 = clipRects[i].y2; - GLfloat verts[4][2], tex[4][2]; - int j; - - /* build vertices for four corners of clip rect */ - verts[0][0] = srcX0; - verts[0][1] = srcY0; - verts[1][0] = srcX1; - verts[1][1] = srcY0; - verts[2][0] = srcX1; - verts[2][1] = srcY1; - verts[3][0] = srcX0; - verts[3][1] = srcY1; - - /* .. and texcoords */ - tex[0][0] = srcX0; - tex[0][1] = srcY0; - tex[1][0] = srcX1; - tex[1][1] = srcY0; - tex[2][0] = srcX1; - tex[2][1] = srcY1; - tex[3][0] = srcX0; - tex[3][1] = srcY1; - - /* transform coords to rotated screen coords */ - - for (j = 0; j < 4; j++) { - matrix23TransformCoordf(&screen->rotMatrix, - &verts[j][0], &verts[j][1]); - } - - /* draw polygon to map source image to dest region */ - intel_meta_draw_poly(intel, 4, verts, 0, 0, tex); - - } /* cliprect loop */ - - intel->vtbl.leave_meta_state(intel); - intel_batchbuffer_flush(intel->batch); - - /* restore original drawing origin and cliprects */ - intel->numClipRects = origNumClipRects; - intel->pClipRects = origRects; - - UNLOCK_HARDWARE(intel); -} -#endif - /** * Called by ctx->Driver.Clear. */ diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.h b/src/mesa/drivers/dri/i915tex/intel_buffers.h index 3c7a9cf139..13d1a15ffb 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.h +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.h @@ -52,10 +52,4 @@ extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); extern void intelInitBufferFuncs(struct dd_function_table *functions); -#if 0 -extern void -intelRotateWindow(struct intel_context *intel, - __DRIdrawablePrivate * dPriv, GLuint srcBuf); -#endif - #endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 5acfd923b9..769287ba6a 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -296,17 +296,7 @@ intelCheckFrontUpdate(GLcontext * ctx) if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT) { __DRIdrawablePrivate *dPriv = intel->driDrawable; -#if 0 - intelScreenPrivate *screen = intel->intelScreen; - if (screen->current_rotation != 0) { - intelRotateWindow(intel, dPriv, BUFFER_BIT_FRONT_LEFT); - } - else { - intelCopyBuffer(dPriv, NULL); - } -#else intelCopyBuffer(dPriv, NULL); -#endif } } @@ -570,6 +560,8 @@ GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv) { struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; + /* XXX UnbindContext is called AFTER the new context is made current. + Hopefully shouldn't be a problem ? */ FLUSH_VERTICES((&intel->ctx), 0); intelFlush(&intel->ctx); return GL_TRUE; diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c index b0c243ebe0..399a280c47 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c @@ -299,25 +299,6 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } -#if 0 -/** - * Called for each hardware renderbuffer when a _window_ is resized. - * Just update fields. - * Not used for user-created renderbuffers! - */ -static GLboolean -intel_alloc_window_storage(GLcontext * ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, GLuint width, GLuint height) -{ - ASSERT(rb->Name == 0); - rb->Width = width; - rb->Height = height; - rb->_ActualFormat = internalFormat; - - return GL_TRUE; -} -#endif - static void intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, GLuint width, GLuint height) @@ -353,107 +334,6 @@ intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } -#if 0 -/** - * Create a new intel_renderbuffer which corresponds to an on-screen window, - * not a user-created renderbuffer. - * \param width the screen width - * \param height the screen height - */ -struct intel_renderbuffer * -intel_create_renderbuffer(GLenum intFormat, GLsizei width, GLsizei height, - int offset, int pitch, int cpp, void *map) -{ - GET_CURRENT_CONTEXT(ctx); - - struct intel_renderbuffer *irb; - const GLuint name = 0; - - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, name); - irb->Base.ClassID = INTEL_RB_CLASS; - - switch (intFormat) { - case GL_RGB5: - irb->Base._ActualFormat = GL_RGB5; - irb->Base._BaseFormat = GL_RGBA; - irb->Base.RedBits = 5; - irb->Base.GreenBits = 6; - irb->Base.BlueBits = 5; - irb->Base.DataType = GL_UNSIGNED_BYTE; - cpp = 2; - break; - case GL_RGBA8: - irb->Base._ActualFormat = GL_RGBA8; - irb->Base._BaseFormat = GL_RGBA; - irb->Base.RedBits = 8; - irb->Base.GreenBits = 8; - irb->Base.BlueBits = 8; - irb->Base.AlphaBits = 8; - irb->Base.DataType = GL_UNSIGNED_BYTE; - cpp = 4; - break; - case GL_STENCIL_INDEX8_EXT: - irb->Base._ActualFormat = GL_STENCIL_INDEX8_EXT; - irb->Base._BaseFormat = GL_STENCIL_INDEX; - irb->Base.StencilBits = 8; - irb->Base.DataType = GL_UNSIGNED_BYTE; - cpp = 1; - break; - case GL_DEPTH_COMPONENT16: - irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - irb->Base.DepthBits = 16; - irb->Base.DataType = GL_UNSIGNED_SHORT; - cpp = 2; - break; - case GL_DEPTH_COMPONENT24: - irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - irb->Base.DepthBits = 24; - irb->Base.DataType = GL_UNSIGNED_INT; - cpp = 4; - break; - case GL_DEPTH24_STENCIL8_EXT: - irb->Base._ActualFormat = GL_DEPTH24_STENCIL8_EXT; - irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; - irb->Base.DepthBits = 24; - irb->Base.StencilBits = 8; - irb->Base.DataType = GL_UNSIGNED_INT_24_8_EXT; - cpp = 4; - break; - default: - _mesa_problem(NULL, - "Unexpected intFormat in intel_create_renderbuffer"); - return NULL; - } - - irb->Base.InternalFormat = intFormat; - - /* intel-specific methods */ - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_alloc_window_storage; - irb->Base.GetPointer = intel_get_pointer; - /* This sets the Get/PutRow/Value functions */ - intel_set_span_functions(&irb->Base); - - irb->pfMap = map; - irb->pfPitch = pitch / cpp; /* in pixels */ - -#if 00 - irb->region = intel_region_create_static(intel, - DRM_MM_TT, - offset, map, cpp, width, height); -#endif - - return irb; -} -#endif struct intel_renderbuffer * intel_new_renderbuffer_fb(GLuint intFormat) diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c index 58c415215e..2ecacc3c09 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c @@ -234,6 +234,8 @@ do_blit_copypixels(GLcontext * ctx, GLint dstx, GLint dsty, GLenum type) { struct intel_context *intel = intel_context(ctx); + struct gl_framebuffer *fb = ctx->DrawBuffer; + struct gl_framebuffer *fbread = ctx->ReadBuffer; struct intel_renderbuffer *irbread; struct intel_renderbuffer *irbdraw; struct intel_region *dst; @@ -249,16 +251,16 @@ do_blit_copypixels(GLcontext * ctx, intelFlush(&intel->ctx); if (type == GL_COLOR) { - irbread = intel_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); - irbdraw = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); + irbread = intel_renderbuffer(fbread->_ColorReadBuffer); + irbdraw = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); if (!irbread || !irbread->region || !irbdraw || !irbdraw->region) return GL_FALSE; } else if (type == GL_DEPTH) { /* Don't think this is really possible execpt at 16bpp, when we have no stencil. */ - irbread = intel_renderbuffer(ctx->ReadBuffer->_DepthBuffer->Wrapped); - irbdraw = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + irbread = intel_renderbuffer(fbread->_DepthBuffer->Wrapped); + irbdraw = intel_renderbuffer(fb->_DepthBuffer->Wrapped); if (!irbread || !irbread->region || !irbdraw || !irbdraw->region || !(irbread->region->cpp == 2)) return GL_FALSE; @@ -266,8 +268,8 @@ do_blit_copypixels(GLcontext * ctx, else if (type == GL_DEPTH_STENCIL_EXT) { /* Does it matter whether it is stencil/depth or depth/stencil? */ - irbread = intel_renderbuffer(ctx->ReadBuffer->_DepthBuffer->Wrapped); - irbdraw = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + irbread = intel_renderbuffer(fbread->_DepthBuffer->Wrapped); + irbdraw = intel_renderbuffer(fb->_DepthBuffer->Wrapped); if (!irbread || !irbread->region || !irbdraw || !irbdraw->region) return GL_FALSE; } @@ -281,61 +283,38 @@ do_blit_copypixels(GLcontext * ctx, dst = irbdraw->region; { - GLint delta_x = 0; - GLint delta_y = 0; + GLint dx = dstx - srcx; + GLint dy = dsty - srcy; - /* Do scissoring in GL coordinates: + /* Clip against dest, including scissor, in GL coordinates: */ - if (ctx->Scissor.Enabled) - { - GLint x = ctx->Scissor.X; - GLint y = ctx->Scissor.Y; - GLuint w = ctx->Scissor.Width; - GLuint h = ctx->Scissor.Height; - GLint dx = dstx - srcx; - GLint dy = dsty - srcy; - - if (!_mesa_clip_to_region(x, y, x+w-1, y+h-1, &dstx, &dsty, &width, &height)) - goto out; - srcx = dstx - dx; - srcy = dsty - dy; - } + if (!_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, + fb->_Xmax - 1, fb->_Ymax - 1, + &dstx, &dsty, &width, &height)) + goto out; + + srcx = dstx - dx; + srcy = dsty - dy; /* Convert from GL to hardware coordinates: */ - dsty = irbdraw->Base.Height - dsty - height; - srcy = irbread->Base.Height - srcy - height; + dsty = fb->Height - dsty - height; + srcy = fbread->Height - srcy - height; /* Clip against the source region: */ - { - delta_x = srcx - dstx; - delta_y = srcy - dsty; + dx = srcx - dstx; + dy = srcy - dsty; - if (!_mesa_clip_to_region(0, 0, irbread->Base.Width - 1, - irbread->Base.Height - 1, - &srcx, &srcy, &width, &height)) - goto out; + if (!_mesa_clip_to_region(0, 0, irbread->Base.Width - 1, + irbread->Base.Height - 1, + &srcx, &srcy, &width, &height)) + goto out; - dstx = srcx - delta_x; - dsty = srcy - delta_y; - } + dstx = srcx - dx; + dsty = srcy - dy; - /* Clip against the dest region: - */ - { - delta_x = dstx - srcx; - delta_y = dsty - srcy; - - if (!_mesa_clip_to_region(0, 0, irbdraw->Base.Width - 1, - irbdraw->Base.Height - 1, - &dstx, &dsty, &width, &height)) - goto out; - - srcx = dstx - delta_x; - srcy = dsty - delta_y; - } { diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c index 4d6b4af589..4139117206 100644 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c +++ b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c @@ -219,33 +219,33 @@ do_blit_drawpixels(GLcontext * ctx, const GLvoid * pixels) { struct intel_context *intel = intel_context(ctx); + struct gl_framebuffer *fb = ctx->DrawBuffer; struct intel_renderbuffer *irbdraw; struct intel_region *dest; struct intel_buffer_object *src = intel_buffer_object(unpack->BufferObj); GLuint src_offset; GLuint rowLength; - GLuint height_orig = height; struct _DriFenceObject *fence = NULL; if (INTEL_DEBUG & DEBUG_PIXEL) _mesa_printf("%s\n", __FUNCTION__); if (type == GL_COLOR) { - irbdraw = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); + irbdraw = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); if (!irbdraw || !irbdraw->region) return GL_FALSE; } else if (type == GL_DEPTH) { /* Don't think this is really possible execpt at 16bpp, when we have no stencil. */ - irbdraw = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + irbdraw = intel_renderbuffer(fb->_DepthBuffer->Wrapped); if (!irbdraw || !irbdraw->region || !(irbdraw->region->cpp == 2)) return GL_FALSE; } else if (type == GL_DEPTH_STENCIL_EXT) { /* Does it matter whether it is stencil/depth or depth/stencil? */ - irbdraw = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + irbdraw = intel_renderbuffer(fb->_DepthBuffer->Wrapped); if (!irbdraw || !irbdraw->region) return GL_FALSE; } @@ -331,40 +331,26 @@ do_blit_drawpixels(GLcontext * ctx, GLuint srcy = 0; GLint dx = dstx; GLint dy = dsty; + GLuint height_orig = height; - /* Do scissoring in GL coordinates: + /* Do scissoring and clipping in GL coordinates, no need to clip against + * pbo src region (note fbo fields include scissor already): */ - if (ctx->Scissor.Enabled) - { - GLint x = ctx->Scissor.X; - GLint y = ctx->Scissor.Y; - GLuint w = ctx->Scissor.Width; - GLuint h = ctx->Scissor.Height; - height_orig = height; - - - if (!_mesa_clip_to_region(x, y, x+w-1, y+h-1, &dstx, &dsty, &width, &height)) - goto out; - - } + height_orig = height; + if (!_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, + fb->_Xmax - 1, fb->_Ymax - 1, + &dstx, &dsty, &width, &height)) + goto out; - /* no need to clip against pbo src region, but clip against dest */ - { - if (!_mesa_clip_to_region(0, 0, irbdraw->Base.Width - 1, - irbdraw->Base.Height - 1, - &dstx, &dsty, &width, &height)) - goto out; - - srcx = dstx - dx; - srcy = dsty - dy; - } + srcx = dstx - dx; + srcy = dsty - dy; struct _DriBufferObject *src_buffer = intel_bufferobj_buffer(intel, src, INTEL_READ); /* Convert from GL to hardware coordinates: */ - dsty = irbdraw->Base.Height - dsty - height; + dsty = fb->Height - dsty - height; srcy = height_orig - srcy - height; { -- cgit v1.2.3 From e0850e52eed52ff4e2acac6b31a7c456eb9716a0 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Thu, 19 Jul 2007 18:11:48 +0200 Subject: remove some more really old ifdefed out code... --- src/mesa/drivers/dri/i915tex/i915_state.c | 48 ------------------------------- 1 file changed, 48 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/i915_state.c b/src/mesa/drivers/dri/i915tex/i915_state.c index f523206bef..1fede6a8d8 100644 --- a/src/mesa/drivers/dri/i915tex/i915_state.c +++ b/src/mesa/drivers/dri/i915tex/i915_state.c @@ -904,44 +904,9 @@ i915_init_packets(struct i915_context *i915) { I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); -#if 0 - /* color buffer offset/stride */ - i915->state.Buffer[I915_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - /* XXX FBO: remove this? Also get set in i915_set_draw_region() */ - i915->state.Buffer[I915_DESTREG_CBUFADDR1] = (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(screen->front.pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - - i915->state.Buffer[I915_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - /* XXX FBO: remove this? Also get set in i915_set_draw_region() */ - i915->state.Buffer[I915_DESTREG_DBUFADDR1] = (BUF_3D_ID_DEPTH | BUF_3D_PITCH(screen->depth.pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); -#endif i915->state.Buffer[I915_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD; - /* XXX FBO: remove this? Also get set in i915_set_draw_region() */ -#if 0 /* seems we don't need this */ - switch (screen->fbFormat) { - case DV_PF_565: - i915->state.Buffer[I915_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - LOD_PRECLAMP_OGL | - TEX_DEFAULT_COLOR_OGL | - DITHER_FULL_ALWAYS | - screen->fbFormat | - DEPTH_FRMT_16_FIXED); - break; - case DV_PF_8888: - i915->state.Buffer[I915_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - LOD_PRECLAMP_OGL | - TEX_DEFAULT_COLOR_OGL | - screen->fbFormat | - DEPTH_FRMT_24_FIXED_8_OTHER); - break; - } -#endif - /* scissor */ i915->state.Buffer[I915_DESTREG_SENABLE] = @@ -952,19 +917,6 @@ i915_init_packets(struct i915_context *i915) } -#if 0 - { - I915_STATECHANGE(i915, I915_UPLOAD_DEFAULTS); - i915->state.Default[I915_DEFREG_C0] = _3DSTATE_DEFAULT_DIFFUSE; - i915->state.Default[I915_DEFREG_C1] = 0; - i915->state.Default[I915_DEFREG_S0] = _3DSTATE_DEFAULT_SPECULAR; - i915->state.Default[I915_DEFREG_S1] = 0; - i915->state.Default[I915_DEFREG_Z0] = _3DSTATE_DEFAULT_Z; - i915->state.Default[I915_DEFREG_Z1] = 0; - } -#endif - - /* These will be emitted every at the head of every buffer, unless * we get hardware contexts working. */ -- cgit v1.2.3 From 4824c342c864e870251a7d343c95e51274e50d23 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 19 Jul 2007 20:24:55 +0100 Subject: Trigger tgsi compilation for fragment programs. Not sure the generated program looks correct though... --- src/mesa/pipe/p_state.h | 5 ++- src/mesa/pipe/softpipe/sp_state_derived.c | 4 +-- src/mesa/pipe/tgsi/core/tgsi_dump.c | 4 +++ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 2 ++ src/mesa/sources | 1 + src/mesa/state_tracker/st_atom.c | 1 + src/mesa/state_tracker/st_atom.h | 1 + src/mesa/state_tracker/st_atom_fs.c | 27 ++++++++++++-- src/mesa/state_tracker/st_atom_vs.c | 49 +++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_program.c | 59 ++++++++++++++----------------- src/mesa/state_tracker/st_context.h | 1 + src/mesa/state_tracker/st_program.h | 38 +++++++++++++++++++- 12 files changed, 152 insertions(+), 40 deletions(-) create mode 100644 src/mesa/state_tracker/st_atom_vs.c (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index fd5e7ad3af..e3f62a80ad 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -121,8 +121,11 @@ struct pipe_clip_state { GLuint nr; }; + struct pipe_fs_state { - struct gl_fragment_program *fp; + GLuint inputs_read; /* FRAG_ATTRIB_* */ + const struct tgsi_token *tokens; + }; struct pipe_constant_buffer { diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 18dfb50e38..fcdedb54a9 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -28,7 +28,6 @@ #include "main/glheader.h" #include "main/macros.h" #include "main/enums.h" -#include "shader/program.h" #include "vf/vf.h" #include "pipe/draw/draw_context.h" @@ -68,8 +67,7 @@ static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - struct gl_fragment_program *fp = softpipe->fs.fp; - const GLuint inputsRead = fp->Base.InputsRead; + const GLuint inputsRead = softpipe->fs.inputs_read; GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; GLbitfield attr_mask = 0x0; GLuint i; diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index fecb246ab1..0345fd93f7 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -400,12 +400,16 @@ tgsi_dump( GLuint deflt = !(flags & TGSI_DUMP_NO_DEFAULT); { +#if 0 static GLuint counter = 0; char buffer[64]; sprintf( buffer, "sbir-dump-%.4u.txt", counter++ ); dump.file = fopen( buffer, "wt" ); +#else + dump.file = stderr; dump.tabs = 0; +#endif } tgsi_parse_init( &parse, tokens ); diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index 4c1141e579..9256318997 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -5,6 +5,8 @@ extern "C" { #endif // defined __cplusplus +struct tgsi_token; + GLboolean tgsi_mesa_compile_fp_program( const struct gl_fragment_program *program, diff --git a/src/mesa/sources b/src/mesa/sources index a76e41bdda..a589ae4373 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -206,6 +206,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_clip.c \ state_tracker/st_atom_depth.c \ state_tracker/st_atom_fs.c \ + state_tracker/st_atom_vs.c \ state_tracker/st_atom_framebuffer.c \ state_tracker/st_atom_sampler.c \ state_tracker/st_atom_scissor.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index dfebfb4768..85c99bc182 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -46,6 +46,7 @@ static const struct st_tracked_state *atoms[] = &st_update_clear_color, &st_update_depth, &st_update_clip, + &st_update_vs, &st_update_fs, &st_update_setup, &st_update_polygon_stipple, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index a56483ac39..1b70e27933 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -49,6 +49,7 @@ const struct st_tracked_state st_update_clip; const struct st_tracked_state st_update_clear_color; const struct st_tracked_state st_update_depth; const struct st_tracked_state st_update_fs; +const struct st_tracked_state st_update_vs; const struct st_tracked_state st_update_setup; const struct st_tracked_state st_update_polygon_stipple; const struct st_tracked_state st_update_viewport; diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 9c6bc1ce2a..6fa4f53c73 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -32,15 +32,38 @@ #include "st_context.h" #include "pipe/p_context.h" #include "st_atom.h" +#include "st_program.h" +#include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "pipe/tgsi/core/tgsi_dump.h" + +static void compile_fs( struct st_context *st, + struct st_fragment_program *fs ) +{ + /* XXX: fix static allocation of tokens: + */ + tgsi_mesa_compile_fp_program( &fs->Base, fs->tokens, ST_FP_MAX_TOKENS ); + + tgsi_dump( fs->tokens, TGSI_DUMP_VERBOSE ); +} static void update_fs( struct st_context *st ) { struct pipe_fs_state fs; + struct st_fragment_program *fp = st_fragment_program(st->ctx->FragmentProgram._Current); + + memset( &fs, 0, sizeof(fs) ); - fs.fp = st->ctx->FragmentProgram._Current; + if (fp->dirty) + compile_fs( st, fp ); + + fs.inputs_read = fp->Base.Base.InputsRead; + fs.tokens = &fp->tokens[0]; - if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0) { + if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0 || + fp->dirty) + { + fp->dirty = 0; st->state.fs = fs; st->pipe->set_fs_state(st->pipe, &fs); } diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c new file mode 100644 index 0000000000..6a26bfdd19 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -0,0 +1,49 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "st_context.h" +#include "pipe/p_context.h" +#include "st_atom.h" + + + +static void update_vs( struct st_context *st ) +{ +} + + +const struct st_tracked_state st_update_vs = { + .dirty = { + .mesa = 0, + .st = ST_NEW_VERTEX_PROGRAM, + }, + .update = update_vs +}; diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 327b627722..18061ca69c 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -52,6 +52,7 @@ static void st_bind_program( GLcontext *ctx, switch (target) { case GL_VERTEX_PROGRAM_ARB: + st->dirty.st |= ST_NEW_VERTEX_PROGRAM; break; case GL_FRAGMENT_PROGRAM_ARB: st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; @@ -66,16 +67,23 @@ static struct gl_program *st_new_program( GLcontext *ctx, struct st_context *st = st_context(ctx); switch (target) { - case GL_VERTEX_PROGRAM_ARB: - return _mesa_init_vertex_program(ctx, - CALLOC_STRUCT(gl_vertex_program), - target, - id); + case GL_VERTEX_PROGRAM_ARB: { + struct st_vertex_program *prog = CALLOC_STRUCT(st_vertex_program); + + prog->id = st->program_id++; + prog->dirty = 1; + + return _mesa_init_vertex_program( ctx, + &prog->Base, + target, + id ); + } case GL_FRAGMENT_PROGRAM_ARB: { struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program); prog->id = st->program_id++; + prog->dirty = 1; return _mesa_init_fragment_program( ctx, &prog->Base, @@ -106,40 +114,25 @@ static void st_program_string_notify( GLcontext *ctx, GLenum target, struct gl_program *prog ) { - if (target == GL_FRAGMENT_PROGRAM_ARB) { - struct st_context *st = st_context(ctx); + struct st_context *st = st_context(ctx); - if (prog == &st->ctx->FragmentProgram._Current->Base) - { - struct st_fragment_program *p = - (struct st_fragment_program *) prog; + if (target == GL_FRAGMENT_PROGRAM_ARB) { + struct st_fragment_program *p = (struct st_fragment_program *)prog; + if (prog == &ctx->FragmentProgram._Current->Base) st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - p->id = st->program_id++; -#if 0 - p->param_state = p->Base.Base.Parameters->StateFlags; - p->translated = 0; -#endif - - /* Gack! do this in the compiler: - */ - if (p->Base.FogOption) { - /* add extra instructions to do fog, then turn off FogOption field */ - _mesa_append_fog_code(ctx, &p->Base); - p->Base.FogOption = GL_NONE; - } - - /* XXX: Not hooked-up yet. */ - { - struct tgsi_token tokens[1024]; - - tgsi_mesa_compile_fp_program( prog, tokens, 1024 ); - tgsi_dump( tokens, TGSI_DUMP_VERBOSE ); - } - } + p->id = st->program_id++; + p->param_state = p->Base.Base.Parameters->StateFlags; } else if (target == GL_VERTEX_PROGRAM_ARB) { + struct st_vertex_program *p = (struct st_vertex_program *)prog; + + if (prog == &ctx->VertexProgram._Current->Base) + st->dirty.st |= ST_NEW_VERTEX_PROGRAM; + + p->id = st->program_id++; + p->param_state = p->Base.Base.Parameters->StateFlags; /* Also tell tnl about it: */ diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index c1d868604c..ef3cdb3b09 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -40,6 +40,7 @@ struct st_fragment_program; #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 +#define ST_NEW_VERTEX_PROGRAM 0x4 struct st_state_flags { GLuint mesa; diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index a47059d7a6..b28887946d 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -34,6 +34,12 @@ #ifndef ST_PROGRAM_H #define ST_PROGRAM_H +#include "mtypes.h" +#include "pipe/tgsi/core/tgsi_token.h" + +#define ST_FP_MAX_TOKENS 1024 + + struct st_fragment_program { struct gl_fragment_program Base; @@ -43,6 +49,11 @@ struct st_fragment_program * ProgramStringNotify changes. */ + + struct tgsi_token tokens[ST_FP_MAX_TOKENS]; + GLboolean dirty; + + #if 0 GLfloat (*cbuffer)[4]; GLuint nr_constants; @@ -56,13 +67,38 @@ struct st_fragment_program const GLfloat *values; /* Pointer to tracked values */ } *param; GLuint nr_params; +#endif GLuint param_state; -#endif }; +struct st_vertex_program +{ + struct gl_vertex_program Base; + GLboolean error; /* If program is malformed for any reason. */ + + GLuint id; /* String id, for tracking + * ProgramStringNotify changes. + */ + + GLboolean dirty; + GLuint param_state; +}; + void st_init_cb_program( struct st_context *st ); void st_destroy_cb_program( struct st_context *st ); +static inline struct st_fragment_program * +st_fragment_program( struct gl_fragment_program *fp ) +{ + return (struct st_fragment_program *)fp; +} + +static inline struct st_vertex_program * +st_vertex_program( struct gl_vertex_program *vp ) +{ + return (struct st_vertex_program *)vp; +} + #endif -- cgit v1.2.3 From 97cd87ef74eb06aa8da52be42871120cca145269 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 20 Jul 2007 11:33:24 +0200 Subject: minor cleanups, disable debugging --- src/mesa/drivers/dri/i915tex/i830_vtbl.c | 19 ------------------- src/mesa/drivers/dri/i915tex/intel_buffers.c | 2 +- src/mesa/drivers/dri/i915tex/intel_context.c | 3 +-- src/mesa/drivers/dri/i915tex/intel_context.h | 2 +- src/mesa/drivers/dri/i915tex/intel_screen.c | 4 ++-- src/mesa/drivers/dri/i915tex/intel_state.c | 2 +- 6 files changed, 6 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/i830_vtbl.c b/src/mesa/drivers/dri/i915tex/i830_vtbl.c index e432648ada..03d4788be1 100644 --- a/src/mesa/drivers/dri/i915tex/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915tex/i830_vtbl.c @@ -597,25 +597,6 @@ i830_set_draw_region(struct intel_context *intel, i830_state_draw_region(intel, &i830->state, color_region, depth_region); } -#if 0 -static void -i830_update_color_z_regions(intelContextPtr intel, - const intelRegion * colorRegion, - const intelRegion * depthRegion) -{ - i830ContextPtr i830 = I830_CONTEXT(intel); - - i830->state.Buffer[I830_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(colorRegion->pitch) | - BUF_3D_USE_FENCE); - i830->state.Buffer[I830_DESTREG_CBUFADDR2] = colorRegion->offset; - - i830->state.Buffer[I830_DESTREG_DBUFADDR1] = - (BUF_3D_ID_DEPTH | BUF_3D_PITCH(depthRegion->pitch) | BUF_3D_USE_FENCE); - i830->state.Buffer[I830_DESTREG_DBUFADDR2] = depthRegion->offset; -} -#endif - /* This isn't really handled at the moment. */ diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index ee3266e09b..4b2a4acaf9 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -645,7 +645,7 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) return ret; } - + void intelSwapBuffers(__DRIdrawablePrivate * dPriv) { diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c index 769287ba6a..f032c97dfb 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ b/src/mesa/drivers/dri/i915tex/intel_context.c @@ -703,8 +703,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) } - -/* Lock the hardware and validate our state. +/* Lock the hardware and validate our state. */ void LOCK_HARDWARE( struct intel_context *intel ) { diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index 628a15f8de..535a93effa 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -335,7 +335,7 @@ __memcpy(void *to, const void *from, size_t n) /* ================================================================ * Debugging: */ -#define DO_DEBUG 1 +#define DO_DEBUG 0 #if DO_DEBUG extern int INTEL_DEBUG; #else diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c index 5ad0932e2e..3c0ae854e4 100644 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ b/src/mesa/drivers/dri/i915tex/intel_screen.c @@ -118,7 +118,7 @@ intel_recreate_static(intelScreenPrivate *intelScreen, * Only used for real front buffer now. * * Note that these don't allocate video memory, just describe - * allocations alread made by the X server. + * allocations already made by the X server. */ static void intel_recreate_static_regions(intelScreenPrivate *intelScreen) @@ -222,7 +222,7 @@ intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, intelScreen->front.handle = sarea->front_handle; intelScreen->front.size = sarea->front_size; - if (1) + if (0) intelPrintSAREA(sarea); } diff --git a/src/mesa/drivers/dri/i915tex/intel_state.c b/src/mesa/drivers/dri/i915tex/intel_state.c index b2773990e4..5c5f2c6de5 100644 --- a/src/mesa/drivers/dri/i915tex/intel_state.c +++ b/src/mesa/drivers/dri/i915tex/intel_state.c @@ -194,7 +194,7 @@ intelClearColor(GLcontext * ctx, const GLfloat color[4]) * Update the viewport transformation matrix. Depends on: * - viewport pos/size * - depthrange - * - window pos/size or FBO size + * - window size or FBO size */ static void intelCalcViewport(GLcontext * ctx) -- cgit v1.2.3 From 118bd626b10d213eb022ac2618bf2efc5fe734d1 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 20 Jul 2007 11:34:26 +0200 Subject: fix bogus recently introduced function param comments --- src/mesa/main/buffers.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index 7764a5d3b2..0e6ca8ea1c 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -496,7 +496,6 @@ set_color_output(GLcontext *ctx, GLuint output, GLenum buffer, * \param destMask array[n] of BUFFER_* bitmasks which correspond to the * colorbuffer names. (i.e. GL_FRONT_AND_BACK => * BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT). - * \param callDriver call driver or not (bad idea sometimes this is called) */ void _mesa_drawbuffers(GLcontext *ctx, GLuint n, const GLenum *buffers, @@ -574,7 +573,6 @@ _mesa_readbuffer_update_fields(GLcontext *ctx, GLenum buffer) /** * Called by glReadBuffer to set the source renderbuffer for reading pixels. * \param mode color buffer such as GL_FRONT, GL_BACK, etc. - * \param callDriver call driver or not (bad idea sometimes this is called) */ void GLAPIENTRY _mesa_ReadBuffer(GLenum buffer) -- cgit v1.2.3 From 36fce66034887009749c3d52587c293da2457820 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Sat, 21 Jul 2007 16:51:15 +0200 Subject: set the _BaseFormat for window-framebuffers, could hit some assertion otherwise --- src/mesa/drivers/dri/i915tex/intel_fbo.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c index 399a280c47..04ca5ae8ba 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c @@ -350,6 +350,21 @@ intel_new_renderbuffer_fb(GLuint intFormat) irb->Base.ClassID = INTEL_RB_CLASS; irb->Base.InternalFormat = intFormat; + switch (intFormat) { + case GL_RGB5: + case GL_RGBA8: + irb->Base._BaseFormat = GL_RGBA; + break; + case GL_DEPTH_COMPONENT16: + irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + break; + case GL_DEPTH24_STENCIL8_EXT: + irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; + break; + default: + assert(0); + } + /* intel-specific methods */ irb->Base.Delete = intel_delete_renderbuffer; irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; -- cgit v1.2.3 From 98eaf5503d0d7c4f18fab6910a08aba7a3d08639 Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 23 Jul 2007 18:26:25 +0200 Subject: Execute fs tokens. Fix align128 bug. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 44 +++++++++++++++++++++++++++++++++++-- src/mesa/pipe/tgsi/core/tgsi_util.c | 15 +++++++++---- 2 files changed, 53 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 7b1c90cba6..536365f219 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -37,6 +37,7 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" +#include "core/tgsi_core.h" struct exec_machine { const struct setup_coefficient *coef; /**< will point to quad->coef */ @@ -180,8 +181,47 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) #endif } -#if 0 - softpipe->run_fs( tri->fp, quad, &tri->outputs ); +#if 1 + /*softpipe->run_fs( tri->fp, quad, &tri->outputs );*/ + + { + struct tgsi_exec_machine machine; + struct tgsi_exec_vector inputs[FRAG_ATTRIB_MAX + 1]; + struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; + struct tgsi_exec_vector *ainputs; + struct tgsi_exec_vector *aoutputs; + GLuint i, total; + + ainputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); + aoutputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs ); + + for( i = total = 0; i < PIPE_ATTRIB_MAX; i++ ) { + GLuint attr; + + attr = softpipe->fp_attr_to_slot[i]; + if( attr ) { + assert( total < FRAG_ATTRIB_MAX ); + assert( attr < FRAG_ATTRIB_MAX ); + assert( sizeof( ainputs[0] ) == sizeof( exec.attr[0] ) ); + + memcpy( + &ainputs[total], + exec.attr[attr], + sizeof( ainputs[0] ) ); + total++; + } + } + + tgsi_exec_machine_init( + &machine, + softpipe->fs.tokens ); + + machine.Inputs = ainputs; + machine.Outputs = aoutputs; + + tgsi_exec_machine_run( + &machine ); + } #else { GLuint attr = softpipe->fp_attr_to_slot[FRAG_ATTRIB_COL0]; diff --git a/src/mesa/pipe/tgsi/core/tgsi_util.c b/src/mesa/pipe/tgsi/core/tgsi_util.c index 2331affdfd..38d6d6e6bc 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_util.c +++ b/src/mesa/pipe/tgsi/core/tgsi_util.c @@ -1,15 +1,22 @@ #include "tgsi_platform.h" #include "tgsi_core.h" +union pointer_hack +{ + void *pointer; + unsigned long long uint64; +}; + void * tgsi_align_128bit( void *unaligned ) { - GLuint *ptr, addr; + union pointer_hack ph; - ptr = (GLuint *) unaligned; - addr = (*(GLuint *) &ptr + 15) & ~15; - return *(void **) &addr; + ph.uint64 = 0; + ph.pointer = unaligned; + ph.uint64 = (ph.uint64 + 15) & ~15; + return ph.pointer; } GLuint -- cgit v1.2.3 From 4477a01372c87b84e859c7522501d55a62f11b88 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 09:57:26 -0600 Subject: call ctx->Driver.NewProgram() instead of _mesa_new_program() --- src/mesa/shader/program.c | 2 +- src/mesa/shader/slang/slang_compile.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 4205919828..1f227390af 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -333,7 +333,7 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog) { struct gl_program *clone; - clone = _mesa_new_program(ctx, prog->Target, prog->Id); + clone = ctx->Driver.NewProgram(ctx, prog->Target, prog->Id); if (!clone) return NULL; diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 70f5aac16d..4e29e8dcc9 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -2135,7 +2135,7 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) progTarget = GL_FRAGMENT_PROGRAM_ARB; shader->Programs = (struct gl_program **) malloc(sizeof(struct gl_program*)); - shader->Programs[0] = _mesa_new_program(ctx, progTarget, 1); + shader->Programs[0] = ctx->Driver.NewProgram(ctx, progTarget, 1); shader->NumPrograms = 1; shader->Programs[0]->Parameters = _mesa_new_parameter_list(); -- cgit v1.2.3 From 47fafcf06f04c0db2d9908f30cfce6cd564d8973 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 09:58:19 -0600 Subject: use enabled/linked fragment shader in preference to a fragment program --- src/mesa/state_tracker/st_atom_fs.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 6fa4f53c73..2397828e4c 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -50,7 +50,15 @@ static void compile_fs( struct st_context *st, static void update_fs( struct st_context *st ) { struct pipe_fs_state fs; - struct st_fragment_program *fp = st_fragment_program(st->ctx->FragmentProgram._Current); + struct st_fragment_program *fp; + + if (st->ctx->Shader.CurrentProgram && + st->ctx->Shader.CurrentProgram->LinkStatus) { + fp = st_fragment_program(st->ctx->Shader.CurrentProgram->FragmentProgram); + } + else if (st->ctx->FragmentProgram._Current) { + fp = st_fragment_program(st->ctx->FragmentProgram._Current); + } memset( &fs, 0, sizeof(fs) ); @@ -72,7 +80,7 @@ static void update_fs( struct st_context *st ) const struct st_tracked_state st_update_fs = { .dirty = { - .mesa = 0, + .mesa = _NEW_PROGRAM, .st = ST_NEW_FRAGMENT_PROGRAM, }, .update = update_fs -- cgit v1.2.3 From ef3849ffa0af7f2aad8a69a7d04a4dd780fa247b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 12:12:52 -0600 Subject: rebuild quad pipeline when framebuffer changes too --- src/mesa/pipe/softpipe/sp_state_derived.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index fcdedb54a9..27220ea8fa 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -142,6 +142,7 @@ void softpipe_update_derived( struct softpipe_context *softpipe ) if (softpipe->dirty & (SP_NEW_BLEND | SP_NEW_DEPTH_TEST | SP_NEW_ALPHA_TEST | + SP_NEW_FRAMEBUFFER | SP_NEW_STENCIL | SP_NEW_SETUP | SP_NEW_FS)) -- cgit v1.2.3 From 43d7c1fe9d2544b851afa2f1c1e245a63860e670 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 12:27:38 -0600 Subject: Basic fragment programs run now. Handling of constants might change. For now, the st_fragment_program struct contains a pipe_constant_buffer (not a pointer). --- src/mesa/pipe/p_state.h | 12 ++++++------ src/mesa/pipe/softpipe/sp_quad_fs.c | 30 ++++++++++++++++++++++++++---- src/mesa/state_tracker/st_atom_fs.c | 25 ++++++++++++++++++++----- src/mesa/state_tracker/st_program.h | 1 + 4 files changed, 53 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index e3f62a80ad..c3704177b3 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -122,18 +122,18 @@ struct pipe_clip_state { }; -struct pipe_fs_state { - GLuint inputs_read; /* FRAG_ATTRIB_* */ - const struct tgsi_token *tokens; - -}; - struct pipe_constant_buffer { GLfloat constant[PIPE_MAX_CONSTANT][4]; GLuint nr_constants; }; +struct pipe_fs_state { + GLuint inputs_read; /* FRAG_ATTRIB_* */ + const struct tgsi_token *tokens; + struct pipe_constant_buffer *constants; /* XXX temporary? */ +}; + struct pipe_depth_state { GLuint enabled:1; /**< depth test enabled? */ diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 536365f219..846c330056 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -190,16 +190,21 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; struct tgsi_exec_vector *ainputs; struct tgsi_exec_vector *aoutputs; - GLuint i, total; + GLuint i /*, total*/; + +#ifdef DEBUG + memset(&machine, 0, sizeof(machine)); +#endif ainputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); aoutputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs ); +#if 0 for( i = total = 0; i < PIPE_ATTRIB_MAX; i++ ) { GLuint attr; attr = softpipe->fp_attr_to_slot[i]; - if( attr ) { + if( attr || total == 0) { assert( total < FRAG_ATTRIB_MAX ); assert( attr < FRAG_ATTRIB_MAX ); assert( sizeof( ainputs[0] ) == sizeof( exec.attr[0] ) ); @@ -211,16 +216,33 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) total++; } } +#else + /* load input registers */ + /* XXX simpler than above, but might not be right... */ + for (i = 0; i < softpipe->nr_attrs; i++) { + memcpy( + &ainputs[i], + exec.attr[i], + sizeof( ainputs[0] ) ); + } +#endif + /* init machine state */ tgsi_exec_machine_init( &machine, softpipe->fs.tokens ); machine.Inputs = ainputs; machine.Outputs = aoutputs; + machine.Consts = softpipe->fs.constants->constant; /* XXX alignment? */ + + /* run shader */ + tgsi_exec_machine_run( &machine ); - tgsi_exec_machine_run( - &machine ); + /* store result color */ + memcpy(quad->outputs.color, + &aoutputs[FRAG_ATTRIB_COL0].xyzw[0].f[0], + sizeof(quad->outputs.color)); } #else { diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 2397828e4c..9ca1807913 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -29,6 +29,7 @@ * Keith Whitwell */ +#include "shader/prog_parameter.h" #include "st_context.h" #include "pipe/p_context.h" #include "st_atom.h" @@ -50,23 +51,37 @@ static void compile_fs( struct st_context *st, static void update_fs( struct st_context *st ) { struct pipe_fs_state fs; - struct st_fragment_program *fp; + struct st_fragment_program *fp = NULL; + struct gl_program_parameter_list *params = NULL; if (st->ctx->Shader.CurrentProgram && - st->ctx->Shader.CurrentProgram->LinkStatus) { - fp = st_fragment_program(st->ctx->Shader.CurrentProgram->FragmentProgram); + st->ctx->Shader.CurrentProgram->LinkStatus && + st->ctx->Shader.CurrentProgram->FragmentProgram) { + struct gl_fragment_program *f + = st->ctx->Shader.CurrentProgram->FragmentProgram; + fp = st_fragment_program(f); + params = f->Base.Parameters; } else if (st->ctx->FragmentProgram._Current) { fp = st_fragment_program(st->ctx->FragmentProgram._Current); + params = st->ctx->FragmentProgram._Current->Base.Parameters; } - memset( &fs, 0, sizeof(fs) ); + if (fp && params) { + /* load program's constants array */ + fp->constants.nr_constants = params->NumParameters; + memcpy(fp->constants.constant, + params->ParameterValues, + params->NumParameters * sizeof(GLfloat) * 4); + } if (fp->dirty) compile_fs( st, fp ); - + + memset( &fs, 0, sizeof(fs) ); fs.inputs_read = fp->Base.Base.InputsRead; fs.tokens = &fp->tokens[0]; + fs.constants = &fp->constants; if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0 || fp->dirty) diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index b28887946d..f6d5f6d76c 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -53,6 +53,7 @@ struct st_fragment_program struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; + struct pipe_constant_buffer constants; #if 0 GLfloat (*cbuffer)[4]; -- cgit v1.2.3 From 67b61520ce1fabe2256e606ea73729f44d57fa99 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 13:48:20 -0600 Subject: added missing OPCODE_ARL to TGSI_OPCODES[], added sanity check assertion --- src/mesa/pipe/tgsi/core/tgsi_dump.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index 0345fd93f7..33a898ea04 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -179,6 +179,7 @@ static const char *TGSI_IMMS[] = static const char *TGSI_OPCODES[] = { + "OPCODE_ARL", "OPCODE_MOV", "OPCODE_LIT", "OPCODE_RCP", @@ -404,7 +405,7 @@ tgsi_dump( static GLuint counter = 0; char buffer[64]; - sprintf( buffer, "sbir-dump-%.4u.txt", counter++ ); + sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); dump.file = fopen( buffer, "wt" ); #else dump.file = stderr; @@ -412,9 +413,12 @@ tgsi_dump( #endif } + /* sanity check */ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); + tgsi_parse_init( &parse, tokens ); - TXT( "sbir-dump" ); + TXT( "tgsi-dump begin" ); CHR( '\n' ); TXT( "\nMajorVersion: " ); @@ -868,6 +872,8 @@ tgsi_dump( } } + TXT( "\ntgsi-dump end\n" ); + tgsi_parse_free( &parse ); fclose( dump.file ); -- cgit v1.2.3 From 2de4c7573efe7bdccea48448c37a9f28124b8df7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 14:31:16 -0600 Subject: Fix a typo (i << i) that was fouling up the vertex layout info. --- src/mesa/pipe/softpipe/sp_state_derived.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 27220ea8fa..a470cc256e 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -63,11 +63,14 @@ static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = }; -/* Derived from: fs, setup states. +/** + * Determine which post-transform / pre-rasterization vertex attributes + * we need. + * Derived from: fs, setup states. */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - const GLuint inputsRead = softpipe->fs.inputs_read; + const GLbitfield inputsRead = softpipe->fs.inputs_read; GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; GLbitfield attr_mask = 0x0; GLuint i; @@ -87,7 +90,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) * as fixed point or ubyte format. */ for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { - if (inputsRead & (i << i)) { + if (inputsRead & (1 << i)) { if (softpipe->setup.flatshade && (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1)) EMIT_ATTR(frag_to_vf[i], i, INTERP_CONSTANT); @@ -97,7 +100,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) } for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { - if (inputsRead & (i << i)) { + if (inputsRead & (1 << i)) { EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); } } -- cgit v1.2.3 From 1475d74cfba634cf24da30e48fe66de9c2d75d55 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 14:44:44 -0600 Subject: Compute need_z and need_w properly and use the correct const/linear/perspective interpolation in shade_quad(). --- src/mesa/pipe/softpipe/sp_quad_fs.c | 32 +++++++++++++++---------------- src/mesa/pipe/softpipe/sp_state_derived.c | 21 +++++++++++++++++++- 2 files changed, 36 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 846c330056..526992408a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -121,12 +121,11 @@ static INLINE void pinterp( struct exec_machine *exec, static void shade_quad( struct quad_stage *qs, struct quad_header *quad ) { - struct softpipe_context *softpipe = qs->softpipe; + const struct softpipe_context *softpipe = qs->softpipe; struct exec_machine exec; - GLfloat fx = quad->x0; - GLfloat fy = quad->y0; + const GLfloat fx = quad->x0; + const GLfloat fy = quad->y0; GLuint i, j; - GLboolean need_z = softpipe->depth_test.enabled; /* XXX hack */ exec.coef = quad->coef; @@ -142,27 +141,22 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0; exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; - /* Z and W are done by linear interpolation: - * XXX we'll probably have to use integers for Z - */ - if (/*softpipe->*/need_z) { + /* Z and W are done by linear interpolation */ + if (softpipe->need_z) { linterp(&exec, 0, 2); /* attr[0].z */ } if (softpipe->need_w) { linterp(&exec, 0, 3); /* attr[0].w */ -// invert(&exec, 0, 3); + /*invert(&exec, 0, 3);*/ } /* Interpolate all the remaining attributes. This will get pushed * into the fragment program's responsibilities at some point. + * Start at 1 to skip fragment position attribute (computed above). */ for (i = 1; i < quad->nr_attrs; i++) { -#if 1 - for (j = 0; j < NUM_CHANNELS; j++) - linterp(&exec, i, j); -#else - switch (quad->interp[i]) { + switch (softpipe->interp[i]) { case INTERP_CONSTANT: for (j = 0; j < NUM_CHANNELS; j++) cinterp(&exec, i, j); @@ -178,7 +172,6 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) pinterp(&exec, i, j); break; } -#endif } #if 1 @@ -243,6 +236,13 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) memcpy(quad->outputs.color, &aoutputs[FRAG_ATTRIB_COL0].xyzw[0].f[0], sizeof(quad->outputs.color)); + if (softpipe->need_z) { + /* XXX temporary */ + quad->outputs.depth[0] = exec.attr[0][2][0]; + quad->outputs.depth[1] = exec.attr[0][2][1]; + quad->outputs.depth[2] = exec.attr[0][2][2]; + quad->outputs.depth[3] = exec.attr[0][2][3]; + } } #else { @@ -253,7 +253,7 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) exec.attr[attr], sizeof(quad->outputs.color)); - if (need_z) { + if (softpipe->need_z) { quad->outputs.depth[0] = exec.attr[0][2][0]; quad->outputs.depth[1] = exec.attr[0][2][1]; quad->outputs.depth[2] = exec.attr[0][2][2]; diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index a470cc256e..84a1ec79f8 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -75,6 +75,24 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) GLbitfield attr_mask = 0x0; GLuint i; + /* Need Z if depth test is enabled or the fragment program uses the + * fragment position (XYZW). + */ + if (softpipe->depth_test.enabled || + (inputsRead & FRAG_ATTRIB_WPOS)) + softpipe->need_z = GL_TRUE; + else + softpipe->need_z = GL_FALSE; + + /* Need W if we do any perspective-corrected interpolation or the + * fragment program uses the fragment position. + */ + if (inputsRead & FRAG_ATTRIB_WPOS) + softpipe->need_w = GL_TRUE; + else + softpipe->need_w = GL_FALSE; + + softpipe->nr_attrs = 0; memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); @@ -101,7 +119,8 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { if (inputsRead & (1 << i)) { - EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); + EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); + softpipe->need_w = GL_TRUE; } } -- cgit v1.2.3 From 8c99f7ba61fcc95ef22d8ccd8cce3e31eaa67757 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 14:45:38 -0600 Subject: s/GLuint/GLbitfield/ --- src/mesa/pipe/p_state.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index c3704177b3..4ae8928018 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -129,7 +129,7 @@ struct pipe_constant_buffer { struct pipe_fs_state { - GLuint inputs_read; /* FRAG_ATTRIB_* */ + GLbitfield inputs_read; /* FRAG_ATTRIB_* */ const struct tgsi_token *tokens; struct pipe_constant_buffer *constants; /* XXX temporary? */ }; -- cgit v1.2.3 From 6a235d3fe5ccaa8cd0d4a2f3356c1513006dadcb Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 14:45:57 -0600 Subject: comments --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 27 ++++++++++++++++++++++----- 1 file changed, 22 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 291e62d262..dc25bac4ea 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -31,22 +31,34 @@ map_register_file( } } -/* +/** * Map mesa register file index to SBIR index. * Take special care when processing input and output indices. + * \param processor either TGSI_PROCESSOR_FRAGMENT or TGSI_PROCESSOR_VERTEX + * \param file one of TGSI_FILE_x + * \param index the mesa register file index + * \param usage_bitmask ??? */ static GLuint map_register_file_index( GLuint processor, GLuint file, GLuint index, - GLuint usage_bitmask ) + GLbitfield usage_bitmask ) { GLuint mapped_index; GLuint i; switch( file ) { case TGSI_FILE_INPUT: + /* + * The fragment/vertex program input indexes (FRAG/VERT_ATTRIB_x) get + * mapped to a packed sequence of integers. + * If a program uses one input attribute, the mapped index will be 1. + * If a program uses two input attribs, the mapped indexes will be 1,2. + * If a program uses 3 input attribs, the mapped indexes will be 1,2,3. + * etc. + */ assert( index < 32 ); assert( usage_bitmask & (1 << index) ); mapped_index = 0; @@ -55,11 +67,15 @@ map_register_file_index( mapped_index++; } } + printf("Map input %d to %d\n", index, mapped_index); break; case TGSI_FILE_OUTPUT: - assert( usage_bitmask == 0 ); + assert( usage_bitmask == 0x0 ); if( processor == TGSI_PROCESSOR_FRAGMENT ) { + /* depth result -> index 0 + * color results -> index 1, 2, ... + */ if( index == FRAG_RESULT_DEPR ) { mapped_index = 0; } @@ -69,6 +85,7 @@ map_register_file_index( } } else { + /* mapped_index = VERT_RESULT_x */ mapped_index = index; } break; @@ -137,7 +154,7 @@ convert_writemask( static GLboolean compile_instruction( - struct prog_instruction *inst, + const struct prog_instruction *inst, struct tgsi_full_instruction *fullinst, GLuint inputs_read, GLuint processor ) @@ -158,7 +175,7 @@ compile_instruction( processor, fulldst->DstRegister.File, inst->DstReg.Index, - 0 ); + 0x0 ); fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); for( i = 0; i < fullinst->Instruction.NumSrcRegs; i++ ) { -- cgit v1.2.3 From 5c2cff44d22c09e018a4b43a51cf6781ce4ca5b9 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 15:38:41 -0600 Subject: added PROGRAM_UNIFORM to map_register_file() --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index dc25bac4ea..c784091985 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -18,6 +18,7 @@ map_register_file( case PROGRAM_STATE_VAR: //case PROGRAM_NAMED_PARAM: case PROGRAM_CONSTANT: + case PROGRAM_UNIFORM: return TGSI_FILE_CONSTANT; case PROGRAM_INPUT: return TGSI_FILE_INPUT; -- cgit v1.2.3 From 496ee38b7130c55cbdddb1456f8ad618897bd711 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 15:39:41 -0600 Subject: Add VF_ATTRIB_VAR0..7 for varying variables. The demos/fslight.c demo runs now. --- src/mesa/pipe/softpipe/sp_state_derived.c | 10 ++++++++++ src/mesa/vf/vf.h | 20 ++++++++++++++------ 2 files changed, 24 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 84a1ec79f8..26083c2962 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -60,6 +60,14 @@ static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = VF_ATTRIB_TEX5, VF_ATTRIB_TEX6, VF_ATTRIB_TEX7, + VF_ATTRIB_VAR0, + VF_ATTRIB_VAR1, + VF_ATTRIB_VAR2, + VF_ATTRIB_VAR3, + VF_ATTRIB_VAR4, + VF_ATTRIB_VAR5, + VF_ATTRIB_VAR6, + VF_ATTRIB_VAR7, }; @@ -109,6 +117,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { if (inputsRead & (1 << i)) { + assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); if (softpipe->setup.flatshade && (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1)) EMIT_ATTR(frag_to_vf[i], i, INTERP_CONSTANT); @@ -119,6 +128,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { if (inputsRead & (1 << i)) { + assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); softpipe->need_w = GL_TRUE; } diff --git a/src/mesa/vf/vf.h b/src/mesa/vf/vf.h index fcbf490ce1..fc988b9fb9 100644 --- a/src/mesa/vf/vf.h +++ b/src/mesa/vf/vf.h @@ -48,12 +48,20 @@ enum { VF_ATTRIB_TEX5 = 13, VF_ATTRIB_TEX6 = 14, VF_ATTRIB_TEX7 = 15, - VF_ATTRIB_POINTSIZE = 16, - VF_ATTRIB_BFC0 = 17, - VF_ATTRIB_BFC1 = 18, - VF_ATTRIB_CLIP_POS = 19, - VF_ATTRIB_VERTEX_HEADER = 20, - VF_ATTRIB_MAX = 21 + VF_ATTRIB_VAR0 = 16, + VF_ATTRIB_VAR1 = 17, + VF_ATTRIB_VAR2 = 18, + VF_ATTRIB_VAR3 = 19, + VF_ATTRIB_VAR4 = 20, + VF_ATTRIB_VAR5 = 21, + VF_ATTRIB_VAR6 = 22, + VF_ATTRIB_VAR7 = 23, + VF_ATTRIB_POINTSIZE = 24, + VF_ATTRIB_BFC0 = 25, + VF_ATTRIB_BFC1 = 26, + VF_ATTRIB_CLIP_POS = 27, + VF_ATTRIB_VERTEX_HEADER = 28, + VF_ATTRIB_MAX = 29 }; -- cgit v1.2.3 From 3ffd30638a3be1fafc976cc5d640022c734b0458 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 16:49:41 -0600 Subject: added GL_FRAGMENT_PROGRAM_NV in st_new_program() --- src/mesa/state_tracker/st_cb_program.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 18061ca69c..fa9eaf446e 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -79,7 +79,9 @@ static struct gl_program *st_new_program( GLcontext *ctx, id ); } - case GL_FRAGMENT_PROGRAM_ARB: { + case GL_FRAGMENT_PROGRAM_ARB: + case GL_FRAGMENT_PROGRAM_NV: + { struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program); prog->id = st->program_id++; -- cgit v1.2.3 From ed52adc64ed13583162a997f13f3bc4624feb5ee Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 16:50:00 -0600 Subject: added PROGRAM_NAMED_PARAM in map_register_file() --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index c784091985..21b488c98f 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -16,7 +16,7 @@ map_register_file( //case PROGRAM_LOCAL_PARAM: //case PROGRAM_ENV_PARAM: case PROGRAM_STATE_VAR: - //case PROGRAM_NAMED_PARAM: + case PROGRAM_NAMED_PARAM: case PROGRAM_CONSTANT: case PROGRAM_UNIFORM: return TGSI_FILE_CONSTANT; -- cgit v1.2.3 From 2eb9061f73523c3cfc2d77db789afa804c68ee7f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 16:50:46 -0600 Subject: check return value of tgsi_parse_init() --- src/mesa/pipe/tgsi/core/tgsi_exec.c | 25 +++++++++++++++++++++---- 1 file changed, 21 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 189a41167a..74734beb39 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -57,12 +57,17 @@ tgsi_exec_machine_init( struct tgsi_exec_machine *mach, struct tgsi_token *tokens ) { - GLuint i; + GLuint i, k; struct tgsi_parse_context parse; mach->Tokens = tokens; - tgsi_parse_init (&parse, mach->Tokens); + k = tgsi_parse_init (&parse, mach->Tokens); + if (k != TGSI_PARSE_OK) { + printf("Problem parsing!\n"); + return; + } + mach->Processor = parse.FullHeader.Processor.Processor; tgsi_parse_free (&parse); @@ -93,11 +98,17 @@ tgsi_exec_prepare( struct tgsi_exec_labels *labels ) { struct tgsi_parse_context parse; + GLuint k; mach->ImmLimit = 0; labels->count = 0; - tgsi_parse_init( &parse, mach->Tokens ); + k = tgsi_parse_init( &parse, mach->Tokens ); + if (k != TGSI_PARSE_OK) { + printf("Problem parsing!\n"); + return; + } + while( !tgsi_parse_end_of_tokens( &parse ) ) { GLuint pointer = parse.Position; GLuint i; @@ -2153,6 +2164,7 @@ tgsi_exec_machine_run2( mach->Temps); #else struct tgsi_parse_context parse; + GLuint k; mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; @@ -2162,7 +2174,12 @@ tgsi_exec_machine_run2( mach->Primitives[0] = 0; } - tgsi_parse_init( &parse, mach->Tokens ); + k = tgsi_parse_init( &parse, mach->Tokens ); + if (k != TGSI_PARSE_OK) { + printf("Problem parsing!\n"); + return; + } + while( !tgsi_parse_end_of_tokens( &parse ) ) { tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { -- cgit v1.2.3 From a502c08a255391311c64fa4c39d5026005206d2b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 24 Jul 2007 17:10:30 -0600 Subject: use consistent indentation --- src/mesa/pipe/tgsi/core/tgsi_exec.c | 734 ++++++++++++++++++------------------ 1 file changed, 361 insertions(+), 373 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 74734beb39..e0387b514d 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -1256,10 +1256,10 @@ exec_instruction( GLuint *programCounter ) { #if MESA - GET_CURRENT_CONTEXT(ctx); + GET_CURRENT_CONTEXT(ctx); #endif - GLuint chan_index; - union tgsi_exec_channel r[8]; + GLuint chan_index; + union tgsi_exec_channel r[8]; switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: @@ -1379,28 +1379,28 @@ exec_instruction( case TGSI_OPCODE_DP4: /* TGSI_OPCODE_DOT4 */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1] ); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); - FETCH(&r[1], 0, CHAN_W); - FETCH(&r[2], 1, CHAN_W); + FETCH(&r[1], 0, CHAN_W); + FETCH(&r[2], 1, CHAN_W); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1430,29 +1430,27 @@ exec_instruction( } break; - case TGSI_OPCODE_MIN: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); + case TGSI_OPCODE_MIN: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); - micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); + micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); - STORE(&r[0], 0, chan_index); - } - break; + STORE(&r[0], 0, chan_index); + } + break; - case TGSI_OPCODE_MAX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); + case TGSI_OPCODE_MAX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); - micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); + micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); - STORE(&r[0], 0, chan_index); - } - break; + STORE(&r[0], 0, chan_index); + } + break; case TGSI_OPCODE_SLT: /* TGSI_OPCODE_SETLT */ @@ -1464,18 +1462,17 @@ exec_instruction( } break; - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE(&r[0], 0, chan_index); - } - break; + STORE(&r[0], 0, chan_index); + } + break; case TGSI_OPCODE_MAD: /* TGSI_OPCODE_MADD */ @@ -1490,53 +1487,51 @@ exec_instruction( break; case TGSI_OPCODE_SUB: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); - micro_sub( &r[0], &r[0], &r[1] ); + micro_sub( &r[0], &r[0], &r[1] ); - STORE(&r[0], 0, chan_index); - } - break; + STORE(&r[0], 0, chan_index); + } + break; - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); - micro_sub( &r[1], &r[1], &r[2] ); - micro_mul( &r[0], &r[0], &r[1] ); - micro_add( &r[0], &r[0], &r[2] ); + micro_sub( &r[1], &r[1], &r[2] ); + micro_mul( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[2] ); - STORE(&r[0], 0, chan_index); - } - break; + STORE(&r[0], 0, chan_index); + } + break; - case TGSI_OPCODE_CND: - assert (0); - break; + case TGSI_OPCODE_CND: + assert (0); + break; - case TGSI_OPCODE_CND0: - assert (0); - break; + case TGSI_OPCODE_CND0: + assert (0); + break; - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - assert (0); - break; + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert (0); + break; - case TGSI_OPCODE_INDEX: - assert (0); - break; + case TGSI_OPCODE_INDEX: + assert (0); + break; - case TGSI_OPCODE_NEGATE: - assert (0); - break; + case TGSI_OPCODE_NEGATE: + assert (0); + break; case TGSI_OPCODE_FRAC: /* TGSI_OPCODE_FRC */ @@ -1547,9 +1542,9 @@ exec_instruction( } break; - case TGSI_OPCODE_CLAMP: - assert (0); - break; + case TGSI_OPCODE_CLAMP: + assert (0); + break; case TGSI_OPCODE_FLOOR: /* TGSI_OPCODE_FLR */ @@ -1568,11 +1563,11 @@ exec_instruction( } break; - case TGSI_OPCODE_EXPBASE2: + case TGSI_OPCODE_EXPBASE2: /* TGSI_OPCODE_EX2 */ - FETCH(&r[0], 0, CHAN_X); + FETCH(&r[0], 0, CHAN_X); - micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); + micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1588,51 +1583,51 @@ exec_instruction( } break; - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); - micro_pow( &r[0], &r[0], &r[1] ); + micro_pow( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); } break; - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - FETCH(&r[0], 0, CHAN_Y); - FETCH(&r[1], 1, CHAN_Z); + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + FETCH(&r[0], 0, CHAN_Y); + FETCH(&r[1], 1, CHAN_Z); - micro_mul( &r[2], &r[0], &r[1] ); + micro_mul( &r[2], &r[0], &r[1] ); - FETCH(&r[3], 0, CHAN_Z); - FETCH(&r[4], 1, CHAN_Y); + FETCH(&r[3], 0, CHAN_Z); + FETCH(&r[4], 1, CHAN_Y); - micro_mul( &r[5], &r[3], &r[4] ); - micro_sub( &r[2], &r[2], &r[5] ); + micro_mul( &r[5], &r[3], &r[4] ); + micro_sub( &r[2], &r[2], &r[5] ); if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { STORE( &r[2], 0, CHAN_X ); } - FETCH(&r[2], 1, CHAN_X); + FETCH(&r[2], 1, CHAN_X); - micro_mul( &r[3], &r[3], &r[2] ); + micro_mul( &r[3], &r[3], &r[2] ); - FETCH(&r[5], 0, CHAN_X); + FETCH(&r[5], 0, CHAN_X); - micro_mul( &r[1], &r[1], &r[5] ); - micro_sub( &r[3], &r[3], &r[1] ); + micro_mul( &r[1], &r[1], &r[5] ); + micro_sub( &r[3], &r[3], &r[1] ); if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { STORE( &r[3], 0, CHAN_Y ); } - micro_mul( &r[5], &r[5], &r[4] ); - micro_mul( &r[0], &r[0], &r[2] ); - micro_sub( &r[5], &r[5], &r[0] ); + micro_mul( &r[5], &r[5], &r[4] ); + micro_mul( &r[0], &r[0], &r[2] ); + micro_sub( &r[5], &r[5], &r[0] ); if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { STORE( &r[5], 0, CHAN_Z ); @@ -1641,292 +1636,285 @@ exec_instruction( if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); } - break; + break; case TGSI_OPCODE_MULTIPLYMATRIX: - assert (0); - break; + assert (0); + break; case TGSI_OPCODE_ABS: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); - micro_abs( &r[0], &r[0] ); + micro_abs( &r[0], &r[0] ); - STORE(&r[0], 0, chan_index); - } - break; + STORE(&r[0], 0, chan_index); + } + break; - case TGSI_OPCODE_RCC: - assert (0); - break; + case TGSI_OPCODE_RCC: + assert (0); + break; - case TGSI_OPCODE_DPH: - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); + case TGSI_OPCODE_DPH: + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1] ); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); - FETCH(&r[1], 1, CHAN_W); + FETCH(&r[1], 1, CHAN_W); - micro_add( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); } break; - case TGSI_OPCODE_COS: - FETCH(&r[0], 0, CHAN_X); + case TGSI_OPCODE_COS: + FETCH(&r[0], 0, CHAN_X); - micro_cos( &r[0], &r[0] ); + micro_cos( &r[0], &r[0] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); } break; - case TGSI_OPCODE_DDX: - assert (0); - break; + case TGSI_OPCODE_DDX: + assert (0); + break; - case TGSI_OPCODE_DDY: - assert (0); - break; + case TGSI_OPCODE_DDY: + assert (0); + break; - case TGSI_OPCODE_KIL: - exec_kil (mach, inst); - break; + case TGSI_OPCODE_KIL: + exec_kil (mach, inst); + break; - case TGSI_OPCODE_PK2H: - assert (0); - break; + case TGSI_OPCODE_PK2H: + assert (0); + break; - case TGSI_OPCODE_PK2US: - assert (0); - break; + case TGSI_OPCODE_PK2US: + assert (0); + break; - case TGSI_OPCODE_PK4B: - assert (0); - break; + case TGSI_OPCODE_PK4B: + assert (0); + break; - case TGSI_OPCODE_PK4UB: - assert (0); - break; + case TGSI_OPCODE_PK4UB: + assert (0); + break; - case TGSI_OPCODE_RFL: - assert (0); - break; + case TGSI_OPCODE_RFL: + assert (0); + break; - case TGSI_OPCODE_SEQ: - assert (0); - break; + case TGSI_OPCODE_SEQ: + assert (0); + break; - case TGSI_OPCODE_SFL: - assert (0); - break; + case TGSI_OPCODE_SFL: + assert (0); + break; - case TGSI_OPCODE_SGT: - assert (0); - break; + case TGSI_OPCODE_SGT: + assert (0); + break; - case TGSI_OPCODE_SIN: - FETCH(&r[0], 0, CHAN_X); + case TGSI_OPCODE_SIN: + FETCH(&r[0], 0, CHAN_X); - micro_sin( &r[0], &r[0] ); + micro_sin( &r[0], &r[0] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); } break; - case TGSI_OPCODE_SLE: - assert (0); - break; - - case TGSI_OPCODE_SNE: - assert (0); - break; + case TGSI_OPCODE_SLE: + assert (0); + break; - case TGSI_OPCODE_STR: - assert (0); - break; + case TGSI_OPCODE_SNE: + assert (0); + break; - case TGSI_OPCODE_TEX: + case TGSI_OPCODE_STR: + assert (0); + break; - switch (inst->InstructionExtTexture.Texture) - { - case TGSI_TEXTURE_1D: + case TGSI_OPCODE_TEX: + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: - FETCH(&r[0], 0, CHAN_X); + FETCH(&r[0], 0, CHAN_X); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) - { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[1], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[1] ); - break; + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[1], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[1] ); + break; - case TGSI_EXTSWIZZLE_ONE: - break; + case TGSI_EXTSWIZZLE_ONE: + break; - default: - assert (0); - } + default: + assert (0); + } #if MESA - fetch_texel_1d (ctx, - &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], - &r[0], - inst->FullSrcRegisters[1].SrcRegister.Index, - &r[0], &r[1], &r[2], &r[3]); + fetch_texel_1d (ctx, + &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], + &r[0], + inst->FullSrcRegisters[1].SrcRegister.Index, + &r[0], &r[1], &r[2], &r[3]); #endif - break; + break; - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) - { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[2], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[2] ); - micro_div( &r[1], &r[1], &r[2] ); - break; + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[2], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[2] ); + micro_div( &r[1], &r[1], &r[2] ); + break; - case TGSI_EXTSWIZZLE_ONE: - break; + case TGSI_EXTSWIZZLE_ONE: + break; - default: - assert (0); - } + default: + assert (0); + } #if MESA - fetch_texel_2d (ctx, - &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], - &r[0], &r[1], - inst->FullSrcRegisters[1].SrcRegister.Index, - &r[0], &r[1], &r[2], &r[3]); + fetch_texel_2d (ctx, + &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], + &r[0], &r[1], + inst->FullSrcRegisters[1].SrcRegister.Index, + &r[0], &r[1], &r[2], &r[3]); #endif - break; + break; - case TGSI_TEXTURE_3D: - case TGSI_TEXTURE_CUBE: + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) - { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - break; + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; - case TGSI_EXTSWIZZLE_ONE: - break; + case TGSI_EXTSWIZZLE_ONE: + break; - default: - assert (0); - } + default: + assert (0); + } #if MESA - fetch_texel_3d (ctx, - &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], - &r[0], &r[1], &r[2], - inst->FullSrcRegisters[1].SrcRegister.Index, - &r[0], &r[1], &r[2], &r[3]); + fetch_texel_3d (ctx, + &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], + &r[0], &r[1], &r[2], + inst->FullSrcRegisters[1].SrcRegister.Index, + &r[0], &r[1], &r[2], &r[3]); #endif - break; + break; - default: - assert (0); - } + default: + assert (0); + } FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[chan_index], 0, chan_index ); } break; - case TGSI_OPCODE_TXD: - assert (0); - break; + case TGSI_OPCODE_TXD: + assert (0); + break; - case TGSI_OPCODE_UP2H: - assert (0); - break; + case TGSI_OPCODE_UP2H: + assert (0); + break; - case TGSI_OPCODE_UP2US: - assert (0); - break; + case TGSI_OPCODE_UP2US: + assert (0); + break; - case TGSI_OPCODE_UP4B: - assert (0); - break; + case TGSI_OPCODE_UP4B: + assert (0); + break; - case TGSI_OPCODE_UP4UB: - assert (0); - break; + case TGSI_OPCODE_UP4UB: + assert (0); + break; - case TGSI_OPCODE_X2D: - assert (0); - break; + case TGSI_OPCODE_X2D: + assert (0); + break; - case TGSI_OPCODE_ARA: - assert (0); - break; + case TGSI_OPCODE_ARA: + assert (0); + break; - case TGSI_OPCODE_ARR: - assert (0); - break; + case TGSI_OPCODE_ARR: + assert (0); + break; - case TGSI_OPCODE_BRA: - assert (0); - break; + case TGSI_OPCODE_BRA: + assert (0); + break; - case TGSI_OPCODE_CAL: - assert (0); - break; + case TGSI_OPCODE_CAL: + assert (0); + break; - case TGSI_OPCODE_RET: - /* XXX: end of shader! */ - /*assert (0);*/ - break; + case TGSI_OPCODE_RET: + /* XXX: end of shader! */ + /*assert (0);*/ + break; - case TGSI_OPCODE_SSG: - assert (0); - break; + case TGSI_OPCODE_SSG: + assert (0); + break; - case TGSI_OPCODE_CMP: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); + case TGSI_OPCODE_CMP: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); - micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); + micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); - STORE(&r[0], 0, chan_index); - } - break; + STORE(&r[0], 0, chan_index); + } + break; case TGSI_OPCODE_SCS: if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { @@ -1948,13 +1936,13 @@ exec_instruction( } break; - case TGSI_OPCODE_TXB: - assert (0); - break; + case TGSI_OPCODE_TXB: + assert (0); + break; - case TGSI_OPCODE_NRM: - assert (0); - break; + case TGSI_OPCODE_NRM: + assert (0); + break; case TGSI_OPCODE_DIV: assert( 0 ); @@ -1975,49 +1963,49 @@ exec_instruction( } break; - case TGSI_OPCODE_TXL: - assert (0); - break; + case TGSI_OPCODE_TXL: + assert (0); + break; - case TGSI_OPCODE_BRK: - assert (0); - break; + case TGSI_OPCODE_BRK: + assert (0); + break; - case TGSI_OPCODE_IF: - assert (0); - break; + case TGSI_OPCODE_IF: + assert (0); + break; - case TGSI_OPCODE_LOOP: - assert (0); - break; + case TGSI_OPCODE_LOOP: + assert (0); + break; - case TGSI_OPCODE_REP: - assert (0); - break; + case TGSI_OPCODE_REP: + assert (0); + break; - case TGSI_OPCODE_ELSE: - assert (0); - break; + case TGSI_OPCODE_ELSE: + assert (0); + break; - case TGSI_OPCODE_ENDIF: - assert (0); - break; + case TGSI_OPCODE_ENDIF: + assert (0); + break; - case TGSI_OPCODE_ENDLOOP: - assert (0); - break; + case TGSI_OPCODE_ENDLOOP: + assert (0); + break; - case TGSI_OPCODE_ENDREP: - assert (0); - break; + case TGSI_OPCODE_ENDREP: + assert (0); + break; - case TGSI_OPCODE_PUSHA: - assert (0); - break; + case TGSI_OPCODE_PUSHA: + assert (0); + break; - case TGSI_OPCODE_POPA: - assert (0); - break; + case TGSI_OPCODE_POPA: + assert (0); + break; case TGSI_OPCODE_CEIL: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { @@ -2043,9 +2031,9 @@ exec_instruction( } break; - case TGSI_OPCODE_TRUNC: - assert (0); - break; + case TGSI_OPCODE_TRUNC: + assert (0); + break; case TGSI_OPCODE_SHL: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { @@ -2083,9 +2071,9 @@ exec_instruction( } break; - case TGSI_OPCODE_MOD: - assert (0); - break; + case TGSI_OPCODE_MOD: + assert (0); + break; case TGSI_OPCODE_XOR: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { @@ -2096,21 +2084,21 @@ exec_instruction( } break; - case TGSI_OPCODE_SAD: - assert (0); - break; + case TGSI_OPCODE_SAD: + assert (0); + break; - case TGSI_OPCODE_TXF: - assert (0); - break; + case TGSI_OPCODE_TXF: + assert (0); + break; - case TGSI_OPCODE_TXQ: - assert (0); - break; + case TGSI_OPCODE_TXQ: + assert (0); + break; - case TGSI_OPCODE_CONT: - assert (0); - break; + case TGSI_OPCODE_CONT: + assert (0); + break; case TGSI_OPCODE_EMIT: mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; @@ -2124,7 +2112,7 @@ exec_instruction( default: assert( 0 ); - } + } } @@ -2147,21 +2135,21 @@ tgsi_exec_machine_run2( struct tgsi_exec_labels *labels ) { #if MESA - GET_CURRENT_CONTEXT(ctx); - GLuint i; + GET_CURRENT_CONTEXT(ctx); + GLuint i; #endif #if XXX_SSE - fp_function function; + fp_function function; - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; - - function = (fp_function) x86_get_func (&mach->Function); + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + + function = (fp_function) x86_get_func (&mach->Function); - function (mach->Inputs, - mach->Outputs, - mach->Consts, - mach->Temps); + function (mach->Inputs, + mach->Outputs, + mach->Consts, + mach->Temps); #else struct tgsi_parse_context parse; GLuint k; -- cgit v1.2.3 From d24e60a6b10ab3cb365e9c8f9dc0b36864bd485d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 25 Jul 2007 12:15:02 -0600 Subject: patch things up enough to allow texture sampling to work --- src/mesa/pipe/tgsi/core/tgsi_exec.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index e0387b514d..db965ccbec 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -1,6 +1,15 @@ #include "tgsi_platform.h" #include "tgsi_core.h" +#define MESA 1 +#if MESA +#include "main/context.h" +#include "main/macros.h" +#include "main/colormac.h" +#include "swrast/swrast.h" +#include "swrast/s_context.h" +#endif + #define TILE_BOTTOM_LEFT 0 #define TILE_BOTTOM_RIGHT 1 #define TILE_TOP_LEFT 2 @@ -1039,6 +1048,10 @@ fetch_texel_1d( GLcontext *ctx, lambdas[3] = lambda; } + if (!swrast->TextureSample[unit]) { + _swrast_update_texture_samplers(ctx); + } + /* XXX use a float-valued TextureSample routine here!!! */ swrast->TextureSample[unit] (ctx, ctx->Texture.Unit[unit]._Current, @@ -1123,6 +1136,10 @@ fetch_texel_2d( GLcontext *ctx, lambdas[3] = lambda; } + if (!swrast->TextureSample[unit]) { + _swrast_update_texture_samplers(ctx); + } + /* XXX use a float-valued TextureSample routine here!!! */ swrast->TextureSample[unit]( ctx, @@ -1214,6 +1231,10 @@ fetch_texel_3d( GLcontext *ctx, lambdas[3] = lambda; } + if (!swrast->TextureSample[unit]) { + _swrast_update_texture_samplers(ctx); + } + /* XXX use a float-valued TextureSample routine here!!! */ swrast->TextureSample[unit] (ctx, ctx->Texture.Unit[unit]._Current, -- cgit v1.2.3 From be8725321ccbc4ca5f52afc9a3e257c91f43a119 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 25 Jul 2007 14:27:38 -0600 Subject: Fix pinterp() to compute 1 / FRAG_ATTRIB_WPOS.w Update comments too. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 526992408a..9b0b5b91e5 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -91,11 +91,11 @@ static INLINE void linterp( struct exec_machine *exec, * * Push into the fp: * - * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx - * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy - * INPUT[attr] = MUL INPUT[attr], INPUT_WPOS.wwww + * INPUT[attr] = MAD COEF_DADX[attr], INPUT_WPOS.xxxx, COEF_A0[attr] + * INPUT[attr] = MAD COEF_DADY[attr], INPUT_WPOS.yyyy, INPUT[attr] + * TMP = RCP INPUT_WPOS.w + * INPUT[attr] = MUL INPUT[attr], TMP.xxxx * - * (Or should that be 1/w ???) */ static INLINE void pinterp( struct exec_machine *exec, GLuint attrib, @@ -106,10 +106,11 @@ static INLINE void pinterp( struct exec_machine *exec, for (j = 0; j < QUAD_SIZE; j++) { const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; - const GLfloat invW = exec->attr[FRAG_ATTRIB_WPOS][3][j]; + /* FRAG_ATTRIB_WPOS.w here is really 1/w */ + const GLfloat w = 1.0 / exec->attr[FRAG_ATTRIB_WPOS][3][j]; exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + exec->coef[attrib].dadx[i] * x + - exec->coef[attrib].dady[i] * y) * invW); + exec->coef[attrib].dady[i] * y) * w); } } -- cgit v1.2.3 From 8b2955b736086103ac0d184a4b3f89d8ab8baab1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 25 Jul 2007 14:28:24 -0600 Subject: implement DDX, DDY instructions --- src/mesa/pipe/tgsi/core/tgsi_exec.c | 34 ++++++++++++++++++++++++++++++++-- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 6 ++++++ 2 files changed, 38 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index db965ccbec..6aaaef98a8 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -228,6 +228,28 @@ micro_cos( dst->f[3] = (GLfloat) cos( (GLdouble) src->f[3] ); } +static void +micro_ddx( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_ddy( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; +} + static void micro_div( union tgsi_exec_channel *dst, @@ -1715,11 +1737,19 @@ exec_instruction( break; case TGSI_OPCODE_DDX: - assert (0); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddx( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } break; case TGSI_OPCODE_DDY: - assert (0); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddy( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } break; case TGSI_OPCODE_KIL: diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 21b488c98f..a9415ac15b 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -261,6 +261,12 @@ compile_instruction( case OPCODE_DP3: fullinst->Instruction.Opcode = TGSI_OPCODE_DP3; break; + case OPCODE_DDX: + fullinst->Instruction.Opcode = TGSI_OPCODE_DDX; + break; + case OPCODE_DDY: + fullinst->Instruction.Opcode = TGSI_OPCODE_DDY; + break; case OPCODE_DP4: fullinst->Instruction.Opcode = TGSI_OPCODE_DP4; break; -- cgit v1.2.3 From 1f6b4b0f75d7cd7306e6aa24c27fd478f6b6c52d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 25 Jul 2007 14:32:28 -0600 Subject: rename some vars --- src/mesa/pipe/softpipe/sp_quad_fs.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 9b0b5b91e5..301bd13aa9 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -126,7 +126,7 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) struct exec_machine exec; const GLfloat fx = quad->x0; const GLfloat fy = quad->y0; - GLuint i, j; + GLuint attr, i; exec.coef = quad->coef; @@ -156,21 +156,21 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) * into the fragment program's responsibilities at some point. * Start at 1 to skip fragment position attribute (computed above). */ - for (i = 1; i < quad->nr_attrs; i++) { - switch (softpipe->interp[i]) { + for (attr = 1; attr < quad->nr_attrs; attr++) { + switch (softpipe->interp[attr]) { case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - cinterp(&exec, i, j); + for (i = 0; i < NUM_CHANNELS; i++) + cinterp(&exec, attr, i); break; case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - linterp(&exec, i, j); + for (i = 0; i < NUM_CHANNELS; i++) + linterp(&exec, attr, i); break; case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - pinterp(&exec, i, j); + for (i = 0; i < NUM_CHANNELS; i++) + pinterp(&exec, attr, i); break; } } -- cgit v1.2.3 From 0360b49afbcd839f99ba0745d01cf9dc5be4d122 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 25 Jul 2007 15:48:09 -0600 Subject: Implement line stippling. Also added draw_stage::reset_line_stipple(). There may be a better way of doing that though. --- src/mesa/pipe/draw/draw_clip.c | 7 +++++++ src/mesa/pipe/draw/draw_cull.c | 6 ++++++ src/mesa/pipe/draw/draw_flatshade.c | 7 +++++++ src/mesa/pipe/draw/draw_offset.c | 7 +++++++ src/mesa/pipe/draw/draw_private.h | 2 ++ src/mesa/pipe/draw/draw_twoside.c | 7 +++++++ src/mesa/pipe/draw/draw_unfilled.c | 7 +++++++ src/mesa/pipe/draw/draw_vb.c | 8 ++++++- src/mesa/pipe/softpipe/sp_context.h | 4 ++++ src/mesa/pipe/softpipe/sp_prim_setup.c | 38 ++++++++++++++++++++++++++++++++-- 10 files changed, 90 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index b220cc3643..f8bacf8a81 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -433,6 +433,12 @@ static void clip_end( struct draw_stage *stage ) } +static void clip_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + /** * Allocate a new clipper stage. * \return pointer to new stage object @@ -449,6 +455,7 @@ struct draw_stage *draw_clip_stage( struct draw_context *draw ) clipper->stage.line = clip_line; clipper->stage.tri = clip_tri; clipper->stage.end = clip_end; + clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; clipper->plane = draw->plane; diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index e563f9f45f..8b2ac5ea5f 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -111,6 +111,11 @@ static void cull_end( struct draw_stage *stage ) } +static void cull_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + /** * Create a new polygon culling stage. */ @@ -127,6 +132,7 @@ struct draw_stage *draw_cull_stage( struct draw_context *draw ) cull->stage.line = cull_line; cull->stage.tri = cull_tri; cull->stage.end = cull_end; + cull->stage.reset_stipple_counter = cull_reset_stipple_counter; return &cull->stage; } diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 09547865fa..cf5e762079 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -130,6 +130,12 @@ static void flatshade_end( struct draw_stage *stage ) } +static void flatshade_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + /** * Create flatshading drawing stage. */ @@ -146,6 +152,7 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) flatshade->stage.line = flatshade_line; flatshade->stage.tri = flatshade_tri; flatshade->stage.end = flatshade_end; + flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; flatshade->lookup = draw->vf_attr_to_slot; diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 1a7f56d0c5..9f66566353 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -145,6 +145,12 @@ static void offset_end( struct draw_stage *stage ) } +static void offset_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + /** * Create polygon offset drawing stage. */ @@ -161,6 +167,7 @@ struct draw_stage *draw_offset_stage( struct draw_context *draw ) offset->stage.line = offset_line; offset->stage.tri = offset_tri; offset->stage.end = offset_end; + offset->stage.reset_stipple_counter = offset_reset_stipple_counter; return &offset->stage; } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 573b5bf588..3dfaa0581d 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -98,6 +98,8 @@ struct draw_stage struct prim_header * ); void (*end)( struct draw_stage * ); + + void (*reset_stipple_counter)( struct draw_stage * ); }; diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 9f26335fb6..fdda6b362f 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -139,6 +139,12 @@ static void twoside_end( struct draw_stage *stage ) } +static void twoside_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + /** * Create twoside pipeline stage. */ @@ -155,6 +161,7 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) twoside->stage.line = twoside_line; twoside->stage.tri = twoside_tri; twoside->stage.end = twoside_end; + twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; twoside->lookup = draw->vf_attr_to_slot; diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index f0aee58333..82e8775f59 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -156,6 +156,12 @@ static void unfilled_end( struct draw_stage *stage ) } +static void unfilled_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + /** * Create unfilled triangle stage. */ @@ -173,6 +179,7 @@ struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) unfilled->stage.line = unfilled_line; unfilled->stage.tri = unfilled_tri; unfilled->stage.end = unfilled_end; + unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; return &unfilled->stage; } diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index ac721c589d..ac126c5baa 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -151,12 +151,14 @@ static void draw_indexed_prim( struct draw_context *draw, prim.v[0] = get_vertex( draw, elts[i + 0] ); prim.v[1] = get_vertex( draw, elts[i + 1] ); + first->reset_stipple_counter( first ); first->line( first, &prim ); } break; case GL_LINE_LOOP: if (count >= 2) { + first->reset_stipple_counter( first ); for (i = 1; i < count; i++) { prim.v[0] = get_vertex( draw, elts[i-1] ); prim.v[1] = get_vertex( draw, elts[i] ); @@ -178,6 +180,7 @@ static void draw_indexed_prim( struct draw_context *draw, * require more complex code here. */ if (count >= 2) { + first->reset_stipple_counter( first ); prim.v[0] = 0; prim.v[1] = get_vertex( draw, elts[0] ); @@ -313,13 +316,15 @@ static void draw_prim( struct draw_context *draw, for (i = 0; i+1 < count; i += 2) { prim.v[0] = get_vertex( draw, start + i + 0 ); prim.v[1] = get_vertex( draw, start + i + 1 ); - + + first->reset_stipple_counter( first ); first->line( first, &prim ); } break; case GL_LINE_LOOP: if (count >= 2) { + first->reset_stipple_counter( first ); for (i = 1; i < count; i++) { prim.v[0] = get_vertex( draw, start + i - 1 ); prim.v[1] = get_vertex( draw, start + i ); @@ -334,6 +339,7 @@ static void draw_prim( struct draw_context *draw, case GL_LINE_STRIP: if (count >= 2) { + first->reset_stipple_counter( first ); prim.v[0] = 0; prim.v[1] = get_vertex( draw, start + 0 ); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index ef1a06ccaf..8871b45d4f 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -111,12 +111,16 @@ struct softpipe_context { GLboolean need_z; /**< produce quad/fragment Z values? */ GLboolean need_w; /**< produce quad/fragment W values? */ +#if 0 /* Stipple derived state: */ GLubyte stipple_masks[16][16]; +#endif GLuint occlusion_counter; + GLuint line_stipple_counter; + /** Software quad rendering pipeline */ struct { struct quad_stage *polygon_stipple; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 6420fc8809..091a2343a9 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -706,6 +706,17 @@ plot(struct setup_stage *setup, GLint x, GLint y) } +/** + * Determine whether or not to emit a line fragment by checking + * line stipple pattern. + */ +static INLINE GLuint +stipple_test(GLint counter, GLushort pattern, GLint factor) +{ + GLint b = (counter / factor) & 0xf; + return (1 << b) & pattern; +} + /** * Do setup for line rasterization, then render the line. @@ -718,6 +729,7 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) const struct vertex_header *v0 = prim->v[0]; const struct vertex_header *v1 = prim->v[1]; struct setup_stage *setup = setup_stage( stage ); + struct softpipe_context *sp = setup->softpipe; GLint x0 = (GLint) v0->data[0][0]; GLint x1 = (GLint) v1->data[0][0]; @@ -763,7 +775,12 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) const GLint errorDec = error - dx; for (i = 0; i < dx; i++) { - plot(setup, x0, y0); + if (!sp->setup.line_stipple_enable || + stipple_test(sp->line_stipple_counter, + sp->setup.line_stipple_pattern, + sp->setup.line_stipple_factor + 1)) { + plot(setup, x0, y0); + } x0 += xstep; if (error < 0) { @@ -773,6 +790,8 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) error += errorDec; y0 += ystep; } + + sp->line_stipple_counter++; } } else { @@ -783,7 +802,12 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) const GLint errorDec = error - dy; for (i = 0; i < dy; i++) { - plot(setup, x0, y0); + if (!sp->setup.line_stipple_enable || + stipple_test(sp->line_stipple_counter, + sp->setup.line_stipple_pattern, + sp->setup.line_stipple_factor + 1)) { + plot(setup, x0, y0); + } y0 += ystep; @@ -794,6 +818,8 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) error += errorDec; x0 += xstep; } + + sp->line_stipple_counter++; } } @@ -984,6 +1010,13 @@ static void setup_end( struct draw_stage *stage ) } +static void reset_stipple_counter( struct draw_stage *stage ) +{ + struct setup_stage *setup = setup_stage(stage); + setup->softpipe->line_stipple_counter = 0; +} + + /** * Create a new primitive setup/render stage. */ @@ -998,6 +1031,7 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) setup->stage.line = setup_line; setup->stage.tri = setup_tri; setup->stage.end = setup_end; + setup->stage.reset_stipple_counter = reset_stipple_counter; setup->quad.coef = setup->coef; -- cgit v1.2.3 From 87555946478f58111013554e5af5c8844f944420 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 27 Jul 2007 12:50:38 +0200 Subject: Avoid unnecessary input attrib copy by aligning exec_machine attribs. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 79 +++++++++++++++++++++---------------- 1 file changed, 44 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 301bd13aa9..d4acf402ce 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -39,10 +39,20 @@ #include "sp_quad.h" #include "core/tgsi_core.h" +#if defined __GNUC__ +#define ALIGNED_ATTRIBS 1 +#else +#define ALIGNED_ATTRIBS 0 +#endif + struct exec_machine { const struct setup_coefficient *coef; /**< will point to quad->coef */ - GLfloat attr[FRAG_ATTRIB_MAX][4][QUAD_SIZE]; +#if ALIGNED_ATTRIBS + GLfloat attr[FRAG_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE] __attribute__(( aligned( 16 ) )); +#else + GLfloat attr[FRAG_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE]; +#endif }; @@ -162,7 +172,7 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) for (i = 0; i < NUM_CHANNELS; i++) cinterp(&exec, attr, i); break; - + case INTERP_LINEAR: for (i = 0; i < NUM_CHANNELS; i++) linterp(&exec, attr, i); @@ -180,56 +190,55 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) { struct tgsi_exec_machine machine; - struct tgsi_exec_vector inputs[FRAG_ATTRIB_MAX + 1]; struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; - struct tgsi_exec_vector *ainputs; struct tgsi_exec_vector *aoutputs; - GLuint i /*, total*/; + GLuint i; + +#if !ALIGNED_ATTRIBS + struct tgsi_exec_vector inputs[FRAG_ATTRIB_MAX + 1]; + struct tgsi_exec_vector *ainputs; +#endif #ifdef DEBUG memset(&machine, 0, sizeof(machine)); #endif - ainputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); + /* init machine state */ + tgsi_exec_machine_init( + &machine, + softpipe->fs.tokens ); + + /* Consts does not require 16 byte alignment. */ + machine.Consts = softpipe->fs.constants->constant; + aoutputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs ); + machine.Outputs = aoutputs; + + assert( sizeof( struct tgsi_exec_vector ) == sizeof( exec.attr[0] ) ); -#if 0 - for( i = total = 0; i < PIPE_ATTRIB_MAX; i++ ) { - GLuint attr; - - attr = softpipe->fp_attr_to_slot[i]; - if( attr || total == 0) { - assert( total < FRAG_ATTRIB_MAX ); - assert( attr < FRAG_ATTRIB_MAX ); - assert( sizeof( ainputs[0] ) == sizeof( exec.attr[0] ) ); - - memcpy( - &ainputs[total], - exec.attr[attr], - sizeof( ainputs[0] ) ); - total++; - } +#if ALIGNED_ATTRIBS + machine.Inputs = (struct tgsi_exec_vector *) exec.attr; + + for (i = 0; i < softpipe->nr_attrs; i++) { + /* Make sure fp_attr_to_slot[] is an identity transform. */ + assert( softpipe->fp_attr_to_slot[i] == i ); } #else + ainputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); + machine.Inputs = ainputs; + /* load input registers */ - /* XXX simpler than above, but might not be right... */ for (i = 0; i < softpipe->nr_attrs; i++) { + /* Make sure fp_attr_to_slot[] is an identity transform. */ + assert( softpipe->fp_attr_to_slot[i] == i ); + memcpy( - &ainputs[i], - exec.attr[i], - sizeof( ainputs[0] ) ); + &ainputs[i], + exec.attr[i], + sizeof( ainputs[0] ) ); } #endif - /* init machine state */ - tgsi_exec_machine_init( - &machine, - softpipe->fs.tokens ); - - machine.Inputs = ainputs; - machine.Outputs = aoutputs; - machine.Consts = softpipe->fs.constants->constant; /* XXX alignment? */ - /* run shader */ tgsi_exec_machine_run( &machine ); -- cgit v1.2.3 From e1a19781ddd855540aa65142891b23769e6da0f9 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 27 Jul 2007 09:06:02 -0600 Subject: init quad.coverage values to 1.0 in case line AA is enabled so that we see something --- src/mesa/pipe/softpipe/sp_prim_setup.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 091a2343a9..bb37bb9a8b 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -766,6 +766,13 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) setup->quad.x0 = setup->quad.y0 = -1; setup->quad.mask = 0x0; setup->quad.prim = PRIM_LINE; + /* XXX temporary: set coverage to 1.0 so the line appears + * if AA mode happens to be enabled. + */ + setup->quad.coverage[0] = + setup->quad.coverage[1] = + setup->quad.coverage[2] = + setup->quad.coverage[3] = 1.0; if (dx > dy) { /*** X-major line ***/ -- cgit v1.2.3 From df23ebbb9fe6aca7263371efc6ab6e6f4c1cf423 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 27 Jul 2007 09:06:15 -0600 Subject: s/SP_TILE_H/SP_QUAD_H/ --- src/mesa/pipe/softpipe/sp_quad.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 2ee53bdecd..72a9981977 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -28,8 +28,8 @@ /* Authors: Keith Whitwell */ -#ifndef SP_TILE_H -#define SP_TILE_H +#ifndef SP_QUAD_H +#define SP_QUAD_H struct softpipe_context; @@ -62,4 +62,4 @@ void sp_build_quad_pipeline(struct softpipe_context *sp); void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad); -#endif /* SP_TILE_H */ +#endif /* SP_QUAD_H */ -- cgit v1.2.3 From 6abdad28f6f6c69158b942fe420a55f12c71fb08 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 27 Jul 2007 09:27:56 -0600 Subject: check scissor state --- src/mesa/state_tracker/st_atom_setup.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index 6ee96f6178..3eac2588df 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -200,6 +200,10 @@ static void update_setup_state( struct st_context *st ) if (ctx->Multisample.Enabled) setup.multisample = 1; + /* _NEW_SCISSOR */ + if (ctx->Scissor.Enabled) + setup.scissor = 1; + if (memcmp(&setup, &st->state.setup, sizeof(setup)) != 0) { st->state.setup = setup; st->pipe->set_setup_state( st->pipe, &setup ); @@ -208,7 +212,7 @@ static void update_setup_state( struct st_context *st ) const struct st_tracked_state st_update_setup = { .dirty = { - .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | + .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | _NEW_SCISSOR | _NEW_POINT | _NEW_BUFFERS | _NEW_MULTISAMPLE), .st = 0, }, -- cgit v1.2.3 From 78ea9cee0d7fd424238ec25f9939d2c955694c0a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 27 Jul 2007 10:10:49 -0600 Subject: Implement point/line quad clipping. Not quite as efficient as it probably could be, but sufficient for now. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 89 ++++++++++++++++++++++++++-------- 1 file changed, 70 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index bb37bb9a8b..8f4fa41f44 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -43,18 +43,6 @@ #include "sp_prim_setup.h" - -/** - * Emit/render a quad. - * This passes the quad to the first stage of per-fragment operations. - */ -static INLINE void -quad_emit(struct softpipe_context *sp, struct quad_header *quad) -{ - sp->quad.first->run(sp->quad.first, quad); -} - - /** * Triangle edge info */ @@ -103,6 +91,8 @@ struct setup_stage { GLuint y_flags; GLuint mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ } span; + + struct pipe_scissor_state scissor; }; @@ -116,6 +106,46 @@ static inline struct setup_stage *setup_stage( struct draw_stage *stage ) } +/** + * Clip setup->quad against the scissor/surface bounds. + */ +static INLINE void +quad_clip(struct setup_stage *setup) +{ + if (setup->quad.x0 >= setup->scissor.maxx || + setup->quad.y0 >= setup->scissor.maxy || + setup->quad.x0 + 1 < setup->scissor.minx || + setup->quad.y0 + 1 < setup->scissor.miny) { + /* totally clipped */ + setup->quad.mask = 0x0; + return; + } + if (setup->quad.x0 < setup->scissor.minx) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + if (setup->quad.y0 < setup->scissor.miny) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + if (setup->quad.x0 == setup->scissor.maxx - 1) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + if (setup->quad.y0 == setup->scissor.maxy - 1) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); +} + + +/** + * Emit/render a quad. + * This passes the quad to the first stage of per-fragment operations. + */ +static INLINE void +quad_emit(struct setup_stage *setup) +{ + quad_clip(setup); + if (setup->quad.mask) { + struct softpipe_context *sp = setup->softpipe; + sp->quad.first->run(sp->quad.first, &setup->quad); + } +} + + /** * Given an X or Y coordinate, return the block/quad coordinate that it * belongs to. @@ -137,7 +167,7 @@ static void run_shader_block( struct setup_stage *setup, setup->quad.y0 = y; setup->quad.mask = mask; - quad_emit(setup->softpipe, &setup->quad); + quad_emit(setup); } @@ -694,8 +724,8 @@ plot(struct setup_stage *setup, GLint x, GLint y) { /* flush prev quad, start new quad */ - if (setup->quad.x0 != -1) - quad_emit(setup->softpipe, &setup->quad); + if (setup->quad.x0 != -1) + quad_emit(setup); setup->quad.x0 = quadX; setup->quad.y0 = quadY; @@ -832,7 +862,7 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) /* draw final quad */ if (setup->quad.mask) { - quad_emit(setup->softpipe, &setup->quad); + quad_emit(setup); } } @@ -889,7 +919,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) setup->quad.x0 = x - ix; setup->quad.y0 = y - iy; setup->quad.mask = (1 << ix) << (2 * iy); - quad_emit(setup->softpipe, &setup->quad); + quad_emit(setup); } else { const GLint ixmin = block((GLint) (x - halfSize)); @@ -951,7 +981,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) if (setup->quad.mask) { setup->quad.x0 = ix; setup->quad.y0 = iy; - quad_emit( setup->softpipe, &setup->quad ); + quad_emit(setup); } } } @@ -985,7 +1015,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) if (setup->quad.mask) { setup->quad.x0 = ix; setup->quad.y0 = iy; - quad_emit( setup->softpipe, &setup->quad ); + quad_emit(setup); } } } @@ -1001,6 +1031,27 @@ static void setup_begin( struct draw_stage *stage ) setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; + /* compute scissor/drawing bounds. + * XXX we should probably move this into the sp_state_derived.c file + * and only compute when scissor or setup state changes. + */ + { + const struct softpipe_context *sp = setup->softpipe; + const struct pipe_surface *surf = sp->cbuf; + if (sp->setup.scissor) { + setup->scissor.minx = MAX2(sp->scissor.minx, 0); + setup->scissor.miny = MAX2(sp->scissor.miny, 0); + setup->scissor.maxx = MIN2(sp->scissor.maxx, surf->width - 1); + setup->scissor.maxy = MIN2(sp->scissor.maxy, surf->height - 1); + } + else { + setup->scissor.minx = 0; + setup->scissor.miny = 0; + setup->scissor.maxx = surf->width - 1; + setup->scissor.maxy = surf->height - 1; + } + } + /* * XXX this is where we might map() the renderbuffers to begin * s/w rendering. -- cgit v1.2.3 From a0c1d591ab8d19377c7ab784a0472fac00f9fcd3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 27 Jul 2007 10:21:34 -0600 Subject: Maintain cliprect (scissor) info in sp_state_derived.c. The cliprect depends on the scissor rect (if enabled), otherwise the drawing surface bounds. --- src/mesa/pipe/softpipe/sp_context.h | 3 +++ src/mesa/pipe/softpipe/sp_prim_setup.c | 40 +++++++------------------------ src/mesa/pipe/softpipe/sp_state_derived.c | 40 +++++++++++++++++++++++++++++++ 3 files changed, 52 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 8871b45d4f..3c379c9091 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -117,6 +117,9 @@ struct softpipe_context { GLubyte stipple_masks[16][16]; #endif + /** Derived from scissor and surface bounds: */ + struct pipe_scissor_state cliprect; + GLuint occlusion_counter; GLuint line_stipple_counter; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 8f4fa41f44..d18f4f3414 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -91,8 +91,6 @@ struct setup_stage { GLuint y_flags; GLuint mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ } span; - - struct pipe_scissor_state scissor; }; @@ -112,21 +110,22 @@ static inline struct setup_stage *setup_stage( struct draw_stage *stage ) static INLINE void quad_clip(struct setup_stage *setup) { - if (setup->quad.x0 >= setup->scissor.maxx || - setup->quad.y0 >= setup->scissor.maxy || - setup->quad.x0 + 1 < setup->scissor.minx || - setup->quad.y0 + 1 < setup->scissor.miny) { + const struct softpipe_context *sp = setup->softpipe; + if (setup->quad.x0 >= sp->cliprect.maxx || + setup->quad.y0 >= sp->cliprect.maxy || + setup->quad.x0 + 1 < sp->cliprect.minx || + setup->quad.y0 + 1 < sp->cliprect.miny) { /* totally clipped */ setup->quad.mask = 0x0; return; } - if (setup->quad.x0 < setup->scissor.minx) + if (setup->quad.x0 < sp->cliprect.minx) setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup->quad.y0 < setup->scissor.miny) + if (setup->quad.y0 < sp->cliprect.miny) setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); - if (setup->quad.x0 == setup->scissor.maxx - 1) + if (setup->quad.x0 == sp->cliprect.maxx - 1) setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup->quad.y0 == setup->scissor.maxy - 1) + if (setup->quad.y0 == sp->cliprect.maxy - 1) setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); } @@ -1031,27 +1030,6 @@ static void setup_begin( struct draw_stage *stage ) setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; - /* compute scissor/drawing bounds. - * XXX we should probably move this into the sp_state_derived.c file - * and only compute when scissor or setup state changes. - */ - { - const struct softpipe_context *sp = setup->softpipe; - const struct pipe_surface *surf = sp->cbuf; - if (sp->setup.scissor) { - setup->scissor.minx = MAX2(sp->scissor.minx, 0); - setup->scissor.miny = MAX2(sp->scissor.miny, 0); - setup->scissor.maxx = MIN2(sp->scissor.maxx, surf->width - 1); - setup->scissor.maxy = MIN2(sp->scissor.maxy, surf->height - 1); - } - else { - setup->scissor.minx = 0; - setup->scissor.miny = 0; - setup->scissor.maxx = surf->width - 1; - setup->scissor.maxy = surf->height - 1; - } - } - /* * XXX this is where we might map() the renderbuffers to begin * s/w rendering. diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 26083c2962..9572890afe 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -163,6 +163,41 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) } +/** + * Recompute cliprect from scissor bounds, scissor enable and surface size. + */ +static void +compute_cliprect(struct softpipe_context *sp) +{ + GLint surfWidth, surfHeight; + + if (sp->framebuffer.num_cbufs > 0) { + surfWidth = sp->framebuffer.cbufs[0]->width; + surfHeight = sp->framebuffer.cbufs[0]->height; + } + else { + /* no surface? */ + surfWidth = sp->scissor.maxx; + surfHeight = sp->scissor.maxy; + } + + if (sp->setup.scissor) { + /* clip to scissor rect */ + sp->cliprect.minx = MAX2(sp->scissor.minx, 0); + sp->cliprect.miny = MAX2(sp->scissor.miny, 0); + sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth - 1); + sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight - 1); + } + else { + /* clip to surface bounds */ + sp->cliprect.minx = 0; + sp->cliprect.miny = 0; + sp->cliprect.maxx = surfWidth - 1; + sp->cliprect.maxy = surfHeight - 1; + } +} + + /* Hopefully this will remain quite simple, otherwise need to pull in * something like the state tracker mechanism. */ @@ -171,6 +206,11 @@ void softpipe_update_derived( struct softpipe_context *softpipe ) if (softpipe->dirty & (SP_NEW_SETUP | SP_NEW_FS)) calculate_vertex_layout( softpipe ); + if (softpipe->dirty & (SP_NEW_SCISSOR | + SP_NEW_STENCIL | + SP_NEW_FRAMEBUFFER)) + compute_cliprect(softpipe); + if (softpipe->dirty & (SP_NEW_BLEND | SP_NEW_DEPTH_TEST | SP_NEW_ALPHA_TEST | -- cgit v1.2.3 From 05bde092f39e16463dfbbe038f22f17da4346527 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 27 Jul 2007 10:25:11 -0600 Subject: remove obsolete comments --- src/mesa/pipe/softpipe/sp_prim_setup.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index d18f4f3414..63568404dc 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -750,7 +750,6 @@ stipple_test(GLint counter, GLushort pattern, GLint factor) /** * Do setup for line rasterization, then render the line. * XXX single-pixel width, no stipple, etc - * XXX no scissoring yet. */ static void setup_line(struct draw_stage *stage, struct prim_header *prim) @@ -909,8 +908,6 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) setup->quad.prim = PRIM_POINT; - /* XXX need to clip against scissor bounds too */ - if (halfSize <= 0.5 && !round) { /* special case for 1-pixel points */ const GLint ix = ((GLint) x) & 1; -- cgit v1.2.3 From 05ba76757b7349f0c3de2d9ce42c543fa7e931c2 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 27 Jul 2007 10:50:35 -0600 Subject: Clip triangles against softpipe->cliprect which includes scissor and surface bounds. This prevents rendering out of bounds when the viewport is partially outside the surface bounds. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 72 ++++++++++++++----------------- src/mesa/pipe/softpipe/sp_state_derived.c | 8 ++-- 2 files changed, 36 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 63568404dc..d07bd5381d 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -110,28 +110,31 @@ static inline struct setup_stage *setup_stage( struct draw_stage *stage ) static INLINE void quad_clip(struct setup_stage *setup) { - const struct softpipe_context *sp = setup->softpipe; - if (setup->quad.x0 >= sp->cliprect.maxx || - setup->quad.y0 >= sp->cliprect.maxy || - setup->quad.x0 + 1 < sp->cliprect.minx || - setup->quad.y0 + 1 < sp->cliprect.miny) { + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + if (setup->quad.x0 >= cliprect->maxx || + setup->quad.y0 >= cliprect->maxy || + setup->quad.x0 + 1 < cliprect->minx || + setup->quad.y0 + 1 < cliprect->miny) { /* totally clipped */ setup->quad.mask = 0x0; return; } - if (setup->quad.x0 < sp->cliprect.minx) + if (setup->quad.x0 < cliprect->minx) setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup->quad.y0 < sp->cliprect.miny) + if (setup->quad.y0 < cliprect->miny) setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); - if (setup->quad.x0 == sp->cliprect.maxx - 1) + if (setup->quad.x0 == cliprect->maxx - 1) setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup->quad.y0 == sp->cliprect.maxy - 1) + if (setup->quad.y0 == cliprect->maxy - 1) setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); } /** * Emit/render a quad. + * Called during point/line rendering. For triangles, we call + * run_shader_block() which doesn't do clipping (since clipping is + * done at a higher level for tris). * This passes the quad to the first stage of per-fragment operations. */ static INLINE void @@ -162,11 +165,11 @@ static inline GLint block( GLint x ) static void run_shader_block( struct setup_stage *setup, GLint x, GLint y, GLuint mask ) { + struct softpipe_context *sp = setup->softpipe; setup->quad.x0 = x; setup->quad.y0 = y; setup->quad.mask = mask; - - quad_emit(setup); + sp->quad.first->run(sp->quad.first, &setup->quad); } @@ -501,38 +504,31 @@ static void setup_tri_edges( struct setup_stage *setup ) /** * Render the upper or lower half of a triangle. - * Scissoring is applied here too. + * Scissoring/cliprect is applied here too. */ static void subtriangle( struct setup_stage *setup, struct edge *eleft, struct edge *eright, GLuint lines ) { + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; GLint y, start_y, finish_y; GLint sy = (GLint)eleft->sy; assert((GLint)eleft->sy == (GLint) eright->sy); - assert((GLint)eleft->sy >= 0); /* catch bug in x64? */ - /* scissor y: - */ - if (setup->softpipe->setup.scissor) { - start_y = sy; - finish_y = start_y + lines; + /* clip top/bottom */ + start_y = sy; + finish_y = sy + lines; - if (start_y < setup->softpipe->scissor.miny) - start_y = setup->softpipe->scissor.miny; + if (start_y < cliprect->miny) + start_y = cliprect->miny; - if (finish_y > setup->softpipe->scissor.maxy) - finish_y = setup->softpipe->scissor.maxy; + if (finish_y > cliprect->maxy) + finish_y = cliprect->maxy; - start_y -= sy; - finish_y -= sy; - } - else { - start_y = 0; - finish_y = lines; - } + start_y -= sy; + finish_y -= sy; /* _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); @@ -549,15 +545,11 @@ static void subtriangle( struct setup_stage *setup, GLint left = (GLint)(eleft->sx + y * eleft->dxdy); GLint right = (GLint)(eright->sx + y * eright->dxdy); - /* scissor x: - */ - if (setup->softpipe->setup.scissor) { - if (left < setup->softpipe->scissor.minx) - left = setup->softpipe->scissor.minx; - - if (right > setup->softpipe->scissor.maxx) - right = setup->softpipe->scissor.maxx; - } + /* clip left/right */ + if (left < cliprect->minx) + left = cliprect->minx; + if (right > cliprect->maxx) + right = cliprect->maxx; if (left < right) { GLint _y = sy+y; @@ -875,8 +867,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); /*XXX this should be a vertex attrib! */ - GLfloat halfSize = 0.5 * setup->softpipe->setup.point_size; - GLboolean round = setup->softpipe->setup.point_smooth; + const GLfloat halfSize = 0.5 * setup->softpipe->setup.point_size; + const GLboolean round = setup->softpipe->setup.point_smooth; const struct vertex_header *v0 = prim->v[0]; const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 9572890afe..e1faaed93c 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -185,15 +185,15 @@ compute_cliprect(struct softpipe_context *sp) /* clip to scissor rect */ sp->cliprect.minx = MAX2(sp->scissor.minx, 0); sp->cliprect.miny = MAX2(sp->scissor.miny, 0); - sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth - 1); - sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight - 1); + sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth); + sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight); } else { /* clip to surface bounds */ sp->cliprect.minx = 0; sp->cliprect.miny = 0; - sp->cliprect.maxx = surfWidth - 1; - sp->cliprect.maxy = surfHeight - 1; + sp->cliprect.maxx = surfWidth; + sp->cliprect.maxy = surfHeight; } } -- cgit v1.2.3 From 75b8a396e5b29e2ee86ab3a5c9590e6db5d28155 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 27 Jul 2007 11:02:54 -0600 Subject: renaming, comments, clean-up --- src/mesa/pipe/softpipe/sp_prim_setup.c | 62 +++++++++++++++------------------- 1 file changed, 27 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index d07bd5381d..57b7d73415 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -47,12 +47,11 @@ * Triangle edge info */ struct edge { - GLfloat dx; /* X(v1) - X(v0), used only during setup */ - GLfloat dy; /* Y(v1) - Y(v0), used only during setup */ - GLfloat dxdy; /* dx/dy */ - GLfloat sx; /* first sample point x coord */ - GLfloat sy; - GLint lines; /* number of lines on this edge */ + GLfloat dx; /**< X(v1) - X(v0), used only during setup */ + GLfloat dy; /**< Y(v1) - Y(v0), used only during setup */ + GLfloat dxdy; /**< dx/dy */ + GLfloat sx, sy; /**< first sample point coord */ + GLint lines; /**< number of lines on this edge */ }; @@ -63,7 +62,6 @@ struct edge { struct setup_stage { struct draw_stage stage; /**< This must be first (base class) */ - /*XXX NEW */ struct softpipe_context *softpipe; /* Vertices are just an array of floats making up each attribute in @@ -98,7 +96,7 @@ struct setup_stage { /** * Basically a cast wrapper. */ -static inline struct setup_stage *setup_stage( struct draw_stage *stage ) +static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) { return (struct setup_stage *)stage; } @@ -131,14 +129,10 @@ quad_clip(struct setup_stage *setup) /** - * Emit/render a quad. - * Called during point/line rendering. For triangles, we call - * run_shader_block() which doesn't do clipping (since clipping is - * done at a higher level for tris). - * This passes the quad to the first stage of per-fragment operations. + * Emit a quad (pass to next stage) with clipping. */ static INLINE void -quad_emit(struct setup_stage *setup) +clip_emit_quad(struct setup_stage *setup) { quad_clip(setup); if (setup->quad.mask) { @@ -149,27 +143,26 @@ quad_emit(struct setup_stage *setup) /** - * Given an X or Y coordinate, return the block/quad coordinate that it - * belongs to. + * Emit a quad (pass to next stage). No clipping is done. */ -static inline GLint block( GLint x ) +static INLINE void +emit_quad( struct setup_stage *setup, GLint x, GLint y, GLuint mask ) { - return x & ~1; + struct softpipe_context *sp = setup->softpipe; + setup->quad.x0 = x; + setup->quad.y0 = y; + setup->quad.mask = mask; + sp->quad.first->run(sp->quad.first, &setup->quad); } - /** - * Run shader on a quad/block. + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. */ -static void run_shader_block( struct setup_stage *setup, - GLint x, GLint y, GLuint mask ) +static INLINE GLint block( GLint x ) { - struct softpipe_context *sp = setup->softpipe; - setup->quad.x0 = x; - setup->quad.y0 = y; - setup->quad.mask = mask; - sp->quad.first->run(sp->quad.first, &setup->quad); + return x & ~1; } @@ -232,9 +225,8 @@ static void flush_spans( struct setup_stage *setup ) for (x = block(minleft); x <= block(maxright); ) { - run_shader_block( setup, x, - setup->span.y, - calculate_mask( setup, x ) ); + emit_quad( setup, x, setup->span.y, + calculate_mask( setup, x ) ); x += 2; } @@ -716,7 +708,7 @@ plot(struct setup_stage *setup, GLint x, GLint y) /* flush prev quad, start new quad */ if (setup->quad.x0 != -1) - quad_emit(setup); + clip_emit_quad(setup); setup->quad.x0 = quadX; setup->quad.y0 = quadY; @@ -852,7 +844,7 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) /* draw final quad */ if (setup->quad.mask) { - quad_emit(setup); + clip_emit_quad(setup); } } @@ -907,7 +899,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) setup->quad.x0 = x - ix; setup->quad.y0 = y - iy; setup->quad.mask = (1 << ix) << (2 * iy); - quad_emit(setup); + clip_emit_quad(setup); } else { const GLint ixmin = block((GLint) (x - halfSize)); @@ -969,7 +961,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) if (setup->quad.mask) { setup->quad.x0 = ix; setup->quad.y0 = iy; - quad_emit(setup); + clip_emit_quad(setup); } } } @@ -1003,7 +995,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) if (setup->quad.mask) { setup->quad.x0 = ix; setup->quad.y0 = iy; - quad_emit(setup); + clip_emit_quad(setup); } } } -- cgit v1.2.3 From 9d2d34b5eabfb7855081ad56fa00b3e0f6b8e631 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Sun, 29 Jul 2007 18:04:28 +0200 Subject: Fix crashes when the frag prog can't be handled in hardware (#11131) Must not change to/from swrast after Render.Start or bad things will happen. (Driver will still somewhat incorrectly report an implementation error, and apps can't really figure out if a prog is natively supported as validation is later - could try doing it earlier to give some hint at least, even though native status may still change later due to fog etc.) --- src/mesa/drivers/dri/i915tex/i830_vtbl.c | 6 ++++++ src/mesa/drivers/dri/i915tex/i915_vtbl.c | 8 +++++++- src/mesa/drivers/dri/i915tex/intel_context.h | 1 + src/mesa/drivers/dri/i915tex/intel_render.c | 2 ++ 4 files changed, 16 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/i830_vtbl.c b/src/mesa/drivers/dri/i915tex/i830_vtbl.c index 03d4788be1..20c6900025 100644 --- a/src/mesa/drivers/dri/i915tex/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915tex/i830_vtbl.c @@ -60,6 +60,11 @@ do { \ #define VRTX_TEX_SET_FMT(n, x) ((x)<<((n)*2)) #define TEXBIND_SET(n, x) ((x)<<((n)*4)) +static void +i830_render_prevalidate(struct intel_context *intel) +{ +} + static void i830_render_start(struct intel_context *intel) { @@ -638,5 +643,6 @@ i830InitVtbl(struct i830_context *i830) i830->intel.vtbl.update_texture_state = i830UpdateTextureState; i830->intel.vtbl.flush_cmd = i830_flush_cmd; i830->intel.vtbl.render_start = i830_render_start; + i830->intel.vtbl.render_prevalidate = i830_render_prevalidate; i830->intel.vtbl.assert_not_dirty = i830_assert_not_dirty; } diff --git a/src/mesa/drivers/dri/i915tex/i915_vtbl.c b/src/mesa/drivers/dri/i915tex/i915_vtbl.c index 51aff68840..13ca9aee6b 100644 --- a/src/mesa/drivers/dri/i915tex/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915tex/i915_vtbl.c @@ -44,13 +44,18 @@ #include "i915_context.h" static void -i915_render_start(struct intel_context *intel) +i915_render_prevalidate(struct intel_context *intel) { struct i915_context *i915 = i915_context(&intel->ctx); i915ValidateFragmentProgram(i915); } +static void +i915_render_start(struct intel_context *intel) +{ +} + static void i915_reduced_primitive_state(struct intel_context *intel, GLenum rprim) @@ -547,6 +552,7 @@ i915InitVtbl(struct i915_context *i915) i915->intel.vtbl.lost_hardware = i915_lost_hardware; i915->intel.vtbl.reduced_primitive_state = i915_reduced_primitive_state; i915->intel.vtbl.render_start = i915_render_start; + i915->intel.vtbl.render_prevalidate = i915_render_prevalidate; i915->intel.vtbl.set_draw_region = i915_set_draw_region; i915->intel.vtbl.update_texture_state = i915UpdateTextureState; i915->intel.vtbl.flush_cmd = i915_flush_cmd; diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h index 535a93effa..ee5be45533 100644 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ b/src/mesa/drivers/dri/i915tex/intel_context.h @@ -131,6 +131,7 @@ struct intel_context void (*update_texture_state) (struct intel_context * intel); void (*render_start) (struct intel_context * intel); + void (*render_prevalidate) (struct intel_context * intel); void (*set_draw_region) (struct intel_context * intel, struct intel_region * draw_region, struct intel_region * depth_region); diff --git a/src/mesa/drivers/dri/i915tex/intel_render.c b/src/mesa/drivers/dri/i915tex/intel_render.c index f9fa55051e..c8b6d308d9 100644 --- a/src/mesa/drivers/dri/i915tex/intel_render.c +++ b/src/mesa/drivers/dri/i915tex/intel_render.c @@ -202,6 +202,8 @@ intel_run_render(GLcontext * ctx, struct tnl_pipeline_stage *stage) struct vertex_buffer *VB = &tnl->vb; GLuint i; + intel->vtbl.render_prevalidate( intel ); + /* Don't handle clipping or indexed vertices. */ if (intel->RenderIndex != 0 || -- cgit v1.2.3 From 7aefdd518578687f961bb512a2b408857f0b9824 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Sun, 29 Jul 2007 19:40:50 +0200 Subject: fix range reduction for sin/cos in i915tex (#11609) --- src/mesa/drivers/dri/i915tex/i915_context.h | 6 ------ src/mesa/drivers/dri/i915tex/i915_fragprog.c | 8 ++++---- 2 files changed, 4 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/i915_context.h b/src/mesa/drivers/dri/i915tex/i915_context.h index d2713e88f9..3a41d66c14 100644 --- a/src/mesa/drivers/dri/i915tex/i915_context.h +++ b/src/mesa/drivers/dri/i915tex/i915_context.h @@ -308,12 +308,6 @@ extern GLboolean i915CreateContext(const __GLcontextModes * mesaVis, void *sharedContextPrivate); -/*====================================================================== - * i915_texprog.c - */ -extern void i915ValidateTextureProgram(struct i915_context *i915); - - /*====================================================================== * i915_debug.c */ diff --git a/src/mesa/drivers/dri/i915tex/i915_fragprog.c b/src/mesa/drivers/dri/i915tex/i915_fragprog.c index a4b22a0c32..95ec50490a 100644 --- a/src/mesa/drivers/dri/i915tex/i915_fragprog.c +++ b/src/mesa/drivers/dri/i915tex/i915_fragprog.c @@ -320,7 +320,7 @@ upload_program(struct i915_fragment_program *p) i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - src0, i915_emit_const1f(p, 1.0 / (M_PI * 2)), 0); + src0, i915_emit_const1f(p, 1.0 / (M_PI)), 0); i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); @@ -329,7 +329,7 @@ upload_program(struct i915_fragment_program *p) i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - tmp, i915_emit_const1f(p, (M_PI * 2)), 0); + tmp, i915_emit_const1f(p, (M_PI)), 0); /* * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 @@ -642,7 +642,7 @@ upload_program(struct i915_fragment_program *p) i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - src0, i915_emit_const1f(p, 1.0 / (M_PI * 2)), 0); + src0, i915_emit_const1f(p, 1.0 / (M_PI)), 0); i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); @@ -651,7 +651,7 @@ upload_program(struct i915_fragment_program *p) i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - tmp, i915_emit_const1f(p, (M_PI * 2)), 0); + tmp, i915_emit_const1f(p, (M_PI)), 0); /* * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 -- cgit v1.2.3 From f38d26fd83a2193c6cba9ce72d912f579b25f5fe Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 13:09:53 -0600 Subject: added map/unmap() stubs --- src/mesa/drivers/x11/xm_surface.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 58081a36a0..9d6db2b5ce 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -69,6 +69,20 @@ xmesa_surface(struct softpipe_surface *sps) } +static void * +map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) +{ + /* no-op */ +} + + +static void +unmap_surface_buffer(struct pipe_buffer *pb) +{ + /* no-op */ +} + + /** * quad reading/writing * These functions are just wrappers around the existing renderbuffer @@ -199,6 +213,9 @@ create_surface(XMesaContext xmctx, struct xmesa_renderbuffer *xrb) xmsurf->sps.write_quad_ub = write_quad_ub; xmsurf->sps.write_mono_row_ub = write_mono_row_ub; + xmsurf->sps.surface.buffer.map = map_surface_buffer; + xmsurf->sps.surface.buffer.unmap = unmap_surface_buffer; + #if 0 if (xrb->ximage) { xmsurf->sps.surface.ptr = (GLubyte *) xrb->ximage->data; -- cgit v1.2.3 From a548d3cb71116f93d6cd9bc818c57b70ac9d9504 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 13:10:12 -0600 Subject: call st_invalidate_state() --- src/mesa/drivers/dri/i915tex/i915_context.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/i915_context.c b/src/mesa/drivers/dri/i915tex/i915_context.c index 49e30141e4..b6d004b258 100644 --- a/src/mesa/drivers/dri/i915tex/i915_context.c +++ b/src/mesa/drivers/dri/i915tex/i915_context.c @@ -71,6 +71,8 @@ i915InvalidateState(GLcontext * ctx, GLuint new_state) _tnl_invalidate_vertex_state(ctx, new_state); intel_context(ctx)->NewGLState |= new_state; + st_invalidate_state(ctx, new_state); + /* Todo: gather state values under which tracked parameters become * invalidated, add callbacks for things like * ProgramLocalParameters, etc. -- cgit v1.2.3 From 737f6e7eb1f90133c81cff83444191df84754e77 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 13:10:52 -0600 Subject: disable ProgramStringNotify assertion --- src/mesa/state_tracker/st_cb_program.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index fa9eaf446e..6da2aeb2f2 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -153,7 +153,9 @@ void st_init_cb_program( struct st_context *st ) st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; st->ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; +#if 0 assert(functions->ProgramStringNotify == _tnl_program_string); +#endif functions->BindProgram = st_bind_program; functions->NewProgram = st_new_program; functions->DeleteProgram = st_delete_program; -- cgit v1.2.3 From 1c8905790be301da6d18d2397fd3bbe4a47f987c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 13:11:09 -0600 Subject: remove old comments --- src/mesa/pipe/softpipe/sp_prim_setup.c | 9 --------- 1 file changed, 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index d07bd5381d..7b09d7940e 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -1018,20 +1018,11 @@ static void setup_begin( struct draw_stage *stage ) struct setup_stage *setup = setup_stage(stage); setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; - - /* - * XXX this is where we might map() the renderbuffers to begin - * s/w rendering. - */ } static void setup_end( struct draw_stage *stage ) { - /* - * XXX this is where we might unmap() the renderbuffers after - * s/w rendering. - */ } -- cgit v1.2.3 From f1569bf0d5fcfd4430e13d6d1bd9c60fb44e9535 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 13:11:27 -0600 Subject: map/unmap surfaces before/after rendering --- src/mesa/pipe/softpipe/sp_context.c | 31 +++++++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 3bb04a6835..6b44fabfa4 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -32,12 +32,40 @@ #include "main/imports.h" #include "main/macros.h" #include "pipe/draw/draw_context.h" +#include "pipe/p_defines.h" #include "sp_context.h" #include "sp_clear.h" #include "sp_state.h" +#include "sp_surface.h" #include "sp_prim_setup.h" +static void map_surfaces(struct softpipe_context *sp) +{ + GLuint i; + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) { + struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); + struct pipe_buffer *buf = &sps->surface.buffer; + buf->map(buf, PIPE_MAP_READ_WRITE); + } + /* XXX depth & stencil bufs */ +} + + +static void unmap_surfaces(struct softpipe_context *sp) +{ + GLuint i; + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) { + struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); + struct pipe_buffer *buf = &sps->surface.buffer; + buf->unmap(buf); + } + /* XXX depth & stencil bufs */ +} + + static void softpipe_destroy( struct pipe_context *pipe ) { struct softpipe_context *softpipe = softpipe_context( pipe ); @@ -56,7 +84,10 @@ static void softpipe_draw_vb( struct pipe_context *pipe, if (softpipe->dirty) softpipe_update_derived( softpipe ); + /* XXX move mapping/unmapping to higher/coarser level? */ + map_surfaces(softpipe); draw_vb( softpipe->draw, VB ); + unmap_surfaces(softpipe); } -- cgit v1.2.3 From 3a35ce336458352653329426c550bfce1ffc3f66 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 13:11:52 -0600 Subject: implement surfaces for softpipe rendering --- src/mesa/drivers/dri/i915tex/Makefile | 1 + src/mesa/drivers/dri/i915tex/intel_fbo.h | 18 +++ src/mesa/drivers/dri/i915tex/intel_surface.c | 163 +++++++++++++++++++++++++++ 3 files changed, 182 insertions(+) create mode 100644 src/mesa/drivers/dri/i915tex/intel_surface.c (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/Makefile b/src/mesa/drivers/dri/i915tex/Makefile index b218929dce..827acc0c46 100644 --- a/src/mesa/drivers/dri/i915tex/Makefile +++ b/src/mesa/drivers/dri/i915tex/Makefile @@ -48,6 +48,7 @@ DRIVER_SOURCES = \ intel_screen.c \ intel_span.c \ intel_state.c \ + intel_surface.c \ intel_tris.c \ intel_fbo.c \ intel_depthstencil.c \ diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.h b/src/mesa/drivers/dri/i915tex/intel_fbo.h index 963f5e706f..1642ce774f 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.h +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.h @@ -29,9 +29,25 @@ #define INTEL_FBO_H +#include "pipe/p_state.h" +#include "pipe/softpipe/sp_surface.h" + + struct intel_context; struct intel_region; + +/** + * Intel "pipe" surface. This is kind of a temporary thing as + * renderbuffers and surfaces should eventually become one. + */ +struct intel_surface +{ + struct softpipe_surface surface; /**< base class */ + struct intel_renderbuffer *rb; /**< ptr back to matching renderbuffer */ +}; + + /** * Intel framebuffer, derived from gl_framebuffer. */ @@ -82,6 +98,8 @@ struct intel_renderbuffer GLuint pf_pending; /**< sequence number of pending flip */ GLuint vbl_pending; /**< vblank sequence number of pending flip */ + + struct intel_surface *surface; }; diff --git a/src/mesa/drivers/dri/i915tex/intel_surface.c b/src/mesa/drivers/dri/i915tex/intel_surface.c new file mode 100644 index 0000000000..29b640e58e --- /dev/null +++ b/src/mesa/drivers/dri/i915tex/intel_surface.c @@ -0,0 +1,163 @@ +#include "glheader.h" +#include "context.h" +#include "framebuffer.h" +#include "renderbuffer.h" +#include "utils.h" +#include "main/macros.h" + + +#include "intel_screen.h" + +#include "intel_context.h" +#include "intel_buffers.h" +#include "intel_regions.h" +#include "intel_span.h" +#include "intel_fbo.h" + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/softpipe/sp_surface.h" + + +/* + * XXX a lof of this is a temporary kludge + */ + +extern void +intel_map_unmap_buffers(struct intel_context *intel, GLboolean map); + + + +static void +read_quad_f_swz(struct softpipe_surface *gs, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) +{ + + +} + + +static void +write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) +{ + struct intel_surface *is = (struct intel_surface *) sps; + struct intel_renderbuffer *irb = is->rb; + const GLfloat *src = (const GLfloat *) rrrr; + GLubyte *dst = (GLubyte *) irb->region->map + + (y * irb->region->pitch + x) * irb->region->cpp; + GLubyte temp[16]; + GLuint i, j; + + for (i = 0; i < 4; i++) { + for (j = 0; j < 4; j++) { + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + i], src[i * 4 + j]); + } + } + + printf("intel_surface::write_quad\n"); + + memcpy(dst, temp, 8); + memcpy(dst + irb->region->pitch * irb->region->cpp, temp + 8, 8); +} + + + +static void * +map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) +{ + struct intel_surface *is = (struct intel_surface *) pb; + struct intel_renderbuffer *irb = is->rb; + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); + + assert(access_mode == PIPE_MAP_READ_WRITE); + + intelFinish(&intel->ctx); + + /*LOCK_HARDWARE(intel);*/ + + if (irb->region) { + intel_region_map(intel->intelScreen, irb->region); + } + pb->ptr = irb->region->map; + + return pb->ptr; +} + + +static void +unmap_surface_buffer(struct pipe_buffer *pb) +{ + struct intel_surface *is = (struct intel_surface *) pb; + struct intel_renderbuffer *irb = is->rb; + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); + + if (irb->region) { + intel_region_unmap(intel->intelScreen, irb->region); + } + pb->ptr = NULL; + + /*UNLOCK_HARDWARE(intel);*/ +} + + +struct pipe_surface * +xmesa_get_color_surface(GLcontext *ctx, GLuint i) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_framebuffer *intel_fb; + struct intel_renderbuffer *intel_rb; + + intel_fb = (struct intel_framebuffer *) ctx->DrawBuffer; + intel_rb = intel_fb->color_rb[1]; + + if (!intel_rb->surface) { + /* create surface and attach to intel_rb */ + struct intel_surface *is; + is = CALLOC_STRUCT(intel_surface); + if (is) { + is->surface.surface.width = intel_rb->Base.Width; + is->surface.surface.height = intel_rb->Base.Height; + + is->surface.read_quad_f_swz = read_quad_f_swz; + is->surface.write_quad_f_swz = write_quad_f_swz; + + is->surface.surface.buffer.map = map_surface_buffer; + is->surface.surface.buffer.unmap = unmap_surface_buffer; + + is->rb = intel_rb; + } + intel_rb->surface = is; + } + else { + /* update surface size */ + struct intel_surface *is = intel_rb->surface; + is->surface.surface.width = intel_rb->Base.Width; + is->surface.surface.height = intel_rb->Base.Height; + /* sanity check */ + assert(is->surface.surface.buffer.map == map_surface_buffer); + } + + return &intel_rb->surface->surface.surface; +} + + +struct pipe_surface * +xmesa_get_z_surface(GLcontext *ctx) +{ + /* XXX fix */ + return NULL; +} + + +struct pipe_surface * +xmesa_get_stencil_surface(GLcontext *ctx) +{ + /* XXX fix */ + return NULL; +} + + + -- cgit v1.2.3 From 09574bee850edec46d967b7cdbc2073f92b8223d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 14:00:27 -0600 Subject: implement read_quad_f_swz() --- src/mesa/drivers/dri/i915tex/intel_surface.c | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_surface.c b/src/mesa/drivers/dri/i915tex/intel_surface.c index 29b640e58e..3be902cf9c 100644 --- a/src/mesa/drivers/dri/i915tex/intel_surface.c +++ b/src/mesa/drivers/dri/i915tex/intel_surface.c @@ -23,17 +23,28 @@ * XXX a lof of this is a temporary kludge */ -extern void -intel_map_unmap_buffers(struct intel_context *intel, GLboolean map); - static void -read_quad_f_swz(struct softpipe_surface *gs, GLint x, GLint y, +read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rrrr)[QUAD_SIZE]) { + struct intel_surface *is = (struct intel_surface *) sps; + struct intel_renderbuffer *irb = is->rb; + const GLubyte *src = (const GLubyte *) irb->region->map + + (y * irb->region->pitch + x) * irb->region->cpp; + GLfloat *dst = (GLfloat *) rrrr; + GLubyte temp[16]; + GLuint i, j; + memcpy(temp, src, 8); + memcpy(temp + 8, src + irb->region->pitch * irb->region->cpp, 8); + for (i = 0; i < 4; i++) { + for (j = 0; j < 4; j++) { + dst[j * 4 + i] = UBYTE_TO_FLOAT(temp[i * 4 + j]); + } + } } @@ -55,8 +66,6 @@ write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, } } - printf("intel_surface::write_quad\n"); - memcpy(dst, temp, 8); memcpy(dst + irb->region->pitch * irb->region->cpp, temp + 8, 8); } -- cgit v1.2.3 From 6558af67d380f2855b112ea3ce4dded9215c7cf7 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 17:15:50 -0600 Subject: compute scale for Z24 buffers --- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index d47c4c42b8..a26bd51d84 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -59,6 +59,8 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) */ if (sps->surface.format == PIPE_FORMAT_U_Z16) scale = 65535.0; + else if (sps->surface.format == PIPE_FORMAT_Z24_S8) + scale = (float) ((1 << 24) - 1); else assert(0); /* XXX fix this someday */ -- cgit v1.2.3 From 4576d754c98e3fb5d413e294d48fb70a893defcf Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 17:17:44 -0600 Subject: Lots of improvements to the surface-related code. Z testing now works with i915 driver. Add gl_renderbuffer::surface pointer (and reverse pointer). Remove intel_surface and xmesa_surface types - no longer used. --- src/mesa/drivers/dri/i915tex/intel_fbo.c | 18 ++ src/mesa/drivers/dri/i915tex/intel_fbo.h | 14 +- src/mesa/drivers/dri/i915tex/intel_surface.c | 231 ++++++++++++++------- src/mesa/drivers/x11/xm_buffer.c | 8 + src/mesa/drivers/x11/xm_surface.c | 288 +++++---------------------- src/mesa/drivers/x11/xmesaP.h | 14 +- src/mesa/main/mtypes.h | 3 + src/mesa/main/renderbuffer.c | 22 +- src/mesa/pipe/p_state.h | 16 +- src/mesa/pipe/softpipe/sp_context.c | 13 ++ src/mesa/pipe/softpipe/sp_z_surface.c | 118 +++++++++++ src/mesa/pipe/softpipe/sp_z_surface.h | 37 ++++ src/mesa/sources | 1 + src/mesa/state_tracker/st_atom_framebuffer.c | 19 +- 14 files changed, 449 insertions(+), 353 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_z_surface.c create mode 100644 src/mesa/pipe/softpipe/sp_z_surface.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c index a09db46163..5a93eb7ad1 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c @@ -289,6 +289,12 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, rb->Width = width; rb->Height = height; +#if 1 + /* update the surface's size too */ + rb->surface->width = width; + rb->surface->height = height; +#endif + /* This sets the Get/PutRow/Value functions */ intel_set_span_functions(&irb->Base); @@ -451,6 +457,12 @@ intel_create_renderbuffer(GLenum intFormat, GLsizei width, GLsizei height, return irb; } + +/** + * Create a new renderbuffer which corresponds to an X window buffer + * (color, depth, stencil, etc) - not a user-created GL renderbuffer. + * The internal format is set at creation time and does not change. + */ struct gl_renderbuffer * intel_new_renderbuffer_fb(GLcontext * ctx, GLuint intFormat) { @@ -472,6 +484,9 @@ intel_new_renderbuffer_fb(GLcontext * ctx, GLuint intFormat) irb->Base.GetPointer = intel_get_pointer; /* span routines set in alloc_storage function */ + irb->Base.surface = intel_new_surface(intFormat); + irb->Base.surface->rb = irb; + return &irb->Base; } @@ -500,6 +515,9 @@ intel_new_renderbuffer(GLcontext * ctx, GLuint name) irb->Base.GetPointer = intel_get_pointer; /* span routines set in alloc_storage function */ + irb->Base.surface = intel_new_surface(0 /*unknown format*/); + irb->Base.surface->rb = irb; + return &irb->Base; } diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.h b/src/mesa/drivers/dri/i915tex/intel_fbo.h index 1642ce774f..86c8106084 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.h +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.h @@ -37,17 +37,6 @@ struct intel_context; struct intel_region; -/** - * Intel "pipe" surface. This is kind of a temporary thing as - * renderbuffers and surfaces should eventually become one. - */ -struct intel_surface -{ - struct softpipe_surface surface; /**< base class */ - struct intel_renderbuffer *rb; /**< ptr back to matching renderbuffer */ -}; - - /** * Intel framebuffer, derived from gl_framebuffer. */ @@ -129,5 +118,8 @@ extern struct intel_region *intel_get_rb_region(struct gl_framebuffer *fb, +extern struct pipe_surface * +intel_new_surface(GLuint intFormat); + #endif /* INTEL_FBO_H */ diff --git a/src/mesa/drivers/dri/i915tex/intel_surface.c b/src/mesa/drivers/dri/i915tex/intel_surface.c index 3be902cf9c..043c5aa5fe 100644 --- a/src/mesa/drivers/dri/i915tex/intel_surface.c +++ b/src/mesa/drivers/dri/i915tex/intel_surface.c @@ -23,27 +23,33 @@ * XXX a lof of this is a temporary kludge */ +/** + * Note: the arithmetic/addressing in these functions is a little + * tricky since we need to invert the Y axis. + */ static void read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rrrr)[QUAD_SIZE]) { - struct intel_surface *is = (struct intel_surface *) sps; - struct intel_renderbuffer *irb = is->rb; - const GLubyte *src = (const GLubyte *) irb->region->map - + (y * irb->region->pitch + x) * irb->region->cpp; + const GLint bytesPerRow = sps->surface.stride * sps->surface.cpp; + const GLint invY = sps->surface.height - y - 1; + const GLubyte *src = sps->surface.ptr + invY * bytesPerRow + x * sps->surface.cpp; GLfloat *dst = (GLfloat *) rrrr; GLubyte temp[16]; - GLuint i, j; + GLuint j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - memcpy(temp, src, 8); - memcpy(temp + 8, src + irb->region->pitch * irb->region->cpp, 8); + memcpy(temp + 8, src, 8); + memcpy(temp + 0, src + bytesPerRow, 8); - for (i = 0; i < 4; i++) { - for (j = 0; j < 4; j++) { - dst[j * 4 + i] = UBYTE_TO_FLOAT(temp[i * 4 + j]); - } + for (j = 0; j < 4; j++) { + dst[0 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 2]); /*R*/ + dst[1 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 1]); /*G*/ + dst[2 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 0]); /*B*/ + dst[3 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 3]); /*A*/ } } @@ -52,45 +58,128 @@ static void write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rrrr)[QUAD_SIZE]) { - struct intel_surface *is = (struct intel_surface *) sps; - struct intel_renderbuffer *irb = is->rb; const GLfloat *src = (const GLfloat *) rrrr; - GLubyte *dst = (GLubyte *) irb->region->map - + (y * irb->region->pitch + x) * irb->region->cpp; + const GLint bytesPerRow = sps->surface.stride * sps->surface.cpp; + const GLint invY = sps->surface.height - y - 1; + GLubyte *dst = sps->surface.ptr + invY * bytesPerRow + x * sps->surface.cpp; GLubyte temp[16]; - GLuint i, j; + GLuint j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - for (i = 0; i < 4; i++) { - for (j = 0; j < 4; j++) { - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + i], src[i * 4 + j]); - } + for (j = 0; j < 4; j++) { + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 2], src[0 * 4 + j]); /*R*/ + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 1], src[1 * 4 + j]); /*G*/ + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 0], src[2 * 4 + j]); /*B*/ + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 3], src[3 * 4 + j]); /*A*/ } - memcpy(dst, temp, 8); - memcpy(dst + irb->region->pitch * irb->region->cpp, temp + 8, 8); + memcpy(dst, temp + 8, 8); + memcpy(dst + bytesPerRow, temp + 0, 8); } +static void +read_quad_z24(struct softpipe_surface *sps, + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) +{ + static const GLuint mask = 0xffffff; + const GLint invY = sps->surface.height - y - 1; + const GLuint *src + = (GLuint *) (sps->surface.ptr + + (invY * sps->surface.stride + x) * sps->surface.cpp); + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + /* extract lower three bytes */ + zzzz[0] = src[0] & mask; + zzzz[1] = src[1] & mask; + zzzz[2] = src[-sps->surface.stride] & mask; + zzzz[3] = src[-sps->surface.stride + 1] & mask; +} + +static void +write_quad_z24(struct softpipe_surface *sps, + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) +{ + static const GLuint mask = 0xff000000; + const GLint invY = sps->surface.height - y - 1; + GLuint *dst + = (GLuint *) (sps->surface.ptr + + (invY * sps->surface.stride + x) * sps->surface.cpp); + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + /* write lower three bytes */ + dst[0] = (dst[0] & mask) | zzzz[0]; + dst[1] = (dst[1] & mask) | zzzz[1]; + dst -= sps->surface.stride; + dst[0] = (dst[0] & mask) | zzzz[2]; + dst[1] = (dst[1] & mask) | zzzz[3]; +} + + +static void +read_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) +{ + const GLint invY = sps->surface.height - y - 1; + const GLuint *src = (const GLuint *) (sps->surface.ptr + + (invY * sps->surface.stride + x) * sps->surface.cpp); + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + /* extract high byte */ + ssss[0] = src[0] >> 24; + ssss[1] = src[1] >> 24; + ssss[2] = src[-sps->surface.width] >> 24; + ssss[3] = src[-sps->surface.width + 1] >> 24; +} + +static void +write_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) +{ + static const GLuint mask = 0x00ffffff; + const GLint invY = sps->surface.height - y - 1; + GLuint *dst = (GLuint *) (sps->surface.ptr + + (invY * sps->surface.stride + x) * sps->surface.cpp); + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + /* write high byte */ + dst[0] = (dst[0] & mask) | (ssss[0] << 24); + dst[1] = (dst[1] & mask) | (ssss[1] << 24); + dst -= sps->surface.stride; + dst[0] = (dst[0] & mask) | (ssss[2] << 24); + dst[1] = (dst[1] & mask) | (ssss[3] << 24); +} + + static void * map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) { - struct intel_surface *is = (struct intel_surface *) pb; - struct intel_renderbuffer *irb = is->rb; - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); - + struct softpipe_surface *sps = (struct softpipe_surface *) pb; + struct intel_renderbuffer *irb = (struct intel_renderbuffer *) sps->surface.rb; assert(access_mode == PIPE_MAP_READ_WRITE); - intelFinish(&intel->ctx); - /*LOCK_HARDWARE(intel);*/ if (irb->region) { + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); +#if 0 + intelFinish(&intel->ctx); /* XXX need this? */ +#endif intel_region_map(intel->intelScreen, irb->region); } pb->ptr = irb->region->map; + sps->surface.stride = irb->region->pitch; + sps->surface.cpp = irb->region->cpp; + sps->surface.ptr = irb->region->map; + return pb->ptr; } @@ -98,69 +187,67 @@ map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) static void unmap_surface_buffer(struct pipe_buffer *pb) { - struct intel_surface *is = (struct intel_surface *) pb; - struct intel_renderbuffer *irb = is->rb; - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); + struct softpipe_surface *sps = (struct softpipe_surface *) pb; + struct intel_renderbuffer *irb = (struct intel_renderbuffer *) sps->surface.rb; if (irb->region) { + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); intel_region_unmap(intel->intelScreen, irb->region); } pb->ptr = NULL; + sps->surface.stride = 0; + sps->surface.cpp = 0; + sps->surface.ptr = NULL; + /*UNLOCK_HARDWARE(intel);*/ } struct pipe_surface * -xmesa_get_color_surface(GLcontext *ctx, GLuint i) +intel_new_surface(GLuint intFormat) { - struct intel_context *intel = intel_context(ctx); - struct intel_framebuffer *intel_fb; - struct intel_renderbuffer *intel_rb; - - intel_fb = (struct intel_framebuffer *) ctx->DrawBuffer; - intel_rb = intel_fb->color_rb[1]; - - if (!intel_rb->surface) { - /* create surface and attach to intel_rb */ - struct intel_surface *is; - is = CALLOC_STRUCT(intel_surface); - if (is) { - is->surface.surface.width = intel_rb->Base.Width; - is->surface.surface.height = intel_rb->Base.Height; - - is->surface.read_quad_f_swz = read_quad_f_swz; - is->surface.write_quad_f_swz = write_quad_f_swz; - - is->surface.surface.buffer.map = map_surface_buffer; - is->surface.surface.buffer.unmap = unmap_surface_buffer; - - is->rb = intel_rb; - } - intel_rb->surface = is; + struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); + if (!sps) + return NULL; + + sps->surface.width = 0; /* set in intel_alloc_renderbuffer_storage() */ + sps->surface.height = 0; + + if (intFormat == GL_RGBA8) { + sps->surface.format = PIPE_FORMAT_U_A8_R8_G8_B8; + sps->read_quad_f_swz = read_quad_f_swz; + sps->write_quad_f_swz = write_quad_f_swz; } - else { - /* update surface size */ - struct intel_surface *is = intel_rb->surface; - is->surface.surface.width = intel_rb->Base.Width; - is->surface.surface.height = intel_rb->Base.Height; - /* sanity check */ - assert(is->surface.surface.buffer.map == map_surface_buffer); + else if (intFormat == GL_RGB5) { + sps->surface.format = PIPE_FORMAT_U_R5_G6_B5; + } + else if (intFormat == GL_DEPTH_COMPONENT16) { + sps->surface.format = PIPE_FORMAT_U_Z16; - return &intel_rb->surface->surface.surface; -} + } + else if (intFormat == GL_DEPTH24_STENCIL8_EXT) { + sps->surface.format = PIPE_FORMAT_Z24_S8; + sps->read_quad_z = read_quad_z24; + sps->write_quad_z = write_quad_z24; + sps->read_quad_stencil = read_quad_stencil; + sps->write_quad_stencil = write_quad_stencil; + } + else { + /* TBD / unknown */ + } -struct pipe_surface * -xmesa_get_z_surface(GLcontext *ctx) -{ - /* XXX fix */ - return NULL; + sps->surface.buffer.map = map_surface_buffer; + sps->surface.buffer.unmap = unmap_surface_buffer; + + return &sps->surface; } + struct pipe_surface * xmesa_get_stencil_surface(GLcontext *ctx) { diff --git a/src/mesa/drivers/x11/xm_buffer.c b/src/mesa/drivers/x11/xm_buffer.c index 51d183bb43..8fbd9a783b 100644 --- a/src/mesa/drivers/x11/xm_buffer.c +++ b/src/mesa/drivers/x11/xm_buffer.c @@ -35,6 +35,7 @@ #include "imports.h" #include "framebuffer.h" #include "renderbuffer.h" +#include "pipe/p_state.h" #if defined(USE_XSHM) && !defined(XFree86Server) @@ -268,6 +269,8 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->Height = height; rb->InternalFormat = internalFormat; + rb->surface->resize(rb->surface, width, height); + return GL_TRUE; } @@ -317,6 +320,8 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xrb->origin4 = NULL; } + rb->surface->resize(rb->surface, width, height); + return GL_TRUE; } @@ -352,6 +357,9 @@ xmesa_new_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, xrb->Base.IndexBits = visual->indexBits; } /* only need to set Red/Green/EtcBits fields for user-created RBs */ + + xrb->Base.surface = xmesa_new_surface(xrb); + } return xrb; } diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 9d6db2b5ce..17f5f28a9d 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -24,7 +24,7 @@ /** - * \file xm_surface.h + * \file xm_surface.c * Code to allow the softpipe code to write to X windows/buffers. * This is a bit of a hack for now. We've basically got two different * abstractions for color buffers: gl_renderbuffer and softpipe_surface. @@ -48,31 +48,11 @@ #include "pipe/softpipe/sp_surface.h" -/** - * An xm_surface is derived from a softpipe_surface - */ -struct xmesa_surface -{ - struct softpipe_surface sps; - struct xmesa_renderbuffer *xrb; /** ptr back to matching xmesa_renderbuffer */ - struct gl_renderbuffer *rb; /* ptr to matching gl_renderbuffer */ -}; - - -/** - * Cast wrapper - */ -static INLINE struct xmesa_surface * -xmesa_surface(struct softpipe_surface *sps) -{ - return (struct xmesa_surface *) sps; -} - - static void * map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) { /* no-op */ + return NULL; } @@ -83,6 +63,13 @@ unmap_surface_buffer(struct pipe_buffer *pb) } +static INLINE struct xmesa_renderbuffer * +xmesa_rb(struct softpipe_surface *sps) +{ + return (struct xmesa_renderbuffer *) sps->surface.rb; +} + + /** * quad reading/writing * These functions are just wrappers around the existing renderbuffer @@ -90,11 +77,10 @@ unmap_surface_buffer(struct pipe_buffer *pb) */ static void -read_quad_f(struct softpipe_surface *gs, GLint x, GLint y, +read_quad_f(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rgba)[NUM_CHANNELS]) { - struct xmesa_surface *xmsurf = xmesa_surface(gs); - struct xmesa_renderbuffer *xrb = xmsurf->xrb; + struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GLubyte temp[16]; GLfloat *dst = (GLfloat *) rgba; GLuint i; @@ -107,11 +93,10 @@ read_quad_f(struct softpipe_surface *gs, GLint x, GLint y, } static void -read_quad_f_swz(struct softpipe_surface *gs, GLint x, GLint y, +read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rrrr)[QUAD_SIZE]) { - struct xmesa_surface *xmsurf = xmesa_surface(gs); - struct xmesa_renderbuffer *xrb = xmsurf->xrb; + struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GLubyte temp[16]; GLfloat *dst = (GLfloat *) rrrr; GLuint i, j; @@ -126,11 +111,10 @@ read_quad_f_swz(struct softpipe_surface *gs, GLint x, GLint y, } static void -write_quad_f(struct softpipe_surface *gs, GLint x, GLint y, +write_quad_f(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rgba)[NUM_CHANNELS]) { - struct xmesa_surface *xmsurf = xmesa_surface(gs); - struct xmesa_renderbuffer *xrb = xmsurf->xrb; + struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GLubyte temp[16]; const GLfloat *src = (const GLfloat *) rgba; GLuint i; @@ -143,11 +127,10 @@ write_quad_f(struct softpipe_surface *gs, GLint x, GLint y, } static void -write_quad_f_swz(struct softpipe_surface *gs, GLint x, GLint y, +write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rrrr)[QUAD_SIZE]) { - struct xmesa_surface *xmsurf = xmesa_surface(gs); - struct xmesa_renderbuffer *xrb = xmsurf->xrb; + struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GLubyte temp[16]; const GLfloat *src = (const GLfloat *) rrrr; GLuint i, j; @@ -162,251 +145,70 @@ write_quad_f_swz(struct softpipe_surface *gs, GLint x, GLint y, } static void -read_quad_ub(struct softpipe_surface *gs, GLint x, GLint y, +read_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, GLubyte (*rgba)[NUM_CHANNELS]) { - struct xmesa_surface *xmsurf = xmesa_surface(gs); - struct xmesa_renderbuffer *xrb = xmsurf->xrb; + struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GET_CURRENT_CONTEXT(ctx); xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, rgba); xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, rgba + 2); } static void -write_quad_ub(struct softpipe_surface *gs, GLint x, GLint y, +write_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, GLubyte (*rgba)[NUM_CHANNELS]) { - struct xmesa_surface *xmsurf = xmesa_surface(gs); - struct xmesa_renderbuffer *xrb = xmsurf->xrb; + struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GET_CURRENT_CONTEXT(ctx); xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, rgba); xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, rgba + 2); } static void -write_mono_row_ub(struct softpipe_surface *gs, GLuint count, GLint x, GLint y, +write_mono_row_ub(struct softpipe_surface *sps, GLuint count, GLint x, GLint y, GLubyte rgba[NUM_CHANNELS]) { - struct xmesa_surface *xmsurf = xmesa_surface(gs); - struct xmesa_renderbuffer *xrb = xmsurf->xrb; + struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GET_CURRENT_CONTEXT(ctx); xrb->Base.PutMonoRow(ctx, &xrb->Base, count, x, y, rgba, NULL); } -static struct xmesa_surface * -create_surface(XMesaContext xmctx, struct xmesa_renderbuffer *xrb) -{ - struct xmesa_surface *xmsurf; - - xmsurf = CALLOC_STRUCT(xmesa_surface); - if (xmsurf) { - xmsurf->xrb = xrb; - xmsurf->sps.surface.width = xrb->Base.Width; - xmsurf->sps.surface.height = xrb->Base.Height; - - xmsurf->sps.read_quad_f = read_quad_f; - xmsurf->sps.read_quad_f_swz = read_quad_f_swz; - xmsurf->sps.read_quad_ub = read_quad_ub; - xmsurf->sps.write_quad_f = write_quad_f; - xmsurf->sps.write_quad_f_swz = write_quad_f_swz; - xmsurf->sps.write_quad_ub = write_quad_ub; - xmsurf->sps.write_mono_row_ub = write_mono_row_ub; - - xmsurf->sps.surface.buffer.map = map_surface_buffer; - xmsurf->sps.surface.buffer.unmap = unmap_surface_buffer; - -#if 0 - if (xrb->ximage) { - xmsurf->sps.surface.ptr = (GLubyte *) xrb->ximage->data; - xmsurf->sps.surface.stride = xrb->ximage->bytes_per_line; - xmsurf->sps.surface.cpp = xrb->ximage->depth; - - } -#endif - } - return xmsurf; -} - - -static void -free_surface(struct softpipe_surface *sps) -{ - /* XXX may need to do more in the future */ - free(sps); -} - - -/** - * Return generic surface pointer corresponding to the current color buffer. - */ -struct pipe_surface * -xmesa_get_color_surface(GLcontext *ctx, GLuint buf) -{ - XMesaContext xmctx = XMESA_CONTEXT(ctx); - struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[0][buf]; - struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); - struct softpipe_surface *sps = (struct softpipe_surface *) xrb->pSurface; - - if (!sps) { - xrb->pSurface = create_surface(xmctx, xrb); - } - else if (sps->surface.width != rb->Width || - sps->surface.height != rb->Height) { - free_surface(sps); - xrb->pSurface = create_surface(xmctx, xrb); - } - - return (struct pipe_surface *) xrb->pSurface; -} - - - - -static void -read_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) -{ - struct xmesa_surface *xmsurf = xmesa_surface(sps); - struct gl_renderbuffer *rb = xmsurf->rb; - GLushort temp[4]; - GLuint i; - GET_CURRENT_CONTEXT(ctx); - rb->GetRow(ctx, rb, 2, x, y, temp); - rb->GetRow(ctx, rb, 2, x, y + 1, temp + 2); - /* convert from GLushort to GLuint */ - for (i = 0; i < 4; i++) { - zzzz[i] = temp[i]; - } -} - -static void -write_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) -{ - struct xmesa_surface *xmsurf = xmesa_surface(sps); - struct gl_renderbuffer *rb = xmsurf->rb; - GLushort temp[4]; - GLuint i; - GET_CURRENT_CONTEXT(ctx); - /* convert from GLuint to GLushort */ - for (i = 0; i < 4; i++) { - temp[i] = zzzz[i]; - } - rb->PutRow(ctx, rb, 2, x, y, temp, NULL); - rb->PutRow(ctx, rb, 2, x, y + 1, temp + 2, NULL); -} - - -static struct xmesa_surface * -create_z_surface(XMesaContext xmctx, struct gl_renderbuffer *rb) -{ - struct xmesa_surface *xmsurf; - - xmsurf = CALLOC_STRUCT(xmesa_surface); - if (xmsurf) { - xmsurf->sps.surface.format = PIPE_FORMAT_U_Z16; - xmsurf->sps.surface.width = rb->Width; - xmsurf->sps.surface.height = rb->Height; - xmsurf->sps.read_quad_z = read_quad_z; - xmsurf->sps.write_quad_z = write_quad_z; - xmsurf->rb = rb; - } - return xmsurf; -} - - - - static void -read_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) +resize_surface(struct pipe_surface *ps, GLuint width, GLuint height) { - struct xmesa_surface *xmsurf = xmesa_surface(sps); - struct gl_renderbuffer *rb = xmsurf->rb; - GET_CURRENT_CONTEXT(ctx); - rb->GetRow(ctx, rb, 2, x, y, ssss); - rb->GetRow(ctx, rb, 2, x, y + 1, ssss + 2); -} - -static void -write_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) -{ - struct xmesa_surface *xmsurf = xmesa_surface(sps); - struct gl_renderbuffer *rb = xmsurf->rb; - GET_CURRENT_CONTEXT(ctx); - rb->PutRow(ctx, rb, 2, x, y, ssss, NULL); - rb->PutRow(ctx, rb, 2, x, y + 1, ssss + 2, NULL); -} - -static struct xmesa_surface * -create_stencil_surface(XMesaContext xmctx, struct gl_renderbuffer *rb) -{ - struct xmesa_surface *xmsurf; - - xmsurf = CALLOC_STRUCT(xmesa_surface); - if (xmsurf) { - xmsurf->sps.surface.format = PIPE_FORMAT_U_S8; - xmsurf->sps.surface.width = rb->Width; - xmsurf->sps.surface.height = rb->Height; - xmsurf->sps.read_quad_stencil = read_quad_stencil; - xmsurf->sps.write_quad_stencil = write_quad_stencil; - xmsurf->rb = rb; - } - return xmsurf; + ps->width = width; + ps->height = height; } - - /** - * Return a pipe_surface that wraps the current Z/depth buffer. - * XXX this is pretty much a total hack until gl_renderbuffers and - * pipe_surfaces are merged... + * Called to create a pipe_surface for each X renderbuffer. */ struct pipe_surface * -xmesa_get_z_surface(GLcontext *ctx) +xmesa_new_surface(struct xmesa_renderbuffer *xrb) { - XMesaContext xmctx = XMESA_CONTEXT(ctx); - struct gl_renderbuffer *rb = ctx->DrawBuffer->_DepthBuffer; - static struct xmesa_surface *xms = NULL; + struct softpipe_surface *sps; - if (!rb) + sps = CALLOC_STRUCT(softpipe_surface); + if (!sps) return NULL; - if (!xms) { - xms = create_z_surface(xmctx, rb); - } - else if (xms->sps.surface.width != rb->Width || - xms->sps.surface.height != rb->Height) { - free_surface(&xms->sps); - xms = create_z_surface(xmctx, rb); - } - - return (struct pipe_surface *) &xms->sps.surface; -} + sps->surface.rb = xrb; + sps->surface.width = xrb->Base.Width; + sps->surface.height = xrb->Base.Height; + sps->read_quad_f = read_quad_f; + sps->read_quad_f_swz = read_quad_f_swz; + sps->read_quad_ub = read_quad_ub; + sps->write_quad_f = write_quad_f; + sps->write_quad_f_swz = write_quad_f_swz; + sps->write_quad_ub = write_quad_ub; + sps->write_mono_row_ub = write_mono_row_ub; -struct pipe_surface * -xmesa_get_stencil_surface(GLcontext *ctx) -{ - XMesaContext xmctx = XMESA_CONTEXT(ctx); - struct gl_renderbuffer *rb = ctx->DrawBuffer->_StencilBuffer; - static struct xmesa_surface *xms = NULL; + sps->surface.buffer.map = map_surface_buffer; + sps->surface.buffer.unmap = unmap_surface_buffer; + sps->surface.resize = resize_surface; - if (!rb) - return NULL; - - if (!xms) { - xms = create_stencil_surface(xmctx, rb); - } - else if (xms->sps.surface.width != rb->Width || - xms->sps.surface.height != rb->Height) { - free_surface(&xms->sps); - xms = create_stencil_surface(xmctx, rb); - } - - return (struct pipe_surface *) &xms->sps.surface; + return &sps->surface; } - diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 8648b19939..daf6a3f942 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -585,23 +585,11 @@ extern void xmesa_register_swrast_functions( GLcontext *ctx ); #define ENABLE_EXT_timer_query 0 /* may not have 64-bit GLuint64EXT */ #endif -#if 0 -GLboolean xmesa_get_cbuf_details( GLcontext *ctx, - void **ptr, - GLuint *cpp, - GLint *stride, - GLuint *format ); -#endif struct pipe_surface; -struct pipe_surface * -xmesa_get_color_surface(GLcontext *ctx, GLuint buf); - -struct pipe_surface * -xmesa_get_z_surface(GLcontext *ctx); struct pipe_surface * -xmesa_get_stencil_surface(GLcontext *ctx); +xmesa_new_surface(struct xmesa_renderbuffer *xrb); #endif diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 52448ee04e..d70df5d945 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -127,6 +127,7 @@ struct gl_texture_format; struct gl_texture_image; struct gl_texture_object; struct st_context; +struct pipe_surface; typedef struct __GLcontextRec GLcontext; typedef struct __GLcontextModesRec GLvisual; typedef struct gl_framebuffer GLframebuffer; @@ -2268,6 +2269,8 @@ struct gl_renderbuffer GLubyte StencilBits; GLvoid *Data; /**< This may not be used by some kinds of RBs */ + struct pipe_surface *surface; + /* Used to wrap one renderbuffer around another: */ struct gl_renderbuffer *Wrapped; diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c index 6f1d7c3960..a1412ef007 100644 --- a/src/mesa/main/renderbuffer.c +++ b/src/mesa/main/renderbuffer.c @@ -49,6 +49,9 @@ #include "rbadaptors.h" +#include "pipe/softpipe/sp_z_surface.h" +#include "pipe/p_state.h" + /* 32-bit color index format. Not a public format. */ #define COLOR_INDEX32 0x424243 @@ -1091,6 +1094,7 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutValues = put_values_ushort; rb->PutMonoValues = put_mono_values_ushort; rb->DepthBits = 8 * sizeof(GLushort); + rb->surface = (struct pipe_surface *) softpipe_new_z_surface(16); pixelSize = sizeof(GLushort); break; case GL_DEPTH_COMPONENT24: @@ -1193,13 +1197,27 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, /* free old buffer storage */ if (rb->Data) { - _mesa_free(rb->Data); + if (rb->surface) { + /* pipe surface */ + } + else { + /* legacy renderbuffer */ + _mesa_free(rb->Data); + } rb->Data = NULL; } if (width > 0 && height > 0) { /* allocate new buffer storage */ - rb->Data = _mesa_malloc(width * height * pixelSize); + if (rb->surface) { + /* pipe surface */ + rb->surface->resize(rb->surface, width, height); + rb->Data = rb->surface->buffer.ptr; + } + else { + /* legacy renderbuffer */ + rb->Data = _mesa_malloc(width * height * pixelSize); + } if (rb->Data == NULL) { rb->Width = 0; rb->Height = 0; diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 4ae8928018..9973a7b8dd 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -239,6 +239,7 @@ struct pipe_sampler_state /** * A mappable buffer (vertex data, pixel data, etc) + * XXX replace with "intel_region". */ struct pipe_buffer { @@ -247,7 +248,7 @@ struct pipe_buffer const void *src); void *(*map)(struct pipe_buffer *pb, GLuint access_mode); void (*unmap)(struct pipe_buffer *pb); - void *ptr; /**< address, only valid while mapped */ + GLubyte *ptr; /**< address, only valid while mapped */ GLuint mode; /**< PIPE_MAP_x, only valid while mapped */ }; @@ -261,12 +262,13 @@ struct pipe_surface struct pipe_buffer buffer; /**< surfaces can be mapped */ GLuint format:5; /**< PIPE_FORMAT_x */ GLuint width, height; -#if 0 - GLubyte *ptr; - GLint stride; - GLuint cpp; - GLuint format; -#endif + + GLint stride, cpp; + GLubyte *ptr; /**< only valid while mapped, may not equal buffer->ptr */ + + void *rb; /**< Ptr back to renderbuffer (temporary?) */ + + void (*resize)(struct pipe_surface *ps, GLuint width, GLuint height); }; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 6b44fabfa4..8655aa83fd 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -49,6 +49,13 @@ static void map_surfaces(struct softpipe_context *sp) struct pipe_buffer *buf = &sps->surface.buffer; buf->map(buf, PIPE_MAP_READ_WRITE); } + + if (sp->framebuffer.zbuf) { + struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.zbuf); + struct pipe_buffer *buf = &sps->surface.buffer; + buf->map(buf, PIPE_MAP_READ_WRITE); + } + /* XXX depth & stencil bufs */ } @@ -62,6 +69,12 @@ static void unmap_surfaces(struct softpipe_context *sp) struct pipe_buffer *buf = &sps->surface.buffer; buf->unmap(buf); } + + if (sp->framebuffer.zbuf) { + struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.zbuf); + struct pipe_buffer *buf = &sps->surface.buffer; + buf->unmap(buf); + } /* XXX depth & stencil bufs */ } diff --git a/src/mesa/pipe/softpipe/sp_z_surface.c b/src/mesa/pipe/softpipe/sp_z_surface.c new file mode 100644 index 0000000000..662a4a15ee --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_z_surface.c @@ -0,0 +1,118 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Software Z buffer/surface. + * + * \author Brian Paul + */ + + +#include "main/imports.h" +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "sp_surface.h" +#include "sp_z_surface.h" + +static void* +z16_map(struct pipe_buffer *pb, GLuint access_mode) +{ + struct softpipe_surface *sps = (struct softpipe_surface *) pb; + sps->surface.ptr = pb->ptr; + return pb->ptr; +} + +static void +z16_unmap(struct pipe_buffer *pb) +{ + struct softpipe_surface *sps = (struct softpipe_surface *) pb; + sps->surface.ptr = NULL; +} + +static void +z16_resize(struct pipe_surface *ps, GLuint width, GLuint height) +{ + struct softpipe_surface *sps = (struct softpipe_surface *) ps; + + if (sps->surface.buffer.ptr) + free(sps->surface.buffer.ptr); + + ps->buffer.ptr = (GLubyte *) malloc(width * height * sizeof(GLushort)); + ps->width = width; + ps->height = height; + + sps->surface.stride = sps->surface.width; + sps->surface.cpp = 2; +} + +static void +z16_read_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) +{ + const GLushort *src + = (GLushort *) sps->surface.ptr + y * sps->surface.stride + x; + + /* converting GLushort to GLuint: */ + zzzz[0] = src[0]; + zzzz[1] = src[1]; + zzzz[2] = src[sps->surface.width]; + zzzz[3] = src[sps->surface.width + 1]; +} + +static void +z16_write_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) +{ + GLushort *dst = (GLushort *) sps->surface.ptr + y * sps->surface.stride + x; + + /* converting GLuint to GLushort: */ + dst[0] = zzzz[0]; + dst[1] = zzzz[1]; + dst[sps->surface.width] = zzzz[2]; + dst[sps->surface.width + 1] = zzzz[3]; +} + +struct softpipe_surface * +softpipe_new_z_surface(GLuint depth) +{ + struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); + if (!sps) + return NULL; + + /* XXX ignoring depth param for now */ + + sps->surface.format = PIPE_FORMAT_U_Z16; + + sps->surface.resize = z16_resize; + sps->surface.buffer.map = z16_map; + sps->surface.buffer.unmap = z16_unmap; + sps->read_quad_z = z16_read_quad_z; + sps->write_quad_z = z16_write_quad_z; + + return sps; +} diff --git a/src/mesa/pipe/softpipe/sp_z_surface.h b/src/mesa/pipe/softpipe/sp_z_surface.h new file mode 100644 index 0000000000..9c3c43ca57 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_z_surface.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SP_Z_SURFACE_H +#define SP_Z_SURFACE_H + + +extern struct softpipe_surface * +softpipe_new_z_surface(GLuint depth); + + +#endif /* SP_Z_SURFACE_H */ diff --git a/src/mesa/sources b/src/mesa/sources index a589ae4373..32c2ff2350 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -176,6 +176,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_state_sampler.c \ pipe/softpipe/sp_state_setup.c \ pipe/softpipe/sp_state_surface.c \ + pipe/softpipe/sp_z_surface.c \ pipe/softpipe/sp_prim_setup.c DRAW_SOURCES = \ diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 595f390b28..f5e3ce8b67 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -54,22 +54,31 @@ static void update_framebuffer_state( struct st_context *st ) { struct pipe_framebuffer_state framebuffer; + struct gl_renderbuffer *rb; GLuint i; + memset(&framebuffer, 0, sizeof(framebuffer)); + /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state * to determine which surfaces to draw to */ framebuffer.num_cbufs = st->ctx->DrawBuffer->_NumColorDrawBuffers[0]; for (i = 0; i < framebuffer.num_cbufs; i++) { - framebuffer.cbufs[i] = xmesa_get_color_surface(st->ctx, i); + rb = st->ctx->DrawBuffer->_ColorDrawBuffers[0][i]; + assert(rb->surface); + framebuffer.cbufs[i] = rb->surface; } - if (st->ctx->DrawBuffer->_DepthBuffer/*Attachment[BUFFER_DEPTH].Renderbuffer*/) { - framebuffer.zbuf = xmesa_get_z_surface(st->ctx); + rb = st->ctx->DrawBuffer->_DepthBuffer; + if (rb) { + assert(rb->surface); + framebuffer.zbuf = rb->Wrapped->surface; } - if (st->ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer) { - framebuffer.sbuf = xmesa_get_stencil_surface(st->ctx); + rb = st->ctx->DrawBuffer->_StencilBuffer; + if (rb) { + assert(rb->surface); + framebuffer.sbuf = rb->Wrapped->surface; } if (memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0) { -- cgit v1.2.3 From 42c817269a4af535225217d09ec416efb085a94e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 17:20:02 -0600 Subject: remove some obsolete xmesa remnants --- src/mesa/state_tracker/st_atom_framebuffer.c | 10 ---------- 1 file changed, 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index f5e3ce8b67..a8d47ea4f4 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -36,16 +36,6 @@ #include "pipe/p_context.h" -extern struct pipe_surface * -xmesa_get_color_surface(GLcontext *ctx, GLuint i); - -extern struct pipe_surface * -xmesa_get_z_surface(GLcontext *ctx); - -extern struct pipe_surface * -xmesa_get_stencil_surface(GLcontext *ctx); - - /** * Update framebuffer state (color, depth, stencil, etc. buffers) * XXX someday: separate draw/read buffers. -- cgit v1.2.3 From 9bc1c92a0b809c6b60d5e4a2c8909f5f98528919 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 30 Jul 2007 21:39:57 -0600 Subject: 32 and z24s8 softpipe buffers --- src/mesa/main/renderbuffer.c | 8 +- src/mesa/pipe/p_defines.h | 28 +++--- src/mesa/pipe/softpipe/sp_z_surface.c | 159 ++++++++++++++++++++++++++++++---- src/mesa/pipe/softpipe/sp_z_surface.h | 2 +- 4 files changed, 164 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c index a1412ef007..a900de169e 100644 --- a/src/mesa/main/renderbuffer.c +++ b/src/mesa/main/renderbuffer.c @@ -51,6 +51,7 @@ #include "pipe/softpipe/sp_z_surface.h" #include "pipe/p_state.h" +#include "pipe/p_defines.h" /* 32-bit color index format. Not a public format. */ @@ -1094,7 +1095,8 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutValues = put_values_ushort; rb->PutMonoValues = put_mono_values_ushort; rb->DepthBits = 8 * sizeof(GLushort); - rb->surface = (struct pipe_surface *) softpipe_new_z_surface(16); + rb->surface + = (struct pipe_surface *) softpipe_new_z_surface(PIPE_FORMAT_U_Z16); pixelSize = sizeof(GLushort); break; case GL_DEPTH_COMPONENT24: @@ -1117,6 +1119,8 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->_ActualFormat = GL_DEPTH_COMPONENT32; rb->DepthBits = 32; } + rb->surface + = (struct pipe_surface *) softpipe_new_z_surface(PIPE_FORMAT_U_Z32); pixelSize = sizeof(GLuint); break; case GL_DEPTH_STENCIL_EXT: @@ -1134,6 +1138,8 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutMonoValues = put_mono_values_uint; rb->DepthBits = 24; rb->StencilBits = 8; + rb->surface + = (struct pipe_surface *) softpipe_new_z_surface(PIPE_FORMAT_Z24_S8); pixelSize = sizeof(GLuint); break; case GL_COLOR_INDEX8_EXT: diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 1b799f1451..58f01758e3 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -132,19 +132,21 @@ /** * Texture/surface image formats (preliminary) */ -#define PIPE_FORMAT_U_R8_G8_B8_A8 0 /**< ubyte[4] RGBA */ -#define PIPE_FORMAT_U_A8_R8_G8_B8 1 /**< ubyte[4] ARGB */ -#define PIPE_FORMAT_U_R5_G6_B5 2 /**< 5/6/5 RGB */ -#define PIPE_FORMAT_U_L8 3 /**< ubyte luminance */ -#define PIPE_FORMAT_U_A8 4 /**< ubyte alpha */ -#define PIPE_FORMAT_U_I8 5 /**< ubyte intensity */ -#define PIPE_FORMAT_U_L8_A8 6 /**< ubyte luminance, alpha */ -#define PIPE_FORMAT_U_Z16 7 /**< ushort Z/depth */ -#define PIPE_FORMAT_F_Z32 8 /**< float Z/depth */ -#define PIPE_FORMAT_YCBCR 9 -#define PIPE_FORMAT_YCBCR_REV 10 -#define PIPE_FORMAT_U_S8 11 /**< 8-bit stencil */ -#define PIPE_FORMAT_Z24_S8 12 /**< 24-bit Z + 8-bit stencil */ +#define PIPE_FORMAT_NONE 0 /**< unstructured */ +#define PIPE_FORMAT_U_R8_G8_B8_A8 1 /**< ubyte[4] RGBA */ +#define PIPE_FORMAT_U_A8_R8_G8_B8 2 /**< ubyte[4] ARGB */ +#define PIPE_FORMAT_U_R5_G6_B5 3 /**< 5/6/5 RGB */ +#define PIPE_FORMAT_U_L8 4 /**< ubyte luminance */ +#define PIPE_FORMAT_U_A8 5 /**< ubyte alpha */ +#define PIPE_FORMAT_U_I8 6 /**< ubyte intensity */ +#define PIPE_FORMAT_U_L8_A8 7 /**< ubyte luminance, alpha */ +#define PIPE_FORMAT_YCBCR 8 +#define PIPE_FORMAT_YCBCR_REV 9 +#define PIPE_FORMAT_U_Z16 10 /**< ushort Z/depth */ +#define PIPE_FORMAT_U_Z32 11 /**< uint Z/depth */ +#define PIPE_FORMAT_F_Z32 12 /**< float Z/depth */ +#define PIPE_FORMAT_Z24_S8 13 /**< 24-bit Z + 8-bit stencil */ +#define PIPE_FORMAT_U_S8 14 /**< 8-bit stencil */ /** diff --git a/src/mesa/pipe/softpipe/sp_z_surface.c b/src/mesa/pipe/softpipe/sp_z_surface.c index 662a4a15ee..744737cb6c 100644 --- a/src/mesa/pipe/softpipe/sp_z_surface.c +++ b/src/mesa/pipe/softpipe/sp_z_surface.c @@ -36,38 +36,48 @@ #include "main/imports.h" #include "pipe/p_state.h" #include "pipe/p_defines.h" +#include "sp_context.h" #include "sp_surface.h" #include "sp_z_surface.h" static void* -z16_map(struct pipe_buffer *pb, GLuint access_mode) +z_map(struct pipe_buffer *pb, GLuint access_mode) { struct softpipe_surface *sps = (struct softpipe_surface *) pb; sps->surface.ptr = pb->ptr; + sps->surface.stride = sps->surface.width; return pb->ptr; } static void -z16_unmap(struct pipe_buffer *pb) +z_unmap(struct pipe_buffer *pb) { struct softpipe_surface *sps = (struct softpipe_surface *) pb; sps->surface.ptr = NULL; + sps->surface.stride = 0; } static void -z16_resize(struct pipe_surface *ps, GLuint width, GLuint height) +z_resize(struct pipe_surface *ps, GLuint width, GLuint height) { struct softpipe_surface *sps = (struct softpipe_surface *) ps; if (sps->surface.buffer.ptr) free(sps->surface.buffer.ptr); - ps->buffer.ptr = (GLubyte *) malloc(width * height * sizeof(GLushort)); + sps->surface.stride = sps->surface.width; + if (sps->surface.format == PIPE_FORMAT_U_Z16) + sps->surface.cpp = 2; + else if (sps->surface.format == PIPE_FORMAT_U_Z32 || + sps->surface.format == PIPE_FORMAT_Z24_S8) + sps->surface.cpp = 4; + else + assert(0); + + ps->buffer.ptr = (GLubyte *) malloc(width * height * sps->surface.cpp); ps->width = width; ps->height = height; - sps->surface.stride = sps->surface.width; - sps->surface.cpp = 2; } static void @@ -77,10 +87,12 @@ z16_read_quad_z(struct softpipe_surface *sps, const GLushort *src = (GLushort *) sps->surface.ptr + y * sps->surface.stride + x; + assert(sps->surface.format == PIPE_FORMAT_U_Z16); + /* converting GLushort to GLuint: */ zzzz[0] = src[0]; zzzz[1] = src[1]; - zzzz[2] = src[sps->surface.width]; + zzzz[2] = src[sps->surface.width + 0]; zzzz[3] = src[sps->surface.width + 1]; } @@ -90,29 +102,140 @@ z16_write_quad_z(struct softpipe_surface *sps, { GLushort *dst = (GLushort *) sps->surface.ptr + y * sps->surface.stride + x; + assert(sps->surface.format == PIPE_FORMAT_U_Z16); + /* converting GLuint to GLushort: */ dst[0] = zzzz[0]; dst[1] = zzzz[1]; - dst[sps->surface.width] = zzzz[2]; + dst[sps->surface.width + 0] = zzzz[2]; + dst[sps->surface.width + 1] = zzzz[3]; +} + +static void +z32_read_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) +{ + const GLuint *src + = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + + assert(sps->surface.format == PIPE_FORMAT_U_Z32); + + zzzz[0] = src[0]; + zzzz[1] = src[1]; + zzzz[2] = src[sps->surface.width + 0]; + zzzz[3] = src[sps->surface.width + 1]; +} + +static void +z32_write_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) +{ + GLuint *dst = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + + assert(sps->surface.format == PIPE_FORMAT_U_Z32); + + dst[0] = zzzz[0]; + dst[1] = zzzz[1]; + dst[sps->surface.width + 0] = zzzz[2]; dst[sps->surface.width + 1] = zzzz[3]; } +static void +z24s8_read_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) +{ + const GLuint *src + = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + zzzz[0] = src[0] >> 8; + zzzz[1] = src[1] >> 8; + zzzz[2] = src[sps->surface.width + 0] >> 8; + zzzz[3] = src[sps->surface.width + 1] >> 8; +} + +static void +z24s8_write_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) +{ + GLuint *dst = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(zzzz[0] <= 0xffffff); + + dst[0] = (dst[0] & 0xff) | (zzzz[0] << 8); + dst[1] = (dst[1] & 0xff) | (zzzz[1] << 8); + dst += sps->surface.width; + dst[0] = (dst[0] & 0xff) | (zzzz[2] << 8); + dst[1] = (dst[1] & 0xff) | (zzzz[3] << 8); +} + +static void +z24s8_read_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) +{ + const GLuint *src + = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + ssss[0] = src[0] & 0xff; + ssss[1] = src[1] & 0xff; + ssss[2] = src[sps->surface.width + 0] & 0xff; + ssss[3] = src[sps->surface.width + 1] & 0xff; +} + +static void +z24s8_write_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) +{ + GLuint *dst = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + dst[0] = (dst[0] & 0xffffff00) | ssss[0]; + dst[1] = (dst[1] & 0xffffff00) | ssss[1]; + dst += sps->surface.width; + dst[0] = (dst[0] & 0xffffff00) | ssss[2]; + dst[1] = (dst[1] & 0xffffff00) | ssss[3]; +} + + + +/** + * Create a new software-based Z buffer. + * \param format one of the PIPE_FORMAT_z* formats + */ struct softpipe_surface * -softpipe_new_z_surface(GLuint depth) +softpipe_new_z_surface(GLuint format) { struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); if (!sps) return NULL; - /* XXX ignoring depth param for now */ - - sps->surface.format = PIPE_FORMAT_U_Z16; - - sps->surface.resize = z16_resize; - sps->surface.buffer.map = z16_map; - sps->surface.buffer.unmap = z16_unmap; - sps->read_quad_z = z16_read_quad_z; - sps->write_quad_z = z16_write_quad_z; + sps->surface.format = format; + sps->surface.resize = z_resize; + sps->surface.buffer.map = z_map; + sps->surface.buffer.unmap = z_unmap; + + if (format == PIPE_FORMAT_U_Z16) { + sps->read_quad_z = z16_read_quad_z; + sps->write_quad_z = z16_write_quad_z; + } + else if (format == PIPE_FORMAT_U_Z32) { + sps->read_quad_z = z32_read_quad_z; + sps->write_quad_z = z32_write_quad_z; + } + else if (format == PIPE_FORMAT_Z24_S8) { + sps->read_quad_z = z24s8_read_quad_z; + sps->write_quad_z = z24s8_write_quad_z; + sps->read_quad_stencil = z24s8_read_quad_stencil; + sps->write_quad_stencil = z24s8_write_quad_stencil; + } + else { + assert(0); + } return sps; } diff --git a/src/mesa/pipe/softpipe/sp_z_surface.h b/src/mesa/pipe/softpipe/sp_z_surface.h index 9c3c43ca57..6a8d89a7c1 100644 --- a/src/mesa/pipe/softpipe/sp_z_surface.h +++ b/src/mesa/pipe/softpipe/sp_z_surface.h @@ -31,7 +31,7 @@ extern struct softpipe_surface * -softpipe_new_z_surface(GLuint depth); +softpipe_new_z_surface(GLuint format); #endif /* SP_Z_SURFACE_H */ -- cgit v1.2.3 From ed83159ff46ed6593a367ab93ce8b310dce91b77 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 08:22:18 -0600 Subject: fix assertions --- src/mesa/state_tracker/st_atom_framebuffer.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index a8d47ea4f4..7edd044ad9 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -61,13 +61,13 @@ update_framebuffer_state( struct st_context *st ) rb = st->ctx->DrawBuffer->_DepthBuffer; if (rb) { - assert(rb->surface); + assert(rb->Wrapped->surface); framebuffer.zbuf = rb->Wrapped->surface; } rb = st->ctx->DrawBuffer->_StencilBuffer; if (rb) { - assert(rb->surface); + assert(rb->Wrapped->surface); framebuffer.sbuf = rb->Wrapped->surface; } -- cgit v1.2.3 From 616c329e81f03183924b57cb19f027b22c9307f8 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 10:42:39 -0600 Subject: fix value returned by intel_new_renderbuffer_fb() --- src/mesa/drivers/dri/i915tex/intel_fbo.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c index 559788f20a..928139b5d9 100644 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ b/src/mesa/drivers/dri/i915tex/intel_fbo.c @@ -382,7 +382,7 @@ intel_new_renderbuffer_fb(GLuint intFormat) irb->Base.surface = intel_new_surface(intFormat); irb->Base.surface->rb = irb; - return &irb->Base; + return irb; } /** -- cgit v1.2.3 From 00afb6719938095bbf7daff3062127e4462bacc9 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 10:43:16 -0600 Subject: fix comments and param names for intel_miptree_image_map() --- src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c | 5 +++-- src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h | 2 +- 2 files changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c index 843a78eb82..ab3ec71f9e 100644 --- a/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c @@ -259,7 +259,8 @@ intel_miptree_image_offset(struct intel_mipmap_tree * mt, /** * Map a teximage in a mipmap tree. * \param row_stride returns row stride in bytes - * \param image_stride returns image stride in bytes (for 3D textures). + * \param image_offsets returns array [texture.depth] of image offsets + * (in bytes) for 3D textures. * \return address of mapping */ GLubyte * @@ -267,7 +268,7 @@ intel_miptree_image_map(struct intel_context * intel, struct intel_mipmap_tree * mt, GLuint face, GLuint level, - GLuint * row_stride, GLuint * image_offsets) + GLuint * row_stride, GLuint image_offsets[]) { DBG("%s \n", __FUNCTION__); diff --git a/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h index ecdb7be244..675a438c20 100644 --- a/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h +++ b/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h @@ -142,7 +142,7 @@ GLubyte *intel_miptree_image_map(struct intel_context *intel, struct intel_mipmap_tree *mt, GLuint face, GLuint level, - GLuint * row_stride, GLuint * image_stride); + GLuint * row_stride, GLuint image_offsets[]); void intel_miptree_image_unmap(struct intel_context *intel, struct intel_mipmap_tree *mt); -- cgit v1.2.3 From da2b82059f56d37893a318d0f7a7f2479db99c59 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 10:43:36 -0600 Subject: remove obsolete xmesa_get_stencil_surface() stub --- src/mesa/drivers/dri/i915tex/intel_surface.c | 11 ----------- 1 file changed, 11 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_surface.c b/src/mesa/drivers/dri/i915tex/intel_surface.c index 043c5aa5fe..6b34d23ac9 100644 --- a/src/mesa/drivers/dri/i915tex/intel_surface.c +++ b/src/mesa/drivers/dri/i915tex/intel_surface.c @@ -246,14 +246,3 @@ intel_new_surface(GLuint intFormat) return &sps->surface; } - - -struct pipe_surface * -xmesa_get_stencil_surface(GLcontext *ctx) -{ - /* XXX fix */ - return NULL; -} - - - -- cgit v1.2.3 From 6490c7d4aadac7d240d62e2d5c02d8973d69b71f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 11:15:30 -0600 Subject: fix stencil addressing bug --- src/mesa/drivers/dri/i915tex/intel_surface.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_surface.c b/src/mesa/drivers/dri/i915tex/intel_surface.c index 6b34d23ac9..3e86835de2 100644 --- a/src/mesa/drivers/dri/i915tex/intel_surface.c +++ b/src/mesa/drivers/dri/i915tex/intel_surface.c @@ -95,8 +95,9 @@ read_quad_z24(struct softpipe_surface *sps, /* extract lower three bytes */ zzzz[0] = src[0] & mask; zzzz[1] = src[1] & mask; - zzzz[2] = src[-sps->surface.stride] & mask; - zzzz[3] = src[-sps->surface.stride + 1] & mask; + src -= sps->surface.stride; + zzzz[2] = src[0] & mask; + zzzz[3] = src[1] & mask; } static void @@ -133,8 +134,9 @@ read_quad_stencil(struct softpipe_surface *sps, /* extract high byte */ ssss[0] = src[0] >> 24; ssss[1] = src[1] >> 24; - ssss[2] = src[-sps->surface.width] >> 24; - ssss[3] = src[-sps->surface.width + 1] >> 24; + src -= sps->surface.stride; + ssss[2] = src[0] >> 24; + ssss[3] = src[1] >> 24; } static void -- cgit v1.2.3 From f0b4d9cb24506d637a2e8ed06626277726afb321 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 12:49:11 -0600 Subject: In i915/i830_emit_state(), check if state->draw_region is non-null. This fixes a problem hit by glClear in the samples/stencil.c program. --- src/mesa/drivers/dri/i915tex/i830_vtbl.c | 15 ++++++++------- src/mesa/drivers/dri/i915tex/i915_vtbl.c | 15 ++++++++------- 2 files changed, 16 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/i830_vtbl.c b/src/mesa/drivers/dri/i915tex/i830_vtbl.c index 20c6900025..5cde80a469 100644 --- a/src/mesa/drivers/dri/i915tex/i830_vtbl.c +++ b/src/mesa/drivers/dri/i915tex/i830_vtbl.c @@ -452,13 +452,14 @@ i830_emit_state(struct intel_context *intel) if (dirty & I830_UPLOAD_BUFFERS) { DBG("I830_UPLOAD_BUFFERS:\n"); BEGIN_BATCH(I830_DEST_SETUP_SIZE + 2, 0); - OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR0]); - OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR1]); - OUT_RELOC(state->draw_region->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - state->draw_region->draw_offset); - + if (state->draw_region->buffer) { + OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR0]); + OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR1]); + OUT_RELOC(state->draw_region->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, + state->draw_region->draw_offset); + } if (state->depth_region) { OUT_BATCH(state->Buffer[I830_DESTREG_DBUFADDR0]); OUT_BATCH(state->Buffer[I830_DESTREG_DBUFADDR1]); diff --git a/src/mesa/drivers/dri/i915tex/i915_vtbl.c b/src/mesa/drivers/dri/i915tex/i915_vtbl.c index cc74ceae18..b836241a63 100644 --- a/src/mesa/drivers/dri/i915tex/i915_vtbl.c +++ b/src/mesa/drivers/dri/i915tex/i915_vtbl.c @@ -328,13 +328,14 @@ i915_emit_state(struct intel_context *intel) if (INTEL_DEBUG & DEBUG_STATE) fprintf(stderr, "I915_UPLOAD_BUFFERS:\n"); BEGIN_BATCH(I915_DEST_SETUP_SIZE + 2, 0); - OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR0]); - OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR1]); - OUT_RELOC(state->draw_region->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - state->draw_region->draw_offset); - + if (state->draw_region) { + OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR0]); + OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR1]); + OUT_RELOC(state->draw_region->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, + state->draw_region->draw_offset); + } if (state->depth_region) { OUT_BATCH(state->Buffer[I915_DESTREG_DBUFADDR0]); OUT_BATCH(state->Buffer[I915_DESTREG_DBUFADDR1]); -- cgit v1.2.3 From 973985b7ef19cd69c6bd5b85cd22e953d21aa8ae Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 13:03:18 -0600 Subject: simplify clear.x1,y2,x2,y2 setup --- src/mesa/drivers/dri/i915tex/intel_buffers.c | 16 ++++------------ 1 file changed, 4 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c index 72c7b5ae45..8944145afa 100644 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ b/src/mesa/drivers/dri/i915tex/intel_buffers.c @@ -308,22 +308,14 @@ intelClearWithTris(struct intel_context *intel, GLbitfield mask) LOCK_HARDWARE(intel); if (intel->numClipRects) { - GLint cx, cy, cw, ch; GLuint buf; intel->vtbl.install_meta_state(intel); - /* Get clear bounds after locking */ - cx = fb->_Xmin; - cy = fb->_Ymin; - ch = fb->_Ymax - cx; - cw = fb->_Xmax - cy; - - /* note: regardless of 'all', cx, cy, cw, ch are now correct */ - clear.x1 = cx; - clear.y1 = cy; - clear.x2 = cx + cw; - clear.y2 = cy + ch; + clear.x1 = fb->_Xmin; + clear.y1 = fb->_Ymin; + clear.x2 = fb->_Xmax; + clear.y2 = fb->_Ymax; /* Back and stencil cliprects are the same. Try and do both * buffers at once: -- cgit v1.2.3 From 280cb3713064ccc17417e1a25398ad615dfc53c8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 31 Jul 2007 20:06:09 +0100 Subject: A version of the i915tex driver with all drawing code removed. This is intended to support the softpipe development work. More code will be removed and pushed into softpipe until this basicially becomes the DRI/GLX interface for that driver. --- src/mesa/drivers/dri/i915pipe/Makefile | 49 ++ src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c | 342 +++++++++ src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h | 124 ++++ src/mesa/drivers/dri/i915pipe/intel_batchpool.c | 418 +++++++++++ src/mesa/drivers/dri/i915pipe/intel_blit.c | 527 ++++++++++++++ src/mesa/drivers/dri/i915pipe/intel_blit.h | 62 ++ .../drivers/dri/i915pipe/intel_buffer_objects.c | 257 +++++++ .../drivers/dri/i915pipe/intel_buffer_objects.h | 86 +++ src/mesa/drivers/dri/i915pipe/intel_buffers.c | 792 +++++++++++++++++++++ src/mesa/drivers/dri/i915pipe/intel_buffers.h | 55 ++ src/mesa/drivers/dri/i915pipe/intel_context.c | 768 ++++++++++++++++++++ src/mesa/drivers/dri/i915pipe/intel_context.h | 426 +++++++++++ src/mesa/drivers/dri/i915pipe/intel_depthstencil.c | 282 ++++++++ src/mesa/drivers/dri/i915pipe/intel_depthstencil.h | 14 + src/mesa/drivers/dri/i915pipe/intel_fbo.c | 623 ++++++++++++++++ src/mesa/drivers/dri/i915pipe/intel_fbo.h | 127 ++++ src/mesa/drivers/dri/i915pipe/intel_ioctl.c | 137 ++++ src/mesa/drivers/dri/i915pipe/intel_ioctl.h | 40 ++ src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c | 358 ++++++++++ src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h | 198 ++++++ src/mesa/drivers/dri/i915pipe/intel_pixel.c | 45 ++ src/mesa/drivers/dri/i915pipe/intel_pixel.h | 63 ++ src/mesa/drivers/dri/i915pipe/intel_reg.h | 88 +++ src/mesa/drivers/dri/i915pipe/intel_regions.c | 480 +++++++++++++ src/mesa/drivers/dri/i915pipe/intel_regions.h | 141 ++++ src/mesa/drivers/dri/i915pipe/intel_render.c | 244 +++++++ src/mesa/drivers/dri/i915pipe/intel_rotate.c | 237 ++++++ src/mesa/drivers/dri/i915pipe/intel_rotate.h | 39 + src/mesa/drivers/dri/i915pipe/intel_screen.c | 708 ++++++++++++++++++ src/mesa/drivers/dri/i915pipe/intel_screen.h | 129 ++++ src/mesa/drivers/dri/i915pipe/intel_span.c | 401 +++++++++++ src/mesa/drivers/dri/i915pipe/intel_span.h | 38 + src/mesa/drivers/dri/i915pipe/intel_state.c | 261 +++++++ src/mesa/drivers/dri/i915pipe/intel_structs.h | 132 ++++ src/mesa/drivers/dri/i915pipe/intel_surface.c | 259 +++++++ src/mesa/drivers/dri/i915pipe/intel_tex.c | 192 +++++ src/mesa/drivers/dri/i915pipe/intel_tex.h | 151 ++++ src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 293 ++++++++ src/mesa/drivers/dri/i915pipe/intel_tex_format.c | 172 +++++ src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 691 ++++++++++++++++++ src/mesa/drivers/dri/i915pipe/intel_tex_layout.c | 1 + src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c | 182 +++++ src/mesa/drivers/dri/i915pipe/intel_tex_validate.c | 272 +++++++ 43 files changed, 10904 insertions(+) create mode 100644 src/mesa/drivers/dri/i915pipe/Makefile create mode 100644 src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_batchpool.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_blit.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_blit.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_buffers.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_buffers.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_context.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_context.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_depthstencil.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_depthstencil.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_fbo.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_fbo.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_ioctl.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_ioctl.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_pixel.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_pixel.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_reg.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_regions.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_regions.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_render.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_rotate.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_rotate.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_screen.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_screen.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_span.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_span.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_state.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_structs.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_surface.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_copy.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_format.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_image.c create mode 120000 src/mesa/drivers/dri/i915pipe/intel_tex_layout.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c create mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_validate.c (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile new file mode 100644 index 0000000000..857d3fc985 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/Makefile @@ -0,0 +1,49 @@ + +TOP = ../../../../.. +include $(TOP)/configs/current + +LIBNAME = i915tex_dri.so + +MINIGLX_SOURCES = server/intel_dri.c + +DRIVER_SOURCES = \ + intel_regions.c \ + intel_buffer_objects.c \ + intel_batchbuffer.c \ + intel_mipmap_tree.c \ + intel_tex_layout.c \ + intel_tex_image.c \ + intel_tex_subimage.c \ + intel_tex_copy.c \ + intel_tex_validate.c \ + intel_tex_format.c \ + intel_tex.c \ + intel_pixel.c \ + intel_buffers.c \ + intel_blit.c \ + intel_context.c \ + intel_ioctl.c \ + intel_rotate.c \ + intel_screen.c \ + intel_span.c \ + intel_state.c \ + intel_surface.c \ + intel_fbo.c \ + intel_depthstencil.c \ + intel_batchpool.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(COMMON_BM_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") + +include ../Makefile.template + +intel_tex_layout.o: ../intel/intel_tex_layout.c + +symlinks: diff --git a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c new file mode 100644 index 0000000000..9259f2ac05 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c @@ -0,0 +1,342 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_batchbuffer.h" +#include "intel_ioctl.h" + +/* Relocations in kernel space: + * - pass dma buffer seperately + * - memory manager knows how to patch + * - pass list of dependent buffers + * - pass relocation list + * + * Either: + * - get back an offset for buffer to fire + * - memory manager knows how to fire buffer + * + * Really want the buffer to be AGP and pinned. + * + */ + +/* Cliprect fence: The highest fence protecting a dma buffer + * containing explicit cliprect information. Like the old drawable + * lock but irq-driven. X server must wait for this fence to expire + * before changing cliprects [and then doing sw rendering?]. For + * other dma buffers, the scheduler will grab current cliprect info + * and mix into buffer. X server must hold the lock while changing + * cliprects??? Make per-drawable. Need cliprects in shared memory + * -- beats storing them with every cmd buffer in the queue. + * + * ==> X server must wait for this fence to expire before touching the + * framebuffer with new cliprects. + * + * ==> Cliprect-dependent buffers associated with a + * cliprect-timestamp. All of the buffers associated with a timestamp + * must go to hardware before any buffer with a newer timestamp. + * + * ==> Dma should be queued per-drawable for correct X/GL + * synchronization. Or can fences be used for this? + * + * Applies to: Blit operations, metaops, X server operations -- X + * server automatically waits on its own dma to complete before + * modifying cliprects ??? + */ + +static void +intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) +{ + int i; + fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); + for (i = 0; i < count / 4; i += 4) + fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", + offset + i * 4, ptr[i], ptr[i + 1], ptr[i + 2], ptr[i + 3]); + fprintf(stderr, "END BATCH\n\n\n"); +} + +void +intel_batchbuffer_reset(struct intel_batchbuffer *batch) +{ + + int i; + + /* + * Get a new, free batchbuffer. + */ + + batch->size = batch->intel->intelScreen->maxBatchSize; + driBOData(batch->buffer, batch->size, NULL, 0); + + driBOResetList(&batch->list); + + /* + * Unreference buffers previously on the relocation list. + */ + + for (i = 0; i < batch->nr_relocs; i++) { + struct buffer_reloc *r = &batch->reloc[i]; + driBOUnReference(r->buf); + } + + batch->list_count = 0; + batch->nr_relocs = 0; + batch->flags = 0; + + /* + * We don't refcount the batchbuffer itself since we can't destroy it + * while it's on the list. + */ + + + driBOAddListItem(&batch->list, batch->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE); + + + batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); + batch->ptr = batch->map; +} + +/*====================================================================== + * Public functions + */ +struct intel_batchbuffer * +intel_batchbuffer_alloc(struct intel_context *intel) +{ + struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); + + batch->intel = intel; + + driGenBuffers(intel->intelScreen->batchPool, "batchbuffer", 1, + &batch->buffer, 4096, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); + batch->last_fence = NULL; + driBOCreateList(20, &batch->list); + intel_batchbuffer_reset(batch); + return batch; +} + +void +intel_batchbuffer_free(struct intel_batchbuffer *batch) +{ + if (batch->last_fence) { + driFenceFinish(batch->last_fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); + driFenceUnReference(batch->last_fence); + batch->last_fence = NULL; + } + if (batch->map) { + driBOUnmap(batch->buffer); + batch->map = NULL; + } + driBOUnReference(batch->buffer); + batch->buffer = NULL; + free(batch); +} + +/* TODO: Push this whole function into bufmgr. + */ +static void +do_flush_locked(struct intel_batchbuffer *batch, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) +{ + GLuint *ptr; + GLuint i; + struct intel_context *intel = batch->intel; + unsigned fenceFlags; + struct _DriFenceObject *fo; + + driBOValidateList(batch->intel->driFd, &batch->list); + + /* Apply the relocations. This nasty map indicates to me that the + * whole task should be done internally by the memory manager, and + * that dma buffers probably need to be pinned within agp space. + */ + ptr = (GLuint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, + DRM_BO_HINT_ALLOW_UNFENCED_MAP); + + + for (i = 0; i < batch->nr_relocs; i++) { + struct buffer_reloc *r = &batch->reloc[i]; + + ptr[r->offset / 4] = driBOOffset(r->buf) + r->delta; + } + + if (INTEL_DEBUG & DEBUG_BATCH) + intel_dump_batchbuffer(0, ptr, used); + + driBOUnmap(batch->buffer); + batch->map = NULL; + + /* Throw away non-effective packets. Won't work once we have + * hardware contexts which would preserve statechanges beyond a + * single buffer. + */ + + if (!(intel->numClipRects == 0 && !ignore_cliprects)) { + intel_batch_ioctl(batch->intel, + driBOOffset(batch->buffer), + used, ignore_cliprects, allow_unlock); + } + + + /* + * Kernel fencing. The flags tells the kernel that we've + * programmed an MI_FLUSH. + */ + + fenceFlags = DRM_I915_FENCE_FLAG_FLUSHED; + fo = driFenceBuffers(batch->intel->driFd, + "Batch fence", fenceFlags); + + /* + * User space fencing. + */ + + driBOFence(batch->buffer, fo); + + if (driFenceType(fo) == DRM_FENCE_TYPE_EXE) { + + /* + * Oops. We only validated a batch buffer. This means we + * didn't do any proper rendering. Discard this fence object. + */ + + driFenceUnReference(fo); + } else { + driFenceUnReference(batch->last_fence); + batch->last_fence = fo; + for (i = 0; i < batch->nr_relocs; i++) { + struct buffer_reloc *r = &batch->reloc[i]; + driBOFence(r->buf, fo); + } + } + + if (intel->numClipRects == 0 && !ignore_cliprects) { + if (allow_unlock) { + UNLOCK_HARDWARE(intel); + sched_yield(); + LOCK_HARDWARE(intel); + } +// intel->vtbl.lost_hardware(intel); + } +} + + +struct _DriFenceObject * +intel_batchbuffer_flush(struct intel_batchbuffer *batch) +{ + struct intel_context *intel = batch->intel; + GLuint used = batch->ptr - batch->map; + GLboolean was_locked = intel->locked; + + if (used == 0) + return batch->last_fence; + +#define MI_FLUSH ((0<<29)|(4<<23)) + + /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a + * performance drain that we would like to avoid. + */ + if (used & 4) { + ((int *) batch->ptr)[0] = MI_FLUSH; + ((int *) batch->ptr)[1] = 0; + ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->ptr)[0] = MI_FLUSH; + ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; + used += 8; + } + + driBOUnmap(batch->buffer); + batch->ptr = NULL; + batch->map = NULL; + + /* TODO: Just pass the relocation list and dma buffer up to the + * kernel. + */ + if (!was_locked) + LOCK_HARDWARE(intel); + + do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), + GL_FALSE); + + if (!was_locked) + UNLOCK_HARDWARE(intel); + + /* Reset the buffer: + */ + intel_batchbuffer_reset(batch); + return batch->last_fence; +} + +void +intel_batchbuffer_finish(struct intel_batchbuffer *batch) +{ + struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); + driFenceReference(fence); + driFenceFinish(fence, 3, GL_FALSE); + driFenceUnReference(fence); +} + + +/* This is the only way buffers get added to the validate list. + */ +GLboolean +intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, + struct _DriBufferObject *buffer, + GLuint flags, GLuint mask, GLuint delta) +{ + assert(batch->nr_relocs < MAX_RELOCS); + + driBOAddListItem(&batch->list, buffer, flags, mask); + + { + struct buffer_reloc *r = &batch->reloc[batch->nr_relocs++]; + driBOReference(buffer); + r->buf = buffer; + r->offset = batch->ptr - batch->map; + r->delta = delta; + } + + batch->ptr += 4; + return GL_TRUE; +} + + + +void +intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, GLuint bytes, GLuint flags) +{ + assert((bytes & 3) == 0); + intel_batchbuffer_require_space(batch, bytes, flags); + __memcpy(batch->ptr, data, bytes); + batch->ptr += bytes; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h new file mode 100644 index 0000000000..212f130101 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h @@ -0,0 +1,124 @@ +#ifndef INTEL_BATCHBUFFER_H +#define INTEL_BATCHBUFFER_H + +#include "mtypes.h" +#include "dri_bufmgr.h" + +struct intel_context; + +#define BATCH_SZ 16384 +#define BATCH_RESERVED 16 + +#define MAX_RELOCS 4096 + +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 + +struct buffer_reloc +{ + struct _DriBufferObject *buf; + GLuint offset; + GLuint delta; /* not needed? */ +}; + +struct intel_batchbuffer +{ + struct bufmgr *bm; + struct intel_context *intel; + + struct _DriBufferObject *buffer; + struct _DriFenceObject *last_fence; + GLuint flags; + + drmBOList list; + GLuint list_count; + GLubyte *map; + GLubyte *ptr; + + struct buffer_reloc reloc[MAX_RELOCS]; + GLuint nr_relocs; + GLuint size; +}; + +struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context + *intel); + +void intel_batchbuffer_free(struct intel_batchbuffer *batch); + + +void intel_batchbuffer_finish(struct intel_batchbuffer *batch); + +struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer + *batch); + +void intel_batchbuffer_reset(struct intel_batchbuffer *batch); + + +/* Unlike bmBufferData, this currently requires the buffer be mapped. + * Consider it a convenience function wrapping multple + * intel_buffer_dword() calls. + */ +void intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, GLuint bytes, GLuint flags); + +void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, + GLuint bytes); + +GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, + struct _DriBufferObject *buffer, + GLuint flags, + GLuint mask, GLuint offset); + +/* Inline functions - might actually be better off with these + * non-inlined. Certainly better off switching all command packets to + * be passed as structs rather than dwords, but that's a little bit of + * work... + */ +static INLINE GLuint +intel_batchbuffer_space(struct intel_batchbuffer *batch) +{ + return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); +} + + +static INLINE void +intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) +{ + assert(batch->map); + assert(intel_batchbuffer_space(batch) >= 4); + *(GLuint *) (batch->ptr) = dword; + batch->ptr += 4; +} + +static INLINE void +intel_batchbuffer_require_space(struct intel_batchbuffer *batch, + GLuint sz, GLuint flags) +{ + assert(sz < batch->size - 8); + if (intel_batchbuffer_space(batch) < sz || + (batch->flags != 0 && flags != 0 && batch->flags != flags)) + intel_batchbuffer_flush(batch); + + batch->flags |= flags; +} + +/* Here are the crusty old macros, to be removed: + */ +#define BATCH_LOCALS + +#define BEGIN_BATCH(n, flags) do { \ + assert(!intel->prim.flush); \ + intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ +} while (0) + +#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) + +#define OUT_RELOC(buf,flags,mask,delta) do { \ + assert((delta) >= 0); \ + intel_batchbuffer_emit_reloc(intel->batch, buf, flags, mask, delta); \ +} while (0) + +#define ADVANCE_BATCH() do { } while(0) + + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_batchpool.c b/src/mesa/drivers/dri/i915pipe/intel_batchpool.c new file mode 100644 index 0000000000..2503b8a62a --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_batchpool.c @@ -0,0 +1,418 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "imports.h" +#include "glthread.h" +#include "dri_bufpool.h" +#include "dri_bufmgr.h" +#include "intel_screen.h" + +typedef struct +{ + drmMMListHead head; + struct _BPool *parent; + struct _DriFenceObject *fence; + unsigned long start; + int unfenced; + int mapped; +} BBuf; + +typedef struct _BPool +{ + _glthread_Mutex mutex; + unsigned long bufSize; + unsigned poolSize; + unsigned numFree; + unsigned numTot; + unsigned numDelayed; + unsigned checkDelayed; + drmMMListHead free; + drmMMListHead delayed; + drmMMListHead head; + drmBO kernelBO; + void *virtual; + BBuf *bufs; +} BPool; + + +static BPool * +createBPool(int fd, unsigned long bufSize, unsigned numBufs, unsigned flags, + unsigned checkDelayed) +{ + BPool *p = (BPool *) malloc(sizeof(*p)); + BBuf *buf; + int i; + + if (!p) + return NULL; + + p->bufs = (BBuf *) malloc(numBufs * sizeof(*p->bufs)); + if (!p->bufs) { + free(p); + return NULL; + } + + DRMINITLISTHEAD(&p->free); + DRMINITLISTHEAD(&p->head); + DRMINITLISTHEAD(&p->delayed); + + p->numTot = numBufs; + p->numFree = numBufs; + p->bufSize = bufSize; + p->numDelayed = 0; + p->checkDelayed = checkDelayed; + + _glthread_INIT_MUTEX(p->mutex); + + if (drmBOCreate(fd, 0, numBufs * bufSize, 0, NULL, drm_bo_type_dc, + flags, DRM_BO_HINT_DONT_FENCE, &p->kernelBO)) { + free(p->bufs); + free(p); + return NULL; + } + if (drmBOMap(fd, &p->kernelBO, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, + &p->virtual)) { + drmBODestroy(fd, &p->kernelBO); + free(p->bufs); + free(p); + return NULL; + } + + /* + * We unmap the buffer so that we can validate it later. Note that this is + * just a synchronizing operation. The buffer will have a virtual mapping + * until it is destroyed. + */ + + drmBOUnmap(fd, &p->kernelBO); + + buf = p->bufs; + for (i = 0; i < numBufs; ++i) { + buf->parent = p; + buf->fence = NULL; + buf->start = i * bufSize; + buf->mapped = 0; + buf->unfenced = 0; + DRMLISTADDTAIL(&buf->head, &p->free); + buf++; + } + + return p; +} + + +static void +pool_checkFree(BPool * p, int wait) +{ + drmMMListHead *list, *prev; + BBuf *buf; + int signaled = 0; + int i; + + list = p->delayed.next; + + if (p->numDelayed > 3) { + for (i = 0; i < p->numDelayed; i += 3) { + list = list->next; + } + } + + prev = list->prev; + for (; list != &p->delayed; list = prev, prev = list->prev) { + + buf = DRMLISTENTRY(BBuf, list, head); + + if (!signaled) { + if (wait) { + driFenceFinish(buf->fence, DRM_FENCE_TYPE_EXE, 1); + signaled = 1; + } + else { + signaled = driFenceSignaled(buf->fence, DRM_FENCE_TYPE_EXE); + } + } + + if (!signaled) + break; + + driFenceUnReference(buf->fence); + buf->fence = NULL; + DRMLISTDEL(list); + p->numDelayed--; + DRMLISTADD(list, &p->free); + p->numFree++; + } +} + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, unsigned flags, unsigned hint, + unsigned alignment) +{ + BPool *p = (BPool *) pool->data; + + drmMMListHead *item; + + if (alignment && (alignment != 4096)) + return NULL; + + _glthread_LOCK_MUTEX(p->mutex); + + if (p->numFree == 0) + pool_checkFree(p, GL_TRUE); + + if (p->numFree == 0) { + fprintf(stderr, "Out of fixed size buffer objects\n"); + BM_CKFATAL(-ENOMEM); + } + + item = p->free.next; + + if (item == &p->free) { + fprintf(stderr, "Fixed size buffer pool corruption\n"); + } + + DRMLISTDEL(item); + --p->numFree; + + _glthread_UNLOCK_MUTEX(p->mutex); + return (void *) DRMLISTENTRY(BBuf, item, head); +} + + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + _glthread_LOCK_MUTEX(p->mutex); + + if (buf->fence) { + DRMLISTADDTAIL(&buf->head, &p->delayed); + p->numDelayed++; + } + else { + buf->unfenced = 0; + DRMLISTADD(&buf->head, &p->free); + p->numFree++; + } + + if ((p->numDelayed % p->checkDelayed) == 0) + pool_checkFree(p, 0); + + _glthread_UNLOCK_MUTEX(p->mutex); + return 0; +} + + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, void **virtual) +{ + + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + _glthread_LOCK_MUTEX(p->mutex); + + /* + * Currently Mesa doesn't have any condition variables to resolve this + * cleanly in a multithreading environment. + * We bail out instead. + */ + + if (buf->mapped) { + fprintf(stderr, "Trying to map already mapped buffer object\n"); + BM_CKFATAL(-EINVAL); + } + +#if 0 + if (buf->unfenced && !(hint & DRM_BO_HINT_ALLOW_UNFENCED_MAP)) { + fprintf(stderr, "Trying to map an unfenced buffer object 0x%08x" + " 0x%08x %d\n", hint, flags, buf->start); + BM_CKFATAL(-EINVAL); + } + +#endif + + if (buf->fence) { + _glthread_UNLOCK_MUTEX(p->mutex); + return -EBUSY; + } + + buf->mapped = GL_TRUE; + *virtual = (unsigned char *) p->virtual + buf->start; + _glthread_UNLOCK_MUTEX(p->mutex); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, int lazy) +{ + BBuf *buf = (BBuf *) private; + driFenceFinish(buf->fence, 0, lazy); + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + + buf->mapped = 0; + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + return p->kernelBO.offset + buf->start; +} + +static unsigned +pool_flags(struct _DriBufferPool *pool, void *private) +{ + BPool *p = (BPool *) pool->data; + + return p->kernelBO.flags; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + BPool *p = (BPool *) pool->data; + + return p->bufSize; +} + + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + _glthread_LOCK_MUTEX(p->mutex); + if (buf->fence) { + driFenceUnReference(buf->fence); + } + buf->fence = fence; + buf->unfenced = 0; + driFenceReference(buf->fence); + _glthread_UNLOCK_MUTEX(p->mutex); + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + return &p->kernelBO; +} + +static int +pool_validate(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + _glthread_LOCK_MUTEX(p->mutex); + buf->unfenced = GL_TRUE; + _glthread_UNLOCK_MUTEX(p->mutex); + return 0; +} + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + BPool *p = (BPool *) pool->data; + + /* + * Wait on outstanding fences. + */ + + _glthread_LOCK_MUTEX(p->mutex); + while ((p->numFree < p->numTot) && p->numDelayed) { + _glthread_UNLOCK_MUTEX(p->mutex); + sched_yield(); + pool_checkFree(p, GL_TRUE); + _glthread_LOCK_MUTEX(p->mutex); + } + + drmBODestroy(pool->fd, &p->kernelBO); + free(p->bufs); + _glthread_UNLOCK_MUTEX(p->mutex); + free(p); + free(pool); +} + + +struct _DriBufferPool * +driBatchPoolInit(int fd, unsigned flags, + unsigned long bufSize, + unsigned numBufs, unsigned checkDelayed) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + if (!pool) + return NULL; + + pool->data = createBPool(fd, bufSize, numBufs, flags, checkDelayed); + if (!pool->data) + return NULL; + + pool->fd = fd; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = &pool_validate; + pool->waitIdle = &pool_waitIdle; + pool->setstatic = NULL; + pool->takeDown = &pool_takedown; + return pool; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.c b/src/mesa/drivers/dri/i915pipe/intel_blit.c new file mode 100644 index 0000000000..28441f4b83 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_blit.c @@ -0,0 +1,527 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include +#include + +#include "mtypes.h" +#include "context.h" +#include "enums.h" + +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_buffers.h" +#include "intel_context.h" +#include "intel_fbo.h" +#include "intel_reg.h" +#include "intel_regions.h" +#include "vblank.h" + +#define FILE_DEBUG_FLAG DEBUG_BLIT + +/** + * Copy the back color buffer to the front color buffer. + * Used for SwapBuffers(). + */ +void +intelCopyBuffer(__DRIdrawablePrivate * dPriv, + const drm_clip_rect_t * rect) +{ + + struct intel_context *intel; + const intelScreenPrivate *intelScreen; + + DBG("%s\n", __FUNCTION__); + + assert(dPriv); + + intel = intelScreenContext(dPriv->driScreenPriv->private); + if (!intel) + return; + + intelScreen = intel->intelScreen; + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + current context (which is what intelScreenContext should return) might + not get a contended lock and thus cliprects not updated (tests/manywin) */ + if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + + + if (dPriv && dPriv->numClipRects) { + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + const struct intel_region *frontRegion + = intelScreen->front_region; + const struct intel_region *backRegion + = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? + intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : + intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); + const int backWidth = intel_fb->Base.Width; + const int backHeight = intel_fb->Base.Height; + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + const int pitch = frontRegion->pitch; + const int srcpitch = backRegion->pitch; + const int cpp = frontRegion->cpp; + int BR13, CMD; + int i; + + ASSERT(intel_fb); + ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ + ASSERT(frontRegion); + ASSERT(backRegion); + ASSERT(frontRegion->cpp == backRegion->cpp); + + DBG("front pitch %d back pitch %d\n", + frontRegion->pitch, backRegion->pitch); + + if (cpp == 2) { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + } + else { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + } + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + drm_clip_rect_t sbox; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->width || pbox->y2 > intelScreen->height) + continue; + + box = *pbox; + + if (rect) { + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > backWidth) + box.x2 = backWidth + box.x1; + if (box.y2 - box.y1 > backHeight) + box.y2 = backHeight + box.y1; + + DBG("box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((box.y1 << 16) | box.x1); + OUT_BATCH((box.y2 << 16) | box.x2); + + OUT_RELOC(frontRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + OUT_BATCH((sbox.y1 << 16) | sbox.x1); + OUT_BATCH((srcpitch * cpp) & 0xffff); + OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + ADVANCE_BATCH(); + } + + if (intel->first_swap_fence) + driFenceUnReference(intel->first_swap_fence); + intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); + driFenceReference(intel->first_swap_fence); + } + + UNLOCK_HARDWARE(intel); + + /* XXX this is bogus. The context here may not even be bound to this drawable! */ + if (intel->lastStamp != dPriv->lastStamp) { + GET_CURRENT_CONTEXT(currctx); + struct intel_context *intelcurrent = intel_context(currctx); + if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { + intelWindowMoved(intel); + intel->lastStamp = dPriv->lastStamp; + } + } + +} + + + + +void +intelEmitFillBlit(struct intel_context *intel, + GLuint cpp, + GLshort dst_pitch, + struct _DriBufferObject *dst_buffer, + GLuint dst_offset, + GLshort x, GLshort y, GLshort w, GLshort h, GLuint color) +{ + GLuint BR13, CMD; + BATCH_LOCALS; + + dst_pitch *= cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); + CMD = XY_COLOR_BLT_CMD; + break; + case 4: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); + CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | + XY_COLOR_BLT_WRITE_RGB); + break; + default: + return; + } + + DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); + + + BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((y << 16) | x); + OUT_BATCH(((y + h) << 16) | (x + w)); + OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); + OUT_BATCH(color); + ADVANCE_BATCH(); +} + + +static GLuint translate_raster_op(GLenum logicop) +{ + switch(logicop) { + case GL_CLEAR: return 0x00; + case GL_AND: return 0x88; + case GL_AND_REVERSE: return 0x44; + case GL_COPY: return 0xCC; + case GL_AND_INVERTED: return 0x22; + case GL_NOOP: return 0xAA; + case GL_XOR: return 0x66; + case GL_OR: return 0xEE; + case GL_NOR: return 0x11; + case GL_EQUIV: return 0x99; + case GL_INVERT: return 0x55; + case GL_OR_REVERSE: return 0xDD; + case GL_COPY_INVERTED: return 0x33; + case GL_OR_INVERTED: return 0xBB; + case GL_NAND: return 0x77; + case GL_SET: return 0xFF; + default: return 0; + } +} + + +/* Copy BitBlt + */ +void +intelEmitCopyBlit(struct intel_context *intel, + GLuint cpp, + GLshort src_pitch, + struct _DriBufferObject *src_buffer, + GLuint src_offset, + GLshort dst_pitch, + struct _DriBufferObject *dst_buffer, + GLuint dst_offset, + GLshort src_x, GLshort src_y, + GLshort dst_x, GLshort dst_y, + GLshort w, GLshort h, + GLenum logic_op) +{ + GLuint CMD, BR13; + int dst_y2 = dst_y + h; + int dst_x2 = dst_x + w; + BATCH_LOCALS; + + + DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + src_buffer, src_pitch, src_offset, src_x, src_y, + dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); + + src_pitch *= cpp; + dst_pitch *= cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = (((GLint) dst_pitch) & 0xffff) | + (translate_raster_op(logic_op) << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + break; + case 4: + BR13 = + (((GLint) dst_pitch) & 0xffff) | + (translate_raster_op(logic_op) << 16) | (1 << 24) | (1 << 25); + CMD = + (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + break; + default: + return; + } + + if (dst_y2 < dst_y || dst_x2 < dst_x) { + return; + } + + /* Initial y values don't seem to work with negative pitches. If + * we adjust the offsets manually (below), it seems to work fine. + * + * On the other hand, if we always adjust, the hardware doesn't + * know which blit directions to use, so overlapping copypixels get + * the wrong result. + */ + if (dst_pitch > 0 && src_pitch > 0) { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((dst_y << 16) | dst_x); + OUT_BATCH((dst_y2 << 16) | dst_x2); + OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); + OUT_BATCH((src_y << 16) | src_x); + OUT_BATCH(((GLint) src_pitch & 0xffff)); + OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, src_offset); + ADVANCE_BATCH(); + } + else { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((0 << 16) | dst_x); + OUT_BATCH((h << 16) | dst_x2); + OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, + dst_offset + dst_y * dst_pitch); + OUT_BATCH((0 << 16) | src_x); + OUT_BATCH(((GLint) src_pitch & 0xffff)); + OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, + src_offset + src_y * src_pitch); + ADVANCE_BATCH(); + } +} + + +/** + * Use blitting to clear the renderbuffers named by 'flags'. + * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferMask field + * since that might include software renderbuffers or renderbuffers + * which we're clearing with triangles. + * \param mask bitmask of BUFFER_BIT_* values indicating buffers to clear + */ +void +intelClearWithBlit(GLcontext * ctx, GLbitfield mask) +{ + struct intel_context *intel = intel_context(ctx); + struct gl_framebuffer *fb = ctx->DrawBuffer; + GLuint clear_depth; + GLbitfield skipBuffers = 0; + BATCH_LOCALS; + + DBG("%s %x\n", __FUNCTION__, mask); + + /* + * Compute values for clearing the buffers. + */ + clear_depth = 0; + if (mask & BUFFER_BIT_DEPTH) { + clear_depth = (GLuint) (fb->_DepthMax * ctx->Depth.Clear); + } + if (mask & BUFFER_BIT_STENCIL) { + clear_depth |= (ctx->Stencil.Clear & 0xff) << 24; + } + + /* If clearing both depth and stencil, skip BUFFER_BIT_STENCIL in + * the loop below. + */ + if ((mask & BUFFER_BIT_DEPTH) && (mask & BUFFER_BIT_STENCIL)) { + skipBuffers = BUFFER_BIT_STENCIL; + } + + /* XXX Move this flush/lock into the following conditional? */ + intelFlush(&intel->ctx); + LOCK_HARDWARE(intel); + + if (intel->numClipRects) { + GLint cx, cy, cw, ch; + drm_clip_rect_t b; + + /* Get clear bounds after locking */ + cx = fb->_Xmin; + cy = fb->_Ymin; + cw = fb->_Xmax - cx; + ch = fb->_Ymax - cy; + + if (fb->Name == 0) { + /* clearing a window */ + /* flip top to bottom */ + b.x1 = cx; + b.y1 = fb->Height - cy - ch; + b.x2 = b.x1 + cw; + b.y2 = b.y1 + ch; + } + else { + /* clearing FBO */ + b.x1 = cx; + b.y1 = cy; + b.x2 = b.x1 + cw; + b.y2 = b.y1 + ch; + /* no change to mask */ + } + + { + GLuint buf; + GLuint clearMask = mask; /* use copy, since we modify it below */ + GLboolean all = (cw == fb->Width && ch == fb->Height); + + DBG("clear %d,%d..%d,%d, mask %x\n", + b.x1, b.y1, b.x2, b.y2, mask); + + /* Loop over all renderbuffers */ + for (buf = 0; buf < BUFFER_COUNT && clearMask; buf++) { + const GLbitfield bufBit = 1 << buf; + if ((clearMask & bufBit) && !(bufBit & skipBuffers)) { + /* OK, clear this renderbuffer */ + struct intel_region *irb_region = + intel_get_rb_region(fb, buf); + struct _DriBufferObject *write_buffer = + intel_region_buffer(intel->intelScreen, irb_region, + all ? INTEL_WRITE_FULL : + INTEL_WRITE_PART); + + GLuint clearVal; + GLint pitch, cpp; + GLuint BR13, CMD; + + ASSERT(irb_region); + + pitch = irb_region->pitch; + cpp = irb_region->cpp; + + DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + irb_region->buffer, (pitch * cpp), + irb_region->draw_offset, + b.x1, b.y1, b.x2 - b.x1, b.y2 - b.y1); + + + /* Setup the blit command */ + if (cpp == 4) { + BR13 = (0xF0 << 16) | (pitch * cpp) | (1 << 24) | (1 << 25); + if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) { + CMD = XY_COLOR_BLT_CMD; + if (clearMask & BUFFER_BIT_DEPTH) + CMD |= XY_COLOR_BLT_WRITE_RGB; + if (clearMask & BUFFER_BIT_STENCIL) + CMD |= XY_COLOR_BLT_WRITE_ALPHA; + } + else { + /* clearing RGBA */ + CMD = (XY_COLOR_BLT_CMD | + XY_COLOR_BLT_WRITE_ALPHA | + XY_COLOR_BLT_WRITE_RGB); + } + } + else { + ASSERT(cpp == 2 || cpp == 0); + BR13 = (0xF0 << 16) | (pitch * cpp) | (1 << 24); + CMD = XY_COLOR_BLT_CMD; + } + + if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) { + clearVal = clear_depth; + } + else { + clearVal = (cpp == 4) + ? intel->ClearColor8888 : intel->ClearColor565; + } + /* + _mesa_debug(ctx, "hardware blit clear buf %d rb id %d\n", + buf, irb->Base.Name); + */ + intel_wait_flips(intel, INTEL_BATCH_NO_CLIPRECTS); + + BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((b.y1 << 16) | b.x1); + OUT_BATCH((b.y2 << 16) | b.x2); + OUT_RELOC(write_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, + irb_region->draw_offset); + OUT_BATCH(clearVal); + ADVANCE_BATCH(); + clearMask &= ~bufBit; /* turn off bit, for faster loop exit */ + } + } + } + intel_batchbuffer_flush(intel->batch); + } + + UNLOCK_HARDWARE(intel); +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.h b/src/mesa/drivers/dri/i915pipe/intel_blit.h new file mode 100644 index 0000000000..77686444fa --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_blit.h @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BLIT_H +#define INTEL_BLIT_H + +#include "intel_context.h" +#include "intel_ioctl.h" +#include "dri_bufmgr.h" + +extern void intelCopyBuffer(__DRIdrawablePrivate * dpriv, + const drm_clip_rect_t * rect); + +extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask); + +extern void intelEmitCopyBlit(struct intel_context *intel, + GLuint cpp, + GLshort src_pitch, + struct _DriBufferObject *src_buffer, + GLuint src_offset, + GLshort dst_pitch, + struct _DriBufferObject *dst_buffer, + GLuint dst_offset, + GLshort srcx, GLshort srcy, + GLshort dstx, GLshort dsty, + GLshort w, GLshort h, + GLenum logicop ); + +extern void intelEmitFillBlit(struct intel_context *intel, + GLuint cpp, + GLshort dst_pitch, + struct _DriBufferObject *dst_buffer, + GLuint dst_offset, + GLshort x, GLshort y, + GLshort w, GLshort h, GLuint color); + + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c new file mode 100644 index 0000000000..91c45ad95b --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c @@ -0,0 +1,257 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "imports.h" +#include "mtypes.h" +#include "bufferobj.h" + +#include "intel_context.h" +#include "intel_buffer_objects.h" +#include "intel_regions.h" +#include "dri_bufmgr.h" + +/** + * There is some duplication between mesa's bufferobjects and our + * bufmgr buffers. Both have an integer handle and a hashtable to + * lookup an opaque structure. It would be nice if the handles and + * internal structure where somehow shared. + */ +static struct gl_buffer_object * +intel_bufferobj_alloc(GLcontext * ctx, GLuint name, GLenum target) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *obj = CALLOC_STRUCT(intel_buffer_object); + + _mesa_initialize_buffer_object(&obj->Base, name, target); + + driGenBuffers(intel->intelScreen->regionPool, + "bufferobj", 1, &obj->buffer, 64, + DRM_BO_FLAG_MEM_LOCAL | + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0); + + return &obj->Base; +} + + +/* Break the COW tie to the region. The region gets to keep the data. + */ +void +intel_bufferobj_release_region(struct intel_context *intel, + struct intel_buffer_object *intel_obj) +{ + assert(intel_obj->region->buffer == intel_obj->buffer); + intel_obj->region->pbo = NULL; + intel_obj->region = NULL; + driBOUnReference(intel_obj->buffer); + intel_obj->buffer = NULL; + + /* This leads to a large number of buffer deletion/creation events. + * Currently the drm doesn't like that: + */ + driGenBuffers(intel->intelScreen->regionPool, + "buffer object", 1, &intel_obj->buffer, 64, 0, 0); + LOCK_HARDWARE(intel); + driBOData(intel_obj->buffer, intel_obj->Base.Size, NULL, 0); + UNLOCK_HARDWARE(intel); +} + +/* Break the COW tie to the region. Both the pbo and the region end + * up with a copy of the data. + */ +void +intel_bufferobj_cow(struct intel_context *intel, + struct intel_buffer_object *intel_obj) +{ + assert(intel_obj->region); + intel_region_cow(intel->intelScreen, intel_obj->region); +} + + +/** + * Deallocate/free a vertex/pixel buffer object. + * Called via glDeleteBuffersARB(). + */ +static void +intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + assert(intel_obj); + + if (intel_obj->region) { + intel_bufferobj_release_region(intel, intel_obj); + } + else if (intel_obj->buffer) { + driDeleteBuffers(1, &intel_obj->buffer); + } + + _mesa_free(intel_obj); +} + + + +/** + * Allocate space for and store data in a buffer object. Any data that was + * previously stored in the buffer object is lost. If data is NULL, + * memory will be allocated, but no copy will occur. + * Called via glBufferDataARB(). + */ +static void +intel_bufferobj_data(GLcontext * ctx, + GLenum target, + GLsizeiptrARB size, + const GLvoid * data, + GLenum usage, struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + intel_obj->Base.Size = size; + intel_obj->Base.Usage = usage; + + if (intel_obj->region) + intel_bufferobj_release_region(intel, intel_obj); + + LOCK_HARDWARE(intel); + driBOData(intel_obj->buffer, size, data, 0); + UNLOCK_HARDWARE(intel); +} + + +/** + * Replace data in a subrange of buffer object. If the data range + * specified by size + offset extends beyond the end of the buffer or + * if data is NULL, no copy is performed. + * Called via glBufferSubDataARB(). + */ +static void +intel_bufferobj_subdata(GLcontext * ctx, + GLenum target, + GLintptrARB offset, + GLsizeiptrARB size, + const GLvoid * data, struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + assert(intel_obj); + + if (intel_obj->region) + intel_bufferobj_cow(intel, intel_obj); + + driBOSubData(intel_obj->buffer, offset, size, data); +} + + +/** + * Called via glGetBufferSubDataARB(). + */ +static void +intel_bufferobj_get_subdata(GLcontext * ctx, + GLenum target, + GLintptrARB offset, + GLsizeiptrARB size, + GLvoid * data, struct gl_buffer_object *obj) +{ + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + assert(intel_obj); + driBOGetSubData(intel_obj->buffer, offset, size, data); +} + + + +/** + * Called via glMapBufferARB(). + */ +static void * +intel_bufferobj_map(GLcontext * ctx, + GLenum target, + GLenum access, struct gl_buffer_object *obj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + /* XXX: Translate access to flags arg below: + */ + assert(intel_obj); + + if (intel_obj->region) + intel_bufferobj_cow(intel, intel_obj); + + obj->Pointer = driBOMap(intel_obj->buffer, + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); + return obj->Pointer; +} + + +/** + * Called via glMapBufferARB(). + */ +static GLboolean +intel_bufferobj_unmap(GLcontext * ctx, + GLenum target, struct gl_buffer_object *obj) +{ + struct intel_buffer_object *intel_obj = intel_buffer_object(obj); + + assert(intel_obj); + assert(obj->Pointer); + driBOUnmap(intel_obj->buffer); + obj->Pointer = NULL; + return GL_TRUE; +} + +struct _DriBufferObject * +intel_bufferobj_buffer(struct intel_context *intel, + struct intel_buffer_object *intel_obj, GLuint flag) +{ + if (intel_obj->region) { + if (flag == INTEL_WRITE_PART) + intel_bufferobj_cow(intel, intel_obj); + else if (flag == INTEL_WRITE_FULL) + intel_bufferobj_release_region(intel, intel_obj); + } + + return intel_obj->buffer; +} + +void +intel_bufferobj_init(struct intel_context *intel) +{ + GLcontext *ctx = &intel->ctx; + + ctx->Driver.NewBufferObject = intel_bufferobj_alloc; + ctx->Driver.DeleteBuffer = intel_bufferobj_free; + ctx->Driver.BufferData = intel_bufferobj_data; + ctx->Driver.BufferSubData = intel_bufferobj_subdata; + ctx->Driver.GetBufferSubData = intel_bufferobj_get_subdata; + ctx->Driver.MapBuffer = intel_bufferobj_map; + ctx->Driver.UnmapBuffer = intel_bufferobj_unmap; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h new file mode 100644 index 0000000000..afe9b2f7cf --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h @@ -0,0 +1,86 @@ + /************************************************************************** + * + * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BUFFEROBJ_H +#define INTEL_BUFFEROBJ_H + +#include "mtypes.h" + +struct intel_context; +struct intel_region; +struct gl_buffer_object; + + +/** + * Intel vertex/pixel buffer object, derived from Mesa's gl_buffer_object. + */ +struct intel_buffer_object +{ + struct gl_buffer_object Base; + struct _DriBufferObject *buffer; /* the low-level buffer manager's buffer handle */ + + struct intel_region *region; /* Is there a zero-copy texture + associated with this (pixel) + buffer object? */ +}; + + +/* Get the bm buffer associated with a GL bufferobject: + */ +struct _DriBufferObject *intel_bufferobj_buffer(struct intel_context *intel, + struct intel_buffer_object + *obj, GLuint flag); + +/* Hook the bufferobject implementation into mesa: + */ +void intel_bufferobj_init(struct intel_context *intel); + + + +/* Are the obj->Name tests necessary? Unfortunately yes, mesa + * allocates a couple of gl_buffer_object structs statically, and + * the Name == 0 test is the only way to identify them and avoid + * casting them erroneously to our structs. + */ +static INLINE struct intel_buffer_object * +intel_buffer_object(struct gl_buffer_object *obj) +{ + if (obj->Name) + return (struct intel_buffer_object *) obj; + else + return NULL; +} + +/* Helpers for zerocopy image uploads. See also intel_regions.h: + */ +void intel_bufferobj_cow(struct intel_context *intel, + struct intel_buffer_object *intel_obj); +void intel_bufferobj_release_region(struct intel_context *intel, + struct intel_buffer_object *intel_obj); + + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.c b/src/mesa/drivers/dri/i915pipe/intel_buffers.c new file mode 100644 index 0000000000..7684e36e3b --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.c @@ -0,0 +1,792 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_blit.h" +#include "intel_buffers.h" +#include "intel_depthstencil.h" +#include "intel_fbo.h" +#include "intel_tris.h" +#include "intel_regions.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" +#include "context.h" +#include "utils.h" +#include "drirenderbuffer.h" +#include "framebuffer.h" +#include "swrast/swrast.h" +#include "vblank.h" + + +/* This block can be removed when libdrm >= 2.3.1 is required */ + +#ifndef DRM_VBLANK_FLIP + +#define DRM_VBLANK_FLIP 0x8000000 + +typedef struct drm_i915_flip { + int pipes; +} drm_i915_flip_t; + +#undef DRM_IOCTL_I915_FLIP +#define DRM_IOCTL_I915_FLIP DRM_IOW(DRM_COMMAND_BASE + DRM_I915_FLIP, \ + drm_i915_flip_t) + +#endif + + +/** + * XXX move this into a new dri/common/cliprects.c file. + */ +GLboolean +intel_intersect_cliprects(drm_clip_rect_t * dst, + const drm_clip_rect_t * a, + const drm_clip_rect_t * b) +{ + GLint bx = b->x1; + GLint by = b->y1; + GLint bw = b->x2 - bx; + GLint bh = b->y2 - by; + + if (bx < a->x1) + bw -= a->x1 - bx, bx = a->x1; + if (by < a->y1) + bh -= a->y1 - by, by = a->y1; + if (bx + bw > a->x2) + bw = a->x2 - bx; + if (by + bh > a->y2) + bh = a->y2 - by; + if (bw <= 0) + return GL_FALSE; + if (bh <= 0) + return GL_FALSE; + + dst->x1 = bx; + dst->y1 = by; + dst->x2 = bx + bw; + dst->y2 = by + bh; + + return GL_TRUE; +} + +/** + * Return pointer to current color drawing region, or NULL. + */ +struct intel_region * +intel_drawbuf_region(struct intel_context *intel) +{ + struct intel_renderbuffer *irbColor = + intel_renderbuffer(intel->ctx.DrawBuffer->_ColorDrawBuffers[0][0]); + if (irbColor) + return irbColor->region; + else + return NULL; +} + +/** + * Return pointer to current color reading region, or NULL. + */ +struct intel_region * +intel_readbuf_region(struct intel_context *intel) +{ + struct intel_renderbuffer *irb + = intel_renderbuffer(intel->ctx.ReadBuffer->_ColorReadBuffer); + if (irb) + return irb->region; + else + return NULL; +} + + + +/** + * Update the following fields for rendering: + * intel->numClipRects + * intel->pClipRects + */ +static void +intelSetRenderbufferClipRects(struct intel_context *intel) +{ + /* zero-sized buffers might be legal? */ + assert(intel->ctx.DrawBuffer->Width > 0); + assert(intel->ctx.DrawBuffer->Height > 0); + intel->fboRect.x1 = 0; + intel->fboRect.y1 = 0; + intel->fboRect.x2 = intel->ctx.DrawBuffer->Width; + intel->fboRect.y2 = intel->ctx.DrawBuffer->Height; + intel->numClipRects = 1; + intel->pClipRects = &intel->fboRect; +} + + +/** + * This will be called whenever the currently bound window is moved/resized. + * XXX: actually, it seems to NOT be called when the window is only moved (BP). + */ +void +intelWindowMoved(struct intel_context *intel) +{ + GLcontext *ctx = &intel->ctx; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + + if (!intel->ctx.DrawBuffer) { + /* when would this happen? -BP */ + assert(0); + intel->numClipRects = 0; + } + + /* Update Mesa's notion of window size */ + driUpdateFramebufferSize(ctx, dPriv); + intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ + + { + drmI830Sarea *sarea = intel->sarea; + drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, + .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; + drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, .y1 = sarea->pipeA_y, + .x2 = sarea->pipeA_x + sarea->pipeA_w, + .y2 = sarea->pipeA_y + sarea->pipeA_h }; + drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, .y1 = sarea->pipeB_y, + .x2 = sarea->pipeB_x + sarea->pipeB_w, + .y2 = sarea->pipeB_y + sarea->pipeB_h }; + GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); + GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); + GLuint flags = intel_fb->vblank_flags; + GLboolean pf_active; + GLint pf_pipes; + + /* Update page flipping info + */ + pf_pipes = 0; + + if (areaA > 0) + pf_pipes |= 1; + + if (areaB > 0) + pf_pipes |= 2; + + intel_fb->pf_current_page = (intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3; + + intel_fb->pf_num_pages = 2 /*intel->intelScreen->third.handle ? 3 : 2*/; + + pf_active = pf_pipes && (pf_pipes & intel->sarea->pf_active) == pf_pipes; + + if (INTEL_DEBUG & DEBUG_LOCK) + if (pf_active != intel_fb->pf_active) + _mesa_printf("%s - Page flipping %sactive\n", __progname, + pf_active ? "" : "in"); + + if (pf_active) { + /* Sync pages between pipes if we're flipping on both at the same time */ + if (pf_pipes == 0x3 && pf_pipes != intel_fb->pf_pipes && + (intel->sarea->pf_current_page & 0x3) != + (((intel->sarea->pf_current_page) >> 2) & 0x3)) { + drm_i915_flip_t flip; + + if (intel_fb->pf_current_page == + (intel->sarea->pf_current_page & 0x3)) { + /* XXX: This is ugly, but emitting two flips 'in a row' can cause + * lockups for unknown reasons. + */ + intel->sarea->pf_current_page = + intel->sarea->pf_current_page & 0x3; + intel->sarea->pf_current_page |= + ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % + intel_fb->pf_num_pages) << 2; + + flip.pipes = 0x2; + } else { + intel->sarea->pf_current_page = + intel->sarea->pf_current_page & (0x3 << 2); + intel->sarea->pf_current_page |= + (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % + intel_fb->pf_num_pages; + + flip.pipes = 0x1; + } + + drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); + } + + intel_fb->pf_pipes = pf_pipes; + } + + intel_fb->pf_active = pf_active; + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); + + /* Update vblank info + */ + if (areaB > areaA || (areaA == areaB && areaB > 0)) { + flags = intel_fb->vblank_flags | VBLANK_FLAG_SECONDARY; + } else { + flags = intel_fb->vblank_flags & ~VBLANK_FLAG_SECONDARY; + } + + if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && + !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { + drmVBlank vbl; + int i; + + vbl.request.type = DRM_VBLANK_ABSOLUTE; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + vbl.request.type |= DRM_VBLANK_SECONDARY; + } + + for (i = 0; i < intel_fb->pf_num_pages; i++) { + if (!intel_fb->color_rb[i] || + (intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <= + (1<<23)) + continue; + + vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending; + drmWaitVBlank(intel->driFd, &vbl); + } + + intel_fb->vblank_flags = flags; + driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); + intel_fb->vbl_waited = intel_fb->vbl_seq; + + for (i = 0; i < intel_fb->pf_num_pages; i++) { + if (intel_fb->color_rb[i]) + intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited; + } + } + } + + /* This will be picked up by looking at the dirty state flags: + */ + + /* Update hardware scissor */ +// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, +// ctx->Scissor.Width, ctx->Scissor.Height); + + /* Re-calculate viewport related state */ +// ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far ); +} + + + + + +/** + * Called by ctx->Driver.Clear. + */ +static void +intelClear(GLcontext *ctx, GLbitfield mask) +{ + const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask); + GLbitfield tri_mask = 0; + GLbitfield blit_mask = 0; + GLbitfield swrast_mask = 0; + struct gl_framebuffer *fb = ctx->DrawBuffer; + GLuint i; + + if (0) + fprintf(stderr, "%s\n", __FUNCTION__); + + /* HW color buffers (front, back, aux, generic FBO, etc) */ + if (colorMask == ~0) { + /* clear all R,G,B,A */ + /* XXX FBO: need to check if colorbuffers are software RBOs! */ + blit_mask |= (mask & BUFFER_BITS_COLOR); + } + else { + /* glColorMask in effect */ + tri_mask |= (mask & BUFFER_BITS_COLOR); + } + + /* HW stencil */ + if (mask & BUFFER_BIT_STENCIL) { + const struct intel_region *stencilRegion + = intel_get_rb_region(fb, BUFFER_STENCIL); + if (stencilRegion) { + /* have hw stencil */ + if ((ctx->Stencil.WriteMask[0] & 0xff) != 0xff) { + /* not clearing all stencil bits, so use triangle clearing */ + tri_mask |= BUFFER_BIT_STENCIL; + } + else { + /* clearing all stencil bits, use blitting */ + blit_mask |= BUFFER_BIT_STENCIL; + } + } + } + + /* HW depth */ + if (mask & BUFFER_BIT_DEPTH) { + /* clear depth with whatever method is used for stencil (see above) */ + if (tri_mask & BUFFER_BIT_STENCIL) + tri_mask |= BUFFER_BIT_DEPTH; + else + blit_mask |= BUFFER_BIT_DEPTH; + } + + /* SW fallback clearing */ + swrast_mask = mask & ~tri_mask & ~blit_mask; + + for (i = 0; i < BUFFER_COUNT; i++) { + GLuint bufBit = 1 << i; + if ((blit_mask | tri_mask) & bufBit) { + if (!fb->Attachment[i].Renderbuffer->ClassID) { + blit_mask &= ~bufBit; + tri_mask &= ~bufBit; + swrast_mask |= bufBit; + } + } + } + + + intelFlush(ctx); /* XXX intelClearWithBlit also does this */ + + if (blit_mask) + intelClearWithBlit(ctx, blit_mask); + + if (swrast_mask | tri_mask) + _swrast_Clear(ctx, swrast_mask | tri_mask); +} + + +/* Emit wait for pending flips */ +void +intel_wait_flips(struct intel_context *intel, GLuint batch_flags) +{ + struct intel_framebuffer *intel_fb = + (struct intel_framebuffer *) intel->ctx.DrawBuffer; + struct intel_renderbuffer *intel_rb = + intel_get_renderbuffer(&intel_fb->Base, + intel_fb->Base._ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : + BUFFER_BACK_LEFT); + + if (intel_fb->Base.Name == 0 && intel_rb->pf_pending == intel_fb->pf_seq) { + GLint pf_pipes = intel_fb->pf_pipes; + BATCH_LOCALS; + + /* Wait for pending flips to take effect */ + BEGIN_BATCH(2, batch_flags); + OUT_BATCH(pf_pipes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) + : 0); + OUT_BATCH(pf_pipes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) + : 0); + ADVANCE_BATCH(); + + intel_rb->pf_pending--; + } +} + +#if 0 +/* Flip the front & back buffers + */ +static GLboolean +intelPageFlip(const __DRIdrawablePrivate * dPriv) +{ + struct intel_context *intel; + int ret; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + + if (INTEL_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + + assert(dPriv); + assert(dPriv->driContextPriv); + assert(dPriv->driContextPriv->driverPrivate); + + intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate; + + if (intel->intelScreen->drmMinor < 9) + return GL_FALSE; + + intelFlush(&intel->ctx); + + ret = 0; + + LOCK_HARDWARE(intel); + + if (dPriv->numClipRects && intel_fb->pf_active) { + drm_i915_flip_t flip; + + flip.pipes = intel_fb->pf_pipes; + + ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); + } + + UNLOCK_HARDWARE(intel); + + if (ret || !intel_fb->pf_active) + return GL_FALSE; + + if (!dPriv->numClipRects) { + usleep(10000); /* throttle invisible client 10ms */ + } + + intel_fb->pf_current_page = (intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3; + + if (dPriv->numClipRects != 0) { + intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending = + intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->pf_pending = + ++intel_fb->pf_seq; + } + + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, &intel_fb->Base); + + if (INTEL_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s: success\n", __FUNCTION__); + + return GL_TRUE; +} +#endif + + +static GLboolean +intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) +{ + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); + struct intel_context *intel = + intelScreenContext(dPriv->driScreenPriv->private); + const intelScreenPrivate *intelScreen = intel->intelScreen; + unsigned int target; + drm_i915_vblank_swap_t swap; + GLboolean ret; + + if (!intel_fb->vblank_flags || + (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || + intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) + return GL_FALSE; + + swap.seqtype = DRM_VBLANK_ABSOLUTE; + + if (intel_fb->vblank_flags & VBLANK_FLAG_SYNC) { + swap.seqtype |= DRM_VBLANK_NEXTONMISS; + } else if (interval == 0) { + return GL_FALSE; + } + + swap.drawable = dPriv->hHWDrawable; + target = swap.sequence = intel_fb->vbl_seq + interval; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + swap.seqtype |= DRM_VBLANK_SECONDARY; + } + + LOCK_HARDWARE(intel); + + intel_batchbuffer_flush(intel->batch); + + if ( intel_fb->pf_active ) { + swap.seqtype |= DRM_VBLANK_FLIP; + + intel_fb->pf_current_page = (((intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3) + 1) % + intel_fb->pf_num_pages; + } + + if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, + sizeof(swap))) { + intel_fb->vbl_seq = swap.sequence; + swap.sequence -= target; + *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); + + intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending = + intel_get_renderbuffer(&intel_fb->Base, + BUFFER_FRONT_LEFT)->vbl_pending = + intel_fb->vbl_seq; + + if (swap.seqtype & DRM_VBLANK_FLIP) { + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); + } + + ret = GL_TRUE; + } else { + if (swap.seqtype & DRM_VBLANK_FLIP) { + intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3) % + intel_fb->pf_num_pages; + } + + ret = GL_FALSE; + } + + UNLOCK_HARDWARE(intel); + + return ret; +} + +void +intelSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel; + + if (ctx == NULL) + return; + + intel = intel_context(ctx); + + if (ctx->Visual.doubleBufferMode) { + GLboolean missed_target; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + int64_t ust; + + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + + if (!intelScheduleSwap(dPriv, &missed_target)) { + driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, + &missed_target); + + intelCopyBuffer(dPriv, NULL); + } + + intel_fb->swap_count++; + (*dri_interface->getUST) (&ust); + if (missed_target) { + intel_fb->swap_missed_count++; + intel_fb->swap_missed_ust = ust - intel_fb->swap_ust; + } + + intel_fb->swap_ust = ust; + } + } + else { + /* XXX this shouldn't be an error but we can't handle it for now */ + fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + } +} + +void +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + struct intel_context *intel = + (struct intel_context *) dPriv->driContextPriv->driverPrivate; + GLcontext *ctx = &intel->ctx; + + if (ctx->Visual.doubleBufferMode) { + drm_clip_rect_t rect; + /* fixup cliprect (driDrawable may have changed?) later */ + rect.x1 = x; + rect.y1 = y; + rect.x2 = w; + rect.y2 = h; + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + intelCopyBuffer(dPriv, &rect); + } + } + else { + /* XXX this shouldn't be an error but we can't handle it for now */ + fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + } +} + + +/** + * Update the hardware state for drawing into a window or framebuffer object. + * + * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other + * places within the driver. + * + * Basically, this needs to be called any time the current framebuffer + * changes, the renderbuffers change, or we need to draw into different + * color buffers. + */ +void +intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_region *colorRegion, *depthRegion = NULL; + struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; + + if (!fb) { + /* this can happen during the initial context initialization */ + return; + } + + /* Do this here, not core Mesa, since this function is called from + * many places within the driver. + */ + if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { + /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */ + _mesa_update_framebuffer(ctx); + /* this updates the DrawBuffer's Width/Height if it's a FBO */ + _mesa_update_draw_buffer_bounds(ctx); + } + + if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { + /* this may occur when we're called by glBindFrameBuffer() during + * the process of someone setting up renderbuffers, etc. + */ + /*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/ + return; + } + + if (fb->Name) + intel_validate_paired_depth_stencil(ctx, fb); + + /* + * How many color buffers are we drawing into? + */ + if (fb->_NumColorDrawBuffers[0] != 1) { + /* writing to 0 or 2 or 4 color buffers */ + /*_mesa_debug(ctx, "Software rendering\n");*/ + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); + } + else { + /* draw to exactly one color buffer */ + /*_mesa_debug(ctx, "Hardware rendering\n");*/ + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); + } + + /* + * Get the intel_renderbuffer for the colorbuffer we're drawing into. + * And set up cliprects. + */ + { + struct intel_renderbuffer *irb; + intelSetRenderbufferClipRects(intel); + irb = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); + colorRegion = (irb && irb->region) ? irb->region : NULL; + } + + /* Update culling direction which changes depending on the + * orientation of the buffer: + */ + if (ctx->Driver.FrontFace) + ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace); + else + ctx->NewState |= _NEW_POLYGON; + + if (!colorRegion) { + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); + } + else { + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); + } + + /*** + *** Get depth buffer region and check if we need a software fallback. + *** Note that the depth buffer is usually a DEPTH_STENCIL buffer. + ***/ + if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) { + irbDepth = intel_renderbuffer(fb->_DepthBuffer->Wrapped); + if (irbDepth && irbDepth->region) { + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); + depthRegion = irbDepth->region; + } + else { + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_TRUE); + depthRegion = NULL; + } + } + else { + /* not using depth buffer */ + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); + depthRegion = NULL; + } + + /*** + *** Stencil buffer + *** This can only be hardware accelerated if we're using a + *** combined DEPTH_STENCIL buffer (for now anyway). + ***/ + if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { + irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); + if (irbStencil && irbStencil->region) { + ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); + /* need to re-compute stencil hw state */ +// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); + if (!depthRegion) + depthRegion = irbStencil->region; + } + else { + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE); + } + } + else { + /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */ + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); + /* need to re-compute stencil hw state */ +// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); + } + + + /** + ** Release old regions, reference new regions + **/ + + // intel->vtbl.set_draw_region(intel, colorRegion, depthRegion); + + /* update viewport since it depends on window size */ +// ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y, +// ctx->Viewport.Width, ctx->Viewport.Height); + + /* Update hardware scissor */ +// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, +// ctx->Scissor.Width, ctx->Scissor.Height); +} + + +static void +intelDrawBuffer(GLcontext * ctx, GLenum mode) +{ + intel_draw_buffer(ctx, ctx->DrawBuffer); +} + + +static void +intelReadBuffer(GLcontext * ctx, GLenum mode) +{ + if (ctx->ReadBuffer == ctx->DrawBuffer) { + /* This will update FBO completeness status. + * A framebuffer will be incomplete if the GL_READ_BUFFER setting + * refers to a missing renderbuffer. Calling glReadBuffer can set + * that straight and can make the drawing buffer complete. + */ + intel_draw_buffer(ctx, ctx->DrawBuffer); + } + /* Generally, functions which read pixels (glReadPixels, glCopyPixels, etc) + * reference ctx->ReadBuffer and do appropriate state checks. + */ +} + + +void +intelInitBufferFuncs(struct dd_function_table *functions) +{ + functions->Clear = intelClear; + functions->DrawBuffer = intelDrawBuffer; + functions->ReadBuffer = intelReadBuffer; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.h b/src/mesa/drivers/dri/i915pipe/intel_buffers.h new file mode 100644 index 0000000000..13d1a15ffb --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.h @@ -0,0 +1,55 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BUFFERS_H +#define INTEL_BUFFERS_H + + +struct intel_context; +struct intel_framebuffer; + + +extern GLboolean +intel_intersect_cliprects(drm_clip_rect_t * dest, + const drm_clip_rect_t * a, + const drm_clip_rect_t * b); + +extern struct intel_region *intel_readbuf_region(struct intel_context *intel); + +extern struct intel_region *intel_drawbuf_region(struct intel_context *intel); + +extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); + +extern void intelWindowMoved(struct intel_context *intel); + +extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); + +extern void intelInitBufferFuncs(struct dd_function_table *functions); + +#endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c new file mode 100644 index 0000000000..e8e00b9606 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -0,0 +1,768 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "glheader.h" +#include "context.h" +#include "matrix.h" +#include "simple_list.h" +#include "extensions.h" +#include "framebuffer.h" +#include "imports.h" +#include "points.h" + +#include "swrast/swrast.h" +#include "swrast_setup/swrast_setup.h" +#include "tnl/tnl.h" + +#include "tnl/t_pipeline.h" +#include "tnl/t_vertex.h" + +#include "drivers/common/driverfuncs.h" + +#include "intel_screen.h" + +#include "i830_dri.h" + +#include "intel_buffers.h" +#include "intel_tex.h" +#include "intel_span.h" +#include "intel_tris.h" +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_pixel.h" +#include "intel_regions.h" +#include "intel_buffer_objects.h" +#include "intel_fbo.h" + +#include "pipe/softpipe/sp_context.h" +#include "state_tracker/st_public.h" + + +#include "drirenderbuffer.h" +#include "vblank.h" +#include "utils.h" +#include "xmlpool.h" /* for symbolic values of enum-type options */ +#ifndef INTEL_DEBUG +int INTEL_DEBUG = (0); +#endif + +#define need_GL_ARB_multisample +#define need_GL_ARB_point_parameters +#define need_GL_ARB_texture_compression +#define need_GL_ARB_vertex_buffer_object +#define need_GL_ARB_vertex_program +#define need_GL_ARB_window_pos +#define need_GL_EXT_blend_color +#define need_GL_EXT_blend_equation_separate +#define need_GL_EXT_blend_func_separate +#define need_GL_EXT_blend_minmax +#define need_GL_EXT_cull_vertex +#define need_GL_EXT_fog_coord +#define need_GL_EXT_framebuffer_object +#define need_GL_EXT_multi_draw_arrays +#define need_GL_EXT_secondary_color +#define need_GL_NV_vertex_program +#include "extension_helper.h" + + +#define DRIVER_DATE "20061102" + +_glthread_Mutex lockMutex; +static GLboolean lockMutexInit = GL_FALSE; + + +static const GLubyte * +intelGetString(GLcontext * ctx, GLenum name) +{ + const char *chipset; + static char buffer[128]; + + switch (name) { + case GL_VENDOR: + return (GLubyte *) "Tungsten Graphics, Inc"; + break; + + case GL_RENDERER: + switch (intel_context(ctx)->intelScreen->deviceID) { + case PCI_CHIP_845_G: + chipset = "Intel(R) 845G"; + break; + case PCI_CHIP_I830_M: + chipset = "Intel(R) 830M"; + break; + case PCI_CHIP_I855_GM: + chipset = "Intel(R) 852GM/855GM"; + break; + case PCI_CHIP_I865_G: + chipset = "Intel(R) 865G"; + break; + case PCI_CHIP_I915_G: + chipset = "Intel(R) 915G"; + break; + case PCI_CHIP_I915_GM: + chipset = "Intel(R) 915GM"; + break; + case PCI_CHIP_I945_G: + chipset = "Intel(R) 945G"; + break; + case PCI_CHIP_I945_GM: + chipset = "Intel(R) 945GM"; + break; + case PCI_CHIP_I945_GME: + chipset = "Intel(R) 945GME"; + break; + case PCI_CHIP_G33_G: + chipset = "Intel(R) G33"; + break; + case PCI_CHIP_Q35_G: + chipset = "Intel(R) Q35"; + break; + case PCI_CHIP_Q33_G: + chipset = "Intel(R) Q33"; + break; + default: + chipset = "Unknown Intel Chipset"; + break; + } + + (void) driGetRendererString(buffer, chipset, DRIVER_DATE, 0); + return (GLubyte *) buffer; + + default: + return NULL; + } +} + + +/** + * Extension strings exported by the intel driver. + * + * \note + * It appears that ARB_texture_env_crossbar has "disappeared" compared to the + * old i830-specific driver. + */ +const struct dri_extension card_extensions[] = { + {"GL_ARB_multisample", GL_ARB_multisample_functions}, + {"GL_ARB_multitexture", NULL}, + {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_texture_border_clamp", NULL}, + {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, + {"GL_ARB_texture_cube_map", NULL}, + {"GL_ARB_texture_env_add", NULL}, + {"GL_ARB_texture_env_combine", NULL}, + {"GL_ARB_texture_env_dot3", NULL}, + {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_rectangle", NULL}, + {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, + {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, + {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, + {"GL_EXT_blend_equation_separate", + GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, + {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_subtract", NULL}, + {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, + {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, +#if 1 /* XXX FBO temporary? */ + {"GL_EXT_packed_depth_stencil", NULL}, +#endif + {"GL_EXT_pixel_buffer_object", NULL}, + {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, + {"GL_EXT_stencil_wrap", NULL}, + {"GL_EXT_texture_edge_clamp", NULL}, + {"GL_EXT_texture_env_combine", NULL}, + {"GL_EXT_texture_env_dot3", NULL}, + {"GL_EXT_texture_filter_anisotropic", NULL}, + {"GL_EXT_texture_lod_bias", NULL}, + {"GL_3DFX_texture_compression_FXT1", NULL}, + {"GL_APPLE_client_storage", NULL}, + {"GL_MESA_pack_invert", NULL}, + {"GL_MESA_ycbcr_texture", NULL}, + {"GL_NV_blend_square", NULL}, + {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, + {"GL_NV_vertex_program1_1", NULL}, +/* { "GL_SGIS_generate_mipmap", NULL }, */ + {NULL, NULL} +}; + + + + +static const struct dri_debug_control debug_control[] = { + {"tex", DEBUG_TEXTURE}, + {"state", DEBUG_STATE}, + {"ioctl", DEBUG_IOCTL}, + {"blit", DEBUG_BLIT}, + {"mip", DEBUG_MIPTREE}, + {"fall", DEBUG_FALLBACKS}, + {"verb", DEBUG_VERBOSE}, + {"bat", DEBUG_BATCH}, + {"pix", DEBUG_PIXEL}, + {"buf", DEBUG_BUFMGR}, + {"reg", DEBUG_REGION}, + {"fbo", DEBUG_FBO}, + {"lock", DEBUG_LOCK}, + {NULL, 0} +}; + + +static void +intelInvalidateState(GLcontext * ctx, GLuint new_state) +{ + _swrast_InvalidateState(ctx, new_state); + _swsetup_InvalidateState(ctx, new_state); + _vbo_InvalidateState(ctx, new_state); + _tnl_InvalidateState(ctx, new_state); + _tnl_invalidate_vertex_state(ctx, new_state); + + st_invalidate_state( ctx, new_state ); + + intel_context(ctx)->NewGLState |= new_state; +} + + +void +intelFlush(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + + if (intel->Fallback) + _swrast_flush(ctx); + + INTEL_FIREVERTICES(intel); + + if (intel->batch->map != intel->batch->ptr) + intel_batchbuffer_flush(intel->batch); + + /* XXX: Need to do an MI_FLUSH here. + */ +} + + +/** + * Check if we need to rotate/warp the front color buffer to the + * rotated screen. We generally need to do this when we get a glFlush + * or glFinish after drawing to the front color buffer. + * If no rotation, just copy the private fake front buffer to the real one. + */ +static void +intelCheckFrontUpdate(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + /* rely on _ColorDrawBufferMask being kept up to date by mesa + even for window-fbos. */ + /* not sure. Might need that for all masks including + BUFFER_BIT_FRONT_LEFT maybe? */ + if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT) { + __DRIdrawablePrivate *dPriv = intel->driDrawable; + intelCopyBuffer(dPriv, NULL); + } +} + + +/** + * Called via glFlush. + */ +static void +intelglFlush(GLcontext * ctx) +{ + intelFlush(ctx); + intelCheckFrontUpdate(ctx); +} + +void +intelFinish(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + intelFlush(ctx); + if (intel->batch->last_fence) { + driFenceFinish(intel->batch->last_fence, + 0, GL_FALSE); + driFenceUnReference(intel->batch->last_fence); + intel->batch->last_fence = NULL; + } + intelCheckFrontUpdate(ctx); +} + + +void +intelInitDriverFunctions(struct dd_function_table *functions) +{ + _mesa_init_driver_functions(functions); + + functions->Flush = intelglFlush; + functions->Finish = intelFinish; + functions->GetString = intelGetString; + functions->UpdateState = intelInvalidateState; + functions->CopyColorTable = _swrast_CopyColorTable; + functions->CopyColorSubTable = _swrast_CopyColorSubTable; + functions->CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D; + functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D; + + intelInitTextureFuncs(functions); + intelInitPixelFuncs(functions); + intelInitStateFuncs(functions); + intelInitBufferFuncs(functions); +} + + + +GLboolean +intelCreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) +{ + struct dd_function_table functions; + + struct intel_context *intel = CALLOC_STRUCT(intel_context); + GLcontext *ctx = &intel->ctx; + + GLcontext *shareCtx = (GLcontext *) sharedContextPrivate; + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + drmI830Sarea *saPriv = (drmI830Sarea *) + (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset); + int fthrottle_mode; + GLboolean havePools; + + DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + havePools = intelCreatePools(intelScreen); + DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + + if (!havePools) + return GL_FALSE; + + intelInitDriverFunctions(&functions); + + if (!_mesa_initialize_context(&intel->ctx, + mesaVis, shareCtx, + &functions, (void *) intel)) + return GL_FALSE; + + driContextPriv->driverPrivate = intel; + intel->intelScreen = intelScreen; + intel->driScreen = sPriv; + intel->sarea = saPriv; + + if (!lockMutexInit) { + lockMutexInit = GL_TRUE; + _glthread_INIT_MUTEX(lockMutex); + } + + driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, + intel->driScreen->myNum, "i915"); + + ctx->Const.MaxTextureMaxAnisotropy = 2.0; + + /* This doesn't yet catch all non-conformant rendering, but it's a + * start. + */ + if (getenv("INTEL_STRICT_CONFORMANCE")) { + intel->strict_conformance = 1; + } + + ctx->Const.MinLineWidth = 1.0; + ctx->Const.MinLineWidthAA = 1.0; + ctx->Const.MaxLineWidth = 3.0; + ctx->Const.MaxLineWidthAA = 3.0; + ctx->Const.LineWidthGranularity = 1.0; + + ctx->Const.MinPointSize = 1.0; + ctx->Const.MinPointSizeAA = 1.0; + ctx->Const.MaxPointSize = 255.0; + ctx->Const.MaxPointSizeAA = 3.0; + ctx->Const.PointSizeGranularity = 1.0; + + /* reinitialize the context point state. + * It depend on constants in __GLcontextRec::Const + */ + _mesa_init_point(ctx); + + ctx->Const.MaxColorAttachments = 4; /* XXX FBO: review this */ + + /* Initialize the software rasterizer and helper modules. */ + _swrast_CreateContext(ctx); + _vbo_CreateContext(ctx); + _tnl_CreateContext(ctx); + _swsetup_CreateContext(ctx); + + + /* Configure swrast to match hardware characteristics: */ + _swrast_allow_pixel_fog(ctx, GL_FALSE); + _swrast_allow_vertex_fog(ctx, GL_TRUE); + + /* Dri stuff */ + intel->hHWContext = driContextPriv->hHWContext; + intel->driFd = sPriv->fd; + intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; + + intel->hw_stipple = 1; + + /* XXX FBO: this doesn't seem to be used anywhere */ + switch (mesaVis->depthBits) { + case 0: /* what to do in this case? */ + case 16: + intel->polygon_offset_scale = 1.0 / 0xffff; + break; + case 24: + intel->polygon_offset_scale = 2.0 / 0xffffff; /* req'd to pass glean */ + break; + default: + assert(0); + break; + } + + /* Initialize swrast, tnl driver tables: */ + intelInitSpanFuncs(ctx); + + TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; + + + intel->RenderIndex = ~0; + + fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); + intel->iw.irq_seq = -1; + intel->irqsEmitted = 0; + + intel->do_irqs = (intel->intelScreen->irq_active && + fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS); + + intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); + + _math_matrix_ctr(&intel->ViewportMatrix); + + /* Disable imaging extension until convolution is working in + * teximage paths: + */ + driInitExtensions(ctx, card_extensions, +/* GL_TRUE, */ + GL_FALSE); + + + intel->batch = intel_batchbuffer_alloc(intel); + intel->last_swap_fence = NULL; + intel->first_swap_fence = NULL; + + intel_bufferobj_init(intel); + intel_fbo_init(intel); + + if (intel->ctx.Mesa_DXTn) { + _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); + _mesa_enable_extension(ctx, "GL_S3_s3tc"); + } + else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) { + _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); + } + + intel->prim.primitive = ~0; + + +#if DO_DEBUG + INTEL_DEBUG = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); +#endif + + if (getenv("INTEL_NO_RAST")) { + fprintf(stderr, "disabling 3D rasterization\n"); + FALLBACK(intel, INTEL_FALLBACK_USER, 1); + } + + + st_create_context( &intel->ctx, + softpipe_create() ); + + + return GL_TRUE; +} + +void +intelDestroyContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = + (struct intel_context *) driContextPriv->driverPrivate; + + assert(intel); /* should never be null */ + if (intel) { + GLboolean release_texture_heaps; + + INTEL_FIREVERTICES(intel); + + //intel->vtbl.destroy(intel); + + release_texture_heaps = (intel->ctx.Shared->RefCount == 1); + _swsetup_DestroyContext(&intel->ctx); + _tnl_DestroyContext(&intel->ctx); + _vbo_DestroyContext(&intel->ctx); + + _swrast_DestroyContext(&intel->ctx); + intel->Fallback = 0; /* don't call _swrast_Flush later */ + + intel_batchbuffer_free(intel->batch); + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + if (intel->first_swap_fence) { + driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->first_swap_fence); + intel->first_swap_fence = NULL; + } + + + if (release_texture_heaps) { + /* This share group is about to go away, free our private + * texture object data. + */ + if (INTEL_DEBUG & DEBUG_TEXTURE) + fprintf(stderr, "do something to free texture heaps\n"); + } + + /* free the Mesa context */ + _mesa_free_context_data(&intel->ctx); + } +} + +GLboolean +intelUnbindContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; + /* XXX UnbindContext is called AFTER the new context is made current. + Hopefully shouldn't be a problem ? */ + FLUSH_VERTICES((&intel->ctx), 0); + intelFlush(&intel->ctx); + return GL_TRUE; +} + +GLboolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) +{ + +#if 0 + if (driDrawPriv) { + fprintf(stderr, "x %d, y %d, width %d, height %d\n", + driDrawPriv->x, driDrawPriv->y, driDrawPriv->w, driDrawPriv->h); + } +#endif + + if (driContextPriv) { + struct intel_context *intel = + (struct intel_context *) driContextPriv->driverPrivate; + struct intel_framebuffer *intel_fb = + (struct intel_framebuffer *) driDrawPriv->driverPrivate; + GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; + + /* this is a hack so we have a valid context when the region allocation + is done. Need a per-screen context? */ + intel->intelScreen->dummyctxptr = intel; + + /* update GLframebuffer size to match window if needed */ + driUpdateFramebufferSize(&intel->ctx, driDrawPriv); + + if (driReadPriv != driDrawPriv) { + driUpdateFramebufferSize(&intel->ctx, driReadPriv); + } + + _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); + + /* The drawbuffer won't always be updated by _mesa_make_current: + */ + if (intel->ctx.DrawBuffer == &intel_fb->Base) { + + if (intel->driDrawable != driDrawPriv) { + if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) { + int i; + + intel_fb->vblank_flags = (intel->intelScreen->irq_active != 0) + ? driGetDefaultVBlankFlags(&intel->optionCache) + : VBLANK_FLAG_NO_IRQ; + + (*dri_interface->getUST) (&intel_fb->swap_ust); + driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags, + &intel_fb->vbl_seq); + intel_fb->vbl_waited = intel_fb->vbl_seq; + + for (i = 0; i < 2; i++) { + if (intel_fb->color_rb[i]) + intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_seq; + } + } + } + } + + if ((intel->driDrawable != driDrawPriv) || + (intel->lastStamp != driDrawPriv->lastStamp)) { + intel->driDrawable = driDrawPriv; + intelWindowMoved(intel); + intel->lastStamp = driDrawPriv->lastStamp; + } + + } + else { + _mesa_make_current(NULL, NULL, NULL); + } + + return GL_TRUE; +} + +static void +intelContendedLock(struct intel_context *intel, GLuint flags) +{ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + __DRIscreenPrivate *sPriv = intel->driScreen; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + drmI830Sarea *sarea = intel->sarea; + + drmGetLock(intel->driFd, intel->hHWContext, flags); + + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("%s - got contended lock\n", __progname); + + /* If the window moved, may need to set a new cliprect now. + * + * NOTE: This releases and regains the hw lock, so all state + * checking must be done *after* this call: + */ + if (dPriv) + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + + if (sarea->width != intelScreen->width || + sarea->height != intelScreen->height) { + + intelUpdateScreenRotation(sPriv, sarea); + } + +#if 0 + if (sarea->width != intel->width || + sarea->height != intel->height || + sarea->rotation != intel->current_rotation) { + int numClipRects = intel->numClipRects; + + /* + * FIXME: Really only need to do this when drawing to a + * common back- or front buffer. + */ + + /* + * This will essentially drop the outstanding batchbuffer on the floor. + */ + intel->numClipRects = 0; + + if (intel->Fallback) + _swrast_flush(&intel->ctx); + + INTEL_FIREVERTICES(intel); + + if (intel->batch->map != intel->batch->ptr) + intel_batchbuffer_flush(intel->batch); + + intel->numClipRects = numClipRects; + + /* force window update */ + intel->lastStamp = 0; + + intel->width = sarea->width; + intel->height = sarea->height; + intel->current_rotation = sarea->rotation; + } +#endif +} + + +/* Lock the hardware and validate our state. + */ +void LOCK_HARDWARE( struct intel_context *intel ) +{ + char __ret=0; + struct intel_framebuffer *intel_fb = NULL; + struct intel_renderbuffer *intel_rb = NULL; + _glthread_LOCK_MUTEX(lockMutex); + assert(!intel->locked); + + if (intel->driDrawable) { + intel_fb = intel->driDrawable->driverPrivate; + + if (intel_fb) + intel_rb = + intel_get_renderbuffer(&intel_fb->Base, + intel_fb->Base._ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : + BUFFER_BACK_LEFT); + } + + if (intel_rb && intel_fb->vblank_flags && + !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && + (intel_fb->vbl_waited - intel_rb->vbl_pending) > (1<<23)) { + drmVBlank vbl; + + vbl.request.type = DRM_VBLANK_ABSOLUTE; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + vbl.request.type |= DRM_VBLANK_SECONDARY; + } + + vbl.request.sequence = intel_rb->vbl_pending; + drmWaitVBlank(intel->driFd, &vbl); + intel_fb->vbl_waited = vbl.reply.sequence; + } + + DRM_CAS(intel->driHwLock, intel->hHWContext, + (DRM_LOCK_HELD|intel->hHWContext), __ret); + + if (__ret) + intelContendedLock( intel, 0 ); + + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("%s - locked\n", __progname); + + intel->locked = 1; +} + + + /* Unlock the hardware using the global current context + */ +void UNLOCK_HARDWARE( struct intel_context *intel ) +{ + intel->locked = 0; + + DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); + + _glthread_UNLOCK_MUTEX(lockMutex); + + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("%s - unlocked\n", __progname); +} + diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.h b/src/mesa/drivers/dri/i915pipe/intel_context.h new file mode 100644 index 0000000000..154be972b9 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_context.h @@ -0,0 +1,426 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTELCONTEXT_INC +#define INTELCONTEXT_INC + + + +#include "mtypes.h" +#include "drm.h" +#include "mm.h" +#include "texmem.h" + +#include "intel_screen.h" +#include "i915_drm.h" +#include "i830_common.h" +#include "tnl/t_vertex.h" + +#define TAG(x) intel##x +#include "tnl_dd/t_dd_vertex.h" +#undef TAG + +#define DV_PF_555 (1<<8) +#define DV_PF_565 (2<<8) +#define DV_PF_8888 (3<<8) + +struct intel_region; +struct intel_context; +struct _DriBufferObject; + +typedef void (*intel_tri_func) (struct intel_context *, intelVertex *, + intelVertex *, intelVertex *); +typedef void (*intel_line_func) (struct intel_context *, intelVertex *, + intelVertex *); +typedef void (*intel_point_func) (struct intel_context *, intelVertex *); + +#define INTEL_FALLBACK_DRAW_BUFFER 0x1 +#define INTEL_FALLBACK_READ_BUFFER 0x2 +#define INTEL_FALLBACK_DEPTH_BUFFER 0x4 +#define INTEL_FALLBACK_STENCIL_BUFFER 0x8 +#define INTEL_FALLBACK_USER 0x10 +#define INTEL_FALLBACK_RENDERMODE 0x20 + +#define FALLBACK( intel, bit, mode ) _mesa_printf("intelFallback not implemented/removed\n") + + +#define INTEL_WRITE_PART 0x1 +#define INTEL_WRITE_FULL 0x2 +#define INTEL_READ 0x4 + +struct intel_texture_object +{ + struct gl_texture_object base; /* The "parent" object */ + + /* The mipmap tree must include at least these levels once + * validated: + */ + GLuint firstLevel; + GLuint lastLevel; + + /* Offset for firstLevel image: + */ + GLuint textureOffset; + + /* On validation any active images held in main memory or in other + * regions will be copied to this region and the old storage freed. + */ + struct intel_mipmap_tree *mt; + + GLboolean imageOverride; + GLint depthOverride; + GLuint pitchOverride; +}; + + + +struct intel_texture_image +{ + struct gl_texture_image base; + + /* These aren't stored in gl_texture_image + */ + GLuint level; + GLuint face; + + /* If intelImage->mt != NULL, image data is stored here. + * Else if intelImage->base.Data != NULL, image is stored there. + * Else there is no image data. + */ + struct intel_mipmap_tree *mt; +}; + + +#define INTEL_MAX_FIXUP 64 + +struct intel_context +{ + GLcontext ctx; /* the parent class */ + + GLint refcount; + GLuint Fallback; + GLuint NewGLState; + + struct _DriFenceObject *last_swap_fence; + struct _DriFenceObject *first_swap_fence; + + struct intel_batchbuffer *batch; + + struct + { + GLuint id; + GLuint primitive; + GLubyte *start_ptr; + void (*flush) (struct intel_context *); + } prim; + + GLboolean locked; + char *prevLockFile; + int prevLockLine; + + GLuint ClearColor565; + GLuint ClearColor8888; + + /* Offsets of fields within the current vertex: + */ + GLuint coloroffset; + GLuint specoffset; + GLuint wpos_offset; + GLuint wpos_size; + + struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; + GLuint vertex_attr_count; + + GLfloat polygon_offset_scale; /* dependent on depth_scale, bpp */ + + GLboolean hw_stipple; + GLboolean strict_conformance; + + /* AGP memory buffer manager: + */ + struct bufmgr *bm; + + + /* State for intelvb.c and inteltris.c. + */ + GLuint RenderIndex; + GLmatrix ViewportMatrix; + GLenum render_primitive; + GLenum reduced_primitive; + GLuint vertex_size; + GLubyte *verts; /* points to tnl->clipspace.vertex_buf */ + + + /* Fallback rasterization functions + */ + intel_point_func draw_point; + intel_line_func draw_line; + intel_tri_func draw_tri; + + /* These refer to the current drawing buffer: + */ + GLuint numClipRects; /**< cliprects for drawing */ + drm_clip_rect_t *pClipRects; + drm_clip_rect_t fboRect; /**< cliprect for rendering */ + + GLuint do_usleeps; + int do_irqs; + GLuint irqsEmitted; + drm_i915_irq_wait_t iw; + + drm_context_t hHWContext; + drmLock *driHwLock; + int driFd; + + __DRIdrawablePrivate *driDrawable; + __DRIscreenPrivate *driScreen; + intelScreenPrivate *intelScreen; + drmI830Sarea *sarea; + + GLuint lastStamp; + + /** + * Configuration cache + */ + driOptionCache optionCache; +}; + +/* These are functions now: + */ +void LOCK_HARDWARE( struct intel_context *intel ); +void UNLOCK_HARDWARE( struct intel_context *intel ); + +extern char *__progname; + + +#define SUBPIXEL_X 0.125 +#define SUBPIXEL_Y 0.125 + +#define INTEL_FIREVERTICES(intel) \ +do { \ + if ((intel)->prim.flush) \ + (intel)->prim.flush(intel); \ +} while (0) + +/* ================================================================ + * Color packing: + */ + +#define INTEL_PACKCOLOR4444(r,g,b,a) \ + ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) + +#define INTEL_PACKCOLOR1555(r,g,b,a) \ + ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ + ((a) ? 0x8000 : 0)) + +#define INTEL_PACKCOLOR565(r,g,b) \ + ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) + +#define INTEL_PACKCOLOR8888(r,g,b,a) \ + ((a<<24) | (r<<16) | (g<<8) | b) + + + +/* ================================================================ + * From linux kernel i386 header files, copes with odd sizes better + * than COPY_DWORDS would: + * XXX Put this in src/mesa/main/imports.h ??? + */ +#if defined(i386) || defined(__i386__) +static INLINE void * +__memcpy(void *to, const void *from, size_t n) +{ + int d0, d1, d2; + __asm__ __volatile__("rep ; movsl\n\t" + "testb $2,%b4\n\t" + "je 1f\n\t" + "movsw\n" + "1:\ttestb $1,%b4\n\t" + "je 2f\n\t" + "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2) + :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from) + :"memory"); + return (to); +} +#else +#define __memcpy(a,b,c) memcpy(a,b,c) +#endif + + + +/* ================================================================ + * Debugging: + */ +#define DO_DEBUG 0 +#if DO_DEBUG +extern int INTEL_DEBUG; +#else +#define INTEL_DEBUG 0 +#endif + +#define DEBUG_TEXTURE 0x1 +#define DEBUG_STATE 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_BLIT 0x8 +#define DEBUG_MIPTREE 0x10 +#define DEBUG_FALLBACKS 0x20 +#define DEBUG_VERBOSE 0x40 +#define DEBUG_BATCH 0x80 +#define DEBUG_PIXEL 0x100 +#define DEBUG_BUFMGR 0x200 +#define DEBUG_REGION 0x400 +#define DEBUG_FBO 0x800 +#define DEBUG_LOCK 0x1000 + +#define DBG(...) do { if (INTEL_DEBUG & FILE_DEBUG_FLAG) _mesa_printf(__VA_ARGS__); } while(0) + + +#define PCI_CHIP_845_G 0x2562 +#define PCI_CHIP_I830_M 0x3577 +#define PCI_CHIP_I855_GM 0x3582 +#define PCI_CHIP_I865_G 0x2572 +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GME 0x27AE +#define PCI_CHIP_G33_G 0x29C2 +#define PCI_CHIP_Q35_G 0x29B2 +#define PCI_CHIP_Q33_G 0x29D2 + + +/* ================================================================ + * intel_context.c: + */ + +extern GLboolean intelInitContext(struct intel_context *intel, + const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate, + struct dd_function_table *functions); + +extern void intelGetLock(struct intel_context *intel, GLuint flags); + +extern void intelFinish(GLcontext * ctx); +extern void intelFlush(GLcontext * ctx); + +extern void intelInitDriverFunctions(struct dd_function_table *functions); + + +/* ================================================================ + * intel_state.c: + */ +extern void intelInitStateFuncs(struct dd_function_table *functions); + +#define COMPAREFUNC_ALWAYS 0 +#define COMPAREFUNC_NEVER 0x1 +#define COMPAREFUNC_LESS 0x2 +#define COMPAREFUNC_EQUAL 0x3 +#define COMPAREFUNC_LEQUAL 0x4 +#define COMPAREFUNC_GREATER 0x5 +#define COMPAREFUNC_NOTEQUAL 0x6 +#define COMPAREFUNC_GEQUAL 0x7 + +#define STENCILOP_KEEP 0 +#define STENCILOP_ZERO 0x1 +#define STENCILOP_REPLACE 0x2 +#define STENCILOP_INCRSAT 0x3 +#define STENCILOP_DECRSAT 0x4 +#define STENCILOP_INCR 0x5 +#define STENCILOP_DECR 0x6 +#define STENCILOP_INVERT 0x7 + +#define LOGICOP_CLEAR 0 +#define LOGICOP_NOR 0x1 +#define LOGICOP_AND_INV 0x2 +#define LOGICOP_COPY_INV 0x3 +#define LOGICOP_AND_RVRSE 0x4 +#define LOGICOP_INV 0x5 +#define LOGICOP_XOR 0x6 +#define LOGICOP_NAND 0x7 +#define LOGICOP_AND 0x8 +#define LOGICOP_EQUIV 0x9 +#define LOGICOP_NOOP 0xa +#define LOGICOP_OR_INV 0xb +#define LOGICOP_COPY 0xc +#define LOGICOP_OR_RVRSE 0xd +#define LOGICOP_OR 0xe +#define LOGICOP_SET 0xf + +#define BLENDFACT_ZERO 0x01 +#define BLENDFACT_ONE 0x02 +#define BLENDFACT_SRC_COLR 0x03 +#define BLENDFACT_INV_SRC_COLR 0x04 +#define BLENDFACT_SRC_ALPHA 0x05 +#define BLENDFACT_INV_SRC_ALPHA 0x06 +#define BLENDFACT_DST_ALPHA 0x07 +#define BLENDFACT_INV_DST_ALPHA 0x08 +#define BLENDFACT_DST_COLR 0x09 +#define BLENDFACT_INV_DST_COLR 0x0a +#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b +#define BLENDFACT_CONST_COLOR 0x0c +#define BLENDFACT_INV_CONST_COLOR 0x0d +#define BLENDFACT_CONST_ALPHA 0x0e +#define BLENDFACT_INV_CONST_ALPHA 0x0f +#define BLENDFACT_MASK 0x0f + +#define MI_BATCH_BUFFER_END (0xA<<23) + + +extern int intel_translate_compare_func(GLenum func); +extern int intel_translate_stencil_op(GLenum op); +extern int intel_translate_blend_factor(GLenum factor); +extern int intel_translate_logic_op(GLenum opcode); + + +/*====================================================================== + * Inline conversion functions. + * These are better-typed than the macros used previously: + */ +static INLINE struct intel_context * +intel_context(GLcontext * ctx) +{ + return (struct intel_context *) ctx; +} + +static INLINE struct intel_texture_object * +intel_texture_object(struct gl_texture_object *obj) +{ + return (struct intel_texture_object *) obj; +} + +static INLINE struct intel_texture_image * +intel_texture_image(struct gl_texture_image *img) +{ + return (struct intel_texture_image *) img; +} + +extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer + *rb); + + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c b/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c new file mode 100644 index 0000000000..d269a85a3c --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c @@ -0,0 +1,282 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "imports.h" +#include "context.h" +#include "depthstencil.h" +#include "fbobject.h" +#include "framebuffer.h" +#include "hash.h" +#include "mtypes.h" +#include "renderbuffer.h" + +#include "intel_context.h" +#include "intel_fbo.h" +#include "intel_depthstencil.h" +#include "intel_regions.h" + + +/** + * The GL_EXT_framebuffer_object allows the user to create their own + * framebuffer objects consisting of color renderbuffers (0 or more), + * depth renderbuffers (0 or 1) and stencil renderbuffers (0 or 1). + * + * The spec considers depth and stencil renderbuffers to be totally independent + * buffers. In reality, most graphics hardware today uses a combined + * depth+stencil buffer (one 32-bit pixel = 24 bits of Z + 8 bits of stencil). + * + * This causes difficulty because the user may create some number of depth + * renderbuffers and some number of stencil renderbuffers and bind them + * together in framebuffers in any combination. + * + * This code manages all that. + * + * 1. Depth renderbuffers are always allocated in hardware as 32bpp + * GL_DEPTH24_STENCIL8 buffers. + * + * 2. Stencil renderbuffers are initially allocated in software as 8bpp + * GL_STENCIL_INDEX8 buffers. + * + * 3. Depth and Stencil renderbuffers use the PairedStencil and PairedDepth + * fields (respectively) to indicate if the buffer's currently paired + * with another stencil or depth buffer (respectively). + * + * 4. When a depth and stencil buffer are initially both attached to the + * current framebuffer, we merge the stencil buffer values into the + * depth buffer (really a depth+stencil buffer). The then hardware uses + * the combined buffer. + * + * 5. Whenever a depth or stencil buffer is reallocated (with + * glRenderbufferStorage) we undo the pairing and copy the stencil values + * from the combined depth/stencil buffer back to the stencil-only buffer. + * + * 6. We also undo the pairing when we find a change in buffer bindings. + * + * 7. If a framebuffer is only using a depth renderbuffer (no stencil), we + * just use the combined depth/stencil buffer and ignore the stencil values. + * + * 8. If a framebuffer is only using a stencil renderbuffer (no depth) we have + * to promote the 8bpp software stencil buffer to a 32bpp hardware + * depth+stencil buffer. + * + */ + + + +static void +map_regions(GLcontext * ctx, + struct intel_renderbuffer *depthRb, + struct intel_renderbuffer *stencilRb) +{ + struct intel_context *intel = intel_context(ctx); + if (depthRb && depthRb->region) { + intel_region_map(intel->intelScreen, depthRb->region); + depthRb->pfMap = depthRb->region->map; + depthRb->pfPitch = depthRb->region->pitch; + } + if (stencilRb && stencilRb->region) { + intel_region_map(intel->intelScreen, stencilRb->region); + stencilRb->pfMap = stencilRb->region->map; + stencilRb->pfPitch = stencilRb->region->pitch; + } +} + +static void +unmap_regions(GLcontext * ctx, + struct intel_renderbuffer *depthRb, + struct intel_renderbuffer *stencilRb) +{ + struct intel_context *intel = intel_context(ctx); + if (depthRb && depthRb->region) { + intel_region_unmap(intel->intelScreen, depthRb->region); + depthRb->pfMap = NULL; + depthRb->pfPitch = 0; + } + if (stencilRb && stencilRb->region) { + intel_region_unmap(intel->intelScreen, stencilRb->region); + stencilRb->pfMap = NULL; + stencilRb->pfPitch = 0; + } +} + + + +/** + * Undo the pairing/interleaving between depth and stencil buffers. + * irb should be a depth/stencil or stencil renderbuffer. + */ +void +intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb) +{ + if (irb->PairedStencil) { + /* irb is a depth/stencil buffer */ + struct gl_renderbuffer *stencilRb; + struct intel_renderbuffer *stencilIrb; + + ASSERT(irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + + stencilRb = _mesa_lookup_renderbuffer(ctx, irb->PairedStencil); + stencilIrb = intel_renderbuffer(stencilRb); + if (stencilIrb) { + /* need to extract stencil values from the depth buffer */ + ASSERT(stencilIrb->PairedDepth == irb->Base.Name); + map_regions(ctx, irb, stencilIrb); + _mesa_extract_stencil(ctx, &irb->Base, &stencilIrb->Base); + unmap_regions(ctx, irb, stencilIrb); + stencilIrb->PairedDepth = 0; + } + irb->PairedStencil = 0; + } + else if (irb->PairedDepth) { + /* irb is a stencil buffer */ + struct gl_renderbuffer *depthRb; + struct intel_renderbuffer *depthIrb; + + ASSERT(irb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || + irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + + depthRb = _mesa_lookup_renderbuffer(ctx, irb->PairedDepth); + depthIrb = intel_renderbuffer(depthRb); + if (depthIrb) { + /* need to extract stencil values from the depth buffer */ + ASSERT(depthIrb->PairedStencil == irb->Base.Name); + map_regions(ctx, depthIrb, irb); + _mesa_extract_stencil(ctx, &depthIrb->Base, &irb->Base); + unmap_regions(ctx, depthIrb, irb); + depthIrb->PairedStencil = 0; + } + irb->PairedDepth = 0; + } + else { + _mesa_problem(ctx, "Problem in undo_depth_stencil_pairing"); + } + + ASSERT(irb->PairedStencil == 0); + ASSERT(irb->PairedDepth == 0); +} + + +/** + * Examine the depth and stencil renderbuffers which are attached to the + * framebuffer. If both depth and stencil are attached, make sure that the + * renderbuffers are 'paired' (combined). If only depth or only stencil is + * attached, undo any previous pairing. + * + * Must be called if NewState & _NEW_BUFFER (when renderbuffer attachments + * change, for example). + */ +void +intel_validate_paired_depth_stencil(GLcontext * ctx, + struct gl_framebuffer *fb) +{ + struct intel_renderbuffer *depthRb, *stencilRb; + + depthRb = intel_get_renderbuffer(fb, BUFFER_DEPTH); + stencilRb = intel_get_renderbuffer(fb, BUFFER_STENCIL); + + if (depthRb && stencilRb) { + if (depthRb == stencilRb) { + /* Using a user-created combined depth/stencil buffer. + * Nothing to do. + */ + ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_STENCIL_EXT); + ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + } + else { + /* Separate depth/stencil buffers, need to interleave now */ + ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_COMPONENT); + ASSERT(stencilRb->Base._BaseFormat == GL_STENCIL_INDEX); + /* may need to interleave depth/stencil now */ + if (depthRb->PairedStencil == stencilRb->Base.Name) { + /* OK, the depth and stencil buffers are already interleaved */ + ASSERT(stencilRb->PairedDepth == depthRb->Base.Name); + } + else { + /* need to setup new pairing/interleaving */ + if (depthRb->PairedStencil) { + intel_unpair_depth_stencil(ctx, depthRb); + } + if (stencilRb->PairedDepth) { + intel_unpair_depth_stencil(ctx, stencilRb); + } + + ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + ASSERT(stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || + stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + + /* establish new pairing: interleave stencil into depth buffer */ + map_regions(ctx, depthRb, stencilRb); + _mesa_insert_stencil(ctx, &depthRb->Base, &stencilRb->Base); + unmap_regions(ctx, depthRb, stencilRb); + depthRb->PairedStencil = stencilRb->Base.Name; + stencilRb->PairedDepth = depthRb->Base.Name; + } + + } + } + else if (depthRb) { + /* Depth buffer but no stencil buffer. + * We'll use a GL_DEPTH24_STENCIL8 buffer and ignore the stencil bits. + */ + /* can't assert this until storage is allocated: + ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + */ + /* intel_undo any previous pairing */ + if (depthRb->PairedStencil) { + intel_unpair_depth_stencil(ctx, depthRb); + } + } + else if (stencilRb) { + /* Stencil buffer but no depth buffer. + * Since h/w doesn't typically support just 8bpp stencil w/out Z, + * we'll use a GL_DEPTH24_STENCIL8 buffer and ignore the depth bits. + */ + /* undo any previous pairing */ + if (stencilRb->PairedDepth) { + intel_unpair_depth_stencil(ctx, stencilRb); + } + if (stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT) { + /* promote buffer to GL_DEPTH24_STENCIL8 for hw rendering */ + _mesa_promote_stencil(ctx, &stencilRb->Base); + ASSERT(stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + } + } + + /* Finally, update the fb->_DepthBuffer and fb->_StencilBuffer fields */ + _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH); + if (depthRb && depthRb->PairedStencil) + _mesa_update_stencil_buffer(ctx, fb, BUFFER_DEPTH); + else + _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL); + + + /* The hardware should use fb->Attachment[BUFFER_DEPTH].Renderbuffer + * first, if present, then fb->Attachment[BUFFER_STENCIL].Renderbuffer + * if present. + */ +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_depthstencil.h b/src/mesa/drivers/dri/i915pipe/intel_depthstencil.h new file mode 100644 index 0000000000..2d3fc48b3a --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_depthstencil.h @@ -0,0 +1,14 @@ + +#ifndef INTEL_DEPTH_STENCIL_H +#define INTEL_DEPTH_STENCIL_H + + +extern void +intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb); + +extern void +intel_validate_paired_depth_stencil(GLcontext * ctx, + struct gl_framebuffer *fb); + + +#endif /* INTEL_DEPTH_STENCIL_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c new file mode 100644 index 0000000000..a2a5c811b1 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -0,0 +1,623 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "imports.h" +#include "mtypes.h" +#include "fbobject.h" +#include "framebuffer.h" +#include "renderbuffer.h" +#include "context.h" +#include "texformat.h" +#include "texrender.h" + +#include "intel_context.h" +#include "intel_buffers.h" +#include "intel_depthstencil.h" +#include "intel_fbo.h" +#include "intel_mipmap_tree.h" +#include "intel_regions.h" +#include "intel_span.h" + + +#define FILE_DEBUG_FLAG DEBUG_FBO + +#define INTEL_RB_CLASS 0x12345678 + + +/* XXX FBO: move this to intel_context.h (inlined) */ +/** + * Return a gl_renderbuffer ptr casted to intel_renderbuffer. + * NULL will be returned if the rb isn't really an intel_renderbuffer. + * This is determiend by checking the ClassID. + */ +struct intel_renderbuffer * +intel_renderbuffer(struct gl_renderbuffer *rb) +{ + struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; + if (irb && irb->Base.ClassID == INTEL_RB_CLASS) { + /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/ + return irb; + } + else + return NULL; +} + + +struct intel_renderbuffer * +intel_get_renderbuffer(struct gl_framebuffer *fb, GLuint attIndex) +{ + return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer); +} + + +void +intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) +{ + int current_page = intel_fb->pf_current_page; + int next_page = (current_page + 1) % intel_fb->pf_num_pages; + struct gl_renderbuffer *tmp_rb; + + /* Exchange renderbuffers if necessary but make sure their reference counts + * are preserved. + */ + if (intel_fb->color_rb[current_page] && + intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer != + &intel_fb->color_rb[current_page]->Base) { + tmp_rb = NULL; + _mesa_reference_renderbuffer(&tmp_rb, + intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + tmp_rb = &intel_fb->color_rb[current_page]->Base; + _mesa_reference_renderbuffer( + &intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer, tmp_rb); + _mesa_reference_renderbuffer(&tmp_rb, NULL); + } + + if (intel_fb->color_rb[next_page] && + intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer != + &intel_fb->color_rb[next_page]->Base) { + tmp_rb = NULL; + _mesa_reference_renderbuffer(&tmp_rb, + intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); + tmp_rb = &intel_fb->color_rb[next_page]->Base; + _mesa_reference_renderbuffer( + &intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer, tmp_rb); + _mesa_reference_renderbuffer(&tmp_rb, NULL); + } +} + + +struct intel_region * +intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) +{ + struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex); + + if (irb) + return irb->region; + else + return NULL; +} + + + +/** + * Create a new framebuffer object. + */ +static struct gl_framebuffer * +intel_new_framebuffer(GLcontext * ctx, GLuint name) +{ + /* Only drawable state in intel_framebuffer at this time, just use Mesa's + * class + */ + return _mesa_new_framebuffer(ctx, name); +} + + +static void +intel_delete_renderbuffer(struct gl_renderbuffer *rb) +{ + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); + struct intel_renderbuffer *irb = intel_renderbuffer(rb); + + ASSERT(irb); + + DBG("freeing renderbuffer\n"); + + if (irb->PairedStencil || irb->PairedDepth) { + intel_unpair_depth_stencil(ctx, irb); + } + + if (intel && irb->region) { + intel_region_release(&irb->region); + } + + _mesa_free(irb); +} + + + +/** + * Return a pointer to a specific pixel in a renderbuffer. + */ +static void * +intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb, + GLint x, GLint y) +{ + /* By returning NULL we force all software rendering to go through + * the span routines. + */ + return NULL; +} + + + +/** + * Called via glRenderbufferStorageEXT() to set the format and allocate + * storage for a user-created (or priv buffer) renderbuffer. + */ +static GLboolean +intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, + GLuint width, GLuint height) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_renderbuffer *irb = intel_renderbuffer(rb); + GLboolean softwareBuffer = GL_FALSE; + int cpp; + + switch (internalFormat) { + case GL_R3_G3_B2: + case GL_RGB4: + case GL_RGB5: + rb->_ActualFormat = GL_RGB5; + rb->DataType = GL_UNSIGNED_BYTE; + rb->RedBits = 5; + rb->GreenBits = 6; + rb->BlueBits = 5; + cpp = 2; + break; + case GL_RGB: + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + case GL_RGBA: + case GL_RGBA2: + case GL_RGBA4: + case GL_RGB5_A1: + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + rb->_ActualFormat = GL_RGBA8; + rb->DataType = GL_UNSIGNED_BYTE; + rb->RedBits = 8; + rb->GreenBits = 8; + rb->BlueBits = 8; + rb->AlphaBits = 8; + cpp = 4; + break; + case GL_STENCIL_INDEX: + case GL_STENCIL_INDEX1_EXT: + case GL_STENCIL_INDEX4_EXT: + case GL_STENCIL_INDEX8_EXT: + case GL_STENCIL_INDEX16_EXT: + /* alloc a depth+stencil buffer */ + rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->DataType = GL_UNSIGNED_INT_24_8_EXT; + rb->StencilBits = 8; + cpp = 4; + break; + case GL_DEPTH_COMPONENT16: + rb->_ActualFormat = GL_DEPTH_COMPONENT16; + rb->DataType = GL_UNSIGNED_SHORT; + rb->DepthBits = 16; + cpp = 2; + break; + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->DataType = GL_UNSIGNED_INT_24_8_EXT; + rb->DepthBits = 24; + cpp = 4; + break; + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->DataType = GL_UNSIGNED_INT_24_8_EXT; + rb->DepthBits = 24; + rb->StencilBits = 8; + cpp = 4; + break; + default: + _mesa_problem(ctx, + "Unexpected format (%x) in intel_alloc_renderbuffer_storage", internalFormat); + return GL_FALSE; + } + + intelFlush(ctx); + + /* free old region */ + if (irb->region) { + intel_region_release(&irb->region); + } + + /* allocate new memory region/renderbuffer */ + if (softwareBuffer) { + return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat, + width, height); + } + else { + /* Choose a pitch to match hardware requirements: + */ + GLuint pitch = ((cpp * width + 63) & ~63) / cpp; + + /* alloc hardware renderbuffer */ + DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, + height, pitch); + + irb->region = intel_region_alloc(intel->intelScreen, cpp, pitch, height); + if (!irb->region) + return GL_FALSE; /* out of memory? */ + + ASSERT(irb->region->buffer); + + rb->Width = width; + rb->Height = height; + +#if 1 + /* update the surface's size too */ + rb->surface->width = width; + rb->surface->height = height; +#endif + + /* This sets the Get/PutRow/Value functions */ + intel_set_span_functions(&irb->Base); + + return GL_TRUE; + } +} + + +static void +intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, + GLuint width, GLuint height) +{ + struct intel_framebuffer *intel_fb = (struct intel_framebuffer*)fb; + int i; + + _mesa_resize_framebuffer(ctx, fb, width, height); + + fb->Initialized = GL_TRUE; /* XXX remove someday */ + + if (fb->Name != 0) { + return; + } + + /* Make sure all window system renderbuffers are up to date */ + for (i = 0; i < 3; i++) { + struct gl_renderbuffer *rb = &intel_fb->color_rb[i]->Base; + + /* only resize if size is changing */ + if (rb && (rb->Width != width || rb->Height != height)) { + rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height); + } + } +} + +static GLboolean +intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, GLuint width, GLuint height) +{ + _mesa_problem(ctx, "intel_op_alloc_storage should never be called."); + return GL_FALSE; +} + + + +struct intel_renderbuffer * +intel_new_renderbuffer_fb(GLuint intFormat) +{ + struct intel_renderbuffer *irb; + + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, 0); + irb->Base.ClassID = INTEL_RB_CLASS; + irb->Base.InternalFormat = intFormat; + + switch (intFormat) { + case GL_RGB5: + case GL_RGBA8: + irb->Base._BaseFormat = GL_RGBA; + break; + case GL_DEPTH_COMPONENT16: + irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + break; + case GL_DEPTH24_STENCIL8_EXT: + irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; + break; + default: + _mesa_problem(NULL, + "Unexpected intFormat in intel_create_renderbuffer"); + return NULL; + } + + /* intel-specific methods */ + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; + irb->Base.GetPointer = intel_get_pointer; + /* span routines set in alloc_storage function */ + + irb->Base.surface = intel_new_surface(intFormat); + irb->Base.surface->rb = irb; + + return &irb->Base; +} + +/** + * Create a new renderbuffer object. + * Typically called via glBindRenderbufferEXT(). + */ +static struct gl_renderbuffer * +intel_new_renderbuffer(GLcontext * ctx, GLuint name) +{ + /*struct intel_context *intel = intel_context(ctx); */ + struct intel_renderbuffer *irb; + + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, name); + irb->Base.ClassID = INTEL_RB_CLASS; + + /* intel-specific methods */ + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; + irb->Base.GetPointer = intel_get_pointer; + /* span routines set in alloc_storage function */ + + irb->Base.surface = intel_new_surface(0 /*unknown format*/); + irb->Base.surface->rb = irb; + + return &irb->Base; +} + + +/** + * Called via glBindFramebufferEXT(). + */ +static void +intel_bind_framebuffer(GLcontext * ctx, GLenum target, + struct gl_framebuffer *fb, struct gl_framebuffer *fbread) +{ + if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) { + intel_draw_buffer(ctx, fb); + /* Integer depth range depends on depth buffer bits */ + /* XXX + */ +// ctx->Driver.DepthRange(ctx, ctx->Viewport.Near, ctx->Viewport.Far); + } + else { + /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */ + } +} + + +/** + * Called via glFramebufferRenderbufferEXT(). + */ +static void +intel_framebuffer_renderbuffer(GLcontext * ctx, + struct gl_framebuffer *fb, + GLenum attachment, struct gl_renderbuffer *rb) +{ + DBG("Intel FramebufferRenderbuffer %u %u\n", fb->Name, rb ? rb->Name : 0); + + intelFlush(ctx); + + _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb); + intel_draw_buffer(ctx, fb); +} + + +/** + * When glFramebufferTexture[123]D is called this function sets up the + * gl_renderbuffer wrapper around the texture image. + * This will have the region info needed for hardware rendering. + */ +static struct intel_renderbuffer * +intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage) +{ + const GLuint name = ~0; /* not significant, but distinct for debugging */ + struct intel_renderbuffer *irb; + + /* make an intel_renderbuffer to wrap the texture image */ + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, name); + irb->Base.ClassID = INTEL_RB_CLASS; + + if (texImage->TexFormat == &_mesa_texformat_argb8888) { + irb->Base._ActualFormat = GL_RGBA8; + irb->Base._BaseFormat = GL_RGBA; + DBG("Render to RGBA8 texture OK\n"); + } + else if (texImage->TexFormat == &_mesa_texformat_rgb565) { + irb->Base._ActualFormat = GL_RGB5; + irb->Base._BaseFormat = GL_RGB; + DBG("Render to RGB5 texture OK\n"); + } + else if (texImage->TexFormat == &_mesa_texformat_z16) { + irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; + irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + DBG("Render to DEPTH16 texture OK\n"); + } + else { + DBG("Render to texture BAD FORMAT %d\n", + texImage->TexFormat->MesaFormat); + _mesa_free(irb); + return NULL; + } + + irb->Base.InternalFormat = irb->Base._ActualFormat; + irb->Base.Width = texImage->Width; + irb->Base.Height = texImage->Height; + irb->Base.DataType = GL_UNSIGNED_BYTE; /* FBO XXX fix */ + irb->Base.RedBits = texImage->TexFormat->RedBits; + irb->Base.GreenBits = texImage->TexFormat->GreenBits; + irb->Base.BlueBits = texImage->TexFormat->BlueBits; + irb->Base.AlphaBits = texImage->TexFormat->AlphaBits; + irb->Base.DepthBits = texImage->TexFormat->DepthBits; + + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_nop_alloc_storage; + intel_set_span_functions(&irb->Base); + + irb->RenderToTexture = GL_TRUE; + + return irb; +} + + +/** + * Called by glFramebufferTexture[123]DEXT() (and other places) to + * prepare for rendering into texture memory. This might be called + * many times to choose different texture levels, cube faces, etc + * before intel_finish_render_texture() is ever called. + */ +static void +intel_render_texture(GLcontext * ctx, + struct gl_framebuffer *fb, + struct gl_renderbuffer_attachment *att) +{ + struct gl_texture_image *newImage + = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; + struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); + struct intel_texture_image *intel_image; + GLuint imageOffset; + + (void) fb; + + ASSERT(newImage); + + if (!irb) { + irb = intel_wrap_texture(ctx, newImage); + if (irb) { + /* bind the wrapper to the attachment point */ + _mesa_reference_renderbuffer(&att->Renderbuffer, &irb->Base); + } + else { + /* fallback to software rendering */ + _mesa_render_texture(ctx, fb, att); + return; + } + } + + DBG("Begin render texture tid %x tex=%u w=%d h=%d refcount=%d\n", + _glthread_GetID(), + att->Texture->Name, newImage->Width, newImage->Height, + irb->Base.RefCount); + + /* point the renderbufer's region to the texture image region */ + intel_image = intel_texture_image(newImage); + if (irb->region != intel_image->mt->region) { + if (irb->region) + intel_region_release(&irb->region); + intel_region_reference(&irb->region, intel_image->mt->region); + } + + /* compute offset of the particular 2D image within the texture region */ + imageOffset = intel_miptree_image_offset(intel_image->mt, + att->CubeMapFace, + att->TextureLevel); + + if (att->Texture->Target == GL_TEXTURE_3D) { + const GLuint *offsets = intel_miptree_depth_offsets(intel_image->mt, + att->TextureLevel); + imageOffset += offsets[att->Zoffset]; + } + + /* store that offset in the region */ + intel_image->mt->region->draw_offset = imageOffset; + + /* update drawing region, etc */ + intel_draw_buffer(ctx, fb); +} + + +/** + * Called by Mesa when rendering to a texture is done. + */ +static void +intel_finish_render_texture(GLcontext * ctx, + struct gl_renderbuffer_attachment *att) +{ + struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); + + DBG("End render texture (tid %x) tex %u\n", _glthread_GetID(), att->Texture->Name); + + if (irb) { + /* just release the region */ + intel_region_release(&irb->region); + } + else if (att->Renderbuffer) { + /* software fallback */ + _mesa_finish_render_texture(ctx, att); + /* XXX FBO: Need to unmap the buffer (or in intelSpanRenderStart???) */ + } +} + + +/** + * Do one-time context initializations related to GL_EXT_framebuffer_object. + * Hook in device driver functions. + */ +void +intel_fbo_init(struct intel_context *intel) +{ + intel->ctx.Driver.NewFramebuffer = intel_new_framebuffer; + intel->ctx.Driver.NewRenderbuffer = intel_new_renderbuffer; + intel->ctx.Driver.BindFramebuffer = intel_bind_framebuffer; + intel->ctx.Driver.FramebufferRenderbuffer = intel_framebuffer_renderbuffer; + intel->ctx.Driver.RenderTexture = intel_render_texture; + intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture; + intel->ctx.Driver.ResizeBuffers = intel_resize_buffers; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.h b/src/mesa/drivers/dri/i915pipe/intel_fbo.h new file mode 100644 index 0000000000..7dc3967533 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.h @@ -0,0 +1,127 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_FBO_H +#define INTEL_FBO_H + + +#include "pipe/p_state.h" +#include "pipe/softpipe/sp_surface.h" + + +struct intel_context; +struct intel_region; + + +/** + * Intel framebuffer, derived from gl_framebuffer. + */ +struct intel_framebuffer +{ + struct gl_framebuffer Base; + + struct intel_renderbuffer *color_rb[3]; + + /* Drawable page flipping state */ + GLboolean pf_active; + GLuint pf_seq; + GLint pf_pipes; + GLint pf_current_page; + GLint pf_num_pages; + + /* VBI + */ + GLuint vbl_seq; + GLuint vblank_flags; + GLuint vbl_waited; + + int64_t swap_ust; + int64_t swap_missed_ust; + + GLuint swap_count; + GLuint swap_missed_count; +}; + + +/** + * Intel renderbuffer, derived from gl_renderbuffer. + * Note: The PairedDepth and PairedStencil fields use renderbuffer IDs, + * not pointers because in some circumstances a deleted renderbuffer could + * result in a dangling pointer here. + */ +struct intel_renderbuffer +{ + struct gl_renderbuffer Base; + struct intel_region *region; + void *pfMap; /* possibly paged flipped map pointer */ + GLuint pfPitch; /* possibly paged flipped pitch */ + GLboolean RenderToTexture; /* RTT? */ + + GLuint PairedDepth; /**< only used if this is a depth renderbuffer */ + GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */ + + GLuint pf_pending; /**< sequence number of pending flip */ + + GLuint vbl_pending; /**< vblank sequence number of pending flip */ + + struct intel_surface *surface; +}; + +#if 0 +extern struct intel_renderbuffer *intel_create_renderbuffer(GLenum intFormat, + GLsizei width, + GLsizei height, + int offset, + int pitch, + int cpp, + void *map); +#endif + +extern struct intel_renderbuffer *intel_new_renderbuffer_fb(GLuint intFormat); + +extern void intel_fbo_init(struct intel_context *intel); + + +/* XXX make inline or macro */ +extern struct intel_renderbuffer *intel_get_renderbuffer(struct gl_framebuffer + *fb, + GLuint attIndex); + +extern void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb); + + +/* XXX make inline or macro */ +extern struct intel_region *intel_get_rb_region(struct gl_framebuffer *fb, + GLuint attIndex); + + + +extern struct pipe_surface * +intel_new_surface(GLuint intFormat); + + +#endif /* INTEL_FBO_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_ioctl.c b/src/mesa/drivers/dri/i915pipe/intel_ioctl.c new file mode 100644 index 0000000000..e349a3ecd9 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_ioctl.c @@ -0,0 +1,137 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include +#include +#include +#include + +#include "mtypes.h" +#include "context.h" +#include "swrast/swrast.h" + +#include "intel_context.h" +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_regions.h" +#include "drm.h" + +#define FILE_DEBUG_FLAG DEBUG_IOCTL + +int +intelEmitIrqLocked(struct intel_context *intel) +{ + drmI830IrqEmit ie; + int ret, seq; + + assert(((*(int *) intel->driHwLock) & ~DRM_LOCK_CONT) == + (DRM_LOCK_HELD | intel->hHWContext)); + + ie.irq_seq = &seq; + + ret = drmCommandWriteRead(intel->driFd, DRM_I830_IRQ_EMIT, + &ie, sizeof(ie)); + if (ret) { + fprintf(stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret); + exit(1); + } + + DBG("%s --> %d\n", __FUNCTION__, seq); + + return seq; +} + +void +intelWaitIrq(struct intel_context *intel, int seq) +{ + int ret; + + DBG("%s %d\n", __FUNCTION__, seq); + + intel->iw.irq_seq = seq; + + do { + ret = + drmCommandWrite(intel->driFd, DRM_I830_IRQ_WAIT, &intel->iw, + sizeof(intel->iw)); + } while (ret == -EAGAIN || ret == -EINTR); + + if (ret) { + fprintf(stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret); + exit(1); + } +} + + +void +intel_batch_ioctl(struct intel_context *intel, + GLuint start_offset, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) +{ + drmI830BatchBuffer batch; + + assert(intel->locked); + assert(used); + + DBG("%s used %d offset %x..%x ignore_cliprects %d\n", + __FUNCTION__, + used, start_offset, start_offset + used, ignore_cliprects); + + /* Throw away non-effective packets. Won't work once we have + * hardware contexts which would preserve statechanges beyond a + * single buffer. + */ + + + + batch.start = start_offset; + batch.used = used; + batch.cliprects = intel->pClipRects; + batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + batch.DR1 = 0; + batch.DR4 = 0; /* still need this ? */ + + DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n", + __FUNCTION__, + batch.start, + batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); + + if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, + sizeof(batch))) { + fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n", -errno); + UNLOCK_HARDWARE(intel); + exit(1); + } + + /* FIXME: use hardware contexts to avoid 'losing' hardware after + * each buffer flush. + */ + //intel->vtbl.lost_hardware(intel); +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_ioctl.h b/src/mesa/drivers/dri/i915pipe/intel_ioctl.h new file mode 100644 index 0000000000..e8d07de893 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_ioctl.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_IOCTL_H +#define INTEL_IOCTL_H + +#include "intel_context.h" + +void intelWaitIrq(struct intel_context *intel, int seq); +int intelEmitIrqLocked(struct intel_context *intel); + +void intel_batch_ioctl(struct intel_context *intel, + GLuint start_offset, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock); +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c new file mode 100644 index 0000000000..88a9277427 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c @@ -0,0 +1,358 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_context.h" +#include "intel_mipmap_tree.h" +#include "intel_regions.h" +#include "enums.h" + +#define FILE_DEBUG_FLAG DEBUG_MIPTREE + +static GLenum +target_to_target(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + return GL_TEXTURE_CUBE_MAP_ARB; + default: + return target; + } +} + +struct intel_mipmap_tree * +intel_miptree_create(struct intel_context *intel, + GLenum target, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, GLuint cpp, GLuint compress_byte) +{ + GLboolean ok; + struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1); + + DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), + _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); + + mt->target = target_to_target(target); + mt->internal_format = internal_format; + mt->first_level = first_level; + mt->last_level = last_level; + mt->width0 = width0; + mt->height0 = height0; + mt->depth0 = depth0; + mt->cpp = compress_byte ? compress_byte : cpp; + mt->compressed = compress_byte ? 1 : 0; + mt->refcount = 1; + + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: +// ok = i945_miptree_layout(mt); + break; + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + case PCI_CHIP_I830_M: + case PCI_CHIP_I855_GM: + case PCI_CHIP_I865_G: + default: + /* All the i830 chips and the i915 use this layout: + */ +// ok = i915_miptree_layout(mt); + break; + } + + ok = 0; /* TODO */ + + if (ok) + mt->region = intel_region_alloc(intel->intelScreen, + mt->cpp, mt->pitch, mt->total_height); + + if (!mt->region) { + free(mt); + return NULL; + } + + return mt; +} + + +void +intel_miptree_reference(struct intel_mipmap_tree **dst, + struct intel_mipmap_tree *src) +{ + src->refcount++; + *dst = src; + DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); +} + +void +intel_miptree_release(struct intel_context *intel, + struct intel_mipmap_tree **mt) +{ + if (!*mt) + return; + + DBG("%s %p refcount will be %d\n", __FUNCTION__, *mt, (*mt)->refcount - 1); + if (--(*mt)->refcount <= 0) { + GLuint i; + + DBG("%s deleting %p\n", __FUNCTION__, *mt); + + intel_region_release(&((*mt)->region)); + + for (i = 0; i < MAX_TEXTURE_LEVELS; i++) + if ((*mt)->level[i].image_offset) + free((*mt)->level[i].image_offset); + + free(*mt); + } + *mt = NULL; +} + + + + +/* Can the image be pulled into a unified mipmap tree. This mirrors + * the completeness test in a lot of ways. + * + * Not sure whether I want to pass gl_texture_image here. + */ +GLboolean +intel_miptree_match_image(struct intel_mipmap_tree *mt, + struct gl_texture_image *image, + GLuint face, GLuint level) +{ + /* Images with borders are never pulled into mipmap trees. + */ + if (image->Border) + return GL_FALSE; + + if (image->InternalFormat != mt->internal_format || + image->IsCompressed != mt->compressed) + return GL_FALSE; + + /* Test image dimensions against the base level image adjusted for + * minification. This will also catch images not present in the + * tree, changed targets, etc. + */ + if (image->Width != mt->level[level].width || + image->Height != mt->level[level].height || + image->Depth != mt->level[level].depth) + return GL_FALSE; + + return GL_TRUE; +} + + +void +intel_miptree_set_level_info(struct intel_mipmap_tree *mt, + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) +{ + + mt->level[level].width = w; + mt->level[level].height = h; + mt->level[level].depth = d; + mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; + mt->level[level].nr_images = nr_images; + + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, mt->level[level].level_offset); + + /* Not sure when this would happen, but anyway: + */ + if (mt->level[level].image_offset) { + free(mt->level[level].image_offset); + mt->level[level].image_offset = NULL; + } + + assert(nr_images); + + mt->level[level].image_offset = malloc(nr_images * sizeof(GLuint)); + mt->level[level].image_offset[0] = 0; +} + + + +void +intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, + GLuint level, GLuint img, GLuint x, GLuint y) +{ + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + + assert(img < mt->level[level].nr_images); + + mt->level[level].image_offset[img] = (x + y * mt->pitch); + + DBG("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); +} + + +/* Although we use the image_offset[] array to store relative offsets + * to cube faces, Mesa doesn't know anything about this and expects + * each cube face to be treated as a separate image. + * + * These functions present that view to mesa: + */ +const GLuint * +intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, GLuint level) +{ + static const GLuint zero = 0; + + if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) + return &zero; + else + return mt->level[level].image_offset; +} + + +GLuint +intel_miptree_image_offset(struct intel_mipmap_tree * mt, + GLuint face, GLuint level) +{ + if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) + return (mt->level[level].level_offset + + mt->level[level].image_offset[face] * mt->cpp); + else + return mt->level[level].level_offset; +} + + + +/** + * Map a teximage in a mipmap tree. + * \param row_stride returns row stride in bytes + * \param image_stride returns image stride in bytes (for 3D textures). + * \return address of mapping + */ +GLubyte * +intel_miptree_image_map(struct intel_context * intel, + struct intel_mipmap_tree * mt, + GLuint face, + GLuint level, + GLuint * row_stride, GLuint * image_offsets) +{ + DBG("%s \n", __FUNCTION__); + + if (row_stride) + *row_stride = mt->pitch * mt->cpp; + + if (image_offsets) + memcpy(image_offsets, mt->level[level].image_offset, + mt->level[level].depth * sizeof(GLuint)); + + return (intel_region_map(intel->intelScreen, mt->region) + + intel_miptree_image_offset(mt, face, level)); +} + +void +intel_miptree_image_unmap(struct intel_context *intel, + struct intel_mipmap_tree *mt) +{ + DBG("%s\n", __FUNCTION__); + intel_region_unmap(intel->intelScreen, mt->region); +} + + + +/* Upload data for a particular image. + */ +void +intel_miptree_image_data(struct intel_context *intel, + struct intel_mipmap_tree *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch) +{ + GLuint depth = dst->level[level].depth; + GLuint dst_offset = intel_miptree_image_offset(dst, face, level); + const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); + GLuint i; + GLuint height = 0; + + DBG("%s\n", __FUNCTION__); + for (i = 0; i < depth; i++) { + height = dst->level[level].height; + if(dst->compressed) + height /= 4; + intel_region_data(intel->intelScreen, dst->region, + dst_offset + dst_depth_offset[i], /* dst_offset */ + 0, 0, /* dstx, dsty */ + src, + src_row_pitch, + 0, 0, /* source x, y */ + dst->level[level].width, height); /* width, height */ + + src += src_image_pitch * dst->cpp; + } +} + +/* Copy mipmap image between trees + */ +void +intel_miptree_image_copy(struct intel_context *intel, + struct intel_mipmap_tree *dst, + GLuint face, GLuint level, + struct intel_mipmap_tree *src) +{ + GLuint width = src->level[level].width; + GLuint height = src->level[level].height; + GLuint depth = src->level[level].depth; + GLuint dst_offset = intel_miptree_image_offset(dst, face, level); + GLuint src_offset = intel_miptree_image_offset(src, face, level); + const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); + const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level); + GLuint i; + + if (dst->compressed) + height /= 4; + for (i = 0; i < depth; i++) { + intel_region_copy(intel->intelScreen, + dst->region, dst_offset + dst_depth_offset[i], + 0, + 0, + src->region, src_offset + src_depth_offset[i], + 0, 0, width, height); + } + +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h new file mode 100644 index 0000000000..ecdb7be244 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h @@ -0,0 +1,198 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_MIPMAP_TREE_H +#define INTEL_MIPMAP_TREE_H + +#include "intel_regions.h" + +/* A layer on top of the intel_regions code which adds: + * + * - Code to size and layout a region to hold a set of mipmaps. + * - Query to determine if a new image fits in an existing tree. + * - More refcounting + * - maybe able to remove refcounting from intel_region? + * - ? + * + * The fixed mipmap layout of intel hardware where one offset + * specifies the position of all images in a mipmap hierachy + * complicates the implementation of GL texture image commands, + * compared to hardware where each image is specified with an + * independent offset. + * + * In an ideal world, each texture object would be associated with a + * single bufmgr buffer or 2d intel_region, and all the images within + * the texture object would slot into the tree as they arrive. The + * reality can be a little messier, as images can arrive from the user + * with sizes that don't fit in the existing tree, or in an order + * where the tree layout cannot be guessed immediately. + * + * This structure encodes an idealized mipmap tree. The GL image + * commands build these where possible, otherwise store the images in + * temporary system buffers. + */ + + +/** + * Describes the location of each texture image within a texture region. + */ +struct intel_mipmap_level +{ + GLuint level_offset; + GLuint width; + GLuint height; + GLuint depth; + GLuint nr_images; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + GLuint *image_offset; +}; + +struct intel_mipmap_tree +{ + /* Effectively the key: + */ + GLenum target; + GLenum internal_format; + + GLuint first_level; + GLuint last_level; + + GLuint width0, height0, depth0; /**< Level zero image dimensions */ + GLuint cpp; + GLboolean compressed; + + /* Derived from the above: + */ + GLuint pitch; + GLuint depth_pitch; /* per-image on i945? */ + GLuint total_height; + + /* Includes image offset tables: + */ + struct intel_mipmap_level level[MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct intel_region *region; + + /* These are also refcounted: + */ + GLuint refcount; +}; + + + +struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel, + GLenum target, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, + GLuint cpp, + GLuint compress_byte); + +void intel_miptree_reference(struct intel_mipmap_tree **dst, + struct intel_mipmap_tree *src); + +void intel_miptree_release(struct intel_context *intel, + struct intel_mipmap_tree **mt); + +/* Check if an image fits an existing mipmap tree layout + */ +GLboolean intel_miptree_match_image(struct intel_mipmap_tree *mt, + struct gl_texture_image *image, + GLuint face, GLuint level); + +/* Return a pointer to an image within a tree. Return image stride as + * well. + */ +GLubyte *intel_miptree_image_map(struct intel_context *intel, + struct intel_mipmap_tree *mt, + GLuint face, + GLuint level, + GLuint * row_stride, GLuint * image_stride); + +void intel_miptree_image_unmap(struct intel_context *intel, + struct intel_mipmap_tree *mt); + + +/* Return the linear offset of an image relative to the start of the + * tree: + */ +GLuint intel_miptree_image_offset(struct intel_mipmap_tree *mt, + GLuint face, GLuint level); + +/* Return pointers to each 2d slice within an image. Indexed by depth + * value. + */ +const GLuint *intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, + GLuint level); + + +void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, + GLuint w, GLuint h, GLuint d); + +void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, + GLuint level, + GLuint img, GLuint x, GLuint y); + + +/* Upload an image into a tree + */ +void intel_miptree_image_data(struct intel_context *intel, + struct intel_mipmap_tree *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch); + +/* Copy an image between two trees + */ +void intel_miptree_image_copy(struct intel_context *intel, + struct intel_mipmap_tree *dst, + GLuint face, GLuint level, + struct intel_mipmap_tree *src); + +/* i915_mipmap_tree.c: + */ +GLboolean i915_miptree_layout(struct intel_mipmap_tree *mt); +GLboolean i945_miptree_layout(struct intel_mipmap_tree *mt); + + + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_pixel.c b/src/mesa/drivers/dri/i915pipe/intel_pixel.c new file mode 100644 index 0000000000..104d288de0 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_pixel.c @@ -0,0 +1,45 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portionsalloc + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "enums.h" +#include "state.h" +#include "swrast/swrast.h" + +#include "intel_context.h" +#include "intel_pixel.h" +#include "intel_regions.h" + + +void +intelInitPixelFuncs(struct dd_function_table *functions) +{ + functions->Accum = _swrast_Accum; + functions->Bitmap = _swrast_Bitmap; + functions->CopyPixels = _swrast_CopyPixels; + functions->ReadPixels = _swrast_ReadPixels; + functions->DrawPixels = _swrast_DrawPixels; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_pixel.h b/src/mesa/drivers/dri/i915pipe/intel_pixel.h new file mode 100644 index 0000000000..a6fcf90ce0 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_pixel.h @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_PIXEL_H +#define INTEL_PIXEL_H + +#include "mtypes.h" + +void intelInitPixelFuncs(struct dd_function_table *functions); + +GLboolean intel_check_blit_fragment_ops(GLcontext * ctx); + +GLboolean intel_check_meta_tex_fragment_ops(GLcontext * ctx); + +GLboolean intel_check_blit_format(struct intel_region *region, + GLenum format, GLenum type); + + +void intelReadPixels(GLcontext * ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *pack, + GLvoid * pixels); + +void intelDrawPixels(GLcontext * ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + GLenum format, + GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid * pixels); + +void intelCopyPixels(GLcontext * ctx, + GLint srcx, GLint srcy, + GLsizei width, GLsizei height, + GLint destx, GLint desty, GLenum type); + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_reg.h b/src/mesa/drivers/dri/i915pipe/intel_reg.h new file mode 100644 index 0000000000..7828ba6ad3 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_reg.h @@ -0,0 +1,88 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef _INTEL_REG_H_ +#define _INTEL_REG_H_ + + + +#define CMD_3D (0x3<<29) + + +#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) +#define PRIM_INDIRECT (1<<23) +#define PRIM_INLINE (0<<23) +#define PRIM_INDIRECT_SEQUENTIAL (0<<17) +#define PRIM_INDIRECT_ELTS (1<<17) + +#define PRIM3D_TRILIST (0x0<<18) +#define PRIM3D_TRISTRIP (0x1<<18) +#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) +#define PRIM3D_TRIFAN (0x3<<18) +#define PRIM3D_POLY (0x4<<18) +#define PRIM3D_LINELIST (0x5<<18) +#define PRIM3D_LINESTRIP (0x6<<18) +#define PRIM3D_RECTLIST (0x7<<18) +#define PRIM3D_POINTLIST (0x8<<18) +#define PRIM3D_DIB (0x9<<18) +#define PRIM3D_MASK (0x1f<<18) + +#define I915PACKCOLOR4444(r,g,b,a) \ + ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) + +#define I915PACKCOLOR1555(r,g,b,a) \ + ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ + ((a) ? 0x8000 : 0)) + +#define I915PACKCOLOR565(r,g,b) \ + ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) + +#define I915PACKCOLOR8888(r,g,b,a) \ + ((a<<24) | (r<<16) | (g<<8) | b) + + + + +#define BR00_BITBLT_CLIENT 0x40000000 +#define BR00_OP_COLOR_BLT 0x10000000 +#define BR00_OP_SRC_COPY_BLT 0x10C00000 +#define BR13_SOLID_PATTERN 0x80000000 + +#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) +#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) +#define XY_COLOR_BLT_WRITE_RGB (1<<20) + +#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) +#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) +#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) + +#define MI_WAIT_FOR_EVENT ((0x3<<23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_regions.c b/src/mesa/drivers/dri/i915pipe/intel_regions.c new file mode 100644 index 0000000000..7d19bd07d3 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_regions.c @@ -0,0 +1,480 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Provide additional functionality on top of bufmgr buffers: + * - 2d semantics and blit operations + * - refcounting of buffers for multiple images in a buffer. + * - refcounting of buffer mappings. + * - some logic for moving the buffers to the best memory pools for + * given operations. + * + * Most of this is to make it easier to implement the fixed-layout + * mipmap tree required by intel hardware in the face of GL's + * programming interface where each image can be specifed in random + * order and it isn't clear what layout the tree should have until the + * last moment. + */ + +#include "intel_context.h" +#include "intel_regions.h" +#include "intel_blit.h" +#include "intel_buffer_objects.h" +#include "dri_bufmgr.h" +#include "intel_batchbuffer.h" + +#define FILE_DEBUG_FLAG DEBUG_REGION + +void +intel_region_idle(intelScreenPrivate *intelScreen, struct intel_region *region) +{ + DBG("%s\n", __FUNCTION__); + if (region && region->buffer) + driBOWaitIdle(region->buffer, GL_FALSE); +} + +/* XXX: Thread safety? + */ +GLubyte * +intel_region_map(intelScreenPrivate *intelScreen, struct intel_region *region) +{ + DBG("%s\n", __FUNCTION__); + if (!region->map_refcount++) { + if (region->pbo) + intel_region_cow(intelScreen, region); + + region->map = driBOMap(region->buffer, + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); + } + + return region->map; +} + +void +intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *region) +{ + DBG("%s\n", __FUNCTION__); + if (!--region->map_refcount) { + driBOUnmap(region->buffer); + region->map = NULL; + } +} + +#undef TEST_CACHED_TEXTURES + +struct intel_region * +intel_region_alloc(intelScreenPrivate *intelScreen, + GLuint cpp, GLuint pitch, GLuint height) +{ + struct intel_region *region = calloc(sizeof(*region), 1); + struct intel_context *intel = intelScreenContext(intelScreen); + + DBG("%s\n", __FUNCTION__); + + region->cpp = cpp; + region->pitch = pitch; + region->height = height; /* needed? */ + region->refcount = 1; + + driGenBuffers(intelScreen->regionPool, + "region", 1, ®ion->buffer, 64, +#ifdef TEST_CACHED_TEXTURES + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_BIND_CACHED | + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, +#else + 0, +#endif + 0); + LOCK_HARDWARE(intel); + driBOData(region->buffer, pitch * cpp * height, NULL, 0); + UNLOCK_HARDWARE(intel); + return region; +} + +void +intel_region_reference(struct intel_region **dst, struct intel_region *src) +{ + assert(*dst == NULL); + if (src) { + src->refcount++; + *dst = src; + } +} + +void +intel_region_release(struct intel_region **region) +{ + if (!*region) + return; + + DBG("%s %d\n", __FUNCTION__, (*region)->refcount - 1); + + ASSERT((*region)->refcount > 0); + (*region)->refcount--; + + if ((*region)->refcount == 0) { + assert((*region)->map_refcount == 0); + + if ((*region)->pbo) + (*region)->pbo->region = NULL; + (*region)->pbo = NULL; + driBOUnReference((*region)->buffer); + free(*region); + } + *region = NULL; +} + + +struct intel_region * +intel_region_create_static(intelScreenPrivate *intelScreen, + GLuint mem_type, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, GLuint height) +{ + struct intel_region *region = calloc(sizeof(*region), 1); + DBG("%s\n", __FUNCTION__); + + region->cpp = cpp; + region->pitch = pitch; + region->height = height; /* needed? */ + region->refcount = 1; + + /* + * We use a "shared" buffer type to indicate buffers created and + * shared by others. + */ + + driGenBuffers(intelScreen->staticPool, "static region", 1, + ®ion->buffer, 64, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); + driBOSetStatic(region->buffer, offset, pitch * cpp * height, virtual, 0); + + return region; +} + + + +void +intel_region_update_static(intelScreenPrivate *intelScreen, + struct intel_region *region, + GLuint mem_type, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, GLuint height) +{ + DBG("%s\n", __FUNCTION__); + + region->cpp = cpp; + region->pitch = pitch; + region->height = height; /* needed? */ + + /* + * We use a "shared" buffer type to indicate buffers created and + * shared by others. + */ + + driDeleteBuffers(1, ®ion->buffer); + driGenBuffers(intelScreen->staticPool, "static region", 1, + ®ion->buffer, 64, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); + driBOSetStatic(region->buffer, offset, pitch * cpp * height, virtual, 0); + +} + + + +/* + * XXX Move this into core Mesa? + */ +static void +_mesa_copy_rect(GLubyte * dst, + GLuint cpp, + GLuint dst_pitch, + GLuint dst_x, + GLuint dst_y, + GLuint width, + GLuint height, + const GLubyte * src, + GLuint src_pitch, GLuint src_x, GLuint src_y) +{ + GLuint i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * dst_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } +} + + +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +void +intel_region_data(intelScreenPrivate *intelScreen, + struct intel_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + const void *src, GLuint src_pitch, + GLuint srcx, GLuint srcy, GLuint width, GLuint height) +{ + struct intel_context *intel = intelScreenContext(intelScreen); + + DBG("%s\n", __FUNCTION__); + + if (intel == NULL) + return; + + if (dst->pbo) { + if (dstx == 0 && + dsty == 0 && width == dst->pitch && height == dst->height) + intel_region_release_pbo(intelScreen, dst); + else + intel_region_cow(intelScreen, dst); + } + + + LOCK_HARDWARE(intel); + + _mesa_copy_rect(intel_region_map(intelScreen, dst) + dst_offset, + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + intel_region_unmap(intelScreen, dst); + + UNLOCK_HARDWARE(intel); + +} + +/* Copy rectangular sub-regions. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +void +intel_region_copy(intelScreenPrivate *intelScreen, + struct intel_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + struct intel_region *src, + GLuint src_offset, + GLuint srcx, GLuint srcy, GLuint width, GLuint height) +{ + struct intel_context *intel = intelScreenContext(intelScreen); + + DBG("%s\n", __FUNCTION__); + + if (intel == NULL) + return; + + if (dst->pbo) { + if (dstx == 0 && + dsty == 0 && width == dst->pitch && height == dst->height) + intel_region_release_pbo(intelScreen, dst); + else + intel_region_cow(intelScreen, dst); + } + + assert(src->cpp == dst->cpp); + + intelEmitCopyBlit(intel, + dst->cpp, + src->pitch, src->buffer, src_offset, + dst->pitch, dst->buffer, dst_offset, + srcx, srcy, dstx, dsty, width, height, + GL_COPY); +} + +/* Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +void +intel_region_fill(intelScreenPrivate *intelScreen, + struct intel_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + GLuint width, GLuint height, GLuint color) +{ + struct intel_context *intel = intelScreenContext(intelScreen); + + DBG("%s\n", __FUNCTION__); + + if (intel == NULL) + return; + + if (dst->pbo) { + if (dstx == 0 && + dsty == 0 && width == dst->pitch && height == dst->height) + intel_region_release_pbo(intelScreen, dst); + else + intel_region_cow(intelScreen, dst); + } + + intelEmitFillBlit(intel, + dst->cpp, + dst->pitch, dst->buffer, dst_offset, + dstx, dsty, width, height, color); +} + +/* Attach to a pbo, discarding our data. Effectively zero-copy upload + * the pbo's data. + */ +void +intel_region_attach_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region, + struct intel_buffer_object *pbo) +{ + if (region->pbo == pbo) + return; + + /* If there is already a pbo attached, break the cow tie now. + * Don't call intel_region_release_pbo() as that would + * unnecessarily allocate a new buffer we would have to immediately + * discard. + */ + if (region->pbo) { + region->pbo->region = NULL; + region->pbo = NULL; + } + + if (region->buffer) { + driDeleteBuffers(1, ®ion->buffer); + region->buffer = NULL; + } + + region->pbo = pbo; + region->pbo->region = region; + region->buffer = driBOReference(pbo->buffer); +} + + +/* Break the COW tie to the pbo. The pbo gets to keep the data. + */ +void +intel_region_release_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region) +{ + struct intel_context *intel = intelScreenContext(intelScreen); + + assert(region->buffer == region->pbo->buffer); + region->pbo->region = NULL; + region->pbo = NULL; + driBOUnReference(region->buffer); + region->buffer = NULL; + + driGenBuffers(intelScreen->regionPool, + "region", 1, ®ion->buffer, 64, 0, 0); + + LOCK_HARDWARE(intel); + driBOData(region->buffer, + region->cpp * region->pitch * region->height, NULL, 0); + UNLOCK_HARDWARE(intel); +} + +/* Break the COW tie to the pbo. Both the pbo and the region end up + * with a copy of the data. + */ +void +intel_region_cow(intelScreenPrivate *intelScreen, struct intel_region *region) +{ + struct intel_context *intel = intelScreenContext(intelScreen); + struct intel_buffer_object *pbo = region->pbo; + + if (intel == NULL) + return; + + intel_region_release_pbo(intelScreen, region); + + assert(region->cpp * region->pitch * region->height == pbo->Base.Size); + + DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size); + + /* Now blit from the texture buffer to the new buffer: + */ + + intel_batchbuffer_flush(intel->batch); + + if (!intel->locked) { + LOCK_HARDWARE(intel); + intelEmitCopyBlit(intel, + region->cpp, + region->pitch, + region->buffer, 0, + region->pitch, + pbo->buffer, 0, + 0, 0, 0, 0, + region->pitch, region->height, + GL_COPY); + + intel_batchbuffer_flush(intel->batch); + UNLOCK_HARDWARE(intel); + } + else { + intelEmitCopyBlit(intel, + region->cpp, + region->pitch, + region->buffer, 0, + region->pitch, + pbo->buffer, 0, + 0, 0, 0, 0, + region->pitch, region->height, + GL_COPY); + + intel_batchbuffer_flush(intel->batch); + } +} + +struct _DriBufferObject * +intel_region_buffer(intelScreenPrivate *intelScreen, + struct intel_region *region, GLuint flag) +{ + if (region->pbo) { + if (flag == INTEL_WRITE_PART) + intel_region_cow(intelScreen, region); + else if (flag == INTEL_WRITE_FULL) + intel_region_release_pbo(intelScreen, region); + } + + return region->buffer; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_regions.h b/src/mesa/drivers/dri/i915pipe/intel_regions.h new file mode 100644 index 0000000000..d938c107a4 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_regions.h @@ -0,0 +1,141 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_REGIONS_H +#define INTEL_REGIONS_H + +#include "mtypes.h" +#include "intel_screen.h" + +struct intel_context; +struct intel_buffer_object; + +/** + * A layer on top of the bufmgr buffers that adds a few useful things: + * + * - Refcounting for local buffer references. + * - Refcounting for buffer maps + * - Buffer dimensions - pitch and height. + * - Blitter commands for copying 2D regions between buffers. (really???) + */ +struct intel_region +{ + struct _DriBufferObject *buffer; /**< buffer manager's buffer ID */ + GLuint refcount; /**< Reference count for region */ + GLuint cpp; /**< bytes per pixel */ + GLuint pitch; /**< in pixels */ + GLuint height; /**< in pixels */ + GLubyte *map; /**< only non-NULL when region is actually mapped */ + GLuint map_refcount; /**< Reference count for mapping */ + + GLuint draw_offset; /**< Offset of drawing address within the region */ + + struct intel_buffer_object *pbo; /* zero-copy uploads */ +}; + + +/* Allocate a refcounted region. Pointers to regions should only be + * copied by calling intel_reference_region(). + */ +struct intel_region *intel_region_alloc(intelScreenPrivate *intelScreen, + GLuint cpp, + GLuint pitch, GLuint height); + +void intel_region_reference(struct intel_region **dst, + struct intel_region *src); + +void intel_region_release(struct intel_region **ib); + +extern struct intel_region +*intel_region_create_static(intelScreenPrivate *intelScreen, + GLuint mem_type, + GLuint offset, + void *virtual, + GLuint cpp, + GLuint pitch, GLuint height); +extern void +intel_region_update_static(intelScreenPrivate *intelScreen, + struct intel_region *region, + GLuint mem_type, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, GLuint height); + + +void intel_region_idle(intelScreenPrivate *intelScreen, + struct intel_region *ib); + +/* Map/unmap regions. This is refcounted also: + */ +GLubyte *intel_region_map(intelScreenPrivate *intelScreen, + struct intel_region *ib); + +void intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *ib); + + +/* Upload data to a rectangular sub-region + */ +void intel_region_data(intelScreenPrivate *intelScreen, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + const void *src, GLuint src_stride, + GLuint srcx, GLuint srcy, GLuint width, GLuint height); + +/* Copy rectangular sub-regions + */ +void intel_region_copy(intelScreenPrivate *intelScreen, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + struct intel_region *src, + GLuint src_offset, + GLuint srcx, GLuint srcy, GLuint width, GLuint height); + +/* Fill a rectangular sub-region + */ +void intel_region_fill(intelScreenPrivate *intelScreen, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + GLuint width, GLuint height, GLuint color); + +/* Helpers for zerocopy uploads, particularly texture image uploads: + */ +void intel_region_attach_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region, + struct intel_buffer_object *pbo); +void intel_region_release_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region); +void intel_region_cow(intelScreenPrivate *intelScreen, + struct intel_region *region); + +struct _DriBufferObject *intel_region_buffer(intelScreenPrivate *intelScreen, + struct intel_region *region, + GLuint flag); + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_render.c b/src/mesa/drivers/dri/i915pipe/intel_render.c new file mode 100644 index 0000000000..c8b6d308d9 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_render.c @@ -0,0 +1,244 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * Render unclipped vertex buffers by emitting vertices directly to + * dma buffers. Use strip/fan hardware acceleration where possible. + * + */ +#include "glheader.h" +#include "context.h" +#include "macros.h" +#include "imports.h" +#include "mtypes.h" +#include "enums.h" + +#include "tnl/t_context.h" +#include "tnl/t_vertex.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_tris.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" + +/* + * Render unclipped vertex buffers by emitting vertices directly to + * dma buffers. Use strip/fan hardware primitives where possible. + * Try to simulate missing primitives with indexed vertices. + */ +#define HAVE_POINTS 0 /* Has it, but can't use because subpixel has to + * be adjusted for points on the INTEL/I845G + */ +#define HAVE_LINES 1 +#define HAVE_LINE_STRIPS 1 +#define HAVE_TRIANGLES 1 +#define HAVE_TRI_STRIPS 1 +#define HAVE_TRI_STRIP_1 0 /* has it, template can't use it yet */ +#define HAVE_TRI_FANS 1 +#define HAVE_POLYGONS 1 +#define HAVE_QUADS 0 +#define HAVE_QUAD_STRIPS 0 + +#define HAVE_ELTS 0 + +static GLuint hw_prim[GL_POLYGON + 1] = { + 0, + PRIM3D_LINELIST, + PRIM3D_LINESTRIP, + PRIM3D_LINESTRIP, + PRIM3D_TRILIST, + PRIM3D_TRISTRIP, + PRIM3D_TRIFAN, + 0, + 0, + PRIM3D_POLY +}; + +static const GLenum reduced_prim[GL_POLYGON + 1] = { + GL_POINTS, + GL_LINES, + GL_LINES, + GL_LINES, + GL_TRIANGLES, + GL_TRIANGLES, + GL_TRIANGLES, + GL_TRIANGLES, + GL_TRIANGLES, + GL_TRIANGLES +}; + +static const int scale_prim[GL_POLYGON + 1] = { + 0, /* fallback case */ + 1, + 2, + 2, + 1, + 3, + 3, + 0, /* fallback case */ + 0, /* fallback case */ + 3 +}; + + +static void +intelDmaPrimitive(struct intel_context *intel, GLenum prim) +{ + if (0) + fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim)); + INTEL_FIREVERTICES(intel); + intel->vtbl.reduced_primitive_state(intel, reduced_prim[prim]); + intelStartInlinePrimitive(intel, hw_prim[prim], INTEL_BATCH_CLIPRECTS); +} + + +#define LOCAL_VARS struct intel_context *intel = intel_context(ctx) +#define INIT( prim ) \ +do { \ + intelDmaPrimitive( intel, prim ); \ +} while (0) + +#define FLUSH() INTEL_FIREVERTICES(intel) + +#define GET_SUBSEQUENT_VB_MAX_VERTS() \ + ((intel->batch->size - 1500) / (intel->vertex_size*4)) +#define GET_CURRENT_VB_MAX_VERTS() GET_SUBSEQUENT_VB_MAX_VERTS() + +#define ALLOC_VERTS( nr ) \ + intelExtendInlinePrimitive( intel, (nr) * intel->vertex_size ) + +#define EMIT_VERTS( ctx, j, nr, buf ) \ + _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf ) + +#define TAG(x) intel_##x +#include "tnl_dd/t_dd_dmatmp.h" + + +/**********************************************************************/ +/* Render pipeline stage */ +/**********************************************************************/ + +/* Heuristic to choose between the two render paths: + */ +static GLboolean +choose_render(struct intel_context *intel, struct vertex_buffer *VB) +{ + int vertsz = intel->vertex_size; + int cost_render = 0; + int cost_fallback = 0; + int nr_prims = 0; + int nr_rprims = 0; + int nr_rverts = 0; + int rprim = intel->reduced_primitive; + int i = 0; + + for (i = 0; i < VB->PrimitiveCount; i++) { + GLuint prim = VB->Primitive[i].mode; + GLuint length = VB->Primitive[i].count; + + if (!length) + continue; + + nr_prims++; + nr_rverts += length * scale_prim[prim & PRIM_MODE_MASK]; + + if (reduced_prim[prim & PRIM_MODE_MASK] != rprim) { + nr_rprims++; + rprim = reduced_prim[prim & PRIM_MODE_MASK]; + } + } + + /* One point for each generated primitive: + */ + cost_render = nr_prims; + cost_fallback = nr_rprims; + + /* One point for every 1024 dwords (4k) of dma: + */ + cost_render += (vertsz * i) / 1024; + cost_fallback += (vertsz * nr_rverts) / 1024; + + if (0) + fprintf(stderr, "cost render: %d fallback: %d\n", + cost_render, cost_fallback); + + if (cost_render > cost_fallback) + return GL_FALSE; + + return GL_TRUE; +} + + +static GLboolean +intel_run_render(GLcontext * ctx, struct tnl_pipeline_stage *stage) +{ + struct intel_context *intel = intel_context(ctx); + TNLcontext *tnl = TNL_CONTEXT(ctx); + struct vertex_buffer *VB = &tnl->vb; + GLuint i; + + intel->vtbl.render_prevalidate( intel ); + + /* Don't handle clipping or indexed vertices. + */ + if (intel->RenderIndex != 0 || + !intel_validate_render(ctx, VB) || !choose_render(intel, VB)) { + return GL_TRUE; + } + + tnl->clipspace.new_inputs |= VERT_BIT_POS; + + tnl->Driver.Render.Start(ctx); + + for (i = 0; i < VB->PrimitiveCount; i++) { + GLuint prim = VB->Primitive[i].mode; + GLuint start = VB->Primitive[i].start; + GLuint length = VB->Primitive[i].count; + + if (!length) + continue; + + intel_render_tab_verts[prim & PRIM_MODE_MASK] (ctx, start, + start + length, prim); + } + + tnl->Driver.Render.Finish(ctx); + + INTEL_FIREVERTICES(intel); + + return GL_FALSE; /* finished the pipe */ +} + +const struct tnl_pipeline_stage _intel_render_stage = { + "intel render", + NULL, + NULL, + NULL, + NULL, + intel_run_render /* run */ +}; diff --git a/src/mesa/drivers/dri/i915pipe/intel_rotate.c b/src/mesa/drivers/dri/i915pipe/intel_rotate.c new file mode 100644 index 0000000000..12d98c4ad2 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_rotate.c @@ -0,0 +1,237 @@ + +/** + * Routines for simple 2D->2D transformations for rotated, flipped screens. + * + * XXX This code is not intel-specific. Move it into a common/utility + * someday. + */ + +#include "intel_rotate.h" + +#define MIN2(A, B) ( ((A) < (B)) ? (A) : (B) ) + +#define ABS(A) ( ((A) < 0) ? -(A) : (A) ) + + +void +matrix23Set(struct matrix23 *m, + int m00, int m01, int m02, int m10, int m11, int m12) +{ + m->m00 = m00; + m->m01 = m01; + m->m02 = m02; + m->m10 = m10; + m->m11 = m11; + m->m12 = m12; +} + + +/* + * Transform (x,y) coordinate by the given matrix. + */ +void +matrix23TransformCoordf(const struct matrix23 *m, float *x, float *y) +{ + const float x0 = *x; + const float y0 = *y; + + *x = m->m00 * x0 + m->m01 * y0 + m->m02; + *y = m->m10 * x0 + m->m11 * y0 + m->m12; +} + + +void +matrix23TransformCoordi(const struct matrix23 *m, int *x, int *y) +{ + const int x0 = *x; + const int y0 = *y; + + *x = m->m00 * x0 + m->m01 * y0 + m->m02; + *y = m->m10 * x0 + m->m11 * y0 + m->m12; +} + + +/* + * Transform a width and height by the given matrix. + * XXX this could be optimized quite a bit. + */ +void +matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist) +{ + int x0 = 0, y0 = 0; + int x1 = *xDist, y1 = 0; + int x2 = 0, y2 = *yDist; + matrix23TransformCoordi(m, &x0, &y0); + matrix23TransformCoordi(m, &x1, &y1); + matrix23TransformCoordi(m, &x2, &y2); + + *xDist = (x1 - x0) + (x2 - x0); + *yDist = (y1 - y0) + (y2 - y0); + + if (*xDist < 0) + *xDist = -*xDist; + if (*yDist < 0) + *yDist = -*yDist; +} + + +/** + * Transform the rect defined by (x, y, w, h) by m. + */ +void +matrix23TransformRect(const struct matrix23 *m, int *x, int *y, int *w, + int *h) +{ + int x0 = *x, y0 = *y; + int x1 = *x + *w, y1 = *y; + int x2 = *x + *w, y2 = *y + *h; + int x3 = *x, y3 = *y + *h; + matrix23TransformCoordi(m, &x0, &y0); + matrix23TransformCoordi(m, &x1, &y1); + matrix23TransformCoordi(m, &x2, &y2); + matrix23TransformCoordi(m, &x3, &y3); + *w = ABS(x1 - x0) + ABS(x2 - x1); + /**w = ABS(*w);*/ + *h = ABS(y1 - y0) + ABS(y2 - y1); + /**h = ABS(*h);*/ + *x = MIN2(x0, x1); + *x = MIN2(*x, x2); + *y = MIN2(y0, y1); + *y = MIN2(*y, y2); +} + + +/* + * Make rotation matrix for width X height screen. + */ +void +matrix23Rotate(struct matrix23 *m, int width, int height, int angle) +{ + switch (angle) { + case 0: + matrix23Set(m, 1, 0, 0, 0, 1, 0); + break; + case 90: + matrix23Set(m, 0, 1, 0, -1, 0, width); + break; + case 180: + matrix23Set(m, -1, 0, width, 0, -1, height); + break; + case 270: + matrix23Set(m, 0, -1, height, 1, 0, 0); + break; + default: + /*abort() */ ; + } +} + + +/* + * Make flip/reflection matrix for width X height screen. + */ +void +matrix23Flip(struct matrix23 *m, int width, int height, int xflip, int yflip) +{ + if (xflip) { + m->m00 = -1; + m->m01 = 0; + m->m02 = width - 1; + } + else { + m->m00 = 1; + m->m01 = 0; + m->m02 = 0; + } + if (yflip) { + m->m10 = 0; + m->m11 = -1; + m->m12 = height - 1; + } + else { + m->m10 = 0; + m->m11 = 1; + m->m12 = 0; + } +} + + +/* + * result = a * b + */ +void +matrix23Multiply(struct matrix23 *result, + const struct matrix23 *a, const struct matrix23 *b) +{ + result->m00 = a->m00 * b->m00 + a->m01 * b->m10; + result->m01 = a->m00 * b->m01 + a->m01 * b->m11; + result->m02 = a->m00 * b->m02 + a->m01 * b->m12 + a->m02; + + result->m10 = a->m10 * b->m00 + a->m11 * b->m10; + result->m11 = a->m10 * b->m01 + a->m11 * b->m11; + result->m12 = a->m10 * b->m02 + a->m11 * b->m12 + a->m12; +} + + +#if 000 + +#include + +int +main(int argc, char *argv[]) +{ + int width = 500, height = 400; + int rot; + int fx = 0, fy = 0; /* flip x and/or y ? */ + int coords[4][2]; + + /* four corner coords to test with */ + coords[0][0] = 0; + coords[0][1] = 0; + coords[1][0] = width - 1; + coords[1][1] = 0; + coords[2][0] = width - 1; + coords[2][1] = height - 1; + coords[3][0] = 0; + coords[3][1] = height - 1; + + + for (rot = 0; rot < 360; rot += 90) { + struct matrix23 rotate, flip, m; + int i; + + printf("Rot %d, xFlip %d, yFlip %d:\n", rot, fx, fy); + + /* make transformation matrix 'm' */ + matrix23Rotate(&rotate, width, height, rot); + matrix23Flip(&flip, width, height, fx, fy); + matrix23Multiply(&m, &rotate, &flip); + + /* xform four coords */ + for (i = 0; i < 4; i++) { + int x = coords[i][0]; + int y = coords[i][1]; + matrix23TransformCoordi(&m, &x, &y); + printf(" %d, %d -> %d %d\n", coords[i][0], coords[i][1], x, y); + } + + /* xform width, height */ + { + int x = width; + int y = height; + matrix23TransformDistance(&m, &x, &y); + printf(" %d x %d -> %d x %d\n", width, height, x, y); + } + + /* xform rect */ + { + int x = 50, y = 10, w = 200, h = 100; + matrix23TransformRect(&m, &x, &y, &w, &h); + printf(" %d,%d %d x %d -> %d, %d %d x %d\n", 50, 10, 200, 100, + x, y, w, h); + } + + } + + return 0; +} +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_rotate.h b/src/mesa/drivers/dri/i915pipe/intel_rotate.h new file mode 100644 index 0000000000..9c8802ca47 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_rotate.h @@ -0,0 +1,39 @@ +#ifndef INTEL_ROTATE_H +#define INTEL_ROTATE_H 1 + +struct matrix23 +{ + int m00, m01, m02; + int m10, m11, m12; +}; + + + +extern void +matrix23Set(struct matrix23 *m, + int m00, int m01, int m02, int m10, int m11, int m12); + +extern void matrix23TransformCoordi(const struct matrix23 *m, int *x, int *y); + +extern void +matrix23TransformCoordf(const struct matrix23 *m, float *x, float *y); + +extern void +matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist); + +extern void +matrix23TransformRect(const struct matrix23 *m, + int *x, int *y, int *w, int *h); + +extern void +matrix23Rotate(struct matrix23 *m, int width, int height, int angle); + +extern void +matrix23Flip(struct matrix23 *m, int width, int height, int xflip, int yflip); + +extern void +matrix23Multiply(struct matrix23 *result, + const struct matrix23 *a, const struct matrix23 *b); + + +#endif /* INTEL_ROTATE_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.c b/src/mesa/drivers/dri/i915pipe/intel_screen.c new file mode 100644 index 0000000000..8d8b2523f3 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.c @@ -0,0 +1,708 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "context.h" +#include "framebuffer.h" +#include "matrix.h" +#include "renderbuffer.h" +#include "simple_list.h" +#include "utils.h" +#include "vblank.h" +#include "xmlpool.h" + + +#include "intel_screen.h" + +#include "intel_buffers.h" +#include "intel_tex.h" +#include "intel_span.h" +#include "intel_tris.h" +#include "intel_ioctl.h" +#include "intel_fbo.h" + +#include "i830_dri.h" +#include "dri_bufpool.h" +#include "intel_regions.h" +#include "intel_batchbuffer.h" + +PUBLIC const char __driConfigOptions[] = + DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE + DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) + DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) + DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY + DRI_CONF_FORCE_S3TC_ENABLE(false) + DRI_CONF_ALLOW_LARGE_TEXTURES(1) + DRI_CONF_SECTION_END DRI_CONF_END; + const GLuint __driNConfigOptions = 4; + +#ifdef USE_NEW_INTERFACE + static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; +#endif /*USE_NEW_INTERFACE */ + + extern const struct dri_extension card_extensions[]; + +/** + * Map all the memory regions described by the screen. + * \return GL_TRUE if success, GL_FALSE if error. + */ +GLboolean +intelMapScreenRegions(__DRIscreenPrivate * sPriv) +{ + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + + if (intelScreen->front.handle) { + if (drmMap(sPriv->fd, + intelScreen->front.handle, + intelScreen->front.size, + (drmAddress *) & intelScreen->front.map) != 0) { + _mesa_problem(NULL, "drmMap(frontbuffer) failed!"); + return GL_FALSE; + } + } + else { + _mesa_warning(NULL, "no front buffer handle in intelMapScreenRegions!"); + } + + if (0) + printf("Mappings: front: %p\n", intelScreen->front.map); + return GL_TRUE; +} + + +static struct intel_region * +intel_recreate_static(intelScreenPrivate *intelScreen, + struct intel_region *region, + GLuint mem_type, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, GLuint height) +{ + if (region) { + intel_region_update_static(intelScreen, region, mem_type, offset, + virtual, cpp, pitch, height); + } else { + region = intel_region_create_static(intelScreen, mem_type, offset, + virtual, cpp, pitch, height); + } + return region; +} + + +/* Create intel_region structs to describe the static front,back,depth + * buffers created by the xserver. + * Only used for real front buffer now. + * + * Note that these don't allocate video memory, just describe + * allocations already made by the X server. + */ +static void +intel_recreate_static_regions(intelScreenPrivate *intelScreen) +{ +/* this is the real front buffer which is only used for blitting to */ + intelScreen->front_region = + intel_recreate_static(intelScreen, + intelScreen->front_region, + DRM_BO_FLAG_MEM_TT, + intelScreen->front.offset, + intelScreen->front.map, + intelScreen->cpp, + intelScreen->front.pitch / intelScreen->cpp, + intelScreen->height); + +} + +/** + * Use the information in the sarea to update the screen parameters + * related to screen rotation. Needs to be called locked. + */ +void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) +{ + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + + intelUnmapScreenRegions(intelScreen); + intelUpdateScreenFromSAREA(intelScreen, sarea); + if (!intelMapScreenRegions(sPriv)) { + fprintf(stderr, "ERROR Remapping screen regions!!!\n"); + } + intel_recreate_static_regions(intelScreen); +} + + +void +intelUnmapScreenRegions(intelScreenPrivate * intelScreen) +{ +#define REALLY_UNMAP 1 + if (intelScreen->front.map) { +#if REALLY_UNMAP + if (drmUnmap(intelScreen->front.map, intelScreen->front.size) != 0) + printf("drmUnmap front failed!\n"); +#endif + intelScreen->front.map = NULL; + } +} + + +static void +intelPrintDRIInfo(intelScreenPrivate * intelScreen, + __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) +{ + fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", + intelScreen->front.size, intelScreen->front.offset, + intelScreen->front.pitch); + fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); +} + + +static void +intelPrintSAREA(const drmI830Sarea * sarea) +{ + fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, + sarea->height); + fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); + fprintf(stderr, + "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->front_offset, sarea->front_size, + (unsigned) sarea->front_handle); + fprintf(stderr, + "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->back_offset, sarea->back_size, + (unsigned) sarea->back_handle); + fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->depth_offset, sarea->depth_size, + (unsigned) sarea->depth_handle); + fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); + fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); + fprintf(stderr, + "SAREA: rotated offset: 0x%08x size: 0x%x\n", + sarea->rotated_offset, sarea->rotated_size); + fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); +} + + +/** + * A number of the screen parameters are obtained/computed from + * information in the SAREA. This function updates those parameters. + */ +void +intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, + drmI830Sarea * sarea) +{ + intelScreen->width = sarea->width; + intelScreen->height = sarea->height; + + intelScreen->front.offset = sarea->front_offset; + intelScreen->front.pitch = sarea->pitch * intelScreen->cpp; + intelScreen->front.handle = sarea->front_handle; + intelScreen->front.size = sarea->front_size; + + if (0) + intelPrintSAREA(sarea); +} + +GLboolean +intelCreatePools(intelScreenPrivate *intelScreen) +{ + unsigned batchPoolSize = 1024*1024; + __DRIscreenPrivate * sPriv = intelScreen->driScrnPriv; + + if (intelScreen->havePools) + return GL_TRUE; + + batchPoolSize /= intelScreen->maxBatchSize; + intelScreen->regionPool = driDRMPoolInit(sPriv->fd); + + if (!intelScreen->regionPool) + return GL_FALSE; + + intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); + + if (!intelScreen->staticPool) + return GL_FALSE; + + intelScreen->texPool = intelScreen->regionPool; + + intelScreen->batchPool = driBatchPoolInit(sPriv->fd, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT | + DRM_BO_FLAG_MEM_LOCAL, + intelScreen->maxBatchSize, + batchPoolSize, 5); + if (!intelScreen->batchPool) { + fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); + return GL_FALSE; + } + + intel_recreate_static_regions(intelScreen); + intelScreen->havePools = GL_TRUE; + + return GL_TRUE; +} + + +static GLboolean +intelInitDriver(__DRIscreenPrivate * sPriv) +{ + intelScreenPrivate *intelScreen; + I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; + drmI830Sarea *sarea; + + PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = + (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> + getProcAddress("glxEnableExtension")); + void *const psc = sPriv->psc->screenConfigs; + + if (sPriv->devPrivSize != sizeof(I830DRIRec)) { + fprintf(stderr, + "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); + return GL_FALSE; + } + + /* Allocate the private area */ + intelScreen = (intelScreenPrivate *) CALLOC(sizeof(intelScreenPrivate)); + if (!intelScreen) { + fprintf(stderr, "\nERROR! Allocating private area failed\n"); + return GL_FALSE; + } + /* parse information in __driConfigOptions */ + driParseOptionInfo(&intelScreen->optionCache, + __driConfigOptions, __driNConfigOptions); + + intelScreen->driScrnPriv = sPriv; + sPriv->private = (void *) intelScreen; + intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset; + sarea = (drmI830Sarea *) + (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset); + + intelScreen->maxBatchSize = BATCH_SZ; + intelScreen->deviceID = gDRIPriv->deviceID; + if (intelScreen->deviceID == PCI_CHIP_I865_G) + intelScreen->maxBatchSize = 4096; + + intelScreen->mem = gDRIPriv->mem; + intelScreen->cpp = gDRIPriv->cpp; + + switch (gDRIPriv->bitsPerPixel) { + case 16: + break; + case 32: + break; + default: + exit(1); + break; + } + + intelUpdateScreenFromSAREA(intelScreen, sarea); + + if (!intelMapScreenRegions(sPriv)) { + fprintf(stderr, "\nERROR! mapping regions\n"); + _mesa_free(intelScreen); + sPriv->private = NULL; + return GL_FALSE; + } + + + intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset; + + if (0) + intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); + + intelScreen->drmMinor = sPriv->drmMinor; + + /* Determine if IRQs are active? */ + { + int ret; + drmI830GetParam gp; + + gp.param = I830_PARAM_IRQ_ACTIVE; + gp.value = &intelScreen->irq_active; + + ret = drmCommandWriteRead(sPriv->fd, DRM_I830_GETPARAM, + &gp, sizeof(gp)); + if (ret) { + fprintf(stderr, "drmI830GetParam: %d\n", ret); + return GL_FALSE; + } + } + + /* Determine if batchbuffers are allowed */ + { + int ret; + drmI830GetParam gp; + + gp.param = I830_PARAM_ALLOW_BATCHBUFFER; + gp.value = &intelScreen->allow_batchbuffer; + + ret = drmCommandWriteRead(sPriv->fd, DRM_I830_GETPARAM, + &gp, sizeof(gp)); + if (ret) { + fprintf(stderr, "drmI830GetParam: (%d) %d\n", gp.param, ret); + return GL_FALSE; + } + } + + if (glx_enable_extension != NULL) { + (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); + (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); + (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); + } + + return GL_TRUE; +} + + +static void +intelDestroyScreen(__DRIscreenPrivate * sPriv) +{ + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + + intelUnmapScreenRegions(intelScreen); + + if (intelScreen->havePools) { + driPoolTakeDown(intelScreen->regionPool); + driPoolTakeDown(intelScreen->staticPool); + driPoolTakeDown(intelScreen->batchPool); + } + FREE(intelScreen); + sPriv->private = NULL; +} + + +/** + * This is called when we need to set up GL rendering to a new X window. + */ +static GLboolean +intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, + __DRIdrawablePrivate * driDrawPriv, + const __GLcontextModes * mesaVis, GLboolean isPixmap) +{ +#if 0 + intelScreenPrivate *screen = (intelScreenPrivate *) driScrnPriv->private; +#endif + + if (isPixmap) { + return GL_FALSE; /* not implemented */ + } + else { + GLboolean swStencil = (mesaVis->stencilBits > 0 && + mesaVis->depthBits != 24); + GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); + + struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); + + if (!intel_fb) + return GL_FALSE; + + _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); + + { + /* fake frontbuffer */ + /* XXX allocation should only happen in the unusual case + it's actually needed */ + intel_fb->color_rb[0] + = intel_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, + &intel_fb->color_rb[0]->Base); + } + + if (mesaVis->doubleBufferMode) { + intel_fb->color_rb[1] + = intel_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, + &intel_fb->color_rb[1]->Base); + +#if 0 + if (screen->third.handle) { + struct gl_renderbuffer *tmp_rb = NULL; + + intel_fb->color_rb[2] + = intel_new_renderbuffer_fb(rgbFormat); + _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base); + } +#endif + } + + if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { + /* combined depth/stencil buffer */ + struct intel_renderbuffer *depthStencilRb + = intel_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + /* note: bind RB to two attachment points */ + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, + &depthStencilRb->Base); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL, + &depthStencilRb->Base); + } + else if (mesaVis->depthBits == 16) { + /* just 16-bit depth buffer, no hw stencil */ + struct intel_renderbuffer *depthRb + = intel_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); + } + + + /* now add any/all software-based renderbuffers we may need */ + _mesa_add_soft_renderbuffers(&intel_fb->Base, + GL_FALSE, /* never sw color */ + GL_FALSE, /* never sw depth */ + swStencil, mesaVis->accumRedBits > 0, + GL_FALSE, /* never sw alpha */ + GL_FALSE /* never sw aux */ ); + driDrawPriv->driverPrivate = (void *) intel_fb; + + return GL_TRUE; + } +} + +static void +intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) +{ + _mesa_unreference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate))); +} + + +/** + * Get information about previous buffer swaps. + */ +static int +intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) +{ + struct intel_framebuffer *intel_fb; + + if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) + || (sInfo == NULL)) { + return -1; + } + + intel_fb = dPriv->driverPrivate; + sInfo->swap_count = intel_fb->swap_count; + sInfo->swap_ust = intel_fb->swap_ust; + sInfo->swap_missed_count = intel_fb->swap_missed_count; + + sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0) + ? driCalculateSwapUsage(dPriv, 0, intel_fb->swap_missed_ust) + : 0.0; + + return 0; +} + + + + +static const struct __DriverAPIRec intelAPI = { + .InitDriver = intelInitDriver, + .DestroyScreen = intelDestroyScreen, + .CreateContext = intelCreateContext, + .DestroyContext = intelDestroyContext, + .CreateBuffer = intelCreateBuffer, + .DestroyBuffer = intelDestroyBuffer, + .SwapBuffers = intelSwapBuffers, + .MakeCurrent = intelMakeCurrent, + .UnbindContext = intelUnbindContext, + .GetSwapInfo = intelGetSwapInfo, + .GetMSC = driGetMSC32, + .WaitForMSC = driWaitForMSC32, + .WaitForSBC = NULL, + .SwapBuffersMSC = NULL, + .CopySubBuffer = intelCopySubBuffer, + .setTexOffset = intelSetTexOffset, +}; + + +static __GLcontextModes * +intelFillInModes(unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, GLboolean have_back_buffer) +{ + __GLcontextModes *modes; + __GLcontextModes *m; + unsigned num_modes; + unsigned depth_buffer_factor; + unsigned back_buffer_factor; + GLenum fb_format; + GLenum fb_type; + + /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't + * support pageflipping at all. + */ + static const GLenum back_buffer_modes[] = { + GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML + }; + + u_int8_t depth_bits_array[3]; + u_int8_t stencil_bits_array[3]; + + + depth_bits_array[0] = 0; + depth_bits_array[1] = depth_bits; + depth_bits_array[2] = depth_bits; + + /* Just like with the accumulation buffer, always provide some modes + * with a stencil buffer. It will be a sw fallback, but some apps won't + * care about that. + */ + stencil_bits_array[0] = 0; + stencil_bits_array[1] = 0; + if (depth_bits == 24) + stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; + + stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; + + depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; + back_buffer_factor = (have_back_buffer) ? 3 : 1; + + num_modes = depth_buffer_factor * back_buffer_factor * 4; + + if (pixel_bits == 16) { + fb_format = GL_RGB; + fb_type = GL_UNSIGNED_SHORT_5_6_5; + } + else { + fb_format = GL_BGRA; + fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; + } + + modes = + (*dri_interface->createContextModes) (num_modes, + sizeof(__GLcontextModes)); + m = modes; + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_TRUE_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_DIRECT_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + + /* Mark the visual as slow if there are "fake" stencil bits. + */ + for (m = modes; m != NULL; m = m->next) { + if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { + m->visualRating = GLX_SLOW_CONFIG; + } + } + + return modes; +} + + +/** + * This is the bootstrap function for the driver. libGL supplies all of the + * requisite information about the system, and the driver initializes itself. + * This routine also fills in the linked list pointed to by \c driver_modes + * with the \c __GLcontextModes that the driver can support for windows or + * pbuffers. + * + * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on + * failure. + */ +PUBLIC void * +__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, + __DRIscreen * psc, + const __GLcontextModes * modes, + const __DRIversion * ddx_version, + const __DRIversion * dri_version, + const __DRIversion * drm_version, + const __DRIframebuffer * frame_buffer, + drmAddress pSAREA, int fd, + int internal_api_version, + const __DRIinterfaceMethods * interface, + __GLcontextModes ** driver_modes) +{ + __DRIscreenPrivate *psp; + static const __DRIversion ddx_expected = { 1, 7, 0 }; + static const __DRIversion dri_expected = { 4, 0, 0 }; + static const __DRIversion drm_expected = { 1, 7, 0 }; + + dri_interface = interface; + + if (!driCheckDriDdxDrmVersions2("i915", + dri_version, &dri_expected, + ddx_version, &ddx_expected, + drm_version, &drm_expected)) { + return NULL; + } + + psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, + ddx_version, dri_version, drm_version, + frame_buffer, pSAREA, fd, + internal_api_version, &intelAPI); + + if (psp != NULL) { + I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; + *driver_modes = intelFillInModes(dri_priv->cpp * 8, + (dri_priv->cpp == 2) ? 16 : 24, + (dri_priv->cpp == 2) ? 0 : 8, 1); + + /* Calling driInitExtensions here, with a NULL context pointer, does not actually + * enable the extensions. It just makes sure that all the dispatch offsets for all + * the extensions that *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create is called, but we can't + * enable the extensions until we have a context pointer. + * + * Hello chicken. Hello egg. How are you two today? + */ + driInitExtensions(NULL, card_extensions, GL_FALSE); + } + + return (void *) psp; +} + +struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) +{ + /* + * This should probably change to have the screen allocate a dummy + * context at screen creation. For now just use the current context. + */ + + GET_CURRENT_CONTEXT(ctx); + if (ctx == NULL) { +/* _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; */ + /* need a context for the first time makecurrent is called (for hw lock + when allocating priv buffers) */ + if (intelScreen->dummyctxptr == NULL) { + _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; + } + return intelScreen->dummyctxptr; + } + return intel_context(ctx); + +} + diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.h b/src/mesa/drivers/dri/i915pipe/intel_screen.h new file mode 100644 index 0000000000..eb4685d6e9 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.h @@ -0,0 +1,129 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _INTEL_INIT_H_ +#define _INTEL_INIT_H_ + +#include +#include "dri_util.h" +#include "intel_rotate.h" +#include "i830_common.h" +#include "xmlconfig.h" +#include "dri_bufpool.h" + +/* XXX: change name or eliminate to avoid conflict with "struct + * intel_region"!!! + */ +typedef struct +{ + drm_handle_t handle; + drmSize size; /* region size in bytes */ + char *map; /* memory map */ + int offset; /* from start of video mem, in bytes */ + int pitch; /* row stride, in bytes */ +} intelRegion; + +typedef struct +{ + intelRegion front; + + struct intel_region *front_region; + + int deviceID; + int width; + int height; + int mem; /* unused */ + + int cpp; /* for front and back buffers */ + + __DRIscreenPrivate *driScrnPriv; + unsigned int sarea_priv_offset; + + int drmMinor; + + int irq_active; + int allow_batchbuffer; + + struct matrix23 rotMatrix; + + + /** + * Configuration cache with default values for all contexts + */ + driOptionCache optionCache; + struct _DriBufferPool *batchPool; + struct _DriBufferPool *texPool; + struct _DriBufferPool *regionPool; + struct _DriBufferPool *staticPool; + unsigned int maxBatchSize; + GLboolean havePools; + struct intel_context *dummyctxptr; +} intelScreenPrivate; + + + +extern GLboolean intelMapScreenRegions(__DRIscreenPrivate * sPriv); + +extern void intelUnmapScreenRegions(intelScreenPrivate * intelScreen); + +extern void +intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, + drmI830Sarea * sarea); + +extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); + +extern GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); + +extern GLboolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); + +extern void +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h); + +extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, + unsigned long bufSize, + unsigned numBufs, + unsigned checkDelayed); + +extern struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen); + +extern void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); +extern GLboolean +intelCreatePools(intelScreenPrivate *intelScreen); + +extern GLboolean +intelCreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); + + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_span.c b/src/mesa/drivers/dri/i915pipe/intel_span.c new file mode 100644 index 0000000000..5a978d9ce2 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_span.c @@ -0,0 +1,401 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "macros.h" +#include "mtypes.h" +#include "colormac.h" + +#include "intel_fbo.h" +#include "intel_screen.h" +#include "intel_span.h" +#include "intel_regions.h" +#include "intel_ioctl.h" +#include "intel_tex.h" + +#include "swrast/swrast.h" + +/* + break intelWriteRGBASpan_ARGB8888 +*/ + +#undef DBG +#define DBG 0 + +#define LOCAL_VARS \ + struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ + const GLint yScale = irb->RenderToTexture ? 1 : -1; \ + const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ + GLubyte *buf = (GLubyte *) irb->pfMap; \ + GLuint p; \ + assert(irb->pfMap);\ + (void) p; + +/* There is just a single cliploop! + */ +#define HW_CLIPLOOP() \ + do { \ + int minx = 0; \ + int miny = 0; \ + int maxx = irb->Base.Width - 1; \ + int maxy = irb->Base.Height - 1; + +#define HW_ENDCLIPLOOP() \ + } while (0) + + +#define Y_FLIP(_y) ((_y) * yScale + yBias) + +#define HW_LOCK() + +#define HW_UNLOCK() + +/* 16 bit, RGB565 color spanline and pixel functions + */ +#define SPANTMP_PIXEL_FMT GL_RGB +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5 + +#define TAG(x) intel##x##_RGB565 +#define TAG2(x,y) intel##x##_RGB565##y +#define GET_PTR(X,Y) (buf + ((Y) * irb->pfPitch + (X)) * 2) +#include "spantmp2.h" + +/* 32 bit, ARGB8888 color spanline and pixel functions + */ +#define SPANTMP_PIXEL_FMT GL_BGRA +#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV + +#define TAG(x) intel##x##_ARGB8888 +#define TAG2(x,y) intel##x##_ARGB8888##y +#define GET_PTR(X,Y) (buf + ((Y) * irb->pfPitch + (X)) * 4) +#include "spantmp2.h" + + +#define LOCAL_DEPTH_VARS \ + struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ + const GLuint pitch = irb->pfPitch/***XXX region->pitch*/; /* in pixels */ \ + const GLint yScale = irb->RenderToTexture ? 1 : -1; \ + const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ + char *buf = (char *) irb->pfMap/*XXX use region->map*/ ; + +#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS + +/** + ** 16-bit depthbuffer functions. + **/ +#define WRITE_DEPTH( _x, _y, d ) \ + ((GLushort *)buf)[(_x) + (_y) * pitch] = d; + +#define READ_DEPTH( d, _x, _y ) \ + d = ((GLushort *)buf)[(_x) + (_y) * pitch]; + + +#define TAG(x) intel##x##_z16 +#include "depthtmp.h" + + +/** + ** 24/8-bit interleaved depth/stencil functions + ** Note: we're actually reading back combined depth+stencil values. + ** The wrappers in main/depthstencil.c are used to extract the depth + ** and stencil values. + **/ +/* Change ZZZS -> SZZZ */ +#define WRITE_DEPTH( _x, _y, d ) { \ + GLuint tmp = ((d) >> 8) | ((d) << 24); \ + ((GLuint *)buf)[(_x) + (_y) * pitch] = tmp; \ +} + +/* Change SZZZ -> ZZZS */ +#define READ_DEPTH( d, _x, _y ) { \ + GLuint tmp = ((GLuint *)buf)[(_x) + (_y) * pitch]; \ + d = (tmp << 8) | (tmp >> 24); \ +} + +#define TAG(x) intel##x##_z24_s8 +#include "depthtmp.h" + + +/** + ** 8-bit stencil function (XXX FBO: This is obsolete) + **/ +#define WRITE_STENCIL( _x, _y, d ) { \ + GLuint tmp = ((GLuint *)buf)[(_x) + (_y) * pitch]; \ + tmp &= 0xffffff; \ + tmp |= ((d) << 24); \ + ((GLuint *) buf)[(_x) + (_y) * pitch] = tmp; \ +} + +#define READ_STENCIL( d, _x, _y ) \ + d = ((GLuint *)buf)[(_x) + (_y) * pitch] >> 24; + +#define TAG(x) intel##x##_z24_s8 +#include "stenciltmp.h" + + + +/** + * Map or unmap all the renderbuffers which we may need during + * software rendering. + * XXX in the future, we could probably convey extra information to + * reduce the number of mappings needed. I.e. if doing a glReadPixels + * from the depth buffer, we really only need one mapping. + * + * XXX Rewrite this function someday. + * We can probably just loop over all the renderbuffer attachments, + * map/unmap all of them, and not worry about the _ColorDrawBuffers + * _ColorReadBuffer, _DepthBuffer or _StencilBuffer fields. + */ +static void +intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) +{ + GLcontext *ctx = &intel->ctx; + GLuint i, j; + struct intel_renderbuffer *irb; + + /* color draw buffers */ + for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { + for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers[i]; j++) { + struct gl_renderbuffer *rb = + ctx->DrawBuffer->_ColorDrawBuffers[i][j]; + irb = intel_renderbuffer(rb); + if (irb) { + /* this is a user-created intel_renderbuffer */ + if (irb->region) { + if (map) + intel_region_map(intel->intelScreen, irb->region); + else + intel_region_unmap(intel->intelScreen, irb->region); + } + irb->pfMap = irb->region->map; + irb->pfPitch = irb->region->pitch; + } + } + } + + /* check for render to textures */ + for (i = 0; i < BUFFER_COUNT; i++) { + struct gl_renderbuffer_attachment *att = + ctx->DrawBuffer->Attachment + i; + struct gl_texture_object *tex = att->Texture; + if (tex) { + /* render to texture */ + ASSERT(att->Renderbuffer); + if (map) { + struct gl_texture_image *texImg; + texImg = tex->Image[att->CubeMapFace][att->TextureLevel]; + intel_tex_map_images(intel, intel_texture_object(tex)); + } + else { + intel_tex_unmap_images(intel, intel_texture_object(tex)); + } + } + } + + /* color read buffers */ + irb = intel_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + if (irb && irb->region) { + if (map) + intel_region_map(intel->intelScreen, irb->region); + else + intel_region_unmap(intel->intelScreen, irb->region); + irb->pfMap = irb->region->map; + irb->pfPitch = irb->region->pitch; + } + + /* Account for front/back color page flipping. + * The span routines use the pfMap and pfPitch fields which will + * swap the front/back region map/pitch if we're page flipped. + * Do this after mapping, above, so the map field is valid. + */ +#if 0 + if (map && ctx->DrawBuffer->Name == 0) { + struct intel_renderbuffer *irbFront + = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_FRONT_LEFT); + struct intel_renderbuffer *irbBack + = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_BACK_LEFT); + if (irbBack) { + /* double buffered */ + if (intel->sarea->pf_current_page == 0) { + irbFront->pfMap = irbFront->region->map; + irbFront->pfPitch = irbFront->region->pitch; + irbBack->pfMap = irbBack->region->map; + irbBack->pfPitch = irbBack->region->pitch; + } + else { + irbFront->pfMap = irbBack->region->map; + irbFront->pfPitch = irbBack->region->pitch; + irbBack->pfMap = irbFront->region->map; + irbBack->pfPitch = irbFront->region->pitch; + } + } + } +#endif + + /* depth buffer (Note wrapper!) */ + if (ctx->DrawBuffer->_DepthBuffer) { + irb = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); + if (irb && irb->region) { + if (map) { + intel_region_map(intel->intelScreen, irb->region); + irb->pfMap = irb->region->map; + irb->pfPitch = irb->region->pitch; + } + else { + intel_region_unmap(intel->intelScreen, irb->region); + irb->pfMap = NULL; + irb->pfPitch = 0; + } + } + } + + /* stencil buffer (Note wrapper!) */ + if (ctx->DrawBuffer->_StencilBuffer) { + irb = intel_renderbuffer(ctx->DrawBuffer->_StencilBuffer->Wrapped); + if (irb && irb->region) { + if (map) { + intel_region_map(intel->intelScreen, irb->region); + irb->pfMap = irb->region->map; + irb->pfPitch = irb->region->pitch; + } + else { + intel_region_unmap(intel->intelScreen, irb->region); + irb->pfMap = NULL; + irb->pfPitch = 0; + } + } + } +} + + + +/** + * Prepare for softare rendering. Map current read/draw framebuffers' + * renderbuffes and all currently bound texture objects. + * + * Old note: Moved locking out to get reasonable span performance. + */ +void +intelSpanRenderStart(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + GLuint i; + + intelFinish(&intel->ctx); + LOCK_HARDWARE(intel); + +#if 0 + /* Just map the framebuffer and all textures. Bufmgr code will + * take care of waiting on the necessary fences: + */ + intel_region_map(intel->intelScreen, intel->front_region); + intel_region_map(intel->intelScreen, intel->back_region); + intel_region_map(intel->intelScreen, intel->intelScreen->depth_region); +#endif + + for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { + struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; + intel_tex_map_images(intel, intel_texture_object(texObj)); + } + } + + intel_map_unmap_buffers(intel, GL_TRUE); +} + +/** + * Called when done softare rendering. Unmap the buffers we mapped in + * the above function. + */ +void +intelSpanRenderFinish(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + GLuint i; + + _swrast_flush(ctx); + + /* Now unmap the framebuffer: + */ +#if 0 + intel_region_unmap(intel, intel->front_region); + intel_region_unmap(intel, intel->back_region); + intel_region_unmap(intel, intel->intelScreen->depth_region); +#endif + + for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { + if (ctx->Texture.Unit[i]._ReallyEnabled) { + struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; + intel_tex_unmap_images(intel, intel_texture_object(texObj)); + } + } + + intel_map_unmap_buffers(intel, GL_FALSE); + + UNLOCK_HARDWARE(intel); +} + + +void +intelInitSpanFuncs(GLcontext * ctx) +{ + struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx); + swdd->SpanRenderStart = intelSpanRenderStart; + swdd->SpanRenderFinish = intelSpanRenderFinish; +} + + +/** + * Plug in appropriate span read/write functions for the given renderbuffer. + * These are used for the software fallbacks. + */ +void +intel_set_span_functions(struct gl_renderbuffer *rb) +{ + if (rb->_ActualFormat == GL_RGB5) { + /* 565 RGB */ + intelInitPointers_RGB565(rb); + } + else if (rb->_ActualFormat == GL_RGBA8) { + /* 8888 RGBA */ + intelInitPointers_ARGB8888(rb); + } + else if (rb->_ActualFormat == GL_DEPTH_COMPONENT16) { + intelInitDepthPointers_z16(rb); + } + else if (rb->_ActualFormat == GL_DEPTH_COMPONENT24 || /* XXX FBO remove */ + rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) { + intelInitDepthPointers_z24_s8(rb); + } + else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) { /* XXX FBO remove */ + intelInitStencilPointers_z24_s8(rb); + } + else { + _mesa_problem(NULL, + "Unexpected _ActualFormat in intelSetSpanFunctions"); + } +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_span.h b/src/mesa/drivers/dri/i915pipe/intel_span.h new file mode 100644 index 0000000000..5201f6d6c6 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_span.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _INTEL_SPAN_H +#define _INTEL_SPAN_H + +extern void intelInitSpanFuncs(GLcontext * ctx); + +extern void intelSpanRenderFinish(GLcontext * ctx); +extern void intelSpanRenderStart(GLcontext * ctx); + +extern void intel_set_span_functions(struct gl_renderbuffer *rb); + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_state.c b/src/mesa/drivers/dri/i915pipe/intel_state.c new file mode 100644 index 0000000000..5c5f2c6de5 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_state.c @@ -0,0 +1,261 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "glheader.h" +#include "context.h" +#include "macros.h" +#include "enums.h" +#include "colormac.h" +#include "dd.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_fbo.h" +#include "intel_regions.h" +#include "swrast/swrast.h" + +int +intel_translate_compare_func(GLenum func) +{ + switch (func) { + case GL_NEVER: + return COMPAREFUNC_NEVER; + case GL_LESS: + return COMPAREFUNC_LESS; + case GL_LEQUAL: + return COMPAREFUNC_LEQUAL; + case GL_GREATER: + return COMPAREFUNC_GREATER; + case GL_GEQUAL: + return COMPAREFUNC_GEQUAL; + case GL_NOTEQUAL: + return COMPAREFUNC_NOTEQUAL; + case GL_EQUAL: + return COMPAREFUNC_EQUAL; + case GL_ALWAYS: + return COMPAREFUNC_ALWAYS; + } + + fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); + return COMPAREFUNC_ALWAYS; +} + +int +intel_translate_stencil_op(GLenum op) +{ + switch (op) { + case GL_KEEP: + return STENCILOP_KEEP; + case GL_ZERO: + return STENCILOP_ZERO; + case GL_REPLACE: + return STENCILOP_REPLACE; + case GL_INCR: + return STENCILOP_INCRSAT; + case GL_DECR: + return STENCILOP_DECRSAT; + case GL_INCR_WRAP: + return STENCILOP_INCR; + case GL_DECR_WRAP: + return STENCILOP_DECR; + case GL_INVERT: + return STENCILOP_INVERT; + default: + return STENCILOP_ZERO; + } +} + +int +intel_translate_blend_factor(GLenum factor) +{ + switch (factor) { + case GL_ZERO: + return BLENDFACT_ZERO; + case GL_SRC_ALPHA: + return BLENDFACT_SRC_ALPHA; + case GL_ONE: + return BLENDFACT_ONE; + case GL_SRC_COLOR: + return BLENDFACT_SRC_COLR; + case GL_ONE_MINUS_SRC_COLOR: + return BLENDFACT_INV_SRC_COLR; + case GL_DST_COLOR: + return BLENDFACT_DST_COLR; + case GL_ONE_MINUS_DST_COLOR: + return BLENDFACT_INV_DST_COLR; + case GL_ONE_MINUS_SRC_ALPHA: + return BLENDFACT_INV_SRC_ALPHA; + case GL_DST_ALPHA: + return BLENDFACT_DST_ALPHA; + case GL_ONE_MINUS_DST_ALPHA: + return BLENDFACT_INV_DST_ALPHA; + case GL_SRC_ALPHA_SATURATE: + return BLENDFACT_SRC_ALPHA_SATURATE; + case GL_CONSTANT_COLOR: + return BLENDFACT_CONST_COLOR; + case GL_ONE_MINUS_CONSTANT_COLOR: + return BLENDFACT_INV_CONST_COLOR; + case GL_CONSTANT_ALPHA: + return BLENDFACT_CONST_ALPHA; + case GL_ONE_MINUS_CONSTANT_ALPHA: + return BLENDFACT_INV_CONST_ALPHA; + } + + fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, factor); + return BLENDFACT_ZERO; +} + +int +intel_translate_logic_op(GLenum opcode) +{ + switch (opcode) { + case GL_CLEAR: + return LOGICOP_CLEAR; + case GL_AND: + return LOGICOP_AND; + case GL_AND_REVERSE: + return LOGICOP_AND_RVRSE; + case GL_COPY: + return LOGICOP_COPY; + case GL_COPY_INVERTED: + return LOGICOP_COPY_INV; + case GL_AND_INVERTED: + return LOGICOP_AND_INV; + case GL_NOOP: + return LOGICOP_NOOP; + case GL_XOR: + return LOGICOP_XOR; + case GL_OR: + return LOGICOP_OR; + case GL_OR_INVERTED: + return LOGICOP_OR_INV; + case GL_NOR: + return LOGICOP_NOR; + case GL_EQUIV: + return LOGICOP_EQUIV; + case GL_INVERT: + return LOGICOP_INV; + case GL_OR_REVERSE: + return LOGICOP_OR_RVRSE; + case GL_NAND: + return LOGICOP_NAND; + case GL_SET: + return LOGICOP_SET; + default: + return LOGICOP_SET; + } +} + + +static void +intelClearColor(GLcontext * ctx, const GLfloat color[4]) +{ + struct intel_context *intel = intel_context(ctx); + GLubyte clear[4]; + + CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]); + CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]); + CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); + CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); + + /* compute both 32 and 16-bit clear values */ + intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1], + clear[2], clear[3]); + intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]); +} + + +/** + * Update the viewport transformation matrix. Depends on: + * - viewport pos/size + * - depthrange + * - window size or FBO size + */ +static void +intelCalcViewport(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + const GLfloat *v = ctx->Viewport._WindowMap.m; + const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF; + GLfloat *m = intel->ViewportMatrix.m; + GLfloat yScale, yBias; + + struct intel_renderbuffer *irb + = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); + if (irb && !irb->RenderToTexture) { + /* y=0=top */ + yScale = -1.0; + yBias = irb->Base.Height; + } + else { + /* y=0=bottom */ + yScale = 1.0; + yBias = 0.0; + } + + m[MAT_SX] = v[MAT_SX]; + m[MAT_TX] = v[MAT_TX] + SUBPIXEL_X; + + m[MAT_SY] = v[MAT_SY] * yScale; + m[MAT_TY] = v[MAT_TY] * yScale + yBias + SUBPIXEL_Y; + + m[MAT_SZ] = v[MAT_SZ] * depthScale; + m[MAT_TZ] = v[MAT_TZ] * depthScale; +} + +static void +intelViewport(GLcontext * ctx, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + intelCalcViewport(ctx); +} + +static void +intelDepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) +{ + intelCalcViewport(ctx); +} + +/* Fallback to swrast for select and feedback. + */ +static void +intelRenderMode(GLcontext * ctx, GLenum mode) +{ + struct intel_context *intel = intel_context(ctx); + FALLBACK(intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER)); +} + + +void +intelInitStateFuncs(struct dd_function_table *functions) +{ + functions->RenderMode = intelRenderMode; + functions->Viewport = intelViewport; + functions->DepthRange = intelDepthRange; + functions->ClearColor = intelClearColor; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_structs.h b/src/mesa/drivers/dri/i915pipe/intel_structs.h new file mode 100644 index 0000000000..522e3bd92c --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_structs.h @@ -0,0 +1,132 @@ +#ifndef INTEL_STRUCTS_H +#define INTEL_STRUCTS_H + +struct br0 { + GLuint length:8; + GLuint pad0:3; + GLuint dst_tiled:1; + GLuint pad1:8; + GLuint write_rgb:1; + GLuint write_alpha:1; + GLuint opcode:7; + GLuint client:3; +}; + + +struct br13 { + GLint dest_pitch:16; + GLuint rop:8; + GLuint color_depth:2; + GLuint pad1:3; + GLuint mono_source_transparency:1; + GLuint clipping_enable:1; + GLuint pad0:1; +}; + + + +/* This is an attempt to move some of the 2D interaction in this + * driver to using structs for packets rather than a bunch of #defines + * and dwords. + */ +struct xy_color_blit { + struct br0 br0; + struct br13 br13; + + struct { + GLuint dest_x1:16; + GLuint dest_y1:16; + } dw2; + + struct { + GLuint dest_x2:16; + GLuint dest_y2:16; + } dw3; + + GLuint dest_base_addr; + GLuint color; +}; + +struct xy_src_copy_blit { + struct br0 br0; + struct br13 br13; + + struct { + GLuint dest_x1:16; + GLuint dest_y1:16; + } dw2; + + struct { + GLuint dest_x2:16; + GLuint dest_y2:16; + } dw3; + + GLuint dest_base_addr; + + struct { + GLuint src_x1:16; + GLuint src_y1:16; + } dw5; + + struct { + GLint src_pitch:16; + GLuint pad:16; + } dw6; + + GLuint src_base_addr; +}; + +struct xy_setup_blit { + struct br0 br0; + struct br13 br13; + + struct { + GLuint clip_x1:16; + GLuint clip_y1:16; + } dw2; + + struct { + GLuint clip_x2:16; + GLuint clip_y2:16; + } dw3; + + GLuint dest_base_addr; + GLuint background_color; + GLuint foreground_color; + GLuint pattern_base_addr; +}; + + +struct xy_text_immediate_blit { + struct { + GLuint length:8; + GLuint pad2:3; + GLuint dst_tiled:1; + GLuint pad1:4; + GLuint byte_packed:1; + GLuint pad0:5; + GLuint opcode:7; + GLuint client:3; + } dw0; + + struct { + GLuint dest_x1:16; + GLuint dest_y1:16; + } dw1; + + struct { + GLuint dest_x2:16; + GLuint dest_y2:16; + } dw2; + + /* Src bitmap data follows as inline dwords. + */ +}; + + +#define CLIENT_2D 0x2 +#define OPCODE_XY_SETUP_BLT 0x1 +#define OPCODE_XY_COLOR_BLT 0x50 +#define OPCODE_XY_TEXT_IMMEDIATE_BLT 0x31 + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_surface.c b/src/mesa/drivers/dri/i915pipe/intel_surface.c new file mode 100644 index 0000000000..043c5aa5fe --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_surface.c @@ -0,0 +1,259 @@ +#include "glheader.h" +#include "context.h" +#include "framebuffer.h" +#include "renderbuffer.h" +#include "utils.h" +#include "main/macros.h" + + +#include "intel_screen.h" + +#include "intel_context.h" +#include "intel_buffers.h" +#include "intel_regions.h" +#include "intel_span.h" +#include "intel_fbo.h" + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/softpipe/sp_surface.h" + + +/* + * XXX a lof of this is a temporary kludge + */ + +/** + * Note: the arithmetic/addressing in these functions is a little + * tricky since we need to invert the Y axis. + */ + + +static void +read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) +{ + const GLint bytesPerRow = sps->surface.stride * sps->surface.cpp; + const GLint invY = sps->surface.height - y - 1; + const GLubyte *src = sps->surface.ptr + invY * bytesPerRow + x * sps->surface.cpp; + GLfloat *dst = (GLfloat *) rrrr; + GLubyte temp[16]; + GLuint j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); + + memcpy(temp + 8, src, 8); + memcpy(temp + 0, src + bytesPerRow, 8); + + for (j = 0; j < 4; j++) { + dst[0 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 2]); /*R*/ + dst[1 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 1]); /*G*/ + dst[2 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 0]); /*B*/ + dst[3 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 3]); /*A*/ + } +} + + +static void +write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) +{ + const GLfloat *src = (const GLfloat *) rrrr; + const GLint bytesPerRow = sps->surface.stride * sps->surface.cpp; + const GLint invY = sps->surface.height - y - 1; + GLubyte *dst = sps->surface.ptr + invY * bytesPerRow + x * sps->surface.cpp; + GLubyte temp[16]; + GLuint j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); + + for (j = 0; j < 4; j++) { + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 2], src[0 * 4 + j]); /*R*/ + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 1], src[1 * 4 + j]); /*G*/ + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 0], src[2 * 4 + j]); /*B*/ + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 3], src[3 * 4 + j]); /*A*/ + } + + memcpy(dst, temp + 8, 8); + memcpy(dst + bytesPerRow, temp + 0, 8); +} + + + +static void +read_quad_z24(struct softpipe_surface *sps, + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) +{ + static const GLuint mask = 0xffffff; + const GLint invY = sps->surface.height - y - 1; + const GLuint *src + = (GLuint *) (sps->surface.ptr + + (invY * sps->surface.stride + x) * sps->surface.cpp); + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + /* extract lower three bytes */ + zzzz[0] = src[0] & mask; + zzzz[1] = src[1] & mask; + zzzz[2] = src[-sps->surface.stride] & mask; + zzzz[3] = src[-sps->surface.stride + 1] & mask; +} + +static void +write_quad_z24(struct softpipe_surface *sps, + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) +{ + static const GLuint mask = 0xff000000; + const GLint invY = sps->surface.height - y - 1; + GLuint *dst + = (GLuint *) (sps->surface.ptr + + (invY * sps->surface.stride + x) * sps->surface.cpp); + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + /* write lower three bytes */ + dst[0] = (dst[0] & mask) | zzzz[0]; + dst[1] = (dst[1] & mask) | zzzz[1]; + dst -= sps->surface.stride; + dst[0] = (dst[0] & mask) | zzzz[2]; + dst[1] = (dst[1] & mask) | zzzz[3]; +} + + +static void +read_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) +{ + const GLint invY = sps->surface.height - y - 1; + const GLuint *src = (const GLuint *) (sps->surface.ptr + + (invY * sps->surface.stride + x) * sps->surface.cpp); + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + /* extract high byte */ + ssss[0] = src[0] >> 24; + ssss[1] = src[1] >> 24; + ssss[2] = src[-sps->surface.width] >> 24; + ssss[3] = src[-sps->surface.width + 1] >> 24; +} + +static void +write_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) +{ + static const GLuint mask = 0x00ffffff; + const GLint invY = sps->surface.height - y - 1; + GLuint *dst = (GLuint *) (sps->surface.ptr + + (invY * sps->surface.stride + x) * sps->surface.cpp); + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + /* write high byte */ + dst[0] = (dst[0] & mask) | (ssss[0] << 24); + dst[1] = (dst[1] & mask) | (ssss[1] << 24); + dst -= sps->surface.stride; + dst[0] = (dst[0] & mask) | (ssss[2] << 24); + dst[1] = (dst[1] & mask) | (ssss[3] << 24); +} + + +static void * +map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) +{ + struct softpipe_surface *sps = (struct softpipe_surface *) pb; + struct intel_renderbuffer *irb = (struct intel_renderbuffer *) sps->surface.rb; + assert(access_mode == PIPE_MAP_READ_WRITE); + + /*LOCK_HARDWARE(intel);*/ + + if (irb->region) { + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); +#if 0 + intelFinish(&intel->ctx); /* XXX need this? */ +#endif + intel_region_map(intel->intelScreen, irb->region); + } + pb->ptr = irb->region->map; + + sps->surface.stride = irb->region->pitch; + sps->surface.cpp = irb->region->cpp; + sps->surface.ptr = irb->region->map; + + return pb->ptr; +} + + +static void +unmap_surface_buffer(struct pipe_buffer *pb) +{ + struct softpipe_surface *sps = (struct softpipe_surface *) pb; + struct intel_renderbuffer *irb = (struct intel_renderbuffer *) sps->surface.rb; + + if (irb->region) { + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); + intel_region_unmap(intel->intelScreen, irb->region); + } + pb->ptr = NULL; + + sps->surface.stride = 0; + sps->surface.cpp = 0; + sps->surface.ptr = NULL; + + /*UNLOCK_HARDWARE(intel);*/ +} + + +struct pipe_surface * +intel_new_surface(GLuint intFormat) +{ + struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); + if (!sps) + return NULL; + + sps->surface.width = 0; /* set in intel_alloc_renderbuffer_storage() */ + sps->surface.height = 0; + + if (intFormat == GL_RGBA8) { + sps->surface.format = PIPE_FORMAT_U_A8_R8_G8_B8; + sps->read_quad_f_swz = read_quad_f_swz; + sps->write_quad_f_swz = write_quad_f_swz; + } + else if (intFormat == GL_RGB5) { + sps->surface.format = PIPE_FORMAT_U_R5_G6_B5; + + } + else if (intFormat == GL_DEPTH_COMPONENT16) { + sps->surface.format = PIPE_FORMAT_U_Z16; + + } + else if (intFormat == GL_DEPTH24_STENCIL8_EXT) { + sps->surface.format = PIPE_FORMAT_Z24_S8; + sps->read_quad_z = read_quad_z24; + sps->write_quad_z = write_quad_z24; + sps->read_quad_stencil = read_quad_stencil; + sps->write_quad_stencil = write_quad_stencil; + } + else { + /* TBD / unknown */ + + } + + sps->surface.buffer.map = map_surface_buffer; + sps->surface.buffer.unmap = unmap_surface_buffer; + + return &sps->surface; +} + + + +struct pipe_surface * +xmesa_get_stencil_surface(GLcontext *ctx) +{ + /* XXX fix */ + return NULL; +} + + + diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.c b/src/mesa/drivers/dri/i915pipe/intel_tex.c new file mode 100644 index 0000000000..b08dee43bc --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.c @@ -0,0 +1,192 @@ +#include "texobj.h" +#include "intel_context.h" +#include "intel_mipmap_tree.h" +#include "intel_tex.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +static GLboolean +intelIsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) +{ +#if 0 + struct intel_context *intel = intel_context(ctx); + struct intel_texture_object *intelObj = intel_texture_object(texObj); + + return + intelObj->mt && + intelObj->mt->region && + intel_is_region_resident(intel, intelObj->mt->region); +#endif + return 1; +} + + + +static struct gl_texture_image * +intelNewTextureImage(GLcontext * ctx) +{ + DBG("%s\n", __FUNCTION__); + (void) ctx; + return (struct gl_texture_image *) CALLOC_STRUCT(intel_texture_image); +} + + +static struct gl_texture_object * +intelNewTextureObject(GLcontext * ctx, GLuint name, GLenum target) +{ + struct intel_texture_object *obj = CALLOC_STRUCT(intel_texture_object); + + DBG("%s\n", __FUNCTION__); + _mesa_initialize_texture_object(&obj->base, name, target); + + return &obj->base; +} + +static void +intelDeleteTextureObject(GLcontext *ctx, + struct gl_texture_object *texObj) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_object *intelObj = intel_texture_object(texObj); + + if (intelObj->mt) + intel_miptree_release(intel, &intelObj->mt); + + _mesa_delete_texture_object(ctx, texObj); +} + + +static void +intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_image *intelImage = intel_texture_image(texImage); + + DBG("%s\n", __FUNCTION__); + + if (intelImage->mt) { + intel_miptree_release(intel, &intelImage->mt); + } + + if (texImage->Data) { + free(texImage->Data); + texImage->Data = NULL; + } +} + + +/* The system memcpy (at least on ubuntu 5.10) has problems copying + * to agp (writecombined) memory from a source which isn't 64-byte + * aligned - there is a 4x performance falloff. + * + * The x86 __memcpy is immune to this but is slightly slower + * (10%-ish) than the system memcpy. + * + * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but + * isn't much faster than x86_memcpy for agp copies. + * + * TODO: switch dynamically. + */ +static void * +do_memcpy(void *dest, const void *src, size_t n) +{ + if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) { + return __memcpy(dest, src, n); + } + else + return memcpy(dest, src, n); +} + + +#if DO_DEBUG + +#ifndef __x86_64__ +static unsigned +fastrdtsc(void) +{ + unsigned eax; + __asm__ volatile ("\t" + "pushl %%ebx\n\t" + "cpuid\n\t" ".byte 0x0f, 0x31\n\t" + "popl %%ebx\n":"=a" (eax) + :"0"(0) + :"ecx", "edx", "cc"); + + return eax; +} +#else +static unsigned +fastrdtsc(void) +{ + unsigned eax; + __asm__ volatile ("\t" "cpuid\n\t" ".byte 0x0f, 0x31\n\t":"=a" (eax) + :"0"(0) + :"ecx", "edx", "ebx", "cc"); + + return eax; +} +#endif + +static unsigned +time_diff(unsigned t, unsigned t2) +{ + return ((t < t2) ? t2 - t : 0xFFFFFFFFU - (t - t2 - 1)); +} + + +static void * +timed_memcpy(void *dest, const void *src, size_t n) +{ + void *ret; + unsigned t1, t2; + double rate; + + if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) + _mesa_printf("Warning - non-aligned texture copy!\n"); + + t1 = fastrdtsc(); + ret = do_memcpy(dest, src, n); + t2 = fastrdtsc(); + + rate = time_diff(t1, t2); + rate /= (double) n; + _mesa_printf("timed_memcpy: %u %u --> %f clocks/byte\n", t1, t2, rate); + return ret; +} +#endif /* DO_DEBUG */ + + +void +intelInitTextureFuncs(struct dd_function_table *functions) +{ + functions->ChooseTextureFormat = intelChooseTextureFormat; + functions->TexImage1D = intelTexImage1D; + functions->TexImage2D = intelTexImage2D; + functions->TexImage3D = intelTexImage3D; + functions->TexSubImage1D = intelTexSubImage1D; + functions->TexSubImage2D = intelTexSubImage2D; + functions->TexSubImage3D = intelTexSubImage3D; + functions->CopyTexImage1D = intelCopyTexImage1D; + functions->CopyTexImage2D = intelCopyTexImage2D; + functions->CopyTexSubImage1D = intelCopyTexSubImage1D; + functions->CopyTexSubImage2D = intelCopyTexSubImage2D; + functions->GetTexImage = intelGetTexImage; + + /* compressed texture functions */ + functions->CompressedTexImage2D = intelCompressedTexImage2D; + functions->GetCompressedTexImage = intelGetCompressedTexImage; + + functions->NewTextureObject = intelNewTextureObject; + functions->NewTextureImage = intelNewTextureImage; + functions->DeleteTexture = intelDeleteTextureObject; + functions->FreeTexImageData = intelFreeTextureImageData; + functions->UpdateTexturePalette = 0; + functions->IsTextureResident = intelIsTextureResident; + +#if DO_DEBUG + if (INTEL_DEBUG & DEBUG_BUFMGR) + functions->TextureMemCpy = timed_memcpy; + else +#endif + functions->TextureMemCpy = do_memcpy; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.h b/src/mesa/drivers/dri/i915pipe/intel_tex.h new file mode 100644 index 0000000000..b77d7a1d8a --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.h @@ -0,0 +1,151 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTELTEX_INC +#define INTELTEX_INC + +#include "mtypes.h" +#include "intel_context.h" +#include "texmem.h" + + +void intelInitTextureFuncs(struct dd_function_table *functions); + +const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx, + GLint internalFormat, + GLenum format, + GLenum type); + + +void intelTexImage3D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexSubImage3D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexImage2D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexSubImage2D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexImage1D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexSubImage1D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLint border); + +void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border); + +void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint x, GLint y, GLsizei width); + +void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, GLsizei width, GLsizei height); + +void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid *data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage ); + +void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, + GLvoid *pixels, + const struct gl_texture_object *texObj, + const struct gl_texture_image *texImage); + +void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, + unsigned long long offset, GLint depth, GLuint pitch); + +GLuint intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit); + +void intel_tex_map_images(struct intel_context *intel, + struct intel_texture_object *intelObj); + +void intel_tex_unmap_images(struct intel_context *intel, + struct intel_texture_object *intelObj); + +int intel_compressed_num_bytes(GLuint mesaFormat); + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c new file mode 100644 index 0000000000..8d7f8f9d91 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c @@ -0,0 +1,293 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "mtypes.h" +#include "enums.h" +#include "image.h" +#include "teximage.h" +#include "swrast/swrast.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_buffers.h" +#include "intel_mipmap_tree.h" +#include "intel_regions.h" +#include "intel_fbo.h" +#include "intel_tex.h" +#include "intel_blit.h" +#include "intel_pixel.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +/** + * Get the intel_region which is the source for any glCopyTex[Sub]Image call. + * + * Do the best we can using the blitter. A future project is to use + * the texture engine and fragment programs for these copies. + */ +static const struct intel_region * +get_teximage_source(struct intel_context *intel, GLenum internalFormat) +{ + struct intel_renderbuffer *irb; + + DBG("%s %s\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(internalFormat)); + + switch (internalFormat) { + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16_ARB: + irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); + if (irb && irb->region && irb->region->cpp == 2) + return irb->region; + return NULL; + case GL_DEPTH24_STENCIL8_EXT: + case GL_DEPTH_STENCIL_EXT: + irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); + if (irb && irb->region && irb->region->cpp == 4) + return irb->region; + return NULL; + case GL_RGBA: + case GL_RGBA8: + return intel_readbuf_region(intel); + case GL_RGB: + if (intel->intelScreen->cpp == 2) + return intel_readbuf_region(intel); + return NULL; + default: + return NULL; + } +} + + +static GLboolean +do_copy_texsubimage(struct intel_context *intel, + struct intel_texture_image *intelImage, + GLenum internalFormat, + GLint dstx, GLint dsty, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + GLcontext *ctx = &intel->ctx; + const struct intel_region *src = + get_teximage_source(intel, internalFormat); + + if (!intelImage->mt || !src) { + DBG("%s fail %p %p\n", __FUNCTION__, intelImage->mt, src); + return GL_FALSE; + } + + intelFlush(ctx); + /* XXX still need the lock ? */ + LOCK_HARDWARE(intel); + { + GLuint image_offset = intel_miptree_image_offset(intelImage->mt, + intelImage->face, + intelImage->level); + const GLint orig_x = x; + const GLint orig_y = y; + const struct gl_framebuffer *fb = ctx->DrawBuffer; + + if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, + &x, &y, &width, &height)) { + /* Update dst for clipped src. Need to also clip the source rect. + */ + dstx += x - orig_x; + dsty += y - orig_y; + + if (!(ctx->ReadBuffer->Name == 0)) { + /* XXX this looks bogus ? */ + /* FBO: invert Y */ + y = ctx->ReadBuffer->Height - y - 1; + } + + /* A bit of fiddling to get the blitter to work with -ve + * pitches. But we get a nice inverted blit this way, so it's + * worth it: + */ + intelEmitCopyBlit(intel, + intelImage->mt->cpp, + -src->pitch, + src->buffer, + src->height * src->pitch * src->cpp, + intelImage->mt->pitch, + intelImage->mt->region->buffer, + image_offset, + x, y + height, dstx, dsty, width, height, + GL_COPY); /* ? */ + + intel_batchbuffer_flush(intel->batch); + } + } + + + UNLOCK_HARDWARE(intel); + +#if 0 + /* GL_SGIS_generate_mipmap -- this can be accelerated now. + * XXX Add a ctx->Driver.GenerateMipmaps() function? + */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif + + return GL_TRUE; +} + + + + + +void +intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLint border) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage1D(ctx, target, level, internalFormat, + width, border, + GL_RGBA, CHAN_TYPE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + if (!do_copy_texsubimage(intel_context(ctx), + intel_texture_image(texImage), + internalFormat, 0, 0, x, y, width, 1)) + goto fail; + + return; + + fail: + _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, + width, border); +} + +void +intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage2D(ctx, target, level, internalFormat, + width, height, border, + GL_RGBA, CHAN_TYPE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + + if (!do_copy_texsubimage(intel_context(ctx), + intel_texture_image(texImage), + internalFormat, 0, 0, x, y, width, height)) + goto fail; + + return; + + fail: + _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, + width, height, border); +} + + +void +intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint x, GLint y, GLsizei width) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + GLenum internalFormat = texImage->InternalFormat; + + /* XXX need to check as in above function? */ + + /* Need to check texture is compatible with source format. + */ + + if (!do_copy_texsubimage(intel_context(ctx), + intel_texture_image(texImage), + internalFormat, xoffset, 0, x, y, width, 1)) { + _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); + } +} + + + +void +intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + GLenum internalFormat = texImage->InternalFormat; + + + /* Need to check texture is compatible with source format. + */ + + if (!do_copy_texsubimage(intel_context(ctx), + intel_texture_image(texImage), + internalFormat, + xoffset, yoffset, x, y, width, height)) { + + DBG("%s - fallback to swrast\n", __FUNCTION__); + + _swrast_copy_texsubimage2d(ctx, target, level, + xoffset, yoffset, x, y, width, height); + } +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_format.c b/src/mesa/drivers/dri/i915pipe/intel_tex_format.c new file mode 100644 index 0000000000..6e058dff69 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_format.c @@ -0,0 +1,172 @@ +#include "intel_context.h" +#include "intel_tex.h" +#include "texformat.h" +#include "enums.h" + +/* It works out that this function is fine for all the supported + * hardware. However, there is still a need to map the formats onto + * hardware descriptors. + */ +/* Note that the i915 can actually support many more formats than + * these if we take the step of simply swizzling the colors + * immediately after sampling... + */ +const struct gl_texture_format * +intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, + GLenum format, GLenum type) +{ + struct intel_context *intel = intel_context(ctx); + const GLboolean do32bpt = (intel->intelScreen->cpp == 4); + + switch (internalFormat) { + case 4: + case GL_RGBA: + case GL_COMPRESSED_RGBA: + if (format == GL_BGRA) { + if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { + return &_mesa_texformat_argb8888; + } + else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { + return &_mesa_texformat_argb4444; + } + else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { + return &_mesa_texformat_argb1555; + } + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case 3: + case GL_RGB: + case GL_COMPRESSED_RGB: + if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { + return &_mesa_texformat_rgb565; + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case GL_RGBA4: + case GL_RGBA2: + return &_mesa_texformat_argb4444; + + case GL_RGB5_A1: + return &_mesa_texformat_argb1555; + + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + return &_mesa_texformat_argb8888; + + case GL_RGB5: + case GL_RGB4: + case GL_R3_G3_B2: + return &_mesa_texformat_rgb565; + + case GL_ALPHA: + case GL_ALPHA4: + case GL_ALPHA8: + case GL_ALPHA12: + case GL_ALPHA16: + case GL_COMPRESSED_ALPHA: + return &_mesa_texformat_a8; + + case 1: + case GL_LUMINANCE: + case GL_LUMINANCE4: + case GL_LUMINANCE8: + case GL_LUMINANCE12: + case GL_LUMINANCE16: + case GL_COMPRESSED_LUMINANCE: + return &_mesa_texformat_l8; + + case 2: + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE4_ALPHA4: + case GL_LUMINANCE6_ALPHA2: + case GL_LUMINANCE8_ALPHA8: + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: + case GL_COMPRESSED_LUMINANCE_ALPHA: + return &_mesa_texformat_al88; + + case GL_INTENSITY: + case GL_INTENSITY4: + case GL_INTENSITY8: + case GL_INTENSITY12: + case GL_INTENSITY16: + case GL_COMPRESSED_INTENSITY: + return &_mesa_texformat_i8; + + case GL_YCBCR_MESA: + if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) + return &_mesa_texformat_ycbcr; + else + return &_mesa_texformat_ycbcr_rev; + + case GL_COMPRESSED_RGB_FXT1_3DFX: + return &_mesa_texformat_rgb_fxt1; + case GL_COMPRESSED_RGBA_FXT1_3DFX: + return &_mesa_texformat_rgba_fxt1; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + return &_mesa_texformat_rgb_dxt1; + + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return &_mesa_texformat_rgba_dxt1; + + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + return &_mesa_texformat_rgba_dxt3; + + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + return &_mesa_texformat_rgba_dxt5; + + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + return &_mesa_texformat_z16; + + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + return &_mesa_texformat_z24_s8; + + default: + fprintf(stderr, "unexpected texture format %s in %s\n", + _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); + return NULL; + } + + return NULL; /* never get here */ +} + +int intel_compressed_num_bytes(GLuint mesaFormat) +{ + int bytes = 0; + switch(mesaFormat) { + + case MESA_FORMAT_RGB_FXT1: + case MESA_FORMAT_RGBA_FXT1: + case MESA_FORMAT_RGB_DXT1: + case MESA_FORMAT_RGBA_DXT1: + bytes = 2; + break; + + case MESA_FORMAT_RGBA_DXT3: + case MESA_FORMAT_RGBA_DXT5: + bytes = 4; + default: + break; + } + + return bytes; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c new file mode 100644 index 0000000000..f790b1e6f7 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c @@ -0,0 +1,691 @@ + +#include +#include + +#include "glheader.h" +#include "macros.h" +#include "mtypes.h" +#include "enums.h" +#include "colortab.h" +#include "convolve.h" +#include "context.h" +#include "simple_list.h" +#include "texcompress.h" +#include "texformat.h" +#include "texobj.h" +#include "texstore.h" + +#include "intel_context.h" +#include "intel_mipmap_tree.h" +#include "intel_buffer_objects.h" +#include "intel_batchbuffer.h" +#include "intel_tex.h" +#include "intel_ioctl.h" +#include "intel_blit.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +/* Functions to store texture images. Where possible, mipmap_tree's + * will be created or further instantiated with image data, otherwise + * images will be stored in malloc'd memory. A validation step is + * required to pull those images into a mipmap tree, or otherwise + * decide a fallback is required. + */ + + +static int +logbase2(int n) +{ + GLint i = 1; + GLint log2 = 0; + + while (n > i) { + i *= 2; + log2++; + } + + return log2; +} + + +/* Otherwise, store it in memory if (Border != 0) or (any dimension == + * 1). + * + * Otherwise, if max_level >= level >= min_level, create tree with + * space for textures from min_level down to max_level. + * + * Otherwise, create tree with space for textures from (level + * 0)..(1x1). Consider pruning this tree at a validation if the + * saving is worth it. + */ +static void +guess_and_alloc_mipmap_tree(struct intel_context *intel, + struct intel_texture_object *intelObj, + struct intel_texture_image *intelImage) +{ + GLuint firstLevel; + GLuint lastLevel; + GLuint width = intelImage->base.Width; + GLuint height = intelImage->base.Height; + GLuint depth = intelImage->base.Depth; + GLuint l2width, l2height, l2depth; + GLuint i, comp_byte = 0; + + DBG("%s\n", __FUNCTION__); + + if (intelImage->base.Border) + return; + + if (intelImage->level > intelObj->base.BaseLevel && + (intelImage->base.Width == 1 || + (intelObj->base.Target != GL_TEXTURE_1D && + intelImage->base.Height == 1) || + (intelObj->base.Target == GL_TEXTURE_3D && + intelImage->base.Depth == 1))) + return; + + /* If this image disrespects BaseLevel, allocate from level zero. + * Usually BaseLevel == 0, so it's unlikely to happen. + */ + if (intelImage->level < intelObj->base.BaseLevel) + firstLevel = 0; + else + firstLevel = intelObj->base.BaseLevel; + + + /* Figure out image dimensions at start level. + */ + for (i = intelImage->level; i > firstLevel; i--) { + width <<= 1; + if (height != 1) + height <<= 1; + if (depth != 1) + depth <<= 1; + } + + /* Guess a reasonable value for lastLevel. This is probably going + * to be wrong fairly often and might mean that we have to look at + * resizable buffers, or require that buffers implement lazy + * pagetable arrangements. + */ + if ((intelObj->base.MinFilter == GL_NEAREST || + intelObj->base.MinFilter == GL_LINEAR) && + intelImage->level == firstLevel) { + lastLevel = firstLevel; + } + else { + l2width = logbase2(width); + l2height = logbase2(height); + l2depth = logbase2(depth); + lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); + } + + assert(!intelObj->mt); + if (intelImage->base.IsCompressed) + comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); + intelObj->mt = intel_miptree_create(intel, + intelObj->base.Target, + intelImage->base.InternalFormat, + firstLevel, + lastLevel, + width, + height, + depth, + intelImage->base.TexFormat->TexelBytes, + comp_byte); + + DBG("%s - success\n", __FUNCTION__); +} + + + + +static GLuint +target_to_face(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + return ((GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X); + default: + return 0; + } +} + +/* There are actually quite a few combinations this will work for, + * more than what I've listed here. + */ +static GLboolean +check_pbo_format(GLint internalFormat, + GLenum format, GLenum type, + const struct gl_texture_format *mesa_format) +{ + switch (internalFormat) { + case 4: + case GL_RGBA: + return (format == GL_BGRA && + (type == GL_UNSIGNED_BYTE || + type == GL_UNSIGNED_INT_8_8_8_8_REV) && + mesa_format == &_mesa_texformat_argb8888); + case 3: + case GL_RGB: + return (format == GL_RGB && + type == GL_UNSIGNED_SHORT_5_6_5 && + mesa_format == &_mesa_texformat_rgb565); + case GL_YCBCR_MESA: + return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); + default: + return GL_FALSE; + } +} + + +/* XXX: Do this for TexSubImage also: + */ +static GLboolean +try_pbo_upload(struct intel_context *intel, + struct intel_texture_image *intelImage, + const struct gl_pixelstore_attrib *unpack, + GLint internalFormat, + GLint width, GLint height, + GLenum format, GLenum type, const void *pixels) +{ + struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); + GLuint src_offset, src_stride; + GLuint dst_offset, dst_stride; + + if (!pbo || + intel->ctx._ImageTransferState || + unpack->SkipPixels || unpack->SkipRows) { + _mesa_printf("%s: failure 1\n", __FUNCTION__); + return GL_FALSE; + } + + src_offset = (GLuint) pixels; + + if (unpack->RowLength > 0) + src_stride = unpack->RowLength; + else + src_stride = width; + + dst_offset = intel_miptree_image_offset(intelImage->mt, + intelImage->face, + intelImage->level); + + dst_stride = intelImage->mt->pitch; + + intelFlush(&intel->ctx); + LOCK_HARDWARE(intel); + { + struct _DriBufferObject *src_buffer = + intel_bufferobj_buffer(intel, pbo, INTEL_READ); + struct _DriBufferObject *dst_buffer = + intel_region_buffer(intel->intelScreen, intelImage->mt->region, + INTEL_WRITE_FULL); + + + intelEmitCopyBlit(intel, + intelImage->mt->cpp, + src_stride, src_buffer, src_offset, + dst_stride, dst_buffer, dst_offset, + 0, 0, 0, 0, width, height, + GL_COPY); + + intel_batchbuffer_flush(intel->batch); + } + UNLOCK_HARDWARE(intel); + + return GL_TRUE; +} + + + +static GLboolean +try_pbo_zcopy(struct intel_context *intel, + struct intel_texture_image *intelImage, + const struct gl_pixelstore_attrib *unpack, + GLint internalFormat, + GLint width, GLint height, + GLenum format, GLenum type, const void *pixels) +{ + struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); + GLuint src_offset, src_stride; + GLuint dst_offset, dst_stride; + + if (!pbo || + intel->ctx._ImageTransferState || + unpack->SkipPixels || unpack->SkipRows) { + _mesa_printf("%s: failure 1\n", __FUNCTION__); + return GL_FALSE; + } + + src_offset = (GLuint) pixels; + + if (unpack->RowLength > 0) + src_stride = unpack->RowLength; + else + src_stride = width; + + dst_offset = intel_miptree_image_offset(intelImage->mt, + intelImage->face, + intelImage->level); + + dst_stride = intelImage->mt->pitch; + + if (src_stride != dst_stride || dst_offset != 0 || src_offset != 0) { + _mesa_printf("%s: failure 2\n", __FUNCTION__); + return GL_FALSE; + } + + intel_region_attach_pbo(intel->intelScreen, intelImage->mt->region, pbo); + + return GL_TRUE; +} + + + + + + +static void +intelTexImage(GLcontext * ctx, + GLint dims, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, GLsizei imageSize, int compressed) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_object *intelObj = intel_texture_object(texObj); + struct intel_texture_image *intelImage = intel_texture_image(texImage); + GLint postConvWidth = width; + GLint postConvHeight = height; + GLint texelBytes, sizeInBytes; + GLuint dstRowStride; + + + DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); + + intelFlush(ctx); + + intelImage->face = target_to_face(target); + intelImage->level = level; + + if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { + _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth, + &postConvHeight); + } + + /* choose the texture format */ + texImage->TexFormat = intelChooseTextureFormat(ctx, internalFormat, + format, type); + + _mesa_set_fetch_functions(texImage, dims); + + if (texImage->TexFormat->TexelBytes == 0) { + /* must be a compressed format */ + texelBytes = 0; + texImage->IsCompressed = GL_TRUE; + texImage->CompressedSize = + ctx->Driver.CompressedTextureSize(ctx, texImage->Width, + texImage->Height, texImage->Depth, + texImage->TexFormat->MesaFormat); + } else { + texelBytes = texImage->TexFormat->TexelBytes; + + /* Minimum pitch of 32 bytes */ + if (postConvWidth * texelBytes < 32) { + postConvWidth = 32 / texelBytes; + texImage->RowStride = postConvWidth; + } + + assert(texImage->RowStride == postConvWidth); + } + + /* Release the reference to a potentially orphaned buffer. + * Release any old malloced memory. + */ + if (intelImage->mt) { + intel_miptree_release(intel, &intelImage->mt); + assert(!texImage->Data); + } + else if (texImage->Data) { + _mesa_align_free(texImage->Data); + } + + /* If this is the only texture image in the tree, could call + * bmBufferData with NULL data to free the old block and avoid + * waiting on any outstanding fences. + */ + if (intelObj->mt && + intelObj->mt->first_level == level && + intelObj->mt->last_level == level && + intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && + !intel_miptree_match_image(intelObj->mt, &intelImage->base, + intelImage->face, intelImage->level)) { + + DBG("release it\n"); + intel_miptree_release(intel, &intelObj->mt); + assert(!intelObj->mt); + } + + if (!intelObj->mt) { + guess_and_alloc_mipmap_tree(intel, intelObj, intelImage); + if (!intelObj->mt) { + DBG("guess_and_alloc_mipmap_tree: failed\n"); + } + } + + assert(!intelImage->mt); + + if (intelObj->mt && + intel_miptree_match_image(intelObj->mt, &intelImage->base, + intelImage->face, intelImage->level)) { + + intel_miptree_reference(&intelImage->mt, intelObj->mt); + assert(intelImage->mt); + } + + if (!intelImage->mt) + DBG("XXX: Image did not fit into tree - storing in local memory!\n"); + + /* PBO fastpaths: + */ + if (dims <= 2 && + intelImage->mt && + intel_buffer_object(unpack->BufferObj) && + check_pbo_format(internalFormat, format, + type, intelImage->base.TexFormat)) { + + DBG("trying pbo upload\n"); + + /* Attempt to texture directly from PBO data (zero copy upload). + * + * Currently disable as it can lead to worse as well as better + * performance (in particular when intel_region_cow() is + * required). + */ + if (intelObj->mt == intelImage->mt && + intelObj->mt->first_level == level && + intelObj->mt->last_level == level) { + + if (try_pbo_zcopy(intel, intelImage, unpack, + internalFormat, + width, height, format, type, pixels)) { + + DBG("pbo zcopy upload succeeded\n"); + return; + } + } + + + /* Otherwise, attempt to use the blitter for PBO image uploads. + */ + if (try_pbo_upload(intel, intelImage, unpack, + internalFormat, + width, height, format, type, pixels)) { + DBG("pbo upload succeeded\n"); + return; + } + + DBG("pbo upload failed\n"); + } + + + + /* intelCopyTexImage calls this function with pixels == NULL, with + * the expectation that the mipmap tree will be set up but nothing + * more will be done. This is where those calls return: + */ + if (compressed) { + pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels, + unpack, + "glCompressedTexImage"); + } else { + pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, + format, type, + pixels, unpack, "glTexImage"); + } + if (!pixels) + return; + + + if (intelImage->mt) + intel_region_idle(intel->intelScreen, intelImage->mt->region); + + LOCK_HARDWARE(intel); + + if (intelImage->mt) { + texImage->Data = intel_miptree_image_map(intel, + intelImage->mt, + intelImage->face, + intelImage->level, + &dstRowStride, + intelImage->base.ImageOffsets); + } + else { + /* Allocate regular memory and store the image there temporarily. */ + if (texImage->IsCompressed) { + sizeInBytes = texImage->CompressedSize; + dstRowStride = + _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); + assert(dims != 3); + } + else { + dstRowStride = postConvWidth * texelBytes; + sizeInBytes = depth * dstRowStride * postConvHeight; + } + + texImage->Data = malloc(sizeInBytes); + } + + DBG("Upload image %dx%dx%d row_len %x " + "pitch %x\n", + width, height, depth, width * texelBytes, dstRowStride); + + /* Copy data. Would like to know when it's ok for us to eg. use + * the blitter to copy. Or, use the hardware to do the format + * conversion and copy: + */ + if (compressed) { + memcpy(texImage->Data, pixels, imageSize); + } else if (!texImage->TexFormat->StoreImage(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, unpack)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + } + + _mesa_unmap_teximage_pbo(ctx, unpack); + + if (intelImage->mt) { + intel_miptree_image_unmap(intel, intelImage->mt); + texImage->Data = NULL; + } + + UNLOCK_HARDWARE(intel); + +#if 0 + /* GL_SGIS_generate_mipmap -- this can be accelerated now. + */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif +} + +void +intelTexImage3D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexImage(ctx, 3, target, level, + internalFormat, width, height, depth, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + + +void +intelTexImage2D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexImage(ctx, 2, target, level, + internalFormat, width, height, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + +void +intelTexImage1D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexImage(ctx, 1, target, level, + internalFormat, width, 1, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + +void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid *data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage ) +{ + intelTexImage(ctx, 2, target, level, + internalFormat, width, height, 1, border, + 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1); +} + +/** + * Need to map texture image into memory before copying image data, + * then unmap it. + */ +static void +intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, int compressed) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_image *intelImage = intel_texture_image(texImage); + + /* Map */ + if (intelImage->mt) { + /* Image is stored in hardware format in a buffer managed by the + * kernel. Need to explicitly map and unmap it. + */ + intelImage->base.Data = + intel_miptree_image_map(intel, + intelImage->mt, + intelImage->face, + intelImage->level, + &intelImage->base.RowStride, + intelImage->base.ImageOffsets); + intelImage->base.RowStride /= intelImage->mt->cpp; + } + else { + /* Otherwise, the image should actually be stored in + * intelImage->base.Data. This is pretty confusing for + * everybody, I'd much prefer to separate the two functions of + * texImage->Data - storage for texture images in main memory + * and access (ie mappings) of images. In other words, we'd + * create a new texImage->Map field and leave Data simply for + * storage. + */ + assert(intelImage->base.Data); + } + + + if (compressed) { + _mesa_get_compressed_teximage(ctx, target, level, pixels, + texObj, texImage); + } else { + _mesa_get_teximage(ctx, target, level, format, type, pixels, + texObj, texImage); + } + + + /* Unmap */ + if (intelImage->mt) { + intel_miptree_image_unmap(intel, intelImage->mt); + intelImage->base.Data = NULL; + } +} + +void +intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intel_get_tex_image(ctx, target, level, format, type, pixels, + texObj, texImage, 0); + + +} + +void +intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, + GLvoid *pixels, + const struct gl_texture_object *texObj, + const struct gl_texture_image *texImage) +{ + intel_get_tex_image(ctx, target, level, 0, 0, pixels, + texObj, texImage, 1); + +} + +void +intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, + unsigned long long offset, GLint depth, GLuint pitch) +{ + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct intel_texture_object *intelObj = intel_texture_object(tObj); + + if (!intelObj) + return; + + if (intelObj->mt) + intel_miptree_release(intel, &intelObj->mt); + + intelObj->imageOverride = GL_TRUE; + intelObj->depthOverride = depth; + intelObj->pitchOverride = pitch; + + if (offset) + intelObj->textureOffset = offset; +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_layout.c b/src/mesa/drivers/dri/i915pipe/intel_tex_layout.c new file mode 120000 index 0000000000..fe61b44194 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_layout.c @@ -0,0 +1 @@ +../intel/intel_tex_layout.c \ No newline at end of file diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c new file mode 100644 index 0000000000..3935787806 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c @@ -0,0 +1,182 @@ + +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "mtypes.h" +#include "texobj.h" +#include "texstore.h" +#include "enums.h" + +#include "intel_context.h" +#include "intel_tex.h" +#include "intel_mipmap_tree.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +static void +intelTexSubimage(GLcontext * ctx, + GLint dims, + GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLint width, GLint height, GLint depth, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_texture_image *intelImage = intel_texture_image(texImage); + GLuint dstRowStride; + + DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), + level, xoffset, yoffset, width, height); + + intelFlush(ctx); + + pixels = + _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format, + type, pixels, packing, "glTexSubImage2D"); + if (!pixels) + return; + + if (intelImage->mt) + intel_region_idle(intel->intelScreen, intelImage->mt->region); + + LOCK_HARDWARE(intel); + + /* Map buffer if necessary. Need to lock to prevent other contexts + * from uploading the buffer under us. + */ + if (intelImage->mt) + texImage->Data = intel_miptree_image_map(intel, + intelImage->mt, + intelImage->face, + intelImage->level, + &dstRowStride, + texImage->ImageOffsets); + + assert(dstRowStride); + + if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, packing)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); + } + +#if 0 + /* GL_SGIS_generate_mipmap */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + _mesa_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif + + _mesa_unmap_teximage_pbo(ctx, packing); + + if (intelImage->mt) { + intel_miptree_image_unmap(intel, intelImage->mt); + texImage->Data = NULL; + } + + UNLOCK_HARDWARE(intel); +} + + + + + +void +intelTexSubImage3D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + + intelTexSubimage(ctx, 3, + target, level, + xoffset, yoffset, zoffset, + width, height, depth, + format, type, pixels, packing, texObj, texImage); + +} + + + +void +intelTexSubImage2D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + + intelTexSubimage(ctx, 2, + target, level, + xoffset, yoffset, 0, + width, height, 1, + format, type, pixels, packing, texObj, texImage); + +} + + +void +intelTexSubImage1D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexSubimage(ctx, 1, + target, level, + xoffset, 0, 0, + width, 1, 1, + format, type, pixels, packing, texObj, texImage); + +} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c new file mode 100644 index 0000000000..af18c26d55 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c @@ -0,0 +1,272 @@ +#include "mtypes.h" +#include "macros.h" + +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_mipmap_tree.h" +#include "intel_tex.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +/** + * Compute which mipmap levels that really need to be sent to the hardware. + * This depends on the base image size, GL_TEXTURE_MIN_LOD, + * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. + */ +static void +intel_calculate_first_last_level(struct intel_texture_object *intelObj) +{ + struct gl_texture_object *tObj = &intelObj->base; + const struct gl_texture_image *const baseImage = + tObj->Image[0][tObj->BaseLevel]; + + /* These must be signed values. MinLod and MaxLod can be negative numbers, + * and having firstLevel and lastLevel as signed prevents the need for + * extra sign checks. + */ + int firstLevel; + int lastLevel; + + /* Yes, this looks overly complicated, but it's all needed. + */ + switch (tObj->Target) { + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_3D: + case GL_TEXTURE_CUBE_MAP: + if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) { + /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL. + */ + firstLevel = lastLevel = tObj->BaseLevel; + } + else { + firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); + firstLevel = MAX2(firstLevel, tObj->BaseLevel); + lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); + lastLevel = MAX2(lastLevel, tObj->BaseLevel); + lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); + lastLevel = MIN2(lastLevel, tObj->MaxLevel); + lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ + } + break; + case GL_TEXTURE_RECTANGLE_NV: + case GL_TEXTURE_4D_SGIS: + firstLevel = lastLevel = 0; + break; + default: + return; + } + + /* save these values */ + intelObj->firstLevel = firstLevel; + intelObj->lastLevel = lastLevel; +} + +static void +copy_image_data_to_tree(struct intel_context *intel, + struct intel_texture_object *intelObj, + struct intel_texture_image *intelImage) +{ + if (intelImage->mt) { + /* Copy potentially with the blitter: + */ + intel_miptree_image_copy(intel, + intelObj->mt, + intelImage->face, + intelImage->level, intelImage->mt); + + intel_miptree_release(intel, &intelImage->mt); + } + else { + assert(intelImage->base.Data != NULL); + + /* More straightforward upload. + */ + intel_miptree_image_data(intel, + intelObj->mt, + intelImage->face, + intelImage->level, + intelImage->base.Data, + intelImage->base.RowStride, + intelImage->base.RowStride * + intelImage->base.Height); + _mesa_align_free(intelImage->base.Data); + intelImage->base.Data = NULL; + } + + intel_miptree_reference(&intelImage->mt, intelObj->mt); +} + + +/* + */ +GLuint +intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) +{ + struct gl_texture_object *tObj = intel->ctx.Texture.Unit[unit]._Current; + struct intel_texture_object *intelObj = intel_texture_object(tObj); + int comp_byte = 0; + int cpp; + + GLuint face, i; + GLuint nr_faces = 0; + struct intel_texture_image *firstImage; + + GLboolean need_flush = GL_FALSE; + + /* We know/require this is true by now: + */ + assert(intelObj->base._Complete); + + /* What levels must the tree include at a minimum? + */ + intel_calculate_first_last_level(intelObj); + firstImage = + intel_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); + + /* Fallback case: + */ + if (firstImage->base.Border) { + if (intelObj->mt) { + intel_miptree_release(intel, &intelObj->mt); + } + return GL_FALSE; + } + + + /* If both firstImage and intelObj have a tree which can contain + * all active images, favour firstImage. Note that because of the + * completeness requirement, we know that the image dimensions + * will match. + */ + if (firstImage->mt && + firstImage->mt != intelObj->mt && + firstImage->mt->first_level <= intelObj->firstLevel && + firstImage->mt->last_level >= intelObj->lastLevel) { + + if (intelObj->mt) + intel_miptree_release(intel, &intelObj->mt); + + intel_miptree_reference(&intelObj->mt, firstImage->mt); + } + + if (firstImage->base.IsCompressed) { + comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); + cpp = comp_byte; + } + else cpp = firstImage->base.TexFormat->TexelBytes; + + /* Check tree can hold all active levels. Check tree matches + * target, imageFormat, etc. + * + * XXX: For some layouts (eg i945?), the test might have to be + * first_level == firstLevel, as the tree isn't valid except at the + * original start level. Hope to get around this by + * programming minLod, maxLod, baseLevel into the hardware and + * leaving the tree alone. + */ + if (intelObj->mt && + (intelObj->mt->target != intelObj->base.Target || + intelObj->mt->internal_format != firstImage->base.InternalFormat || + intelObj->mt->first_level != intelObj->firstLevel || + intelObj->mt->last_level != intelObj->lastLevel || + intelObj->mt->width0 != firstImage->base.Width || + intelObj->mt->height0 != firstImage->base.Height || + intelObj->mt->depth0 != firstImage->base.Depth || + intelObj->mt->cpp != cpp || + intelObj->mt->compressed != firstImage->base.IsCompressed)) { + intel_miptree_release(intel, &intelObj->mt); + } + + + /* May need to create a new tree: + */ + if (!intelObj->mt) { + intelObj->mt = intel_miptree_create(intel, + intelObj->base.Target, + firstImage->base.InternalFormat, + intelObj->firstLevel, + intelObj->lastLevel, + firstImage->base.Width, + firstImage->base.Height, + firstImage->base.Depth, + cpp, + comp_byte); + } + + /* Pull in any images not in the object's tree: + */ + nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct intel_texture_image *intelImage = + intel_texture_image(intelObj->base.Image[face][i]); + + /* Need to import images in main memory or held in other trees. + */ + if (intelObj->mt != intelImage->mt) { + copy_image_data_to_tree(intel, intelObj, intelImage); + need_flush = GL_TRUE; + } + } + } + + if (need_flush) + intel_batchbuffer_flush(intel->batch); + + return GL_TRUE; +} + + + +void +intel_tex_map_images(struct intel_context *intel, + struct intel_texture_object *intelObj) +{ + GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + + DBG("%s\n", __FUNCTION__); + + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct intel_texture_image *intelImage = + intel_texture_image(intelObj->base.Image[face][i]); + + if (intelImage->mt) { + intelImage->base.Data = + intel_miptree_image_map(intel, + intelImage->mt, + intelImage->face, + intelImage->level, + &intelImage->base.RowStride, + intelImage->base.ImageOffsets); + /* convert stride to texels, not bytes */ + intelImage->base.RowStride /= intelImage->mt->cpp; +/* intelImage->base.ImageStride /= intelImage->mt->cpp; */ + } + } + } +} + + + +void +intel_tex_unmap_images(struct intel_context *intel, + struct intel_texture_object *intelObj) +{ + GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct intel_texture_image *intelImage = + intel_texture_image(intelObj->base.Image[face][i]); + + if (intelImage->mt) { + intel_miptree_image_unmap(intel, intelImage->mt); + intelImage->base.Data = NULL; + } + } + } +} -- cgit v1.2.3 From 43d07d8697a0392a57a676d8e70a0b126b124899 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 31 Jul 2007 20:28:26 +0100 Subject: Remove references to intel_tris.h --- src/mesa/drivers/dri/i915pipe/intel_buffers.c | 1 - src/mesa/drivers/dri/i915pipe/intel_context.c | 1 - src/mesa/drivers/dri/i915pipe/intel_screen.c | 1 - 3 files changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.c b/src/mesa/drivers/dri/i915pipe/intel_buffers.c index 7684e36e3b..cc9a4262df 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.c +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.c @@ -31,7 +31,6 @@ #include "intel_buffers.h" #include "intel_depthstencil.h" #include "intel_fbo.h" -#include "intel_tris.h" #include "intel_regions.h" #include "intel_batchbuffer.h" #include "intel_reg.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index e8e00b9606..395b0f63be 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -51,7 +51,6 @@ #include "intel_buffers.h" #include "intel_tex.h" #include "intel_span.h" -#include "intel_tris.h" #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.c b/src/mesa/drivers/dri/i915pipe/intel_screen.c index 8d8b2523f3..c7c5ed8bbd 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.c +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.c @@ -41,7 +41,6 @@ #include "intel_buffers.h" #include "intel_tex.h" #include "intel_span.h" -#include "intel_tris.h" #include "intel_ioctl.h" #include "intel_fbo.h" -- cgit v1.2.3 From 58c1297b22c61d2c9d9c0ae7a6caf97545fe520e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 31 Jul 2007 20:34:17 +0100 Subject: Add missing files --- src/mesa/drivers/dri/i915pipe/server/i830_common.h | 226 ++++ src/mesa/drivers/dri/i915pipe/server/i830_dri.h | 63 + src/mesa/drivers/dri/i915pipe/server/intel.h | 331 +++++ src/mesa/drivers/dri/i915pipe/server/intel_dri.c | 1306 ++++++++++++++++++++ 4 files changed, 1926 insertions(+) create mode 100644 src/mesa/drivers/dri/i915pipe/server/i830_common.h create mode 100644 src/mesa/drivers/dri/i915pipe/server/i830_dri.h create mode 100644 src/mesa/drivers/dri/i915pipe/server/intel.h create mode 100644 src/mesa/drivers/dri/i915pipe/server/intel_dri.c (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/server/i830_common.h b/src/mesa/drivers/dri/i915pipe/server/i830_common.h new file mode 100644 index 0000000000..d4d58886ce --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/server/i830_common.h @@ -0,0 +1,226 @@ +/************************************************************************** + +Copyright 2001 VA Linux Systems Inc., Fremont, California. +Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas. + +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the "Software"), +to deal in the Software without restriction, including without limitation +on the rights to use, copy, modify, merge, publish, distribute, sub +license, and/or sell copies of the Software, and to permit persons to whom +the Software is furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice (including the next +paragraph) shall be included in all copies or substantial portions of the +Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL +ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, +DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR +OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE +USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ + +#ifndef _I830_COMMON_H_ +#define _I830_COMMON_H_ + + +#define I830_NR_TEX_REGIONS 255 /* maximum due to use of chars for next/prev */ +#define I830_LOG_MIN_TEX_REGION_SIZE 14 + + +/* Driver specific DRM command indices + * NOTE: these are not OS specific, but they are driver specific + */ +#define DRM_I830_INIT 0x00 +#define DRM_I830_FLUSH 0x01 +#define DRM_I830_FLIP 0x02 +#define DRM_I830_BATCHBUFFER 0x03 +#define DRM_I830_IRQ_EMIT 0x04 +#define DRM_I830_IRQ_WAIT 0x05 +#define DRM_I830_GETPARAM 0x06 +#define DRM_I830_SETPARAM 0x07 +#define DRM_I830_ALLOC 0x08 +#define DRM_I830_FREE 0x09 +#define DRM_I830_INIT_HEAP 0x0a +#define DRM_I830_CMDBUFFER 0x0b +#define DRM_I830_DESTROY_HEAP 0x0c +#define DRM_I830_SET_VBLANK_PIPE 0x0d +#define DRM_I830_GET_VBLANK_PIPE 0x0e + +typedef struct { + enum { + I830_INIT_DMA = 0x01, + I830_CLEANUP_DMA = 0x02, + I830_RESUME_DMA = 0x03 + } func; + unsigned int mmio_offset; + int sarea_priv_offset; + unsigned int ring_start; + unsigned int ring_end; + unsigned int ring_size; + unsigned int front_offset; + unsigned int back_offset; + unsigned int depth_offset; + unsigned int w; + unsigned int h; + unsigned int pitch; + unsigned int pitch_bits; + unsigned int back_pitch; + unsigned int depth_pitch; + unsigned int cpp; + unsigned int chipset; +} drmI830Init; + +typedef struct { + drmTextureRegion texList[I830_NR_TEX_REGIONS+1]; + int last_upload; /* last time texture was uploaded */ + int last_enqueue; /* last time a buffer was enqueued */ + int last_dispatch; /* age of the most recently dispatched buffer */ + int ctxOwner; /* last context to upload state */ + int texAge; + int pf_enabled; /* is pageflipping allowed? */ + int pf_active; + int pf_current_page; /* which buffer is being displayed? */ + int perf_boxes; /* performance boxes to be displayed */ + int width, height; /* screen size in pixels */ + + drm_handle_t front_handle; + int front_offset; + int front_size; + + drm_handle_t back_handle; + int back_offset; + int back_size; + + drm_handle_t depth_handle; + int depth_offset; + int depth_size; + + drm_handle_t tex_handle; + int tex_offset; + int tex_size; + int log_tex_granularity; + int pitch; + int rotation; /* 0, 90, 180 or 270 */ + int rotated_offset; + int rotated_size; + int rotated_pitch; + int virtualX, virtualY; + + unsigned int front_tiled; + unsigned int back_tiled; + unsigned int depth_tiled; + unsigned int rotated_tiled; + unsigned int rotated2_tiled; + + int pipeA_x; + int pipeA_y; + int pipeA_w; + int pipeA_h; + int pipeB_x; + int pipeB_y; + int pipeB_w; + int pipeB_h; + + /* Triple buffering */ + drm_handle_t third_handle; + int third_offset; + int third_size; + unsigned int third_tiled; +} drmI830Sarea; + +/* Flags for perf_boxes + */ +#define I830_BOX_RING_EMPTY 0x1 /* populated by kernel */ +#define I830_BOX_FLIP 0x2 /* populated by kernel */ +#define I830_BOX_WAIT 0x4 /* populated by kernel & client */ +#define I830_BOX_TEXTURE_LOAD 0x8 /* populated by kernel */ +#define I830_BOX_LOST_CONTEXT 0x10 /* populated by client */ + + +typedef struct { + int start; /* agp offset */ + int used; /* nr bytes in use */ + int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ + int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ + int num_cliprects; /* mulitpass with multiple cliprects? */ + drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ +} drmI830BatchBuffer; + +typedef struct { + char *buf; /* agp offset */ + int sz; /* nr bytes in use */ + int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ + int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ + int num_cliprects; /* mulitpass with multiple cliprects? */ + drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ +} drmI830CmdBuffer; + +typedef struct { + int *irq_seq; +} drmI830IrqEmit; + +typedef struct { + int irq_seq; +} drmI830IrqWait; + +typedef struct { + int param; + int *value; +} drmI830GetParam; + +#define I830_PARAM_IRQ_ACTIVE 1 +#define I830_PARAM_ALLOW_BATCHBUFFER 2 + +typedef struct { + int param; + int value; +} drmI830SetParam; + +#define I830_SETPARAM_USE_MI_BATCHBUFFER_START 1 +#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY 2 +#define I830_SETPARAM_ALLOW_BATCHBUFFER 3 + + +/* A memory manager for regions of shared memory: + */ +#define I830_MEM_REGION_AGP 1 + +typedef struct { + int region; + int alignment; + int size; + int *region_offset; /* offset from start of fb or agp */ +} drmI830MemAlloc; + +typedef struct { + int region; + int region_offset; +} drmI830MemFree; + +typedef struct { + int region; + int size; + int start; +} drmI830MemInitHeap; + +typedef struct { + int region; +} drmI830MemDestroyHeap; + +#define DRM_I830_VBLANK_PIPE_A 1 +#define DRM_I830_VBLANK_PIPE_B 2 + +typedef struct { + int pipe; +} drmI830VBlankPipe; + +#endif /* _I830_DRM_H_ */ diff --git a/src/mesa/drivers/dri/i915pipe/server/i830_dri.h b/src/mesa/drivers/dri/i915pipe/server/i830_dri.h new file mode 100644 index 0000000000..c2a3af8cbf --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/server/i830_dri.h @@ -0,0 +1,63 @@ +/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ + +#ifndef _I830_DRI_H +#define _I830_DRI_H + +#include "xf86drm.h" +#include "i830_common.h" + +#define I830_MAX_DRAWABLES 256 + +#define I830_MAJOR_VERSION 1 +#define I830_MINOR_VERSION 7 +#define I830_PATCHLEVEL 2 + +#define I830_REG_SIZE 0x80000 + +typedef struct _I830DRIRec { + drm_handle_t regs; + drmSize regsSize; + + drmSize unused1; /* backbufferSize */ + drm_handle_t unused2; /* backbuffer */ + + drmSize unused3; /* depthbufferSize */ + drm_handle_t unused4; /* depthbuffer */ + + drmSize unused5; /* rotatedSize */ + drm_handle_t unused6; /* rotatedbuffer */ + + drm_handle_t unused7; /* textures */ + int unused8; /* textureSize */ + + drm_handle_t unused9; /* agp_buffers */ + drmSize unused10; /* agp_buf_size */ + + int deviceID; + int width; + int height; + int mem; + int cpp; + int bitsPerPixel; + + int unused11[8]; /* was front/back/depth/rotated offset/pitch */ + + int unused12; /* logTextureGranularity */ + int unused13; /* textureOffset */ + + int irq; + int sarea_priv_offset; +} I830DRIRec, *I830DRIPtr; + +typedef struct { + /* Nothing here yet */ + int dummy; +} I830ConfigPrivRec, *I830ConfigPrivPtr; + +typedef struct { + /* Nothing here yet */ + int dummy; +} I830DRIContextRec, *I830DRIContextPtr; + + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/server/intel.h b/src/mesa/drivers/dri/i915pipe/server/intel.h new file mode 100644 index 0000000000..6ea72499c1 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/server/intel.h @@ -0,0 +1,331 @@ +#ifndef _INTEL_H_ +#define _INTEL_H_ + +#include "xf86drm.h" /* drm_handle_t, etc */ + +/* Intel */ +#ifndef PCI_CHIP_I810 +#define PCI_CHIP_I810 0x7121 +#define PCI_CHIP_I810_DC100 0x7123 +#define PCI_CHIP_I810_E 0x7125 +#define PCI_CHIP_I815 0x1132 +#define PCI_CHIP_I810_BRIDGE 0x7120 +#define PCI_CHIP_I810_DC100_BRIDGE 0x7122 +#define PCI_CHIP_I810_E_BRIDGE 0x7124 +#define PCI_CHIP_I815_BRIDGE 0x1130 +#endif + +#define PCI_CHIP_845_G 0x2562 +#define PCI_CHIP_I830_M 0x3577 + +#ifndef PCI_CHIP_I855_GM +#define PCI_CHIP_I855_GM 0x3582 +#define PCI_CHIP_I855_GM_BRIDGE 0x3580 +#endif + +#ifndef PCI_CHIP_I865_G +#define PCI_CHIP_I865_G 0x2572 +#define PCI_CHIP_I865_G_BRIDGE 0x2570 +#endif + +#ifndef PCI_CHIP_I915_G +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_G_BRIDGE 0x2580 +#endif + +#ifndef PCI_CHIP_I915_GM +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I915_GM_BRIDGE 0x2590 +#endif + +#ifndef PCI_CHIP_E7221_G +#define PCI_CHIP_E7221_G 0x258A +/* Same as I915_G_BRIDGE */ +#define PCI_CHIP_E7221_G_BRIDGE 0x2580 +#endif + +#ifndef PCI_CHIP_I945_G +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_G_BRIDGE 0x2770 +#endif + +#ifndef PCI_CHIP_I945_GM +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GM_BRIDGE 0x27A0 +#endif + +#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 || \ + pI810->Chipset == PCI_CHIP_I810_DC100 || \ + pI810->Chipset == PCI_CHIP_I810_E) +#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815) +#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M) +#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G) +#define IS_I85X(pI810) (pI810->Chipset == PCI_CHIP_I855_GM) +#define IS_I852(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME)) +#define IS_I855(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME)) +#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G) + +#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G) +#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM) +#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G) +#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM) +#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810)) + +#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810)) + +#define I830_GMCH_CTRL 0x52 + +#define I830_GMCH_MEM_MASK 0x1 +#define I830_GMCH_MEM_64M 0x1 +#define I830_GMCH_MEM_128M 0 + +#define I830_GMCH_GMS_MASK 0x70 +#define I830_GMCH_GMS_DISABLED 0x00 +#define I830_GMCH_GMS_LOCAL 0x10 +#define I830_GMCH_GMS_STOLEN_512 0x20 +#define I830_GMCH_GMS_STOLEN_1024 0x30 +#define I830_GMCH_GMS_STOLEN_8192 0x40 + +#define I855_GMCH_GMS_MASK (0x7 << 4) +#define I855_GMCH_GMS_DISABLED 0x00 +#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) +#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) +#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) +#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) +#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) +#define I915G_GMCH_GMS_STOLEN_48M (0x6 << 4) +#define I915G_GMCH_GMS_STOLEN_64M (0x7 << 4) + +typedef unsigned char Bool; +#define TRUE 1 +#define FALSE 0 + +#define PIPE_NONE 0<<0 +#define PIPE_CRT 1<<0 +#define PIPE_TV 1<<1 +#define PIPE_DFP 1<<2 +#define PIPE_LFP 1<<3 +#define PIPE_CRT2 1<<4 +#define PIPE_TV2 1<<5 +#define PIPE_DFP2 1<<6 +#define PIPE_LFP2 1<<7 + +typedef struct _I830MemPool *I830MemPoolPtr; +typedef struct _I830MemRange *I830MemRangePtr; +typedef struct _I830MemRange { + long Start; + long End; + long Size; + unsigned long Physical; + unsigned long Offset; /* Offset of AGP-allocated portion */ + unsigned long Alignment; + drm_handle_t Key; + unsigned long Pitch; // add pitch + I830MemPoolPtr Pool; +} I830MemRange; + +typedef struct _I830MemPool { + I830MemRange Total; + I830MemRange Free; + I830MemRange Fixed; + I830MemRange Allocated; +} I830MemPool; + +typedef struct { + int tail_mask; + I830MemRange mem; + unsigned char *virtual_start; + int head; + int tail; + int space; +} I830RingBuffer; + +typedef struct _I830Rec { + unsigned char *MMIOBase; + unsigned char *FbBase; + int cpp; + uint32_t aper_size; + unsigned int bios_version; + + /* These are set in PreInit and never changed. */ + long FbMapSize; + long TotalVideoRam; + I830MemRange StolenMemory; /* pre-allocated memory */ + long BIOSMemorySize; /* min stolen pool size */ + int BIOSMemSizeLoc; + + /* These change according to what has been allocated. */ + long FreeMemory; + I830MemRange MemoryAperture; + I830MemPool StolenPool; + long allocatedMemory; + + /* Regions allocated either from the above pools, or from agpgart. */ + /* for single and dual head configurations */ + I830MemRange FrontBuffer; + I830MemRange FrontBuffer2; + I830MemRange Scratch; + I830MemRange Scratch2; + + I830RingBuffer *LpRing; + + I830MemRange BackBuffer; + I830MemRange DepthBuffer; + I830MemRange TexMem; + int TexGranularity; + I830MemRange ContextMem; + int drmMinor; + Bool have3DWindows; + + Bool NeedRingBufferLow; + Bool allowPageFlip; + Bool disableTiling; + + int Chipset; + unsigned long LinearAddr; + unsigned long MMIOAddr; + + drmSize registerSize; /**< \brief MMIO register map size */ + drm_handle_t registerHandle; /**< \brief MMIO register map handle */ + // IOADDRESS ioBase; + int irq; /**< \brief IRQ number */ + int GttBound; + + drm_handle_t ring_map; + unsigned int Fence[8]; + +} I830Rec; + +/* + * 12288 is set as the maximum, chosen because it is enough for + * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare. + */ +#define I830_MAXIMUM_VBIOS_MEM 12288 +#define I830_DEFAULT_VIDEOMEM_2D (MB(32) / 1024) +#define I830_DEFAULT_VIDEOMEM_3D (MB(64) / 1024) + +/* Flags for memory allocation function */ +#define FROM_ANYWHERE 0x00000000 +#define FROM_POOL_ONLY 0x00000001 +#define FROM_NEW_ONLY 0x00000002 +#define FROM_MASK 0x0000000f + +#define ALLOCATE_AT_TOP 0x00000010 +#define ALLOCATE_AT_BOTTOM 0x00000020 +#define FORCE_GAPS 0x00000040 + +#define NEED_PHYSICAL_ADDR 0x00000100 +#define ALIGN_BOTH_ENDS 0x00000200 +#define FORCE_LOW 0x00000400 + +#define ALLOC_NO_TILING 0x00001000 +#define ALLOC_INITIAL 0x00002000 + +#define ALLOCATE_DRY_RUN 0x80000000 + +/* Chipset registers for VIDEO BIOS memory RW access */ +#define _855_DRAM_RW_CONTROL 0x58 +#define _845_DRAM_RW_CONTROL 0x90 +#define DRAM_WRITE 0x33330000 + +#define KB(x) ((x) * 1024) +#define MB(x) ((x) * KB(1024)) + +#define GTT_PAGE_SIZE KB(4) +#define ROUND_TO(x, y) (((x) + (y) - 1) / (y) * (y)) +#define ROUND_DOWN_TO(x, y) ((x) / (y) * (y)) +#define ROUND_TO_PAGE(x) ROUND_TO((x), GTT_PAGE_SIZE) +#define ROUND_TO_MB(x) ROUND_TO((x), MB(1)) +#define PRIMARY_RINGBUFFER_SIZE KB(128) + + +/* Ring buffer registers, p277, overview p19 + */ +#define LP_RING 0x2030 +#define HP_RING 0x2040 + +#define RING_TAIL 0x00 +#define TAIL_ADDR 0x000FFFF8 +#define I830_TAIL_MASK 0x001FFFF8 + +#define RING_HEAD 0x04 +#define HEAD_WRAP_COUNT 0xFFE00000 +#define HEAD_WRAP_ONE 0x00200000 +#define HEAD_ADDR 0x001FFFFC +#define I830_HEAD_MASK 0x001FFFFC + +#define RING_START 0x08 +#define START_ADDR 0x03FFFFF8 +#define I830_RING_START_MASK 0xFFFFF000 + +#define RING_LEN 0x0C +#define RING_NR_PAGES 0x001FF000 +#define I830_RING_NR_PAGES 0x001FF000 +#define RING_REPORT_MASK 0x00000006 +#define RING_REPORT_64K 0x00000002 +#define RING_REPORT_128K 0x00000004 +#define RING_NO_REPORT 0x00000000 +#define RING_VALID_MASK 0x00000001 +#define RING_VALID 0x00000001 +#define RING_INVALID 0x00000000 + + +/* Fence/Tiling ranges [0..7] + */ +#define FENCE 0x2000 +#define FENCE_NR 8 + +#define I915G_FENCE_START_MASK 0x0ff00000 + +#define I830_FENCE_START_MASK 0x07f80000 + +#define FENCE_START_MASK 0x03F80000 +#define FENCE_X_MAJOR 0x00000000 +#define FENCE_Y_MAJOR 0x00001000 +#define FENCE_SIZE_MASK 0x00000700 +#define FENCE_SIZE_512K 0x00000000 +#define FENCE_SIZE_1M 0x00000100 +#define FENCE_SIZE_2M 0x00000200 +#define FENCE_SIZE_4M 0x00000300 +#define FENCE_SIZE_8M 0x00000400 +#define FENCE_SIZE_16M 0x00000500 +#define FENCE_SIZE_32M 0x00000600 +#define FENCE_SIZE_64M 0x00000700 +#define I915G_FENCE_SIZE_1M 0x00000000 +#define I915G_FENCE_SIZE_2M 0x00000100 +#define I915G_FENCE_SIZE_4M 0x00000200 +#define I915G_FENCE_SIZE_8M 0x00000300 +#define I915G_FENCE_SIZE_16M 0x00000400 +#define I915G_FENCE_SIZE_32M 0x00000500 +#define I915G_FENCE_SIZE_64M 0x00000600 +#define I915G_FENCE_SIZE_128M 0x00000700 +#define FENCE_PITCH_1 0x00000000 +#define FENCE_PITCH_2 0x00000010 +#define FENCE_PITCH_4 0x00000020 +#define FENCE_PITCH_8 0x00000030 +#define FENCE_PITCH_16 0x00000040 +#define FENCE_PITCH_32 0x00000050 +#define FENCE_PITCH_64 0x00000060 +#define FENCE_VALID 0x00000001 + +#include + +# define MMIO_IN8(base, offset) \ + *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) +# define MMIO_IN32(base, offset) \ + read_MMIO_LE32(base, offset) +# define MMIO_OUT8(base, offset, val) \ + *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val) +# define MMIO_OUT32(base, offset, val) \ + *(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val) + + + /* Memory mapped register access macros */ +#define INREG8(addr) MMIO_IN8(MMIO, addr) +#define INREG(addr) MMIO_IN32(MMIO, addr) +#define OUTREG8(addr, val) MMIO_OUT8(MMIO, addr, val) +#define OUTREG(addr, val) MMIO_OUT32(MMIO, addr, val) + +#define DSPABASE 0x70184 + +#endif diff --git a/src/mesa/drivers/dri/i915pipe/server/intel_dri.c b/src/mesa/drivers/dri/i915pipe/server/intel_dri.c new file mode 100644 index 0000000000..e49c4214ad --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/server/intel_dri.c @@ -0,0 +1,1306 @@ +/** + * \file server/intel_dri.c + * \brief File to perform the device-specific initialization tasks typically + * done in the X server. + * + * Here they are converted to run in the client (or perhaps a standalone + * process), and to work with the frame buffer device rather than the X + * server infrastructure. + * + * Copyright (C) 2006 Dave Airlie (airlied@linux.ie) + + Permission is hereby granted, free of charge, to any person obtaining a + copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sub license, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial portions + of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR + ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +*/ + +#include +#include +#include +#include +#include + +#include "driver.h" +#include "drm.h" + +#include "intel.h" +#include "i830_dri.h" + +#include "memops.h" +#include "pciaccess.h" + +static size_t drm_page_size; +static int nextTile = 0; +#define xf86DrvMsg(...) do {} while(0) + +static const int pitches[] = { + 128 * 8, + 128 * 16, + 128 * 32, + 128 * 64, + 0 +}; + +static Bool I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea); + +static unsigned long +GetBestTileAlignment(unsigned long size) +{ + unsigned long i; + + for (i = KB(512); i < size; i <<= 1) + ; + + if (i > MB(64)) + i = MB(64); + + return i; +} + +static void SetFenceRegs(const DRIDriverContext *ctx, I830Rec *pI830) +{ + int i; + unsigned char *MMIO = ctx->MMIOAddress; + + for (i = 0; i < 8; i++) { + OUTREG(FENCE + i * 4, pI830->Fence[i]); + // if (I810_DEBUG & DEBUG_VERBOSE_VGA) + fprintf(stderr,"Fence Register : %x\n", pI830->Fence[i]); + } +} + +/* Tiled memory is good... really, really good... + * + * Need to make it less likely that we miss out on this - probably + * need to move the frontbuffer away from the 'guarenteed' alignment + * of the first memory segment, or perhaps allocate a discontigous + * framebuffer to get more alignment 'sweet spots'. + */ +static void +SetFence(const DRIDriverContext *ctx, I830Rec *pI830, + int nr, unsigned int start, unsigned int pitch, + unsigned int size) +{ + unsigned int val; + unsigned int fence_mask = 0; + unsigned int fence_pitch; + + if (nr < 0 || nr > 7) { + fprintf(stderr, + "SetFence: fence %d out of range\n",nr); + return; + } + + pI830->Fence[nr] = 0; + + if (IS_I9XX(pI830)) + fence_mask = ~I915G_FENCE_START_MASK; + else + fence_mask = ~I830_FENCE_START_MASK; + + if (start & fence_mask) { + fprintf(stderr, + "SetFence: %d: start (0x%08x) is not %s aligned\n", + nr, start, (IS_I9XX(pI830)) ? "1MB" : "512k"); + return; + } + + if (start % size) { + fprintf(stderr, + "SetFence: %d: start (0x%08x) is not size (%dk) aligned\n", + nr, start, size / 1024); + return; + } + + if (pitch & 127) { + fprintf(stderr, + "SetFence: %d: pitch (%d) not a multiple of 128 bytes\n", + nr, pitch); + return; + } + + val = (start | FENCE_X_MAJOR | FENCE_VALID); + + if (IS_I9XX(pI830)) { + switch (size) { + case MB(1): + val |= I915G_FENCE_SIZE_1M; + break; + case MB(2): + val |= I915G_FENCE_SIZE_2M; + break; + case MB(4): + val |= I915G_FENCE_SIZE_4M; + break; + case MB(8): + val |= I915G_FENCE_SIZE_8M; + break; + case MB(16): + val |= I915G_FENCE_SIZE_16M; + break; + case MB(32): + val |= I915G_FENCE_SIZE_32M; + break; + case MB(64): + val |= I915G_FENCE_SIZE_64M; + break; + default: + fprintf(stderr, + "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); + return; + } + } else { + switch (size) { + case KB(512): + val |= FENCE_SIZE_512K; + break; + case MB(1): + val |= FENCE_SIZE_1M; + break; + case MB(2): + val |= FENCE_SIZE_2M; + break; + case MB(4): + val |= FENCE_SIZE_4M; + break; + case MB(8): + val |= FENCE_SIZE_8M; + break; + case MB(16): + val |= FENCE_SIZE_16M; + break; + case MB(32): + val |= FENCE_SIZE_32M; + break; + case MB(64): + val |= FENCE_SIZE_64M; + break; + default: + fprintf(stderr, + "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); + return; + } + } + + if (IS_I9XX(pI830)) + fence_pitch = pitch / 512; + else + fence_pitch = pitch / 128; + + switch (fence_pitch) { + case 1: + val |= FENCE_PITCH_1; + break; + case 2: + val |= FENCE_PITCH_2; + break; + case 4: + val |= FENCE_PITCH_4; + break; + case 8: + val |= FENCE_PITCH_8; + break; + case 16: + val |= FENCE_PITCH_16; + break; + case 32: + val |= FENCE_PITCH_32; + break; + case 64: + val |= FENCE_PITCH_64; + break; + default: + fprintf(stderr, + "SetFence: %d: illegal pitch (%d)\n", nr, pitch); + return; + } + + pI830->Fence[nr] = val; +} + +static Bool +MakeTiles(const DRIDriverContext *ctx, I830Rec *pI830, I830MemRange *pMem) +{ + int pitch, ntiles, i; + + pitch = pMem->Pitch * ctx->cpp; + /* + * Simply try to break the region up into at most four pieces of size + * equal to the alignment. + */ + ntiles = ROUND_TO(pMem->Size, pMem->Alignment) / pMem->Alignment; + if (ntiles >= 4) { + return FALSE; + } + + for (i = 0; i < ntiles; i++, nextTile++) { + SetFence(ctx, pI830, nextTile, pMem->Start + i * pMem->Alignment, + pitch, pMem->Alignment); + } + return TRUE; +} + +static void I830SetupMemoryTiling(const DRIDriverContext *ctx, I830Rec *pI830) +{ + int i; + + /* Clear out */ + for (i = 0; i < 8; i++) + pI830->Fence[i] = 0; + + nextTile = 0; + + if (pI830->BackBuffer.Alignment >= KB(512)) { + if (MakeTiles(ctx, pI830, &(pI830->BackBuffer))) { + fprintf(stderr, + "Activating tiled memory for the back buffer.\n"); + } else { + fprintf(stderr, + "MakeTiles failed for the back buffer.\n"); + pI830->allowPageFlip = FALSE; + } + } + + if (pI830->DepthBuffer.Alignment >= KB(512)) { + if (MakeTiles(ctx, pI830, &(pI830->DepthBuffer))) { + fprintf(stderr, + "Activating tiled memory for the depth buffer.\n"); + } else { + fprintf(stderr, + "MakeTiles failed for the depth buffer.\n"); + } + } + + return; +} + +static int I830DetectMemory(const DRIDriverContext *ctx, I830Rec *pI830) +{ + struct pci_device host_bridge, ig_dev; + uint32_t gmch_ctrl; + int memsize = 0; + int range; + uint32_t aper_size; + uint32_t membase2 = 0; + + memset(&host_bridge, 0, sizeof(host_bridge)); + memset(&ig_dev, 0, sizeof(ig_dev)); + + ig_dev.dev = 2; + + pci_device_cfg_read_u32(&host_bridge, &gmch_ctrl, I830_GMCH_CTRL); + + if (IS_I830(pI830) || IS_845G(pI830)) { + if ((gmch_ctrl & I830_GMCH_MEM_MASK) == I830_GMCH_MEM_128M) { + aper_size = 0x80000000; + } else { + aper_size = 0x40000000; + } + } else { + if (IS_I9XX(pI830)) { + int ret; + ret = pci_device_cfg_read_u32(&ig_dev, &membase2, 0x18); + if (membase2 & 0x08000000) + aper_size = 0x8000000; + else + aper_size = 0x10000000; + + fprintf(stderr,"aper size is %08X %08x %d\n", aper_size, membase2, ret); + } else + aper_size = 0x8000000; + } + + pI830->aper_size = aper_size; + + + /* We need to reduce the stolen size, by the GTT and the popup. + * The GTT varying according the the FbMapSize and the popup is 4KB */ + range = (ctx->shared.fbSize / (1024*1024)) + 4; + + if (IS_I85X(pI830) || IS_I865G(pI830) || IS_I9XX(pI830)) { + switch (gmch_ctrl & I830_GMCH_GMS_MASK) { + case I855_GMCH_GMS_STOLEN_1M: + memsize = MB(1) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_4M: + memsize = MB(4) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_8M: + memsize = MB(8) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_16M: + memsize = MB(16) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_32M: + memsize = MB(32) - KB(range); + break; + case I915G_GMCH_GMS_STOLEN_48M: + if (IS_I9XX(pI830)) + memsize = MB(48) - KB(range); + break; + case I915G_GMCH_GMS_STOLEN_64M: + if (IS_I9XX(pI830)) + memsize = MB(64) - KB(range); + break; + } + } else { + switch (gmch_ctrl & I830_GMCH_GMS_MASK) { + case I830_GMCH_GMS_STOLEN_512: + memsize = KB(512) - KB(range); + break; + case I830_GMCH_GMS_STOLEN_1024: + memsize = MB(1) - KB(range); + break; + case I830_GMCH_GMS_STOLEN_8192: + memsize = MB(8) - KB(range); + break; + case I830_GMCH_GMS_LOCAL: + memsize = 0; + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "Local memory found, but won't be used.\n"); + break; + } + } + if (memsize > 0) { + fprintf(stderr, + "detected %d kB stolen memory.\n", memsize / 1024); + } else { + fprintf(stderr, + "no video memory detected.\n"); + } + return memsize; +} + +static int AgpInit(const DRIDriverContext *ctx, I830Rec *info) +{ + unsigned long mode = 0x4; + + if (drmAgpAcquire(ctx->drmFD) < 0) { + fprintf(stderr, "[gart] AGP not available\n"); + return 0; + } + + if (drmAgpEnable(ctx->drmFD, mode) < 0) { + fprintf(stderr, "[gart] AGP not enabled\n"); + drmAgpRelease(ctx->drmFD); + return 0; + } + else + fprintf(stderr, "[gart] AGP enabled at %dx\n", ctx->agpmode); + + return 1; +} + +/* + * Allocate memory from the given pool. Grow the pool if needed and if + * possible. + */ +static unsigned long +AllocFromPool(const DRIDriverContext *ctx, I830Rec *pI830, + I830MemRange *result, I830MemPool *pool, + long size, unsigned long alignment, int flags) +{ + long needed, start, end; + + if (!result || !pool || !size) + return 0; + + /* Calculate how much space is needed. */ + if (alignment <= GTT_PAGE_SIZE) + needed = size; + else { + start = ROUND_TO(pool->Free.Start, alignment); + end = ROUND_TO(start + size, alignment); + needed = end - pool->Free.Start; + } + if (needed > pool->Free.Size) { + return 0; + } + + result->Start = ROUND_TO(pool->Free.Start, alignment); + pool->Free.Start += needed; + result->End = pool->Free.Start; + + pool->Free.Size = pool->Free.End - pool->Free.Start; + result->Size = result->End - result->Start; + result->Pool = pool; + result->Alignment = alignment; + return needed; +} + +static unsigned long AllocFromAGP(const DRIDriverContext *ctx, I830Rec *pI830, long size, unsigned long alignment, I830MemRange *result) +{ + unsigned long start, end; + unsigned long newApStart, newApEnd; + int ret; + if (!result || !size) + return 0; + + if (!alignment) + alignment = 4; + + start = ROUND_TO(pI830->MemoryAperture.Start, alignment); + end = ROUND_TO(start + size, alignment); + newApStart = end; + newApEnd = pI830->MemoryAperture.End; + + ret=drmAgpAlloc(ctx->drmFD, size, 0, &(result->Physical), (drm_handle_t *)&(result->Key)); + + if (ret) + { + fprintf(stderr,"drmAgpAlloc failed %d\n", ret); + return 0; + } + pI830->allocatedMemory += size; + pI830->MemoryAperture.Start = newApStart; + pI830->MemoryAperture.End = newApEnd; + pI830->MemoryAperture.Size = newApEnd - newApStart; + // pI830->FreeMemory -= size; + result->Start = start; + result->End = start + size; + result->Size = size; + result->Offset = start; + result->Alignment = alignment; + result->Pool = NULL; + + return size; +} + +unsigned long +I830AllocVidMem(const DRIDriverContext *ctx, I830Rec *pI830, + I830MemRange *result, I830MemPool *pool, long size, + unsigned long alignment, int flags) +{ + unsigned long ret; + + if (!result) + return 0; + + /* Make sure these are initialised. */ + result->Size = 0; + result->Key = -1; + + if (!size) { + return 0; + } + + if (pool->Free.Size < size) { + ret = AllocFromAGP(ctx, pI830, size, alignment, result); + } + else { + ret = AllocFromPool(ctx, pI830, result, pool, size, alignment, flags); + if (ret == 0) + ret = AllocFromAGP(ctx, pI830, size, alignment, result); + } + return ret; +} + +static Bool BindAgpRange(const DRIDriverContext *ctx, I830MemRange *mem) +{ + if (!mem) + return FALSE; + + if (mem->Key == -1) + return TRUE; + + return !drmAgpBind(ctx->drmFD, mem->Key, mem->Offset); +} + +/* simple memory allocation routines needed */ +/* put ring buffer in low memory */ +/* need to allocate front, back, depth buffers aligned correctly, + allocate ring buffer, +*/ + +/* */ +static Bool +I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned long size, ret; + unsigned long lines, lineSize, align; + + /* allocate ring buffer */ + memset(pI830->LpRing, 0, sizeof(I830RingBuffer)); + pI830->LpRing->mem.Key = -1; + + size = PRIMARY_RINGBUFFER_SIZE; + + ret = I830AllocVidMem(ctx, pI830, &pI830->LpRing->mem, &pI830->StolenPool, size, 0x1000, 0); + + if (ret != size) + { + fprintf(stderr,"unable to allocate ring buffer %ld\n", ret); + return FALSE; + } + + pI830->LpRing->tail_mask = pI830->LpRing->mem.Size - 1; + + + /* allocate front buffer */ + memset(&(pI830->FrontBuffer), 0, sizeof(pI830->FrontBuffer)); + pI830->FrontBuffer.Key = -1; + pI830->FrontBuffer.Pitch = ctx->shared.virtualWidth; + + align = KB(512); + + lineSize = ctx->shared.virtualWidth * ctx->cpp; + lines = (ctx->shared.virtualHeight + 15) / 16 * 16; + size = lineSize * lines; + size = ROUND_TO_PAGE(size); + + align = GetBestTileAlignment(size); + + ret = I830AllocVidMem(ctx, pI830, &pI830->FrontBuffer, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate front buffer %ld\n", ret); + return FALSE; + } + + memset(&(pI830->BackBuffer), 0, sizeof(pI830->BackBuffer)); + pI830->BackBuffer.Key = -1; + pI830->BackBuffer.Pitch = ctx->shared.virtualWidth; + + ret = I830AllocVidMem(ctx, pI830, &pI830->BackBuffer, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate back buffer %ld\n", ret); + return FALSE; + } + + memset(&(pI830->DepthBuffer), 0, sizeof(pI830->DepthBuffer)); + pI830->DepthBuffer.Key = -1; + pI830->DepthBuffer.Pitch = ctx->shared.virtualWidth; + + ret = I830AllocVidMem(ctx, pI830, &pI830->DepthBuffer, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate depth buffer %ld\n", ret); + return FALSE; + } + + memset(&(pI830->ContextMem), 0, sizeof(pI830->ContextMem)); + pI830->ContextMem.Key = -1; + size = KB(32); + + ret = I830AllocVidMem(ctx, pI830, &pI830->ContextMem, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate context buffer %ld\n", ret); + return FALSE; + } + +#if 0 + memset(&(pI830->TexMem), 0, sizeof(pI830->TexMem)); + pI830->TexMem.Key = -1; + + size = 32768 * 1024; + ret = AllocFromAGP(ctx, pI830, size, align, &pI830->TexMem); + if (ret < size) + { + fprintf(stderr,"unable to allocate texture memory %ld\n", ret); + return FALSE; + } +#endif + + return TRUE; +} + +static Bool +I830BindMemory(const DRIDriverContext *ctx, I830Rec *pI830) +{ + if (!BindAgpRange(ctx, &pI830->LpRing->mem)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->FrontBuffer)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->BackBuffer)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->DepthBuffer)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->ContextMem)) + return FALSE; +#if 0 + if (!BindAgpRange(ctx, &pI830->TexMem)) + return FALSE; +#endif + return TRUE; +} + +static void SetupDRIMM(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned long aperEnd = ROUND_DOWN_TO(pI830->aper_size, GTT_PAGE_SIZE) / GTT_PAGE_SIZE; + unsigned long aperStart = ROUND_TO(pI830->aper_size - KB(32768), GTT_PAGE_SIZE) / GTT_PAGE_SIZE; + + fprintf(stderr, "aper size is %08X\n", ctx->shared.fbSize); + if (drmMMInit(ctx->drmFD, aperStart, aperEnd - aperStart, DRM_BO_MEM_TT)) { + fprintf(stderr, + "DRM MM Initialization Failed\n"); + } else { + fprintf(stderr, + "DRM MM Initialized at offset 0x%lx length %d page\n", aperStart, aperEnd-aperStart); + } + +} + +static Bool +I830CleanupDma(const DRIDriverContext *ctx) +{ + drmI830Init info; + + memset(&info, 0, sizeof(drmI830Init)); + info.func = I830_CLEANUP_DMA; + + if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, + &info, sizeof(drmI830Init))) { + fprintf(stderr, "I830 Dma Cleanup Failed\n"); + return FALSE; + } + + return TRUE; +} + +static Bool +I830InitDma(const DRIDriverContext *ctx, I830Rec *pI830) +{ + I830RingBuffer *ring = pI830->LpRing; + drmI830Init info; + + memset(&info, 0, sizeof(drmI830Init)); + info.func = I830_INIT_DMA; + + info.ring_start = ring->mem.Start + pI830->LinearAddr; + info.ring_end = ring->mem.End + pI830->LinearAddr; + info.ring_size = ring->mem.Size; + + info.mmio_offset = (unsigned int)ctx->MMIOStart; + + info.sarea_priv_offset = sizeof(drm_sarea_t); + + info.front_offset = pI830->FrontBuffer.Start; + info.back_offset = pI830->BackBuffer.Start; + info.depth_offset = pI830->DepthBuffer.Start; + info.w = ctx->shared.virtualWidth; + info.h = ctx->shared.virtualHeight; + info.pitch = ctx->shared.virtualWidth; + info.back_pitch = pI830->BackBuffer.Pitch; + info.depth_pitch = pI830->DepthBuffer.Pitch; + info.cpp = ctx->cpp; + + if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, + &info, sizeof(drmI830Init))) { + fprintf(stderr, + "I830 Dma Initialization Failed\n"); + return FALSE; + } + + return TRUE; +} + +static int I830CheckDRMVersion( const DRIDriverContext *ctx, + I830Rec *pI830 ) +{ + drmVersionPtr version; + + version = drmGetVersion(ctx->drmFD); + + if (version) { + int req_minor, req_patch; + + req_minor = 4; + req_patch = 0; + + if (version->version_major != 1 || + version->version_minor < req_minor || + (version->version_minor == req_minor && + version->version_patchlevel < req_patch)) { + /* Incompatible drm version */ + fprintf(stderr, + "[dri] I830DRIScreenInit failed because of a version " + "mismatch.\n" + "[dri] i915.o kernel module version is %d.%d.%d " + "but version 1.%d.%d or newer is needed.\n" + "[dri] Disabling DRI.\n", + version->version_major, + version->version_minor, + version->version_patchlevel, + req_minor, + req_patch); + drmFreeVersion(version); + return 0; + } + + pI830->drmMinor = version->version_minor; + drmFreeVersion(version); + } + return 1; +} + +static void +I830SetRingRegs(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned int itemp; + unsigned char *MMIO = ctx->MMIOAddress; + + OUTREG(LP_RING + RING_LEN, 0); + OUTREG(LP_RING + RING_TAIL, 0); + OUTREG(LP_RING + RING_HEAD, 0); + + if ((long)(pI830->LpRing->mem.Start & I830_RING_START_MASK) != + pI830->LpRing->mem.Start) { + fprintf(stderr, + "I830SetRingRegs: Ring buffer start (%lx) violates its " + "mask (%x)\n", pI830->LpRing->mem.Start, I830_RING_START_MASK); + } + /* Don't care about the old value. Reserved bits must be zero anyway. */ + itemp = pI830->LpRing->mem.Start & I830_RING_START_MASK; + OUTREG(LP_RING + RING_START, itemp); + + if (((pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES) != + pI830->LpRing->mem.Size - 4096) { + fprintf(stderr, + "I830SetRingRegs: Ring buffer size - 4096 (%lx) violates its " + "mask (%x)\n", pI830->LpRing->mem.Size - 4096, + I830_RING_NR_PAGES); + } + /* Don't care about the old value. Reserved bits must be zero anyway. */ + itemp = (pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES; + itemp |= (RING_NO_REPORT | RING_VALID); + OUTREG(LP_RING + RING_LEN, itemp); + + pI830->LpRing->head = INREG(LP_RING + RING_HEAD) & I830_HEAD_MASK; + pI830->LpRing->tail = INREG(LP_RING + RING_TAIL); + pI830->LpRing->space = pI830->LpRing->head - (pI830->LpRing->tail + 8); + if (pI830->LpRing->space < 0) + pI830->LpRing->space += pI830->LpRing->mem.Size; + + SetFenceRegs(ctx, pI830); + + /* RESET THE DISPLAY PIPE TO POINT TO THE FRONTBUFFER - hacky + hacky hacky */ + OUTREG(DSPABASE, pI830->FrontBuffer.Start + pI830->LinearAddr); + +} + +static Bool +I830SetParam(const DRIDriverContext *ctx, int param, int value) +{ + drmI830SetParam sp; + + memset(&sp, 0, sizeof(sp)); + sp.param = param; + sp.value = value; + + if (drmCommandWrite(ctx->drmFD, DRM_I830_SETPARAM, &sp, sizeof(sp))) { + fprintf(stderr, "I830 SetParam Failed\n"); + return FALSE; + } + + return TRUE; +} + +static Bool +I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ + fprintf(stderr, + "[drm] Mapping front buffer\n"); + + if (drmAddMap(ctx->drmFD, + (drm_handle_t)(sarea->front_offset + pI830->LinearAddr), + sarea->front_size, + DRM_FRAME_BUFFER, /*DRM_AGP,*/ + 0, + &sarea->front_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(front_handle) failed. Disabling DRI\n"); + return FALSE; + } + ctx->shared.hFrameBuffer = sarea->front_handle; + ctx->shared.fbSize = sarea->front_size; + fprintf(stderr, "[drm] Front Buffer = 0x%08x\n", + sarea->front_handle); + + if (drmAddMap(ctx->drmFD, + (drm_handle_t)(sarea->back_offset), + sarea->back_size, DRM_AGP, 0, + &sarea->back_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(back_handle) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] Back Buffer = 0x%08x\n", + sarea->back_handle); + + if (drmAddMap(ctx->drmFD, + (drm_handle_t)sarea->depth_offset, + sarea->depth_size, DRM_AGP, 0, + &sarea->depth_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(depth_handle) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] Depth Buffer = 0x%08x\n", + sarea->depth_handle); + +#if 0 + if (drmAddMap(ctx->drmFD, + (drm_handle_t)sarea->tex_offset, + sarea->tex_size, DRM_AGP, 0, + &sarea->tex_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(tex_handle) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] textures = 0x%08x\n", + sarea->tex_handle); +#endif + return TRUE; +} + + +static void +I830DRIUnmapScreenRegions(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ +#if 1 + if (sarea->front_handle) { + drmRmMap(ctx->drmFD, sarea->front_handle); + sarea->front_handle = 0; + } +#endif + if (sarea->back_handle) { + drmRmMap(ctx->drmFD, sarea->back_handle); + sarea->back_handle = 0; + } + if (sarea->depth_handle) { + drmRmMap(ctx->drmFD, sarea->depth_handle); + sarea->depth_handle = 0; + } + if (sarea->tex_handle) { + drmRmMap(ctx->drmFD, sarea->tex_handle); + sarea->tex_handle = 0; + } +} + +static Bool +I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ + if (drmAddMap(ctx->drmFD, + (drm_handle_t)pI830->LpRing->mem.Start, + pI830->LpRing->mem.Size, DRM_AGP, 0, + &pI830->ring_map) < 0) { + fprintf(stderr, + "[drm] drmAddMap(ring_map) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] ring buffer = 0x%08x\n", + pI830->ring_map); + + if (I830InitDma(ctx, pI830) == FALSE) { + return FALSE; + } + + /* init to zero to be safe */ + + I830DRIMapScreenRegions(ctx, pI830, sarea); + SetupDRIMM(ctx, pI830); + + if (ctx->pciDevice != PCI_CHIP_845_G && + ctx->pciDevice != PCI_CHIP_I830_M) { + I830SetParam(ctx, I830_SETPARAM_USE_MI_BATCHBUFFER_START, 1 ); + } + + /* Okay now initialize the dma engine */ + { + pI830->irq = drmGetInterruptFromBusID(ctx->drmFD, + ctx->pciBus, + ctx->pciDevice, + ctx->pciFunc); + + if (drmCtlInstHandler(ctx->drmFD, pI830->irq)) { + fprintf(stderr, + "[drm] failure adding irq handler\n"); + pI830->irq = 0; + return FALSE; + } + else + fprintf(stderr, + "[drm] dma control initialized, using IRQ %d\n", + pI830->irq); + } + + fprintf(stderr, "[dri] visual configs initialized\n"); + + return TRUE; +} + +static Bool +I830ClearScreen(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ + /* need to drmMap front and back buffers and zero them */ + drmAddress map_addr; + int ret; + + ret = drmMap(ctx->drmFD, + sarea->front_handle, + sarea->front_size, + &map_addr); + + if (ret) + { + fprintf(stderr, "Unable to map front buffer\n"); + return FALSE; + } + + drimemsetio((char *)map_addr, + 0, + sarea->front_size); + drmUnmap(map_addr, sarea->front_size); + + + ret = drmMap(ctx->drmFD, + sarea->back_handle, + sarea->back_size, + &map_addr); + + if (ret) + { + fprintf(stderr, "Unable to map back buffer\n"); + return FALSE; + } + + drimemsetio((char *)map_addr, + 0, + sarea->back_size); + drmUnmap(map_addr, sarea->back_size); + + return TRUE; +} + +static Bool +I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) + +{ + I830DRIPtr pI830DRI; + drmI830Sarea *pSAREAPriv; + int err; + + drm_page_size = getpagesize(); + + pI830->registerSize = ctx->MMIOSize; + /* This is a hack for now. We have to have more than a 4k page here + * because of the size of the state. However, the state should be + * in a per-context mapping. This will be added in the Mesa 3.5 port + * of the I830 driver. + */ + ctx->shared.SAREASize = SAREA_MAX; + + /* Note that drmOpen will try to load the kernel module, if needed. */ + ctx->drmFD = drmOpen("i915", NULL ); + if (ctx->drmFD < 0) { + fprintf(stderr, "[drm] drmOpen failed\n"); + return 0; + } + + if ((err = drmSetBusid(ctx->drmFD, ctx->pciBusID)) < 0) { + fprintf(stderr, "[drm] drmSetBusid failed (%d, %s), %s\n", + ctx->drmFD, ctx->pciBusID, strerror(-err)); + return 0; + } + + if (drmAddMap( ctx->drmFD, + 0, + ctx->shared.SAREASize, + DRM_SHM, + DRM_CONTAINS_LOCK, + &ctx->shared.hSAREA) < 0) + { + fprintf(stderr, "[drm] drmAddMap failed\n"); + return 0; + } + + fprintf(stderr, "[drm] added %d byte SAREA at 0x%08x\n", + ctx->shared.SAREASize, ctx->shared.hSAREA); + + if (drmMap( ctx->drmFD, + ctx->shared.hSAREA, + ctx->shared.SAREASize, + (drmAddressPtr)(&ctx->pSAREA)) < 0) + { + fprintf(stderr, "[drm] drmMap failed\n"); + return 0; + + } + + memset(ctx->pSAREA, 0, ctx->shared.SAREASize); + fprintf(stderr, "[drm] mapped SAREA 0x%08x to %p, size %d\n", + ctx->shared.hSAREA, ctx->pSAREA, ctx->shared.SAREASize); + + + if (drmAddMap(ctx->drmFD, + ctx->MMIOStart, + ctx->MMIOSize, + DRM_REGISTERS, + DRM_READ_ONLY, + &pI830->registerHandle) < 0) { + fprintf(stderr, "[drm] drmAddMap mmio failed\n"); + return 0; + } + fprintf(stderr, + "[drm] register handle = 0x%08x\n", pI830->registerHandle); + + + if (!I830CheckDRMVersion(ctx, pI830)) { + return FALSE; + } + + /* Create a 'server' context so we can grab the lock for + * initialization ioctls. + */ + if ((err = drmCreateContext(ctx->drmFD, &ctx->serverContext)) != 0) { + fprintf(stderr, "%s: drmCreateContext failed %d\n", __FUNCTION__, err); + return 0; + } + + DRM_LOCK(ctx->drmFD, ctx->pSAREA, ctx->serverContext, 0); + + /* Initialize the SAREA private data structure */ + pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + + sizeof(drm_sarea_t)); + memset(pSAREAPriv, 0, sizeof(*pSAREAPriv)); + + pI830->StolenMemory.Size = I830DetectMemory(ctx, pI830); + pI830->StolenMemory.Start = 0; + pI830->StolenMemory.End = pI830->StolenMemory.Size; + + pI830->MemoryAperture.Start = pI830->StolenMemory.End; + pI830->MemoryAperture.End = KB(40000); + pI830->MemoryAperture.Size = pI830->MemoryAperture.End - pI830->MemoryAperture.Start; + + pI830->StolenPool.Fixed = pI830->StolenMemory; + pI830->StolenPool.Total = pI830->StolenMemory; + pI830->StolenPool.Free = pI830->StolenPool.Total; + pI830->FreeMemory = pI830->StolenPool.Total.Size; + + if (!AgpInit(ctx, pI830)) + return FALSE; + + if (I830AllocateMemory(ctx, pI830) == FALSE) + { + return FALSE; + } + + if (I830BindMemory(ctx, pI830) == FALSE) + { + return FALSE; + } + + pSAREAPriv->rotated_offset = -1; + pSAREAPriv->rotated_size = 0; + pSAREAPriv->rotated_pitch = ctx->shared.virtualWidth; + + pSAREAPriv->front_offset = pI830->FrontBuffer.Start; + pSAREAPriv->front_size = pI830->FrontBuffer.Size; + pSAREAPriv->width = ctx->shared.virtualWidth; + pSAREAPriv->height = ctx->shared.virtualHeight; + pSAREAPriv->pitch = ctx->shared.virtualWidth; + pSAREAPriv->virtualX = ctx->shared.virtualWidth; + pSAREAPriv->virtualY = ctx->shared.virtualHeight; + pSAREAPriv->back_offset = pI830->BackBuffer.Start; + pSAREAPriv->back_size = pI830->BackBuffer.Size; + pSAREAPriv->depth_offset = pI830->DepthBuffer.Start; + pSAREAPriv->depth_size = pI830->DepthBuffer.Size; +#if 0 + pSAREAPriv->tex_offset = pI830->TexMem.Start; + pSAREAPriv->tex_size = pI830->TexMem.Size; +#endif + pSAREAPriv->log_tex_granularity = pI830->TexGranularity; + + ctx->driverClientMsg = malloc(sizeof(I830DRIRec)); + ctx->driverClientMsgSize = sizeof(I830DRIRec); + pI830DRI = (I830DRIPtr)ctx->driverClientMsg; + pI830DRI->deviceID = pI830->Chipset; + pI830DRI->regsSize = I830_REG_SIZE; + pI830DRI->width = ctx->shared.virtualWidth; + pI830DRI->height = ctx->shared.virtualHeight; + pI830DRI->mem = ctx->shared.fbSize; + pI830DRI->cpp = ctx->cpp; + + pI830DRI->bitsPerPixel = ctx->bpp; + pI830DRI->sarea_priv_offset = sizeof(drm_sarea_t); + + err = I830DRIDoMappings(ctx, pI830, pSAREAPriv); + if (err == FALSE) + return FALSE; + + I830SetupMemoryTiling(ctx, pI830); + + /* Quick hack to clear the front & back buffers. Could also use + * the clear ioctl to do this, but would need to setup hw state + * first. + */ + I830ClearScreen(ctx, pI830, pSAREAPriv); + + I830SetRingRegs(ctx, pI830); + + return TRUE; +} + + +/** + * \brief Validate the fbdev mode. + * + * \param ctx display handle. + * + * \return one on success, or zero on failure. + * + * Saves some registers and returns 1. + * + * \sa radeonValidateMode(). + */ +static int i830ValidateMode( const DRIDriverContext *ctx ) +{ + return 1; +} + +/** + * \brief Examine mode returned by fbdev. + * + * \param ctx display handle. + * + * \return one on success, or zero on failure. + * + * Restores registers that fbdev has clobbered and returns 1. + * + * \sa i810ValidateMode(). + */ +static int i830PostValidateMode( const DRIDriverContext *ctx ) +{ + I830Rec *pI830 = ctx->driverPrivate; + + I830SetRingRegs(ctx, pI830); + return 1; +} + + +/** + * \brief Initialize the framebuffer device mode + * + * \param ctx display handle. + * + * \return one on success, or zero on failure. + * + * Fills in \p info with some default values and some information from \p ctx + * and then calls I810ScreenInit() for the screen initialization. + * + * Before exiting clears the framebuffer memory accessing it directly. + */ +static int i830InitFBDev( DRIDriverContext *ctx ) +{ + I830Rec *pI830 = calloc(1, sizeof(I830Rec)); + int i; + + { + int dummy = ctx->shared.virtualWidth; + + switch (ctx->bpp / 8) { + case 1: dummy = (ctx->shared.virtualWidth + 127) & ~127; break; + case 2: dummy = (ctx->shared.virtualWidth + 31) & ~31; break; + case 3: + case 4: dummy = (ctx->shared.virtualWidth + 15) & ~15; break; + } + + ctx->shared.virtualWidth = dummy; + ctx->shared.Width = ctx->shared.virtualWidth; + } + + + for (i = 0; pitches[i] != 0; i++) { + if (pitches[i] >= ctx->shared.virtualWidth) { + ctx->shared.virtualWidth = pitches[i]; + break; + } + } + + ctx->driverPrivate = (void *)pI830; + + pI830->LpRing = calloc(1, sizeof(I830RingBuffer)); + pI830->Chipset = ctx->chipset; + pI830->LinearAddr = ctx->FBStart; + + if (!I830ScreenInit( ctx, pI830 )) + return 0; + + + return 1; +} + + +/** + * \brief The screen is being closed, so clean up any state and free any + * resources used by the DRI. + * + * \param ctx display handle. + * + * Unmaps the SAREA, closes the DRM device file descriptor and frees the driver + * private data. + */ +static void i830HaltFBDev( DRIDriverContext *ctx ) +{ + drmI830Sarea *pSAREAPriv; + I830Rec *pI830 = ctx->driverPrivate; + + if (pI830->irq) { + drmCtlUninstHandler(ctx->drmFD); + pI830->irq = 0; } + + I830CleanupDma(ctx); + + pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + + sizeof(drm_sarea_t)); + + I830DRIUnmapScreenRegions(ctx, pI830, pSAREAPriv); + drmUnmap( ctx->pSAREA, ctx->shared.SAREASize ); + drmClose(ctx->drmFD); + + if (ctx->driverPrivate) { + free(ctx->driverPrivate); + ctx->driverPrivate = 0; + } +} + + +extern void i810NotifyFocus( int ); + +/** + * \brief Exported driver interface for Mini GLX. + * + * \sa DRIDriverRec. + */ +const struct DRIDriverRec __driDriver = { + i830ValidateMode, + i830PostValidateMode, + i830InitFBDev, + i830HaltFBDev, + NULL,//I830EngineShutdown, + NULL, //I830EngineRestore, +#ifndef _EMBEDDED + 0, +#else + i810NotifyFocus, +#endif +}; -- cgit v1.2.3 From 33891b64a9a00ddfd7b9c57a2020e83449af62e5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 13:42:23 -0600 Subject: re-fix stencil addressing bug --- src/mesa/drivers/dri/i915pipe/intel_surface.c | 17 +++-------------- 1 file changed, 3 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_surface.c b/src/mesa/drivers/dri/i915pipe/intel_surface.c index 043c5aa5fe..e64e5f8a32 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_surface.c +++ b/src/mesa/drivers/dri/i915pipe/intel_surface.c @@ -133,8 +133,9 @@ read_quad_stencil(struct softpipe_surface *sps, /* extract high byte */ ssss[0] = src[0] >> 24; ssss[1] = src[1] >> 24; - ssss[2] = src[-sps->surface.width] >> 24; - ssss[3] = src[-sps->surface.width + 1] >> 24; + src -= sps->surface.stride; + ssss[2] = src[0] >> 24; + ssss[3] = src[1] >> 24; } static void @@ -245,15 +246,3 @@ intel_new_surface(GLuint intFormat) return &sps->surface; } - - - -struct pipe_surface * -xmesa_get_stencil_surface(GLcontext *ctx) -{ - /* XXX fix */ - return NULL; -} - - - -- cgit v1.2.3 From 2f245bce420c7a6c6928c4927d0f9a5701cde17f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 15:44:50 -0600 Subject: Lift region-related functions up to the pipe interface. Some of these functions probably should be driver-private. Note: intel_buffer_object is in p_state.h and should be fixed/removed. There are just a few i915 dependencies in intel_region.c --- src/mesa/drivers/dri/i915pipe/intel_blit.c | 15 +- .../drivers/dri/i915pipe/intel_buffer_objects.c | 7 +- .../drivers/dri/i915pipe/intel_buffer_objects.h | 6 +- src/mesa/drivers/dri/i915pipe/intel_buffers.c | 11 +- src/mesa/drivers/dri/i915pipe/intel_buffers.h | 4 +- src/mesa/drivers/dri/i915pipe/intel_context.c | 37 +++-- src/mesa/drivers/dri/i915pipe/intel_context.h | 7 +- src/mesa/drivers/dri/i915pipe/intel_depthstencil.c | 10 +- src/mesa/drivers/dri/i915pipe/intel_fbo.c | 21 +-- src/mesa/drivers/dri/i915pipe/intel_fbo.h | 6 +- src/mesa/drivers/dri/i915pipe/intel_ioctl.c | 1 - src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c | 21 +-- src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h | 14 +- src/mesa/drivers/dri/i915pipe/intel_pixel.c | 2 - src/mesa/drivers/dri/i915pipe/intel_pixel.h | 2 +- src/mesa/drivers/dri/i915pipe/intel_regions.c | 160 +++++++++++++-------- src/mesa/drivers/dri/i915pipe/intel_regions.h | 105 -------------- src/mesa/drivers/dri/i915pipe/intel_screen.c | 28 ++-- src/mesa/drivers/dri/i915pipe/intel_screen.h | 11 +- src/mesa/drivers/dri/i915pipe/intel_span.c | 31 ++-- src/mesa/drivers/dri/i915pipe/intel_state.c | 1 - src/mesa/drivers/dri/i915pipe/intel_surface.c | 6 +- src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 7 +- src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 12 +- src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c | 5 +- src/mesa/pipe/p_context.h | 81 ++++++++++- src/mesa/pipe/p_state.h | 18 +++ 27 files changed, 342 insertions(+), 287 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.c b/src/mesa/drivers/dri/i915pipe/intel_blit.c index 28441f4b83..8e7f294b57 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_blit.c +++ b/src/mesa/drivers/dri/i915pipe/intel_blit.c @@ -26,9 +26,6 @@ **************************************************************************/ -#include -#include - #include "mtypes.h" #include "context.h" #include "enums.h" @@ -39,9 +36,11 @@ #include "intel_context.h" #include "intel_fbo.h" #include "intel_reg.h" -#include "intel_regions.h" #include "vblank.h" +#include "pipe/p_context.h" + + #define FILE_DEBUG_FLAG DEBUG_BLIT /** @@ -87,9 +86,9 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, if (dPriv && dPriv->numClipRects) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - const struct intel_region *frontRegion + const struct pipe_region *frontRegion = intelScreen->front_region; - const struct intel_region *backRegion + const struct pipe_region *backRegion = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); @@ -447,10 +446,10 @@ intelClearWithBlit(GLcontext * ctx, GLbitfield mask) const GLbitfield bufBit = 1 << buf; if ((clearMask & bufBit) && !(bufBit & skipBuffers)) { /* OK, clear this renderbuffer */ - struct intel_region *irb_region = + struct pipe_region *irb_region = intel_get_rb_region(fb, buf); struct _DriBufferObject *write_buffer = - intel_region_buffer(intel->intelScreen, irb_region, + intel->pipe->region_buffer(intel->pipe, irb_region, all ? INTEL_WRITE_FULL : INTEL_WRITE_PART); diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c index 91c45ad95b..eee5fa9bf4 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c +++ b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c @@ -32,9 +32,12 @@ #include "intel_context.h" #include "intel_buffer_objects.h" -#include "intel_regions.h" #include "dri_bufmgr.h" +#include "pipe/p_state.h" +#include "pipe/p_context.h" + + /** * There is some duplication between mesa's bufferobjects and our * bufmgr buffers. Both have an integer handle and a hashtable to @@ -89,7 +92,7 @@ intel_bufferobj_cow(struct intel_context *intel, struct intel_buffer_object *intel_obj) { assert(intel_obj->region); - intel_region_cow(intel->intelScreen, intel_obj->region); + intel->pipe->region_cow(intel->pipe, intel_obj->region); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h index afe9b2f7cf..a1f63c147b 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h +++ b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h @@ -31,7 +31,7 @@ #include "mtypes.h" struct intel_context; -struct intel_region; +struct pipe_region; struct gl_buffer_object; @@ -43,7 +43,7 @@ struct intel_buffer_object struct gl_buffer_object Base; struct _DriBufferObject *buffer; /* the low-level buffer manager's buffer handle */ - struct intel_region *region; /* Is there a zero-copy texture + struct pipe_region *region; /* Is there a zero-copy texture associated with this (pixel) buffer object? */ }; @@ -75,7 +75,7 @@ intel_buffer_object(struct gl_buffer_object *obj) return NULL; } -/* Helpers for zerocopy image uploads. See also intel_regions.h: +/* Helpers for zerocopy image uploads. See also pipe_regions.h: */ void intel_bufferobj_cow(struct intel_context *intel, struct intel_buffer_object *intel_obj); diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.c b/src/mesa/drivers/dri/i915pipe/intel_buffers.c index cc9a4262df..c03c009a3a 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.c +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.c @@ -31,7 +31,6 @@ #include "intel_buffers.h" #include "intel_depthstencil.h" #include "intel_fbo.h" -#include "intel_regions.h" #include "intel_batchbuffer.h" #include "intel_reg.h" #include "context.h" @@ -96,7 +95,7 @@ intel_intersect_cliprects(drm_clip_rect_t * dst, /** * Return pointer to current color drawing region, or NULL. */ -struct intel_region * +struct pipe_region * intel_drawbuf_region(struct intel_context *intel) { struct intel_renderbuffer *irbColor = @@ -110,7 +109,7 @@ intel_drawbuf_region(struct intel_context *intel) /** * Return pointer to current color reading region, or NULL. */ -struct intel_region * +struct pipe_region * intel_readbuf_region(struct intel_context *intel) { struct intel_renderbuffer *irb @@ -325,7 +324,7 @@ intelClear(GLcontext *ctx, GLbitfield mask) /* HW stencil */ if (mask & BUFFER_BIT_STENCIL) { - const struct intel_region *stencilRegion + const struct pipe_region *stencilRegion = intel_get_rb_region(fb, BUFFER_STENCIL); if (stencilRegion) { /* have hw stencil */ @@ -369,8 +368,10 @@ intelClear(GLcontext *ctx, GLbitfield mask) if (blit_mask) intelClearWithBlit(ctx, blit_mask); +#if 1 if (swrast_mask | tri_mask) _swrast_Clear(ctx, swrast_mask | tri_mask); +#endif } @@ -625,7 +626,7 @@ void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) { struct intel_context *intel = intel_context(ctx); - struct intel_region *colorRegion, *depthRegion = NULL; + struct pipe_region *colorRegion, *depthRegion = NULL; struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; if (!fb) { diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.h b/src/mesa/drivers/dri/i915pipe/intel_buffers.h index 13d1a15ffb..5834e39501 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.h +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.h @@ -38,9 +38,9 @@ intel_intersect_cliprects(drm_clip_rect_t * dest, const drm_clip_rect_t * a, const drm_clip_rect_t * b); -extern struct intel_region *intel_readbuf_region(struct intel_context *intel); +extern struct pipe_region *intel_readbuf_region(struct intel_context *intel); -extern struct intel_region *intel_drawbuf_region(struct intel_context *intel); +extern struct pipe_region *intel_drawbuf_region(struct intel_context *intel); extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index 395b0f63be..c8129c9e5f 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -55,12 +55,12 @@ #include "intel_batchbuffer.h" #include "intel_blit.h" #include "intel_pixel.h" -#include "intel_regions.h" #include "intel_buffer_objects.h" #include "intel_fbo.h" #include "pipe/softpipe/sp_context.h" #include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" #include "drirenderbuffer.h" @@ -90,7 +90,7 @@ int INTEL_DEBUG = (0); #include "extension_helper.h" -#define DRIVER_DATE "20061102" +#define DRIVER_DATE "20070731" _glthread_Mutex lockMutex; static GLboolean lockMutexInit = GL_FALSE; @@ -355,13 +355,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, int fthrottle_mode; GLboolean havePools; - DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - havePools = intelCreatePools(intelScreen); - DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - - if (!havePools) - return GL_FALSE; - intelInitDriverFunctions(&functions); if (!_mesa_initialize_context(&intel->ctx, @@ -416,11 +409,31 @@ intelCreateContext(const __GLcontextModes * mesaVis, _tnl_CreateContext(ctx); _swsetup_CreateContext(ctx); - /* Configure swrast to match hardware characteristics: */ _swrast_allow_pixel_fog(ctx, GL_FALSE); _swrast_allow_vertex_fog(ctx, GL_TRUE); + /* + * Pipe-related setup + */ + st_create_context( &intel->ctx, + softpipe_create() ); + + intel->pipe = intel->ctx.st->pipe; + intel->pipe->screen = intelScreen; + intelScreen->pipe = intel->pipe; + intel_init_region_functions(intel->pipe); + + /* + * memory pools + */ + DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + havePools = intelCreatePools(intelScreen); + DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + if (!havePools) + return GL_FALSE; + + /* Dri stuff */ intel->hHWContext = driContextPriv->hHWContext; intel->driFd = sPriv->fd; @@ -497,10 +510,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, } - st_create_context( &intel->ctx, - softpipe_create() ); - - return GL_TRUE; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.h b/src/mesa/drivers/dri/i915pipe/intel_context.h index 154be972b9..9d331e41b5 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.h +++ b/src/mesa/drivers/dri/i915pipe/intel_context.h @@ -48,7 +48,8 @@ #define DV_PF_565 (2<<8) #define DV_PF_8888 (3<<8) -struct intel_region; +struct pipe_context; +struct pipe_region; struct intel_context; struct _DriBufferObject; @@ -121,6 +122,8 @@ struct intel_context { GLcontext ctx; /* the parent class */ + struct pipe_context *pipe; + GLint refcount; GLuint Fallback; GLuint NewGLState; @@ -422,5 +425,7 @@ intel_texture_image(struct gl_texture_image *img) extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer *rb); +extern void intel_init_region_functions(struct pipe_context *pipe); + #endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c b/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c index d269a85a3c..81e09698cc 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c +++ b/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c @@ -38,8 +38,8 @@ #include "intel_context.h" #include "intel_fbo.h" #include "intel_depthstencil.h" -#include "intel_regions.h" +#include "pipe/p_context.h" /** * The GL_EXT_framebuffer_object allows the user to create their own @@ -95,12 +95,12 @@ map_regions(GLcontext * ctx, { struct intel_context *intel = intel_context(ctx); if (depthRb && depthRb->region) { - intel_region_map(intel->intelScreen, depthRb->region); + intel->pipe->region_map(intel->pipe, depthRb->region); depthRb->pfMap = depthRb->region->map; depthRb->pfPitch = depthRb->region->pitch; } if (stencilRb && stencilRb->region) { - intel_region_map(intel->intelScreen, stencilRb->region); + intel->pipe->region_map(intel->pipe, stencilRb->region); stencilRb->pfMap = stencilRb->region->map; stencilRb->pfPitch = stencilRb->region->pitch; } @@ -113,12 +113,12 @@ unmap_regions(GLcontext * ctx, { struct intel_context *intel = intel_context(ctx); if (depthRb && depthRb->region) { - intel_region_unmap(intel->intelScreen, depthRb->region); + intel->pipe->region_unmap(intel->pipe, depthRb->region); depthRb->pfMap = NULL; depthRb->pfPitch = 0; } if (stencilRb && stencilRb->region) { - intel_region_unmap(intel->intelScreen, stencilRb->region); + intel->pipe->region_unmap(intel->pipe, stencilRb->region); stencilRb->pfMap = NULL; stencilRb->pfPitch = 0; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index a2a5c811b1..bac2ef8467 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -40,9 +40,9 @@ #include "intel_depthstencil.h" #include "intel_fbo.h" #include "intel_mipmap_tree.h" -#include "intel_regions.h" #include "intel_span.h" +#include "pipe/p_context.h" #define FILE_DEBUG_FLAG DEBUG_FBO @@ -111,7 +111,7 @@ intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) } -struct intel_region * +struct pipe_region * intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) { struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex); @@ -153,7 +153,7 @@ intel_delete_renderbuffer(struct gl_renderbuffer *rb) } if (intel && irb->region) { - intel_region_release(&irb->region); + intel->pipe->region_release(intel->pipe, &irb->region); } _mesa_free(irb); @@ -265,7 +265,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, /* free old region */ if (irb->region) { - intel_region_release(&irb->region); + intel->pipe->region_release(intel->pipe, &irb->region); } /* allocate new memory region/renderbuffer */ @@ -282,7 +282,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, height, pitch); - irb->region = intel_region_alloc(intel->intelScreen, cpp, pitch, height); + irb->region = intel->pipe->region_alloc(intel->pipe, cpp, pitch, height); if (!irb->region) return GL_FALSE; /* out of memory? */ @@ -382,7 +382,7 @@ intel_new_renderbuffer_fb(GLuint intFormat) irb->Base.surface = intel_new_surface(intFormat); irb->Base.surface->rb = irb; - return &irb->Base; + return irb; } /** @@ -509,6 +509,7 @@ intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage) irb->Base.Delete = intel_delete_renderbuffer; irb->Base.AllocStorage = intel_nop_alloc_storage; + intel_set_span_functions(&irb->Base); irb->RenderToTexture = GL_TRUE; @@ -528,6 +529,7 @@ intel_render_texture(GLcontext * ctx, struct gl_framebuffer *fb, struct gl_renderbuffer_attachment *att) { + struct intel_context *intel = intel_context(ctx); struct gl_texture_image *newImage = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); @@ -560,8 +562,8 @@ intel_render_texture(GLcontext * ctx, intel_image = intel_texture_image(newImage); if (irb->region != intel_image->mt->region) { if (irb->region) - intel_region_release(&irb->region); - intel_region_reference(&irb->region, intel_image->mt->region); + intel->pipe->region_release(intel->pipe, &irb->region); + pipe_region_reference(&irb->region, intel_image->mt->region); } /* compute offset of the particular 2D image within the texture region */ @@ -590,13 +592,14 @@ static void intel_finish_render_texture(GLcontext * ctx, struct gl_renderbuffer_attachment *att) { + struct intel_context *intel = intel_context(ctx); struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); DBG("End render texture (tid %x) tex %u\n", _glthread_GetID(), att->Texture->Name); if (irb) { /* just release the region */ - intel_region_release(&irb->region); + intel->pipe->region_release(intel->pipe, &irb->region); } else if (att->Renderbuffer) { /* software fallback */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.h b/src/mesa/drivers/dri/i915pipe/intel_fbo.h index 7dc3967533..0f99a3e98d 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.h +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.h @@ -34,7 +34,7 @@ struct intel_context; -struct intel_region; +struct pipe_region; /** @@ -76,7 +76,7 @@ struct intel_framebuffer struct intel_renderbuffer { struct gl_renderbuffer Base; - struct intel_region *region; + struct pipe_region *region; void *pfMap; /* possibly paged flipped map pointer */ GLuint pfPitch; /* possibly paged flipped pitch */ GLboolean RenderToTexture; /* RTT? */ @@ -115,7 +115,7 @@ extern void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb); /* XXX make inline or macro */ -extern struct intel_region *intel_get_rb_region(struct gl_framebuffer *fb, +extern struct pipe_region *intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex); diff --git a/src/mesa/drivers/dri/i915pipe/intel_ioctl.c b/src/mesa/drivers/dri/i915pipe/intel_ioctl.c index e349a3ecd9..fb5b518903 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_ioctl.c +++ b/src/mesa/drivers/dri/i915pipe/intel_ioctl.c @@ -39,7 +39,6 @@ #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" -#include "intel_regions.h" #include "drm.h" #define FILE_DEBUG_FLAG DEBUG_IOCTL diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c index 88a9277427..6717984f7d 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c @@ -27,9 +27,12 @@ #include "intel_context.h" #include "intel_mipmap_tree.h" -#include "intel_regions.h" #include "enums.h" +#include "pipe/p_state.h" +#include "pipe/p_context.h" + + #define FILE_DEBUG_FLAG DEBUG_MIPTREE static GLenum @@ -100,7 +103,7 @@ intel_miptree_create(struct intel_context *intel, ok = 0; /* TODO */ if (ok) - mt->region = intel_region_alloc(intel->intelScreen, + mt->region = intel->pipe->region_alloc(intel->pipe, mt->cpp, mt->pitch, mt->total_height); if (!mt->region) { @@ -134,7 +137,7 @@ intel_miptree_release(struct intel_context *intel, DBG("%s deleting %p\n", __FUNCTION__, *mt); - intel_region_release(&((*mt)->region)); + intel->pipe->region_release(intel->pipe, &((*mt)->region)); for (i = 0; i < MAX_TEXTURE_LEVELS; i++) if ((*mt)->level[i].image_offset) @@ -271,6 +274,7 @@ intel_miptree_image_map(struct intel_context * intel, GLuint level, GLuint * row_stride, GLuint * image_offsets) { + GLubyte *ptr; DBG("%s \n", __FUNCTION__); if (row_stride) @@ -280,8 +284,9 @@ intel_miptree_image_map(struct intel_context * intel, memcpy(image_offsets, mt->level[level].image_offset, mt->level[level].depth * sizeof(GLuint)); - return (intel_region_map(intel->intelScreen, mt->region) + - intel_miptree_image_offset(mt, face, level)); + ptr = intel->pipe->region_map(intel->pipe, mt->region); + + return ptr + intel_miptree_image_offset(mt, face, level); } void @@ -289,7 +294,7 @@ intel_miptree_image_unmap(struct intel_context *intel, struct intel_mipmap_tree *mt) { DBG("%s\n", __FUNCTION__); - intel_region_unmap(intel->intelScreen, mt->region); + intel->pipe->region_unmap(intel->pipe, mt->region); } @@ -315,7 +320,7 @@ intel_miptree_image_data(struct intel_context *intel, height = dst->level[level].height; if(dst->compressed) height /= 4; - intel_region_data(intel->intelScreen, dst->region, + intel->pipe->region_data(intel->pipe, dst->region, dst_offset + dst_depth_offset[i], /* dst_offset */ 0, 0, /* dstx, dsty */ src, @@ -347,7 +352,7 @@ intel_miptree_image_copy(struct intel_context *intel, if (dst->compressed) height /= 4; for (i = 0; i < depth; i++) { - intel_region_copy(intel->intelScreen, + intel->pipe->region_copy(intel->pipe, dst->region, dst_offset + dst_depth_offset[i], 0, 0, diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h index ecdb7be244..09b2e362fc 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h @@ -28,14 +28,18 @@ #ifndef INTEL_MIPMAP_TREE_H #define INTEL_MIPMAP_TREE_H -#include "intel_regions.h" +#include "intel_context.h" +#include "main/glheader.h" -/* A layer on top of the intel_regions code which adds: +struct pipe_region; + + +/* A layer on top of the pipe_regions code which adds: * * - Code to size and layout a region to hold a set of mipmaps. * - Query to determine if a new image fits in an existing tree. * - More refcounting - * - maybe able to remove refcounting from intel_region? + * - maybe able to remove refcounting from pipe_region? * - ? * * The fixed mipmap layout of intel hardware where one offset @@ -45,7 +49,7 @@ * independent offset. * * In an ideal world, each texture object would be associated with a - * single bufmgr buffer or 2d intel_region, and all the images within + * single bufmgr buffer or 2d pipe_region, and all the images within * the texture object would slot into the tree as they arrive. The * reality can be a little messier, as images can arrive from the user * with sizes that don't fit in the existing tree, or in an order @@ -103,7 +107,7 @@ struct intel_mipmap_tree /* The data is held here: */ - struct intel_region *region; + struct pipe_region *region; /* These are also refcounted: */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_pixel.c b/src/mesa/drivers/dri/i915pipe/intel_pixel.c index 104d288de0..74d4ce7f2a 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_pixel.c +++ b/src/mesa/drivers/dri/i915pipe/intel_pixel.c @@ -25,13 +25,11 @@ * **************************************************************************/ -#include "enums.h" #include "state.h" #include "swrast/swrast.h" #include "intel_context.h" #include "intel_pixel.h" -#include "intel_regions.h" void diff --git a/src/mesa/drivers/dri/i915pipe/intel_pixel.h b/src/mesa/drivers/dri/i915pipe/intel_pixel.h index a6fcf90ce0..73a06c11ec 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_pixel.h +++ b/src/mesa/drivers/dri/i915pipe/intel_pixel.h @@ -36,7 +36,7 @@ GLboolean intel_check_blit_fragment_ops(GLcontext * ctx); GLboolean intel_check_meta_tex_fragment_ops(GLcontext * ctx); -GLboolean intel_check_blit_format(struct intel_region *region, +GLboolean intel_check_blit_format(struct pipe_region *region, GLenum format, GLenum type); diff --git a/src/mesa/drivers/dri/i915pipe/intel_regions.c b/src/mesa/drivers/dri/i915pipe/intel_regions.c index 7d19bd07d3..e95e745adc 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_regions.c +++ b/src/mesa/drivers/dri/i915pipe/intel_regions.c @@ -39,17 +39,29 @@ * last moment. */ +#include "pipe/p_state.h" +#include "pipe/p_context.h" + #include "intel_context.h" -#include "intel_regions.h" #include "intel_blit.h" #include "intel_buffer_objects.h" #include "dri_bufmgr.h" #include "intel_batchbuffer.h" + #define FILE_DEBUG_FLAG DEBUG_REGION -void -intel_region_idle(intelScreenPrivate *intelScreen, struct intel_region *region) + +/** XXX temporary helper */ +static intelScreenPrivate * +pipe_screen(struct pipe_context *pipe) +{ + return (intelScreenPrivate *) pipe->screen; +} + + +static void +intel_region_idle(struct pipe_context *pipe, struct pipe_region *region) { DBG("%s\n", __FUNCTION__); if (region && region->buffer) @@ -58,13 +70,14 @@ intel_region_idle(intelScreenPrivate *intelScreen, struct intel_region *region) /* XXX: Thread safety? */ -GLubyte * -intel_region_map(intelScreenPrivate *intelScreen, struct intel_region *region) +static GLubyte * +intel_region_map(struct pipe_context *pipe, struct pipe_region *region) { DBG("%s\n", __FUNCTION__); if (!region->map_refcount++) { - if (region->pbo) - intel_region_cow(intelScreen, region); + if (region->pbo) { + pipe->region_cow(pipe, region); + } region->map = driBOMap(region->buffer, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); @@ -73,8 +86,8 @@ intel_region_map(intelScreenPrivate *intelScreen, struct intel_region *region) return region->map; } -void -intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *region) +static void +intel_region_unmap(struct pipe_context *pipe, struct pipe_region *region) { DBG("%s\n", __FUNCTION__); if (!--region->map_refcount) { @@ -85,11 +98,12 @@ intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *region) #undef TEST_CACHED_TEXTURES -struct intel_region * -intel_region_alloc(intelScreenPrivate *intelScreen, +static struct pipe_region * +intel_region_alloc(struct pipe_context *pipe, GLuint cpp, GLuint pitch, GLuint height) { - struct intel_region *region = calloc(sizeof(*region), 1); + intelScreenPrivate *intelScreen = pipe_screen(pipe); + struct pipe_region *region = calloc(sizeof(*region), 1); struct intel_context *intel = intelScreenContext(intelScreen); DBG("%s\n", __FUNCTION__); @@ -114,18 +128,8 @@ intel_region_alloc(intelScreenPrivate *intelScreen, return region; } -void -intel_region_reference(struct intel_region **dst, struct intel_region *src) -{ - assert(*dst == NULL); - if (src) { - src->refcount++; - *dst = src; - } -} - -void -intel_region_release(struct intel_region **region) +static void +intel_region_release(struct pipe_context *pipe, struct pipe_region **region) { if (!*region) return; @@ -148,14 +152,15 @@ intel_region_release(struct intel_region **region) } -struct intel_region * -intel_region_create_static(intelScreenPrivate *intelScreen, +static struct pipe_region * +intel_region_create_static(struct pipe_context *pipe, GLuint mem_type, GLuint offset, void *virtual, GLuint cpp, GLuint pitch, GLuint height) { - struct intel_region *region = calloc(sizeof(*region), 1); + intelScreenPrivate *intelScreen = pipe_screen(pipe); + struct pipe_region *region = calloc(sizeof(*region), 1); DBG("%s\n", __FUNCTION__); region->cpp = cpp; @@ -179,14 +184,16 @@ intel_region_create_static(intelScreenPrivate *intelScreen, -void -intel_region_update_static(intelScreenPrivate *intelScreen, - struct intel_region *region, +static void +intel_region_update_static(struct pipe_context *pipe, + struct pipe_region *region, GLuint mem_type, GLuint offset, void *virtual, GLuint cpp, GLuint pitch, GLuint height) { + intelScreenPrivate *intelScreen = pipe_screen(pipe); + DBG("%s\n", __FUNCTION__); region->cpp = cpp; @@ -252,14 +259,15 @@ _mesa_copy_rect(GLubyte * dst, * * Currently always memcpy. */ -void -intel_region_data(intelScreenPrivate *intelScreen, - struct intel_region *dst, +static void +intel_region_data(struct pipe_context *pipe, + struct pipe_region *dst, GLuint dst_offset, GLuint dstx, GLuint dsty, const void *src, GLuint src_pitch, GLuint srcx, GLuint srcy, GLuint width, GLuint height) { + intelScreenPrivate *intelScreen = pipe_screen(pipe); struct intel_context *intel = intelScreenContext(intelScreen); DBG("%s\n", __FUNCTION__); @@ -270,20 +278,20 @@ intel_region_data(intelScreenPrivate *intelScreen, if (dst->pbo) { if (dstx == 0 && dsty == 0 && width == dst->pitch && height == dst->height) - intel_region_release_pbo(intelScreen, dst); + pipe->region_release_pbo(pipe, dst); else - intel_region_cow(intelScreen, dst); + pipe->region_cow(pipe, dst); } LOCK_HARDWARE(intel); - _mesa_copy_rect(intel_region_map(intelScreen, dst) + dst_offset, + _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, dst->cpp, dst->pitch, dstx, dsty, width, height, src, src_pitch, srcx, srcy); - intel_region_unmap(intelScreen, dst); + pipe->region_unmap(pipe, dst); UNLOCK_HARDWARE(intel); @@ -292,15 +300,16 @@ intel_region_data(intelScreenPrivate *intelScreen, /* Copy rectangular sub-regions. Need better logic about when to * push buffers into AGP - will currently do so whenever possible. */ -void -intel_region_copy(intelScreenPrivate *intelScreen, - struct intel_region *dst, +static void +intel_region_copy(struct pipe_context *pipe, + struct pipe_region *dst, GLuint dst_offset, GLuint dstx, GLuint dsty, - struct intel_region *src, + const struct pipe_region *src, GLuint src_offset, GLuint srcx, GLuint srcy, GLuint width, GLuint height) { + intelScreenPrivate *intelScreen = pipe_screen(pipe); struct intel_context *intel = intelScreenContext(intelScreen); DBG("%s\n", __FUNCTION__); @@ -311,9 +320,9 @@ intel_region_copy(intelScreenPrivate *intelScreen, if (dst->pbo) { if (dstx == 0 && dsty == 0 && width == dst->pitch && height == dst->height) - intel_region_release_pbo(intelScreen, dst); + pipe->region_release_pbo(pipe, dst); else - intel_region_cow(intelScreen, dst); + pipe->region_cow(pipe, dst); } assert(src->cpp == dst->cpp); @@ -329,13 +338,14 @@ intel_region_copy(intelScreenPrivate *intelScreen, /* Fill a rectangular sub-region. Need better logic about when to * push buffers into AGP - will currently do so whenever possible. */ -void -intel_region_fill(intelScreenPrivate *intelScreen, - struct intel_region *dst, +static void +intel_region_fill(struct pipe_context *pipe, + struct pipe_region *dst, GLuint dst_offset, GLuint dstx, GLuint dsty, GLuint width, GLuint height, GLuint color) { + intelScreenPrivate *intelScreen = pipe_screen(pipe); struct intel_context *intel = intelScreenContext(intelScreen); DBG("%s\n", __FUNCTION__); @@ -346,9 +356,9 @@ intel_region_fill(intelScreenPrivate *intelScreen, if (dst->pbo) { if (dstx == 0 && dsty == 0 && width == dst->pitch && height == dst->height) - intel_region_release_pbo(intelScreen, dst); + pipe->region_release_pbo(pipe, dst); else - intel_region_cow(intelScreen, dst); + pipe->region_cow(pipe, dst); } intelEmitFillBlit(intel, @@ -360,16 +370,16 @@ intel_region_fill(intelScreenPrivate *intelScreen, /* Attach to a pbo, discarding our data. Effectively zero-copy upload * the pbo's data. */ -void -intel_region_attach_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region, +static void +intel_region_attach_pbo(struct pipe_context *pipe, + struct pipe_region *region, struct intel_buffer_object *pbo) { if (region->pbo == pbo) return; /* If there is already a pbo attached, break the cow tie now. - * Don't call intel_region_release_pbo() as that would + * Don't call pipe_region_release_pbo() as that would * unnecessarily allocate a new buffer we would have to immediately * discard. */ @@ -391,10 +401,11 @@ intel_region_attach_pbo(intelScreenPrivate *intelScreen, /* Break the COW tie to the pbo. The pbo gets to keep the data. */ -void -intel_region_release_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region) +static void +intel_region_release_pbo(struct pipe_context *pipe, + struct pipe_region *region) { + intelScreenPrivate *intelScreen = pipe_screen(pipe); struct intel_context *intel = intelScreenContext(intelScreen); assert(region->buffer == region->pbo->buffer); @@ -415,16 +426,17 @@ intel_region_release_pbo(intelScreenPrivate *intelScreen, /* Break the COW tie to the pbo. Both the pbo and the region end up * with a copy of the data. */ -void -intel_region_cow(intelScreenPrivate *intelScreen, struct intel_region *region) +static void +intel_region_cow(struct pipe_context *pipe, struct pipe_region *region) { + intelScreenPrivate *intelScreen = pipe_screen(pipe); struct intel_context *intel = intelScreenContext(intelScreen); struct intel_buffer_object *pbo = region->pbo; if (intel == NULL) return; - intel_region_release_pbo(intelScreen, region); + pipe->region_release_pbo(pipe, region); assert(region->cpp * region->pitch * region->height == pbo->Base.Size); @@ -465,16 +477,38 @@ intel_region_cow(intelScreenPrivate *intelScreen, struct intel_region *region) } } -struct _DriBufferObject * -intel_region_buffer(intelScreenPrivate *intelScreen, - struct intel_region *region, GLuint flag) +static struct _DriBufferObject * +intel_region_buffer(struct pipe_context *pipe, + struct pipe_region *region, GLuint flag) { if (region->pbo) { if (flag == INTEL_WRITE_PART) - intel_region_cow(intelScreen, region); + pipe->region_cow(pipe, region); else if (flag == INTEL_WRITE_FULL) - intel_region_release_pbo(intelScreen, region); + pipe->region_release_pbo(pipe, region); } return region->buffer; } + + + +void +intel_init_region_functions(struct pipe_context *pipe) +{ + pipe->region_idle = intel_region_idle; + pipe->region_map = intel_region_map; + pipe->region_unmap = intel_region_unmap; + pipe->region_alloc = intel_region_alloc; + pipe->region_release = intel_region_release; + pipe->region_create_static = intel_region_create_static; + pipe->region_update_static = intel_region_update_static; + pipe->region_data = intel_region_data; + pipe->region_copy = intel_region_copy; + pipe->region_fill = intel_region_fill; + pipe->region_cow = intel_region_cow; + pipe->region_attach_pbo = intel_region_attach_pbo; + pipe->region_release_pbo = intel_region_release_pbo; + pipe->region_buffer = intel_region_buffer; +} + diff --git a/src/mesa/drivers/dri/i915pipe/intel_regions.h b/src/mesa/drivers/dri/i915pipe/intel_regions.h index d938c107a4..2a3b5b4025 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_regions.h +++ b/src/mesa/drivers/dri/i915pipe/intel_regions.h @@ -31,111 +31,6 @@ #include "mtypes.h" #include "intel_screen.h" -struct intel_context; -struct intel_buffer_object; -/** - * A layer on top of the bufmgr buffers that adds a few useful things: - * - * - Refcounting for local buffer references. - * - Refcounting for buffer maps - * - Buffer dimensions - pitch and height. - * - Blitter commands for copying 2D regions between buffers. (really???) - */ -struct intel_region -{ - struct _DriBufferObject *buffer; /**< buffer manager's buffer ID */ - GLuint refcount; /**< Reference count for region */ - GLuint cpp; /**< bytes per pixel */ - GLuint pitch; /**< in pixels */ - GLuint height; /**< in pixels */ - GLubyte *map; /**< only non-NULL when region is actually mapped */ - GLuint map_refcount; /**< Reference count for mapping */ - - GLuint draw_offset; /**< Offset of drawing address within the region */ - - struct intel_buffer_object *pbo; /* zero-copy uploads */ -}; - - -/* Allocate a refcounted region. Pointers to regions should only be - * copied by calling intel_reference_region(). - */ -struct intel_region *intel_region_alloc(intelScreenPrivate *intelScreen, - GLuint cpp, - GLuint pitch, GLuint height); - -void intel_region_reference(struct intel_region **dst, - struct intel_region *src); - -void intel_region_release(struct intel_region **ib); - -extern struct intel_region -*intel_region_create_static(intelScreenPrivate *intelScreen, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, - GLuint pitch, GLuint height); -extern void -intel_region_update_static(intelScreenPrivate *intelScreen, - struct intel_region *region, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height); - - -void intel_region_idle(intelScreenPrivate *intelScreen, - struct intel_region *ib); - -/* Map/unmap regions. This is refcounted also: - */ -GLubyte *intel_region_map(intelScreenPrivate *intelScreen, - struct intel_region *ib); - -void intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *ib); - - -/* Upload data to a rectangular sub-region - */ -void intel_region_data(intelScreenPrivate *intelScreen, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - const void *src, GLuint src_stride, - GLuint srcx, GLuint srcy, GLuint width, GLuint height); - -/* Copy rectangular sub-regions - */ -void intel_region_copy(intelScreenPrivate *intelScreen, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - struct intel_region *src, - GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height); - -/* Fill a rectangular sub-region - */ -void intel_region_fill(intelScreenPrivate *intelScreen, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - GLuint width, GLuint height, GLuint color); - -/* Helpers for zerocopy uploads, particularly texture image uploads: - */ -void intel_region_attach_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region, - struct intel_buffer_object *pbo); -void intel_region_release_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region); -void intel_region_cow(intelScreenPrivate *intelScreen, - struct intel_region *region); - -struct _DriBufferObject *intel_region_buffer(intelScreenPrivate *intelScreen, - struct intel_region *region, - GLuint flag); #endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.c b/src/mesa/drivers/dri/i915pipe/intel_screen.c index c7c5ed8bbd..4f3a20a819 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.c +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.c @@ -35,9 +35,8 @@ #include "vblank.h" #include "xmlpool.h" - #include "intel_screen.h" - +#include "intel_batchbuffer.h" #include "intel_buffers.h" #include "intel_tex.h" #include "intel_span.h" @@ -46,8 +45,10 @@ #include "i830_dri.h" #include "dri_bufpool.h" -#include "intel_regions.h" -#include "intel_batchbuffer.h" + +#include "pipe/p_context.h" + + PUBLIC const char __driConfigOptions[] = DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE @@ -93,26 +94,27 @@ intelMapScreenRegions(__DRIscreenPrivate * sPriv) } -static struct intel_region * -intel_recreate_static(intelScreenPrivate *intelScreen, - struct intel_region *region, +static struct pipe_region * +intel_recreate_static(struct pipe_context *pipe, + struct pipe_region *region, GLuint mem_type, GLuint offset, void *virtual, GLuint cpp, GLuint pitch, GLuint height) { + struct intel_context *intel = 0; if (region) { - intel_region_update_static(intelScreen, region, mem_type, offset, - virtual, cpp, pitch, height); + pipe->region_update_static(pipe, region, mem_type, offset, + virtual, cpp, pitch, height); } else { - region = intel_region_create_static(intelScreen, mem_type, offset, - virtual, cpp, pitch, height); + region = pipe->region_create_static(pipe, mem_type, offset, + virtual, cpp, pitch, height); } return region; } -/* Create intel_region structs to describe the static front,back,depth +/* Create pipe_region structs to describe the static front,back,depth * buffers created by the xserver. * Only used for real front buffer now. * @@ -124,7 +126,7 @@ intel_recreate_static_regions(intelScreenPrivate *intelScreen) { /* this is the real front buffer which is only used for blitting to */ intelScreen->front_region = - intel_recreate_static(intelScreen, + intel_recreate_static(intelScreen->pipe, intelScreen->front_region, DRM_BO_FLAG_MEM_TT, intelScreen->front.offset, diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.h b/src/mesa/drivers/dri/i915pipe/intel_screen.h index eb4685d6e9..d5866ac0ec 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.h +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.h @@ -25,19 +25,15 @@ * **************************************************************************/ -#ifndef _INTEL_INIT_H_ -#define _INTEL_INIT_H_ +#ifndef _INTEL_SCREEN_H_ +#define _INTEL_SCREEN_H_ -#include #include "dri_util.h" #include "intel_rotate.h" #include "i830_common.h" #include "xmlconfig.h" #include "dri_bufpool.h" -/* XXX: change name or eliminate to avoid conflict with "struct - * intel_region"!!! - */ typedef struct { drm_handle_t handle; @@ -49,9 +45,10 @@ typedef struct typedef struct { + struct pipe_context *pipe; /** for accessing region functions */ intelRegion front; - struct intel_region *front_region; + struct pipe_region *front_region; int deviceID; int width; diff --git a/src/mesa/drivers/dri/i915pipe/intel_span.c b/src/mesa/drivers/dri/i915pipe/intel_span.c index 5a978d9ce2..c635ebe0af 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_span.c +++ b/src/mesa/drivers/dri/i915pipe/intel_span.c @@ -33,12 +33,13 @@ #include "intel_fbo.h" #include "intel_screen.h" #include "intel_span.h" -#include "intel_regions.h" #include "intel_ioctl.h" #include "intel_tex.h" #include "swrast/swrast.h" +#include "pipe/p_context.h" + /* break intelWriteRGBASpan_ARGB8888 */ @@ -187,9 +188,9 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) /* this is a user-created intel_renderbuffer */ if (irb->region) { if (map) - intel_region_map(intel->intelScreen, irb->region); + intel->pipe->region_map(intel->pipe, irb->region); else - intel_region_unmap(intel->intelScreen, irb->region); + intel->pipe->region_unmap(intel->pipe, irb->region); } irb->pfMap = irb->region->map; irb->pfPitch = irb->region->pitch; @@ -220,9 +221,9 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) irb = intel_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); if (irb && irb->region) { if (map) - intel_region_map(intel->intelScreen, irb->region); + intel->pipe->region_map(intel->pipe, irb->region); else - intel_region_unmap(intel->intelScreen, irb->region); + intel->pipe->region_unmap(intel->pipe, irb->region); irb->pfMap = irb->region->map; irb->pfPitch = irb->region->pitch; } @@ -261,12 +262,12 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) irb = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); if (irb && irb->region) { if (map) { - intel_region_map(intel->intelScreen, irb->region); + intel->pipe->region_map(intel->pipe, irb->region); irb->pfMap = irb->region->map; irb->pfPitch = irb->region->pitch; } else { - intel_region_unmap(intel->intelScreen, irb->region); + intel->pipe->region_unmap(intel->pipe, irb->region); irb->pfMap = NULL; irb->pfPitch = 0; } @@ -278,12 +279,12 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) irb = intel_renderbuffer(ctx->DrawBuffer->_StencilBuffer->Wrapped); if (irb && irb->region) { if (map) { - intel_region_map(intel->intelScreen, irb->region); + intel->pipe->region_map(intel->pipe, irb->region); irb->pfMap = irb->region->map; irb->pfPitch = irb->region->pitch; } else { - intel_region_unmap(intel->intelScreen, irb->region); + intel->pipe->region_unmap(intel->pipe, irb->region); irb->pfMap = NULL; irb->pfPitch = 0; } @@ -312,9 +313,9 @@ intelSpanRenderStart(GLcontext * ctx) /* Just map the framebuffer and all textures. Bufmgr code will * take care of waiting on the necessary fences: */ - intel_region_map(intel->intelScreen, intel->front_region); - intel_region_map(intel->intelScreen, intel->back_region); - intel_region_map(intel->intelScreen, intel->intelScreen->depth_region); + intel->pipe->region_map(intel->intelScreen, intel->front_region); + intel->pipe->region_map(intel->intelScreen, intel->back_region); + intel->pipe->region_map(intel->intelScreen, intel->intelScreen->depth_region); #endif for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { @@ -342,9 +343,9 @@ intelSpanRenderFinish(GLcontext * ctx) /* Now unmap the framebuffer: */ #if 0 - intel_region_unmap(intel, intel->front_region); - intel_region_unmap(intel, intel->back_region); - intel_region_unmap(intel, intel->intelScreen->depth_region); + intel->pipe->region_unmap(intel, intel->front_region); + intel->pipe->region_unmap(intel, intel->back_region); + intel->pipe->region_unmap(intel, intel->intelScreen->depth_region); #endif for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { diff --git a/src/mesa/drivers/dri/i915pipe/intel_state.c b/src/mesa/drivers/dri/i915pipe/intel_state.c index 5c5f2c6de5..fc98b6a12b 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_state.c +++ b/src/mesa/drivers/dri/i915pipe/intel_state.c @@ -36,7 +36,6 @@ #include "intel_screen.h" #include "intel_context.h" #include "intel_fbo.h" -#include "intel_regions.h" #include "swrast/swrast.h" int diff --git a/src/mesa/drivers/dri/i915pipe/intel_surface.c b/src/mesa/drivers/dri/i915pipe/intel_surface.c index e64e5f8a32..58f5cb5f96 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_surface.c +++ b/src/mesa/drivers/dri/i915pipe/intel_surface.c @@ -10,11 +10,11 @@ #include "intel_context.h" #include "intel_buffers.h" -#include "intel_regions.h" #include "intel_span.h" #include "intel_fbo.h" #include "pipe/p_state.h" +#include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/softpipe/sp_surface.h" @@ -173,7 +173,7 @@ map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) #if 0 intelFinish(&intel->ctx); /* XXX need this? */ #endif - intel_region_map(intel->intelScreen, irb->region); + intel->pipe->region_map(intel->pipe, irb->region); } pb->ptr = irb->region->map; @@ -194,7 +194,7 @@ unmap_surface_buffer(struct pipe_buffer *pb) if (irb->region) { GET_CURRENT_CONTEXT(ctx); struct intel_context *intel = intel_context(ctx); - intel_region_unmap(intel->intelScreen, irb->region); + intel->pipe->region_unmap(intel->pipe, irb->region); } pb->ptr = NULL; diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c index 8d7f8f9d91..6deb6abd4d 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c @@ -36,7 +36,6 @@ #include "intel_batchbuffer.h" #include "intel_buffers.h" #include "intel_mipmap_tree.h" -#include "intel_regions.h" #include "intel_fbo.h" #include "intel_tex.h" #include "intel_blit.h" @@ -45,12 +44,12 @@ #define FILE_DEBUG_FLAG DEBUG_TEXTURE /** - * Get the intel_region which is the source for any glCopyTex[Sub]Image call. + * Get the pipe_region which is the source for any glCopyTex[Sub]Image call. * * Do the best we can using the blitter. A future project is to use * the texture engine and fragment programs for these copies. */ -static const struct intel_region * +static const struct pipe_region * get_teximage_source(struct intel_context *intel, GLenum internalFormat) { struct intel_renderbuffer *irb; @@ -92,7 +91,7 @@ do_copy_texsubimage(struct intel_context *intel, GLint x, GLint y, GLsizei width, GLsizei height) { GLcontext *ctx = &intel->ctx; - const struct intel_region *src = + const struct pipe_region *src = get_teximage_source(intel, internalFormat); if (!intelImage->mt || !src) { diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c index f790b1e6f7..37307b49c5 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c @@ -23,6 +23,8 @@ #include "intel_ioctl.h" #include "intel_blit.h" +#include "pipe/p_context.h" + #define FILE_DEBUG_FLAG DEBUG_TEXTURE /* Functions to store texture images. Where possible, mipmap_tree's @@ -224,8 +226,8 @@ try_pbo_upload(struct intel_context *intel, struct _DriBufferObject *src_buffer = intel_bufferobj_buffer(intel, pbo, INTEL_READ); struct _DriBufferObject *dst_buffer = - intel_region_buffer(intel->intelScreen, intelImage->mt->region, - INTEL_WRITE_FULL); + intel->pipe->region_buffer(intel->pipe, intelImage->mt->region, + INTEL_WRITE_FULL); intelEmitCopyBlit(intel, @@ -281,7 +283,7 @@ try_pbo_zcopy(struct intel_context *intel, return GL_FALSE; } - intel_region_attach_pbo(intel->intelScreen, intelImage->mt->region, pbo); + intel->pipe->region_attach_pbo(intel->pipe, intelImage->mt->region, pbo); return GL_TRUE; } @@ -411,7 +413,7 @@ intelTexImage(GLcontext * ctx, /* Attempt to texture directly from PBO data (zero copy upload). * * Currently disable as it can lead to worse as well as better - * performance (in particular when intel_region_cow() is + * performance (in particular when pipe_region_cow() is * required). */ if (intelObj->mt == intelImage->mt && @@ -460,7 +462,7 @@ intelTexImage(GLcontext * ctx, if (intelImage->mt) - intel_region_idle(intel->intelScreen, intelImage->mt->region); + intel->pipe->region_idle(intel->pipe, intelImage->mt->region); LOCK_HARDWARE(intel); diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c index 3935787806..d660c2dcee 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c @@ -35,6 +35,9 @@ #include "intel_tex.h" #include "intel_mipmap_tree.h" +#include "pipe/p_context.h" + + #define FILE_DEBUG_FLAG DEBUG_TEXTURE static void @@ -65,7 +68,7 @@ intelTexSubimage(GLcontext * ctx, return; if (intelImage->mt) - intel_region_idle(intel->intelScreen, intelImage->mt->region); + intel->pipe->region_idle(intel->pipe, intelImage->mt->region); LOCK_HARDWARE(intel); diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 05a175c8dc..bfae55a4ba 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -114,7 +114,86 @@ struct pipe_context { void (*set_viewport_state)( struct pipe_context *, const struct pipe_viewport_state * ); + + + /* + * Memory region functions + * Some of these may go away... + */ + struct pipe_region *(*region_alloc)(struct pipe_context *pipe, + GLuint cpp, GLuint pitch, GLuint height); + + void (*region_release)(struct pipe_context *pipe, struct pipe_region **r); + + struct pipe_region *(*region_create_static)(struct pipe_context *pipe, + GLuint mem_type, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, + GLuint height); + + void (*region_update_static)(struct pipe_context *pipe, + struct pipe_region *region, + GLuint mem_type, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, GLuint height); + + void (*region_idle)(struct pipe_context *pipe, struct pipe_region *region); + + GLubyte *(*region_map)(struct pipe_context *pipe, struct pipe_region *r); + + void (*region_unmap)(struct pipe_context *pipe, struct pipe_region *r); + + void (*region_data)(struct pipe_context *pipe, + struct pipe_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + const void *src, GLuint src_stride, + GLuint srcx, GLuint srcy, GLuint width, GLuint height); + + void (*region_copy)(struct pipe_context *pipe, + struct pipe_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + const struct pipe_region *src, + GLuint src_offset, + GLuint srcx, GLuint srcy, GLuint width, GLuint height); + + void (*region_fill)(struct pipe_context *pipe, + struct pipe_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + GLuint width, GLuint height, GLuint color); + + void (*region_cow)(struct pipe_context *pipe, struct pipe_region *region); + + void (*region_attach_pbo)(struct pipe_context *pipe, + struct pipe_region *region, + struct intel_buffer_object *pbo); + + void (*region_release_pbo)(struct pipe_context *pipe, + struct pipe_region *region); + + struct _DriBufferObject *(*region_buffer)(struct pipe_context *pipe, + struct pipe_region *region, + GLuint flag); + + void *screen; /**< temporary */ }; -#endif + +static INLINE void +pipe_region_reference(struct pipe_region **dst, struct pipe_region *src) +{ + assert(*dst == NULL); + if (src) { + src->refcount++; + *dst = src; + } +} + + + +#endif /* PIPE_CONTEXT_H */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 9973a7b8dd..d81c3f1778 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -272,6 +272,24 @@ struct pipe_surface }; +struct _DriBufferObject; +struct intel_buffer_object; + +struct pipe_region +{ + struct _DriBufferObject *buffer; /**< buffer manager's buffer ID */ + GLuint refcount; /**< Reference count for region */ + GLuint cpp; /**< bytes per pixel */ + GLuint pitch; /**< in pixels */ + GLuint height; /**< in pixels */ + GLubyte *map; /**< only non-NULL when region is actually mapped */ + GLuint map_refcount; /**< Reference count for mapping */ + + GLuint draw_offset; /**< Offset of drawing address within the region */ + + struct intel_buffer_object *pbo; /* zero-copy uploads */ +}; + /** * Texture object. * Mipmap levels, cube faces, 3D slices can be accessed as surfaces. -- cgit v1.2.3 From 20adf45c23dd9ec86a1439ad87c1473395bbb1a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 17:42:03 -0600 Subject: Redesign pipe_surface in terms of pipe_region. struct pipe_buffer goes away. Added basic region functions to softpipe to allocate/release malloc'd regions. Surface-related code is fairly coherent now. --- src/mesa/drivers/dri/i915pipe/intel_fbo.c | 3 +- src/mesa/drivers/dri/i915pipe/intel_surface.c | 85 ++-------- src/mesa/drivers/x11/xm_buffer.c | 13 +- src/mesa/drivers/x11/xm_surface.c | 42 ++--- src/mesa/drivers/x11/xmesaP.h | 2 +- src/mesa/main/renderbuffer.c | 54 +++--- src/mesa/pipe/p_context.h | 8 + src/mesa/pipe/p_state.h | 53 ++---- src/mesa/pipe/softpipe/sp_context.c | 19 ++- src/mesa/pipe/softpipe/sp_region.c | 105 ++++++++++++ src/mesa/pipe/softpipe/sp_region.h | 40 +++++ src/mesa/pipe/softpipe/sp_surface.c | 236 ++++++++++++++++++++++++-- src/mesa/pipe/softpipe/sp_surface.h | 7 +- src/mesa/pipe/softpipe/sp_z_surface.c | 138 ++++++--------- src/mesa/sources | 3 +- 15 files changed, 538 insertions(+), 270 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_region.c create mode 100644 src/mesa/pipe/softpipe/sp_region.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index bac2ef8467..1470ce7d82 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -291,11 +291,10 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, rb->Width = width; rb->Height = height; -#if 1 /* update the surface's size too */ rb->surface->width = width; rb->surface->height = height; -#endif + rb->surface->region = irb->region; /* This sets the Get/PutRow/Value functions */ intel_set_span_functions(&irb->Base); diff --git a/src/mesa/drivers/dri/i915pipe/intel_surface.c b/src/mesa/drivers/dri/i915pipe/intel_surface.c index 58f5cb5f96..936c9cb5e7 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_surface.c +++ b/src/mesa/drivers/dri/i915pipe/intel_surface.c @@ -33,9 +33,9 @@ static void read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rrrr)[QUAD_SIZE]) { - const GLint bytesPerRow = sps->surface.stride * sps->surface.cpp; + const GLint bytesPerRow = sps->surface.region->pitch * sps->surface.region->cpp; const GLint invY = sps->surface.height - y - 1; - const GLubyte *src = sps->surface.ptr + invY * bytesPerRow + x * sps->surface.cpp; + const GLubyte *src = sps->surface.region->map + invY * bytesPerRow + x * sps->surface.region->cpp; GLfloat *dst = (GLfloat *) rrrr; GLubyte temp[16]; GLuint j; @@ -59,9 +59,9 @@ write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rrrr)[QUAD_SIZE]) { const GLfloat *src = (const GLfloat *) rrrr; - const GLint bytesPerRow = sps->surface.stride * sps->surface.cpp; + const GLint bytesPerRow = sps->surface.region->pitch * sps->surface.region->cpp; const GLint invY = sps->surface.height - y - 1; - GLubyte *dst = sps->surface.ptr + invY * bytesPerRow + x * sps->surface.cpp; + GLubyte *dst = sps->surface.region->map + invY * bytesPerRow + x * sps->surface.region->cpp; GLubyte temp[16]; GLuint j; @@ -87,16 +87,16 @@ read_quad_z24(struct softpipe_surface *sps, static const GLuint mask = 0xffffff; const GLint invY = sps->surface.height - y - 1; const GLuint *src - = (GLuint *) (sps->surface.ptr - + (invY * sps->surface.stride + x) * sps->surface.cpp); + = (GLuint *) (sps->surface.region->map + + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); assert(sps->surface.format == PIPE_FORMAT_Z24_S8); /* extract lower three bytes */ zzzz[0] = src[0] & mask; zzzz[1] = src[1] & mask; - zzzz[2] = src[-sps->surface.stride] & mask; - zzzz[3] = src[-sps->surface.stride + 1] & mask; + zzzz[2] = src[-sps->surface.region->pitch] & mask; + zzzz[3] = src[-sps->surface.region->pitch + 1] & mask; } static void @@ -106,15 +106,15 @@ write_quad_z24(struct softpipe_surface *sps, static const GLuint mask = 0xff000000; const GLint invY = sps->surface.height - y - 1; GLuint *dst - = (GLuint *) (sps->surface.ptr - + (invY * sps->surface.stride + x) * sps->surface.cpp); + = (GLuint *) (sps->surface.region->map + + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); assert(sps->surface.format == PIPE_FORMAT_Z24_S8); /* write lower three bytes */ dst[0] = (dst[0] & mask) | zzzz[0]; dst[1] = (dst[1] & mask) | zzzz[1]; - dst -= sps->surface.stride; + dst -= sps->surface.region->pitch; dst[0] = (dst[0] & mask) | zzzz[2]; dst[1] = (dst[1] & mask) | zzzz[3]; } @@ -125,15 +125,15 @@ read_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) { const GLint invY = sps->surface.height - y - 1; - const GLuint *src = (const GLuint *) (sps->surface.ptr - + (invY * sps->surface.stride + x) * sps->surface.cpp); + const GLuint *src = (const GLuint *) (sps->surface.region->map + + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); assert(sps->surface.format == PIPE_FORMAT_Z24_S8); /* extract high byte */ ssss[0] = src[0] >> 24; ssss[1] = src[1] >> 24; - src -= sps->surface.stride; + src -= sps->surface.region->pitch; ssss[2] = src[0] >> 24; ssss[3] = src[1] >> 24; } @@ -144,68 +144,20 @@ write_quad_stencil(struct softpipe_surface *sps, { static const GLuint mask = 0x00ffffff; const GLint invY = sps->surface.height - y - 1; - GLuint *dst = (GLuint *) (sps->surface.ptr - + (invY * sps->surface.stride + x) * sps->surface.cpp); + GLuint *dst = (GLuint *) (sps->surface.region->map + + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); assert(sps->surface.format == PIPE_FORMAT_Z24_S8); /* write high byte */ dst[0] = (dst[0] & mask) | (ssss[0] << 24); dst[1] = (dst[1] & mask) | (ssss[1] << 24); - dst -= sps->surface.stride; + dst -= sps->surface.region->pitch; dst[0] = (dst[0] & mask) | (ssss[2] << 24); dst[1] = (dst[1] & mask) | (ssss[3] << 24); } -static void * -map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) -{ - struct softpipe_surface *sps = (struct softpipe_surface *) pb; - struct intel_renderbuffer *irb = (struct intel_renderbuffer *) sps->surface.rb; - assert(access_mode == PIPE_MAP_READ_WRITE); - - /*LOCK_HARDWARE(intel);*/ - - if (irb->region) { - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); -#if 0 - intelFinish(&intel->ctx); /* XXX need this? */ -#endif - intel->pipe->region_map(intel->pipe, irb->region); - } - pb->ptr = irb->region->map; - - sps->surface.stride = irb->region->pitch; - sps->surface.cpp = irb->region->cpp; - sps->surface.ptr = irb->region->map; - - return pb->ptr; -} - - -static void -unmap_surface_buffer(struct pipe_buffer *pb) -{ - struct softpipe_surface *sps = (struct softpipe_surface *) pb; - struct intel_renderbuffer *irb = (struct intel_renderbuffer *) sps->surface.rb; - - if (irb->region) { - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); - intel->pipe->region_unmap(intel->pipe, irb->region); - } - pb->ptr = NULL; - - sps->surface.stride = 0; - sps->surface.cpp = 0; - sps->surface.ptr = NULL; - - /*UNLOCK_HARDWARE(intel);*/ -} - - struct pipe_surface * intel_new_surface(GLuint intFormat) { @@ -241,8 +193,5 @@ intel_new_surface(GLuint intFormat) } - sps->surface.buffer.map = map_surface_buffer; - sps->surface.buffer.unmap = unmap_surface_buffer; - return &sps->surface; } diff --git a/src/mesa/drivers/x11/xm_buffer.c b/src/mesa/drivers/x11/xm_buffer.c index 8fbd9a783b..5bba52da48 100644 --- a/src/mesa/drivers/x11/xm_buffer.c +++ b/src/mesa/drivers/x11/xm_buffer.c @@ -269,7 +269,10 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->Height = height; rb->InternalFormat = internalFormat; - rb->surface->resize(rb->surface, width, height); + if (!xrb->Base.surface) + xrb->Base.surface = xmesa_new_surface(ctx, xrb); + xrb->Base.surface->width = width; + xrb->Base.surface->height = height; return GL_TRUE; } @@ -320,7 +323,10 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xrb->origin4 = NULL; } - rb->surface->resize(rb->surface, width, height); + if (!xrb->Base.surface) + xrb->Base.surface = xmesa_new_surface(ctx, xrb); + xrb->Base.surface->width = width; + xrb->Base.surface->height = height; return GL_TRUE; } @@ -357,9 +363,6 @@ xmesa_new_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, xrb->Base.IndexBits = visual->indexBits; } /* only need to set Red/Green/EtcBits fields for user-created RBs */ - - xrb->Base.surface = xmesa_new_surface(xrb); - } return xrb; } diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 17f5f28a9d..e8e795c00f 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -42,25 +42,11 @@ #include "framebuffer.h" #include "renderbuffer.h" -#include "pipe/p_state.h" +#include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/softpipe/sp_context.h" #include "pipe/softpipe/sp_surface.h" - - -static void * -map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) -{ - /* no-op */ - return NULL; -} - - -static void -unmap_surface_buffer(struct pipe_buffer *pb) -{ - /* no-op */ -} +#include "state_tracker/st_context.h" static INLINE struct xmesa_renderbuffer * @@ -174,27 +160,22 @@ write_mono_row_ub(struct softpipe_surface *sps, GLuint count, GLint x, GLint y, } -static void -resize_surface(struct pipe_surface *ps, GLuint width, GLuint height) -{ - ps->width = width; - ps->height = height; -} - - /** * Called to create a pipe_surface for each X renderbuffer. + * Note: this is being used instead of pipe->surface_alloc() since we + * have special/unique quad read/write functions for X. */ struct pipe_surface * -xmesa_new_surface(struct xmesa_renderbuffer *xrb) +xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb) { + struct pipe_context *pipe = ctx->st->pipe; struct softpipe_surface *sps; sps = CALLOC_STRUCT(softpipe_surface); if (!sps) return NULL; - sps->surface.rb = xrb; + sps->surface.rb = xrb; /* XXX only needed for quad funcs above */ sps->surface.width = xrb->Base.Width; sps->surface.height = xrb->Base.Height; @@ -206,9 +187,12 @@ xmesa_new_surface(struct xmesa_renderbuffer *xrb) sps->write_quad_ub = write_quad_ub; sps->write_mono_row_ub = write_mono_row_ub; - sps->surface.buffer.map = map_surface_buffer; - sps->surface.buffer.unmap = unmap_surface_buffer; - sps->surface.resize = resize_surface; + /* Note, the region we allocate doesn't actually have any storage + * since we're drawing into an XImage or Pixmap. + * The region's size will get set in the xmesa_alloc_front/back_storage() + * functions. + */ + sps->surface.region = pipe->region_alloc(pipe, 0, 0, 0); return &sps->surface; } diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index daf6a3f942..098b9218ae 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -589,7 +589,7 @@ extern void xmesa_register_swrast_functions( GLcontext *ctx ); struct pipe_surface; struct pipe_surface * -xmesa_new_surface(struct xmesa_renderbuffer *xrb); +xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb); #endif diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c index a900de169e..7c35575d12 100644 --- a/src/mesa/main/renderbuffer.c +++ b/src/mesa/main/renderbuffer.c @@ -51,7 +51,9 @@ #include "pipe/softpipe/sp_z_surface.h" #include "pipe/p_state.h" +#include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "state_tracker/st_context.h" /* 32-bit color index format. Not a public format. */ @@ -949,6 +951,7 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, GLenum internalFormat, GLuint width, GLuint height) { + struct pipe_context *pipe = ctx->st->pipe; GLuint pixelSize; /* first clear these fields */ @@ -1065,6 +1068,9 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutMonoValues = put_mono_values_ubyte; rb->StencilBits = 8 * sizeof(GLubyte); pixelSize = sizeof(GLubyte); + if (!rb->surface) + rb->surface = (struct pipe_surface *) + pipe->surface_alloc(pipe, PIPE_FORMAT_U_S8); break; case GL_STENCIL_INDEX16_EXT: rb->_ActualFormat = GL_STENCIL_INDEX16_EXT; @@ -1095,8 +1101,9 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutValues = put_values_ushort; rb->PutMonoValues = put_mono_values_ushort; rb->DepthBits = 8 * sizeof(GLushort); - rb->surface - = (struct pipe_surface *) softpipe_new_z_surface(PIPE_FORMAT_U_Z16); + if (!rb->surface) + rb->surface = (struct pipe_surface *) + pipe->surface_alloc(pipe, PIPE_FORMAT_U_Z16); pixelSize = sizeof(GLushort); break; case GL_DEPTH_COMPONENT24: @@ -1119,8 +1126,9 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->_ActualFormat = GL_DEPTH_COMPONENT32; rb->DepthBits = 32; } - rb->surface - = (struct pipe_surface *) softpipe_new_z_surface(PIPE_FORMAT_U_Z32); + if (!rb->surface) + rb->surface = (struct pipe_surface *) + pipe->surface_alloc(pipe, PIPE_FORMAT_U_Z32); pixelSize = sizeof(GLuint); break; case GL_DEPTH_STENCIL_EXT: @@ -1138,8 +1146,9 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutMonoValues = put_mono_values_uint; rb->DepthBits = 24; rb->StencilBits = 8; - rb->surface - = (struct pipe_surface *) softpipe_new_z_surface(PIPE_FORMAT_Z24_S8); + if (!rb->surface) + rb->surface = (struct pipe_surface *) + pipe->surface_alloc(pipe, PIPE_FORMAT_Z24_S8); pixelSize = sizeof(GLuint); break; case GL_COLOR_INDEX8_EXT: @@ -1202,28 +1211,33 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, ASSERT(rb->PutMonoValues); /* free old buffer storage */ - if (rb->Data) { - if (rb->surface) { - /* pipe surface */ - } - else { - /* legacy renderbuffer */ - _mesa_free(rb->Data); - } - rb->Data = NULL; + if (rb->surface) { + /* pipe_surface/region */ } + else if (rb->Data) { + /* legacy renderbuffer (this will go away) */ + _mesa_free(rb->Data); + } + rb->Data = NULL; if (width > 0 && height > 0) { /* allocate new buffer storage */ if (rb->surface) { - /* pipe surface */ - rb->surface->resize(rb->surface, width, height); - rb->Data = rb->surface->buffer.ptr; + /* pipe_surface/region */ + if (rb->surface->region) { + pipe->region_unmap(pipe, rb->surface->region); + pipe->region_release(pipe, &rb->surface->region); + } + rb->surface->region = pipe->region_alloc(pipe, pixelSize, width, height); + /* XXX probably don't want to really map here */ + pipe->region_map(pipe, rb->surface->region); + rb->Data = rb->surface->region->map; } else { - /* legacy renderbuffer */ - rb->Data = _mesa_malloc(width * height * pixelSize); + /* legacy renderbuffer (this will go away) */ + rb->Data = malloc(width * height * pixelSize); } + if (rb->Data == NULL) { rb->Width = 0; rb->Height = 0; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index bfae55a4ba..b48c7775de 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -116,6 +116,14 @@ struct pipe_context { const struct pipe_viewport_state * ); + /* + * Surface functions + * This might go away... + */ + struct pipe_surface *(*surface_alloc)(struct pipe_context *pipe, + GLuint format); + + /* * Memory region functions * Some of these may go away... diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index d81c3f1778..0767fc2fcb 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -233,51 +233,17 @@ struct pipe_sampler_state /*** - *** Non-state Objects + *** Resource Objects ***/ -/** - * A mappable buffer (vertex data, pixel data, etc) - * XXX replace with "intel_region". - */ -struct pipe_buffer -{ - void (*buffer_data)(struct pipe_buffer *pb, GLuint size, const void *src); - void (*buffer_sub_data)(struct pipe_buffer *pb, GLuint offset, GLuint size, - const void *src); - void *(*map)(struct pipe_buffer *pb, GLuint access_mode); - void (*unmap)(struct pipe_buffer *pb); - GLubyte *ptr; /**< address, only valid while mapped */ - GLuint mode; /**< PIPE_MAP_x, only valid while mapped */ -}; - - -/** - * 2D surface. - * May be a renderbuffer, texture mipmap level, etc. - */ -struct pipe_surface -{ - struct pipe_buffer buffer; /**< surfaces can be mapped */ - GLuint format:5; /**< PIPE_FORMAT_x */ - GLuint width, height; - - GLint stride, cpp; - GLubyte *ptr; /**< only valid while mapped, may not equal buffer->ptr */ - - void *rb; /**< Ptr back to renderbuffer (temporary?) */ - - void (*resize)(struct pipe_surface *ps, GLuint width, GLuint height); -}; - - struct _DriBufferObject; struct intel_buffer_object; struct pipe_region { struct _DriBufferObject *buffer; /**< buffer manager's buffer ID */ + GLuint refcount; /**< Reference count for region */ GLuint cpp; /**< bytes per pixel */ GLuint pitch; /**< in pixels */ @@ -290,6 +256,21 @@ struct pipe_region struct intel_buffer_object *pbo; /* zero-copy uploads */ }; + +/** + * 2D surface. + * May be a renderbuffer, texture mipmap level, etc. + */ +struct pipe_surface +{ + struct pipe_region *region; + GLuint format:5; /**< PIPE_FORMAT_x */ + GLuint width, height; + + void *rb; /**< Ptr back to renderbuffer (temporary?) */ +}; + + /** * Texture object. * Mipmap levels, cube faces, 3D slices can be accessed as surfaces. diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 8655aa83fd..002fe73b59 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -35,6 +35,7 @@ #include "pipe/p_defines.h" #include "sp_context.h" #include "sp_clear.h" +#include "sp_region.h" #include "sp_state.h" #include "sp_surface.h" #include "sp_prim_setup.h" @@ -42,18 +43,16 @@ static void map_surfaces(struct softpipe_context *sp) { + struct pipe_context *pipe = &sp->pipe; GLuint i; for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); - struct pipe_buffer *buf = &sps->surface.buffer; - buf->map(buf, PIPE_MAP_READ_WRITE); + struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); pipe->region_map(pipe, sps->surface.region); } if (sp->framebuffer.zbuf) { struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.zbuf); - struct pipe_buffer *buf = &sps->surface.buffer; - buf->map(buf, PIPE_MAP_READ_WRITE); + pipe->region_map(pipe, sps->surface.region); } /* XXX depth & stencil bufs */ @@ -62,18 +61,17 @@ static void map_surfaces(struct softpipe_context *sp) static void unmap_surfaces(struct softpipe_context *sp) { + struct pipe_context *pipe = &sp->pipe; GLuint i; for (i = 0; i < sp->framebuffer.num_cbufs; i++) { struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); - struct pipe_buffer *buf = &sps->surface.buffer; - buf->unmap(buf); + pipe->region_unmap(pipe, sps->surface.region); } if (sp->framebuffer.zbuf) { struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.zbuf); - struct pipe_buffer *buf = &sps->surface.buffer; - buf->unmap(buf); + pipe->region_unmap(pipe, sps->surface.region); } /* XXX depth & stencil bufs */ } @@ -161,6 +159,9 @@ struct pipe_context *softpipe_create( void ) softpipe->draw = draw_create(); draw_set_setup_stage(softpipe->draw, sp_draw_render_stage(softpipe)); + sp_init_region_functions(softpipe); + sp_init_surface_functions(softpipe); + /* * XXX we could plug GL selection/feedback into the drawing pipeline * by specifying a different setup/render stage. diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c new file mode 100644 index 0000000000..04ae5e94f9 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -0,0 +1,105 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Software-based region allocation and management. + * A hardware driver would override these functions. + */ + + +#include "sp_context.h" +#include "sp_region.h" +#include "sp_surface.h" +#include "main/imports.h" + + +static struct pipe_region * +sp_region_alloc(struct pipe_context *pipe, + GLuint cpp, GLuint pitch, GLuint height) +{ + struct pipe_region *region = CALLOC_STRUCT(pipe_region); + if (!region) + return NULL; + + region->refcount = 1; + region->cpp = cpp; + region->pitch = pitch; + region->height = height; + region->map = malloc(cpp * pitch * height); + + return region; +} + + +static void +sp_region_release(struct pipe_context *pipe, struct pipe_region **region) +{ + assert((*region)->refcount > 0); + (*region)->refcount--; + + if ((*region)->refcount == 0) { + assert((*region)->map_refcount == 0); + +#if 0 + if ((*region)->pbo) + (*region)->pbo->region = NULL; + (*region)->pbo = NULL; +#endif + + free(*region); + } + *region = NULL; +} + + + +static GLubyte * +sp_region_map(struct pipe_context *pipe, struct pipe_region *region) +{ + region->map_refcount++; + return region->map; +} + + +static void +sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) +{ + region->map_refcount--; +} + + +void +sp_init_region_functions(struct softpipe_context *sp) +{ + sp->pipe.region_alloc = sp_region_alloc; + sp->pipe.region_release = sp_region_release; + + sp->pipe.region_map = sp_region_map; + sp->pipe.region_unmap = sp_region_unmap; + + /* XXX lots of other region functions needed... */ +} diff --git a/src/mesa/pipe/softpipe/sp_region.h b/src/mesa/pipe/softpipe/sp_region.h new file mode 100644 index 0000000000..432746b27f --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_region.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SP_REGION_H +#define SP_REGION_H + + +struct softpipe_context; + + +extern void +sp_init_region_functions(struct softpipe_context *sp); + + +#endif /* SP_REGION_H */ diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 16bbacb12b..cf89d28941 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -28,8 +28,22 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_surface.h" -#include "sp_headers.h" +#include "pipe/p_defines.h" +#include "main/imports.h" + +/** + * Softpipe surface functions. + * Basically, create surface of a particular type, then plug in default + * read/write_quad functions. + * Note that these quad funcs assume the buffer/region is in a linear + * layout with Y=0=bottom. + * If we had swizzled/AOS buffers the read/write functions could be + * simplified a lot.... + */ + + +#if 000 /* OLD... should be recycled... */ static void rgba8_read_quad_f( struct softpipe_surface *gs, GLint x, GLint y, GLfloat (*rgba)[NUM_CHANNELS] ) @@ -98,9 +112,6 @@ static void rgba8_write_quad_f_swz( struct softpipe_surface *gs, } } - - - static void rgba8_read_quad_ub( struct softpipe_surface *gs, GLint x, GLint y, GLubyte (*rgba)[NUM_CHANNELS] ) @@ -118,7 +129,6 @@ static void rgba8_read_quad_ub( struct softpipe_surface *gs, } } - static void rgba8_write_quad_ub( struct softpipe_surface *gs, GLint x, GLint y, GLubyte (*rgba)[NUM_CHANNELS] ) @@ -136,18 +146,216 @@ static void rgba8_write_quad_ub( struct softpipe_surface *gs, } } +#endif + + + +static void +z16_read_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) +{ + const GLushort *src + = (GLushort *) sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_U_Z16); + + /* converting GLushort to GLuint: */ + zzzz[0] = src[0]; + zzzz[1] = src[1]; + src += sps->surface.region->pitch; + zzzz[2] = src[0]; + zzzz[3] = src[1]; +} + +static void +z16_write_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) +{ + GLushort *dst = (GLushort *) sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_U_Z16); + + /* converting GLuint to GLushort: */ + dst[0] = zzzz[0]; + dst[1] = zzzz[1]; + dst += sps->surface.region->pitch; + dst[0] = zzzz[2]; + dst[1] = zzzz[3]; +} + +static void +z32_read_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) +{ + const GLuint *src + = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_U_Z32); + + zzzz[0] = src[0]; + zzzz[1] = src[1]; + src += sps->surface.region->pitch; + zzzz[2] = src[0]; + zzzz[3] = src[1]; +} + +static void +z32_write_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) +{ + GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_U_Z32); + + dst[0] = zzzz[0]; + dst[1] = zzzz[1]; + dst += sps->surface.region->pitch; + dst[0] = zzzz[2]; + dst[1] = zzzz[3]; +} + +static void +z24s8_read_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) +{ + const GLuint *src + = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + zzzz[0] = src[0] >> 8; + zzzz[1] = src[1] >> 8; + src += sps->surface.region->pitch; + zzzz[2] = src[0] >> 8; + zzzz[3] = src[1] >> 8; +} + +static void +z24s8_write_quad_z(struct softpipe_surface *sps, + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) +{ + GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(zzzz[0] <= 0xffffff); + + dst[0] = (dst[0] & 0xff) | (zzzz[0] << 8); + dst[1] = (dst[1] & 0xff) | (zzzz[1] << 8); + dst += sps->surface.region->pitch; + dst[0] = (dst[0] & 0xff) | (zzzz[2] << 8); + dst[1] = (dst[1] & 0xff) | (zzzz[3] << 8); +} + +static void +z24s8_read_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) +{ + const GLuint *src + = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + + ssss[0] = src[0] & 0xff; + ssss[1] = src[1] & 0xff; + src += sps->surface.region->pitch; + ssss[2] = src[0] & 0xff; + ssss[3] = src[1] & 0xff; +} +static void +z24s8_write_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) +{ + GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + dst[0] = (dst[0] & 0xffffff00) | ssss[0]; + dst[1] = (dst[1] & 0xffffff00) | ssss[1]; + dst += sps->surface.region->pitch; + dst[0] = (dst[0] & 0xffffff00) | ssss[2]; + dst[1] = (dst[1] & 0xffffff00) | ssss[3]; +} -struct softpipe_surface_type gs_rgba8 = { - G_SURFACE_RGBA_8888, - rgba8_read_quad_f, - rgba8_read_quad_f_swz, - rgba8_read_quad_ub, - rgba8_write_quad_f, - rgba8_write_quad_f_swz, - rgba8_write_quad_ub, -}; +static void +s8_read_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) +{ + const GLubyte *src + = sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_U_S8); + + ssss[0] = src[0]; + ssss[1] = src[1]; + src += sps->surface.region->pitch; + ssss[2] = src[0]; + ssss[3] = src[1]; +} + +static void +s8_write_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) +{ + GLubyte *dst + = sps->surface.region->map + y * sps->surface.region->pitch + x; + assert(sps->surface.format == PIPE_FORMAT_U_S8); + dst[0] = ssss[0]; + dst[1] = ssss[1]; + dst += sps->surface.region->pitch; + dst[0] = ssss[2]; + dst[1] = ssss[3]; +} + + + +static void +init_quad_funcs(struct softpipe_surface *sps) +{ + switch (sps->surface.format) { + case PIPE_FORMAT_U_Z16: + sps->read_quad_z = z16_read_quad_z; + sps->write_quad_z = z16_write_quad_z; + break; + case PIPE_FORMAT_U_Z32: + sps->read_quad_z = z32_read_quad_z; + sps->write_quad_z = z32_write_quad_z; + break; + case PIPE_FORMAT_Z24_S8: + sps->read_quad_z = z24s8_read_quad_z; + sps->write_quad_z = z24s8_write_quad_z; + sps->read_quad_stencil = z24s8_read_quad_stencil; + sps->write_quad_stencil = z24s8_write_quad_stencil; + break; + case PIPE_FORMAT_U_S8: + sps->read_quad_stencil = s8_read_quad_stencil; + sps->write_quad_stencil = s8_write_quad_stencil; + break; + default: + assert(0); + } +} + + +static struct pipe_surface * +sp_surface_alloc(struct pipe_context *pipe, GLenum format) +{ + struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); + if (!sps) + return NULL; + + sps->surface.format = format; + init_quad_funcs(sps); + + return &sps->surface; +} + + +void +sp_init_surface_functions(struct softpipe_context *sp) +{ + sp->pipe.surface_alloc = sp_surface_alloc; +} diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 3ba732cebe..e8466256db 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -35,8 +35,7 @@ #include "sp_headers.h" struct softpipe_surface; - -#define G_SURFACE_RGBA_8888 0x1 +struct softpipe_context; /** @@ -98,4 +97,8 @@ softpipe_surface(struct pipe_surface *ps) } +extern void +sp_init_surface_functions(struct softpipe_context *sp); + + #endif /* SP_SURFACE_H */ diff --git a/src/mesa/pipe/softpipe/sp_z_surface.c b/src/mesa/pipe/softpipe/sp_z_surface.c index 744737cb6c..7ff1a0cbc8 100644 --- a/src/mesa/pipe/softpipe/sp_z_surface.c +++ b/src/mesa/pipe/softpipe/sp_z_surface.c @@ -40,75 +40,35 @@ #include "sp_surface.h" #include "sp_z_surface.h" -static void* -z_map(struct pipe_buffer *pb, GLuint access_mode) -{ - struct softpipe_surface *sps = (struct softpipe_surface *) pb; - sps->surface.ptr = pb->ptr; - sps->surface.stride = sps->surface.width; - return pb->ptr; -} - -static void -z_unmap(struct pipe_buffer *pb) -{ - struct softpipe_surface *sps = (struct softpipe_surface *) pb; - sps->surface.ptr = NULL; - sps->surface.stride = 0; -} - -static void -z_resize(struct pipe_surface *ps, GLuint width, GLuint height) -{ - struct softpipe_surface *sps = (struct softpipe_surface *) ps; - - if (sps->surface.buffer.ptr) - free(sps->surface.buffer.ptr); - - sps->surface.stride = sps->surface.width; - if (sps->surface.format == PIPE_FORMAT_U_Z16) - sps->surface.cpp = 2; - else if (sps->surface.format == PIPE_FORMAT_U_Z32 || - sps->surface.format == PIPE_FORMAT_Z24_S8) - sps->surface.cpp = 4; - else - assert(0); - - ps->buffer.ptr = (GLubyte *) malloc(width * height * sps->surface.cpp); - ps->width = width; - ps->height = height; - -} - static void z16_read_quad_z(struct softpipe_surface *sps, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) { const GLushort *src - = (GLushort *) sps->surface.ptr + y * sps->surface.stride + x; + = (GLushort *) sps->surface.region->map + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z16); /* converting GLushort to GLuint: */ zzzz[0] = src[0]; zzzz[1] = src[1]; - zzzz[2] = src[sps->surface.width + 0]; - zzzz[3] = src[sps->surface.width + 1]; + zzzz[2] = src[sps->surface.region->pitch + 0]; + zzzz[3] = src[sps->surface.region->pitch + 1]; } static void z16_write_quad_z(struct softpipe_surface *sps, GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) { - GLushort *dst = (GLushort *) sps->surface.ptr + y * sps->surface.stride + x; + GLushort *dst = (GLushort *) sps->surface.region->map + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z16); /* converting GLuint to GLushort: */ dst[0] = zzzz[0]; dst[1] = zzzz[1]; - dst[sps->surface.width + 0] = zzzz[2]; - dst[sps->surface.width + 1] = zzzz[3]; + dst[sps->surface.region->pitch + 0] = zzzz[2]; + dst[sps->surface.region->pitch + 1] = zzzz[3]; } static void @@ -116,28 +76,28 @@ z32_read_quad_z(struct softpipe_surface *sps, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) { const GLuint *src - = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z32); zzzz[0] = src[0]; zzzz[1] = src[1]; - zzzz[2] = src[sps->surface.width + 0]; - zzzz[3] = src[sps->surface.width + 1]; + zzzz[2] = src[sps->surface.region->pitch + 0]; + zzzz[3] = src[sps->surface.region->pitch + 1]; } static void z32_write_quad_z(struct softpipe_surface *sps, GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) { - GLuint *dst = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z32); dst[0] = zzzz[0]; dst[1] = zzzz[1]; - dst[sps->surface.width + 0] = zzzz[2]; - dst[sps->surface.width + 1] = zzzz[3]; + dst[sps->surface.region->pitch + 0] = zzzz[2]; + dst[sps->surface.region->pitch + 1] = zzzz[3]; } static void @@ -145,28 +105,28 @@ z24s8_read_quad_z(struct softpipe_surface *sps, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) { const GLuint *src - = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_Z24_S8); zzzz[0] = src[0] >> 8; zzzz[1] = src[1] >> 8; - zzzz[2] = src[sps->surface.width + 0] >> 8; - zzzz[3] = src[sps->surface.width + 1] >> 8; + zzzz[2] = src[sps->surface.region->pitch + 0] >> 8; + zzzz[3] = src[sps->surface.region->pitch + 1] >> 8; } static void z24s8_write_quad_z(struct softpipe_surface *sps, GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) { - GLuint *dst = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_Z24_S8); assert(zzzz[0] <= 0xffffff); dst[0] = (dst[0] & 0xff) | (zzzz[0] << 8); dst[1] = (dst[1] & 0xff) | (zzzz[1] << 8); - dst += sps->surface.width; + dst += sps->surface.region->pitch; dst[0] = (dst[0] & 0xff) | (zzzz[2] << 8); dst[1] = (dst[1] & 0xff) | (zzzz[3] << 8); } @@ -176,32 +136,65 @@ z24s8_read_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) { const GLuint *src - = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_Z24_S8); ssss[0] = src[0] & 0xff; ssss[1] = src[1] & 0xff; - ssss[2] = src[sps->surface.width + 0] & 0xff; - ssss[3] = src[sps->surface.width + 1] & 0xff; + ssss[2] = src[sps->surface.region->pitch + 0] & 0xff; + ssss[3] = src[sps->surface.region->pitch + 1] & 0xff; } static void z24s8_write_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) { - GLuint *dst = (GLuint *) sps->surface.ptr + y * sps->surface.stride + x; + GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_Z24_S8); dst[0] = (dst[0] & 0xffffff00) | ssss[0]; dst[1] = (dst[1] & 0xffffff00) | ssss[1]; - dst += sps->surface.width; + dst += sps->surface.region->pitch; dst[0] = (dst[0] & 0xffffff00) | ssss[2]; dst[1] = (dst[1] & 0xffffff00) | ssss[3]; } +static void +s8_read_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) +{ + const GLubyte *src + = sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_U_S8); + + ssss[0] = src[0]; + ssss[1] = src[1]; + src += sps->surface.region->pitch; + ssss[2] = src[0]; + ssss[3] = src[1]; +} + +static void +s8_write_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) +{ + GLubyte *dst + = sps->surface.region->map + y * sps->surface.region->pitch + x; + + assert(sps->surface.format == PIPE_FORMAT_U_S8); + + dst[0] = ssss[0]; + dst[1] = ssss[1]; + dst += sps->surface.region->pitch; + dst[0] = ssss[2]; + dst[1] = ssss[3]; +} + + /** * Create a new software-based Z buffer. @@ -215,27 +208,6 @@ softpipe_new_z_surface(GLuint format) return NULL; sps->surface.format = format; - sps->surface.resize = z_resize; - sps->surface.buffer.map = z_map; - sps->surface.buffer.unmap = z_unmap; - - if (format == PIPE_FORMAT_U_Z16) { - sps->read_quad_z = z16_read_quad_z; - sps->write_quad_z = z16_write_quad_z; - } - else if (format == PIPE_FORMAT_U_Z32) { - sps->read_quad_z = z32_read_quad_z; - sps->write_quad_z = z32_write_quad_z; - } - else if (format == PIPE_FORMAT_Z24_S8) { - sps->read_quad_z = z24s8_read_quad_z; - sps->write_quad_z = z24s8_write_quad_z; - sps->read_quad_stencil = z24s8_read_quad_stencil; - sps->write_quad_stencil = z24s8_write_quad_stencil; - } - else { - assert(0); - } return sps; } diff --git a/src/mesa/sources b/src/mesa/sources index 32c2ff2350..3b820b71f0 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -157,6 +157,7 @@ VF_SOURCES = \ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_clear.c \ pipe/softpipe/sp_context.c \ + pipe/softpipe/sp_region.c \ pipe/softpipe/sp_quad.c \ pipe/softpipe/sp_quad_alpha_test.c \ pipe/softpipe/sp_quad_blend.c \ @@ -176,7 +177,7 @@ SOFTPIPE_SOURCES = \ pipe/softpipe/sp_state_sampler.c \ pipe/softpipe/sp_state_setup.c \ pipe/softpipe/sp_state_surface.c \ - pipe/softpipe/sp_z_surface.c \ + pipe/softpipe/sp_surface.c \ pipe/softpipe/sp_prim_setup.c DRAW_SOURCES = \ -- cgit v1.2.3 From 18fbf48a1e71cc26bd2d82a7c26275cf4a81c020 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 17:54:56 -0600 Subject: Obsolete. --- src/mesa/pipe/softpipe/sp_z_surface.c | 213 ---------------------------------- src/mesa/pipe/softpipe/sp_z_surface.h | 37 ------ 2 files changed, 250 deletions(-) delete mode 100644 src/mesa/pipe/softpipe/sp_z_surface.c delete mode 100644 src/mesa/pipe/softpipe/sp_z_surface.h (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_z_surface.c b/src/mesa/pipe/softpipe/sp_z_surface.c deleted file mode 100644 index 7ff1a0cbc8..0000000000 --- a/src/mesa/pipe/softpipe/sp_z_surface.c +++ /dev/null @@ -1,213 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * Software Z buffer/surface. - * - * \author Brian Paul - */ - - -#include "main/imports.h" -#include "pipe/p_state.h" -#include "pipe/p_defines.h" -#include "sp_context.h" -#include "sp_surface.h" -#include "sp_z_surface.h" - -static void -z16_read_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) -{ - const GLushort *src - = (GLushort *) sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_Z16); - - /* converting GLushort to GLuint: */ - zzzz[0] = src[0]; - zzzz[1] = src[1]; - zzzz[2] = src[sps->surface.region->pitch + 0]; - zzzz[3] = src[sps->surface.region->pitch + 1]; -} - -static void -z16_write_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) -{ - GLushort *dst = (GLushort *) sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_Z16); - - /* converting GLuint to GLushort: */ - dst[0] = zzzz[0]; - dst[1] = zzzz[1]; - dst[sps->surface.region->pitch + 0] = zzzz[2]; - dst[sps->surface.region->pitch + 1] = zzzz[3]; -} - -static void -z32_read_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) -{ - const GLuint *src - = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_Z32); - - zzzz[0] = src[0]; - zzzz[1] = src[1]; - zzzz[2] = src[sps->surface.region->pitch + 0]; - zzzz[3] = src[sps->surface.region->pitch + 1]; -} - -static void -z32_write_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) -{ - GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_Z32); - - dst[0] = zzzz[0]; - dst[1] = zzzz[1]; - dst[sps->surface.region->pitch + 0] = zzzz[2]; - dst[sps->surface.region->pitch + 1] = zzzz[3]; -} - -static void -z24s8_read_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) -{ - const GLuint *src - = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); - - zzzz[0] = src[0] >> 8; - zzzz[1] = src[1] >> 8; - zzzz[2] = src[sps->surface.region->pitch + 0] >> 8; - zzzz[3] = src[sps->surface.region->pitch + 1] >> 8; -} - -static void -z24s8_write_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) -{ - GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); - assert(zzzz[0] <= 0xffffff); - - dst[0] = (dst[0] & 0xff) | (zzzz[0] << 8); - dst[1] = (dst[1] & 0xff) | (zzzz[1] << 8); - dst += sps->surface.region->pitch; - dst[0] = (dst[0] & 0xff) | (zzzz[2] << 8); - dst[1] = (dst[1] & 0xff) | (zzzz[3] << 8); -} - -static void -z24s8_read_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) -{ - const GLuint *src - = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); - - ssss[0] = src[0] & 0xff; - ssss[1] = src[1] & 0xff; - ssss[2] = src[sps->surface.region->pitch + 0] & 0xff; - ssss[3] = src[sps->surface.region->pitch + 1] & 0xff; -} - -static void -z24s8_write_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) -{ - GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); - - dst[0] = (dst[0] & 0xffffff00) | ssss[0]; - dst[1] = (dst[1] & 0xffffff00) | ssss[1]; - dst += sps->surface.region->pitch; - dst[0] = (dst[0] & 0xffffff00) | ssss[2]; - dst[1] = (dst[1] & 0xffffff00) | ssss[3]; -} - - -static void -s8_read_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) -{ - const GLubyte *src - = sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_S8); - - ssss[0] = src[0]; - ssss[1] = src[1]; - src += sps->surface.region->pitch; - ssss[2] = src[0]; - ssss[3] = src[1]; -} - -static void -s8_write_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) -{ - GLubyte *dst - = sps->surface.region->map + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_S8); - - dst[0] = ssss[0]; - dst[1] = ssss[1]; - dst += sps->surface.region->pitch; - dst[0] = ssss[2]; - dst[1] = ssss[3]; -} - - - -/** - * Create a new software-based Z buffer. - * \param format one of the PIPE_FORMAT_z* formats - */ -struct softpipe_surface * -softpipe_new_z_surface(GLuint format) -{ - struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); - if (!sps) - return NULL; - - sps->surface.format = format; - - return sps; -} diff --git a/src/mesa/pipe/softpipe/sp_z_surface.h b/src/mesa/pipe/softpipe/sp_z_surface.h deleted file mode 100644 index 6a8d89a7c1..0000000000 --- a/src/mesa/pipe/softpipe/sp_z_surface.h +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef SP_Z_SURFACE_H -#define SP_Z_SURFACE_H - - -extern struct softpipe_surface * -softpipe_new_z_surface(GLuint format); - - -#endif /* SP_Z_SURFACE_H */ -- cgit v1.2.3 From d28661870a92f0beccd018855030146e01efb02e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 31 Jul 2007 17:55:32 -0600 Subject: sp_z_surface.h is dead --- src/mesa/main/renderbuffer.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c index 7c35575d12..9509df3159 100644 --- a/src/mesa/main/renderbuffer.c +++ b/src/mesa/main/renderbuffer.c @@ -49,7 +49,6 @@ #include "rbadaptors.h" -#include "pipe/softpipe/sp_z_surface.h" #include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -- cgit v1.2.3 From c5802911df668fb15cbe550219a8f361fbc83b82 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 31 Jul 2007 21:05:24 +0100 Subject: Remove intel_span.[ch] --- src/mesa/drivers/dri/i915pipe/Makefile | 1 - src/mesa/drivers/dri/i915pipe/intel_context.c | 3 - src/mesa/drivers/dri/i915pipe/intel_fbo.c | 6 +- src/mesa/drivers/dri/i915pipe/intel_screen.c | 1 - src/mesa/drivers/dri/i915pipe/intel_span.c | 402 -------------------------- src/mesa/drivers/dri/i915pipe/intel_span.h | 38 --- src/mesa/drivers/dri/i915pipe/intel_surface.c | 2 +- 7 files changed, 4 insertions(+), 449 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_span.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_span.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile index 857d3fc985..bf801f4bdd 100644 --- a/src/mesa/drivers/dri/i915pipe/Makefile +++ b/src/mesa/drivers/dri/i915pipe/Makefile @@ -25,7 +25,6 @@ DRIVER_SOURCES = \ intel_ioctl.c \ intel_rotate.c \ intel_screen.c \ - intel_span.c \ intel_state.c \ intel_surface.c \ intel_fbo.c \ diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index c8129c9e5f..0ccd22a4d0 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -50,7 +50,6 @@ #include "intel_buffers.h" #include "intel_tex.h" -#include "intel_span.h" #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" @@ -455,8 +454,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, break; } - /* Initialize swrast, tnl driver tables: */ - intelInitSpanFuncs(ctx); TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index 1470ce7d82..7b6906dea6 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -40,7 +40,7 @@ #include "intel_depthstencil.h" #include "intel_fbo.h" #include "intel_mipmap_tree.h" -#include "intel_span.h" +#include "intel_regions.h" #include "pipe/p_context.h" @@ -297,7 +297,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, rb->surface->region = irb->region; /* This sets the Get/PutRow/Value functions */ - intel_set_span_functions(&irb->Base); + // intel_set_span_functions(&irb->Base); return GL_TRUE; } @@ -509,7 +509,7 @@ intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage) irb->Base.Delete = intel_delete_renderbuffer; irb->Base.AllocStorage = intel_nop_alloc_storage; - intel_set_span_functions(&irb->Base); +// intel_set_span_functions(&irb->Base); irb->RenderToTexture = GL_TRUE; diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.c b/src/mesa/drivers/dri/i915pipe/intel_screen.c index 4f3a20a819..fad3a2c29b 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.c +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.c @@ -39,7 +39,6 @@ #include "intel_batchbuffer.h" #include "intel_buffers.h" #include "intel_tex.h" -#include "intel_span.h" #include "intel_ioctl.h" #include "intel_fbo.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_span.c b/src/mesa/drivers/dri/i915pipe/intel_span.c deleted file mode 100644 index c635ebe0af..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_span.c +++ /dev/null @@ -1,402 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "macros.h" -#include "mtypes.h" -#include "colormac.h" - -#include "intel_fbo.h" -#include "intel_screen.h" -#include "intel_span.h" -#include "intel_ioctl.h" -#include "intel_tex.h" - -#include "swrast/swrast.h" - -#include "pipe/p_context.h" - -/* - break intelWriteRGBASpan_ARGB8888 -*/ - -#undef DBG -#define DBG 0 - -#define LOCAL_VARS \ - struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ - const GLint yScale = irb->RenderToTexture ? 1 : -1; \ - const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ - GLubyte *buf = (GLubyte *) irb->pfMap; \ - GLuint p; \ - assert(irb->pfMap);\ - (void) p; - -/* There is just a single cliploop! - */ -#define HW_CLIPLOOP() \ - do { \ - int minx = 0; \ - int miny = 0; \ - int maxx = irb->Base.Width - 1; \ - int maxy = irb->Base.Height - 1; - -#define HW_ENDCLIPLOOP() \ - } while (0) - - -#define Y_FLIP(_y) ((_y) * yScale + yBias) - -#define HW_LOCK() - -#define HW_UNLOCK() - -/* 16 bit, RGB565 color spanline and pixel functions - */ -#define SPANTMP_PIXEL_FMT GL_RGB -#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5 - -#define TAG(x) intel##x##_RGB565 -#define TAG2(x,y) intel##x##_RGB565##y -#define GET_PTR(X,Y) (buf + ((Y) * irb->pfPitch + (X)) * 2) -#include "spantmp2.h" - -/* 32 bit, ARGB8888 color spanline and pixel functions - */ -#define SPANTMP_PIXEL_FMT GL_BGRA -#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV - -#define TAG(x) intel##x##_ARGB8888 -#define TAG2(x,y) intel##x##_ARGB8888##y -#define GET_PTR(X,Y) (buf + ((Y) * irb->pfPitch + (X)) * 4) -#include "spantmp2.h" - - -#define LOCAL_DEPTH_VARS \ - struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ - const GLuint pitch = irb->pfPitch/***XXX region->pitch*/; /* in pixels */ \ - const GLint yScale = irb->RenderToTexture ? 1 : -1; \ - const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ - char *buf = (char *) irb->pfMap/*XXX use region->map*/ ; - -#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS - -/** - ** 16-bit depthbuffer functions. - **/ -#define WRITE_DEPTH( _x, _y, d ) \ - ((GLushort *)buf)[(_x) + (_y) * pitch] = d; - -#define READ_DEPTH( d, _x, _y ) \ - d = ((GLushort *)buf)[(_x) + (_y) * pitch]; - - -#define TAG(x) intel##x##_z16 -#include "depthtmp.h" - - -/** - ** 24/8-bit interleaved depth/stencil functions - ** Note: we're actually reading back combined depth+stencil values. - ** The wrappers in main/depthstencil.c are used to extract the depth - ** and stencil values. - **/ -/* Change ZZZS -> SZZZ */ -#define WRITE_DEPTH( _x, _y, d ) { \ - GLuint tmp = ((d) >> 8) | ((d) << 24); \ - ((GLuint *)buf)[(_x) + (_y) * pitch] = tmp; \ -} - -/* Change SZZZ -> ZZZS */ -#define READ_DEPTH( d, _x, _y ) { \ - GLuint tmp = ((GLuint *)buf)[(_x) + (_y) * pitch]; \ - d = (tmp << 8) | (tmp >> 24); \ -} - -#define TAG(x) intel##x##_z24_s8 -#include "depthtmp.h" - - -/** - ** 8-bit stencil function (XXX FBO: This is obsolete) - **/ -#define WRITE_STENCIL( _x, _y, d ) { \ - GLuint tmp = ((GLuint *)buf)[(_x) + (_y) * pitch]; \ - tmp &= 0xffffff; \ - tmp |= ((d) << 24); \ - ((GLuint *) buf)[(_x) + (_y) * pitch] = tmp; \ -} - -#define READ_STENCIL( d, _x, _y ) \ - d = ((GLuint *)buf)[(_x) + (_y) * pitch] >> 24; - -#define TAG(x) intel##x##_z24_s8 -#include "stenciltmp.h" - - - -/** - * Map or unmap all the renderbuffers which we may need during - * software rendering. - * XXX in the future, we could probably convey extra information to - * reduce the number of mappings needed. I.e. if doing a glReadPixels - * from the depth buffer, we really only need one mapping. - * - * XXX Rewrite this function someday. - * We can probably just loop over all the renderbuffer attachments, - * map/unmap all of them, and not worry about the _ColorDrawBuffers - * _ColorReadBuffer, _DepthBuffer or _StencilBuffer fields. - */ -static void -intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) -{ - GLcontext *ctx = &intel->ctx; - GLuint i, j; - struct intel_renderbuffer *irb; - - /* color draw buffers */ - for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { - for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers[i]; j++) { - struct gl_renderbuffer *rb = - ctx->DrawBuffer->_ColorDrawBuffers[i][j]; - irb = intel_renderbuffer(rb); - if (irb) { - /* this is a user-created intel_renderbuffer */ - if (irb->region) { - if (map) - intel->pipe->region_map(intel->pipe, irb->region); - else - intel->pipe->region_unmap(intel->pipe, irb->region); - } - irb->pfMap = irb->region->map; - irb->pfPitch = irb->region->pitch; - } - } - } - - /* check for render to textures */ - for (i = 0; i < BUFFER_COUNT; i++) { - struct gl_renderbuffer_attachment *att = - ctx->DrawBuffer->Attachment + i; - struct gl_texture_object *tex = att->Texture; - if (tex) { - /* render to texture */ - ASSERT(att->Renderbuffer); - if (map) { - struct gl_texture_image *texImg; - texImg = tex->Image[att->CubeMapFace][att->TextureLevel]; - intel_tex_map_images(intel, intel_texture_object(tex)); - } - else { - intel_tex_unmap_images(intel, intel_texture_object(tex)); - } - } - } - - /* color read buffers */ - irb = intel_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); - if (irb && irb->region) { - if (map) - intel->pipe->region_map(intel->pipe, irb->region); - else - intel->pipe->region_unmap(intel->pipe, irb->region); - irb->pfMap = irb->region->map; - irb->pfPitch = irb->region->pitch; - } - - /* Account for front/back color page flipping. - * The span routines use the pfMap and pfPitch fields which will - * swap the front/back region map/pitch if we're page flipped. - * Do this after mapping, above, so the map field is valid. - */ -#if 0 - if (map && ctx->DrawBuffer->Name == 0) { - struct intel_renderbuffer *irbFront - = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_FRONT_LEFT); - struct intel_renderbuffer *irbBack - = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_BACK_LEFT); - if (irbBack) { - /* double buffered */ - if (intel->sarea->pf_current_page == 0) { - irbFront->pfMap = irbFront->region->map; - irbFront->pfPitch = irbFront->region->pitch; - irbBack->pfMap = irbBack->region->map; - irbBack->pfPitch = irbBack->region->pitch; - } - else { - irbFront->pfMap = irbBack->region->map; - irbFront->pfPitch = irbBack->region->pitch; - irbBack->pfMap = irbFront->region->map; - irbBack->pfPitch = irbFront->region->pitch; - } - } - } -#endif - - /* depth buffer (Note wrapper!) */ - if (ctx->DrawBuffer->_DepthBuffer) { - irb = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); - if (irb && irb->region) { - if (map) { - intel->pipe->region_map(intel->pipe, irb->region); - irb->pfMap = irb->region->map; - irb->pfPitch = irb->region->pitch; - } - else { - intel->pipe->region_unmap(intel->pipe, irb->region); - irb->pfMap = NULL; - irb->pfPitch = 0; - } - } - } - - /* stencil buffer (Note wrapper!) */ - if (ctx->DrawBuffer->_StencilBuffer) { - irb = intel_renderbuffer(ctx->DrawBuffer->_StencilBuffer->Wrapped); - if (irb && irb->region) { - if (map) { - intel->pipe->region_map(intel->pipe, irb->region); - irb->pfMap = irb->region->map; - irb->pfPitch = irb->region->pitch; - } - else { - intel->pipe->region_unmap(intel->pipe, irb->region); - irb->pfMap = NULL; - irb->pfPitch = 0; - } - } - } -} - - - -/** - * Prepare for softare rendering. Map current read/draw framebuffers' - * renderbuffes and all currently bound texture objects. - * - * Old note: Moved locking out to get reasonable span performance. - */ -void -intelSpanRenderStart(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - GLuint i; - - intelFinish(&intel->ctx); - LOCK_HARDWARE(intel); - -#if 0 - /* Just map the framebuffer and all textures. Bufmgr code will - * take care of waiting on the necessary fences: - */ - intel->pipe->region_map(intel->intelScreen, intel->front_region); - intel->pipe->region_map(intel->intelScreen, intel->back_region); - intel->pipe->region_map(intel->intelScreen, intel->intelScreen->depth_region); -#endif - - for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { - struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; - intel_tex_map_images(intel, intel_texture_object(texObj)); - } - } - - intel_map_unmap_buffers(intel, GL_TRUE); -} - -/** - * Called when done softare rendering. Unmap the buffers we mapped in - * the above function. - */ -void -intelSpanRenderFinish(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - GLuint i; - - _swrast_flush(ctx); - - /* Now unmap the framebuffer: - */ -#if 0 - intel->pipe->region_unmap(intel, intel->front_region); - intel->pipe->region_unmap(intel, intel->back_region); - intel->pipe->region_unmap(intel, intel->intelScreen->depth_region); -#endif - - for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { - struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; - intel_tex_unmap_images(intel, intel_texture_object(texObj)); - } - } - - intel_map_unmap_buffers(intel, GL_FALSE); - - UNLOCK_HARDWARE(intel); -} - - -void -intelInitSpanFuncs(GLcontext * ctx) -{ - struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx); - swdd->SpanRenderStart = intelSpanRenderStart; - swdd->SpanRenderFinish = intelSpanRenderFinish; -} - - -/** - * Plug in appropriate span read/write functions for the given renderbuffer. - * These are used for the software fallbacks. - */ -void -intel_set_span_functions(struct gl_renderbuffer *rb) -{ - if (rb->_ActualFormat == GL_RGB5) { - /* 565 RGB */ - intelInitPointers_RGB565(rb); - } - else if (rb->_ActualFormat == GL_RGBA8) { - /* 8888 RGBA */ - intelInitPointers_ARGB8888(rb); - } - else if (rb->_ActualFormat == GL_DEPTH_COMPONENT16) { - intelInitDepthPointers_z16(rb); - } - else if (rb->_ActualFormat == GL_DEPTH_COMPONENT24 || /* XXX FBO remove */ - rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) { - intelInitDepthPointers_z24_s8(rb); - } - else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) { /* XXX FBO remove */ - intelInitStencilPointers_z24_s8(rb); - } - else { - _mesa_problem(NULL, - "Unexpected _ActualFormat in intelSetSpanFunctions"); - } -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_span.h b/src/mesa/drivers/dri/i915pipe/intel_span.h deleted file mode 100644 index 5201f6d6c6..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_span.h +++ /dev/null @@ -1,38 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef _INTEL_SPAN_H -#define _INTEL_SPAN_H - -extern void intelInitSpanFuncs(GLcontext * ctx); - -extern void intelSpanRenderFinish(GLcontext * ctx); -extern void intelSpanRenderStart(GLcontext * ctx); - -extern void intel_set_span_functions(struct gl_renderbuffer *rb); - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_surface.c b/src/mesa/drivers/dri/i915pipe/intel_surface.c index 936c9cb5e7..d1a8960cc6 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_surface.c +++ b/src/mesa/drivers/dri/i915pipe/intel_surface.c @@ -10,7 +10,7 @@ #include "intel_context.h" #include "intel_buffers.h" -#include "intel_span.h" +#include "intel_regions.h" #include "intel_fbo.h" #include "pipe/p_state.h" -- cgit v1.2.3 From e5df28f39a776442d8aeb9530e7cfcfe351d675f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 31 Jul 2007 21:06:02 +0100 Subject: Remove unused file --- src/mesa/drivers/dri/i915pipe/intel_structs.h | 132 -------------------------- 1 file changed, 132 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_structs.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_structs.h b/src/mesa/drivers/dri/i915pipe/intel_structs.h deleted file mode 100644 index 522e3bd92c..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_structs.h +++ /dev/null @@ -1,132 +0,0 @@ -#ifndef INTEL_STRUCTS_H -#define INTEL_STRUCTS_H - -struct br0 { - GLuint length:8; - GLuint pad0:3; - GLuint dst_tiled:1; - GLuint pad1:8; - GLuint write_rgb:1; - GLuint write_alpha:1; - GLuint opcode:7; - GLuint client:3; -}; - - -struct br13 { - GLint dest_pitch:16; - GLuint rop:8; - GLuint color_depth:2; - GLuint pad1:3; - GLuint mono_source_transparency:1; - GLuint clipping_enable:1; - GLuint pad0:1; -}; - - - -/* This is an attempt to move some of the 2D interaction in this - * driver to using structs for packets rather than a bunch of #defines - * and dwords. - */ -struct xy_color_blit { - struct br0 br0; - struct br13 br13; - - struct { - GLuint dest_x1:16; - GLuint dest_y1:16; - } dw2; - - struct { - GLuint dest_x2:16; - GLuint dest_y2:16; - } dw3; - - GLuint dest_base_addr; - GLuint color; -}; - -struct xy_src_copy_blit { - struct br0 br0; - struct br13 br13; - - struct { - GLuint dest_x1:16; - GLuint dest_y1:16; - } dw2; - - struct { - GLuint dest_x2:16; - GLuint dest_y2:16; - } dw3; - - GLuint dest_base_addr; - - struct { - GLuint src_x1:16; - GLuint src_y1:16; - } dw5; - - struct { - GLint src_pitch:16; - GLuint pad:16; - } dw6; - - GLuint src_base_addr; -}; - -struct xy_setup_blit { - struct br0 br0; - struct br13 br13; - - struct { - GLuint clip_x1:16; - GLuint clip_y1:16; - } dw2; - - struct { - GLuint clip_x2:16; - GLuint clip_y2:16; - } dw3; - - GLuint dest_base_addr; - GLuint background_color; - GLuint foreground_color; - GLuint pattern_base_addr; -}; - - -struct xy_text_immediate_blit { - struct { - GLuint length:8; - GLuint pad2:3; - GLuint dst_tiled:1; - GLuint pad1:4; - GLuint byte_packed:1; - GLuint pad0:5; - GLuint opcode:7; - GLuint client:3; - } dw0; - - struct { - GLuint dest_x1:16; - GLuint dest_y1:16; - } dw1; - - struct { - GLuint dest_x2:16; - GLuint dest_y2:16; - } dw2; - - /* Src bitmap data follows as inline dwords. - */ -}; - - -#define CLIENT_2D 0x2 -#define OPCODE_XY_SETUP_BLT 0x1 -#define OPCODE_XY_COLOR_BLT 0x50 -#define OPCODE_XY_TEXT_IMMEDIATE_BLT 0x31 - -#endif -- cgit v1.2.3 From 028a08f563ea9b9039986bafbfaf095fd4a72818 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 31 Jul 2007 21:06:31 +0100 Subject: Remove unused file intel_render.c --- src/mesa/drivers/dri/i915pipe/intel_render.c | 244 --------------------------- 1 file changed, 244 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_render.c (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_render.c b/src/mesa/drivers/dri/i915pipe/intel_render.c deleted file mode 100644 index c8b6d308d9..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_render.c +++ /dev/null @@ -1,244 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* - * Render unclipped vertex buffers by emitting vertices directly to - * dma buffers. Use strip/fan hardware acceleration where possible. - * - */ -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" -#include "enums.h" - -#include "tnl/t_context.h" -#include "tnl/t_vertex.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_tris.h" -#include "intel_batchbuffer.h" -#include "intel_reg.h" - -/* - * Render unclipped vertex buffers by emitting vertices directly to - * dma buffers. Use strip/fan hardware primitives where possible. - * Try to simulate missing primitives with indexed vertices. - */ -#define HAVE_POINTS 0 /* Has it, but can't use because subpixel has to - * be adjusted for points on the INTEL/I845G - */ -#define HAVE_LINES 1 -#define HAVE_LINE_STRIPS 1 -#define HAVE_TRIANGLES 1 -#define HAVE_TRI_STRIPS 1 -#define HAVE_TRI_STRIP_1 0 /* has it, template can't use it yet */ -#define HAVE_TRI_FANS 1 -#define HAVE_POLYGONS 1 -#define HAVE_QUADS 0 -#define HAVE_QUAD_STRIPS 0 - -#define HAVE_ELTS 0 - -static GLuint hw_prim[GL_POLYGON + 1] = { - 0, - PRIM3D_LINELIST, - PRIM3D_LINESTRIP, - PRIM3D_LINESTRIP, - PRIM3D_TRILIST, - PRIM3D_TRISTRIP, - PRIM3D_TRIFAN, - 0, - 0, - PRIM3D_POLY -}; - -static const GLenum reduced_prim[GL_POLYGON + 1] = { - GL_POINTS, - GL_LINES, - GL_LINES, - GL_LINES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES -}; - -static const int scale_prim[GL_POLYGON + 1] = { - 0, /* fallback case */ - 1, - 2, - 2, - 1, - 3, - 3, - 0, /* fallback case */ - 0, /* fallback case */ - 3 -}; - - -static void -intelDmaPrimitive(struct intel_context *intel, GLenum prim) -{ - if (0) - fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim)); - INTEL_FIREVERTICES(intel); - intel->vtbl.reduced_primitive_state(intel, reduced_prim[prim]); - intelStartInlinePrimitive(intel, hw_prim[prim], INTEL_BATCH_CLIPRECTS); -} - - -#define LOCAL_VARS struct intel_context *intel = intel_context(ctx) -#define INIT( prim ) \ -do { \ - intelDmaPrimitive( intel, prim ); \ -} while (0) - -#define FLUSH() INTEL_FIREVERTICES(intel) - -#define GET_SUBSEQUENT_VB_MAX_VERTS() \ - ((intel->batch->size - 1500) / (intel->vertex_size*4)) -#define GET_CURRENT_VB_MAX_VERTS() GET_SUBSEQUENT_VB_MAX_VERTS() - -#define ALLOC_VERTS( nr ) \ - intelExtendInlinePrimitive( intel, (nr) * intel->vertex_size ) - -#define EMIT_VERTS( ctx, j, nr, buf ) \ - _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf ) - -#define TAG(x) intel_##x -#include "tnl_dd/t_dd_dmatmp.h" - - -/**********************************************************************/ -/* Render pipeline stage */ -/**********************************************************************/ - -/* Heuristic to choose between the two render paths: - */ -static GLboolean -choose_render(struct intel_context *intel, struct vertex_buffer *VB) -{ - int vertsz = intel->vertex_size; - int cost_render = 0; - int cost_fallback = 0; - int nr_prims = 0; - int nr_rprims = 0; - int nr_rverts = 0; - int rprim = intel->reduced_primitive; - int i = 0; - - for (i = 0; i < VB->PrimitiveCount; i++) { - GLuint prim = VB->Primitive[i].mode; - GLuint length = VB->Primitive[i].count; - - if (!length) - continue; - - nr_prims++; - nr_rverts += length * scale_prim[prim & PRIM_MODE_MASK]; - - if (reduced_prim[prim & PRIM_MODE_MASK] != rprim) { - nr_rprims++; - rprim = reduced_prim[prim & PRIM_MODE_MASK]; - } - } - - /* One point for each generated primitive: - */ - cost_render = nr_prims; - cost_fallback = nr_rprims; - - /* One point for every 1024 dwords (4k) of dma: - */ - cost_render += (vertsz * i) / 1024; - cost_fallback += (vertsz * nr_rverts) / 1024; - - if (0) - fprintf(stderr, "cost render: %d fallback: %d\n", - cost_render, cost_fallback); - - if (cost_render > cost_fallback) - return GL_FALSE; - - return GL_TRUE; -} - - -static GLboolean -intel_run_render(GLcontext * ctx, struct tnl_pipeline_stage *stage) -{ - struct intel_context *intel = intel_context(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *VB = &tnl->vb; - GLuint i; - - intel->vtbl.render_prevalidate( intel ); - - /* Don't handle clipping or indexed vertices. - */ - if (intel->RenderIndex != 0 || - !intel_validate_render(ctx, VB) || !choose_render(intel, VB)) { - return GL_TRUE; - } - - tnl->clipspace.new_inputs |= VERT_BIT_POS; - - tnl->Driver.Render.Start(ctx); - - for (i = 0; i < VB->PrimitiveCount; i++) { - GLuint prim = VB->Primitive[i].mode; - GLuint start = VB->Primitive[i].start; - GLuint length = VB->Primitive[i].count; - - if (!length) - continue; - - intel_render_tab_verts[prim & PRIM_MODE_MASK] (ctx, start, - start + length, prim); - } - - tnl->Driver.Render.Finish(ctx); - - INTEL_FIREVERTICES(intel); - - return GL_FALSE; /* finished the pipe */ -} - -const struct tnl_pipeline_stage _intel_render_stage = { - "intel render", - NULL, - NULL, - NULL, - NULL, - intel_run_render /* run */ -}; -- cgit v1.2.3 From e99b673cb062a2fead92d1d7d373926d148ade71 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 1 Aug 2007 15:56:23 +0100 Subject: Build libsoftpipe.a Each pipe driver will build to a .a library, as these will optionally be included in the various DRI drivers (this will make more sense once there is at least one hardware driver...). Not strictly necessary for softpipe, but want to minimize the differences between it and actual hw implementations. --- src/mesa/drivers/dri/Makefile.template | 4 +-- src/mesa/drivers/dri/i915pipe/Makefile | 3 ++ src/mesa/pipe/Makefile | 4 +++ src/mesa/pipe/Makefile.template | 58 ++++++++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/Makefile | 45 ++++++++++++++++++++++++-- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_prim_setup.h | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/sources | 26 --------------- src/mesa/vf/vf.h | 2 +- 10 files changed, 114 insertions(+), 33 deletions(-) create mode 100644 src/mesa/pipe/Makefile.template (limited to 'src') diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template index 43c0e912bf..00b09613ec 100644 --- a/src/mesa/drivers/dri/Makefile.template +++ b/src/mesa/drivers/dri/Makefile.template @@ -73,9 +73,9 @@ SHARED_INCLUDES = \ default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) -$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template +$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template $(TOP)/bin/mklib -noprefix -o $@ \ - $(OBJECTS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) + $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile index bf801f4bdd..8cc8d91264 100644 --- a/src/mesa/drivers/dri/i915pipe/Makefile +++ b/src/mesa/drivers/dri/i915pipe/Makefile @@ -6,6 +6,9 @@ LIBNAME = i915tex_dri.so MINIGLX_SOURCES = server/intel_dri.c +PIPE_DRIVERS = \ + $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a + DRIVER_SOURCES = \ intel_regions.c \ intel_buffer_objects.c \ diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile index 451911a354..ab7e09c991 100644 --- a/src/mesa/pipe/Makefile +++ b/src/mesa/pipe/Makefile @@ -1,2 +1,6 @@ default: cd .. ; make + cd softpipe ; make + +clean: + rm -f `find . -name \*.[oa]` \ No newline at end of file diff --git a/src/mesa/pipe/Makefile.template b/src/mesa/pipe/Makefile.template new file mode 100644 index 0000000000..7c64981d7b --- /dev/null +++ b/src/mesa/pipe/Makefile.template @@ -0,0 +1,58 @@ +# -*-makefile-*- + + +# We still have a dependency on the "dri" buffer manager. Most likely +# the interface can be reused in non-dri environments, and also as a +# frontend to simpler memory managers. +# +COMMON_SOURCES = + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(ASM_SOURCES:.S=.o) + + +### Include directories +INCLUDES = \ + -I. \ + -I$(TOP)/src/mesa/drivers/dri/common \ + -I$(TOP)/src/mesa/main \ + -I$(TOP)/src/mesa/pipe \ + -I$(TOP)/src/mesa + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + + +##### TARGETS ##### + +default:: depend symlinks $(LIBNAME) + + +$(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/mesa/pipe/Makefile.template + $(TOP)/bin/mklib -o $@ -static $(OBJECTS) + + +depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ + $(ASM_SOURCES) 2> /dev/null + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean: + -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) + -rm -f depend depend.bak + + +include depend diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 12a8bd0409..947983baf0 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -1,3 +1,44 @@ -default: - cd ../.. ; make +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = softpipe + +DRIVER_SOURCES = \ + sp_clear.c \ + sp_context.c \ + sp_prim_setup.c \ + sp_quad.c \ + sp_quad_alpha_test.c \ + sp_quad_blend.c \ + sp_quad_bufloop.c \ + sp_quad_colormask.c \ + sp_quad_coverage.c \ + sp_quad_depth_test.c \ + sp_quad_fs.c \ + sp_quad_occlusion.c \ + sp_quad_output.c \ + sp_quad_stencil.c \ + sp_quad_stipple.c \ + sp_region.c \ + sp_state_blend.c \ + sp_state_clip.c \ + sp_state_derived.c \ + sp_state_fs.c \ + sp_state_sampler.c \ + sp_state_setup.c \ + sp_state_surface.c \ + sp_surface.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../Makefile.template + +symlinks: + +default:: + cd ../.. ; make diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 002fe73b59..434e18308a 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -157,6 +157,7 @@ struct pipe_context *softpipe_create( void ) * Create drawing context and plug our rendering stage into it. */ softpipe->draw = draw_create(); + assert(softpipe->draw); draw_set_setup_stage(softpipe->draw, sp_draw_render_stage(softpipe)); sp_init_region_functions(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h index e5abefcd49..0180454a8d 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.h +++ b/src/mesa/pipe/softpipe/sp_prim_setup.h @@ -41,8 +41,8 @@ #include "imports.h" #if 0 #include "s_tri_public.h" -#endif #include "s_context.h" +#endif extern struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index d4acf402ce..1676d47d51 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -37,7 +37,7 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" -#include "core/tgsi_core.h" +#include "tgsi/core/tgsi_core.h" #if defined __GNUC__ #define ALIGNED_ATTRIBS 1 diff --git a/src/mesa/sources b/src/mesa/sources index 3b820b71f0..d9ee7266e5 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -154,31 +154,6 @@ VF_SOURCES = \ vf/vf_generic.c \ vf/vf_sse.c -SOFTPIPE_SOURCES = \ - pipe/softpipe/sp_clear.c \ - pipe/softpipe/sp_context.c \ - pipe/softpipe/sp_region.c \ - pipe/softpipe/sp_quad.c \ - pipe/softpipe/sp_quad_alpha_test.c \ - pipe/softpipe/sp_quad_blend.c \ - pipe/softpipe/sp_quad_bufloop.c \ - pipe/softpipe/sp_quad_colormask.c \ - pipe/softpipe/sp_quad_coverage.c \ - pipe/softpipe/sp_quad_depth_test.c \ - pipe/softpipe/sp_quad_fs.c \ - pipe/softpipe/sp_quad_occlusion.c \ - pipe/softpipe/sp_quad_output.c \ - pipe/softpipe/sp_quad_stipple.c \ - pipe/softpipe/sp_quad_stencil.c \ - pipe/softpipe/sp_state_blend.c \ - pipe/softpipe/sp_state_clip.c \ - pipe/softpipe/sp_state_derived.c \ - pipe/softpipe/sp_state_fs.c \ - pipe/softpipe/sp_state_sampler.c \ - pipe/softpipe/sp_state_setup.c \ - pipe/softpipe/sp_state_surface.c \ - pipe/softpipe/sp_surface.c \ - pipe/softpipe/sp_prim_setup.c DRAW_SOURCES = \ pipe/draw/draw_clip.c \ @@ -367,7 +342,6 @@ SOLO_SOURCES = \ $(MATH_SOURCES) \ $(VBO_SOURCES) \ $(VF_SOURCES) \ - $(SOFTPIPE_SOURCES) \ $(DRAW_SOURCES) \ $(TGSICORE_SOURCES) \ $(TGSIMESA_SOURCES) \ diff --git a/src/mesa/vf/vf.h b/src/mesa/vf/vf.h index fc988b9fb9..8ca3ab6d4a 100644 --- a/src/mesa/vf/vf.h +++ b/src/mesa/vf/vf.h @@ -29,7 +29,7 @@ #define VF_VERTEX_H #include "mtypes.h" -#include "m_vector.h" +#include "math/m_vector.h" enum { VF_ATTRIB_POS = 0, -- cgit v1.2.3 From fb206809ba2a131fd9034e10a00592f2d0d81fce Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 1 Aug 2007 12:58:38 -0600 Subject: Checkpoint: glClear changes - working, bug very rough. --- src/mesa/Makefile | 9 ++- src/mesa/drivers/dri/i915pipe/intel_buffers.c | 29 ++++++++- src/mesa/drivers/dri/i915pipe/intel_buffers.h | 5 ++ src/mesa/drivers/dri/i915pipe/intel_context.c | 2 + src/mesa/drivers/x11/xm_api.c | 2 + src/mesa/drivers/x11/xm_dd.c | 26 ++++++++ src/mesa/drivers/x11/xmesaP.h | 6 +- src/mesa/main/buffers.c | 10 +++ src/mesa/pipe/p_context.h | 1 + src/mesa/pipe/softpipe/sp_clear.c | 90 ++++++++++++++++++++++----- src/mesa/pipe/softpipe/sp_region.c | 55 ++++++++++++++++ 11 files changed, 213 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 6943219036..3055564341 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -11,6 +11,8 @@ GL_MINOR = 5 GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) +SOFTPIPE_LIB = $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a + .SUFFIXES : .cpp .c.o: @@ -110,11 +112,12 @@ stand-alone: depend subdirs $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) $(TOP)/$(LIB_DIR)/$ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ - $(MKLIB_OPTIONS) $(GL_LIB_DEPS) $(STAND_ALONE_OBJECTS) + $(MKLIB_OPTIONS) $(GL_LIB_DEPS) $(STAND_ALONE_OBJECTS) \ + $(SOFTPIPE_LIB) # Make the OSMesa library $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECTS) @@ -146,7 +149,7 @@ depend: $(ALL_SOURCES) subdirs: @ (cd x86 ; $(MAKE)) @ (cd x86-64 ; $(MAKE)) - + #(cd pipe/softpipe ; $(MAKE)) install: default $(INSTALL) -d $(INSTALL_DIR)/include/GL diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.c b/src/mesa/drivers/dri/i915pipe/intel_buffers.c index c03c009a3a..fb93151430 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.c +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.c @@ -40,6 +40,7 @@ #include "swrast/swrast.h" #include "vblank.h" +#include "pipe/p_context.h" /* This block can be removed when libdrm >= 2.3.1 is required */ @@ -298,9 +299,17 @@ intelWindowMoved(struct intel_context *intel) /** * Called by ctx->Driver.Clear. */ +#if 0 static void intelClear(GLcontext *ctx, GLbitfield mask) +#else +void +intelClear(struct pipe_context *pipe, + GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum) +#endif { + GLcontext *ctx = (GLcontext *) pipe->glctx; const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask); GLbitfield tri_mask = 0; GLbitfield blit_mask = 0; @@ -308,6 +317,13 @@ intelClear(GLcontext *ctx, GLbitfield mask) struct gl_framebuffer *fb = ctx->DrawBuffer; GLuint i; + GLbitfield mask; + + if (color) + mask = ctx->DrawBuffer->_ColorDrawBufferMask[0]; /*XXX temporary*/ + else + mask = 0x0; + if (0) fprintf(stderr, "%s\n", __FUNCTION__); @@ -323,7 +339,7 @@ intelClear(GLcontext *ctx, GLbitfield mask) } /* HW stencil */ - if (mask & BUFFER_BIT_STENCIL) { + if (stencil) { const struct pipe_region *stencilRegion = intel_get_rb_region(fb, BUFFER_STENCIL); if (stencilRegion) { @@ -340,7 +356,7 @@ intelClear(GLcontext *ctx, GLbitfield mask) } /* HW depth */ - if (mask & BUFFER_BIT_DEPTH) { + if (depth) { /* clear depth with whatever method is used for stencil (see above) */ if (tri_mask & BUFFER_BIT_STENCIL) tri_mask |= BUFFER_BIT_DEPTH; @@ -368,9 +384,14 @@ intelClear(GLcontext *ctx, GLbitfield mask) if (blit_mask) intelClearWithBlit(ctx, blit_mask); -#if 1 +#if 0 if (swrast_mask | tri_mask) _swrast_Clear(ctx, swrast_mask | tri_mask); +#else + softpipe_clear(pipe, GL_FALSE, + (swrast_mask | tri_mask) & BUFFER_BIT_DEPTH, + (swrast_mask | tri_mask) & BUFFER_BIT_STENCIL, + (swrast_mask | tri_mask) & BUFFER_BIT_ACCUM); #endif } @@ -786,7 +807,9 @@ intelReadBuffer(GLcontext * ctx, GLenum mode) void intelInitBufferFuncs(struct dd_function_table *functions) { +#if 0 functions->Clear = intelClear; +#endif functions->DrawBuffer = intelDrawBuffer; functions->ReadBuffer = intelReadBuffer; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.h b/src/mesa/drivers/dri/i915pipe/intel_buffers.h index 5834e39501..f0602eebae 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.h +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.h @@ -52,4 +52,9 @@ extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); extern void intelInitBufferFuncs(struct dd_function_table *functions); +extern void +intelClear(struct pipe_context *pipe, + GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum); + #endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index 0ccd22a4d0..6121f6bc60 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -420,6 +420,8 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->pipe = intel->ctx.st->pipe; intel->pipe->screen = intelScreen; + intel->pipe->glctx = ctx; + intel->pipe->clear = intelClear; intelScreen->pipe = intel->pipe; intel_init_region_functions(intel->pipe); diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index f20e8104fb..92d37085d1 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -82,6 +82,7 @@ #include "drivers/common/driverfuncs.h" #include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" #include "pipe/softpipe/sp_context.h" /** @@ -1572,6 +1573,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) st_create_context( mesaCtx, softpipe_create() ); + mesaCtx->st->pipe->clear = xmesa_clear; return c; } diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index 5725414856..0aa47d55e4 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -435,6 +435,32 @@ clear_buffers(GLcontext *ctx, GLbitfield buffers) } +void +xmesa_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum) +{ + struct softpipe_context *sp = (struct softpipe_context *) pipe; + if (color) { + GET_CURRENT_CONTEXT(ctx); + GLuint i; + softpipe_update_derived(sp); + for (i = 0; i < sp->framebuffer.num_cbufs; i++) { + struct pipe_surface *ps = sp->framebuffer.cbufs[i]; + struct xmesa_renderbuffer *xrb = (struct xmesa_renderbuffer *) ps->rb; + const GLint x = sp->cliprect.minx; + const GLint y = sp->cliprect.miny; + const GLint w = sp->cliprect.maxx - x; + const GLint h = sp->cliprect.maxy - y; + xrb->clearFunc(ctx, xrb, x, y, w, h); + } + color = GL_FALSE; + } + + softpipe_clear(pipe, color, depth, stencil, accum); +} + + + #ifndef XFree86Server /* XXX this was never tested in the Xserver environment */ diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 098b9218ae..fb1c1f8c3b 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -587,9 +587,13 @@ extern void xmesa_register_swrast_functions( GLcontext *ctx ); struct pipe_surface; +struct pipe_context; -struct pipe_surface * +extern struct pipe_surface * xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb); +extern void +xmesa_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum); #endif diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index 0e6ca8ea1c..bb019b5998 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -38,6 +38,8 @@ #include "fbobject.h" #include "state.h" +#include "state_tracker/st_draw.h" + #define BAD_MASK ~0u @@ -176,7 +178,15 @@ _mesa_Clear( GLbitfield mask ) } ASSERT(ctx->Driver.Clear); +#if 0 ctx->Driver.Clear(ctx, bufferMask); +#else + st_clear(ctx->st, + (mask & GL_COLOR_BUFFER_BIT) ? GL_TRUE : GL_FALSE, + (bufferMask & BUFFER_BIT_DEPTH) ? GL_TRUE : GL_FALSE, + (bufferMask & BUFFER_BIT_STENCIL) ? GL_TRUE : GL_FALSE, + (bufferMask & BUFFER_BIT_ACCUM) ? GL_TRUE : GL_FALSE); +#endif } } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index b48c7775de..8517d7ab68 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -188,6 +188,7 @@ struct pipe_context { GLuint flag); void *screen; /**< temporary */ + void *glctx; /**< temporary */ }; diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index e83bc053ef..40b1156715 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -30,42 +30,102 @@ */ +#include "pipe/p_defines.h" #include "sp_clear.h" #include "sp_context.h" #include "sp_surface.h" #include "colormac.h" +static GLuint +color_value(GLuint format, const GLfloat color[4]) +{ + GLubyte r, g, b, a; + + UNCLAMPED_FLOAT_TO_UBYTE(r, color[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, color[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, color[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, color[3]); + + switch (format) { + case PIPE_FORMAT_U_R8_G8_B8_A8: + return (r << 24) | (g << 16) | (b << 8) | a; + case PIPE_FORMAT_U_A8_R8_G8_B8: + return (a << 24) | (r << 16) | (g << 8) | b; + case PIPE_FORMAT_U_R5_G6_B5: + return ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); + default: + return 0; + } +} + + void softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) { const struct softpipe_context *softpipe = softpipe_context(pipe); - const GLint x = softpipe->scissor.minx; - const GLint y = softpipe->scissor.miny; - const GLint w = softpipe->scissor.maxx - x; - const GLint h = softpipe->scissor.maxy - y; + const GLint x = softpipe->cliprect.minx; + const GLint y = softpipe->cliprect.miny; + const GLint w = softpipe->cliprect.maxx - x; + const GLint h = softpipe->cliprect.maxy - y; if (color) { GLuint i; - GLubyte clr[4]; - - UNCLAMPED_FLOAT_TO_UBYTE(clr[0], softpipe->clear_color.color[0]); - UNCLAMPED_FLOAT_TO_UBYTE(clr[1], softpipe->clear_color.color[1]); - UNCLAMPED_FLOAT_TO_UBYTE(clr[2], softpipe->clear_color.color[2]); - UNCLAMPED_FLOAT_TO_UBYTE(clr[3], softpipe->clear_color.color[3]); - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { struct pipe_surface *ps = softpipe->framebuffer.cbufs[i]; - struct softpipe_surface *sps = softpipe_surface(ps); - GLint j; - for (j = 0; j < h; j++) { - sps->write_mono_row_ub(sps, w, x, y + j, clr); + + if (softpipe->blend.colormask == (PIPE_MASK_R | PIPE_MASK_G | + PIPE_MASK_B | PIPE_MASK_A)) { + /* no masking */ + GLuint clearVal = color_value(ps->format, + softpipe->clear_color.color); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal); + } + else { + /* masking */ + + /* + for (j = 0; j < h; j++) { + sps->write_mono_row_ub(sps, w, x, y + j, clr); + } + */ } } } if (depth) { + struct pipe_surface *ps = softpipe->framebuffer.zbuf; + GLuint clearVal; + + switch (ps->format) { + case PIPE_FORMAT_U_Z16: + clearVal = (GLuint) (softpipe->depth_test.clear * 65535.0); + break; + case PIPE_FORMAT_U_Z32: + clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffffff); + break; + case PIPE_FORMAT_Z24_S8: + clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffff); + break; + default: + assert(0); + } + + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal); } + if (stencil) { + struct pipe_surface *ps = softpipe->framebuffer.sbuf; + GLuint clearVal = softpipe->stencil.clear_value; + if (softpipe->stencil.write_mask[0] /*== 0xff*/) { + /* no masking */ + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal); + } + else if (softpipe->stencil.write_mask[0] != 0x0) { + /* masking */ + /* fill with quad funcs */ + assert(0); + } + } } diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 04ae5e94f9..4d8b35d3e7 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -92,6 +92,59 @@ sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) } + +static GLubyte * +get_pointer(struct pipe_region *dst, GLuint x, GLuint y) +{ + return dst->map + (y * dst->pitch + x) * dst->cpp; +} + + +static void +sp_region_fill(struct pipe_context *pipe, + struct pipe_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + GLuint width, GLuint height, GLuint value) +{ + GLuint i, j; + switch (dst->cpp) { + case 1: + { + GLubyte *row = get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: + { + GLushort *row = (GLushort *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + case 4: + { + GLuint *row = (GLuint *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + default: + assert(0); + + } +} + + void sp_init_region_functions(struct softpipe_context *sp) { @@ -101,5 +154,7 @@ sp_init_region_functions(struct softpipe_context *sp) sp->pipe.region_map = sp_region_map; sp->pipe.region_unmap = sp_region_unmap; + sp->pipe.region_fill = sp_region_fill; + /* XXX lots of other region functions needed... */ } -- cgit v1.2.3 From 5fd46065915d3958569ebb590104b69886352157 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 1 Aug 2007 13:04:58 -0600 Subject: s/Z24_S8/S8_Z24/ (stencil is in the high byte) --- src/mesa/main/renderbuffer.c | 2 +- src/mesa/pipe/p_defines.h | 2 +- src/mesa/pipe/softpipe/sp_clear.c | 2 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 2 +- src/mesa/pipe/softpipe/sp_surface.c | 61 +++++++++++++++-------------- 5 files changed, 36 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c index 9509df3159..d89704196a 100644 --- a/src/mesa/main/renderbuffer.c +++ b/src/mesa/main/renderbuffer.c @@ -1147,7 +1147,7 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->StencilBits = 8; if (!rb->surface) rb->surface = (struct pipe_surface *) - pipe->surface_alloc(pipe, PIPE_FORMAT_Z24_S8); + pipe->surface_alloc(pipe, PIPE_FORMAT_S8_Z24); pixelSize = sizeof(GLuint); break; case GL_COLOR_INDEX8_EXT: diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 58f01758e3..c92895342e 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -145,7 +145,7 @@ #define PIPE_FORMAT_U_Z16 10 /**< ushort Z/depth */ #define PIPE_FORMAT_U_Z32 11 /**< uint Z/depth */ #define PIPE_FORMAT_F_Z32 12 /**< float Z/depth */ -#define PIPE_FORMAT_Z24_S8 13 /**< 24-bit Z + 8-bit stencil */ +#define PIPE_FORMAT_S8_Z24 13 /**< 8-bit stencil + 24-bit Z */ #define PIPE_FORMAT_U_S8 14 /**< 8-bit stencil */ diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 40b1156715..6266d124be 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -105,7 +105,7 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, case PIPE_FORMAT_U_Z32: clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffffff); break; - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_S8_Z24: clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffff); break; default: diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index a26bd51d84..3a8df33e67 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -59,7 +59,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) */ if (sps->surface.format == PIPE_FORMAT_U_Z16) scale = 65535.0; - else if (sps->surface.format == PIPE_FORMAT_Z24_S8) + else if (sps->surface.format == PIPE_FORMAT_S8_Z24) scale = (float) ((1 << 24) - 1); else assert(0); /* XXX fix this someday */ diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index cf89d28941..819243ae6e 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -215,66 +215,69 @@ z32_write_quad_z(struct softpipe_surface *sps, } static void -z24s8_read_quad_z(struct softpipe_surface *sps, +s8z24_read_quad_z(struct softpipe_surface *sps, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) { + static const GLuint mask = 0x00ffffff; const GLuint *src = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - zzzz[0] = src[0] >> 8; - zzzz[1] = src[1] >> 8; + zzzz[0] = src[0] & mask; + zzzz[1] = src[1] & mask; src += sps->surface.region->pitch; - zzzz[2] = src[0] >> 8; - zzzz[3] = src[1] >> 8; + zzzz[2] = src[0] & mask; + zzzz[3] = src[1] & mask; } static void -z24s8_write_quad_z(struct softpipe_surface *sps, +s8z24_write_quad_z(struct softpipe_surface *sps, GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) { + static const GLuint mask = 0xff000000; GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); assert(zzzz[0] <= 0xffffff); - dst[0] = (dst[0] & 0xff) | (zzzz[0] << 8); - dst[1] = (dst[1] & 0xff) | (zzzz[1] << 8); + dst[0] = (dst[0] & mask) | zzzz[0]; + dst[1] = (dst[1] & mask) | zzzz[1]; dst += sps->surface.region->pitch; - dst[0] = (dst[0] & 0xff) | (zzzz[2] << 8); - dst[1] = (dst[1] & 0xff) | (zzzz[3] << 8); + dst[0] = (dst[0] & mask) | zzzz[2]; + dst[1] = (dst[1] & mask) | zzzz[3]; } static void -z24s8_read_quad_stencil(struct softpipe_surface *sps, +s8z24_read_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) { const GLuint *src = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - ssss[0] = src[0] & 0xff; - ssss[1] = src[1] & 0xff; + ssss[0] = src[0] >> 24; + ssss[1] = src[1] >> 24; src += sps->surface.region->pitch; - ssss[2] = src[0] & 0xff; - ssss[3] = src[1] & 0xff; + ssss[2] = src[0] >> 24; + ssss[3] = src[1] >> 24; } static void -z24s8_write_quad_stencil(struct softpipe_surface *sps, +s8z24_write_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) { + static const GLuint mask = 0x00ffffff; GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - dst[0] = (dst[0] & 0xffffff00) | ssss[0]; - dst[1] = (dst[1] & 0xffffff00) | ssss[1]; + dst[0] = (dst[0] & mask) | (ssss[0] << 24); + dst[1] = (dst[1] & mask) | (ssss[1] << 24); dst += sps->surface.region->pitch; - dst[0] = (dst[0] & 0xffffff00) | ssss[2]; - dst[1] = (dst[1] & 0xffffff00) | ssss[3]; + dst[0] = (dst[0] & mask) | (ssss[2] << 24); + dst[1] = (dst[1] & mask) | (ssss[3] << 24); } @@ -324,11 +327,11 @@ init_quad_funcs(struct softpipe_surface *sps) sps->read_quad_z = z32_read_quad_z; sps->write_quad_z = z32_write_quad_z; break; - case PIPE_FORMAT_Z24_S8: - sps->read_quad_z = z24s8_read_quad_z; - sps->write_quad_z = z24s8_write_quad_z; - sps->read_quad_stencil = z24s8_read_quad_stencil; - sps->write_quad_stencil = z24s8_write_quad_stencil; + case PIPE_FORMAT_S8_Z24: + sps->read_quad_z = s8z24_read_quad_z; + sps->write_quad_z = s8z24_write_quad_z; + sps->read_quad_stencil = s8z24_read_quad_stencil; + sps->write_quad_stencil = s8z24_write_quad_stencil; break; case PIPE_FORMAT_U_S8: sps->read_quad_stencil = s8_read_quad_stencil; -- cgit v1.2.3 From cddce60e7b14171f7afb204be5c131ce73ce01c5 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 1 Aug 2007 13:11:34 -0600 Subject: s/Z24_S8/S8_Z24/ --- src/mesa/drivers/dri/i915pipe/intel_surface.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_surface.c b/src/mesa/drivers/dri/i915pipe/intel_surface.c index d1a8960cc6..78ed995794 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_surface.c +++ b/src/mesa/drivers/dri/i915pipe/intel_surface.c @@ -90,7 +90,7 @@ read_quad_z24(struct softpipe_surface *sps, = (GLuint *) (sps->surface.region->map + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); /* extract lower three bytes */ zzzz[0] = src[0] & mask; @@ -109,7 +109,7 @@ write_quad_z24(struct softpipe_surface *sps, = (GLuint *) (sps->surface.region->map + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); /* write lower three bytes */ dst[0] = (dst[0] & mask) | zzzz[0]; @@ -128,7 +128,7 @@ read_quad_stencil(struct softpipe_surface *sps, const GLuint *src = (const GLuint *) (sps->surface.region->map + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); /* extract high byte */ ssss[0] = src[0] >> 24; @@ -147,7 +147,7 @@ write_quad_stencil(struct softpipe_surface *sps, GLuint *dst = (GLuint *) (sps->surface.region->map + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); /* write high byte */ dst[0] = (dst[0] & mask) | (ssss[0] << 24); @@ -182,7 +182,7 @@ intel_new_surface(GLuint intFormat) } else if (intFormat == GL_DEPTH24_STENCIL8_EXT) { - sps->surface.format = PIPE_FORMAT_Z24_S8; + sps->surface.format = PIPE_FORMAT_S8_Z24; sps->read_quad_z = read_quad_z24; sps->write_quad_z = write_quad_z24; sps->read_quad_stencil = read_quad_stencil; -- cgit v1.2.3 From 0eb02a1963a25f6994b730147d0613b03424c11e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 1 Aug 2007 14:46:07 -0600 Subject: More work on glClear. Add a 'mask' param to region_fill() to help with clearing combined Z/stencil buffers, glColorMask, etc. --- src/mesa/drivers/dri/i915pipe/intel_blit.c | 22 ++++- src/mesa/drivers/dri/i915pipe/intel_blit.h | 2 +- src/mesa/drivers/dri/i915pipe/intel_regions.c | 5 +- src/mesa/pipe/p_context.h | 3 +- src/mesa/pipe/softpipe/sp_clear.c | 127 +++++++++++++++++--------- src/mesa/pipe/softpipe/sp_region.c | 2 +- 6 files changed, 111 insertions(+), 50 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.c b/src/mesa/drivers/dri/i915pipe/intel_blit.c index 8e7f294b57..109d4face2 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_blit.c +++ b/src/mesa/drivers/dri/i915pipe/intel_blit.c @@ -209,9 +209,11 @@ intelEmitFillBlit(struct intel_context *intel, GLshort dst_pitch, struct _DriBufferObject *dst_buffer, GLuint dst_offset, - GLshort x, GLshort y, GLshort w, GLshort h, GLuint color) + GLshort x, GLshort y, GLshort w, GLshort h, + GLuint value, GLuint mask) { GLuint BR13, CMD; + GLboolean badMask = GL_FALSE; BATCH_LOCALS; dst_pitch *= cpp; @@ -222,16 +224,32 @@ intelEmitFillBlit(struct intel_context *intel, case 3: BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); CMD = XY_COLOR_BLT_CMD; + if ((mask & 0xffff) != 0xffff) + badMask = GL_TRUE; break; case 4: BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); +#if 0 CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | XY_COLOR_BLT_WRITE_RGB); +#else + CMD = XY_COLOR_BLT_CMD; + if ((mask & 0xff000000) == 0xff000000) + CMD |= XY_COLOR_BLT_WRITE_ALPHA; + else if (mask & 0xff000000) + badMask = GL_TRUE; + if ((mask & 0x00ffffff) == 0x00ffffff) + CMD |= XY_COLOR_BLT_WRITE_RGB; + else if (mask & 0x00ffffff) + badMask = GL_TRUE; +#endif break; default: return; } + assert(!badMask); + DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); @@ -243,7 +261,7 @@ intelEmitFillBlit(struct intel_context *intel, OUT_BATCH(((y + h) << 16) | (x + w)); OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); - OUT_BATCH(color); + OUT_BATCH(value); ADVANCE_BATCH(); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.h b/src/mesa/drivers/dri/i915pipe/intel_blit.h index 77686444fa..46c2594477 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_blit.h +++ b/src/mesa/drivers/dri/i915pipe/intel_blit.h @@ -56,7 +56,7 @@ extern void intelEmitFillBlit(struct intel_context *intel, struct _DriBufferObject *dst_buffer, GLuint dst_offset, GLshort x, GLshort y, - GLshort w, GLshort h, GLuint color); + GLshort w, GLshort h, GLuint value, GLuint mask); #endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_regions.c b/src/mesa/drivers/dri/i915pipe/intel_regions.c index e95e745adc..bdbc59e7ac 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_regions.c +++ b/src/mesa/drivers/dri/i915pipe/intel_regions.c @@ -343,7 +343,8 @@ intel_region_fill(struct pipe_context *pipe, struct pipe_region *dst, GLuint dst_offset, GLuint dstx, GLuint dsty, - GLuint width, GLuint height, GLuint color) + GLuint width, GLuint height, + GLuint value, GLuint mask) { intelScreenPrivate *intelScreen = pipe_screen(pipe); struct intel_context *intel = intelScreenContext(intelScreen); @@ -364,7 +365,7 @@ intel_region_fill(struct pipe_context *pipe, intelEmitFillBlit(intel, dst->cpp, dst->pitch, dst->buffer, dst_offset, - dstx, dsty, width, height, color); + dstx, dsty, width, height, value, mask); } /* Attach to a pbo, discarding our data. Effectively zero-copy upload diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 8517d7ab68..8e51daae06 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -172,7 +172,8 @@ struct pipe_context { struct pipe_region *dst, GLuint dst_offset, GLuint dstx, GLuint dsty, - GLuint width, GLuint height, GLuint color); + GLuint width, GLuint height, + GLuint value, GLuint mask); void (*region_cow)(struct pipe_context *pipe, struct pipe_region *region); diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 6266d124be..aa7601ab8c 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -59,7 +59,39 @@ color_value(GLuint format, const GLfloat color[4]) } } +static GLuint +color_mask(GLuint format, GLuint pipeMask) +{ + GLuint mask = 0x0; + switch (format) { + case PIPE_FORMAT_U_R8_G8_B8_A8: + if (pipeMask & PIPE_MASK_R) mask |= 0xff000000; + if (pipeMask & PIPE_MASK_G) mask |= 0x00ff0000; + if (pipeMask & PIPE_MASK_B) mask |= 0x0000ff00; + if (pipeMask & PIPE_MASK_A) mask |= 0x000000ff; + break; + case PIPE_FORMAT_U_A8_R8_G8_B8: + if (pipeMask & PIPE_MASK_R) mask |= 0x00ff0000; + if (pipeMask & PIPE_MASK_G) mask |= 0x0000ff00; + if (pipeMask & PIPE_MASK_B) mask |= 0x000000ff; + if (pipeMask & PIPE_MASK_A) mask |= 0xff000000; + break; + case PIPE_FORMAT_U_R5_G6_B5: + if (pipeMask & PIPE_MASK_R) mask |= 0xf800; + if (pipeMask & PIPE_MASK_G) mask |= 0x07e0; + if (pipeMask & PIPE_MASK_B) mask |= 0x001f; + if (pipeMask & PIPE_MASK_A) mask |= 0; + break; + default: + return 0; + } + return mask; +} + +/** + * XXX maybe this belongs in the GL state tracker... + */ void softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) @@ -74,58 +106,67 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLuint i; for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { struct pipe_surface *ps = softpipe->framebuffer.cbufs[i]; - - if (softpipe->blend.colormask == (PIPE_MASK_R | PIPE_MASK_G | - PIPE_MASK_B | PIPE_MASK_A)) { - /* no masking */ - GLuint clearVal = color_value(ps->format, - softpipe->clear_color.color); - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal); - } - else { - /* masking */ - - /* - for (j = 0; j < h; j++) { - sps->write_mono_row_ub(sps, w, x, y + j, clr); - } - */ - } + GLuint clearVal = color_value(ps->format, + softpipe->clear_color.color); + GLuint mask = color_mask(ps->format, softpipe->blend.colormask); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); } } - if (depth) { + if (depth && stencil && + softpipe->framebuffer.zbuf == softpipe->framebuffer.sbuf) { + /* clear Z and stencil together */ struct pipe_surface *ps = softpipe->framebuffer.zbuf; - GLuint clearVal; - - switch (ps->format) { - case PIPE_FORMAT_U_Z16: - clearVal = (GLuint) (softpipe->depth_test.clear * 65535.0); - break; - case PIPE_FORMAT_U_Z32: - clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffffff); - break; - case PIPE_FORMAT_S8_Z24: - clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffff); - break; - default: + if (ps->format == PIPE_FORMAT_S8_Z24) { + GLuint mask = (softpipe->stencil.write_mask[0] << 8) | 0xffffff; + GLuint clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffff); + clearVal |= (softpipe->stencil.clear_value << 24); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); + } + else { + /* XXX Z24_S8 format? */ assert(0); } - - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal); } + else { + /* separate Z and stencil */ + if (depth) { + struct pipe_surface *ps = softpipe->framebuffer.zbuf; + GLuint mask, clearVal; + + switch (ps->format) { + case PIPE_FORMAT_U_Z16: + clearVal = (GLuint) (softpipe->depth_test.clear * 65535.0); + mask = 0xffff; + break; + case PIPE_FORMAT_U_Z32: + clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffffff); + mask = 0xffffffff; + break; + case PIPE_FORMAT_S8_Z24: + clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffff); + mask = 0xffffff; + break; + default: + assert(0); + } - if (stencil) { - struct pipe_surface *ps = softpipe->framebuffer.sbuf; - GLuint clearVal = softpipe->stencil.clear_value; - if (softpipe->stencil.write_mask[0] /*== 0xff*/) { - /* no masking */ - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); } - else if (softpipe->stencil.write_mask[0] != 0x0) { - /* masking */ - /* fill with quad funcs */ - assert(0); + + if (stencil) { + struct pipe_surface *ps = softpipe->framebuffer.sbuf; + GLuint clearVal = softpipe->stencil.clear_value; + GLuint mask = 0xff; + if (softpipe->stencil.write_mask[0] /*== 0xff*/) { + /* no masking */ + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); + } + else if (softpipe->stencil.write_mask[0] != 0x0) { + /* masking */ + /* fill with quad funcs */ + assert(0); + } } } } diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 4d8b35d3e7..34fbcce8ca 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -105,7 +105,7 @@ sp_region_fill(struct pipe_context *pipe, struct pipe_region *dst, GLuint dst_offset, GLuint dstx, GLuint dsty, - GLuint width, GLuint height, GLuint value) + GLuint width, GLuint height, GLuint value, GLuint mask) { GLuint i, j; switch (dst->cpp) { -- cgit v1.2.3 From 0c4acfe15a0555a6f3da02ab17e58cc379df11fd Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 1 Aug 2007 15:11:59 -0600 Subject: Re-implement intelClear() in terms of softpipe_clear(). Pretty simple/small now. Note: softpipe_clear() should really be renamed to something like pipe_clear_with_blits() and put into a driver-indepedent module... --- src/mesa/drivers/dri/i915pipe/intel_blit.c | 4 +++ src/mesa/drivers/dri/i915pipe/intel_buffers.c | 34 +++++++++++++++++++++++--- src/mesa/drivers/dri/i915pipe/intel_context.c | 1 + src/mesa/pipe/softpipe/sp_clear.c | 35 ++++++++++++++++++++------- 4 files changed, 62 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.c b/src/mesa/drivers/dri/i915pipe/intel_blit.c index 109d4face2..4a5f58ed9e 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_blit.c +++ b/src/mesa/drivers/dri/i915pipe/intel_blit.c @@ -216,6 +216,10 @@ intelEmitFillBlit(struct intel_context *intel, GLboolean badMask = GL_FALSE; BATCH_LOCALS; + /* + printf("Emit fill blit value=0x%x mask=0x%x\n", value, mask); + */ + dst_pitch *= cpp; switch (cpp) { diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.c b/src/mesa/drivers/dri/i915pipe/intel_buffers.c index fb93151430..e39220fe47 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.c +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.c @@ -295,16 +295,16 @@ intelWindowMoved(struct intel_context *intel) - /** * Called by ctx->Driver.Clear. + * XXX NO LONGER USED - REMOVE IN NEAR FUTURE */ #if 0 static void intelClear(GLcontext *ctx, GLbitfield mask) #else -void -intelClear(struct pipe_context *pipe, +static void +OLD_intelClear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) #endif @@ -396,6 +396,34 @@ intelClear(struct pipe_context *pipe, } +/** + * Clear buffers. Called via pipe->clear(). + */ +void +intelClear(struct pipe_context *pipe, + GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum) +{ + GLcontext *ctx = (GLcontext *) pipe->glctx; + struct intel_context *intel = intel_context(ctx); + + /* XXX + * Examine stencil and color writemasks to determine if we can clear + * with blits. + */ + + intelFlush(&intel->ctx); + LOCK_HARDWARE(intel); + + softpipe_clear(pipe, color, depth, stencil, accum); + + intel_batchbuffer_flush(intel->batch); + + UNLOCK_HARDWARE(intel); +} + + + /* Emit wait for pending flips */ void intel_wait_flips(struct intel_context *intel, GLuint batch_flags) diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index 6121f6bc60..be8235d7d1 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -763,6 +763,7 @@ void LOCK_HARDWARE( struct intel_context *intel ) */ void UNLOCK_HARDWARE( struct intel_context *intel ) { + assert(intel->locked); intel->locked = 0; DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index aa7601ab8c..b3bbc96f30 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -90,7 +90,8 @@ color_mask(GLuint format, GLuint pipeMask) /** - * XXX maybe this belongs in the GL state tracker... + * XXX This should probaby be renamed to something like pipe_clear_with_blits() + * and moved into a device-independent pipe file. */ void softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, @@ -157,16 +158,32 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, if (stencil) { struct pipe_surface *ps = softpipe->framebuffer.sbuf; GLuint clearVal = softpipe->stencil.clear_value; - GLuint mask = 0xff; - if (softpipe->stencil.write_mask[0] /*== 0xff*/) { - /* no masking */ - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); - } - else if (softpipe->stencil.write_mask[0] != 0x0) { - /* masking */ - /* fill with quad funcs */ + GLuint mask = softpipe->stencil.write_mask[0]; + + switch (ps->format) { + case PIPE_FORMAT_S8_Z24: + clearVal = clearVal << 24; + mask = mask << 24; + break; + case PIPE_FORMAT_U_S8: + /* nothing */ + break; + default: assert(0); } + + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); } } + + if (accum) { + /* XXX there might be no notion of accum buffers in 'pipe'. + * Just implement them with a deep RGBA surface format... + */ + struct pipe_surface *ps = softpipe->framebuffer.abuf; + GLuint clearVal = 0x0; /* XXX FIX */ + GLuint mask = !0; + assert(ps); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); + } } -- cgit v1.2.3 From 54135597b8a23a7349f5fa3193642758c229777e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 1 Aug 2007 15:38:59 -0600 Subject: get cliprect bounds after softpipe_update_derived() --- src/mesa/pipe/softpipe/sp_clear.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index b3bbc96f30..a133b48891 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -34,6 +34,7 @@ #include "sp_clear.h" #include "sp_context.h" #include "sp_surface.h" +#include "sp_state.h" #include "colormac.h" @@ -98,10 +99,14 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) { const struct softpipe_context *softpipe = softpipe_context(pipe); - const GLint x = softpipe->cliprect.minx; - const GLint y = softpipe->cliprect.miny; - const GLint w = softpipe->cliprect.maxx - x; - const GLint h = softpipe->cliprect.maxy - y; + GLint x, y, w, h; + + softpipe_update_derived(softpipe); + + x = softpipe->cliprect.minx; + y = softpipe->cliprect.miny; + w = softpipe->cliprect.maxx - x; + h = softpipe->cliprect.maxy - y; if (color) { GLuint i; -- cgit v1.2.3 From 1ecc648398a51f734ef1e3b729595f41cedf29f9 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 1 Aug 2007 15:39:19 -0600 Subject: rearrange things in xmesa_clear() a bit --- src/mesa/drivers/x11/xm_dd.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index 0aa47d55e4..fec4f5d038 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -440,10 +440,15 @@ xmesa_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) { struct softpipe_context *sp = (struct softpipe_context *) pipe; + + /* Clear non-color buffers first. This will cause softpipe to + * re-validate the scissor/surface bounds. + */ + softpipe_clear(pipe, GL_FALSE, depth, stencil, accum); + if (color) { GET_CURRENT_CONTEXT(ctx); GLuint i; - softpipe_update_derived(sp); for (i = 0; i < sp->framebuffer.num_cbufs; i++) { struct pipe_surface *ps = sp->framebuffer.cbufs[i]; struct xmesa_renderbuffer *xrb = (struct xmesa_renderbuffer *) ps->rb; @@ -453,10 +458,8 @@ xmesa_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, const GLint h = sp->cliprect.maxy - y; xrb->clearFunc(ctx, xrb, x, y, w, h); } - color = GL_FALSE; } - softpipe_clear(pipe, color, depth, stencil, accum); } -- cgit v1.2.3 From 1b0e92b91a66c0188a870fb3ed6c20a8466b6ae9 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 1 Aug 2007 16:15:30 -0600 Subject: implement masking in sp_region_fill() --- src/mesa/pipe/softpipe/sp_region.c | 59 +++++++++++++++++++++++++++++++------- 1 file changed, 48 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 34fbcce8ca..1dbd1609e3 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -112,29 +112,66 @@ sp_region_fill(struct pipe_context *pipe, case 1: { GLubyte *row = get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; + if ((mask & 0xff) == 0xff) { + /* no masking */ + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + else { + value &= mask; + mask = ~mask; + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) { + row[j] = (row[j] & mask) | value; + } + row += dst->pitch; + } } } break; case 2: { GLushort *row = (GLushort *) get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; + if ((mask & 0xffff) == 0xffff) { + /* no masking */ + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + else { + value &= mask; + mask = ~mask; + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (row[j] & mask) | value; + row += dst->pitch; + } } } break; case 4: { GLuint *row = (GLuint *) get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; + if (mask == 0xffffffff) { + /* no masking */ + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + else { + value &= mask; + mask = ~mask; + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (row[j] & mask) | value; + row += dst->pitch; + } } } break; -- cgit v1.2.3 From 4f442d9ef5db42867c99a7288b4114a0340f73e6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 2 Aug 2007 13:59:31 +0100 Subject: Reroute some clear functionality. Still require the intelClear() call to flush batchbuffers. That will be removed later... --- src/mesa/drivers/dri/i915pipe/intel_buffers.c | 120 ++------------------------ src/mesa/drivers/dri/i915pipe/intel_context.c | 7 +- src/mesa/main/buffers.c | 8 -- src/mesa/pipe/softpipe/sp_clear.c | 70 +++++---------- src/mesa/sources | 1 + src/mesa/state_tracker/st_cb_clear.c | 89 +++++++++++++++++++ src/mesa/state_tracker/st_context.c | 1 + src/mesa/state_tracker/st_draw.c | 11 --- 8 files changed, 125 insertions(+), 182 deletions(-) create mode 100644 src/mesa/state_tracker/st_cb_clear.c (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.c b/src/mesa/drivers/dri/i915pipe/intel_buffers.c index e39220fe47..5e68a869bf 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.c +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.c @@ -295,110 +295,14 @@ intelWindowMoved(struct intel_context *intel) -/** - * Called by ctx->Driver.Clear. - * XXX NO LONGER USED - REMOVE IN NEAR FUTURE - */ -#if 0 -static void -intelClear(GLcontext *ctx, GLbitfield mask) -#else -static void -OLD_intelClear(struct pipe_context *pipe, - GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum) -#endif -{ - GLcontext *ctx = (GLcontext *) pipe->glctx; - const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask); - GLbitfield tri_mask = 0; - GLbitfield blit_mask = 0; - GLbitfield swrast_mask = 0; - struct gl_framebuffer *fb = ctx->DrawBuffer; - GLuint i; - - GLbitfield mask; - - if (color) - mask = ctx->DrawBuffer->_ColorDrawBufferMask[0]; /*XXX temporary*/ - else - mask = 0x0; - - if (0) - fprintf(stderr, "%s\n", __FUNCTION__); - - /* HW color buffers (front, back, aux, generic FBO, etc) */ - if (colorMask == ~0) { - /* clear all R,G,B,A */ - /* XXX FBO: need to check if colorbuffers are software RBOs! */ - blit_mask |= (mask & BUFFER_BITS_COLOR); - } - else { - /* glColorMask in effect */ - tri_mask |= (mask & BUFFER_BITS_COLOR); - } - - /* HW stencil */ - if (stencil) { - const struct pipe_region *stencilRegion - = intel_get_rb_region(fb, BUFFER_STENCIL); - if (stencilRegion) { - /* have hw stencil */ - if ((ctx->Stencil.WriteMask[0] & 0xff) != 0xff) { - /* not clearing all stencil bits, so use triangle clearing */ - tri_mask |= BUFFER_BIT_STENCIL; - } - else { - /* clearing all stencil bits, use blitting */ - blit_mask |= BUFFER_BIT_STENCIL; - } - } - } - - /* HW depth */ - if (depth) { - /* clear depth with whatever method is used for stencil (see above) */ - if (tri_mask & BUFFER_BIT_STENCIL) - tri_mask |= BUFFER_BIT_DEPTH; - else - blit_mask |= BUFFER_BIT_DEPTH; - } - - /* SW fallback clearing */ - swrast_mask = mask & ~tri_mask & ~blit_mask; - - for (i = 0; i < BUFFER_COUNT; i++) { - GLuint bufBit = 1 << i; - if ((blit_mask | tri_mask) & bufBit) { - if (!fb->Attachment[i].Renderbuffer->ClassID) { - blit_mask &= ~bufBit; - tri_mask &= ~bufBit; - swrast_mask |= bufBit; - } - } - } - - - intelFlush(ctx); /* XXX intelClearWithBlit also does this */ - - if (blit_mask) - intelClearWithBlit(ctx, blit_mask); - -#if 0 - if (swrast_mask | tri_mask) - _swrast_Clear(ctx, swrast_mask | tri_mask); -#else - softpipe_clear(pipe, GL_FALSE, - (swrast_mask | tri_mask) & BUFFER_BIT_DEPTH, - (swrast_mask | tri_mask) & BUFFER_BIT_STENCIL, - (swrast_mask | tri_mask) & BUFFER_BIT_ACCUM); -#endif -} - -/** - * Clear buffers. Called via pipe->clear(). - */ +/* XXX - kludge required because softpipe_clear uses + * region->fill(), which still calls intelBlit(!), but doesn't + * flush the batchbuffer. + * + * One way or another, that behaviour should stop, and then this + * function can go aawy. + */ void intelClear(struct pipe_context *pipe, GLboolean color, GLboolean depth, @@ -407,19 +311,9 @@ intelClear(struct pipe_context *pipe, GLcontext *ctx = (GLcontext *) pipe->glctx; struct intel_context *intel = intel_context(ctx); - /* XXX - * Examine stencil and color writemasks to determine if we can clear - * with blits. - */ - - intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); - softpipe_clear(pipe, color, depth, stencil, accum); intel_batchbuffer_flush(intel->batch); - - UNLOCK_HARDWARE(intel); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index be8235d7d1..0fc24c3b5a 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -417,7 +417,12 @@ intelCreateContext(const __GLcontextModes * mesaVis, */ st_create_context( &intel->ctx, softpipe_create() ); - + + /* KW: Not sure I like this - we should only be talking to the + * state_tracker. The pipe code will need some way of talking to + * us, eg for batchbuffer ioctls, and there will need to be a + * buffer manager interface. So, this is a temporary hack, right? + */ intel->pipe = intel->ctx.st->pipe; intel->pipe->screen = intelScreen; intel->pipe->glctx = ctx; diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index bb019b5998..eea443c03c 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -178,15 +178,7 @@ _mesa_Clear( GLbitfield mask ) } ASSERT(ctx->Driver.Clear); -#if 0 ctx->Driver.Clear(ctx, bufferMask); -#else - st_clear(ctx->st, - (mask & GL_COLOR_BUFFER_BIT) ? GL_TRUE : GL_FALSE, - (bufferMask & BUFFER_BIT_DEPTH) ? GL_TRUE : GL_FALSE, - (bufferMask & BUFFER_BIT_STENCIL) ? GL_TRUE : GL_FALSE, - (bufferMask & BUFFER_BIT_ACCUM) ? GL_TRUE : GL_FALSE); -#endif } } diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index a133b48891..09cc643003 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -60,53 +60,20 @@ color_value(GLuint format, const GLfloat color[4]) } } -static GLuint -color_mask(GLuint format, GLuint pipeMask) -{ - GLuint mask = 0x0; - switch (format) { - case PIPE_FORMAT_U_R8_G8_B8_A8: - if (pipeMask & PIPE_MASK_R) mask |= 0xff000000; - if (pipeMask & PIPE_MASK_G) mask |= 0x00ff0000; - if (pipeMask & PIPE_MASK_B) mask |= 0x0000ff00; - if (pipeMask & PIPE_MASK_A) mask |= 0x000000ff; - break; - case PIPE_FORMAT_U_A8_R8_G8_B8: - if (pipeMask & PIPE_MASK_R) mask |= 0x00ff0000; - if (pipeMask & PIPE_MASK_G) mask |= 0x0000ff00; - if (pipeMask & PIPE_MASK_B) mask |= 0x000000ff; - if (pipeMask & PIPE_MASK_A) mask |= 0xff000000; - break; - case PIPE_FORMAT_U_R5_G6_B5: - if (pipeMask & PIPE_MASK_R) mask |= 0xf800; - if (pipeMask & PIPE_MASK_G) mask |= 0x07e0; - if (pipeMask & PIPE_MASK_B) mask |= 0x001f; - if (pipeMask & PIPE_MASK_A) mask |= 0; - break; - default: - return 0; - } - return mask; -} - -/** - * XXX This should probaby be renamed to something like pipe_clear_with_blits() - * and moved into a device-independent pipe file. - */ void softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) { - const struct softpipe_context *softpipe = softpipe_context(pipe); + struct softpipe_context *softpipe = softpipe_context(pipe); GLint x, y, w, h; - softpipe_update_derived(softpipe); + softpipe_update_derived(softpipe); /* not needed?? */ - x = softpipe->cliprect.minx; - y = softpipe->cliprect.miny; - w = softpipe->cliprect.maxx - x; - h = softpipe->cliprect.maxy - y; + x = 0; + y = 0; + w = softpipe->framebuffer.cbufs[0]->width; + h = softpipe->framebuffer.cbufs[0]->height; if (color) { GLuint i; @@ -114,8 +81,7 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, struct pipe_surface *ps = softpipe->framebuffer.cbufs[i]; GLuint clearVal = color_value(ps->format, softpipe->clear_color.color); - GLuint mask = color_mask(ps->format, softpipe->blend.colormask); - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, ~0); } } @@ -124,10 +90,13 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, /* clear Z and stencil together */ struct pipe_surface *ps = softpipe->framebuffer.zbuf; if (ps->format == PIPE_FORMAT_S8_Z24) { - GLuint mask = (softpipe->stencil.write_mask[0] << 8) | 0xffffff; + GLuint mask = (softpipe->stencil.write_mask[0] << 8) | 0xffffff; GLuint clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffff); + + assert (mask == ~0); + clearVal |= (softpipe->stencil.clear_value << 24); - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, ~0); } else { /* XXX Z24_S8 format? */ @@ -138,33 +107,36 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, /* separate Z and stencil */ if (depth) { struct pipe_surface *ps = softpipe->framebuffer.zbuf; - GLuint mask, clearVal; + GLuint clearVal; switch (ps->format) { case PIPE_FORMAT_U_Z16: clearVal = (GLuint) (softpipe->depth_test.clear * 65535.0); - mask = 0xffff; break; case PIPE_FORMAT_U_Z32: clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffffff); - mask = 0xffffffff; break; case PIPE_FORMAT_S8_Z24: clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffff); - mask = 0xffffff; break; default: assert(0); } - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, ~0); } if (stencil) { struct pipe_surface *ps = softpipe->framebuffer.sbuf; GLuint clearVal = softpipe->stencil.clear_value; + + /* If this is not ~0, we shouldn't get here - clear should be + * done with geometry instead. + */ GLuint mask = softpipe->stencil.write_mask[0]; + assert((mask & 0xff) == 0xff); + switch (ps->format) { case PIPE_FORMAT_S8_Z24: clearVal = clearVal << 24; @@ -187,7 +159,7 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, */ struct pipe_surface *ps = softpipe->framebuffer.abuf; GLuint clearVal = 0x0; /* XXX FIX */ - GLuint mask = !0; + GLuint mask = ~0; assert(ps); pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); } diff --git a/src/mesa/sources b/src/mesa/sources index d9ee7266e5..d0fe3a979f 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -191,6 +191,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_stencil.c \ state_tracker/st_atom_stipple.c \ state_tracker/st_atom_viewport.c \ + state_tracker/st_cb_clear.c \ state_tracker/st_cb_program.c \ state_tracker/st_draw.c \ state_tracker/st_context.c \ diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c new file mode 100644 index 0000000000..d9cb83b8ad --- /dev/null +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -0,0 +1,89 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "st_context.h" +#include "glheader.h" +#include "macros.h" +#include "enums.h" +#include "st_context.h" +#include "st_atom.h" +#include "pipe/p_context.h" + + + +/* XXX: doesn't pick up the differences between front/back/left/right + * clears. Need to sort that out... + */ +static void st_clear(GLcontext *ctx, GLbitfield mask) +{ + struct st_context *st = ctx->st; + GLboolean color = (mask & BUFFER_BITS_COLOR) ? GL_TRUE : GL_FALSE; + GLboolean depth = (mask & BUFFER_BIT_DEPTH) ? GL_TRUE : GL_FALSE; + GLboolean stencil = (mask & BUFFER_BIT_STENCIL) ? GL_TRUE : GL_FALSE; + GLboolean accum = (mask & BUFFER_BIT_ACCUM) ? GL_TRUE : GL_FALSE; + GLboolean fullscreen = 1; /* :-) */ + + /* This makes sure the softpipe has the latest scissor, etc values */ + st_validate_state( st ); + + if (fullscreen) { + /* pipe->clear() should clear a particular surface, so that we + * can iterate over render buffers at this level and clear the + * ones GL is asking for. + * + * Will probably need something like pipe->clear_z_stencil() to + * cope with the special case of paired and unpaired z/stencil + * buffers, though could perhaps deal with them explicitly at + * this level. + */ + st->pipe->clear(st->pipe, color, depth, stencil, accum); + } + else { + /* Convert to geometry, etc: + */ + } +} + + +void st_init_cb_clear( struct st_context *st ) +{ + struct dd_function_table *functions = &st->ctx->Driver; + + functions->Clear = st_clear; +} + + +void st_destroy_cb_clear( struct st_context *st ) +{ +} + diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 6308e81a61..4d3f0ec4d3 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -58,6 +58,7 @@ struct st_context *st_create_context( GLcontext *ctx, st_init_atoms( st ); st_init_draw( st ); st_init_cb_program( st ); + st_init_cb_clear( st ); return st; } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index a424d1dd05..95fa43df4d 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -107,14 +107,3 @@ void st_destroy_draw( struct st_context *st ) } -/** XXX temporary here */ -void -st_clear(struct st_context *st, GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum) -{ - /* This makes sure the softpipe has the latest scissor, etc values */ - st_validate_state( st ); - - st->pipe->clear(st->pipe, color, depth, stencil, accum); -} - -- cgit v1.2.3 From d2bffbb7b5a0b6fe4af17a82cf3e58cce34394b0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 2 Aug 2007 14:50:08 +0100 Subject: Remove remnants of i915 texture-from-pbo code. --- .../drivers/dri/i915pipe/intel_buffer_objects.c | 57 +------- .../drivers/dri/i915pipe/intel_buffer_objects.h | 4 - src/mesa/drivers/dri/i915pipe/intel_regions.c | 158 +-------------------- src/mesa/drivers/dri/i915pipe/intel_screen.c | 1 - src/mesa/drivers/dri/i915pipe/intel_state.c | 2 - src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 33 +---- src/mesa/pipe/p_context.h | 9 -- src/mesa/pipe/p_state.h | 3 - 8 files changed, 3 insertions(+), 264 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c index eee5fa9bf4..fdcfdca9bb 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c +++ b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c @@ -62,39 +62,6 @@ intel_bufferobj_alloc(GLcontext * ctx, GLuint name, GLenum target) } -/* Break the COW tie to the region. The region gets to keep the data. - */ -void -intel_bufferobj_release_region(struct intel_context *intel, - struct intel_buffer_object *intel_obj) -{ - assert(intel_obj->region->buffer == intel_obj->buffer); - intel_obj->region->pbo = NULL; - intel_obj->region = NULL; - driBOUnReference(intel_obj->buffer); - intel_obj->buffer = NULL; - - /* This leads to a large number of buffer deletion/creation events. - * Currently the drm doesn't like that: - */ - driGenBuffers(intel->intelScreen->regionPool, - "buffer object", 1, &intel_obj->buffer, 64, 0, 0); - LOCK_HARDWARE(intel); - driBOData(intel_obj->buffer, intel_obj->Base.Size, NULL, 0); - UNLOCK_HARDWARE(intel); -} - -/* Break the COW tie to the region. Both the pbo and the region end - * up with a copy of the data. - */ -void -intel_bufferobj_cow(struct intel_context *intel, - struct intel_buffer_object *intel_obj) -{ - assert(intel_obj->region); - intel->pipe->region_cow(intel->pipe, intel_obj->region); -} - /** * Deallocate/free a vertex/pixel buffer object. @@ -103,15 +70,11 @@ intel_bufferobj_cow(struct intel_context *intel, static void intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj) { - struct intel_context *intel = intel_context(ctx); struct intel_buffer_object *intel_obj = intel_buffer_object(obj); assert(intel_obj); - if (intel_obj->region) { - intel_bufferobj_release_region(intel, intel_obj); - } - else if (intel_obj->buffer) { + if (intel_obj->buffer) { driDeleteBuffers(1, &intel_obj->buffer); } @@ -139,9 +102,6 @@ intel_bufferobj_data(GLcontext * ctx, intel_obj->Base.Size = size; intel_obj->Base.Usage = usage; - if (intel_obj->region) - intel_bufferobj_release_region(intel, intel_obj); - LOCK_HARDWARE(intel); driBOData(intel_obj->buffer, size, data, 0); UNLOCK_HARDWARE(intel); @@ -161,14 +121,10 @@ intel_bufferobj_subdata(GLcontext * ctx, GLsizeiptrARB size, const GLvoid * data, struct gl_buffer_object *obj) { - struct intel_context *intel = intel_context(ctx); struct intel_buffer_object *intel_obj = intel_buffer_object(obj); assert(intel_obj); - if (intel_obj->region) - intel_bufferobj_cow(intel, intel_obj); - driBOSubData(intel_obj->buffer, offset, size, data); } @@ -199,16 +155,12 @@ intel_bufferobj_map(GLcontext * ctx, GLenum target, GLenum access, struct gl_buffer_object *obj) { - struct intel_context *intel = intel_context(ctx); struct intel_buffer_object *intel_obj = intel_buffer_object(obj); /* XXX: Translate access to flags arg below: */ assert(intel_obj); - if (intel_obj->region) - intel_bufferobj_cow(intel, intel_obj); - obj->Pointer = driBOMap(intel_obj->buffer, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); return obj->Pointer; @@ -235,13 +187,6 @@ struct _DriBufferObject * intel_bufferobj_buffer(struct intel_context *intel, struct intel_buffer_object *intel_obj, GLuint flag) { - if (intel_obj->region) { - if (flag == INTEL_WRITE_PART) - intel_bufferobj_cow(intel, intel_obj); - else if (flag == INTEL_WRITE_FULL) - intel_bufferobj_release_region(intel, intel_obj); - } - return intel_obj->buffer; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h index a1f63c147b..0f648104ee 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h +++ b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h @@ -42,10 +42,6 @@ struct intel_buffer_object { struct gl_buffer_object Base; struct _DriBufferObject *buffer; /* the low-level buffer manager's buffer handle */ - - struct pipe_region *region; /* Is there a zero-copy texture - associated with this (pixel) - buffer object? */ }; diff --git a/src/mesa/drivers/dri/i915pipe/intel_regions.c b/src/mesa/drivers/dri/i915pipe/intel_regions.c index bdbc59e7ac..f2280ceb51 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_regions.c +++ b/src/mesa/drivers/dri/i915pipe/intel_regions.c @@ -75,10 +75,6 @@ intel_region_map(struct pipe_context *pipe, struct pipe_region *region) { DBG("%s\n", __FUNCTION__); if (!region->map_refcount++) { - if (region->pbo) { - pipe->region_cow(pipe, region); - } - region->map = driBOMap(region->buffer, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); } @@ -96,8 +92,6 @@ intel_region_unmap(struct pipe_context *pipe, struct pipe_region *region) } } -#undef TEST_CACHED_TEXTURES - static struct pipe_region * intel_region_alloc(struct pipe_context *pipe, GLuint cpp, GLuint pitch, GLuint height) @@ -115,13 +109,9 @@ intel_region_alloc(struct pipe_context *pipe, driGenBuffers(intelScreen->regionPool, "region", 1, ®ion->buffer, 64, -#ifdef TEST_CACHED_TEXTURES - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_BIND_CACHED | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, -#else 0, -#endif 0); + LOCK_HARDWARE(intel); driBOData(region->buffer, pitch * cpp * height, NULL, 0); UNLOCK_HARDWARE(intel); @@ -142,9 +132,6 @@ intel_region_release(struct pipe_context *pipe, struct pipe_region **region) if ((*region)->refcount == 0) { assert((*region)->map_refcount == 0); - if ((*region)->pbo) - (*region)->pbo->region = NULL; - (*region)->pbo = NULL; driBOUnReference((*region)->buffer); free(*region); } @@ -275,15 +262,6 @@ intel_region_data(struct pipe_context *pipe, if (intel == NULL) return; - if (dst->pbo) { - if (dstx == 0 && - dsty == 0 && width == dst->pitch && height == dst->height) - pipe->region_release_pbo(pipe, dst); - else - pipe->region_cow(pipe, dst); - } - - LOCK_HARDWARE(intel); _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, @@ -317,14 +295,6 @@ intel_region_copy(struct pipe_context *pipe, if (intel == NULL) return; - if (dst->pbo) { - if (dstx == 0 && - dsty == 0 && width == dst->pitch && height == dst->height) - pipe->region_release_pbo(pipe, dst); - else - pipe->region_cow(pipe, dst); - } - assert(src->cpp == dst->cpp); intelEmitCopyBlit(intel, @@ -354,141 +324,18 @@ intel_region_fill(struct pipe_context *pipe, if (intel == NULL) return; - if (dst->pbo) { - if (dstx == 0 && - dsty == 0 && width == dst->pitch && height == dst->height) - pipe->region_release_pbo(pipe, dst); - else - pipe->region_cow(pipe, dst); - } - intelEmitFillBlit(intel, dst->cpp, dst->pitch, dst->buffer, dst_offset, dstx, dsty, width, height, value, mask); } -/* Attach to a pbo, discarding our data. Effectively zero-copy upload - * the pbo's data. - */ -static void -intel_region_attach_pbo(struct pipe_context *pipe, - struct pipe_region *region, - struct intel_buffer_object *pbo) -{ - if (region->pbo == pbo) - return; - - /* If there is already a pbo attached, break the cow tie now. - * Don't call pipe_region_release_pbo() as that would - * unnecessarily allocate a new buffer we would have to immediately - * discard. - */ - if (region->pbo) { - region->pbo->region = NULL; - region->pbo = NULL; - } - - if (region->buffer) { - driDeleteBuffers(1, ®ion->buffer); - region->buffer = NULL; - } - - region->pbo = pbo; - region->pbo->region = region; - region->buffer = driBOReference(pbo->buffer); -} - -/* Break the COW tie to the pbo. The pbo gets to keep the data. - */ -static void -intel_region_release_pbo(struct pipe_context *pipe, - struct pipe_region *region) -{ - intelScreenPrivate *intelScreen = pipe_screen(pipe); - struct intel_context *intel = intelScreenContext(intelScreen); - - assert(region->buffer == region->pbo->buffer); - region->pbo->region = NULL; - region->pbo = NULL; - driBOUnReference(region->buffer); - region->buffer = NULL; - - driGenBuffers(intelScreen->regionPool, - "region", 1, ®ion->buffer, 64, 0, 0); - - LOCK_HARDWARE(intel); - driBOData(region->buffer, - region->cpp * region->pitch * region->height, NULL, 0); - UNLOCK_HARDWARE(intel); -} - -/* Break the COW tie to the pbo. Both the pbo and the region end up - * with a copy of the data. - */ -static void -intel_region_cow(struct pipe_context *pipe, struct pipe_region *region) -{ - intelScreenPrivate *intelScreen = pipe_screen(pipe); - struct intel_context *intel = intelScreenContext(intelScreen); - struct intel_buffer_object *pbo = region->pbo; - - if (intel == NULL) - return; - - pipe->region_release_pbo(pipe, region); - - assert(region->cpp * region->pitch * region->height == pbo->Base.Size); - - DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size); - - /* Now blit from the texture buffer to the new buffer: - */ - - intel_batchbuffer_flush(intel->batch); - - if (!intel->locked) { - LOCK_HARDWARE(intel); - intelEmitCopyBlit(intel, - region->cpp, - region->pitch, - region->buffer, 0, - region->pitch, - pbo->buffer, 0, - 0, 0, 0, 0, - region->pitch, region->height, - GL_COPY); - - intel_batchbuffer_flush(intel->batch); - UNLOCK_HARDWARE(intel); - } - else { - intelEmitCopyBlit(intel, - region->cpp, - region->pitch, - region->buffer, 0, - region->pitch, - pbo->buffer, 0, - 0, 0, 0, 0, - region->pitch, region->height, - GL_COPY); - - intel_batchbuffer_flush(intel->batch); - } -} static struct _DriBufferObject * intel_region_buffer(struct pipe_context *pipe, struct pipe_region *region, GLuint flag) { - if (region->pbo) { - if (flag == INTEL_WRITE_PART) - pipe->region_cow(pipe, region); - else if (flag == INTEL_WRITE_FULL) - pipe->region_release_pbo(pipe, region); - } - return region->buffer; } @@ -507,9 +354,6 @@ intel_init_region_functions(struct pipe_context *pipe) pipe->region_data = intel_region_data; pipe->region_copy = intel_region_copy; pipe->region_fill = intel_region_fill; - pipe->region_cow = intel_region_cow; - pipe->region_attach_pbo = intel_region_attach_pbo; - pipe->region_release_pbo = intel_region_release_pbo; pipe->region_buffer = intel_region_buffer; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.c b/src/mesa/drivers/dri/i915pipe/intel_screen.c index fad3a2c29b..565bfc5243 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.c +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.c @@ -101,7 +101,6 @@ intel_recreate_static(struct pipe_context *pipe, void *virtual, GLuint cpp, GLuint pitch, GLuint height) { - struct intel_context *intel = 0; if (region) { pipe->region_update_static(pipe, region, mem_type, offset, virtual, cpp, pitch, height); diff --git a/src/mesa/drivers/dri/i915pipe/intel_state.c b/src/mesa/drivers/dri/i915pipe/intel_state.c index fc98b6a12b..2455f177da 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_state.c +++ b/src/mesa/drivers/dri/i915pipe/intel_state.c @@ -245,8 +245,6 @@ intelDepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) static void intelRenderMode(GLcontext * ctx, GLenum mode) { - struct intel_context *intel = intel_context(ctx); - FALLBACK(intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER)); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c index 37307b49c5..b4e440e49a 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c @@ -254,38 +254,7 @@ try_pbo_zcopy(struct intel_context *intel, GLint width, GLint height, GLenum format, GLenum type, const void *pixels) { - struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); - GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; - - if (!pbo || - intel->ctx._ImageTransferState || - unpack->SkipPixels || unpack->SkipRows) { - _mesa_printf("%s: failure 1\n", __FUNCTION__); - return GL_FALSE; - } - - src_offset = (GLuint) pixels; - - if (unpack->RowLength > 0) - src_stride = unpack->RowLength; - else - src_stride = width; - - dst_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); - - dst_stride = intelImage->mt->pitch; - - if (src_stride != dst_stride || dst_offset != 0 || src_offset != 0) { - _mesa_printf("%s: failure 2\n", __FUNCTION__); - return GL_FALSE; - } - - intel->pipe->region_attach_pbo(intel->pipe, intelImage->mt->region, pbo); - - return GL_TRUE; + return GL_FALSE; } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 8e51daae06..4f5937b8f3 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -175,15 +175,6 @@ struct pipe_context { GLuint width, GLuint height, GLuint value, GLuint mask); - void (*region_cow)(struct pipe_context *pipe, struct pipe_region *region); - - void (*region_attach_pbo)(struct pipe_context *pipe, - struct pipe_region *region, - struct intel_buffer_object *pbo); - - void (*region_release_pbo)(struct pipe_context *pipe, - struct pipe_region *region); - struct _DriBufferObject *(*region_buffer)(struct pipe_context *pipe, struct pipe_region *region, GLuint flag); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 0767fc2fcb..d6ed514fb1 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -238,7 +238,6 @@ struct pipe_sampler_state struct _DriBufferObject; -struct intel_buffer_object; struct pipe_region { @@ -252,8 +251,6 @@ struct pipe_region GLuint map_refcount; /**< Reference count for mapping */ GLuint draw_offset; /**< Offset of drawing address within the region */ - - struct intel_buffer_object *pbo; /* zero-copy uploads */ }; -- cgit v1.2.3 From a4a0189ed6dcf232f544e2ca3ea0fc7f71a68862 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 2 Aug 2007 14:53:08 +0100 Subject: Remove intelClear() hack. Have added intel_batchbuffer_flush calls to the blit functions. We still shouldn't be calling back into this remnant intel code from the softpipe driver, though, so that will go too at some point. --- src/mesa/drivers/dri/i915pipe/intel_blit.c | 4 ++++ src/mesa/drivers/dri/i915pipe/intel_buffers.c | 25 ------------------------- src/mesa/drivers/dri/i915pipe/intel_buffers.h | 5 ----- src/mesa/drivers/dri/i915pipe/intel_context.c | 1 - 4 files changed, 4 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.c b/src/mesa/drivers/dri/i915pipe/intel_blit.c index 4a5f58ed9e..062cd2632c 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_blit.c +++ b/src/mesa/drivers/dri/i915pipe/intel_blit.c @@ -267,6 +267,8 @@ intelEmitFillBlit(struct intel_context *intel, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); OUT_BATCH(value); ADVANCE_BATCH(); + + intel_batchbuffer_flush(intel->batch); } @@ -385,6 +387,8 @@ intelEmitCopyBlit(struct intel_context *intel, src_offset + src_y * src_pitch); ADVANCE_BATCH(); } + + intel_batchbuffer_flush( intel->batch ); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.c b/src/mesa/drivers/dri/i915pipe/intel_buffers.c index 5e68a869bf..8fd785d20d 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.c +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.c @@ -296,28 +296,6 @@ intelWindowMoved(struct intel_context *intel) -/* XXX - kludge required because softpipe_clear uses - * region->fill(), which still calls intelBlit(!), but doesn't - * flush the batchbuffer. - * - * One way or another, that behaviour should stop, and then this - * function can go aawy. - */ -void -intelClear(struct pipe_context *pipe, - GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum) -{ - GLcontext *ctx = (GLcontext *) pipe->glctx; - struct intel_context *intel = intel_context(ctx); - - softpipe_clear(pipe, color, depth, stencil, accum); - - intel_batchbuffer_flush(intel->batch); -} - - - /* Emit wait for pending flips */ void intel_wait_flips(struct intel_context *intel, GLuint batch_flags) @@ -729,9 +707,6 @@ intelReadBuffer(GLcontext * ctx, GLenum mode) void intelInitBufferFuncs(struct dd_function_table *functions) { -#if 0 - functions->Clear = intelClear; -#endif functions->DrawBuffer = intelDrawBuffer; functions->ReadBuffer = intelReadBuffer; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.h b/src/mesa/drivers/dri/i915pipe/intel_buffers.h index f0602eebae..5834e39501 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.h +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.h @@ -52,9 +52,4 @@ extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); extern void intelInitBufferFuncs(struct dd_function_table *functions); -extern void -intelClear(struct pipe_context *pipe, - GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum); - #endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index 0fc24c3b5a..9c32ab0ddf 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -426,7 +426,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->pipe = intel->ctx.st->pipe; intel->pipe->screen = intelScreen; intel->pipe->glctx = ctx; - intel->pipe->clear = intelClear; intelScreen->pipe = intel->pipe; intel_init_region_functions(intel->pipe); -- cgit v1.2.3 From a6dd731656c7daa9a3d2a020f13001ac4d391a44 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 10:01:14 -0600 Subject: comment follow-up --- src/mesa/drivers/dri/i915pipe/intel_context.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index 9c32ab0ddf..9229ec35cc 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -422,6 +422,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, * state_tracker. The pipe code will need some way of talking to * us, eg for batchbuffer ioctls, and there will need to be a * buffer manager interface. So, this is a temporary hack, right? + * BP: Yes, a temporary hack so we can make jumps between old/new code. */ intel->pipe = intel->ctx.st->pipe; intel->pipe->screen = intelScreen; -- cgit v1.2.3 From 5936b439e88308a2eabb8a5570e91dd8fff89397 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 10:29:04 -0600 Subject: add PIPE_MASK_RGBA --- src/mesa/pipe/p_defines.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index c92895342e..4c29ec2c0e 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -75,6 +75,8 @@ #define PIPE_MASK_G 0x2 #define PIPE_MASK_B 0x4 #define PIPE_MASK_A 0x8 +#define PIPE_MASK_RGBA 0xf + /** * Inequality functions. Used for depth test, stencil compare, alpha -- cgit v1.2.3 From e415dced44bd716185e63df0c8c17f57de2b29e2 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 10:29:42 -0600 Subject: sketch out clearing with quads --- src/mesa/state_tracker/st_cb_clear.c | 84 +++++++++++++++++++++++++++++++++--- 1 file changed, 79 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index d9cb83b8ad..523a9b682c 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -31,17 +31,85 @@ * Brian Paul */ -#include "st_context.h" #include "glheader.h" #include "macros.h" #include "enums.h" -#include "st_context.h" #include "st_atom.h" +#include "st_context.h" +#include "st_cb_clear.h" +#include "st_public.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" + + + +/** + * Do glClear by drawing a quadrilateral. + */ +static void +clear_with_quad(GLcontext *ctx, + GLboolean color, GLboolean depth, + GLboolean stencil, GLboolean accum) +{ + struct st_context *st = ctx->st; + struct pipe_blend_state blend; + struct pipe_depth_state depth_test; + struct pipe_stencil_state stencil_test; + GLfloat z = ctx->Depth.Clear; + + /* depth state: always pass */ + memset(&depth_test, 0, sizeof(depth)); + if (depth) { + depth_test.enabled = 1; + depth_test.writemask = 1; + depth_test.func = PIPE_FUNC_ALWAYS; + } + st->pipe->set_depth_state(st->pipe, &depth_test); + + /* stencil state: always set to ref value */ + memset(&stencil_test, 0, sizeof(stencil)); + if (stencil) { + stencil_test.front_enabled = 1; + stencil_test.front_func = PIPE_FUNC_ALWAYS; + stencil_test.front_fail_op = PIPE_STENCIL_OP_REPLACE; + stencil_test.front_zpass_op = PIPE_STENCIL_OP_REPLACE; + stencil_test.front_zfail_op = PIPE_STENCIL_OP_REPLACE; + stencil_test.ref_value[0] = ctx->Stencil.Clear; + stencil_test.value_mask[0] = 0xff; + stencil_test.write_mask[0] = ctx->Stencil.WriteMask[0]; + } + st->pipe->set_stencil_state(st->pipe, &stencil_test); + + /* blend state: RGBA masking */ + memset(&blend, 0, sizeof(blend)); + if (color) { + if (ctx->Color.ColorMask[0]) + blend.colormask |= PIPE_MASK_R; + if (ctx->Color.ColorMask[1]) + blend.colormask |= PIPE_MASK_G; + if (ctx->Color.ColorMask[2]) + blend.colormask |= PIPE_MASK_B; + if (ctx->Color.ColorMask[3]) + blend.colormask |= PIPE_MASK_A; + if (st->ctx->Color.DitherFlag) + blend.dither = 1; + } + st->pipe->set_blend_state(st->pipe, &blend); + /* + * XXX Render quad here + */ -/* XXX: doesn't pick up the differences between front/back/left/right + /* Restore GL state */ + st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); +} + + + +/** + * Called via ctx->Driver.Clear() + * XXX: doesn't pick up the differences between front/back/left/right * clears. Need to sort that out... */ static void st_clear(GLcontext *ctx, GLbitfield mask) @@ -51,12 +119,17 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) GLboolean depth = (mask & BUFFER_BIT_DEPTH) ? GL_TRUE : GL_FALSE; GLboolean stencil = (mask & BUFFER_BIT_STENCIL) ? GL_TRUE : GL_FALSE; GLboolean accum = (mask & BUFFER_BIT_ACCUM) ? GL_TRUE : GL_FALSE; + GLboolean maskColor, maskStencil; GLboolean fullscreen = 1; /* :-) */ + GLuint stencilMax = 1 << ctx->DrawBuffer->_StencilBuffer->StencilBits; /* This makes sure the softpipe has the latest scissor, etc values */ st_validate_state( st ); - if (fullscreen) { + maskColor = st->state.blend.colormask != PIPE_MASK_RGBA; + maskStencil = ctx->Stencil.WriteMask[0] != stencilMax; + + if (fullscreen && !maskColor) { /* pipe->clear() should clear a particular surface, so that we * can iterate over render buffers at this level and clear the * ones GL is asking for. @@ -71,6 +144,7 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) else { /* Convert to geometry, etc: */ + clear_with_quad(ctx, color, depth, stencil, accum); } } -- cgit v1.2.3 From 51b300c54da63199103c2d7caaf27760d7243621 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 10:29:50 -0600 Subject: include st_cb_clear.h --- src/mesa/state_tracker/st_context.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 4d3f0ec4d3..cde1dd11ca 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -28,6 +28,7 @@ #include "imports.h" #include "st_public.h" #include "st_context.h" +#include "st_cb_clear.h" #include "st_atom.h" #include "st_draw.h" #include "st_program.h" -- cgit v1.2.3 From 107206bbbe6f110e5d280ecb306bec0f764c834d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 10:30:32 -0600 Subject: New header file. --- src/mesa/state_tracker/st_cb_clear.h | 37 ++++++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_clear.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.h b/src/mesa/state_tracker/st_cb_clear.h new file mode 100644 index 0000000000..32086971b5 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_clear.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_CLEAR_H +#define ST_CB_CLEAR_H + +extern void st_init_cb_clear( struct st_context *st ); + +extern void st_destroy_cb_clear( struct st_context *st ); + +#endif /* ST_CB_CLEAR_H */ + -- cgit v1.2.3 From 3dfe125861df183fcc325c516639f80b86bfbc19 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 2 Aug 2007 14:56:44 +0100 Subject: Remove intel_state.c, intel_rotate.[ch] --- src/mesa/drivers/dri/i915pipe/Makefile | 2 - src/mesa/drivers/dri/i915pipe/intel_context.c | 1 - src/mesa/drivers/dri/i915pipe/intel_context.h | 62 ------- src/mesa/drivers/dri/i915pipe/intel_rotate.c | 237 ----------------------- src/mesa/drivers/dri/i915pipe/intel_rotate.h | 39 ---- src/mesa/drivers/dri/i915pipe/intel_screen.h | 4 - src/mesa/drivers/dri/i915pipe/intel_state.c | 258 -------------------------- 7 files changed, 603 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_rotate.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_rotate.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_state.c (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile index 8cc8d91264..2331e81b1c 100644 --- a/src/mesa/drivers/dri/i915pipe/Makefile +++ b/src/mesa/drivers/dri/i915pipe/Makefile @@ -26,9 +26,7 @@ DRIVER_SOURCES = \ intel_blit.c \ intel_context.c \ intel_ioctl.c \ - intel_rotate.c \ intel_screen.c \ - intel_state.c \ intel_surface.c \ intel_fbo.c \ intel_depthstencil.c \ diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index 9229ec35cc..b66dbe71e5 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -330,7 +330,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) intelInitTextureFuncs(functions); intelInitPixelFuncs(functions); - intelInitStateFuncs(functions); intelInitBufferFuncs(functions); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.h b/src/mesa/drivers/dri/i915pipe/intel_context.h index 9d331e41b5..cac9fd7b49 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.h +++ b/src/mesa/drivers/dri/i915pipe/intel_context.h @@ -334,72 +334,10 @@ extern void intelFlush(GLcontext * ctx); extern void intelInitDriverFunctions(struct dd_function_table *functions); -/* ================================================================ - * intel_state.c: - */ -extern void intelInitStateFuncs(struct dd_function_table *functions); - -#define COMPAREFUNC_ALWAYS 0 -#define COMPAREFUNC_NEVER 0x1 -#define COMPAREFUNC_LESS 0x2 -#define COMPAREFUNC_EQUAL 0x3 -#define COMPAREFUNC_LEQUAL 0x4 -#define COMPAREFUNC_GREATER 0x5 -#define COMPAREFUNC_NOTEQUAL 0x6 -#define COMPAREFUNC_GEQUAL 0x7 - -#define STENCILOP_KEEP 0 -#define STENCILOP_ZERO 0x1 -#define STENCILOP_REPLACE 0x2 -#define STENCILOP_INCRSAT 0x3 -#define STENCILOP_DECRSAT 0x4 -#define STENCILOP_INCR 0x5 -#define STENCILOP_DECR 0x6 -#define STENCILOP_INVERT 0x7 - -#define LOGICOP_CLEAR 0 -#define LOGICOP_NOR 0x1 -#define LOGICOP_AND_INV 0x2 -#define LOGICOP_COPY_INV 0x3 -#define LOGICOP_AND_RVRSE 0x4 -#define LOGICOP_INV 0x5 -#define LOGICOP_XOR 0x6 -#define LOGICOP_NAND 0x7 -#define LOGICOP_AND 0x8 -#define LOGICOP_EQUIV 0x9 -#define LOGICOP_NOOP 0xa -#define LOGICOP_OR_INV 0xb -#define LOGICOP_COPY 0xc -#define LOGICOP_OR_RVRSE 0xd -#define LOGICOP_OR 0xe -#define LOGICOP_SET 0xf - -#define BLENDFACT_ZERO 0x01 -#define BLENDFACT_ONE 0x02 -#define BLENDFACT_SRC_COLR 0x03 -#define BLENDFACT_INV_SRC_COLR 0x04 -#define BLENDFACT_SRC_ALPHA 0x05 -#define BLENDFACT_INV_SRC_ALPHA 0x06 -#define BLENDFACT_DST_ALPHA 0x07 -#define BLENDFACT_INV_DST_ALPHA 0x08 -#define BLENDFACT_DST_COLR 0x09 -#define BLENDFACT_INV_DST_COLR 0x0a -#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b -#define BLENDFACT_CONST_COLOR 0x0c -#define BLENDFACT_INV_CONST_COLOR 0x0d -#define BLENDFACT_CONST_ALPHA 0x0e -#define BLENDFACT_INV_CONST_ALPHA 0x0f -#define BLENDFACT_MASK 0x0f #define MI_BATCH_BUFFER_END (0xA<<23) -extern int intel_translate_compare_func(GLenum func); -extern int intel_translate_stencil_op(GLenum op); -extern int intel_translate_blend_factor(GLenum factor); -extern int intel_translate_logic_op(GLenum opcode); - - /*====================================================================== * Inline conversion functions. * These are better-typed than the macros used previously: diff --git a/src/mesa/drivers/dri/i915pipe/intel_rotate.c b/src/mesa/drivers/dri/i915pipe/intel_rotate.c deleted file mode 100644 index 12d98c4ad2..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_rotate.c +++ /dev/null @@ -1,237 +0,0 @@ - -/** - * Routines for simple 2D->2D transformations for rotated, flipped screens. - * - * XXX This code is not intel-specific. Move it into a common/utility - * someday. - */ - -#include "intel_rotate.h" - -#define MIN2(A, B) ( ((A) < (B)) ? (A) : (B) ) - -#define ABS(A) ( ((A) < 0) ? -(A) : (A) ) - - -void -matrix23Set(struct matrix23 *m, - int m00, int m01, int m02, int m10, int m11, int m12) -{ - m->m00 = m00; - m->m01 = m01; - m->m02 = m02; - m->m10 = m10; - m->m11 = m11; - m->m12 = m12; -} - - -/* - * Transform (x,y) coordinate by the given matrix. - */ -void -matrix23TransformCoordf(const struct matrix23 *m, float *x, float *y) -{ - const float x0 = *x; - const float y0 = *y; - - *x = m->m00 * x0 + m->m01 * y0 + m->m02; - *y = m->m10 * x0 + m->m11 * y0 + m->m12; -} - - -void -matrix23TransformCoordi(const struct matrix23 *m, int *x, int *y) -{ - const int x0 = *x; - const int y0 = *y; - - *x = m->m00 * x0 + m->m01 * y0 + m->m02; - *y = m->m10 * x0 + m->m11 * y0 + m->m12; -} - - -/* - * Transform a width and height by the given matrix. - * XXX this could be optimized quite a bit. - */ -void -matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist) -{ - int x0 = 0, y0 = 0; - int x1 = *xDist, y1 = 0; - int x2 = 0, y2 = *yDist; - matrix23TransformCoordi(m, &x0, &y0); - matrix23TransformCoordi(m, &x1, &y1); - matrix23TransformCoordi(m, &x2, &y2); - - *xDist = (x1 - x0) + (x2 - x0); - *yDist = (y1 - y0) + (y2 - y0); - - if (*xDist < 0) - *xDist = -*xDist; - if (*yDist < 0) - *yDist = -*yDist; -} - - -/** - * Transform the rect defined by (x, y, w, h) by m. - */ -void -matrix23TransformRect(const struct matrix23 *m, int *x, int *y, int *w, - int *h) -{ - int x0 = *x, y0 = *y; - int x1 = *x + *w, y1 = *y; - int x2 = *x + *w, y2 = *y + *h; - int x3 = *x, y3 = *y + *h; - matrix23TransformCoordi(m, &x0, &y0); - matrix23TransformCoordi(m, &x1, &y1); - matrix23TransformCoordi(m, &x2, &y2); - matrix23TransformCoordi(m, &x3, &y3); - *w = ABS(x1 - x0) + ABS(x2 - x1); - /**w = ABS(*w);*/ - *h = ABS(y1 - y0) + ABS(y2 - y1); - /**h = ABS(*h);*/ - *x = MIN2(x0, x1); - *x = MIN2(*x, x2); - *y = MIN2(y0, y1); - *y = MIN2(*y, y2); -} - - -/* - * Make rotation matrix for width X height screen. - */ -void -matrix23Rotate(struct matrix23 *m, int width, int height, int angle) -{ - switch (angle) { - case 0: - matrix23Set(m, 1, 0, 0, 0, 1, 0); - break; - case 90: - matrix23Set(m, 0, 1, 0, -1, 0, width); - break; - case 180: - matrix23Set(m, -1, 0, width, 0, -1, height); - break; - case 270: - matrix23Set(m, 0, -1, height, 1, 0, 0); - break; - default: - /*abort() */ ; - } -} - - -/* - * Make flip/reflection matrix for width X height screen. - */ -void -matrix23Flip(struct matrix23 *m, int width, int height, int xflip, int yflip) -{ - if (xflip) { - m->m00 = -1; - m->m01 = 0; - m->m02 = width - 1; - } - else { - m->m00 = 1; - m->m01 = 0; - m->m02 = 0; - } - if (yflip) { - m->m10 = 0; - m->m11 = -1; - m->m12 = height - 1; - } - else { - m->m10 = 0; - m->m11 = 1; - m->m12 = 0; - } -} - - -/* - * result = a * b - */ -void -matrix23Multiply(struct matrix23 *result, - const struct matrix23 *a, const struct matrix23 *b) -{ - result->m00 = a->m00 * b->m00 + a->m01 * b->m10; - result->m01 = a->m00 * b->m01 + a->m01 * b->m11; - result->m02 = a->m00 * b->m02 + a->m01 * b->m12 + a->m02; - - result->m10 = a->m10 * b->m00 + a->m11 * b->m10; - result->m11 = a->m10 * b->m01 + a->m11 * b->m11; - result->m12 = a->m10 * b->m02 + a->m11 * b->m12 + a->m12; -} - - -#if 000 - -#include - -int -main(int argc, char *argv[]) -{ - int width = 500, height = 400; - int rot; - int fx = 0, fy = 0; /* flip x and/or y ? */ - int coords[4][2]; - - /* four corner coords to test with */ - coords[0][0] = 0; - coords[0][1] = 0; - coords[1][0] = width - 1; - coords[1][1] = 0; - coords[2][0] = width - 1; - coords[2][1] = height - 1; - coords[3][0] = 0; - coords[3][1] = height - 1; - - - for (rot = 0; rot < 360; rot += 90) { - struct matrix23 rotate, flip, m; - int i; - - printf("Rot %d, xFlip %d, yFlip %d:\n", rot, fx, fy); - - /* make transformation matrix 'm' */ - matrix23Rotate(&rotate, width, height, rot); - matrix23Flip(&flip, width, height, fx, fy); - matrix23Multiply(&m, &rotate, &flip); - - /* xform four coords */ - for (i = 0; i < 4; i++) { - int x = coords[i][0]; - int y = coords[i][1]; - matrix23TransformCoordi(&m, &x, &y); - printf(" %d, %d -> %d %d\n", coords[i][0], coords[i][1], x, y); - } - - /* xform width, height */ - { - int x = width; - int y = height; - matrix23TransformDistance(&m, &x, &y); - printf(" %d x %d -> %d x %d\n", width, height, x, y); - } - - /* xform rect */ - { - int x = 50, y = 10, w = 200, h = 100; - matrix23TransformRect(&m, &x, &y, &w, &h); - printf(" %d,%d %d x %d -> %d, %d %d x %d\n", 50, 10, 200, 100, - x, y, w, h); - } - - } - - return 0; -} -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_rotate.h b/src/mesa/drivers/dri/i915pipe/intel_rotate.h deleted file mode 100644 index 9c8802ca47..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_rotate.h +++ /dev/null @@ -1,39 +0,0 @@ -#ifndef INTEL_ROTATE_H -#define INTEL_ROTATE_H 1 - -struct matrix23 -{ - int m00, m01, m02; - int m10, m11, m12; -}; - - - -extern void -matrix23Set(struct matrix23 *m, - int m00, int m01, int m02, int m10, int m11, int m12); - -extern void matrix23TransformCoordi(const struct matrix23 *m, int *x, int *y); - -extern void -matrix23TransformCoordf(const struct matrix23 *m, float *x, float *y); - -extern void -matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist); - -extern void -matrix23TransformRect(const struct matrix23 *m, - int *x, int *y, int *w, int *h); - -extern void -matrix23Rotate(struct matrix23 *m, int width, int height, int angle); - -extern void -matrix23Flip(struct matrix23 *m, int width, int height, int xflip, int yflip); - -extern void -matrix23Multiply(struct matrix23 *result, - const struct matrix23 *a, const struct matrix23 *b); - - -#endif /* INTEL_ROTATE_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.h b/src/mesa/drivers/dri/i915pipe/intel_screen.h index d5866ac0ec..de0bea6609 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.h +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.h @@ -29,7 +29,6 @@ #define _INTEL_SCREEN_H_ #include "dri_util.h" -#include "intel_rotate.h" #include "i830_common.h" #include "xmlconfig.h" #include "dri_bufpool.h" @@ -65,7 +64,6 @@ typedef struct int irq_active; int allow_batchbuffer; - struct matrix23 rotMatrix; /** @@ -112,8 +110,6 @@ extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, extern struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen); -extern void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); extern GLboolean intelCreatePools(intelScreenPrivate *intelScreen); diff --git a/src/mesa/drivers/dri/i915pipe/intel_state.c b/src/mesa/drivers/dri/i915pipe/intel_state.c deleted file mode 100644 index 2455f177da..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_state.c +++ /dev/null @@ -1,258 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "enums.h" -#include "colormac.h" -#include "dd.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_fbo.h" -#include "swrast/swrast.h" - -int -intel_translate_compare_func(GLenum func) -{ - switch (func) { - case GL_NEVER: - return COMPAREFUNC_NEVER; - case GL_LESS: - return COMPAREFUNC_LESS; - case GL_LEQUAL: - return COMPAREFUNC_LEQUAL; - case GL_GREATER: - return COMPAREFUNC_GREATER; - case GL_GEQUAL: - return COMPAREFUNC_GEQUAL; - case GL_NOTEQUAL: - return COMPAREFUNC_NOTEQUAL; - case GL_EQUAL: - return COMPAREFUNC_EQUAL; - case GL_ALWAYS: - return COMPAREFUNC_ALWAYS; - } - - fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); - return COMPAREFUNC_ALWAYS; -} - -int -intel_translate_stencil_op(GLenum op) -{ - switch (op) { - case GL_KEEP: - return STENCILOP_KEEP; - case GL_ZERO: - return STENCILOP_ZERO; - case GL_REPLACE: - return STENCILOP_REPLACE; - case GL_INCR: - return STENCILOP_INCRSAT; - case GL_DECR: - return STENCILOP_DECRSAT; - case GL_INCR_WRAP: - return STENCILOP_INCR; - case GL_DECR_WRAP: - return STENCILOP_DECR; - case GL_INVERT: - return STENCILOP_INVERT; - default: - return STENCILOP_ZERO; - } -} - -int -intel_translate_blend_factor(GLenum factor) -{ - switch (factor) { - case GL_ZERO: - return BLENDFACT_ZERO; - case GL_SRC_ALPHA: - return BLENDFACT_SRC_ALPHA; - case GL_ONE: - return BLENDFACT_ONE; - case GL_SRC_COLOR: - return BLENDFACT_SRC_COLR; - case GL_ONE_MINUS_SRC_COLOR: - return BLENDFACT_INV_SRC_COLR; - case GL_DST_COLOR: - return BLENDFACT_DST_COLR; - case GL_ONE_MINUS_DST_COLOR: - return BLENDFACT_INV_DST_COLR; - case GL_ONE_MINUS_SRC_ALPHA: - return BLENDFACT_INV_SRC_ALPHA; - case GL_DST_ALPHA: - return BLENDFACT_DST_ALPHA; - case GL_ONE_MINUS_DST_ALPHA: - return BLENDFACT_INV_DST_ALPHA; - case GL_SRC_ALPHA_SATURATE: - return BLENDFACT_SRC_ALPHA_SATURATE; - case GL_CONSTANT_COLOR: - return BLENDFACT_CONST_COLOR; - case GL_ONE_MINUS_CONSTANT_COLOR: - return BLENDFACT_INV_CONST_COLOR; - case GL_CONSTANT_ALPHA: - return BLENDFACT_CONST_ALPHA; - case GL_ONE_MINUS_CONSTANT_ALPHA: - return BLENDFACT_INV_CONST_ALPHA; - } - - fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, factor); - return BLENDFACT_ZERO; -} - -int -intel_translate_logic_op(GLenum opcode) -{ - switch (opcode) { - case GL_CLEAR: - return LOGICOP_CLEAR; - case GL_AND: - return LOGICOP_AND; - case GL_AND_REVERSE: - return LOGICOP_AND_RVRSE; - case GL_COPY: - return LOGICOP_COPY; - case GL_COPY_INVERTED: - return LOGICOP_COPY_INV; - case GL_AND_INVERTED: - return LOGICOP_AND_INV; - case GL_NOOP: - return LOGICOP_NOOP; - case GL_XOR: - return LOGICOP_XOR; - case GL_OR: - return LOGICOP_OR; - case GL_OR_INVERTED: - return LOGICOP_OR_INV; - case GL_NOR: - return LOGICOP_NOR; - case GL_EQUIV: - return LOGICOP_EQUIV; - case GL_INVERT: - return LOGICOP_INV; - case GL_OR_REVERSE: - return LOGICOP_OR_RVRSE; - case GL_NAND: - return LOGICOP_NAND; - case GL_SET: - return LOGICOP_SET; - default: - return LOGICOP_SET; - } -} - - -static void -intelClearColor(GLcontext * ctx, const GLfloat color[4]) -{ - struct intel_context *intel = intel_context(ctx); - GLubyte clear[4]; - - CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - - /* compute both 32 and 16-bit clear values */ - intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1], - clear[2], clear[3]); - intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]); -} - - -/** - * Update the viewport transformation matrix. Depends on: - * - viewport pos/size - * - depthrange - * - window size or FBO size - */ -static void -intelCalcViewport(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - const GLfloat *v = ctx->Viewport._WindowMap.m; - const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF; - GLfloat *m = intel->ViewportMatrix.m; - GLfloat yScale, yBias; - - struct intel_renderbuffer *irb - = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); - if (irb && !irb->RenderToTexture) { - /* y=0=top */ - yScale = -1.0; - yBias = irb->Base.Height; - } - else { - /* y=0=bottom */ - yScale = 1.0; - yBias = 0.0; - } - - m[MAT_SX] = v[MAT_SX]; - m[MAT_TX] = v[MAT_TX] + SUBPIXEL_X; - - m[MAT_SY] = v[MAT_SY] * yScale; - m[MAT_TY] = v[MAT_TY] * yScale + yBias + SUBPIXEL_Y; - - m[MAT_SZ] = v[MAT_SZ] * depthScale; - m[MAT_TZ] = v[MAT_TZ] * depthScale; -} - -static void -intelViewport(GLcontext * ctx, - GLint x, GLint y, GLsizei width, GLsizei height) -{ - intelCalcViewport(ctx); -} - -static void -intelDepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) -{ - intelCalcViewport(ctx); -} - -/* Fallback to swrast for select and feedback. - */ -static void -intelRenderMode(GLcontext * ctx, GLenum mode) -{ -} - - -void -intelInitStateFuncs(struct dd_function_table *functions) -{ - functions->RenderMode = intelRenderMode; - functions->Viewport = intelViewport; - functions->DepthRange = intelDepthRange; - functions->ClearColor = intelClearColor; -} -- cgit v1.2.3 From 0e067f1fb20094417e84e1b18f2302251cece2ca Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 2 Aug 2007 18:25:10 +0100 Subject: Remove references to accum buffers in softpipe. Also some minor clear fixes. --- src/mesa/pipe/p_context.h | 2 +- src/mesa/pipe/p_state.h | 1 - src/mesa/pipe/softpipe/sp_clear.c | 13 +------------ src/mesa/pipe/softpipe/sp_clear.h | 2 +- src/mesa/state_tracker/st_cb_clear.c | 20 +++++++++++++------- src/mesa/state_tracker/st_draw.h | 2 +- 6 files changed, 17 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 4f5937b8f3..2ce2781771 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -57,7 +57,7 @@ struct pipe_context { /** Clear framebuffer */ void (*clear)(struct pipe_context *pipe, GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum); + GLboolean stencil); /** occlusion counting (XXX this may be temporary - we should probably * have generic query objects with begin/end methods) diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index d6ed514fb1..699efd5877 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -202,7 +202,6 @@ struct pipe_framebuffer_state struct pipe_surface *zbuf; /**< Z buffer */ struct pipe_surface *sbuf; /**< Stencil buffer */ - struct pipe_surface *abuf; /**< Accum buffer */ }; diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 09cc643003..e9b142e780 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -63,7 +63,7 @@ color_value(GLuint format, const GLfloat color[4]) void softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum) + GLboolean stencil) { struct softpipe_context *softpipe = softpipe_context(pipe); GLint x, y, w, h; @@ -152,15 +152,4 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); } } - - if (accum) { - /* XXX there might be no notion of accum buffers in 'pipe'. - * Just implement them with a deep RGBA surface format... - */ - struct pipe_surface *ps = softpipe->framebuffer.abuf; - GLuint clearVal = 0x0; /* XXX FIX */ - GLuint mask = ~0; - assert(ps); - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); - } } diff --git a/src/mesa/pipe/softpipe/sp_clear.h b/src/mesa/pipe/softpipe/sp_clear.h index f9db99dd32..d41cc1d070 100644 --- a/src/mesa/pipe/softpipe/sp_clear.h +++ b/src/mesa/pipe/softpipe/sp_clear.h @@ -37,7 +37,7 @@ struct pipe_context; extern void softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum); + GLboolean stencil); #endif /* SP_CLEAR_H */ diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 523a9b682c..a7a3b5eba1 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -49,7 +49,7 @@ static void clear_with_quad(GLcontext *ctx, GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum) + GLboolean stencil) { struct st_context *st = ctx->st; struct pipe_blend_state blend; @@ -119,17 +119,18 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) GLboolean depth = (mask & BUFFER_BIT_DEPTH) ? GL_TRUE : GL_FALSE; GLboolean stencil = (mask & BUFFER_BIT_STENCIL) ? GL_TRUE : GL_FALSE; GLboolean accum = (mask & BUFFER_BIT_ACCUM) ? GL_TRUE : GL_FALSE; + GLboolean maskColor, maskStencil; GLboolean fullscreen = 1; /* :-) */ - GLuint stencilMax = 1 << ctx->DrawBuffer->_StencilBuffer->StencilBits; + GLuint stencilMax = stencil ? (1 << ctx->DrawBuffer->_StencilBuffer->StencilBits) : 0; /* This makes sure the softpipe has the latest scissor, etc values */ st_validate_state( st ); - maskColor = st->state.blend.colormask != PIPE_MASK_RGBA; - maskStencil = ctx->Stencil.WriteMask[0] != stencilMax; + maskColor = color && st->state.blend.colormask != PIPE_MASK_RGBA; + maskStencil = stencil && ctx->Stencil.WriteMask[0] != stencilMax; - if (fullscreen && !maskColor) { + if (fullscreen && !maskColor && !maskStencil) { /* pipe->clear() should clear a particular surface, so that we * can iterate over render buffers at this level and clear the * ones GL is asking for. @@ -139,12 +140,17 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) * buffers, though could perhaps deal with them explicitly at * this level. */ - st->pipe->clear(st->pipe, color, depth, stencil, accum); + st->pipe->clear(st->pipe, color, depth, stencil); + + /* And here we would do a clear on whatever surface we are using + * to implement accum buffers: + */ + assert(!accum); } else { /* Convert to geometry, etc: */ - clear_with_quad(ctx, color, depth, stencil, accum); + clear_with_quad(ctx, color, depth, stencil); } } diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h index 7a3ba52130..0afadab577 100644 --- a/src/mesa/state_tracker/st_draw.h +++ b/src/mesa/state_tracker/st_draw.h @@ -39,6 +39,6 @@ void st_destroy_draw( struct st_context *st ); /** XXX temporary here */ void st_clear(struct st_context *st, GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum); + GLboolean stencil); #endif -- cgit v1.2.3 From 55314f8f311bff065f609ff17c8421a8d5216b84 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 12:12:48 -0600 Subject: Implement new draw_vertices() path for simple vertex array drawing, use it for glClear. --- src/mesa/pipe/draw/draw_context.h | 5 +++ src/mesa/pipe/draw/draw_vb.c | 61 ++++++++++++++++++++++++++++++-- src/mesa/pipe/p_context.h | 5 +++ src/mesa/pipe/softpipe/sp_context.c | 20 +++++++++++ src/mesa/state_tracker/st_cb_clear.c | 67 ++++++++++++++++++++++++++++++------ 5 files changed, 146 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index c298d4f46d..be0a18d6d7 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -70,5 +70,10 @@ void draw_set_vertex_attributes( struct draw_context *draw, void draw_vb(struct draw_context *draw, struct vertex_buffer *VB ); +void draw_vertices(struct draw_context *draw, + GLuint mode, + GLuint numVertex, const GLfloat *verts, + GLuint numAttribs, const GLuint attribs[]); + #endif /* DRAW_CONTEXT_H */ diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index ac126c5baa..f9c10e5f97 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -57,7 +57,7 @@ static void draw_allocate_vertices( struct draw_context *draw, GLuint nr_vertices ) { draw->nr_vertices = nr_vertices; - draw->verts = MALLOC( nr_vertices * draw->vertex_size ); + draw->verts = (GLubyte *) malloc( nr_vertices * draw->vertex_size ); draw->pipeline.first->begin( draw->pipeline.first ); } @@ -453,7 +453,7 @@ static void draw_release_vertices( struct draw_context *draw ) { draw->pipeline.first->end( draw->pipeline.first ); - FREE(draw->verts); + free(draw->verts); draw->verts = NULL; } @@ -646,6 +646,63 @@ void draw_vb(struct draw_context *draw, } +/** + * XXX Temporary mechanism to draw simple vertex arrays. + * All attribs are GLfloat[4]. Arrays are interleaved, in GL-speak. + */ +void +draw_vertices(struct draw_context *draw, + GLuint mode, + GLuint numVerts, const GLfloat *vertices, + GLuint numAttrs, const GLuint attribs[]) +{ + /*GLuint first, incr;*/ + GLuint i, j; + + assert(mode <= GL_POLYGON); + + draw->vertex_size + = sizeof(struct vertex_header) + numAttrs * 4 * sizeof(GLfloat); + + /*draw_prim_info(mode, &first, &incr);*/ + draw_allocate_vertices( draw, numVerts ); + if (draw->prim != mode) + draw_set_prim( draw, mode ); + + /* setup attr info */ + draw->nr_attrs = numAttrs + 2; + draw->attrs[0].attrib = VF_ATTRIB_VERTEX_HEADER; + draw->attrs[0].format = EMIT_1F; + draw->attrs[1].attrib = VF_ATTRIB_CLIP_POS; + draw->attrs[1].format = EMIT_4F; + for (j = 0; j < numAttrs; j++) { + draw->vf_attr_to_slot[attribs[j]] = 2+j; + draw->attrs[2+j].attrib = attribs[j]; + draw->attrs[2+j].format = EMIT_4F; + } + + /* build vertices */ + for (i = 0; i < numVerts; i++) { + struct vertex_header *v + = (struct vertex_header *) (draw->verts + i * draw->vertex_size); + v->clipmask = 0x0; + v->edgeflag = 0; + for (j = 0; j < numAttrs; j++) { + COPY_4FV(v->data[j], vertices + (i * numAttrs + j) * 4); + } + } + + /* draw */ + draw_prim(draw, 0, numVerts); + + /* clean up */ + draw_release_vertices( draw ); + draw->verts = NULL; + draw->in_vb = 0; +} + + + /** * Accumulate another attribute's info. * Note the "- 2" factor here. We need this because the vertex->data[] diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 2ce2781771..0972fd58b5 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -55,6 +55,11 @@ struct pipe_context { void (*draw_vb)( struct pipe_context *pipe, struct vertex_buffer *VB ); + void (*draw_vertices)( struct pipe_context *pipe, + GLuint mode, + GLuint numVertex, const GLfloat *verts, + GLuint numAttribs, const GLuint attribs[]); + /** Clear framebuffer */ void (*clear)(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 434e18308a..9fba9605e8 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -102,6 +102,25 @@ static void softpipe_draw_vb( struct pipe_context *pipe, } +static void +softpipe_draw_vertices(struct pipe_context *pipe, + GLuint mode, + GLuint numVertex, const GLfloat *verts, + GLuint numAttribs, const GLuint attribs[]) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + + if (softpipe->dirty) + softpipe_update_derived( softpipe ); + + /* XXX move mapping/unmapping to higher/coarser level? */ + map_surfaces(softpipe); + draw_vertices(softpipe->draw, mode, numVertex, verts, numAttribs, attribs); + unmap_surfaces(softpipe); +} + + + static void softpipe_reset_occlusion_counter(struct pipe_context *pipe) { struct softpipe_context *softpipe = softpipe_context( pipe ); @@ -137,6 +156,7 @@ struct pipe_context *softpipe_create( void ) softpipe->pipe.set_texture_state = softpipe_set_texture_state; softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; softpipe->pipe.draw_vb = softpipe_draw_vb; + softpipe->pipe.draw_vertices = softpipe_draw_vertices; softpipe->pipe.clear = softpipe_clear; softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index a7a3b5eba1..8e2e30253e 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -40,6 +40,51 @@ #include "st_public.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "vf/vf.h" + + +/** + * Draw a screen-aligned quadrilateral. + * Coords are window coords. + */ +static void +draw_quad(GLcontext *ctx, + float x0, float y0, float x1, float y1, GLfloat z, + const GLfloat color[4]) +{ + static const GLuint attribs[2] = { + VF_ATTRIB_POS, + VF_ATTRIB_COLOR0 + }; + GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ + GLuint i; + + /* positions */ + verts[0][0][0] = x0; + verts[0][0][1] = y0; + + verts[1][0][0] = x1; + verts[1][0][1] = y0; + + verts[2][0][0] = x1; + verts[2][0][1] = y1; + + verts[3][0][0] = x0; + verts[3][0][1] = y1; + + /* same for all verts: */ + for (i = 0; i < 4; i++) { + verts[i][0][2] = z; + verts[i][0][3] = 1.0; + verts[i][1][0] = color[0]; + verts[i][1][1] = color[1]; + verts[i][1][2] = color[2]; + verts[i][1][3] = color[3]; + } + + ctx->st->pipe->draw_vertices(ctx->st->pipe, GL_QUADS, + 4, (GLfloat *) verts, 2, attribs); +} @@ -48,14 +93,12 @@ */ static void clear_with_quad(GLcontext *ctx, - GLboolean color, GLboolean depth, - GLboolean stencil) + GLboolean color, GLboolean depth, GLboolean stencil) { struct st_context *st = ctx->st; struct pipe_blend_state blend; struct pipe_depth_state depth_test; struct pipe_stencil_state stencil_test; - GLfloat z = ctx->Depth.Clear; /* depth state: always pass */ memset(&depth_test, 0, sizeof(depth)); @@ -96,13 +139,19 @@ clear_with_quad(GLcontext *ctx, } st->pipe->set_blend_state(st->pipe, &blend); - - /* - * XXX Render quad here - */ + draw_quad(ctx, + ctx->Scissor.X, ctx->Scissor.Y, + ctx->Scissor.X + ctx->Scissor.Width, + ctx->Scissor.Y + ctx->Scissor.Height, + ctx->Depth.Clear, ctx->Color.ClearColor); /* Restore GL state */ + st->pipe->set_blend_state(st->pipe, &st->state.blend); + st->pipe->set_depth_state(st->pipe, &st->state.depth); + st->pipe->set_stencil_state(st->pipe, &st->state.stencil); + /* OR: st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); + */ } @@ -121,7 +170,7 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) GLboolean accum = (mask & BUFFER_BIT_ACCUM) ? GL_TRUE : GL_FALSE; GLboolean maskColor, maskStencil; - GLboolean fullscreen = 1; /* :-) */ + GLboolean fullscreen = !ctx->Scissor.Enabled; GLuint stencilMax = stencil ? (1 << ctx->DrawBuffer->_StencilBuffer->StencilBits) : 0; /* This makes sure the softpipe has the latest scissor, etc values */ @@ -148,8 +197,6 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) assert(!accum); } else { - /* Convert to geometry, etc: - */ clear_with_quad(ctx, color, depth, stencil); } } -- cgit v1.2.3 From d878bdc7a107991075080f9d6d2f1489b966981f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 2 Aug 2007 20:33:58 +0100 Subject: Fix make recursion. Unfortunately means you can't just type make in softpipe any more. --- src/mesa/Makefile | 3 ++- src/mesa/pipe/softpipe/Makefile | 2 -- 2 files changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 3055564341..ee4cbd0d38 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -149,7 +149,7 @@ depend: $(ALL_SOURCES) subdirs: @ (cd x86 ; $(MAKE)) @ (cd x86-64 ; $(MAKE)) - #(cd pipe/softpipe ; $(MAKE)) + (cd pipe/softpipe ; $(MAKE)) install: default $(INSTALL) -d $(INSTALL_DIR)/include/GL @@ -182,6 +182,7 @@ clean: (cd drivers/dri && $(MAKE) clean) (cd x86 && $(MAKE) clean) (cd x86-64 && $(MAKE) clean) + (cd pipe/softpipe ; $(MAKE) clean ) include depend diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 947983baf0..04464154e7 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -40,5 +40,3 @@ include ../Makefile.template symlinks: -default:: - cd ../.. ; make -- cgit v1.2.3 From 877478e6fc659e2e40b8ba6bb7d78ee512c7c0d5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 2 Aug 2007 20:35:50 +0100 Subject: Remove all references to swrast. The one place the functionality continues to be needed is as last-ditch implementations of TexCopyImage, etc. TBD what to do about that, but that will be an issue for state_tracker, not for here. --- src/mesa/drivers/dri/i915pipe/intel_buffers.c | 1 - src/mesa/drivers/dri/i915pipe/intel_context.c | 22 ---------------------- src/mesa/drivers/dri/i915pipe/intel_ioctl.c | 1 - src/mesa/drivers/dri/i915pipe/intel_pixel.c | 6 ------ src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 12 +++++++++--- 5 files changed, 9 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.c b/src/mesa/drivers/dri/i915pipe/intel_buffers.c index 8fd785d20d..fc9d60c88e 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.c +++ b/src/mesa/drivers/dri/i915pipe/intel_buffers.c @@ -37,7 +37,6 @@ #include "utils.h" #include "drirenderbuffer.h" #include "framebuffer.h" -#include "swrast/swrast.h" #include "vblank.h" #include "pipe/p_context.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index b66dbe71e5..bab94e2697 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -35,8 +35,6 @@ #include "imports.h" #include "points.h" -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" #include "tnl/tnl.h" #include "tnl/t_pipeline.h" @@ -237,8 +235,6 @@ static const struct dri_debug_control debug_control[] = { static void intelInvalidateState(GLcontext * ctx, GLuint new_state) { - _swrast_InvalidateState(ctx, new_state); - _swsetup_InvalidateState(ctx, new_state); _vbo_InvalidateState(ctx, new_state); _tnl_InvalidateState(ctx, new_state); _tnl_invalidate_vertex_state(ctx, new_state); @@ -254,9 +250,6 @@ intelFlush(GLcontext * ctx) { struct intel_context *intel = intel_context(ctx); - if (intel->Fallback) - _swrast_flush(ctx); - INTEL_FIREVERTICES(intel); if (intel->batch->map != intel->batch->ptr) @@ -323,10 +316,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) functions->Finish = intelFinish; functions->GetString = intelGetString; functions->UpdateState = intelInvalidateState; - functions->CopyColorTable = _swrast_CopyColorTable; - functions->CopyColorSubTable = _swrast_CopyColorSubTable; - functions->CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D; - functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D; intelInitTextureFuncs(functions); intelInitPixelFuncs(functions); @@ -402,14 +391,8 @@ intelCreateContext(const __GLcontextModes * mesaVis, ctx->Const.MaxColorAttachments = 4; /* XXX FBO: review this */ /* Initialize the software rasterizer and helper modules. */ - _swrast_CreateContext(ctx); _vbo_CreateContext(ctx); _tnl_CreateContext(ctx); - _swsetup_CreateContext(ctx); - - /* Configure swrast to match hardware characteristics: */ - _swrast_allow_pixel_fog(ctx, GL_FALSE); - _swrast_allow_vertex_fog(ctx, GL_TRUE); /* * Pipe-related setup @@ -531,11 +514,9 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) //intel->vtbl.destroy(intel); release_texture_heaps = (intel->ctx.Shared->RefCount == 1); - _swsetup_DestroyContext(&intel->ctx); _tnl_DestroyContext(&intel->ctx); _vbo_DestroyContext(&intel->ctx); - _swrast_DestroyContext(&intel->ctx); intel->Fallback = 0; /* don't call _swrast_Flush later */ intel_batchbuffer_free(intel->batch); @@ -692,9 +673,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags) */ intel->numClipRects = 0; - if (intel->Fallback) - _swrast_flush(&intel->ctx); - INTEL_FIREVERTICES(intel); if (intel->batch->map != intel->batch->ptr) diff --git a/src/mesa/drivers/dri/i915pipe/intel_ioctl.c b/src/mesa/drivers/dri/i915pipe/intel_ioctl.c index fb5b518903..154a3e289f 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_ioctl.c +++ b/src/mesa/drivers/dri/i915pipe/intel_ioctl.c @@ -33,7 +33,6 @@ #include "mtypes.h" #include "context.h" -#include "swrast/swrast.h" #include "intel_context.h" #include "intel_ioctl.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_pixel.c b/src/mesa/drivers/dri/i915pipe/intel_pixel.c index 74d4ce7f2a..27de2a6636 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_pixel.c +++ b/src/mesa/drivers/dri/i915pipe/intel_pixel.c @@ -26,7 +26,6 @@ **************************************************************************/ #include "state.h" -#include "swrast/swrast.h" #include "intel_context.h" #include "intel_pixel.h" @@ -35,9 +34,4 @@ void intelInitPixelFuncs(struct dd_function_table *functions) { - functions->Accum = _swrast_Accum; - functions->Bitmap = _swrast_Bitmap; - functions->CopyPixels = _swrast_CopyPixels; - functions->ReadPixels = _swrast_ReadPixels; - functions->DrawPixels = _swrast_DrawPixels; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c index 6deb6abd4d..6f4f75079e 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c @@ -231,8 +231,11 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, return; fail: +#if 0 _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, width, height, border); +#endif + assert(0); } @@ -256,7 +259,10 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, if (!do_copy_texsubimage(intel_context(ctx), intel_texture_image(texImage), internalFormat, xoffset, 0, x, y, width, 1)) { +#if 0 _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); +#endif + assert(0); } } @@ -283,10 +289,10 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, intel_texture_image(texImage), internalFormat, xoffset, yoffset, x, y, width, height)) { - - DBG("%s - fallback to swrast\n", __FUNCTION__); - +#if 0 _swrast_copy_texsubimage2d(ctx, target, level, xoffset, yoffset, x, y, width, height); +#endif + assert(0); } } -- cgit v1.2.3 From 3ab0210c3152e101ebb5af586c837978919f49d0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 13:38:38 -0600 Subject: get rid of accum param --- src/mesa/drivers/x11/xm_dd.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index fec4f5d038..87f8ede50b 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -425,8 +425,7 @@ clear_buffers(GLcontext *ctx, GLbitfield buffers) GLboolean color = (buffers & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)) ? 1: 0; GLboolean depth = (buffers & BUFFER_BIT_DEPTH) ? 1 : 0; GLboolean stencil = (buffers & BUFFER_BIT_STENCIL) ? 1 : 0; - GLboolean accum = (buffers & BUFFER_BIT_ACCUM) ? 1 : 0; - st_clear(st, color, depth, stencil, accum); + st_clear(st, color, depth, stencil); #endif } } -- cgit v1.2.3 From 406da44da31f71afd98fc45e2bbfbf41deff1d12 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 13:48:02 -0600 Subject: setup more state for clear_with_quad() --- src/mesa/state_tracker/st_cb_clear.c | 49 +++++++++++++++++++++++------------- 1 file changed, 31 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 8e2e30253e..85df549404 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -96,12 +96,34 @@ clear_with_quad(GLcontext *ctx, GLboolean color, GLboolean depth, GLboolean stencil) { struct st_context *st = ctx->st; + struct pipe_alpha_test_state alpha_test; struct pipe_blend_state blend; struct pipe_depth_state depth_test; struct pipe_stencil_state stencil_test; + struct pipe_setup_state setup; + + /* alpha state: disabled */ + memset(&alpha_test, 0, sizeof(alpha_test)); + st->pipe->set_alpha_test_state(st->pipe, &alpha_test); + + /* blend state: RGBA masking */ + memset(&blend, 0, sizeof(blend)); + if (color) { + if (ctx->Color.ColorMask[0]) + blend.colormask |= PIPE_MASK_R; + if (ctx->Color.ColorMask[1]) + blend.colormask |= PIPE_MASK_G; + if (ctx->Color.ColorMask[2]) + blend.colormask |= PIPE_MASK_B; + if (ctx->Color.ColorMask[3]) + blend.colormask |= PIPE_MASK_A; + if (st->ctx->Color.DitherFlag) + blend.dither = 1; + } + st->pipe->set_blend_state(st->pipe, &blend); /* depth state: always pass */ - memset(&depth_test, 0, sizeof(depth)); + memset(&depth_test, 0, sizeof(depth_test)); if (depth) { depth_test.enabled = 1; depth_test.writemask = 1; @@ -109,8 +131,12 @@ clear_with_quad(GLcontext *ctx, } st->pipe->set_depth_state(st->pipe, &depth_test); + /* setup state: nothing */ + memset(&setup, 0, sizeof(setup)); + st->pipe->set_setup_state(st->pipe, &setup); + /* stencil state: always set to ref value */ - memset(&stencil_test, 0, sizeof(stencil)); + memset(&stencil_test, 0, sizeof(stencil_test)); if (stencil) { stencil_test.front_enabled = 1; stencil_test.front_func = PIPE_FUNC_ALWAYS; @@ -123,22 +149,7 @@ clear_with_quad(GLcontext *ctx, } st->pipe->set_stencil_state(st->pipe, &stencil_test); - /* blend state: RGBA masking */ - memset(&blend, 0, sizeof(blend)); - if (color) { - if (ctx->Color.ColorMask[0]) - blend.colormask |= PIPE_MASK_R; - if (ctx->Color.ColorMask[1]) - blend.colormask |= PIPE_MASK_G; - if (ctx->Color.ColorMask[2]) - blend.colormask |= PIPE_MASK_B; - if (ctx->Color.ColorMask[3]) - blend.colormask |= PIPE_MASK_A; - if (st->ctx->Color.DitherFlag) - blend.dither = 1; - } - st->pipe->set_blend_state(st->pipe, &blend); - + /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, ctx->Scissor.X, ctx->Scissor.Y, ctx->Scissor.X + ctx->Scissor.Width, @@ -146,8 +157,10 @@ clear_with_quad(GLcontext *ctx, ctx->Depth.Clear, ctx->Color.ClearColor); /* Restore GL state */ + st->pipe->set_alpha_test_state(st->pipe, &st->state.alpha_test); st->pipe->set_blend_state(st->pipe, &st->state.blend); st->pipe->set_depth_state(st->pipe, &st->state.depth); + st->pipe->set_setup_state(st->pipe, &st->state.setup); st->pipe->set_stencil_state(st->pipe, &st->state.stencil); /* OR: st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); -- cgit v1.2.3 From 566ae9196b267492d9fcfb2be9065fb8017702f9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 14:20:13 -0600 Subject: remove st_draw.h include --- src/mesa/main/buffers.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index eea443c03c..0e6ca8ea1c 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -38,8 +38,6 @@ #include "fbobject.h" #include "state.h" -#include "state_tracker/st_draw.h" - #define BAD_MASK ~0u -- cgit v1.2.3 From d09f4e2007b3043ceb9dd3cb9cfab0dcd0226ba6 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 14:20:40 -0600 Subject: initial work for textured-quad glDrawPixels --- src/mesa/state_tracker/st_cb_drawpixels.c | 278 ++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_drawpixels.h | 38 ++++ 2 files changed, 316 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_drawpixels.c create mode 100644 src/mesa/state_tracker/st_cb_drawpixels.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c new file mode 100644 index 0000000000..7209db00a3 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -0,0 +1,278 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Brian Paul + */ + +#include "imports.h" + +#include "tnl/t_context.h" +#include "tnl/t_pipeline.h" + +#include "st_context.h" +#include "st_atom.h" +#include "st_draw.h" +#include "st_cb_drawpixels.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "vf/vf.h" + + +static void +draw_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, + GLsizei width, GLsizei height) +{ + static const GLuint attribs[2] = { + VF_ATTRIB_POS, + VF_ATTRIB_TEX0 + }; + GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ + GLuint i; + + /* lower-left */ + verts[0][0][0] = x; + verts[0][0][1] = y; + verts[0][1][0] = 0.0; + verts[0][1][1] = 0.0; + + /* lower-right */ + verts[1][0][0] = x + width; + verts[1][0][1] = y; + verts[1][1][0] = 1.0; + verts[1][1][1] = 0.0; + + /* upper-right */ + verts[2][0][0] = x + width; + verts[2][0][1] = y + height; + verts[2][1][0] = 1.0; + verts[2][1][1] = 1.0; + + /* upper-left */ + verts[3][0][0] = x; + verts[3][0][1] = y + height; + verts[3][1][0] = 0.0; + verts[3][1][1] = 11.0; + + /* same for all verts: */ + for (i = 0; i < 4; i++) { + verts[i][0][2] = z; /*Z*/ + verts[i][0][3] = 1.0; /*W*/ + verts[i][1][2] = 0.0; /*R*/ + verts[i][1][3] = 1.0; /*Q*/ + } + + st->pipe->draw_vertices(st->pipe, GL_QUADS, + 4, (GLfloat *) verts, 2, attribs); +} + + +static void +draw_textured_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, + GLsizei width, GLsizei height, GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels) +{ + struct pipe_setup_state setup; + + /* setup state: just scissor */ + memset(&setup, 0, sizeof(setup)); + if (st->ctx->Scissor.Enabled) + setup.scissor = 1; + st->pipe->set_setup_state(st->pipe, &setup); + + + /* XXX upload the texture image */ + + draw_quad(st, x, y, z, width, height); + + /* restore GL state */ + st->pipe->set_setup_state(st->pipe, &st->state.setup); +} + + +/** + * Check if a GL format/type combination is a match to the given pipe format. + * XXX probably move this to a re-usable place. + */ +static GLboolean +compatible_formats(GLenum format, GLenum type, GLuint pipeFormat) +{ + static const GLuint one = 1; + GLubyte littleEndian = *((GLubyte *) one); + + if (pipeFormat == PIPE_FORMAT_U_R8_G8_B8_A8 && + format == GL_RGBA && + type == GL_UNSIGNED_BYTE && + !littleEndian) { + return GL_TRUE; + } + else if (pipeFormat == PIPE_FORMAT_U_R8_G8_B8_A8 && + format == GL_ABGR_EXT && + type == GL_UNSIGNED_BYTE && + littleEndian) { + return GL_TRUE; + } + else if (pipeFormat == PIPE_FORMAT_U_A8_R8_G8_B8 && + format == GL_BGRA && + type == GL_UNSIGNED_BYTE && + littleEndian) { + return GL_TRUE; + } + else if (pipeFormat == PIPE_FORMAT_U_R5_G6_B5 && + format == GL_RGB && + type == GL_UNSIGNED_SHORT_5_6_5) { + /* endian don't care */ + return GL_TRUE; + } + else if (pipeFormat == PIPE_FORMAT_U_R5_G6_B5 && + format == GL_BGR && + type == GL_UNSIGNED_SHORT_5_6_5_REV) { + /* endian don't care */ + return GL_TRUE; + } + else if (pipeFormat == PIPE_FORMAT_U_S8 && + format == GL_STENCIL_INDEX && + type == GL_UNSIGNED_BYTE) { + return GL_TRUE; + } + else if (pipeFormat == PIPE_FORMAT_U_Z32 && + format == GL_DEPTH_COMPONENT && + type == GL_UNSIGNED_INT) { + return GL_TRUE; + } + /* XXX add more cases */ + else { + return GL_FALSE; + } +} + + +/** + * Check if any per-fragment ops are enabled. + * XXX probably move this to a re-usable place. + */ +static GLboolean +any_fragment_ops(const struct st_context *st) +{ + if (st->state.alpha_test.enabled || + st->state.blend.blend_enable || + st->state.blend.logicop_enable || + st->state.depth.enabled) + /* XXX more checks */ + return GL_TRUE; + else + return GL_FALSE; +} + + +/** + * Check if any pixel transfer ops are enabled. + * XXX probably move this to a re-usable place. + */ +static GLboolean +any_pixel_transfer_ops(const struct st_context *st) +{ + if (st->ctx->Pixel.RedScale != 1.0 || + st->ctx->Pixel.RedBias != 0.0 || + st->ctx->Pixel.GreenScale != 1.0 || + st->ctx->Pixel.GreenBias != 0.0 || + st->ctx->Pixel.BlueScale != 1.0 || + st->ctx->Pixel.BlueBias != 0.0 || + st->ctx->Pixel.AlphaScale != 1.0 || + st->ctx->Pixel.AlphaBias != 0.0 || + st->ctx->Pixel.MapColorFlag) + /* XXX more checks */ + return GL_TRUE; + else + return GL_FALSE; +} + + +/** + * Draw image with a blit, or other non-textured quad method. + */ +static void +draw_blit(struct st_context *st, + GLsizei width, GLsizei height, + GLenum format, GLenum type, const GLvoid *pixels) +{ + + +} + + +/** + * Called via ctx->Driver.DrawPixels() + */ +static void +st_drawpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) +{ + struct st_context *st = ctx->st; + struct pipe_surface *ps; + GLuint bufferFormat; + + if (format == GL_DEPTH_COMPONENT) { + ps = st->state.framebuffer.zbuf; + } + else if (format == GL_STENCIL_INDEX) { + ps = st->state.framebuffer.sbuf; + } + else { + ps = st->state.framebuffer.cbufs[0]; + } + + bufferFormat = ps->format; + + if (any_fragment_ops(st) || + any_pixel_transfer_ops(st) || + !compatible_formats(format, type, ps->format)) { + /* textured quad */ + draw_textured_quad(st, x, y, ctx->Current.RasterPos[2], width, height, + format, type, unpack, pixels); + } + else { + /* blit */ + draw_blit(st, width, height, format, type, pixels); + } +} + + +void st_init_cb_drawpixels( struct st_context *st ) +{ + struct dd_function_table *functions = &st->ctx->Driver; + + functions->DrawPixels = st_drawpixels; +} + + +void st_destroy_cb_drawpixels( struct st_context *st ) +{ +} diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h new file mode 100644 index 0000000000..8c36aaa931 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_drawpixels.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_DRAWPIXELS_H +#define ST_CB_DRAWPIXELS_H + + +void st_init_cb_drawpixels( struct st_context *st ); + +void st_destroy_cb_drawpixels( struct st_context *st ); + + +#endif /* ST_CB_DRAWPIXELS_H */ -- cgit v1.2.3 From 4df8e90a01b395801c47ea897b80637a749c246b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 14:21:02 -0600 Subject: added st_cb_drawpixels.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index d0fe3a979f..a7c31001c2 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -192,6 +192,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_stipple.c \ state_tracker/st_atom_viewport.c \ state_tracker/st_cb_clear.c \ + state_tracker/st_cb_drawpixels.c \ state_tracker/st_cb_program.c \ state_tracker/st_draw.c \ state_tracker/st_context.c \ -- cgit v1.2.3 From 184b6a1f04e060e73858438561a0261d32eeb506 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 14:21:16 -0600 Subject: call st_init_cb_drawpixels --- src/mesa/state_tracker/st_context.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index cde1dd11ca..168ad33d6e 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -29,6 +29,7 @@ #include "st_public.h" #include "st_context.h" #include "st_cb_clear.h" +#include "st_cb_drawpixels.h" #include "st_atom.h" #include "st_draw.h" #include "st_program.h" @@ -60,6 +61,7 @@ struct st_context *st_create_context( GLcontext *ctx, st_init_draw( st ); st_init_cb_program( st ); st_init_cb_clear( st ); + st_init_cb_drawpixels( st ); return st; } -- cgit v1.2.3 From 47fdaf0ed9ef2f89cdaa97d0d48b1f1194d710c6 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 16:08:18 -0600 Subject: pipe->clear() now takes a surface, rather than color/depth/stencil flags. pipe->clear() only used to clear whole buffers (no scissor) w/out masking. Draw a colored quadrilateral in all other cases. --- src/mesa/drivers/x11/xm_api.c | 5 +- src/mesa/drivers/x11/xm_dd.c | 23 ++- src/mesa/drivers/x11/xmesaP.h | 7 +- src/mesa/pipe/p_context.h | 6 +- src/mesa/pipe/softpipe/sp_clear.c | 109 +------------- src/mesa/pipe/softpipe/sp_clear.h | 4 +- src/mesa/state_tracker/st_cb_clear.c | 274 ++++++++++++++++++++++++++++++----- 7 files changed, 270 insertions(+), 158 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 92d37085d1..7fb99df5f4 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -1573,8 +1573,11 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) st_create_context( mesaCtx, softpipe_create() ); + mesaCtx->Driver.Clear = xmesa_clear_buffers; + /* mesaCtx->st->pipe->clear = xmesa_clear; - + */ + return c; } diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index 87f8ede50b..f64f8780cf 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -381,8 +381,8 @@ clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, -static void -clear_buffers(GLcontext *ctx, GLbitfield buffers) +void +xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers) { if (ctx->DrawBuffer->Name == 0) { /* this is a window system framebuffer */ @@ -395,7 +395,6 @@ clear_buffers(GLcontext *ctx, GLbitfield buffers) /* we can't handle color or index masking */ if (*colorMask == 0xffffffff && ctx->Color.IndexMask == 0xffffffff) { -#if 0 if (buffers & BUFFER_BIT_FRONT_LEFT) { /* clear front color buffer */ struct gl_renderbuffer *frontRb @@ -419,14 +418,6 @@ clear_buffers(GLcontext *ctx, GLbitfield buffers) buffers &= ~BUFFER_BIT_BACK_LEFT; } } -#else - /* Clear with state-tracker/pipe interface */ - struct st_context *st = st_context(ctx); - GLboolean color = (buffers & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)) ? 1: 0; - GLboolean depth = (buffers & BUFFER_BIT_DEPTH) ? 1 : 0; - GLboolean stencil = (buffers & BUFFER_BIT_STENCIL) ? 1 : 0; - st_clear(st, color, depth, stencil); -#endif } } if (buffers) @@ -434,6 +425,7 @@ clear_buffers(GLcontext *ctx, GLbitfield buffers) } +#if 0 void xmesa_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, GLboolean stencil, GLboolean accum) @@ -458,9 +450,14 @@ xmesa_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, xrb->clearFunc(ctx, xrb, x, y, w, h); } } - } +#endif + +void +xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) +{ +} #ifndef XFree86Server @@ -1113,7 +1110,7 @@ xmesa_init_driver_functions( XMesaVisual xmvisual, driver->IndexMask = index_mask; driver->ColorMask = color_mask; driver->Enable = enable; - driver->Clear = clear_buffers; + driver->Clear = xmesa_clear_buffers; driver->Viewport = xmesa_viewport; #ifndef XFree86Server driver->CopyPixels = xmesa_CopyPixels; diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index fb1c1f8c3b..dd95aed4d0 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -593,7 +593,10 @@ extern struct pipe_surface * xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb); extern void -xmesa_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum); +xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value); + +extern void +xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers); + #endif diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 0972fd58b5..4f8bdae140 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -60,9 +60,9 @@ struct pipe_context { GLuint numVertex, const GLfloat *verts, GLuint numAttribs, const GLuint attribs[]); - /** Clear framebuffer */ - void (*clear)(struct pipe_context *pipe, GLboolean color, GLboolean depth, - GLboolean stencil); + /** Clear a surface to given value (no scissor; clear whole surface) */ + void (*clear)(struct pipe_context *pipe, struct pipe_surface *ps, + GLuint clearValue); /** occlusion counting (XXX this may be temporary - we should probably * have generic query objects with begin/end methods) diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index e9b142e780..d7684d2044 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -38,32 +38,13 @@ #include "colormac.h" -static GLuint -color_value(GLuint format, const GLfloat color[4]) -{ - GLubyte r, g, b, a; - - UNCLAMPED_FLOAT_TO_UBYTE(r, color[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, color[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, color[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, color[3]); - - switch (format) { - case PIPE_FORMAT_U_R8_G8_B8_A8: - return (r << 24) | (g << 16) | (b << 8) | a; - case PIPE_FORMAT_U_A8_R8_G8_B8: - return (a << 24) | (r << 16) | (g << 8) | b; - case PIPE_FORMAT_U_R5_G6_B5: - return ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); - default: - return 0; - } -} - - +/** + * Clear the given surface to the specified value. + * No masking, no scissor (clear entire buffer). + */ void -softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, - GLboolean stencil) +softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, + GLuint clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); GLint x, y, w, h; @@ -75,81 +56,5 @@ softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, w = softpipe->framebuffer.cbufs[0]->width; h = softpipe->framebuffer.cbufs[0]->height; - if (color) { - GLuint i; - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - struct pipe_surface *ps = softpipe->framebuffer.cbufs[i]; - GLuint clearVal = color_value(ps->format, - softpipe->clear_color.color); - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, ~0); - } - } - - if (depth && stencil && - softpipe->framebuffer.zbuf == softpipe->framebuffer.sbuf) { - /* clear Z and stencil together */ - struct pipe_surface *ps = softpipe->framebuffer.zbuf; - if (ps->format == PIPE_FORMAT_S8_Z24) { - GLuint mask = (softpipe->stencil.write_mask[0] << 8) | 0xffffff; - GLuint clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffff); - - assert (mask == ~0); - - clearVal |= (softpipe->stencil.clear_value << 24); - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, ~0); - } - else { - /* XXX Z24_S8 format? */ - assert(0); - } - } - else { - /* separate Z and stencil */ - if (depth) { - struct pipe_surface *ps = softpipe->framebuffer.zbuf; - GLuint clearVal; - - switch (ps->format) { - case PIPE_FORMAT_U_Z16: - clearVal = (GLuint) (softpipe->depth_test.clear * 65535.0); - break; - case PIPE_FORMAT_U_Z32: - clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffffff); - break; - case PIPE_FORMAT_S8_Z24: - clearVal = (GLuint) (softpipe->depth_test.clear * 0xffffff); - break; - default: - assert(0); - } - - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, ~0); - } - - if (stencil) { - struct pipe_surface *ps = softpipe->framebuffer.sbuf; - GLuint clearVal = softpipe->stencil.clear_value; - - /* If this is not ~0, we shouldn't get here - clear should be - * done with geometry instead. - */ - GLuint mask = softpipe->stencil.write_mask[0]; - - assert((mask & 0xff) == 0xff); - - switch (ps->format) { - case PIPE_FORMAT_S8_Z24: - clearVal = clearVal << 24; - mask = mask << 24; - break; - case PIPE_FORMAT_U_S8: - /* nothing */ - break; - default: - assert(0); - } - - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearVal, mask); - } - } + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue, ~0); } diff --git a/src/mesa/pipe/softpipe/sp_clear.h b/src/mesa/pipe/softpipe/sp_clear.h index d41cc1d070..e706e731c2 100644 --- a/src/mesa/pipe/softpipe/sp_clear.h +++ b/src/mesa/pipe/softpipe/sp_clear.h @@ -36,8 +36,8 @@ struct pipe_context; extern void -softpipe_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, - GLboolean stencil); +softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, + GLuint clearValue); #endif /* SP_CLEAR_H */ diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 85df549404..bd58abc4d1 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -43,6 +43,65 @@ #include "vf/vf.h" + +static GLuint +color_value(GLuint pipeFormat, const GLfloat color[4]) +{ + GLubyte r, g, b, a; + + UNCLAMPED_FLOAT_TO_UBYTE(r, color[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, color[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, color[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, color[3]); + + switch (pipeFormat) { + case PIPE_FORMAT_U_R8_G8_B8_A8: + return (r << 24) | (g << 16) | (b << 8) | a; + case PIPE_FORMAT_U_A8_R8_G8_B8: + return (a << 24) | (r << 16) | (g << 8) | b; + case PIPE_FORMAT_U_R5_G6_B5: + return ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); + default: + return 0; + } +} + + +static GLuint +depth_value(GLuint pipeFormat, GLfloat value) +{ + GLuint val; + switch (pipeFormat) { + case PIPE_FORMAT_U_Z16: + val = (GLuint) (value * 0xffffff); + break; + case PIPE_FORMAT_U_Z32: + val = (GLuint) (value * 0xffffffff); + break; + case PIPE_FORMAT_S8_Z24: + /*case PIPE_FORMAT_Z24_S8:*/ + val = (GLuint) (value * 0xffffff); + break; + default: + assert(0); + } + return val; +} + + +static GLboolean +is_depth_stencil_format(GLuint pipeFormat) +{ + switch (pipeFormat) { + case PIPE_FORMAT_S8_Z24: + /*case PIPE_FORMAT_Z24_S8:*/ + return GL_TRUE; + default: + return GL_FALSE; + } +} + + /** * Draw a screen-aligned quadrilateral. * Coords are window coords. @@ -92,7 +151,8 @@ draw_quad(GLcontext *ctx, * Do glClear by drawing a quadrilateral. */ static void -clear_with_quad(GLcontext *ctx, +clear_with_quad(GLcontext *ctx, GLuint x0, GLuint y0, + GLuint x1, GLuint y1, GLboolean color, GLboolean depth, GLboolean stencil) { struct st_context *st = ctx->st; @@ -133,6 +193,8 @@ clear_with_quad(GLcontext *ctx, /* setup state: nothing */ memset(&setup, 0, sizeof(setup)); + if (ctx->Scissor.Enabled) + setup.scissor = 1; st->pipe->set_setup_state(st->pipe, &setup); /* stencil state: always set to ref value */ @@ -145,18 +207,14 @@ clear_with_quad(GLcontext *ctx, stencil_test.front_zfail_op = PIPE_STENCIL_OP_REPLACE; stencil_test.ref_value[0] = ctx->Stencil.Clear; stencil_test.value_mask[0] = 0xff; - stencil_test.write_mask[0] = ctx->Stencil.WriteMask[0]; + stencil_test.write_mask[0] = ctx->Stencil.WriteMask[0] & 0xff; } st->pipe->set_stencil_state(st->pipe, &stencil_test); /* draw quad matching scissor rect (XXX verify coord round-off) */ - draw_quad(ctx, - ctx->Scissor.X, ctx->Scissor.Y, - ctx->Scissor.X + ctx->Scissor.Width, - ctx->Scissor.Y + ctx->Scissor.Height, - ctx->Depth.Clear, ctx->Color.ClearColor); + draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); - /* Restore GL state */ + /* Restore pipe state */ st->pipe->set_alpha_test_state(st->pipe, &st->state.alpha_test); st->pipe->set_blend_state(st->pipe, &st->state.blend); st->pipe->set_depth_state(st->pipe, &st->state.depth); @@ -168,6 +226,141 @@ clear_with_quad(GLcontext *ctx, } +static void +clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + if (ctx->Color.ColorMask[0] && + ctx->Color.ColorMask[1] && + ctx->Color.ColorMask[2] && + ctx->Color.ColorMask[3] && + !ctx->Scissor.Enabled) + { + /* clear whole buffer w/out masking */ + GLuint clearValue + = color_value(rb->surface->format, ctx->Color.ClearColor); + ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + } + else { + /* masking or scissoring */ + clear_with_quad(ctx, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmax, + ctx->DrawBuffer->_Ymax, + GL_TRUE, GL_FALSE, GL_FALSE); + } +} + + +static void +clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + if (!ctx->Scissor.Enabled) { + /* clear whole buffer w/out masking */ + GLuint clearValue + = color_value(rb->surface->format, ctx->Accum.ClearColor); + /* Note that clearValue is 32 bits but the accum buffer will + * typically be 64bpp... + */ + ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + } + else { + /* scissoring */ + /* XXX point framebuffer.cbufs[0] at the accum buffer */ + clear_with_quad(ctx, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmax, + ctx->DrawBuffer->_Ymax, + GL_TRUE, GL_FALSE, GL_FALSE); + } +} + + +static void +clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + if (!ctx->Scissor.Enabled && + !is_depth_stencil_format(rb->surface->format)) { + /* clear whole depth buffer w/out masking */ + GLuint clearValue = depth_value(rb->surface->format, ctx->Depth.Clear); + ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + } + else { + /* masking or scissoring or combined z/stencil buffer */ + clear_with_quad(ctx, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmax, + ctx->DrawBuffer->_Ymax, + GL_FALSE, GL_TRUE, GL_FALSE); + } +} + + +static void +clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + const GLuint stencilMax = (1 << rb->StencilBits) - 1; + GLboolean maskStencil = ctx->Stencil.WriteMask[0] != stencilMax; + + if (!maskStencil && !ctx->Scissor.Enabled && + !is_depth_stencil_format(rb->surface->format)) { + /* clear whole stencil buffer w/out masking */ + GLuint clearValue = ctx->Stencil.Clear; + ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + } + else { + /* masking or scissoring */ + clear_with_quad(ctx, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmax, + ctx->DrawBuffer->_Ymax, + GL_FALSE, GL_FALSE, GL_TRUE); + } +} + + +static void +clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + const GLuint stencilMax = 1 << rb->StencilBits; + GLboolean maskStencil = ctx->Stencil.WriteMask[0] != stencilMax; + + assert(is_depth_stencil_format(rb->surface->format)); + + if (!maskStencil && !ctx->Scissor.Enabled) { + /* clear whole buffer w/out masking */ + GLuint clearValue = depth_value(rb->surface->format, ctx->Depth.Clear); + + switch (rb->surface->format) { + case PIPE_FORMAT_S8_Z24: + clearValue |= ctx->Stencil.Clear << 24; + break; +#if 0 + case PIPE_FORMAT_Z24_S8: + clearValue = (clearValue << 8) | clearVal; + break; +#endif + default: + assert(0); + } + + ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + } + else { + /* masking or scissoring */ + clear_with_quad(ctx, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmin, + ctx->DrawBuffer->_Xmax, + ctx->DrawBuffer->_Ymax, + GL_FALSE, GL_TRUE, GL_TRUE); + } +} + + /** * Called via ctx->Driver.Clear() @@ -176,41 +369,52 @@ clear_with_quad(GLcontext *ctx, */ static void st_clear(GLcontext *ctx, GLbitfield mask) { + static const GLbitfield BUFFER_BITS_DS + = (BUFFER_BIT_DEPTH | BUFFER_BIT_STENCIL); struct st_context *st = ctx->st; - GLboolean color = (mask & BUFFER_BITS_COLOR) ? GL_TRUE : GL_FALSE; - GLboolean depth = (mask & BUFFER_BIT_DEPTH) ? GL_TRUE : GL_FALSE; - GLboolean stencil = (mask & BUFFER_BIT_STENCIL) ? GL_TRUE : GL_FALSE; - GLboolean accum = (mask & BUFFER_BIT_ACCUM) ? GL_TRUE : GL_FALSE; - - GLboolean maskColor, maskStencil; - GLboolean fullscreen = !ctx->Scissor.Enabled; - GLuint stencilMax = stencil ? (1 << ctx->DrawBuffer->_StencilBuffer->StencilBits) : 0; + struct gl_renderbuffer *depthRb + = ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer; + struct gl_renderbuffer *stencilRb + = ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer; /* This makes sure the softpipe has the latest scissor, etc values */ st_validate_state( st ); - maskColor = color && st->state.blend.colormask != PIPE_MASK_RGBA; - maskStencil = stencil && ctx->Stencil.WriteMask[0] != stencilMax; - - if (fullscreen && !maskColor && !maskStencil) { - /* pipe->clear() should clear a particular surface, so that we - * can iterate over render buffers at this level and clear the - * ones GL is asking for. - * - * Will probably need something like pipe->clear_z_stencil() to - * cope with the special case of paired and unpaired z/stencil - * buffers, though could perhaps deal with them explicitly at - * this level. - */ - st->pipe->clear(st->pipe, color, depth, stencil); + /* + * XXX TO-DO: + * If we're going to use clear_with_quad() for any reason, use it to + * clear as many other buffers as possible. + * As it is now, we sometimes call clear_with_quad() three times to clear + * color/depth/stencil individually... + */ + + if (mask & BUFFER_BITS_COLOR) { + GLuint b; + for (b = 0; b < BUFFER_COUNT; b++) { + if (BUFFER_BITS_COLOR & mask & (1 << b)) { + clear_color_buffer(ctx, + ctx->DrawBuffer->Attachment[b].Renderbuffer); + } + } + } - /* And here we would do a clear on whatever surface we are using - * to implement accum buffers: - */ - assert(!accum); + if (mask & BUFFER_BIT_ACCUM) { + clear_accum_buffer(ctx, + ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer); + } + + if ((mask & BUFFER_BITS_DS) == BUFFER_BITS_DS && depthRb == stencilRb) { + /* clearing combined depth + stencil */ + clear_depth_stencil_buffer(ctx, depthRb); } else { - clear_with_quad(ctx, color, depth, stencil); + /* separate depth/stencil clears */ + if (mask & BUFFER_BIT_DEPTH) { + clear_depth_buffer(ctx, depthRb); + } + if (mask & BUFFER_BIT_STENCIL) { + clear_stencil_buffer(ctx, stencilRb); + } } } -- cgit v1.2.3 From 4ab3bbba295a5b39aacae8015130a9098954ad68 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 20:36:06 -0600 Subject: beginings of teximage functions --- src/mesa/state_tracker/st_cb_teximage.c | 265 ++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_teximage.h | 38 +++++ 2 files changed, 303 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_teximage.c create mode 100644 src/mesa/state_tracker/st_cb_teximage.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_teximage.c b/src/mesa/state_tracker/st_cb_teximage.c new file mode 100644 index 0000000000..5d651d96bc --- /dev/null +++ b/src/mesa/state_tracker/st_cb_teximage.c @@ -0,0 +1,265 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Texture Image-related functions. + * \author Brian Paul + */ + +#include "main/imports.h" +#include "main/context.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "st_context.h" +#include "st_cb_teximage.h" + + +/** + * Search list of formats for first RGBA format. + */ +static GLuint +default_rgba_format(const GLuint formats[], GLuint num) +{ + GLuint i; + for (i = 0; i < num; i++) { + if (formats[i] == PIPE_FORMAT_U_R8_G8_B8_A8 || + formats[i] == PIPE_FORMAT_U_A8_R8_G8_B8 || + formats[i] == PIPE_FORMAT_U_R5_G6_B5) { + return formats[i]; + } + } + return PIPE_FORMAT_NONE; +} + + +/** + * Search list of formats for first depth/Z format. + */ +static GLuint +default_depth_format(const GLuint formats[], GLuint num) +{ + GLuint i; + for (i = 0; i < num; i++) { + if (formats[i] == PIPE_FORMAT_U_Z16 || + formats[i] == PIPE_FORMAT_U_Z32 || + formats[i] == PIPE_FORMAT_S8_Z24) { + return formats[i]; + } + } + return PIPE_FORMAT_NONE; +} + + +/** + * Choose the PIPE_FORMAT_ to use for storing a texture image based + * on the user's internalFormat, format and type parameters. + * We query the pipe device for a list of formats which it supports + * and choose from them. + * If we find a device that needs a more intricate selection mechanism, + * this function _could_ get pushed down into the pipe device. + * + * \return PIPE_FORMAT_NONE if error/problem. + */ +static GLuint +choose_tex_format(struct pipe_context *pipe, GLint internalFormat, + GLenum format, GLenum type) +{ + const GLuint *supported; + GLboolean allow[PIPE_FORMAT_COUNT]; + GLuint i, n; + + /* query supported formats and fill in bool allow[] table */ + supported = pipe->supported_formats(pipe, &n); + assert(n < PIPE_FORMAT_COUNT); /* sanity check */ + memset(allow, 0, sizeof(allow)); + for (i = 0; i < n; i++) { + allow[supported[i]] = 1; + } + + switch (internalFormat) { + case 4: + case GL_RGBA: + case GL_COMPRESSED_RGBA: + if (format == GL_BGRA) { + if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { + if (allow[PIPE_FORMAT_U_A8_R8_G8_B8]) + return PIPE_FORMAT_U_A8_R8_G8_B8; + } + else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { + /* + if (allow[PIPE_FORMAT_U_A4_R4_G4_B4]) + return PIPE_FORMAT_U_A4_R4_G4_B4; + */ + } + else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { + /* + if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) + return PIPE_FORMAT_U_A1_R5_G5_B5; + */ + } + } + return default_rgba_format(supported, n); +#if 0 + case 3: + case GL_RGB: + case GL_COMPRESSED_RGB: + if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { + return &_mesa_texformat_rgb565; + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case GL_RGBA4: + case GL_RGBA2: + return &_mesa_texformat_argb4444; + + case GL_RGB5_A1: + return &_mesa_texformat_argb1555; + + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + return &_mesa_texformat_argb8888; + + case GL_RGB5: + case GL_RGB4: + case GL_R3_G3_B2: + return &_mesa_texformat_rgb565; + + case GL_ALPHA: + case GL_ALPHA4: + case GL_ALPHA8: + case GL_ALPHA12: + case GL_ALPHA16: + case GL_COMPRESSED_ALPHA: + return &_mesa_texformat_a8; + + case 1: + case GL_LUMINANCE: + case GL_LUMINANCE4: + case GL_LUMINANCE8: + case GL_LUMINANCE12: + case GL_LUMINANCE16: + case GL_COMPRESSED_LUMINANCE: + return &_mesa_texformat_l8; + + case 2: + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE4_ALPHA4: + case GL_LUMINANCE6_ALPHA2: + case GL_LUMINANCE8_ALPHA8: + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: + case GL_COMPRESSED_LUMINANCE_ALPHA: + return &_mesa_texformat_al88; + + case GL_INTENSITY: + case GL_INTENSITY4: + case GL_INTENSITY8: + case GL_INTENSITY12: + case GL_INTENSITY16: + case GL_COMPRESSED_INTENSITY: + return &_mesa_texformat_i8; + + case GL_YCBCR_MESA: + if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) + return &_mesa_texformat_ycbcr; + else + return &_mesa_texformat_ycbcr_rev; + + case GL_COMPRESSED_RGB_FXT1_3DFX: + return &_mesa_texformat_rgb_fxt1; + case GL_COMPRESSED_RGBA_FXT1_3DFX: + return &_mesa_texformat_rgba_fxt1; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + return &_mesa_texformat_rgb_dxt1; + + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return &_mesa_texformat_rgba_dxt1; + + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + return &_mesa_texformat_rgba_dxt3; + + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + return &_mesa_texformat_rgba_dxt5; +#endif + + case GL_DEPTH_COMPONENT16: + if (allow[PIPE_FORMAT_U_Z16]) + return PIPE_FORMAT_U_Z16; + /* fall-through */ + case GL_DEPTH_COMPONENT24: + if (allow[PIPE_FORMAT_S8_Z24]) + return PIPE_FORMAT_S8_Z24; + /* fall-through */ + case GL_DEPTH_COMPONENT32: + if (allow[PIPE_FORMAT_U_Z32]) + return PIPE_FORMAT_U_Z32; + /* fall-through */ + case GL_DEPTH_COMPONENT: + return default_depth_format(supported, n); + + +#if 0 + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + return &_mesa_texformat_z24_s8; +#endif + default: + return PIPE_FORMAT_NONE; + } +} + + + +void st_init_cb_teximage( struct st_context *st ) +{ + /* + struct dd_function_table *functions = &st->ctx->Driver; + + functions->TexImage2D = st_teximage2d; + */ +} + + +void st_destroy_cb_teximage( struct st_context *st ) +{ +} diff --git a/src/mesa/state_tracker/st_cb_teximage.h b/src/mesa/state_tracker/st_cb_teximage.h new file mode 100644 index 0000000000..6e62a2e3c9 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_teximage.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_TEXIMAGE_H +#define ST_CB_TEXIMAGE_H + + +extern void st_init_cb_teximage( struct st_context *st ); + +extern void st_destroy_cb_teximage( struct st_context *st ); + + +#endif /* ST_CB_TEXIMAGE_H */ -- cgit v1.2.3 From fc9ff31298952f0a7e4cdfae95059144a5be6e1b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 20:36:36 -0600 Subject: trim #includes --- src/mesa/state_tracker/st_cb_clear.c | 1 - src/mesa/state_tracker/st_cb_drawpixels.c | 3 --- 2 files changed, 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index bd58abc4d1..c907b0ed22 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -33,7 +33,6 @@ #include "glheader.h" #include "macros.h" -#include "enums.h" #include "st_atom.h" #include "st_context.h" #include "st_cb_clear.h" diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 7209db00a3..13f5c5f3c7 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -32,9 +32,6 @@ #include "imports.h" -#include "tnl/t_context.h" -#include "tnl/t_pipeline.h" - #include "st_context.h" #include "st_atom.h" #include "st_draw.h" -- cgit v1.2.3 From 66ae07854e7e6e9f56e0beb6b56a8c1429bfa3c6 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 20:39:45 -0600 Subject: more formats --- src/mesa/pipe/p_defines.h | 34 +++++++++++++++++++--------------- 1 file changed, 19 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 4c29ec2c0e..a3288134db 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -134,21 +134,25 @@ /** * Texture/surface image formats (preliminary) */ -#define PIPE_FORMAT_NONE 0 /**< unstructured */ -#define PIPE_FORMAT_U_R8_G8_B8_A8 1 /**< ubyte[4] RGBA */ -#define PIPE_FORMAT_U_A8_R8_G8_B8 2 /**< ubyte[4] ARGB */ -#define PIPE_FORMAT_U_R5_G6_B5 3 /**< 5/6/5 RGB */ -#define PIPE_FORMAT_U_L8 4 /**< ubyte luminance */ -#define PIPE_FORMAT_U_A8 5 /**< ubyte alpha */ -#define PIPE_FORMAT_U_I8 6 /**< ubyte intensity */ -#define PIPE_FORMAT_U_L8_A8 7 /**< ubyte luminance, alpha */ -#define PIPE_FORMAT_YCBCR 8 -#define PIPE_FORMAT_YCBCR_REV 9 -#define PIPE_FORMAT_U_Z16 10 /**< ushort Z/depth */ -#define PIPE_FORMAT_U_Z32 11 /**< uint Z/depth */ -#define PIPE_FORMAT_F_Z32 12 /**< float Z/depth */ -#define PIPE_FORMAT_S8_Z24 13 /**< 8-bit stencil + 24-bit Z */ -#define PIPE_FORMAT_U_S8 14 /**< 8-bit stencil */ +#define PIPE_FORMAT_NONE 0 /**< unstructured */ +#define PIPE_FORMAT_U_R8_G8_B8_A8 1 /**< ubyte[4] RGBA */ +#define PIPE_FORMAT_U_A8_R8_G8_B8 2 /**< ubyte[4] ARGB */ +#define PIPE_FORMAT_U_A1_R5_G5_B5 3 /**< 16-bit packed RGBA */ +#define PIPE_FORMAT_U_A4_R4_G4_B4 4 /**< 16-bit packed RGBA */ +#define PIPE_FORMAT_U_R5_G6_B5 5 /**< 16-bit packed RGB */ +#define PIPE_FORMAT_U_L8 6 /**< ubyte luminance */ +#define PIPE_FORMAT_U_A8 7 /**< ubyte alpha */ +#define PIPE_FORMAT_U_I8 8 /**< ubyte intensity */ +#define PIPE_FORMAT_U_L8_A8 9 /**< ubyte luminance, alpha */ +#define PIPE_FORMAT_S_R16_G16_B16_A16 10 /**< signed 16-bit RGBA (accum) */ +#define PIPE_FORMAT_YCBCR 11 +#define PIPE_FORMAT_YCBCR_REV 12 +#define PIPE_FORMAT_U_Z16 13 /**< ushort Z/depth */ +#define PIPE_FORMAT_U_Z32 14 /**< uint Z/depth */ +#define PIPE_FORMAT_F_Z32 15 /**< float Z/depth */ +#define PIPE_FORMAT_S8_Z24 16 /**< 8-bit stencil + 24-bit Z */ +#define PIPE_FORMAT_U_S8 17 /**< 8-bit stencil */ +#define PIPE_FORMAT_COUNT 18 /**< number of formats */ /** -- cgit v1.2.3 From 61d0215d7b35bff980acbf1d61c764f8b80c0e71 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 20:40:19 -0600 Subject: hook in teximage bits --- src/mesa/sources | 1 + src/mesa/state_tracker/st_context.c | 10 +++++++++- 2 files changed, 10 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index a7c31001c2..8074a43c85 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -194,6 +194,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_drawpixels.c \ state_tracker/st_cb_program.c \ + state_tracker/st_cb_teximage.c \ state_tracker/st_draw.c \ state_tracker/st_context.c \ state_tracker/st_texobj.c diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 168ad33d6e..6814eac2dc 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -30,6 +30,7 @@ #include "st_context.h" #include "st_cb_clear.h" #include "st_cb_drawpixels.h" +#include "st_cb_teximage.h" #include "st_atom.h" #include "st_draw.h" #include "st_program.h" @@ -59,9 +60,11 @@ struct st_context *st_create_context( GLcontext *ctx, st_init_atoms( st ); st_init_draw( st ); - st_init_cb_program( st ); + st_init_cb_clear( st ); + st_init_cb_program( st ); st_init_cb_drawpixels( st ); + st_init_cb_teximage( st ); return st; } @@ -71,7 +74,12 @@ void st_destroy_context( struct st_context *st ) { st_destroy_atoms( st ); st_destroy_draw( st ); + + st_destroy_cb_clear( st ); st_destroy_cb_program( st ); + st_destroy_cb_drawpixels( st ); + st_destroy_cb_teximage( st ); + st->pipe->destroy( st->pipe ); FREE( st ); } -- cgit v1.2.3 From c358a39af5968faf27c5fdc85bf64ac76fa2486b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 2 Aug 2007 20:40:33 -0600 Subject: added pipe->supported_formats() --- src/mesa/pipe/p_context.h | 6 ++++++ src/mesa/pipe/softpipe/sp_context.c | 36 ++++++++++++++++++++++++++++++++++++ 2 files changed, 42 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 4f8bdae140..7acaabbdfd 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -49,6 +49,12 @@ struct pipe_context { void (*destroy)( struct pipe_context * ); + /* + * Queries + */ + const GLuint *(*supported_formats)(struct pipe_context *pipe, + GLuint *numFormats); + /* * Drawing */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 9fba9605e8..fe7f2f351a 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -41,6 +41,39 @@ #include "sp_prim_setup.h" + +/** + * Return list of supported surface/texture formats. + * If we find texture and drawable support differs, add a selector + * parameter or another function. + */ +static const GLuint * +softpipe_supported_formats(struct pipe_context *pipe, GLuint *numFormats) +{ + static const GLuint supported[] = { + PIPE_FORMAT_U_R8_G8_B8_A8, + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_U_L8, + PIPE_FORMAT_U_A8, + PIPE_FORMAT_U_I8, + PIPE_FORMAT_U_L8_A8, + PIPE_FORMAT_S_R16_G16_B16_A16, + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + PIPE_FORMAT_U_Z16, + PIPE_FORMAT_U_Z32, + PIPE_FORMAT_F_Z32, + PIPE_FORMAT_S8_Z24, + PIPE_FORMAT_U_S8 + }; + + *numFormats = sizeof(supported)/sizeof(supported[0]); + return supported; +} + + + static void map_surfaces(struct softpipe_context *sp) { struct pipe_context *pipe = &sp->pipe; @@ -140,6 +173,9 @@ struct pipe_context *softpipe_create( void ) struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); softpipe->pipe.destroy = softpipe_destroy; + + softpipe->pipe.supported_formats = softpipe_supported_formats; + softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; softpipe->pipe.set_blend_state = softpipe_set_blend_state; -- cgit v1.2.3 From 389b1d1a3fe742004165d7dcb381314dd9c6806f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 3 Aug 2007 12:28:03 +0100 Subject: Workaround wierd oops on gutsy when building mesa. --- src/mesa/Makefile | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index ee4cbd0d38..7d4bd22203 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -141,8 +141,9 @@ $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECT depend: $(ALL_SOURCES) @ echo "running $(MKDEP)" + @ rm -f depend # workaround oops on gutsy?!? @ touch depend - @$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \ + @ $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \ > /dev/null 2>/dev/null -- cgit v1.2.3 From b8846a5be9443ed5d4027ca164c159f93e707824 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 3 Aug 2007 13:14:27 +0100 Subject: Simplify frontbuffer / sarea / rotation management. Remove lots of old cruft. --- src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_blit.c | 17 +- src/mesa/drivers/dri/i915pipe/intel_context.c | 17 +- src/mesa/drivers/dri/i915pipe/intel_screen.c | 251 +++++----------------- src/mesa/drivers/dri/i915pipe/intel_screen.h | 59 +++-- src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_tex_format.c | 2 +- 7 files changed, 97 insertions(+), 253 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c index 9259f2ac05..97e3882bc9 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c @@ -87,7 +87,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) * Get a new, free batchbuffer. */ - batch->size = batch->intel->intelScreen->maxBatchSize; + batch->size = BATCH_SZ; driBOData(batch->buffer, batch->size, NULL, 0); driBOResetList(&batch->list); diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.c b/src/mesa/drivers/dri/i915pipe/intel_blit.c index 062cd2632c..5d9cb0cf0f 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_blit.c +++ b/src/mesa/drivers/dri/i915pipe/intel_blit.c @@ -86,8 +86,6 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, if (dPriv && dPriv->numClipRects) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - const struct pipe_region *frontRegion - = intelScreen->front_region; const struct pipe_region *backRegion = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : @@ -96,20 +94,19 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, const int backHeight = intel_fb->Base.Height; const int nbox = dPriv->numClipRects; const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = frontRegion->pitch; + const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; const int srcpitch = backRegion->pitch; - const int cpp = frontRegion->cpp; + const int cpp = intelScreen->front.cpp; int BR13, CMD; int i; ASSERT(intel_fb); ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ - ASSERT(frontRegion); ASSERT(backRegion); - ASSERT(frontRegion->cpp == backRegion->cpp); + ASSERT(backRegion->cpp == cpp); DBG("front pitch %d back pitch %d\n", - frontRegion->pitch, backRegion->pitch); + pitch, backRegion->pitch); if (cpp == 2) { BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); @@ -127,7 +124,8 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, if (pbox->x1 > pbox->x2 || pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->width || pbox->y2 > intelScreen->height) + pbox->x2 > intelScreen->front.width || + pbox->y2 > intelScreen->front.height) continue; box = *pbox; @@ -170,7 +168,8 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, OUT_BATCH((box.y1 << 16) | box.x1); OUT_BATCH((box.y2 << 16) | box.x2); - OUT_RELOC(frontRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + OUT_RELOC(intelScreen->front.buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); OUT_BATCH((sbox.y1 << 16) | sbox.x1); OUT_BATCH((srcpitch * cpp) & 0xffff); diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index bab94e2697..f0ca5c75c7 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -337,8 +337,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, GLcontext *shareCtx = (GLcontext *) sharedContextPrivate; __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - drmI830Sarea *saPriv = (drmI830Sarea *) - (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset); + drmI830Sarea *saPriv = intelScreen->sarea; int fthrottle_mode; GLboolean havePools; @@ -409,7 +408,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->pipe = intel->ctx.st->pipe; intel->pipe->screen = intelScreen; intel->pipe->glctx = ctx; - intelScreen->pipe = intel->pipe; intel_init_region_functions(intel->pipe); /* @@ -453,11 +451,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->iw.irq_seq = -1; intel->irqsEmitted = 0; - intel->do_irqs = (intel->intelScreen->irq_active && - fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS); - - intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); - _math_matrix_ctr(&intel->ViewportMatrix); /* Disable imaging extension until convolution is working in @@ -598,9 +591,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) { int i; - intel_fb->vblank_flags = (intel->intelScreen->irq_active != 0) - ? driGetDefaultVBlankFlags(&intel->optionCache) - : VBLANK_FLAG_NO_IRQ; + intel_fb->vblank_flags = driGetDefaultVBlankFlags(&intel->optionCache); (*dri_interface->getUST) (&intel_fb->swap_ust); driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags, @@ -651,8 +642,8 @@ intelContendedLock(struct intel_context *intel, GLuint flags) if (dPriv) DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - if (sarea->width != intelScreen->width || - sarea->height != intelScreen->height) { + if (sarea->width != intelScreen->front.width || + sarea->height != intelScreen->front.height) { intelUpdateScreenRotation(sPriv, sarea); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.c b/src/mesa/drivers/dri/i915pipe/intel_screen.c index 565bfc5243..5b8d6256b5 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.c +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.c @@ -65,106 +65,8 @@ PUBLIC const char __driConfigOptions[] = extern const struct dri_extension card_extensions[]; -/** - * Map all the memory regions described by the screen. - * \return GL_TRUE if success, GL_FALSE if error. - */ -GLboolean -intelMapScreenRegions(__DRIscreenPrivate * sPriv) -{ - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - - if (intelScreen->front.handle) { - if (drmMap(sPriv->fd, - intelScreen->front.handle, - intelScreen->front.size, - (drmAddress *) & intelScreen->front.map) != 0) { - _mesa_problem(NULL, "drmMap(frontbuffer) failed!"); - return GL_FALSE; - } - } - else { - _mesa_warning(NULL, "no front buffer handle in intelMapScreenRegions!"); - } - - if (0) - printf("Mappings: front: %p\n", intelScreen->front.map); - return GL_TRUE; -} - - -static struct pipe_region * -intel_recreate_static(struct pipe_context *pipe, - struct pipe_region *region, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height) -{ - if (region) { - pipe->region_update_static(pipe, region, mem_type, offset, - virtual, cpp, pitch, height); - } else { - region = pipe->region_create_static(pipe, mem_type, offset, - virtual, cpp, pitch, height); - } - return region; -} - - -/* Create pipe_region structs to describe the static front,back,depth - * buffers created by the xserver. - * Only used for real front buffer now. - * - * Note that these don't allocate video memory, just describe - * allocations already made by the X server. - */ -static void -intel_recreate_static_regions(intelScreenPrivate *intelScreen) -{ -/* this is the real front buffer which is only used for blitting to */ - intelScreen->front_region = - intel_recreate_static(intelScreen->pipe, - intelScreen->front_region, - DRM_BO_FLAG_MEM_TT, - intelScreen->front.offset, - intelScreen->front.map, - intelScreen->cpp, - intelScreen->front.pitch / intelScreen->cpp, - intelScreen->height); - -} - -/** - * Use the information in the sarea to update the screen parameters - * related to screen rotation. Needs to be called locked. - */ -void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) -{ - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - - intelUnmapScreenRegions(intelScreen); - intelUpdateScreenFromSAREA(intelScreen, sarea); - if (!intelMapScreenRegions(sPriv)) { - fprintf(stderr, "ERROR Remapping screen regions!!!\n"); - } - intel_recreate_static_regions(intelScreen); -} -void -intelUnmapScreenRegions(intelScreenPrivate * intelScreen) -{ -#define REALLY_UNMAP 1 - if (intelScreen->front.map) { -#if REALLY_UNMAP - if (drmUnmap(intelScreen->front.map, intelScreen->front.size) != 0) - printf("drmUnmap front failed!\n"); -#endif - intelScreen->front.map = NULL; - } -} static void @@ -205,26 +107,60 @@ intelPrintSAREA(const drmI830Sarea * sarea) } + /** - * A number of the screen parameters are obtained/computed from - * information in the SAREA. This function updates those parameters. + * Use the information in the sarea to update the screen parameters + * related to screen rotation. Needs to be called locked. */ void -intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, - drmI830Sarea * sarea) +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) { - intelScreen->width = sarea->width; - intelScreen->height = sarea->height; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + if (intelScreen->front.map) { + drmUnmap(intelScreen->front.map, intelScreen->front.size); + intelScreen->front.map = NULL; + } + + if (intelScreen->front.buffer) + driDeleteBuffers(1, &intelScreen->front.buffer); + + intelScreen->front.width = sarea->width; + intelScreen->front.height = sarea->height; intelScreen->front.offset = sarea->front_offset; - intelScreen->front.pitch = sarea->pitch * intelScreen->cpp; - intelScreen->front.handle = sarea->front_handle; + intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; intelScreen->front.size = sarea->front_size; + intelScreen->front.handle = sarea->front_handle; - if (0) - intelPrintSAREA(sarea); + assert( sarea->front_size >= + intelScreen->front.pitch * intelScreen->front.height ); + + if (!sarea->front_handle) + return; + + if (drmMap(sPriv->fd, + sarea->front_handle, + intelScreen->front.size, + (drmAddress *) & intelScreen->front.map) != 0) { + _mesa_problem(NULL, "drmMap(frontbuffer) failed!"); + return; + } + + if (intelScreen->staticPool) { + driGenBuffers(intelScreen->staticPool, "static region", 1, + &intelScreen->front.buffer, 64, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); + + driBOSetStatic(intelScreen->front.buffer, + intelScreen->front.offset, + intelScreen->front.pitch * intelScreen->front.height, + intelScreen->front.map, 0); + } } + + GLboolean intelCreatePools(intelScreenPrivate *intelScreen) { @@ -234,7 +170,7 @@ intelCreatePools(intelScreenPrivate *intelScreen) if (intelScreen->havePools) return GL_TRUE; - batchPoolSize /= intelScreen->maxBatchSize; + batchPoolSize /= BATCH_SZ; intelScreen->regionPool = driDRMPoolInit(sPriv->fd); if (!intelScreen->regionPool) @@ -251,16 +187,17 @@ intelCreatePools(intelScreenPrivate *intelScreen) DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_LOCAL, - intelScreen->maxBatchSize, + BATCH_SZ, batchPoolSize, 5); if (!intelScreen->batchPool) { fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); return GL_FALSE; } - intel_recreate_static_regions(intelScreen); intelScreen->havePools = GL_TRUE; + intelUpdateScreenRotation(sPriv, intelScreen->sarea); + return GL_TRUE; } @@ -270,7 +207,6 @@ intelInitDriver(__DRIscreenPrivate * sPriv) { intelScreenPrivate *intelScreen; I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; - drmI830Sarea *sarea; PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> @@ -285,87 +221,30 @@ intelInitDriver(__DRIscreenPrivate * sPriv) /* Allocate the private area */ intelScreen = (intelScreenPrivate *) CALLOC(sizeof(intelScreenPrivate)); - if (!intelScreen) { - fprintf(stderr, "\nERROR! Allocating private area failed\n"); + if (!intelScreen) return GL_FALSE; - } + /* parse information in __driConfigOptions */ driParseOptionInfo(&intelScreen->optionCache, __driConfigOptions, __driNConfigOptions); intelScreen->driScrnPriv = sPriv; sPriv->private = (void *) intelScreen; - intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset; - sarea = (drmI830Sarea *) - (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset); - intelScreen->maxBatchSize = BATCH_SZ; + intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + + gDRIPriv->sarea_priv_offset); intelScreen->deviceID = gDRIPriv->deviceID; - if (intelScreen->deviceID == PCI_CHIP_I865_G) - intelScreen->maxBatchSize = 4096; - - intelScreen->mem = gDRIPriv->mem; - intelScreen->cpp = gDRIPriv->cpp; - - switch (gDRIPriv->bitsPerPixel) { - case 16: - break; - case 32: - break; - default: - exit(1); - break; - } - - intelUpdateScreenFromSAREA(intelScreen, sarea); - - if (!intelMapScreenRegions(sPriv)) { - fprintf(stderr, "\nERROR! mapping regions\n"); - _mesa_free(intelScreen); - sPriv->private = NULL; - return GL_FALSE; - } + intelScreen->front.cpp = gDRIPriv->cpp; + intelScreen->drmMinor = sPriv->drmMinor; + assert(gDRIPriv->bitsPerPixel == 16 || + gDRIPriv->bitsPerPixel == 32); - intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset; + intelUpdateScreenRotation(sPriv, intelScreen->sarea); if (0) intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); - intelScreen->drmMinor = sPriv->drmMinor; - - /* Determine if IRQs are active? */ - { - int ret; - drmI830GetParam gp; - - gp.param = I830_PARAM_IRQ_ACTIVE; - gp.value = &intelScreen->irq_active; - - ret = drmCommandWriteRead(sPriv->fd, DRM_I830_GETPARAM, - &gp, sizeof(gp)); - if (ret) { - fprintf(stderr, "drmI830GetParam: %d\n", ret); - return GL_FALSE; - } - } - - /* Determine if batchbuffers are allowed */ - { - int ret; - drmI830GetParam gp; - - gp.param = I830_PARAM_ALLOW_BATCHBUFFER; - gp.value = &intelScreen->allow_batchbuffer; - - ret = drmCommandWriteRead(sPriv->fd, DRM_I830_GETPARAM, - &gp, sizeof(gp)); - if (ret) { - fprintf(stderr, "drmI830GetParam: (%d) %d\n", gp.param, ret); - return GL_FALSE; - } - } - if (glx_enable_extension != NULL) { (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); @@ -383,7 +262,7 @@ intelDestroyScreen(__DRIscreenPrivate * sPriv) { intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - intelUnmapScreenRegions(intelScreen); +// intelUnmapScreenRegions(intelScreen); if (intelScreen->havePools) { driPoolTakeDown(intelScreen->regionPool); @@ -403,10 +282,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, __DRIdrawablePrivate * driDrawPriv, const __GLcontextModes * mesaVis, GLboolean isPixmap) { -#if 0 - intelScreenPrivate *screen = (intelScreenPrivate *) driScrnPriv->private; -#endif - if (isPixmap) { return GL_FALSE; /* not implemented */ } @@ -437,16 +312,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, = intel_new_renderbuffer_fb(rgbFormat); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, &intel_fb->color_rb[1]->Base); - -#if 0 - if (screen->third.handle) { - struct gl_renderbuffer *tmp_rb = NULL; - - intel_fb->color_rb[2] - = intel_new_renderbuffer_fb(rgbFormat); - _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base); - } -#endif } if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.h b/src/mesa/drivers/dri/i915pipe/intel_screen.h index de0bea6609..549587a6f2 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.h +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.h @@ -33,37 +33,30 @@ #include "xmlconfig.h" #include "dri_bufpool.h" -typedef struct -{ - drm_handle_t handle; - drmSize size; /* region size in bytes */ - char *map; /* memory map */ - int offset; /* from start of video mem, in bytes */ - int pitch; /* row stride, in bytes */ -} intelRegion; - -typedef struct -{ - struct pipe_context *pipe; /** for accessing region functions */ - intelRegion front; - struct pipe_region *front_region; +struct intel_screen +{ + struct { + drm_handle_t handle; + + /* We create a static dri buffer for the frontbuffer. + */ + struct _DriBufferObject *buffer; + + char *map; /* memory map */ + int offset; /* from start of video mem, in bytes */ + int pitch; /* row stride, in bytes */ + int width; + int height; + int size; + int cpp; /* for front and back buffers */ + } front; int deviceID; - int width; - int height; - int mem; /* unused */ - - int cpp; /* for front and back buffers */ - - __DRIscreenPrivate *driScrnPriv; - unsigned int sarea_priv_offset; - int drmMinor; - int irq_active; - int allow_batchbuffer; - + __DRIscreenPrivate *driScrnPriv; + drmI830Sarea *sarea; /** @@ -74,20 +67,16 @@ typedef struct struct _DriBufferPool *texPool; struct _DriBufferPool *regionPool; struct _DriBufferPool *staticPool; - unsigned int maxBatchSize; GLboolean havePools; - struct intel_context *dummyctxptr; -} intelScreenPrivate; - + struct intel_context *dummyctxptr; +}; -extern GLboolean intelMapScreenRegions(__DRIscreenPrivate * sPriv); - -extern void intelUnmapScreenRegions(intelScreenPrivate * intelScreen); +typedef struct intel_screen intelScreenPrivate; extern void -intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, - drmI830Sarea * sarea); +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); + extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c index 6f4f75079e..6c49c60966 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c @@ -74,7 +74,7 @@ get_teximage_source(struct intel_context *intel, GLenum internalFormat) case GL_RGBA8: return intel_readbuf_region(intel); case GL_RGB: - if (intel->intelScreen->cpp == 2) + if (intel->intelScreen->front.cpp == 2) return intel_readbuf_region(intel); return NULL; default: diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_format.c b/src/mesa/drivers/dri/i915pipe/intel_tex_format.c index 6e058dff69..a11718bb07 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_format.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_format.c @@ -16,7 +16,7 @@ intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, GLenum format, GLenum type) { struct intel_context *intel = intel_context(ctx); - const GLboolean do32bpt = (intel->intelScreen->cpp == 4); + const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); switch (internalFormat) { case 4: -- cgit v1.2.3 From 7d5c25c8a70f42a16b682f4366247cdb8457a631 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 3 Aug 2007 13:22:30 +0100 Subject: Remove "static region" support. The frontbuffer/driBufMgr interactions are handled as a special case in the intel_screen code. --- src/mesa/drivers/dri/i915pipe/intel_regions.c | 61 --------------------------- src/mesa/pipe/p_context.h | 14 ------ 2 files changed, 75 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_regions.c b/src/mesa/drivers/dri/i915pipe/intel_regions.c index f2280ceb51..75ba1f9cdb 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_regions.c +++ b/src/mesa/drivers/dri/i915pipe/intel_regions.c @@ -139,68 +139,9 @@ intel_region_release(struct pipe_context *pipe, struct pipe_region **region) } -static struct pipe_region * -intel_region_create_static(struct pipe_context *pipe, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height) -{ - intelScreenPrivate *intelScreen = pipe_screen(pipe); - struct pipe_region *region = calloc(sizeof(*region), 1); - DBG("%s\n", __FUNCTION__); - - region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ - region->refcount = 1; - - /* - * We use a "shared" buffer type to indicate buffers created and - * shared by others. - */ - - driGenBuffers(intelScreen->staticPool, "static region", 1, - ®ion->buffer, 64, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - driBOSetStatic(region->buffer, offset, pitch * cpp * height, virtual, 0); - - return region; -} -static void -intel_region_update_static(struct pipe_context *pipe, - struct pipe_region *region, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height) -{ - intelScreenPrivate *intelScreen = pipe_screen(pipe); - - DBG("%s\n", __FUNCTION__); - - region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ - - /* - * We use a "shared" buffer type to indicate buffers created and - * shared by others. - */ - - driDeleteBuffers(1, ®ion->buffer); - driGenBuffers(intelScreen->staticPool, "static region", 1, - ®ion->buffer, 64, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - driBOSetStatic(region->buffer, offset, pitch * cpp * height, virtual, 0); - -} - /* @@ -349,8 +290,6 @@ intel_init_region_functions(struct pipe_context *pipe) pipe->region_unmap = intel_region_unmap; pipe->region_alloc = intel_region_alloc; pipe->region_release = intel_region_release; - pipe->region_create_static = intel_region_create_static; - pipe->region_update_static = intel_region_update_static; pipe->region_data = intel_region_data; pipe->region_copy = intel_region_copy; pipe->region_fill = intel_region_fill; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 7acaabbdfd..ac6567f009 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -144,20 +144,6 @@ struct pipe_context { void (*region_release)(struct pipe_context *pipe, struct pipe_region **r); - struct pipe_region *(*region_create_static)(struct pipe_context *pipe, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, - GLuint height); - - void (*region_update_static)(struct pipe_context *pipe, - struct pipe_region *region, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height); - void (*region_idle)(struct pipe_context *pipe, struct pipe_region *region); GLubyte *(*region_map)(struct pipe_context *pipe, struct pipe_region *r); -- cgit v1.2.3 From db70498462344680ff884e7ed576543175833fc0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 3 Aug 2007 18:56:12 +0100 Subject: Rip out more dead drawing-related code. --- src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h | 1 - src/mesa/drivers/dri/i915pipe/intel_context.c | 59 -------- src/mesa/drivers/dri/i915pipe/intel_context.h | 157 +--------------------- src/mesa/drivers/dri/i915pipe/intel_fbo.c | 1 + src/mesa/drivers/dri/i915pipe/intel_tex.c | 28 ++++ src/mesa/drivers/dri/i915pipe/intel_tex.h | 56 ++++++++ 7 files changed, 90 insertions(+), 214 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c index 97e3882bc9..60bd8eaec2 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c @@ -337,6 +337,6 @@ intel_batchbuffer_data(struct intel_batchbuffer *batch, { assert((bytes & 3) == 0); intel_batchbuffer_require_space(batch, bytes, flags); - __memcpy(batch->ptr, data, bytes); + memcpy(batch->ptr, data, bytes); batch->ptr += bytes; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h index 212f130101..6d4d05e0bb 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h @@ -107,7 +107,6 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch, #define BATCH_LOCALS #define BEGIN_BATCH(n, flags) do { \ - assert(!intel->prim.flush); \ intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ } while (0) diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index f0ca5c75c7..395ffd086b 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -240,8 +240,6 @@ intelInvalidateState(GLcontext * ctx, GLuint new_state) _tnl_invalidate_vertex_state(ctx, new_state); st_invalidate_state( ctx, new_state ); - - intel_context(ctx)->NewGLState |= new_state; } @@ -361,33 +359,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, intel->driScreen->myNum, "i915"); - ctx->Const.MaxTextureMaxAnisotropy = 2.0; - - /* This doesn't yet catch all non-conformant rendering, but it's a - * start. - */ - if (getenv("INTEL_STRICT_CONFORMANCE")) { - intel->strict_conformance = 1; - } - - ctx->Const.MinLineWidth = 1.0; - ctx->Const.MinLineWidthAA = 1.0; - ctx->Const.MaxLineWidth = 3.0; - ctx->Const.MaxLineWidthAA = 3.0; - ctx->Const.LineWidthGranularity = 1.0; - - ctx->Const.MinPointSize = 1.0; - ctx->Const.MinPointSizeAA = 1.0; - ctx->Const.MaxPointSize = 255.0; - ctx->Const.MaxPointSizeAA = 3.0; - ctx->Const.PointSizeGranularity = 1.0; - - /* reinitialize the context point state. - * It depend on constants in __GLcontextRec::Const - */ - _mesa_init_point(ctx); - - ctx->Const.MaxColorAttachments = 4; /* XXX FBO: review this */ /* Initialize the software rasterizer and helper modules. */ _vbo_CreateContext(ctx); @@ -425,34 +396,14 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->driFd = sPriv->fd; intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; - intel->hw_stipple = 1; - - /* XXX FBO: this doesn't seem to be used anywhere */ - switch (mesaVis->depthBits) { - case 0: /* what to do in this case? */ - case 16: - intel->polygon_offset_scale = 1.0 / 0xffff; - break; - case 24: - intel->polygon_offset_scale = 2.0 / 0xffffff; /* req'd to pass glean */ - break; - default: - assert(0); - break; - } - - TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; - intel->RenderIndex = ~0; fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); intel->iw.irq_seq = -1; intel->irqsEmitted = 0; - _math_matrix_ctr(&intel->ViewportMatrix); - /* Disable imaging extension until convolution is working in * teximage paths: */ @@ -476,18 +427,10 @@ intelCreateContext(const __GLcontextModes * mesaVis, _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); } - intel->prim.primitive = ~0; - - #if DO_DEBUG INTEL_DEBUG = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); #endif - if (getenv("INTEL_NO_RAST")) { - fprintf(stderr, "disabling 3D rasterization\n"); - FALLBACK(intel, INTEL_FALLBACK_USER, 1); - } - return GL_TRUE; } @@ -510,8 +453,6 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) _tnl_DestroyContext(&intel->ctx); _vbo_DestroyContext(&intel->ctx); - intel->Fallback = 0; /* don't call _swrast_Flush later */ - intel_batchbuffer_free(intel->batch); if (intel->last_swap_fence) { diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.h b/src/mesa/drivers/dri/i915pipe/intel_context.h index cac9fd7b49..6c728dfe73 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.h +++ b/src/mesa/drivers/dri/i915pipe/intel_context.h @@ -38,11 +38,7 @@ #include "intel_screen.h" #include "i915_drm.h" #include "i830_common.h" -#include "tnl/t_vertex.h" -#define TAG(x) intel##x -#include "tnl_dd/t_dd_vertex.h" -#undef TAG #define DV_PF_555 (1<<8) #define DV_PF_565 (2<<8) @@ -53,67 +49,11 @@ struct pipe_region; struct intel_context; struct _DriBufferObject; -typedef void (*intel_tri_func) (struct intel_context *, intelVertex *, - intelVertex *, intelVertex *); -typedef void (*intel_line_func) (struct intel_context *, intelVertex *, - intelVertex *); -typedef void (*intel_point_func) (struct intel_context *, intelVertex *); - -#define INTEL_FALLBACK_DRAW_BUFFER 0x1 -#define INTEL_FALLBACK_READ_BUFFER 0x2 -#define INTEL_FALLBACK_DEPTH_BUFFER 0x4 -#define INTEL_FALLBACK_STENCIL_BUFFER 0x8 -#define INTEL_FALLBACK_USER 0x10 -#define INTEL_FALLBACK_RENDERMODE 0x20 - -#define FALLBACK( intel, bit, mode ) _mesa_printf("intelFallback not implemented/removed\n") - #define INTEL_WRITE_PART 0x1 #define INTEL_WRITE_FULL 0x2 #define INTEL_READ 0x4 -struct intel_texture_object -{ - struct gl_texture_object base; /* The "parent" object */ - - /* The mipmap tree must include at least these levels once - * validated: - */ - GLuint firstLevel; - GLuint lastLevel; - - /* Offset for firstLevel image: - */ - GLuint textureOffset; - - /* On validation any active images held in main memory or in other - * regions will be copied to this region and the old storage freed. - */ - struct intel_mipmap_tree *mt; - - GLboolean imageOverride; - GLint depthOverride; - GLuint pitchOverride; -}; - - - -struct intel_texture_image -{ - struct gl_texture_image base; - - /* These aren't stored in gl_texture_image - */ - GLuint level; - GLuint face; - - /* If intelImage->mt != NULL, image data is stored here. - * Else if intelImage->base.Data != NULL, image is stored there. - * Else there is no image data. - */ - struct intel_mipmap_tree *mt; -}; #define INTEL_MAX_FIXUP 64 @@ -125,21 +65,12 @@ struct intel_context struct pipe_context *pipe; GLint refcount; - GLuint Fallback; - GLuint NewGLState; struct _DriFenceObject *last_swap_fence; struct _DriFenceObject *first_swap_fence; struct intel_batchbuffer *batch; - struct - { - GLuint id; - GLuint primitive; - GLubyte *start_ptr; - void (*flush) (struct intel_context *); - } prim; GLboolean locked; char *prevLockFile; @@ -148,41 +79,6 @@ struct intel_context GLuint ClearColor565; GLuint ClearColor8888; - /* Offsets of fields within the current vertex: - */ - GLuint coloroffset; - GLuint specoffset; - GLuint wpos_offset; - GLuint wpos_size; - - struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; - GLuint vertex_attr_count; - - GLfloat polygon_offset_scale; /* dependent on depth_scale, bpp */ - - GLboolean hw_stipple; - GLboolean strict_conformance; - - /* AGP memory buffer manager: - */ - struct bufmgr *bm; - - - /* State for intelvb.c and inteltris.c. - */ - GLuint RenderIndex; - GLmatrix ViewportMatrix; - GLenum render_primitive; - GLenum reduced_primitive; - GLuint vertex_size; - GLubyte *verts; /* points to tnl->clipspace.vertex_buf */ - - - /* Fallback rasterization functions - */ - intel_point_func draw_point; - intel_line_func draw_line; - intel_tri_func draw_tri; /* These refer to the current drawing buffer: */ @@ -190,8 +86,6 @@ struct intel_context drm_clip_rect_t *pClipRects; drm_clip_rect_t fboRect; /**< cliprect for rendering */ - GLuint do_usleeps; - int do_irqs; GLuint irqsEmitted; drm_i915_irq_wait_t iw; @@ -219,15 +113,9 @@ void UNLOCK_HARDWARE( struct intel_context *intel ); extern char *__progname; - -#define SUBPIXEL_X 0.125 -#define SUBPIXEL_Y 0.125 - -#define INTEL_FIREVERTICES(intel) \ -do { \ - if ((intel)->prim.flush) \ - (intel)->prim.flush(intel); \ -} while (0) +/* Will become a call into state_tracker: + */ +#define INTEL_FIREVERTICES(intel) /* ================================================================ * Color packing: @@ -248,32 +136,6 @@ do { \ -/* ================================================================ - * From linux kernel i386 header files, copes with odd sizes better - * than COPY_DWORDS would: - * XXX Put this in src/mesa/main/imports.h ??? - */ -#if defined(i386) || defined(__i386__) -static INLINE void * -__memcpy(void *to, const void *from, size_t n) -{ - int d0, d1, d2; - __asm__ __volatile__("rep ; movsl\n\t" - "testb $2,%b4\n\t" - "je 1f\n\t" - "movsw\n" - "1:\ttestb $1,%b4\n\t" - "je 2f\n\t" - "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2) - :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from) - :"memory"); - return (to); -} -#else -#define __memcpy(a,b,c) memcpy(a,b,c) -#endif - - /* ================================================================ * Debugging: @@ -337,6 +199,7 @@ extern void intelInitDriverFunctions(struct dd_function_table *functions); #define MI_BATCH_BUFFER_END (0xA<<23) +#define FALLBACK( ctx, bit, mode ) /*====================================================================== * Inline conversion functions. @@ -348,18 +211,6 @@ intel_context(GLcontext * ctx) return (struct intel_context *) ctx; } -static INLINE struct intel_texture_object * -intel_texture_object(struct gl_texture_object *obj) -{ - return (struct intel_texture_object *) obj; -} - -static INLINE struct intel_texture_image * -intel_texture_image(struct gl_texture_image *img) -{ - return (struct intel_texture_image *) img; -} - extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer *rb); diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index 7b6906dea6..9eee5cce45 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -41,6 +41,7 @@ #include "intel_fbo.h" #include "intel_mipmap_tree.h" #include "intel_regions.h" +#include "intel_tex.h" #include "pipe/p_context.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.c b/src/mesa/drivers/dri/i915pipe/intel_tex.c index b08dee43bc..9e494d21ba 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.c @@ -75,6 +75,34 @@ intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) } + + +/* ================================================================ + * From linux kernel i386 header files, copes with odd sizes better + * than COPY_DWORDS would: + * XXX Put this in src/mesa/main/imports.h ??? + */ +#if defined(i386) || defined(__i386__) +static INLINE void * +__memcpy(void *to, const void *from, size_t n) +{ + int d0, d1, d2; + __asm__ __volatile__("rep ; movsl\n\t" + "testb $2,%b4\n\t" + "je 1f\n\t" + "movsw\n" + "1:\ttestb $1,%b4\n\t" + "je 2f\n\t" + "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2) + :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from) + :"memory"); + return (to); +} +#else +#define __memcpy(a,b,c) memcpy(a,b,c) +#endif + + /* The system memcpy (at least on ubuntu 5.10) has problems copying * to agp (writecombined) memory from a source which isn't 64-byte * aligned - there is a 4x performance falloff. diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.h b/src/mesa/drivers/dri/i915pipe/intel_tex.h index b77d7a1d8a..4a5081eee7 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.h +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.h @@ -32,6 +32,48 @@ #include "intel_context.h" #include "texmem.h" +struct intel_texture_object +{ + struct gl_texture_object base; /* The "parent" object */ + + /* The mipmap tree must include at least these levels once + * validated: + */ + GLuint firstLevel; + GLuint lastLevel; + + /* Offset for firstLevel image: + */ + GLuint textureOffset; + + /* On validation any active images held in main memory or in other + * regions will be copied to this region and the old storage freed. + */ + struct intel_mipmap_tree *mt; + + GLboolean imageOverride; + GLint depthOverride; + GLuint pitchOverride; +}; + + + +struct intel_texture_image +{ + struct gl_texture_image base; + + /* These aren't stored in gl_texture_image + */ + GLuint level; + GLuint face; + + /* If intelImage->mt != NULL, image data is stored here. + * Else if intelImage->base.Data != NULL, image is stored there. + * Else there is no image data. + */ + struct intel_mipmap_tree *mt; +}; + void intelInitTextureFuncs(struct dd_function_table *functions); @@ -148,4 +190,18 @@ void intel_tex_unmap_images(struct intel_context *intel, int intel_compressed_num_bytes(GLuint mesaFormat); + +static INLINE struct intel_texture_object * +intel_texture_object(struct gl_texture_object *obj) +{ + return (struct intel_texture_object *) obj; +} + +static INLINE struct intel_texture_image * +intel_texture_image(struct gl_texture_image *img) +{ + return (struct intel_texture_image *) img; +} + + #endif -- cgit v1.2.3 From ec471e9f0723c2823d1fb882cde72ad1265c0916 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 3 Aug 2007 18:58:41 +0100 Subject: Remove dead files --- src/mesa/drivers/dri/i915pipe/intel_context.c | 2 - src/mesa/drivers/dri/i915pipe/intel_pixel.c | 37 --------------- src/mesa/drivers/dri/i915pipe/intel_pixel.h | 63 -------------------------- src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 1 - 4 files changed, 103 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_pixel.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_pixel.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index 395ffd086b..ac04d4e8c2 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -51,7 +51,6 @@ #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" -#include "intel_pixel.h" #include "intel_buffer_objects.h" #include "intel_fbo.h" @@ -316,7 +315,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) functions->UpdateState = intelInvalidateState; intelInitTextureFuncs(functions); - intelInitPixelFuncs(functions); intelInitBufferFuncs(functions); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_pixel.c b/src/mesa/drivers/dri/i915pipe/intel_pixel.c deleted file mode 100644 index 27de2a6636..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_pixel.c +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portionsalloc - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "state.h" - -#include "intel_context.h" -#include "intel_pixel.h" - - -void -intelInitPixelFuncs(struct dd_function_table *functions) -{ -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_pixel.h b/src/mesa/drivers/dri/i915pipe/intel_pixel.h deleted file mode 100644 index 73a06c11ec..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_pixel.h +++ /dev/null @@ -1,63 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_PIXEL_H -#define INTEL_PIXEL_H - -#include "mtypes.h" - -void intelInitPixelFuncs(struct dd_function_table *functions); - -GLboolean intel_check_blit_fragment_ops(GLcontext * ctx); - -GLboolean intel_check_meta_tex_fragment_ops(GLcontext * ctx); - -GLboolean intel_check_blit_format(struct pipe_region *region, - GLenum format, GLenum type); - - -void intelReadPixels(GLcontext * ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - GLvoid * pixels); - -void intelDrawPixels(GLcontext * ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - GLenum format, - GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid * pixels); - -void intelCopyPixels(GLcontext * ctx, - GLint srcx, GLint srcy, - GLsizei width, GLsizei height, - GLint destx, GLint desty, GLenum type); - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c index 6c49c60966..0a3f153d56 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c @@ -39,7 +39,6 @@ #include "intel_fbo.h" #include "intel_tex.h" #include "intel_blit.h" -#include "intel_pixel.h" #define FILE_DEBUG_FLAG DEBUG_TEXTURE -- cgit v1.2.3 From 4b87d37e90b6d2d9562a192540833b599bda5a35 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 3 Aug 2007 13:28:35 -0600 Subject: framebuffer object functions --- src/mesa/state_tracker/st_cb_fbo.c | 343 +++++++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_fbo.h | 38 ++++ 2 files changed, 381 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_fbo.c create mode 100644 src/mesa/state_tracker/st_cb_fbo.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c new file mode 100644 index 0000000000..6b9ae88dbe --- /dev/null +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -0,0 +1,343 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Framebuffer/renderbuffer functions. + * + * \author Brian Paul + */ + + +#include "main/imports.h" +#include "main/context.h" +#include "main/fbobject.h" +#include "main/framebuffer.h" +#include "main/renderbuffer.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "st_context.h" +#include "st_cb_fbo.h" +#include "st_cb_teximage.h" + + +/** + * Derived renderbuffer class. Just need to add a pointer to the + * pipe surface. + */ +struct st_renderbuffer +{ + struct gl_renderbuffer Base; + struct pipe_surface *surface; +}; + + +/** + * Cast wrapper. + */ +static INLINE struct st_renderbuffer * +st_renderbuffer(struct gl_renderbuffer *rb) +{ + return (struct st_renderbuffer *) rb; +} + + +struct pipe_format_info +{ + GLuint format; + GLenum base_format; + GLubyte red_bits; + GLubyte green_bits; + GLubyte blue_bits; + GLubyte alpha_bits; + GLubyte luminance_bits; + GLubyte intensity_bits; + GLubyte depth_bits; + GLubyte stencil_bits; + GLubyte size; /**< in bytes */ +}; + + +/* + * XXX temporary here + */ +static const struct pipe_format_info * +pipe_get_format_info(GLuint format) +{ + static const struct pipe_format_info info[] = { + { + PIPE_FORMAT_U_R8_G8_B8_A8, /* format */ + GL_RGBA, /* base_format */ + 4, 4, 4, 4, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 4 /* size in bytes */ + }, + { + PIPE_FORMAT_U_A8_R8_G8_B8, + GL_RGBA, /* base_format */ + 4, 4, 4, 4, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 4 /* size in bytes */ + }, + { + PIPE_FORMAT_U_A1_R5_G5_B5, + GL_RGBA, /* base_format */ + 5, 5, 5, 1, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 2 /* size in bytes */ + }, + { + PIPE_FORMAT_U_R5_G6_B5, + GL_RGBA, /* base_format */ + 5, 6, 5, 0, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 2 /* size in bytes */ + }, + /* XXX lots more */ + { + PIPE_FORMAT_S8_Z24, + GL_DEPTH_STENCIL_EXT, /* base_format */ + 0, 0, 0, 0, 0, 0, /* color bits */ + 24, 8, /* depth, stencil */ + 4 /* size in bytes */ + } + }; + GLuint i; + + for (i = 0; i < sizeof(info) / sizeof(info[0]); i++) { + if (info[i].format == format) + return info + i; + } + return NULL; +} + + +/** + * gl_renderbuffer::AllocStorage() + */ +static GLboolean +st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, + GLuint width, GLuint height) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_renderbuffer *strb = st_renderbuffer(rb); + const GLuint pipeFormat + = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE); + const struct pipe_format_info *info = pipe_get_format_info(pipeFormat); + GLuint cpp, pitch; + + if (!info) + return GL_FALSE; + + strb->Base._ActualFormat = info->base_format; + strb->Base.DataType = GL_UNSIGNED_BYTE; /* XXX fix */ + strb->Base.RedBits = info->red_bits; + strb->Base.GreenBits = info->green_bits; + strb->Base.BlueBits = info->blue_bits; + strb->Base.AlphaBits = info->alpha_bits; + strb->Base.DepthBits = info->depth_bits; + strb->Base.StencilBits = info->stencil_bits; + + cpp = info->size; + + if (!strb->surface) { + strb->surface = pipe->surface_alloc(pipe, pipeFormat); + if (!strb->surface) + return GL_FALSE; + } + + /* free old region */ + if (strb->surface->region) { + pipe->region_release(pipe, &strb->surface->region); + } + + /* Choose a pitch to match hardware requirements: + */ + pitch = ((cpp * width + 63) & ~63) / cpp; /* XXX fix: device-specific */ + + strb->surface->region = pipe->region_alloc(pipe, cpp, pitch, height); + if (!strb->surface->region) + return GL_FALSE; /* out of memory, try s/w buffer? */ + + ASSERT(strb->surface->region->buffer); + + strb->Base.Width = strb->surface->width = width; + strb->Base.Height = strb->surface->height = height; + + return GL_TRUE; +} + + +/** + * gl_renderbuffer::Delete() + */ +static void +st_renderbuffer_delete(struct gl_renderbuffer *rb) +{ + GET_CURRENT_CONTEXT(ctx); + struct pipe_context *pipe = ctx->st->pipe; + struct st_renderbuffer *strb = st_renderbuffer(rb); + ASSERT(strb); + if (strb && strb->surface) { + if (rb->surface->region) { + pipe->region_release(pipe, &strb->surface->region); + } + free(strb->surface); + } + free(strb); +} + + +/** + * gl_renderbuffer::GetPointer() + */ +static void * +null_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb, + GLint x, GLint y) +{ + /* By returning NULL we force all software rendering to go through + * the span routines. + */ + assert(0); /* Should never get called with softpipe */ + return NULL; +} + + +/** + * Called via ctx->Driver.NewFramebuffer() + */ +static struct gl_framebuffer * +st_new_framebuffer(GLcontext *ctx, GLuint name) +{ + /* XXX not sure we need to subclass gl_framebuffer for pipe */ + return _mesa_new_framebuffer(ctx, name); +} + + +/** + * Called via ctx->Driver.NewRenderbuffer() + */ +static struct gl_renderbuffer * +st_new_renderbuffer(GLcontext *ctx, GLuint name) +{ + struct st_renderbuffer *strb = CALLOC_STRUCT(st_renderbuffer); + if (strb) { + _mesa_init_renderbuffer(&strb->Base, name); + strb->Base.Delete = st_renderbuffer_delete; + strb->Base.AllocStorage = st_renderbuffer_alloc_storage; + strb->Base.GetPointer = null_get_pointer; + return &strb->Base; + } + return NULL; +} + +/** + * Called via ctx->Driver.BindFramebufferEXT(). + */ +static void +st_bind_framebuffer(GLcontext *ctx, GLenum target, + struct gl_framebuffer *fb, struct gl_framebuffer *fbread) +{ + +} + +/** + * Called by ctx->Driver.FramebufferRenderbuffer + */ +static void +st_framebuffer_renderbuffer(GLcontext *ctx, + struct gl_framebuffer *fb, + GLenum attachment, + struct gl_renderbuffer *rb) +{ + /* XXX no need for derivation? */ + _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb); +} + + +/** + * Called by ctx->Driver.RenderTexture + */ +static void +st_render_texture(GLcontext *ctx, + struct gl_framebuffer *fb, + struct gl_renderbuffer_attachment *att) +{ + struct st_context *st = ctx->st; + struct pipe_framebuffer_state framebuffer; + struct pipe_surface *texsurface; + + texsurface = NULL; /* find the mipmap level, cube face, etc */ + + /* + * XXX basically like this... set the current color (or depth) + * drawing surface to be the given texture renderbuffer. + */ + memset(&framebuffer, 0, sizeof(framebuffer)); + framebuffer.num_cbufs = 1; + framebuffer.cbufs[0] = texsurface; + + if (memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0) { + st->state.framebuffer = framebuffer; + st->pipe->set_framebuffer_state( st->pipe, &framebuffer ); + } +} + + +/** + * Called via ctx->Driver.FinishRenderTexture. + */ +static void +st_finish_render_texture(GLcontext *ctx, + struct gl_renderbuffer_attachment *att) +{ + /* restore drawing to normal framebuffer. may be a no-op */ +} + + + +void st_init_cb_fbo( struct st_context *st ) +{ + struct dd_function_table *functions = &st->ctx->Driver; + + functions->NewFramebuffer = st_new_framebuffer; + functions->NewRenderbuffer = st_new_renderbuffer; + functions->BindFramebuffer = st_bind_framebuffer; + functions->FramebufferRenderbuffer = st_framebuffer_renderbuffer; + functions->RenderTexture = st_render_texture; + functions->FinishRenderTexture = st_finish_render_texture; + /* no longer needed by core Mesa, drivers handle resizes... + functions->ResizeBuffers = st_resize_buffers; + */ +} + + +void st_destroy_cb_fbo( struct st_context *st ) +{ +} diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h new file mode 100644 index 0000000000..f4fa66df59 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_FBO_H +#define ST_CB_FBO_H + + +extern void st_init_cb_fbo( struct st_context *st ); + +extern void st_destroy_cb_fbo( struct st_context *st ); + + +#endif /* ST_CB_FBO_H */ -- cgit v1.2.3 From fdedc4fec035b9dd474a7966be60d2ab85b316e3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 3 Aug 2007 13:29:02 -0600 Subject: more work on st_choose_pipe_format() --- src/mesa/state_tracker/st_cb_teximage.c | 133 +++++++++++++++++++++++++------- src/mesa/state_tracker/st_cb_teximage.h | 5 ++ 2 files changed, 110 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_teximage.c b/src/mesa/state_tracker/st_cb_teximage.c index 5d651d96bc..39c9367695 100644 --- a/src/mesa/state_tracker/st_cb_teximage.c +++ b/src/mesa/state_tracker/st_cb_teximage.c @@ -33,6 +33,7 @@ #include "main/imports.h" #include "main/context.h" +#include "main/texstore.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -40,6 +41,8 @@ #include "st_cb_teximage.h" + + /** * Search list of formats for first RGBA format. */ @@ -84,11 +87,15 @@ default_depth_format(const GLuint formats[], GLuint num) * If we find a device that needs a more intricate selection mechanism, * this function _could_ get pushed down into the pipe device. * + * Note: also used for glRenderbufferStorageEXT() + * + * Note: format and type may be GL_NONE (see renderbuffers) + * * \return PIPE_FORMAT_NONE if error/problem. */ -static GLuint -choose_tex_format(struct pipe_context *pipe, GLint internalFormat, - GLenum format, GLenum type) +GLuint +st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, + GLenum format, GLenum type) { const GLuint *supported; GLboolean allow[PIPE_FORMAT_COUNT]; @@ -112,51 +119,54 @@ choose_tex_format(struct pipe_context *pipe, GLint internalFormat, return PIPE_FORMAT_U_A8_R8_G8_B8; } else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - /* if (allow[PIPE_FORMAT_U_A4_R4_G4_B4]) return PIPE_FORMAT_U_A4_R4_G4_B4; - */ } else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - /* if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) return PIPE_FORMAT_U_A1_R5_G5_B5; - */ } } return default_rgba_format(supported, n); -#if 0 + case 3: case GL_RGB: case GL_COMPRESSED_RGB: if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return &_mesa_texformat_rgb565; + if (allow[PIPE_FORMAT_U_R5_G6_B5]) + return PIPE_FORMAT_U_R5_G6_B5; } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + return default_rgba_format(supported, n); case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return default_rgba_format(supported, n); case GL_RGBA4: case GL_RGBA2: - return &_mesa_texformat_argb4444; + if (allow[PIPE_FORMAT_U_A4_R4_G4_B4]) + return PIPE_FORMAT_U_A4_R4_G4_B4; + return default_rgba_format(supported, n); case GL_RGB5_A1: - return &_mesa_texformat_argb1555; + if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) + return PIPE_FORMAT_U_A1_R5_G5_B5; + return default_rgba_format(supported, n); case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return &_mesa_texformat_argb8888; + return default_rgba_format(supported, n); case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; + if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) + return PIPE_FORMAT_U_A1_R5_G5_B5; + return default_rgba_format(supported, n); case GL_ALPHA: case GL_ALPHA4: @@ -164,7 +174,9 @@ choose_tex_format(struct pipe_context *pipe, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; + if (allow[PIPE_FORMAT_U_A8]) + return PIPE_FORMAT_U_A8; + return default_rgba_format(supported, n); case 1: case GL_LUMINANCE: @@ -173,7 +185,9 @@ choose_tex_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; + if (allow[PIPE_FORMAT_U_A8]) + return PIPE_FORMAT_U_A8; + return default_rgba_format(supported, n); case 2: case GL_LUMINANCE_ALPHA: @@ -184,7 +198,9 @@ choose_tex_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; + if (allow[PIPE_FORMAT_U_L8_A8]) + return PIPE_FORMAT_U_L8_A8; + return default_rgba_format(supported, n); case GL_INTENSITY: case GL_INTENSITY4: @@ -192,14 +208,22 @@ choose_tex_format(struct pipe_context *pipe, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; + if (allow[PIPE_FORMAT_U_I8]) + return PIPE_FORMAT_U_I8; + return default_rgba_format(supported, n); case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; + if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) { + if (allow[PIPE_FORMAT_YCBCR]) + return PIPE_FORMAT_YCBCR; + } + else { + if (allow[PIPE_FORMAT_YCBCR_REV]) + return PIPE_FORMAT_YCBCR_REV; + } + return PIPE_FORMAT_NONE; +#if 0 case GL_COMPRESSED_RGB_FXT1_3DFX: return &_mesa_texformat_rgb_fxt1; case GL_COMPRESSED_RGBA_FXT1_3DFX: @@ -237,12 +261,23 @@ choose_tex_format(struct pipe_context *pipe, GLint internalFormat, case GL_DEPTH_COMPONENT: return default_depth_format(supported, n); + case GL_STENCIL_INDEX: + case GL_STENCIL_INDEX1_EXT: + case GL_STENCIL_INDEX4_EXT: + case GL_STENCIL_INDEX8_EXT: + case GL_STENCIL_INDEX16_EXT: + if (allow[PIPE_FORMAT_U_S8]) + return PIPE_FORMAT_U_S8; + if (allow[PIPE_FORMAT_S8_Z24]) + return PIPE_FORMAT_S8_Z24; + return PIPE_FORMAT_NONE; -#if 0 case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_z24_s8; -#endif + if (allow[PIPE_FORMAT_S8_Z24]) + return PIPE_FORMAT_S8_Z24; + return PIPE_FORMAT_NONE; + default: return PIPE_FORMAT_NONE; } @@ -250,13 +285,55 @@ choose_tex_format(struct pipe_context *pipe, GLint internalFormat, +static void +st_teximage2d(GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + /* probably nothing here: use core Mesa TexImage2D fallback to + * save teximage in main memory. + * Later, when we have a complete texobj and are ready to render, + * create the pipe texture object / mipmap-tree. + */ + + _mesa_store_teximage2d(ctx, target, level, internalFormat, + width, height, border, format, type, pixels, + packing, texObj, texImage); +} + + +static void +st_texsubimage2d(GLcontext *ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ +#if 0 + struct pipe_mipmap_tree *mt = texObj->DriverData; +#endif + /* 1. find the region which stores this texture object. + * 2. convert texels to pipe format if needed. + * 3. replace texdata in the texture region. + */ + +} + + + void st_init_cb_teximage( struct st_context *st ) { - /* struct dd_function_table *functions = &st->ctx->Driver; functions->TexImage2D = st_teximage2d; - */ + functions->TexSubImage2D = st_texsubimage2d; } diff --git a/src/mesa/state_tracker/st_cb_teximage.h b/src/mesa/state_tracker/st_cb_teximage.h index 6e62a2e3c9..462128a844 100644 --- a/src/mesa/state_tracker/st_cb_teximage.h +++ b/src/mesa/state_tracker/st_cb_teximage.h @@ -30,6 +30,11 @@ #define ST_CB_TEXIMAGE_H +extern GLuint +st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, + GLenum format, GLenum type); + + extern void st_init_cb_teximage( struct st_context *st ); extern void st_destroy_cb_teximage( struct st_context *st ); -- cgit v1.2.3 From 16682fee0a582ace8aa101704280d6839e9f0f23 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 3 Aug 2007 13:30:55 -0600 Subject: added st_cb_fbo.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 8074a43c85..4e3e44f15a 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -193,6 +193,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_viewport.c \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_drawpixels.c \ + state_tracker/st_cb_fbo.c \ state_tracker/st_cb_program.c \ state_tracker/st_cb_teximage.c \ state_tracker/st_draw.c \ -- cgit v1.2.3 From 5e7921b77874ce0142f0ab032c710a379d86ff09 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 3 Aug 2007 13:31:56 -0600 Subject: remove intel_pixel.c --- src/mesa/drivers/dri/i915pipe/Makefile | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile index 2331e81b1c..617b2c7bce 100644 --- a/src/mesa/drivers/dri/i915pipe/Makefile +++ b/src/mesa/drivers/dri/i915pipe/Makefile @@ -21,7 +21,6 @@ DRIVER_SOURCES = \ intel_tex_validate.c \ intel_tex_format.c \ intel_tex.c \ - intel_pixel.c \ intel_buffers.c \ intel_blit.c \ intel_context.c \ -- cgit v1.2.3 From 1d1b9e6be45e75ad12a01c82e3c0d55ff9da4183 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 5 Aug 2007 17:23:38 +0100 Subject: Add a new interface between softpipe and the window system / buffer manager. This interface is defined by softpipe and any window system (eg i915pipe) wishing to use softpipe is required to implement the interface. Currently the interface is all about buffer management. Generalizing, each pipe driver will advertise an interface in a similar spirit to this, and again any window system driver wishing to use that rendering pipeline will have to implement the interface it defines. It clearly isn't a one-way street however, as softpipe could just do its own buffer management with malloc. The interaction with a buffer manager is desired to allow us to exercise the hardware swapbuffers functionality of the i915pipe driver, and also to get a feel for the way hardware drivers which really need a buffer manager will work. --- src/mesa/drivers/dri/i915pipe/Makefile | 2 +- src/mesa/drivers/dri/i915pipe/intel_context.c | 10 +- src/mesa/drivers/dri/i915pipe/intel_context.h | 7 + src/mesa/drivers/dri/i915pipe/intel_regions.c | 298 ------------------------- src/mesa/drivers/dri/i915pipe/intel_regions.h | 36 --- src/mesa/drivers/dri/i915pipe/intel_softpipe.c | 182 +++++++++++++++ src/mesa/pipe/p_context.h | 12 +- src/mesa/pipe/p_state.h | 5 +- src/mesa/pipe/softpipe/sp_clear.c | 2 +- src/mesa/pipe/softpipe/sp_context.c | 5 +- src/mesa/pipe/softpipe/sp_context.h | 4 +- src/mesa/pipe/softpipe/sp_region.c | 287 +++++++++++++++--------- src/mesa/pipe/softpipe/sp_winsys.h | 100 +++++++++ 13 files changed, 493 insertions(+), 457 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_regions.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_regions.h create mode 100644 src/mesa/drivers/dri/i915pipe/intel_softpipe.c create mode 100644 src/mesa/pipe/softpipe/sp_winsys.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile index 617b2c7bce..4e8d020196 100644 --- a/src/mesa/drivers/dri/i915pipe/Makefile +++ b/src/mesa/drivers/dri/i915pipe/Makefile @@ -10,7 +10,7 @@ PIPE_DRIVERS = \ $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a DRIVER_SOURCES = \ - intel_regions.c \ + intel_softpipe.c \ intel_buffer_objects.c \ intel_batchbuffer.c \ intel_mipmap_tree.c \ diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index ac04d4e8c2..d43b784112 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -54,11 +54,9 @@ #include "intel_buffer_objects.h" #include "intel_fbo.h" -#include "pipe/softpipe/sp_context.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" - #include "drirenderbuffer.h" #include "vblank.h" #include "utils.h" @@ -366,7 +364,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, * Pipe-related setup */ st_create_context( &intel->ctx, - softpipe_create() ); + intel_create_softpipe( intel ) ); /* KW: Not sure I like this - we should only be talking to the * state_tracker. The pipe code will need some way of talking to @@ -375,9 +373,9 @@ intelCreateContext(const __GLcontextModes * mesaVis, * BP: Yes, a temporary hack so we can make jumps between old/new code. */ intel->pipe = intel->ctx.st->pipe; - intel->pipe->screen = intelScreen; - intel->pipe->glctx = ctx; - intel_init_region_functions(intel->pipe); +// intel->pipe->screen = intelScreen; +// intel->pipe->glctx = ctx; +// intel_init_region_functions(intel->pipe); /* * memory pools diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.h b/src/mesa/drivers/dri/i915pipe/intel_context.h index 6c728dfe73..ad55e7eef2 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.h +++ b/src/mesa/drivers/dri/i915pipe/intel_context.h @@ -196,6 +196,13 @@ extern void intelFlush(GLcontext * ctx); extern void intelInitDriverFunctions(struct dd_function_table *functions); +/* ================================================================ + * intel_softpipe.c: + */ + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel ); + #define MI_BATCH_BUFFER_END (0xA<<23) diff --git a/src/mesa/drivers/dri/i915pipe/intel_regions.c b/src/mesa/drivers/dri/i915pipe/intel_regions.c deleted file mode 100644 index 75ba1f9cdb..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_regions.c +++ /dev/null @@ -1,298 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Provide additional functionality on top of bufmgr buffers: - * - 2d semantics and blit operations - * - refcounting of buffers for multiple images in a buffer. - * - refcounting of buffer mappings. - * - some logic for moving the buffers to the best memory pools for - * given operations. - * - * Most of this is to make it easier to implement the fixed-layout - * mipmap tree required by intel hardware in the face of GL's - * programming interface where each image can be specifed in random - * order and it isn't clear what layout the tree should have until the - * last moment. - */ - -#include "pipe/p_state.h" -#include "pipe/p_context.h" - -#include "intel_context.h" -#include "intel_blit.h" -#include "intel_buffer_objects.h" -#include "dri_bufmgr.h" -#include "intel_batchbuffer.h" - - -#define FILE_DEBUG_FLAG DEBUG_REGION - - -/** XXX temporary helper */ -static intelScreenPrivate * -pipe_screen(struct pipe_context *pipe) -{ - return (intelScreenPrivate *) pipe->screen; -} - - -static void -intel_region_idle(struct pipe_context *pipe, struct pipe_region *region) -{ - DBG("%s\n", __FUNCTION__); - if (region && region->buffer) - driBOWaitIdle(region->buffer, GL_FALSE); -} - -/* XXX: Thread safety? - */ -static GLubyte * -intel_region_map(struct pipe_context *pipe, struct pipe_region *region) -{ - DBG("%s\n", __FUNCTION__); - if (!region->map_refcount++) { - region->map = driBOMap(region->buffer, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - } - - return region->map; -} - -static void -intel_region_unmap(struct pipe_context *pipe, struct pipe_region *region) -{ - DBG("%s\n", __FUNCTION__); - if (!--region->map_refcount) { - driBOUnmap(region->buffer); - region->map = NULL; - } -} - -static struct pipe_region * -intel_region_alloc(struct pipe_context *pipe, - GLuint cpp, GLuint pitch, GLuint height) -{ - intelScreenPrivate *intelScreen = pipe_screen(pipe); - struct pipe_region *region = calloc(sizeof(*region), 1); - struct intel_context *intel = intelScreenContext(intelScreen); - - DBG("%s\n", __FUNCTION__); - - region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ - region->refcount = 1; - - driGenBuffers(intelScreen->regionPool, - "region", 1, ®ion->buffer, 64, - 0, - 0); - - LOCK_HARDWARE(intel); - driBOData(region->buffer, pitch * cpp * height, NULL, 0); - UNLOCK_HARDWARE(intel); - return region; -} - -static void -intel_region_release(struct pipe_context *pipe, struct pipe_region **region) -{ - if (!*region) - return; - - DBG("%s %d\n", __FUNCTION__, (*region)->refcount - 1); - - ASSERT((*region)->refcount > 0); - (*region)->refcount--; - - if ((*region)->refcount == 0) { - assert((*region)->map_refcount == 0); - - driBOUnReference((*region)->buffer); - free(*region); - } - *region = NULL; -} - - - - - - - -/* - * XXX Move this into core Mesa? - */ -static void -_mesa_copy_rect(GLubyte * dst, - GLuint cpp, - GLuint dst_pitch, - GLuint dst_x, - GLuint dst_y, - GLuint width, - GLuint height, - const GLubyte * src, - GLuint src_pitch, GLuint src_x, GLuint src_y) -{ - GLuint i; - - dst_pitch *= cpp; - src_pitch *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * dst_pitch; - width *= cpp; - - if (width == dst_pitch && width == src_pitch) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; - } - } -} - - -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -static void -intel_region_data(struct pipe_context *pipe, - struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - const void *src, GLuint src_pitch, - GLuint srcx, GLuint srcy, GLuint width, GLuint height) -{ - intelScreenPrivate *intelScreen = pipe_screen(pipe); - struct intel_context *intel = intelScreenContext(intelScreen); - - DBG("%s\n", __FUNCTION__); - - if (intel == NULL) - return; - - LOCK_HARDWARE(intel); - - _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - pipe->region_unmap(pipe, dst); - - UNLOCK_HARDWARE(intel); - -} - -/* Copy rectangular sub-regions. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static void -intel_region_copy(struct pipe_context *pipe, - struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - const struct pipe_region *src, - GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height) -{ - intelScreenPrivate *intelScreen = pipe_screen(pipe); - struct intel_context *intel = intelScreenContext(intelScreen); - - DBG("%s\n", __FUNCTION__); - - if (intel == NULL) - return; - - assert(src->cpp == dst->cpp); - - intelEmitCopyBlit(intel, - dst->cpp, - src->pitch, src->buffer, src_offset, - dst->pitch, dst->buffer, dst_offset, - srcx, srcy, dstx, dsty, width, height, - GL_COPY); -} - -/* Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static void -intel_region_fill(struct pipe_context *pipe, - struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - GLuint width, GLuint height, - GLuint value, GLuint mask) -{ - intelScreenPrivate *intelScreen = pipe_screen(pipe); - struct intel_context *intel = intelScreenContext(intelScreen); - - DBG("%s\n", __FUNCTION__); - - if (intel == NULL) - return; - - intelEmitFillBlit(intel, - dst->cpp, - dst->pitch, dst->buffer, dst_offset, - dstx, dsty, width, height, value, mask); -} - - - -static struct _DriBufferObject * -intel_region_buffer(struct pipe_context *pipe, - struct pipe_region *region, GLuint flag) -{ - return region->buffer; -} - - - -void -intel_init_region_functions(struct pipe_context *pipe) -{ - pipe->region_idle = intel_region_idle; - pipe->region_map = intel_region_map; - pipe->region_unmap = intel_region_unmap; - pipe->region_alloc = intel_region_alloc; - pipe->region_release = intel_region_release; - pipe->region_data = intel_region_data; - pipe->region_copy = intel_region_copy; - pipe->region_fill = intel_region_fill; - pipe->region_buffer = intel_region_buffer; -} - diff --git a/src/mesa/drivers/dri/i915pipe/intel_regions.h b/src/mesa/drivers/dri/i915pipe/intel_regions.h deleted file mode 100644 index 2a3b5b4025..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_regions.h +++ /dev/null @@ -1,36 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_REGIONS_H -#define INTEL_REGIONS_H - -#include "mtypes.h" -#include "intel_screen.h" - - - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_softpipe.c b/src/mesa/drivers/dri/i915pipe/intel_softpipe.c new file mode 100644 index 0000000000..4591730a02 --- /dev/null +++ b/src/mesa/drivers/dri/i915pipe/intel_softpipe.c @@ -0,0 +1,182 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +#include "dri_bufpool.h" +#include "dri_bufmgr.h" + +/* #include "errno.h" */ +/* #include "string.h" */ +/* #include "imports.h" */ + +#include "intel_context.h" + +#include "pipe/softpipe/sp_winsys.h" + + +struct intel_softpipe_winsys { + struct softpipe_winsys sws; + struct intel_context *intel; +}; + + + +/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static inline struct _DriBufferObject * +dri_bo( struct softpipe_buffer_handle *bo ) +{ + return (struct _DriBufferObject *)bo; +} + +static inline struct softpipe_buffer_handle * +pipe_bo( struct _DriBufferObject *bo ) +{ + return (struct softpipe_buffer_handle *)bo; +} + +/* Turn a softpipe winsys into an intel/softpipe winsys: + */ +static inline struct intel_softpipe_winsys * +intel_softpipe_winsys( struct softpipe_winsys *sws ) +{ + return (struct intel_softpipe_winsys *)sws; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void *intel_buffer_map(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf ) +{ + return driBOMap( dri_bo(buf), + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0 ); +} + +static void intel_buffer_unmap(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf) +{ + driBOUnmap( dri_bo(buf) ); +} + + +static struct softpipe_buffer_handle * +intel_buffer_reference(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf) +{ + return pipe_bo( driBOReference( dri_bo(buf) ) ); +} + +static void intel_buffer_unreference(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf) +{ + driBOUnReference( dri_bo(buf) ); +} + +/* Grabs the hardware lock! + */ +static void intel_buffer_data(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct intel_context *intel = intel_softpipe_winsys(sws)->intel; + + LOCK_HARDWARE( intel ); + driBOData( dri_bo(buf), size, data, 0 ); + UNLOCK_HARDWARE( intel ); +} + +static void intel_buffer_subdata(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) +{ + driBOSubData( dri_bo(buf), offset, size, data ); +} + +static void intel_buffer_get_subdata(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) +{ + driBOGetSubData( dri_bo(buf), offset, size, data ); +} + +/* Softpipe has no concept of pools. We choose the tex/region pool + * for all buffers. + */ +static struct softpipe_buffer_handle * +intel_create_buffer(struct softpipe_winsys *sws, + const char *name, + unsigned alignment) +{ + struct intel_context *intel = intel_softpipe_winsys(sws)->intel; + struct _DriBufferObject *buffer; + + LOCK_HARDWARE( intel ); + driGenBuffers( intel->intelScreen->regionPool, + name, 1, &buffer, alignment, 0, 0 ); + UNLOCK_HARDWARE( intel ); + + return pipe_bo(buffer); +} + + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel ) +{ + struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); + + /* Fill in this struct with callbacks that softpipe will need to + * communicate with the window system, buffer manager, etc. + * + * Softpipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + isws->sws.create_buffer = intel_create_buffer; + isws->sws.buffer_map = intel_buffer_map; + isws->sws.buffer_unmap = intel_buffer_unmap; + isws->sws.buffer_reference = intel_buffer_reference; + isws->sws.buffer_unreference = intel_buffer_unreference; + isws->sws.buffer_data = intel_buffer_data; + isws->sws.buffer_subdata = intel_buffer_subdata; + isws->sws.buffer_get_subdata = intel_buffer_get_subdata; + isws->intel = intel; + + /* Create the softpipe context: + */ + return softpipe_create( &isws->sws ); +} diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index ac6567f009..92670c7733 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -32,10 +32,6 @@ #include "p_state.h" -/* Kludge: - */ -extern struct pipe_context *softpipe_create( void ); - /* Drawing currently kludged up via the existing tnl/ module. */ struct vertex_buffer; @@ -161,7 +157,8 @@ struct pipe_context { struct pipe_region *dest, GLuint dest_offset, GLuint destx, GLuint desty, - const struct pipe_region *src, + struct pipe_region *src, /* don't make this const - + need to map/unmap */ GLuint src_offset, GLuint srcx, GLuint srcy, GLuint width, GLuint height); @@ -170,14 +167,11 @@ struct pipe_context { GLuint dst_offset, GLuint dstx, GLuint dsty, GLuint width, GLuint height, - GLuint value, GLuint mask); + GLuint value); struct _DriBufferObject *(*region_buffer)(struct pipe_context *pipe, struct pipe_region *region, GLuint flag); - - void *screen; /**< temporary */ - void *glctx; /**< temporary */ }; diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 699efd5877..3be1010007 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -235,12 +235,9 @@ struct pipe_sampler_state *** Resource Objects ***/ - -struct _DriBufferObject; - struct pipe_region { - struct _DriBufferObject *buffer; /**< buffer manager's buffer ID */ + void *buffer; /**< driver private buffer handle */ GLuint refcount; /**< Reference count for region */ GLuint cpp; /**< bytes per pixel */ diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index d7684d2044..2d69b6e0dc 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -56,5 +56,5 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, w = softpipe->framebuffer.cbufs[0]->width; h = softpipe->framebuffer.cbufs[0]->height; - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue, ~0); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue); } diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index fe7f2f351a..22928f7bd6 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -39,6 +39,7 @@ #include "sp_state.h" #include "sp_surface.h" #include "sp_prim_setup.h" +#include "sp_winsys.h" @@ -168,7 +169,7 @@ static GLuint softpipe_get_occlusion_counter(struct pipe_context *pipe) } -struct pipe_context *softpipe_create( void ) +struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); @@ -209,6 +210,8 @@ struct pipe_context *softpipe_create( void ) softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); + softpipe->winsys = sws; + /* * Create drawing context and plug our rendering stage into it. */ diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 3c379c9091..43fa8acd76 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -38,8 +38,8 @@ #include "sp_quad.h" - struct softpipe_surface; +struct softpipe_winsys; struct draw_context; struct draw_stage; @@ -68,6 +68,8 @@ enum interp_mode { struct softpipe_context { struct pipe_context pipe; /**< base class */ + struct softpipe_winsys *winsys; /**< window system interface */ + /* The most recent drawing state as set by the driver: */ diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 1dbd1609e3..78a0919ad5 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -25,74 +25,185 @@ * **************************************************************************/ -/** - * Software-based region allocation and management. - * A hardware driver would override these functions. +/* Provide additional functionality on top of bufmgr buffers: + * - 2d semantics and blit operations (XXX: remove/simplify blits??) + * - refcounting of buffers for multiple images in a buffer. + * - refcounting of buffer mappings. */ - #include "sp_context.h" +#include "sp_winsys.h" #include "sp_region.h" -#include "sp_surface.h" -#include "main/imports.h" +static void +sp_region_idle(struct pipe_context *pipe, struct pipe_region *region) +{ + +} + + +static GLubyte * +sp_region_map(struct pipe_context *pipe, struct pipe_region *region) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + + if (!region->map_refcount++) { + region->map = sp->winsys->buffer_map( sp->winsys, + region->buffer ); + } + + return region->map; +} + +static void +sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + + if (!--region->map_refcount) { + sp->winsys->buffer_unmap( sp->winsys, + region->buffer ); + region->map = NULL; + } +} + static struct pipe_region * sp_region_alloc(struct pipe_context *pipe, - GLuint cpp, GLuint pitch, GLuint height) + GLuint cpp, GLuint pitch, GLuint height) { - struct pipe_region *region = CALLOC_STRUCT(pipe_region); - if (!region) - return NULL; + struct softpipe_context *sp = softpipe_context( pipe ); + struct pipe_region *region = calloc(sizeof(*region), 1); - region->refcount = 1; region->cpp = cpp; region->pitch = pitch; - region->height = height; - region->map = malloc(cpp * pitch * height); + region->height = height; /* needed? */ + region->refcount = 1; + + region->buffer = sp->winsys->create_buffer(sp->winsys, "region", 64 ); + + sp->winsys->buffer_data( sp->winsys, + region->buffer, + pitch * cpp * height, + NULL ); return region; } - static void sp_region_release(struct pipe_context *pipe, struct pipe_region **region) { - assert((*region)->refcount > 0); + struct softpipe_context *sp = softpipe_context( pipe ); + + if (!*region) + return; + + ASSERT((*region)->refcount > 0); (*region)->refcount--; if ((*region)->refcount == 0) { assert((*region)->map_refcount == 0); -#if 0 - if ((*region)->pbo) - (*region)->pbo->region = NULL; - (*region)->pbo = NULL; -#endif - + sp->winsys->buffer_unreference( sp->winsys, + (*region)->buffer ); free(*region); } *region = NULL; } - -static GLubyte * -sp_region_map(struct pipe_context *pipe, struct pipe_region *region) +/* + * XXX Move this into core Mesa? + */ +static void +_mesa_copy_rect(GLubyte * dst, + GLuint cpp, + GLuint dst_pitch, + GLuint dst_x, + GLuint dst_y, + GLuint width, + GLuint height, + const GLubyte * src, + GLuint src_pitch, + GLuint src_x, + GLuint src_y) { - region->map_refcount++; - return region->map; + GLuint i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * dst_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } } +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ static void -sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) +sp_region_data(struct pipe_context *pipe, + struct pipe_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + const void *src, GLuint src_pitch, + GLuint srcx, GLuint srcy, GLuint width, GLuint height) { - region->map_refcount--; + _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + pipe->region_unmap(pipe, dst); } +/* Assumes all values are within bounds -- no checking at this level - + * do it higher up if required. + */ +static void +sp_region_copy(struct pipe_context *pipe, + struct pipe_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + struct pipe_region *src, + GLuint src_offset, + GLuint srcx, GLuint srcy, GLuint width, GLuint height) +{ + assert( dst != src ); + assert( dst->cpp == src->cpp ); + _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe->region_map(pipe, src) + src_offset, + src->pitch, + srcx, srcy); + pipe->region_unmap(pipe, src); + pipe->region_unmap(pipe, dst); +} + +/* Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ static GLubyte * get_pointer(struct pipe_region *dst, GLuint x, GLuint y) { @@ -105,93 +216,69 @@ sp_region_fill(struct pipe_context *pipe, struct pipe_region *dst, GLuint dst_offset, GLuint dstx, GLuint dsty, - GLuint width, GLuint height, GLuint value, GLuint mask) + GLuint width, GLuint height, GLuint value) { GLuint i, j; + + (void)pipe->region_map(pipe, dst); + switch (dst->cpp) { - case 1: - { - GLubyte *row = get_pointer(dst, dstx, dsty); - if ((mask & 0xff) == 0xff) { - /* no masking */ - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - else { - value &= mask; - mask = ~mask; - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) { - row[j] = (row[j] & mask) | value; - } - row += dst->pitch; - } - } + case 1: { + GLubyte *row = get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; } - break; - case 2: - { - GLushort *row = (GLushort *) get_pointer(dst, dstx, dsty); - if ((mask & 0xffff) == 0xffff) { - /* no masking */ - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - else { - value &= mask; - mask = ~mask; - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (row[j] & mask) | value; - row += dst->pitch; - } - } + } + break; + case 2: { + GLushort *row = (GLushort *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; } - break; - case 4: - { - GLuint *row = (GLuint *) get_pointer(dst, dstx, dsty); - if (mask == 0xffffffff) { - /* no masking */ - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - else { - value &= mask; - mask = ~mask; - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (row[j] & mask) | value; - row += dst->pitch; - } - } + } + break; + case 4: { + GLuint *row = (GLuint *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; } - break; + } + break; default: assert(0); - + break; } + + pipe->region_unmap( pipe, dst ); +} + + + +static struct _DriBufferObject * +sp_region_buffer(struct pipe_context *pipe, + struct pipe_region *region, GLuint flag) +{ + return region->buffer; } + void sp_init_region_functions(struct softpipe_context *sp) { - sp->pipe.region_alloc = sp_region_alloc; - sp->pipe.region_release = sp_region_release; - + sp->pipe.region_idle = sp_region_idle; sp->pipe.region_map = sp_region_map; sp->pipe.region_unmap = sp_region_unmap; - + sp->pipe.region_alloc = sp_region_alloc; + sp->pipe.region_release = sp_region_release; + sp->pipe.region_data = sp_region_data; + sp->pipe.region_copy = sp_region_copy; sp->pipe.region_fill = sp_region_fill; - - /* XXX lots of other region functions needed... */ + sp->pipe.region_buffer = sp_region_buffer; } + diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h new file mode 100644 index 0000000000..43953c648b --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_winsys.h @@ -0,0 +1,100 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_WINSYS_H +#define SP_WINSYS_H + +#include "main/mtypes.h" + +/* This is the interface that softpipe requires any window system + * hosting it to implement. This is the only include file in softpipe + * which is public. + */ + + +/* Pipe drivers are (meant to be!) independent of both GL and the + * window system. The window system provides a buffer manager and a + * set of additional hooks for things like command buffer submission, + * etc. + * + * There clearly has to be some agreement between the window system + * driver and the hardware driver about the format of command buffers, + * etc. + */ + +struct softpipe_buffer_handle; + +struct softpipe_winsys { + + + /* The buffer manager is modeled after the dri_bugmgr interface, + * but this is the subset that softpipe cares about. Remember that + * softpipe gets to choose the interface it needs, and the window + * systems must then implement that interface (rather than the + * other way around...). + * + * Softpipe only really wants to make system memory allocations, + * right?? + */ + struct softpipe_buffer_handle *(*create_buffer)(struct softpipe_winsys *sws, + const char *name, + unsigned alignment ); + + void *(*buffer_map)( struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf ); + + void (*buffer_unmap)( struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf ); + + struct softpipe_buffer_handle *(*buffer_reference)( struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf ); + + void (*buffer_unreference)( struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf ); + + void (*buffer_data)(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf, + unsigned size, const void *data ); + + void (*buffer_subdata)(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data); + + void (*buffer_get_subdata)(struct softpipe_winsys *sws, + struct softpipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data); +}; + + +struct pipe_context *softpipe_create( struct softpipe_winsys * ); + + +#endif /* SP_WINSYS_H */ -- cgit v1.2.3 From 95794abec4bdc5cda9f2e7d139a70c3acf372fe3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 6 Aug 2007 10:15:30 +0100 Subject: Add pipe buffer managment functions. The state_tracker driver needs these to implement, eg. pixel buffer objects, vertex buffer objects. --- src/mesa/drivers/dri/i915pipe/intel_blit.c | 157 ------------------ src/mesa/drivers/dri/i915pipe/intel_fbo.c | 1 - src/mesa/drivers/dri/i915pipe/intel_softpipe.c | 32 ++-- src/mesa/drivers/dri/i915pipe/intel_surface.c | 1 - src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 6 +- src/mesa/pipe/Makefile.template | 1 + src/mesa/pipe/p_context.h | 41 ++++- src/mesa/pipe/p_state.h | 13 ++ src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_buffer.c | 119 ++++++++++++++ src/mesa/pipe/softpipe/sp_buffer.h | 40 +++++ src/mesa/pipe/softpipe/sp_context.c | 2 + src/mesa/pipe/softpipe/sp_region.c | 10 +- src/mesa/pipe/softpipe/sp_winsys.h | 23 ++- src/mesa/sources | 1 + src/mesa/state_tracker/st_cb_bufferobjects.c | 206 ++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_bufferobjects.h | 66 ++++++++ 17 files changed, 520 insertions(+), 200 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_buffer.c create mode 100644 src/mesa/pipe/softpipe/sp_buffer.h create mode 100644 src/mesa/state_tracker/st_cb_bufferobjects.c create mode 100644 src/mesa/state_tracker/st_cb_bufferobjects.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.c b/src/mesa/drivers/dri/i915pipe/intel_blit.c index 5d9cb0cf0f..8e878f0088 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_blit.c +++ b/src/mesa/drivers/dri/i915pipe/intel_blit.c @@ -391,161 +391,4 @@ intelEmitCopyBlit(struct intel_context *intel, } -/** - * Use blitting to clear the renderbuffers named by 'flags'. - * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferMask field - * since that might include software renderbuffers or renderbuffers - * which we're clearing with triangles. - * \param mask bitmask of BUFFER_BIT_* values indicating buffers to clear - */ -void -intelClearWithBlit(GLcontext * ctx, GLbitfield mask) -{ - struct intel_context *intel = intel_context(ctx); - struct gl_framebuffer *fb = ctx->DrawBuffer; - GLuint clear_depth; - GLbitfield skipBuffers = 0; - BATCH_LOCALS; - - DBG("%s %x\n", __FUNCTION__, mask); - /* - * Compute values for clearing the buffers. - */ - clear_depth = 0; - if (mask & BUFFER_BIT_DEPTH) { - clear_depth = (GLuint) (fb->_DepthMax * ctx->Depth.Clear); - } - if (mask & BUFFER_BIT_STENCIL) { - clear_depth |= (ctx->Stencil.Clear & 0xff) << 24; - } - - /* If clearing both depth and stencil, skip BUFFER_BIT_STENCIL in - * the loop below. - */ - if ((mask & BUFFER_BIT_DEPTH) && (mask & BUFFER_BIT_STENCIL)) { - skipBuffers = BUFFER_BIT_STENCIL; - } - - /* XXX Move this flush/lock into the following conditional? */ - intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); - - if (intel->numClipRects) { - GLint cx, cy, cw, ch; - drm_clip_rect_t b; - - /* Get clear bounds after locking */ - cx = fb->_Xmin; - cy = fb->_Ymin; - cw = fb->_Xmax - cx; - ch = fb->_Ymax - cy; - - if (fb->Name == 0) { - /* clearing a window */ - /* flip top to bottom */ - b.x1 = cx; - b.y1 = fb->Height - cy - ch; - b.x2 = b.x1 + cw; - b.y2 = b.y1 + ch; - } - else { - /* clearing FBO */ - b.x1 = cx; - b.y1 = cy; - b.x2 = b.x1 + cw; - b.y2 = b.y1 + ch; - /* no change to mask */ - } - - { - GLuint buf; - GLuint clearMask = mask; /* use copy, since we modify it below */ - GLboolean all = (cw == fb->Width && ch == fb->Height); - - DBG("clear %d,%d..%d,%d, mask %x\n", - b.x1, b.y1, b.x2, b.y2, mask); - - /* Loop over all renderbuffers */ - for (buf = 0; buf < BUFFER_COUNT && clearMask; buf++) { - const GLbitfield bufBit = 1 << buf; - if ((clearMask & bufBit) && !(bufBit & skipBuffers)) { - /* OK, clear this renderbuffer */ - struct pipe_region *irb_region = - intel_get_rb_region(fb, buf); - struct _DriBufferObject *write_buffer = - intel->pipe->region_buffer(intel->pipe, irb_region, - all ? INTEL_WRITE_FULL : - INTEL_WRITE_PART); - - GLuint clearVal; - GLint pitch, cpp; - GLuint BR13, CMD; - - ASSERT(irb_region); - - pitch = irb_region->pitch; - cpp = irb_region->cpp; - - DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, - irb_region->buffer, (pitch * cpp), - irb_region->draw_offset, - b.x1, b.y1, b.x2 - b.x1, b.y2 - b.y1); - - - /* Setup the blit command */ - if (cpp == 4) { - BR13 = (0xF0 << 16) | (pitch * cpp) | (1 << 24) | (1 << 25); - if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) { - CMD = XY_COLOR_BLT_CMD; - if (clearMask & BUFFER_BIT_DEPTH) - CMD |= XY_COLOR_BLT_WRITE_RGB; - if (clearMask & BUFFER_BIT_STENCIL) - CMD |= XY_COLOR_BLT_WRITE_ALPHA; - } - else { - /* clearing RGBA */ - CMD = (XY_COLOR_BLT_CMD | - XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); - } - } - else { - ASSERT(cpp == 2 || cpp == 0); - BR13 = (0xF0 << 16) | (pitch * cpp) | (1 << 24); - CMD = XY_COLOR_BLT_CMD; - } - - if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) { - clearVal = clear_depth; - } - else { - clearVal = (cpp == 4) - ? intel->ClearColor8888 : intel->ClearColor565; - } - /* - _mesa_debug(ctx, "hardware blit clear buf %d rb id %d\n", - buf, irb->Base.Name); - */ - intel_wait_flips(intel, INTEL_BATCH_NO_CLIPRECTS); - - BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((b.y1 << 16) | b.x1); - OUT_BATCH((b.y2 << 16) | b.x2); - OUT_RELOC(write_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - irb_region->draw_offset); - OUT_BATCH(clearVal); - ADVANCE_BATCH(); - clearMask &= ~bufBit; /* turn off bit, for faster loop exit */ - } - } - } - intel_batchbuffer_flush(intel->batch); - } - - UNLOCK_HARDWARE(intel); -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index 9eee5cce45..267c12fd4b 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -40,7 +40,6 @@ #include "intel_depthstencil.h" #include "intel_fbo.h" #include "intel_mipmap_tree.h" -#include "intel_regions.h" #include "intel_tex.h" #include "pipe/p_context.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_softpipe.c b/src/mesa/drivers/dri/i915pipe/intel_softpipe.c index 4591730a02..ef47744358 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_softpipe.c +++ b/src/mesa/drivers/dri/i915pipe/intel_softpipe.c @@ -54,15 +54,15 @@ struct intel_softpipe_winsys { * buffer pointer... */ static inline struct _DriBufferObject * -dri_bo( struct softpipe_buffer_handle *bo ) +dri_bo( struct pipe_buffer_handle *bo ) { return (struct _DriBufferObject *)bo; } -static inline struct softpipe_buffer_handle * +static inline struct pipe_buffer_handle * pipe_bo( struct _DriBufferObject *bo ) { - return (struct softpipe_buffer_handle *)bo; + return (struct pipe_buffer_handle *)bo; } /* Turn a softpipe winsys into an intel/softpipe winsys: @@ -77,36 +77,39 @@ intel_softpipe_winsys( struct softpipe_winsys *sws ) /* Most callbacks map direcly onto dri_bufmgr operations: */ static void *intel_buffer_map(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf ) + struct pipe_buffer_handle *buf ) { return driBOMap( dri_bo(buf), DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0 ); } static void intel_buffer_unmap(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf) + struct pipe_buffer_handle *buf) { driBOUnmap( dri_bo(buf) ); } -static struct softpipe_buffer_handle * +static struct pipe_buffer_handle * intel_buffer_reference(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf) + struct pipe_buffer_handle *buf) { return pipe_bo( driBOReference( dri_bo(buf) ) ); } static void intel_buffer_unreference(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf) + struct pipe_buffer_handle **buf) { - driBOUnReference( dri_bo(buf) ); + if (*buf) { + driBOUnReference( dri_bo(*buf) ); + *buf = NULL; + } } /* Grabs the hardware lock! */ static void intel_buffer_data(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf, + struct pipe_buffer_handle *buf, unsigned size, const void *data ) { struct intel_context *intel = intel_softpipe_winsys(sws)->intel; @@ -117,7 +120,7 @@ static void intel_buffer_data(struct softpipe_winsys *sws, } static void intel_buffer_subdata(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf, + struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, const void *data) @@ -126,7 +129,7 @@ static void intel_buffer_subdata(struct softpipe_winsys *sws, } static void intel_buffer_get_subdata(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf, + struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, void *data) @@ -137,9 +140,8 @@ static void intel_buffer_get_subdata(struct softpipe_winsys *sws, /* Softpipe has no concept of pools. We choose the tex/region pool * for all buffers. */ -static struct softpipe_buffer_handle * +static struct pipe_buffer_handle * intel_create_buffer(struct softpipe_winsys *sws, - const char *name, unsigned alignment) { struct intel_context *intel = intel_softpipe_winsys(sws)->intel; @@ -147,7 +149,7 @@ intel_create_buffer(struct softpipe_winsys *sws, LOCK_HARDWARE( intel ); driGenBuffers( intel->intelScreen->regionPool, - name, 1, &buffer, alignment, 0, 0 ); + "softpipe buffer", 1, &buffer, alignment, 0, 0 ); UNLOCK_HARDWARE( intel ); return pipe_bo(buffer); diff --git a/src/mesa/drivers/dri/i915pipe/intel_surface.c b/src/mesa/drivers/dri/i915pipe/intel_surface.c index 78ed995794..3fa40271c8 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_surface.c +++ b/src/mesa/drivers/dri/i915pipe/intel_surface.c @@ -10,7 +10,6 @@ #include "intel_context.h" #include "intel_buffers.h" -#include "intel_regions.h" #include "intel_fbo.h" #include "pipe/p_state.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c index b4e440e49a..9f8e115533 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c @@ -225,9 +225,11 @@ try_pbo_upload(struct intel_context *intel, { struct _DriBufferObject *src_buffer = intel_bufferobj_buffer(intel, pbo, INTEL_READ); + + /* Temporary hack: cast to _DriBufferObject: + */ struct _DriBufferObject *dst_buffer = - intel->pipe->region_buffer(intel->pipe, intelImage->mt->region, - INTEL_WRITE_FULL); + (struct _DriBufferObject *)intelImage->mt->region->buffer; intelEmitCopyBlit(intel, diff --git a/src/mesa/pipe/Makefile.template b/src/mesa/pipe/Makefile.template index 7c64981d7b..84b4dcd13a 100644 --- a/src/mesa/pipe/Makefile.template +++ b/src/mesa/pipe/Makefile.template @@ -39,6 +39,7 @@ $(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/mesa/pipe/Makefile.template depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ $(ASM_SOURCES) 2> /dev/null diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 92670c7733..76a2593567 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -169,9 +169,44 @@ struct pipe_context { GLuint width, GLuint height, GLuint value); - struct _DriBufferObject *(*region_buffer)(struct pipe_context *pipe, - struct pipe_region *region, - GLuint flag); + + /* Buffer management functions need to be exposed as well. A pipe + * buffer may be used as a texture, render target or vertex/index + * buffer, or some combination according to flags. + */ + + struct pipe_buffer_handle *(*create_buffer)(struct pipe_context *pipe, + unsigned alignment, + unsigned flags ); + + void *(*buffer_map)( struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned flags ); + + void (*buffer_unmap)( struct pipe_context *pipe, + struct pipe_buffer_handle *buf ); + + struct pipe_buffer_handle *(*buffer_reference)( struct pipe_context *pipe, + struct pipe_buffer_handle *buf ); + + void (*buffer_unreference)( struct pipe_context *pipe, + struct pipe_buffer_handle **buf ); + + void (*buffer_data)(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ); + + void (*buffer_subdata)(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data); + + void (*buffer_get_subdata)(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data); }; diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 3be1010007..518110d447 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -292,4 +292,17 @@ struct pipe_texture_object }; +struct pipe_buffer_handle; + +#define PIPE_BUFFER_FLAG_READ 0x1 +#define PIPE_BUFFER_FLAG_WRITE 0x2 + +#define PIPE_BUFFER_USE_TEXTURE 0x1 +#define PIPE_BUFFER_USE_VERTEX_BUFFER 0x2 +#define PIPE_BUFFER_USE_INDEX_BUFFER 0x4 +#define PIPE_BUFFER_USE_RENDER_TARGET 0x8 + + + + #endif diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 04464154e7..0eacfa2182 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -5,6 +5,7 @@ include $(TOP)/configs/current LIBNAME = softpipe DRIVER_SOURCES = \ + sp_buffer.c \ sp_clear.c \ sp_context.c \ sp_prim_setup.c \ diff --git a/src/mesa/pipe/softpipe/sp_buffer.c b/src/mesa/pipe/softpipe/sp_buffer.c new file mode 100644 index 0000000000..f3a4718449 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_buffer.c @@ -0,0 +1,119 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include "sp_context.h" +#include "sp_winsys.h" + + + +/* Most callbacks map direcly onto winsys operations: + */ +static struct pipe_buffer_handle * +sp_create_buffer(struct pipe_context *pipe, + unsigned alignment, + unsigned flags) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + return sp->winsys->create_buffer( sp->winsys, alignment ); +} + +static void *sp_buffer_map(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned flags ) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + return sp->winsys->buffer_map( sp->winsys, buf ); +} + +static void sp_buffer_unmap(struct pipe_context *pipe, + struct pipe_buffer_handle *buf) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + sp->winsys->buffer_unmap( sp->winsys, buf ); +} + +static struct pipe_buffer_handle * +sp_buffer_reference(struct pipe_context *pipe, + struct pipe_buffer_handle *buf) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + return sp->winsys->buffer_reference( sp->winsys, buf ); +} + +static void sp_buffer_unreference(struct pipe_context *pipe, + struct pipe_buffer_handle **buf) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + sp->winsys->buffer_unreference( sp->winsys, buf ); +} + +static void sp_buffer_data(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + sp->winsys->buffer_data( sp->winsys, buf, size, data ); +} + +static void sp_buffer_subdata(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + sp->winsys->buffer_subdata( sp->winsys, buf, offset, size, data ); +} + +static void sp_buffer_get_subdata(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) +{ + struct softpipe_context *sp = softpipe_context( pipe ); + sp->winsys->buffer_get_subdata( sp->winsys, buf, offset, size, data ); +} + + +void +sp_init_buffer_functions( struct softpipe_context *sp ) +{ + sp->pipe.create_buffer = sp_create_buffer; + sp->pipe.buffer_map = sp_buffer_map; + sp->pipe.buffer_unmap = sp_buffer_unmap; + sp->pipe.buffer_reference = sp_buffer_reference; + sp->pipe.buffer_unreference = sp_buffer_unreference; + sp->pipe.buffer_data = sp_buffer_data; + sp->pipe.buffer_subdata = sp_buffer_subdata; + sp->pipe.buffer_get_subdata = sp_buffer_get_subdata; +} diff --git a/src/mesa/pipe/softpipe/sp_buffer.h b/src/mesa/pipe/softpipe/sp_buffer.h new file mode 100644 index 0000000000..9805c5142a --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_buffer.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SP_BUFFER_H +#define SP_BUFFER_H + + +struct softpipe_context; + + +extern void +sp_init_buffer_functions(struct softpipe_context *sp); + + +#endif /* SP_BUFFER_H */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 22928f7bd6..9352c7afa7 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -36,6 +36,7 @@ #include "sp_context.h" #include "sp_clear.h" #include "sp_region.h" +#include "sp_buffer.h" #include "sp_state.h" #include "sp_surface.h" #include "sp_prim_setup.h" @@ -219,6 +220,7 @@ struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) assert(softpipe->draw); draw_set_setup_stage(softpipe->draw, sp_draw_render_stage(softpipe)); + sp_init_buffer_functions(softpipe); sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 78a0919ad5..a0ced3eb86 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -80,7 +80,7 @@ sp_region_alloc(struct pipe_context *pipe, region->height = height; /* needed? */ region->refcount = 1; - region->buffer = sp->winsys->create_buffer(sp->winsys, "region", 64 ); + region->buffer = sp->winsys->create_buffer( sp->winsys, 64 ); sp->winsys->buffer_data( sp->winsys, region->buffer, @@ -259,13 +259,6 @@ sp_region_fill(struct pipe_context *pipe, -static struct _DriBufferObject * -sp_region_buffer(struct pipe_context *pipe, - struct pipe_region *region, GLuint flag) -{ - return region->buffer; -} - void @@ -279,6 +272,5 @@ sp_init_region_functions(struct softpipe_context *sp) sp->pipe.region_data = sp_region_data; sp->pipe.region_copy = sp_region_copy; sp->pipe.region_fill = sp_region_fill; - sp->pipe.region_buffer = sp_region_buffer; } diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h index 43953c648b..bc6db15d32 100644 --- a/src/mesa/pipe/softpipe/sp_winsys.h +++ b/src/mesa/pipe/softpipe/sp_winsys.h @@ -46,7 +46,7 @@ * etc. */ -struct softpipe_buffer_handle; +struct pipe_buffer_handle; struct softpipe_winsys { @@ -60,34 +60,33 @@ struct softpipe_winsys { * Softpipe only really wants to make system memory allocations, * right?? */ - struct softpipe_buffer_handle *(*create_buffer)(struct softpipe_winsys *sws, - const char *name, - unsigned alignment ); + struct pipe_buffer_handle *(*create_buffer)(struct softpipe_winsys *sws, + unsigned alignment ); void *(*buffer_map)( struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf ); + struct pipe_buffer_handle *buf ); void (*buffer_unmap)( struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf ); + struct pipe_buffer_handle *buf ); - struct softpipe_buffer_handle *(*buffer_reference)( struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf ); + struct pipe_buffer_handle *(*buffer_reference)( struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf ); void (*buffer_unreference)( struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf ); + struct pipe_buffer_handle **buf ); void (*buffer_data)(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf, + struct pipe_buffer_handle *buf, unsigned size, const void *data ); void (*buffer_subdata)(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf, + struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, const void *data); void (*buffer_get_subdata)(struct softpipe_winsys *sws, - struct softpipe_buffer_handle *buf, + struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, void *data); diff --git a/src/mesa/sources b/src/mesa/sources index 4e3e44f15a..bd1bbd4299 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -196,6 +196,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_program.c \ state_tracker/st_cb_teximage.c \ + state_tracker/st_cb_bufferobjects.c \ state_tracker/st_draw.c \ state_tracker/st_context.c \ state_tracker/st_texobj.c diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c new file mode 100644 index 0000000000..a667b3e775 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -0,0 +1,206 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "imports.h" +#include "mtypes.h" +#include "bufferobj.h" + +#include "st_context.h" +#include "st_cb_bufferobjects.h" + +#include "pipe/p_context.h" + +/* Pixel buffers and Vertex/index buffers are handled through these + * mesa callbacks. Framebuffer/Renderbuffer objects are + * created/managed elsewhere. + */ + + + +/** + * There is some duplication between mesa's bufferobjects and our + * bufmgr buffers. Both have an integer handle and a hashtable to + * lookup an opaque structure. It would be nice if the handles and + * internal structure where somehow shared. + */ +static struct gl_buffer_object * +st_bufferobj_alloc(GLcontext *ctx, GLuint name, GLenum target) +{ + struct st_context *st = st_context(ctx); + struct st_buffer_object *st_obj = CALLOC_STRUCT(st_buffer_object); + + _mesa_initialize_buffer_object(&st_obj->Base, name, target); + + st_obj->buffer = st->pipe->create_buffer( st->pipe, 32, 0 ); + + return &st_obj->Base; +} + + + +/** + * Deallocate/free a vertex/pixel buffer object. + * Called via glDeleteBuffersARB(). + */ +static void +st_bufferobj_free(GLcontext *ctx, struct gl_buffer_object *obj) +{ + struct pipe_context *pipe = st_context(ctx)->pipe; + struct st_buffer_object *st_obj = st_buffer_object(obj); + + if (st_obj->buffer) + pipe->buffer_unreference(pipe, &st_obj->buffer); + + FREE(st_obj); +} + + + +/** + * Allocate space for and store data in a buffer object. Any data that was + * previously stored in the buffer object is lost. If data is NULL, + * memory will be allocated, but no copy will occur. + * Called via glBufferDataARB(). + */ +static void +st_bufferobj_data(GLcontext *ctx, + GLenum target, + GLsizeiptrARB size, + const GLvoid * data, + GLenum usage, + struct gl_buffer_object *obj) +{ + struct pipe_context *pipe = st_context(ctx)->pipe; + struct st_buffer_object *st_obj = st_buffer_object(obj); + + st_obj->Base.Size = size; + st_obj->Base.Usage = usage; + + pipe->buffer_data( pipe, st_obj->buffer, size, data ); +} + + +/** + * Replace data in a subrange of buffer object. If the data range + * specified by size + offset extends beyond the end of the buffer or + * if data is NULL, no copy is performed. + * Called via glBufferSubDataARB(). + */ +static void +st_bufferobj_subdata(GLcontext *ctx, + GLenum target, + GLintptrARB offset, + GLsizeiptrARB size, + const GLvoid * data, struct gl_buffer_object *obj) +{ + struct pipe_context *pipe = st_context(ctx)->pipe; + struct st_buffer_object *st_obj = st_buffer_object(obj); + + pipe->buffer_subdata(pipe, st_obj->buffer, offset, size, data); +} + + +/** + * Called via glGetBufferSubDataARB(). + */ +static void +st_bufferobj_get_subdata(GLcontext *ctx, + GLenum target, + GLintptrARB offset, + GLsizeiptrARB size, + GLvoid * data, struct gl_buffer_object *obj) +{ + struct pipe_context *pipe = st_context(ctx)->pipe; + struct st_buffer_object *st_obj = st_buffer_object(obj); + + pipe->buffer_get_subdata(pipe, st_obj->buffer, offset, size, data); +} + + + +/** + * Called via glMapBufferARB(). + */ +static void * +st_bufferobj_map(GLcontext *ctx, + GLenum target, + GLenum access, struct gl_buffer_object *obj) +{ + struct pipe_context *pipe = st_context(ctx)->pipe; + struct st_buffer_object *st_obj = st_buffer_object(obj); + GLuint flags; + + switch (access) { + case GL_WRITE_ONLY: + flags = PIPE_BUFFER_FLAG_WRITE; + break; + + + flags = PIPE_BUFFER_FLAG_READ; + break; + + case GL_READ_WRITE: + default: + flags = PIPE_BUFFER_FLAG_READ | PIPE_BUFFER_FLAG_WRITE; + break; + } + + obj->Pointer = pipe->buffer_map(pipe, st_obj->buffer, flags); + return obj->Pointer; +} + + +/** + * Called via glMapBufferARB(). + */ +static GLboolean +st_bufferobj_unmap(GLcontext *ctx, + GLenum target, struct gl_buffer_object *obj) +{ + struct pipe_context *pipe = st_context(ctx)->pipe; + struct st_buffer_object *st_obj = st_buffer_object(obj); + + pipe->buffer_unmap(pipe, st_obj->buffer); + obj->Pointer = NULL; + return GL_TRUE; +} + + +void +st_init_cb_bufferobjects( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + + ctx->Driver.NewBufferObject = st_bufferobj_alloc; + ctx->Driver.DeleteBuffer = st_bufferobj_free; + ctx->Driver.BufferData = st_bufferobj_data; + ctx->Driver.BufferSubData = st_bufferobj_subdata; + ctx->Driver.GetBufferSubData = st_bufferobj_get_subdata; + ctx->Driver.MapBuffer = st_bufferobj_map; + ctx->Driver.UnmapBuffer = st_bufferobj_unmap; +} diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.h b/src/mesa/state_tracker/st_cb_bufferobjects.h new file mode 100644 index 0000000000..2787411c5f --- /dev/null +++ b/src/mesa/state_tracker/st_cb_bufferobjects.h @@ -0,0 +1,66 @@ + /************************************************************************** + * + * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ST_CB_BUFFEROBJECTS_H +#define ST_CB_BUFFEROBJECTS_H + +struct st_context; +struct gl_buffer_object; +struct pipe_buffer_handle; + +/** + * State_tracker vertex/pixel buffer object, derived from Mesa's + * gl_buffer_object. + */ +struct st_buffer_object +{ + struct gl_buffer_object Base; + struct pipe_buffer_handle *buffer; +}; + + +/* Hook the bufferobject implementation into mesa: + */ +void st_init_cb_bufferobjects( struct st_context *st ); + + +/* Are the obj->Name tests necessary? Unfortunately yes, mesa + * allocates a couple of gl_buffer_object structs statically, and the + * Name == 0 test is the only way to identify them and avoid casting + * them erroneously to our structs. + */ +static INLINE struct st_buffer_object * +st_buffer_object(struct gl_buffer_object *obj) +{ + if (obj->Name) + return (struct st_buffer_object *) obj; + else + return NULL; +} + + +#endif -- cgit v1.2.3 From d9605cdf7cbbd3c81c506d01eb8e88a11ccfc09b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 11:22:00 -0600 Subject: Checkpoint lifting of intel_mipmap_tree (intel_mipmap_tree -> pipe_mipmap_tree and move some code) --- src/mesa/drivers/dri/i915pipe/intel_context.c | 26 ++ src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c | 54 +-- src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h | 114 +------ src/mesa/drivers/dri/i915pipe/intel_tex.h | 4 +- src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_tex_validate.c | 2 + src/mesa/drivers/dri/intel/intel_tex_layout.c | 362 ++++++++++++++++++++- src/mesa/drivers/dri/intel/intel_tex_layout.h | 7 +- src/mesa/pipe/p_context.h | 8 + src/mesa/pipe/p_state.h | 55 ++++ 10 files changed, 485 insertions(+), 149 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index d43b784112..df34c360e5 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -53,6 +53,7 @@ #include "intel_blit.h" #include "intel_buffer_objects.h" #include "intel_fbo.h" +#include "intel_mipmap_tree.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" @@ -61,6 +62,11 @@ #include "vblank.h" #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ + +#include "pipe/p_context.h" + + + #ifndef INTEL_DEBUG int INTEL_DEBUG = (0); #endif @@ -377,6 +383,26 @@ intelCreateContext(const __GLcontextModes * mesaVis, // intel->pipe->glctx = ctx; // intel_init_region_functions(intel->pipe); + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + intel->pipe->mipmap_tree_layout = i945_miptree_layout; + break; + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + case PCI_CHIP_I830_M: + case PCI_CHIP_I855_GM: + case PCI_CHIP_I865_G: + intel->pipe->mipmap_tree_layout = i915_miptree_layout; + default: + assert(0); /*FIX*/ + } + + /* * memory pools */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c index 6717984f7d..50e830281c 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c @@ -51,7 +51,7 @@ target_to_target(GLenum target) } } -struct intel_mipmap_tree * +struct pipe_mipmap_tree * intel_miptree_create(struct intel_context *intel, GLenum target, GLenum internal_format, @@ -62,7 +62,7 @@ intel_miptree_create(struct intel_context *intel, GLuint depth0, GLuint cpp, GLuint compress_byte) { GLboolean ok; - struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1); + struct pipe_mipmap_tree *mt = calloc(sizeof(*mt), 1); DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), @@ -79,29 +79,7 @@ intel_miptree_create(struct intel_context *intel, mt->compressed = compress_byte ? 1 : 0; mt->refcount = 1; - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: -// ok = i945_miptree_layout(mt); - break; - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - case PCI_CHIP_I830_M: - case PCI_CHIP_I855_GM: - case PCI_CHIP_I865_G: - default: - /* All the i830 chips and the i915 use this layout: - */ -// ok = i915_miptree_layout(mt); - break; - } - - ok = 0; /* TODO */ - + ok = intel->pipe->mipmap_tree_layout(intel->pipe, mt); if (ok) mt->region = intel->pipe->region_alloc(intel->pipe, mt->cpp, mt->pitch, mt->total_height); @@ -116,8 +94,8 @@ intel_miptree_create(struct intel_context *intel, void -intel_miptree_reference(struct intel_mipmap_tree **dst, - struct intel_mipmap_tree *src) +intel_miptree_reference(struct pipe_mipmap_tree **dst, + struct pipe_mipmap_tree *src) { src->refcount++; *dst = src; @@ -126,7 +104,7 @@ intel_miptree_reference(struct intel_mipmap_tree **dst, void intel_miptree_release(struct intel_context *intel, - struct intel_mipmap_tree **mt) + struct pipe_mipmap_tree **mt) { if (!*mt) return; @@ -157,7 +135,7 @@ intel_miptree_release(struct intel_context *intel, * Not sure whether I want to pass gl_texture_image here. */ GLboolean -intel_miptree_match_image(struct intel_mipmap_tree *mt, +intel_miptree_match_image(struct pipe_mipmap_tree *mt, struct gl_texture_image *image, GLuint face, GLuint level) { @@ -184,7 +162,7 @@ intel_miptree_match_image(struct intel_mipmap_tree *mt, void -intel_miptree_set_level_info(struct intel_mipmap_tree *mt, +intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, GLuint level, GLuint nr_images, GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) @@ -215,7 +193,7 @@ intel_miptree_set_level_info(struct intel_mipmap_tree *mt, void -intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, +intel_miptree_set_image_offset(struct pipe_mipmap_tree *mt, GLuint level, GLuint img, GLuint x, GLuint y) { if (img == 0 && level == 0) @@ -237,7 +215,7 @@ intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, * These functions present that view to mesa: */ const GLuint * -intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, GLuint level) +intel_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) { static const GLuint zero = 0; @@ -249,7 +227,7 @@ intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, GLuint level) GLuint -intel_miptree_image_offset(struct intel_mipmap_tree * mt, +intel_miptree_image_offset(struct pipe_mipmap_tree * mt, GLuint face, GLuint level) { if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) @@ -269,7 +247,7 @@ intel_miptree_image_offset(struct intel_mipmap_tree * mt, */ GLubyte * intel_miptree_image_map(struct intel_context * intel, - struct intel_mipmap_tree * mt, + struct pipe_mipmap_tree * mt, GLuint face, GLuint level, GLuint * row_stride, GLuint * image_offsets) @@ -291,7 +269,7 @@ intel_miptree_image_map(struct intel_context * intel, void intel_miptree_image_unmap(struct intel_context *intel, - struct intel_mipmap_tree *mt) + struct pipe_mipmap_tree *mt) { DBG("%s\n", __FUNCTION__); intel->pipe->region_unmap(intel->pipe, mt->region); @@ -303,7 +281,7 @@ intel_miptree_image_unmap(struct intel_context *intel, */ void intel_miptree_image_data(struct intel_context *intel, - struct intel_mipmap_tree *dst, + struct pipe_mipmap_tree *dst, GLuint face, GLuint level, void *src, @@ -336,9 +314,9 @@ intel_miptree_image_data(struct intel_context *intel, */ void intel_miptree_image_copy(struct intel_context *intel, - struct intel_mipmap_tree *dst, + struct pipe_mipmap_tree *dst, GLuint face, GLuint level, - struct intel_mipmap_tree *src) + struct pipe_mipmap_tree *src) { GLuint width = src->level[level].width; GLuint height = src->level[level].height; diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h index 09b2e362fc..94fb21372c 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h @@ -34,89 +34,7 @@ struct pipe_region; -/* A layer on top of the pipe_regions code which adds: - * - * - Code to size and layout a region to hold a set of mipmaps. - * - Query to determine if a new image fits in an existing tree. - * - More refcounting - * - maybe able to remove refcounting from pipe_region? - * - ? - * - * The fixed mipmap layout of intel hardware where one offset - * specifies the position of all images in a mipmap hierachy - * complicates the implementation of GL texture image commands, - * compared to hardware where each image is specified with an - * independent offset. - * - * In an ideal world, each texture object would be associated with a - * single bufmgr buffer or 2d pipe_region, and all the images within - * the texture object would slot into the tree as they arrive. The - * reality can be a little messier, as images can arrive from the user - * with sizes that don't fit in the existing tree, or in an order - * where the tree layout cannot be guessed immediately. - * - * This structure encodes an idealized mipmap tree. The GL image - * commands build these where possible, otherwise store the images in - * temporary system buffers. - */ - - -/** - * Describes the location of each texture image within a texture region. - */ -struct intel_mipmap_level -{ - GLuint level_offset; - GLuint width; - GLuint height; - GLuint depth; - GLuint nr_images; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - GLuint *image_offset; -}; - -struct intel_mipmap_tree -{ - /* Effectively the key: - */ - GLenum target; - GLenum internal_format; - - GLuint first_level; - GLuint last_level; - - GLuint width0, height0, depth0; /**< Level zero image dimensions */ - GLuint cpp; - GLboolean compressed; - - /* Derived from the above: - */ - GLuint pitch; - GLuint depth_pitch; /* per-image on i945? */ - GLuint total_height; - - /* Includes image offset tables: - */ - struct intel_mipmap_level level[MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_region *region; - - /* These are also refcounted: - */ - GLuint refcount; -}; - - - -struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel, +struct pipe_mipmap_tree *intel_miptree_create(struct intel_context *intel, GLenum target, GLenum internal_format, GLuint first_level, @@ -127,15 +45,15 @@ struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel, GLuint cpp, GLuint compress_byte); -void intel_miptree_reference(struct intel_mipmap_tree **dst, - struct intel_mipmap_tree *src); +void intel_miptree_reference(struct pipe_mipmap_tree **dst, + struct pipe_mipmap_tree *src); void intel_miptree_release(struct intel_context *intel, - struct intel_mipmap_tree **mt); + struct pipe_mipmap_tree **mt); /* Check if an image fits an existing mipmap tree layout */ -GLboolean intel_miptree_match_image(struct intel_mipmap_tree *mt, +GLboolean intel_miptree_match_image(struct pipe_mipmap_tree *mt, struct gl_texture_image *image, GLuint face, GLuint level); @@ -143,35 +61,35 @@ GLboolean intel_miptree_match_image(struct intel_mipmap_tree *mt, * well. */ GLubyte *intel_miptree_image_map(struct intel_context *intel, - struct intel_mipmap_tree *mt, + struct pipe_mipmap_tree *mt, GLuint face, GLuint level, GLuint * row_stride, GLuint * image_stride); void intel_miptree_image_unmap(struct intel_context *intel, - struct intel_mipmap_tree *mt); + struct pipe_mipmap_tree *mt); /* Return the linear offset of an image relative to the start of the * tree: */ -GLuint intel_miptree_image_offset(struct intel_mipmap_tree *mt, +GLuint intel_miptree_image_offset(struct pipe_mipmap_tree *mt, GLuint face, GLuint level); /* Return pointers to each 2d slice within an image. Indexed by depth * value. */ -const GLuint *intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, +const GLuint *intel_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level); -void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, +void intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, GLuint level, GLuint nr_images, GLuint x, GLuint y, GLuint w, GLuint h, GLuint d); -void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, +void intel_miptree_set_image_offset(struct pipe_mipmap_tree *mt, GLuint level, GLuint img, GLuint x, GLuint y); @@ -179,7 +97,7 @@ void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, /* Upload an image into a tree */ void intel_miptree_image_data(struct intel_context *intel, - struct intel_mipmap_tree *dst, + struct pipe_mipmap_tree *dst, GLuint face, GLuint level, void *src, @@ -188,14 +106,14 @@ void intel_miptree_image_data(struct intel_context *intel, /* Copy an image between two trees */ void intel_miptree_image_copy(struct intel_context *intel, - struct intel_mipmap_tree *dst, + struct pipe_mipmap_tree *dst, GLuint face, GLuint level, - struct intel_mipmap_tree *src); + struct pipe_mipmap_tree *src); /* i915_mipmap_tree.c: */ -GLboolean i915_miptree_layout(struct intel_mipmap_tree *mt); -GLboolean i945_miptree_layout(struct intel_mipmap_tree *mt); +GLboolean i915_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); +GLboolean i945_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.h b/src/mesa/drivers/dri/i915pipe/intel_tex.h index 4a5081eee7..1ee239402e 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.h +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.h @@ -49,7 +49,7 @@ struct intel_texture_object /* On validation any active images held in main memory or in other * regions will be copied to this region and the old storage freed. */ - struct intel_mipmap_tree *mt; + struct pipe_mipmap_tree *mt; GLboolean imageOverride; GLint depthOverride; @@ -71,7 +71,7 @@ struct intel_texture_image * Else if intelImage->base.Data != NULL, image is stored there. * Else there is no image data. */ - struct intel_mipmap_tree *mt; + struct pipe_mipmap_tree *mt; }; diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c index 9f8e115533..596493f45a 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c @@ -16,9 +16,9 @@ #include "texstore.h" #include "intel_context.h" -#include "intel_mipmap_tree.h" #include "intel_buffer_objects.h" #include "intel_batchbuffer.h" +#include "intel_mipmap_tree.h" #include "intel_tex.h" #include "intel_ioctl.h" #include "intel_blit.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c index af18c26d55..242fe1237b 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c @@ -6,6 +6,8 @@ #include "intel_mipmap_tree.h" #include "intel_tex.h" +#include "pipe/p_state.h" + #define FILE_DEBUG_FLAG DEBUG_TEXTURE /** diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c index fcb5cc3906..bedd835cbb 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.c +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c @@ -30,17 +30,57 @@ * Michel Dänzer */ -#include "intel_mipmap_tree.h" -#include "intel_tex_layout.h" #include "macros.h" +#include "pipe/p_state.h" +#include "intel_mipmap_tree.h" + +static GLuint minify( GLuint d ) +{ + return MAX2(1, d>>1); +} static int align(int value, int alignment) { return (value + alignment - 1) & ~(alignment - 1); } -void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ) + +static void +mipmaptree_set_level_info(struct pipe_mipmap_tree *mt, + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) +{ + assert(level < MAX_TEXTURE_LEVELS); + + mt->level[level].width = w; + mt->level[level].height = h; + mt->level[level].depth = d; + mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; + mt->level[level].nr_images = nr_images; + + /* + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, mt->level[level].level_offset); + */ + + /* Not sure when this would happen, but anyway: + */ + if (mt->level[level].image_offset) { + free(mt->level[level].image_offset); + mt->level[level].image_offset = NULL; + } + + assert(nr_images); + + mt->level[level].image_offset = malloc(nr_images * sizeof(GLuint)); + mt->level[level].image_offset[0] = 0; +} + + +static void +i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) { GLint align_h = 2, align_w = 4; GLuint level; @@ -73,8 +113,7 @@ void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ) for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { GLuint img_height; - intel_miptree_set_level_info(mt, level, 1, x, y, width, - height, 1); + mipmaptree_set_level_info(mt, level, 1, x, y, width, height, 1); if (mt->compressed) img_height = MAX2(1, height/4); @@ -100,3 +139,316 @@ void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ) height = minify(height); } } + + +static const GLint initial_offsets[6][2] = { + {0, 0}, + {0, 2}, + {1, 0}, + {1, 2}, + {1, 1}, + {1, 3} +}; + +static const GLint step_offsets[6][2] = { + {0, 2}, + {0, 2}, + {-1, 2}, + {-1, 2}, + {-1, 1}, + {-1, 1} +}; + + +GLboolean +i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) +{ + GLint level; + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP:{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* double pitch for cube layouts */ + mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; + mt->total_height = dim * 4; + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_level_info(mt, level, 6, + 0, 0, + /*OLD: mt->pitch, mt->total_height,*/ + lvlWidth, lvlHeight, + 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + for (face = 0; face < 6; face++) { + GLuint x = initial_offsets[face][0] * dim; + GLuint y = initial_offsets[face][1] * dim; + GLuint d = dim; + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_image_offset(mt, level, face, x, y); + + if (d == 0) + _mesa_printf("cube mipmap %d/%d (%d..%d) is 0x0\n", + face, level, mt->first_level, mt->last_level); + + d >>= 1; + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + } + } + break; + } + case GL_TEXTURE_3D:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint stack_height = 0; + + /* Calculate the size of a single slice. + */ + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + + /* XXX: hardware expects/requires 9 levels at minimum. + */ + for (level = mt->first_level; level <= MAX2(8, mt->last_level); + level++) { + intel_miptree_set_level_info(mt, level, depth, 0, mt->total_height, + width, height, depth); + + + stack_height += MAX2(2, height); + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + + /* Fixup depth image_offsets: + */ + depth = mt->depth0; + for (level = mt->first_level; level <= mt->last_level; level++) { + GLuint i; + for (i = 0; i < depth; i++) + intel_miptree_set_image_offset(mt, level, i, + 0, i * stack_height); + + depth = minify(depth); + } + + + /* Multiply slice size by texture depth for total size. It's + * remarkable how wasteful of memory the i915 texture layouts + * are. They are largely fixed in the i945. + */ + mt->total_height = stack_height * mt->depth0; + break; + } + + default:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint img_height; + + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->total_height = 0; + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_level_info(mt, level, 1, + 0, mt->total_height, + width, height, 1); + + if (mt->compressed) + img_height = MAX2(1, height / 4); + else + img_height = (MAX2(2, height) + 1) & ~1; + + mt->total_height += img_height; + + width = minify(width); + height = minify(height); + } + break; + } + } + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + mt->pitch, + mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); + */ + + return GL_TRUE; +} + + +GLboolean +i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) +{ + GLint level; + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP:{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* Depending on the size of the largest images, pitch can be + * determined either by the old-style packing of cubemap faces, + * or the final row of 4x4, 2x2 and 1x1 faces below this. + */ + if (dim > 32) + mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; + else + mt->pitch = 14 * 8; + + mt->total_height = dim * 4 + 4; + + /* Set all the levels to effectively occupy the whole rectangular region. + */ + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_level_info(mt, level, 6, + 0, 0, + lvlWidth, lvlHeight, 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + + for (face = 0; face < 6; face++) { + GLuint x = initial_offsets[face][0] * dim; + GLuint y = initial_offsets[face][1] * dim; + GLuint d = dim; + + if (dim == 4 && face >= 4) { + y = mt->total_height - 4; + x = (face - 4) * 8; + } + else if (dim < 4 && (face > 0 || mt->first_level > 0)) { + y = mt->total_height - 4; + x = face * 8; + } + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_image_offset(mt, level, face, x, y); + + d >>= 1; + + switch (d) { + case 4: + switch (face) { + case FACE_POS_X: + case FACE_NEG_X: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + case FACE_POS_Y: + case FACE_NEG_Y: + y += 12; + x -= 8; + break; + case FACE_POS_Z: + case FACE_NEG_Z: + y = mt->total_height - 4; + x = (face - 4) * 8; + break; + } + + case 2: + y = mt->total_height - 4; + x = 16 + face * 8; + break; + + case 1: + x += 48; + break; + + default: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + } + } + } + break; + } + case GL_TEXTURE_3D:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint pack_x_pitch, pack_x_nr; + GLuint pack_y_pitch; + GLuint level; + + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->total_height = 0; + + pack_y_pitch = MAX2(mt->height0, 2); + pack_x_pitch = mt->pitch; + pack_x_nr = 1; + + for (level = mt->first_level; level <= mt->last_level; level++) { + GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; + GLint x = 0; + GLint y = 0; + GLint q, j; + + intel_miptree_set_level_info(mt, level, nr_images, + 0, mt->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + intel_miptree_set_image_offset(mt, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + mt->total_height += y; + + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= mt->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + } + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + break; + } + + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_RECTANGLE_ARB: + i945_miptree_layout_2d(mt); + break; + default: + _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()"); + } + + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + mt->pitch, + mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); + */ + + return GL_TRUE; +} diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h index 1e37f8f525..cbd4b8e27f 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.h +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h @@ -33,9 +33,6 @@ #include "macros.h" -static GLuint minify( GLuint d ) -{ - return MAX2(1, d>>1); -} +extern void i915_miptree_layout_2d( struct pipe_context *, struct pipe_mipmap_tree *mt ); -extern void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ); +extern void i945_miptree_layout_2d( struct pipe_context *, struct pipe_mipmap_tree *mt ); diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 76a2593567..db9429f8d3 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -207,6 +207,14 @@ struct pipe_context { unsigned long offset, unsigned long size, void *data); + + /* + * Texture functions + */ + GLboolean (*mipmap_tree_layout)( struct pipe_context *pipe, + struct pipe_mipmap_tree *mt ); + + }; diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 518110d447..13b8813f4a 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -292,6 +292,61 @@ struct pipe_texture_object }; +/** + * Describes the location of each texture image within a texture region. + */ +struct pipe_mipmap_level +{ + GLuint level_offset; + GLuint width; + GLuint height; + GLuint depth; + GLuint nr_images; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + GLuint *image_offset; +}; + +struct pipe_mipmap_tree +{ + /* Effectively the key: + */ + GLenum target; + GLenum internal_format; + + GLuint first_level; + GLuint last_level; + + GLuint width0, height0, depth0; /**< Level zero image dimensions */ + GLuint cpp; + GLboolean compressed; + + /* Derived from the above: + */ + GLuint pitch; + GLuint depth_pitch; /* per-image on i945? */ + GLuint total_height; + + /* Includes image offset tables: + */ + struct pipe_mipmap_level level[MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_region *region; + + /* These are also refcounted: + */ + GLuint refcount; +}; + + + struct pipe_buffer_handle; #define PIPE_BUFFER_FLAG_READ 0x1 -- cgit v1.2.3 From e6c8691b1a3b8bd44206b9aaa8d75c7c5ea819ed Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 11:51:16 -0600 Subject: checkpoint: move some texture layout code --- src/mesa/drivers/dri/i915pipe/intel_context.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c | 47 ----------------------- src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h | 17 -------- 3 files changed, 1 insertion(+), 65 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index df34c360e5..e267135973 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -53,7 +53,7 @@ #include "intel_blit.h" #include "intel_buffer_objects.h" #include "intel_fbo.h" -#include "intel_mipmap_tree.h" +#include "intel_tex_layout.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c index 50e830281c..990e5c0e40 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c @@ -161,53 +161,6 @@ intel_miptree_match_image(struct pipe_mipmap_tree *mt, } -void -intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, - GLuint level, - GLuint nr_images, - GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) -{ - - mt->level[level].width = w; - mt->level[level].height = h; - mt->level[level].depth = d; - mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; - mt->level[level].nr_images = nr_images; - - DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - level, w, h, d, x, y, mt->level[level].level_offset); - - /* Not sure when this would happen, but anyway: - */ - if (mt->level[level].image_offset) { - free(mt->level[level].image_offset); - mt->level[level].image_offset = NULL; - } - - assert(nr_images); - - mt->level[level].image_offset = malloc(nr_images * sizeof(GLuint)); - mt->level[level].image_offset[0] = 0; -} - - - -void -intel_miptree_set_image_offset(struct pipe_mipmap_tree *mt, - GLuint level, GLuint img, GLuint x, GLuint y) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < mt->level[level].nr_images); - - mt->level[level].image_offset[img] = (x + y * mt->pitch); - - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); -} - - /* Although we use the image_offset[] array to store relative offsets * to cube faces, Mesa doesn't know anything about this and expects * each cube face to be treated as a separate image. diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h index 94fb21372c..f453462f51 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h @@ -83,17 +83,6 @@ const GLuint *intel_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level); -void intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, - GLuint level, - GLuint nr_images, - GLuint x, GLuint y, - GLuint w, GLuint h, GLuint d); - -void intel_miptree_set_image_offset(struct pipe_mipmap_tree *mt, - GLuint level, - GLuint img, GLuint x, GLuint y); - - /* Upload an image into a tree */ void intel_miptree_image_data(struct intel_context *intel, @@ -110,11 +99,5 @@ void intel_miptree_image_copy(struct intel_context *intel, GLuint face, GLuint level, struct pipe_mipmap_tree *src); -/* i915_mipmap_tree.c: - */ -GLboolean i915_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); -GLboolean i945_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); - - #endif -- cgit v1.2.3 From 681f04f6d3d4f5a01a9fa0886bc6caecf4657aef Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 12:07:48 -0600 Subject: remove intel dependencies from intel_mipmap_tree.c --- src/mesa/drivers/dri/i915pipe/intel_fbo.c | 4 +- src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c | 53 +++++++++++----------- src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h | 24 +++++----- src/mesa/drivers/dri/i915pipe/intel_tex.c | 4 +- src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 24 +++++----- src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c | 4 +- src/mesa/drivers/dri/i915pipe/intel_tex_validate.c | 22 ++++----- 8 files changed, 68 insertions(+), 69 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index 267c12fd4b..fe5b1a37df 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -566,12 +566,12 @@ intel_render_texture(GLcontext * ctx, } /* compute offset of the particular 2D image within the texture region */ - imageOffset = intel_miptree_image_offset(intel_image->mt, + imageOffset = st_miptree_image_offset(intel_image->mt, att->CubeMapFace, att->TextureLevel); if (att->Texture->Target == GL_TEXTURE_3D) { - const GLuint *offsets = intel_miptree_depth_offsets(intel_image->mt, + const GLuint *offsets = st_miptree_depth_offsets(intel_image->mt, att->TextureLevel); imageOffset += offsets[att->Zoffset]; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c index 990e5c0e40..6a91ddbe5e 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c @@ -25,7 +25,6 @@ * **************************************************************************/ -#include "intel_context.h" #include "intel_mipmap_tree.h" #include "enums.h" @@ -52,7 +51,7 @@ target_to_target(GLenum target) } struct pipe_mipmap_tree * -intel_miptree_create(struct intel_context *intel, +st_miptree_create(struct pipe_context *pipe, GLenum target, GLenum internal_format, GLuint first_level, @@ -79,9 +78,9 @@ intel_miptree_create(struct intel_context *intel, mt->compressed = compress_byte ? 1 : 0; mt->refcount = 1; - ok = intel->pipe->mipmap_tree_layout(intel->pipe, mt); + ok = pipe->mipmap_tree_layout(pipe, mt); if (ok) - mt->region = intel->pipe->region_alloc(intel->pipe, + mt->region = pipe->region_alloc(pipe, mt->cpp, mt->pitch, mt->total_height); if (!mt->region) { @@ -94,7 +93,7 @@ intel_miptree_create(struct intel_context *intel, void -intel_miptree_reference(struct pipe_mipmap_tree **dst, +st_miptree_reference(struct pipe_mipmap_tree **dst, struct pipe_mipmap_tree *src) { src->refcount++; @@ -103,8 +102,8 @@ intel_miptree_reference(struct pipe_mipmap_tree **dst, } void -intel_miptree_release(struct intel_context *intel, - struct pipe_mipmap_tree **mt) +st_miptree_release(struct pipe_context *pipe, + struct pipe_mipmap_tree **mt) { if (!*mt) return; @@ -115,7 +114,7 @@ intel_miptree_release(struct intel_context *intel, DBG("%s deleting %p\n", __FUNCTION__, *mt); - intel->pipe->region_release(intel->pipe, &((*mt)->region)); + pipe->region_release(pipe, &((*mt)->region)); for (i = 0; i < MAX_TEXTURE_LEVELS; i++) if ((*mt)->level[i].image_offset) @@ -135,7 +134,7 @@ intel_miptree_release(struct intel_context *intel, * Not sure whether I want to pass gl_texture_image here. */ GLboolean -intel_miptree_match_image(struct pipe_mipmap_tree *mt, +st_miptree_match_image(struct pipe_mipmap_tree *mt, struct gl_texture_image *image, GLuint face, GLuint level) { @@ -168,7 +167,7 @@ intel_miptree_match_image(struct pipe_mipmap_tree *mt, * These functions present that view to mesa: */ const GLuint * -intel_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) +st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) { static const GLuint zero = 0; @@ -180,8 +179,8 @@ intel_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) GLuint -intel_miptree_image_offset(struct pipe_mipmap_tree * mt, - GLuint face, GLuint level) +st_miptree_image_offset(struct pipe_mipmap_tree * mt, + GLuint face, GLuint level) { if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) return (mt->level[level].level_offset + @@ -199,7 +198,7 @@ intel_miptree_image_offset(struct pipe_mipmap_tree * mt, * \return address of mapping */ GLubyte * -intel_miptree_image_map(struct intel_context * intel, +st_miptree_image_map(struct pipe_context *pipe, struct pipe_mipmap_tree * mt, GLuint face, GLuint level, @@ -215,17 +214,17 @@ intel_miptree_image_map(struct intel_context * intel, memcpy(image_offsets, mt->level[level].image_offset, mt->level[level].depth * sizeof(GLuint)); - ptr = intel->pipe->region_map(intel->pipe, mt->region); + ptr = pipe->region_map(pipe, mt->region); - return ptr + intel_miptree_image_offset(mt, face, level); + return ptr + st_miptree_image_offset(mt, face, level); } void -intel_miptree_image_unmap(struct intel_context *intel, +st_miptree_image_unmap(struct pipe_context *pipe, struct pipe_mipmap_tree *mt) { DBG("%s\n", __FUNCTION__); - intel->pipe->region_unmap(intel->pipe, mt->region); + pipe->region_unmap(pipe, mt->region); } @@ -233,7 +232,7 @@ intel_miptree_image_unmap(struct intel_context *intel, /* Upload data for a particular image. */ void -intel_miptree_image_data(struct intel_context *intel, +st_miptree_image_data(struct pipe_context *pipe, struct pipe_mipmap_tree *dst, GLuint face, GLuint level, @@ -241,8 +240,8 @@ intel_miptree_image_data(struct intel_context *intel, GLuint src_row_pitch, GLuint src_image_pitch) { GLuint depth = dst->level[level].depth; - GLuint dst_offset = intel_miptree_image_offset(dst, face, level); - const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); + GLuint dst_offset = st_miptree_image_offset(dst, face, level); + const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); GLuint i; GLuint height = 0; @@ -251,7 +250,7 @@ intel_miptree_image_data(struct intel_context *intel, height = dst->level[level].height; if(dst->compressed) height /= 4; - intel->pipe->region_data(intel->pipe, dst->region, + pipe->region_data(pipe, dst->region, dst_offset + dst_depth_offset[i], /* dst_offset */ 0, 0, /* dstx, dsty */ src, @@ -266,7 +265,7 @@ intel_miptree_image_data(struct intel_context *intel, /* Copy mipmap image between trees */ void -intel_miptree_image_copy(struct intel_context *intel, +st_miptree_image_copy(struct pipe_context *pipe, struct pipe_mipmap_tree *dst, GLuint face, GLuint level, struct pipe_mipmap_tree *src) @@ -274,16 +273,16 @@ intel_miptree_image_copy(struct intel_context *intel, GLuint width = src->level[level].width; GLuint height = src->level[level].height; GLuint depth = src->level[level].depth; - GLuint dst_offset = intel_miptree_image_offset(dst, face, level); - GLuint src_offset = intel_miptree_image_offset(src, face, level); - const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); - const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level); + GLuint dst_offset = st_miptree_image_offset(dst, face, level); + GLuint src_offset = st_miptree_image_offset(src, face, level); + const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); + const GLuint *src_depth_offset = st_miptree_depth_offsets(src, level); GLuint i; if (dst->compressed) height /= 4; for (i = 0; i < depth; i++) { - intel->pipe->region_copy(intel->pipe, + pipe->region_copy(pipe, dst->region, dst_offset + dst_depth_offset[i], 0, 0, diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h index f453462f51..59f1bfcab4 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h @@ -34,7 +34,7 @@ struct pipe_region; -struct pipe_mipmap_tree *intel_miptree_create(struct intel_context *intel, +struct pipe_mipmap_tree *st_miptree_create(struct pipe_context *pipe, GLenum target, GLenum internal_format, GLuint first_level, @@ -45,47 +45,47 @@ struct pipe_mipmap_tree *intel_miptree_create(struct intel_context *intel, GLuint cpp, GLuint compress_byte); -void intel_miptree_reference(struct pipe_mipmap_tree **dst, +void st_miptree_reference(struct pipe_mipmap_tree **dst, struct pipe_mipmap_tree *src); -void intel_miptree_release(struct intel_context *intel, +void st_miptree_release(struct pipe_context *pipe, struct pipe_mipmap_tree **mt); /* Check if an image fits an existing mipmap tree layout */ -GLboolean intel_miptree_match_image(struct pipe_mipmap_tree *mt, +GLboolean st_miptree_match_image(struct pipe_mipmap_tree *mt, struct gl_texture_image *image, GLuint face, GLuint level); /* Return a pointer to an image within a tree. Return image stride as * well. */ -GLubyte *intel_miptree_image_map(struct intel_context *intel, +GLubyte *st_miptree_image_map(struct pipe_context *pipe, struct pipe_mipmap_tree *mt, GLuint face, GLuint level, GLuint * row_stride, GLuint * image_stride); -void intel_miptree_image_unmap(struct intel_context *intel, +void st_miptree_image_unmap(struct pipe_context *pipe, struct pipe_mipmap_tree *mt); /* Return the linear offset of an image relative to the start of the * tree: */ -GLuint intel_miptree_image_offset(struct pipe_mipmap_tree *mt, - GLuint face, GLuint level); +GLuint st_miptree_image_offset(struct pipe_mipmap_tree *mt, + GLuint face, GLuint level); /* Return pointers to each 2d slice within an image. Indexed by depth * value. */ -const GLuint *intel_miptree_depth_offsets(struct pipe_mipmap_tree *mt, - GLuint level); +const GLuint *st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, + GLuint level); /* Upload an image into a tree */ -void intel_miptree_image_data(struct intel_context *intel, +void st_miptree_image_data(struct pipe_context *pipe, struct pipe_mipmap_tree *dst, GLuint face, GLuint level, @@ -94,7 +94,7 @@ void intel_miptree_image_data(struct intel_context *intel, /* Copy an image between two trees */ -void intel_miptree_image_copy(struct intel_context *intel, +void st_miptree_image_copy(struct pipe_context *pipe, struct pipe_mipmap_tree *dst, GLuint face, GLuint level, struct pipe_mipmap_tree *src); diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.c b/src/mesa/drivers/dri/i915pipe/intel_tex.c index 9e494d21ba..5e9ab1c9ca 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.c @@ -50,7 +50,7 @@ intelDeleteTextureObject(GLcontext *ctx, struct intel_texture_object *intelObj = intel_texture_object(texObj); if (intelObj->mt) - intel_miptree_release(intel, &intelObj->mt); + st_miptree_release(intel->pipe, &intelObj->mt); _mesa_delete_texture_object(ctx, texObj); } @@ -65,7 +65,7 @@ intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) DBG("%s\n", __FUNCTION__); if (intelImage->mt) { - intel_miptree_release(intel, &intelImage->mt); + st_miptree_release(intel->pipe, &intelImage->mt); } if (texImage->Data) { diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c index 0a3f153d56..bf06b5fed5 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c @@ -102,7 +102,7 @@ do_copy_texsubimage(struct intel_context *intel, /* XXX still need the lock ? */ LOCK_HARDWARE(intel); { - GLuint image_offset = intel_miptree_image_offset(intelImage->mt, + GLuint image_offset = st_miptree_image_offset(intelImage->mt, intelImage->face, intelImage->level); const GLint orig_x = x; diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c index 596493f45a..19aa61ae83 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c @@ -125,7 +125,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, assert(!intelObj->mt); if (intelImage->base.IsCompressed) comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); - intelObj->mt = intel_miptree_create(intel, + intelObj->mt = st_miptree_create(intel->pipe, intelObj->base.Target, intelImage->base.InternalFormat, firstLevel, @@ -214,7 +214,7 @@ try_pbo_upload(struct intel_context *intel, else src_stride = width; - dst_offset = intel_miptree_image_offset(intelImage->mt, + dst_offset = st_miptree_image_offset(intelImage->mt, intelImage->face, intelImage->level); @@ -328,7 +328,7 @@ intelTexImage(GLcontext * ctx, * Release any old malloced memory. */ if (intelImage->mt) { - intel_miptree_release(intel, &intelImage->mt); + st_miptree_release(intel->pipe, &intelImage->mt); assert(!texImage->Data); } else if (texImage->Data) { @@ -343,11 +343,11 @@ intelTexImage(GLcontext * ctx, intelObj->mt->first_level == level && intelObj->mt->last_level == level && intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && - !intel_miptree_match_image(intelObj->mt, &intelImage->base, + !st_miptree_match_image(intelObj->mt, &intelImage->base, intelImage->face, intelImage->level)) { DBG("release it\n"); - intel_miptree_release(intel, &intelObj->mt); + st_miptree_release(intel->pipe, &intelObj->mt); assert(!intelObj->mt); } @@ -361,10 +361,10 @@ intelTexImage(GLcontext * ctx, assert(!intelImage->mt); if (intelObj->mt && - intel_miptree_match_image(intelObj->mt, &intelImage->base, + st_miptree_match_image(intelObj->mt, &intelImage->base, intelImage->face, intelImage->level)) { - intel_miptree_reference(&intelImage->mt, intelObj->mt); + st_miptree_reference(&intelImage->mt, intelObj->mt); assert(intelImage->mt); } @@ -438,7 +438,7 @@ intelTexImage(GLcontext * ctx, LOCK_HARDWARE(intel); if (intelImage->mt) { - texImage->Data = intel_miptree_image_map(intel, + texImage->Data = st_miptree_image_map(intel->pipe, intelImage->mt, intelImage->face, intelImage->level, @@ -485,7 +485,7 @@ intelTexImage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, unpack); if (intelImage->mt) { - intel_miptree_image_unmap(intel, intelImage->mt); + st_miptree_image_unmap(intel->pipe, intelImage->mt); texImage->Data = NULL; } @@ -580,7 +580,7 @@ intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, * kernel. Need to explicitly map and unmap it. */ intelImage->base.Data = - intel_miptree_image_map(intel, + st_miptree_image_map(intel->pipe, intelImage->mt, intelImage->face, intelImage->level, @@ -612,7 +612,7 @@ intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, /* Unmap */ if (intelImage->mt) { - intel_miptree_image_unmap(intel, intelImage->mt); + st_miptree_image_unmap(intel->pipe, intelImage->mt); intelImage->base.Data = NULL; } } @@ -653,7 +653,7 @@ intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, return; if (intelObj->mt) - intel_miptree_release(intel, &intelObj->mt); + st_miptree_release(intel->pipe, &intelObj->mt); intelObj->imageOverride = GL_TRUE; intelObj->depthOverride = depth; diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c index d660c2dcee..50ffd0ed1d 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c @@ -76,7 +76,7 @@ intelTexSubimage(GLcontext * ctx, * from uploading the buffer under us. */ if (intelImage->mt) - texImage->Data = intel_miptree_image_map(intel, + texImage->Data = st_miptree_image_map(intel->pipe, intelImage->mt, intelImage->face, intelImage->level, @@ -108,7 +108,7 @@ intelTexSubimage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, packing); if (intelImage->mt) { - intel_miptree_image_unmap(intel, intelImage->mt); + st_miptree_image_unmap(intel->pipe, intelImage->mt); texImage->Data = NULL; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c index 242fe1237b..d366970b26 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c @@ -72,19 +72,19 @@ copy_image_data_to_tree(struct intel_context *intel, if (intelImage->mt) { /* Copy potentially with the blitter: */ - intel_miptree_image_copy(intel, + st_miptree_image_copy(intel->pipe, intelObj->mt, intelImage->face, intelImage->level, intelImage->mt); - intel_miptree_release(intel, &intelImage->mt); + st_miptree_release(intel->pipe, &intelImage->mt); } else { assert(intelImage->base.Data != NULL); /* More straightforward upload. */ - intel_miptree_image_data(intel, + st_miptree_image_data(intel->pipe, intelObj->mt, intelImage->face, intelImage->level, @@ -96,7 +96,7 @@ copy_image_data_to_tree(struct intel_context *intel, intelImage->base.Data = NULL; } - intel_miptree_reference(&intelImage->mt, intelObj->mt); + st_miptree_reference(&intelImage->mt, intelObj->mt); } @@ -130,7 +130,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) */ if (firstImage->base.Border) { if (intelObj->mt) { - intel_miptree_release(intel, &intelObj->mt); + st_miptree_release(intel->pipe, &intelObj->mt); } return GL_FALSE; } @@ -147,9 +147,9 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) firstImage->mt->last_level >= intelObj->lastLevel) { if (intelObj->mt) - intel_miptree_release(intel, &intelObj->mt); + st_miptree_release(intel->pipe, &intelObj->mt); - intel_miptree_reference(&intelObj->mt, firstImage->mt); + st_miptree_reference(&intelObj->mt, firstImage->mt); } if (firstImage->base.IsCompressed) { @@ -177,14 +177,14 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) intelObj->mt->depth0 != firstImage->base.Depth || intelObj->mt->cpp != cpp || intelObj->mt->compressed != firstImage->base.IsCompressed)) { - intel_miptree_release(intel, &intelObj->mt); + st_miptree_release(intel->pipe, &intelObj->mt); } /* May need to create a new tree: */ if (!intelObj->mt) { - intelObj->mt = intel_miptree_create(intel, + intelObj->mt = st_miptree_create(intel->pipe, intelObj->base.Target, firstImage->base.InternalFormat, intelObj->firstLevel, @@ -237,7 +237,7 @@ intel_tex_map_images(struct intel_context *intel, if (intelImage->mt) { intelImage->base.Data = - intel_miptree_image_map(intel, + st_miptree_image_map(intel->pipe, intelImage->mt, intelImage->face, intelImage->level, @@ -266,7 +266,7 @@ intel_tex_unmap_images(struct intel_context *intel, intel_texture_image(intelObj->base.Image[face][i]); if (intelImage->mt) { - intel_miptree_image_unmap(intel, intelImage->mt); + st_miptree_image_unmap(intel->pipe, intelImage->mt); intelImage->base.Data = NULL; } } -- cgit v1.2.3 From 24ddf1169dc2dc2438934957b0fd858042cef9cb Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 12:28:33 -0600 Subject: remove more intel dependencies --- src/mesa/drivers/dri/i915pipe/intel_tex.h | 11 ++++-- src/mesa/drivers/dri/i915pipe/intel_tex_validate.c | 43 ++++++++++++---------- 2 files changed, 31 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.h b/src/mesa/drivers/dri/i915pipe/intel_tex.h index 1ee239402e..d90a80ffa9 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.h +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.h @@ -180,13 +180,18 @@ void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, unsigned long long offset, GLint depth, GLuint pitch); -GLuint intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit); +GLuint intel_finalize_mipmap_tree(GLcontext *ctx, + struct pipe_context *pipe, GLuint unit, + GLboolean *needFlush); -void intel_tex_map_images(struct intel_context *intel, + +#if 0 +void intel_tex_map_images(struct pipe_context *pipe, struct intel_texture_object *intelObj); -void intel_tex_unmap_images(struct intel_context *intel, +void intel_tex_unmap_images(struct pipe_context *pipe, struct intel_texture_object *intelObj); +#endif int intel_compressed_num_bytes(GLuint mesaFormat); diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c index d366970b26..4b06e1803c 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c @@ -1,8 +1,6 @@ #include "mtypes.h" #include "macros.h" -#include "intel_context.h" -#include "intel_batchbuffer.h" #include "intel_mipmap_tree.h" #include "intel_tex.h" @@ -65,26 +63,26 @@ intel_calculate_first_last_level(struct intel_texture_object *intelObj) } static void -copy_image_data_to_tree(struct intel_context *intel, +copy_image_data_to_tree(struct pipe_context *pipe, struct intel_texture_object *intelObj, struct intel_texture_image *intelImage) { if (intelImage->mt) { /* Copy potentially with the blitter: */ - st_miptree_image_copy(intel->pipe, + st_miptree_image_copy(pipe, intelObj->mt, intelImage->face, intelImage->level, intelImage->mt); - st_miptree_release(intel->pipe, &intelImage->mt); + st_miptree_release(pipe, &intelImage->mt); } else { assert(intelImage->base.Data != NULL); /* More straightforward upload. */ - st_miptree_image_data(intel->pipe, + st_miptree_image_data(pipe, intelObj->mt, intelImage->face, intelImage->level, @@ -103,9 +101,11 @@ copy_image_data_to_tree(struct intel_context *intel, /* */ GLuint -intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) +intel_finalize_mipmap_tree(GLcontext *ctx, + struct pipe_context *pipe, GLuint unit, + GLboolean *needFlush) { - struct gl_texture_object *tObj = intel->ctx.Texture.Unit[unit]._Current; + struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; struct intel_texture_object *intelObj = intel_texture_object(tObj); int comp_byte = 0; int cpp; @@ -114,7 +114,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) GLuint nr_faces = 0; struct intel_texture_image *firstImage; - GLboolean need_flush = GL_FALSE; + *needFlush = GL_FALSE; /* We know/require this is true by now: */ @@ -130,7 +130,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) */ if (firstImage->base.Border) { if (intelObj->mt) { - st_miptree_release(intel->pipe, &intelObj->mt); + st_miptree_release(pipe, &intelObj->mt); } return GL_FALSE; } @@ -147,7 +147,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) firstImage->mt->last_level >= intelObj->lastLevel) { if (intelObj->mt) - st_miptree_release(intel->pipe, &intelObj->mt); + st_miptree_release(pipe, &intelObj->mt); st_miptree_reference(&intelObj->mt, firstImage->mt); } @@ -177,14 +177,14 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) intelObj->mt->depth0 != firstImage->base.Depth || intelObj->mt->cpp != cpp || intelObj->mt->compressed != firstImage->base.IsCompressed)) { - st_miptree_release(intel->pipe, &intelObj->mt); + st_miptree_release(pipe, &intelObj->mt); } /* May need to create a new tree: */ if (!intelObj->mt) { - intelObj->mt = st_miptree_create(intel->pipe, + intelObj->mt = st_miptree_create(pipe, intelObj->base.Target, firstImage->base.InternalFormat, intelObj->firstLevel, @@ -207,22 +207,24 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) /* Need to import images in main memory or held in other trees. */ if (intelObj->mt != intelImage->mt) { - copy_image_data_to_tree(intel, intelObj, intelImage); - need_flush = GL_TRUE; + copy_image_data_to_tree(pipe, intelObj, intelImage); + *needFlush = GL_TRUE; } } } + /** if (need_flush) intel_batchbuffer_flush(intel->batch); + **/ return GL_TRUE; } - +#if 0 /* unused? */ void -intel_tex_map_images(struct intel_context *intel, +intel_tex_map_images(struct pipe_context *pipe, struct intel_texture_object *intelObj) { GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; @@ -237,7 +239,7 @@ intel_tex_map_images(struct intel_context *intel, if (intelImage->mt) { intelImage->base.Data = - st_miptree_image_map(intel->pipe, + st_miptree_image_map(pipe, intelImage->mt, intelImage->face, intelImage->level, @@ -254,7 +256,7 @@ intel_tex_map_images(struct intel_context *intel, void -intel_tex_unmap_images(struct intel_context *intel, +intel_tex_unmap_images(struct pipe_context *pipe, struct intel_texture_object *intelObj) { GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; @@ -266,9 +268,10 @@ intel_tex_unmap_images(struct intel_context *intel, intel_texture_image(intelObj->base.Image[face][i]); if (intelImage->mt) { - st_miptree_image_unmap(intel->pipe, intelImage->mt); + st_miptree_image_unmap(pipe, intelImage->mt); intelImage->base.Data = NULL; } } } } +#endif -- cgit v1.2.3 From bdd3ccf95982e4043f2b57b42fd5d63ff8de4b08 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 12:32:16 -0600 Subject: More code movement, removal of intel dependencies --- src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h | 6 ++++-- src/mesa/drivers/dri/intel/intel_tex_layout.c | 23 +++++++++++++++++++++-- src/mesa/drivers/dri/intel/intel_tex_layout.h | 4 ++++ 4 files changed, 30 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c index 6a91ddbe5e..1a9aa10115 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c @@ -32,7 +32,7 @@ #include "pipe/p_context.h" -#define FILE_DEBUG_FLAG DEBUG_MIPTREE +#define DBG if(0) printf static GLenum target_to_target(GLenum target) diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h index 59f1bfcab4..eafe47c029 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h +++ b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h @@ -28,9 +28,11 @@ #ifndef INTEL_MIPMAP_TREE_H #define INTEL_MIPMAP_TREE_H -#include "intel_context.h" -#include "main/glheader.h" +#include "main/mtypes.h" + +struct pipe_context; +struct pipe_mipmap_tree; struct pipe_region; diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c index bedd835cbb..61b8fb77c5 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.c +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c @@ -32,6 +32,7 @@ #include "macros.h" #include "pipe/p_state.h" +#include "intel_tex_layout.h" #include "intel_mipmap_tree.h" @@ -47,7 +48,7 @@ static int align(int value, int alignment) static void -mipmaptree_set_level_info(struct pipe_mipmap_tree *mt, +intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, GLuint level, GLuint nr_images, GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) @@ -79,6 +80,24 @@ mipmaptree_set_level_info(struct pipe_mipmap_tree *mt, } +static void +intel_miptree_set_image_offset(struct pipe_mipmap_tree *mt, + GLuint level, GLuint img, GLuint x, GLuint y) +{ + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + + assert(img < mt->level[level].nr_images); + + mt->level[level].image_offset[img] = (x + y * mt->pitch); + + /* + DBG("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); + */ +} + + static void i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) { @@ -113,7 +132,7 @@ i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { GLuint img_height; - mipmaptree_set_level_info(mt, level, 1, x, y, width, height, 1); + intel_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); if (mt->compressed) img_height = MAX2(1, height/4); diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h index cbd4b8e27f..ed30e2ef76 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.h +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h @@ -33,6 +33,10 @@ #include "macros.h" +struct pipe_context; +struct pipe_mipmap_tree; + + extern void i915_miptree_layout_2d( struct pipe_context *, struct pipe_mipmap_tree *mt ); extern void i945_miptree_layout_2d( struct pipe_context *, struct pipe_mipmap_tree *mt ); -- cgit v1.2.3 From 307fe0702edb8fb79bc8f7f8830af440f2779a10 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 12:40:51 -0600 Subject: Lift intel_mipmap_tree.c to st_mipmap_tree.c --- src/mesa/drivers/dri/i915pipe/Makefile | 1 - src/mesa/drivers/dri/i915pipe/intel_fbo.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_tex.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_tex_validate.c | 2 +- src/mesa/drivers/dri/intel/intel_tex_layout.c | 2 +- src/mesa/sources | 1 + src/mesa/state_tracker/st_mipmap_tree.c | 293 +++++++++++++++++++++ src/mesa/state_tracker/st_mipmap_tree.h | 105 ++++++++ 11 files changed, 406 insertions(+), 8 deletions(-) create mode 100644 src/mesa/state_tracker/st_mipmap_tree.c create mode 100644 src/mesa/state_tracker/st_mipmap_tree.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile index 4e8d020196..f9403abd73 100644 --- a/src/mesa/drivers/dri/i915pipe/Makefile +++ b/src/mesa/drivers/dri/i915pipe/Makefile @@ -13,7 +13,6 @@ DRIVER_SOURCES = \ intel_softpipe.c \ intel_buffer_objects.c \ intel_batchbuffer.c \ - intel_mipmap_tree.c \ intel_tex_layout.c \ intel_tex_image.c \ intel_tex_subimage.c \ diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index fe5b1a37df..061c6bfa1f 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -39,7 +39,7 @@ #include "intel_buffers.h" #include "intel_depthstencil.h" #include "intel_fbo.h" -#include "intel_mipmap_tree.h" +#include "state_tracker/st_mipmap_tree.h" #include "intel_tex.h" #include "pipe/p_context.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.c b/src/mesa/drivers/dri/i915pipe/intel_tex.c index 5e9ab1c9ca..c2f5727e6b 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.c @@ -1,6 +1,6 @@ #include "texobj.h" #include "intel_context.h" -#include "intel_mipmap_tree.h" +#include "state_tracker/st_mipmap_tree.h" #include "intel_tex.h" #define FILE_DEBUG_FLAG DEBUG_TEXTURE diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c index bf06b5fed5..88cdee4583 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c @@ -35,7 +35,7 @@ #include "intel_context.h" #include "intel_batchbuffer.h" #include "intel_buffers.h" -#include "intel_mipmap_tree.h" +#include "state_tracker/st_mipmap_tree.h" #include "intel_fbo.h" #include "intel_tex.h" #include "intel_blit.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c index 19aa61ae83..ea7b26b7f1 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c @@ -18,7 +18,7 @@ #include "intel_context.h" #include "intel_buffer_objects.h" #include "intel_batchbuffer.h" -#include "intel_mipmap_tree.h" +#include "state_tracker/st_mipmap_tree.h" #include "intel_tex.h" #include "intel_ioctl.h" #include "intel_blit.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c index 50ffd0ed1d..27b56e907d 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c @@ -33,7 +33,7 @@ #include "intel_context.h" #include "intel_tex.h" -#include "intel_mipmap_tree.h" +#include "state_tracker/st_mipmap_tree.h" #include "pipe/p_context.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c index 4b06e1803c..2083a2434d 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c @@ -1,7 +1,7 @@ #include "mtypes.h" #include "macros.h" -#include "intel_mipmap_tree.h" +#include "state_tracker/st_mipmap_tree.h" #include "intel_tex.h" #include "pipe/p_state.h" diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c index 61b8fb77c5..433a2c5f47 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.c +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c @@ -33,7 +33,7 @@ #include "macros.h" #include "pipe/p_state.h" #include "intel_tex_layout.h" -#include "intel_mipmap_tree.h" +#include "state_tracker/st_mipmap_tree.h" static GLuint minify( GLuint d ) diff --git a/src/mesa/sources b/src/mesa/sources index bd1bbd4299..bef9929f34 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -199,6 +199,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_draw.c \ state_tracker/st_context.c \ + state_tracker/st_mipmap_tree.c \ state_tracker/st_texobj.c SHADER_SOURCES = \ diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c new file mode 100644 index 0000000000..1a9aa10115 --- /dev/null +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -0,0 +1,293 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_mipmap_tree.h" +#include "enums.h" + +#include "pipe/p_state.h" +#include "pipe/p_context.h" + + +#define DBG if(0) printf + +static GLenum +target_to_target(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + return GL_TEXTURE_CUBE_MAP_ARB; + default: + return target; + } +} + +struct pipe_mipmap_tree * +st_miptree_create(struct pipe_context *pipe, + GLenum target, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, GLuint cpp, GLuint compress_byte) +{ + GLboolean ok; + struct pipe_mipmap_tree *mt = calloc(sizeof(*mt), 1); + + DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), + _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); + + mt->target = target_to_target(target); + mt->internal_format = internal_format; + mt->first_level = first_level; + mt->last_level = last_level; + mt->width0 = width0; + mt->height0 = height0; + mt->depth0 = depth0; + mt->cpp = compress_byte ? compress_byte : cpp; + mt->compressed = compress_byte ? 1 : 0; + mt->refcount = 1; + + ok = pipe->mipmap_tree_layout(pipe, mt); + if (ok) + mt->region = pipe->region_alloc(pipe, + mt->cpp, mt->pitch, mt->total_height); + + if (!mt->region) { + free(mt); + return NULL; + } + + return mt; +} + + +void +st_miptree_reference(struct pipe_mipmap_tree **dst, + struct pipe_mipmap_tree *src) +{ + src->refcount++; + *dst = src; + DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); +} + +void +st_miptree_release(struct pipe_context *pipe, + struct pipe_mipmap_tree **mt) +{ + if (!*mt) + return; + + DBG("%s %p refcount will be %d\n", __FUNCTION__, *mt, (*mt)->refcount - 1); + if (--(*mt)->refcount <= 0) { + GLuint i; + + DBG("%s deleting %p\n", __FUNCTION__, *mt); + + pipe->region_release(pipe, &((*mt)->region)); + + for (i = 0; i < MAX_TEXTURE_LEVELS; i++) + if ((*mt)->level[i].image_offset) + free((*mt)->level[i].image_offset); + + free(*mt); + } + *mt = NULL; +} + + + + +/* Can the image be pulled into a unified mipmap tree. This mirrors + * the completeness test in a lot of ways. + * + * Not sure whether I want to pass gl_texture_image here. + */ +GLboolean +st_miptree_match_image(struct pipe_mipmap_tree *mt, + struct gl_texture_image *image, + GLuint face, GLuint level) +{ + /* Images with borders are never pulled into mipmap trees. + */ + if (image->Border) + return GL_FALSE; + + if (image->InternalFormat != mt->internal_format || + image->IsCompressed != mt->compressed) + return GL_FALSE; + + /* Test image dimensions against the base level image adjusted for + * minification. This will also catch images not present in the + * tree, changed targets, etc. + */ + if (image->Width != mt->level[level].width || + image->Height != mt->level[level].height || + image->Depth != mt->level[level].depth) + return GL_FALSE; + + return GL_TRUE; +} + + +/* Although we use the image_offset[] array to store relative offsets + * to cube faces, Mesa doesn't know anything about this and expects + * each cube face to be treated as a separate image. + * + * These functions present that view to mesa: + */ +const GLuint * +st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) +{ + static const GLuint zero = 0; + + if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) + return &zero; + else + return mt->level[level].image_offset; +} + + +GLuint +st_miptree_image_offset(struct pipe_mipmap_tree * mt, + GLuint face, GLuint level) +{ + if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) + return (mt->level[level].level_offset + + mt->level[level].image_offset[face] * mt->cpp); + else + return mt->level[level].level_offset; +} + + + +/** + * Map a teximage in a mipmap tree. + * \param row_stride returns row stride in bytes + * \param image_stride returns image stride in bytes (for 3D textures). + * \return address of mapping + */ +GLubyte * +st_miptree_image_map(struct pipe_context *pipe, + struct pipe_mipmap_tree * mt, + GLuint face, + GLuint level, + GLuint * row_stride, GLuint * image_offsets) +{ + GLubyte *ptr; + DBG("%s \n", __FUNCTION__); + + if (row_stride) + *row_stride = mt->pitch * mt->cpp; + + if (image_offsets) + memcpy(image_offsets, mt->level[level].image_offset, + mt->level[level].depth * sizeof(GLuint)); + + ptr = pipe->region_map(pipe, mt->region); + + return ptr + st_miptree_image_offset(mt, face, level); +} + +void +st_miptree_image_unmap(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt) +{ + DBG("%s\n", __FUNCTION__); + pipe->region_unmap(pipe, mt->region); +} + + + +/* Upload data for a particular image. + */ +void +st_miptree_image_data(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch) +{ + GLuint depth = dst->level[level].depth; + GLuint dst_offset = st_miptree_image_offset(dst, face, level); + const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); + GLuint i; + GLuint height = 0; + + DBG("%s\n", __FUNCTION__); + for (i = 0; i < depth; i++) { + height = dst->level[level].height; + if(dst->compressed) + height /= 4; + pipe->region_data(pipe, dst->region, + dst_offset + dst_depth_offset[i], /* dst_offset */ + 0, 0, /* dstx, dsty */ + src, + src_row_pitch, + 0, 0, /* source x, y */ + dst->level[level].width, height); /* width, height */ + + src += src_image_pitch * dst->cpp; + } +} + +/* Copy mipmap image between trees + */ +void +st_miptree_image_copy(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, GLuint level, + struct pipe_mipmap_tree *src) +{ + GLuint width = src->level[level].width; + GLuint height = src->level[level].height; + GLuint depth = src->level[level].depth; + GLuint dst_offset = st_miptree_image_offset(dst, face, level); + GLuint src_offset = st_miptree_image_offset(src, face, level); + const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); + const GLuint *src_depth_offset = st_miptree_depth_offsets(src, level); + GLuint i; + + if (dst->compressed) + height /= 4; + for (i = 0; i < depth; i++) { + pipe->region_copy(pipe, + dst->region, dst_offset + dst_depth_offset[i], + 0, + 0, + src->region, src_offset + src_depth_offset[i], + 0, 0, width, height); + } + +} diff --git a/src/mesa/state_tracker/st_mipmap_tree.h b/src/mesa/state_tracker/st_mipmap_tree.h new file mode 100644 index 0000000000..eafe47c029 --- /dev/null +++ b/src/mesa/state_tracker/st_mipmap_tree.h @@ -0,0 +1,105 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_MIPMAP_TREE_H +#define INTEL_MIPMAP_TREE_H + + +#include "main/mtypes.h" + +struct pipe_context; +struct pipe_mipmap_tree; +struct pipe_region; + + +struct pipe_mipmap_tree *st_miptree_create(struct pipe_context *pipe, + GLenum target, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, + GLuint cpp, + GLuint compress_byte); + +void st_miptree_reference(struct pipe_mipmap_tree **dst, + struct pipe_mipmap_tree *src); + +void st_miptree_release(struct pipe_context *pipe, + struct pipe_mipmap_tree **mt); + +/* Check if an image fits an existing mipmap tree layout + */ +GLboolean st_miptree_match_image(struct pipe_mipmap_tree *mt, + struct gl_texture_image *image, + GLuint face, GLuint level); + +/* Return a pointer to an image within a tree. Return image stride as + * well. + */ +GLubyte *st_miptree_image_map(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt, + GLuint face, + GLuint level, + GLuint * row_stride, GLuint * image_stride); + +void st_miptree_image_unmap(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt); + + +/* Return the linear offset of an image relative to the start of the + * tree: + */ +GLuint st_miptree_image_offset(struct pipe_mipmap_tree *mt, + GLuint face, GLuint level); + +/* Return pointers to each 2d slice within an image. Indexed by depth + * value. + */ +const GLuint *st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, + GLuint level); + + +/* Upload an image into a tree + */ +void st_miptree_image_data(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch); + +/* Copy an image between two trees + */ +void st_miptree_image_copy(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, GLuint level, + struct pipe_mipmap_tree *src); + + +#endif -- cgit v1.2.3 From bc3b608317fcda659e7d910b4fab06426ae5f334 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 12:41:39 -0600 Subject: minor renaming --- src/mesa/state_tracker/st_mipmap_tree.c | 4 ++-- src/mesa/state_tracker/st_mipmap_tree.h | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index 1a9aa10115..e99cd987e8 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "intel_mipmap_tree.h" +#include "st_mipmap_tree.h" #include "enums.h" #include "pipe/p_state.h" diff --git a/src/mesa/state_tracker/st_mipmap_tree.h b/src/mesa/state_tracker/st_mipmap_tree.h index eafe47c029..ea3b52ca42 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.h +++ b/src/mesa/state_tracker/st_mipmap_tree.h @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -25,8 +25,8 @@ * **************************************************************************/ -#ifndef INTEL_MIPMAP_TREE_H -#define INTEL_MIPMAP_TREE_H +#ifndef ST_MIPMAP_TREE_H +#define ST_MIPMAP_TREE_H #include "main/mtypes.h" -- cgit v1.2.3 From b9f4b0ce4e4465f149b64a0188e304f73dfa6e94 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 13:12:24 -0600 Subject: Replace intel_texture_object with st_texture_object, intel_texture_image with st_texture_image --- src/mesa/drivers/dri/i915pipe/intel_fbo.c | 14 ++++----- src/mesa/drivers/dri/i915pipe/intel_tex.c | 24 +++++++-------- src/mesa/drivers/dri/i915pipe/intel_tex.h | 24 +++++++-------- src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 10 +++--- src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 36 +++++++++++----------- src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c | 2 +- src/mesa/drivers/dri/i915pipe/intel_tex_validate.c | 28 ++++++++--------- 7 files changed, 69 insertions(+), 69 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index 061c6bfa1f..2f1aa00038 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -532,7 +532,7 @@ intel_render_texture(GLcontext * ctx, struct gl_texture_image *newImage = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); - struct intel_texture_image *intel_image; + struct st_texture_image *st_image; GLuint imageOffset; (void) fb; @@ -558,26 +558,26 @@ intel_render_texture(GLcontext * ctx, irb->Base.RefCount); /* point the renderbufer's region to the texture image region */ - intel_image = intel_texture_image(newImage); - if (irb->region != intel_image->mt->region) { + st_image = st_texture_image(newImage); + if (irb->region != st_image->mt->region) { if (irb->region) intel->pipe->region_release(intel->pipe, &irb->region); - pipe_region_reference(&irb->region, intel_image->mt->region); + pipe_region_reference(&irb->region, st_image->mt->region); } /* compute offset of the particular 2D image within the texture region */ - imageOffset = st_miptree_image_offset(intel_image->mt, + imageOffset = st_miptree_image_offset(st_image->mt, att->CubeMapFace, att->TextureLevel); if (att->Texture->Target == GL_TEXTURE_3D) { - const GLuint *offsets = st_miptree_depth_offsets(intel_image->mt, + const GLuint *offsets = st_miptree_depth_offsets(st_image->mt, att->TextureLevel); imageOffset += offsets[att->Zoffset]; } /* store that offset in the region */ - intel_image->mt->region->draw_offset = imageOffset; + st_image->mt->region->draw_offset = imageOffset; /* update drawing region, etc */ intel_draw_buffer(ctx, fb); diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.c b/src/mesa/drivers/dri/i915pipe/intel_tex.c index c2f5727e6b..94583cba97 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.c @@ -10,12 +10,12 @@ intelIsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) { #if 0 struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); + struct st_texture_object *stObj = st_texture_object(texObj); return - intelObj->mt && - intelObj->mt->region && - intel_is_region_resident(intel, intelObj->mt->region); + stObj->mt && + stObj->mt->region && + intel_is_region_resident(intel, stObj->mt->region); #endif return 1; } @@ -27,14 +27,14 @@ intelNewTextureImage(GLcontext * ctx) { DBG("%s\n", __FUNCTION__); (void) ctx; - return (struct gl_texture_image *) CALLOC_STRUCT(intel_texture_image); + return (struct gl_texture_image *) CALLOC_STRUCT(st_texture_image); } static struct gl_texture_object * intelNewTextureObject(GLcontext * ctx, GLuint name, GLenum target) { - struct intel_texture_object *obj = CALLOC_STRUCT(intel_texture_object); + struct st_texture_object *obj = CALLOC_STRUCT(st_texture_object); DBG("%s\n", __FUNCTION__); _mesa_initialize_texture_object(&obj->base, name, target); @@ -47,10 +47,10 @@ intelDeleteTextureObject(GLcontext *ctx, struct gl_texture_object *texObj) { struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); + struct st_texture_object *stObj = st_texture_object(texObj); - if (intelObj->mt) - st_miptree_release(intel->pipe, &intelObj->mt); + if (stObj->mt) + st_miptree_release(intel->pipe, &stObj->mt); _mesa_delete_texture_object(ctx, texObj); } @@ -60,12 +60,12 @@ static void intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) { struct intel_context *intel = intel_context(ctx); - struct intel_texture_image *intelImage = intel_texture_image(texImage); + struct st_texture_image *stImage = st_texture_image(texImage); DBG("%s\n", __FUNCTION__); - if (intelImage->mt) { - st_miptree_release(intel->pipe, &intelImage->mt); + if (stImage->mt) { + st_miptree_release(intel->pipe, &stImage->mt); } if (texImage->Data) { diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.h b/src/mesa/drivers/dri/i915pipe/intel_tex.h index d90a80ffa9..7a1cc91974 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.h +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.h @@ -32,7 +32,7 @@ #include "intel_context.h" #include "texmem.h" -struct intel_texture_object +struct st_texture_object { struct gl_texture_object base; /* The "parent" object */ @@ -58,7 +58,7 @@ struct intel_texture_object -struct intel_texture_image +struct st_texture_image { struct gl_texture_image base; @@ -67,8 +67,8 @@ struct intel_texture_image GLuint level; GLuint face; - /* If intelImage->mt != NULL, image data is stored here. - * Else if intelImage->base.Data != NULL, image is stored there. + /* If stImage->mt != NULL, image data is stored here. + * Else if stImage->base.Data != NULL, image is stored there. * Else there is no image data. */ struct pipe_mipmap_tree *mt; @@ -187,25 +187,25 @@ GLuint intel_finalize_mipmap_tree(GLcontext *ctx, #if 0 void intel_tex_map_images(struct pipe_context *pipe, - struct intel_texture_object *intelObj); + struct st_texture_object *stObj); void intel_tex_unmap_images(struct pipe_context *pipe, - struct intel_texture_object *intelObj); + struct st_texture_object *stObj); #endif int intel_compressed_num_bytes(GLuint mesaFormat); -static INLINE struct intel_texture_object * -intel_texture_object(struct gl_texture_object *obj) +static INLINE struct st_texture_object * +st_texture_object(struct gl_texture_object *obj) { - return (struct intel_texture_object *) obj; + return (struct st_texture_object *) obj; } -static INLINE struct intel_texture_image * -intel_texture_image(struct gl_texture_image *img) +static INLINE struct st_texture_image * +st_texture_image(struct gl_texture_image *img) { - return (struct intel_texture_image *) img; + return (struct st_texture_image *) img; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c index 88cdee4583..b62aaaeebb 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c @@ -84,7 +84,7 @@ get_teximage_source(struct intel_context *intel, GLenum internalFormat) static GLboolean do_copy_texsubimage(struct intel_context *intel, - struct intel_texture_image *intelImage, + struct st_texture_image *intelImage, GLenum internalFormat, GLint dstx, GLint dsty, GLint x, GLint y, GLsizei width, GLsizei height) @@ -186,7 +186,7 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, &ctx->DefaultPacking, texObj, texImage); if (!do_copy_texsubimage(intel_context(ctx), - intel_texture_image(texImage), + st_texture_image(texImage), internalFormat, 0, 0, x, y, width, 1)) goto fail; @@ -223,7 +223,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, if (!do_copy_texsubimage(intel_context(ctx), - intel_texture_image(texImage), + st_texture_image(texImage), internalFormat, 0, 0, x, y, width, height)) goto fail; @@ -256,7 +256,7 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, */ if (!do_copy_texsubimage(intel_context(ctx), - intel_texture_image(texImage), + st_texture_image(texImage), internalFormat, xoffset, 0, x, y, width, 1)) { #if 0 _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); @@ -285,7 +285,7 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, */ if (!do_copy_texsubimage(intel_context(ctx), - intel_texture_image(texImage), + st_texture_image(texImage), internalFormat, xoffset, yoffset, x, y, width, height)) { #if 0 diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c index ea7b26b7f1..2825df37f6 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c @@ -61,9 +61,9 @@ logbase2(int n) * saving is worth it. */ static void -guess_and_alloc_mipmap_tree(struct intel_context *intel, - struct intel_texture_object *intelObj, - struct intel_texture_image *intelImage) +guess_and_alloc_mipmap_tree(struct pipe_context *pipe, + struct st_texture_object *intelObj, + struct st_texture_image *intelImage) { GLuint firstLevel; GLuint lastLevel; @@ -125,7 +125,7 @@ guess_and_alloc_mipmap_tree(struct intel_context *intel, assert(!intelObj->mt); if (intelImage->base.IsCompressed) comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); - intelObj->mt = st_miptree_create(intel->pipe, + intelObj->mt = st_miptree_create(pipe, intelObj->base.Target, intelImage->base.InternalFormat, firstLevel, @@ -190,7 +190,7 @@ check_pbo_format(GLint internalFormat, */ static GLboolean try_pbo_upload(struct intel_context *intel, - struct intel_texture_image *intelImage, + struct st_texture_image *intelImage, const struct gl_pixelstore_attrib *unpack, GLint internalFormat, GLint width, GLint height, @@ -250,7 +250,7 @@ try_pbo_upload(struct intel_context *intel, static GLboolean try_pbo_zcopy(struct intel_context *intel, - struct intel_texture_image *intelImage, + struct st_texture_image *intelImage, const struct gl_pixelstore_attrib *unpack, GLint internalFormat, GLint width, GLint height, @@ -277,8 +277,8 @@ intelTexImage(GLcontext * ctx, struct gl_texture_image *texImage, GLsizei imageSize, int compressed) { struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); - struct intel_texture_image *intelImage = intel_texture_image(texImage); + struct st_texture_object *intelObj = st_texture_object(texObj); + struct st_texture_image *intelImage = st_texture_image(texImage); GLint postConvWidth = width; GLint postConvHeight = height; GLint texelBytes, sizeInBytes; @@ -352,7 +352,7 @@ intelTexImage(GLcontext * ctx, } if (!intelObj->mt) { - guess_and_alloc_mipmap_tree(intel, intelObj, intelImage); + guess_and_alloc_mipmap_tree(intel->pipe, intelObj, intelImage); if (!intelObj->mt) { DBG("guess_and_alloc_mipmap_tree: failed\n"); } @@ -572,7 +572,7 @@ intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, struct gl_texture_image *texImage, int compressed) { struct intel_context *intel = intel_context(ctx); - struct intel_texture_image *intelImage = intel_texture_image(texImage); + struct st_texture_image *intelImage = st_texture_image(texImage); /* Map */ if (intelImage->mt) { @@ -647,18 +647,18 @@ intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, struct intel_context *intel = (struct intel_context*) ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct intel_texture_object *intelObj = intel_texture_object(tObj); + struct st_texture_object *stObj = st_texture_object(tObj); - if (!intelObj) + if (!stObj) return; - if (intelObj->mt) - st_miptree_release(intel->pipe, &intelObj->mt); + if (stObj->mt) + st_miptree_release(intel->pipe, &stObj->mt); - intelObj->imageOverride = GL_TRUE; - intelObj->depthOverride = depth; - intelObj->pitchOverride = pitch; + stObj->imageOverride = GL_TRUE; + stObj->depthOverride = depth; + stObj->pitchOverride = pitch; if (offset) - intelObj->textureOffset = offset; + stObj->textureOffset = offset; } diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c index 27b56e907d..c5aab96ea7 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c @@ -52,7 +52,7 @@ intelTexSubimage(GLcontext * ctx, struct gl_texture_image *texImage) { struct intel_context *intel = intel_context(ctx); - struct intel_texture_image *intelImage = intel_texture_image(texImage); + struct st_texture_image *intelImage = st_texture_image(texImage); GLuint dstRowStride; DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c index 2083a2434d..d0631a88f1 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c @@ -14,7 +14,7 @@ * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. */ static void -intel_calculate_first_last_level(struct intel_texture_object *intelObj) +intel_calculate_first_last_level(struct st_texture_object *intelObj) { struct gl_texture_object *tObj = &intelObj->base; const struct gl_texture_image *const baseImage = @@ -64,8 +64,8 @@ intel_calculate_first_last_level(struct intel_texture_object *intelObj) static void copy_image_data_to_tree(struct pipe_context *pipe, - struct intel_texture_object *intelObj, - struct intel_texture_image *intelImage) + struct st_texture_object *intelObj, + struct st_texture_image *intelImage) { if (intelImage->mt) { /* Copy potentially with the blitter: @@ -106,13 +106,13 @@ intel_finalize_mipmap_tree(GLcontext *ctx, GLboolean *needFlush) { struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; - struct intel_texture_object *intelObj = intel_texture_object(tObj); + struct st_texture_object *intelObj = st_texture_object(tObj); int comp_byte = 0; int cpp; GLuint face, i; GLuint nr_faces = 0; - struct intel_texture_image *firstImage; + struct st_texture_image *firstImage; *needFlush = GL_FALSE; @@ -124,7 +124,7 @@ intel_finalize_mipmap_tree(GLcontext *ctx, */ intel_calculate_first_last_level(intelObj); firstImage = - intel_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); + st_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); /* Fallback case: */ @@ -201,8 +201,8 @@ intel_finalize_mipmap_tree(GLcontext *ctx, nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; for (face = 0; face < nr_faces; face++) { for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct intel_texture_image *intelImage = - intel_texture_image(intelObj->base.Image[face][i]); + struct st_texture_image *intelImage = + st_texture_image(intelObj->base.Image[face][i]); /* Need to import images in main memory or held in other trees. */ @@ -225,7 +225,7 @@ intel_finalize_mipmap_tree(GLcontext *ctx, #if 0 /* unused? */ void intel_tex_map_images(struct pipe_context *pipe, - struct intel_texture_object *intelObj) + struct st_texture_object *intelObj) { GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; GLuint face, i; @@ -234,8 +234,8 @@ intel_tex_map_images(struct pipe_context *pipe, for (face = 0; face < nr_faces; face++) { for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct intel_texture_image *intelImage = - intel_texture_image(intelObj->base.Image[face][i]); + struct st_texture_image *intelImage = + st_texture_image(intelObj->base.Image[face][i]); if (intelImage->mt) { intelImage->base.Data = @@ -257,15 +257,15 @@ intel_tex_map_images(struct pipe_context *pipe, void intel_tex_unmap_images(struct pipe_context *pipe, - struct intel_texture_object *intelObj) + struct st_texture_object *intelObj) { GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; GLuint face, i; for (face = 0; face < nr_faces; face++) { for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct intel_texture_image *intelImage = - intel_texture_image(intelObj->base.Image[face][i]); + struct st_texture_image *intelImage = + st_texture_image(intelObj->base.Image[face][i]); if (intelImage->mt) { st_miptree_image_unmap(pipe, intelImage->mt); -- cgit v1.2.3 From e3ad58975be8599ef557988ffd18ca0236dd58ec Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 13:17:38 -0600 Subject: remove timing code --- src/mesa/drivers/dri/i915pipe/intel_tex.c | 65 +------------------------------ 1 file changed, 1 insertion(+), 64 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.c b/src/mesa/drivers/dri/i915pipe/intel_tex.c index 94583cba97..f87d342366 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.c @@ -126,64 +126,6 @@ do_memcpy(void *dest, const void *src, size_t n) } -#if DO_DEBUG - -#ifndef __x86_64__ -static unsigned -fastrdtsc(void) -{ - unsigned eax; - __asm__ volatile ("\t" - "pushl %%ebx\n\t" - "cpuid\n\t" ".byte 0x0f, 0x31\n\t" - "popl %%ebx\n":"=a" (eax) - :"0"(0) - :"ecx", "edx", "cc"); - - return eax; -} -#else -static unsigned -fastrdtsc(void) -{ - unsigned eax; - __asm__ volatile ("\t" "cpuid\n\t" ".byte 0x0f, 0x31\n\t":"=a" (eax) - :"0"(0) - :"ecx", "edx", "ebx", "cc"); - - return eax; -} -#endif - -static unsigned -time_diff(unsigned t, unsigned t2) -{ - return ((t < t2) ? t2 - t : 0xFFFFFFFFU - (t - t2 - 1)); -} - - -static void * -timed_memcpy(void *dest, const void *src, size_t n) -{ - void *ret; - unsigned t1, t2; - double rate; - - if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) - _mesa_printf("Warning - non-aligned texture copy!\n"); - - t1 = fastrdtsc(); - ret = do_memcpy(dest, src, n); - t2 = fastrdtsc(); - - rate = time_diff(t1, t2); - rate /= (double) n; - _mesa_printf("timed_memcpy: %u %u --> %f clocks/byte\n", t1, t2, rate); - return ret; -} -#endif /* DO_DEBUG */ - - void intelInitTextureFuncs(struct dd_function_table *functions) { @@ -211,10 +153,5 @@ intelInitTextureFuncs(struct dd_function_table *functions) functions->UpdateTexturePalette = 0; functions->IsTextureResident = intelIsTextureResident; -#if DO_DEBUG - if (INTEL_DEBUG & DEBUG_BUFMGR) - functions->TextureMemCpy = timed_memcpy; - else -#endif - functions->TextureMemCpy = do_memcpy; + functions->TextureMemCpy = do_memcpy; } -- cgit v1.2.3 From f42feca07569213e24a09f640311b93937d0bd9e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 13:24:54 -0600 Subject: disable some PBO code, remove intel dependencies --- src/mesa/drivers/dri/i915pipe/intel_tex.c | 27 ++++++++++++++ src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 49 ++++++++++++++++++++----- 2 files changed, 67 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.c b/src/mesa/drivers/dri/i915pipe/intel_tex.c index f87d342366..0990c499b5 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.c @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #include "texobj.h" #include "intel_context.h" #include "state_tracker/st_mipmap_tree.h" diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c index 2825df37f6..93a59b6896 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c @@ -1,6 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ -#include -#include #include "glheader.h" #include "macros.h" @@ -15,15 +39,12 @@ #include "texobj.h" #include "texstore.h" -#include "intel_context.h" -#include "intel_buffer_objects.h" -#include "intel_batchbuffer.h" +#include "pipe/p_context.h" #include "state_tracker/st_mipmap_tree.h" + #include "intel_tex.h" -#include "intel_ioctl.h" -#include "intel_blit.h" -#include "pipe/p_context.h" + #define FILE_DEBUG_FLAG DEBUG_TEXTURE @@ -158,6 +179,8 @@ target_to_face(GLenum target) } } + + /* There are actually quite a few combinations this will work for, * more than what I've listed here. */ @@ -196,6 +219,8 @@ try_pbo_upload(struct intel_context *intel, GLint width, GLint height, GLenum format, GLenum type, const void *pixels) { + return GL_FALSE; /* XXX fix flushing/locking/blitting below */ +#if 000 struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); GLuint src_offset, src_stride; GLuint dst_offset, dst_stride; @@ -244,6 +269,7 @@ try_pbo_upload(struct intel_context *intel, UNLOCK_HARDWARE(intel); return GL_TRUE; +#endif } @@ -371,6 +397,7 @@ intelTexImage(GLcontext * ctx, if (!intelImage->mt) DBG("XXX: Image did not fit into tree - storing in local memory!\n"); +#if 0 /* XXX FIX when st_buffer_objects are in place */ /* PBO fastpaths: */ if (dims <= 2 && @@ -412,7 +439,11 @@ intelTexImage(GLcontext * ctx, DBG("pbo upload failed\n"); } - +#else + (void) try_pbo_upload; + (void) check_pbo_format; + (void) try_pbo_zcopy; +#endif /* intelCopyTexImage calls this function with pixels == NULL, with -- cgit v1.2.3 From d49cee1a748c513325249a4cbaa982a60897fa77 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 13:31:23 -0600 Subject: more intel_context removal --- src/mesa/drivers/dri/i915pipe/intel_tex.c | 12 ++++++------ src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c | 13 +++++++++---- 2 files changed, 15 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.c b/src/mesa/drivers/dri/i915pipe/intel_tex.c index 0990c499b5..5270094cc5 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex.c @@ -25,8 +25,8 @@ * **************************************************************************/ -#include "texobj.h" -#include "intel_context.h" +#include "main/texobj.h" +#include "state_tracker/st_context.h" #include "state_tracker/st_mipmap_tree.h" #include "intel_tex.h" @@ -73,11 +73,11 @@ static void intelDeleteTextureObject(GLcontext *ctx, struct gl_texture_object *texObj) { - struct intel_context *intel = intel_context(ctx); + struct pipe_context *pipe = ctx->st->pipe; struct st_texture_object *stObj = st_texture_object(texObj); if (stObj->mt) - st_miptree_release(intel->pipe, &stObj->mt); + st_miptree_release(pipe, &stObj->mt); _mesa_delete_texture_object(ctx, texObj); } @@ -86,13 +86,13 @@ intelDeleteTextureObject(GLcontext *ctx, static void intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) { - struct intel_context *intel = intel_context(ctx); + struct pipe_context *pipe = ctx->st->pipe; struct st_texture_image *stImage = st_texture_image(texImage); DBG("%s\n", __FUNCTION__); if (stImage->mt) { - st_miptree_release(intel->pipe, &stImage->mt); + st_miptree_release(pipe, &stImage->mt); } if (texImage->Data) { diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c index c5aab96ea7..5b25b2a43b 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c @@ -33,6 +33,7 @@ #include "intel_context.h" #include "intel_tex.h" +#include "state_tracker/st_context.h" #include "state_tracker/st_mipmap_tree.h" #include "pipe/p_context.h" @@ -51,7 +52,7 @@ intelTexSubimage(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - struct intel_context *intel = intel_context(ctx); + struct pipe_context *pipe = ctx->st->pipe; struct st_texture_image *intelImage = st_texture_image(texImage); GLuint dstRowStride; @@ -68,15 +69,17 @@ intelTexSubimage(GLcontext * ctx, return; if (intelImage->mt) - intel->pipe->region_idle(intel->pipe, intelImage->mt->region); + pipe->region_idle(pipe, intelImage->mt->region); +#if 0 LOCK_HARDWARE(intel); +#endif /* Map buffer if necessary. Need to lock to prevent other contexts * from uploading the buffer under us. */ if (intelImage->mt) - texImage->Data = st_miptree_image_map(intel->pipe, + texImage->Data = st_miptree_image_map(pipe, intelImage->mt, intelImage->face, intelImage->level, @@ -108,11 +111,13 @@ intelTexSubimage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, packing); if (intelImage->mt) { - st_miptree_image_unmap(intel->pipe, intelImage->mt); + st_miptree_image_unmap(pipe, intelImage->mt); texImage->Data = NULL; } +#if 0 UNLOCK_HARDWARE(intel); +#endif } -- cgit v1.2.3 From 2dd4506de35fbf41b656fbdda65fe7f8585158cd Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 13:42:18 -0600 Subject: disable/remove intel dependencies --- src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 48 ++++++++++++++++++-------- 1 file changed, 33 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c index b62aaaeebb..d2cf6f4669 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c @@ -25,20 +25,23 @@ * **************************************************************************/ -#include "mtypes.h" -#include "enums.h" -#include "image.h" -#include "teximage.h" -#include "swrast/swrast.h" +#include "main/mtypes.h" +#include "main/enums.h" +#include "main/image.h" +#include "main/teximage.h" +#include "state_tracker/st_mipmap_tree.h" + +#if 0 #include "intel_screen.h" #include "intel_context.h" #include "intel_batchbuffer.h" #include "intel_buffers.h" -#include "state_tracker/st_mipmap_tree.h" #include "intel_fbo.h" -#include "intel_tex.h" #include "intel_blit.h" +#endif +#include "intel_tex.h" + #define FILE_DEBUG_FLAG DEBUG_TEXTURE @@ -51,6 +54,7 @@ static const struct pipe_region * get_teximage_source(struct intel_context *intel, GLenum internalFormat) { +#if 00 struct intel_renderbuffer *irb; DBG("%s %s\n", __FUNCTION__, @@ -79,17 +83,20 @@ get_teximage_source(struct intel_context *intel, GLenum internalFormat) default: return NULL; } +#else + return NULL; +#endif } static GLboolean -do_copy_texsubimage(struct intel_context *intel, +do_copy_texsubimage(GLcontext *ctx, struct st_texture_image *intelImage, GLenum internalFormat, GLint dstx, GLint dsty, GLint x, GLint y, GLsizei width, GLsizei height) { - GLcontext *ctx = &intel->ctx; + struct intel_context *intel = intel_context(ctx); const struct pipe_region *src = get_teximage_source(intel, internalFormat); @@ -98,9 +105,12 @@ do_copy_texsubimage(struct intel_context *intel, return GL_FALSE; } +#if 00 /* XXX FIX flush/locking */ intelFlush(ctx); /* XXX still need the lock ? */ LOCK_HARDWARE(intel); +#endif + { GLuint image_offset = st_miptree_image_offset(intelImage->mt, intelImage->face, @@ -126,6 +136,7 @@ do_copy_texsubimage(struct intel_context *intel, * pitches. But we get a nice inverted blit this way, so it's * worth it: */ +#if 0 intelEmitCopyBlit(intel, intelImage->mt->cpp, -src->pitch, @@ -136,13 +147,17 @@ do_copy_texsubimage(struct intel_context *intel, image_offset, x, y + height, dstx, dsty, width, height, GL_COPY); /* ? */ - intel_batchbuffer_flush(intel->batch); +#else + /* XXX use pipe->region_copy() ??? */ + (void) image_offset; +#endif } } - +#if 0 UNLOCK_HARDWARE(intel); +#endif #if 0 /* GL_SGIS_generate_mipmap -- this can be accelerated now. @@ -185,7 +200,7 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, GL_RGBA, CHAN_TYPE, NULL, &ctx->DefaultPacking, texObj, texImage); - if (!do_copy_texsubimage(intel_context(ctx), + if (!do_copy_texsubimage(ctx, st_texture_image(texImage), internalFormat, 0, 0, x, y, width, 1)) goto fail; @@ -193,8 +208,11 @@ intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, return; fail: +#if 0 _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, width, border); +#endif + ; } void @@ -222,7 +240,7 @@ intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, &ctx->DefaultPacking, texObj, texImage); - if (!do_copy_texsubimage(intel_context(ctx), + if (!do_copy_texsubimage(ctx, st_texture_image(texImage), internalFormat, 0, 0, x, y, width, height)) goto fail; @@ -255,7 +273,7 @@ intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, /* Need to check texture is compatible with source format. */ - if (!do_copy_texsubimage(intel_context(ctx), + if (!do_copy_texsubimage(ctx, st_texture_image(texImage), internalFormat, xoffset, 0, x, y, width, 1)) { #if 0 @@ -284,7 +302,7 @@ intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, /* Need to check texture is compatible with source format. */ - if (!do_copy_texsubimage(intel_context(ctx), + if (!do_copy_texsubimage(ctx, st_texture_image(texImage), internalFormat, xoffset, yoffset, x, y, width, height)) { -- cgit v1.2.3 From fe52cce0e8e72fcf4c11c226246ea7c24015fef6 Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 15:57:26 +0200 Subject: Add GLSL opcodes. --- src/mesa/pipe/tgsi/core/tgsi_token.h | 59 ++++++++++++++++++++++++++++++++++-- 1 file changed, 57 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index becdd480fb..40297f2575 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -268,7 +268,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_RCC */ /* TGSI_OPCODE_RCP */ #define TGSI_OPCODE_RET 63 -/* TGSI_OPCODE_RSQNV */ +/* TGSI_OPCODE_RSQNV - use TGSI_OPCODE_RSQ */ /* TGSI_OPCODE_SEQ */ /* TGSI_OPCODE_SFL */ /* TGSI_OPCODE_SGE */ @@ -677,7 +677,62 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_EMIT 96 #define TGSI_OPCODE_ENDPRIM 97 -#define TGSI_OPCODE_LAST 98 +/* + * GLSL + */ +#define TGSI_OPCODE_NOP 98 +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_BGNLOOP 99 +#define TGSI_OPCODE_BGNSUB 100 +/* TGSI_OPCODE_BRA */ +/* TGSI_OPCODE_BRK */ +/* TGSI_OPCODE_CONT */ +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_DDX */ +/* TGSI_OPCODE_DDY */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +#define TGSI_OPCODE_ENDLOOP2 101 +#define TGSI_OPCODE_ENDSUB 102 +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_IF */ +#define TGSI_OPCODE_INT TGSI_OPCODE_TRUNC +/* TGSI_OPCODE_KIL */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +#define TGSI_OPCODE_NOISE1 103 +#define TGSI_OPCODE_NOISE2 104 +#define TGSI_OPCODE_NOISE3 105 +#define TGSI_OPCODE_NOISE4 106 +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXB */ +/* TGSI_OPCODE_TXD */ +/* TGSI_OPCODE_TXL */ +/* TGSI_OPCODE_XPD */ + +#define TGSI_OPCODE_LAST 107 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ -- cgit v1.2.3 From 3e147fe8e73519074de5ba6f5bc8eaea7c788968 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 15:48:08 -0600 Subject: comments --- src/mesa/pipe/p_state.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 13b8813f4a..be4b0ebdc7 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -309,15 +309,15 @@ struct pipe_mipmap_level * compute the offsets of depth/cube images within a mipmap level, * so have to store them as a lookup table: */ - GLuint *image_offset; + GLuint *image_offset; /**< array [depth] of offsets */ }; struct pipe_mipmap_tree { /* Effectively the key: */ - GLenum target; - GLenum internal_format; + GLenum target; /* XXX convert to PIPE_TEXTURE_x */ + GLenum internal_format; /* XXX convert to PIPE_FORMAT_x */ GLuint first_level; GLuint last_level; -- cgit v1.2.3 From 24df8f895fe8807aa2ba058e71bd40adfc01d21e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 15:48:42 -0600 Subject: new texture functions --- src/mesa/state_tracker/st_cb_texture.c | 1763 ++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_texture.h | 19 + 2 files changed, 1782 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_texture.c create mode 100644 src/mesa/state_tracker/st_cb_texture.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c new file mode 100644 index 0000000000..a0245b553f --- /dev/null +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -0,0 +1,1763 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "main/imports.h" +#include "main/convolve.h" +#include "main/enums.h" +#include "main/image.h" +#include "main/macros.h" +#include "main/texcompress.h" +#include "main/texformat.h" +#include "main/teximage.h" +#include "main/texobj.h" +#include "main/texstore.h" + +#include "state_tracker/st_context.h" +#include "state_tracker/st_cb_texture.h" +#include "state_tracker/st_mipmap_tree.h" + +#include "pipe/p_context.h" + + +#define DBG if (0) printf + + +struct st_texture_object +{ + struct gl_texture_object base; /* The "parent" object */ + + /* The mipmap tree must include at least these levels once + * validated: + */ + GLuint firstLevel; + GLuint lastLevel; + + /* Offset for firstLevel image: + */ + GLuint textureOffset; + + /* On validation any active images held in main memory or in other + * regions will be copied to this region and the old storage freed. + */ + struct pipe_mipmap_tree *mt; + + GLboolean imageOverride; + GLint depthOverride; + GLuint pitchOverride; +}; + + + +struct st_texture_image +{ + struct gl_texture_image base; + + /* These aren't stored in gl_texture_image + */ + GLuint level; + GLuint face; + + /* If stImage->mt != NULL, image data is stored here. + * Else if stImage->base.Data != NULL, image is stored there. + * Else there is no image data. + */ + struct pipe_mipmap_tree *mt; +}; + + + + +static INLINE struct st_texture_object * +st_texture_object(struct gl_texture_object *obj) +{ + return (struct st_texture_object *) obj; +} + +static INLINE struct st_texture_image * +st_texture_image(struct gl_texture_image *img) +{ + return (struct st_texture_image *) img; +} + + +static int +intel_compressed_num_bytes(GLuint mesaFormat) +{ + int bytes = 0; + switch(mesaFormat) { + + case MESA_FORMAT_RGB_FXT1: + case MESA_FORMAT_RGBA_FXT1: + case MESA_FORMAT_RGB_DXT1: + case MESA_FORMAT_RGBA_DXT1: + bytes = 2; + break; + + case MESA_FORMAT_RGBA_DXT3: + case MESA_FORMAT_RGBA_DXT5: + bytes = 4; + default: + break; + } + + return bytes; +} + + +/* It works out that this function is fine for all the supported + * hardware. However, there is still a need to map the formats onto + * hardware descriptors. + */ +/* Note that the i915 can actually support many more formats than + * these if we take the step of simply swizzling the colors + * immediately after sampling... + */ +static const struct gl_texture_format * +st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, + GLenum format, GLenum type) +{ +#if 0 + struct intel_context *intel = intel_context(ctx); + const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); +#else + const GLboolean do32bpt = 1; +#endif + + switch (internalFormat) { + case 4: + case GL_RGBA: + case GL_COMPRESSED_RGBA: + if (format == GL_BGRA) { + if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { + return &_mesa_texformat_argb8888; + } + else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { + return &_mesa_texformat_argb4444; + } + else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { + return &_mesa_texformat_argb1555; + } + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case 3: + case GL_RGB: + case GL_COMPRESSED_RGB: + if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { + return &_mesa_texformat_rgb565; + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case GL_RGBA4: + case GL_RGBA2: + return &_mesa_texformat_argb4444; + + case GL_RGB5_A1: + return &_mesa_texformat_argb1555; + + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + return &_mesa_texformat_argb8888; + + case GL_RGB5: + case GL_RGB4: + case GL_R3_G3_B2: + return &_mesa_texformat_rgb565; + + case GL_ALPHA: + case GL_ALPHA4: + case GL_ALPHA8: + case GL_ALPHA12: + case GL_ALPHA16: + case GL_COMPRESSED_ALPHA: + return &_mesa_texformat_a8; + + case 1: + case GL_LUMINANCE: + case GL_LUMINANCE4: + case GL_LUMINANCE8: + case GL_LUMINANCE12: + case GL_LUMINANCE16: + case GL_COMPRESSED_LUMINANCE: + return &_mesa_texformat_l8; + + case 2: + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE4_ALPHA4: + case GL_LUMINANCE6_ALPHA2: + case GL_LUMINANCE8_ALPHA8: + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: + case GL_COMPRESSED_LUMINANCE_ALPHA: + return &_mesa_texformat_al88; + + case GL_INTENSITY: + case GL_INTENSITY4: + case GL_INTENSITY8: + case GL_INTENSITY12: + case GL_INTENSITY16: + case GL_COMPRESSED_INTENSITY: + return &_mesa_texformat_i8; + + case GL_YCBCR_MESA: + if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) + return &_mesa_texformat_ycbcr; + else + return &_mesa_texformat_ycbcr_rev; + + case GL_COMPRESSED_RGB_FXT1_3DFX: + return &_mesa_texformat_rgb_fxt1; + case GL_COMPRESSED_RGBA_FXT1_3DFX: + return &_mesa_texformat_rgba_fxt1; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + return &_mesa_texformat_rgb_dxt1; + + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return &_mesa_texformat_rgba_dxt1; + + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + return &_mesa_texformat_rgba_dxt3; + + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + return &_mesa_texformat_rgba_dxt5; + + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + return &_mesa_texformat_z16; + + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + return &_mesa_texformat_z24_s8; + + default: + fprintf(stderr, "unexpected texture format %s in %s\n", + _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); + return NULL; + } + + return NULL; /* never get here */ +} + + +static GLboolean +st_IsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) +{ +#if 0 + struct intel_context *intel = intel_context(ctx); + struct st_texture_object *stObj = st_texture_object(texObj); + + return + stObj->mt && + stObj->mt->region && + intel_is_region_resident(intel, stObj->mt->region); +#endif + return 1; +} + + + +static struct gl_texture_image * +st_NewTextureImage(GLcontext * ctx) +{ + DBG("%s\n", __FUNCTION__); + (void) ctx; + return (struct gl_texture_image *) CALLOC_STRUCT(st_texture_image); +} + + +static struct gl_texture_object * +st_NewTextureObject(GLcontext * ctx, GLuint name, GLenum target) +{ + struct st_texture_object *obj = CALLOC_STRUCT(st_texture_object); + + DBG("%s\n", __FUNCTION__); + _mesa_initialize_texture_object(&obj->base, name, target); + + return &obj->base; +} + +static void +st_DeleteTextureObject(GLcontext *ctx, + struct gl_texture_object *texObj) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_texture_object *stObj = st_texture_object(texObj); + + if (stObj->mt) + st_miptree_release(pipe, &stObj->mt); + + _mesa_delete_texture_object(ctx, texObj); +} + + +static void +st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_texture_image *stImage = st_texture_image(texImage); + + DBG("%s\n", __FUNCTION__); + + if (stImage->mt) { + st_miptree_release(pipe, &stImage->mt); + } + + if (texImage->Data) { + free(texImage->Data); + texImage->Data = NULL; + } +} + + + + +/* ================================================================ + * From linux kernel i386 header files, copes with odd sizes better + * than COPY_DWORDS would: + * XXX Put this in src/mesa/main/imports.h ??? + */ +#if defined(i386) || defined(__i386__) +static INLINE void * +__memcpy(void *to, const void *from, size_t n) +{ + int d0, d1, d2; + __asm__ __volatile__("rep ; movsl\n\t" + "testb $2,%b4\n\t" + "je 1f\n\t" + "movsw\n" + "1:\ttestb $1,%b4\n\t" + "je 2f\n\t" + "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2) + :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from) + :"memory"); + return (to); +} +#else +#define __memcpy(a,b,c) memcpy(a,b,c) +#endif + + +/* The system memcpy (at least on ubuntu 5.10) has problems copying + * to agp (writecombined) memory from a source which isn't 64-byte + * aligned - there is a 4x performance falloff. + * + * The x86 __memcpy is immune to this but is slightly slower + * (10%-ish) than the system memcpy. + * + * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but + * isn't much faster than x86_memcpy for agp copies. + * + * TODO: switch dynamically. + */ +static void * +do_memcpy(void *dest, const void *src, size_t n) +{ + if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) { + return __memcpy(dest, src, n); + } + else + return memcpy(dest, src, n); +} + + +/* Functions to store texture images. Where possible, mipmap_tree's + * will be created or further instantiated with image data, otherwise + * images will be stored in malloc'd memory. A validation step is + * required to pull those images into a mipmap tree, or otherwise + * decide a fallback is required. + */ + + +static int +logbase2(int n) +{ + GLint i = 1; + GLint log2 = 0; + + while (n > i) { + i *= 2; + log2++; + } + + return log2; +} + + +/* Otherwise, store it in memory if (Border != 0) or (any dimension == + * 1). + * + * Otherwise, if max_level >= level >= min_level, create tree with + * space for textures from min_level down to max_level. + * + * Otherwise, create tree with space for textures from (level + * 0)..(1x1). Consider pruning this tree at a validation if the + * saving is worth it. + */ +static void +guess_and_alloc_mipmap_tree(struct pipe_context *pipe, + struct st_texture_object *intelObj, + struct st_texture_image *intelImage) +{ + GLuint firstLevel; + GLuint lastLevel; + GLuint width = intelImage->base.Width; + GLuint height = intelImage->base.Height; + GLuint depth = intelImage->base.Depth; + GLuint l2width, l2height, l2depth; + GLuint i, comp_byte = 0; + + DBG("%s\n", __FUNCTION__); + + if (intelImage->base.Border) + return; + + if (intelImage->level > intelObj->base.BaseLevel && + (intelImage->base.Width == 1 || + (intelObj->base.Target != GL_TEXTURE_1D && + intelImage->base.Height == 1) || + (intelObj->base.Target == GL_TEXTURE_3D && + intelImage->base.Depth == 1))) + return; + + /* If this image disrespects BaseLevel, allocate from level zero. + * Usually BaseLevel == 0, so it's unlikely to happen. + */ + if (intelImage->level < intelObj->base.BaseLevel) + firstLevel = 0; + else + firstLevel = intelObj->base.BaseLevel; + + + /* Figure out image dimensions at start level. + */ + for (i = intelImage->level; i > firstLevel; i--) { + width <<= 1; + if (height != 1) + height <<= 1; + if (depth != 1) + depth <<= 1; + } + + /* Guess a reasonable value for lastLevel. This is probably going + * to be wrong fairly often and might mean that we have to look at + * resizable buffers, or require that buffers implement lazy + * pagetable arrangements. + */ + if ((intelObj->base.MinFilter == GL_NEAREST || + intelObj->base.MinFilter == GL_LINEAR) && + intelImage->level == firstLevel) { + lastLevel = firstLevel; + } + else { + l2width = logbase2(width); + l2height = logbase2(height); + l2depth = logbase2(depth); + lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); + } + + assert(!intelObj->mt); + if (intelImage->base.IsCompressed) + comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); + intelObj->mt = st_miptree_create(pipe, + intelObj->base.Target, + intelImage->base.InternalFormat, + firstLevel, + lastLevel, + width, + height, + depth, + intelImage->base.TexFormat->TexelBytes, + comp_byte); + + DBG("%s - success\n", __FUNCTION__); +} + + + + +static GLuint +target_to_face(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + return ((GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X); + default: + return 0; + } +} + + + +/* There are actually quite a few combinations this will work for, + * more than what I've listed here. + */ +static GLboolean +check_pbo_format(GLint internalFormat, + GLenum format, GLenum type, + const struct gl_texture_format *mesa_format) +{ + switch (internalFormat) { + case 4: + case GL_RGBA: + return (format == GL_BGRA && + (type == GL_UNSIGNED_BYTE || + type == GL_UNSIGNED_INT_8_8_8_8_REV) && + mesa_format == &_mesa_texformat_argb8888); + case 3: + case GL_RGB: + return (format == GL_RGB && + type == GL_UNSIGNED_SHORT_5_6_5 && + mesa_format == &_mesa_texformat_rgb565); + case GL_YCBCR_MESA: + return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); + default: + return GL_FALSE; + } +} + + +/* XXX: Do this for TexSubImage also: + */ +static GLboolean +try_pbo_upload(GLcontext *ctx, + struct st_texture_image *intelImage, + const struct gl_pixelstore_attrib *unpack, + GLint internalFormat, + GLint width, GLint height, + GLenum format, GLenum type, const void *pixels) +{ + return GL_FALSE; /* XXX fix flushing/locking/blitting below */ +#if 000 + struct intel_context *intel = intel_context(ctx); + struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); + GLuint src_offset, src_stride; + GLuint dst_offset, dst_stride; + + if (!pbo || + ctx._ImageTransferState || + unpack->SkipPixels || unpack->SkipRows) { + _mesa_printf("%s: failure 1\n", __FUNCTION__); + return GL_FALSE; + } + + src_offset = (GLuint) pixels; + + if (unpack->RowLength > 0) + src_stride = unpack->RowLength; + else + src_stride = width; + + dst_offset = st_miptree_image_offset(intelImage->mt, + intelImage->face, + intelImage->level); + + dst_stride = intelImage->mt->pitch; + + intelFlush(&intel->ctx); + LOCK_HARDWARE(intel); + { + struct _DriBufferObject *src_buffer = + intel_bufferobj_buffer(intel, pbo, INTEL_READ); + + /* Temporary hack: cast to _DriBufferObject: + */ + struct _DriBufferObject *dst_buffer = + (struct _DriBufferObject *)intelImage->mt->region->buffer; + + + intelEmitCopyBlit(intel, + intelImage->mt->cpp, + src_stride, src_buffer, src_offset, + dst_stride, dst_buffer, dst_offset, + 0, 0, 0, 0, width, height, + GL_COPY); + + intel_batchbuffer_flush(intel->batch); + } + UNLOCK_HARDWARE(intel); + + return GL_TRUE; +#endif +} + + + +static GLboolean +try_pbo_zcopy(GLcontext *ctx, + struct st_texture_image *intelImage, + const struct gl_pixelstore_attrib *unpack, + GLint internalFormat, + GLint width, GLint height, + GLenum format, GLenum type, const void *pixels) +{ + return GL_FALSE; +} + + + + + + +static void +st_TexImage(GLcontext * ctx, + GLint dims, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, GLsizei imageSize, int compressed) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_texture_object *intelObj = st_texture_object(texObj); + struct st_texture_image *intelImage = st_texture_image(texImage); + GLint postConvWidth = width; + GLint postConvHeight = height; + GLint texelBytes, sizeInBytes; + GLuint dstRowStride; + + + DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); + +#if 0 + intelFlush(ctx); +#endif + + intelImage->face = target_to_face(target); + intelImage->level = level; + + if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { + _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth, + &postConvHeight); + } + + /* choose the texture format */ + texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat, + format, type); + + _mesa_set_fetch_functions(texImage, dims); + + if (texImage->TexFormat->TexelBytes == 0) { + /* must be a compressed format */ + texelBytes = 0; + texImage->IsCompressed = GL_TRUE; + texImage->CompressedSize = + ctx->Driver.CompressedTextureSize(ctx, texImage->Width, + texImage->Height, texImage->Depth, + texImage->TexFormat->MesaFormat); + } else { + texelBytes = texImage->TexFormat->TexelBytes; + + /* Minimum pitch of 32 bytes */ + if (postConvWidth * texelBytes < 32) { + postConvWidth = 32 / texelBytes; + texImage->RowStride = postConvWidth; + } + + assert(texImage->RowStride == postConvWidth); + } + + /* Release the reference to a potentially orphaned buffer. + * Release any old malloced memory. + */ + if (intelImage->mt) { + st_miptree_release(pipe, &intelImage->mt); + assert(!texImage->Data); + } + else if (texImage->Data) { + _mesa_align_free(texImage->Data); + } + + /* If this is the only texture image in the tree, could call + * bmBufferData with NULL data to free the old block and avoid + * waiting on any outstanding fences. + */ + if (intelObj->mt && + intelObj->mt->first_level == level && + intelObj->mt->last_level == level && + intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && + !st_miptree_match_image(intelObj->mt, &intelImage->base, + intelImage->face, intelImage->level)) { + + DBG("release it\n"); + st_miptree_release(pipe, &intelObj->mt); + assert(!intelObj->mt); + } + + if (!intelObj->mt) { + guess_and_alloc_mipmap_tree(pipe, intelObj, intelImage); + if (!intelObj->mt) { + DBG("guess_and_alloc_mipmap_tree: failed\n"); + } + } + + assert(!intelImage->mt); + + if (intelObj->mt && + st_miptree_match_image(intelObj->mt, &intelImage->base, + intelImage->face, intelImage->level)) { + + st_miptree_reference(&intelImage->mt, intelObj->mt); + assert(intelImage->mt); + } + + if (!intelImage->mt) + DBG("XXX: Image did not fit into tree - storing in local memory!\n"); + +#if 0 /* XXX FIX when st_buffer_objects are in place */ + /* PBO fastpaths: + */ + if (dims <= 2 && + intelImage->mt && + intel_buffer_object(unpack->BufferObj) && + check_pbo_format(internalFormat, format, + type, intelImage->base.TexFormat)) { + + DBG("trying pbo upload\n"); + + /* Attempt to texture directly from PBO data (zero copy upload). + * + * Currently disable as it can lead to worse as well as better + * performance (in particular when pipe_region_cow() is + * required). + */ + if (intelObj->mt == intelImage->mt && + intelObj->mt->first_level == level && + intelObj->mt->last_level == level) { + + if (try_pbo_zcopy(intel, intelImage, unpack, + internalFormat, + width, height, format, type, pixels)) { + + DBG("pbo zcopy upload succeeded\n"); + return; + } + } + + + /* Otherwise, attempt to use the blitter for PBO image uploads. + */ + if (try_pbo_upload(intel, intelImage, unpack, + internalFormat, + width, height, format, type, pixels)) { + DBG("pbo upload succeeded\n"); + return; + } + + DBG("pbo upload failed\n"); + } +#else + (void) try_pbo_upload; + (void) check_pbo_format; + (void) try_pbo_zcopy; +#endif + + + /* intelCopyTexImage calls this function with pixels == NULL, with + * the expectation that the mipmap tree will be set up but nothing + * more will be done. This is where those calls return: + */ + if (compressed) { + pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels, + unpack, + "glCompressedTexImage"); + } else { + pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, + format, type, + pixels, unpack, "glTexImage"); + } + if (!pixels) + return; + + + if (intelImage->mt) + pipe->region_idle(pipe, intelImage->mt->region); + +#if 0 + LOCK_HARDWARE(intel); +#endif + + if (intelImage->mt) { + texImage->Data = st_miptree_image_map(pipe, + intelImage->mt, + intelImage->face, + intelImage->level, + &dstRowStride, + intelImage->base.ImageOffsets); + } + else { + /* Allocate regular memory and store the image there temporarily. */ + if (texImage->IsCompressed) { + sizeInBytes = texImage->CompressedSize; + dstRowStride = + _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); + assert(dims != 3); + } + else { + dstRowStride = postConvWidth * texelBytes; + sizeInBytes = depth * dstRowStride * postConvHeight; + } + + texImage->Data = malloc(sizeInBytes); + } + + DBG("Upload image %dx%dx%d row_len %x pitch %x\n", + width, height, depth, width * texelBytes, dstRowStride); + + /* Copy data. Would like to know when it's ok for us to eg. use + * the blitter to copy. Or, use the hardware to do the format + * conversion and copy: + */ + if (compressed) { + memcpy(texImage->Data, pixels, imageSize); + } + else if (!texImage->TexFormat->StoreImage(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, unpack)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + } + + _mesa_unmap_teximage_pbo(ctx, unpack); + + if (intelImage->mt) { + st_miptree_image_unmap(pipe, intelImage->mt); + texImage->Data = NULL; + } + +#if 0 + UNLOCK_HARDWARE(intel); +#endif + +#if 0 + /* GL_SGIS_generate_mipmap -- this can be accelerated now. + */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif +} + + +static void +st_TexImage3D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + st_TexImage(ctx, 3, target, level, + internalFormat, width, height, depth, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + + +static void +st_TexImage2D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + st_TexImage(ctx, 2, target, level, + internalFormat, width, height, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + + +static void +st_TexImage1D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + st_TexImage(ctx, 1, target, level, + internalFormat, width, 1, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + + +static void +st_CompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid *data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage ) +{ + st_TexImage(ctx, 2, target, level, + internalFormat, width, height, 1, border, + 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1); +} + + +/** + * Need to map texture image into memory before copying image data, + * then unmap it. + */ +static void +st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, int compressed) +{ + /* + struct intel_context *intel = intel_context(ctx); + */ + struct pipe_context *pipe = ctx->st->pipe; + struct st_texture_image *stImage = st_texture_image(texImage); + + /* Map */ + if (stImage->mt) { + /* Image is stored in hardware format in a buffer managed by the + * kernel. Need to explicitly map and unmap it. + */ + stImage->base.Data = + st_miptree_image_map(pipe, + stImage->mt, + stImage->face, + stImage->level, + &stImage->base.RowStride, + stImage->base.ImageOffsets); + stImage->base.RowStride /= stImage->mt->cpp; + } + else { + /* Otherwise, the image should actually be stored in + * stImage->base.Data. This is pretty confusing for + * everybody, I'd much prefer to separate the two functions of + * texImage->Data - storage for texture images in main memory + * and access (ie mappings) of images. In other words, we'd + * create a new texImage->Map field and leave Data simply for + * storage. + */ + assert(stImage->base.Data); + } + + + if (compressed) { + _mesa_get_compressed_teximage(ctx, target, level, pixels, + texObj, texImage); + } else { + _mesa_get_teximage(ctx, target, level, format, type, pixels, + texObj, texImage); + } + + + /* Unmap */ + if (stImage->mt) { + st_miptree_image_unmap(pipe, stImage->mt); + stImage->base.Data = NULL; + } +} + + +static void +st_GetTexImage(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + st_get_tex_image(ctx, target, level, format, type, pixels, + texObj, texImage, 0); +} + + +static void +st_GetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, + GLvoid *pixels, + const struct gl_texture_object *texObj, + const struct gl_texture_image *texImage) +{ + st_get_tex_image(ctx, target, level, 0, 0, pixels, + (struct gl_texture_object *) texObj, + (struct gl_texture_image *) texImage, 1); +} + + + +static void +st_TexSubimage(GLcontext * ctx, + GLint dims, + GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLint width, GLint height, GLint depth, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_texture_image *stImage = st_texture_image(texImage); + GLuint dstRowStride; + + DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), + level, xoffset, yoffset, width, height); + +#if 0 + intelFlush(ctx); +#endif + + pixels = + _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format, + type, pixels, packing, "glTexSubImage2D"); + if (!pixels) + return; + + if (stImage->mt) + pipe->region_idle(pipe, stImage->mt->region); + +#if 0 + LOCK_HARDWARE(intel); +#endif + + /* Map buffer if necessary. Need to lock to prevent other contexts + * from uploading the buffer under us. + */ + if (stImage->mt) + texImage->Data = st_miptree_image_map(pipe, + stImage->mt, + stImage->face, + stImage->level, + &dstRowStride, + texImage->ImageOffsets); + + assert(dstRowStride); + + if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, packing)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); + } + +#if 0 + /* GL_SGIS_generate_mipmap */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + _mesa_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif + + _mesa_unmap_teximage_pbo(ctx, packing); + + if (stImage->mt) { + st_miptree_image_unmap(pipe, stImage->mt); + texImage->Data = NULL; + } + +#if 0 + UNLOCK_HARDWARE(intel); +#endif +} + + + +static void +st_TexSubImage3D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + st_TexSubimage(ctx, 3, + target, level, + xoffset, yoffset, zoffset, + width, height, depth, + format, type, pixels, packing, texObj, texImage); +} + + + +static void +st_TexSubImage2D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + st_TexSubimage(ctx, 2, + target, level, + xoffset, yoffset, 0, + width, height, 1, + format, type, pixels, packing, texObj, texImage); +} + + +static void +st_TexSubImage1D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + st_TexSubimage(ctx, 1, + target, level, + xoffset, 0, 0, + width, 1, 1, + format, type, pixels, packing, texObj, texImage); +} + + + +/** + * Get the pipe_region which is the source for any glCopyTex[Sub]Image call. + * + * Do the best we can using the blitter. A future project is to use + * the texture engine and fragment programs for these copies. + */ +static const struct pipe_region * +get_teximage_source(GLcontext *ctx, GLenum internalFormat) +{ +#if 00 + struct intel_renderbuffer *irb; + + DBG("%s %s\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(internalFormat)); + + switch (internalFormat) { + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16_ARB: + irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); + if (irb && irb->region && irb->region->cpp == 2) + return irb->region; + return NULL; + case GL_DEPTH24_STENCIL8_EXT: + case GL_DEPTH_STENCIL_EXT: + irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); + if (irb && irb->region && irb->region->cpp == 4) + return irb->region; + return NULL; + case GL_RGBA: + case GL_RGBA8: + return intel_readbuf_region(intel); + case GL_RGB: + if (intel->intelScreen->front.cpp == 2) + return intel_readbuf_region(intel); + return NULL; + default: + return NULL; + } +#else + return NULL; +#endif +} + + +static GLboolean +do_copy_texsubimage(GLcontext *ctx, + struct st_texture_image *stImage, + GLenum internalFormat, + GLint dstx, GLint dsty, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + const struct pipe_region *src = + get_teximage_source(ctx, internalFormat); + + if (!stImage->mt || !src) { + DBG("%s fail %p %p\n", __FUNCTION__, stImage->mt, src); + return GL_FALSE; + } + +#if 00 /* XXX FIX flush/locking */ + intelFlush(ctx); + /* XXX still need the lock ? */ + LOCK_HARDWARE(intel); +#endif + + { + GLuint image_offset = st_miptree_image_offset(stImage->mt, + stImage->face, + stImage->level); + const GLint orig_x = x; + const GLint orig_y = y; + const struct gl_framebuffer *fb = ctx->DrawBuffer; + + if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, + &x, &y, &width, &height)) { + /* Update dst for clipped src. Need to also clip the source rect. + */ + dstx += x - orig_x; + dsty += y - orig_y; + + if (!(ctx->ReadBuffer->Name == 0)) { + /* XXX this looks bogus ? */ + /* FBO: invert Y */ + y = ctx->ReadBuffer->Height - y - 1; + } + + /* A bit of fiddling to get the blitter to work with -ve + * pitches. But we get a nice inverted blit this way, so it's + * worth it: + */ +#if 0 + intelEmitCopyBlit(intel, + stImage->mt->cpp, + -src->pitch, + src->buffer, + src->height * src->pitch * src->cpp, + stImage->mt->pitch, + stImage->mt->region->buffer, + image_offset, + x, y + height, dstx, dsty, width, height, + GL_COPY); /* ? */ + intel_batchbuffer_flush(intel->batch); +#else + /* XXX use pipe->region_copy() ??? */ + (void) image_offset; +#endif + } + } + +#if 0 + UNLOCK_HARDWARE(intel); +#endif + +#if 0 + /* GL_SGIS_generate_mipmap -- this can be accelerated now. + * XXX Add a ctx->Driver.GenerateMipmaps() function? + */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif + + return GL_TRUE; +} + + +static void +st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLint border) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage1D(ctx, target, level, internalFormat, + width, border, + GL_RGBA, CHAN_TYPE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + if (!do_copy_texsubimage(ctx, + st_texture_image(texImage), + internalFormat, 0, 0, x, y, width, 1)) + goto fail; + + return; + + fail: +#if 0 + _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, + width, border); +#endif + ; +} + + +static void +st_CopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage2D(ctx, target, level, internalFormat, + width, height, border, + GL_RGBA, CHAN_TYPE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + + if (!do_copy_texsubimage(ctx, + st_texture_image(texImage), + internalFormat, 0, 0, x, y, width, height)) + goto fail; + + return; + + fail: +#if 0 + _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, + width, height, border); +#endif + assert(0); +} + + +static void +st_CopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint x, GLint y, GLsizei width) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + GLenum internalFormat = texImage->InternalFormat; + + /* XXX need to check as in above function? */ + + /* Need to check texture is compatible with source format. + */ + + if (!do_copy_texsubimage(ctx, + st_texture_image(texImage), + internalFormat, xoffset, 0, x, y, width, 1)) { +#if 0 + _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); +#endif + assert(0); + } +} + + +static void +st_CopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + GLenum internalFormat = texImage->InternalFormat; + + + /* Need to check texture is compatible with source format. + */ + + if (!do_copy_texsubimage(ctx, + st_texture_image(texImage), + internalFormat, + xoffset, yoffset, x, y, width, height)) { +#if 0 + _swrast_copy_texsubimage2d(ctx, target, level, + xoffset, yoffset, x, y, width, height); +#endif + assert(0); + } +} + + + + +/** + * Compute which mipmap levels that really need to be sent to the hardware. + * This depends on the base image size, GL_TEXTURE_MIN_LOD, + * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. + */ +static void +intel_calculate_first_last_level(struct st_texture_object *intelObj) +{ + struct gl_texture_object *tObj = &intelObj->base; + const struct gl_texture_image *const baseImage = + tObj->Image[0][tObj->BaseLevel]; + + /* These must be signed values. MinLod and MaxLod can be negative numbers, + * and having firstLevel and lastLevel as signed prevents the need for + * extra sign checks. + */ + int firstLevel; + int lastLevel; + + /* Yes, this looks overly complicated, but it's all needed. + */ + switch (tObj->Target) { + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_3D: + case GL_TEXTURE_CUBE_MAP: + if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) { + /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL. + */ + firstLevel = lastLevel = tObj->BaseLevel; + } + else { + firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); + firstLevel = MAX2(firstLevel, tObj->BaseLevel); + lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); + lastLevel = MAX2(lastLevel, tObj->BaseLevel); + lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); + lastLevel = MIN2(lastLevel, tObj->MaxLevel); + lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ + } + break; + case GL_TEXTURE_RECTANGLE_NV: + case GL_TEXTURE_4D_SGIS: + firstLevel = lastLevel = 0; + break; + default: + return; + } + + /* save these values */ + intelObj->firstLevel = firstLevel; + intelObj->lastLevel = lastLevel; +} + + +static void +copy_image_data_to_tree(struct pipe_context *pipe, + struct st_texture_object *intelObj, + struct st_texture_image *stImage) +{ + if (stImage->mt) { + /* Copy potentially with the blitter: + */ + st_miptree_image_copy(pipe, + intelObj->mt, + stImage->face, + stImage->level, stImage->mt); + + st_miptree_release(pipe, &stImage->mt); + } + else { + assert(stImage->base.Data != NULL); + + /* More straightforward upload. + */ + st_miptree_image_data(pipe, + intelObj->mt, + stImage->face, + stImage->level, + stImage->base.Data, + stImage->base.RowStride, + stImage->base.RowStride * + stImage->base.Height); + _mesa_align_free(stImage->base.Data); + stImage->base.Data = NULL; + } + + st_miptree_reference(&stImage->mt, intelObj->mt); +} + + +/* + */ +GLuint +st_finalize_mipmap_tree(GLcontext *ctx, + struct pipe_context *pipe, GLuint unit, + GLboolean *needFlush) +{ + struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; + struct st_texture_object *intelObj = st_texture_object(tObj); + int comp_byte = 0; + int cpp; + + GLuint face, i; + GLuint nr_faces = 0; + struct st_texture_image *firstImage; + + *needFlush = GL_FALSE; + + /* We know/require this is true by now: + */ + assert(intelObj->base._Complete); + + /* What levels must the tree include at a minimum? + */ + intel_calculate_first_last_level(intelObj); + firstImage = + st_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); + + /* Fallback case: + */ + if (firstImage->base.Border) { + if (intelObj->mt) { + st_miptree_release(pipe, &intelObj->mt); + } + return GL_FALSE; + } + + + /* If both firstImage and intelObj have a tree which can contain + * all active images, favour firstImage. Note that because of the + * completeness requirement, we know that the image dimensions + * will match. + */ + if (firstImage->mt && + firstImage->mt != intelObj->mt && + firstImage->mt->first_level <= intelObj->firstLevel && + firstImage->mt->last_level >= intelObj->lastLevel) { + + if (intelObj->mt) + st_miptree_release(pipe, &intelObj->mt); + + st_miptree_reference(&intelObj->mt, firstImage->mt); + } + + if (firstImage->base.IsCompressed) { + comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); + cpp = comp_byte; + } + else cpp = firstImage->base.TexFormat->TexelBytes; + + /* Check tree can hold all active levels. Check tree matches + * target, imageFormat, etc. + * + * XXX: For some layouts (eg i945?), the test might have to be + * first_level == firstLevel, as the tree isn't valid except at the + * original start level. Hope to get around this by + * programming minLod, maxLod, baseLevel into the hardware and + * leaving the tree alone. + */ + if (intelObj->mt && + (intelObj->mt->target != intelObj->base.Target || + intelObj->mt->internal_format != firstImage->base.InternalFormat || + intelObj->mt->first_level != intelObj->firstLevel || + intelObj->mt->last_level != intelObj->lastLevel || + intelObj->mt->width0 != firstImage->base.Width || + intelObj->mt->height0 != firstImage->base.Height || + intelObj->mt->depth0 != firstImage->base.Depth || + intelObj->mt->cpp != cpp || + intelObj->mt->compressed != firstImage->base.IsCompressed)) { + st_miptree_release(pipe, &intelObj->mt); + } + + + /* May need to create a new tree: + */ + if (!intelObj->mt) { + intelObj->mt = st_miptree_create(pipe, + intelObj->base.Target, + firstImage->base.InternalFormat, + intelObj->firstLevel, + intelObj->lastLevel, + firstImage->base.Width, + firstImage->base.Height, + firstImage->base.Depth, + cpp, + comp_byte); + } + + /* Pull in any images not in the object's tree: + */ + nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct st_texture_image *stImage = + st_texture_image(intelObj->base.Image[face][i]); + + /* Need to import images in main memory or held in other trees. + */ + if (intelObj->mt != stImage->mt) { + copy_image_data_to_tree(pipe, intelObj, stImage); + *needFlush = GL_TRUE; + } + } + } + + /** + if (need_flush) + intel_batchbuffer_flush(intel->batch); + **/ + + return GL_TRUE; +} + + +#if 0 /* unused? */ +void +st_tex_map_images(struct pipe_context *pipe, + struct st_texture_object *intelObj) +{ + GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + + DBG("%s\n", __FUNCTION__); + + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct st_texture_image *stImage = + st_texture_image(intelObj->base.Image[face][i]); + + if (stImage->mt) { + stImage->base.Data = + st_miptree_image_map(pipe, + stImage->mt, + stImage->face, + stImage->level, + &stImage->base.RowStride, + stImage->base.ImageOffsets); + /* convert stride to texels, not bytes */ + stImage->base.RowStride /= stImage->mt->cpp; +/* stImage->base.ImageStride /= stImage->mt->cpp; */ + } + } + } +} + + + +void +st_tex_unmap_images(struct pipe_context *pipe, + struct st_texture_object *intelObj) +{ + GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct st_texture_image *stImage = + st_texture_image(intelObj->base.Image[face][i]); + + if (stImage->mt) { + st_miptree_image_unmap(pipe, stImage->mt); + stImage->base.Data = NULL; + } + } + } +} +#endif + + + + +void st_init_cb_texture( struct st_context *st ) +{ + struct dd_function_table *functions = &st->ctx->Driver; + + functions->ChooseTextureFormat = st_ChooseTextureFormat; + functions->TexImage1D = st_TexImage1D; + functions->TexImage2D = st_TexImage2D; + functions->TexImage3D = st_TexImage3D; + functions->TexSubImage1D = st_TexSubImage1D; + functions->TexSubImage2D = st_TexSubImage2D; + functions->TexSubImage3D = st_TexSubImage3D; + functions->CopyTexImage1D = st_CopyTexImage1D; + functions->CopyTexImage2D = st_CopyTexImage2D; + functions->CopyTexSubImage1D = st_CopyTexSubImage1D; + functions->CopyTexSubImage2D = st_CopyTexSubImage2D; + functions->GetTexImage = st_GetTexImage; + + /* compressed texture functions */ + functions->CompressedTexImage2D = st_CompressedTexImage2D; + functions->GetCompressedTexImage = st_GetCompressedTexImage; + + functions->NewTextureObject = st_NewTextureObject; + functions->NewTextureImage = st_NewTextureImage; + functions->DeleteTexture = st_DeleteTextureObject; + functions->FreeTexImageData = st_FreeTextureImageData; + functions->UpdateTexturePalette = 0; + functions->IsTextureResident = st_IsTextureResident; + + functions->TextureMemCpy = do_memcpy; +} + + +void st_destroy_cb_texture( struct st_context *st ) +{ +} diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h new file mode 100644 index 0000000000..c474d16465 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_texture.h @@ -0,0 +1,19 @@ +#ifndef ST_CB_TEXTURE_H +#define ST_CB_TEXTURE_H + + +extern GLuint +st_finalize_mipmap_tree(GLcontext *ctx, + struct pipe_context *pipe, GLuint unit, + GLboolean *needFlush); + + +extern void +st_init_cb_texture( struct st_context *st ); + + +extern void +st_destroy_cb_texture( struct st_context *st ); + + +#endif /* ST_CB_TEXTURE_H */ -- cgit v1.2.3 From 0790d9a074ee3c161d5e4395093d62f4d573f39b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 15:49:00 -0600 Subject: clean-ups --- src/mesa/state_tracker/st_mipmap_tree.c | 13 ++++- src/mesa/state_tracker/st_mipmap_tree.h | 99 +++++++++++++++++++-------------- 2 files changed, 68 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index e99cd987e8..6293cc8f51 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -179,7 +179,7 @@ st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) GLuint -st_miptree_image_offset(struct pipe_mipmap_tree * mt, +st_miptree_image_offset(const struct pipe_mipmap_tree * mt, GLuint face, GLuint level) { if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) @@ -190,6 +190,17 @@ st_miptree_image_offset(struct pipe_mipmap_tree * mt, } +GLuint +st_miptree_texel_offset(const struct pipe_mipmap_tree * mt, + GLuint face, GLuint level, + GLuint col, GLuint row, GLuint img) +{ + GLuint imgOffset = st_miptree_image_offset(mt, face, level); + + return imgOffset + row * (mt->pitch + col) * mt->cpp; +} + + /** * Map a teximage in a mipmap tree. diff --git a/src/mesa/state_tracker/st_mipmap_tree.h b/src/mesa/state_tracker/st_mipmap_tree.h index ea3b52ca42..3e7fd7fa0c 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.h +++ b/src/mesa/state_tracker/st_mipmap_tree.h @@ -36,70 +36,83 @@ struct pipe_mipmap_tree; struct pipe_region; -struct pipe_mipmap_tree *st_miptree_create(struct pipe_context *pipe, - GLenum target, - GLenum internal_format, - GLuint first_level, - GLuint last_level, - GLuint width0, - GLuint height0, - GLuint depth0, - GLuint cpp, - GLuint compress_byte); - -void st_miptree_reference(struct pipe_mipmap_tree **dst, - struct pipe_mipmap_tree *src); - -void st_miptree_release(struct pipe_context *pipe, - struct pipe_mipmap_tree **mt); +extern struct pipe_mipmap_tree * +st_miptree_create(struct pipe_context *pipe, + GLenum target, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, + GLuint cpp, + GLuint compress_byte); + +extern void +st_miptree_reference(struct pipe_mipmap_tree **dst, + struct pipe_mipmap_tree *src); + +extern void +st_miptree_release(struct pipe_context *pipe, struct pipe_mipmap_tree **mt); + /* Check if an image fits an existing mipmap tree layout */ -GLboolean st_miptree_match_image(struct pipe_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level); +extern GLboolean +st_miptree_match_image(struct pipe_mipmap_tree *mt, + struct gl_texture_image *image, + GLuint face, GLuint level); /* Return a pointer to an image within a tree. Return image stride as * well. */ -GLubyte *st_miptree_image_map(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt, - GLuint face, - GLuint level, - GLuint * row_stride, GLuint * image_stride); +extern GLubyte * +st_miptree_image_map(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt, + GLuint face, GLuint level, + GLuint * row_stride, GLuint * image_stride); + +extern void +st_miptree_image_unmap(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt); + -void st_miptree_image_unmap(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt); +/* Return pointers to each 2d slice within an image. Indexed by depth + * value. + */ +extern const GLuint * +st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level); /* Return the linear offset of an image relative to the start of the * tree: */ -GLuint st_miptree_image_offset(struct pipe_mipmap_tree *mt, - GLuint face, GLuint level); +extern GLuint +st_miptree_image_offset(const struct pipe_mipmap_tree *mt, + GLuint face, GLuint level); -/* Return pointers to each 2d slice within an image. Indexed by depth - * value. - */ -const GLuint *st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, - GLuint level); +extern GLuint +st_miptree_texel_offset(const struct pipe_mipmap_tree * mt, + GLuint face, GLuint level, + GLuint col, GLuint row, GLuint img); /* Upload an image into a tree */ -void st_miptree_image_data(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, - GLuint level, - void *src, - GLuint src_row_pitch, GLuint src_image_pitch); +extern void +st_miptree_image_data(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, GLuint level, void *src, + GLuint src_row_pitch, GLuint src_image_pitch); + /* Copy an image between two trees */ -void st_miptree_image_copy(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, GLuint level, - struct pipe_mipmap_tree *src); +extern void +st_miptree_image_copy(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, GLuint level, + struct pipe_mipmap_tree *src); #endif -- cgit v1.2.3 From 4435baef18ccd86ebb9797bf271b6b7182cc777e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 15:49:11 -0600 Subject: use new texture funcs --- src/mesa/state_tracker/st_context.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 6814eac2dc..2b96286770 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -30,7 +30,7 @@ #include "st_context.h" #include "st_cb_clear.h" #include "st_cb_drawpixels.h" -#include "st_cb_teximage.h" +#include "st_cb_texture.h" #include "st_atom.h" #include "st_draw.h" #include "st_program.h" @@ -64,7 +64,7 @@ struct st_context *st_create_context( GLcontext *ctx, st_init_cb_clear( st ); st_init_cb_program( st ); st_init_cb_drawpixels( st ); - st_init_cb_teximage( st ); + st_init_cb_texture( st ); return st; } @@ -78,7 +78,8 @@ void st_destroy_context( struct st_context *st ) st_destroy_cb_clear( st ); st_destroy_cb_program( st ); st_destroy_cb_drawpixels( st ); - st_destroy_cb_teximage( st ); + /*st_destroy_cb_teximage( st );*/ + st_destroy_cb_texture( st ); st->pipe->destroy( st->pipe ); FREE( st ); -- cgit v1.2.3 From 1cc774f8da4ce9d7defb5419b07e1bdb8b8275a1 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 15:49:44 -0600 Subject: switch to new texture functions in state tracker --- src/mesa/drivers/dri/i915pipe/Makefile | 15 +++++++----- src/mesa/drivers/dri/i915pipe/intel_context.c | 4 +++- src/mesa/drivers/dri/i915pipe/intel_fbo.c | 4 +++- src/mesa/drivers/dri/i915pipe/intel_screen.c | 27 +++++++++++++++++++++- src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c | 2 ++ 5 files changed, 43 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile index f9403abd73..3f3e04e978 100644 --- a/src/mesa/drivers/dri/i915pipe/Makefile +++ b/src/mesa/drivers/dri/i915pipe/Makefile @@ -14,12 +14,6 @@ DRIVER_SOURCES = \ intel_buffer_objects.c \ intel_batchbuffer.c \ intel_tex_layout.c \ - intel_tex_image.c \ - intel_tex_subimage.c \ - intel_tex_copy.c \ - intel_tex_validate.c \ - intel_tex_format.c \ - intel_tex.c \ intel_buffers.c \ intel_blit.c \ intel_context.c \ @@ -30,6 +24,15 @@ DRIVER_SOURCES = \ intel_depthstencil.c \ intel_batchpool.c +OLD_TEX =\ + intel_tex_image.c \ + intel_tex_subimage.c \ + intel_tex_copy.c \ + intel_tex_validate.c \ + intel_tex_format.c \ + intel_tex.c + + C_SOURCES = \ $(COMMON_SOURCES) \ $(COMMON_BM_SOURCES) \ diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index e267135973..ced4366009 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -47,7 +47,7 @@ #include "i830_dri.h" #include "intel_buffers.h" -#include "intel_tex.h" +/*#include "intel_tex.h"*/ #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" @@ -318,7 +318,9 @@ intelInitDriverFunctions(struct dd_function_table *functions) functions->GetString = intelGetString; functions->UpdateState = intelInvalidateState; + /* intelInitTextureFuncs(functions); + */ intelInitBufferFuncs(functions); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index 2f1aa00038..a53563f52b 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -40,7 +40,7 @@ #include "intel_depthstencil.h" #include "intel_fbo.h" #include "state_tracker/st_mipmap_tree.h" -#include "intel_tex.h" +/*#include "intel_tex.h"*/ #include "pipe/p_context.h" @@ -528,6 +528,7 @@ intel_render_texture(GLcontext * ctx, struct gl_framebuffer *fb, struct gl_renderbuffer_attachment *att) { +#if 0 struct intel_context *intel = intel_context(ctx); struct gl_texture_image *newImage = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; @@ -581,6 +582,7 @@ intel_render_texture(GLcontext * ctx, /* update drawing region, etc */ intel_draw_buffer(ctx, fb); +#endif } diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.c b/src/mesa/drivers/dri/i915pipe/intel_screen.c index 5b8d6256b5..81ab435f42 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.c +++ b/src/mesa/drivers/dri/i915pipe/intel_screen.c @@ -38,7 +38,7 @@ #include "intel_screen.h" #include "intel_batchbuffer.h" #include "intel_buffers.h" -#include "intel_tex.h" +/*#include "intel_tex.h"*/ #include "intel_ioctl.h" #include "intel_fbo.h" @@ -378,6 +378,31 @@ intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) } +static void +intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, + unsigned long long offset, GLint depth, GLuint pitch) +{ + abort(); +#if 0 + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct st_texture_object *stObj = st_texture_object(tObj); + + if (!stObj) + return; + + if (stObj->mt) + st_miptree_release(intel->pipe, &stObj->mt); + + stObj->imageOverride = GL_TRUE; + stObj->depthOverride = depth; + stObj->pitchOverride = pitch; + + if (offset) + stObj->textureOffset = offset; +#endif +} static const struct __DriverAPIRec intelAPI = { diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c index 5b25b2a43b..0937114c7f 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c +++ b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c @@ -31,7 +31,9 @@ #include "texstore.h" #include "enums.h" +/* #include "intel_context.h" +*/ #include "intel_tex.h" #include "state_tracker/st_context.h" #include "state_tracker/st_mipmap_tree.h" -- cgit v1.2.3 From 9ee61c97003ed78579d1238f552a3d815738c211 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 15:50:03 -0600 Subject: clean-ups --- src/mesa/drivers/dri/intel/intel_tex_layout.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c index 433a2c5f47..344aff396f 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.c +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c @@ -49,9 +49,9 @@ static int align(int value, int alignment) static void intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, - GLuint level, - GLuint nr_images, - GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) { assert(level < MAX_TEXTURE_LEVELS); @@ -74,8 +74,9 @@ intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, } assert(nr_images); + assert(!mt->level[level].image_offset); - mt->level[level].image_offset = malloc(nr_images * sizeof(GLuint)); + mt->level[level].image_offset = (GLuint *) malloc(nr_images * sizeof(GLuint)); mt->level[level].image_offset[0] = 0; } -- cgit v1.2.3 From ddece00faaf293602b3e0e1b28066b1bb9b91bb1 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 15:50:16 -0600 Subject: added st_cb_texture.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index bef9929f34..6590313aac 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -196,6 +196,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_program.c \ state_tracker/st_cb_teximage.c \ + state_tracker/st_cb_texture.c \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_draw.c \ state_tracker/st_context.c \ -- cgit v1.2.3 From f2034954a7ba360e742a325fe7eb4c2d7c9f9ea5 Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 16:00:54 +0100 Subject: Add path for GL includes in mesa tree. --- src/mesa/pipe/Makefile.template | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/Makefile.template b/src/mesa/pipe/Makefile.template index 84b4dcd13a..f2a9591de5 100644 --- a/src/mesa/pipe/Makefile.template +++ b/src/mesa/pipe/Makefile.template @@ -17,6 +17,7 @@ INCLUDES = \ -I$(TOP)/src/mesa/drivers/dri/common \ -I$(TOP)/src/mesa/main \ -I$(TOP)/src/mesa/pipe \ + -I$(TOP)/include \ -I$(TOP)/src/mesa -- cgit v1.2.3 From f79b90bc1962c6c40f806540c0bd1f867a9b80c7 Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 16:18:37 +0100 Subject: Translate MESA GLSL opcodes to TGSI opcodes. --- src/mesa/pipe/tgsi/core/tgsi_token.h | 2 +- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 27 +++++++++++++++++++++++++++ 2 files changed, 28 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index 40297f2575..12fc6df2ee 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -683,7 +683,7 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_NOP 98 /* TGSI_OPCODE_ABS */ /* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_BGNLOOP 99 +#define TGSI_OPCODE_BGNLOOP2 99 #define TGSI_OPCODE_BGNSUB 100 /* TGSI_OPCODE_BRA */ /* TGSI_OPCODE_BRK */ diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index a9415ac15b..aa14577642 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -369,6 +369,33 @@ compile_instruction( fullinst->Instruction.Opcode = TGSI_OPCODE_XPD; fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; break; + case OPCODE_NOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOP; + break; + case OPCODE_BGNLOOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2; + break; + case OPCODE_BGNSUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB; + break; + case OPCODE_ENDLOOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2; + break; + case OPCODE_ENDSUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB; + break; + case OPCODE_NOISE1: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE1; + break; + case OPCODE_NOISE2: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE2; + break; + case OPCODE_NOISE3: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE3; + break; + case OPCODE_NOISE4: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE4; + break; case OPCODE_END: return GL_TRUE; default: -- cgit v1.2.3 From a96d20eac81964cae98202c18e8368f04ec213d3 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 16:17:16 -0600 Subject: softpipe/winsys interface --- src/mesa/drivers/x11/xm_softpipe.c | 198 +++++++++++++++++++++++++++++++++++++ 1 file changed, 198 insertions(+) create mode 100644 src/mesa/drivers/x11/xm_softpipe.c (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_softpipe.c b/src/mesa/drivers/x11/xm_softpipe.c new file mode 100644 index 0000000000..6e8ef21b50 --- /dev/null +++ b/src/mesa/drivers/x11/xm_softpipe.c @@ -0,0 +1,198 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "glxheader.h" +#include "xmesaP.h" +#include "main/macros.h" + +#include "pipe/softpipe/sp_winsys.h" + + +struct xm_softpipe_winsys +{ + struct softpipe_winsys sws; + XMesaContext xmctx; /* not really needed */ +}; + + +struct xm_buffer +{ + int refcount; + void *data; + void *mapped; +}; + + + +/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static inline struct xm_buffer * +xm_bo( struct pipe_buffer_handle *bo ) +{ + return (struct xm_buffer *)bo; +} + +static inline struct pipe_buffer_handle * +pipe_bo( struct xm_buffer *bo ) +{ + return (struct pipe_buffer_handle *)bo; +} + +/* Turn a softpipe winsys into an xm/softpipe winsys: + */ +static inline struct xm_softpipe_winsys * +xm_softpipe_winsys( struct softpipe_winsys *sws ) +{ + return (struct xm_softpipe_winsys *)sws; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void *xm_buffer_map(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf ) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + xm_buf->mapped = xm_buf->data; + return xm_buf->mapped; +} + +static void xm_buffer_unmap(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + xm_buf->mapped = NULL; +} + + +static struct pipe_buffer_handle * +xm_buffer_reference(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + xm_buf->refcount++; + return buf; +} + +static void xm_buffer_unreference(struct softpipe_winsys *sws, + struct pipe_buffer_handle **buf) +{ + struct xm_buffer *xm_buf = xm_bo(*buf); + xm_buf->refcount--; + if (xm_buf->refcount == 0) { + if (xm_buf->data) { + free(xm_buf->data); + xm_buf->data = NULL; + } + free(xm_buf); + } + *buf = NULL; +} + + +static void xm_buffer_data(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + assert(!xm_buf->data); + xm_buf->data = malloc(size); +} + +static void xm_buffer_subdata(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + GLubyte *b = (GLubyte *) xm_buf->data; + assert(b); + memcpy(b + offset, data, size); +} + +static void xm_buffer_get_subdata(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) +{ + const struct xm_buffer *xm_buf = xm_bo(buf); + const GLubyte *b = (GLubyte *) xm_buf->data; + assert(b); + memcpy(data, b + offset, size); +} + + +/* Softpipe has no concept of pools. We choose the tex/region pool + * for all buffers. + */ +static struct pipe_buffer_handle * +xm_create_buffer(struct softpipe_winsys *sws, + unsigned alignment) +{ + struct xm_buffer *buffer; + + buffer = CALLOC_STRUCT(xm_buffer); + return pipe_bo(buffer); +} + + +struct pipe_context * +xmesa_create_softpipe(XMesaContext xmesa) +{ + struct xm_softpipe_winsys *isws = CALLOC_STRUCT( xm_softpipe_winsys ); + + /* Fill in this struct with callbacks that softpipe will need to + * communicate with the window system, buffer manager, etc. + * + * Softpipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate xm_buffer. + */ + isws->sws.create_buffer = xm_create_buffer; + isws->sws.buffer_map = xm_buffer_map; + isws->sws.buffer_unmap = xm_buffer_unmap; + isws->sws.buffer_reference = xm_buffer_reference; + isws->sws.buffer_unreference = xm_buffer_unreference; + isws->sws.buffer_data = xm_buffer_data; + isws->sws.buffer_subdata = xm_buffer_subdata; + isws->sws.buffer_get_subdata = xm_buffer_get_subdata; + + /* Create the softpipe context: + */ + return softpipe_create( &isws->sws ); +} -- cgit v1.2.3 From b1ad6289f83ce5c91da2ebc64d07506c25efa7e2 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 16:17:26 -0600 Subject: use new xmesa_create_softpipe() --- src/mesa/drivers/x11/xm_api.c | 3 ++- src/mesa/drivers/x11/xmesaP.h | 3 +++ 2 files changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 7fb99df5f4..6fbbb94f0f 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -1572,7 +1572,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) st_create_context( mesaCtx, - softpipe_create() ); + xmesa_create_softpipe( c ) ); + mesaCtx->Driver.Clear = xmesa_clear_buffers; /* mesaCtx->st->pipe->clear = xmesa_clear; diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index dd95aed4d0..99c87da3c5 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -598,5 +598,8 @@ xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value); extern void xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers); +extern struct pipe_context * +xmesa_create_softpipe(XMesaContext xm); + #endif -- cgit v1.2.3 From 8929c90a7173262fa74f1efcdc639a644da514c6 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 16:17:41 -0600 Subject: added xm_softpipe.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 6590313aac..0a69da9efc 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -302,6 +302,7 @@ X11_DRIVER_SOURCES = \ drivers/x11/xm_dd.c \ drivers/x11/xm_glide.c \ drivers/x11/xm_line.c \ + drivers/x11/xm_softpipe.c \ drivers/x11/xm_span.c \ drivers/x11/xm_surface.c \ drivers/x11/xm_tri.c -- cgit v1.2.3 From 2f35238394f366156e0ba413df5b757ad89e3c8d Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 17:00:55 +0100 Subject: Translate the remaining ops. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 113 ++++++++++++++++++++++++--------- 1 file changed, 83 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index aa14577642..8ce3931f5e 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -252,21 +252,36 @@ compile_instruction( case OPCODE_ADD: fullinst->Instruction.Opcode = TGSI_OPCODE_ADD; break; + case OPCODE_BGNLOOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2; + break; + case OPCODE_BGNSUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB; + break; + case OPCODE_BRA: + fullinst->Instruction.Opcode = TGSI_OPCODE_BRA; + break; + case OPCODE_BRK: + fullinst->Instruction.Opcode = TGSI_OPCODE_BRK; + break; case OPCODE_CMP: fullinst->Instruction.Opcode = TGSI_OPCODE_CMP; break; + case OPCODE_CONT: + fullinst->Instruction.Opcode = TGSI_OPCODE_CONT; + break; case OPCODE_COS: fullinst->Instruction.Opcode = TGSI_OPCODE_COS; break; - case OPCODE_DP3: - fullinst->Instruction.Opcode = TGSI_OPCODE_DP3; - break; case OPCODE_DDX: fullinst->Instruction.Opcode = TGSI_OPCODE_DDX; break; case OPCODE_DDY: fullinst->Instruction.Opcode = TGSI_OPCODE_DDY; break; + case OPCODE_DP3: + fullinst->Instruction.Opcode = TGSI_OPCODE_DP3; + break; case OPCODE_DP4: fullinst->Instruction.Opcode = TGSI_OPCODE_DP4; break; @@ -276,21 +291,45 @@ compile_instruction( case OPCODE_DST: fullinst->Instruction.Opcode = TGSI_OPCODE_DST; break; + case OPCODE_ELSE: + fullinst->Instruction.Opcode = TGSI_OPCODE_ELSE; + break; + case OPCODE_ENDIF: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF; + break; + case OPCODE_ENDLOOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2; + break; + case OPCODE_ENDSUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB; + break; case OPCODE_EX2: fullinst->Instruction.Opcode = TGSI_OPCODE_EX2; break; + case OPCODE_EXP: + fullinst->Instruction.Opcode = TGSI_OPCODE_EXP; + break; case OPCODE_FLR: fullinst->Instruction.Opcode = TGSI_OPCODE_FLR; break; case OPCODE_FRC: fullinst->Instruction.Opcode = TGSI_OPCODE_FRC; break; + case OPCODE_IF: + fullinst->Instruction.Opcode = TGSI_OPCODE_IF; + break; + case OPCODE_INT: + fullinst->Instruction.Opcode = TGSI_OPCODE_INT; + break; case OPCODE_KIL: fullinst->Instruction.Opcode = TGSI_OPCODE_KIL; break; case OPCODE_LG2: fullinst->Instruction.Opcode = TGSI_OPCODE_LG2; break; + case OPCODE_LOG: + fullinst->Instruction.Opcode = TGSI_OPCODE_LOG; + break; case OPCODE_LIT: fullinst->Instruction.Opcode = TGSI_OPCODE_LIT; break; @@ -312,6 +351,21 @@ compile_instruction( case OPCODE_MUL: fullinst->Instruction.Opcode = TGSI_OPCODE_MUL; break; + case OPCODE_NOISE1: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE1; + break; + case OPCODE_NOISE2: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE2; + break; + case OPCODE_NOISE3: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE3; + break; + case OPCODE_NOISE4: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE4; + break; + case OPCODE_NOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOP; + break; case OPCODE_POW: fullinst->Instruction.Opcode = TGSI_OPCODE_POW; break; @@ -328,15 +382,27 @@ compile_instruction( fullinst->Instruction.Opcode = TGSI_OPCODE_SCS; fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XY; break; + case OPCODE_SEQ: + fullinst->Instruction.Opcode = TGSI_OPCODE_SEQ; + break; case OPCODE_SGE: fullinst->Instruction.Opcode = TGSI_OPCODE_SGE; break; + case OPCODE_SGT: + fullinst->Instruction.Opcode = TGSI_OPCODE_SGT; + break; case OPCODE_SIN: fullinst->Instruction.Opcode = TGSI_OPCODE_SIN; break; + case OPCODE_SLE: + fullinst->Instruction.Opcode = TGSI_OPCODE_SLE; + break; case OPCODE_SLT: fullinst->Instruction.Opcode = TGSI_OPCODE_SLT; break; + case OPCODE_SNE: + fullinst->Instruction.Opcode = TGSI_OPCODE_SNE; + break; case OPCODE_SUB: fullinst->Instruction.Opcode = TGSI_OPCODE_SUB; break; @@ -357,6 +423,20 @@ compile_instruction( fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; break; + case OPCODE_TXD: + fullinst->Instruction.Opcode = TGSI_OPCODE_TXD; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXL: + fullinst->Instruction.Opcode = TGSI_OPCODE_TXL; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; case OPCODE_TXP: fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; fullinst->Instruction.NumSrcRegs = 2; @@ -369,33 +449,6 @@ compile_instruction( fullinst->Instruction.Opcode = TGSI_OPCODE_XPD; fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; break; - case OPCODE_NOP: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOP; - break; - case OPCODE_BGNLOOP: - fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2; - break; - case OPCODE_BGNSUB: - fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB; - break; - case OPCODE_ENDLOOP: - fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2; - break; - case OPCODE_ENDSUB: - fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB; - break; - case OPCODE_NOISE1: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE1; - break; - case OPCODE_NOISE2: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE2; - break; - case OPCODE_NOISE3: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE3; - break; - case OPCODE_NOISE4: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE4; - break; case OPCODE_END: return GL_TRUE; default: -- cgit v1.2.3 From 1d45082f6e8fb2416fd37f6f1fa48fc21576cd1d Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 17:03:51 +0100 Subject: Reorder opcodes. --- src/mesa/pipe/tgsi/core/tgsi_token.h | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index 12fc6df2ee..c442457cd5 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -680,11 +680,10 @@ struct tgsi_immediate_float32 /* * GLSL */ -#define TGSI_OPCODE_NOP 98 /* TGSI_OPCODE_ABS */ /* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_BGNLOOP2 99 -#define TGSI_OPCODE_BGNSUB 100 +#define TGSI_OPCODE_BGNLOOP2 98 +#define TGSI_OPCODE_BGNSUB 99 /* TGSI_OPCODE_BRA */ /* TGSI_OPCODE_BRK */ /* TGSI_OPCODE_CONT */ @@ -695,8 +694,8 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_DP4 */ /* TGSI_OPCODE_ELSE */ /* TGSI_OPCODE_ENDIF */ -#define TGSI_OPCODE_ENDLOOP2 101 -#define TGSI_OPCODE_ENDSUB 102 +#define TGSI_OPCODE_ENDLOOP2 100 +#define TGSI_OPCODE_ENDSUB 101 /* TGSI_OPCODE_EX2 */ /* TGSI_OPCODE_EXP */ /* TGSI_OPCODE_FLR */ @@ -711,10 +710,11 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MIN */ /* TGSI_OPCODE_MOV */ /* TGSI_OPCODE_MUL */ -#define TGSI_OPCODE_NOISE1 103 -#define TGSI_OPCODE_NOISE2 104 -#define TGSI_OPCODE_NOISE3 105 -#define TGSI_OPCODE_NOISE4 106 +#define TGSI_OPCODE_NOISE1 102 +#define TGSI_OPCODE_NOISE2 103 +#define TGSI_OPCODE_NOISE3 104 +#define TGSI_OPCODE_NOISE4 105 +#define TGSI_OPCODE_NOP 106 /* TGSI_OPCODE_POW */ /* TGSI_OPCODE_RCP */ /* TGSI_OPCODE_RSQ */ @@ -764,7 +764,7 @@ struct tgsi_instruction GLuint Extended : 1; /* BOOL */ }; -/* +/*1 * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows. * * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow. -- cgit v1.2.3 From 555ff452602fc572420b590eac690d2e3849ca3c Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 17:04:34 +0100 Subject: Use fp_attr_to_slot[]. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 1676d47d51..66513bb436 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -39,11 +39,15 @@ #include "sp_quad.h" #include "tgsi/core/tgsi_core.h" +#if 0 #if defined __GNUC__ #define ALIGNED_ATTRIBS 1 #else #define ALIGNED_ATTRIBS 0 #endif +#else +#define ALIGNED_ATTRIBS 0 +#endif struct exec_machine { const struct setup_coefficient *coef; /**< will point to quad->coef */ @@ -229,6 +233,7 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) /* load input registers */ for (i = 0; i < softpipe->nr_attrs; i++) { +#if 0 /* Make sure fp_attr_to_slot[] is an identity transform. */ assert( softpipe->fp_attr_to_slot[i] == i ); @@ -236,6 +241,12 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) &ainputs[i], exec.attr[i], sizeof( ainputs[0] ) ); +#else + memcpy( + &ainputs[i], + exec.attr[softpipe->fp_attr_to_slot[i]], + sizeof( ainputs[0] ) ); +#endif } #endif -- cgit v1.2.3 From 10fd2742c698d8f4a43b3a7b9506e8d95bf56f4f Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 17:06:03 +0100 Subject: Remove garbage text. --- src/mesa/pipe/tgsi/core/tgsi_token.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index c442457cd5..632d678818 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -764,7 +764,7 @@ struct tgsi_instruction GLuint Extended : 1; /* BOOL */ }; -/*1 +/* * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows. * * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow. -- cgit v1.2.3 From 76478a470660becaceb09c7217b8c73601b1e92a Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 17:15:37 +0100 Subject: Make SGE, SGT, SLE working. --- src/mesa/pipe/tgsi/core/tgsi_exec.c | 44 +++++++++++++++++++++++++++---------- 1 file changed, 32 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 6aaaef98a8..9a42878e0a 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -357,6 +357,20 @@ micro_frc( dst->f[3] = src->f[3] - (GLfloat) floor( (GLdouble) src->f[3] ); } +static void +micro_ge( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; +} + static void micro_i2f( union tgsi_exec_channel *dst, @@ -1508,12 +1522,10 @@ exec_instruction( case TGSI_OPCODE_SGE: /* TGSI_OPCODE_SETGE */ FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - - STORE(&r[0], 0, chan_index); + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); } break; @@ -1785,21 +1797,29 @@ exec_instruction( break; case TGSI_OPCODE_SGT: - assert (0); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } break; case TGSI_OPCODE_SIN: - FETCH(&r[0], 0, CHAN_X); - + FETCH( &r[0], 0, CHAN_X ); micro_sin( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); } break; case TGSI_OPCODE_SLE: - assert (0); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } break; case TGSI_OPCODE_SNE: -- cgit v1.2.3 From 2cd6291c624625368050d7c6104e11f01c605ebb Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 17:18:25 +0100 Subject: Add GLSL ops. --- src/mesa/pipe/tgsi/core/tgsi_exec.c | 36 ++++++++++++++++++++++++++++++++++++ 1 file changed, 36 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 9a42878e0a..7cf48021de 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -2103,6 +2103,7 @@ exec_instruction( break; case TGSI_OPCODE_TRUNC: + /* TGSI_OPCODE_INT */ assert (0); break; @@ -2181,6 +2182,41 @@ exec_instruction( mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; break; + case TGSI_OPCODE_BGNLOOP2: + assert( 0 ); + break; + + case TGSI_OPCODE_BGNSUB: + assert( 0 ); + break; + + case TGSI_OPCODE_ENDLOOP2: + assert( 0 ); + break; + + case TGSI_OPCODE_ENDSUB: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE1: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE2: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE3: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE4: + assert( 0 ); + break; + + case TGSI_OPCODE_NOP: + break; + default: assert( 0 ); } -- cgit v1.2.3 From 3f26f6167874ba59a132cc12d77a2548c8d1d309 Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 6 Aug 2007 17:20:31 +0100 Subject: Add GLSL opcodes. --- src/mesa/pipe/tgsi/core/tgsi_dump.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index 33a898ea04..bdffc59d39 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -276,7 +276,16 @@ static const char *TGSI_OPCODES[] = "OPCODE_TXQ", "OPCODE_CONT", "OPCODE_EMIT", - "OPCODE_ENDPRIM" + "OPCODE_ENDPRIM", + "OPCODE_BGNLOOP2", + "OPCODE_BGNSUB", + "OPCODE_ENDLOOP2", + "OPCODE_ENDSUB", + "OPCODE_NOISE1", + "OPCODE_NOISE2", + "OPCODE_NOISE3", + "OPCODE_NOISE4", + "OPCODE_NOP" }; static const char *TGSI_SATS[] = -- cgit v1.2.3 From a2e3def40c0750db158e2cc0e1e90bfb30bd5a5c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 20:49:44 +0100 Subject: fix asst. warnings --- src/mesa/state_tracker/st_mipmap_tree.c | 22 ++++++++++++---------- 1 file changed, 12 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index 6293cc8f51..ac74335ba1 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -98,7 +98,7 @@ st_miptree_reference(struct pipe_mipmap_tree **dst, { src->refcount++; *dst = src; - DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); + DBG("%s %p refcount now %d\n", __FUNCTION__, (void *) src, src->refcount); } void @@ -108,11 +108,12 @@ st_miptree_release(struct pipe_context *pipe, if (!*mt) return; - DBG("%s %p refcount will be %d\n", __FUNCTION__, *mt, (*mt)->refcount - 1); + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *mt, (*mt)->refcount - 1); if (--(*mt)->refcount <= 0) { GLuint i; - DBG("%s deleting %p\n", __FUNCTION__, *mt); + DBG("%s deleting %p\n", __FUNCTION__, (void *) *mt); pipe->region_release(pipe, &((*mt)->region)); @@ -244,17 +245,18 @@ st_miptree_image_unmap(struct pipe_context *pipe, */ void st_miptree_image_data(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, - GLuint level, - void *src, - GLuint src_row_pitch, GLuint src_image_pitch) + struct pipe_mipmap_tree *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch) { GLuint depth = dst->level[level].depth; GLuint dst_offset = st_miptree_image_offset(dst, face, level); const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); GLuint i; GLuint height = 0; + const GLubyte *srcUB = src; DBG("%s\n", __FUNCTION__); for (i = 0; i < depth; i++) { @@ -264,12 +266,12 @@ st_miptree_image_data(struct pipe_context *pipe, pipe->region_data(pipe, dst->region, dst_offset + dst_depth_offset[i], /* dst_offset */ 0, 0, /* dstx, dsty */ - src, + srcUB, src_row_pitch, 0, 0, /* source x, y */ dst->level[level].width, height); /* width, height */ - src += src_image_pitch * dst->cpp; + srcUB += src_image_pitch * dst->cpp; } } -- cgit v1.2.3 From 8ebf91d15eeb6f19d19580f6319477aaad4df3fd Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 20:52:04 +0100 Subject: new header --- src/mesa/state_tracker/st_cb_program.h | 10 ++++++++++ 1 file changed, 10 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_program.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.h b/src/mesa/state_tracker/st_cb_program.h new file mode 100644 index 0000000000..38ec4e0a59 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_program.h @@ -0,0 +1,10 @@ + +#ifndef ST_CB_PROGRAM_H +#define ST_CB_PROGRAM_H + + +extern void +st_init_program_functions(struct dd_function_table *functions); + + +#endif -- cgit v1.2.3 From 6da9234fd437f97267e7831f034c78b31156d939 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 20:53:28 +0100 Subject: New st_init_*_functions() to initialize the driver functions table. We need to do these initializations before initializing the Mesa context because context init involves creating texture/program/etc objects. --- src/mesa/state_tracker/st_cb_bufferobjects.c | 18 ++++++++--------- src/mesa/state_tracker/st_cb_bufferobjects.h | 11 +++++------ src/mesa/state_tracker/st_cb_clear.c | 10 +--------- src/mesa/state_tracker/st_cb_clear.h | 5 +++-- src/mesa/state_tracker/st_cb_drawpixels.c | 8 +------- src/mesa/state_tracker/st_cb_drawpixels.h | 4 +--- src/mesa/state_tracker/st_cb_fbo.c | 9 +-------- src/mesa/state_tracker/st_cb_fbo.h | 5 ++--- src/mesa/state_tracker/st_cb_program.c | 29 ++++++++++------------------ src/mesa/state_tracker/st_cb_texture.c | 12 +++--------- src/mesa/state_tracker/st_cb_texture.h | 6 +----- src/mesa/state_tracker/st_context.c | 18 +++++++++++++++++ src/mesa/state_tracker/st_context.h | 7 +++---- src/mesa/state_tracker/st_program.h | 5 +++-- 14 files changed, 60 insertions(+), 87 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index a667b3e775..d020eb2007 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -192,15 +192,13 @@ st_bufferobj_unmap(GLcontext *ctx, void -st_init_cb_bufferobjects( struct st_context *st ) +st_init_bufferobject_functions(struct dd_function_table *functions) { - GLcontext *ctx = st->ctx; - - ctx->Driver.NewBufferObject = st_bufferobj_alloc; - ctx->Driver.DeleteBuffer = st_bufferobj_free; - ctx->Driver.BufferData = st_bufferobj_data; - ctx->Driver.BufferSubData = st_bufferobj_subdata; - ctx->Driver.GetBufferSubData = st_bufferobj_get_subdata; - ctx->Driver.MapBuffer = st_bufferobj_map; - ctx->Driver.UnmapBuffer = st_bufferobj_unmap; + functions->NewBufferObject = st_bufferobj_alloc; + functions->DeleteBuffer = st_bufferobj_free; + functions->BufferData = st_bufferobj_data; + functions->BufferSubData = st_bufferobj_subdata; + functions->GetBufferSubData = st_bufferobj_get_subdata; + functions->MapBuffer = st_bufferobj_map; + functions->UnmapBuffer = st_bufferobj_unmap; } diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.h b/src/mesa/state_tracker/st_cb_bufferobjects.h index 2787411c5f..2090a743e0 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.h +++ b/src/mesa/state_tracker/st_cb_bufferobjects.h @@ -1,4 +1,4 @@ - /************************************************************************** +/************************************************************************** * * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. @@ -43,11 +43,6 @@ struct st_buffer_object }; -/* Hook the bufferobject implementation into mesa: - */ -void st_init_cb_bufferobjects( struct st_context *st ); - - /* Are the obj->Name tests necessary? Unfortunately yes, mesa * allocates a couple of gl_buffer_object structs statically, and the * Name == 0 test is the only way to identify them and avoid casting @@ -63,4 +58,8 @@ st_buffer_object(struct gl_buffer_object *obj) } +extern void +st_init_bufferobject_functions(struct dd_function_table *functions); + + #endif diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index c907b0ed22..0ec7784d84 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -418,15 +418,7 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) } -void st_init_cb_clear( struct st_context *st ) +void st_init_clear_functions(struct dd_function_table *functions) { - struct dd_function_table *functions = &st->ctx->Driver; - functions->Clear = st_clear; } - - -void st_destroy_cb_clear( struct st_context *st ) -{ -} - diff --git a/src/mesa/state_tracker/st_cb_clear.h b/src/mesa/state_tracker/st_cb_clear.h index 32086971b5..c715e56bd5 100644 --- a/src/mesa/state_tracker/st_cb_clear.h +++ b/src/mesa/state_tracker/st_cb_clear.h @@ -29,9 +29,10 @@ #ifndef ST_CB_CLEAR_H #define ST_CB_CLEAR_H -extern void st_init_cb_clear( struct st_context *st ); -extern void st_destroy_cb_clear( struct st_context *st ); +extern void +st_init_clear_functions(struct dd_function_table *functions); + #endif /* ST_CB_CLEAR_H */ diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 13f5c5f3c7..92a4e305d1 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -262,14 +262,8 @@ st_drawpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } -void st_init_cb_drawpixels( struct st_context *st ) +void st_init_drawpixels_functions(struct dd_function_table *functions) { - struct dd_function_table *functions = &st->ctx->Driver; - functions->DrawPixels = st_drawpixels; } - -void st_destroy_cb_drawpixels( struct st_context *st ) -{ -} diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h index 8c36aaa931..71ba487020 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.h +++ b/src/mesa/state_tracker/st_cb_drawpixels.h @@ -30,9 +30,7 @@ #define ST_CB_DRAWPIXELS_H -void st_init_cb_drawpixels( struct st_context *st ); - -void st_destroy_cb_drawpixels( struct st_context *st ); +extern void st_init_drawpixels_functions(struct dd_function_table *functions); #endif /* ST_CB_DRAWPIXELS_H */ diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 6b9ae88dbe..d0205fd635 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -322,10 +322,8 @@ st_finish_render_texture(GLcontext *ctx, -void st_init_cb_fbo( struct st_context *st ) +void st_init_fbo_functions(struct dd_function_table *functions) { - struct dd_function_table *functions = &st->ctx->Driver; - functions->NewFramebuffer = st_new_framebuffer; functions->NewRenderbuffer = st_new_renderbuffer; functions->BindFramebuffer = st_bind_framebuffer; @@ -336,8 +334,3 @@ void st_init_cb_fbo( struct st_context *st ) functions->ResizeBuffers = st_resize_buffers; */ } - - -void st_destroy_cb_fbo( struct st_context *st ) -{ -} diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index f4fa66df59..6142434ec6 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -30,9 +30,8 @@ #define ST_CB_FBO_H -extern void st_init_cb_fbo( struct st_context *st ); - -extern void st_destroy_cb_fbo( struct st_context *st ); +extern void +st_init_fbo_functions(struct dd_function_table *functions); #endif /* ST_CB_FBO_H */ diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 6da2aeb2f2..ed47c12066 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -32,7 +32,6 @@ #include "st_context.h" #include "st_program.h" - #include "glheader.h" #include "macros.h" #include "enums.h" @@ -44,6 +43,11 @@ #include "pipe/tgsi/mesa/tgsi_mesa.h" +/* Counter to track program string changes: + */ +static GLuint program_id = 0; + + static void st_bind_program( GLcontext *ctx, GLenum target, struct gl_program *prog ) @@ -70,7 +74,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *prog = CALLOC_STRUCT(st_vertex_program); - prog->id = st->program_id++; + prog->id = program_id++; prog->dirty = 1; return _mesa_init_vertex_program( ctx, @@ -84,7 +88,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, { struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program); - prog->id = st->program_id++; + prog->id = program_id++; prog->dirty = 1; return _mesa_init_fragment_program( ctx, @@ -124,7 +128,7 @@ static void st_program_string_notify( GLcontext *ctx, if (prog == &ctx->FragmentProgram._Current->Base) st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - p->id = st->program_id++; + p->id = program_id++; p->param_state = p->Base.Base.Parameters->StateFlags; } else if (target == GL_VERTEX_PROGRAM_ARB) { @@ -133,7 +137,7 @@ static void st_program_string_notify( GLcontext *ctx, if (prog == &ctx->VertexProgram._Current->Base) st->dirty.st |= ST_NEW_VERTEX_PROGRAM; - p->id = st->program_id++; + p->id = program_id++; p->param_state = p->Base.Base.Parameters->StateFlags; /* Also tell tnl about it: @@ -144,15 +148,8 @@ static void st_program_string_notify( GLcontext *ctx, -void st_init_cb_program( struct st_context *st ) +void st_init_program_functions(struct dd_function_table *functions) { - struct dd_function_table *functions = &st->ctx->Driver; - - /* Need these flags: - */ - st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; - st->ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; - #if 0 assert(functions->ProgramStringNotify == _tnl_program_string); #endif @@ -162,9 +159,3 @@ void st_init_cb_program( struct st_context *st ) functions->IsProgramNative = st_is_program_native; functions->ProgramStringNotify = st_program_string_notify; } - - -void st_destroy_cb_program( struct st_context *st ) -{ -} - diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index a0245b553f..5872ae3e74 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1241,7 +1241,7 @@ do_copy_texsubimage(GLcontext *ctx, get_teximage_source(ctx, internalFormat); if (!stImage->mt || !src) { - DBG("%s fail %p %p\n", __FUNCTION__, stImage->mt, src); + DBG("%s fail %p %p\n", __FUNCTION__, (void *) stImage->mt, (void *) src); return GL_FALSE; } @@ -1726,10 +1726,9 @@ st_tex_unmap_images(struct pipe_context *pipe, -void st_init_cb_texture( struct st_context *st ) +void +st_init_texture_functions(struct dd_function_table *functions) { - struct dd_function_table *functions = &st->ctx->Driver; - functions->ChooseTextureFormat = st_ChooseTextureFormat; functions->TexImage1D = st_TexImage1D; functions->TexImage2D = st_TexImage2D; @@ -1756,8 +1755,3 @@ void st_init_cb_texture( struct st_context *st ) functions->TextureMemCpy = do_memcpy; } - - -void st_destroy_cb_texture( struct st_context *st ) -{ -} diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h index c474d16465..c732881c39 100644 --- a/src/mesa/state_tracker/st_cb_texture.h +++ b/src/mesa/state_tracker/st_cb_texture.h @@ -9,11 +9,7 @@ st_finalize_mipmap_tree(GLcontext *ctx, extern void -st_init_cb_texture( struct st_context *st ); - - -extern void -st_destroy_cb_texture( struct st_context *st ); +st_init_texture_functions(struct dd_function_table *functions); #endif /* ST_CB_TEXTURE_H */ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 2b96286770..0ea06c692d 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -28,6 +28,7 @@ #include "imports.h" #include "st_public.h" #include "st_context.h" +#include "st_cb_bufferobjects.h" #include "st_cb_clear.h" #include "st_cb_drawpixels.h" #include "st_cb_texture.h" @@ -61,10 +62,17 @@ struct st_context *st_create_context( GLcontext *ctx, st_init_atoms( st ); st_init_draw( st ); + /* Need these flags: + */ + st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; + st->ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; + +#if 0 st_init_cb_clear( st ); st_init_cb_program( st ); st_init_cb_drawpixels( st ); st_init_cb_texture( st ); +#endif return st; } @@ -75,11 +83,13 @@ void st_destroy_context( struct st_context *st ) st_destroy_atoms( st ); st_destroy_draw( st ); +#if 0 st_destroy_cb_clear( st ); st_destroy_cb_program( st ); st_destroy_cb_drawpixels( st ); /*st_destroy_cb_teximage( st );*/ st_destroy_cb_texture( st ); +#endif st->pipe->destroy( st->pipe ); FREE( st ); @@ -87,3 +97,11 @@ void st_destroy_context( struct st_context *st ) +void st_init_driver_functions(struct dd_function_table *functions) +{ + st_init_bufferobject_functions(functions); + st_init_clear_functions(functions); + st_init_drawpixels_functions(functions); + st_init_program_functions(functions); + st_init_texture_functions(functions); +} diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index ef3cdb3b09..fe73630c75 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -97,10 +97,6 @@ struct st_context struct st_state_flags dirty; - /* Counter to track program string changes: - */ - GLuint program_id; - GLfloat polygon_offset_scale; /* ?? */ }; @@ -113,4 +109,7 @@ static INLINE struct st_context *st_context(GLcontext *ctx) } +extern void st_init_driver_functions(struct dd_function_table *functions); + + #endif diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index f6d5f6d76c..8dcb2ceb48 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -87,8 +87,9 @@ struct st_vertex_program GLuint param_state; }; -void st_init_cb_program( struct st_context *st ); -void st_destroy_cb_program( struct st_context *st ); + +extern void st_init_program_functions(struct dd_function_table *functions); + static inline struct st_fragment_program * st_fragment_program( struct gl_fragment_program *fp ) -- cgit v1.2.3 From 07633371ff4987d3149e78a1f0791da720af4e60 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 20:54:03 +0100 Subject: need a softpipe func for texture layout --- src/mesa/pipe/softpipe/sp_tex_layout.c | 13 +++++++++++++ src/mesa/pipe/softpipe/sp_tex_layout.h | 16 ++++++++++++++++ 2 files changed, 29 insertions(+) create mode 100644 src/mesa/pipe/softpipe/sp_tex_layout.c create mode 100644 src/mesa/pipe/softpipe/sp_tex_layout.h (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c new file mode 100644 index 0000000000..2bbdca0f58 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tex_layout.c @@ -0,0 +1,13 @@ + +#include "pipe/p_state.h" +#include "sp_tex_layout.h" + + +GLboolean +softpipe_mipmap_tree_layout(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt) +{ + return GL_TRUE; +} + + diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.h b/src/mesa/pipe/softpipe/sp_tex_layout.h new file mode 100644 index 0000000000..be85e4be58 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tex_layout.h @@ -0,0 +1,16 @@ +#ifndef SP_TEX_LAYOUT_H +#define SP_TEX_LAYOUT_H + + +struct pipe_context; +struct pipe_mipmap_tree; + + +extern GLboolean +softpipe_mipmap_tree_layout(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt); + + +#endif /* SP_TEX_LAYOUT_H */ + + -- cgit v1.2.3 From 4a60b94701a54a7110766a184b048106f8fe4628 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 20:54:18 +0100 Subject: added softpipe_mipmap_tree_layout --- src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_context.c | 9 ++++++--- 2 files changed, 7 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 0eacfa2182..3e815aa467 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -29,6 +29,7 @@ DRIVER_SOURCES = \ sp_state_sampler.c \ sp_state_setup.c \ sp_state_surface.c \ + sp_tex_layout.c \ sp_surface.c C_SOURCES = \ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 9352c7afa7..139afde376 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -33,13 +33,14 @@ #include "main/macros.h" #include "pipe/draw/draw_context.h" #include "pipe/p_defines.h" -#include "sp_context.h" +#include "sp_buffer.h" #include "sp_clear.h" +#include "sp_context.h" +#include "sp_prim_setup.h" #include "sp_region.h" -#include "sp_buffer.h" #include "sp_state.h" #include "sp_surface.h" -#include "sp_prim_setup.h" +#include "sp_tex_layout.h" #include "sp_winsys.h" @@ -199,6 +200,8 @@ struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; + softpipe->pipe.mipmap_tree_layout = softpipe_mipmap_tree_layout; + softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); softpipe->quad.shade = sp_quad_shade_stage(softpipe); softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); -- cgit v1.2.3 From 34ff87e46af23919cf53e9a4cf551fa154535f0b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 20:54:41 +0100 Subject: call st_init_driver_functions() --- src/mesa/drivers/x11/xm_api.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 6fbbb94f0f..88e6bf8d6d 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -1515,6 +1515,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) /* initialize with default driver functions, then plug in XMesa funcs */ _mesa_init_driver_functions(&functions); xmesa_init_driver_functions(v, &functions); + st_init_driver_functions(&functions); + if (!_mesa_initialize_context(mesaCtx, &v->mesa_visual, share_list ? &(share_list->mesa) : (GLcontext *) NULL, &functions, (void *) c)) { -- cgit v1.2.3 From 647afc3179b6c2b3f35a9c4d388f7e0289a86404 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 6 Aug 2007 20:56:29 +0100 Subject: call st_init_driver_functions(), make intelInitDriverFunctions() static --- src/mesa/drivers/dri/i915pipe/intel_context.c | 4 +++- src/mesa/drivers/dri/i915pipe/intel_context.h | 2 -- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index ced4366009..a70b0ad729 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -308,7 +308,7 @@ intelFinish(GLcontext * ctx) } -void +static void intelInitDriverFunctions(struct dd_function_table *functions) { _mesa_init_driver_functions(functions); @@ -322,6 +322,8 @@ intelInitDriverFunctions(struct dd_function_table *functions) intelInitTextureFuncs(functions); */ intelInitBufferFuncs(functions); + + st_init_driver_functions(functions); } diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.h b/src/mesa/drivers/dri/i915pipe/intel_context.h index ad55e7eef2..1a4e7ca1cd 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.h +++ b/src/mesa/drivers/dri/i915pipe/intel_context.h @@ -193,8 +193,6 @@ extern void intelGetLock(struct intel_context *intel, GLuint flags); extern void intelFinish(GLcontext * ctx); extern void intelFlush(GLcontext * ctx); -extern void intelInitDriverFunctions(struct dd_function_table *functions); - /* ================================================================ * intel_softpipe.c: -- cgit v1.2.3 From e43bc8da7a03462f1f3027fa308f7577fdb10156 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 09:55:13 -0600 Subject: get rid of temp write_mono_row_ub() function --- src/mesa/drivers/x11/xm_surface.c | 3 ++- src/mesa/pipe/softpipe/sp_surface.h | 4 ---- 2 files changed, 2 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index e8e795c00f..6b52b33bf2 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -150,6 +150,7 @@ write_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, rgba + 2); } +#if 0 static void write_mono_row_ub(struct softpipe_surface *sps, GLuint count, GLint x, GLint y, GLubyte rgba[NUM_CHANNELS]) @@ -158,6 +159,7 @@ write_mono_row_ub(struct softpipe_surface *sps, GLuint count, GLint x, GLint y, GET_CURRENT_CONTEXT(ctx); xrb->Base.PutMonoRow(ctx, &xrb->Base, count, x, y, rgba, NULL); } +#endif /** @@ -185,7 +187,6 @@ xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb) sps->write_quad_f = write_quad_f; sps->write_quad_f_swz = write_quad_f_swz; sps->write_quad_ub = write_quad_ub; - sps->write_mono_row_ub = write_mono_row_ub; /* Note, the region we allocate doesn't actually have any storage * since we're drawing into an XImage or Pixmap. diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index e8466256db..6782bba376 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -73,10 +73,6 @@ struct softpipe_surface { GLint x, GLint y, GLubyte (*rgba)[NUM_CHANNELS] ); - void (*write_mono_row_ub)( struct softpipe_surface *, - GLuint count, GLint x, GLint y, - GLubyte rgba[NUM_CHANNELS] ); - void (*read_quad_z)(struct softpipe_surface *, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]); void (*write_quad_z)(struct softpipe_surface *, -- cgit v1.2.3 From 958a1ad4200521f4fefa87196ee56be0f0c61613 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 10:07:23 -0600 Subject: prototypes --- src/mesa/drivers/dri/intel/intel_tex_layout.h | 44 ++++++--------------------- 1 file changed, 9 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h index ed30e2ef76..83913b89cb 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.h +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h @@ -1,42 +1,16 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Michel Dänzer - */ - -#include "macros.h" +#ifndef INTEL_TEX_LAYOUT_H +#define INTEL_TEX_LAYOUT_H struct pipe_context; struct pipe_mipmap_tree; -extern void i915_miptree_layout_2d( struct pipe_context *, struct pipe_mipmap_tree *mt ); +extern GLboolean +i915_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); + +extern GLboolean +i945_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); + -extern void i945_miptree_layout_2d( struct pipe_context *, struct pipe_mipmap_tree *mt ); +#endif /* INTEL_TEX_LAYOUT_H */ -- cgit v1.2.3 From 285c7d593170506409ef78adb40f3bb849c93e5d Mon Sep 17 00:00:00 2001 From: michal Date: Tue, 7 Aug 2007 10:14:50 +0100 Subject: Labels are always label targets. Translate label for GLSL IF, ELSE. --- src/mesa/pipe/tgsi/core/tgsi_build.c | 4 ---- src/mesa/pipe/tgsi/core/tgsi_build.h | 1 - src/mesa/pipe/tgsi/core/tgsi_dump.c | 4 ---- src/mesa/pipe/tgsi/core/tgsi_exec.c | 3 +-- src/mesa/pipe/tgsi/core/tgsi_token.h | 3 +-- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 ++ 6 files changed, 4 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.c b/src/mesa/pipe/tgsi/core/tgsi_build.c index 2a482a78dc..bbf4a89a5e 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_build.c +++ b/src/mesa/pipe/tgsi/core/tgsi_build.c @@ -528,7 +528,6 @@ tgsi_build_full_instruction( *instruction_ext_label = tgsi_build_instruction_ext_label( full_inst->InstructionExtLabel.Label, - full_inst->InstructionExtLabel.Target, prev_token, instruction, header ); @@ -835,7 +834,6 @@ tgsi_default_instruction_ext_label( void ) instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; instruction_ext_label.Label = 0; - instruction_ext_label.Target = 0; instruction_ext_label.Padding = 0; instruction_ext_label.Extended = 0; @@ -855,7 +853,6 @@ tgsi_compare_instruction_ext_label( struct tgsi_instruction_ext_label tgsi_build_instruction_ext_label( GLuint label, - GLuint target, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ) @@ -864,7 +861,6 @@ tgsi_build_instruction_ext_label( instruction_ext_label = tgsi_default_instruction_ext_label(); instruction_ext_label.Label = label; - instruction_ext_label.Target = target; prev_token->Extended = 1; instruction_grow( instruction, header ); diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.h b/src/mesa/pipe/tgsi/core/tgsi_build.h index db25956519..35c1a2506d 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_build.h +++ b/src/mesa/pipe/tgsi/core/tgsi_build.h @@ -160,7 +160,6 @@ tgsi_compare_instruction_ext_label( struct tgsi_instruction_ext_label tgsi_build_instruction_ext_label( GLuint label, - GLuint target, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index bdffc59d39..882771da2e 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -618,10 +618,6 @@ tgsi_dump( TXT( "\nLabel : " ); UID( inst->InstructionExtLabel.Label ); } - if( deflt || fi.InstructionExtLabel.Target != inst->InstructionExtLabel.Target ) { - TXT( "\nTarget : " ); - UID( inst->InstructionExtLabel.Target ); - } if( ignored ) { TXT( "\nPadding : " ); UIX( inst->InstructionExtLabel.Padding ); diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 7cf48021de..37ca7c1b2a 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -134,8 +134,7 @@ tgsi_exec_prepare( mach->ImmLimit += (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4; break; case TGSI_TOKEN_TYPE_INSTRUCTION: - if( parse.FullToken.FullInstruction.InstructionExtLabel.Label && - parse.FullToken.FullInstruction.InstructionExtLabel.Target ) { + if( parse.FullToken.FullInstruction.InstructionExtLabel.Label ) { assert( labels->count < 128 ); labels->labels[labels->count][0] = parse.FullToken.FullInstruction.InstructionExtLabel.Label; labels->labels[labels->count][1] = pointer; diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index 632d678818..fd59ea1e2e 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -855,8 +855,7 @@ struct tgsi_instruction_ext_label { GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ GLuint Label : 24; /* UINT */ - GLuint Target : 1; /* BOOL */ - GLuint Padding : 2; + GLuint Padding : 3; GLuint Extended : 1; /* BOOL */ }; diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 8ce3931f5e..47fd5791e1 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -293,6 +293,7 @@ compile_instruction( break; case OPCODE_ELSE: fullinst->Instruction.Opcode = TGSI_OPCODE_ELSE; + fullinst->InstructionExtLabel.Label = inst->BranchTarget; break; case OPCODE_ENDIF: fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF; @@ -317,6 +318,7 @@ compile_instruction( break; case OPCODE_IF: fullinst->Instruction.Opcode = TGSI_OPCODE_IF; + fullinst->InstructionExtLabel.Label = inst->BranchTarget; break; case OPCODE_INT: fullinst->Instruction.Opcode = TGSI_OPCODE_INT; -- cgit v1.2.3 From 87a8f5643dc228b0e98e15dd42db20d54bc65a62 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 7 Aug 2007 10:17:22 +0100 Subject: Remove old intel_buffer_objects code - there is a state_tracker version now. --- src/mesa/drivers/dri/i915pipe/Makefile | 1 - .../drivers/dri/i915pipe/intel_buffer_objects.c | 205 --------------------- .../drivers/dri/i915pipe/intel_buffer_objects.h | 82 --------- src/mesa/drivers/dri/i915pipe/intel_context.c | 1 - 4 files changed, 289 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile index 3f3e04e978..e4e9cf17b0 100644 --- a/src/mesa/drivers/dri/i915pipe/Makefile +++ b/src/mesa/drivers/dri/i915pipe/Makefile @@ -11,7 +11,6 @@ PIPE_DRIVERS = \ DRIVER_SOURCES = \ intel_softpipe.c \ - intel_buffer_objects.c \ intel_batchbuffer.c \ intel_tex_layout.c \ intel_buffers.c \ diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c deleted file mode 100644 index fdcfdca9bb..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.c +++ /dev/null @@ -1,205 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "imports.h" -#include "mtypes.h" -#include "bufferobj.h" - -#include "intel_context.h" -#include "intel_buffer_objects.h" -#include "dri_bufmgr.h" - -#include "pipe/p_state.h" -#include "pipe/p_context.h" - - -/** - * There is some duplication between mesa's bufferobjects and our - * bufmgr buffers. Both have an integer handle and a hashtable to - * lookup an opaque structure. It would be nice if the handles and - * internal structure where somehow shared. - */ -static struct gl_buffer_object * -intel_bufferobj_alloc(GLcontext * ctx, GLuint name, GLenum target) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_buffer_object *obj = CALLOC_STRUCT(intel_buffer_object); - - _mesa_initialize_buffer_object(&obj->Base, name, target); - - driGenBuffers(intel->intelScreen->regionPool, - "bufferobj", 1, &obj->buffer, 64, - DRM_BO_FLAG_MEM_LOCAL | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0); - - return &obj->Base; -} - - - -/** - * Deallocate/free a vertex/pixel buffer object. - * Called via glDeleteBuffersARB(). - */ -static void -intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj) -{ - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - assert(intel_obj); - - if (intel_obj->buffer) { - driDeleteBuffers(1, &intel_obj->buffer); - } - - _mesa_free(intel_obj); -} - - - -/** - * Allocate space for and store data in a buffer object. Any data that was - * previously stored in the buffer object is lost. If data is NULL, - * memory will be allocated, but no copy will occur. - * Called via glBufferDataARB(). - */ -static void -intel_bufferobj_data(GLcontext * ctx, - GLenum target, - GLsizeiptrARB size, - const GLvoid * data, - GLenum usage, struct gl_buffer_object *obj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - intel_obj->Base.Size = size; - intel_obj->Base.Usage = usage; - - LOCK_HARDWARE(intel); - driBOData(intel_obj->buffer, size, data, 0); - UNLOCK_HARDWARE(intel); -} - - -/** - * Replace data in a subrange of buffer object. If the data range - * specified by size + offset extends beyond the end of the buffer or - * if data is NULL, no copy is performed. - * Called via glBufferSubDataARB(). - */ -static void -intel_bufferobj_subdata(GLcontext * ctx, - GLenum target, - GLintptrARB offset, - GLsizeiptrARB size, - const GLvoid * data, struct gl_buffer_object *obj) -{ - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - assert(intel_obj); - - driBOSubData(intel_obj->buffer, offset, size, data); -} - - -/** - * Called via glGetBufferSubDataARB(). - */ -static void -intel_bufferobj_get_subdata(GLcontext * ctx, - GLenum target, - GLintptrARB offset, - GLsizeiptrARB size, - GLvoid * data, struct gl_buffer_object *obj) -{ - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - assert(intel_obj); - driBOGetSubData(intel_obj->buffer, offset, size, data); -} - - - -/** - * Called via glMapBufferARB(). - */ -static void * -intel_bufferobj_map(GLcontext * ctx, - GLenum target, - GLenum access, struct gl_buffer_object *obj) -{ - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - /* XXX: Translate access to flags arg below: - */ - assert(intel_obj); - - obj->Pointer = driBOMap(intel_obj->buffer, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - return obj->Pointer; -} - - -/** - * Called via glMapBufferARB(). - */ -static GLboolean -intel_bufferobj_unmap(GLcontext * ctx, - GLenum target, struct gl_buffer_object *obj) -{ - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - assert(intel_obj); - assert(obj->Pointer); - driBOUnmap(intel_obj->buffer); - obj->Pointer = NULL; - return GL_TRUE; -} - -struct _DriBufferObject * -intel_bufferobj_buffer(struct intel_context *intel, - struct intel_buffer_object *intel_obj, GLuint flag) -{ - return intel_obj->buffer; -} - -void -intel_bufferobj_init(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - - ctx->Driver.NewBufferObject = intel_bufferobj_alloc; - ctx->Driver.DeleteBuffer = intel_bufferobj_free; - ctx->Driver.BufferData = intel_bufferobj_data; - ctx->Driver.BufferSubData = intel_bufferobj_subdata; - ctx->Driver.GetBufferSubData = intel_bufferobj_get_subdata; - ctx->Driver.MapBuffer = intel_bufferobj_map; - ctx->Driver.UnmapBuffer = intel_bufferobj_unmap; -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h b/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h deleted file mode 100644 index 0f648104ee..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_buffer_objects.h +++ /dev/null @@ -1,82 +0,0 @@ - /************************************************************************** - * - * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BUFFEROBJ_H -#define INTEL_BUFFEROBJ_H - -#include "mtypes.h" - -struct intel_context; -struct pipe_region; -struct gl_buffer_object; - - -/** - * Intel vertex/pixel buffer object, derived from Mesa's gl_buffer_object. - */ -struct intel_buffer_object -{ - struct gl_buffer_object Base; - struct _DriBufferObject *buffer; /* the low-level buffer manager's buffer handle */ -}; - - -/* Get the bm buffer associated with a GL bufferobject: - */ -struct _DriBufferObject *intel_bufferobj_buffer(struct intel_context *intel, - struct intel_buffer_object - *obj, GLuint flag); - -/* Hook the bufferobject implementation into mesa: - */ -void intel_bufferobj_init(struct intel_context *intel); - - - -/* Are the obj->Name tests necessary? Unfortunately yes, mesa - * allocates a couple of gl_buffer_object structs statically, and - * the Name == 0 test is the only way to identify them and avoid - * casting them erroneously to our structs. - */ -static INLINE struct intel_buffer_object * -intel_buffer_object(struct gl_buffer_object *obj) -{ - if (obj->Name) - return (struct intel_buffer_object *) obj; - else - return NULL; -} - -/* Helpers for zerocopy image uploads. See also pipe_regions.h: - */ -void intel_bufferobj_cow(struct intel_context *intel, - struct intel_buffer_object *intel_obj); -void intel_bufferobj_release_region(struct intel_context *intel, - struct intel_buffer_object *intel_obj); - - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index a70b0ad729..4e50356eb7 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -442,7 +442,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->last_swap_fence = NULL; intel->first_swap_fence = NULL; - intel_bufferobj_init(intel); intel_fbo_init(intel); if (intel->ctx.Mesa_DXTn) { -- cgit v1.2.3 From 90ee7a41443a971d6ff2e0a6e90997806c28d11b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 7 Aug 2007 10:18:04 +0100 Subject: Gutsy oopses on touch of existing file. Workaround. --- src/egl/drivers/dri/Makefile | 1 + src/egl/main/Makefile | 1 + src/glut/beos/Makefile | 1 + src/glut/directfb/Makefile | 1 + src/glut/fbdev/Makefile | 1 + src/glut/ggi/Makefile | 1 + src/glut/glx/Makefile | 1 + src/glut/mini/Makefile | 1 + src/glw/Makefile | 1 + src/glx/mini/Makefile | 1 + src/glx/x11/Makefile | 1 + src/mesa/drivers/beos/Makefile | 1 + src/mesa/drivers/dri/Makefile.template | 1 + src/mesa/drivers/dri/glcore/Makefile | 1 + 14 files changed, 14 insertions(+) (limited to 'src') diff --git a/src/egl/drivers/dri/Makefile b/src/egl/drivers/dri/Makefile index cb11971c64..fa3720a16b 100644 --- a/src/egl/drivers/dri/Makefile +++ b/src/egl/drivers/dri/Makefile @@ -52,6 +52,7 @@ clean: depend: $(SOURCES) $(HEADERS) @ echo "running $(MKDEP)" + @ rm -f depend @ touch depend $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) \ $(SOURCES) $(HEADERS) > /dev/null diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index 431eda4197..e6844d4852 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -58,6 +58,7 @@ clean: depend: $(SOURCES) $(HEADERS) @ echo "running $(MKDEP)" + @ rm -f depend @ touch depend $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) \ $(SOURCES) $(HEADERS) > /dev/null diff --git a/src/glut/beos/Makefile b/src/glut/beos/Makefile index 3ff928bfd7..a7ba512cb1 100644 --- a/src/glut/beos/Makefile +++ b/src/glut/beos/Makefile @@ -91,6 +91,7 @@ clean: -rm -f *.o depend: $(SOURCES) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) > /dev/null diff --git a/src/glut/directfb/Makefile b/src/glut/directfb/Makefile index 3791b16df8..e881b8aacd 100644 --- a/src/glut/directfb/Makefile +++ b/src/glut/directfb/Makefile @@ -70,6 +70,7 @@ $(TOP)/$(LIB_DIR)/$(GLUT_LIB_NAME): depend $(OBJECTS) # Run 'make -f Makefile.solo dep' to update the dependencies if you change # what's included by any source file. depend: $(SOURCES) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) > /dev/null diff --git a/src/glut/fbdev/Makefile b/src/glut/fbdev/Makefile index 254ff8c098..b32b4141e8 100644 --- a/src/glut/fbdev/Makefile +++ b/src/glut/fbdev/Makefile @@ -77,6 +77,7 @@ install: # Run 'make -f Makefile.solo dep' to update the dependencies if you change # what's included by any source file. depend: $(SOURCES) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) > /dev/null diff --git a/src/glut/ggi/Makefile b/src/glut/ggi/Makefile index 1bef6c23a3..748c83c228 100644 --- a/src/glut/ggi/Makefile +++ b/src/glut/ggi/Makefile @@ -46,6 +46,7 @@ clean: depend: $(SOURCES) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) -I$(TOP)/include $(SOURCES) diff --git a/src/glut/glx/Makefile b/src/glut/glx/Makefile index 7e1d56b327..910c6fa28a 100644 --- a/src/glut/glx/Makefile +++ b/src/glut/glx/Makefile @@ -113,6 +113,7 @@ clean: depend: $(SOURCES) @ echo "running $(MKDEP)" + @ rm -f depend @ touch depend @ $(MKDEP) $(MKDEP_OPTIONS) -I$(TOP)/include $(SOURCES) \ > /dev/null diff --git a/src/glut/mini/Makefile b/src/glut/mini/Makefile index 177bfe2d2b..980306d0e2 100644 --- a/src/glut/mini/Makefile +++ b/src/glut/mini/Makefile @@ -64,6 +64,7 @@ $(TOP)/$(LIB_DIR)/$(GLUT_LIB_NAME): depend $(OBJECTS) # Run 'make -f Makefile.solo dep' to update the dependencies if you change # what's included by any source file. depend: $(SOURCES) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) > /dev/null diff --git a/src/glw/Makefile b/src/glw/Makefile index 5228cbbb72..12ab2facd4 100644 --- a/src/glw/Makefile +++ b/src/glw/Makefile @@ -49,6 +49,7 @@ $(TOP)/$(LIB_DIR)/$(GLW_LIB_NAME): $(OBJECTS) # by any source file. # depend: $(GLW_SOURCES) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) -I$(TOP)/include $(GLW_SOURCES) \ > /dev/null diff --git a/src/glx/mini/Makefile b/src/glx/mini/Makefile index 71f085d5cd..9545505cbf 100644 --- a/src/glx/mini/Makefile +++ b/src/glx/mini/Makefile @@ -65,6 +65,7 @@ drmtest: xf86drm.o drmtest.o depend: $(C_SOURCES) $(ASM_SOURCES) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(C_SOURCES) $(ASM_SOURCES) \ > /dev/null diff --git a/src/glx/x11/Makefile b/src/glx/x11/Makefile index 296747e0dd..c7b5874faa 100644 --- a/src/glx/x11/Makefile +++ b/src/glx/x11/Makefile @@ -70,6 +70,7 @@ $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(OBJECTS) Makefile depend: $(SOURCES) $(MESA_GLAPI_SOURCES) $(MESA_ASM_API) Makefile + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) \ $(MESA_GLAPI_SOURCES) $(MESA_ASM_API) diff --git a/src/mesa/drivers/beos/Makefile b/src/mesa/drivers/beos/Makefile index 1897d6aa3e..f8b7eb69a4 100644 --- a/src/mesa/drivers/beos/Makefile +++ b/src/mesa/drivers/beos/Makefile @@ -186,6 +186,7 @@ $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(OBJECTS) $(MESA_MODULES) $(GLU_MODULES) # cd $(GLU_DIR); $(MAKE) $< ; depend: $(DRIVER_SOURCES) $(GLU_SOURCES) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(DRIVER_SOURCES) $(GLU_SOURCES) > /dev/null diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template index 00b09613ec..37bf0a4977 100644 --- a/src/mesa/drivers/dri/Makefile.template +++ b/src/mesa/drivers/dri/Makefile.template @@ -83,6 +83,7 @@ $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ $(ASM_SOURCES) 2> /dev/null diff --git a/src/mesa/drivers/dri/glcore/Makefile b/src/mesa/drivers/dri/glcore/Makefile index a9e96970fa..ac7e1de928 100644 --- a/src/mesa/drivers/dri/glcore/Makefile +++ b/src/mesa/drivers/dri/glcore/Makefile @@ -66,6 +66,7 @@ $(TOP)/$(LIB_DIR)/$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(WINOBJ) Makefile depend: $(C_SOURCES) $(ASM_SOURCES) + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(C_SOURCES) $(ASM_SOURCES) \ > /dev/null -- cgit v1.2.3 From c61927a0cbbedc2fd77c151ef81a3600c2e64d53 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 10:56:01 -0600 Subject: don't include non-existant intel_buffer_objects.h --- src/mesa/drivers/dri/i915pipe/intel_context.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c index 4e50356eb7..e6f0d4a0fa 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ b/src/mesa/drivers/dri/i915pipe/intel_context.c @@ -51,7 +51,9 @@ #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" +/* #include "intel_buffer_objects.h" +*/ #include "intel_fbo.h" #include "intel_tex_layout.h" -- cgit v1.2.3 From cec49ee954287b7f34664ea9e8137c674c26dcf1 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 10:56:41 -0600 Subject: s/pipe_texture_object/pipe_mipmap_tree/, drop st_texobj.c --- src/mesa/pipe/p_context.h | 2 +- src/mesa/pipe/p_state.h | 2 ++ src/mesa/pipe/softpipe/sp_context.h | 2 +- src/mesa/pipe/softpipe/sp_state.h | 2 +- src/mesa/pipe/softpipe/sp_state_sampler.c | 2 +- src/mesa/sources | 3 +-- 6 files changed, 7 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index db9429f8d3..08b9b18393 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -117,7 +117,7 @@ struct pipe_context { void (*set_texture_state)( struct pipe_context *, GLuint unit, - struct pipe_texture_object * ); + struct pipe_mipmap_tree * ); void (*set_viewport_state)( struct pipe_context *, const struct pipe_viewport_state * ); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index be4b0ebdc7..f75d8f1957 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -264,6 +264,7 @@ struct pipe_surface }; +#if 0 /** * Texture object. * Mipmap levels, cube faces, 3D slices can be accessed as surfaces. @@ -290,6 +291,7 @@ struct pipe_texture_object void (*release_surface)(struct pipe_texture_object *pto, struct pipe_surface *ps); }; +#endif /** diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 43fa8acd76..a99c083caa 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -86,7 +86,7 @@ struct softpipe_context { struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_setup_state setup; struct pipe_stencil_state stencil; - struct pipe_texture_object *texture[PIPE_MAX_SAMPLERS]; + struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; GLuint dirty; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 71c1a2d9ba..3c572cdb6a 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -77,7 +77,7 @@ void softpipe_set_stencil_state( struct pipe_context *, void softpipe_set_texture_state( struct pipe_context *, GLuint unit, - struct pipe_texture_object * ); + struct pipe_mipmap_tree * ); void softpipe_set_viewport_state( struct pipe_context *, const struct pipe_viewport_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 9ef71f73cb..34cf210834 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -53,7 +53,7 @@ softpipe_set_sampler_state(struct pipe_context *pipe, void softpipe_set_texture_state(struct pipe_context *pipe, GLuint unit, - struct pipe_texture_object *texture) + struct pipe_mipmap_tree *texture) { struct softpipe_context *softpipe = softpipe_context(pipe); diff --git a/src/mesa/sources b/src/mesa/sources index 0a69da9efc..b648b6b8de 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -200,8 +200,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_draw.c \ state_tracker/st_context.c \ - state_tracker/st_mipmap_tree.c \ - state_tracker/st_texobj.c + state_tracker/st_mipmap_tree.c SHADER_SOURCES = \ shader/arbprogparse.c \ -- cgit v1.2.3 From c7e6a2eb83bbf86aa6d558aeb3f58cbc76a3983f Mon Sep 17 00:00:00 2001 From: michal Date: Tue, 7 Aug 2007 10:24:27 +0100 Subject: Dump tokens in more compact form. --- src/mesa/pipe/tgsi/core/tgsi_dump.c | 38 +++++++++++++++++++++++++++---------- src/mesa/pipe/tgsi/core/tgsi_dump.h | 6 +++--- 2 files changed, 31 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index 882771da2e..3e2eb4999f 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -137,7 +137,15 @@ text_dump_untab( static const char *TGSI_PROCESSOR_TYPES[] = { "PROCESSOR_FRAGMENT", - "PROCESSOR_VERTEX" + "PROCESSOR_VERTEX", + "PROCESSOR_GEOMETRY" +}; + +static const char *TGSI_PROCESSOR_TYPES_SHORT[] = +{ + "FRAG", + "VERT", + "GEOM" }; static const char *TGSI_TOKEN_TYPES[] = @@ -406,6 +414,7 @@ tgsi_dump( struct tgsi_parse_context parse; struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; + GLuint verbose = flags & TGSI_DUMP_VERBOSE; GLuint ignored = !(flags & TGSI_DUMP_NO_IGNORED); GLuint deflt = !(flags & TGSI_DUMP_NO_DEFAULT); @@ -430,18 +439,27 @@ tgsi_dump( TXT( "tgsi-dump begin" ); CHR( '\n' ); - TXT( "\nMajorVersion: " ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); + CHR( ' ' ); UID( parse.FullVersion.Version.MajorVersion ); - TXT( "\nMinorVersion: " ); + CHR( '.' ); UID( parse.FullVersion.Version.MinorVersion ); - CHR( '\n' ); - TXT( "\nHeaderSize: " ); - UID( parse.FullHeader.Header.HeaderSize ); - TXT( "\nBodySize : " ); - UID( parse.FullHeader.Header.BodySize ); - TXT( "\nProcessor : " ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + if( verbose ) { + CHR( '\n' ); + TXT( "\nMajorVersion: " ); + UID( parse.FullVersion.Version.MajorVersion ); + TXT( "\nMinorVersion: " ); + UID( parse.FullVersion.Version.MinorVersion ); + + CHR( '\n' ); + TXT( "\nHeaderSize: " ); + UID( parse.FullHeader.Header.HeaderSize ); + TXT( "\nBodySize : " ); + UID( parse.FullHeader.Header.BodySize ); + TXT( "\nProcessor : " ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + } fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.h b/src/mesa/pipe/tgsi/core/tgsi_dump.h index dc34a840de..8553bdff59 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.h +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.h @@ -5,9 +5,9 @@ extern "C" { #endif // defined __cplusplus -#define TGSI_DUMP_VERBOSE 0 -#define TGSI_DUMP_NO_IGNORED 1 -#define TGSI_DUMP_NO_DEFAULT 2 +#define TGSI_DUMP_VERBOSE 1 +#define TGSI_DUMP_NO_IGNORED 2 +#define TGSI_DUMP_NO_DEFAULT 4 void tgsi_dump( -- cgit v1.2.3 From f141399031981d6af19a32025ae7c1b947196480 Mon Sep 17 00:00:00 2001 From: michal Date: Tue, 7 Aug 2007 11:49:33 +0100 Subject: Enable compact TGSI dump. --- src/mesa/pipe/tgsi/core/tgsi_dump.c | 1222 +++++++++++++++++++++----------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 +- 2 files changed, 794 insertions(+), 430 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index 3e2eb4999f..68b81b7ffe 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -124,15 +124,15 @@ text_dump_untab( --dump->tabs; } -#define TXT(S) text_dump_str( &dump, S ) -#define CHR(C) text_dump_chr( &dump, C ) -#define UIX(I) text_dump_uix( &dump, I ) -#define UID(I) text_dump_uid( &dump, I ) -#define SID(I) text_dump_sid( &dump, I ) -#define FLT(F) text_dump_flt( &dump, F ) -#define TAB() text_dump_tab( &dump ) -#define UNT() text_dump_untab( &dump ) -#define ENM(E,ENUMS) text_dump_enum( &dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) +#define TXT(S) text_dump_str( dump, S ) +#define CHR(C) text_dump_chr( dump, C ) +#define UIX(I) text_dump_uix( dump, I ) +#define UID(I) text_dump_uid( dump, I ) +#define SID(I) text_dump_sid( dump, I ) +#define FLT(F) text_dump_flt( dump, F ) +#define TAB() text_dump_tab( dump ) +#define UNT() text_dump_untab( dump ) +#define ENM(E,ENUMS) text_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) static const char *TGSI_PROCESSOR_TYPES[] = { @@ -167,6 +167,18 @@ static const char *TGSI_FILES[] = "FILE_IMMEDIATE" }; +static const char *TGSI_FILES_SHORT[] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + static const char *TGSI_DECLARES[] = { "DECLARE_RANGE", @@ -180,11 +192,23 @@ static const char *TGSI_INTERPOLATES[] = "INTERPOLATE_PERSPECTIVE" }; +static const char *TGSI_INTERPOLATES_SHORT[] = +{ + "CONSTANT", + "LINEAR", + "PERSPECTIVE" +}; + static const char *TGSI_IMMS[] = { "IMM_FLOAT32" }; +static const char *TGSI_IMMS_SHORT[] = +{ + "FLT32" +}; + static const char *TGSI_OPCODES[] = { "OPCODE_ARL", @@ -296,6 +320,117 @@ static const char *TGSI_OPCODES[] = "OPCODE_NOP" }; +static const char *TGSI_OPCODES_SHORT[] = +{ + "ARL", + "MOV", + "LIT", + "RCP", + "RSQ", + "EXP", + "LOG", + "MUL", + "ADD", + "DP3", + "DP4", + "DST", + "MIN", + "MAX", + "SLT", + "SGE", + "MAD", + "SUB", + "LERP", + "CND", + "CND0", + "DOT2ADD", + "INDEX", + "NEGATE", + "FRAC", + "CLAMP", + "FLOOR", + "ROUND", + "EXPBASE2", + "LOGBASE2", + "POWER", + "CROSSPRODUCT", + "MULTIPLYMATRIX", + "ABS", + "RCC", + "DPH", + "COS", + "DDX", + "DDY", + "KIL", + "PK2H", + "PK2US", + "PK4B", + "PK4UB", + "RFL", + "SEQ", + "SFL", + "SGT", + "SIN", + "SLE", + "SNE", + "STR", + "TEX", + "TXD", + "UP2H", + "UP2US", + "UP4B", + "UP4UB", + "X2D", + "ARA", + "ARR", + "BRA", + "CAL", + "RET", + "SSG", + "CMP", + "SCS", + "TXB", + "NRM", + "DIV", + "DP2", + "TXL", + "BRK", + "IF", + "LOOP", + "REP", + "ELSE", + "ENDIF", + "ENDLOOP", + "ENDREP", + "PUSHA", + "POPA", + "CEIL", + "I2F", + "NOT", + "TRUNC", + "SHL", + "SHR", + "AND", + "OR", + "MOD", + "XOR", + "SAD", + "TXF", + "TXQ", + "CONT", + "EMIT", + "ENDPRIM", + "BGNLOOP2", + "BGNSUB", + "ENDLOOP2", + "ENDSUB", + "NOISE1", + "NOISE2", + "NOISE3", + "NOISE4", + "NOP" +}; + static const char *TGSI_SATS[] = { "SAT_NONE", @@ -339,6 +474,14 @@ static const char *TGSI_SWIZZLES[] = "SWIZZLE_W" }; +static const char *TGSI_SWIZZLES_SHORT[] = +{ + "x", + "y", + "z", + "w" +}; + static const char *TGSI_TEXTURES[] = { "TEXTURE_UNKNOWN", @@ -405,12 +548,592 @@ static const char *TGSI_MODULATES[] = "MODULATE_EIGHTH" }; +static void +dump_declaration_short( + struct text_dump *dump, + struct tgsi_full_declaration *decl ) +{ + TXT( "\nDCL " ); + ENM( decl->Declaration.File, TGSI_FILES_SHORT ); + + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + CHR( '[' ); + UID( decl->u.DeclarationRange.First ); + if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { + TXT( ".." ); + UID( decl->u.DeclarationRange.Last ); + } + CHR( ']' ); + break; + default: + assert( 0 ); + } + + if( decl->Declaration.Interpolate ) { + TXT( ", " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); + } +} + +static void +dump_declaration_verbose( + struct text_dump *dump, + struct tgsi_full_declaration *decl, + GLuint ignored, + GLuint deflt, + struct tgsi_full_declaration *fd ) +{ + TXT( "\nFile : " ); + ENM( decl->Declaration.File, TGSI_FILES ); + TXT( "\nDeclare : " ); + ENM( decl->Declaration.Declare, TGSI_DECLARES ); + if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { + TXT( "\nInterpolate: " ); + UID( decl->Declaration.Interpolate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Declaration.Padding ); + } + + CHR( '\n' ); + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + TXT( "\nFirst: " ); + UID( decl->u.DeclarationRange.First ); + TXT( "\nLast : " ); + UID( decl->u.DeclarationRange.Last ); + break; + + case TGSI_DECLARE_MASK: + TXT( "\nMask: " ); + UIX( decl->u.DeclarationMask.Mask ); + break; + + default: + assert( 0 ); + } + + if( decl->Declaration.Interpolate ) { + CHR( '\n' ); + TXT( "\nInterpolate: " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Interpolation.Padding ); + } + } +} + +static void +dump_immediate_short( + struct text_dump *dump, + struct tgsi_full_immediate *imm ) +{ + GLuint i; + + TXT( "\nIMM " ); + ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); + + TXT( " { " ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + + if( i < imm->Immediate.Size - 2 ) { + TXT( ", " ); + } + } + TXT( " }" ); +} + +static void +dump_immediate_verbose( + struct text_dump *dump, + struct tgsi_full_immediate *imm, + GLuint ignored ) +{ + GLuint i; + + TXT( "\nDataType : " ); + ENM( imm->Immediate.DataType, TGSI_IMMS ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( imm->Immediate.Padding ); + } + + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + CHR( '\n' ); + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + TXT( "\nFloat: " ); + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + } +} + +static void +dump_instruction_short( + struct text_dump *dump, + struct tgsi_full_instruction *inst ) +{ + GLuint i; + GLboolean first_reg = GL_TRUE; + + CHR( '\n' ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); + + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + TXT( "_SAT" ); + break; + case TGSI_SAT_MINUS_PLUS_ONE: + TXT( "_SAT[-1,1]" ); + break; + default: + assert( 0 ); + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( dst->DstRegister.Index ); + CHR( ']' ); + + if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + + first_reg = GL_FALSE; + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + if( src->SrcRegisterExtMod.Negate ) { + CHR( '-' ); + } + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + if( src->SrcRegister.Negate ) { + CHR( '-' ); + } + + ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( src->SrcRegister.Index ); + CHR( ']' ); + + if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { + CHR( '.' ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); + } + + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + + first_reg = GL_FALSE; + } +} + +static void +dump_instruction_verbose( + struct text_dump *dump, + struct tgsi_full_instruction *inst, + GLuint ignored, + GLuint deflt, + struct tgsi_full_instruction *fi ) +{ + GLuint i; + + TXT( "\nOpcode : " ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES ); + if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { + TXT( "\nSaturate : " ); + ENM( inst->Instruction.Saturate, TGSI_SATS ); + } + if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { + TXT( "\nNumDstRegs : " ); + UID( inst->Instruction.NumDstRegs ); + } + if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { + TXT( "\nNumSrcRegs : " ); + UID( inst->Instruction.NumSrcRegs ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->Instruction.Padding ); + } + + if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { + TXT( "\nPrecision : " ); + ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); + } + if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { + TXT( "\nCondDstIndex : " ); + UID( inst->InstructionExtNv.CondDstIndex ); + } + if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { + TXT( "\nCondFlowIndex : " ); + UID( inst->InstructionExtNv.CondFlowIndex ); + } + if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { + TXT( "\nCondMask : " ); + ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { + TXT( "\nCondSwizzleX : " ); + ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { + TXT( "\nCondSwizzleY : " ); + ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ : " ); + ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { + TXT( "\nCondSwizzleW : " ); + ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { + TXT( "\nCondDstUpdate : " ); + UID( inst->InstructionExtNv.CondDstUpdate ); + } + if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { + TXT( "\nCondFlowEnable: " ); + UID( inst->InstructionExtNv.CondFlowEnable ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtNv.Padding ); + if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { + TXT( "\nExtended : " ); + UID( inst->InstructionExtNv.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { + TXT( "\nLabel : " ); + UID( inst->InstructionExtLabel.Label ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtLabel.Padding ); + if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtLabel.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { + TXT( "\nTexture : " ); + ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtTexture.Padding ); + if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtTexture.Extended ); + } + } + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : " ); + ENM( dst->DstRegister.File, TGSI_FILES ); + if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { + TXT( "\nWriteMask: " ); + ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); + } + if( ignored ) { + if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( dst->DstRegister.Indirect ); + } + if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( dst->DstRegister.Dimension ); + } + } + if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { + TXT( "\nIndex : " ); + SID( dst->DstRegister.Index ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegister.Padding ); + if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegister.Extended ); + } + } + + if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { + TXT( "\nCondMask : " ); + ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { + TXT( "\nCondSwizzleX: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { + TXT( "\nCondSwizzleY: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { + TXT( "\nCondSwizzleW: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { + TXT( "\nCondSrcIndex: " ); + UID( dst->DstRegisterExtConcode.CondSrcIndex ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtConcode.Padding ); + if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegisterExtConcode.Extended ); + } + } + } + + if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { + TXT( "\nModulate: " ); + ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtModulate.Padding ); + if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { + TXT( "\nExtended: " ); + UID( dst->DstRegisterExtModulate.Extended ); + } + } + } + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : "); + ENM( src->SrcRegister.File, TGSI_FILES ); + if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { + TXT( "\nSwizzleX : " ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { + TXT( "\nSwizzleY : " ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { + TXT( "\nSwizzleZ : " ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { + TXT( "\nSwizzleW : " ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegister.Negate ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( src->SrcRegister.Indirect ); + } + if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( src->SrcRegister.Dimension ); + } + } + if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { + TXT( "\nIndex : " ); + SID( src->SrcRegister.Index ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegister.Extended ); + } + } + + if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { + TXT( "\nExtSwizzleX: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { + TXT( "\nExtSwizzleY: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { + TXT( "\nExtSwizzleZ: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { + TXT( "\nExtSwizzleW: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { + TXT( "\nNegateX : " ); + UID( src->SrcRegisterExtSwz.NegateX ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { + TXT( "\nNegateY : " ); + UID( src->SrcRegisterExtSwz.NegateY ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { + TXT( "\nNegateZ : " ); + UID( src->SrcRegisterExtSwz.NegateZ ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { + TXT( "\nNegateW : " ); + UID( src->SrcRegisterExtSwz.NegateW ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { + TXT( "\nExtDivide : " ); + ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtSwz.Padding ); + if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtSwz.Extended ); + } + } + } + + if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { + TXT( "\nComplement: " ); + UID( src->SrcRegisterExtMod.Complement ); + } + if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { + TXT( "\nBias : " ); + UID( src->SrcRegisterExtMod.Bias ); + } + if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { + TXT( "\nScale2X : " ); + UID( src->SrcRegisterExtMod.Scale2X ); + } + if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { + TXT( "\nAbsolute : " ); + UID( src->SrcRegisterExtMod.Absolute ); + } + if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegisterExtMod.Negate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtMod.Padding ); + if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtMod.Extended ); + } + } + } + } +} + void tgsi_dump( const struct tgsi_token *tokens, GLuint flags ) { - struct text_dump dump; + struct text_dump _dump; + struct text_dump *dump = &_dump; struct tgsi_parse_context parse; struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; @@ -424,10 +1147,10 @@ tgsi_dump( char buffer[64]; sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); - dump.file = fopen( buffer, "wt" ); + dump->file = fopen( buffer, "wt" ); #else - dump.file = stderr; - dump.tabs = 0; + dump->file = stderr; + dump->tabs = 0; #endif } @@ -436,7 +1159,7 @@ tgsi_dump( tgsi_parse_init( &parse, tokens ); - TXT( "tgsi-dump begin" ); + TXT( "tgsi-dump begin -----------------" ); CHR( '\n' ); ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); @@ -446,459 +1169,100 @@ tgsi_dump( UID( parse.FullVersion.Version.MinorVersion ); if( verbose ) { - CHR( '\n' ); TXT( "\nMajorVersion: " ); UID( parse.FullVersion.Version.MajorVersion ); TXT( "\nMinorVersion: " ); UID( parse.FullVersion.Version.MinorVersion ); - CHR( '\n' ); + TXT( "\nHeaderSize: " ); UID( parse.FullHeader.Header.HeaderSize ); TXT( "\nBodySize : " ); UID( parse.FullHeader.Header.BodySize ); TXT( "\nProcessor : " ); ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + CHR( '\n' ); } fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); while( !tgsi_parse_end_of_tokens( &parse ) ) { - GLuint i; - tgsi_parse_token( &parse ); - CHR( '\n' ); - TXT( "\nType : " ); - ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); - if( ignored ) { - TXT( "\nSize : " ); - UID( parse.FullToken.Token.Size ); - if( deflt || parse.FullToken.Token.Extended ) { - TXT( "\nExtended : " ); - UID( parse.FullToken.Token.Extended ); - } - } - switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; - - TXT( "\nFile : " ); - ENM( decl->Declaration.File, TGSI_FILES ); - TXT( "\nDeclare : " ); - ENM( decl->Declaration.Declare, TGSI_DECLARES ); - if( deflt || fd.Declaration.Interpolate != decl->Declaration.Interpolate ) { - TXT( "\nInterpolate: " ); - UID( decl->Declaration.Interpolate ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Declaration.Padding ); - } - - CHR( '\n' ); - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - TXT( "\nFirst: " ); - UID( decl->u.DeclarationRange.First ); - TXT( "\nLast : " ); - UID( decl->u.DeclarationRange.Last ); - break; - - case TGSI_DECLARE_MASK: - TXT( "\nMask: " ); - UIX( decl->u.DeclarationMask.Mask ); - break; - - default: - assert( 0 ); - } - - if( decl->Declaration.Interpolate ) { - CHR( '\n' ); - TXT( "\nInterpolate: " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Interpolation.Padding ); - } - } - } + dump_declaration_short( + dump, + &parse.FullToken.FullDeclaration ); break; case TGSI_TOKEN_TYPE_IMMEDIATE: - TXT( "\nDataType : " ); - ENM( parse.FullToken.FullImmediate.Immediate.DataType, TGSI_IMMS ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( parse.FullToken.FullImmediate.Immediate.Padding ); - } - - for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size - 1; i++ ) { - CHR( '\n' ); - switch( parse.FullToken.FullImmediate.Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - TXT( "\nFloat: " ); - FLT( parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - } + dump_immediate_short( + dump, + &parse.FullToken.FullImmediate ); break; case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction; - - TXT( "\nOpcode : " ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES ); - if( deflt || fi.Instruction.Saturate != inst->Instruction.Saturate ) { - TXT( "\nSaturate : " ); - ENM( inst->Instruction.Saturate, TGSI_SATS ); - } - if( deflt || fi.Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { - TXT( "\nNumDstRegs : " ); - UID( inst->Instruction.NumDstRegs ); - } - if( deflt || fi.Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { - TXT( "\nNumSrcRegs : " ); - UID( inst->Instruction.NumSrcRegs ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->Instruction.Padding ); - } - - if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi.InstructionExtNv)) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi.InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { - TXT( "\nPrecision : " ); - ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); - } - if( deflt || fi.InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { - TXT( "\nCondDstIndex : " ); - UID( inst->InstructionExtNv.CondDstIndex ); - } - if( deflt || fi.InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { - TXT( "\nCondFlowIndex : " ); - UID( inst->InstructionExtNv.CondFlowIndex ); - } - if( deflt || fi.InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { - TXT( "\nCondMask : " ); - ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); - } - if( deflt || fi.InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { - TXT( "\nCondSwizzleX : " ); - ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fi.InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { - TXT( "\nCondSwizzleY : " ); - ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fi.InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ : " ); - ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fi.InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { - TXT( "\nCondSwizzleW : " ); - ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fi.InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { - TXT( "\nCondDstUpdate : " ); - UID( inst->InstructionExtNv.CondDstUpdate ); - } - if( deflt || fi.InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { - TXT( "\nCondFlowEnable: " ); - UID( inst->InstructionExtNv.CondFlowEnable ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtNv.Padding ); - if( deflt || fi.InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { - TXT( "\nExtended : " ); - UID( inst->InstructionExtNv.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi.InstructionExtLabel ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi.InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { - TXT( "\nLabel : " ); - UID( inst->InstructionExtLabel.Label ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtLabel.Padding ); - if( deflt || fi.InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtLabel.Extended ); - } - } - } + dump_instruction_short( + dump, + &parse.FullToken.FullInstruction ); + break; - if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi.InstructionExtTexture ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi.InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { - TXT( "\nTexture : " ); - ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtTexture.Padding ); - if( deflt || fi.InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtTexture.Extended ); - } - } - } + default: + assert( 0 ); + } - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - struct tgsi_full_dst_register *fd = &fi.FullDstRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : " ); - ENM( dst->DstRegister.File, TGSI_FILES ); - if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { - TXT( "\nWriteMask: " ); - ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); - } - if( ignored ) { - if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( dst->DstRegister.Indirect ); - } - if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( dst->DstRegister.Dimension ); - } - } - if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { - TXT( "\nIndex : " ); - SID( dst->DstRegister.Index ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegister.Padding ); - if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegister.Extended ); - } - } - - if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { - TXT( "\nCondMask : " ); - ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { - TXT( "\nCondSwizzleX: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { - TXT( "\nCondSwizzleY: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { - TXT( "\nCondSwizzleW: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { - TXT( "\nCondSrcIndex: " ); - UID( dst->DstRegisterExtConcode.CondSrcIndex ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtConcode.Padding ); - if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegisterExtConcode.Extended ); - } - } - } - - if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { - TXT( "\nModulate: " ); - ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtModulate.Padding ); - if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { - TXT( "\nExtended: " ); - UID( dst->DstRegisterExtModulate.Extended ); - } - } - } + if( verbose ) { + TXT( "\nType : " ); + ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); + if( ignored ) { + TXT( "\nSize : " ); + UID( parse.FullToken.Token.Size ); + if( deflt || parse.FullToken.Token.Extended ) { + TXT( "\nExtended : " ); + UID( parse.FullToken.Token.Extended ); } + } - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - struct tgsi_full_src_register *fs = &fi.FullSrcRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : "); - ENM( src->SrcRegister.File, TGSI_FILES ); - if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { - TXT( "\nSwizzleX : " ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { - TXT( "\nSwizzleY : " ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { - TXT( "\nSwizzleZ : " ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { - TXT( "\nSwizzleW : " ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegister.Negate ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( src->SrcRegister.Indirect ); - } - if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( src->SrcRegister.Dimension ); - } - } - if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { - TXT( "\nIndex : " ); - SID( src->SrcRegister.Index ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegister.Extended ); - } - } - - if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { - TXT( "\nExtSwizzleX: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { - TXT( "\nExtSwizzleY: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { - TXT( "\nExtSwizzleZ: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { - TXT( "\nExtSwizzleW: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { - TXT( "\nNegateX : " ); - UID( src->SrcRegisterExtSwz.NegateX ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { - TXT( "\nNegateY : " ); - UID( src->SrcRegisterExtSwz.NegateY ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { - TXT( "\nNegateZ : " ); - UID( src->SrcRegisterExtSwz.NegateZ ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { - TXT( "\nNegateW : " ); - UID( src->SrcRegisterExtSwz.NegateW ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { - TXT( "\nExtDivide : " ); - ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtSwz.Padding ); - if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtSwz.Extended ); - } - } - } - - if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { - TXT( "\nComplement: " ); - UID( src->SrcRegisterExtMod.Complement ); - } - if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { - TXT( "\nBias : " ); - UID( src->SrcRegisterExtMod.Bias ); - } - if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { - TXT( "\nScale2X : " ); - UID( src->SrcRegisterExtMod.Scale2X ); - } - if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { - TXT( "\nAbsolute : " ); - UID( src->SrcRegisterExtMod.Absolute ); - } - if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegisterExtMod.Negate ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtMod.Padding ); - if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtMod.Extended ); - } - } - } - } + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_verbose( + dump, + &parse.FullToken.FullDeclaration, + ignored, + deflt, + &fd ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_verbose( + dump, + &parse.FullToken.FullImmediate, + ignored ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_verbose( + dump, + &parse.FullToken.FullInstruction, + ignored, + deflt, + &fi ); + break; + + default: + assert( 0 ); } - break; - default: - assert( 0 ); + CHR( '\n' ); } } - TXT( "\ntgsi-dump end\n" ); + TXT( "\ntgsi-dump end -------------------\n" ); tgsi_parse_free( &parse ); - fclose( dump.file ); + fclose( dump->file ); } diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 47fd5791e1..15f4d1e197 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -569,7 +569,7 @@ tgsi_mesa_compile_fp_program( assert( i == program->Base.NumInstructions - 1 ); tgsi_dump( tokens, - TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT ); + 0/*TGSI_DUMP_VERBOSE | TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT*/ ); break; } -- cgit v1.2.3 From 281dad225947b4f00bfc5e785d92d2a7c2f06afe Mon Sep 17 00:00:00 2001 From: michal Date: Tue, 7 Aug 2007 12:50:12 +0100 Subject: Handle labels for IF/ELSE correctly --- src/mesa/pipe/tgsi/core/tgsi_dump.c | 18 ++++++++++++++++-- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 9 +++++++-- 2 files changed, 23 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index 68b81b7ffe..641339c896 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -686,12 +686,15 @@ dump_immediate_verbose( static void dump_instruction_short( struct text_dump *dump, - struct tgsi_full_instruction *inst ) + struct tgsi_full_instruction *inst, + GLuint instno ) { GLuint i; GLboolean first_reg = GL_TRUE; CHR( '\n' ); + UID( instno ); + CHR( ':' ); ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); switch( inst->Instruction.Saturate ) { @@ -781,6 +784,14 @@ dump_instruction_short( first_reg = GL_FALSE; } + + switch( inst->Instruction.Opcode ) { + case TGSI_OPCODE_IF: + case TGSI_OPCODE_ELSE: + TXT( " : " ); + UID( inst->InstructionExtLabel.Label ); + break; + } } static void @@ -1140,6 +1151,7 @@ tgsi_dump( GLuint verbose = flags & TGSI_DUMP_VERBOSE; GLuint ignored = !(flags & TGSI_DUMP_NO_IGNORED); GLuint deflt = !(flags & TGSI_DUMP_NO_DEFAULT); + GLuint instno = 0; { #if 0 @@ -1206,7 +1218,9 @@ tgsi_dump( case TGSI_TOKEN_TYPE_INSTRUCTION: dump_instruction_short( dump, - &parse.FullToken.FullInstruction ); + &parse.FullToken.FullInstruction, + instno ); + instno++; break; default: diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 15f4d1e197..654ec9ac68 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -158,6 +158,7 @@ compile_instruction( const struct prog_instruction *inst, struct tgsi_full_instruction *fullinst, GLuint inputs_read, + GLuint preamble_size, GLuint processor ) { GLuint i; @@ -293,7 +294,7 @@ compile_instruction( break; case OPCODE_ELSE: fullinst->Instruction.Opcode = TGSI_OPCODE_ELSE; - fullinst->InstructionExtLabel.Label = inst->BranchTarget; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; break; case OPCODE_ENDIF: fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF; @@ -318,7 +319,7 @@ compile_instruction( break; case OPCODE_IF: fullinst->Instruction.Opcode = TGSI_OPCODE_IF; - fullinst->InstructionExtLabel.Label = inst->BranchTarget; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; break; case OPCODE_INT: fullinst->Instruction.Opcode = TGSI_OPCODE_INT; @@ -473,6 +474,7 @@ tgsi_mesa_compile_fp_program( struct tgsi_full_dst_register *fulldst; struct tgsi_full_src_register *fullsrc; GLuint inputs_read; + GLuint preamble_size = 0; *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -559,12 +561,14 @@ tgsi_mesa_compile_fp_program( &tokens[ti], header, maxTokens - ti ); + preamble_size++; for( i = 0; i < program->Base.NumInstructions; i++ ) { if( compile_instruction( &program->Base.Instructions[i], &fullinst, inputs_read, + preamble_size, TGSI_PROCESSOR_FRAGMENT ) ) { assert( i == program->Base.NumInstructions - 1 ); tgsi_dump( @@ -610,6 +614,7 @@ tgsi_mesa_compile_vp_program( &program->Base.Instructions[ii], &fullinst, inputs_read, + 0, TGSI_PROCESSOR_VERTEX ) ) { assert( ii == program->Base.NumInstructions - 1 ); tgsi_dump( tokens, TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT ); -- cgit v1.2.3 From b23f358cbc36a2b6a9e7609290a7458fa48f7ccb Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 13:13:41 -0600 Subject: sketch out new pipe surface/sampler types --- src/mesa/drivers/dri/i915pipe/intel_fbo.c | 2 ++ src/mesa/drivers/dri/intel/intel_tex_layout.c | 2 ++ src/mesa/pipe/p_context.h | 15 +++++++++ src/mesa/pipe/p_state.h | 45 +++++++++++---------------- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_surface.c | 42 +++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_surface.h | 6 ++++ 7 files changed, 86 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c index a53563f52b..0b6c6a94ee 100644 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ b/src/mesa/drivers/dri/i915pipe/intel_fbo.c @@ -578,7 +578,9 @@ intel_render_texture(GLcontext * ctx, } /* store that offset in the region */ +#if 0 st_image->mt->region->draw_offset = imageOffset; +#endif /* update drawing region, etc */ intel_draw_buffer(ctx, fb); diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c index 344aff396f..882d9e04a2 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.c +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c @@ -32,6 +32,7 @@ #include "macros.h" #include "pipe/p_state.h" +#include "pipe/p_context.h" #include "intel_tex_layout.h" #include "state_tracker/st_mipmap_tree.h" @@ -472,3 +473,4 @@ i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) return GL_TRUE; } + diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 08b9b18393..3062eda329 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -130,6 +130,10 @@ struct pipe_context { struct pipe_surface *(*surface_alloc)(struct pipe_context *pipe, GLuint format); + struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, + struct pipe_mipmap_tree *texture, + GLuint face, GLuint level, + GLuint zslice); /* * Memory region functions @@ -230,5 +234,16 @@ pipe_region_reference(struct pipe_region **dst, struct pipe_region *src) } +static INLINE void +pipe_surface_reference(struct pipe_surface **dst, struct pipe_surface *src) +{ + assert(*dst == NULL); + if (src) { + src->refcount++; + *dst = src; + } +} + + #endif /* PIPE_CONTEXT_H */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index f75d8f1957..e51b9c460f 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -245,13 +245,11 @@ struct pipe_region GLuint height; /**< in pixels */ GLubyte *map; /**< only non-NULL when region is actually mapped */ GLuint map_refcount; /**< Reference count for mapping */ - - GLuint draw_offset; /**< Offset of drawing address within the region */ }; /** - * 2D surface. + * 2D surface. This is basically a view into a pipe_region (memory buffer). * May be a renderbuffer, texture mipmap level, etc. */ struct pipe_surface @@ -259,39 +257,32 @@ struct pipe_surface struct pipe_region *region; GLuint format:5; /**< PIPE_FORMAT_x */ GLuint width, height; + GLuint offset; /**< offset from start of region, in bytes */ + GLint refcount; void *rb; /**< Ptr back to renderbuffer (temporary?) */ + + /** get block/tile of pixels from surface */ + void (*get_tile)(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p); + + /** put block/tile of pixels into surface */ + void (*put_tile)(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, const GLfloat *p); }; -#if 0 /** - * Texture object. - * Mipmap levels, cube faces, 3D slices can be accessed as surfaces. + * surface sampler object + * XXX prototype */ -struct pipe_texture_object +struct pipe_surface_sampler { - GLuint type:2; /**< PIPE_TEXTURE_x */ - GLuint format:5; /**< PIPE_FORMAT_x */ - GLuint width:13; /**< 13 bits = 8K max size */ - GLuint height:13; - GLuint depth:13; - GLuint mipmapped:1; - - /** to access a 1D or 2D texture object as a surface */ - struct pipe_surface *(*get_2d_surface)(struct pipe_texture_object *pto, - GLuint level); - /** to access a 3D texture object as a surface */ - struct pipe_surface *(*get_3d_surface)(struct pipe_texture_object *pto, - GLuint level, GLuint slice); - /** to access a cube texture object as a surface */ - struct pipe_surface *(*get_cube_surface)(struct pipe_texture_object *pto, - GLuint face, GLuint level); - /** when finished with surface: */ - void (*release_surface)(struct pipe_texture_object *pto, - struct pipe_surface *ps); + struct pipe_sampler_state state; + struct pipe_mipmap_tree *texture; + void (*get_sample)(struct pipe_surface_sampler *sampler, + const GLfloat strq[4], GLfloat rgba[4]); }; -#endif /** diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 139afde376..264f3d6e58 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -201,6 +201,7 @@ struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; softpipe->pipe.mipmap_tree_layout = softpipe_mipmap_tree_layout; + softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); softpipe->quad.shade = sp_quad_shade_stage(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 819243ae6e..6512f1d789 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -351,12 +351,54 @@ sp_surface_alloc(struct pipe_context *pipe, GLenum format) return NULL; sps->surface.format = format; + sps->surface.refcount = 1; init_quad_funcs(sps); return &sps->surface; } + + + +/** + * Called via pipe->get_tex_surface() + * XXX is this in the right place? + */ +struct pipe_surface * +softpipe_get_tex_surface(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt, + GLuint face, GLuint level, GLuint zslice) +{ + struct pipe_surface *ps; + GLuint offset; /* in bytes */ + + offset = mt->level[level].level_offset; + + if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) { + offset += mt->level[level].image_offset[face] * mt->cpp; + } + else if (mt->target == GL_TEXTURE_3D) { + offset += mt->level[level].image_offset[zslice] * mt->cpp; + } + else { + assert(face == 0); + assert(zslice == 0); + } + + ps = pipe->surface_alloc(pipe, mt->internal_format); + if (ps) { + assert(ps->format); + assert(ps->refcount); + ps->region = mt->region; + ps->width = mt->level[level].width; + ps->height = mt->level[level].height; + ps->offset = offset; + } + return ps; +} + + void sp_init_surface_functions(struct softpipe_context *sp) { diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 6782bba376..90086c327f 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -85,6 +85,12 @@ struct softpipe_surface { }; +extern struct pipe_surface * +softpipe_get_tex_surface(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt, + GLuint face, GLuint level, GLuint zslice); + + /** Cast wrapper */ static INLINE struct softpipe_surface * softpipe_surface(struct pipe_surface *ps) -- cgit v1.2.3 From 28d64d398bb3e04d20a9c03c9a0b92c0a7b9d6d4 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 15:11:20 -0600 Subject: texture image state handling --- src/mesa/state_tracker/st_atom_texture.c | 76 ++++++++++++++++++++++++++++++++ 1 file changed, 76 insertions(+) create mode 100644 src/mesa/state_tracker/st_atom_texture.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c new file mode 100644 index 0000000000..bb83f7d121 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "st_context.h" +#include "st_atom.h" +#include "st_cb_texture.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + + + +static void +update_textures(struct st_context *st) +{ + GLuint u; + + for (u = 0; u < st->ctx->Const.MaxTextureImageUnits; u++) { + struct gl_texture_object *texObj + = st->ctx->Texture.Unit[u]._Current; + if (texObj) { + struct pipe_mipmap_tree *mt; + + mt = st_get_texobj_mipmap_tree(texObj); + + st->pipe->set_texture_state(st->pipe, u, mt); + } + else { + st->pipe->set_texture_state(st->pipe, u, NULL); + } + } +} + + +const struct st_tracked_state st_update_texture = { + .dirty = { + .mesa = _NEW_TEXTURE, + .st = 0, + }, + .update = update_textures +}; + + + + + -- cgit v1.2.3 From d78dab126724e6e9d475289a086fb6f85adc3985 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 15:12:22 -0600 Subject: plug in texture/sampler state update --- src/mesa/state_tracker/st_atom.c | 2 ++ src/mesa/state_tracker/st_atom.h | 2 ++ src/mesa/state_tracker/st_atom_sampler.c | 14 ++++++++------ src/mesa/state_tracker/st_atom_texture.c | 5 ++++- src/mesa/state_tracker/st_cb_texture.c | 8 ++++++++ src/mesa/state_tracker/st_cb_texture.h | 4 ++++ 6 files changed, 28 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 85c99bc182..32b8b8f277 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -54,6 +54,8 @@ static const struct st_tracked_state *atoms[] = &st_update_scissor, &st_update_blend, &st_update_stencil, + &st_update_sampler, + &st_update_texture, /* will be patched out at runtime */ /* &st_update_constants */ }; diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 1b70e27933..2f628206ca 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -57,6 +57,8 @@ const struct st_tracked_state st_update_constants; const struct st_tracked_state st_update_scissor; const struct st_tracked_state st_update_blend; const struct st_tracked_state st_update_stencil; +const struct st_tracked_state st_update_sampler; +const struct st_tracked_state st_update_texture; #endif diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 1aa9da8484..a49698cda4 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -103,14 +103,16 @@ update_samplers(struct st_context *st) memset(&sampler, 0, sizeof(sampler)); - sampler.wrap_s = gl_wrap_to_sp(texobj->WrapS); - sampler.wrap_t = gl_wrap_to_sp(texobj->WrapT); - sampler.wrap_r = gl_wrap_to_sp(texobj->WrapR); + if (texobj) { + sampler.wrap_s = gl_wrap_to_sp(texobj->WrapS); + sampler.wrap_t = gl_wrap_to_sp(texobj->WrapT); + sampler.wrap_r = gl_wrap_to_sp(texobj->WrapR); - sampler.min_filter = gl_filter_to_sp(texobj->MinFilter); - sampler.mag_filter = gl_filter_to_sp(texobj->MagFilter); + sampler.min_filter = gl_filter_to_sp(texobj->MinFilter); + sampler.mag_filter = gl_filter_to_sp(texobj->MagFilter); - /* XXX more sampler state here */ + /* XXX more sampler state here */ + } if (memcmp(&sampler, &st->state.sampler[u], sizeof(sampler)) != 0) { /* state has changed */ diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index bb83f7d121..f82c33e572 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -39,7 +39,10 @@ #include "pipe/p_defines.h" - +/** + * XXX This needs some work yet.... + * Need to "upload" texture images at appropriate times. + */ static void update_textures(struct st_context *st) { diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 5872ae3e74..42d6b75cb3 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -104,6 +104,14 @@ st_texture_image(struct gl_texture_image *img) } +struct pipe_mipmap_tree * +st_get_texobj_mipmap_tree(struct gl_texture_object *texObj) +{ + struct st_texture_object *stObj = st_texture_object(texObj); + return stObj->mt; +} + + static int intel_compressed_num_bytes(GLuint mesaFormat) { diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h index c732881c39..dc68aa3d97 100644 --- a/src/mesa/state_tracker/st_cb_texture.h +++ b/src/mesa/state_tracker/st_cb_texture.h @@ -2,6 +2,10 @@ #define ST_CB_TEXTURE_H +extern struct pipe_mipmap_tree * +st_get_texobj_mipmap_tree(struct gl_texture_object *texObj); + + extern GLuint st_finalize_mipmap_tree(GLcontext *ctx, struct pipe_context *pipe, GLuint unit, -- cgit v1.2.3 From d805d46a5fe1d163b299ea5934308e9264a57a29 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 15:12:40 -0600 Subject: added st_atom_texture.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index b648b6b8de..9bcb07c6b4 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -190,6 +190,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_setup.c \ state_tracker/st_atom_stencil.c \ state_tracker/st_atom_stipple.c \ + state_tracker/st_atom_texture.c \ state_tracker/st_atom_viewport.c \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_drawpixels.c \ -- cgit v1.2.3 From ab80be0df126066c441cef5e2042b77fa6e0bdde Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 15:27:03 -0600 Subject: remove pipe_surface_sampler --- src/mesa/pipe/p_state.h | 13 ------------- 1 file changed, 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index e51b9c460f..68616cf767 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -272,19 +272,6 @@ struct pipe_surface }; -/** - * surface sampler object - * XXX prototype - */ -struct pipe_surface_sampler -{ - struct pipe_sampler_state state; - struct pipe_mipmap_tree *texture; - void (*get_sample)(struct pipe_surface_sampler *sampler, - const GLfloat strq[4], GLfloat rgba[4]); -}; - - /** * Describes the location of each texture image within a texture region. */ -- cgit v1.2.3 From adf5560ea5f6955425939df7e75ce9d11ea36511 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 15:28:12 -0600 Subject: checkpoint new tgsi_sampler_state struct --- src/mesa/pipe/softpipe/sp_quad_fs.c | 20 ++- src/mesa/pipe/tgsi/core/tgsi_exec.c | 270 +++++++++++++++--------------------- src/mesa/pipe/tgsi/core/tgsi_exec.h | 13 +- 3 files changed, 138 insertions(+), 165 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 66513bb436..8eedb2ecc8 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -129,6 +129,16 @@ static INLINE void pinterp( struct exec_machine *exec, } +static void +get_sample(const struct tgsi_sampler_state *sampler, + const GLfloat strq[4], GLfloat rgba[4]) +{ + rgba[0] = 1; + rgba[1] = 1; + rgba[2] = 0; + rgba[3] = 0; +} + /* This should be done by the fragment shader execution unit (code * generated from the decl instructions). Do it here for now. @@ -196,6 +206,7 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) struct tgsi_exec_machine machine; struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; struct tgsi_exec_vector *aoutputs; + struct tgsi_sampler_state samplers[8]; GLuint i; #if !ALIGNED_ATTRIBS @@ -207,10 +218,17 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) memset(&machine, 0, sizeof(machine)); #endif +#if 11 /* temp sampler setup */ + samplers[0].state = &softpipe->sampler[0]; + samplers[0].texture = softpipe->texture[0]; + samplers[0].get_sample = get_sample; +#endif + /* init machine state */ tgsi_exec_machine_init( &machine, - softpipe->fs.tokens ); + softpipe->fs.tokens, + 8, samplers); /* Consts does not require 16 byte alignment. */ machine.Consts = softpipe->fs.constants->constant; diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 37ca7c1b2a..2bd2e7ce05 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -1,13 +1,12 @@ #include "tgsi_platform.h" #include "tgsi_core.h" +#include "pipe/p_state.h" #define MESA 1 #if MESA #include "main/context.h" #include "main/macros.h" #include "main/colormac.h" -#include "swrast/swrast.h" -#include "swrast/s_context.h" #endif #define TILE_BOTTOM_LEFT 0 @@ -64,13 +63,17 @@ void tgsi_exec_machine_init( struct tgsi_exec_machine *mach, - struct tgsi_token *tokens ) + struct tgsi_token *tokens, + GLuint numSamplers, + const struct tgsi_sampler_state *samplers) { GLuint i, k; struct tgsi_parse_context parse; mach->Tokens = tokens; + mach->Samplers = samplers; + k = tgsi_parse_init (&parse, mach->Tokens); if (k != TGSI_PARSE_OK) { printf("Problem parsing!\n"); @@ -1032,7 +1035,7 @@ exec_kil (struct tgsi_exec_machine *mach, */ static void fetch_texel_1d( GLcontext *ctx, - struct tgsi_sampler_state *sampler, + const struct tgsi_sampler_state *sampler, const union tgsi_exec_channel *s, GLuint unit, union tgsi_exec_channel *r, @@ -1040,68 +1043,48 @@ fetch_texel_1d( GLcontext *ctx, union tgsi_exec_channel *b, union tgsi_exec_channel *a ) { - SWcontext *swrast = SWRAST_CONTEXT(ctx); - GLuint fragment_index; - GLfloat stpq[4][4]; - GLfloat lambdas[4]; - GLchan rgba[4][4]; - - for (fragment_index = 0; fragment_index < 4; fragment_index++) - { - stpq[fragment_index][0] = s->f[fragment_index]; - } - - if (sampler->NeedLambda) - { - GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; - GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; - - GLfloat rho, lambda; - - dsdx = FABSF(dsdx); - dsdy = FABSF(dsdy); + GLuint fragment_index; + GLfloat stpq[4][4]; + GLfloat lambdas[4]; - rho = MAX2(dsdx, dsdy) * sampler->ImageWidth; + for (fragment_index = 0; fragment_index < 4; fragment_index++) { + stpq[fragment_index][0] = s->f[fragment_index]; + } - lambda = LOG2(rho); + if (sampler->state->min_filter != sampler->state->mag_filter) { + GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; + GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; + GLfloat rho, lambda; - if (sampler->NeedLodBias) - lambda += sampler->LodBias; + dsdx = FABSF(dsdx); + dsdy = FABSF(dsdy); - if (sampler->NeedLambdaClamp) - lambda = CLAMP(lambda, sampler->MinLod, sampler->MaxLod); + rho = MAX2(dsdx, dsdy) * sampler->texture->width0; - /* XXX: Use the same lambda value throughout the tile. Could - * end up with four unique values by recalculating partial - * derivs in the other row and column, and calculating lambda - * using the dx and dy values appropriate for each fragment in - * the tile. - */ - lambdas[0] = - lambdas[1] = - lambdas[2] = - lambdas[3] = lambda; - } + lambda = LOG2(rho); + lambda += sampler->state->lod_bias; + lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); - if (!swrast->TextureSample[unit]) { - _swrast_update_texture_samplers(ctx); - } - - /* XXX use a float-valued TextureSample routine here!!! */ - swrast->TextureSample[unit] (ctx, - ctx->Texture.Unit[unit]._Current, - 4, - (const GLfloat (*)[4])stpq, - lambdas, - rgba); + /* XXX: Use the same lambda value throughout the tile. Could + * end up with four unique values by recalculating partial + * derivs in the other row and column, and calculating lambda + * using the dx and dy values appropriate for each fragment in + * the tile. + */ + lambdas[0] = + lambdas[1] = + lambdas[2] = + lambdas[3] = lambda; + } - for (fragment_index = 0; fragment_index < 4; fragment_index++) - { - r->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][0]); - g->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][1]); - b->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][2]); - a->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][3]); - } + for (fragment_index = 0; fragment_index < 4; fragment_index++) { + GLfloat rgba[4]; + sampler->get_sample(sampler, stpq[fragment_index], rgba); + r->f[fragment_index] = rgba[0]; + g->f[fragment_index] = rgba[1]; + b->f[fragment_index] = rgba[2]; + a->f[fragment_index] = rgba[3]; + } } /* @@ -1109,7 +1092,7 @@ fetch_texel_1d( GLcontext *ctx, */ static void fetch_texel_2d( GLcontext *ctx, - struct tgsi_sampler_state *sampler, + const struct tgsi_sampler_state *sampler, const union tgsi_exec_channel *s, const union tgsi_exec_channel *t, GLuint unit, @@ -1118,18 +1101,16 @@ fetch_texel_2d( GLcontext *ctx, union tgsi_exec_channel *b, union tgsi_exec_channel *a ) { - SWcontext *swrast = SWRAST_CONTEXT( ctx ); GLuint fragment_index; GLfloat stpq[4][4]; GLfloat lambdas[4]; - GLchan rgba[4][4]; for (fragment_index = 0; fragment_index < 4; fragment_index++) { stpq[fragment_index][0] = s->f[fragment_index]; stpq[fragment_index][1] = t->f[fragment_index]; } - if (sampler->NeedLambda) { + if (sampler->state->min_filter != sampler->state->mag_filter) { GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; @@ -1143,21 +1124,15 @@ fetch_texel_2d( GLcontext *ctx, dtdx = FABSF( dtdx ); dtdy = FABSF( dtdy ); - maxU = MAX2( dsdx, dsdy ) * sampler->ImageWidth; - maxV = MAX2( dtdx, dtdy ) * sampler->ImageHeight; + maxU = MAX2( dsdx, dsdy ) * sampler->texture->width0; + maxV = MAX2( dtdx, dtdy ) * sampler->texture->height0; rho = MAX2( maxU, maxV ); lambda = LOG2( rho ); - if (sampler->NeedLodBias) - lambda += sampler->LodBias; - - if (sampler->NeedLambdaClamp) - lambda = CLAMP( - lambda, - sampler->MinLod, - sampler->MaxLod ); + lambda += sampler->state->lod_bias; + lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); /* XXX: Use the same lambda value throughout the tile. Could * end up with four unique values by recalculating partial @@ -1171,24 +1146,13 @@ fetch_texel_2d( GLcontext *ctx, lambdas[3] = lambda; } - if (!swrast->TextureSample[unit]) { - _swrast_update_texture_samplers(ctx); - } - - /* XXX use a float-valued TextureSample routine here!!! */ - swrast->TextureSample[unit]( - ctx, - ctx->Texture.Unit[unit]._Current, - 4, - (const GLfloat (*)[4]) stpq, - lambdas, - rgba ); - for (fragment_index = 0; fragment_index < 4; fragment_index++) { - r->f[fragment_index] = CHAN_TO_FLOAT( rgba[fragment_index][0] ); - g->f[fragment_index] = CHAN_TO_FLOAT( rgba[fragment_index][1] ); - b->f[fragment_index] = CHAN_TO_FLOAT( rgba[fragment_index][2] ); - a->f[fragment_index] = CHAN_TO_FLOAT( rgba[fragment_index][3] ); + GLfloat rgba[4]; + sampler->get_sample(sampler, stpq[fragment_index], rgba); + r->f[fragment_index] = rgba[0]; + g->f[fragment_index] = rgba[1]; + b->f[fragment_index] = rgba[2]; + a->f[fragment_index] = rgba[3]; } } @@ -1197,7 +1161,7 @@ fetch_texel_2d( GLcontext *ctx, */ static void fetch_texel_3d( GLcontext *ctx, - struct tgsi_sampler_state *sampler, + const struct tgsi_sampler_state *sampler, const union tgsi_exec_channel *s, const union tgsi_exec_channel *t, const union tgsi_exec_channel *p, @@ -1207,84 +1171,65 @@ fetch_texel_3d( GLcontext *ctx, union tgsi_exec_channel *b, union tgsi_exec_channel *a ) { - SWcontext *swrast = SWRAST_CONTEXT(ctx); - GLuint fragment_index; - GLfloat stpq[4][4]; - GLfloat lambdas[4]; - GLchan rgba[4][4]; - - for (fragment_index = 0; fragment_index < 4; fragment_index++) - { - stpq[fragment_index][0] = s->f[fragment_index]; - stpq[fragment_index][1] = t->f[fragment_index]; - stpq[fragment_index][2] = p->f[fragment_index]; - } - - if (sampler->NeedLambda) - { - GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; - GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; - - GLfloat dtdx = t->f[TILE_BOTTOM_RIGHT] - t->f[TILE_BOTTOM_LEFT]; - GLfloat dtdy = t->f[TILE_TOP_LEFT] - t->f[TILE_BOTTOM_LEFT]; - - GLfloat dpdx = p->f[TILE_BOTTOM_RIGHT] - p->f[TILE_BOTTOM_LEFT]; - GLfloat dpdy = p->f[TILE_TOP_LEFT] - p->f[TILE_BOTTOM_LEFT]; + GLuint fragment_index; + GLfloat stpq[4][4]; + GLfloat lambdas[4]; - GLfloat maxU, maxV, maxW, rho, lambda; + for (fragment_index = 0; fragment_index < 4; fragment_index++) { + stpq[fragment_index][0] = s->f[fragment_index]; + stpq[fragment_index][1] = t->f[fragment_index]; + stpq[fragment_index][2] = p->f[fragment_index]; + } - dsdx = FABSF(dsdx); - dsdy = FABSF(dsdy); - dtdx = FABSF(dtdx); - dtdy = FABSF(dtdy); - dpdx = FABSF(dpdx); - dpdy = FABSF(dpdy); + if (sampler->state->min_filter != sampler->state->mag_filter) { + GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; + GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; - maxU = MAX2(dsdx, dsdy) * sampler->ImageWidth; - maxV = MAX2(dtdx, dtdy) * sampler->ImageHeight; - maxW = MAX2(dpdx, dpdy) * sampler->ImageDepth; + GLfloat dtdx = t->f[TILE_BOTTOM_RIGHT] - t->f[TILE_BOTTOM_LEFT]; + GLfloat dtdy = t->f[TILE_TOP_LEFT] - t->f[TILE_BOTTOM_LEFT]; - rho = MAX2(maxU, MAX2(maxV, maxW)); + GLfloat dpdx = p->f[TILE_BOTTOM_RIGHT] - p->f[TILE_BOTTOM_LEFT]; + GLfloat dpdy = p->f[TILE_TOP_LEFT] - p->f[TILE_BOTTOM_LEFT]; - lambda = LOG2(rho); + GLfloat maxU, maxV, maxW, rho, lambda; - if (sampler->NeedLodBias) - lambda += sampler->LodBias; + dsdx = FABSF(dsdx); + dsdy = FABSF(dsdy); + dtdx = FABSF(dtdx); + dtdy = FABSF(dtdy); + dpdx = FABSF(dpdx); + dpdy = FABSF(dpdy); - if (sampler->NeedLambdaClamp) - lambda = CLAMP(lambda, sampler->MinLod, sampler->MaxLod); + maxU = MAX2(dsdx, dsdy) * sampler->texture->width0; + maxV = MAX2(dtdx, dtdy) * sampler->texture->height0; + maxW = MAX2(dpdx, dpdy) * sampler->texture->depth0; - /* XXX: Use the same lambda value throughout the tile. Could - * end up with four unique values by recalculating partial - * derivs in the other row and column, and calculating lambda - * using the dx and dy values appropriate for each fragment in - * the tile. - */ - lambdas[0] = - lambdas[1] = - lambdas[2] = - lambdas[3] = lambda; - } + rho = MAX2(maxU, MAX2(maxV, maxW)); - if (!swrast->TextureSample[unit]) { - _swrast_update_texture_samplers(ctx); - } + lambda = LOG2(rho); + lambda += sampler->state->lod_bias; + lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); - /* XXX use a float-valued TextureSample routine here!!! */ - swrast->TextureSample[unit] (ctx, - ctx->Texture.Unit[unit]._Current, - 4, - (const GLfloat (*)[4])stpq, - lambdas, - rgba); + /* XXX: Use the same lambda value throughout the tile. Could + * end up with four unique values by recalculating partial + * derivs in the other row and column, and calculating lambda + * using the dx and dy values appropriate for each fragment in + * the tile. + */ + lambdas[0] = + lambdas[1] = + lambdas[2] = + lambdas[3] = lambda; + } - for (fragment_index = 0; fragment_index < 4; fragment_index++) - { - r->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][0]); - g->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][1]); - b->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][2]); - a->f[fragment_index] = CHAN_TO_FLOAT(rgba[fragment_index][3]); - } + for (fragment_index = 0; fragment_index < 4; fragment_index++) { + GLfloat rgba[4]; + sampler->get_sample(sampler, stpq[fragment_index], rgba); + r->f[fragment_index] = rgba[0]; + g->f[fragment_index] = rgba[1]; + b->f[fragment_index] = rgba[2]; + a->f[fragment_index] = rgba[3]; + } } #endif @@ -1314,7 +1259,7 @@ exec_instruction( #if MESA GET_CURRENT_CONTEXT(ctx); #endif - GLuint chan_index; + GLuint chan_index, unit; union tgsi_exec_channel r[8]; switch (inst->Instruction.Opcode) { @@ -1830,6 +1775,7 @@ exec_instruction( break; case TGSI_OPCODE_TEX: + unit = inst->FullSrcRegisters[1].SrcRegister.Index; switch (inst->InstructionExtTexture.Texture) { case TGSI_TEXTURE_1D: @@ -1849,7 +1795,7 @@ exec_instruction( } #if MESA fetch_texel_1d (ctx, - &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], + &mach->Samplers[unit], &r[0], inst->FullSrcRegisters[1].SrcRegister.Index, &r[0], &r[1], &r[2], &r[3]); @@ -1878,7 +1824,7 @@ exec_instruction( #if MESA fetch_texel_2d (ctx, - &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], + &mach->Samplers[unit], &r[0], &r[1], inst->FullSrcRegisters[1].SrcRegister.Index, &r[0], &r[1], &r[2], &r[3]); @@ -1909,7 +1855,7 @@ exec_instruction( #if MESA fetch_texel_3d (ctx, - &mach->Samplers[inst->FullSrcRegisters[1].SrcRegister.Index], + &mach->Samplers[unit], &r[0], &r[1], &r[2], inst->FullSrcRegisters[1].SrcRegister.Index, &r[0], &r[1], &r[2], &r[3]); diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index abce6ee924..d2ff40cabb 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -23,6 +23,12 @@ struct tgsi_exec_vector struct tgsi_sampler_state { + const struct pipe_sampler_state *state; + const struct pipe_mipmap_tree *texture; + void (*get_sample)(const struct tgsi_sampler_state *sampler, + const GLfloat strq[4], GLfloat rgba[4]); + +#if 0 GLboolean NeedLambda; GLboolean NeedLodBias; /* if NeedLambda */ GLboolean NeedLambdaClamp; /* if NeedLambda */ @@ -32,6 +38,7 @@ struct tgsi_sampler_state GLfloat ImageWidth; GLfloat ImageHeight; GLfloat ImageDepth; +#endif }; struct tgsi_exec_labels @@ -94,7 +101,7 @@ struct tgsi_exec_machine struct tgsi_exec_vector *Temps; struct tgsi_exec_vector *Addrs; - struct tgsi_sampler_state Samplers[16]; + const struct tgsi_sampler_state *Samplers; GLfloat Imms[256][4]; GLuint ImmLimit; @@ -113,7 +120,9 @@ struct tgsi_exec_machine void tgsi_exec_machine_init( struct tgsi_exec_machine *mach, - struct tgsi_token *tokens ); + struct tgsi_token *tokens, + GLuint numSamplers, + const struct tgsi_sampler_state *samplers); void tgsi_exec_prepare( -- cgit v1.2.3 From 4a470f63c9c3f5aeef1e10ae42eee24c772f74fc Mon Sep 17 00:00:00 2001 From: michal Date: Tue, 7 Aug 2007 15:34:11 +0100 Subject: Naive implementation of IF/ELSE/ENDIF. Handle TGSI labels correctly. Enhance MESA opcode info queries. --- src/mesa/pipe/tgsi/core/tgsi_dump.c | 2 +- src/mesa/pipe/tgsi/core/tgsi_exec.c | 174 ++++++++++++++++++++++++++----- src/mesa/pipe/tgsi/core/tgsi_exec.h | 25 +++++ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 +- src/mesa/shader/prog_instruction.c | 183 ++++++++++++++++++--------------- src/mesa/shader/prog_instruction.h | 3 + src/mesa/shader/shader_api.c | 2 +- 7 files changed, 277 insertions(+), 114 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index 641339c896..f1773ac908 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -788,7 +788,7 @@ dump_instruction_short( switch( inst->Instruction.Opcode ) { case TGSI_OPCODE_IF: case TGSI_OPCODE_ELSE: - TXT( " : " ); + TXT( " :" ); UID( inst->InstructionExtLabel.Label ); break; } diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 2bd2e7ce05..71dfbc10f3 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -111,6 +111,7 @@ tgsi_exec_prepare( { struct tgsi_parse_context parse; GLuint k; + GLuint instno = 0; mach->ImmLimit = 0; labels->count = 0; @@ -124,26 +125,30 @@ tgsi_exec_prepare( while( !tgsi_parse_end_of_tokens( &parse ) ) { GLuint pointer = parse.Position; GLuint i; + tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: break; + case TGSI_TOKEN_TYPE_IMMEDIATE: assert( (parse.FullToken.FullImmediate.Immediate.Size - 1) % 4 == 0 ); assert( mach->ImmLimit + (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4 <= 256 ); + for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size - 1; i++ ) { mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; } mach->ImmLimit += (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4; break; + case TGSI_TOKEN_TYPE_INSTRUCTION: - if( parse.FullToken.FullInstruction.InstructionExtLabel.Label ) { - assert( labels->count < 128 ); - labels->labels[labels->count][0] = parse.FullToken.FullInstruction.InstructionExtLabel.Label; - labels->labels[labels->count][1] = pointer; - labels->count++; - } + assert( labels->count < 128 ); + + labels->labels[labels->count][0] = instno; + labels->labels[labels->count][1] = pointer; + labels->count++; break; + default: assert( 0 ); } @@ -1233,22 +1238,6 @@ fetch_texel_3d( GLcontext *ctx, } #endif -static GLuint -map_label( - GLuint label, - struct tgsi_exec_labels *labels ) -{ - GLuint i; - - for( i = 0; i < labels->count; i++ ) { - if( labels->labels[i][0] == label ) { - return labels->labels[i][1]; - } - } - assert( 0 ); - return 0; -} - static void exec_instruction( struct tgsi_exec_machine *mach, @@ -1988,7 +1977,43 @@ exec_instruction( break; case TGSI_OPCODE_IF: - assert (0); + { + GLuint cond = 0; + struct tgsi_exec_cond_state *state; + + /* Allocate condition state. */ + assert( mach->CondStack.Index > 0 ); + mach->CondStack.Index--; + + /* Evaluate the condition mask. */ + FETCH( &r[0], 0, CHAN_X ); + if( r[0].u[0] ) { + cond |= 1; + } + if( r[0].u[1] ) { + cond |= 2; + } + if( r[0].u[2] ) { + cond |= 4; + } + if( r[0].u[3] ) { + cond |= 8; + } + + state = &mach->CondStack.States[mach->CondStack.Index]; + + /* Initialize the If portion of condition state. */ + memcpy( + state->IfPortion.TempsAddrs, + mach->Temps, + sizeof( state->IfPortion.TempsAddrs ) ); + memcpy( + state->IfPortion.Outputs, + mach->Outputs, + sizeof( state->IfPortion.Outputs ) ); + state->Condition = cond; + state->WasElse = GL_FALSE; + } break; case TGSI_OPCODE_LOOP: @@ -2000,12 +2025,107 @@ exec_instruction( break; case TGSI_OPCODE_ELSE: - assert (0); + { + struct tgsi_exec_cond_state *state; + struct tgsi_exec_cond_regs temp; + + state = &mach->CondStack.States[mach->CondStack.Index]; + + /* Copy the results of the If portion to temporary storage. */ + memcpy( + temp.TempsAddrs, + mach->Temps, + sizeof( temp.TempsAddrs ) ); + memcpy( + temp.Outputs, + mach->Outputs, + sizeof( temp.Outputs ) ); + + /* Restore the state of registers from before the If statement. */ + memcpy( + mach->Temps, + state->IfPortion.TempsAddrs, + sizeof( state->IfPortion.TempsAddrs ) ); + memcpy( + mach->Outputs, + state->IfPortion.Outputs, + sizeof( state->IfPortion.Outputs ) ); + + /* Save the results of If portion. */ + memcpy( + &state->IfPortion, + &temp, + sizeof( state->IfPortion ) ); + state->WasElse = GL_TRUE; + } break; case TGSI_OPCODE_ENDIF: - assert (0); - break; + { + struct tgsi_exec_cond_state *state; + GLuint i; + + state = &mach->CondStack.States[mach->CondStack.Index]; + + if( state->WasElse ) { + /* Save the results of Else portion. */ + memcpy( + state->ElsePortion.TempsAddrs, + mach->Temps, + sizeof( state->ElsePortion.TempsAddrs ) ); + memcpy( + state->ElsePortion.Outputs, + mach->Outputs, + sizeof( state->ElsePortion.Outputs ) ); + } + else { + /* Copy the state of registers from before the If statement to Else portion. */ + memcpy( + &state->ElsePortion, + &state->IfPortion, + sizeof( state->ElsePortion ) ); + + /* Save the results of the If portion. */ + memcpy( + state->IfPortion.TempsAddrs, + mach->Temps, + sizeof( state->IfPortion.TempsAddrs ) ); + memcpy( + state->IfPortion.Outputs, + mach->Outputs, + sizeof( state->IfPortion.Outputs ) ); + } + + /* Mix the If and Else portions based on condition mask. */ + for( i = 0; i < 4; i++ ) { + struct tgsi_exec_cond_regs *regs; + GLuint j; + + if( state->Condition & (1 << i) ) { + regs = &state->IfPortion; + } + else { + regs = &state->ElsePortion; + } + + for( j = 0; j < TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS; j++ ) { + mach->Temps[j].xyzw[0].u[i] = regs->TempsAddrs[j].xyzw[0].u[i]; + mach->Temps[j].xyzw[1].u[i] = regs->TempsAddrs[j].xyzw[1].u[i]; + mach->Temps[j].xyzw[2].u[i] = regs->TempsAddrs[j].xyzw[2].u[i]; + mach->Temps[j].xyzw[3].u[i] = regs->TempsAddrs[j].xyzw[3].u[i]; + } + for( j = 0; j < 2; j++ ) { + mach->Outputs[j].xyzw[0].u[i] = regs->Outputs[j].xyzw[0].u[i]; + mach->Outputs[j].xyzw[1].u[i] = regs->Outputs[j].xyzw[1].u[i]; + mach->Outputs[j].xyzw[2].u[i] = regs->Outputs[j].xyzw[2].u[i]; + mach->Outputs[j].xyzw[3].u[i] = regs->Outputs[j].xyzw[3].u[i]; + } + } + + /* Release condition state. */ + mach->CondStack.Index++; + } + break; case TGSI_OPCODE_ENDLOOP: assert (0); @@ -2214,6 +2334,8 @@ tgsi_exec_machine_run2( mach->Primitives[0] = 0; } + mach->CondStack.Index = 8; + k = tgsi_parse_init( &parse, mach->Tokens ); if (k != TGSI_PARSE_OK) { printf("Problem parsing!\n"); diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index d2ff40cabb..b92d1b5918 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -85,6 +85,29 @@ struct tgsi_exec_labels #define TGSI_EXEC_NUM_TEMPS (32 + 4) #define TGSI_EXEC_NUM_ADDRS 1 +/* XXX: This is temporary */ +struct tgsi_exec_cond_regs +{ + struct tgsi_exec_vector TempsAddrs[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS]; + struct tgsi_exec_vector Outputs[2]; /* XXX: That's just enough for fragment shader only! */ +}; + +/* XXX: This is temporary */ +struct tgsi_exec_cond_state +{ + struct tgsi_exec_cond_regs IfPortion; + struct tgsi_exec_cond_regs ElsePortion; + GLuint Condition; + GLboolean WasElse; +}; + +/* XXX: This is temporary */ +struct tgsi_exec_cond_stack +{ + struct tgsi_exec_cond_state States[8]; + GLuint Index; /* into States[] */ +}; + struct tgsi_exec_machine { /* @@ -112,6 +135,8 @@ struct tgsi_exec_machine GLuint Processor; GLuint *Primitives; + + struct tgsi_exec_cond_stack CondStack; #if XXX_SSE struct x86_function Function; #endif diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 654ec9ac68..107b6bd2a3 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -168,7 +168,7 @@ compile_instruction( *fullinst = tgsi_default_full_instruction(); fullinst->Instruction.Saturate = convert_sat( inst->SaturateMode ); - fullinst->Instruction.NumDstRegs = 1; + fullinst->Instruction.NumDstRegs = _mesa_num_inst_dst_regs( inst->Opcode ); fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode ); fulldst = &fullinst->FullDstRegisters[0]; diff --git a/src/mesa/shader/prog_instruction.c b/src/mesa/shader/prog_instruction.c index d6b5652a27..c84c76fd5b 100644 --- a/src/mesa/shader/prog_instruction.c +++ b/src/mesa/shader/prog_instruction.c @@ -126,6 +126,7 @@ struct instruction_info gl_inst_opcode Opcode; const char *Name; GLuint NumSrcRegs; + GLuint NumDstRegs; }; /** @@ -133,91 +134,91 @@ struct instruction_info * \note Opcode should equal array index! */ static const struct instruction_info InstInfo[MAX_OPCODE] = { - { OPCODE_NOP, "NOP", 0 }, - { OPCODE_ABS, "ABS", 1 }, - { OPCODE_ADD, "ADD", 2 }, - { OPCODE_ARA, "ARA", 1 }, - { OPCODE_ARL, "ARL", 1 }, - { OPCODE_ARL_NV, "ARL", 1 }, - { OPCODE_ARR, "ARL", 1 }, - { OPCODE_BGNLOOP,"BGNLOOP", 0 }, - { OPCODE_BGNSUB, "BGNSUB", 0 }, - { OPCODE_BRA, "BRA", 0 }, - { OPCODE_BRK, "BRK", 0 }, - { OPCODE_CAL, "CAL", 0 }, - { OPCODE_CMP, "CMP", 3 }, - { OPCODE_CONT, "CONT", 0 }, - { OPCODE_COS, "COS", 1 }, - { OPCODE_DDX, "DDX", 1 }, - { OPCODE_DDY, "DDY", 1 }, - { OPCODE_DP3, "DP3", 2 }, - { OPCODE_DP4, "DP4", 2 }, - { OPCODE_DPH, "DPH", 2 }, - { OPCODE_DST, "DST", 2 }, - { OPCODE_ELSE, "ELSE", 0 }, - { OPCODE_END, "END", 0 }, - { OPCODE_ENDIF, "ENDIF", 0 }, - { OPCODE_ENDLOOP,"ENDLOOP", 0 }, - { OPCODE_ENDSUB, "ENDSUB", 0 }, - { OPCODE_EX2, "EX2", 1 }, - { OPCODE_EXP, "EXP", 1 }, - { OPCODE_FLR, "FLR", 1 }, - { OPCODE_FRC, "FRC", 1 }, - { OPCODE_IF, "IF", 0 }, - { OPCODE_INT, "INT", 1 }, - { OPCODE_KIL, "KIL", 1 }, - { OPCODE_KIL_NV, "KIL", 0 }, - { OPCODE_LG2, "LG2", 1 }, - { OPCODE_LIT, "LIT", 1 }, - { OPCODE_LOG, "LOG", 1 }, - { OPCODE_LRP, "LRP", 3 }, - { OPCODE_MAD, "MAD", 3 }, - { OPCODE_MAX, "MAX", 2 }, - { OPCODE_MIN, "MIN", 2 }, - { OPCODE_MOV, "MOV", 1 }, - { OPCODE_MUL, "MUL", 2 }, - { OPCODE_NOISE1, "NOISE1", 1 }, - { OPCODE_NOISE2, "NOISE2", 1 }, - { OPCODE_NOISE3, "NOISE3", 1 }, - { OPCODE_NOISE4, "NOISE4", 1 }, - { OPCODE_PK2H, "PK2H", 1 }, - { OPCODE_PK2US, "PK2US", 1 }, - { OPCODE_PK4B, "PK4B", 1 }, - { OPCODE_PK4UB, "PK4UB", 1 }, - { OPCODE_POW, "POW", 2 }, - { OPCODE_POPA, "POPA", 0 }, - { OPCODE_PRINT, "PRINT", 1 }, - { OPCODE_PUSHA, "PUSHA", 0 }, - { OPCODE_RCC, "RCC", 1 }, - { OPCODE_RCP, "RCP", 1 }, - { OPCODE_RET, "RET", 0 }, - { OPCODE_RFL, "RFL", 1 }, - { OPCODE_RSQ, "RSQ", 1 }, - { OPCODE_SCS, "SCS", 1 }, - { OPCODE_SEQ, "SEQ", 2 }, - { OPCODE_SFL, "SFL", 0 }, - { OPCODE_SGE, "SGE", 2 }, - { OPCODE_SGT, "SGT", 2 }, - { OPCODE_SIN, "SIN", 1 }, - { OPCODE_SLE, "SLE", 2 }, - { OPCODE_SLT, "SLT", 2 }, - { OPCODE_SNE, "SNE", 2 }, - { OPCODE_SSG, "SSG", 1 }, - { OPCODE_STR, "STR", 0 }, - { OPCODE_SUB, "SUB", 2 }, - { OPCODE_SWZ, "SWZ", 1 }, - { OPCODE_TEX, "TEX", 1 }, - { OPCODE_TXB, "TXB", 1 }, - { OPCODE_TXD, "TXD", 3 }, - { OPCODE_TXL, "TXL", 1 }, - { OPCODE_TXP, "TXP", 1 }, - { OPCODE_TXP_NV, "TXP", 1 }, - { OPCODE_UP2H, "UP2H", 1 }, - { OPCODE_UP2US, "UP2US", 1 }, - { OPCODE_UP4B, "UP4B", 1 }, - { OPCODE_UP4UB, "UP4UB", 1 }, - { OPCODE_X2D, "X2D", 3 }, - { OPCODE_XPD, "XPD", 2 } + { OPCODE_NOP, "NOP", 0, 0 }, + { OPCODE_ABS, "ABS", 1, 1 }, + { OPCODE_ADD, "ADD", 2, 1 }, + { OPCODE_ARA, "ARA", 1, 1 }, + { OPCODE_ARL, "ARL", 1, 1 }, + { OPCODE_ARL_NV, "ARL", 1, 1 }, + { OPCODE_ARR, "ARL", 1, 1 }, + { OPCODE_BGNLOOP,"BGNLOOP", 0, 0 }, + { OPCODE_BGNSUB, "BGNSUB", 0, 0 }, + { OPCODE_BRA, "BRA", 0, 0 }, + { OPCODE_BRK, "BRK", 0, 0 }, + { OPCODE_CAL, "CAL", 0, 0 }, + { OPCODE_CMP, "CMP", 3, 1 }, + { OPCODE_CONT, "CONT", 0, 0 }, + { OPCODE_COS, "COS", 1, 1 }, + { OPCODE_DDX, "DDX", 1, 1 }, + { OPCODE_DDY, "DDY", 1, 1 }, + { OPCODE_DP3, "DP3", 2, 1 }, + { OPCODE_DP4, "DP4", 2, 1 }, + { OPCODE_DPH, "DPH", 2, 1 }, + { OPCODE_DST, "DST", 2, 1 }, + { OPCODE_ELSE, "ELSE", 0, 0 }, + { OPCODE_END, "END", 0, 0 }, + { OPCODE_ENDIF, "ENDIF", 0, 0 }, + { OPCODE_ENDLOOP,"ENDLOOP", 0, 0 }, + { OPCODE_ENDSUB, "ENDSUB", 0, 0 }, + { OPCODE_EX2, "EX2", 1, 1 }, + { OPCODE_EXP, "EXP", 1, 1 }, + { OPCODE_FLR, "FLR", 1, 1 }, + { OPCODE_FRC, "FRC", 1, 1 }, + { OPCODE_IF, "IF", 1, 0 }, + { OPCODE_INT, "INT", 1, 1 }, + { OPCODE_KIL, "KIL", 1, 0 }, + { OPCODE_KIL_NV, "KIL", 0, 0 }, + { OPCODE_LG2, "LG2", 1, 1 }, + { OPCODE_LIT, "LIT", 1, 1 }, + { OPCODE_LOG, "LOG", 1, 1 }, + { OPCODE_LRP, "LRP", 3, 1 }, + { OPCODE_MAD, "MAD", 3, 1 }, + { OPCODE_MAX, "MAX", 2, 1 }, + { OPCODE_MIN, "MIN", 2, 1 }, + { OPCODE_MOV, "MOV", 1, 1 }, + { OPCODE_MUL, "MUL", 2, 1 }, + { OPCODE_NOISE1, "NOISE1", 1, 1 }, + { OPCODE_NOISE2, "NOISE2", 1, 1 }, + { OPCODE_NOISE3, "NOISE3", 1, 1 }, + { OPCODE_NOISE4, "NOISE4", 1, 1 }, + { OPCODE_PK2H, "PK2H", 1, 1 }, + { OPCODE_PK2US, "PK2US", 1, 1 }, + { OPCODE_PK4B, "PK4B", 1, 1 }, + { OPCODE_PK4UB, "PK4UB", 1, 1 }, + { OPCODE_POW, "POW", 2, 1 }, + { OPCODE_POPA, "POPA", 0, 0 }, + { OPCODE_PRINT, "PRINT", 1, 0 }, + { OPCODE_PUSHA, "PUSHA", 0, 0 }, + { OPCODE_RCC, "RCC", 1, 1 }, + { OPCODE_RCP, "RCP", 1, 1 }, + { OPCODE_RET, "RET", 0, 0 }, + { OPCODE_RFL, "RFL", 1, 1 }, + { OPCODE_RSQ, "RSQ", 1, 1 }, + { OPCODE_SCS, "SCS", 1, 1 }, + { OPCODE_SEQ, "SEQ", 2, 1 }, + { OPCODE_SFL, "SFL", 0, 1 }, + { OPCODE_SGE, "SGE", 2, 1 }, + { OPCODE_SGT, "SGT", 2, 1 }, + { OPCODE_SIN, "SIN", 1, 1 }, + { OPCODE_SLE, "SLE", 2, 1 }, + { OPCODE_SLT, "SLT", 2, 1 }, + { OPCODE_SNE, "SNE", 2, 1 }, + { OPCODE_SSG, "SSG", 1, 1 }, + { OPCODE_STR, "STR", 0, 1 }, + { OPCODE_SUB, "SUB", 2, 1 }, + { OPCODE_SWZ, "SWZ", 1, 1 }, + { OPCODE_TEX, "TEX", 1, 1 }, + { OPCODE_TXB, "TXB", 1, 1 }, + { OPCODE_TXD, "TXD", 3, 1 }, + { OPCODE_TXL, "TXL", 1, 1 }, + { OPCODE_TXP, "TXP", 1, 1 }, + { OPCODE_TXP_NV, "TXP", 1, 1 }, + { OPCODE_UP2H, "UP2H", 1, 1 }, + { OPCODE_UP2US, "UP2US", 1, 1 }, + { OPCODE_UP4B, "UP4B", 1, 1 }, + { OPCODE_UP4UB, "UP4UB", 1, 1 }, + { OPCODE_X2D, "X2D", 3, 1 }, + { OPCODE_XPD, "XPD", 2, 1 } }; @@ -233,6 +234,18 @@ _mesa_num_inst_src_regs(gl_inst_opcode opcode) } +/** + * Return the number of dst registers for the given instruction/opcode. + */ +GLuint +_mesa_num_inst_dst_regs(gl_inst_opcode opcode) +{ + ASSERT(opcode == InstInfo[opcode].Opcode); + ASSERT(OPCODE_XPD == InstInfo[OPCODE_XPD].Opcode); + return InstInfo[opcode].NumDstRegs; +} + + /** * Return string name for given program opcode. */ diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h index c800757aa0..643969b367 100644 --- a/src/mesa/shader/prog_instruction.h +++ b/src/mesa/shader/prog_instruction.h @@ -440,6 +440,9 @@ _mesa_copy_instructions(struct prog_instruction *dest, extern GLuint _mesa_num_inst_src_regs(gl_inst_opcode opcode); +extern GLuint +_mesa_num_inst_dst_regs(gl_inst_opcode opcode); + extern const char * _mesa_opcode_string(gl_inst_opcode opcode); diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 1a931326af..9f6c54dd4c 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -332,7 +332,7 @@ _mesa_init_shader_state(GLcontext * ctx) * are generated by the GLSL compiler. */ ctx->Shader.EmitHighLevelInstructions = GL_TRUE; - ctx->Shader.EmitCondCodes = GL_TRUE; /* XXX probably want GL_FALSE... */ + ctx->Shader.EmitCondCodes = GL_FALSE;/*GL_TRUE;*/ /* XXX probably want GL_FALSE... */ ctx->Shader.EmitComments = GL_FALSE; } -- cgit v1.2.3 From 14b98343309fdcff3514f05020303f7b40e83a4a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 16:42:08 -0600 Subject: s/intel/st/ --- src/mesa/state_tracker/st_cb_texture.c | 252 ++++++++++++++++----------------- src/mesa/state_tracker/st_cb_texture.h | 2 +- 2 files changed, 127 insertions(+), 127 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 42d6b75cb3..b37e0cf0af 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -113,7 +113,7 @@ st_get_texobj_mipmap_tree(struct gl_texture_object *texObj) static int -intel_compressed_num_bytes(GLuint mesaFormat) +compressed_num_bytes(GLuint mesaFormat) { int bytes = 0; switch(mesaFormat) { @@ -443,42 +443,42 @@ logbase2(int n) */ static void guess_and_alloc_mipmap_tree(struct pipe_context *pipe, - struct st_texture_object *intelObj, - struct st_texture_image *intelImage) + struct st_texture_object *stObj, + struct st_texture_image *stImage) { GLuint firstLevel; GLuint lastLevel; - GLuint width = intelImage->base.Width; - GLuint height = intelImage->base.Height; - GLuint depth = intelImage->base.Depth; + GLuint width = stImage->base.Width; + GLuint height = stImage->base.Height; + GLuint depth = stImage->base.Depth; GLuint l2width, l2height, l2depth; GLuint i, comp_byte = 0; DBG("%s\n", __FUNCTION__); - if (intelImage->base.Border) + if (stImage->base.Border) return; - if (intelImage->level > intelObj->base.BaseLevel && - (intelImage->base.Width == 1 || - (intelObj->base.Target != GL_TEXTURE_1D && - intelImage->base.Height == 1) || - (intelObj->base.Target == GL_TEXTURE_3D && - intelImage->base.Depth == 1))) + if (stImage->level > stObj->base.BaseLevel && + (stImage->base.Width == 1 || + (stObj->base.Target != GL_TEXTURE_1D && + stImage->base.Height == 1) || + (stObj->base.Target == GL_TEXTURE_3D && + stImage->base.Depth == 1))) return; /* If this image disrespects BaseLevel, allocate from level zero. * Usually BaseLevel == 0, so it's unlikely to happen. */ - if (intelImage->level < intelObj->base.BaseLevel) + if (stImage->level < stObj->base.BaseLevel) firstLevel = 0; else - firstLevel = intelObj->base.BaseLevel; + firstLevel = stObj->base.BaseLevel; /* Figure out image dimensions at start level. */ - for (i = intelImage->level; i > firstLevel; i--) { + for (i = stImage->level; i > firstLevel; i--) { width <<= 1; if (height != 1) height <<= 1; @@ -491,9 +491,9 @@ guess_and_alloc_mipmap_tree(struct pipe_context *pipe, * resizable buffers, or require that buffers implement lazy * pagetable arrangements. */ - if ((intelObj->base.MinFilter == GL_NEAREST || - intelObj->base.MinFilter == GL_LINEAR) && - intelImage->level == firstLevel) { + if ((stObj->base.MinFilter == GL_NEAREST || + stObj->base.MinFilter == GL_LINEAR) && + stImage->level == firstLevel) { lastLevel = firstLevel; } else { @@ -503,18 +503,18 @@ guess_and_alloc_mipmap_tree(struct pipe_context *pipe, lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); } - assert(!intelObj->mt); - if (intelImage->base.IsCompressed) - comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); - intelObj->mt = st_miptree_create(pipe, - intelObj->base.Target, - intelImage->base.InternalFormat, + assert(!stObj->mt); + if (stImage->base.IsCompressed) + comp_byte = compressed_num_bytes(stImage->base.TexFormat->MesaFormat); + stObj->mt = st_miptree_create(pipe, + stObj->base.Target, + stImage->base.InternalFormat, firstLevel, lastLevel, width, height, depth, - intelImage->base.TexFormat->TexelBytes, + stImage->base.TexFormat->TexelBytes, comp_byte); DBG("%s - success\n", __FUNCTION__); @@ -573,7 +573,7 @@ check_pbo_format(GLint internalFormat, */ static GLboolean try_pbo_upload(GLcontext *ctx, - struct st_texture_image *intelImage, + struct st_texture_image *stImage, const struct gl_pixelstore_attrib *unpack, GLint internalFormat, GLint width, GLint height, @@ -600,11 +600,11 @@ try_pbo_upload(GLcontext *ctx, else src_stride = width; - dst_offset = st_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); + dst_offset = st_miptree_image_offset(stImage->mt, + stImage->face, + stImage->level); - dst_stride = intelImage->mt->pitch; + dst_stride = stImage->mt->pitch; intelFlush(&intel->ctx); LOCK_HARDWARE(intel); @@ -615,11 +615,11 @@ try_pbo_upload(GLcontext *ctx, /* Temporary hack: cast to _DriBufferObject: */ struct _DriBufferObject *dst_buffer = - (struct _DriBufferObject *)intelImage->mt->region->buffer; + (struct _DriBufferObject *)stImage->mt->region->buffer; intelEmitCopyBlit(intel, - intelImage->mt->cpp, + stImage->mt->cpp, src_stride, src_buffer, src_offset, dst_stride, dst_buffer, dst_offset, 0, 0, 0, 0, width, height, @@ -637,7 +637,7 @@ try_pbo_upload(GLcontext *ctx, static GLboolean try_pbo_zcopy(GLcontext *ctx, - struct st_texture_image *intelImage, + struct st_texture_image *stImage, const struct gl_pixelstore_attrib *unpack, GLint internalFormat, GLint width, GLint height, @@ -664,8 +664,8 @@ st_TexImage(GLcontext * ctx, struct gl_texture_image *texImage, GLsizei imageSize, int compressed) { struct pipe_context *pipe = ctx->st->pipe; - struct st_texture_object *intelObj = st_texture_object(texObj); - struct st_texture_image *intelImage = st_texture_image(texImage); + struct st_texture_object *stObj = st_texture_object(texObj); + struct st_texture_image *stImage = st_texture_image(texImage); GLint postConvWidth = width; GLint postConvHeight = height; GLint texelBytes, sizeInBytes; @@ -679,8 +679,8 @@ st_TexImage(GLcontext * ctx, intelFlush(ctx); #endif - intelImage->face = target_to_face(target); - intelImage->level = level; + stImage->face = target_to_face(target); + stImage->level = level; if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth, @@ -716,8 +716,8 @@ st_TexImage(GLcontext * ctx, /* Release the reference to a potentially orphaned buffer. * Release any old malloced memory. */ - if (intelImage->mt) { - st_miptree_release(pipe, &intelImage->mt); + if (stImage->mt) { + st_miptree_release(pipe, &stImage->mt); assert(!texImage->Data); } else if (texImage->Data) { @@ -728,46 +728,46 @@ st_TexImage(GLcontext * ctx, * bmBufferData with NULL data to free the old block and avoid * waiting on any outstanding fences. */ - if (intelObj->mt && - intelObj->mt->first_level == level && - intelObj->mt->last_level == level && - intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && - !st_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { + if (stObj->mt && + stObj->mt->first_level == level && + stObj->mt->last_level == level && + stObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && + !st_miptree_match_image(stObj->mt, &stImage->base, + stImage->face, stImage->level)) { DBG("release it\n"); - st_miptree_release(pipe, &intelObj->mt); - assert(!intelObj->mt); + st_miptree_release(pipe, &stObj->mt); + assert(!stObj->mt); } - if (!intelObj->mt) { - guess_and_alloc_mipmap_tree(pipe, intelObj, intelImage); - if (!intelObj->mt) { + if (!stObj->mt) { + guess_and_alloc_mipmap_tree(pipe, stObj, stImage); + if (!stObj->mt) { DBG("guess_and_alloc_mipmap_tree: failed\n"); } } - assert(!intelImage->mt); + assert(!stImage->mt); - if (intelObj->mt && - st_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { + if (stObj->mt && + st_miptree_match_image(stObj->mt, &stImage->base, + stImage->face, stImage->level)) { - st_miptree_reference(&intelImage->mt, intelObj->mt); - assert(intelImage->mt); + st_miptree_reference(&stImage->mt, stObj->mt); + assert(stImage->mt); } - if (!intelImage->mt) + if (!stImage->mt) DBG("XXX: Image did not fit into tree - storing in local memory!\n"); #if 0 /* XXX FIX when st_buffer_objects are in place */ /* PBO fastpaths: */ if (dims <= 2 && - intelImage->mt && + stImage->mt && intel_buffer_object(unpack->BufferObj) && check_pbo_format(internalFormat, format, - type, intelImage->base.TexFormat)) { + type, stImage->base.TexFormat)) { DBG("trying pbo upload\n"); @@ -777,11 +777,11 @@ st_TexImage(GLcontext * ctx, * performance (in particular when pipe_region_cow() is * required). */ - if (intelObj->mt == intelImage->mt && - intelObj->mt->first_level == level && - intelObj->mt->last_level == level) { + if (stObj->mt == stImage->mt && + stObj->mt->first_level == level && + stObj->mt->last_level == level) { - if (try_pbo_zcopy(intel, intelImage, unpack, + if (try_pbo_zcopy(intel, stImage, unpack, internalFormat, width, height, format, type, pixels)) { @@ -793,7 +793,7 @@ st_TexImage(GLcontext * ctx, /* Otherwise, attempt to use the blitter for PBO image uploads. */ - if (try_pbo_upload(intel, intelImage, unpack, + if (try_pbo_upload(intel, stImage, unpack, internalFormat, width, height, format, type, pixels)) { DBG("pbo upload succeeded\n"); @@ -826,20 +826,20 @@ st_TexImage(GLcontext * ctx, return; - if (intelImage->mt) - pipe->region_idle(pipe, intelImage->mt->region); + if (stImage->mt) + pipe->region_idle(pipe, stImage->mt->region); #if 0 LOCK_HARDWARE(intel); #endif - if (intelImage->mt) { + if (stImage->mt) { texImage->Data = st_miptree_image_map(pipe, - intelImage->mt, - intelImage->face, - intelImage->level, + stImage->mt, + stImage->face, + stImage->level, &dstRowStride, - intelImage->base.ImageOffsets); + stImage->base.ImageOffsets); } else { /* Allocate regular memory and store the image there temporarily. */ @@ -881,8 +881,8 @@ st_TexImage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, unpack); - if (intelImage->mt) { - st_miptree_image_unmap(pipe, intelImage->mt); + if (stImage->mt) { + st_miptree_image_unmap(pipe, stImage->mt); texImage->Data = NULL; } @@ -1468,9 +1468,9 @@ st_CopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. */ static void -intel_calculate_first_last_level(struct st_texture_object *intelObj) +calculate_first_last_level(struct st_texture_object *stObj) { - struct gl_texture_object *tObj = &intelObj->base; + struct gl_texture_object *tObj = &stObj->base; const struct gl_texture_image *const baseImage = tObj->Image[0][tObj->BaseLevel]; @@ -1512,21 +1512,21 @@ intel_calculate_first_last_level(struct st_texture_object *intelObj) } /* save these values */ - intelObj->firstLevel = firstLevel; - intelObj->lastLevel = lastLevel; + stObj->firstLevel = firstLevel; + stObj->lastLevel = lastLevel; } static void copy_image_data_to_tree(struct pipe_context *pipe, - struct st_texture_object *intelObj, + struct st_texture_object *stObj, struct st_texture_image *stImage) { if (stImage->mt) { /* Copy potentially with the blitter: */ st_miptree_image_copy(pipe, - intelObj->mt, + stObj->mt, stImage->face, stImage->level, stImage->mt); @@ -1538,7 +1538,7 @@ copy_image_data_to_tree(struct pipe_context *pipe, /* More straightforward upload. */ st_miptree_image_data(pipe, - intelObj->mt, + stObj->mt, stImage->face, stImage->level, stImage->base.Data, @@ -1549,19 +1549,19 @@ copy_image_data_to_tree(struct pipe_context *pipe, stImage->base.Data = NULL; } - st_miptree_reference(&stImage->mt, intelObj->mt); + st_miptree_reference(&stImage->mt, stObj->mt); } /* */ -GLuint +GLboolean st_finalize_mipmap_tree(GLcontext *ctx, struct pipe_context *pipe, GLuint unit, GLboolean *needFlush) { struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; - struct st_texture_object *intelObj = st_texture_object(tObj); + struct st_texture_object *stObj = st_texture_object(tObj); int comp_byte = 0; int cpp; @@ -1573,42 +1573,42 @@ st_finalize_mipmap_tree(GLcontext *ctx, /* We know/require this is true by now: */ - assert(intelObj->base._Complete); + assert(stObj->base._Complete); /* What levels must the tree include at a minimum? */ - intel_calculate_first_last_level(intelObj); + calculate_first_last_level(stObj); firstImage = - st_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); + st_texture_image(stObj->base.Image[0][stObj->firstLevel]); /* Fallback case: */ if (firstImage->base.Border) { - if (intelObj->mt) { - st_miptree_release(pipe, &intelObj->mt); + if (stObj->mt) { + st_miptree_release(pipe, &stObj->mt); } return GL_FALSE; } - /* If both firstImage and intelObj have a tree which can contain + /* If both firstImage and stObj have a tree which can contain * all active images, favour firstImage. Note that because of the * completeness requirement, we know that the image dimensions * will match. */ if (firstImage->mt && - firstImage->mt != intelObj->mt && - firstImage->mt->first_level <= intelObj->firstLevel && - firstImage->mt->last_level >= intelObj->lastLevel) { + firstImage->mt != stObj->mt && + firstImage->mt->first_level <= stObj->firstLevel && + firstImage->mt->last_level >= stObj->lastLevel) { - if (intelObj->mt) - st_miptree_release(pipe, &intelObj->mt); + if (stObj->mt) + st_miptree_release(pipe, &stObj->mt); - st_miptree_reference(&intelObj->mt, firstImage->mt); + st_miptree_reference(&stObj->mt, firstImage->mt); } if (firstImage->base.IsCompressed) { - comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); + comp_byte = compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); cpp = comp_byte; } else cpp = firstImage->base.TexFormat->TexelBytes; @@ -1622,28 +1622,28 @@ st_finalize_mipmap_tree(GLcontext *ctx, * programming minLod, maxLod, baseLevel into the hardware and * leaving the tree alone. */ - if (intelObj->mt && - (intelObj->mt->target != intelObj->base.Target || - intelObj->mt->internal_format != firstImage->base.InternalFormat || - intelObj->mt->first_level != intelObj->firstLevel || - intelObj->mt->last_level != intelObj->lastLevel || - intelObj->mt->width0 != firstImage->base.Width || - intelObj->mt->height0 != firstImage->base.Height || - intelObj->mt->depth0 != firstImage->base.Depth || - intelObj->mt->cpp != cpp || - intelObj->mt->compressed != firstImage->base.IsCompressed)) { - st_miptree_release(pipe, &intelObj->mt); + if (stObj->mt && + (stObj->mt->target != stObj->base.Target || + stObj->mt->internal_format != firstImage->base.InternalFormat || + stObj->mt->first_level != stObj->firstLevel || + stObj->mt->last_level != stObj->lastLevel || + stObj->mt->width0 != firstImage->base.Width || + stObj->mt->height0 != firstImage->base.Height || + stObj->mt->depth0 != firstImage->base.Depth || + stObj->mt->cpp != cpp || + stObj->mt->compressed != firstImage->base.IsCompressed)) { + st_miptree_release(pipe, &stObj->mt); } /* May need to create a new tree: */ - if (!intelObj->mt) { - intelObj->mt = st_miptree_create(pipe, - intelObj->base.Target, + if (!stObj->mt) { + stObj->mt = st_miptree_create(pipe, + stObj->base.Target, firstImage->base.InternalFormat, - intelObj->firstLevel, - intelObj->lastLevel, + stObj->firstLevel, + stObj->lastLevel, firstImage->base.Width, firstImage->base.Height, firstImage->base.Depth, @@ -1653,16 +1653,16 @@ st_finalize_mipmap_tree(GLcontext *ctx, /* Pull in any images not in the object's tree: */ - nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + for (i = stObj->firstLevel; i <= stObj->lastLevel; i++) { struct st_texture_image *stImage = - st_texture_image(intelObj->base.Image[face][i]); + st_texture_image(stObj->base.Image[face][i]); /* Need to import images in main memory or held in other trees. */ - if (intelObj->mt != stImage->mt) { - copy_image_data_to_tree(pipe, intelObj, stImage); + if (stObj->mt != stImage->mt) { + copy_image_data_to_tree(pipe, stObj, stImage); *needFlush = GL_TRUE; } } @@ -1680,17 +1680,17 @@ st_finalize_mipmap_tree(GLcontext *ctx, #if 0 /* unused? */ void st_tex_map_images(struct pipe_context *pipe, - struct st_texture_object *intelObj) + struct st_texture_object *stObj) { - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; GLuint face, i; DBG("%s\n", __FUNCTION__); for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + for (i = stObj->firstLevel; i <= stObj->lastLevel; i++) { struct st_texture_image *stImage = - st_texture_image(intelObj->base.Image[face][i]); + st_texture_image(stObj->base.Image[face][i]); if (stImage->mt) { stImage->base.Data = @@ -1712,15 +1712,15 @@ st_tex_map_images(struct pipe_context *pipe, void st_tex_unmap_images(struct pipe_context *pipe, - struct st_texture_object *intelObj) + struct st_texture_object *stObj) { - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; GLuint face, i; for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + for (i = stObj->firstLevel; i <= stObj->lastLevel; i++) { struct st_texture_image *stImage = - st_texture_image(intelObj->base.Image[face][i]); + st_texture_image(stObj->base.Image[face][i]); if (stImage->mt) { st_miptree_image_unmap(pipe, stImage->mt); diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h index dc68aa3d97..7a1867dc58 100644 --- a/src/mesa/state_tracker/st_cb_texture.h +++ b/src/mesa/state_tracker/st_cb_texture.h @@ -6,7 +6,7 @@ extern struct pipe_mipmap_tree * st_get_texobj_mipmap_tree(struct gl_texture_object *texObj); -extern GLuint +extern GLboolean st_finalize_mipmap_tree(GLcontext *ctx, struct pipe_context *pipe, GLuint unit, GLboolean *needFlush); -- cgit v1.2.3 From 6cdff693a7feb51492451f4a04f3736842d2268b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:18:08 -0600 Subject: added pipe_mipmap_tree::format --- src/mesa/pipe/p_state.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 68616cf767..2dcd2db868 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -299,6 +299,7 @@ struct pipe_mipmap_tree GLenum target; /* XXX convert to PIPE_TEXTURE_x */ GLenum internal_format; /* XXX convert to PIPE_FORMAT_x */ + GLuint format; /**< PIPE_FORMAT_x */ GLuint first_level; GLuint last_level; -- cgit v1.2.3 From b245840b86cf877c9b8d666edf229364a84f1dea Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:18:54 -0600 Subject: set mt->format --- src/mesa/state_tracker/st_cb_texture.c | 81 +++++++++++++++++++++------------- 1 file changed, 50 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index b37e0cf0af..6b361b0a4b 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -41,6 +41,7 @@ #include "state_tracker/st_mipmap_tree.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" #define DBG if (0) printf @@ -112,6 +113,21 @@ st_get_texobj_mipmap_tree(struct gl_texture_object *texObj) } +static GLuint +mesa_format_to_pipe_format(GLuint mesaFormat) +{ + switch (mesaFormat) { + /* fix this */ + case MESA_FORMAT_ARGB8888_REV: + case MESA_FORMAT_ARGB8888: + return PIPE_FORMAT_U_A8_R8_G8_B8; + default: + assert(0); + return 0; + } +} + + static int compressed_num_bytes(GLuint mesaFormat) { @@ -517,6 +533,9 @@ guess_and_alloc_mipmap_tree(struct pipe_context *pipe, stImage->base.TexFormat->TexelBytes, comp_byte); + stObj->mt->format + = mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat); + DBG("%s - success\n", __FUNCTION__); } @@ -1144,11 +1163,10 @@ st_TexSubImage3D(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - st_TexSubimage(ctx, 3, - target, level, - xoffset, yoffset, zoffset, - width, height, depth, - format, type, pixels, packing, texObj, texImage); + st_TexSubimage(ctx, 3, target, level, + xoffset, yoffset, zoffset, + width, height, depth, + format, type, pixels, packing, texObj, texImage); } @@ -1165,11 +1183,10 @@ st_TexSubImage2D(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - st_TexSubimage(ctx, 2, - target, level, - xoffset, yoffset, 0, - width, height, 1, - format, type, pixels, packing, texObj, texImage); + st_TexSubimage(ctx, 2, target, level, + xoffset, yoffset, 0, + width, height, 1, + format, type, pixels, packing, texObj, texImage); } @@ -1185,11 +1202,10 @@ st_TexSubImage1D(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - st_TexSubimage(ctx, 1, - target, level, - xoffset, 0, 0, - width, 1, 1, - format, type, pixels, packing, texObj, texImage); + st_TexSubimage(ctx, 1, target, level, + xoffset, 0, 0, + width, 1, 1, + format, type, pixels, packing, texObj, texImage); } @@ -1640,15 +1656,18 @@ st_finalize_mipmap_tree(GLcontext *ctx, */ if (!stObj->mt) { stObj->mt = st_miptree_create(pipe, - stObj->base.Target, - firstImage->base.InternalFormat, - stObj->firstLevel, - stObj->lastLevel, - firstImage->base.Width, - firstImage->base.Height, - firstImage->base.Depth, - cpp, - comp_byte); + stObj->base.Target, + firstImage->base.InternalFormat, + stObj->firstLevel, + stObj->lastLevel, + firstImage->base.Width, + firstImage->base.Height, + firstImage->base.Depth, + cpp, + comp_byte); + + stObj->mt->format + = mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat); } /* Pull in any images not in the object's tree: @@ -1693,13 +1712,13 @@ st_tex_map_images(struct pipe_context *pipe, st_texture_image(stObj->base.Image[face][i]); if (stImage->mt) { - stImage->base.Data = - st_miptree_image_map(pipe, - stImage->mt, - stImage->face, - stImage->level, - &stImage->base.RowStride, - stImage->base.ImageOffsets); + stImage->base.Data + = st_miptree_image_map(pipe, + stImage->mt, + stImage->face, + stImage->level, + &stImage->base.RowStride, + stImage->base.ImageOffsets); /* convert stride to texels, not bytes */ stImage->base.RowStride /= stImage->mt->cpp; /* stImage->base.ImageStride /= stImage->mt->cpp; */ -- cgit v1.2.3 From b4784862b4d03727d1d50408837961c776c53481 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:19:06 -0600 Subject: debug info --- src/mesa/state_tracker/st_atom_texture.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index f82c33e572..eeaf68b001 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -53,6 +53,11 @@ update_textures(struct st_context *st) = st->ctx->Texture.Unit[u]._Current; if (texObj) { struct pipe_mipmap_tree *mt; + GLboolean flush, retval; + + retval = st_finalize_mipmap_tree(st->ctx, st->pipe, u, &flush); + printf("finalize_mipmap_tree returned %d, flush = %d\n", + retval, flush); mt = st_get_texobj_mipmap_tree(texObj); -- cgit v1.2.3 From 204d35c0efa2e1e96cc318185acfe01c4a8edfbb Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:19:34 -0600 Subject: remove dst!=src assertion --- src/mesa/pipe/softpipe/sp_region.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index a0ced3eb86..58749492ec 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -185,7 +185,6 @@ sp_region_copy(struct pipe_context *pipe, GLuint src_offset, GLuint srcx, GLuint srcy, GLuint width, GLuint height) { - assert( dst != src ); assert( dst->cpp == src->cpp ); _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, -- cgit v1.2.3 From 11c557d3cab41e15a5b03715feffc7f920e4b661 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:19:49 -0600 Subject: map/unmap textures --- src/mesa/pipe/softpipe/sp_context.c | 17 +++++++++++++++++ 1 file changed, 17 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 264f3d6e58..ac5651f9a7 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -91,6 +91,14 @@ static void map_surfaces(struct softpipe_context *sp) pipe->region_map(pipe, sps->surface.region); } + /* textures */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + struct pipe_mipmap_tree *mt = sp->texture[i]; + if (mt) { + pipe->region_map(pipe, mt->region); + } + } + /* XXX depth & stencil bufs */ } @@ -109,6 +117,15 @@ static void unmap_surfaces(struct softpipe_context *sp) struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.zbuf); pipe->region_unmap(pipe, sps->surface.region); } + + /* textures */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + struct pipe_mipmap_tree *mt = sp->texture[i]; + if (mt) { + pipe->region_unmap(pipe, mt->region); + } + } + /* XXX depth & stencil bufs */ } -- cgit v1.2.3 From 85675db0fc7fcc9151f47ab7a5ca8643569d2d1d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:20:30 -0600 Subject: added a get_tile() func --- src/mesa/pipe/softpipe/sp_surface.c | 26 +++++++++++++++++++++++++- 1 file changed, 25 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 6512f1d789..87bb64af1a 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -30,6 +30,7 @@ #include "sp_surface.h" #include "pipe/p_defines.h" #include "main/imports.h" +#include "main/macros.h" /** @@ -315,6 +316,23 @@ s8_write_quad_stencil(struct softpipe_surface *sps, +static void +a8r8g8b8_get_tile(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) +{ + const GLuint *src + = ((const GLuint *) ps->region->map) + y * ps->region->pitch + x; + assert(w == 1); + assert(h == 1); + p[0] = UBYTE_TO_FLOAT((src[0] >> 16) & 0xff); + p[1] = UBYTE_TO_FLOAT((src[0] >> 8) & 0xff); + p[2] = UBYTE_TO_FLOAT((src[0] >> 0) & 0xff); + p[3] = UBYTE_TO_FLOAT((src[0] >> 24) & 0xff); +} + + + + static void init_quad_funcs(struct softpipe_surface *sps) { @@ -337,8 +355,14 @@ init_quad_funcs(struct softpipe_surface *sps) sps->read_quad_stencil = s8_read_quad_stencil; sps->write_quad_stencil = s8_write_quad_stencil; break; + case PIPE_FORMAT_U_A8_R8_G8_B8: + sps->surface.get_tile = a8r8g8b8_get_tile; + break; default: + /* assert(0); + */ + ; } } @@ -386,7 +410,7 @@ softpipe_get_tex_surface(struct pipe_context *pipe, assert(zslice == 0); } - ps = pipe->surface_alloc(pipe, mt->internal_format); + ps = pipe->surface_alloc(pipe, mt->format); if (ps) { assert(ps->format); assert(ps->refcount); -- cgit v1.2.3 From 283ed438b6357f6152d7ee002380e049724ffed4 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:20:55 -0600 Subject: include p_state.h --- src/mesa/pipe/softpipe/sp_surface.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 90086c327f..837e0c4ba1 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -33,7 +33,9 @@ #include "glheader.h" #include "sp_headers.h" +#include "pipe/p_state.h" +struct pipe_context; struct softpipe_surface; struct softpipe_context; -- cgit v1.2.3 From 0dc4eea64f56cc93e5359372b08b99a2d600273c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:22:03 -0600 Subject: initial texture filtering code --- src/mesa/pipe/softpipe/sp_tex_sample.c | 166 +++++++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_tex_sample.h | 13 +++ 2 files changed, 179 insertions(+) create mode 100644 src/mesa/pipe/softpipe/sp_tex_sample.c create mode 100644 src/mesa/pipe/softpipe/sp_tex_sample.h (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c new file mode 100644 index 0000000000..e501cea5ad --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -0,0 +1,166 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Texture sampling + * + * Authors: + * Brian Paul + */ + + +#include "main/macros.h" +#include "sp_surface.h" +#include "sp_surface.h" +#include "sp_tex_sample.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/tgsi/core/tgsi_exec.h" + + +#define LERP(T, A, B) ( (A) + (T) * ((B) - (A)) ) + + +/** + * Do 2D/biliner interpolation of float values. + * v00, v10, v01 and v11 are typically four texture samples in a square/box. + * a and b are the horizontal and vertical interpolants. + * It's important that this function is inlined when compiled with + * optimization! If we find that's not true on some systems, convert + * to a macro. + */ +static INLINE GLfloat +lerp_2d(GLfloat a, GLfloat b, + GLfloat v00, GLfloat v10, GLfloat v01, GLfloat v11) +{ + const GLfloat temp0 = LERP(a, v00, v10); + const GLfloat temp1 = LERP(a, v01, v11); + return LERP(b, temp0, temp1); +} + + +/** + * Apply texture coord wrapping mode and return integer texture + * row/column/img index. + */ +static INLINE void +nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint width, GLint *k) +{ + /* XXX this arithmetic isn't exactly right, add more wrap modes */ + switch (wrapMode) { + case PIPE_TEX_WRAP_REPEAT: + *k = (int) (s * width) % width; + break; + case PIPE_TEX_WRAP_CLAMP: + *k = (int) CLAMP(s, 0.0, 1.0) * width; + break; + default: + assert(0); + *k = 0; + } +} + +static INLINE void +linear_texcoord(GLuint wrapMode, GLfloat s, GLuint width, GLint *k0, GLint *k1, + GLfloat *a) +{ + /* XXX this arithmetic isn't exactly right, add more wrap modes */ + switch (wrapMode) { + case PIPE_TEX_WRAP_REPEAT: + *k0 = (int) (s * width) % width; + *k1 = (*k0 + 1) % width; + break; + case PIPE_TEX_WRAP_CLAMP: + *k0 = (int) CLAMP(s, 0.0, 1.0) * width; + *k1 = (*k0 + 1); + if (*k1 >= width) + *k1 = width - 1; + break; + default: + assert(0); + *k0 = *k1 = 0; + } + /* kludge the lerp weight: */ + *a = (s * width) - (int) (s * width); +} + + +/** + * Called via tgsi_sampler::get_sample() + * Use the sampler's state setting to get a filtered RGBA value + * from the sampler's texture (mipmap tree). + * + * XXX we can implement many versions of this function, each + * tightly coded for a specific combination of sampler state + * (nearest + repeat), (bilinear mipmap + clamp), etc. + * + * The update_samplers() function in st_atom_sampler.c could create + * a new tgsi_sampler object for each state combo it finds.... + */ +void +sp_get_sample(struct tgsi_sampler *sampler, + const GLfloat strq[4], GLfloat rgba[4]) +{ + struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; + struct pipe_surface *ps + = pipe->get_tex_surface(pipe, sampler->texture, 0, 0, 0); + + switch (sampler->state->min_filter) { + case PIPE_TEX_FILTER_NEAREST: + { + GLint x, y; + nearest_texcoord(sampler->state->wrap_s, strq[0], + sampler->texture->width0, &x); + nearest_texcoord(sampler->state->wrap_t, strq[1], + sampler->texture->height0, &y); + ps->get_tile(ps, x, y, 1, 1, rgba); + } + break; + case PIPE_TEX_FILTER_LINEAR: + { + GLfloat t00[4], t01[4], t10[4], t11[4]; + GLint x0, y0, x1, y1; + GLfloat a, b; + linear_texcoord(sampler->state->wrap_s, strq[0], + sampler->texture->width0, &x0, &x1, &a); + linear_texcoord(sampler->state->wrap_t, strq[1], + sampler->texture->height0, &y0, &y1, &b); + ps->get_tile(ps, x0, y0, 1, 1, t00); + ps->get_tile(ps, x1, y0, 1, 1, t10); + ps->get_tile(ps, x0, y1, 1, 1, t01); + ps->get_tile(ps, x1, y1, 1, 1, t11); + + rgba[0] = lerp_2d(a, b, t00[0], t10[0], t01[0], t11[0]); + rgba[1] = lerp_2d(a, b, t00[1], t10[1], t01[1], t11[1]); + rgba[2] = lerp_2d(a, b, t00[2], t10[2], t01[2], t11[2]); + rgba[3] = lerp_2d(a, b, t00[3], t10[3], t01[3], t11[3]); + } + break; + default: + assert(0); + } +} diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.h b/src/mesa/pipe/softpipe/sp_tex_sample.h new file mode 100644 index 0000000000..d1d40e4a2f --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tex_sample.h @@ -0,0 +1,13 @@ +#ifndef SP_TEX_SAMPLE_H +#define SP_TEX_SAMPLE_H + + +struct tgsi_sampler; + + +extern void +sp_get_sample(struct tgsi_sampler *sampler, + const GLfloat strq[4], GLfloat rgba[4]); + + +#endif /* SP_TEX_SAMPLE_H */ -- cgit v1.2.3 From e4b5265487c3d912ea84511064681e02f3e2a5ba Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:22:16 -0600 Subject: use new tex filtering code --- src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_quad_fs.c | 24 ++++++++---------------- 2 files changed, 9 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 3e815aa467..feb167405a 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -30,6 +30,7 @@ DRIVER_SOURCES = \ sp_state_setup.c \ sp_state_surface.c \ sp_tex_layout.c \ + sp_tex_sample.c \ sp_surface.c C_SOURCES = \ diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 8eedb2ecc8..e55bfd9725 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -37,6 +37,7 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" +#include "sp_tex_sample.h" #include "tgsi/core/tgsi_core.h" #if 0 @@ -129,24 +130,13 @@ static INLINE void pinterp( struct exec_machine *exec, } -static void -get_sample(const struct tgsi_sampler_state *sampler, - const GLfloat strq[4], GLfloat rgba[4]) -{ - rgba[0] = 1; - rgba[1] = 1; - rgba[2] = 0; - rgba[3] = 0; -} - - /* This should be done by the fragment shader execution unit (code * generated from the decl instructions). Do it here for now. */ static void shade_quad( struct quad_stage *qs, struct quad_header *quad ) { - const struct softpipe_context *softpipe = qs->softpipe; + struct softpipe_context *softpipe = qs->softpipe; struct exec_machine exec; const GLfloat fx = quad->x0; const GLfloat fy = quad->y0; @@ -206,7 +196,7 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) struct tgsi_exec_machine machine; struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; struct tgsi_exec_vector *aoutputs; - struct tgsi_sampler_state samplers[8]; + struct tgsi_sampler samplers[8]; GLuint i; #if !ALIGNED_ATTRIBS @@ -221,7 +211,8 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) #if 11 /* temp sampler setup */ samplers[0].state = &softpipe->sampler[0]; samplers[0].texture = softpipe->texture[0]; - samplers[0].get_sample = get_sample; + samplers[0].get_sample = sp_get_sample; + samplers[0].pipe = &softpipe->pipe; #endif /* init machine state */ @@ -251,10 +242,11 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) /* load input registers */ for (i = 0; i < softpipe->nr_attrs; i++) { -#if 0 +#if 01 /* Make sure fp_attr_to_slot[] is an identity transform. */ + /* assert( softpipe->fp_attr_to_slot[i] == i ); - + */ memcpy( &ainputs[i], exec.attr[i], -- cgit v1.2.3 From ddd30d8160dc7db5f24a5ac823fd1c64c43b50be Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 18:22:40 -0600 Subject: clean up tgsi_sampler-related bits --- src/mesa/pipe/tgsi/core/tgsi_exec.c | 8 ++++---- src/mesa/pipe/tgsi/core/tgsi_exec.h | 23 ++++++----------------- 2 files changed, 10 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 71dfbc10f3..87dab9956f 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -65,7 +65,7 @@ tgsi_exec_machine_init( struct tgsi_exec_machine *mach, struct tgsi_token *tokens, GLuint numSamplers, - const struct tgsi_sampler_state *samplers) + const struct tgsi_sampler *samplers) { GLuint i, k; struct tgsi_parse_context parse; @@ -1040,7 +1040,7 @@ exec_kil (struct tgsi_exec_machine *mach, */ static void fetch_texel_1d( GLcontext *ctx, - const struct tgsi_sampler_state *sampler, + const struct tgsi_sampler *sampler, const union tgsi_exec_channel *s, GLuint unit, union tgsi_exec_channel *r, @@ -1097,7 +1097,7 @@ fetch_texel_1d( GLcontext *ctx, */ static void fetch_texel_2d( GLcontext *ctx, - const struct tgsi_sampler_state *sampler, + const struct tgsi_sampler *sampler, const union tgsi_exec_channel *s, const union tgsi_exec_channel *t, GLuint unit, @@ -1166,7 +1166,7 @@ fetch_texel_2d( GLcontext *ctx, */ static void fetch_texel_3d( GLcontext *ctx, - const struct tgsi_sampler_state *sampler, + const struct tgsi_sampler *sampler, const union tgsi_exec_channel *s, const union tgsi_exec_channel *t, const union tgsi_exec_channel *p, diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index b92d1b5918..1c714ad553 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -21,24 +21,13 @@ struct tgsi_exec_vector union tgsi_exec_channel xyzw[4]; }; -struct tgsi_sampler_state +struct tgsi_sampler { const struct pipe_sampler_state *state; - const struct pipe_mipmap_tree *texture; - void (*get_sample)(const struct tgsi_sampler_state *sampler, + struct pipe_mipmap_tree *texture; + void (*get_sample)(struct tgsi_sampler *sampler, const GLfloat strq[4], GLfloat rgba[4]); - -#if 0 - GLboolean NeedLambda; - GLboolean NeedLodBias; /* if NeedLambda */ - GLboolean NeedLambdaClamp; /* if NeedLambda */ - GLfloat LodBias; /* if NeedLodBias */ - GLfloat MinLod; /* if NeedLambdaClamp */ - GLfloat MaxLod; /* if NeedLambdaClamp */ - GLfloat ImageWidth; - GLfloat ImageHeight; - GLfloat ImageDepth; -#endif + void *pipe; /*XXX temporary*/ }; struct tgsi_exec_labels @@ -124,7 +113,7 @@ struct tgsi_exec_machine struct tgsi_exec_vector *Temps; struct tgsi_exec_vector *Addrs; - const struct tgsi_sampler_state *Samplers; + struct tgsi_sampler *Samplers; GLfloat Imms[256][4]; GLuint ImmLimit; @@ -147,7 +136,7 @@ tgsi_exec_machine_init( struct tgsi_exec_machine *mach, struct tgsi_token *tokens, GLuint numSamplers, - const struct tgsi_sampler_state *samplers); + const struct tgsi_sampler *samplers); void tgsi_exec_prepare( -- cgit v1.2.3 From 01c35b52a1fccbe4253a96b9c0506ef6c991e44f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 09:28:57 +0100 Subject: Checkpoint of a super-simplified i915 driver in the pipe model. Doesn't yet compile, but when it does, it will only draw gouraud tris and even those will be as simple as possible. Needs some 'winsys' support also before that can happen. --- src/mesa/pipe/i915simple/Makefile | 31 + src/mesa/pipe/i915simple/i915_batch.h | 44 ++ src/mesa/pipe/i915simple/i915_blit.c | 157 ++++ src/mesa/pipe/i915simple/i915_blit.h | 54 ++ src/mesa/pipe/i915simple/i915_buffer.c | 121 ++++ src/mesa/pipe/i915simple/i915_clear.c | 55 ++ src/mesa/pipe/i915simple/i915_context.c | 180 +++++ src/mesa/pipe/i915simple/i915_context.h | 122 ++++ src/mesa/pipe/i915simple/i915_debug.c | 337 +++++++++ src/mesa/pipe/i915simple/i915_debug.h | 39 + src/mesa/pipe/i915simple/i915_prim_emit.c | 201 +++++ src/mesa/pipe/i915simple/i915_reg.h | 965 +++++++++++++++++++++++++ src/mesa/pipe/i915simple/i915_regions.c | 304 ++++++++ src/mesa/pipe/i915simple/i915_regions.h | 141 ++++ src/mesa/pipe/i915simple/i915_state.c | 238 ++++++ src/mesa/pipe/i915simple/i915_state.h | 40 + src/mesa/pipe/i915simple/i915_state_derived.c | 186 +++++ src/mesa/pipe/i915simple/i915_state_draw.c | 37 + src/mesa/pipe/i915simple/i915_state_emit.c | 241 ++++++ src/mesa/pipe/i915simple/i915_state_fragprog.c | 67 ++ src/mesa/pipe/i915simple/i915_state_vertex.c | 145 ++++ src/mesa/pipe/i915simple/i915_surface.c | 52 ++ src/mesa/pipe/i915simple/i915_winsys.h | 128 ++++ 23 files changed, 3885 insertions(+) create mode 100644 src/mesa/pipe/i915simple/Makefile create mode 100644 src/mesa/pipe/i915simple/i915_batch.h create mode 100644 src/mesa/pipe/i915simple/i915_blit.c create mode 100644 src/mesa/pipe/i915simple/i915_blit.h create mode 100644 src/mesa/pipe/i915simple/i915_buffer.c create mode 100644 src/mesa/pipe/i915simple/i915_clear.c create mode 100644 src/mesa/pipe/i915simple/i915_context.c create mode 100644 src/mesa/pipe/i915simple/i915_context.h create mode 100644 src/mesa/pipe/i915simple/i915_debug.c create mode 100644 src/mesa/pipe/i915simple/i915_debug.h create mode 100644 src/mesa/pipe/i915simple/i915_prim_emit.c create mode 100644 src/mesa/pipe/i915simple/i915_reg.h create mode 100644 src/mesa/pipe/i915simple/i915_regions.c create mode 100644 src/mesa/pipe/i915simple/i915_regions.h create mode 100644 src/mesa/pipe/i915simple/i915_state.c create mode 100644 src/mesa/pipe/i915simple/i915_state.h create mode 100644 src/mesa/pipe/i915simple/i915_state_derived.c create mode 100644 src/mesa/pipe/i915simple/i915_state_draw.c create mode 100644 src/mesa/pipe/i915simple/i915_state_emit.c create mode 100644 src/mesa/pipe/i915simple/i915_state_fragprog.c create mode 100644 src/mesa/pipe/i915simple/i915_state_vertex.c create mode 100644 src/mesa/pipe/i915simple/i915_surface.c create mode 100644 src/mesa/pipe/i915simple/i915_winsys.h (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile new file mode 100644 index 0000000000..4170349893 --- /dev/null +++ b/src/mesa/pipe/i915simple/Makefile @@ -0,0 +1,31 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = i915simple + +DRIVER_SOURCES = \ + i915_blit.c \ + i915_buffer.c \ + i915_clear.c \ + i915_context.c \ + i915_context.c \ + i915_debug.c \ + i915_regions.c \ + i915_state.c \ + i915_state_derived.c \ + i915_state_emit.c \ + i915_state_fragprog.c \ + i915_prim_emit.c \ + i915_surface.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../Makefile.template + +symlinks: + diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h new file mode 100644 index 0000000000..77008eee14 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_batch.h @@ -0,0 +1,44 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_BATCH_H +#define I915_BATCH_H + +#define BATCH_LOCALS + +#define BEGIN_BATCH( dwords, relocs ) \ + i915->winsys->batch_start( i915->winsys, dwords, relocs ) + +#define OUT_BATCH( dword ) \ + i915->winsys->batch_dword( i915->winsys, dword ) + +#define OUT_RELOC( buf, flags, delta ) \ + i915->winsys->batch_reloc( i915->winsys, buf, flags, delta ) + +#define ADVANCE_BATCH() + +#endif diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/mesa/pipe/i915simple/i915_blit.c new file mode 100644 index 0000000000..c4eafa71d4 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_blit.c @@ -0,0 +1,157 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include + +#include "mtypes.h" + +#include "i915_context.h" +#include "i915_winsys.h" +#include "i915_blit.h" +#include "i915_reg.h" +#include "i915_batch.h" + + +void +i915_fill_blit(struct i915_context *i915, + GLuint cpp, + GLshort dst_pitch, + struct pipe_buffer_handle *dst_buffer, + GLuint dst_offset, + GLshort x, GLshort y, + GLshort w, GLshort h, + GLuint color) +{ + GLuint BR13, CMD; + BATCH_LOCALS; + + dst_pitch *= cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); + CMD = XY_COLOR_BLT_CMD; + break; + case 4: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); + CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | + XY_COLOR_BLT_WRITE_RGB); + break; + default: + return; + } + +// DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", +// __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); + + + BEGIN_BATCH(6, 1); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((y << 16) | x); + OUT_BATCH(((y + h) << 16) | (x + w)); + OUT_RELOC( dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); + OUT_BATCH(color); + ADVANCE_BATCH(); +} + + +void +i915_copy_blit( struct i915_context *i915, + GLuint cpp, + GLshort src_pitch, + struct pipe_buffer_handle *src_buffer, + GLuint src_offset, + GLshort dst_pitch, + struct pipe_buffer_handle *dst_buffer, + GLuint dst_offset, + GLshort src_x, GLshort src_y, + GLshort dst_x, GLshort dst_y, + GLshort w, GLshort h ) +{ + GLuint CMD, BR13; + int dst_y2 = dst_y + h; + int dst_x2 = dst_x + w; + BATCH_LOCALS; + + + printf("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + src_buffer, src_pitch, src_offset, src_x, src_y, + dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); + + src_pitch *= cpp; + dst_pitch *= cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = (((GLint) dst_pitch) & 0xffff) | + (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + break; + case 4: + BR13 = + (((GLint) dst_pitch) & 0xffff) | + (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = + (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + break; + default: + return; + } + + if (dst_y2 < dst_y || + dst_x2 < dst_x) { + return; + } + + /* Hardware can handle negative pitches but loses the ability to do + * proper overlapping blits in that case. We don't really have a + * need for either at this stage. + */ + assert (dst_pitch > 0 && src_pitch > 0); + + + BEGIN_BATCH(8, 2); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((dst_y << 16) | dst_x); + OUT_BATCH((dst_y2 << 16) | dst_x2); + OUT_RELOC(dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); + OUT_BATCH((src_y << 16) | src_x); + OUT_BATCH(((GLint) src_pitch & 0xffff)); + OUT_RELOC(src_buffer, I915_BUFFER_ACCESS_READ, src_offset); + ADVANCE_BATCH(); +} + + diff --git a/src/mesa/pipe/i915simple/i915_blit.h b/src/mesa/pipe/i915simple/i915_blit.h new file mode 100644 index 0000000000..b1131a005c --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_blit.h @@ -0,0 +1,54 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_BLIT_H +#define I915_BLIT_H + +#include "i915_context.h" + +extern void i915_copy_blit(struct i915_context *i915, + GLuint cpp, + GLshort src_pitch, + struct pipe_buffer_handle *src_buffer, + GLuint src_offset, + GLshort dst_pitch, + struct pipe_buffer_handle *dst_buffer, + GLuint dst_offset, + GLshort srcx, GLshort srcy, + GLshort dstx, GLshort dsty, + GLshort w, GLshort h ); + +extern void i915_fill_blit(struct i915_context *i915, + GLuint cpp, + GLshort dst_pitch, + struct pipe_buffer_handle *dst_buffer, + GLuint dst_offset, + GLshort x, GLshort y, + GLshort w, GLshort h, GLuint color); + + +#endif diff --git a/src/mesa/pipe/i915simple/i915_buffer.c b/src/mesa/pipe/i915simple/i915_buffer.c new file mode 100644 index 0000000000..eaaf3c1beb --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_buffer.c @@ -0,0 +1,121 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include "i915_context.h" +#include "i915_winsys.h" + + + +/* Most callbacks map direcly onto winsys operations at the moment, + * but this may change, especially as state_trackers and winsys's + * evolve in separate directions... Don't try and remove this yet. + */ +static struct pipe_buffer_handle * +i915_create_buffer(struct pipe_context *pipe, + unsigned alignment, + unsigned flags) +{ + struct i915_context *i915 = i915_context( pipe ); + return i915->winsys->create_buffer( i915->winsys, alignment ); +} + +static void *i915_buffer_map(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned flags ) +{ + struct i915_context *i915 = i915_context( pipe ); + return i915->winsys->buffer_map( i915->winsys, buf ); +} + +static void i915_buffer_unmap(struct pipe_context *pipe, + struct pipe_buffer_handle *buf) +{ + struct i915_context *i915 = i915_context( pipe ); + i915->winsys->buffer_unmap( i915->winsys, buf ); +} + +static struct pipe_buffer_handle * +i915_buffer_reference(struct pipe_context *pipe, + struct pipe_buffer_handle *buf) +{ + struct i915_context *i915 = i915_context( pipe ); + return i915->winsys->buffer_reference( i915->winsys, buf ); +} + +static void i915_buffer_unreference(struct pipe_context *pipe, + struct pipe_buffer_handle **buf) +{ + struct i915_context *i915 = i915_context( pipe ); + i915->winsys->buffer_unreference( i915->winsys, buf ); +} + +static void i915_buffer_data(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct i915_context *i915 = i915_context( pipe ); + i915->winsys->buffer_data( i915->winsys, buf, size, data ); +} + +static void i915_buffer_subdata(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) +{ + struct i915_context *i915 = i915_context( pipe ); + i915->winsys->buffer_subdata( i915->winsys, buf, offset, size, data ); +} + +static void i915_buffer_get_subdata(struct pipe_context *pipe, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) +{ + struct i915_context *i915 = i915_context( pipe ); + i915->winsys->buffer_get_subdata( i915->winsys, buf, offset, size, data ); +} + + +void +i915_init_buffer_functions( struct i915_context *i915 ) +{ + i915->pipe.create_buffer = i915_create_buffer; + i915->pipe.buffer_map = i915_buffer_map; + i915->pipe.buffer_unmap = i915_buffer_unmap; + i915->pipe.buffer_reference = i915_buffer_reference; + i915->pipe.buffer_unreference = i915_buffer_unreference; + i915->pipe.buffer_data = i915_buffer_data; + i915->pipe.buffer_subdata = i915_buffer_subdata; + i915->pipe.buffer_get_subdata = i915_buffer_get_subdata; +} diff --git a/src/mesa/pipe/i915simple/i915_clear.c b/src/mesa/pipe/i915simple/i915_clear.c new file mode 100644 index 0000000000..11aa55b64b --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_clear.c @@ -0,0 +1,55 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + */ + + +#include "pipe/p_defines.h" +#include "i915_context.h" +#include "i915_state.h" +#include "colormac.h" + + +/** + * Clear the given surface to the specified value. + * No masking, no scissor (clear entire buffer). + */ +void +i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, + GLuint clearValue) +{ + GLint x, y, w, h; + + x = 0; + y = 0; + w = ps->width; + h = ps->height; + + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue); +} diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c new file mode 100644 index 0000000000..1a21e931ae --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -0,0 +1,180 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "main/imports.h" /* CALLOC */ +#include "i915_context.h" +#include "i915_winsys.h" +#include "i915_state.h" + +#include "pipe/draw/draw_context.h" +#include "pipe/p_defines.h" + + + +/** + * Return list of supported surface/texture formats. + * If we find texture and drawable support differs, add a selector + * parameter or another function. + */ +static const GLuint * +i915_supported_formats(struct pipe_context *pipe, +// GLuint type, + GLuint *numFormats) +{ +#if 0 + static const GLuint tex_supported[] = { + PIPE_FORMAT_U_R8_G8_B8_A8, + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_U_L8, + PIPE_FORMAT_U_A8, + PIPE_FORMAT_U_I8, + PIPE_FORMAT_U_L8_A8, + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + PIPE_FORMAT_S8_Z24, + }; + + + /* Actually a lot more than this - add later: + */ + static const GLuint render_supported[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + }; + + /* + */ + static const GLuint z_stencil_supported[] = { + PIPE_FORMAT_U_Z16, + PIPE_FORMAT_U_Z32, + PIPE_FORMAT_S8_Z24, + }; + + switch (type) { + case PIPE_RENDER_FORMAT: + *numFormats = Elements(render_supported); + return render_supported; + + case PIPE_TEX_FORMAT: + *numFormats = Elements(tex_supported); + return render_supported; + + case PIPE_Z_STENCIL_FORMAT: + *numFormats = Elements(render_supported); + return render_supported; + + default: + *numFormats = 0; + return NULL; + } +#else + static const GLuint render_supported[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_S8_Z24, + }; + *numFormats = 2; + return render_supported; +#endif +} + + + +static void i915_destroy( struct pipe_context *pipe ) +{ + struct i915_context *i915 = i915_context( pipe ); + + draw_destroy( i915->draw ); + + free( i915 ); +} + +static void i915_draw_vb( struct pipe_context *pipe, + struct vertex_buffer *VB ) +{ + struct i915_context *i915 = i915_context( pipe ); + +// if (i915->dirty) +// i915_update_derived( i915 ); + + draw_vb( i915->draw, VB ); +} + + +static void +i915_draw_vertices(struct pipe_context *pipe, + GLuint mode, + GLuint numVertex, const GLfloat *verts, + GLuint numAttribs, const GLuint attribs[]) +{ + struct i915_context *i915 = i915_context( pipe ); + + if (i915->dirty) + i915_update_derived( i915 ); + + draw_vertices(i915->draw, mode, numVertex, verts, numAttribs, attribs); +} + + + + +struct pipe_context *i915_create( struct i915_winsys *winsys ) +{ + struct i915_context *i915 = CALLOC_STRUCT(i915_context); + + i915->pipe.destroy = i915_destroy; + + i915->pipe.supported_formats = i915_supported_formats; + + i915->pipe.draw_vb = i915_draw_vb; + i915->pipe.draw_vertices = i915_draw_vertices; + i915->pipe.clear = i915_clear; + i915->pipe.reset_occlusion_counter = NULL; /* no support */ + i915->pipe.get_occlusion_counter = NULL; + + i915->winsys = winsys; + + /* + * Create drawing context and plug our rendering stage into it. + */ + i915->draw = draw_create(); + assert(i915->draw); + draw_set_setup_stage(i915->draw, i915_draw_render_stage(i915)); + + i915_init_buffer_functions(i915); + i915_init_region_functions(i915); + i915_init_surface_functions(i915); + + /* + * XXX we could plug GL selection/feedback into the drawing pipeline + * by specifying a different setup/render stage. + */ + + return &i915->pipe; +} + diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h new file mode 100644 index 0000000000..060997bfa4 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -0,0 +1,122 @@ + /************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_CONTEXT_H +#define I915_CONTEXT_H + + +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +struct i915_context +{ + struct pipe_context pipe; + struct i915_winsys *winsys; + struct draw_context *draw; + + /* The most recent drawing state as set by the driver: + */ + struct pipe_alpha_test_state alpha_test; + struct pipe_blend_state blend; + struct pipe_blend_color blend_color; + struct pipe_clear_color_state clear_color; + struct pipe_clip_state clip; + struct pipe_depth_state depth_test; + struct pipe_framebuffer_state framebuffer; + struct pipe_fs_state fs; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; + struct pipe_setup_state setup; + struct pipe_stencil_state stencil; + struct pipe_texture_object *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + GLuint dirty; + + struct pipe_scissor_state cliprect; + +}; + +#define I915_NEW_VIEWPORT 0x1 +#define I915_NEW_SETUP 0x2 +#define I915_NEW_FS 0x4 +#define I915_NEW_BLEND 0x8 +#define I915_NEW_CLIP 0x10 +#define I915_NEW_SCISSOR 0x20 +#define I915_NEW_STIPPLE 0x40 +#define I915_NEW_FRAMEBUFFER 0x80 +#define I915_NEW_ALPHA_TEST 0x100 +#define I915_NEW_DEPTH_TEST 0x200 +#define I915_NEW_SAMPLER 0x400 +#define I915_NEW_TEXTURE 0x800 +#define I915_NEW_STENCIL 0x1000 + + +/*********************************************************************** + * i915_prim_emit.c: + */ +struct draw_stage *i915_draw_render_stage( struct i915_context *i915 ); + + +/*********************************************************************** + * i915_state_emit.c: + */ +void i915_emit_hardware_state(struct i915_context *i915 ); +unsigned *i915_passthrough_program( unsigned *dwords ); + + + +/*********************************************************************** + * i915_clear.c: + */ +void i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, + GLuint clearValue); + + +/*********************************************************************** + * i915_buffer.c: + */ +void i915_init_buffer_functions( struct i915_context *i915 ); +void i915_init_region_functions( struct i915_context *i915 ); +void i915_init_surface_functions( struct i915_context *i915 ); +void i915_init_state_functions( struct i915_context *i915 ); + + + +/*********************************************************************** + * Inline conversion functions. These are better-typed than the + * macros used previously: + */ +static INLINE struct i915_context * +i915_context( struct pipe_context *pipe ) +{ + return (struct i915_context *)pipe; +} + + + +#endif diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c new file mode 100644 index 0000000000..8d80590396 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -0,0 +1,337 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include +#include + +#include "i915_reg.h" +#include "i915_debug.h" + + + +static const char *opcodes[0x20] = { + "NOP", + "ADD", + "MOV", + "MUL", + "MAD", + "DP2ADD", + "DP3", + "DP4", + "FRC", + "RCP", + "RSQ", + "EXP", + "LOG", + "CMP", + "MIN", + "MAX", + "FLR", + "MOD", + "TRC", + "SGE", + "SLT", + "TEXLD", + "TEXLDP", + "TEXLDB", + "TEXKILL", + "DCL", + "0x1a", + "0x1b", + "0x1c", + "0x1d", + "0x1e", + "0x1f", +}; + + +static const int args[0x20] = { + 0, /* 0 nop */ + 2, /* 1 add */ + 1, /* 2 mov */ + 2, /* 3 m ul */ + 3, /* 4 mad */ + 3, /* 5 dp2add */ + 2, /* 6 dp3 */ + 2, /* 7 dp4 */ + 1, /* 8 frc */ + 1, /* 9 rcp */ + 1, /* a rsq */ + 1, /* b exp */ + 1, /* c log */ + 3, /* d cmp */ + 2, /* e min */ + 2, /* f max */ + 1, /* 10 flr */ + 1, /* 11 mod */ + 1, /* 12 trc */ + 2, /* 13 sge */ + 2, /* 14 slt */ + 1, + 1, + 1, + 1, + 0, + 0, + 0, + 0, + 0, + 0, + 0, +}; + + +static const char *regname[0x8] = { + "R", + "T", + "CONST", + "S", + "OC", + "OD", + "U", + "UNKNOWN", +}; + +static void +print_reg_type_nr(unsigned type, unsigned nr) +{ + switch (type) { + case REG_TYPE_T: + switch (nr) { + case T_DIFFUSE: + fprintf(stderr, "T_DIFFUSE"); + return; + case T_SPECULAR: + fprintf(stderr, "T_SPECULAR"); + return; + case T_FOG_W: + fprintf(stderr, "T_FOG_W"); + return; + default: + fprintf(stderr, "T_TEX%d", nr); + return; + } + case REG_TYPE_OC: + if (nr == 0) { + fprintf(stderr, "oC"); + return; + } + break; + case REG_TYPE_OD: + if (nr == 0) { + fprintf(stderr, "oD"); + return; + } + break; + default: + break; + } + + fprintf(stderr, "%s[%d]", regname[type], nr); +} + +#define REG_SWIZZLE_MASK 0x7777 +#define REG_NEGATE_MASK 0x8888 + +#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ + (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ + (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ + (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) + + +static void +print_reg_neg_swizzle(unsigned reg) +{ + int i; + + if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && + (reg & REG_NEGATE_MASK) == 0) + return; + + fprintf(stderr, "."); + + for (i = 3; i >= 0; i--) { + if (reg & (1 << ((i * 4) + 3))) + fprintf(stderr, "-"); + + switch ((reg >> (i * 4)) & 0x7) { + case 0: + fprintf(stderr, "x"); + break; + case 1: + fprintf(stderr, "y"); + break; + case 2: + fprintf(stderr, "z"); + break; + case 3: + fprintf(stderr, "w"); + break; + case 4: + fprintf(stderr, "0"); + break; + case 5: + fprintf(stderr, "1"); + break; + default: + fprintf(stderr, "?"); + break; + } + } +} + + +static void +print_src_reg(unsigned dword) +{ + unsigned nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; + unsigned type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(type, nr); + print_reg_neg_swizzle(dword); +} + +void +i915_print_ureg(const char *msg, unsigned ureg) +{ + fprintf(stderr, "%s: ", msg); + print_src_reg(ureg >> 8); + fprintf(stderr, "\n"); +} + +static void +print_dest_reg(unsigned dword) +{ + unsigned nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; + unsigned type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(type, nr); + if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) + return; + fprintf(stderr, "."); + if (dword & A0_DEST_CHANNEL_X) + fprintf(stderr, "x"); + if (dword & A0_DEST_CHANNEL_Y) + fprintf(stderr, "y"); + if (dword & A0_DEST_CHANNEL_Z) + fprintf(stderr, "z"); + if (dword & A0_DEST_CHANNEL_W) + fprintf(stderr, "w"); +} + + +#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) +#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) +#define GET_SRC2_REG(r) (r) + + +static void +print_arith_op(unsigned opcode, const unsigned * program) +{ + if (opcode != A0_NOP) { + print_dest_reg(program[0]); + if (program[0] & A0_DEST_SATURATE) + fprintf(stderr, " = SATURATE "); + else + fprintf(stderr, " = "); + } + + fprintf(stderr, "%s ", opcodes[opcode]); + + print_src_reg(GET_SRC0_REG(program[0], program[1])); + if (args[opcode] == 1) { + fprintf(stderr, "\n"); + return; + } + + fprintf(stderr, ", "); + print_src_reg(GET_SRC1_REG(program[1], program[2])); + if (args[opcode] == 2) { + fprintf(stderr, "\n"); + return; + } + + fprintf(stderr, ", "); + print_src_reg(GET_SRC2_REG(program[2])); + fprintf(stderr, "\n"); + return; +} + + +static void +print_tex_op(unsigned opcode, const unsigned * program) +{ + print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); + fprintf(stderr, " = "); + + fprintf(stderr, "%s ", opcodes[opcode]); + + fprintf(stderr, "S[%d],", program[0] & T0_SAMPLER_NR_MASK); + + print_reg_type_nr((program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & + REG_TYPE_MASK, + (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); + fprintf(stderr, "\n"); +} + +static void +print_dcl_op(unsigned opcode, const unsigned * program) +{ + fprintf(stderr, "%s ", opcodes[opcode]); + print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); + fprintf(stderr, "\n"); +} + + +void +i915_disassemble_program(const unsigned * program, unsigned sz) +{ + unsigned size = program[0] & 0x1ff; + unsigned i; + + fprintf(stderr, "BEGIN\n"); + + if (size + 2 != sz) { + fprintf(stderr, "%s: program size mismatch %d/%d\n", __FUNCTION__, + size + 2, sz); + assert(0); + } + + program++; + for (i = 1; i < sz; i += 3, program += 3) { + unsigned opcode = program[0] & (0x1f << 24); + + if ((unsigned) opcode >= A0_NOP && opcode <= A0_SLT) + print_arith_op(opcode >> 24, program); + else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) + print_tex_op(opcode >> 24, program); + else if (opcode == D0_DCL) + print_dcl_op(opcode >> 24, program); + else + fprintf(stderr, "Unknown opcode 0x%x\n", opcode); + } + + fprintf(stderr, "END\n\n"); +} diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h new file mode 100644 index 0000000000..47a02401ee --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_debug.h @@ -0,0 +1,39 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef I915_DEBUG_H +#define I915_DEBUG_H + + +extern void i915_disassemble_program(const unsigned *program, unsigned sz); +extern void i915_print_ureg(const char *msg, unsigned ureg); + + +#endif diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c new file mode 100644 index 0000000000..97e337451e --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -0,0 +1,201 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "imports.h" +#include "macros.h" + +#include "pipe/draw/draw_private.h" + +#include "i915_context.h" +#include "i915_winsys.h" +#include "i915_reg.h" +#include "i915_state.h" + + + +/** + * Primitive emit to hardware. No support for vertex buffers or any + * nice fast paths. + */ +struct setup_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct i915_context *i915; +}; + + + +/** + * Basically a cast wrapper. + */ +static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) +{ + return (struct setup_stage *)stage; +} + +static inline unsigned pack_ub4( unsigned char b0, + unsigned char b1, + unsigned char b2, + unsigned char b3 ) +{ + return ((((unsigned int)b0) << 0) | + (((unsigned int)b1) << 8) | + (((unsigned int)b2) << 16) | + (((unsigned int)b3) << 24)); +} + +static inline unsigned fui( float f ) +{ + union { + float f; + unsigned ui; + } fi; + + fi.f = f; + return fi.ui; +} + +static inline unsigned char float_to_ubyte( float f ) +{ + unsigned char ub; + UNCLAMPED_FLOAT_TO_UBYTE(ub, f); + return ub; +} + + +/* Hardcoded vertex format: xyz/rgba + */ +static inline void +emit_hw_vertex( unsigned *ptr, + struct vertex_header *vertex ) +{ + ptr[0] = fui( vertex->data[0][0] ); + ptr[1] = fui( vertex->data[0][1] ); + ptr[2] = fui( vertex->data[0][2] ); + + ptr[3] = pack_ub4( float_to_ubyte( vertex->data[1][0] ), + float_to_ubyte( vertex->data[1][1] ), + float_to_ubyte( vertex->data[1][2] ), + float_to_ubyte( vertex->data[1][3] ) ); +} + + + + +static inline void +emit_prim( struct draw_stage *stage, + struct prim_header *prim, + unsigned hwprim, + unsigned nr ) +{ + struct i915_context *i915 = setup_stage(stage)->i915; + struct i915_winsys *winsys = i915->winsys; + unsigned vertex_size = 4 * sizeof(int); + unsigned *ptr; + unsigned i; + + i915_emit_hardware_state( i915 ); + + ptr = winsys->batch_start( winsys, nr * vertex_size, 0 ); + if (ptr == 0) { + winsys->batch_flush( winsys ); + ptr = winsys->batch_start( winsys, nr * vertex_size, 0 ); + if (ptr == 0) { + assert(0); + return; + } + } + + /* Emit each triangle as a single primitive. I told you this was + * simple. + */ + *ptr++ = (_3DPRIMITIVE | + hwprim | + ((4 + vertex_size * nr)/4 - 2)); + + for (i = 0; i < nr; i++) { + emit_hw_vertex(ptr, prim->v[i]); + ptr += vertex_size / sizeof(int); + } +} + + +static void +setup_tri( struct draw_stage *stage, struct prim_header *prim ) +{ + emit_prim( stage, prim, PRIM3D_TRILIST, 3 ); +} + + +static void +setup_line(struct draw_stage *stage, struct prim_header *prim) +{ + emit_prim( stage, prim, PRIM3D_LINELIST, 2 ); +} + + +static void +setup_point(struct draw_stage *stage, struct prim_header *prim) +{ + emit_prim( stage, prim, PRIM3D_POINTLIST, 1 ); +} + + + +static void setup_begin( struct draw_stage *stage ) +{ +} + + +static void setup_end( struct draw_stage *stage ) +{ +} + +static void reset_stipple_counter( struct draw_stage *stage ) +{ +} + + +/** + * Create a new primitive setup/render stage. + */ +struct draw_stage *i915_draw_render_stage( struct i915_context *i915 ) +{ + struct setup_stage *setup = CALLOC_STRUCT(setup_stage); + + setup->i915 = i915; + setup->stage.draw = i915->draw; + setup->stage.begin = setup_begin; + setup->stage.point = setup_point; + setup->stage.line = setup_line; + setup->stage.tri = setup_tri; + setup->stage.end = setup_end; + setup->stage.reset_stipple_counter = reset_stipple_counter; + + return &setup->stage; +} diff --git a/src/mesa/pipe/i915simple/i915_reg.h b/src/mesa/pipe/i915simple/i915_reg.h new file mode 100644 index 0000000000..0776b75ef9 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_reg.h @@ -0,0 +1,965 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef I915_REG_H +#define I915_REG_H + + +#define I915_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value) + +#define CMD_3D (0x3<<29) + +#define PRIM3D_INLINE (CMD_3D | (0x1f<<24)) +#define PRIM3D_TRILIST (0x0<<18) +#define PRIM3D_TRISTRIP (0x1<<18) +#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) +#define PRIM3D_TRIFAN (0x3<<18) +#define PRIM3D_POLY (0x4<<18) +#define PRIM3D_LINELIST (0x5<<18) +#define PRIM3D_LINESTRIP (0x6<<18) +#define PRIM3D_RECTLIST (0x7<<18) +#define PRIM3D_POINTLIST (0x8<<18) +#define PRIM3D_DIB (0x9<<18) +#define PRIM3D_CLEAR_RECT (0xa<<18) +#define PRIM3D_ZONE_INIT (0xd<<18) +#define PRIM3D_MASK (0x1f<<18) + +/* p137 */ +#define _3DSTATE_AA_CMD (CMD_3D | (0x06<<24)) +#define AA_LINE_ECAAR_WIDTH_ENABLE (1<<16) +#define AA_LINE_ECAAR_WIDTH_0_5 0 +#define AA_LINE_ECAAR_WIDTH_1_0 (1<<14) +#define AA_LINE_ECAAR_WIDTH_2_0 (2<<14) +#define AA_LINE_ECAAR_WIDTH_4_0 (3<<14) +#define AA_LINE_REGION_WIDTH_ENABLE (1<<8) +#define AA_LINE_REGION_WIDTH_0_5 0 +#define AA_LINE_REGION_WIDTH_1_0 (1<<6) +#define AA_LINE_REGION_WIDTH_2_0 (2<<6) +#define AA_LINE_REGION_WIDTH_4_0 (3<<6) + +/* 3DSTATE_BACKFACE_STENCIL_OPS, p138*/ +#define _3DSTATE_BACKFACE_STENCIL_OPS (CMD_3D | (0x8<<24)) +#define BFO_ENABLE_STENCIL_REF (1<<23) +#define BFO_STENCIL_REF_SHIFT 15 +#define BFO_STENCIL_REF_MASK (0xff<<15) +#define BFO_ENABLE_STENCIL_FUNCS (1<<14) +#define BFO_STENCIL_TEST_SHIFT 11 +#define BFO_STENCIL_TEST_MASK (0x7<<11) +#define BFO_STENCIL_FAIL_SHIFT 8 +#define BFO_STENCIL_FAIL_MASK (0x7<<8) +#define BFO_STENCIL_PASS_Z_FAIL_SHIFT 5 +#define BFO_STENCIL_PASS_Z_FAIL_MASK (0x7<<5) +#define BFO_STENCIL_PASS_Z_PASS_SHIFT 2 +#define BFO_STENCIL_PASS_Z_PASS_MASK (0x7<<2) +#define BFO_ENABLE_STENCIL_TWO_SIDE (1<<1) +#define BFO_STENCIL_TWO_SIDE (1<<0) + + +/* 3DSTATE_BACKFACE_STENCIL_MASKS, p140 */ +#define _3DSTATE_BACKFACE_STENCIL_MASKS (CMD_3D | (0x9<<24)) +#define BFM_ENABLE_STENCIL_TEST_MASK (1<<17) +#define BFM_ENABLE_STENCIL_WRITE_MASK (1<<16) +#define BFM_STENCIL_TEST_MASK_SHIFT 8 +#define BFM_STENCIL_TEST_MASK_MASK (0xff<<8) +#define BFM_STENCIL_WRITE_MASK_SHIFT 0 +#define BFM_STENCIL_WRITE_MASK_MASK (0xff<<0) + + + +/* 3DSTATE_BIN_CONTROL p141 */ + +/* p143 */ +#define _3DSTATE_BUF_INFO_CMD (CMD_3D | (0x1d<<24) | (0x8e<<16) | 1) +/* Dword 1 */ +#define BUF_3D_ID_COLOR_BACK (0x3<<24) +#define BUF_3D_ID_DEPTH (0x7<<24) +#define BUF_3D_USE_FENCE (1<<23) +#define BUF_3D_TILED_SURFACE (1<<22) +#define BUF_3D_TILE_WALK_X 0 +#define BUF_3D_TILE_WALK_Y (1<<21) +#define BUF_3D_PITCH(x) (((x)/4)<<2) +/* Dword 2 */ +#define BUF_3D_ADDR(x) ((x) & ~0x3) + + +/* 3DSTATE_CHROMA_KEY */ + +/* 3DSTATE_CLEAR_PARAMETERS, p150 */ +#define _3DSTATE_CLEAR_PARAMETERS (CMD_3D | (0x1d<<24) | (0x9c<<16) | 5) +/* Dword 1 */ +#define CLEARPARAM_CLEAR_RECT (1 << 16) +#define CLEARPARAM_ZONE_INIT (0 << 16) +#define CLEARPARAM_WRITE_COLOR (1 << 2) +#define CLEARPARAM_WRITE_DEPTH (1 << 1) +#define CLEARPARAM_WRITE_STENCIL (1 << 0) + +/* 3DSTATE_CONSTANT_BLEND_COLOR, p153 */ +#define _3DSTATE_CONST_BLEND_COLOR_CMD (CMD_3D | (0x1d<<24) | (0x88<<16)) + + + +/* 3DSTATE_COORD_SET_BINDINGS, p154 */ +#define _3DSTATE_COORD_SET_BINDINGS (CMD_3D | (0x16<<24)) +#define CSB_TCB(iunit, eunit) ((eunit)<<(iunit*3)) + +/* p156 */ +#define _3DSTATE_DFLT_DIFFUSE_CMD (CMD_3D | (0x1d<<24) | (0x99<<16)) + +/* p157 */ +#define _3DSTATE_DFLT_SPEC_CMD (CMD_3D | (0x1d<<24) | (0x9a<<16)) + +/* p158 */ +#define _3DSTATE_DFLT_Z_CMD (CMD_3D | (0x1d<<24) | (0x98<<16)) + + +/* 3DSTATE_DEPTH_OFFSET_SCALE, p159 */ +#define _3DSTATE_DEPTH_OFFSET_SCALE (CMD_3D | (0x1d<<24) | (0x97<<16)) +/* scale in dword 1 */ + + +/* 3DSTATE_DEPTH_SUBRECT_DISABLE, p160 */ +#define _3DSTATE_DEPTH_SUBRECT_DISABLE (CMD_3D | (0x1c<<24) | (0x11<<19) | 0x2) + +/* p161 */ +#define _3DSTATE_DST_BUF_VARS_CMD (CMD_3D | (0x1d<<24) | (0x85<<16)) +/* Dword 1 */ +#define TEX_DEFAULT_COLOR_OGL (0<<30) +#define TEX_DEFAULT_COLOR_D3D (1<<30) +#define ZR_EARLY_DEPTH (1<<29) +#define LOD_PRECLAMP_OGL (1<<28) +#define LOD_PRECLAMP_D3D (0<<28) +#define DITHER_FULL_ALWAYS (0<<26) +#define DITHER_FULL_ON_FB_BLEND (1<<26) +#define DITHER_CLAMPED_ALWAYS (2<<26) +#define LINEAR_GAMMA_BLEND_32BPP (1<<25) +#define DEBUG_DISABLE_ENH_DITHER (1<<24) +#define DSTORG_HORT_BIAS(x) ((x)<<20) +#define DSTORG_VERT_BIAS(x) ((x)<<16) +#define COLOR_4_2_2_CHNL_WRT_ALL 0 +#define COLOR_4_2_2_CHNL_WRT_Y (1<<12) +#define COLOR_4_2_2_CHNL_WRT_CR (2<<12) +#define COLOR_4_2_2_CHNL_WRT_CB (3<<12) +#define COLOR_4_2_2_CHNL_WRT_CRCB (4<<12) +#define COLOR_BUF_8BIT 0 +#define COLOR_BUF_RGB555 (1<<8) +#define COLOR_BUF_RGB565 (2<<8) +#define COLOR_BUF_ARGB8888 (3<<8) +#define DEPTH_FRMT_16_FIXED 0 +#define DEPTH_FRMT_16_FLOAT (1<<2) +#define DEPTH_FRMT_24_FIXED_8_OTHER (2<<2) +#define VERT_LINE_STRIDE_1 (1<<1) +#define VERT_LINE_STRIDE_0 (0<<1) +#define VERT_LINE_STRIDE_OFS_1 1 +#define VERT_LINE_STRIDE_OFS_0 0 + +/* p166 */ +#define _3DSTATE_DRAW_RECT_CMD (CMD_3D|(0x1d<<24)|(0x80<<16)|3) +/* Dword 1 */ +#define DRAW_RECT_DIS_DEPTH_OFS (1<<30) +#define DRAW_DITHER_OFS_X(x) ((x)<<26) +#define DRAW_DITHER_OFS_Y(x) ((x)<<24) +/* Dword 2 */ +#define DRAW_YMIN(x) ((x)<<16) +#define DRAW_XMIN(x) (x) +/* Dword 3 */ +#define DRAW_YMAX(x) ((x)<<16) +#define DRAW_XMAX(x) (x) +/* Dword 4 */ +#define DRAW_YORG(x) ((x)<<16) +#define DRAW_XORG(x) (x) + + +/* 3DSTATE_FILTER_COEFFICIENTS_4X4, p170 */ + +/* 3DSTATE_FILTER_COEFFICIENTS_6X5, p172 */ + + +/* _3DSTATE_FOG_COLOR, p173 */ +#define _3DSTATE_FOG_COLOR_CMD (CMD_3D|(0x15<<24)) +#define FOG_COLOR_RED(x) ((x)<<16) +#define FOG_COLOR_GREEN(x) ((x)<<8) +#define FOG_COLOR_BLUE(x) (x) + +/* _3DSTATE_FOG_MODE, p174 */ +#define _3DSTATE_FOG_MODE_CMD (CMD_3D|(0x1d<<24)|(0x89<<16)|2) +/* Dword 1 */ +#define FMC1_FOGFUNC_MODIFY_ENABLE (1<<31) +#define FMC1_FOGFUNC_VERTEX (0<<28) +#define FMC1_FOGFUNC_PIXEL_EXP (1<<28) +#define FMC1_FOGFUNC_PIXEL_EXP2 (2<<28) +#define FMC1_FOGFUNC_PIXEL_LINEAR (3<<28) +#define FMC1_FOGFUNC_MASK (3<<28) +#define FMC1_FOGINDEX_MODIFY_ENABLE (1<<27) +#define FMC1_FOGINDEX_Z (0<<25) +#define FMC1_FOGINDEX_W (1<<25) +#define FMC1_C1_C2_MODIFY_ENABLE (1<<24) +#define FMC1_DENSITY_MODIFY_ENABLE (1<<23) +#define FMC1_C1_ONE (1<<13) +#define FMC1_C1_MASK (0xffff<<4) +/* Dword 2 */ +#define FMC2_C2_ONE (1<<16) +/* Dword 3 */ +#define FMC3_D_ONE (1<<16) + + + +/* _3DSTATE_INDEPENDENT_ALPHA_BLEND, p177 */ +#define _3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD (CMD_3D|(0x0b<<24)) +#define IAB_MODIFY_ENABLE (1<<23) +#define IAB_ENABLE (1<<22) +#define IAB_MODIFY_FUNC (1<<21) +#define IAB_FUNC_SHIFT 16 +#define IAB_MODIFY_SRC_FACTOR (1<<11) +#define IAB_SRC_FACTOR_SHIFT 6 +#define IAB_SRC_FACTOR_MASK (BLENDFACT_MASK<<6) +#define IAB_MODIFY_DST_FACTOR (1<<5) +#define IAB_DST_FACTOR_SHIFT 0 +#define IAB_DST_FACTOR_MASK (BLENDFACT_MASK<<0) + + +#define BLENDFUNC_ADD 0x0 +#define BLENDFUNC_SUBTRACT 0x1 +#define BLENDFUNC_REVERSE_SUBTRACT 0x2 +#define BLENDFUNC_MIN 0x3 +#define BLENDFUNC_MAX 0x4 +#define BLENDFUNC_MASK 0x7 + +/* 3DSTATE_LOAD_INDIRECT, p180 */ + +#define _3DSTATE_LOAD_INDIRECT (CMD_3D|(0x1d<<24)|(0x7<<16)) +#define LI0_STATE_STATIC_INDIRECT (0x01<<8) +#define LI0_STATE_DYNAMIC_INDIRECT (0x02<<8) +#define LI0_STATE_SAMPLER (0x04<<8) +#define LI0_STATE_MAP (0x08<<8) +#define LI0_STATE_PROGRAM (0x10<<8) +#define LI0_STATE_CONSTANTS (0x20<<8) + +#define SIS0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define SIS0_FORCE_LOAD (1<<1) +#define SIS0_BUFFER_VALID (1<<0) +#define SIS1_BUFFER_LENGTH(x) ((x)&0xff) + +#define DIS0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define DIS0_BUFFER_RESET (1<<1) +#define DIS0_BUFFER_VALID (1<<0) + +#define SSB0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define SSB0_FORCE_LOAD (1<<1) +#define SSB0_BUFFER_VALID (1<<0) +#define SSB1_BUFFER_LENGTH(x) ((x)&0xff) + +#define MSB0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define MSB0_FORCE_LOAD (1<<1) +#define MSB0_BUFFER_VALID (1<<0) +#define MSB1_BUFFER_LENGTH(x) ((x)&0xff) + +#define PSP0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define PSP0_FORCE_LOAD (1<<1) +#define PSP0_BUFFER_VALID (1<<0) +#define PSP1_BUFFER_LENGTH(x) ((x)&0xff) + +#define PSC0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define PSC0_FORCE_LOAD (1<<1) +#define PSC0_BUFFER_VALID (1<<0) +#define PSC1_BUFFER_LENGTH(x) ((x)&0xff) + + + + + +/* _3DSTATE_RASTERIZATION_RULES */ +#define _3DSTATE_RASTER_RULES_CMD (CMD_3D|(0x07<<24)) +#define ENABLE_POINT_RASTER_RULE (1<<15) +#define OGL_POINT_RASTER_RULE (1<<13) +#define ENABLE_TEXKILL_3D_4D (1<<10) +#define TEXKILL_3D (0<<9) +#define TEXKILL_4D (1<<9) +#define ENABLE_LINE_STRIP_PROVOKE_VRTX (1<<8) +#define ENABLE_TRI_FAN_PROVOKE_VRTX (1<<5) +#define LINE_STRIP_PROVOKE_VRTX(x) ((x)<<6) +#define TRI_FAN_PROVOKE_VRTX(x) ((x)<<3) + +/* _3DSTATE_SCISSOR_ENABLE, p256 */ +#define _3DSTATE_SCISSOR_ENABLE_CMD (CMD_3D|(0x1c<<24)|(0x10<<19)) +#define ENABLE_SCISSOR_RECT ((1<<1) | 1) +#define DISABLE_SCISSOR_RECT (1<<1) + +/* _3DSTATE_SCISSOR_RECTANGLE_0, p257 */ +#define _3DSTATE_SCISSOR_RECT_0_CMD (CMD_3D|(0x1d<<24)|(0x81<<16)|1) +/* Dword 1 */ +#define SCISSOR_RECT_0_YMIN(x) ((x)<<16) +#define SCISSOR_RECT_0_XMIN(x) (x) +/* Dword 2 */ +#define SCISSOR_RECT_0_YMAX(x) ((x)<<16) +#define SCISSOR_RECT_0_XMAX(x) (x) + +/* p189 */ +#define _3DSTATE_LOAD_STATE_IMMEDIATE_1 ((0x3<<29)|(0x1d<<24)|(0x04<<16)) +#define I1_LOAD_S(n) (1<<(4+n)) + +#define S0_VB_OFFSET_MASK 0xffffffc +#define S0_AUTO_CACHE_INV_DISABLE (1<<0) + +#define S1_VERTEX_WIDTH_SHIFT 24 +#define S1_VERTEX_WIDTH_MASK (0x3f<<24) +#define S1_VERTEX_PITCH_SHIFT 16 +#define S1_VERTEX_PITCH_MASK (0x3f<<16) + +#define TEXCOORDFMT_2D 0x0 +#define TEXCOORDFMT_3D 0x1 +#define TEXCOORDFMT_4D 0x2 +#define TEXCOORDFMT_1D 0x3 +#define TEXCOORDFMT_2D_16 0x4 +#define TEXCOORDFMT_4D_16 0x5 +#define TEXCOORDFMT_NOT_PRESENT 0xf +#define S2_TEXCOORD_FMT0_MASK 0xf +#define S2_TEXCOORD_FMT1_SHIFT 4 +#define S2_TEXCOORD_FMT(unit, type) ((type)<<(unit*4)) +#define S2_TEXCOORD_NONE (~0) + +/* S3 not interesting */ + +#define S4_POINT_WIDTH_SHIFT 23 +#define S4_POINT_WIDTH_MASK (0x1ff<<23) +#define S4_LINE_WIDTH_SHIFT 19 +#define S4_LINE_WIDTH_ONE (0x2<<19) +#define S4_LINE_WIDTH_MASK (0xf<<19) +#define S4_FLATSHADE_ALPHA (1<<18) +#define S4_FLATSHADE_FOG (1<<17) +#define S4_FLATSHADE_SPECULAR (1<<16) +#define S4_FLATSHADE_COLOR (1<<15) +#define S4_CULLMODE_BOTH (0<<13) +#define S4_CULLMODE_NONE (1<<13) +#define S4_CULLMODE_CW (2<<13) +#define S4_CULLMODE_CCW (3<<13) +#define S4_CULLMODE_MASK (3<<13) +#define S4_VFMT_POINT_WIDTH (1<<12) +#define S4_VFMT_SPEC_FOG (1<<11) +#define S4_VFMT_COLOR (1<<10) +#define S4_VFMT_DEPTH_OFFSET (1<<9) +#define S4_VFMT_XYZ (1<<6) +#define S4_VFMT_XYZW (2<<6) +#define S4_VFMT_XY (3<<6) +#define S4_VFMT_XYW (4<<6) +#define S4_VFMT_XYZW_MASK (7<<6) +#define S4_FORCE_DEFAULT_DIFFUSE (1<<5) +#define S4_FORCE_DEFAULT_SPECULAR (1<<4) +#define S4_LOCAL_DEPTH_OFFSET_ENABLE (1<<3) +#define S4_VFMT_FOG_PARAM (1<<2) +#define S4_SPRITE_POINT_ENABLE (1<<1) +#define S4_LINE_ANTIALIAS_ENABLE (1<<0) + +#define S4_VFMT_MASK (S4_VFMT_POINT_WIDTH | \ + S4_VFMT_SPEC_FOG | \ + S4_VFMT_COLOR | \ + S4_VFMT_DEPTH_OFFSET | \ + S4_VFMT_XYZW_MASK | \ + S4_VFMT_FOG_PARAM) + + +#define S5_WRITEDISABLE_ALPHA (1<<31) +#define S5_WRITEDISABLE_RED (1<<30) +#define S5_WRITEDISABLE_GREEN (1<<29) +#define S5_WRITEDISABLE_BLUE (1<<28) +#define S5_WRITEDISABLE_MASK (0xf<<28) +#define S5_FORCE_DEFAULT_POINT_SIZE (1<<27) +#define S5_LAST_PIXEL_ENABLE (1<<26) +#define S5_GLOBAL_DEPTH_OFFSET_ENABLE (1<<25) +#define S5_FOG_ENABLE (1<<24) +#define S5_STENCIL_REF_SHIFT 16 +#define S5_STENCIL_REF_MASK (0xff<<16) +#define S5_STENCIL_TEST_FUNC_SHIFT 13 +#define S5_STENCIL_TEST_FUNC_MASK (0x7<<13) +#define S5_STENCIL_FAIL_SHIFT 10 +#define S5_STENCIL_FAIL_MASK (0x7<<10) +#define S5_STENCIL_PASS_Z_FAIL_SHIFT 7 +#define S5_STENCIL_PASS_Z_FAIL_MASK (0x7<<7) +#define S5_STENCIL_PASS_Z_PASS_SHIFT 4 +#define S5_STENCIL_PASS_Z_PASS_MASK (0x7<<4) +#define S5_STENCIL_WRITE_ENABLE (1<<3) +#define S5_STENCIL_TEST_ENABLE (1<<2) +#define S5_COLOR_DITHER_ENABLE (1<<1) +#define S5_LOGICOP_ENABLE (1<<0) + + +#define S6_ALPHA_TEST_ENABLE (1<<31) +#define S6_ALPHA_TEST_FUNC_SHIFT 28 +#define S6_ALPHA_TEST_FUNC_MASK (0x7<<28) +#define S6_ALPHA_REF_SHIFT 20 +#define S6_ALPHA_REF_MASK (0xff<<20) +#define S6_DEPTH_TEST_ENABLE (1<<19) +#define S6_DEPTH_TEST_FUNC_SHIFT 16 +#define S6_DEPTH_TEST_FUNC_MASK (0x7<<16) +#define S6_CBUF_BLEND_ENABLE (1<<15) +#define S6_CBUF_BLEND_FUNC_SHIFT 12 +#define S6_CBUF_BLEND_FUNC_MASK (0x7<<12) +#define S6_CBUF_SRC_BLEND_FACT_SHIFT 8 +#define S6_CBUF_SRC_BLEND_FACT_MASK (0xf<<8) +#define S6_CBUF_DST_BLEND_FACT_SHIFT 4 +#define S6_CBUF_DST_BLEND_FACT_MASK (0xf<<4) +#define S6_DEPTH_WRITE_ENABLE (1<<3) +#define S6_COLOR_WRITE_ENABLE (1<<2) +#define S6_TRISTRIP_PV_SHIFT 0 +#define S6_TRISTRIP_PV_MASK (0x3<<0) + +#define S7_DEPTH_OFFSET_CONST_MASK ~0 + + + +#define DST_BLND_FACT(f) ((f)<= 0.0) ? src1 : src2 */ +#define A0_MIN (0xe<<24) /* dst = (src0 < src1) ? src0 : src1 */ +#define A0_MAX (0xf<<24) /* dst = (src0 >= src1) ? src0 : src1 */ +#define A0_FLR (0x10<<24) /* dst = floor(src0) */ +#define A0_MOD (0x11<<24) /* dst = src0 fmod 1.0 */ +#define A0_TRC (0x12<<24) /* dst = int(src0) */ +#define A0_SGE (0x13<<24) /* dst = src0 >= src1 ? 1.0 : 0.0 */ +#define A0_SLT (0x14<<24) /* dst = src0 < src1 ? 1.0 : 0.0 */ +#define A0_DEST_SATURATE (1<<22) +#define A0_DEST_TYPE_SHIFT 19 +/* Allow: R, OC, OD, U */ +#define A0_DEST_NR_SHIFT 14 +/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ +#define A0_DEST_CHANNEL_X (1<<10) +#define A0_DEST_CHANNEL_Y (2<<10) +#define A0_DEST_CHANNEL_Z (4<<10) +#define A0_DEST_CHANNEL_W (8<<10) +#define A0_DEST_CHANNEL_ALL (0xf<<10) +#define A0_DEST_CHANNEL_SHIFT 10 +#define A0_SRC0_TYPE_SHIFT 7 +#define A0_SRC0_NR_SHIFT 2 + +#define A0_DEST_CHANNEL_XY (A0_DEST_CHANNEL_X|A0_DEST_CHANNEL_Y) +#define A0_DEST_CHANNEL_XYZ (A0_DEST_CHANNEL_XY|A0_DEST_CHANNEL_Z) + + +#define SRC_X 0 +#define SRC_Y 1 +#define SRC_Z 2 +#define SRC_W 3 +#define SRC_ZERO 4 +#define SRC_ONE 5 + +#define A1_SRC0_CHANNEL_X_NEGATE (1<<31) +#define A1_SRC0_CHANNEL_X_SHIFT 28 +#define A1_SRC0_CHANNEL_Y_NEGATE (1<<27) +#define A1_SRC0_CHANNEL_Y_SHIFT 24 +#define A1_SRC0_CHANNEL_Z_NEGATE (1<<23) +#define A1_SRC0_CHANNEL_Z_SHIFT 20 +#define A1_SRC0_CHANNEL_W_NEGATE (1<<19) +#define A1_SRC0_CHANNEL_W_SHIFT 16 +#define A1_SRC1_TYPE_SHIFT 13 +#define A1_SRC1_NR_SHIFT 8 +#define A1_SRC1_CHANNEL_X_NEGATE (1<<7) +#define A1_SRC1_CHANNEL_X_SHIFT 4 +#define A1_SRC1_CHANNEL_Y_NEGATE (1<<3) +#define A1_SRC1_CHANNEL_Y_SHIFT 0 + +#define A2_SRC1_CHANNEL_Z_NEGATE (1<<31) +#define A2_SRC1_CHANNEL_Z_SHIFT 28 +#define A2_SRC1_CHANNEL_W_NEGATE (1<<27) +#define A2_SRC1_CHANNEL_W_SHIFT 24 +#define A2_SRC2_TYPE_SHIFT 21 +#define A2_SRC2_NR_SHIFT 16 +#define A2_SRC2_CHANNEL_X_NEGATE (1<<15) +#define A2_SRC2_CHANNEL_X_SHIFT 12 +#define A2_SRC2_CHANNEL_Y_NEGATE (1<<11) +#define A2_SRC2_CHANNEL_Y_SHIFT 8 +#define A2_SRC2_CHANNEL_Z_NEGATE (1<<7) +#define A2_SRC2_CHANNEL_Z_SHIFT 4 +#define A2_SRC2_CHANNEL_W_NEGATE (1<<3) +#define A2_SRC2_CHANNEL_W_SHIFT 0 + + + +/* Texture instructions */ +#define T0_TEXLD (0x15<<24) /* Sample texture using predeclared + * sampler and address, and output + * filtered texel data to destination + * register */ +#define T0_TEXLDP (0x16<<24) /* Same as texld but performs a + * perspective divide of the texture + * coordinate .xyz values by .w before + * sampling. */ +#define T0_TEXLDB (0x17<<24) /* Same as texld but biases the + * computed LOD by w. Only S4.6 two's + * comp is used. This implies that a + * float to fixed conversion is + * done. */ +#define T0_TEXKILL (0x18<<24) /* Does not perform a sampling + * operation. Simply kills the pixel + * if any channel of the address + * register is < 0.0. */ +#define T0_DEST_TYPE_SHIFT 19 +/* Allow: R, OC, OD, U */ +/* Note: U (unpreserved) regs do not retain their values between + * phases (cannot be used for feedback) + * + * Note: oC and OD registers can only be used as the destination of a + * texture instruction once per phase (this is an implementation + * restriction). + */ +#define T0_DEST_NR_SHIFT 14 +/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ +#define T0_SAMPLER_NR_SHIFT 0 /* This field ignored for TEXKILL */ +#define T0_SAMPLER_NR_MASK (0xf<<0) + +#define T1_ADDRESS_REG_TYPE_SHIFT 24 /* Reg to use as texture coord */ +/* Allow R, T, OC, OD -- R, OC, OD are 'dependent' reads, new program phase */ +#define T1_ADDRESS_REG_NR_SHIFT 17 +#define T2_MBZ 0 + +/* Declaration instructions */ +#define D0_DCL (0x19<<24) /* Declare a t (interpolated attrib) + * register or an s (sampler) + * register. */ +#define D0_SAMPLE_TYPE_SHIFT 22 +#define D0_SAMPLE_TYPE_2D (0x0<<22) +#define D0_SAMPLE_TYPE_CUBE (0x1<<22) +#define D0_SAMPLE_TYPE_VOLUME (0x2<<22) +#define D0_SAMPLE_TYPE_MASK (0x3<<22) + +#define D0_TYPE_SHIFT 19 +/* Allow: T, S */ +#define D0_NR_SHIFT 14 +/* Allow T: 0..10, S: 0..15 */ +#define D0_CHANNEL_X (1<<10) +#define D0_CHANNEL_Y (2<<10) +#define D0_CHANNEL_Z (4<<10) +#define D0_CHANNEL_W (8<<10) +#define D0_CHANNEL_ALL (0xf<<10) +#define D0_CHANNEL_NONE (0<<10) + +#define D0_CHANNEL_XY (D0_CHANNEL_X|D0_CHANNEL_Y) +#define D0_CHANNEL_XYZ (D0_CHANNEL_XY|D0_CHANNEL_Z) + +/* I915 Errata: Do not allow (xz), (xw), (xzw) combinations for diffuse + * or specular declarations. + * + * For T dcls, only allow: (x), (xy), (xyz), (w), (xyzw) + * + * Must be zero for S (sampler) dcls + */ +#define D1_MBZ 0 +#define D2_MBZ 0 + + + +/* p207 */ +#define _3DSTATE_MAP_STATE (CMD_3D|(0x1d<<24)|(0x0<<16)) + +#define MS1_MAPMASK_SHIFT 0 +#define MS1_MAPMASK_MASK (0x8fff<<0) + +#define MS2_UNTRUSTED_SURFACE (1<<31) +#define MS2_ADDRESS_MASK 0xfffffffc +#define MS2_VERTICAL_LINE_STRIDE (1<<1) +#define MS2_VERTICAL_OFFSET (1<<1) + +#define MS3_HEIGHT_SHIFT 21 +#define MS3_WIDTH_SHIFT 10 +#define MS3_PALETTE_SELECT (1<<9) +#define MS3_MAPSURF_FORMAT_SHIFT 7 +#define MS3_MAPSURF_FORMAT_MASK (0x7<<7) +#define MAPSURF_8BIT (1<<7) +#define MAPSURF_16BIT (2<<7) +#define MAPSURF_32BIT (3<<7) +#define MAPSURF_422 (5<<7) +#define MAPSURF_COMPRESSED (6<<7) +#define MAPSURF_4BIT_INDEXED (7<<7) +#define MS3_MT_FORMAT_MASK (0x7 << 3) +#define MS3_MT_FORMAT_SHIFT 3 +#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ +#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ +#define MT_8BIT_L8 (1<<3) +#define MT_8BIT_A8 (4<<3) +#define MT_8BIT_MONO8 (5<<3) +#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ +#define MT_16BIT_ARGB1555 (1<<3) +#define MT_16BIT_ARGB4444 (2<<3) +#define MT_16BIT_AY88 (3<<3) +#define MT_16BIT_88DVDU (5<<3) +#define MT_16BIT_BUMP_655LDVDU (6<<3) +#define MT_16BIT_I16 (7<<3) +#define MT_16BIT_L16 (8<<3) +#define MT_16BIT_A16 (9<<3) +#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ +#define MT_32BIT_ABGR8888 (1<<3) +#define MT_32BIT_XRGB8888 (2<<3) +#define MT_32BIT_XBGR8888 (3<<3) +#define MT_32BIT_QWVU8888 (4<<3) +#define MT_32BIT_AXVU8888 (5<<3) +#define MT_32BIT_LXVU8888 (6<<3) +#define MT_32BIT_XLVU8888 (7<<3) +#define MT_32BIT_ARGB2101010 (8<<3) +#define MT_32BIT_ABGR2101010 (9<<3) +#define MT_32BIT_AWVU2101010 (0xA<<3) +#define MT_32BIT_GR1616 (0xB<<3) +#define MT_32BIT_VU1616 (0xC<<3) +#define MT_32BIT_xI824 (0xD<<3) +#define MT_32BIT_xA824 (0xE<<3) +#define MT_32BIT_xL824 (0xF<<3) +#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ +#define MT_422_YCRCB_NORMAL (1<<3) +#define MT_422_YCRCB_SWAPUV (2<<3) +#define MT_422_YCRCB_SWAPUVY (3<<3) +#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ +#define MT_COMPRESS_DXT2_3 (1<<3) +#define MT_COMPRESS_DXT4_5 (2<<3) +#define MT_COMPRESS_FXT1 (3<<3) +#define MT_COMPRESS_DXT1_RGB (4<<3) +#define MS3_USE_FENCE_REGS (1<<2) +#define MS3_TILED_SURFACE (1<<1) +#define MS3_TILE_WALK (1<<0) + +#define MS4_PITCH_SHIFT 21 +#define MS4_CUBE_FACE_ENA_NEGX (1<<20) +#define MS4_CUBE_FACE_ENA_POSX (1<<19) +#define MS4_CUBE_FACE_ENA_NEGY (1<<18) +#define MS4_CUBE_FACE_ENA_POSY (1<<17) +#define MS4_CUBE_FACE_ENA_NEGZ (1<<16) +#define MS4_CUBE_FACE_ENA_POSZ (1<<15) +#define MS4_CUBE_FACE_ENA_MASK (0x3f<<15) +#define MS4_MAX_LOD_SHIFT 9 +#define MS4_MAX_LOD_MASK (0x3f<<9) +#define MS4_MIP_LAYOUT_LEGACY (0<<8) +#define MS4_MIP_LAYOUT_BELOW_LPT (0<<8) +#define MS4_MIP_LAYOUT_RIGHT_LPT (1<<8) +#define MS4_VOLUME_DEPTH_SHIFT 0 +#define MS4_VOLUME_DEPTH_MASK (0xff<<0) + +/* p244 */ +#define _3DSTATE_SAMPLER_STATE (CMD_3D|(0x1d<<24)|(0x1<<16)) + +#define SS1_MAPMASK_SHIFT 0 +#define SS1_MAPMASK_MASK (0x8fff<<0) + +#define SS2_REVERSE_GAMMA_ENABLE (1<<31) +#define SS2_PACKED_TO_PLANAR_ENABLE (1<<30) +#define SS2_COLORSPACE_CONVERSION (1<<29) +#define SS2_CHROMAKEY_SHIFT 27 +#define SS2_BASE_MIP_LEVEL_SHIFT 22 +#define SS2_BASE_MIP_LEVEL_MASK (0x1f<<22) +#define SS2_MIP_FILTER_SHIFT 20 +#define SS2_MIP_FILTER_MASK (0x3<<20) +#define MIPFILTER_NONE 0 +#define MIPFILTER_NEAREST 1 +#define MIPFILTER_LINEAR 3 +#define SS2_MAG_FILTER_SHIFT 17 +#define SS2_MAG_FILTER_MASK (0x7<<17) +#define FILTER_NEAREST 0 +#define FILTER_LINEAR 1 +#define FILTER_ANISOTROPIC 2 +#define FILTER_4X4_1 3 +#define FILTER_4X4_2 4 +#define FILTER_4X4_FLAT 5 +#define FILTER_6X5_MONO 6 /* XXX - check */ +#define SS2_MIN_FILTER_SHIFT 14 +#define SS2_MIN_FILTER_MASK (0x7<<14) +#define SS2_LOD_BIAS_SHIFT 5 +#define SS2_LOD_BIAS_ONE (0x10<<5) +#define SS2_LOD_BIAS_MASK (0x1ff<<5) +/* Shadow requires: + * MT_X8{I,L,A}24 or MT_{I,L,A}16 texture format + * FILTER_4X4_x MIN and MAG filters + */ +#define SS2_SHADOW_ENABLE (1<<4) +#define SS2_MAX_ANISO_MASK (1<<3) +#define SS2_MAX_ANISO_2 (0<<3) +#define SS2_MAX_ANISO_4 (1<<3) +#define SS2_SHADOW_FUNC_SHIFT 0 +#define SS2_SHADOW_FUNC_MASK (0x7<<0) +/* SS2_SHADOW_FUNC values: see COMPAREFUNC_* */ + +#define SS3_MIN_LOD_SHIFT 24 +#define SS3_MIN_LOD_ONE (0x10<<24) +#define SS3_MIN_LOD_MASK (0xff<<24) +#define SS3_KILL_PIXEL_ENABLE (1<<17) +#define SS3_TCX_ADDR_MODE_SHIFT 12 +#define SS3_TCX_ADDR_MODE_MASK (0x7<<12) +#define TEXCOORDMODE_WRAP 0 +#define TEXCOORDMODE_MIRROR 1 +#define TEXCOORDMODE_CLAMP_EDGE 2 +#define TEXCOORDMODE_CUBE 3 +#define TEXCOORDMODE_CLAMP_BORDER 4 +#define TEXCOORDMODE_MIRROR_ONCE 5 +#define SS3_TCY_ADDR_MODE_SHIFT 9 +#define SS3_TCY_ADDR_MODE_MASK (0x7<<9) +#define SS3_TCZ_ADDR_MODE_SHIFT 6 +#define SS3_TCZ_ADDR_MODE_MASK (0x7<<6) +#define SS3_NORMALIZED_COORDS (1<<5) +#define SS3_TEXTUREMAP_INDEX_SHIFT 1 +#define SS3_TEXTUREMAP_INDEX_MASK (0xf<<1) +#define SS3_DEINTERLACER_ENABLE (1<<0) + +#define SS4_BORDER_COLOR_MASK (~0) + +/* 3DSTATE_SPAN_STIPPLE, p258 + */ +#define _3DSTATE_STIPPLE ((0x3<<29)|(0x1d<<24)|(0x83<<16)) +#define ST1_ENABLE (1<<16) +#define ST1_MASK (0xffff) + +#define _3DSTATE_DEFAULT_Z ((0x3<<29)|(0x1d<<24)|(0x98<<16)) +#define _3DSTATE_DEFAULT_DIFFUSE ((0x3<<29)|(0x1d<<24)|(0x99<<16)) +#define _3DSTATE_DEFAULT_SPECULAR ((0x3<<29)|(0x1d<<24)|(0x9a<<16)) + + +#define MI_FLUSH ((0<<29)|(4<<23)) +#define FLUSH_MAP_CACHE (1<<0) +#define INHIBIT_FLUSH_RENDER_CACHE (1<<2) + + +#define CMD_3D (0x3<<29) + + +#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) +#define PRIM_INDIRECT (1<<23) +#define PRIM_INLINE (0<<23) +#define PRIM_INDIRECT_SEQUENTIAL (0<<17) +#define PRIM_INDIRECT_ELTS (1<<17) + +#define PRIM3D_TRILIST (0x0<<18) +#define PRIM3D_TRISTRIP (0x1<<18) +#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) +#define PRIM3D_TRIFAN (0x3<<18) +#define PRIM3D_POLY (0x4<<18) +#define PRIM3D_LINELIST (0x5<<18) +#define PRIM3D_LINESTRIP (0x6<<18) +#define PRIM3D_RECTLIST (0x7<<18) +#define PRIM3D_POINTLIST (0x8<<18) +#define PRIM3D_DIB (0x9<<18) +#define PRIM3D_MASK (0x1f<<18) + +#define I915PACKCOLOR4444(r,g,b,a) \ + ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) + +#define I915PACKCOLOR1555(r,g,b,a) \ + ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ + ((a) ? 0x8000 : 0)) + +#define I915PACKCOLOR565(r,g,b) \ + ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) + +#define I915PACKCOLOR8888(r,g,b,a) \ + ((a<<24) | (r<<16) | (g<<8) | b) + + + + +#define BR00_BITBLT_CLIENT 0x40000000 +#define BR00_OP_COLOR_BLT 0x10000000 +#define BR00_OP_SRC_COPY_BLT 0x10C00000 +#define BR13_SOLID_PATTERN 0x80000000 + +#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) +#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) +#define XY_COLOR_BLT_WRITE_RGB (1<<20) + +#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) +#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) +#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) + +#define MI_WAIT_FOR_EVENT ((0x3<<23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + + + +#define COMPAREFUNC_ALWAYS 0 +#define COMPAREFUNC_NEVER 0x1 +#define COMPAREFUNC_LESS 0x2 +#define COMPAREFUNC_EQUAL 0x3 +#define COMPAREFUNC_LEQUAL 0x4 +#define COMPAREFUNC_GREATER 0x5 +#define COMPAREFUNC_NOTEQUAL 0x6 +#define COMPAREFUNC_GEQUAL 0x7 + +#define STENCILOP_KEEP 0 +#define STENCILOP_ZERO 0x1 +#define STENCILOP_REPLACE 0x2 +#define STENCILOP_INCRSAT 0x3 +#define STENCILOP_DECRSAT 0x4 +#define STENCILOP_INCR 0x5 +#define STENCILOP_DECR 0x6 +#define STENCILOP_INVERT 0x7 + +#define LOGICOP_CLEAR 0 +#define LOGICOP_NOR 0x1 +#define LOGICOP_AND_INV 0x2 +#define LOGICOP_COPY_INV 0x3 +#define LOGICOP_AND_RVRSE 0x4 +#define LOGICOP_INV 0x5 +#define LOGICOP_XOR 0x6 +#define LOGICOP_NAND 0x7 +#define LOGICOP_AND 0x8 +#define LOGICOP_EQUIV 0x9 +#define LOGICOP_NOOP 0xa +#define LOGICOP_OR_INV 0xb +#define LOGICOP_COPY 0xc +#define LOGICOP_OR_RVRSE 0xd +#define LOGICOP_OR 0xe +#define LOGICOP_SET 0xf + +#define BLENDFACT_ZERO 0x01 +#define BLENDFACT_ONE 0x02 +#define BLENDFACT_SRC_COLR 0x03 +#define BLENDFACT_INV_SRC_COLR 0x04 +#define BLENDFACT_SRC_ALPHA 0x05 +#define BLENDFACT_INV_SRC_ALPHA 0x06 +#define BLENDFACT_DST_ALPHA 0x07 +#define BLENDFACT_INV_DST_ALPHA 0x08 +#define BLENDFACT_DST_COLR 0x09 +#define BLENDFACT_INV_DST_COLR 0x0a +#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b +#define BLENDFACT_CONST_COLOR 0x0c +#define BLENDFACT_INV_CONST_COLOR 0x0d +#define BLENDFACT_CONST_ALPHA 0x0e +#define BLENDFACT_INV_CONST_ALPHA 0x0f +#define BLENDFACT_MASK 0x0f + + +#endif diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c new file mode 100644 index 0000000000..52181cf9f4 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -0,0 +1,304 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Provide additional functionality on top of bufmgr buffers: + * - 2d semantics and blit operations (XXX: remove/simplify blits??) + * - refcounting of buffers for multiple images in a buffer. + * - refcounting of buffer mappings. + */ + +#include "i915_context.h" +#include "i915_winsys.h" +#include "i915_blit.h" + + +static void +i915_region_idle(struct pipe_context *pipe, struct pipe_region *region) +{ + +} + + +static GLubyte * +i915_region_map(struct pipe_context *pipe, struct pipe_region *region) +{ + struct i915_context *i915 = i915_context( pipe ); + + if (!region->map_refcount++) { + region->map = i915->winsys->buffer_map( i915->winsys, + region->buffer ); + } + + return region->map; +} + +static void +i915_region_unmap(struct pipe_context *pipe, struct pipe_region *region) +{ + struct i915_context *i915 = i915_context( pipe ); + + if (!--region->map_refcount) { + i915->winsys->buffer_unmap( i915->winsys, + region->buffer ); + region->map = NULL; + } +} + +static struct pipe_region * +i915_region_alloc(struct pipe_context *pipe, + GLuint cpp, GLuint width, GLuint height) +{ + struct i915_context *i915 = i915_context( pipe ); + struct pipe_region *region = calloc(sizeof(*region), 1); + + /* Choose a pitch to match hardware requirements - requires 64 byte + * alignment of render targets. + * + * XXX: is this ok for textures?? + * clearly want to be able to render to textures under some + * circumstances, but maybe not always a requirement. + */ + unsigned pitch = ((cpp * width + 63) & ~63) / cpp; + + region->cpp = cpp; + region->pitch = pitch; + region->height = height; /* needed? */ + region->refcount = 1; + + region->buffer = i915->winsys->create_buffer( i915->winsys, 64 ); + + i915->winsys->buffer_data( i915->winsys, + region->buffer, + pitch * cpp * height, + NULL ); + + return region; +} + +static void +i915_region_release(struct pipe_context *pipe, struct pipe_region **region) +{ + struct i915_context *i915 = i915_context( pipe ); + + if (!*region) + return; + + ASSERT((*region)->refcount > 0); + (*region)->refcount--; + + if ((*region)->refcount == 0) { + assert((*region)->map_refcount == 0); + + i915->winsys->buffer_unreference( i915->winsys, + (*region)->buffer ); + free(*region); + } + *region = NULL; +} + + +/* + * XXX Move this into core Mesa? + */ +static void +_mesa_copy_rect(GLubyte * dst, + GLuint cpp, + GLuint dst_pitch, + GLuint dst_x, + GLuint dst_y, + GLuint width, + GLuint height, + const GLubyte * src, + GLuint src_pitch, + GLuint src_x, + GLuint src_y) +{ + GLuint i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * dst_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } +} + + +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +static void +i915_region_data(struct pipe_context *pipe, + struct pipe_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + const void *src, GLuint src_pitch, + GLuint srcx, GLuint srcy, GLuint width, GLuint height) +{ + _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + pipe->region_unmap(pipe, dst); +} + + +/* Assumes all values are within bounds -- no checking at this level - + * do it higher up if required. + */ +static void +i915_region_copy(struct pipe_context *pipe, + struct pipe_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + struct pipe_region *src, + GLuint src_offset, + GLuint srcx, GLuint srcy, GLuint width, GLuint height) +{ + assert( dst != src ); + assert( dst->cpp == src->cpp ); + + if (0) { + _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe->region_map(pipe, src) + src_offset, + src->pitch, + srcx, srcy); + + pipe->region_unmap(pipe, src); + pipe->region_unmap(pipe, dst); + } + else { + i915_copy_blit( i915_context(pipe), + dst->cpp, + src->pitch, src->buffer, src_offset, + dst->pitch, dst->buffer, dst_offset, + srcx, srcy, dstx, dsty, width, height ); + } +} + +/* Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +static GLubyte * +get_pointer(struct pipe_region *dst, GLuint x, GLuint y) +{ + return dst->map + (y * dst->pitch + x) * dst->cpp; +} + + +static void +i915_region_fill(struct pipe_context *pipe, + struct pipe_region *dst, + GLuint dst_offset, + GLuint dstx, GLuint dsty, + GLuint width, GLuint height, GLuint value) +{ + if (0) { + GLuint i, j; + + (void)pipe->region_map(pipe, dst); + + switch (dst->cpp) { + case 1: { + GLubyte *row = get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: { + GLushort *row = (GLushort *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + case 4: { + GLuint *row = (GLuint *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + default: + assert(0); + break; + } + } + else { + i915_fill_blit( i915_context(pipe), + dst->cpp, + dst->pitch, + dst->buffer, dst_offset, + dstx, dsty, + width, height, + value ); + } +} + + + + + +void +i915_init_region_functions(struct i915_context *i915) +{ + i915->pipe.region_idle = i915_region_idle; + i915->pipe.region_map = i915_region_map; + i915->pipe.region_unmap = i915_region_unmap; + i915->pipe.region_alloc = i915_region_alloc; + i915->pipe.region_release = i915_region_release; + i915->pipe.region_data = i915_region_data; + i915->pipe.region_copy = i915_region_copy; + i915->pipe.region_fill = i915_region_fill; +} + diff --git a/src/mesa/pipe/i915simple/i915_regions.h b/src/mesa/pipe/i915simple/i915_regions.h new file mode 100644 index 0000000000..d938c107a4 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_regions.h @@ -0,0 +1,141 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_REGIONS_H +#define INTEL_REGIONS_H + +#include "mtypes.h" +#include "intel_screen.h" + +struct intel_context; +struct intel_buffer_object; + +/** + * A layer on top of the bufmgr buffers that adds a few useful things: + * + * - Refcounting for local buffer references. + * - Refcounting for buffer maps + * - Buffer dimensions - pitch and height. + * - Blitter commands for copying 2D regions between buffers. (really???) + */ +struct intel_region +{ + struct _DriBufferObject *buffer; /**< buffer manager's buffer ID */ + GLuint refcount; /**< Reference count for region */ + GLuint cpp; /**< bytes per pixel */ + GLuint pitch; /**< in pixels */ + GLuint height; /**< in pixels */ + GLubyte *map; /**< only non-NULL when region is actually mapped */ + GLuint map_refcount; /**< Reference count for mapping */ + + GLuint draw_offset; /**< Offset of drawing address within the region */ + + struct intel_buffer_object *pbo; /* zero-copy uploads */ +}; + + +/* Allocate a refcounted region. Pointers to regions should only be + * copied by calling intel_reference_region(). + */ +struct intel_region *intel_region_alloc(intelScreenPrivate *intelScreen, + GLuint cpp, + GLuint pitch, GLuint height); + +void intel_region_reference(struct intel_region **dst, + struct intel_region *src); + +void intel_region_release(struct intel_region **ib); + +extern struct intel_region +*intel_region_create_static(intelScreenPrivate *intelScreen, + GLuint mem_type, + GLuint offset, + void *virtual, + GLuint cpp, + GLuint pitch, GLuint height); +extern void +intel_region_update_static(intelScreenPrivate *intelScreen, + struct intel_region *region, + GLuint mem_type, + GLuint offset, + void *virtual, + GLuint cpp, GLuint pitch, GLuint height); + + +void intel_region_idle(intelScreenPrivate *intelScreen, + struct intel_region *ib); + +/* Map/unmap regions. This is refcounted also: + */ +GLubyte *intel_region_map(intelScreenPrivate *intelScreen, + struct intel_region *ib); + +void intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *ib); + + +/* Upload data to a rectangular sub-region + */ +void intel_region_data(intelScreenPrivate *intelScreen, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + const void *src, GLuint src_stride, + GLuint srcx, GLuint srcy, GLuint width, GLuint height); + +/* Copy rectangular sub-regions + */ +void intel_region_copy(intelScreenPrivate *intelScreen, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + struct intel_region *src, + GLuint src_offset, + GLuint srcx, GLuint srcy, GLuint width, GLuint height); + +/* Fill a rectangular sub-region + */ +void intel_region_fill(intelScreenPrivate *intelScreen, + struct intel_region *dest, + GLuint dest_offset, + GLuint destx, GLuint desty, + GLuint width, GLuint height, GLuint color); + +/* Helpers for zerocopy uploads, particularly texture image uploads: + */ +void intel_region_attach_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region, + struct intel_buffer_object *pbo); +void intel_region_release_pbo(intelScreenPrivate *intelScreen, + struct intel_region *region); +void intel_region_cow(intelScreenPrivate *intelScreen, + struct intel_region *region); + +struct _DriBufferObject *intel_region_buffer(intelScreenPrivate *intelScreen, + struct intel_region *region, + GLuint flag); + +#endif diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c new file mode 100644 index 0000000000..d732e879db --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -0,0 +1,238 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "pipe/draw/draw_context.h" + +#include "i915_context.h" +#include "i915_state.h" + +/* None of this state is actually used for anything yet. + */ +static void i915_set_blend_state( struct pipe_context *pipe, + const struct pipe_blend_state *blend ) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->blend = *blend; + + i915->dirty |= I915_NEW_BLEND; +} + + +static void i915_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->blend_color = *blend_color; + + i915->dirty |= I915_NEW_BLEND; +} + + +/** XXX move someday? Or consolidate all these simple state setters + * into one file. + */ +static void i915_set_depth_test_state(struct pipe_context *pipe, + const struct pipe_depth_state *depth) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->depth_test = *depth; + + i915->dirty |= I915_NEW_DEPTH_TEST; +} + +static void i915_set_alpha_test_state(struct pipe_context *pipe, + const struct pipe_alpha_test_state *alpha) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->alpha_test = *alpha; + + i915->dirty |= I915_NEW_ALPHA_TEST; +} + +static void i915_set_stencil_state(struct pipe_context *pipe, + const struct pipe_stencil_state *stencil) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->stencil = *stencil; + + i915->dirty |= I915_NEW_STENCIL; +} + + + +static void i915_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct i915_context *i915 = i915_context(pipe); + + memcpy( &i915->scissor, scissor, sizeof(*scissor) ); + i915->dirty |= I915_NEW_SCISSOR; +} + + +static void i915_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ +} + + + +static void i915_set_fs_state( struct pipe_context *pipe, + const struct pipe_fs_state *fs ) +{ + struct i915_context *i915 = i915_context(pipe); + + memcpy(&i915->fs, fs, sizeof(*fs)); + + i915->dirty |= I915_NEW_FS; +} + + +static void i915_set_sampler_state(struct pipe_context *pipe, + GLuint unit, + const struct pipe_sampler_state *sampler) +{ + struct i915_context *i915 = i915_context(pipe); + + assert(unit < PIPE_MAX_SAMPLERS); + i915->sampler[unit] = *sampler; + + i915->dirty |= I915_NEW_SAMPLER; +} + + +static void i915_set_texture_state(struct pipe_context *pipe, + GLuint unit, + struct pipe_texture_object *texture) +{ + struct i915_context *i915 = i915_context(pipe); + + assert(unit < PIPE_MAX_SAMPLERS); + i915->texture[unit] = texture; /* ptr, not struct */ + + i915->dirty |= I915_NEW_TEXTURE; +} + + + +static void i915_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->framebuffer = *fb; /* struct copy */ + + i915->dirty |= I915_NEW_FRAMEBUFFER; +} + + + + +static void i915_set_clear_color_state(struct pipe_context *pipe, + const struct pipe_clear_color_state *clear) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->clear_color = *clear; /* struct copy */ +} + + + +static void i915_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct i915_context *i915 = i915_context(pipe); + + draw_set_clip_state(i915->draw, clip); + + i915->dirty |= I915_NEW_CLIP; +} + + + +/* Called when driver state tracker notices changes to the viewport + * matrix: + */ +static void i915_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->viewport = *viewport; /* struct copy */ + + /* pass the viewport info to the draw module */ + draw_set_viewport_state(i915->draw, viewport); + + /* Using tnl/ and vf/ modules is temporary while getting started. + * Full pipe will have vertex shader, vertex fetch of its own. + */ + i915->dirty |= I915_NEW_VIEWPORT; + +} + +static void i915_set_setup_state( struct pipe_context *pipe, + const struct pipe_setup_state *setup ) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->setup = *setup; + + /* pass-through to draw module */ + draw_set_setup_state(i915->draw, setup); + + i915->dirty |= I915_NEW_SETUP; +} + + +void +i915_init_state_functions( struct i915_context *i915 ) +{ + i915->pipe.set_alpha_test_state = i915_set_alpha_test_state; + i915->pipe.set_blend_color = i915_set_blend_color; + i915->pipe.set_blend_state = i915_set_blend_state; + i915->pipe.set_clip_state = i915_set_clip_state; + i915->pipe.set_clear_color_state = i915_set_clear_color_state; + i915->pipe.set_depth_state = i915_set_depth_test_state; + i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; + i915->pipe.set_fs_state = i915_set_fs_state; + i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; + i915->pipe.set_sampler_state = i915_set_sampler_state; + i915->pipe.set_scissor_state = i915_set_scissor_state; + i915->pipe.set_setup_state = i915_set_setup_state; + i915->pipe.set_stencil_state = i915_set_stencil_state; + i915->pipe.set_texture_state = i915_set_texture_state; + i915->pipe.set_viewport_state = i915_set_viewport_state; +} diff --git a/src/mesa/pipe/i915simple/i915_state.h b/src/mesa/pipe/i915simple/i915_state.h new file mode 100644 index 0000000000..c9b9f15d39 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef I915_STATE_H +#define I915_STATE_H + +struct i915_context; + +void i915_update_derived( struct i915_context *i915 ); + +void i915_emit_hardware_state( struct i915_context *i915 ); + +#endif diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c new file mode 100644 index 0000000000..392ac09b40 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -0,0 +1,186 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "main/glheader.h" +#include "main/macros.h" +#include "main/enums.h" + +#include "vf/vf.h" +#include "pipe/draw/draw_context.h" +#include "i915_context.h" +#include "i915_state.h" + + +#define EMIT_ATTR( VF_ATTR, FRAG_ATTR, INTERP ) \ +do { \ + slot_to_vf_attr[nr_attrs] = VF_ATTR; \ + nr_attrs++; \ + attr_mask |= (1 << (VF_ATTR)); \ +} while (0) + + +static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = +{ + VF_ATTRIB_POS, + VF_ATTRIB_COLOR0, + VF_ATTRIB_COLOR1, + VF_ATTRIB_FOG, + VF_ATTRIB_TEX0, + VF_ATTRIB_TEX1, + VF_ATTRIB_TEX2, + VF_ATTRIB_TEX3, + VF_ATTRIB_TEX4, + VF_ATTRIB_TEX5, + VF_ATTRIB_TEX6, + VF_ATTRIB_TEX7, + VF_ATTRIB_VAR0, + VF_ATTRIB_VAR1, + VF_ATTRIB_VAR2, + VF_ATTRIB_VAR3, + VF_ATTRIB_VAR4, + VF_ATTRIB_VAR5, + VF_ATTRIB_VAR6, + VF_ATTRIB_VAR7, +}; + + +/** + * Determine which post-transform / pre-rasterization vertex attributes + * we need. + * Derived from: fs, setup states. + */ +static void calculate_vertex_layout( struct i915_context *i915 ) +{ +// const GLbitfield inputsRead = i915->fs.inputs_read; + const GLbitfield inputsRead = (FRAG_ATTRIB_WPOS | FRAG_ATTRIB_COL0); + GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; + GLbitfield attr_mask = 0x0; + GLuint nr_attrs = 0; + GLuint i; + + memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); + + + /* TODO - Figure out if we need to do perspective divide, etc. + */ + EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); + + /* Pull in the rest of the attributes. They are all in float4 + * format. Future optimizations could be to keep some attributes + * as fixed point or ubyte format. + */ + for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { + if (inputsRead & (1 << i)) { + assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); + if (i915->setup.flatshade + && (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1)) + EMIT_ATTR(frag_to_vf[i], i, INTERP_CONSTANT); + else + EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); + } + } + + for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { + if (inputsRead & (1 << i)) { + assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); + EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); + } + } + + /* Additional attributes required for setup: Just twosided + * lighting. Edgeflag is dealt with specially by setting bits in + * the vertex header. + */ + if (i915->setup.light_twoside) { + if (inputsRead & FRAG_BIT_COL0) { + EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ + } + + if (inputsRead & FRAG_BIT_COL1) { + EMIT_ATTR(VF_ATTRIB_BFC1, FRAG_ATTRIB_MAX, 0); /* XXX: discard after setup */ + } + } + + /* If the attributes have changed, tell the draw module (which in turn + * tells the vf module) about the new vertex layout. + */ + draw_set_vertex_attributes( i915->draw, + slot_to_vf_attr, + nr_attrs ); +} + + +/** + * Recompute cliprect from scissor bounds, scissor enable and surface size. + */ +static void +compute_cliprect(struct i915_context *sp) +{ + GLint surfWidth, surfHeight; + + if (sp->framebuffer.num_cbufs > 0) { + surfWidth = sp->framebuffer.cbufs[0]->width; + surfHeight = sp->framebuffer.cbufs[0]->height; + } + else { + /* no surface? */ + surfWidth = sp->scissor.maxx; + surfHeight = sp->scissor.maxy; + } + + if (sp->setup.scissor) { + /* clip to scissor rect */ + sp->cliprect.minx = MAX2(sp->scissor.minx, 0); + sp->cliprect.miny = MAX2(sp->scissor.miny, 0); + sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth); + sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight); + } + else { + /* clip to surface bounds */ + sp->cliprect.minx = 0; + sp->cliprect.miny = 0; + sp->cliprect.maxx = surfWidth; + sp->cliprect.maxy = surfHeight; + } +} + + +/* Hopefully this will remain quite simple, otherwise need to pull in + * something like the state tracker mechanism. + */ +void i915_update_derived( struct i915_context *i915 ) +{ + if (i915->dirty & (I915_NEW_SETUP | I915_NEW_FS)) + calculate_vertex_layout( i915 ); + + if (i915->dirty & (I915_NEW_SCISSOR | + I915_NEW_STENCIL | + I915_NEW_FRAMEBUFFER)) + compute_cliprect(i915); + + i915->dirty = 0; +} diff --git a/src/mesa/pipe/i915simple/i915_state_draw.c b/src/mesa/pipe/i915simple/i915_state_draw.c new file mode 100644 index 0000000000..6c436b4c06 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state_draw.c @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "imports.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "pipe/draw/draw_context.h" + + + diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c new file mode 100644 index 0000000000..dd7932bc0d --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -0,0 +1,241 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + + +#include "glheader.h" +#include "mtypes.h" + +#include "i915_reg.h" +#include "i915_context.h" +#include "i915_winsys.h" +#include "i915_batch.h" +#include "i915_reg.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + +static unsigned translate_format( unsigned format ) +{ + switch (format) { + case PIPE_FORMAT_U_R8_G8_B8_A8: + return COLOR_BUF_ARGB8888; + case PIPE_FORMAT_U_R5_G6_B5: + return COLOR_BUF_RGB565; + default: + assert(0); + return 0; + } +} + +static unsigned translate_depth_format( unsigned zformat ) +{ + assert(zformat == PIPE_FORMAT_S8_Z24); + return DEPTH_FRMT_24_FIXED_8_OTHER; +} + + +/* Push the state into the sarea and/or texture memory. + */ +void +i915_emit_hardware_state(struct i915_context *i915 ) +{ + BEGIN_BATCH(100, 10); + + { + OUT_BATCH(_3DSTATE_AA_CMD | + AA_LINE_ECAAR_WIDTH_ENABLE | + AA_LINE_ECAAR_WIDTH_1_0 | + AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0); + } + + { + OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); + OUT_BATCH(0); + + OUT_BATCH(_3DSTATE_DFLT_SPEC_CMD); + OUT_BATCH(0); + + OUT_BATCH(_3DSTATE_DFLT_Z_CMD); + OUT_BATCH(0); + } + + + { + OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS | + CSB_TCB(0, 0) | + CSB_TCB(1, 1) | + CSB_TCB(2, 2) | + CSB_TCB(3, 3) | + CSB_TCB(4, 4) | + CSB_TCB(5, 5) | + CSB_TCB(6, 6) | + CSB_TCB(7, 7)); + } + + { + OUT_BATCH(_3DSTATE_RASTER_RULES_CMD | + ENABLE_POINT_RASTER_RULE | + OGL_POINT_RASTER_RULE | + ENABLE_LINE_STRIP_PROVOKE_VRTX | + ENABLE_TRI_FAN_PROVOKE_VRTX | + LINE_STRIP_PROVOKE_VRTX(1) | + TRI_FAN_PROVOKE_VRTX(2) | + ENABLE_TEXKILL_3D_4D | + TEXKILL_4D); + } + + /* Need to initialize this to zero. + */ + { + OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (1)); + OUT_BATCH(0); + } + + + { + OUT_BATCH(_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); + + OUT_BATCH(_3DSTATE_SCISSOR_RECT_0_CMD); + OUT_BATCH(0); + OUT_BATCH(0); + } + + { + OUT_BATCH(_3DSTATE_DEPTH_SUBRECT_DISABLE); + } + + { + OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0); /* disable indirect state */ + OUT_BATCH(0); + } + + + { + /* Don't support twosided stencil yet */ + OUT_BATCH(_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_TWO_SIDE | 0); + } + + + + { + OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | + I1_LOAD_S(2) | + I1_LOAD_S(4) | + I1_LOAD_S(5) | + I1_LOAD_S(6) | + (3)); + + OUT_BATCH(0); + OUT_BATCH(S4_VFMT_XYZ | S4_VFMT_COLOR); + OUT_BATCH(0); + OUT_BATCH(S6_COLOR_WRITE_ENABLE | + (2 << S6_TRISTRIP_PV_SHIFT)); + } + + { + OUT_BATCH(_3DSTATE_MODES_4_CMD | + ENABLE_LOGIC_OP_FUNC | + LOGIC_OP_FUNC(LOGICOP_COPY) | + ENABLE_STENCIL_TEST_MASK | + STENCIL_TEST_MASK(0xff) | + ENABLE_STENCIL_WRITE_MASK | + STENCIL_WRITE_MASK(0xff)); + } + + { + OUT_BATCH(_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | + IAB_MODIFY_ENABLE | + IAB_MODIFY_FUNC | + IAB_MODIFY_SRC_FACTOR | + IAB_MODIFY_DST_FACTOR); + } + + + if (i915->framebuffer.cbufs[0]) { + struct pipe_region *cbuf_region = i915->framebuffer.cbufs[0]->region; + unsigned pitch = (cbuf_region->pitch * + cbuf_region->cpp); + + OUT_BATCH(_3DSTATE_BUF_INFO_CMD); + + OUT_BATCH(BUF_3D_ID_COLOR_BACK | + BUF_3D_PITCH(pitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); + + OUT_RELOC(cbuf_region->buffer, + I915_BUFFER_ACCESS_WRITE, + cbuf_region->draw_offset); + } + + /* What happens if no zbuf?? + */ + if (i915->framebuffer.zbuf) { + struct pipe_region *depth_region = i915->framebuffer.zbuf->region; + unsigned zpitch = (depth_region->pitch * + depth_region->cpp); + + OUT_BATCH(_3DSTATE_BUF_INFO_CMD); + + OUT_BATCH(BUF_3D_ID_DEPTH | + BUF_3D_PITCH(zpitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); + + OUT_RELOC(depth_region->buffer, + I915_BUFFER_ACCESS_WRITE, + depth_region->draw_offset); + } + + + { + unsigned cformat = i915->framebuffer.cbufs[0]->format; + unsigned zformat = i915->framebuffer.zbuf->format; + + OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); + + OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | /* .5 */ + LOD_PRECLAMP_OGL | + TEX_DEFAULT_COLOR_OGL | + translate_format( cformat ) | + translate_depth_format( zformat )); + } + + { + OUT_BATCH(_3DSTATE_STIPPLE); + OUT_BATCH(0); + } + + { + GLuint i, dwords; + GLuint *prog = i915_passthrough_program( &dwords ); + + for (i = 0; i < dwords; i++) + OUT_BATCH( prog[i] ); + } +} + diff --git a/src/mesa/pipe/i915simple/i915_state_fragprog.c b/src/mesa/pipe/i915simple/i915_state_fragprog.c new file mode 100644 index 0000000000..e4639bc223 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state_fragprog.c @@ -0,0 +1,67 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_reg.h" +#include "i915_context.h" +#include "i915_debug.h" + + +static unsigned passthrough[] = +{ + _3DSTATE_PIXEL_SHADER_PROGRAM | ((2*3)-1), + + /* declare input color: + */ + (D0_DCL | + (REG_TYPE_T << D0_TYPE_SHIFT) | + (T_DIFFUSE << D0_NR_SHIFT) | + D0_CHANNEL_ALL), + 0, + 0, + + /* move to output color: + */ + (A0_MOV | + (REG_TYPE_OC << A0_DEST_TYPE_SHIFT) | + A0_DEST_CHANNEL_ALL | + (REG_TYPE_T << A0_SRC0_TYPE_SHIFT) | + (T_DIFFUSE << A0_SRC0_NR_SHIFT)), + 0, + 0 +}; + +unsigned *i915_passthrough_program( unsigned *dwords ) +{ + i915_disassemble_program( passthrough, Elements(passthrough) ); + + *dwords = Elements(passthrough); + return passthrough; +} + + + + diff --git a/src/mesa/pipe/i915simple/i915_state_vertex.c b/src/mesa/pipe/i915simple/i915_state_vertex.c new file mode 100644 index 0000000000..053e39deb2 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state_vertex.c @@ -0,0 +1,145 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "main/glheader.h" +#include "main/macros.h" +#include "main/enums.h" + +#include "vf/vf.h" +#include "pipe/draw/draw_context.h" +#include "i915_context.h" +#include "i915_state.h" + + + + +/*********************************************************************** + * + */ + +#define SZ_TO_HW(sz) ((sz-2)&0x3) +#define EMIT_SZ(sz) (EMIT_1F + (sz) - 1) +#define EMIT_ATTR( ATTR, STYLE, S4, SZ ) \ +do { \ + i915->vertex_attrs[i915->vertex_attr_count].attrib = (ATTR); \ + i915->vertex_attrs[i915->vertex_attr_count].format = (STYLE); \ + i915->vertex_attr_count++; \ + s4 |= S4; \ + offset += (SZ); \ +} while (0) + +#define EMIT_PAD( N ) \ +do { \ + i915->vertex_attrs[i915->vertex_attr_count].attrib = 0; \ + i915->vertex_attrs[i915->vertex_attr_count].format = EMIT_PAD; \ + i915->vertex_attrs[i915->vertex_attr_count].offset = (N); \ + i915->vertex_attr_count++; \ + offset += (N); \ +} while (0) + + + +/** + * Determine which post-transform / pre-rasterization vertex attributes + * we need. + * Derived from: fs, setup states. + */ +static void calculate_vertex_layout( struct i915_context *i915 ) +{ + const GLbitfield inputsRead = i915->fs.inputs_read; + GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; + GLuint nr_attrs = 0; + GLuint i; + + + + memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); + + EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); + EMIT_ATTR(VF_ATTRIB_COLOR0, FRAG_ATTRIB_COL0, INTERP_LINEAR); + + i915->hw_vertex_size = i915->nr_attrs; + + /* Additional attributes required for setup: Just twosided + * lighting. Edgeflag is dealt with specially by setting bits in + * the vertex header. + */ + if (i915->setup.light_twoside) { + EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ + } + + /* If the attributes have changed, tell the draw module (which in turn + * tells the vf module) about the new vertex layout. + */ + draw_set_vertex_attributes( i915->draw, + slot_to_vf_attr, + i915->nr_attrs ); +} + + + +/*********************************************************************** + * + */ +static inline GLuint attr_size(GLuint sizes, GLuint attr) +{ + return ((sizes >> (attr*2)) & 0x3) + 1; +} + +static void i915_calculate_vertex_format( struct intel_context *intel ) +{ + struct i915_context *i915 = i915_context( &intel->ctx ); + + GLuint s2 = S2_TEXCOORD_NONE; + GLuint s4 = 0; + GLuint offset = 0; + GLuint i; + i915->vertex_attr_count = 0; + + EMIT_ATTR(VF_ATTRIB_POS, EMIT_4F_VIEWPORT, S4_VFMT_XYZW, 16); + EMIT_ATTR(VF_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, S4_VFMT_COLOR, 4); + + + + if (s2 != i915->vertex_format.LIS2 || + s4 != i915->vertex_format.LIS4) { + + clip_set_hw_vertex_format( intel->clip, + i915->vertex_attrs, + i915->vertex_attr_count, + offset ); + + /* Needed? This does raise the INTEL_NEW_VERTEX_SIZE flag: + */ + intel_vb_set_vertex_size( intel->vb, offset ); + + i915->vertex_format.LIS2 = s2; + i915->vertex_format.LIS4 = s4; + intel->state.dirty.intel |= I915_NEW_VERTEX_FORMAT; + } +} + diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c new file mode 100644 index 0000000000..73d33570f6 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -0,0 +1,52 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_context.h" +#include "i915_state.h" +#include "pipe/p_defines.h" +#include "main/imports.h" + + +static struct pipe_surface * +i915_surface_alloc(struct pipe_context *pipe, GLenum format) +{ + struct i915_surface *surf = CALLOC_STRUCT(i915_surface); + + if (!surf) + return NULL; + + surf->surface.format = format; + + return &surf->surface; +} + + +void +i915_init_surface_functions(struct i915_context *i915) +{ + i915->pipe.surface_alloc = i915_surface_alloc; +} diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h new file mode 100644 index 0000000000..514c53c7b2 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -0,0 +1,128 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_WINSYS_H +#define I915_WINSYS_H + +#include "main/mtypes.h" + +/* This is the interface that softpipe requires any window system + * hosting it to implement. This is the only include file in softpipe + * which is public. + */ + + +/* Pipe drivers are (meant to be!) independent of both GL and the + * window system. The window system provides a buffer manager and a + * set of additional hooks for things like command buffer submission, + * etc. + * + * There clearly has to be some agreement between the window system + * driver and the hardware driver about the format of command buffers, + * etc. + */ + +struct pipe_buffer_handle; + +struct i915_winsys { + + + /* The buffer manager is modeled after the dri_bugmgr interface, + * but this is the subset that i915 cares about. Remember that + * i915 gets to choose the interface it needs, and the window + * systems must then implement that interface (rather than the + * other way around...). + * + * I915 only really wants to make system memory allocations, + * right?? + */ + struct pipe_buffer_handle *(*create_buffer)(struct i915_winsys *sws, + unsigned alignment ); + + void *(*buffer_map)( struct i915_winsys *sws, + struct pipe_buffer_handle *buf ); + + void (*buffer_unmap)( struct i915_winsys *sws, + struct pipe_buffer_handle *buf ); + + struct pipe_buffer_handle *(*buffer_reference)( struct i915_winsys *sws, + struct pipe_buffer_handle *buf ); + + void (*buffer_unreference)( struct i915_winsys *sws, + struct pipe_buffer_handle **buf ); + + void (*buffer_data)(struct i915_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ); + + void (*buffer_subdata)(struct i915_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data); + + void (*buffer_get_subdata)(struct i915_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data); + + + /* An over-simple batchbuffer mechanism. Will want to improve the + * performance of this, perhaps based on the cmdstream stuff. It + * would be pretty impossible to implement swz on top of this + * interface. + * + * Will also need additions/changes to implement static/dynamic + * indirect state. + */ + unsigned *(*batch_start)( struct i915_winsys *sws, + unsigned dwords, + unsigned relocs ); + void (*batch_dword)( struct i915_winsys *sws, + unsigned dword ); + void (*batch_reloc)( struct i915_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned access_flags, + unsigned delta ); + void (*batch_flush)( struct i915_winsys *sws ); + + + /* Printf??? + */ + void (*dpf)( const char *fmt, ... ); + +}; + +#define I915_BUFFER_ACCESS_WRITE 0x1 +#define I915_BUFFER_ACCESS_READ 0x2 + + +struct pipe_context *i915_create( struct i915_winsys * ); + + +#endif -- cgit v1.2.3 From 08f33a025100dea2d951e6d628891fe294b18082 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 09:35:03 +0100 Subject: fully support all texcoord wrap modes --- src/mesa/pipe/softpipe/sp_tex_sample.c | 284 +++++++++++++++++++++++++++++---- 1 file changed, 257 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index e501cea5ad..de8e01ea07 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -34,7 +34,7 @@ #include "main/macros.h" -#include "sp_surface.h" +#include "sp_context.h" #include "sp_surface.h" #include "sp_tex_sample.h" #include "pipe/p_context.h" @@ -42,6 +42,19 @@ #include "pipe/tgsi/core/tgsi_exec.h" +/* + * Note, the FRAC macro has to work perfectly. Otherwise you'll sometimes + * see 1-pixel bands of improperly weighted linear-filtered textures. + * The tests/texwrap.c demo is a good test. + * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0. + * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x). + */ +#define FRAC(f) ((f) - IFLOOR(f)) + + +/** + * Linear interpolation macro + */ #define LERP(T, A, B) ( (A) + (T) * ((B) - (A)) ) @@ -64,51 +77,268 @@ lerp_2d(GLfloat a, GLfloat b, /** - * Apply texture coord wrapping mode and return integer texture - * row/column/img index. + * Compute the remainder of a divided by b, but be careful with + * negative values so that REPEAT mode works right. */ -static INLINE void -nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint width, GLint *k) +static INLINE GLint +repeat_remainder(GLint a, GLint b) +{ + if (a >= 0) + return a % b; + else + return (a + 1) % b + b - 1; +} + + +/** + * Apply texture coord wrapping mode and return integer texture index. + * \param wrapMode PIPE_TEX_WRAP_x + * \param s the texcoord + * \param size the texture image size + * \return integer texture index + */ +static INLINE GLint +nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint size) { - /* XXX this arithmetic isn't exactly right, add more wrap modes */ + GLint i; switch (wrapMode) { case PIPE_TEX_WRAP_REPEAT: - *k = (int) (s * width) % width; - break; + /* s limited to [0,1) */ + /* i limited to [0,size-1] */ + i = IFLOOR(s * size); + i = repeat_remainder(i, size); + return i; case PIPE_TEX_WRAP_CLAMP: - *k = (int) CLAMP(s, 0.0, 1.0) * width; - break; + /* s limited to [0,1] */ + /* i limited to [0,size-1] */ + if (s <= 0.0F) + i = 0; + else if (s >= 1.0F) + i = size - 1; + else + i = IFLOOR(s * size); + return i; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + { + /* s limited to [min,max] */ + /* i limited to [0, size-1] */ + const GLfloat min = 1.0F / (2.0F * size); + const GLfloat max = 1.0F - min; + if (s < min) + i = 0; + else if (s > max) + i = size - 1; + else + i = IFLOOR(s * size); + } + return i; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + { + /* s limited to [min,max] */ + /* i limited to [-1, size] */ + const GLfloat min = -1.0F / (2.0F * size); + const GLfloat max = 1.0F - min; + if (s <= min) + i = -1; + else if (s >= max) + i = size; + else + i = IFLOOR(s * size); + } + return i; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + { + const GLfloat min = 1.0F / (2.0F * size); + const GLfloat max = 1.0F - min; + const GLint flr = IFLOOR(s); + GLfloat u; + if (flr & 1) + u = 1.0F - (s - (GLfloat) flr); + else + u = s - (GLfloat) flr; + if (u < min) + i = 0; + else if (u > max) + i = size - 1; + else + i = IFLOOR(u * size); + } + return i; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + { + /* s limited to [0,1] */ + /* i limited to [0,size-1] */ + const GLfloat u = FABSF(s); + if (u <= 0.0F) + i = 0; + else if (u >= 1.0F) + i = size - 1; + else + i = IFLOOR(u * size); + } + return i; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + { + /* s limited to [min,max] */ + /* i limited to [0, size-1] */ + const GLfloat min = 1.0F / (2.0F * size); + const GLfloat max = 1.0F - min; + const GLfloat u = FABSF(s); + if (u < min) + i = 0; + else if (u > max) + i = size - 1; + else + i = IFLOOR(u * size); + } + return i; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + { + /* s limited to [min,max] */ + /* i limited to [0, size-1] */ + const GLfloat min = -1.0F / (2.0F * size); + const GLfloat max = 1.0F - min; + const GLfloat u = FABSF(s); + if (u < min) + i = -1; + else if (u > max) + i = size; + else + i = IFLOOR(u * size); + } + return i; default: assert(0); - *k = 0; + return 0; } } + +/** + * Used to compute texel locations for linear sampling. + * \param wrapMode PIPE_TEX_WRAP_x + * \param s the texcoord + * \param size the texture image size + * \param i0 returns first texture index + * \param i1 returns second texture index (usually *i0 + 1) + * \param a returns blend factor/weight between texture indexes + */ static INLINE void -linear_texcoord(GLuint wrapMode, GLfloat s, GLuint width, GLint *k0, GLint *k1, - GLfloat *a) +linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, + GLint *i0, GLint *i1, GLfloat *a) { - /* XXX this arithmetic isn't exactly right, add more wrap modes */ + GLfloat u; switch (wrapMode) { case PIPE_TEX_WRAP_REPEAT: - *k0 = (int) (s * width) % width; - *k1 = (*k0 + 1) % width; + u = s * size - 0.5F; + *i0 = repeat_remainder(IFLOOR(u), size); + *i1 = repeat_remainder(*i0 + 1, size); break; case PIPE_TEX_WRAP_CLAMP: - *k0 = (int) CLAMP(s, 0.0, 1.0) * width; - *k1 = (*k0 + 1); - if (*k1 >= width) - *k1 = width - 1; + if (s <= 0.0F) + u = 0.0F; + else if (s >= 1.0F) + u = (GLfloat) size; + else + u = s * size; + u -= 0.5F; + *i0 = IFLOOR(u); + *i1 = *i0 + 1; + break; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + if (s <= 0.0F) + u = 0.0F; + else if (s >= 1.0F) + u = (GLfloat) size; + else + u = s * size; + u -= 0.5F; + *i0 = IFLOOR(u); + *i1 = *i0 + 1; + if (*i0 < 0) + *i0 = 0; + if (*i1 >= (GLint) size) + *i1 = size - 1; + break; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + { + const GLfloat min = -1.0F / (2.0F * size); + const GLfloat max = 1.0F - min; + if (s <= min) + u = min * size; + else if (s >= max) + u = max * size; + else + u = s * size; + u -= 0.5F; + *i0 = IFLOOR(u); + *i1 = *i0 + 1; + } + break; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + { + const GLint flr = IFLOOR(s); + if (flr & 1) + u = 1.0F - (s - (GLfloat) flr); + else + u = s - (GLfloat) flr; + u = (u * size) - 0.5F; + *i0 = IFLOOR(u); + *i1 = *i0 + 1; + if (*i0 < 0) + *i0 = 0; + if (*i1 >= (GLint) size) + *i1 = size - 1; + } + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + u = FABSF(s); + if (u >= 1.0F) + u = (GLfloat) size; + else + u *= size; + u -= 0.5F; + *i0 = IFLOOR(u); + *i1 = *i0 + 1; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + u = FABSF(s); + if (u >= 1.0F) + u = (GLfloat) size; + else + u *= size; + u -= 0.5F; + *i0 = IFLOOR(u); + *i1 = *i0 + 1; + if (*i0 < 0) + *i0 = 0; + if (*i1 >= (GLint) size) + *i1 = size - 1; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + { + const GLfloat min = -1.0F / (2.0F * size); + const GLfloat max = 1.0F - min; + u = FABSF(s); + if (u <= min) + u = min * size; + else if (u >= max) + u = max * size; + else + u *= size; + u -= 0.5F; + *i0 = IFLOOR(u); + *i1 = *i0 + 1; + } break; default: assert(0); - *k0 = *k1 = 0; } - /* kludge the lerp weight: */ - *a = (s * width) - (int) (s * width); + *a = FRAC(u); } + /** * Called via tgsi_sampler::get_sample() * Use the sampler's state setting to get a filtered RGBA value @@ -133,10 +363,10 @@ sp_get_sample(struct tgsi_sampler *sampler, case PIPE_TEX_FILTER_NEAREST: { GLint x, y; - nearest_texcoord(sampler->state->wrap_s, strq[0], - sampler->texture->width0, &x); - nearest_texcoord(sampler->state->wrap_t, strq[1], - sampler->texture->height0, &y); + x = nearest_texcoord(sampler->state->wrap_s, strq[0], + sampler->texture->width0); + y = nearest_texcoord(sampler->state->wrap_t, strq[1], + sampler->texture->height0); ps->get_tile(ps, x, y, 1, 1, rgba); } break; -- cgit v1.2.3 From f62b573922dd5274c01250fa941f9b1a151e6c45 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 09:35:16 +0100 Subject: more get_tile() work --- src/mesa/pipe/softpipe/sp_surface.c | 40 +++++++++++++++++++++++++++++++------ 1 file changed, 34 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 87bb64af1a..85e1fcfa38 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -322,12 +322,37 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, { const GLuint *src = ((const GLuint *) ps->region->map) + y * ps->region->pitch + x; - assert(w == 1); - assert(h == 1); - p[0] = UBYTE_TO_FLOAT((src[0] >> 16) & 0xff); - p[1] = UBYTE_TO_FLOAT((src[0] >> 8) & 0xff); - p[2] = UBYTE_TO_FLOAT((src[0] >> 0) & 0xff); - p[3] = UBYTE_TO_FLOAT((src[0] >> 24) & 0xff); + GLuint i, j; + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + p[0] = UBYTE_TO_FLOAT((src[j] >> 16) & 0xff); + p[1] = UBYTE_TO_FLOAT((src[j] >> 8) & 0xff); + p[2] = UBYTE_TO_FLOAT((src[j] >> 0) & 0xff); + p[3] = UBYTE_TO_FLOAT((src[j] >> 24) & 0xff); + p += 4; + } + src += ps->region->pitch; + } +} + + +static void +a1r5g5b5_get_tile(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) +{ + const GLushort *src + = ((const GLushort *) ps->region->map) + y * ps->region->pitch + x; + GLuint i, j; + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + p[0] = ((src[j] >> 10) & 0x1f) * (1.0 / 31); + p[1] = ((src[j] >> 5) & 0x1f) * (1.0 / 31); + p[2] = ((src[j] ) & 0x1f) * (1.0 / 31); + p[3] = src[j] >> 15; + p += 4; + } + src += ps->region->pitch; + } } @@ -358,6 +383,9 @@ init_quad_funcs(struct softpipe_surface *sps) case PIPE_FORMAT_U_A8_R8_G8_B8: sps->surface.get_tile = a8r8g8b8_get_tile; break; + case PIPE_FORMAT_U_A1_R5_G5_B5: + sps->surface.get_tile = a1r5g5b5_get_tile; + break; default: /* assert(0); -- cgit v1.2.3 From 8a475cb791a0e66fcdc3fcd12c517fcb98d75957 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 09:41:52 +0100 Subject: Rename drivers/dri/i915pipe --> drivers/dri/intel_winsys. This is appropriate as this is not a "pipe driver for the i915" as the old name would suggest, but rather a dri driver that can provide "winsys" backends to softpipe/i915/i965/etc pipe drivers, running under the intel DDX, drm, etc. It also frees up the i915pipe name for something more appropriate, ie mesa/pipe/i915pipe. --- src/mesa/drivers/dri/i915pipe/Makefile | 49 - src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c | 342 ----- src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h | 123 -- src/mesa/drivers/dri/i915pipe/intel_batchpool.c | 418 ------- src/mesa/drivers/dri/i915pipe/intel_blit.c | 394 ------ src/mesa/drivers/dri/i915pipe/intel_blit.h | 62 - src/mesa/drivers/dri/i915pipe/intel_buffers.c | 711 ----------- src/mesa/drivers/dri/i915pipe/intel_buffers.h | 55 - src/mesa/drivers/dri/i915pipe/intel_context.c | 717 ----------- src/mesa/drivers/dri/i915pipe/intel_context.h | 225 ---- src/mesa/drivers/dri/i915pipe/intel_depthstencil.c | 282 ----- src/mesa/drivers/dri/i915pipe/intel_depthstencil.h | 14 - src/mesa/drivers/dri/i915pipe/intel_fbo.c | 629 ---------- src/mesa/drivers/dri/i915pipe/intel_fbo.h | 127 -- src/mesa/drivers/dri/i915pipe/intel_ioctl.c | 135 -- src/mesa/drivers/dri/i915pipe/intel_ioctl.h | 40 - src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c | 293 ----- src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h | 105 -- src/mesa/drivers/dri/i915pipe/intel_reg.h | 88 -- src/mesa/drivers/dri/i915pipe/intel_screen.c | 597 --------- src/mesa/drivers/dri/i915pipe/intel_screen.h | 111 -- src/mesa/drivers/dri/i915pipe/intel_softpipe.c | 184 --- src/mesa/drivers/dri/i915pipe/intel_surface.c | 196 --- src/mesa/drivers/dri/i915pipe/intel_tex.c | 184 --- src/mesa/drivers/dri/i915pipe/intel_tex.h | 212 ---- src/mesa/drivers/dri/i915pipe/intel_tex_copy.c | 315 ----- src/mesa/drivers/dri/i915pipe/intel_tex_format.c | 172 --- src/mesa/drivers/dri/i915pipe/intel_tex_image.c | 695 ----------- src/mesa/drivers/dri/i915pipe/intel_tex_layout.c | 1 - src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c | 192 --- src/mesa/drivers/dri/i915pipe/intel_tex_validate.c | 277 ----- src/mesa/drivers/dri/i915pipe/server/i830_common.h | 226 ---- src/mesa/drivers/dri/i915pipe/server/i830_dri.h | 63 - src/mesa/drivers/dri/i915pipe/server/intel.h | 331 ----- src/mesa/drivers/dri/i915pipe/server/intel_dri.c | 1306 -------------------- src/mesa/drivers/dri/intel_winsys/Makefile | 49 + .../drivers/dri/intel_winsys/intel_batchbuffer.c | 342 +++++ .../drivers/dri/intel_winsys/intel_batchbuffer.h | 123 ++ .../drivers/dri/intel_winsys/intel_batchpool.c | 418 +++++++ src/mesa/drivers/dri/intel_winsys/intel_blit.c | 394 ++++++ src/mesa/drivers/dri/intel_winsys/intel_blit.h | 62 + src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 711 +++++++++++ src/mesa/drivers/dri/intel_winsys/intel_buffers.h | 55 + src/mesa/drivers/dri/intel_winsys/intel_context.c | 717 +++++++++++ src/mesa/drivers/dri/intel_winsys/intel_context.h | 225 ++++ .../drivers/dri/intel_winsys/intel_depthstencil.c | 282 +++++ .../drivers/dri/intel_winsys/intel_depthstencil.h | 14 + src/mesa/drivers/dri/intel_winsys/intel_fbo.c | 629 ++++++++++ src/mesa/drivers/dri/intel_winsys/intel_fbo.h | 127 ++ src/mesa/drivers/dri/intel_winsys/intel_ioctl.c | 135 ++ src/mesa/drivers/dri/intel_winsys/intel_ioctl.h | 40 + .../drivers/dri/intel_winsys/intel_mipmap_tree.c | 293 +++++ .../drivers/dri/intel_winsys/intel_mipmap_tree.h | 105 ++ src/mesa/drivers/dri/intel_winsys/intel_reg.h | 88 ++ src/mesa/drivers/dri/intel_winsys/intel_screen.c | 597 +++++++++ src/mesa/drivers/dri/intel_winsys/intel_screen.h | 111 ++ src/mesa/drivers/dri/intel_winsys/intel_softpipe.c | 184 +++ src/mesa/drivers/dri/intel_winsys/intel_surface.c | 196 +++ src/mesa/drivers/dri/intel_winsys/intel_tex.c | 184 +++ src/mesa/drivers/dri/intel_winsys/intel_tex.h | 212 ++++ src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c | 315 +++++ .../drivers/dri/intel_winsys/intel_tex_format.c | 172 +++ .../drivers/dri/intel_winsys/intel_tex_image.c | 695 +++++++++++ .../drivers/dri/intel_winsys/intel_tex_layout.c | 1 + .../drivers/dri/intel_winsys/intel_tex_subimage.c | 192 +++ .../drivers/dri/intel_winsys/intel_tex_validate.c | 277 +++++ .../drivers/dri/intel_winsys/server/i830_common.h | 226 ++++ .../drivers/dri/intel_winsys/server/i830_dri.h | 63 + src/mesa/drivers/dri/intel_winsys/server/intel.h | 331 +++++ .../drivers/dri/intel_winsys/server/intel_dri.c | 1306 ++++++++++++++++++++ 70 files changed, 9871 insertions(+), 9871 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915pipe/Makefile delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_batchpool.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_blit.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_blit.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_buffers.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_buffers.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_context.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_context.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_depthstencil.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_depthstencil.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_fbo.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_fbo.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_ioctl.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_ioctl.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_reg.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_screen.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_screen.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_softpipe.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_surface.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex.h delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_copy.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_format.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_image.c delete mode 120000 src/mesa/drivers/dri/i915pipe/intel_tex_layout.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c delete mode 100644 src/mesa/drivers/dri/i915pipe/intel_tex_validate.c delete mode 100644 src/mesa/drivers/dri/i915pipe/server/i830_common.h delete mode 100644 src/mesa/drivers/dri/i915pipe/server/i830_dri.h delete mode 100644 src/mesa/drivers/dri/i915pipe/server/intel.h delete mode 100644 src/mesa/drivers/dri/i915pipe/server/intel_dri.c create mode 100644 src/mesa/drivers/dri/intel_winsys/Makefile create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_batchpool.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_blit.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_blit.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_buffers.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_buffers.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_context.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_context.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_depthstencil.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_depthstencil.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_fbo.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_fbo.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_ioctl.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_ioctl.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_reg.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_screen.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_screen.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_softpipe.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_surface.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_format.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_image.c create mode 120000 src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_subimage.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_validate.c create mode 100644 src/mesa/drivers/dri/intel_winsys/server/i830_common.h create mode 100644 src/mesa/drivers/dri/intel_winsys/server/i830_dri.h create mode 100644 src/mesa/drivers/dri/intel_winsys/server/intel.h create mode 100644 src/mesa/drivers/dri/intel_winsys/server/intel_dri.c (limited to 'src') diff --git a/src/mesa/drivers/dri/i915pipe/Makefile b/src/mesa/drivers/dri/i915pipe/Makefile deleted file mode 100644 index e4e9cf17b0..0000000000 --- a/src/mesa/drivers/dri/i915pipe/Makefile +++ /dev/null @@ -1,49 +0,0 @@ - -TOP = ../../../../.. -include $(TOP)/configs/current - -LIBNAME = i915tex_dri.so - -MINIGLX_SOURCES = server/intel_dri.c - -PIPE_DRIVERS = \ - $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a - -DRIVER_SOURCES = \ - intel_softpipe.c \ - intel_batchbuffer.c \ - intel_tex_layout.c \ - intel_buffers.c \ - intel_blit.c \ - intel_context.c \ - intel_ioctl.c \ - intel_screen.c \ - intel_surface.c \ - intel_fbo.c \ - intel_depthstencil.c \ - intel_batchpool.c - -OLD_TEX =\ - intel_tex_image.c \ - intel_tex_subimage.c \ - intel_tex_copy.c \ - intel_tex_validate.c \ - intel_tex_format.c \ - intel_tex.c - - -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(COMMON_BM_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") - -include ../Makefile.template - -intel_tex_layout.o: ../intel/intel_tex_layout.c - -symlinks: diff --git a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c deleted file mode 100644 index 60bd8eaec2..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.c +++ /dev/null @@ -1,342 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_batchbuffer.h" -#include "intel_ioctl.h" - -/* Relocations in kernel space: - * - pass dma buffer seperately - * - memory manager knows how to patch - * - pass list of dependent buffers - * - pass relocation list - * - * Either: - * - get back an offset for buffer to fire - * - memory manager knows how to fire buffer - * - * Really want the buffer to be AGP and pinned. - * - */ - -/* Cliprect fence: The highest fence protecting a dma buffer - * containing explicit cliprect information. Like the old drawable - * lock but irq-driven. X server must wait for this fence to expire - * before changing cliprects [and then doing sw rendering?]. For - * other dma buffers, the scheduler will grab current cliprect info - * and mix into buffer. X server must hold the lock while changing - * cliprects??? Make per-drawable. Need cliprects in shared memory - * -- beats storing them with every cmd buffer in the queue. - * - * ==> X server must wait for this fence to expire before touching the - * framebuffer with new cliprects. - * - * ==> Cliprect-dependent buffers associated with a - * cliprect-timestamp. All of the buffers associated with a timestamp - * must go to hardware before any buffer with a newer timestamp. - * - * ==> Dma should be queued per-drawable for correct X/GL - * synchronization. Or can fences be used for this? - * - * Applies to: Blit operations, metaops, X server operations -- X - * server automatically waits on its own dma to complete before - * modifying cliprects ??? - */ - -static void -intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) -{ - int i; - fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); - for (i = 0; i < count / 4; i += 4) - fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", - offset + i * 4, ptr[i], ptr[i + 1], ptr[i + 2], ptr[i + 3]); - fprintf(stderr, "END BATCH\n\n\n"); -} - -void -intel_batchbuffer_reset(struct intel_batchbuffer *batch) -{ - - int i; - - /* - * Get a new, free batchbuffer. - */ - - batch->size = BATCH_SZ; - driBOData(batch->buffer, batch->size, NULL, 0); - - driBOResetList(&batch->list); - - /* - * Unreference buffers previously on the relocation list. - */ - - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - driBOUnReference(r->buf); - } - - batch->list_count = 0; - batch->nr_relocs = 0; - batch->flags = 0; - - /* - * We don't refcount the batchbuffer itself since we can't destroy it - * while it's on the list. - */ - - - driBOAddListItem(&batch->list, batch->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE); - - - batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); - batch->ptr = batch->map; -} - -/*====================================================================== - * Public functions - */ -struct intel_batchbuffer * -intel_batchbuffer_alloc(struct intel_context *intel) -{ - struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); - - batch->intel = intel; - - driGenBuffers(intel->intelScreen->batchPool, "batchbuffer", 1, - &batch->buffer, 4096, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); - batch->last_fence = NULL; - driBOCreateList(20, &batch->list); - intel_batchbuffer_reset(batch); - return batch; -} - -void -intel_batchbuffer_free(struct intel_batchbuffer *batch) -{ - if (batch->last_fence) { - driFenceFinish(batch->last_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); - driFenceUnReference(batch->last_fence); - batch->last_fence = NULL; - } - if (batch->map) { - driBOUnmap(batch->buffer); - batch->map = NULL; - } - driBOUnReference(batch->buffer); - batch->buffer = NULL; - free(batch); -} - -/* TODO: Push this whole function into bufmgr. - */ -static void -do_flush_locked(struct intel_batchbuffer *batch, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) -{ - GLuint *ptr; - GLuint i; - struct intel_context *intel = batch->intel; - unsigned fenceFlags; - struct _DriFenceObject *fo; - - driBOValidateList(batch->intel->driFd, &batch->list); - - /* Apply the relocations. This nasty map indicates to me that the - * whole task should be done internally by the memory manager, and - * that dma buffers probably need to be pinned within agp space. - */ - ptr = (GLuint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, - DRM_BO_HINT_ALLOW_UNFENCED_MAP); - - - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - - ptr[r->offset / 4] = driBOOffset(r->buf) + r->delta; - } - - if (INTEL_DEBUG & DEBUG_BATCH) - intel_dump_batchbuffer(0, ptr, used); - - driBOUnmap(batch->buffer); - batch->map = NULL; - - /* Throw away non-effective packets. Won't work once we have - * hardware contexts which would preserve statechanges beyond a - * single buffer. - */ - - if (!(intel->numClipRects == 0 && !ignore_cliprects)) { - intel_batch_ioctl(batch->intel, - driBOOffset(batch->buffer), - used, ignore_cliprects, allow_unlock); - } - - - /* - * Kernel fencing. The flags tells the kernel that we've - * programmed an MI_FLUSH. - */ - - fenceFlags = DRM_I915_FENCE_FLAG_FLUSHED; - fo = driFenceBuffers(batch->intel->driFd, - "Batch fence", fenceFlags); - - /* - * User space fencing. - */ - - driBOFence(batch->buffer, fo); - - if (driFenceType(fo) == DRM_FENCE_TYPE_EXE) { - - /* - * Oops. We only validated a batch buffer. This means we - * didn't do any proper rendering. Discard this fence object. - */ - - driFenceUnReference(fo); - } else { - driFenceUnReference(batch->last_fence); - batch->last_fence = fo; - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - driBOFence(r->buf, fo); - } - } - - if (intel->numClipRects == 0 && !ignore_cliprects) { - if (allow_unlock) { - UNLOCK_HARDWARE(intel); - sched_yield(); - LOCK_HARDWARE(intel); - } -// intel->vtbl.lost_hardware(intel); - } -} - - -struct _DriFenceObject * -intel_batchbuffer_flush(struct intel_batchbuffer *batch) -{ - struct intel_context *intel = batch->intel; - GLuint used = batch->ptr - batch->map; - GLboolean was_locked = intel->locked; - - if (used == 0) - return batch->last_fence; - -#define MI_FLUSH ((0<<29)|(4<<23)) - - /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a - * performance drain that we would like to avoid. - */ - if (used & 4) { - ((int *) batch->ptr)[0] = MI_FLUSH; - ((int *) batch->ptr)[1] = 0; - ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->ptr)[0] = MI_FLUSH; - ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; - used += 8; - } - - driBOUnmap(batch->buffer); - batch->ptr = NULL; - batch->map = NULL; - - /* TODO: Just pass the relocation list and dma buffer up to the - * kernel. - */ - if (!was_locked) - LOCK_HARDWARE(intel); - - do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), - GL_FALSE); - - if (!was_locked) - UNLOCK_HARDWARE(intel); - - /* Reset the buffer: - */ - intel_batchbuffer_reset(batch); - return batch->last_fence; -} - -void -intel_batchbuffer_finish(struct intel_batchbuffer *batch) -{ - struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); - driFenceReference(fence); - driFenceFinish(fence, 3, GL_FALSE); - driFenceUnReference(fence); -} - - -/* This is the only way buffers get added to the validate list. - */ -GLboolean -intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, - struct _DriBufferObject *buffer, - GLuint flags, GLuint mask, GLuint delta) -{ - assert(batch->nr_relocs < MAX_RELOCS); - - driBOAddListItem(&batch->list, buffer, flags, mask); - - { - struct buffer_reloc *r = &batch->reloc[batch->nr_relocs++]; - driBOReference(buffer); - r->buf = buffer; - r->offset = batch->ptr - batch->map; - r->delta = delta; - } - - batch->ptr += 4; - return GL_TRUE; -} - - - -void -intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags) -{ - assert((bytes & 3) == 0); - intel_batchbuffer_require_space(batch, bytes, flags); - memcpy(batch->ptr, data, bytes); - batch->ptr += bytes; -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h b/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h deleted file mode 100644 index 6d4d05e0bb..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_batchbuffer.h +++ /dev/null @@ -1,123 +0,0 @@ -#ifndef INTEL_BATCHBUFFER_H -#define INTEL_BATCHBUFFER_H - -#include "mtypes.h" -#include "dri_bufmgr.h" - -struct intel_context; - -#define BATCH_SZ 16384 -#define BATCH_RESERVED 16 - -#define MAX_RELOCS 4096 - -#define INTEL_BATCH_NO_CLIPRECTS 0x1 -#define INTEL_BATCH_CLIPRECTS 0x2 - -struct buffer_reloc -{ - struct _DriBufferObject *buf; - GLuint offset; - GLuint delta; /* not needed? */ -}; - -struct intel_batchbuffer -{ - struct bufmgr *bm; - struct intel_context *intel; - - struct _DriBufferObject *buffer; - struct _DriFenceObject *last_fence; - GLuint flags; - - drmBOList list; - GLuint list_count; - GLubyte *map; - GLubyte *ptr; - - struct buffer_reloc reloc[MAX_RELOCS]; - GLuint nr_relocs; - GLuint size; -}; - -struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context - *intel); - -void intel_batchbuffer_free(struct intel_batchbuffer *batch); - - -void intel_batchbuffer_finish(struct intel_batchbuffer *batch); - -struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer - *batch); - -void intel_batchbuffer_reset(struct intel_batchbuffer *batch); - - -/* Unlike bmBufferData, this currently requires the buffer be mapped. - * Consider it a convenience function wrapping multple - * intel_buffer_dword() calls. - */ -void intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags); - -void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, - GLuint bytes); - -GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, - struct _DriBufferObject *buffer, - GLuint flags, - GLuint mask, GLuint offset); - -/* Inline functions - might actually be better off with these - * non-inlined. Certainly better off switching all command packets to - * be passed as structs rather than dwords, but that's a little bit of - * work... - */ -static INLINE GLuint -intel_batchbuffer_space(struct intel_batchbuffer *batch) -{ - return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); -} - - -static INLINE void -intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) -{ - assert(batch->map); - assert(intel_batchbuffer_space(batch) >= 4); - *(GLuint *) (batch->ptr) = dword; - batch->ptr += 4; -} - -static INLINE void -intel_batchbuffer_require_space(struct intel_batchbuffer *batch, - GLuint sz, GLuint flags) -{ - assert(sz < batch->size - 8); - if (intel_batchbuffer_space(batch) < sz || - (batch->flags != 0 && flags != 0 && batch->flags != flags)) - intel_batchbuffer_flush(batch); - - batch->flags |= flags; -} - -/* Here are the crusty old macros, to be removed: - */ -#define BATCH_LOCALS - -#define BEGIN_BATCH(n, flags) do { \ - intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ -} while (0) - -#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) - -#define OUT_RELOC(buf,flags,mask,delta) do { \ - assert((delta) >= 0); \ - intel_batchbuffer_emit_reloc(intel->batch, buf, flags, mask, delta); \ -} while (0) - -#define ADVANCE_BATCH() do { } while(0) - - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_batchpool.c b/src/mesa/drivers/dri/i915pipe/intel_batchpool.c deleted file mode 100644 index 2503b8a62a..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_batchpool.c +++ /dev/null @@ -1,418 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#include -#include -#include -#include "imports.h" -#include "glthread.h" -#include "dri_bufpool.h" -#include "dri_bufmgr.h" -#include "intel_screen.h" - -typedef struct -{ - drmMMListHead head; - struct _BPool *parent; - struct _DriFenceObject *fence; - unsigned long start; - int unfenced; - int mapped; -} BBuf; - -typedef struct _BPool -{ - _glthread_Mutex mutex; - unsigned long bufSize; - unsigned poolSize; - unsigned numFree; - unsigned numTot; - unsigned numDelayed; - unsigned checkDelayed; - drmMMListHead free; - drmMMListHead delayed; - drmMMListHead head; - drmBO kernelBO; - void *virtual; - BBuf *bufs; -} BPool; - - -static BPool * -createBPool(int fd, unsigned long bufSize, unsigned numBufs, unsigned flags, - unsigned checkDelayed) -{ - BPool *p = (BPool *) malloc(sizeof(*p)); - BBuf *buf; - int i; - - if (!p) - return NULL; - - p->bufs = (BBuf *) malloc(numBufs * sizeof(*p->bufs)); - if (!p->bufs) { - free(p); - return NULL; - } - - DRMINITLISTHEAD(&p->free); - DRMINITLISTHEAD(&p->head); - DRMINITLISTHEAD(&p->delayed); - - p->numTot = numBufs; - p->numFree = numBufs; - p->bufSize = bufSize; - p->numDelayed = 0; - p->checkDelayed = checkDelayed; - - _glthread_INIT_MUTEX(p->mutex); - - if (drmBOCreate(fd, 0, numBufs * bufSize, 0, NULL, drm_bo_type_dc, - flags, DRM_BO_HINT_DONT_FENCE, &p->kernelBO)) { - free(p->bufs); - free(p); - return NULL; - } - if (drmBOMap(fd, &p->kernelBO, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, - &p->virtual)) { - drmBODestroy(fd, &p->kernelBO); - free(p->bufs); - free(p); - return NULL; - } - - /* - * We unmap the buffer so that we can validate it later. Note that this is - * just a synchronizing operation. The buffer will have a virtual mapping - * until it is destroyed. - */ - - drmBOUnmap(fd, &p->kernelBO); - - buf = p->bufs; - for (i = 0; i < numBufs; ++i) { - buf->parent = p; - buf->fence = NULL; - buf->start = i * bufSize; - buf->mapped = 0; - buf->unfenced = 0; - DRMLISTADDTAIL(&buf->head, &p->free); - buf++; - } - - return p; -} - - -static void -pool_checkFree(BPool * p, int wait) -{ - drmMMListHead *list, *prev; - BBuf *buf; - int signaled = 0; - int i; - - list = p->delayed.next; - - if (p->numDelayed > 3) { - for (i = 0; i < p->numDelayed; i += 3) { - list = list->next; - } - } - - prev = list->prev; - for (; list != &p->delayed; list = prev, prev = list->prev) { - - buf = DRMLISTENTRY(BBuf, list, head); - - if (!signaled) { - if (wait) { - driFenceFinish(buf->fence, DRM_FENCE_TYPE_EXE, 1); - signaled = 1; - } - else { - signaled = driFenceSignaled(buf->fence, DRM_FENCE_TYPE_EXE); - } - } - - if (!signaled) - break; - - driFenceUnReference(buf->fence); - buf->fence = NULL; - DRMLISTDEL(list); - p->numDelayed--; - DRMLISTADD(list, &p->free); - p->numFree++; - } -} - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, unsigned flags, unsigned hint, - unsigned alignment) -{ - BPool *p = (BPool *) pool->data; - - drmMMListHead *item; - - if (alignment && (alignment != 4096)) - return NULL; - - _glthread_LOCK_MUTEX(p->mutex); - - if (p->numFree == 0) - pool_checkFree(p, GL_TRUE); - - if (p->numFree == 0) { - fprintf(stderr, "Out of fixed size buffer objects\n"); - BM_CKFATAL(-ENOMEM); - } - - item = p->free.next; - - if (item == &p->free) { - fprintf(stderr, "Fixed size buffer pool corruption\n"); - } - - DRMLISTDEL(item); - --p->numFree; - - _glthread_UNLOCK_MUTEX(p->mutex); - return (void *) DRMLISTENTRY(BBuf, item, head); -} - - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - - if (buf->fence) { - DRMLISTADDTAIL(&buf->head, &p->delayed); - p->numDelayed++; - } - else { - buf->unfenced = 0; - DRMLISTADD(&buf->head, &p->free); - p->numFree++; - } - - if ((p->numDelayed % p->checkDelayed) == 0) - pool_checkFree(p, 0); - - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, void **virtual) -{ - - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - - /* - * Currently Mesa doesn't have any condition variables to resolve this - * cleanly in a multithreading environment. - * We bail out instead. - */ - - if (buf->mapped) { - fprintf(stderr, "Trying to map already mapped buffer object\n"); - BM_CKFATAL(-EINVAL); - } - -#if 0 - if (buf->unfenced && !(hint & DRM_BO_HINT_ALLOW_UNFENCED_MAP)) { - fprintf(stderr, "Trying to map an unfenced buffer object 0x%08x" - " 0x%08x %d\n", hint, flags, buf->start); - BM_CKFATAL(-EINVAL); - } - -#endif - - if (buf->fence) { - _glthread_UNLOCK_MUTEX(p->mutex); - return -EBUSY; - } - - buf->mapped = GL_TRUE; - *virtual = (unsigned char *) p->virtual + buf->start; - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, int lazy) -{ - BBuf *buf = (BBuf *) private; - driFenceFinish(buf->fence, 0, lazy); - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - - buf->mapped = 0; - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - return p->kernelBO.offset + buf->start; -} - -static unsigned -pool_flags(struct _DriBufferPool *pool, void *private) -{ - BPool *p = (BPool *) pool->data; - - return p->kernelBO.flags; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - BPool *p = (BPool *) pool->data; - - return p->bufSize; -} - - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - if (buf->fence) { - driFenceUnReference(buf->fence); - } - buf->fence = fence; - buf->unfenced = 0; - driFenceReference(buf->fence); - _glthread_UNLOCK_MUTEX(p->mutex); - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - return &p->kernelBO; -} - -static int -pool_validate(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - _glthread_LOCK_MUTEX(p->mutex); - buf->unfenced = GL_TRUE; - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - BPool *p = (BPool *) pool->data; - - /* - * Wait on outstanding fences. - */ - - _glthread_LOCK_MUTEX(p->mutex); - while ((p->numFree < p->numTot) && p->numDelayed) { - _glthread_UNLOCK_MUTEX(p->mutex); - sched_yield(); - pool_checkFree(p, GL_TRUE); - _glthread_LOCK_MUTEX(p->mutex); - } - - drmBODestroy(pool->fd, &p->kernelBO); - free(p->bufs); - _glthread_UNLOCK_MUTEX(p->mutex); - free(p); - free(pool); -} - - -struct _DriBufferPool * -driBatchPoolInit(int fd, unsigned flags, - unsigned long bufSize, - unsigned numBufs, unsigned checkDelayed) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - if (!pool) - return NULL; - - pool->data = createBPool(fd, bufSize, numBufs, flags, checkDelayed); - if (!pool->data) - return NULL; - - pool->fd = fd; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = &pool_validate; - pool->waitIdle = &pool_waitIdle; - pool->setstatic = NULL; - pool->takeDown = &pool_takedown; - return pool; -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.c b/src/mesa/drivers/dri/i915pipe/intel_blit.c deleted file mode 100644 index 8e878f0088..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_blit.c +++ /dev/null @@ -1,394 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "mtypes.h" -#include "context.h" -#include "enums.h" - -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "intel_buffers.h" -#include "intel_context.h" -#include "intel_fbo.h" -#include "intel_reg.h" -#include "vblank.h" - -#include "pipe/p_context.h" - - -#define FILE_DEBUG_FLAG DEBUG_BLIT - -/** - * Copy the back color buffer to the front color buffer. - * Used for SwapBuffers(). - */ -void -intelCopyBuffer(__DRIdrawablePrivate * dPriv, - const drm_clip_rect_t * rect) -{ - - struct intel_context *intel; - const intelScreenPrivate *intelScreen; - - DBG("%s\n", __FUNCTION__); - - assert(dPriv); - - intel = intelScreenContext(dPriv->driScreenPriv->private); - if (!intel) - return; - - intelScreen = intel->intelScreen; - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - current context (which is what intelScreenContext should return) might - not get a contended lock and thus cliprects not updated (tests/manywin) */ - if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - - if (dPriv && dPriv->numClipRects) { - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - const struct pipe_region *backRegion - = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? - intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : - intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); - const int backWidth = intel_fb->Base.Width; - const int backHeight = intel_fb->Base.Height; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; - const int srcpitch = backRegion->pitch; - const int cpp = intelScreen->front.cpp; - int BR13, CMD; - int i; - - ASSERT(intel_fb); - ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ - ASSERT(backRegion); - ASSERT(backRegion->cpp == cpp); - - DBG("front pitch %d back pitch %d\n", - pitch, backRegion->pitch); - - if (cpp == 2) { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - } - else { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - } - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) - continue; - - box = *pbox; - - if (rect) { - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > backWidth) - box.x2 = backWidth + box.x1; - if (box.y2 - box.y1 > backHeight) - box.y2 = backHeight + box.y1; - - DBG("box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((box.y1 << 16) | box.x1); - OUT_BATCH((box.y2 << 16) | box.x2); - - OUT_RELOC(intelScreen->front.buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((sbox.y1 << 16) | sbox.x1); - OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - ADVANCE_BATCH(); - } - - if (intel->first_swap_fence) - driFenceUnReference(intel->first_swap_fence); - intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); - driFenceReference(intel->first_swap_fence); - } - - UNLOCK_HARDWARE(intel); - - /* XXX this is bogus. The context here may not even be bound to this drawable! */ - if (intel->lastStamp != dPriv->lastStamp) { - GET_CURRENT_CONTEXT(currctx); - struct intel_context *intelcurrent = intel_context(currctx); - if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { - intelWindowMoved(intel); - intel->lastStamp = dPriv->lastStamp; - } - } - -} - - - - -void -intelEmitFillBlit(struct intel_context *intel, - GLuint cpp, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort x, GLshort y, GLshort w, GLshort h, - GLuint value, GLuint mask) -{ - GLuint BR13, CMD; - GLboolean badMask = GL_FALSE; - BATCH_LOCALS; - - /* - printf("Emit fill blit value=0x%x mask=0x%x\n", value, mask); - */ - - dst_pitch *= cpp; - - switch (cpp) { - case 1: - case 2: - case 3: - BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); - CMD = XY_COLOR_BLT_CMD; - if ((mask & 0xffff) != 0xffff) - badMask = GL_TRUE; - break; - case 4: - BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); -#if 0 - CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); -#else - CMD = XY_COLOR_BLT_CMD; - if ((mask & 0xff000000) == 0xff000000) - CMD |= XY_COLOR_BLT_WRITE_ALPHA; - else if (mask & 0xff000000) - badMask = GL_TRUE; - if ((mask & 0x00ffffff) == 0x00ffffff) - CMD |= XY_COLOR_BLT_WRITE_RGB; - else if (mask & 0x00ffffff) - badMask = GL_TRUE; -#endif - break; - default: - return; - } - - assert(!badMask); - - DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); - - - BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((y << 16) | x); - OUT_BATCH(((y + h) << 16) | (x + w)); - OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); - OUT_BATCH(value); - ADVANCE_BATCH(); - - intel_batchbuffer_flush(intel->batch); -} - - -static GLuint translate_raster_op(GLenum logicop) -{ - switch(logicop) { - case GL_CLEAR: return 0x00; - case GL_AND: return 0x88; - case GL_AND_REVERSE: return 0x44; - case GL_COPY: return 0xCC; - case GL_AND_INVERTED: return 0x22; - case GL_NOOP: return 0xAA; - case GL_XOR: return 0x66; - case GL_OR: return 0xEE; - case GL_NOR: return 0x11; - case GL_EQUIV: return 0x99; - case GL_INVERT: return 0x55; - case GL_OR_REVERSE: return 0xDD; - case GL_COPY_INVERTED: return 0x33; - case GL_OR_INVERTED: return 0xBB; - case GL_NAND: return 0x77; - case GL_SET: return 0xFF; - default: return 0; - } -} - - -/* Copy BitBlt - */ -void -intelEmitCopyBlit(struct intel_context *intel, - GLuint cpp, - GLshort src_pitch, - struct _DriBufferObject *src_buffer, - GLuint src_offset, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort src_x, GLshort src_y, - GLshort dst_x, GLshort dst_y, - GLshort w, GLshort h, - GLenum logic_op) -{ - GLuint CMD, BR13; - int dst_y2 = dst_y + h; - int dst_x2 = dst_x + w; - BATCH_LOCALS; - - - DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, - src_buffer, src_pitch, src_offset, src_x, src_y, - dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); - - src_pitch *= cpp; - dst_pitch *= cpp; - - switch (cpp) { - case 1: - case 2: - case 3: - BR13 = (((GLint) dst_pitch) & 0xffff) | - (translate_raster_op(logic_op) << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - break; - case 4: - BR13 = - (((GLint) dst_pitch) & 0xffff) | - (translate_raster_op(logic_op) << 16) | (1 << 24) | (1 << 25); - CMD = - (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - break; - default: - return; - } - - if (dst_y2 < dst_y || dst_x2 < dst_x) { - return; - } - - /* Initial y values don't seem to work with negative pitches. If - * we adjust the offsets manually (below), it seems to work fine. - * - * On the other hand, if we always adjust, the hardware doesn't - * know which blit directions to use, so overlapping copypixels get - * the wrong result. - */ - if (dst_pitch > 0 && src_pitch > 0) { - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((dst_y << 16) | dst_x); - OUT_BATCH((dst_y2 << 16) | dst_x2); - OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); - OUT_BATCH((src_y << 16) | src_x); - OUT_BATCH(((GLint) src_pitch & 0xffff)); - OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, src_offset); - ADVANCE_BATCH(); - } - else { - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((0 << 16) | dst_x); - OUT_BATCH((h << 16) | dst_x2); - OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - dst_offset + dst_y * dst_pitch); - OUT_BATCH((0 << 16) | src_x); - OUT_BATCH(((GLint) src_pitch & 0xffff)); - OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, - src_offset + src_y * src_pitch); - ADVANCE_BATCH(); - } - - intel_batchbuffer_flush( intel->batch ); -} - - - diff --git a/src/mesa/drivers/dri/i915pipe/intel_blit.h b/src/mesa/drivers/dri/i915pipe/intel_blit.h deleted file mode 100644 index 46c2594477..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_blit.h +++ /dev/null @@ -1,62 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BLIT_H -#define INTEL_BLIT_H - -#include "intel_context.h" -#include "intel_ioctl.h" -#include "dri_bufmgr.h" - -extern void intelCopyBuffer(__DRIdrawablePrivate * dpriv, - const drm_clip_rect_t * rect); - -extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask); - -extern void intelEmitCopyBlit(struct intel_context *intel, - GLuint cpp, - GLshort src_pitch, - struct _DriBufferObject *src_buffer, - GLuint src_offset, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort srcx, GLshort srcy, - GLshort dstx, GLshort dsty, - GLshort w, GLshort h, - GLenum logicop ); - -extern void intelEmitFillBlit(struct intel_context *intel, - GLuint cpp, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort x, GLshort y, - GLshort w, GLshort h, GLuint value, GLuint mask); - - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.c b/src/mesa/drivers/dri/i915pipe/intel_buffers.c deleted file mode 100644 index fc9d60c88e..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.c +++ /dev/null @@ -1,711 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_blit.h" -#include "intel_buffers.h" -#include "intel_depthstencil.h" -#include "intel_fbo.h" -#include "intel_batchbuffer.h" -#include "intel_reg.h" -#include "context.h" -#include "utils.h" -#include "drirenderbuffer.h" -#include "framebuffer.h" -#include "vblank.h" - -#include "pipe/p_context.h" - -/* This block can be removed when libdrm >= 2.3.1 is required */ - -#ifndef DRM_VBLANK_FLIP - -#define DRM_VBLANK_FLIP 0x8000000 - -typedef struct drm_i915_flip { - int pipes; -} drm_i915_flip_t; - -#undef DRM_IOCTL_I915_FLIP -#define DRM_IOCTL_I915_FLIP DRM_IOW(DRM_COMMAND_BASE + DRM_I915_FLIP, \ - drm_i915_flip_t) - -#endif - - -/** - * XXX move this into a new dri/common/cliprects.c file. - */ -GLboolean -intel_intersect_cliprects(drm_clip_rect_t * dst, - const drm_clip_rect_t * a, - const drm_clip_rect_t * b) -{ - GLint bx = b->x1; - GLint by = b->y1; - GLint bw = b->x2 - bx; - GLint bh = b->y2 - by; - - if (bx < a->x1) - bw -= a->x1 - bx, bx = a->x1; - if (by < a->y1) - bh -= a->y1 - by, by = a->y1; - if (bx + bw > a->x2) - bw = a->x2 - bx; - if (by + bh > a->y2) - bh = a->y2 - by; - if (bw <= 0) - return GL_FALSE; - if (bh <= 0) - return GL_FALSE; - - dst->x1 = bx; - dst->y1 = by; - dst->x2 = bx + bw; - dst->y2 = by + bh; - - return GL_TRUE; -} - -/** - * Return pointer to current color drawing region, or NULL. - */ -struct pipe_region * -intel_drawbuf_region(struct intel_context *intel) -{ - struct intel_renderbuffer *irbColor = - intel_renderbuffer(intel->ctx.DrawBuffer->_ColorDrawBuffers[0][0]); - if (irbColor) - return irbColor->region; - else - return NULL; -} - -/** - * Return pointer to current color reading region, or NULL. - */ -struct pipe_region * -intel_readbuf_region(struct intel_context *intel) -{ - struct intel_renderbuffer *irb - = intel_renderbuffer(intel->ctx.ReadBuffer->_ColorReadBuffer); - if (irb) - return irb->region; - else - return NULL; -} - - - -/** - * Update the following fields for rendering: - * intel->numClipRects - * intel->pClipRects - */ -static void -intelSetRenderbufferClipRects(struct intel_context *intel) -{ - /* zero-sized buffers might be legal? */ - assert(intel->ctx.DrawBuffer->Width > 0); - assert(intel->ctx.DrawBuffer->Height > 0); - intel->fboRect.x1 = 0; - intel->fboRect.y1 = 0; - intel->fboRect.x2 = intel->ctx.DrawBuffer->Width; - intel->fboRect.y2 = intel->ctx.DrawBuffer->Height; - intel->numClipRects = 1; - intel->pClipRects = &intel->fboRect; -} - - -/** - * This will be called whenever the currently bound window is moved/resized. - * XXX: actually, it seems to NOT be called when the window is only moved (BP). - */ -void -intelWindowMoved(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - - if (!intel->ctx.DrawBuffer) { - /* when would this happen? -BP */ - assert(0); - intel->numClipRects = 0; - } - - /* Update Mesa's notion of window size */ - driUpdateFramebufferSize(ctx, dPriv); - intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ - - { - drmI830Sarea *sarea = intel->sarea; - drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, - .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; - drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, .y1 = sarea->pipeA_y, - .x2 = sarea->pipeA_x + sarea->pipeA_w, - .y2 = sarea->pipeA_y + sarea->pipeA_h }; - drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, .y1 = sarea->pipeB_y, - .x2 = sarea->pipeB_x + sarea->pipeB_w, - .y2 = sarea->pipeB_y + sarea->pipeB_h }; - GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); - GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); - GLuint flags = intel_fb->vblank_flags; - GLboolean pf_active; - GLint pf_pipes; - - /* Update page flipping info - */ - pf_pipes = 0; - - if (areaA > 0) - pf_pipes |= 1; - - if (areaB > 0) - pf_pipes |= 2; - - intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3; - - intel_fb->pf_num_pages = 2 /*intel->intelScreen->third.handle ? 3 : 2*/; - - pf_active = pf_pipes && (pf_pipes & intel->sarea->pf_active) == pf_pipes; - - if (INTEL_DEBUG & DEBUG_LOCK) - if (pf_active != intel_fb->pf_active) - _mesa_printf("%s - Page flipping %sactive\n", __progname, - pf_active ? "" : "in"); - - if (pf_active) { - /* Sync pages between pipes if we're flipping on both at the same time */ - if (pf_pipes == 0x3 && pf_pipes != intel_fb->pf_pipes && - (intel->sarea->pf_current_page & 0x3) != - (((intel->sarea->pf_current_page) >> 2) & 0x3)) { - drm_i915_flip_t flip; - - if (intel_fb->pf_current_page == - (intel->sarea->pf_current_page & 0x3)) { - /* XXX: This is ugly, but emitting two flips 'in a row' can cause - * lockups for unknown reasons. - */ - intel->sarea->pf_current_page = - intel->sarea->pf_current_page & 0x3; - intel->sarea->pf_current_page |= - ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % - intel_fb->pf_num_pages) << 2; - - flip.pipes = 0x2; - } else { - intel->sarea->pf_current_page = - intel->sarea->pf_current_page & (0x3 << 2); - intel->sarea->pf_current_page |= - (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % - intel_fb->pf_num_pages; - - flip.pipes = 0x1; - } - - drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); - } - - intel_fb->pf_pipes = pf_pipes; - } - - intel_fb->pf_active = pf_active; - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); - - /* Update vblank info - */ - if (areaB > areaA || (areaA == areaB && areaB > 0)) { - flags = intel_fb->vblank_flags | VBLANK_FLAG_SECONDARY; - } else { - flags = intel_fb->vblank_flags & ~VBLANK_FLAG_SECONDARY; - } - - if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && - !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { - drmVBlank vbl; - int i; - - vbl.request.type = DRM_VBLANK_ABSOLUTE; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - vbl.request.type |= DRM_VBLANK_SECONDARY; - } - - for (i = 0; i < intel_fb->pf_num_pages; i++) { - if (!intel_fb->color_rb[i] || - (intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <= - (1<<23)) - continue; - - vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending; - drmWaitVBlank(intel->driFd, &vbl); - } - - intel_fb->vblank_flags = flags; - driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); - intel_fb->vbl_waited = intel_fb->vbl_seq; - - for (i = 0; i < intel_fb->pf_num_pages; i++) { - if (intel_fb->color_rb[i]) - intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited; - } - } - } - - /* This will be picked up by looking at the dirty state flags: - */ - - /* Update hardware scissor */ -// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, -// ctx->Scissor.Width, ctx->Scissor.Height); - - /* Re-calculate viewport related state */ -// ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far ); -} - - - - - -/* Emit wait for pending flips */ -void -intel_wait_flips(struct intel_context *intel, GLuint batch_flags) -{ - struct intel_framebuffer *intel_fb = - (struct intel_framebuffer *) intel->ctx.DrawBuffer; - struct intel_renderbuffer *intel_rb = - intel_get_renderbuffer(&intel_fb->Base, - intel_fb->Base._ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : - BUFFER_BACK_LEFT); - - if (intel_fb->Base.Name == 0 && intel_rb->pf_pending == intel_fb->pf_seq) { - GLint pf_pipes = intel_fb->pf_pipes; - BATCH_LOCALS; - - /* Wait for pending flips to take effect */ - BEGIN_BATCH(2, batch_flags); - OUT_BATCH(pf_pipes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) - : 0); - OUT_BATCH(pf_pipes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) - : 0); - ADVANCE_BATCH(); - - intel_rb->pf_pending--; - } -} - -#if 0 -/* Flip the front & back buffers - */ -static GLboolean -intelPageFlip(const __DRIdrawablePrivate * dPriv) -{ - struct intel_context *intel; - int ret; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate; - - if (intel->intelScreen->drmMinor < 9) - return GL_FALSE; - - intelFlush(&intel->ctx); - - ret = 0; - - LOCK_HARDWARE(intel); - - if (dPriv->numClipRects && intel_fb->pf_active) { - drm_i915_flip_t flip; - - flip.pipes = intel_fb->pf_pipes; - - ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); - } - - UNLOCK_HARDWARE(intel); - - if (ret || !intel_fb->pf_active) - return GL_FALSE; - - if (!dPriv->numClipRects) { - usleep(10000); /* throttle invisible client 10ms */ - } - - intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3; - - if (dPriv->numClipRects != 0) { - intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending = - intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->pf_pending = - ++intel_fb->pf_seq; - } - - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, &intel_fb->Base); - - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s: success\n", __FUNCTION__); - - return GL_TRUE; -} -#endif - - -static GLboolean -intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) -{ - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); - struct intel_context *intel = - intelScreenContext(dPriv->driScreenPriv->private); - const intelScreenPrivate *intelScreen = intel->intelScreen; - unsigned int target; - drm_i915_vblank_swap_t swap; - GLboolean ret; - - if (!intel_fb->vblank_flags || - (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || - intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) - return GL_FALSE; - - swap.seqtype = DRM_VBLANK_ABSOLUTE; - - if (intel_fb->vblank_flags & VBLANK_FLAG_SYNC) { - swap.seqtype |= DRM_VBLANK_NEXTONMISS; - } else if (interval == 0) { - return GL_FALSE; - } - - swap.drawable = dPriv->hHWDrawable; - target = swap.sequence = intel_fb->vbl_seq + interval; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - swap.seqtype |= DRM_VBLANK_SECONDARY; - } - - LOCK_HARDWARE(intel); - - intel_batchbuffer_flush(intel->batch); - - if ( intel_fb->pf_active ) { - swap.seqtype |= DRM_VBLANK_FLIP; - - intel_fb->pf_current_page = (((intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3) + 1) % - intel_fb->pf_num_pages; - } - - if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, - sizeof(swap))) { - intel_fb->vbl_seq = swap.sequence; - swap.sequence -= target; - *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); - - intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending = - intel_get_renderbuffer(&intel_fb->Base, - BUFFER_FRONT_LEFT)->vbl_pending = - intel_fb->vbl_seq; - - if (swap.seqtype & DRM_VBLANK_FLIP) { - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); - } - - ret = GL_TRUE; - } else { - if (swap.seqtype & DRM_VBLANK_FLIP) { - intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3) % - intel_fb->pf_num_pages; - } - - ret = GL_FALSE; - } - - UNLOCK_HARDWARE(intel); - - return ret; -} - -void -intelSwapBuffers(__DRIdrawablePrivate * dPriv) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel; - - if (ctx == NULL) - return; - - intel = intel_context(ctx); - - if (ctx->Visual.doubleBufferMode) { - GLboolean missed_target; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - int64_t ust; - - _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - - if (!intelScheduleSwap(dPriv, &missed_target)) { - driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, - &missed_target); - - intelCopyBuffer(dPriv, NULL); - } - - intel_fb->swap_count++; - (*dri_interface->getUST) (&ust); - if (missed_target) { - intel_fb->swap_missed_count++; - intel_fb->swap_missed_ust = ust - intel_fb->swap_ust; - } - - intel_fb->swap_ust = ust; - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); - } -} - -void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - struct intel_context *intel = - (struct intel_context *) dPriv->driContextPriv->driverPrivate; - GLcontext *ctx = &intel->ctx; - - if (ctx->Visual.doubleBufferMode) { - drm_clip_rect_t rect; - /* fixup cliprect (driDrawable may have changed?) later */ - rect.x1 = x; - rect.y1 = y; - rect.x2 = w; - rect.y2 = h; - _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - intelCopyBuffer(dPriv, &rect); - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); - } -} - - -/** - * Update the hardware state for drawing into a window or framebuffer object. - * - * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other - * places within the driver. - * - * Basically, this needs to be called any time the current framebuffer - * changes, the renderbuffers change, or we need to draw into different - * color buffers. - */ -void -intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) -{ - struct intel_context *intel = intel_context(ctx); - struct pipe_region *colorRegion, *depthRegion = NULL; - struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; - - if (!fb) { - /* this can happen during the initial context initialization */ - return; - } - - /* Do this here, not core Mesa, since this function is called from - * many places within the driver. - */ - if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { - /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */ - _mesa_update_framebuffer(ctx); - /* this updates the DrawBuffer's Width/Height if it's a FBO */ - _mesa_update_draw_buffer_bounds(ctx); - } - - if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { - /* this may occur when we're called by glBindFrameBuffer() during - * the process of someone setting up renderbuffers, etc. - */ - /*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/ - return; - } - - if (fb->Name) - intel_validate_paired_depth_stencil(ctx, fb); - - /* - * How many color buffers are we drawing into? - */ - if (fb->_NumColorDrawBuffers[0] != 1) { - /* writing to 0 or 2 or 4 color buffers */ - /*_mesa_debug(ctx, "Software rendering\n");*/ - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); - } - else { - /* draw to exactly one color buffer */ - /*_mesa_debug(ctx, "Hardware rendering\n");*/ - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); - } - - /* - * Get the intel_renderbuffer for the colorbuffer we're drawing into. - * And set up cliprects. - */ - { - struct intel_renderbuffer *irb; - intelSetRenderbufferClipRects(intel); - irb = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); - colorRegion = (irb && irb->region) ? irb->region : NULL; - } - - /* Update culling direction which changes depending on the - * orientation of the buffer: - */ - if (ctx->Driver.FrontFace) - ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace); - else - ctx->NewState |= _NEW_POLYGON; - - if (!colorRegion) { - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); - } - else { - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); - } - - /*** - *** Get depth buffer region and check if we need a software fallback. - *** Note that the depth buffer is usually a DEPTH_STENCIL buffer. - ***/ - if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) { - irbDepth = intel_renderbuffer(fb->_DepthBuffer->Wrapped); - if (irbDepth && irbDepth->region) { - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); - depthRegion = irbDepth->region; - } - else { - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_TRUE); - depthRegion = NULL; - } - } - else { - /* not using depth buffer */ - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); - depthRegion = NULL; - } - - /*** - *** Stencil buffer - *** This can only be hardware accelerated if we're using a - *** combined DEPTH_STENCIL buffer (for now anyway). - ***/ - if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { - irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); - if (irbStencil && irbStencil->region) { - ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); - /* need to re-compute stencil hw state */ -// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); - if (!depthRegion) - depthRegion = irbStencil->region; - } - else { - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE); - } - } - else { - /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */ - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); - /* need to re-compute stencil hw state */ -// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); - } - - - /** - ** Release old regions, reference new regions - **/ - - // intel->vtbl.set_draw_region(intel, colorRegion, depthRegion); - - /* update viewport since it depends on window size */ -// ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y, -// ctx->Viewport.Width, ctx->Viewport.Height); - - /* Update hardware scissor */ -// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, -// ctx->Scissor.Width, ctx->Scissor.Height); -} - - -static void -intelDrawBuffer(GLcontext * ctx, GLenum mode) -{ - intel_draw_buffer(ctx, ctx->DrawBuffer); -} - - -static void -intelReadBuffer(GLcontext * ctx, GLenum mode) -{ - if (ctx->ReadBuffer == ctx->DrawBuffer) { - /* This will update FBO completeness status. - * A framebuffer will be incomplete if the GL_READ_BUFFER setting - * refers to a missing renderbuffer. Calling glReadBuffer can set - * that straight and can make the drawing buffer complete. - */ - intel_draw_buffer(ctx, ctx->DrawBuffer); - } - /* Generally, functions which read pixels (glReadPixels, glCopyPixels, etc) - * reference ctx->ReadBuffer and do appropriate state checks. - */ -} - - -void -intelInitBufferFuncs(struct dd_function_table *functions) -{ - functions->DrawBuffer = intelDrawBuffer; - functions->ReadBuffer = intelReadBuffer; -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_buffers.h b/src/mesa/drivers/dri/i915pipe/intel_buffers.h deleted file mode 100644 index 5834e39501..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_buffers.h +++ /dev/null @@ -1,55 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BUFFERS_H -#define INTEL_BUFFERS_H - - -struct intel_context; -struct intel_framebuffer; - - -extern GLboolean -intel_intersect_cliprects(drm_clip_rect_t * dest, - const drm_clip_rect_t * a, - const drm_clip_rect_t * b); - -extern struct pipe_region *intel_readbuf_region(struct intel_context *intel); - -extern struct pipe_region *intel_drawbuf_region(struct intel_context *intel); - -extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); - -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void intelWindowMoved(struct intel_context *intel); - -extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); - -extern void intelInitBufferFuncs(struct dd_function_table *functions); - -#endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.c b/src/mesa/drivers/dri/i915pipe/intel_context.c deleted file mode 100644 index e6f0d4a0fa..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_context.c +++ /dev/null @@ -1,717 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "glheader.h" -#include "context.h" -#include "matrix.h" -#include "simple_list.h" -#include "extensions.h" -#include "framebuffer.h" -#include "imports.h" -#include "points.h" - -#include "tnl/tnl.h" - -#include "tnl/t_pipeline.h" -#include "tnl/t_vertex.h" - -#include "drivers/common/driverfuncs.h" - -#include "intel_screen.h" - -#include "i830_dri.h" - -#include "intel_buffers.h" -/*#include "intel_tex.h"*/ -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "intel_blit.h" -/* -#include "intel_buffer_objects.h" -*/ -#include "intel_fbo.h" -#include "intel_tex_layout.h" - -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" - -#include "drirenderbuffer.h" -#include "vblank.h" -#include "utils.h" -#include "xmlpool.h" /* for symbolic values of enum-type options */ - -#include "pipe/p_context.h" - - - -#ifndef INTEL_DEBUG -int INTEL_DEBUG = (0); -#endif - -#define need_GL_ARB_multisample -#define need_GL_ARB_point_parameters -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_buffer_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_window_pos -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_equation_separate -#define need_GL_EXT_blend_func_separate -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_cull_vertex -#define need_GL_EXT_fog_coord -#define need_GL_EXT_framebuffer_object -#define need_GL_EXT_multi_draw_arrays -#define need_GL_EXT_secondary_color -#define need_GL_NV_vertex_program -#include "extension_helper.h" - - -#define DRIVER_DATE "20070731" - -_glthread_Mutex lockMutex; -static GLboolean lockMutexInit = GL_FALSE; - - -static const GLubyte * -intelGetString(GLcontext * ctx, GLenum name) -{ - const char *chipset; - static char buffer[128]; - - switch (name) { - case GL_VENDOR: - return (GLubyte *) "Tungsten Graphics, Inc"; - break; - - case GL_RENDERER: - switch (intel_context(ctx)->intelScreen->deviceID) { - case PCI_CHIP_845_G: - chipset = "Intel(R) 845G"; - break; - case PCI_CHIP_I830_M: - chipset = "Intel(R) 830M"; - break; - case PCI_CHIP_I855_GM: - chipset = "Intel(R) 852GM/855GM"; - break; - case PCI_CHIP_I865_G: - chipset = "Intel(R) 865G"; - break; - case PCI_CHIP_I915_G: - chipset = "Intel(R) 915G"; - break; - case PCI_CHIP_I915_GM: - chipset = "Intel(R) 915GM"; - break; - case PCI_CHIP_I945_G: - chipset = "Intel(R) 945G"; - break; - case PCI_CHIP_I945_GM: - chipset = "Intel(R) 945GM"; - break; - case PCI_CHIP_I945_GME: - chipset = "Intel(R) 945GME"; - break; - case PCI_CHIP_G33_G: - chipset = "Intel(R) G33"; - break; - case PCI_CHIP_Q35_G: - chipset = "Intel(R) Q35"; - break; - case PCI_CHIP_Q33_G: - chipset = "Intel(R) Q33"; - break; - default: - chipset = "Unknown Intel Chipset"; - break; - } - - (void) driGetRendererString(buffer, chipset, DRIVER_DATE, 0); - return (GLubyte *) buffer; - - default: - return NULL; - } -} - - -/** - * Extension strings exported by the intel driver. - * - * \note - * It appears that ARB_texture_env_crossbar has "disappeared" compared to the - * old i830-specific driver. - */ -const struct dri_extension card_extensions[] = { - {"GL_ARB_multisample", GL_ARB_multisample_functions}, - {"GL_ARB_multitexture", NULL}, - {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, - {"GL_ARB_texture_border_clamp", NULL}, - {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, - {"GL_ARB_texture_cube_map", NULL}, - {"GL_ARB_texture_env_add", NULL}, - {"GL_ARB_texture_env_combine", NULL}, - {"GL_ARB_texture_env_dot3", NULL}, - {"GL_ARB_texture_mirrored_repeat", NULL}, - {"GL_ARB_texture_rectangle", NULL}, - {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, - {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, - {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, - {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", - GL_EXT_blend_equation_separate_functions}, - {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, - {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_subtract", NULL}, - {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, - {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, -#if 1 /* XXX FBO temporary? */ - {"GL_EXT_packed_depth_stencil", NULL}, -#endif - {"GL_EXT_pixel_buffer_object", NULL}, - {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, - {"GL_EXT_stencil_wrap", NULL}, - {"GL_EXT_texture_edge_clamp", NULL}, - {"GL_EXT_texture_env_combine", NULL}, - {"GL_EXT_texture_env_dot3", NULL}, - {"GL_EXT_texture_filter_anisotropic", NULL}, - {"GL_EXT_texture_lod_bias", NULL}, - {"GL_3DFX_texture_compression_FXT1", NULL}, - {"GL_APPLE_client_storage", NULL}, - {"GL_MESA_pack_invert", NULL}, - {"GL_MESA_ycbcr_texture", NULL}, - {"GL_NV_blend_square", NULL}, - {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, - {"GL_NV_vertex_program1_1", NULL}, -/* { "GL_SGIS_generate_mipmap", NULL }, */ - {NULL, NULL} -}; - - - - -static const struct dri_debug_control debug_control[] = { - {"tex", DEBUG_TEXTURE}, - {"state", DEBUG_STATE}, - {"ioctl", DEBUG_IOCTL}, - {"blit", DEBUG_BLIT}, - {"mip", DEBUG_MIPTREE}, - {"fall", DEBUG_FALLBACKS}, - {"verb", DEBUG_VERBOSE}, - {"bat", DEBUG_BATCH}, - {"pix", DEBUG_PIXEL}, - {"buf", DEBUG_BUFMGR}, - {"reg", DEBUG_REGION}, - {"fbo", DEBUG_FBO}, - {"lock", DEBUG_LOCK}, - {NULL, 0} -}; - - -static void -intelInvalidateState(GLcontext * ctx, GLuint new_state) -{ - _vbo_InvalidateState(ctx, new_state); - _tnl_InvalidateState(ctx, new_state); - _tnl_invalidate_vertex_state(ctx, new_state); - - st_invalidate_state( ctx, new_state ); -} - - -void -intelFlush(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - - INTEL_FIREVERTICES(intel); - - if (intel->batch->map != intel->batch->ptr) - intel_batchbuffer_flush(intel->batch); - - /* XXX: Need to do an MI_FLUSH here. - */ -} - - -/** - * Check if we need to rotate/warp the front color buffer to the - * rotated screen. We generally need to do this when we get a glFlush - * or glFinish after drawing to the front color buffer. - * If no rotation, just copy the private fake front buffer to the real one. - */ -static void -intelCheckFrontUpdate(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - /* rely on _ColorDrawBufferMask being kept up to date by mesa - even for window-fbos. */ - /* not sure. Might need that for all masks including - BUFFER_BIT_FRONT_LEFT maybe? */ - if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - intelCopyBuffer(dPriv, NULL); - } -} - - -/** - * Called via glFlush. - */ -static void -intelglFlush(GLcontext * ctx) -{ - intelFlush(ctx); - intelCheckFrontUpdate(ctx); -} - -void -intelFinish(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - intelFlush(ctx); - if (intel->batch->last_fence) { - driFenceFinish(intel->batch->last_fence, - 0, GL_FALSE); - driFenceUnReference(intel->batch->last_fence); - intel->batch->last_fence = NULL; - } - intelCheckFrontUpdate(ctx); -} - - -static void -intelInitDriverFunctions(struct dd_function_table *functions) -{ - _mesa_init_driver_functions(functions); - - functions->Flush = intelglFlush; - functions->Finish = intelFinish; - functions->GetString = intelGetString; - functions->UpdateState = intelInvalidateState; - - /* - intelInitTextureFuncs(functions); - */ - intelInitBufferFuncs(functions); - - st_init_driver_functions(functions); -} - - - -GLboolean -intelCreateContext(const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate) -{ - struct dd_function_table functions; - - struct intel_context *intel = CALLOC_STRUCT(intel_context); - GLcontext *ctx = &intel->ctx; - - GLcontext *shareCtx = (GLcontext *) sharedContextPrivate; - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - drmI830Sarea *saPriv = intelScreen->sarea; - int fthrottle_mode; - GLboolean havePools; - - intelInitDriverFunctions(&functions); - - if (!_mesa_initialize_context(&intel->ctx, - mesaVis, shareCtx, - &functions, (void *) intel)) - return GL_FALSE; - - driContextPriv->driverPrivate = intel; - intel->intelScreen = intelScreen; - intel->driScreen = sPriv; - intel->sarea = saPriv; - - if (!lockMutexInit) { - lockMutexInit = GL_TRUE; - _glthread_INIT_MUTEX(lockMutex); - } - - driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, - intel->driScreen->myNum, "i915"); - - - /* Initialize the software rasterizer and helper modules. */ - _vbo_CreateContext(ctx); - _tnl_CreateContext(ctx); - - /* - * Pipe-related setup - */ - st_create_context( &intel->ctx, - intel_create_softpipe( intel ) ); - - /* KW: Not sure I like this - we should only be talking to the - * state_tracker. The pipe code will need some way of talking to - * us, eg for batchbuffer ioctls, and there will need to be a - * buffer manager interface. So, this is a temporary hack, right? - * BP: Yes, a temporary hack so we can make jumps between old/new code. - */ - intel->pipe = intel->ctx.st->pipe; -// intel->pipe->screen = intelScreen; -// intel->pipe->glctx = ctx; -// intel_init_region_functions(intel->pipe); - - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - intel->pipe->mipmap_tree_layout = i945_miptree_layout; - break; - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - case PCI_CHIP_I830_M: - case PCI_CHIP_I855_GM: - case PCI_CHIP_I865_G: - intel->pipe->mipmap_tree_layout = i915_miptree_layout; - default: - assert(0); /*FIX*/ - } - - - /* - * memory pools - */ - DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - havePools = intelCreatePools(intelScreen); - DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - if (!havePools) - return GL_FALSE; - - - /* Dri stuff */ - intel->hHWContext = driContextPriv->hHWContext; - intel->driFd = sPriv->fd; - intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; - - TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; - - - - fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->iw.irq_seq = -1; - intel->irqsEmitted = 0; - - /* Disable imaging extension until convolution is working in - * teximage paths: - */ - driInitExtensions(ctx, card_extensions, -/* GL_TRUE, */ - GL_FALSE); - - - intel->batch = intel_batchbuffer_alloc(intel); - intel->last_swap_fence = NULL; - intel->first_swap_fence = NULL; - - intel_fbo_init(intel); - - if (intel->ctx.Mesa_DXTn) { - _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); - _mesa_enable_extension(ctx, "GL_S3_s3tc"); - } - else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) { - _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); - } - -#if DO_DEBUG - INTEL_DEBUG = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); -#endif - - - return GL_TRUE; -} - -void -intelDestroyContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = - (struct intel_context *) driContextPriv->driverPrivate; - - assert(intel); /* should never be null */ - if (intel) { - GLboolean release_texture_heaps; - - INTEL_FIREVERTICES(intel); - - //intel->vtbl.destroy(intel); - - release_texture_heaps = (intel->ctx.Shared->RefCount == 1); - _tnl_DestroyContext(&intel->ctx); - _vbo_DestroyContext(&intel->ctx); - - intel_batchbuffer_free(intel->batch); - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - if (intel->first_swap_fence) { - driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->first_swap_fence); - intel->first_swap_fence = NULL; - } - - - if (release_texture_heaps) { - /* This share group is about to go away, free our private - * texture object data. - */ - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "do something to free texture heaps\n"); - } - - /* free the Mesa context */ - _mesa_free_context_data(&intel->ctx); - } -} - -GLboolean -intelUnbindContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; - /* XXX UnbindContext is called AFTER the new context is made current. - Hopefully shouldn't be a problem ? */ - FLUSH_VERTICES((&intel->ctx), 0); - intelFlush(&intel->ctx); - return GL_TRUE; -} - -GLboolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv) -{ - -#if 0 - if (driDrawPriv) { - fprintf(stderr, "x %d, y %d, width %d, height %d\n", - driDrawPriv->x, driDrawPriv->y, driDrawPriv->w, driDrawPriv->h); - } -#endif - - if (driContextPriv) { - struct intel_context *intel = - (struct intel_context *) driContextPriv->driverPrivate; - struct intel_framebuffer *intel_fb = - (struct intel_framebuffer *) driDrawPriv->driverPrivate; - GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; - - /* this is a hack so we have a valid context when the region allocation - is done. Need a per-screen context? */ - intel->intelScreen->dummyctxptr = intel; - - /* update GLframebuffer size to match window if needed */ - driUpdateFramebufferSize(&intel->ctx, driDrawPriv); - - if (driReadPriv != driDrawPriv) { - driUpdateFramebufferSize(&intel->ctx, driReadPriv); - } - - _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); - - /* The drawbuffer won't always be updated by _mesa_make_current: - */ - if (intel->ctx.DrawBuffer == &intel_fb->Base) { - - if (intel->driDrawable != driDrawPriv) { - if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) { - int i; - - intel_fb->vblank_flags = driGetDefaultVBlankFlags(&intel->optionCache); - - (*dri_interface->getUST) (&intel_fb->swap_ust); - driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags, - &intel_fb->vbl_seq); - intel_fb->vbl_waited = intel_fb->vbl_seq; - - for (i = 0; i < 2; i++) { - if (intel_fb->color_rb[i]) - intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_seq; - } - } - } - } - - if ((intel->driDrawable != driDrawPriv) || - (intel->lastStamp != driDrawPriv->lastStamp)) { - intel->driDrawable = driDrawPriv; - intelWindowMoved(intel); - intel->lastStamp = driDrawPriv->lastStamp; - } - - } - else { - _mesa_make_current(NULL, NULL, NULL); - } - - return GL_TRUE; -} - -static void -intelContendedLock(struct intel_context *intel, GLuint flags) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - __DRIscreenPrivate *sPriv = intel->driScreen; - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - drmI830Sarea *sarea = intel->sarea; - - drmGetLock(intel->driFd, intel->hHWContext, flags); - - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("%s - got contended lock\n", __progname); - - /* If the window moved, may need to set a new cliprect now. - * - * NOTE: This releases and regains the hw lock, so all state - * checking must be done *after* this call: - */ - if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - - if (sarea->width != intelScreen->front.width || - sarea->height != intelScreen->front.height) { - - intelUpdateScreenRotation(sPriv, sarea); - } - -#if 0 - if (sarea->width != intel->width || - sarea->height != intel->height || - sarea->rotation != intel->current_rotation) { - int numClipRects = intel->numClipRects; - - /* - * FIXME: Really only need to do this when drawing to a - * common back- or front buffer. - */ - - /* - * This will essentially drop the outstanding batchbuffer on the floor. - */ - intel->numClipRects = 0; - - INTEL_FIREVERTICES(intel); - - if (intel->batch->map != intel->batch->ptr) - intel_batchbuffer_flush(intel->batch); - - intel->numClipRects = numClipRects; - - /* force window update */ - intel->lastStamp = 0; - - intel->width = sarea->width; - intel->height = sarea->height; - intel->current_rotation = sarea->rotation; - } -#endif -} - - -/* Lock the hardware and validate our state. - */ -void LOCK_HARDWARE( struct intel_context *intel ) -{ - char __ret=0; - struct intel_framebuffer *intel_fb = NULL; - struct intel_renderbuffer *intel_rb = NULL; - _glthread_LOCK_MUTEX(lockMutex); - assert(!intel->locked); - - if (intel->driDrawable) { - intel_fb = intel->driDrawable->driverPrivate; - - if (intel_fb) - intel_rb = - intel_get_renderbuffer(&intel_fb->Base, - intel_fb->Base._ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : - BUFFER_BACK_LEFT); - } - - if (intel_rb && intel_fb->vblank_flags && - !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && - (intel_fb->vbl_waited - intel_rb->vbl_pending) > (1<<23)) { - drmVBlank vbl; - - vbl.request.type = DRM_VBLANK_ABSOLUTE; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - vbl.request.type |= DRM_VBLANK_SECONDARY; - } - - vbl.request.sequence = intel_rb->vbl_pending; - drmWaitVBlank(intel->driFd, &vbl); - intel_fb->vbl_waited = vbl.reply.sequence; - } - - DRM_CAS(intel->driHwLock, intel->hHWContext, - (DRM_LOCK_HELD|intel->hHWContext), __ret); - - if (__ret) - intelContendedLock( intel, 0 ); - - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("%s - locked\n", __progname); - - intel->locked = 1; -} - - - /* Unlock the hardware using the global current context - */ -void UNLOCK_HARDWARE( struct intel_context *intel ) -{ - assert(intel->locked); - intel->locked = 0; - - DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); - - _glthread_UNLOCK_MUTEX(lockMutex); - - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("%s - unlocked\n", __progname); -} - diff --git a/src/mesa/drivers/dri/i915pipe/intel_context.h b/src/mesa/drivers/dri/i915pipe/intel_context.h deleted file mode 100644 index 1a4e7ca1cd..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_context.h +++ /dev/null @@ -1,225 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTELCONTEXT_INC -#define INTELCONTEXT_INC - - - -#include "mtypes.h" -#include "drm.h" -#include "mm.h" -#include "texmem.h" - -#include "intel_screen.h" -#include "i915_drm.h" -#include "i830_common.h" - - -#define DV_PF_555 (1<<8) -#define DV_PF_565 (2<<8) -#define DV_PF_8888 (3<<8) - -struct pipe_context; -struct pipe_region; -struct intel_context; -struct _DriBufferObject; - - -#define INTEL_WRITE_PART 0x1 -#define INTEL_WRITE_FULL 0x2 -#define INTEL_READ 0x4 - - - -#define INTEL_MAX_FIXUP 64 - -struct intel_context -{ - GLcontext ctx; /* the parent class */ - - struct pipe_context *pipe; - - GLint refcount; - - struct _DriFenceObject *last_swap_fence; - struct _DriFenceObject *first_swap_fence; - - struct intel_batchbuffer *batch; - - - GLboolean locked; - char *prevLockFile; - int prevLockLine; - - GLuint ClearColor565; - GLuint ClearColor8888; - - - /* These refer to the current drawing buffer: - */ - GLuint numClipRects; /**< cliprects for drawing */ - drm_clip_rect_t *pClipRects; - drm_clip_rect_t fboRect; /**< cliprect for rendering */ - - GLuint irqsEmitted; - drm_i915_irq_wait_t iw; - - drm_context_t hHWContext; - drmLock *driHwLock; - int driFd; - - __DRIdrawablePrivate *driDrawable; - __DRIscreenPrivate *driScreen; - intelScreenPrivate *intelScreen; - drmI830Sarea *sarea; - - GLuint lastStamp; - - /** - * Configuration cache - */ - driOptionCache optionCache; -}; - -/* These are functions now: - */ -void LOCK_HARDWARE( struct intel_context *intel ); -void UNLOCK_HARDWARE( struct intel_context *intel ); - -extern char *__progname; - -/* Will become a call into state_tracker: - */ -#define INTEL_FIREVERTICES(intel) - -/* ================================================================ - * Color packing: - */ - -#define INTEL_PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define INTEL_PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define INTEL_PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define INTEL_PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - - -/* ================================================================ - * Debugging: - */ -#define DO_DEBUG 0 -#if DO_DEBUG -extern int INTEL_DEBUG; -#else -#define INTEL_DEBUG 0 -#endif - -#define DEBUG_TEXTURE 0x1 -#define DEBUG_STATE 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_BLIT 0x8 -#define DEBUG_MIPTREE 0x10 -#define DEBUG_FALLBACKS 0x20 -#define DEBUG_VERBOSE 0x40 -#define DEBUG_BATCH 0x80 -#define DEBUG_PIXEL 0x100 -#define DEBUG_BUFMGR 0x200 -#define DEBUG_REGION 0x400 -#define DEBUG_FBO 0x800 -#define DEBUG_LOCK 0x1000 - -#define DBG(...) do { if (INTEL_DEBUG & FILE_DEBUG_FLAG) _mesa_printf(__VA_ARGS__); } while(0) - - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GME 0x27AE -#define PCI_CHIP_G33_G 0x29C2 -#define PCI_CHIP_Q35_G 0x29B2 -#define PCI_CHIP_Q33_G 0x29D2 - - -/* ================================================================ - * intel_context.c: - */ - -extern GLboolean intelInitContext(struct intel_context *intel, - const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate, - struct dd_function_table *functions); - -extern void intelGetLock(struct intel_context *intel, GLuint flags); - -extern void intelFinish(GLcontext * ctx); -extern void intelFlush(GLcontext * ctx); - - -/* ================================================================ - * intel_softpipe.c: - */ - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel ); - - -#define MI_BATCH_BUFFER_END (0xA<<23) - -#define FALLBACK( ctx, bit, mode ) - -/*====================================================================== - * Inline conversion functions. - * These are better-typed than the macros used previously: - */ -static INLINE struct intel_context * -intel_context(GLcontext * ctx) -{ - return (struct intel_context *) ctx; -} - -extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer - *rb); - -extern void intel_init_region_functions(struct pipe_context *pipe); - - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c b/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c deleted file mode 100644 index 81e09698cc..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_depthstencil.c +++ /dev/null @@ -1,282 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "imports.h" -#include "context.h" -#include "depthstencil.h" -#include "fbobject.h" -#include "framebuffer.h" -#include "hash.h" -#include "mtypes.h" -#include "renderbuffer.h" - -#include "intel_context.h" -#include "intel_fbo.h" -#include "intel_depthstencil.h" - -#include "pipe/p_context.h" - -/** - * The GL_EXT_framebuffer_object allows the user to create their own - * framebuffer objects consisting of color renderbuffers (0 or more), - * depth renderbuffers (0 or 1) and stencil renderbuffers (0 or 1). - * - * The spec considers depth and stencil renderbuffers to be totally independent - * buffers. In reality, most graphics hardware today uses a combined - * depth+stencil buffer (one 32-bit pixel = 24 bits of Z + 8 bits of stencil). - * - * This causes difficulty because the user may create some number of depth - * renderbuffers and some number of stencil renderbuffers and bind them - * together in framebuffers in any combination. - * - * This code manages all that. - * - * 1. Depth renderbuffers are always allocated in hardware as 32bpp - * GL_DEPTH24_STENCIL8 buffers. - * - * 2. Stencil renderbuffers are initially allocated in software as 8bpp - * GL_STENCIL_INDEX8 buffers. - * - * 3. Depth and Stencil renderbuffers use the PairedStencil and PairedDepth - * fields (respectively) to indicate if the buffer's currently paired - * with another stencil or depth buffer (respectively). - * - * 4. When a depth and stencil buffer are initially both attached to the - * current framebuffer, we merge the stencil buffer values into the - * depth buffer (really a depth+stencil buffer). The then hardware uses - * the combined buffer. - * - * 5. Whenever a depth or stencil buffer is reallocated (with - * glRenderbufferStorage) we undo the pairing and copy the stencil values - * from the combined depth/stencil buffer back to the stencil-only buffer. - * - * 6. We also undo the pairing when we find a change in buffer bindings. - * - * 7. If a framebuffer is only using a depth renderbuffer (no stencil), we - * just use the combined depth/stencil buffer and ignore the stencil values. - * - * 8. If a framebuffer is only using a stencil renderbuffer (no depth) we have - * to promote the 8bpp software stencil buffer to a 32bpp hardware - * depth+stencil buffer. - * - */ - - - -static void -map_regions(GLcontext * ctx, - struct intel_renderbuffer *depthRb, - struct intel_renderbuffer *stencilRb) -{ - struct intel_context *intel = intel_context(ctx); - if (depthRb && depthRb->region) { - intel->pipe->region_map(intel->pipe, depthRb->region); - depthRb->pfMap = depthRb->region->map; - depthRb->pfPitch = depthRb->region->pitch; - } - if (stencilRb && stencilRb->region) { - intel->pipe->region_map(intel->pipe, stencilRb->region); - stencilRb->pfMap = stencilRb->region->map; - stencilRb->pfPitch = stencilRb->region->pitch; - } -} - -static void -unmap_regions(GLcontext * ctx, - struct intel_renderbuffer *depthRb, - struct intel_renderbuffer *stencilRb) -{ - struct intel_context *intel = intel_context(ctx); - if (depthRb && depthRb->region) { - intel->pipe->region_unmap(intel->pipe, depthRb->region); - depthRb->pfMap = NULL; - depthRb->pfPitch = 0; - } - if (stencilRb && stencilRb->region) { - intel->pipe->region_unmap(intel->pipe, stencilRb->region); - stencilRb->pfMap = NULL; - stencilRb->pfPitch = 0; - } -} - - - -/** - * Undo the pairing/interleaving between depth and stencil buffers. - * irb should be a depth/stencil or stencil renderbuffer. - */ -void -intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb) -{ - if (irb->PairedStencil) { - /* irb is a depth/stencil buffer */ - struct gl_renderbuffer *stencilRb; - struct intel_renderbuffer *stencilIrb; - - ASSERT(irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - - stencilRb = _mesa_lookup_renderbuffer(ctx, irb->PairedStencil); - stencilIrb = intel_renderbuffer(stencilRb); - if (stencilIrb) { - /* need to extract stencil values from the depth buffer */ - ASSERT(stencilIrb->PairedDepth == irb->Base.Name); - map_regions(ctx, irb, stencilIrb); - _mesa_extract_stencil(ctx, &irb->Base, &stencilIrb->Base); - unmap_regions(ctx, irb, stencilIrb); - stencilIrb->PairedDepth = 0; - } - irb->PairedStencil = 0; - } - else if (irb->PairedDepth) { - /* irb is a stencil buffer */ - struct gl_renderbuffer *depthRb; - struct intel_renderbuffer *depthIrb; - - ASSERT(irb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || - irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - - depthRb = _mesa_lookup_renderbuffer(ctx, irb->PairedDepth); - depthIrb = intel_renderbuffer(depthRb); - if (depthIrb) { - /* need to extract stencil values from the depth buffer */ - ASSERT(depthIrb->PairedStencil == irb->Base.Name); - map_regions(ctx, depthIrb, irb); - _mesa_extract_stencil(ctx, &depthIrb->Base, &irb->Base); - unmap_regions(ctx, depthIrb, irb); - depthIrb->PairedStencil = 0; - } - irb->PairedDepth = 0; - } - else { - _mesa_problem(ctx, "Problem in undo_depth_stencil_pairing"); - } - - ASSERT(irb->PairedStencil == 0); - ASSERT(irb->PairedDepth == 0); -} - - -/** - * Examine the depth and stencil renderbuffers which are attached to the - * framebuffer. If both depth and stencil are attached, make sure that the - * renderbuffers are 'paired' (combined). If only depth or only stencil is - * attached, undo any previous pairing. - * - * Must be called if NewState & _NEW_BUFFER (when renderbuffer attachments - * change, for example). - */ -void -intel_validate_paired_depth_stencil(GLcontext * ctx, - struct gl_framebuffer *fb) -{ - struct intel_renderbuffer *depthRb, *stencilRb; - - depthRb = intel_get_renderbuffer(fb, BUFFER_DEPTH); - stencilRb = intel_get_renderbuffer(fb, BUFFER_STENCIL); - - if (depthRb && stencilRb) { - if (depthRb == stencilRb) { - /* Using a user-created combined depth/stencil buffer. - * Nothing to do. - */ - ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_STENCIL_EXT); - ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - } - else { - /* Separate depth/stencil buffers, need to interleave now */ - ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_COMPONENT); - ASSERT(stencilRb->Base._BaseFormat == GL_STENCIL_INDEX); - /* may need to interleave depth/stencil now */ - if (depthRb->PairedStencil == stencilRb->Base.Name) { - /* OK, the depth and stencil buffers are already interleaved */ - ASSERT(stencilRb->PairedDepth == depthRb->Base.Name); - } - else { - /* need to setup new pairing/interleaving */ - if (depthRb->PairedStencil) { - intel_unpair_depth_stencil(ctx, depthRb); - } - if (stencilRb->PairedDepth) { - intel_unpair_depth_stencil(ctx, stencilRb); - } - - ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - ASSERT(stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || - stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - - /* establish new pairing: interleave stencil into depth buffer */ - map_regions(ctx, depthRb, stencilRb); - _mesa_insert_stencil(ctx, &depthRb->Base, &stencilRb->Base); - unmap_regions(ctx, depthRb, stencilRb); - depthRb->PairedStencil = stencilRb->Base.Name; - stencilRb->PairedDepth = depthRb->Base.Name; - } - - } - } - else if (depthRb) { - /* Depth buffer but no stencil buffer. - * We'll use a GL_DEPTH24_STENCIL8 buffer and ignore the stencil bits. - */ - /* can't assert this until storage is allocated: - ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - */ - /* intel_undo any previous pairing */ - if (depthRb->PairedStencil) { - intel_unpair_depth_stencil(ctx, depthRb); - } - } - else if (stencilRb) { - /* Stencil buffer but no depth buffer. - * Since h/w doesn't typically support just 8bpp stencil w/out Z, - * we'll use a GL_DEPTH24_STENCIL8 buffer and ignore the depth bits. - */ - /* undo any previous pairing */ - if (stencilRb->PairedDepth) { - intel_unpair_depth_stencil(ctx, stencilRb); - } - if (stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT) { - /* promote buffer to GL_DEPTH24_STENCIL8 for hw rendering */ - _mesa_promote_stencil(ctx, &stencilRb->Base); - ASSERT(stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - } - } - - /* Finally, update the fb->_DepthBuffer and fb->_StencilBuffer fields */ - _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH); - if (depthRb && depthRb->PairedStencil) - _mesa_update_stencil_buffer(ctx, fb, BUFFER_DEPTH); - else - _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL); - - - /* The hardware should use fb->Attachment[BUFFER_DEPTH].Renderbuffer - * first, if present, then fb->Attachment[BUFFER_STENCIL].Renderbuffer - * if present. - */ -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_depthstencil.h b/src/mesa/drivers/dri/i915pipe/intel_depthstencil.h deleted file mode 100644 index 2d3fc48b3a..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_depthstencil.h +++ /dev/null @@ -1,14 +0,0 @@ - -#ifndef INTEL_DEPTH_STENCIL_H -#define INTEL_DEPTH_STENCIL_H - - -extern void -intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb); - -extern void -intel_validate_paired_depth_stencil(GLcontext * ctx, - struct gl_framebuffer *fb); - - -#endif /* INTEL_DEPTH_STENCIL_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.c b/src/mesa/drivers/dri/i915pipe/intel_fbo.c deleted file mode 100644 index 0b6c6a94ee..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.c +++ /dev/null @@ -1,629 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "imports.h" -#include "mtypes.h" -#include "fbobject.h" -#include "framebuffer.h" -#include "renderbuffer.h" -#include "context.h" -#include "texformat.h" -#include "texrender.h" - -#include "intel_context.h" -#include "intel_buffers.h" -#include "intel_depthstencil.h" -#include "intel_fbo.h" -#include "state_tracker/st_mipmap_tree.h" -/*#include "intel_tex.h"*/ - -#include "pipe/p_context.h" - -#define FILE_DEBUG_FLAG DEBUG_FBO - -#define INTEL_RB_CLASS 0x12345678 - - -/* XXX FBO: move this to intel_context.h (inlined) */ -/** - * Return a gl_renderbuffer ptr casted to intel_renderbuffer. - * NULL will be returned if the rb isn't really an intel_renderbuffer. - * This is determiend by checking the ClassID. - */ -struct intel_renderbuffer * -intel_renderbuffer(struct gl_renderbuffer *rb) -{ - struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; - if (irb && irb->Base.ClassID == INTEL_RB_CLASS) { - /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/ - return irb; - } - else - return NULL; -} - - -struct intel_renderbuffer * -intel_get_renderbuffer(struct gl_framebuffer *fb, GLuint attIndex) -{ - return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer); -} - - -void -intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) -{ - int current_page = intel_fb->pf_current_page; - int next_page = (current_page + 1) % intel_fb->pf_num_pages; - struct gl_renderbuffer *tmp_rb; - - /* Exchange renderbuffers if necessary but make sure their reference counts - * are preserved. - */ - if (intel_fb->color_rb[current_page] && - intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer != - &intel_fb->color_rb[current_page]->Base) { - tmp_rb = NULL; - _mesa_reference_renderbuffer(&tmp_rb, - intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); - tmp_rb = &intel_fb->color_rb[current_page]->Base; - _mesa_reference_renderbuffer( - &intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer, tmp_rb); - _mesa_reference_renderbuffer(&tmp_rb, NULL); - } - - if (intel_fb->color_rb[next_page] && - intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer != - &intel_fb->color_rb[next_page]->Base) { - tmp_rb = NULL; - _mesa_reference_renderbuffer(&tmp_rb, - intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); - tmp_rb = &intel_fb->color_rb[next_page]->Base; - _mesa_reference_renderbuffer( - &intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer, tmp_rb); - _mesa_reference_renderbuffer(&tmp_rb, NULL); - } -} - - -struct pipe_region * -intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) -{ - struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex); - - if (irb) - return irb->region; - else - return NULL; -} - - - -/** - * Create a new framebuffer object. - */ -static struct gl_framebuffer * -intel_new_framebuffer(GLcontext * ctx, GLuint name) -{ - /* Only drawable state in intel_framebuffer at this time, just use Mesa's - * class - */ - return _mesa_new_framebuffer(ctx, name); -} - - -static void -intel_delete_renderbuffer(struct gl_renderbuffer *rb) -{ - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); - struct intel_renderbuffer *irb = intel_renderbuffer(rb); - - ASSERT(irb); - - DBG("freeing renderbuffer\n"); - - if (irb->PairedStencil || irb->PairedDepth) { - intel_unpair_depth_stencil(ctx, irb); - } - - if (intel && irb->region) { - intel->pipe->region_release(intel->pipe, &irb->region); - } - - _mesa_free(irb); -} - - - -/** - * Return a pointer to a specific pixel in a renderbuffer. - */ -static void * -intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb, - GLint x, GLint y) -{ - /* By returning NULL we force all software rendering to go through - * the span routines. - */ - return NULL; -} - - - -/** - * Called via glRenderbufferStorageEXT() to set the format and allocate - * storage for a user-created (or priv buffer) renderbuffer. - */ -static GLboolean -intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, - GLuint width, GLuint height) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_renderbuffer *irb = intel_renderbuffer(rb); - GLboolean softwareBuffer = GL_FALSE; - int cpp; - - switch (internalFormat) { - case GL_R3_G3_B2: - case GL_RGB4: - case GL_RGB5: - rb->_ActualFormat = GL_RGB5; - rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 5; - rb->GreenBits = 6; - rb->BlueBits = 5; - cpp = 2; - break; - case GL_RGB: - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - case GL_RGBA: - case GL_RGBA2: - case GL_RGBA4: - case GL_RGB5_A1: - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - rb->_ActualFormat = GL_RGBA8; - rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 8; - rb->GreenBits = 8; - rb->BlueBits = 8; - rb->AlphaBits = 8; - cpp = 4; - break; - case GL_STENCIL_INDEX: - case GL_STENCIL_INDEX1_EXT: - case GL_STENCIL_INDEX4_EXT: - case GL_STENCIL_INDEX8_EXT: - case GL_STENCIL_INDEX16_EXT: - /* alloc a depth+stencil buffer */ - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->StencilBits = 8; - cpp = 4; - break; - case GL_DEPTH_COMPONENT16: - rb->_ActualFormat = GL_DEPTH_COMPONENT16; - rb->DataType = GL_UNSIGNED_SHORT; - rb->DepthBits = 16; - cpp = 2; - break; - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - cpp = 4; - break; - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - rb->StencilBits = 8; - cpp = 4; - break; - default: - _mesa_problem(ctx, - "Unexpected format (%x) in intel_alloc_renderbuffer_storage", internalFormat); - return GL_FALSE; - } - - intelFlush(ctx); - - /* free old region */ - if (irb->region) { - intel->pipe->region_release(intel->pipe, &irb->region); - } - - /* allocate new memory region/renderbuffer */ - if (softwareBuffer) { - return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat, - width, height); - } - else { - /* Choose a pitch to match hardware requirements: - */ - GLuint pitch = ((cpp * width + 63) & ~63) / cpp; - - /* alloc hardware renderbuffer */ - DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, - height, pitch); - - irb->region = intel->pipe->region_alloc(intel->pipe, cpp, pitch, height); - if (!irb->region) - return GL_FALSE; /* out of memory? */ - - ASSERT(irb->region->buffer); - - rb->Width = width; - rb->Height = height; - - /* update the surface's size too */ - rb->surface->width = width; - rb->surface->height = height; - rb->surface->region = irb->region; - - /* This sets the Get/PutRow/Value functions */ - // intel_set_span_functions(&irb->Base); - - return GL_TRUE; - } -} - - -static void -intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, - GLuint width, GLuint height) -{ - struct intel_framebuffer *intel_fb = (struct intel_framebuffer*)fb; - int i; - - _mesa_resize_framebuffer(ctx, fb, width, height); - - fb->Initialized = GL_TRUE; /* XXX remove someday */ - - if (fb->Name != 0) { - return; - } - - /* Make sure all window system renderbuffers are up to date */ - for (i = 0; i < 3; i++) { - struct gl_renderbuffer *rb = &intel_fb->color_rb[i]->Base; - - /* only resize if size is changing */ - if (rb && (rb->Width != width || rb->Height != height)) { - rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height); - } - } -} - -static GLboolean -intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, GLuint width, GLuint height) -{ - _mesa_problem(ctx, "intel_op_alloc_storage should never be called."); - return GL_FALSE; -} - - - -struct intel_renderbuffer * -intel_new_renderbuffer_fb(GLuint intFormat) -{ - struct intel_renderbuffer *irb; - - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, 0); - irb->Base.ClassID = INTEL_RB_CLASS; - irb->Base.InternalFormat = intFormat; - - switch (intFormat) { - case GL_RGB5: - case GL_RGBA8: - irb->Base._BaseFormat = GL_RGBA; - break; - case GL_DEPTH_COMPONENT16: - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - break; - case GL_DEPTH24_STENCIL8_EXT: - irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; - break; - default: - _mesa_problem(NULL, - "Unexpected intFormat in intel_create_renderbuffer"); - return NULL; - } - - /* intel-specific methods */ - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; - irb->Base.GetPointer = intel_get_pointer; - /* span routines set in alloc_storage function */ - - irb->Base.surface = intel_new_surface(intFormat); - irb->Base.surface->rb = irb; - - return irb; -} - -/** - * Create a new renderbuffer object. - * Typically called via glBindRenderbufferEXT(). - */ -static struct gl_renderbuffer * -intel_new_renderbuffer(GLcontext * ctx, GLuint name) -{ - /*struct intel_context *intel = intel_context(ctx); */ - struct intel_renderbuffer *irb; - - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, name); - irb->Base.ClassID = INTEL_RB_CLASS; - - /* intel-specific methods */ - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; - irb->Base.GetPointer = intel_get_pointer; - /* span routines set in alloc_storage function */ - - irb->Base.surface = intel_new_surface(0 /*unknown format*/); - irb->Base.surface->rb = irb; - - return &irb->Base; -} - - -/** - * Called via glBindFramebufferEXT(). - */ -static void -intel_bind_framebuffer(GLcontext * ctx, GLenum target, - struct gl_framebuffer *fb, struct gl_framebuffer *fbread) -{ - if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) { - intel_draw_buffer(ctx, fb); - /* Integer depth range depends on depth buffer bits */ - /* XXX - */ -// ctx->Driver.DepthRange(ctx, ctx->Viewport.Near, ctx->Viewport.Far); - } - else { - /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */ - } -} - - -/** - * Called via glFramebufferRenderbufferEXT(). - */ -static void -intel_framebuffer_renderbuffer(GLcontext * ctx, - struct gl_framebuffer *fb, - GLenum attachment, struct gl_renderbuffer *rb) -{ - DBG("Intel FramebufferRenderbuffer %u %u\n", fb->Name, rb ? rb->Name : 0); - - intelFlush(ctx); - - _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb); - intel_draw_buffer(ctx, fb); -} - - -/** - * When glFramebufferTexture[123]D is called this function sets up the - * gl_renderbuffer wrapper around the texture image. - * This will have the region info needed for hardware rendering. - */ -static struct intel_renderbuffer * -intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage) -{ - const GLuint name = ~0; /* not significant, but distinct for debugging */ - struct intel_renderbuffer *irb; - - /* make an intel_renderbuffer to wrap the texture image */ - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, name); - irb->Base.ClassID = INTEL_RB_CLASS; - - if (texImage->TexFormat == &_mesa_texformat_argb8888) { - irb->Base._ActualFormat = GL_RGBA8; - irb->Base._BaseFormat = GL_RGBA; - DBG("Render to RGBA8 texture OK\n"); - } - else if (texImage->TexFormat == &_mesa_texformat_rgb565) { - irb->Base._ActualFormat = GL_RGB5; - irb->Base._BaseFormat = GL_RGB; - DBG("Render to RGB5 texture OK\n"); - } - else if (texImage->TexFormat == &_mesa_texformat_z16) { - irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - DBG("Render to DEPTH16 texture OK\n"); - } - else { - DBG("Render to texture BAD FORMAT %d\n", - texImage->TexFormat->MesaFormat); - _mesa_free(irb); - return NULL; - } - - irb->Base.InternalFormat = irb->Base._ActualFormat; - irb->Base.Width = texImage->Width; - irb->Base.Height = texImage->Height; - irb->Base.DataType = GL_UNSIGNED_BYTE; /* FBO XXX fix */ - irb->Base.RedBits = texImage->TexFormat->RedBits; - irb->Base.GreenBits = texImage->TexFormat->GreenBits; - irb->Base.BlueBits = texImage->TexFormat->BlueBits; - irb->Base.AlphaBits = texImage->TexFormat->AlphaBits; - irb->Base.DepthBits = texImage->TexFormat->DepthBits; - - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_nop_alloc_storage; - -// intel_set_span_functions(&irb->Base); - - irb->RenderToTexture = GL_TRUE; - - return irb; -} - - -/** - * Called by glFramebufferTexture[123]DEXT() (and other places) to - * prepare for rendering into texture memory. This might be called - * many times to choose different texture levels, cube faces, etc - * before intel_finish_render_texture() is ever called. - */ -static void -intel_render_texture(GLcontext * ctx, - struct gl_framebuffer *fb, - struct gl_renderbuffer_attachment *att) -{ -#if 0 - struct intel_context *intel = intel_context(ctx); - struct gl_texture_image *newImage - = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; - struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); - struct st_texture_image *st_image; - GLuint imageOffset; - - (void) fb; - - ASSERT(newImage); - - if (!irb) { - irb = intel_wrap_texture(ctx, newImage); - if (irb) { - /* bind the wrapper to the attachment point */ - _mesa_reference_renderbuffer(&att->Renderbuffer, &irb->Base); - } - else { - /* fallback to software rendering */ - _mesa_render_texture(ctx, fb, att); - return; - } - } - - DBG("Begin render texture tid %x tex=%u w=%d h=%d refcount=%d\n", - _glthread_GetID(), - att->Texture->Name, newImage->Width, newImage->Height, - irb->Base.RefCount); - - /* point the renderbufer's region to the texture image region */ - st_image = st_texture_image(newImage); - if (irb->region != st_image->mt->region) { - if (irb->region) - intel->pipe->region_release(intel->pipe, &irb->region); - pipe_region_reference(&irb->region, st_image->mt->region); - } - - /* compute offset of the particular 2D image within the texture region */ - imageOffset = st_miptree_image_offset(st_image->mt, - att->CubeMapFace, - att->TextureLevel); - - if (att->Texture->Target == GL_TEXTURE_3D) { - const GLuint *offsets = st_miptree_depth_offsets(st_image->mt, - att->TextureLevel); - imageOffset += offsets[att->Zoffset]; - } - - /* store that offset in the region */ -#if 0 - st_image->mt->region->draw_offset = imageOffset; -#endif - - /* update drawing region, etc */ - intel_draw_buffer(ctx, fb); -#endif -} - - -/** - * Called by Mesa when rendering to a texture is done. - */ -static void -intel_finish_render_texture(GLcontext * ctx, - struct gl_renderbuffer_attachment *att) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); - - DBG("End render texture (tid %x) tex %u\n", _glthread_GetID(), att->Texture->Name); - - if (irb) { - /* just release the region */ - intel->pipe->region_release(intel->pipe, &irb->region); - } - else if (att->Renderbuffer) { - /* software fallback */ - _mesa_finish_render_texture(ctx, att); - /* XXX FBO: Need to unmap the buffer (or in intelSpanRenderStart???) */ - } -} - - -/** - * Do one-time context initializations related to GL_EXT_framebuffer_object. - * Hook in device driver functions. - */ -void -intel_fbo_init(struct intel_context *intel) -{ - intel->ctx.Driver.NewFramebuffer = intel_new_framebuffer; - intel->ctx.Driver.NewRenderbuffer = intel_new_renderbuffer; - intel->ctx.Driver.BindFramebuffer = intel_bind_framebuffer; - intel->ctx.Driver.FramebufferRenderbuffer = intel_framebuffer_renderbuffer; - intel->ctx.Driver.RenderTexture = intel_render_texture; - intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture; - intel->ctx.Driver.ResizeBuffers = intel_resize_buffers; -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_fbo.h b/src/mesa/drivers/dri/i915pipe/intel_fbo.h deleted file mode 100644 index 0f99a3e98d..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_fbo.h +++ /dev/null @@ -1,127 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_FBO_H -#define INTEL_FBO_H - - -#include "pipe/p_state.h" -#include "pipe/softpipe/sp_surface.h" - - -struct intel_context; -struct pipe_region; - - -/** - * Intel framebuffer, derived from gl_framebuffer. - */ -struct intel_framebuffer -{ - struct gl_framebuffer Base; - - struct intel_renderbuffer *color_rb[3]; - - /* Drawable page flipping state */ - GLboolean pf_active; - GLuint pf_seq; - GLint pf_pipes; - GLint pf_current_page; - GLint pf_num_pages; - - /* VBI - */ - GLuint vbl_seq; - GLuint vblank_flags; - GLuint vbl_waited; - - int64_t swap_ust; - int64_t swap_missed_ust; - - GLuint swap_count; - GLuint swap_missed_count; -}; - - -/** - * Intel renderbuffer, derived from gl_renderbuffer. - * Note: The PairedDepth and PairedStencil fields use renderbuffer IDs, - * not pointers because in some circumstances a deleted renderbuffer could - * result in a dangling pointer here. - */ -struct intel_renderbuffer -{ - struct gl_renderbuffer Base; - struct pipe_region *region; - void *pfMap; /* possibly paged flipped map pointer */ - GLuint pfPitch; /* possibly paged flipped pitch */ - GLboolean RenderToTexture; /* RTT? */ - - GLuint PairedDepth; /**< only used if this is a depth renderbuffer */ - GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */ - - GLuint pf_pending; /**< sequence number of pending flip */ - - GLuint vbl_pending; /**< vblank sequence number of pending flip */ - - struct intel_surface *surface; -}; - -#if 0 -extern struct intel_renderbuffer *intel_create_renderbuffer(GLenum intFormat, - GLsizei width, - GLsizei height, - int offset, - int pitch, - int cpp, - void *map); -#endif - -extern struct intel_renderbuffer *intel_new_renderbuffer_fb(GLuint intFormat); - -extern void intel_fbo_init(struct intel_context *intel); - - -/* XXX make inline or macro */ -extern struct intel_renderbuffer *intel_get_renderbuffer(struct gl_framebuffer - *fb, - GLuint attIndex); - -extern void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb); - - -/* XXX make inline or macro */ -extern struct pipe_region *intel_get_rb_region(struct gl_framebuffer *fb, - GLuint attIndex); - - - -extern struct pipe_surface * -intel_new_surface(GLuint intFormat); - - -#endif /* INTEL_FBO_H */ diff --git a/src/mesa/drivers/dri/i915pipe/intel_ioctl.c b/src/mesa/drivers/dri/i915pipe/intel_ioctl.c deleted file mode 100644 index 154a3e289f..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_ioctl.c +++ /dev/null @@ -1,135 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include -#include -#include -#include - -#include "mtypes.h" -#include "context.h" - -#include "intel_context.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "drm.h" - -#define FILE_DEBUG_FLAG DEBUG_IOCTL - -int -intelEmitIrqLocked(struct intel_context *intel) -{ - drmI830IrqEmit ie; - int ret, seq; - - assert(((*(int *) intel->driHwLock) & ~DRM_LOCK_CONT) == - (DRM_LOCK_HELD | intel->hHWContext)); - - ie.irq_seq = &seq; - - ret = drmCommandWriteRead(intel->driFd, DRM_I830_IRQ_EMIT, - &ie, sizeof(ie)); - if (ret) { - fprintf(stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret); - exit(1); - } - - DBG("%s --> %d\n", __FUNCTION__, seq); - - return seq; -} - -void -intelWaitIrq(struct intel_context *intel, int seq) -{ - int ret; - - DBG("%s %d\n", __FUNCTION__, seq); - - intel->iw.irq_seq = seq; - - do { - ret = - drmCommandWrite(intel->driFd, DRM_I830_IRQ_WAIT, &intel->iw, - sizeof(intel->iw)); - } while (ret == -EAGAIN || ret == -EINTR); - - if (ret) { - fprintf(stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret); - exit(1); - } -} - - -void -intel_batch_ioctl(struct intel_context *intel, - GLuint start_offset, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) -{ - drmI830BatchBuffer batch; - - assert(intel->locked); - assert(used); - - DBG("%s used %d offset %x..%x ignore_cliprects %d\n", - __FUNCTION__, - used, start_offset, start_offset + used, ignore_cliprects); - - /* Throw away non-effective packets. Won't work once we have - * hardware contexts which would preserve statechanges beyond a - * single buffer. - */ - - - - batch.start = start_offset; - batch.used = used; - batch.cliprects = intel->pClipRects; - batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; - batch.DR1 = 0; - batch.DR4 = 0; /* still need this ? */ - - DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n", - __FUNCTION__, - batch.start, - batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); - - if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, - sizeof(batch))) { - fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n", -errno); - UNLOCK_HARDWARE(intel); - exit(1); - } - - /* FIXME: use hardware contexts to avoid 'losing' hardware after - * each buffer flush. - */ - //intel->vtbl.lost_hardware(intel); -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_ioctl.h b/src/mesa/drivers/dri/i915pipe/intel_ioctl.h deleted file mode 100644 index e8d07de893..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_ioctl.h +++ /dev/null @@ -1,40 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_IOCTL_H -#define INTEL_IOCTL_H - -#include "intel_context.h" - -void intelWaitIrq(struct intel_context *intel, int seq); -int intelEmitIrqLocked(struct intel_context *intel); - -void intel_batch_ioctl(struct intel_context *intel, - GLuint start_offset, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock); -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c deleted file mode 100644 index 1a9aa10115..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.c +++ /dev/null @@ -1,293 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_mipmap_tree.h" -#include "enums.h" - -#include "pipe/p_state.h" -#include "pipe/p_context.h" - - -#define DBG if(0) printf - -static GLenum -target_to_target(GLenum target) -{ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - return GL_TEXTURE_CUBE_MAP_ARB; - default: - return target; - } -} - -struct pipe_mipmap_tree * -st_miptree_create(struct pipe_context *pipe, - GLenum target, - GLenum internal_format, - GLuint first_level, - GLuint last_level, - GLuint width0, - GLuint height0, - GLuint depth0, GLuint cpp, GLuint compress_byte) -{ - GLboolean ok; - struct pipe_mipmap_tree *mt = calloc(sizeof(*mt), 1); - - DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), - _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); - - mt->target = target_to_target(target); - mt->internal_format = internal_format; - mt->first_level = first_level; - mt->last_level = last_level; - mt->width0 = width0; - mt->height0 = height0; - mt->depth0 = depth0; - mt->cpp = compress_byte ? compress_byte : cpp; - mt->compressed = compress_byte ? 1 : 0; - mt->refcount = 1; - - ok = pipe->mipmap_tree_layout(pipe, mt); - if (ok) - mt->region = pipe->region_alloc(pipe, - mt->cpp, mt->pitch, mt->total_height); - - if (!mt->region) { - free(mt); - return NULL; - } - - return mt; -} - - -void -st_miptree_reference(struct pipe_mipmap_tree **dst, - struct pipe_mipmap_tree *src) -{ - src->refcount++; - *dst = src; - DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); -} - -void -st_miptree_release(struct pipe_context *pipe, - struct pipe_mipmap_tree **mt) -{ - if (!*mt) - return; - - DBG("%s %p refcount will be %d\n", __FUNCTION__, *mt, (*mt)->refcount - 1); - if (--(*mt)->refcount <= 0) { - GLuint i; - - DBG("%s deleting %p\n", __FUNCTION__, *mt); - - pipe->region_release(pipe, &((*mt)->region)); - - for (i = 0; i < MAX_TEXTURE_LEVELS; i++) - if ((*mt)->level[i].image_offset) - free((*mt)->level[i].image_offset); - - free(*mt); - } - *mt = NULL; -} - - - - -/* Can the image be pulled into a unified mipmap tree. This mirrors - * the completeness test in a lot of ways. - * - * Not sure whether I want to pass gl_texture_image here. - */ -GLboolean -st_miptree_match_image(struct pipe_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level) -{ - /* Images with borders are never pulled into mipmap trees. - */ - if (image->Border) - return GL_FALSE; - - if (image->InternalFormat != mt->internal_format || - image->IsCompressed != mt->compressed) - return GL_FALSE; - - /* Test image dimensions against the base level image adjusted for - * minification. This will also catch images not present in the - * tree, changed targets, etc. - */ - if (image->Width != mt->level[level].width || - image->Height != mt->level[level].height || - image->Depth != mt->level[level].depth) - return GL_FALSE; - - return GL_TRUE; -} - - -/* Although we use the image_offset[] array to store relative offsets - * to cube faces, Mesa doesn't know anything about this and expects - * each cube face to be treated as a separate image. - * - * These functions present that view to mesa: - */ -const GLuint * -st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) -{ - static const GLuint zero = 0; - - if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) - return &zero; - else - return mt->level[level].image_offset; -} - - -GLuint -st_miptree_image_offset(struct pipe_mipmap_tree * mt, - GLuint face, GLuint level) -{ - if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) - return (mt->level[level].level_offset + - mt->level[level].image_offset[face] * mt->cpp); - else - return mt->level[level].level_offset; -} - - - -/** - * Map a teximage in a mipmap tree. - * \param row_stride returns row stride in bytes - * \param image_stride returns image stride in bytes (for 3D textures). - * \return address of mapping - */ -GLubyte * -st_miptree_image_map(struct pipe_context *pipe, - struct pipe_mipmap_tree * mt, - GLuint face, - GLuint level, - GLuint * row_stride, GLuint * image_offsets) -{ - GLubyte *ptr; - DBG("%s \n", __FUNCTION__); - - if (row_stride) - *row_stride = mt->pitch * mt->cpp; - - if (image_offsets) - memcpy(image_offsets, mt->level[level].image_offset, - mt->level[level].depth * sizeof(GLuint)); - - ptr = pipe->region_map(pipe, mt->region); - - return ptr + st_miptree_image_offset(mt, face, level); -} - -void -st_miptree_image_unmap(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt) -{ - DBG("%s\n", __FUNCTION__); - pipe->region_unmap(pipe, mt->region); -} - - - -/* Upload data for a particular image. - */ -void -st_miptree_image_data(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, - GLuint level, - void *src, - GLuint src_row_pitch, GLuint src_image_pitch) -{ - GLuint depth = dst->level[level].depth; - GLuint dst_offset = st_miptree_image_offset(dst, face, level); - const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); - GLuint i; - GLuint height = 0; - - DBG("%s\n", __FUNCTION__); - for (i = 0; i < depth; i++) { - height = dst->level[level].height; - if(dst->compressed) - height /= 4; - pipe->region_data(pipe, dst->region, - dst_offset + dst_depth_offset[i], /* dst_offset */ - 0, 0, /* dstx, dsty */ - src, - src_row_pitch, - 0, 0, /* source x, y */ - dst->level[level].width, height); /* width, height */ - - src += src_image_pitch * dst->cpp; - } -} - -/* Copy mipmap image between trees - */ -void -st_miptree_image_copy(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, GLuint level, - struct pipe_mipmap_tree *src) -{ - GLuint width = src->level[level].width; - GLuint height = src->level[level].height; - GLuint depth = src->level[level].depth; - GLuint dst_offset = st_miptree_image_offset(dst, face, level); - GLuint src_offset = st_miptree_image_offset(src, face, level); - const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); - const GLuint *src_depth_offset = st_miptree_depth_offsets(src, level); - GLuint i; - - if (dst->compressed) - height /= 4; - for (i = 0; i < depth; i++) { - pipe->region_copy(pipe, - dst->region, dst_offset + dst_depth_offset[i], - 0, - 0, - src->region, src_offset + src_depth_offset[i], - 0, 0, width, height); - } - -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h deleted file mode 100644 index eafe47c029..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_mipmap_tree.h +++ /dev/null @@ -1,105 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_MIPMAP_TREE_H -#define INTEL_MIPMAP_TREE_H - - -#include "main/mtypes.h" - -struct pipe_context; -struct pipe_mipmap_tree; -struct pipe_region; - - -struct pipe_mipmap_tree *st_miptree_create(struct pipe_context *pipe, - GLenum target, - GLenum internal_format, - GLuint first_level, - GLuint last_level, - GLuint width0, - GLuint height0, - GLuint depth0, - GLuint cpp, - GLuint compress_byte); - -void st_miptree_reference(struct pipe_mipmap_tree **dst, - struct pipe_mipmap_tree *src); - -void st_miptree_release(struct pipe_context *pipe, - struct pipe_mipmap_tree **mt); - -/* Check if an image fits an existing mipmap tree layout - */ -GLboolean st_miptree_match_image(struct pipe_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level); - -/* Return a pointer to an image within a tree. Return image stride as - * well. - */ -GLubyte *st_miptree_image_map(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt, - GLuint face, - GLuint level, - GLuint * row_stride, GLuint * image_stride); - -void st_miptree_image_unmap(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt); - - -/* Return the linear offset of an image relative to the start of the - * tree: - */ -GLuint st_miptree_image_offset(struct pipe_mipmap_tree *mt, - GLuint face, GLuint level); - -/* Return pointers to each 2d slice within an image. Indexed by depth - * value. - */ -const GLuint *st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, - GLuint level); - - -/* Upload an image into a tree - */ -void st_miptree_image_data(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, - GLuint level, - void *src, - GLuint src_row_pitch, GLuint src_image_pitch); - -/* Copy an image between two trees - */ -void st_miptree_image_copy(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, GLuint level, - struct pipe_mipmap_tree *src); - - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_reg.h b/src/mesa/drivers/dri/i915pipe/intel_reg.h deleted file mode 100644 index 7828ba6ad3..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_reg.h +++ /dev/null @@ -1,88 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef _INTEL_REG_H_ -#define _INTEL_REG_H_ - - - -#define CMD_3D (0x3<<29) - - -#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) -#define PRIM_INDIRECT (1<<23) -#define PRIM_INLINE (0<<23) -#define PRIM_INDIRECT_SEQUENTIAL (0<<17) -#define PRIM_INDIRECT_ELTS (1<<17) - -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_MASK (0x1f<<18) - -#define I915PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define I915PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define I915PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define I915PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - - -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 - -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) - -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) - -#define MI_WAIT_FOR_EVENT ((0x3<<23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.c b/src/mesa/drivers/dri/i915pipe/intel_screen.c deleted file mode 100644 index 81ab435f42..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.c +++ /dev/null @@ -1,597 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "context.h" -#include "framebuffer.h" -#include "matrix.h" -#include "renderbuffer.h" -#include "simple_list.h" -#include "utils.h" -#include "vblank.h" -#include "xmlpool.h" - -#include "intel_screen.h" -#include "intel_batchbuffer.h" -#include "intel_buffers.h" -/*#include "intel_tex.h"*/ -#include "intel_ioctl.h" -#include "intel_fbo.h" - -#include "i830_dri.h" -#include "dri_bufpool.h" - -#include "pipe/p_context.h" - - - -PUBLIC const char __driConfigOptions[] = - DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE - DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) - DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) - DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY - DRI_CONF_FORCE_S3TC_ENABLE(false) - DRI_CONF_ALLOW_LARGE_TEXTURES(1) - DRI_CONF_SECTION_END DRI_CONF_END; - const GLuint __driNConfigOptions = 4; - -#ifdef USE_NEW_INTERFACE - static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; -#endif /*USE_NEW_INTERFACE */ - - extern const struct dri_extension card_extensions[]; - - - - - -static void -intelPrintDRIInfo(intelScreenPrivate * intelScreen, - __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) -{ - fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", - intelScreen->front.size, intelScreen->front.offset, - intelScreen->front.pitch); - fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); -} - - -static void -intelPrintSAREA(const drmI830Sarea * sarea) -{ - fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, - sarea->height); - fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); - fprintf(stderr, - "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->front_offset, sarea->front_size, - (unsigned) sarea->front_handle); - fprintf(stderr, - "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->back_offset, sarea->back_size, - (unsigned) sarea->back_handle); - fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->depth_offset, sarea->depth_size, - (unsigned) sarea->depth_handle); - fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); - fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); - fprintf(stderr, - "SAREA: rotated offset: 0x%08x size: 0x%x\n", - sarea->rotated_offset, sarea->rotated_size); - fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); -} - - - -/** - * Use the information in the sarea to update the screen parameters - * related to screen rotation. Needs to be called locked. - */ -void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) -{ - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - - if (intelScreen->front.map) { - drmUnmap(intelScreen->front.map, intelScreen->front.size); - intelScreen->front.map = NULL; - } - - if (intelScreen->front.buffer) - driDeleteBuffers(1, &intelScreen->front.buffer); - - intelScreen->front.width = sarea->width; - intelScreen->front.height = sarea->height; - intelScreen->front.offset = sarea->front_offset; - intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; - intelScreen->front.size = sarea->front_size; - intelScreen->front.handle = sarea->front_handle; - - assert( sarea->front_size >= - intelScreen->front.pitch * intelScreen->front.height ); - - if (!sarea->front_handle) - return; - - if (drmMap(sPriv->fd, - sarea->front_handle, - intelScreen->front.size, - (drmAddress *) & intelScreen->front.map) != 0) { - _mesa_problem(NULL, "drmMap(frontbuffer) failed!"); - return; - } - - if (intelScreen->staticPool) { - driGenBuffers(intelScreen->staticPool, "static region", 1, - &intelScreen->front.buffer, 64, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - - driBOSetStatic(intelScreen->front.buffer, - intelScreen->front.offset, - intelScreen->front.pitch * intelScreen->front.height, - intelScreen->front.map, 0); - } -} - - - -GLboolean -intelCreatePools(intelScreenPrivate *intelScreen) -{ - unsigned batchPoolSize = 1024*1024; - __DRIscreenPrivate * sPriv = intelScreen->driScrnPriv; - - if (intelScreen->havePools) - return GL_TRUE; - - batchPoolSize /= BATCH_SZ; - intelScreen->regionPool = driDRMPoolInit(sPriv->fd); - - if (!intelScreen->regionPool) - return GL_FALSE; - - intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); - - if (!intelScreen->staticPool) - return GL_FALSE; - - intelScreen->texPool = intelScreen->regionPool; - - intelScreen->batchPool = driBatchPoolInit(sPriv->fd, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT | - DRM_BO_FLAG_MEM_LOCAL, - BATCH_SZ, - batchPoolSize, 5); - if (!intelScreen->batchPool) { - fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); - return GL_FALSE; - } - - intelScreen->havePools = GL_TRUE; - - intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - return GL_TRUE; -} - - -static GLboolean -intelInitDriver(__DRIscreenPrivate * sPriv) -{ - intelScreenPrivate *intelScreen; - I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; - - PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = - (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> - getProcAddress("glxEnableExtension")); - void *const psc = sPriv->psc->screenConfigs; - - if (sPriv->devPrivSize != sizeof(I830DRIRec)) { - fprintf(stderr, - "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); - return GL_FALSE; - } - - /* Allocate the private area */ - intelScreen = (intelScreenPrivate *) CALLOC(sizeof(intelScreenPrivate)); - if (!intelScreen) - return GL_FALSE; - - /* parse information in __driConfigOptions */ - driParseOptionInfo(&intelScreen->optionCache, - __driConfigOptions, __driNConfigOptions); - - intelScreen->driScrnPriv = sPriv; - sPriv->private = (void *) intelScreen; - - intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + - gDRIPriv->sarea_priv_offset); - intelScreen->deviceID = gDRIPriv->deviceID; - intelScreen->front.cpp = gDRIPriv->cpp; - intelScreen->drmMinor = sPriv->drmMinor; - - assert(gDRIPriv->bitsPerPixel == 16 || - gDRIPriv->bitsPerPixel == 32); - - intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - if (0) - intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); - - if (glx_enable_extension != NULL) { - (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); - (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); - (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); - } - - return GL_TRUE; -} - - -static void -intelDestroyScreen(__DRIscreenPrivate * sPriv) -{ - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - -// intelUnmapScreenRegions(intelScreen); - - if (intelScreen->havePools) { - driPoolTakeDown(intelScreen->regionPool); - driPoolTakeDown(intelScreen->staticPool); - driPoolTakeDown(intelScreen->batchPool); - } - FREE(intelScreen); - sPriv->private = NULL; -} - - -/** - * This is called when we need to set up GL rendering to a new X window. - */ -static GLboolean -intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, - __DRIdrawablePrivate * driDrawPriv, - const __GLcontextModes * mesaVis, GLboolean isPixmap) -{ - if (isPixmap) { - return GL_FALSE; /* not implemented */ - } - else { - GLboolean swStencil = (mesaVis->stencilBits > 0 && - mesaVis->depthBits != 24); - GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); - - struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); - - if (!intel_fb) - return GL_FALSE; - - _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); - - { - /* fake frontbuffer */ - /* XXX allocation should only happen in the unusual case - it's actually needed */ - intel_fb->color_rb[0] - = intel_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, - &intel_fb->color_rb[0]->Base); - } - - if (mesaVis->doubleBufferMode) { - intel_fb->color_rb[1] - = intel_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, - &intel_fb->color_rb[1]->Base); - } - - if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { - /* combined depth/stencil buffer */ - struct intel_renderbuffer *depthStencilRb - = intel_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); - /* note: bind RB to two attachment points */ - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, - &depthStencilRb->Base); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL, - &depthStencilRb->Base); - } - else if (mesaVis->depthBits == 16) { - /* just 16-bit depth buffer, no hw stencil */ - struct intel_renderbuffer *depthRb - = intel_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); - } - - - /* now add any/all software-based renderbuffers we may need */ - _mesa_add_soft_renderbuffers(&intel_fb->Base, - GL_FALSE, /* never sw color */ - GL_FALSE, /* never sw depth */ - swStencil, mesaVis->accumRedBits > 0, - GL_FALSE, /* never sw alpha */ - GL_FALSE /* never sw aux */ ); - driDrawPriv->driverPrivate = (void *) intel_fb; - - return GL_TRUE; - } -} - -static void -intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) -{ - _mesa_unreference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate))); -} - - -/** - * Get information about previous buffer swaps. - */ -static int -intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) -{ - struct intel_framebuffer *intel_fb; - - if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) - || (sInfo == NULL)) { - return -1; - } - - intel_fb = dPriv->driverPrivate; - sInfo->swap_count = intel_fb->swap_count; - sInfo->swap_ust = intel_fb->swap_ust; - sInfo->swap_missed_count = intel_fb->swap_missed_count; - - sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0) - ? driCalculateSwapUsage(dPriv, 0, intel_fb->swap_missed_ust) - : 0.0; - - return 0; -} - - -static void -intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, - unsigned long long offset, GLint depth, GLuint pitch) -{ - abort(); -#if 0 - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct st_texture_object *stObj = st_texture_object(tObj); - - if (!stObj) - return; - - if (stObj->mt) - st_miptree_release(intel->pipe, &stObj->mt); - - stObj->imageOverride = GL_TRUE; - stObj->depthOverride = depth; - stObj->pitchOverride = pitch; - - if (offset) - stObj->textureOffset = offset; -#endif -} - - -static const struct __DriverAPIRec intelAPI = { - .InitDriver = intelInitDriver, - .DestroyScreen = intelDestroyScreen, - .CreateContext = intelCreateContext, - .DestroyContext = intelDestroyContext, - .CreateBuffer = intelCreateBuffer, - .DestroyBuffer = intelDestroyBuffer, - .SwapBuffers = intelSwapBuffers, - .MakeCurrent = intelMakeCurrent, - .UnbindContext = intelUnbindContext, - .GetSwapInfo = intelGetSwapInfo, - .GetMSC = driGetMSC32, - .WaitForMSC = driWaitForMSC32, - .WaitForSBC = NULL, - .SwapBuffersMSC = NULL, - .CopySubBuffer = intelCopySubBuffer, - .setTexOffset = intelSetTexOffset, -}; - - -static __GLcontextModes * -intelFillInModes(unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, GLboolean have_back_buffer) -{ - __GLcontextModes *modes; - __GLcontextModes *m; - unsigned num_modes; - unsigned depth_buffer_factor; - unsigned back_buffer_factor; - GLenum fb_format; - GLenum fb_type; - - /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't - * support pageflipping at all. - */ - static const GLenum back_buffer_modes[] = { - GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML - }; - - u_int8_t depth_bits_array[3]; - u_int8_t stencil_bits_array[3]; - - - depth_bits_array[0] = 0; - depth_bits_array[1] = depth_bits; - depth_bits_array[2] = depth_bits; - - /* Just like with the accumulation buffer, always provide some modes - * with a stencil buffer. It will be a sw fallback, but some apps won't - * care about that. - */ - stencil_bits_array[0] = 0; - stencil_bits_array[1] = 0; - if (depth_bits == 24) - stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; - - stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; - - depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; - back_buffer_factor = (have_back_buffer) ? 3 : 1; - - num_modes = depth_buffer_factor * back_buffer_factor * 4; - - if (pixel_bits == 16) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; - } - - modes = - (*dri_interface->createContextModes) (num_modes, - sizeof(__GLcontextModes)); - m = modes; - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_TRUE_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_DIRECT_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - - /* Mark the visual as slow if there are "fake" stencil bits. - */ - for (m = modes; m != NULL; m = m->next) { - if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { - m->visualRating = GLX_SLOW_CONFIG; - } - } - - return modes; -} - - -/** - * This is the bootstrap function for the driver. libGL supplies all of the - * requisite information about the system, and the driver initializes itself. - * This routine also fills in the linked list pointed to by \c driver_modes - * with the \c __GLcontextModes that the driver can support for windows or - * pbuffers. - * - * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on - * failure. - */ -PUBLIC void * -__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, - __DRIscreen * psc, - const __GLcontextModes * modes, - const __DRIversion * ddx_version, - const __DRIversion * dri_version, - const __DRIversion * drm_version, - const __DRIframebuffer * frame_buffer, - drmAddress pSAREA, int fd, - int internal_api_version, - const __DRIinterfaceMethods * interface, - __GLcontextModes ** driver_modes) -{ - __DRIscreenPrivate *psp; - static const __DRIversion ddx_expected = { 1, 7, 0 }; - static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 7, 0 }; - - dri_interface = interface; - - if (!driCheckDriDdxDrmVersions2("i915", - dri_version, &dri_expected, - ddx_version, &ddx_expected, - drm_version, &drm_expected)) { - return NULL; - } - - psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, - ddx_version, dri_version, drm_version, - frame_buffer, pSAREA, fd, - internal_api_version, &intelAPI); - - if (psp != NULL) { - I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; - *driver_modes = intelFillInModes(dri_priv->cpp * 8, - (dri_priv->cpp == 2) ? 16 : 24, - (dri_priv->cpp == 2) ? 0 : 8, 1); - - /* Calling driInitExtensions here, with a NULL context pointer, does not actually - * enable the extensions. It just makes sure that all the dispatch offsets for all - * the extensions that *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is called, but we can't - * enable the extensions until we have a context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions(NULL, card_extensions, GL_FALSE); - } - - return (void *) psp; -} - -struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) -{ - /* - * This should probably change to have the screen allocate a dummy - * context at screen creation. For now just use the current context. - */ - - GET_CURRENT_CONTEXT(ctx); - if (ctx == NULL) { -/* _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; */ - /* need a context for the first time makecurrent is called (for hw lock - when allocating priv buffers) */ - if (intelScreen->dummyctxptr == NULL) { - _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; - } - return intelScreen->dummyctxptr; - } - return intel_context(ctx); - -} - diff --git a/src/mesa/drivers/dri/i915pipe/intel_screen.h b/src/mesa/drivers/dri/i915pipe/intel_screen.h deleted file mode 100644 index 549587a6f2..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_screen.h +++ /dev/null @@ -1,111 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef _INTEL_SCREEN_H_ -#define _INTEL_SCREEN_H_ - -#include "dri_util.h" -#include "i830_common.h" -#include "xmlconfig.h" -#include "dri_bufpool.h" - - -struct intel_screen -{ - struct { - drm_handle_t handle; - - /* We create a static dri buffer for the frontbuffer. - */ - struct _DriBufferObject *buffer; - - char *map; /* memory map */ - int offset; /* from start of video mem, in bytes */ - int pitch; /* row stride, in bytes */ - int width; - int height; - int size; - int cpp; /* for front and back buffers */ - } front; - - int deviceID; - int drmMinor; - - __DRIscreenPrivate *driScrnPriv; - drmI830Sarea *sarea; - - - /** - * Configuration cache with default values for all contexts - */ - driOptionCache optionCache; - struct _DriBufferPool *batchPool; - struct _DriBufferPool *texPool; - struct _DriBufferPool *regionPool; - struct _DriBufferPool *staticPool; - GLboolean havePools; - - struct intel_context *dummyctxptr; -}; - -typedef struct intel_screen intelScreenPrivate; - -extern void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); - - -extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); - -extern GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); - -extern GLboolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); - -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h); - -extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, - unsigned long bufSize, - unsigned numBufs, - unsigned checkDelayed); - -extern struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen); - -extern GLboolean -intelCreatePools(intelScreenPrivate *intelScreen); - -extern GLboolean -intelCreateContext(const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_softpipe.c b/src/mesa/drivers/dri/i915pipe/intel_softpipe.c deleted file mode 100644 index ef47744358..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_softpipe.c +++ /dev/null @@ -1,184 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include -#include "dri_bufpool.h" -#include "dri_bufmgr.h" - -/* #include "errno.h" */ -/* #include "string.h" */ -/* #include "imports.h" */ - -#include "intel_context.h" - -#include "pipe/softpipe/sp_winsys.h" - - -struct intel_softpipe_winsys { - struct softpipe_winsys sws; - struct intel_context *intel; -}; - - - -/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque - * buffer pointer... - */ -static inline struct _DriBufferObject * -dri_bo( struct pipe_buffer_handle *bo ) -{ - return (struct _DriBufferObject *)bo; -} - -static inline struct pipe_buffer_handle * -pipe_bo( struct _DriBufferObject *bo ) -{ - return (struct pipe_buffer_handle *)bo; -} - -/* Turn a softpipe winsys into an intel/softpipe winsys: - */ -static inline struct intel_softpipe_winsys * -intel_softpipe_winsys( struct softpipe_winsys *sws ) -{ - return (struct intel_softpipe_winsys *)sws; -} - - -/* Most callbacks map direcly onto dri_bufmgr operations: - */ -static void *intel_buffer_map(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf ) -{ - return driBOMap( dri_bo(buf), - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0 ); -} - -static void intel_buffer_unmap(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf) -{ - driBOUnmap( dri_bo(buf) ); -} - - -static struct pipe_buffer_handle * -intel_buffer_reference(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf) -{ - return pipe_bo( driBOReference( dri_bo(buf) ) ); -} - -static void intel_buffer_unreference(struct softpipe_winsys *sws, - struct pipe_buffer_handle **buf) -{ - if (*buf) { - driBOUnReference( dri_bo(*buf) ); - *buf = NULL; - } -} - -/* Grabs the hardware lock! - */ -static void intel_buffer_data(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ) -{ - struct intel_context *intel = intel_softpipe_winsys(sws)->intel; - - LOCK_HARDWARE( intel ); - driBOData( dri_bo(buf), size, data, 0 ); - UNLOCK_HARDWARE( intel ); -} - -static void intel_buffer_subdata(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) -{ - driBOSubData( dri_bo(buf), offset, size, data ); -} - -static void intel_buffer_get_subdata(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) -{ - driBOGetSubData( dri_bo(buf), offset, size, data ); -} - -/* Softpipe has no concept of pools. We choose the tex/region pool - * for all buffers. - */ -static struct pipe_buffer_handle * -intel_create_buffer(struct softpipe_winsys *sws, - unsigned alignment) -{ - struct intel_context *intel = intel_softpipe_winsys(sws)->intel; - struct _DriBufferObject *buffer; - - LOCK_HARDWARE( intel ); - driGenBuffers( intel->intelScreen->regionPool, - "softpipe buffer", 1, &buffer, alignment, 0, 0 ); - UNLOCK_HARDWARE( intel ); - - return pipe_bo(buffer); -} - - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel ) -{ - struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); - - /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. - * - * Softpipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate _DriBufferObject. - */ - isws->sws.create_buffer = intel_create_buffer; - isws->sws.buffer_map = intel_buffer_map; - isws->sws.buffer_unmap = intel_buffer_unmap; - isws->sws.buffer_reference = intel_buffer_reference; - isws->sws.buffer_unreference = intel_buffer_unreference; - isws->sws.buffer_data = intel_buffer_data; - isws->sws.buffer_subdata = intel_buffer_subdata; - isws->sws.buffer_get_subdata = intel_buffer_get_subdata; - isws->intel = intel; - - /* Create the softpipe context: - */ - return softpipe_create( &isws->sws ); -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_surface.c b/src/mesa/drivers/dri/i915pipe/intel_surface.c deleted file mode 100644 index 3fa40271c8..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_surface.c +++ /dev/null @@ -1,196 +0,0 @@ -#include "glheader.h" -#include "context.h" -#include "framebuffer.h" -#include "renderbuffer.h" -#include "utils.h" -#include "main/macros.h" - - -#include "intel_screen.h" - -#include "intel_context.h" -#include "intel_buffers.h" -#include "intel_fbo.h" - -#include "pipe/p_state.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/softpipe/sp_surface.h" - - -/* - * XXX a lof of this is a temporary kludge - */ - -/** - * Note: the arithmetic/addressing in these functions is a little - * tricky since we need to invert the Y axis. - */ - - -static void -read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) -{ - const GLint bytesPerRow = sps->surface.region->pitch * sps->surface.region->cpp; - const GLint invY = sps->surface.height - y - 1; - const GLubyte *src = sps->surface.region->map + invY * bytesPerRow + x * sps->surface.region->cpp; - GLfloat *dst = (GLfloat *) rrrr; - GLubyte temp[16]; - GLuint j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - - memcpy(temp + 8, src, 8); - memcpy(temp + 0, src + bytesPerRow, 8); - - for (j = 0; j < 4; j++) { - dst[0 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 2]); /*R*/ - dst[1 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 1]); /*G*/ - dst[2 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 0]); /*B*/ - dst[3 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 3]); /*A*/ - } -} - - -static void -write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) -{ - const GLfloat *src = (const GLfloat *) rrrr; - const GLint bytesPerRow = sps->surface.region->pitch * sps->surface.region->cpp; - const GLint invY = sps->surface.height - y - 1; - GLubyte *dst = sps->surface.region->map + invY * bytesPerRow + x * sps->surface.region->cpp; - GLubyte temp[16]; - GLuint j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - - for (j = 0; j < 4; j++) { - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 2], src[0 * 4 + j]); /*R*/ - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 1], src[1 * 4 + j]); /*G*/ - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 0], src[2 * 4 + j]); /*B*/ - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 3], src[3 * 4 + j]); /*A*/ - } - - memcpy(dst, temp + 8, 8); - memcpy(dst + bytesPerRow, temp + 0, 8); -} - - - -static void -read_quad_z24(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) -{ - static const GLuint mask = 0xffffff; - const GLint invY = sps->surface.height - y - 1; - const GLuint *src - = (GLuint *) (sps->surface.region->map - + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - /* extract lower three bytes */ - zzzz[0] = src[0] & mask; - zzzz[1] = src[1] & mask; - zzzz[2] = src[-sps->surface.region->pitch] & mask; - zzzz[3] = src[-sps->surface.region->pitch + 1] & mask; -} - -static void -write_quad_z24(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) -{ - static const GLuint mask = 0xff000000; - const GLint invY = sps->surface.height - y - 1; - GLuint *dst - = (GLuint *) (sps->surface.region->map - + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - /* write lower three bytes */ - dst[0] = (dst[0] & mask) | zzzz[0]; - dst[1] = (dst[1] & mask) | zzzz[1]; - dst -= sps->surface.region->pitch; - dst[0] = (dst[0] & mask) | zzzz[2]; - dst[1] = (dst[1] & mask) | zzzz[3]; -} - - -static void -read_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) -{ - const GLint invY = sps->surface.height - y - 1; - const GLuint *src = (const GLuint *) (sps->surface.region->map - + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - /* extract high byte */ - ssss[0] = src[0] >> 24; - ssss[1] = src[1] >> 24; - src -= sps->surface.region->pitch; - ssss[2] = src[0] >> 24; - ssss[3] = src[1] >> 24; -} - -static void -write_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) -{ - static const GLuint mask = 0x00ffffff; - const GLint invY = sps->surface.height - y - 1; - GLuint *dst = (GLuint *) (sps->surface.region->map - + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - /* write high byte */ - dst[0] = (dst[0] & mask) | (ssss[0] << 24); - dst[1] = (dst[1] & mask) | (ssss[1] << 24); - dst -= sps->surface.region->pitch; - dst[0] = (dst[0] & mask) | (ssss[2] << 24); - dst[1] = (dst[1] & mask) | (ssss[3] << 24); -} - - -struct pipe_surface * -intel_new_surface(GLuint intFormat) -{ - struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); - if (!sps) - return NULL; - - sps->surface.width = 0; /* set in intel_alloc_renderbuffer_storage() */ - sps->surface.height = 0; - - if (intFormat == GL_RGBA8) { - sps->surface.format = PIPE_FORMAT_U_A8_R8_G8_B8; - sps->read_quad_f_swz = read_quad_f_swz; - sps->write_quad_f_swz = write_quad_f_swz; - } - else if (intFormat == GL_RGB5) { - sps->surface.format = PIPE_FORMAT_U_R5_G6_B5; - - } - else if (intFormat == GL_DEPTH_COMPONENT16) { - sps->surface.format = PIPE_FORMAT_U_Z16; - - } - else if (intFormat == GL_DEPTH24_STENCIL8_EXT) { - sps->surface.format = PIPE_FORMAT_S8_Z24; - sps->read_quad_z = read_quad_z24; - sps->write_quad_z = write_quad_z24; - sps->read_quad_stencil = read_quad_stencil; - sps->write_quad_stencil = write_quad_stencil; - } - else { - /* TBD / unknown */ - - } - - return &sps->surface; -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.c b/src/mesa/drivers/dri/i915pipe/intel_tex.c deleted file mode 100644 index 5270094cc5..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.c +++ /dev/null @@ -1,184 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "main/texobj.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_mipmap_tree.h" -#include "intel_tex.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -static GLboolean -intelIsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) -{ -#if 0 - struct intel_context *intel = intel_context(ctx); - struct st_texture_object *stObj = st_texture_object(texObj); - - return - stObj->mt && - stObj->mt->region && - intel_is_region_resident(intel, stObj->mt->region); -#endif - return 1; -} - - - -static struct gl_texture_image * -intelNewTextureImage(GLcontext * ctx) -{ - DBG("%s\n", __FUNCTION__); - (void) ctx; - return (struct gl_texture_image *) CALLOC_STRUCT(st_texture_image); -} - - -static struct gl_texture_object * -intelNewTextureObject(GLcontext * ctx, GLuint name, GLenum target) -{ - struct st_texture_object *obj = CALLOC_STRUCT(st_texture_object); - - DBG("%s\n", __FUNCTION__); - _mesa_initialize_texture_object(&obj->base, name, target); - - return &obj->base; -} - -static void -intelDeleteTextureObject(GLcontext *ctx, - struct gl_texture_object *texObj) -{ - struct pipe_context *pipe = ctx->st->pipe; - struct st_texture_object *stObj = st_texture_object(texObj); - - if (stObj->mt) - st_miptree_release(pipe, &stObj->mt); - - _mesa_delete_texture_object(ctx, texObj); -} - - -static void -intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) -{ - struct pipe_context *pipe = ctx->st->pipe; - struct st_texture_image *stImage = st_texture_image(texImage); - - DBG("%s\n", __FUNCTION__); - - if (stImage->mt) { - st_miptree_release(pipe, &stImage->mt); - } - - if (texImage->Data) { - free(texImage->Data); - texImage->Data = NULL; - } -} - - - - -/* ================================================================ - * From linux kernel i386 header files, copes with odd sizes better - * than COPY_DWORDS would: - * XXX Put this in src/mesa/main/imports.h ??? - */ -#if defined(i386) || defined(__i386__) -static INLINE void * -__memcpy(void *to, const void *from, size_t n) -{ - int d0, d1, d2; - __asm__ __volatile__("rep ; movsl\n\t" - "testb $2,%b4\n\t" - "je 1f\n\t" - "movsw\n" - "1:\ttestb $1,%b4\n\t" - "je 2f\n\t" - "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2) - :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from) - :"memory"); - return (to); -} -#else -#define __memcpy(a,b,c) memcpy(a,b,c) -#endif - - -/* The system memcpy (at least on ubuntu 5.10) has problems copying - * to agp (writecombined) memory from a source which isn't 64-byte - * aligned - there is a 4x performance falloff. - * - * The x86 __memcpy is immune to this but is slightly slower - * (10%-ish) than the system memcpy. - * - * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but - * isn't much faster than x86_memcpy for agp copies. - * - * TODO: switch dynamically. - */ -static void * -do_memcpy(void *dest, const void *src, size_t n) -{ - if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) { - return __memcpy(dest, src, n); - } - else - return memcpy(dest, src, n); -} - - -void -intelInitTextureFuncs(struct dd_function_table *functions) -{ - functions->ChooseTextureFormat = intelChooseTextureFormat; - functions->TexImage1D = intelTexImage1D; - functions->TexImage2D = intelTexImage2D; - functions->TexImage3D = intelTexImage3D; - functions->TexSubImage1D = intelTexSubImage1D; - functions->TexSubImage2D = intelTexSubImage2D; - functions->TexSubImage3D = intelTexSubImage3D; - functions->CopyTexImage1D = intelCopyTexImage1D; - functions->CopyTexImage2D = intelCopyTexImage2D; - functions->CopyTexSubImage1D = intelCopyTexSubImage1D; - functions->CopyTexSubImage2D = intelCopyTexSubImage2D; - functions->GetTexImage = intelGetTexImage; - - /* compressed texture functions */ - functions->CompressedTexImage2D = intelCompressedTexImage2D; - functions->GetCompressedTexImage = intelGetCompressedTexImage; - - functions->NewTextureObject = intelNewTextureObject; - functions->NewTextureImage = intelNewTextureImage; - functions->DeleteTexture = intelDeleteTextureObject; - functions->FreeTexImageData = intelFreeTextureImageData; - functions->UpdateTexturePalette = 0; - functions->IsTextureResident = intelIsTextureResident; - - functions->TextureMemCpy = do_memcpy; -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex.h b/src/mesa/drivers/dri/i915pipe/intel_tex.h deleted file mode 100644 index 7a1cc91974..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_tex.h +++ /dev/null @@ -1,212 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTELTEX_INC -#define INTELTEX_INC - -#include "mtypes.h" -#include "intel_context.h" -#include "texmem.h" - -struct st_texture_object -{ - struct gl_texture_object base; /* The "parent" object */ - - /* The mipmap tree must include at least these levels once - * validated: - */ - GLuint firstLevel; - GLuint lastLevel; - - /* Offset for firstLevel image: - */ - GLuint textureOffset; - - /* On validation any active images held in main memory or in other - * regions will be copied to this region and the old storage freed. - */ - struct pipe_mipmap_tree *mt; - - GLboolean imageOverride; - GLint depthOverride; - GLuint pitchOverride; -}; - - - -struct st_texture_image -{ - struct gl_texture_image base; - - /* These aren't stored in gl_texture_image - */ - GLuint level; - GLuint face; - - /* If stImage->mt != NULL, image data is stored here. - * Else if stImage->base.Data != NULL, image is stored there. - * Else there is no image data. - */ - struct pipe_mipmap_tree *mt; -}; - - -void intelInitTextureFuncs(struct dd_function_table *functions); - -const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type); - - -void intelTexImage3D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage3D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexImage2D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage2D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexImage1D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage1D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLint border); - -void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint border); - -void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint x, GLint y, GLsizei width); - -void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLint x, GLint y, GLsizei width, GLsizei height); - -void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ); - -void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, - GLvoid *pixels, - const struct gl_texture_object *texObj, - const struct gl_texture_image *texImage); - -void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, - unsigned long long offset, GLint depth, GLuint pitch); - -GLuint intel_finalize_mipmap_tree(GLcontext *ctx, - struct pipe_context *pipe, GLuint unit, - GLboolean *needFlush); - - -#if 0 -void intel_tex_map_images(struct pipe_context *pipe, - struct st_texture_object *stObj); - -void intel_tex_unmap_images(struct pipe_context *pipe, - struct st_texture_object *stObj); -#endif - -int intel_compressed_num_bytes(GLuint mesaFormat); - - -static INLINE struct st_texture_object * -st_texture_object(struct gl_texture_object *obj) -{ - return (struct st_texture_object *) obj; -} - -static INLINE struct st_texture_image * -st_texture_image(struct gl_texture_image *img) -{ - return (struct st_texture_image *) img; -} - - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c b/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c deleted file mode 100644 index d2cf6f4669..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_copy.c +++ /dev/null @@ -1,315 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "main/mtypes.h" -#include "main/enums.h" -#include "main/image.h" -#include "main/teximage.h" - -#include "state_tracker/st_mipmap_tree.h" - -#if 0 -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_buffers.h" -#include "intel_fbo.h" -#include "intel_blit.h" -#endif -#include "intel_tex.h" - - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -/** - * Get the pipe_region which is the source for any glCopyTex[Sub]Image call. - * - * Do the best we can using the blitter. A future project is to use - * the texture engine and fragment programs for these copies. - */ -static const struct pipe_region * -get_teximage_source(struct intel_context *intel, GLenum internalFormat) -{ -#if 00 - struct intel_renderbuffer *irb; - - DBG("%s %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(internalFormat)); - - switch (internalFormat) { - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16_ARB: - irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); - if (irb && irb->region && irb->region->cpp == 2) - return irb->region; - return NULL; - case GL_DEPTH24_STENCIL8_EXT: - case GL_DEPTH_STENCIL_EXT: - irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); - if (irb && irb->region && irb->region->cpp == 4) - return irb->region; - return NULL; - case GL_RGBA: - case GL_RGBA8: - return intel_readbuf_region(intel); - case GL_RGB: - if (intel->intelScreen->front.cpp == 2) - return intel_readbuf_region(intel); - return NULL; - default: - return NULL; - } -#else - return NULL; -#endif -} - - -static GLboolean -do_copy_texsubimage(GLcontext *ctx, - struct st_texture_image *intelImage, - GLenum internalFormat, - GLint dstx, GLint dsty, - GLint x, GLint y, GLsizei width, GLsizei height) -{ - struct intel_context *intel = intel_context(ctx); - const struct pipe_region *src = - get_teximage_source(intel, internalFormat); - - if (!intelImage->mt || !src) { - DBG("%s fail %p %p\n", __FUNCTION__, intelImage->mt, src); - return GL_FALSE; - } - -#if 00 /* XXX FIX flush/locking */ - intelFlush(ctx); - /* XXX still need the lock ? */ - LOCK_HARDWARE(intel); -#endif - - { - GLuint image_offset = st_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); - const GLint orig_x = x; - const GLint orig_y = y; - const struct gl_framebuffer *fb = ctx->DrawBuffer; - - if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, - &x, &y, &width, &height)) { - /* Update dst for clipped src. Need to also clip the source rect. - */ - dstx += x - orig_x; - dsty += y - orig_y; - - if (!(ctx->ReadBuffer->Name == 0)) { - /* XXX this looks bogus ? */ - /* FBO: invert Y */ - y = ctx->ReadBuffer->Height - y - 1; - } - - /* A bit of fiddling to get the blitter to work with -ve - * pitches. But we get a nice inverted blit this way, so it's - * worth it: - */ -#if 0 - intelEmitCopyBlit(intel, - intelImage->mt->cpp, - -src->pitch, - src->buffer, - src->height * src->pitch * src->cpp, - intelImage->mt->pitch, - intelImage->mt->region->buffer, - image_offset, - x, y + height, dstx, dsty, width, height, - GL_COPY); /* ? */ - intel_batchbuffer_flush(intel->batch); -#else - /* XXX use pipe->region_copy() ??? */ - (void) image_offset; -#endif - } - } - -#if 0 - UNLOCK_HARDWARE(intel); -#endif - -#if 0 - /* GL_SGIS_generate_mipmap -- this can be accelerated now. - * XXX Add a ctx->Driver.GenerateMipmaps() function? - */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); - } -#endif - - return GL_TRUE; -} - - - - - -void -intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLint border) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - - if (border) - goto fail; - - /* Setup or redefine the texture object, mipmap tree and texture - * image. Don't populate yet. - */ - ctx->Driver.TexImage1D(ctx, target, level, internalFormat, - width, border, - GL_RGBA, CHAN_TYPE, NULL, - &ctx->DefaultPacking, texObj, texImage); - - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, 0, 0, x, y, width, 1)) - goto fail; - - return; - - fail: -#if 0 - _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, - width, border); -#endif - ; -} - -void -intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint border) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - - if (border) - goto fail; - - /* Setup or redefine the texture object, mipmap tree and texture - * image. Don't populate yet. - */ - ctx->Driver.TexImage2D(ctx, target, level, internalFormat, - width, height, border, - GL_RGBA, CHAN_TYPE, NULL, - &ctx->DefaultPacking, texObj, texImage); - - - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, 0, 0, x, y, width, height)) - goto fail; - - return; - - fail: -#if 0 - _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, - width, height, border); -#endif - assert(0); -} - - -void -intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint x, GLint y, GLsizei width) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - GLenum internalFormat = texImage->InternalFormat; - - /* XXX need to check as in above function? */ - - /* Need to check texture is compatible with source format. - */ - - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, xoffset, 0, x, y, width, 1)) { -#if 0 - _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); -#endif - assert(0); - } -} - - - -void -intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLint x, GLint y, GLsizei width, GLsizei height) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - GLenum internalFormat = texImage->InternalFormat; - - - /* Need to check texture is compatible with source format. - */ - - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, - xoffset, yoffset, x, y, width, height)) { -#if 0 - _swrast_copy_texsubimage2d(ctx, target, level, - xoffset, yoffset, x, y, width, height); -#endif - assert(0); - } -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_format.c b/src/mesa/drivers/dri/i915pipe/intel_tex_format.c deleted file mode 100644 index a11718bb07..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_format.c +++ /dev/null @@ -1,172 +0,0 @@ -#include "intel_context.h" -#include "intel_tex.h" -#include "texformat.h" -#include "enums.h" - -/* It works out that this function is fine for all the supported - * hardware. However, there is still a need to map the formats onto - * hardware descriptors. - */ -/* Note that the i915 can actually support many more formats than - * these if we take the step of simply swizzling the colors - * immediately after sampling... - */ -const struct gl_texture_format * -intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, - GLenum format, GLenum type) -{ - struct intel_context *intel = intel_context(ctx); - const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); - - switch (internalFormat) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - if (format == GL_BGRA) { - if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - return &_mesa_texformat_argb8888; - } - else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - return &_mesa_texformat_argb4444; - } - else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - return &_mesa_texformat_argb1555; - } - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return &_mesa_texformat_rgb565; - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case GL_RGBA4: - case GL_RGBA2: - return &_mesa_texformat_argb4444; - - case GL_RGB5_A1: - return &_mesa_texformat_argb1555; - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - return &_mesa_texformat_argb8888; - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; - - case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; - - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - return &_mesa_texformat_z16; - - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_z24_s8; - - default: - fprintf(stderr, "unexpected texture format %s in %s\n", - _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); - return NULL; - } - - return NULL; /* never get here */ -} - -int intel_compressed_num_bytes(GLuint mesaFormat) -{ - int bytes = 0; - switch(mesaFormat) { - - case MESA_FORMAT_RGB_FXT1: - case MESA_FORMAT_RGBA_FXT1: - case MESA_FORMAT_RGB_DXT1: - case MESA_FORMAT_RGBA_DXT1: - bytes = 2; - break; - - case MESA_FORMAT_RGBA_DXT3: - case MESA_FORMAT_RGBA_DXT5: - bytes = 4; - default: - break; - } - - return bytes; -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c b/src/mesa/drivers/dri/i915pipe/intel_tex_image.c deleted file mode 100644 index 93a59b6896..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_image.c +++ /dev/null @@ -1,695 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "glheader.h" -#include "macros.h" -#include "mtypes.h" -#include "enums.h" -#include "colortab.h" -#include "convolve.h" -#include "context.h" -#include "simple_list.h" -#include "texcompress.h" -#include "texformat.h" -#include "texobj.h" -#include "texstore.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_mipmap_tree.h" - -#include "intel_tex.h" - - - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -/* Functions to store texture images. Where possible, mipmap_tree's - * will be created or further instantiated with image data, otherwise - * images will be stored in malloc'd memory. A validation step is - * required to pull those images into a mipmap tree, or otherwise - * decide a fallback is required. - */ - - -static int -logbase2(int n) -{ - GLint i = 1; - GLint log2 = 0; - - while (n > i) { - i *= 2; - log2++; - } - - return log2; -} - - -/* Otherwise, store it in memory if (Border != 0) or (any dimension == - * 1). - * - * Otherwise, if max_level >= level >= min_level, create tree with - * space for textures from min_level down to max_level. - * - * Otherwise, create tree with space for textures from (level - * 0)..(1x1). Consider pruning this tree at a validation if the - * saving is worth it. - */ -static void -guess_and_alloc_mipmap_tree(struct pipe_context *pipe, - struct st_texture_object *intelObj, - struct st_texture_image *intelImage) -{ - GLuint firstLevel; - GLuint lastLevel; - GLuint width = intelImage->base.Width; - GLuint height = intelImage->base.Height; - GLuint depth = intelImage->base.Depth; - GLuint l2width, l2height, l2depth; - GLuint i, comp_byte = 0; - - DBG("%s\n", __FUNCTION__); - - if (intelImage->base.Border) - return; - - if (intelImage->level > intelObj->base.BaseLevel && - (intelImage->base.Width == 1 || - (intelObj->base.Target != GL_TEXTURE_1D && - intelImage->base.Height == 1) || - (intelObj->base.Target == GL_TEXTURE_3D && - intelImage->base.Depth == 1))) - return; - - /* If this image disrespects BaseLevel, allocate from level zero. - * Usually BaseLevel == 0, so it's unlikely to happen. - */ - if (intelImage->level < intelObj->base.BaseLevel) - firstLevel = 0; - else - firstLevel = intelObj->base.BaseLevel; - - - /* Figure out image dimensions at start level. - */ - for (i = intelImage->level; i > firstLevel; i--) { - width <<= 1; - if (height != 1) - height <<= 1; - if (depth != 1) - depth <<= 1; - } - - /* Guess a reasonable value for lastLevel. This is probably going - * to be wrong fairly often and might mean that we have to look at - * resizable buffers, or require that buffers implement lazy - * pagetable arrangements. - */ - if ((intelObj->base.MinFilter == GL_NEAREST || - intelObj->base.MinFilter == GL_LINEAR) && - intelImage->level == firstLevel) { - lastLevel = firstLevel; - } - else { - l2width = logbase2(width); - l2height = logbase2(height); - l2depth = logbase2(depth); - lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); - } - - assert(!intelObj->mt); - if (intelImage->base.IsCompressed) - comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); - intelObj->mt = st_miptree_create(pipe, - intelObj->base.Target, - intelImage->base.InternalFormat, - firstLevel, - lastLevel, - width, - height, - depth, - intelImage->base.TexFormat->TexelBytes, - comp_byte); - - DBG("%s - success\n", __FUNCTION__); -} - - - - -static GLuint -target_to_face(GLenum target) -{ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - return ((GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X); - default: - return 0; - } -} - - - -/* There are actually quite a few combinations this will work for, - * more than what I've listed here. - */ -static GLboolean -check_pbo_format(GLint internalFormat, - GLenum format, GLenum type, - const struct gl_texture_format *mesa_format) -{ - switch (internalFormat) { - case 4: - case GL_RGBA: - return (format == GL_BGRA && - (type == GL_UNSIGNED_BYTE || - type == GL_UNSIGNED_INT_8_8_8_8_REV) && - mesa_format == &_mesa_texformat_argb8888); - case 3: - case GL_RGB: - return (format == GL_RGB && - type == GL_UNSIGNED_SHORT_5_6_5 && - mesa_format == &_mesa_texformat_rgb565); - case GL_YCBCR_MESA: - return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); - default: - return GL_FALSE; - } -} - - -/* XXX: Do this for TexSubImage also: - */ -static GLboolean -try_pbo_upload(struct intel_context *intel, - struct st_texture_image *intelImage, - const struct gl_pixelstore_attrib *unpack, - GLint internalFormat, - GLint width, GLint height, - GLenum format, GLenum type, const void *pixels) -{ - return GL_FALSE; /* XXX fix flushing/locking/blitting below */ -#if 000 - struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); - GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; - - if (!pbo || - intel->ctx._ImageTransferState || - unpack->SkipPixels || unpack->SkipRows) { - _mesa_printf("%s: failure 1\n", __FUNCTION__); - return GL_FALSE; - } - - src_offset = (GLuint) pixels; - - if (unpack->RowLength > 0) - src_stride = unpack->RowLength; - else - src_stride = width; - - dst_offset = st_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); - - dst_stride = intelImage->mt->pitch; - - intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); - { - struct _DriBufferObject *src_buffer = - intel_bufferobj_buffer(intel, pbo, INTEL_READ); - - /* Temporary hack: cast to _DriBufferObject: - */ - struct _DriBufferObject *dst_buffer = - (struct _DriBufferObject *)intelImage->mt->region->buffer; - - - intelEmitCopyBlit(intel, - intelImage->mt->cpp, - src_stride, src_buffer, src_offset, - dst_stride, dst_buffer, dst_offset, - 0, 0, 0, 0, width, height, - GL_COPY); - - intel_batchbuffer_flush(intel->batch); - } - UNLOCK_HARDWARE(intel); - - return GL_TRUE; -#endif -} - - - -static GLboolean -try_pbo_zcopy(struct intel_context *intel, - struct st_texture_image *intelImage, - const struct gl_pixelstore_attrib *unpack, - GLint internalFormat, - GLint width, GLint height, - GLenum format, GLenum type, const void *pixels) -{ - return GL_FALSE; -} - - - - - - -static void -intelTexImage(GLcontext * ctx, - GLint dims, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage, GLsizei imageSize, int compressed) -{ - struct intel_context *intel = intel_context(ctx); - struct st_texture_object *intelObj = st_texture_object(texObj); - struct st_texture_image *intelImage = st_texture_image(texImage); - GLint postConvWidth = width; - GLint postConvHeight = height; - GLint texelBytes, sizeInBytes; - GLuint dstRowStride; - - - DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); - - intelFlush(ctx); - - intelImage->face = target_to_face(target); - intelImage->level = level; - - if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { - _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth, - &postConvHeight); - } - - /* choose the texture format */ - texImage->TexFormat = intelChooseTextureFormat(ctx, internalFormat, - format, type); - - _mesa_set_fetch_functions(texImage, dims); - - if (texImage->TexFormat->TexelBytes == 0) { - /* must be a compressed format */ - texelBytes = 0; - texImage->IsCompressed = GL_TRUE; - texImage->CompressedSize = - ctx->Driver.CompressedTextureSize(ctx, texImage->Width, - texImage->Height, texImage->Depth, - texImage->TexFormat->MesaFormat); - } else { - texelBytes = texImage->TexFormat->TexelBytes; - - /* Minimum pitch of 32 bytes */ - if (postConvWidth * texelBytes < 32) { - postConvWidth = 32 / texelBytes; - texImage->RowStride = postConvWidth; - } - - assert(texImage->RowStride == postConvWidth); - } - - /* Release the reference to a potentially orphaned buffer. - * Release any old malloced memory. - */ - if (intelImage->mt) { - st_miptree_release(intel->pipe, &intelImage->mt); - assert(!texImage->Data); - } - else if (texImage->Data) { - _mesa_align_free(texImage->Data); - } - - /* If this is the only texture image in the tree, could call - * bmBufferData with NULL data to free the old block and avoid - * waiting on any outstanding fences. - */ - if (intelObj->mt && - intelObj->mt->first_level == level && - intelObj->mt->last_level == level && - intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && - !st_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { - - DBG("release it\n"); - st_miptree_release(intel->pipe, &intelObj->mt); - assert(!intelObj->mt); - } - - if (!intelObj->mt) { - guess_and_alloc_mipmap_tree(intel->pipe, intelObj, intelImage); - if (!intelObj->mt) { - DBG("guess_and_alloc_mipmap_tree: failed\n"); - } - } - - assert(!intelImage->mt); - - if (intelObj->mt && - st_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { - - st_miptree_reference(&intelImage->mt, intelObj->mt); - assert(intelImage->mt); - } - - if (!intelImage->mt) - DBG("XXX: Image did not fit into tree - storing in local memory!\n"); - -#if 0 /* XXX FIX when st_buffer_objects are in place */ - /* PBO fastpaths: - */ - if (dims <= 2 && - intelImage->mt && - intel_buffer_object(unpack->BufferObj) && - check_pbo_format(internalFormat, format, - type, intelImage->base.TexFormat)) { - - DBG("trying pbo upload\n"); - - /* Attempt to texture directly from PBO data (zero copy upload). - * - * Currently disable as it can lead to worse as well as better - * performance (in particular when pipe_region_cow() is - * required). - */ - if (intelObj->mt == intelImage->mt && - intelObj->mt->first_level == level && - intelObj->mt->last_level == level) { - - if (try_pbo_zcopy(intel, intelImage, unpack, - internalFormat, - width, height, format, type, pixels)) { - - DBG("pbo zcopy upload succeeded\n"); - return; - } - } - - - /* Otherwise, attempt to use the blitter for PBO image uploads. - */ - if (try_pbo_upload(intel, intelImage, unpack, - internalFormat, - width, height, format, type, pixels)) { - DBG("pbo upload succeeded\n"); - return; - } - - DBG("pbo upload failed\n"); - } -#else - (void) try_pbo_upload; - (void) check_pbo_format; - (void) try_pbo_zcopy; -#endif - - - /* intelCopyTexImage calls this function with pixels == NULL, with - * the expectation that the mipmap tree will be set up but nothing - * more will be done. This is where those calls return: - */ - if (compressed) { - pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels, - unpack, - "glCompressedTexImage"); - } else { - pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, - format, type, - pixels, unpack, "glTexImage"); - } - if (!pixels) - return; - - - if (intelImage->mt) - intel->pipe->region_idle(intel->pipe, intelImage->mt->region); - - LOCK_HARDWARE(intel); - - if (intelImage->mt) { - texImage->Data = st_miptree_image_map(intel->pipe, - intelImage->mt, - intelImage->face, - intelImage->level, - &dstRowStride, - intelImage->base.ImageOffsets); - } - else { - /* Allocate regular memory and store the image there temporarily. */ - if (texImage->IsCompressed) { - sizeInBytes = texImage->CompressedSize; - dstRowStride = - _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); - assert(dims != 3); - } - else { - dstRowStride = postConvWidth * texelBytes; - sizeInBytes = depth * dstRowStride * postConvHeight; - } - - texImage->Data = malloc(sizeInBytes); - } - - DBG("Upload image %dx%dx%d row_len %x " - "pitch %x\n", - width, height, depth, width * texelBytes, dstRowStride); - - /* Copy data. Would like to know when it's ok for us to eg. use - * the blitter to copy. Or, use the hardware to do the format - * conversion and copy: - */ - if (compressed) { - memcpy(texImage->Data, pixels, imageSize); - } else if (!texImage->TexFormat->StoreImage(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, unpack)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); - } - - _mesa_unmap_teximage_pbo(ctx, unpack); - - if (intelImage->mt) { - st_miptree_image_unmap(intel->pipe, intelImage->mt); - texImage->Data = NULL; - } - - UNLOCK_HARDWARE(intel); - -#if 0 - /* GL_SGIS_generate_mipmap -- this can be accelerated now. - */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); - } -#endif -} - -void -intelTexImage3D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexImage(ctx, 3, target, level, - internalFormat, width, height, depth, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); -} - - -void -intelTexImage2D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexImage(ctx, 2, target, level, - internalFormat, width, height, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); -} - -void -intelTexImage1D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexImage(ctx, 1, target, level, - internalFormat, width, 1, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); -} - -void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - intelTexImage(ctx, 2, target, level, - internalFormat, width, height, 1, border, - 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1); -} - -/** - * Need to map texture image into memory before copying image data, - * then unmap it. - */ -static void -intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage, int compressed) -{ - struct intel_context *intel = intel_context(ctx); - struct st_texture_image *intelImage = st_texture_image(texImage); - - /* Map */ - if (intelImage->mt) { - /* Image is stored in hardware format in a buffer managed by the - * kernel. Need to explicitly map and unmap it. - */ - intelImage->base.Data = - st_miptree_image_map(intel->pipe, - intelImage->mt, - intelImage->face, - intelImage->level, - &intelImage->base.RowStride, - intelImage->base.ImageOffsets); - intelImage->base.RowStride /= intelImage->mt->cpp; - } - else { - /* Otherwise, the image should actually be stored in - * intelImage->base.Data. This is pretty confusing for - * everybody, I'd much prefer to separate the two functions of - * texImage->Data - storage for texture images in main memory - * and access (ie mappings) of images. In other words, we'd - * create a new texImage->Map field and leave Data simply for - * storage. - */ - assert(intelImage->base.Data); - } - - - if (compressed) { - _mesa_get_compressed_teximage(ctx, target, level, pixels, - texObj, texImage); - } else { - _mesa_get_teximage(ctx, target, level, format, type, pixels, - texObj, texImage); - } - - - /* Unmap */ - if (intelImage->mt) { - st_miptree_image_unmap(intel->pipe, intelImage->mt); - intelImage->base.Data = NULL; - } -} - -void -intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intel_get_tex_image(ctx, target, level, format, type, pixels, - texObj, texImage, 0); - - -} - -void -intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, - GLvoid *pixels, - const struct gl_texture_object *texObj, - const struct gl_texture_image *texImage) -{ - intel_get_tex_image(ctx, target, level, 0, 0, pixels, - texObj, texImage, 1); - -} - -void -intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, - unsigned long long offset, GLint depth, GLuint pitch) -{ - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct st_texture_object *stObj = st_texture_object(tObj); - - if (!stObj) - return; - - if (stObj->mt) - st_miptree_release(intel->pipe, &stObj->mt); - - stObj->imageOverride = GL_TRUE; - stObj->depthOverride = depth; - stObj->pitchOverride = pitch; - - if (offset) - stObj->textureOffset = offset; -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_layout.c b/src/mesa/drivers/dri/i915pipe/intel_tex_layout.c deleted file mode 120000 index fe61b44194..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_layout.c +++ /dev/null @@ -1 +0,0 @@ -../intel/intel_tex_layout.c \ No newline at end of file diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c b/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c deleted file mode 100644 index 0937114c7f..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_subimage.c +++ /dev/null @@ -1,192 +0,0 @@ - -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "mtypes.h" -#include "texobj.h" -#include "texstore.h" -#include "enums.h" - -/* -#include "intel_context.h" -*/ -#include "intel_tex.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_mipmap_tree.h" - -#include "pipe/p_context.h" - - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -static void -intelTexSubimage(GLcontext * ctx, - GLint dims, - GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLint width, GLint height, GLint depth, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - struct pipe_context *pipe = ctx->st->pipe; - struct st_texture_image *intelImage = st_texture_image(texImage); - GLuint dstRowStride; - - DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), - level, xoffset, yoffset, width, height); - - intelFlush(ctx); - - pixels = - _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format, - type, pixels, packing, "glTexSubImage2D"); - if (!pixels) - return; - - if (intelImage->mt) - pipe->region_idle(pipe, intelImage->mt->region); - -#if 0 - LOCK_HARDWARE(intel); -#endif - - /* Map buffer if necessary. Need to lock to prevent other contexts - * from uploading the buffer under us. - */ - if (intelImage->mt) - texImage->Data = st_miptree_image_map(pipe, - intelImage->mt, - intelImage->face, - intelImage->level, - &dstRowStride, - texImage->ImageOffsets); - - assert(dstRowStride); - - if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); - } - -#if 0 - /* GL_SGIS_generate_mipmap */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); - } -#endif - - _mesa_unmap_teximage_pbo(ctx, packing); - - if (intelImage->mt) { - st_miptree_image_unmap(pipe, intelImage->mt); - texImage->Data = NULL; - } - -#if 0 - UNLOCK_HARDWARE(intel); -#endif -} - - - - - -void -intelTexSubImage3D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - - intelTexSubimage(ctx, 3, - target, level, - xoffset, yoffset, zoffset, - width, height, depth, - format, type, pixels, packing, texObj, texImage); - -} - - - -void -intelTexSubImage2D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - - intelTexSubimage(ctx, 2, - target, level, - xoffset, yoffset, 0, - width, height, 1, - format, type, pixels, packing, texObj, texImage); - -} - - -void -intelTexSubImage1D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexSubimage(ctx, 1, - target, level, - xoffset, 0, 0, - width, 1, 1, - format, type, pixels, packing, texObj, texImage); - -} diff --git a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c b/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c deleted file mode 100644 index d0631a88f1..0000000000 --- a/src/mesa/drivers/dri/i915pipe/intel_tex_validate.c +++ /dev/null @@ -1,277 +0,0 @@ -#include "mtypes.h" -#include "macros.h" - -#include "state_tracker/st_mipmap_tree.h" -#include "intel_tex.h" - -#include "pipe/p_state.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -/** - * Compute which mipmap levels that really need to be sent to the hardware. - * This depends on the base image size, GL_TEXTURE_MIN_LOD, - * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. - */ -static void -intel_calculate_first_last_level(struct st_texture_object *intelObj) -{ - struct gl_texture_object *tObj = &intelObj->base; - const struct gl_texture_image *const baseImage = - tObj->Image[0][tObj->BaseLevel]; - - /* These must be signed values. MinLod and MaxLod can be negative numbers, - * and having firstLevel and lastLevel as signed prevents the need for - * extra sign checks. - */ - int firstLevel; - int lastLevel; - - /* Yes, this looks overly complicated, but it's all needed. - */ - switch (tObj->Target) { - case GL_TEXTURE_1D: - case GL_TEXTURE_2D: - case GL_TEXTURE_3D: - case GL_TEXTURE_CUBE_MAP: - if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) { - /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL. - */ - firstLevel = lastLevel = tObj->BaseLevel; - } - else { - firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); - firstLevel = MAX2(firstLevel, tObj->BaseLevel); - lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); - lastLevel = MAX2(lastLevel, tObj->BaseLevel); - lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); - lastLevel = MIN2(lastLevel, tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ - } - break; - case GL_TEXTURE_RECTANGLE_NV: - case GL_TEXTURE_4D_SGIS: - firstLevel = lastLevel = 0; - break; - default: - return; - } - - /* save these values */ - intelObj->firstLevel = firstLevel; - intelObj->lastLevel = lastLevel; -} - -static void -copy_image_data_to_tree(struct pipe_context *pipe, - struct st_texture_object *intelObj, - struct st_texture_image *intelImage) -{ - if (intelImage->mt) { - /* Copy potentially with the blitter: - */ - st_miptree_image_copy(pipe, - intelObj->mt, - intelImage->face, - intelImage->level, intelImage->mt); - - st_miptree_release(pipe, &intelImage->mt); - } - else { - assert(intelImage->base.Data != NULL); - - /* More straightforward upload. - */ - st_miptree_image_data(pipe, - intelObj->mt, - intelImage->face, - intelImage->level, - intelImage->base.Data, - intelImage->base.RowStride, - intelImage->base.RowStride * - intelImage->base.Height); - _mesa_align_free(intelImage->base.Data); - intelImage->base.Data = NULL; - } - - st_miptree_reference(&intelImage->mt, intelObj->mt); -} - - -/* - */ -GLuint -intel_finalize_mipmap_tree(GLcontext *ctx, - struct pipe_context *pipe, GLuint unit, - GLboolean *needFlush) -{ - struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; - struct st_texture_object *intelObj = st_texture_object(tObj); - int comp_byte = 0; - int cpp; - - GLuint face, i; - GLuint nr_faces = 0; - struct st_texture_image *firstImage; - - *needFlush = GL_FALSE; - - /* We know/require this is true by now: - */ - assert(intelObj->base._Complete); - - /* What levels must the tree include at a minimum? - */ - intel_calculate_first_last_level(intelObj); - firstImage = - st_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); - - /* Fallback case: - */ - if (firstImage->base.Border) { - if (intelObj->mt) { - st_miptree_release(pipe, &intelObj->mt); - } - return GL_FALSE; - } - - - /* If both firstImage and intelObj have a tree which can contain - * all active images, favour firstImage. Note that because of the - * completeness requirement, we know that the image dimensions - * will match. - */ - if (firstImage->mt && - firstImage->mt != intelObj->mt && - firstImage->mt->first_level <= intelObj->firstLevel && - firstImage->mt->last_level >= intelObj->lastLevel) { - - if (intelObj->mt) - st_miptree_release(pipe, &intelObj->mt); - - st_miptree_reference(&intelObj->mt, firstImage->mt); - } - - if (firstImage->base.IsCompressed) { - comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); - cpp = comp_byte; - } - else cpp = firstImage->base.TexFormat->TexelBytes; - - /* Check tree can hold all active levels. Check tree matches - * target, imageFormat, etc. - * - * XXX: For some layouts (eg i945?), the test might have to be - * first_level == firstLevel, as the tree isn't valid except at the - * original start level. Hope to get around this by - * programming minLod, maxLod, baseLevel into the hardware and - * leaving the tree alone. - */ - if (intelObj->mt && - (intelObj->mt->target != intelObj->base.Target || - intelObj->mt->internal_format != firstImage->base.InternalFormat || - intelObj->mt->first_level != intelObj->firstLevel || - intelObj->mt->last_level != intelObj->lastLevel || - intelObj->mt->width0 != firstImage->base.Width || - intelObj->mt->height0 != firstImage->base.Height || - intelObj->mt->depth0 != firstImage->base.Depth || - intelObj->mt->cpp != cpp || - intelObj->mt->compressed != firstImage->base.IsCompressed)) { - st_miptree_release(pipe, &intelObj->mt); - } - - - /* May need to create a new tree: - */ - if (!intelObj->mt) { - intelObj->mt = st_miptree_create(pipe, - intelObj->base.Target, - firstImage->base.InternalFormat, - intelObj->firstLevel, - intelObj->lastLevel, - firstImage->base.Width, - firstImage->base.Height, - firstImage->base.Depth, - cpp, - comp_byte); - } - - /* Pull in any images not in the object's tree: - */ - nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct st_texture_image *intelImage = - st_texture_image(intelObj->base.Image[face][i]); - - /* Need to import images in main memory or held in other trees. - */ - if (intelObj->mt != intelImage->mt) { - copy_image_data_to_tree(pipe, intelObj, intelImage); - *needFlush = GL_TRUE; - } - } - } - - /** - if (need_flush) - intel_batchbuffer_flush(intel->batch); - **/ - - return GL_TRUE; -} - - -#if 0 /* unused? */ -void -intel_tex_map_images(struct pipe_context *pipe, - struct st_texture_object *intelObj) -{ - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - GLuint face, i; - - DBG("%s\n", __FUNCTION__); - - for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct st_texture_image *intelImage = - st_texture_image(intelObj->base.Image[face][i]); - - if (intelImage->mt) { - intelImage->base.Data = - st_miptree_image_map(pipe, - intelImage->mt, - intelImage->face, - intelImage->level, - &intelImage->base.RowStride, - intelImage->base.ImageOffsets); - /* convert stride to texels, not bytes */ - intelImage->base.RowStride /= intelImage->mt->cpp; -/* intelImage->base.ImageStride /= intelImage->mt->cpp; */ - } - } - } -} - - - -void -intel_tex_unmap_images(struct pipe_context *pipe, - struct st_texture_object *intelObj) -{ - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - GLuint face, i; - - for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct st_texture_image *intelImage = - st_texture_image(intelObj->base.Image[face][i]); - - if (intelImage->mt) { - st_miptree_image_unmap(pipe, intelImage->mt); - intelImage->base.Data = NULL; - } - } - } -} -#endif diff --git a/src/mesa/drivers/dri/i915pipe/server/i830_common.h b/src/mesa/drivers/dri/i915pipe/server/i830_common.h deleted file mode 100644 index d4d58886ce..0000000000 --- a/src/mesa/drivers/dri/i915pipe/server/i830_common.h +++ /dev/null @@ -1,226 +0,0 @@ -/************************************************************************** - -Copyright 2001 VA Linux Systems Inc., Fremont, California. -Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas. - -All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining a -copy of this software and associated documentation files (the "Software"), -to deal in the Software without restriction, including without limitation -on the rights to use, copy, modify, merge, publish, distribute, sub -license, and/or sell copies of the Software, and to permit persons to whom -the Software is furnished to do so, subject to the following conditions: - -The above copyright notice and this permission notice (including the next -paragraph) shall be included in all copies or substantial portions of the -Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL -ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, -DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR -OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE -USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ - -#ifndef _I830_COMMON_H_ -#define _I830_COMMON_H_ - - -#define I830_NR_TEX_REGIONS 255 /* maximum due to use of chars for next/prev */ -#define I830_LOG_MIN_TEX_REGION_SIZE 14 - - -/* Driver specific DRM command indices - * NOTE: these are not OS specific, but they are driver specific - */ -#define DRM_I830_INIT 0x00 -#define DRM_I830_FLUSH 0x01 -#define DRM_I830_FLIP 0x02 -#define DRM_I830_BATCHBUFFER 0x03 -#define DRM_I830_IRQ_EMIT 0x04 -#define DRM_I830_IRQ_WAIT 0x05 -#define DRM_I830_GETPARAM 0x06 -#define DRM_I830_SETPARAM 0x07 -#define DRM_I830_ALLOC 0x08 -#define DRM_I830_FREE 0x09 -#define DRM_I830_INIT_HEAP 0x0a -#define DRM_I830_CMDBUFFER 0x0b -#define DRM_I830_DESTROY_HEAP 0x0c -#define DRM_I830_SET_VBLANK_PIPE 0x0d -#define DRM_I830_GET_VBLANK_PIPE 0x0e - -typedef struct { - enum { - I830_INIT_DMA = 0x01, - I830_CLEANUP_DMA = 0x02, - I830_RESUME_DMA = 0x03 - } func; - unsigned int mmio_offset; - int sarea_priv_offset; - unsigned int ring_start; - unsigned int ring_end; - unsigned int ring_size; - unsigned int front_offset; - unsigned int back_offset; - unsigned int depth_offset; - unsigned int w; - unsigned int h; - unsigned int pitch; - unsigned int pitch_bits; - unsigned int back_pitch; - unsigned int depth_pitch; - unsigned int cpp; - unsigned int chipset; -} drmI830Init; - -typedef struct { - drmTextureRegion texList[I830_NR_TEX_REGIONS+1]; - int last_upload; /* last time texture was uploaded */ - int last_enqueue; /* last time a buffer was enqueued */ - int last_dispatch; /* age of the most recently dispatched buffer */ - int ctxOwner; /* last context to upload state */ - int texAge; - int pf_enabled; /* is pageflipping allowed? */ - int pf_active; - int pf_current_page; /* which buffer is being displayed? */ - int perf_boxes; /* performance boxes to be displayed */ - int width, height; /* screen size in pixels */ - - drm_handle_t front_handle; - int front_offset; - int front_size; - - drm_handle_t back_handle; - int back_offset; - int back_size; - - drm_handle_t depth_handle; - int depth_offset; - int depth_size; - - drm_handle_t tex_handle; - int tex_offset; - int tex_size; - int log_tex_granularity; - int pitch; - int rotation; /* 0, 90, 180 or 270 */ - int rotated_offset; - int rotated_size; - int rotated_pitch; - int virtualX, virtualY; - - unsigned int front_tiled; - unsigned int back_tiled; - unsigned int depth_tiled; - unsigned int rotated_tiled; - unsigned int rotated2_tiled; - - int pipeA_x; - int pipeA_y; - int pipeA_w; - int pipeA_h; - int pipeB_x; - int pipeB_y; - int pipeB_w; - int pipeB_h; - - /* Triple buffering */ - drm_handle_t third_handle; - int third_offset; - int third_size; - unsigned int third_tiled; -} drmI830Sarea; - -/* Flags for perf_boxes - */ -#define I830_BOX_RING_EMPTY 0x1 /* populated by kernel */ -#define I830_BOX_FLIP 0x2 /* populated by kernel */ -#define I830_BOX_WAIT 0x4 /* populated by kernel & client */ -#define I830_BOX_TEXTURE_LOAD 0x8 /* populated by kernel */ -#define I830_BOX_LOST_CONTEXT 0x10 /* populated by client */ - - -typedef struct { - int start; /* agp offset */ - int used; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830BatchBuffer; - -typedef struct { - char *buf; /* agp offset */ - int sz; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830CmdBuffer; - -typedef struct { - int *irq_seq; -} drmI830IrqEmit; - -typedef struct { - int irq_seq; -} drmI830IrqWait; - -typedef struct { - int param; - int *value; -} drmI830GetParam; - -#define I830_PARAM_IRQ_ACTIVE 1 -#define I830_PARAM_ALLOW_BATCHBUFFER 2 - -typedef struct { - int param; - int value; -} drmI830SetParam; - -#define I830_SETPARAM_USE_MI_BATCHBUFFER_START 1 -#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY 2 -#define I830_SETPARAM_ALLOW_BATCHBUFFER 3 - - -/* A memory manager for regions of shared memory: - */ -#define I830_MEM_REGION_AGP 1 - -typedef struct { - int region; - int alignment; - int size; - int *region_offset; /* offset from start of fb or agp */ -} drmI830MemAlloc; - -typedef struct { - int region; - int region_offset; -} drmI830MemFree; - -typedef struct { - int region; - int size; - int start; -} drmI830MemInitHeap; - -typedef struct { - int region; -} drmI830MemDestroyHeap; - -#define DRM_I830_VBLANK_PIPE_A 1 -#define DRM_I830_VBLANK_PIPE_B 2 - -typedef struct { - int pipe; -} drmI830VBlankPipe; - -#endif /* _I830_DRM_H_ */ diff --git a/src/mesa/drivers/dri/i915pipe/server/i830_dri.h b/src/mesa/drivers/dri/i915pipe/server/i830_dri.h deleted file mode 100644 index c2a3af8cbf..0000000000 --- a/src/mesa/drivers/dri/i915pipe/server/i830_dri.h +++ /dev/null @@ -1,63 +0,0 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ - -#ifndef _I830_DRI_H -#define _I830_DRI_H - -#include "xf86drm.h" -#include "i830_common.h" - -#define I830_MAX_DRAWABLES 256 - -#define I830_MAJOR_VERSION 1 -#define I830_MINOR_VERSION 7 -#define I830_PATCHLEVEL 2 - -#define I830_REG_SIZE 0x80000 - -typedef struct _I830DRIRec { - drm_handle_t regs; - drmSize regsSize; - - drmSize unused1; /* backbufferSize */ - drm_handle_t unused2; /* backbuffer */ - - drmSize unused3; /* depthbufferSize */ - drm_handle_t unused4; /* depthbuffer */ - - drmSize unused5; /* rotatedSize */ - drm_handle_t unused6; /* rotatedbuffer */ - - drm_handle_t unused7; /* textures */ - int unused8; /* textureSize */ - - drm_handle_t unused9; /* agp_buffers */ - drmSize unused10; /* agp_buf_size */ - - int deviceID; - int width; - int height; - int mem; - int cpp; - int bitsPerPixel; - - int unused11[8]; /* was front/back/depth/rotated offset/pitch */ - - int unused12; /* logTextureGranularity */ - int unused13; /* textureOffset */ - - int irq; - int sarea_priv_offset; -} I830DRIRec, *I830DRIPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830ConfigPrivRec, *I830ConfigPrivPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830DRIContextRec, *I830DRIContextPtr; - - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/server/intel.h b/src/mesa/drivers/dri/i915pipe/server/intel.h deleted file mode 100644 index 6ea72499c1..0000000000 --- a/src/mesa/drivers/dri/i915pipe/server/intel.h +++ /dev/null @@ -1,331 +0,0 @@ -#ifndef _INTEL_H_ -#define _INTEL_H_ - -#include "xf86drm.h" /* drm_handle_t, etc */ - -/* Intel */ -#ifndef PCI_CHIP_I810 -#define PCI_CHIP_I810 0x7121 -#define PCI_CHIP_I810_DC100 0x7123 -#define PCI_CHIP_I810_E 0x7125 -#define PCI_CHIP_I815 0x1132 -#define PCI_CHIP_I810_BRIDGE 0x7120 -#define PCI_CHIP_I810_DC100_BRIDGE 0x7122 -#define PCI_CHIP_I810_E_BRIDGE 0x7124 -#define PCI_CHIP_I815_BRIDGE 0x1130 -#endif - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 - -#ifndef PCI_CHIP_I855_GM -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I855_GM_BRIDGE 0x3580 -#endif - -#ifndef PCI_CHIP_I865_G -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I865_G_BRIDGE 0x2570 -#endif - -#ifndef PCI_CHIP_I915_G -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I915_GM -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I915_GM_BRIDGE 0x2590 -#endif - -#ifndef PCI_CHIP_E7221_G -#define PCI_CHIP_E7221_G 0x258A -/* Same as I915_G_BRIDGE */ -#define PCI_CHIP_E7221_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I945_G -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_G_BRIDGE 0x2770 -#endif - -#ifndef PCI_CHIP_I945_GM -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GM_BRIDGE 0x27A0 -#endif - -#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 || \ - pI810->Chipset == PCI_CHIP_I810_DC100 || \ - pI810->Chipset == PCI_CHIP_I810_E) -#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815) -#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M) -#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G) -#define IS_I85X(pI810) (pI810->Chipset == PCI_CHIP_I855_GM) -#define IS_I852(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME)) -#define IS_I855(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME)) -#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G) - -#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G) -#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM) -#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G) -#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM) -#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810)) - -#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810)) - -#define I830_GMCH_CTRL 0x52 - -#define I830_GMCH_MEM_MASK 0x1 -#define I830_GMCH_MEM_64M 0x1 -#define I830_GMCH_MEM_128M 0 - -#define I830_GMCH_GMS_MASK 0x70 -#define I830_GMCH_GMS_DISABLED 0x00 -#define I830_GMCH_GMS_LOCAL 0x10 -#define I830_GMCH_GMS_STOLEN_512 0x20 -#define I830_GMCH_GMS_STOLEN_1024 0x30 -#define I830_GMCH_GMS_STOLEN_8192 0x40 - -#define I855_GMCH_GMS_MASK (0x7 << 4) -#define I855_GMCH_GMS_DISABLED 0x00 -#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) -#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) -#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) -#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) -#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) -#define I915G_GMCH_GMS_STOLEN_48M (0x6 << 4) -#define I915G_GMCH_GMS_STOLEN_64M (0x7 << 4) - -typedef unsigned char Bool; -#define TRUE 1 -#define FALSE 0 - -#define PIPE_NONE 0<<0 -#define PIPE_CRT 1<<0 -#define PIPE_TV 1<<1 -#define PIPE_DFP 1<<2 -#define PIPE_LFP 1<<3 -#define PIPE_CRT2 1<<4 -#define PIPE_TV2 1<<5 -#define PIPE_DFP2 1<<6 -#define PIPE_LFP2 1<<7 - -typedef struct _I830MemPool *I830MemPoolPtr; -typedef struct _I830MemRange *I830MemRangePtr; -typedef struct _I830MemRange { - long Start; - long End; - long Size; - unsigned long Physical; - unsigned long Offset; /* Offset of AGP-allocated portion */ - unsigned long Alignment; - drm_handle_t Key; - unsigned long Pitch; // add pitch - I830MemPoolPtr Pool; -} I830MemRange; - -typedef struct _I830MemPool { - I830MemRange Total; - I830MemRange Free; - I830MemRange Fixed; - I830MemRange Allocated; -} I830MemPool; - -typedef struct { - int tail_mask; - I830MemRange mem; - unsigned char *virtual_start; - int head; - int tail; - int space; -} I830RingBuffer; - -typedef struct _I830Rec { - unsigned char *MMIOBase; - unsigned char *FbBase; - int cpp; - uint32_t aper_size; - unsigned int bios_version; - - /* These are set in PreInit and never changed. */ - long FbMapSize; - long TotalVideoRam; - I830MemRange StolenMemory; /* pre-allocated memory */ - long BIOSMemorySize; /* min stolen pool size */ - int BIOSMemSizeLoc; - - /* These change according to what has been allocated. */ - long FreeMemory; - I830MemRange MemoryAperture; - I830MemPool StolenPool; - long allocatedMemory; - - /* Regions allocated either from the above pools, or from agpgart. */ - /* for single and dual head configurations */ - I830MemRange FrontBuffer; - I830MemRange FrontBuffer2; - I830MemRange Scratch; - I830MemRange Scratch2; - - I830RingBuffer *LpRing; - - I830MemRange BackBuffer; - I830MemRange DepthBuffer; - I830MemRange TexMem; - int TexGranularity; - I830MemRange ContextMem; - int drmMinor; - Bool have3DWindows; - - Bool NeedRingBufferLow; - Bool allowPageFlip; - Bool disableTiling; - - int Chipset; - unsigned long LinearAddr; - unsigned long MMIOAddr; - - drmSize registerSize; /**< \brief MMIO register map size */ - drm_handle_t registerHandle; /**< \brief MMIO register map handle */ - // IOADDRESS ioBase; - int irq; /**< \brief IRQ number */ - int GttBound; - - drm_handle_t ring_map; - unsigned int Fence[8]; - -} I830Rec; - -/* - * 12288 is set as the maximum, chosen because it is enough for - * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare. - */ -#define I830_MAXIMUM_VBIOS_MEM 12288 -#define I830_DEFAULT_VIDEOMEM_2D (MB(32) / 1024) -#define I830_DEFAULT_VIDEOMEM_3D (MB(64) / 1024) - -/* Flags for memory allocation function */ -#define FROM_ANYWHERE 0x00000000 -#define FROM_POOL_ONLY 0x00000001 -#define FROM_NEW_ONLY 0x00000002 -#define FROM_MASK 0x0000000f - -#define ALLOCATE_AT_TOP 0x00000010 -#define ALLOCATE_AT_BOTTOM 0x00000020 -#define FORCE_GAPS 0x00000040 - -#define NEED_PHYSICAL_ADDR 0x00000100 -#define ALIGN_BOTH_ENDS 0x00000200 -#define FORCE_LOW 0x00000400 - -#define ALLOC_NO_TILING 0x00001000 -#define ALLOC_INITIAL 0x00002000 - -#define ALLOCATE_DRY_RUN 0x80000000 - -/* Chipset registers for VIDEO BIOS memory RW access */ -#define _855_DRAM_RW_CONTROL 0x58 -#define _845_DRAM_RW_CONTROL 0x90 -#define DRAM_WRITE 0x33330000 - -#define KB(x) ((x) * 1024) -#define MB(x) ((x) * KB(1024)) - -#define GTT_PAGE_SIZE KB(4) -#define ROUND_TO(x, y) (((x) + (y) - 1) / (y) * (y)) -#define ROUND_DOWN_TO(x, y) ((x) / (y) * (y)) -#define ROUND_TO_PAGE(x) ROUND_TO((x), GTT_PAGE_SIZE) -#define ROUND_TO_MB(x) ROUND_TO((x), MB(1)) -#define PRIMARY_RINGBUFFER_SIZE KB(128) - - -/* Ring buffer registers, p277, overview p19 - */ -#define LP_RING 0x2030 -#define HP_RING 0x2040 - -#define RING_TAIL 0x00 -#define TAIL_ADDR 0x000FFFF8 -#define I830_TAIL_MASK 0x001FFFF8 - -#define RING_HEAD 0x04 -#define HEAD_WRAP_COUNT 0xFFE00000 -#define HEAD_WRAP_ONE 0x00200000 -#define HEAD_ADDR 0x001FFFFC -#define I830_HEAD_MASK 0x001FFFFC - -#define RING_START 0x08 -#define START_ADDR 0x03FFFFF8 -#define I830_RING_START_MASK 0xFFFFF000 - -#define RING_LEN 0x0C -#define RING_NR_PAGES 0x001FF000 -#define I830_RING_NR_PAGES 0x001FF000 -#define RING_REPORT_MASK 0x00000006 -#define RING_REPORT_64K 0x00000002 -#define RING_REPORT_128K 0x00000004 -#define RING_NO_REPORT 0x00000000 -#define RING_VALID_MASK 0x00000001 -#define RING_VALID 0x00000001 -#define RING_INVALID 0x00000000 - - -/* Fence/Tiling ranges [0..7] - */ -#define FENCE 0x2000 -#define FENCE_NR 8 - -#define I915G_FENCE_START_MASK 0x0ff00000 - -#define I830_FENCE_START_MASK 0x07f80000 - -#define FENCE_START_MASK 0x03F80000 -#define FENCE_X_MAJOR 0x00000000 -#define FENCE_Y_MAJOR 0x00001000 -#define FENCE_SIZE_MASK 0x00000700 -#define FENCE_SIZE_512K 0x00000000 -#define FENCE_SIZE_1M 0x00000100 -#define FENCE_SIZE_2M 0x00000200 -#define FENCE_SIZE_4M 0x00000300 -#define FENCE_SIZE_8M 0x00000400 -#define FENCE_SIZE_16M 0x00000500 -#define FENCE_SIZE_32M 0x00000600 -#define FENCE_SIZE_64M 0x00000700 -#define I915G_FENCE_SIZE_1M 0x00000000 -#define I915G_FENCE_SIZE_2M 0x00000100 -#define I915G_FENCE_SIZE_4M 0x00000200 -#define I915G_FENCE_SIZE_8M 0x00000300 -#define I915G_FENCE_SIZE_16M 0x00000400 -#define I915G_FENCE_SIZE_32M 0x00000500 -#define I915G_FENCE_SIZE_64M 0x00000600 -#define I915G_FENCE_SIZE_128M 0x00000700 -#define FENCE_PITCH_1 0x00000000 -#define FENCE_PITCH_2 0x00000010 -#define FENCE_PITCH_4 0x00000020 -#define FENCE_PITCH_8 0x00000030 -#define FENCE_PITCH_16 0x00000040 -#define FENCE_PITCH_32 0x00000050 -#define FENCE_PITCH_64 0x00000060 -#define FENCE_VALID 0x00000001 - -#include - -# define MMIO_IN8(base, offset) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) -# define MMIO_IN32(base, offset) \ - read_MMIO_LE32(base, offset) -# define MMIO_OUT8(base, offset, val) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val) -# define MMIO_OUT32(base, offset, val) \ - *(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val) - - - /* Memory mapped register access macros */ -#define INREG8(addr) MMIO_IN8(MMIO, addr) -#define INREG(addr) MMIO_IN32(MMIO, addr) -#define OUTREG8(addr, val) MMIO_OUT8(MMIO, addr, val) -#define OUTREG(addr, val) MMIO_OUT32(MMIO, addr, val) - -#define DSPABASE 0x70184 - -#endif diff --git a/src/mesa/drivers/dri/i915pipe/server/intel_dri.c b/src/mesa/drivers/dri/i915pipe/server/intel_dri.c deleted file mode 100644 index e49c4214ad..0000000000 --- a/src/mesa/drivers/dri/i915pipe/server/intel_dri.c +++ /dev/null @@ -1,1306 +0,0 @@ -/** - * \file server/intel_dri.c - * \brief File to perform the device-specific initialization tasks typically - * done in the X server. - * - * Here they are converted to run in the client (or perhaps a standalone - * process), and to work with the frame buffer device rather than the X - * server infrastructure. - * - * Copyright (C) 2006 Dave Airlie (airlied@linux.ie) - - Permission is hereby granted, free of charge, to any person obtaining a - copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sub license, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial portions - of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR - ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. -*/ - -#include -#include -#include -#include -#include - -#include "driver.h" -#include "drm.h" - -#include "intel.h" -#include "i830_dri.h" - -#include "memops.h" -#include "pciaccess.h" - -static size_t drm_page_size; -static int nextTile = 0; -#define xf86DrvMsg(...) do {} while(0) - -static const int pitches[] = { - 128 * 8, - 128 * 16, - 128 * 32, - 128 * 64, - 0 -}; - -static Bool I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea); - -static unsigned long -GetBestTileAlignment(unsigned long size) -{ - unsigned long i; - - for (i = KB(512); i < size; i <<= 1) - ; - - if (i > MB(64)) - i = MB(64); - - return i; -} - -static void SetFenceRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - unsigned char *MMIO = ctx->MMIOAddress; - - for (i = 0; i < 8; i++) { - OUTREG(FENCE + i * 4, pI830->Fence[i]); - // if (I810_DEBUG & DEBUG_VERBOSE_VGA) - fprintf(stderr,"Fence Register : %x\n", pI830->Fence[i]); - } -} - -/* Tiled memory is good... really, really good... - * - * Need to make it less likely that we miss out on this - probably - * need to move the frontbuffer away from the 'guarenteed' alignment - * of the first memory segment, or perhaps allocate a discontigous - * framebuffer to get more alignment 'sweet spots'. - */ -static void -SetFence(const DRIDriverContext *ctx, I830Rec *pI830, - int nr, unsigned int start, unsigned int pitch, - unsigned int size) -{ - unsigned int val; - unsigned int fence_mask = 0; - unsigned int fence_pitch; - - if (nr < 0 || nr > 7) { - fprintf(stderr, - "SetFence: fence %d out of range\n",nr); - return; - } - - pI830->Fence[nr] = 0; - - if (IS_I9XX(pI830)) - fence_mask = ~I915G_FENCE_START_MASK; - else - fence_mask = ~I830_FENCE_START_MASK; - - if (start & fence_mask) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not %s aligned\n", - nr, start, (IS_I9XX(pI830)) ? "1MB" : "512k"); - return; - } - - if (start % size) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not size (%dk) aligned\n", - nr, start, size / 1024); - return; - } - - if (pitch & 127) { - fprintf(stderr, - "SetFence: %d: pitch (%d) not a multiple of 128 bytes\n", - nr, pitch); - return; - } - - val = (start | FENCE_X_MAJOR | FENCE_VALID); - - if (IS_I9XX(pI830)) { - switch (size) { - case MB(1): - val |= I915G_FENCE_SIZE_1M; - break; - case MB(2): - val |= I915G_FENCE_SIZE_2M; - break; - case MB(4): - val |= I915G_FENCE_SIZE_4M; - break; - case MB(8): - val |= I915G_FENCE_SIZE_8M; - break; - case MB(16): - val |= I915G_FENCE_SIZE_16M; - break; - case MB(32): - val |= I915G_FENCE_SIZE_32M; - break; - case MB(64): - val |= I915G_FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } else { - switch (size) { - case KB(512): - val |= FENCE_SIZE_512K; - break; - case MB(1): - val |= FENCE_SIZE_1M; - break; - case MB(2): - val |= FENCE_SIZE_2M; - break; - case MB(4): - val |= FENCE_SIZE_4M; - break; - case MB(8): - val |= FENCE_SIZE_8M; - break; - case MB(16): - val |= FENCE_SIZE_16M; - break; - case MB(32): - val |= FENCE_SIZE_32M; - break; - case MB(64): - val |= FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } - - if (IS_I9XX(pI830)) - fence_pitch = pitch / 512; - else - fence_pitch = pitch / 128; - - switch (fence_pitch) { - case 1: - val |= FENCE_PITCH_1; - break; - case 2: - val |= FENCE_PITCH_2; - break; - case 4: - val |= FENCE_PITCH_4; - break; - case 8: - val |= FENCE_PITCH_8; - break; - case 16: - val |= FENCE_PITCH_16; - break; - case 32: - val |= FENCE_PITCH_32; - break; - case 64: - val |= FENCE_PITCH_64; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal pitch (%d)\n", nr, pitch); - return; - } - - pI830->Fence[nr] = val; -} - -static Bool -MakeTiles(const DRIDriverContext *ctx, I830Rec *pI830, I830MemRange *pMem) -{ - int pitch, ntiles, i; - - pitch = pMem->Pitch * ctx->cpp; - /* - * Simply try to break the region up into at most four pieces of size - * equal to the alignment. - */ - ntiles = ROUND_TO(pMem->Size, pMem->Alignment) / pMem->Alignment; - if (ntiles >= 4) { - return FALSE; - } - - for (i = 0; i < ntiles; i++, nextTile++) { - SetFence(ctx, pI830, nextTile, pMem->Start + i * pMem->Alignment, - pitch, pMem->Alignment); - } - return TRUE; -} - -static void I830SetupMemoryTiling(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - - /* Clear out */ - for (i = 0; i < 8; i++) - pI830->Fence[i] = 0; - - nextTile = 0; - - if (pI830->BackBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->BackBuffer))) { - fprintf(stderr, - "Activating tiled memory for the back buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the back buffer.\n"); - pI830->allowPageFlip = FALSE; - } - } - - if (pI830->DepthBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->DepthBuffer))) { - fprintf(stderr, - "Activating tiled memory for the depth buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the depth buffer.\n"); - } - } - - return; -} - -static int I830DetectMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - struct pci_device host_bridge, ig_dev; - uint32_t gmch_ctrl; - int memsize = 0; - int range; - uint32_t aper_size; - uint32_t membase2 = 0; - - memset(&host_bridge, 0, sizeof(host_bridge)); - memset(&ig_dev, 0, sizeof(ig_dev)); - - ig_dev.dev = 2; - - pci_device_cfg_read_u32(&host_bridge, &gmch_ctrl, I830_GMCH_CTRL); - - if (IS_I830(pI830) || IS_845G(pI830)) { - if ((gmch_ctrl & I830_GMCH_MEM_MASK) == I830_GMCH_MEM_128M) { - aper_size = 0x80000000; - } else { - aper_size = 0x40000000; - } - } else { - if (IS_I9XX(pI830)) { - int ret; - ret = pci_device_cfg_read_u32(&ig_dev, &membase2, 0x18); - if (membase2 & 0x08000000) - aper_size = 0x8000000; - else - aper_size = 0x10000000; - - fprintf(stderr,"aper size is %08X %08x %d\n", aper_size, membase2, ret); - } else - aper_size = 0x8000000; - } - - pI830->aper_size = aper_size; - - - /* We need to reduce the stolen size, by the GTT and the popup. - * The GTT varying according the the FbMapSize and the popup is 4KB */ - range = (ctx->shared.fbSize / (1024*1024)) + 4; - - if (IS_I85X(pI830) || IS_I865G(pI830) || IS_I9XX(pI830)) { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I855_GMCH_GMS_STOLEN_1M: - memsize = MB(1) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_4M: - memsize = MB(4) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_8M: - memsize = MB(8) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_16M: - memsize = MB(16) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_32M: - memsize = MB(32) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_48M: - if (IS_I9XX(pI830)) - memsize = MB(48) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_64M: - if (IS_I9XX(pI830)) - memsize = MB(64) - KB(range); - break; - } - } else { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I830_GMCH_GMS_STOLEN_512: - memsize = KB(512) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_1024: - memsize = MB(1) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_8192: - memsize = MB(8) - KB(range); - break; - case I830_GMCH_GMS_LOCAL: - memsize = 0; - xf86DrvMsg(pScrn->scrnIndex, X_WARNING, - "Local memory found, but won't be used.\n"); - break; - } - } - if (memsize > 0) { - fprintf(stderr, - "detected %d kB stolen memory.\n", memsize / 1024); - } else { - fprintf(stderr, - "no video memory detected.\n"); - } - return memsize; -} - -static int AgpInit(const DRIDriverContext *ctx, I830Rec *info) -{ - unsigned long mode = 0x4; - - if (drmAgpAcquire(ctx->drmFD) < 0) { - fprintf(stderr, "[gart] AGP not available\n"); - return 0; - } - - if (drmAgpEnable(ctx->drmFD, mode) < 0) { - fprintf(stderr, "[gart] AGP not enabled\n"); - drmAgpRelease(ctx->drmFD); - return 0; - } - else - fprintf(stderr, "[gart] AGP enabled at %dx\n", ctx->agpmode); - - return 1; -} - -/* - * Allocate memory from the given pool. Grow the pool if needed and if - * possible. - */ -static unsigned long -AllocFromPool(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, - long size, unsigned long alignment, int flags) -{ - long needed, start, end; - - if (!result || !pool || !size) - return 0; - - /* Calculate how much space is needed. */ - if (alignment <= GTT_PAGE_SIZE) - needed = size; - else { - start = ROUND_TO(pool->Free.Start, alignment); - end = ROUND_TO(start + size, alignment); - needed = end - pool->Free.Start; - } - if (needed > pool->Free.Size) { - return 0; - } - - result->Start = ROUND_TO(pool->Free.Start, alignment); - pool->Free.Start += needed; - result->End = pool->Free.Start; - - pool->Free.Size = pool->Free.End - pool->Free.Start; - result->Size = result->End - result->Start; - result->Pool = pool; - result->Alignment = alignment; - return needed; -} - -static unsigned long AllocFromAGP(const DRIDriverContext *ctx, I830Rec *pI830, long size, unsigned long alignment, I830MemRange *result) -{ - unsigned long start, end; - unsigned long newApStart, newApEnd; - int ret; - if (!result || !size) - return 0; - - if (!alignment) - alignment = 4; - - start = ROUND_TO(pI830->MemoryAperture.Start, alignment); - end = ROUND_TO(start + size, alignment); - newApStart = end; - newApEnd = pI830->MemoryAperture.End; - - ret=drmAgpAlloc(ctx->drmFD, size, 0, &(result->Physical), (drm_handle_t *)&(result->Key)); - - if (ret) - { - fprintf(stderr,"drmAgpAlloc failed %d\n", ret); - return 0; - } - pI830->allocatedMemory += size; - pI830->MemoryAperture.Start = newApStart; - pI830->MemoryAperture.End = newApEnd; - pI830->MemoryAperture.Size = newApEnd - newApStart; - // pI830->FreeMemory -= size; - result->Start = start; - result->End = start + size; - result->Size = size; - result->Offset = start; - result->Alignment = alignment; - result->Pool = NULL; - - return size; -} - -unsigned long -I830AllocVidMem(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, long size, - unsigned long alignment, int flags) -{ - unsigned long ret; - - if (!result) - return 0; - - /* Make sure these are initialised. */ - result->Size = 0; - result->Key = -1; - - if (!size) { - return 0; - } - - if (pool->Free.Size < size) { - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - else { - ret = AllocFromPool(ctx, pI830, result, pool, size, alignment, flags); - if (ret == 0) - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - return ret; -} - -static Bool BindAgpRange(const DRIDriverContext *ctx, I830MemRange *mem) -{ - if (!mem) - return FALSE; - - if (mem->Key == -1) - return TRUE; - - return !drmAgpBind(ctx->drmFD, mem->Key, mem->Offset); -} - -/* simple memory allocation routines needed */ -/* put ring buffer in low memory */ -/* need to allocate front, back, depth buffers aligned correctly, - allocate ring buffer, -*/ - -/* */ -static Bool -I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned long size, ret; - unsigned long lines, lineSize, align; - - /* allocate ring buffer */ - memset(pI830->LpRing, 0, sizeof(I830RingBuffer)); - pI830->LpRing->mem.Key = -1; - - size = PRIMARY_RINGBUFFER_SIZE; - - ret = I830AllocVidMem(ctx, pI830, &pI830->LpRing->mem, &pI830->StolenPool, size, 0x1000, 0); - - if (ret != size) - { - fprintf(stderr,"unable to allocate ring buffer %ld\n", ret); - return FALSE; - } - - pI830->LpRing->tail_mask = pI830->LpRing->mem.Size - 1; - - - /* allocate front buffer */ - memset(&(pI830->FrontBuffer), 0, sizeof(pI830->FrontBuffer)); - pI830->FrontBuffer.Key = -1; - pI830->FrontBuffer.Pitch = ctx->shared.virtualWidth; - - align = KB(512); - - lineSize = ctx->shared.virtualWidth * ctx->cpp; - lines = (ctx->shared.virtualHeight + 15) / 16 * 16; - size = lineSize * lines; - size = ROUND_TO_PAGE(size); - - align = GetBestTileAlignment(size); - - ret = I830AllocVidMem(ctx, pI830, &pI830->FrontBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate front buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->BackBuffer), 0, sizeof(pI830->BackBuffer)); - pI830->BackBuffer.Key = -1; - pI830->BackBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->BackBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate back buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->DepthBuffer), 0, sizeof(pI830->DepthBuffer)); - pI830->DepthBuffer.Key = -1; - pI830->DepthBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->DepthBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate depth buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->ContextMem), 0, sizeof(pI830->ContextMem)); - pI830->ContextMem.Key = -1; - size = KB(32); - - ret = I830AllocVidMem(ctx, pI830, &pI830->ContextMem, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate context buffer %ld\n", ret); - return FALSE; - } - -#if 0 - memset(&(pI830->TexMem), 0, sizeof(pI830->TexMem)); - pI830->TexMem.Key = -1; - - size = 32768 * 1024; - ret = AllocFromAGP(ctx, pI830, size, align, &pI830->TexMem); - if (ret < size) - { - fprintf(stderr,"unable to allocate texture memory %ld\n", ret); - return FALSE; - } -#endif - - return TRUE; -} - -static Bool -I830BindMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - if (!BindAgpRange(ctx, &pI830->LpRing->mem)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->FrontBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->BackBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->DepthBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->ContextMem)) - return FALSE; -#if 0 - if (!BindAgpRange(ctx, &pI830->TexMem)) - return FALSE; -#endif - return TRUE; -} - -static void SetupDRIMM(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned long aperEnd = ROUND_DOWN_TO(pI830->aper_size, GTT_PAGE_SIZE) / GTT_PAGE_SIZE; - unsigned long aperStart = ROUND_TO(pI830->aper_size - KB(32768), GTT_PAGE_SIZE) / GTT_PAGE_SIZE; - - fprintf(stderr, "aper size is %08X\n", ctx->shared.fbSize); - if (drmMMInit(ctx->drmFD, aperStart, aperEnd - aperStart, DRM_BO_MEM_TT)) { - fprintf(stderr, - "DRM MM Initialization Failed\n"); - } else { - fprintf(stderr, - "DRM MM Initialized at offset 0x%lx length %d page\n", aperStart, aperEnd-aperStart); - } - -} - -static Bool -I830CleanupDma(const DRIDriverContext *ctx) -{ - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_CLEANUP_DMA; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, "I830 Dma Cleanup Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830InitDma(const DRIDriverContext *ctx, I830Rec *pI830) -{ - I830RingBuffer *ring = pI830->LpRing; - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_INIT_DMA; - - info.ring_start = ring->mem.Start + pI830->LinearAddr; - info.ring_end = ring->mem.End + pI830->LinearAddr; - info.ring_size = ring->mem.Size; - - info.mmio_offset = (unsigned int)ctx->MMIOStart; - - info.sarea_priv_offset = sizeof(drm_sarea_t); - - info.front_offset = pI830->FrontBuffer.Start; - info.back_offset = pI830->BackBuffer.Start; - info.depth_offset = pI830->DepthBuffer.Start; - info.w = ctx->shared.virtualWidth; - info.h = ctx->shared.virtualHeight; - info.pitch = ctx->shared.virtualWidth; - info.back_pitch = pI830->BackBuffer.Pitch; - info.depth_pitch = pI830->DepthBuffer.Pitch; - info.cpp = ctx->cpp; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, - "I830 Dma Initialization Failed\n"); - return FALSE; - } - - return TRUE; -} - -static int I830CheckDRMVersion( const DRIDriverContext *ctx, - I830Rec *pI830 ) -{ - drmVersionPtr version; - - version = drmGetVersion(ctx->drmFD); - - if (version) { - int req_minor, req_patch; - - req_minor = 4; - req_patch = 0; - - if (version->version_major != 1 || - version->version_minor < req_minor || - (version->version_minor == req_minor && - version->version_patchlevel < req_patch)) { - /* Incompatible drm version */ - fprintf(stderr, - "[dri] I830DRIScreenInit failed because of a version " - "mismatch.\n" - "[dri] i915.o kernel module version is %d.%d.%d " - "but version 1.%d.%d or newer is needed.\n" - "[dri] Disabling DRI.\n", - version->version_major, - version->version_minor, - version->version_patchlevel, - req_minor, - req_patch); - drmFreeVersion(version); - return 0; - } - - pI830->drmMinor = version->version_minor; - drmFreeVersion(version); - } - return 1; -} - -static void -I830SetRingRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned int itemp; - unsigned char *MMIO = ctx->MMIOAddress; - - OUTREG(LP_RING + RING_LEN, 0); - OUTREG(LP_RING + RING_TAIL, 0); - OUTREG(LP_RING + RING_HEAD, 0); - - if ((long)(pI830->LpRing->mem.Start & I830_RING_START_MASK) != - pI830->LpRing->mem.Start) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer start (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Start, I830_RING_START_MASK); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = pI830->LpRing->mem.Start & I830_RING_START_MASK; - OUTREG(LP_RING + RING_START, itemp); - - if (((pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES) != - pI830->LpRing->mem.Size - 4096) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer size - 4096 (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Size - 4096, - I830_RING_NR_PAGES); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = (pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES; - itemp |= (RING_NO_REPORT | RING_VALID); - OUTREG(LP_RING + RING_LEN, itemp); - - pI830->LpRing->head = INREG(LP_RING + RING_HEAD) & I830_HEAD_MASK; - pI830->LpRing->tail = INREG(LP_RING + RING_TAIL); - pI830->LpRing->space = pI830->LpRing->head - (pI830->LpRing->tail + 8); - if (pI830->LpRing->space < 0) - pI830->LpRing->space += pI830->LpRing->mem.Size; - - SetFenceRegs(ctx, pI830); - - /* RESET THE DISPLAY PIPE TO POINT TO THE FRONTBUFFER - hacky - hacky hacky */ - OUTREG(DSPABASE, pI830->FrontBuffer.Start + pI830->LinearAddr); - -} - -static Bool -I830SetParam(const DRIDriverContext *ctx, int param, int value) -{ - drmI830SetParam sp; - - memset(&sp, 0, sizeof(sp)); - sp.param = param; - sp.value = value; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_SETPARAM, &sp, sizeof(sp))) { - fprintf(stderr, "I830 SetParam Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - fprintf(stderr, - "[drm] Mapping front buffer\n"); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->front_offset + pI830->LinearAddr), - sarea->front_size, - DRM_FRAME_BUFFER, /*DRM_AGP,*/ - 0, - &sarea->front_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(front_handle) failed. Disabling DRI\n"); - return FALSE; - } - ctx->shared.hFrameBuffer = sarea->front_handle; - ctx->shared.fbSize = sarea->front_size; - fprintf(stderr, "[drm] Front Buffer = 0x%08x\n", - sarea->front_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->back_offset), - sarea->back_size, DRM_AGP, 0, - &sarea->back_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(back_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Back Buffer = 0x%08x\n", - sarea->back_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->depth_offset, - sarea->depth_size, DRM_AGP, 0, - &sarea->depth_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(depth_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Depth Buffer = 0x%08x\n", - sarea->depth_handle); - -#if 0 - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->tex_offset, - sarea->tex_size, DRM_AGP, 0, - &sarea->tex_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(tex_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] textures = 0x%08x\n", - sarea->tex_handle); -#endif - return TRUE; -} - - -static void -I830DRIUnmapScreenRegions(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ -#if 1 - if (sarea->front_handle) { - drmRmMap(ctx->drmFD, sarea->front_handle); - sarea->front_handle = 0; - } -#endif - if (sarea->back_handle) { - drmRmMap(ctx->drmFD, sarea->back_handle); - sarea->back_handle = 0; - } - if (sarea->depth_handle) { - drmRmMap(ctx->drmFD, sarea->depth_handle); - sarea->depth_handle = 0; - } - if (sarea->tex_handle) { - drmRmMap(ctx->drmFD, sarea->tex_handle); - sarea->tex_handle = 0; - } -} - -static Bool -I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - if (drmAddMap(ctx->drmFD, - (drm_handle_t)pI830->LpRing->mem.Start, - pI830->LpRing->mem.Size, DRM_AGP, 0, - &pI830->ring_map) < 0) { - fprintf(stderr, - "[drm] drmAddMap(ring_map) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] ring buffer = 0x%08x\n", - pI830->ring_map); - - if (I830InitDma(ctx, pI830) == FALSE) { - return FALSE; - } - - /* init to zero to be safe */ - - I830DRIMapScreenRegions(ctx, pI830, sarea); - SetupDRIMM(ctx, pI830); - - if (ctx->pciDevice != PCI_CHIP_845_G && - ctx->pciDevice != PCI_CHIP_I830_M) { - I830SetParam(ctx, I830_SETPARAM_USE_MI_BATCHBUFFER_START, 1 ); - } - - /* Okay now initialize the dma engine */ - { - pI830->irq = drmGetInterruptFromBusID(ctx->drmFD, - ctx->pciBus, - ctx->pciDevice, - ctx->pciFunc); - - if (drmCtlInstHandler(ctx->drmFD, pI830->irq)) { - fprintf(stderr, - "[drm] failure adding irq handler\n"); - pI830->irq = 0; - return FALSE; - } - else - fprintf(stderr, - "[drm] dma control initialized, using IRQ %d\n", - pI830->irq); - } - - fprintf(stderr, "[dri] visual configs initialized\n"); - - return TRUE; -} - -static Bool -I830ClearScreen(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - /* need to drmMap front and back buffers and zero them */ - drmAddress map_addr; - int ret; - - ret = drmMap(ctx->drmFD, - sarea->front_handle, - sarea->front_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map front buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->front_size); - drmUnmap(map_addr, sarea->front_size); - - - ret = drmMap(ctx->drmFD, - sarea->back_handle, - sarea->back_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map back buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->back_size); - drmUnmap(map_addr, sarea->back_size); - - return TRUE; -} - -static Bool -I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) - -{ - I830DRIPtr pI830DRI; - drmI830Sarea *pSAREAPriv; - int err; - - drm_page_size = getpagesize(); - - pI830->registerSize = ctx->MMIOSize; - /* This is a hack for now. We have to have more than a 4k page here - * because of the size of the state. However, the state should be - * in a per-context mapping. This will be added in the Mesa 3.5 port - * of the I830 driver. - */ - ctx->shared.SAREASize = SAREA_MAX; - - /* Note that drmOpen will try to load the kernel module, if needed. */ - ctx->drmFD = drmOpen("i915", NULL ); - if (ctx->drmFD < 0) { - fprintf(stderr, "[drm] drmOpen failed\n"); - return 0; - } - - if ((err = drmSetBusid(ctx->drmFD, ctx->pciBusID)) < 0) { - fprintf(stderr, "[drm] drmSetBusid failed (%d, %s), %s\n", - ctx->drmFD, ctx->pciBusID, strerror(-err)); - return 0; - } - - if (drmAddMap( ctx->drmFD, - 0, - ctx->shared.SAREASize, - DRM_SHM, - DRM_CONTAINS_LOCK, - &ctx->shared.hSAREA) < 0) - { - fprintf(stderr, "[drm] drmAddMap failed\n"); - return 0; - } - - fprintf(stderr, "[drm] added %d byte SAREA at 0x%08x\n", - ctx->shared.SAREASize, ctx->shared.hSAREA); - - if (drmMap( ctx->drmFD, - ctx->shared.hSAREA, - ctx->shared.SAREASize, - (drmAddressPtr)(&ctx->pSAREA)) < 0) - { - fprintf(stderr, "[drm] drmMap failed\n"); - return 0; - - } - - memset(ctx->pSAREA, 0, ctx->shared.SAREASize); - fprintf(stderr, "[drm] mapped SAREA 0x%08x to %p, size %d\n", - ctx->shared.hSAREA, ctx->pSAREA, ctx->shared.SAREASize); - - - if (drmAddMap(ctx->drmFD, - ctx->MMIOStart, - ctx->MMIOSize, - DRM_REGISTERS, - DRM_READ_ONLY, - &pI830->registerHandle) < 0) { - fprintf(stderr, "[drm] drmAddMap mmio failed\n"); - return 0; - } - fprintf(stderr, - "[drm] register handle = 0x%08x\n", pI830->registerHandle); - - - if (!I830CheckDRMVersion(ctx, pI830)) { - return FALSE; - } - - /* Create a 'server' context so we can grab the lock for - * initialization ioctls. - */ - if ((err = drmCreateContext(ctx->drmFD, &ctx->serverContext)) != 0) { - fprintf(stderr, "%s: drmCreateContext failed %d\n", __FUNCTION__, err); - return 0; - } - - DRM_LOCK(ctx->drmFD, ctx->pSAREA, ctx->serverContext, 0); - - /* Initialize the SAREA private data structure */ - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - memset(pSAREAPriv, 0, sizeof(*pSAREAPriv)); - - pI830->StolenMemory.Size = I830DetectMemory(ctx, pI830); - pI830->StolenMemory.Start = 0; - pI830->StolenMemory.End = pI830->StolenMemory.Size; - - pI830->MemoryAperture.Start = pI830->StolenMemory.End; - pI830->MemoryAperture.End = KB(40000); - pI830->MemoryAperture.Size = pI830->MemoryAperture.End - pI830->MemoryAperture.Start; - - pI830->StolenPool.Fixed = pI830->StolenMemory; - pI830->StolenPool.Total = pI830->StolenMemory; - pI830->StolenPool.Free = pI830->StolenPool.Total; - pI830->FreeMemory = pI830->StolenPool.Total.Size; - - if (!AgpInit(ctx, pI830)) - return FALSE; - - if (I830AllocateMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - if (I830BindMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - pSAREAPriv->rotated_offset = -1; - pSAREAPriv->rotated_size = 0; - pSAREAPriv->rotated_pitch = ctx->shared.virtualWidth; - - pSAREAPriv->front_offset = pI830->FrontBuffer.Start; - pSAREAPriv->front_size = pI830->FrontBuffer.Size; - pSAREAPriv->width = ctx->shared.virtualWidth; - pSAREAPriv->height = ctx->shared.virtualHeight; - pSAREAPriv->pitch = ctx->shared.virtualWidth; - pSAREAPriv->virtualX = ctx->shared.virtualWidth; - pSAREAPriv->virtualY = ctx->shared.virtualHeight; - pSAREAPriv->back_offset = pI830->BackBuffer.Start; - pSAREAPriv->back_size = pI830->BackBuffer.Size; - pSAREAPriv->depth_offset = pI830->DepthBuffer.Start; - pSAREAPriv->depth_size = pI830->DepthBuffer.Size; -#if 0 - pSAREAPriv->tex_offset = pI830->TexMem.Start; - pSAREAPriv->tex_size = pI830->TexMem.Size; -#endif - pSAREAPriv->log_tex_granularity = pI830->TexGranularity; - - ctx->driverClientMsg = malloc(sizeof(I830DRIRec)); - ctx->driverClientMsgSize = sizeof(I830DRIRec); - pI830DRI = (I830DRIPtr)ctx->driverClientMsg; - pI830DRI->deviceID = pI830->Chipset; - pI830DRI->regsSize = I830_REG_SIZE; - pI830DRI->width = ctx->shared.virtualWidth; - pI830DRI->height = ctx->shared.virtualHeight; - pI830DRI->mem = ctx->shared.fbSize; - pI830DRI->cpp = ctx->cpp; - - pI830DRI->bitsPerPixel = ctx->bpp; - pI830DRI->sarea_priv_offset = sizeof(drm_sarea_t); - - err = I830DRIDoMappings(ctx, pI830, pSAREAPriv); - if (err == FALSE) - return FALSE; - - I830SetupMemoryTiling(ctx, pI830); - - /* Quick hack to clear the front & back buffers. Could also use - * the clear ioctl to do this, but would need to setup hw state - * first. - */ - I830ClearScreen(ctx, pI830, pSAREAPriv); - - I830SetRingRegs(ctx, pI830); - - return TRUE; -} - - -/** - * \brief Validate the fbdev mode. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Saves some registers and returns 1. - * - * \sa radeonValidateMode(). - */ -static int i830ValidateMode( const DRIDriverContext *ctx ) -{ - return 1; -} - -/** - * \brief Examine mode returned by fbdev. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Restores registers that fbdev has clobbered and returns 1. - * - * \sa i810ValidateMode(). - */ -static int i830PostValidateMode( const DRIDriverContext *ctx ) -{ - I830Rec *pI830 = ctx->driverPrivate; - - I830SetRingRegs(ctx, pI830); - return 1; -} - - -/** - * \brief Initialize the framebuffer device mode - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Fills in \p info with some default values and some information from \p ctx - * and then calls I810ScreenInit() for the screen initialization. - * - * Before exiting clears the framebuffer memory accessing it directly. - */ -static int i830InitFBDev( DRIDriverContext *ctx ) -{ - I830Rec *pI830 = calloc(1, sizeof(I830Rec)); - int i; - - { - int dummy = ctx->shared.virtualWidth; - - switch (ctx->bpp / 8) { - case 1: dummy = (ctx->shared.virtualWidth + 127) & ~127; break; - case 2: dummy = (ctx->shared.virtualWidth + 31) & ~31; break; - case 3: - case 4: dummy = (ctx->shared.virtualWidth + 15) & ~15; break; - } - - ctx->shared.virtualWidth = dummy; - ctx->shared.Width = ctx->shared.virtualWidth; - } - - - for (i = 0; pitches[i] != 0; i++) { - if (pitches[i] >= ctx->shared.virtualWidth) { - ctx->shared.virtualWidth = pitches[i]; - break; - } - } - - ctx->driverPrivate = (void *)pI830; - - pI830->LpRing = calloc(1, sizeof(I830RingBuffer)); - pI830->Chipset = ctx->chipset; - pI830->LinearAddr = ctx->FBStart; - - if (!I830ScreenInit( ctx, pI830 )) - return 0; - - - return 1; -} - - -/** - * \brief The screen is being closed, so clean up any state and free any - * resources used by the DRI. - * - * \param ctx display handle. - * - * Unmaps the SAREA, closes the DRM device file descriptor and frees the driver - * private data. - */ -static void i830HaltFBDev( DRIDriverContext *ctx ) -{ - drmI830Sarea *pSAREAPriv; - I830Rec *pI830 = ctx->driverPrivate; - - if (pI830->irq) { - drmCtlUninstHandler(ctx->drmFD); - pI830->irq = 0; } - - I830CleanupDma(ctx); - - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - - I830DRIUnmapScreenRegions(ctx, pI830, pSAREAPriv); - drmUnmap( ctx->pSAREA, ctx->shared.SAREASize ); - drmClose(ctx->drmFD); - - if (ctx->driverPrivate) { - free(ctx->driverPrivate); - ctx->driverPrivate = 0; - } -} - - -extern void i810NotifyFocus( int ); - -/** - * \brief Exported driver interface for Mini GLX. - * - * \sa DRIDriverRec. - */ -const struct DRIDriverRec __driDriver = { - i830ValidateMode, - i830PostValidateMode, - i830InitFBDev, - i830HaltFBDev, - NULL,//I830EngineShutdown, - NULL, //I830EngineRestore, -#ifndef _EMBEDDED - 0, -#else - i810NotifyFocus, -#endif -}; diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile new file mode 100644 index 0000000000..e4e9cf17b0 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -0,0 +1,49 @@ + +TOP = ../../../../.. +include $(TOP)/configs/current + +LIBNAME = i915tex_dri.so + +MINIGLX_SOURCES = server/intel_dri.c + +PIPE_DRIVERS = \ + $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a + +DRIVER_SOURCES = \ + intel_softpipe.c \ + intel_batchbuffer.c \ + intel_tex_layout.c \ + intel_buffers.c \ + intel_blit.c \ + intel_context.c \ + intel_ioctl.c \ + intel_screen.c \ + intel_surface.c \ + intel_fbo.c \ + intel_depthstencil.c \ + intel_batchpool.c + +OLD_TEX =\ + intel_tex_image.c \ + intel_tex_subimage.c \ + intel_tex_copy.c \ + intel_tex_validate.c \ + intel_tex_format.c \ + intel_tex.c + + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(COMMON_BM_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") + +include ../Makefile.template + +intel_tex_layout.o: ../intel/intel_tex_layout.c + +symlinks: diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c new file mode 100644 index 0000000000..60bd8eaec2 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -0,0 +1,342 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_batchbuffer.h" +#include "intel_ioctl.h" + +/* Relocations in kernel space: + * - pass dma buffer seperately + * - memory manager knows how to patch + * - pass list of dependent buffers + * - pass relocation list + * + * Either: + * - get back an offset for buffer to fire + * - memory manager knows how to fire buffer + * + * Really want the buffer to be AGP and pinned. + * + */ + +/* Cliprect fence: The highest fence protecting a dma buffer + * containing explicit cliprect information. Like the old drawable + * lock but irq-driven. X server must wait for this fence to expire + * before changing cliprects [and then doing sw rendering?]. For + * other dma buffers, the scheduler will grab current cliprect info + * and mix into buffer. X server must hold the lock while changing + * cliprects??? Make per-drawable. Need cliprects in shared memory + * -- beats storing them with every cmd buffer in the queue. + * + * ==> X server must wait for this fence to expire before touching the + * framebuffer with new cliprects. + * + * ==> Cliprect-dependent buffers associated with a + * cliprect-timestamp. All of the buffers associated with a timestamp + * must go to hardware before any buffer with a newer timestamp. + * + * ==> Dma should be queued per-drawable for correct X/GL + * synchronization. Or can fences be used for this? + * + * Applies to: Blit operations, metaops, X server operations -- X + * server automatically waits on its own dma to complete before + * modifying cliprects ??? + */ + +static void +intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) +{ + int i; + fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); + for (i = 0; i < count / 4; i += 4) + fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", + offset + i * 4, ptr[i], ptr[i + 1], ptr[i + 2], ptr[i + 3]); + fprintf(stderr, "END BATCH\n\n\n"); +} + +void +intel_batchbuffer_reset(struct intel_batchbuffer *batch) +{ + + int i; + + /* + * Get a new, free batchbuffer. + */ + + batch->size = BATCH_SZ; + driBOData(batch->buffer, batch->size, NULL, 0); + + driBOResetList(&batch->list); + + /* + * Unreference buffers previously on the relocation list. + */ + + for (i = 0; i < batch->nr_relocs; i++) { + struct buffer_reloc *r = &batch->reloc[i]; + driBOUnReference(r->buf); + } + + batch->list_count = 0; + batch->nr_relocs = 0; + batch->flags = 0; + + /* + * We don't refcount the batchbuffer itself since we can't destroy it + * while it's on the list. + */ + + + driBOAddListItem(&batch->list, batch->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE); + + + batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); + batch->ptr = batch->map; +} + +/*====================================================================== + * Public functions + */ +struct intel_batchbuffer * +intel_batchbuffer_alloc(struct intel_context *intel) +{ + struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); + + batch->intel = intel; + + driGenBuffers(intel->intelScreen->batchPool, "batchbuffer", 1, + &batch->buffer, 4096, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); + batch->last_fence = NULL; + driBOCreateList(20, &batch->list); + intel_batchbuffer_reset(batch); + return batch; +} + +void +intel_batchbuffer_free(struct intel_batchbuffer *batch) +{ + if (batch->last_fence) { + driFenceFinish(batch->last_fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); + driFenceUnReference(batch->last_fence); + batch->last_fence = NULL; + } + if (batch->map) { + driBOUnmap(batch->buffer); + batch->map = NULL; + } + driBOUnReference(batch->buffer); + batch->buffer = NULL; + free(batch); +} + +/* TODO: Push this whole function into bufmgr. + */ +static void +do_flush_locked(struct intel_batchbuffer *batch, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) +{ + GLuint *ptr; + GLuint i; + struct intel_context *intel = batch->intel; + unsigned fenceFlags; + struct _DriFenceObject *fo; + + driBOValidateList(batch->intel->driFd, &batch->list); + + /* Apply the relocations. This nasty map indicates to me that the + * whole task should be done internally by the memory manager, and + * that dma buffers probably need to be pinned within agp space. + */ + ptr = (GLuint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, + DRM_BO_HINT_ALLOW_UNFENCED_MAP); + + + for (i = 0; i < batch->nr_relocs; i++) { + struct buffer_reloc *r = &batch->reloc[i]; + + ptr[r->offset / 4] = driBOOffset(r->buf) + r->delta; + } + + if (INTEL_DEBUG & DEBUG_BATCH) + intel_dump_batchbuffer(0, ptr, used); + + driBOUnmap(batch->buffer); + batch->map = NULL; + + /* Throw away non-effective packets. Won't work once we have + * hardware contexts which would preserve statechanges beyond a + * single buffer. + */ + + if (!(intel->numClipRects == 0 && !ignore_cliprects)) { + intel_batch_ioctl(batch->intel, + driBOOffset(batch->buffer), + used, ignore_cliprects, allow_unlock); + } + + + /* + * Kernel fencing. The flags tells the kernel that we've + * programmed an MI_FLUSH. + */ + + fenceFlags = DRM_I915_FENCE_FLAG_FLUSHED; + fo = driFenceBuffers(batch->intel->driFd, + "Batch fence", fenceFlags); + + /* + * User space fencing. + */ + + driBOFence(batch->buffer, fo); + + if (driFenceType(fo) == DRM_FENCE_TYPE_EXE) { + + /* + * Oops. We only validated a batch buffer. This means we + * didn't do any proper rendering. Discard this fence object. + */ + + driFenceUnReference(fo); + } else { + driFenceUnReference(batch->last_fence); + batch->last_fence = fo; + for (i = 0; i < batch->nr_relocs; i++) { + struct buffer_reloc *r = &batch->reloc[i]; + driBOFence(r->buf, fo); + } + } + + if (intel->numClipRects == 0 && !ignore_cliprects) { + if (allow_unlock) { + UNLOCK_HARDWARE(intel); + sched_yield(); + LOCK_HARDWARE(intel); + } +// intel->vtbl.lost_hardware(intel); + } +} + + +struct _DriFenceObject * +intel_batchbuffer_flush(struct intel_batchbuffer *batch) +{ + struct intel_context *intel = batch->intel; + GLuint used = batch->ptr - batch->map; + GLboolean was_locked = intel->locked; + + if (used == 0) + return batch->last_fence; + +#define MI_FLUSH ((0<<29)|(4<<23)) + + /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a + * performance drain that we would like to avoid. + */ + if (used & 4) { + ((int *) batch->ptr)[0] = MI_FLUSH; + ((int *) batch->ptr)[1] = 0; + ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->ptr)[0] = MI_FLUSH; + ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; + used += 8; + } + + driBOUnmap(batch->buffer); + batch->ptr = NULL; + batch->map = NULL; + + /* TODO: Just pass the relocation list and dma buffer up to the + * kernel. + */ + if (!was_locked) + LOCK_HARDWARE(intel); + + do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), + GL_FALSE); + + if (!was_locked) + UNLOCK_HARDWARE(intel); + + /* Reset the buffer: + */ + intel_batchbuffer_reset(batch); + return batch->last_fence; +} + +void +intel_batchbuffer_finish(struct intel_batchbuffer *batch) +{ + struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); + driFenceReference(fence); + driFenceFinish(fence, 3, GL_FALSE); + driFenceUnReference(fence); +} + + +/* This is the only way buffers get added to the validate list. + */ +GLboolean +intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, + struct _DriBufferObject *buffer, + GLuint flags, GLuint mask, GLuint delta) +{ + assert(batch->nr_relocs < MAX_RELOCS); + + driBOAddListItem(&batch->list, buffer, flags, mask); + + { + struct buffer_reloc *r = &batch->reloc[batch->nr_relocs++]; + driBOReference(buffer); + r->buf = buffer; + r->offset = batch->ptr - batch->map; + r->delta = delta; + } + + batch->ptr += 4; + return GL_TRUE; +} + + + +void +intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, GLuint bytes, GLuint flags) +{ + assert((bytes & 3) == 0); + intel_batchbuffer_require_space(batch, bytes, flags); + memcpy(batch->ptr, data, bytes); + batch->ptr += bytes; +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h new file mode 100644 index 0000000000..6d4d05e0bb --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h @@ -0,0 +1,123 @@ +#ifndef INTEL_BATCHBUFFER_H +#define INTEL_BATCHBUFFER_H + +#include "mtypes.h" +#include "dri_bufmgr.h" + +struct intel_context; + +#define BATCH_SZ 16384 +#define BATCH_RESERVED 16 + +#define MAX_RELOCS 4096 + +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 + +struct buffer_reloc +{ + struct _DriBufferObject *buf; + GLuint offset; + GLuint delta; /* not needed? */ +}; + +struct intel_batchbuffer +{ + struct bufmgr *bm; + struct intel_context *intel; + + struct _DriBufferObject *buffer; + struct _DriFenceObject *last_fence; + GLuint flags; + + drmBOList list; + GLuint list_count; + GLubyte *map; + GLubyte *ptr; + + struct buffer_reloc reloc[MAX_RELOCS]; + GLuint nr_relocs; + GLuint size; +}; + +struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context + *intel); + +void intel_batchbuffer_free(struct intel_batchbuffer *batch); + + +void intel_batchbuffer_finish(struct intel_batchbuffer *batch); + +struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer + *batch); + +void intel_batchbuffer_reset(struct intel_batchbuffer *batch); + + +/* Unlike bmBufferData, this currently requires the buffer be mapped. + * Consider it a convenience function wrapping multple + * intel_buffer_dword() calls. + */ +void intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, GLuint bytes, GLuint flags); + +void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, + GLuint bytes); + +GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, + struct _DriBufferObject *buffer, + GLuint flags, + GLuint mask, GLuint offset); + +/* Inline functions - might actually be better off with these + * non-inlined. Certainly better off switching all command packets to + * be passed as structs rather than dwords, but that's a little bit of + * work... + */ +static INLINE GLuint +intel_batchbuffer_space(struct intel_batchbuffer *batch) +{ + return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); +} + + +static INLINE void +intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) +{ + assert(batch->map); + assert(intel_batchbuffer_space(batch) >= 4); + *(GLuint *) (batch->ptr) = dword; + batch->ptr += 4; +} + +static INLINE void +intel_batchbuffer_require_space(struct intel_batchbuffer *batch, + GLuint sz, GLuint flags) +{ + assert(sz < batch->size - 8); + if (intel_batchbuffer_space(batch) < sz || + (batch->flags != 0 && flags != 0 && batch->flags != flags)) + intel_batchbuffer_flush(batch); + + batch->flags |= flags; +} + +/* Here are the crusty old macros, to be removed: + */ +#define BATCH_LOCALS + +#define BEGIN_BATCH(n, flags) do { \ + intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ +} while (0) + +#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) + +#define OUT_RELOC(buf,flags,mask,delta) do { \ + assert((delta) >= 0); \ + intel_batchbuffer_emit_reloc(intel->batch, buf, flags, mask, delta); \ +} while (0) + +#define ADVANCE_BATCH() do { } while(0) + + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c new file mode 100644 index 0000000000..2503b8a62a --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c @@ -0,0 +1,418 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "imports.h" +#include "glthread.h" +#include "dri_bufpool.h" +#include "dri_bufmgr.h" +#include "intel_screen.h" + +typedef struct +{ + drmMMListHead head; + struct _BPool *parent; + struct _DriFenceObject *fence; + unsigned long start; + int unfenced; + int mapped; +} BBuf; + +typedef struct _BPool +{ + _glthread_Mutex mutex; + unsigned long bufSize; + unsigned poolSize; + unsigned numFree; + unsigned numTot; + unsigned numDelayed; + unsigned checkDelayed; + drmMMListHead free; + drmMMListHead delayed; + drmMMListHead head; + drmBO kernelBO; + void *virtual; + BBuf *bufs; +} BPool; + + +static BPool * +createBPool(int fd, unsigned long bufSize, unsigned numBufs, unsigned flags, + unsigned checkDelayed) +{ + BPool *p = (BPool *) malloc(sizeof(*p)); + BBuf *buf; + int i; + + if (!p) + return NULL; + + p->bufs = (BBuf *) malloc(numBufs * sizeof(*p->bufs)); + if (!p->bufs) { + free(p); + return NULL; + } + + DRMINITLISTHEAD(&p->free); + DRMINITLISTHEAD(&p->head); + DRMINITLISTHEAD(&p->delayed); + + p->numTot = numBufs; + p->numFree = numBufs; + p->bufSize = bufSize; + p->numDelayed = 0; + p->checkDelayed = checkDelayed; + + _glthread_INIT_MUTEX(p->mutex); + + if (drmBOCreate(fd, 0, numBufs * bufSize, 0, NULL, drm_bo_type_dc, + flags, DRM_BO_HINT_DONT_FENCE, &p->kernelBO)) { + free(p->bufs); + free(p); + return NULL; + } + if (drmBOMap(fd, &p->kernelBO, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, + &p->virtual)) { + drmBODestroy(fd, &p->kernelBO); + free(p->bufs); + free(p); + return NULL; + } + + /* + * We unmap the buffer so that we can validate it later. Note that this is + * just a synchronizing operation. The buffer will have a virtual mapping + * until it is destroyed. + */ + + drmBOUnmap(fd, &p->kernelBO); + + buf = p->bufs; + for (i = 0; i < numBufs; ++i) { + buf->parent = p; + buf->fence = NULL; + buf->start = i * bufSize; + buf->mapped = 0; + buf->unfenced = 0; + DRMLISTADDTAIL(&buf->head, &p->free); + buf++; + } + + return p; +} + + +static void +pool_checkFree(BPool * p, int wait) +{ + drmMMListHead *list, *prev; + BBuf *buf; + int signaled = 0; + int i; + + list = p->delayed.next; + + if (p->numDelayed > 3) { + for (i = 0; i < p->numDelayed; i += 3) { + list = list->next; + } + } + + prev = list->prev; + for (; list != &p->delayed; list = prev, prev = list->prev) { + + buf = DRMLISTENTRY(BBuf, list, head); + + if (!signaled) { + if (wait) { + driFenceFinish(buf->fence, DRM_FENCE_TYPE_EXE, 1); + signaled = 1; + } + else { + signaled = driFenceSignaled(buf->fence, DRM_FENCE_TYPE_EXE); + } + } + + if (!signaled) + break; + + driFenceUnReference(buf->fence); + buf->fence = NULL; + DRMLISTDEL(list); + p->numDelayed--; + DRMLISTADD(list, &p->free); + p->numFree++; + } +} + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, unsigned flags, unsigned hint, + unsigned alignment) +{ + BPool *p = (BPool *) pool->data; + + drmMMListHead *item; + + if (alignment && (alignment != 4096)) + return NULL; + + _glthread_LOCK_MUTEX(p->mutex); + + if (p->numFree == 0) + pool_checkFree(p, GL_TRUE); + + if (p->numFree == 0) { + fprintf(stderr, "Out of fixed size buffer objects\n"); + BM_CKFATAL(-ENOMEM); + } + + item = p->free.next; + + if (item == &p->free) { + fprintf(stderr, "Fixed size buffer pool corruption\n"); + } + + DRMLISTDEL(item); + --p->numFree; + + _glthread_UNLOCK_MUTEX(p->mutex); + return (void *) DRMLISTENTRY(BBuf, item, head); +} + + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + _glthread_LOCK_MUTEX(p->mutex); + + if (buf->fence) { + DRMLISTADDTAIL(&buf->head, &p->delayed); + p->numDelayed++; + } + else { + buf->unfenced = 0; + DRMLISTADD(&buf->head, &p->free); + p->numFree++; + } + + if ((p->numDelayed % p->checkDelayed) == 0) + pool_checkFree(p, 0); + + _glthread_UNLOCK_MUTEX(p->mutex); + return 0; +} + + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, void **virtual) +{ + + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + _glthread_LOCK_MUTEX(p->mutex); + + /* + * Currently Mesa doesn't have any condition variables to resolve this + * cleanly in a multithreading environment. + * We bail out instead. + */ + + if (buf->mapped) { + fprintf(stderr, "Trying to map already mapped buffer object\n"); + BM_CKFATAL(-EINVAL); + } + +#if 0 + if (buf->unfenced && !(hint & DRM_BO_HINT_ALLOW_UNFENCED_MAP)) { + fprintf(stderr, "Trying to map an unfenced buffer object 0x%08x" + " 0x%08x %d\n", hint, flags, buf->start); + BM_CKFATAL(-EINVAL); + } + +#endif + + if (buf->fence) { + _glthread_UNLOCK_MUTEX(p->mutex); + return -EBUSY; + } + + buf->mapped = GL_TRUE; + *virtual = (unsigned char *) p->virtual + buf->start; + _glthread_UNLOCK_MUTEX(p->mutex); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, int lazy) +{ + BBuf *buf = (BBuf *) private; + driFenceFinish(buf->fence, 0, lazy); + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + + buf->mapped = 0; + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + return p->kernelBO.offset + buf->start; +} + +static unsigned +pool_flags(struct _DriBufferPool *pool, void *private) +{ + BPool *p = (BPool *) pool->data; + + return p->kernelBO.flags; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + BPool *p = (BPool *) pool->data; + + return p->bufSize; +} + + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + _glthread_LOCK_MUTEX(p->mutex); + if (buf->fence) { + driFenceUnReference(buf->fence); + } + buf->fence = fence; + buf->unfenced = 0; + driFenceReference(buf->fence); + _glthread_UNLOCK_MUTEX(p->mutex); + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + return &p->kernelBO; +} + +static int +pool_validate(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + _glthread_LOCK_MUTEX(p->mutex); + buf->unfenced = GL_TRUE; + _glthread_UNLOCK_MUTEX(p->mutex); + return 0; +} + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + BPool *p = (BPool *) pool->data; + + /* + * Wait on outstanding fences. + */ + + _glthread_LOCK_MUTEX(p->mutex); + while ((p->numFree < p->numTot) && p->numDelayed) { + _glthread_UNLOCK_MUTEX(p->mutex); + sched_yield(); + pool_checkFree(p, GL_TRUE); + _glthread_LOCK_MUTEX(p->mutex); + } + + drmBODestroy(pool->fd, &p->kernelBO); + free(p->bufs); + _glthread_UNLOCK_MUTEX(p->mutex); + free(p); + free(pool); +} + + +struct _DriBufferPool * +driBatchPoolInit(int fd, unsigned flags, + unsigned long bufSize, + unsigned numBufs, unsigned checkDelayed) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + if (!pool) + return NULL; + + pool->data = createBPool(fd, bufSize, numBufs, flags, checkDelayed); + if (!pool->data) + return NULL; + + pool->fd = fd; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = &pool_validate; + pool->waitIdle = &pool_waitIdle; + pool->setstatic = NULL; + pool->takeDown = &pool_takedown; + return pool; +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_blit.c b/src/mesa/drivers/dri/intel_winsys/intel_blit.c new file mode 100644 index 0000000000..8e878f0088 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_blit.c @@ -0,0 +1,394 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "mtypes.h" +#include "context.h" +#include "enums.h" + +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_buffers.h" +#include "intel_context.h" +#include "intel_fbo.h" +#include "intel_reg.h" +#include "vblank.h" + +#include "pipe/p_context.h" + + +#define FILE_DEBUG_FLAG DEBUG_BLIT + +/** + * Copy the back color buffer to the front color buffer. + * Used for SwapBuffers(). + */ +void +intelCopyBuffer(__DRIdrawablePrivate * dPriv, + const drm_clip_rect_t * rect) +{ + + struct intel_context *intel; + const intelScreenPrivate *intelScreen; + + DBG("%s\n", __FUNCTION__); + + assert(dPriv); + + intel = intelScreenContext(dPriv->driScreenPriv->private); + if (!intel) + return; + + intelScreen = intel->intelScreen; + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + current context (which is what intelScreenContext should return) might + not get a contended lock and thus cliprects not updated (tests/manywin) */ + if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + + + if (dPriv && dPriv->numClipRects) { + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + const struct pipe_region *backRegion + = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? + intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : + intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); + const int backWidth = intel_fb->Base.Width; + const int backHeight = intel_fb->Base.Height; + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; + const int srcpitch = backRegion->pitch; + const int cpp = intelScreen->front.cpp; + int BR13, CMD; + int i; + + ASSERT(intel_fb); + ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ + ASSERT(backRegion); + ASSERT(backRegion->cpp == cpp); + + DBG("front pitch %d back pitch %d\n", + pitch, backRegion->pitch); + + if (cpp == 2) { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + } + else { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + } + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + drm_clip_rect_t sbox; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->front.width || + pbox->y2 > intelScreen->front.height) + continue; + + box = *pbox; + + if (rect) { + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > backWidth) + box.x2 = backWidth + box.x1; + if (box.y2 - box.y1 > backHeight) + box.y2 = backHeight + box.y1; + + DBG("box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((box.y1 << 16) | box.x1); + OUT_BATCH((box.y2 << 16) | box.x2); + + OUT_RELOC(intelScreen->front.buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + OUT_BATCH((sbox.y1 << 16) | sbox.x1); + OUT_BATCH((srcpitch * cpp) & 0xffff); + OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + ADVANCE_BATCH(); + } + + if (intel->first_swap_fence) + driFenceUnReference(intel->first_swap_fence); + intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); + driFenceReference(intel->first_swap_fence); + } + + UNLOCK_HARDWARE(intel); + + /* XXX this is bogus. The context here may not even be bound to this drawable! */ + if (intel->lastStamp != dPriv->lastStamp) { + GET_CURRENT_CONTEXT(currctx); + struct intel_context *intelcurrent = intel_context(currctx); + if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { + intelWindowMoved(intel); + intel->lastStamp = dPriv->lastStamp; + } + } + +} + + + + +void +intelEmitFillBlit(struct intel_context *intel, + GLuint cpp, + GLshort dst_pitch, + struct _DriBufferObject *dst_buffer, + GLuint dst_offset, + GLshort x, GLshort y, GLshort w, GLshort h, + GLuint value, GLuint mask) +{ + GLuint BR13, CMD; + GLboolean badMask = GL_FALSE; + BATCH_LOCALS; + + /* + printf("Emit fill blit value=0x%x mask=0x%x\n", value, mask); + */ + + dst_pitch *= cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); + CMD = XY_COLOR_BLT_CMD; + if ((mask & 0xffff) != 0xffff) + badMask = GL_TRUE; + break; + case 4: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); +#if 0 + CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | + XY_COLOR_BLT_WRITE_RGB); +#else + CMD = XY_COLOR_BLT_CMD; + if ((mask & 0xff000000) == 0xff000000) + CMD |= XY_COLOR_BLT_WRITE_ALPHA; + else if (mask & 0xff000000) + badMask = GL_TRUE; + if ((mask & 0x00ffffff) == 0x00ffffff) + CMD |= XY_COLOR_BLT_WRITE_RGB; + else if (mask & 0x00ffffff) + badMask = GL_TRUE; +#endif + break; + default: + return; + } + + assert(!badMask); + + DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); + + + BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((y << 16) | x); + OUT_BATCH(((y + h) << 16) | (x + w)); + OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); + OUT_BATCH(value); + ADVANCE_BATCH(); + + intel_batchbuffer_flush(intel->batch); +} + + +static GLuint translate_raster_op(GLenum logicop) +{ + switch(logicop) { + case GL_CLEAR: return 0x00; + case GL_AND: return 0x88; + case GL_AND_REVERSE: return 0x44; + case GL_COPY: return 0xCC; + case GL_AND_INVERTED: return 0x22; + case GL_NOOP: return 0xAA; + case GL_XOR: return 0x66; + case GL_OR: return 0xEE; + case GL_NOR: return 0x11; + case GL_EQUIV: return 0x99; + case GL_INVERT: return 0x55; + case GL_OR_REVERSE: return 0xDD; + case GL_COPY_INVERTED: return 0x33; + case GL_OR_INVERTED: return 0xBB; + case GL_NAND: return 0x77; + case GL_SET: return 0xFF; + default: return 0; + } +} + + +/* Copy BitBlt + */ +void +intelEmitCopyBlit(struct intel_context *intel, + GLuint cpp, + GLshort src_pitch, + struct _DriBufferObject *src_buffer, + GLuint src_offset, + GLshort dst_pitch, + struct _DriBufferObject *dst_buffer, + GLuint dst_offset, + GLshort src_x, GLshort src_y, + GLshort dst_x, GLshort dst_y, + GLshort w, GLshort h, + GLenum logic_op) +{ + GLuint CMD, BR13; + int dst_y2 = dst_y + h; + int dst_x2 = dst_x + w; + BATCH_LOCALS; + + + DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + src_buffer, src_pitch, src_offset, src_x, src_y, + dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); + + src_pitch *= cpp; + dst_pitch *= cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = (((GLint) dst_pitch) & 0xffff) | + (translate_raster_op(logic_op) << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + break; + case 4: + BR13 = + (((GLint) dst_pitch) & 0xffff) | + (translate_raster_op(logic_op) << 16) | (1 << 24) | (1 << 25); + CMD = + (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + break; + default: + return; + } + + if (dst_y2 < dst_y || dst_x2 < dst_x) { + return; + } + + /* Initial y values don't seem to work with negative pitches. If + * we adjust the offsets manually (below), it seems to work fine. + * + * On the other hand, if we always adjust, the hardware doesn't + * know which blit directions to use, so overlapping copypixels get + * the wrong result. + */ + if (dst_pitch > 0 && src_pitch > 0) { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((dst_y << 16) | dst_x); + OUT_BATCH((dst_y2 << 16) | dst_x2); + OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); + OUT_BATCH((src_y << 16) | src_x); + OUT_BATCH(((GLint) src_pitch & 0xffff)); + OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, src_offset); + ADVANCE_BATCH(); + } + else { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((0 << 16) | dst_x); + OUT_BATCH((h << 16) | dst_x2); + OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, + dst_offset + dst_y * dst_pitch); + OUT_BATCH((0 << 16) | src_x); + OUT_BATCH(((GLint) src_pitch & 0xffff)); + OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, + src_offset + src_y * src_pitch); + ADVANCE_BATCH(); + } + + intel_batchbuffer_flush( intel->batch ); +} + + + diff --git a/src/mesa/drivers/dri/intel_winsys/intel_blit.h b/src/mesa/drivers/dri/intel_winsys/intel_blit.h new file mode 100644 index 0000000000..46c2594477 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_blit.h @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BLIT_H +#define INTEL_BLIT_H + +#include "intel_context.h" +#include "intel_ioctl.h" +#include "dri_bufmgr.h" + +extern void intelCopyBuffer(__DRIdrawablePrivate * dpriv, + const drm_clip_rect_t * rect); + +extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask); + +extern void intelEmitCopyBlit(struct intel_context *intel, + GLuint cpp, + GLshort src_pitch, + struct _DriBufferObject *src_buffer, + GLuint src_offset, + GLshort dst_pitch, + struct _DriBufferObject *dst_buffer, + GLuint dst_offset, + GLshort srcx, GLshort srcy, + GLshort dstx, GLshort dsty, + GLshort w, GLshort h, + GLenum logicop ); + +extern void intelEmitFillBlit(struct intel_context *intel, + GLuint cpp, + GLshort dst_pitch, + struct _DriBufferObject *dst_buffer, + GLuint dst_offset, + GLshort x, GLshort y, + GLshort w, GLshort h, GLuint value, GLuint mask); + + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c new file mode 100644 index 0000000000..fc9d60c88e --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -0,0 +1,711 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_blit.h" +#include "intel_buffers.h" +#include "intel_depthstencil.h" +#include "intel_fbo.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" +#include "context.h" +#include "utils.h" +#include "drirenderbuffer.h" +#include "framebuffer.h" +#include "vblank.h" + +#include "pipe/p_context.h" + +/* This block can be removed when libdrm >= 2.3.1 is required */ + +#ifndef DRM_VBLANK_FLIP + +#define DRM_VBLANK_FLIP 0x8000000 + +typedef struct drm_i915_flip { + int pipes; +} drm_i915_flip_t; + +#undef DRM_IOCTL_I915_FLIP +#define DRM_IOCTL_I915_FLIP DRM_IOW(DRM_COMMAND_BASE + DRM_I915_FLIP, \ + drm_i915_flip_t) + +#endif + + +/** + * XXX move this into a new dri/common/cliprects.c file. + */ +GLboolean +intel_intersect_cliprects(drm_clip_rect_t * dst, + const drm_clip_rect_t * a, + const drm_clip_rect_t * b) +{ + GLint bx = b->x1; + GLint by = b->y1; + GLint bw = b->x2 - bx; + GLint bh = b->y2 - by; + + if (bx < a->x1) + bw -= a->x1 - bx, bx = a->x1; + if (by < a->y1) + bh -= a->y1 - by, by = a->y1; + if (bx + bw > a->x2) + bw = a->x2 - bx; + if (by + bh > a->y2) + bh = a->y2 - by; + if (bw <= 0) + return GL_FALSE; + if (bh <= 0) + return GL_FALSE; + + dst->x1 = bx; + dst->y1 = by; + dst->x2 = bx + bw; + dst->y2 = by + bh; + + return GL_TRUE; +} + +/** + * Return pointer to current color drawing region, or NULL. + */ +struct pipe_region * +intel_drawbuf_region(struct intel_context *intel) +{ + struct intel_renderbuffer *irbColor = + intel_renderbuffer(intel->ctx.DrawBuffer->_ColorDrawBuffers[0][0]); + if (irbColor) + return irbColor->region; + else + return NULL; +} + +/** + * Return pointer to current color reading region, or NULL. + */ +struct pipe_region * +intel_readbuf_region(struct intel_context *intel) +{ + struct intel_renderbuffer *irb + = intel_renderbuffer(intel->ctx.ReadBuffer->_ColorReadBuffer); + if (irb) + return irb->region; + else + return NULL; +} + + + +/** + * Update the following fields for rendering: + * intel->numClipRects + * intel->pClipRects + */ +static void +intelSetRenderbufferClipRects(struct intel_context *intel) +{ + /* zero-sized buffers might be legal? */ + assert(intel->ctx.DrawBuffer->Width > 0); + assert(intel->ctx.DrawBuffer->Height > 0); + intel->fboRect.x1 = 0; + intel->fboRect.y1 = 0; + intel->fboRect.x2 = intel->ctx.DrawBuffer->Width; + intel->fboRect.y2 = intel->ctx.DrawBuffer->Height; + intel->numClipRects = 1; + intel->pClipRects = &intel->fboRect; +} + + +/** + * This will be called whenever the currently bound window is moved/resized. + * XXX: actually, it seems to NOT be called when the window is only moved (BP). + */ +void +intelWindowMoved(struct intel_context *intel) +{ + GLcontext *ctx = &intel->ctx; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + + if (!intel->ctx.DrawBuffer) { + /* when would this happen? -BP */ + assert(0); + intel->numClipRects = 0; + } + + /* Update Mesa's notion of window size */ + driUpdateFramebufferSize(ctx, dPriv); + intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ + + { + drmI830Sarea *sarea = intel->sarea; + drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, + .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; + drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, .y1 = sarea->pipeA_y, + .x2 = sarea->pipeA_x + sarea->pipeA_w, + .y2 = sarea->pipeA_y + sarea->pipeA_h }; + drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, .y1 = sarea->pipeB_y, + .x2 = sarea->pipeB_x + sarea->pipeB_w, + .y2 = sarea->pipeB_y + sarea->pipeB_h }; + GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); + GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); + GLuint flags = intel_fb->vblank_flags; + GLboolean pf_active; + GLint pf_pipes; + + /* Update page flipping info + */ + pf_pipes = 0; + + if (areaA > 0) + pf_pipes |= 1; + + if (areaB > 0) + pf_pipes |= 2; + + intel_fb->pf_current_page = (intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3; + + intel_fb->pf_num_pages = 2 /*intel->intelScreen->third.handle ? 3 : 2*/; + + pf_active = pf_pipes && (pf_pipes & intel->sarea->pf_active) == pf_pipes; + + if (INTEL_DEBUG & DEBUG_LOCK) + if (pf_active != intel_fb->pf_active) + _mesa_printf("%s - Page flipping %sactive\n", __progname, + pf_active ? "" : "in"); + + if (pf_active) { + /* Sync pages between pipes if we're flipping on both at the same time */ + if (pf_pipes == 0x3 && pf_pipes != intel_fb->pf_pipes && + (intel->sarea->pf_current_page & 0x3) != + (((intel->sarea->pf_current_page) >> 2) & 0x3)) { + drm_i915_flip_t flip; + + if (intel_fb->pf_current_page == + (intel->sarea->pf_current_page & 0x3)) { + /* XXX: This is ugly, but emitting two flips 'in a row' can cause + * lockups for unknown reasons. + */ + intel->sarea->pf_current_page = + intel->sarea->pf_current_page & 0x3; + intel->sarea->pf_current_page |= + ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % + intel_fb->pf_num_pages) << 2; + + flip.pipes = 0x2; + } else { + intel->sarea->pf_current_page = + intel->sarea->pf_current_page & (0x3 << 2); + intel->sarea->pf_current_page |= + (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % + intel_fb->pf_num_pages; + + flip.pipes = 0x1; + } + + drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); + } + + intel_fb->pf_pipes = pf_pipes; + } + + intel_fb->pf_active = pf_active; + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); + + /* Update vblank info + */ + if (areaB > areaA || (areaA == areaB && areaB > 0)) { + flags = intel_fb->vblank_flags | VBLANK_FLAG_SECONDARY; + } else { + flags = intel_fb->vblank_flags & ~VBLANK_FLAG_SECONDARY; + } + + if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && + !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { + drmVBlank vbl; + int i; + + vbl.request.type = DRM_VBLANK_ABSOLUTE; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + vbl.request.type |= DRM_VBLANK_SECONDARY; + } + + for (i = 0; i < intel_fb->pf_num_pages; i++) { + if (!intel_fb->color_rb[i] || + (intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <= + (1<<23)) + continue; + + vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending; + drmWaitVBlank(intel->driFd, &vbl); + } + + intel_fb->vblank_flags = flags; + driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); + intel_fb->vbl_waited = intel_fb->vbl_seq; + + for (i = 0; i < intel_fb->pf_num_pages; i++) { + if (intel_fb->color_rb[i]) + intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited; + } + } + } + + /* This will be picked up by looking at the dirty state flags: + */ + + /* Update hardware scissor */ +// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, +// ctx->Scissor.Width, ctx->Scissor.Height); + + /* Re-calculate viewport related state */ +// ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far ); +} + + + + + +/* Emit wait for pending flips */ +void +intel_wait_flips(struct intel_context *intel, GLuint batch_flags) +{ + struct intel_framebuffer *intel_fb = + (struct intel_framebuffer *) intel->ctx.DrawBuffer; + struct intel_renderbuffer *intel_rb = + intel_get_renderbuffer(&intel_fb->Base, + intel_fb->Base._ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : + BUFFER_BACK_LEFT); + + if (intel_fb->Base.Name == 0 && intel_rb->pf_pending == intel_fb->pf_seq) { + GLint pf_pipes = intel_fb->pf_pipes; + BATCH_LOCALS; + + /* Wait for pending flips to take effect */ + BEGIN_BATCH(2, batch_flags); + OUT_BATCH(pf_pipes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) + : 0); + OUT_BATCH(pf_pipes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) + : 0); + ADVANCE_BATCH(); + + intel_rb->pf_pending--; + } +} + +#if 0 +/* Flip the front & back buffers + */ +static GLboolean +intelPageFlip(const __DRIdrawablePrivate * dPriv) +{ + struct intel_context *intel; + int ret; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + + if (INTEL_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s\n", __FUNCTION__); + + assert(dPriv); + assert(dPriv->driContextPriv); + assert(dPriv->driContextPriv->driverPrivate); + + intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate; + + if (intel->intelScreen->drmMinor < 9) + return GL_FALSE; + + intelFlush(&intel->ctx); + + ret = 0; + + LOCK_HARDWARE(intel); + + if (dPriv->numClipRects && intel_fb->pf_active) { + drm_i915_flip_t flip; + + flip.pipes = intel_fb->pf_pipes; + + ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); + } + + UNLOCK_HARDWARE(intel); + + if (ret || !intel_fb->pf_active) + return GL_FALSE; + + if (!dPriv->numClipRects) { + usleep(10000); /* throttle invisible client 10ms */ + } + + intel_fb->pf_current_page = (intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3; + + if (dPriv->numClipRects != 0) { + intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending = + intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->pf_pending = + ++intel_fb->pf_seq; + } + + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, &intel_fb->Base); + + if (INTEL_DEBUG & DEBUG_IOCTL) + fprintf(stderr, "%s: success\n", __FUNCTION__); + + return GL_TRUE; +} +#endif + + +static GLboolean +intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) +{ + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); + struct intel_context *intel = + intelScreenContext(dPriv->driScreenPriv->private); + const intelScreenPrivate *intelScreen = intel->intelScreen; + unsigned int target; + drm_i915_vblank_swap_t swap; + GLboolean ret; + + if (!intel_fb->vblank_flags || + (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || + intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) + return GL_FALSE; + + swap.seqtype = DRM_VBLANK_ABSOLUTE; + + if (intel_fb->vblank_flags & VBLANK_FLAG_SYNC) { + swap.seqtype |= DRM_VBLANK_NEXTONMISS; + } else if (interval == 0) { + return GL_FALSE; + } + + swap.drawable = dPriv->hHWDrawable; + target = swap.sequence = intel_fb->vbl_seq + interval; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + swap.seqtype |= DRM_VBLANK_SECONDARY; + } + + LOCK_HARDWARE(intel); + + intel_batchbuffer_flush(intel->batch); + + if ( intel_fb->pf_active ) { + swap.seqtype |= DRM_VBLANK_FLIP; + + intel_fb->pf_current_page = (((intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3) + 1) % + intel_fb->pf_num_pages; + } + + if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, + sizeof(swap))) { + intel_fb->vbl_seq = swap.sequence; + swap.sequence -= target; + *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); + + intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending = + intel_get_renderbuffer(&intel_fb->Base, + BUFFER_FRONT_LEFT)->vbl_pending = + intel_fb->vbl_seq; + + if (swap.seqtype & DRM_VBLANK_FLIP) { + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); + } + + ret = GL_TRUE; + } else { + if (swap.seqtype & DRM_VBLANK_FLIP) { + intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3) % + intel_fb->pf_num_pages; + } + + ret = GL_FALSE; + } + + UNLOCK_HARDWARE(intel); + + return ret; +} + +void +intelSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel; + + if (ctx == NULL) + return; + + intel = intel_context(ctx); + + if (ctx->Visual.doubleBufferMode) { + GLboolean missed_target; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + int64_t ust; + + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + + if (!intelScheduleSwap(dPriv, &missed_target)) { + driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, + &missed_target); + + intelCopyBuffer(dPriv, NULL); + } + + intel_fb->swap_count++; + (*dri_interface->getUST) (&ust); + if (missed_target) { + intel_fb->swap_missed_count++; + intel_fb->swap_missed_ust = ust - intel_fb->swap_ust; + } + + intel_fb->swap_ust = ust; + } + } + else { + /* XXX this shouldn't be an error but we can't handle it for now */ + fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + } +} + +void +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + struct intel_context *intel = + (struct intel_context *) dPriv->driContextPriv->driverPrivate; + GLcontext *ctx = &intel->ctx; + + if (ctx->Visual.doubleBufferMode) { + drm_clip_rect_t rect; + /* fixup cliprect (driDrawable may have changed?) later */ + rect.x1 = x; + rect.y1 = y; + rect.x2 = w; + rect.y2 = h; + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + intelCopyBuffer(dPriv, &rect); + } + } + else { + /* XXX this shouldn't be an error but we can't handle it for now */ + fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + } +} + + +/** + * Update the hardware state for drawing into a window or framebuffer object. + * + * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other + * places within the driver. + * + * Basically, this needs to be called any time the current framebuffer + * changes, the renderbuffers change, or we need to draw into different + * color buffers. + */ +void +intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) +{ + struct intel_context *intel = intel_context(ctx); + struct pipe_region *colorRegion, *depthRegion = NULL; + struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; + + if (!fb) { + /* this can happen during the initial context initialization */ + return; + } + + /* Do this here, not core Mesa, since this function is called from + * many places within the driver. + */ + if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { + /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */ + _mesa_update_framebuffer(ctx); + /* this updates the DrawBuffer's Width/Height if it's a FBO */ + _mesa_update_draw_buffer_bounds(ctx); + } + + if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { + /* this may occur when we're called by glBindFrameBuffer() during + * the process of someone setting up renderbuffers, etc. + */ + /*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/ + return; + } + + if (fb->Name) + intel_validate_paired_depth_stencil(ctx, fb); + + /* + * How many color buffers are we drawing into? + */ + if (fb->_NumColorDrawBuffers[0] != 1) { + /* writing to 0 or 2 or 4 color buffers */ + /*_mesa_debug(ctx, "Software rendering\n");*/ + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); + } + else { + /* draw to exactly one color buffer */ + /*_mesa_debug(ctx, "Hardware rendering\n");*/ + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); + } + + /* + * Get the intel_renderbuffer for the colorbuffer we're drawing into. + * And set up cliprects. + */ + { + struct intel_renderbuffer *irb; + intelSetRenderbufferClipRects(intel); + irb = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); + colorRegion = (irb && irb->region) ? irb->region : NULL; + } + + /* Update culling direction which changes depending on the + * orientation of the buffer: + */ + if (ctx->Driver.FrontFace) + ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace); + else + ctx->NewState |= _NEW_POLYGON; + + if (!colorRegion) { + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); + } + else { + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); + } + + /*** + *** Get depth buffer region and check if we need a software fallback. + *** Note that the depth buffer is usually a DEPTH_STENCIL buffer. + ***/ + if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) { + irbDepth = intel_renderbuffer(fb->_DepthBuffer->Wrapped); + if (irbDepth && irbDepth->region) { + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); + depthRegion = irbDepth->region; + } + else { + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_TRUE); + depthRegion = NULL; + } + } + else { + /* not using depth buffer */ + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); + depthRegion = NULL; + } + + /*** + *** Stencil buffer + *** This can only be hardware accelerated if we're using a + *** combined DEPTH_STENCIL buffer (for now anyway). + ***/ + if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { + irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); + if (irbStencil && irbStencil->region) { + ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); + /* need to re-compute stencil hw state */ +// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); + if (!depthRegion) + depthRegion = irbStencil->region; + } + else { + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE); + } + } + else { + /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */ + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); + /* need to re-compute stencil hw state */ +// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); + } + + + /** + ** Release old regions, reference new regions + **/ + + // intel->vtbl.set_draw_region(intel, colorRegion, depthRegion); + + /* update viewport since it depends on window size */ +// ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y, +// ctx->Viewport.Width, ctx->Viewport.Height); + + /* Update hardware scissor */ +// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, +// ctx->Scissor.Width, ctx->Scissor.Height); +} + + +static void +intelDrawBuffer(GLcontext * ctx, GLenum mode) +{ + intel_draw_buffer(ctx, ctx->DrawBuffer); +} + + +static void +intelReadBuffer(GLcontext * ctx, GLenum mode) +{ + if (ctx->ReadBuffer == ctx->DrawBuffer) { + /* This will update FBO completeness status. + * A framebuffer will be incomplete if the GL_READ_BUFFER setting + * refers to a missing renderbuffer. Calling glReadBuffer can set + * that straight and can make the drawing buffer complete. + */ + intel_draw_buffer(ctx, ctx->DrawBuffer); + } + /* Generally, functions which read pixels (glReadPixels, glCopyPixels, etc) + * reference ctx->ReadBuffer and do appropriate state checks. + */ +} + + +void +intelInitBufferFuncs(struct dd_function_table *functions) +{ + functions->DrawBuffer = intelDrawBuffer; + functions->ReadBuffer = intelReadBuffer; +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.h b/src/mesa/drivers/dri/intel_winsys/intel_buffers.h new file mode 100644 index 0000000000..5834e39501 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.h @@ -0,0 +1,55 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BUFFERS_H +#define INTEL_BUFFERS_H + + +struct intel_context; +struct intel_framebuffer; + + +extern GLboolean +intel_intersect_cliprects(drm_clip_rect_t * dest, + const drm_clip_rect_t * a, + const drm_clip_rect_t * b); + +extern struct pipe_region *intel_readbuf_region(struct intel_context *intel); + +extern struct pipe_region *intel_drawbuf_region(struct intel_context *intel); + +extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); + +extern void intelWindowMoved(struct intel_context *intel); + +extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); + +extern void intelInitBufferFuncs(struct dd_function_table *functions); + +#endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c new file mode 100644 index 0000000000..e6f0d4a0fa --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -0,0 +1,717 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "glheader.h" +#include "context.h" +#include "matrix.h" +#include "simple_list.h" +#include "extensions.h" +#include "framebuffer.h" +#include "imports.h" +#include "points.h" + +#include "tnl/tnl.h" + +#include "tnl/t_pipeline.h" +#include "tnl/t_vertex.h" + +#include "drivers/common/driverfuncs.h" + +#include "intel_screen.h" + +#include "i830_dri.h" + +#include "intel_buffers.h" +/*#include "intel_tex.h"*/ +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" +/* +#include "intel_buffer_objects.h" +*/ +#include "intel_fbo.h" +#include "intel_tex_layout.h" + +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" + +#include "drirenderbuffer.h" +#include "vblank.h" +#include "utils.h" +#include "xmlpool.h" /* for symbolic values of enum-type options */ + +#include "pipe/p_context.h" + + + +#ifndef INTEL_DEBUG +int INTEL_DEBUG = (0); +#endif + +#define need_GL_ARB_multisample +#define need_GL_ARB_point_parameters +#define need_GL_ARB_texture_compression +#define need_GL_ARB_vertex_buffer_object +#define need_GL_ARB_vertex_program +#define need_GL_ARB_window_pos +#define need_GL_EXT_blend_color +#define need_GL_EXT_blend_equation_separate +#define need_GL_EXT_blend_func_separate +#define need_GL_EXT_blend_minmax +#define need_GL_EXT_cull_vertex +#define need_GL_EXT_fog_coord +#define need_GL_EXT_framebuffer_object +#define need_GL_EXT_multi_draw_arrays +#define need_GL_EXT_secondary_color +#define need_GL_NV_vertex_program +#include "extension_helper.h" + + +#define DRIVER_DATE "20070731" + +_glthread_Mutex lockMutex; +static GLboolean lockMutexInit = GL_FALSE; + + +static const GLubyte * +intelGetString(GLcontext * ctx, GLenum name) +{ + const char *chipset; + static char buffer[128]; + + switch (name) { + case GL_VENDOR: + return (GLubyte *) "Tungsten Graphics, Inc"; + break; + + case GL_RENDERER: + switch (intel_context(ctx)->intelScreen->deviceID) { + case PCI_CHIP_845_G: + chipset = "Intel(R) 845G"; + break; + case PCI_CHIP_I830_M: + chipset = "Intel(R) 830M"; + break; + case PCI_CHIP_I855_GM: + chipset = "Intel(R) 852GM/855GM"; + break; + case PCI_CHIP_I865_G: + chipset = "Intel(R) 865G"; + break; + case PCI_CHIP_I915_G: + chipset = "Intel(R) 915G"; + break; + case PCI_CHIP_I915_GM: + chipset = "Intel(R) 915GM"; + break; + case PCI_CHIP_I945_G: + chipset = "Intel(R) 945G"; + break; + case PCI_CHIP_I945_GM: + chipset = "Intel(R) 945GM"; + break; + case PCI_CHIP_I945_GME: + chipset = "Intel(R) 945GME"; + break; + case PCI_CHIP_G33_G: + chipset = "Intel(R) G33"; + break; + case PCI_CHIP_Q35_G: + chipset = "Intel(R) Q35"; + break; + case PCI_CHIP_Q33_G: + chipset = "Intel(R) Q33"; + break; + default: + chipset = "Unknown Intel Chipset"; + break; + } + + (void) driGetRendererString(buffer, chipset, DRIVER_DATE, 0); + return (GLubyte *) buffer; + + default: + return NULL; + } +} + + +/** + * Extension strings exported by the intel driver. + * + * \note + * It appears that ARB_texture_env_crossbar has "disappeared" compared to the + * old i830-specific driver. + */ +const struct dri_extension card_extensions[] = { + {"GL_ARB_multisample", GL_ARB_multisample_functions}, + {"GL_ARB_multitexture", NULL}, + {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_texture_border_clamp", NULL}, + {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, + {"GL_ARB_texture_cube_map", NULL}, + {"GL_ARB_texture_env_add", NULL}, + {"GL_ARB_texture_env_combine", NULL}, + {"GL_ARB_texture_env_dot3", NULL}, + {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_rectangle", NULL}, + {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, + {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, + {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, + {"GL_EXT_blend_equation_separate", + GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, + {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_subtract", NULL}, + {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, + {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, +#if 1 /* XXX FBO temporary? */ + {"GL_EXT_packed_depth_stencil", NULL}, +#endif + {"GL_EXT_pixel_buffer_object", NULL}, + {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, + {"GL_EXT_stencil_wrap", NULL}, + {"GL_EXT_texture_edge_clamp", NULL}, + {"GL_EXT_texture_env_combine", NULL}, + {"GL_EXT_texture_env_dot3", NULL}, + {"GL_EXT_texture_filter_anisotropic", NULL}, + {"GL_EXT_texture_lod_bias", NULL}, + {"GL_3DFX_texture_compression_FXT1", NULL}, + {"GL_APPLE_client_storage", NULL}, + {"GL_MESA_pack_invert", NULL}, + {"GL_MESA_ycbcr_texture", NULL}, + {"GL_NV_blend_square", NULL}, + {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, + {"GL_NV_vertex_program1_1", NULL}, +/* { "GL_SGIS_generate_mipmap", NULL }, */ + {NULL, NULL} +}; + + + + +static const struct dri_debug_control debug_control[] = { + {"tex", DEBUG_TEXTURE}, + {"state", DEBUG_STATE}, + {"ioctl", DEBUG_IOCTL}, + {"blit", DEBUG_BLIT}, + {"mip", DEBUG_MIPTREE}, + {"fall", DEBUG_FALLBACKS}, + {"verb", DEBUG_VERBOSE}, + {"bat", DEBUG_BATCH}, + {"pix", DEBUG_PIXEL}, + {"buf", DEBUG_BUFMGR}, + {"reg", DEBUG_REGION}, + {"fbo", DEBUG_FBO}, + {"lock", DEBUG_LOCK}, + {NULL, 0} +}; + + +static void +intelInvalidateState(GLcontext * ctx, GLuint new_state) +{ + _vbo_InvalidateState(ctx, new_state); + _tnl_InvalidateState(ctx, new_state); + _tnl_invalidate_vertex_state(ctx, new_state); + + st_invalidate_state( ctx, new_state ); +} + + +void +intelFlush(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + + INTEL_FIREVERTICES(intel); + + if (intel->batch->map != intel->batch->ptr) + intel_batchbuffer_flush(intel->batch); + + /* XXX: Need to do an MI_FLUSH here. + */ +} + + +/** + * Check if we need to rotate/warp the front color buffer to the + * rotated screen. We generally need to do this when we get a glFlush + * or glFinish after drawing to the front color buffer. + * If no rotation, just copy the private fake front buffer to the real one. + */ +static void +intelCheckFrontUpdate(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + /* rely on _ColorDrawBufferMask being kept up to date by mesa + even for window-fbos. */ + /* not sure. Might need that for all masks including + BUFFER_BIT_FRONT_LEFT maybe? */ + if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT) { + __DRIdrawablePrivate *dPriv = intel->driDrawable; + intelCopyBuffer(dPriv, NULL); + } +} + + +/** + * Called via glFlush. + */ +static void +intelglFlush(GLcontext * ctx) +{ + intelFlush(ctx); + intelCheckFrontUpdate(ctx); +} + +void +intelFinish(GLcontext * ctx) +{ + struct intel_context *intel = intel_context(ctx); + intelFlush(ctx); + if (intel->batch->last_fence) { + driFenceFinish(intel->batch->last_fence, + 0, GL_FALSE); + driFenceUnReference(intel->batch->last_fence); + intel->batch->last_fence = NULL; + } + intelCheckFrontUpdate(ctx); +} + + +static void +intelInitDriverFunctions(struct dd_function_table *functions) +{ + _mesa_init_driver_functions(functions); + + functions->Flush = intelglFlush; + functions->Finish = intelFinish; + functions->GetString = intelGetString; + functions->UpdateState = intelInvalidateState; + + /* + intelInitTextureFuncs(functions); + */ + intelInitBufferFuncs(functions); + + st_init_driver_functions(functions); +} + + + +GLboolean +intelCreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) +{ + struct dd_function_table functions; + + struct intel_context *intel = CALLOC_STRUCT(intel_context); + GLcontext *ctx = &intel->ctx; + + GLcontext *shareCtx = (GLcontext *) sharedContextPrivate; + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + drmI830Sarea *saPriv = intelScreen->sarea; + int fthrottle_mode; + GLboolean havePools; + + intelInitDriverFunctions(&functions); + + if (!_mesa_initialize_context(&intel->ctx, + mesaVis, shareCtx, + &functions, (void *) intel)) + return GL_FALSE; + + driContextPriv->driverPrivate = intel; + intel->intelScreen = intelScreen; + intel->driScreen = sPriv; + intel->sarea = saPriv; + + if (!lockMutexInit) { + lockMutexInit = GL_TRUE; + _glthread_INIT_MUTEX(lockMutex); + } + + driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, + intel->driScreen->myNum, "i915"); + + + /* Initialize the software rasterizer and helper modules. */ + _vbo_CreateContext(ctx); + _tnl_CreateContext(ctx); + + /* + * Pipe-related setup + */ + st_create_context( &intel->ctx, + intel_create_softpipe( intel ) ); + + /* KW: Not sure I like this - we should only be talking to the + * state_tracker. The pipe code will need some way of talking to + * us, eg for batchbuffer ioctls, and there will need to be a + * buffer manager interface. So, this is a temporary hack, right? + * BP: Yes, a temporary hack so we can make jumps between old/new code. + */ + intel->pipe = intel->ctx.st->pipe; +// intel->pipe->screen = intelScreen; +// intel->pipe->glctx = ctx; +// intel_init_region_functions(intel->pipe); + + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + intel->pipe->mipmap_tree_layout = i945_miptree_layout; + break; + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + case PCI_CHIP_I830_M: + case PCI_CHIP_I855_GM: + case PCI_CHIP_I865_G: + intel->pipe->mipmap_tree_layout = i915_miptree_layout; + default: + assert(0); /*FIX*/ + } + + + /* + * memory pools + */ + DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + havePools = intelCreatePools(intelScreen); + DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + if (!havePools) + return GL_FALSE; + + + /* Dri stuff */ + intel->hHWContext = driContextPriv->hHWContext; + intel->driFd = sPriv->fd; + intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; + + TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; + + + + fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); + intel->iw.irq_seq = -1; + intel->irqsEmitted = 0; + + /* Disable imaging extension until convolution is working in + * teximage paths: + */ + driInitExtensions(ctx, card_extensions, +/* GL_TRUE, */ + GL_FALSE); + + + intel->batch = intel_batchbuffer_alloc(intel); + intel->last_swap_fence = NULL; + intel->first_swap_fence = NULL; + + intel_fbo_init(intel); + + if (intel->ctx.Mesa_DXTn) { + _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); + _mesa_enable_extension(ctx, "GL_S3_s3tc"); + } + else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) { + _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); + } + +#if DO_DEBUG + INTEL_DEBUG = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); +#endif + + + return GL_TRUE; +} + +void +intelDestroyContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = + (struct intel_context *) driContextPriv->driverPrivate; + + assert(intel); /* should never be null */ + if (intel) { + GLboolean release_texture_heaps; + + INTEL_FIREVERTICES(intel); + + //intel->vtbl.destroy(intel); + + release_texture_heaps = (intel->ctx.Shared->RefCount == 1); + _tnl_DestroyContext(&intel->ctx); + _vbo_DestroyContext(&intel->ctx); + + intel_batchbuffer_free(intel->batch); + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + if (intel->first_swap_fence) { + driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->first_swap_fence); + intel->first_swap_fence = NULL; + } + + + if (release_texture_heaps) { + /* This share group is about to go away, free our private + * texture object data. + */ + if (INTEL_DEBUG & DEBUG_TEXTURE) + fprintf(stderr, "do something to free texture heaps\n"); + } + + /* free the Mesa context */ + _mesa_free_context_data(&intel->ctx); + } +} + +GLboolean +intelUnbindContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; + /* XXX UnbindContext is called AFTER the new context is made current. + Hopefully shouldn't be a problem ? */ + FLUSH_VERTICES((&intel->ctx), 0); + intelFlush(&intel->ctx); + return GL_TRUE; +} + +GLboolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) +{ + +#if 0 + if (driDrawPriv) { + fprintf(stderr, "x %d, y %d, width %d, height %d\n", + driDrawPriv->x, driDrawPriv->y, driDrawPriv->w, driDrawPriv->h); + } +#endif + + if (driContextPriv) { + struct intel_context *intel = + (struct intel_context *) driContextPriv->driverPrivate; + struct intel_framebuffer *intel_fb = + (struct intel_framebuffer *) driDrawPriv->driverPrivate; + GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; + + /* this is a hack so we have a valid context when the region allocation + is done. Need a per-screen context? */ + intel->intelScreen->dummyctxptr = intel; + + /* update GLframebuffer size to match window if needed */ + driUpdateFramebufferSize(&intel->ctx, driDrawPriv); + + if (driReadPriv != driDrawPriv) { + driUpdateFramebufferSize(&intel->ctx, driReadPriv); + } + + _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); + + /* The drawbuffer won't always be updated by _mesa_make_current: + */ + if (intel->ctx.DrawBuffer == &intel_fb->Base) { + + if (intel->driDrawable != driDrawPriv) { + if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) { + int i; + + intel_fb->vblank_flags = driGetDefaultVBlankFlags(&intel->optionCache); + + (*dri_interface->getUST) (&intel_fb->swap_ust); + driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags, + &intel_fb->vbl_seq); + intel_fb->vbl_waited = intel_fb->vbl_seq; + + for (i = 0; i < 2; i++) { + if (intel_fb->color_rb[i]) + intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_seq; + } + } + } + } + + if ((intel->driDrawable != driDrawPriv) || + (intel->lastStamp != driDrawPriv->lastStamp)) { + intel->driDrawable = driDrawPriv; + intelWindowMoved(intel); + intel->lastStamp = driDrawPriv->lastStamp; + } + + } + else { + _mesa_make_current(NULL, NULL, NULL); + } + + return GL_TRUE; +} + +static void +intelContendedLock(struct intel_context *intel, GLuint flags) +{ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + __DRIscreenPrivate *sPriv = intel->driScreen; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + drmI830Sarea *sarea = intel->sarea; + + drmGetLock(intel->driFd, intel->hHWContext, flags); + + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("%s - got contended lock\n", __progname); + + /* If the window moved, may need to set a new cliprect now. + * + * NOTE: This releases and regains the hw lock, so all state + * checking must be done *after* this call: + */ + if (dPriv) + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + + if (sarea->width != intelScreen->front.width || + sarea->height != intelScreen->front.height) { + + intelUpdateScreenRotation(sPriv, sarea); + } + +#if 0 + if (sarea->width != intel->width || + sarea->height != intel->height || + sarea->rotation != intel->current_rotation) { + int numClipRects = intel->numClipRects; + + /* + * FIXME: Really only need to do this when drawing to a + * common back- or front buffer. + */ + + /* + * This will essentially drop the outstanding batchbuffer on the floor. + */ + intel->numClipRects = 0; + + INTEL_FIREVERTICES(intel); + + if (intel->batch->map != intel->batch->ptr) + intel_batchbuffer_flush(intel->batch); + + intel->numClipRects = numClipRects; + + /* force window update */ + intel->lastStamp = 0; + + intel->width = sarea->width; + intel->height = sarea->height; + intel->current_rotation = sarea->rotation; + } +#endif +} + + +/* Lock the hardware and validate our state. + */ +void LOCK_HARDWARE( struct intel_context *intel ) +{ + char __ret=0; + struct intel_framebuffer *intel_fb = NULL; + struct intel_renderbuffer *intel_rb = NULL; + _glthread_LOCK_MUTEX(lockMutex); + assert(!intel->locked); + + if (intel->driDrawable) { + intel_fb = intel->driDrawable->driverPrivate; + + if (intel_fb) + intel_rb = + intel_get_renderbuffer(&intel_fb->Base, + intel_fb->Base._ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : + BUFFER_BACK_LEFT); + } + + if (intel_rb && intel_fb->vblank_flags && + !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && + (intel_fb->vbl_waited - intel_rb->vbl_pending) > (1<<23)) { + drmVBlank vbl; + + vbl.request.type = DRM_VBLANK_ABSOLUTE; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + vbl.request.type |= DRM_VBLANK_SECONDARY; + } + + vbl.request.sequence = intel_rb->vbl_pending; + drmWaitVBlank(intel->driFd, &vbl); + intel_fb->vbl_waited = vbl.reply.sequence; + } + + DRM_CAS(intel->driHwLock, intel->hHWContext, + (DRM_LOCK_HELD|intel->hHWContext), __ret); + + if (__ret) + intelContendedLock( intel, 0 ); + + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("%s - locked\n", __progname); + + intel->locked = 1; +} + + + /* Unlock the hardware using the global current context + */ +void UNLOCK_HARDWARE( struct intel_context *intel ) +{ + assert(intel->locked); + intel->locked = 0; + + DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); + + _glthread_UNLOCK_MUTEX(lockMutex); + + if (INTEL_DEBUG & DEBUG_LOCK) + _mesa_printf("%s - unlocked\n", __progname); +} + diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h new file mode 100644 index 0000000000..1a4e7ca1cd --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -0,0 +1,225 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTELCONTEXT_INC +#define INTELCONTEXT_INC + + + +#include "mtypes.h" +#include "drm.h" +#include "mm.h" +#include "texmem.h" + +#include "intel_screen.h" +#include "i915_drm.h" +#include "i830_common.h" + + +#define DV_PF_555 (1<<8) +#define DV_PF_565 (2<<8) +#define DV_PF_8888 (3<<8) + +struct pipe_context; +struct pipe_region; +struct intel_context; +struct _DriBufferObject; + + +#define INTEL_WRITE_PART 0x1 +#define INTEL_WRITE_FULL 0x2 +#define INTEL_READ 0x4 + + + +#define INTEL_MAX_FIXUP 64 + +struct intel_context +{ + GLcontext ctx; /* the parent class */ + + struct pipe_context *pipe; + + GLint refcount; + + struct _DriFenceObject *last_swap_fence; + struct _DriFenceObject *first_swap_fence; + + struct intel_batchbuffer *batch; + + + GLboolean locked; + char *prevLockFile; + int prevLockLine; + + GLuint ClearColor565; + GLuint ClearColor8888; + + + /* These refer to the current drawing buffer: + */ + GLuint numClipRects; /**< cliprects for drawing */ + drm_clip_rect_t *pClipRects; + drm_clip_rect_t fboRect; /**< cliprect for rendering */ + + GLuint irqsEmitted; + drm_i915_irq_wait_t iw; + + drm_context_t hHWContext; + drmLock *driHwLock; + int driFd; + + __DRIdrawablePrivate *driDrawable; + __DRIscreenPrivate *driScreen; + intelScreenPrivate *intelScreen; + drmI830Sarea *sarea; + + GLuint lastStamp; + + /** + * Configuration cache + */ + driOptionCache optionCache; +}; + +/* These are functions now: + */ +void LOCK_HARDWARE( struct intel_context *intel ); +void UNLOCK_HARDWARE( struct intel_context *intel ); + +extern char *__progname; + +/* Will become a call into state_tracker: + */ +#define INTEL_FIREVERTICES(intel) + +/* ================================================================ + * Color packing: + */ + +#define INTEL_PACKCOLOR4444(r,g,b,a) \ + ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) + +#define INTEL_PACKCOLOR1555(r,g,b,a) \ + ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ + ((a) ? 0x8000 : 0)) + +#define INTEL_PACKCOLOR565(r,g,b) \ + ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) + +#define INTEL_PACKCOLOR8888(r,g,b,a) \ + ((a<<24) | (r<<16) | (g<<8) | b) + + + + +/* ================================================================ + * Debugging: + */ +#define DO_DEBUG 0 +#if DO_DEBUG +extern int INTEL_DEBUG; +#else +#define INTEL_DEBUG 0 +#endif + +#define DEBUG_TEXTURE 0x1 +#define DEBUG_STATE 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_BLIT 0x8 +#define DEBUG_MIPTREE 0x10 +#define DEBUG_FALLBACKS 0x20 +#define DEBUG_VERBOSE 0x40 +#define DEBUG_BATCH 0x80 +#define DEBUG_PIXEL 0x100 +#define DEBUG_BUFMGR 0x200 +#define DEBUG_REGION 0x400 +#define DEBUG_FBO 0x800 +#define DEBUG_LOCK 0x1000 + +#define DBG(...) do { if (INTEL_DEBUG & FILE_DEBUG_FLAG) _mesa_printf(__VA_ARGS__); } while(0) + + +#define PCI_CHIP_845_G 0x2562 +#define PCI_CHIP_I830_M 0x3577 +#define PCI_CHIP_I855_GM 0x3582 +#define PCI_CHIP_I865_G 0x2572 +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GME 0x27AE +#define PCI_CHIP_G33_G 0x29C2 +#define PCI_CHIP_Q35_G 0x29B2 +#define PCI_CHIP_Q33_G 0x29D2 + + +/* ================================================================ + * intel_context.c: + */ + +extern GLboolean intelInitContext(struct intel_context *intel, + const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate, + struct dd_function_table *functions); + +extern void intelGetLock(struct intel_context *intel, GLuint flags); + +extern void intelFinish(GLcontext * ctx); +extern void intelFlush(GLcontext * ctx); + + +/* ================================================================ + * intel_softpipe.c: + */ + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel ); + + +#define MI_BATCH_BUFFER_END (0xA<<23) + +#define FALLBACK( ctx, bit, mode ) + +/*====================================================================== + * Inline conversion functions. + * These are better-typed than the macros used previously: + */ +static INLINE struct intel_context * +intel_context(GLcontext * ctx) +{ + return (struct intel_context *) ctx; +} + +extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer + *rb); + +extern void intel_init_region_functions(struct pipe_context *pipe); + + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.c b/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.c new file mode 100644 index 0000000000..81e09698cc --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.c @@ -0,0 +1,282 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "imports.h" +#include "context.h" +#include "depthstencil.h" +#include "fbobject.h" +#include "framebuffer.h" +#include "hash.h" +#include "mtypes.h" +#include "renderbuffer.h" + +#include "intel_context.h" +#include "intel_fbo.h" +#include "intel_depthstencil.h" + +#include "pipe/p_context.h" + +/** + * The GL_EXT_framebuffer_object allows the user to create their own + * framebuffer objects consisting of color renderbuffers (0 or more), + * depth renderbuffers (0 or 1) and stencil renderbuffers (0 or 1). + * + * The spec considers depth and stencil renderbuffers to be totally independent + * buffers. In reality, most graphics hardware today uses a combined + * depth+stencil buffer (one 32-bit pixel = 24 bits of Z + 8 bits of stencil). + * + * This causes difficulty because the user may create some number of depth + * renderbuffers and some number of stencil renderbuffers and bind them + * together in framebuffers in any combination. + * + * This code manages all that. + * + * 1. Depth renderbuffers are always allocated in hardware as 32bpp + * GL_DEPTH24_STENCIL8 buffers. + * + * 2. Stencil renderbuffers are initially allocated in software as 8bpp + * GL_STENCIL_INDEX8 buffers. + * + * 3. Depth and Stencil renderbuffers use the PairedStencil and PairedDepth + * fields (respectively) to indicate if the buffer's currently paired + * with another stencil or depth buffer (respectively). + * + * 4. When a depth and stencil buffer are initially both attached to the + * current framebuffer, we merge the stencil buffer values into the + * depth buffer (really a depth+stencil buffer). The then hardware uses + * the combined buffer. + * + * 5. Whenever a depth or stencil buffer is reallocated (with + * glRenderbufferStorage) we undo the pairing and copy the stencil values + * from the combined depth/stencil buffer back to the stencil-only buffer. + * + * 6. We also undo the pairing when we find a change in buffer bindings. + * + * 7. If a framebuffer is only using a depth renderbuffer (no stencil), we + * just use the combined depth/stencil buffer and ignore the stencil values. + * + * 8. If a framebuffer is only using a stencil renderbuffer (no depth) we have + * to promote the 8bpp software stencil buffer to a 32bpp hardware + * depth+stencil buffer. + * + */ + + + +static void +map_regions(GLcontext * ctx, + struct intel_renderbuffer *depthRb, + struct intel_renderbuffer *stencilRb) +{ + struct intel_context *intel = intel_context(ctx); + if (depthRb && depthRb->region) { + intel->pipe->region_map(intel->pipe, depthRb->region); + depthRb->pfMap = depthRb->region->map; + depthRb->pfPitch = depthRb->region->pitch; + } + if (stencilRb && stencilRb->region) { + intel->pipe->region_map(intel->pipe, stencilRb->region); + stencilRb->pfMap = stencilRb->region->map; + stencilRb->pfPitch = stencilRb->region->pitch; + } +} + +static void +unmap_regions(GLcontext * ctx, + struct intel_renderbuffer *depthRb, + struct intel_renderbuffer *stencilRb) +{ + struct intel_context *intel = intel_context(ctx); + if (depthRb && depthRb->region) { + intel->pipe->region_unmap(intel->pipe, depthRb->region); + depthRb->pfMap = NULL; + depthRb->pfPitch = 0; + } + if (stencilRb && stencilRb->region) { + intel->pipe->region_unmap(intel->pipe, stencilRb->region); + stencilRb->pfMap = NULL; + stencilRb->pfPitch = 0; + } +} + + + +/** + * Undo the pairing/interleaving between depth and stencil buffers. + * irb should be a depth/stencil or stencil renderbuffer. + */ +void +intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb) +{ + if (irb->PairedStencil) { + /* irb is a depth/stencil buffer */ + struct gl_renderbuffer *stencilRb; + struct intel_renderbuffer *stencilIrb; + + ASSERT(irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + + stencilRb = _mesa_lookup_renderbuffer(ctx, irb->PairedStencil); + stencilIrb = intel_renderbuffer(stencilRb); + if (stencilIrb) { + /* need to extract stencil values from the depth buffer */ + ASSERT(stencilIrb->PairedDepth == irb->Base.Name); + map_regions(ctx, irb, stencilIrb); + _mesa_extract_stencil(ctx, &irb->Base, &stencilIrb->Base); + unmap_regions(ctx, irb, stencilIrb); + stencilIrb->PairedDepth = 0; + } + irb->PairedStencil = 0; + } + else if (irb->PairedDepth) { + /* irb is a stencil buffer */ + struct gl_renderbuffer *depthRb; + struct intel_renderbuffer *depthIrb; + + ASSERT(irb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || + irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + + depthRb = _mesa_lookup_renderbuffer(ctx, irb->PairedDepth); + depthIrb = intel_renderbuffer(depthRb); + if (depthIrb) { + /* need to extract stencil values from the depth buffer */ + ASSERT(depthIrb->PairedStencil == irb->Base.Name); + map_regions(ctx, depthIrb, irb); + _mesa_extract_stencil(ctx, &depthIrb->Base, &irb->Base); + unmap_regions(ctx, depthIrb, irb); + depthIrb->PairedStencil = 0; + } + irb->PairedDepth = 0; + } + else { + _mesa_problem(ctx, "Problem in undo_depth_stencil_pairing"); + } + + ASSERT(irb->PairedStencil == 0); + ASSERT(irb->PairedDepth == 0); +} + + +/** + * Examine the depth and stencil renderbuffers which are attached to the + * framebuffer. If both depth and stencil are attached, make sure that the + * renderbuffers are 'paired' (combined). If only depth or only stencil is + * attached, undo any previous pairing. + * + * Must be called if NewState & _NEW_BUFFER (when renderbuffer attachments + * change, for example). + */ +void +intel_validate_paired_depth_stencil(GLcontext * ctx, + struct gl_framebuffer *fb) +{ + struct intel_renderbuffer *depthRb, *stencilRb; + + depthRb = intel_get_renderbuffer(fb, BUFFER_DEPTH); + stencilRb = intel_get_renderbuffer(fb, BUFFER_STENCIL); + + if (depthRb && stencilRb) { + if (depthRb == stencilRb) { + /* Using a user-created combined depth/stencil buffer. + * Nothing to do. + */ + ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_STENCIL_EXT); + ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + } + else { + /* Separate depth/stencil buffers, need to interleave now */ + ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_COMPONENT); + ASSERT(stencilRb->Base._BaseFormat == GL_STENCIL_INDEX); + /* may need to interleave depth/stencil now */ + if (depthRb->PairedStencil == stencilRb->Base.Name) { + /* OK, the depth and stencil buffers are already interleaved */ + ASSERT(stencilRb->PairedDepth == depthRb->Base.Name); + } + else { + /* need to setup new pairing/interleaving */ + if (depthRb->PairedStencil) { + intel_unpair_depth_stencil(ctx, depthRb); + } + if (stencilRb->PairedDepth) { + intel_unpair_depth_stencil(ctx, stencilRb); + } + + ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + ASSERT(stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || + stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + + /* establish new pairing: interleave stencil into depth buffer */ + map_regions(ctx, depthRb, stencilRb); + _mesa_insert_stencil(ctx, &depthRb->Base, &stencilRb->Base); + unmap_regions(ctx, depthRb, stencilRb); + depthRb->PairedStencil = stencilRb->Base.Name; + stencilRb->PairedDepth = depthRb->Base.Name; + } + + } + } + else if (depthRb) { + /* Depth buffer but no stencil buffer. + * We'll use a GL_DEPTH24_STENCIL8 buffer and ignore the stencil bits. + */ + /* can't assert this until storage is allocated: + ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + */ + /* intel_undo any previous pairing */ + if (depthRb->PairedStencil) { + intel_unpair_depth_stencil(ctx, depthRb); + } + } + else if (stencilRb) { + /* Stencil buffer but no depth buffer. + * Since h/w doesn't typically support just 8bpp stencil w/out Z, + * we'll use a GL_DEPTH24_STENCIL8 buffer and ignore the depth bits. + */ + /* undo any previous pairing */ + if (stencilRb->PairedDepth) { + intel_unpair_depth_stencil(ctx, stencilRb); + } + if (stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT) { + /* promote buffer to GL_DEPTH24_STENCIL8 for hw rendering */ + _mesa_promote_stencil(ctx, &stencilRb->Base); + ASSERT(stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + } + } + + /* Finally, update the fb->_DepthBuffer and fb->_StencilBuffer fields */ + _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH); + if (depthRb && depthRb->PairedStencil) + _mesa_update_stencil_buffer(ctx, fb, BUFFER_DEPTH); + else + _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL); + + + /* The hardware should use fb->Attachment[BUFFER_DEPTH].Renderbuffer + * first, if present, then fb->Attachment[BUFFER_STENCIL].Renderbuffer + * if present. + */ +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.h b/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.h new file mode 100644 index 0000000000..2d3fc48b3a --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.h @@ -0,0 +1,14 @@ + +#ifndef INTEL_DEPTH_STENCIL_H +#define INTEL_DEPTH_STENCIL_H + + +extern void +intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb); + +extern void +intel_validate_paired_depth_stencil(GLcontext * ctx, + struct gl_framebuffer *fb); + + +#endif /* INTEL_DEPTH_STENCIL_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c new file mode 100644 index 0000000000..0b6c6a94ee --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c @@ -0,0 +1,629 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "imports.h" +#include "mtypes.h" +#include "fbobject.h" +#include "framebuffer.h" +#include "renderbuffer.h" +#include "context.h" +#include "texformat.h" +#include "texrender.h" + +#include "intel_context.h" +#include "intel_buffers.h" +#include "intel_depthstencil.h" +#include "intel_fbo.h" +#include "state_tracker/st_mipmap_tree.h" +/*#include "intel_tex.h"*/ + +#include "pipe/p_context.h" + +#define FILE_DEBUG_FLAG DEBUG_FBO + +#define INTEL_RB_CLASS 0x12345678 + + +/* XXX FBO: move this to intel_context.h (inlined) */ +/** + * Return a gl_renderbuffer ptr casted to intel_renderbuffer. + * NULL will be returned if the rb isn't really an intel_renderbuffer. + * This is determiend by checking the ClassID. + */ +struct intel_renderbuffer * +intel_renderbuffer(struct gl_renderbuffer *rb) +{ + struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; + if (irb && irb->Base.ClassID == INTEL_RB_CLASS) { + /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/ + return irb; + } + else + return NULL; +} + + +struct intel_renderbuffer * +intel_get_renderbuffer(struct gl_framebuffer *fb, GLuint attIndex) +{ + return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer); +} + + +void +intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) +{ + int current_page = intel_fb->pf_current_page; + int next_page = (current_page + 1) % intel_fb->pf_num_pages; + struct gl_renderbuffer *tmp_rb; + + /* Exchange renderbuffers if necessary but make sure their reference counts + * are preserved. + */ + if (intel_fb->color_rb[current_page] && + intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer != + &intel_fb->color_rb[current_page]->Base) { + tmp_rb = NULL; + _mesa_reference_renderbuffer(&tmp_rb, + intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + tmp_rb = &intel_fb->color_rb[current_page]->Base; + _mesa_reference_renderbuffer( + &intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer, tmp_rb); + _mesa_reference_renderbuffer(&tmp_rb, NULL); + } + + if (intel_fb->color_rb[next_page] && + intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer != + &intel_fb->color_rb[next_page]->Base) { + tmp_rb = NULL; + _mesa_reference_renderbuffer(&tmp_rb, + intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); + tmp_rb = &intel_fb->color_rb[next_page]->Base; + _mesa_reference_renderbuffer( + &intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer, tmp_rb); + _mesa_reference_renderbuffer(&tmp_rb, NULL); + } +} + + +struct pipe_region * +intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) +{ + struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex); + + if (irb) + return irb->region; + else + return NULL; +} + + + +/** + * Create a new framebuffer object. + */ +static struct gl_framebuffer * +intel_new_framebuffer(GLcontext * ctx, GLuint name) +{ + /* Only drawable state in intel_framebuffer at this time, just use Mesa's + * class + */ + return _mesa_new_framebuffer(ctx, name); +} + + +static void +intel_delete_renderbuffer(struct gl_renderbuffer *rb) +{ + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel = intel_context(ctx); + struct intel_renderbuffer *irb = intel_renderbuffer(rb); + + ASSERT(irb); + + DBG("freeing renderbuffer\n"); + + if (irb->PairedStencil || irb->PairedDepth) { + intel_unpair_depth_stencil(ctx, irb); + } + + if (intel && irb->region) { + intel->pipe->region_release(intel->pipe, &irb->region); + } + + _mesa_free(irb); +} + + + +/** + * Return a pointer to a specific pixel in a renderbuffer. + */ +static void * +intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb, + GLint x, GLint y) +{ + /* By returning NULL we force all software rendering to go through + * the span routines. + */ + return NULL; +} + + + +/** + * Called via glRenderbufferStorageEXT() to set the format and allocate + * storage for a user-created (or priv buffer) renderbuffer. + */ +static GLboolean +intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, + GLuint width, GLuint height) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_renderbuffer *irb = intel_renderbuffer(rb); + GLboolean softwareBuffer = GL_FALSE; + int cpp; + + switch (internalFormat) { + case GL_R3_G3_B2: + case GL_RGB4: + case GL_RGB5: + rb->_ActualFormat = GL_RGB5; + rb->DataType = GL_UNSIGNED_BYTE; + rb->RedBits = 5; + rb->GreenBits = 6; + rb->BlueBits = 5; + cpp = 2; + break; + case GL_RGB: + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + case GL_RGBA: + case GL_RGBA2: + case GL_RGBA4: + case GL_RGB5_A1: + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + rb->_ActualFormat = GL_RGBA8; + rb->DataType = GL_UNSIGNED_BYTE; + rb->RedBits = 8; + rb->GreenBits = 8; + rb->BlueBits = 8; + rb->AlphaBits = 8; + cpp = 4; + break; + case GL_STENCIL_INDEX: + case GL_STENCIL_INDEX1_EXT: + case GL_STENCIL_INDEX4_EXT: + case GL_STENCIL_INDEX8_EXT: + case GL_STENCIL_INDEX16_EXT: + /* alloc a depth+stencil buffer */ + rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->DataType = GL_UNSIGNED_INT_24_8_EXT; + rb->StencilBits = 8; + cpp = 4; + break; + case GL_DEPTH_COMPONENT16: + rb->_ActualFormat = GL_DEPTH_COMPONENT16; + rb->DataType = GL_UNSIGNED_SHORT; + rb->DepthBits = 16; + cpp = 2; + break; + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->DataType = GL_UNSIGNED_INT_24_8_EXT; + rb->DepthBits = 24; + cpp = 4; + break; + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; + rb->DataType = GL_UNSIGNED_INT_24_8_EXT; + rb->DepthBits = 24; + rb->StencilBits = 8; + cpp = 4; + break; + default: + _mesa_problem(ctx, + "Unexpected format (%x) in intel_alloc_renderbuffer_storage", internalFormat); + return GL_FALSE; + } + + intelFlush(ctx); + + /* free old region */ + if (irb->region) { + intel->pipe->region_release(intel->pipe, &irb->region); + } + + /* allocate new memory region/renderbuffer */ + if (softwareBuffer) { + return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat, + width, height); + } + else { + /* Choose a pitch to match hardware requirements: + */ + GLuint pitch = ((cpp * width + 63) & ~63) / cpp; + + /* alloc hardware renderbuffer */ + DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, + height, pitch); + + irb->region = intel->pipe->region_alloc(intel->pipe, cpp, pitch, height); + if (!irb->region) + return GL_FALSE; /* out of memory? */ + + ASSERT(irb->region->buffer); + + rb->Width = width; + rb->Height = height; + + /* update the surface's size too */ + rb->surface->width = width; + rb->surface->height = height; + rb->surface->region = irb->region; + + /* This sets the Get/PutRow/Value functions */ + // intel_set_span_functions(&irb->Base); + + return GL_TRUE; + } +} + + +static void +intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, + GLuint width, GLuint height) +{ + struct intel_framebuffer *intel_fb = (struct intel_framebuffer*)fb; + int i; + + _mesa_resize_framebuffer(ctx, fb, width, height); + + fb->Initialized = GL_TRUE; /* XXX remove someday */ + + if (fb->Name != 0) { + return; + } + + /* Make sure all window system renderbuffers are up to date */ + for (i = 0; i < 3; i++) { + struct gl_renderbuffer *rb = &intel_fb->color_rb[i]->Base; + + /* only resize if size is changing */ + if (rb && (rb->Width != width || rb->Height != height)) { + rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height); + } + } +} + +static GLboolean +intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, GLuint width, GLuint height) +{ + _mesa_problem(ctx, "intel_op_alloc_storage should never be called."); + return GL_FALSE; +} + + + +struct intel_renderbuffer * +intel_new_renderbuffer_fb(GLuint intFormat) +{ + struct intel_renderbuffer *irb; + + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, 0); + irb->Base.ClassID = INTEL_RB_CLASS; + irb->Base.InternalFormat = intFormat; + + switch (intFormat) { + case GL_RGB5: + case GL_RGBA8: + irb->Base._BaseFormat = GL_RGBA; + break; + case GL_DEPTH_COMPONENT16: + irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + break; + case GL_DEPTH24_STENCIL8_EXT: + irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; + break; + default: + _mesa_problem(NULL, + "Unexpected intFormat in intel_create_renderbuffer"); + return NULL; + } + + /* intel-specific methods */ + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; + irb->Base.GetPointer = intel_get_pointer; + /* span routines set in alloc_storage function */ + + irb->Base.surface = intel_new_surface(intFormat); + irb->Base.surface->rb = irb; + + return irb; +} + +/** + * Create a new renderbuffer object. + * Typically called via glBindRenderbufferEXT(). + */ +static struct gl_renderbuffer * +intel_new_renderbuffer(GLcontext * ctx, GLuint name) +{ + /*struct intel_context *intel = intel_context(ctx); */ + struct intel_renderbuffer *irb; + + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, name); + irb->Base.ClassID = INTEL_RB_CLASS; + + /* intel-specific methods */ + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; + irb->Base.GetPointer = intel_get_pointer; + /* span routines set in alloc_storage function */ + + irb->Base.surface = intel_new_surface(0 /*unknown format*/); + irb->Base.surface->rb = irb; + + return &irb->Base; +} + + +/** + * Called via glBindFramebufferEXT(). + */ +static void +intel_bind_framebuffer(GLcontext * ctx, GLenum target, + struct gl_framebuffer *fb, struct gl_framebuffer *fbread) +{ + if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) { + intel_draw_buffer(ctx, fb); + /* Integer depth range depends on depth buffer bits */ + /* XXX + */ +// ctx->Driver.DepthRange(ctx, ctx->Viewport.Near, ctx->Viewport.Far); + } + else { + /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */ + } +} + + +/** + * Called via glFramebufferRenderbufferEXT(). + */ +static void +intel_framebuffer_renderbuffer(GLcontext * ctx, + struct gl_framebuffer *fb, + GLenum attachment, struct gl_renderbuffer *rb) +{ + DBG("Intel FramebufferRenderbuffer %u %u\n", fb->Name, rb ? rb->Name : 0); + + intelFlush(ctx); + + _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb); + intel_draw_buffer(ctx, fb); +} + + +/** + * When glFramebufferTexture[123]D is called this function sets up the + * gl_renderbuffer wrapper around the texture image. + * This will have the region info needed for hardware rendering. + */ +static struct intel_renderbuffer * +intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage) +{ + const GLuint name = ~0; /* not significant, but distinct for debugging */ + struct intel_renderbuffer *irb; + + /* make an intel_renderbuffer to wrap the texture image */ + irb = CALLOC_STRUCT(intel_renderbuffer); + if (!irb) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, name); + irb->Base.ClassID = INTEL_RB_CLASS; + + if (texImage->TexFormat == &_mesa_texformat_argb8888) { + irb->Base._ActualFormat = GL_RGBA8; + irb->Base._BaseFormat = GL_RGBA; + DBG("Render to RGBA8 texture OK\n"); + } + else if (texImage->TexFormat == &_mesa_texformat_rgb565) { + irb->Base._ActualFormat = GL_RGB5; + irb->Base._BaseFormat = GL_RGB; + DBG("Render to RGB5 texture OK\n"); + } + else if (texImage->TexFormat == &_mesa_texformat_z16) { + irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; + irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + DBG("Render to DEPTH16 texture OK\n"); + } + else { + DBG("Render to texture BAD FORMAT %d\n", + texImage->TexFormat->MesaFormat); + _mesa_free(irb); + return NULL; + } + + irb->Base.InternalFormat = irb->Base._ActualFormat; + irb->Base.Width = texImage->Width; + irb->Base.Height = texImage->Height; + irb->Base.DataType = GL_UNSIGNED_BYTE; /* FBO XXX fix */ + irb->Base.RedBits = texImage->TexFormat->RedBits; + irb->Base.GreenBits = texImage->TexFormat->GreenBits; + irb->Base.BlueBits = texImage->TexFormat->BlueBits; + irb->Base.AlphaBits = texImage->TexFormat->AlphaBits; + irb->Base.DepthBits = texImage->TexFormat->DepthBits; + + irb->Base.Delete = intel_delete_renderbuffer; + irb->Base.AllocStorage = intel_nop_alloc_storage; + +// intel_set_span_functions(&irb->Base); + + irb->RenderToTexture = GL_TRUE; + + return irb; +} + + +/** + * Called by glFramebufferTexture[123]DEXT() (and other places) to + * prepare for rendering into texture memory. This might be called + * many times to choose different texture levels, cube faces, etc + * before intel_finish_render_texture() is ever called. + */ +static void +intel_render_texture(GLcontext * ctx, + struct gl_framebuffer *fb, + struct gl_renderbuffer_attachment *att) +{ +#if 0 + struct intel_context *intel = intel_context(ctx); + struct gl_texture_image *newImage + = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; + struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); + struct st_texture_image *st_image; + GLuint imageOffset; + + (void) fb; + + ASSERT(newImage); + + if (!irb) { + irb = intel_wrap_texture(ctx, newImage); + if (irb) { + /* bind the wrapper to the attachment point */ + _mesa_reference_renderbuffer(&att->Renderbuffer, &irb->Base); + } + else { + /* fallback to software rendering */ + _mesa_render_texture(ctx, fb, att); + return; + } + } + + DBG("Begin render texture tid %x tex=%u w=%d h=%d refcount=%d\n", + _glthread_GetID(), + att->Texture->Name, newImage->Width, newImage->Height, + irb->Base.RefCount); + + /* point the renderbufer's region to the texture image region */ + st_image = st_texture_image(newImage); + if (irb->region != st_image->mt->region) { + if (irb->region) + intel->pipe->region_release(intel->pipe, &irb->region); + pipe_region_reference(&irb->region, st_image->mt->region); + } + + /* compute offset of the particular 2D image within the texture region */ + imageOffset = st_miptree_image_offset(st_image->mt, + att->CubeMapFace, + att->TextureLevel); + + if (att->Texture->Target == GL_TEXTURE_3D) { + const GLuint *offsets = st_miptree_depth_offsets(st_image->mt, + att->TextureLevel); + imageOffset += offsets[att->Zoffset]; + } + + /* store that offset in the region */ +#if 0 + st_image->mt->region->draw_offset = imageOffset; +#endif + + /* update drawing region, etc */ + intel_draw_buffer(ctx, fb); +#endif +} + + +/** + * Called by Mesa when rendering to a texture is done. + */ +static void +intel_finish_render_texture(GLcontext * ctx, + struct gl_renderbuffer_attachment *att) +{ + struct intel_context *intel = intel_context(ctx); + struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); + + DBG("End render texture (tid %x) tex %u\n", _glthread_GetID(), att->Texture->Name); + + if (irb) { + /* just release the region */ + intel->pipe->region_release(intel->pipe, &irb->region); + } + else if (att->Renderbuffer) { + /* software fallback */ + _mesa_finish_render_texture(ctx, att); + /* XXX FBO: Need to unmap the buffer (or in intelSpanRenderStart???) */ + } +} + + +/** + * Do one-time context initializations related to GL_EXT_framebuffer_object. + * Hook in device driver functions. + */ +void +intel_fbo_init(struct intel_context *intel) +{ + intel->ctx.Driver.NewFramebuffer = intel_new_framebuffer; + intel->ctx.Driver.NewRenderbuffer = intel_new_renderbuffer; + intel->ctx.Driver.BindFramebuffer = intel_bind_framebuffer; + intel->ctx.Driver.FramebufferRenderbuffer = intel_framebuffer_renderbuffer; + intel->ctx.Driver.RenderTexture = intel_render_texture; + intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture; + intel->ctx.Driver.ResizeBuffers = intel_resize_buffers; +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h new file mode 100644 index 0000000000..0f99a3e98d --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h @@ -0,0 +1,127 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_FBO_H +#define INTEL_FBO_H + + +#include "pipe/p_state.h" +#include "pipe/softpipe/sp_surface.h" + + +struct intel_context; +struct pipe_region; + + +/** + * Intel framebuffer, derived from gl_framebuffer. + */ +struct intel_framebuffer +{ + struct gl_framebuffer Base; + + struct intel_renderbuffer *color_rb[3]; + + /* Drawable page flipping state */ + GLboolean pf_active; + GLuint pf_seq; + GLint pf_pipes; + GLint pf_current_page; + GLint pf_num_pages; + + /* VBI + */ + GLuint vbl_seq; + GLuint vblank_flags; + GLuint vbl_waited; + + int64_t swap_ust; + int64_t swap_missed_ust; + + GLuint swap_count; + GLuint swap_missed_count; +}; + + +/** + * Intel renderbuffer, derived from gl_renderbuffer. + * Note: The PairedDepth and PairedStencil fields use renderbuffer IDs, + * not pointers because in some circumstances a deleted renderbuffer could + * result in a dangling pointer here. + */ +struct intel_renderbuffer +{ + struct gl_renderbuffer Base; + struct pipe_region *region; + void *pfMap; /* possibly paged flipped map pointer */ + GLuint pfPitch; /* possibly paged flipped pitch */ + GLboolean RenderToTexture; /* RTT? */ + + GLuint PairedDepth; /**< only used if this is a depth renderbuffer */ + GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */ + + GLuint pf_pending; /**< sequence number of pending flip */ + + GLuint vbl_pending; /**< vblank sequence number of pending flip */ + + struct intel_surface *surface; +}; + +#if 0 +extern struct intel_renderbuffer *intel_create_renderbuffer(GLenum intFormat, + GLsizei width, + GLsizei height, + int offset, + int pitch, + int cpp, + void *map); +#endif + +extern struct intel_renderbuffer *intel_new_renderbuffer_fb(GLuint intFormat); + +extern void intel_fbo_init(struct intel_context *intel); + + +/* XXX make inline or macro */ +extern struct intel_renderbuffer *intel_get_renderbuffer(struct gl_framebuffer + *fb, + GLuint attIndex); + +extern void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb); + + +/* XXX make inline or macro */ +extern struct pipe_region *intel_get_rb_region(struct gl_framebuffer *fb, + GLuint attIndex); + + + +extern struct pipe_surface * +intel_new_surface(GLuint intFormat); + + +#endif /* INTEL_FBO_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c new file mode 100644 index 0000000000..154a3e289f --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c @@ -0,0 +1,135 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include +#include +#include +#include + +#include "mtypes.h" +#include "context.h" + +#include "intel_context.h" +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "drm.h" + +#define FILE_DEBUG_FLAG DEBUG_IOCTL + +int +intelEmitIrqLocked(struct intel_context *intel) +{ + drmI830IrqEmit ie; + int ret, seq; + + assert(((*(int *) intel->driHwLock) & ~DRM_LOCK_CONT) == + (DRM_LOCK_HELD | intel->hHWContext)); + + ie.irq_seq = &seq; + + ret = drmCommandWriteRead(intel->driFd, DRM_I830_IRQ_EMIT, + &ie, sizeof(ie)); + if (ret) { + fprintf(stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret); + exit(1); + } + + DBG("%s --> %d\n", __FUNCTION__, seq); + + return seq; +} + +void +intelWaitIrq(struct intel_context *intel, int seq) +{ + int ret; + + DBG("%s %d\n", __FUNCTION__, seq); + + intel->iw.irq_seq = seq; + + do { + ret = + drmCommandWrite(intel->driFd, DRM_I830_IRQ_WAIT, &intel->iw, + sizeof(intel->iw)); + } while (ret == -EAGAIN || ret == -EINTR); + + if (ret) { + fprintf(stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret); + exit(1); + } +} + + +void +intel_batch_ioctl(struct intel_context *intel, + GLuint start_offset, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) +{ + drmI830BatchBuffer batch; + + assert(intel->locked); + assert(used); + + DBG("%s used %d offset %x..%x ignore_cliprects %d\n", + __FUNCTION__, + used, start_offset, start_offset + used, ignore_cliprects); + + /* Throw away non-effective packets. Won't work once we have + * hardware contexts which would preserve statechanges beyond a + * single buffer. + */ + + + + batch.start = start_offset; + batch.used = used; + batch.cliprects = intel->pClipRects; + batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + batch.DR1 = 0; + batch.DR4 = 0; /* still need this ? */ + + DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n", + __FUNCTION__, + batch.start, + batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); + + if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, + sizeof(batch))) { + fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n", -errno); + UNLOCK_HARDWARE(intel); + exit(1); + } + + /* FIXME: use hardware contexts to avoid 'losing' hardware after + * each buffer flush. + */ + //intel->vtbl.lost_hardware(intel); +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h new file mode 100644 index 0000000000..e8d07de893 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_IOCTL_H +#define INTEL_IOCTL_H + +#include "intel_context.h" + +void intelWaitIrq(struct intel_context *intel, int seq); +int intelEmitIrqLocked(struct intel_context *intel); + +void intel_batch_ioctl(struct intel_context *intel, + GLuint start_offset, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock); +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.c new file mode 100644 index 0000000000..1a9aa10115 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.c @@ -0,0 +1,293 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_mipmap_tree.h" +#include "enums.h" + +#include "pipe/p_state.h" +#include "pipe/p_context.h" + + +#define DBG if(0) printf + +static GLenum +target_to_target(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + return GL_TEXTURE_CUBE_MAP_ARB; + default: + return target; + } +} + +struct pipe_mipmap_tree * +st_miptree_create(struct pipe_context *pipe, + GLenum target, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, GLuint cpp, GLuint compress_byte) +{ + GLboolean ok; + struct pipe_mipmap_tree *mt = calloc(sizeof(*mt), 1); + + DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), + _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); + + mt->target = target_to_target(target); + mt->internal_format = internal_format; + mt->first_level = first_level; + mt->last_level = last_level; + mt->width0 = width0; + mt->height0 = height0; + mt->depth0 = depth0; + mt->cpp = compress_byte ? compress_byte : cpp; + mt->compressed = compress_byte ? 1 : 0; + mt->refcount = 1; + + ok = pipe->mipmap_tree_layout(pipe, mt); + if (ok) + mt->region = pipe->region_alloc(pipe, + mt->cpp, mt->pitch, mt->total_height); + + if (!mt->region) { + free(mt); + return NULL; + } + + return mt; +} + + +void +st_miptree_reference(struct pipe_mipmap_tree **dst, + struct pipe_mipmap_tree *src) +{ + src->refcount++; + *dst = src; + DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); +} + +void +st_miptree_release(struct pipe_context *pipe, + struct pipe_mipmap_tree **mt) +{ + if (!*mt) + return; + + DBG("%s %p refcount will be %d\n", __FUNCTION__, *mt, (*mt)->refcount - 1); + if (--(*mt)->refcount <= 0) { + GLuint i; + + DBG("%s deleting %p\n", __FUNCTION__, *mt); + + pipe->region_release(pipe, &((*mt)->region)); + + for (i = 0; i < MAX_TEXTURE_LEVELS; i++) + if ((*mt)->level[i].image_offset) + free((*mt)->level[i].image_offset); + + free(*mt); + } + *mt = NULL; +} + + + + +/* Can the image be pulled into a unified mipmap tree. This mirrors + * the completeness test in a lot of ways. + * + * Not sure whether I want to pass gl_texture_image here. + */ +GLboolean +st_miptree_match_image(struct pipe_mipmap_tree *mt, + struct gl_texture_image *image, + GLuint face, GLuint level) +{ + /* Images with borders are never pulled into mipmap trees. + */ + if (image->Border) + return GL_FALSE; + + if (image->InternalFormat != mt->internal_format || + image->IsCompressed != mt->compressed) + return GL_FALSE; + + /* Test image dimensions against the base level image adjusted for + * minification. This will also catch images not present in the + * tree, changed targets, etc. + */ + if (image->Width != mt->level[level].width || + image->Height != mt->level[level].height || + image->Depth != mt->level[level].depth) + return GL_FALSE; + + return GL_TRUE; +} + + +/* Although we use the image_offset[] array to store relative offsets + * to cube faces, Mesa doesn't know anything about this and expects + * each cube face to be treated as a separate image. + * + * These functions present that view to mesa: + */ +const GLuint * +st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) +{ + static const GLuint zero = 0; + + if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) + return &zero; + else + return mt->level[level].image_offset; +} + + +GLuint +st_miptree_image_offset(struct pipe_mipmap_tree * mt, + GLuint face, GLuint level) +{ + if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) + return (mt->level[level].level_offset + + mt->level[level].image_offset[face] * mt->cpp); + else + return mt->level[level].level_offset; +} + + + +/** + * Map a teximage in a mipmap tree. + * \param row_stride returns row stride in bytes + * \param image_stride returns image stride in bytes (for 3D textures). + * \return address of mapping + */ +GLubyte * +st_miptree_image_map(struct pipe_context *pipe, + struct pipe_mipmap_tree * mt, + GLuint face, + GLuint level, + GLuint * row_stride, GLuint * image_offsets) +{ + GLubyte *ptr; + DBG("%s \n", __FUNCTION__); + + if (row_stride) + *row_stride = mt->pitch * mt->cpp; + + if (image_offsets) + memcpy(image_offsets, mt->level[level].image_offset, + mt->level[level].depth * sizeof(GLuint)); + + ptr = pipe->region_map(pipe, mt->region); + + return ptr + st_miptree_image_offset(mt, face, level); +} + +void +st_miptree_image_unmap(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt) +{ + DBG("%s\n", __FUNCTION__); + pipe->region_unmap(pipe, mt->region); +} + + + +/* Upload data for a particular image. + */ +void +st_miptree_image_data(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch) +{ + GLuint depth = dst->level[level].depth; + GLuint dst_offset = st_miptree_image_offset(dst, face, level); + const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); + GLuint i; + GLuint height = 0; + + DBG("%s\n", __FUNCTION__); + for (i = 0; i < depth; i++) { + height = dst->level[level].height; + if(dst->compressed) + height /= 4; + pipe->region_data(pipe, dst->region, + dst_offset + dst_depth_offset[i], /* dst_offset */ + 0, 0, /* dstx, dsty */ + src, + src_row_pitch, + 0, 0, /* source x, y */ + dst->level[level].width, height); /* width, height */ + + src += src_image_pitch * dst->cpp; + } +} + +/* Copy mipmap image between trees + */ +void +st_miptree_image_copy(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, GLuint level, + struct pipe_mipmap_tree *src) +{ + GLuint width = src->level[level].width; + GLuint height = src->level[level].height; + GLuint depth = src->level[level].depth; + GLuint dst_offset = st_miptree_image_offset(dst, face, level); + GLuint src_offset = st_miptree_image_offset(src, face, level); + const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); + const GLuint *src_depth_offset = st_miptree_depth_offsets(src, level); + GLuint i; + + if (dst->compressed) + height /= 4; + for (i = 0; i < depth; i++) { + pipe->region_copy(pipe, + dst->region, dst_offset + dst_depth_offset[i], + 0, + 0, + src->region, src_offset + src_depth_offset[i], + 0, 0, width, height); + } + +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.h b/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.h new file mode 100644 index 0000000000..eafe47c029 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.h @@ -0,0 +1,105 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_MIPMAP_TREE_H +#define INTEL_MIPMAP_TREE_H + + +#include "main/mtypes.h" + +struct pipe_context; +struct pipe_mipmap_tree; +struct pipe_region; + + +struct pipe_mipmap_tree *st_miptree_create(struct pipe_context *pipe, + GLenum target, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, + GLuint cpp, + GLuint compress_byte); + +void st_miptree_reference(struct pipe_mipmap_tree **dst, + struct pipe_mipmap_tree *src); + +void st_miptree_release(struct pipe_context *pipe, + struct pipe_mipmap_tree **mt); + +/* Check if an image fits an existing mipmap tree layout + */ +GLboolean st_miptree_match_image(struct pipe_mipmap_tree *mt, + struct gl_texture_image *image, + GLuint face, GLuint level); + +/* Return a pointer to an image within a tree. Return image stride as + * well. + */ +GLubyte *st_miptree_image_map(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt, + GLuint face, + GLuint level, + GLuint * row_stride, GLuint * image_stride); + +void st_miptree_image_unmap(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt); + + +/* Return the linear offset of an image relative to the start of the + * tree: + */ +GLuint st_miptree_image_offset(struct pipe_mipmap_tree *mt, + GLuint face, GLuint level); + +/* Return pointers to each 2d slice within an image. Indexed by depth + * value. + */ +const GLuint *st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, + GLuint level); + + +/* Upload an image into a tree + */ +void st_miptree_image_data(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch); + +/* Copy an image between two trees + */ +void st_miptree_image_copy(struct pipe_context *pipe, + struct pipe_mipmap_tree *dst, + GLuint face, GLuint level, + struct pipe_mipmap_tree *src); + + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_reg.h b/src/mesa/drivers/dri/intel_winsys/intel_reg.h new file mode 100644 index 0000000000..7828ba6ad3 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_reg.h @@ -0,0 +1,88 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef _INTEL_REG_H_ +#define _INTEL_REG_H_ + + + +#define CMD_3D (0x3<<29) + + +#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) +#define PRIM_INDIRECT (1<<23) +#define PRIM_INLINE (0<<23) +#define PRIM_INDIRECT_SEQUENTIAL (0<<17) +#define PRIM_INDIRECT_ELTS (1<<17) + +#define PRIM3D_TRILIST (0x0<<18) +#define PRIM3D_TRISTRIP (0x1<<18) +#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) +#define PRIM3D_TRIFAN (0x3<<18) +#define PRIM3D_POLY (0x4<<18) +#define PRIM3D_LINELIST (0x5<<18) +#define PRIM3D_LINESTRIP (0x6<<18) +#define PRIM3D_RECTLIST (0x7<<18) +#define PRIM3D_POINTLIST (0x8<<18) +#define PRIM3D_DIB (0x9<<18) +#define PRIM3D_MASK (0x1f<<18) + +#define I915PACKCOLOR4444(r,g,b,a) \ + ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) + +#define I915PACKCOLOR1555(r,g,b,a) \ + ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ + ((a) ? 0x8000 : 0)) + +#define I915PACKCOLOR565(r,g,b) \ + ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) + +#define I915PACKCOLOR8888(r,g,b,a) \ + ((a<<24) | (r<<16) | (g<<8) | b) + + + + +#define BR00_BITBLT_CLIENT 0x40000000 +#define BR00_OP_COLOR_BLT 0x10000000 +#define BR00_OP_SRC_COPY_BLT 0x10C00000 +#define BR13_SOLID_PATTERN 0x80000000 + +#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) +#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) +#define XY_COLOR_BLT_WRITE_RGB (1<<20) + +#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) +#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) +#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) + +#define MI_WAIT_FOR_EVENT ((0x3<<23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c new file mode 100644 index 0000000000..81ab435f42 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -0,0 +1,597 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "context.h" +#include "framebuffer.h" +#include "matrix.h" +#include "renderbuffer.h" +#include "simple_list.h" +#include "utils.h" +#include "vblank.h" +#include "xmlpool.h" + +#include "intel_screen.h" +#include "intel_batchbuffer.h" +#include "intel_buffers.h" +/*#include "intel_tex.h"*/ +#include "intel_ioctl.h" +#include "intel_fbo.h" + +#include "i830_dri.h" +#include "dri_bufpool.h" + +#include "pipe/p_context.h" + + + +PUBLIC const char __driConfigOptions[] = + DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE + DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) + DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) + DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY + DRI_CONF_FORCE_S3TC_ENABLE(false) + DRI_CONF_ALLOW_LARGE_TEXTURES(1) + DRI_CONF_SECTION_END DRI_CONF_END; + const GLuint __driNConfigOptions = 4; + +#ifdef USE_NEW_INTERFACE + static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; +#endif /*USE_NEW_INTERFACE */ + + extern const struct dri_extension card_extensions[]; + + + + + +static void +intelPrintDRIInfo(intelScreenPrivate * intelScreen, + __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) +{ + fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", + intelScreen->front.size, intelScreen->front.offset, + intelScreen->front.pitch); + fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); +} + + +static void +intelPrintSAREA(const drmI830Sarea * sarea) +{ + fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, + sarea->height); + fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); + fprintf(stderr, + "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->front_offset, sarea->front_size, + (unsigned) sarea->front_handle); + fprintf(stderr, + "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->back_offset, sarea->back_size, + (unsigned) sarea->back_handle); + fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->depth_offset, sarea->depth_size, + (unsigned) sarea->depth_handle); + fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); + fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); + fprintf(stderr, + "SAREA: rotated offset: 0x%08x size: 0x%x\n", + sarea->rotated_offset, sarea->rotated_size); + fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); +} + + + +/** + * Use the information in the sarea to update the screen parameters + * related to screen rotation. Needs to be called locked. + */ +void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) +{ + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + + if (intelScreen->front.map) { + drmUnmap(intelScreen->front.map, intelScreen->front.size); + intelScreen->front.map = NULL; + } + + if (intelScreen->front.buffer) + driDeleteBuffers(1, &intelScreen->front.buffer); + + intelScreen->front.width = sarea->width; + intelScreen->front.height = sarea->height; + intelScreen->front.offset = sarea->front_offset; + intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; + intelScreen->front.size = sarea->front_size; + intelScreen->front.handle = sarea->front_handle; + + assert( sarea->front_size >= + intelScreen->front.pitch * intelScreen->front.height ); + + if (!sarea->front_handle) + return; + + if (drmMap(sPriv->fd, + sarea->front_handle, + intelScreen->front.size, + (drmAddress *) & intelScreen->front.map) != 0) { + _mesa_problem(NULL, "drmMap(frontbuffer) failed!"); + return; + } + + if (intelScreen->staticPool) { + driGenBuffers(intelScreen->staticPool, "static region", 1, + &intelScreen->front.buffer, 64, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); + + driBOSetStatic(intelScreen->front.buffer, + intelScreen->front.offset, + intelScreen->front.pitch * intelScreen->front.height, + intelScreen->front.map, 0); + } +} + + + +GLboolean +intelCreatePools(intelScreenPrivate *intelScreen) +{ + unsigned batchPoolSize = 1024*1024; + __DRIscreenPrivate * sPriv = intelScreen->driScrnPriv; + + if (intelScreen->havePools) + return GL_TRUE; + + batchPoolSize /= BATCH_SZ; + intelScreen->regionPool = driDRMPoolInit(sPriv->fd); + + if (!intelScreen->regionPool) + return GL_FALSE; + + intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); + + if (!intelScreen->staticPool) + return GL_FALSE; + + intelScreen->texPool = intelScreen->regionPool; + + intelScreen->batchPool = driBatchPoolInit(sPriv->fd, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT | + DRM_BO_FLAG_MEM_LOCAL, + BATCH_SZ, + batchPoolSize, 5); + if (!intelScreen->batchPool) { + fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); + return GL_FALSE; + } + + intelScreen->havePools = GL_TRUE; + + intelUpdateScreenRotation(sPriv, intelScreen->sarea); + + return GL_TRUE; +} + + +static GLboolean +intelInitDriver(__DRIscreenPrivate * sPriv) +{ + intelScreenPrivate *intelScreen; + I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; + + PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = + (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> + getProcAddress("glxEnableExtension")); + void *const psc = sPriv->psc->screenConfigs; + + if (sPriv->devPrivSize != sizeof(I830DRIRec)) { + fprintf(stderr, + "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); + return GL_FALSE; + } + + /* Allocate the private area */ + intelScreen = (intelScreenPrivate *) CALLOC(sizeof(intelScreenPrivate)); + if (!intelScreen) + return GL_FALSE; + + /* parse information in __driConfigOptions */ + driParseOptionInfo(&intelScreen->optionCache, + __driConfigOptions, __driNConfigOptions); + + intelScreen->driScrnPriv = sPriv; + sPriv->private = (void *) intelScreen; + + intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + + gDRIPriv->sarea_priv_offset); + intelScreen->deviceID = gDRIPriv->deviceID; + intelScreen->front.cpp = gDRIPriv->cpp; + intelScreen->drmMinor = sPriv->drmMinor; + + assert(gDRIPriv->bitsPerPixel == 16 || + gDRIPriv->bitsPerPixel == 32); + + intelUpdateScreenRotation(sPriv, intelScreen->sarea); + + if (0) + intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); + + if (glx_enable_extension != NULL) { + (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); + (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); + (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); + } + + return GL_TRUE; +} + + +static void +intelDestroyScreen(__DRIscreenPrivate * sPriv) +{ + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + +// intelUnmapScreenRegions(intelScreen); + + if (intelScreen->havePools) { + driPoolTakeDown(intelScreen->regionPool); + driPoolTakeDown(intelScreen->staticPool); + driPoolTakeDown(intelScreen->batchPool); + } + FREE(intelScreen); + sPriv->private = NULL; +} + + +/** + * This is called when we need to set up GL rendering to a new X window. + */ +static GLboolean +intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, + __DRIdrawablePrivate * driDrawPriv, + const __GLcontextModes * mesaVis, GLboolean isPixmap) +{ + if (isPixmap) { + return GL_FALSE; /* not implemented */ + } + else { + GLboolean swStencil = (mesaVis->stencilBits > 0 && + mesaVis->depthBits != 24); + GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); + + struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); + + if (!intel_fb) + return GL_FALSE; + + _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); + + { + /* fake frontbuffer */ + /* XXX allocation should only happen in the unusual case + it's actually needed */ + intel_fb->color_rb[0] + = intel_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, + &intel_fb->color_rb[0]->Base); + } + + if (mesaVis->doubleBufferMode) { + intel_fb->color_rb[1] + = intel_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, + &intel_fb->color_rb[1]->Base); + } + + if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { + /* combined depth/stencil buffer */ + struct intel_renderbuffer *depthStencilRb + = intel_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + /* note: bind RB to two attachment points */ + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, + &depthStencilRb->Base); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL, + &depthStencilRb->Base); + } + else if (mesaVis->depthBits == 16) { + /* just 16-bit depth buffer, no hw stencil */ + struct intel_renderbuffer *depthRb + = intel_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); + } + + + /* now add any/all software-based renderbuffers we may need */ + _mesa_add_soft_renderbuffers(&intel_fb->Base, + GL_FALSE, /* never sw color */ + GL_FALSE, /* never sw depth */ + swStencil, mesaVis->accumRedBits > 0, + GL_FALSE, /* never sw alpha */ + GL_FALSE /* never sw aux */ ); + driDrawPriv->driverPrivate = (void *) intel_fb; + + return GL_TRUE; + } +} + +static void +intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) +{ + _mesa_unreference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate))); +} + + +/** + * Get information about previous buffer swaps. + */ +static int +intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) +{ + struct intel_framebuffer *intel_fb; + + if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) + || (sInfo == NULL)) { + return -1; + } + + intel_fb = dPriv->driverPrivate; + sInfo->swap_count = intel_fb->swap_count; + sInfo->swap_ust = intel_fb->swap_ust; + sInfo->swap_missed_count = intel_fb->swap_missed_count; + + sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0) + ? driCalculateSwapUsage(dPriv, 0, intel_fb->swap_missed_ust) + : 0.0; + + return 0; +} + + +static void +intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, + unsigned long long offset, GLint depth, GLuint pitch) +{ + abort(); +#if 0 + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct st_texture_object *stObj = st_texture_object(tObj); + + if (!stObj) + return; + + if (stObj->mt) + st_miptree_release(intel->pipe, &stObj->mt); + + stObj->imageOverride = GL_TRUE; + stObj->depthOverride = depth; + stObj->pitchOverride = pitch; + + if (offset) + stObj->textureOffset = offset; +#endif +} + + +static const struct __DriverAPIRec intelAPI = { + .InitDriver = intelInitDriver, + .DestroyScreen = intelDestroyScreen, + .CreateContext = intelCreateContext, + .DestroyContext = intelDestroyContext, + .CreateBuffer = intelCreateBuffer, + .DestroyBuffer = intelDestroyBuffer, + .SwapBuffers = intelSwapBuffers, + .MakeCurrent = intelMakeCurrent, + .UnbindContext = intelUnbindContext, + .GetSwapInfo = intelGetSwapInfo, + .GetMSC = driGetMSC32, + .WaitForMSC = driWaitForMSC32, + .WaitForSBC = NULL, + .SwapBuffersMSC = NULL, + .CopySubBuffer = intelCopySubBuffer, + .setTexOffset = intelSetTexOffset, +}; + + +static __GLcontextModes * +intelFillInModes(unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, GLboolean have_back_buffer) +{ + __GLcontextModes *modes; + __GLcontextModes *m; + unsigned num_modes; + unsigned depth_buffer_factor; + unsigned back_buffer_factor; + GLenum fb_format; + GLenum fb_type; + + /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't + * support pageflipping at all. + */ + static const GLenum back_buffer_modes[] = { + GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML + }; + + u_int8_t depth_bits_array[3]; + u_int8_t stencil_bits_array[3]; + + + depth_bits_array[0] = 0; + depth_bits_array[1] = depth_bits; + depth_bits_array[2] = depth_bits; + + /* Just like with the accumulation buffer, always provide some modes + * with a stencil buffer. It will be a sw fallback, but some apps won't + * care about that. + */ + stencil_bits_array[0] = 0; + stencil_bits_array[1] = 0; + if (depth_bits == 24) + stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; + + stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; + + depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; + back_buffer_factor = (have_back_buffer) ? 3 : 1; + + num_modes = depth_buffer_factor * back_buffer_factor * 4; + + if (pixel_bits == 16) { + fb_format = GL_RGB; + fb_type = GL_UNSIGNED_SHORT_5_6_5; + } + else { + fb_format = GL_BGRA; + fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; + } + + modes = + (*dri_interface->createContextModes) (num_modes, + sizeof(__GLcontextModes)); + m = modes; + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_TRUE_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_DIRECT_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + + /* Mark the visual as slow if there are "fake" stencil bits. + */ + for (m = modes; m != NULL; m = m->next) { + if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { + m->visualRating = GLX_SLOW_CONFIG; + } + } + + return modes; +} + + +/** + * This is the bootstrap function for the driver. libGL supplies all of the + * requisite information about the system, and the driver initializes itself. + * This routine also fills in the linked list pointed to by \c driver_modes + * with the \c __GLcontextModes that the driver can support for windows or + * pbuffers. + * + * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on + * failure. + */ +PUBLIC void * +__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, + __DRIscreen * psc, + const __GLcontextModes * modes, + const __DRIversion * ddx_version, + const __DRIversion * dri_version, + const __DRIversion * drm_version, + const __DRIframebuffer * frame_buffer, + drmAddress pSAREA, int fd, + int internal_api_version, + const __DRIinterfaceMethods * interface, + __GLcontextModes ** driver_modes) +{ + __DRIscreenPrivate *psp; + static const __DRIversion ddx_expected = { 1, 7, 0 }; + static const __DRIversion dri_expected = { 4, 0, 0 }; + static const __DRIversion drm_expected = { 1, 7, 0 }; + + dri_interface = interface; + + if (!driCheckDriDdxDrmVersions2("i915", + dri_version, &dri_expected, + ddx_version, &ddx_expected, + drm_version, &drm_expected)) { + return NULL; + } + + psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, + ddx_version, dri_version, drm_version, + frame_buffer, pSAREA, fd, + internal_api_version, &intelAPI); + + if (psp != NULL) { + I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; + *driver_modes = intelFillInModes(dri_priv->cpp * 8, + (dri_priv->cpp == 2) ? 16 : 24, + (dri_priv->cpp == 2) ? 0 : 8, 1); + + /* Calling driInitExtensions here, with a NULL context pointer, does not actually + * enable the extensions. It just makes sure that all the dispatch offsets for all + * the extensions that *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create is called, but we can't + * enable the extensions until we have a context pointer. + * + * Hello chicken. Hello egg. How are you two today? + */ + driInitExtensions(NULL, card_extensions, GL_FALSE); + } + + return (void *) psp; +} + +struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) +{ + /* + * This should probably change to have the screen allocate a dummy + * context at screen creation. For now just use the current context. + */ + + GET_CURRENT_CONTEXT(ctx); + if (ctx == NULL) { +/* _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; */ + /* need a context for the first time makecurrent is called (for hw lock + when allocating priv buffers) */ + if (intelScreen->dummyctxptr == NULL) { + _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; + } + return intelScreen->dummyctxptr; + } + return intel_context(ctx); + +} + diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h new file mode 100644 index 0000000000..549587a6f2 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -0,0 +1,111 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _INTEL_SCREEN_H_ +#define _INTEL_SCREEN_H_ + +#include "dri_util.h" +#include "i830_common.h" +#include "xmlconfig.h" +#include "dri_bufpool.h" + + +struct intel_screen +{ + struct { + drm_handle_t handle; + + /* We create a static dri buffer for the frontbuffer. + */ + struct _DriBufferObject *buffer; + + char *map; /* memory map */ + int offset; /* from start of video mem, in bytes */ + int pitch; /* row stride, in bytes */ + int width; + int height; + int size; + int cpp; /* for front and back buffers */ + } front; + + int deviceID; + int drmMinor; + + __DRIscreenPrivate *driScrnPriv; + drmI830Sarea *sarea; + + + /** + * Configuration cache with default values for all contexts + */ + driOptionCache optionCache; + struct _DriBufferPool *batchPool; + struct _DriBufferPool *texPool; + struct _DriBufferPool *regionPool; + struct _DriBufferPool *staticPool; + GLboolean havePools; + + struct intel_context *dummyctxptr; +}; + +typedef struct intel_screen intelScreenPrivate; + +extern void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); + + +extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); + +extern GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); + +extern GLboolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); + +extern void +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h); + +extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, + unsigned long bufSize, + unsigned numBufs, + unsigned checkDelayed); + +extern struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen); + +extern GLboolean +intelCreatePools(intelScreenPrivate *intelScreen); + +extern GLboolean +intelCreateContext(const __GLcontextModes * mesaVis, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); + + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_softpipe.c new file mode 100644 index 0000000000..ef47744358 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_softpipe.c @@ -0,0 +1,184 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +#include "dri_bufpool.h" +#include "dri_bufmgr.h" + +/* #include "errno.h" */ +/* #include "string.h" */ +/* #include "imports.h" */ + +#include "intel_context.h" + +#include "pipe/softpipe/sp_winsys.h" + + +struct intel_softpipe_winsys { + struct softpipe_winsys sws; + struct intel_context *intel; +}; + + + +/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static inline struct _DriBufferObject * +dri_bo( struct pipe_buffer_handle *bo ) +{ + return (struct _DriBufferObject *)bo; +} + +static inline struct pipe_buffer_handle * +pipe_bo( struct _DriBufferObject *bo ) +{ + return (struct pipe_buffer_handle *)bo; +} + +/* Turn a softpipe winsys into an intel/softpipe winsys: + */ +static inline struct intel_softpipe_winsys * +intel_softpipe_winsys( struct softpipe_winsys *sws ) +{ + return (struct intel_softpipe_winsys *)sws; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void *intel_buffer_map(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf ) +{ + return driBOMap( dri_bo(buf), + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0 ); +} + +static void intel_buffer_unmap(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf) +{ + driBOUnmap( dri_bo(buf) ); +} + + +static struct pipe_buffer_handle * +intel_buffer_reference(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf) +{ + return pipe_bo( driBOReference( dri_bo(buf) ) ); +} + +static void intel_buffer_unreference(struct softpipe_winsys *sws, + struct pipe_buffer_handle **buf) +{ + if (*buf) { + driBOUnReference( dri_bo(*buf) ); + *buf = NULL; + } +} + +/* Grabs the hardware lock! + */ +static void intel_buffer_data(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct intel_context *intel = intel_softpipe_winsys(sws)->intel; + + LOCK_HARDWARE( intel ); + driBOData( dri_bo(buf), size, data, 0 ); + UNLOCK_HARDWARE( intel ); +} + +static void intel_buffer_subdata(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) +{ + driBOSubData( dri_bo(buf), offset, size, data ); +} + +static void intel_buffer_get_subdata(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) +{ + driBOGetSubData( dri_bo(buf), offset, size, data ); +} + +/* Softpipe has no concept of pools. We choose the tex/region pool + * for all buffers. + */ +static struct pipe_buffer_handle * +intel_create_buffer(struct softpipe_winsys *sws, + unsigned alignment) +{ + struct intel_context *intel = intel_softpipe_winsys(sws)->intel; + struct _DriBufferObject *buffer; + + LOCK_HARDWARE( intel ); + driGenBuffers( intel->intelScreen->regionPool, + "softpipe buffer", 1, &buffer, alignment, 0, 0 ); + UNLOCK_HARDWARE( intel ); + + return pipe_bo(buffer); +} + + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel ) +{ + struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); + + /* Fill in this struct with callbacks that softpipe will need to + * communicate with the window system, buffer manager, etc. + * + * Softpipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + isws->sws.create_buffer = intel_create_buffer; + isws->sws.buffer_map = intel_buffer_map; + isws->sws.buffer_unmap = intel_buffer_unmap; + isws->sws.buffer_reference = intel_buffer_reference; + isws->sws.buffer_unreference = intel_buffer_unreference; + isws->sws.buffer_data = intel_buffer_data; + isws->sws.buffer_subdata = intel_buffer_subdata; + isws->sws.buffer_get_subdata = intel_buffer_get_subdata; + isws->intel = intel; + + /* Create the softpipe context: + */ + return softpipe_create( &isws->sws ); +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.c b/src/mesa/drivers/dri/intel_winsys/intel_surface.c new file mode 100644 index 0000000000..3fa40271c8 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.c @@ -0,0 +1,196 @@ +#include "glheader.h" +#include "context.h" +#include "framebuffer.h" +#include "renderbuffer.h" +#include "utils.h" +#include "main/macros.h" + + +#include "intel_screen.h" + +#include "intel_context.h" +#include "intel_buffers.h" +#include "intel_fbo.h" + +#include "pipe/p_state.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/softpipe/sp_surface.h" + + +/* + * XXX a lof of this is a temporary kludge + */ + +/** + * Note: the arithmetic/addressing in these functions is a little + * tricky since we need to invert the Y axis. + */ + + +static void +read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) +{ + const GLint bytesPerRow = sps->surface.region->pitch * sps->surface.region->cpp; + const GLint invY = sps->surface.height - y - 1; + const GLubyte *src = sps->surface.region->map + invY * bytesPerRow + x * sps->surface.region->cpp; + GLfloat *dst = (GLfloat *) rrrr; + GLubyte temp[16]; + GLuint j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); + + memcpy(temp + 8, src, 8); + memcpy(temp + 0, src + bytesPerRow, 8); + + for (j = 0; j < 4; j++) { + dst[0 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 2]); /*R*/ + dst[1 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 1]); /*G*/ + dst[2 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 0]); /*B*/ + dst[3 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 3]); /*A*/ + } +} + + +static void +write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) +{ + const GLfloat *src = (const GLfloat *) rrrr; + const GLint bytesPerRow = sps->surface.region->pitch * sps->surface.region->cpp; + const GLint invY = sps->surface.height - y - 1; + GLubyte *dst = sps->surface.region->map + invY * bytesPerRow + x * sps->surface.region->cpp; + GLubyte temp[16]; + GLuint j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); + + for (j = 0; j < 4; j++) { + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 2], src[0 * 4 + j]); /*R*/ + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 1], src[1 * 4 + j]); /*G*/ + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 0], src[2 * 4 + j]); /*B*/ + UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 3], src[3 * 4 + j]); /*A*/ + } + + memcpy(dst, temp + 8, 8); + memcpy(dst + bytesPerRow, temp + 0, 8); +} + + + +static void +read_quad_z24(struct softpipe_surface *sps, + GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) +{ + static const GLuint mask = 0xffffff; + const GLint invY = sps->surface.height - y - 1; + const GLuint *src + = (GLuint *) (sps->surface.region->map + + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); + + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); + + /* extract lower three bytes */ + zzzz[0] = src[0] & mask; + zzzz[1] = src[1] & mask; + zzzz[2] = src[-sps->surface.region->pitch] & mask; + zzzz[3] = src[-sps->surface.region->pitch + 1] & mask; +} + +static void +write_quad_z24(struct softpipe_surface *sps, + GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) +{ + static const GLuint mask = 0xff000000; + const GLint invY = sps->surface.height - y - 1; + GLuint *dst + = (GLuint *) (sps->surface.region->map + + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); + + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); + + /* write lower three bytes */ + dst[0] = (dst[0] & mask) | zzzz[0]; + dst[1] = (dst[1] & mask) | zzzz[1]; + dst -= sps->surface.region->pitch; + dst[0] = (dst[0] & mask) | zzzz[2]; + dst[1] = (dst[1] & mask) | zzzz[3]; +} + + +static void +read_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) +{ + const GLint invY = sps->surface.height - y - 1; + const GLuint *src = (const GLuint *) (sps->surface.region->map + + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); + + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); + + /* extract high byte */ + ssss[0] = src[0] >> 24; + ssss[1] = src[1] >> 24; + src -= sps->surface.region->pitch; + ssss[2] = src[0] >> 24; + ssss[3] = src[1] >> 24; +} + +static void +write_quad_stencil(struct softpipe_surface *sps, + GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) +{ + static const GLuint mask = 0x00ffffff; + const GLint invY = sps->surface.height - y - 1; + GLuint *dst = (GLuint *) (sps->surface.region->map + + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); + + assert(sps->surface.format == PIPE_FORMAT_S8_Z24); + + /* write high byte */ + dst[0] = (dst[0] & mask) | (ssss[0] << 24); + dst[1] = (dst[1] & mask) | (ssss[1] << 24); + dst -= sps->surface.region->pitch; + dst[0] = (dst[0] & mask) | (ssss[2] << 24); + dst[1] = (dst[1] & mask) | (ssss[3] << 24); +} + + +struct pipe_surface * +intel_new_surface(GLuint intFormat) +{ + struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); + if (!sps) + return NULL; + + sps->surface.width = 0; /* set in intel_alloc_renderbuffer_storage() */ + sps->surface.height = 0; + + if (intFormat == GL_RGBA8) { + sps->surface.format = PIPE_FORMAT_U_A8_R8_G8_B8; + sps->read_quad_f_swz = read_quad_f_swz; + sps->write_quad_f_swz = write_quad_f_swz; + } + else if (intFormat == GL_RGB5) { + sps->surface.format = PIPE_FORMAT_U_R5_G6_B5; + + } + else if (intFormat == GL_DEPTH_COMPONENT16) { + sps->surface.format = PIPE_FORMAT_U_Z16; + + } + else if (intFormat == GL_DEPTH24_STENCIL8_EXT) { + sps->surface.format = PIPE_FORMAT_S8_Z24; + sps->read_quad_z = read_quad_z24; + sps->write_quad_z = write_quad_z24; + sps->read_quad_stencil = read_quad_stencil; + sps->write_quad_stencil = write_quad_stencil; + } + else { + /* TBD / unknown */ + + } + + return &sps->surface; +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex.c b/src/mesa/drivers/dri/intel_winsys/intel_tex.c new file mode 100644 index 0000000000..5270094cc5 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex.c @@ -0,0 +1,184 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "main/texobj.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_mipmap_tree.h" +#include "intel_tex.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +static GLboolean +intelIsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) +{ +#if 0 + struct intel_context *intel = intel_context(ctx); + struct st_texture_object *stObj = st_texture_object(texObj); + + return + stObj->mt && + stObj->mt->region && + intel_is_region_resident(intel, stObj->mt->region); +#endif + return 1; +} + + + +static struct gl_texture_image * +intelNewTextureImage(GLcontext * ctx) +{ + DBG("%s\n", __FUNCTION__); + (void) ctx; + return (struct gl_texture_image *) CALLOC_STRUCT(st_texture_image); +} + + +static struct gl_texture_object * +intelNewTextureObject(GLcontext * ctx, GLuint name, GLenum target) +{ + struct st_texture_object *obj = CALLOC_STRUCT(st_texture_object); + + DBG("%s\n", __FUNCTION__); + _mesa_initialize_texture_object(&obj->base, name, target); + + return &obj->base; +} + +static void +intelDeleteTextureObject(GLcontext *ctx, + struct gl_texture_object *texObj) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_texture_object *stObj = st_texture_object(texObj); + + if (stObj->mt) + st_miptree_release(pipe, &stObj->mt); + + _mesa_delete_texture_object(ctx, texObj); +} + + +static void +intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_texture_image *stImage = st_texture_image(texImage); + + DBG("%s\n", __FUNCTION__); + + if (stImage->mt) { + st_miptree_release(pipe, &stImage->mt); + } + + if (texImage->Data) { + free(texImage->Data); + texImage->Data = NULL; + } +} + + + + +/* ================================================================ + * From linux kernel i386 header files, copes with odd sizes better + * than COPY_DWORDS would: + * XXX Put this in src/mesa/main/imports.h ??? + */ +#if defined(i386) || defined(__i386__) +static INLINE void * +__memcpy(void *to, const void *from, size_t n) +{ + int d0, d1, d2; + __asm__ __volatile__("rep ; movsl\n\t" + "testb $2,%b4\n\t" + "je 1f\n\t" + "movsw\n" + "1:\ttestb $1,%b4\n\t" + "je 2f\n\t" + "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2) + :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from) + :"memory"); + return (to); +} +#else +#define __memcpy(a,b,c) memcpy(a,b,c) +#endif + + +/* The system memcpy (at least on ubuntu 5.10) has problems copying + * to agp (writecombined) memory from a source which isn't 64-byte + * aligned - there is a 4x performance falloff. + * + * The x86 __memcpy is immune to this but is slightly slower + * (10%-ish) than the system memcpy. + * + * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but + * isn't much faster than x86_memcpy for agp copies. + * + * TODO: switch dynamically. + */ +static void * +do_memcpy(void *dest, const void *src, size_t n) +{ + if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) { + return __memcpy(dest, src, n); + } + else + return memcpy(dest, src, n); +} + + +void +intelInitTextureFuncs(struct dd_function_table *functions) +{ + functions->ChooseTextureFormat = intelChooseTextureFormat; + functions->TexImage1D = intelTexImage1D; + functions->TexImage2D = intelTexImage2D; + functions->TexImage3D = intelTexImage3D; + functions->TexSubImage1D = intelTexSubImage1D; + functions->TexSubImage2D = intelTexSubImage2D; + functions->TexSubImage3D = intelTexSubImage3D; + functions->CopyTexImage1D = intelCopyTexImage1D; + functions->CopyTexImage2D = intelCopyTexImage2D; + functions->CopyTexSubImage1D = intelCopyTexSubImage1D; + functions->CopyTexSubImage2D = intelCopyTexSubImage2D; + functions->GetTexImage = intelGetTexImage; + + /* compressed texture functions */ + functions->CompressedTexImage2D = intelCompressedTexImage2D; + functions->GetCompressedTexImage = intelGetCompressedTexImage; + + functions->NewTextureObject = intelNewTextureObject; + functions->NewTextureImage = intelNewTextureImage; + functions->DeleteTexture = intelDeleteTextureObject; + functions->FreeTexImageData = intelFreeTextureImageData; + functions->UpdateTexturePalette = 0; + functions->IsTextureResident = intelIsTextureResident; + + functions->TextureMemCpy = do_memcpy; +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex.h b/src/mesa/drivers/dri/intel_winsys/intel_tex.h new file mode 100644 index 0000000000..7a1cc91974 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex.h @@ -0,0 +1,212 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTELTEX_INC +#define INTELTEX_INC + +#include "mtypes.h" +#include "intel_context.h" +#include "texmem.h" + +struct st_texture_object +{ + struct gl_texture_object base; /* The "parent" object */ + + /* The mipmap tree must include at least these levels once + * validated: + */ + GLuint firstLevel; + GLuint lastLevel; + + /* Offset for firstLevel image: + */ + GLuint textureOffset; + + /* On validation any active images held in main memory or in other + * regions will be copied to this region and the old storage freed. + */ + struct pipe_mipmap_tree *mt; + + GLboolean imageOverride; + GLint depthOverride; + GLuint pitchOverride; +}; + + + +struct st_texture_image +{ + struct gl_texture_image base; + + /* These aren't stored in gl_texture_image + */ + GLuint level; + GLuint face; + + /* If stImage->mt != NULL, image data is stored here. + * Else if stImage->base.Data != NULL, image is stored there. + * Else there is no image data. + */ + struct pipe_mipmap_tree *mt; +}; + + +void intelInitTextureFuncs(struct dd_function_table *functions); + +const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx, + GLint internalFormat, + GLenum format, + GLenum type); + + +void intelTexImage3D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexSubImage3D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexImage2D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexSubImage2D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexImage1D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelTexSubImage1D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLint border); + +void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border); + +void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint x, GLint y, GLsizei width); + +void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, GLsizei width, GLsizei height); + +void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage); + +void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid *data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage ); + +void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, + GLvoid *pixels, + const struct gl_texture_object *texObj, + const struct gl_texture_image *texImage); + +void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, + unsigned long long offset, GLint depth, GLuint pitch); + +GLuint intel_finalize_mipmap_tree(GLcontext *ctx, + struct pipe_context *pipe, GLuint unit, + GLboolean *needFlush); + + +#if 0 +void intel_tex_map_images(struct pipe_context *pipe, + struct st_texture_object *stObj); + +void intel_tex_unmap_images(struct pipe_context *pipe, + struct st_texture_object *stObj); +#endif + +int intel_compressed_num_bytes(GLuint mesaFormat); + + +static INLINE struct st_texture_object * +st_texture_object(struct gl_texture_object *obj) +{ + return (struct st_texture_object *) obj; +} + +static INLINE struct st_texture_image * +st_texture_image(struct gl_texture_image *img) +{ + return (struct st_texture_image *) img; +} + + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c new file mode 100644 index 0000000000..d2cf6f4669 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c @@ -0,0 +1,315 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "main/mtypes.h" +#include "main/enums.h" +#include "main/image.h" +#include "main/teximage.h" + +#include "state_tracker/st_mipmap_tree.h" + +#if 0 +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_buffers.h" +#include "intel_fbo.h" +#include "intel_blit.h" +#endif +#include "intel_tex.h" + + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +/** + * Get the pipe_region which is the source for any glCopyTex[Sub]Image call. + * + * Do the best we can using the blitter. A future project is to use + * the texture engine and fragment programs for these copies. + */ +static const struct pipe_region * +get_teximage_source(struct intel_context *intel, GLenum internalFormat) +{ +#if 00 + struct intel_renderbuffer *irb; + + DBG("%s %s\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(internalFormat)); + + switch (internalFormat) { + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16_ARB: + irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); + if (irb && irb->region && irb->region->cpp == 2) + return irb->region; + return NULL; + case GL_DEPTH24_STENCIL8_EXT: + case GL_DEPTH_STENCIL_EXT: + irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); + if (irb && irb->region && irb->region->cpp == 4) + return irb->region; + return NULL; + case GL_RGBA: + case GL_RGBA8: + return intel_readbuf_region(intel); + case GL_RGB: + if (intel->intelScreen->front.cpp == 2) + return intel_readbuf_region(intel); + return NULL; + default: + return NULL; + } +#else + return NULL; +#endif +} + + +static GLboolean +do_copy_texsubimage(GLcontext *ctx, + struct st_texture_image *intelImage, + GLenum internalFormat, + GLint dstx, GLint dsty, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + struct intel_context *intel = intel_context(ctx); + const struct pipe_region *src = + get_teximage_source(intel, internalFormat); + + if (!intelImage->mt || !src) { + DBG("%s fail %p %p\n", __FUNCTION__, intelImage->mt, src); + return GL_FALSE; + } + +#if 00 /* XXX FIX flush/locking */ + intelFlush(ctx); + /* XXX still need the lock ? */ + LOCK_HARDWARE(intel); +#endif + + { + GLuint image_offset = st_miptree_image_offset(intelImage->mt, + intelImage->face, + intelImage->level); + const GLint orig_x = x; + const GLint orig_y = y; + const struct gl_framebuffer *fb = ctx->DrawBuffer; + + if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, + &x, &y, &width, &height)) { + /* Update dst for clipped src. Need to also clip the source rect. + */ + dstx += x - orig_x; + dsty += y - orig_y; + + if (!(ctx->ReadBuffer->Name == 0)) { + /* XXX this looks bogus ? */ + /* FBO: invert Y */ + y = ctx->ReadBuffer->Height - y - 1; + } + + /* A bit of fiddling to get the blitter to work with -ve + * pitches. But we get a nice inverted blit this way, so it's + * worth it: + */ +#if 0 + intelEmitCopyBlit(intel, + intelImage->mt->cpp, + -src->pitch, + src->buffer, + src->height * src->pitch * src->cpp, + intelImage->mt->pitch, + intelImage->mt->region->buffer, + image_offset, + x, y + height, dstx, dsty, width, height, + GL_COPY); /* ? */ + intel_batchbuffer_flush(intel->batch); +#else + /* XXX use pipe->region_copy() ??? */ + (void) image_offset; +#endif + } + } + +#if 0 + UNLOCK_HARDWARE(intel); +#endif + +#if 0 + /* GL_SGIS_generate_mipmap -- this can be accelerated now. + * XXX Add a ctx->Driver.GenerateMipmaps() function? + */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif + + return GL_TRUE; +} + + + + + +void +intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLint border) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage1D(ctx, target, level, internalFormat, + width, border, + GL_RGBA, CHAN_TYPE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + if (!do_copy_texsubimage(ctx, + st_texture_image(texImage), + internalFormat, 0, 0, x, y, width, 1)) + goto fail; + + return; + + fail: +#if 0 + _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, + width, border); +#endif + ; +} + +void +intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, + GLenum internalFormat, + GLint x, GLint y, GLsizei width, GLsizei height, + GLint border) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + + if (border) + goto fail; + + /* Setup or redefine the texture object, mipmap tree and texture + * image. Don't populate yet. + */ + ctx->Driver.TexImage2D(ctx, target, level, internalFormat, + width, height, border, + GL_RGBA, CHAN_TYPE, NULL, + &ctx->DefaultPacking, texObj, texImage); + + + if (!do_copy_texsubimage(ctx, + st_texture_image(texImage), + internalFormat, 0, 0, x, y, width, height)) + goto fail; + + return; + + fail: +#if 0 + _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, + width, height, border); +#endif + assert(0); +} + + +void +intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint x, GLint y, GLsizei width) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + GLenum internalFormat = texImage->InternalFormat; + + /* XXX need to check as in above function? */ + + /* Need to check texture is compatible with source format. + */ + + if (!do_copy_texsubimage(ctx, + st_texture_image(texImage), + internalFormat, xoffset, 0, x, y, width, 1)) { +#if 0 + _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); +#endif + assert(0); + } +} + + + +void +intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + GLenum internalFormat = texImage->InternalFormat; + + + /* Need to check texture is compatible with source format. + */ + + if (!do_copy_texsubimage(ctx, + st_texture_image(texImage), + internalFormat, + xoffset, yoffset, x, y, width, height)) { +#if 0 + _swrast_copy_texsubimage2d(ctx, target, level, + xoffset, yoffset, x, y, width, height); +#endif + assert(0); + } +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_format.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_format.c new file mode 100644 index 0000000000..a11718bb07 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex_format.c @@ -0,0 +1,172 @@ +#include "intel_context.h" +#include "intel_tex.h" +#include "texformat.h" +#include "enums.h" + +/* It works out that this function is fine for all the supported + * hardware. However, there is still a need to map the formats onto + * hardware descriptors. + */ +/* Note that the i915 can actually support many more formats than + * these if we take the step of simply swizzling the colors + * immediately after sampling... + */ +const struct gl_texture_format * +intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, + GLenum format, GLenum type) +{ + struct intel_context *intel = intel_context(ctx); + const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); + + switch (internalFormat) { + case 4: + case GL_RGBA: + case GL_COMPRESSED_RGBA: + if (format == GL_BGRA) { + if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { + return &_mesa_texformat_argb8888; + } + else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { + return &_mesa_texformat_argb4444; + } + else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { + return &_mesa_texformat_argb1555; + } + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case 3: + case GL_RGB: + case GL_COMPRESSED_RGB: + if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { + return &_mesa_texformat_rgb565; + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case GL_RGBA4: + case GL_RGBA2: + return &_mesa_texformat_argb4444; + + case GL_RGB5_A1: + return &_mesa_texformat_argb1555; + + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + return &_mesa_texformat_argb8888; + + case GL_RGB5: + case GL_RGB4: + case GL_R3_G3_B2: + return &_mesa_texformat_rgb565; + + case GL_ALPHA: + case GL_ALPHA4: + case GL_ALPHA8: + case GL_ALPHA12: + case GL_ALPHA16: + case GL_COMPRESSED_ALPHA: + return &_mesa_texformat_a8; + + case 1: + case GL_LUMINANCE: + case GL_LUMINANCE4: + case GL_LUMINANCE8: + case GL_LUMINANCE12: + case GL_LUMINANCE16: + case GL_COMPRESSED_LUMINANCE: + return &_mesa_texformat_l8; + + case 2: + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE4_ALPHA4: + case GL_LUMINANCE6_ALPHA2: + case GL_LUMINANCE8_ALPHA8: + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: + case GL_COMPRESSED_LUMINANCE_ALPHA: + return &_mesa_texformat_al88; + + case GL_INTENSITY: + case GL_INTENSITY4: + case GL_INTENSITY8: + case GL_INTENSITY12: + case GL_INTENSITY16: + case GL_COMPRESSED_INTENSITY: + return &_mesa_texformat_i8; + + case GL_YCBCR_MESA: + if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) + return &_mesa_texformat_ycbcr; + else + return &_mesa_texformat_ycbcr_rev; + + case GL_COMPRESSED_RGB_FXT1_3DFX: + return &_mesa_texformat_rgb_fxt1; + case GL_COMPRESSED_RGBA_FXT1_3DFX: + return &_mesa_texformat_rgba_fxt1; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + return &_mesa_texformat_rgb_dxt1; + + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return &_mesa_texformat_rgba_dxt1; + + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + return &_mesa_texformat_rgba_dxt3; + + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + return &_mesa_texformat_rgba_dxt5; + + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + return &_mesa_texformat_z16; + + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + return &_mesa_texformat_z24_s8; + + default: + fprintf(stderr, "unexpected texture format %s in %s\n", + _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); + return NULL; + } + + return NULL; /* never get here */ +} + +int intel_compressed_num_bytes(GLuint mesaFormat) +{ + int bytes = 0; + switch(mesaFormat) { + + case MESA_FORMAT_RGB_FXT1: + case MESA_FORMAT_RGBA_FXT1: + case MESA_FORMAT_RGB_DXT1: + case MESA_FORMAT_RGBA_DXT1: + bytes = 2; + break; + + case MESA_FORMAT_RGBA_DXT3: + case MESA_FORMAT_RGBA_DXT5: + bytes = 4; + default: + break; + } + + return bytes; +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_image.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_image.c new file mode 100644 index 0000000000..93a59b6896 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex_image.c @@ -0,0 +1,695 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "glheader.h" +#include "macros.h" +#include "mtypes.h" +#include "enums.h" +#include "colortab.h" +#include "convolve.h" +#include "context.h" +#include "simple_list.h" +#include "texcompress.h" +#include "texformat.h" +#include "texobj.h" +#include "texstore.h" + +#include "pipe/p_context.h" +#include "state_tracker/st_mipmap_tree.h" + +#include "intel_tex.h" + + + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +/* Functions to store texture images. Where possible, mipmap_tree's + * will be created or further instantiated with image data, otherwise + * images will be stored in malloc'd memory. A validation step is + * required to pull those images into a mipmap tree, or otherwise + * decide a fallback is required. + */ + + +static int +logbase2(int n) +{ + GLint i = 1; + GLint log2 = 0; + + while (n > i) { + i *= 2; + log2++; + } + + return log2; +} + + +/* Otherwise, store it in memory if (Border != 0) or (any dimension == + * 1). + * + * Otherwise, if max_level >= level >= min_level, create tree with + * space for textures from min_level down to max_level. + * + * Otherwise, create tree with space for textures from (level + * 0)..(1x1). Consider pruning this tree at a validation if the + * saving is worth it. + */ +static void +guess_and_alloc_mipmap_tree(struct pipe_context *pipe, + struct st_texture_object *intelObj, + struct st_texture_image *intelImage) +{ + GLuint firstLevel; + GLuint lastLevel; + GLuint width = intelImage->base.Width; + GLuint height = intelImage->base.Height; + GLuint depth = intelImage->base.Depth; + GLuint l2width, l2height, l2depth; + GLuint i, comp_byte = 0; + + DBG("%s\n", __FUNCTION__); + + if (intelImage->base.Border) + return; + + if (intelImage->level > intelObj->base.BaseLevel && + (intelImage->base.Width == 1 || + (intelObj->base.Target != GL_TEXTURE_1D && + intelImage->base.Height == 1) || + (intelObj->base.Target == GL_TEXTURE_3D && + intelImage->base.Depth == 1))) + return; + + /* If this image disrespects BaseLevel, allocate from level zero. + * Usually BaseLevel == 0, so it's unlikely to happen. + */ + if (intelImage->level < intelObj->base.BaseLevel) + firstLevel = 0; + else + firstLevel = intelObj->base.BaseLevel; + + + /* Figure out image dimensions at start level. + */ + for (i = intelImage->level; i > firstLevel; i--) { + width <<= 1; + if (height != 1) + height <<= 1; + if (depth != 1) + depth <<= 1; + } + + /* Guess a reasonable value for lastLevel. This is probably going + * to be wrong fairly often and might mean that we have to look at + * resizable buffers, or require that buffers implement lazy + * pagetable arrangements. + */ + if ((intelObj->base.MinFilter == GL_NEAREST || + intelObj->base.MinFilter == GL_LINEAR) && + intelImage->level == firstLevel) { + lastLevel = firstLevel; + } + else { + l2width = logbase2(width); + l2height = logbase2(height); + l2depth = logbase2(depth); + lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); + } + + assert(!intelObj->mt); + if (intelImage->base.IsCompressed) + comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); + intelObj->mt = st_miptree_create(pipe, + intelObj->base.Target, + intelImage->base.InternalFormat, + firstLevel, + lastLevel, + width, + height, + depth, + intelImage->base.TexFormat->TexelBytes, + comp_byte); + + DBG("%s - success\n", __FUNCTION__); +} + + + + +static GLuint +target_to_face(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + return ((GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X); + default: + return 0; + } +} + + + +/* There are actually quite a few combinations this will work for, + * more than what I've listed here. + */ +static GLboolean +check_pbo_format(GLint internalFormat, + GLenum format, GLenum type, + const struct gl_texture_format *mesa_format) +{ + switch (internalFormat) { + case 4: + case GL_RGBA: + return (format == GL_BGRA && + (type == GL_UNSIGNED_BYTE || + type == GL_UNSIGNED_INT_8_8_8_8_REV) && + mesa_format == &_mesa_texformat_argb8888); + case 3: + case GL_RGB: + return (format == GL_RGB && + type == GL_UNSIGNED_SHORT_5_6_5 && + mesa_format == &_mesa_texformat_rgb565); + case GL_YCBCR_MESA: + return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); + default: + return GL_FALSE; + } +} + + +/* XXX: Do this for TexSubImage also: + */ +static GLboolean +try_pbo_upload(struct intel_context *intel, + struct st_texture_image *intelImage, + const struct gl_pixelstore_attrib *unpack, + GLint internalFormat, + GLint width, GLint height, + GLenum format, GLenum type, const void *pixels) +{ + return GL_FALSE; /* XXX fix flushing/locking/blitting below */ +#if 000 + struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); + GLuint src_offset, src_stride; + GLuint dst_offset, dst_stride; + + if (!pbo || + intel->ctx._ImageTransferState || + unpack->SkipPixels || unpack->SkipRows) { + _mesa_printf("%s: failure 1\n", __FUNCTION__); + return GL_FALSE; + } + + src_offset = (GLuint) pixels; + + if (unpack->RowLength > 0) + src_stride = unpack->RowLength; + else + src_stride = width; + + dst_offset = st_miptree_image_offset(intelImage->mt, + intelImage->face, + intelImage->level); + + dst_stride = intelImage->mt->pitch; + + intelFlush(&intel->ctx); + LOCK_HARDWARE(intel); + { + struct _DriBufferObject *src_buffer = + intel_bufferobj_buffer(intel, pbo, INTEL_READ); + + /* Temporary hack: cast to _DriBufferObject: + */ + struct _DriBufferObject *dst_buffer = + (struct _DriBufferObject *)intelImage->mt->region->buffer; + + + intelEmitCopyBlit(intel, + intelImage->mt->cpp, + src_stride, src_buffer, src_offset, + dst_stride, dst_buffer, dst_offset, + 0, 0, 0, 0, width, height, + GL_COPY); + + intel_batchbuffer_flush(intel->batch); + } + UNLOCK_HARDWARE(intel); + + return GL_TRUE; +#endif +} + + + +static GLboolean +try_pbo_zcopy(struct intel_context *intel, + struct st_texture_image *intelImage, + const struct gl_pixelstore_attrib *unpack, + GLint internalFormat, + GLint width, GLint height, + GLenum format, GLenum type, const void *pixels) +{ + return GL_FALSE; +} + + + + + + +static void +intelTexImage(GLcontext * ctx, + GLint dims, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, GLsizei imageSize, int compressed) +{ + struct intel_context *intel = intel_context(ctx); + struct st_texture_object *intelObj = st_texture_object(texObj); + struct st_texture_image *intelImage = st_texture_image(texImage); + GLint postConvWidth = width; + GLint postConvHeight = height; + GLint texelBytes, sizeInBytes; + GLuint dstRowStride; + + + DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); + + intelFlush(ctx); + + intelImage->face = target_to_face(target); + intelImage->level = level; + + if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { + _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth, + &postConvHeight); + } + + /* choose the texture format */ + texImage->TexFormat = intelChooseTextureFormat(ctx, internalFormat, + format, type); + + _mesa_set_fetch_functions(texImage, dims); + + if (texImage->TexFormat->TexelBytes == 0) { + /* must be a compressed format */ + texelBytes = 0; + texImage->IsCompressed = GL_TRUE; + texImage->CompressedSize = + ctx->Driver.CompressedTextureSize(ctx, texImage->Width, + texImage->Height, texImage->Depth, + texImage->TexFormat->MesaFormat); + } else { + texelBytes = texImage->TexFormat->TexelBytes; + + /* Minimum pitch of 32 bytes */ + if (postConvWidth * texelBytes < 32) { + postConvWidth = 32 / texelBytes; + texImage->RowStride = postConvWidth; + } + + assert(texImage->RowStride == postConvWidth); + } + + /* Release the reference to a potentially orphaned buffer. + * Release any old malloced memory. + */ + if (intelImage->mt) { + st_miptree_release(intel->pipe, &intelImage->mt); + assert(!texImage->Data); + } + else if (texImage->Data) { + _mesa_align_free(texImage->Data); + } + + /* If this is the only texture image in the tree, could call + * bmBufferData with NULL data to free the old block and avoid + * waiting on any outstanding fences. + */ + if (intelObj->mt && + intelObj->mt->first_level == level && + intelObj->mt->last_level == level && + intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && + !st_miptree_match_image(intelObj->mt, &intelImage->base, + intelImage->face, intelImage->level)) { + + DBG("release it\n"); + st_miptree_release(intel->pipe, &intelObj->mt); + assert(!intelObj->mt); + } + + if (!intelObj->mt) { + guess_and_alloc_mipmap_tree(intel->pipe, intelObj, intelImage); + if (!intelObj->mt) { + DBG("guess_and_alloc_mipmap_tree: failed\n"); + } + } + + assert(!intelImage->mt); + + if (intelObj->mt && + st_miptree_match_image(intelObj->mt, &intelImage->base, + intelImage->face, intelImage->level)) { + + st_miptree_reference(&intelImage->mt, intelObj->mt); + assert(intelImage->mt); + } + + if (!intelImage->mt) + DBG("XXX: Image did not fit into tree - storing in local memory!\n"); + +#if 0 /* XXX FIX when st_buffer_objects are in place */ + /* PBO fastpaths: + */ + if (dims <= 2 && + intelImage->mt && + intel_buffer_object(unpack->BufferObj) && + check_pbo_format(internalFormat, format, + type, intelImage->base.TexFormat)) { + + DBG("trying pbo upload\n"); + + /* Attempt to texture directly from PBO data (zero copy upload). + * + * Currently disable as it can lead to worse as well as better + * performance (in particular when pipe_region_cow() is + * required). + */ + if (intelObj->mt == intelImage->mt && + intelObj->mt->first_level == level && + intelObj->mt->last_level == level) { + + if (try_pbo_zcopy(intel, intelImage, unpack, + internalFormat, + width, height, format, type, pixels)) { + + DBG("pbo zcopy upload succeeded\n"); + return; + } + } + + + /* Otherwise, attempt to use the blitter for PBO image uploads. + */ + if (try_pbo_upload(intel, intelImage, unpack, + internalFormat, + width, height, format, type, pixels)) { + DBG("pbo upload succeeded\n"); + return; + } + + DBG("pbo upload failed\n"); + } +#else + (void) try_pbo_upload; + (void) check_pbo_format; + (void) try_pbo_zcopy; +#endif + + + /* intelCopyTexImage calls this function with pixels == NULL, with + * the expectation that the mipmap tree will be set up but nothing + * more will be done. This is where those calls return: + */ + if (compressed) { + pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels, + unpack, + "glCompressedTexImage"); + } else { + pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, + format, type, + pixels, unpack, "glTexImage"); + } + if (!pixels) + return; + + + if (intelImage->mt) + intel->pipe->region_idle(intel->pipe, intelImage->mt->region); + + LOCK_HARDWARE(intel); + + if (intelImage->mt) { + texImage->Data = st_miptree_image_map(intel->pipe, + intelImage->mt, + intelImage->face, + intelImage->level, + &dstRowStride, + intelImage->base.ImageOffsets); + } + else { + /* Allocate regular memory and store the image there temporarily. */ + if (texImage->IsCompressed) { + sizeInBytes = texImage->CompressedSize; + dstRowStride = + _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); + assert(dims != 3); + } + else { + dstRowStride = postConvWidth * texelBytes; + sizeInBytes = depth * dstRowStride * postConvHeight; + } + + texImage->Data = malloc(sizeInBytes); + } + + DBG("Upload image %dx%dx%d row_len %x " + "pitch %x\n", + width, height, depth, width * texelBytes, dstRowStride); + + /* Copy data. Would like to know when it's ok for us to eg. use + * the blitter to copy. Or, use the hardware to do the format + * conversion and copy: + */ + if (compressed) { + memcpy(texImage->Data, pixels, imageSize); + } else if (!texImage->TexFormat->StoreImage(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, unpack)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + } + + _mesa_unmap_teximage_pbo(ctx, unpack); + + if (intelImage->mt) { + st_miptree_image_unmap(intel->pipe, intelImage->mt); + texImage->Data = NULL; + } + + UNLOCK_HARDWARE(intel); + +#if 0 + /* GL_SGIS_generate_mipmap -- this can be accelerated now. + */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + intel_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif +} + +void +intelTexImage3D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexImage(ctx, 3, target, level, + internalFormat, width, height, depth, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + + +void +intelTexImage2D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexImage(ctx, 2, target, level, + internalFormat, width, height, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + +void +intelTexImage1D(GLcontext * ctx, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexImage(ctx, 1, target, level, + internalFormat, width, 1, 1, border, + format, type, pixels, unpack, texObj, texImage, 0, 0); +} + +void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid *data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage ) +{ + intelTexImage(ctx, 2, target, level, + internalFormat, width, height, 1, border, + 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1); +} + +/** + * Need to map texture image into memory before copying image data, + * then unmap it. + */ +static void +intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, int compressed) +{ + struct intel_context *intel = intel_context(ctx); + struct st_texture_image *intelImage = st_texture_image(texImage); + + /* Map */ + if (intelImage->mt) { + /* Image is stored in hardware format in a buffer managed by the + * kernel. Need to explicitly map and unmap it. + */ + intelImage->base.Data = + st_miptree_image_map(intel->pipe, + intelImage->mt, + intelImage->face, + intelImage->level, + &intelImage->base.RowStride, + intelImage->base.ImageOffsets); + intelImage->base.RowStride /= intelImage->mt->cpp; + } + else { + /* Otherwise, the image should actually be stored in + * intelImage->base.Data. This is pretty confusing for + * everybody, I'd much prefer to separate the two functions of + * texImage->Data - storage for texture images in main memory + * and access (ie mappings) of images. In other words, we'd + * create a new texImage->Map field and leave Data simply for + * storage. + */ + assert(intelImage->base.Data); + } + + + if (compressed) { + _mesa_get_compressed_teximage(ctx, target, level, pixels, + texObj, texImage); + } else { + _mesa_get_teximage(ctx, target, level, format, type, pixels, + texObj, texImage); + } + + + /* Unmap */ + if (intelImage->mt) { + st_miptree_image_unmap(intel->pipe, intelImage->mt); + intelImage->base.Data = NULL; + } +} + +void +intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intel_get_tex_image(ctx, target, level, format, type, pixels, + texObj, texImage, 0); + + +} + +void +intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, + GLvoid *pixels, + const struct gl_texture_object *texObj, + const struct gl_texture_image *texImage) +{ + intel_get_tex_image(ctx, target, level, 0, 0, pixels, + texObj, texImage, 1); + +} + +void +intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, + unsigned long long offset, GLint depth, GLuint pitch) +{ + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct st_texture_object *stObj = st_texture_object(tObj); + + if (!stObj) + return; + + if (stObj->mt) + st_miptree_release(intel->pipe, &stObj->mt); + + stObj->imageOverride = GL_TRUE; + stObj->depthOverride = depth; + stObj->pitchOverride = pitch; + + if (offset) + stObj->textureOffset = offset; +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c new file mode 120000 index 0000000000..fe61b44194 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c @@ -0,0 +1 @@ +../intel/intel_tex_layout.c \ No newline at end of file diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_subimage.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_subimage.c new file mode 100644 index 0000000000..0937114c7f --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex_subimage.c @@ -0,0 +1,192 @@ + +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "mtypes.h" +#include "texobj.h" +#include "texstore.h" +#include "enums.h" + +/* +#include "intel_context.h" +*/ +#include "intel_tex.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_mipmap_tree.h" + +#include "pipe/p_context.h" + + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +static void +intelTexSubimage(GLcontext * ctx, + GLint dims, + GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLint width, GLint height, GLint depth, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_texture_image *intelImage = st_texture_image(texImage); + GLuint dstRowStride; + + DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), + level, xoffset, yoffset, width, height); + + intelFlush(ctx); + + pixels = + _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format, + type, pixels, packing, "glTexSubImage2D"); + if (!pixels) + return; + + if (intelImage->mt) + pipe->region_idle(pipe, intelImage->mt->region); + +#if 0 + LOCK_HARDWARE(intel); +#endif + + /* Map buffer if necessary. Need to lock to prevent other contexts + * from uploading the buffer under us. + */ + if (intelImage->mt) + texImage->Data = st_miptree_image_map(pipe, + intelImage->mt, + intelImage->face, + intelImage->level, + &dstRowStride, + texImage->ImageOffsets); + + assert(dstRowStride); + + if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + xoffset, yoffset, zoffset, + dstRowStride, + texImage->ImageOffsets, + width, height, depth, + format, type, pixels, packing)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); + } + +#if 0 + /* GL_SGIS_generate_mipmap */ + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { + _mesa_generate_mipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); + } +#endif + + _mesa_unmap_teximage_pbo(ctx, packing); + + if (intelImage->mt) { + st_miptree_image_unmap(pipe, intelImage->mt); + texImage->Data = NULL; + } + +#if 0 + UNLOCK_HARDWARE(intel); +#endif +} + + + + + +void +intelTexSubImage3D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + + intelTexSubimage(ctx, 3, + target, level, + xoffset, yoffset, zoffset, + width, height, depth, + format, type, pixels, packing, texObj, texImage); + +} + + + +void +intelTexSubImage2D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + + intelTexSubimage(ctx, 2, + target, level, + xoffset, yoffset, 0, + width, height, 1, + format, type, pixels, packing, texObj, texImage); + +} + + +void +intelTexSubImage1D(GLcontext * ctx, + GLenum target, + GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLenum type, + const GLvoid * pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) +{ + intelTexSubimage(ctx, 1, + target, level, + xoffset, 0, 0, + width, 1, 1, + format, type, pixels, packing, texObj, texImage); + +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_validate.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_validate.c new file mode 100644 index 0000000000..d0631a88f1 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex_validate.c @@ -0,0 +1,277 @@ +#include "mtypes.h" +#include "macros.h" + +#include "state_tracker/st_mipmap_tree.h" +#include "intel_tex.h" + +#include "pipe/p_state.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +/** + * Compute which mipmap levels that really need to be sent to the hardware. + * This depends on the base image size, GL_TEXTURE_MIN_LOD, + * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. + */ +static void +intel_calculate_first_last_level(struct st_texture_object *intelObj) +{ + struct gl_texture_object *tObj = &intelObj->base; + const struct gl_texture_image *const baseImage = + tObj->Image[0][tObj->BaseLevel]; + + /* These must be signed values. MinLod and MaxLod can be negative numbers, + * and having firstLevel and lastLevel as signed prevents the need for + * extra sign checks. + */ + int firstLevel; + int lastLevel; + + /* Yes, this looks overly complicated, but it's all needed. + */ + switch (tObj->Target) { + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_3D: + case GL_TEXTURE_CUBE_MAP: + if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) { + /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL. + */ + firstLevel = lastLevel = tObj->BaseLevel; + } + else { + firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); + firstLevel = MAX2(firstLevel, tObj->BaseLevel); + lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); + lastLevel = MAX2(lastLevel, tObj->BaseLevel); + lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); + lastLevel = MIN2(lastLevel, tObj->MaxLevel); + lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ + } + break; + case GL_TEXTURE_RECTANGLE_NV: + case GL_TEXTURE_4D_SGIS: + firstLevel = lastLevel = 0; + break; + default: + return; + } + + /* save these values */ + intelObj->firstLevel = firstLevel; + intelObj->lastLevel = lastLevel; +} + +static void +copy_image_data_to_tree(struct pipe_context *pipe, + struct st_texture_object *intelObj, + struct st_texture_image *intelImage) +{ + if (intelImage->mt) { + /* Copy potentially with the blitter: + */ + st_miptree_image_copy(pipe, + intelObj->mt, + intelImage->face, + intelImage->level, intelImage->mt); + + st_miptree_release(pipe, &intelImage->mt); + } + else { + assert(intelImage->base.Data != NULL); + + /* More straightforward upload. + */ + st_miptree_image_data(pipe, + intelObj->mt, + intelImage->face, + intelImage->level, + intelImage->base.Data, + intelImage->base.RowStride, + intelImage->base.RowStride * + intelImage->base.Height); + _mesa_align_free(intelImage->base.Data); + intelImage->base.Data = NULL; + } + + st_miptree_reference(&intelImage->mt, intelObj->mt); +} + + +/* + */ +GLuint +intel_finalize_mipmap_tree(GLcontext *ctx, + struct pipe_context *pipe, GLuint unit, + GLboolean *needFlush) +{ + struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; + struct st_texture_object *intelObj = st_texture_object(tObj); + int comp_byte = 0; + int cpp; + + GLuint face, i; + GLuint nr_faces = 0; + struct st_texture_image *firstImage; + + *needFlush = GL_FALSE; + + /* We know/require this is true by now: + */ + assert(intelObj->base._Complete); + + /* What levels must the tree include at a minimum? + */ + intel_calculate_first_last_level(intelObj); + firstImage = + st_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); + + /* Fallback case: + */ + if (firstImage->base.Border) { + if (intelObj->mt) { + st_miptree_release(pipe, &intelObj->mt); + } + return GL_FALSE; + } + + + /* If both firstImage and intelObj have a tree which can contain + * all active images, favour firstImage. Note that because of the + * completeness requirement, we know that the image dimensions + * will match. + */ + if (firstImage->mt && + firstImage->mt != intelObj->mt && + firstImage->mt->first_level <= intelObj->firstLevel && + firstImage->mt->last_level >= intelObj->lastLevel) { + + if (intelObj->mt) + st_miptree_release(pipe, &intelObj->mt); + + st_miptree_reference(&intelObj->mt, firstImage->mt); + } + + if (firstImage->base.IsCompressed) { + comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); + cpp = comp_byte; + } + else cpp = firstImage->base.TexFormat->TexelBytes; + + /* Check tree can hold all active levels. Check tree matches + * target, imageFormat, etc. + * + * XXX: For some layouts (eg i945?), the test might have to be + * first_level == firstLevel, as the tree isn't valid except at the + * original start level. Hope to get around this by + * programming minLod, maxLod, baseLevel into the hardware and + * leaving the tree alone. + */ + if (intelObj->mt && + (intelObj->mt->target != intelObj->base.Target || + intelObj->mt->internal_format != firstImage->base.InternalFormat || + intelObj->mt->first_level != intelObj->firstLevel || + intelObj->mt->last_level != intelObj->lastLevel || + intelObj->mt->width0 != firstImage->base.Width || + intelObj->mt->height0 != firstImage->base.Height || + intelObj->mt->depth0 != firstImage->base.Depth || + intelObj->mt->cpp != cpp || + intelObj->mt->compressed != firstImage->base.IsCompressed)) { + st_miptree_release(pipe, &intelObj->mt); + } + + + /* May need to create a new tree: + */ + if (!intelObj->mt) { + intelObj->mt = st_miptree_create(pipe, + intelObj->base.Target, + firstImage->base.InternalFormat, + intelObj->firstLevel, + intelObj->lastLevel, + firstImage->base.Width, + firstImage->base.Height, + firstImage->base.Depth, + cpp, + comp_byte); + } + + /* Pull in any images not in the object's tree: + */ + nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct st_texture_image *intelImage = + st_texture_image(intelObj->base.Image[face][i]); + + /* Need to import images in main memory or held in other trees. + */ + if (intelObj->mt != intelImage->mt) { + copy_image_data_to_tree(pipe, intelObj, intelImage); + *needFlush = GL_TRUE; + } + } + } + + /** + if (need_flush) + intel_batchbuffer_flush(intel->batch); + **/ + + return GL_TRUE; +} + + +#if 0 /* unused? */ +void +intel_tex_map_images(struct pipe_context *pipe, + struct st_texture_object *intelObj) +{ + GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + + DBG("%s\n", __FUNCTION__); + + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct st_texture_image *intelImage = + st_texture_image(intelObj->base.Image[face][i]); + + if (intelImage->mt) { + intelImage->base.Data = + st_miptree_image_map(pipe, + intelImage->mt, + intelImage->face, + intelImage->level, + &intelImage->base.RowStride, + intelImage->base.ImageOffsets); + /* convert stride to texels, not bytes */ + intelImage->base.RowStride /= intelImage->mt->cpp; +/* intelImage->base.ImageStride /= intelImage->mt->cpp; */ + } + } + } +} + + + +void +intel_tex_unmap_images(struct pipe_context *pipe, + struct st_texture_object *intelObj) +{ + GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; + GLuint face, i; + + for (face = 0; face < nr_faces; face++) { + for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { + struct st_texture_image *intelImage = + st_texture_image(intelObj->base.Image[face][i]); + + if (intelImage->mt) { + st_miptree_image_unmap(pipe, intelImage->mt); + intelImage->base.Data = NULL; + } + } + } +} +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/server/i830_common.h b/src/mesa/drivers/dri/intel_winsys/server/i830_common.h new file mode 100644 index 0000000000..d4d58886ce --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/server/i830_common.h @@ -0,0 +1,226 @@ +/************************************************************************** + +Copyright 2001 VA Linux Systems Inc., Fremont, California. +Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas. + +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the "Software"), +to deal in the Software without restriction, including without limitation +on the rights to use, copy, modify, merge, publish, distribute, sub +license, and/or sell copies of the Software, and to permit persons to whom +the Software is furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice (including the next +paragraph) shall be included in all copies or substantial portions of the +Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL +ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, +DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR +OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE +USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ + +#ifndef _I830_COMMON_H_ +#define _I830_COMMON_H_ + + +#define I830_NR_TEX_REGIONS 255 /* maximum due to use of chars for next/prev */ +#define I830_LOG_MIN_TEX_REGION_SIZE 14 + + +/* Driver specific DRM command indices + * NOTE: these are not OS specific, but they are driver specific + */ +#define DRM_I830_INIT 0x00 +#define DRM_I830_FLUSH 0x01 +#define DRM_I830_FLIP 0x02 +#define DRM_I830_BATCHBUFFER 0x03 +#define DRM_I830_IRQ_EMIT 0x04 +#define DRM_I830_IRQ_WAIT 0x05 +#define DRM_I830_GETPARAM 0x06 +#define DRM_I830_SETPARAM 0x07 +#define DRM_I830_ALLOC 0x08 +#define DRM_I830_FREE 0x09 +#define DRM_I830_INIT_HEAP 0x0a +#define DRM_I830_CMDBUFFER 0x0b +#define DRM_I830_DESTROY_HEAP 0x0c +#define DRM_I830_SET_VBLANK_PIPE 0x0d +#define DRM_I830_GET_VBLANK_PIPE 0x0e + +typedef struct { + enum { + I830_INIT_DMA = 0x01, + I830_CLEANUP_DMA = 0x02, + I830_RESUME_DMA = 0x03 + } func; + unsigned int mmio_offset; + int sarea_priv_offset; + unsigned int ring_start; + unsigned int ring_end; + unsigned int ring_size; + unsigned int front_offset; + unsigned int back_offset; + unsigned int depth_offset; + unsigned int w; + unsigned int h; + unsigned int pitch; + unsigned int pitch_bits; + unsigned int back_pitch; + unsigned int depth_pitch; + unsigned int cpp; + unsigned int chipset; +} drmI830Init; + +typedef struct { + drmTextureRegion texList[I830_NR_TEX_REGIONS+1]; + int last_upload; /* last time texture was uploaded */ + int last_enqueue; /* last time a buffer was enqueued */ + int last_dispatch; /* age of the most recently dispatched buffer */ + int ctxOwner; /* last context to upload state */ + int texAge; + int pf_enabled; /* is pageflipping allowed? */ + int pf_active; + int pf_current_page; /* which buffer is being displayed? */ + int perf_boxes; /* performance boxes to be displayed */ + int width, height; /* screen size in pixels */ + + drm_handle_t front_handle; + int front_offset; + int front_size; + + drm_handle_t back_handle; + int back_offset; + int back_size; + + drm_handle_t depth_handle; + int depth_offset; + int depth_size; + + drm_handle_t tex_handle; + int tex_offset; + int tex_size; + int log_tex_granularity; + int pitch; + int rotation; /* 0, 90, 180 or 270 */ + int rotated_offset; + int rotated_size; + int rotated_pitch; + int virtualX, virtualY; + + unsigned int front_tiled; + unsigned int back_tiled; + unsigned int depth_tiled; + unsigned int rotated_tiled; + unsigned int rotated2_tiled; + + int pipeA_x; + int pipeA_y; + int pipeA_w; + int pipeA_h; + int pipeB_x; + int pipeB_y; + int pipeB_w; + int pipeB_h; + + /* Triple buffering */ + drm_handle_t third_handle; + int third_offset; + int third_size; + unsigned int third_tiled; +} drmI830Sarea; + +/* Flags for perf_boxes + */ +#define I830_BOX_RING_EMPTY 0x1 /* populated by kernel */ +#define I830_BOX_FLIP 0x2 /* populated by kernel */ +#define I830_BOX_WAIT 0x4 /* populated by kernel & client */ +#define I830_BOX_TEXTURE_LOAD 0x8 /* populated by kernel */ +#define I830_BOX_LOST_CONTEXT 0x10 /* populated by client */ + + +typedef struct { + int start; /* agp offset */ + int used; /* nr bytes in use */ + int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ + int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ + int num_cliprects; /* mulitpass with multiple cliprects? */ + drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ +} drmI830BatchBuffer; + +typedef struct { + char *buf; /* agp offset */ + int sz; /* nr bytes in use */ + int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ + int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ + int num_cliprects; /* mulitpass with multiple cliprects? */ + drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ +} drmI830CmdBuffer; + +typedef struct { + int *irq_seq; +} drmI830IrqEmit; + +typedef struct { + int irq_seq; +} drmI830IrqWait; + +typedef struct { + int param; + int *value; +} drmI830GetParam; + +#define I830_PARAM_IRQ_ACTIVE 1 +#define I830_PARAM_ALLOW_BATCHBUFFER 2 + +typedef struct { + int param; + int value; +} drmI830SetParam; + +#define I830_SETPARAM_USE_MI_BATCHBUFFER_START 1 +#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY 2 +#define I830_SETPARAM_ALLOW_BATCHBUFFER 3 + + +/* A memory manager for regions of shared memory: + */ +#define I830_MEM_REGION_AGP 1 + +typedef struct { + int region; + int alignment; + int size; + int *region_offset; /* offset from start of fb or agp */ +} drmI830MemAlloc; + +typedef struct { + int region; + int region_offset; +} drmI830MemFree; + +typedef struct { + int region; + int size; + int start; +} drmI830MemInitHeap; + +typedef struct { + int region; +} drmI830MemDestroyHeap; + +#define DRM_I830_VBLANK_PIPE_A 1 +#define DRM_I830_VBLANK_PIPE_B 2 + +typedef struct { + int pipe; +} drmI830VBlankPipe; + +#endif /* _I830_DRM_H_ */ diff --git a/src/mesa/drivers/dri/intel_winsys/server/i830_dri.h b/src/mesa/drivers/dri/intel_winsys/server/i830_dri.h new file mode 100644 index 0000000000..c2a3af8cbf --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/server/i830_dri.h @@ -0,0 +1,63 @@ +/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ + +#ifndef _I830_DRI_H +#define _I830_DRI_H + +#include "xf86drm.h" +#include "i830_common.h" + +#define I830_MAX_DRAWABLES 256 + +#define I830_MAJOR_VERSION 1 +#define I830_MINOR_VERSION 7 +#define I830_PATCHLEVEL 2 + +#define I830_REG_SIZE 0x80000 + +typedef struct _I830DRIRec { + drm_handle_t regs; + drmSize regsSize; + + drmSize unused1; /* backbufferSize */ + drm_handle_t unused2; /* backbuffer */ + + drmSize unused3; /* depthbufferSize */ + drm_handle_t unused4; /* depthbuffer */ + + drmSize unused5; /* rotatedSize */ + drm_handle_t unused6; /* rotatedbuffer */ + + drm_handle_t unused7; /* textures */ + int unused8; /* textureSize */ + + drm_handle_t unused9; /* agp_buffers */ + drmSize unused10; /* agp_buf_size */ + + int deviceID; + int width; + int height; + int mem; + int cpp; + int bitsPerPixel; + + int unused11[8]; /* was front/back/depth/rotated offset/pitch */ + + int unused12; /* logTextureGranularity */ + int unused13; /* textureOffset */ + + int irq; + int sarea_priv_offset; +} I830DRIRec, *I830DRIPtr; + +typedef struct { + /* Nothing here yet */ + int dummy; +} I830ConfigPrivRec, *I830ConfigPrivPtr; + +typedef struct { + /* Nothing here yet */ + int dummy; +} I830DRIContextRec, *I830DRIContextPtr; + + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/server/intel.h b/src/mesa/drivers/dri/intel_winsys/server/intel.h new file mode 100644 index 0000000000..6ea72499c1 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/server/intel.h @@ -0,0 +1,331 @@ +#ifndef _INTEL_H_ +#define _INTEL_H_ + +#include "xf86drm.h" /* drm_handle_t, etc */ + +/* Intel */ +#ifndef PCI_CHIP_I810 +#define PCI_CHIP_I810 0x7121 +#define PCI_CHIP_I810_DC100 0x7123 +#define PCI_CHIP_I810_E 0x7125 +#define PCI_CHIP_I815 0x1132 +#define PCI_CHIP_I810_BRIDGE 0x7120 +#define PCI_CHIP_I810_DC100_BRIDGE 0x7122 +#define PCI_CHIP_I810_E_BRIDGE 0x7124 +#define PCI_CHIP_I815_BRIDGE 0x1130 +#endif + +#define PCI_CHIP_845_G 0x2562 +#define PCI_CHIP_I830_M 0x3577 + +#ifndef PCI_CHIP_I855_GM +#define PCI_CHIP_I855_GM 0x3582 +#define PCI_CHIP_I855_GM_BRIDGE 0x3580 +#endif + +#ifndef PCI_CHIP_I865_G +#define PCI_CHIP_I865_G 0x2572 +#define PCI_CHIP_I865_G_BRIDGE 0x2570 +#endif + +#ifndef PCI_CHIP_I915_G +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_G_BRIDGE 0x2580 +#endif + +#ifndef PCI_CHIP_I915_GM +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I915_GM_BRIDGE 0x2590 +#endif + +#ifndef PCI_CHIP_E7221_G +#define PCI_CHIP_E7221_G 0x258A +/* Same as I915_G_BRIDGE */ +#define PCI_CHIP_E7221_G_BRIDGE 0x2580 +#endif + +#ifndef PCI_CHIP_I945_G +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_G_BRIDGE 0x2770 +#endif + +#ifndef PCI_CHIP_I945_GM +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GM_BRIDGE 0x27A0 +#endif + +#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 || \ + pI810->Chipset == PCI_CHIP_I810_DC100 || \ + pI810->Chipset == PCI_CHIP_I810_E) +#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815) +#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M) +#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G) +#define IS_I85X(pI810) (pI810->Chipset == PCI_CHIP_I855_GM) +#define IS_I852(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME)) +#define IS_I855(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME)) +#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G) + +#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G) +#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM) +#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G) +#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM) +#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810)) + +#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810)) + +#define I830_GMCH_CTRL 0x52 + +#define I830_GMCH_MEM_MASK 0x1 +#define I830_GMCH_MEM_64M 0x1 +#define I830_GMCH_MEM_128M 0 + +#define I830_GMCH_GMS_MASK 0x70 +#define I830_GMCH_GMS_DISABLED 0x00 +#define I830_GMCH_GMS_LOCAL 0x10 +#define I830_GMCH_GMS_STOLEN_512 0x20 +#define I830_GMCH_GMS_STOLEN_1024 0x30 +#define I830_GMCH_GMS_STOLEN_8192 0x40 + +#define I855_GMCH_GMS_MASK (0x7 << 4) +#define I855_GMCH_GMS_DISABLED 0x00 +#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) +#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) +#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) +#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) +#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) +#define I915G_GMCH_GMS_STOLEN_48M (0x6 << 4) +#define I915G_GMCH_GMS_STOLEN_64M (0x7 << 4) + +typedef unsigned char Bool; +#define TRUE 1 +#define FALSE 0 + +#define PIPE_NONE 0<<0 +#define PIPE_CRT 1<<0 +#define PIPE_TV 1<<1 +#define PIPE_DFP 1<<2 +#define PIPE_LFP 1<<3 +#define PIPE_CRT2 1<<4 +#define PIPE_TV2 1<<5 +#define PIPE_DFP2 1<<6 +#define PIPE_LFP2 1<<7 + +typedef struct _I830MemPool *I830MemPoolPtr; +typedef struct _I830MemRange *I830MemRangePtr; +typedef struct _I830MemRange { + long Start; + long End; + long Size; + unsigned long Physical; + unsigned long Offset; /* Offset of AGP-allocated portion */ + unsigned long Alignment; + drm_handle_t Key; + unsigned long Pitch; // add pitch + I830MemPoolPtr Pool; +} I830MemRange; + +typedef struct _I830MemPool { + I830MemRange Total; + I830MemRange Free; + I830MemRange Fixed; + I830MemRange Allocated; +} I830MemPool; + +typedef struct { + int tail_mask; + I830MemRange mem; + unsigned char *virtual_start; + int head; + int tail; + int space; +} I830RingBuffer; + +typedef struct _I830Rec { + unsigned char *MMIOBase; + unsigned char *FbBase; + int cpp; + uint32_t aper_size; + unsigned int bios_version; + + /* These are set in PreInit and never changed. */ + long FbMapSize; + long TotalVideoRam; + I830MemRange StolenMemory; /* pre-allocated memory */ + long BIOSMemorySize; /* min stolen pool size */ + int BIOSMemSizeLoc; + + /* These change according to what has been allocated. */ + long FreeMemory; + I830MemRange MemoryAperture; + I830MemPool StolenPool; + long allocatedMemory; + + /* Regions allocated either from the above pools, or from agpgart. */ + /* for single and dual head configurations */ + I830MemRange FrontBuffer; + I830MemRange FrontBuffer2; + I830MemRange Scratch; + I830MemRange Scratch2; + + I830RingBuffer *LpRing; + + I830MemRange BackBuffer; + I830MemRange DepthBuffer; + I830MemRange TexMem; + int TexGranularity; + I830MemRange ContextMem; + int drmMinor; + Bool have3DWindows; + + Bool NeedRingBufferLow; + Bool allowPageFlip; + Bool disableTiling; + + int Chipset; + unsigned long LinearAddr; + unsigned long MMIOAddr; + + drmSize registerSize; /**< \brief MMIO register map size */ + drm_handle_t registerHandle; /**< \brief MMIO register map handle */ + // IOADDRESS ioBase; + int irq; /**< \brief IRQ number */ + int GttBound; + + drm_handle_t ring_map; + unsigned int Fence[8]; + +} I830Rec; + +/* + * 12288 is set as the maximum, chosen because it is enough for + * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare. + */ +#define I830_MAXIMUM_VBIOS_MEM 12288 +#define I830_DEFAULT_VIDEOMEM_2D (MB(32) / 1024) +#define I830_DEFAULT_VIDEOMEM_3D (MB(64) / 1024) + +/* Flags for memory allocation function */ +#define FROM_ANYWHERE 0x00000000 +#define FROM_POOL_ONLY 0x00000001 +#define FROM_NEW_ONLY 0x00000002 +#define FROM_MASK 0x0000000f + +#define ALLOCATE_AT_TOP 0x00000010 +#define ALLOCATE_AT_BOTTOM 0x00000020 +#define FORCE_GAPS 0x00000040 + +#define NEED_PHYSICAL_ADDR 0x00000100 +#define ALIGN_BOTH_ENDS 0x00000200 +#define FORCE_LOW 0x00000400 + +#define ALLOC_NO_TILING 0x00001000 +#define ALLOC_INITIAL 0x00002000 + +#define ALLOCATE_DRY_RUN 0x80000000 + +/* Chipset registers for VIDEO BIOS memory RW access */ +#define _855_DRAM_RW_CONTROL 0x58 +#define _845_DRAM_RW_CONTROL 0x90 +#define DRAM_WRITE 0x33330000 + +#define KB(x) ((x) * 1024) +#define MB(x) ((x) * KB(1024)) + +#define GTT_PAGE_SIZE KB(4) +#define ROUND_TO(x, y) (((x) + (y) - 1) / (y) * (y)) +#define ROUND_DOWN_TO(x, y) ((x) / (y) * (y)) +#define ROUND_TO_PAGE(x) ROUND_TO((x), GTT_PAGE_SIZE) +#define ROUND_TO_MB(x) ROUND_TO((x), MB(1)) +#define PRIMARY_RINGBUFFER_SIZE KB(128) + + +/* Ring buffer registers, p277, overview p19 + */ +#define LP_RING 0x2030 +#define HP_RING 0x2040 + +#define RING_TAIL 0x00 +#define TAIL_ADDR 0x000FFFF8 +#define I830_TAIL_MASK 0x001FFFF8 + +#define RING_HEAD 0x04 +#define HEAD_WRAP_COUNT 0xFFE00000 +#define HEAD_WRAP_ONE 0x00200000 +#define HEAD_ADDR 0x001FFFFC +#define I830_HEAD_MASK 0x001FFFFC + +#define RING_START 0x08 +#define START_ADDR 0x03FFFFF8 +#define I830_RING_START_MASK 0xFFFFF000 + +#define RING_LEN 0x0C +#define RING_NR_PAGES 0x001FF000 +#define I830_RING_NR_PAGES 0x001FF000 +#define RING_REPORT_MASK 0x00000006 +#define RING_REPORT_64K 0x00000002 +#define RING_REPORT_128K 0x00000004 +#define RING_NO_REPORT 0x00000000 +#define RING_VALID_MASK 0x00000001 +#define RING_VALID 0x00000001 +#define RING_INVALID 0x00000000 + + +/* Fence/Tiling ranges [0..7] + */ +#define FENCE 0x2000 +#define FENCE_NR 8 + +#define I915G_FENCE_START_MASK 0x0ff00000 + +#define I830_FENCE_START_MASK 0x07f80000 + +#define FENCE_START_MASK 0x03F80000 +#define FENCE_X_MAJOR 0x00000000 +#define FENCE_Y_MAJOR 0x00001000 +#define FENCE_SIZE_MASK 0x00000700 +#define FENCE_SIZE_512K 0x00000000 +#define FENCE_SIZE_1M 0x00000100 +#define FENCE_SIZE_2M 0x00000200 +#define FENCE_SIZE_4M 0x00000300 +#define FENCE_SIZE_8M 0x00000400 +#define FENCE_SIZE_16M 0x00000500 +#define FENCE_SIZE_32M 0x00000600 +#define FENCE_SIZE_64M 0x00000700 +#define I915G_FENCE_SIZE_1M 0x00000000 +#define I915G_FENCE_SIZE_2M 0x00000100 +#define I915G_FENCE_SIZE_4M 0x00000200 +#define I915G_FENCE_SIZE_8M 0x00000300 +#define I915G_FENCE_SIZE_16M 0x00000400 +#define I915G_FENCE_SIZE_32M 0x00000500 +#define I915G_FENCE_SIZE_64M 0x00000600 +#define I915G_FENCE_SIZE_128M 0x00000700 +#define FENCE_PITCH_1 0x00000000 +#define FENCE_PITCH_2 0x00000010 +#define FENCE_PITCH_4 0x00000020 +#define FENCE_PITCH_8 0x00000030 +#define FENCE_PITCH_16 0x00000040 +#define FENCE_PITCH_32 0x00000050 +#define FENCE_PITCH_64 0x00000060 +#define FENCE_VALID 0x00000001 + +#include + +# define MMIO_IN8(base, offset) \ + *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) +# define MMIO_IN32(base, offset) \ + read_MMIO_LE32(base, offset) +# define MMIO_OUT8(base, offset, val) \ + *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val) +# define MMIO_OUT32(base, offset, val) \ + *(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val) + + + /* Memory mapped register access macros */ +#define INREG8(addr) MMIO_IN8(MMIO, addr) +#define INREG(addr) MMIO_IN32(MMIO, addr) +#define OUTREG8(addr, val) MMIO_OUT8(MMIO, addr, val) +#define OUTREG(addr, val) MMIO_OUT32(MMIO, addr, val) + +#define DSPABASE 0x70184 + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/server/intel_dri.c b/src/mesa/drivers/dri/intel_winsys/server/intel_dri.c new file mode 100644 index 0000000000..e49c4214ad --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/server/intel_dri.c @@ -0,0 +1,1306 @@ +/** + * \file server/intel_dri.c + * \brief File to perform the device-specific initialization tasks typically + * done in the X server. + * + * Here they are converted to run in the client (or perhaps a standalone + * process), and to work with the frame buffer device rather than the X + * server infrastructure. + * + * Copyright (C) 2006 Dave Airlie (airlied@linux.ie) + + Permission is hereby granted, free of charge, to any person obtaining a + copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sub license, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial portions + of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR + ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +*/ + +#include +#include +#include +#include +#include + +#include "driver.h" +#include "drm.h" + +#include "intel.h" +#include "i830_dri.h" + +#include "memops.h" +#include "pciaccess.h" + +static size_t drm_page_size; +static int nextTile = 0; +#define xf86DrvMsg(...) do {} while(0) + +static const int pitches[] = { + 128 * 8, + 128 * 16, + 128 * 32, + 128 * 64, + 0 +}; + +static Bool I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea); + +static unsigned long +GetBestTileAlignment(unsigned long size) +{ + unsigned long i; + + for (i = KB(512); i < size; i <<= 1) + ; + + if (i > MB(64)) + i = MB(64); + + return i; +} + +static void SetFenceRegs(const DRIDriverContext *ctx, I830Rec *pI830) +{ + int i; + unsigned char *MMIO = ctx->MMIOAddress; + + for (i = 0; i < 8; i++) { + OUTREG(FENCE + i * 4, pI830->Fence[i]); + // if (I810_DEBUG & DEBUG_VERBOSE_VGA) + fprintf(stderr,"Fence Register : %x\n", pI830->Fence[i]); + } +} + +/* Tiled memory is good... really, really good... + * + * Need to make it less likely that we miss out on this - probably + * need to move the frontbuffer away from the 'guarenteed' alignment + * of the first memory segment, or perhaps allocate a discontigous + * framebuffer to get more alignment 'sweet spots'. + */ +static void +SetFence(const DRIDriverContext *ctx, I830Rec *pI830, + int nr, unsigned int start, unsigned int pitch, + unsigned int size) +{ + unsigned int val; + unsigned int fence_mask = 0; + unsigned int fence_pitch; + + if (nr < 0 || nr > 7) { + fprintf(stderr, + "SetFence: fence %d out of range\n",nr); + return; + } + + pI830->Fence[nr] = 0; + + if (IS_I9XX(pI830)) + fence_mask = ~I915G_FENCE_START_MASK; + else + fence_mask = ~I830_FENCE_START_MASK; + + if (start & fence_mask) { + fprintf(stderr, + "SetFence: %d: start (0x%08x) is not %s aligned\n", + nr, start, (IS_I9XX(pI830)) ? "1MB" : "512k"); + return; + } + + if (start % size) { + fprintf(stderr, + "SetFence: %d: start (0x%08x) is not size (%dk) aligned\n", + nr, start, size / 1024); + return; + } + + if (pitch & 127) { + fprintf(stderr, + "SetFence: %d: pitch (%d) not a multiple of 128 bytes\n", + nr, pitch); + return; + } + + val = (start | FENCE_X_MAJOR | FENCE_VALID); + + if (IS_I9XX(pI830)) { + switch (size) { + case MB(1): + val |= I915G_FENCE_SIZE_1M; + break; + case MB(2): + val |= I915G_FENCE_SIZE_2M; + break; + case MB(4): + val |= I915G_FENCE_SIZE_4M; + break; + case MB(8): + val |= I915G_FENCE_SIZE_8M; + break; + case MB(16): + val |= I915G_FENCE_SIZE_16M; + break; + case MB(32): + val |= I915G_FENCE_SIZE_32M; + break; + case MB(64): + val |= I915G_FENCE_SIZE_64M; + break; + default: + fprintf(stderr, + "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); + return; + } + } else { + switch (size) { + case KB(512): + val |= FENCE_SIZE_512K; + break; + case MB(1): + val |= FENCE_SIZE_1M; + break; + case MB(2): + val |= FENCE_SIZE_2M; + break; + case MB(4): + val |= FENCE_SIZE_4M; + break; + case MB(8): + val |= FENCE_SIZE_8M; + break; + case MB(16): + val |= FENCE_SIZE_16M; + break; + case MB(32): + val |= FENCE_SIZE_32M; + break; + case MB(64): + val |= FENCE_SIZE_64M; + break; + default: + fprintf(stderr, + "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); + return; + } + } + + if (IS_I9XX(pI830)) + fence_pitch = pitch / 512; + else + fence_pitch = pitch / 128; + + switch (fence_pitch) { + case 1: + val |= FENCE_PITCH_1; + break; + case 2: + val |= FENCE_PITCH_2; + break; + case 4: + val |= FENCE_PITCH_4; + break; + case 8: + val |= FENCE_PITCH_8; + break; + case 16: + val |= FENCE_PITCH_16; + break; + case 32: + val |= FENCE_PITCH_32; + break; + case 64: + val |= FENCE_PITCH_64; + break; + default: + fprintf(stderr, + "SetFence: %d: illegal pitch (%d)\n", nr, pitch); + return; + } + + pI830->Fence[nr] = val; +} + +static Bool +MakeTiles(const DRIDriverContext *ctx, I830Rec *pI830, I830MemRange *pMem) +{ + int pitch, ntiles, i; + + pitch = pMem->Pitch * ctx->cpp; + /* + * Simply try to break the region up into at most four pieces of size + * equal to the alignment. + */ + ntiles = ROUND_TO(pMem->Size, pMem->Alignment) / pMem->Alignment; + if (ntiles >= 4) { + return FALSE; + } + + for (i = 0; i < ntiles; i++, nextTile++) { + SetFence(ctx, pI830, nextTile, pMem->Start + i * pMem->Alignment, + pitch, pMem->Alignment); + } + return TRUE; +} + +static void I830SetupMemoryTiling(const DRIDriverContext *ctx, I830Rec *pI830) +{ + int i; + + /* Clear out */ + for (i = 0; i < 8; i++) + pI830->Fence[i] = 0; + + nextTile = 0; + + if (pI830->BackBuffer.Alignment >= KB(512)) { + if (MakeTiles(ctx, pI830, &(pI830->BackBuffer))) { + fprintf(stderr, + "Activating tiled memory for the back buffer.\n"); + } else { + fprintf(stderr, + "MakeTiles failed for the back buffer.\n"); + pI830->allowPageFlip = FALSE; + } + } + + if (pI830->DepthBuffer.Alignment >= KB(512)) { + if (MakeTiles(ctx, pI830, &(pI830->DepthBuffer))) { + fprintf(stderr, + "Activating tiled memory for the depth buffer.\n"); + } else { + fprintf(stderr, + "MakeTiles failed for the depth buffer.\n"); + } + } + + return; +} + +static int I830DetectMemory(const DRIDriverContext *ctx, I830Rec *pI830) +{ + struct pci_device host_bridge, ig_dev; + uint32_t gmch_ctrl; + int memsize = 0; + int range; + uint32_t aper_size; + uint32_t membase2 = 0; + + memset(&host_bridge, 0, sizeof(host_bridge)); + memset(&ig_dev, 0, sizeof(ig_dev)); + + ig_dev.dev = 2; + + pci_device_cfg_read_u32(&host_bridge, &gmch_ctrl, I830_GMCH_CTRL); + + if (IS_I830(pI830) || IS_845G(pI830)) { + if ((gmch_ctrl & I830_GMCH_MEM_MASK) == I830_GMCH_MEM_128M) { + aper_size = 0x80000000; + } else { + aper_size = 0x40000000; + } + } else { + if (IS_I9XX(pI830)) { + int ret; + ret = pci_device_cfg_read_u32(&ig_dev, &membase2, 0x18); + if (membase2 & 0x08000000) + aper_size = 0x8000000; + else + aper_size = 0x10000000; + + fprintf(stderr,"aper size is %08X %08x %d\n", aper_size, membase2, ret); + } else + aper_size = 0x8000000; + } + + pI830->aper_size = aper_size; + + + /* We need to reduce the stolen size, by the GTT and the popup. + * The GTT varying according the the FbMapSize and the popup is 4KB */ + range = (ctx->shared.fbSize / (1024*1024)) + 4; + + if (IS_I85X(pI830) || IS_I865G(pI830) || IS_I9XX(pI830)) { + switch (gmch_ctrl & I830_GMCH_GMS_MASK) { + case I855_GMCH_GMS_STOLEN_1M: + memsize = MB(1) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_4M: + memsize = MB(4) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_8M: + memsize = MB(8) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_16M: + memsize = MB(16) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_32M: + memsize = MB(32) - KB(range); + break; + case I915G_GMCH_GMS_STOLEN_48M: + if (IS_I9XX(pI830)) + memsize = MB(48) - KB(range); + break; + case I915G_GMCH_GMS_STOLEN_64M: + if (IS_I9XX(pI830)) + memsize = MB(64) - KB(range); + break; + } + } else { + switch (gmch_ctrl & I830_GMCH_GMS_MASK) { + case I830_GMCH_GMS_STOLEN_512: + memsize = KB(512) - KB(range); + break; + case I830_GMCH_GMS_STOLEN_1024: + memsize = MB(1) - KB(range); + break; + case I830_GMCH_GMS_STOLEN_8192: + memsize = MB(8) - KB(range); + break; + case I830_GMCH_GMS_LOCAL: + memsize = 0; + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "Local memory found, but won't be used.\n"); + break; + } + } + if (memsize > 0) { + fprintf(stderr, + "detected %d kB stolen memory.\n", memsize / 1024); + } else { + fprintf(stderr, + "no video memory detected.\n"); + } + return memsize; +} + +static int AgpInit(const DRIDriverContext *ctx, I830Rec *info) +{ + unsigned long mode = 0x4; + + if (drmAgpAcquire(ctx->drmFD) < 0) { + fprintf(stderr, "[gart] AGP not available\n"); + return 0; + } + + if (drmAgpEnable(ctx->drmFD, mode) < 0) { + fprintf(stderr, "[gart] AGP not enabled\n"); + drmAgpRelease(ctx->drmFD); + return 0; + } + else + fprintf(stderr, "[gart] AGP enabled at %dx\n", ctx->agpmode); + + return 1; +} + +/* + * Allocate memory from the given pool. Grow the pool if needed and if + * possible. + */ +static unsigned long +AllocFromPool(const DRIDriverContext *ctx, I830Rec *pI830, + I830MemRange *result, I830MemPool *pool, + long size, unsigned long alignment, int flags) +{ + long needed, start, end; + + if (!result || !pool || !size) + return 0; + + /* Calculate how much space is needed. */ + if (alignment <= GTT_PAGE_SIZE) + needed = size; + else { + start = ROUND_TO(pool->Free.Start, alignment); + end = ROUND_TO(start + size, alignment); + needed = end - pool->Free.Start; + } + if (needed > pool->Free.Size) { + return 0; + } + + result->Start = ROUND_TO(pool->Free.Start, alignment); + pool->Free.Start += needed; + result->End = pool->Free.Start; + + pool->Free.Size = pool->Free.End - pool->Free.Start; + result->Size = result->End - result->Start; + result->Pool = pool; + result->Alignment = alignment; + return needed; +} + +static unsigned long AllocFromAGP(const DRIDriverContext *ctx, I830Rec *pI830, long size, unsigned long alignment, I830MemRange *result) +{ + unsigned long start, end; + unsigned long newApStart, newApEnd; + int ret; + if (!result || !size) + return 0; + + if (!alignment) + alignment = 4; + + start = ROUND_TO(pI830->MemoryAperture.Start, alignment); + end = ROUND_TO(start + size, alignment); + newApStart = end; + newApEnd = pI830->MemoryAperture.End; + + ret=drmAgpAlloc(ctx->drmFD, size, 0, &(result->Physical), (drm_handle_t *)&(result->Key)); + + if (ret) + { + fprintf(stderr,"drmAgpAlloc failed %d\n", ret); + return 0; + } + pI830->allocatedMemory += size; + pI830->MemoryAperture.Start = newApStart; + pI830->MemoryAperture.End = newApEnd; + pI830->MemoryAperture.Size = newApEnd - newApStart; + // pI830->FreeMemory -= size; + result->Start = start; + result->End = start + size; + result->Size = size; + result->Offset = start; + result->Alignment = alignment; + result->Pool = NULL; + + return size; +} + +unsigned long +I830AllocVidMem(const DRIDriverContext *ctx, I830Rec *pI830, + I830MemRange *result, I830MemPool *pool, long size, + unsigned long alignment, int flags) +{ + unsigned long ret; + + if (!result) + return 0; + + /* Make sure these are initialised. */ + result->Size = 0; + result->Key = -1; + + if (!size) { + return 0; + } + + if (pool->Free.Size < size) { + ret = AllocFromAGP(ctx, pI830, size, alignment, result); + } + else { + ret = AllocFromPool(ctx, pI830, result, pool, size, alignment, flags); + if (ret == 0) + ret = AllocFromAGP(ctx, pI830, size, alignment, result); + } + return ret; +} + +static Bool BindAgpRange(const DRIDriverContext *ctx, I830MemRange *mem) +{ + if (!mem) + return FALSE; + + if (mem->Key == -1) + return TRUE; + + return !drmAgpBind(ctx->drmFD, mem->Key, mem->Offset); +} + +/* simple memory allocation routines needed */ +/* put ring buffer in low memory */ +/* need to allocate front, back, depth buffers aligned correctly, + allocate ring buffer, +*/ + +/* */ +static Bool +I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned long size, ret; + unsigned long lines, lineSize, align; + + /* allocate ring buffer */ + memset(pI830->LpRing, 0, sizeof(I830RingBuffer)); + pI830->LpRing->mem.Key = -1; + + size = PRIMARY_RINGBUFFER_SIZE; + + ret = I830AllocVidMem(ctx, pI830, &pI830->LpRing->mem, &pI830->StolenPool, size, 0x1000, 0); + + if (ret != size) + { + fprintf(stderr,"unable to allocate ring buffer %ld\n", ret); + return FALSE; + } + + pI830->LpRing->tail_mask = pI830->LpRing->mem.Size - 1; + + + /* allocate front buffer */ + memset(&(pI830->FrontBuffer), 0, sizeof(pI830->FrontBuffer)); + pI830->FrontBuffer.Key = -1; + pI830->FrontBuffer.Pitch = ctx->shared.virtualWidth; + + align = KB(512); + + lineSize = ctx->shared.virtualWidth * ctx->cpp; + lines = (ctx->shared.virtualHeight + 15) / 16 * 16; + size = lineSize * lines; + size = ROUND_TO_PAGE(size); + + align = GetBestTileAlignment(size); + + ret = I830AllocVidMem(ctx, pI830, &pI830->FrontBuffer, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate front buffer %ld\n", ret); + return FALSE; + } + + memset(&(pI830->BackBuffer), 0, sizeof(pI830->BackBuffer)); + pI830->BackBuffer.Key = -1; + pI830->BackBuffer.Pitch = ctx->shared.virtualWidth; + + ret = I830AllocVidMem(ctx, pI830, &pI830->BackBuffer, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate back buffer %ld\n", ret); + return FALSE; + } + + memset(&(pI830->DepthBuffer), 0, sizeof(pI830->DepthBuffer)); + pI830->DepthBuffer.Key = -1; + pI830->DepthBuffer.Pitch = ctx->shared.virtualWidth; + + ret = I830AllocVidMem(ctx, pI830, &pI830->DepthBuffer, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate depth buffer %ld\n", ret); + return FALSE; + } + + memset(&(pI830->ContextMem), 0, sizeof(pI830->ContextMem)); + pI830->ContextMem.Key = -1; + size = KB(32); + + ret = I830AllocVidMem(ctx, pI830, &pI830->ContextMem, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate context buffer %ld\n", ret); + return FALSE; + } + +#if 0 + memset(&(pI830->TexMem), 0, sizeof(pI830->TexMem)); + pI830->TexMem.Key = -1; + + size = 32768 * 1024; + ret = AllocFromAGP(ctx, pI830, size, align, &pI830->TexMem); + if (ret < size) + { + fprintf(stderr,"unable to allocate texture memory %ld\n", ret); + return FALSE; + } +#endif + + return TRUE; +} + +static Bool +I830BindMemory(const DRIDriverContext *ctx, I830Rec *pI830) +{ + if (!BindAgpRange(ctx, &pI830->LpRing->mem)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->FrontBuffer)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->BackBuffer)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->DepthBuffer)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->ContextMem)) + return FALSE; +#if 0 + if (!BindAgpRange(ctx, &pI830->TexMem)) + return FALSE; +#endif + return TRUE; +} + +static void SetupDRIMM(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned long aperEnd = ROUND_DOWN_TO(pI830->aper_size, GTT_PAGE_SIZE) / GTT_PAGE_SIZE; + unsigned long aperStart = ROUND_TO(pI830->aper_size - KB(32768), GTT_PAGE_SIZE) / GTT_PAGE_SIZE; + + fprintf(stderr, "aper size is %08X\n", ctx->shared.fbSize); + if (drmMMInit(ctx->drmFD, aperStart, aperEnd - aperStart, DRM_BO_MEM_TT)) { + fprintf(stderr, + "DRM MM Initialization Failed\n"); + } else { + fprintf(stderr, + "DRM MM Initialized at offset 0x%lx length %d page\n", aperStart, aperEnd-aperStart); + } + +} + +static Bool +I830CleanupDma(const DRIDriverContext *ctx) +{ + drmI830Init info; + + memset(&info, 0, sizeof(drmI830Init)); + info.func = I830_CLEANUP_DMA; + + if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, + &info, sizeof(drmI830Init))) { + fprintf(stderr, "I830 Dma Cleanup Failed\n"); + return FALSE; + } + + return TRUE; +} + +static Bool +I830InitDma(const DRIDriverContext *ctx, I830Rec *pI830) +{ + I830RingBuffer *ring = pI830->LpRing; + drmI830Init info; + + memset(&info, 0, sizeof(drmI830Init)); + info.func = I830_INIT_DMA; + + info.ring_start = ring->mem.Start + pI830->LinearAddr; + info.ring_end = ring->mem.End + pI830->LinearAddr; + info.ring_size = ring->mem.Size; + + info.mmio_offset = (unsigned int)ctx->MMIOStart; + + info.sarea_priv_offset = sizeof(drm_sarea_t); + + info.front_offset = pI830->FrontBuffer.Start; + info.back_offset = pI830->BackBuffer.Start; + info.depth_offset = pI830->DepthBuffer.Start; + info.w = ctx->shared.virtualWidth; + info.h = ctx->shared.virtualHeight; + info.pitch = ctx->shared.virtualWidth; + info.back_pitch = pI830->BackBuffer.Pitch; + info.depth_pitch = pI830->DepthBuffer.Pitch; + info.cpp = ctx->cpp; + + if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, + &info, sizeof(drmI830Init))) { + fprintf(stderr, + "I830 Dma Initialization Failed\n"); + return FALSE; + } + + return TRUE; +} + +static int I830CheckDRMVersion( const DRIDriverContext *ctx, + I830Rec *pI830 ) +{ + drmVersionPtr version; + + version = drmGetVersion(ctx->drmFD); + + if (version) { + int req_minor, req_patch; + + req_minor = 4; + req_patch = 0; + + if (version->version_major != 1 || + version->version_minor < req_minor || + (version->version_minor == req_minor && + version->version_patchlevel < req_patch)) { + /* Incompatible drm version */ + fprintf(stderr, + "[dri] I830DRIScreenInit failed because of a version " + "mismatch.\n" + "[dri] i915.o kernel module version is %d.%d.%d " + "but version 1.%d.%d or newer is needed.\n" + "[dri] Disabling DRI.\n", + version->version_major, + version->version_minor, + version->version_patchlevel, + req_minor, + req_patch); + drmFreeVersion(version); + return 0; + } + + pI830->drmMinor = version->version_minor; + drmFreeVersion(version); + } + return 1; +} + +static void +I830SetRingRegs(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned int itemp; + unsigned char *MMIO = ctx->MMIOAddress; + + OUTREG(LP_RING + RING_LEN, 0); + OUTREG(LP_RING + RING_TAIL, 0); + OUTREG(LP_RING + RING_HEAD, 0); + + if ((long)(pI830->LpRing->mem.Start & I830_RING_START_MASK) != + pI830->LpRing->mem.Start) { + fprintf(stderr, + "I830SetRingRegs: Ring buffer start (%lx) violates its " + "mask (%x)\n", pI830->LpRing->mem.Start, I830_RING_START_MASK); + } + /* Don't care about the old value. Reserved bits must be zero anyway. */ + itemp = pI830->LpRing->mem.Start & I830_RING_START_MASK; + OUTREG(LP_RING + RING_START, itemp); + + if (((pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES) != + pI830->LpRing->mem.Size - 4096) { + fprintf(stderr, + "I830SetRingRegs: Ring buffer size - 4096 (%lx) violates its " + "mask (%x)\n", pI830->LpRing->mem.Size - 4096, + I830_RING_NR_PAGES); + } + /* Don't care about the old value. Reserved bits must be zero anyway. */ + itemp = (pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES; + itemp |= (RING_NO_REPORT | RING_VALID); + OUTREG(LP_RING + RING_LEN, itemp); + + pI830->LpRing->head = INREG(LP_RING + RING_HEAD) & I830_HEAD_MASK; + pI830->LpRing->tail = INREG(LP_RING + RING_TAIL); + pI830->LpRing->space = pI830->LpRing->head - (pI830->LpRing->tail + 8); + if (pI830->LpRing->space < 0) + pI830->LpRing->space += pI830->LpRing->mem.Size; + + SetFenceRegs(ctx, pI830); + + /* RESET THE DISPLAY PIPE TO POINT TO THE FRONTBUFFER - hacky + hacky hacky */ + OUTREG(DSPABASE, pI830->FrontBuffer.Start + pI830->LinearAddr); + +} + +static Bool +I830SetParam(const DRIDriverContext *ctx, int param, int value) +{ + drmI830SetParam sp; + + memset(&sp, 0, sizeof(sp)); + sp.param = param; + sp.value = value; + + if (drmCommandWrite(ctx->drmFD, DRM_I830_SETPARAM, &sp, sizeof(sp))) { + fprintf(stderr, "I830 SetParam Failed\n"); + return FALSE; + } + + return TRUE; +} + +static Bool +I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ + fprintf(stderr, + "[drm] Mapping front buffer\n"); + + if (drmAddMap(ctx->drmFD, + (drm_handle_t)(sarea->front_offset + pI830->LinearAddr), + sarea->front_size, + DRM_FRAME_BUFFER, /*DRM_AGP,*/ + 0, + &sarea->front_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(front_handle) failed. Disabling DRI\n"); + return FALSE; + } + ctx->shared.hFrameBuffer = sarea->front_handle; + ctx->shared.fbSize = sarea->front_size; + fprintf(stderr, "[drm] Front Buffer = 0x%08x\n", + sarea->front_handle); + + if (drmAddMap(ctx->drmFD, + (drm_handle_t)(sarea->back_offset), + sarea->back_size, DRM_AGP, 0, + &sarea->back_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(back_handle) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] Back Buffer = 0x%08x\n", + sarea->back_handle); + + if (drmAddMap(ctx->drmFD, + (drm_handle_t)sarea->depth_offset, + sarea->depth_size, DRM_AGP, 0, + &sarea->depth_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(depth_handle) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] Depth Buffer = 0x%08x\n", + sarea->depth_handle); + +#if 0 + if (drmAddMap(ctx->drmFD, + (drm_handle_t)sarea->tex_offset, + sarea->tex_size, DRM_AGP, 0, + &sarea->tex_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(tex_handle) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] textures = 0x%08x\n", + sarea->tex_handle); +#endif + return TRUE; +} + + +static void +I830DRIUnmapScreenRegions(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ +#if 1 + if (sarea->front_handle) { + drmRmMap(ctx->drmFD, sarea->front_handle); + sarea->front_handle = 0; + } +#endif + if (sarea->back_handle) { + drmRmMap(ctx->drmFD, sarea->back_handle); + sarea->back_handle = 0; + } + if (sarea->depth_handle) { + drmRmMap(ctx->drmFD, sarea->depth_handle); + sarea->depth_handle = 0; + } + if (sarea->tex_handle) { + drmRmMap(ctx->drmFD, sarea->tex_handle); + sarea->tex_handle = 0; + } +} + +static Bool +I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ + if (drmAddMap(ctx->drmFD, + (drm_handle_t)pI830->LpRing->mem.Start, + pI830->LpRing->mem.Size, DRM_AGP, 0, + &pI830->ring_map) < 0) { + fprintf(stderr, + "[drm] drmAddMap(ring_map) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] ring buffer = 0x%08x\n", + pI830->ring_map); + + if (I830InitDma(ctx, pI830) == FALSE) { + return FALSE; + } + + /* init to zero to be safe */ + + I830DRIMapScreenRegions(ctx, pI830, sarea); + SetupDRIMM(ctx, pI830); + + if (ctx->pciDevice != PCI_CHIP_845_G && + ctx->pciDevice != PCI_CHIP_I830_M) { + I830SetParam(ctx, I830_SETPARAM_USE_MI_BATCHBUFFER_START, 1 ); + } + + /* Okay now initialize the dma engine */ + { + pI830->irq = drmGetInterruptFromBusID(ctx->drmFD, + ctx->pciBus, + ctx->pciDevice, + ctx->pciFunc); + + if (drmCtlInstHandler(ctx->drmFD, pI830->irq)) { + fprintf(stderr, + "[drm] failure adding irq handler\n"); + pI830->irq = 0; + return FALSE; + } + else + fprintf(stderr, + "[drm] dma control initialized, using IRQ %d\n", + pI830->irq); + } + + fprintf(stderr, "[dri] visual configs initialized\n"); + + return TRUE; +} + +static Bool +I830ClearScreen(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ + /* need to drmMap front and back buffers and zero them */ + drmAddress map_addr; + int ret; + + ret = drmMap(ctx->drmFD, + sarea->front_handle, + sarea->front_size, + &map_addr); + + if (ret) + { + fprintf(stderr, "Unable to map front buffer\n"); + return FALSE; + } + + drimemsetio((char *)map_addr, + 0, + sarea->front_size); + drmUnmap(map_addr, sarea->front_size); + + + ret = drmMap(ctx->drmFD, + sarea->back_handle, + sarea->back_size, + &map_addr); + + if (ret) + { + fprintf(stderr, "Unable to map back buffer\n"); + return FALSE; + } + + drimemsetio((char *)map_addr, + 0, + sarea->back_size); + drmUnmap(map_addr, sarea->back_size); + + return TRUE; +} + +static Bool +I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) + +{ + I830DRIPtr pI830DRI; + drmI830Sarea *pSAREAPriv; + int err; + + drm_page_size = getpagesize(); + + pI830->registerSize = ctx->MMIOSize; + /* This is a hack for now. We have to have more than a 4k page here + * because of the size of the state. However, the state should be + * in a per-context mapping. This will be added in the Mesa 3.5 port + * of the I830 driver. + */ + ctx->shared.SAREASize = SAREA_MAX; + + /* Note that drmOpen will try to load the kernel module, if needed. */ + ctx->drmFD = drmOpen("i915", NULL ); + if (ctx->drmFD < 0) { + fprintf(stderr, "[drm] drmOpen failed\n"); + return 0; + } + + if ((err = drmSetBusid(ctx->drmFD, ctx->pciBusID)) < 0) { + fprintf(stderr, "[drm] drmSetBusid failed (%d, %s), %s\n", + ctx->drmFD, ctx->pciBusID, strerror(-err)); + return 0; + } + + if (drmAddMap( ctx->drmFD, + 0, + ctx->shared.SAREASize, + DRM_SHM, + DRM_CONTAINS_LOCK, + &ctx->shared.hSAREA) < 0) + { + fprintf(stderr, "[drm] drmAddMap failed\n"); + return 0; + } + + fprintf(stderr, "[drm] added %d byte SAREA at 0x%08x\n", + ctx->shared.SAREASize, ctx->shared.hSAREA); + + if (drmMap( ctx->drmFD, + ctx->shared.hSAREA, + ctx->shared.SAREASize, + (drmAddressPtr)(&ctx->pSAREA)) < 0) + { + fprintf(stderr, "[drm] drmMap failed\n"); + return 0; + + } + + memset(ctx->pSAREA, 0, ctx->shared.SAREASize); + fprintf(stderr, "[drm] mapped SAREA 0x%08x to %p, size %d\n", + ctx->shared.hSAREA, ctx->pSAREA, ctx->shared.SAREASize); + + + if (drmAddMap(ctx->drmFD, + ctx->MMIOStart, + ctx->MMIOSize, + DRM_REGISTERS, + DRM_READ_ONLY, + &pI830->registerHandle) < 0) { + fprintf(stderr, "[drm] drmAddMap mmio failed\n"); + return 0; + } + fprintf(stderr, + "[drm] register handle = 0x%08x\n", pI830->registerHandle); + + + if (!I830CheckDRMVersion(ctx, pI830)) { + return FALSE; + } + + /* Create a 'server' context so we can grab the lock for + * initialization ioctls. + */ + if ((err = drmCreateContext(ctx->drmFD, &ctx->serverContext)) != 0) { + fprintf(stderr, "%s: drmCreateContext failed %d\n", __FUNCTION__, err); + return 0; + } + + DRM_LOCK(ctx->drmFD, ctx->pSAREA, ctx->serverContext, 0); + + /* Initialize the SAREA private data structure */ + pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + + sizeof(drm_sarea_t)); + memset(pSAREAPriv, 0, sizeof(*pSAREAPriv)); + + pI830->StolenMemory.Size = I830DetectMemory(ctx, pI830); + pI830->StolenMemory.Start = 0; + pI830->StolenMemory.End = pI830->StolenMemory.Size; + + pI830->MemoryAperture.Start = pI830->StolenMemory.End; + pI830->MemoryAperture.End = KB(40000); + pI830->MemoryAperture.Size = pI830->MemoryAperture.End - pI830->MemoryAperture.Start; + + pI830->StolenPool.Fixed = pI830->StolenMemory; + pI830->StolenPool.Total = pI830->StolenMemory; + pI830->StolenPool.Free = pI830->StolenPool.Total; + pI830->FreeMemory = pI830->StolenPool.Total.Size; + + if (!AgpInit(ctx, pI830)) + return FALSE; + + if (I830AllocateMemory(ctx, pI830) == FALSE) + { + return FALSE; + } + + if (I830BindMemory(ctx, pI830) == FALSE) + { + return FALSE; + } + + pSAREAPriv->rotated_offset = -1; + pSAREAPriv->rotated_size = 0; + pSAREAPriv->rotated_pitch = ctx->shared.virtualWidth; + + pSAREAPriv->front_offset = pI830->FrontBuffer.Start; + pSAREAPriv->front_size = pI830->FrontBuffer.Size; + pSAREAPriv->width = ctx->shared.virtualWidth; + pSAREAPriv->height = ctx->shared.virtualHeight; + pSAREAPriv->pitch = ctx->shared.virtualWidth; + pSAREAPriv->virtualX = ctx->shared.virtualWidth; + pSAREAPriv->virtualY = ctx->shared.virtualHeight; + pSAREAPriv->back_offset = pI830->BackBuffer.Start; + pSAREAPriv->back_size = pI830->BackBuffer.Size; + pSAREAPriv->depth_offset = pI830->DepthBuffer.Start; + pSAREAPriv->depth_size = pI830->DepthBuffer.Size; +#if 0 + pSAREAPriv->tex_offset = pI830->TexMem.Start; + pSAREAPriv->tex_size = pI830->TexMem.Size; +#endif + pSAREAPriv->log_tex_granularity = pI830->TexGranularity; + + ctx->driverClientMsg = malloc(sizeof(I830DRIRec)); + ctx->driverClientMsgSize = sizeof(I830DRIRec); + pI830DRI = (I830DRIPtr)ctx->driverClientMsg; + pI830DRI->deviceID = pI830->Chipset; + pI830DRI->regsSize = I830_REG_SIZE; + pI830DRI->width = ctx->shared.virtualWidth; + pI830DRI->height = ctx->shared.virtualHeight; + pI830DRI->mem = ctx->shared.fbSize; + pI830DRI->cpp = ctx->cpp; + + pI830DRI->bitsPerPixel = ctx->bpp; + pI830DRI->sarea_priv_offset = sizeof(drm_sarea_t); + + err = I830DRIDoMappings(ctx, pI830, pSAREAPriv); + if (err == FALSE) + return FALSE; + + I830SetupMemoryTiling(ctx, pI830); + + /* Quick hack to clear the front & back buffers. Could also use + * the clear ioctl to do this, but would need to setup hw state + * first. + */ + I830ClearScreen(ctx, pI830, pSAREAPriv); + + I830SetRingRegs(ctx, pI830); + + return TRUE; +} + + +/** + * \brief Validate the fbdev mode. + * + * \param ctx display handle. + * + * \return one on success, or zero on failure. + * + * Saves some registers and returns 1. + * + * \sa radeonValidateMode(). + */ +static int i830ValidateMode( const DRIDriverContext *ctx ) +{ + return 1; +} + +/** + * \brief Examine mode returned by fbdev. + * + * \param ctx display handle. + * + * \return one on success, or zero on failure. + * + * Restores registers that fbdev has clobbered and returns 1. + * + * \sa i810ValidateMode(). + */ +static int i830PostValidateMode( const DRIDriverContext *ctx ) +{ + I830Rec *pI830 = ctx->driverPrivate; + + I830SetRingRegs(ctx, pI830); + return 1; +} + + +/** + * \brief Initialize the framebuffer device mode + * + * \param ctx display handle. + * + * \return one on success, or zero on failure. + * + * Fills in \p info with some default values and some information from \p ctx + * and then calls I810ScreenInit() for the screen initialization. + * + * Before exiting clears the framebuffer memory accessing it directly. + */ +static int i830InitFBDev( DRIDriverContext *ctx ) +{ + I830Rec *pI830 = calloc(1, sizeof(I830Rec)); + int i; + + { + int dummy = ctx->shared.virtualWidth; + + switch (ctx->bpp / 8) { + case 1: dummy = (ctx->shared.virtualWidth + 127) & ~127; break; + case 2: dummy = (ctx->shared.virtualWidth + 31) & ~31; break; + case 3: + case 4: dummy = (ctx->shared.virtualWidth + 15) & ~15; break; + } + + ctx->shared.virtualWidth = dummy; + ctx->shared.Width = ctx->shared.virtualWidth; + } + + + for (i = 0; pitches[i] != 0; i++) { + if (pitches[i] >= ctx->shared.virtualWidth) { + ctx->shared.virtualWidth = pitches[i]; + break; + } + } + + ctx->driverPrivate = (void *)pI830; + + pI830->LpRing = calloc(1, sizeof(I830RingBuffer)); + pI830->Chipset = ctx->chipset; + pI830->LinearAddr = ctx->FBStart; + + if (!I830ScreenInit( ctx, pI830 )) + return 0; + + + return 1; +} + + +/** + * \brief The screen is being closed, so clean up any state and free any + * resources used by the DRI. + * + * \param ctx display handle. + * + * Unmaps the SAREA, closes the DRM device file descriptor and frees the driver + * private data. + */ +static void i830HaltFBDev( DRIDriverContext *ctx ) +{ + drmI830Sarea *pSAREAPriv; + I830Rec *pI830 = ctx->driverPrivate; + + if (pI830->irq) { + drmCtlUninstHandler(ctx->drmFD); + pI830->irq = 0; } + + I830CleanupDma(ctx); + + pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + + sizeof(drm_sarea_t)); + + I830DRIUnmapScreenRegions(ctx, pI830, pSAREAPriv); + drmUnmap( ctx->pSAREA, ctx->shared.SAREASize ); + drmClose(ctx->drmFD); + + if (ctx->driverPrivate) { + free(ctx->driverPrivate); + ctx->driverPrivate = 0; + } +} + + +extern void i810NotifyFocus( int ); + +/** + * \brief Exported driver interface for Mini GLX. + * + * \sa DRIDriverRec. + */ +const struct DRIDriverRec __driDriver = { + i830ValidateMode, + i830PostValidateMode, + i830InitFBDev, + i830HaltFBDev, + NULL,//I830EngineShutdown, + NULL, //I830EngineRestore, +#ifndef _EMBEDDED + 0, +#else + i810NotifyFocus, +#endif +}; -- cgit v1.2.3 From 7523f490b0166b00e24429c33f8958c3be72ccb5 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 09:50:49 -0600 Subject: define i915_surface struct --- src/mesa/pipe/i915simple/i915_surface.c | 30 +++++++++++++++++++++++++++++- 1 file changed, 29 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 73d33570f6..12b39bb1e4 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -31,8 +31,32 @@ #include "main/imports.h" +struct i915_surface +{ + struct pipe_surface surface; + /* anything else? */ +}; + + +static void +i915_get_tile(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) +{ + /* any need to get tiles from i915 surfaces? */ +} + + +static void +i915_put_tile(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, const GLfloat *p) +{ + /* any need to put tiles into i915 surfaces? */ +} + + + static struct pipe_surface * -i915_surface_alloc(struct pipe_context *pipe, GLenum format) +i915_surface_alloc(struct pipe_context *pipe, GLuint format) { struct i915_surface *surf = CALLOC_STRUCT(i915_surface); @@ -40,6 +64,10 @@ i915_surface_alloc(struct pipe_context *pipe, GLenum format) return NULL; surf->surface.format = format; + surf->surface.refcount = 1; + + surf->surface.get_tile = i915_get_tile; + surf->surface.put_tile = i915_put_tile; return &surf->surface; } -- cgit v1.2.3 From 40bc47f5780333e75c35a16bfec1f9670b2a646a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 09:52:35 -0600 Subject: replace draw_offset w/ 0 for now --- src/mesa/pipe/i915simple/i915_state_emit.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index dd7932bc0d..ba8cbdacf3 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -189,7 +189,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_RELOC(cbuf_region->buffer, I915_BUFFER_ACCESS_WRITE, - cbuf_region->draw_offset); + /*cbuf_region->draw_offset*/0); } /* What happens if no zbuf?? @@ -207,7 +207,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_RELOC(depth_region->buffer, I915_BUFFER_ACCESS_WRITE, - depth_region->draw_offset); + /*depth_region->draw_offset*/0); } -- cgit v1.2.3 From eb147ed775bedec9d849cdc817f208b5fe5eade3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 10:26:16 -0600 Subject: added PIPE_TEX_FACE_* --- src/mesa/pipe/p_defines.h | 22 ++++++++++++++-------- 1 file changed, 14 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index a3288134db..77512f8818 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -112,6 +112,19 @@ #define PIPE_STENCIL_OP_DECR_WRAP 6 #define PIPE_STENCIL_OP_INVERT 7 +/** Texture types */ +#define PIPE_TEXTURE_1D 0 +#define PIPE_TEXTURE_2D 1 +#define PIPE_TEXTURE_3D 2 +#define PIPE_TEXTURE_CUBE 3 + +#define PIPE_TEX_FACE_POS_X 0 +#define PIPE_TEX_FACE_NEG_X 1 +#define PIPE_TEX_FACE_POS_Y 2 +#define PIPE_TEX_FACE_NEG_Y 3 +#define PIPE_TEX_FACE_POS_Z 4 +#define PIPE_TEX_FACE_NEG_Z 5 + #define PIPE_TEX_WRAP_REPEAT 0 #define PIPE_TEX_WRAP_CLAMP 1 #define PIPE_TEX_WRAP_CLAMP_TO_EDGE 2 @@ -131,6 +144,7 @@ #define PIPE_TEX_COMPARE_NONE 0 #define PIPE_TEX_COMPARE_R_TO_TEXTURE 1 + /** * Texture/surface image formats (preliminary) */ @@ -155,14 +169,6 @@ #define PIPE_FORMAT_COUNT 18 /**< number of formats */ -/** - * Texture typess - */ -#define PIPE_TEXTURE_1D 0 -#define PIPE_TEXTURE_2D 1 -#define PIPE_TEXTURE_3D 2 -#define PIPE_TEXTURE_CUBE 3 - /** * Buffer mapping access modes */ -- cgit v1.2.3 From 34a48abd5ff82ce9748fc29191e35a0985d47c5f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 10:26:51 -0600 Subject: sketch out 1D/3D/cube texture sampling --- src/mesa/pipe/softpipe/sp_tex_sample.c | 156 ++++++++++++++++++++++++++++++++- 1 file changed, 154 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index de8e01ea07..df64c81769 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -338,6 +338,115 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, } +static GLuint +choose_cube_face(const GLfloat texcoord[4], GLfloat newCoord[4]) +{ + /* + major axis + direction target sc tc ma + ---------- ------------------------------- --- --- --- + +rx TEXTURE_CUBE_MAP_POSITIVE_X_EXT -rz -ry rx + -rx TEXTURE_CUBE_MAP_NEGATIVE_X_EXT +rz -ry rx + +ry TEXTURE_CUBE_MAP_POSITIVE_Y_EXT +rx +rz ry + -ry TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT +rx -rz ry + +rz TEXTURE_CUBE_MAP_POSITIVE_Z_EXT +rx -ry rz + -rz TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT -rx -ry rz + */ + const GLfloat rx = texcoord[0]; + const GLfloat ry = texcoord[1]; + const GLfloat rz = texcoord[2]; + const GLfloat arx = FABSF(rx), ary = FABSF(ry), arz = FABSF(rz); + GLuint face; + GLfloat sc, tc, ma; + + if (arx > ary && arx > arz) { + if (rx >= 0.0F) { + face = PIPE_TEX_FACE_POS_X; + sc = -rz; + tc = -ry; + ma = arx; + } + else { + face = PIPE_TEX_FACE_NEG_X; + sc = rz; + tc = -ry; + ma = arx; + } + } + else if (ary > arx && ary > arz) { + if (ry >= 0.0F) { + face = PIPE_TEX_FACE_POS_Y; + sc = rx; + tc = rz; + ma = ary; + } + else { + face = PIPE_TEX_FACE_NEG_Y; + sc = rx; + tc = -rz; + ma = ary; + } + } + else { + if (rz > 0.0F) { + face = PIPE_TEX_FACE_POS_Z; + sc = rx; + tc = -ry; + ma = arz; + } + else { + face = PIPE_TEX_FACE_NEG_Z; + sc = -rx; + tc = -ry; + ma = arz; + } + } + + newCoord[0] = ( sc / ma + 1.0F ) * 0.5F; + newCoord[1] = ( tc / ma + 1.0F ) * 0.5F; + + return face; +} + + +static void +sp_get_sample_1d(struct tgsi_sampler *sampler, + const GLfloat strq[4], GLfloat rgba[4]) +{ + struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; + struct pipe_surface *ps + = pipe->get_tex_surface(pipe, sampler->texture, 0, 0, 0); + + switch (sampler->state->min_filter) { + case PIPE_TEX_FILTER_NEAREST: + { + GLint x; + x = nearest_texcoord(sampler->state->wrap_s, strq[0], + sampler->texture->width0); + ps->get_tile(ps, x, 0, 1, 1, rgba); + } + break; + case PIPE_TEX_FILTER_LINEAR: + { + GLfloat t0[4], t1[4]; + GLint x0, x1; + GLfloat a; + linear_texcoord(sampler->state->wrap_s, strq[0], + sampler->texture->width0, &x0, &x1, &a); + ps->get_tile(ps, x0, 0, 1, 1, t0); + ps->get_tile(ps, x1, 0, 1, 1, t1); + + rgba[0] = LERP(a, t0[0], t1[0]); + rgba[1] = LERP(a, t0[1], t1[1]); + rgba[2] = LERP(a, t0[2], t1[2]); + rgba[3] = LERP(a, t0[3], t1[3]); + } + break; + default: + assert(0); + } +} + /** * Called via tgsi_sampler::get_sample() @@ -352,8 +461,8 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, * a new tgsi_sampler object for each state combo it finds.... */ void -sp_get_sample(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat rgba[4]) +sp_get_sample_2d(struct tgsi_sampler *sampler, + const GLfloat strq[4], GLfloat rgba[4]) { struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; struct pipe_surface *ps @@ -394,3 +503,46 @@ sp_get_sample(struct tgsi_sampler *sampler, assert(0); } } + + +static void +sp_get_sample_3d(struct tgsi_sampler *sampler, + const GLfloat strq[4], GLfloat rgba[4]) +{ + /* get/map pipe_surfaces corresponding to 3D tex slices */ +} + + +static void +sp_get_sample_cube(struct tgsi_sampler *sampler, + const GLfloat strq[4], GLfloat rgba[4]) +{ + GLfloat st[4]; + GLuint face = choose_cube_face(strq, st); + + /* get/map surface corresponding to the face */ +} + + +void +sp_get_sample(struct tgsi_sampler *sampler, + const GLfloat strq[4], GLfloat rgba[4]) +{ + switch (sampler->texture->target) { + case GL_TEXTURE_1D: + sp_get_sample_1d(sampler, strq, rgba); + break; + case GL_TEXTURE_2D: + sp_get_sample_2d(sampler, strq, rgba); + break; + case GL_TEXTURE_3D: + sp_get_sample_3d(sampler, strq, rgba); + break; + case GL_TEXTURE_CUBE_MAP: + sp_get_sample_cube(sampler, strq, rgba); + break; + default: + assert(0); + } +} + -- cgit v1.2.3 From e12810d92ffb3547680b227bf88937c03018112b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 11:09:02 -0600 Subject: add lambda param to get_sample() --- src/mesa/pipe/softpipe/sp_tex_sample.c | 20 ++++++++++---------- src/mesa/pipe/softpipe/sp_tex_sample.h | 2 +- 2 files changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index df64c81769..1d274d5038 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -411,7 +411,7 @@ choose_cube_face(const GLfloat texcoord[4], GLfloat newCoord[4]) static void sp_get_sample_1d(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat rgba[4]) + const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]) { struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; struct pipe_surface *ps @@ -460,9 +460,9 @@ sp_get_sample_1d(struct tgsi_sampler *sampler, * The update_samplers() function in st_atom_sampler.c could create * a new tgsi_sampler object for each state combo it finds.... */ -void +static void sp_get_sample_2d(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat rgba[4]) + const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]) { struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; struct pipe_surface *ps @@ -507,7 +507,7 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, static void sp_get_sample_3d(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat rgba[4]) + const GLfloat strq[4], GLfloat lamba, GLfloat rgba[4]) { /* get/map pipe_surfaces corresponding to 3D tex slices */ } @@ -515,7 +515,7 @@ sp_get_sample_3d(struct tgsi_sampler *sampler, static void sp_get_sample_cube(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat rgba[4]) + const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]) { GLfloat st[4]; GLuint face = choose_cube_face(strq, st); @@ -526,20 +526,20 @@ sp_get_sample_cube(struct tgsi_sampler *sampler, void sp_get_sample(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat rgba[4]) + const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]) { switch (sampler->texture->target) { case GL_TEXTURE_1D: - sp_get_sample_1d(sampler, strq, rgba); + sp_get_sample_1d(sampler, strq, lambda, rgba); break; case GL_TEXTURE_2D: - sp_get_sample_2d(sampler, strq, rgba); + sp_get_sample_2d(sampler, strq, lambda, rgba); break; case GL_TEXTURE_3D: - sp_get_sample_3d(sampler, strq, rgba); + sp_get_sample_3d(sampler, strq, lambda, rgba); break; case GL_TEXTURE_CUBE_MAP: - sp_get_sample_cube(sampler, strq, rgba); + sp_get_sample_cube(sampler, strq, lambda, rgba); break; default: assert(0); diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.h b/src/mesa/pipe/softpipe/sp_tex_sample.h index d1d40e4a2f..55ae49094b 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.h +++ b/src/mesa/pipe/softpipe/sp_tex_sample.h @@ -7,7 +7,7 @@ struct tgsi_sampler; extern void sp_get_sample(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat rgba[4]); + const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]); #endif /* SP_TEX_SAMPLE_H */ -- cgit v1.2.3 From 7e78b9e4d0117d7438fa1a07dff2bc261a9100d9 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 11:09:53 -0600 Subject: rewrite texture sampling code (smaller, simpler) --- src/mesa/pipe/tgsi/core/tgsi_exec.c | 369 ++++++++++++------------------------ src/mesa/pipe/tgsi/core/tgsi_exec.h | 4 +- 2 files changed, 127 insertions(+), 246 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 87dab9956f..64d413b1b7 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -6,7 +6,6 @@ #if MESA #include "main/context.h" #include "main/macros.h" -#include "main/colormac.h" #endif #define TILE_BOTTOM_LEFT 0 @@ -65,7 +64,7 @@ tgsi_exec_machine_init( struct tgsi_exec_machine *mach, struct tgsi_token *tokens, GLuint numSamplers, - const struct tgsi_sampler *samplers) + struct tgsi_sampler *samplers) { GLuint i, k; struct tgsi_parse_context parse; @@ -1034,209 +1033,94 @@ exec_kil (struct tgsi_exec_machine *mach, mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; } -#if MESA -/* - * Fetch a texel using S texture coordinate. - */ -static void -fetch_texel_1d( GLcontext *ctx, - const struct tgsi_sampler *sampler, - const union tgsi_exec_channel *s, - GLuint unit, - union tgsi_exec_channel *r, - union tgsi_exec_channel *g, - union tgsi_exec_channel *b, - union tgsi_exec_channel *a ) -{ - GLuint fragment_index; - GLfloat stpq[4][4]; - GLfloat lambdas[4]; - for (fragment_index = 0; fragment_index < 4; fragment_index++) { - stpq[fragment_index][0] = s->f[fragment_index]; - } +static GLfloat +compute_lambda(struct tgsi_sampler *sampler, + const union tgsi_exec_channel *s, + const union tgsi_exec_channel *t, + const union tgsi_exec_channel *p) +{ + GLfloat rho, lambda; - if (sampler->state->min_filter != sampler->state->mag_filter) { + assert(s); + { GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; - GLfloat rho, lambda; - dsdx = FABSF(dsdx); dsdy = FABSF(dsdy); - rho = MAX2(dsdx, dsdy) * sampler->texture->width0; - - lambda = LOG2(rho); - lambda += sampler->state->lod_bias; - lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); - - /* XXX: Use the same lambda value throughout the tile. Could - * end up with four unique values by recalculating partial - * derivs in the other row and column, and calculating lambda - * using the dx and dy values appropriate for each fragment in - * the tile. - */ - lambdas[0] = - lambdas[1] = - lambdas[2] = - lambdas[3] = lambda; } - - for (fragment_index = 0; fragment_index < 4; fragment_index++) { - GLfloat rgba[4]; - sampler->get_sample(sampler, stpq[fragment_index], rgba); - r->f[fragment_index] = rgba[0]; - g->f[fragment_index] = rgba[1]; - b->f[fragment_index] = rgba[2]; - a->f[fragment_index] = rgba[3]; - } -} - -/* - * Fetch a texel using ST texture coordinates. - */ -static void -fetch_texel_2d( GLcontext *ctx, - const struct tgsi_sampler *sampler, - const union tgsi_exec_channel *s, - const union tgsi_exec_channel *t, - GLuint unit, - union tgsi_exec_channel *r, - union tgsi_exec_channel *g, - union tgsi_exec_channel *b, - union tgsi_exec_channel *a ) -{ - GLuint fragment_index; - GLfloat stpq[4][4]; - GLfloat lambdas[4]; - - for (fragment_index = 0; fragment_index < 4; fragment_index++) { - stpq[fragment_index][0] = s->f[fragment_index]; - stpq[fragment_index][1] = t->f[fragment_index]; - } - - if (sampler->state->min_filter != sampler->state->mag_filter) { - GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; - GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; - + if (t) { GLfloat dtdx = t->f[TILE_BOTTOM_RIGHT] - t->f[TILE_BOTTOM_LEFT]; GLfloat dtdy = t->f[TILE_TOP_LEFT] - t->f[TILE_BOTTOM_LEFT]; - - GLfloat maxU, maxV, rho, lambda; - - dsdx = FABSF( dsdx ); - dsdy = FABSF( dsdy ); - dtdx = FABSF( dtdx ); - dtdy = FABSF( dtdy ); - - maxU = MAX2( dsdx, dsdy ) * sampler->texture->width0; - maxV = MAX2( dtdx, dtdy ) * sampler->texture->height0; - - rho = MAX2( maxU, maxV ); - - lambda = LOG2( rho ); - - lambda += sampler->state->lod_bias; - lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); - - /* XXX: Use the same lambda value throughout the tile. Could - * end up with four unique values by recalculating partial - * derivs in the other row and column, and calculating lambda - * using the dx and dy values appropriate for each fragment in - * the tile. - */ - lambdas[0] = - lambdas[1] = - lambdas[2] = - lambdas[3] = lambda; + GLfloat max; + dtdx = FABSF(dtdx); + dtdy = FABSF(dtdy); + max = MAX2(dtdx, dtdy) * sampler->texture->height0; + rho = MAX2(rho, max); } - - for (fragment_index = 0; fragment_index < 4; fragment_index++) { - GLfloat rgba[4]; - sampler->get_sample(sampler, stpq[fragment_index], rgba); - r->f[fragment_index] = rgba[0]; - g->f[fragment_index] = rgba[1]; - b->f[fragment_index] = rgba[2]; - a->f[fragment_index] = rgba[3]; + if (p) { + GLfloat dpdx = p->f[TILE_BOTTOM_RIGHT] - p->f[TILE_BOTTOM_LEFT]; + GLfloat dpdy = p->f[TILE_TOP_LEFT] - p->f[TILE_BOTTOM_LEFT]; + GLfloat max; + dpdx = FABSF(dpdx); + dpdy = FABSF(dpdy); + max = MAX2(dpdx, dpdy) * sampler->texture->depth0; + rho = MAX2(rho, max); } + + lambda = LOG2(rho); + return lambda; } + /* * Fetch a texel using STR texture coordinates. */ static void -fetch_texel_3d( GLcontext *ctx, - const struct tgsi_sampler *sampler, - const union tgsi_exec_channel *s, - const union tgsi_exec_channel *t, - const union tgsi_exec_channel *p, - GLuint unit, - union tgsi_exec_channel *r, - union tgsi_exec_channel *g, - union tgsi_exec_channel *b, - union tgsi_exec_channel *a ) +fetch_texel( struct tgsi_sampler *sampler, + GLfloat lambda, + const union tgsi_exec_channel *s, + const union tgsi_exec_channel *t, + const union tgsi_exec_channel *p, + GLuint unit, + union tgsi_exec_channel *r, + union tgsi_exec_channel *g, + union tgsi_exec_channel *b, + union tgsi_exec_channel *a ) { GLuint fragment_index; GLfloat stpq[4][4]; - GLfloat lambdas[4]; for (fragment_index = 0; fragment_index < 4; fragment_index++) { stpq[fragment_index][0] = s->f[fragment_index]; - stpq[fragment_index][1] = t->f[fragment_index]; - stpq[fragment_index][2] = p->f[fragment_index]; + if (t) + stpq[fragment_index][1] = t->f[fragment_index]; + if (p) + stpq[fragment_index][2] = p->f[fragment_index]; } - if (sampler->state->min_filter != sampler->state->mag_filter) { - GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; - GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; - - GLfloat dtdx = t->f[TILE_BOTTOM_RIGHT] - t->f[TILE_BOTTOM_LEFT]; - GLfloat dtdy = t->f[TILE_TOP_LEFT] - t->f[TILE_BOTTOM_LEFT]; - - GLfloat dpdx = p->f[TILE_BOTTOM_RIGHT] - p->f[TILE_BOTTOM_LEFT]; - GLfloat dpdy = p->f[TILE_TOP_LEFT] - p->f[TILE_BOTTOM_LEFT]; - - GLfloat maxU, maxV, maxW, rho, lambda; - - dsdx = FABSF(dsdx); - dsdy = FABSF(dsdy); - dtdx = FABSF(dtdx); - dtdy = FABSF(dtdy); - dpdx = FABSF(dpdx); - dpdy = FABSF(dpdy); - - maxU = MAX2(dsdx, dsdy) * sampler->texture->width0; - maxV = MAX2(dtdx, dtdy) * sampler->texture->height0; - maxW = MAX2(dpdx, dpdy) * sampler->texture->depth0; + lambda += sampler->state->lod_bias; + lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); - rho = MAX2(maxU, MAX2(maxV, maxW)); - - lambda = LOG2(rho); - lambda += sampler->state->lod_bias; - lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); - - /* XXX: Use the same lambda value throughout the tile. Could - * end up with four unique values by recalculating partial - * derivs in the other row and column, and calculating lambda - * using the dx and dy values appropriate for each fragment in - * the tile. - */ - lambdas[0] = - lambdas[1] = - lambdas[2] = - lambdas[3] = lambda; - } + /* XXX: Use the same lambda value throughout the tile. Could + * end up with four unique values by recalculating partial + * derivs in the other row and column, and calculating lambda + * using the dx and dy values appropriate for each fragment in + * the tile. + */ for (fragment_index = 0; fragment_index < 4; fragment_index++) { GLfloat rgba[4]; - sampler->get_sample(sampler, stpq[fragment_index], rgba); + sampler->get_sample(sampler, stpq[fragment_index], lambda, rgba); r->f[fragment_index] = rgba[0]; g->f[fragment_index] = rgba[1]; b->f[fragment_index] = rgba[2]; a->f[fragment_index] = rgba[3]; } } -#endif + + static void exec_instruction( @@ -1245,10 +1129,7 @@ exec_instruction( struct tgsi_exec_labels *labels, GLuint *programCounter ) { -#if MESA - GET_CURRENT_CONTEXT(ctx); -#endif - GLuint chan_index, unit; + GLuint chan_index; union tgsi_exec_channel r[8]; switch (inst->Instruction.Opcode) { @@ -1764,99 +1645,99 @@ exec_instruction( break; case TGSI_OPCODE_TEX: - unit = inst->FullSrcRegisters[1].SrcRegister.Index; - switch (inst->InstructionExtTexture.Texture) { - case TGSI_TEXTURE_1D: + { + const GLuint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + GLfloat lambda; + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: - FETCH(&r[0], 0, CHAN_X); + FETCH(&r[0], 0, CHAN_X); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[1], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[1] ); - break; + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[1], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[1] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } - case TGSI_EXTSWIZZLE_ONE: + lambda = compute_lambda(&mach->Samplers[unit], &r[0], NULL, NULL); + fetch_texel(&mach->Samplers[unit], lambda, + &r[0], NULL, NULL, + inst->FullSrcRegisters[1].SrcRegister.Index, + &r[0], &r[1], &r[2], &r[3]); break; - default: - assert (0); - } -#if MESA - fetch_texel_1d (ctx, - &mach->Samplers[unit], - &r[0], - inst->FullSrcRegisters[1].SrcRegister.Index, - &r[0], &r[1], &r[2], &r[3]); -#endif - break; + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[2], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[2] ); + micro_div( &r[1], &r[1], &r[2] ); + break; - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[2], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[2] ); - micro_div( &r[1], &r[1], &r[2] ); - break; + case TGSI_EXTSWIZZLE_ONE: + break; - case TGSI_EXTSWIZZLE_ONE: + default: + assert (0); + } + + lambda = compute_lambda(&mach->Samplers[unit], &r[0], &r[1], NULL); + fetch_texel(&mach->Samplers[unit], + lambda, + &r[0], &r[1], NULL, + inst->FullSrcRegisters[1].SrcRegister.Index, + &r[0], &r[1], &r[2], &r[3]); break; - default: - assert (0); - } + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: -#if MESA - fetch_texel_2d (ctx, - &mach->Samplers[unit], - &r[0], &r[1], - inst->FullSrcRegisters[1].SrcRegister.Index, - &r[0], &r[1], &r[2], &r[3]); -#endif - break; + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); - case TGSI_TEXTURE_3D: - case TGSI_TEXTURE_CUBE: + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); + case TGSI_EXTSWIZZLE_ONE: + break; - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - break; + default: + assert (0); + } - case TGSI_EXTSWIZZLE_ONE: + lambda = compute_lambda(&mach->Samplers[unit], &r[0], &r[1], &r[2]); + fetch_texel(&mach->Samplers[unit], + lambda, + &r[0], &r[1], &r[2], + inst->FullSrcRegisters[1].SrcRegister.Index, + &r[0], &r[1], &r[2], &r[3]); break; default: assert (0); } -#if MESA - fetch_texel_3d (ctx, - &mach->Samplers[unit], - &r[0], &r[1], &r[2], - inst->FullSrcRegisters[1].SrcRegister.Index, - &r[0], &r[1], &r[2], &r[3]); -#endif - break; - - default: - assert (0); - } - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[chan_index], 0, chan_index ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } } break; diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index 1c714ad553..2ad4b965bd 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -26,7 +26,7 @@ struct tgsi_sampler const struct pipe_sampler_state *state; struct pipe_mipmap_tree *texture; void (*get_sample)(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat rgba[4]); + const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]); void *pipe; /*XXX temporary*/ }; @@ -136,7 +136,7 @@ tgsi_exec_machine_init( struct tgsi_exec_machine *mach, struct tgsi_token *tokens, GLuint numSamplers, - const struct tgsi_sampler *samplers); + struct tgsi_sampler *samplers); void tgsi_exec_prepare( -- cgit v1.2.3 From dbb297fac6a9637cc7ee7ab2cf40f196048ba626 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 11:10:37 +0100 Subject: Hook up i915simple. Doesn't compile. --- src/mesa/Makefile | 4 +- src/mesa/drivers/dri/intel_winsys/Makefile | 3 +- src/mesa/drivers/dri/intel_winsys/intel_context.c | 45 ++-- src/mesa/drivers/dri/intel_winsys/intel_context.h | 8 - src/mesa/drivers/dri/intel_winsys/intel_pipe.h | 42 ++++ .../dri/intel_winsys/intel_pipe_i915simple.c | 248 +++++++++++++++++++++ .../drivers/dri/intel_winsys/intel_pipe_softpipe.c | 184 +++++++++++++++ src/mesa/drivers/dri/intel_winsys/intel_softpipe.c | 184 --------------- src/mesa/pipe/Makefile | 2 +- src/mesa/pipe/i915simple/i915_buffer.c | 6 +- src/mesa/pipe/i915simple/i915_regions.c | 2 +- src/mesa/pipe/i915simple/i915_state.c | 5 +- src/mesa/pipe/i915simple/i915_state_emit.c | 4 +- src/mesa/pipe/i915simple/i915_winsys.h | 15 +- 14 files changed, 521 insertions(+), 231 deletions(-) create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_pipe.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_softpipe.c (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 7d4bd22203..9dde96ee21 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -150,7 +150,7 @@ depend: $(ALL_SOURCES) subdirs: @ (cd x86 ; $(MAKE)) @ (cd x86-64 ; $(MAKE)) - (cd pipe/softpipe ; $(MAKE)) + (cd pipe ; $(MAKE)) install: default $(INSTALL) -d $(INSTALL_DIR)/include/GL @@ -183,7 +183,7 @@ clean: (cd drivers/dri && $(MAKE) clean) (cd x86 && $(MAKE) clean) (cd x86-64 && $(MAKE) clean) - (cd pipe/softpipe ; $(MAKE) clean ) + (cd pipe ; $(MAKE) clean ) include depend diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index e4e9cf17b0..c9dc15e088 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -10,7 +10,8 @@ PIPE_DRIVERS = \ $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a DRIVER_SOURCES = \ - intel_softpipe.c \ + intel_pipe_i915simple.c \ + intel_pipe_softpipe.c \ intel_batchbuffer.c \ intel_tex_layout.c \ intel_buffers.c \ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index e6f0d4a0fa..d748c08f3c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -47,13 +47,10 @@ #include "i830_dri.h" #include "intel_buffers.h" -/*#include "intel_tex.h"*/ +#include "intel_pipe.h" #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" -/* -#include "intel_buffer_objects.h" -*/ #include "intel_fbo.h" #include "intel_tex_layout.h" @@ -375,20 +372,34 @@ intelCreateContext(const __GLcontextModes * mesaVis, /* * Pipe-related setup */ - st_create_context( &intel->ctx, - intel_create_softpipe( intel ) ); - - /* KW: Not sure I like this - we should only be talking to the - * state_tracker. The pipe code will need some way of talking to - * us, eg for batchbuffer ioctls, and there will need to be a - * buffer manager interface. So, this is a temporary hack, right? - * BP: Yes, a temporary hack so we can make jumps between old/new code. - */ - intel->pipe = intel->ctx.st->pipe; -// intel->pipe->screen = intelScreen; -// intel->pipe->glctx = ctx; -// intel_init_region_functions(intel->pipe); + if (getenv("INTEL_SOFTPIPE")) { + intel->pipe = intel_create_softpipe( intel ); + } + else { + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + intel->pipe = intel_create_i915simple( intel ); + default: + _mesa_printf("Unknown PCIID %x in %s, using software driver\n", + intel->intelScreen->deviceID, __FUNCTION__); + intel->pipe = intel_create_softpipe( intel ); + break; + } + } + + st_create_context( &intel->ctx, intel->pipe ); + + + /* TODO: Push this down into the pipe driver: + */ switch (intel->intelScreen->deviceID) { case PCI_CHIP_I945_G: case PCI_CHIP_I945_GM: diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 1a4e7ca1cd..727f3b7474 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -194,14 +194,6 @@ extern void intelFinish(GLcontext * ctx); extern void intelFlush(GLcontext * ctx); -/* ================================================================ - * intel_softpipe.c: - */ - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel ); - - #define MI_BATCH_BUFFER_END (0xA<<23) #define FALLBACK( ctx, bit, mode ) diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe.h b/src/mesa/drivers/dri/intel_winsys/intel_pipe.h new file mode 100644 index 0000000000..5d3e25aea2 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_PIPE_H +#define INTEL_PIPE_H + +struct intel_context; +struct pipe_context; + + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel ); + +struct pipe_context * +intel_create_i915simple( struct intel_context *intel ); + + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c new file mode 100644 index 0000000000..b456550a8c --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c @@ -0,0 +1,248 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +#include "dri_bufpool.h" +#include "dri_bufmgr.h" + +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_pipe.h" + +#include "pipe/i915simple/i915_winsys.h" + + +struct intel_i915_winsys { + struct i915_winsys winsys; + struct intel_context *intel; +}; + + +/* Turn a i915simple winsys into an intel/i915simple winsys: + */ +static inline struct intel_i915_winsys * +intel_i915_winsys( struct i915_winsys *sws ) +{ + return (struct intel_i915_winsys *)sws; +} + + + +/* Many of the winsys's are probably going to have a similar + * buffer-manager interface, as something almost identical is + * currently exposed in the pipe interface. Probably want to avoid + * endless repetition of this code somehow. + */ +static inline struct _DriBufferObject * +dri_bo( struct pipe_buffer_handle *bo ) +{ + return (struct _DriBufferObject *)bo; +} + +static inline struct pipe_buffer_handle * +pipe_bo( struct _DriBufferObject *bo ) +{ + return (struct pipe_buffer_handle *)bo; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void *intel_i915_buffer_map(struct i915_winsys *sws, + struct pipe_buffer_handle *buf ) +{ + return driBOMap( dri_bo(buf), + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0 ); +} + +static void intel_i915_buffer_unmap(struct i915_winsys *sws, + struct pipe_buffer_handle *buf) +{ + driBOUnmap( dri_bo(buf) ); +} + + +static struct pipe_buffer_handle * +intel_i915_buffer_reference(struct i915_winsys *sws, + struct pipe_buffer_handle *buf) +{ + return pipe_bo( driBOReference( dri_bo(buf) ) ); +} + +static void intel_i915_buffer_unreference(struct i915_winsys *sws, + struct pipe_buffer_handle **buf) +{ + if (*buf) { + driBOUnReference( dri_bo(*buf) ); + *buf = NULL; + } +} + +/* Grabs the hardware lock! + */ +static void intel_i915_buffer_data(struct i915_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + LOCK_HARDWARE( intel ); + driBOData( dri_bo(buf), size, data, 0 ); + UNLOCK_HARDWARE( intel ); +} + +static void intel_i915_buffer_subdata(struct i915_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) +{ + driBOSubData( dri_bo(buf), offset, size, data ); +} + +static void intel_i915_buffer_get_subdata(struct i915_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) +{ + driBOGetSubData( dri_bo(buf), offset, size, data ); +} + +/* I915simple has no concept of pools. We choose the tex/region pool + * for all buffers. + */ +static struct pipe_buffer_handle * +intel_i915_buffer_create(struct i915_winsys *sws, + unsigned alignment) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + struct _DriBufferObject *buffer; + + LOCK_HARDWARE( intel ); + driGenBuffers( intel->intelScreen->regionPool, + "i915simple buffer", 1, &buffer, alignment, 0, 0 ); + UNLOCK_HARDWARE( intel ); + + return pipe_bo(buffer); +} + + +/* Simple batchbuffer interface: + */ + +static unsigned *intel_i915_batch_start( struct i915_winsys *sws, + unsigned dwords, + unsigned relocs ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + /* XXX: check relocs. + */ + if (intel_batchbuffer_space( intel->batch ) < dwords * 4) { + /* XXX: Hmm, the driver can't really do much with this pointer: + */ + //return intel->batch->ptr; + return (unsigned *)~0; + } + else + return NULL; +} + +static void intel_i915_batch_dword( struct i915_winsys *sws, + unsigned dword ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + intel_batchbuffer_emit_dword( intel->batch, dword ); +} + +static void intel_i915_batch_reloc( struct i915_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned access_flags, + unsigned delta ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + unsigned flags = 0; + unsigned mask = 0; + + intel_batchbuffer_emit_reloc( intel->batch, + dri_bo( buf ), + flags, mask, + delta ); +} + +static void intel_i915_batch_flush( struct i915_winsys *sws ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + intel_batchbuffer_flush( intel->batch ); +} + +static void intel_i915_batch_wait_idle( struct i915_winsys *sws ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + if (intel->batch->last_fence) { + driFenceFinish(intel->batch->last_fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); + driFenceUnReference(intel->batch->last_fence); + intel->batch->last_fence = NULL; + } +} + +struct pipe_context * +intel_create_i915simple( struct intel_context *intel ) +{ + struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); + + /* Fill in this struct with callbacks that i915simple will need to + * communicate with the window system, buffer manager, etc. + */ + iws->winsys.buffer_create = intel_i915_buffer_create; + iws->winsys.buffer_map = intel_i915_buffer_map; + iws->winsys.buffer_unmap = intel_i915_buffer_unmap; + iws->winsys.buffer_reference = intel_i915_buffer_reference; + iws->winsys.buffer_unreference = intel_i915_buffer_unreference; + iws->winsys.buffer_data = intel_i915_buffer_data; + iws->winsys.buffer_subdata = intel_i915_buffer_subdata; + iws->winsys.buffer_get_subdata = intel_i915_buffer_get_subdata; + iws->winsys.batch_start = intel_i915_batch_start; + iws->winsys.batch_dword = intel_i915_batch_dword; + iws->winsys.batch_reloc = intel_i915_batch_reloc; + iws->winsys.batch_flush = intel_i915_batch_flush; + iws->winsys.batch_wait_idle = intel_i915_batch_wait_idle; + iws->intel = intel; + + /* Create the i915simple context: + */ + return i915_create( &iws->winsys ); +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c new file mode 100644 index 0000000000..ef47744358 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c @@ -0,0 +1,184 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +#include "dri_bufpool.h" +#include "dri_bufmgr.h" + +/* #include "errno.h" */ +/* #include "string.h" */ +/* #include "imports.h" */ + +#include "intel_context.h" + +#include "pipe/softpipe/sp_winsys.h" + + +struct intel_softpipe_winsys { + struct softpipe_winsys sws; + struct intel_context *intel; +}; + + + +/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static inline struct _DriBufferObject * +dri_bo( struct pipe_buffer_handle *bo ) +{ + return (struct _DriBufferObject *)bo; +} + +static inline struct pipe_buffer_handle * +pipe_bo( struct _DriBufferObject *bo ) +{ + return (struct pipe_buffer_handle *)bo; +} + +/* Turn a softpipe winsys into an intel/softpipe winsys: + */ +static inline struct intel_softpipe_winsys * +intel_softpipe_winsys( struct softpipe_winsys *sws ) +{ + return (struct intel_softpipe_winsys *)sws; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void *intel_buffer_map(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf ) +{ + return driBOMap( dri_bo(buf), + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0 ); +} + +static void intel_buffer_unmap(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf) +{ + driBOUnmap( dri_bo(buf) ); +} + + +static struct pipe_buffer_handle * +intel_buffer_reference(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf) +{ + return pipe_bo( driBOReference( dri_bo(buf) ) ); +} + +static void intel_buffer_unreference(struct softpipe_winsys *sws, + struct pipe_buffer_handle **buf) +{ + if (*buf) { + driBOUnReference( dri_bo(*buf) ); + *buf = NULL; + } +} + +/* Grabs the hardware lock! + */ +static void intel_buffer_data(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct intel_context *intel = intel_softpipe_winsys(sws)->intel; + + LOCK_HARDWARE( intel ); + driBOData( dri_bo(buf), size, data, 0 ); + UNLOCK_HARDWARE( intel ); +} + +static void intel_buffer_subdata(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) +{ + driBOSubData( dri_bo(buf), offset, size, data ); +} + +static void intel_buffer_get_subdata(struct softpipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) +{ + driBOGetSubData( dri_bo(buf), offset, size, data ); +} + +/* Softpipe has no concept of pools. We choose the tex/region pool + * for all buffers. + */ +static struct pipe_buffer_handle * +intel_create_buffer(struct softpipe_winsys *sws, + unsigned alignment) +{ + struct intel_context *intel = intel_softpipe_winsys(sws)->intel; + struct _DriBufferObject *buffer; + + LOCK_HARDWARE( intel ); + driGenBuffers( intel->intelScreen->regionPool, + "softpipe buffer", 1, &buffer, alignment, 0, 0 ); + UNLOCK_HARDWARE( intel ); + + return pipe_bo(buffer); +} + + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel ) +{ + struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); + + /* Fill in this struct with callbacks that softpipe will need to + * communicate with the window system, buffer manager, etc. + * + * Softpipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + isws->sws.create_buffer = intel_create_buffer; + isws->sws.buffer_map = intel_buffer_map; + isws->sws.buffer_unmap = intel_buffer_unmap; + isws->sws.buffer_reference = intel_buffer_reference; + isws->sws.buffer_unreference = intel_buffer_unreference; + isws->sws.buffer_data = intel_buffer_data; + isws->sws.buffer_subdata = intel_buffer_subdata; + isws->sws.buffer_get_subdata = intel_buffer_get_subdata; + isws->intel = intel; + + /* Create the softpipe context: + */ + return softpipe_create( &isws->sws ); +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_softpipe.c deleted file mode 100644 index ef47744358..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_softpipe.c +++ /dev/null @@ -1,184 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include -#include "dri_bufpool.h" -#include "dri_bufmgr.h" - -/* #include "errno.h" */ -/* #include "string.h" */ -/* #include "imports.h" */ - -#include "intel_context.h" - -#include "pipe/softpipe/sp_winsys.h" - - -struct intel_softpipe_winsys { - struct softpipe_winsys sws; - struct intel_context *intel; -}; - - - -/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque - * buffer pointer... - */ -static inline struct _DriBufferObject * -dri_bo( struct pipe_buffer_handle *bo ) -{ - return (struct _DriBufferObject *)bo; -} - -static inline struct pipe_buffer_handle * -pipe_bo( struct _DriBufferObject *bo ) -{ - return (struct pipe_buffer_handle *)bo; -} - -/* Turn a softpipe winsys into an intel/softpipe winsys: - */ -static inline struct intel_softpipe_winsys * -intel_softpipe_winsys( struct softpipe_winsys *sws ) -{ - return (struct intel_softpipe_winsys *)sws; -} - - -/* Most callbacks map direcly onto dri_bufmgr operations: - */ -static void *intel_buffer_map(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf ) -{ - return driBOMap( dri_bo(buf), - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0 ); -} - -static void intel_buffer_unmap(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf) -{ - driBOUnmap( dri_bo(buf) ); -} - - -static struct pipe_buffer_handle * -intel_buffer_reference(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf) -{ - return pipe_bo( driBOReference( dri_bo(buf) ) ); -} - -static void intel_buffer_unreference(struct softpipe_winsys *sws, - struct pipe_buffer_handle **buf) -{ - if (*buf) { - driBOUnReference( dri_bo(*buf) ); - *buf = NULL; - } -} - -/* Grabs the hardware lock! - */ -static void intel_buffer_data(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ) -{ - struct intel_context *intel = intel_softpipe_winsys(sws)->intel; - - LOCK_HARDWARE( intel ); - driBOData( dri_bo(buf), size, data, 0 ); - UNLOCK_HARDWARE( intel ); -} - -static void intel_buffer_subdata(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) -{ - driBOSubData( dri_bo(buf), offset, size, data ); -} - -static void intel_buffer_get_subdata(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) -{ - driBOGetSubData( dri_bo(buf), offset, size, data ); -} - -/* Softpipe has no concept of pools. We choose the tex/region pool - * for all buffers. - */ -static struct pipe_buffer_handle * -intel_create_buffer(struct softpipe_winsys *sws, - unsigned alignment) -{ - struct intel_context *intel = intel_softpipe_winsys(sws)->intel; - struct _DriBufferObject *buffer; - - LOCK_HARDWARE( intel ); - driGenBuffers( intel->intelScreen->regionPool, - "softpipe buffer", 1, &buffer, alignment, 0, 0 ); - UNLOCK_HARDWARE( intel ); - - return pipe_bo(buffer); -} - - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel ) -{ - struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); - - /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. - * - * Softpipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate _DriBufferObject. - */ - isws->sws.create_buffer = intel_create_buffer; - isws->sws.buffer_map = intel_buffer_map; - isws->sws.buffer_unmap = intel_buffer_unmap; - isws->sws.buffer_reference = intel_buffer_reference; - isws->sws.buffer_unreference = intel_buffer_unreference; - isws->sws.buffer_data = intel_buffer_data; - isws->sws.buffer_subdata = intel_buffer_subdata; - isws->sws.buffer_get_subdata = intel_buffer_get_subdata; - isws->intel = intel; - - /* Create the softpipe context: - */ - return softpipe_create( &isws->sws ); -} diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile index ab7e09c991..f5e884e3a0 100644 --- a/src/mesa/pipe/Makefile +++ b/src/mesa/pipe/Makefile @@ -1,6 +1,6 @@ default: - cd .. ; make cd softpipe ; make + cd i915simple ; make clean: rm -f `find . -name \*.[oa]` \ No newline at end of file diff --git a/src/mesa/pipe/i915simple/i915_buffer.c b/src/mesa/pipe/i915simple/i915_buffer.c index eaaf3c1beb..680213182b 100644 --- a/src/mesa/pipe/i915simple/i915_buffer.c +++ b/src/mesa/pipe/i915simple/i915_buffer.c @@ -40,12 +40,12 @@ * evolve in separate directions... Don't try and remove this yet. */ static struct pipe_buffer_handle * -i915_create_buffer(struct pipe_context *pipe, +i915_buffer_create(struct pipe_context *pipe, unsigned alignment, unsigned flags) { struct i915_context *i915 = i915_context( pipe ); - return i915->winsys->create_buffer( i915->winsys, alignment ); + return i915->winsys->buffer_create( i915->winsys, alignment ); } static void *i915_buffer_map(struct pipe_context *pipe, @@ -110,7 +110,7 @@ static void i915_buffer_get_subdata(struct pipe_context *pipe, void i915_init_buffer_functions( struct i915_context *i915 ) { - i915->pipe.create_buffer = i915_create_buffer; + i915->pipe.create_buffer = i915_buffer_create; i915->pipe.buffer_map = i915_buffer_map; i915->pipe.buffer_unmap = i915_buffer_unmap; i915->pipe.buffer_reference = i915_buffer_reference; diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index 52181cf9f4..e2b807f959 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -89,7 +89,7 @@ i915_region_alloc(struct pipe_context *pipe, region->height = height; /* needed? */ region->refcount = 1; - region->buffer = i915->winsys->create_buffer( i915->winsys, 64 ); + region->buffer = i915->winsys->buffer_create( i915->winsys, 64 ); i915->winsys->buffer_data( i915->winsys, region->buffer, diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index d732e879db..ebc55eb634 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -135,12 +135,11 @@ static void i915_set_sampler_state(struct pipe_context *pipe, static void i915_set_texture_state(struct pipe_context *pipe, - GLuint unit, - struct pipe_texture_object *texture) + GLuint unit, + struct pipe_mipmap_tree *texture) { struct i915_context *i915 = i915_context(pipe); - assert(unit < PIPE_MAX_SAMPLERS); i915->texture[unit] = texture; /* ptr, not struct */ i915->dirty |= I915_NEW_TEXTURE; diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index ba8cbdacf3..c0539daa6e 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -189,7 +189,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_RELOC(cbuf_region->buffer, I915_BUFFER_ACCESS_WRITE, - /*cbuf_region->draw_offset*/0); + 0); } /* What happens if no zbuf?? @@ -207,7 +207,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_RELOC(depth_region->buffer, I915_BUFFER_ACCESS_WRITE, - /*depth_region->draw_offset*/0); + 0); } diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 514c53c7b2..50cb231422 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -51,16 +51,12 @@ struct pipe_buffer_handle; struct i915_winsys { - /* The buffer manager is modeled after the dri_bugmgr interface, - * but this is the subset that i915 cares about. Remember that - * i915 gets to choose the interface it needs, and the window - * systems must then implement that interface (rather than the - * other way around...). - * - * I915 only really wants to make system memory allocations, - * right?? + /* Many of the winsys's are probably going to have a similar + * buffer-manager interface, as something almost identical is + * currently exposed in the pipe interface. Probably want to avoid + * endless repetition of this code somehow. */ - struct pipe_buffer_handle *(*create_buffer)(struct i915_winsys *sws, + struct pipe_buffer_handle *(*buffer_create)(struct i915_winsys *sws, unsigned alignment ); void *(*buffer_map)( struct i915_winsys *sws, @@ -110,6 +106,7 @@ struct i915_winsys { unsigned access_flags, unsigned delta ); void (*batch_flush)( struct i915_winsys *sws ); + void (*batch_wait_idle)( struct i915_winsys *sws ); /* Printf??? -- cgit v1.2.3 From ef8b68a8a84eeac238fffa98f729d4809b6e2de4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 11:13:36 +0100 Subject: Pull in i915simple.a --- src/mesa/drivers/dri/intel_winsys/Makefile | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index c9dc15e088..786e8c02a2 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -7,7 +7,8 @@ LIBNAME = i915tex_dri.so MINIGLX_SOURCES = server/intel_dri.c PIPE_DRIVERS = \ - $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a + $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a \ + $(TOP)/src/mesa/pipe/i915simple/libi915simple.a DRIVER_SOURCES = \ intel_pipe_i915simple.c \ -- cgit v1.2.3 From 173e2c6a678ff35cde8057caf15aa8a0056dfdf8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 11:17:48 +0100 Subject: Remove dead files. --- src/mesa/drivers/dri/intel_winsys/Makefile | 9 - src/mesa/drivers/dri/intel_winsys/intel_tex.c | 184 ------ src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c | 315 ---------- .../drivers/dri/intel_winsys/intel_tex_format.c | 172 ----- .../drivers/dri/intel_winsys/intel_tex_image.c | 695 --------------------- .../drivers/dri/intel_winsys/intel_tex_subimage.c | 192 ------ .../drivers/dri/intel_winsys/intel_tex_validate.c | 277 -------- 7 files changed, 1844 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_format.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_image.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_subimage.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_validate.c (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index 786e8c02a2..3d3d525c30 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -25,15 +25,6 @@ DRIVER_SOURCES = \ intel_depthstencil.c \ intel_batchpool.c -OLD_TEX =\ - intel_tex_image.c \ - intel_tex_subimage.c \ - intel_tex_copy.c \ - intel_tex_validate.c \ - intel_tex_format.c \ - intel_tex.c - - C_SOURCES = \ $(COMMON_SOURCES) \ $(COMMON_BM_SOURCES) \ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex.c b/src/mesa/drivers/dri/intel_winsys/intel_tex.c deleted file mode 100644 index 5270094cc5..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex.c +++ /dev/null @@ -1,184 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "main/texobj.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_mipmap_tree.h" -#include "intel_tex.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -static GLboolean -intelIsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) -{ -#if 0 - struct intel_context *intel = intel_context(ctx); - struct st_texture_object *stObj = st_texture_object(texObj); - - return - stObj->mt && - stObj->mt->region && - intel_is_region_resident(intel, stObj->mt->region); -#endif - return 1; -} - - - -static struct gl_texture_image * -intelNewTextureImage(GLcontext * ctx) -{ - DBG("%s\n", __FUNCTION__); - (void) ctx; - return (struct gl_texture_image *) CALLOC_STRUCT(st_texture_image); -} - - -static struct gl_texture_object * -intelNewTextureObject(GLcontext * ctx, GLuint name, GLenum target) -{ - struct st_texture_object *obj = CALLOC_STRUCT(st_texture_object); - - DBG("%s\n", __FUNCTION__); - _mesa_initialize_texture_object(&obj->base, name, target); - - return &obj->base; -} - -static void -intelDeleteTextureObject(GLcontext *ctx, - struct gl_texture_object *texObj) -{ - struct pipe_context *pipe = ctx->st->pipe; - struct st_texture_object *stObj = st_texture_object(texObj); - - if (stObj->mt) - st_miptree_release(pipe, &stObj->mt); - - _mesa_delete_texture_object(ctx, texObj); -} - - -static void -intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) -{ - struct pipe_context *pipe = ctx->st->pipe; - struct st_texture_image *stImage = st_texture_image(texImage); - - DBG("%s\n", __FUNCTION__); - - if (stImage->mt) { - st_miptree_release(pipe, &stImage->mt); - } - - if (texImage->Data) { - free(texImage->Data); - texImage->Data = NULL; - } -} - - - - -/* ================================================================ - * From linux kernel i386 header files, copes with odd sizes better - * than COPY_DWORDS would: - * XXX Put this in src/mesa/main/imports.h ??? - */ -#if defined(i386) || defined(__i386__) -static INLINE void * -__memcpy(void *to, const void *from, size_t n) -{ - int d0, d1, d2; - __asm__ __volatile__("rep ; movsl\n\t" - "testb $2,%b4\n\t" - "je 1f\n\t" - "movsw\n" - "1:\ttestb $1,%b4\n\t" - "je 2f\n\t" - "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2) - :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from) - :"memory"); - return (to); -} -#else -#define __memcpy(a,b,c) memcpy(a,b,c) -#endif - - -/* The system memcpy (at least on ubuntu 5.10) has problems copying - * to agp (writecombined) memory from a source which isn't 64-byte - * aligned - there is a 4x performance falloff. - * - * The x86 __memcpy is immune to this but is slightly slower - * (10%-ish) than the system memcpy. - * - * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but - * isn't much faster than x86_memcpy for agp copies. - * - * TODO: switch dynamically. - */ -static void * -do_memcpy(void *dest, const void *src, size_t n) -{ - if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) { - return __memcpy(dest, src, n); - } - else - return memcpy(dest, src, n); -} - - -void -intelInitTextureFuncs(struct dd_function_table *functions) -{ - functions->ChooseTextureFormat = intelChooseTextureFormat; - functions->TexImage1D = intelTexImage1D; - functions->TexImage2D = intelTexImage2D; - functions->TexImage3D = intelTexImage3D; - functions->TexSubImage1D = intelTexSubImage1D; - functions->TexSubImage2D = intelTexSubImage2D; - functions->TexSubImage3D = intelTexSubImage3D; - functions->CopyTexImage1D = intelCopyTexImage1D; - functions->CopyTexImage2D = intelCopyTexImage2D; - functions->CopyTexSubImage1D = intelCopyTexSubImage1D; - functions->CopyTexSubImage2D = intelCopyTexSubImage2D; - functions->GetTexImage = intelGetTexImage; - - /* compressed texture functions */ - functions->CompressedTexImage2D = intelCompressedTexImage2D; - functions->GetCompressedTexImage = intelGetCompressedTexImage; - - functions->NewTextureObject = intelNewTextureObject; - functions->NewTextureImage = intelNewTextureImage; - functions->DeleteTexture = intelDeleteTextureObject; - functions->FreeTexImageData = intelFreeTextureImageData; - functions->UpdateTexturePalette = 0; - functions->IsTextureResident = intelIsTextureResident; - - functions->TextureMemCpy = do_memcpy; -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c deleted file mode 100644 index d2cf6f4669..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex_copy.c +++ /dev/null @@ -1,315 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "main/mtypes.h" -#include "main/enums.h" -#include "main/image.h" -#include "main/teximage.h" - -#include "state_tracker/st_mipmap_tree.h" - -#if 0 -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_buffers.h" -#include "intel_fbo.h" -#include "intel_blit.h" -#endif -#include "intel_tex.h" - - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -/** - * Get the pipe_region which is the source for any glCopyTex[Sub]Image call. - * - * Do the best we can using the blitter. A future project is to use - * the texture engine and fragment programs for these copies. - */ -static const struct pipe_region * -get_teximage_source(struct intel_context *intel, GLenum internalFormat) -{ -#if 00 - struct intel_renderbuffer *irb; - - DBG("%s %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(internalFormat)); - - switch (internalFormat) { - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16_ARB: - irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); - if (irb && irb->region && irb->region->cpp == 2) - return irb->region; - return NULL; - case GL_DEPTH24_STENCIL8_EXT: - case GL_DEPTH_STENCIL_EXT: - irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); - if (irb && irb->region && irb->region->cpp == 4) - return irb->region; - return NULL; - case GL_RGBA: - case GL_RGBA8: - return intel_readbuf_region(intel); - case GL_RGB: - if (intel->intelScreen->front.cpp == 2) - return intel_readbuf_region(intel); - return NULL; - default: - return NULL; - } -#else - return NULL; -#endif -} - - -static GLboolean -do_copy_texsubimage(GLcontext *ctx, - struct st_texture_image *intelImage, - GLenum internalFormat, - GLint dstx, GLint dsty, - GLint x, GLint y, GLsizei width, GLsizei height) -{ - struct intel_context *intel = intel_context(ctx); - const struct pipe_region *src = - get_teximage_source(intel, internalFormat); - - if (!intelImage->mt || !src) { - DBG("%s fail %p %p\n", __FUNCTION__, intelImage->mt, src); - return GL_FALSE; - } - -#if 00 /* XXX FIX flush/locking */ - intelFlush(ctx); - /* XXX still need the lock ? */ - LOCK_HARDWARE(intel); -#endif - - { - GLuint image_offset = st_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); - const GLint orig_x = x; - const GLint orig_y = y; - const struct gl_framebuffer *fb = ctx->DrawBuffer; - - if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, - &x, &y, &width, &height)) { - /* Update dst for clipped src. Need to also clip the source rect. - */ - dstx += x - orig_x; - dsty += y - orig_y; - - if (!(ctx->ReadBuffer->Name == 0)) { - /* XXX this looks bogus ? */ - /* FBO: invert Y */ - y = ctx->ReadBuffer->Height - y - 1; - } - - /* A bit of fiddling to get the blitter to work with -ve - * pitches. But we get a nice inverted blit this way, so it's - * worth it: - */ -#if 0 - intelEmitCopyBlit(intel, - intelImage->mt->cpp, - -src->pitch, - src->buffer, - src->height * src->pitch * src->cpp, - intelImage->mt->pitch, - intelImage->mt->region->buffer, - image_offset, - x, y + height, dstx, dsty, width, height, - GL_COPY); /* ? */ - intel_batchbuffer_flush(intel->batch); -#else - /* XXX use pipe->region_copy() ??? */ - (void) image_offset; -#endif - } - } - -#if 0 - UNLOCK_HARDWARE(intel); -#endif - -#if 0 - /* GL_SGIS_generate_mipmap -- this can be accelerated now. - * XXX Add a ctx->Driver.GenerateMipmaps() function? - */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); - } -#endif - - return GL_TRUE; -} - - - - - -void -intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLint border) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - - if (border) - goto fail; - - /* Setup or redefine the texture object, mipmap tree and texture - * image. Don't populate yet. - */ - ctx->Driver.TexImage1D(ctx, target, level, internalFormat, - width, border, - GL_RGBA, CHAN_TYPE, NULL, - &ctx->DefaultPacking, texObj, texImage); - - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, 0, 0, x, y, width, 1)) - goto fail; - - return; - - fail: -#if 0 - _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, - width, border); -#endif - ; -} - -void -intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint border) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - - if (border) - goto fail; - - /* Setup or redefine the texture object, mipmap tree and texture - * image. Don't populate yet. - */ - ctx->Driver.TexImage2D(ctx, target, level, internalFormat, - width, height, border, - GL_RGBA, CHAN_TYPE, NULL, - &ctx->DefaultPacking, texObj, texImage); - - - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, 0, 0, x, y, width, height)) - goto fail; - - return; - - fail: -#if 0 - _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, - width, height, border); -#endif - assert(0); -} - - -void -intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint x, GLint y, GLsizei width) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - GLenum internalFormat = texImage->InternalFormat; - - /* XXX need to check as in above function? */ - - /* Need to check texture is compatible with source format. - */ - - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, xoffset, 0, x, y, width, 1)) { -#if 0 - _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); -#endif - assert(0); - } -} - - - -void -intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLint x, GLint y, GLsizei width, GLsizei height) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - GLenum internalFormat = texImage->InternalFormat; - - - /* Need to check texture is compatible with source format. - */ - - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, - xoffset, yoffset, x, y, width, height)) { -#if 0 - _swrast_copy_texsubimage2d(ctx, target, level, - xoffset, yoffset, x, y, width, height); -#endif - assert(0); - } -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_format.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_format.c deleted file mode 100644 index a11718bb07..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex_format.c +++ /dev/null @@ -1,172 +0,0 @@ -#include "intel_context.h" -#include "intel_tex.h" -#include "texformat.h" -#include "enums.h" - -/* It works out that this function is fine for all the supported - * hardware. However, there is still a need to map the formats onto - * hardware descriptors. - */ -/* Note that the i915 can actually support many more formats than - * these if we take the step of simply swizzling the colors - * immediately after sampling... - */ -const struct gl_texture_format * -intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, - GLenum format, GLenum type) -{ - struct intel_context *intel = intel_context(ctx); - const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); - - switch (internalFormat) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - if (format == GL_BGRA) { - if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - return &_mesa_texformat_argb8888; - } - else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - return &_mesa_texformat_argb4444; - } - else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - return &_mesa_texformat_argb1555; - } - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return &_mesa_texformat_rgb565; - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case GL_RGBA4: - case GL_RGBA2: - return &_mesa_texformat_argb4444; - - case GL_RGB5_A1: - return &_mesa_texformat_argb1555; - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - return &_mesa_texformat_argb8888; - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; - - case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; - - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - return &_mesa_texformat_z16; - - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_z24_s8; - - default: - fprintf(stderr, "unexpected texture format %s in %s\n", - _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); - return NULL; - } - - return NULL; /* never get here */ -} - -int intel_compressed_num_bytes(GLuint mesaFormat) -{ - int bytes = 0; - switch(mesaFormat) { - - case MESA_FORMAT_RGB_FXT1: - case MESA_FORMAT_RGBA_FXT1: - case MESA_FORMAT_RGB_DXT1: - case MESA_FORMAT_RGBA_DXT1: - bytes = 2; - break; - - case MESA_FORMAT_RGBA_DXT3: - case MESA_FORMAT_RGBA_DXT5: - bytes = 4; - default: - break; - } - - return bytes; -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_image.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_image.c deleted file mode 100644 index 93a59b6896..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex_image.c +++ /dev/null @@ -1,695 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "glheader.h" -#include "macros.h" -#include "mtypes.h" -#include "enums.h" -#include "colortab.h" -#include "convolve.h" -#include "context.h" -#include "simple_list.h" -#include "texcompress.h" -#include "texformat.h" -#include "texobj.h" -#include "texstore.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_mipmap_tree.h" - -#include "intel_tex.h" - - - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -/* Functions to store texture images. Where possible, mipmap_tree's - * will be created or further instantiated with image data, otherwise - * images will be stored in malloc'd memory. A validation step is - * required to pull those images into a mipmap tree, or otherwise - * decide a fallback is required. - */ - - -static int -logbase2(int n) -{ - GLint i = 1; - GLint log2 = 0; - - while (n > i) { - i *= 2; - log2++; - } - - return log2; -} - - -/* Otherwise, store it in memory if (Border != 0) or (any dimension == - * 1). - * - * Otherwise, if max_level >= level >= min_level, create tree with - * space for textures from min_level down to max_level. - * - * Otherwise, create tree with space for textures from (level - * 0)..(1x1). Consider pruning this tree at a validation if the - * saving is worth it. - */ -static void -guess_and_alloc_mipmap_tree(struct pipe_context *pipe, - struct st_texture_object *intelObj, - struct st_texture_image *intelImage) -{ - GLuint firstLevel; - GLuint lastLevel; - GLuint width = intelImage->base.Width; - GLuint height = intelImage->base.Height; - GLuint depth = intelImage->base.Depth; - GLuint l2width, l2height, l2depth; - GLuint i, comp_byte = 0; - - DBG("%s\n", __FUNCTION__); - - if (intelImage->base.Border) - return; - - if (intelImage->level > intelObj->base.BaseLevel && - (intelImage->base.Width == 1 || - (intelObj->base.Target != GL_TEXTURE_1D && - intelImage->base.Height == 1) || - (intelObj->base.Target == GL_TEXTURE_3D && - intelImage->base.Depth == 1))) - return; - - /* If this image disrespects BaseLevel, allocate from level zero. - * Usually BaseLevel == 0, so it's unlikely to happen. - */ - if (intelImage->level < intelObj->base.BaseLevel) - firstLevel = 0; - else - firstLevel = intelObj->base.BaseLevel; - - - /* Figure out image dimensions at start level. - */ - for (i = intelImage->level; i > firstLevel; i--) { - width <<= 1; - if (height != 1) - height <<= 1; - if (depth != 1) - depth <<= 1; - } - - /* Guess a reasonable value for lastLevel. This is probably going - * to be wrong fairly often and might mean that we have to look at - * resizable buffers, or require that buffers implement lazy - * pagetable arrangements. - */ - if ((intelObj->base.MinFilter == GL_NEAREST || - intelObj->base.MinFilter == GL_LINEAR) && - intelImage->level == firstLevel) { - lastLevel = firstLevel; - } - else { - l2width = logbase2(width); - l2height = logbase2(height); - l2depth = logbase2(depth); - lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); - } - - assert(!intelObj->mt); - if (intelImage->base.IsCompressed) - comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); - intelObj->mt = st_miptree_create(pipe, - intelObj->base.Target, - intelImage->base.InternalFormat, - firstLevel, - lastLevel, - width, - height, - depth, - intelImage->base.TexFormat->TexelBytes, - comp_byte); - - DBG("%s - success\n", __FUNCTION__); -} - - - - -static GLuint -target_to_face(GLenum target) -{ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - return ((GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X); - default: - return 0; - } -} - - - -/* There are actually quite a few combinations this will work for, - * more than what I've listed here. - */ -static GLboolean -check_pbo_format(GLint internalFormat, - GLenum format, GLenum type, - const struct gl_texture_format *mesa_format) -{ - switch (internalFormat) { - case 4: - case GL_RGBA: - return (format == GL_BGRA && - (type == GL_UNSIGNED_BYTE || - type == GL_UNSIGNED_INT_8_8_8_8_REV) && - mesa_format == &_mesa_texformat_argb8888); - case 3: - case GL_RGB: - return (format == GL_RGB && - type == GL_UNSIGNED_SHORT_5_6_5 && - mesa_format == &_mesa_texformat_rgb565); - case GL_YCBCR_MESA: - return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); - default: - return GL_FALSE; - } -} - - -/* XXX: Do this for TexSubImage also: - */ -static GLboolean -try_pbo_upload(struct intel_context *intel, - struct st_texture_image *intelImage, - const struct gl_pixelstore_attrib *unpack, - GLint internalFormat, - GLint width, GLint height, - GLenum format, GLenum type, const void *pixels) -{ - return GL_FALSE; /* XXX fix flushing/locking/blitting below */ -#if 000 - struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); - GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; - - if (!pbo || - intel->ctx._ImageTransferState || - unpack->SkipPixels || unpack->SkipRows) { - _mesa_printf("%s: failure 1\n", __FUNCTION__); - return GL_FALSE; - } - - src_offset = (GLuint) pixels; - - if (unpack->RowLength > 0) - src_stride = unpack->RowLength; - else - src_stride = width; - - dst_offset = st_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); - - dst_stride = intelImage->mt->pitch; - - intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); - { - struct _DriBufferObject *src_buffer = - intel_bufferobj_buffer(intel, pbo, INTEL_READ); - - /* Temporary hack: cast to _DriBufferObject: - */ - struct _DriBufferObject *dst_buffer = - (struct _DriBufferObject *)intelImage->mt->region->buffer; - - - intelEmitCopyBlit(intel, - intelImage->mt->cpp, - src_stride, src_buffer, src_offset, - dst_stride, dst_buffer, dst_offset, - 0, 0, 0, 0, width, height, - GL_COPY); - - intel_batchbuffer_flush(intel->batch); - } - UNLOCK_HARDWARE(intel); - - return GL_TRUE; -#endif -} - - - -static GLboolean -try_pbo_zcopy(struct intel_context *intel, - struct st_texture_image *intelImage, - const struct gl_pixelstore_attrib *unpack, - GLint internalFormat, - GLint width, GLint height, - GLenum format, GLenum type, const void *pixels) -{ - return GL_FALSE; -} - - - - - - -static void -intelTexImage(GLcontext * ctx, - GLint dims, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage, GLsizei imageSize, int compressed) -{ - struct intel_context *intel = intel_context(ctx); - struct st_texture_object *intelObj = st_texture_object(texObj); - struct st_texture_image *intelImage = st_texture_image(texImage); - GLint postConvWidth = width; - GLint postConvHeight = height; - GLint texelBytes, sizeInBytes; - GLuint dstRowStride; - - - DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); - - intelFlush(ctx); - - intelImage->face = target_to_face(target); - intelImage->level = level; - - if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { - _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth, - &postConvHeight); - } - - /* choose the texture format */ - texImage->TexFormat = intelChooseTextureFormat(ctx, internalFormat, - format, type); - - _mesa_set_fetch_functions(texImage, dims); - - if (texImage->TexFormat->TexelBytes == 0) { - /* must be a compressed format */ - texelBytes = 0; - texImage->IsCompressed = GL_TRUE; - texImage->CompressedSize = - ctx->Driver.CompressedTextureSize(ctx, texImage->Width, - texImage->Height, texImage->Depth, - texImage->TexFormat->MesaFormat); - } else { - texelBytes = texImage->TexFormat->TexelBytes; - - /* Minimum pitch of 32 bytes */ - if (postConvWidth * texelBytes < 32) { - postConvWidth = 32 / texelBytes; - texImage->RowStride = postConvWidth; - } - - assert(texImage->RowStride == postConvWidth); - } - - /* Release the reference to a potentially orphaned buffer. - * Release any old malloced memory. - */ - if (intelImage->mt) { - st_miptree_release(intel->pipe, &intelImage->mt); - assert(!texImage->Data); - } - else if (texImage->Data) { - _mesa_align_free(texImage->Data); - } - - /* If this is the only texture image in the tree, could call - * bmBufferData with NULL data to free the old block and avoid - * waiting on any outstanding fences. - */ - if (intelObj->mt && - intelObj->mt->first_level == level && - intelObj->mt->last_level == level && - intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && - !st_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { - - DBG("release it\n"); - st_miptree_release(intel->pipe, &intelObj->mt); - assert(!intelObj->mt); - } - - if (!intelObj->mt) { - guess_and_alloc_mipmap_tree(intel->pipe, intelObj, intelImage); - if (!intelObj->mt) { - DBG("guess_and_alloc_mipmap_tree: failed\n"); - } - } - - assert(!intelImage->mt); - - if (intelObj->mt && - st_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { - - st_miptree_reference(&intelImage->mt, intelObj->mt); - assert(intelImage->mt); - } - - if (!intelImage->mt) - DBG("XXX: Image did not fit into tree - storing in local memory!\n"); - -#if 0 /* XXX FIX when st_buffer_objects are in place */ - /* PBO fastpaths: - */ - if (dims <= 2 && - intelImage->mt && - intel_buffer_object(unpack->BufferObj) && - check_pbo_format(internalFormat, format, - type, intelImage->base.TexFormat)) { - - DBG("trying pbo upload\n"); - - /* Attempt to texture directly from PBO data (zero copy upload). - * - * Currently disable as it can lead to worse as well as better - * performance (in particular when pipe_region_cow() is - * required). - */ - if (intelObj->mt == intelImage->mt && - intelObj->mt->first_level == level && - intelObj->mt->last_level == level) { - - if (try_pbo_zcopy(intel, intelImage, unpack, - internalFormat, - width, height, format, type, pixels)) { - - DBG("pbo zcopy upload succeeded\n"); - return; - } - } - - - /* Otherwise, attempt to use the blitter for PBO image uploads. - */ - if (try_pbo_upload(intel, intelImage, unpack, - internalFormat, - width, height, format, type, pixels)) { - DBG("pbo upload succeeded\n"); - return; - } - - DBG("pbo upload failed\n"); - } -#else - (void) try_pbo_upload; - (void) check_pbo_format; - (void) try_pbo_zcopy; -#endif - - - /* intelCopyTexImage calls this function with pixels == NULL, with - * the expectation that the mipmap tree will be set up but nothing - * more will be done. This is where those calls return: - */ - if (compressed) { - pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels, - unpack, - "glCompressedTexImage"); - } else { - pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, - format, type, - pixels, unpack, "glTexImage"); - } - if (!pixels) - return; - - - if (intelImage->mt) - intel->pipe->region_idle(intel->pipe, intelImage->mt->region); - - LOCK_HARDWARE(intel); - - if (intelImage->mt) { - texImage->Data = st_miptree_image_map(intel->pipe, - intelImage->mt, - intelImage->face, - intelImage->level, - &dstRowStride, - intelImage->base.ImageOffsets); - } - else { - /* Allocate regular memory and store the image there temporarily. */ - if (texImage->IsCompressed) { - sizeInBytes = texImage->CompressedSize; - dstRowStride = - _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); - assert(dims != 3); - } - else { - dstRowStride = postConvWidth * texelBytes; - sizeInBytes = depth * dstRowStride * postConvHeight; - } - - texImage->Data = malloc(sizeInBytes); - } - - DBG("Upload image %dx%dx%d row_len %x " - "pitch %x\n", - width, height, depth, width * texelBytes, dstRowStride); - - /* Copy data. Would like to know when it's ok for us to eg. use - * the blitter to copy. Or, use the hardware to do the format - * conversion and copy: - */ - if (compressed) { - memcpy(texImage->Data, pixels, imageSize); - } else if (!texImage->TexFormat->StoreImage(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, unpack)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); - } - - _mesa_unmap_teximage_pbo(ctx, unpack); - - if (intelImage->mt) { - st_miptree_image_unmap(intel->pipe, intelImage->mt); - texImage->Data = NULL; - } - - UNLOCK_HARDWARE(intel); - -#if 0 - /* GL_SGIS_generate_mipmap -- this can be accelerated now. - */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); - } -#endif -} - -void -intelTexImage3D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexImage(ctx, 3, target, level, - internalFormat, width, height, depth, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); -} - - -void -intelTexImage2D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexImage(ctx, 2, target, level, - internalFormat, width, height, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); -} - -void -intelTexImage1D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexImage(ctx, 1, target, level, - internalFormat, width, 1, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); -} - -void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - intelTexImage(ctx, 2, target, level, - internalFormat, width, height, 1, border, - 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1); -} - -/** - * Need to map texture image into memory before copying image data, - * then unmap it. - */ -static void -intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage, int compressed) -{ - struct intel_context *intel = intel_context(ctx); - struct st_texture_image *intelImage = st_texture_image(texImage); - - /* Map */ - if (intelImage->mt) { - /* Image is stored in hardware format in a buffer managed by the - * kernel. Need to explicitly map and unmap it. - */ - intelImage->base.Data = - st_miptree_image_map(intel->pipe, - intelImage->mt, - intelImage->face, - intelImage->level, - &intelImage->base.RowStride, - intelImage->base.ImageOffsets); - intelImage->base.RowStride /= intelImage->mt->cpp; - } - else { - /* Otherwise, the image should actually be stored in - * intelImage->base.Data. This is pretty confusing for - * everybody, I'd much prefer to separate the two functions of - * texImage->Data - storage for texture images in main memory - * and access (ie mappings) of images. In other words, we'd - * create a new texImage->Map field and leave Data simply for - * storage. - */ - assert(intelImage->base.Data); - } - - - if (compressed) { - _mesa_get_compressed_teximage(ctx, target, level, pixels, - texObj, texImage); - } else { - _mesa_get_teximage(ctx, target, level, format, type, pixels, - texObj, texImage); - } - - - /* Unmap */ - if (intelImage->mt) { - st_miptree_image_unmap(intel->pipe, intelImage->mt); - intelImage->base.Data = NULL; - } -} - -void -intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intel_get_tex_image(ctx, target, level, format, type, pixels, - texObj, texImage, 0); - - -} - -void -intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, - GLvoid *pixels, - const struct gl_texture_object *texObj, - const struct gl_texture_image *texImage) -{ - intel_get_tex_image(ctx, target, level, 0, 0, pixels, - texObj, texImage, 1); - -} - -void -intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, - unsigned long long offset, GLint depth, GLuint pitch) -{ - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct st_texture_object *stObj = st_texture_object(tObj); - - if (!stObj) - return; - - if (stObj->mt) - st_miptree_release(intel->pipe, &stObj->mt); - - stObj->imageOverride = GL_TRUE; - stObj->depthOverride = depth; - stObj->pitchOverride = pitch; - - if (offset) - stObj->textureOffset = offset; -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_subimage.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_subimage.c deleted file mode 100644 index 0937114c7f..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex_subimage.c +++ /dev/null @@ -1,192 +0,0 @@ - -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "mtypes.h" -#include "texobj.h" -#include "texstore.h" -#include "enums.h" - -/* -#include "intel_context.h" -*/ -#include "intel_tex.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_mipmap_tree.h" - -#include "pipe/p_context.h" - - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -static void -intelTexSubimage(GLcontext * ctx, - GLint dims, - GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLint width, GLint height, GLint depth, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - struct pipe_context *pipe = ctx->st->pipe; - struct st_texture_image *intelImage = st_texture_image(texImage); - GLuint dstRowStride; - - DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), - level, xoffset, yoffset, width, height); - - intelFlush(ctx); - - pixels = - _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format, - type, pixels, packing, "glTexSubImage2D"); - if (!pixels) - return; - - if (intelImage->mt) - pipe->region_idle(pipe, intelImage->mt->region); - -#if 0 - LOCK_HARDWARE(intel); -#endif - - /* Map buffer if necessary. Need to lock to prevent other contexts - * from uploading the buffer under us. - */ - if (intelImage->mt) - texImage->Data = st_miptree_image_map(pipe, - intelImage->mt, - intelImage->face, - intelImage->level, - &dstRowStride, - texImage->ImageOffsets); - - assert(dstRowStride); - - if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); - } - -#if 0 - /* GL_SGIS_generate_mipmap */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); - } -#endif - - _mesa_unmap_teximage_pbo(ctx, packing); - - if (intelImage->mt) { - st_miptree_image_unmap(pipe, intelImage->mt); - texImage->Data = NULL; - } - -#if 0 - UNLOCK_HARDWARE(intel); -#endif -} - - - - - -void -intelTexSubImage3D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - - intelTexSubimage(ctx, 3, - target, level, - xoffset, yoffset, zoffset, - width, height, depth, - format, type, pixels, packing, texObj, texImage); - -} - - - -void -intelTexSubImage2D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - - intelTexSubimage(ctx, 2, - target, level, - xoffset, yoffset, 0, - width, height, 1, - format, type, pixels, packing, texObj, texImage); - -} - - -void -intelTexSubImage1D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexSubimage(ctx, 1, - target, level, - xoffset, 0, 0, - width, 1, 1, - format, type, pixels, packing, texObj, texImage); - -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_validate.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_validate.c deleted file mode 100644 index d0631a88f1..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex_validate.c +++ /dev/null @@ -1,277 +0,0 @@ -#include "mtypes.h" -#include "macros.h" - -#include "state_tracker/st_mipmap_tree.h" -#include "intel_tex.h" - -#include "pipe/p_state.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -/** - * Compute which mipmap levels that really need to be sent to the hardware. - * This depends on the base image size, GL_TEXTURE_MIN_LOD, - * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. - */ -static void -intel_calculate_first_last_level(struct st_texture_object *intelObj) -{ - struct gl_texture_object *tObj = &intelObj->base; - const struct gl_texture_image *const baseImage = - tObj->Image[0][tObj->BaseLevel]; - - /* These must be signed values. MinLod and MaxLod can be negative numbers, - * and having firstLevel and lastLevel as signed prevents the need for - * extra sign checks. - */ - int firstLevel; - int lastLevel; - - /* Yes, this looks overly complicated, but it's all needed. - */ - switch (tObj->Target) { - case GL_TEXTURE_1D: - case GL_TEXTURE_2D: - case GL_TEXTURE_3D: - case GL_TEXTURE_CUBE_MAP: - if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) { - /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL. - */ - firstLevel = lastLevel = tObj->BaseLevel; - } - else { - firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); - firstLevel = MAX2(firstLevel, tObj->BaseLevel); - lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); - lastLevel = MAX2(lastLevel, tObj->BaseLevel); - lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); - lastLevel = MIN2(lastLevel, tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ - } - break; - case GL_TEXTURE_RECTANGLE_NV: - case GL_TEXTURE_4D_SGIS: - firstLevel = lastLevel = 0; - break; - default: - return; - } - - /* save these values */ - intelObj->firstLevel = firstLevel; - intelObj->lastLevel = lastLevel; -} - -static void -copy_image_data_to_tree(struct pipe_context *pipe, - struct st_texture_object *intelObj, - struct st_texture_image *intelImage) -{ - if (intelImage->mt) { - /* Copy potentially with the blitter: - */ - st_miptree_image_copy(pipe, - intelObj->mt, - intelImage->face, - intelImage->level, intelImage->mt); - - st_miptree_release(pipe, &intelImage->mt); - } - else { - assert(intelImage->base.Data != NULL); - - /* More straightforward upload. - */ - st_miptree_image_data(pipe, - intelObj->mt, - intelImage->face, - intelImage->level, - intelImage->base.Data, - intelImage->base.RowStride, - intelImage->base.RowStride * - intelImage->base.Height); - _mesa_align_free(intelImage->base.Data); - intelImage->base.Data = NULL; - } - - st_miptree_reference(&intelImage->mt, intelObj->mt); -} - - -/* - */ -GLuint -intel_finalize_mipmap_tree(GLcontext *ctx, - struct pipe_context *pipe, GLuint unit, - GLboolean *needFlush) -{ - struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; - struct st_texture_object *intelObj = st_texture_object(tObj); - int comp_byte = 0; - int cpp; - - GLuint face, i; - GLuint nr_faces = 0; - struct st_texture_image *firstImage; - - *needFlush = GL_FALSE; - - /* We know/require this is true by now: - */ - assert(intelObj->base._Complete); - - /* What levels must the tree include at a minimum? - */ - intel_calculate_first_last_level(intelObj); - firstImage = - st_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); - - /* Fallback case: - */ - if (firstImage->base.Border) { - if (intelObj->mt) { - st_miptree_release(pipe, &intelObj->mt); - } - return GL_FALSE; - } - - - /* If both firstImage and intelObj have a tree which can contain - * all active images, favour firstImage. Note that because of the - * completeness requirement, we know that the image dimensions - * will match. - */ - if (firstImage->mt && - firstImage->mt != intelObj->mt && - firstImage->mt->first_level <= intelObj->firstLevel && - firstImage->mt->last_level >= intelObj->lastLevel) { - - if (intelObj->mt) - st_miptree_release(pipe, &intelObj->mt); - - st_miptree_reference(&intelObj->mt, firstImage->mt); - } - - if (firstImage->base.IsCompressed) { - comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); - cpp = comp_byte; - } - else cpp = firstImage->base.TexFormat->TexelBytes; - - /* Check tree can hold all active levels. Check tree matches - * target, imageFormat, etc. - * - * XXX: For some layouts (eg i945?), the test might have to be - * first_level == firstLevel, as the tree isn't valid except at the - * original start level. Hope to get around this by - * programming minLod, maxLod, baseLevel into the hardware and - * leaving the tree alone. - */ - if (intelObj->mt && - (intelObj->mt->target != intelObj->base.Target || - intelObj->mt->internal_format != firstImage->base.InternalFormat || - intelObj->mt->first_level != intelObj->firstLevel || - intelObj->mt->last_level != intelObj->lastLevel || - intelObj->mt->width0 != firstImage->base.Width || - intelObj->mt->height0 != firstImage->base.Height || - intelObj->mt->depth0 != firstImage->base.Depth || - intelObj->mt->cpp != cpp || - intelObj->mt->compressed != firstImage->base.IsCompressed)) { - st_miptree_release(pipe, &intelObj->mt); - } - - - /* May need to create a new tree: - */ - if (!intelObj->mt) { - intelObj->mt = st_miptree_create(pipe, - intelObj->base.Target, - firstImage->base.InternalFormat, - intelObj->firstLevel, - intelObj->lastLevel, - firstImage->base.Width, - firstImage->base.Height, - firstImage->base.Depth, - cpp, - comp_byte); - } - - /* Pull in any images not in the object's tree: - */ - nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct st_texture_image *intelImage = - st_texture_image(intelObj->base.Image[face][i]); - - /* Need to import images in main memory or held in other trees. - */ - if (intelObj->mt != intelImage->mt) { - copy_image_data_to_tree(pipe, intelObj, intelImage); - *needFlush = GL_TRUE; - } - } - } - - /** - if (need_flush) - intel_batchbuffer_flush(intel->batch); - **/ - - return GL_TRUE; -} - - -#if 0 /* unused? */ -void -intel_tex_map_images(struct pipe_context *pipe, - struct st_texture_object *intelObj) -{ - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - GLuint face, i; - - DBG("%s\n", __FUNCTION__); - - for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct st_texture_image *intelImage = - st_texture_image(intelObj->base.Image[face][i]); - - if (intelImage->mt) { - intelImage->base.Data = - st_miptree_image_map(pipe, - intelImage->mt, - intelImage->face, - intelImage->level, - &intelImage->base.RowStride, - intelImage->base.ImageOffsets); - /* convert stride to texels, not bytes */ - intelImage->base.RowStride /= intelImage->mt->cpp; -/* intelImage->base.ImageStride /= intelImage->mt->cpp; */ - } - } - } -} - - - -void -intel_tex_unmap_images(struct pipe_context *pipe, - struct st_texture_object *intelObj) -{ - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - GLuint face, i; - - for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct st_texture_image *intelImage = - st_texture_image(intelObj->base.Image[face][i]); - - if (intelImage->mt) { - st_miptree_image_unmap(pipe, intelImage->mt); - intelImage->base.Data = NULL; - } - } - } -} -#endif -- cgit v1.2.3 From 8daa738cf95a771d2217fe3dd9de9334143348e0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 11:19:17 +0100 Subject: Remove more dead files. --- .../drivers/dri/intel_winsys/intel_mipmap_tree.c | 293 --------------------- .../drivers/dri/intel_winsys/intel_mipmap_tree.h | 105 -------- 2 files changed, 398 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.c b/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.c deleted file mode 100644 index 1a9aa10115..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.c +++ /dev/null @@ -1,293 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_mipmap_tree.h" -#include "enums.h" - -#include "pipe/p_state.h" -#include "pipe/p_context.h" - - -#define DBG if(0) printf - -static GLenum -target_to_target(GLenum target) -{ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - return GL_TEXTURE_CUBE_MAP_ARB; - default: - return target; - } -} - -struct pipe_mipmap_tree * -st_miptree_create(struct pipe_context *pipe, - GLenum target, - GLenum internal_format, - GLuint first_level, - GLuint last_level, - GLuint width0, - GLuint height0, - GLuint depth0, GLuint cpp, GLuint compress_byte) -{ - GLboolean ok; - struct pipe_mipmap_tree *mt = calloc(sizeof(*mt), 1); - - DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), - _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); - - mt->target = target_to_target(target); - mt->internal_format = internal_format; - mt->first_level = first_level; - mt->last_level = last_level; - mt->width0 = width0; - mt->height0 = height0; - mt->depth0 = depth0; - mt->cpp = compress_byte ? compress_byte : cpp; - mt->compressed = compress_byte ? 1 : 0; - mt->refcount = 1; - - ok = pipe->mipmap_tree_layout(pipe, mt); - if (ok) - mt->region = pipe->region_alloc(pipe, - mt->cpp, mt->pitch, mt->total_height); - - if (!mt->region) { - free(mt); - return NULL; - } - - return mt; -} - - -void -st_miptree_reference(struct pipe_mipmap_tree **dst, - struct pipe_mipmap_tree *src) -{ - src->refcount++; - *dst = src; - DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); -} - -void -st_miptree_release(struct pipe_context *pipe, - struct pipe_mipmap_tree **mt) -{ - if (!*mt) - return; - - DBG("%s %p refcount will be %d\n", __FUNCTION__, *mt, (*mt)->refcount - 1); - if (--(*mt)->refcount <= 0) { - GLuint i; - - DBG("%s deleting %p\n", __FUNCTION__, *mt); - - pipe->region_release(pipe, &((*mt)->region)); - - for (i = 0; i < MAX_TEXTURE_LEVELS; i++) - if ((*mt)->level[i].image_offset) - free((*mt)->level[i].image_offset); - - free(*mt); - } - *mt = NULL; -} - - - - -/* Can the image be pulled into a unified mipmap tree. This mirrors - * the completeness test in a lot of ways. - * - * Not sure whether I want to pass gl_texture_image here. - */ -GLboolean -st_miptree_match_image(struct pipe_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level) -{ - /* Images with borders are never pulled into mipmap trees. - */ - if (image->Border) - return GL_FALSE; - - if (image->InternalFormat != mt->internal_format || - image->IsCompressed != mt->compressed) - return GL_FALSE; - - /* Test image dimensions against the base level image adjusted for - * minification. This will also catch images not present in the - * tree, changed targets, etc. - */ - if (image->Width != mt->level[level].width || - image->Height != mt->level[level].height || - image->Depth != mt->level[level].depth) - return GL_FALSE; - - return GL_TRUE; -} - - -/* Although we use the image_offset[] array to store relative offsets - * to cube faces, Mesa doesn't know anything about this and expects - * each cube face to be treated as a separate image. - * - * These functions present that view to mesa: - */ -const GLuint * -st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) -{ - static const GLuint zero = 0; - - if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) - return &zero; - else - return mt->level[level].image_offset; -} - - -GLuint -st_miptree_image_offset(struct pipe_mipmap_tree * mt, - GLuint face, GLuint level) -{ - if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) - return (mt->level[level].level_offset + - mt->level[level].image_offset[face] * mt->cpp); - else - return mt->level[level].level_offset; -} - - - -/** - * Map a teximage in a mipmap tree. - * \param row_stride returns row stride in bytes - * \param image_stride returns image stride in bytes (for 3D textures). - * \return address of mapping - */ -GLubyte * -st_miptree_image_map(struct pipe_context *pipe, - struct pipe_mipmap_tree * mt, - GLuint face, - GLuint level, - GLuint * row_stride, GLuint * image_offsets) -{ - GLubyte *ptr; - DBG("%s \n", __FUNCTION__); - - if (row_stride) - *row_stride = mt->pitch * mt->cpp; - - if (image_offsets) - memcpy(image_offsets, mt->level[level].image_offset, - mt->level[level].depth * sizeof(GLuint)); - - ptr = pipe->region_map(pipe, mt->region); - - return ptr + st_miptree_image_offset(mt, face, level); -} - -void -st_miptree_image_unmap(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt) -{ - DBG("%s\n", __FUNCTION__); - pipe->region_unmap(pipe, mt->region); -} - - - -/* Upload data for a particular image. - */ -void -st_miptree_image_data(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, - GLuint level, - void *src, - GLuint src_row_pitch, GLuint src_image_pitch) -{ - GLuint depth = dst->level[level].depth; - GLuint dst_offset = st_miptree_image_offset(dst, face, level); - const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); - GLuint i; - GLuint height = 0; - - DBG("%s\n", __FUNCTION__); - for (i = 0; i < depth; i++) { - height = dst->level[level].height; - if(dst->compressed) - height /= 4; - pipe->region_data(pipe, dst->region, - dst_offset + dst_depth_offset[i], /* dst_offset */ - 0, 0, /* dstx, dsty */ - src, - src_row_pitch, - 0, 0, /* source x, y */ - dst->level[level].width, height); /* width, height */ - - src += src_image_pitch * dst->cpp; - } -} - -/* Copy mipmap image between trees - */ -void -st_miptree_image_copy(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, GLuint level, - struct pipe_mipmap_tree *src) -{ - GLuint width = src->level[level].width; - GLuint height = src->level[level].height; - GLuint depth = src->level[level].depth; - GLuint dst_offset = st_miptree_image_offset(dst, face, level); - GLuint src_offset = st_miptree_image_offset(src, face, level); - const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); - const GLuint *src_depth_offset = st_miptree_depth_offsets(src, level); - GLuint i; - - if (dst->compressed) - height /= 4; - for (i = 0; i < depth; i++) { - pipe->region_copy(pipe, - dst->region, dst_offset + dst_depth_offset[i], - 0, - 0, - src->region, src_offset + src_depth_offset[i], - 0, 0, width, height); - } - -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.h b/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.h deleted file mode 100644 index eafe47c029..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_mipmap_tree.h +++ /dev/null @@ -1,105 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_MIPMAP_TREE_H -#define INTEL_MIPMAP_TREE_H - - -#include "main/mtypes.h" - -struct pipe_context; -struct pipe_mipmap_tree; -struct pipe_region; - - -struct pipe_mipmap_tree *st_miptree_create(struct pipe_context *pipe, - GLenum target, - GLenum internal_format, - GLuint first_level, - GLuint last_level, - GLuint width0, - GLuint height0, - GLuint depth0, - GLuint cpp, - GLuint compress_byte); - -void st_miptree_reference(struct pipe_mipmap_tree **dst, - struct pipe_mipmap_tree *src); - -void st_miptree_release(struct pipe_context *pipe, - struct pipe_mipmap_tree **mt); - -/* Check if an image fits an existing mipmap tree layout - */ -GLboolean st_miptree_match_image(struct pipe_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level); - -/* Return a pointer to an image within a tree. Return image stride as - * well. - */ -GLubyte *st_miptree_image_map(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt, - GLuint face, - GLuint level, - GLuint * row_stride, GLuint * image_stride); - -void st_miptree_image_unmap(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt); - - -/* Return the linear offset of an image relative to the start of the - * tree: - */ -GLuint st_miptree_image_offset(struct pipe_mipmap_tree *mt, - GLuint face, GLuint level); - -/* Return pointers to each 2d slice within an image. Indexed by depth - * value. - */ -const GLuint *st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, - GLuint level); - - -/* Upload an image into a tree - */ -void st_miptree_image_data(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, - GLuint level, - void *src, - GLuint src_row_pitch, GLuint src_image_pitch); - -/* Copy an image between two trees - */ -void st_miptree_image_copy(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, GLuint level, - struct pipe_mipmap_tree *src); - - -#endif -- cgit v1.2.3 From f00179f9b47e17087d546940e1d57ffb2e2a8e42 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 11:37:00 -0600 Subject: Add a quad_stage::begin() method to do per-primitive preparations (like setting up samplers). --- src/mesa/pipe/softpipe/sp_prim_setup.c | 3 ++ src/mesa/pipe/softpipe/sp_quad.h | 2 ++ src/mesa/pipe/softpipe/sp_quad_fs.c | 56 ++++++++++++++++++++++++++-------- 3 files changed, 48 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 4e66f40e47..c8046befa7 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -1008,8 +1008,11 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) static void setup_begin( struct draw_stage *stage ) { struct setup_stage *setup = setup_stage(stage); + struct softpipe_context *sp = setup->softpipe; setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; + + sp->quad.first->begin(sp->quad.first); } diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 72a9981977..cca7dbdcbe 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -41,6 +41,8 @@ struct quad_stage { struct quad_stage *next; + void (*begin)(struct quad_stage *qs); + /** the stage action */ void (*run)(struct quad_stage *qs, struct quad_header *quad); }; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index e55bfd9725..f5240b67f8 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -50,6 +50,23 @@ #define ALIGNED_ATTRIBS 0 #endif + +struct quad_shade_stage +{ + struct quad_stage stage; + struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; +}; + + +/** cast wrapper */ +static INLINE struct quad_shade_stage * +quad_shade_stage(struct quad_stage *qs) +{ + return (struct quad_shade_stage *) qs; +} + + + struct exec_machine { const struct setup_coefficient *coef; /**< will point to quad->coef */ @@ -136,6 +153,7 @@ static INLINE void pinterp( struct exec_machine *exec, static void shade_quad( struct quad_stage *qs, struct quad_header *quad ) { + struct quad_shade_stage *qss = quad_shade_stage(qs); struct softpipe_context *softpipe = qs->softpipe; struct exec_machine exec; const GLfloat fx = quad->x0; @@ -196,7 +214,6 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) struct tgsi_exec_machine machine; struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; struct tgsi_exec_vector *aoutputs; - struct tgsi_sampler samplers[8]; GLuint i; #if !ALIGNED_ATTRIBS @@ -208,18 +225,11 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) memset(&machine, 0, sizeof(machine)); #endif -#if 11 /* temp sampler setup */ - samplers[0].state = &softpipe->sampler[0]; - samplers[0].texture = softpipe->texture[0]; - samplers[0].get_sample = sp_get_sample; - samplers[0].pipe = &softpipe->pipe; -#endif - /* init machine state */ tgsi_exec_machine_init( &machine, softpipe->fs.tokens, - 8, samplers); + PIPE_MAX_SAMPLERS, qss->samplers); /* Consts does not require 16 byte alignment. */ machine.Consts = softpipe->fs.constants->constant; @@ -299,13 +309,33 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) } +/** + * Per-primitive (or per-begin?) setup + */ +static void shade_begin(struct quad_stage *qs) +{ + struct quad_shade_stage *qss = quad_shade_stage(qs); + struct softpipe_context *softpipe = qs->softpipe; + GLuint i; + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + qss->samplers[i].state = &softpipe->sampler[i]; + qss->samplers[i].texture = softpipe->texture[i]; + qss->samplers[i].get_sample = sp_get_sample; + qss->samplers[i].pipe = &softpipe->pipe; + } + + if (qs->next->begin) + qs->next->begin(qs->next); +} + struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) { - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + struct quad_shade_stage *stage = CALLOC_STRUCT(quad_shade_stage); - stage->softpipe = softpipe; - stage->run = shade_quad; + stage->stage.softpipe = softpipe; + stage->stage.begin = shade_begin; + stage->stage.run = shade_quad; - return stage; + return &stage->stage; } -- cgit v1.2.3 From e4eb97318cbce238c5aaaf11af42c33229274859 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 11:44:06 -0600 Subject: add quad_stage::begin() funcs --- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 8 ++++++++ src/mesa/pipe/softpipe/sp_quad_blend.c | 6 ++++++ src/mesa/pipe/softpipe/sp_quad_bufloop.c | 8 ++++++++ src/mesa/pipe/softpipe/sp_quad_colormask.c | 6 ++++++ src/mesa/pipe/softpipe/sp_quad_coverage.c | 8 ++++++++ src/mesa/pipe/softpipe/sp_quad_depth_test.c | 6 ++++++ src/mesa/pipe/softpipe/sp_quad_fs.c | 1 + src/mesa/pipe/softpipe/sp_quad_occlusion.c | 9 +++++++++ src/mesa/pipe/softpipe/sp_quad_output.c | 8 ++++++++ src/mesa/pipe/softpipe/sp_quad_stencil.c | 6 ++++++ src/mesa/pipe/softpipe/sp_quad_stipple.c | 8 ++++++++ 11 files changed, 74 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index 8c28a824be..585adb41fe 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -82,12 +82,20 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) } +static void alpha_test_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} + + struct quad_stage * sp_quad_alpha_test_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = alpha_test_begin; stage->run = alpha_test_quad; return stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 3d097aea65..ad21c27b2f 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -382,6 +382,11 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) } +static void blend_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ) @@ -389,6 +394,7 @@ struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ) struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = blend_begin; stage->run = blend_quad; return stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c index be32d02a46..91c4b70794 100644 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -45,6 +45,13 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) } +static void cbuf_loop_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} + + /** * Create the colorbuffer loop stage. * This is used to implement multiple render targets and GL_FRONT_AND_BACK @@ -55,6 +62,7 @@ struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ) struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = cbuf_loop_begin; stage->run = cbuf_loop_quad; return stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index fff6efa8f6..da5ed8f27c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -71,6 +71,11 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) } +static void colormask_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ) @@ -78,6 +83,7 @@ struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ) struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = colormask_begin; stage->run = colormask_quad; return stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c index cdd8890c7f..f166276e84 100644 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -63,11 +63,19 @@ coverage_quad(struct quad_stage *qs, struct quad_header *quad) } +static void coverage_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} + + struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = coverage_begin; stage->run = coverage_quad; return stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 3a8df33e67..904d93a614 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -156,6 +156,11 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } +static void depth_test_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ) @@ -163,6 +168,7 @@ struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ) struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = depth_test_begin; stage->run = depth_test_quad; return stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index f5240b67f8..5e4782ca48 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -322,6 +322,7 @@ static void shade_begin(struct quad_stage *qs) qss->samplers[i].texture = softpipe->texture[i]; qss->samplers[i].get_sample = sp_get_sample; qss->samplers[i].pipe = &softpipe->pipe; + /* init cache info here */ } if (qs->next->begin) diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c index 843c462d48..0fc1acbf2f 100644 --- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c +++ b/src/mesa/pipe/softpipe/sp_quad_occlusion.c @@ -56,11 +56,20 @@ occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad) } +static void occlusion_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} + + + struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = occlusion_begin; stage->run = occlusion_count_quad; return stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index 12ab1eca1c..49161993e6 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -84,11 +84,19 @@ output_quad(struct quad_stage *qs, struct quad_header *quad) } +static void output_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} + + struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = output_begin; stage->run = output_quad; return stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 0b37474c1a..6a9ea9069d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -275,6 +275,11 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) } +static void stencil_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ) @@ -282,6 +287,7 @@ struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = stencil_begin; stage->run = stencil_test_quad; return stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index cad1a1400c..286eb1cd95 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -36,12 +36,20 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) } +static void stipple_begin(struct quad_stage *qs) +{ + if (qs->next->begin) + qs->next->begin(qs->next); +} + + struct quad_stage * sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); stage->softpipe = softpipe; + stage->begin = stipple_begin; stage->run = stipple_quad; return stage; -- cgit v1.2.3 From c8cc1e86f68a54198e3136dcd198a252f6d7e8f2 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 12:02:18 -0600 Subject: fix qs->next tests --- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 2 +- src/mesa/pipe/softpipe/sp_quad_blend.c | 2 +- src/mesa/pipe/softpipe/sp_quad_bufloop.c | 2 +- src/mesa/pipe/softpipe/sp_quad_colormask.c | 2 +- src/mesa/pipe/softpipe/sp_quad_coverage.c | 2 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 4 +++- src/mesa/pipe/softpipe/sp_quad_occlusion.c | 2 +- src/mesa/pipe/softpipe/sp_quad_output.c | 2 +- src/mesa/pipe/softpipe/sp_quad_stencil.c | 2 +- src/mesa/pipe/softpipe/sp_quad_stipple.c | 2 +- 11 files changed, 13 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index 585adb41fe..1c51907078 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -84,7 +84,7 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) static void alpha_test_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index ad21c27b2f..09d0aa258c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -384,7 +384,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) static void blend_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c index 91c4b70794..c459a9e8b0 100644 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -47,7 +47,7 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) static void cbuf_loop_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index da5ed8f27c..5102a000b5 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -73,7 +73,7 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) static void colormask_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c index f166276e84..efeb85aab9 100644 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -65,7 +65,7 @@ coverage_quad(struct quad_stage *qs, struct quad_header *quad) static void coverage_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 904d93a614..28f264b3c4 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -158,7 +158,7 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) static void depth_test_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 5e4782ca48..e5ab0ebc16 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -323,9 +323,11 @@ static void shade_begin(struct quad_stage *qs) qss->samplers[i].get_sample = sp_get_sample; qss->samplers[i].pipe = &softpipe->pipe; /* init cache info here */ + qss->samplers[i].cache_x = + qss->samplers[i].cache_y = -1; } - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c index 0fc1acbf2f..8073f94553 100644 --- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c +++ b/src/mesa/pipe/softpipe/sp_quad_occlusion.c @@ -58,7 +58,7 @@ occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad) static void occlusion_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index 49161993e6..62f466be5d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -86,7 +86,7 @@ output_quad(struct quad_stage *qs, struct quad_header *quad) static void output_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 6a9ea9069d..857f1a5989 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -277,7 +277,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) static void stencil_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index 286eb1cd95..6e93bf81f1 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -38,7 +38,7 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) static void stipple_begin(struct quad_stage *qs) { - if (qs->next->begin) + if (qs->next) qs->next->begin(qs->next); } -- cgit v1.2.3 From 267cf16e066f548a33c7b5f7c73deb59039d9b3d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 12:02:52 -0600 Subject: very simple texture cache implementation --- src/mesa/pipe/softpipe/sp_tex_sample.c | 26 ++++++++++++++++++++++++-- src/mesa/pipe/tgsi/core/tgsi_exec.h | 5 +++++ 2 files changed, 29 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 1d274d5038..7f8bf0d99c 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -471,12 +471,34 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, switch (sampler->state->min_filter) { case PIPE_TEX_FILTER_NEAREST: { - GLint x, y; + GLint x, y, cx, cy; + const GLfloat *src; x = nearest_texcoord(sampler->state->wrap_s, strq[0], sampler->texture->width0); y = nearest_texcoord(sampler->state->wrap_t, strq[1], sampler->texture->height0); - ps->get_tile(ps, x, y, 1, 1, rgba); + + cx = x / SAMPLER_CACHE_SIZE; + cy = y / SAMPLER_CACHE_SIZE; + if (cx != sampler->cache_x || cy != sampler->cache_y) { + /* cache miss, replace cache with new tile */ + sampler->cache_x = cx; + sampler->cache_y = cy; + ps->get_tile(ps, + cx * SAMPLER_CACHE_SIZE, + cy * SAMPLER_CACHE_SIZE, + SAMPLER_CACHE_SIZE, SAMPLER_CACHE_SIZE, + (GLfloat *) sampler->cache); + /*printf("cache miss (%d, %d)\n", x, y);*/ + } + else { + /*printf("cache hit (%d, %d)\n", x, y);*/ + } + + cx = x % SAMPLER_CACHE_SIZE; + cy = y % SAMPLER_CACHE_SIZE; + src = sampler->cache[cy][cx]; + COPY_4V(rgba, src); } break; case PIPE_TEX_FILTER_LINEAR: diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index 2ad4b965bd..fb11eec777 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -21,6 +21,8 @@ struct tgsi_exec_vector union tgsi_exec_channel xyzw[4]; }; +#define SAMPLER_CACHE_SIZE 8 + struct tgsi_sampler { const struct pipe_sampler_state *state; @@ -28,6 +30,9 @@ struct tgsi_sampler void (*get_sample)(struct tgsi_sampler *sampler, const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]); void *pipe; /*XXX temporary*/ + + GLint cache_x, cache_y; + GLfloat cache[SAMPLER_CACHE_SIZE][SAMPLER_CACHE_SIZE][4]; }; struct tgsi_exec_labels -- cgit v1.2.3 From 7f74538173b78281b69f8b18a975e64d2c934245 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 11:25:28 +0100 Subject: Remove unused code, definitions. --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 1 + src/mesa/drivers/dri/intel_winsys/intel_context.h | 40 --------------------- .../drivers/dri/intel_winsys/intel_pipe_softpipe.c | 5 +-- src/mesa/drivers/dri/intel_winsys/intel_reg.h | 41 ++-------------------- 4 files changed, 5 insertions(+), 82 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 60bd8eaec2..a85dc0b613 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -27,6 +27,7 @@ #include "intel_batchbuffer.h" #include "intel_ioctl.h" +#include "intel_reg.h" /* Relocations in kernel space: * - pass dma buffer seperately diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 727f3b7474..0e040f7922 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -40,21 +40,12 @@ #include "i830_common.h" -#define DV_PF_555 (1<<8) -#define DV_PF_565 (2<<8) -#define DV_PF_8888 (3<<8) - struct pipe_context; struct pipe_region; struct intel_context; struct _DriBufferObject; -#define INTEL_WRITE_PART 0x1 -#define INTEL_WRITE_FULL 0x2 -#define INTEL_READ 0x4 - - #define INTEL_MAX_FIXUP 64 @@ -76,9 +67,6 @@ struct intel_context char *prevLockFile; int prevLockLine; - GLuint ClearColor565; - GLuint ClearColor8888; - /* These refer to the current drawing buffer: */ @@ -117,25 +105,6 @@ extern char *__progname; */ #define INTEL_FIREVERTICES(intel) -/* ================================================================ - * Color packing: - */ - -#define INTEL_PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define INTEL_PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define INTEL_PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define INTEL_PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - /* ================================================================ * Debugging: @@ -182,20 +151,11 @@ extern int INTEL_DEBUG; * intel_context.c: */ -extern GLboolean intelInitContext(struct intel_context *intel, - const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate, - struct dd_function_table *functions); - extern void intelGetLock(struct intel_context *intel, GLuint flags); extern void intelFinish(GLcontext * ctx); extern void intelFlush(GLcontext * ctx); - -#define MI_BATCH_BUFFER_END (0xA<<23) - #define FALLBACK( ctx, bit, mode ) /*====================================================================== diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c index ef47744358..5edbca1098 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c @@ -34,11 +34,8 @@ #include "dri_bufpool.h" #include "dri_bufmgr.h" -/* #include "errno.h" */ -/* #include "string.h" */ -/* #include "imports.h" */ - #include "intel_context.h" +#include "intel_pipe.h" #include "pipe/softpipe/sp_winsys.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_reg.h b/src/mesa/drivers/dri/intel_winsys/intel_reg.h index 7828ba6ad3..f37c24fda9 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_reg.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_reg.h @@ -30,44 +30,6 @@ #define _INTEL_REG_H_ - -#define CMD_3D (0x3<<29) - - -#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) -#define PRIM_INDIRECT (1<<23) -#define PRIM_INLINE (0<<23) -#define PRIM_INDIRECT_SEQUENTIAL (0<<17) -#define PRIM_INDIRECT_ELTS (1<<17) - -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_MASK (0x1f<<18) - -#define I915PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define I915PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define I915PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define I915PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - - #define BR00_BITBLT_CLIENT 0x40000000 #define BR00_OP_COLOR_BLT 0x10000000 #define BR00_OP_SRC_COPY_BLT 0x10C00000 @@ -85,4 +47,7 @@ #define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) #define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) +#define MI_BATCH_BUFFER_END (0xA<<23) + + #endif -- cgit v1.2.3 From db068269fcba8febf92be9d6cd54a9315e6bbf04 Mon Sep 17 00:00:00 2001 From: keithw Date: Wed, 8 Aug 2007 11:32:24 +0100 Subject: remove unused file --- src/mesa/pipe/i915simple/i915_state_draw.c | 37 ------------------------------ 1 file changed, 37 deletions(-) delete mode 100644 src/mesa/pipe/i915simple/i915_state_draw.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_draw.c b/src/mesa/pipe/i915simple/i915_state_draw.c deleted file mode 100644 index 6c436b4c06..0000000000 --- a/src/mesa/pipe/i915simple/i915_state_draw.c +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "imports.h" - -#include "sp_context.h" -#include "sp_state.h" -#include "pipe/draw/draw_context.h" - - - -- cgit v1.2.3 From c28d2e6702f8ea165309b999388b9e34253ceb29 Mon Sep 17 00:00:00 2001 From: keithw Date: Wed, 8 Aug 2007 11:32:56 +0100 Subject: remove unused file --- src/mesa/pipe/i915simple/i915_state_vertex.c | 145 --------------------------- 1 file changed, 145 deletions(-) delete mode 100644 src/mesa/pipe/i915simple/i915_state_vertex.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_vertex.c b/src/mesa/pipe/i915simple/i915_state_vertex.c deleted file mode 100644 index 053e39deb2..0000000000 --- a/src/mesa/pipe/i915simple/i915_state_vertex.c +++ /dev/null @@ -1,145 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "main/glheader.h" -#include "main/macros.h" -#include "main/enums.h" - -#include "vf/vf.h" -#include "pipe/draw/draw_context.h" -#include "i915_context.h" -#include "i915_state.h" - - - - -/*********************************************************************** - * - */ - -#define SZ_TO_HW(sz) ((sz-2)&0x3) -#define EMIT_SZ(sz) (EMIT_1F + (sz) - 1) -#define EMIT_ATTR( ATTR, STYLE, S4, SZ ) \ -do { \ - i915->vertex_attrs[i915->vertex_attr_count].attrib = (ATTR); \ - i915->vertex_attrs[i915->vertex_attr_count].format = (STYLE); \ - i915->vertex_attr_count++; \ - s4 |= S4; \ - offset += (SZ); \ -} while (0) - -#define EMIT_PAD( N ) \ -do { \ - i915->vertex_attrs[i915->vertex_attr_count].attrib = 0; \ - i915->vertex_attrs[i915->vertex_attr_count].format = EMIT_PAD; \ - i915->vertex_attrs[i915->vertex_attr_count].offset = (N); \ - i915->vertex_attr_count++; \ - offset += (N); \ -} while (0) - - - -/** - * Determine which post-transform / pre-rasterization vertex attributes - * we need. - * Derived from: fs, setup states. - */ -static void calculate_vertex_layout( struct i915_context *i915 ) -{ - const GLbitfield inputsRead = i915->fs.inputs_read; - GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; - GLuint nr_attrs = 0; - GLuint i; - - - - memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); - - EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); - EMIT_ATTR(VF_ATTRIB_COLOR0, FRAG_ATTRIB_COL0, INTERP_LINEAR); - - i915->hw_vertex_size = i915->nr_attrs; - - /* Additional attributes required for setup: Just twosided - * lighting. Edgeflag is dealt with specially by setting bits in - * the vertex header. - */ - if (i915->setup.light_twoside) { - EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ - } - - /* If the attributes have changed, tell the draw module (which in turn - * tells the vf module) about the new vertex layout. - */ - draw_set_vertex_attributes( i915->draw, - slot_to_vf_attr, - i915->nr_attrs ); -} - - - -/*********************************************************************** - * - */ -static inline GLuint attr_size(GLuint sizes, GLuint attr) -{ - return ((sizes >> (attr*2)) & 0x3) + 1; -} - -static void i915_calculate_vertex_format( struct intel_context *intel ) -{ - struct i915_context *i915 = i915_context( &intel->ctx ); - - GLuint s2 = S2_TEXCOORD_NONE; - GLuint s4 = 0; - GLuint offset = 0; - GLuint i; - i915->vertex_attr_count = 0; - - EMIT_ATTR(VF_ATTRIB_POS, EMIT_4F_VIEWPORT, S4_VFMT_XYZW, 16); - EMIT_ATTR(VF_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, S4_VFMT_COLOR, 4); - - - - if (s2 != i915->vertex_format.LIS2 || - s4 != i915->vertex_format.LIS4) { - - clip_set_hw_vertex_format( intel->clip, - i915->vertex_attrs, - i915->vertex_attr_count, - offset ); - - /* Needed? This does raise the INTEL_NEW_VERTEX_SIZE flag: - */ - intel_vb_set_vertex_size( intel->vb, offset ); - - i915->vertex_format.LIS2 = s2; - i915->vertex_format.LIS4 = s4; - intel->state.dirty.intel |= I915_NEW_VERTEX_FORMAT; - } -} - -- cgit v1.2.3 From 7bf2e9e38329a332281011520413de720906447c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 11:35:14 +0100 Subject: Fix typo so that i915simple actually gets used. --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index d748c08f3c..e8b330e503 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -386,6 +386,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, case PCI_CHIP_I915_G: case PCI_CHIP_I915_GM: intel->pipe = intel_create_i915simple( intel ); + break; default: _mesa_printf("Unknown PCIID %x in %s, using software driver\n", intel->intelScreen->deviceID, __FUNCTION__); -- cgit v1.2.3 From cf3fb4e3ef16ec1ab08fdbe418afe15c48df73fc Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 11:35:47 +0100 Subject: Call init_state_funcs --- src/mesa/pipe/i915simple/i915_context.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 1a21e931ae..87fccee9e5 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -169,6 +169,7 @@ struct pipe_context *i915_create( struct i915_winsys *winsys ) i915_init_buffer_functions(i915); i915_init_region_functions(i915); i915_init_surface_functions(i915); + i915_init_state_functions(i915); /* * XXX we could plug GL selection/feedback into the drawing pipeline -- cgit v1.2.3 From 715acc7622255f7cc99d632c57aac14af4bac89a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 12:28:21 +0100 Subject: The i915simple driver now runs well enough to lock up hardware. --- .../dri/intel_winsys/intel_pipe_i915simple.c | 16 ++++++++++++--- src/mesa/pipe/i915simple/i915_context.c | 4 ++-- src/mesa/pipe/i915simple/i915_prim_emit.c | 23 +++++++++++----------- src/mesa/pipe/i915simple/i915_state_derived.c | 2 +- src/mesa/pipe/i915simple/i915_state_emit.c | 13 +++++++----- 5 files changed, 36 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c index b456550a8c..244e32c885 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c @@ -168,7 +168,7 @@ static unsigned *intel_i915_batch_start( struct i915_winsys *sws, /* XXX: check relocs. */ - if (intel_batchbuffer_space( intel->batch ) < dwords * 4) { + if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { /* XXX: Hmm, the driver can't really do much with this pointer: */ //return intel->batch->ptr; @@ -191,8 +191,18 @@ static void intel_i915_batch_reloc( struct i915_winsys *sws, unsigned delta ) { struct intel_context *intel = intel_i915_winsys(sws)->intel; - unsigned flags = 0; - unsigned mask = 0; + unsigned flags = DRM_BO_FLAG_MEM_TT; + unsigned mask = DRM_BO_MASK_MEM; + + if (access_flags & I915_BUFFER_ACCESS_WRITE) { + flags |= DRM_BO_FLAG_WRITE; + mask |= DRM_BO_FLAG_WRITE; + } + + if (access_flags & I915_BUFFER_ACCESS_READ) { + flags |= DRM_BO_FLAG_READ; + mask |= DRM_BO_FLAG_READ; + } intel_batchbuffer_emit_reloc( intel->batch, dri_bo( buf ), diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 87fccee9e5..f39c282981 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -119,8 +119,8 @@ static void i915_draw_vb( struct pipe_context *pipe, { struct i915_context *i915 = i915_context( pipe ); -// if (i915->dirty) -// i915_update_derived( i915 ); + if (i915->dirty) + i915_update_derived( i915 ); draw_vb( i915->draw, VB ); } diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index 97e337451e..6d7cfe2e3e 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -35,6 +35,7 @@ #include "i915_winsys.h" #include "i915_reg.h" #include "i915_state.h" +#include "i915_batch.h" @@ -91,17 +92,17 @@ static inline unsigned char float_to_ubyte( float f ) /* Hardcoded vertex format: xyz/rgba */ static inline void -emit_hw_vertex( unsigned *ptr, +emit_hw_vertex( struct i915_context *i915, struct vertex_header *vertex ) { - ptr[0] = fui( vertex->data[0][0] ); - ptr[1] = fui( vertex->data[0][1] ); - ptr[2] = fui( vertex->data[0][2] ); - - ptr[3] = pack_ub4( float_to_ubyte( vertex->data[1][0] ), - float_to_ubyte( vertex->data[1][1] ), - float_to_ubyte( vertex->data[1][2] ), - float_to_ubyte( vertex->data[1][3] ) ); + OUT_BATCH( fui(vertex->data[0][0]) ); + OUT_BATCH( fui(vertex->data[0][1]) ); + OUT_BATCH( fui(vertex->data[0][2]) ); + + OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[1][0] ), + float_to_ubyte( vertex->data[1][1] ), + float_to_ubyte( vertex->data[1][2] ), + float_to_ubyte( vertex->data[1][3] )) ); } @@ -134,12 +135,12 @@ emit_prim( struct draw_stage *stage, /* Emit each triangle as a single primitive. I told you this was * simple. */ - *ptr++ = (_3DPRIMITIVE | + OUT_BATCH(_3DPRIMITIVE | hwprim | ((4 + vertex_size * nr)/4 - 2)); for (i = 0; i < nr; i++) { - emit_hw_vertex(ptr, prim->v[i]); + emit_hw_vertex(i915, prim->v[i]); ptr += vertex_size / sizeof(int); } } diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 392ac09b40..6e91e4448f 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -76,7 +76,7 @@ static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = static void calculate_vertex_layout( struct i915_context *i915 ) { // const GLbitfield inputsRead = i915->fs.inputs_read; - const GLbitfield inputsRead = (FRAG_ATTRIB_WPOS | FRAG_ATTRIB_COL0); + const GLbitfield inputsRead = (FRAG_BIT_WPOS | FRAG_BIT_COL0); GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; GLbitfield attr_mask = 0x0; GLuint nr_attrs = 0; diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index c0539daa6e..4aa7e28a86 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -42,7 +42,7 @@ static unsigned translate_format( unsigned format ) { switch (format) { - case PIPE_FORMAT_U_R8_G8_B8_A8: + case PIPE_FORMAT_U_A8_R8_G8_B8: return COLOR_BUF_ARGB8888; case PIPE_FORMAT_U_R5_G6_B5: return COLOR_BUF_RGB565; @@ -212,8 +212,11 @@ i915_emit_hardware_state(struct i915_context *i915 ) { - unsigned cformat = i915->framebuffer.cbufs[0]->format; - unsigned zformat = i915->framebuffer.zbuf->format; + unsigned cformat = translate_format( i915->framebuffer.cbufs[0]->format ); + unsigned zformat = 0; + + if (i915->framebuffer.zbuf) + zformat = translate_depth_format( i915->framebuffer.zbuf->format ); OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); @@ -221,8 +224,8 @@ i915_emit_hardware_state(struct i915_context *i915 ) DSTORG_VERT_BIAS(0x8) | /* .5 */ LOD_PRECLAMP_OGL | TEX_DEFAULT_COLOR_OGL | - translate_format( cformat ) | - translate_depth_format( zformat )); + cformat | + zformat ); } { -- cgit v1.2.3 From 65f7b3834b62f3ab7e83a9f290cc6a66cda60c05 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 12:32:06 +0100 Subject: Don't use hardware driver by default, until its working a bit better. --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index e8b330e503..66be3af72b 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -372,7 +372,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, /* * Pipe-related setup */ - if (getenv("INTEL_SOFTPIPE")) { + if (!getenv("INTEL_HW")) { intel->pipe = intel_create_softpipe( intel ); } else { -- cgit v1.2.3 From 960bf97c9f9cd98c0a26f26a2fa23960ae96aa6e Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 8 Aug 2007 13:55:56 +0100 Subject: Add D3D9 opcodes up to ps_2_x and vs_2_x. --- src/mesa/pipe/tgsi/core/tgsi_token.h | 317 ++++++++++++++++++++++++++++++++++- 1 file changed, 316 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index fd59ea1e2e..1950dfc0ad 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -732,7 +732,322 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_TXL */ /* TGSI_OPCODE_XPD */ -#define TGSI_OPCODE_LAST 107 +/* + * ps_1_1 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_LRP */ +#define TGSI_OPCODE_TEXCOORD TGSI_OPCODE_NOP +#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL +/* TGSI_OPCODE_TEX */ +#define TGSI_OPCODE_TEXBEM 107 +#define TGSI_OPCODE_TEXBEML 108 +#define TGSI_OPCODE_TEXREG2AR 109 +#define TGSI_OPCODE_TEXM3X2PAD 110 +#define TGSI_OPCODE_TEXM3X2TEX 111 +#define TGSI_OPCODE_TEXM3X3PAD 112 +#define TGSI_OPCODE_TEXM3X3TEX 113 +#define TGSI_OPCODE_TEXM3X3SPEC 114 +#define TGSI_OPCODE_TEXM3X3VSPEC 115 +/* TGSI_OPCODE_CND */ + +/* + * ps_1_2 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_TEXCOORD */ +/* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TEXBEM */ +/* TGSI_OPCODE_TEXBEML */ +/* TGSI_OPCODE_TEXREG2AR */ +#define TGSI_OPCODE_TEXREG2GB 116 +/* TGSI_OPCODE_TEXM3X2PAD */ +/* TGSI_OPCODE_TEXM3X2TEX */ +/* TGSI_OPCODE_TEXM3X3PAD */ +/* TGSI_OPCODE_TEXM3X3TEX */ +/* TGSI_OPCODE_TEXM3X3SPEC */ +/* TGSI_OPCODE_TEXM3X3VSPEC */ +/* TGSI_OPCODE_CND */ +#define TGSI_OPCODE_TEXREG2RGB 117 +#define TGSI_OPCODE_TEXDP3TEX 118 +#define TGSI_OPCODE_TEXDP3 119 +#define TGSI_OPCODE_TEXM3X3 120 +/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ + +/* + * ps_1_3 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_TEXCOORD */ +/* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TEXBEM */ +/* TGSI_OPCODE_TEXBEML */ +/* TGSI_OPCODE_TEXREG2AR */ +/* TGSI_OPCODE_TEXREG2GB */ +/* TGSI_OPCODE_TEXM3X2PAD */ +/* TGSI_OPCODE_TEXM3X2TEX */ +/* TGSI_OPCODE_TEXM3X3PAD */ +/* TGSI_OPCODE_TEXM3X3TEX */ +/* TGSI_OPCODE_TEXM3X3SPEC */ +/* TGSI_OPCODE_TEXM3X3VSPEC */ +/* TGSI_OPCODE_CND */ +/* TGSI_OPCODE_TEXREG2RGB */ +/* TGSI_OPCODE_TEXDP3TEX */ +#define TEXM3X2DEPTH 121 +/* TGSI_OPCODE_TEXDP3 */ +/* TGSI_OPCODE_TEXM3X3 */ +/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ + +/* + * ps_1_4 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_CND */ +#define TGSI_OPCODE_TEXDEPTH 122 +/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ +#define TGSI_OPCODE_BEM 123 + +/* + * ps_2_0 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_FRC */ +#define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX +#define TGSI_OPCODE_M4X3 124 +#define TGSI_OPCODE_M3X4 125 +#define TGSI_OPCODE_M3X3 126 +#define TGSI_OPCODE_M3X2 127 +/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ +#define TGSI_OPCODE_CRS TGSI_OPCODE_XPD +/* TGSI_OPCODE_ABS */ +#define TGSI_OPCODE_NRM4 128 +#define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS +/* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ +#define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A + +/* + * ps_2_x + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_M4X4 */ +/* TGSI_OPCODE_M4X3 */ +/* TGSI_OPCODE_M3X4 */ +/* TGSI_OPCODE_M3X3 */ +/* TGSI_OPCODE_M3X2 */ +#define TGSI_OPCODE_CALL TGSI_OPCODE_CAL +#define TGSI_OPCODE_CALLNZ 129 +/* TGSI_OPCODE_RET */ +/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ +/* TGSI_OPCODE_CRS */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_NRM4 */ +/* TGSI_OPCODE_SINCOS */ +/* TGSI_OPCODE_REP */ +/* TGSI_OPCODE_ENDREP */ +/* TGSI_OPCODE_IF */ +#define TGSI_OPCODE_IFC 130 +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +/* TGSI_OPCODE_BREAK */ +#define TGSI_OPCODE_BREAKC 131 +/* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ +/* TGSI_OPCODE_DP2ADD */ +#define TGSI_OPCODE_DSX TGSI_OPCODE_DDX +#define TGSI_OPCODE_DSY TGSI_OPCODE_DDY +#define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD + +/* + * vs_1_1 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_M4X4 */ +/* TGSI_OPCODE_M4X3 */ +/* TGSI_OPCODE_M3X4 */ +/* TGSI_OPCODE_M3X3 */ +/* TGSI_OPCODE_M3X2 */ +#define TGSI_OPCODE_EXPP 132 +#define TGSI_OPCODE_LOGP TGSI_OPCODE_LG2 + +/* + * vs_2_0 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_M4X4 */ +/* TGSI_OPCODE_M4X3 */ +/* TGSI_OPCODE_M3X4 */ +/* TGSI_OPCODE_M3X3 */ +/* TGSI_OPCODE_M3X2 */ +/* TGSI_OPCODE_CALL */ +/* TGSI_OPCODE_CALLNZ */ +/* TGSI_OPCODE_LOOP */ +/* TGSI_OPCODE_RET */ +/* TGSI_OPCODE_ENDLOOP */ +/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ +/* TGSI_OPCODE_CRS */ +/* TGSI_OPCODE_SGN */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_NRM4 */ +/* TGSI_OPCODE_SINCOS */ +/* TGSI_OPCODE_REP */ +/* TGSI_OPCODE_ENDREP */ +/* TGSI_OPCODE_IF */ +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +#define TGSI_OPCODE_MOVA TGSI_OPCODE_ARR +/* TGSI_OPCODE_LOGP */ + +/* + * vs_2_x + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_M4X4 */ +/* TGSI_OPCODE_M4X3 */ +/* TGSI_OPCODE_M3X4 */ +/* TGSI_OPCODE_M3X3 */ +/* TGSI_OPCODE_M3X2 */ +/* TGSI_OPCODE_CALL */ +/* TGSI_OPCODE_CALLNZ */ +/* TGSI_OPCODE_LOOP */ +/* TGSI_OPCODE_RET */ +/* TGSI_OPCODE_ENDLOOP */ +/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ +/* TGSI_OPCODE_CRS */ +/* TGSI_OPCODE_SGN */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_NRM4 */ +/* TGSI_OPCODE_SINCOS */ +/* TGSI_OPCODE_REP */ +/* TGSI_OPCODE_ENDREP */ +/* TGSI_OPCODE_IF */ +/* TGSI_OPCODE_IFC */ +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +/* TGSI_OPCODE_BREAK */ +/* TGSI_OPCODE_BREAKC */ +/* TGSI_OPCODE_MOVA */ +/* TGSI_OPCODE_LOGP */ + +#define TGSI_OPCODE_LAST 133 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ -- cgit v1.2.3 From 31132546891e0fd0f0fc889506fb5e46c0075c4c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 16:06:48 -0600 Subject: new assertions --- src/mesa/drivers/dri/intel_winsys/intel_surface.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.c b/src/mesa/drivers/dri/intel_winsys/intel_surface.c index 3fa40271c8..d2e2dabade 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.c @@ -40,6 +40,8 @@ read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLuint j; assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); + assert(x < sps->surface.width - 1); + assert(y < sps->surface.height - 1); memcpy(temp + 8, src, 8); memcpy(temp + 0, src + bytesPerRow, 8); -- cgit v1.2.3 From 6dc212ce49bd2939e905ea0bb8f2fa162cd67df6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 16:07:07 -0600 Subject: print_vertex func for debugging --- src/mesa/pipe/softpipe/sp_prim_setup.c | 20 +++++++++++++++++++- 1 file changed, 19 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index c8046befa7..7a3d011b7b 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -236,6 +236,18 @@ static void flush_spans( struct setup_stage *setup ) setup->span.right[1] = 0; } +#if 0 +static void print_vertex(const struct setup_stage *setup, + const struct vertex_header *v) +{ + int i; + printf("Vertex:\n"); + for (i = 0; i < setup->softpipe->nr_attrs; i++) { + printf(" %d: %f %f %f\n", i, + v->data[i][0], v->data[i][1], v->data[i][2]); + } +} +#endif static GLboolean setup_sort_vertices( struct setup_stage *setup, const struct prim_header *prim ) @@ -244,6 +256,13 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, const struct vertex_header *v1 = prim->v[1]; const struct vertex_header *v2 = prim->v[2]; +#if 0 + printf("Triangle:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); + print_vertex(setup, v2); +#endif + setup->vprovoke = v2; /* determine bottom to top order of vertices */ @@ -429,7 +448,6 @@ static void tri_persp_coeff( struct setup_stage *setup, } - /** * Compute the setup->coef[] array dadx, dady, a0 values. * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. -- cgit v1.2.3 From ed227b7b8ddc3aad6ed764124e94caae6b98d182 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 16:07:28 -0600 Subject: added assertion --- src/mesa/pipe/softpipe/sp_surface.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 85e1fcfa38..174dcc0169 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -323,6 +323,8 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, const GLuint *src = ((const GLuint *) ps->region->map) + y * ps->region->pitch + x; GLuint i, j; + assert(x + w <= ps->width); + assert(y + h <= ps->height); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { p[0] = UBYTE_TO_FLOAT((src[j] >> 16) & 0xff); -- cgit v1.2.3 From 7faa3542f062dfa32e1596f5ce2b531cb8b4eeef Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 16:08:45 -0600 Subject: setup frag shader state: color pass-through program --- src/mesa/state_tracker/st_cb_clear.c | 64 ++++++++++++++++++++++++++++++++++-- 1 file changed, 62 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 0ec7784d84..d862f7ba46 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -31,18 +31,22 @@ * Brian Paul */ -#include "glheader.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/macros.h" +#include "shader/prog_instruction.h" #include "st_atom.h" #include "st_context.h" #include "st_cb_clear.h" +#include "st_program.h" #include "st_public.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "vf/vf.h" + static GLuint color_value(GLuint pipeFormat, const GLfloat color[4]) { @@ -101,6 +105,49 @@ is_depth_stencil_format(GLuint pipeFormat) } + +/** + * Create a simple fragment shader that just passes through the fragment color. + */ +static struct st_fragment_program * +make_color_shader(struct st_context *st) +{ + GLcontext *ctx = st->ctx; + struct st_fragment_program *stfp; + struct gl_program *p; + GLboolean b; + + p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + if (!p) + return NULL; + + p->NumInstructions = 2; + p->Instructions = _mesa_alloc_instructions(2); + if (!p->Instructions) { + ctx->Driver.DeleteProgram(ctx, p); + return NULL; + } + _mesa_init_instructions(p->Instructions, 2); + p->Instructions[0].Opcode = OPCODE_MOV; + p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[0].DstReg.Index = 0; + p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[0].SrcReg[0].Index = FRAG_ATTRIB_COL0; + p->Instructions[1].Opcode = OPCODE_END; + + p->InputsRead = FRAG_BIT_COL0; + p->OutputsWritten = (1 << FRAG_RESULT_COLR); + + stfp = (struct st_fragment_program *) p; + /* compile into tgsi format */ + b = tgsi_mesa_compile_fp_program(&stfp->Base, + stfp->tokens, ST_FP_MAX_TOKENS); + assert(b); + + return stfp; +} + + /** * Draw a screen-aligned quadrilateral. * Coords are window coords. @@ -154,12 +201,14 @@ clear_with_quad(GLcontext *ctx, GLuint x0, GLuint y0, GLuint x1, GLuint y1, GLboolean color, GLboolean depth, GLboolean stencil) { + static struct st_fragment_program *stfp = NULL; struct st_context *st = ctx->st; struct pipe_alpha_test_state alpha_test; struct pipe_blend_state blend; struct pipe_depth_state depth_test; struct pipe_stencil_state stencil_test; struct pipe_setup_state setup; + struct pipe_fs_state fs; /* alpha state: disabled */ memset(&alpha_test, 0, sizeof(alpha_test)); @@ -210,6 +259,16 @@ clear_with_quad(GLcontext *ctx, GLuint x0, GLuint y0, } st->pipe->set_stencil_state(st->pipe, &stencil_test); + /* fragment shader state: color pass-through program */ + if (!stfp) { + stfp = make_color_shader(st); + } + memset(&fs, 0, sizeof(fs)); + fs.inputs_read = stfp->Base.Base.InputsRead; + fs.tokens = &stfp->tokens[0]; + fs.constants = NULL; + st->pipe->set_fs_state(st->pipe, &fs); + /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); @@ -217,6 +276,7 @@ clear_with_quad(GLcontext *ctx, GLuint x0, GLuint y0, st->pipe->set_alpha_test_state(st->pipe, &st->state.alpha_test); st->pipe->set_blend_state(st->pipe, &st->state.blend); st->pipe->set_depth_state(st->pipe, &st->state.depth); + st->pipe->set_fs_state(st->pipe, &st->state.fs); st->pipe->set_setup_state(st->pipe, &st->state.setup); st->pipe->set_stencil_state(st->pipe, &st->state.stencil); /* OR: -- cgit v1.2.3 From 09a1b912605ff48c8782dcc5aae55ac77e27037b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 16:09:13 -0600 Subject: initial lod/mipmap texture supported --- src/mesa/pipe/softpipe/sp_quad_fs.c | 1 + src/mesa/pipe/softpipe/sp_tex_sample.c | 53 +++++++++++++++++++++++++++++----- src/mesa/pipe/tgsi/core/tgsi_exec.h | 2 +- 3 files changed, 47 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index e5ab0ebc16..b68d0f95ac 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -325,6 +325,7 @@ static void shade_begin(struct quad_stage *qs) /* init cache info here */ qss->samplers[i].cache_x = qss->samplers[i].cache_y = -1; + qss->samplers[i].cache_level = -1; } if (qs->next) diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 7f8bf0d99c..062d61dea4 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -447,6 +447,22 @@ sp_get_sample_1d(struct tgsi_sampler *sampler, } } +static GLuint +choose_mipmap_level(struct tgsi_sampler *sampler, GLfloat lambda) +{ + if (sampler->state->min_filter == sampler->state->mag_filter) { + assert(sampler->state->min_filter == PIPE_TEX_FILTER_LINEAR || + sampler->state->min_filter == PIPE_TEX_FILTER_NEAREST); + return 0; + } + else { + GLint level = (int) lambda; + level = CLAMP(level, sampler->texture->first_level, + sampler->texture->last_level); + return level; + } +} + /** * Called via tgsi_sampler::get_sample() @@ -465,14 +481,22 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]) { struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; - struct pipe_surface *ps - = pipe->get_tex_surface(pipe, sampler->texture, 0, 0, 0); + GLuint filter; + GLint level0; - switch (sampler->state->min_filter) { + if (lambda < 0.0) + filter = sampler->state->mag_filter; + else + filter = sampler->state->min_filter; + + level0 = choose_mipmap_level(sampler, lambda); + + switch (filter) { case PIPE_TEX_FILTER_NEAREST: + case PIPE_TEX_FILTER_NEAREST_MIPMAP_NEAREST: { - GLint x, y, cx, cy; - const GLfloat *src; + GLint x, y, cx, cy; + x = nearest_texcoord(sampler->state->wrap_s, strq[0], sampler->texture->width0); y = nearest_texcoord(sampler->state->wrap_t, strq[1], @@ -480,8 +504,12 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, cx = x / SAMPLER_CACHE_SIZE; cy = y / SAMPLER_CACHE_SIZE; - if (cx != sampler->cache_x || cy != sampler->cache_y) { + if (cx != sampler->cache_x || cy != sampler->cache_y || + level0 != sampler->cache_level) { /* cache miss, replace cache with new tile */ + struct pipe_surface *ps + = pipe->get_tex_surface(pipe, sampler->texture, 0, level0, 0); + sampler->cache_level = level0; sampler->cache_x = cx; sampler->cache_y = cy; ps->get_tile(ps, @@ -497,8 +525,7 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, cx = x % SAMPLER_CACHE_SIZE; cy = y % SAMPLER_CACHE_SIZE; - src = sampler->cache[cy][cx]; - COPY_4V(rgba, src); + COPY_4V(rgba, sampler->cache[cy][cx]); } break; case PIPE_TEX_FILTER_LINEAR: @@ -506,6 +533,9 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, GLfloat t00[4], t01[4], t10[4], t11[4]; GLint x0, y0, x1, y1; GLfloat a, b; + struct pipe_surface *ps + = pipe->get_tex_surface(pipe, sampler->texture, 0, level0, 0); + linear_texcoord(sampler->state->wrap_s, strq[0], sampler->texture->width0, &x0, &x1, &a); linear_texcoord(sampler->state->wrap_t, strq[1], @@ -521,6 +551,13 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, rgba[3] = lerp_2d(a, b, t00[3], t10[3], t01[3], t11[3]); } break; + /* + { + GLuint level0, level1; + level0 = choose_mipmap_level(sampler, lambda); + } + break; + */ default: assert(0); } diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index fb11eec777..01dd7f11b8 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -31,7 +31,7 @@ struct tgsi_sampler const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]); void *pipe; /*XXX temporary*/ - GLint cache_x, cache_y; + GLint cache_x, cache_y, cache_level; GLfloat cache[SAMPLER_CACHE_SIZE][SAMPLER_CACHE_SIZE][4]; }; -- cgit v1.2.3 From 8140642f5335eba5a524a7782fa40e6d095b5ada Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 16:46:32 -0600 Subject: use surface offset value in get_tile() --- src/mesa/pipe/softpipe/sp_surface.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 174dcc0169..3260bcce75 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -321,10 +321,19 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) { const GLuint *src - = ((const GLuint *) ps->region->map) + y * ps->region->pitch + x; + = ((const GLuint *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; GLuint i, j; +#if 0 assert(x + w <= ps->width); assert(y + h <= ps->height); +#else + /* temp hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { p[0] = UBYTE_TO_FLOAT((src[j] >> 16) & 0xff); @@ -343,7 +352,8 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) { const GLushort *src - = ((const GLushort *) ps->region->map) + y * ps->region->pitch + x; + = ((const GLushort *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; GLuint i, j; for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { -- cgit v1.2.3 From 5671bba1cb103c00b44905dfa015c3338cb4af56 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 16:46:59 -0600 Subject: clean-up --- src/mesa/pipe/softpipe/sp_tex_sample.c | 19 +++++++++---------- 1 file changed, 9 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 062d61dea4..dcc9485cf6 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -495,20 +495,19 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, case PIPE_TEX_FILTER_NEAREST: case PIPE_TEX_FILTER_NEAREST_MIPMAP_NEAREST: { - GLint x, y, cx, cy; - - x = nearest_texcoord(sampler->state->wrap_s, strq[0], - sampler->texture->width0); - y = nearest_texcoord(sampler->state->wrap_t, strq[1], - sampler->texture->height0); - - cx = x / SAMPLER_CACHE_SIZE; - cy = y / SAMPLER_CACHE_SIZE; + GLint x = nearest_texcoord(sampler->state->wrap_s, strq[0], + sampler->texture->level[level0].width); + GLint y = nearest_texcoord(sampler->state->wrap_t, strq[1], + sampler->texture->level[level0].height); + GLint cx = x / SAMPLER_CACHE_SIZE; + GLint cy = y / SAMPLER_CACHE_SIZE; if (cx != sampler->cache_x || cy != sampler->cache_y || level0 != sampler->cache_level) { /* cache miss, replace cache with new tile */ struct pipe_surface *ps = pipe->get_tex_surface(pipe, sampler->texture, 0, level0, 0); + assert(ps->width == sampler->texture->level[level0].width); + assert(ps->height == sampler->texture->level[level0].height); sampler->cache_level = level0; sampler->cache_x = cx; sampler->cache_y = cy; @@ -522,7 +521,7 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, else { /*printf("cache hit (%d, %d)\n", x, y);*/ } - + /* get texel from cache */ cx = x % SAMPLER_CACHE_SIZE; cy = y % SAMPLER_CACHE_SIZE; COPY_4V(rgba, sampler->cache[cy][cx]); -- cgit v1.2.3 From 562fdca12de5bbba491fafc7a26d86b23b5f7c93 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 8 Aug 2007 16:47:51 -0600 Subject: update lod bias, limits --- src/mesa/state_tracker/st_atom_sampler.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index a49698cda4..066ce7860f 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -111,6 +111,10 @@ update_samplers(struct st_context *st) sampler.min_filter = gl_filter_to_sp(texobj->MinFilter); sampler.mag_filter = gl_filter_to_sp(texobj->MagFilter); + sampler.lod_bias = st->ctx->Texture.Unit[u].LodBias; + sampler.min_lod = texobj->MinLod; + sampler.max_lod = texobj->MaxLod; + /* XXX more sampler state here */ } -- cgit v1.2.3 From 64f4268e85db54bae213d758b7ede676550d8185 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 9 Aug 2007 10:43:56 +0100 Subject: Declare missing opcode aliases. Declare SGN opcode. --- src/mesa/pipe/tgsi/core/tgsi_token.h | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index 1950dfc0ad..8848e99808 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -834,7 +834,9 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_DP3 */ /* TGSI_OPCODE_DP4 */ /* TGSI_OPCODE_LRP */ +#define TGSI_OPCODE_TEXCRD TGSI_OPCODE_TEXCOORD /* TGSI_OPCODE_TEXKILL */ +#define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX /* TGSI_OPCODE_CND */ #define TGSI_OPCODE_TEXDEPTH 122 /* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ @@ -870,6 +872,9 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_NRM4 128 #define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS /* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_TEXLD */ +#define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB +#define TGSI_OPCODE_TEXLDP TGSI_OPCODE_TEX /* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ #define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A @@ -911,14 +916,17 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_IFC 130 /* TGSI_OPCODE_ELSE */ /* TGSI_OPCODE_ENDIF */ -/* TGSI_OPCODE_BREAK */ +#define TGSI_OPCODE_BREAK TGSI_OPCODE_BRK #define TGSI_OPCODE_BREAKC 131 /* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_TEXLD */ +/* TGSI_OPCODE_TEXLDB */ /* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ /* TGSI_OPCODE_DP2ADD */ #define TGSI_OPCODE_DSX TGSI_OPCODE_DDX #define TGSI_OPCODE_DSY TGSI_OPCODE_DDY #define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD +/* TGSI_OPCODE_TEXLDP */ /* * vs_1_1 @@ -985,7 +993,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_ENDLOOP */ /* TGSI_OPCODE_POW */ /* XXX: takes ABS */ /* TGSI_OPCODE_CRS */ -/* TGSI_OPCODE_SGN */ +#define TGSI_OPCODE_SGN 133 /* TGSI_OPCODE_ABS */ /* TGSI_OPCODE_NRM4 */ /* TGSI_OPCODE_SINCOS */ @@ -1047,7 +1055,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MOVA */ /* TGSI_OPCODE_LOGP */ -#define TGSI_OPCODE_LAST 133 +#define TGSI_OPCODE_LAST 134 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ -- cgit v1.2.3 From d0521ebd516617f4471279e69af5dc0f4e4992af Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 16:40:45 +0100 Subject: Split swapbuffers and read/draw buffers functionality into two files. --- src/mesa/drivers/dri/intel_winsys/Makefile | 1 + src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 199 ----------------- .../dri/intel_winsys/intel_read_draw_buffer.c | 243 +++++++++++++++++++++ 3 files changed, 244 insertions(+), 199 deletions(-) create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_read_draw_buffer.c (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index 3d3d525c30..ff635eb44d 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -16,6 +16,7 @@ DRIVER_SOURCES = \ intel_batchbuffer.c \ intel_tex_layout.c \ intel_buffers.c \ + intel_read_draw_buffer.c \ intel_blit.c \ intel_context.c \ intel_ioctl.c \ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index fc9d60c88e..3cece98079 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -121,27 +121,6 @@ intel_readbuf_region(struct intel_context *intel) } - -/** - * Update the following fields for rendering: - * intel->numClipRects - * intel->pClipRects - */ -static void -intelSetRenderbufferClipRects(struct intel_context *intel) -{ - /* zero-sized buffers might be legal? */ - assert(intel->ctx.DrawBuffer->Width > 0); - assert(intel->ctx.DrawBuffer->Height > 0); - intel->fboRect.x1 = 0; - intel->fboRect.y1 = 0; - intel->fboRect.x2 = intel->ctx.DrawBuffer->Width; - intel->fboRect.y2 = intel->ctx.DrawBuffer->Height; - intel->numClipRects = 1; - intel->pClipRects = &intel->fboRect; -} - - /** * This will be called whenever the currently bound window is moved/resized. * XXX: actually, it seems to NOT be called when the window is only moved (BP). @@ -531,181 +510,3 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) } } - -/** - * Update the hardware state for drawing into a window or framebuffer object. - * - * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other - * places within the driver. - * - * Basically, this needs to be called any time the current framebuffer - * changes, the renderbuffers change, or we need to draw into different - * color buffers. - */ -void -intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) -{ - struct intel_context *intel = intel_context(ctx); - struct pipe_region *colorRegion, *depthRegion = NULL; - struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; - - if (!fb) { - /* this can happen during the initial context initialization */ - return; - } - - /* Do this here, not core Mesa, since this function is called from - * many places within the driver. - */ - if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { - /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */ - _mesa_update_framebuffer(ctx); - /* this updates the DrawBuffer's Width/Height if it's a FBO */ - _mesa_update_draw_buffer_bounds(ctx); - } - - if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { - /* this may occur when we're called by glBindFrameBuffer() during - * the process of someone setting up renderbuffers, etc. - */ - /*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/ - return; - } - - if (fb->Name) - intel_validate_paired_depth_stencil(ctx, fb); - - /* - * How many color buffers are we drawing into? - */ - if (fb->_NumColorDrawBuffers[0] != 1) { - /* writing to 0 or 2 or 4 color buffers */ - /*_mesa_debug(ctx, "Software rendering\n");*/ - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); - } - else { - /* draw to exactly one color buffer */ - /*_mesa_debug(ctx, "Hardware rendering\n");*/ - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); - } - - /* - * Get the intel_renderbuffer for the colorbuffer we're drawing into. - * And set up cliprects. - */ - { - struct intel_renderbuffer *irb; - intelSetRenderbufferClipRects(intel); - irb = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); - colorRegion = (irb && irb->region) ? irb->region : NULL; - } - - /* Update culling direction which changes depending on the - * orientation of the buffer: - */ - if (ctx->Driver.FrontFace) - ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace); - else - ctx->NewState |= _NEW_POLYGON; - - if (!colorRegion) { - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); - } - else { - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); - } - - /*** - *** Get depth buffer region and check if we need a software fallback. - *** Note that the depth buffer is usually a DEPTH_STENCIL buffer. - ***/ - if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) { - irbDepth = intel_renderbuffer(fb->_DepthBuffer->Wrapped); - if (irbDepth && irbDepth->region) { - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); - depthRegion = irbDepth->region; - } - else { - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_TRUE); - depthRegion = NULL; - } - } - else { - /* not using depth buffer */ - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); - depthRegion = NULL; - } - - /*** - *** Stencil buffer - *** This can only be hardware accelerated if we're using a - *** combined DEPTH_STENCIL buffer (for now anyway). - ***/ - if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { - irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); - if (irbStencil && irbStencil->region) { - ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); - /* need to re-compute stencil hw state */ -// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); - if (!depthRegion) - depthRegion = irbStencil->region; - } - else { - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE); - } - } - else { - /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */ - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); - /* need to re-compute stencil hw state */ -// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); - } - - - /** - ** Release old regions, reference new regions - **/ - - // intel->vtbl.set_draw_region(intel, colorRegion, depthRegion); - - /* update viewport since it depends on window size */ -// ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y, -// ctx->Viewport.Width, ctx->Viewport.Height); - - /* Update hardware scissor */ -// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, -// ctx->Scissor.Width, ctx->Scissor.Height); -} - - -static void -intelDrawBuffer(GLcontext * ctx, GLenum mode) -{ - intel_draw_buffer(ctx, ctx->DrawBuffer); -} - - -static void -intelReadBuffer(GLcontext * ctx, GLenum mode) -{ - if (ctx->ReadBuffer == ctx->DrawBuffer) { - /* This will update FBO completeness status. - * A framebuffer will be incomplete if the GL_READ_BUFFER setting - * refers to a missing renderbuffer. Calling glReadBuffer can set - * that straight and can make the drawing buffer complete. - */ - intel_draw_buffer(ctx, ctx->DrawBuffer); - } - /* Generally, functions which read pixels (glReadPixels, glCopyPixels, etc) - * reference ctx->ReadBuffer and do appropriate state checks. - */ -} - - -void -intelInitBufferFuncs(struct dd_function_table *functions) -{ - functions->DrawBuffer = intelDrawBuffer; - functions->ReadBuffer = intelReadBuffer; -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_read_draw_buffer.c b/src/mesa/drivers/dri/intel_winsys/intel_read_draw_buffer.c new file mode 100644 index 0000000000..0ac246426e --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_read_draw_buffer.c @@ -0,0 +1,243 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_blit.h" +#include "intel_buffers.h" +#include "intel_depthstencil.h" +#include "intel_fbo.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" +#include "context.h" +#include "utils.h" +#include "drirenderbuffer.h" +#include "framebuffer.h" +#include "vblank.h" + +#include "pipe/p_context.h" + + + +/** + * Update the following fields for rendering: + * intel->numClipRects + * intel->pClipRects + */ +static void +intelSetRenderbufferClipRects(struct intel_context *intel) +{ + /* zero-sized buffers might be legal? */ + assert(intel->ctx.DrawBuffer->Width > 0); + assert(intel->ctx.DrawBuffer->Height > 0); + intel->fboRect.x1 = 0; + intel->fboRect.y1 = 0; + intel->fboRect.x2 = intel->ctx.DrawBuffer->Width; + intel->fboRect.y2 = intel->ctx.DrawBuffer->Height; + intel->numClipRects = 1; + intel->pClipRects = &intel->fboRect; +} + + + +/** + * Update the hardware state for drawing into a window or framebuffer object. + * + * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other + * places within the driver. + * + * Basically, this needs to be called any time the current framebuffer + * changes, the renderbuffers change, or we need to draw into different + * color buffers. + */ +void +intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) +{ + struct intel_context *intel = intel_context(ctx); + struct pipe_region *colorRegion, *depthRegion = NULL; + struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; + + if (!fb) { + /* this can happen during the initial context initialization */ + return; + } + + /* Do this here, not core Mesa, since this function is called from + * many places within the driver. + */ + if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { + /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */ + _mesa_update_framebuffer(ctx); + /* this updates the DrawBuffer's Width/Height if it's a FBO */ + _mesa_update_draw_buffer_bounds(ctx); + } + + if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { + /* this may occur when we're called by glBindFrameBuffer() during + * the process of someone setting up renderbuffers, etc. + */ + /*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/ + return; + } + + if (fb->Name) + intel_validate_paired_depth_stencil(ctx, fb); + + /* + * How many color buffers are we drawing into? + */ + if (fb->_NumColorDrawBuffers[0] != 1) { + /* writing to 0 or 2 or 4 color buffers */ + /*_mesa_debug(ctx, "Software rendering\n");*/ + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); + } + else { + /* draw to exactly one color buffer */ + /*_mesa_debug(ctx, "Hardware rendering\n");*/ + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); + } + + /* + * Get the intel_renderbuffer for the colorbuffer we're drawing into. + * And set up cliprects. + */ + { + struct intel_renderbuffer *irb; + intelSetRenderbufferClipRects(intel); + irb = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); + colorRegion = (irb && irb->region) ? irb->region : NULL; + } + + /* Update culling direction which changes depending on the + * orientation of the buffer: + */ + if (ctx->Driver.FrontFace) + ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace); + else + ctx->NewState |= _NEW_POLYGON; + + if (!colorRegion) { + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); + } + else { + FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); + } + + /*** + *** Get depth buffer region and check if we need a software fallback. + *** Note that the depth buffer is usually a DEPTH_STENCIL buffer. + ***/ + if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) { + irbDepth = intel_renderbuffer(fb->_DepthBuffer->Wrapped); + if (irbDepth && irbDepth->region) { + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); + depthRegion = irbDepth->region; + } + else { + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_TRUE); + depthRegion = NULL; + } + } + else { + /* not using depth buffer */ + FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); + depthRegion = NULL; + } + + /*** + *** Stencil buffer + *** This can only be hardware accelerated if we're using a + *** combined DEPTH_STENCIL buffer (for now anyway). + ***/ + if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { + irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); + if (irbStencil && irbStencil->region) { + ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); + /* need to re-compute stencil hw state */ +// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); + if (!depthRegion) + depthRegion = irbStencil->region; + } + else { + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE); + } + } + else { + /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */ + FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); + /* need to re-compute stencil hw state */ +// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); + } + + + /** + ** Release old regions, reference new regions + **/ + + // intel->vtbl.set_draw_region(intel, colorRegion, depthRegion); + + /* update viewport since it depends on window size */ +// ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y, +// ctx->Viewport.Width, ctx->Viewport.Height); + + /* Update hardware scissor */ +// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, +// ctx->Scissor.Width, ctx->Scissor.Height); +} + + +static void +intelDrawBuffer(GLcontext * ctx, GLenum mode) +{ + intel_draw_buffer(ctx, ctx->DrawBuffer); +} + + +static void +intelReadBuffer(GLcontext * ctx, GLenum mode) +{ + if (ctx->ReadBuffer == ctx->DrawBuffer) { + /* This will update FBO completeness status. + * A framebuffer will be incomplete if the GL_READ_BUFFER setting + * refers to a missing renderbuffer. Calling glReadBuffer can set + * that straight and can make the drawing buffer complete. + */ + intel_draw_buffer(ctx, ctx->DrawBuffer); + } + /* Generally, functions which read pixels (glReadPixels, glCopyPixels, etc) + * reference ctx->ReadBuffer and do appropriate state checks. + */ +} + + +void +intelInitBufferFuncs(struct dd_function_table *functions) +{ + functions->DrawBuffer = intelDrawBuffer; + functions->ReadBuffer = intelReadBuffer; +} -- cgit v1.2.3 From 93d5cfd8b7c641c5656aed38ad2f9559e09502f0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 16:42:01 +0100 Subject: Add flush/finish functionality to pipe. Not sure if finish() is the right interface yet. --- src/mesa/pipe/i915simple/Makefile | 1 + src/mesa/pipe/i915simple/i915_batch.h | 9 ++++ src/mesa/pipe/i915simple/i915_context.c | 7 +++ src/mesa/pipe/i915simple/i915_context.h | 6 ++- src/mesa/pipe/i915simple/i915_flush.c | 82 +++++++++++++++++++++++++++++++++ src/mesa/pipe/p_context.h | 4 ++ src/mesa/pipe/p_defines.h | 7 +++ src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_buffer.c | 1 + src/mesa/pipe/softpipe/sp_context.c | 3 ++ src/mesa/pipe/softpipe/sp_flush.c | 57 +++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_flush.h | 36 +++++++++++++++ 12 files changed, 212 insertions(+), 2 deletions(-) create mode 100644 src/mesa/pipe/i915simple/i915_flush.c create mode 100644 src/mesa/pipe/softpipe/sp_flush.c create mode 100644 src/mesa/pipe/softpipe/sp_flush.h (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 4170349893..f8b14220fb 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -8,6 +8,7 @@ DRIVER_SOURCES = \ i915_blit.c \ i915_buffer.c \ i915_clear.c \ + i915_flush.c \ i915_context.c \ i915_context.c \ i915_debug.c \ diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h index 77008eee14..b4ca8c38da 100644 --- a/src/mesa/pipe/i915simple/i915_batch.h +++ b/src/mesa/pipe/i915simple/i915_batch.h @@ -28,6 +28,9 @@ #ifndef I915_BATCH_H #define I915_BATCH_H +#include "i915_winsys.h" +#include "i915_debug.h" + #define BATCH_LOCALS #define BEGIN_BATCH( dwords, relocs ) \ @@ -41,4 +44,10 @@ #define ADVANCE_BATCH() +#define FLUSH_BATCH() do { \ +/* i915_dump_batchbuffer( i915, i915->batch_start, BEGIN_BATCH(0, 0) ); */ \ + i915->winsys->batch_flush( i915->winsys ); \ + i915->batch_start = BEGIN_BATCH(0, 0); \ +} while (0) + #endif diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index f39c282981..02b51bea43 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -170,6 +170,13 @@ struct pipe_context *i915_create( struct i915_winsys *winsys ) i915_init_region_functions(i915); i915_init_surface_functions(i915); i915_init_state_functions(i915); + i915_init_flush_functions(i915); + + + /* Batch stream debugging is a bit hacked up at the moment: + */ + i915->batch_start = winsys->batch_start( winsys, 0, 0 ); + /* * XXX we could plug GL selection/feedback into the drawing pipeline diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 060997bfa4..dab5423190 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -53,12 +53,13 @@ struct i915_context struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_setup_state setup; struct pipe_stencil_state stencil; - struct pipe_texture_object *texture[PIPE_MAX_SAMPLERS]; + struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; GLuint dirty; - struct pipe_scissor_state cliprect; + GLuint *batch_start; + struct pipe_scissor_state cliprect; }; #define I915_NEW_VIEWPORT 0x1 @@ -104,6 +105,7 @@ void i915_init_buffer_functions( struct i915_context *i915 ); void i915_init_region_functions( struct i915_context *i915 ); void i915_init_surface_functions( struct i915_context *i915 ); void i915_init_state_functions( struct i915_context *i915 ); +void i915_init_flush_functions( struct i915_context *i915 ); diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c new file mode 100644 index 0000000000..1cf945e9a2 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_flush.c @@ -0,0 +1,82 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "i915_context.h" +#include "i915_reg.h" +#include "i915_batch.h" + +/* There will be actual work to do here. In future we may want a + * fence-like interface instead of finish, and perhaps flush will take + * flags to indicate what type of flush is required. + */ +static void i915_flush( struct pipe_context *pipe, + unsigned flags ) +{ + struct i915_context *i915 = i915_context(pipe); + + /* Do we need to emit an MI_FLUSH command to flush the hardware + * caches? + */ + if (flags) { + unsigned flush = MI_FLUSH; + + if (!(flags & PIPE_FLUSH_RENDER_CACHE)) + flush |= INHIBIT_FLUSH_RENDER_CACHE; + + if (flags & PIPE_FLUSH_TEXTURE_CACHE) + flush |= FLUSH_MAP_CACHE; + + BEGIN_BATCH( 1, 0 ); + OUT_BATCH( flush ); + ADVANCE_BATCH(); + } + + /* If there are no flags, just flush pending commands to hardware: + */ + FLUSH_BATCH(); +} + +static void i915_finish(struct pipe_context *pipe) +{ + struct i915_context *i915 = i915_context(pipe); + + i915_flush( pipe, 0 ); + i915->winsys->batch_wait_idle( i915->winsys ); +} + + +void i915_init_flush_functions( struct i915_context *i915 ) +{ + i915->pipe.flush = i915_flush; + i915->pipe.finish = i915_finish; +} diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 3062eda329..23961f4b8f 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -219,6 +219,10 @@ struct pipe_context { struct pipe_mipmap_tree *mt ); + void (*flush)( struct pipe_context *pipe, + unsigned flags ); + + void (*finish)( struct pipe_context *pipe ); }; diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 77512f8818..9eb4c37434 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -176,4 +176,11 @@ #define PIPE_MAP_WRITE 2 #define PIPE_MAP_READ_WRITE 3 + +/** + * Flush types: + */ +#define PIPE_FLUSH_RENDER_CACHE 0x1 +#define PIPE_FLUSH_TEXTURE_CACHE 0x2 + #endif diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index feb167405a..ac83da451d 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -7,6 +7,7 @@ LIBNAME = softpipe DRIVER_SOURCES = \ sp_buffer.c \ sp_clear.c \ + sp_flush.c \ sp_context.c \ sp_prim_setup.c \ sp_quad.c \ diff --git a/src/mesa/pipe/softpipe/sp_buffer.c b/src/mesa/pipe/softpipe/sp_buffer.c index f3a4718449..27443421a3 100644 --- a/src/mesa/pipe/softpipe/sp_buffer.c +++ b/src/mesa/pipe/softpipe/sp_buffer.c @@ -32,6 +32,7 @@ #include #include "sp_context.h" #include "sp_winsys.h" +#include "sp_buffer.h" diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index ac5651f9a7..ebc994670a 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -36,6 +36,7 @@ #include "sp_buffer.h" #include "sp_clear.h" #include "sp_context.h" +#include "sp_flush.h" #include "sp_prim_setup.h" #include "sp_region.h" #include "sp_state.h" @@ -214,6 +215,8 @@ struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.draw_vertices = softpipe_draw_vertices; softpipe->pipe.clear = softpipe_clear; + softpipe->pipe.flush = softpipe_flush; + softpipe->pipe.finish = softpipe_finish; softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c new file mode 100644 index 0000000000..aa609469a6 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -0,0 +1,57 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "sp_flush.h" +#include "sp_context.h" + +/* There will be actual work to do here. In future we may want a + * fence-like interface instead of finish, and perhaps flush will take + * flags to indicate what type of flush is required. + */ +void +softpipe_flush( struct pipe_context *pipe, + unsigned flags ) +{ + /* - flush the quad pipeline + * - flush the texture cache + * - flush the render cache + */ +} + +void +softpipe_finish(struct pipe_context *pipe) +{ + /* Just calls into flush() + */ + softpipe_flush( pipe, 0 ); +} diff --git a/src/mesa/pipe/softpipe/sp_flush.h b/src/mesa/pipe/softpipe/sp_flush.h new file mode 100644 index 0000000000..5e204f87a5 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_flush.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_FLUSH_H +#define SP_FLUSH_H + +struct pipe_context; + +void softpipe_finish(struct pipe_context *pipe); +void softpipe_flush(struct pipe_context *pipe, unsigned flags ); + +#endif -- cgit v1.2.3 From b0e57f789c2ae266fc51874c480d033309075468 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 8 Aug 2007 16:42:29 +0100 Subject: Hook up pipe flush functionality --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 1 + src/mesa/drivers/dri/intel_winsys/intel_context.c | 115 ++++++++++----------- 2 files changed, 56 insertions(+), 60 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index a85dc0b613..80a3163d55 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -324,6 +324,7 @@ intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, r->buf = buffer; r->offset = batch->ptr - batch->map; r->delta = delta; + *(GLuint *) batch->ptr = 0x12345678; } batch->ptr += 4; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 66be3af72b..c1d541199f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -250,13 +250,7 @@ intelFlush(GLcontext * ctx) { struct intel_context *intel = intel_context(ctx); - INTEL_FIREVERTICES(intel); - - if (intel->batch->map != intel->batch->ptr) - intel_batchbuffer_flush(intel->batch); - - /* XXX: Need to do an MI_FLUSH here. - */ + intel->pipe->flush( intel->pipe, 0 ); } @@ -317,9 +311,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) functions->GetString = intelGetString; functions->UpdateState = intelInvalidateState; - /* - intelInitTextureFuncs(functions); - */ intelInitBufferFuncs(functions); st_init_driver_functions(functions); @@ -369,56 +360,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, _vbo_CreateContext(ctx); _tnl_CreateContext(ctx); - /* - * Pipe-related setup - */ - if (!getenv("INTEL_HW")) { - intel->pipe = intel_create_softpipe( intel ); - } - else { - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - intel->pipe = intel_create_i915simple( intel ); - break; - default: - _mesa_printf("Unknown PCIID %x in %s, using software driver\n", - intel->intelScreen->deviceID, __FUNCTION__); - - intel->pipe = intel_create_softpipe( intel ); - break; - } - } - - st_create_context( &intel->ctx, intel->pipe ); - - - /* TODO: Push this down into the pipe driver: - */ - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - intel->pipe->mipmap_tree_layout = i945_miptree_layout; - break; - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - case PCI_CHIP_I830_M: - case PCI_CHIP_I855_GM: - case PCI_CHIP_I865_G: - intel->pipe->mipmap_tree_layout = i915_miptree_layout; - default: - assert(0); /*FIX*/ - } /* @@ -471,6 +412,60 @@ intelCreateContext(const __GLcontextModes * mesaVis, #endif + + + /* + * Pipe-related setup + */ + if (!getenv("INTEL_HW")) { + intel->pipe = intel_create_softpipe( intel ); + } + else { + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + intel->pipe = intel_create_i915simple( intel ); + break; + default: + _mesa_printf("Unknown PCIID %x in %s, using software driver\n", + intel->intelScreen->deviceID, __FUNCTION__); + + intel->pipe = intel_create_softpipe( intel ); + break; + } + } + + st_create_context( &intel->ctx, intel->pipe ); + + + /* TODO: Push this down into the pipe driver: + */ + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + intel->pipe->mipmap_tree_layout = i945_miptree_layout; + break; + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + case PCI_CHIP_I830_M: + case PCI_CHIP_I855_GM: + case PCI_CHIP_I865_G: + intel->pipe->mipmap_tree_layout = i915_miptree_layout; + default: + assert(0); /*FIX*/ + } + + return GL_TRUE; } -- cgit v1.2.3 From dc2c20b6250915548c7fbfd369fdc890bae2b98c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 10:34:35 +0100 Subject: Put back the old versions of intel_tex_layout.* in this shared directory. Unbreak the original i915, i915tex drivers. We will have our own version of this functionality in the pipe driver and not attempt to share with the older code. --- src/mesa/drivers/dri/intel/intel_tex_layout.c | 384 +------------------------- src/mesa/drivers/dri/intel/intel_tex_layout.h | 49 +++- 2 files changed, 42 insertions(+), 391 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c index 882d9e04a2..fcb5cc3906 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.c +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c @@ -30,78 +30,17 @@ * Michel Dänzer */ -#include "macros.h" -#include "pipe/p_state.h" -#include "pipe/p_context.h" +#include "intel_mipmap_tree.h" #include "intel_tex_layout.h" -#include "state_tracker/st_mipmap_tree.h" - +#include "macros.h" -static GLuint minify( GLuint d ) -{ - return MAX2(1, d>>1); -} static int align(int value, int alignment) { return (value + alignment - 1) & ~(alignment - 1); } - -static void -intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, - GLuint level, - GLuint nr_images, - GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) -{ - assert(level < MAX_TEXTURE_LEVELS); - - mt->level[level].width = w; - mt->level[level].height = h; - mt->level[level].depth = d; - mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; - mt->level[level].nr_images = nr_images; - - /* - DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - level, w, h, d, x, y, mt->level[level].level_offset); - */ - - /* Not sure when this would happen, but anyway: - */ - if (mt->level[level].image_offset) { - free(mt->level[level].image_offset); - mt->level[level].image_offset = NULL; - } - - assert(nr_images); - assert(!mt->level[level].image_offset); - - mt->level[level].image_offset = (GLuint *) malloc(nr_images * sizeof(GLuint)); - mt->level[level].image_offset[0] = 0; -} - - -static void -intel_miptree_set_image_offset(struct pipe_mipmap_tree *mt, - GLuint level, GLuint img, GLuint x, GLuint y) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < mt->level[level].nr_images); - - mt->level[level].image_offset[img] = (x + y * mt->pitch); - - /* - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); - */ -} - - -static void -i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) +void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ) { GLint align_h = 2, align_w = 4; GLuint level; @@ -134,7 +73,8 @@ i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { GLuint img_height; - intel_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); + intel_miptree_set_level_info(mt, level, 1, x, y, width, + height, 1); if (mt->compressed) img_height = MAX2(1, height/4); @@ -160,317 +100,3 @@ i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) height = minify(height); } } - - -static const GLint initial_offsets[6][2] = { - {0, 0}, - {0, 2}, - {1, 0}, - {1, 2}, - {1, 1}, - {1, 3} -}; - -static const GLint step_offsets[6][2] = { - {0, 2}, - {0, 2}, - {-1, 2}, - {-1, 2}, - {-1, 1}, - {-1, 1} -}; - - -GLboolean -i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) -{ - GLint level; - - switch (mt->target) { - case GL_TEXTURE_CUBE_MAP:{ - const GLuint dim = mt->width0; - GLuint face; - GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* double pitch for cube layouts */ - mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; - mt->total_height = dim * 4; - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_level_info(mt, level, 6, - 0, 0, - /*OLD: mt->pitch, mt->total_height,*/ - lvlWidth, lvlHeight, - 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - for (face = 0; face < 6; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_image_offset(mt, level, face, x, y); - - if (d == 0) - _mesa_printf("cube mipmap %d/%d (%d..%d) is 0x0\n", - face, level, mt->first_level, mt->last_level); - - d >>= 1; - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - } - } - break; - } - case GL_TEXTURE_3D:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint stack_height = 0; - - /* Calculate the size of a single slice. - */ - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - - /* XXX: hardware expects/requires 9 levels at minimum. - */ - for (level = mt->first_level; level <= MAX2(8, mt->last_level); - level++) { - intel_miptree_set_level_info(mt, level, depth, 0, mt->total_height, - width, height, depth); - - - stack_height += MAX2(2, height); - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - - /* Fixup depth image_offsets: - */ - depth = mt->depth0; - for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint i; - for (i = 0; i < depth; i++) - intel_miptree_set_image_offset(mt, level, i, - 0, i * stack_height); - - depth = minify(depth); - } - - - /* Multiply slice size by texture depth for total size. It's - * remarkable how wasteful of memory the i915 texture layouts - * are. They are largely fixed in the i945. - */ - mt->total_height = stack_height * mt->depth0; - break; - } - - default:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint img_height; - - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - mt->total_height = 0; - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_level_info(mt, level, 1, - 0, mt->total_height, - width, height, 1); - - if (mt->compressed) - img_height = MAX2(1, height / 4); - else - img_height = (MAX2(2, height) + 1) & ~1; - - mt->total_height += img_height; - - width = minify(width); - height = minify(height); - } - break; - } - } - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); - */ - - return GL_TRUE; -} - - -GLboolean -i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) -{ - GLint level; - - switch (mt->target) { - case GL_TEXTURE_CUBE_MAP:{ - const GLuint dim = mt->width0; - GLuint face; - GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* Depending on the size of the largest images, pitch can be - * determined either by the old-style packing of cubemap faces, - * or the final row of 4x4, 2x2 and 1x1 faces below this. - */ - if (dim > 32) - mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; - else - mt->pitch = 14 * 8; - - mt->total_height = dim * 4 + 4; - - /* Set all the levels to effectively occupy the whole rectangular region. - */ - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_level_info(mt, level, 6, - 0, 0, - lvlWidth, lvlHeight, 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - - for (face = 0; face < 6; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; - - if (dim == 4 && face >= 4) { - y = mt->total_height - 4; - x = (face - 4) * 8; - } - else if (dim < 4 && (face > 0 || mt->first_level > 0)) { - y = mt->total_height - 4; - x = face * 8; - } - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_image_offset(mt, level, face, x, y); - - d >>= 1; - - switch (d) { - case 4: - switch (face) { - case FACE_POS_X: - case FACE_NEG_X: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - case FACE_POS_Y: - case FACE_NEG_Y: - y += 12; - x -= 8; - break; - case FACE_POS_Z: - case FACE_NEG_Z: - y = mt->total_height - 4; - x = (face - 4) * 8; - break; - } - - case 2: - y = mt->total_height - 4; - x = 16 + face * 8; - break; - - case 1: - x += 48; - break; - - default: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - } - } - } - break; - } - case GL_TEXTURE_3D:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint pack_x_pitch, pack_x_nr; - GLuint pack_y_pitch; - GLuint level; - - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - mt->total_height = 0; - - pack_y_pitch = MAX2(mt->height0, 2); - pack_x_pitch = mt->pitch; - pack_x_nr = 1; - - for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; - GLint x = 0; - GLint y = 0; - GLint q, j; - - intel_miptree_set_level_info(mt, level, nr_images, - 0, mt->total_height, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - intel_miptree_set_image_offset(mt, level, q, x, y); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - mt->total_height += y; - - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= mt->pitch); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - } - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - break; - } - - case GL_TEXTURE_1D: - case GL_TEXTURE_2D: - case GL_TEXTURE_RECTANGLE_ARB: - i945_miptree_layout_2d(mt); - break; - default: - _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()"); - } - - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); - */ - - return GL_TRUE; -} - diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h index 83913b89cb..1e37f8f525 100644 --- a/src/mesa/drivers/dri/intel/intel_tex_layout.h +++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h @@ -1,16 +1,41 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ -#ifndef INTEL_TEX_LAYOUT_H -#define INTEL_TEX_LAYOUT_H +#include "macros.h" -struct pipe_context; -struct pipe_mipmap_tree; +static GLuint minify( GLuint d ) +{ + return MAX2(1, d>>1); +} -extern GLboolean -i915_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); - -extern GLboolean -i945_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); - - -#endif /* INTEL_TEX_LAYOUT_H */ +extern void i945_miptree_layout_2d( struct intel_mipmap_tree *mt ); -- cgit v1.2.3 From 9526f7445e9acea8553769bf8f86e3c95655cca1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 10:35:48 +0100 Subject: The pipe version of these files. Can probably relocate to pipe/i915simple. --- .../drivers/dri/intel_winsys/intel_tex_layout.c | 477 ++++++++++++++++++++- .../drivers/dri/intel_winsys/intel_tex_layout.h | 16 + 2 files changed, 492 insertions(+), 1 deletion(-) mode change 120000 => 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_layout.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c deleted file mode 120000 index fe61b44194..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c +++ /dev/null @@ -1 +0,0 @@ -../intel/intel_tex_layout.c \ No newline at end of file diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c new file mode 100644 index 0000000000..882d9e04a2 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c @@ -0,0 +1,476 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "macros.h" +#include "pipe/p_state.h" +#include "pipe/p_context.h" +#include "intel_tex_layout.h" +#include "state_tracker/st_mipmap_tree.h" + + +static GLuint minify( GLuint d ) +{ + return MAX2(1, d>>1); +} + +static int align(int value, int alignment) +{ + return (value + alignment - 1) & ~(alignment - 1); +} + + +static void +intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) +{ + assert(level < MAX_TEXTURE_LEVELS); + + mt->level[level].width = w; + mt->level[level].height = h; + mt->level[level].depth = d; + mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; + mt->level[level].nr_images = nr_images; + + /* + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, mt->level[level].level_offset); + */ + + /* Not sure when this would happen, but anyway: + */ + if (mt->level[level].image_offset) { + free(mt->level[level].image_offset); + mt->level[level].image_offset = NULL; + } + + assert(nr_images); + assert(!mt->level[level].image_offset); + + mt->level[level].image_offset = (GLuint *) malloc(nr_images * sizeof(GLuint)); + mt->level[level].image_offset[0] = 0; +} + + +static void +intel_miptree_set_image_offset(struct pipe_mipmap_tree *mt, + GLuint level, GLuint img, GLuint x, GLuint y) +{ + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + + assert(img < mt->level[level].nr_images); + + mt->level[level].image_offset[img] = (x + y * mt->pitch); + + /* + DBG("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); + */ +} + + +static void +i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) +{ + GLint align_h = 2, align_w = 4; + GLuint level; + GLuint x = 0; + GLuint y = 0; + GLuint width = mt->width0; + GLuint height = mt->height0; + + mt->pitch = mt->width0; + + /* May need to adjust pitch to accomodate the placement of + * the 2nd mipmap. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap out past the width of its parent. + */ + if (mt->first_level != mt->last_level) { + GLuint mip1_width = align(minify(mt->width0), align_w) + + minify(minify(mt->width0)); + + if (mip1_width > mt->width0) + mt->pitch = mip1_width; + } + + /* Pitch must be a whole number of dwords, even though we + * express it in texels. + */ + mt->pitch = align(mt->pitch * mt->cpp, 4) / mt->cpp; + mt->total_height = 0; + + for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { + GLuint img_height; + + intel_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); + + if (mt->compressed) + img_height = MAX2(1, height/4); + else + img_height = align(height, align_h); + + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + mt->total_height = MAX2(mt->total_height, y + img_height); + + /* Layout_below: step right after second mipmap. + */ + if (level == mt->first_level + 1) { + x += align(width, align_w); + } + else { + y += img_height; + } + + width = minify(width); + height = minify(height); + } +} + + +static const GLint initial_offsets[6][2] = { + {0, 0}, + {0, 2}, + {1, 0}, + {1, 2}, + {1, 1}, + {1, 3} +}; + +static const GLint step_offsets[6][2] = { + {0, 2}, + {0, 2}, + {-1, 2}, + {-1, 2}, + {-1, 1}, + {-1, 1} +}; + + +GLboolean +i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) +{ + GLint level; + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP:{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* double pitch for cube layouts */ + mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; + mt->total_height = dim * 4; + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_level_info(mt, level, 6, + 0, 0, + /*OLD: mt->pitch, mt->total_height,*/ + lvlWidth, lvlHeight, + 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + for (face = 0; face < 6; face++) { + GLuint x = initial_offsets[face][0] * dim; + GLuint y = initial_offsets[face][1] * dim; + GLuint d = dim; + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_image_offset(mt, level, face, x, y); + + if (d == 0) + _mesa_printf("cube mipmap %d/%d (%d..%d) is 0x0\n", + face, level, mt->first_level, mt->last_level); + + d >>= 1; + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + } + } + break; + } + case GL_TEXTURE_3D:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint stack_height = 0; + + /* Calculate the size of a single slice. + */ + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + + /* XXX: hardware expects/requires 9 levels at minimum. + */ + for (level = mt->first_level; level <= MAX2(8, mt->last_level); + level++) { + intel_miptree_set_level_info(mt, level, depth, 0, mt->total_height, + width, height, depth); + + + stack_height += MAX2(2, height); + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + + /* Fixup depth image_offsets: + */ + depth = mt->depth0; + for (level = mt->first_level; level <= mt->last_level; level++) { + GLuint i; + for (i = 0; i < depth; i++) + intel_miptree_set_image_offset(mt, level, i, + 0, i * stack_height); + + depth = minify(depth); + } + + + /* Multiply slice size by texture depth for total size. It's + * remarkable how wasteful of memory the i915 texture layouts + * are. They are largely fixed in the i945. + */ + mt->total_height = stack_height * mt->depth0; + break; + } + + default:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint img_height; + + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->total_height = 0; + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_level_info(mt, level, 1, + 0, mt->total_height, + width, height, 1); + + if (mt->compressed) + img_height = MAX2(1, height / 4); + else + img_height = (MAX2(2, height) + 1) & ~1; + + mt->total_height += img_height; + + width = minify(width); + height = minify(height); + } + break; + } + } + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + mt->pitch, + mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); + */ + + return GL_TRUE; +} + + +GLboolean +i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) +{ + GLint level; + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP:{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* Depending on the size of the largest images, pitch can be + * determined either by the old-style packing of cubemap faces, + * or the final row of 4x4, 2x2 and 1x1 faces below this. + */ + if (dim > 32) + mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; + else + mt->pitch = 14 * 8; + + mt->total_height = dim * 4 + 4; + + /* Set all the levels to effectively occupy the whole rectangular region. + */ + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_level_info(mt, level, 6, + 0, 0, + lvlWidth, lvlHeight, 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + + for (face = 0; face < 6; face++) { + GLuint x = initial_offsets[face][0] * dim; + GLuint y = initial_offsets[face][1] * dim; + GLuint d = dim; + + if (dim == 4 && face >= 4) { + y = mt->total_height - 4; + x = (face - 4) * 8; + } + else if (dim < 4 && (face > 0 || mt->first_level > 0)) { + y = mt->total_height - 4; + x = face * 8; + } + + for (level = mt->first_level; level <= mt->last_level; level++) { + intel_miptree_set_image_offset(mt, level, face, x, y); + + d >>= 1; + + switch (d) { + case 4: + switch (face) { + case FACE_POS_X: + case FACE_NEG_X: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + case FACE_POS_Y: + case FACE_NEG_Y: + y += 12; + x -= 8; + break; + case FACE_POS_Z: + case FACE_NEG_Z: + y = mt->total_height - 4; + x = (face - 4) * 8; + break; + } + + case 2: + y = mt->total_height - 4; + x = 16 + face * 8; + break; + + case 1: + x += 48; + break; + + default: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + } + } + } + break; + } + case GL_TEXTURE_3D:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint pack_x_pitch, pack_x_nr; + GLuint pack_y_pitch; + GLuint level; + + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->total_height = 0; + + pack_y_pitch = MAX2(mt->height0, 2); + pack_x_pitch = mt->pitch; + pack_x_nr = 1; + + for (level = mt->first_level; level <= mt->last_level; level++) { + GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; + GLint x = 0; + GLint y = 0; + GLint q, j; + + intel_miptree_set_level_info(mt, level, nr_images, + 0, mt->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + intel_miptree_set_image_offset(mt, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + mt->total_height += y; + + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= mt->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + } + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + break; + } + + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_RECTANGLE_ARB: + i945_miptree_layout_2d(mt); + break; + default: + _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()"); + } + + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + mt->pitch, + mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); + */ + + return GL_TRUE; +} + diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.h b/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.h new file mode 100644 index 0000000000..83913b89cb --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.h @@ -0,0 +1,16 @@ + +#ifndef INTEL_TEX_LAYOUT_H +#define INTEL_TEX_LAYOUT_H + +struct pipe_context; +struct pipe_mipmap_tree; + + +extern GLboolean +i915_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); + +extern GLboolean +i945_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); + + +#endif /* INTEL_TEX_LAYOUT_H */ -- cgit v1.2.3 From c046174eb047c95e25f31390019f04f310ae2b8e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 11:18:27 +0100 Subject: Pull in debug code for human-readable batchbuffer dumps. --- src/mesa/pipe/i915simple/i915_debug.c | 610 ++++++++++++++++++------------- src/mesa/pipe/i915simple/i915_debug.h | 17 + src/mesa/pipe/i915simple/i915_debug_fp.c | 331 +++++++++++++++++ 3 files changed, 704 insertions(+), 254 deletions(-) create mode 100644 src/mesa/pipe/i915simple/i915_debug_fp.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index 8d80590396..0951319d38 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -25,313 +25,415 @@ * **************************************************************************/ -#include -#include +#include "imports.h" #include "i915_reg.h" +#include "i915_context.h" #include "i915_debug.h" -static const char *opcodes[0x20] = { - "NOP", - "ADD", - "MOV", - "MUL", - "MAD", - "DP2ADD", - "DP3", - "DP4", - "FRC", - "RCP", - "RSQ", - "EXP", - "LOG", - "CMP", - "MIN", - "MAX", - "FLR", - "MOD", - "TRC", - "SGE", - "SLT", - "TEXLD", - "TEXLDP", - "TEXLDB", - "TEXKILL", - "DCL", - "0x1a", - "0x1b", - "0x1c", - "0x1d", - "0x1e", - "0x1f", -}; - - -static const int args[0x20] = { - 0, /* 0 nop */ - 2, /* 1 add */ - 1, /* 2 mov */ - 2, /* 3 m ul */ - 3, /* 4 mad */ - 3, /* 5 dp2add */ - 2, /* 6 dp3 */ - 2, /* 7 dp4 */ - 1, /* 8 frc */ - 1, /* 9 rcp */ - 1, /* a rsq */ - 1, /* b exp */ - 1, /* c log */ - 3, /* d cmp */ - 2, /* e min */ - 2, /* f max */ - 1, /* 10 flr */ - 1, /* 11 mod */ - 1, /* 12 trc */ - 2, /* 13 sge */ - 2, /* 14 slt */ - 1, - 1, - 1, - 1, - 0, - 0, - 0, - 0, - 0, - 0, - 0, -}; - - -static const char *regname[0x8] = { - "R", - "T", - "CONST", - "S", - "OC", - "OD", - "U", - "UNKNOWN", -}; - -static void -print_reg_type_nr(unsigned type, unsigned nr) +static GLboolean debug( struct debug_stream *stream, const char *name, GLuint len ) { - switch (type) { - case REG_TYPE_T: - switch (nr) { - case T_DIFFUSE: - fprintf(stderr, "T_DIFFUSE"); - return; - case T_SPECULAR: - fprintf(stderr, "T_SPECULAR"); - return; - case T_FOG_W: - fprintf(stderr, "T_FOG_W"); - return; - default: - fprintf(stderr, "T_TEX%d", nr); - return; - } - case REG_TYPE_OC: - if (nr == 0) { - fprintf(stderr, "oC"); - return; - } - break; - case REG_TYPE_OD: - if (nr == 0) { - fprintf(stderr, "oD"); - return; - } - break; - default: - break; + GLuint i; + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + + if (len == 0) { + _mesa_printf("Error - zero length packet (0x%08x)\n", stream->ptr[0]); + assert(0); + return GL_FALSE; } - fprintf(stderr, "%s[%d]", regname[type], nr); -} + if (stream->print_addresses) + _mesa_printf("%08x: ", stream->offset); + -#define REG_SWIZZLE_MASK 0x7777 -#define REG_NEGATE_MASK 0x8888 + _mesa_printf("%s (%d dwords):\n", name, len); + for (i = 0; i < len; i++) + _mesa_printf("\t\t0x%08x\n", ptr[i]); + _mesa_printf("\n"); -#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ - (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ - (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ - (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) + stream->offset += len * sizeof(GLuint); + + return GL_TRUE; +} -static void -print_reg_neg_swizzle(unsigned reg) +static const char *get_prim_name( GLuint val ) { - int i; - - if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && - (reg & REG_NEGATE_MASK) == 0) - return; - - fprintf(stderr, "."); - - for (i = 3; i >= 0; i--) { - if (reg & (1 << ((i * 4) + 3))) - fprintf(stderr, "-"); - - switch ((reg >> (i * 4)) & 0x7) { - case 0: - fprintf(stderr, "x"); - break; - case 1: - fprintf(stderr, "y"); - break; - case 2: - fprintf(stderr, "z"); - break; - case 3: - fprintf(stderr, "w"); - break; - case 4: - fprintf(stderr, "0"); - break; - case 5: - fprintf(stderr, "1"); - break; - default: - fprintf(stderr, "?"); - break; - } + switch (val & PRIM3D_MASK) { + case PRIM3D_TRILIST: return "TRILIST"; break; + case PRIM3D_TRISTRIP: return "TRISTRIP"; break; + case PRIM3D_TRISTRIP_RVRSE: return "TRISTRIP_RVRSE"; break; + case PRIM3D_TRIFAN: return "TRIFAN"; break; + case PRIM3D_POLY: return "POLY"; break; + case PRIM3D_LINELIST: return "LINELIST"; break; + case PRIM3D_LINESTRIP: return "LINESTRIP"; break; + case PRIM3D_RECTLIST: return "RECTLIST"; break; + case PRIM3D_POINTLIST: return "POINTLIST"; break; + case PRIM3D_DIB: return "DIB"; break; + case PRIM3D_CLEAR_RECT: return "CLEAR_RECT"; break; + case PRIM3D_ZONE_INIT: return "ZONE_INIT"; break; + default: return "????"; break; } } - -static void -print_src_reg(unsigned dword) +static GLboolean debug_prim( struct debug_stream *stream, const char *name, + GLboolean dump_floats, + GLuint len ) { - unsigned nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; - unsigned type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(type, nr); - print_reg_neg_swizzle(dword); + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + const char *prim = get_prim_name( ptr[0] ); + GLuint i; + + + + _mesa_printf("%s %s (%d dwords):\n", name, prim, len); + _mesa_printf("\t\t0x%08x\n", ptr[0]); + for (i = 1; i < len; i++) { + if (dump_floats) + _mesa_printf("\t\t0x%08x // %f\n", ptr[i], *(GLfloat *)&ptr[i]); + else + _mesa_printf("\t\t0x%08x\n", ptr[i]); + } + + + _mesa_printf("\n"); + + stream->offset += len * sizeof(GLuint); + + return GL_TRUE; } + -void -i915_print_ureg(const char *msg, unsigned ureg) + + +static GLboolean debug_program( struct debug_stream *stream, const char *name, GLuint len ) { - fprintf(stderr, "%s: ", msg); - print_src_reg(ureg >> 8); - fprintf(stderr, "\n"); + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + + if (len == 0) { + _mesa_printf("Error - zero length packet (0x%08x)\n", stream->ptr[0]); + assert(0); + return GL_FALSE; + } + + if (stream->print_addresses) + _mesa_printf("%08x: ", stream->offset); + + _mesa_printf("%s (%d dwords):\n", name, len); + i915_disassemble_program( ptr, len ); + + stream->offset += len * sizeof(GLuint); + return GL_TRUE; } -static void -print_dest_reg(unsigned dword) + +static GLboolean debug_chain( struct debug_stream *stream, const char *name, GLuint len ) { - unsigned nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; - unsigned type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(type, nr); - if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) - return; - fprintf(stderr, "."); - if (dword & A0_DEST_CHANNEL_X) - fprintf(stderr, "x"); - if (dword & A0_DEST_CHANNEL_Y) - fprintf(stderr, "y"); - if (dword & A0_DEST_CHANNEL_Z) - fprintf(stderr, "z"); - if (dword & A0_DEST_CHANNEL_W) - fprintf(stderr, "w"); -} + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + GLuint old_offset = stream->offset + len * sizeof(GLuint); + GLuint i; + + _mesa_printf("%s (%d dwords):\n", name, len); + for (i = 0; i < len; i++) + _mesa_printf("\t\t0x%08x\n", ptr[i]); + stream->offset = ptr[1] & ~0x3; + + if (stream->offset < old_offset) + _mesa_printf("\n... skipping backwards from 0x%x --> 0x%x ...\n\n", + old_offset, stream->offset ); + else + _mesa_printf("\n... skipping from 0x%x --> 0x%x ...\n\n", + old_offset, stream->offset ); -#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) -#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) -#define GET_SRC2_REG(r) (r) + + return GL_TRUE; +} -static void -print_arith_op(unsigned opcode, const unsigned * program) +static GLboolean debug_variable_length_prim( struct debug_stream *stream ) { - if (opcode != A0_NOP) { - print_dest_reg(program[0]); - if (program[0] & A0_DEST_SATURATE) - fprintf(stderr, " = SATURATE "); - else - fprintf(stderr, " = "); - } + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + const char *prim = get_prim_name( ptr[0] ); + GLuint i, len; - fprintf(stderr, "%s ", opcodes[opcode]); + GLushort *idx = (GLushort *)(ptr+1); + for (i = 0; idx[i] != 0xffff; i++) + ; - print_src_reg(GET_SRC0_REG(program[0], program[1])); - if (args[opcode] == 1) { - fprintf(stderr, "\n"); - return; - } + len = 1+(i+2)/2; - fprintf(stderr, ", "); - print_src_reg(GET_SRC1_REG(program[1], program[2])); - if (args[opcode] == 2) { - fprintf(stderr, "\n"); - return; + _mesa_printf("3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); + for (i = 0; i < len; i++) + _mesa_printf("\t\t0x%08x\n", ptr[i]); + _mesa_printf("\n"); + + stream->offset += len * sizeof(GLuint); + return GL_TRUE; +} + + +static GLboolean debug_load_immediate( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + GLuint bits = (ptr[0] >> 4) & 0xff; + GLuint i, j = 0; + + _mesa_printf("%s (%d dwords, flags: %x):\n", name, len, bits); + _mesa_printf("\t\t0x%08x\n", ptr[j++]); + + for (i = 0; i < 8; i++) { + if (bits & (1<offset += len * sizeof(GLuint); + + return GL_TRUE; } + -static void -print_tex_op(unsigned opcode, const unsigned * program) +static GLboolean debug_load_indirect( struct debug_stream *stream, + const char *name, + GLuint len ) { - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - fprintf(stderr, " = "); + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + GLuint bits = (ptr[0] >> 8) & 0x3f; + GLuint i, j = 0; + + _mesa_printf("%s (%d dwords):\n", name, len); + _mesa_printf("\t\t0x%08x\n", ptr[j++]); + + for (i = 0; i < 6; i++) { + if (bits & (1<> T1_ADDRESS_REG_TYPE_SHIFT) & - REG_TYPE_MASK, - (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); - fprintf(stderr, "\n"); + + assert(j == len); + + stream->offset += len * sizeof(GLuint); + + return GL_TRUE; } + -static void -print_dcl_op(unsigned opcode, const unsigned * program) +static GLboolean i915_debug_packet( struct debug_stream *stream ) { - fprintf(stderr, "%s ", opcodes[opcode]); - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - fprintf(stderr, "\n"); + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + GLuint cmd = *ptr; + + switch (((cmd >> 29) & 0x7)) { + case 0x0: + switch ((cmd >> 23) & 0x3f) { + case 0x0: + return debug(stream, "MI_NOOP", 1); + case 0x3: + return debug(stream, "MI_WAIT_FOR_EVENT", 1); + case 0x4: + return debug(stream, "MI_FLUSH", 1); + case 0xA: + debug(stream, "MI_BATCH_BUFFER_END", 1); + return GL_FALSE; + case 0x22: + return debug(stream, "MI_LOAD_REGISTER_IMM", 3); + case 0x31: + return debug_chain(stream, "MI_BATCH_BUFFER_START", 2); + default: + break; + } + break; + case 0x1: + break; + case 0x2: + switch ((cmd >> 22) & 0xff) { + case 0x50: + return debug(stream, "XY_COLOR_BLT", (cmd & 0xff) + 2); + case 0x53: + return debug(stream, "XY_SRC_COPY_BLT", (cmd & 0xff) + 2); + default: + return debug(stream, "blit command", (cmd & 0xff) + 2); + } + break; + case 0x3: + switch ((cmd >> 24) & 0x1f) { + case 0x6: + return debug(stream, "3DSTATE_ANTI_ALIASING", 1); + case 0x7: + return debug(stream, "3DSTATE_RASTERIZATION_RULES", 1); + case 0x8: + return debug(stream, "3DSTATE_BACKFACE_STENCIL_OPS", 2); + case 0x9: + return debug(stream, "3DSTATE_BACKFACE_STENCIL_MASKS", 1); + case 0xb: + return debug(stream, "3DSTATE_INDEPENDENT_ALPHA_BLEND", 1); + case 0xc: + return debug(stream, "3DSTATE_MODES5", 1); + case 0xd: + return debug(stream, "3DSTATE_MODES4", 1); + case 0x15: + return debug(stream, "3DSTATE_FOG_COLOR", 1); + case 0x16: + return debug(stream, "3DSTATE_COORD_SET_BINDINGS", 1); + case 0x1c: + /* 3DState16NP */ + switch((cmd >> 19) & 0x1f) { + case 0x10: + return debug(stream, "3DSTATE_SCISSOR_ENABLE", 1); + case 0x11: + return debug(stream, "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE", 1); + default: + break; + } + break; + case 0x1d: + /* 3DStateMW */ + switch ((cmd >> 16) & 0xff) { + case 0x0: + return debug(stream, "3DSTATE_MAP_STATE", (cmd & 0x1f) + 2); + case 0x1: + return debug(stream, "3DSTATE_SAMPLER_STATE", (cmd & 0x1f) + 2); + case 0x4: + return debug_load_immediate(stream, "3DSTATE_LOAD_STATE_IMMEDIATE", (cmd & 0xf) + 2); + case 0x5: + return debug_program(stream, "3DSTATE_PIXEL_SHADER_PROGRAM", (cmd & 0x1ff) + 2); + case 0x6: + return debug(stream, "3DSTATE_PIXEL_SHADER_CONSTANTS", (cmd & 0xff) + 2); + case 0x7: + return debug_load_indirect(stream, "3DSTATE_LOAD_INDIRECT", (cmd & 0xff) + 2); + case 0x80: + return debug(stream, "3DSTATE_DRAWING_RECTANGLE", (cmd & 0xffff) + 2); + case 0x81: + return debug(stream, "3DSTATE_SCISSOR_RECTANGLE", (cmd & 0xffff) + 2); + case 0x83: + return debug(stream, "3DSTATE_SPAN_STIPPLE", (cmd & 0xffff) + 2); + case 0x85: + return debug(stream, "3DSTATE_DEST_BUFFER_VARS", (cmd & 0xffff) + 2); + case 0x88: + return debug(stream, "3DSTATE_CONSTANT_BLEND_COLOR", (cmd & 0xffff) + 2); + case 0x89: + return debug(stream, "3DSTATE_FOG_MODE", (cmd & 0xffff) + 2); + case 0x8e: + return debug(stream, "3DSTATE_BUFFER_INFO", (cmd & 0xffff) + 2); + case 0x97: + return debug(stream, "3DSTATE_DEPTH_OFFSET_SCALE", (cmd & 0xffff) + 2); + case 0x98: + return debug(stream, "3DSTATE_DEFAULT_Z", (cmd & 0xffff) + 2); + case 0x99: + return debug(stream, "3DSTATE_DEFAULT_DIFFUSE", (cmd & 0xffff) + 2); + case 0x9a: + return debug(stream, "3DSTATE_DEFAULT_SPECULAR", (cmd & 0xffff) + 2); + case 0x9c: + return debug(stream, "3DSTATE_CLEAR_PARAMETERS", (cmd & 0xffff) + 2); + default: + assert(0); + return 0; + } + break; + case 0x1e: + if (cmd & (1 << 23)) + return debug(stream, "???", (cmd & 0xffff) + 1); + else + return debug(stream, "", 1); + break; + case 0x1f: + if ((cmd & (1 << 23)) == 0) + return debug_prim(stream, "3DPRIM (inline)", 1, (cmd & 0x1ffff) + 2); + else if (cmd & (1 << 17)) + { + if ((cmd & 0xffff) == 0) + return debug_variable_length_prim(stream); + else + return debug_prim(stream, "3DPRIM (indexed)", 0, (((cmd & 0xffff) + 1) / 2) + 1); + } + else + return debug_prim(stream, "3DPRIM (indirect sequential)", 0, 2); + break; + default: + return debug(stream, "", 0); + } + default: + assert(0); + return 0; + } + + assert(0); + return 0; } + void -i915_disassemble_program(const unsigned * program, unsigned sz) +i915_dump_batchbuffer( struct i915_context *i915, + GLuint *start, + GLuint *end ) { - unsigned size = program[0] & 0x1ff; - unsigned i; + struct debug_stream stream; + GLuint bytes = (end - start) * 4; + GLboolean done = GL_FALSE; - fprintf(stderr, "BEGIN\n"); + fprintf(stderr, "\n\nBATCH: (%d)\n", bytes / 4); - if (size + 2 != sz) { - fprintf(stderr, "%s: program size mismatch %d/%d\n", __FUNCTION__, - size + 2, sz); - assert(0); - } + stream.offset = 0; + stream.ptr = (char *)start; + stream.print_addresses = 0; - program++; - for (i = 1; i < sz; i += 3, program += 3) { - unsigned opcode = program[0] & (0x1f << 24); + while (!done && + stream.offset < bytes && + stream.offset >= 0) + { + if (!i915_debug_packet( &stream )) + break; - if ((unsigned) opcode >= A0_NOP && opcode <= A0_SLT) - print_arith_op(opcode >> 24, program); - else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) - print_tex_op(opcode >> 24, program); - else if (opcode == D0_DCL) - print_dcl_op(opcode >> 24, program); - else - fprintf(stderr, "Unknown opcode 0x%x\n", opcode); + assert(stream.offset <= bytes && + stream.offset >= 0); } - fprintf(stderr, "END\n\n"); + fprintf(stderr, "END-BATCH\n\n\n"); } + + diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h index 47a02401ee..3e01db6434 100644 --- a/src/mesa/pipe/i915simple/i915_debug.h +++ b/src/mesa/pipe/i915simple/i915_debug.h @@ -31,9 +31,26 @@ #ifndef I915_DEBUG_H #define I915_DEBUG_H +struct i915_context; + +struct debug_stream +{ + unsigned offset; /* current gtt offset */ + char *ptr; /* pointer to gtt offset zero */ + char *end; /* pointer to gtt offset zero */ + unsigned print_addresses; +}; + + extern void i915_disassemble_program(const unsigned *program, unsigned sz); extern void i915_print_ureg(const char *msg, unsigned ureg); +void +i915_dump_batchbuffer( struct i915_context *i915, + unsigned *start, + unsigned *end ); + + #endif diff --git a/src/mesa/pipe/i915simple/i915_debug_fp.c b/src/mesa/pipe/i915simple/i915_debug_fp.c new file mode 100644 index 0000000000..f4e8ae95ad --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_debug_fp.c @@ -0,0 +1,331 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include + +#include "i915_reg.h" +#include "i915_debug.h" +//#include "i915_fpc.h" +#include "shader/program.h" +#include "shader/prog_instruction.h" +#include "shader/prog_print.h" + +static const char *opcodes[0x20] = { + "NOP", + "ADD", + "MOV", + "MUL", + "MAD", + "DP2ADD", + "DP3", + "DP4", + "FRC", + "RCP", + "RSQ", + "EXP", + "LOG", + "CMP", + "MIN", + "MAX", + "FLR", + "MOD", + "TRC", + "SGE", + "SLT", + "TEXLD", + "TEXLDP", + "TEXLDB", + "TEXKILL", + "DCL", + "0x1a", + "0x1b", + "0x1c", + "0x1d", + "0x1e", + "0x1f", +}; + + +static const int args[0x20] = { + 0, /* 0 nop */ + 2, /* 1 add */ + 1, /* 2 mov */ + 2, /* 3 m ul */ + 3, /* 4 mad */ + 3, /* 5 dp2add */ + 2, /* 6 dp3 */ + 2, /* 7 dp4 */ + 1, /* 8 frc */ + 1, /* 9 rcp */ + 1, /* a rsq */ + 1, /* b exp */ + 1, /* c log */ + 3, /* d cmp */ + 2, /* e min */ + 2, /* f max */ + 1, /* 10 flr */ + 1, /* 11 mod */ + 1, /* 12 trc */ + 2, /* 13 sge */ + 2, /* 14 slt */ + 1, + 1, + 1, + 1, + 0, + 0, + 0, + 0, + 0, + 0, + 0, +}; + + +static const char *regname[0x8] = { + "R", + "T", + "CONST", + "S", + "OC", + "OD", + "U", + "UNKNOWN", +}; + +static void +print_reg_type_nr(GLuint type, GLuint nr) +{ + switch (type) { + case REG_TYPE_T: + switch (nr) { + case T_DIFFUSE: + _mesa_printf("T_DIFFUSE"); + return; + case T_SPECULAR: + _mesa_printf("T_SPECULAR"); + return; + case T_FOG_W: + _mesa_printf("T_FOG_W"); + return; + default: + _mesa_printf("T_TEX%d", nr); + return; + } + case REG_TYPE_OC: + if (nr == 0) { + _mesa_printf("oC"); + return; + } + break; + case REG_TYPE_OD: + if (nr == 0) { + _mesa_printf("oD"); + return; + } + break; + default: + break; + } + + _mesa_printf("%s[%d]", regname[type], nr); +} + +#define REG_SWIZZLE_MASK 0x7777 +#define REG_NEGATE_MASK 0x8888 + +#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ + (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ + (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ + (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) + + +static void +print_reg_neg_swizzle(GLuint reg) +{ + int i; + + if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && + (reg & REG_NEGATE_MASK) == 0) + return; + + _mesa_printf("."); + + for (i = 3; i >= 0; i--) { + if (reg & (1 << ((i * 4) + 3))) + _mesa_printf("-"); + + switch ((reg >> (i * 4)) & 0x7) { + case 0: + _mesa_printf("x"); + break; + case 1: + _mesa_printf("y"); + break; + case 2: + _mesa_printf("z"); + break; + case 3: + _mesa_printf("w"); + break; + case 4: + _mesa_printf("0"); + break; + case 5: + _mesa_printf("1"); + break; + default: + _mesa_printf("?"); + break; + } + } +} + + +static void +print_src_reg(GLuint dword) +{ + GLuint nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; + GLuint type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(type, nr); + print_reg_neg_swizzle(dword); +} + + +static void +print_dest_reg(GLuint dword) +{ + GLuint nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; + GLuint type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(type, nr); + if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) + return; + _mesa_printf("."); + if (dword & A0_DEST_CHANNEL_X) + _mesa_printf("x"); + if (dword & A0_DEST_CHANNEL_Y) + _mesa_printf("y"); + if (dword & A0_DEST_CHANNEL_Z) + _mesa_printf("z"); + if (dword & A0_DEST_CHANNEL_W) + _mesa_printf("w"); +} + + +#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) +#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) +#define GET_SRC2_REG(r) (r) + + +static void +print_arith_op(GLuint opcode, const GLuint * program) +{ + if (opcode != A0_NOP) { + print_dest_reg(program[0]); + if (program[0] & A0_DEST_SATURATE) + _mesa_printf(" = SATURATE "); + else + _mesa_printf(" = "); + } + + _mesa_printf("%s ", opcodes[opcode]); + + print_src_reg(GET_SRC0_REG(program[0], program[1])); + if (args[opcode] == 1) { + _mesa_printf("\n"); + return; + } + + _mesa_printf(", "); + print_src_reg(GET_SRC1_REG(program[1], program[2])); + if (args[opcode] == 2) { + _mesa_printf("\n"); + return; + } + + _mesa_printf(", "); + print_src_reg(GET_SRC2_REG(program[2])); + _mesa_printf("\n"); + return; +} + + +static void +print_tex_op(GLuint opcode, const GLuint * program) +{ + print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); + _mesa_printf(" = "); + + _mesa_printf("%s ", opcodes[opcode]); + + _mesa_printf("S[%d],", program[0] & T0_SAMPLER_NR_MASK); + + print_reg_type_nr((program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & + REG_TYPE_MASK, + (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); + _mesa_printf("\n"); +} + +static void +print_dcl_op(GLuint opcode, const GLuint * program) +{ + _mesa_printf("%s ", opcodes[opcode]); + print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); + _mesa_printf("\n"); +} + + +void +i915_disassemble_program(const GLuint * program, GLuint sz) +{ + GLuint size = program[0] & 0x1ff; + GLint i; + + _mesa_printf("\t\tBEGIN\n"); + + assert(size + 2 == sz); + + program++; + for (i = 1; i < sz; i += 3, program += 3) { + GLuint opcode = program[0] & (0x1f << 24); + + _mesa_printf("\t\t"); + + if ((GLint) opcode >= A0_NOP && opcode <= A0_SLT) + print_arith_op(opcode >> 24, program); + else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) + print_tex_op(opcode >> 24, program); + else if (opcode == D0_DCL) + print_dcl_op(opcode >> 24, program); + else + _mesa_printf("Unknown opcode 0x%x\n", opcode); + } + + _mesa_printf("\t\tEND\n\n"); +} + + -- cgit v1.2.3 From 93b6e8aeac8ee410ee49cecd28a785efe5c2b1e3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 11:19:09 +0100 Subject: Tweaks to state emit, acheives first hw triangles from a pipe driver. --- src/mesa/pipe/i915simple/Makefile | 1 + src/mesa/pipe/i915simple/i915_batch.h | 2 +- src/mesa/pipe/i915simple/i915_state_emit.c | 24 +++++++++++++++++------- src/mesa/pipe/i915simple/i915_state_fragprog.c | 2 +- 4 files changed, 20 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index f8b14220fb..93d9c6cf01 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -12,6 +12,7 @@ DRIVER_SOURCES = \ i915_context.c \ i915_context.c \ i915_debug.c \ + i915_debug_fp.c \ i915_regions.c \ i915_state.c \ i915_state_derived.c \ diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h index b4ca8c38da..8a35ff6a3a 100644 --- a/src/mesa/pipe/i915simple/i915_batch.h +++ b/src/mesa/pipe/i915simple/i915_batch.h @@ -45,7 +45,7 @@ #define ADVANCE_BATCH() #define FLUSH_BATCH() do { \ -/* i915_dump_batchbuffer( i915, i915->batch_start, BEGIN_BATCH(0, 0) ); */ \ + i915_dump_batchbuffer( i915, i915->batch_start, BEGIN_BATCH(0, 0) ); \ i915->winsys->batch_flush( i915->winsys ); \ i915->batch_start = BEGIN_BATCH(0, 0); \ } while (0) diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 4aa7e28a86..aeb2d6753f 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -112,7 +112,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) /* Need to initialize this to zero. */ { - OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (1)); + OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (0)); OUT_BATCH(0); } @@ -138,6 +138,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) { /* Don't support twosided stencil yet */ OUT_BATCH(_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_TWO_SIDE | 0); + OUT_BATCH(0); } @@ -150,11 +151,10 @@ i915_emit_hardware_state(struct i915_context *i915 ) I1_LOAD_S(6) | (3)); - OUT_BATCH(0); - OUT_BATCH(S4_VFMT_XYZ | S4_VFMT_COLOR); - OUT_BATCH(0); - OUT_BATCH(S6_COLOR_WRITE_ENABLE | - (2 << S6_TRISTRIP_PV_SHIFT)); + OUT_BATCH(0xffffffff); + OUT_BATCH(0x00902440); // OUT_BATCH(S4_VFMT_XYZ | S4_VFMT_COLOR); + OUT_BATCH(0x00000002); + OUT_BATCH(0x00020216); // OUT_BATCH( S6_COLOR_WRITE_ENABLE | (2 << S6_TRISTRIP_PV_SHIFT)); } { @@ -167,7 +167,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) STENCIL_WRITE_MASK(0xff)); } - { + if (0) { OUT_BATCH(_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | IAB_MODIFY_ENABLE | IAB_MODIFY_FUNC | @@ -175,6 +175,16 @@ i915_emit_hardware_state(struct i915_context *i915 ) IAB_MODIFY_DST_FACTOR); } + { + //3DSTATE_INDEPENDENT_ALPHA_BLEND (1 dwords): + OUT_BATCH(0x6ba008a1); + + //3DSTATE_CONSTANT_BLEND_COLOR (2 dwords): + OUT_BATCH(0x7d880000); + OUT_BATCH(0x00000000); + } + + if (i915->framebuffer.cbufs[0]) { struct pipe_region *cbuf_region = i915->framebuffer.cbufs[0]->region; diff --git a/src/mesa/pipe/i915simple/i915_state_fragprog.c b/src/mesa/pipe/i915simple/i915_state_fragprog.c index e4639bc223..83f43befaf 100644 --- a/src/mesa/pipe/i915simple/i915_state_fragprog.c +++ b/src/mesa/pipe/i915simple/i915_state_fragprog.c @@ -50,7 +50,7 @@ static unsigned passthrough[] = A0_DEST_CHANNEL_ALL | (REG_TYPE_T << A0_SRC0_TYPE_SHIFT) | (T_DIFFUSE << A0_SRC0_NR_SHIFT)), - 0, + 0x01230000, /* .xyzw */ 0 }; -- cgit v1.2.3 From 13eaf04b56218506bb7967df9aa37b692c44338f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 11:20:38 +0100 Subject: Remove unreferenced file. --- src/mesa/drivers/dri/intel_winsys/intel_tex.h | 212 -------------------------- 1 file changed, 212 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex.h b/src/mesa/drivers/dri/intel_winsys/intel_tex.h deleted file mode 100644 index 7a1cc91974..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex.h +++ /dev/null @@ -1,212 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTELTEX_INC -#define INTELTEX_INC - -#include "mtypes.h" -#include "intel_context.h" -#include "texmem.h" - -struct st_texture_object -{ - struct gl_texture_object base; /* The "parent" object */ - - /* The mipmap tree must include at least these levels once - * validated: - */ - GLuint firstLevel; - GLuint lastLevel; - - /* Offset for firstLevel image: - */ - GLuint textureOffset; - - /* On validation any active images held in main memory or in other - * regions will be copied to this region and the old storage freed. - */ - struct pipe_mipmap_tree *mt; - - GLboolean imageOverride; - GLint depthOverride; - GLuint pitchOverride; -}; - - - -struct st_texture_image -{ - struct gl_texture_image base; - - /* These aren't stored in gl_texture_image - */ - GLuint level; - GLuint face; - - /* If stImage->mt != NULL, image data is stored here. - * Else if stImage->base.Data != NULL, image is stored there. - * Else there is no image data. - */ - struct pipe_mipmap_tree *mt; -}; - - -void intelInitTextureFuncs(struct dd_function_table *functions); - -const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type); - - -void intelTexImage3D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage3D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexImage2D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage2D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexImage1D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage1D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLint border); - -void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint border); - -void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint x, GLint y, GLsizei width); - -void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLint x, GLint y, GLsizei width, GLsizei height); - -void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ); - -void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, - GLvoid *pixels, - const struct gl_texture_object *texObj, - const struct gl_texture_image *texImage); - -void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, - unsigned long long offset, GLint depth, GLuint pitch); - -GLuint intel_finalize_mipmap_tree(GLcontext *ctx, - struct pipe_context *pipe, GLuint unit, - GLboolean *needFlush); - - -#if 0 -void intel_tex_map_images(struct pipe_context *pipe, - struct st_texture_object *stObj); - -void intel_tex_unmap_images(struct pipe_context *pipe, - struct st_texture_object *stObj); -#endif - -int intel_compressed_num_bytes(GLuint mesaFormat); - - -static INLINE struct st_texture_object * -st_texture_object(struct gl_texture_object *obj) -{ - return (struct st_texture_object *) obj; -} - -static INLINE struct st_texture_image * -st_texture_image(struct gl_texture_image *img) -{ - return (struct st_texture_image *) img; -} - - -#endif -- cgit v1.2.3 From 4e517f9b0aed27b49c6050a3d7c279efc625f352 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 11:21:22 +0100 Subject: Supporting batchbuffer changes for 1st hwpipe triangles... --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 12 +++++++++--- .../dri/intel_winsys/intel_pipe_i915simple.c | 21 +++++++++++++-------- 2 files changed, 22 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 80a3163d55..d1fb2907a9 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -72,9 +72,9 @@ intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) { int i; fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); - for (i = 0; i < count / 4; i += 4) - fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", - offset + i * 4, ptr[i], ptr[i + 1], ptr[i + 2], ptr[i + 3]); + for (i = 0; i < count / 4; i += 1) + fprintf(stderr, "\t0x%08x\n", + ptr[i]); fprintf(stderr, "END BATCH\n\n\n"); } @@ -84,6 +84,12 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) int i; + if (batch->map) { + driBOUnmap(batch->buffer); + batch->map = NULL; + } + + /* * Get a new, free batchbuffer. */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c index 244e32c885..98d446ba90 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c @@ -171,8 +171,7 @@ static unsigned *intel_i915_batch_start( struct i915_winsys *sws, if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { /* XXX: Hmm, the driver can't really do much with this pointer: */ - //return intel->batch->ptr; - return (unsigned *)~0; + return intel->batch->ptr; } else return NULL; @@ -210,12 +209,6 @@ static void intel_i915_batch_reloc( struct i915_winsys *sws, delta ); } -static void intel_i915_batch_flush( struct i915_winsys *sws ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - - intel_batchbuffer_flush( intel->batch ); -} static void intel_i915_batch_wait_idle( struct i915_winsys *sws ) { @@ -229,6 +222,18 @@ static void intel_i915_batch_wait_idle( struct i915_winsys *sws ) } } + +static void intel_i915_batch_flush( struct i915_winsys *sws ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + _mesa_printf("%s: start\n"); + intel_batchbuffer_flush( intel->batch ); + intel_i915_batch_wait_idle( sws ); + _mesa_printf("%s: done\n"); +} + + struct pipe_context * intel_create_i915simple( struct intel_context *intel ) { -- cgit v1.2.3 From 17dde8dbb131a07f4a1109615fa80b60a7857a60 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 11:22:29 +0100 Subject: Quieten TGSI --- src/mesa/pipe/tgsi/core/tgsi_dump.c | 4 +++- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 13 +++++++++---- 2 files changed, 12 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index f1773ac908..c3d796afee 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -1277,6 +1277,8 @@ tgsi_dump( tgsi_parse_free( &parse ); - fclose( dump->file ); + if (dump->file != stderr && + dump->file != stdout) + fclose( dump->file ); } diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 107b6bd2a3..c3201ece2c 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -1,6 +1,8 @@ #include "tgsi_platform.h" #include "tgsi_mesa.h" +#define TGSI_DEBUG 0 + /* * Map mesa register file to SBIR register file. */ @@ -571,9 +573,11 @@ tgsi_mesa_compile_fp_program( preamble_size, TGSI_PROCESSOR_FRAGMENT ) ) { assert( i == program->Base.NumInstructions - 1 ); - tgsi_dump( - tokens, - 0/*TGSI_DUMP_VERBOSE | TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT*/ ); + + if (TGSI_DEBUG) + tgsi_dump( + tokens, + 0/*TGSI_DUMP_VERBOSE | TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT*/ ); break; } @@ -617,7 +621,8 @@ tgsi_mesa_compile_vp_program( 0, TGSI_PROCESSOR_VERTEX ) ) { assert( ii == program->Base.NumInstructions - 1 ); - tgsi_dump( tokens, TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT ); + if (TGSI_DEBUG) + tgsi_dump( tokens, TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT ); break; } -- cgit v1.2.3 From 5279b267688ecf0eb8a9b9873d80aa399538303e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 11:22:45 +0100 Subject: Quieten TGSI --- src/mesa/state_tracker/st_atom_fs.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 9ca1807913..58875de3f9 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -37,6 +37,8 @@ #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/core/tgsi_dump.h" +#define TGSI_DEBUG 0 + static void compile_fs( struct st_context *st, struct st_fragment_program *fs ) { @@ -44,7 +46,8 @@ static void compile_fs( struct st_context *st, */ tgsi_mesa_compile_fp_program( &fs->Base, fs->tokens, ST_FP_MAX_TOKENS ); - tgsi_dump( fs->tokens, TGSI_DUMP_VERBOSE ); + if (TGSI_DEBUG) + tgsi_dump( fs->tokens, TGSI_DUMP_VERBOSE ); } -- cgit v1.2.3 From 81e595dfe2b9e2d7a27d49da365907bcea850f14 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 9 Aug 2007 12:49:46 +0100 Subject: Further opcode list reorg. Add TXP opcode. Alias SGN to SSG. Alias EXPP to EXP. --- src/mesa/pipe/tgsi/core/tgsi_token.h | 41 ++++++++++++++++++------------------ 1 file changed, 21 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index 8848e99808..8e82c9b275 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -231,7 +231,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_SUB */ #define TGSI_OPCODE_TEX 52 #define TGSI_OPCODE_TXD 53 -/* TGSI_OPCODE_TXP - use TGSI_OPCODE_TEX */ +#define TGSI_OPCODE_TXP 132 #define TGSI_OPCODE_UP2H 54 #define TGSI_OPCODE_UP2US 55 #define TGSI_OPCODE_UP4B 56 @@ -730,6 +730,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_TXB */ /* TGSI_OPCODE_TXD */ /* TGSI_OPCODE_TXL */ +/* TGSI_OPCODE_TXP */ /* TGSI_OPCODE_XPD */ /* @@ -787,7 +788,7 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_TEXDP3TEX 118 #define TGSI_OPCODE_TEXDP3 119 #define TGSI_OPCODE_TEXM3X3 120 -/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ +/* CMP - use TGSI_OPCODE_CND0 */ /* * ps_1_3 @@ -820,7 +821,7 @@ struct tgsi_immediate_float32 #define TEXM3X2DEPTH 121 /* TGSI_OPCODE_TEXDP3 */ /* TGSI_OPCODE_TEXM3X3 */ -/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ +/* CMP - use TGSI_OPCODE_CND0 */ /* * ps_1_4 @@ -839,7 +840,7 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX /* TGSI_OPCODE_CND */ #define TGSI_OPCODE_TEXDEPTH 122 -/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ +/* CMP - use TGSI_OPCODE_CND0 */ #define TGSI_OPCODE_BEM 123 /* @@ -857,8 +858,8 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_DP4 */ /* TGSI_OPCODE_MIN */ /* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ /* TGSI_OPCODE_LRP */ /* TGSI_OPCODE_FRC */ #define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX @@ -874,8 +875,8 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_TEXKILL */ /* TGSI_OPCODE_TEXLD */ #define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB -#define TGSI_OPCODE_TEXLDP TGSI_OPCODE_TEX -/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ +#define TGSI_OPCODE_TEXLDP TGSI_OPCODE_TXP +/* CMP - use TGSI_OPCODE_CND0 */ #define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A /* @@ -893,8 +894,8 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_DP4 */ /* TGSI_OPCODE_MIN */ /* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ /* TGSI_OPCODE_LRP */ /* TGSI_OPCODE_FRC */ /* TGSI_OPCODE_M4X4 */ @@ -921,7 +922,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_TEXKILL */ /* TGSI_OPCODE_TEXLD */ /* TGSI_OPCODE_TEXLDB */ -/* TGSI_OPCODE_CMPDX - use TGSI_OPCODE_CND0 */ +/* CMP - use TGSI_OPCODE_CND0 */ /* TGSI_OPCODE_DP2ADD */ #define TGSI_OPCODE_DSX TGSI_OPCODE_DDX #define TGSI_OPCODE_DSY TGSI_OPCODE_DDY @@ -945,8 +946,8 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MAX */ /* TGSI_OPCODE_SLT */ /* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ /* TGSI_OPCODE_LIT */ /* TGSI_OPCODE_DST */ /* TGSI_OPCODE_FRC */ @@ -955,7 +956,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_M3X4 */ /* TGSI_OPCODE_M3X3 */ /* TGSI_OPCODE_M3X2 */ -#define TGSI_OPCODE_EXPP 132 +#define TGSI_OPCODE_EXPP TGSI_OPCODE_EXP #define TGSI_OPCODE_LOGP TGSI_OPCODE_LG2 /* @@ -975,8 +976,8 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MAX */ /* TGSI_OPCODE_SLT */ /* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ /* TGSI_OPCODE_LIT */ /* TGSI_OPCODE_DST */ /* TGSI_OPCODE_LRP */ @@ -993,7 +994,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_ENDLOOP */ /* TGSI_OPCODE_POW */ /* XXX: takes ABS */ /* TGSI_OPCODE_CRS */ -#define TGSI_OPCODE_SGN 133 +#define TGSI_OPCODE_SGN TGSI_OPCODE_SSG /* TGSI_OPCODE_ABS */ /* TGSI_OPCODE_NRM4 */ /* TGSI_OPCODE_SINCOS */ @@ -1022,8 +1023,8 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MAX */ /* TGSI_OPCODE_SLT */ /* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_EXPDX - use TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LOGDX - use TGSI_OPCODE_LG2 */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ /* TGSI_OPCODE_LIT */ /* TGSI_OPCODE_DST */ /* TGSI_OPCODE_LRP */ @@ -1055,7 +1056,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MOVA */ /* TGSI_OPCODE_LOGP */ -#define TGSI_OPCODE_LAST 134 +#define TGSI_OPCODE_LAST 133 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ -- cgit v1.2.3 From ba75e41e41f70d8a4ab01399fe2cd0ab15f09b39 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 9 Aug 2007 12:55:50 +0100 Subject: Update opcode list. --- src/mesa/pipe/tgsi/core/tgsi_dump.c | 55 ++++++++++++++++++++++++++++++++++-- src/mesa/pipe/tgsi/core/tgsi_token.h | 2 +- 2 files changed, 54 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index c3d796afee..51d3bb703d 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -317,7 +317,33 @@ static const char *TGSI_OPCODES[] = "OPCODE_NOISE2", "OPCODE_NOISE3", "OPCODE_NOISE4", - "OPCODE_NOP" + "OPCODE_NOP", + "OPCODE_TEXBEM", + "OPCODE_TEXBEML", + "OPCODE_TEXREG2AR", + "OPCODE_TEXM3X2PAD", + "OPCODE_TEXM3X2TEX", + "OPCODE_TEXM3X3PAD", + "OPCODE_TEXM3X3TEX", + "OPCODE_TEXM3X3SPEC", + "OPCODE_TEXM3X3VSPEC", + "OPCODE_TEXREG2GB", + "OPCODE_TEXREG2RGB", + "OPCODE_TEXDP3TEX", + "OPCODE_TEXDP3", + "OPCODE_TEXM3X3", + "OPCODE_TEXM3X2DEPTH", + "OPCODE_TEXDEPTH", + "OPCODE_BEM", + "OPCODE_M4X3", + "OPCODE_M3X4", + "OPCODE_M3X3", + "OPCODE_M3X2", + "OPCODE_NRM4", + "OPCODE_CALLNZ", + "OPCODE_IFC", + "OPCODE_BREAKC", + "OPCODE_TXP" }; static const char *TGSI_OPCODES_SHORT[] = @@ -428,7 +454,32 @@ static const char *TGSI_OPCODES_SHORT[] = "NOISE2", "NOISE3", "NOISE4", - "NOP" + "NOP", + "TEXBEM", + "TEXBEML", + "TEXREG2AR", + "TEXM3X2PAD", + "TEXM3X2TEX", + "TEXM3X3PAD", + "TEXM3X3TEX", + "TEXM3X3SPEC", + "TEXM3X3VSPEC", + "TEXREG2GB", + "TEXREG2RGB", + "TEXDP3TEX", + "TEXDP3", + "TEXM3X3", + "TEXM3X2DEPTH", + "TEXDEPTH", + "BEM", + "M4X3", + "M3X4", + "M3X3", + "M3X2", + "NRM4", + "CALLNZ", + "IFC", + "BREAKC", }; static const char *TGSI_SATS[] = diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index 8e82c9b275..3552331e79 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -818,7 +818,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_CND */ /* TGSI_OPCODE_TEXREG2RGB */ /* TGSI_OPCODE_TEXDP3TEX */ -#define TEXM3X2DEPTH 121 +#define TGSI_OPCODE_TEXM3X2DEPTH 121 /* TGSI_OPCODE_TEXDP3 */ /* TGSI_OPCODE_TEXM3X3 */ /* CMP - use TGSI_OPCODE_CND0 */ -- cgit v1.2.3 From 3614a9a8ae2badceca40f626c4a80f386c5d0177 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 10:33:18 -0600 Subject: checkpoint: intel_renderbuffer removal --- src/mesa/drivers/dri/intel_winsys/Makefile | 3 ++- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 3 +++ src/mesa/drivers/dri/intel_winsys/intel_fbo.c | 5 ++++- src/mesa/drivers/dri/intel_winsys/intel_fbo.h | 2 ++ src/mesa/drivers/dri/intel_winsys/intel_screen.c | 8 ++++++++ 5 files changed, 19 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index ff635eb44d..8d3d17e13e 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -23,9 +23,10 @@ DRIVER_SOURCES = \ intel_screen.c \ intel_surface.c \ intel_fbo.c \ - intel_depthstencil.c \ intel_batchpool.c +UNUSED = intel_depthstencil.c + C_SOURCES = \ $(COMMON_SOURCES) \ $(COMMON_BM_SOURCES) \ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index 3cece98079..ce217ceb56 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -41,6 +41,7 @@ #include "pipe/p_context.h" + /* This block can be removed when libdrm >= 2.3.1 is required */ #ifndef DRM_VBLANK_FLIP @@ -275,6 +276,7 @@ intelWindowMoved(struct intel_context *intel) /* Emit wait for pending flips */ +#if 0 void intel_wait_flips(struct intel_context *intel, GLuint batch_flags) { @@ -301,6 +303,7 @@ intel_wait_flips(struct intel_context *intel, GLuint batch_flags) intel_rb->pf_pending--; } } +#endif #if 0 /* Flip the front & back buffers diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c index 0b6c6a94ee..7b9bef73bd 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c @@ -148,10 +148,11 @@ intel_delete_renderbuffer(struct gl_renderbuffer *rb) DBG("freeing renderbuffer\n"); +#if 0 if (irb->PairedStencil || irb->PairedDepth) { intel_unpair_depth_stencil(ctx, irb); } - +#endif if (intel && irb->region) { intel->pipe->region_release(intel->pipe, &irb->region); } @@ -511,7 +512,9 @@ intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage) // intel_set_span_functions(&irb->Base); +#if 0 irb->RenderToTexture = GL_TRUE; +#endif return irb; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h index 0f99a3e98d..35587f364a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h @@ -77,6 +77,7 @@ struct intel_renderbuffer { struct gl_renderbuffer Base; struct pipe_region *region; +#if 0 void *pfMap; /* possibly paged flipped map pointer */ GLuint pfPitch; /* possibly paged flipped pitch */ GLboolean RenderToTexture; /* RTT? */ @@ -85,6 +86,7 @@ struct intel_renderbuffer GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */ GLuint pf_pending; /**< sequence number of pending flip */ +#endif GLuint vbl_pending; /**< vblank sequence number of pending flip */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 81ab435f42..fba0c47e59 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -46,6 +46,7 @@ #include "dri_bufpool.h" #include "pipe/p_context.h" +#include "state_tracker/st_cb_fbo.h" @@ -308,10 +309,17 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, } if (mesaVis->doubleBufferMode) { +#if 01 intel_fb->color_rb[1] = intel_new_renderbuffer_fb(rgbFormat); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, &intel_fb->color_rb[1]->Base); +#else + intel_fb->color_rb[1] + = st_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, + &intel_fb->color_rb[1]->Base); +#endif } if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { -- cgit v1.2.3 From 4ab52c412d738bd26164236e69935c2f82035b1f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 10:39:19 -0600 Subject: checkpoint: intel_renderbuffer removal (move vbl_pending field into intel_framebuffer --- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 10 +++++++--- src/mesa/drivers/dri/intel_winsys/intel_context.c | 19 +++++++++++++++---- src/mesa/drivers/dri/intel_winsys/intel_fbo.h | 4 +++- 3 files changed, 25 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index ce217ceb56..33e4b16a5c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -241,11 +241,11 @@ intelWindowMoved(struct intel_context *intel) for (i = 0; i < intel_fb->pf_num_pages; i++) { if (!intel_fb->color_rb[i] || - (intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <= + (intel_fb->vbl_waited - intel_fb->vbl_pending[i]) <= (1<<23)) continue; - vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending; + vbl.request.sequence = intel_fb->vbl_pending[i]; drmWaitVBlank(intel->driFd, &vbl); } @@ -255,7 +255,7 @@ intelWindowMoved(struct intel_context *intel) for (i = 0; i < intel_fb->pf_num_pages; i++) { if (intel_fb->color_rb[i]) - intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited; + intel_fb->vbl_pending[i] = intel_fb->vbl_waited; } } } @@ -420,10 +420,14 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) swap.sequence -= target; *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); +#if 1 + intel_fb->vbl_pending[1] = intel_fb->vbl_pending[0] = intel_fb->vbl_seq; +#else intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending = intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->vbl_pending = intel_fb->vbl_seq; +#endif if (swap.seqtype & DRM_VBLANK_FLIP) { intel_flip_renderbuffers(intel_fb); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index c1d541199f..5c184d5c8d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -574,8 +574,12 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, intel_fb->vbl_waited = intel_fb->vbl_seq; for (i = 0; i < 2; i++) { +#if 1 + intel_fb->vbl_pending[i] = intel_fb->vbl_seq; +#else if (intel_fb->color_rb[i]) intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_seq; +#endif } } } @@ -663,24 +667,31 @@ void LOCK_HARDWARE( struct intel_context *intel ) { char __ret=0; struct intel_framebuffer *intel_fb = NULL; +#if 0 struct intel_renderbuffer *intel_rb = NULL; +#else + int curbuf; +#endif _glthread_LOCK_MUTEX(lockMutex); assert(!intel->locked); if (intel->driDrawable) { intel_fb = intel->driDrawable->driverPrivate; - +#if 0 if (intel_fb) intel_rb = intel_get_renderbuffer(&intel_fb->Base, intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : BUFFER_BACK_LEFT); +#endif } - if (intel_rb && intel_fb->vblank_flags && + curbuf = 0; /* current draw buf: 0 = front, 1 = back */ + + if (intel_fb && intel_fb->vblank_flags && !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && - (intel_fb->vbl_waited - intel_rb->vbl_pending) > (1<<23)) { + (intel_fb->vbl_waited - intel_fb->vbl_pending[curbuf]) > (1<<23)) { drmVBlank vbl; vbl.request.type = DRM_VBLANK_ABSOLUTE; @@ -689,7 +700,7 @@ void LOCK_HARDWARE( struct intel_context *intel ) vbl.request.type |= DRM_VBLANK_SECONDARY; } - vbl.request.sequence = intel_rb->vbl_pending; + vbl.request.sequence = intel_fb->vbl_pending[curbuf]; drmWaitVBlank(intel->driFd, &vbl); intel_fb->vbl_waited = vbl.reply.sequence; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h index 35587f364a..63afa6df89 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h @@ -64,6 +64,8 @@ struct intel_framebuffer GLuint swap_count; GLuint swap_missed_count; + + GLuint vbl_pending[3]; /**< [number of color buffers] */ }; @@ -86,9 +88,9 @@ struct intel_renderbuffer GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */ GLuint pf_pending; /**< sequence number of pending flip */ -#endif GLuint vbl_pending; /**< vblank sequence number of pending flip */ +#endif struct intel_surface *surface; }; -- cgit v1.2.3 From d177a00e5bcf6bffb3fc0810b720d8534995c752 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 10:52:03 -0600 Subject: checkpoint: intel_renderbuffer removal (intel_renderbuffer now == gl_renderbuffer --- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 4 ++-- src/mesa/drivers/dri/intel_winsys/intel_fbo.c | 20 ++++++++++---------- src/mesa/drivers/dri/intel_winsys/intel_fbo.h | 5 +++-- 3 files changed, 15 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index 33e4b16a5c..33fa4b9b40 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -102,7 +102,7 @@ intel_drawbuf_region(struct intel_context *intel) struct intel_renderbuffer *irbColor = intel_renderbuffer(intel->ctx.DrawBuffer->_ColorDrawBuffers[0][0]); if (irbColor) - return irbColor->region; + return irbColor->Base.surface->region; else return NULL; } @@ -116,7 +116,7 @@ intel_readbuf_region(struct intel_context *intel) struct intel_renderbuffer *irb = intel_renderbuffer(intel->ctx.ReadBuffer->_ColorReadBuffer); if (irb) - return irb->region; + return irb->Base.surface->region; else return NULL; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c index 7b9bef73bd..c80fd69fdb 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c @@ -117,7 +117,7 @@ intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex); if (irb) - return irb->region; + return irb->Base.surface->region; else return NULL; } @@ -153,8 +153,8 @@ intel_delete_renderbuffer(struct gl_renderbuffer *rb) intel_unpair_depth_stencil(ctx, irb); } #endif - if (intel && irb->region) { - intel->pipe->region_release(intel->pipe, &irb->region); + if (intel && irb->Base.surface->region) { + intel->pipe->region_release(intel->pipe, &irb->Base.surface->region); } _mesa_free(irb); @@ -265,8 +265,8 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, intelFlush(ctx); /* free old region */ - if (irb->region) { - intel->pipe->region_release(intel->pipe, &irb->region); + if (irb->Base.surface->region) { + intel->pipe->region_release(intel->pipe, &irb->Base.surface->region); } /* allocate new memory region/renderbuffer */ @@ -283,11 +283,11 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, height, pitch); - irb->region = intel->pipe->region_alloc(intel->pipe, cpp, pitch, height); - if (!irb->region) + irb->Base.surface->region = intel->pipe->region_alloc(intel->pipe, cpp, pitch, height); + if (!irb->Base.surface->region) return GL_FALSE; /* out of memory? */ - ASSERT(irb->region->buffer); + ASSERT(irb->Base.surface->region->buffer); rb->Width = width; rb->Height = height; @@ -295,7 +295,7 @@ intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, /* update the surface's size too */ rb->surface->width = width; rb->surface->height = height; - rb->surface->region = irb->region; + rb->surface->region = irb->Base.surface->region;/*XXX no-op*/ /* This sets the Get/PutRow/Value functions */ // intel_set_span_functions(&irb->Base); @@ -605,7 +605,7 @@ intel_finish_render_texture(GLcontext * ctx, if (irb) { /* just release the region */ - intel->pipe->region_release(intel->pipe, &irb->region); + intel->pipe->region_release(intel->pipe, &irb->Base.surface->region); } else if (att->Renderbuffer) { /* software fallback */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h index 63afa6df89..63d8e839c5 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h @@ -78,8 +78,9 @@ struct intel_framebuffer struct intel_renderbuffer { struct gl_renderbuffer Base; - struct pipe_region *region; #if 0 + struct pipe_region *region; + void *pfMap; /* possibly paged flipped map pointer */ GLuint pfPitch; /* possibly paged flipped pitch */ GLboolean RenderToTexture; /* RTT? */ @@ -90,9 +91,9 @@ struct intel_renderbuffer GLuint pf_pending; /**< sequence number of pending flip */ GLuint vbl_pending; /**< vblank sequence number of pending flip */ -#endif struct intel_surface *surface; +#endif }; #if 0 -- cgit v1.2.3 From 334d3650a95786a7f4e8212eb84fe9cf70206d8f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 11:11:11 -0600 Subject: checkpoint: more intel_renderbuffer removal --- src/mesa/drivers/dri/intel_winsys/intel_blit.c | 17 +++++++++++++++++ src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 6 ++++++ src/mesa/drivers/dri/intel_winsys/intel_fbo.c | 5 ++++- 3 files changed, 27 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_blit.c b/src/mesa/drivers/dri/intel_winsys/intel_blit.c index 8e878f0088..7e30c7d034 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_blit.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_blit.c @@ -86,20 +86,37 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, if (dPriv && dPriv->numClipRects) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; +#if 0 const struct pipe_region *backRegion = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); +#endif const int backWidth = intel_fb->Base.Width; const int backHeight = intel_fb->Base.Height; const int nbox = dPriv->numClipRects; const drm_clip_rect_t *pbox = dPriv->pClipRects; const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; +#if 0 const int srcpitch = backRegion->pitch; +#endif const int cpp = intelScreen->front.cpp; int BR13, CMD; int i; + const struct pipe_surface *backSurf; + const struct pipe_region *backRegion; + int srcpitch; + + /* blit from back color buffer if it exists, else front buffer */ + if (intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer) + backSurf = intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer->surface; + else + backSurf = intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer->surface; + + backRegion = backSurf->region; + srcpitch = backRegion->pitch; + ASSERT(intel_fb); ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ ASSERT(backRegion); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index 33fa4b9b40..c4db446788 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -217,8 +217,10 @@ intelWindowMoved(struct intel_context *intel) } intel_fb->pf_active = pf_active; +#if 0 intel_flip_renderbuffers(intel_fb); intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); +#endif /* Update vblank info */ @@ -359,8 +361,10 @@ intelPageFlip(const __DRIdrawablePrivate * dPriv) ++intel_fb->pf_seq; } +#if 0 intel_flip_renderbuffers(intel_fb); intel_draw_buffer(&intel->ctx, &intel_fb->Base); +#endif if (INTEL_DEBUG & DEBUG_IOCTL) fprintf(stderr, "%s: success\n", __FUNCTION__); @@ -430,8 +434,10 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) #endif if (swap.seqtype & DRM_VBLANK_FLIP) { +#if 0 intel_flip_renderbuffers(intel_fb); intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); +#endif } ret = GL_TRUE; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c index c80fd69fdb..ff786de1e3 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c @@ -74,7 +74,7 @@ intel_get_renderbuffer(struct gl_framebuffer *fb, GLuint attIndex) return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer); } - +#if 0 void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) { @@ -109,6 +109,7 @@ intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) _mesa_reference_renderbuffer(&tmp_rb, NULL); } } +#endif struct pipe_region * @@ -321,6 +322,7 @@ intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, } /* Make sure all window system renderbuffers are up to date */ +#if 0 for (i = 0; i < 3; i++) { struct gl_renderbuffer *rb = &intel_fb->color_rb[i]->Base; @@ -329,6 +331,7 @@ intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height); } } +#endif } static GLboolean -- cgit v1.2.3 From 0481f22fdbed715dea9e36dad871606d5fd285bb Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 11:16:13 -0600 Subject: disable intel_get_rb_region() --- src/mesa/drivers/dri/intel_winsys/intel_fbo.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c index ff786de1e3..5382ab342c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c @@ -112,6 +112,7 @@ intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) #endif +#if 0 struct pipe_region * intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) { @@ -122,7 +123,7 @@ intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) else return NULL; } - +#endif /** @@ -331,6 +332,8 @@ intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height); } } +#else + #endif } -- cgit v1.2.3 From b6e4efb451061a4e124f6bf575e528cbbff551c2 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 11:30:40 -0600 Subject: remove color_rb[] and lots of dead code --- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 7 +-- src/mesa/drivers/dri/intel_winsys/intel_context.c | 5 -- src/mesa/drivers/dri/intel_winsys/intel_fbo.c | 58 ----------------------- src/mesa/drivers/dri/intel_winsys/intel_fbo.h | 41 ---------------- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 15 ++---- 5 files changed, 6 insertions(+), 120 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index c4db446788..9266fb62af 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -242,9 +242,7 @@ intelWindowMoved(struct intel_context *intel) } for (i = 0; i < intel_fb->pf_num_pages; i++) { - if (!intel_fb->color_rb[i] || - (intel_fb->vbl_waited - intel_fb->vbl_pending[i]) <= - (1<<23)) + if ((intel_fb->vbl_waited - intel_fb->vbl_pending[i]) <= (1<<23)) continue; vbl.request.sequence = intel_fb->vbl_pending[i]; @@ -256,8 +254,7 @@ intelWindowMoved(struct intel_context *intel) intel_fb->vbl_waited = intel_fb->vbl_seq; for (i = 0; i < intel_fb->pf_num_pages; i++) { - if (intel_fb->color_rb[i]) - intel_fb->vbl_pending[i] = intel_fb->vbl_waited; + intel_fb->vbl_pending[i] = intel_fb->vbl_waited; } } } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 5c184d5c8d..9defcc1aae 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -574,12 +574,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, intel_fb->vbl_waited = intel_fb->vbl_seq; for (i = 0; i < 2; i++) { -#if 1 intel_fb->vbl_pending[i] = intel_fb->vbl_seq; -#else - if (intel_fb->color_rb[i]) - intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_seq; -#endif } } } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c index 5382ab342c..3dd99747da 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c @@ -68,64 +68,6 @@ intel_renderbuffer(struct gl_renderbuffer *rb) } -struct intel_renderbuffer * -intel_get_renderbuffer(struct gl_framebuffer *fb, GLuint attIndex) -{ - return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer); -} - -#if 0 -void -intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) -{ - int current_page = intel_fb->pf_current_page; - int next_page = (current_page + 1) % intel_fb->pf_num_pages; - struct gl_renderbuffer *tmp_rb; - - /* Exchange renderbuffers if necessary but make sure their reference counts - * are preserved. - */ - if (intel_fb->color_rb[current_page] && - intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer != - &intel_fb->color_rb[current_page]->Base) { - tmp_rb = NULL; - _mesa_reference_renderbuffer(&tmp_rb, - intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); - tmp_rb = &intel_fb->color_rb[current_page]->Base; - _mesa_reference_renderbuffer( - &intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer, tmp_rb); - _mesa_reference_renderbuffer(&tmp_rb, NULL); - } - - if (intel_fb->color_rb[next_page] && - intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer != - &intel_fb->color_rb[next_page]->Base) { - tmp_rb = NULL; - _mesa_reference_renderbuffer(&tmp_rb, - intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); - tmp_rb = &intel_fb->color_rb[next_page]->Base; - _mesa_reference_renderbuffer( - &intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer, tmp_rb); - _mesa_reference_renderbuffer(&tmp_rb, NULL); - } -} -#endif - - -#if 0 -struct pipe_region * -intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) -{ - struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex); - - if (irb) - return irb->Base.surface->region; - else - return NULL; -} -#endif - - /** * Create a new framebuffer object. */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h index 63d8e839c5..578e1013c2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h @@ -44,8 +44,6 @@ struct intel_framebuffer { struct gl_framebuffer Base; - struct intel_renderbuffer *color_rb[3]; - /* Drawable page flipping state */ GLboolean pf_active; GLuint pf_seq; @@ -78,53 +76,14 @@ struct intel_framebuffer struct intel_renderbuffer { struct gl_renderbuffer Base; -#if 0 - struct pipe_region *region; - - void *pfMap; /* possibly paged flipped map pointer */ - GLuint pfPitch; /* possibly paged flipped pitch */ - GLboolean RenderToTexture; /* RTT? */ - - GLuint PairedDepth; /**< only used if this is a depth renderbuffer */ - GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */ - - GLuint pf_pending; /**< sequence number of pending flip */ - - GLuint vbl_pending; /**< vblank sequence number of pending flip */ - - struct intel_surface *surface; -#endif }; -#if 0 -extern struct intel_renderbuffer *intel_create_renderbuffer(GLenum intFormat, - GLsizei width, - GLsizei height, - int offset, - int pitch, - int cpp, - void *map); -#endif extern struct intel_renderbuffer *intel_new_renderbuffer_fb(GLuint intFormat); extern void intel_fbo_init(struct intel_context *intel); -/* XXX make inline or macro */ -extern struct intel_renderbuffer *intel_get_renderbuffer(struct gl_framebuffer - *fb, - GLuint attIndex); - -extern void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb); - - -/* XXX make inline or macro */ -extern struct pipe_region *intel_get_rb_region(struct gl_framebuffer *fb, - GLuint attIndex); - - - extern struct pipe_surface * intel_new_surface(GLuint intFormat); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index fba0c47e59..74cd714cf3 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -302,24 +302,17 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, /* fake frontbuffer */ /* XXX allocation should only happen in the unusual case it's actually needed */ - intel_fb->color_rb[0] + struct intel_renderbuffer *irb = intel_new_renderbuffer_fb(rgbFormat); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, - &intel_fb->color_rb[0]->Base); + &irb->Base); } if (mesaVis->doubleBufferMode) { -#if 01 - intel_fb->color_rb[1] + struct intel_renderbuffer *irb = intel_new_renderbuffer_fb(rgbFormat); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, - &intel_fb->color_rb[1]->Base); -#else - intel_fb->color_rb[1] - = st_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, - &intel_fb->color_rb[1]->Base); -#endif + &irb->Base); } if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { -- cgit v1.2.3 From 64da7515009f3551796c90acc74eb0a2ffdb68a0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 12:27:44 -0600 Subject: checkpoint: no longer using intel_fbo.c --- src/mesa/drivers/dri/intel_winsys/Makefile | 4 +- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 3 +- src/mesa/drivers/dri/intel_winsys/intel_context.c | 54 ++++++++++++ src/mesa/drivers/dri/intel_winsys/intel_fbo.h | 11 ++- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 28 +++--- src/mesa/drivers/dri/intel_winsys/intel_surface.c | 82 +++++++++++++---- src/mesa/state_tracker/st_cb_fbo.c | 102 ++++++++++++++++++---- src/mesa/state_tracker/st_cb_fbo.h | 8 ++ src/mesa/state_tracker/st_context.c | 2 + 9 files changed, 241 insertions(+), 53 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index 8d3d17e13e..32c2c3bb24 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -22,10 +22,10 @@ DRIVER_SOURCES = \ intel_ioctl.c \ intel_screen.c \ intel_surface.c \ - intel_fbo.c \ intel_batchpool.c -UNUSED = intel_depthstencil.c +UNUSED = intel_depthstencil.c \ + intel_fbo.c C_SOURCES = \ $(COMMON_SOURCES) \ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index 9266fb62af..d0139d5b2a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -96,6 +96,7 @@ intel_intersect_cliprects(drm_clip_rect_t * dst, /** * Return pointer to current color drawing region, or NULL. */ +#if 0 struct pipe_region * intel_drawbuf_region(struct intel_context *intel) { @@ -120,7 +121,7 @@ intel_readbuf_region(struct intel_context *intel) else return NULL; } - +#endif /** * This will be called whenever the currently bound window is moved/resized. diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 9defcc1aae..0d83d4efd6 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -360,6 +360,58 @@ intelCreateContext(const __GLcontextModes * mesaVis, _vbo_CreateContext(ctx); _tnl_CreateContext(ctx); + /* + * Pipe-related setup + */ + if (!getenv("INTEL_HW")) { + intel->pipe = intel_create_softpipe( intel ); + intel->pipe->surface_alloc = intel_new_surface; + intel->pipe->supported_formats = intel_supported_formats; + } + else { + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + intel->pipe = intel_create_i915simple( intel ); + break; + default: + _mesa_printf("Unknown PCIID %x in %s, using software driver\n", + intel->intelScreen->deviceID, __FUNCTION__); + + intel->pipe = intel_create_softpipe( intel ); + break; + } + } + + st_create_context( &intel->ctx, intel->pipe ); + + + /* TODO: Push this down into the pipe driver: + */ + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + intel->pipe->mipmap_tree_layout = i945_miptree_layout; + break; + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + case PCI_CHIP_I830_M: + case PCI_CHIP_I855_GM: + case PCI_CHIP_I865_G: + intel->pipe->mipmap_tree_layout = i915_miptree_layout; + default: + assert(0); /*FIX*/ + } /* @@ -397,7 +449,9 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->last_swap_fence = NULL; intel->first_swap_fence = NULL; +#if 00 intel_fbo_init(intel); +#endif if (intel->ctx.Mesa_DXTn) { _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h index 578e1013c2..2801c7345d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h @@ -73,19 +73,26 @@ struct intel_framebuffer * not pointers because in some circumstances a deleted renderbuffer could * result in a dangling pointer here. */ +#if 0 struct intel_renderbuffer { struct gl_renderbuffer Base; }; +#endif - +#if 0 extern struct intel_renderbuffer *intel_new_renderbuffer_fb(GLuint intFormat); +#endif extern void intel_fbo_init(struct intel_context *intel); extern struct pipe_surface * -intel_new_surface(GLuint intFormat); +intel_new_surface(struct pipe_context *pipe, GLuint pipeFormat); + + +extern const GLuint * +intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats); #endif /* INTEL_FBO_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 74cd714cf3..e5f7b36527 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -302,34 +302,32 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, /* fake frontbuffer */ /* XXX allocation should only happen in the unusual case it's actually needed */ - struct intel_renderbuffer *irb - = intel_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, - &irb->Base); + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, rb); } if (mesaVis->doubleBufferMode) { - struct intel_renderbuffer *irb - = intel_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, - &irb->Base); + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, rb); } if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { /* combined depth/stencil buffer */ - struct intel_renderbuffer *depthStencilRb - = intel_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + struct gl_renderbuffer *depthStencilRb + = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); /* note: bind RB to two attachment points */ _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, - &depthStencilRb->Base); + depthStencilRb); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL, - &depthStencilRb->Base); + depthStencilRb); } else if (mesaVis->depthBits == 16) { /* just 16-bit depth buffer, no hw stencil */ - struct intel_renderbuffer *depthRb - = intel_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); + struct gl_renderbuffer *depthRb + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, depthRb); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.c b/src/mesa/drivers/dri/intel_winsys/intel_surface.c index d2e2dabade..20b6bda83f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.c @@ -159,8 +159,43 @@ write_quad_stencil(struct softpipe_surface *sps, } + + +static void +a8r8g8b8_get_tile(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) +{ + const GLuint *src + = ((const GLuint *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + GLuint i, j; +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + p[0] = UBYTE_TO_FLOAT((src[j] >> 16) & 0xff); + p[1] = UBYTE_TO_FLOAT((src[j] >> 8) & 0xff); + p[2] = UBYTE_TO_FLOAT((src[j] >> 0) & 0xff); + p[3] = UBYTE_TO_FLOAT((src[j] >> 24) & 0xff); + p += 4; + } + src += ps->region->pitch; + } +} + + + + struct pipe_surface * -intel_new_surface(GLuint intFormat) +intel_new_surface(struct pipe_context *pipe, GLuint pipeFormat) { struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); if (!sps) @@ -168,31 +203,42 @@ intel_new_surface(GLuint intFormat) sps->surface.width = 0; /* set in intel_alloc_renderbuffer_storage() */ sps->surface.height = 0; + sps->surface.refcount = 1; + sps->surface.format = pipeFormat; - if (intFormat == GL_RGBA8) { - sps->surface.format = PIPE_FORMAT_U_A8_R8_G8_B8; + switch (pipeFormat) { + case PIPE_FORMAT_U_A8_R8_G8_B8: sps->read_quad_f_swz = read_quad_f_swz; sps->write_quad_f_swz = write_quad_f_swz; - } - else if (intFormat == GL_RGB5) { - sps->surface.format = PIPE_FORMAT_U_R5_G6_B5; - - } - else if (intFormat == GL_DEPTH_COMPONENT16) { - sps->surface.format = PIPE_FORMAT_U_Z16; - - } - else if (intFormat == GL_DEPTH24_STENCIL8_EXT) { - sps->surface.format = PIPE_FORMAT_S8_Z24; + sps->surface.get_tile = a8r8g8b8_get_tile; + break; + case PIPE_FORMAT_U_R5_G6_B5: + break; + case PIPE_FORMAT_U_Z16: + break; + case PIPE_FORMAT_S8_Z24: sps->read_quad_z = read_quad_z24; sps->write_quad_z = write_quad_z24; sps->read_quad_stencil = read_quad_stencil; sps->write_quad_stencil = write_quad_stencil; - } - else { - /* TBD / unknown */ - + break; } return &sps->surface; } + + + +const GLuint * +intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats) +{ + static const GLuint formats[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_S8_Z24, + }; + + *numFormats = sizeof(formats) / sizeof(formats[0]); + return formats; +} + diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index d0205fd635..3cd1fbe851 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -53,7 +53,9 @@ struct st_renderbuffer { struct gl_renderbuffer Base; +#if 0 struct pipe_surface *surface; +#endif }; @@ -93,14 +95,14 @@ pipe_get_format_info(GLuint format) { PIPE_FORMAT_U_R8_G8_B8_A8, /* format */ GL_RGBA, /* base_format */ - 4, 4, 4, 4, 0, 0, /* color bits */ + 8, 8, 8, 8, 0, 0, /* color bits */ 0, 0, /* depth, stencil */ 4 /* size in bytes */ }, { PIPE_FORMAT_U_A8_R8_G8_B8, GL_RGBA, /* base_format */ - 4, 4, 4, 4, 0, 0, /* color bits */ + 8, 8, 8, 8, 0, 0, /* color bits */ 0, 0, /* depth, stencil */ 4 /* size in bytes */ }, @@ -166,29 +168,29 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, cpp = info->size; - if (!strb->surface) { - strb->surface = pipe->surface_alloc(pipe, pipeFormat); - if (!strb->surface) + if (!strb->Base.surface) { + strb->Base.surface = pipe->surface_alloc(pipe, pipeFormat); + if (!strb->Base.surface) return GL_FALSE; } /* free old region */ - if (strb->surface->region) { - pipe->region_release(pipe, &strb->surface->region); + if (strb->Base.surface->region) { + pipe->region_release(pipe, &strb->Base.surface->region); } /* Choose a pitch to match hardware requirements: */ pitch = ((cpp * width + 63) & ~63) / cpp; /* XXX fix: device-specific */ - strb->surface->region = pipe->region_alloc(pipe, cpp, pitch, height); - if (!strb->surface->region) + strb->Base.surface->region = pipe->region_alloc(pipe, cpp, pitch, height); + if (!strb->Base.surface->region) return GL_FALSE; /* out of memory, try s/w buffer? */ - ASSERT(strb->surface->region->buffer); + ASSERT(strb->Base.surface->region->buffer); - strb->Base.Width = strb->surface->width = width; - strb->Base.Height = strb->surface->height = height; + strb->Base.Width = strb->Base.surface->width = width; + strb->Base.Height = strb->Base.surface->height = height; return GL_TRUE; } @@ -204,11 +206,11 @@ st_renderbuffer_delete(struct gl_renderbuffer *rb) struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); ASSERT(strb); - if (strb && strb->surface) { + if (strb && strb->Base.surface) { if (rb->surface->region) { - pipe->region_release(pipe, &strb->surface->region); + pipe->region_release(pipe, &strb->Base.surface->region); } - free(strb->surface); + free(strb->Base.surface); } free(strb); } @@ -257,6 +259,76 @@ st_new_renderbuffer(GLcontext *ctx, GLuint name) return NULL; } + +#if 000 +struct gl_renderbuffer * +st_new_renderbuffer_fb(struct pipe_region *region, GLuint width, GLuint height) +{ + struct st_renderbuffer *strb = CALLOC_STRUCT(st_renderbuffer); + if (!strb) + return; + + _mesa_init_renderbuffer(&strb->Base, name); + strb->Base.Delete = st_renderbuffer_delete; + strb->Base.AllocStorage = st_renderbuffer_alloc_storage; + strb->Base.GetPointer = null_get_pointer; + strb->Base.Width = width; + strb->Base.Heigth = height; + + strb->region = region; + + return &strb->Base; +} + +#else + +struct gl_renderbuffer * +st_new_renderbuffer_fb(GLuint intFormat) +{ + struct st_renderbuffer *irb; + + irb = CALLOC_STRUCT(st_renderbuffer); + if (!irb) { + _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer"); + return NULL; + } + + _mesa_init_renderbuffer(&irb->Base, 0); + irb->Base.ClassID = 0x42; /* XXX temp */ + irb->Base.InternalFormat = intFormat; + + switch (intFormat) { + case GL_RGB5: + case GL_RGBA8: + irb->Base._BaseFormat = GL_RGBA; + break; + case GL_DEPTH_COMPONENT16: + irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + break; + case GL_DEPTH24_STENCIL8_EXT: + irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; + break; + default: + _mesa_problem(NULL, + "Unexpected intFormat in st_new_renderbuffer"); + return NULL; + } + + /* st-specific methods */ + irb->Base.Delete = st_renderbuffer_delete; + irb->Base.AllocStorage = st_renderbuffer_alloc_storage; + irb->Base.GetPointer = null_get_pointer; + /* span routines set in alloc_storage function */ + + irb->Base.surface = NULL;/*intel_new_surface(intFormat);*/ + /*irb->Base.surface->rb = irb;*/ + + return &irb->Base; +} +#endif + + + /** * Called via ctx->Driver.BindFramebufferEXT(). */ diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index 6142434ec6..7f52ab10d7 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -30,6 +30,14 @@ #define ST_CB_FBO_H +/* +extern struct gl_renderbuffer * +st_new_renderbuffer_fb(struct pipe_region *region, GLuint width, GLuint height); +*/ +extern struct gl_renderbuffer * +st_new_renderbuffer_fb(GLuint intFormat); + + extern void st_init_fbo_functions(struct dd_function_table *functions); diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 0ea06c692d..9f7916e40e 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -31,6 +31,7 @@ #include "st_cb_bufferobjects.h" #include "st_cb_clear.h" #include "st_cb_drawpixels.h" +#include "st_cb_fbo.h" #include "st_cb_texture.h" #include "st_atom.h" #include "st_draw.h" @@ -102,6 +103,7 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_bufferobject_functions(functions); st_init_clear_functions(functions); st_init_drawpixels_functions(functions); + st_init_fbo_functions(functions); st_init_program_functions(functions); st_init_texture_functions(functions); } -- cgit v1.2.3 From 01ee12cc5faa447d5542d00970e8bc3867bd5c60 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 12:36:19 -0600 Subject: surface-related prototypes --- src/mesa/drivers/dri/intel_winsys/intel_surface.h | 14 ++++++++++++++ 1 file changed, 14 insertions(+) create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_surface.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.h b/src/mesa/drivers/dri/intel_winsys/intel_surface.h new file mode 100644 index 0000000000..90a1841efb --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.h @@ -0,0 +1,14 @@ + +#ifndef INTEL_SURFACE_H +#define INTEL_SURFACE_H + + +extern struct pipe_surface * +intel_new_surface(struct pipe_context *pipe, GLuint pipeFormat); + + +extern const GLuint * +intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats); + + +#endif /* INTEL_SURFACE_H */ -- cgit v1.2.3 From 36fc648ddd937962e659d6215ed6cfb864fb9f05 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 12:36:47 -0600 Subject: remove dead code, remove intel_fbo.h includes --- src/mesa/drivers/dri/intel_winsys/intel_blit.c | 1 - src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 8 +++-- src/mesa/drivers/dri/intel_winsys/intel_buffers.h | 37 ++++++++++++++++++----- src/mesa/drivers/dri/intel_winsys/intel_context.c | 4 +-- src/mesa/drivers/dri/intel_winsys/intel_fbo.h | 3 +- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 2 -- src/mesa/drivers/dri/intel_winsys/intel_surface.c | 2 +- 7 files changed, 39 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_blit.c b/src/mesa/drivers/dri/intel_winsys/intel_blit.c index 7e30c7d034..48bbbbeac9 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_blit.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_blit.c @@ -34,7 +34,6 @@ #include "intel_blit.h" #include "intel_buffers.h" #include "intel_context.h" -#include "intel_fbo.h" #include "intel_reg.h" #include "vblank.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index d0139d5b2a..cbf6ea38b3 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -30,7 +30,6 @@ #include "intel_blit.h" #include "intel_buffers.h" #include "intel_depthstencil.h" -#include "intel_fbo.h" #include "intel_batchbuffer.h" #include "intel_reg.h" #include "context.h" @@ -93,6 +92,7 @@ intel_intersect_cliprects(drm_clip_rect_t * dst, return GL_TRUE; } +<<<<<<< HEAD:src/mesa/drivers/dri/intel_winsys/intel_buffers.c /** * Return pointer to current color drawing region, or NULL. */ @@ -107,7 +107,10 @@ intel_drawbuf_region(struct intel_context *intel) else return NULL; } +======= +>>>>>>> remove dead code, remove intel_fbo.h includes:src/mesa/drivers/dri/intel_winsys/intel_buffers.c +<<<<<<< HEAD:src/mesa/drivers/dri/intel_winsys/intel_buffers.c /** * Return pointer to current color reading region, or NULL. */ @@ -122,6 +125,8 @@ intel_readbuf_region(struct intel_context *intel) return NULL; } #endif +======= +>>>>>>> remove dead code, remove intel_fbo.h includes:src/mesa/drivers/dri/intel_winsys/intel_buffers.c /** * This will be called whenever the currently bound window is moved/resized. @@ -520,4 +525,3 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); } } - diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.h b/src/mesa/drivers/dri/intel_winsys/intel_buffers.h index 5834e39501..3e1597e3c4 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.h @@ -32,24 +32,45 @@ struct intel_context; struct intel_framebuffer; +/** + * Intel framebuffer, derived from gl_framebuffer. + */ +struct intel_framebuffer +{ + struct gl_framebuffer Base; + + /* Drawable page flipping state */ + GLboolean pf_active; + GLuint pf_seq; + GLint pf_pipes; + GLint pf_current_page; + GLint pf_num_pages; + + /* VBI + */ + GLuint vbl_seq; + GLuint vblank_flags; + GLuint vbl_waited; + + int64_t swap_ust; + int64_t swap_missed_ust; + + GLuint swap_count; + GLuint swap_missed_count; + + GLuint vbl_pending[3]; /**< [number of color buffers] */ +}; + extern GLboolean intel_intersect_cliprects(drm_clip_rect_t * dest, const drm_clip_rect_t * a, const drm_clip_rect_t * b); -extern struct pipe_region *intel_readbuf_region(struct intel_context *intel); - -extern struct pipe_region *intel_drawbuf_region(struct intel_context *intel); - extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); extern void intelWindowMoved(struct intel_context *intel); -extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); - -extern void intelInitBufferFuncs(struct dd_function_table *functions); - #endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 0d83d4efd6..5aba4ef584 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -51,8 +51,8 @@ #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" -#include "intel_fbo.h" #include "intel_tex_layout.h" +#include "intel_surface.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" @@ -311,8 +311,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) functions->GetString = intelGetString; functions->UpdateState = intelInvalidateState; - intelInitBufferFuncs(functions); - st_init_driver_functions(functions); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h index 2801c7345d..0487c27600 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h @@ -36,7 +36,7 @@ struct intel_context; struct pipe_region; - +#if 0 /** * Intel framebuffer, derived from gl_framebuffer. */ @@ -65,6 +65,7 @@ struct intel_framebuffer GLuint vbl_pending[3]; /**< [number of color buffers] */ }; +#endif /** diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index e5f7b36527..b765c2b4b2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -38,9 +38,7 @@ #include "intel_screen.h" #include "intel_batchbuffer.h" #include "intel_buffers.h" -/*#include "intel_tex.h"*/ #include "intel_ioctl.h" -#include "intel_fbo.h" #include "i830_dri.h" #include "dri_bufpool.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.c b/src/mesa/drivers/dri/intel_winsys/intel_surface.c index 20b6bda83f..13c34e6c8e 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.c @@ -10,7 +10,7 @@ #include "intel_context.h" #include "intel_buffers.h" -#include "intel_fbo.h" +#include "intel_surface.h" #include "pipe/p_state.h" #include "pipe/p_context.h" -- cgit v1.2.3 From f5713c7d2e7ba8e1170fd9b1dd95379662ab6117 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 12:59:11 -0600 Subject: Checkpoint intel_renderbuffer removal. Remove surface ptr from gl_renderbuffer. Use st_renderbuffer in most places. More clean-up. --- src/mesa/drivers/dri/intel_winsys/intel_blit.c | 14 +++-- src/mesa/main/mtypes.h | 2 - src/mesa/main/renderbuffer.c | 14 ++++- src/mesa/pipe/p_state.h | 2 - src/mesa/state_tracker/st_atom_framebuffer.c | 27 +++++---- src/mesa/state_tracker/st_cb_clear.c | 35 +++++++---- src/mesa/state_tracker/st_cb_fbo.c | 82 +++++++++----------------- src/mesa/state_tracker/st_cb_fbo.h | 23 ++++++-- 8 files changed, 107 insertions(+), 92 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_blit.c b/src/mesa/drivers/dri/intel_winsys/intel_blit.c index 48bbbbeac9..aa4135ed2d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_blit.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_blit.c @@ -38,6 +38,7 @@ #include "vblank.h" #include "pipe/p_context.h" +#include "state_tracker/st_cb_fbo.h" #define FILE_DEBUG_FLAG DEBUG_BLIT @@ -106,12 +107,17 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, const struct pipe_surface *backSurf; const struct pipe_region *backRegion; int srcpitch; + struct st_renderbuffer *strb; /* blit from back color buffer if it exists, else front buffer */ - if (intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer) - backSurf = intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer->surface; - else - backSurf = intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer->surface; + strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); + if (strb) { + backSurf = strb->surface; + } + else { + strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + backSurf = strb->surface; + } backRegion = backSurf->region; srcpitch = backRegion->pitch; diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index d70df5d945..0a64e0c58c 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2269,8 +2269,6 @@ struct gl_renderbuffer GLubyte StencilBits; GLvoid *Data; /**< This may not be used by some kinds of RBs */ - struct pipe_surface *surface; - /* Used to wrap one renderbuffer around another: */ struct gl_renderbuffer *Wrapped; diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c index d89704196a..e7aeea8e3e 100644 --- a/src/mesa/main/renderbuffer.c +++ b/src/mesa/main/renderbuffer.c @@ -1067,9 +1067,11 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutMonoValues = put_mono_values_ubyte; rb->StencilBits = 8 * sizeof(GLubyte); pixelSize = sizeof(GLubyte); +#if 0 if (!rb->surface) rb->surface = (struct pipe_surface *) pipe->surface_alloc(pipe, PIPE_FORMAT_U_S8); +#endif break; case GL_STENCIL_INDEX16_EXT: rb->_ActualFormat = GL_STENCIL_INDEX16_EXT; @@ -1100,9 +1102,11 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutValues = put_values_ushort; rb->PutMonoValues = put_mono_values_ushort; rb->DepthBits = 8 * sizeof(GLushort); +#if 0 if (!rb->surface) rb->surface = (struct pipe_surface *) pipe->surface_alloc(pipe, PIPE_FORMAT_U_Z16); +#endif pixelSize = sizeof(GLushort); break; case GL_DEPTH_COMPONENT24: @@ -1125,9 +1129,11 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->_ActualFormat = GL_DEPTH_COMPONENT32; rb->DepthBits = 32; } +#if 0 if (!rb->surface) rb->surface = (struct pipe_surface *) pipe->surface_alloc(pipe, PIPE_FORMAT_U_Z32); +#endif pixelSize = sizeof(GLuint); break; case GL_DEPTH_STENCIL_EXT: @@ -1145,9 +1151,11 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutMonoValues = put_mono_values_uint; rb->DepthBits = 24; rb->StencilBits = 8; +#if 0 if (!rb->surface) rb->surface = (struct pipe_surface *) pipe->surface_alloc(pipe, PIPE_FORMAT_S8_Z24); +#endif pixelSize = sizeof(GLuint); break; case GL_COLOR_INDEX8_EXT: @@ -1210,7 +1218,7 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, ASSERT(rb->PutMonoValues); /* free old buffer storage */ - if (rb->surface) { + if (0/**rb->surface**/) { /* pipe_surface/region */ } else if (rb->Data) { @@ -1221,8 +1229,9 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, if (width > 0 && height > 0) { /* allocate new buffer storage */ - if (rb->surface) { + if (0/**rb->surface**/) { /* pipe_surface/region */ +#if 0 if (rb->surface->region) { pipe->region_unmap(pipe, rb->surface->region); pipe->region_release(pipe, &rb->surface->region); @@ -1231,6 +1240,7 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, /* XXX probably don't want to really map here */ pipe->region_map(pipe, rb->surface->region); rb->Data = rb->surface->region->map; +#endif } else { /* legacy renderbuffer (this will go away) */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 2dcd2db868..ee29e38a48 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -260,8 +260,6 @@ struct pipe_surface GLuint offset; /**< offset from start of region, in bytes */ GLint refcount; - void *rb; /**< Ptr back to renderbuffer (temporary?) */ - /** get block/tile of pixels from surface */ void (*get_tile)(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p); diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 7edd044ad9..f054eb8f21 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -33,6 +33,7 @@ #include "st_context.h" #include "st_atom.h" +#include "st_cb_fbo.h" #include "pipe/p_context.h" @@ -44,7 +45,7 @@ static void update_framebuffer_state( struct st_context *st ) { struct pipe_framebuffer_state framebuffer; - struct gl_renderbuffer *rb; + struct st_renderbuffer *strb; GLuint i; memset(&framebuffer, 0, sizeof(framebuffer)); @@ -54,21 +55,23 @@ update_framebuffer_state( struct st_context *st ) */ framebuffer.num_cbufs = st->ctx->DrawBuffer->_NumColorDrawBuffers[0]; for (i = 0; i < framebuffer.num_cbufs; i++) { - rb = st->ctx->DrawBuffer->_ColorDrawBuffers[0][i]; - assert(rb->surface); - framebuffer.cbufs[i] = rb->surface; + strb = st_renderbuffer(st->ctx->DrawBuffer->_ColorDrawBuffers[0][i]); + assert(strb->surface); + framebuffer.cbufs[i] = strb->surface; } - rb = st->ctx->DrawBuffer->_DepthBuffer; - if (rb) { - assert(rb->Wrapped->surface); - framebuffer.zbuf = rb->Wrapped->surface; + strb = st_renderbuffer(st->ctx->DrawBuffer->_DepthBuffer); + if (strb) { + strb = st_renderbuffer(strb->Base.Wrapped); + assert(strb->surface); + framebuffer.zbuf = strb->surface; } - rb = st->ctx->DrawBuffer->_StencilBuffer; - if (rb) { - assert(rb->Wrapped->surface); - framebuffer.sbuf = rb->Wrapped->surface; + strb = st_renderbuffer(st->ctx->DrawBuffer->_StencilBuffer); + if (strb) { + strb = st_renderbuffer(strb->Base.Wrapped); + assert(strb->surface); + framebuffer.sbuf = strb->surface; } if (memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0) { diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index d862f7ba46..4da3a2500d 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -37,6 +37,7 @@ #include "st_atom.h" #include "st_context.h" #include "st_cb_clear.h" +#include "st_cb_fbo.h" #include "st_program.h" #include "st_public.h" #include "pipe/p_context.h" @@ -288,6 +289,8 @@ clear_with_quad(GLcontext *ctx, GLuint x0, GLuint y0, static void clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { + struct st_renderbuffer *strb = st_renderbuffer(rb); + if (ctx->Color.ColorMask[0] && ctx->Color.ColorMask[1] && ctx->Color.ColorMask[2] && @@ -296,8 +299,8 @@ clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { /* clear whole buffer w/out masking */ GLuint clearValue - = color_value(rb->surface->format, ctx->Color.ClearColor); - ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + = color_value(strb->surface->format, ctx->Color.ClearColor); + ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } else { /* masking or scissoring */ @@ -314,14 +317,16 @@ clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) static void clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { + struct st_renderbuffer *strb = st_renderbuffer(rb); + if (!ctx->Scissor.Enabled) { /* clear whole buffer w/out masking */ GLuint clearValue - = color_value(rb->surface->format, ctx->Accum.ClearColor); + = color_value(strb->surface->format, ctx->Accum.ClearColor); /* Note that clearValue is 32 bits but the accum buffer will * typically be 64bpp... */ - ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } else { /* scissoring */ @@ -339,11 +344,13 @@ clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) static void clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { + struct st_renderbuffer *strb = st_renderbuffer(rb); + if (!ctx->Scissor.Enabled && - !is_depth_stencil_format(rb->surface->format)) { + !is_depth_stencil_format(strb->surface->format)) { /* clear whole depth buffer w/out masking */ - GLuint clearValue = depth_value(rb->surface->format, ctx->Depth.Clear); - ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); + ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } else { /* masking or scissoring or combined z/stencil buffer */ @@ -360,14 +367,15 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) static void clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { + struct st_renderbuffer *strb = st_renderbuffer(rb); const GLuint stencilMax = (1 << rb->StencilBits) - 1; GLboolean maskStencil = ctx->Stencil.WriteMask[0] != stencilMax; if (!maskStencil && !ctx->Scissor.Enabled && - !is_depth_stencil_format(rb->surface->format)) { + !is_depth_stencil_format(strb->surface->format)) { /* clear whole stencil buffer w/out masking */ GLuint clearValue = ctx->Stencil.Clear; - ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } else { /* masking or scissoring */ @@ -384,16 +392,17 @@ clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) static void clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { + struct st_renderbuffer *strb = st_renderbuffer(rb); const GLuint stencilMax = 1 << rb->StencilBits; GLboolean maskStencil = ctx->Stencil.WriteMask[0] != stencilMax; - assert(is_depth_stencil_format(rb->surface->format)); + assert(is_depth_stencil_format(strb->surface->format)); if (!maskStencil && !ctx->Scissor.Enabled) { /* clear whole buffer w/out masking */ - GLuint clearValue = depth_value(rb->surface->format, ctx->Depth.Clear); + GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); - switch (rb->surface->format) { + switch (strb->surface->format) { case PIPE_FORMAT_S8_Z24: clearValue |= ctx->Stencil.Clear << 24; break; @@ -406,7 +415,7 @@ clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) assert(0); } - ctx->st->pipe->clear(ctx->st->pipe, rb->surface, clearValue); + ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } else { /* masking or scissoring */ diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 3cd1fbe851..02bdb5aba5 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -46,29 +46,6 @@ #include "st_cb_teximage.h" -/** - * Derived renderbuffer class. Just need to add a pointer to the - * pipe surface. - */ -struct st_renderbuffer -{ - struct gl_renderbuffer Base; -#if 0 - struct pipe_surface *surface; -#endif -}; - - -/** - * Cast wrapper. - */ -static INLINE struct st_renderbuffer * -st_renderbuffer(struct gl_renderbuffer *rb) -{ - return (struct st_renderbuffer *) rb; -} - - struct pipe_format_info { GLuint format; @@ -168,29 +145,29 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, cpp = info->size; - if (!strb->Base.surface) { - strb->Base.surface = pipe->surface_alloc(pipe, pipeFormat); - if (!strb->Base.surface) + if (!strb->surface) { + strb->surface = pipe->surface_alloc(pipe, pipeFormat); + if (!strb->surface) return GL_FALSE; } /* free old region */ - if (strb->Base.surface->region) { - pipe->region_release(pipe, &strb->Base.surface->region); + if (strb->surface->region) { + pipe->region_release(pipe, &strb->surface->region); } /* Choose a pitch to match hardware requirements: */ pitch = ((cpp * width + 63) & ~63) / cpp; /* XXX fix: device-specific */ - strb->Base.surface->region = pipe->region_alloc(pipe, cpp, pitch, height); - if (!strb->Base.surface->region) + strb->surface->region = pipe->region_alloc(pipe, cpp, pitch, height); + if (!strb->surface->region) return GL_FALSE; /* out of memory, try s/w buffer? */ - ASSERT(strb->Base.surface->region->buffer); + ASSERT(strb->surface->region->buffer); - strb->Base.Width = strb->Base.surface->width = width; - strb->Base.Height = strb->Base.surface->height = height; + strb->Base.Width = strb->surface->width = width; + strb->Base.Height = strb->surface->height = height; return GL_TRUE; } @@ -206,11 +183,11 @@ st_renderbuffer_delete(struct gl_renderbuffer *rb) struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); ASSERT(strb); - if (strb && strb->Base.surface) { - if (rb->surface->region) { - pipe->region_release(pipe, &strb->Base.surface->region); + if (strb && strb->surface) { + if (strb->surface->region) { + pipe->region_release(pipe, &strb->surface->region); } - free(strb->Base.surface); + free(strb->surface); } free(strb); } @@ -285,28 +262,28 @@ st_new_renderbuffer_fb(struct pipe_region *region, GLuint width, GLuint height) struct gl_renderbuffer * st_new_renderbuffer_fb(GLuint intFormat) { - struct st_renderbuffer *irb; + struct st_renderbuffer *strb; - irb = CALLOC_STRUCT(st_renderbuffer); - if (!irb) { + strb = CALLOC_STRUCT(st_renderbuffer); + if (!strb) { _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer"); return NULL; } - _mesa_init_renderbuffer(&irb->Base, 0); - irb->Base.ClassID = 0x42; /* XXX temp */ - irb->Base.InternalFormat = intFormat; + _mesa_init_renderbuffer(&strb->Base, 0); + strb->Base.ClassID = 0x42; /* XXX temp */ + strb->Base.InternalFormat = intFormat; switch (intFormat) { case GL_RGB5: case GL_RGBA8: - irb->Base._BaseFormat = GL_RGBA; + strb->Base._BaseFormat = GL_RGBA; break; case GL_DEPTH_COMPONENT16: - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; + strb->Base._BaseFormat = GL_DEPTH_COMPONENT; break; case GL_DEPTH24_STENCIL8_EXT: - irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; + strb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; break; default: _mesa_problem(NULL, @@ -315,15 +292,14 @@ st_new_renderbuffer_fb(GLuint intFormat) } /* st-specific methods */ - irb->Base.Delete = st_renderbuffer_delete; - irb->Base.AllocStorage = st_renderbuffer_alloc_storage; - irb->Base.GetPointer = null_get_pointer; - /* span routines set in alloc_storage function */ + strb->Base.Delete = st_renderbuffer_delete; + strb->Base.AllocStorage = st_renderbuffer_alloc_storage; + strb->Base.GetPointer = null_get_pointer; - irb->Base.surface = NULL;/*intel_new_surface(intFormat);*/ - /*irb->Base.surface->rb = irb;*/ + /* surface is allocate in alloc_renderbuffer_storage() */ + strb->surface = NULL; - return &irb->Base; + return &strb->Base; } #endif diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index 7f52ab10d7..b2e7ba810c 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -30,10 +30,25 @@ #define ST_CB_FBO_H -/* -extern struct gl_renderbuffer * -st_new_renderbuffer_fb(struct pipe_region *region, GLuint width, GLuint height); -*/ + +/** + * Derived renderbuffer class. Just need to add a pointer to the + * pipe surface. + */ +struct st_renderbuffer +{ + struct gl_renderbuffer Base; + struct pipe_surface *surface; +}; + + +static INLINE struct st_renderbuffer * +st_renderbuffer(struct gl_renderbuffer *rb) +{ + return (struct st_renderbuffer *) rb; +} + + extern struct gl_renderbuffer * st_new_renderbuffer_fb(GLuint intFormat); -- cgit v1.2.3 From 74b5cb0e6fd0cdb6306fa7dd32e5be1aa4e3fb69 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 13:04:53 -0600 Subject: remove some includes --- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 1 - src/mesa/drivers/dri/intel_winsys/intel_context.c | 6 ---- src/mesa/drivers/dri/intel_winsys/intel_surface.c | 42 ++++++++++++++++++----- src/mesa/drivers/dri/intel_winsys/intel_surface.h | 27 +++++++++++++++ 4 files changed, 60 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index cbf6ea38b3..c149a5012b 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -35,7 +35,6 @@ #include "context.h" #include "utils.h" #include "drirenderbuffer.h" -#include "framebuffer.h" #include "vblank.h" #include "pipe/p_context.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 5aba4ef584..c4e8e0f95c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -28,15 +28,9 @@ #include "glheader.h" #include "context.h" -#include "matrix.h" -#include "simple_list.h" #include "extensions.h" -#include "framebuffer.h" -#include "imports.h" -#include "points.h" #include "tnl/tnl.h" - #include "tnl/t_pipeline.h" #include "tnl/t_vertex.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.c b/src/mesa/drivers/dri/intel_winsys/intel_surface.c index 13c34e6c8e..3a5d52b6d7 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.c @@ -1,13 +1,40 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Surface-related functions for intel drivers. + * \author Brian Paul + */ + + #include "glheader.h" -#include "context.h" -#include "framebuffer.h" -#include "renderbuffer.h" -#include "utils.h" #include "main/macros.h" - #include "intel_screen.h" - #include "intel_context.h" #include "intel_buffers.h" #include "intel_surface.h" @@ -18,9 +45,6 @@ #include "pipe/softpipe/sp_surface.h" -/* - * XXX a lof of this is a temporary kludge - */ /** * Note: the arithmetic/addressing in these functions is a little diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.h b/src/mesa/drivers/dri/intel_winsys/intel_surface.h index 90a1841efb..83a9bff2db 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.h @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #ifndef INTEL_SURFACE_H #define INTEL_SURFACE_H -- cgit v1.2.3 From 3a035d0d23d2ee7edd48be144c6619d56fbd5469 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 13:08:43 -0600 Subject: don't include intel_depthstencil.h --- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index c149a5012b..375f51c705 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -29,7 +29,6 @@ #include "intel_context.h" #include "intel_blit.h" #include "intel_buffers.h" -#include "intel_depthstencil.h" #include "intel_batchbuffer.h" #include "intel_reg.h" #include "context.h" -- cgit v1.2.3 From a47c387793bf99ad533a36082c262aadab7e322a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 13:09:27 -0600 Subject: Obsolete files --- .../drivers/dri/intel_winsys/intel_depthstencil.c | 282 ---------- .../drivers/dri/intel_winsys/intel_depthstencil.h | 14 - src/mesa/drivers/dri/intel_winsys/intel_fbo.c | 580 --------------------- src/mesa/drivers/dri/intel_winsys/intel_fbo.h | 99 ---- 4 files changed, 975 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_depthstencil.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_depthstencil.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_fbo.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_fbo.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.c b/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.c deleted file mode 100644 index 81e09698cc..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.c +++ /dev/null @@ -1,282 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "imports.h" -#include "context.h" -#include "depthstencil.h" -#include "fbobject.h" -#include "framebuffer.h" -#include "hash.h" -#include "mtypes.h" -#include "renderbuffer.h" - -#include "intel_context.h" -#include "intel_fbo.h" -#include "intel_depthstencil.h" - -#include "pipe/p_context.h" - -/** - * The GL_EXT_framebuffer_object allows the user to create their own - * framebuffer objects consisting of color renderbuffers (0 or more), - * depth renderbuffers (0 or 1) and stencil renderbuffers (0 or 1). - * - * The spec considers depth and stencil renderbuffers to be totally independent - * buffers. In reality, most graphics hardware today uses a combined - * depth+stencil buffer (one 32-bit pixel = 24 bits of Z + 8 bits of stencil). - * - * This causes difficulty because the user may create some number of depth - * renderbuffers and some number of stencil renderbuffers and bind them - * together in framebuffers in any combination. - * - * This code manages all that. - * - * 1. Depth renderbuffers are always allocated in hardware as 32bpp - * GL_DEPTH24_STENCIL8 buffers. - * - * 2. Stencil renderbuffers are initially allocated in software as 8bpp - * GL_STENCIL_INDEX8 buffers. - * - * 3. Depth and Stencil renderbuffers use the PairedStencil and PairedDepth - * fields (respectively) to indicate if the buffer's currently paired - * with another stencil or depth buffer (respectively). - * - * 4. When a depth and stencil buffer are initially both attached to the - * current framebuffer, we merge the stencil buffer values into the - * depth buffer (really a depth+stencil buffer). The then hardware uses - * the combined buffer. - * - * 5. Whenever a depth or stencil buffer is reallocated (with - * glRenderbufferStorage) we undo the pairing and copy the stencil values - * from the combined depth/stencil buffer back to the stencil-only buffer. - * - * 6. We also undo the pairing when we find a change in buffer bindings. - * - * 7. If a framebuffer is only using a depth renderbuffer (no stencil), we - * just use the combined depth/stencil buffer and ignore the stencil values. - * - * 8. If a framebuffer is only using a stencil renderbuffer (no depth) we have - * to promote the 8bpp software stencil buffer to a 32bpp hardware - * depth+stencil buffer. - * - */ - - - -static void -map_regions(GLcontext * ctx, - struct intel_renderbuffer *depthRb, - struct intel_renderbuffer *stencilRb) -{ - struct intel_context *intel = intel_context(ctx); - if (depthRb && depthRb->region) { - intel->pipe->region_map(intel->pipe, depthRb->region); - depthRb->pfMap = depthRb->region->map; - depthRb->pfPitch = depthRb->region->pitch; - } - if (stencilRb && stencilRb->region) { - intel->pipe->region_map(intel->pipe, stencilRb->region); - stencilRb->pfMap = stencilRb->region->map; - stencilRb->pfPitch = stencilRb->region->pitch; - } -} - -static void -unmap_regions(GLcontext * ctx, - struct intel_renderbuffer *depthRb, - struct intel_renderbuffer *stencilRb) -{ - struct intel_context *intel = intel_context(ctx); - if (depthRb && depthRb->region) { - intel->pipe->region_unmap(intel->pipe, depthRb->region); - depthRb->pfMap = NULL; - depthRb->pfPitch = 0; - } - if (stencilRb && stencilRb->region) { - intel->pipe->region_unmap(intel->pipe, stencilRb->region); - stencilRb->pfMap = NULL; - stencilRb->pfPitch = 0; - } -} - - - -/** - * Undo the pairing/interleaving between depth and stencil buffers. - * irb should be a depth/stencil or stencil renderbuffer. - */ -void -intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb) -{ - if (irb->PairedStencil) { - /* irb is a depth/stencil buffer */ - struct gl_renderbuffer *stencilRb; - struct intel_renderbuffer *stencilIrb; - - ASSERT(irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - - stencilRb = _mesa_lookup_renderbuffer(ctx, irb->PairedStencil); - stencilIrb = intel_renderbuffer(stencilRb); - if (stencilIrb) { - /* need to extract stencil values from the depth buffer */ - ASSERT(stencilIrb->PairedDepth == irb->Base.Name); - map_regions(ctx, irb, stencilIrb); - _mesa_extract_stencil(ctx, &irb->Base, &stencilIrb->Base); - unmap_regions(ctx, irb, stencilIrb); - stencilIrb->PairedDepth = 0; - } - irb->PairedStencil = 0; - } - else if (irb->PairedDepth) { - /* irb is a stencil buffer */ - struct gl_renderbuffer *depthRb; - struct intel_renderbuffer *depthIrb; - - ASSERT(irb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || - irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - - depthRb = _mesa_lookup_renderbuffer(ctx, irb->PairedDepth); - depthIrb = intel_renderbuffer(depthRb); - if (depthIrb) { - /* need to extract stencil values from the depth buffer */ - ASSERT(depthIrb->PairedStencil == irb->Base.Name); - map_regions(ctx, depthIrb, irb); - _mesa_extract_stencil(ctx, &depthIrb->Base, &irb->Base); - unmap_regions(ctx, depthIrb, irb); - depthIrb->PairedStencil = 0; - } - irb->PairedDepth = 0; - } - else { - _mesa_problem(ctx, "Problem in undo_depth_stencil_pairing"); - } - - ASSERT(irb->PairedStencil == 0); - ASSERT(irb->PairedDepth == 0); -} - - -/** - * Examine the depth and stencil renderbuffers which are attached to the - * framebuffer. If both depth and stencil are attached, make sure that the - * renderbuffers are 'paired' (combined). If only depth or only stencil is - * attached, undo any previous pairing. - * - * Must be called if NewState & _NEW_BUFFER (when renderbuffer attachments - * change, for example). - */ -void -intel_validate_paired_depth_stencil(GLcontext * ctx, - struct gl_framebuffer *fb) -{ - struct intel_renderbuffer *depthRb, *stencilRb; - - depthRb = intel_get_renderbuffer(fb, BUFFER_DEPTH); - stencilRb = intel_get_renderbuffer(fb, BUFFER_STENCIL); - - if (depthRb && stencilRb) { - if (depthRb == stencilRb) { - /* Using a user-created combined depth/stencil buffer. - * Nothing to do. - */ - ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_STENCIL_EXT); - ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - } - else { - /* Separate depth/stencil buffers, need to interleave now */ - ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_COMPONENT); - ASSERT(stencilRb->Base._BaseFormat == GL_STENCIL_INDEX); - /* may need to interleave depth/stencil now */ - if (depthRb->PairedStencil == stencilRb->Base.Name) { - /* OK, the depth and stencil buffers are already interleaved */ - ASSERT(stencilRb->PairedDepth == depthRb->Base.Name); - } - else { - /* need to setup new pairing/interleaving */ - if (depthRb->PairedStencil) { - intel_unpair_depth_stencil(ctx, depthRb); - } - if (stencilRb->PairedDepth) { - intel_unpair_depth_stencil(ctx, stencilRb); - } - - ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - ASSERT(stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || - stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - - /* establish new pairing: interleave stencil into depth buffer */ - map_regions(ctx, depthRb, stencilRb); - _mesa_insert_stencil(ctx, &depthRb->Base, &stencilRb->Base); - unmap_regions(ctx, depthRb, stencilRb); - depthRb->PairedStencil = stencilRb->Base.Name; - stencilRb->PairedDepth = depthRb->Base.Name; - } - - } - } - else if (depthRb) { - /* Depth buffer but no stencil buffer. - * We'll use a GL_DEPTH24_STENCIL8 buffer and ignore the stencil bits. - */ - /* can't assert this until storage is allocated: - ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - */ - /* intel_undo any previous pairing */ - if (depthRb->PairedStencil) { - intel_unpair_depth_stencil(ctx, depthRb); - } - } - else if (stencilRb) { - /* Stencil buffer but no depth buffer. - * Since h/w doesn't typically support just 8bpp stencil w/out Z, - * we'll use a GL_DEPTH24_STENCIL8 buffer and ignore the depth bits. - */ - /* undo any previous pairing */ - if (stencilRb->PairedDepth) { - intel_unpair_depth_stencil(ctx, stencilRb); - } - if (stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT) { - /* promote buffer to GL_DEPTH24_STENCIL8 for hw rendering */ - _mesa_promote_stencil(ctx, &stencilRb->Base); - ASSERT(stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - } - } - - /* Finally, update the fb->_DepthBuffer and fb->_StencilBuffer fields */ - _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH); - if (depthRb && depthRb->PairedStencil) - _mesa_update_stencil_buffer(ctx, fb, BUFFER_DEPTH); - else - _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL); - - - /* The hardware should use fb->Attachment[BUFFER_DEPTH].Renderbuffer - * first, if present, then fb->Attachment[BUFFER_STENCIL].Renderbuffer - * if present. - */ -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.h b/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.h deleted file mode 100644 index 2d3fc48b3a..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_depthstencil.h +++ /dev/null @@ -1,14 +0,0 @@ - -#ifndef INTEL_DEPTH_STENCIL_H -#define INTEL_DEPTH_STENCIL_H - - -extern void -intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb); - -extern void -intel_validate_paired_depth_stencil(GLcontext * ctx, - struct gl_framebuffer *fb); - - -#endif /* INTEL_DEPTH_STENCIL_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c b/src/mesa/drivers/dri/intel_winsys/intel_fbo.c deleted file mode 100644 index 3dd99747da..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.c +++ /dev/null @@ -1,580 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "imports.h" -#include "mtypes.h" -#include "fbobject.h" -#include "framebuffer.h" -#include "renderbuffer.h" -#include "context.h" -#include "texformat.h" -#include "texrender.h" - -#include "intel_context.h" -#include "intel_buffers.h" -#include "intel_depthstencil.h" -#include "intel_fbo.h" -#include "state_tracker/st_mipmap_tree.h" -/*#include "intel_tex.h"*/ - -#include "pipe/p_context.h" - -#define FILE_DEBUG_FLAG DEBUG_FBO - -#define INTEL_RB_CLASS 0x12345678 - - -/* XXX FBO: move this to intel_context.h (inlined) */ -/** - * Return a gl_renderbuffer ptr casted to intel_renderbuffer. - * NULL will be returned if the rb isn't really an intel_renderbuffer. - * This is determiend by checking the ClassID. - */ -struct intel_renderbuffer * -intel_renderbuffer(struct gl_renderbuffer *rb) -{ - struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; - if (irb && irb->Base.ClassID == INTEL_RB_CLASS) { - /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/ - return irb; - } - else - return NULL; -} - - -/** - * Create a new framebuffer object. - */ -static struct gl_framebuffer * -intel_new_framebuffer(GLcontext * ctx, GLuint name) -{ - /* Only drawable state in intel_framebuffer at this time, just use Mesa's - * class - */ - return _mesa_new_framebuffer(ctx, name); -} - - -static void -intel_delete_renderbuffer(struct gl_renderbuffer *rb) -{ - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); - struct intel_renderbuffer *irb = intel_renderbuffer(rb); - - ASSERT(irb); - - DBG("freeing renderbuffer\n"); - -#if 0 - if (irb->PairedStencil || irb->PairedDepth) { - intel_unpair_depth_stencil(ctx, irb); - } -#endif - if (intel && irb->Base.surface->region) { - intel->pipe->region_release(intel->pipe, &irb->Base.surface->region); - } - - _mesa_free(irb); -} - - - -/** - * Return a pointer to a specific pixel in a renderbuffer. - */ -static void * -intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb, - GLint x, GLint y) -{ - /* By returning NULL we force all software rendering to go through - * the span routines. - */ - return NULL; -} - - - -/** - * Called via glRenderbufferStorageEXT() to set the format and allocate - * storage for a user-created (or priv buffer) renderbuffer. - */ -static GLboolean -intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, - GLuint width, GLuint height) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_renderbuffer *irb = intel_renderbuffer(rb); - GLboolean softwareBuffer = GL_FALSE; - int cpp; - - switch (internalFormat) { - case GL_R3_G3_B2: - case GL_RGB4: - case GL_RGB5: - rb->_ActualFormat = GL_RGB5; - rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 5; - rb->GreenBits = 6; - rb->BlueBits = 5; - cpp = 2; - break; - case GL_RGB: - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - case GL_RGBA: - case GL_RGBA2: - case GL_RGBA4: - case GL_RGB5_A1: - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - rb->_ActualFormat = GL_RGBA8; - rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 8; - rb->GreenBits = 8; - rb->BlueBits = 8; - rb->AlphaBits = 8; - cpp = 4; - break; - case GL_STENCIL_INDEX: - case GL_STENCIL_INDEX1_EXT: - case GL_STENCIL_INDEX4_EXT: - case GL_STENCIL_INDEX8_EXT: - case GL_STENCIL_INDEX16_EXT: - /* alloc a depth+stencil buffer */ - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->StencilBits = 8; - cpp = 4; - break; - case GL_DEPTH_COMPONENT16: - rb->_ActualFormat = GL_DEPTH_COMPONENT16; - rb->DataType = GL_UNSIGNED_SHORT; - rb->DepthBits = 16; - cpp = 2; - break; - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - cpp = 4; - break; - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - rb->StencilBits = 8; - cpp = 4; - break; - default: - _mesa_problem(ctx, - "Unexpected format (%x) in intel_alloc_renderbuffer_storage", internalFormat); - return GL_FALSE; - } - - intelFlush(ctx); - - /* free old region */ - if (irb->Base.surface->region) { - intel->pipe->region_release(intel->pipe, &irb->Base.surface->region); - } - - /* allocate new memory region/renderbuffer */ - if (softwareBuffer) { - return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat, - width, height); - } - else { - /* Choose a pitch to match hardware requirements: - */ - GLuint pitch = ((cpp * width + 63) & ~63) / cpp; - - /* alloc hardware renderbuffer */ - DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, - height, pitch); - - irb->Base.surface->region = intel->pipe->region_alloc(intel->pipe, cpp, pitch, height); - if (!irb->Base.surface->region) - return GL_FALSE; /* out of memory? */ - - ASSERT(irb->Base.surface->region->buffer); - - rb->Width = width; - rb->Height = height; - - /* update the surface's size too */ - rb->surface->width = width; - rb->surface->height = height; - rb->surface->region = irb->Base.surface->region;/*XXX no-op*/ - - /* This sets the Get/PutRow/Value functions */ - // intel_set_span_functions(&irb->Base); - - return GL_TRUE; - } -} - - -static void -intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, - GLuint width, GLuint height) -{ - struct intel_framebuffer *intel_fb = (struct intel_framebuffer*)fb; - int i; - - _mesa_resize_framebuffer(ctx, fb, width, height); - - fb->Initialized = GL_TRUE; /* XXX remove someday */ - - if (fb->Name != 0) { - return; - } - - /* Make sure all window system renderbuffers are up to date */ -#if 0 - for (i = 0; i < 3; i++) { - struct gl_renderbuffer *rb = &intel_fb->color_rb[i]->Base; - - /* only resize if size is changing */ - if (rb && (rb->Width != width || rb->Height != height)) { - rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height); - } - } -#else - -#endif -} - -static GLboolean -intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, GLuint width, GLuint height) -{ - _mesa_problem(ctx, "intel_op_alloc_storage should never be called."); - return GL_FALSE; -} - - - -struct intel_renderbuffer * -intel_new_renderbuffer_fb(GLuint intFormat) -{ - struct intel_renderbuffer *irb; - - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, 0); - irb->Base.ClassID = INTEL_RB_CLASS; - irb->Base.InternalFormat = intFormat; - - switch (intFormat) { - case GL_RGB5: - case GL_RGBA8: - irb->Base._BaseFormat = GL_RGBA; - break; - case GL_DEPTH_COMPONENT16: - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - break; - case GL_DEPTH24_STENCIL8_EXT: - irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; - break; - default: - _mesa_problem(NULL, - "Unexpected intFormat in intel_create_renderbuffer"); - return NULL; - } - - /* intel-specific methods */ - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; - irb->Base.GetPointer = intel_get_pointer; - /* span routines set in alloc_storage function */ - - irb->Base.surface = intel_new_surface(intFormat); - irb->Base.surface->rb = irb; - - return irb; -} - -/** - * Create a new renderbuffer object. - * Typically called via glBindRenderbufferEXT(). - */ -static struct gl_renderbuffer * -intel_new_renderbuffer(GLcontext * ctx, GLuint name) -{ - /*struct intel_context *intel = intel_context(ctx); */ - struct intel_renderbuffer *irb; - - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, name); - irb->Base.ClassID = INTEL_RB_CLASS; - - /* intel-specific methods */ - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; - irb->Base.GetPointer = intel_get_pointer; - /* span routines set in alloc_storage function */ - - irb->Base.surface = intel_new_surface(0 /*unknown format*/); - irb->Base.surface->rb = irb; - - return &irb->Base; -} - - -/** - * Called via glBindFramebufferEXT(). - */ -static void -intel_bind_framebuffer(GLcontext * ctx, GLenum target, - struct gl_framebuffer *fb, struct gl_framebuffer *fbread) -{ - if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) { - intel_draw_buffer(ctx, fb); - /* Integer depth range depends on depth buffer bits */ - /* XXX - */ -// ctx->Driver.DepthRange(ctx, ctx->Viewport.Near, ctx->Viewport.Far); - } - else { - /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */ - } -} - - -/** - * Called via glFramebufferRenderbufferEXT(). - */ -static void -intel_framebuffer_renderbuffer(GLcontext * ctx, - struct gl_framebuffer *fb, - GLenum attachment, struct gl_renderbuffer *rb) -{ - DBG("Intel FramebufferRenderbuffer %u %u\n", fb->Name, rb ? rb->Name : 0); - - intelFlush(ctx); - - _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb); - intel_draw_buffer(ctx, fb); -} - - -/** - * When glFramebufferTexture[123]D is called this function sets up the - * gl_renderbuffer wrapper around the texture image. - * This will have the region info needed for hardware rendering. - */ -static struct intel_renderbuffer * -intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage) -{ - const GLuint name = ~0; /* not significant, but distinct for debugging */ - struct intel_renderbuffer *irb; - - /* make an intel_renderbuffer to wrap the texture image */ - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, name); - irb->Base.ClassID = INTEL_RB_CLASS; - - if (texImage->TexFormat == &_mesa_texformat_argb8888) { - irb->Base._ActualFormat = GL_RGBA8; - irb->Base._BaseFormat = GL_RGBA; - DBG("Render to RGBA8 texture OK\n"); - } - else if (texImage->TexFormat == &_mesa_texformat_rgb565) { - irb->Base._ActualFormat = GL_RGB5; - irb->Base._BaseFormat = GL_RGB; - DBG("Render to RGB5 texture OK\n"); - } - else if (texImage->TexFormat == &_mesa_texformat_z16) { - irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - DBG("Render to DEPTH16 texture OK\n"); - } - else { - DBG("Render to texture BAD FORMAT %d\n", - texImage->TexFormat->MesaFormat); - _mesa_free(irb); - return NULL; - } - - irb->Base.InternalFormat = irb->Base._ActualFormat; - irb->Base.Width = texImage->Width; - irb->Base.Height = texImage->Height; - irb->Base.DataType = GL_UNSIGNED_BYTE; /* FBO XXX fix */ - irb->Base.RedBits = texImage->TexFormat->RedBits; - irb->Base.GreenBits = texImage->TexFormat->GreenBits; - irb->Base.BlueBits = texImage->TexFormat->BlueBits; - irb->Base.AlphaBits = texImage->TexFormat->AlphaBits; - irb->Base.DepthBits = texImage->TexFormat->DepthBits; - - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_nop_alloc_storage; - -// intel_set_span_functions(&irb->Base); - -#if 0 - irb->RenderToTexture = GL_TRUE; -#endif - - return irb; -} - - -/** - * Called by glFramebufferTexture[123]DEXT() (and other places) to - * prepare for rendering into texture memory. This might be called - * many times to choose different texture levels, cube faces, etc - * before intel_finish_render_texture() is ever called. - */ -static void -intel_render_texture(GLcontext * ctx, - struct gl_framebuffer *fb, - struct gl_renderbuffer_attachment *att) -{ -#if 0 - struct intel_context *intel = intel_context(ctx); - struct gl_texture_image *newImage - = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; - struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); - struct st_texture_image *st_image; - GLuint imageOffset; - - (void) fb; - - ASSERT(newImage); - - if (!irb) { - irb = intel_wrap_texture(ctx, newImage); - if (irb) { - /* bind the wrapper to the attachment point */ - _mesa_reference_renderbuffer(&att->Renderbuffer, &irb->Base); - } - else { - /* fallback to software rendering */ - _mesa_render_texture(ctx, fb, att); - return; - } - } - - DBG("Begin render texture tid %x tex=%u w=%d h=%d refcount=%d\n", - _glthread_GetID(), - att->Texture->Name, newImage->Width, newImage->Height, - irb->Base.RefCount); - - /* point the renderbufer's region to the texture image region */ - st_image = st_texture_image(newImage); - if (irb->region != st_image->mt->region) { - if (irb->region) - intel->pipe->region_release(intel->pipe, &irb->region); - pipe_region_reference(&irb->region, st_image->mt->region); - } - - /* compute offset of the particular 2D image within the texture region */ - imageOffset = st_miptree_image_offset(st_image->mt, - att->CubeMapFace, - att->TextureLevel); - - if (att->Texture->Target == GL_TEXTURE_3D) { - const GLuint *offsets = st_miptree_depth_offsets(st_image->mt, - att->TextureLevel); - imageOffset += offsets[att->Zoffset]; - } - - /* store that offset in the region */ -#if 0 - st_image->mt->region->draw_offset = imageOffset; -#endif - - /* update drawing region, etc */ - intel_draw_buffer(ctx, fb); -#endif -} - - -/** - * Called by Mesa when rendering to a texture is done. - */ -static void -intel_finish_render_texture(GLcontext * ctx, - struct gl_renderbuffer_attachment *att) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); - - DBG("End render texture (tid %x) tex %u\n", _glthread_GetID(), att->Texture->Name); - - if (irb) { - /* just release the region */ - intel->pipe->region_release(intel->pipe, &irb->Base.surface->region); - } - else if (att->Renderbuffer) { - /* software fallback */ - _mesa_finish_render_texture(ctx, att); - /* XXX FBO: Need to unmap the buffer (or in intelSpanRenderStart???) */ - } -} - - -/** - * Do one-time context initializations related to GL_EXT_framebuffer_object. - * Hook in device driver functions. - */ -void -intel_fbo_init(struct intel_context *intel) -{ - intel->ctx.Driver.NewFramebuffer = intel_new_framebuffer; - intel->ctx.Driver.NewRenderbuffer = intel_new_renderbuffer; - intel->ctx.Driver.BindFramebuffer = intel_bind_framebuffer; - intel->ctx.Driver.FramebufferRenderbuffer = intel_framebuffer_renderbuffer; - intel->ctx.Driver.RenderTexture = intel_render_texture; - intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture; - intel->ctx.Driver.ResizeBuffers = intel_resize_buffers; -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h b/src/mesa/drivers/dri/intel_winsys/intel_fbo.h deleted file mode 100644 index 0487c27600..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_fbo.h +++ /dev/null @@ -1,99 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_FBO_H -#define INTEL_FBO_H - - -#include "pipe/p_state.h" -#include "pipe/softpipe/sp_surface.h" - - -struct intel_context; -struct pipe_region; - -#if 0 -/** - * Intel framebuffer, derived from gl_framebuffer. - */ -struct intel_framebuffer -{ - struct gl_framebuffer Base; - - /* Drawable page flipping state */ - GLboolean pf_active; - GLuint pf_seq; - GLint pf_pipes; - GLint pf_current_page; - GLint pf_num_pages; - - /* VBI - */ - GLuint vbl_seq; - GLuint vblank_flags; - GLuint vbl_waited; - - int64_t swap_ust; - int64_t swap_missed_ust; - - GLuint swap_count; - GLuint swap_missed_count; - - GLuint vbl_pending[3]; /**< [number of color buffers] */ -}; -#endif - - -/** - * Intel renderbuffer, derived from gl_renderbuffer. - * Note: The PairedDepth and PairedStencil fields use renderbuffer IDs, - * not pointers because in some circumstances a deleted renderbuffer could - * result in a dangling pointer here. - */ -#if 0 -struct intel_renderbuffer -{ - struct gl_renderbuffer Base; -}; -#endif - -#if 0 -extern struct intel_renderbuffer *intel_new_renderbuffer_fb(GLuint intFormat); -#endif - -extern void intel_fbo_init(struct intel_context *intel); - - -extern struct pipe_surface * -intel_new_surface(struct pipe_context *pipe, GLuint pipeFormat); - - -extern const GLuint * -intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats); - - -#endif /* INTEL_FBO_H */ -- cgit v1.2.3 From a4517215ca98066c70155432ebfc578e6412d6de Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 13:09:58 -0600 Subject: remove intel_depthstencil.c and intel_fbo.c --- src/mesa/drivers/dri/intel_winsys/Makefile | 3 --- 1 file changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index 32c2c3bb24..39ae6f2f33 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -24,9 +24,6 @@ DRIVER_SOURCES = \ intel_surface.c \ intel_batchpool.c -UNUSED = intel_depthstencil.c \ - intel_fbo.c - C_SOURCES = \ $(COMMON_SOURCES) \ $(COMMON_BM_SOURCES) \ -- cgit v1.2.3 From 3e930c3f1fe1deebd0cb976070155f5c55d55b6f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 13:17:49 -0600 Subject: minor clean-ups --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 13 ++++--------- 1 file changed, 4 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index b765c2b4b2..4c99f2ed5f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -290,7 +290,6 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); - if (!intel_fb) return GL_FALSE; @@ -300,14 +299,12 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, /* fake frontbuffer */ /* XXX allocation should only happen in the unusual case it's actually needed */ - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(rgbFormat); + struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, rb); } if (mesaVis->doubleBufferMode) { - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(rgbFormat); + struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, rb); } @@ -316,10 +313,8 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, struct gl_renderbuffer *depthStencilRb = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); /* note: bind RB to two attachment points */ - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, - depthStencilRb); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL, - depthStencilRb); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, depthStencilRb); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL,depthStencilRb); } else if (mesaVis->depthBits == 16) { /* just 16-bit depth buffer, no hw stencil */ -- cgit v1.2.3 From 807ecbd344853288f4cadff95445d5677b28c93c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 15:39:31 -0600 Subject: *numFormats = 3 --- src/mesa/pipe/i915simple/i915_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 02b51bea43..3e8d1fbb33 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -98,7 +98,7 @@ i915_supported_formats(struct pipe_context *pipe, PIPE_FORMAT_U_R5_G6_B5, PIPE_FORMAT_S8_Z24, }; - *numFormats = 2; + *numFormats = 3; return render_supported; #endif } -- cgit v1.2.3 From 3c07bbb1982da5629685f569285529f02b35e48c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 15:39:54 -0600 Subject: assertions --- src/mesa/state_tracker/st_cb_fbo.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 02bdb5aba5..15816bee9e 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -131,6 +131,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, const struct pipe_format_info *info = pipe_get_format_info(pipeFormat); GLuint cpp, pitch; + assert(info); if (!info) return GL_FALSE; @@ -147,6 +148,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, if (!strb->surface) { strb->surface = pipe->surface_alloc(pipe, pipeFormat); + assert(strb->surface); if (!strb->surface) return GL_FALSE; } -- cgit v1.2.3 From 0c4c86cf2ce65d1284819dd14c649ac4c33cae33 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 15:41:08 -0600 Subject: don't use intel_read_draw_buffer.c anymore --- src/mesa/drivers/dri/intel_winsys/Makefile | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index 39ae6f2f33..786bf28a39 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -16,7 +16,6 @@ DRIVER_SOURCES = \ intel_batchbuffer.c \ intel_tex_layout.c \ intel_buffers.c \ - intel_read_draw_buffer.c \ intel_blit.c \ intel_context.c \ intel_ioctl.c \ -- cgit v1.2.3 From fb8caffea8ac9a8f8e2ee63959264abe8eb36c59 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 15:42:12 -0600 Subject: fix merge conflicts --- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 35 ----------------------- src/mesa/drivers/dri/intel_winsys/intel_context.c | 6 ++-- 2 files changed, 4 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index 375f51c705..eb42be98e5 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -90,41 +90,6 @@ intel_intersect_cliprects(drm_clip_rect_t * dst, return GL_TRUE; } -<<<<<<< HEAD:src/mesa/drivers/dri/intel_winsys/intel_buffers.c -/** - * Return pointer to current color drawing region, or NULL. - */ -#if 0 -struct pipe_region * -intel_drawbuf_region(struct intel_context *intel) -{ - struct intel_renderbuffer *irbColor = - intel_renderbuffer(intel->ctx.DrawBuffer->_ColorDrawBuffers[0][0]); - if (irbColor) - return irbColor->Base.surface->region; - else - return NULL; -} -======= ->>>>>>> remove dead code, remove intel_fbo.h includes:src/mesa/drivers/dri/intel_winsys/intel_buffers.c - -<<<<<<< HEAD:src/mesa/drivers/dri/intel_winsys/intel_buffers.c -/** - * Return pointer to current color reading region, or NULL. - */ -struct pipe_region * -intel_readbuf_region(struct intel_context *intel) -{ - struct intel_renderbuffer *irb - = intel_renderbuffer(intel->ctx.ReadBuffer->_ColorReadBuffer); - if (irb) - return irb->Base.surface->region; - else - return NULL; -} -#endif -======= ->>>>>>> remove dead code, remove intel_fbo.h includes:src/mesa/drivers/dri/intel_winsys/intel_buffers.c /** * This will be called whenever the currently bound window is moved/resized. diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index c4e8e0f95c..9d6ef058a3 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -351,7 +351,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, /* Initialize the software rasterizer and helper modules. */ _vbo_CreateContext(ctx); _tnl_CreateContext(ctx); - +#if 0 /* * Pipe-related setup */ @@ -404,7 +404,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, default: assert(0); /*FIX*/ } - +#endif /* * memory pools @@ -465,6 +465,8 @@ intelCreateContext(const __GLcontextModes * mesaVis, */ if (!getenv("INTEL_HW")) { intel->pipe = intel_create_softpipe( intel ); + intel->pipe->surface_alloc = intel_new_surface; + intel->pipe->supported_formats = intel_supported_formats; } else { switch (intel->intelScreen->deviceID) { -- cgit v1.2.3 From 277e4989f348913e3852f3d8c4efb82ba1380fcb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 12:02:45 +0100 Subject: Don't reemit hardware state *every* primitive. --- src/mesa/pipe/i915simple/i915_context.h | 1 + src/mesa/pipe/i915simple/i915_prim_emit.c | 6 +++++- src/mesa/pipe/i915simple/i915_state_derived.c | 3 +++ src/mesa/pipe/i915simple/i915_state_emit.c | 2 ++ 4 files changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index dab5423190..c298f0f581 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -56,6 +56,7 @@ struct i915_context struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; GLuint dirty; + GLuint hw_dirty; GLuint *batch_start; diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index 6d7cfe2e3e..a1e0e85196 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -120,7 +120,11 @@ emit_prim( struct draw_stage *stage, unsigned *ptr; unsigned i; - i915_emit_hardware_state( i915 ); + if (i915->dirty) + i915_update_derived( i915 ); + + if (i915->hw_dirty) + i915_emit_hardware_state( i915 ); ptr = winsys->batch_start( winsys, nr * vertex_size, 0 ); if (ptr == 0) { diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 6e91e4448f..f1f798b506 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -182,5 +182,8 @@ void i915_update_derived( struct i915_context *i915 ) I915_NEW_FRAMEBUFFER)) compute_cliprect(i915); + if (i915->dirty) + i915->hw_dirty = 1; + i915->dirty = 0; } diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index aeb2d6753f..3553d565c5 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -250,5 +250,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) for (i = 0; i < dwords; i++) OUT_BATCH( prog[i] ); } + + i915->hw_dirty = 0; } -- cgit v1.2.3 From bb142746bdf2c08a924471c53c660b1dea60cd8b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 15:52:38 +0100 Subject: Remove unused blit functions. --- src/mesa/drivers/dri/intel_winsys/intel_blit.c | 187 ------------------------- 1 file changed, 187 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_blit.c b/src/mesa/drivers/dri/intel_winsys/intel_blit.c index aa4135ed2d..f099eb2483 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_blit.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_blit.c @@ -224,193 +224,6 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, -void -intelEmitFillBlit(struct intel_context *intel, - GLuint cpp, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort x, GLshort y, GLshort w, GLshort h, - GLuint value, GLuint mask) -{ - GLuint BR13, CMD; - GLboolean badMask = GL_FALSE; - BATCH_LOCALS; - - /* - printf("Emit fill blit value=0x%x mask=0x%x\n", value, mask); - */ - - dst_pitch *= cpp; - - switch (cpp) { - case 1: - case 2: - case 3: - BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); - CMD = XY_COLOR_BLT_CMD; - if ((mask & 0xffff) != 0xffff) - badMask = GL_TRUE; - break; - case 4: - BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); -#if 0 - CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); -#else - CMD = XY_COLOR_BLT_CMD; - if ((mask & 0xff000000) == 0xff000000) - CMD |= XY_COLOR_BLT_WRITE_ALPHA; - else if (mask & 0xff000000) - badMask = GL_TRUE; - if ((mask & 0x00ffffff) == 0x00ffffff) - CMD |= XY_COLOR_BLT_WRITE_RGB; - else if (mask & 0x00ffffff) - badMask = GL_TRUE; -#endif - break; - default: - return; - } - - assert(!badMask); - - DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); - - - BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((y << 16) | x); - OUT_BATCH(((y + h) << 16) | (x + w)); - OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); - OUT_BATCH(value); - ADVANCE_BATCH(); - - intel_batchbuffer_flush(intel->batch); -} - - -static GLuint translate_raster_op(GLenum logicop) -{ - switch(logicop) { - case GL_CLEAR: return 0x00; - case GL_AND: return 0x88; - case GL_AND_REVERSE: return 0x44; - case GL_COPY: return 0xCC; - case GL_AND_INVERTED: return 0x22; - case GL_NOOP: return 0xAA; - case GL_XOR: return 0x66; - case GL_OR: return 0xEE; - case GL_NOR: return 0x11; - case GL_EQUIV: return 0x99; - case GL_INVERT: return 0x55; - case GL_OR_REVERSE: return 0xDD; - case GL_COPY_INVERTED: return 0x33; - case GL_OR_INVERTED: return 0xBB; - case GL_NAND: return 0x77; - case GL_SET: return 0xFF; - default: return 0; - } -} - - -/* Copy BitBlt - */ -void -intelEmitCopyBlit(struct intel_context *intel, - GLuint cpp, - GLshort src_pitch, - struct _DriBufferObject *src_buffer, - GLuint src_offset, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort src_x, GLshort src_y, - GLshort dst_x, GLshort dst_y, - GLshort w, GLshort h, - GLenum logic_op) -{ - GLuint CMD, BR13; - int dst_y2 = dst_y + h; - int dst_x2 = dst_x + w; - BATCH_LOCALS; - - - DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, - src_buffer, src_pitch, src_offset, src_x, src_y, - dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); - - src_pitch *= cpp; - dst_pitch *= cpp; - - switch (cpp) { - case 1: - case 2: - case 3: - BR13 = (((GLint) dst_pitch) & 0xffff) | - (translate_raster_op(logic_op) << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - break; - case 4: - BR13 = - (((GLint) dst_pitch) & 0xffff) | - (translate_raster_op(logic_op) << 16) | (1 << 24) | (1 << 25); - CMD = - (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - break; - default: - return; - } - - if (dst_y2 < dst_y || dst_x2 < dst_x) { - return; - } - - /* Initial y values don't seem to work with negative pitches. If - * we adjust the offsets manually (below), it seems to work fine. - * - * On the other hand, if we always adjust, the hardware doesn't - * know which blit directions to use, so overlapping copypixels get - * the wrong result. - */ - if (dst_pitch > 0 && src_pitch > 0) { - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((dst_y << 16) | dst_x); - OUT_BATCH((dst_y2 << 16) | dst_x2); - OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); - OUT_BATCH((src_y << 16) | src_x); - OUT_BATCH(((GLint) src_pitch & 0xffff)); - OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, src_offset); - ADVANCE_BATCH(); - } - else { - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((0 << 16) | dst_x); - OUT_BATCH((h << 16) | dst_x2); - OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - dst_offset + dst_y * dst_pitch); - OUT_BATCH((0 << 16) | src_x); - OUT_BATCH(((GLint) src_pitch & 0xffff)); - OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, - src_offset + src_y * src_pitch); - ADVANCE_BATCH(); - } - - intel_batchbuffer_flush( intel->batch ); -} -- cgit v1.2.3 From 94fda5695e3e05d93dfa20e8d193c2a784c2ae65 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 16:00:43 +0100 Subject: Emit drawrect command as we no longer ask the kernel to do this. --- src/mesa/pipe/i915simple/i915_state_emit.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 3553d565c5..3b2ab4775b 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -251,6 +251,19 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH( prog[i] ); } + { + int w = i915->framebuffer.cbufs[0]->width; + int h = i915->framebuffer.cbufs[0]->height; + + OUT_BATCH(_3DSTATE_DRAW_RECT_CMD); + OUT_BATCH(0); + OUT_BATCH(0); + OUT_BATCH(((w - 1) & 0xffff) | ((h - 1) << 16)); + OUT_BATCH(0); + OUT_BATCH(0); + } + + i915->hw_dirty = 0; } -- cgit v1.2.3 From e24c696ca8035bc867899ee6f8ab39f5f276ac82 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 16:04:39 -0600 Subject: dead code removal --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 94 +++-------------------- 1 file changed, 11 insertions(+), 83 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 9d6ef058a3..4ba8394f88 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -351,60 +351,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, /* Initialize the software rasterizer and helper modules. */ _vbo_CreateContext(ctx); _tnl_CreateContext(ctx); -#if 0 - /* - * Pipe-related setup - */ - if (!getenv("INTEL_HW")) { - intel->pipe = intel_create_softpipe( intel ); - intel->pipe->surface_alloc = intel_new_surface; - intel->pipe->supported_formats = intel_supported_formats; - } - else { - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - intel->pipe = intel_create_i915simple( intel ); - break; - default: - _mesa_printf("Unknown PCIID %x in %s, using software driver\n", - intel->intelScreen->deviceID, __FUNCTION__); - - intel->pipe = intel_create_softpipe( intel ); - break; - } - } - - st_create_context( &intel->ctx, intel->pipe ); - - /* TODO: Push this down into the pipe driver: - */ - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - intel->pipe->mipmap_tree_layout = i945_miptree_layout; - break; - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - case PCI_CHIP_I830_M: - case PCI_CHIP_I855_GM: - case PCI_CHIP_I865_G: - intel->pipe->mipmap_tree_layout = i915_miptree_layout; - default: - assert(0); /*FIX*/ - } -#endif /* * memory pools @@ -424,7 +371,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; - fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); intel->iw.irq_seq = -1; intel->irqsEmitted = 0; @@ -441,10 +387,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->last_swap_fence = NULL; intel->first_swap_fence = NULL; -#if 00 - intel_fbo_init(intel); -#endif - if (intel->ctx.Mesa_DXTn) { _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); _mesa_enable_extension(ctx, "GL_S3_s3tc"); @@ -458,8 +400,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, #endif - - /* * Pipe-related setup */ @@ -513,7 +453,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, assert(0); /*FIX*/ } - return GL_TRUE; } @@ -710,24 +649,13 @@ void LOCK_HARDWARE( struct intel_context *intel ) { char __ret=0; struct intel_framebuffer *intel_fb = NULL; -#if 0 - struct intel_renderbuffer *intel_rb = NULL; -#else int curbuf; -#endif + _glthread_LOCK_MUTEX(lockMutex); assert(!intel->locked); if (intel->driDrawable) { intel_fb = intel->driDrawable->driverPrivate; -#if 0 - if (intel_fb) - intel_rb = - intel_get_renderbuffer(&intel_fb->Base, - intel_fb->Base._ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : - BUFFER_BACK_LEFT); -#endif } curbuf = 0; /* current draw buf: 0 = front, 1 = back */ @@ -735,24 +663,24 @@ void LOCK_HARDWARE( struct intel_context *intel ) if (intel_fb && intel_fb->vblank_flags && !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && (intel_fb->vbl_waited - intel_fb->vbl_pending[curbuf]) > (1<<23)) { - drmVBlank vbl; + drmVBlank vbl; - vbl.request.type = DRM_VBLANK_ABSOLUTE; + vbl.request.type = DRM_VBLANK_ABSOLUTE; - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - vbl.request.type |= DRM_VBLANK_SECONDARY; - } + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + vbl.request.type |= DRM_VBLANK_SECONDARY; + } - vbl.request.sequence = intel_fb->vbl_pending[curbuf]; - drmWaitVBlank(intel->driFd, &vbl); - intel_fb->vbl_waited = vbl.reply.sequence; + vbl.request.sequence = intel_fb->vbl_pending[curbuf]; + drmWaitVBlank(intel->driFd, &vbl); + intel_fb->vbl_waited = vbl.reply.sequence; } DRM_CAS(intel->driHwLock, intel->hHWContext, - (DRM_LOCK_HELD|intel->hHWContext), __ret); + (DRM_LOCK_HELD|intel->hHWContext), __ret); if (__ret) - intelContendedLock( intel, 0 ); + intelContendedLock( intel, 0 ); if (INTEL_DEBUG & DEBUG_LOCK) _mesa_printf("%s - locked\n", __progname); -- cgit v1.2.3 From ef5d99e08386b19bb7e693ebd4970fe2412783de Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 16:05:08 -0600 Subject: Obsolete --- .../dri/intel_winsys/intel_read_draw_buffer.c | 243 --------------------- 1 file changed, 243 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_read_draw_buffer.c (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_read_draw_buffer.c b/src/mesa/drivers/dri/intel_winsys/intel_read_draw_buffer.c deleted file mode 100644 index 0ac246426e..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_read_draw_buffer.c +++ /dev/null @@ -1,243 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_blit.h" -#include "intel_buffers.h" -#include "intel_depthstencil.h" -#include "intel_fbo.h" -#include "intel_batchbuffer.h" -#include "intel_reg.h" -#include "context.h" -#include "utils.h" -#include "drirenderbuffer.h" -#include "framebuffer.h" -#include "vblank.h" - -#include "pipe/p_context.h" - - - -/** - * Update the following fields for rendering: - * intel->numClipRects - * intel->pClipRects - */ -static void -intelSetRenderbufferClipRects(struct intel_context *intel) -{ - /* zero-sized buffers might be legal? */ - assert(intel->ctx.DrawBuffer->Width > 0); - assert(intel->ctx.DrawBuffer->Height > 0); - intel->fboRect.x1 = 0; - intel->fboRect.y1 = 0; - intel->fboRect.x2 = intel->ctx.DrawBuffer->Width; - intel->fboRect.y2 = intel->ctx.DrawBuffer->Height; - intel->numClipRects = 1; - intel->pClipRects = &intel->fboRect; -} - - - -/** - * Update the hardware state for drawing into a window or framebuffer object. - * - * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other - * places within the driver. - * - * Basically, this needs to be called any time the current framebuffer - * changes, the renderbuffers change, or we need to draw into different - * color buffers. - */ -void -intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) -{ - struct intel_context *intel = intel_context(ctx); - struct pipe_region *colorRegion, *depthRegion = NULL; - struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; - - if (!fb) { - /* this can happen during the initial context initialization */ - return; - } - - /* Do this here, not core Mesa, since this function is called from - * many places within the driver. - */ - if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { - /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */ - _mesa_update_framebuffer(ctx); - /* this updates the DrawBuffer's Width/Height if it's a FBO */ - _mesa_update_draw_buffer_bounds(ctx); - } - - if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { - /* this may occur when we're called by glBindFrameBuffer() during - * the process of someone setting up renderbuffers, etc. - */ - /*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/ - return; - } - - if (fb->Name) - intel_validate_paired_depth_stencil(ctx, fb); - - /* - * How many color buffers are we drawing into? - */ - if (fb->_NumColorDrawBuffers[0] != 1) { - /* writing to 0 or 2 or 4 color buffers */ - /*_mesa_debug(ctx, "Software rendering\n");*/ - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); - } - else { - /* draw to exactly one color buffer */ - /*_mesa_debug(ctx, "Hardware rendering\n");*/ - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); - } - - /* - * Get the intel_renderbuffer for the colorbuffer we're drawing into. - * And set up cliprects. - */ - { - struct intel_renderbuffer *irb; - intelSetRenderbufferClipRects(intel); - irb = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); - colorRegion = (irb && irb->region) ? irb->region : NULL; - } - - /* Update culling direction which changes depending on the - * orientation of the buffer: - */ - if (ctx->Driver.FrontFace) - ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace); - else - ctx->NewState |= _NEW_POLYGON; - - if (!colorRegion) { - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); - } - else { - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); - } - - /*** - *** Get depth buffer region and check if we need a software fallback. - *** Note that the depth buffer is usually a DEPTH_STENCIL buffer. - ***/ - if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) { - irbDepth = intel_renderbuffer(fb->_DepthBuffer->Wrapped); - if (irbDepth && irbDepth->region) { - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); - depthRegion = irbDepth->region; - } - else { - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_TRUE); - depthRegion = NULL; - } - } - else { - /* not using depth buffer */ - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); - depthRegion = NULL; - } - - /*** - *** Stencil buffer - *** This can only be hardware accelerated if we're using a - *** combined DEPTH_STENCIL buffer (for now anyway). - ***/ - if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { - irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); - if (irbStencil && irbStencil->region) { - ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); - /* need to re-compute stencil hw state */ -// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); - if (!depthRegion) - depthRegion = irbStencil->region; - } - else { - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE); - } - } - else { - /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */ - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); - /* need to re-compute stencil hw state */ -// ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); - } - - - /** - ** Release old regions, reference new regions - **/ - - // intel->vtbl.set_draw_region(intel, colorRegion, depthRegion); - - /* update viewport since it depends on window size */ -// ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y, -// ctx->Viewport.Width, ctx->Viewport.Height); - - /* Update hardware scissor */ -// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, -// ctx->Scissor.Width, ctx->Scissor.Height); -} - - -static void -intelDrawBuffer(GLcontext * ctx, GLenum mode) -{ - intel_draw_buffer(ctx, ctx->DrawBuffer); -} - - -static void -intelReadBuffer(GLcontext * ctx, GLenum mode) -{ - if (ctx->ReadBuffer == ctx->DrawBuffer) { - /* This will update FBO completeness status. - * A framebuffer will be incomplete if the GL_READ_BUFFER setting - * refers to a missing renderbuffer. Calling glReadBuffer can set - * that straight and can make the drawing buffer complete. - */ - intel_draw_buffer(ctx, ctx->DrawBuffer); - } - /* Generally, functions which read pixels (glReadPixels, glCopyPixels, etc) - * reference ctx->ReadBuffer and do appropriate state checks. - */ -} - - -void -intelInitBufferFuncs(struct dd_function_table *functions) -{ - functions->DrawBuffer = intelDrawBuffer; - functions->ReadBuffer = intelReadBuffer; -} -- cgit v1.2.3 From df338be6d7c26fdef184d97fee726d32cb1c04a1 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 16:31:48 -0600 Subject: comments, assertions, etc --- src/mesa/drivers/dri/intel_winsys/intel_surface.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.c b/src/mesa/drivers/dri/intel_winsys/intel_surface.c index 3a5d52b6d7..ef2e62c809 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.c @@ -217,7 +217,11 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, - +/** + * Create a new surface of the specified format. + * This also plugs in the quad/tile read/write functions that + * will be used to access the buffer for rendering/texturing/etc. + */ struct pipe_surface * intel_new_surface(struct pipe_context *pipe, GLuint pipeFormat) { @@ -225,7 +229,7 @@ intel_new_surface(struct pipe_context *pipe, GLuint pipeFormat) if (!sps) return NULL; - sps->surface.width = 0; /* set in intel_alloc_renderbuffer_storage() */ + sps->surface.width = 0; /* set by renderbuffer::AllocStorage() */ sps->surface.height = 0; sps->surface.refcount = 1; sps->surface.format = pipeFormat; @@ -246,13 +250,17 @@ intel_new_surface(struct pipe_context *pipe, GLuint pipeFormat) sps->read_quad_stencil = read_quad_stencil; sps->write_quad_stencil = write_quad_stencil; break; + default: + assert(0); } return &sps->surface; } - +/** + * Return list of surface formats supported by this driver. + */ const GLuint * intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats) { -- cgit v1.2.3 From 3bf25e5516c3c2688d5e1b3b1e120c970b8076c2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 16:06:16 +0100 Subject: Remove references to fallbacks from this directory --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 1 - src/mesa/drivers/dri/intel_winsys/intel_context.h | 3 --- 2 files changed, 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 4ba8394f88..054f2c8421 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -216,7 +216,6 @@ static const struct dri_debug_control debug_control[] = { {"ioctl", DEBUG_IOCTL}, {"blit", DEBUG_BLIT}, {"mip", DEBUG_MIPTREE}, - {"fall", DEBUG_FALLBACKS}, {"verb", DEBUG_VERBOSE}, {"bat", DEBUG_BATCH}, {"pix", DEBUG_PIXEL}, diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 0e040f7922..e84f53d1dc 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -121,7 +121,6 @@ extern int INTEL_DEBUG; #define DEBUG_IOCTL 0x4 #define DEBUG_BLIT 0x8 #define DEBUG_MIPTREE 0x10 -#define DEBUG_FALLBACKS 0x20 #define DEBUG_VERBOSE 0x40 #define DEBUG_BATCH 0x80 #define DEBUG_PIXEL 0x100 @@ -156,8 +155,6 @@ extern void intelGetLock(struct intel_context *intel, GLuint flags); extern void intelFinish(GLcontext * ctx); extern void intelFlush(GLcontext * ctx); -#define FALLBACK( ctx, bit, mode ) - /*====================================================================== * Inline conversion functions. * These are better-typed than the macros used previously: -- cgit v1.2.3 From e23966f3de81342f7bd11f1e79c04022b5f0193f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 16:43:44 +0100 Subject: Some basic state - blend and ztest sort of work. --- src/mesa/pipe/i915simple/Makefile | 1 + src/mesa/pipe/i915simple/i915_context.h | 75 ++++++++++++++++++++++++++- src/mesa/pipe/i915simple/i915_prim_emit.c | 2 +- src/mesa/pipe/i915simple/i915_state.h | 1 + src/mesa/pipe/i915simple/i915_state_derived.c | 7 ++- src/mesa/pipe/i915simple/i915_state_emit.c | 10 ++-- 6 files changed, 88 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 93d9c6cf01..1adffbbfe1 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -15,6 +15,7 @@ DRIVER_SOURCES = \ i915_debug_fp.c \ i915_regions.c \ i915_state.c \ + i915_state_immediate.c \ i915_state_derived.c \ i915_state_emit.c \ i915_state_fragprog.c \ diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index c298f0f581..0f0728cde0 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -32,6 +32,63 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" + + +#define I915_TEX_UNITS 8 + +#define I915_DYNAMIC_MODES4 0 +#define I915_DYNAMIC_DEPTHSCALE_0 1 /* just the header */ +#define I915_DYNAMIC_DEPTHSCALE_1 2 +#define I915_DYNAMIC_IAB 3 +#define I915_DYNAMIC_BC_0 4 /* just the header */ +#define I915_DYNAMIC_BC_1 5 +#define I915_DYNAMIC_BFO_0 6 +#define I915_DYNAMIC_BFO_1 7 +#define I915_DYNAMIC_STP_0 8 +#define I915_DYNAMIC_STP_1 9 +#define I915_DYNAMIC_SC_0 10 +#define I915_DYNAMIC_SC_1 11 +#define I915_DYNAMIC_SC_2 12 +#define I915_DYNAMIC_SC_3 13 +#define I915_MAX_DYNAMIC 14 + + +#define I915_IMMEDIATE_S0 0 +#define I915_IMMEDIATE_S1 1 +#define I915_IMMEDIATE_S2 2 +#define I915_IMMEDIATE_S3 3 +#define I915_IMMEDIATE_S4 4 +#define I915_IMMEDIATE_S5 5 +#define I915_IMMEDIATE_S6 6 +#define I915_IMMEDIATE_S7 7 +#define I915_MAX_IMMEDIATE 8 + +/* These must mach the order of LI0_STATE_* bits, as they will be used + * to generate hardware packets: + */ +#define I915_CACHE_STATIC 0 +#define I915_CACHE_DYNAMIC 1 /* handled specially */ +#define I915_CACHE_SAMPLER 2 +#define I915_CACHE_MAP 3 +#define I915_CACHE_PROGRAM 4 +#define I915_CACHE_CONSTANTS 5 +#define I915_MAX_CACHE 6 + + +struct i915_cache_context; + +/* Use to calculate differences between state emitted to hardware and + * current driver-calculated state. + */ +struct i915_state +{ + GLuint immediate[I915_MAX_IMMEDIATE]; + + GLuint id; /* track lost context events */ +}; + + + struct i915_context { struct pipe_context pipe; @@ -56,13 +113,19 @@ struct i915_context struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; GLuint dirty; - GLuint hw_dirty; GLuint *batch_start; + struct i915_state current; + GLuint hardware_dirty; + + + struct pipe_scissor_state cliprect; }; +/* A flag for each state_tracker state object: + */ #define I915_NEW_VIEWPORT 0x1 #define I915_NEW_SETUP 0x2 #define I915_NEW_FS 0x4 @@ -77,6 +140,16 @@ struct i915_context #define I915_NEW_TEXTURE 0x800 #define I915_NEW_STENCIL 0x1000 +/* Driver's internally generated state flags: + */ +#define I915_NEW_VERTEX_FORMAT 0x10000 + + +/* Dirty flags for hardware emit + */ +#define I915_HW_INDIRECT (1<<0) +#define I915_HW_IMMEDIATE (1<<1) + /*********************************************************************** * i915_prim_emit.c: diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index a1e0e85196..708fc41d31 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -123,7 +123,7 @@ emit_prim( struct draw_stage *stage, if (i915->dirty) i915_update_derived( i915 ); - if (i915->hw_dirty) + if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); ptr = winsys->batch_start( winsys, nr * vertex_size, 0 ); diff --git a/src/mesa/pipe/i915simple/i915_state.h b/src/mesa/pipe/i915simple/i915_state.h index c9b9f15d39..e26e8f6b0b 100644 --- a/src/mesa/pipe/i915simple/i915_state.h +++ b/src/mesa/pipe/i915simple/i915_state.h @@ -33,6 +33,7 @@ struct i915_context; +void i915_update_immediate( struct i915_context *i915 ); void i915_update_derived( struct i915_context *i915 ); void i915_emit_hardware_state( struct i915_context *i915 ); diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index f1f798b506..32b8b6c808 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -183,7 +183,12 @@ void i915_update_derived( struct i915_context *i915 ) compute_cliprect(i915); if (i915->dirty) - i915->hw_dirty = 1; + i915_update_immediate( i915 ); + + /* HW emit currently references framebuffer state directly: + */ + if (i915->dirty & I915_NEW_FRAMEBUFFER) + i915->hardware_dirty = 1; i915->dirty = 0; } diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 3b2ab4775b..1a50470fbd 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -151,10 +151,10 @@ i915_emit_hardware_state(struct i915_context *i915 ) I1_LOAD_S(6) | (3)); - OUT_BATCH(0xffffffff); - OUT_BATCH(0x00902440); // OUT_BATCH(S4_VFMT_XYZ | S4_VFMT_COLOR); - OUT_BATCH(0x00000002); - OUT_BATCH(0x00020216); // OUT_BATCH( S6_COLOR_WRITE_ENABLE | (2 << S6_TRISTRIP_PV_SHIFT)); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S2]); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S4]); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S5]); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S6]); } { @@ -264,6 +264,6 @@ i915_emit_hardware_state(struct i915_context *i915 ) } - i915->hw_dirty = 0; + i915->hardware_dirty = 0; } -- cgit v1.2.3 From debb4884e3bdf1e632ff79e728a59c1920cc2d15 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 16:45:20 +0100 Subject: Missing files --- src/mesa/pipe/i915simple/i915_state_immediate.c | 285 ++++++++++++++++++++++++ src/mesa/pipe/i915simple/i915_state_inlines.h | 230 +++++++++++++++++++ 2 files changed, 515 insertions(+) create mode 100644 src/mesa/pipe/i915simple/i915_state_immediate.c create mode 100644 src/mesa/pipe/i915simple/i915_state_inlines.h (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c new file mode 100644 index 0000000000..8b93ca7f62 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -0,0 +1,285 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "macros.h" + +#include "i915_state_inlines.h" +#include "i915_context.h" +#include "i915_state.h" +#include "i915_reg.h" + + +/* All state expressable with the LOAD_STATE_IMMEDIATE_1 packet. + * Would like to opportunistically recombine all these fragments into + * a single packet containing only what has changed, but for now emit + * as multiple packets. + */ + + +struct i915_tracked_state { + unsigned dirty; + void (*update)( struct i915_context * ); +}; + + + +/*********************************************************************** + * S4: Vertex format, rasterization state + */ +static void upload_S2S4(struct i915_context *i915) +{ + GLuint LIS2, LIS4; + + /* I915_NEW_VERTEX_FORMAT */ + LIS2 = 0xffffffff; + LIS4 = (S4_VFMT_XYZ | S4_VFMT_COLOR); + + + /* I915_NEW_SETUP */ + switch (i915->setup.cull_mode) { + case PIPE_WINDING_NONE: + LIS4 |= S4_CULLMODE_NONE; + break; + case PIPE_WINDING_CW: + LIS4 |= S4_CULLMODE_CW; + break; + case PIPE_WINDING_CCW: + LIS4 |= S4_CULLMODE_CCW; + break; + case PIPE_WINDING_BOTH: + LIS4 |= S4_CULLMODE_BOTH; + break; + } + + /* I915_NEW_SETUP */ + { + int line_width = CLAMP((int)(i915->setup.line_width * 2), 1, 0xf); + + LIS4 |= line_width << S4_LINE_WIDTH_SHIFT; + + if (i915->setup.line_smooth) + LIS4 |= S4_LINE_ANTIALIAS_ENABLE; + } + + /* I915_NEW_SETUP */ + { + GLint point_size = CLAMP((int) i915->setup.point_size, 1, 0xff); + + LIS4 |= point_size << S4_POINT_WIDTH_SHIFT; + } + + /* I915_NEW_SETUP */ + if (i915->setup.flatshade) { + LIS4 |= (S4_FLATSHADE_ALPHA | + S4_FLATSHADE_COLOR | + S4_FLATSHADE_SPECULAR); + } + + + if (LIS2 != i915->current.immediate[I915_IMMEDIATE_S2] || + LIS4 != i915->current.immediate[I915_IMMEDIATE_S4]) { + + i915->current.immediate[I915_IMMEDIATE_S2] = LIS2; + i915->current.immediate[I915_IMMEDIATE_S4] = LIS4; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + + +const struct i915_tracked_state i915_upload_S2S4 = { + .dirty = I915_NEW_SETUP | I915_NEW_VERTEX_FORMAT, + .update = upload_S2S4 +}; + + + +/*********************************************************************** + * + */ +static void upload_S5( struct i915_context *i915 ) +{ + GLuint LIS5 = 0; + + /* I915_NEW_STENCIL */ + if (i915->stencil.front_enabled) { + GLint test = i915_translate_compare_func(i915->stencil.front_func); + GLint fop = i915_translate_stencil_op(i915->stencil.front_fail_op); + GLint dfop = i915_translate_stencil_op(i915->stencil.front_zfail_op); + GLint dpop = i915_translate_stencil_op(i915->stencil.front_zpass_op); + GLint ref = i915->stencil.ref_value[0] & 0xff; + + LIS5 |= (S5_STENCIL_TEST_ENABLE | + S5_STENCIL_WRITE_ENABLE | + (ref << S5_STENCIL_REF_SHIFT) | + (test << S5_STENCIL_TEST_FUNC_SHIFT) | + (fop << S5_STENCIL_FAIL_SHIFT) | + (dfop << S5_STENCIL_PASS_Z_FAIL_SHIFT) | + (dpop << S5_STENCIL_PASS_Z_PASS_SHIFT)); + } + + /* I915_NEW_BLEND */ + if (i915->blend.logicop_enable) + LIS5 |= S5_LOGICOP_ENABLE; + + if (i915->blend.dither) + LIS5 |= S5_COLOR_DITHER_ENABLE; + + if ((i915->blend.colormask & PIPE_MASK_R) == 0) + LIS5 |= S5_WRITEDISABLE_RED; + + if ((i915->blend.colormask & PIPE_MASK_G) == 0) + LIS5 |= S5_WRITEDISABLE_GREEN; + + if ((i915->blend.colormask & PIPE_MASK_B) == 0) + LIS5 |= S5_WRITEDISABLE_BLUE; + + if ((i915->blend.colormask & PIPE_MASK_A) == 0) + LIS5 |= S5_WRITEDISABLE_ALPHA; + + +#if 0 + /* I915_NEW_SETUP */ + if (i915->state.Polygon->OffsetFill) { + LIS5 |= S5_GLOBAL_DEPTH_OFFSET_ENABLE; + } +#endif + + + if (LIS5 != i915->current.immediate[I915_IMMEDIATE_S5]) { + i915->current.immediate[I915_IMMEDIATE_S5] = LIS5; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + +const struct i915_tracked_state i915_upload_S5 = { + .dirty = (I915_NEW_STENCIL | I915_NEW_BLEND | I915_NEW_SETUP), + .update = upload_S5 +}; + + +/*********************************************************************** + */ +static void upload_S6( struct i915_context *i915 ) +{ + GLuint LIS6 = (S6_COLOR_WRITE_ENABLE | + (2 << S6_TRISTRIP_PV_SHIFT)); + + /* I915_NEW_ALPHA_TEST + */ + if (i915->alpha_test.enabled) { + int test = i915_translate_compare_func(i915->alpha_test.func); + GLubyte refByte; + + CLAMPED_FLOAT_TO_UBYTE(refByte, i915->alpha_test.ref); + + + LIS6 |= (S6_ALPHA_TEST_ENABLE | + (test << S6_ALPHA_TEST_FUNC_SHIFT) | + (((unsigned) refByte) << S6_ALPHA_REF_SHIFT)); + } + + /* I915_NEW_BLEND + */ + if (i915->blend.blend_enable) + { + GLuint funcRGB = i915->blend.rgb_func; + GLuint srcRGB = i915->blend.rgb_src_factor; + GLuint dstRGB = i915->blend.rgb_dst_factor; + + LIS6 |= (S6_CBUF_BLEND_ENABLE | + SRC_BLND_FACT(i915_translate_blend_factor(srcRGB)) | + DST_BLND_FACT(i915_translate_blend_factor(dstRGB)) | + (i915_translate_blend_func(funcRGB) << S6_CBUF_BLEND_FUNC_SHIFT)); + } + + /* I915_NEW_DEPTH + */ + if (i915->depth_test.enabled) { + GLint func = i915_translate_compare_func(i915->depth_test.func); + + LIS6 |= (S6_DEPTH_TEST_ENABLE | + (func << S6_DEPTH_TEST_FUNC_SHIFT)); + + if (i915->depth_test.writemask) + LIS6 |= S6_DEPTH_WRITE_ENABLE; + } + + if (LIS6 != i915->current.immediate[I915_IMMEDIATE_S6]) { + i915->current.immediate[I915_IMMEDIATE_S6] = LIS6; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + +const struct i915_tracked_state i915_upload_S6 = { + .dirty = I915_NEW_ALPHA_TEST | I915_NEW_BLEND | I915_NEW_DEPTH_TEST, + .update = upload_S6 +}; + + +/*********************************************************************** + */ +static void upload_S7( struct i915_context *i915 ) +{ + GLfloat LIS7; + + /* I915_NEW_SETUP + */ + LIS7 = i915->setup.offset_units; /* probably incorrect */ + + if (LIS7 != i915->current.immediate[I915_IMMEDIATE_S7]) { + i915->current.immediate[I915_IMMEDIATE_S7] = LIS7; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + +const struct i915_tracked_state i915_upload_S7 = { + .dirty = I915_NEW_SETUP, + .update = upload_S7 +}; + + +static const struct i915_tracked_state *atoms[] = { + &i915_upload_S2S4, + &i915_upload_S5, + &i915_upload_S6, + &i915_upload_S7 +}; + +/* + */ +void i915_update_immediate( struct i915_context *i915 ) +{ + int i; + + for (i = 0; i < Elements(atoms); i++) + if (i915->dirty & atoms[i]->dirty) + atoms[i]->update( i915 ); +} diff --git a/src/mesa/pipe/i915simple/i915_state_inlines.h b/src/mesa/pipe/i915simple/i915_state_inlines.h new file mode 100644 index 0000000000..28aff0ebd4 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state_inlines.h @@ -0,0 +1,230 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_STATE_INLINES_H +#define I915_STATE_INLINES_H + +#include "p_defines.h" +#include "i915_reg.h" + + +static INLINE unsigned +i915_translate_compare_func(unsigned func) +{ + switch (func) { + case PIPE_FUNC_NEVER: + return COMPAREFUNC_NEVER; + case PIPE_FUNC_LESS: + return COMPAREFUNC_LESS; + case PIPE_FUNC_LEQUAL: + return COMPAREFUNC_LEQUAL; + case PIPE_FUNC_GREATER: + return COMPAREFUNC_GREATER; + case PIPE_FUNC_GEQUAL: + return COMPAREFUNC_GEQUAL; + case PIPE_FUNC_NOTEQUAL: + return COMPAREFUNC_NOTEQUAL; + case PIPE_FUNC_EQUAL: + return COMPAREFUNC_EQUAL; + case PIPE_FUNC_ALWAYS: + return COMPAREFUNC_ALWAYS; + default: + return COMPAREFUNC_ALWAYS; + } +} + +static INLINE unsigned +i915_translate_stencil_op(unsigned op) +{ + switch (op) { + case PIPE_STENCIL_OP_KEEP: + return STENCILOP_KEEP; + case PIPE_STENCIL_OP_ZERO: + return STENCILOP_ZERO; + case PIPE_STENCIL_OP_REPLACE: + return STENCILOP_REPLACE; + case PIPE_STENCIL_OP_INCR: + return STENCILOP_INCRSAT; + case PIPE_STENCIL_OP_DECR: + return STENCILOP_DECRSAT; + case PIPE_STENCIL_OP_INCR_WRAP: + return STENCILOP_INCR; + case PIPE_STENCIL_OP_DECR_WRAP: + return STENCILOP_DECR; + case PIPE_STENCIL_OP_INVERT: + return STENCILOP_INVERT; + default: + return STENCILOP_ZERO; + } +} + +static INLINE unsigned +i915_translate_blend_factor(unsigned factor) +{ + switch (factor) { + case PIPE_BLENDFACTOR_ZERO: + return BLENDFACT_ZERO; + case PIPE_BLENDFACTOR_SRC_ALPHA: + return BLENDFACT_SRC_ALPHA; + case PIPE_BLENDFACTOR_ONE: + return BLENDFACT_ONE; + case PIPE_BLENDFACTOR_SRC_COLOR: + return BLENDFACT_SRC_COLR; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + return BLENDFACT_INV_SRC_COLR; + case PIPE_BLENDFACTOR_DST_COLOR: + return BLENDFACT_DST_COLR; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + return BLENDFACT_INV_DST_COLR; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + return BLENDFACT_INV_SRC_ALPHA; + case PIPE_BLENDFACTOR_DST_ALPHA: + return BLENDFACT_DST_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + return BLENDFACT_INV_DST_ALPHA; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + return BLENDFACT_SRC_ALPHA_SATURATE; + case PIPE_BLENDFACTOR_CONST_COLOR: + return BLENDFACT_CONST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + return BLENDFACT_INV_CONST_COLOR; + case PIPE_BLENDFACTOR_CONST_ALPHA: + return BLENDFACT_CONST_ALPHA; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + return BLENDFACT_INV_CONST_ALPHA; + default: + return BLENDFACT_ZERO; + } +} + +static INLINE unsigned +i915_translate_blend_func(unsigned mode) +{ + switch (mode) { + case PIPE_BLEND_ADD: + return BLENDFUNC_ADD; + case PIPE_BLEND_MIN: + return BLENDFUNC_MIN; + case PIPE_BLEND_MAX: + return BLENDFUNC_MAX; + case PIPE_BLEND_SUBTRACT: + return BLENDFUNC_SUBTRACT; + case PIPE_BLEND_REVERSE_SUBTRACT: + return BLENDFUNC_REVERSE_SUBTRACT; + default: + return 0; + } +} + + +static INLINE unsigned +i915_translate_logic_op(unsigned opcode) +{ + switch (opcode) { + case PIPE_LOGICOP_CLEAR: + return LOGICOP_CLEAR; + case PIPE_LOGICOP_AND: + return LOGICOP_AND; + case PIPE_LOGICOP_AND_REVERSE: + return LOGICOP_AND_RVRSE; + case PIPE_LOGICOP_COPY: + return LOGICOP_COPY; + case PIPE_LOGICOP_COPY_INVERTED: + return LOGICOP_COPY_INV; + case PIPE_LOGICOP_AND_INVERTED: + return LOGICOP_AND_INV; + case PIPE_LOGICOP_NOOP: + return LOGICOP_NOOP; + case PIPE_LOGICOP_XOR: + return LOGICOP_XOR; + case PIPE_LOGICOP_OR: + return LOGICOP_OR; + case PIPE_LOGICOP_OR_INVERTED: + return LOGICOP_OR_INV; + case PIPE_LOGICOP_NOR: + return LOGICOP_NOR; + case PIPE_LOGICOP_EQUIV: + return LOGICOP_EQUIV; + case PIPE_LOGICOP_INVERT: + return LOGICOP_INV; + case PIPE_LOGICOP_OR_REVERSE: + return LOGICOP_OR_RVRSE; + case PIPE_LOGICOP_NAND: + return LOGICOP_NAND; + case PIPE_LOGICOP_SET: + return LOGICOP_SET; + default: + return LOGICOP_SET; + } +} + + + +static INLINE GLboolean i915_validate_vertices( GLuint hw_prim, GLuint nr ) +{ + GLboolean ok; + + switch (hw_prim) { + case PRIM3D_POINTLIST: + ok = (nr >= 1); + assert(ok); + break; + case PRIM3D_LINELIST: + ok = (nr >= 2) && (nr % 2) == 0; + assert(ok); + break; + case PRIM3D_LINESTRIP: + ok = (nr >= 2); + assert(ok); + break; + case PRIM3D_TRILIST: + ok = (nr >= 3) && (nr % 3) == 0; + assert(ok); + break; + case PRIM3D_TRISTRIP: + ok = (nr >= 3); + assert(ok); + break; + case PRIM3D_TRIFAN: + ok = (nr >= 3); + assert(ok); + break; + case PRIM3D_POLY: + ok = (nr >= 3); + assert(ok); + break; + default: + assert(0); + ok = 0; + break; + } + + return ok; +} + + +#endif -- cgit v1.2.3 From 584c0762010a467e391d7dcdb48aae4d7809f6b5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 17:58:28 -0600 Subject: vertex colors are ARGB --- src/mesa/pipe/i915simple/i915_prim_emit.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index 708fc41d31..18918459a8 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -59,7 +59,7 @@ static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) return (struct setup_stage *)stage; } -static inline unsigned pack_ub4( unsigned char b0, +static INLINE unsigned pack_ub4( unsigned char b0, unsigned char b1, unsigned char b2, unsigned char b3 ) @@ -70,7 +70,7 @@ static inline unsigned pack_ub4( unsigned char b0, (((unsigned int)b3) << 24)); } -static inline unsigned fui( float f ) +static INLINE unsigned fui( float f ) { union { float f; @@ -81,7 +81,7 @@ static inline unsigned fui( float f ) return fi.ui; } -static inline unsigned char float_to_ubyte( float f ) +static INLINE unsigned char float_to_ubyte( float f ) { unsigned char ub; UNCLAMPED_FLOAT_TO_UBYTE(ub, f); @@ -91,7 +91,7 @@ static inline unsigned char float_to_ubyte( float f ) /* Hardcoded vertex format: xyz/rgba */ -static inline void +static INLINE void emit_hw_vertex( struct i915_context *i915, struct vertex_header *vertex ) { @@ -99,16 +99,17 @@ emit_hw_vertex( struct i915_context *i915, OUT_BATCH( fui(vertex->data[0][1]) ); OUT_BATCH( fui(vertex->data[0][2]) ); - OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[1][0] ), + /* colors are ARGB (MSB to LSB) */ + OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[1][2] ), float_to_ubyte( vertex->data[1][1] ), - float_to_ubyte( vertex->data[1][2] ), + float_to_ubyte( vertex->data[1][0] ), float_to_ubyte( vertex->data[1][3] )) ); } -static inline void +static INLINE void emit_prim( struct draw_stage *stage, struct prim_header *prim, unsigned hwprim, -- cgit v1.2.3 From f1637e89ded7312a97dc0a8fdac74886d70766c0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 17:59:13 -0600 Subject: invert Y axis with the viewport transformation to put things right side up --- src/mesa/pipe/i915simple/i915_state.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index ebc55eb634..a176165b74 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -189,11 +189,24 @@ static void i915_set_viewport_state( struct pipe_context *pipe, const struct pipe_viewport_state *viewport ) { struct i915_context *i915 = i915_context(pipe); + float sy, ty; i915->viewport = *viewport; /* struct copy */ + /* Negate Y scale to flip image vertically. + * The NDC Y coords prior to viewport transformation are in the range + * [y=-1=bottom, y=1=top] + * Intel window coords are in the range [y=0=top, y=H-1=bottom] where H + * is the window height. + * Use the viewport transformation to invert Y. + */ + sy = viewport->scale[1]; + ty = viewport->translate[1]; + i915->viewport.scale[1] = -sy; + i915->viewport.translate[1] = i915->framebuffer.cbufs[0]->height - ty; + /* pass the viewport info to the draw module */ - draw_set_viewport_state(i915->draw, viewport); + draw_set_viewport_state(i915->draw, &i915->viewport); /* Using tnl/ and vf/ modules is temporary while getting started. * Full pipe will have vertex shader, vertex fetch of its own. -- cgit v1.2.3 From b3efc75f1a3496cbd0b93b5eff497a92fa31dc3b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 17:40:18 +0100 Subject: Remove unused file --- src/mesa/pipe/i915simple/i915_regions.h | 141 -------------------------------- 1 file changed, 141 deletions(-) delete mode 100644 src/mesa/pipe/i915simple/i915_regions.h (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_regions.h b/src/mesa/pipe/i915simple/i915_regions.h deleted file mode 100644 index d938c107a4..0000000000 --- a/src/mesa/pipe/i915simple/i915_regions.h +++ /dev/null @@ -1,141 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_REGIONS_H -#define INTEL_REGIONS_H - -#include "mtypes.h" -#include "intel_screen.h" - -struct intel_context; -struct intel_buffer_object; - -/** - * A layer on top of the bufmgr buffers that adds a few useful things: - * - * - Refcounting for local buffer references. - * - Refcounting for buffer maps - * - Buffer dimensions - pitch and height. - * - Blitter commands for copying 2D regions between buffers. (really???) - */ -struct intel_region -{ - struct _DriBufferObject *buffer; /**< buffer manager's buffer ID */ - GLuint refcount; /**< Reference count for region */ - GLuint cpp; /**< bytes per pixel */ - GLuint pitch; /**< in pixels */ - GLuint height; /**< in pixels */ - GLubyte *map; /**< only non-NULL when region is actually mapped */ - GLuint map_refcount; /**< Reference count for mapping */ - - GLuint draw_offset; /**< Offset of drawing address within the region */ - - struct intel_buffer_object *pbo; /* zero-copy uploads */ -}; - - -/* Allocate a refcounted region. Pointers to regions should only be - * copied by calling intel_reference_region(). - */ -struct intel_region *intel_region_alloc(intelScreenPrivate *intelScreen, - GLuint cpp, - GLuint pitch, GLuint height); - -void intel_region_reference(struct intel_region **dst, - struct intel_region *src); - -void intel_region_release(struct intel_region **ib); - -extern struct intel_region -*intel_region_create_static(intelScreenPrivate *intelScreen, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, - GLuint pitch, GLuint height); -extern void -intel_region_update_static(intelScreenPrivate *intelScreen, - struct intel_region *region, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height); - - -void intel_region_idle(intelScreenPrivate *intelScreen, - struct intel_region *ib); - -/* Map/unmap regions. This is refcounted also: - */ -GLubyte *intel_region_map(intelScreenPrivate *intelScreen, - struct intel_region *ib); - -void intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *ib); - - -/* Upload data to a rectangular sub-region - */ -void intel_region_data(intelScreenPrivate *intelScreen, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - const void *src, GLuint src_stride, - GLuint srcx, GLuint srcy, GLuint width, GLuint height); - -/* Copy rectangular sub-regions - */ -void intel_region_copy(intelScreenPrivate *intelScreen, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - struct intel_region *src, - GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height); - -/* Fill a rectangular sub-region - */ -void intel_region_fill(intelScreenPrivate *intelScreen, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - GLuint width, GLuint height, GLuint color); - -/* Helpers for zerocopy uploads, particularly texture image uploads: - */ -void intel_region_attach_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region, - struct intel_buffer_object *pbo); -void intel_region_release_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region); -void intel_region_cow(intelScreenPrivate *intelScreen, - struct intel_region *region); - -struct _DriBufferObject *intel_region_buffer(intelScreenPrivate *intelScreen, - struct intel_region *region, - GLuint flag); - -#endif -- cgit v1.2.3 From 7c9a1357beabd6d0df88142d31cd5d77a012e85a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 18:07:42 +0100 Subject: Add more i915 state packets. These correspond to the dynamic indirect state, though they are currently just pushed straight to the batch buffer. --- src/mesa/pipe/i915simple/Makefile | 1 + src/mesa/pipe/i915simple/i915_batch.h | 1 + src/mesa/pipe/i915simple/i915_context.c | 4 +- src/mesa/pipe/i915simple/i915_context.h | 19 +- src/mesa/pipe/i915simple/i915_prim_emit.c | 13 +- src/mesa/pipe/i915simple/i915_state.h | 7 + src/mesa/pipe/i915simple/i915_state_derived.c | 44 +-- src/mesa/pipe/i915simple/i915_state_dynamic.c | 429 ++++++++++++++++++++++++ src/mesa/pipe/i915simple/i915_state_emit.c | 160 ++++----- src/mesa/pipe/i915simple/i915_state_immediate.c | 5 - 10 files changed, 527 insertions(+), 156 deletions(-) create mode 100644 src/mesa/pipe/i915simple/i915_state_dynamic.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 1adffbbfe1..96e16a948f 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -16,6 +16,7 @@ DRIVER_SOURCES = \ i915_regions.c \ i915_state.c \ i915_state_immediate.c \ + i915_state_dynamic.c \ i915_state_derived.c \ i915_state_emit.c \ i915_state_fragprog.c \ diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h index 8a35ff6a3a..b11bbeedbd 100644 --- a/src/mesa/pipe/i915simple/i915_batch.h +++ b/src/mesa/pipe/i915simple/i915_batch.h @@ -48,6 +48,7 @@ i915_dump_batchbuffer( i915, i915->batch_start, BEGIN_BATCH(0, 0) ); \ i915->winsys->batch_flush( i915->winsys ); \ i915->batch_start = BEGIN_BATCH(0, 0); \ + i915->hardware_dirty = ~0; \ } while (0) #endif diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 3e8d1fbb33..eb7f3804d3 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -173,11 +173,13 @@ struct pipe_context *i915_create( struct i915_winsys *winsys ) i915_init_flush_functions(i915); + i915->dirty = ~0; + i915->hardware_dirty = ~0; + /* Batch stream debugging is a bit hacked up at the moment: */ i915->batch_start = winsys->batch_start( winsys, 0, 0 ); - /* * XXX we could plug GL selection/feedback into the drawing pipeline * by specifying a different setup/render stage. diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 0f0728cde0..c9fd2e9a3f 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -46,10 +46,10 @@ #define I915_DYNAMIC_BFO_1 7 #define I915_DYNAMIC_STP_0 8 #define I915_DYNAMIC_STP_1 9 -#define I915_DYNAMIC_SC_0 10 -#define I915_DYNAMIC_SC_1 11 -#define I915_DYNAMIC_SC_2 12 -#define I915_DYNAMIC_SC_3 13 +#define I915_DYNAMIC_SC_ENA_0 10 +#define I915_DYNAMIC_SC_RECT_0 11 +#define I915_DYNAMIC_SC_RECT_1 12 +#define I915_DYNAMIC_SC_RECT_2 13 #define I915_MAX_DYNAMIC 14 @@ -83,6 +83,7 @@ struct i915_cache_context; struct i915_state { GLuint immediate[I915_MAX_IMMEDIATE]; + GLuint dynamic[I915_MAX_DYNAMIC]; GLuint id; /* track lost context events */ }; @@ -147,8 +148,14 @@ struct i915_context /* Dirty flags for hardware emit */ -#define I915_HW_INDIRECT (1<<0) -#define I915_HW_IMMEDIATE (1<<1) +#define I915_HW_STATIC (1<i915; - struct i915_winsys *winsys = i915->winsys; unsigned vertex_size = 4 * sizeof(int); unsigned *ptr; unsigned i; @@ -127,10 +126,16 @@ emit_prim( struct draw_stage *stage, if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); - ptr = winsys->batch_start( winsys, nr * vertex_size, 0 ); + ptr = BEGIN_BATCH( nr * vertex_size, 0 ); if (ptr == 0) { - winsys->batch_flush( winsys ); - ptr = winsys->batch_start( winsys, nr * vertex_size, 0 ); + FLUSH_BATCH(); + + /* Make sure state is re-emitted after a flush: + */ + i915_update_derived( i915 ); + i915_emit_hardware_state( i915 ); + + ptr = BEGIN_BATCH( nr * vertex_size, 0 ); if (ptr == 0) { assert(0); return; diff --git a/src/mesa/pipe/i915simple/i915_state.h b/src/mesa/pipe/i915simple/i915_state.h index e26e8f6b0b..794a251ad8 100644 --- a/src/mesa/pipe/i915simple/i915_state.h +++ b/src/mesa/pipe/i915simple/i915_state.h @@ -33,7 +33,14 @@ struct i915_context; + +struct i915_tracked_state { + unsigned dirty; + void (*update)( struct i915_context * ); +}; + void i915_update_immediate( struct i915_context *i915 ); +void i915_update_dynamic( struct i915_context *i915 ); void i915_update_derived( struct i915_context *i915 ); void i915_emit_hardware_state( struct i915_context *i915 ); diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 32b8b6c808..7d03ed5567 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -134,40 +134,6 @@ static void calculate_vertex_layout( struct i915_context *i915 ) } -/** - * Recompute cliprect from scissor bounds, scissor enable and surface size. - */ -static void -compute_cliprect(struct i915_context *sp) -{ - GLint surfWidth, surfHeight; - - if (sp->framebuffer.num_cbufs > 0) { - surfWidth = sp->framebuffer.cbufs[0]->width; - surfHeight = sp->framebuffer.cbufs[0]->height; - } - else { - /* no surface? */ - surfWidth = sp->scissor.maxx; - surfHeight = sp->scissor.maxy; - } - - if (sp->setup.scissor) { - /* clip to scissor rect */ - sp->cliprect.minx = MAX2(sp->scissor.minx, 0); - sp->cliprect.miny = MAX2(sp->scissor.miny, 0); - sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth); - sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight); - } - else { - /* clip to surface bounds */ - sp->cliprect.minx = 0; - sp->cliprect.miny = 0; - sp->cliprect.maxx = surfWidth; - sp->cliprect.maxy = surfHeight; - } -} - /* Hopefully this will remain quite simple, otherwise need to pull in * something like the state tracker mechanism. @@ -177,18 +143,16 @@ void i915_update_derived( struct i915_context *i915 ) if (i915->dirty & (I915_NEW_SETUP | I915_NEW_FS)) calculate_vertex_layout( i915 ); - if (i915->dirty & (I915_NEW_SCISSOR | - I915_NEW_STENCIL | - I915_NEW_FRAMEBUFFER)) - compute_cliprect(i915); - if (i915->dirty) i915_update_immediate( i915 ); + if (i915->dirty) + i915_update_dynamic( i915 ); + /* HW emit currently references framebuffer state directly: */ if (i915->dirty & I915_NEW_FRAMEBUFFER) - i915->hardware_dirty = 1; + i915->hardware_dirty |= I915_HW_STATIC; i915->dirty = 0; } diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c new file mode 100644 index 0000000000..8d5a398890 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -0,0 +1,429 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "glheader.h" +#include "context.h" +#include "macros.h" +#include "enums.h" + +#include "i915_batch.h" +#include "i915_state_inlines.h" +#include "i915_context.h" +#include "i915_reg.h" +#include "i915_state.h" + +#define FILE_DEBUG_FLAG DEBUG_STATE + +/* State that we have chosen to store in the DYNAMIC segment of the + * i915 indirect state mechanism. + * + * Can't cache these in the way we do the static state, as there is no + * start/size in the command packet, instead an 'end' value that gets + * incremented. + * + * Additionally, there seems to be a requirement to re-issue the full + * (active) state every time a 4kb boundary is crossed. + */ + +static inline void set_dynamic_indirect( struct i915_context *i915, + GLuint offset, + const GLuint *src, + GLuint dwords ) +{ + int i; + + for (i = 0; i < dwords; i++) + i915->current.dynamic[offset + i] = src[i]; + + i915->hardware_dirty |= I915_HW_DYNAMIC; +} + + +/*********************************************************************** + * Modes4: stencil masks and logicop + */ +static void upload_MODES4( struct i915_context *i915 ) +{ + GLuint modes4 = 0; + + /* I915_NEW_STENCIL */ + { + GLint testmask = i915->stencil.value_mask[0] & 0xff; + GLint writemask = i915->stencil.write_mask[0] & 0xff; + + modes4 |= (_3DSTATE_MODES_4_CMD | + ENABLE_STENCIL_TEST_MASK | + STENCIL_TEST_MASK(testmask) | + ENABLE_STENCIL_WRITE_MASK | + STENCIL_WRITE_MASK(writemask)); + } + + /* I915_NEW_BLEND */ + { + modes4 |= (_3DSTATE_MODES_4_CMD | + ENABLE_LOGIC_OP_FUNC | + LOGIC_OP_FUNC(i915_translate_logic_op(i915->blend.logicop_func))); + } + + /* Always, so that we know when state is in-active: + */ + set_dynamic_indirect( i915, + I915_DYNAMIC_MODES4, + &modes4, + 1 ); +} + +const struct i915_tracked_state i915_upload_MODES4 = { + .dirty = I915_NEW_BLEND | I915_NEW_STENCIL, + .update = upload_MODES4 +}; + + + + +/*********************************************************************** + */ + +static void upload_BFO( struct i915_context *i915 ) +{ + GLuint bf[2]; + + memset( bf, 0, sizeof(bf) ); + + /* _NEW_STENCIL + */ + if (i915->stencil.back_enabled) { + GLint test = i915_translate_compare_func(i915->stencil.back_func); + GLint fop = i915_translate_stencil_op(i915->stencil.back_fail_op); + GLint dfop = i915_translate_stencil_op(i915->stencil.back_zfail_op); + GLint dpop = i915_translate_stencil_op(i915->stencil.back_zpass_op); + GLint ref = i915->stencil.ref_value[1] & 0xff; + GLint tmask = i915->stencil.value_mask[1] & 0xff; + GLint wmask = i915->stencil.write_mask[1] & 0xff; + + bf[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | + BFO_ENABLE_STENCIL_FUNCS | + BFO_ENABLE_STENCIL_TWO_SIDE | + BFO_ENABLE_STENCIL_REF | + BFO_STENCIL_TWO_SIDE | + (ref << BFO_STENCIL_REF_SHIFT) | + (test << BFO_STENCIL_TEST_SHIFT) | + (fop << BFO_STENCIL_FAIL_SHIFT) | + (dfop << BFO_STENCIL_PASS_Z_FAIL_SHIFT) | + (dpop << BFO_STENCIL_PASS_Z_PASS_SHIFT)); + + bf[1] = (_3DSTATE_BACKFACE_STENCIL_MASKS | + BFM_ENABLE_STENCIL_TEST_MASK | + BFM_ENABLE_STENCIL_WRITE_MASK | + (tmask << BFM_STENCIL_TEST_MASK_SHIFT) | + (wmask << BFM_STENCIL_WRITE_MASK_SHIFT)); + } + else { + /* This actually disables two-side stencil: The bit set is a + * modify-enable bit to indicate we are changing the two-side + * setting. Then there is a symbolic zero to show that we are + * setting the flag to zero/off. + */ + bf[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | + BFO_ENABLE_STENCIL_TWO_SIDE | + 0); + bf[1] = 0; + } + + set_dynamic_indirect( i915, + I915_DYNAMIC_BFO_0, + &bf[0], + 2 ); +} + +const struct i915_tracked_state i915_upload_BFO = { + .dirty = I915_NEW_STENCIL, + .update = upload_BFO +}; + + +/*********************************************************************** + */ + + +static void upload_BLENDCOLOR( struct i915_context *i915 ) +{ + GLuint bc[2]; + + memset( bc, 0, sizeof(bc) ); + + /* I915_NEW_BLEND {_COLOR} + */ + { + const GLfloat *color = i915->blend_color.color; + GLubyte r, g, b, a; + + UNCLAMPED_FLOAT_TO_UBYTE(r, color[RCOMP]); + UNCLAMPED_FLOAT_TO_UBYTE(g, color[GCOMP]); + UNCLAMPED_FLOAT_TO_UBYTE(b, color[BCOMP]); + UNCLAMPED_FLOAT_TO_UBYTE(a, color[ACOMP]); + + bc[0] = (_3DSTATE_CONST_BLEND_COLOR_CMD); + bc[1] = (a << 24) | (r << 16) | (g << 8) | b; + } + + set_dynamic_indirect( i915, + I915_DYNAMIC_BC_0, + bc, + 2 ); +} + +const struct i915_tracked_state i915_upload_BLENDCOLOR = { + .dirty = I915_NEW_BLEND, + .update = upload_BLENDCOLOR +}; + +/*********************************************************************** + */ + + +static void upload_IAB( struct i915_context *i915 ) +{ + GLuint iab = 0; + + { + GLuint eqRGB = i915->blend.rgb_func; + GLuint srcRGB = i915->blend.rgb_src_factor; + GLuint dstRGB = i915->blend.rgb_dst_factor; + + GLuint eqA = i915->blend.alpha_func; + GLuint srcA = i915->blend.alpha_src_factor; + GLuint dstA = i915->blend.alpha_dst_factor; + + if (eqA == GL_MIN || eqA == GL_MAX) { + srcA = dstA = GL_ONE; + } + + if (eqRGB == GL_MIN || eqRGB == GL_MAX) { + srcRGB = dstRGB = GL_ONE; + } + + if (srcA != srcRGB || + dstA != dstRGB || + eqA != eqRGB) { + + iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | + IAB_MODIFY_ENABLE | + IAB_ENABLE | + IAB_MODIFY_FUNC | + IAB_MODIFY_SRC_FACTOR | + IAB_MODIFY_DST_FACTOR | + SRC_ABLND_FACT(i915_translate_blend_factor(srcA)) | + DST_ABLND_FACT(i915_translate_blend_factor(dstA)) | + (i915_translate_blend_func(eqA) << IAB_FUNC_SHIFT)); + } + else { + iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | + IAB_MODIFY_ENABLE | + 0); + } + } + + + set_dynamic_indirect( i915, + I915_DYNAMIC_IAB, + &iab, + 1 ); +} + +const struct i915_tracked_state i915_upload_IAB = { + .dirty = I915_NEW_BLEND, + .update = upload_IAB +}; + + +/*********************************************************************** + */ + + + +static void upload_DEPTHSCALE( struct i915_context *i915 ) +{ + union { GLfloat f; GLuint u; } ds[2]; + + memset( ds, 0, sizeof(ds) ); + + /* I915_NEW_SETUP + */ + ds[0].u = _3DSTATE_DEPTH_OFFSET_SCALE; + ds[1].f = i915->setup.offset_scale; + + set_dynamic_indirect( i915, + I915_DYNAMIC_DEPTHSCALE_0, + &ds[0].u, + 2 ); +} + +const struct i915_tracked_state i915_upload_DEPTHSCALE = { + .dirty = I915_NEW_SETUP, + .update = upload_DEPTHSCALE +}; + + + +/*********************************************************************** + * Polygon stipple + * + * The i915 supports a 4x4 stipple natively, GL wants 32x32. + * Fortunately stipple is usually a repeating pattern. + * + * XXX: does stipple pattern need to be adjusted according to + * the window position? + * + * XXX: possibly need workaround for conform paths test. + */ + +static void upload_STIPPLE( struct i915_context *i915 ) +{ + GLuint st[2]; + + st[0] = _3DSTATE_STIPPLE; + st[1] = 0; + + /* I915_NEW_SETUP + */ + if (i915->setup.poly_stipple_enable) { + st[1] |= ST1_ENABLE; + } + + + /* I915_NEW_STIPPLE + */ + { + const GLubyte *mask = (const GLubyte *)i915->poly_stipple.stipple; + GLubyte p[4]; + + p[0] = mask[12] & 0xf; + p[1] = mask[8] & 0xf; + p[2] = mask[4] & 0xf; + p[3] = mask[0] & 0xf; + + /* Not sure what to do about fallbacks, so for now just dont: + */ + st[1] |= ((p[0] << 0) | + (p[1] << 4) | + (p[2] << 8) | + (p[3] << 12)); + } + + + set_dynamic_indirect( i915, + I915_DYNAMIC_STP_0, + &st[0], + 2 ); +} + + +const struct i915_tracked_state i915_upload_STIPPLE = { + .dirty = I915_NEW_SETUP | I915_NEW_STIPPLE, + .update = upload_STIPPLE +}; + + + +/*********************************************************************** + * Scissor. + */ +static void upload_SCISSOR_ENABLE( struct i915_context *i915 ) +{ + unsigned sc[1]; + + if (i915->setup.scissor) + sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT; + else + sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT; + + set_dynamic_indirect( i915, + I915_DYNAMIC_SC_ENA_0, + &sc[0], + 1 ); +} + +const struct i915_tracked_state i915_upload_SCISSOR_ENABLE = { + .dirty = I915_NEW_SETUP, + .update = upload_SCISSOR_ENABLE +}; + + + +static void upload_SCISSOR_RECT( struct i915_context *i915 ) +{ + unsigned x1 = i915->scissor.minx; + unsigned y1 = i915->scissor.miny; + unsigned x2 = i915->scissor.maxx; + unsigned y2 = i915->scissor.maxy; + unsigned sc[3]; + + sc[0] = _3DSTATE_SCISSOR_RECT_0_CMD; + sc[1] = (y1 << 16) | (x1 & 0xffff); + sc[2] = (y2 << 16) | (x2 & 0xffff); + + set_dynamic_indirect( i915, + I915_DYNAMIC_SC_RECT_0, + &sc[0], + 3 ); +} + + +const struct i915_tracked_state i915_upload_SCISSOR_RECT = { + .dirty = I915_NEW_SCISSOR, + .update = upload_SCISSOR_RECT +}; + + + + + + +static const struct i915_tracked_state *atoms[] = { + &i915_upload_MODES4, + &i915_upload_BFO, + &i915_upload_BLENDCOLOR, + &i915_upload_IAB, + &i915_upload_DEPTHSCALE, + &i915_upload_STIPPLE, + &i915_upload_SCISSOR_ENABLE, + &i915_upload_SCISSOR_RECT +}; + +/* These will be dynamic indirect state commands, but for now just end + * up on the batch buffer with everything else. + */ +void i915_update_dynamic( struct i915_context *i915 ) +{ + int i; + + for (i = 0; i < Elements(atoms); i++) + if (i915->dirty & atoms[i]->dirty) + atoms[i]->update( i915 ); +} + diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 1a50470fbd..2ce079fbaa 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -66,14 +66,13 @@ i915_emit_hardware_state(struct i915_context *i915 ) { BEGIN_BATCH(100, 10); + if (i915->hardware_dirty & I915_HW_INVARIENT) { OUT_BATCH(_3DSTATE_AA_CMD | AA_LINE_ECAAR_WIDTH_ENABLE | AA_LINE_ECAAR_WIDTH_1_0 | AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0); - } - { OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); OUT_BATCH(0); @@ -82,10 +81,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(_3DSTATE_DFLT_Z_CMD); OUT_BATCH(0); - } - - { OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS | CSB_TCB(0, 0) | CSB_TCB(1, 1) | @@ -95,9 +91,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) CSB_TCB(5, 5) | CSB_TCB(6, 6) | CSB_TCB(7, 7)); - } - { OUT_BATCH(_3DSTATE_RASTER_RULES_CMD | ENABLE_POINT_RASTER_RULE | OGL_POINT_RASTER_RULE | @@ -107,42 +101,25 @@ i915_emit_hardware_state(struct i915_context *i915 ) TRI_FAN_PROVOKE_VRTX(2) | ENABLE_TEXKILL_3D_4D | TEXKILL_4D); - } - /* Need to initialize this to zero. - */ - { + /* Need to initialize this to zero. + */ OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (0)); OUT_BATCH(0); - } - - { - OUT_BATCH(_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); - - OUT_BATCH(_3DSTATE_SCISSOR_RECT_0_CMD); - OUT_BATCH(0); - OUT_BATCH(0); - } - - { OUT_BATCH(_3DSTATE_DEPTH_SUBRECT_DISABLE); - } - { - OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0); /* disable indirect state */ + /* disable indirect state for now + */ + OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0); OUT_BATCH(0); } - - { - /* Don't support twosided stencil yet */ - OUT_BATCH(_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_TWO_SIDE | 0); - OUT_BATCH(0); - } + + if (i915->hardware_dirty & I915_HW_IMMEDIATE) { OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(2) | @@ -156,93 +133,76 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S5]); OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S6]); } - - { - OUT_BATCH(_3DSTATE_MODES_4_CMD | - ENABLE_LOGIC_OP_FUNC | - LOGIC_OP_FUNC(LOGICOP_COPY) | - ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(0xff) | - ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(0xff)); - } - if (0) { - OUT_BATCH(_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | - IAB_MODIFY_ENABLE | - IAB_MODIFY_FUNC | - IAB_MODIFY_SRC_FACTOR | - IAB_MODIFY_DST_FACTOR); - } + if (i915->hardware_dirty & I915_HW_DYNAMIC) { - //3DSTATE_INDEPENDENT_ALPHA_BLEND (1 dwords): - OUT_BATCH(0x6ba008a1); - - //3DSTATE_CONSTANT_BLEND_COLOR (2 dwords): - OUT_BATCH(0x7d880000); - OUT_BATCH(0x00000000); + int i; + for (i = 0; i < I915_MAX_DYNAMIC; i++) { + OUT_BATCH(i915->current.dynamic[i]); + } } - if (i915->framebuffer.cbufs[0]) { - struct pipe_region *cbuf_region = i915->framebuffer.cbufs[0]->region; - unsigned pitch = (cbuf_region->pitch * - cbuf_region->cpp); - - OUT_BATCH(_3DSTATE_BUF_INFO_CMD); - - OUT_BATCH(BUF_3D_ID_COLOR_BACK | - BUF_3D_PITCH(pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - - OUT_RELOC(cbuf_region->buffer, - I915_BUFFER_ACCESS_WRITE, - 0); - } - - /* What happens if no zbuf?? - */ - if (i915->framebuffer.zbuf) { - struct pipe_region *depth_region = i915->framebuffer.zbuf->region; - unsigned zpitch = (depth_region->pitch * - depth_region->cpp); + if (i915->hardware_dirty & I915_HW_STATIC) + { + if (i915->framebuffer.cbufs[0]) { + struct pipe_region *cbuf_region = i915->framebuffer.cbufs[0]->region; + unsigned pitch = (cbuf_region->pitch * + cbuf_region->cpp); + + OUT_BATCH(_3DSTATE_BUF_INFO_CMD); + + OUT_BATCH(BUF_3D_ID_COLOR_BACK | + BUF_3D_PITCH(pitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); + + OUT_RELOC(cbuf_region->buffer, + I915_BUFFER_ACCESS_WRITE, + 0); + } + + /* What happens if no zbuf?? + */ + if (i915->framebuffer.zbuf) { + struct pipe_region *depth_region = i915->framebuffer.zbuf->region; + unsigned zpitch = (depth_region->pitch * + depth_region->cpp); - OUT_BATCH(_3DSTATE_BUF_INFO_CMD); + OUT_BATCH(_3DSTATE_BUF_INFO_CMD); - OUT_BATCH(BUF_3D_ID_DEPTH | - BUF_3D_PITCH(zpitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); + OUT_BATCH(BUF_3D_ID_DEPTH | + BUF_3D_PITCH(zpitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); - OUT_RELOC(depth_region->buffer, - I915_BUFFER_ACCESS_WRITE, - 0); - } + OUT_RELOC(depth_region->buffer, + I915_BUFFER_ACCESS_WRITE, + 0); + } - { - unsigned cformat = translate_format( i915->framebuffer.cbufs[0]->format ); - unsigned zformat = 0; + { + unsigned cformat = translate_format( i915->framebuffer.cbufs[0]->format ); + unsigned zformat = 0; - if (i915->framebuffer.zbuf) - zformat = translate_depth_format( i915->framebuffer.zbuf->format ); + if (i915->framebuffer.zbuf) + zformat = translate_depth_format( i915->framebuffer.zbuf->format ); - OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); + OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); - OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - LOD_PRECLAMP_OGL | - TEX_DEFAULT_COLOR_OGL | - cformat | - zformat ); - } + OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | /* .5 */ + LOD_PRECLAMP_OGL | + TEX_DEFAULT_COLOR_OGL | + cformat | + zformat ); + } - { - OUT_BATCH(_3DSTATE_STIPPLE); - OUT_BATCH(0); } + + { GLuint i, dwords; GLuint *prog = i915_passthrough_program( &dwords ); diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 8b93ca7f62..2f64d9dd90 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -44,11 +44,6 @@ */ -struct i915_tracked_state { - unsigned dirty; - void (*update)( struct i915_context * ); -}; - /*********************************************************************** -- cgit v1.2.3 From 3d31252d44fb5983a089b9f3488745757772adea Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 18:24:37 -0600 Subject: move viewport Y inversion to state tracker --- src/mesa/pipe/i915simple/i915_state.c | 13 ------- src/mesa/state_tracker/st_atom_viewport.c | 57 ++++++++++++------------------- 2 files changed, 21 insertions(+), 49 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index a176165b74..ab00cbc822 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -189,22 +189,9 @@ static void i915_set_viewport_state( struct pipe_context *pipe, const struct pipe_viewport_state *viewport ) { struct i915_context *i915 = i915_context(pipe); - float sy, ty; i915->viewport = *viewport; /* struct copy */ - /* Negate Y scale to flip image vertically. - * The NDC Y coords prior to viewport transformation are in the range - * [y=-1=bottom, y=1=top] - * Intel window coords are in the range [y=0=top, y=H-1=bottom] where H - * is the window height. - * Use the viewport transformation to invert Y. - */ - sy = viewport->scale[1]; - ty = viewport->translate[1]; - i915->viewport.scale[1] = -sy; - i915->viewport.translate[1] = i915->framebuffer.cbufs[0]->height - ty; - /* pass the viewport info to the draw module */ draw_set_viewport_state(i915->draw, &i915->viewport); diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index ac91f628aa..1307cbb6d2 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -33,9 +33,6 @@ #include "st_atom.h" - - - /** * Update the viewport transformation matrix. Depends on: * - viewport pos/size @@ -45,49 +42,37 @@ static void update_viewport( struct st_context *st ) { GLcontext *ctx = st->ctx; - const GLframebuffer *DrawBuffer = ctx->DrawBuffer; - GLfloat yScale = 1.0; - GLfloat yBias = 0.0; + GLfloat yScale, yBias; - /* _NEW_BUFFERS + /* Negate Y scale to flip image vertically. + * The NDC Y coords prior to viewport transformation are in the range + * [y=-1=bottom, y=1=top] + * Hardware window coords are in the range [y=0=top, y=H-1=bottom] where H + * is the window height. + * Use the viewport transformation to invert Y. */ - if (DrawBuffer) { -#if 0 - if (DrawBuffer->Name) { - /* User created FBO */ - struct st_renderbuffer *irb - = st_renderbuffer(DrawBuffer->_ColorDrawBuffers[0][0]); - if (irb && !irb->RenderToTexture) { - /* y=0=top */ - yScale = -1.0; - yBias = irb->Base.Height; - } - else { - /* y=0=bottom */ - yScale = 1.0; - yBias = 0.0; - } - } - else - { - /* window buffer, y=0=top */ - yScale = -1.0; - yBias = DrawBuffer->Height; - } -#endif + /* _NEW_BUFFERS + */ + if (ctx->DrawBuffer) { + yScale = -1; + yBias = ctx->DrawBuffer->Height ; + } + else { + /* we won't be rendering anything */ + yScale = 1.0; + yBias = 0.0; } + /* _NEW_VIEWPORT + */ { - /* _NEW_VIEWPORT - */ GLfloat x = ctx->Viewport.X; GLfloat y = ctx->Viewport.Y; GLfloat z = ctx->Viewport.Near; GLfloat half_width = ctx->Viewport.Width / 2.0; GLfloat half_height = ctx->Viewport.Height / 2.0; GLfloat half_depth = (ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; - struct pipe_viewport_state vp; vp.scale[0] = half_width; @@ -95,9 +80,9 @@ static void update_viewport( struct st_context *st ) vp.scale[2] = half_depth; vp.scale[3] = 1.0; - vp.translate[0] = (half_width + x); + vp.translate[0] = half_width + x; vp.translate[1] = (half_height + y) * yScale + yBias; - vp.translate[2] = (half_depth + z); + vp.translate[2] = half_depth + z; vp.translate[3] = 0.0; if (memcmp(&vp, &st->state.viewport, sizeof(vp)) != 0) { -- cgit v1.2.3 From be57c1aac46e6af2f1dd8ce1f10334d034ac7464 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 19:04:28 -0600 Subject: comments --- src/mesa/state_tracker/st_cb_clear.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 4da3a2500d..e3392101a8 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -129,11 +129,13 @@ make_color_shader(struct st_context *st) return NULL; } _mesa_init_instructions(p->Instructions, 2); + /* MOV result.color, fragment.color; */ p->Instructions[0].Opcode = OPCODE_MOV; p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[0].DstReg.Index = 0; + p->Instructions[0].DstReg.Index = FRAG_RESULT_COLR; p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; p->Instructions[0].SrcReg[0].Index = FRAG_ATTRIB_COL0; + /* END; */ p->Instructions[1].Opcode = OPCODE_END; p->InputsRead = FRAG_BIT_COL0; -- cgit v1.2.3 From c19a95510f89401125ac3641725749c957f35d96 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 19:04:50 -0600 Subject: more work toward texture-based glDrawPixels --- src/mesa/state_tracker/st_cb_drawpixels.c | 130 +++++++++++++++++++++++++++++- 1 file changed, 129 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 92a4e305d1..6b7cef2b2a 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -35,12 +35,122 @@ #include "st_context.h" #include "st_atom.h" #include "st_draw.h" +#include "st_program.h" #include "st_cb_drawpixels.h" +#include "st_cb_texture.h" +#include "st_cb_teximage.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "shader/prog_instruction.h" #include "vf/vf.h" + +/** + * Create a simple fragment shader that passes the texture color through + * to the fragment color. + */ +static struct st_fragment_program * +make_drawpixels_shader(struct st_context *st) +{ + GLcontext *ctx = st->ctx; + struct st_fragment_program *stfp; + struct gl_program *p; + GLboolean b; + + p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + if (!p) + return NULL; + + p->NumInstructions = 2; + p->Instructions = _mesa_alloc_instructions(2); + if (!p->Instructions) { + ctx->Driver.DeleteProgram(ctx, p); + return NULL; + } + _mesa_init_instructions(p->Instructions, 2); + /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ + p->Instructions[0].Opcode = OPCODE_TEX; + p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[0].DstReg.Index = FRAG_RESULT_COLR; + p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[0].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + p->Instructions[0].TexSrcUnit = 0; + p->Instructions[0].TexSrcTarget = TEXTURE_2D_INDEX; + /* END; */ + p->Instructions[1].Opcode = OPCODE_END; + + p->InputsRead = FRAG_BIT_COL0; + p->OutputsWritten = (1 << FRAG_RESULT_COLR); + + stfp = (struct st_fragment_program *) p; + /* compile into tgsi format */ + b = tgsi_mesa_compile_fp_program(&stfp->Base, + stfp->tokens, ST_FP_MAX_TOKENS); + assert(b); + + return stfp; +} + + +static struct pipe_mipmap_tree * +make_mipmap_tree(struct st_context *st, + GLsizei width, GLsizei height, GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels) +{ + GLuint pipeFormat = st_choose_pipe_format(st->pipe, format, type, GL_RGBA); + int cpp = 4, pitch; + struct pipe_mipmap_tree *mt = CALLOC_STRUCT(pipe_mipmap_tree); + + assert(pipeFormat); + + pitch = width; /* XXX pad */ + + if (unpack->BufferObj) { + /* + mt->region = buffer_object_region(unpack->BufferObj); + */ + } + else { + mt->region = st->pipe->region_alloc(st->pipe, cpp, pitch, height); + /* XXX do texstore() here */ + } + + mt->target = GL_TEXTURE_2D; + mt->internal_format = GL_RGBA; + mt->format = pipeFormat; + mt->first_level = 0; + mt->last_level = 0; + mt->width0 = width; + mt->height0 = height; + mt->depth0 = 1; + mt->cpp = cpp; + mt->compressed = 0; + mt->pitch = pitch; + mt->depth_pitch = 0; + mt->total_height = height; + mt->level[0].level_offset = 0; + mt->level[0].width = width; + mt->level[0].height = height; + mt->level[0].depth = 1; + mt->level[0].nr_images = 1; + mt->level[0].image_offset = NULL; + mt->refcount = 1; + + return mt; +} + + +static void +free_mipmap_tree(struct pipe_context *pipe, struct pipe_mipmap_tree *mt) +{ + pipe->region_release(pipe, &mt->region); +} + + + static void draw_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, GLsizei width, GLsizei height) @@ -95,7 +205,10 @@ draw_textured_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { + static struct st_fragment_program *stfp = NULL; + struct pipe_mipmap_tree *mt; struct pipe_setup_state setup; + struct pipe_fs_state fs; /* setup state: just scissor */ memset(&setup, 0, sizeof(setup)); @@ -103,13 +216,28 @@ draw_textured_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, setup.scissor = 1; st->pipe->set_setup_state(st->pipe, &setup); + /* fragment shader state: color pass-through program */ + if (!stfp) { + stfp = make_drawpixels_shader(st); + } + memset(&fs, 0, sizeof(fs)); + fs.inputs_read = stfp->Base.Base.InputsRead; + fs.tokens = &stfp->tokens[0]; + fs.constants = NULL; + st->pipe->set_fs_state(st->pipe, &fs); - /* XXX upload the texture image */ + /* mipmap tree state: */ + mt = make_mipmap_tree(st, width, height, format, type, unpack, pixels); + st->pipe->set_texture_state(st->pipe, 0, mt); + /* draw! */ draw_quad(st, x, y, z, width, height); /* restore GL state */ st->pipe->set_setup_state(st->pipe, &st->state.setup); + st->pipe->set_fs_state(st->pipe, &st->state.fs); + + free_mipmap_tree(st->pipe, mt); } -- cgit v1.2.3 From 1ac106eefd148fa47a1a77f652089bfe5ed92824 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 19:08:23 +0100 Subject: silence warning --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 4c99f2ed5f..90a432a041 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -78,7 +78,7 @@ intelPrintDRIInfo(intelScreenPrivate * intelScreen, fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); } - +#if 0 static void intelPrintSAREA(const drmI830Sarea * sarea) { @@ -104,6 +104,7 @@ intelPrintSAREA(const drmI830Sarea * sarea) sarea->rotated_offset, sarea->rotated_size); fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); } +#endif -- cgit v1.2.3 From 00677fb67c44a671f866cbd351fc6f183bcd83bb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 19:08:52 +0100 Subject: silence warning --- src/mesa/state_tracker/st_cb_program.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index ed47c12066..d0344efa0c 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -68,7 +68,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, GLenum target, GLuint id ) { - struct st_context *st = st_context(ctx); +// struct st_context *st = st_context(ctx); switch (target) { case GL_VERTEX_PROGRAM_ARB: { -- cgit v1.2.3 From 78b1a29a0da8d1877408421df5012d37084a96de Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 9 Aug 2007 19:09:19 +0100 Subject: Split texfilter enums to match common hardware usage. --- src/mesa/pipe/p_defines.h | 18 ++++++++----- src/mesa/pipe/p_state.h | 5 ++-- src/mesa/pipe/softpipe/sp_tex_sample.c | 11 +++----- src/mesa/state_tracker/st_atom_sampler.c | 46 ++++++++++++++++++++++++-------- 4 files changed, 54 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 9eb4c37434..efbb3239a2 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -134,12 +134,18 @@ #define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE 6 #define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER 7 -#define PIPE_TEX_FILTER_NEAREST 0 -#define PIPE_TEX_FILTER_LINEAR 1 -#define PIPE_TEX_FILTER_NEAREST_MIPMAP_NEAREST 2 -#define PIPE_TEX_FILTER_NEAREST_MIPMAP_LINEAR 3 -#define PIPE_TEX_FILTER_LINEAR_MIPMAP_NEAREST 4 -#define PIPE_TEX_FILTER_LINEAR_MIPMAP_LINEAR 5 +/* Between mipmaps, ie mipfilter + */ +#define PIPE_TEX_MIPFILTER_NEAREST 0 +#define PIPE_TEX_MIPFILTER_LINEAR 1 +#define PIPE_TEX_MIPFILTER_NONE 2 + +/* Within a mipmap, ie min/mag filter + */ +#define PIPE_TEX_FILTER_NEAREST 0 +#define PIPE_TEX_FILTER_LINEAR 1 +//#define PIPE_TEX_FILTER_ANISO 2 + #define PIPE_TEX_COMPARE_NONE 0 #define PIPE_TEX_COMPARE_R_TO_TEXTURE 1 diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index ee29e38a48..64a475ba00 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -213,8 +213,9 @@ struct pipe_sampler_state GLuint wrap_s:3; /**< PIPE_TEX_WRAP_x */ GLuint wrap_t:3; /**< PIPE_TEX_WRAP_x */ GLuint wrap_r:3; /**< PIPE_TEX_WRAP_x */ - GLuint min_filter:3; /**< PIPE_TEX_FILTER_x */ - GLuint mag_filter:1; /**< PIPE_TEX_FILTER_LINEAR or _NEAREST */ + GLuint min_img_filter:2; /**< PIPE_TEX_FILTER_x */ + GLuint min_mip_filter:2; /**< PIPE_TEX_MIPFILTER_x */ + GLuint mag_img_filter:2; /**< PIPE_TEX_FILTER_x */ GLuint compare:1; /**< shadow/depth compare enabled? */ GLenum compare_mode:1; /**< PIPE_TEX_COMPARE_x */ GLenum compare_func:3; /**< PIPE_FUNC_x */ diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index dcc9485cf6..40350eec75 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -417,7 +417,7 @@ sp_get_sample_1d(struct tgsi_sampler *sampler, struct pipe_surface *ps = pipe->get_tex_surface(pipe, sampler->texture, 0, 0, 0); - switch (sampler->state->min_filter) { + switch (sampler->state->min_img_filter) { case PIPE_TEX_FILTER_NEAREST: { GLint x; @@ -450,9 +450,7 @@ sp_get_sample_1d(struct tgsi_sampler *sampler, static GLuint choose_mipmap_level(struct tgsi_sampler *sampler, GLfloat lambda) { - if (sampler->state->min_filter == sampler->state->mag_filter) { - assert(sampler->state->min_filter == PIPE_TEX_FILTER_LINEAR || - sampler->state->min_filter == PIPE_TEX_FILTER_NEAREST); + if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { return 0; } else { @@ -485,15 +483,14 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, GLint level0; if (lambda < 0.0) - filter = sampler->state->mag_filter; + filter = sampler->state->mag_img_filter; else - filter = sampler->state->min_filter; + filter = sampler->state->min_img_filter; level0 = choose_mipmap_level(sampler, lambda); switch (filter) { case PIPE_TEX_FILTER_NEAREST: - case PIPE_TEX_FILTER_NEAREST_MIPMAP_NEAREST: { GLint x = nearest_texcoord(sampler->state->wrap_s, strq[0], sampler->texture->level[level0].width); diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 066ce7860f..7d568baf9e 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -69,28 +69,50 @@ gl_wrap_to_sp(GLenum wrap) static GLuint -gl_filter_to_sp(GLenum filter) +gl_filter_to_mip_filter(GLenum filter) { switch (filter) { case GL_NEAREST: - return PIPE_TEX_FILTER_NEAREST; case GL_LINEAR: - return PIPE_TEX_FILTER_LINEAR; + return PIPE_TEX_MIPFILTER_NONE; + case GL_NEAREST_MIPMAP_NEAREST: - return PIPE_TEX_FILTER_NEAREST_MIPMAP_NEAREST; + case GL_LINEAR_MIPMAP_NEAREST: + return PIPE_TEX_MIPFILTER_NEAREST; + case GL_NEAREST_MIPMAP_LINEAR: - return PIPE_TEX_FILTER_NEAREST_MIPMAP_LINEAR; + case GL_LINEAR_MIPMAP_LINEAR: + return PIPE_TEX_MIPFILTER_LINEAR; + + default: + assert(0); + return PIPE_TEX_MIPFILTER_NONE; + } +} + + +static GLuint +gl_filter_to_img_filter(GLenum filter) +{ + switch (filter) { + case GL_NEAREST: + case GL_NEAREST_MIPMAP_NEAREST: + case GL_NEAREST_MIPMAP_LINEAR: + return PIPE_TEX_FILTER_NEAREST; + + case GL_LINEAR: case GL_LINEAR_MIPMAP_NEAREST: - return PIPE_TEX_FILTER_LINEAR_MIPMAP_NEAREST; case GL_LINEAR_MIPMAP_LINEAR: - return PIPE_TEX_FILTER_LINEAR_MIPMAP_LINEAR; + return PIPE_TEX_FILTER_LINEAR; + default: - abort(); - return 0; + assert(0); + return PIPE_TEX_FILTER_NEAREST; } } + static void update_samplers(struct st_context *st) { @@ -108,12 +130,14 @@ update_samplers(struct st_context *st) sampler.wrap_t = gl_wrap_to_sp(texobj->WrapT); sampler.wrap_r = gl_wrap_to_sp(texobj->WrapR); - sampler.min_filter = gl_filter_to_sp(texobj->MinFilter); - sampler.mag_filter = gl_filter_to_sp(texobj->MagFilter); + sampler.min_img_filter = gl_filter_to_img_filter(texobj->MinFilter); + sampler.min_mip_filter = gl_filter_to_mip_filter(texobj->MinFilter); + sampler.mag_img_filter = gl_filter_to_img_filter(texobj->MagFilter); sampler.lod_bias = st->ctx->Texture.Unit[u].LodBias; sampler.min_lod = texobj->MinLod; sampler.max_lod = texobj->MaxLod; + sampler.max_anisotropy = texobj->MaxAnisotropy; /* XXX more sampler state here */ } -- cgit v1.2.3 From ac36cdbcc1ce4d78df50150117b11861c6ea672b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 22:55:15 +0100 Subject: added null ptr checks --- src/mesa/pipe/softpipe/sp_context.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index ebc994670a..0794d9a888 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -84,12 +84,15 @@ static void map_surfaces(struct softpipe_context *sp) GLuint i; for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); pipe->region_map(pipe, sps->surface.region); + struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); + if (sps->surface.region) + pipe->region_map(pipe, sps->surface.region); } if (sp->framebuffer.zbuf) { struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.zbuf); - pipe->region_map(pipe, sps->surface.region); + if (sps->surface.region) + pipe->region_map(pipe, sps->surface.region); } /* textures */ @@ -111,12 +114,14 @@ static void unmap_surfaces(struct softpipe_context *sp) for (i = 0; i < sp->framebuffer.num_cbufs; i++) { struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); - pipe->region_unmap(pipe, sps->surface.region); + if (sps->surface.region) + pipe->region_unmap(pipe, sps->surface.region); } if (sp->framebuffer.zbuf) { struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.zbuf); - pipe->region_unmap(pipe, sps->surface.region); + if (sps->surface.region) + pipe->region_unmap(pipe, sps->surface.region); } /* textures */ -- cgit v1.2.3 From 2f4f703af1f1d99f84adc9d8db30f3725eacd77f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 22:55:28 +0100 Subject: export softpipe_init_surface_funcs() --- src/mesa/pipe/softpipe/sp_surface.c | 10 +++------- src/mesa/pipe/softpipe/sp_surface.h | 4 ++++ 2 files changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 3260bcce75..852aeeeed1 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -368,10 +368,8 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, } - - -static void -init_quad_funcs(struct softpipe_surface *sps) +void +softpipe_init_surface_funcs(struct softpipe_surface *sps) { switch (sps->surface.format) { case PIPE_FORMAT_U_Z16: @@ -399,9 +397,7 @@ init_quad_funcs(struct softpipe_surface *sps) sps->surface.get_tile = a1r5g5b5_get_tile; break; default: - /* assert(0); - */ ; } } @@ -416,7 +412,7 @@ sp_surface_alloc(struct pipe_context *pipe, GLenum format) sps->surface.format = format; sps->surface.refcount = 1; - init_quad_funcs(sps); + softpipe_init_surface_funcs(sps); return &sps->surface; } diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 837e0c4ba1..00b5edcd92 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -93,6 +93,10 @@ softpipe_get_tex_surface(struct pipe_context *pipe, GLuint face, GLuint level, GLuint zslice); +extern void +softpipe_init_surface_funcs(struct softpipe_surface *sps); + + /** Cast wrapper */ static INLINE struct softpipe_surface * softpipe_surface(struct pipe_surface *ps) -- cgit v1.2.3 From 6883930f61c67924789f2c7cbc274627407cc784 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 22:56:21 +0100 Subject: init strb->Base.DataType appropriately, clean-ups --- src/mesa/state_tracker/st_cb_fbo.c | 14 ++++++++++++-- src/mesa/state_tracker/st_cb_fbo.h | 2 +- 2 files changed, 13 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 15816bee9e..67e239fd6d 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -135,8 +135,15 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, if (!info) return GL_FALSE; + switch (pipeFormat) { + case PIPE_FORMAT_S8_Z24: + strb->Base.DataType = GL_UNSIGNED_INT; + break; + default: + strb->Base.DataType = GL_UNSIGNED_BYTE; /* XXX fix */ + } + strb->Base._ActualFormat = info->base_format; - strb->Base.DataType = GL_UNSIGNED_BYTE; /* XXX fix */ strb->Base.RedBits = info->red_bits; strb->Base.GreenBits = info->green_bits; strb->Base.BlueBits = info->blue_bits; @@ -205,7 +212,9 @@ null_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb, /* By returning NULL we force all software rendering to go through * the span routines. */ +#if 0 assert(0); /* Should never get called with softpipe */ +#endif return NULL; } @@ -262,7 +271,7 @@ st_new_renderbuffer_fb(struct pipe_region *region, GLuint width, GLuint height) #else struct gl_renderbuffer * -st_new_renderbuffer_fb(GLuint intFormat) +st_new_renderbuffer_fb(GLenum intFormat) { struct st_renderbuffer *strb; @@ -282,6 +291,7 @@ st_new_renderbuffer_fb(GLuint intFormat) strb->Base._BaseFormat = GL_RGBA; break; case GL_DEPTH_COMPONENT16: + case GL_DEPTH_COMPONENT32: strb->Base._BaseFormat = GL_DEPTH_COMPONENT; break; case GL_DEPTH24_STENCIL8_EXT: diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index b2e7ba810c..2280441db5 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -50,7 +50,7 @@ st_renderbuffer(struct gl_renderbuffer *rb) extern struct gl_renderbuffer * -st_new_renderbuffer_fb(GLuint intFormat); +st_new_renderbuffer_fb(GLenum intFormat); extern void -- cgit v1.2.3 From 3a9eca5cc0a3a0c7f6198b37f61cc0d0ad45da1d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 9 Aug 2007 22:56:50 +0100 Subject: asst changes to get softpipe rendering again (no zbuf support for now) --- src/mesa/drivers/x11/xm_api.c | 31 +++- src/mesa/drivers/x11/xm_buffer.c | 102 +++++++++--- src/mesa/drivers/x11/xm_dd.c | 9 +- src/mesa/drivers/x11/xm_span.c | 325 +++++++++++++++++++------------------- src/mesa/drivers/x11/xm_surface.c | 94 +++++++++-- src/mesa/drivers/x11/xmesaP.h | 24 ++- 6 files changed, 369 insertions(+), 216 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 88e6bf8d6d..cea790dc60 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -84,6 +84,7 @@ #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" #include "pipe/softpipe/sp_context.h" +#include "pipe/p_defines.h" /** * Global X driver lock @@ -385,7 +386,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, /* * Front renderbuffer */ - b->frontxrb = xmesa_new_renderbuffer(NULL, 0, &vis->mesa_visual, GL_FALSE); + b->frontxrb = xmesa_create_renderbuffer(NULL, 0, &vis->mesa_visual, GL_FALSE); if (!b->frontxrb) { _mesa_free(b); return NULL; @@ -394,13 +395,13 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, b->frontxrb->drawable = d; b->frontxrb->pixmap = (XMesaPixmap) d; _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_FRONT_LEFT, - &b->frontxrb->Base); + &b->frontxrb->St.Base); /* * Back renderbuffer */ if (vis->mesa_visual.doubleBufferMode) { - b->backxrb = xmesa_new_renderbuffer(NULL, 0, &vis->mesa_visual, GL_TRUE); + b->backxrb = xmesa_create_renderbuffer(NULL, 0, &vis->mesa_visual, GL_TRUE); if (!b->backxrb) { /* XXX free front xrb too */ _mesa_free(b); @@ -411,7 +412,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, b->db_mode = vis->ximage_flag ? BACK_XIMAGE : BACK_PIXMAP; _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_BACK_LEFT, - &b->backxrb->Base); + &b->backxrb->St.Base); } /* @@ -429,12 +430,19 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, b->swAlpha = GL_FALSE; } + if (vis->mesa_visual.depthBits > 0) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_DEPTH, rb); + } + + /* * Other renderbuffer (depth, stencil, etc) */ _mesa_add_soft_renderbuffers(&b->mesa_buffer, GL_FALSE, /* color */ - vis->mesa_visual.haveDepthBuffer, + GL_FALSE,/*vis->mesa_visual.haveDepthBuffer,*/ vis->mesa_visual.haveStencilBuffer, vis->mesa_visual.haveAccumBuffer, b->swAlpha, @@ -1517,6 +1525,10 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) xmesa_init_driver_functions(v, &functions); st_init_driver_functions(&functions); + /* + functions.NewRenderbuffer = xmesa_new_renderbuffer; + */ + if (!_mesa_initialize_context(mesaCtx, &v->mesa_visual, share_list ? &(share_list->mesa) : (GLcontext *) NULL, &functions, (void *) c)) { @@ -1576,10 +1588,15 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) st_create_context( mesaCtx, xmesa_create_softpipe( c ) ); + mesaCtx->st->pipe->surface_alloc = xmesa_surface_alloc; + mesaCtx->st->pipe->supported_formats = xmesa_supported_formats; + +#if 1 mesaCtx->Driver.Clear = xmesa_clear_buffers; - /* +#endif +#if 0 mesaCtx->st->pipe->clear = xmesa_clear; - */ +#endif return c; } diff --git a/src/mesa/drivers/x11/xm_buffer.c b/src/mesa/drivers/x11/xm_buffer.c index 5bba52da48..fb6815ae1e 100644 --- a/src/mesa/drivers/x11/xm_buffer.c +++ b/src/mesa/drivers/x11/xm_buffer.c @@ -36,6 +36,8 @@ #include "framebuffer.h" #include "renderbuffer.h" #include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "state_tracker/st_context.h" #if defined(USE_XSHM) && !defined(XFree86Server) @@ -269,10 +271,10 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->Height = height; rb->InternalFormat = internalFormat; - if (!xrb->Base.surface) - xrb->Base.surface = xmesa_new_surface(ctx, xrb); - xrb->Base.surface->width = width; - xrb->Base.surface->height = height; + if (!xrb->St.surface) + xrb->St.surface = xmesa_new_surface(ctx, xrb); + xrb->St.surface->width = width; + xrb->St.surface->height = height; return GL_TRUE; } @@ -323,49 +325,101 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xrb->origin4 = NULL; } - if (!xrb->Base.surface) - xrb->Base.surface = xmesa_new_surface(ctx, xrb); - xrb->Base.surface->width = width; - xrb->Base.surface->height = height; + if (!xrb->St.surface) + xrb->St.surface = xmesa_new_surface(ctx, xrb); + xrb->St.surface->width = width; + xrb->St.surface->height = height; return GL_TRUE; } +/** + * Called to create the front/back color renderbuffers, not user-created + * renderbuffers. + */ struct xmesa_renderbuffer * -xmesa_new_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, +xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, GLboolean backBuffer) +{ + struct xmesa_renderbuffer *xrb = CALLOC_STRUCT(xmesa_renderbuffer); + struct pipe_context *pipe = NULL;/*ctx->st->pipe;*/ + if (xrb) { + GLuint name = 0; + GLuint pipeFormat = 0; + struct xmesa_surface *xms; + + _mesa_init_renderbuffer(&xrb->St.Base, name); + + xrb->St.Base.Delete = xmesa_delete_renderbuffer; + if (backBuffer) + xrb->St.Base.AllocStorage = xmesa_alloc_back_storage; + else + xrb->St.Base.AllocStorage = xmesa_alloc_front_storage; + + if (visual->rgbMode) { + xrb->St.Base.InternalFormat = GL_RGBA; + xrb->St.Base._BaseFormat = GL_RGBA; + xrb->St.Base.DataType = GL_UNSIGNED_BYTE; + xrb->St.Base.RedBits = visual->redBits; + xrb->St.Base.GreenBits = visual->greenBits; + xrb->St.Base.BlueBits = visual->blueBits; + xrb->St.Base.AlphaBits = visual->alphaBits; + pipeFormat = PIPE_FORMAT_U_A8_R8_G8_B8; + } + else { + xrb->St.Base.InternalFormat = GL_COLOR_INDEX; + xrb->St.Base._BaseFormat = GL_COLOR_INDEX; + xrb->St.Base.DataType = GL_UNSIGNED_INT; + xrb->St.Base.IndexBits = visual->indexBits; + } + /* only need to set Red/Green/EtcBits fields for user-created RBs */ + + xrb->St.surface = xmesa_surface_alloc(pipe, pipeFormat); + xms = (struct xmesa_surface *) xrb->St.surface; + xms->xrb = xrb; + + } + return xrb; +} + + +#if 0 +struct gl_renderbuffer * +xmesa_new_renderbuffer(GLcontext *ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, GLuint width, GLuint height) { struct xmesa_renderbuffer *xrb = CALLOC_STRUCT(xmesa_renderbuffer); if (xrb) { GLuint name = 0; - _mesa_init_renderbuffer(&xrb->Base, name); + _mesa_init_renderbuffer(&xrb->St.Base, name); - xrb->Base.Delete = xmesa_delete_renderbuffer; + xrb->St.Base.Delete = xmesa_delete_renderbuffer; if (backBuffer) - xrb->Base.AllocStorage = xmesa_alloc_back_storage; + xrb->St.Base.AllocStorage = xmesa_alloc_back_storage; else - xrb->Base.AllocStorage = xmesa_alloc_front_storage; + xrb->St.Base.AllocStorage = xmesa_alloc_front_storage; if (visual->rgbMode) { - xrb->Base.InternalFormat = GL_RGBA; - xrb->Base._BaseFormat = GL_RGBA; - xrb->Base.DataType = GL_UNSIGNED_BYTE; - xrb->Base.RedBits = visual->redBits; - xrb->Base.GreenBits = visual->greenBits; - xrb->Base.BlueBits = visual->blueBits; - xrb->Base.AlphaBits = visual->alphaBits; + xrb->St.Base.InternalFormat = GL_RGBA; + xrb->St.Base._BaseFormat = GL_RGBA; + xrb->St.Base.DataType = GL_UNSIGNED_BYTE; + xrb->St.Base.RedBits = visual->redBits; + xrb->St.Base.GreenBits = visual->greenBits; + xrb->St.Base.BlueBits = visual->blueBits; + xrb->St.Base.AlphaBits = visual->alphaBits; } else { - xrb->Base.InternalFormat = GL_COLOR_INDEX; - xrb->Base._BaseFormat = GL_COLOR_INDEX; - xrb->Base.DataType = GL_UNSIGNED_INT; - xrb->Base.IndexBits = visual->indexBits; + xrb->St.Base.InternalFormat = GL_COLOR_INDEX; + xrb->St.Base._BaseFormat = GL_COLOR_INDEX; + xrb->St.Base.DataType = GL_UNSIGNED_INT; + xrb->St.Base.IndexBits = visual->indexBits; } /* only need to set Red/Green/EtcBits fields for user-created RBs */ } return xrb; } +#endif /** diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index f64f8780cf..6eb6d32d19 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -224,7 +224,7 @@ clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, assert(xmbuf->cleargc); XMesaFillRectangle( xmesa->display, xrb->pixmap, xmbuf->cleargc, - x, xrb->Base.Height - y - height, + x, xrb->St.Base.Height - y - height, width, height ); } @@ -335,9 +335,9 @@ clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, | ((pixel << 24) & 0xff000000); } - if (width == xrb->Base.Width && height == xrb->Base.Height) { + if (width == xrb->St.Base.Width && height == xrb->St.Base.Height) { /* clearing whole buffer */ - const GLuint n = xrb->Base.Width * xrb->Base.Height; + const GLuint n = xrb->St.Base.Width * xrb->St.Base.Height; GLuint *ptr4 = (GLuint *) xrb->ximage->data; if (pixel == 0) { /* common case */ @@ -490,7 +490,7 @@ xmesa_DrawPixels_8R8G8B( GLcontext *ctx, ctx->Pixel.ZoomX == 1.0 && /* no zooming */ ctx->Pixel.ZoomY == 1.0 && xrb->pixmap && - xrb->Base.AlphaBits == 0) + xrb->St.Base.AlphaBits == 0) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); @@ -1134,6 +1134,7 @@ xmesa_init_driver_functions( XMesaVisual xmvisual, driver->BeginQuery = xmesa_begin_query; driver->EndQuery = xmesa_end_query; #endif + } diff --git a/src/mesa/drivers/x11/xm_span.c b/src/mesa/drivers/x11/xm_span.c index 3776891e2e..ce54a18a27 100644 --- a/src/mesa/drivers/x11/xm_span.c +++ b/src/mesa/drivers/x11/xm_span.c @@ -4539,259 +4539,260 @@ xmesa_set_renderbuffer_funcs(struct xmesa_renderbuffer *xrb, enum pixel_format pixelformat, GLint depth) { const GLboolean pixmap = xrb->pixmap ? GL_TRUE : GL_FALSE; + struct gl_renderbuffer *rb = &xrb->St.Base; switch (pixelformat) { case PF_Index: - ASSERT(xrb->Base.DataType == GL_UNSIGNED_INT); + ASSERT(rb->DataType == GL_UNSIGNED_INT); if (pixmap) { - xrb->Base.PutRow = put_row_ci_pixmap; - xrb->Base.PutRowRGB = NULL; - xrb->Base.PutMonoRow = put_mono_row_ci_pixmap; - xrb->Base.PutValues = put_values_ci_pixmap; - xrb->Base.PutMonoValues = put_mono_values_ci_pixmap; + rb->PutRow = put_row_ci_pixmap; + rb->PutRowRGB = NULL; + rb->PutMonoRow = put_mono_row_ci_pixmap; + rb->PutValues = put_values_ci_pixmap; + rb->PutMonoValues = put_mono_values_ci_pixmap; } else { - xrb->Base.PutRow = put_row_ci_ximage; - xrb->Base.PutRowRGB = NULL; - xrb->Base.PutMonoRow = put_mono_row_ci_ximage; - xrb->Base.PutValues = put_values_ci_ximage; - xrb->Base.PutMonoValues = put_mono_values_ci_ximage; + rb->PutRow = put_row_ci_ximage; + rb->PutRowRGB = NULL; + rb->PutMonoRow = put_mono_row_ci_ximage; + rb->PutValues = put_values_ci_ximage; + rb->PutMonoValues = put_mono_values_ci_ximage; } break; case PF_Truecolor: if (pixmap) { - xrb->Base.PutRow = put_row_TRUECOLOR_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_TRUECOLOR_pixmap; - xrb->Base.PutMonoRow = put_mono_row_pixmap; - xrb->Base.PutValues = put_values_TRUECOLOR_pixmap; - xrb->Base.PutMonoValues = put_mono_values_pixmap; + rb->PutRow = put_row_TRUECOLOR_pixmap; + rb->PutRowRGB = put_row_rgb_TRUECOLOR_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_TRUECOLOR_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; } else { - xrb->Base.PutRow = put_row_TRUECOLOR_ximage; - xrb->Base.PutRowRGB = put_row_rgb_TRUECOLOR_ximage; - xrb->Base.PutMonoRow = put_mono_row_ximage; - xrb->Base.PutValues = put_values_TRUECOLOR_ximage; - xrb->Base.PutMonoValues = put_mono_values_ximage; + rb->PutRow = put_row_TRUECOLOR_ximage; + rb->PutRowRGB = put_row_rgb_TRUECOLOR_ximage; + rb->PutMonoRow = put_mono_row_ximage; + rb->PutValues = put_values_TRUECOLOR_ximage; + rb->PutMonoValues = put_mono_values_ximage; } break; case PF_Dither_True: if (pixmap) { - xrb->Base.PutRow = put_row_TRUEDITHER_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_TRUEDITHER_pixmap; - xrb->Base.PutMonoRow = put_mono_row_TRUEDITHER_pixmap; - xrb->Base.PutValues = put_values_TRUEDITHER_pixmap; - xrb->Base.PutMonoValues = put_mono_values_TRUEDITHER_pixmap; + rb->PutRow = put_row_TRUEDITHER_pixmap; + rb->PutRowRGB = put_row_rgb_TRUEDITHER_pixmap; + rb->PutMonoRow = put_mono_row_TRUEDITHER_pixmap; + rb->PutValues = put_values_TRUEDITHER_pixmap; + rb->PutMonoValues = put_mono_values_TRUEDITHER_pixmap; } else { - xrb->Base.PutRow = put_row_TRUEDITHER_ximage; - xrb->Base.PutRowRGB = put_row_rgb_TRUEDITHER_ximage; - xrb->Base.PutMonoRow = put_mono_row_TRUEDITHER_ximage; - xrb->Base.PutValues = put_values_TRUEDITHER_ximage; - xrb->Base.PutMonoValues = put_mono_values_TRUEDITHER_ximage; + rb->PutRow = put_row_TRUEDITHER_ximage; + rb->PutRowRGB = put_row_rgb_TRUEDITHER_ximage; + rb->PutMonoRow = put_mono_row_TRUEDITHER_ximage; + rb->PutValues = put_values_TRUEDITHER_ximage; + rb->PutMonoValues = put_mono_values_TRUEDITHER_ximage; } break; case PF_8A8B8G8R: if (pixmap) { - xrb->Base.PutRow = put_row_8A8B8G8R_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_8A8B8G8R_pixmap; - xrb->Base.PutMonoRow = put_mono_row_pixmap; - xrb->Base.PutValues = put_values_8A8B8G8R_pixmap; - xrb->Base.PutMonoValues = put_mono_values_pixmap; + rb->PutRow = put_row_8A8B8G8R_pixmap; + rb->PutRowRGB = put_row_rgb_8A8B8G8R_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_8A8B8G8R_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; } else { - xrb->Base.PutRow = put_row_8A8B8G8R_ximage; - xrb->Base.PutRowRGB = put_row_rgb_8A8B8G8R_ximage; - xrb->Base.PutMonoRow = put_mono_row_8A8B8G8R_ximage; - xrb->Base.PutValues = put_values_8A8B8G8R_ximage; - xrb->Base.PutMonoValues = put_mono_values_8A8B8G8R_ximage; - xrb->Base.GetPointer = get_pointer_4_ximage; + rb->PutRow = put_row_8A8B8G8R_ximage; + rb->PutRowRGB = put_row_rgb_8A8B8G8R_ximage; + rb->PutMonoRow = put_mono_row_8A8B8G8R_ximage; + rb->PutValues = put_values_8A8B8G8R_ximage; + rb->PutMonoValues = put_mono_values_8A8B8G8R_ximage; + rb->GetPointer = get_pointer_4_ximage; } break; case PF_8A8R8G8B: if (pixmap) { - xrb->Base.PutRow = put_row_8A8R8G8B_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_8A8R8G8B_pixmap; - xrb->Base.PutMonoRow = put_mono_row_pixmap; - xrb->Base.PutValues = put_values_8A8R8G8B_pixmap; - xrb->Base.PutMonoValues = put_mono_values_pixmap; + rb->PutRow = put_row_8A8R8G8B_pixmap; + rb->PutRowRGB = put_row_rgb_8A8R8G8B_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_8A8R8G8B_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; } else { - xrb->Base.PutRow = put_row_8A8R8G8B_ximage; - xrb->Base.PutRowRGB = put_row_rgb_8A8R8G8B_ximage; - xrb->Base.PutMonoRow = put_mono_row_8A8R8G8B_ximage; - xrb->Base.PutValues = put_values_8A8R8G8B_ximage; - xrb->Base.PutMonoValues = put_mono_values_8A8R8G8B_ximage; - xrb->Base.GetPointer = get_pointer_4_ximage; + rb->PutRow = put_row_8A8R8G8B_ximage; + rb->PutRowRGB = put_row_rgb_8A8R8G8B_ximage; + rb->PutMonoRow = put_mono_row_8A8R8G8B_ximage; + rb->PutValues = put_values_8A8R8G8B_ximage; + rb->PutMonoValues = put_mono_values_8A8R8G8B_ximage; + rb->GetPointer = get_pointer_4_ximage; } break; case PF_8R8G8B: if (pixmap) { - xrb->Base.PutRow = put_row_8R8G8B_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_8R8G8B_pixmap; - xrb->Base.PutMonoRow = put_mono_row_pixmap; - xrb->Base.PutValues = put_values_8R8G8B_pixmap; - xrb->Base.PutMonoValues = put_mono_values_pixmap; + rb->PutRow = put_row_8R8G8B_pixmap; + rb->PutRowRGB = put_row_rgb_8R8G8B_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_8R8G8B_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; } else { - xrb->Base.PutRow = put_row_8R8G8B_ximage; - xrb->Base.PutRowRGB = put_row_rgb_8R8G8B_ximage; - xrb->Base.PutMonoRow = put_mono_row_8R8G8B_ximage; - xrb->Base.PutValues = put_values_8R8G8B_ximage; - xrb->Base.PutMonoValues = put_mono_values_8R8G8B_ximage; + rb->PutRow = put_row_8R8G8B_ximage; + rb->PutRowRGB = put_row_rgb_8R8G8B_ximage; + rb->PutMonoRow = put_mono_row_8R8G8B_ximage; + rb->PutValues = put_values_8R8G8B_ximage; + rb->PutMonoValues = put_mono_values_8R8G8B_ximage; } break; case PF_8R8G8B24: if (pixmap) { - xrb->Base.PutRow = put_row_8R8G8B24_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_8R8G8B24_pixmap; - xrb->Base.PutMonoRow = put_mono_row_pixmap; - xrb->Base.PutValues = put_values_8R8G8B24_pixmap; - xrb->Base.PutMonoValues = put_mono_values_pixmap; + rb->PutRow = put_row_8R8G8B24_pixmap; + rb->PutRowRGB = put_row_rgb_8R8G8B24_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_8R8G8B24_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; } else { - xrb->Base.PutRow = put_row_8R8G8B24_ximage; - xrb->Base.PutRowRGB = put_row_rgb_8R8G8B24_ximage; - xrb->Base.PutMonoRow = put_mono_row_8R8G8B24_ximage; - xrb->Base.PutValues = put_values_8R8G8B24_ximage; - xrb->Base.PutMonoValues = put_mono_values_8R8G8B24_ximage; + rb->PutRow = put_row_8R8G8B24_ximage; + rb->PutRowRGB = put_row_rgb_8R8G8B24_ximage; + rb->PutMonoRow = put_mono_row_8R8G8B24_ximage; + rb->PutValues = put_values_8R8G8B24_ximage; + rb->PutMonoValues = put_mono_values_8R8G8B24_ximage; } break; case PF_5R6G5B: if (pixmap) { - xrb->Base.PutRow = put_row_5R6G5B_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_5R6G5B_pixmap; - xrb->Base.PutMonoRow = put_mono_row_pixmap; - xrb->Base.PutValues = put_values_5R6G5B_pixmap; - xrb->Base.PutMonoValues = put_mono_values_pixmap; + rb->PutRow = put_row_5R6G5B_pixmap; + rb->PutRowRGB = put_row_rgb_5R6G5B_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_5R6G5B_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; } else { - xrb->Base.PutRow = put_row_5R6G5B_ximage; - xrb->Base.PutRowRGB = put_row_rgb_5R6G5B_ximage; - xrb->Base.PutMonoRow = put_mono_row_ximage; - xrb->Base.PutValues = put_values_5R6G5B_ximage; - xrb->Base.PutMonoValues = put_mono_values_ximage; + rb->PutRow = put_row_5R6G5B_ximage; + rb->PutRowRGB = put_row_rgb_5R6G5B_ximage; + rb->PutMonoRow = put_mono_row_ximage; + rb->PutValues = put_values_5R6G5B_ximage; + rb->PutMonoValues = put_mono_values_ximage; } break; case PF_Dither_5R6G5B: if (pixmap) { - xrb->Base.PutRow = put_row_DITHER_5R6G5B_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_DITHER_5R6G5B_pixmap; - xrb->Base.PutMonoRow = put_mono_row_TRUEDITHER_pixmap; - xrb->Base.PutValues = put_values_DITHER_5R6G5B_pixmap; - xrb->Base.PutMonoValues = put_mono_values_TRUEDITHER_pixmap; + rb->PutRow = put_row_DITHER_5R6G5B_pixmap; + rb->PutRowRGB = put_row_rgb_DITHER_5R6G5B_pixmap; + rb->PutMonoRow = put_mono_row_TRUEDITHER_pixmap; + rb->PutValues = put_values_DITHER_5R6G5B_pixmap; + rb->PutMonoValues = put_mono_values_TRUEDITHER_pixmap; } else { - xrb->Base.PutRow = put_row_DITHER_5R6G5B_ximage; - xrb->Base.PutRowRGB = put_row_rgb_DITHER_5R6G5B_ximage; - xrb->Base.PutMonoRow = put_mono_row_DITHER_5R6G5B_ximage; - xrb->Base.PutValues = put_values_DITHER_5R6G5B_ximage; - xrb->Base.PutMonoValues = put_mono_values_DITHER_5R6G5B_ximage; + rb->PutRow = put_row_DITHER_5R6G5B_ximage; + rb->PutRowRGB = put_row_rgb_DITHER_5R6G5B_ximage; + rb->PutMonoRow = put_mono_row_DITHER_5R6G5B_ximage; + rb->PutValues = put_values_DITHER_5R6G5B_ximage; + rb->PutMonoValues = put_mono_values_DITHER_5R6G5B_ximage; } break; case PF_Dither: if (pixmap) { - xrb->Base.PutRow = put_row_DITHER_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_DITHER_pixmap; - xrb->Base.PutMonoRow = put_mono_row_DITHER_pixmap; - xrb->Base.PutValues = put_values_DITHER_pixmap; - xrb->Base.PutMonoValues = put_mono_values_DITHER_pixmap; + rb->PutRow = put_row_DITHER_pixmap; + rb->PutRowRGB = put_row_rgb_DITHER_pixmap; + rb->PutMonoRow = put_mono_row_DITHER_pixmap; + rb->PutValues = put_values_DITHER_pixmap; + rb->PutMonoValues = put_mono_values_DITHER_pixmap; } else { if (depth == 8) { - xrb->Base.PutRow = put_row_DITHER8_ximage; - xrb->Base.PutRowRGB = put_row_rgb_DITHER8_ximage; - xrb->Base.PutMonoRow = put_mono_row_DITHER8_ximage; - xrb->Base.PutValues = put_values_DITHER8_ximage; - xrb->Base.PutMonoValues = put_mono_values_DITHER8_ximage; + rb->PutRow = put_row_DITHER8_ximage; + rb->PutRowRGB = put_row_rgb_DITHER8_ximage; + rb->PutMonoRow = put_mono_row_DITHER8_ximage; + rb->PutValues = put_values_DITHER8_ximage; + rb->PutMonoValues = put_mono_values_DITHER8_ximage; } else { - xrb->Base.PutRow = put_row_DITHER_ximage; - xrb->Base.PutRowRGB = put_row_rgb_DITHER_ximage; - xrb->Base.PutMonoRow = put_mono_row_DITHER_ximage; - xrb->Base.PutValues = put_values_DITHER_ximage; - xrb->Base.PutMonoValues = put_mono_values_DITHER_ximage; + rb->PutRow = put_row_DITHER_ximage; + rb->PutRowRGB = put_row_rgb_DITHER_ximage; + rb->PutMonoRow = put_mono_row_DITHER_ximage; + rb->PutValues = put_values_DITHER_ximage; + rb->PutMonoValues = put_mono_values_DITHER_ximage; } } break; case PF_1Bit: if (pixmap) { - xrb->Base.PutRow = put_row_1BIT_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_1BIT_pixmap; - xrb->Base.PutMonoRow = put_mono_row_1BIT_pixmap; - xrb->Base.PutValues = put_values_1BIT_pixmap; - xrb->Base.PutMonoValues = put_mono_values_1BIT_pixmap; + rb->PutRow = put_row_1BIT_pixmap; + rb->PutRowRGB = put_row_rgb_1BIT_pixmap; + rb->PutMonoRow = put_mono_row_1BIT_pixmap; + rb->PutValues = put_values_1BIT_pixmap; + rb->PutMonoValues = put_mono_values_1BIT_pixmap; } else { - xrb->Base.PutRow = put_row_1BIT_ximage; - xrb->Base.PutRowRGB = put_row_rgb_1BIT_ximage; - xrb->Base.PutMonoRow = put_mono_row_1BIT_ximage; - xrb->Base.PutValues = put_values_1BIT_ximage; - xrb->Base.PutMonoValues = put_mono_values_1BIT_ximage; + rb->PutRow = put_row_1BIT_ximage; + rb->PutRowRGB = put_row_rgb_1BIT_ximage; + rb->PutMonoRow = put_mono_row_1BIT_ximage; + rb->PutValues = put_values_1BIT_ximage; + rb->PutMonoValues = put_mono_values_1BIT_ximage; } break; case PF_HPCR: if (pixmap) { - xrb->Base.PutRow = put_row_HPCR_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_HPCR_pixmap; - xrb->Base.PutMonoRow = put_mono_row_pixmap; - xrb->Base.PutValues = put_values_HPCR_pixmap; - xrb->Base.PutMonoValues = put_mono_values_pixmap; + rb->PutRow = put_row_HPCR_pixmap; + rb->PutRowRGB = put_row_rgb_HPCR_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_HPCR_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; } else { - xrb->Base.PutRow = put_row_HPCR_ximage; - xrb->Base.PutRowRGB = put_row_rgb_HPCR_ximage; - xrb->Base.PutMonoRow = put_mono_row_HPCR_ximage; - xrb->Base.PutValues = put_values_HPCR_ximage; - xrb->Base.PutMonoValues = put_mono_values_HPCR_ximage; + rb->PutRow = put_row_HPCR_ximage; + rb->PutRowRGB = put_row_rgb_HPCR_ximage; + rb->PutMonoRow = put_mono_row_HPCR_ximage; + rb->PutValues = put_values_HPCR_ximage; + rb->PutMonoValues = put_mono_values_HPCR_ximage; } break; case PF_Lookup: if (pixmap) { - xrb->Base.PutRow = put_row_LOOKUP_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_LOOKUP_pixmap; - xrb->Base.PutMonoRow = put_mono_row_pixmap; - xrb->Base.PutValues = put_values_LOOKUP_pixmap; - xrb->Base.PutMonoValues = put_mono_values_pixmap; + rb->PutRow = put_row_LOOKUP_pixmap; + rb->PutRowRGB = put_row_rgb_LOOKUP_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_LOOKUP_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; } else { if (depth==8) { - xrb->Base.PutRow = put_row_LOOKUP8_ximage; - xrb->Base.PutRowRGB = put_row_rgb_LOOKUP8_ximage; - xrb->Base.PutMonoRow = put_mono_row_LOOKUP8_ximage; - xrb->Base.PutValues = put_values_LOOKUP8_ximage; - xrb->Base.PutMonoValues = put_mono_values_LOOKUP8_ximage; + rb->PutRow = put_row_LOOKUP8_ximage; + rb->PutRowRGB = put_row_rgb_LOOKUP8_ximage; + rb->PutMonoRow = put_mono_row_LOOKUP8_ximage; + rb->PutValues = put_values_LOOKUP8_ximage; + rb->PutMonoValues = put_mono_values_LOOKUP8_ximage; } else { - xrb->Base.PutRow = put_row_LOOKUP_ximage; - xrb->Base.PutRowRGB = put_row_rgb_LOOKUP_ximage; - xrb->Base.PutMonoRow = put_mono_row_ximage; - xrb->Base.PutValues = put_values_LOOKUP_ximage; - xrb->Base.PutMonoValues = put_mono_values_ximage; + rb->PutRow = put_row_LOOKUP_ximage; + rb->PutRowRGB = put_row_rgb_LOOKUP_ximage; + rb->PutMonoRow = put_mono_row_ximage; + rb->PutValues = put_values_LOOKUP_ximage; + rb->PutMonoValues = put_mono_values_ximage; } } break; case PF_Grayscale: if (pixmap) { - xrb->Base.PutRow = put_row_GRAYSCALE_pixmap; - xrb->Base.PutRowRGB = put_row_rgb_GRAYSCALE_pixmap; - xrb->Base.PutMonoRow = put_mono_row_pixmap; - xrb->Base.PutValues = put_values_GRAYSCALE_pixmap; - xrb->Base.PutMonoValues = put_mono_values_pixmap; + rb->PutRow = put_row_GRAYSCALE_pixmap; + rb->PutRowRGB = put_row_rgb_GRAYSCALE_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_GRAYSCALE_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; } else { if (depth == 8) { - xrb->Base.PutRow = put_row_GRAYSCALE8_ximage; - xrb->Base.PutRowRGB = put_row_rgb_GRAYSCALE8_ximage; - xrb->Base.PutMonoRow = put_mono_row_GRAYSCALE8_ximage; - xrb->Base.PutValues = put_values_GRAYSCALE8_ximage; - xrb->Base.PutMonoValues = put_mono_values_GRAYSCALE8_ximage; + rb->PutRow = put_row_GRAYSCALE8_ximage; + rb->PutRowRGB = put_row_rgb_GRAYSCALE8_ximage; + rb->PutMonoRow = put_mono_row_GRAYSCALE8_ximage; + rb->PutValues = put_values_GRAYSCALE8_ximage; + rb->PutMonoValues = put_mono_values_GRAYSCALE8_ximage; } else { - xrb->Base.PutRow = put_row_GRAYSCALE_ximage; - xrb->Base.PutRowRGB = put_row_rgb_GRAYSCALE_ximage; - xrb->Base.PutMonoRow = put_mono_row_ximage; - xrb->Base.PutValues = put_values_GRAYSCALE_ximage; - xrb->Base.PutMonoValues = put_mono_values_ximage; + rb->PutRow = put_row_GRAYSCALE_ximage; + rb->PutRowRGB = put_row_rgb_GRAYSCALE_ximage; + rb->PutMonoRow = put_mono_row_ximage; + rb->PutValues = put_values_GRAYSCALE_ximage; + rb->PutMonoValues = put_mono_values_ximage; } } break; @@ -4803,12 +4804,12 @@ xmesa_set_renderbuffer_funcs(struct xmesa_renderbuffer *xrb, /* Get functions */ if (pixelformat == PF_Index) { - xrb->Base.GetRow = get_row_ci; - xrb->Base.GetValues = get_values_ci; + rb->GetRow = get_row_ci; + rb->GetValues = get_values_ci; } else { - xrb->Base.GetRow = get_row_rgba; - xrb->Base.GetValues = get_values_rgba; + rb->GetRow = get_row_rgba; + rb->GetValues = get_values_rgba; } } diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 6b52b33bf2..9d7d25667e 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -45,14 +45,21 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/softpipe/sp_context.h" -#include "pipe/softpipe/sp_surface.h" #include "state_tracker/st_context.h" +static INLINE struct xmesa_surface * +xmesa_surf(struct softpipe_surface *sps) +{ + return (struct xmesa_surface *) sps; +} + + static INLINE struct xmesa_renderbuffer * xmesa_rb(struct softpipe_surface *sps) { - return (struct xmesa_renderbuffer *) sps->surface.rb; + struct xmesa_surface *xms = xmesa_surf(sps); + return xms->xrb; } @@ -66,13 +73,14 @@ static void read_quad_f(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rgba)[NUM_CHANNELS]) { + struct xmesa_surface *xms = xmesa_surf(sps); struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GLubyte temp[16]; GLfloat *dst = (GLfloat *) rgba; GLuint i; GET_CURRENT_CONTEXT(ctx); - xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, temp); - xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, temp + 8); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, temp); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y + 1, temp + 8); for (i = 0; i < 16; i++) { dst[i] = UBYTE_TO_FLOAT(temp[i]); } @@ -82,13 +90,14 @@ static void read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rrrr)[QUAD_SIZE]) { + struct xmesa_surface *xms = xmesa_surf(sps); struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GLubyte temp[16]; GLfloat *dst = (GLfloat *) rrrr; GLuint i, j; GET_CURRENT_CONTEXT(ctx); - xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, temp); - xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, temp + 8); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, temp); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y + 1, temp + 8); for (i = 0; i < 4; i++) { for (j = 0; j < 4; j++) { dst[j * 4 + i] = UBYTE_TO_FLOAT(temp[i * 4 + j]); @@ -108,8 +117,8 @@ write_quad_f(struct softpipe_surface *sps, GLint x, GLint y, for (i = 0; i < 16; i++) { UNCLAMPED_FLOAT_TO_UBYTE(temp[i], src[i]); } - xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y, temp, NULL); - xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y + 1, temp + 8, NULL); + xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y, temp, NULL); + xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y + 1, temp + 8, NULL); } static void @@ -126,8 +135,8 @@ write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + i], src[i * 4 + j]); } } - xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y, temp, NULL); - xrb->Base.PutRow(ctx, &xrb->Base, 2, x, y + 1, temp + 8, NULL); + xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y, temp, NULL); + xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y + 1, temp + 8, NULL); } static void @@ -136,8 +145,8 @@ read_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, { struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GET_CURRENT_CONTEXT(ctx); - xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, rgba); - xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, rgba + 2); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, rgba); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y + 1, rgba + 2); } static void @@ -146,8 +155,8 @@ write_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, { struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GET_CURRENT_CONTEXT(ctx); - xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y, rgba); - xrb->Base.GetRow(ctx, &xrb->Base, 2, x, y + 1, rgba + 2); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, rgba); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y + 1, rgba + 2); } #if 0 @@ -157,7 +166,7 @@ write_mono_row_ub(struct softpipe_surface *sps, GLuint count, GLint x, GLint y, { struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GET_CURRENT_CONTEXT(ctx); - xrb->Base.PutMonoRow(ctx, &xrb->Base, count, x, y, rgba, NULL); + xrb->St.Base.PutMonoRow(ctx, &xrb->St.Base, count, x, y, rgba, NULL); } #endif @@ -177,9 +186,11 @@ xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb) if (!sps) return NULL; +#if 0 sps->surface.rb = xrb; /* XXX only needed for quad funcs above */ - sps->surface.width = xrb->Base.Width; - sps->surface.height = xrb->Base.Height; +#endif + sps->surface.width = xrb->St.Base.Width; + sps->surface.height = xrb->St.Base.Height; sps->read_quad_f = read_quad_f; sps->read_quad_f_swz = read_quad_f_swz; @@ -197,3 +208,52 @@ xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb) return &sps->surface; } + + +struct pipe_surface * +xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) +{ + struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + + assert(pipeFormat); + + xms->surface.surface.format = pipeFormat; + + switch (pipeFormat) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + xms->surface.read_quad_f_swz = read_quad_f; + xms->surface.read_quad_f = read_quad_f; + xms->surface.read_quad_f_swz = read_quad_f_swz; + xms->surface.read_quad_ub = read_quad_ub; + xms->surface.write_quad_f = write_quad_f; + xms->surface.write_quad_f_swz = write_quad_f_swz; + xms->surface.write_quad_ub = write_quad_ub; + break; + case PIPE_FORMAT_S8_Z24: + softpipe_init_surface_funcs(&xms->surface); + /* + xms->surface.read_quad_z = 1; + xms->surface.write_quad_z = 1; + */ + break; + default: + abort(); + } + + return &xms->surface.surface; +} + + +const GLuint * +xmesa_supported_formats(struct pipe_context *pipe, GLuint *numFormats) +{ + static const GLuint formats[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_S8_Z24 + }; + + *numFormats = 2; + + return formats; +} + diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 99c87da3c5..361c657354 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -36,6 +36,9 @@ #ifdef XFree86Server #include "xm_image.h" #endif +#include "state_tracker/st_cb_fbo.h" +#include "pipe/softpipe/sp_context.h" +#include "pipe/softpipe/sp_surface.h" extern _glthread_Mutex _xmesa_lock; @@ -177,7 +180,11 @@ typedef enum { */ struct xmesa_renderbuffer { +#if 0 struct gl_renderbuffer Base; /* Base class */ +#else + struct st_renderbuffer St; /**< Base class */ +#endif XMesaBuffer Parent; /**< The XMesaBuffer this renderbuffer belongs to */ XMesaDrawable drawable; /* Usually the X window ID */ @@ -493,8 +500,8 @@ extern const int xmesa_kernel1[16]; */ extern struct xmesa_renderbuffer * -xmesa_new_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, - GLboolean backBuffer); +xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, + GLboolean backBuffer); extern void xmesa_delete_framebuffer(struct gl_framebuffer *fb); @@ -589,6 +596,13 @@ extern void xmesa_register_swrast_functions( GLcontext *ctx ); struct pipe_surface; struct pipe_context; +struct xmesa_surface +{ + struct softpipe_surface surface; + struct xmesa_renderbuffer *xrb; +}; + + extern struct pipe_surface * xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb); @@ -601,5 +615,11 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers); extern struct pipe_context * xmesa_create_softpipe(XMesaContext xm); +extern struct pipe_surface * +xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); + +extern const GLuint * +xmesa_supported_formats(struct pipe_context *pipe, GLuint *numFormats); + #endif -- cgit v1.2.3 From d2bfdae860022b2cef2fff2de25440ba0e8f1d2b Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 08:33:43 +0100 Subject: invert Y axis --- src/mesa/drivers/x11/xm_surface.c | 35 ++++++++++++++++------------------- 1 file changed, 16 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 9d7d25667e..1915151ba4 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -63,24 +63,28 @@ xmesa_rb(struct softpipe_surface *sps) } +#define FLIP(Y) Y = xrb->St.Base.Height - (Y) - 1; + + /** * quad reading/writing * These functions are just wrappers around the existing renderbuffer * functions. + * Note that Y=0 is the top of the surface. */ static void read_quad_f(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rgba)[NUM_CHANNELS]) { - struct xmesa_surface *xms = xmesa_surf(sps); struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GLubyte temp[16]; GLfloat *dst = (GLfloat *) rgba; GLuint i; GET_CURRENT_CONTEXT(ctx); + FLIP(y); xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, temp); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y + 1, temp + 8); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y - 1, temp + 8); for (i = 0; i < 16; i++) { dst[i] = UBYTE_TO_FLOAT(temp[i]); } @@ -90,14 +94,14 @@ static void read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, GLfloat (*rrrr)[QUAD_SIZE]) { - struct xmesa_surface *xms = xmesa_surf(sps); struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GLubyte temp[16]; GLfloat *dst = (GLfloat *) rrrr; GLuint i, j; GET_CURRENT_CONTEXT(ctx); + FLIP(y); xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, temp); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y + 1, temp + 8); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y - 1, temp + 8); for (i = 0; i < 4; i++) { for (j = 0; j < 4; j++) { dst[j * 4 + i] = UBYTE_TO_FLOAT(temp[i * 4 + j]); @@ -114,11 +118,12 @@ write_quad_f(struct softpipe_surface *sps, GLint x, GLint y, const GLfloat *src = (const GLfloat *) rgba; GLuint i; GET_CURRENT_CONTEXT(ctx); + FLIP(y); for (i = 0; i < 16; i++) { UNCLAMPED_FLOAT_TO_UBYTE(temp[i], src[i]); } xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y, temp, NULL); - xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y + 1, temp + 8, NULL); + xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y - 1, temp + 8, NULL); } static void @@ -130,13 +135,14 @@ write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, const GLfloat *src = (const GLfloat *) rrrr; GLuint i, j; GET_CURRENT_CONTEXT(ctx); + FLIP(y); for (i = 0; i < 4; i++) { for (j = 0; j < 4; j++) { UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + i], src[i * 4 + j]); } } xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y, temp, NULL); - xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y + 1, temp + 8, NULL); + xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y - 1, temp + 8, NULL); } static void @@ -145,8 +151,9 @@ read_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, { struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GET_CURRENT_CONTEXT(ctx); + FLIP(y); xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, rgba); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y + 1, rgba + 2); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y - 1, rgba + 2); } static void @@ -155,20 +162,10 @@ write_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, { struct xmesa_renderbuffer *xrb = xmesa_rb(sps); GET_CURRENT_CONTEXT(ctx); + FLIP(y); xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, rgba); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y + 1, rgba + 2); -} - -#if 0 -static void -write_mono_row_ub(struct softpipe_surface *sps, GLuint count, GLint x, GLint y, - GLubyte rgba[NUM_CHANNELS]) -{ - struct xmesa_renderbuffer *xrb = xmesa_rb(sps); - GET_CURRENT_CONTEXT(ctx); - xrb->St.Base.PutMonoRow(ctx, &xrb->St.Base, count, x, y, rgba, NULL); + xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y - 1, rgba + 2); } -#endif /** -- cgit v1.2.3 From a25dd4d407a9ed797d4b9841c62f33efdfa07847 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 08:34:18 +0100 Subject: code movement --- src/mesa/state_tracker/st_cb_fbo.c | 71 +--------- src/mesa/state_tracker/st_cb_teximage.c | 238 ++++++++++++++++++++++++++------ src/mesa/state_tracker/st_cb_teximage.h | 26 +++- src/mesa/state_tracker/st_cb_texture.c | 150 +------------------- 4 files changed, 219 insertions(+), 266 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 67e239fd6d..8443c9ea1b 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -46,75 +46,6 @@ #include "st_cb_teximage.h" -struct pipe_format_info -{ - GLuint format; - GLenum base_format; - GLubyte red_bits; - GLubyte green_bits; - GLubyte blue_bits; - GLubyte alpha_bits; - GLubyte luminance_bits; - GLubyte intensity_bits; - GLubyte depth_bits; - GLubyte stencil_bits; - GLubyte size; /**< in bytes */ -}; - - -/* - * XXX temporary here - */ -static const struct pipe_format_info * -pipe_get_format_info(GLuint format) -{ - static const struct pipe_format_info info[] = { - { - PIPE_FORMAT_U_R8_G8_B8_A8, /* format */ - GL_RGBA, /* base_format */ - 8, 8, 8, 8, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 4 /* size in bytes */ - }, - { - PIPE_FORMAT_U_A8_R8_G8_B8, - GL_RGBA, /* base_format */ - 8, 8, 8, 8, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 4 /* size in bytes */ - }, - { - PIPE_FORMAT_U_A1_R5_G5_B5, - GL_RGBA, /* base_format */ - 5, 5, 5, 1, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 2 /* size in bytes */ - }, - { - PIPE_FORMAT_U_R5_G6_B5, - GL_RGBA, /* base_format */ - 5, 6, 5, 0, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 2 /* size in bytes */ - }, - /* XXX lots more */ - { - PIPE_FORMAT_S8_Z24, - GL_DEPTH_STENCIL_EXT, /* base_format */ - 0, 0, 0, 0, 0, 0, /* color bits */ - 24, 8, /* depth, stencil */ - 4 /* size in bytes */ - } - }; - GLuint i; - - for (i = 0; i < sizeof(info) / sizeof(info[0]); i++) { - if (info[i].format == format) - return info + i; - } - return NULL; -} - /** * gl_renderbuffer::AllocStorage() @@ -128,7 +59,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, struct st_renderbuffer *strb = st_renderbuffer(rb); const GLuint pipeFormat = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE); - const struct pipe_format_info *info = pipe_get_format_info(pipeFormat); + const struct pipe_format_info *info = st_get_format_info(pipeFormat); GLuint cpp, pitch; assert(info); diff --git a/src/mesa/state_tracker/st_cb_teximage.c b/src/mesa/state_tracker/st_cb_teximage.c index 39c9367695..c54602db5d 100644 --- a/src/mesa/state_tracker/st_cb_teximage.c +++ b/src/mesa/state_tracker/st_cb_teximage.c @@ -34,6 +34,8 @@ #include "main/imports.h" #include "main/context.h" #include "main/texstore.h" +#include "main/texformat.h" +#include "main/enums.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -42,6 +44,60 @@ +/* + * XXX temporary here + */ +const struct pipe_format_info * +st_get_format_info(GLuint format) +{ + static const struct pipe_format_info info[] = { + { + PIPE_FORMAT_U_R8_G8_B8_A8, /* format */ + GL_RGBA, /* base_format */ + 8, 8, 8, 8, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 4 /* size in bytes */ + }, + { + PIPE_FORMAT_U_A8_R8_G8_B8, + GL_RGBA, /* base_format */ + 8, 8, 8, 8, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 4 /* size in bytes */ + }, + { + PIPE_FORMAT_U_A1_R5_G5_B5, + GL_RGBA, /* base_format */ + 5, 5, 5, 1, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 2 /* size in bytes */ + }, + { + PIPE_FORMAT_U_R5_G6_B5, + GL_RGBA, /* base_format */ + 5, 6, 5, 0, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 2 /* size in bytes */ + }, + /* XXX lots more */ + { + PIPE_FORMAT_S8_Z24, + GL_DEPTH_STENCIL_EXT, /* base_format */ + 0, 0, 0, 0, 0, 0, /* color bits */ + 24, 8, /* depth, stencil */ + 4 /* size in bytes */ + } + }; + GLuint i; + + for (i = 0; i < sizeof(info) / sizeof(info[0]); i++) { + if (info[i].format == format) + return info + i; + } + return NULL; +} + + /** * Search list of formats for first RGBA format. @@ -285,58 +341,152 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, -static void -st_teximage2d(GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - /* probably nothing here: use core Mesa TexImage2D fallback to - * save teximage in main memory. - * Later, when we have a complete texobj and are ready to render, - * create the pipe texture object / mipmap-tree. - */ - - _mesa_store_teximage2d(ctx, target, level, internalFormat, - width, height, border, format, type, pixels, - packing, texObj, texImage); -} - - -static void -st_texsubimage2d(GLcontext *ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) +/* It works out that this function is fine for all the supported + * hardware. However, there is still a need to map the formats onto + * hardware descriptors. + */ +/* Note that the i915 can actually support many more formats than + * these if we take the step of simply swizzling the colors + * immediately after sampling... + */ +const struct gl_texture_format * +st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, + GLenum format, GLenum type) { #if 0 - struct pipe_mipmap_tree *mt = texObj->DriverData; + struct intel_context *intel = intel_context(ctx); + const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); +#else + const GLboolean do32bpt = 1; #endif - /* 1. find the region which stores this texture object. - * 2. convert texels to pipe format if needed. - * 3. replace texdata in the texture region. - */ -} + switch (internalFormat) { + case 4: + case GL_RGBA: + case GL_COMPRESSED_RGBA: + if (format == GL_BGRA) { + if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { + return &_mesa_texformat_argb8888; + } + else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { + return &_mesa_texformat_argb4444; + } + else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { + return &_mesa_texformat_argb1555; + } + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + case 3: + case GL_RGB: + case GL_COMPRESSED_RGB: + if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { + return &_mesa_texformat_rgb565; + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; -void st_init_cb_teximage( struct st_context *st ) -{ - struct dd_function_table *functions = &st->ctx->Driver; + case GL_RGBA4: + case GL_RGBA2: + return &_mesa_texformat_argb4444; - functions->TexImage2D = st_teximage2d; - functions->TexSubImage2D = st_texsubimage2d; -} + case GL_RGB5_A1: + return &_mesa_texformat_argb1555; + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + return &_mesa_texformat_argb8888; -void st_destroy_cb_teximage( struct st_context *st ) -{ + case GL_RGB5: + case GL_RGB4: + case GL_R3_G3_B2: + return &_mesa_texformat_rgb565; + + case GL_ALPHA: + case GL_ALPHA4: + case GL_ALPHA8: + case GL_ALPHA12: + case GL_ALPHA16: + case GL_COMPRESSED_ALPHA: + return &_mesa_texformat_a8; + + case 1: + case GL_LUMINANCE: + case GL_LUMINANCE4: + case GL_LUMINANCE8: + case GL_LUMINANCE12: + case GL_LUMINANCE16: + case GL_COMPRESSED_LUMINANCE: + return &_mesa_texformat_l8; + + case 2: + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE4_ALPHA4: + case GL_LUMINANCE6_ALPHA2: + case GL_LUMINANCE8_ALPHA8: + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: + case GL_COMPRESSED_LUMINANCE_ALPHA: + return &_mesa_texformat_al88; + + case GL_INTENSITY: + case GL_INTENSITY4: + case GL_INTENSITY8: + case GL_INTENSITY12: + case GL_INTENSITY16: + case GL_COMPRESSED_INTENSITY: + return &_mesa_texformat_i8; + + case GL_YCBCR_MESA: + if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) + return &_mesa_texformat_ycbcr; + else + return &_mesa_texformat_ycbcr_rev; + + case GL_COMPRESSED_RGB_FXT1_3DFX: + return &_mesa_texformat_rgb_fxt1; + case GL_COMPRESSED_RGBA_FXT1_3DFX: + return &_mesa_texformat_rgba_fxt1; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + return &_mesa_texformat_rgb_dxt1; + + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return &_mesa_texformat_rgba_dxt1; + + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + return &_mesa_texformat_rgba_dxt3; + + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + return &_mesa_texformat_rgba_dxt5; + + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + return &_mesa_texformat_z16; + + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + return &_mesa_texformat_z24_s8; + + default: + fprintf(stderr, "unexpected texture format %s in %s\n", + _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); + return NULL; + } + + return NULL; /* never get here */ } diff --git a/src/mesa/state_tracker/st_cb_teximage.h b/src/mesa/state_tracker/st_cb_teximage.h index 462128a844..452ae59d6d 100644 --- a/src/mesa/state_tracker/st_cb_teximage.h +++ b/src/mesa/state_tracker/st_cb_teximage.h @@ -30,14 +30,34 @@ #define ST_CB_TEXIMAGE_H +struct pipe_format_info +{ + GLuint format; + GLenum base_format; + GLubyte red_bits; + GLubyte green_bits; + GLubyte blue_bits; + GLubyte alpha_bits; + GLubyte luminance_bits; + GLubyte intensity_bits; + GLubyte depth_bits; + GLubyte stencil_bits; + GLubyte size; /**< in bytes */ +}; + + +extern const struct pipe_format_info * +st_get_format_info(GLuint format); + + extern GLuint st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, GLenum format, GLenum type); -extern void st_init_cb_teximage( struct st_context *st ); - -extern void st_destroy_cb_teximage( struct st_context *st ); +extern const struct gl_texture_format * +st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, + GLenum format, GLenum type); #endif /* ST_CB_TEXIMAGE_H */ diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 6b361b0a4b..6850d602e2 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -38,6 +38,7 @@ #include "state_tracker/st_context.h" #include "state_tracker/st_cb_texture.h" +#include "state_tracker/st_cb_teximage.h" #include "state_tracker/st_mipmap_tree.h" #include "pipe/p_context.h" @@ -152,155 +153,6 @@ compressed_num_bytes(GLuint mesaFormat) } -/* It works out that this function is fine for all the supported - * hardware. However, there is still a need to map the formats onto - * hardware descriptors. - */ -/* Note that the i915 can actually support many more formats than - * these if we take the step of simply swizzling the colors - * immediately after sampling... - */ -static const struct gl_texture_format * -st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, - GLenum format, GLenum type) -{ -#if 0 - struct intel_context *intel = intel_context(ctx); - const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); -#else - const GLboolean do32bpt = 1; -#endif - - switch (internalFormat) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - if (format == GL_BGRA) { - if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - return &_mesa_texformat_argb8888; - } - else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - return &_mesa_texformat_argb4444; - } - else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - return &_mesa_texformat_argb1555; - } - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return &_mesa_texformat_rgb565; - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case GL_RGBA4: - case GL_RGBA2: - return &_mesa_texformat_argb4444; - - case GL_RGB5_A1: - return &_mesa_texformat_argb1555; - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - return &_mesa_texformat_argb8888; - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; - - case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; - - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - return &_mesa_texformat_z16; - - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_z24_s8; - - default: - fprintf(stderr, "unexpected texture format %s in %s\n", - _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); - return NULL; - } - - return NULL; /* never get here */ -} static GLboolean -- cgit v1.2.3 From 1463c00d4e436b1904011001cea288d03be8226f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 08:35:13 +0100 Subject: Rename (file contains surface format-related functions) --- src/mesa/state_tracker/st_cb_teximage.c | 492 -------------------------------- src/mesa/state_tracker/st_cb_teximage.h | 63 ---- src/mesa/state_tracker/st_format.c | 492 ++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_format.h | 63 ++++ 4 files changed, 555 insertions(+), 555 deletions(-) delete mode 100644 src/mesa/state_tracker/st_cb_teximage.c delete mode 100644 src/mesa/state_tracker/st_cb_teximage.h create mode 100644 src/mesa/state_tracker/st_format.c create mode 100644 src/mesa/state_tracker/st_format.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_teximage.c b/src/mesa/state_tracker/st_cb_teximage.c deleted file mode 100644 index c54602db5d..0000000000 --- a/src/mesa/state_tracker/st_cb_teximage.c +++ /dev/null @@ -1,492 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * Texture Image-related functions. - * \author Brian Paul - */ - -#include "main/imports.h" -#include "main/context.h" -#include "main/texstore.h" -#include "main/texformat.h" -#include "main/enums.h" - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "st_context.h" -#include "st_cb_teximage.h" - - - -/* - * XXX temporary here - */ -const struct pipe_format_info * -st_get_format_info(GLuint format) -{ - static const struct pipe_format_info info[] = { - { - PIPE_FORMAT_U_R8_G8_B8_A8, /* format */ - GL_RGBA, /* base_format */ - 8, 8, 8, 8, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 4 /* size in bytes */ - }, - { - PIPE_FORMAT_U_A8_R8_G8_B8, - GL_RGBA, /* base_format */ - 8, 8, 8, 8, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 4 /* size in bytes */ - }, - { - PIPE_FORMAT_U_A1_R5_G5_B5, - GL_RGBA, /* base_format */ - 5, 5, 5, 1, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 2 /* size in bytes */ - }, - { - PIPE_FORMAT_U_R5_G6_B5, - GL_RGBA, /* base_format */ - 5, 6, 5, 0, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 2 /* size in bytes */ - }, - /* XXX lots more */ - { - PIPE_FORMAT_S8_Z24, - GL_DEPTH_STENCIL_EXT, /* base_format */ - 0, 0, 0, 0, 0, 0, /* color bits */ - 24, 8, /* depth, stencil */ - 4 /* size in bytes */ - } - }; - GLuint i; - - for (i = 0; i < sizeof(info) / sizeof(info[0]); i++) { - if (info[i].format == format) - return info + i; - } - return NULL; -} - - - -/** - * Search list of formats for first RGBA format. - */ -static GLuint -default_rgba_format(const GLuint formats[], GLuint num) -{ - GLuint i; - for (i = 0; i < num; i++) { - if (formats[i] == PIPE_FORMAT_U_R8_G8_B8_A8 || - formats[i] == PIPE_FORMAT_U_A8_R8_G8_B8 || - formats[i] == PIPE_FORMAT_U_R5_G6_B5) { - return formats[i]; - } - } - return PIPE_FORMAT_NONE; -} - - -/** - * Search list of formats for first depth/Z format. - */ -static GLuint -default_depth_format(const GLuint formats[], GLuint num) -{ - GLuint i; - for (i = 0; i < num; i++) { - if (formats[i] == PIPE_FORMAT_U_Z16 || - formats[i] == PIPE_FORMAT_U_Z32 || - formats[i] == PIPE_FORMAT_S8_Z24) { - return formats[i]; - } - } - return PIPE_FORMAT_NONE; -} - - -/** - * Choose the PIPE_FORMAT_ to use for storing a texture image based - * on the user's internalFormat, format and type parameters. - * We query the pipe device for a list of formats which it supports - * and choose from them. - * If we find a device that needs a more intricate selection mechanism, - * this function _could_ get pushed down into the pipe device. - * - * Note: also used for glRenderbufferStorageEXT() - * - * Note: format and type may be GL_NONE (see renderbuffers) - * - * \return PIPE_FORMAT_NONE if error/problem. - */ -GLuint -st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, - GLenum format, GLenum type) -{ - const GLuint *supported; - GLboolean allow[PIPE_FORMAT_COUNT]; - GLuint i, n; - - /* query supported formats and fill in bool allow[] table */ - supported = pipe->supported_formats(pipe, &n); - assert(n < PIPE_FORMAT_COUNT); /* sanity check */ - memset(allow, 0, sizeof(allow)); - for (i = 0; i < n; i++) { - allow[supported[i]] = 1; - } - - switch (internalFormat) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - if (format == GL_BGRA) { - if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - if (allow[PIPE_FORMAT_U_A8_R8_G8_B8]) - return PIPE_FORMAT_U_A8_R8_G8_B8; - } - else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - if (allow[PIPE_FORMAT_U_A4_R4_G4_B4]) - return PIPE_FORMAT_U_A4_R4_G4_B4; - } - else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) - return PIPE_FORMAT_U_A1_R5_G5_B5; - } - } - return default_rgba_format(supported, n); - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - if (allow[PIPE_FORMAT_U_R5_G6_B5]) - return PIPE_FORMAT_U_R5_G6_B5; - } - return default_rgba_format(supported, n); - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return default_rgba_format(supported, n); - - case GL_RGBA4: - case GL_RGBA2: - if (allow[PIPE_FORMAT_U_A4_R4_G4_B4]) - return PIPE_FORMAT_U_A4_R4_G4_B4; - return default_rgba_format(supported, n); - - case GL_RGB5_A1: - if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) - return PIPE_FORMAT_U_A1_R5_G5_B5; - return default_rgba_format(supported, n); - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - return default_rgba_format(supported, n); - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: - if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) - return PIPE_FORMAT_U_A1_R5_G5_B5; - return default_rgba_format(supported, n); - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: - if (allow[PIPE_FORMAT_U_A8]) - return PIPE_FORMAT_U_A8; - return default_rgba_format(supported, n); - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: - if (allow[PIPE_FORMAT_U_A8]) - return PIPE_FORMAT_U_A8; - return default_rgba_format(supported, n); - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - if (allow[PIPE_FORMAT_U_L8_A8]) - return PIPE_FORMAT_U_L8_A8; - return default_rgba_format(supported, n); - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: - if (allow[PIPE_FORMAT_U_I8]) - return PIPE_FORMAT_U_I8; - return default_rgba_format(supported, n); - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) { - if (allow[PIPE_FORMAT_YCBCR]) - return PIPE_FORMAT_YCBCR; - } - else { - if (allow[PIPE_FORMAT_YCBCR_REV]) - return PIPE_FORMAT_YCBCR_REV; - } - return PIPE_FORMAT_NONE; - -#if 0 - case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; -#endif - - case GL_DEPTH_COMPONENT16: - if (allow[PIPE_FORMAT_U_Z16]) - return PIPE_FORMAT_U_Z16; - /* fall-through */ - case GL_DEPTH_COMPONENT24: - if (allow[PIPE_FORMAT_S8_Z24]) - return PIPE_FORMAT_S8_Z24; - /* fall-through */ - case GL_DEPTH_COMPONENT32: - if (allow[PIPE_FORMAT_U_Z32]) - return PIPE_FORMAT_U_Z32; - /* fall-through */ - case GL_DEPTH_COMPONENT: - return default_depth_format(supported, n); - - case GL_STENCIL_INDEX: - case GL_STENCIL_INDEX1_EXT: - case GL_STENCIL_INDEX4_EXT: - case GL_STENCIL_INDEX8_EXT: - case GL_STENCIL_INDEX16_EXT: - if (allow[PIPE_FORMAT_U_S8]) - return PIPE_FORMAT_U_S8; - if (allow[PIPE_FORMAT_S8_Z24]) - return PIPE_FORMAT_S8_Z24; - return PIPE_FORMAT_NONE; - - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - if (allow[PIPE_FORMAT_S8_Z24]) - return PIPE_FORMAT_S8_Z24; - return PIPE_FORMAT_NONE; - - default: - return PIPE_FORMAT_NONE; - } -} - - - -/* It works out that this function is fine for all the supported - * hardware. However, there is still a need to map the formats onto - * hardware descriptors. - */ -/* Note that the i915 can actually support many more formats than - * these if we take the step of simply swizzling the colors - * immediately after sampling... - */ -const struct gl_texture_format * -st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, - GLenum format, GLenum type) -{ -#if 0 - struct intel_context *intel = intel_context(ctx); - const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); -#else - const GLboolean do32bpt = 1; -#endif - - switch (internalFormat) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - if (format == GL_BGRA) { - if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - return &_mesa_texformat_argb8888; - } - else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - return &_mesa_texformat_argb4444; - } - else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - return &_mesa_texformat_argb1555; - } - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return &_mesa_texformat_rgb565; - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case GL_RGBA4: - case GL_RGBA2: - return &_mesa_texformat_argb4444; - - case GL_RGB5_A1: - return &_mesa_texformat_argb1555; - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - return &_mesa_texformat_argb8888; - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; - - case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; - - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - return &_mesa_texformat_z16; - - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_z24_s8; - - default: - fprintf(stderr, "unexpected texture format %s in %s\n", - _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); - return NULL; - } - - return NULL; /* never get here */ -} diff --git a/src/mesa/state_tracker/st_cb_teximage.h b/src/mesa/state_tracker/st_cb_teximage.h deleted file mode 100644 index 452ae59d6d..0000000000 --- a/src/mesa/state_tracker/st_cb_teximage.h +++ /dev/null @@ -1,63 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef ST_CB_TEXIMAGE_H -#define ST_CB_TEXIMAGE_H - - -struct pipe_format_info -{ - GLuint format; - GLenum base_format; - GLubyte red_bits; - GLubyte green_bits; - GLubyte blue_bits; - GLubyte alpha_bits; - GLubyte luminance_bits; - GLubyte intensity_bits; - GLubyte depth_bits; - GLubyte stencil_bits; - GLubyte size; /**< in bytes */ -}; - - -extern const struct pipe_format_info * -st_get_format_info(GLuint format); - - -extern GLuint -st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, - GLenum format, GLenum type); - - -extern const struct gl_texture_format * -st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, - GLenum format, GLenum type); - - -#endif /* ST_CB_TEXIMAGE_H */ diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c new file mode 100644 index 0000000000..c54602db5d --- /dev/null +++ b/src/mesa/state_tracker/st_format.c @@ -0,0 +1,492 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Texture Image-related functions. + * \author Brian Paul + */ + +#include "main/imports.h" +#include "main/context.h" +#include "main/texstore.h" +#include "main/texformat.h" +#include "main/enums.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "st_context.h" +#include "st_cb_teximage.h" + + + +/* + * XXX temporary here + */ +const struct pipe_format_info * +st_get_format_info(GLuint format) +{ + static const struct pipe_format_info info[] = { + { + PIPE_FORMAT_U_R8_G8_B8_A8, /* format */ + GL_RGBA, /* base_format */ + 8, 8, 8, 8, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 4 /* size in bytes */ + }, + { + PIPE_FORMAT_U_A8_R8_G8_B8, + GL_RGBA, /* base_format */ + 8, 8, 8, 8, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 4 /* size in bytes */ + }, + { + PIPE_FORMAT_U_A1_R5_G5_B5, + GL_RGBA, /* base_format */ + 5, 5, 5, 1, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 2 /* size in bytes */ + }, + { + PIPE_FORMAT_U_R5_G6_B5, + GL_RGBA, /* base_format */ + 5, 6, 5, 0, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 2 /* size in bytes */ + }, + /* XXX lots more */ + { + PIPE_FORMAT_S8_Z24, + GL_DEPTH_STENCIL_EXT, /* base_format */ + 0, 0, 0, 0, 0, 0, /* color bits */ + 24, 8, /* depth, stencil */ + 4 /* size in bytes */ + } + }; + GLuint i; + + for (i = 0; i < sizeof(info) / sizeof(info[0]); i++) { + if (info[i].format == format) + return info + i; + } + return NULL; +} + + + +/** + * Search list of formats for first RGBA format. + */ +static GLuint +default_rgba_format(const GLuint formats[], GLuint num) +{ + GLuint i; + for (i = 0; i < num; i++) { + if (formats[i] == PIPE_FORMAT_U_R8_G8_B8_A8 || + formats[i] == PIPE_FORMAT_U_A8_R8_G8_B8 || + formats[i] == PIPE_FORMAT_U_R5_G6_B5) { + return formats[i]; + } + } + return PIPE_FORMAT_NONE; +} + + +/** + * Search list of formats for first depth/Z format. + */ +static GLuint +default_depth_format(const GLuint formats[], GLuint num) +{ + GLuint i; + for (i = 0; i < num; i++) { + if (formats[i] == PIPE_FORMAT_U_Z16 || + formats[i] == PIPE_FORMAT_U_Z32 || + formats[i] == PIPE_FORMAT_S8_Z24) { + return formats[i]; + } + } + return PIPE_FORMAT_NONE; +} + + +/** + * Choose the PIPE_FORMAT_ to use for storing a texture image based + * on the user's internalFormat, format and type parameters. + * We query the pipe device for a list of formats which it supports + * and choose from them. + * If we find a device that needs a more intricate selection mechanism, + * this function _could_ get pushed down into the pipe device. + * + * Note: also used for glRenderbufferStorageEXT() + * + * Note: format and type may be GL_NONE (see renderbuffers) + * + * \return PIPE_FORMAT_NONE if error/problem. + */ +GLuint +st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, + GLenum format, GLenum type) +{ + const GLuint *supported; + GLboolean allow[PIPE_FORMAT_COUNT]; + GLuint i, n; + + /* query supported formats and fill in bool allow[] table */ + supported = pipe->supported_formats(pipe, &n); + assert(n < PIPE_FORMAT_COUNT); /* sanity check */ + memset(allow, 0, sizeof(allow)); + for (i = 0; i < n; i++) { + allow[supported[i]] = 1; + } + + switch (internalFormat) { + case 4: + case GL_RGBA: + case GL_COMPRESSED_RGBA: + if (format == GL_BGRA) { + if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { + if (allow[PIPE_FORMAT_U_A8_R8_G8_B8]) + return PIPE_FORMAT_U_A8_R8_G8_B8; + } + else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { + if (allow[PIPE_FORMAT_U_A4_R4_G4_B4]) + return PIPE_FORMAT_U_A4_R4_G4_B4; + } + else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { + if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) + return PIPE_FORMAT_U_A1_R5_G5_B5; + } + } + return default_rgba_format(supported, n); + + case 3: + case GL_RGB: + case GL_COMPRESSED_RGB: + if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { + if (allow[PIPE_FORMAT_U_R5_G6_B5]) + return PIPE_FORMAT_U_R5_G6_B5; + } + return default_rgba_format(supported, n); + + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + return default_rgba_format(supported, n); + + case GL_RGBA4: + case GL_RGBA2: + if (allow[PIPE_FORMAT_U_A4_R4_G4_B4]) + return PIPE_FORMAT_U_A4_R4_G4_B4; + return default_rgba_format(supported, n); + + case GL_RGB5_A1: + if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) + return PIPE_FORMAT_U_A1_R5_G5_B5; + return default_rgba_format(supported, n); + + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + return default_rgba_format(supported, n); + + case GL_RGB5: + case GL_RGB4: + case GL_R3_G3_B2: + if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) + return PIPE_FORMAT_U_A1_R5_G5_B5; + return default_rgba_format(supported, n); + + case GL_ALPHA: + case GL_ALPHA4: + case GL_ALPHA8: + case GL_ALPHA12: + case GL_ALPHA16: + case GL_COMPRESSED_ALPHA: + if (allow[PIPE_FORMAT_U_A8]) + return PIPE_FORMAT_U_A8; + return default_rgba_format(supported, n); + + case 1: + case GL_LUMINANCE: + case GL_LUMINANCE4: + case GL_LUMINANCE8: + case GL_LUMINANCE12: + case GL_LUMINANCE16: + case GL_COMPRESSED_LUMINANCE: + if (allow[PIPE_FORMAT_U_A8]) + return PIPE_FORMAT_U_A8; + return default_rgba_format(supported, n); + + case 2: + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE4_ALPHA4: + case GL_LUMINANCE6_ALPHA2: + case GL_LUMINANCE8_ALPHA8: + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: + case GL_COMPRESSED_LUMINANCE_ALPHA: + if (allow[PIPE_FORMAT_U_L8_A8]) + return PIPE_FORMAT_U_L8_A8; + return default_rgba_format(supported, n); + + case GL_INTENSITY: + case GL_INTENSITY4: + case GL_INTENSITY8: + case GL_INTENSITY12: + case GL_INTENSITY16: + case GL_COMPRESSED_INTENSITY: + if (allow[PIPE_FORMAT_U_I8]) + return PIPE_FORMAT_U_I8; + return default_rgba_format(supported, n); + + case GL_YCBCR_MESA: + if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) { + if (allow[PIPE_FORMAT_YCBCR]) + return PIPE_FORMAT_YCBCR; + } + else { + if (allow[PIPE_FORMAT_YCBCR_REV]) + return PIPE_FORMAT_YCBCR_REV; + } + return PIPE_FORMAT_NONE; + +#if 0 + case GL_COMPRESSED_RGB_FXT1_3DFX: + return &_mesa_texformat_rgb_fxt1; + case GL_COMPRESSED_RGBA_FXT1_3DFX: + return &_mesa_texformat_rgba_fxt1; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + return &_mesa_texformat_rgb_dxt1; + + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return &_mesa_texformat_rgba_dxt1; + + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + return &_mesa_texformat_rgba_dxt3; + + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + return &_mesa_texformat_rgba_dxt5; +#endif + + case GL_DEPTH_COMPONENT16: + if (allow[PIPE_FORMAT_U_Z16]) + return PIPE_FORMAT_U_Z16; + /* fall-through */ + case GL_DEPTH_COMPONENT24: + if (allow[PIPE_FORMAT_S8_Z24]) + return PIPE_FORMAT_S8_Z24; + /* fall-through */ + case GL_DEPTH_COMPONENT32: + if (allow[PIPE_FORMAT_U_Z32]) + return PIPE_FORMAT_U_Z32; + /* fall-through */ + case GL_DEPTH_COMPONENT: + return default_depth_format(supported, n); + + case GL_STENCIL_INDEX: + case GL_STENCIL_INDEX1_EXT: + case GL_STENCIL_INDEX4_EXT: + case GL_STENCIL_INDEX8_EXT: + case GL_STENCIL_INDEX16_EXT: + if (allow[PIPE_FORMAT_U_S8]) + return PIPE_FORMAT_U_S8; + if (allow[PIPE_FORMAT_S8_Z24]) + return PIPE_FORMAT_S8_Z24; + return PIPE_FORMAT_NONE; + + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + if (allow[PIPE_FORMAT_S8_Z24]) + return PIPE_FORMAT_S8_Z24; + return PIPE_FORMAT_NONE; + + default: + return PIPE_FORMAT_NONE; + } +} + + + +/* It works out that this function is fine for all the supported + * hardware. However, there is still a need to map the formats onto + * hardware descriptors. + */ +/* Note that the i915 can actually support many more formats than + * these if we take the step of simply swizzling the colors + * immediately after sampling... + */ +const struct gl_texture_format * +st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, + GLenum format, GLenum type) +{ +#if 0 + struct intel_context *intel = intel_context(ctx); + const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); +#else + const GLboolean do32bpt = 1; +#endif + + switch (internalFormat) { + case 4: + case GL_RGBA: + case GL_COMPRESSED_RGBA: + if (format == GL_BGRA) { + if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { + return &_mesa_texformat_argb8888; + } + else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { + return &_mesa_texformat_argb4444; + } + else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { + return &_mesa_texformat_argb1555; + } + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case 3: + case GL_RGB: + case GL_COMPRESSED_RGB: + if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { + return &_mesa_texformat_rgb565; + } + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; + + case GL_RGBA8: + case GL_RGB10_A2: + case GL_RGBA12: + case GL_RGBA16: + return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + + case GL_RGBA4: + case GL_RGBA2: + return &_mesa_texformat_argb4444; + + case GL_RGB5_A1: + return &_mesa_texformat_argb1555; + + case GL_RGB8: + case GL_RGB10: + case GL_RGB12: + case GL_RGB16: + return &_mesa_texformat_argb8888; + + case GL_RGB5: + case GL_RGB4: + case GL_R3_G3_B2: + return &_mesa_texformat_rgb565; + + case GL_ALPHA: + case GL_ALPHA4: + case GL_ALPHA8: + case GL_ALPHA12: + case GL_ALPHA16: + case GL_COMPRESSED_ALPHA: + return &_mesa_texformat_a8; + + case 1: + case GL_LUMINANCE: + case GL_LUMINANCE4: + case GL_LUMINANCE8: + case GL_LUMINANCE12: + case GL_LUMINANCE16: + case GL_COMPRESSED_LUMINANCE: + return &_mesa_texformat_l8; + + case 2: + case GL_LUMINANCE_ALPHA: + case GL_LUMINANCE4_ALPHA4: + case GL_LUMINANCE6_ALPHA2: + case GL_LUMINANCE8_ALPHA8: + case GL_LUMINANCE12_ALPHA4: + case GL_LUMINANCE12_ALPHA12: + case GL_LUMINANCE16_ALPHA16: + case GL_COMPRESSED_LUMINANCE_ALPHA: + return &_mesa_texformat_al88; + + case GL_INTENSITY: + case GL_INTENSITY4: + case GL_INTENSITY8: + case GL_INTENSITY12: + case GL_INTENSITY16: + case GL_COMPRESSED_INTENSITY: + return &_mesa_texformat_i8; + + case GL_YCBCR_MESA: + if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) + return &_mesa_texformat_ycbcr; + else + return &_mesa_texformat_ycbcr_rev; + + case GL_COMPRESSED_RGB_FXT1_3DFX: + return &_mesa_texformat_rgb_fxt1; + case GL_COMPRESSED_RGBA_FXT1_3DFX: + return &_mesa_texformat_rgba_fxt1; + + case GL_RGB_S3TC: + case GL_RGB4_S3TC: + case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + return &_mesa_texformat_rgb_dxt1; + + case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + return &_mesa_texformat_rgba_dxt1; + + case GL_RGBA_S3TC: + case GL_RGBA4_S3TC: + case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + return &_mesa_texformat_rgba_dxt3; + + case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + return &_mesa_texformat_rgba_dxt5; + + case GL_DEPTH_COMPONENT: + case GL_DEPTH_COMPONENT16: + case GL_DEPTH_COMPONENT24: + case GL_DEPTH_COMPONENT32: + return &_mesa_texformat_z16; + + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + return &_mesa_texformat_z24_s8; + + default: + fprintf(stderr, "unexpected texture format %s in %s\n", + _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); + return NULL; + } + + return NULL; /* never get here */ +} diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h new file mode 100644 index 0000000000..452ae59d6d --- /dev/null +++ b/src/mesa/state_tracker/st_format.h @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_TEXIMAGE_H +#define ST_CB_TEXIMAGE_H + + +struct pipe_format_info +{ + GLuint format; + GLenum base_format; + GLubyte red_bits; + GLubyte green_bits; + GLubyte blue_bits; + GLubyte alpha_bits; + GLubyte luminance_bits; + GLubyte intensity_bits; + GLubyte depth_bits; + GLubyte stencil_bits; + GLubyte size; /**< in bytes */ +}; + + +extern const struct pipe_format_info * +st_get_format_info(GLuint format); + + +extern GLuint +st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, + GLenum format, GLenum type); + + +extern const struct gl_texture_format * +st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, + GLenum format, GLenum type); + + +#endif /* ST_CB_TEXIMAGE_H */ -- cgit v1.2.3 From f8ab24760d0d3f07e9ee81c98207ddf92dfe74da Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 08:37:15 +0100 Subject: rename st_cb_teximage.h st_format.h --- src/mesa/sources | 4 ++-- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_cb_fbo.c | 2 +- src/mesa/state_tracker/st_cb_texture.c | 2 +- src/mesa/state_tracker/st_format.c | 2 +- 5 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 9bcb07c6b4..be81e102ed 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -196,11 +196,11 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_drawpixels.c \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_program.c \ - state_tracker/st_cb_teximage.c \ state_tracker/st_cb_texture.c \ state_tracker/st_cb_bufferobjects.c \ - state_tracker/st_draw.c \ state_tracker/st_context.c \ + state_tracker/st_draw.c \ + state_tracker/st_format.c \ state_tracker/st_mipmap_tree.c SHADER_SOURCES = \ diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 6b7cef2b2a..20990fe80a 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -38,7 +38,7 @@ #include "st_program.h" #include "st_cb_drawpixels.h" #include "st_cb_texture.h" -#include "st_cb_teximage.h" +#include "st_format.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 8443c9ea1b..bee7474e38 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -43,7 +43,7 @@ #include "pipe/p_defines.h" #include "st_context.h" #include "st_cb_fbo.h" -#include "st_cb_teximage.h" +#include "st_format.h" diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 6850d602e2..22c1b4baff 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -38,7 +38,7 @@ #include "state_tracker/st_context.h" #include "state_tracker/st_cb_texture.h" -#include "state_tracker/st_cb_teximage.h" +#include "state_tracker/st_format.h" #include "state_tracker/st_mipmap_tree.h" #include "pipe/p_context.h" diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index c54602db5d..4f36c84d4e 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -40,7 +40,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "st_context.h" -#include "st_cb_teximage.h" +#include "st_format.h" -- cgit v1.2.3 From c7dd9bf870c7d3002524abab58981f89b8f434a8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 08:38:31 +0100 Subject: return (softpipe) in GL_RENDERER string - temporarily --- src/mesa/drivers/x11/xm_dd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index 6eb6d32d19..22eae51ca9 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -740,7 +740,7 @@ get_string( GLcontext *ctx, GLenum name ) #ifdef XFree86Server return (const GLubyte *) "Mesa GLX Indirect"; #else - return (const GLubyte *) "Mesa X11"; + return (const GLubyte *) "Mesa X11 (softpipe)"; #endif case GL_VENDOR: #ifdef XFree86Server -- cgit v1.2.3 From 3654922b333faf07a0543b2b4d2b19a3d4e4f1fd Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 10:06:00 -0600 Subject: flesh out w/ more formats, y=0=top convention --- src/mesa/pipe/softpipe/sp_surface.c | 264 ++++++++++++++++-------------------- 1 file changed, 116 insertions(+), 148 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 852aeeeed1..054e4188e9 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -36,120 +36,130 @@ /** * Softpipe surface functions. * Basically, create surface of a particular type, then plug in default - * read/write_quad functions. + * read/write_quad and get/put_tile() functions. * Note that these quad funcs assume the buffer/region is in a linear - * layout with Y=0=bottom. - * If we had swizzled/AOS buffers the read/write functions could be + * layout with Y=0=top. + * If we had swizzled/AOS buffers the read/write quad functions could be * simplified a lot.... */ -#if 000 /* OLD... should be recycled... */ -static void rgba8_read_quad_f( struct softpipe_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - rgba[k][0] = ptr[0] * (1.0 / 255.0); - rgba[k][1] = ptr[1] * (1.0 / 255.0); - rgba[k][2] = ptr[2] * (1.0 / 255.0); - rgba[k][3] = ptr[3] * (1.0 / 255.0); - } - } -} -static void rgba8_read_quad_f_swz( struct softpipe_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) -{ - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - rrrr[0][k] = ptr[0] * (1.0 / 255.0); - rrrr[1][k] = ptr[1] * (1.0 / 255.0); - rrrr[2][k] = ptr[2] * (1.0 / 255.0); - rrrr[3][k] = ptr[3] * (1.0 / 255.0); - } - } -} +/*** PIPE_FORMAT_U_A8_R8_G8_B8 ***/ -static void rgba8_write_quad_f( struct softpipe_surface *gs, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ) +static void +a8r8g8b8_read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) { - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - ptr[0] = rgba[k][0] * 255.0; - ptr[1] = rgba[k][1] * 255.0; - ptr[2] = rgba[k][2] * 255.0; - ptr[3] = rgba[k][3] * 255.0; + const GLuint *src + = ((const GLuint *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + GLuint i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); + assert(x < sps->surface.width - 1); + assert(y < sps->surface.height - 1); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + const GLuint p = src[j]; + rrrr[0][i * 2 + j] = UBYTE_TO_FLOAT((p >> 16) & 0xff); /*R*/ + rrrr[1][i * 2 + j] = UBYTE_TO_FLOAT((p >> 8) & 0xff); /*G*/ + rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT((p ) & 0xff); /*B*/ + rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT((p >> 24) & 0xff); /*A*/ } + src += sps->surface.region->pitch; } } -static void rgba8_write_quad_f_swz( struct softpipe_surface *gs, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ) +static void +a8r8g8b8_write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, + GLfloat (*rrrr)[QUAD_SIZE]) { - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - ptr[0] = rrrr[0][k] * 255.0; - ptr[1] = rrrr[1][k] * 255.0; - ptr[2] = rrrr[2][k] * 255.0; - ptr[3] = rrrr[3][k] * 255.0; + GLuint *dst + = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + GLuint i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + GLubyte r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/ + UNCLAMPED_FLOAT_TO_UBYTE(g, rrrr[1][i * 2 + j]); /*G*/ + UNCLAMPED_FLOAT_TO_UBYTE(b, rrrr[2][i * 2 + j]); /*B*/ + UNCLAMPED_FLOAT_TO_UBYTE(a, rrrr[3][i * 2 + j]); /*A*/ + dst[j] = (a << 24) | (r << 16) | (g << 8) | b; } + dst += sps->surface.region->pitch; } } -static void rgba8_read_quad_ub( struct softpipe_surface *gs, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ) +static void +a8r8g8b8_get_tile(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) { - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - rgba[k][0] = ptr[0]; - rgba[k][1] = ptr[1]; - rgba[k][2] = ptr[2]; - rgba[k][3] = ptr[3]; + const GLuint *src + = ((const GLuint *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + GLuint i, j; + + assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const GLuint pixel = src[j]; + p[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + p[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + p[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + p[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + p += 4; } + src += ps->region->pitch; } } -static void rgba8_write_quad_ub( struct softpipe_surface *gs, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ) + +/*** PIPE_FORMAT_U_A1_R5_G5_B5 ***/ + +static void +a1r5g5b5_get_tile(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) { - GLuint i, j, k = 0; - - for (i = 0; i < 2; i++) { - for (j = 0; j < 2; j++, k++) { - GLubyte *ptr = gs->surface.ptr + (y+i) * gs->surface.stride + (x+j) * 4; - ptr[0] = rgba[k][0]; - ptr[1] = rgba[k][1]; - ptr[2] = rgba[k][2]; - ptr[3] = rgba[k][3]; + const GLushort *src + = ((const GLushort *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + GLuint i, j; + + assert(ps->format == PIPE_FORMAT_U_A1_R5_G5_B5); + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const GLushort pixel = src[j]; + p[0] = ((pixel >> 10) & 0x1f) * (1.0 / 31); + p[1] = ((pixel >> 5) & 0x1f) * (1.0 / 31); + p[2] = ((pixel ) & 0x1f) * (1.0 / 31); + p[3] = ((pixel >> 15) ); + p += 4; } + src += ps->region->pitch; } } -#endif +/*** PIPE_FORMAT_U_Z16 ***/ static void z16_read_quad_z(struct softpipe_surface *sps, @@ -184,6 +194,9 @@ z16_write_quad_z(struct softpipe_surface *sps, dst[1] = zzzz[3]; } + +/*** PIPE_FORMAT_U_Z32 ***/ + static void z32_read_quad_z(struct softpipe_surface *sps, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) @@ -215,6 +228,9 @@ z32_write_quad_z(struct softpipe_surface *sps, dst[1] = zzzz[3]; } + +/*** PIPE_FORMAT_S8_Z24 ***/ + static void s8z24_read_quad_z(struct softpipe_surface *sps, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) @@ -225,6 +241,7 @@ s8z24_read_quad_z(struct softpipe_surface *sps, assert(sps->surface.format == PIPE_FORMAT_S8_Z24); + /* extract lower three bytes */ zzzz[0] = src[0] & mask; zzzz[1] = src[1] & mask; src += sps->surface.region->pitch; @@ -282,6 +299,8 @@ s8z24_write_quad_stencil(struct softpipe_surface *sps, } +/*** PIPE_FORMAT_U_S8 ***/ + static void s8_read_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) @@ -315,63 +334,17 @@ s8_write_quad_stencil(struct softpipe_surface *sps, } - -static void -a8r8g8b8_get_tile(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) -{ - const GLuint *src - = ((const GLuint *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; - GLuint i, j; -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - p[0] = UBYTE_TO_FLOAT((src[j] >> 16) & 0xff); - p[1] = UBYTE_TO_FLOAT((src[j] >> 8) & 0xff); - p[2] = UBYTE_TO_FLOAT((src[j] >> 0) & 0xff); - p[3] = UBYTE_TO_FLOAT((src[j] >> 24) & 0xff); - p += 4; - } - src += ps->region->pitch; - } -} - - -static void -a1r5g5b5_get_tile(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) -{ - const GLushort *src - = ((const GLushort *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; - GLuint i, j; - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - p[0] = ((src[j] >> 10) & 0x1f) * (1.0 / 31); - p[1] = ((src[j] >> 5) & 0x1f) * (1.0 / 31); - p[2] = ((src[j] ) & 0x1f) * (1.0 / 31); - p[3] = src[j] >> 15; - p += 4; - } - src += ps->region->pitch; - } -} - - void softpipe_init_surface_funcs(struct softpipe_surface *sps) { switch (sps->surface.format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + sps->read_quad_f_swz = a8r8g8b8_read_quad_f_swz; + sps->write_quad_f_swz = a8r8g8b8_write_quad_f_swz; + sps->surface.get_tile = a8r8g8b8_get_tile; + case PIPE_FORMAT_U_A1_R5_G5_B5: + sps->surface.get_tile = a1r5g5b5_get_tile; + break; case PIPE_FORMAT_U_Z16: sps->read_quad_z = z16_read_quad_z; sps->write_quad_z = z16_write_quad_z; @@ -390,27 +363,22 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) sps->read_quad_stencil = s8_read_quad_stencil; sps->write_quad_stencil = s8_write_quad_stencil; break; - case PIPE_FORMAT_U_A8_R8_G8_B8: - sps->surface.get_tile = a8r8g8b8_get_tile; - break; - case PIPE_FORMAT_U_A1_R5_G5_B5: - sps->surface.get_tile = a1r5g5b5_get_tile; - break; default: assert(0); - ; } } static struct pipe_surface * -sp_surface_alloc(struct pipe_context *pipe, GLenum format) +softpipe_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) { struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); if (!sps) return NULL; - sps->surface.format = format; + assert(pipeFormat < PIPE_FORMAT_COUNT); + + sps->surface.format = pipeFormat; sps->surface.refcount = 1; softpipe_init_surface_funcs(sps); @@ -462,5 +430,5 @@ softpipe_get_tex_surface(struct pipe_context *pipe, void sp_init_surface_functions(struct softpipe_context *sp) { - sp->pipe.surface_alloc = sp_surface_alloc; + sp->pipe.surface_alloc = softpipe_surface_alloc; } -- cgit v1.2.3 From 12b0aa129e95920774a46583cf7e40236d380007 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 10:06:57 -0600 Subject: we can now just rely on softpipe for surface creation, quad_read/write, etc --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 +- src/mesa/drivers/dri/intel_winsys/intel_surface.c | 218 ---------------------- src/mesa/drivers/dri/intel_winsys/intel_surface.h | 5 - 3 files changed, 1 insertion(+), 224 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 054f2c8421..d418796d3c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -404,7 +404,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, */ if (!getenv("INTEL_HW")) { intel->pipe = intel_create_softpipe( intel ); - intel->pipe->surface_alloc = intel_new_surface; + /* use default softpipe function for surface_alloc() */ intel->pipe->supported_formats = intel_supported_formats; } else { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.c b/src/mesa/drivers/dri/intel_winsys/intel_surface.c index ef2e62c809..683a81fb63 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.c @@ -31,233 +31,15 @@ */ -#include "glheader.h" -#include "main/macros.h" - -#include "intel_screen.h" #include "intel_context.h" -#include "intel_buffers.h" #include "intel_surface.h" #include "pipe/p_state.h" -#include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/softpipe/sp_surface.h" -/** - * Note: the arithmetic/addressing in these functions is a little - * tricky since we need to invert the Y axis. - */ - - -static void -read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) -{ - const GLint bytesPerRow = sps->surface.region->pitch * sps->surface.region->cpp; - const GLint invY = sps->surface.height - y - 1; - const GLubyte *src = sps->surface.region->map + invY * bytesPerRow + x * sps->surface.region->cpp; - GLfloat *dst = (GLfloat *) rrrr; - GLubyte temp[16]; - GLuint j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - assert(x < sps->surface.width - 1); - assert(y < sps->surface.height - 1); - - memcpy(temp + 8, src, 8); - memcpy(temp + 0, src + bytesPerRow, 8); - - for (j = 0; j < 4; j++) { - dst[0 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 2]); /*R*/ - dst[1 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 1]); /*G*/ - dst[2 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 0]); /*B*/ - dst[3 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 3]); /*A*/ - } -} - - -static void -write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) -{ - const GLfloat *src = (const GLfloat *) rrrr; - const GLint bytesPerRow = sps->surface.region->pitch * sps->surface.region->cpp; - const GLint invY = sps->surface.height - y - 1; - GLubyte *dst = sps->surface.region->map + invY * bytesPerRow + x * sps->surface.region->cpp; - GLubyte temp[16]; - GLuint j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - - for (j = 0; j < 4; j++) { - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 2], src[0 * 4 + j]); /*R*/ - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 1], src[1 * 4 + j]); /*G*/ - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 0], src[2 * 4 + j]); /*B*/ - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 3], src[3 * 4 + j]); /*A*/ - } - - memcpy(dst, temp + 8, 8); - memcpy(dst + bytesPerRow, temp + 0, 8); -} - - - -static void -read_quad_z24(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) -{ - static const GLuint mask = 0xffffff; - const GLint invY = sps->surface.height - y - 1; - const GLuint *src - = (GLuint *) (sps->surface.region->map - + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - /* extract lower three bytes */ - zzzz[0] = src[0] & mask; - zzzz[1] = src[1] & mask; - zzzz[2] = src[-sps->surface.region->pitch] & mask; - zzzz[3] = src[-sps->surface.region->pitch + 1] & mask; -} - -static void -write_quad_z24(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) -{ - static const GLuint mask = 0xff000000; - const GLint invY = sps->surface.height - y - 1; - GLuint *dst - = (GLuint *) (sps->surface.region->map - + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - /* write lower three bytes */ - dst[0] = (dst[0] & mask) | zzzz[0]; - dst[1] = (dst[1] & mask) | zzzz[1]; - dst -= sps->surface.region->pitch; - dst[0] = (dst[0] & mask) | zzzz[2]; - dst[1] = (dst[1] & mask) | zzzz[3]; -} - - -static void -read_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) -{ - const GLint invY = sps->surface.height - y - 1; - const GLuint *src = (const GLuint *) (sps->surface.region->map - + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - /* extract high byte */ - ssss[0] = src[0] >> 24; - ssss[1] = src[1] >> 24; - src -= sps->surface.region->pitch; - ssss[2] = src[0] >> 24; - ssss[3] = src[1] >> 24; -} - -static void -write_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) -{ - static const GLuint mask = 0x00ffffff; - const GLint invY = sps->surface.height - y - 1; - GLuint *dst = (GLuint *) (sps->surface.region->map - + (invY * sps->surface.region->pitch + x) * sps->surface.region->cpp); - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - /* write high byte */ - dst[0] = (dst[0] & mask) | (ssss[0] << 24); - dst[1] = (dst[1] & mask) | (ssss[1] << 24); - dst -= sps->surface.region->pitch; - dst[0] = (dst[0] & mask) | (ssss[2] << 24); - dst[1] = (dst[1] & mask) | (ssss[3] << 24); -} - - - - -static void -a8r8g8b8_get_tile(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) -{ - const GLuint *src - = ((const GLuint *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; - GLuint i, j; -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - p[0] = UBYTE_TO_FLOAT((src[j] >> 16) & 0xff); - p[1] = UBYTE_TO_FLOAT((src[j] >> 8) & 0xff); - p[2] = UBYTE_TO_FLOAT((src[j] >> 0) & 0xff); - p[3] = UBYTE_TO_FLOAT((src[j] >> 24) & 0xff); - p += 4; - } - src += ps->region->pitch; - } -} - - - -/** - * Create a new surface of the specified format. - * This also plugs in the quad/tile read/write functions that - * will be used to access the buffer for rendering/texturing/etc. - */ -struct pipe_surface * -intel_new_surface(struct pipe_context *pipe, GLuint pipeFormat) -{ - struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); - if (!sps) - return NULL; - - sps->surface.width = 0; /* set by renderbuffer::AllocStorage() */ - sps->surface.height = 0; - sps->surface.refcount = 1; - sps->surface.format = pipeFormat; - - switch (pipeFormat) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - sps->read_quad_f_swz = read_quad_f_swz; - sps->write_quad_f_swz = write_quad_f_swz; - sps->surface.get_tile = a8r8g8b8_get_tile; - break; - case PIPE_FORMAT_U_R5_G6_B5: - break; - case PIPE_FORMAT_U_Z16: - break; - case PIPE_FORMAT_S8_Z24: - sps->read_quad_z = read_quad_z24; - sps->write_quad_z = write_quad_z24; - sps->read_quad_stencil = read_quad_stencil; - sps->write_quad_stencil = write_quad_stencil; - break; - default: - assert(0); - } - - return &sps->surface; -} - - /** * Return list of surface formats supported by this driver. */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.h b/src/mesa/drivers/dri/intel_winsys/intel_surface.h index 83a9bff2db..0987e963de 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_surface.h @@ -29,11 +29,6 @@ #ifndef INTEL_SURFACE_H #define INTEL_SURFACE_H - -extern struct pipe_surface * -intel_new_surface(struct pipe_context *pipe, GLuint pipeFormat); - - extern const GLuint * intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats); -- cgit v1.2.3 From 8639a543193b65ef5b39da2cce843431bb77a618 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 10:09:12 -0600 Subject: move intel_supported_formats() into intel_context.c --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 19 ++++++++++++++++++- 1 file changed, 18 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index d418796d3c..8c61f0cf3c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -46,10 +46,10 @@ #include "intel_batchbuffer.h" #include "intel_blit.h" #include "intel_tex_layout.h" -#include "intel_surface.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" +#include "pipe/p_defines.h" #include "drirenderbuffer.h" #include "vblank.h" @@ -309,6 +309,23 @@ intelInitDriverFunctions(struct dd_function_table *functions) +/** + * Return list of surface formats supported by this driver. + */ +static const GLuint * +intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats) +{ + static const GLuint formats[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_S8_Z24, + }; + + *numFormats = sizeof(formats) / sizeof(formats[0]); + return formats; +} + + GLboolean intelCreateContext(const __GLcontextModes * mesaVis, __DRIcontextPrivate * driContextPriv, -- cgit v1.2.3 From 0643ea07caf72663a4620f254a5b60344810ce7f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 10:09:29 -0600 Subject: no longer need intel_surface.c --- src/mesa/drivers/dri/intel_winsys/Makefile | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index 786bf28a39..da971aca83 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -20,7 +20,6 @@ DRIVER_SOURCES = \ intel_context.c \ intel_ioctl.c \ intel_screen.c \ - intel_surface.c \ intel_batchpool.c C_SOURCES = \ -- cgit v1.2.3 From 6e3c01578e8669ce4ede85d5c7b94b784d136625 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 10:09:52 -0600 Subject: obsolete files --- src/mesa/drivers/dri/intel_winsys/intel_surface.c | 58 ----------------------- src/mesa/drivers/dri/intel_winsys/intel_surface.h | 36 -------------- 2 files changed, 94 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_surface.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_surface.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.c b/src/mesa/drivers/dri/intel_winsys/intel_surface.c deleted file mode 100644 index 683a81fb63..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.c +++ /dev/null @@ -1,58 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Surface-related functions for intel drivers. - * \author Brian Paul - */ - - -#include "intel_context.h" -#include "intel_surface.h" - -#include "pipe/p_state.h" -#include "pipe/p_defines.h" -#include "pipe/softpipe/sp_surface.h" - - - -/** - * Return list of surface formats supported by this driver. - */ -const GLuint * -intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats) -{ - static const GLuint formats[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_S8_Z24, - }; - - *numFormats = sizeof(formats) / sizeof(formats[0]); - return formats; -} - diff --git a/src/mesa/drivers/dri/intel_winsys/intel_surface.h b/src/mesa/drivers/dri/intel_winsys/intel_surface.h deleted file mode 100644 index 0987e963de..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_surface.h +++ /dev/null @@ -1,36 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef INTEL_SURFACE_H -#define INTEL_SURFACE_H - -extern const GLuint * -intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats); - - -#endif /* INTEL_SURFACE_H */ -- cgit v1.2.3 From af25ca6e69cef93e19bfb118111ffbaedf354f03 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 09:55:40 +0100 Subject: Add a printf method to winsys. Not currently used. --- src/mesa/pipe/softpipe/sp_winsys.h | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h index bc6db15d32..6f6d0f2446 100644 --- a/src/mesa/pipe/softpipe/sp_winsys.h +++ b/src/mesa/pipe/softpipe/sp_winsys.h @@ -50,6 +50,10 @@ struct pipe_buffer_handle; struct softpipe_winsys { + /* debug output + */ + void (*printf)( const char *, ... ); + /* The buffer manager is modeled after the dri_bugmgr interface, * but this is the subset that softpipe cares about. Remember that -- cgit v1.2.3 From cb2d95ba68affe665619cc0ec7b74fd0aaae7fc2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 09:58:17 +0100 Subject: Add a printf method to winsys and use for all debugging. --- src/mesa/pipe/i915simple/i915_blit.c | 4 +- src/mesa/pipe/i915simple/i915_context.h | 1 + src/mesa/pipe/i915simple/i915_debug.c | 83 +++++++-------- src/mesa/pipe/i915simple/i915_debug.h | 45 +++++++-- src/mesa/pipe/i915simple/i915_debug_fp.c | 134 ++++++++++++++----------- src/mesa/pipe/i915simple/i915_state_fragprog.c | 2 - src/mesa/pipe/i915simple/i915_winsys.h | 4 + 7 files changed, 164 insertions(+), 109 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/mesa/pipe/i915simple/i915_blit.c index c4eafa71d4..4fc216dd33 100644 --- a/src/mesa/pipe/i915simple/i915_blit.c +++ b/src/mesa/pipe/i915simple/i915_blit.c @@ -36,6 +36,7 @@ #include "i915_reg.h" #include "i915_batch.h" +#define FILE_DEBUG_FLAG DEBUG_BLIT void i915_fill_blit(struct i915_context *i915, @@ -102,7 +103,8 @@ i915_copy_blit( struct i915_context *i915, BATCH_LOCALS; - printf("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + DBG(i915, + "%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", __FUNCTION__, src_buffer, src_pitch, src_offset, src_x, src_y, dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index c9fd2e9a3f..550d9c012f 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -120,6 +120,7 @@ struct i915_context struct i915_state current; GLuint hardware_dirty; + GLuint debug; struct pipe_scissor_state cliprect; diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index 0951319d38..0ea6f03e49 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -29,8 +29,10 @@ #include "i915_reg.h" #include "i915_context.h" +#include "i915_winsys.h" #include "i915_debug.h" +#define PRINTF( stream, ... ) (stream)->winsys->printf( (stream)->winsys, __VA_ARGS__ ) static GLboolean debug( struct debug_stream *stream, const char *name, GLuint len ) @@ -39,19 +41,19 @@ static GLboolean debug( struct debug_stream *stream, const char *name, GLuint le GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); if (len == 0) { - _mesa_printf("Error - zero length packet (0x%08x)\n", stream->ptr[0]); + PRINTF( stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0] ); assert(0); return GL_FALSE; } if (stream->print_addresses) - _mesa_printf("%08x: ", stream->offset); + PRINTF(stream, "%08x: ", stream->offset); - _mesa_printf("%s (%d dwords):\n", name, len); + PRINTF(stream, "%s (%d dwords):\n", name, len); for (i = 0; i < len; i++) - _mesa_printf("\t\t0x%08x\n", ptr[i]); - _mesa_printf("\n"); + PRINTF(stream, "\t\t0x%08x\n", ptr[i]); + PRINTF(stream, "\n"); stream->offset += len * sizeof(GLuint); @@ -88,17 +90,17 @@ static GLboolean debug_prim( struct debug_stream *stream, const char *name, - _mesa_printf("%s %s (%d dwords):\n", name, prim, len); - _mesa_printf("\t\t0x%08x\n", ptr[0]); + PRINTF(stream, "%s %s (%d dwords):\n", name, prim, len); + PRINTF(stream, "\t\t0x%08x\n", ptr[0]); for (i = 1; i < len; i++) { if (dump_floats) - _mesa_printf("\t\t0x%08x // %f\n", ptr[i], *(GLfloat *)&ptr[i]); + PRINTF(stream, "\t\t0x%08x // %f\n", ptr[i], *(GLfloat *)&ptr[i]); else - _mesa_printf("\t\t0x%08x\n", ptr[i]); + PRINTF(stream, "\t\t0x%08x\n", ptr[i]); } - _mesa_printf("\n"); + PRINTF(stream, "\n"); stream->offset += len * sizeof(GLuint); @@ -113,16 +115,16 @@ static GLboolean debug_program( struct debug_stream *stream, const char *name, G GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); if (len == 0) { - _mesa_printf("Error - zero length packet (0x%08x)\n", stream->ptr[0]); + PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]); assert(0); return GL_FALSE; } if (stream->print_addresses) - _mesa_printf("%08x: ", stream->offset); + PRINTF(stream, "%08x: ", stream->offset); - _mesa_printf("%s (%d dwords):\n", name, len); - i915_disassemble_program( ptr, len ); + PRINTF(stream, "%s (%d dwords):\n", name, len); + i915_disassemble_program( stream, ptr, len ); stream->offset += len * sizeof(GLuint); return GL_TRUE; @@ -135,17 +137,17 @@ static GLboolean debug_chain( struct debug_stream *stream, const char *name, GLu GLuint old_offset = stream->offset + len * sizeof(GLuint); GLuint i; - _mesa_printf("%s (%d dwords):\n", name, len); + PRINTF(stream, "%s (%d dwords):\n", name, len); for (i = 0; i < len; i++) - _mesa_printf("\t\t0x%08x\n", ptr[i]); + PRINTF(stream, "\t\t0x%08x\n", ptr[i]); stream->offset = ptr[1] & ~0x3; if (stream->offset < old_offset) - _mesa_printf("\n... skipping backwards from 0x%x --> 0x%x ...\n\n", + PRINTF(stream, "\n... skipping backwards from 0x%x --> 0x%x ...\n\n", old_offset, stream->offset ); else - _mesa_printf("\n... skipping from 0x%x --> 0x%x ...\n\n", + PRINTF(stream, "\n... skipping from 0x%x --> 0x%x ...\n\n", old_offset, stream->offset ); @@ -165,10 +167,10 @@ static GLboolean debug_variable_length_prim( struct debug_stream *stream ) len = 1+(i+2)/2; - _mesa_printf("3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); + PRINTF(stream, "3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); for (i = 0; i < len; i++) - _mesa_printf("\t\t0x%08x\n", ptr[i]); - _mesa_printf("\n"); + PRINTF(stream, "\t\t0x%08x\n", ptr[i]); + PRINTF(stream, "\n"); stream->offset += len * sizeof(GLuint); return GL_TRUE; @@ -183,16 +185,16 @@ static GLboolean debug_load_immediate( struct debug_stream *stream, GLuint bits = (ptr[0] >> 4) & 0xff; GLuint i, j = 0; - _mesa_printf("%s (%d dwords, flags: %x):\n", name, len, bits); - _mesa_printf("\t\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords, flags: %x):\n", name, len, bits); + PRINTF(stream, "\t\t0x%08x\n", ptr[j++]); for (i = 0; i < 8; i++) { if (bits & (1<> 8) & 0x3f; GLuint i, j = 0; - _mesa_printf("%s (%d dwords):\n", name, len); - _mesa_printf("\t\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t\t0x%08x\n", ptr[j++]); for (i = 0; i < 6; i++) { if (bits & (1<winsys; while (!done && stream.offset < bytes && diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h index 3e01db6434..f0f72780d3 100644 --- a/src/mesa/pipe/i915simple/i915_debug.h +++ b/src/mesa/pipe/i915simple/i915_debug.h @@ -39,18 +39,51 @@ struct debug_stream char *ptr; /* pointer to gtt offset zero */ char *end; /* pointer to gtt offset zero */ unsigned print_addresses; + struct i915_winsys *winsys; }; +/* Internal functions + */ +void i915_disassemble_program(struct debug_stream *stream, + const unsigned *program, unsigned sz); + +void i915_print_ureg(const char *msg, unsigned ureg); + + +#define DEBUG_BATCH 0x1 +#define DEBUG_BLIT 0x2 +#define DEBUG_BUFFER 0x4 +#define DEBUG_CONSTANTS 0x8 +#define DEBUG_CONTEXT 0x10 +#define DEBUG_DRAW 0x20 +#define DEBUG_DYNAMIC 0x40 +#define DEBUG_FLUSH 0x80 +#define DEBUG_MAP 0x100 +#define DEBUG_PROGRAM 0x200 +#define DEBUG_REGIONS 0x400 +#define DEBUG_SAMPLER 0x800 +#define DEBUG_STATIC 0x1000 +#define DEBUG_SURFACE 0x2000 +#define DEBUG_WINSYS 0x4000 + +#ifdef DEBUG +#include "i915_winsys.h" +#define DBG( i915, ... ) \ + if ((i915)->debug & FILE_DEBUG_FLAG) (i915)->winsys->printf( (i915)->winsys, __VA_ARGS__ ) +#else +#define DBG( i915, ... ) \ + (void)i915 +#endif + + +void i915_dump_batchbuffer( struct i915_context *i915, + unsigned *start, + unsigned *end ); -extern void i915_disassemble_program(const unsigned *program, unsigned sz); -extern void i915_print_ureg(const char *msg, unsigned ureg); -void -i915_dump_batchbuffer( struct i915_context *i915, - unsigned *start, - unsigned *end ); +void i915_debug_init( struct i915_context *i915 ); #endif diff --git a/src/mesa/pipe/i915simple/i915_debug_fp.c b/src/mesa/pipe/i915simple/i915_debug_fp.c index f4e8ae95ad..95476d3c55 100644 --- a/src/mesa/pipe/i915simple/i915_debug_fp.c +++ b/src/mesa/pipe/i915simple/i915_debug_fp.c @@ -29,11 +29,19 @@ #include "i915_reg.h" #include "i915_debug.h" +#include "i915_winsys.h" //#include "i915_fpc.h" #include "shader/program.h" #include "shader/prog_instruction.h" #include "shader/prog_print.h" + + +#define PRINTF( stream, ... ) (stream)->winsys->printf( (stream)->winsys, __VA_ARGS__ ) + + + + static const char *opcodes[0x20] = { "NOP", "ADD", @@ -118,33 +126,33 @@ static const char *regname[0x8] = { }; static void -print_reg_type_nr(GLuint type, GLuint nr) +print_reg_type_nr(struct debug_stream *stream, GLuint type, GLuint nr) { switch (type) { case REG_TYPE_T: switch (nr) { case T_DIFFUSE: - _mesa_printf("T_DIFFUSE"); + PRINTF(stream, "T_DIFFUSE"); return; case T_SPECULAR: - _mesa_printf("T_SPECULAR"); + PRINTF(stream, "T_SPECULAR"); return; case T_FOG_W: - _mesa_printf("T_FOG_W"); + PRINTF(stream, "T_FOG_W"); return; default: - _mesa_printf("T_TEX%d", nr); + PRINTF(stream, "T_TEX%d", nr); return; } case REG_TYPE_OC: if (nr == 0) { - _mesa_printf("oC"); + PRINTF(stream, "oC"); return; } break; case REG_TYPE_OD: if (nr == 0) { - _mesa_printf("oD"); + PRINTF(stream, "oD"); return; } break; @@ -152,7 +160,7 @@ print_reg_type_nr(GLuint type, GLuint nr) break; } - _mesa_printf("%s[%d]", regname[type], nr); + PRINTF(stream, "%s[%d]", regname[type], nr); } #define REG_SWIZZLE_MASK 0x7777 @@ -165,7 +173,7 @@ print_reg_type_nr(GLuint type, GLuint nr) static void -print_reg_neg_swizzle(GLuint reg) +print_reg_neg_swizzle(struct debug_stream *stream, GLuint reg) { int i; @@ -173,33 +181,33 @@ print_reg_neg_swizzle(GLuint reg) (reg & REG_NEGATE_MASK) == 0) return; - _mesa_printf("."); + PRINTF(stream, "."); for (i = 3; i >= 0; i--) { if (reg & (1 << ((i * 4) + 3))) - _mesa_printf("-"); + PRINTF(stream, "-"); switch ((reg >> (i * 4)) & 0x7) { case 0: - _mesa_printf("x"); + PRINTF(stream, "x"); break; case 1: - _mesa_printf("y"); + PRINTF(stream, "y"); break; case 2: - _mesa_printf("z"); + PRINTF(stream, "z"); break; case 3: - _mesa_printf("w"); + PRINTF(stream, "w"); break; case 4: - _mesa_printf("0"); + PRINTF(stream, "0"); break; case 5: - _mesa_printf("1"); + PRINTF(stream, "1"); break; default: - _mesa_printf("?"); + PRINTF(stream, "?"); break; } } @@ -207,32 +215,32 @@ print_reg_neg_swizzle(GLuint reg) static void -print_src_reg(GLuint dword) +print_src_reg(struct debug_stream *stream, GLuint dword) { GLuint nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; GLuint type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(type, nr); - print_reg_neg_swizzle(dword); + print_reg_type_nr(stream, type, nr); + print_reg_neg_swizzle(stream, dword); } static void -print_dest_reg(GLuint dword) +print_dest_reg(struct debug_stream *stream, GLuint dword) { GLuint nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; GLuint type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(type, nr); + print_reg_type_nr(stream, type, nr); if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) return; - _mesa_printf("."); + PRINTF(stream, "."); if (dword & A0_DEST_CHANNEL_X) - _mesa_printf("x"); + PRINTF(stream, "x"); if (dword & A0_DEST_CHANNEL_Y) - _mesa_printf("y"); + PRINTF(stream, "y"); if (dword & A0_DEST_CHANNEL_Z) - _mesa_printf("z"); + PRINTF(stream, "z"); if (dword & A0_DEST_CHANNEL_W) - _mesa_printf("w"); + PRINTF(stream, "w"); } @@ -242,70 +250,76 @@ print_dest_reg(GLuint dword) static void -print_arith_op(GLuint opcode, const GLuint * program) +print_arith_op(struct debug_stream *stream, + GLuint opcode, const GLuint * program) { if (opcode != A0_NOP) { - print_dest_reg(program[0]); + print_dest_reg(stream, program[0]); if (program[0] & A0_DEST_SATURATE) - _mesa_printf(" = SATURATE "); + PRINTF(stream, " = SATURATE "); else - _mesa_printf(" = "); + PRINTF(stream, " = "); } - _mesa_printf("%s ", opcodes[opcode]); + PRINTF(stream, "%s ", opcodes[opcode]); - print_src_reg(GET_SRC0_REG(program[0], program[1])); + print_src_reg(stream, GET_SRC0_REG(program[0], program[1])); if (args[opcode] == 1) { - _mesa_printf("\n"); + PRINTF(stream, "\n"); return; } - _mesa_printf(", "); - print_src_reg(GET_SRC1_REG(program[1], program[2])); + PRINTF(stream, ", "); + print_src_reg(stream, GET_SRC1_REG(program[1], program[2])); if (args[opcode] == 2) { - _mesa_printf("\n"); + PRINTF(stream, "\n"); return; } - _mesa_printf(", "); - print_src_reg(GET_SRC2_REG(program[2])); - _mesa_printf("\n"); + PRINTF(stream, ", "); + print_src_reg(stream, GET_SRC2_REG(program[2])); + PRINTF(stream, "\n"); return; } static void -print_tex_op(GLuint opcode, const GLuint * program) +print_tex_op(struct debug_stream *stream, + GLuint opcode, const GLuint * program) { - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - _mesa_printf(" = "); + print_dest_reg(stream, program[0] | A0_DEST_CHANNEL_ALL); + PRINTF(stream, " = "); - _mesa_printf("%s ", opcodes[opcode]); + PRINTF(stream, "%s ", opcodes[opcode]); - _mesa_printf("S[%d],", program[0] & T0_SAMPLER_NR_MASK); + PRINTF(stream, "S[%d],", program[0] & T0_SAMPLER_NR_MASK); - print_reg_type_nr((program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & + print_reg_type_nr(stream, + (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & REG_TYPE_MASK, (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); - _mesa_printf("\n"); + PRINTF(stream, "\n"); } static void -print_dcl_op(GLuint opcode, const GLuint * program) +print_dcl_op(struct debug_stream *stream, + GLuint opcode, const GLuint * program) { - _mesa_printf("%s ", opcodes[opcode]); - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - _mesa_printf("\n"); + PRINTF(stream, "%s ", opcodes[opcode]); + print_dest_reg(stream, + program[0] | A0_DEST_CHANNEL_ALL); + PRINTF(stream, "\n"); } void -i915_disassemble_program(const GLuint * program, GLuint sz) +i915_disassemble_program(struct debug_stream *stream, + const GLuint * program, GLuint sz) { GLuint size = program[0] & 0x1ff; GLint i; - _mesa_printf("\t\tBEGIN\n"); + PRINTF(stream, "\t\tBEGIN\n"); assert(size + 2 == sz); @@ -313,19 +327,19 @@ i915_disassemble_program(const GLuint * program, GLuint sz) for (i = 1; i < sz; i += 3, program += 3) { GLuint opcode = program[0] & (0x1f << 24); - _mesa_printf("\t\t"); + PRINTF(stream, "\t\t"); if ((GLint) opcode >= A0_NOP && opcode <= A0_SLT) - print_arith_op(opcode >> 24, program); + print_arith_op(stream, opcode >> 24, program); else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) - print_tex_op(opcode >> 24, program); + print_tex_op(stream, opcode >> 24, program); else if (opcode == D0_DCL) - print_dcl_op(opcode >> 24, program); + print_dcl_op(stream, opcode >> 24, program); else - _mesa_printf("Unknown opcode 0x%x\n", opcode); + PRINTF(stream, "Unknown opcode 0x%x\n", opcode); } - _mesa_printf("\t\tEND\n\n"); + PRINTF(stream, "\t\tEND\n\n"); } diff --git a/src/mesa/pipe/i915simple/i915_state_fragprog.c b/src/mesa/pipe/i915simple/i915_state_fragprog.c index 83f43befaf..163883b07a 100644 --- a/src/mesa/pipe/i915simple/i915_state_fragprog.c +++ b/src/mesa/pipe/i915simple/i915_state_fragprog.c @@ -56,8 +56,6 @@ static unsigned passthrough[] = unsigned *i915_passthrough_program( unsigned *dwords ) { - i915_disassemble_program( passthrough, Elements(passthrough) ); - *dwords = Elements(passthrough); return passthrough; } diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 50cb231422..887308f5dd 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -50,6 +50,10 @@ struct pipe_buffer_handle; struct i915_winsys { + /* debug output + */ + void (*printf)( struct i915_winsys *sws, + const char *, ... ); /* Many of the winsys's are probably going to have a similar * buffer-manager interface, as something almost identical is -- cgit v1.2.3 From c12b71ef93ed71a78473568025d2b4d8d8fd2003 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 10:01:15 +0100 Subject: Pass pci_id to i915_create() --- src/mesa/pipe/i915simple/i915_context.c | 42 +++++++++++++++++++++++++++++---- src/mesa/pipe/i915simple/i915_context.h | 5 ++-- src/mesa/pipe/i915simple/i915_winsys.h | 3 ++- 3 files changed, 43 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index eb7f3804d3..a15d3505d7 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -33,6 +33,14 @@ #include "pipe/draw/draw_context.h" #include "pipe/p_defines.h" +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GME 0x27AE +#define PCI_CHIP_G33_G 0x29C2 +#define PCI_CHIP_Q35_G 0x29B2 +#define PCI_CHIP_Q33_G 0x29D2 /** @@ -143,14 +151,40 @@ i915_draw_vertices(struct pipe_context *pipe, -struct pipe_context *i915_create( struct i915_winsys *winsys ) +struct pipe_context *i915_create( struct i915_winsys *winsys, + unsigned pci_id ) { - struct i915_context *i915 = CALLOC_STRUCT(i915_context); + struct i915_context *i915; + unsigned is_i945 = 0; - i915->pipe.destroy = i915_destroy; + /* TODO: Push this down into the pipe driver: + */ + switch (pci_id) { + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + break; + + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + is_i945 = 1; + break; - i915->pipe.supported_formats = i915_supported_formats; + default: + winsys->printf(winsys, "%s: unknown pci id 0x%x, cannot create context\n", + __FUNCTION__, pci_id); + return NULL; + } + + i915 = CALLOC_STRUCT(i915_context); + if (i915 == NULL) + return NULL; + i915->pipe.destroy = i915_destroy; + i915->pipe.supported_formats = i915_supported_formats; i915->pipe.draw_vb = i915_draw_vb; i915->pipe.draw_vertices = i915_draw_vertices; i915->pipe.clear = i915_clear; diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 550d9c012f..e8db2b7c36 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -122,8 +122,9 @@ struct i915_context GLuint debug; - - struct pipe_scissor_state cliprect; + struct { + unsigned is_i945:1; + } flags; }; /* A flag for each state_tracker state object: diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 887308f5dd..9802148aa1 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -123,7 +123,8 @@ struct i915_winsys { #define I915_BUFFER_ACCESS_READ 0x2 -struct pipe_context *i915_create( struct i915_winsys * ); +struct pipe_context *i915_create( struct i915_winsys *, + unsigned pci_id ); #endif -- cgit v1.2.3 From 938c307e4526298d2703818d5fa848a31b076905 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 10:02:34 +0100 Subject: Add printf handlers, pass pci id and move texlayout code to driver. --- src/mesa/drivers/dri/intel_winsys/Makefile | 1 - src/mesa/drivers/dri/intel_winsys/intel_context.c | 26 +- .../dri/intel_winsys/intel_pipe_i915simple.c | 17 +- .../drivers/dri/intel_winsys/intel_pipe_softpipe.c | 14 +- .../drivers/dri/intel_winsys/intel_tex_layout.c | 476 --------------------- .../drivers/dri/intel_winsys/intel_tex_layout.h | 16 - src/mesa/main/imports.c | 4 +- src/mesa/pipe/i915simple/Makefile | 1 + src/mesa/pipe/i915simple/i915_context.c | 7 + 9 files changed, 37 insertions(+), 525 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_tex_layout.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index da971aca83..ae08afccb3 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -14,7 +14,6 @@ DRIVER_SOURCES = \ intel_pipe_i915simple.c \ intel_pipe_softpipe.c \ intel_batchbuffer.c \ - intel_tex_layout.c \ intel_buffers.c \ intel_blit.c \ intel_context.c \ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 8c61f0cf3c..b88c785fb6 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -243,6 +243,8 @@ intelFlush(GLcontext * ctx) { struct intel_context *intel = intel_context(ctx); + /* Hmm: + */ intel->pipe->flush( intel->pipe, 0 ); } @@ -421,8 +423,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, */ if (!getenv("INTEL_HW")) { intel->pipe = intel_create_softpipe( intel ); - /* use default softpipe function for surface_alloc() */ - intel->pipe->supported_formats = intel_supported_formats; } else { switch (intel->intelScreen->deviceID) { @@ -447,28 +447,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, st_create_context( &intel->ctx, intel->pipe ); - - /* TODO: Push this down into the pipe driver: - */ - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - intel->pipe->mipmap_tree_layout = i945_miptree_layout; - break; - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - case PCI_CHIP_I830_M: - case PCI_CHIP_I855_GM: - case PCI_CHIP_I865_G: - intel->pipe->mipmap_tree_layout = i915_miptree_layout; - default: - assert(0); /*FIX*/ - } - return GL_TRUE; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c index 98d446ba90..c0e8c2349c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c @@ -227,10 +227,18 @@ static void intel_i915_batch_flush( struct i915_winsys *sws ) { struct intel_context *intel = intel_i915_winsys(sws)->intel; - _mesa_printf("%s: start\n"); intel_batchbuffer_flush( intel->batch ); - intel_i915_batch_wait_idle( sws ); - _mesa_printf("%s: done\n"); + if (0) intel_i915_batch_wait_idle( sws ); +} + + +static void intel_i915_printf( struct i915_winsys *sws, + const char *fmtString, ... ) +{ + va_list args; + va_start( args, fmtString ); + vfprintf(stderr, fmtString, args); + va_end( args ); } @@ -242,6 +250,7 @@ intel_create_i915simple( struct intel_context *intel ) /* Fill in this struct with callbacks that i915simple will need to * communicate with the window system, buffer manager, etc. */ + iws->winsys.printf = intel_i915_printf; iws->winsys.buffer_create = intel_i915_buffer_create; iws->winsys.buffer_map = intel_i915_buffer_map; iws->winsys.buffer_unmap = intel_i915_buffer_unmap; @@ -259,5 +268,5 @@ intel_create_i915simple( struct intel_context *intel ) /* Create the i915simple context: */ - return i915_create( &iws->winsys ); + return i915_create( &iws->winsys, intel->intelScreen->deviceID ); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c index 5edbca1098..fdd7acf747 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c @@ -36,9 +36,14 @@ #include "intel_context.h" #include "intel_pipe.h" +#include "intel_surface.h" #include "pipe/softpipe/sp_winsys.h" +/* Shouldn't really need this: + */ +#include "pipe/p_context.h" + struct intel_softpipe_winsys { struct softpipe_winsys sws; @@ -157,6 +162,7 @@ struct pipe_context * intel_create_softpipe( struct intel_context *intel ) { struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); + struct pipe_context *pipe; /* Fill in this struct with callbacks that softpipe will need to * communicate with the window system, buffer manager, etc. @@ -177,5 +183,11 @@ intel_create_softpipe( struct intel_context *intel ) /* Create the softpipe context: */ - return softpipe_create( &isws->sws ); + pipe = softpipe_create( &isws->sws ); + + /* XXX: This should probably be a parameter to softpipe_create() + */ + pipe->supported_formats = intel_supported_formats; + + return pipe; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c b/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c deleted file mode 100644 index 882d9e04a2..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.c +++ /dev/null @@ -1,476 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Michel Dänzer - */ - -#include "macros.h" -#include "pipe/p_state.h" -#include "pipe/p_context.h" -#include "intel_tex_layout.h" -#include "state_tracker/st_mipmap_tree.h" - - -static GLuint minify( GLuint d ) -{ - return MAX2(1, d>>1); -} - -static int align(int value, int alignment) -{ - return (value + alignment - 1) & ~(alignment - 1); -} - - -static void -intel_miptree_set_level_info(struct pipe_mipmap_tree *mt, - GLuint level, - GLuint nr_images, - GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) -{ - assert(level < MAX_TEXTURE_LEVELS); - - mt->level[level].width = w; - mt->level[level].height = h; - mt->level[level].depth = d; - mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; - mt->level[level].nr_images = nr_images; - - /* - DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - level, w, h, d, x, y, mt->level[level].level_offset); - */ - - /* Not sure when this would happen, but anyway: - */ - if (mt->level[level].image_offset) { - free(mt->level[level].image_offset); - mt->level[level].image_offset = NULL; - } - - assert(nr_images); - assert(!mt->level[level].image_offset); - - mt->level[level].image_offset = (GLuint *) malloc(nr_images * sizeof(GLuint)); - mt->level[level].image_offset[0] = 0; -} - - -static void -intel_miptree_set_image_offset(struct pipe_mipmap_tree *mt, - GLuint level, GLuint img, GLuint x, GLuint y) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < mt->level[level].nr_images); - - mt->level[level].image_offset[img] = (x + y * mt->pitch); - - /* - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); - */ -} - - -static void -i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) -{ - GLint align_h = 2, align_w = 4; - GLuint level; - GLuint x = 0; - GLuint y = 0; - GLuint width = mt->width0; - GLuint height = mt->height0; - - mt->pitch = mt->width0; - - /* May need to adjust pitch to accomodate the placement of - * the 2nd mipmap. This occurs when the alignment - * constraints of mipmap placement push the right edge of the - * 2nd mipmap out past the width of its parent. - */ - if (mt->first_level != mt->last_level) { - GLuint mip1_width = align(minify(mt->width0), align_w) - + minify(minify(mt->width0)); - - if (mip1_width > mt->width0) - mt->pitch = mip1_width; - } - - /* Pitch must be a whole number of dwords, even though we - * express it in texels. - */ - mt->pitch = align(mt->pitch * mt->cpp, 4) / mt->cpp; - mt->total_height = 0; - - for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { - GLuint img_height; - - intel_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); - - if (mt->compressed) - img_height = MAX2(1, height/4); - else - img_height = align(height, align_h); - - - /* Because the images are packed better, the final offset - * might not be the maximal one: - */ - mt->total_height = MAX2(mt->total_height, y + img_height); - - /* Layout_below: step right after second mipmap. - */ - if (level == mt->first_level + 1) { - x += align(width, align_w); - } - else { - y += img_height; - } - - width = minify(width); - height = minify(height); - } -} - - -static const GLint initial_offsets[6][2] = { - {0, 0}, - {0, 2}, - {1, 0}, - {1, 2}, - {1, 1}, - {1, 3} -}; - -static const GLint step_offsets[6][2] = { - {0, 2}, - {0, 2}, - {-1, 2}, - {-1, 2}, - {-1, 1}, - {-1, 1} -}; - - -GLboolean -i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) -{ - GLint level; - - switch (mt->target) { - case GL_TEXTURE_CUBE_MAP:{ - const GLuint dim = mt->width0; - GLuint face; - GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* double pitch for cube layouts */ - mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; - mt->total_height = dim * 4; - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_level_info(mt, level, 6, - 0, 0, - /*OLD: mt->pitch, mt->total_height,*/ - lvlWidth, lvlHeight, - 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - for (face = 0; face < 6; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_image_offset(mt, level, face, x, y); - - if (d == 0) - _mesa_printf("cube mipmap %d/%d (%d..%d) is 0x0\n", - face, level, mt->first_level, mt->last_level); - - d >>= 1; - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - } - } - break; - } - case GL_TEXTURE_3D:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint stack_height = 0; - - /* Calculate the size of a single slice. - */ - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - - /* XXX: hardware expects/requires 9 levels at minimum. - */ - for (level = mt->first_level; level <= MAX2(8, mt->last_level); - level++) { - intel_miptree_set_level_info(mt, level, depth, 0, mt->total_height, - width, height, depth); - - - stack_height += MAX2(2, height); - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - - /* Fixup depth image_offsets: - */ - depth = mt->depth0; - for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint i; - for (i = 0; i < depth; i++) - intel_miptree_set_image_offset(mt, level, i, - 0, i * stack_height); - - depth = minify(depth); - } - - - /* Multiply slice size by texture depth for total size. It's - * remarkable how wasteful of memory the i915 texture layouts - * are. They are largely fixed in the i945. - */ - mt->total_height = stack_height * mt->depth0; - break; - } - - default:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint img_height; - - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - mt->total_height = 0; - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_level_info(mt, level, 1, - 0, mt->total_height, - width, height, 1); - - if (mt->compressed) - img_height = MAX2(1, height / 4); - else - img_height = (MAX2(2, height) + 1) & ~1; - - mt->total_height += img_height; - - width = minify(width); - height = minify(height); - } - break; - } - } - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); - */ - - return GL_TRUE; -} - - -GLboolean -i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) -{ - GLint level; - - switch (mt->target) { - case GL_TEXTURE_CUBE_MAP:{ - const GLuint dim = mt->width0; - GLuint face; - GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* Depending on the size of the largest images, pitch can be - * determined either by the old-style packing of cubemap faces, - * or the final row of 4x4, 2x2 and 1x1 faces below this. - */ - if (dim > 32) - mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; - else - mt->pitch = 14 * 8; - - mt->total_height = dim * 4 + 4; - - /* Set all the levels to effectively occupy the whole rectangular region. - */ - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_level_info(mt, level, 6, - 0, 0, - lvlWidth, lvlHeight, 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - - for (face = 0; face < 6; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; - - if (dim == 4 && face >= 4) { - y = mt->total_height - 4; - x = (face - 4) * 8; - } - else if (dim < 4 && (face > 0 || mt->first_level > 0)) { - y = mt->total_height - 4; - x = face * 8; - } - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_image_offset(mt, level, face, x, y); - - d >>= 1; - - switch (d) { - case 4: - switch (face) { - case FACE_POS_X: - case FACE_NEG_X: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - case FACE_POS_Y: - case FACE_NEG_Y: - y += 12; - x -= 8; - break; - case FACE_POS_Z: - case FACE_NEG_Z: - y = mt->total_height - 4; - x = (face - 4) * 8; - break; - } - - case 2: - y = mt->total_height - 4; - x = 16 + face * 8; - break; - - case 1: - x += 48; - break; - - default: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - } - } - } - break; - } - case GL_TEXTURE_3D:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint pack_x_pitch, pack_x_nr; - GLuint pack_y_pitch; - GLuint level; - - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - mt->total_height = 0; - - pack_y_pitch = MAX2(mt->height0, 2); - pack_x_pitch = mt->pitch; - pack_x_nr = 1; - - for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; - GLint x = 0; - GLint y = 0; - GLint q, j; - - intel_miptree_set_level_info(mt, level, nr_images, - 0, mt->total_height, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - intel_miptree_set_image_offset(mt, level, q, x, y); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - mt->total_height += y; - - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= mt->pitch); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - } - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - break; - } - - case GL_TEXTURE_1D: - case GL_TEXTURE_2D: - case GL_TEXTURE_RECTANGLE_ARB: - i945_miptree_layout_2d(mt); - break; - default: - _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()"); - } - - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); - */ - - return GL_TRUE; -} - diff --git a/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.h b/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.h deleted file mode 100644 index 83913b89cb..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_tex_layout.h +++ /dev/null @@ -1,16 +0,0 @@ - -#ifndef INTEL_TEX_LAYOUT_H -#define INTEL_TEX_LAYOUT_H - -struct pipe_context; -struct pipe_mipmap_tree; - - -extern GLboolean -i915_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); - -extern GLboolean -i945_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); - - -#endif /* INTEL_TEX_LAYOUT_H */ diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c index 3ae56c8b0b..d8d35af15e 100644 --- a/src/mesa/main/imports.c +++ b/src/mesa/main/imports.c @@ -902,12 +902,10 @@ _mesa_sprintf( char *str, const char *fmt, ... ) void _mesa_printf( const char *fmtString, ... ) { - char s[MAXSTRING]; va_list args; va_start( args, fmtString ); - vsnprintf(s, MAXSTRING, fmtString, args); + vfprintf(stderr, fmtString, args); va_end( args ); - fprintf(stderr,"%s", s); } /** Wrapper around vsprintf() */ diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 96e16a948f..5b919e3bed 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -21,6 +21,7 @@ DRIVER_SOURCES = \ i915_state_emit.c \ i915_state_fragprog.c \ i915_prim_emit.c \ + i915_tex_layout.c \ i915_surface.c C_SOURCES = \ diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index a15d3505d7..d8e54f02ee 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -29,6 +29,7 @@ #include "i915_context.h" #include "i915_winsys.h" #include "i915_state.h" +#include "i915_tex_layout.h" #include "pipe/draw/draw_context.h" #include "pipe/p_defines.h" @@ -207,6 +208,12 @@ struct pipe_context *i915_create( struct i915_winsys *winsys, i915_init_flush_functions(i915); + if (i915->flags.is_i945) + i915->pipe.mipmap_tree_layout = i945_miptree_layout; + else + i915->pipe.mipmap_tree_layout = i945_miptree_layout; + + i915->dirty = ~0; i915->hardware_dirty = ~0; -- cgit v1.2.3 From fa6815f0566339eb0879842369d3aea25dc6233f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 10:15:23 +0100 Subject: Give softpipe a native texture layout of its own. Basically the same as the 945's for now. --- src/mesa/pipe/softpipe/sp_tex_layout.c | 347 ++++++++++++++++++++++++++++++++- 1 file changed, 344 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c index 2bbdca0f58..37969c84de 100644 --- a/src/mesa/pipe/softpipe/sp_tex_layout.c +++ b/src/mesa/pipe/softpipe/sp_tex_layout.c @@ -1,13 +1,354 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ +#include "macros.h" #include "pipe/p_state.h" +#include "pipe/p_context.h" #include "sp_tex_layout.h" +/* At the moment, just make softpipe use the same layout for its + * textures as the i945. Softpipe needs some sort of texture layout, + * this one was handy. May be worthwhile to simplify this code a + * little. + */ + +static GLuint minify( GLuint d ) +{ + return MAX2(1, d>>1); +} + +static int align(int value, int alignment) +{ + return (value + alignment - 1) & ~(alignment - 1); +} + + +static void +sp_miptree_set_level_info(struct pipe_mipmap_tree *mt, + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) +{ + assert(level < MAX_TEXTURE_LEVELS); + + mt->level[level].width = w; + mt->level[level].height = h; + mt->level[level].depth = d; + mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; + mt->level[level].nr_images = nr_images; + + /* + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, mt->level[level].level_offset); + */ + + /* Not sure when this would happen, but anyway: + */ + if (mt->level[level].image_offset) { + free(mt->level[level].image_offset); + mt->level[level].image_offset = NULL; + } + + assert(nr_images); + assert(!mt->level[level].image_offset); + + mt->level[level].image_offset = (GLuint *) malloc(nr_images * sizeof(GLuint)); + mt->level[level].image_offset[0] = 0; +} + + +static void +sp_miptree_set_image_offset(struct pipe_mipmap_tree *mt, + GLuint level, GLuint img, GLuint x, GLuint y) +{ + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + + assert(img < mt->level[level].nr_images); + + mt->level[level].image_offset[img] = (x + y * mt->pitch); + + /* + DBG("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); + */ +} + + +static void +sp_miptree_layout_2d( struct pipe_mipmap_tree *mt ) +{ + GLint align_h = 2, align_w = 4; + GLuint level; + GLuint x = 0; + GLuint y = 0; + GLuint width = mt->width0; + GLuint height = mt->height0; + + mt->pitch = mt->width0; + + /* May need to adjust pitch to accomodate the placement of + * the 2nd mipmap. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap out past the width of its parent. + */ + if (mt->first_level != mt->last_level) { + GLuint mip1_width = align(minify(mt->width0), align_w) + + minify(minify(mt->width0)); + + if (mip1_width > mt->width0) + mt->pitch = mip1_width; + } + + /* Pitch must be a whole number of dwords, even though we + * express it in texels. + */ + mt->pitch = align(mt->pitch * mt->cpp, 4) / mt->cpp; + mt->total_height = 0; + + for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { + GLuint img_height; + + sp_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); + + if (mt->compressed) + img_height = MAX2(1, height/4); + else + img_height = align(height, align_h); + + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + mt->total_height = MAX2(mt->total_height, y + img_height); + + /* Layout_below: step right after second mipmap. + */ + if (level == mt->first_level + 1) { + x += align(width, align_w); + } + else { + y += img_height; + } + + width = minify(width); + height = minify(height); + } +} + + +static const GLint initial_offsets[6][2] = { + {0, 0}, + {0, 2}, + {1, 0}, + {1, 2}, + {1, 1}, + {1, 3} +}; + +static const GLint step_offsets[6][2] = { + {0, 2}, + {0, 2}, + {-1, 2}, + {-1, 2}, + {-1, 1}, + {-1, 1} +}; + + + GLboolean -softpipe_mipmap_tree_layout(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt) +softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) { + GLint level; + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP:{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* Depending on the size of the largest images, pitch can be + * determined either by the old-style packing of cubemap faces, + * or the final row of 4x4, 2x2 and 1x1 faces below this. + */ + if (dim > 32) + mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; + else + mt->pitch = 14 * 8; + + mt->total_height = dim * 4 + 4; + + /* Set all the levels to effectively occupy the whole rectangular region. + */ + for (level = mt->first_level; level <= mt->last_level; level++) { + sp_miptree_set_level_info(mt, level, 6, + 0, 0, + lvlWidth, lvlHeight, 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + + for (face = 0; face < 6; face++) { + GLuint x = initial_offsets[face][0] * dim; + GLuint y = initial_offsets[face][1] * dim; + GLuint d = dim; + + if (dim == 4 && face >= 4) { + y = mt->total_height - 4; + x = (face - 4) * 8; + } + else if (dim < 4 && (face > 0 || mt->first_level > 0)) { + y = mt->total_height - 4; + x = face * 8; + } + + for (level = mt->first_level; level <= mt->last_level; level++) { + sp_miptree_set_image_offset(mt, level, face, x, y); + + d >>= 1; + + switch (d) { + case 4: + switch (face) { + case FACE_POS_X: + case FACE_NEG_X: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + case FACE_POS_Y: + case FACE_NEG_Y: + y += 12; + x -= 8; + break; + case FACE_POS_Z: + case FACE_NEG_Z: + y = mt->total_height - 4; + x = (face - 4) * 8; + break; + } + + case 2: + y = mt->total_height - 4; + x = 16 + face * 8; + break; + + case 1: + x += 48; + break; + + default: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + } + } + } + break; + } + case GL_TEXTURE_3D:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint pack_x_pitch, pack_x_nr; + GLuint pack_y_pitch; + GLuint level; + + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->total_height = 0; + + pack_y_pitch = MAX2(mt->height0, 2); + pack_x_pitch = mt->pitch; + pack_x_nr = 1; + + for (level = mt->first_level; level <= mt->last_level; level++) { + GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; + GLint x = 0; + GLint y = 0; + GLint q, j; + + sp_miptree_set_level_info(mt, level, nr_images, + 0, mt->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + sp_miptree_set_image_offset(mt, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + mt->total_height += y; + + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= mt->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + } + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + break; + } + + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_RECTANGLE_ARB: + sp_miptree_layout_2d(mt); + break; + default: + _mesa_problem(NULL, "Unexpected tex target in sp_miptree_layout()"); + } + + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + mt->pitch, + mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); + */ + return GL_TRUE; } - -- cgit v1.2.3 From 07fd9ad7bbd6de8d2beaa5202daecb7dac861899 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 10:20:35 +0100 Subject: Fix merge issue --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 15 --------------- .../drivers/dri/intel_winsys/intel_pipe_softpipe.c | 20 +++++++++++++++++++- 2 files changed, 19 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index b88c785fb6..fea6995b07 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -311,21 +311,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) -/** - * Return list of surface formats supported by this driver. - */ -static const GLuint * -intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats) -{ - static const GLuint formats[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_S8_Z24, - }; - - *numFormats = sizeof(formats) / sizeof(formats[0]); - return formats; -} GLboolean diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c index fdd7acf747..06db5c9f43 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c @@ -36,9 +36,9 @@ #include "intel_context.h" #include "intel_pipe.h" -#include "intel_surface.h" #include "pipe/softpipe/sp_winsys.h" +#include "pipe/p_defines.h" /* Shouldn't really need this: */ @@ -158,6 +158,24 @@ intel_create_buffer(struct softpipe_winsys *sws, } +/** + * Return list of surface formats supported by this driver. + */ +static const GLuint * +intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats) +{ + static const GLuint formats[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_S8_Z24, + }; + + *numFormats = sizeof(formats) / sizeof(formats[0]); + return formats; +} + + + struct pipe_context * intel_create_softpipe( struct intel_context *intel ) { -- cgit v1.2.3 From c26d635e225c150af400be085754f4230524ea44 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 10:21:42 +0100 Subject: Remove dead include line --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index fea6995b07..3782148001 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -45,7 +45,6 @@ #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" -#include "intel_tex_layout.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" -- cgit v1.2.3 From a27c0db260fb289ec7994a8c66b002443bc8a96b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 10:23:24 +0100 Subject: Missing files --- src/mesa/pipe/i915simple/i915_tex_layout.c | 475 +++++++++++++++++++++++++++++ src/mesa/pipe/i915simple/i915_tex_layout.h | 16 + 2 files changed, 491 insertions(+) create mode 100644 src/mesa/pipe/i915simple/i915_tex_layout.c create mode 100644 src/mesa/pipe/i915simple/i915_tex_layout.h (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_tex_layout.c b/src/mesa/pipe/i915simple/i915_tex_layout.c new file mode 100644 index 0000000000..d0c58de691 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_tex_layout.c @@ -0,0 +1,475 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "macros.h" +#include "pipe/p_state.h" +#include "pipe/p_context.h" +#include "i915_tex_layout.h" + + +static GLuint minify( GLuint d ) +{ + return MAX2(1, d>>1); +} + +static int align(int value, int alignment) +{ + return (value + alignment - 1) & ~(alignment - 1); +} + + +static void +i915_miptree_set_level_info(struct pipe_mipmap_tree *mt, + GLuint level, + GLuint nr_images, + GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) +{ + assert(level < MAX_TEXTURE_LEVELS); + + mt->level[level].width = w; + mt->level[level].height = h; + mt->level[level].depth = d; + mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; + mt->level[level].nr_images = nr_images; + + /* + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, mt->level[level].level_offset); + */ + + /* Not sure when this would happen, but anyway: + */ + if (mt->level[level].image_offset) { + free(mt->level[level].image_offset); + mt->level[level].image_offset = NULL; + } + + assert(nr_images); + assert(!mt->level[level].image_offset); + + mt->level[level].image_offset = (GLuint *) malloc(nr_images * sizeof(GLuint)); + mt->level[level].image_offset[0] = 0; +} + + +static void +i915_miptree_set_image_offset(struct pipe_mipmap_tree *mt, + GLuint level, GLuint img, GLuint x, GLuint y) +{ + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + + assert(img < mt->level[level].nr_images); + + mt->level[level].image_offset[img] = (x + y * mt->pitch); + + /* + DBG("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); + */ +} + + +static void +i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) +{ + GLint align_h = 2, align_w = 4; + GLuint level; + GLuint x = 0; + GLuint y = 0; + GLuint width = mt->width0; + GLuint height = mt->height0; + + mt->pitch = mt->width0; + + /* May need to adjust pitch to accomodate the placement of + * the 2nd mipmap. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap out past the width of its parent. + */ + if (mt->first_level != mt->last_level) { + GLuint mip1_width = align(minify(mt->width0), align_w) + + minify(minify(mt->width0)); + + if (mip1_width > mt->width0) + mt->pitch = mip1_width; + } + + /* Pitch must be a whole number of dwords, even though we + * express it in texels. + */ + mt->pitch = align(mt->pitch * mt->cpp, 4) / mt->cpp; + mt->total_height = 0; + + for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { + GLuint img_height; + + i915_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); + + if (mt->compressed) + img_height = MAX2(1, height/4); + else + img_height = align(height, align_h); + + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + mt->total_height = MAX2(mt->total_height, y + img_height); + + /* Layout_below: step right after second mipmap. + */ + if (level == mt->first_level + 1) { + x += align(width, align_w); + } + else { + y += img_height; + } + + width = minify(width); + height = minify(height); + } +} + + +static const GLint initial_offsets[6][2] = { + {0, 0}, + {0, 2}, + {1, 0}, + {1, 2}, + {1, 1}, + {1, 3} +}; + +static const GLint step_offsets[6][2] = { + {0, 2}, + {0, 2}, + {-1, 2}, + {-1, 2}, + {-1, 1}, + {-1, 1} +}; + + +GLboolean +i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) +{ + GLint level; + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP:{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* double pitch for cube layouts */ + mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; + mt->total_height = dim * 4; + + for (level = mt->first_level; level <= mt->last_level; level++) { + i915_miptree_set_level_info(mt, level, 6, + 0, 0, + /*OLD: mt->pitch, mt->total_height,*/ + lvlWidth, lvlHeight, + 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + for (face = 0; face < 6; face++) { + GLuint x = initial_offsets[face][0] * dim; + GLuint y = initial_offsets[face][1] * dim; + GLuint d = dim; + + for (level = mt->first_level; level <= mt->last_level; level++) { + i915_miptree_set_image_offset(mt, level, face, x, y); + + if (d == 0) + _mesa_printf("cube mipmap %d/%d (%d..%d) is 0x0\n", + face, level, mt->first_level, mt->last_level); + + d >>= 1; + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + } + } + break; + } + case GL_TEXTURE_3D:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint stack_height = 0; + + /* Calculate the size of a single slice. + */ + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + + /* XXX: hardware expects/requires 9 levels at minimum. + */ + for (level = mt->first_level; level <= MAX2(8, mt->last_level); + level++) { + i915_miptree_set_level_info(mt, level, depth, 0, mt->total_height, + width, height, depth); + + + stack_height += MAX2(2, height); + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + + /* Fixup depth image_offsets: + */ + depth = mt->depth0; + for (level = mt->first_level; level <= mt->last_level; level++) { + GLuint i; + for (i = 0; i < depth; i++) + i915_miptree_set_image_offset(mt, level, i, + 0, i * stack_height); + + depth = minify(depth); + } + + + /* Multiply slice size by texture depth for total size. It's + * remarkable how wasteful of memory the i915 texture layouts + * are. They are largely fixed in the i945. + */ + mt->total_height = stack_height * mt->depth0; + break; + } + + default:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint img_height; + + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->total_height = 0; + + for (level = mt->first_level; level <= mt->last_level; level++) { + i915_miptree_set_level_info(mt, level, 1, + 0, mt->total_height, + width, height, 1); + + if (mt->compressed) + img_height = MAX2(1, height / 4); + else + img_height = (MAX2(2, height) + 1) & ~1; + + mt->total_height += img_height; + + width = minify(width); + height = minify(height); + } + break; + } + } + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + mt->pitch, + mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); + */ + + return GL_TRUE; +} + + +GLboolean +i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) +{ + GLint level; + + switch (mt->target) { + case GL_TEXTURE_CUBE_MAP:{ + const GLuint dim = mt->width0; + GLuint face; + GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* Depending on the size of the largest images, pitch can be + * determined either by the old-style packing of cubemap faces, + * or the final row of 4x4, 2x2 and 1x1 faces below this. + */ + if (dim > 32) + mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; + else + mt->pitch = 14 * 8; + + mt->total_height = dim * 4 + 4; + + /* Set all the levels to effectively occupy the whole rectangular region. + */ + for (level = mt->first_level; level <= mt->last_level; level++) { + i915_miptree_set_level_info(mt, level, 6, + 0, 0, + lvlWidth, lvlHeight, 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + + for (face = 0; face < 6; face++) { + GLuint x = initial_offsets[face][0] * dim; + GLuint y = initial_offsets[face][1] * dim; + GLuint d = dim; + + if (dim == 4 && face >= 4) { + y = mt->total_height - 4; + x = (face - 4) * 8; + } + else if (dim < 4 && (face > 0 || mt->first_level > 0)) { + y = mt->total_height - 4; + x = face * 8; + } + + for (level = mt->first_level; level <= mt->last_level; level++) { + i915_miptree_set_image_offset(mt, level, face, x, y); + + d >>= 1; + + switch (d) { + case 4: + switch (face) { + case FACE_POS_X: + case FACE_NEG_X: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + case FACE_POS_Y: + case FACE_NEG_Y: + y += 12; + x -= 8; + break; + case FACE_POS_Z: + case FACE_NEG_Z: + y = mt->total_height - 4; + x = (face - 4) * 8; + break; + } + + case 2: + y = mt->total_height - 4; + x = 16 + face * 8; + break; + + case 1: + x += 48; + break; + + default: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + } + } + } + break; + } + case GL_TEXTURE_3D:{ + GLuint width = mt->width0; + GLuint height = mt->height0; + GLuint depth = mt->depth0; + GLuint pack_x_pitch, pack_x_nr; + GLuint pack_y_pitch; + GLuint level; + + mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; + mt->total_height = 0; + + pack_y_pitch = MAX2(mt->height0, 2); + pack_x_pitch = mt->pitch; + pack_x_nr = 1; + + for (level = mt->first_level; level <= mt->last_level; level++) { + GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; + GLint x = 0; + GLint y = 0; + GLint q, j; + + i915_miptree_set_level_info(mt, level, nr_images, + 0, mt->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + i915_miptree_set_image_offset(mt, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + mt->total_height += y; + + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= mt->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + } + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + break; + } + + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_RECTANGLE_ARB: + i945_miptree_layout_2d(mt); + break; + default: + _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()"); + } + + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + mt->pitch, + mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); + */ + + return GL_TRUE; +} + diff --git a/src/mesa/pipe/i915simple/i915_tex_layout.h b/src/mesa/pipe/i915simple/i915_tex_layout.h new file mode 100644 index 0000000000..b1555f1dd8 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_tex_layout.h @@ -0,0 +1,16 @@ + +#ifndef I915_TEX_LAYOUT_H +#define I915_TEX_LAYOUT_H + +struct pipe_context; +struct pipe_mipmap_tree; + + +extern GLboolean +i915_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); + +extern GLboolean +i945_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); + + +#endif /* I915_TEX_LAYOUT_H */ -- cgit v1.2.3 From 30e506eb04ec9d82dbb879f27547475a875d50d7 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 10:37:20 +0100 Subject: Clean up / sanitize debugging --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 9 +++--- src/mesa/drivers/dri/intel_winsys/intel_blit.c | 12 ++++---- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 13 ++++----- src/mesa/drivers/dri/intel_winsys/intel_context.c | 29 ++++++-------------- src/mesa/drivers/dri/intel_winsys/intel_context.h | 32 ++++++++++------------ src/mesa/drivers/dri/intel_winsys/intel_ioctl.c | 9 +++--- 6 files changed, 40 insertions(+), 64 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index d1fb2907a9..30511bcba1 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -71,11 +71,10 @@ static void intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) { int i; - fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); + _mesa_printf("\n\n\nSTART BATCH (%d dwords):\n", count / 4); for (i = 0; i < count / 4; i += 1) - fprintf(stderr, "\t0x%08x\n", - ptr[i]); - fprintf(stderr, "END BATCH\n\n\n"); + _mesa_printf("\t0x%08x\n", ptr[i]); + _mesa_printf("END BATCH\n\n\n"); } void @@ -193,7 +192,7 @@ do_flush_locked(struct intel_batchbuffer *batch, ptr[r->offset / 4] = driBOOffset(r->buf) + r->delta; } - if (INTEL_DEBUG & DEBUG_BATCH) + if (0) intel_dump_batchbuffer(0, ptr, used); driBOUnmap(batch->buffer); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_blit.c b/src/mesa/drivers/dri/intel_winsys/intel_blit.c index f099eb2483..1578340e17 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_blit.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_blit.c @@ -41,8 +41,6 @@ #include "state_tracker/st_cb_fbo.h" -#define FILE_DEBUG_FLAG DEBUG_BLIT - /** * Copy the back color buffer to the front color buffer. * Used for SwapBuffers(). @@ -55,7 +53,7 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, struct intel_context *intel; const intelScreenPrivate *intelScreen; - DBG("%s\n", __FUNCTION__); + DBG(SWAP, "%s\n", __FUNCTION__); assert(dPriv); @@ -127,8 +125,8 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, ASSERT(backRegion); ASSERT(backRegion->cpp == cpp); - DBG("front pitch %d back pitch %d\n", - pitch, backRegion->pitch); + DBG(SWAP, "front pitch %d back pitch %d\n", + pitch, backRegion->pitch); if (cpp == 2) { BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); @@ -178,8 +176,8 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, if (box.y2 - box.y1 > backHeight) box.y2 = backHeight + box.y1; - DBG("box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); + DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); sbox.x1 = box.x1 - dPriv->x; sbox.y1 = box.y1 - dPriv->y; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index eb42be98e5..1bd2a36738 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -145,10 +145,9 @@ intelWindowMoved(struct intel_context *intel) pf_active = pf_pipes && (pf_pipes & intel->sarea->pf_active) == pf_pipes; - if (INTEL_DEBUG & DEBUG_LOCK) - if (pf_active != intel_fb->pf_active) - _mesa_printf("%s - Page flipping %sactive\n", __progname, - pf_active ? "" : "in"); + if (pf_active != intel_fb->pf_active) + DBG(LOCK, "%s - Page flipping %sactive\n", + __progname, pf_active ? "" : "in"); if (pf_active) { /* Sync pages between pipes if we're flipping on both at the same time */ @@ -283,8 +282,7 @@ intelPageFlip(const __DRIdrawablePrivate * dPriv) int ret; struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); + DBG(SWAP, "%s\n", __FUNCTION__); assert(dPriv); assert(dPriv->driContextPriv); @@ -332,8 +330,7 @@ intelPageFlip(const __DRIdrawablePrivate * dPriv) intel_draw_buffer(&intel->ctx, &intel_fb->Base); #endif - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s: success\n", __FUNCTION__); + DBG(SWAP, "%s: success\n", __FUNCTION__); return GL_TRUE; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 3782148001..1032fc2d0f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -59,8 +59,8 @@ -#ifndef INTEL_DEBUG -int INTEL_DEBUG = (0); +#if DEBUG +int __intel_debug = 0; #endif #define need_GL_ARB_multisample @@ -210,18 +210,10 @@ const struct dri_extension card_extensions[] = { static const struct dri_debug_control debug_control[] = { - {"tex", DEBUG_TEXTURE}, - {"state", DEBUG_STATE}, {"ioctl", DEBUG_IOCTL}, - {"blit", DEBUG_BLIT}, - {"mip", DEBUG_MIPTREE}, - {"verb", DEBUG_VERBOSE}, {"bat", DEBUG_BATCH}, - {"pix", DEBUG_PIXEL}, - {"buf", DEBUG_BUFMGR}, - {"reg", DEBUG_REGION}, - {"fbo", DEBUG_FBO}, {"lock", DEBUG_LOCK}, + {"swap", DEBUG_SWAP}, {NULL, 0} }; @@ -397,8 +389,8 @@ intelCreateContext(const __GLcontextModes * mesaVis, _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); } -#if DO_DEBUG - INTEL_DEBUG = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); +#if DEBUG + __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); #endif @@ -470,8 +462,6 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) /* This share group is about to go away, free our private * texture object data. */ - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "do something to free texture heaps\n"); } /* free the Mesa context */ @@ -570,8 +560,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) drmGetLock(intel->driFd, intel->hHWContext, flags); - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("%s - got contended lock\n", __progname); + DBG(LOCK, "%s - got contended lock\n", __progname); /* If the window moved, may need to set a new cliprect now. * @@ -660,8 +649,7 @@ void LOCK_HARDWARE( struct intel_context *intel ) if (__ret) intelContendedLock( intel, 0 ); - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("%s - locked\n", __progname); + DBG(LOCK, "%s - locked\n", __progname); intel->locked = 1; } @@ -678,7 +666,6 @@ void UNLOCK_HARDWARE( struct intel_context *intel ) _glthread_UNLOCK_MUTEX(lockMutex); - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("%s - unlocked\n", __progname); + DBG(LOCK, "%s - unlocked\n", __progname); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index e84f53d1dc..924b7ca05a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -109,27 +109,23 @@ extern char *__progname; /* ================================================================ * Debugging: */ -#define DO_DEBUG 0 -#if DO_DEBUG -extern int INTEL_DEBUG; +#if DEBUG +extern int __intel_debug; + +#define DEBUG_SWAP 0x1 +#define DEBUG_LOCK 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_BATCH 0x8 + +#define DBG(flag, ...) do { \ + if (__intel_debug & (DEBUG_##flag)) \ + _mesa_printf(__VA_ARGS__); \ +} while(0) + #else -#define INTEL_DEBUG 0 +#define DBG(flag, ...) #endif -#define DEBUG_TEXTURE 0x1 -#define DEBUG_STATE 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_BLIT 0x8 -#define DEBUG_MIPTREE 0x10 -#define DEBUG_VERBOSE 0x40 -#define DEBUG_BATCH 0x80 -#define DEBUG_PIXEL 0x100 -#define DEBUG_BUFMGR 0x200 -#define DEBUG_REGION 0x400 -#define DEBUG_FBO 0x800 -#define DEBUG_LOCK 0x1000 - -#define DBG(...) do { if (INTEL_DEBUG & FILE_DEBUG_FLAG) _mesa_printf(__VA_ARGS__); } while(0) #define PCI_CHIP_845_G 0x2562 diff --git a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c index 154a3e289f..6e4d134b80 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c @@ -40,7 +40,6 @@ #include "intel_blit.h" #include "drm.h" -#define FILE_DEBUG_FLAG DEBUG_IOCTL int intelEmitIrqLocked(struct intel_context *intel) @@ -60,7 +59,7 @@ intelEmitIrqLocked(struct intel_context *intel) exit(1); } - DBG("%s --> %d\n", __FUNCTION__, seq); + DBG(IOCTL, "%s --> %d\n", __FUNCTION__, seq); return seq; } @@ -70,7 +69,7 @@ intelWaitIrq(struct intel_context *intel, int seq) { int ret; - DBG("%s %d\n", __FUNCTION__, seq); + DBG(IOCTL, "%s %d\n", __FUNCTION__, seq); intel->iw.irq_seq = seq; @@ -98,7 +97,7 @@ intel_batch_ioctl(struct intel_context *intel, assert(intel->locked); assert(used); - DBG("%s used %d offset %x..%x ignore_cliprects %d\n", + DBG(IOCTL, "%s used %d offset %x..%x ignore_cliprects %d\n", __FUNCTION__, used, start_offset, start_offset + used, ignore_cliprects); @@ -116,7 +115,7 @@ intel_batch_ioctl(struct intel_context *intel, batch.DR1 = 0; batch.DR4 = 0; /* still need this ? */ - DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n", + DBG(IOCTL, "%s: 0x%x..0x%x DR4: %x cliprects: %d\n", __FUNCTION__, batch.start, batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); -- cgit v1.2.3 From 3cc5b1645b6f04f546b8535ba46bb418b8199f49 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:15:15 -0600 Subject: include surface.offset in address calculations --- src/mesa/pipe/softpipe/sp_surface.c | 34 ++++++++++++++++++++++++---------- 1 file changed, 24 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 054e4188e9..2ad168cc90 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -166,7 +166,8 @@ z16_read_quad_z(struct softpipe_surface *sps, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) { const GLushort *src - = (GLushort *) sps->surface.region->map + y * sps->surface.region->pitch + x; + = ((const GLushort *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z16); @@ -182,7 +183,9 @@ static void z16_write_quad_z(struct softpipe_surface *sps, GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) { - GLushort *dst = (GLushort *) sps->surface.region->map + y * sps->surface.region->pitch + x; + GLushort *dst + = ((GLushort *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z16); @@ -202,7 +205,8 @@ z32_read_quad_z(struct softpipe_surface *sps, GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) { const GLuint *src - = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z32); @@ -217,7 +221,9 @@ static void z32_write_quad_z(struct softpipe_surface *sps, GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) { - GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + GLuint *dst + = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z32); @@ -237,7 +243,8 @@ s8z24_read_quad_z(struct softpipe_surface *sps, { static const GLuint mask = 0x00ffffff; const GLuint *src - = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_S8_Z24); @@ -254,7 +261,9 @@ s8z24_write_quad_z(struct softpipe_surface *sps, GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) { static const GLuint mask = 0xff000000; - GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + GLuint *dst + = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_S8_Z24); assert(zzzz[0] <= 0xffffff); @@ -271,7 +280,8 @@ s8z24_read_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) { const GLuint *src - = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_S8_Z24); @@ -287,7 +297,9 @@ s8z24_write_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) { static const GLuint mask = 0x00ffffff; - GLuint *dst = (GLuint *) sps->surface.region->map + y * sps->surface.region->pitch + x; + GLuint *dst + = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_S8_Z24); @@ -306,7 +318,8 @@ s8_read_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) { const GLubyte *src - = sps->surface.region->map + y * sps->surface.region->pitch + x; + = sps->surface.region->map + sps->surface.offset + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_S8); @@ -322,7 +335,8 @@ s8_write_quad_stencil(struct softpipe_surface *sps, GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) { GLubyte *dst - = sps->surface.region->map + y * sps->surface.region->pitch + x; + = sps->surface.region->map + sps->surface.offset + + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_S8); -- cgit v1.2.3 From 0aa8b1643b6b32860a38285e886d60c6f38fccf5 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:15:33 -0600 Subject: added pipe_surface_unreference() --- src/mesa/pipe/p_context.h | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 23961f4b8f..b5dd149603 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -248,6 +248,17 @@ pipe_surface_reference(struct pipe_surface **dst, struct pipe_surface *src) } } +static INLINE void +pipe_surface_unreference(struct pipe_surface **ps) +{ + assert(*ps); + (*ps)->refcount--; + if ((*ps)->refcount <= 0) { + /* XXX need a proper surface->free method */ + free(*ps); + } + *ps = NULL; +} #endif /* PIPE_CONTEXT_H */ -- cgit v1.2.3 From 6349bd3112116841326885550188224af87ec15c Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:16:53 -0600 Subject: remove some temporary hacks --- src/mesa/main/renderbuffer.c | 50 +++----------------------------------------- 1 file changed, 3 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c index e7aeea8e3e..d576a7a121 100644 --- a/src/mesa/main/renderbuffer.c +++ b/src/mesa/main/renderbuffer.c @@ -49,9 +49,6 @@ #include "rbadaptors.h" -#include "pipe/p_state.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" #include "state_tracker/st_context.h" @@ -950,7 +947,6 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, GLenum internalFormat, GLuint width, GLuint height) { - struct pipe_context *pipe = ctx->st->pipe; GLuint pixelSize; /* first clear these fields */ @@ -1067,11 +1063,6 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutMonoValues = put_mono_values_ubyte; rb->StencilBits = 8 * sizeof(GLubyte); pixelSize = sizeof(GLubyte); -#if 0 - if (!rb->surface) - rb->surface = (struct pipe_surface *) - pipe->surface_alloc(pipe, PIPE_FORMAT_U_S8); -#endif break; case GL_STENCIL_INDEX16_EXT: rb->_ActualFormat = GL_STENCIL_INDEX16_EXT; @@ -1102,11 +1093,6 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutValues = put_values_ushort; rb->PutMonoValues = put_mono_values_ushort; rb->DepthBits = 8 * sizeof(GLushort); -#if 0 - if (!rb->surface) - rb->surface = (struct pipe_surface *) - pipe->surface_alloc(pipe, PIPE_FORMAT_U_Z16); -#endif pixelSize = sizeof(GLushort); break; case GL_DEPTH_COMPONENT24: @@ -1129,11 +1115,6 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->_ActualFormat = GL_DEPTH_COMPONENT32; rb->DepthBits = 32; } -#if 0 - if (!rb->surface) - rb->surface = (struct pipe_surface *) - pipe->surface_alloc(pipe, PIPE_FORMAT_U_Z32); -#endif pixelSize = sizeof(GLuint); break; case GL_DEPTH_STENCIL_EXT: @@ -1151,11 +1132,6 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->PutMonoValues = put_mono_values_uint; rb->DepthBits = 24; rb->StencilBits = 8; -#if 0 - if (!rb->surface) - rb->surface = (struct pipe_surface *) - pipe->surface_alloc(pipe, PIPE_FORMAT_S8_Z24); -#endif pixelSize = sizeof(GLuint); break; case GL_COLOR_INDEX8_EXT: @@ -1218,34 +1194,14 @@ _mesa_soft_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb, ASSERT(rb->PutMonoValues); /* free old buffer storage */ - if (0/**rb->surface**/) { - /* pipe_surface/region */ - } - else if (rb->Data) { - /* legacy renderbuffer (this will go away) */ + if (rb->Data) { _mesa_free(rb->Data); + rb->Data = NULL; } - rb->Data = NULL; if (width > 0 && height > 0) { /* allocate new buffer storage */ - if (0/**rb->surface**/) { - /* pipe_surface/region */ -#if 0 - if (rb->surface->region) { - pipe->region_unmap(pipe, rb->surface->region); - pipe->region_release(pipe, &rb->surface->region); - } - rb->surface->region = pipe->region_alloc(pipe, pixelSize, width, height); - /* XXX probably don't want to really map here */ - pipe->region_map(pipe, rb->surface->region); - rb->Data = rb->surface->region->map; -#endif - } - else { - /* legacy renderbuffer (this will go away) */ - rb->Data = malloc(width * height * pixelSize); - } + rb->Data = malloc(width * height * pixelSize); if (rb->Data == NULL) { rb->Width = 0; -- cgit v1.2.3 From 137ec509946bba39940d3a7932bf196450cb951e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:17:26 -0600 Subject: sketch out some render-to-texture code --- src/mesa/state_tracker/st_cb_fbo.c | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index bee7474e38..99c00916f4 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -43,7 +43,9 @@ #include "pipe/p_defines.h" #include "st_context.h" #include "st_cb_fbo.h" +#include "st_cb_texture.h" #include "st_format.h" +#include "st_public.h" @@ -281,10 +283,17 @@ st_render_texture(GLcontext *ctx, struct gl_renderbuffer_attachment *att) { struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; struct pipe_framebuffer_state framebuffer; struct pipe_surface *texsurface; + struct pipe_mipmap_tree *mt; - texsurface = NULL; /* find the mipmap level, cube face, etc */ + mt = st_get_texobj_mipmap_tree(att->Texture); + + texsurface = pipe->get_tex_surface(pipe, mt, + att->CubeMapFace, + att->TextureLevel, + att->Zoffset); /* * XXX basically like this... set the current color (or depth) @@ -308,7 +317,12 @@ static void st_finish_render_texture(GLcontext *ctx, struct gl_renderbuffer_attachment *att) { - /* restore drawing to normal framebuffer. may be a no-op */ + struct st_renderbuffer *strb = st_renderbuffer(att->Renderbuffer); + + pipe_surface_unreference(&strb->surface); + + /* restore previous framebuffer state */ + st_invalidate_state(ctx, _NEW_BUFFERS); } -- cgit v1.2.3 From 3738606a52b585108d709ce6be8565fc6144cbc7 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:35:57 -0600 Subject: more render to texture work --- src/mesa/state_tracker/st_cb_fbo.c | 19 ++++++++----------- 1 file changed, 8 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 99c00916f4..64ec802216 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -283,26 +283,23 @@ st_render_texture(GLcontext *ctx, struct gl_renderbuffer_attachment *att) { struct st_context *st = ctx->st; + struct st_renderbuffer *strb = st_renderbuffer(att->Renderbuffer); struct pipe_context *pipe = st->pipe; struct pipe_framebuffer_state framebuffer; - struct pipe_surface *texsurface; struct pipe_mipmap_tree *mt; mt = st_get_texobj_mipmap_tree(att->Texture); - texsurface = pipe->get_tex_surface(pipe, mt, - att->CubeMapFace, - att->TextureLevel, - att->Zoffset); + /* the renderbuffer's surface is inside the mipmap_tree: */ + strb->surface = pipe->get_tex_surface(pipe, mt, + att->CubeMapFace, + att->TextureLevel, + att->Zoffset); - /* - * XXX basically like this... set the current color (or depth) - * drawing surface to be the given texture renderbuffer. - */ + /* update pipe's framebuffer state */ memset(&framebuffer, 0, sizeof(framebuffer)); framebuffer.num_cbufs = 1; - framebuffer.cbufs[0] = texsurface; - + framebuffer.cbufs[0] = strb->surface; if (memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0) { st->state.framebuffer = framebuffer; st->pipe->set_framebuffer_state( st->pipe, &framebuffer ); -- cgit v1.2.3 From 9879ca019b1ed99ca37e7b485ef4daae0fded066 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:36:23 -0600 Subject: add missing break stmt --- src/mesa/pipe/softpipe/sp_surface.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 2ad168cc90..5ef8ec935f 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -356,6 +356,7 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) sps->read_quad_f_swz = a8r8g8b8_read_quad_f_swz; sps->write_quad_f_swz = a8r8g8b8_write_quad_f_swz; sps->surface.get_tile = a8r8g8b8_get_tile; + break; case PIPE_FORMAT_U_A1_R5_G5_B5: sps->surface.get_tile = a1r5g5b5_get_tile; break; -- cgit v1.2.3 From c943dab9a36ac1b48a3da570d44128caa67c9f51 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:37:05 -0600 Subject: glReadPixels --- src/mesa/state_tracker/st_cb_readpixels.c | 75 +++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_readpixels.h | 36 +++++++++++++++ 2 files changed, 111 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_readpixels.c create mode 100644 src/mesa/state_tracker/st_cb_readpixels.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c new file mode 100644 index 0000000000..9738aa3c70 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -0,0 +1,75 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * glReadPixels interface to pipe + * + * \author Brian Paul + */ + + +#include "main/imports.h" +#include "main/context.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "st_context.h" +#include "st_cb_readpixels.h" +#include "st_cb_fbo.h" +#include "st_format.h" +#include "st_public.h" + + +static void +st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + GLvoid *dest) +{ + struct pipe_context *pipe = ctx->st->pipe; + + if (format == GL_RGBA && type == GL_FLOAT) { + struct st_renderbuffer *strb + = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + if (!strb) + return; + + pipe->region_map(pipe, strb->surface->region); + strb->surface->get_tile(strb->surface, x, y, width, height, dest); + pipe->region_unmap(pipe, strb->surface->region); + } + else { + assert(0); + } + +} + +void st_init_readpixels_functions(struct dd_function_table *functions) +{ + functions->ReadPixels = st_readpixels; +} diff --git a/src/mesa/state_tracker/st_cb_readpixels.h b/src/mesa/state_tracker/st_cb_readpixels.h new file mode 100644 index 0000000000..1dbe9727a5 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_readpixels.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_READPIXELS_H +#define ST_CB_READPIXELS_H + +extern void +st_init_readpixels_functions(struct dd_function_table *functions); + + +#endif /* ST_CB_READPIXELS_H */ -- cgit v1.2.3 From 6f27aabe49f45d1c4915d4d5e12723774c6bbdeb Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:37:21 -0600 Subject: use st_cb_readpixels.c --- src/mesa/state_tracker/st_context.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 9f7916e40e..80935442ae 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -32,12 +32,14 @@ #include "st_cb_clear.h" #include "st_cb_drawpixels.h" #include "st_cb_fbo.h" +#include "st_cb_readpixels.h" #include "st_cb_texture.h" #include "st_atom.h" #include "st_draw.h" #include "st_program.h" #include "pipe/p_context.h" + void st_invalidate_state(GLcontext * ctx, GLuint new_state) { struct st_context *st = st_context(ctx); @@ -105,5 +107,6 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_drawpixels_functions(functions); st_init_fbo_functions(functions); st_init_program_functions(functions); + st_init_readpixels_functions(functions); st_init_texture_functions(functions); } -- cgit v1.2.3 From 900b0807b80dc79c99afbd063fa014d76c8e7692 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:37:41 -0600 Subject: added st_cb_readpixels.c --- src/mesa/sources | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index be81e102ed..54a0978715 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -192,12 +192,13 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_stipple.c \ state_tracker/st_atom_texture.c \ state_tracker/st_atom_viewport.c \ + state_tracker/st_cb_bufferobjects.c \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_drawpixels.c \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_program.c \ + state_tracker/st_cb_readpixels.c \ state_tracker/st_cb_texture.c \ - state_tracker/st_cb_bufferobjects.c \ state_tracker/st_context.c \ state_tracker/st_draw.c \ state_tracker/st_format.c \ -- cgit v1.2.3 From 9cad4dde9745d2342e0694879a84e1a7006b9b0d Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 10 Aug 2007 11:45:06 +0100 Subject: Support predicates. Predicates, as per D3D9 spec, are a complete subset of NVIDIA's condition codes. However, due to its broad support in non-NVIDIA hardware, it is being exposed in TGSI to ease hw driver development. Two decomposition functions are expected in the future. 1. GetRidOfPredicates() - predicates are implemented with condition codes. 2. There is no second function, it's just this pesky VI that does not allow me to move with arrow keys and so I cannot edit previous paragraphs. --- src/mesa/pipe/tgsi/core/tgsi_token.h | 71 +++++++++++++++++++++++++++--------- 1 file changed, 54 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index 3552331e79..a94595ca43 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -1102,6 +1102,7 @@ struct tgsi_instruction #define TGSI_INSTRUCTION_EXT_TYPE_NV 0 #define TGSI_INSTRUCTION_EXT_TYPE_LABEL 1 #define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE 2 +#define TGSI_INSTRUCTION_EXT_TYPE_PREDICATE 3 struct tgsi_instruction_ext { @@ -1120,6 +1121,9 @@ struct tgsi_instruction_ext * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it * should be cast to tgsi_instruction_ext_texture. * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_PREDICATE, it + * should be cast to tgsi_instruction_ext_predicate. + * * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext * follows. */ @@ -1201,6 +1205,32 @@ struct tgsi_instruction_ext_texture GLuint Extended : 1; /* BOOL */ }; +#define TGSI_WRITEMASK_NONE 0x00 +#define TGSI_WRITEMASK_X 0x01 +#define TGSI_WRITEMASK_Y 0x02 +#define TGSI_WRITEMASK_XY 0x03 +#define TGSI_WRITEMASK_Z 0x04 +#define TGSI_WRITEMASK_XZ 0x05 +#define TGSI_WRITEMASK_YZ 0x06 +#define TGSI_WRITEMASK_XYZ 0x07 +#define TGSI_WRITEMASK_W 0x08 +#define TGSI_WRITEMASK_XW 0x09 +#define TGSI_WRITEMASK_YW 0x0A +#define TGSI_WRITEMASK_XYW 0x0B +#define TGSI_WRITEMASK_ZW 0x0C +#define TGSI_WRITEMASK_XZW 0x0D +#define TGSI_WRITEMASK_YZW 0x0E +#define TGSI_WRITEMASK_XYZW 0x0F + +struct tgsi_instruction_ext_predicate +{ + GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ + GLuint PredDstIndex : 4; /* UINT */ + GLuint PredWriteMask : 4; /* TGSI_WRITEMASK_ */ + GLuint Padding : 19; + GLuint Extended : 1; /* BOOL */ +}; + /* * File specifies the register array to access. * @@ -1331,23 +1361,6 @@ struct tgsi_dimension GLuint Extended : 1; /* BOOL */ }; -#define TGSI_WRITEMASK_NONE 0x00 -#define TGSI_WRITEMASK_X 0x01 -#define TGSI_WRITEMASK_Y 0x02 -#define TGSI_WRITEMASK_XY 0x03 -#define TGSI_WRITEMASK_Z 0x04 -#define TGSI_WRITEMASK_XZ 0x05 -#define TGSI_WRITEMASK_YZ 0x06 -#define TGSI_WRITEMASK_XYZ 0x07 -#define TGSI_WRITEMASK_W 0x08 -#define TGSI_WRITEMASK_XW 0x09 -#define TGSI_WRITEMASK_YW 0x0A -#define TGSI_WRITEMASK_XYW 0x0B -#define TGSI_WRITEMASK_ZW 0x0C -#define TGSI_WRITEMASK_XZW 0x0D -#define TGSI_WRITEMASK_YZW 0x0E -#define TGSI_WRITEMASK_XYZW 0x0F - struct tgsi_dst_register { GLuint File : 4; /* TGSI_FILE_ */ @@ -1367,6 +1380,7 @@ struct tgsi_dst_register #define TGSI_DST_REGISTER_EXT_TYPE_CONDCODE 0 #define TGSI_DST_REGISTER_EXT_TYPE_MODULATE 1 +#define TGSI_DST_REGISTER_EXT_TYPE_PREDICATE 2 struct tgsi_dst_register_ext { @@ -1382,6 +1396,9 @@ struct tgsi_dst_register_ext * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_MODULATE, * it should be cast to tgsi_dst_register_ext_modulate. * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_PREDICATE, + * it should be cast to tgsi_dst_register_ext_predicate. + * * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext * follows. */ @@ -1415,6 +1432,26 @@ struct tgsi_dst_register_ext_modulate GLuint Extended : 1; /* BOOL */ }; +/* + * Currently, the following constraints apply. + * + * - PredSwizzleXYZW is either set to identity or replicate. + * - PredSrcIndex is 0. + */ + +struct tgsi_dst_register_ext_predicate +{ + GLuint Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */ + GLuint PredSwizzleX : 2; /* TGSI_SWIZZLE_ */ + GLuint PredSwizzleY : 2; /* TGSI_SWIZZLE_ */ + GLuint PredSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + GLuint PredSwizzleW : 2; /* TGSI_SWIZZLE_ */ + GLuint PredSrcIndex : 4; /* UINT */ + GLuint Negate : 1; /* BOOL */ + GLuint Padding : 14; + GLuint Extended : 1; /* BOOL */ +}; + #if defined __cplusplus } // extern "C" #endif // defined __cplusplus -- cgit v1.2.3 From 80e8538a5d7f99b59f4577defdb7023e39dc5f69 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 10 Aug 2007 11:54:39 +0100 Subject: Add SGT/LT/GE/LE/EQ/NE opcodes for ps/vs_2_x. --- src/mesa/pipe/tgsi/core/tgsi_token.h | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index a94595ca43..ec62836ef3 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -894,6 +894,12 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_DP4 */ /* TGSI_OPCODE_MIN */ /* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SNE */ /* EXP - use TGSI_OPCODE_EX2 */ /* LOG - use TGSI_OPCODE_LG2 */ /* TGSI_OPCODE_LRP */ @@ -1023,6 +1029,10 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MAX */ /* TGSI_OPCODE_SLT */ /* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SNE */ /* EXP - use TGSI_OPCODE_EX2 */ /* LOG - use TGSI_OPCODE_LG2 */ /* TGSI_OPCODE_LIT */ -- cgit v1.2.3 From 519aacef031e3271e16693308ca462346a8a160c Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 11:43:04 -0600 Subject: fix asst glitches --- src/mesa/state_tracker/st_cb_drawpixels.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 20990fe80a..f0c7a2bfc5 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -81,7 +81,7 @@ make_drawpixels_shader(struct st_context *st) /* END; */ p->Instructions[1].Opcode = OPCODE_END; - p->InputsRead = FRAG_BIT_COL0; + p->InputsRead = FRAG_BIT_TEX0; p->OutputsWritten = (1 << FRAG_RESULT_COLR); stfp = (struct st_fragment_program *) p; @@ -100,7 +100,7 @@ make_mipmap_tree(struct st_context *st, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { - GLuint pipeFormat = st_choose_pipe_format(st->pipe, format, type, GL_RGBA); + GLuint pipeFormat = st_choose_pipe_format(st->pipe, GL_RGBA, format, type); int cpp = 4, pitch; struct pipe_mipmap_tree *mt = CALLOC_STRUCT(pipe_mipmap_tree); @@ -249,7 +249,7 @@ static GLboolean compatible_formats(GLenum format, GLenum type, GLuint pipeFormat) { static const GLuint one = 1; - GLubyte littleEndian = *((GLubyte *) one); + GLubyte littleEndian = *((GLubyte *) &one); if (pipeFormat == PIPE_FORMAT_U_R8_G8_B8_A8 && format == GL_RGBA && -- cgit v1.2.3 From 9ac1a8d416c2bd50ca10186ca09f5e86f6fa4ce6 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 12:13:48 -0600 Subject: pipe->region_alloc() now takes width instead of pitch, plus a flags param --- src/mesa/pipe/i915simple/i915_regions.c | 11 +++++++++-- src/mesa/pipe/p_context.h | 3 ++- src/mesa/pipe/p_defines.h | 19 +++++++++++++++---- src/mesa/pipe/p_state.h | 10 ---------- src/mesa/pipe/softpipe/sp_region.c | 2 +- src/mesa/state_tracker/st_cb_bufferobjects.c | 5 ++++- src/mesa/state_tracker/st_cb_drawpixels.c | 9 ++++----- src/mesa/state_tracker/st_cb_fbo.c | 9 +++------ src/mesa/state_tracker/st_mipmap_tree.c | 9 ++++++--- 9 files changed, 44 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index e2b807f959..b3bcae547d 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -31,6 +31,7 @@ * - refcounting of buffer mappings. */ +#include "pipe/p_defines.h" #include "i915_context.h" #include "i915_winsys.h" #include "i915_blit.h" @@ -70,7 +71,7 @@ i915_region_unmap(struct pipe_context *pipe, struct pipe_region *region) static struct pipe_region * i915_region_alloc(struct pipe_context *pipe, - GLuint cpp, GLuint width, GLuint height) + GLuint cpp, GLuint width, GLuint height, GLbitfield flags) { struct i915_context *i915 = i915_context( pipe ); struct pipe_region *region = calloc(sizeof(*region), 1); @@ -82,7 +83,13 @@ i915_region_alloc(struct pipe_context *pipe, * clearly want to be able to render to textures under some * circumstances, but maybe not always a requirement. */ - unsigned pitch = ((cpp * width + 63) & ~63) / cpp; + unsigned pitch; + + /* XXX is the pitch different for textures vs. drawables? */ + if (flags & PIPE_SURFACE_FLAG_TEXTURE) /* or PIPE_SURFACE_FLAG_RENDER? */ + pitch = ((cpp * width + 63) & ~63) / cpp; + else + pitch = ((cpp * width + 63) & ~63) / cpp; region->cpp = cpp; region->pitch = pitch; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index b5dd149603..7eb492816b 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -140,7 +140,8 @@ struct pipe_context { * Some of these may go away... */ struct pipe_region *(*region_alloc)(struct pipe_context *pipe, - GLuint cpp, GLuint pitch, GLuint height); + GLuint cpp, GLuint width, GLuint height, + GLbitfield flags); void (*region_release)(struct pipe_context *pipe, struct pipe_region **r); diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index efbb3239a2..52ecd5b119 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -176,11 +176,22 @@ /** - * Buffer mapping access modes + * Surface flags */ -#define PIPE_MAP_READ 1 -#define PIPE_MAP_WRITE 2 -#define PIPE_MAP_READ_WRITE 3 +#define PIPE_SURFACE_FLAG_TEXTURE 0x1 +#define PIPE_SURFACE_FLAG_RENDER 0x2 + + +/** + * Buffer flags + */ +#define PIPE_BUFFER_FLAG_READ 0x1 +#define PIPE_BUFFER_FLAG_WRITE 0x2 + +#define PIPE_BUFFER_USE_TEXTURE 0x1 +#define PIPE_BUFFER_USE_VERTEX_BUFFER 0x2 +#define PIPE_BUFFER_USE_INDEX_BUFFER 0x4 +#define PIPE_BUFFER_USE_RENDER_TARGET 0x8 /** diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 64a475ba00..df456cc2ed 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -329,15 +329,5 @@ struct pipe_mipmap_tree struct pipe_buffer_handle; -#define PIPE_BUFFER_FLAG_READ 0x1 -#define PIPE_BUFFER_FLAG_WRITE 0x2 - -#define PIPE_BUFFER_USE_TEXTURE 0x1 -#define PIPE_BUFFER_USE_VERTEX_BUFFER 0x2 -#define PIPE_BUFFER_USE_INDEX_BUFFER 0x4 -#define PIPE_BUFFER_USE_RENDER_TARGET 0x8 - - - #endif diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 58749492ec..1db508f028 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -70,7 +70,7 @@ sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) static struct pipe_region * sp_region_alloc(struct pipe_context *pipe, - GLuint cpp, GLuint pitch, GLuint height) + GLuint cpp, GLuint pitch, GLuint height, GLbitfield flags) { struct softpipe_context *sp = softpipe_context( pipe ); struct pipe_region *region = calloc(sizeof(*region), 1); diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index d020eb2007..78fc18a49a 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -34,6 +34,9 @@ #include "st_cb_bufferobjects.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" + + /* Pixel buffers and Vertex/index buffers are handled through these * mesa callbacks. Framebuffer/Renderbuffer objects are @@ -160,7 +163,7 @@ st_bufferobj_map(GLcontext *ctx, flags = PIPE_BUFFER_FLAG_WRITE; break; - + case GL_READ_ONLY: flags = PIPE_BUFFER_FLAG_READ; break; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index f0c7a2bfc5..6981097ef4 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -101,20 +101,19 @@ make_mipmap_tree(struct st_context *st, const GLvoid *pixels) { GLuint pipeFormat = st_choose_pipe_format(st->pipe, GL_RGBA, format, type); - int cpp = 4, pitch; + int cpp = 4; struct pipe_mipmap_tree *mt = CALLOC_STRUCT(pipe_mipmap_tree); + GLbitfield flags = PIPE_SURFACE_FLAG_TEXTURE; assert(pipeFormat); - pitch = width; /* XXX pad */ - if (unpack->BufferObj) { /* mt->region = buffer_object_region(unpack->BufferObj); */ } else { - mt->region = st->pipe->region_alloc(st->pipe, cpp, pitch, height); + mt->region = st->pipe->region_alloc(st->pipe, cpp, width, height, flags); /* XXX do texstore() here */ } @@ -128,7 +127,7 @@ make_mipmap_tree(struct st_context *st, mt->depth0 = 1; mt->cpp = cpp; mt->compressed = 0; - mt->pitch = pitch; + mt->pitch = mt->region->pitch; mt->depth_pitch = 0; mt->total_height = height; mt->level[0].level_offset = 0; diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 64ec802216..bb588d10ea 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -62,7 +62,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, const GLuint pipeFormat = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE); const struct pipe_format_info *info = st_get_format_info(pipeFormat); - GLuint cpp, pitch; + GLuint cpp; + GLbitfield flags = PIPE_SURFACE_FLAG_RENDER; /* want to render to surface */ assert(info); if (!info) @@ -98,11 +99,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe->region_release(pipe, &strb->surface->region); } - /* Choose a pitch to match hardware requirements: - */ - pitch = ((cpp * width + 63) & ~63) / cpp; /* XXX fix: device-specific */ - - strb->surface->region = pipe->region_alloc(pipe, cpp, pitch, height); + strb->surface->region = pipe->region_alloc(pipe, cpp, width, height, flags); if (!strb->surface->region) return GL_FALSE; /* out of memory, try s/w buffer? */ diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index ac74335ba1..3cbe697ab3 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -62,6 +62,7 @@ st_miptree_create(struct pipe_context *pipe, { GLboolean ok; struct pipe_mipmap_tree *mt = calloc(sizeof(*mt), 1); + GLbitfield flags = 0x0; DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), @@ -79,9 +80,11 @@ st_miptree_create(struct pipe_context *pipe, mt->refcount = 1; ok = pipe->mipmap_tree_layout(pipe, mt); - if (ok) - mt->region = pipe->region_alloc(pipe, - mt->cpp, mt->pitch, mt->total_height); + if (ok) { + /* note: it's OK to pass 'pitch' as 'width' here: */ + mt->region = pipe->region_alloc(pipe, mt->cpp, mt->pitch, + mt->total_height, flags); + } if (!mt->region) { free(mt); -- cgit v1.2.3 From 5c2c05600081f811e001a81a600778de0fcab85d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 12:57:14 +0100 Subject: Handle glFlush/glFinish through the state tracker. --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 46 ------------ .../dri/intel_winsys/intel_pipe_i915simple.c | 14 +++- .../drivers/dri/intel_winsys/intel_pipe_softpipe.c | 31 +++++++++ src/mesa/pipe/i915simple/i915_flush.c | 5 +- src/mesa/pipe/i915simple/i915_winsys.h | 6 ++ src/mesa/pipe/p_context.h | 6 +- src/mesa/pipe/softpipe/sp_context.c | 2 +- src/mesa/pipe/softpipe/sp_flush.c | 19 ++++- src/mesa/pipe/softpipe/sp_flush.h | 2 +- src/mesa/pipe/softpipe/sp_winsys.h | 12 +++- src/mesa/sources | 1 + src/mesa/state_tracker/st_cb_flush.c | 81 ++++++++++++++++++++++ src/mesa/state_tracker/st_cb_flush.h | 38 ++++++++++ src/mesa/state_tracker/st_context.c | 2 + src/mesa/state_tracker/st_context.h | 4 ++ 15 files changed, 212 insertions(+), 57 deletions(-) create mode 100644 src/mesa/state_tracker/st_cb_flush.c create mode 100644 src/mesa/state_tracker/st_cb_flush.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 1032fc2d0f..a1746e0965 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -240,51 +240,7 @@ intelFlush(GLcontext * ctx) } -/** - * Check if we need to rotate/warp the front color buffer to the - * rotated screen. We generally need to do this when we get a glFlush - * or glFinish after drawing to the front color buffer. - * If no rotation, just copy the private fake front buffer to the real one. - */ -static void -intelCheckFrontUpdate(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - /* rely on _ColorDrawBufferMask being kept up to date by mesa - even for window-fbos. */ - /* not sure. Might need that for all masks including - BUFFER_BIT_FRONT_LEFT maybe? */ - if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - intelCopyBuffer(dPriv, NULL); - } -} - -/** - * Called via glFlush. - */ -static void -intelglFlush(GLcontext * ctx) -{ - intelFlush(ctx); - intelCheckFrontUpdate(ctx); -} - -void -intelFinish(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - intelFlush(ctx); - if (intel->batch->last_fence) { - driFenceFinish(intel->batch->last_fence, - 0, GL_FALSE); - driFenceUnReference(intel->batch->last_fence); - intel->batch->last_fence = NULL; - } - intelCheckFrontUpdate(ctx); -} static void @@ -292,8 +248,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) { _mesa_init_driver_functions(functions); - functions->Flush = intelglFlush; - functions->Finish = intelFinish; functions->GetString = intelGetString; functions->UpdateState = intelInvalidateState; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c index c0e8c2349c..d78d81aa1c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c @@ -37,6 +37,7 @@ #include "intel_context.h" #include "intel_batchbuffer.h" #include "intel_pipe.h" +#include "intel_blit.h" #include "pipe/i915simple/i915_winsys.h" @@ -171,7 +172,7 @@ static unsigned *intel_i915_batch_start( struct i915_winsys *sws, if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { /* XXX: Hmm, the driver can't really do much with this pointer: */ - return intel->batch->ptr; + return (unsigned *)intel->batch->ptr; } else return NULL; @@ -242,6 +243,16 @@ static void intel_i915_printf( struct i915_winsys *sws, } +static void +intel_i915_flush_frontbuffer( struct i915_winsys *sws ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelCopyBuffer(dPriv, NULL); +} + + struct pipe_context * intel_create_i915simple( struct intel_context *intel ) { @@ -264,6 +275,7 @@ intel_create_i915simple( struct intel_context *intel ) iws->winsys.batch_reloc = intel_i915_batch_reloc; iws->winsys.batch_flush = intel_i915_batch_flush; iws->winsys.batch_wait_idle = intel_i915_batch_wait_idle; + iws->winsys.flush_frontbuffer = intel_i915_flush_frontbuffer; iws->intel = intel; /* Create the i915simple context: diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c index 06db5c9f43..439bb372e1 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c @@ -36,6 +36,8 @@ #include "intel_context.h" #include "intel_pipe.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" #include "pipe/softpipe/sp_winsys.h" #include "pipe/p_defines.h" @@ -175,6 +177,33 @@ intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats) } +static void intel_wait_idle( struct softpipe_winsys *sws ) +{ + struct intel_context *intel = intel_softpipe_winsys(sws)->intel; + + if (intel->batch->last_fence) { + driFenceFinish(intel->batch->last_fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); + driFenceUnReference(intel->batch->last_fence); + intel->batch->last_fence = NULL; + } +} + + +/* The state tracker (should!) keep track of whether the fake + * frontbuffer has been touched by any rendering since the last time + * we copied its contents to the real frontbuffer. Our task is easy: + */ +static void +intel_flush_frontbuffer( struct softpipe_winsys *sws ) +{ + struct intel_context *intel = intel_softpipe_winsys(sws)->intel; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelCopyBuffer(dPriv, NULL); +} + + struct pipe_context * intel_create_softpipe( struct intel_context *intel ) @@ -197,6 +226,8 @@ intel_create_softpipe( struct intel_context *intel ) isws->sws.buffer_data = intel_buffer_data; isws->sws.buffer_subdata = intel_buffer_subdata; isws->sws.buffer_get_subdata = intel_buffer_get_subdata; + isws->sws.flush_frontbuffer = intel_flush_frontbuffer; + isws->sws.wait_idle = intel_wait_idle; isws->intel = intel; /* Create the softpipe context: diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c index 1cf945e9a2..8af4ce770c 100644 --- a/src/mesa/pipe/i915simple/i915_flush.c +++ b/src/mesa/pipe/i915simple/i915_flush.c @@ -66,11 +66,10 @@ static void i915_flush( struct pipe_context *pipe, FLUSH_BATCH(); } -static void i915_finish(struct pipe_context *pipe) +static void i915_wait_idle(struct pipe_context *pipe) { struct i915_context *i915 = i915_context(pipe); - i915_flush( pipe, 0 ); i915->winsys->batch_wait_idle( i915->winsys ); } @@ -78,5 +77,5 @@ static void i915_finish(struct pipe_context *pipe) void i915_init_flush_functions( struct i915_context *i915 ) { i915->pipe.flush = i915_flush; - i915->pipe.finish = i915_finish; + i915->pipe.wait_idle = i915_wait_idle; } diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 9802148aa1..a3dadbfd3d 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -50,6 +50,12 @@ struct pipe_buffer_handle; struct i915_winsys { + /* Do any special operations to ensure frontbuffer contents are + * displayed, eg copy fake frontbuffer. + */ + void (*flush_frontbuffer)( struct i915_winsys *sws ); + + /* debug output */ void (*printf)( struct i915_winsys *sws, diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 7eb492816b..533840c555 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -220,10 +220,14 @@ struct pipe_context { struct pipe_mipmap_tree *mt ); + /* Simple flush/finish support: + */ void (*flush)( struct pipe_context *pipe, unsigned flags ); - void (*finish)( struct pipe_context *pipe ); + void (*wait_idle)( struct pipe_context *pipe ); + + void (*flush_frontbuffer)( struct pipe_context *pipe ); }; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 0794d9a888..db572f169d 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -221,7 +221,7 @@ struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) softpipe->pipe.draw_vertices = softpipe_draw_vertices; softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; - softpipe->pipe.finish = softpipe_finish; + softpipe->pipe.wait_idle = softpipe_wait_idle; softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index aa609469a6..a0bce200ed 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -33,6 +33,7 @@ #include "pipe/p_defines.h" #include "sp_flush.h" #include "sp_context.h" +#include "sp_winsys.h" /* There will be actual work to do here. In future we may want a * fence-like interface instead of finish, and perhaps flush will take @@ -49,9 +50,21 @@ softpipe_flush( struct pipe_context *pipe, } void -softpipe_finish(struct pipe_context *pipe) +softpipe_wait_idle(struct pipe_context *pipe) { - /* Just calls into flush() + /* Nothing to do. + * XXX: What about swapbuffers. + * XXX: Even more so - what about fake frontbuffer copies?? */ - softpipe_flush( pipe, 0 ); + struct softpipe_context *softpipe = softpipe_context(pipe); + softpipe->winsys->wait_idle( softpipe->winsys ); +} + + +void +softpipe_flush_frontbuffer( struct pipe_context *pipe ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->winsys->flush_frontbuffer( softpipe->winsys ); } diff --git a/src/mesa/pipe/softpipe/sp_flush.h b/src/mesa/pipe/softpipe/sp_flush.h index 5e204f87a5..03c0010623 100644 --- a/src/mesa/pipe/softpipe/sp_flush.h +++ b/src/mesa/pipe/softpipe/sp_flush.h @@ -30,7 +30,7 @@ struct pipe_context; -void softpipe_finish(struct pipe_context *pipe); void softpipe_flush(struct pipe_context *pipe, unsigned flags ); +void softpipe_wait_idle(struct pipe_context *pipe); #endif diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h index 6f6d0f2446..73b0659067 100644 --- a/src/mesa/pipe/softpipe/sp_winsys.h +++ b/src/mesa/pipe/softpipe/sp_winsys.h @@ -50,9 +50,19 @@ struct pipe_buffer_handle; struct softpipe_winsys { + /* Do any special operations to ensure frontbuffer contents are + * displayed, eg copy fake frontbuffer. + */ + void (*flush_frontbuffer)( struct softpipe_winsys *sws ); + + /* Wait for any hw swapbuffers, etc. to finish: + */ + void (*wait_idle)( struct softpipe_winsys *sws ); + /* debug output */ - void (*printf)( const char *, ... ); + void (*printf)( struct softpipe_winsys *sws, + const char *, ... ); /* The buffer manager is modeled after the dri_bugmgr interface, diff --git a/src/mesa/sources b/src/mesa/sources index 54a0978715..a2b9fed82b 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -194,6 +194,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_viewport.c \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_cb_clear.c \ + state_tracker/st_cb_flush.c \ state_tracker/st_cb_drawpixels.c \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_program.c \ diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c new file mode 100644 index 0000000000..a6a8f8d90f --- /dev/null +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -0,0 +1,81 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "main/glheader.h" +#include "main/macros.h" +#include "st_context.h" +#include "st_cb_flush.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + + +static void st_flush(GLcontext *ctx) +{ + struct st_context *st = ctx->st; + + /* If there has been no rendering to the frontbuffer, consider + * short-circuiting this, or perhaps pass an "optional" flag down + * to the driver so that it can make the decision. + */ + st->pipe->flush( st->pipe, 0 ); + + + /* XXX: temporary hack. This flag should only be set if we do any + * rendering to the front buffer. + */ + st->flags.frontbuffer_dirty = (ctx->DrawBuffer->_ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT); + + + if (st->flags.frontbuffer_dirty) { + /* Hook for copying "fake" frontbuffer if necessary: + */ + st->pipe->flush_frontbuffer( st->pipe ); + st->flags.frontbuffer_dirty = 0; + } +} + +static void st_finish(GLcontext *ctx) +{ + struct st_context *st = ctx->st; + + st_flush( ctx ); + st->pipe->wait_idle( st->pipe ); +} + + +void st_init_flush_functions(struct dd_function_table *functions) +{ + functions->Flush = st_flush; + functions->Finish = st_finish; +} diff --git a/src/mesa/state_tracker/st_cb_flush.h b/src/mesa/state_tracker/st_cb_flush.h new file mode 100644 index 0000000000..29ceab3a56 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_flush.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_CLEAR_H +#define ST_CB_CLEAR_H + + +extern void +st_init_flush_functions(struct dd_function_table *functions); + + +#endif /* ST_CB_CLEAR_H */ + diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 80935442ae..2bdcc6caf3 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -34,6 +34,7 @@ #include "st_cb_fbo.h" #include "st_cb_readpixels.h" #include "st_cb_texture.h" +#include "st_cb_flush.h" #include "st_atom.h" #include "st_draw.h" #include "st_program.h" @@ -109,4 +110,5 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_program_functions(functions); st_init_readpixels_functions(functions); st_init_texture_functions(functions); + st_init_flush_functions(functions); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index fe73630c75..88e2701fc3 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -90,6 +90,10 @@ struct st_context struct gl_fragment_program *fragment_program; } cb; + struct { + GLuint frontbuffer_dirty:1; + } flags; + /* State to be validated: */ struct st_tracked_state **atoms; -- cgit v1.2.3 From 48f853f73f7bb46e8c977e551fd66a39e50935ee Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 12:24:38 -0600 Subject: minor clean-ups --- src/mesa/state_tracker/st_cb_bufferobjects.c | 33 ++++++++++++++-------------- 1 file changed, 16 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index 78fc18a49a..9e407de34f 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -26,9 +26,9 @@ **************************************************************************/ -#include "imports.h" -#include "mtypes.h" -#include "bufferobj.h" +#include "main/imports.h" +#include "main/mtypes.h" +#include "main/bufferobj.h" #include "st_context.h" #include "st_cb_bufferobjects.h" @@ -57,6 +57,9 @@ st_bufferobj_alloc(GLcontext *ctx, GLuint name, GLenum target) struct st_context *st = st_context(ctx); struct st_buffer_object *st_obj = CALLOC_STRUCT(st_buffer_object); + if (!st_obj) + return NULL; + _mesa_initialize_buffer_object(&st_obj->Base, name, target); st_obj->buffer = st->pipe->create_buffer( st->pipe, 32, 0 ); @@ -79,7 +82,7 @@ st_bufferobj_free(GLcontext *ctx, struct gl_buffer_object *obj) if (st_obj->buffer) pipe->buffer_unreference(pipe, &st_obj->buffer); - FREE(st_obj); + free(st_obj); } @@ -133,10 +136,10 @@ st_bufferobj_subdata(GLcontext *ctx, */ static void st_bufferobj_get_subdata(GLcontext *ctx, - GLenum target, - GLintptrARB offset, - GLsizeiptrARB size, - GLvoid * data, struct gl_buffer_object *obj) + GLenum target, + GLintptrARB offset, + GLsizeiptrARB size, + GLvoid * data, struct gl_buffer_object *obj) { struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); @@ -145,14 +148,12 @@ st_bufferobj_get_subdata(GLcontext *ctx, } - /** * Called via glMapBufferARB(). */ static void * -st_bufferobj_map(GLcontext *ctx, - GLenum target, - GLenum access, struct gl_buffer_object *obj) +st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access, + struct gl_buffer_object *obj) { struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); @@ -162,12 +163,11 @@ st_bufferobj_map(GLcontext *ctx, case GL_WRITE_ONLY: flags = PIPE_BUFFER_FLAG_WRITE; break; - case GL_READ_ONLY: flags = PIPE_BUFFER_FLAG_READ; break; - case GL_READ_WRITE: + /* fall-through */ default: flags = PIPE_BUFFER_FLAG_READ | PIPE_BUFFER_FLAG_WRITE; break; @@ -182,8 +182,7 @@ st_bufferobj_map(GLcontext *ctx, * Called via glMapBufferARB(). */ static GLboolean -st_bufferobj_unmap(GLcontext *ctx, - GLenum target, struct gl_buffer_object *obj) +st_bufferobj_unmap(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj) { struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); -- cgit v1.2.3 From da9178c73088982b9f422187cf782ae5ab6b64ce Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 12:37:20 -0600 Subject: alignment/pitch/width clean-ups --- src/mesa/pipe/i915simple/i915_regions.c | 3 ++- src/mesa/pipe/softpipe/sp_region.c | 23 +++++++++++++++++------ 2 files changed, 19 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index b3bcae547d..886b089506 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -75,6 +75,7 @@ i915_region_alloc(struct pipe_context *pipe, { struct i915_context *i915 = i915_context( pipe ); struct pipe_region *region = calloc(sizeof(*region), 1); + const unsigned alignment = 64; /* Choose a pitch to match hardware requirements - requires 64 byte * alignment of render targets. @@ -96,7 +97,7 @@ i915_region_alloc(struct pipe_context *pipe, region->height = height; /* needed? */ region->refcount = 1; - region->buffer = i915->winsys->buffer_create( i915->winsys, 64 ); + region->buffer = i915->winsys->buffer_create( i915->winsys, alignment ); i915->winsys->buffer_data( i915->winsys, region->buffer, diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 1db508f028..142d121047 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -36,6 +36,16 @@ #include "sp_region.h" +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + + static void sp_region_idle(struct pipe_context *pipe, struct pipe_region *region) { @@ -70,23 +80,24 @@ sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) static struct pipe_region * sp_region_alloc(struct pipe_context *pipe, - GLuint cpp, GLuint pitch, GLuint height, GLbitfield flags) + GLuint cpp, GLuint width, GLuint height, GLbitfield flags) { struct softpipe_context *sp = softpipe_context( pipe ); struct pipe_region *region = calloc(sizeof(*region), 1); + const unsigned alignment = 64; region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ + region->pitch = round_up(width, alignment / cpp); + region->height = height; region->refcount = 1; - region->buffer = sp->winsys->create_buffer( sp->winsys, 64 ); + region->buffer = sp->winsys->create_buffer( sp->winsys, alignment ); + /* NULL data --> just allocate the space */ sp->winsys->buffer_data( sp->winsys, region->buffer, - pitch * cpp * height, + region->pitch * cpp * height, NULL ); - return region; } -- cgit v1.2.3 From 24e510ea3796d3136c5eacf2c96a1dddded12aa2 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 13:02:25 -0600 Subject: added st_mesa_format_to_pipe_format() --- src/mesa/state_tracker/st_format.c | 14 ++++++++++++++ src/mesa/state_tracker/st_format.h | 4 ++++ 2 files changed, 18 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 4f36c84d4e..4aa5066b8c 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -98,6 +98,20 @@ st_get_format_info(GLuint format) } +GLuint +st_mesa_format_to_pipe_format(GLuint mesaFormat) +{ + switch (mesaFormat) { + /* fix this */ + case MESA_FORMAT_ARGB8888_REV: + case MESA_FORMAT_ARGB8888: + return PIPE_FORMAT_U_A8_R8_G8_B8; + default: + assert(0); + return 0; + } +} + /** * Search list of formats for first RGBA format. diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index 452ae59d6d..dde4731e94 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -50,6 +50,10 @@ extern const struct pipe_format_info * st_get_format_info(GLuint format); +extern GLuint +st_mesa_format_to_pipe_format(GLuint mesaFormat); + + extern GLuint st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, GLenum format, GLenum type); -- cgit v1.2.3 From 9cf9aa1ea2c3f40a09316975410a4b0e202e82ba Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 13:02:51 -0600 Subject: use st_mesa_format_to_pipe_format() --- src/mesa/state_tracker/st_cb_texture.c | 19 ++----------------- 1 file changed, 2 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 22c1b4baff..f633bd328f 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -114,21 +114,6 @@ st_get_texobj_mipmap_tree(struct gl_texture_object *texObj) } -static GLuint -mesa_format_to_pipe_format(GLuint mesaFormat) -{ - switch (mesaFormat) { - /* fix this */ - case MESA_FORMAT_ARGB8888_REV: - case MESA_FORMAT_ARGB8888: - return PIPE_FORMAT_U_A8_R8_G8_B8; - default: - assert(0); - return 0; - } -} - - static int compressed_num_bytes(GLuint mesaFormat) { @@ -386,7 +371,7 @@ guess_and_alloc_mipmap_tree(struct pipe_context *pipe, comp_byte); stObj->mt->format - = mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat); + = st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat); DBG("%s - success\n", __FUNCTION__); } @@ -1519,7 +1504,7 @@ st_finalize_mipmap_tree(GLcontext *ctx, comp_byte); stObj->mt->format - = mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat); + = st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat); } /* Pull in any images not in the object's tree: -- cgit v1.2.3 From e51aa572934c39fe3c99470343f776be4e783f42 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 13:03:13 -0600 Subject: basic texture-based glDrawPixels path works --- src/mesa/state_tracker/st_cb_drawpixels.c | 81 +++++++++++++++++++++++-------- 1 file changed, 60 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 6981097ef4..afe0342957 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -100,21 +100,54 @@ make_mipmap_tree(struct st_context *st, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { + struct pipe_context *pipe = st->pipe; + const struct gl_texture_format *mformat; GLuint pipeFormat = st_choose_pipe_format(st->pipe, GL_RGBA, format, type); int cpp = 4; struct pipe_mipmap_tree *mt = CALLOC_STRUCT(pipe_mipmap_tree); GLbitfield flags = PIPE_SURFACE_FLAG_TEXTURE; + mformat = st_ChooseTextureFormat(st->ctx, GL_RGBA, format, type); + assert(format); + + pipeFormat = st_mesa_format_to_pipe_format(mformat->MesaFormat); assert(pipeFormat); - if (unpack->BufferObj) { + if (unpack->BufferObj && unpack->BufferObj->Name) { /* mt->region = buffer_object_region(unpack->BufferObj); */ } else { + static const GLuint dstImageOffsets = 0; + GLboolean success; + GLubyte *dest; + GLuint pitch; + + /* allocate texture region/storage */ mt->region = st->pipe->region_alloc(st->pipe, cpp, width, height, flags); - /* XXX do texstore() here */ + pitch = mt->region->pitch; + + /* map texture region */ + dest = pipe->region_map(pipe, mt->region); + + /* put image into texture region */ + success = mformat->StoreImage(st->ctx, 2, /* dims */ + GL_RGBA, /* baseInternalFormat */ + mformat, /* gl_texture_format */ + dest, /* dest */ + 0, 0, 0, /* dstX/Y/Zoffset */ + pitch * cpp, /* dstRowStride */ + &dstImageOffsets, /* dstImageOffsets */ + width, height, 1, /* size */ + format, type, /* src format/type */ + pixels, /* data source */ + unpack); + + /* unmap */ + pipe->region_unmap(pipe, mt->region); + + assert(success); } mt->target = GL_TEXTURE_2D; @@ -146,10 +179,10 @@ static void free_mipmap_tree(struct pipe_context *pipe, struct pipe_mipmap_tree *mt) { pipe->region_release(pipe, &mt->region); + free(mt); } - static void draw_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, GLsizei width, GLsizei height) @@ -183,7 +216,7 @@ draw_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, verts[3][0][0] = x; verts[3][0][1] = y + height; verts[3][1][0] = 0.0; - verts[3][1][1] = 11.0; + verts[3][1][1] = 1.0; /* same for all verts: */ for (i = 0; i < 4; i++) { @@ -204,32 +237,38 @@ draw_textured_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { - static struct st_fragment_program *stfp = NULL; struct pipe_mipmap_tree *mt; - struct pipe_setup_state setup; - struct pipe_fs_state fs; /* setup state: just scissor */ - memset(&setup, 0, sizeof(setup)); - if (st->ctx->Scissor.Enabled) - setup.scissor = 1; - st->pipe->set_setup_state(st->pipe, &setup); + { + struct pipe_setup_state setup; + memset(&setup, 0, sizeof(setup)); + if (st->ctx->Scissor.Enabled) + setup.scissor = 1; + st->pipe->set_setup_state(st->pipe, &setup); + } /* fragment shader state: color pass-through program */ - if (!stfp) { - stfp = make_drawpixels_shader(st); + { + static struct st_fragment_program *stfp = NULL; + struct pipe_fs_state fs; + if (!stfp) { + stfp = make_drawpixels_shader(st); + } + memset(&fs, 0, sizeof(fs)); + fs.inputs_read = stfp->Base.Base.InputsRead; + fs.tokens = &stfp->tokens[0]; + fs.constants = NULL; + st->pipe->set_fs_state(st->pipe, &fs); } - memset(&fs, 0, sizeof(fs)); - fs.inputs_read = stfp->Base.Base.InputsRead; - fs.tokens = &stfp->tokens[0]; - fs.constants = NULL; - st->pipe->set_fs_state(st->pipe, &fs); /* mipmap tree state: */ - mt = make_mipmap_tree(st, width, height, format, type, unpack, pixels); - st->pipe->set_texture_state(st->pipe, 0, mt); + { + mt = make_mipmap_tree(st, width, height, format, type, unpack, pixels); + st->pipe->set_texture_state(st->pipe, 0, mt); + } - /* draw! */ + /* draw textured quad */ draw_quad(st, x, y, z, width, height); /* restore GL state */ -- cgit v1.2.3 From ce02b16dac0891a570a3a42f26463ea5c68f1618 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 13:11:31 -0600 Subject: added st_sizeof_format() --- src/mesa/state_tracker/st_format.c | 9 +++++++++ src/mesa/state_tracker/st_format.h | 4 ++++ 2 files changed, 13 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 4aa5066b8c..a89dfb10fb 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -98,6 +98,15 @@ st_get_format_info(GLuint format) } +GLuint +st_sizeof_format(GLuint pipeFormat) +{ + const struct pipe_format_info *info = st_get_format_info(pipeFormat); + assert(info); + return info->size; +} + + GLuint st_mesa_format_to_pipe_format(GLuint mesaFormat) { diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index dde4731e94..6b3cba0f63 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -50,6 +50,10 @@ extern const struct pipe_format_info * st_get_format_info(GLuint format); +extern GLuint +st_sizeof_format(GLuint pipeFormat); + + extern GLuint st_mesa_format_to_pipe_format(GLuint mesaFormat); -- cgit v1.2.3 From 6ccdd5bb15d03b7eadc0f1e3ddba8848de4f146e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 13:11:47 -0600 Subject: don't hardcode cpp --- src/mesa/state_tracker/st_cb_drawpixels.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index afe0342957..81b48859f0 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -102,16 +102,18 @@ make_mipmap_tree(struct st_context *st, { struct pipe_context *pipe = st->pipe; const struct gl_texture_format *mformat; - GLuint pipeFormat = st_choose_pipe_format(st->pipe, GL_RGBA, format, type); - int cpp = 4; - struct pipe_mipmap_tree *mt = CALLOC_STRUCT(pipe_mipmap_tree); - GLbitfield flags = PIPE_SURFACE_FLAG_TEXTURE; + const GLbitfield flags = PIPE_SURFACE_FLAG_TEXTURE; + struct pipe_mipmap_tree *mt; + GLuint pipeFormat, cpp; mformat = st_ChooseTextureFormat(st->ctx, GL_RGBA, format, type); - assert(format); + assert(mformat); pipeFormat = st_mesa_format_to_pipe_format(mformat->MesaFormat); assert(pipeFormat); + cpp = st_sizeof_format(pipeFormat); + + mt = CALLOC_STRUCT(pipe_mipmap_tree); if (unpack->BufferObj && unpack->BufferObj->Name) { /* @@ -137,7 +139,7 @@ make_mipmap_tree(struct st_context *st, mformat, /* gl_texture_format */ dest, /* dest */ 0, 0, 0, /* dstX/Y/Zoffset */ - pitch * cpp, /* dstRowStride */ + pitch * cpp, /* dstRowStride, bytes */ &dstImageOffsets, /* dstImageOffsets */ width, height, 1, /* size */ format, type, /* src format/type */ -- cgit v1.2.3 From bc4aa83794c6336358793c5f428973fb22184050 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 13:30:22 -0600 Subject: setup sampler state, computed wincoords with pixel zoom factors --- src/mesa/state_tracker/st_cb_drawpixels.c | 89 ++++++++++++++++++++----------- 1 file changed, 59 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 81b48859f0..d177f2d6be 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -185,9 +185,13 @@ free_mipmap_tree(struct pipe_context *pipe, struct pipe_mipmap_tree *mt) } +/** + * Draw textured quad. + * Y=0=top + */ static void -draw_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, - GLsizei width, GLsizei height) +draw_quad(struct st_context *st, GLfloat x0, GLfloat y0, GLfloat z, + GLfloat x1, GLfloat y1) { static const GLuint attribs[2] = { VF_ATTRIB_POS, @@ -196,29 +200,29 @@ draw_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ GLuint i; - /* lower-left */ - verts[0][0][0] = x; - verts[0][0][1] = y; + /* upper-left */ + verts[0][0][0] = x0; + verts[0][0][1] = y0; verts[0][1][0] = 0.0; - verts[0][1][1] = 0.0; + verts[0][1][1] = 1.0; - /* lower-right */ - verts[1][0][0] = x + width; - verts[1][0][1] = y; + /* upper-right */ + verts[1][0][0] = x1; + verts[1][0][1] = y0; verts[1][1][0] = 1.0; - verts[1][1][1] = 0.0; + verts[1][1][1] = 1.0; - /* upper-right */ - verts[2][0][0] = x + width; - verts[2][0][1] = y + height; + /* lower-right */ + verts[2][0][0] = x1; + verts[2][0][1] = y1; verts[2][1][0] = 1.0; - verts[2][1][1] = 1.0; + verts[2][1][1] = 0.0; - /* upper-left */ - verts[3][0][0] = x; - verts[3][0][1] = y + height; + /* lower-left */ + verts[3][0][0] = x0; + verts[3][0][1] = y1; verts[3][1][0] = 0.0; - verts[3][1][1] = 1.0; + verts[3][1][1] = 0.0; /* same for all verts: */ for (i = 0; i < 4; i++) { @@ -234,20 +238,23 @@ draw_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, static void -draw_textured_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, +draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, GLsizei width, GLsizei height, GLenum format, GLenum type, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { + const GLuint unit = 0; + struct pipe_context *pipe = ctx->st->pipe; struct pipe_mipmap_tree *mt; + GLfloat x0, y0, x1, y1; /* setup state: just scissor */ { struct pipe_setup_state setup; memset(&setup, 0, sizeof(setup)); - if (st->ctx->Scissor.Enabled) + if (ctx->Scissor.Enabled) setup.scissor = 1; - st->pipe->set_setup_state(st->pipe, &setup); + pipe->set_setup_state(pipe, &setup); } /* fragment shader state: color pass-through program */ @@ -255,29 +262,51 @@ draw_textured_quad(struct st_context *st, GLfloat x, GLfloat y, GLfloat z, static struct st_fragment_program *stfp = NULL; struct pipe_fs_state fs; if (!stfp) { - stfp = make_drawpixels_shader(st); + stfp = make_drawpixels_shader(ctx->st); } memset(&fs, 0, sizeof(fs)); fs.inputs_read = stfp->Base.Base.InputsRead; fs.tokens = &stfp->tokens[0]; fs.constants = NULL; - st->pipe->set_fs_state(st->pipe, &fs); + pipe->set_fs_state(pipe, &fs); + } + + /* texture sampling state: */ + { + struct pipe_sampler_state sampler; + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_REPEAT; + sampler.wrap_t = PIPE_TEX_WRAP_REPEAT; + sampler.wrap_r = PIPE_TEX_WRAP_REPEAT; + sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; + pipe->set_sampler_state(pipe, unit, &sampler); } /* mipmap tree state: */ { - mt = make_mipmap_tree(st, width, height, format, type, unpack, pixels); - st->pipe->set_texture_state(st->pipe, 0, mt); + mt = make_mipmap_tree(ctx->st, width, height, format, type, + unpack, pixels); + pipe->set_texture_state(pipe, unit, mt); } + /* compute window coords (y=0=top) with pixel zoom */ + x0 = x; + y0 = ctx->DrawBuffer->Height - 1 - y; + x1 = x + width * ctx->Pixel.ZoomX; + y1 = ctx->DrawBuffer->Height - 1 - (y + height * ctx->Pixel.ZoomY); + /* draw textured quad */ - draw_quad(st, x, y, z, width, height); + draw_quad(ctx->st, x0, y0, z, x1, y1); /* restore GL state */ - st->pipe->set_setup_state(st->pipe, &st->state.setup); - st->pipe->set_fs_state(st->pipe, &st->state.fs); + pipe->set_setup_state(pipe, &ctx->st->state.setup); + pipe->set_fs_state(pipe, &ctx->st->state.fs); + /* XXX FIX: pipe->set_texture_state(pipe, unit, ???); */ + pipe->set_sampler_state(pipe, unit, &ctx->st->state.sampler[0]); - free_mipmap_tree(st->pipe, mt); + free_mipmap_tree(pipe, mt); } @@ -420,7 +449,7 @@ st_drawpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, any_pixel_transfer_ops(st) || !compatible_formats(format, type, ps->format)) { /* textured quad */ - draw_textured_quad(st, x, y, ctx->Current.RasterPos[2], width, height, + draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, format, type, unpack, pixels); } else { -- cgit v1.2.3 From 04ee3cecdf8ef39695bdc66120a4e3bbb9f40aca Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 15:29:41 -0600 Subject: readpixels handles any format/type/packing/transfer ops now --- src/mesa/state_tracker/st_cb_readpixels.c | 66 ++++++++++++++++++++++++++----- 1 file changed, 56 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 9738aa3c70..273ed5770b 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -35,6 +35,7 @@ #include "main/imports.h" #include "main/context.h" +#include "main/image.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -45,30 +46,75 @@ #include "st_public.h" +/** + * Do glReadPixels by getting rows from the framebuffer surface with + * get_tile(). Convert to requested format/type with Mesa image routines. + * Image transfer ops are done in software too. + */ static void st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, + const struct gl_pixelstore_attrib *pack, GLvoid *dest) { struct pipe_context *pipe = ctx->st->pipe; + GLfloat temp[MAX_WIDTH][4]; + const GLbitfield transferOps = ctx->_ImageTransferState; + GLint i, yInv, dfStride; + GLfloat *df; + struct st_renderbuffer *strb; + struct gl_pixelstore_attrib clippedPacking = *pack; + + /* XXX convolution not done yet */ + assert((transferOps & IMAGE_CONVOLUTION_BIT) == 0); + + /* Do all needed clipping here, so that we can forget about it later */ + if (!_mesa_clip_readpixels(ctx, &x, &y, &width, &height, &clippedPacking)) { + /* The ReadPixels region is totally outside the window bounds */ + return; + } + + + /* XXX check pack->BufferObj !!! */ + + + strb = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + if (!strb) + return; + + pipe->region_map(pipe, strb->surface->region); if (format == GL_RGBA && type == GL_FLOAT) { - struct st_renderbuffer *strb - = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); - if (!strb) - return; - - pipe->region_map(pipe, strb->surface->region); - strb->surface->get_tile(strb->surface, x, y, width, height, dest); - pipe->region_unmap(pipe, strb->surface->region); + /* write tile(row) directly into user's buffer */ + df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width, + height, format, type, 0, 0); + dfStride = width * 4; } else { - assert(0); + /* write tile(row) into temp row buffer */ + df = (GLfloat *) temp; + dfStride = 0; } + /* Do a row at a time to flip image data vertically */ + yInv = strb->Base.Height - 1 - y; + for (i = 0; i < height; i++) { + strb->surface->get_tile(strb->surface, x, yInv, width, 1, df); + yInv--; + df += dfStride; + if (!dfStride) { + /* convert GLfloat to user's format/type */ + GLvoid *dst = _mesa_image_address2d(&clippedPacking, dest, width, + height, format, type, i, 0); + _mesa_pack_rgba_span_float(ctx, width, temp, format, type, dst, + &clippedPacking, transferOps); + } + } + + pipe->region_unmap(pipe, strb->surface->region); } + void st_init_readpixels_functions(struct dd_function_table *functions) { functions->ReadPixels = st_readpixels; -- cgit v1.2.3 From 12e3bb1a65bbff82dabc64110249c57a711501c1 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 15:29:56 -0600 Subject: fix image orientation --- src/mesa/state_tracker/st_cb_drawpixels.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index d177f2d6be..a5ba614429 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -133,7 +133,10 @@ make_mipmap_tree(struct st_context *st, /* map texture region */ dest = pipe->region_map(pipe, mt->region); - /* put image into texture region */ + /* Put image into texture region. + * Note that the image is actually going to be upside down in + * the texture. We deal with that with texcoords. + */ success = mformat->StoreImage(st->ctx, 2, /* dims */ GL_RGBA, /* baseInternalFormat */ mformat, /* gl_texture_format */ @@ -204,25 +207,25 @@ draw_quad(struct st_context *st, GLfloat x0, GLfloat y0, GLfloat z, verts[0][0][0] = x0; verts[0][0][1] = y0; verts[0][1][0] = 0.0; - verts[0][1][1] = 1.0; + verts[0][1][1] = 0.0; /* upper-right */ verts[1][0][0] = x1; verts[1][0][1] = y0; verts[1][1][0] = 1.0; - verts[1][1][1] = 1.0; + verts[1][1][1] = 0.0; /* lower-right */ verts[2][0][0] = x1; verts[2][0][1] = y1; verts[2][1][0] = 1.0; - verts[2][1][1] = 0.0; + verts[2][1][1] = 1.0; /* lower-left */ verts[3][0][0] = x0; verts[3][0][1] = y1; verts[3][1][0] = 0.0; - verts[3][1][1] = 0.0; + verts[3][1][1] = 1.0; /* same for all verts: */ for (i = 0; i < 4; i++) { -- cgit v1.2.3 From 47fc2c4349746997704a7f81dffadd22363e0ff1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 15:31:26 +0100 Subject: Lift common winsys functions into pipe's new p_winsys. --- src/mesa/drivers/dri/i915tex/intel_render.c | 7 +- src/mesa/drivers/dri/i915tex/intel_tris.c | 12 + src/mesa/drivers/dri/intel_winsys/Makefile | 5 +- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 +- src/mesa/drivers/dri/intel_winsys/intel_pipe.h | 42 --- .../dri/intel_winsys/intel_pipe_i915simple.c | 284 --------------------- .../drivers/dri/intel_winsys/intel_pipe_softpipe.c | 242 ------------------ src/mesa/drivers/dri/intel_winsys/intel_winsys.h | 45 ++++ .../drivers/dri/intel_winsys/intel_winsys_i915.c | 144 +++++++++++ .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 221 ++++++++++++++++ .../dri/intel_winsys/intel_winsys_softpipe.c | 79 ++++++ src/mesa/pipe/i915simple/Makefile | 1 - src/mesa/pipe/i915simple/i915_buffer.c | 121 --------- src/mesa/pipe/i915simple/i915_context.c | 17 +- src/mesa/pipe/i915simple/i915_context.h | 3 +- src/mesa/pipe/i915simple/i915_debug.c | 2 +- src/mesa/pipe/i915simple/i915_debug.h | 10 +- src/mesa/pipe/i915simple/i915_debug_fp.c | 2 +- src/mesa/pipe/i915simple/i915_flush.c | 7 - src/mesa/pipe/i915simple/i915_regions.c | 26 +- src/mesa/pipe/i915simple/i915_state.c | 2 +- src/mesa/pipe/i915simple/i915_winsys.h | 58 +---- src/mesa/pipe/p_context.h | 47 +--- src/mesa/pipe/p_winsys.h | 114 +++++++++ src/mesa/pipe/softpipe/Makefile | 1 - src/mesa/pipe/softpipe/sp_buffer.c | 120 --------- src/mesa/pipe/softpipe/sp_buffer.h | 40 --- src/mesa/pipe/softpipe/sp_context.c | 9 +- src/mesa/pipe/softpipe/sp_flush.c | 19 -- src/mesa/pipe/softpipe/sp_flush.h | 1 - src/mesa/pipe/softpipe/sp_region.c | 23 +- src/mesa/pipe/softpipe/sp_winsys.h | 76 +----- src/mesa/state_tracker/st_cb_bufferobjects.c | 15 +- src/mesa/state_tracker/st_cb_flush.c | 5 +- src/mesa/state_tracker/st_cb_flush.h | 6 +- 35 files changed, 703 insertions(+), 1105 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_pipe.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_winsys.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c delete mode 100644 src/mesa/pipe/i915simple/i915_buffer.c create mode 100644 src/mesa/pipe/p_winsys.h delete mode 100644 src/mesa/pipe/softpipe/sp_buffer.c delete mode 100644 src/mesa/pipe/softpipe/sp_buffer.h (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/intel_render.c b/src/mesa/drivers/dri/i915tex/intel_render.c index c8b6d308d9..dddbe9906e 100644 --- a/src/mesa/drivers/dri/i915tex/intel_render.c +++ b/src/mesa/drivers/dri/i915tex/intel_render.c @@ -129,8 +129,9 @@ do { \ ((intel->batch->size - 1500) / (intel->vertex_size*4)) #define GET_CURRENT_VB_MAX_VERTS() GET_SUBSEQUENT_VB_MAX_VERTS() -#define ALLOC_VERTS( nr ) \ - intelExtendInlinePrimitive( intel, (nr) * intel->vertex_size ) +#define ALLOC_VERTS( nr ) NULL + +// intelExtendInlinePrimitive( intel, (nr) * intel->vertex_size ) #define EMIT_VERTS( ctx, j, nr, buf ) \ _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf ) @@ -204,6 +205,8 @@ intel_run_render(GLcontext * ctx, struct tnl_pipeline_stage *stage) intel->vtbl.render_prevalidate( intel ); + return GL_TRUE; + /* Don't handle clipping or indexed vertices. */ if (intel->RenderIndex != 0 || diff --git a/src/mesa/drivers/dri/i915tex/intel_tris.c b/src/mesa/drivers/dri/i915tex/intel_tris.c index 5fe3d4561f..48dae9d720 100644 --- a/src/mesa/drivers/dri/i915tex/intel_tris.c +++ b/src/mesa/drivers/dri/i915tex/intel_tris.c @@ -185,6 +185,7 @@ intel_draw_quad(struct intel_context *intel, intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3) { +#ifndef NO_REND GLuint vertsize = intel->vertex_size; GLuint *vb = intelExtendInlinePrimitive(intel, 6 * vertsize); int j; @@ -207,12 +208,14 @@ intel_draw_quad(struct intel_context *intel, COPY_DWORDS(j, vb, vertsize, v2); COPY_DWORDS(j, vb, vertsize, v3); +#endif } static void intel_draw_triangle(struct intel_context *intel, intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2) { +#ifndef NO_REND GLuint vertsize = intel->vertex_size; GLuint *vb = intelExtendInlinePrimitive(intel, 3 * vertsize); int j; @@ -220,6 +223,7 @@ intel_draw_triangle(struct intel_context *intel, COPY_DWORDS(j, vb, vertsize, v0); COPY_DWORDS(j, vb, vertsize, v1); COPY_DWORDS(j, vb, vertsize, v2); +#endif } @@ -227,18 +231,21 @@ static void intel_draw_line(struct intel_context *intel, intelVertexPtr v0, intelVertexPtr v1) { +#ifndef NO_REND GLuint vertsize = intel->vertex_size; GLuint *vb = intelExtendInlinePrimitive(intel, 2 * vertsize); int j; COPY_DWORDS(j, vb, vertsize, v0); COPY_DWORDS(j, vb, vertsize, v1); +#endif } static void intel_draw_point(struct intel_context *intel, intelVertexPtr v0) { +#ifndef NO_REND GLuint vertsize = intel->vertex_size; GLuint *vb = intelExtendInlinePrimitive(intel, vertsize); int j; @@ -248,6 +255,7 @@ intel_draw_point(struct intel_context *intel, intelVertexPtr v0) *(float *) &vb[1] = v0->v.y - 0.125; for (j = 2; j < vertsize; j++) vb[j] = v0->ui[j]; +#endif } @@ -732,6 +740,7 @@ intelRenderClippedLine(GLcontext * ctx, GLuint ii, GLuint jj) static void intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n) { +#ifndef NO_REND struct intel_context *intel = intel_context(ctx); const GLuint vertsize = intel->vertex_size; GLuint *vb = intelExtendInlinePrimitive(intel, (n - 2) * 3 * vertsize); @@ -744,6 +753,7 @@ intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n) COPY_DWORDS(j, vb, vertsize, V(elts[i])); COPY_DWORDS(j, vb, vertsize, start); } +#endif } /**********************************************************************/ @@ -1069,6 +1079,7 @@ intel_meta_draw_poly(struct intel_context *intel, GLfloat xy[][2], GLfloat z, GLuint color, GLfloat tex[][2]) { +#ifndef NO_REND union fi *vb; GLint i; @@ -1089,6 +1100,7 @@ intel_meta_draw_poly(struct intel_context *intel, } INTEL_FIREVERTICES(intel); +#endif } void diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index ae08afccb3..a5db6547a9 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -11,8 +11,9 @@ PIPE_DRIVERS = \ $(TOP)/src/mesa/pipe/i915simple/libi915simple.a DRIVER_SOURCES = \ - intel_pipe_i915simple.c \ - intel_pipe_softpipe.c \ + intel_winsys_pipe.c \ + intel_winsys_softpipe.c \ + intel_winsys_i915.c \ intel_batchbuffer.c \ intel_buffers.c \ intel_blit.c \ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index a1746e0965..6d92ce7e19 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -41,7 +41,7 @@ #include "i830_dri.h" #include "intel_buffers.h" -#include "intel_pipe.h" +#include "intel_winsys.h" #include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "intel_blit.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe.h b/src/mesa/drivers/dri/intel_winsys/intel_pipe.h deleted file mode 100644 index 5d3e25aea2..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe.h +++ /dev/null @@ -1,42 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_PIPE_H -#define INTEL_PIPE_H - -struct intel_context; -struct pipe_context; - - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel ); - -struct pipe_context * -intel_create_i915simple( struct intel_context *intel ); - - -#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c deleted file mode 100644 index d78d81aa1c..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_i915simple.c +++ /dev/null @@ -1,284 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include -#include "dri_bufpool.h" -#include "dri_bufmgr.h" - -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_pipe.h" -#include "intel_blit.h" - -#include "pipe/i915simple/i915_winsys.h" - - -struct intel_i915_winsys { - struct i915_winsys winsys; - struct intel_context *intel; -}; - - -/* Turn a i915simple winsys into an intel/i915simple winsys: - */ -static inline struct intel_i915_winsys * -intel_i915_winsys( struct i915_winsys *sws ) -{ - return (struct intel_i915_winsys *)sws; -} - - - -/* Many of the winsys's are probably going to have a similar - * buffer-manager interface, as something almost identical is - * currently exposed in the pipe interface. Probably want to avoid - * endless repetition of this code somehow. - */ -static inline struct _DriBufferObject * -dri_bo( struct pipe_buffer_handle *bo ) -{ - return (struct _DriBufferObject *)bo; -} - -static inline struct pipe_buffer_handle * -pipe_bo( struct _DriBufferObject *bo ) -{ - return (struct pipe_buffer_handle *)bo; -} - - -/* Most callbacks map direcly onto dri_bufmgr operations: - */ -static void *intel_i915_buffer_map(struct i915_winsys *sws, - struct pipe_buffer_handle *buf ) -{ - return driBOMap( dri_bo(buf), - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0 ); -} - -static void intel_i915_buffer_unmap(struct i915_winsys *sws, - struct pipe_buffer_handle *buf) -{ - driBOUnmap( dri_bo(buf) ); -} - - -static struct pipe_buffer_handle * -intel_i915_buffer_reference(struct i915_winsys *sws, - struct pipe_buffer_handle *buf) -{ - return pipe_bo( driBOReference( dri_bo(buf) ) ); -} - -static void intel_i915_buffer_unreference(struct i915_winsys *sws, - struct pipe_buffer_handle **buf) -{ - if (*buf) { - driBOUnReference( dri_bo(*buf) ); - *buf = NULL; - } -} - -/* Grabs the hardware lock! - */ -static void intel_i915_buffer_data(struct i915_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - - LOCK_HARDWARE( intel ); - driBOData( dri_bo(buf), size, data, 0 ); - UNLOCK_HARDWARE( intel ); -} - -static void intel_i915_buffer_subdata(struct i915_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) -{ - driBOSubData( dri_bo(buf), offset, size, data ); -} - -static void intel_i915_buffer_get_subdata(struct i915_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) -{ - driBOGetSubData( dri_bo(buf), offset, size, data ); -} - -/* I915simple has no concept of pools. We choose the tex/region pool - * for all buffers. - */ -static struct pipe_buffer_handle * -intel_i915_buffer_create(struct i915_winsys *sws, - unsigned alignment) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - struct _DriBufferObject *buffer; - - LOCK_HARDWARE( intel ); - driGenBuffers( intel->intelScreen->regionPool, - "i915simple buffer", 1, &buffer, alignment, 0, 0 ); - UNLOCK_HARDWARE( intel ); - - return pipe_bo(buffer); -} - - -/* Simple batchbuffer interface: - */ - -static unsigned *intel_i915_batch_start( struct i915_winsys *sws, - unsigned dwords, - unsigned relocs ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - - /* XXX: check relocs. - */ - if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { - /* XXX: Hmm, the driver can't really do much with this pointer: - */ - return (unsigned *)intel->batch->ptr; - } - else - return NULL; -} - -static void intel_i915_batch_dword( struct i915_winsys *sws, - unsigned dword ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - intel_batchbuffer_emit_dword( intel->batch, dword ); -} - -static void intel_i915_batch_reloc( struct i915_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned access_flags, - unsigned delta ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - unsigned flags = DRM_BO_FLAG_MEM_TT; - unsigned mask = DRM_BO_MASK_MEM; - - if (access_flags & I915_BUFFER_ACCESS_WRITE) { - flags |= DRM_BO_FLAG_WRITE; - mask |= DRM_BO_FLAG_WRITE; - } - - if (access_flags & I915_BUFFER_ACCESS_READ) { - flags |= DRM_BO_FLAG_READ; - mask |= DRM_BO_FLAG_READ; - } - - intel_batchbuffer_emit_reloc( intel->batch, - dri_bo( buf ), - flags, mask, - delta ); -} - - -static void intel_i915_batch_wait_idle( struct i915_winsys *sws ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - - if (intel->batch->last_fence) { - driFenceFinish(intel->batch->last_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); - driFenceUnReference(intel->batch->last_fence); - intel->batch->last_fence = NULL; - } -} - - -static void intel_i915_batch_flush( struct i915_winsys *sws ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - - intel_batchbuffer_flush( intel->batch ); - if (0) intel_i915_batch_wait_idle( sws ); -} - - -static void intel_i915_printf( struct i915_winsys *sws, - const char *fmtString, ... ) -{ - va_list args; - va_start( args, fmtString ); - vfprintf(stderr, fmtString, args); - va_end( args ); -} - - -static void -intel_i915_flush_frontbuffer( struct i915_winsys *sws ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelCopyBuffer(dPriv, NULL); -} - - -struct pipe_context * -intel_create_i915simple( struct intel_context *intel ) -{ - struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); - - /* Fill in this struct with callbacks that i915simple will need to - * communicate with the window system, buffer manager, etc. - */ - iws->winsys.printf = intel_i915_printf; - iws->winsys.buffer_create = intel_i915_buffer_create; - iws->winsys.buffer_map = intel_i915_buffer_map; - iws->winsys.buffer_unmap = intel_i915_buffer_unmap; - iws->winsys.buffer_reference = intel_i915_buffer_reference; - iws->winsys.buffer_unreference = intel_i915_buffer_unreference; - iws->winsys.buffer_data = intel_i915_buffer_data; - iws->winsys.buffer_subdata = intel_i915_buffer_subdata; - iws->winsys.buffer_get_subdata = intel_i915_buffer_get_subdata; - iws->winsys.batch_start = intel_i915_batch_start; - iws->winsys.batch_dword = intel_i915_batch_dword; - iws->winsys.batch_reloc = intel_i915_batch_reloc; - iws->winsys.batch_flush = intel_i915_batch_flush; - iws->winsys.batch_wait_idle = intel_i915_batch_wait_idle; - iws->winsys.flush_frontbuffer = intel_i915_flush_frontbuffer; - iws->intel = intel; - - /* Create the i915simple context: - */ - return i915_create( &iws->winsys, intel->intelScreen->deviceID ); -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c deleted file mode 100644 index 439bb372e1..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_pipe_softpipe.c +++ /dev/null @@ -1,242 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include -#include "dri_bufpool.h" -#include "dri_bufmgr.h" - -#include "intel_context.h" -#include "intel_pipe.h" -#include "intel_batchbuffer.h" -#include "intel_blit.h" - -#include "pipe/softpipe/sp_winsys.h" -#include "pipe/p_defines.h" - -/* Shouldn't really need this: - */ -#include "pipe/p_context.h" - - -struct intel_softpipe_winsys { - struct softpipe_winsys sws; - struct intel_context *intel; -}; - - - -/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque - * buffer pointer... - */ -static inline struct _DriBufferObject * -dri_bo( struct pipe_buffer_handle *bo ) -{ - return (struct _DriBufferObject *)bo; -} - -static inline struct pipe_buffer_handle * -pipe_bo( struct _DriBufferObject *bo ) -{ - return (struct pipe_buffer_handle *)bo; -} - -/* Turn a softpipe winsys into an intel/softpipe winsys: - */ -static inline struct intel_softpipe_winsys * -intel_softpipe_winsys( struct softpipe_winsys *sws ) -{ - return (struct intel_softpipe_winsys *)sws; -} - - -/* Most callbacks map direcly onto dri_bufmgr operations: - */ -static void *intel_buffer_map(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf ) -{ - return driBOMap( dri_bo(buf), - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0 ); -} - -static void intel_buffer_unmap(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf) -{ - driBOUnmap( dri_bo(buf) ); -} - - -static struct pipe_buffer_handle * -intel_buffer_reference(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf) -{ - return pipe_bo( driBOReference( dri_bo(buf) ) ); -} - -static void intel_buffer_unreference(struct softpipe_winsys *sws, - struct pipe_buffer_handle **buf) -{ - if (*buf) { - driBOUnReference( dri_bo(*buf) ); - *buf = NULL; - } -} - -/* Grabs the hardware lock! - */ -static void intel_buffer_data(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ) -{ - struct intel_context *intel = intel_softpipe_winsys(sws)->intel; - - LOCK_HARDWARE( intel ); - driBOData( dri_bo(buf), size, data, 0 ); - UNLOCK_HARDWARE( intel ); -} - -static void intel_buffer_subdata(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) -{ - driBOSubData( dri_bo(buf), offset, size, data ); -} - -static void intel_buffer_get_subdata(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) -{ - driBOGetSubData( dri_bo(buf), offset, size, data ); -} - -/* Softpipe has no concept of pools. We choose the tex/region pool - * for all buffers. - */ -static struct pipe_buffer_handle * -intel_create_buffer(struct softpipe_winsys *sws, - unsigned alignment) -{ - struct intel_context *intel = intel_softpipe_winsys(sws)->intel; - struct _DriBufferObject *buffer; - - LOCK_HARDWARE( intel ); - driGenBuffers( intel->intelScreen->regionPool, - "softpipe buffer", 1, &buffer, alignment, 0, 0 ); - UNLOCK_HARDWARE( intel ); - - return pipe_bo(buffer); -} - - -/** - * Return list of surface formats supported by this driver. - */ -static const GLuint * -intel_supported_formats(struct pipe_context *pipe, GLuint *numFormats) -{ - static const GLuint formats[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_S8_Z24, - }; - - *numFormats = sizeof(formats) / sizeof(formats[0]); - return formats; -} - - -static void intel_wait_idle( struct softpipe_winsys *sws ) -{ - struct intel_context *intel = intel_softpipe_winsys(sws)->intel; - - if (intel->batch->last_fence) { - driFenceFinish(intel->batch->last_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); - driFenceUnReference(intel->batch->last_fence); - intel->batch->last_fence = NULL; - } -} - - -/* The state tracker (should!) keep track of whether the fake - * frontbuffer has been touched by any rendering since the last time - * we copied its contents to the real frontbuffer. Our task is easy: - */ -static void -intel_flush_frontbuffer( struct softpipe_winsys *sws ) -{ - struct intel_context *intel = intel_softpipe_winsys(sws)->intel; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelCopyBuffer(dPriv, NULL); -} - - - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel ) -{ - struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); - struct pipe_context *pipe; - - /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. - * - * Softpipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate _DriBufferObject. - */ - isws->sws.create_buffer = intel_create_buffer; - isws->sws.buffer_map = intel_buffer_map; - isws->sws.buffer_unmap = intel_buffer_unmap; - isws->sws.buffer_reference = intel_buffer_reference; - isws->sws.buffer_unreference = intel_buffer_unreference; - isws->sws.buffer_data = intel_buffer_data; - isws->sws.buffer_subdata = intel_buffer_subdata; - isws->sws.buffer_get_subdata = intel_buffer_get_subdata; - isws->sws.flush_frontbuffer = intel_flush_frontbuffer; - isws->sws.wait_idle = intel_wait_idle; - isws->intel = intel; - - /* Create the softpipe context: - */ - pipe = softpipe_create( &isws->sws ); - - /* XXX: This should probably be a parameter to softpipe_create() - */ - pipe->supported_formats = intel_supported_formats; - - return pipe; -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h new file mode 100644 index 0000000000..6d4a5a4df3 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h @@ -0,0 +1,45 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_PIPE_H +#define INTEL_PIPE_H + +struct intel_context; +struct pipe_context; +struct pipe_winsys; + +struct pipe_winsys * +intel_create_pipe_winsys( struct intel_context *intel ); + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel ); + +struct pipe_context * +intel_create_i915simple( struct intel_context *intel ); + + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c new file mode 100644 index 0000000000..51e3055cb0 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -0,0 +1,144 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +#include "dri_bufpool.h" +#include "dri_bufmgr.h" + +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_winsys.h" +#include "intel_blit.h" + +#include "pipe/i915simple/i915_winsys.h" + + +struct intel_i915_winsys { + struct i915_winsys winsys; + struct intel_context *intel; +}; + + +/* Turn a i915simple winsys into an intel/i915simple winsys: + */ +static inline struct intel_i915_winsys * +intel_i915_winsys( struct i915_winsys *sws ) +{ + return (struct intel_i915_winsys *)sws; +} + + +/* Simple batchbuffer interface: + */ + +static unsigned *intel_i915_batch_start( struct i915_winsys *sws, + unsigned dwords, + unsigned relocs ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + /* XXX: check relocs. + */ + if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { + /* XXX: Hmm, the driver can't really do much with this pointer: + */ + return (unsigned *)intel->batch->ptr; + } + else + return NULL; +} + +static void intel_i915_batch_dword( struct i915_winsys *sws, + unsigned dword ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + intel_batchbuffer_emit_dword( intel->batch, dword ); +} + +static void intel_i915_batch_reloc( struct i915_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned access_flags, + unsigned delta ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + unsigned flags = DRM_BO_FLAG_MEM_TT; + unsigned mask = DRM_BO_MASK_MEM; + + if (access_flags & I915_BUFFER_ACCESS_WRITE) { + flags |= DRM_BO_FLAG_WRITE; + mask |= DRM_BO_FLAG_WRITE; + } + + if (access_flags & I915_BUFFER_ACCESS_READ) { + flags |= DRM_BO_FLAG_READ; + mask |= DRM_BO_FLAG_READ; + } + + intel_batchbuffer_emit_reloc( intel->batch, + dri_bo( buf ), + flags, mask, + delta ); +} + + + +static void intel_i915_batch_flush( struct i915_winsys *sws ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + intel_batchbuffer_flush( intel->batch ); + if (0) intel_i915_batch_wait_idle( sws ); +} + + + + +struct pipe_context * +intel_create_i915simple( struct intel_context *intel ) +{ + struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); + + /* Fill in this struct with callbacks that i915simple will need to + * communicate with the window system, buffer manager, etc. + */ + iws->winsys.batch_start = intel_i915_batch_start; + iws->winsys.batch_dword = intel_i915_batch_dword; + iws->winsys.batch_reloc = intel_i915_batch_reloc; + iws->winsys.batch_flush = intel_i915_batch_flush; + iws->intel = intel; + + /* Create the i915simple context: + */ + return i915_create( intel_create_pipe_winsys(intel), + &iws->winsys, + intel->intelScreen->deviceID ); +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c new file mode 100644 index 0000000000..22daac6ea4 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -0,0 +1,221 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +#include "dri_bufpool.h" +#include "dri_bufmgr.h" + +#include "intel_context.h" +#include "intel_winsys.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" + +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" + + + +struct intel_pipe_winsys { + struct pipe_winsys winsys; + struct intel_context *intel; +}; + + + +/* Turn the pipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static inline struct _DriBufferObject * +dri_bo( struct pipe_buffer_handle *bo ) +{ + return (struct _DriBufferObject *)bo; +} + +static inline struct pipe_buffer_handle * +pipe_bo( struct _DriBufferObject *bo ) +{ + return (struct pipe_buffer_handle *)bo; +} + +/* Turn a pipe winsys into an intel/pipe winsys: + */ +static inline struct intel_pipe_winsys * +intel_pipe_winsys( struct pipe_winsys *sws ) +{ + return (struct intel_pipe_winsys *)sws; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void *intel_buffer_map(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned flags ) +{ + unsigned drm_flags = 0; + + if (flags & PIPE_BUFFER_FLAG_WRITE) + drm_flags |= DRM_BO_FLAG_WRITE; + + if (flags & PIPE_BUFFER_FLAG_READ) + drm_flags |= DRM_BO_FLAG_READ; + + return driBOMap( dri_bo(buf), drm_flags, 0 ); +} + +static void intel_buffer_unmap(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf) +{ + driBOUnmap( dri_bo(buf) ); +} + + +static struct pipe_buffer_handle * +intel_buffer_reference(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf) +{ + return pipe_bo( driBOReference( dri_bo(buf) ) ); +} + +static void intel_buffer_unreference(struct pipe_winsys *sws, + struct pipe_buffer_handle **buf) +{ + if (*buf) { + driBOUnReference( dri_bo(*buf) ); + *buf = NULL; + } +} + +/* Grabs the hardware lock! + */ +static void intel_buffer_data(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct intel_context *intel = intel_pipe_winsys(sws)->intel; + + LOCK_HARDWARE( intel ); + driBOData( dri_bo(buf), size, data, 0 ); + UNLOCK_HARDWARE( intel ); +} + +static void intel_buffer_subdata(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) +{ + driBOSubData( dri_bo(buf), offset, size, data ); +} + +static void intel_buffer_get_subdata(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) +{ + driBOGetSubData( dri_bo(buf), offset, size, data ); +} + +/* Pipe has no concept of pools. We choose the tex/region pool + * for all buffers. + */ +static struct pipe_buffer_handle * +intel_buffer_create(struct pipe_winsys *sws, + unsigned alignment) +{ + struct intel_context *intel = intel_pipe_winsys(sws)->intel; + struct _DriBufferObject *buffer; + + LOCK_HARDWARE( intel ); + driGenBuffers( intel->intelScreen->regionPool, + "pipe buffer", 1, &buffer, alignment, 0, 0 ); + UNLOCK_HARDWARE( intel ); + + return pipe_bo(buffer); +} + + +static void intel_wait_idle( struct pipe_winsys *sws ) +{ + struct intel_context *intel = intel_pipe_winsys(sws)->intel; + + if (intel->batch->last_fence) { + driFenceFinish(intel->batch->last_fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); + driFenceUnReference(intel->batch->last_fence); + intel->batch->last_fence = NULL; + } +} + + +/* The state tracker (should!) keep track of whether the fake + * frontbuffer has been touched by any rendering since the last time + * we copied its contents to the real frontbuffer. Our task is easy: + */ +static void +intel_flush_frontbuffer( struct pipe_winsys *sws ) +{ + struct intel_context *intel = intel_pipe_winsys(sws)->intel; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelCopyBuffer(dPriv, NULL); +} + + + +struct pipe_winsys * +intel_create_pipe_winsys( struct intel_context *intel ) +{ + struct intel_pipe_winsys *iws = CALLOC_STRUCT( intel_pipe_winsys ); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + * + * Pipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + iws->winsys.buffer_create = intel_buffer_create; + iws->winsys.buffer_map = intel_buffer_map; + iws->winsys.buffer_unmap = intel_buffer_unmap; + iws->winsys.buffer_reference = intel_buffer_reference; + iws->winsys.buffer_unreference = intel_buffer_unreference; + iws->winsys.buffer_data = intel_buffer_data; + iws->winsys.buffer_subdata = intel_buffer_subdata; + iws->winsys.buffer_get_subdata = intel_buffer_get_subdata; + iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; + iws->winsys.wait_idle = intel_wait_idle; + iws->intel = intel; + + return &iws->winsys; +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c new file mode 100644 index 0000000000..917e933641 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c @@ -0,0 +1,79 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include "imports.h" +#include "intel_context.h" +#include "intel_winsys.h" +#include "pipe/softpipe/sp_winsys.h" +#include "pipe/p_defines.h" + + + +struct intel_softpipe_winsys { + struct softpipe_winsys sws; + struct intel_context *intel; +}; + +/** + * Return list of surface formats supported by this driver. + */ +static const unsigned * +intel_supported_formats(struct softpipe_winsys *sws, + unsigned *numFormats) +{ + static const GLuint formats[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_S8_Z24, + }; + + *numFormats = sizeof(formats) / sizeof(formats[0]); + return formats; +} + + + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel ) +{ + struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); + + /* Fill in this struct with callbacks that softpipe will need to + * communicate with the window system, buffer manager, etc. + */ + isws->sws.supported_formats = intel_supported_formats; + isws->intel = intel; + + /* Create the softpipe context: + */ + return softpipe_create( intel_create_pipe_winsys( intel ), + &isws->sws ); +} diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 5b919e3bed..28fe70d069 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -6,7 +6,6 @@ LIBNAME = i915simple DRIVER_SOURCES = \ i915_blit.c \ - i915_buffer.c \ i915_clear.c \ i915_flush.c \ i915_context.c \ diff --git a/src/mesa/pipe/i915simple/i915_buffer.c b/src/mesa/pipe/i915simple/i915_buffer.c deleted file mode 100644 index 680213182b..0000000000 --- a/src/mesa/pipe/i915simple/i915_buffer.c +++ /dev/null @@ -1,121 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include "i915_context.h" -#include "i915_winsys.h" - - - -/* Most callbacks map direcly onto winsys operations at the moment, - * but this may change, especially as state_trackers and winsys's - * evolve in separate directions... Don't try and remove this yet. - */ -static struct pipe_buffer_handle * -i915_buffer_create(struct pipe_context *pipe, - unsigned alignment, - unsigned flags) -{ - struct i915_context *i915 = i915_context( pipe ); - return i915->winsys->buffer_create( i915->winsys, alignment ); -} - -static void *i915_buffer_map(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned flags ) -{ - struct i915_context *i915 = i915_context( pipe ); - return i915->winsys->buffer_map( i915->winsys, buf ); -} - -static void i915_buffer_unmap(struct pipe_context *pipe, - struct pipe_buffer_handle *buf) -{ - struct i915_context *i915 = i915_context( pipe ); - i915->winsys->buffer_unmap( i915->winsys, buf ); -} - -static struct pipe_buffer_handle * -i915_buffer_reference(struct pipe_context *pipe, - struct pipe_buffer_handle *buf) -{ - struct i915_context *i915 = i915_context( pipe ); - return i915->winsys->buffer_reference( i915->winsys, buf ); -} - -static void i915_buffer_unreference(struct pipe_context *pipe, - struct pipe_buffer_handle **buf) -{ - struct i915_context *i915 = i915_context( pipe ); - i915->winsys->buffer_unreference( i915->winsys, buf ); -} - -static void i915_buffer_data(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ) -{ - struct i915_context *i915 = i915_context( pipe ); - i915->winsys->buffer_data( i915->winsys, buf, size, data ); -} - -static void i915_buffer_subdata(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) -{ - struct i915_context *i915 = i915_context( pipe ); - i915->winsys->buffer_subdata( i915->winsys, buf, offset, size, data ); -} - -static void i915_buffer_get_subdata(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) -{ - struct i915_context *i915 = i915_context( pipe ); - i915->winsys->buffer_get_subdata( i915->winsys, buf, offset, size, data ); -} - - -void -i915_init_buffer_functions( struct i915_context *i915 ) -{ - i915->pipe.create_buffer = i915_buffer_create; - i915->pipe.buffer_map = i915_buffer_map; - i915->pipe.buffer_unmap = i915_buffer_unmap; - i915->pipe.buffer_reference = i915_buffer_reference; - i915->pipe.buffer_unreference = i915_buffer_unreference; - i915->pipe.buffer_data = i915_buffer_data; - i915->pipe.buffer_subdata = i915_buffer_subdata; - i915->pipe.buffer_get_subdata = i915_buffer_get_subdata; -} diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index d8e54f02ee..ee6cfe2ed4 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -29,10 +29,12 @@ #include "i915_context.h" #include "i915_winsys.h" #include "i915_state.h" +#include "i915_batch.h" #include "i915_tex_layout.h" #include "pipe/draw/draw_context.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #define PCI_CHIP_I915_G 0x2582 #define PCI_CHIP_I915_GM 0x2592 @@ -152,7 +154,8 @@ i915_draw_vertices(struct pipe_context *pipe, -struct pipe_context *i915_create( struct i915_winsys *winsys, +struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, + struct i915_winsys *i915_winsys, unsigned pci_id ) { struct i915_context *i915; @@ -175,8 +178,9 @@ struct pipe_context *i915_create( struct i915_winsys *winsys, break; default: - winsys->printf(winsys, "%s: unknown pci id 0x%x, cannot create context\n", - __FUNCTION__, pci_id); + pipe_winsys->printf(pipe_winsys, + "%s: unknown pci id 0x%x, cannot create context\n", + __FUNCTION__, pci_id); return NULL; } @@ -184,6 +188,9 @@ struct pipe_context *i915_create( struct i915_winsys *winsys, if (i915 == NULL) return NULL; + i915->winsys = i915_winsys; + i915->pipe.winsys = pipe_winsys; + i915->pipe.destroy = i915_destroy; i915->pipe.supported_formats = i915_supported_formats; i915->pipe.draw_vb = i915_draw_vb; @@ -192,7 +199,6 @@ struct pipe_context *i915_create( struct i915_winsys *winsys, i915->pipe.reset_occlusion_counter = NULL; /* no support */ i915->pipe.get_occlusion_counter = NULL; - i915->winsys = winsys; /* * Create drawing context and plug our rendering stage into it. @@ -201,7 +207,6 @@ struct pipe_context *i915_create( struct i915_winsys *winsys, assert(i915->draw); draw_set_setup_stage(i915->draw, i915_draw_render_stage(i915)); - i915_init_buffer_functions(i915); i915_init_region_functions(i915); i915_init_surface_functions(i915); i915_init_state_functions(i915); @@ -219,7 +224,7 @@ struct pipe_context *i915_create( struct i915_winsys *winsys, /* Batch stream debugging is a bit hacked up at the moment: */ - i915->batch_start = winsys->batch_start( winsys, 0, 0 ); + i915->batch_start = BEGIN_BATCH(0, 0); /* * XXX we could plug GL selection/feedback into the drawing pipeline diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index e8db2b7c36..7a73a8d8d0 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -182,9 +182,8 @@ void i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, /*********************************************************************** - * i915_buffer.c: + * i915_region.c: */ -void i915_init_buffer_functions( struct i915_context *i915 ); void i915_init_region_functions( struct i915_context *i915 ); void i915_init_surface_functions( struct i915_context *i915 ); void i915_init_state_functions( struct i915_context *i915 ); diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index 0ea6f03e49..8050eb0bf5 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -423,7 +423,7 @@ i915_dump_batchbuffer( struct i915_context *i915, stream.offset = 0; stream.ptr = (char *)start; stream.print_addresses = 0; - stream.winsys = i915->winsys; + stream.winsys = i915->pipe.winsys; while (!done && stream.offset < bytes && diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h index f0f72780d3..0ea131171e 100644 --- a/src/mesa/pipe/i915simple/i915_debug.h +++ b/src/mesa/pipe/i915simple/i915_debug.h @@ -39,7 +39,7 @@ struct debug_stream char *ptr; /* pointer to gtt offset zero */ char *end; /* pointer to gtt offset zero */ unsigned print_addresses; - struct i915_winsys *winsys; + struct pipe_winsys *winsys; }; @@ -68,9 +68,11 @@ void i915_print_ureg(const char *msg, unsigned ureg); #define DEBUG_WINSYS 0x4000 #ifdef DEBUG -#include "i915_winsys.h" -#define DBG( i915, ... ) \ - if ((i915)->debug & FILE_DEBUG_FLAG) (i915)->winsys->printf( (i915)->winsys, __VA_ARGS__ ) +#include "pipe/p_winsys.h" +#define DBG( i915, ... ) do { \ + if ((i915)->debug & FILE_DEBUG_FLAG) \ + (i915)->pipe.winsys->printf( (i915)->pipe.winsys, __VA_ARGS__ ); \ +} while(0) #else #define DBG( i915, ... ) \ (void)i915 diff --git a/src/mesa/pipe/i915simple/i915_debug_fp.c b/src/mesa/pipe/i915simple/i915_debug_fp.c index 95476d3c55..d99c609d48 100644 --- a/src/mesa/pipe/i915simple/i915_debug_fp.c +++ b/src/mesa/pipe/i915simple/i915_debug_fp.c @@ -29,7 +29,7 @@ #include "i915_reg.h" #include "i915_debug.h" -#include "i915_winsys.h" +#include "pipe/p_winsys.h" //#include "i915_fpc.h" #include "shader/program.h" #include "shader/prog_instruction.h" diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c index 8af4ce770c..9a31342cbd 100644 --- a/src/mesa/pipe/i915simple/i915_flush.c +++ b/src/mesa/pipe/i915simple/i915_flush.c @@ -66,16 +66,9 @@ static void i915_flush( struct pipe_context *pipe, FLUSH_BATCH(); } -static void i915_wait_idle(struct pipe_context *pipe) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->winsys->batch_wait_idle( i915->winsys ); -} void i915_init_flush_functions( struct i915_context *i915 ) { i915->pipe.flush = i915_flush; - i915->pipe.wait_idle = i915_wait_idle; } diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index 886b089506..9b7cddb58d 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -32,8 +32,8 @@ */ #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #include "i915_context.h" -#include "i915_winsys.h" #include "i915_blit.h" @@ -50,8 +50,10 @@ i915_region_map(struct pipe_context *pipe, struct pipe_region *region) struct i915_context *i915 = i915_context( pipe ); if (!region->map_refcount++) { - region->map = i915->winsys->buffer_map( i915->winsys, - region->buffer ); + region->map = i915->pipe.winsys->buffer_map( i915->pipe.winsys, + region->buffer, + PIPE_BUFFER_FLAG_WRITE | + PIPE_BUFFER_FLAG_READ); } return region->map; @@ -63,8 +65,8 @@ i915_region_unmap(struct pipe_context *pipe, struct pipe_region *region) struct i915_context *i915 = i915_context( pipe ); if (!--region->map_refcount) { - i915->winsys->buffer_unmap( i915->winsys, - region->buffer ); + i915->pipe.winsys->buffer_unmap( i915->pipe.winsys, + region->buffer ); region->map = NULL; } } @@ -97,12 +99,12 @@ i915_region_alloc(struct pipe_context *pipe, region->height = height; /* needed? */ region->refcount = 1; - region->buffer = i915->winsys->buffer_create( i915->winsys, alignment ); + region->buffer = i915->pipe.winsys->buffer_create( i915->pipe.winsys, alignment ); - i915->winsys->buffer_data( i915->winsys, - region->buffer, - pitch * cpp * height, - NULL ); + i915->pipe.winsys->buffer_data( i915->pipe.winsys, + region->buffer, + pitch * cpp * height, + NULL ); return region; } @@ -121,8 +123,8 @@ i915_region_release(struct pipe_context *pipe, struct pipe_region **region) if ((*region)->refcount == 0) { assert((*region)->map_refcount == 0); - i915->winsys->buffer_unreference( i915->winsys, - (*region)->buffer ); + i915->pipe.winsys->buffer_unreference( i915->pipe.winsys, + (*region)->buffer ); free(*region); } *region = NULL; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index ab00cbc822..22a5bf68b4 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -148,7 +148,7 @@ static void i915_set_texture_state(struct pipe_context *pipe, static void i915_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb) + const struct pipe_framebuffer_state *fb) { struct i915_context *i915 = i915_context(pipe); diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index a3dadbfd3d..803ee9073c 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -47,57 +47,10 @@ */ struct pipe_buffer_handle; +struct pipe_winsys; struct i915_winsys { - /* Do any special operations to ensure frontbuffer contents are - * displayed, eg copy fake frontbuffer. - */ - void (*flush_frontbuffer)( struct i915_winsys *sws ); - - - /* debug output - */ - void (*printf)( struct i915_winsys *sws, - const char *, ... ); - - /* Many of the winsys's are probably going to have a similar - * buffer-manager interface, as something almost identical is - * currently exposed in the pipe interface. Probably want to avoid - * endless repetition of this code somehow. - */ - struct pipe_buffer_handle *(*buffer_create)(struct i915_winsys *sws, - unsigned alignment ); - - void *(*buffer_map)( struct i915_winsys *sws, - struct pipe_buffer_handle *buf ); - - void (*buffer_unmap)( struct i915_winsys *sws, - struct pipe_buffer_handle *buf ); - - struct pipe_buffer_handle *(*buffer_reference)( struct i915_winsys *sws, - struct pipe_buffer_handle *buf ); - - void (*buffer_unreference)( struct i915_winsys *sws, - struct pipe_buffer_handle **buf ); - - void (*buffer_data)(struct i915_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ); - - void (*buffer_subdata)(struct i915_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data); - - void (*buffer_get_subdata)(struct i915_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data); - - /* An over-simple batchbuffer mechanism. Will want to improve the * performance of this, perhaps based on the cmdstream stuff. It * would be pretty impossible to implement swz on top of this @@ -116,20 +69,15 @@ struct i915_winsys { unsigned access_flags, unsigned delta ); void (*batch_flush)( struct i915_winsys *sws ); - void (*batch_wait_idle)( struct i915_winsys *sws ); - - /* Printf??? - */ - void (*dpf)( const char *fmt, ... ); - }; #define I915_BUFFER_ACCESS_WRITE 0x1 #define I915_BUFFER_ACCESS_READ 0x2 -struct pipe_context *i915_create( struct i915_winsys *, +struct pipe_context *i915_create( struct pipe_winsys *, + struct i915_winsys *, unsigned pci_id ); diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 533840c555..89d9b36af7 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -42,7 +42,8 @@ struct vertex_buffer; * state setting functions, plus VBO drawing entrypoint. */ struct pipe_context { - + struct pipe_winsys *winsys; + void (*destroy)( struct pipe_context * ); /* @@ -175,44 +176,6 @@ struct pipe_context { GLuint value); - /* Buffer management functions need to be exposed as well. A pipe - * buffer may be used as a texture, render target or vertex/index - * buffer, or some combination according to flags. - */ - - struct pipe_buffer_handle *(*create_buffer)(struct pipe_context *pipe, - unsigned alignment, - unsigned flags ); - - void *(*buffer_map)( struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned flags ); - - void (*buffer_unmap)( struct pipe_context *pipe, - struct pipe_buffer_handle *buf ); - - struct pipe_buffer_handle *(*buffer_reference)( struct pipe_context *pipe, - struct pipe_buffer_handle *buf ); - - void (*buffer_unreference)( struct pipe_context *pipe, - struct pipe_buffer_handle **buf ); - - void (*buffer_data)(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ); - - void (*buffer_subdata)(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data); - - void (*buffer_get_subdata)(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data); - /* * Texture functions */ @@ -220,14 +183,10 @@ struct pipe_context { struct pipe_mipmap_tree *mt ); - /* Simple flush/finish support: + /* Flush rendering: */ void (*flush)( struct pipe_context *pipe, unsigned flags ); - - void (*wait_idle)( struct pipe_context *pipe ); - - void (*flush_frontbuffer)( struct pipe_context *pipe ); }; diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h new file mode 100644 index 0000000000..c455ebdbbe --- /dev/null +++ b/src/mesa/pipe/p_winsys.h @@ -0,0 +1,114 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_WINSYS_H +#define P_WINSYS_H + +#include "main/mtypes.h" + +/* This is the interface that softpipe requires any window system + * hosting it to implement. This is the only include file in softpipe + * which is public. + */ + + +/* Pipe drivers are (meant to be!) independent of both GL and the + * window system. The window system provides a buffer manager and a + * set of additional hooks for things like command buffer submission, + * etc. + * + * There clearly has to be some agreement between the window system + * driver and the hardware driver about the format of command buffers, + * etc. + */ + +struct pipe_buffer_handle; + +struct pipe_winsys { + + /* Do any special operations to ensure frontbuffer contents are + * displayed, eg copy fake frontbuffer. + */ + void (*flush_frontbuffer)( struct pipe_winsys *sws ); + + /* debug output + */ + void (*printf)( struct pipe_winsys *sws, + const char *, ... ); + + + /* The buffer manager is modeled after the dri_bugmgr interface, + * but this is the subset that softpipe cares about. Remember that + * softpipe gets to choose the interface it needs, and the window + * systems must then implement that interface (rather than the + * other way around...). + * + * Softpipe only really wants to make system memory allocations, + * right?? + */ + struct pipe_buffer_handle *(*buffer_create)(struct pipe_winsys *sws, + unsigned alignment ); + + void *(*buffer_map)( struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned flags ); + + void (*buffer_unmap)( struct pipe_winsys *sws, + struct pipe_buffer_handle *buf ); + + struct pipe_buffer_handle *(*buffer_reference)( struct pipe_winsys *sws, + struct pipe_buffer_handle *buf ); + + void (*buffer_unreference)( struct pipe_winsys *sws, + struct pipe_buffer_handle **buf ); + + void (*buffer_data)(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned size, const void *data ); + + void (*buffer_subdata)(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data); + + void (*buffer_get_subdata)(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data); + + + /* Wait for any hw swapbuffers, etc. to finish: + */ + void (*wait_idle)( struct pipe_winsys *sws ); + +}; + + + +#endif /* SP_WINSYS_H */ diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index ac83da451d..615c612e9c 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -5,7 +5,6 @@ include $(TOP)/configs/current LIBNAME = softpipe DRIVER_SOURCES = \ - sp_buffer.c \ sp_clear.c \ sp_flush.c \ sp_context.c \ diff --git a/src/mesa/pipe/softpipe/sp_buffer.c b/src/mesa/pipe/softpipe/sp_buffer.c deleted file mode 100644 index 27443421a3..0000000000 --- a/src/mesa/pipe/softpipe/sp_buffer.c +++ /dev/null @@ -1,120 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include "sp_context.h" -#include "sp_winsys.h" -#include "sp_buffer.h" - - - -/* Most callbacks map direcly onto winsys operations: - */ -static struct pipe_buffer_handle * -sp_create_buffer(struct pipe_context *pipe, - unsigned alignment, - unsigned flags) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - return sp->winsys->create_buffer( sp->winsys, alignment ); -} - -static void *sp_buffer_map(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned flags ) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - return sp->winsys->buffer_map( sp->winsys, buf ); -} - -static void sp_buffer_unmap(struct pipe_context *pipe, - struct pipe_buffer_handle *buf) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - sp->winsys->buffer_unmap( sp->winsys, buf ); -} - -static struct pipe_buffer_handle * -sp_buffer_reference(struct pipe_context *pipe, - struct pipe_buffer_handle *buf) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - return sp->winsys->buffer_reference( sp->winsys, buf ); -} - -static void sp_buffer_unreference(struct pipe_context *pipe, - struct pipe_buffer_handle **buf) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - sp->winsys->buffer_unreference( sp->winsys, buf ); -} - -static void sp_buffer_data(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - sp->winsys->buffer_data( sp->winsys, buf, size, data ); -} - -static void sp_buffer_subdata(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - sp->winsys->buffer_subdata( sp->winsys, buf, offset, size, data ); -} - -static void sp_buffer_get_subdata(struct pipe_context *pipe, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - sp->winsys->buffer_get_subdata( sp->winsys, buf, offset, size, data ); -} - - -void -sp_init_buffer_functions( struct softpipe_context *sp ) -{ - sp->pipe.create_buffer = sp_create_buffer; - sp->pipe.buffer_map = sp_buffer_map; - sp->pipe.buffer_unmap = sp_buffer_unmap; - sp->pipe.buffer_reference = sp_buffer_reference; - sp->pipe.buffer_unreference = sp_buffer_unreference; - sp->pipe.buffer_data = sp_buffer_data; - sp->pipe.buffer_subdata = sp_buffer_subdata; - sp->pipe.buffer_get_subdata = sp_buffer_get_subdata; -} diff --git a/src/mesa/pipe/softpipe/sp_buffer.h b/src/mesa/pipe/softpipe/sp_buffer.h deleted file mode 100644 index 9805c5142a..0000000000 --- a/src/mesa/pipe/softpipe/sp_buffer.h +++ /dev/null @@ -1,40 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef SP_BUFFER_H -#define SP_BUFFER_H - - -struct softpipe_context; - - -extern void -sp_init_buffer_functions(struct softpipe_context *sp); - - -#endif /* SP_BUFFER_H */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index db572f169d..91b8ae5086 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -33,7 +33,6 @@ #include "main/macros.h" #include "pipe/draw/draw_context.h" #include "pipe/p_defines.h" -#include "sp_buffer.h" #include "sp_clear.h" #include "sp_context.h" #include "sp_flush.h" @@ -194,10 +193,12 @@ static GLuint softpipe_get_occlusion_counter(struct pipe_context *pipe) } -struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) +struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, + struct softpipe_winsys *softpipe_winsys ) { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); + softpipe->pipe.winsys = pipe_winsys; softpipe->pipe.destroy = softpipe_destroy; softpipe->pipe.supported_formats = softpipe_supported_formats; @@ -221,7 +222,6 @@ struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) softpipe->pipe.draw_vertices = softpipe_draw_vertices; softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; - softpipe->pipe.wait_idle = softpipe_wait_idle; softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; @@ -240,7 +240,7 @@ struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); - softpipe->winsys = sws; + softpipe->winsys = softpipe_winsys; /* * Create drawing context and plug our rendering stage into it. @@ -249,7 +249,6 @@ struct pipe_context *softpipe_create( struct softpipe_winsys *sws ) assert(softpipe->draw); draw_set_setup_stage(softpipe->draw, sp_draw_render_stage(softpipe)); - sp_init_buffer_functions(softpipe); sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index a0bce200ed..cdf4a53c83 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -49,22 +49,3 @@ softpipe_flush( struct pipe_context *pipe, */ } -void -softpipe_wait_idle(struct pipe_context *pipe) -{ - /* Nothing to do. - * XXX: What about swapbuffers. - * XXX: Even more so - what about fake frontbuffer copies?? - */ - struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->winsys->wait_idle( softpipe->winsys ); -} - - -void -softpipe_flush_frontbuffer( struct pipe_context *pipe ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->winsys->flush_frontbuffer( softpipe->winsys ); -} diff --git a/src/mesa/pipe/softpipe/sp_flush.h b/src/mesa/pipe/softpipe/sp_flush.h index 03c0010623..34ec617866 100644 --- a/src/mesa/pipe/softpipe/sp_flush.h +++ b/src/mesa/pipe/softpipe/sp_flush.h @@ -31,6 +31,5 @@ struct pipe_context; void softpipe_flush(struct pipe_context *pipe, unsigned flags ); -void softpipe_wait_idle(struct pipe_context *pipe); #endif diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 142d121047..115f3ab826 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -32,8 +32,9 @@ */ #include "sp_context.h" -#include "sp_winsys.h" #include "sp_region.h" +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" /** @@ -59,8 +60,10 @@ sp_region_map(struct pipe_context *pipe, struct pipe_region *region) struct softpipe_context *sp = softpipe_context( pipe ); if (!region->map_refcount++) { - region->map = sp->winsys->buffer_map( sp->winsys, - region->buffer ); + region->map = sp->pipe.winsys->buffer_map( sp->pipe.winsys, + region->buffer, + PIPE_BUFFER_FLAG_WRITE | + PIPE_BUFFER_FLAG_READ); } return region->map; @@ -72,7 +75,7 @@ sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) struct softpipe_context *sp = softpipe_context( pipe ); if (!--region->map_refcount) { - sp->winsys->buffer_unmap( sp->winsys, + sp->pipe.winsys->buffer_unmap( sp->pipe.winsys, region->buffer ); region->map = NULL; } @@ -91,13 +94,13 @@ sp_region_alloc(struct pipe_context *pipe, region->height = height; region->refcount = 1; - region->buffer = sp->winsys->create_buffer( sp->winsys, alignment ); + region->buffer = sp->pipe.winsys->buffer_create( sp->pipe.winsys, alignment ); /* NULL data --> just allocate the space */ - sp->winsys->buffer_data( sp->winsys, - region->buffer, - region->pitch * cpp * height, - NULL ); + sp->pipe.winsys->buffer_data( sp->pipe.winsys, + region->buffer, + region->pitch * cpp * height, + NULL ); return region; } @@ -115,7 +118,7 @@ sp_region_release(struct pipe_context *pipe, struct pipe_region **region) if ((*region)->refcount == 0) { assert((*region)->map_refcount == 0); - sp->winsys->buffer_unreference( sp->winsys, + sp->pipe.winsys->buffer_unreference( sp->pipe.winsys, (*region)->buffer ); free(*region); } diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h index 73b0659067..726e4c8bb6 100644 --- a/src/mesa/pipe/softpipe/sp_winsys.h +++ b/src/mesa/pipe/softpipe/sp_winsys.h @@ -28,86 +28,24 @@ #ifndef SP_WINSYS_H #define SP_WINSYS_H -#include "main/mtypes.h" /* This is the interface that softpipe requires any window system * hosting it to implement. This is the only include file in softpipe * which is public. */ - -/* Pipe drivers are (meant to be!) independent of both GL and the - * window system. The window system provides a buffer manager and a - * set of additional hooks for things like command buffer submission, - * etc. - * - * There clearly has to be some agreement between the window system - * driver and the hardware driver about the format of command buffers, - * etc. - */ - -struct pipe_buffer_handle; - struct softpipe_winsys { + const unsigned *(*supported_formats)(struct softpipe_winsys *sws, + unsigned *numFormats); - /* Do any special operations to ensure frontbuffer contents are - * displayed, eg copy fake frontbuffer. - */ - void (*flush_frontbuffer)( struct softpipe_winsys *sws ); - - /* Wait for any hw swapbuffers, etc. to finish: - */ - void (*wait_idle)( struct softpipe_winsys *sws ); - - /* debug output - */ - void (*printf)( struct softpipe_winsys *sws, - const char *, ... ); - - - /* The buffer manager is modeled after the dri_bugmgr interface, - * but this is the subset that softpipe cares about. Remember that - * softpipe gets to choose the interface it needs, and the window - * systems must then implement that interface (rather than the - * other way around...). - * - * Softpipe only really wants to make system memory allocations, - * right?? - */ - struct pipe_buffer_handle *(*create_buffer)(struct softpipe_winsys *sws, - unsigned alignment ); - - void *(*buffer_map)( struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf ); - - void (*buffer_unmap)( struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf ); - - struct pipe_buffer_handle *(*buffer_reference)( struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf ); - - void (*buffer_unreference)( struct softpipe_winsys *sws, - struct pipe_buffer_handle **buf ); - - void (*buffer_data)(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ); - - void (*buffer_subdata)(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data); - - void (*buffer_get_subdata)(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data); }; +struct pipe_winsys; +struct pipe_context; + -struct pipe_context *softpipe_create( struct softpipe_winsys * ); +struct pipe_context *softpipe_create( struct pipe_winsys *, + struct softpipe_winsys * ); #endif /* SP_WINSYS_H */ diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index 9e407de34f..f1dc4fbaf4 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -35,6 +35,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" @@ -62,7 +63,7 @@ st_bufferobj_alloc(GLcontext *ctx, GLuint name, GLenum target) _mesa_initialize_buffer_object(&st_obj->Base, name, target); - st_obj->buffer = st->pipe->create_buffer( st->pipe, 32, 0 ); + st_obj->buffer = st->pipe->winsys->buffer_create( st->pipe->winsys, 32 ); return &st_obj->Base; } @@ -80,7 +81,7 @@ st_bufferobj_free(GLcontext *ctx, struct gl_buffer_object *obj) struct st_buffer_object *st_obj = st_buffer_object(obj); if (st_obj->buffer) - pipe->buffer_unreference(pipe, &st_obj->buffer); + pipe->winsys->buffer_unreference(pipe->winsys, &st_obj->buffer); free(st_obj); } @@ -107,7 +108,7 @@ st_bufferobj_data(GLcontext *ctx, st_obj->Base.Size = size; st_obj->Base.Usage = usage; - pipe->buffer_data( pipe, st_obj->buffer, size, data ); + pipe->winsys->buffer_data( pipe->winsys, st_obj->buffer, size, data ); } @@ -127,7 +128,7 @@ st_bufferobj_subdata(GLcontext *ctx, struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); - pipe->buffer_subdata(pipe, st_obj->buffer, offset, size, data); + pipe->winsys->buffer_subdata(pipe->winsys, st_obj->buffer, offset, size, data); } @@ -144,7 +145,7 @@ st_bufferobj_get_subdata(GLcontext *ctx, struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); - pipe->buffer_get_subdata(pipe, st_obj->buffer, offset, size, data); + pipe->winsys->buffer_get_subdata(pipe->winsys, st_obj->buffer, offset, size, data); } @@ -173,7 +174,7 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access, break; } - obj->Pointer = pipe->buffer_map(pipe, st_obj->buffer, flags); + obj->Pointer = pipe->winsys->buffer_map(pipe->winsys, st_obj->buffer, flags); return obj->Pointer; } @@ -187,7 +188,7 @@ st_bufferobj_unmap(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj) struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); - pipe->buffer_unmap(pipe, st_obj->buffer); + pipe->winsys->buffer_unmap(pipe->winsys, st_obj->buffer); obj->Pointer = NULL; return GL_TRUE; } diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index a6a8f8d90f..41f21c361c 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -37,6 +37,7 @@ #include "st_cb_flush.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" static void st_flush(GLcontext *ctx) @@ -60,7 +61,7 @@ static void st_flush(GLcontext *ctx) if (st->flags.frontbuffer_dirty) { /* Hook for copying "fake" frontbuffer if necessary: */ - st->pipe->flush_frontbuffer( st->pipe ); + st->pipe->winsys->flush_frontbuffer( st->pipe->winsys ); st->flags.frontbuffer_dirty = 0; } } @@ -70,7 +71,7 @@ static void st_finish(GLcontext *ctx) struct st_context *st = ctx->st; st_flush( ctx ); - st->pipe->wait_idle( st->pipe ); + st->pipe->winsys->wait_idle( st->pipe->winsys ); } diff --git a/src/mesa/state_tracker/st_cb_flush.h b/src/mesa/state_tracker/st_cb_flush.h index 29ceab3a56..c26f779225 100644 --- a/src/mesa/state_tracker/st_cb_flush.h +++ b/src/mesa/state_tracker/st_cb_flush.h @@ -26,13 +26,13 @@ **************************************************************************/ -#ifndef ST_CB_CLEAR_H -#define ST_CB_CLEAR_H +#ifndef ST_CB_FLUSH_H +#define ST_CB_FLUSH_H extern void st_init_flush_functions(struct dd_function_table *functions); -#endif /* ST_CB_CLEAR_H */ +#endif /* ST_CB_FLUSH_H */ -- cgit v1.2.3 From 8fbd81b4ed59e371aa616b87296e4263d8992bff Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 15:47:45 -0600 Subject: added pipe->max_texture_size() query, use it in st_drawpixels() --- src/mesa/pipe/i915simple/i915_context.c | 30 +++++++++++++++++++++++++++++ src/mesa/pipe/p_context.h | 5 +++++ src/mesa/pipe/softpipe/sp_context.c | 32 +++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_drawpixels.c | 9 +++++++++ 4 files changed, 76 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index ee6cfe2ed4..d9fbada1ff 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -115,6 +115,35 @@ i915_supported_formats(struct pipe_context *pipe, } +/** + * We might want to return max texture levels instead... + */ +static void +i915_max_texture_size(struct pipe_context *pipe, GLuint textureType, + GLuint *maxWidth, GLuint *maxHeight, GLuint *maxDepth) +{ + switch (textureType) { + case PIPE_TEXTURE_1D: + *maxWidth = 2048; + break; + case PIPE_TEXTURE_2D: + *maxWidth = + *maxHeight = 2048; + break; + case PIPE_TEXTURE_3D: + *maxWidth = + *maxHeight = + *maxDepth = 256; + break; + case PIPE_TEXTURE_CUBE: + *maxWidth = + *maxHeight = 2048; + break; + default: + assert(0); + } +} + static void i915_destroy( struct pipe_context *pipe ) { @@ -193,6 +222,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.destroy = i915_destroy; i915->pipe.supported_formats = i915_supported_formats; + i915->pipe.max_texture_size = i915_max_texture_size; i915->pipe.draw_vb = i915_draw_vb; i915->pipe.draw_vertices = i915_draw_vertices; i915->pipe.clear = i915_clear; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 89d9b36af7..b303cee5cc 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -51,6 +51,11 @@ struct pipe_context { */ const GLuint *(*supported_formats)(struct pipe_context *pipe, GLuint *numFormats); + void (*max_texture_size)(struct pipe_context *pipe, + GLuint textureType, /* PIPE_TEXTURE_x */ + GLuint *maxWidth, + GLuint *maxHeight, + GLuint *maxDepth); /* * Drawing diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 91b8ae5086..53990637ba 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -76,6 +76,33 @@ softpipe_supported_formats(struct pipe_context *pipe, GLuint *numFormats) } +static void +softpipe_max_texture_size(struct pipe_context *pipe, GLuint textureType, + GLuint *maxWidth, GLuint *maxHeight, + GLuint *maxDepth) +{ + switch (textureType) { + case PIPE_TEXTURE_1D: + *maxWidth = 1 << (MAX_TEXTURE_LEVELS - 1); + break; + case PIPE_TEXTURE_2D: + *maxWidth = + *maxHeight = 1 << (MAX_TEXTURE_LEVELS - 1); + break; + case PIPE_TEXTURE_3D: + *maxWidth = + *maxHeight = + *maxDepth = 1 << (MAX_3D_TEXTURE_LEVELS - 1); + break; + case PIPE_TEXTURE_CUBE: + *maxWidth = + *maxHeight = MAX_TEXTURE_RECT_SIZE; + break; + default: + assert(0); + } +} + static void map_surfaces(struct softpipe_context *sp) { @@ -201,8 +228,11 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.winsys = pipe_winsys; softpipe->pipe.destroy = softpipe_destroy; + /* queries */ softpipe->pipe.supported_formats = softpipe_supported_formats; + softpipe->pipe.max_texture_size = softpipe_max_texture_size; + /* state setters */ softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; softpipe->pipe.set_blend_state = softpipe_set_blend_state; @@ -225,9 +255,11 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; + /* textures */ softpipe->pipe.mipmap_tree_layout = softpipe_mipmap_tree_layout; softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; + /* setup quad rendering stages */ softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); softpipe->quad.shade = sp_quad_shade_stage(softpipe); softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index a5ba614429..ce417b9b7e 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -250,6 +250,15 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, struct pipe_context *pipe = ctx->st->pipe; struct pipe_mipmap_tree *mt; GLfloat x0, y0, x1, y1; + GLuint maxWidth, maxHeight; + + /* limit checks */ + /* XXX if DrawPixels image is larger than max texture size, break + * it up into chunks. + */ + pipe->max_texture_size(pipe, PIPE_TEXTURE_2D, &maxWidth, &maxHeight, NULL); + assert(width <= maxWidth); + assert(height <= maxHeight); /* setup state: just scissor */ { -- cgit v1.2.3 From 03a9994280c14e18be4bae14f4a9b667f8444d67 Mon Sep 17 00:00:00 2001 From: keithw Date: Fri, 10 Aug 2007 16:03:38 +0100 Subject: move dri_bo() to a header file --- src/mesa/drivers/dri/intel_winsys/intel_winsys.h | 20 ++++++++++++++++++++ .../drivers/dri/intel_winsys/intel_winsys_i915.c | 2 +- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 15 --------------- 3 files changed, 21 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h index 6d4a5a4df3..2c8aa20c2a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h @@ -31,6 +31,8 @@ struct intel_context; struct pipe_context; struct pipe_winsys; +struct pipe_buffer_handle; +struct _DriBufferObject; struct pipe_winsys * intel_create_pipe_winsys( struct intel_context *intel ); @@ -42,4 +44,22 @@ struct pipe_context * intel_create_i915simple( struct intel_context *intel ); + +/* Turn the pipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static INLINE struct _DriBufferObject * +dri_bo( struct pipe_buffer_handle *bo ) +{ + return (struct _DriBufferObject *)bo; +} + +static INLINE struct pipe_buffer_handle * +pipe_bo( struct _DriBufferObject *bo ) +{ + return (struct pipe_buffer_handle *)bo; +} + + + #endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index 51e3055cb0..56a27a9c31 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -116,7 +116,7 @@ static void intel_i915_batch_flush( struct i915_winsys *sws ) struct intel_context *intel = intel_i915_winsys(sws)->intel; intel_batchbuffer_flush( intel->batch ); - if (0) intel_i915_batch_wait_idle( sws ); +// if (0) intel_i915_batch_wait_idle( sws ); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 22daac6ea4..7f47b93f09 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -51,21 +51,6 @@ struct intel_pipe_winsys { -/* Turn the pipe opaque buffer pointer into a dri_bufmgr opaque - * buffer pointer... - */ -static inline struct _DriBufferObject * -dri_bo( struct pipe_buffer_handle *bo ) -{ - return (struct _DriBufferObject *)bo; -} - -static inline struct pipe_buffer_handle * -pipe_bo( struct _DriBufferObject *bo ) -{ - return (struct pipe_buffer_handle *)bo; -} - /* Turn a pipe winsys into an intel/pipe winsys: */ static inline struct intel_pipe_winsys * -- cgit v1.2.3 From 5b301132364dd78f5477cc2028bbde36f524fcf1 Mon Sep 17 00:00:00 2001 From: keithw Date: Fri, 10 Aug 2007 16:08:45 +0100 Subject: add missing printf --- src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 7f47b93f09..015d8fc2bb 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -176,6 +176,14 @@ intel_flush_frontbuffer( struct pipe_winsys *sws ) intelCopyBuffer(dPriv, NULL); } +static void +intel_printf( struct pipe_winsys *sws, const char *fmtString, ... ) +{ + va_list args; + va_start( args, fmtString ); + vfprintf(stderr, fmtString, args); + va_end( args ); +} struct pipe_winsys * @@ -200,6 +208,7 @@ intel_create_pipe_winsys( struct intel_context *intel ) iws->winsys.buffer_get_subdata = intel_buffer_get_subdata; iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; iws->winsys.wait_idle = intel_wait_idle; + iws->winsys.printf = intel_printf; iws->intel = intel; return &iws->winsys; -- cgit v1.2.3 From fb4ff8a2a8c0adbe1925d42eaaedfa52707c7ffb Mon Sep 17 00:00:00 2001 From: keithw Date: Fri, 10 Aug 2007 16:12:09 +0100 Subject: use winsys supported_formats() query --- src/mesa/pipe/softpipe/sp_context.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 53990637ba..b35bd66081 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -53,6 +53,7 @@ static const GLuint * softpipe_supported_formats(struct pipe_context *pipe, GLuint *numFormats) { +#if 0 static const GLuint supported[] = { PIPE_FORMAT_U_R8_G8_B8_A8, PIPE_FORMAT_U_A8_R8_G8_B8, @@ -73,6 +74,10 @@ softpipe_supported_formats(struct pipe_context *pipe, GLuint *numFormats) *numFormats = sizeof(supported)/sizeof(supported[0]); return supported; +#else + struct softpipe_context *softpipe = softpipe_context( pipe ); + return softpipe->winsys->supported_formats( softpipe->winsys, numFormats ); +#endif } -- cgit v1.2.3 From 717831ea712c07e18f2c2e797c49754bf2500e5e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 16:24:55 -0600 Subject: invert Y for scissor --- src/mesa/state_tracker/st_atom_scissor.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_scissor.c b/src/mesa/state_tracker/st_atom_scissor.c index 05a9f3eed1..2bf633828e 100644 --- a/src/mesa/state_tracker/st_atom_scissor.c +++ b/src/mesa/state_tracker/st_atom_scissor.c @@ -44,6 +44,7 @@ update_scissor( struct st_context *st ) { struct pipe_scissor_state scissor; const struct gl_framebuffer *fb = st->ctx->DrawBuffer; + GLint miny, maxy; scissor.minx = 0; scissor.miny = 0; @@ -66,6 +67,13 @@ update_scissor( struct st_context *st ) scissor.minx = scissor.miny = scissor.maxx = scissor.maxy = 0; } + /* Now invert Y. Pipe drivers use the convention Y=0=top for surfaces + */ + miny = fb->Height - scissor.maxy; + maxy = fb->Height - scissor.miny; + scissor.miny = miny; + scissor.maxy = maxy; + if (memcmp(&scissor, &st->state.scissor, sizeof(scissor)) != 0) { /* state has changed */ st->state.scissor = scissor; /* struct copy */ -- cgit v1.2.3 From 1c8bcc733d695732ca704565b3a10ac5f4172ea3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 10 Aug 2007 16:25:21 -0600 Subject: Y invert, clean-up --- src/mesa/state_tracker/st_cb_clear.c | 156 +++++++++++++++++------------------ 1 file changed, 74 insertions(+), 82 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index e3392101a8..2583b4f1bd 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -198,79 +198,96 @@ draw_quad(GLcontext *ctx, /** * Do glClear by drawing a quadrilateral. + * The vertices of the quad will be computed from the + * ctx->DrawBuffer->_X/Ymin/max fields. */ static void -clear_with_quad(GLcontext *ctx, GLuint x0, GLuint y0, - GLuint x1, GLuint y1, +clear_with_quad(GLcontext *ctx, GLboolean color, GLboolean depth, GLboolean stencil) { - static struct st_fragment_program *stfp = NULL; struct st_context *st = ctx->st; - struct pipe_alpha_test_state alpha_test; - struct pipe_blend_state blend; - struct pipe_depth_state depth_test; - struct pipe_stencil_state stencil_test; - struct pipe_setup_state setup; - struct pipe_fs_state fs; + const GLfloat x0 = ctx->DrawBuffer->_Xmin; + const GLfloat y0 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax; + const GLfloat x1 = ctx->DrawBuffer->_Xmax; + const GLfloat y1 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin; /* alpha state: disabled */ - memset(&alpha_test, 0, sizeof(alpha_test)); - st->pipe->set_alpha_test_state(st->pipe, &alpha_test); + { + struct pipe_alpha_test_state alpha_test; + memset(&alpha_test, 0, sizeof(alpha_test)); + st->pipe->set_alpha_test_state(st->pipe, &alpha_test); + } /* blend state: RGBA masking */ - memset(&blend, 0, sizeof(blend)); - if (color) { - if (ctx->Color.ColorMask[0]) - blend.colormask |= PIPE_MASK_R; - if (ctx->Color.ColorMask[1]) - blend.colormask |= PIPE_MASK_G; - if (ctx->Color.ColorMask[2]) - blend.colormask |= PIPE_MASK_B; - if (ctx->Color.ColorMask[3]) - blend.colormask |= PIPE_MASK_A; - if (st->ctx->Color.DitherFlag) - blend.dither = 1; + { + struct pipe_blend_state blend; + memset(&blend, 0, sizeof(blend)); + if (color) { + if (ctx->Color.ColorMask[0]) + blend.colormask |= PIPE_MASK_R; + if (ctx->Color.ColorMask[1]) + blend.colormask |= PIPE_MASK_G; + if (ctx->Color.ColorMask[2]) + blend.colormask |= PIPE_MASK_B; + if (ctx->Color.ColorMask[3]) + blend.colormask |= PIPE_MASK_A; + if (st->ctx->Color.DitherFlag) + blend.dither = 1; + } + st->pipe->set_blend_state(st->pipe, &blend); } - st->pipe->set_blend_state(st->pipe, &blend); /* depth state: always pass */ - memset(&depth_test, 0, sizeof(depth_test)); - if (depth) { - depth_test.enabled = 1; - depth_test.writemask = 1; - depth_test.func = PIPE_FUNC_ALWAYS; + { + struct pipe_depth_state depth_test; + memset(&depth_test, 0, sizeof(depth_test)); + if (depth) { + depth_test.enabled = 1; + depth_test.writemask = 1; + depth_test.func = PIPE_FUNC_ALWAYS; + } + st->pipe->set_depth_state(st->pipe, &depth_test); } - st->pipe->set_depth_state(st->pipe, &depth_test); /* setup state: nothing */ - memset(&setup, 0, sizeof(setup)); - if (ctx->Scissor.Enabled) - setup.scissor = 1; - st->pipe->set_setup_state(st->pipe, &setup); + { + struct pipe_setup_state setup; + memset(&setup, 0, sizeof(setup)); + if (ctx->Scissor.Enabled) + setup.scissor = 1; + st->pipe->set_setup_state(st->pipe, &setup); + } /* stencil state: always set to ref value */ - memset(&stencil_test, 0, sizeof(stencil_test)); - if (stencil) { - stencil_test.front_enabled = 1; - stencil_test.front_func = PIPE_FUNC_ALWAYS; - stencil_test.front_fail_op = PIPE_STENCIL_OP_REPLACE; - stencil_test.front_zpass_op = PIPE_STENCIL_OP_REPLACE; - stencil_test.front_zfail_op = PIPE_STENCIL_OP_REPLACE; - stencil_test.ref_value[0] = ctx->Stencil.Clear; - stencil_test.value_mask[0] = 0xff; - stencil_test.write_mask[0] = ctx->Stencil.WriteMask[0] & 0xff; + { + struct pipe_stencil_state stencil_test; + memset(&stencil_test, 0, sizeof(stencil_test)); + if (stencil) { + stencil_test.front_enabled = 1; + stencil_test.front_func = PIPE_FUNC_ALWAYS; + stencil_test.front_fail_op = PIPE_STENCIL_OP_REPLACE; + stencil_test.front_zpass_op = PIPE_STENCIL_OP_REPLACE; + stencil_test.front_zfail_op = PIPE_STENCIL_OP_REPLACE; + stencil_test.ref_value[0] = ctx->Stencil.Clear; + stencil_test.value_mask[0] = 0xff; + stencil_test.write_mask[0] = ctx->Stencil.WriteMask[0] & 0xff; + } + st->pipe->set_stencil_state(st->pipe, &stencil_test); } - st->pipe->set_stencil_state(st->pipe, &stencil_test); /* fragment shader state: color pass-through program */ - if (!stfp) { - stfp = make_color_shader(st); + { + static struct st_fragment_program *stfp = NULL; + struct pipe_fs_state fs; + if (!stfp) { + stfp = make_color_shader(st); + } + memset(&fs, 0, sizeof(fs)); + fs.inputs_read = stfp->Base.Base.InputsRead; + fs.tokens = &stfp->tokens[0]; + fs.constants = NULL; + st->pipe->set_fs_state(st->pipe, &fs); } - memset(&fs, 0, sizeof(fs)); - fs.inputs_read = stfp->Base.Base.InputsRead; - fs.tokens = &stfp->tokens[0]; - fs.constants = NULL; - st->pipe->set_fs_state(st->pipe, &fs); /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); @@ -306,12 +323,7 @@ clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) } else { /* masking or scissoring */ - clear_with_quad(ctx, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmax, - ctx->DrawBuffer->_Ymax, - GL_TRUE, GL_FALSE, GL_FALSE); + clear_with_quad(ctx, GL_TRUE, GL_FALSE, GL_FALSE); } } @@ -333,12 +345,7 @@ clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) else { /* scissoring */ /* XXX point framebuffer.cbufs[0] at the accum buffer */ - clear_with_quad(ctx, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmax, - ctx->DrawBuffer->_Ymax, - GL_TRUE, GL_FALSE, GL_FALSE); + clear_with_quad(ctx, GL_TRUE, GL_FALSE, GL_FALSE); } } @@ -356,12 +363,7 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) } else { /* masking or scissoring or combined z/stencil buffer */ - clear_with_quad(ctx, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmax, - ctx->DrawBuffer->_Ymax, - GL_FALSE, GL_TRUE, GL_FALSE); + clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_FALSE); } } @@ -381,12 +383,7 @@ clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) } else { /* masking or scissoring */ - clear_with_quad(ctx, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmax, - ctx->DrawBuffer->_Ymax, - GL_FALSE, GL_FALSE, GL_TRUE); + clear_with_quad(ctx, GL_FALSE, GL_FALSE, GL_TRUE); } } @@ -421,12 +418,7 @@ clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) } else { /* masking or scissoring */ - clear_with_quad(ctx, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmin, - ctx->DrawBuffer->_Xmax, - ctx->DrawBuffer->_Ymax, - GL_FALSE, GL_TRUE, GL_TRUE); + clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_TRUE); } } -- cgit v1.2.3 From 39407fd85467141fceafbedf52d9e55e008eb011 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 10 Aug 2007 16:42:26 +0100 Subject: Move string functions to state_tracker, add queries to pipe, winsys. --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 7 ++ src/mesa/pipe/i915simple/Makefile | 1 + src/mesa/pipe/i915simple/i915_context.c | 15 +--- src/mesa/pipe/i915simple/i915_context.h | 2 + src/mesa/pipe/i915simple/i915_reg.h | 9 +++ src/mesa/pipe/i915simple/i915_strings.c | 83 +++++++++++++++++++++ src/mesa/pipe/p_context.h | 7 ++ src/mesa/pipe/p_winsys.h | 4 + src/mesa/pipe/softpipe/sp_context.c | 12 +++ src/mesa/sources | 1 + src/mesa/state_tracker/st_cb_strings.c | 86 ++++++++++++++++++++++ src/mesa/state_tracker/st_cb_strings.h | 38 ++++++++++ src/mesa/state_tracker/st_context.c | 2 + src/mesa/state_tracker/st_context.h | 2 + 14 files changed, 258 insertions(+), 11 deletions(-) create mode 100644 src/mesa/pipe/i915simple/i915_strings.c create mode 100644 src/mesa/state_tracker/st_cb_strings.c create mode 100644 src/mesa/state_tracker/st_cb_strings.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 015d8fc2bb..eb64f5aa20 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -185,6 +185,12 @@ intel_printf( struct pipe_winsys *sws, const char *fmtString, ... ) va_end( args ); } +static const char * +intel_get_name( struct pipe_winsys *sws ) +{ + return "Intel/DRI/ttm"; +} + struct pipe_winsys * intel_create_pipe_winsys( struct intel_context *intel ) @@ -209,6 +215,7 @@ intel_create_pipe_winsys( struct intel_context *intel ) iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; iws->winsys.wait_idle = intel_wait_idle; iws->winsys.printf = intel_printf; + iws->winsys.get_name = intel_get_name; iws->intel = intel; return &iws->winsys; diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 28fe70d069..9a0d40a8d4 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -19,6 +19,7 @@ DRIVER_SOURCES = \ i915_state_derived.c \ i915_state_emit.c \ i915_state_fragprog.c \ + i915_strings.c \ i915_prim_emit.c \ i915_tex_layout.c \ i915_surface.c diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index d9fbada1ff..199352cd70 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -31,20 +31,12 @@ #include "i915_state.h" #include "i915_batch.h" #include "i915_tex_layout.h" +#include "i915_reg.h" #include "pipe/draw/draw_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GME 0x27AE -#define PCI_CHIP_G33_G 0x29C2 -#define PCI_CHIP_Q35_G 0x29B2 -#define PCI_CHIP_Q33_G 0x29D2 - /** * Return list of supported surface/texture formats. @@ -190,8 +182,6 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, struct i915_context *i915; unsigned is_i945 = 0; - /* TODO: Push this down into the pipe driver: - */ switch (pci_id) { case PCI_CHIP_I915_G: case PCI_CHIP_I915_GM: @@ -241,7 +231,10 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915_init_surface_functions(i915); i915_init_state_functions(i915); i915_init_flush_functions(i915); + i915_init_string_functions(i915); + i915->pci_id = pci_id; + i915->flags.is_i945 = is_i945; if (i915->flags.is_i945) i915->pipe.mipmap_tree_layout = i945_miptree_layout; diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 7a73a8d8d0..93b5e2546c 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -121,6 +121,7 @@ struct i915_context GLuint hardware_dirty; GLuint debug; + GLuint pci_id; struct { unsigned is_i945:1; @@ -188,6 +189,7 @@ void i915_init_region_functions( struct i915_context *i915 ); void i915_init_surface_functions( struct i915_context *i915 ); void i915_init_state_functions( struct i915_context *i915 ); void i915_init_flush_functions( struct i915_context *i915 ); +void i915_init_string_functions( struct i915_context *i915 ); diff --git a/src/mesa/pipe/i915simple/i915_reg.h b/src/mesa/pipe/i915simple/i915_reg.h index 0776b75ef9..f4070ef93e 100644 --- a/src/mesa/pipe/i915simple/i915_reg.h +++ b/src/mesa/pipe/i915simple/i915_reg.h @@ -961,5 +961,14 @@ #define BLENDFACT_INV_CONST_ALPHA 0x0f #define BLENDFACT_MASK 0x0f +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GME 0x27AE +#define PCI_CHIP_G33_G 0x29C2 +#define PCI_CHIP_Q35_G 0x29B2 +#define PCI_CHIP_Q33_G 0x29D2 + #endif diff --git a/src/mesa/pipe/i915simple/i915_strings.c b/src/mesa/pipe/i915simple/i915_strings.c new file mode 100644 index 0000000000..a9b0bfc7c2 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_strings.c @@ -0,0 +1,83 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_context.h" +#include "i915_reg.h" + + +static const char *i915_get_vendor( struct pipe_context *pipe ) +{ + return "Tungsten Graphics, Inc."; +} + + +static const char *i915_get_name( struct pipe_context *pipe ) +{ + static char buffer[128]; + const char *chipset; + + switch (i915_context(pipe)->pci_id) { + case PCI_CHIP_I915_G: + chipset = "915G"; + break; + case PCI_CHIP_I915_GM: + chipset = "915GM"; + break; + case PCI_CHIP_I945_G: + chipset = "945G"; + break; + case PCI_CHIP_I945_GM: + chipset = "945GM"; + break; + case PCI_CHIP_I945_GME: + chipset = "945GME"; + break; + case PCI_CHIP_G33_G: + chipset = "G33"; + break; + case PCI_CHIP_Q35_G: + chipset = "Q35"; + break; + case PCI_CHIP_Q33_G: + chipset = "Q33"; + break; + default: + chipset = "unknown"; + break; + } + + snprintf(buffer, sizeof(buffer), "pipe/i915 (chipset: %s)", chipset); + return buffer; +} + + +void +i915_init_string_functions(struct i915_context *i915) +{ + i915->pipe.get_name = i915_get_name; + i915->pipe.get_vendor = i915_get_vendor; +} diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index b303cee5cc..84557b9a4a 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -51,12 +51,19 @@ struct pipe_context { */ const GLuint *(*supported_formats)(struct pipe_context *pipe, GLuint *numFormats); + void (*max_texture_size)(struct pipe_context *pipe, GLuint textureType, /* PIPE_TEXTURE_x */ GLuint *maxWidth, GLuint *maxHeight, GLuint *maxDepth); + const char *(*get_name)( struct pipe_context *pipe ); + + const char *(*get_vendor)( struct pipe_context *pipe ); + + + /* * Drawing */ diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index c455ebdbbe..b7495a205f 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -107,6 +107,10 @@ struct pipe_winsys { */ void (*wait_idle)( struct pipe_winsys *sws ); + /* Queries: + */ + const char *(*get_name)( struct pipe_winsys *sws ); + }; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index b35bd66081..d67ad2c01d 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -224,6 +224,16 @@ static GLuint softpipe_get_occlusion_counter(struct pipe_context *pipe) return softpipe->occlusion_counter; } +static const char *softpipe_get_name( struct pipe_context *pipe ) +{ + return "softpipe"; +} + +static const char *softpipe_get_vendor( struct pipe_context *pipe ) +{ + return "Tungsten Graphics, Inc."; +} + struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, struct softpipe_winsys *softpipe_winsys ) @@ -259,6 +269,8 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.flush = softpipe_flush; softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; + softpipe->pipe.get_name = softpipe_get_name; + softpipe->pipe.get_vendor = softpipe_get_vendor; /* textures */ softpipe->pipe.mipmap_tree_layout = softpipe_mipmap_tree_layout; diff --git a/src/mesa/sources b/src/mesa/sources index a2b9fed82b..c94cd3de26 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -199,6 +199,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_program.c \ state_tracker/st_cb_readpixels.c \ + state_tracker/st_cb_strings.c \ state_tracker/st_cb_texture.c \ state_tracker/st_context.c \ state_tracker/st_draw.c \ diff --git a/src/mesa/state_tracker/st_cb_strings.c b/src/mesa/state_tracker/st_cb_strings.c new file mode 100644 index 0000000000..ac1a8b1422 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_strings.c @@ -0,0 +1,86 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "main/glheader.h" +#include "main/macros.h" +#include "main/version.h" +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" +#include "st_context.h" +#include "st_cb_strings.h" + +#define ST_VERSION_STRING "0.1" + +static const GLubyte * +st_get_string(GLcontext * ctx, GLenum name) +{ + struct st_context *st = st_context(ctx); + struct pipe_context *pipe = st->pipe; + static char buffer[128]; + + + switch (name) { + case GL_VENDOR: { + const char *vendor = pipe->get_vendor( pipe ); + const char *tungsten = "Tungsten Graphics, Inc."; + + /* Tungsten developed the state_tracker module (and much of + * Mesa), but the driver itself may come from elsewhere. The + * additional string allows "and XyzCorp" to reflect this. + */ + if (vendor && strcmp(vendor, tungsten) != 0) + snprintf(buffer, sizeof(buffer), "%s and %s", tungsten, vendor); + else + snprintf(buffer, sizeof(buffer), "%s", tungsten); + + return (GLubyte *) buffer; + break; + } + + case GL_RENDERER: + snprintf(buffer, sizeof(buffer), "TG3D, %s on %s", + pipe->get_name( pipe ), + pipe->winsys->get_name( pipe->winsys )); + + return (GLubyte *) buffer; + + default: + return NULL; + } +} + + +void st_init_string_functions(struct dd_function_table *functions) +{ + functions->GetString = st_get_string; +} diff --git a/src/mesa/state_tracker/st_cb_strings.h b/src/mesa/state_tracker/st_cb_strings.h new file mode 100644 index 0000000000..3b765aaa59 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_strings.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_STRINGS_H +#define ST_CB_STRINGS_H + + +extern void +st_init_string_functions(struct dd_function_table *functions); + + +#endif /* ST_CB_CLEAR_H */ + diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 2bdcc6caf3..7bb9cbda45 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -35,6 +35,7 @@ #include "st_cb_readpixels.h" #include "st_cb_texture.h" #include "st_cb_flush.h" +#include "st_cb_strings.h" #include "st_atom.h" #include "st_draw.h" #include "st_program.h" @@ -111,4 +112,5 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_readpixels_functions(functions); st_init_texture_functions(functions); st_init_flush_functions(functions); + st_init_string_functions(functions); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 88e2701fc3..8ce85cddaa 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -116,4 +116,6 @@ static INLINE struct st_context *st_context(GLcontext *ctx) extern void st_init_driver_functions(struct dd_function_table *functions); + + #endif -- cgit v1.2.3 From ecea02fd705b8029b4ac79e638dc6af4a33bddcf Mon Sep 17 00:00:00 2001 From: keithw Date: Sat, 11 Aug 2007 00:09:21 +0100 Subject: remove intelGetString() --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 64 ----------------------- 1 file changed, 64 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 6d92ce7e19..5aa59a9454 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -44,7 +44,6 @@ #include "intel_winsys.h" #include "intel_ioctl.h" #include "intel_batchbuffer.h" -#include "intel_blit.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" @@ -88,68 +87,6 @@ _glthread_Mutex lockMutex; static GLboolean lockMutexInit = GL_FALSE; -static const GLubyte * -intelGetString(GLcontext * ctx, GLenum name) -{ - const char *chipset; - static char buffer[128]; - - switch (name) { - case GL_VENDOR: - return (GLubyte *) "Tungsten Graphics, Inc"; - break; - - case GL_RENDERER: - switch (intel_context(ctx)->intelScreen->deviceID) { - case PCI_CHIP_845_G: - chipset = "Intel(R) 845G"; - break; - case PCI_CHIP_I830_M: - chipset = "Intel(R) 830M"; - break; - case PCI_CHIP_I855_GM: - chipset = "Intel(R) 852GM/855GM"; - break; - case PCI_CHIP_I865_G: - chipset = "Intel(R) 865G"; - break; - case PCI_CHIP_I915_G: - chipset = "Intel(R) 915G"; - break; - case PCI_CHIP_I915_GM: - chipset = "Intel(R) 915GM"; - break; - case PCI_CHIP_I945_G: - chipset = "Intel(R) 945G"; - break; - case PCI_CHIP_I945_GM: - chipset = "Intel(R) 945GM"; - break; - case PCI_CHIP_I945_GME: - chipset = "Intel(R) 945GME"; - break; - case PCI_CHIP_G33_G: - chipset = "Intel(R) G33"; - break; - case PCI_CHIP_Q35_G: - chipset = "Intel(R) Q35"; - break; - case PCI_CHIP_Q33_G: - chipset = "Intel(R) Q33"; - break; - default: - chipset = "Unknown Intel Chipset"; - break; - } - - (void) driGetRendererString(buffer, chipset, DRIVER_DATE, 0); - return (GLubyte *) buffer; - - default: - return NULL; - } -} - /** * Extension strings exported by the intel driver. @@ -248,7 +185,6 @@ intelInitDriverFunctions(struct dd_function_table *functions) { _mesa_init_driver_functions(functions); - functions->GetString = intelGetString; functions->UpdateState = intelInvalidateState; st_init_driver_functions(functions); -- cgit v1.2.3 From acff996c447cf3502bf5f287d661d15b22325304 Mon Sep 17 00:00:00 2001 From: keithw Date: Sat, 11 Aug 2007 00:10:30 +0100 Subject: move locking to new file --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 119 +-------------- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 169 ++++++++++++++++++++++ 2 files changed, 170 insertions(+), 118 deletions(-) create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_lock.c (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 5aa59a9454..af5deb4dc6 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -440,122 +440,5 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, return GL_TRUE; } -static void -intelContendedLock(struct intel_context *intel, GLuint flags) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - __DRIscreenPrivate *sPriv = intel->driScreen; - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - drmI830Sarea *sarea = intel->sarea; - - drmGetLock(intel->driFd, intel->hHWContext, flags); - - DBG(LOCK, "%s - got contended lock\n", __progname); - - /* If the window moved, may need to set a new cliprect now. - * - * NOTE: This releases and regains the hw lock, so all state - * checking must be done *after* this call: - */ - if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - - if (sarea->width != intelScreen->front.width || - sarea->height != intelScreen->front.height) { - - intelUpdateScreenRotation(sPriv, sarea); - } - -#if 0 - if (sarea->width != intel->width || - sarea->height != intel->height || - sarea->rotation != intel->current_rotation) { - int numClipRects = intel->numClipRects; - - /* - * FIXME: Really only need to do this when drawing to a - * common back- or front buffer. - */ - - /* - * This will essentially drop the outstanding batchbuffer on the floor. - */ - intel->numClipRects = 0; - - INTEL_FIREVERTICES(intel); - - if (intel->batch->map != intel->batch->ptr) - intel_batchbuffer_flush(intel->batch); - - intel->numClipRects = numClipRects; - - /* force window update */ - intel->lastStamp = 0; - - intel->width = sarea->width; - intel->height = sarea->height; - intel->current_rotation = sarea->rotation; - } -#endif -} - - -/* Lock the hardware and validate our state. - */ -void LOCK_HARDWARE( struct intel_context *intel ) -{ - char __ret=0; - struct intel_framebuffer *intel_fb = NULL; - int curbuf; - - _glthread_LOCK_MUTEX(lockMutex); - assert(!intel->locked); - - if (intel->driDrawable) { - intel_fb = intel->driDrawable->driverPrivate; - } - - curbuf = 0; /* current draw buf: 0 = front, 1 = back */ - - if (intel_fb && intel_fb->vblank_flags && - !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && - (intel_fb->vbl_waited - intel_fb->vbl_pending[curbuf]) > (1<<23)) { - drmVBlank vbl; - - vbl.request.type = DRM_VBLANK_ABSOLUTE; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - vbl.request.type |= DRM_VBLANK_SECONDARY; - } - - vbl.request.sequence = intel_fb->vbl_pending[curbuf]; - drmWaitVBlank(intel->driFd, &vbl); - intel_fb->vbl_waited = vbl.reply.sequence; - } - - DRM_CAS(intel->driHwLock, intel->hHWContext, - (DRM_LOCK_HELD|intel->hHWContext), __ret); - - if (__ret) - intelContendedLock( intel, 0 ); - - DBG(LOCK, "%s - locked\n", __progname); - - intel->locked = 1; -} - - - /* Unlock the hardware using the global current context - */ -void UNLOCK_HARDWARE( struct intel_context *intel ) -{ - assert(intel->locked); - intel->locked = 0; - - DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); - - _glthread_UNLOCK_MUTEX(lockMutex); - - DBG(LOCK, "%s - unlocked\n", __progname); -} + diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c new file mode 100644 index 0000000000..34319545d9 --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -0,0 +1,169 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "glheader.h" +#include "context.h" +#include "extensions.h" + +#include "intel_screen.h" + +#include "i830_dri.h" + +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" + +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" + +#include "drirenderbuffer.h" +#include "vblank.h" +#include "utils.h" +#include "xmlpool.h" /* for symbolic values of enum-type options */ + + + +_glthread_DECLARE_STATIC_MUTEX( lockMutex ); + + +static void +intelContendedLock(struct intel_context *intel, GLuint flags) +{ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + __DRIscreenPrivate *sPriv = intel->driScreen; + intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + drmI830Sarea *sarea = intel->sarea; + + drmGetLock(intel->driFd, intel->hHWContext, flags); + + DBG(LOCK, "%s - got contended lock\n", __progname); + + /* If the window moved, may need to set a new cliprect now. + * + * NOTE: This releases and regains the hw lock, so all state + * checking must be done *after* this call: + */ + if (dPriv) + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + + if (sarea->width != intelScreen->front.width || + sarea->height != intelScreen->front.height) { + + intelUpdateScreenRotation(sPriv, sarea); + } + +#if 0 + if (sarea->width != intel->width || + sarea->height != intel->height || + sarea->rotation != intel->current_rotation) { + int numClipRects = intel->numClipRects; + + /* + * FIXME: Really only need to do this when drawing to a + * common back- or front buffer. + */ + + /* + * This will essentially drop the outstanding batchbuffer on the floor. + */ + intel->numClipRects = 0; + + INTEL_FIREVERTICES(intel); + + if (intel->batch->map != intel->batch->ptr) + intel_batchbuffer_flush(intel->batch); + + intel->numClipRects = numClipRects; + + /* force window update */ + intel->lastStamp = 0; + + intel->width = sarea->width; + intel->height = sarea->height; + intel->current_rotation = sarea->rotation; + } +#endif +} + +/* Lock the hardware and validate our state. + */ +void LOCK_HARDWARE( struct intel_context *intel ) +{ + char __ret=0; + struct intel_framebuffer *intel_fb = NULL; + int curbuf; + + _glthread_LOCK_MUTEX(lockMutex); + assert(!intel->locked); + + if (intel->driDrawable) { + intel_fb = intel->driDrawable->driverPrivate; + } + + curbuf = 0; /* current draw buf: 0 = front, 1 = back */ + + if (intel_fb && intel_fb->vblank_flags && + !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && + (intel_fb->vbl_waited - intel_fb->vbl_pending[curbuf]) > (1<<23)) { + drmVBlank vbl; + + vbl.request.type = DRM_VBLANK_ABSOLUTE; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + vbl.request.type |= DRM_VBLANK_SECONDARY; + } + + vbl.request.sequence = intel_fb->vbl_pending[curbuf]; + drmWaitVBlank(intel->driFd, &vbl); + intel_fb->vbl_waited = vbl.reply.sequence; + } + + DRM_CAS(intel->driHwLock, intel->hHWContext, + (DRM_LOCK_HELD|intel->hHWContext), __ret); + + if (__ret) + intelContendedLock( intel, 0 ); + + DBG(LOCK, "%s - locked\n", __progname); + + intel->locked = 1; +} + + + /* Unlock the hardware using the global current context + */ +void UNLOCK_HARDWARE( struct intel_context *intel ) +{ + assert(intel->locked); + intel->locked = 0; + + DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); + + _glthread_UNLOCK_MUTEX(lockMutex); + + DBG(LOCK, "%s - unlocked\n", __progname); +} -- cgit v1.2.3 From 20995b74c9bc12a157f9ab13e70dd22ab39c8bbf Mon Sep 17 00:00:00 2001 From: keithw Date: Sat, 11 Aug 2007 00:13:40 +0100 Subject: remove unused irq functions --- src/mesa/drivers/dri/intel_winsys/intel_ioctl.c | 44 ------------------------- src/mesa/drivers/dri/intel_winsys/intel_ioctl.h | 2 -- 2 files changed, 46 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c index 6e4d134b80..cf6806b245 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c @@ -37,53 +37,9 @@ #include "intel_context.h" #include "intel_ioctl.h" #include "intel_batchbuffer.h" -#include "intel_blit.h" #include "drm.h" -int -intelEmitIrqLocked(struct intel_context *intel) -{ - drmI830IrqEmit ie; - int ret, seq; - - assert(((*(int *) intel->driHwLock) & ~DRM_LOCK_CONT) == - (DRM_LOCK_HELD | intel->hHWContext)); - - ie.irq_seq = &seq; - - ret = drmCommandWriteRead(intel->driFd, DRM_I830_IRQ_EMIT, - &ie, sizeof(ie)); - if (ret) { - fprintf(stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret); - exit(1); - } - - DBG(IOCTL, "%s --> %d\n", __FUNCTION__, seq); - - return seq; -} - -void -intelWaitIrq(struct intel_context *intel, int seq) -{ - int ret; - - DBG(IOCTL, "%s %d\n", __FUNCTION__, seq); - - intel->iw.irq_seq = seq; - - do { - ret = - drmCommandWrite(intel->driFd, DRM_I830_IRQ_WAIT, &intel->iw, - sizeof(intel->iw)); - } while (ret == -EAGAIN || ret == -EINTR); - - if (ret) { - fprintf(stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret); - exit(1); - } -} void diff --git a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h index e8d07de893..10b811d7f3 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h @@ -30,8 +30,6 @@ #include "intel_context.h" -void intelWaitIrq(struct intel_context *intel, int seq); -int intelEmitIrqLocked(struct intel_context *intel); void intel_batch_ioctl(struct intel_context *intel, GLuint start_offset, -- cgit v1.2.3 From 4bc932ab292d505700e830520de7f9c9eb385086 Mon Sep 17 00:00:00 2001 From: keithw Date: Sat, 11 Aug 2007 00:16:36 +0100 Subject: move batch ioctl to intel_batchbuffer.c --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 35 +++++++++ src/mesa/drivers/dri/intel_winsys/intel_ioctl.c | 90 ---------------------- src/mesa/drivers/dri/intel_winsys/intel_ioctl.h | 38 --------- 3 files changed, 35 insertions(+), 128 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_ioctl.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_ioctl.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 30511bcba1..4d0759fc32 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -28,6 +28,7 @@ #include "intel_batchbuffer.h" #include "intel_ioctl.h" #include "intel_reg.h" +#include "drm.h" /* Relocations in kernel space: * - pass dma buffer seperately @@ -163,6 +164,40 @@ intel_batchbuffer_free(struct intel_batchbuffer *batch) free(batch); } + + + +static void +intel_batch_ioctl(struct intel_context *intel, + GLuint start_offset, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) +{ + drmI830BatchBuffer batch; + + batch.start = start_offset; + batch.used = used; + batch.cliprects = intel->pClipRects; + batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + batch.DR1 = 0; + batch.DR4 = 0; /* still need this ? */ + + DBG(IOCTL, "%s: 0x%x..0x%x DR4: %x cliprects: %d\n", + __FUNCTION__, + batch.start, + batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); + + if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, + sizeof(batch))) { + _mesa_printf("DRM_I830_BATCHBUFFER: %d\n", -errno); + UNLOCK_HARDWARE(intel); + exit(1); + } +} + + + + /* TODO: Push this whole function into bufmgr. */ static void diff --git a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c deleted file mode 100644 index cf6806b245..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.c +++ /dev/null @@ -1,90 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include -#include -#include -#include - -#include "mtypes.h" -#include "context.h" - -#include "intel_context.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "drm.h" - - - - -void -intel_batch_ioctl(struct intel_context *intel, - GLuint start_offset, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) -{ - drmI830BatchBuffer batch; - - assert(intel->locked); - assert(used); - - DBG(IOCTL, "%s used %d offset %x..%x ignore_cliprects %d\n", - __FUNCTION__, - used, start_offset, start_offset + used, ignore_cliprects); - - /* Throw away non-effective packets. Won't work once we have - * hardware contexts which would preserve statechanges beyond a - * single buffer. - */ - - - - batch.start = start_offset; - batch.used = used; - batch.cliprects = intel->pClipRects; - batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; - batch.DR1 = 0; - batch.DR4 = 0; /* still need this ? */ - - DBG(IOCTL, "%s: 0x%x..0x%x DR4: %x cliprects: %d\n", - __FUNCTION__, - batch.start, - batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); - - if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, - sizeof(batch))) { - fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n", -errno); - UNLOCK_HARDWARE(intel); - exit(1); - } - - /* FIXME: use hardware contexts to avoid 'losing' hardware after - * each buffer flush. - */ - //intel->vtbl.lost_hardware(intel); -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h b/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h deleted file mode 100644 index 10b811d7f3..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_ioctl.h +++ /dev/null @@ -1,38 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_IOCTL_H -#define INTEL_IOCTL_H - -#include "intel_context.h" - - -void intel_batch_ioctl(struct intel_context *intel, - GLuint start_offset, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock); -#endif -- cgit v1.2.3 From 37863c3c426da5069c3cba977634b6d9c2aa6714 Mon Sep 17 00:00:00 2001 From: keithw Date: Sat, 11 Aug 2007 00:18:47 +0100 Subject: remove intel_ioctl.c --- src/mesa/drivers/dri/intel_winsys/Makefile | 2 - src/mesa/drivers/dri/intel_winsys/intel_blit.c | 227 --------------------- src/mesa/drivers/dri/intel_winsys/intel_blit.h | 62 ------ src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 203 +++++++++++++++--- src/mesa/drivers/dri/intel_winsys/intel_buffers.h | 6 +- .../drivers/dri/intel_winsys/intel_winsys_i915.c | 1 - .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 2 +- 7 files changed, 178 insertions(+), 325 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_blit.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_blit.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index a5db6547a9..aca31f712d 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -16,9 +16,7 @@ DRIVER_SOURCES = \ intel_winsys_i915.c \ intel_batchbuffer.c \ intel_buffers.c \ - intel_blit.c \ intel_context.c \ - intel_ioctl.c \ intel_screen.c \ intel_batchpool.c diff --git a/src/mesa/drivers/dri/intel_winsys/intel_blit.c b/src/mesa/drivers/dri/intel_winsys/intel_blit.c deleted file mode 100644 index 1578340e17..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_blit.c +++ /dev/null @@ -1,227 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "mtypes.h" -#include "context.h" -#include "enums.h" - -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "intel_buffers.h" -#include "intel_context.h" -#include "intel_reg.h" -#include "vblank.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_cb_fbo.h" - - -/** - * Copy the back color buffer to the front color buffer. - * Used for SwapBuffers(). - */ -void -intelCopyBuffer(__DRIdrawablePrivate * dPriv, - const drm_clip_rect_t * rect) -{ - - struct intel_context *intel; - const intelScreenPrivate *intelScreen; - - DBG(SWAP, "%s\n", __FUNCTION__); - - assert(dPriv); - - intel = intelScreenContext(dPriv->driScreenPriv->private); - if (!intel) - return; - - intelScreen = intel->intelScreen; - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - current context (which is what intelScreenContext should return) might - not get a contended lock and thus cliprects not updated (tests/manywin) */ - if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - - if (dPriv && dPriv->numClipRects) { - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; -#if 0 - const struct pipe_region *backRegion - = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? - intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : - intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); -#endif - const int backWidth = intel_fb->Base.Width; - const int backHeight = intel_fb->Base.Height; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; -#if 0 - const int srcpitch = backRegion->pitch; -#endif - const int cpp = intelScreen->front.cpp; - int BR13, CMD; - int i; - - const struct pipe_surface *backSurf; - const struct pipe_region *backRegion; - int srcpitch; - struct st_renderbuffer *strb; - - /* blit from back color buffer if it exists, else front buffer */ - strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); - if (strb) { - backSurf = strb->surface; - } - else { - strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); - backSurf = strb->surface; - } - - backRegion = backSurf->region; - srcpitch = backRegion->pitch; - - ASSERT(intel_fb); - ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ - ASSERT(backRegion); - ASSERT(backRegion->cpp == cpp); - - DBG(SWAP, "front pitch %d back pitch %d\n", - pitch, backRegion->pitch); - - if (cpp == 2) { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - } - else { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - } - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) - continue; - - box = *pbox; - - if (rect) { - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > backWidth) - box.x2 = backWidth + box.x1; - if (box.y2 - box.y1 > backHeight) - box.y2 = backHeight + box.y1; - - DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((box.y1 << 16) | box.x1); - OUT_BATCH((box.y2 << 16) | box.x2); - - OUT_RELOC(intelScreen->front.buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((sbox.y1 << 16) | sbox.x1); - OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - ADVANCE_BATCH(); - } - - if (intel->first_swap_fence) - driFenceUnReference(intel->first_swap_fence); - intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); - driFenceReference(intel->first_swap_fence); - } - - UNLOCK_HARDWARE(intel); - - /* XXX this is bogus. The context here may not even be bound to this drawable! */ - if (intel->lastStamp != dPriv->lastStamp) { - GET_CURRENT_CONTEXT(currctx); - struct intel_context *intelcurrent = intel_context(currctx); - if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { - intelWindowMoved(intel); - intel->lastStamp = dPriv->lastStamp; - } - } - -} - - - - - - - diff --git a/src/mesa/drivers/dri/intel_winsys/intel_blit.h b/src/mesa/drivers/dri/intel_winsys/intel_blit.h deleted file mode 100644 index 46c2594477..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_blit.h +++ /dev/null @@ -1,62 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BLIT_H -#define INTEL_BLIT_H - -#include "intel_context.h" -#include "intel_ioctl.h" -#include "dri_bufmgr.h" - -extern void intelCopyBuffer(__DRIdrawablePrivate * dpriv, - const drm_clip_rect_t * rect); - -extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask); - -extern void intelEmitCopyBlit(struct intel_context *intel, - GLuint cpp, - GLshort src_pitch, - struct _DriBufferObject *src_buffer, - GLuint src_offset, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort srcx, GLshort srcy, - GLshort dstx, GLshort dsty, - GLshort w, GLshort h, - GLenum logicop ); - -extern void intelEmitFillBlit(struct intel_context *intel, - GLuint cpp, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort x, GLshort y, - GLshort w, GLshort h, GLuint value, GLuint mask); - - -#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c index 1bd2a36738..0823c5c8ea 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c @@ -27,7 +27,6 @@ #include "intel_screen.h" #include "intel_context.h" -#include "intel_blit.h" #include "intel_buffers.h" #include "intel_batchbuffer.h" #include "intel_reg.h" @@ -37,6 +36,7 @@ #include "vblank.h" #include "pipe/p_context.h" +#include "state_tracker/st_cb_fbo.h" /* This block can be removed when libdrm >= 2.3.1 is required */ @@ -56,41 +56,188 @@ typedef struct drm_i915_flip { #endif + + /** - * XXX move this into a new dri/common/cliprects.c file. + * Copy the back color buffer to the front color buffer. + * Used for SwapBuffers(). */ -GLboolean -intel_intersect_cliprects(drm_clip_rect_t * dst, - const drm_clip_rect_t * a, - const drm_clip_rect_t * b) +void +intelCopyBuffer(__DRIdrawablePrivate * dPriv, + const drm_clip_rect_t * rect) { - GLint bx = b->x1; - GLint by = b->y1; - GLint bw = b->x2 - bx; - GLint bh = b->y2 - by; - - if (bx < a->x1) - bw -= a->x1 - bx, bx = a->x1; - if (by < a->y1) - bh -= a->y1 - by, by = a->y1; - if (bx + bw > a->x2) - bw = a->x2 - bx; - if (by + bh > a->y2) - bh = a->y2 - by; - if (bw <= 0) - return GL_FALSE; - if (bh <= 0) - return GL_FALSE; - dst->x1 = bx; - dst->y1 = by; - dst->x2 = bx + bw; - dst->y2 = by + bh; + struct intel_context *intel; + const intelScreenPrivate *intelScreen; + + DBG(SWAP, "%s\n", __FUNCTION__); + + assert(dPriv); + + intel = intelScreenContext(dPriv->driScreenPriv->private); + if (!intel) + return; + + intelScreen = intel->intelScreen; + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + current context (which is what intelScreenContext should return) might + not get a contended lock and thus cliprects not updated (tests/manywin) */ + if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + + + if (dPriv && dPriv->numClipRects) { + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; +#if 0 + const struct pipe_region *backRegion + = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? + intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : + intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); +#endif + const int backWidth = intel_fb->Base.Width; + const int backHeight = intel_fb->Base.Height; + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; +#if 0 + const int srcpitch = backRegion->pitch; +#endif + const int cpp = intelScreen->front.cpp; + int BR13, CMD; + int i; + + const struct pipe_surface *backSurf; + const struct pipe_region *backRegion; + int srcpitch; + struct st_renderbuffer *strb; + + /* blit from back color buffer if it exists, else front buffer */ + strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); + if (strb) { + backSurf = strb->surface; + } + else { + strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + backSurf = strb->surface; + } + + backRegion = backSurf->region; + srcpitch = backRegion->pitch; + + ASSERT(intel_fb); + ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ + ASSERT(backRegion); + ASSERT(backRegion->cpp == cpp); + + DBG(SWAP, "front pitch %d back pitch %d\n", + pitch, backRegion->pitch); + + if (cpp == 2) { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + } + else { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + } + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + drm_clip_rect_t sbox; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->front.width || + pbox->y2 > intelScreen->front.height) + continue; + + box = *pbox; + + if (rect) { + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > backWidth) + box.x2 = backWidth + box.x1; + if (box.y2 - box.y1 > backHeight) + box.y2 = backHeight + box.y1; + + DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((box.y1 << 16) | box.x1); + OUT_BATCH((box.y2 << 16) | box.x2); + + OUT_RELOC(intelScreen->front.buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + OUT_BATCH((sbox.y1 << 16) | sbox.x1); + OUT_BATCH((srcpitch * cpp) & 0xffff); + OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + ADVANCE_BATCH(); + } + + if (intel->first_swap_fence) + driFenceUnReference(intel->first_swap_fence); + intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); + driFenceReference(intel->first_swap_fence); + } + + UNLOCK_HARDWARE(intel); + + /* XXX this is bogus. The context here may not even be bound to this drawable! */ + if (intel->lastStamp != dPriv->lastStamp) { + GET_CURRENT_CONTEXT(currctx); + struct intel_context *intelcurrent = intel_context(currctx); + if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { + intelWindowMoved(intel); + intel->lastStamp = dPriv->lastStamp; + } + } - return GL_TRUE; } + /** * This will be called whenever the currently bound window is moved/resized. * XXX: actually, it seems to NOT be called when the window is only moved (BP). diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.h b/src/mesa/drivers/dri/intel_winsys/intel_buffers.h index 3e1597e3c4..d12b5ef96e 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_buffers.h @@ -62,10 +62,8 @@ struct intel_framebuffer }; -extern GLboolean -intel_intersect_cliprects(drm_clip_rect_t * dest, - const drm_clip_rect_t * a, - const drm_clip_rect_t * b); +void intelCopyBuffer(__DRIdrawablePrivate * dPriv, + const drm_clip_rect_t * rect); extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index 56a27a9c31..f6eb050152 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -37,7 +37,6 @@ #include "intel_context.h" #include "intel_batchbuffer.h" #include "intel_winsys.h" -#include "intel_blit.h" #include "pipe/i915simple/i915_winsys.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index eb64f5aa20..bedc7d14ab 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -36,8 +36,8 @@ #include "intel_context.h" #include "intel_winsys.h" +#include "intel_buffers.h" #include "intel_batchbuffer.h" -#include "intel_blit.h" #include "pipe/p_winsys.h" #include "pipe/p_defines.h" -- cgit v1.2.3 From eada74dc543348e03ef4eaf557c9b84afec7824a Mon Sep 17 00:00:00 2001 From: keithw Date: Sat, 11 Aug 2007 00:20:30 +0100 Subject: add intel_lock.c to build --- src/mesa/drivers/dri/intel_winsys/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index aca31f712d..7c37812172 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -17,6 +17,7 @@ DRIVER_SOURCES = \ intel_batchbuffer.c \ intel_buffers.c \ intel_context.c \ + intel_lock.c \ intel_screen.c \ intel_batchpool.c -- cgit v1.2.3 From 2eb839ce1dc9a7737925d69d7b628fc839fa752d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 11 Aug 2007 13:06:24 +0100 Subject: make it easier to compare debug code with non-pipe version --- src/mesa/pipe/i915simple/i915_debug.c | 80 ++++++++++++++++---------------- src/mesa/pipe/i915simple/i915_debug_fp.c | 80 ++++++++++++++++---------------- 2 files changed, 80 insertions(+), 80 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index 8050eb0bf5..10d7817816 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -32,7 +32,7 @@ #include "i915_winsys.h" #include "i915_debug.h" -#define PRINTF( stream, ... ) (stream)->winsys->printf( (stream)->winsys, __VA_ARGS__ ) +#define PRINTF( ... ) (stream)->winsys->printf( (stream)->winsys, __VA_ARGS__ ) static GLboolean debug( struct debug_stream *stream, const char *name, GLuint len ) @@ -41,19 +41,19 @@ static GLboolean debug( struct debug_stream *stream, const char *name, GLuint le GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); if (len == 0) { - PRINTF( stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0] ); + PRINTF( "Error - zero length packet (0x%08x)\n", stream->ptr[0] ); assert(0); return GL_FALSE; } if (stream->print_addresses) - PRINTF(stream, "%08x: ", stream->offset); + PRINTF("%08x: ", stream->offset); - PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF("%s (%d dwords):\n", name, len); for (i = 0; i < len; i++) - PRINTF(stream, "\t\t0x%08x\n", ptr[i]); - PRINTF(stream, "\n"); + PRINTF("\t\t0x%08x\n", ptr[i]); + PRINTF("\n"); stream->offset += len * sizeof(GLuint); @@ -90,17 +90,17 @@ static GLboolean debug_prim( struct debug_stream *stream, const char *name, - PRINTF(stream, "%s %s (%d dwords):\n", name, prim, len); - PRINTF(stream, "\t\t0x%08x\n", ptr[0]); + PRINTF("%s %s (%d dwords):\n", name, prim, len); + PRINTF("\t\t0x%08x\n", ptr[0]); for (i = 1; i < len; i++) { if (dump_floats) - PRINTF(stream, "\t\t0x%08x // %f\n", ptr[i], *(GLfloat *)&ptr[i]); + PRINTF("\t\t0x%08x // %f\n", ptr[i], *(GLfloat *)&ptr[i]); else - PRINTF(stream, "\t\t0x%08x\n", ptr[i]); + PRINTF("\t\t0x%08x\n", ptr[i]); } - PRINTF(stream, "\n"); + PRINTF("\n"); stream->offset += len * sizeof(GLuint); @@ -115,15 +115,15 @@ static GLboolean debug_program( struct debug_stream *stream, const char *name, G GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); if (len == 0) { - PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]); + PRINTF("Error - zero length packet (0x%08x)\n", stream->ptr[0]); assert(0); return GL_FALSE; } if (stream->print_addresses) - PRINTF(stream, "%08x: ", stream->offset); + PRINTF("%08x: ", stream->offset); - PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF("%s (%d dwords):\n", name, len); i915_disassemble_program( stream, ptr, len ); stream->offset += len * sizeof(GLuint); @@ -137,17 +137,17 @@ static GLboolean debug_chain( struct debug_stream *stream, const char *name, GLu GLuint old_offset = stream->offset + len * sizeof(GLuint); GLuint i; - PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF("%s (%d dwords):\n", name, len); for (i = 0; i < len; i++) - PRINTF(stream, "\t\t0x%08x\n", ptr[i]); + PRINTF("\t\t0x%08x\n", ptr[i]); stream->offset = ptr[1] & ~0x3; if (stream->offset < old_offset) - PRINTF(stream, "\n... skipping backwards from 0x%x --> 0x%x ...\n\n", + PRINTF("\n... skipping backwards from 0x%x --> 0x%x ...\n\n", old_offset, stream->offset ); else - PRINTF(stream, "\n... skipping from 0x%x --> 0x%x ...\n\n", + PRINTF("\n... skipping from 0x%x --> 0x%x ...\n\n", old_offset, stream->offset ); @@ -167,10 +167,10 @@ static GLboolean debug_variable_length_prim( struct debug_stream *stream ) len = 1+(i+2)/2; - PRINTF(stream, "3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); + PRINTF("3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); for (i = 0; i < len; i++) - PRINTF(stream, "\t\t0x%08x\n", ptr[i]); - PRINTF(stream, "\n"); + PRINTF("\t\t0x%08x\n", ptr[i]); + PRINTF("\n"); stream->offset += len * sizeof(GLuint); return GL_TRUE; @@ -185,16 +185,16 @@ static GLboolean debug_load_immediate( struct debug_stream *stream, GLuint bits = (ptr[0] >> 4) & 0xff; GLuint i, j = 0; - PRINTF(stream, "%s (%d dwords, flags: %x):\n", name, len, bits); - PRINTF(stream, "\t\t0x%08x\n", ptr[j++]); + PRINTF("%s (%d dwords, flags: %x):\n", name, len, bits); + PRINTF("\t\t0x%08x\n", ptr[j++]); for (i = 0; i < 8; i++) { if (bits & (1<> 8) & 0x3f; GLuint i, j = 0; - PRINTF(stream, "%s (%d dwords):\n", name, len); - PRINTF(stream, "\t\t0x%08x\n", ptr[j++]); + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t\t0x%08x\n", ptr[j++]); for (i = 0; i < 6; i++) { if (bits & (1<winsys->printf( (stream)->winsys, __VA_ARGS__ ) +#define PRINTF( ... ) (stream)->winsys->printf( (stream)->winsys, __VA_ARGS__ ) @@ -132,27 +132,27 @@ print_reg_type_nr(struct debug_stream *stream, GLuint type, GLuint nr) case REG_TYPE_T: switch (nr) { case T_DIFFUSE: - PRINTF(stream, "T_DIFFUSE"); + PRINTF("T_DIFFUSE"); return; case T_SPECULAR: - PRINTF(stream, "T_SPECULAR"); + PRINTF("T_SPECULAR"); return; case T_FOG_W: - PRINTF(stream, "T_FOG_W"); + PRINTF("T_FOG_W"); return; default: - PRINTF(stream, "T_TEX%d", nr); + PRINTF("T_TEX%d", nr); return; } case REG_TYPE_OC: if (nr == 0) { - PRINTF(stream, "oC"); + PRINTF("oC"); return; } break; case REG_TYPE_OD: if (nr == 0) { - PRINTF(stream, "oD"); + PRINTF("oD"); return; } break; @@ -160,7 +160,7 @@ print_reg_type_nr(struct debug_stream *stream, GLuint type, GLuint nr) break; } - PRINTF(stream, "%s[%d]", regname[type], nr); + PRINTF("%s[%d]", regname[type], nr); } #define REG_SWIZZLE_MASK 0x7777 @@ -181,33 +181,33 @@ print_reg_neg_swizzle(struct debug_stream *stream, GLuint reg) (reg & REG_NEGATE_MASK) == 0) return; - PRINTF(stream, "."); + PRINTF("."); for (i = 3; i >= 0; i--) { if (reg & (1 << ((i * 4) + 3))) - PRINTF(stream, "-"); + PRINTF("-"); switch ((reg >> (i * 4)) & 0x7) { case 0: - PRINTF(stream, "x"); + PRINTF("x"); break; case 1: - PRINTF(stream, "y"); + PRINTF("y"); break; case 2: - PRINTF(stream, "z"); + PRINTF("z"); break; case 3: - PRINTF(stream, "w"); + PRINTF("w"); break; case 4: - PRINTF(stream, "0"); + PRINTF("0"); break; case 5: - PRINTF(stream, "1"); + PRINTF("1"); break; default: - PRINTF(stream, "?"); + PRINTF("?"); break; } } @@ -232,15 +232,15 @@ print_dest_reg(struct debug_stream *stream, GLuint dword) print_reg_type_nr(stream, type, nr); if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) return; - PRINTF(stream, "."); + PRINTF("."); if (dword & A0_DEST_CHANNEL_X) - PRINTF(stream, "x"); + PRINTF("x"); if (dword & A0_DEST_CHANNEL_Y) - PRINTF(stream, "y"); + PRINTF("y"); if (dword & A0_DEST_CHANNEL_Z) - PRINTF(stream, "z"); + PRINTF("z"); if (dword & A0_DEST_CHANNEL_W) - PRINTF(stream, "w"); + PRINTF("w"); } @@ -256,29 +256,29 @@ print_arith_op(struct debug_stream *stream, if (opcode != A0_NOP) { print_dest_reg(stream, program[0]); if (program[0] & A0_DEST_SATURATE) - PRINTF(stream, " = SATURATE "); + PRINTF(" = SATURATE "); else - PRINTF(stream, " = "); + PRINTF(" = "); } - PRINTF(stream, "%s ", opcodes[opcode]); + PRINTF("%s ", opcodes[opcode]); print_src_reg(stream, GET_SRC0_REG(program[0], program[1])); if (args[opcode] == 1) { - PRINTF(stream, "\n"); + PRINTF("\n"); return; } - PRINTF(stream, ", "); + PRINTF(", "); print_src_reg(stream, GET_SRC1_REG(program[1], program[2])); if (args[opcode] == 2) { - PRINTF(stream, "\n"); + PRINTF("\n"); return; } - PRINTF(stream, ", "); + PRINTF(", "); print_src_reg(stream, GET_SRC2_REG(program[2])); - PRINTF(stream, "\n"); + PRINTF("\n"); return; } @@ -288,27 +288,27 @@ print_tex_op(struct debug_stream *stream, GLuint opcode, const GLuint * program) { print_dest_reg(stream, program[0] | A0_DEST_CHANNEL_ALL); - PRINTF(stream, " = "); + PRINTF(" = "); - PRINTF(stream, "%s ", opcodes[opcode]); + PRINTF("%s ", opcodes[opcode]); - PRINTF(stream, "S[%d],", program[0] & T0_SAMPLER_NR_MASK); + PRINTF("S[%d],", program[0] & T0_SAMPLER_NR_MASK); print_reg_type_nr(stream, (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & REG_TYPE_MASK, (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); - PRINTF(stream, "\n"); + PRINTF("\n"); } static void print_dcl_op(struct debug_stream *stream, GLuint opcode, const GLuint * program) { - PRINTF(stream, "%s ", opcodes[opcode]); + PRINTF("%s ", opcodes[opcode]); print_dest_reg(stream, program[0] | A0_DEST_CHANNEL_ALL); - PRINTF(stream, "\n"); + PRINTF("\n"); } @@ -319,7 +319,7 @@ i915_disassemble_program(struct debug_stream *stream, GLuint size = program[0] & 0x1ff; GLint i; - PRINTF(stream, "\t\tBEGIN\n"); + PRINTF("\t\tBEGIN\n"); assert(size + 2 == sz); @@ -327,7 +327,7 @@ i915_disassemble_program(struct debug_stream *stream, for (i = 1; i < sz; i += 3, program += 3) { GLuint opcode = program[0] & (0x1f << 24); - PRINTF(stream, "\t\t"); + PRINTF("\t\t"); if ((GLint) opcode >= A0_NOP && opcode <= A0_SLT) print_arith_op(stream, opcode >> 24, program); @@ -336,10 +336,10 @@ i915_disassemble_program(struct debug_stream *stream, else if (opcode == D0_DCL) print_dcl_op(stream, opcode >> 24, program); else - PRINTF(stream, "Unknown opcode 0x%x\n", opcode); + PRINTF("Unknown opcode 0x%x\n", opcode); } - PRINTF(stream, "\t\tEND\n\n"); + PRINTF("\t\tEND\n\n"); } -- cgit v1.2.3 From a0e48b1ff9884ed66a156bd2d11b475ff177f166 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 11 Aug 2007 14:00:19 +0100 Subject: Pull in improved debug from master --- src/mesa/pipe/i915simple/i915_debug.c | 452 ++++++++++++++++++++++++++++++++-- 1 file changed, 429 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index 10d7817816..89257e9409 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -41,7 +41,7 @@ static GLboolean debug( struct debug_stream *stream, const char *name, GLuint le GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); if (len == 0) { - PRINTF( "Error - zero length packet (0x%08x)\n", stream->ptr[0] ); + PRINTF("Error - zero length packet (0x%08x)\n", stream->ptr[0]); assert(0); return GL_FALSE; } @@ -52,7 +52,7 @@ static GLboolean debug( struct debug_stream *stream, const char *name, GLuint le PRINTF("%s (%d dwords):\n", name, len); for (i = 0; i < len; i++) - PRINTF("\t\t0x%08x\n", ptr[i]); + PRINTF("\t0x%08x\n", ptr[i]); PRINTF("\n"); stream->offset += len * sizeof(GLuint); @@ -91,12 +91,12 @@ static GLboolean debug_prim( struct debug_stream *stream, const char *name, PRINTF("%s %s (%d dwords):\n", name, prim, len); - PRINTF("\t\t0x%08x\n", ptr[0]); + PRINTF("\t0x%08x\n", ptr[0]); for (i = 1; i < len; i++) { if (dump_floats) - PRINTF("\t\t0x%08x // %f\n", ptr[i], *(GLfloat *)&ptr[i]); + PRINTF("\t0x%08x // %f\n", ptr[i], *(GLfloat *)&ptr[i]); else - PRINTF("\t\t0x%08x\n", ptr[i]); + PRINTF("\t0x%08x\n", ptr[i]); } @@ -139,7 +139,7 @@ static GLboolean debug_chain( struct debug_stream *stream, const char *name, GLu PRINTF("%s (%d dwords):\n", name, len); for (i = 0; i < len; i++) - PRINTF("\t\t0x%08x\n", ptr[i]); + PRINTF("\t0x%08x\n", ptr[i]); stream->offset = ptr[1] & ~0x3; @@ -169,7 +169,7 @@ static GLboolean debug_variable_length_prim( struct debug_stream *stream ) PRINTF("3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); for (i = 0; i < len; i++) - PRINTF("\t\t0x%08x\n", ptr[i]); + PRINTF("\t0x%08x\n", ptr[i]); PRINTF("\n"); stream->offset += len * sizeof(GLuint); @@ -177,22 +177,126 @@ static GLboolean debug_variable_length_prim( struct debug_stream *stream ) } +#define BITS( dw, hi, lo, ... ) \ +do { \ + unsigned himask = ~0UL >> (31 - (hi)); \ + PRINTF("\t\t "); \ + PRINTF(__VA_ARGS__); \ + PRINTF(": 0x%x\n", ((dw) & himask) >> (lo)); \ +} while (0) + +#define MBZ( dw, hi, lo) do { \ + unsigned x = (dw) >> (lo); \ + unsigned lomask = (1 << (lo)) - 1; \ + unsigned himask; \ + himask = (1UL << (hi)) - 1; \ + assert ((x & himask & ~lomask) == 0); \ +} while (0) + +#define FLAG( dw, bit, ... ) \ +do { \ + if (((dw) >> (bit)) & 1) { \ + PRINTF("\t\t "); \ + PRINTF(__VA_ARGS__); \ + PRINTF("\n"); \ + } \ +} while (0) + static GLboolean debug_load_immediate( struct debug_stream *stream, const char *name, GLuint len ) { GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); GLuint bits = (ptr[0] >> 4) & 0xff; - GLuint i, j = 0; + GLuint j = 0; PRINTF("%s (%d dwords, flags: %x):\n", name, len, bits); - PRINTF("\t\t0x%08x\n", ptr[j++]); + PRINTF("\t0x%08x\n", ptr[j++]); - for (i = 0; i < 8; i++) { - if (bits & (1<> (i * 4)) & 0xf; + if (tc != 0xf) + BITS(tc, 3, 0, "tex coord %d", i); } + j++; + } + if (bits & (1<<3)) { + PRINTF("\t LIS3: 0x%08x\n", ptr[j]); + j++; + } + if (bits & (1<<4)) { + PRINTF("\t LIS4: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 23, "point width"); + BITS(ptr[j], 22, 19, "line width"); + FLAG(ptr[j], 18, "alpha flatshade"); + FLAG(ptr[j], 17, "fog flatshade"); + FLAG(ptr[j], 16, "spec flatshade"); + FLAG(ptr[j], 15, "rgb flatshade"); + BITS(ptr[j], 14, 13, "cull mode"); + FLAG(ptr[j], 12, "vfmt: point width"); + FLAG(ptr[j], 11, "vfmt: specular/fog"); + FLAG(ptr[j], 10, "vfmt: rgba"); + FLAG(ptr[j], 9, "vfmt: depth offset"); + BITS(ptr[j], 8, 6, "vfmt: position (2==xyzw)"); + FLAG(ptr[j], 5, "force dflt diffuse"); + FLAG(ptr[j], 4, "force dflt specular"); + FLAG(ptr[j], 3, "local depth offset enable"); + FLAG(ptr[j], 2, "vfmt: fp32 fog coord"); + FLAG(ptr[j], 1, "sprite point"); + FLAG(ptr[j], 0, "antialiasing"); + j++; + } + if (bits & (1<<5)) { + PRINTF("\t LIS5: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 28, "rgba write disables"); + FLAG(ptr[j], 27, "force dflt point width"); + FLAG(ptr[j], 26, "last pixel enable"); + FLAG(ptr[j], 25, "global z offset enable"); + FLAG(ptr[j], 24, "fog enable"); + BITS(ptr[j], 23, 16, "stencil ref"); + BITS(ptr[j], 15, 13, "stencil test"); + BITS(ptr[j], 12, 10, "stencil fail op"); + BITS(ptr[j], 9, 7, "stencil pass z fail op"); + BITS(ptr[j], 6, 4, "stencil pass z pass op"); + FLAG(ptr[j], 3, "stencil write enable"); + FLAG(ptr[j], 2, "stencil test enable"); + FLAG(ptr[j], 1, "color dither enable"); + FLAG(ptr[j], 0, "logiop enable"); + j++; } + if (bits & (1<<6)) { + PRINTF("\t LIS6: 0x%08x\n", ptr[j]); + FLAG(ptr[j], 31, "alpha test enable"); + BITS(ptr[j], 30, 28, "alpha func"); + BITS(ptr[j], 27, 20, "alpha ref"); + FLAG(ptr[j], 19, "depth test enable"); + BITS(ptr[j], 18, 16, "depth func"); + FLAG(ptr[j], 15, "blend enable"); + BITS(ptr[j], 14, 12, "blend func"); + BITS(ptr[j], 11, 8, "blend src factor"); + BITS(ptr[j], 7, 4, "blend dst factor"); + FLAG(ptr[j], 3, "depth write enable"); + FLAG(ptr[j], 2, "color write enable"); + BITS(ptr[j], 1, 0, "provoking vertex"); + j++; + } + PRINTF("\n"); @@ -214,7 +318,7 @@ static GLboolean debug_load_indirect( struct debug_stream *stream, GLuint i, j = 0; PRINTF("%s (%d dwords):\n", name, len); - PRINTF("\t\t0x%08x\n", ptr[j++]); + PRINTF("\t0x%08x\n", ptr[j++]); for (i = 0; i < 6; i++) { if (bits & (1<ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + BR13(stream, ptr[j++]); + BR22(stream, ptr[j++]); + BR23(stream, ptr[j++]); + BR09(stream, ptr[j++]); + BR26(stream, ptr[j++]); + BR11(stream, ptr[j++]); + BR12(stream, ptr[j++]); + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_color_blit( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + BR13(stream, ptr[j++]); + BR22(stream, ptr[j++]); + BR23(stream, ptr[j++]); + BR09(stream, ptr[j++]); + BR16(stream, ptr[j++]); + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_modes4( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j]); + BITS(ptr[j], 21, 18, "logicop func"); + FLAG(ptr[j], 17, "stencil test mask modify-enable"); + FLAG(ptr[j], 16, "stencil write mask modify-enable"); + BITS(ptr[j], 15, 8, "stencil test mask"); + BITS(ptr[j], 7, 0, "stencil write mask"); + j++; + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_map_state( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + { + PRINTF("\t0x%08x\n", ptr[j]); + BITS(ptr[j], 15, 0, "map mask"); + j++; + } + + while (j < len) { + { + PRINTF("\t TMn.0: 0x%08x\n", ptr[j]); + PRINTF("\t map address: 0x%08x\n", (ptr[j] & ~0x3)); + FLAG(ptr[j], 1, "vertical line stride"); + FLAG(ptr[j], 0, "vertical line stride offset"); + j++; + } + + { + PRINTF("\t TMn.1: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 21, "height"); + BITS(ptr[j], 20, 10, "width"); + BITS(ptr[j], 9, 7, "surface format"); + BITS(ptr[j], 6, 3, "texel format"); + FLAG(ptr[j], 2, "use fence regs"); + FLAG(ptr[j], 1, "tiled surface"); + FLAG(ptr[j], 0, "tile walk ymajor"); + j++; + } + { + PRINTF("\t TMn.2: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 21, "dword pitch"); + BITS(ptr[j], 20, 15, "cube face enables"); + BITS(ptr[j], 14, 9, "max lod"); + FLAG(ptr[j], 8, "mip layout right"); + BITS(ptr[j], 7, 0, "depth"); + j++; + } + } + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_sampler_state( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + { + PRINTF("\t0x%08x\n", ptr[j]); + BITS(ptr[j], 15, 0, "sampler mask"); + j++; + } + + while (j < len) { + { + PRINTF("\t TSn.0: 0x%08x\n", ptr[j]); + FLAG(ptr[j], 31, "reverse gamma"); + FLAG(ptr[j], 30, "planar to packed"); + FLAG(ptr[j], 29, "yuv->rgb"); + BITS(ptr[j], 28, 27, "chromakey index"); + BITS(ptr[j], 26, 22, "base mip level"); + BITS(ptr[j], 21, 20, "mip mode filter"); + BITS(ptr[j], 19, 17, "mag mode filter"); + BITS(ptr[j], 16, 14, "min mode filter"); + BITS(ptr[j], 13, 5, "lod bias (s4.4)"); + FLAG(ptr[j], 4, "shadow enable"); + FLAG(ptr[j], 3, "max-aniso-4"); + BITS(ptr[j], 2, 0, "shadow func"); + j++; + } + + { + PRINTF("\t TSn.1: 0x%08x\n", ptr[j]); + BITS(ptr[j], 31, 24, "min lod"); + MBZ( ptr[j], 23, 18 ); + FLAG(ptr[j], 17, "kill pixel enable"); + FLAG(ptr[j], 16, "keyed tex filter mode"); + FLAG(ptr[j], 15, "chromakey enable"); + BITS(ptr[j], 14, 12, "tcx wrap mode"); + BITS(ptr[j], 11, 9, "tcy wrap mode"); + BITS(ptr[j], 8, 6, "tcz wrap mode"); + FLAG(ptr[j], 5, "normalized coords"); + BITS(ptr[j], 4, 1, "map (surface) index"); + FLAG(ptr[j], 0, "EAST deinterlacer enable"); + j++; + } + { + PRINTF("\t TSn.2: 0x%08x (default color)\n", ptr[j]); + j++; + } + } + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_dest_vars( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + { + PRINTF("\t0x%08x\n", ptr[j]); + FLAG(ptr[j], 31, "early classic ztest"); + FLAG(ptr[j], 30, "opengl tex default color"); + FLAG(ptr[j], 29, "bypass iz"); + FLAG(ptr[j], 28, "lod preclamp"); + BITS(ptr[j], 27, 26, "dither pattern"); + FLAG(ptr[j], 25, "linear gamma blend"); + FLAG(ptr[j], 24, "debug dither"); + BITS(ptr[j], 23, 20, "dstorg x"); + BITS(ptr[j], 19, 16, "dstorg y"); + MBZ (ptr[j], 15, 15 ); + BITS(ptr[j], 14, 12, "422 write select"); + BITS(ptr[j], 11, 8, "cbuf format"); + BITS(ptr[j], 3, 2, "zbuf format"); + FLAG(ptr[j], 1, "vert line stride"); + FLAG(ptr[j], 1, "vert line stride offset"); + j++; + } + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} + +static GLboolean debug_buf_info( struct debug_stream *stream, + const char *name, + GLuint len ) +{ + GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF("%s (%d dwords):\n", name, len); + PRINTF("\t0x%08x\n", ptr[j++]); + + { + PRINTF("\t0x%08x\n", ptr[j]); + BITS(ptr[j], 28, 28, "aux buffer id"); + BITS(ptr[j], 27, 24, "buffer id (7=depth, 3=back)"); + FLAG(ptr[j], 23, "use fence regs"); + FLAG(ptr[j], 22, "tiled surface"); + FLAG(ptr[j], 21, "tile walk ymajor"); + MBZ (ptr[j], 20, 14); + BITS(ptr[j], 13, 2, "dword pitch"); + MBZ (ptr[j], 2, 0); + j++; + } + + PRINTF("\t0x%08x -- buffer base address\n", ptr[j++]); + + stream->offset += len * sizeof(GLuint); + assert(j == len); + return GL_TRUE; +} static GLboolean i915_debug_packet( struct debug_stream *stream ) { @@ -294,9 +699,9 @@ static GLboolean i915_debug_packet( struct debug_stream *stream ) case 0x2: switch ((cmd >> 22) & 0xff) { case 0x50: - return debug(stream, "XY_COLOR_BLT", (cmd & 0xff) + 2); + return debug_color_blit(stream, "XY_COLOR_BLT", (cmd & 0xff) + 2); case 0x53: - return debug(stream, "XY_SRC_COPY_BLT", (cmd & 0xff) + 2); + return debug_copy_blit(stream, "XY_SRC_COPY_BLT", (cmd & 0xff) + 2); default: return debug(stream, "blit command", (cmd & 0xff) + 2); } @@ -316,7 +721,7 @@ static GLboolean i915_debug_packet( struct debug_stream *stream ) case 0xc: return debug(stream, "3DSTATE_MODES5", 1); case 0xd: - return debug(stream, "3DSTATE_MODES4", 1); + return debug_modes4(stream, "3DSTATE_MODES4", 1); case 0x15: return debug(stream, "3DSTATE_FOG_COLOR", 1); case 0x16: @@ -336,9 +741,9 @@ static GLboolean i915_debug_packet( struct debug_stream *stream ) /* 3DStateMW */ switch ((cmd >> 16) & 0xff) { case 0x0: - return debug(stream, "3DSTATE_MAP_STATE", (cmd & 0x1f) + 2); + return debug_map_state(stream, "3DSTATE_MAP_STATE", (cmd & 0x1f) + 2); case 0x1: - return debug(stream, "3DSTATE_SAMPLER_STATE", (cmd & 0x1f) + 2); + return debug_sampler_state(stream, "3DSTATE_SAMPLER_STATE", (cmd & 0x1f) + 2); case 0x4: return debug_load_immediate(stream, "3DSTATE_LOAD_STATE_IMMEDIATE", (cmd & 0xf) + 2); case 0x5: @@ -354,13 +759,13 @@ static GLboolean i915_debug_packet( struct debug_stream *stream ) case 0x83: return debug(stream, "3DSTATE_SPAN_STIPPLE", (cmd & 0xffff) + 2); case 0x85: - return debug(stream, "3DSTATE_DEST_BUFFER_VARS", (cmd & 0xffff) + 2); + return debug_dest_vars(stream, "3DSTATE_DEST_BUFFER_VARS", (cmd & 0xffff) + 2); case 0x88: return debug(stream, "3DSTATE_CONSTANT_BLEND_COLOR", (cmd & 0xffff) + 2); case 0x89: return debug(stream, "3DSTATE_FOG_MODE", (cmd & 0xffff) + 2); case 0x8e: - return debug(stream, "3DSTATE_BUFFER_INFO", (cmd & 0xffff) + 2); + return debug_buf_info(stream, "3DSTATE_BUFFER_INFO", (cmd & 0xffff) + 2); case 0x97: return debug(stream, "3DSTATE_DEPTH_OFFSET_SCALE", (cmd & 0xffff) + 2); case 0x98: @@ -418,13 +823,14 @@ i915_dump_batchbuffer( struct i915_context *i915, GLuint bytes = (end - start) * 4; GLboolean done = GL_FALSE; - fprintf(stderr, "\n\nBATCH: (%d)\n", bytes / 4); stream.offset = 0; stream.ptr = (char *)start; stream.print_addresses = 0; stream.winsys = i915->pipe.winsys; + stream.winsys->printf( stream.winsys, "\n\nBATCH: (%d)\n", bytes / 4); + while (!done && stream.offset < bytes && stream.offset >= 0) @@ -436,7 +842,7 @@ i915_dump_batchbuffer( struct i915_context *i915, stream.offset >= 0); } - fprintf(stderr, "END-BATCH\n\n\n"); + stream.winsys->printf( stream.winsys, "END-BATCH\n\n\n"); } -- cgit v1.2.3 From 8c4409690d30b5fc5c048b90351c14522119fa9c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 11 Aug 2007 14:01:37 +0100 Subject: Fix some build failures. Move intelScreenContext to intel_context.c --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 5 ++- src/mesa/drivers/dri/intel_winsys/intel_context.c | 38 +++++++++++++++++----- src/mesa/drivers/dri/intel_winsys/intel_context.h | 2 +- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 5 +-- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 24 -------------- 5 files changed, 38 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 4d0759fc32..6fe9ecc539 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -25,8 +25,11 @@ * **************************************************************************/ +#include +#include "main/imports.c" #include "intel_batchbuffer.h" -#include "intel_ioctl.h" +#include "intel_context.h" +#include "intel_screen.h" #include "intel_reg.h" #include "drm.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index af5deb4dc6..435d17421c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -36,29 +36,28 @@ #include "drivers/common/driverfuncs.h" -#include "intel_screen.h" - #include "i830_dri.h" +#include "intel_screen.h" +#include "intel_context.h" #include "intel_buffers.h" #include "intel_winsys.h" -#include "intel_ioctl.h" #include "intel_batchbuffer.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" #include "pipe/p_defines.h" +#include "pipe/p_context.h" #include "drirenderbuffer.h" #include "vblank.h" #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ -#include "pipe/p_context.h" -#if DEBUG +#ifdef DEBUG int __intel_debug = 0; #endif @@ -145,7 +144,7 @@ const struct dri_extension card_extensions[] = { - +#ifdef DEBUG static const struct dri_debug_control debug_control[] = { {"ioctl", DEBUG_IOCTL}, {"bat", DEBUG_BATCH}, @@ -153,7 +152,7 @@ static const struct dri_debug_control debug_control[] = { {"swap", DEBUG_SWAP}, {NULL, 0} }; - +#endif static void intelInvalidateState(GLcontext * ctx, GLuint new_state) @@ -279,7 +278,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); } -#if DEBUG +#ifdef DEBUG __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); #endif @@ -442,3 +441,26 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, +struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) +{ + /* + * This should probably change to have the screen allocate a dummy + * context at screen creation. For now just use the current context. + */ + + GET_CURRENT_CONTEXT(ctx); + if (ctx == NULL) { +/* _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; */ + /* need a context for the first time makecurrent is called (for hw lock + when allocating priv buffers) */ + if (intelScreen->dummyctxptr == NULL) { + _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; + } + return intelScreen->dummyctxptr; + } + + return intel_context(ctx); +} + diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 924b7ca05a..76c900157a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -109,7 +109,7 @@ extern char *__progname; /* ================================================================ * Debugging: */ -#if DEBUG +#ifdef DEBUG extern int __intel_debug; #define DEBUG_SWAP 0x1 diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 34319545d9..e9e4bc0645 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -31,11 +31,12 @@ #include "extensions.h" #include "intel_screen.h" +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_buffers.h" #include "i830_dri.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 90a432a041..cf8d190d34 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -38,7 +38,6 @@ #include "intel_screen.h" #include "intel_batchbuffer.h" #include "intel_buffers.h" -#include "intel_ioctl.h" #include "i830_dri.h" #include "dri_bufpool.h" @@ -565,26 +564,3 @@ __driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, return (void *) psp; } -struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) -{ - /* - * This should probably change to have the screen allocate a dummy - * context at screen creation. For now just use the current context. - */ - - GET_CURRENT_CONTEXT(ctx); - if (ctx == NULL) { -/* _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; */ - /* need a context for the first time makecurrent is called (for hw lock - when allocating priv buffers) */ - if (intelScreen->dummyctxptr == NULL) { - _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; - } - return intelScreen->dummyctxptr; - } - return intel_context(ctx); - -} - -- cgit v1.2.3 From 91c2216165e87c7dee665808ac175b186c7c5c9b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 11 Aug 2007 14:19:48 +0100 Subject: Rename intel_buffers.[ch] to intel_swapbuffers.[ch] --- src/mesa/drivers/dri/intel_winsys/Makefile | 2 +- src/mesa/drivers/dri/intel_winsys/intel_buffers.c | 634 --------------------- src/mesa/drivers/dri/intel_winsys/intel_buffers.h | 74 --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 10 +- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 2 +- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 2 +- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 634 +++++++++++++++++++++ .../drivers/dri/intel_winsys/intel_swapbuffers.h | 74 +++ .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 2 +- 9 files changed, 713 insertions(+), 721 deletions(-) delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_buffers.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_buffers.h create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index 7c37812172..46b9541d7d 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -15,7 +15,7 @@ DRIVER_SOURCES = \ intel_winsys_softpipe.c \ intel_winsys_i915.c \ intel_batchbuffer.c \ - intel_buffers.c \ + intel_swapbuffers.c \ intel_context.c \ intel_lock.c \ intel_screen.c \ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c b/src/mesa/drivers/dri/intel_winsys/intel_buffers.c deleted file mode 100644 index 0823c5c8ea..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.c +++ /dev/null @@ -1,634 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_buffers.h" -#include "intel_batchbuffer.h" -#include "intel_reg.h" -#include "context.h" -#include "utils.h" -#include "drirenderbuffer.h" -#include "vblank.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_cb_fbo.h" - - -/* This block can be removed when libdrm >= 2.3.1 is required */ - -#ifndef DRM_VBLANK_FLIP - -#define DRM_VBLANK_FLIP 0x8000000 - -typedef struct drm_i915_flip { - int pipes; -} drm_i915_flip_t; - -#undef DRM_IOCTL_I915_FLIP -#define DRM_IOCTL_I915_FLIP DRM_IOW(DRM_COMMAND_BASE + DRM_I915_FLIP, \ - drm_i915_flip_t) - -#endif - - - - -/** - * Copy the back color buffer to the front color buffer. - * Used for SwapBuffers(). - */ -void -intelCopyBuffer(__DRIdrawablePrivate * dPriv, - const drm_clip_rect_t * rect) -{ - - struct intel_context *intel; - const intelScreenPrivate *intelScreen; - - DBG(SWAP, "%s\n", __FUNCTION__); - - assert(dPriv); - - intel = intelScreenContext(dPriv->driScreenPriv->private); - if (!intel) - return; - - intelScreen = intel->intelScreen; - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - current context (which is what intelScreenContext should return) might - not get a contended lock and thus cliprects not updated (tests/manywin) */ - if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - - if (dPriv && dPriv->numClipRects) { - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; -#if 0 - const struct pipe_region *backRegion - = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? - intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : - intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); -#endif - const int backWidth = intel_fb->Base.Width; - const int backHeight = intel_fb->Base.Height; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; -#if 0 - const int srcpitch = backRegion->pitch; -#endif - const int cpp = intelScreen->front.cpp; - int BR13, CMD; - int i; - - const struct pipe_surface *backSurf; - const struct pipe_region *backRegion; - int srcpitch; - struct st_renderbuffer *strb; - - /* blit from back color buffer if it exists, else front buffer */ - strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); - if (strb) { - backSurf = strb->surface; - } - else { - strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); - backSurf = strb->surface; - } - - backRegion = backSurf->region; - srcpitch = backRegion->pitch; - - ASSERT(intel_fb); - ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ - ASSERT(backRegion); - ASSERT(backRegion->cpp == cpp); - - DBG(SWAP, "front pitch %d back pitch %d\n", - pitch, backRegion->pitch); - - if (cpp == 2) { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - } - else { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - } - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) - continue; - - box = *pbox; - - if (rect) { - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > backWidth) - box.x2 = backWidth + box.x1; - if (box.y2 - box.y1 > backHeight) - box.y2 = backHeight + box.y1; - - DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((box.y1 << 16) | box.x1); - OUT_BATCH((box.y2 << 16) | box.x2); - - OUT_RELOC(intelScreen->front.buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((sbox.y1 << 16) | sbox.x1); - OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - ADVANCE_BATCH(); - } - - if (intel->first_swap_fence) - driFenceUnReference(intel->first_swap_fence); - intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); - driFenceReference(intel->first_swap_fence); - } - - UNLOCK_HARDWARE(intel); - - /* XXX this is bogus. The context here may not even be bound to this drawable! */ - if (intel->lastStamp != dPriv->lastStamp) { - GET_CURRENT_CONTEXT(currctx); - struct intel_context *intelcurrent = intel_context(currctx); - if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { - intelWindowMoved(intel); - intel->lastStamp = dPriv->lastStamp; - } - } - -} - - - -/** - * This will be called whenever the currently bound window is moved/resized. - * XXX: actually, it seems to NOT be called when the window is only moved (BP). - */ -void -intelWindowMoved(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - - if (!intel->ctx.DrawBuffer) { - /* when would this happen? -BP */ - assert(0); - intel->numClipRects = 0; - } - - /* Update Mesa's notion of window size */ - driUpdateFramebufferSize(ctx, dPriv); - intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ - - { - drmI830Sarea *sarea = intel->sarea; - drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, - .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; - drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, .y1 = sarea->pipeA_y, - .x2 = sarea->pipeA_x + sarea->pipeA_w, - .y2 = sarea->pipeA_y + sarea->pipeA_h }; - drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, .y1 = sarea->pipeB_y, - .x2 = sarea->pipeB_x + sarea->pipeB_w, - .y2 = sarea->pipeB_y + sarea->pipeB_h }; - GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); - GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); - GLuint flags = intel_fb->vblank_flags; - GLboolean pf_active; - GLint pf_pipes; - - /* Update page flipping info - */ - pf_pipes = 0; - - if (areaA > 0) - pf_pipes |= 1; - - if (areaB > 0) - pf_pipes |= 2; - - intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3; - - intel_fb->pf_num_pages = 2 /*intel->intelScreen->third.handle ? 3 : 2*/; - - pf_active = pf_pipes && (pf_pipes & intel->sarea->pf_active) == pf_pipes; - - if (pf_active != intel_fb->pf_active) - DBG(LOCK, "%s - Page flipping %sactive\n", - __progname, pf_active ? "" : "in"); - - if (pf_active) { - /* Sync pages between pipes if we're flipping on both at the same time */ - if (pf_pipes == 0x3 && pf_pipes != intel_fb->pf_pipes && - (intel->sarea->pf_current_page & 0x3) != - (((intel->sarea->pf_current_page) >> 2) & 0x3)) { - drm_i915_flip_t flip; - - if (intel_fb->pf_current_page == - (intel->sarea->pf_current_page & 0x3)) { - /* XXX: This is ugly, but emitting two flips 'in a row' can cause - * lockups for unknown reasons. - */ - intel->sarea->pf_current_page = - intel->sarea->pf_current_page & 0x3; - intel->sarea->pf_current_page |= - ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % - intel_fb->pf_num_pages) << 2; - - flip.pipes = 0x2; - } else { - intel->sarea->pf_current_page = - intel->sarea->pf_current_page & (0x3 << 2); - intel->sarea->pf_current_page |= - (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % - intel_fb->pf_num_pages; - - flip.pipes = 0x1; - } - - drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); - } - - intel_fb->pf_pipes = pf_pipes; - } - - intel_fb->pf_active = pf_active; -#if 0 - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); -#endif - - /* Update vblank info - */ - if (areaB > areaA || (areaA == areaB && areaB > 0)) { - flags = intel_fb->vblank_flags | VBLANK_FLAG_SECONDARY; - } else { - flags = intel_fb->vblank_flags & ~VBLANK_FLAG_SECONDARY; - } - - if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && - !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { - drmVBlank vbl; - int i; - - vbl.request.type = DRM_VBLANK_ABSOLUTE; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - vbl.request.type |= DRM_VBLANK_SECONDARY; - } - - for (i = 0; i < intel_fb->pf_num_pages; i++) { - if ((intel_fb->vbl_waited - intel_fb->vbl_pending[i]) <= (1<<23)) - continue; - - vbl.request.sequence = intel_fb->vbl_pending[i]; - drmWaitVBlank(intel->driFd, &vbl); - } - - intel_fb->vblank_flags = flags; - driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); - intel_fb->vbl_waited = intel_fb->vbl_seq; - - for (i = 0; i < intel_fb->pf_num_pages; i++) { - intel_fb->vbl_pending[i] = intel_fb->vbl_waited; - } - } - } - - /* This will be picked up by looking at the dirty state flags: - */ - - /* Update hardware scissor */ -// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, -// ctx->Scissor.Width, ctx->Scissor.Height); - - /* Re-calculate viewport related state */ -// ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far ); -} - - - - - -/* Emit wait for pending flips */ -#if 0 -void -intel_wait_flips(struct intel_context *intel, GLuint batch_flags) -{ - struct intel_framebuffer *intel_fb = - (struct intel_framebuffer *) intel->ctx.DrawBuffer; - struct intel_renderbuffer *intel_rb = - intel_get_renderbuffer(&intel_fb->Base, - intel_fb->Base._ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : - BUFFER_BACK_LEFT); - - if (intel_fb->Base.Name == 0 && intel_rb->pf_pending == intel_fb->pf_seq) { - GLint pf_pipes = intel_fb->pf_pipes; - BATCH_LOCALS; - - /* Wait for pending flips to take effect */ - BEGIN_BATCH(2, batch_flags); - OUT_BATCH(pf_pipes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) - : 0); - OUT_BATCH(pf_pipes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) - : 0); - ADVANCE_BATCH(); - - intel_rb->pf_pending--; - } -} -#endif - -#if 0 -/* Flip the front & back buffers - */ -static GLboolean -intelPageFlip(const __DRIdrawablePrivate * dPriv) -{ - struct intel_context *intel; - int ret; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - - DBG(SWAP, "%s\n", __FUNCTION__); - - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate; - - if (intel->intelScreen->drmMinor < 9) - return GL_FALSE; - - intelFlush(&intel->ctx); - - ret = 0; - - LOCK_HARDWARE(intel); - - if (dPriv->numClipRects && intel_fb->pf_active) { - drm_i915_flip_t flip; - - flip.pipes = intel_fb->pf_pipes; - - ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); - } - - UNLOCK_HARDWARE(intel); - - if (ret || !intel_fb->pf_active) - return GL_FALSE; - - if (!dPriv->numClipRects) { - usleep(10000); /* throttle invisible client 10ms */ - } - - intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3; - - if (dPriv->numClipRects != 0) { - intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending = - intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->pf_pending = - ++intel_fb->pf_seq; - } - -#if 0 - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, &intel_fb->Base); -#endif - - DBG(SWAP, "%s: success\n", __FUNCTION__); - - return GL_TRUE; -} -#endif - - -static GLboolean -intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) -{ - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); - struct intel_context *intel = - intelScreenContext(dPriv->driScreenPriv->private); - const intelScreenPrivate *intelScreen = intel->intelScreen; - unsigned int target; - drm_i915_vblank_swap_t swap; - GLboolean ret; - - if (!intel_fb->vblank_flags || - (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || - intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) - return GL_FALSE; - - swap.seqtype = DRM_VBLANK_ABSOLUTE; - - if (intel_fb->vblank_flags & VBLANK_FLAG_SYNC) { - swap.seqtype |= DRM_VBLANK_NEXTONMISS; - } else if (interval == 0) { - return GL_FALSE; - } - - swap.drawable = dPriv->hHWDrawable; - target = swap.sequence = intel_fb->vbl_seq + interval; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - swap.seqtype |= DRM_VBLANK_SECONDARY; - } - - LOCK_HARDWARE(intel); - - intel_batchbuffer_flush(intel->batch); - - if ( intel_fb->pf_active ) { - swap.seqtype |= DRM_VBLANK_FLIP; - - intel_fb->pf_current_page = (((intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3) + 1) % - intel_fb->pf_num_pages; - } - - if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, - sizeof(swap))) { - intel_fb->vbl_seq = swap.sequence; - swap.sequence -= target; - *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); - -#if 1 - intel_fb->vbl_pending[1] = intel_fb->vbl_pending[0] = intel_fb->vbl_seq; -#else - intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending = - intel_get_renderbuffer(&intel_fb->Base, - BUFFER_FRONT_LEFT)->vbl_pending = - intel_fb->vbl_seq; -#endif - - if (swap.seqtype & DRM_VBLANK_FLIP) { -#if 0 - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); -#endif - } - - ret = GL_TRUE; - } else { - if (swap.seqtype & DRM_VBLANK_FLIP) { - intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3) % - intel_fb->pf_num_pages; - } - - ret = GL_FALSE; - } - - UNLOCK_HARDWARE(intel); - - return ret; -} - -void -intelSwapBuffers(__DRIdrawablePrivate * dPriv) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel; - - if (ctx == NULL) - return; - - intel = intel_context(ctx); - - if (ctx->Visual.doubleBufferMode) { - GLboolean missed_target; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - int64_t ust; - - _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - - if (!intelScheduleSwap(dPriv, &missed_target)) { - driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, - &missed_target); - - intelCopyBuffer(dPriv, NULL); - } - - intel_fb->swap_count++; - (*dri_interface->getUST) (&ust); - if (missed_target) { - intel_fb->swap_missed_count++; - intel_fb->swap_missed_ust = ust - intel_fb->swap_ust; - } - - intel_fb->swap_ust = ust; - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); - } -} - -void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - struct intel_context *intel = - (struct intel_context *) dPriv->driContextPriv->driverPrivate; - GLcontext *ctx = &intel->ctx; - - if (ctx->Visual.doubleBufferMode) { - drm_clip_rect_t rect; - /* fixup cliprect (driDrawable may have changed?) later */ - rect.x1 = x; - rect.y1 = y; - rect.x2 = w; - rect.y2 = h; - _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - intelCopyBuffer(dPriv, &rect); - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); - } -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_buffers.h b/src/mesa/drivers/dri/intel_winsys/intel_buffers.h deleted file mode 100644 index d12b5ef96e..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_buffers.h +++ /dev/null @@ -1,74 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BUFFERS_H -#define INTEL_BUFFERS_H - - -struct intel_context; -struct intel_framebuffer; - -/** - * Intel framebuffer, derived from gl_framebuffer. - */ -struct intel_framebuffer -{ - struct gl_framebuffer Base; - - /* Drawable page flipping state */ - GLboolean pf_active; - GLuint pf_seq; - GLint pf_pipes; - GLint pf_current_page; - GLint pf_num_pages; - - /* VBI - */ - GLuint vbl_seq; - GLuint vblank_flags; - GLuint vbl_waited; - - int64_t swap_ust; - int64_t swap_missed_ust; - - GLuint swap_count; - GLuint swap_missed_count; - - GLuint vbl_pending[3]; /**< [number of color buffers] */ -}; - - -void intelCopyBuffer(__DRIdrawablePrivate * dPriv, - const drm_clip_rect_t * rect); - -extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); - -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void intelWindowMoved(struct intel_context *intel); - -#endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 435d17421c..71bc5003d2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -40,7 +40,7 @@ #include "intel_screen.h" #include "intel_context.h" -#include "intel_buffers.h" +#include "intel_swapbuffers.h" #include "intel_winsys.h" #include "intel_batchbuffer.h" @@ -82,9 +82,6 @@ int __intel_debug = 0; #define DRIVER_DATE "20070731" -_glthread_Mutex lockMutex; -static GLboolean lockMutexInit = GL_FALSE; - /** @@ -222,11 +219,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->driScreen = sPriv; intel->sarea = saPriv; - if (!lockMutexInit) { - lockMutexInit = GL_TRUE; - _glthread_INIT_MUTEX(lockMutex); - } - driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, intel->driScreen->myNum, "i915"); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index e9e4bc0645..38f0a4e2d7 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -33,7 +33,7 @@ #include "intel_screen.h" #include "intel_context.h" #include "intel_batchbuffer.h" -#include "intel_buffers.h" +#include "intel_swapbuffers.h" #include "i830_dri.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index cf8d190d34..e6bd87a4ea 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -37,7 +37,7 @@ #include "intel_screen.h" #include "intel_batchbuffer.h" -#include "intel_buffers.h" +#include "intel_swapbuffers.h" #include "i830_dri.h" #include "dri_bufpool.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c new file mode 100644 index 0000000000..56d220ab9e --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -0,0 +1,634 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_swapbuffers.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" +#include "context.h" +#include "utils.h" +#include "drirenderbuffer.h" +#include "vblank.h" + +#include "pipe/p_context.h" +#include "state_tracker/st_cb_fbo.h" + + +/* This block can be removed when libdrm >= 2.3.1 is required */ + +#ifndef DRM_VBLANK_FLIP + +#define DRM_VBLANK_FLIP 0x8000000 + +typedef struct drm_i915_flip { + int pipes; +} drm_i915_flip_t; + +#undef DRM_IOCTL_I915_FLIP +#define DRM_IOCTL_I915_FLIP DRM_IOW(DRM_COMMAND_BASE + DRM_I915_FLIP, \ + drm_i915_flip_t) + +#endif + + + + +/** + * Copy the back color buffer to the front color buffer. + * Used for SwapBuffers(). + */ +void +intelCopyBuffer(__DRIdrawablePrivate * dPriv, + const drm_clip_rect_t * rect) +{ + + struct intel_context *intel; + const intelScreenPrivate *intelScreen; + + DBG(SWAP, "%s\n", __FUNCTION__); + + assert(dPriv); + + intel = intelScreenContext(dPriv->driScreenPriv->private); + if (!intel) + return; + + intelScreen = intel->intelScreen; + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + current context (which is what intelScreenContext should return) might + not get a contended lock and thus cliprects not updated (tests/manywin) */ + if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + + + if (dPriv && dPriv->numClipRects) { + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; +#if 0 + const struct pipe_region *backRegion + = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? + intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : + intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); +#endif + const int backWidth = intel_fb->Base.Width; + const int backHeight = intel_fb->Base.Height; + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; +#if 0 + const int srcpitch = backRegion->pitch; +#endif + const int cpp = intelScreen->front.cpp; + int BR13, CMD; + int i; + + const struct pipe_surface *backSurf; + const struct pipe_region *backRegion; + int srcpitch; + struct st_renderbuffer *strb; + + /* blit from back color buffer if it exists, else front buffer */ + strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); + if (strb) { + backSurf = strb->surface; + } + else { + strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + backSurf = strb->surface; + } + + backRegion = backSurf->region; + srcpitch = backRegion->pitch; + + ASSERT(intel_fb); + ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ + ASSERT(backRegion); + ASSERT(backRegion->cpp == cpp); + + DBG(SWAP, "front pitch %d back pitch %d\n", + pitch, backRegion->pitch); + + if (cpp == 2) { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + } + else { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + } + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + drm_clip_rect_t sbox; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->front.width || + pbox->y2 > intelScreen->front.height) + continue; + + box = *pbox; + + if (rect) { + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > backWidth) + box.x2 = backWidth + box.x1; + if (box.y2 - box.y1 > backHeight) + box.y2 = backHeight + box.y1; + + DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((box.y1 << 16) | box.x1); + OUT_BATCH((box.y2 << 16) | box.x2); + + OUT_RELOC(intelScreen->front.buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + OUT_BATCH((sbox.y1 << 16) | sbox.x1); + OUT_BATCH((srcpitch * cpp) & 0xffff); + OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + ADVANCE_BATCH(); + } + + if (intel->first_swap_fence) + driFenceUnReference(intel->first_swap_fence); + intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); + driFenceReference(intel->first_swap_fence); + } + + UNLOCK_HARDWARE(intel); + + /* XXX this is bogus. The context here may not even be bound to this drawable! */ + if (intel->lastStamp != dPriv->lastStamp) { + GET_CURRENT_CONTEXT(currctx); + struct intel_context *intelcurrent = intel_context(currctx); + if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { + intelWindowMoved(intel); + intel->lastStamp = dPriv->lastStamp; + } + } + +} + + + +/** + * This will be called whenever the currently bound window is moved/resized. + * XXX: actually, it seems to NOT be called when the window is only moved (BP). + */ +void +intelWindowMoved(struct intel_context *intel) +{ + GLcontext *ctx = &intel->ctx; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + + if (!intel->ctx.DrawBuffer) { + /* when would this happen? -BP */ + assert(0); + intel->numClipRects = 0; + } + + /* Update Mesa's notion of window size */ + driUpdateFramebufferSize(ctx, dPriv); + intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ + + { + drmI830Sarea *sarea = intel->sarea; + drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, + .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; + drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, .y1 = sarea->pipeA_y, + .x2 = sarea->pipeA_x + sarea->pipeA_w, + .y2 = sarea->pipeA_y + sarea->pipeA_h }; + drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, .y1 = sarea->pipeB_y, + .x2 = sarea->pipeB_x + sarea->pipeB_w, + .y2 = sarea->pipeB_y + sarea->pipeB_h }; + GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); + GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); + GLuint flags = intel_fb->vblank_flags; + GLboolean pf_active; + GLint pf_pipes; + + /* Update page flipping info + */ + pf_pipes = 0; + + if (areaA > 0) + pf_pipes |= 1; + + if (areaB > 0) + pf_pipes |= 2; + + intel_fb->pf_current_page = (intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3; + + intel_fb->pf_num_pages = 2 /*intel->intelScreen->third.handle ? 3 : 2*/; + + pf_active = pf_pipes && (pf_pipes & intel->sarea->pf_active) == pf_pipes; + + if (pf_active != intel_fb->pf_active) + DBG(LOCK, "%s - Page flipping %sactive\n", + __progname, pf_active ? "" : "in"); + + if (pf_active) { + /* Sync pages between pipes if we're flipping on both at the same time */ + if (pf_pipes == 0x3 && pf_pipes != intel_fb->pf_pipes && + (intel->sarea->pf_current_page & 0x3) != + (((intel->sarea->pf_current_page) >> 2) & 0x3)) { + drm_i915_flip_t flip; + + if (intel_fb->pf_current_page == + (intel->sarea->pf_current_page & 0x3)) { + /* XXX: This is ugly, but emitting two flips 'in a row' can cause + * lockups for unknown reasons. + */ + intel->sarea->pf_current_page = + intel->sarea->pf_current_page & 0x3; + intel->sarea->pf_current_page |= + ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % + intel_fb->pf_num_pages) << 2; + + flip.pipes = 0x2; + } else { + intel->sarea->pf_current_page = + intel->sarea->pf_current_page & (0x3 << 2); + intel->sarea->pf_current_page |= + (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % + intel_fb->pf_num_pages; + + flip.pipes = 0x1; + } + + drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); + } + + intel_fb->pf_pipes = pf_pipes; + } + + intel_fb->pf_active = pf_active; +#if 0 + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); +#endif + + /* Update vblank info + */ + if (areaB > areaA || (areaA == areaB && areaB > 0)) { + flags = intel_fb->vblank_flags | VBLANK_FLAG_SECONDARY; + } else { + flags = intel_fb->vblank_flags & ~VBLANK_FLAG_SECONDARY; + } + + if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && + !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { + drmVBlank vbl; + int i; + + vbl.request.type = DRM_VBLANK_ABSOLUTE; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + vbl.request.type |= DRM_VBLANK_SECONDARY; + } + + for (i = 0; i < intel_fb->pf_num_pages; i++) { + if ((intel_fb->vbl_waited - intel_fb->vbl_pending[i]) <= (1<<23)) + continue; + + vbl.request.sequence = intel_fb->vbl_pending[i]; + drmWaitVBlank(intel->driFd, &vbl); + } + + intel_fb->vblank_flags = flags; + driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); + intel_fb->vbl_waited = intel_fb->vbl_seq; + + for (i = 0; i < intel_fb->pf_num_pages; i++) { + intel_fb->vbl_pending[i] = intel_fb->vbl_waited; + } + } + } + + /* This will be picked up by looking at the dirty state flags: + */ + + /* Update hardware scissor */ +// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, +// ctx->Scissor.Width, ctx->Scissor.Height); + + /* Re-calculate viewport related state */ +// ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far ); +} + + + + + +/* Emit wait for pending flips */ +#if 0 +void +intel_wait_flips(struct intel_context *intel, GLuint batch_flags) +{ + struct intel_framebuffer *intel_fb = + (struct intel_framebuffer *) intel->ctx.DrawBuffer; + struct intel_renderbuffer *intel_rb = + intel_get_renderbuffer(&intel_fb->Base, + intel_fb->Base._ColorDrawBufferMask[0] == + BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : + BUFFER_BACK_LEFT); + + if (intel_fb->Base.Name == 0 && intel_rb->pf_pending == intel_fb->pf_seq) { + GLint pf_pipes = intel_fb->pf_pipes; + BATCH_LOCALS; + + /* Wait for pending flips to take effect */ + BEGIN_BATCH(2, batch_flags); + OUT_BATCH(pf_pipes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) + : 0); + OUT_BATCH(pf_pipes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) + : 0); + ADVANCE_BATCH(); + + intel_rb->pf_pending--; + } +} +#endif + +#if 0 +/* Flip the front & back buffers + */ +static GLboolean +intelPageFlip(const __DRIdrawablePrivate * dPriv) +{ + struct intel_context *intel; + int ret; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + + DBG(SWAP, "%s\n", __FUNCTION__); + + assert(dPriv); + assert(dPriv->driContextPriv); + assert(dPriv->driContextPriv->driverPrivate); + + intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate; + + if (intel->intelScreen->drmMinor < 9) + return GL_FALSE; + + intelFlush(&intel->ctx); + + ret = 0; + + LOCK_HARDWARE(intel); + + if (dPriv->numClipRects && intel_fb->pf_active) { + drm_i915_flip_t flip; + + flip.pipes = intel_fb->pf_pipes; + + ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); + } + + UNLOCK_HARDWARE(intel); + + if (ret || !intel_fb->pf_active) + return GL_FALSE; + + if (!dPriv->numClipRects) { + usleep(10000); /* throttle invisible client 10ms */ + } + + intel_fb->pf_current_page = (intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3; + + if (dPriv->numClipRects != 0) { + intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending = + intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->pf_pending = + ++intel_fb->pf_seq; + } + +#if 0 + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, &intel_fb->Base); +#endif + + DBG(SWAP, "%s: success\n", __FUNCTION__); + + return GL_TRUE; +} +#endif + + +static GLboolean +intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) +{ + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); + struct intel_context *intel = + intelScreenContext(dPriv->driScreenPriv->private); + const intelScreenPrivate *intelScreen = intel->intelScreen; + unsigned int target; + drm_i915_vblank_swap_t swap; + GLboolean ret; + + if (!intel_fb->vblank_flags || + (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || + intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) + return GL_FALSE; + + swap.seqtype = DRM_VBLANK_ABSOLUTE; + + if (intel_fb->vblank_flags & VBLANK_FLAG_SYNC) { + swap.seqtype |= DRM_VBLANK_NEXTONMISS; + } else if (interval == 0) { + return GL_FALSE; + } + + swap.drawable = dPriv->hHWDrawable; + target = swap.sequence = intel_fb->vbl_seq + interval; + + if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { + swap.seqtype |= DRM_VBLANK_SECONDARY; + } + + LOCK_HARDWARE(intel); + + intel_batchbuffer_flush(intel->batch); + + if ( intel_fb->pf_active ) { + swap.seqtype |= DRM_VBLANK_FLIP; + + intel_fb->pf_current_page = (((intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3) + 1) % + intel_fb->pf_num_pages; + } + + if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, + sizeof(swap))) { + intel_fb->vbl_seq = swap.sequence; + swap.sequence -= target; + *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); + +#if 1 + intel_fb->vbl_pending[1] = intel_fb->vbl_pending[0] = intel_fb->vbl_seq; +#else + intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending = + intel_get_renderbuffer(&intel_fb->Base, + BUFFER_FRONT_LEFT)->vbl_pending = + intel_fb->vbl_seq; +#endif + + if (swap.seqtype & DRM_VBLANK_FLIP) { +#if 0 + intel_flip_renderbuffers(intel_fb); + intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); +#endif + } + + ret = GL_TRUE; + } else { + if (swap.seqtype & DRM_VBLANK_FLIP) { + intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> + (intel_fb->pf_pipes & 0x2)) & 0x3) % + intel_fb->pf_num_pages; + } + + ret = GL_FALSE; + } + + UNLOCK_HARDWARE(intel); + + return ret; +} + +void +intelSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + GET_CURRENT_CONTEXT(ctx); + struct intel_context *intel; + + if (ctx == NULL) + return; + + intel = intel_context(ctx); + + if (ctx->Visual.doubleBufferMode) { + GLboolean missed_target; + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + int64_t ust; + + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + + if (!intelScheduleSwap(dPriv, &missed_target)) { + driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, + &missed_target); + + intelCopyBuffer(dPriv, NULL); + } + + intel_fb->swap_count++; + (*dri_interface->getUST) (&ust); + if (missed_target) { + intel_fb->swap_missed_count++; + intel_fb->swap_missed_ust = ust - intel_fb->swap_ust; + } + + intel_fb->swap_ust = ust; + } + } + else { + /* XXX this shouldn't be an error but we can't handle it for now */ + fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + } +} + +void +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) +{ + if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { + struct intel_context *intel = + (struct intel_context *) dPriv->driContextPriv->driverPrivate; + GLcontext *ctx = &intel->ctx; + + if (ctx->Visual.doubleBufferMode) { + drm_clip_rect_t rect; + /* fixup cliprect (driDrawable may have changed?) later */ + rect.x1 = x; + rect.y1 = y; + rect.x2 = w; + rect.y2 = h; + _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + intelCopyBuffer(dPriv, &rect); + } + } + else { + /* XXX this shouldn't be an error but we can't handle it for now */ + fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + } +} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h new file mode 100644 index 0000000000..d12b5ef96e --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -0,0 +1,74 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BUFFERS_H +#define INTEL_BUFFERS_H + + +struct intel_context; +struct intel_framebuffer; + +/** + * Intel framebuffer, derived from gl_framebuffer. + */ +struct intel_framebuffer +{ + struct gl_framebuffer Base; + + /* Drawable page flipping state */ + GLboolean pf_active; + GLuint pf_seq; + GLint pf_pipes; + GLint pf_current_page; + GLint pf_num_pages; + + /* VBI + */ + GLuint vbl_seq; + GLuint vblank_flags; + GLuint vbl_waited; + + int64_t swap_ust; + int64_t swap_missed_ust; + + GLuint swap_count; + GLuint swap_missed_count; + + GLuint vbl_pending[3]; /**< [number of color buffers] */ +}; + + +void intelCopyBuffer(__DRIdrawablePrivate * dPriv, + const drm_clip_rect_t * rect); + +extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); + +extern void intelWindowMoved(struct intel_context *intel); + +#endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index bedc7d14ab..33aacf2a43 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -36,7 +36,7 @@ #include "intel_context.h" #include "intel_winsys.h" -#include "intel_buffers.h" +#include "intel_swapbuffers.h" #include "intel_batchbuffer.h" #include "pipe/p_winsys.h" -- cgit v1.2.3 From 0b13336540b887601d5d9c3106aeece848bf589f Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 16:59:15 +0100 Subject: remove #include mtypes.h --- src/mesa/pipe/p_winsys.h | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index b7495a205f..09baef020e 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -28,7 +28,6 @@ #ifndef P_WINSYS_H #define P_WINSYS_H -#include "main/mtypes.h" /* This is the interface that softpipe requires any window system * hosting it to implement. This is the only include file in softpipe -- cgit v1.2.3 From bec2230514ad2e8e52625ecb339fcadfabffd4df Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 16:59:54 +0100 Subject: move pipe_buffer_handle definition to top of file, use it in pipe_region --- src/mesa/pipe/p_state.h | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index df456cc2ed..6799fdb31c 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -55,6 +55,9 @@ /* fwd decl */ struct pipe_surface; +/* opaque type */ +struct pipe_buffer_handle; + /*** *** State objects @@ -238,7 +241,7 @@ struct pipe_sampler_state struct pipe_region { - void *buffer; /**< driver private buffer handle */ + struct pipe_buffer_handle *buffer; /**< driver private buffer handle */ GLuint refcount; /**< Reference count for region */ GLuint cpp; /**< bytes per pixel */ @@ -326,8 +329,4 @@ struct pipe_mipmap_tree }; - -struct pipe_buffer_handle; - - #endif -- cgit v1.2.3 From 83521ee66ad3f3cd3ec8212d6f746caecdc6c407 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 17:00:56 +0100 Subject: notes/asserts for get/put_tile() --- src/mesa/pipe/i915simple/i915_surface.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 12b39bb1e4..a67784b70c 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -43,6 +43,8 @@ i915_get_tile(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) { /* any need to get tiles from i915 surfaces? */ + /* Yes, for glReadPixels (for a while at least). */ + assert(0); } @@ -51,6 +53,7 @@ i915_put_tile(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, const GLfloat *p) { /* any need to put tiles into i915 surfaces? */ + assert(0); } -- cgit v1.2.3 From 37fdc69a99a7692d337fb4efaea97df842490f21 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 17:03:51 +0100 Subject: fix invalid usage of buffer_unreference() Need to pass the address of the buffer ptr, not the buffer ptr. Before, the region->buffer type was void * so it wasn't effectively type-checked. Changing the type to pipe_buffer_object* allows the compiler to detect the error. Fixing this solves a segfault. --- src/mesa/pipe/i915simple/i915_regions.c | 2 +- src/mesa/pipe/softpipe/sp_region.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index 9b7cddb58d..e27443a58c 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -124,7 +124,7 @@ i915_region_release(struct pipe_context *pipe, struct pipe_region **region) assert((*region)->map_refcount == 0); i915->pipe.winsys->buffer_unreference( i915->pipe.winsys, - (*region)->buffer ); + &((*region)->buffer) ); free(*region); } *region = NULL; diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 115f3ab826..10b7479e97 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -119,7 +119,7 @@ sp_region_release(struct pipe_context *pipe, struct pipe_region **region) assert((*region)->map_refcount == 0); sp->pipe.winsys->buffer_unreference( sp->pipe.winsys, - (*region)->buffer ); + &((*region)->buffer) ); free(*region); } *region = NULL; -- cgit v1.2.3 From bf383d1f5496f6c2ce318bbe1c02e7c7953a9153 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 17:04:12 +0100 Subject: repair breakage from recent pipe_winsys re-org --- src/mesa/drivers/x11/xm_buffer.c | 26 ++++++-- src/mesa/drivers/x11/xm_softpipe.c | 127 +++++++++++++++++++++++++++---------- src/mesa/drivers/x11/xm_surface.c | 34 +++++++++- 3 files changed, 149 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_buffer.c b/src/mesa/drivers/x11/xm_buffer.c index fb6815ae1e..4cd2ab7c6b 100644 --- a/src/mesa/drivers/x11/xm_buffer.c +++ b/src/mesa/drivers/x11/xm_buffer.c @@ -248,6 +248,20 @@ xmesa_delete_renderbuffer(struct gl_renderbuffer *rb) } +static void +finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) +{ + struct pipe_context *pipe = ctx->st->pipe; + + if (!xrb->St.surface) { + xrb->St.surface = xmesa_new_surface(ctx, xrb); + } + if (!xrb->St.surface->region) { + xrb->St.surface->region = pipe->region_alloc(pipe, 1, 0, 0, 0x0); + } +} + + /** * Reallocate renderbuffer storage for front color buffer. * Called via gl_renderbuffer::AllocStorage() @@ -271,8 +285,9 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->Height = height; rb->InternalFormat = internalFormat; - if (!xrb->St.surface) - xrb->St.surface = xmesa_new_surface(ctx, xrb); + if (!xrb->St.surface || !xrb->St.surface->region) + finish_surface_init(ctx, xrb); + xrb->St.surface->width = width; xrb->St.surface->height = height; @@ -325,8 +340,9 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xrb->origin4 = NULL; } - if (!xrb->St.surface) - xrb->St.surface = xmesa_new_surface(ctx, xrb); + if (!xrb->St.surface || !xrb->St.surface->region) + finish_surface_init(ctx, xrb); + xrb->St.surface->width = width; xrb->St.surface->height = height; @@ -340,7 +356,7 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, */ struct xmesa_renderbuffer * xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, - GLboolean backBuffer) + GLboolean backBuffer) { struct xmesa_renderbuffer *xrb = CALLOC_STRUCT(xmesa_renderbuffer); struct pipe_context *pipe = NULL;/*ctx->st->pipe;*/ diff --git a/src/mesa/drivers/x11/xm_softpipe.c b/src/mesa/drivers/x11/xm_softpipe.c index 6e8ef21b50..2f93ff3309 100644 --- a/src/mesa/drivers/x11/xm_softpipe.c +++ b/src/mesa/drivers/x11/xm_softpipe.c @@ -37,6 +37,7 @@ #include "xmesaP.h" #include "main/macros.h" +#include "pipe/p_winsys.h" #include "pipe/softpipe/sp_winsys.h" @@ -62,36 +63,37 @@ struct xm_buffer static inline struct xm_buffer * xm_bo( struct pipe_buffer_handle *bo ) { - return (struct xm_buffer *)bo; + return (struct xm_buffer *) bo; } static inline struct pipe_buffer_handle * pipe_bo( struct xm_buffer *bo ) { - return (struct pipe_buffer_handle *)bo; + return (struct pipe_buffer_handle *) bo; } /* Turn a softpipe winsys into an xm/softpipe winsys: */ static inline struct xm_softpipe_winsys * -xm_softpipe_winsys( struct softpipe_winsys *sws ) +xm_softpipe_winsys(struct softpipe_winsys *sws) { - return (struct xm_softpipe_winsys *)sws; + return (struct xm_softpipe_winsys *) sws; } /* Most callbacks map direcly onto dri_bufmgr operations: */ -static void *xm_buffer_map(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf ) +static void * +xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned flags) { struct xm_buffer *xm_buf = xm_bo(buf); xm_buf->mapped = xm_buf->data; return xm_buf->mapped; } -static void xm_buffer_unmap(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf) +static void +xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) { struct xm_buffer *xm_buf = xm_bo(buf); xm_buf->mapped = NULL; @@ -99,16 +101,15 @@ static void xm_buffer_unmap(struct softpipe_winsys *sws, static struct pipe_buffer_handle * -xm_buffer_reference(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf) +xm_buffer_reference(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) { struct xm_buffer *xm_buf = xm_bo(buf); xm_buf->refcount++; return buf; } -static void xm_buffer_unreference(struct softpipe_winsys *sws, - struct pipe_buffer_handle **buf) +static void +xm_buffer_unreference(struct pipe_winsys *pws, struct pipe_buffer_handle **buf) { struct xm_buffer *xm_buf = xm_bo(*buf); xm_buf->refcount--; @@ -122,21 +123,18 @@ static void xm_buffer_unreference(struct softpipe_winsys *sws, *buf = NULL; } - -static void xm_buffer_data(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned size, const void *data ) +static void +xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned size, const void *data ) { struct xm_buffer *xm_buf = xm_bo(buf); assert(!xm_buf->data); xm_buf->data = malloc(size); } -static void xm_buffer_subdata(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) +static void +xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned long offset, unsigned long size, const void *data) { struct xm_buffer *xm_buf = xm_bo(buf); GLubyte *b = (GLubyte *) xm_buf->data; @@ -144,11 +142,9 @@ static void xm_buffer_subdata(struct softpipe_winsys *sws, memcpy(b + offset, data, size); } -static void xm_buffer_get_subdata(struct softpipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) +static void +xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned long offset, unsigned long size, void *data) { const struct xm_buffer *xm_buf = xm_bo(buf); const GLubyte *b = (GLubyte *) xm_buf->data; @@ -156,21 +152,86 @@ static void xm_buffer_get_subdata(struct softpipe_winsys *sws, memcpy(data, b + offset, size); } +static void +xm_flush_frontbuffer(struct pipe_winsys *pws) +{ + /* + struct intel_context *intel = intel_pipe_winsys(sws)->intel; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelCopyBuffer(dPriv, NULL); + */ +} + +static void +xm_wait_idle(struct pipe_winsys *pws) +{ + /* no-op */ +} + +static void +xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) +{ + va_list args; + va_start( args, fmtString ); + vfprintf(stderr, fmtString, args); + va_end( args ); +} + +static const char * +xm_get_name(struct pipe_winsys *pws) +{ + return "Xlib"; +} + /* Softpipe has no concept of pools. We choose the tex/region pool * for all buffers. */ static struct pipe_buffer_handle * -xm_create_buffer(struct softpipe_winsys *sws, - unsigned alignment) +xm_buffer_create(struct pipe_winsys *pws, unsigned alignment) { - struct xm_buffer *buffer; - - buffer = CALLOC_STRUCT(xm_buffer); + struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); return pipe_bo(buffer); } +struct xmesa_pipe_winsys +{ + struct pipe_winsys winsys; + XMesaContext xmesa; +}; + +static struct pipe_winsys * +xmesa_create_pipe_winsys( XMesaContext xmesa ) +{ + struct xmesa_pipe_winsys *xws = CALLOC_STRUCT(xmesa_pipe_winsys); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + * + * Pipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + xws->winsys.buffer_create = xm_buffer_create; + xws->winsys.buffer_map = xm_buffer_map; + xws->winsys.buffer_unmap = xm_buffer_unmap; + xws->winsys.buffer_reference = xm_buffer_reference; + xws->winsys.buffer_unreference = xm_buffer_unreference; + xws->winsys.buffer_data = xm_buffer_data; + xws->winsys.buffer_subdata = xm_buffer_subdata; + xws->winsys.buffer_get_subdata = xm_buffer_get_subdata; + xws->winsys.flush_frontbuffer = xm_flush_frontbuffer; + xws->winsys.wait_idle = xm_wait_idle; + xws->winsys.printf = xm_printf; + xws->winsys.get_name = xm_get_name; + xws->xmesa = xmesa; + + return &xws->winsys; +} + + struct pipe_context * xmesa_create_softpipe(XMesaContext xmesa) { @@ -183,6 +244,7 @@ xmesa_create_softpipe(XMesaContext xmesa) * the SwapBuffers implementation in this winsys driver requires * that rendering be done to an appropriate xm_buffer. */ +#if 0 isws->sws.create_buffer = xm_create_buffer; isws->sws.buffer_map = xm_buffer_map; isws->sws.buffer_unmap = xm_buffer_unmap; @@ -191,8 +253,9 @@ xmesa_create_softpipe(XMesaContext xmesa) isws->sws.buffer_data = xm_buffer_data; isws->sws.buffer_subdata = xm_buffer_subdata; isws->sws.buffer_get_subdata = xm_buffer_get_subdata; +#endif /* Create the softpipe context: */ - return softpipe_create( &isws->sws ); + return softpipe_create( xmesa_create_pipe_winsys(xmesa), &isws->sws ); } diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 1915151ba4..c6057e5154 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -168,6 +168,23 @@ write_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, } +static void +get_tile(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) +{ + +} + + +static void +put_tile(struct pipe_surface *ps, + GLuint x, GLuint y, GLuint w, GLuint h, const GLfloat *p) +{ + +} + + + /** * Called to create a pipe_surface for each X renderbuffer. * Note: this is being used instead of pipe->surface_alloc() since we @@ -186,6 +203,8 @@ xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb) #if 0 sps->surface.rb = xrb; /* XXX only needed for quad funcs above */ #endif + sps->surface.refcount = 1; + sps->surface.width = xrb->St.Base.Width; sps->surface.height = xrb->St.Base.Height; @@ -195,13 +214,15 @@ xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb) sps->write_quad_f = write_quad_f; sps->write_quad_f_swz = write_quad_f_swz; sps->write_quad_ub = write_quad_ub; + sps->surface.get_tile = get_tile; + sps->surface.put_tile = put_tile; /* Note, the region we allocate doesn't actually have any storage * since we're drawing into an XImage or Pixmap. * The region's size will get set in the xmesa_alloc_front/back_storage() * functions. */ - sps->surface.region = pipe->region_alloc(pipe, 0, 0, 0); + sps->surface.region = pipe->region_alloc(pipe, 0, 0, 0, 0x0); return &sps->surface; } @@ -215,6 +236,7 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) assert(pipeFormat); xms->surface.surface.format = pipeFormat; + xms->surface.surface.refcount = 1; switch (pipeFormat) { case PIPE_FORMAT_U_A8_R8_G8_B8: @@ -225,6 +247,8 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) xms->surface.write_quad_f = write_quad_f; xms->surface.write_quad_f_swz = write_quad_f_swz; xms->surface.write_quad_ub = write_quad_ub; + xms->surface.surface.get_tile = get_tile; + xms->surface.surface.put_tile = put_tile; break; case PIPE_FORMAT_S8_Z24: softpipe_init_surface_funcs(&xms->surface); @@ -237,6 +261,14 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) abort(); } + /* Note, the region we allocate doesn't actually have any storage + * since we're drawing into an XImage or Pixmap. + * The region's size will get set in the xmesa_alloc_front/back_storage() + * functions. + */ + if (pipe) + xms->surface.surface.region = pipe->region_alloc(pipe, 1, 0, 0, 0x0); + return &xms->surface.surface; } -- cgit v1.2.3 From 01efb1dd8f5e14e8c034e0100f76472d89f811af Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 19:56:30 +0100 Subject: Begin/EndQuery functions --- src/mesa/state_tracker/st_cb_queryobj.c | 76 +++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_queryobj.h | 36 ++++++++++++++++ 2 files changed, 112 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_queryobj.c create mode 100644 src/mesa/state_tracker/st_cb_queryobj.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_queryobj.c b/src/mesa/state_tracker/st_cb_queryobj.c new file mode 100644 index 0000000000..3a8fbde8ab --- /dev/null +++ b/src/mesa/state_tracker/st_cb_queryobj.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * glBegin/EndQuery interface to pipe + * + * \author Brian Paul + */ + + +#include "main/imports.h" +#include "main/context.h" +#include "main/image.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "st_context.h" +#include "st_cb_queryobj.h" +#include "st_public.h" + + +/** + * Do glReadPixels by getting rows from the framebuffer surface with + * get_tile(). Convert to requested format/type with Mesa image routines. + * Image transfer ops are done in software too. + */ +static void +st_BeginQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q) +{ + struct pipe_context *pipe = ctx->st->pipe; + if (target == GL_SAMPLES_PASSED_ARB) { + pipe->reset_occlusion_counter(pipe); + } +} + + +static void +st_EndQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q) +{ + struct pipe_context *pipe = ctx->st->pipe; + if (target == GL_SAMPLES_PASSED_ARB) { + q->Result = pipe->get_occlusion_counter(pipe); + } +} + + +void st_init_query_functions(struct dd_function_table *functions) +{ + functions->BeginQuery = st_BeginQuery; + functions->EndQuery = st_EndQuery; +} diff --git a/src/mesa/state_tracker/st_cb_queryobj.h b/src/mesa/state_tracker/st_cb_queryobj.h new file mode 100644 index 0000000000..9220a212b6 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_queryobj.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ST_CB_QUERYOBJ_H +#define ST_CB_QUERYOBJ_H + + +extern void +st_init_query_functions(struct dd_function_table *functions); + + +#endif -- cgit v1.2.3 From 24864741c58f9e40450f1ea05f7fc50d06050441 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 19:57:37 +0100 Subject: added st_cb_queryobj.c --- src/mesa/sources | 1 + src/mesa/state_tracker/st_context.c | 2 ++ 2 files changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index c94cd3de26..a3b8ac34b1 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -198,6 +198,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_drawpixels.c \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_program.c \ + state_tracker/st_cb_queryobj.c \ state_tracker/st_cb_readpixels.c \ state_tracker/st_cb_strings.c \ state_tracker/st_cb_texture.c \ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 7bb9cbda45..2a85ce6baa 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -32,6 +32,7 @@ #include "st_cb_clear.h" #include "st_cb_drawpixels.h" #include "st_cb_fbo.h" +#include "st_cb_queryobj.h" #include "st_cb_readpixels.h" #include "st_cb_texture.h" #include "st_cb_flush.h" @@ -109,6 +110,7 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_drawpixels_functions(functions); st_init_fbo_functions(functions); st_init_program_functions(functions); + st_init_query_functions(functions); st_init_readpixels_functions(functions); st_init_texture_functions(functions); st_init_flush_functions(functions); -- cgit v1.2.3 From 24e21301e0cc16f0a3a81bfd7ac7ae8765174da8 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 19:57:50 +0100 Subject: remove some temp pipe hacks --- src/mesa/main/queryobj.c | 15 --------------- 1 file changed, 15 deletions(-) (limited to 'src') diff --git a/src/mesa/main/queryobj.c b/src/mesa/main/queryobj.c index fc04dde3f4..0e59ba615a 100644 --- a/src/mesa/main/queryobj.c +++ b/src/mesa/main/queryobj.c @@ -30,11 +30,6 @@ #include "queryobj.h" #include "mtypes.h" -#if 1 /*PIPE*/ -#include "pipe/p_context.h" -#include "state_tracker/st_context.h" -#endif - /** * Allocate a new query object. This is a fallback routine called via @@ -225,10 +220,6 @@ _mesa_BeginQueryARB(GLenum target, GLuint id) q->Result = 0; q->Ready = GL_FALSE; -#if 1 /*PIPE*/ - ctx->st->pipe->reset_occlusion_counter(ctx->st->pipe); -#endif - if (target == GL_SAMPLES_PASSED_ARB) { ctx->Query.CurrentOcclusionObject = q; } @@ -291,12 +282,6 @@ _mesa_EndQueryARB(GLenum target) /* if we're using software rendering/querying */ q->Ready = GL_TRUE; } - -#if 1 /*PIPE*/ - if (target == GL_SAMPLES_PASSED_ARB) { - q->Result = ctx->st->pipe->get_occlusion_counter(ctx->st->pipe); - } -#endif } -- cgit v1.2.3 From 498c9e9782186a572885ff9927114706c3d93a22 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 20:03:24 +0100 Subject: added Bitmap() stub, call st_validate_state() in DrawPixels --- src/mesa/state_tracker/st_cb_drawpixels.c | 21 ++++++++++++++++++--- 1 file changed, 18 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index ce417b9b7e..cdb9ebc3ff 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -437,7 +437,7 @@ draw_blit(struct st_context *st, * Called via ctx->Driver.DrawPixels() */ static void -st_drawpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, +st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { @@ -445,6 +445,8 @@ st_drawpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct pipe_surface *ps; GLuint bufferFormat; + st_validate_state(st); + if (format == GL_DEPTH_COMPONENT) { ps = st->state.framebuffer.zbuf; } @@ -471,8 +473,21 @@ st_drawpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } -void st_init_drawpixels_functions(struct dd_function_table *functions) +static void +st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap ) { - functions->DrawPixels = st_drawpixels; + struct st_context *st = ctx->st; + + st_validate_state(st); + + /* XXX to do */ } + + +void st_init_drawpixels_functions(struct dd_function_table *functions) +{ + functions->DrawPixels = st_DrawPixels; + functions->Bitmap = st_Bitmap; +} -- cgit v1.2.3 From e209ca866dc1787b9eeece6ae3b53e1e47811d5f Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 21:18:33 +0100 Subject: do clipping in get_tile() --- src/mesa/pipe/softpipe/sp_surface.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 5ef8ec935f..1919947abf 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -104,6 +104,7 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, = ((const GLuint *) (ps->region->map + ps->offset)) + y * ps->region->pitch + x; GLuint i, j; + GLuint w0 = w; assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); @@ -111,22 +112,24 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, assert(x + w <= ps->width); assert(y + h <= ps->height); #else - /* temp hack */ + /* temp clipping hack */ if (x + w > ps->width) w = ps->width - x; if (y + h > ps->height) h = ps->height -y; #endif for (i = 0; i < h; i++) { + GLfloat *pRow = p; for (j = 0; j < w; j++) { const GLuint pixel = src[j]; - p[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - p[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - p[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - p[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - p += 4; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow += 4; } src += ps->region->pitch; + p += w0 * 4; } } @@ -348,9 +351,14 @@ s8_write_quad_stencil(struct softpipe_surface *sps, } +/** + * Initialize the quad_read/write and get/put_tile() methods. + */ void softpipe_init_surface_funcs(struct softpipe_surface *sps) { + assert(sps->surface.format); + switch (sps->surface.format) { case PIPE_FORMAT_U_A8_R8_G8_B8: sps->read_quad_f_swz = a8r8g8b8_read_quad_f_swz; -- cgit v1.2.3 From 5d42fdb9b70a53938cd29d7ebeaa4b9cedd2e8e3 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 21:19:04 +0100 Subject: mt->pitch must be at least 16 because of 64-byte alignment in sp_region.c (fix that someday) --- src/mesa/pipe/softpipe/sp_tex_layout.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c index 37969c84de..bd5f5e930d 100644 --- a/src/mesa/pipe/softpipe/sp_tex_layout.c +++ b/src/mesa/pipe/softpipe/sp_tex_layout.c @@ -116,6 +116,12 @@ sp_miptree_layout_2d( struct pipe_mipmap_tree *mt ) GLuint height = mt->height0; mt->pitch = mt->width0; + /* XXX FIX THIS: + * we use alignment=64 bytes in sp_region_alloc(). If we change + * that, change this too. + */ + if (mt->pitch < 16) + mt->pitch = 16; /* May need to adjust pitch to accomodate the placement of * the 2nd mipmap. This occurs when the alignment -- cgit v1.2.3 From 0095be534d633848bc12d73ed9dcc1b9aa41f00a Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 11 Aug 2007 21:21:12 +0100 Subject: Change/fix surface allocation functions. Use xmesa_new_color_surface() for front/back renderbuffer surfaces. Use xmesa_surface_alloc() for everything else (textures, other renderbuffers) --- src/mesa/drivers/x11/xm_buffer.c | 7 +--- src/mesa/drivers/x11/xm_softpipe.c | 2 + src/mesa/drivers/x11/xm_surface.c | 80 +++++++++++++++++++------------------- src/mesa/drivers/x11/xmesaP.h | 6 +-- 4 files changed, 45 insertions(+), 50 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_buffer.c b/src/mesa/drivers/x11/xm_buffer.c index 4cd2ab7c6b..52629aca18 100644 --- a/src/mesa/drivers/x11/xm_buffer.c +++ b/src/mesa/drivers/x11/xm_buffer.c @@ -252,10 +252,6 @@ static void finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) { struct pipe_context *pipe = ctx->st->pipe; - - if (!xrb->St.surface) { - xrb->St.surface = xmesa_new_surface(ctx, xrb); - } if (!xrb->St.surface->region) { xrb->St.surface->region = pipe->region_alloc(pipe, 1, 0, 0, 0x0); } @@ -391,10 +387,9 @@ xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, } /* only need to set Red/Green/EtcBits fields for user-created RBs */ - xrb->St.surface = xmesa_surface_alloc(pipe, pipeFormat); + xrb->St.surface = xmesa_new_color_surface(pipe, pipeFormat); xms = (struct xmesa_surface *) xrb->St.surface; xms->xrb = xrb; - } return xrb; } diff --git a/src/mesa/drivers/x11/xm_softpipe.c b/src/mesa/drivers/x11/xm_softpipe.c index 2f93ff3309..a08673444f 100644 --- a/src/mesa/drivers/x11/xm_softpipe.c +++ b/src/mesa/drivers/x11/xm_softpipe.c @@ -130,6 +130,8 @@ xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, struct xm_buffer *xm_buf = xm_bo(buf); assert(!xm_buf->data); xm_buf->data = malloc(size); + if (data) + memcpy(xm_buf->data, data, size); } static void diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index c6057e5154..6f6c549c07 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -172,7 +172,18 @@ static void get_tile(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) { - + struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); + GLubyte tmp[MAX_WIDTH * 4]; + GLuint i, j; + GET_CURRENT_CONTEXT(ctx); + FLIP(y); + for (i = 0; i < h; i++) { + xrb->St.Base.GetRow(ctx, &xrb->St.Base, w, x, y - i, tmp); + for (j = 0; j < w * 4; j++) { + p[j] = UBYTE_TO_FLOAT(tmp[j]); + } + p += w * 4; + } } @@ -180,7 +191,7 @@ static void put_tile(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, const GLfloat *p) { - + assert(0); } @@ -191,45 +202,7 @@ put_tile(struct pipe_surface *ps, * have special/unique quad read/write functions for X. */ struct pipe_surface * -xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb) -{ - struct pipe_context *pipe = ctx->st->pipe; - struct softpipe_surface *sps; - - sps = CALLOC_STRUCT(softpipe_surface); - if (!sps) - return NULL; - -#if 0 - sps->surface.rb = xrb; /* XXX only needed for quad funcs above */ -#endif - sps->surface.refcount = 1; - - sps->surface.width = xrb->St.Base.Width; - sps->surface.height = xrb->St.Base.Height; - - sps->read_quad_f = read_quad_f; - sps->read_quad_f_swz = read_quad_f_swz; - sps->read_quad_ub = read_quad_ub; - sps->write_quad_f = write_quad_f; - sps->write_quad_f_swz = write_quad_f_swz; - sps->write_quad_ub = write_quad_ub; - sps->surface.get_tile = get_tile; - sps->surface.put_tile = put_tile; - - /* Note, the region we allocate doesn't actually have any storage - * since we're drawing into an XImage or Pixmap. - * The region's size will get set in the xmesa_alloc_front/back_storage() - * functions. - */ - sps->surface.region = pipe->region_alloc(pipe, 0, 0, 0, 0x0); - - return &sps->surface; -} - - -struct pipe_surface * -xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) +xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) { struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); @@ -273,6 +246,31 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) } +/** + * Called via pipe->surface_alloc() to create new surfaces (textures, + * renderbuffers, etc. + */ +struct pipe_surface * +xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) +{ + struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + + assert(pipeFormat); + + xms->surface.surface.format = pipeFormat; + xms->surface.surface.refcount = 1; + /* + * This is really just a softpipe surface, not an XImage/Pixmap surface. + */ + softpipe_init_surface_funcs(&xms->surface); + + assert(pipe); + xms->surface.surface.region = pipe->region_alloc(pipe, 1, 0, 0, 0x0); + + return &xms->surface.surface; +} + + const GLuint * xmesa_supported_formats(struct pipe_context *pipe, GLuint *numFormats) { diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 361c657354..9cbe8670f9 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -603,9 +603,6 @@ struct xmesa_surface }; -extern struct pipe_surface * -xmesa_new_surface(GLcontext *ctx, struct xmesa_renderbuffer *xrb); - extern void xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value); @@ -618,6 +615,9 @@ xmesa_create_softpipe(XMesaContext xm); extern struct pipe_surface * xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); +extern struct pipe_surface * +xmesa_new_color_surface(struct pipe_context *pipe, GLuint format); + extern const GLuint * xmesa_supported_formats(struct pipe_context *pipe, GLuint *numFormats); -- cgit v1.2.3 From eb450bbb40bd9cccde74b1e2d660908e154f9cc0 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 13:56:30 -0600 Subject: flush pipe before reading image --- src/mesa/state_tracker/st_cb_readpixels.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 273ed5770b..22abc104e2 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -74,9 +74,17 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, return; } + /* make sure rendering has completed */ + pipe->flush(pipe, 0x0); - /* XXX check pack->BufferObj !!! */ + if (pack->BufferObj && pack->BufferObj->Name) { + /* reading into a PBO */ + } + else { + /* reading into user memory/buffer */ + + } strb = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); if (!strb) -- cgit v1.2.3 From 21b9b8b74ec0975296d22493254875c0deb7b6da Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 13:56:54 -0600 Subject: fill in the get_tile() function so ReadPixels can work --- src/mesa/pipe/i915simple/i915_surface.c | 39 ++++++++++++++++++++++++++++++--- 1 file changed, 36 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index a67784b70c..cc0a87012a 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -29,6 +29,7 @@ #include "i915_state.h" #include "pipe/p_defines.h" #include "main/imports.h" +#include "main/macros.h" struct i915_surface @@ -38,13 +39,45 @@ struct i915_surface }; +/** + * Note: this is exactly like a8r8g8b8_get_tile() in sp_surface.c + * Share it someday. + */ static void i915_get_tile(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) { - /* any need to get tiles from i915 surfaces? */ - /* Yes, for glReadPixels (for a while at least). */ - assert(0); + const GLuint *src + = ((const GLuint *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + GLuint i, j; + GLuint w0 = w; + + assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp clipping hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + GLfloat *pRow = p; + for (j = 0; j < w; j++) { + const GLuint pixel = src[j]; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow += 4; + } + src += ps->region->pitch; + p += w0 * 4; + } } -- cgit v1.2.3 From faa14a9ded2ddb784302b4634f0ca40c51ec1c4b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 14:53:27 -0600 Subject: check for ctx==NULL in st_renderbuffer_delete() --- src/mesa/state_tracker/st_cb_fbo.c | 19 ++++++++++++------- 1 file changed, 12 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index bb588d10ea..3048765114 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -118,15 +118,20 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, static void st_renderbuffer_delete(struct gl_renderbuffer *rb) { - GET_CURRENT_CONTEXT(ctx); - struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); - ASSERT(strb); - if (strb && strb->surface) { - if (strb->surface->region) { - pipe->region_release(pipe, &strb->surface->region); + GET_CURRENT_CONTEXT(ctx); + if (ctx) { + struct pipe_context *pipe = ctx->st->pipe; + ASSERT(strb); + if (strb && strb->surface) { + if (strb->surface->region) { + pipe->region_release(pipe, &strb->surface->region); + } + free(strb->surface); } - free(strb->surface); + } + else { + _mesa_warning(NULL, "st_renderbuffer_delete() called, but no current context"); } free(strb); } -- cgit v1.2.3 From c271078048770a31028836eda684a6dbffc13cf5 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 15:02:50 -0600 Subject: with Y=0=top, front/back determination is negated --- src/mesa/pipe/draw/draw_cull.c | 4 ++-- src/mesa/pipe/draw/draw_twoside.c | 4 ++-- src/mesa/pipe/draw/draw_unfilled.c | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index 8b2ac5ea5f..48a7f5dab8 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -78,10 +78,10 @@ static void cull_tri( struct draw_stage *stage, header->det = ex * fy - ey * fx; if (header->det != 0) { - /* if (det > 0 then Z points toward camera and triangle is + /* if (det < 0 then Z points toward camera and triangle is * counter-clockwise winding. */ - GLuint winding = (header->det > 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; + GLuint winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; if ((winding & cull_stage(stage)->winding) == 0) { /* triangle is not culled, pass to next stage */ diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index fdda6b362f..20b2b4a8aa 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -53,9 +53,9 @@ static void twoside_begin( struct draw_stage *stage ) /* * We'll multiply the primitive's determinant by this sign to determine * if the triangle is back-facing (negative). - * sign = 1 for CCW, -1 for CW + * sign = -1 for CCW, +1 for CW */ - twoside->sign = (stage->draw->setup.front_winding == PIPE_WINDING_CCW) ? 1 : -1; + twoside->sign = (stage->draw->setup.front_winding == PIPE_WINDING_CCW) ? -1 : 1; stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index 82e8775f59..05242d8d10 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -119,7 +119,7 @@ static void unfilled_tri( struct draw_stage *stage, struct prim_header *header ) { struct unfilled_stage *unfilled = unfilled_stage(stage); - GLuint mode = unfilled->mode[header->det > 0.0]; + GLuint mode = unfilled->mode[header->det < 0.0]; switch (mode) { case PIPE_POLYGON_MODE_FILL: -- cgit v1.2.3 From 8cc668a4a8d67af21af6883e18fe7423f28999f9 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 15:20:08 -0600 Subject: fix some issues with texture/mipmap_tree state tracking --- src/mesa/state_tracker/st_atom_texture.c | 9 +++++---- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- src/mesa/state_tracker/st_context.h | 1 + 3 files changed, 8 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index eeaf68b001..bafd38695f 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -51,8 +51,8 @@ update_textures(struct st_context *st) for (u = 0; u < st->ctx->Const.MaxTextureImageUnits; u++) { struct gl_texture_object *texObj = st->ctx->Texture.Unit[u]._Current; + struct pipe_mipmap_tree *mt; if (texObj) { - struct pipe_mipmap_tree *mt; GLboolean flush, retval; retval = st_finalize_mipmap_tree(st->ctx, st->pipe, u, &flush); @@ -60,12 +60,13 @@ update_textures(struct st_context *st) retval, flush); mt = st_get_texobj_mipmap_tree(texObj); - - st->pipe->set_texture_state(st->pipe, u, mt); } else { - st->pipe->set_texture_state(st->pipe, u, NULL); + mt = NULL; } + + st->state.texture[u] = mt; + st->pipe->set_texture_state(st->pipe, u, mt); } } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index cdb9ebc3ff..7a89e853ee 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -315,8 +315,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* restore GL state */ pipe->set_setup_state(pipe, &ctx->st->state.setup); pipe->set_fs_state(pipe, &ctx->st->state.fs); - /* XXX FIX: pipe->set_texture_state(pipe, unit, ???); */ - pipe->set_sampler_state(pipe, unit, &ctx->st->state.sampler[0]); + pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); + pipe->set_sampler_state(pipe, unit, &ctx->st->state.sampler[unit]); free_mipmap_tree(pipe, mt); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 8ce85cddaa..8d82d53133 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -79,6 +79,7 @@ struct st_context struct pipe_scissor_state scissor; struct pipe_setup_state setup; struct pipe_stencil_state stencil; + struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; } state; -- cgit v1.2.3 From d46779103b38aeab61701759ed7a0b30cc71c0ef Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 15:21:01 -0600 Subject: check if we have a zbuf before enabling depth_test stage --- src/mesa/pipe/softpipe/sp_quad.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 0053b16e68..0f0736479f 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -48,7 +48,8 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.stencil_test->next = sp->quad.first; sp->quad.first = sp->quad.stencil_test; } - else if (sp->depth_test.enabled) { + else if (sp->depth_test.enabled && + sp->framebuffer.zbuf) { sp->quad.depth_test->next = sp->quad.first; sp->quad.first = sp->quad.depth_test; } -- cgit v1.2.3 From 40a86b20478024ca7c55400019c536cb5ff631d1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 13 Aug 2007 16:07:11 +0100 Subject: Start breaking the #include dependencies between pipe drivers and mesa. Pipe drivers shouldn't really know much about mesa and certainly shouldn't be #including files from src/mesa/main and the like. I've also (in i915simple especially) moved over from GL types to more conventional int/unsigned usage. This probably isn't really the ultimate desired set of types to use - possibly C99 would be better. It may even be that a subset of the GL types is preferable. --- src/mesa/pipe/i915simple/i915_blit.c | 40 ++-- src/mesa/pipe/i915simple/i915_blit.h | 26 +-- src/mesa/pipe/i915simple/i915_clear.c | 6 +- src/mesa/pipe/i915simple/i915_context.c | 27 +-- src/mesa/pipe/i915simple/i915_context.h | 18 +- src/mesa/pipe/i915simple/i915_debug.c | 194 +++++++++--------- src/mesa/pipe/i915simple/i915_debug_fp.c | 39 ++-- src/mesa/pipe/i915simple/i915_prim_emit.c | 33 +--- src/mesa/pipe/i915simple/i915_regions.c | 64 +++--- src/mesa/pipe/i915simple/i915_state.c | 6 +- src/mesa/pipe/i915simple/i915_state_derived.c | 14 +- src/mesa/pipe/i915simple/i915_state_dynamic.c | 86 ++++---- src/mesa/pipe/i915simple/i915_state_emit.c | 8 +- src/mesa/pipe/i915simple/i915_state_fragprog.c | 2 +- src/mesa/pipe/i915simple/i915_state_immediate.c | 35 ++-- src/mesa/pipe/i915simple/i915_state_inlines.h | 6 +- src/mesa/pipe/i915simple/i915_surface.c | 22 +-- src/mesa/pipe/i915simple/i915_tex_layout.c | 146 +++++++------- src/mesa/pipe/i915simple/i915_tex_layout.h | 4 +- src/mesa/pipe/i915simple/i915_winsys.h | 2 +- src/mesa/pipe/p_context.h | 69 +++---- src/mesa/pipe/p_defines.h | 10 + src/mesa/pipe/p_state.h | 251 ++++++++++++------------ src/mesa/pipe/softpipe/sp_clear.c | 2 +- src/mesa/pipe/softpipe/sp_clear.h | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 12 +- src/mesa/pipe/softpipe/sp_tex_layout.c | 13 +- 27 files changed, 553 insertions(+), 584 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/mesa/pipe/i915simple/i915_blit.c index 4fc216dd33..5044cf17f7 100644 --- a/src/mesa/pipe/i915simple/i915_blit.c +++ b/src/mesa/pipe/i915simple/i915_blit.c @@ -28,7 +28,7 @@ #include -#include "mtypes.h" +//#include "mtypes.h" #include "i915_context.h" #include "i915_winsys.h" @@ -40,15 +40,15 @@ void i915_fill_blit(struct i915_context *i915, - GLuint cpp, - GLshort dst_pitch, + unsigned cpp, + short dst_pitch, struct pipe_buffer_handle *dst_buffer, - GLuint dst_offset, - GLshort x, GLshort y, - GLshort w, GLshort h, - GLuint color) + unsigned dst_offset, + short x, short y, + short w, short h, + unsigned color) { - GLuint BR13, CMD; + unsigned BR13, CMD; BATCH_LOCALS; dst_pitch *= cpp; @@ -86,18 +86,18 @@ i915_fill_blit(struct i915_context *i915, void i915_copy_blit( struct i915_context *i915, - GLuint cpp, - GLshort src_pitch, + unsigned cpp, + short src_pitch, struct pipe_buffer_handle *src_buffer, - GLuint src_offset, - GLshort dst_pitch, + unsigned src_offset, + short dst_pitch, struct pipe_buffer_handle *dst_buffer, - GLuint dst_offset, - GLshort src_x, GLshort src_y, - GLshort dst_x, GLshort dst_y, - GLshort w, GLshort h ) + unsigned dst_offset, + short src_x, short src_y, + short dst_x, short dst_y, + short w, short h ) { - GLuint CMD, BR13; + unsigned CMD, BR13; int dst_y2 = dst_y + h; int dst_x2 = dst_x + w; BATCH_LOCALS; @@ -116,13 +116,13 @@ i915_copy_blit( struct i915_context *i915, case 1: case 2: case 3: - BR13 = (((GLint) dst_pitch) & 0xffff) | + BR13 = (((int) dst_pitch) & 0xffff) | (0xCC << 16) | (1 << 24); CMD = XY_SRC_COPY_BLT_CMD; break; case 4: BR13 = - (((GLint) dst_pitch) & 0xffff) | + (((int) dst_pitch) & 0xffff) | (0xCC << 16) | (1 << 24) | (1 << 25); CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | @@ -151,7 +151,7 @@ i915_copy_blit( struct i915_context *i915, OUT_BATCH((dst_y2 << 16) | dst_x2); OUT_RELOC(dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); OUT_BATCH((src_y << 16) | src_x); - OUT_BATCH(((GLint) src_pitch & 0xffff)); + OUT_BATCH(((int) src_pitch & 0xffff)); OUT_RELOC(src_buffer, I915_BUFFER_ACCESS_READ, src_offset); ADVANCE_BATCH(); } diff --git a/src/mesa/pipe/i915simple/i915_blit.h b/src/mesa/pipe/i915simple/i915_blit.h index b1131a005c..7ea4e979ec 100644 --- a/src/mesa/pipe/i915simple/i915_blit.h +++ b/src/mesa/pipe/i915simple/i915_blit.h @@ -31,24 +31,24 @@ #include "i915_context.h" extern void i915_copy_blit(struct i915_context *i915, - GLuint cpp, - GLshort src_pitch, + unsigned cpp, + short src_pitch, struct pipe_buffer_handle *src_buffer, - GLuint src_offset, - GLshort dst_pitch, + unsigned src_offset, + short dst_pitch, struct pipe_buffer_handle *dst_buffer, - GLuint dst_offset, - GLshort srcx, GLshort srcy, - GLshort dstx, GLshort dsty, - GLshort w, GLshort h ); + unsigned dst_offset, + short srcx, short srcy, + short dstx, short dsty, + short w, short h ); extern void i915_fill_blit(struct i915_context *i915, - GLuint cpp, - GLshort dst_pitch, + unsigned cpp, + short dst_pitch, struct pipe_buffer_handle *dst_buffer, - GLuint dst_offset, - GLshort x, GLshort y, - GLshort w, GLshort h, GLuint color); + unsigned dst_offset, + short x, short y, + short w, short h, unsigned color); #endif diff --git a/src/mesa/pipe/i915simple/i915_clear.c b/src/mesa/pipe/i915simple/i915_clear.c index 11aa55b64b..a4dba1a764 100644 --- a/src/mesa/pipe/i915simple/i915_clear.c +++ b/src/mesa/pipe/i915simple/i915_clear.c @@ -33,7 +33,7 @@ #include "pipe/p_defines.h" #include "i915_context.h" #include "i915_state.h" -#include "colormac.h" +//#include "colormac.h" /** @@ -42,9 +42,9 @@ */ void i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, - GLuint clearValue) + unsigned clearValue) { - GLint x, y, w, h; + int x, y, w, h; x = 0; y = 0; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 199352cd70..8a8582e0b2 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "main/imports.h" /* CALLOC */ +//#include "main/imports.h" /* CALLOC */ #include "i915_context.h" #include "i915_winsys.h" #include "i915_state.h" @@ -36,6 +36,7 @@ #include "pipe/draw/draw_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" +#include "pipe/p_util.h" /** @@ -43,13 +44,13 @@ * If we find texture and drawable support differs, add a selector * parameter or another function. */ -static const GLuint * +static const unsigned * i915_supported_formats(struct pipe_context *pipe, -// GLuint type, - GLuint *numFormats) +// unsigned type, + unsigned *numFormats) { #if 0 - static const GLuint tex_supported[] = { + static const unsigned tex_supported[] = { PIPE_FORMAT_U_R8_G8_B8_A8, PIPE_FORMAT_U_A8_R8_G8_B8, PIPE_FORMAT_U_R5_G6_B5, @@ -65,14 +66,14 @@ i915_supported_formats(struct pipe_context *pipe, /* Actually a lot more than this - add later: */ - static const GLuint render_supported[] = { + static const unsigned render_supported[] = { PIPE_FORMAT_U_A8_R8_G8_B8, PIPE_FORMAT_U_R5_G6_B5, }; /* */ - static const GLuint z_stencil_supported[] = { + static const unsigned z_stencil_supported[] = { PIPE_FORMAT_U_Z16, PIPE_FORMAT_U_Z32, PIPE_FORMAT_S8_Z24, @@ -96,7 +97,7 @@ i915_supported_formats(struct pipe_context *pipe, return NULL; } #else - static const GLuint render_supported[] = { + static const unsigned render_supported[] = { PIPE_FORMAT_U_A8_R8_G8_B8, PIPE_FORMAT_U_R5_G6_B5, PIPE_FORMAT_S8_Z24, @@ -111,8 +112,8 @@ i915_supported_formats(struct pipe_context *pipe, * We might want to return max texture levels instead... */ static void -i915_max_texture_size(struct pipe_context *pipe, GLuint textureType, - GLuint *maxWidth, GLuint *maxHeight, GLuint *maxDepth) +i915_max_texture_size(struct pipe_context *pipe, unsigned textureType, + unsigned *maxWidth, unsigned *maxHeight, unsigned *maxDepth) { switch (textureType) { case PIPE_TEXTURE_1D: @@ -160,9 +161,9 @@ static void i915_draw_vb( struct pipe_context *pipe, static void i915_draw_vertices(struct pipe_context *pipe, - GLuint mode, - GLuint numVertex, const GLfloat *verts, - GLuint numAttribs, const GLuint attribs[]) + unsigned mode, + unsigned numVertex, const float *verts, + unsigned numAttribs, const unsigned attribs[]) { struct i915_context *i915 = i915_context( pipe ); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 93b5e2546c..d69657daa9 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -82,10 +82,10 @@ struct i915_cache_context; */ struct i915_state { - GLuint immediate[I915_MAX_IMMEDIATE]; - GLuint dynamic[I915_MAX_DYNAMIC]; + unsigned immediate[I915_MAX_IMMEDIATE]; + unsigned dynamic[I915_MAX_DYNAMIC]; - GLuint id; /* track lost context events */ + unsigned id; /* track lost context events */ }; @@ -113,15 +113,15 @@ struct i915_context struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; - GLuint dirty; + unsigned dirty; - GLuint *batch_start; + unsigned *batch_start; struct i915_state current; - GLuint hardware_dirty; + unsigned hardware_dirty; - GLuint debug; - GLuint pci_id; + unsigned debug; + unsigned pci_id; struct { unsigned is_i945:1; @@ -179,7 +179,7 @@ unsigned *i915_passthrough_program( unsigned *dwords ); * i915_clear.c: */ void i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, - GLuint clearValue); + unsigned clearValue); /*********************************************************************** diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index 89257e9409..8b7e3628f3 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "imports.h" +//#include "imports.h" #include "i915_reg.h" #include "i915_context.h" @@ -35,15 +35,15 @@ #define PRINTF( ... ) (stream)->winsys->printf( (stream)->winsys, __VA_ARGS__ ) -static GLboolean debug( struct debug_stream *stream, const char *name, GLuint len ) +static boolean debug( struct debug_stream *stream, const char *name, unsigned len ) { - GLuint i; - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned i; + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); if (len == 0) { PRINTF("Error - zero length packet (0x%08x)\n", stream->ptr[0]); assert(0); - return GL_FALSE; + return FALSE; } if (stream->print_addresses) @@ -55,13 +55,13 @@ static GLboolean debug( struct debug_stream *stream, const char *name, GLuint le PRINTF("\t0x%08x\n", ptr[i]); PRINTF("\n"); - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); - return GL_TRUE; + return TRUE; } -static const char *get_prim_name( GLuint val ) +static const char *get_prim_name( unsigned val ) { switch (val & PRIM3D_MASK) { case PRIM3D_TRILIST: return "TRILIST"; break; @@ -80,13 +80,13 @@ static const char *get_prim_name( GLuint val ) } } -static GLboolean debug_prim( struct debug_stream *stream, const char *name, - GLboolean dump_floats, - GLuint len ) +static boolean debug_prim( struct debug_stream *stream, const char *name, + boolean dump_floats, + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); const char *prim = get_prim_name( ptr[0] ); - GLuint i; + unsigned i; @@ -94,7 +94,7 @@ static GLboolean debug_prim( struct debug_stream *stream, const char *name, PRINTF("\t0x%08x\n", ptr[0]); for (i = 1; i < len; i++) { if (dump_floats) - PRINTF("\t0x%08x // %f\n", ptr[i], *(GLfloat *)&ptr[i]); + PRINTF("\t0x%08x // %f\n", ptr[i], *(float *)&ptr[i]); else PRINTF("\t0x%08x\n", ptr[i]); } @@ -102,22 +102,22 @@ static GLboolean debug_prim( struct debug_stream *stream, const char *name, PRINTF("\n"); - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); - return GL_TRUE; + return TRUE; } -static GLboolean debug_program( struct debug_stream *stream, const char *name, GLuint len ) +static boolean debug_program( struct debug_stream *stream, const char *name, unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); if (len == 0) { PRINTF("Error - zero length packet (0x%08x)\n", stream->ptr[0]); assert(0); - return GL_FALSE; + return FALSE; } if (stream->print_addresses) @@ -126,16 +126,16 @@ static GLboolean debug_program( struct debug_stream *stream, const char *name, G PRINTF("%s (%d dwords):\n", name, len); i915_disassemble_program( stream, ptr, len ); - stream->offset += len * sizeof(GLuint); - return GL_TRUE; + stream->offset += len * sizeof(unsigned); + return TRUE; } -static GLboolean debug_chain( struct debug_stream *stream, const char *name, GLuint len ) +static boolean debug_chain( struct debug_stream *stream, const char *name, unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); - GLuint old_offset = stream->offset + len * sizeof(GLuint); - GLuint i; + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned old_offset = stream->offset + len * sizeof(unsigned); + unsigned i; PRINTF("%s (%d dwords):\n", name, len); for (i = 0; i < len; i++) @@ -151,17 +151,17 @@ static GLboolean debug_chain( struct debug_stream *stream, const char *name, GLu old_offset, stream->offset ); - return GL_TRUE; + return TRUE; } -static GLboolean debug_variable_length_prim( struct debug_stream *stream ) +static boolean debug_variable_length_prim( struct debug_stream *stream ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); const char *prim = get_prim_name( ptr[0] ); - GLuint i, len; + unsigned i, len; - GLushort *idx = (GLushort *)(ptr+1); + ushort *idx = (ushort *)(ptr+1); for (i = 0; idx[i] != 0xffff; i++) ; @@ -172,8 +172,8 @@ static GLboolean debug_variable_length_prim( struct debug_stream *stream ) PRINTF("\t0x%08x\n", ptr[i]); PRINTF("\n"); - stream->offset += len * sizeof(GLuint); - return GL_TRUE; + stream->offset += len * sizeof(unsigned); + return TRUE; } @@ -202,13 +202,13 @@ do { \ } \ } while (0) -static GLboolean debug_load_immediate( struct debug_stream *stream, +static boolean debug_load_immediate( struct debug_stream *stream, const char *name, - GLuint len ) + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); - GLuint bits = (ptr[0] >> 4) & 0xff; - GLuint j = 0; + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned bits = (ptr[0] >> 4) & 0xff; + unsigned j = 0; PRINTF("%s (%d dwords, flags: %x):\n", name, len, bits); PRINTF("\t0x%08x\n", ptr[j++]); @@ -302,20 +302,20 @@ static GLboolean debug_load_immediate( struct debug_stream *stream, assert(j == len); - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); - return GL_TRUE; + return TRUE; } -static GLboolean debug_load_indirect( struct debug_stream *stream, +static boolean debug_load_indirect( struct debug_stream *stream, const char *name, - GLuint len ) + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); - GLuint bits = (ptr[0] >> 8) & 0x3f; - GLuint i, j = 0; + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned bits = (ptr[0] >> 8) & 0x3f; + unsigned i, j = 0; PRINTF("%s (%d dwords):\n", name, len); PRINTF("\t0x%08x\n", ptr[j++]); @@ -362,13 +362,13 @@ static GLboolean debug_load_indirect( struct debug_stream *stream, assert(j == len); - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); - return GL_TRUE; + return TRUE; } static void BR13( struct debug_stream *stream, - GLuint val ) + unsigned val ) { PRINTF("\t0x%08x\n", val); FLAG(val, 30, "clipping enable"); @@ -379,7 +379,7 @@ static void BR13( struct debug_stream *stream, static void BR22( struct debug_stream *stream, - GLuint val ) + unsigned val ) { PRINTF("\t0x%08x\n", val); BITS(val, 31, 16, "dest y1"); @@ -387,7 +387,7 @@ static void BR22( struct debug_stream *stream, } static void BR23( struct debug_stream *stream, - GLuint val ) + unsigned val ) { PRINTF("\t0x%08x\n", val); BITS(val, 31, 16, "dest y2"); @@ -395,13 +395,13 @@ static void BR23( struct debug_stream *stream, } static void BR09( struct debug_stream *stream, - GLuint val ) + unsigned val ) { PRINTF("\t0x%08x -- dest address\n", val); } static void BR26( struct debug_stream *stream, - GLuint val ) + unsigned val ) { PRINTF("\t0x%08x\n", val); BITS(val, 31, 16, "src y1"); @@ -409,29 +409,29 @@ static void BR26( struct debug_stream *stream, } static void BR11( struct debug_stream *stream, - GLuint val ) + unsigned val ) { PRINTF("\t0x%08x\n", val); BITS(val, 15, 0, "src pitch"); } static void BR12( struct debug_stream *stream, - GLuint val ) + unsigned val ) { PRINTF("\t0x%08x -- src address\n", val); } static void BR16( struct debug_stream *stream, - GLuint val ) + unsigned val ) { PRINTF("\t0x%08x -- color\n", val); } -static GLboolean debug_copy_blit( struct debug_stream *stream, +static boolean debug_copy_blit( struct debug_stream *stream, const char *name, - GLuint len ) + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; PRINTF("%s (%d dwords):\n", name, len); @@ -445,16 +445,16 @@ static GLboolean debug_copy_blit( struct debug_stream *stream, BR11(stream, ptr[j++]); BR12(stream, ptr[j++]); - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); assert(j == len); - return GL_TRUE; + return TRUE; } -static GLboolean debug_color_blit( struct debug_stream *stream, +static boolean debug_color_blit( struct debug_stream *stream, const char *name, - GLuint len ) + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; PRINTF("%s (%d dwords):\n", name, len); @@ -466,16 +466,16 @@ static GLboolean debug_color_blit( struct debug_stream *stream, BR09(stream, ptr[j++]); BR16(stream, ptr[j++]); - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); assert(j == len); - return GL_TRUE; + return TRUE; } -static GLboolean debug_modes4( struct debug_stream *stream, +static boolean debug_modes4( struct debug_stream *stream, const char *name, - GLuint len ) + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; PRINTF("%s (%d dwords):\n", name, len); @@ -487,16 +487,16 @@ static GLboolean debug_modes4( struct debug_stream *stream, BITS(ptr[j], 7, 0, "stencil write mask"); j++; - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); assert(j == len); - return GL_TRUE; + return TRUE; } -static GLboolean debug_map_state( struct debug_stream *stream, +static boolean debug_map_state( struct debug_stream *stream, const char *name, - GLuint len ) + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; PRINTF("%s (%d dwords):\n", name, len); @@ -539,16 +539,16 @@ static GLboolean debug_map_state( struct debug_stream *stream, } } - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); assert(j == len); - return GL_TRUE; + return TRUE; } -static GLboolean debug_sampler_state( struct debug_stream *stream, +static boolean debug_sampler_state( struct debug_stream *stream, const char *name, - GLuint len ) + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; PRINTF("%s (%d dwords):\n", name, len); @@ -599,16 +599,16 @@ static GLboolean debug_sampler_state( struct debug_stream *stream, } } - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); assert(j == len); - return GL_TRUE; + return TRUE; } -static GLboolean debug_dest_vars( struct debug_stream *stream, +static boolean debug_dest_vars( struct debug_stream *stream, const char *name, - GLuint len ) + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; PRINTF("%s (%d dwords):\n", name, len); @@ -634,16 +634,16 @@ static GLboolean debug_dest_vars( struct debug_stream *stream, j++; } - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); assert(j == len); - return GL_TRUE; + return TRUE; } -static GLboolean debug_buf_info( struct debug_stream *stream, +static boolean debug_buf_info( struct debug_stream *stream, const char *name, - GLuint len ) + unsigned len ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; PRINTF("%s (%d dwords):\n", name, len); @@ -664,15 +664,15 @@ static GLboolean debug_buf_info( struct debug_stream *stream, PRINTF("\t0x%08x -- buffer base address\n", ptr[j++]); - stream->offset += len * sizeof(GLuint); + stream->offset += len * sizeof(unsigned); assert(j == len); - return GL_TRUE; + return TRUE; } -static GLboolean i915_debug_packet( struct debug_stream *stream ) +static boolean i915_debug_packet( struct debug_stream *stream ) { - GLuint *ptr = (GLuint *)(stream->ptr + stream->offset); - GLuint cmd = *ptr; + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned cmd = *ptr; switch (((cmd >> 29) & 0x7)) { case 0x0: @@ -685,7 +685,7 @@ static GLboolean i915_debug_packet( struct debug_stream *stream ) return debug(stream, "MI_FLUSH", 1); case 0xA: debug(stream, "MI_BATCH_BUFFER_END", 1); - return GL_FALSE; + return FALSE; case 0x22: return debug(stream, "MI_LOAD_REGISTER_IMM", 3); case 0x31: @@ -816,12 +816,12 @@ static GLboolean i915_debug_packet( struct debug_stream *stream ) void i915_dump_batchbuffer( struct i915_context *i915, - GLuint *start, - GLuint *end ) + unsigned *start, + unsigned *end ) { struct debug_stream stream; - GLuint bytes = (end - start) * 4; - GLboolean done = GL_FALSE; + unsigned bytes = (end - start) * 4; + boolean done = FALSE; stream.offset = 0; diff --git a/src/mesa/pipe/i915simple/i915_debug_fp.c b/src/mesa/pipe/i915simple/i915_debug_fp.c index c54d36f294..a108f1ad4f 100644 --- a/src/mesa/pipe/i915simple/i915_debug_fp.c +++ b/src/mesa/pipe/i915simple/i915_debug_fp.c @@ -25,15 +25,12 @@ * **************************************************************************/ -#include #include "i915_reg.h" #include "i915_debug.h" #include "pipe/p_winsys.h" -//#include "i915_fpc.h" -#include "shader/program.h" -#include "shader/prog_instruction.h" -#include "shader/prog_print.h" +#include "pipe/p_util.h" + @@ -126,7 +123,7 @@ static const char *regname[0x8] = { }; static void -print_reg_type_nr(struct debug_stream *stream, GLuint type, GLuint nr) +print_reg_type_nr(struct debug_stream *stream, unsigned type, unsigned nr) { switch (type) { case REG_TYPE_T: @@ -173,7 +170,7 @@ print_reg_type_nr(struct debug_stream *stream, GLuint type, GLuint nr) static void -print_reg_neg_swizzle(struct debug_stream *stream, GLuint reg) +print_reg_neg_swizzle(struct debug_stream *stream, unsigned reg) { int i; @@ -215,20 +212,20 @@ print_reg_neg_swizzle(struct debug_stream *stream, GLuint reg) static void -print_src_reg(struct debug_stream *stream, GLuint dword) +print_src_reg(struct debug_stream *stream, unsigned dword) { - GLuint nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; - GLuint type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; + unsigned nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; + unsigned type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; print_reg_type_nr(stream, type, nr); print_reg_neg_swizzle(stream, dword); } static void -print_dest_reg(struct debug_stream *stream, GLuint dword) +print_dest_reg(struct debug_stream *stream, unsigned dword) { - GLuint nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; - GLuint type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; + unsigned nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; + unsigned type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; print_reg_type_nr(stream, type, nr); if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) return; @@ -251,7 +248,7 @@ print_dest_reg(struct debug_stream *stream, GLuint dword) static void print_arith_op(struct debug_stream *stream, - GLuint opcode, const GLuint * program) + unsigned opcode, const unsigned * program) { if (opcode != A0_NOP) { print_dest_reg(stream, program[0]); @@ -285,7 +282,7 @@ print_arith_op(struct debug_stream *stream, static void print_tex_op(struct debug_stream *stream, - GLuint opcode, const GLuint * program) + unsigned opcode, const unsigned * program) { print_dest_reg(stream, program[0] | A0_DEST_CHANNEL_ALL); PRINTF(" = "); @@ -303,7 +300,7 @@ print_tex_op(struct debug_stream *stream, static void print_dcl_op(struct debug_stream *stream, - GLuint opcode, const GLuint * program) + unsigned opcode, const unsigned * program) { PRINTF("%s ", opcodes[opcode]); print_dest_reg(stream, @@ -314,10 +311,10 @@ print_dcl_op(struct debug_stream *stream, void i915_disassemble_program(struct debug_stream *stream, - const GLuint * program, GLuint sz) + const unsigned * program, unsigned sz) { - GLuint size = program[0] & 0x1ff; - GLint i; + unsigned size = program[0] & 0x1ff; + int i; PRINTF("\t\tBEGIN\n"); @@ -325,11 +322,11 @@ i915_disassemble_program(struct debug_stream *stream, program++; for (i = 1; i < sz; i += 3, program += 3) { - GLuint opcode = program[0] & (0x1f << 24); + unsigned opcode = program[0] & (0x1f << 24); PRINTF("\t\t"); - if ((GLint) opcode >= A0_NOP && opcode <= A0_SLT) + if ((int) opcode >= A0_NOP && opcode <= A0_SLT) print_arith_op(stream, opcode >> 24, program); else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) print_tex_op(stream, opcode >> 24, program); diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index 5173fc0b38..5a18d03e50 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -26,10 +26,11 @@ **************************************************************************/ -#include "imports.h" -#include "macros.h" +//#include "imports.h" +//#include "macros.h" #include "pipe/draw/draw_private.h" +#include "pipe/p_util.h" #include "i915_context.h" #include "i915_winsys.h" @@ -59,34 +60,6 @@ static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) return (struct setup_stage *)stage; } -static INLINE unsigned pack_ub4( unsigned char b0, - unsigned char b1, - unsigned char b2, - unsigned char b3 ) -{ - return ((((unsigned int)b0) << 0) | - (((unsigned int)b1) << 8) | - (((unsigned int)b2) << 16) | - (((unsigned int)b3) << 24)); -} - -static INLINE unsigned fui( float f ) -{ - union { - float f; - unsigned ui; - } fi; - - fi.f = f; - return fi.ui; -} - -static INLINE unsigned char float_to_ubyte( float f ) -{ - unsigned char ub; - UNCLAMPED_FLOAT_TO_UBYTE(ub, f); - return ub; -} /* Hardcoded vertex format: xyz/rgba diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index e27443a58c..aad4a6a537 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -44,7 +44,7 @@ i915_region_idle(struct pipe_context *pipe, struct pipe_region *region) } -static GLubyte * +static ubyte * i915_region_map(struct pipe_context *pipe, struct pipe_region *region) { struct i915_context *i915 = i915_context( pipe ); @@ -73,7 +73,7 @@ i915_region_unmap(struct pipe_context *pipe, struct pipe_region *region) static struct pipe_region * i915_region_alloc(struct pipe_context *pipe, - GLuint cpp, GLuint width, GLuint height, GLbitfield flags) + unsigned cpp, unsigned width, unsigned height, unsigned flags) { struct i915_context *i915 = i915_context( pipe ); struct pipe_region *region = calloc(sizeof(*region), 1); @@ -117,7 +117,7 @@ i915_region_release(struct pipe_context *pipe, struct pipe_region **region) if (!*region) return; - ASSERT((*region)->refcount > 0); + assert((*region)->refcount > 0); (*region)->refcount--; if ((*region)->refcount == 0) { @@ -135,19 +135,19 @@ i915_region_release(struct pipe_context *pipe, struct pipe_region **region) * XXX Move this into core Mesa? */ static void -_mesa_copy_rect(GLubyte * dst, - GLuint cpp, - GLuint dst_pitch, - GLuint dst_x, - GLuint dst_y, - GLuint width, - GLuint height, - const GLubyte * src, - GLuint src_pitch, - GLuint src_x, - GLuint src_y) +_mesa_copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + unsigned src_pitch, + unsigned src_x, + unsigned src_y) { - GLuint i; + unsigned i; dst_pitch *= cpp; src_pitch *= cpp; @@ -179,10 +179,10 @@ _mesa_copy_rect(GLubyte * dst, static void i915_region_data(struct pipe_context *pipe, struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - const void *src, GLuint src_pitch, - GLuint srcx, GLuint srcy, GLuint width, GLuint height) + unsigned dst_offset, + unsigned dstx, unsigned dsty, + const void *src, unsigned src_pitch, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) { _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, dst->cpp, @@ -199,11 +199,11 @@ i915_region_data(struct pipe_context *pipe, static void i915_region_copy(struct pipe_context *pipe, struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, + unsigned dst_offset, + unsigned dstx, unsigned dsty, struct pipe_region *src, - GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height) + unsigned src_offset, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) { assert( dst != src ); assert( dst->cpp == src->cpp ); @@ -233,8 +233,8 @@ i915_region_copy(struct pipe_context *pipe, /* Fill a rectangular sub-region. Need better logic about when to * push buffers into AGP - will currently do so whenever possible. */ -static GLubyte * -get_pointer(struct pipe_region *dst, GLuint x, GLuint y) +static ubyte * +get_pointer(struct pipe_region *dst, unsigned x, unsigned y) { return dst->map + (y * dst->pitch + x) * dst->cpp; } @@ -243,18 +243,18 @@ get_pointer(struct pipe_region *dst, GLuint x, GLuint y) static void i915_region_fill(struct pipe_context *pipe, struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - GLuint width, GLuint height, GLuint value) + unsigned dst_offset, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) { if (0) { - GLuint i, j; + unsigned i, j; (void)pipe->region_map(pipe, dst); switch (dst->cpp) { case 1: { - GLubyte *row = get_pointer(dst, dstx, dsty); + ubyte *row = get_pointer(dst, dstx, dsty); for (i = 0; i < height; i++) { memset(row, value, width); row += dst->pitch; @@ -262,7 +262,7 @@ i915_region_fill(struct pipe_context *pipe, } break; case 2: { - GLushort *row = (GLushort *) get_pointer(dst, dstx, dsty); + ushort *row = (ushort *) get_pointer(dst, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = value; @@ -271,7 +271,7 @@ i915_region_fill(struct pipe_context *pipe, } break; case 4: { - GLuint *row = (GLuint *) get_pointer(dst, dstx, dsty); + unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = value; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 22a5bf68b4..f874b21016 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -27,7 +27,7 @@ /* Authors: Keith Whitwell */ -#include "imports.h" +//#include "imports.h" #include "pipe/draw/draw_context.h" @@ -122,7 +122,7 @@ static void i915_set_fs_state( struct pipe_context *pipe, static void i915_set_sampler_state(struct pipe_context *pipe, - GLuint unit, + unsigned unit, const struct pipe_sampler_state *sampler) { struct i915_context *i915 = i915_context(pipe); @@ -135,7 +135,7 @@ static void i915_set_sampler_state(struct pipe_context *pipe, static void i915_set_texture_state(struct pipe_context *pipe, - GLuint unit, + unsigned unit, struct pipe_mipmap_tree *texture) { struct i915_context *i915 = i915_context(pipe); diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 7d03ed5567..a01b193676 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -43,7 +43,7 @@ do { \ } while (0) -static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = +static const unsigned frag_to_vf[FRAG_ATTRIB_MAX] = { VF_ATTRIB_POS, VF_ATTRIB_COLOR0, @@ -75,12 +75,12 @@ static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = */ static void calculate_vertex_layout( struct i915_context *i915 ) { -// const GLbitfield inputsRead = i915->fs.inputs_read; - const GLbitfield inputsRead = (FRAG_BIT_WPOS | FRAG_BIT_COL0); - GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; - GLbitfield attr_mask = 0x0; - GLuint nr_attrs = 0; - GLuint i; +// const unsigned inputsRead = i915->fs.inputs_read; + const unsigned inputsRead = (FRAG_BIT_WPOS | FRAG_BIT_COL0); + unsigned slot_to_vf_attr[VF_ATTRIB_MAX]; + unsigned attr_mask = 0x0; + unsigned nr_attrs = 0; + unsigned i; memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index 8d5a398890..84bff17b2f 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -25,16 +25,17 @@ * **************************************************************************/ -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "enums.h" +//#include "glheader.h" +//#include "context.h" +//#include "macros.h" +//#include "enums.h" #include "i915_batch.h" #include "i915_state_inlines.h" #include "i915_context.h" #include "i915_reg.h" #include "i915_state.h" +#include "pipe/p_util.h" #define FILE_DEBUG_FLAG DEBUG_STATE @@ -50,9 +51,9 @@ */ static inline void set_dynamic_indirect( struct i915_context *i915, - GLuint offset, - const GLuint *src, - GLuint dwords ) + unsigned offset, + const unsigned *src, + unsigned dwords ) { int i; @@ -68,12 +69,12 @@ static inline void set_dynamic_indirect( struct i915_context *i915, */ static void upload_MODES4( struct i915_context *i915 ) { - GLuint modes4 = 0; + unsigned modes4 = 0; /* I915_NEW_STENCIL */ { - GLint testmask = i915->stencil.value_mask[0] & 0xff; - GLint writemask = i915->stencil.write_mask[0] & 0xff; + int testmask = i915->stencil.value_mask[0] & 0xff; + int writemask = i915->stencil.write_mask[0] & 0xff; modes4 |= (_3DSTATE_MODES_4_CMD | ENABLE_STENCIL_TEST_MASK | @@ -110,20 +111,20 @@ const struct i915_tracked_state i915_upload_MODES4 = { static void upload_BFO( struct i915_context *i915 ) { - GLuint bf[2]; + unsigned bf[2]; memset( bf, 0, sizeof(bf) ); /* _NEW_STENCIL */ if (i915->stencil.back_enabled) { - GLint test = i915_translate_compare_func(i915->stencil.back_func); - GLint fop = i915_translate_stencil_op(i915->stencil.back_fail_op); - GLint dfop = i915_translate_stencil_op(i915->stencil.back_zfail_op); - GLint dpop = i915_translate_stencil_op(i915->stencil.back_zpass_op); - GLint ref = i915->stencil.ref_value[1] & 0xff; - GLint tmask = i915->stencil.value_mask[1] & 0xff; - GLint wmask = i915->stencil.write_mask[1] & 0xff; + int test = i915_translate_compare_func(i915->stencil.back_func); + int fop = i915_translate_stencil_op(i915->stencil.back_fail_op); + int dfop = i915_translate_stencil_op(i915->stencil.back_zfail_op); + int dpop = i915_translate_stencil_op(i915->stencil.back_zpass_op); + int ref = i915->stencil.ref_value[1] & 0xff; + int tmask = i915->stencil.value_mask[1] & 0xff; + int wmask = i915->stencil.write_mask[1] & 0xff; bf[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_FUNCS | @@ -172,23 +173,20 @@ const struct i915_tracked_state i915_upload_BFO = { static void upload_BLENDCOLOR( struct i915_context *i915 ) { - GLuint bc[2]; + unsigned bc[2]; memset( bc, 0, sizeof(bc) ); /* I915_NEW_BLEND {_COLOR} */ { - const GLfloat *color = i915->blend_color.color; - GLubyte r, g, b, a; + const float *color = i915->blend_color.color; - UNCLAMPED_FLOAT_TO_UBYTE(r, color[RCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(g, color[GCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(b, color[BCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(a, color[ACOMP]); - - bc[0] = (_3DSTATE_CONST_BLEND_COLOR_CMD); - bc[1] = (a << 24) | (r << 16) | (g << 8) | b; + bc[0] = _3DSTATE_CONST_BLEND_COLOR_CMD; + bc[1] = pack_ui32_float4( color[0], + color[1], + color[2], + color[3] ); } set_dynamic_indirect( i915, @@ -208,25 +206,21 @@ const struct i915_tracked_state i915_upload_BLENDCOLOR = { static void upload_IAB( struct i915_context *i915 ) { - GLuint iab = 0; + unsigned iab = 0; { - GLuint eqRGB = i915->blend.rgb_func; - GLuint srcRGB = i915->blend.rgb_src_factor; - GLuint dstRGB = i915->blend.rgb_dst_factor; + unsigned eqRGB = i915->blend.rgb_func; + unsigned srcRGB = i915->blend.rgb_src_factor; + unsigned dstRGB = i915->blend.rgb_dst_factor; - GLuint eqA = i915->blend.alpha_func; - GLuint srcA = i915->blend.alpha_src_factor; - GLuint dstA = i915->blend.alpha_dst_factor; + unsigned eqA = i915->blend.alpha_func; + unsigned srcA = i915->blend.alpha_src_factor; + unsigned dstA = i915->blend.alpha_dst_factor; - if (eqA == GL_MIN || eqA == GL_MAX) { - srcA = dstA = GL_ONE; - } + /* Special handling for MIN/MAX filter modes handled at + * state_tracker level. + */ - if (eqRGB == GL_MIN || eqRGB == GL_MAX) { - srcRGB = dstRGB = GL_ONE; - } - if (srcA != srcRGB || dstA != dstRGB || eqA != eqRGB) { @@ -268,7 +262,7 @@ const struct i915_tracked_state i915_upload_IAB = { static void upload_DEPTHSCALE( struct i915_context *i915 ) { - union { GLfloat f; GLuint u; } ds[2]; + union { float f; unsigned u; } ds[2]; memset( ds, 0, sizeof(ds) ); @@ -304,7 +298,7 @@ const struct i915_tracked_state i915_upload_DEPTHSCALE = { static void upload_STIPPLE( struct i915_context *i915 ) { - GLuint st[2]; + unsigned st[2]; st[0] = _3DSTATE_STIPPLE; st[1] = 0; @@ -319,8 +313,8 @@ static void upload_STIPPLE( struct i915_context *i915 ) /* I915_NEW_STIPPLE */ { - const GLubyte *mask = (const GLubyte *)i915->poly_stipple.stipple; - GLubyte p[4]; + const ubyte *mask = (const ubyte *)i915->poly_stipple.stipple; + ubyte p[4]; p[0] = mask[12] & 0xf; p[1] = mask[8] & 0xf; diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 2ce079fbaa..a7ae92d93c 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -27,8 +27,8 @@ -#include "glheader.h" -#include "mtypes.h" +//#include "glheader.h" +//#include "mtypes.h" #include "i915_reg.h" #include "i915_context.h" @@ -204,8 +204,8 @@ i915_emit_hardware_state(struct i915_context *i915 ) { - GLuint i, dwords; - GLuint *prog = i915_passthrough_program( &dwords ); + unsigned i, dwords; + unsigned *prog = i915_passthrough_program( &dwords ); for (i = 0; i < dwords; i++) OUT_BATCH( prog[i] ); diff --git a/src/mesa/pipe/i915simple/i915_state_fragprog.c b/src/mesa/pipe/i915simple/i915_state_fragprog.c index 163883b07a..51df86e0b1 100644 --- a/src/mesa/pipe/i915simple/i915_state_fragprog.c +++ b/src/mesa/pipe/i915simple/i915_state_fragprog.c @@ -28,7 +28,7 @@ #include "i915_reg.h" #include "i915_context.h" #include "i915_debug.h" - +#include "pipe/p_util.h" static unsigned passthrough[] = { diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 2f64d9dd90..d06fac2e4c 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -29,12 +29,13 @@ * Keith Whitwell */ -#include "macros.h" +//#include "macros.h" #include "i915_state_inlines.h" #include "i915_context.h" #include "i915_state.h" #include "i915_reg.h" +#include "p_util.h" /* All state expressable with the LOAD_STATE_IMMEDIATE_1 packet. @@ -51,7 +52,7 @@ */ static void upload_S2S4(struct i915_context *i915) { - GLuint LIS2, LIS4; + unsigned LIS2, LIS4; /* I915_NEW_VERTEX_FORMAT */ LIS2 = 0xffffffff; @@ -86,7 +87,7 @@ static void upload_S2S4(struct i915_context *i915) /* I915_NEW_SETUP */ { - GLint point_size = CLAMP((int) i915->setup.point_size, 1, 0xff); + int point_size = CLAMP((int) i915->setup.point_size, 1, 0xff); LIS4 |= point_size << S4_POINT_WIDTH_SHIFT; } @@ -121,15 +122,15 @@ const struct i915_tracked_state i915_upload_S2S4 = { */ static void upload_S5( struct i915_context *i915 ) { - GLuint LIS5 = 0; + unsigned LIS5 = 0; /* I915_NEW_STENCIL */ if (i915->stencil.front_enabled) { - GLint test = i915_translate_compare_func(i915->stencil.front_func); - GLint fop = i915_translate_stencil_op(i915->stencil.front_fail_op); - GLint dfop = i915_translate_stencil_op(i915->stencil.front_zfail_op); - GLint dpop = i915_translate_stencil_op(i915->stencil.front_zpass_op); - GLint ref = i915->stencil.ref_value[0] & 0xff; + int test = i915_translate_compare_func(i915->stencil.front_func); + int fop = i915_translate_stencil_op(i915->stencil.front_fail_op); + int dfop = i915_translate_stencil_op(i915->stencil.front_zfail_op); + int dpop = i915_translate_stencil_op(i915->stencil.front_zpass_op); + int ref = i915->stencil.ref_value[0] & 0xff; LIS5 |= (S5_STENCIL_TEST_ENABLE | S5_STENCIL_WRITE_ENABLE | @@ -184,16 +185,14 @@ const struct i915_tracked_state i915_upload_S5 = { */ static void upload_S6( struct i915_context *i915 ) { - GLuint LIS6 = (S6_COLOR_WRITE_ENABLE | + unsigned LIS6 = (S6_COLOR_WRITE_ENABLE | (2 << S6_TRISTRIP_PV_SHIFT)); /* I915_NEW_ALPHA_TEST */ if (i915->alpha_test.enabled) { int test = i915_translate_compare_func(i915->alpha_test.func); - GLubyte refByte; - - CLAMPED_FLOAT_TO_UBYTE(refByte, i915->alpha_test.ref); + ubyte refByte = float_to_ubyte(i915->alpha_test.ref); LIS6 |= (S6_ALPHA_TEST_ENABLE | @@ -205,9 +204,9 @@ static void upload_S6( struct i915_context *i915 ) */ if (i915->blend.blend_enable) { - GLuint funcRGB = i915->blend.rgb_func; - GLuint srcRGB = i915->blend.rgb_src_factor; - GLuint dstRGB = i915->blend.rgb_dst_factor; + unsigned funcRGB = i915->blend.rgb_func; + unsigned srcRGB = i915->blend.rgb_src_factor; + unsigned dstRGB = i915->blend.rgb_dst_factor; LIS6 |= (S6_CBUF_BLEND_ENABLE | SRC_BLND_FACT(i915_translate_blend_factor(srcRGB)) | @@ -218,7 +217,7 @@ static void upload_S6( struct i915_context *i915 ) /* I915_NEW_DEPTH */ if (i915->depth_test.enabled) { - GLint func = i915_translate_compare_func(i915->depth_test.func); + int func = i915_translate_compare_func(i915->depth_test.func); LIS6 |= (S6_DEPTH_TEST_ENABLE | (func << S6_DEPTH_TEST_FUNC_SHIFT)); @@ -243,7 +242,7 @@ const struct i915_tracked_state i915_upload_S6 = { */ static void upload_S7( struct i915_context *i915 ) { - GLfloat LIS7; + float LIS7; /* I915_NEW_SETUP */ diff --git a/src/mesa/pipe/i915simple/i915_state_inlines.h b/src/mesa/pipe/i915simple/i915_state_inlines.h index 28aff0ebd4..0934ac79a4 100644 --- a/src/mesa/pipe/i915simple/i915_state_inlines.h +++ b/src/mesa/pipe/i915simple/i915_state_inlines.h @@ -28,6 +28,7 @@ #ifndef I915_STATE_INLINES_H #define I915_STATE_INLINES_H +#include "p_compiler.h" #include "p_defines.h" #include "i915_reg.h" @@ -184,9 +185,9 @@ i915_translate_logic_op(unsigned opcode) -static INLINE GLboolean i915_validate_vertices( GLuint hw_prim, GLuint nr ) +static INLINE boolean i915_validate_vertices( unsigned hw_prim, unsigned nr ) { - GLboolean ok; + boolean ok; switch (hw_prim) { case PRIM3D_POINTLIST: @@ -226,5 +227,4 @@ static INLINE GLboolean i915_validate_vertices( GLuint hw_prim, GLuint nr ) return ok; } - #endif diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index cc0a87012a..c094cf9ec4 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -28,8 +28,8 @@ #include "i915_context.h" #include "i915_state.h" #include "pipe/p_defines.h" -#include "main/imports.h" -#include "main/macros.h" +#include "pipe/p_util.h" +//#include "main/imports.h" struct i915_surface @@ -45,13 +45,13 @@ struct i915_surface */ static void i915_get_tile(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) + unsigned x, unsigned y, unsigned w, unsigned h, float *p) { - const GLuint *src - = ((const GLuint *) (ps->region->map + ps->offset)) + const unsigned *src + = ((const unsigned *) (ps->region->map + ps->offset)) + y * ps->region->pitch + x; - GLuint i, j; - GLuint w0 = w; + unsigned i, j; + unsigned w0 = w; assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); @@ -66,9 +66,9 @@ i915_get_tile(struct pipe_surface *ps, h = ps->height -y; #endif for (i = 0; i < h; i++) { - GLfloat *pRow = p; + float *pRow = p; for (j = 0; j < w; j++) { - const GLuint pixel = src[j]; + const unsigned pixel = src[j]; pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); @@ -83,7 +83,7 @@ i915_get_tile(struct pipe_surface *ps, static void i915_put_tile(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, const GLfloat *p) + unsigned x, unsigned y, unsigned w, unsigned h, const float *p) { /* any need to put tiles into i915 surfaces? */ assert(0); @@ -92,7 +92,7 @@ i915_put_tile(struct pipe_surface *ps, static struct pipe_surface * -i915_surface_alloc(struct pipe_context *pipe, GLuint format) +i915_surface_alloc(struct pipe_context *pipe, unsigned format) { struct i915_surface *surf = CALLOC_STRUCT(i915_surface); diff --git a/src/mesa/pipe/i915simple/i915_tex_layout.c b/src/mesa/pipe/i915simple/i915_tex_layout.c index d0c58de691..ec1fb9532d 100644 --- a/src/mesa/pipe/i915simple/i915_tex_layout.c +++ b/src/mesa/pipe/i915simple/i915_tex_layout.c @@ -30,13 +30,17 @@ * Michel Dänzer */ -#include "macros.h" +//#include "macros.h" #include "pipe/p_state.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" + #include "i915_tex_layout.h" +#include "i915_debug.h" -static GLuint minify( GLuint d ) +static unsigned minify( unsigned d ) { return MAX2(1, d>>1); } @@ -49,11 +53,11 @@ static int align(int value, int alignment) static void i915_miptree_set_level_info(struct pipe_mipmap_tree *mt, - GLuint level, - GLuint nr_images, - GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) + unsigned level, + unsigned nr_images, + unsigned x, unsigned y, unsigned w, unsigned h, unsigned d) { - assert(level < MAX_TEXTURE_LEVELS); + assert(level < PIPE_MAX_TEXTURE_LEVELS); mt->level[level].width = w; mt->level[level].height = h; @@ -76,14 +80,14 @@ i915_miptree_set_level_info(struct pipe_mipmap_tree *mt, assert(nr_images); assert(!mt->level[level].image_offset); - mt->level[level].image_offset = (GLuint *) malloc(nr_images * sizeof(GLuint)); + mt->level[level].image_offset = (unsigned *) malloc(nr_images * sizeof(unsigned)); mt->level[level].image_offset[0] = 0; } static void i915_miptree_set_image_offset(struct pipe_mipmap_tree *mt, - GLuint level, GLuint img, GLuint x, GLuint y) + unsigned level, unsigned img, unsigned x, unsigned y) { if (img == 0 && level == 0) assert(x == 0 && y == 0); @@ -102,12 +106,12 @@ i915_miptree_set_image_offset(struct pipe_mipmap_tree *mt, static void i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) { - GLint align_h = 2, align_w = 4; - GLuint level; - GLuint x = 0; - GLuint y = 0; - GLuint width = mt->width0; - GLuint height = mt->height0; + int align_h = 2, align_w = 4; + unsigned level; + unsigned x = 0; + unsigned y = 0; + unsigned width = mt->width0; + unsigned height = mt->height0; mt->pitch = mt->width0; @@ -117,7 +121,7 @@ i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) * 2nd mipmap out past the width of its parent. */ if (mt->first_level != mt->last_level) { - GLuint mip1_width = align(minify(mt->width0), align_w) + unsigned mip1_width = align(minify(mt->width0), align_w) + minify(minify(mt->width0)); if (mip1_width > mt->width0) @@ -131,7 +135,7 @@ i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) mt->total_height = 0; for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { - GLuint img_height; + unsigned img_height; i915_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); @@ -161,7 +165,7 @@ i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) } -static const GLint initial_offsets[6][2] = { +static const int initial_offsets[6][2] = { {0, 0}, {0, 2}, {1, 0}, @@ -170,7 +174,7 @@ static const GLint initial_offsets[6][2] = { {1, 3} }; -static const GLint step_offsets[6][2] = { +static const int step_offsets[6][2] = { {0, 2}, {0, 2}, {-1, 2}, @@ -180,16 +184,16 @@ static const GLint step_offsets[6][2] = { }; -GLboolean +boolean i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) { - GLint level; + int level; switch (mt->target) { - case GL_TEXTURE_CUBE_MAP:{ - const GLuint dim = mt->width0; - GLuint face; - GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + case PIPE_TEXTURE_CUBE: { + const unsigned dim = mt->width0; + unsigned face; + unsigned lvlWidth = mt->width0, lvlHeight = mt->height0; assert(lvlWidth == lvlHeight); /* cubemap images are square */ @@ -208,17 +212,12 @@ i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) } for (face = 0; face < 6; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; + unsigned x = initial_offsets[face][0] * dim; + unsigned y = initial_offsets[face][1] * dim; + unsigned d = dim; for (level = mt->first_level; level <= mt->last_level; level++) { i915_miptree_set_image_offset(mt, level, face, x, y); - - if (d == 0) - _mesa_printf("cube mipmap %d/%d (%d..%d) is 0x0\n", - face, level, mt->first_level, mt->last_level); - d >>= 1; x += step_offsets[face][0] * d; y += step_offsets[face][1] * d; @@ -226,11 +225,11 @@ i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) } break; } - case GL_TEXTURE_3D:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint stack_height = 0; + case PIPE_TEXTURE_3D:{ + unsigned width = mt->width0; + unsigned height = mt->height0; + unsigned depth = mt->depth0; + unsigned stack_height = 0; /* Calculate the size of a single slice. */ @@ -255,7 +254,7 @@ i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) */ depth = mt->depth0; for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint i; + unsigned i; for (i = 0; i < depth; i++) i915_miptree_set_image_offset(mt, level, i, 0, i * stack_height); @@ -273,9 +272,9 @@ i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) } default:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint img_height; + unsigned width = mt->width0; + unsigned height = mt->height0; + unsigned img_height; mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; mt->total_height = 0; @@ -304,20 +303,20 @@ i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); */ - return GL_TRUE; + return TRUE; } -GLboolean +boolean i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) { - GLint level; + int level; switch (mt->target) { - case GL_TEXTURE_CUBE_MAP:{ - const GLuint dim = mt->width0; - GLuint face; - GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + case PIPE_TEXTURE_CUBE:{ + const unsigned dim = mt->width0; + unsigned face; + unsigned lvlWidth = mt->width0, lvlHeight = mt->height0; assert(lvlWidth == lvlHeight); /* cubemap images are square */ @@ -344,9 +343,9 @@ i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) for (face = 0; face < 6; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; + unsigned x = initial_offsets[face][0] * dim; + unsigned y = initial_offsets[face][1] * dim; + unsigned d = dim; if (dim == 4 && face >= 4) { y = mt->total_height - 4; @@ -365,18 +364,18 @@ i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) switch (d) { case 4: switch (face) { - case FACE_POS_X: - case FACE_NEG_X: + case PIPE_TEX_FACE_POS_X: + case PIPE_TEX_FACE_NEG_X: x += step_offsets[face][0] * d; y += step_offsets[face][1] * d; break; - case FACE_POS_Y: - case FACE_NEG_Y: + case PIPE_TEX_FACE_POS_Y: + case PIPE_TEX_FACE_NEG_Y: y += 12; x -= 8; break; - case FACE_POS_Z: - case FACE_NEG_Z: + case PIPE_TEX_FACE_POS_Z: + case PIPE_TEX_FACE_NEG_Z: y = mt->total_height - 4; x = (face - 4) * 8; break; @@ -400,13 +399,13 @@ i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) } break; } - case GL_TEXTURE_3D:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint pack_x_pitch, pack_x_nr; - GLuint pack_y_pitch; - GLuint level; + case PIPE_TEXTURE_3D:{ + unsigned width = mt->width0; + unsigned height = mt->height0; + unsigned depth = mt->depth0; + unsigned pack_x_pitch, pack_x_nr; + unsigned pack_y_pitch; + unsigned level; mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; mt->total_height = 0; @@ -416,10 +415,10 @@ i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) pack_x_nr = 1; for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; - GLint x = 0; - GLint y = 0; - GLint q, j; + unsigned nr_images = mt->target == PIPE_TEXTURE_3D ? depth : 6; + int x = 0; + int y = 0; + int q, j; i915_miptree_set_level_info(mt, level, nr_images, 0, mt->total_height, @@ -455,13 +454,14 @@ i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) break; } - case GL_TEXTURE_1D: - case GL_TEXTURE_2D: - case GL_TEXTURE_RECTANGLE_ARB: + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_2D: +// case PIPE_TEXTURE_RECTANGLE: i945_miptree_layout_2d(mt); break; default: - _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()"); + assert(0); + return FALSE; } /* @@ -470,6 +470,6 @@ i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); */ - return GL_TRUE; + return TRUE; } diff --git a/src/mesa/pipe/i915simple/i915_tex_layout.h b/src/mesa/pipe/i915simple/i915_tex_layout.h index b1555f1dd8..e033786381 100644 --- a/src/mesa/pipe/i915simple/i915_tex_layout.h +++ b/src/mesa/pipe/i915simple/i915_tex_layout.h @@ -6,10 +6,10 @@ struct pipe_context; struct pipe_mipmap_tree; -extern GLboolean +extern boolean i915_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); -extern GLboolean +extern boolean i945_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 803ee9073c..10a8283035 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -28,7 +28,7 @@ #ifndef I915_WINSYS_H #define I915_WINSYS_H -#include "main/mtypes.h" +//#include "main/mtypes.h" /* This is the interface that softpipe requires any window system * hosting it to implement. This is the only include file in softpipe diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 84557b9a4a..7635755947 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -28,8 +28,9 @@ #ifndef PIPE_CONTEXT_H #define PIPE_CONTEXT_H -#include "main/mtypes.h" +//#include "main/mtypes.h" #include "p_state.h" +#include "p_compiler.h" /* Drawing currently kludged up via the existing tnl/ module. @@ -49,14 +50,14 @@ struct pipe_context { /* * Queries */ - const GLuint *(*supported_formats)(struct pipe_context *pipe, - GLuint *numFormats); + const unsigned *(*supported_formats)(struct pipe_context *pipe, + unsigned *numFormats); void (*max_texture_size)(struct pipe_context *pipe, - GLuint textureType, /* PIPE_TEXTURE_x */ - GLuint *maxWidth, - GLuint *maxHeight, - GLuint *maxDepth); + unsigned textureType, /* PIPE_TEXTURE_x */ + unsigned *maxWidth, + unsigned *maxHeight, + unsigned *maxDepth); const char *(*get_name)( struct pipe_context *pipe ); @@ -71,19 +72,19 @@ struct pipe_context { struct vertex_buffer *VB ); void (*draw_vertices)( struct pipe_context *pipe, - GLuint mode, - GLuint numVertex, const GLfloat *verts, - GLuint numAttribs, const GLuint attribs[]); + unsigned mode, + unsigned numVertex, const float *verts, + unsigned numAttribs, const unsigned attribs[]); /** Clear a surface to given value (no scissor; clear whole surface) */ void (*clear)(struct pipe_context *pipe, struct pipe_surface *ps, - GLuint clearValue); + unsigned clearValue); /** occlusion counting (XXX this may be temporary - we should probably * have generic query objects with begin/end methods) */ void (*reset_occlusion_counter)(struct pipe_context *pipe); - GLuint (*get_occlusion_counter)(struct pipe_context *pipe); + unsigned (*get_occlusion_counter)(struct pipe_context *pipe); /* * State functions @@ -125,11 +126,11 @@ struct pipe_context { const struct pipe_stencil_state * ); void (*set_sampler_state)( struct pipe_context *, - GLuint unit, + unsigned unit, const struct pipe_sampler_state * ); void (*set_texture_state)( struct pipe_context *, - GLuint unit, + unsigned unit, struct pipe_mipmap_tree * ); void (*set_viewport_state)( struct pipe_context *, @@ -141,58 +142,58 @@ struct pipe_context { * This might go away... */ struct pipe_surface *(*surface_alloc)(struct pipe_context *pipe, - GLuint format); + unsigned format); struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, struct pipe_mipmap_tree *texture, - GLuint face, GLuint level, - GLuint zslice); + unsigned face, unsigned level, + unsigned zslice); /* * Memory region functions * Some of these may go away... */ struct pipe_region *(*region_alloc)(struct pipe_context *pipe, - GLuint cpp, GLuint width, GLuint height, - GLbitfield flags); + unsigned cpp, unsigned width, unsigned height, + unsigned flags); void (*region_release)(struct pipe_context *pipe, struct pipe_region **r); void (*region_idle)(struct pipe_context *pipe, struct pipe_region *region); - GLubyte *(*region_map)(struct pipe_context *pipe, struct pipe_region *r); + ubyte *(*region_map)(struct pipe_context *pipe, struct pipe_region *r); void (*region_unmap)(struct pipe_context *pipe, struct pipe_region *r); void (*region_data)(struct pipe_context *pipe, struct pipe_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - const void *src, GLuint src_stride, - GLuint srcx, GLuint srcy, GLuint width, GLuint height); + unsigned dest_offset, + unsigned destx, unsigned desty, + const void *src, unsigned src_stride, + unsigned srcx, unsigned srcy, unsigned width, unsigned height); void (*region_copy)(struct pipe_context *pipe, struct pipe_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, + unsigned dest_offset, + unsigned destx, unsigned desty, struct pipe_region *src, /* don't make this const - need to map/unmap */ - GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height); + unsigned src_offset, + unsigned srcx, unsigned srcy, unsigned width, unsigned height); void (*region_fill)(struct pipe_context *pipe, struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - GLuint width, GLuint height, - GLuint value); + unsigned dst_offset, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, + unsigned value); /* * Texture functions */ - GLboolean (*mipmap_tree_layout)( struct pipe_context *pipe, - struct pipe_mipmap_tree *mt ); + boolean (*mipmap_tree_layout)( struct pipe_context *pipe, + struct pipe_mipmap_tree *mt ); /* Flush rendering: diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 52ecd5b119..e6aff6fb49 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -150,6 +150,14 @@ #define PIPE_TEX_COMPARE_NONE 0 #define PIPE_TEX_COMPARE_R_TO_TEXTURE 1 +#define PIPE_TEX_FACE_POS_X 0 +#define PIPE_TEX_FACE_NEG_X 1 +#define PIPE_TEX_FACE_POS_Y 2 +#define PIPE_TEX_FACE_NEG_Y 3 +#define PIPE_TEX_FACE_POS_Z 4 +#define PIPE_TEX_FACE_NEG_Z 5 +#define PIPE_TEX_FACE_MAX 6 + /** * Texture/surface image formats (preliminary) @@ -200,4 +208,6 @@ #define PIPE_FLUSH_RENDER_CACHE 0x1 #define PIPE_FLUSH_TEXTURE_CACHE 0x2 + + #endif diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 6799fdb31c..dc2b589e55 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -38,9 +38,7 @@ #ifndef PIPE_STATE_H #define PIPE_STATE_H -#include "mtypes.h" - - +#include "p_compiler.h" /** * Implementation limits @@ -50,6 +48,7 @@ #define PIPE_MAX_CONSTANT 32 #define PIPE_ATTRIB_MAX 32 #define PIPE_MAX_COLOR_BUFS 8 +#define PIPE_MAX_TEXTURE_LEVELS 16 /* fwd decl */ @@ -69,138 +68,129 @@ struct pipe_buffer_handle; */ struct pipe_setup_state { - GLuint flatshade:1; - GLuint light_twoside:1; - - GLuint front_winding:2; /**< PIPE_WINDING_x */ - - GLuint cull_mode:2; /**< PIPE_WINDING_x */ - - GLuint fill_cw:2; /**< PIPE_POLYGON_MODE_x */ - GLuint fill_ccw:2; /**< PIPE_POLYGON_MODE_x */ - - GLuint offset_cw:1; - GLuint offset_ccw:1; - - GLuint scissor:1; - - GLuint poly_smooth:1; - GLuint poly_stipple_enable:1; - - GLuint line_smooth:1; - GLuint line_stipple_enable:1; - - GLuint point_smooth:1; - - GLuint multisample:1; /* XXX maybe more ms state in future */ - - GLubyte line_stipple_factor; /**< [1..256] actually */ - GLushort line_stipple_pattern; - GLfloat line_width; - GLfloat point_size; /**< used when no per-vertex size */ - GLfloat offset_units; - GLfloat offset_scale; + unsigned flatshade:1; + unsigned light_twoside:1; + unsigned front_winding:2; /**< PIPE_WINDING_x */ + unsigned cull_mode:2; /**< PIPE_WINDING_x */ + unsigned fill_cw:2; /**< PIPE_POLYGON_MODE_x */ + unsigned fill_ccw:2; /**< PIPE_POLYGON_MODE_x */ + unsigned offset_cw:1; + unsigned offset_ccw:1; + unsigned scissor:1; + unsigned poly_smooth:1; + unsigned poly_stipple_enable:1; + unsigned point_smooth:1; + unsigned multisample:1; /* XXX maybe more ms state in future */ + unsigned line_smooth:1; + unsigned line_stipple_enable:1; + unsigned line_stipple_factor:8; /**< [1..256] actually */ + unsigned line_stipple_pattern:16; + + float line_width; + float point_size; /**< used when no per-vertex size */ + float offset_units; + float offset_scale; }; struct pipe_poly_stipple { - GLuint stipple[32]; + unsigned stipple[32]; }; struct pipe_viewport_state { - GLfloat scale[4]; - GLfloat translate[4]; + float scale[4]; + float translate[4]; }; struct pipe_scissor_state { - GLshort minx; - GLshort miny; - GLshort maxx; - GLshort maxy; + unsigned minx:16; + unsigned miny:16; + unsigned maxx:16; + unsigned maxy:16; }; struct pipe_clip_state { - GLfloat ucp[PIPE_MAX_CLIP_PLANES][4]; - GLuint nr; + float ucp[PIPE_MAX_CLIP_PLANES][4]; + unsigned nr; }; struct pipe_constant_buffer { - GLfloat constant[PIPE_MAX_CONSTANT][4]; - GLuint nr_constants; + float constant[PIPE_MAX_CONSTANT][4]; + unsigned nr_constants; }; struct pipe_fs_state { - GLbitfield inputs_read; /* FRAG_ATTRIB_* */ + unsigned inputs_read; /* FRAG_ATTRIB_* */ const struct tgsi_token *tokens; struct pipe_constant_buffer *constants; /* XXX temporary? */ }; struct pipe_depth_state { - GLuint enabled:1; /**< depth test enabled? */ - GLuint writemask:1; /**< allow depth buffer writes? */ - GLuint func:3; /**< depth test func (PIPE_FUNC_x) */ - GLuint occlusion_count:1; /**< XXX move this elsewhere? */ - GLfloat clear; /**< Clear value in [0,1] (XXX correct place?) */ + unsigned enabled:1; /**< depth test enabled? */ + unsigned writemask:1; /**< allow depth buffer writes? */ + unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ + unsigned occlusion_count:1; /**< XXX move this elsewhere? */ + float clear; /**< Clear value in [0,1] (XXX correct place?) */ }; struct pipe_alpha_test_state { - GLuint enabled:1; - GLuint func:3; /**< PIPE_FUNC_x */ - GLfloat ref; /**< reference value */ + unsigned enabled:1; + unsigned func:3; /**< PIPE_FUNC_x */ + float ref; /**< reference value */ }; struct pipe_blend_state { - GLuint blend_enable:1; + unsigned blend_enable:1; - GLuint rgb_func:3; /**< PIPE_BLEND_x */ - GLuint rgb_src_factor:5; /**< PIPE_BLENDFACTOR_x */ - GLuint rgb_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ + unsigned rgb_func:3; /**< PIPE_BLEND_x */ + unsigned rgb_src_factor:5; /**< PIPE_BLENDFACTOR_x */ + unsigned rgb_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ - GLuint alpha_func:3; /**< PIPE_BLEND_x */ - GLuint alpha_src_factor:5; /**< PIPE_BLENDFACTOR_x */ - GLuint alpha_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ + unsigned alpha_func:3; /**< PIPE_BLEND_x */ + unsigned alpha_src_factor:5; /**< PIPE_BLENDFACTOR_x */ + unsigned alpha_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ - GLuint logicop_enable:1; - GLuint logicop_func:4; /**< PIPE_LOGICOP_x */ + unsigned logicop_enable:1; + unsigned logicop_func:4; /**< PIPE_LOGICOP_x */ - GLuint colormask:4; /**< bitmask of PIPE_MASK_R/G/B/A */ - GLuint dither:1; + unsigned colormask:4; /**< bitmask of PIPE_MASK_R/G/B/A */ + unsigned dither:1; }; struct pipe_blend_color { - GLfloat color[4]; + float color[4]; }; struct pipe_clear_color_state { - GLfloat color[4]; + float color[4]; }; struct pipe_stencil_state { - GLuint front_enabled:1; - GLuint front_func:3; /**< PIPE_FUNC_x */ - GLuint front_fail_op:3; /**< PIPE_STENCIL_OP_x */ - GLuint front_zpass_op:3; /**< PIPE_STENCIL_OP_x */ - GLuint front_zfail_op:3; /**< PIPE_STENCIL_OP_x */ - GLuint back_enabled:1; - GLuint back_func:3; /**< PIPE_FUNC_x */ - GLuint back_fail_op:3; /**< PIPE_STENCIL_OP_x */ - GLuint back_zpass_op:3; /**< PIPE_STENCIL_OP_x */ - GLuint back_zfail_op:3; /**< PIPE_STENCIL_OP_x */ - GLubyte ref_value[2]; /**< [0] = front, [1] = back */ - GLubyte value_mask[2]; - GLubyte write_mask[2]; - GLubyte clear_value; + unsigned front_enabled:1; + unsigned front_func:3; /**< PIPE_FUNC_x */ + unsigned front_fail_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned front_zpass_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned front_zfail_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned back_enabled:1; + unsigned back_func:3; /**< PIPE_FUNC_x */ + unsigned back_fail_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned back_zpass_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned back_zfail_op:3; /**< PIPE_STENCIL_OP_x */ + ubyte ref_value[2]; /**< [0] = front, [1] = back */ + ubyte value_mask[2]; + ubyte write_mask[2]; + ubyte clear_value; }; struct pipe_framebuffer_state { /** multiple colorbuffers for multiple render targets */ - GLuint num_cbufs; + unsigned num_cbufs; struct pipe_surface *cbufs[PIPE_MAX_COLOR_BUFS]; struct pipe_surface *zbuf; /**< Z buffer */ @@ -213,25 +203,25 @@ struct pipe_framebuffer_state */ struct pipe_sampler_state { - GLuint wrap_s:3; /**< PIPE_TEX_WRAP_x */ - GLuint wrap_t:3; /**< PIPE_TEX_WRAP_x */ - GLuint wrap_r:3; /**< PIPE_TEX_WRAP_x */ - GLuint min_img_filter:2; /**< PIPE_TEX_FILTER_x */ - GLuint min_mip_filter:2; /**< PIPE_TEX_MIPFILTER_x */ - GLuint mag_img_filter:2; /**< PIPE_TEX_FILTER_x */ - GLuint compare:1; /**< shadow/depth compare enabled? */ - GLenum compare_mode:1; /**< PIPE_TEX_COMPARE_x */ - GLenum compare_func:3; /**< PIPE_FUNC_x */ - GLfloat shadow_ambient; /**< shadow test fail color/intensity */ - GLfloat min_lod; - GLfloat max_lod; - GLfloat lod_bias; + unsigned wrap_s:3; /**< PIPE_TEX_WRAP_x */ + unsigned wrap_t:3; /**< PIPE_TEX_WRAP_x */ + unsigned wrap_r:3; /**< PIPE_TEX_WRAP_x */ + unsigned min_img_filter:2; /**< PIPE_TEX_FILTER_x */ + unsigned min_mip_filter:2; /**< PIPE_TEX_MIPFILTER_x */ + unsigned mag_img_filter:2; /**< PIPE_TEX_FILTER_x */ + unsigned compare:1; /**< shadow/depth compare enabled? */ + unsigned compare_mode:1; /**< PIPE_TEX_COMPARE_x */ + unsigned compare_func:3; /**< PIPE_FUNC_x */ + float shadow_ambient; /**< shadow test fail color/intensity */ + float min_lod; + float max_lod; + float lod_bias; #if 0 /* need these? */ - GLint BaseLevel; /**< min mipmap level, OpenGL 1.2 */ - GLint MaxLevel; /**< max mipmap level, OpenGL 1.2 */ - GLfloat border_color[4]; + int BaseLevel; /**< min mipmap level, OpenGL 1.2 */ + int MaxLevel; /**< max mipmap level, OpenGL 1.2 */ + float border_color[4]; #endif - GLfloat max_anisotropy; + float max_anisotropy; }; @@ -243,12 +233,12 @@ struct pipe_region { struct pipe_buffer_handle *buffer; /**< driver private buffer handle */ - GLuint refcount; /**< Reference count for region */ - GLuint cpp; /**< bytes per pixel */ - GLuint pitch; /**< in pixels */ - GLuint height; /**< in pixels */ - GLubyte *map; /**< only non-NULL when region is actually mapped */ - GLuint map_refcount; /**< Reference count for mapping */ + unsigned refcount; /**< Reference count for region */ + unsigned cpp; /**< bytes per pixel */ + unsigned pitch; /**< in pixels */ + unsigned height; /**< in pixels */ + ubyte *map; /**< only non-NULL when region is actually mapped */ + unsigned map_refcount; /**< Reference count for mapping */ }; @@ -259,18 +249,18 @@ struct pipe_region struct pipe_surface { struct pipe_region *region; - GLuint format:5; /**< PIPE_FORMAT_x */ - GLuint width, height; - GLuint offset; /**< offset from start of region, in bytes */ - GLint refcount; + unsigned format:5; /**< PIPE_FORMAT_x */ + unsigned width, height; + unsigned offset; /**< offset from start of region, in bytes */ + unsigned refcount; /** get block/tile of pixels from surface */ void (*get_tile)(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p); + unsigned x, unsigned y, unsigned w, unsigned h, float *p); /** put block/tile of pixels into surface */ void (*put_tile)(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, const GLfloat *p); + unsigned x, unsigned y, unsigned w, unsigned h, const float *p); }; @@ -279,11 +269,11 @@ struct pipe_surface */ struct pipe_mipmap_level { - GLuint level_offset; - GLuint width; - GLuint height; - GLuint depth; - GLuint nr_images; + unsigned level_offset; + unsigned width; + unsigned height; + unsigned depth; + unsigned nr_images; /* Explicitly store the offset of each image for each cube face or * depth value. Pretty much have to accept that hardware formats @@ -291,33 +281,34 @@ struct pipe_mipmap_level * compute the offsets of depth/cube images within a mipmap level, * so have to store them as a lookup table: */ - GLuint *image_offset; /**< array [depth] of offsets */ + unsigned *image_offset; /**< array [depth] of offsets */ }; struct pipe_mipmap_tree { /* Effectively the key: */ - GLenum target; /* XXX convert to PIPE_TEXTURE_x */ - GLenum internal_format; /* XXX convert to PIPE_FORMAT_x */ + unsigned target; /* XXX convert to PIPE_TEXTURE_x */ + unsigned internal_format; /* XXX convert to PIPE_FORMAT_x */ + + unsigned format; /**< PIPE_FORMAT_x */ + unsigned first_level; + unsigned last_level; - GLuint format; /**< PIPE_FORMAT_x */ - GLuint first_level; - GLuint last_level; + unsigned width0, height0, depth0; /**< Level zero image dimensions */ + unsigned cpp; - GLuint width0, height0, depth0; /**< Level zero image dimensions */ - GLuint cpp; - GLboolean compressed; + unsigned compressed:1; /* Derived from the above: */ - GLuint pitch; - GLuint depth_pitch; /* per-image on i945? */ - GLuint total_height; + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; /* Includes image offset tables: */ - struct pipe_mipmap_level level[MAX_TEXTURE_LEVELS]; + struct pipe_mipmap_level level[PIPE_MAX_TEXTURE_LEVELS]; /* The data is held here: */ @@ -325,7 +316,7 @@ struct pipe_mipmap_tree /* These are also refcounted: */ - GLuint refcount; + unsigned refcount; }; diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 2d69b6e0dc..14ddf0c306 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -44,7 +44,7 @@ */ void softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, - GLuint clearValue) + unsigned clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); GLint x, y, w, h; diff --git a/src/mesa/pipe/softpipe/sp_clear.h b/src/mesa/pipe/softpipe/sp_clear.h index e706e731c2..a8ed1c4ecc 100644 --- a/src/mesa/pipe/softpipe/sp_clear.h +++ b/src/mesa/pipe/softpipe/sp_clear.h @@ -37,7 +37,7 @@ struct pipe_context; extern void softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, - GLuint clearValue); + unsigned clearValue); #endif /* SP_CLEAR_H */ diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index b68d0f95ac..70a80aff88 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -32,13 +32,15 @@ * all the enabled attributes run contiguously. */ -#include "glheader.h" -#include "imports.h" +#include "main/mtypes.h" + +#include "pipe/p_util.h" +#include "tgsi/core/tgsi_core.h" + #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" #include "sp_tex_sample.h" -#include "tgsi/core/tgsi_core.h" #if 0 #if defined __GNUC__ @@ -71,9 +73,9 @@ struct exec_machine { const struct setup_coefficient *coef; /**< will point to quad->coef */ #if ALIGNED_ATTRIBS - GLfloat attr[FRAG_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE] __attribute__(( aligned( 16 ) )); + GLfloat attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE] __attribute__(( aligned( 16 ) )); #else - GLfloat attr[FRAG_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE]; + GLfloat attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE]; #endif }; diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c index bd5f5e930d..fda1b46ab4 100644 --- a/src/mesa/pipe/softpipe/sp_tex_layout.c +++ b/src/mesa/pipe/softpipe/sp_tex_layout.c @@ -33,6 +33,7 @@ #include "macros.h" #include "pipe/p_state.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" #include "sp_tex_layout.h" @@ -250,18 +251,18 @@ softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * switch (d) { case 4: switch (face) { - case FACE_POS_X: - case FACE_NEG_X: + case PIPE_TEX_FACE_POS_X: + case PIPE_TEX_FACE_NEG_X: x += step_offsets[face][0] * d; y += step_offsets[face][1] * d; break; - case FACE_POS_Y: - case FACE_NEG_Y: + case PIPE_TEX_FACE_POS_Y: + case PIPE_TEX_FACE_NEG_Y: y += 12; x -= 8; break; - case FACE_POS_Z: - case FACE_NEG_Z: + case PIPE_TEX_FACE_POS_Z: + case PIPE_TEX_FACE_NEG_Z: y = mt->total_height - 4; x = (face - 4) * 8; break; -- cgit v1.2.3 From 76818b67092579585d9fcb7df1a8c133e6c94a6c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 13 Aug 2007 16:24:24 +0100 Subject: Missing files --- src/mesa/pipe/p_compiler.h | 63 ++++++++++++++++++++++++ src/mesa/pipe/p_util.h | 117 +++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 180 insertions(+) create mode 100644 src/mesa/pipe/p_compiler.h create mode 100644 src/mesa/pipe/p_util.h (limited to 'src') diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h new file mode 100644 index 0000000000..51a78bf4f5 --- /dev/null +++ b/src/mesa/pipe/p_compiler.h @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_COMPILER_H +#define P_COMPILER_H + +#include +#include +#include +#include + +typedef unsigned int uint; +typedef unsigned char ubyte; +typedef unsigned char boolean; +typedef unsigned short ushort; + + +#define TRUE 1 +#define FALSE 0 + +/* Function inlining */ +#if defined(__GNUC__) +# define INLINE __inline__ +#elif defined(__MSC__) +# define INLINE __inline +#elif defined(_MSC_VER) +# define INLINE __inline +#elif defined(__ICL) +# define INLINE __inline +#elif defined(__INTEL_COMPILER) +# define INLINE inline +#elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100) +# define INLINE __inline +#else +# define INLINE +#endif + + +#endif /* P_COMPILER_H */ diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h new file mode 100644 index 0000000000..979befb194 --- /dev/null +++ b/src/mesa/pipe/p_util.h @@ -0,0 +1,117 @@ + +/* + * Mesa 3-D graphics library + * Version: 6.5.2 + * + * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef P_UTIL_H +#define P_UTIL_H + +#include "p_compiler.h" + +#define CALLOC_STRUCT(T) (struct T *) calloc(1, sizeof(struct T)) + + + +/** Clamp X to [MIN,MAX] */ +#define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) + +/** Assign X to CLAMP(X, MIN, MAX) */ +#define CLAMP_SELF(x, mn, mx) \ + ( (x)<(mn) ? ((x) = (mn)) : ((x)>(mx) ? ((x)=(mx)) : (x)) ) + +/** Minimum of two values: */ +#define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) + +/** Maximum of two values: */ +#define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) + + +#define Elements(x) sizeof(x)/sizeof(*(x)) + +union fi { + float f; + int i; + unsigned ui; +}; + +#define UBYTE_TO_FLOAT( ub ) ((float)(ub) / 255.0F) + +#define IEEE_0996 0x3f7f0000 /* 0.996 or so */ + +/* This function/macro is sensitive to precision. Test very carefully + * if you change it! + */ +#define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \ + do { \ + union fi __tmp; \ + __tmp.f = (F); \ + if (__tmp.i < 0) \ + UB = (ubyte) 0; \ + else if (__tmp.i >= IEEE_0996) \ + UB = (ubyte) 255; \ + else { \ + __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F; \ + UB = (ubyte) __tmp.i; \ + } \ + } while (0) + + + +static INLINE unsigned pack_ub4( unsigned char b0, + unsigned char b1, + unsigned char b2, + unsigned char b3 ) +{ + return ((((unsigned int)b0) << 0) | + (((unsigned int)b1) << 8) | + (((unsigned int)b2) << 16) | + (((unsigned int)b3) << 24)); +} + +static INLINE unsigned fui( float f ) +{ + union fi fi; + fi.f = f; + return fi.ui; +} + +static INLINE unsigned char float_to_ubyte( float f ) +{ + unsigned char ub; + UNCLAMPED_FLOAT_TO_UBYTE(ub, f); + return ub; +} + +static INLINE unsigned pack_ui32_float4( float a, + float b, + float d, + float c ) +{ + return pack_ub4( float_to_ubyte(a), + float_to_ubyte(b), + float_to_ubyte(c), + float_to_ubyte(d) ); +} + +#endif -- cgit v1.2.3 From 8571c4babf82b2fe78dba06458070eb4010cfc18 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 16:16:42 -0600 Subject: added an assertion --- src/mesa/state_tracker/st_cb_clear.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 2583b4f1bd..80f136e0a6 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -454,8 +454,10 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) GLuint b; for (b = 0; b < BUFFER_COUNT; b++) { if (BUFFER_BITS_COLOR & mask & (1 << b)) { - clear_color_buffer(ctx, - ctx->DrawBuffer->Attachment[b].Renderbuffer); + struct gl_renderbuffer *rb + = ctx->DrawBuffer->Attachment[b].Renderbuffer; + assert(rb); + clear_color_buffer(ctx, rb); } } } -- cgit v1.2.3 From 5e1bfe426f3e2bc7a13b3814f1fa732141f15a9a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 16:18:03 -0600 Subject: fix up some render-to-texture code ... almost there --- src/mesa/state_tracker/st_cb_fbo.c | 47 ++++++++++++++++++++++++++++++-------- 1 file changed, 38 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 3048765114..c0b88404b6 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -285,27 +285,48 @@ st_render_texture(GLcontext *ctx, struct gl_renderbuffer_attachment *att) { struct st_context *st = ctx->st; - struct st_renderbuffer *strb = st_renderbuffer(att->Renderbuffer); + struct st_renderbuffer *strb; + struct gl_renderbuffer *rb; struct pipe_context *pipe = st->pipe; struct pipe_framebuffer_state framebuffer; struct pipe_mipmap_tree *mt; + assert(!att->Renderbuffer); + + /* create new renderbuffer which wraps the texture image */ + rb = st_new_renderbuffer(ctx, 0); + if (!rb) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture()"); + return; + } + + _mesa_reference_renderbuffer(&att->Renderbuffer, rb); + assert(rb->RefCount == 1); + rb->AllocStorage = NULL; /* should not get called */ + strb = st_renderbuffer(rb); + + /* get the mipmap tree for the texture */ mt = st_get_texobj_mipmap_tree(att->Texture); + assert(mt); + assert(mt->level[0].width); /* the renderbuffer's surface is inside the mipmap_tree: */ strb->surface = pipe->get_tex_surface(pipe, mt, att->CubeMapFace, att->TextureLevel, att->Zoffset); + assert(strb->surface); - /* update pipe's framebuffer state */ - memset(&framebuffer, 0, sizeof(framebuffer)); - framebuffer.num_cbufs = 1; - framebuffer.cbufs[0] = strb->surface; - if (memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0) { - st->state.framebuffer = framebuffer; - st->pipe->set_framebuffer_state( st->pipe, &framebuffer ); - } + /* + printf("RENDER TO TEXTURE mt=%p surf=%p\n", mt, strb->surface); + */ + + /* Invalidate buffer state so that the pipe's framebuffer state + * gets updated. + * That's where the new renderbuffer (which we just created) gets + * passed to the pipe as a (color/depth) render target. + */ + st_invalidate_state(ctx, _NEW_BUFFERS); } @@ -318,8 +339,16 @@ st_finish_render_texture(GLcontext *ctx, { struct st_renderbuffer *strb = st_renderbuffer(att->Renderbuffer); + assert(strb); + + /* + printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface); + */ + pipe_surface_unreference(&strb->surface); + _mesa_reference_renderbuffer(&att->Renderbuffer, NULL); + /* restore previous framebuffer state */ st_invalidate_state(ctx, _NEW_BUFFERS); } -- cgit v1.2.3 From 612cfb749c3526eeb446bbc631bf24716522f373 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 16:18:32 -0600 Subject: added assertion to check that texture width is non-zero --- src/mesa/pipe/softpipe/sp_tex_sample.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 40350eec75..fb4f7930ac 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -489,6 +489,8 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, level0 = choose_mipmap_level(sampler, lambda); + assert(sampler->texture->level[level0].width); + switch (filter) { case PIPE_TEX_FILTER_NEAREST: { -- cgit v1.2.3 From cd308dfa794a1be5d933a4b89539d53929537f88 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 18:14:54 -0600 Subject: map stencil buffer too --- src/mesa/pipe/softpipe/sp_context.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index d67ad2c01d..34ef00f619 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -126,6 +126,12 @@ static void map_surfaces(struct softpipe_context *sp) pipe->region_map(pipe, sps->surface.region); } + if (sp->framebuffer.sbuf) { + struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.sbuf); + if (sps->surface.region) + pipe->region_map(pipe, sps->surface.region); + } + /* textures */ for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { struct pipe_mipmap_tree *mt = sp->texture[i]; @@ -155,6 +161,12 @@ static void unmap_surfaces(struct softpipe_context *sp) pipe->region_unmap(pipe, sps->surface.region); } + if (sp->framebuffer.sbuf) { + struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.sbuf); + if (sps->surface.region) + pipe->region_unmap(pipe, sps->surface.region); + } + /* textures */ for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { struct pipe_mipmap_tree *mt = sp->texture[i]; -- cgit v1.2.3 From 4698483f849ba8dfde20a5d649f1ea099291f241 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 18:16:11 -0600 Subject: Added st_fb_orientation() function to determine the up/down orientation of the framebuffer. --- src/mesa/state_tracker/st_atom_viewport.c | 14 +++----------- src/mesa/state_tracker/st_cb_readpixels.c | 16 ++++++++++++---- src/mesa/state_tracker/st_context.h | 27 +++++++++++++++++++++++++++ 3 files changed, 42 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index 1307cbb6d2..a70f4c7434 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -39,27 +39,19 @@ * - depthrange * - window pos/size or FBO size */ -static void update_viewport( struct st_context *st ) +static void +update_viewport( struct st_context *st ) { GLcontext *ctx = st->ctx; GLfloat yScale, yBias; - /* Negate Y scale to flip image vertically. - * The NDC Y coords prior to viewport transformation are in the range - * [y=-1=bottom, y=1=top] - * Hardware window coords are in the range [y=0=top, y=H-1=bottom] where H - * is the window height. - * Use the viewport transformation to invert Y. - */ - /* _NEW_BUFFERS */ - if (ctx->DrawBuffer) { + if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { yScale = -1; yBias = ctx->DrawBuffer->Height ; } else { - /* we won't be rendering anything */ yScale = 1.0; yBias = 0.0; } diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 22abc104e2..98604e5b6b 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -60,7 +60,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct pipe_context *pipe = ctx->st->pipe; GLfloat temp[MAX_WIDTH][4]; const GLbitfield transferOps = ctx->_ImageTransferState; - GLint i, yInv, dfStride; + GLint i, yStep, dfStride; GLfloat *df; struct st_renderbuffer *strb; struct gl_pixelstore_attrib clippedPacking = *pack; @@ -104,11 +104,19 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, dfStride = 0; } + /* determine bottom-to-top vs. top-to-bottom order */ + if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { + y = strb->Base.Height - 1 - y; + yStep = -1; + } + else { + yStep = 1; + } + /* Do a row at a time to flip image data vertically */ - yInv = strb->Base.Height - 1 - y; for (i = 0; i < height; i++) { - strb->surface->get_tile(strb->surface, x, yInv, width, 1, df); - yInv--; + strb->surface->get_tile(strb->surface, x, y, width, 1, df); + y += yStep; df += dfStride; if (!dfStride) { /* convert GLfloat to user's format/type */ diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 8d82d53133..13843d9b7f 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -118,5 +118,32 @@ extern void st_init_driver_functions(struct dd_function_table *functions); +#define Y_0_TOP 1 +#define Y_0_BOTTOM 2 + +static INLINE GLuint +st_fb_orientation(const struct gl_framebuffer *fb) +{ + if (fb && fb->Name == 0) { + /* Drawing into a window (on-screen buffer). + * + * Negate Y scale to flip image vertically. + * The NDC Y coords prior to viewport transformation are in the range + * [y=-1=bottom, y=1=top] + * Hardware window coords are in the range [y=0=top, y=H-1=bottom] where + * H is the window height. + * Use the viewport transformation to invert Y. + */ + return Y_0_TOP; + } + else { + /* Drawing into user-created FBO (very likely a texture). + * + * For textures, T=0=Bottom, so by extension Y=0=Bottom for rendering. + */ + return Y_0_BOTTOM; + } +} + #endif -- cgit v1.2.3 From 2f605fd457ccd8763ce5b0acc8d2906a59ea22bc Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 13 Aug 2007 18:16:47 -0600 Subject: set renderbuffer Width/Height = texture size --- src/mesa/state_tracker/st_cb_fbo.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index c0b88404b6..7336327330 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -288,7 +288,6 @@ st_render_texture(GLcontext *ctx, struct st_renderbuffer *strb; struct gl_renderbuffer *rb; struct pipe_context *pipe = st->pipe; - struct pipe_framebuffer_state framebuffer; struct pipe_mipmap_tree *mt; assert(!att->Renderbuffer); @@ -310,6 +309,9 @@ st_render_texture(GLcontext *ctx, assert(mt); assert(mt->level[0].width); + rb->Width = mt->level[0].width; + rb->Height = mt->level[0].height; + /* the renderbuffer's surface is inside the mipmap_tree: */ strb->surface = pipe->get_tex_surface(pipe, mt, att->CubeMapFace, @@ -317,9 +319,8 @@ st_render_texture(GLcontext *ctx, att->Zoffset); assert(strb->surface); - /* - printf("RENDER TO TEXTURE mt=%p surf=%p\n", mt, strb->surface); - */ + printf("RENDER TO TEXTURE obj=%p mt=%p surf=%p %d x %d\n", + att->Texture, mt, strb->surface, rb->Width, rb->Height); /* Invalidate buffer state so that the pipe's framebuffer state * gets updated. @@ -341,9 +342,9 @@ st_finish_render_texture(GLcontext *ctx, assert(strb); - /* + ctx->st->pipe->flush(ctx->st->pipe, 0x0); + printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface); - */ pipe_surface_unreference(&strb->surface); -- cgit v1.2.3 From 70af238b494ed1b6da4841c2065c33ee0f0f37c9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 13 Aug 2007 17:02:27 +0100 Subject: Continue reducing dependencies on core mesa include files. Mainly down to the support for legacy TNL processing now. --- src/mesa/pipe/p_util.h | 83 +++++++-- src/mesa/pipe/softpipe/sp_clear.c | 4 +- src/mesa/pipe/softpipe/sp_context.c | 27 ++- src/mesa/pipe/softpipe/sp_context.h | 24 +-- src/mesa/pipe/softpipe/sp_headers.h | 24 +-- src/mesa/pipe/softpipe/sp_prim_setup.c | 219 +++++++++++----------- src/mesa/pipe/softpipe/sp_prim_setup.h | 78 +------- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 7 +- src/mesa/pipe/softpipe/sp_quad_blend.c | 36 ++-- src/mesa/pipe/softpipe/sp_quad_bufloop.c | 8 +- src/mesa/pipe/softpipe/sp_quad_colormask.c | 14 +- src/mesa/pipe/softpipe/sp_quad_coverage.c | 5 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 15 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 48 ++--- src/mesa/pipe/softpipe/sp_quad_occlusion.c | 3 +- src/mesa/pipe/softpipe/sp_quad_output.c | 14 +- src/mesa/pipe/softpipe/sp_quad_stencil.c | 35 ++-- src/mesa/pipe/softpipe/sp_quad_stipple.c | 11 +- src/mesa/pipe/softpipe/sp_region.c | 64 +++---- src/mesa/pipe/softpipe/sp_state.h | 5 +- src/mesa/pipe/softpipe/sp_state_blend.c | 2 - src/mesa/pipe/softpipe/sp_state_clip.c | 2 - src/mesa/pipe/softpipe/sp_state_derived.c | 22 +-- src/mesa/pipe/softpipe/sp_state_sampler.c | 6 +- src/mesa/pipe/softpipe/sp_state_surface.c | 2 - src/mesa/pipe/softpipe/sp_surface.c | 127 +++++++------ src/mesa/pipe/softpipe/sp_surface.h | 35 ++-- src/mesa/pipe/softpipe/sp_tex_layout.c | 87 ++++----- src/mesa/pipe/softpipe/sp_tex_layout.h | 2 +- src/mesa/pipe/softpipe/sp_tex_sample.c | 72 ++++---- src/mesa/pipe/softpipe/sp_tex_sample.h | 8 +- src/mesa/pipe/tgsi/core/tgsi_build.h | 146 +++++++-------- src/mesa/pipe/tgsi/core/tgsi_dump.h | 2 +- src/mesa/pipe/tgsi/core/tgsi_exec.h | 42 +++-- src/mesa/pipe/tgsi/core/tgsi_parse.h | 6 +- src/mesa/pipe/tgsi/core/tgsi_token.h | 272 ++++++++++++++-------------- src/mesa/pipe/tgsi/core/tgsi_util.h | 34 ++-- 37 files changed, 777 insertions(+), 814 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 979befb194..50d57a2ed0 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -28,25 +28,14 @@ #define P_UTIL_H #include "p_compiler.h" +#include #define CALLOC_STRUCT(T) (struct T *) calloc(1, sizeof(struct T)) - - -/** Clamp X to [MIN,MAX] */ #define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) - -/** Assign X to CLAMP(X, MIN, MAX) */ -#define CLAMP_SELF(x, mn, mx) \ - ( (x)<(mn) ? ((x) = (mn)) : ((x)>(mx) ? ((x)=(mx)) : (x)) ) - -/** Minimum of two values: */ #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) - -/** Maximum of two values: */ #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) - #define Elements(x) sizeof(x)/sizeof(*(x)) union fi { @@ -114,4 +103,74 @@ static INLINE unsigned pack_ui32_float4( float a, float_to_ubyte(d) ); } +#define COPY_4V( DST, SRC ) \ +do { \ + (DST)[0] = (SRC)[0]; \ + (DST)[1] = (SRC)[1]; \ + (DST)[2] = (SRC)[2]; \ + (DST)[3] = (SRC)[3]; \ +} while (0) + + +static INLINE int ifloor(float f) +{ + int ai, bi; + double af, bf; + union fi u; + + af = (3 << 22) + 0.5 + (double)f; + bf = (3 << 22) + 0.5 - (double)f; + u.f = (float) af; ai = u.i; + u.f = (float) bf; bi = u.i; + return (ai - bi) >> 1; +} + + +#if defined(__GNUC__) && defined(__i386__) +static INLINE int iround(float f) +{ + int r; + __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st"); + return r; +} +#elif defined(__MSC__) && defined(__WIN32__) +static INLINE int iround(float f) +{ + int r; + _asm { + fld f + fistp r + } + return r; +} +#else +#define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F))) +#endif + + +/* Could maybe have an inline version of this? + */ +#if defined(__GNUC__) +#define FABSF(x) fabsf(x) +#else +#define FABSF(x) ((GLfloat) fabs(x)) +#endif + +/* Pretty fast, and accurate. + * Based on code from http://www.flipcode.com/totd/ + */ +static INLINE float LOG2(float val) +{ + union fi num; + int log_2; + + num.f = val; + log_2 = ((num.i >> 23) & 255) - 128; + num.i &= ~(255 << 23); + num.i += 127 << 23; + num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3; + return num.f + log_2; +} + + #endif diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 14ddf0c306..2774413720 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -35,7 +35,7 @@ #include "sp_context.h" #include "sp_surface.h" #include "sp_state.h" -#include "colormac.h" +//#include "colormac.h" /** @@ -47,7 +47,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); - GLint x, y, w, h; + int x, y, w, h; softpipe_update_derived(softpipe); /* not needed?? */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 34ef00f619..1ea6bbbf7f 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -29,10 +29,9 @@ * Keith Whitwell */ -#include "main/imports.h" -#include "main/macros.h" #include "pipe/draw/draw_context.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_clear.h" #include "sp_context.h" #include "sp_flush.h" @@ -50,11 +49,11 @@ * If we find texture and drawable support differs, add a selector * parameter or another function. */ -static const GLuint * -softpipe_supported_formats(struct pipe_context *pipe, GLuint *numFormats) +static const unsigned * +softpipe_supported_formats(struct pipe_context *pipe, unsigned *numFormats) { #if 0 - static const GLuint supported[] = { + static const unsigned supported[] = { PIPE_FORMAT_U_R8_G8_B8_A8, PIPE_FORMAT_U_A8_R8_G8_B8, PIPE_FORMAT_U_R5_G6_B5, @@ -82,9 +81,9 @@ softpipe_supported_formats(struct pipe_context *pipe, GLuint *numFormats) static void -softpipe_max_texture_size(struct pipe_context *pipe, GLuint textureType, - GLuint *maxWidth, GLuint *maxHeight, - GLuint *maxDepth) +softpipe_max_texture_size(struct pipe_context *pipe, unsigned textureType, + unsigned *maxWidth, unsigned *maxHeight, + unsigned *maxDepth) { switch (textureType) { case PIPE_TEXTURE_1D: @@ -112,7 +111,7 @@ softpipe_max_texture_size(struct pipe_context *pipe, GLuint textureType, static void map_surfaces(struct softpipe_context *sp) { struct pipe_context *pipe = &sp->pipe; - GLuint i; + unsigned i; for (i = 0; i < sp->framebuffer.num_cbufs; i++) { struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); @@ -147,7 +146,7 @@ static void map_surfaces(struct softpipe_context *sp) static void unmap_surfaces(struct softpipe_context *sp) { struct pipe_context *pipe = &sp->pipe; - GLuint i; + unsigned i; for (i = 0; i < sp->framebuffer.num_cbufs; i++) { struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); @@ -206,9 +205,9 @@ static void softpipe_draw_vb( struct pipe_context *pipe, static void softpipe_draw_vertices(struct pipe_context *pipe, - GLuint mode, - GLuint numVertex, const GLfloat *verts, - GLuint numAttribs, const GLuint attribs[]) + unsigned mode, + unsigned numVertex, const float *verts, + unsigned numAttribs, const unsigned attribs[]) { struct softpipe_context *softpipe = softpipe_context( pipe ); @@ -230,7 +229,7 @@ static void softpipe_reset_occlusion_counter(struct pipe_context *pipe) } /* XXX pipe param should be const */ -static GLuint softpipe_get_occlusion_counter(struct pipe_context *pipe) +static unsigned softpipe_get_occlusion_counter(struct pipe_context *pipe) { struct softpipe_context *softpipe = softpipe_context( pipe ); return softpipe->occlusion_counter; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index a99c083caa..8887c576c1 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -31,7 +31,7 @@ #ifndef SP_CONTEXT_H #define SP_CONTEXT_H -#include "glheader.h" +//#include "glheader.h" #include "pipe/p_state.h" #include "pipe/p_context.h" @@ -88,7 +88,7 @@ struct softpipe_context { struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; - GLuint dirty; + unsigned dirty; /* Setup derived state. TODO: this should be passed in the program * tokens as parameters to DECL instructions. @@ -103,28 +103,28 @@ struct softpipe_context { */ /** Map fragment program attribute to quad/coef array slot */ - GLuint fp_attr_to_slot[PIPE_ATTRIB_MAX]; + unsigned fp_attr_to_slot[PIPE_ATTRIB_MAX]; /** Map vertex format attribute to a vertex attribute slot */ - GLuint vf_attr_to_slot[PIPE_ATTRIB_MAX]; - GLuint nr_attrs; - GLuint nr_frag_attrs; /**< number of active fragment attribs */ - GLbitfield attr_mask; /**< bitfield of VF_ATTRIB_ indexes/bits */ + unsigned vf_attr_to_slot[PIPE_ATTRIB_MAX]; + unsigned nr_attrs; + unsigned nr_frag_attrs; /**< number of active fragment attribs */ + unsigned attr_mask; /**< bitfield of VF_ATTRIB_ indexes/bits */ - GLboolean need_z; /**< produce quad/fragment Z values? */ - GLboolean need_w; /**< produce quad/fragment W values? */ + boolean need_z; /**< produce quad/fragment Z values? */ + boolean need_w; /**< produce quad/fragment W values? */ #if 0 /* Stipple derived state: */ - GLubyte stipple_masks[16][16]; + ubyte stipple_masks[16][16]; #endif /** Derived from scissor and surface bounds: */ struct pipe_scissor_state cliprect; - GLuint occlusion_counter; + unsigned occlusion_counter; - GLuint line_stipple_counter; + unsigned line_stipple_counter; /** Software quad rendering pipeline */ struct { diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index 68a84621f2..b7f46cb658 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -57,9 +57,9 @@ struct setup_coefficient { - GLfloat a0[NUM_CHANNELS]; /* in an xyzw layout */ - GLfloat dadx[NUM_CHANNELS]; - GLfloat dady[NUM_CHANNELS]; + float a0[NUM_CHANNELS]; /* in an xyzw layout */ + float dadx[NUM_CHANNELS]; + float dady[NUM_CHANNELS]; }; @@ -69,18 +69,18 @@ struct setup_coefficient { * "Channel-Serial" or "SoA" layout. */ struct quad_header { - GLint x0; - GLint y0; - GLuint mask:4; - GLuint facing:1; /**< Front (0) or back (1) facing? */ - GLuint prim:2; /**< PRIM_POINT, LINE, TRI */ + int x0; + int y0; + unsigned mask:4; + unsigned facing:1; /**< Front (0) or back (1) facing? */ + unsigned prim:2; /**< PRIM_POINT, LINE, TRI */ struct { - GLfloat color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ - GLfloat depth[QUAD_SIZE]; + float color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ + float depth[QUAD_SIZE]; } outputs; - GLfloat coverage[QUAD_SIZE]; /** fragment coverage for antialiasing */ + float coverage[QUAD_SIZE]; /** fragment coverage for antialiasing */ const struct setup_coefficient *coef; @@ -88,7 +88,7 @@ struct quad_header { * encoded in fragment program DECL * statements. */ - GLuint nr_attrs; + unsigned nr_attrs; }; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 7a3d011b7b..45d09860c3 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -33,25 +33,26 @@ */ -#include "imports.h" -#include "macros.h" +//#include "imports.h" +//#include "macros.h" #include "sp_context.h" #include "sp_headers.h" -#include "pipe/draw/draw_private.h" #include "sp_quad.h" #include "sp_prim_setup.h" +#include "pipe/draw/draw_private.h" +#include "pipe/p_util.h" /** * Triangle edge info */ struct edge { - GLfloat dx; /**< X(v1) - X(v0), used only during setup */ - GLfloat dy; /**< Y(v1) - Y(v0), used only during setup */ - GLfloat dxdy; /**< dx/dy */ - GLfloat sx, sy; /**< first sample point coord */ - GLint lines; /**< number of lines on this edge */ + float dx; /**< X(v1) - X(v0), used only during setup */ + float dy; /**< Y(v1) - Y(v0), used only during setup */ + float dxdy; /**< dx/dy */ + float sx, sy; /**< first sample point coord */ + int lines; /**< number of lines on this edge */ }; @@ -77,17 +78,17 @@ struct setup_stage { struct edge etop; struct edge emaj; - GLfloat oneoverarea; + float oneoverarea; struct setup_coefficient coef[FRAG_ATTRIB_MAX]; struct quad_header quad; struct { - GLint left[2]; /**< [0] = row0, [1] = row1 */ - GLint right[2]; - GLint y; - GLuint y_flags; - GLuint mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ + int left[2]; /**< [0] = row0, [1] = row1 */ + int right[2]; + int y; + unsigned y_flags; + unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ } span; }; @@ -146,7 +147,7 @@ clip_emit_quad(struct setup_stage *setup) * Emit a quad (pass to next stage). No clipping is done. */ static INLINE void -emit_quad( struct setup_stage *setup, GLint x, GLint y, GLuint mask ) +emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) { struct softpipe_context *sp = setup->softpipe; setup->quad.x0 = x; @@ -160,7 +161,7 @@ emit_quad( struct setup_stage *setup, GLint x, GLint y, GLuint mask ) * Given an X or Y coordinate, return the block/quad coordinate that it * belongs to. */ -static INLINE GLint block( GLint x ) +static INLINE int block( int x ) { return x & ~1; } @@ -173,10 +174,10 @@ static INLINE GLint block( GLint x ) * this is pretty nasty... may need to rework flush_spans again to * fix it, if possible. */ -static GLuint calculate_mask( struct setup_stage *setup, - GLint x ) +static unsigned calculate_mask( struct setup_stage *setup, + int x ) { - GLuint mask = 0; + unsigned mask = 0; if (x >= setup->span.left[0] && x < setup->span.right[0]) mask |= MASK_BOTTOM_LEFT; @@ -199,8 +200,8 @@ static GLuint calculate_mask( struct setup_stage *setup, */ static void flush_spans( struct setup_stage *setup ) { - GLint minleft, maxright; - GLint x; + int minleft, maxright; + int x; switch (setup->span.y_flags) { case 3: @@ -249,7 +250,7 @@ static void print_vertex(const struct setup_stage *setup, } #endif -static GLboolean setup_sort_vertices( struct setup_stage *setup, +static boolean setup_sort_vertices( struct setup_stage *setup, const struct prim_header *prim ) { const struct vertex_header *v0 = prim->v[0]; @@ -267,9 +268,9 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, /* determine bottom to top order of vertices */ { - GLfloat y0 = v0->data[0][1]; - GLfloat y1 = v1->data[0][1]; - GLfloat y2 = v2->data[0][1]; + float y0 = v0->data[0][1]; + float y1 = v1->data[0][1]; + float y2 = v2->data[0][1]; if (y0 <= y1) { if (y1 <= y2) { /* y0<=y1<=y2 */ @@ -330,7 +331,7 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, * use the prim->det value because its sign is correct. */ { - const GLfloat area = (setup->emaj.dx * setup->ebot.dy - + const float area = (setup->emaj.dx * setup->ebot.dy - setup->ebot.dx * setup->emaj.dy); setup->oneoverarea = 1.0 / area; @@ -346,7 +347,7 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, */ setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->setup.front_winding == PIPE_WINDING_CW); - return GL_TRUE; + return TRUE; } @@ -358,8 +359,8 @@ static GLboolean setup_sort_vertices( struct setup_stage *setup, * \param i which component of the slot (0..3) */ static void const_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i ) + unsigned slot, + unsigned i ) { assert(slot < FRAG_ATTRIB_MAX); assert(i <= 3); @@ -378,13 +379,13 @@ static void const_coeff( struct setup_stage *setup, * for a triangle. */ static void tri_linear_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i) + unsigned slot, + unsigned i) { - GLfloat botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; - GLfloat majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; - GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; + float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; assert(slot < FRAG_ATTRIB_MAX); assert(i <= 3); @@ -423,19 +424,19 @@ static void tri_linear_coeff( struct setup_stage *setup, * for a triangle. */ static void tri_persp_coeff( struct setup_stage *setup, - GLuint slot, - GLuint i ) + unsigned slot, + unsigned i ) { /* premultiply by 1/w: */ - GLfloat mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; - GLfloat mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; - GLfloat maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; - - GLfloat botda = mida - mina; - GLfloat majda = maxa - mina; - GLfloat a = setup->ebot.dy * majda - botda * setup->emaj.dy; - GLfloat b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; + float mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; + float maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; + + float botda = mida - mina; + float majda = maxa - mina; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; assert(slot < FRAG_ATTRIB_MAX); assert(i <= 3); @@ -455,7 +456,7 @@ static void tri_persp_coeff( struct setup_stage *setup, static void setup_tri_coefficients( struct setup_stage *setup ) { const enum interp_mode *interp = setup->softpipe->interp; - GLuint slot, j; + unsigned slot, j; /* z and w are done by linear interpolation: */ @@ -488,25 +489,25 @@ static void setup_tri_coefficients( struct setup_stage *setup ) static void setup_tri_edges( struct setup_stage *setup ) { - GLfloat vmin_x = setup->vmin->data[0][0] + 0.5; - GLfloat vmid_x = setup->vmid->data[0][0] + 0.5; + float vmin_x = setup->vmin->data[0][0] + 0.5; + float vmid_x = setup->vmid->data[0][0] + 0.5; - GLfloat vmin_y = setup->vmin->data[0][1] - 0.5; - GLfloat vmid_y = setup->vmid->data[0][1] - 0.5; - GLfloat vmax_y = setup->vmax->data[0][1] - 0.5; + float vmin_y = setup->vmin->data[0][1] - 0.5; + float vmid_y = setup->vmid->data[0][1] - 0.5; + float vmax_y = setup->vmax->data[0][1] - 0.5; setup->emaj.sy = ceilf(vmin_y); - setup->emaj.lines = (GLint) ceilf(vmax_y - setup->emaj.sy); + setup->emaj.lines = (int) ceilf(vmax_y - setup->emaj.sy); setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; setup->etop.sy = ceilf(vmid_y); - setup->etop.lines = (GLint) ceilf(vmax_y - setup->etop.sy); + setup->etop.lines = (int) ceilf(vmax_y - setup->etop.sy); setup->etop.dxdy = setup->etop.dx / setup->etop.dy; setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; setup->ebot.sy = ceilf(vmin_y); - setup->ebot.lines = (GLint) ceilf(vmid_y - setup->ebot.sy); + setup->ebot.lines = (int) ceilf(vmid_y - setup->ebot.sy); setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; } @@ -519,13 +520,13 @@ static void setup_tri_edges( struct setup_stage *setup ) static void subtriangle( struct setup_stage *setup, struct edge *eleft, struct edge *eright, - GLuint lines ) + unsigned lines ) { const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - GLint y, start_y, finish_y; - GLint sy = (GLint)eleft->sy; + int y, start_y, finish_y; + int sy = (int)eleft->sy; - assert((GLint)eleft->sy == (GLint) eright->sy); + assert((int)eleft->sy == (int) eright->sy); /* clip top/bottom */ start_y = sy; @@ -552,8 +553,8 @@ static void subtriangle( struct setup_stage *setup, * * this is all drowned out by the attribute interpolation anyway. */ - GLint left = (GLint)(eleft->sx + y * eleft->dxdy); - GLint right = (GLint)(eright->sx + y * eright->dxdy); + int left = (int)(eleft->sx + y * eleft->dxdy); + int right = (int)(eright->sx + y * eright->dxdy); /* clip left/right */ if (left < cliprect->minx) @@ -562,7 +563,7 @@ static void subtriangle( struct setup_stage *setup, right = cliprect->maxx; if (left < right) { - GLint _y = sy+y; + int _y = sy+y; if (block(_y) != setup->span.y) { flush_spans(setup); setup->span.y = block(_y); @@ -633,11 +634,11 @@ static void setup_tri( struct draw_stage *stage, * for a line. */ static void -line_linear_coeff(struct setup_stage *setup, GLuint slot, GLuint i) +line_linear_coeff(struct setup_stage *setup, unsigned slot, unsigned i) { - const GLfloat dz = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - const GLfloat dadx = dz * setup->emaj.dx * setup->oneoverarea; - const GLfloat dady = dz * setup->emaj.dy * setup->oneoverarea; + const float dz = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + const float dadx = dz * setup->emaj.dx * setup->oneoverarea; + const float dady = dz * setup->emaj.dy * setup->oneoverarea; setup->coef[slot].dadx[i] = dadx; setup->coef[slot].dady[i] = dady; setup->coef[slot].a0[i] @@ -652,7 +653,7 @@ line_linear_coeff(struct setup_stage *setup, GLuint slot, GLuint i) * for a line. */ static void -line_persp_coeff(struct setup_stage *setup, GLuint slot, GLuint i) +line_persp_coeff(struct setup_stage *setup, unsigned slot, unsigned i) { /* XXX to do */ line_linear_coeff(setup, slot, i); /* XXX temporary */ @@ -667,7 +668,7 @@ static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { const enum interp_mode *interp = setup->softpipe->interp; - GLuint slot, j; + unsigned slot, j; /* use setup->vmin, vmax to point to vertices */ setup->vprovoke = prim->v[1]; @@ -712,13 +713,13 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) * Plot a pixel in a line segment. */ static INLINE void -plot(struct setup_stage *setup, GLint x, GLint y) +plot(struct setup_stage *setup, int x, int y) { - const GLint iy = y & 1; - const GLint ix = x & 1; - const GLint quadX = x - ix; - const GLint quadY = y - iy; - const GLint mask = (1 << ix) << (2 * iy); + const int iy = y & 1; + const int ix = x & 1; + const int quadX = x - ix; + const int quadY = y - iy; + const int mask = (1 << ix) << (2 * iy); if (quadX != setup->quad.x0 || quadY != setup->quad.y0) @@ -741,10 +742,10 @@ plot(struct setup_stage *setup, GLint x, GLint y) * Determine whether or not to emit a line fragment by checking * line stipple pattern. */ -static INLINE GLuint -stipple_test(GLint counter, GLushort pattern, GLint factor) +static INLINE unsigned +stipple_test(int counter, ushort pattern, int factor) { - GLint b = (counter / factor) & 0xf; + int b = (counter / factor) & 0xf; return (1 << b) & pattern; } @@ -761,13 +762,13 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) struct setup_stage *setup = setup_stage( stage ); struct softpipe_context *sp = setup->softpipe; - GLint x0 = (GLint) v0->data[0][0]; - GLint x1 = (GLint) v1->data[0][0]; - GLint y0 = (GLint) v0->data[0][1]; - GLint y1 = (GLint) v1->data[0][1]; - GLint dx = x1 - x0; - GLint dy = y1 - y0; - GLint xstep, ystep; + int x0 = (int) v0->data[0][0]; + int x1 = (int) v1->data[0][0]; + int y0 = (int) v0->data[0][1]; + int y1 = (int) v1->data[0][1]; + int dx = x1 - x0; + int dy = y1 - y0; + int xstep, ystep; if (dx == 0 && dy == 0) return; @@ -806,10 +807,10 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) if (dx > dy) { /*** X-major line ***/ - GLint i; - const GLint errorInc = dy + dy; - GLint error = errorInc - dx; - const GLint errorDec = error - dx; + int i; + const int errorInc = dy + dy; + int error = errorInc - dx; + const int errorDec = error - dx; for (i = 0; i < dx; i++) { if (!sp->setup.line_stipple_enable || @@ -833,10 +834,10 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) } else { /*** Y-major line ***/ - GLint i; - const GLint errorInc = dx + dx; - GLint error = errorInc - dy; - const GLint errorDec = error - dy; + int i; + const int errorInc = dx + dx; + int error = errorInc - dy; + const int errorDec = error - dy; for (i = 0; i < dy; i++) { if (!sp->setup.line_stipple_enable || @@ -877,12 +878,12 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); /*XXX this should be a vertex attrib! */ - const GLfloat halfSize = 0.5 * setup->softpipe->setup.point_size; - const GLboolean round = setup->softpipe->setup.point_smooth; + const float halfSize = 0.5 * setup->softpipe->setup.point_size; + const boolean round = setup->softpipe->setup.point_smooth; const struct vertex_header *v0 = prim->v[0]; - const GLfloat x = v0->data[FRAG_ATTRIB_WPOS][0]; - const GLfloat y = v0->data[FRAG_ATTRIB_WPOS][1]; - GLuint slot, j; + const float x = v0->data[FRAG_ATTRIB_WPOS][0]; + const float y = v0->data[FRAG_ATTRIB_WPOS][1]; + unsigned slot, j; /* For points, all interpolants are constant-valued. * However, for point sprites, we'll need to setup texcoords appropriately. @@ -912,31 +913,31 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) if (halfSize <= 0.5 && !round) { /* special case for 1-pixel points */ - const GLint ix = ((GLint) x) & 1; - const GLint iy = ((GLint) y) & 1; + const int ix = ((int) x) & 1; + const int iy = ((int) y) & 1; setup->quad.x0 = x - ix; setup->quad.y0 = y - iy; setup->quad.mask = (1 << ix) << (2 * iy); clip_emit_quad(setup); } else { - const GLint ixmin = block((GLint) (x - halfSize)); - const GLint ixmax = block((GLint) (x + halfSize)); - const GLint iymin = block((GLint) (y - halfSize)); - const GLint iymax = block((GLint) (y + halfSize)); - GLint ix, iy; + const int ixmin = block((int) (x - halfSize)); + const int ixmax = block((int) (x + halfSize)); + const int iymin = block((int) (y - halfSize)); + const int iymax = block((int) (y + halfSize)); + int ix, iy; if (round) { /* rounded points */ - const GLfloat rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ - const GLfloat rmax = halfSize + 0.7071F; - const GLfloat rmin2 = MAX2(0.0F, rmin * rmin); - const GLfloat rmax2 = rmax * rmax; - const GLfloat cscale = 1.0F / (rmax2 - rmin2); + const float rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ + const float rmax = halfSize + 0.7071F; + const float rmin2 = MAX2(0.0F, rmin * rmin); + const float rmax2 = rmax * rmax; + const float cscale = 1.0F / (rmax2 - rmin2); for (iy = iymin; iy <= iymax; iy += 2) { for (ix = ixmin; ix <= ixmax; ix += 2) { - GLfloat dx, dy, dist2, cover; + float dx, dy, dist2, cover; setup->quad.mask = 0x0; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h index 0180454a8d..dece42fe45 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.h +++ b/src/mesa/pipe/softpipe/sp_prim_setup.h @@ -37,86 +37,12 @@ * all the enabled attributes run contiguously. */ -#include "glheader.h" -#include "imports.h" -#if 0 -#include "s_tri_public.h" -#include "s_context.h" -#endif +struct draw_stage; +struct softpipe_context; extern struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ); -#if 0 /* UNUSED? */ -struct tri_context; -struct fp_context; -struct be_context; - -/* Note the rasterizer does not take a GLcontext argument. This is - * deliberate. - */ -struct tri_context *tri_create_context( GLcontext *ctx ); - -void tri_destroy_context( struct tri_context *tri ); - -void tri_set_fp_context( struct tri_context *tri, - struct fp_context *fp, - void (*fp_run)( struct fp_context *fp, - const struct fp_inputs *, - struct fp_outputs * )); - - -void tri_set_be_context( struct tri_context *tri, - struct be_context *be, - void (*be_run)( struct be_context *be, - const struct fp_outputs * )); - -void tri_set_attribs( struct tri_context *tri, - const struct attr_info *info, - GLuint nr_attrib ); - -void tri_set_backface( struct tri_context *tri, - GLfloat backface ); - -void tri_set_scissor( struct tri_context *tri, - GLint x, - GLint y, - GLuint width, - GLuint height, - GLboolean enabled ); - -void tri_set_stipple( struct tri_context *tri, - const GLuint *pattern, - GLboolean enabled ); - -/* Unfilled triangles will be handled elsewhere (higher in the - * pipeline), as will things like stipple (lower in the pipeline). - */ - -void tri_triangle( struct tri_context *tri, - const struct vertex *v0, - const struct vertex *v1, - const struct vertex *v2 ); - -/* TODO: rasterize_line, rasterize_point?? - * How will linestipple work? - */ - - -#ifdef SETUP_PRIVATE - -GLboolean tri_setup( struct tri_context *tri, - const struct vertex *v0, - const struct vertex *v1, - const struct vertex *v2 ); - -void tri_rasterize( struct tri_context *tri ); -void tri_rasterize_spans( struct tri_context *tri ); - -#endif - - -#endif #endif /* SP_PRIM_SETUP_H */ diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index 1c51907078..64c1624bd0 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -3,20 +3,19 @@ * quad alpha test */ -#include "glheader.h" -#include "imports.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" static void alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - const GLfloat ref = softpipe->alpha_test.ref; - GLuint passMask = 0x0, j; + const float ref = softpipe->alpha_test.ref; + unsigned passMask = 0x0, j; switch (softpipe->alpha_test.func) { case PIPE_FUNC_NEVER: diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 09d0aa258c..7e8851d214 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -30,10 +30,8 @@ * \author Brian Paul */ -#include "glheader.h" -#include "imports.h" -#include "macros.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" @@ -101,11 +99,11 @@ do { \ static void blend_quad(struct quad_stage *qs, struct quad_header *quad) { - static const GLfloat zero[4] = { 0, 0, 0, 0 }; - static const GLfloat one[4] = { 1, 1, 1, 1 }; + static const float zero[4] = { 0, 0, 0, 0 }; + static const float one[4] = { 1, 1, 1, 1 }; struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); - GLfloat source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; + float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; /* get colors from framebuffer */ sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); @@ -126,7 +124,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_SRC_ALPHA: { - const GLfloat *alpha = quad->outputs.color[3]; + const float *alpha = quad->outputs.color[3]; VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ @@ -139,7 +137,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_DST_ALPHA: { - const GLfloat *alpha = dest[3]; + const float *alpha = dest[3]; VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ @@ -150,7 +148,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_CONST_COLOR: { - GLfloat comp[4]; + float comp[4]; VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ VEC4_MUL(source[0], quad->outputs.color[0], comp); /* R */ VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ @@ -161,7 +159,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_CONST_ALPHA: { - GLfloat alpha[4]; + float alpha[4]; VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ @@ -181,7 +179,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_INV_SRC_COLOR: { - GLfloat inv_comp[4]; + float inv_comp[4]; VEC4_SUB(inv_comp, one, quad->outputs.color[0]); /* R */ VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ VEC4_SUB(inv_comp, one, quad->outputs.color[1]); /* G */ @@ -192,7 +190,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_INV_SRC_ALPHA: { - GLfloat inv_alpha[4]; + float inv_alpha[4]; VEC4_SUB(inv_alpha, one, quad->outputs.color[3]); VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ @@ -201,7 +199,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_INV_DST_ALPHA: { - GLfloat inv_alpha[4]; + float inv_alpha[4]; VEC4_SUB(inv_alpha, one, dest[3]); VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ @@ -210,7 +208,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_INV_DST_COLOR: { - GLfloat inv_comp[4]; + float inv_comp[4]; VEC4_SUB(inv_comp, one, dest[0]); /* R */ VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ VEC4_SUB(inv_comp, one, dest[1]); /* G */ @@ -221,7 +219,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_INV_CONST_COLOR: { - GLfloat inv_comp[4]; + float inv_comp[4]; /* R */ VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[0]); VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); @@ -235,7 +233,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_INV_CONST_ALPHA: { - GLfloat alpha[4], inv_alpha[4]; + float alpha[4], inv_alpha[4]; VEC4_SCALAR(alpha, 1.0 - softpipe->blend_color.color[3]); VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ @@ -261,7 +259,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_SRC_ALPHA: { - const GLfloat *alpha = quad->outputs.color[3]; + const float *alpha = quad->outputs.color[3]; VEC4_MUL(source[3], quad->outputs.color[3], alpha); /* A */ } break; @@ -283,7 +281,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_INV_SRC_ALPHA: { - GLfloat one_minus_alpha[QUAD_SIZE]; + float one_minus_alpha[QUAD_SIZE]; VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ @@ -309,7 +307,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLENDFACTOR_INV_SRC_ALPHA: { - GLfloat one_minus_alpha[QUAD_SIZE]; + float one_minus_alpha[QUAD_SIZE]; VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ } diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c index c459a9e8b0..ae2fb5ef97 100644 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -1,7 +1,5 @@ - -#include "main/glheader.h" -#include "main/imports.h" +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" @@ -15,8 +13,8 @@ static void cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - GLfloat tmp[4][QUAD_SIZE]; - GLuint i; + float tmp[4][QUAD_SIZE]; + unsigned i; assert(sizeof(quad->outputs.color) == sizeof(tmp)); assert(softpipe->framebuffer.num_cbufs <= PIPE_MAX_COLOR_BUFS); diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index 5102a000b5..7bb65bf8c8 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -30,10 +30,8 @@ * \author Brian Paul */ -#include "main/glheader.h" -#include "main/imports.h" -#include "main/macros.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" @@ -46,25 +44,25 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); - GLfloat dest[4][QUAD_SIZE]; + float dest[4][QUAD_SIZE]; sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); /* R */ if (!(softpipe->blend.colormask & PIPE_MASK_R)) - COPY_4FV(quad->outputs.color[0], dest[0]); + COPY_4V(quad->outputs.color[0], dest[0]); /* G */ if (!(softpipe->blend.colormask & PIPE_MASK_G)) - COPY_4FV(quad->outputs.color[1], dest[1]); + COPY_4V(quad->outputs.color[1], dest[1]); /* B */ if (!(softpipe->blend.colormask & PIPE_MASK_B)) - COPY_4FV(quad->outputs.color[2], dest[2]); + COPY_4V(quad->outputs.color[2], dest[2]); /* A */ if (!(softpipe->blend.colormask & PIPE_MASK_A)) - COPY_4FV(quad->outputs.color[3], dest[3]); + COPY_4V(quad->outputs.color[3], dest[3]); /* pass quad to next stage */ qs->next->run(qs->next, quad); diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c index efeb85aab9..8dfec59350 100644 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -32,9 +32,8 @@ */ -#include "main/glheader.h" -#include "main/macros.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" @@ -51,7 +50,7 @@ coverage_quad(struct quad_stage *qs, struct quad_header *quad) if ((softpipe->setup.poly_smooth && quad->prim == PRIM_TRI) || (softpipe->setup.line_smooth && quad->prim == PRIM_LINE) || (softpipe->setup.point_smooth && quad->prim == PRIM_POINT)) { - GLuint j; + unsigned j; for (j = 0; j < QUAD_SIZE; j++) { assert(quad->coverage[j] >= 0.0); assert(quad->coverage[j] <= 1.0); diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 28f264b3c4..5d46e70393 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -26,9 +26,8 @@ * \brief Quad depth testing */ -#include "main/glheader.h" -#include "main/imports.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" @@ -44,11 +43,11 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); - GLuint bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ - GLuint qzzzz[QUAD_SIZE]; /**< Z values from the quad */ - GLuint zmask = 0; - GLuint j; - GLfloat scale; + unsigned bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ + unsigned qzzzz[QUAD_SIZE]; /**< Z values from the quad */ + unsigned zmask = 0; + unsigned j; + float scale; assert(sps); /* shouldn't get here if there's no zbuffer */ @@ -72,7 +71,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) * Z-fighting errors. */ for (j = 0; j < QUAD_SIZE; j++) { - qzzzz[j] = (GLuint) (quad->outputs.depth[j] * scale); + qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); } /* get zquad from zbuffer */ diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index e767d0a470..edb3245cba 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -32,8 +32,6 @@ * all the enabled attributes run contiguously. */ -#include "main/mtypes.h" - #include "pipe/p_util.h" #include "tgsi/core/tgsi_core.h" @@ -42,6 +40,8 @@ #include "sp_quad.h" #include "sp_tex_sample.h" +#include "main/mtypes.h" + #if 0 #if defined __GNUC__ #define ALIGNED_ATTRIBS 1 @@ -73,9 +73,9 @@ struct exec_machine { const struct setup_coefficient *coef; /**< will point to quad->coef */ #if ALIGNED_ATTRIBS - GLfloat attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE] __attribute__(( aligned( 16 ) )); + float attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE] __attribute__(( aligned( 16 ) )); #else - GLfloat attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE]; + float attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE]; #endif }; @@ -84,10 +84,10 @@ struct exec_machine { * Compute quad's attributes values, as constants (GL_FLAT shading). */ static INLINE void cinterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) + unsigned attrib, + unsigned i ) { - GLuint j; + unsigned j; for (j = 0; j < QUAD_SIZE; j++) { exec->attr[attrib][i][j] = exec->coef[attrib].a0[i]; @@ -104,14 +104,14 @@ static INLINE void cinterp( struct exec_machine *exec, * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy */ static INLINE void linterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) + unsigned attrib, + unsigned i ) { - GLuint j; + unsigned j; for (j = 0; j < QUAD_SIZE; j++) { - const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; + const float x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const float y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + exec->coef[attrib].dadx[i] * x + exec->coef[attrib].dady[i] * y); @@ -132,16 +132,16 @@ static INLINE void linterp( struct exec_machine *exec, * */ static INLINE void pinterp( struct exec_machine *exec, - GLuint attrib, - GLuint i ) + unsigned attrib, + unsigned i ) { - GLuint j; + unsigned j; for (j = 0; j < QUAD_SIZE; j++) { - const GLfloat x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const GLfloat y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; + const float x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; + const float y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; /* FRAG_ATTRIB_WPOS.w here is really 1/w */ - const GLfloat w = 1.0 / exec->attr[FRAG_ATTRIB_WPOS][3][j]; + const float w = 1.0 / exec->attr[FRAG_ATTRIB_WPOS][3][j]; exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + exec->coef[attrib].dadx[i] * x + exec->coef[attrib].dady[i] * y) * w); @@ -158,9 +158,9 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) struct quad_shade_stage *qss = quad_shade_stage(qs); struct softpipe_context *softpipe = qs->softpipe; struct exec_machine exec; - const GLfloat fx = quad->x0; - const GLfloat fy = quad->y0; - GLuint attr, i; + const float fx = quad->x0; + const float fy = quad->y0; + unsigned attr, i; exec.coef = quad->coef; @@ -216,7 +216,7 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) struct tgsi_exec_machine machine; struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; struct tgsi_exec_vector *aoutputs; - GLuint i; + unsigned i; #if !ALIGNED_ATTRIBS struct tgsi_exec_vector inputs[FRAG_ATTRIB_MAX + 1]; @@ -289,7 +289,7 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) } #else { - GLuint attr = softpipe->fp_attr_to_slot[FRAG_ATTRIB_COL0]; + unsigned attr = softpipe->fp_attr_to_slot[FRAG_ATTRIB_COL0]; assert(attr); memcpy(quad->outputs.color, @@ -318,7 +318,7 @@ static void shade_begin(struct quad_stage *qs) { struct quad_shade_stage *qss = quad_shade_stage(qs); struct softpipe_context *softpipe = qs->softpipe; - GLuint i, entry; + unsigned i, entry; for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { qss->samplers[i].state = &softpipe->sampler[i]; diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c index 8073f94553..6b094a5bef 100644 --- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c +++ b/src/mesa/pipe/softpipe/sp_quad_occlusion.c @@ -32,9 +32,8 @@ */ -#include "main/glheader.h" -#include "main/imports.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index 62f466be5d..49420c4fe7 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -31,20 +31,18 @@ * Attributes are assumed to be 4 floats wide but are packed so that * all the enabled attributes run contiguously. */ - -#include "glheader.h" -#include "imports.h" +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" #include "sp_quad.h" -static void mask_copy( GLfloat (*dest)[4], - GLfloat (*src)[4], - GLuint mask ) +static void mask_copy( float (*dest)[4], + float (*src)[4], + unsigned mask ) { - GLuint i, j; + unsigned i, j; for (i = 0; i < 4; i++) { if (mask & (1<cbuf); if (quad->mask != MASK_ALL) { - GLfloat tmp[4][QUAD_SIZE]; + float tmp[4][QUAD_SIZE]; /* XXX probably add a masked-write function someday */ diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 857f1a5989..47b3b4f089 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -4,13 +4,12 @@ */ -#include "main/glheader.h" -#include "main/imports.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" #include "sp_quad.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" /** Only 8-bit stencil supported */ @@ -28,12 +27,12 @@ * values and ref value are to be used. * \return mask indicating which pixels passed the stencil test */ -static GLbitfield -do_stencil_test(const GLubyte stencilVals[QUAD_SIZE], GLuint func, - GLbitfield ref, GLbitfield valMask) +static unsigned +do_stencil_test(const ubyte stencilVals[QUAD_SIZE], unsigned func, + unsigned ref, unsigned valMask) { - GLbitfield passMask = 0x0; - GLuint j; + unsigned passMask = 0x0; + unsigned j; ref &= valMask; @@ -105,11 +104,11 @@ do_stencil_test(const GLubyte stencilVals[QUAD_SIZE], GLuint func, * stencil values */ static void -apply_stencil_op(GLubyte stencilVals[QUAD_SIZE], - GLbitfield mask, GLuint op, GLubyte ref, GLubyte wrtMask) +apply_stencil_op(ubyte stencilVals[QUAD_SIZE], + unsigned mask, unsigned op, ubyte ref, ubyte wrtMask) { - GLuint j; - GLubyte newstencil[QUAD_SIZE]; + unsigned j; + ubyte newstencil[QUAD_SIZE]; for (j = 0; j < QUAD_SIZE; j++) { newstencil[j] = stencilVals[j]; @@ -202,9 +201,9 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *s_surf = softpipe_surface(softpipe->framebuffer.sbuf); - GLuint func, zFailOp, zPassOp, failOp; - GLubyte ref, wrtMask, valMask; - GLubyte stencilVals[QUAD_SIZE]; + unsigned func, zFailOp, zPassOp, failOp; + ubyte ref, wrtMask, valMask; + ubyte stencilVals[QUAD_SIZE]; /* choose front or back face function, operator, etc */ /* XXX we could do these initializations once per primitive */ @@ -232,7 +231,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) /* do the stencil test first */ { - GLbitfield passMask, failMask; + unsigned passMask, failMask; passMask = do_stencil_test(stencilVals, func, ref, valMask); failMask = quad->mask & ~passMask; quad->mask &= passMask; @@ -246,18 +245,18 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) /* now the pixels that passed the stencil test are depth tested */ if (softpipe->depth_test.enabled) { - const GLbitfield origMask = quad->mask; + const unsigned origMask = quad->mask; sp_depth_test_quad(qs, quad); /* quad->mask is updated */ /* update stencil buffer values according to z pass/fail result */ if (zFailOp != PIPE_STENCIL_OP_KEEP) { - const GLbitfield failMask = origMask & ~quad->mask; + const unsigned failMask = origMask & ~quad->mask; apply_stencil_op(stencilVals, failMask, zFailOp, ref, wrtMask); } if (zPassOp != PIPE_STENCIL_OP_KEEP) { - const GLbitfield passMask = origMask & quad->mask; + const unsigned passMask = origMask & quad->mask; apply_stencil_op(stencilVals, passMask, zPassOp, ref, wrtMask); } } diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index 6e93bf81f1..cb127095d7 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -3,12 +3,11 @@ * quad polygon stipple stage */ -#include "glheader.h" -#include "imports.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" /** @@ -19,10 +18,10 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) { if (quad->prim == PRIM_TRI) { struct softpipe_context *softpipe = qs->softpipe; - const GLint col0 = quad->x0 % 32; - const GLint row0 = quad->y0 % 32; - const GLuint stipple0 = softpipe->poly_stipple.stipple[row0]; - const GLuint stipple1 = softpipe->poly_stipple.stipple[row0 + 1]; + const int col0 = quad->x0 % 32; + const int row0 = quad->y0 % 32; + const unsigned stipple0 = softpipe->poly_stipple.stipple[row0]; + const unsigned stipple1 = softpipe->poly_stipple.stipple[row0 + 1]; /* XXX there may be a better way to lay out the stored stipple * values to further simplify this computation. diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 10b7479e97..231ea1ea22 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -54,7 +54,7 @@ sp_region_idle(struct pipe_context *pipe, struct pipe_region *region) } -static GLubyte * +static ubyte * sp_region_map(struct pipe_context *pipe, struct pipe_region *region) { struct softpipe_context *sp = softpipe_context( pipe ); @@ -83,7 +83,7 @@ sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) static struct pipe_region * sp_region_alloc(struct pipe_context *pipe, - GLuint cpp, GLuint width, GLuint height, GLbitfield flags) + unsigned cpp, unsigned width, unsigned height, unsigned flags) { struct softpipe_context *sp = softpipe_context( pipe ); struct pipe_region *region = calloc(sizeof(*region), 1); @@ -112,7 +112,7 @@ sp_region_release(struct pipe_context *pipe, struct pipe_region **region) if (!*region) return; - ASSERT((*region)->refcount > 0); + assert((*region)->refcount > 0); (*region)->refcount--; if ((*region)->refcount == 0) { @@ -130,19 +130,19 @@ sp_region_release(struct pipe_context *pipe, struct pipe_region **region) * XXX Move this into core Mesa? */ static void -_mesa_copy_rect(GLubyte * dst, - GLuint cpp, - GLuint dst_pitch, - GLuint dst_x, - GLuint dst_y, - GLuint width, - GLuint height, - const GLubyte * src, - GLuint src_pitch, - GLuint src_x, - GLuint src_y) +_mesa_copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + unsigned src_pitch, + unsigned src_x, + unsigned src_y) { - GLuint i; + unsigned i; dst_pitch *= cpp; src_pitch *= cpp; @@ -174,10 +174,10 @@ _mesa_copy_rect(GLubyte * dst, static void sp_region_data(struct pipe_context *pipe, struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - const void *src, GLuint src_pitch, - GLuint srcx, GLuint srcy, GLuint width, GLuint height) + unsigned dst_offset, + unsigned dstx, unsigned dsty, + const void *src, unsigned src_pitch, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) { _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, dst->cpp, @@ -193,11 +193,11 @@ sp_region_data(struct pipe_context *pipe, static void sp_region_copy(struct pipe_context *pipe, struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, + unsigned dst_offset, + unsigned dstx, unsigned dsty, struct pipe_region *src, - GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height) + unsigned src_offset, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) { assert( dst->cpp == src->cpp ); @@ -217,8 +217,8 @@ sp_region_copy(struct pipe_context *pipe, /* Fill a rectangular sub-region. Need better logic about when to * push buffers into AGP - will currently do so whenever possible. */ -static GLubyte * -get_pointer(struct pipe_region *dst, GLuint x, GLuint y) +static ubyte * +get_pointer(struct pipe_region *dst, unsigned x, unsigned y) { return dst->map + (y * dst->pitch + x) * dst->cpp; } @@ -227,17 +227,17 @@ get_pointer(struct pipe_region *dst, GLuint x, GLuint y) static void sp_region_fill(struct pipe_context *pipe, struct pipe_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - GLuint width, GLuint height, GLuint value) + unsigned dst_offset, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) { - GLuint i, j; + unsigned i, j; (void)pipe->region_map(pipe, dst); switch (dst->cpp) { case 1: { - GLubyte *row = get_pointer(dst, dstx, dsty); + ubyte *row = get_pointer(dst, dstx, dsty); for (i = 0; i < height; i++) { memset(row, value, width); row += dst->pitch; @@ -245,7 +245,7 @@ sp_region_fill(struct pipe_context *pipe, } break; case 2: { - GLushort *row = (GLushort *) get_pointer(dst, dstx, dsty); + ushort *row = (ushort *) get_pointer(dst, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = value; @@ -254,7 +254,7 @@ sp_region_fill(struct pipe_context *pipe, } break; case 4: { - GLuint *row = (GLuint *) get_pointer(dst, dstx, dsty); + unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = value; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 3c572cdb6a..5adf5930b8 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -31,7 +31,6 @@ #ifndef SP_STATE_H #define SP_STATE_H -#include "glheader.h" #include "pipe/p_state.h" @@ -69,14 +68,14 @@ void softpipe_set_setup_state( struct pipe_context *, const struct pipe_setup_state * ); void softpipe_set_sampler_state( struct pipe_context *, - GLuint unit, + unsigned unit, const struct pipe_sampler_state * ); void softpipe_set_stencil_state( struct pipe_context *, const struct pipe_stencil_state * ); void softpipe_set_texture_state( struct pipe_context *, - GLuint unit, + unsigned unit, struct pipe_mipmap_tree * ); void softpipe_set_viewport_state( struct pipe_context *, diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 8bc22b0efc..b2e85d86cc 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -27,8 +27,6 @@ /* Authors: Keith Whitwell */ -#include "imports.h" - #include "sp_context.h" #include "sp_state.h" diff --git a/src/mesa/pipe/softpipe/sp_state_clip.c b/src/mesa/pipe/softpipe/sp_state_clip.c index 8cf4383d3f..08c5f06d05 100644 --- a/src/mesa/pipe/softpipe/sp_state_clip.c +++ b/src/mesa/pipe/softpipe/sp_state_clip.c @@ -27,8 +27,6 @@ /* Authors: Keith Whitwell */ -#include "imports.h" - #include "sp_context.h" #include "sp_state.h" #include "pipe/draw/draw_context.h" diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index e1faaed93c..0a2cfbb7d1 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -46,7 +46,7 @@ do { \ } while (0) -static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = +static const unsigned frag_to_vf[FRAG_ATTRIB_MAX] = { VF_ATTRIB_POS, VF_ATTRIB_COLOR0, @@ -78,27 +78,27 @@ static const GLuint frag_to_vf[FRAG_ATTRIB_MAX] = */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - const GLbitfield inputsRead = softpipe->fs.inputs_read; - GLuint slot_to_vf_attr[VF_ATTRIB_MAX]; - GLbitfield attr_mask = 0x0; - GLuint i; + const unsigned inputsRead = softpipe->fs.inputs_read; + unsigned slot_to_vf_attr[VF_ATTRIB_MAX]; + unsigned attr_mask = 0x0; + unsigned i; /* Need Z if depth test is enabled or the fragment program uses the * fragment position (XYZW). */ if (softpipe->depth_test.enabled || (inputsRead & FRAG_ATTRIB_WPOS)) - softpipe->need_z = GL_TRUE; + softpipe->need_z = TRUE; else - softpipe->need_z = GL_FALSE; + softpipe->need_z = FALSE; /* Need W if we do any perspective-corrected interpolation or the * fragment program uses the fragment position. */ if (inputsRead & FRAG_ATTRIB_WPOS) - softpipe->need_w = GL_TRUE; + softpipe->need_w = TRUE; else - softpipe->need_w = GL_FALSE; + softpipe->need_w = FALSE; softpipe->nr_attrs = 0; @@ -130,7 +130,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) if (inputsRead & (1 << i)) { assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); - softpipe->need_w = GL_TRUE; + softpipe->need_w = TRUE; } } @@ -169,7 +169,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) static void compute_cliprect(struct softpipe_context *sp) { - GLint surfWidth, surfHeight; + int surfWidth, surfHeight; if (sp->framebuffer.num_cbufs > 0) { surfWidth = sp->framebuffer.cbufs[0]->width; diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 34cf210834..2e3d0c3d65 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -29,8 +29,6 @@ * Brian Paul */ -#include "imports.h" - #include "sp_context.h" #include "sp_state.h" @@ -38,7 +36,7 @@ void softpipe_set_sampler_state(struct pipe_context *pipe, - GLuint unit, + unsigned unit, const struct pipe_sampler_state *sampler) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -52,7 +50,7 @@ softpipe_set_sampler_state(struct pipe_context *pipe, void softpipe_set_texture_state(struct pipe_context *pipe, - GLuint unit, + unsigned unit, struct pipe_mipmap_tree *texture) { struct softpipe_context *softpipe = softpipe_context(pipe); diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index 8ce81eb2b0..a534ffb2c2 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -27,8 +27,6 @@ /* Authors: Keith Whitwell */ -#include "imports.h" - #include "sp_context.h" #include "sp_state.h" #include "sp_surface.h" diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 1919947abf..4152e3510d 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -29,8 +29,7 @@ #include "sp_state.h" #include "sp_surface.h" #include "pipe/p_defines.h" -#include "main/imports.h" -#include "main/macros.h" +#include "pipe/p_util.h" /** @@ -48,13 +47,13 @@ /*** PIPE_FORMAT_U_A8_R8_G8_B8 ***/ static void -a8r8g8b8_read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) +a8r8g8b8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) { - const GLuint *src - = ((const GLuint *) (sps->surface.region->map + sps->surface.offset)) + const unsigned *src + = ((const unsigned *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; - GLuint i, j; + unsigned i, j; assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); assert(x < sps->surface.width - 1); @@ -62,7 +61,7 @@ a8r8g8b8_read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, for (i = 0; i < 2; i++) { /* loop over pixel row */ for (j = 0; j < 2; j++) { /* loop over pixel column */ - const GLuint p = src[j]; + const unsigned p = src[j]; rrrr[0][i * 2 + j] = UBYTE_TO_FLOAT((p >> 16) & 0xff); /*R*/ rrrr[1][i * 2 + j] = UBYTE_TO_FLOAT((p >> 8) & 0xff); /*G*/ rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT((p ) & 0xff); /*B*/ @@ -73,19 +72,19 @@ a8r8g8b8_read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, } static void -a8r8g8b8_write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) +a8r8g8b8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) { - GLuint *dst - = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + unsigned *dst + = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; - GLuint i, j; + unsigned i, j; assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); for (i = 0; i < 2; i++) { /* loop over pixel row */ for (j = 0; j < 2; j++) { /* loop over pixel column */ - GLubyte r, g, b, a; + ubyte r, g, b, a; UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/ UNCLAMPED_FLOAT_TO_UBYTE(g, rrrr[1][i * 2 + j]); /*G*/ UNCLAMPED_FLOAT_TO_UBYTE(b, rrrr[2][i * 2 + j]); /*B*/ @@ -98,13 +97,13 @@ a8r8g8b8_write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, static void a8r8g8b8_get_tile(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) + unsigned x, unsigned y, unsigned w, unsigned h, float *p) { - const GLuint *src - = ((const GLuint *) (ps->region->map + ps->offset)) + const unsigned *src + = ((const unsigned *) (ps->region->map + ps->offset)) + y * ps->region->pitch + x; - GLuint i, j; - GLuint w0 = w; + unsigned i, j; + unsigned w0 = w; assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); @@ -119,9 +118,9 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, h = ps->height -y; #endif for (i = 0; i < h; i++) { - GLfloat *pRow = p; + float *pRow = p; for (j = 0; j < w; j++) { - const GLuint pixel = src[j]; + const unsigned pixel = src[j]; pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); @@ -138,18 +137,18 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, static void a1r5g5b5_get_tile(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) + unsigned x, unsigned y, unsigned w, unsigned h, float *p) { - const GLushort *src - = ((const GLushort *) (ps->region->map + ps->offset)) + const ushort *src + = ((const ushort *) (ps->region->map + ps->offset)) + y * ps->region->pitch + x; - GLuint i, j; + unsigned i, j; assert(ps->format == PIPE_FORMAT_U_A1_R5_G5_B5); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { - const GLushort pixel = src[j]; + const ushort pixel = src[j]; p[0] = ((pixel >> 10) & 0x1f) * (1.0 / 31); p[1] = ((pixel >> 5) & 0x1f) * (1.0 / 31); p[2] = ((pixel ) & 0x1f) * (1.0 / 31); @@ -166,15 +165,15 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, static void z16_read_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) + int x, int y, unsigned zzzz[QUAD_SIZE]) { - const GLushort *src - = ((const GLushort *) (sps->surface.region->map + sps->surface.offset)) + const ushort *src + = ((const ushort *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z16); - /* converting GLushort to GLuint: */ + /* converting ushort to unsigned: */ zzzz[0] = src[0]; zzzz[1] = src[1]; src += sps->surface.region->pitch; @@ -184,15 +183,15 @@ z16_read_quad_z(struct softpipe_surface *sps, static void z16_write_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) + int x, int y, const unsigned zzzz[QUAD_SIZE]) { - GLushort *dst - = ((GLushort *) (sps->surface.region->map + sps->surface.offset)) + ushort *dst + = ((ushort *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z16); - /* converting GLuint to GLushort: */ + /* converting unsigned to ushort: */ dst[0] = zzzz[0]; dst[1] = zzzz[1]; dst += sps->surface.region->pitch; @@ -205,10 +204,10 @@ z16_write_quad_z(struct softpipe_surface *sps, static void z32_read_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) + int x, int y, unsigned zzzz[QUAD_SIZE]) { - const GLuint *src - = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + const unsigned *src + = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z32); @@ -222,10 +221,10 @@ z32_read_quad_z(struct softpipe_surface *sps, static void z32_write_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) + int x, int y, const unsigned zzzz[QUAD_SIZE]) { - GLuint *dst - = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + unsigned *dst + = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_U_Z32); @@ -242,11 +241,11 @@ z32_write_quad_z(struct softpipe_surface *sps, static void s8z24_read_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) + int x, int y, unsigned zzzz[QUAD_SIZE]) { - static const GLuint mask = 0x00ffffff; - const GLuint *src - = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + static const unsigned mask = 0x00ffffff; + const unsigned *src + = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_S8_Z24); @@ -261,11 +260,11 @@ s8z24_read_quad_z(struct softpipe_surface *sps, static void s8z24_write_quad_z(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) + int x, int y, const unsigned zzzz[QUAD_SIZE]) { - static const GLuint mask = 0xff000000; - GLuint *dst - = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + static const unsigned mask = 0xff000000; + unsigned *dst + = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_S8_Z24); @@ -280,10 +279,10 @@ s8z24_write_quad_z(struct softpipe_surface *sps, static void s8z24_read_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) + int x, int y, ubyte ssss[QUAD_SIZE]) { - const GLuint *src - = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + const unsigned *src + = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_S8_Z24); @@ -297,11 +296,11 @@ s8z24_read_quad_stencil(struct softpipe_surface *sps, static void s8z24_write_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) + int x, int y, const ubyte ssss[QUAD_SIZE]) { - static const GLuint mask = 0x00ffffff; - GLuint *dst - = ((GLuint *) (sps->surface.region->map + sps->surface.offset)) + static const unsigned mask = 0x00ffffff; + unsigned *dst + = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) + y * sps->surface.region->pitch + x; assert(sps->surface.format == PIPE_FORMAT_S8_Z24); @@ -318,9 +317,9 @@ s8z24_write_quad_stencil(struct softpipe_surface *sps, static void s8_read_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) + int x, int y, ubyte ssss[QUAD_SIZE]) { - const GLubyte *src + const ubyte *src = sps->surface.region->map + sps->surface.offset + y * sps->surface.region->pitch + x; @@ -335,9 +334,9 @@ s8_read_quad_stencil(struct softpipe_surface *sps, static void s8_write_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) + int x, int y, const ubyte ssss[QUAD_SIZE]) { - GLubyte *dst + ubyte *dst = sps->surface.region->map + sps->surface.offset + y * sps->surface.region->pitch + x; @@ -393,7 +392,7 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) static struct pipe_surface * -softpipe_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) +softpipe_surface_alloc(struct pipe_context *pipe, unsigned pipeFormat) { struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); if (!sps) @@ -419,17 +418,17 @@ softpipe_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) struct pipe_surface * softpipe_get_tex_surface(struct pipe_context *pipe, struct pipe_mipmap_tree *mt, - GLuint face, GLuint level, GLuint zslice) + unsigned face, unsigned level, unsigned zslice) { struct pipe_surface *ps; - GLuint offset; /* in bytes */ + unsigned offset; /* in bytes */ offset = mt->level[level].level_offset; - if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) { + if (mt->target == PIPE_TEXTURE_CUBE) { offset += mt->level[level].image_offset[face] * mt->cpp; } - else if (mt->target == GL_TEXTURE_3D) { + else if (mt->target == PIPE_TEXTURE_3D) { offset += mt->level[level].image_offset[zslice] * mt->cpp; } else { diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 00b5edcd92..545983f431 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -31,7 +31,6 @@ #ifndef SP_SURFACE_H #define SP_SURFACE_H -#include "glheader.h" #include "sp_headers.h" #include "pipe/p_state.h" @@ -50,47 +49,47 @@ struct softpipe_surface { * Functions for read/writing surface data */ void (*read_quad_f)( struct softpipe_surface *, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ); + int x, int y, + float (*rgba)[NUM_CHANNELS] ); void (*read_quad_f_swz)( struct softpipe_surface *, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ); + int x, int y, + float (*rrrr)[QUAD_SIZE] ); void (*read_quad_ub)( struct softpipe_surface *, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ); + int x, int y, + ubyte (*rgba)[NUM_CHANNELS] ); void (*write_quad_f)( struct softpipe_surface *, - GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS] ); + int x, int y, + float (*rgba)[NUM_CHANNELS] ); void (*write_quad_f_swz)( struct softpipe_surface *, - GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE] ); + int x, int y, + float (*rrrr)[QUAD_SIZE] ); void (*write_quad_ub)( struct softpipe_surface *, - GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS] ); + int x, int y, + ubyte (*rgba)[NUM_CHANNELS] ); void (*read_quad_z)(struct softpipe_surface *, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]); + int x, int y, unsigned zzzz[QUAD_SIZE]); void (*write_quad_z)(struct softpipe_surface *, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]); + int x, int y, const unsigned zzzz[QUAD_SIZE]); void (*read_quad_stencil)(struct softpipe_surface *, - GLint x, GLint y, GLubyte ssss[QUAD_SIZE]); + int x, int y, ubyte ssss[QUAD_SIZE]); void (*write_quad_stencil)(struct softpipe_surface *, - GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]); + int x, int y, const ubyte ssss[QUAD_SIZE]); }; extern struct pipe_surface * softpipe_get_tex_surface(struct pipe_context *pipe, struct pipe_mipmap_tree *mt, - GLuint face, GLuint level, GLuint zslice); + unsigned face, unsigned level, unsigned zslice); extern void diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c index fda1b46ab4..0517c01b5d 100644 --- a/src/mesa/pipe/softpipe/sp_tex_layout.c +++ b/src/mesa/pipe/softpipe/sp_tex_layout.c @@ -30,10 +30,10 @@ * Michel Dänzer */ -#include "macros.h" #include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_tex_layout.h" @@ -43,7 +43,7 @@ * little. */ -static GLuint minify( GLuint d ) +static unsigned minify( unsigned d ) { return MAX2(1, d>>1); } @@ -56,11 +56,11 @@ static int align(int value, int alignment) static void sp_miptree_set_level_info(struct pipe_mipmap_tree *mt, - GLuint level, - GLuint nr_images, - GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) + unsigned level, + unsigned nr_images, + unsigned x, unsigned y, unsigned w, unsigned h, unsigned d) { - assert(level < MAX_TEXTURE_LEVELS); + assert(level < PIPE_MAX_TEXTURE_LEVELS); mt->level[level].width = w; mt->level[level].height = h; @@ -83,14 +83,14 @@ sp_miptree_set_level_info(struct pipe_mipmap_tree *mt, assert(nr_images); assert(!mt->level[level].image_offset); - mt->level[level].image_offset = (GLuint *) malloc(nr_images * sizeof(GLuint)); + mt->level[level].image_offset = (unsigned *) malloc(nr_images * sizeof(unsigned)); mt->level[level].image_offset[0] = 0; } static void sp_miptree_set_image_offset(struct pipe_mipmap_tree *mt, - GLuint level, GLuint img, GLuint x, GLuint y) + unsigned level, unsigned img, unsigned x, unsigned y) { if (img == 0 && level == 0) assert(x == 0 && y == 0); @@ -109,12 +109,12 @@ sp_miptree_set_image_offset(struct pipe_mipmap_tree *mt, static void sp_miptree_layout_2d( struct pipe_mipmap_tree *mt ) { - GLint align_h = 2, align_w = 4; - GLuint level; - GLuint x = 0; - GLuint y = 0; - GLuint width = mt->width0; - GLuint height = mt->height0; + int align_h = 2, align_w = 4; + unsigned level; + unsigned x = 0; + unsigned y = 0; + unsigned width = mt->width0; + unsigned height = mt->height0; mt->pitch = mt->width0; /* XXX FIX THIS: @@ -130,7 +130,7 @@ sp_miptree_layout_2d( struct pipe_mipmap_tree *mt ) * 2nd mipmap out past the width of its parent. */ if (mt->first_level != mt->last_level) { - GLuint mip1_width = align(minify(mt->width0), align_w) + unsigned mip1_width = align(minify(mt->width0), align_w) + minify(minify(mt->width0)); if (mip1_width > mt->width0) @@ -144,7 +144,7 @@ sp_miptree_layout_2d( struct pipe_mipmap_tree *mt ) mt->total_height = 0; for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { - GLuint img_height; + unsigned img_height; sp_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); @@ -174,7 +174,7 @@ sp_miptree_layout_2d( struct pipe_mipmap_tree *mt ) } -static const GLint initial_offsets[6][2] = { +static const int initial_offsets[6][2] = { {0, 0}, {0, 2}, {1, 0}, @@ -183,7 +183,7 @@ static const GLint initial_offsets[6][2] = { {1, 3} }; -static const GLint step_offsets[6][2] = { +static const int step_offsets[6][2] = { {0, 2}, {0, 2}, {-1, 2}, @@ -194,16 +194,16 @@ static const GLint step_offsets[6][2] = { -GLboolean +boolean softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) { - GLint level; + int level; switch (mt->target) { - case GL_TEXTURE_CUBE_MAP:{ - const GLuint dim = mt->width0; - GLuint face; - GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; + case PIPE_TEXTURE_CUBE:{ + const unsigned dim = mt->width0; + unsigned face; + unsigned lvlWidth = mt->width0, lvlHeight = mt->height0; assert(lvlWidth == lvlHeight); /* cubemap images are square */ @@ -230,9 +230,9 @@ softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * for (face = 0; face < 6; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; + unsigned x = initial_offsets[face][0] * dim; + unsigned y = initial_offsets[face][1] * dim; + unsigned d = dim; if (dim == 4 && face >= 4) { y = mt->total_height - 4; @@ -286,13 +286,13 @@ softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * } break; } - case GL_TEXTURE_3D:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint pack_x_pitch, pack_x_nr; - GLuint pack_y_pitch; - GLuint level; + case PIPE_TEXTURE_3D:{ + unsigned width = mt->width0; + unsigned height = mt->height0; + unsigned depth = mt->depth0; + unsigned pack_x_pitch, pack_x_nr; + unsigned pack_y_pitch; + unsigned level; mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; mt->total_height = 0; @@ -302,10 +302,10 @@ softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * pack_x_nr = 1; for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; - GLint x = 0; - GLint y = 0; - GLint q, j; + unsigned nr_images = mt->target == PIPE_TEXTURE_3D ? depth : 6; + int x = 0; + int y = 0; + int q, j; sp_miptree_set_level_info(mt, level, nr_images, 0, mt->total_height, @@ -341,13 +341,14 @@ softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * break; } - case GL_TEXTURE_1D: - case GL_TEXTURE_2D: - case GL_TEXTURE_RECTANGLE_ARB: + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_2D: +// case PIPE_TEXTURE_RECTANGLE: sp_miptree_layout_2d(mt); break; default: - _mesa_problem(NULL, "Unexpected tex target in sp_miptree_layout()"); + assert(0); + break; } /* @@ -356,6 +357,6 @@ softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); */ - return GL_TRUE; + return TRUE; } diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.h b/src/mesa/pipe/softpipe/sp_tex_layout.h index be85e4be58..ea19c13b23 100644 --- a/src/mesa/pipe/softpipe/sp_tex_layout.h +++ b/src/mesa/pipe/softpipe/sp_tex_layout.h @@ -6,7 +6,7 @@ struct pipe_context; struct pipe_mipmap_tree; -extern GLboolean +extern boolean softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree *mt); diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 03fb539e7d..3130c5cdae 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -33,12 +33,12 @@ */ -#include "main/macros.h" #include "sp_context.h" #include "sp_surface.h" #include "sp_tex_sample.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "pipe/tgsi/core/tgsi_exec.h" @@ -49,7 +49,7 @@ * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0. * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x). */ -#define FRAC(f) ((f) - IFLOOR(f)) +#define FRAC(f) ((f) - ifloor(f)) /** @@ -80,8 +80,8 @@ lerp_2d(float a, float b, * Compute the remainder of a divided by b, but be careful with * negative values so that REPEAT mode works right. */ -static INLINE GLint -repeat_remainder(GLint a, GLint b) +static INLINE int +repeat_remainder(int a, int b) { if (a >= 0) return a % b; @@ -97,15 +97,15 @@ repeat_remainder(GLint a, GLint b) * \param size the texture image size * \return integer texture index */ -static INLINE GLint -nearest_texcoord(GLuint wrapMode, float s, GLuint size) +static INLINE int +nearest_texcoord(unsigned wrapMode, float s, unsigned size) { - GLint i; + int i; switch (wrapMode) { case PIPE_TEX_WRAP_REPEAT: /* s limited to [0,1) */ /* i limited to [0,size-1] */ - i = IFLOOR(s * size); + i = ifloor(s * size); i = repeat_remainder(i, size); return i; case PIPE_TEX_WRAP_CLAMP: @@ -116,7 +116,7 @@ nearest_texcoord(GLuint wrapMode, float s, GLuint size) else if (s >= 1.0F) i = size - 1; else - i = IFLOOR(s * size); + i = ifloor(s * size); return i; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: { @@ -129,7 +129,7 @@ nearest_texcoord(GLuint wrapMode, float s, GLuint size) else if (s > max) i = size - 1; else - i = IFLOOR(s * size); + i = ifloor(s * size); } return i; case PIPE_TEX_WRAP_CLAMP_TO_BORDER: @@ -143,14 +143,14 @@ nearest_texcoord(GLuint wrapMode, float s, GLuint size) else if (s >= max) i = size; else - i = IFLOOR(s * size); + i = ifloor(s * size); } return i; case PIPE_TEX_WRAP_MIRROR_REPEAT: { const float min = 1.0F / (2.0F * size); const float max = 1.0F - min; - const GLint flr = IFLOOR(s); + const int flr = ifloor(s); float u; if (flr & 1) u = 1.0F - (s - (float) flr); @@ -161,7 +161,7 @@ nearest_texcoord(GLuint wrapMode, float s, GLuint size) else if (u > max) i = size - 1; else - i = IFLOOR(u * size); + i = ifloor(u * size); } return i; case PIPE_TEX_WRAP_MIRROR_CLAMP: @@ -174,7 +174,7 @@ nearest_texcoord(GLuint wrapMode, float s, GLuint size) else if (u >= 1.0F) i = size - 1; else - i = IFLOOR(u * size); + i = ifloor(u * size); } return i; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: @@ -189,7 +189,7 @@ nearest_texcoord(GLuint wrapMode, float s, GLuint size) else if (u > max) i = size - 1; else - i = IFLOOR(u * size); + i = ifloor(u * size); } return i; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: @@ -204,7 +204,7 @@ nearest_texcoord(GLuint wrapMode, float s, GLuint size) else if (u > max) i = size; else - i = IFLOOR(u * size); + i = ifloor(u * size); } return i; default: @@ -224,14 +224,14 @@ nearest_texcoord(GLuint wrapMode, float s, GLuint size) * \param a returns blend factor/weight between texture indexes */ static INLINE void -linear_texcoord(GLuint wrapMode, float s, GLuint size, - GLint *i0, GLint *i1, float *a) +linear_texcoord(unsigned wrapMode, float s, unsigned size, + int *i0, int *i1, float *a) { float u; switch (wrapMode) { case PIPE_TEX_WRAP_REPEAT: u = s * size - 0.5F; - *i0 = repeat_remainder(IFLOOR(u), size); + *i0 = repeat_remainder(ifloor(u), size); *i1 = repeat_remainder(*i0 + 1, size); break; case PIPE_TEX_WRAP_CLAMP: @@ -242,7 +242,7 @@ linear_texcoord(GLuint wrapMode, float s, GLuint size, else u = s * size; u -= 0.5F; - *i0 = IFLOOR(u); + *i0 = ifloor(u); *i1 = *i0 + 1; break; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: @@ -253,11 +253,11 @@ linear_texcoord(GLuint wrapMode, float s, GLuint size, else u = s * size; u -= 0.5F; - *i0 = IFLOOR(u); + *i0 = ifloor(u); *i1 = *i0 + 1; if (*i0 < 0) *i0 = 0; - if (*i1 >= (GLint) size) + if (*i1 >= (int) size) *i1 = size - 1; break; case PIPE_TEX_WRAP_CLAMP_TO_BORDER: @@ -271,23 +271,23 @@ linear_texcoord(GLuint wrapMode, float s, GLuint size, else u = s * size; u -= 0.5F; - *i0 = IFLOOR(u); + *i0 = ifloor(u); *i1 = *i0 + 1; } break; case PIPE_TEX_WRAP_MIRROR_REPEAT: { - const GLint flr = IFLOOR(s); + const int flr = ifloor(s); if (flr & 1) u = 1.0F - (s - (float) flr); else u = s - (float) flr; u = (u * size) - 0.5F; - *i0 = IFLOOR(u); + *i0 = ifloor(u); *i1 = *i0 + 1; if (*i0 < 0) *i0 = 0; - if (*i1 >= (GLint) size) + if (*i1 >= (int) size) *i1 = size - 1; } break; @@ -298,7 +298,7 @@ linear_texcoord(GLuint wrapMode, float s, GLuint size, else u *= size; u -= 0.5F; - *i0 = IFLOOR(u); + *i0 = ifloor(u); *i1 = *i0 + 1; break; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: @@ -308,11 +308,11 @@ linear_texcoord(GLuint wrapMode, float s, GLuint size, else u *= size; u -= 0.5F; - *i0 = IFLOOR(u); + *i0 = ifloor(u); *i1 = *i0 + 1; if (*i0 < 0) *i0 = 0; - if (*i1 >= (GLint) size) + if (*i1 >= (int) size) *i1 = size - 1; break; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: @@ -327,7 +327,7 @@ linear_texcoord(GLuint wrapMode, float s, GLuint size, else u *= size; u -= 0.5F; - *i0 = IFLOOR(u); + *i0 = ifloor(u); *i1 = *i0 + 1; } break; @@ -338,7 +338,7 @@ linear_texcoord(GLuint wrapMode, float s, GLuint size, } -static GLuint +static unsigned choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) { /* @@ -353,7 +353,7 @@ choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) -rz TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT -rx -ry rz */ const float arx = FABSF(rx), ary = FABSF(ry), arz = FABSF(rz); - GLuint face; + unsigned face; float sc, tc, ma; if (arx > ary && arx > arz) { @@ -779,16 +779,16 @@ sp_get_samples(struct tgsi_sampler *sampler, float rgba[NUM_CHANNELS][QUAD_SIZE]) { switch (sampler->texture->target) { - case GL_TEXTURE_1D: + case PIPE_TEXTURE_1D: sp_get_samples_1d(sampler, s, t, p, lodbias, rgba); break; - case GL_TEXTURE_2D: + case PIPE_TEXTURE_2D: sp_get_samples_2d(sampler, s, t, p, lodbias, rgba); break; - case GL_TEXTURE_3D: + case PIPE_TEXTURE_3D: sp_get_samples_3d(sampler, s, t, p, lodbias, rgba); break; - case GL_TEXTURE_CUBE_MAP: + case PIPE_TEXTURE_CUBE: sp_get_samples_cube(sampler, s, t, p, lodbias, rgba); break; default: diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.h b/src/mesa/pipe/softpipe/sp_tex_sample.h index 43aae87283..404bfd0c36 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.h +++ b/src/mesa/pipe/softpipe/sp_tex_sample.h @@ -7,11 +7,11 @@ struct tgsi_sampler; extern void sp_get_samples(struct tgsi_sampler *sampler, - const GLfloat s[QUAD_SIZE], - const GLfloat t[QUAD_SIZE], - const GLfloat p[QUAD_SIZE], + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], float lodbias, - GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]); + float rgba[NUM_CHANNELS][QUAD_SIZE]); #endif /* SP_TEX_SAMPLE_H */ diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.h b/src/mesa/pipe/tgsi/core/tgsi_build.h index 35c1a2506d..8b5db5662c 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_build.h +++ b/src/mesa/pipe/tgsi/core/tgsi_build.h @@ -24,7 +24,7 @@ tgsi_default_processor( void ); struct tgsi_processor tgsi_build_processor( - GLuint processor, + unsigned processor, struct tgsi_header *header ); /* @@ -36,31 +36,31 @@ tgsi_default_declaration( void ); struct tgsi_declaration tgsi_build_declaration( - GLuint file, - GLuint declare, - GLuint interpolate, + unsigned file, + unsigned declare, + unsigned interpolate, struct tgsi_header *header ); struct tgsi_full_declaration tgsi_default_full_declaration( void ); -GLuint +unsigned tgsi_build_full_declaration( const struct tgsi_full_declaration *full_decl, struct tgsi_token *tokens, struct tgsi_header *header, - GLuint maxsize ); + unsigned maxsize ); struct tgsi_declaration_range tgsi_build_declaration_range( - GLuint first, - GLuint last, + unsigned first, + unsigned last, struct tgsi_declaration *declaration, struct tgsi_header *header ); struct tgsi_declaration_mask tgsi_build_declaration_mask( - GLuint mask, + unsigned mask, struct tgsi_declaration *declaration, struct tgsi_header *header ); @@ -69,7 +69,7 @@ tgsi_default_declaration_interpolation( void ); struct tgsi_declaration_interpolation tgsi_build_declaration_interpolation( - GLuint interpolate, + unsigned interpolate, struct tgsi_declaration *declaration, struct tgsi_header *header ); @@ -89,16 +89,16 @@ tgsi_default_full_immediate( void ); struct tgsi_immediate_float32 tgsi_build_immediate_float32( - GLfloat value, + float value, struct tgsi_immediate *immediate, struct tgsi_header *header ); -GLuint +unsigned tgsi_build_full_immediate( const struct tgsi_full_immediate *full_imm, struct tgsi_token *tokens, struct tgsi_header *header, - GLuint maxsize ); + unsigned maxsize ); /* * instruction @@ -109,42 +109,42 @@ tgsi_default_instruction( void ); struct tgsi_instruction tgsi_build_instruction( - GLuint opcode, - GLuint saturate, - GLuint num_dst_regs, - GLuint num_src_regs, + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, struct tgsi_header *header ); struct tgsi_full_instruction tgsi_default_full_instruction( void ); -GLuint +unsigned tgsi_build_full_instruction( const struct tgsi_full_instruction *full_inst, struct tgsi_token *tokens, struct tgsi_header *header, - GLuint maxsize ); + unsigned maxsize ); struct tgsi_instruction_ext_nv tgsi_default_instruction_ext_nv( void ); -GLuint +unsigned tgsi_compare_instruction_ext_nv( struct tgsi_instruction_ext_nv a, struct tgsi_instruction_ext_nv b ); struct tgsi_instruction_ext_nv tgsi_build_instruction_ext_nv( - GLuint precision, - GLuint cond_dst_index, - GLuint cond_flow_index, - GLuint cond_mask, - GLuint cond_swizzle_x, - GLuint cond_swizzle_y, - GLuint cond_swizzle_z, - GLuint cond_swizzle_w, - GLuint cond_dst_update, - GLuint cond_flow_update, + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); @@ -152,14 +152,14 @@ tgsi_build_instruction_ext_nv( struct tgsi_instruction_ext_label tgsi_default_instruction_ext_label( void ); -GLuint +unsigned tgsi_compare_instruction_ext_label( struct tgsi_instruction_ext_label a, struct tgsi_instruction_ext_label b ); struct tgsi_instruction_ext_label tgsi_build_instruction_ext_label( - GLuint label, + unsigned label, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); @@ -167,14 +167,14 @@ tgsi_build_instruction_ext_label( struct tgsi_instruction_ext_texture tgsi_default_instruction_ext_texture( void ); -GLuint +unsigned tgsi_compare_instruction_ext_texture( struct tgsi_instruction_ext_texture a, struct tgsi_instruction_ext_texture b ); struct tgsi_instruction_ext_texture tgsi_build_instruction_ext_texture( - GLuint texture, + unsigned texture, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); @@ -184,15 +184,15 @@ tgsi_default_src_register( void ); struct tgsi_src_register tgsi_build_src_register( - GLuint file, - GLuint swizzle_x, - GLuint swizzle_y, - GLuint swizzle_z, - GLuint swizzle_w, - GLuint negate, - GLuint indirect, - GLuint dimension, - GLint index, + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, struct tgsi_instruction *instruction, struct tgsi_header *header ); @@ -202,22 +202,22 @@ tgsi_default_full_src_register( void ); struct tgsi_src_register_ext_swz tgsi_default_src_register_ext_swz( void ); -GLuint +unsigned tgsi_compare_src_register_ext_swz( struct tgsi_src_register_ext_swz a, struct tgsi_src_register_ext_swz b ); struct tgsi_src_register_ext_swz tgsi_build_src_register_ext_swz( - GLuint ext_swizzle_x, - GLuint ext_swizzle_y, - GLuint ext_swizzle_z, - GLuint ext_swizzle_w, - GLuint negate_x, - GLuint negate_y, - GLuint negate_z, - GLuint negate_w, - GLuint ext_divide, + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + unsigned ext_divide, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); @@ -225,18 +225,18 @@ tgsi_build_src_register_ext_swz( struct tgsi_src_register_ext_mod tgsi_default_src_register_ext_mod( void ); -GLuint +unsigned tgsi_compare_src_register_ext_mod( struct tgsi_src_register_ext_mod a, struct tgsi_src_register_ext_mod b ); struct tgsi_src_register_ext_mod tgsi_build_src_register_ext_mod( - GLuint complement, - GLuint bias, - GLuint scale_2x, - GLuint absolute, - GLuint negate, + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); @@ -246,8 +246,8 @@ tgsi_default_dimension( void ); struct tgsi_dimension tgsi_build_dimension( - GLuint indirect, - GLuint index, + unsigned indirect, + unsigned index, struct tgsi_instruction *instruction, struct tgsi_header *header ); @@ -256,9 +256,9 @@ tgsi_default_dst_register( void ); struct tgsi_dst_register tgsi_build_dst_register( - GLuint file, - GLuint mask, - GLint index, + unsigned file, + unsigned mask, + int index, struct tgsi_instruction *instruction, struct tgsi_header *header ); @@ -268,19 +268,19 @@ tgsi_default_full_dst_register( void ); struct tgsi_dst_register_ext_concode tgsi_default_dst_register_ext_concode( void ); -GLuint +unsigned tgsi_compare_dst_register_ext_concode( struct tgsi_dst_register_ext_concode a, struct tgsi_dst_register_ext_concode b ); struct tgsi_dst_register_ext_concode tgsi_build_dst_register_ext_concode( - GLuint cc, - GLuint swizzle_x, - GLuint swizzle_y, - GLuint swizzle_z, - GLuint swizzle_w, - GLint index, + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); @@ -288,14 +288,14 @@ tgsi_build_dst_register_ext_concode( struct tgsi_dst_register_ext_modulate tgsi_default_dst_register_ext_modulate( void ); -GLuint +unsigned tgsi_compare_dst_register_ext_modulate( struct tgsi_dst_register_ext_modulate a, struct tgsi_dst_register_ext_modulate b ); struct tgsi_dst_register_ext_modulate tgsi_build_dst_register_ext_modulate( - GLuint modulate, + unsigned modulate, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.h b/src/mesa/pipe/tgsi/core/tgsi_dump.h index 8553bdff59..70860c0885 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.h +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.h @@ -12,7 +12,7 @@ extern "C" { void tgsi_dump( const struct tgsi_token *tokens, - GLuint flags ); + unsigned flags ); #if defined __cplusplus } // extern "C" diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index 5e07e18a31..eed2207d7d 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -1,6 +1,8 @@ #if !defined TGSI_EXEC_H #define TGSI_EXEC_H +#include "pipe/p_compiler.h" + #if 0 #include "x86/rtasm/x86sse.h" #endif @@ -11,9 +13,9 @@ extern "C" { union tgsi_exec_channel { - GLfloat f[4]; - GLint i[4]; - GLuint u[4]; + float f[4]; + int i[4]; + unsigned u[4]; }; struct tgsi_exec_vector @@ -33,7 +35,7 @@ struct tgsi_exec_vector struct tgsi_texture_cache_entry { int x, y, face, level, zslice; - GLfloat data[TEX_CACHE_TILE_SIZE][TEX_CACHE_TILE_SIZE][4]; + float data[TEX_CACHE_TILE_SIZE][TEX_CACHE_TILE_SIZE][4]; }; struct tgsi_sampler @@ -42,19 +44,19 @@ struct tgsi_sampler struct pipe_mipmap_tree *texture; /** Get samples for four fragments in a quad */ void (*get_samples)(struct tgsi_sampler *sampler, - const GLfloat s[QUAD_SIZE], - const GLfloat t[QUAD_SIZE], - const GLfloat p[QUAD_SIZE], - GLfloat lodbias, - GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]); + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); void *pipe; /*XXX temporary*/ struct tgsi_texture_cache_entry cache[TEX_CACHE_NUM_ENTRIES]; }; struct tgsi_exec_labels { - GLuint labels[128][2]; - GLuint count; + unsigned labels[128][2]; + unsigned count; }; #define TGSI_EXEC_TEMP_00000000_I 32 @@ -107,15 +109,15 @@ struct tgsi_exec_cond_state { struct tgsi_exec_cond_regs IfPortion; struct tgsi_exec_cond_regs ElsePortion; - GLuint Condition; - GLboolean WasElse; + unsigned Condition; + boolean WasElse; }; /* XXX: This is temporary */ struct tgsi_exec_cond_stack { struct tgsi_exec_cond_state States[8]; - GLuint Index; /* into States[] */ + unsigned Index; /* into States[] */ }; struct tgsi_exec_machine @@ -136,15 +138,15 @@ struct tgsi_exec_machine struct tgsi_sampler *Samplers; - GLfloat Imms[256][4]; - GLuint ImmLimit; - GLfloat (*Consts)[4]; + float Imms[256][4]; + unsigned ImmLimit; + float (*Consts)[4]; const struct tgsi_exec_vector *Inputs; struct tgsi_exec_vector *Outputs; struct tgsi_token *Tokens; - GLuint Processor; + unsigned Processor; - GLuint *Primitives; + unsigned *Primitives; struct tgsi_exec_cond_stack CondStack; #if XXX_SSE @@ -156,7 +158,7 @@ void tgsi_exec_machine_init( struct tgsi_exec_machine *mach, struct tgsi_token *tokens, - GLuint numSamplers, + unsigned numSamplers, struct tgsi_sampler *samplers); void diff --git a/src/mesa/pipe/tgsi/core/tgsi_parse.h b/src/mesa/pipe/tgsi/core/tgsi_parse.h index 61ad0669b1..bba01431fa 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_parse.h +++ b/src/mesa/pipe/tgsi/core/tgsi_parse.h @@ -86,7 +86,7 @@ tgsi_full_token_free( struct tgsi_parse_context { const struct tgsi_token *Tokens; - GLuint Position; + unsigned Position; struct tgsi_full_version FullVersion; struct tgsi_full_header FullHeader; union tgsi_full_token FullToken; @@ -95,7 +95,7 @@ struct tgsi_parse_context #define TGSI_PARSE_OK 0 #define TGSI_PARSE_ERROR 1 -GLuint +unsigned tgsi_parse_init( struct tgsi_parse_context *ctx, const struct tgsi_token *tokens ); @@ -104,7 +104,7 @@ void tgsi_parse_free( struct tgsi_parse_context *ctx ); -GLuint +unsigned tgsi_parse_end_of_tokens( struct tgsi_parse_context *ctx ); diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index ec62836ef3..dc9301ed37 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -7,15 +7,15 @@ extern "C" { struct tgsi_version { - GLuint MajorVersion : 8; - GLuint MinorVersion : 8; - GLuint Padding : 16; + unsigned MajorVersion : 8; + unsigned MinorVersion : 8; + unsigned Padding : 16; }; struct tgsi_header { - GLuint HeaderSize : 8; - GLuint BodySize : 24; + unsigned HeaderSize : 8; + unsigned BodySize : 24; }; #define TGSI_PROCESSOR_FRAGMENT 0 @@ -24,8 +24,8 @@ struct tgsi_header struct tgsi_processor { - GLuint Processor : 4; /* TGSI_PROCESSOR_ */ - GLuint Padding : 28; + unsigned Processor : 4; /* TGSI_PROCESSOR_ */ + unsigned Padding : 28; }; #define TGSI_TOKEN_TYPE_DECLARATION 0 @@ -34,10 +34,10 @@ struct tgsi_processor struct tgsi_token { - GLuint Type : 4; /* TGSI_TOKEN_TYPE_ */ - GLuint Size : 8; /* UINT */ - GLuint Padding : 19; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_ */ + unsigned Size : 8; /* UINT */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ }; #define TGSI_FILE_NULL 0 @@ -54,24 +54,24 @@ struct tgsi_token struct tgsi_declaration { - GLuint Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ - GLuint Size : 8; /* UINT */ - GLuint File : 4; /* TGSI_FILE_ */ - GLuint Declare : 4; /* TGSI_DECLARE_ */ - GLuint Interpolate : 1; /* BOOL */ - GLuint Padding : 10; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ + unsigned Size : 8; /* UINT */ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned Declare : 4; /* TGSI_DECLARE_ */ + unsigned Interpolate : 1; /* BOOL */ + unsigned Padding : 10; + unsigned Extended : 1; /* BOOL */ }; struct tgsi_declaration_range { - GLuint First : 16; /* UINT */ - GLuint Last : 16; /* UINT */ + unsigned First : 16; /* UINT */ + unsigned Last : 16; /* UINT */ }; struct tgsi_declaration_mask { - GLuint Mask : 32; /* UINT */ + unsigned Mask : 32; /* UINT */ }; #define TGSI_INTERPOLATE_CONSTANT 0 @@ -80,24 +80,24 @@ struct tgsi_declaration_mask struct tgsi_declaration_interpolation { - GLuint Interpolate : 4; /* TGSI_INTERPOLATE_ */ - GLuint Padding : 28; + unsigned Interpolate : 4; /* TGSI_INTERPOLATE_ */ + unsigned Padding : 28; }; #define TGSI_IMM_FLOAT32 0 struct tgsi_immediate { - GLuint Type : 4; /* TGSI_TOKEN_TYPE_IMMEDIATE */ - GLuint Size : 8; /* UINT */ - GLuint DataType : 4; /* TGSI_IMM_ */ - GLuint Padding : 15; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_IMMEDIATE */ + unsigned Size : 8; /* UINT */ + unsigned DataType : 4; /* TGSI_IMM_ */ + unsigned Padding : 15; + unsigned Extended : 1; /* BOOL */ }; struct tgsi_immediate_float32 { - GLfloat Float; + float Float; }; /* @@ -1088,14 +1088,14 @@ struct tgsi_immediate_float32 struct tgsi_instruction { - GLuint Type : 4; /* TGSI_TOKEN_TYPE_INSTRUCTION */ - GLuint Size : 8; /* UINT */ - GLuint Opcode : 8; /* TGSI_OPCODE_ */ - GLuint Saturate : 2; /* TGSI_SAT_ */ - GLuint NumDstRegs : 2; /* UINT */ - GLuint NumSrcRegs : 4; /* UINT */ - GLuint Padding : 3; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_INSTRUCTION */ + unsigned Size : 8; /* UINT */ + unsigned Opcode : 8; /* TGSI_OPCODE_ */ + unsigned Saturate : 2; /* TGSI_SAT_ */ + unsigned NumDstRegs : 2; /* UINT */ + unsigned NumSrcRegs : 4; /* UINT */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ }; /* @@ -1116,9 +1116,9 @@ struct tgsi_instruction struct tgsi_instruction_ext { - GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_ */ - GLuint Padding : 27; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ }; /* @@ -1174,27 +1174,27 @@ struct tgsi_instruction_ext struct tgsi_instruction_ext_nv { - GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_NV */ - GLuint Precision : 4; /* TGSI_PRECISION_ */ - GLuint CondDstIndex : 4; /* UINT */ - GLuint CondFlowIndex : 4; /* UINT */ - GLuint CondMask : 4; /* TGSI_CC_ */ - GLuint CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ - GLuint CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ - GLuint CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - GLuint CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ - GLuint CondDstUpdate : 1; /* BOOL */ - GLuint CondFlowEnable : 1; /* BOOL */ - GLuint Padding : 1; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_NV */ + unsigned Precision : 4; /* TGSI_PRECISION_ */ + unsigned CondDstIndex : 4; /* UINT */ + unsigned CondFlowIndex : 4; /* UINT */ + unsigned CondMask : 4; /* TGSI_CC_ */ + unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned CondDstUpdate : 1; /* BOOL */ + unsigned CondFlowEnable : 1; /* BOOL */ + unsigned Padding : 1; + unsigned Extended : 1; /* BOOL */ }; struct tgsi_instruction_ext_label { - GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ - GLuint Label : 24; /* UINT */ - GLuint Padding : 3; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ + unsigned Label : 24; /* UINT */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ }; #define TGSI_TEXTURE_UNKNOWN 0 @@ -1209,10 +1209,10 @@ struct tgsi_instruction_ext_label struct tgsi_instruction_ext_texture { - GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */ - GLuint Texture : 8; /* TGSI_TEXTURE_ */ - GLuint Padding : 19; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */ + unsigned Texture : 8; /* TGSI_TEXTURE_ */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ }; #define TGSI_WRITEMASK_NONE 0x00 @@ -1234,11 +1234,11 @@ struct tgsi_instruction_ext_texture struct tgsi_instruction_ext_predicate { - GLuint Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ - GLuint PredDstIndex : 4; /* UINT */ - GLuint PredWriteMask : 4; /* TGSI_WRITEMASK_ */ - GLuint Padding : 19; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ + unsigned PredDstIndex : 4; /* UINT */ + unsigned PredWriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ }; /* @@ -1261,16 +1261,16 @@ struct tgsi_instruction_ext_predicate struct tgsi_src_register { - GLuint File : 4; /* TGSI_FILE_ */ - GLuint SwizzleX : 2; /* TGSI_SWIZZLE_ */ - GLuint SwizzleY : 2; /* TGSI_SWIZZLE_ */ - GLuint SwizzleZ : 2; /* TGSI_SWIZZLE_ */ - GLuint SwizzleW : 2; /* TGSI_SWIZZLE_ */ - GLuint Negate : 1; /* BOOL */ - GLuint Indirect : 1; /* BOOL */ - GLuint Dimension : 1; /* BOOL */ - GLint Index : 16; /* SINT */ - GLuint Extended : 1; /* BOOL */ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned Negate : 1; /* BOOL */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ }; /* @@ -1287,9 +1287,9 @@ struct tgsi_src_register struct tgsi_src_register_ext { - GLuint Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_ */ - GLuint Padding : 27; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ }; /* @@ -1323,18 +1323,18 @@ struct tgsi_src_register_ext struct tgsi_src_register_ext_swz { - GLuint Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */ - GLuint ExtSwizzleX : 4; /* TGSI_EXTSWIZZLE_ */ - GLuint ExtSwizzleY : 4; /* TGSI_EXTSWIZZLE_ */ - GLuint ExtSwizzleZ : 4; /* TGSI_EXTSWIZZLE_ */ - GLuint ExtSwizzleW : 4; /* TGSI_EXTSWIZZLE_ */ - GLuint NegateX : 1; /* BOOL */ - GLuint NegateY : 1; /* BOOL */ - GLuint NegateZ : 1; /* BOOL */ - GLuint NegateW : 1; /* BOOL */ - GLuint ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ - GLuint Padding : 3; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */ + unsigned ExtSwizzleX : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleY : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleZ : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleW : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned NegateX : 1; /* BOOL */ + unsigned NegateY : 1; /* BOOL */ + unsigned NegateZ : 1; /* BOOL */ + unsigned NegateW : 1; /* BOOL */ + unsigned ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ }; /* @@ -1352,34 +1352,34 @@ struct tgsi_src_register_ext_swz struct tgsi_src_register_ext_mod { - GLuint Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */ - GLuint Complement : 1; /* BOOL */ - GLuint Bias : 1; /* BOOL */ - GLuint Scale2X : 1; /* BOOL */ - GLuint Absolute : 1; /* BOOL */ - GLuint Negate : 1; /* BOOL */ - GLuint Padding : 22; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */ + unsigned Complement : 1; /* BOOL */ + unsigned Bias : 1; /* BOOL */ + unsigned Scale2X : 1; /* BOOL */ + unsigned Absolute : 1; /* BOOL */ + unsigned Negate : 1; /* BOOL */ + unsigned Padding : 22; + unsigned Extended : 1; /* BOOL */ }; struct tgsi_dimension { - GLuint Indirect : 1; /* BOOL */ - GLuint Dimension : 1; /* BOOL */ - GLuint Padding : 13; - GLint Index : 16; /* SINT */ - GLuint Extended : 1; /* BOOL */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + unsigned Padding : 13; + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ }; struct tgsi_dst_register { - GLuint File : 4; /* TGSI_FILE_ */ - GLuint WriteMask : 4; /* TGSI_WRITEMASK_ */ - GLuint Indirect : 1; /* BOOL */ - GLuint Dimension : 1; /* BOOL */ - GLint Index : 16; /* SINT */ - GLuint Padding : 5; - GLuint Extended : 1; /* BOOL */ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Padding : 5; + unsigned Extended : 1; /* BOOL */ }; /* @@ -1394,9 +1394,9 @@ struct tgsi_dst_register struct tgsi_dst_register_ext { - GLuint Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_ */ - GLuint Padding : 27; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ }; /* @@ -1415,15 +1415,15 @@ struct tgsi_dst_register_ext struct tgsi_dst_register_ext_concode { - GLuint Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ - GLuint CondMask : 4; /* TGSI_CC_ */ - GLuint CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ - GLuint CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ - GLuint CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - GLuint CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ - GLuint CondSrcIndex : 4; /* UINT */ - GLuint Padding : 11; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ + unsigned CondMask : 4; /* TGSI_CC_ */ + unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSrcIndex : 4; /* UINT */ + unsigned Padding : 11; + unsigned Extended : 1; /* BOOL */ }; #define TGSI_MODULATE_1X 0 @@ -1436,10 +1436,10 @@ struct tgsi_dst_register_ext_concode struct tgsi_dst_register_ext_modulate { - GLuint Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */ - GLuint Modulate : 4; /* TGSI_MODULATE_ */ - GLuint Padding : 23; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */ + unsigned Modulate : 4; /* TGSI_MODULATE_ */ + unsigned Padding : 23; + unsigned Extended : 1; /* BOOL */ }; /* @@ -1451,15 +1451,15 @@ struct tgsi_dst_register_ext_modulate struct tgsi_dst_register_ext_predicate { - GLuint Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */ - GLuint PredSwizzleX : 2; /* TGSI_SWIZZLE_ */ - GLuint PredSwizzleY : 2; /* TGSI_SWIZZLE_ */ - GLuint PredSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - GLuint PredSwizzleW : 2; /* TGSI_SWIZZLE_ */ - GLuint PredSrcIndex : 4; /* UINT */ - GLuint Negate : 1; /* BOOL */ - GLuint Padding : 14; - GLuint Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */ + unsigned PredSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSrcIndex : 4; /* UINT */ + unsigned Negate : 1; /* BOOL */ + unsigned Padding : 14; + unsigned Extended : 1; /* BOOL */ }; #if defined __cplusplus diff --git a/src/mesa/pipe/tgsi/core/tgsi_util.h b/src/mesa/pipe/tgsi/core/tgsi_util.h index 70c48690c5..ef14446f0e 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_util.h +++ b/src/mesa/pipe/tgsi/core/tgsi_util.h @@ -9,58 +9,58 @@ void * tgsi_align_128bit( void *unaligned ); -GLuint +unsigned tgsi_util_get_src_register_swizzle( const struct tgsi_src_register *reg, - GLuint component ); + unsigned component ); -GLuint +unsigned tgsi_util_get_src_register_extswizzle( const struct tgsi_src_register_ext_swz *reg, - GLuint component); + unsigned component); -GLuint +unsigned tgsi_util_get_full_src_register_extswizzle( const struct tgsi_full_src_register *reg, - GLuint component ); + unsigned component ); void tgsi_util_set_src_register_swizzle( struct tgsi_src_register *reg, - GLuint swizzle, - GLuint component ); + unsigned swizzle, + unsigned component ); void tgsi_util_set_src_register_extswizzle( struct tgsi_src_register_ext_swz *reg, - GLuint swizzle, - GLuint component ); + unsigned swizzle, + unsigned component ); -GLuint +unsigned tgsi_util_get_src_register_extnegate( const struct tgsi_src_register_ext_swz *reg, - GLuint component ); + unsigned component ); void tgsi_util_set_src_register_extnegate( struct tgsi_src_register_ext_swz *reg, - GLuint negate, - GLuint component ); + unsigned negate, + unsigned component ); #define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ #define TGSI_UTIL_SIGN_SET 1 /* Force negative */ #define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ #define TGSI_UTIL_SIGN_KEEP 3 /* No change */ -GLuint +unsigned tgsi_util_get_full_src_register_sign_mode( const struct tgsi_full_src_register *reg, - GLuint component ); + unsigned component ); void tgsi_util_set_full_src_register_sign_mode( struct tgsi_full_src_register *reg, - GLuint sign_mode ); + unsigned sign_mode ); #if defined __cplusplus } // extern "C" -- cgit v1.2.3 From 2691b187473ddb9c96541cd154dde19c35cbaabe Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 14 Aug 2007 15:41:26 +0100 Subject: Add surface formats to be used for specifying vertex element layouts. --- src/mesa/pipe/p_defines.h | 71 +++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 69 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index e6aff6fb49..addbb026bf 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -162,8 +162,15 @@ /** * Texture/surface image formats (preliminary) */ + +/* KW: Added lots of surface formats to support vertex element layout + * definitions, and eventually render-to-vertex-buffer. Could + * consider making float/int/uint/scaled/normalized a separate + * parameter, but on the other hand there are special cases like + * z24s8, compressed textures, ycbcr, etc that won't fit that model. + */ + #define PIPE_FORMAT_NONE 0 /**< unstructured */ -#define PIPE_FORMAT_U_R8_G8_B8_A8 1 /**< ubyte[4] RGBA */ #define PIPE_FORMAT_U_A8_R8_G8_B8 2 /**< ubyte[4] ARGB */ #define PIPE_FORMAT_U_A1_R5_G5_B5 3 /**< 16-bit packed RGBA */ #define PIPE_FORMAT_U_A4_R4_G4_B4 4 /**< 16-bit packed RGBA */ @@ -180,8 +187,68 @@ #define PIPE_FORMAT_F_Z32 15 /**< float Z/depth */ #define PIPE_FORMAT_S8_Z24 16 /**< 8-bit stencil + 24-bit Z */ #define PIPE_FORMAT_U_S8 17 /**< 8-bit stencil */ -#define PIPE_FORMAT_COUNT 18 /**< number of formats */ +#define PIPE_FORMAT_R64_FLOAT 0x20 +#define PIPE_FORMAT_R64G64_FLOAT 0x21 +#define PIPE_FORMAT_R64G64B64_FLOAT 0x22 +#define PIPE_FORMAT_R64G64B64A64_FLOAT 0x23 +#define PIPE_FORMAT_R32_FLOAT 0x24 +#define PIPE_FORMAT_R32G32_FLOAT 0x25 +#define PIPE_FORMAT_R32G32B32_FLOAT 0x26 +#define PIPE_FORMAT_R32G32B32A32_FLOAT 0x27 +#define PIPE_FORMAT_R32_UNORM 0x28 +#define PIPE_FORMAT_R32G32_UNORM 0x29 +#define PIPE_FORMAT_R32G32B32_UNORM 0x2a +#define PIPE_FORMAT_R32G32B32A32_UNORM 0x2b +#define PIPE_FORMAT_R32_USCALED 0x2c +#define PIPE_FORMAT_R32G32_USCALED 0x2d +#define PIPE_FORMAT_R32G32B32_USCALED 0x2e +#define PIPE_FORMAT_R32G32B32A32_USCALED 0x2f +#define PIPE_FORMAT_R32_SNORM 0x30 +#define PIPE_FORMAT_R32G32_SNORM 0x31 +#define PIPE_FORMAT_R32G32B32_SNORM 0x32 +#define PIPE_FORMAT_R32G32B32A32_SNORM 0x33 +#define PIPE_FORMAT_R32_SSCALED 0x34 +#define PIPE_FORMAT_R32G32_SSCALED 0x35 +#define PIPE_FORMAT_R32G32B32_SSCALED 0x36 +#define PIPE_FORMAT_R32G32B32A32_SSCALED 0x37 +#define PIPE_FORMAT_R16_UNORM 0x38 +#define PIPE_FORMAT_R16G16_UNORM 0x39 +#define PIPE_FORMAT_R16G16B16_UNORM 0x3a +#define PIPE_FORMAT_R16G16B16A16_UNORM 0x3b +#define PIPE_FORMAT_R16_USCALED 0x3c +#define PIPE_FORMAT_R16G16_USCALED 0x3d +#define PIPE_FORMAT_R16G16B16_USCALED 0x3e +#define PIPE_FORMAT_R16G16B16A16_USCALED 0x3f +#define PIPE_FORMAT_R16_SNORM 0x40 +#define PIPE_FORMAT_R16G16_SNORM 0x41 +#define PIPE_FORMAT_R16G16B16_SNORM 0x42 +#define PIPE_FORMAT_R16G16B16A16_SNORM 0x43 +#define PIPE_FORMAT_R16_SSCALED 0x44 +#define PIPE_FORMAT_R16G16_SSCALED 0x45 +#define PIPE_FORMAT_R16G16B16_SSCALED 0x46 +#define PIPE_FORMAT_R16G16B16A16_SSCALED 0x47 +#define PIPE_FORMAT_R8_UNORM 0x48 +#define PIPE_FORMAT_R8G8_UNORM 0x49 +#define PIPE_FORMAT_R8G8B8_UNORM 0x4a +#define PIPE_FORMAT_R8G8B8A8_UNORM 0x4b +#define PIPE_FORMAT_R8_USCALED 0x4c +#define PIPE_FORMAT_R8G8_USCALED 0x4d +#define PIPE_FORMAT_R8G8B8_USCALED 0x4e +#define PIPE_FORMAT_R8G8B8A8_USCALED 0x4f +#define PIPE_FORMAT_R8_SNORM 0x50 +#define PIPE_FORMAT_R8G8_SNORM 0x51 +#define PIPE_FORMAT_R8G8B8_SNORM 0x52 +#define PIPE_FORMAT_R8G8B8A8_SNORM 0x53 +#define PIPE_FORMAT_R8_SSCALED 0x54 +#define PIPE_FORMAT_R8G8_SSCALED 0x55 +#define PIPE_FORMAT_R8G8B8_SSCALED 0x56 +#define PIPE_FORMAT_R8G8B8A8_SSCALED 0x57 +#define PIPE_FORMAT_COUNT 0x58 /**< number of formats */ + +/* Duplicated formats: + */ +#define PIPE_FORMAT_U_R8_G8_B8_A8 PIPE_FORMAT_R8G8B8A8_UNORM /** * Surface flags -- cgit v1.2.3 From 8269bc48d8fafaa432b58f4adf5e0dddd81d979d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 14 Aug 2007 15:42:47 +0100 Subject: Add structs and set-functions for vertex buffer, element state. Not currently used. --- src/mesa/pipe/p_context.h | 10 +++++++++- src/mesa/pipe/p_state.h | 21 +++++++++++++++++++++ 2 files changed, 30 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 7635755947..c0685fa8cb 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -136,7 +136,15 @@ struct pipe_context { void (*set_viewport_state)( struct pipe_context *, const struct pipe_viewport_state * ); - + void (*set_vertex_buffer)( struct pipe_context *, + unsigned index, + struct pipe_vertex_buffer * ); + + void (*set_vertex_element)( struct pipe_context *, + unsigned index, + struct pipe_vertex_element * ); + + /* * Surface functions * This might go away... diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index dc2b589e55..5051ebbfc0 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -320,4 +320,25 @@ struct pipe_mipmap_tree }; + +struct pipe_vertex_buffer +{ + unsigned pitch:11; + unsigned max_index; + struct pipe_buffer_handle *buffer; + unsigned buffer_offset; +}; + + + +struct pipe_vertex_element +{ + unsigned src_offset:11; + unsigned vertex_buffer_index:5; + unsigned dst_offset:8; + unsigned src_format:8; /* PIPE_FORMAT_* */ +}; + + + #endif -- cgit v1.2.3 From 4bb213423941fb12801a734ad2d952a6d8f2347e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 14 Aug 2007 15:44:41 +0100 Subject: Beginnings of a demand-filled post-tnl vertex cache. Probably breaks a bit of stuff, eg unfilled clipping, edgeflags, etc. --- src/mesa/pipe/draw/draw_context.c | 11 + src/mesa/pipe/draw/draw_private.h | 50 +++ src/mesa/pipe/draw/draw_unfilled.c | 12 +- src/mesa/pipe/draw/draw_vb.c | 621 +++++++++++++++++++------------------ 4 files changed, 392 insertions(+), 302 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index a97f488387..4335b47e09 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -59,6 +59,16 @@ struct draw_context *draw_create( void ) draw->vf = vf_create( GL_TRUE ); + /* Statically allocate maximum sized vertices for the cache - could be cleverer... + */ + { + int i; + char *tmp = malloc(Elements(draw->vcache.vertex) * MAX_VERTEX_SIZE); + + for (i = 0; i < Elements(draw->vcache.vertex); i++) + draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * MAX_VERTEX_SIZE); + } + return draw; } @@ -70,6 +80,7 @@ void draw_destroy( struct draw_context *draw ) vf_destroy( draw->vf ); + FREE( draw->vcache.vertex[0] ); /* Frees all the vertices. */ FREE( draw ); } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 3dfaa0581d..597393afdc 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -61,12 +61,18 @@ struct vertex_header { GLfloat data[][4]; /* Note variable size */ }; +#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) + + /** * Basic info for a point/line/triangle primitive. */ struct prim_header { GLfloat det; /**< front/back face determinant */ + GLuint reset_line_stipple:1; + GLuint edgeflags:3; + GLuint pad:28; struct vertex_header *v[3]; /**< 1 to 3 vertex pointers */ }; @@ -103,6 +109,12 @@ struct draw_stage }; +#define PRIM_QUEUE_LENGTH 16 +#define VCACHE_SIZE 32 +#define VCACHE_OVERFLOW 4 +#define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ + + /** * Private context for the drawing module. */ @@ -141,7 +153,45 @@ struct draw_context GLuint nr_vertices; GLboolean in_vb; + void *elts; + + struct vertex_header *(*get_vertex)( struct draw_context *draw, + GLuint i ); + + /* Post-tnl vertex cache: + */ + struct { + GLuint referenced; + GLuint idx[VCACHE_SIZE + VCACHE_OVERFLOW]; + struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; + GLuint overflow; + } vcache; + + /* Vertex shader queue: + */ + struct { + struct { + GLuint elt; + struct vertex_header *dest; + } queue[VS_QUEUE_LENGTH]; + GLuint queue_nr; + } vs; + + /* Prim pipeline queue: + */ + struct { + + /* Need to queue up primitives until their vertices have been + * transformed by a vs queue flush. + */ + struct prim_header queue[PRIM_QUEUE_LENGTH]; + GLuint queue_nr; + } pq; + + + GLenum prim; /**< GL_POINTS, GL_LINE_STRIP, GL_QUADS, etc */ + unsigned reduced_prim; /* Helper for tnl: */ diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index 05242d8d10..e0e486ebe6 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -91,9 +91,9 @@ static void points( struct draw_stage *stage, struct vertex_header *v1 = header->v[1]; struct vertex_header *v2 = header->v[2]; - if (v0->edgeflag) point( stage, v0 ); - if (v1->edgeflag) point( stage, v1 ); - if (v2->edgeflag) point( stage, v2 ); + if (header->edgeflags & 0x1) point( stage, v0 ); + if (header->edgeflags & 0x2) point( stage, v1 ); + if (header->edgeflags & 0x4) point( stage, v2 ); } @@ -104,9 +104,9 @@ static void lines( struct draw_stage *stage, struct vertex_header *v1 = header->v[1]; struct vertex_header *v2 = header->v[2]; - if (v0->edgeflag) line( stage, v0, v1 ); - if (v1->edgeflag) line( stage, v1, v2 ); - if (v2->edgeflag) line( stage, v2, v0 ); + if (header->edgeflags & 0x1) line( stage, v0, v1 ); + if (header->edgeflags & 0x2) line( stage, v1, v2 ); + if (header->edgeflags & 0x4) line( stage, v2, v0 ); } diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index f9c10e5f97..ef32f02ec1 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -40,356 +40,377 @@ #include "draw_context.h" -/* This file is a temporary set of hooks to allow us to use the tnl/ - * and vf/ modules until we have replacements in pipe. - */ +#define RP_NONE 0 +#define RP_POINT 1 +#define RP_LINE 2 +#define RP_TRI 3 + +static unsigned reduced_prim[GL_POLYGON + 1] = { + RP_POINT, + RP_LINE, + RP_LINE, + RP_LINE, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI +}; -static struct vertex_header *get_vertex( struct draw_context *pipe, - GLuint i ) -{ - return (struct vertex_header *)(pipe->verts + i * pipe->vertex_size); -} +/* This file is a temporary set of hooks to allow us to use the tnl/ + * and vf/ modules until we have replacements in pipe. + */ -static void draw_allocate_vertices( struct draw_context *draw, - GLuint nr_vertices ) +static void vs_flush( struct draw_context *draw ) { - draw->nr_vertices = nr_vertices; - draw->verts = (GLubyte *) malloc( nr_vertices * draw->vertex_size ); + unsigned i; - draw->pipeline.first->begin( draw->pipeline.first ); + /* We're not really running a vertex shader yet, so flushing the vs + * queue is just a matter of building the vertices and returning. + */ + /* Actually, I'm cheating even more and pre-building them still + * with the mesa/vf module. So it's very easy... + */ + for (i = 0; i < draw->vs.queue_nr; i++) { + /* Would do the following steps here: + * + * 1) Loop over vertex element descriptors, fetch data from each + * to build the pre-tnl vertex. This might require a new struct + * to represent the pre-tnl vertex. + * + * 2) Bundle groups of upto 4 pre-tnl vertices together and pass + * to vertex shader. + * + * 3) Do any necessary unswizzling, make sure vertex headers are + * correctly populated, store resulting post-transformed + * vertices in vcache. + * + * In this version, just do the last step: + */ + unsigned elt = draw->vs.queue[i].elt; + struct vertex_header *dest = draw->vs.queue[i].dest; + + /* Magic: + */ + memcpy(dest, + draw->verts + elt * draw->vertex_size, + draw->vertex_size); + + } + draw->vs.queue_nr = 0; } -static void draw_set_prim( struct draw_context *draw, - GLenum prim ) + +static void draw_flush( struct draw_context *draw ) { - draw->prim = prim; + struct draw_stage *first = draw->pipeline.first; + unsigned i; - /* Not done yet - need to force edgeflags to 1 in strip/fan - * primitives. + /* Make sure all vertices are available: */ -#if 0 - switch (prim) { - case GL_TRIANGLES: - case GL_POLYGON: - case GL_QUADS: - case GL_QUAD_STRIP: /* yes, we need this */ - respect_edgeflags( pipe, GL_TRUE ); + vs_flush( draw ); + + + switch (draw->reduced_prim) { + case RP_TRI: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + first->reset_stipple_counter( first ); + + first->tri( first, &draw->pq.queue[i] ); + } break; + case RP_LINE: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + first->reset_stipple_counter( first ); - default: - respect_edgeflags( pipe, GL_FALSE ); + first->line( first, &draw->pq.queue[i] ); + } + break; + case RP_POINT: + first->reset_stipple_counter( first ); + for (i = 0; i < draw->pq.queue_nr; i++) + first->point( first, &draw->pq.queue[i] ); break; } -#endif -} - + draw->pq.queue_nr = 0; + draw->vcache.referenced = 0; + draw->vcache.overflow = 0; +} -static void do_quad( struct draw_stage *first, - struct vertex_header *v0, - struct vertex_header *v1, - struct vertex_header *v2, - struct vertex_header *v3 ) +static void draw_invalidate_vcache( struct draw_context *draw ) { - struct prim_header prim; - - { - GLuint tmp = v1->edgeflag; - v1->edgeflag = 0; - - prim.v[0] = v0; - prim.v[1] = v1; - prim.v[2] = v3; - first->tri( first, &prim ); + unsigned i; + + assert(draw->pq.queue_nr == 0); + assert(draw->vs.queue_nr == 0); + assert(draw->vcache.referenced == 0); + + for (i = 0; i < Elements( draw->vcache.idx ); i++) + draw->vcache.idx[i] = ~0; +} - v1->edgeflag = tmp; - } - { - GLuint tmp = v3->edgeflag; - v3->edgeflag = 0; +/* Return a pointer to a freshly queued primitive header. Ensure that + * there is room in the vertex cache for a maximum of "nr_verts" new + * vertices. Flush primitive and/or vertex queues if necessary to + * make space. + */ +static struct prim_header *get_queued_prim( struct draw_context *draw, + GLuint nr_verts ) +{ + if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH || + draw->vcache.overflow + nr_verts >= VCACHE_OVERFLOW) + draw_flush( draw ); - prim.v[0] = v1; - prim.v[1] = v2; - prim.v[2] = v3; - first->tri( first, &prim ); + /* The vs queue is sized so that this can never happen: + */ + assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); - v3->edgeflag = tmp; - } + return &draw->pq.queue[draw->pq.queue_nr++]; } - - -static void draw_indexed_prim( struct draw_context *draw, - const GLuint *elts, - GLuint count ) +/* Check if vertex is in cache, otherwise add it. It won't go through + * VS yet, not until there is a flush operation or the VS queue fills up. + */ +static struct vertex_header *get_vertex( struct draw_context *draw, + GLuint i ) { - struct draw_stage * const first = draw->pipeline.first; - struct prim_header prim; - GLuint i; + unsigned slot = (i + (i>>5)) & 31; + + /* Cache miss? + */ + if (draw->vcache.idx[slot] != i) { - prim.det = 0; /* valid from cull stage onwards */ - prim.v[0] = 0; - prim.v[1] = 0; - prim.v[2] = 0; + /* If slot is in use, use the overflow area: + */ + if (draw->vcache.referenced & (1<vcache.overflow++; - switch (draw->prim) { - case GL_POINTS: - for (i = 0; i < count; i ++) { - prim.v[0] = get_vertex( draw, elts[i] ); + draw->vcache.idx[slot] = i; - first->point( first, &prim ); - } - break; + /* Add to vertex shader queue: + */ + draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; + draw->vs.queue[draw->vs.queue_nr].elt = i; + draw->vs.queue_nr++; + } - case GL_LINES: - for (i = 0; i+1 < count; i += 2) { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - - first->reset_stipple_counter( first ); - first->line( first, &prim ); - } - break; + /* Mark slot as in-use: + */ + draw->vcache.referenced |= (1<vcache.vertex[slot]; +} - case GL_LINE_LOOP: - if (count >= 2) { - first->reset_stipple_counter( first ); - for (i = 1; i < count; i++) { - prim.v[0] = get_vertex( draw, elts[i-1] ); - prim.v[1] = get_vertex( draw, elts[i] ); - first->line( first, &prim ); - } - prim.v[0] = get_vertex( draw, elts[count-1] ); - prim.v[1] = get_vertex( draw, elts[0] ); - first->line( first, &prim ); - } - break; +static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, + GLuint i ) +{ + const GLuint *elts = (const GLuint *)draw->elts; + return get_vertex( draw, elts[i] ); +} - case GL_LINE_STRIP: - /* I'm guessing it will be necessary to have something like a - * render->reset_line_stipple() method to properly support - * splitting strips into primitives like this. Alternately we - * could just scan ahead to find individual clipped lines and - * otherwise leave the strip intact - that might be better, but - * require more complex code here. - */ - if (count >= 2) { - first->reset_stipple_counter( first ); - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, elts[0] ); - - for (i = 1; i < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, elts[i] ); - - first->line( first, &prim ); - } - } - break; +#if 0 +static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, + const void *elts, + GLuint i ) +{ + const GLushort *elts = (const GLushort *)draw->elts; + return get_vertex( draw, elts[i] ); +} - case GL_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - - first->tri( first, &prim ); - } - break; +static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, + const void *elts, + GLuint i ) +{ + const GLubyte *elts = (const GLubyte *)draw->elts; + return get_vertex( draw, elts[i] ); +} +#endif - case GL_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - prim.v[0] = get_vertex( draw, elts[i + 1] ); - prim.v[1] = get_vertex( draw, elts[i + 0] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - } - else { - prim.v[0] = get_vertex( draw, elts[i + 0] ); - prim.v[1] = get_vertex( draw, elts[i + 1] ); - prim.v[2] = get_vertex( draw, elts[i + 2] ); - } - - first->tri( first, &prim ); - } - break; - case GL_TRIANGLE_FAN: - if (count >= 3) { - prim.v[0] = get_vertex( draw, elts[0] ); - prim.v[1] = 0; - prim.v[2] = get_vertex( draw, elts[1] ); - - for (i = 0; i+2 < count; i++) { - prim.v[1] = prim.v[2]; - prim.v[2] = get_vertex( draw, elts[i+2] ); - - first->tri( first, &prim ); - } - } - break; +static void draw_set_prim( struct draw_context *draw, + GLenum prim ) +{ + if (reduced_prim[prim] != draw->reduced_prim) { + draw_flush( draw ); + draw->reduced_prim = reduced_prim[prim]; + } - case GL_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( first, - get_vertex( draw, elts[i + 0] ), - get_vertex( draw, elts[i + 1] ), - get_vertex( draw, elts[i + 2] ), - get_vertex( draw, elts[i + 3] )); - } - break; + draw->prim = prim; +} - case GL_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( first, - get_vertex( draw, elts[i + 2] ), - get_vertex( draw, elts[i + 0] ), - get_vertex( draw, elts[i + 1] ), - get_vertex( draw, elts[i + 3] )); - } - break; +static void do_point( struct draw_context *draw, + GLuint i0 ) +{ + struct prim_header *prim = get_queued_prim( draw, 1 ); + + prim->reset_line_stipple = 0; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->get_vertex( draw, i0 ); +} - case GL_POLYGON: - if (count >= 3) { - int e1save, e2save; - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, elts[1] ); - prim.v[2] = get_vertex( draw, elts[0] ); - e2save = prim.v[2]->edgeflag; - - for (i = 0; i+2 < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, elts[i+2] ); - - /* save v1 edge flag, and clear if not last triangle */ - e1save = prim.v[1]->edgeflag; - if (i + 3 < count) - prim.v[1]->edgeflag = 0; - /* draw */ - first->tri( first, &prim ); +static void do_line( struct draw_context *draw, + GLboolean reset_stipple, + GLuint i0, + GLuint i1 ) +{ + struct prim_header *prim = get_queued_prim( draw, 2 ); + + prim->reset_line_stipple = reset_stipple; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->get_vertex( draw, i0 ); + prim->v[1] = draw->get_vertex( draw, i1 ); +} - prim.v[1]->edgeflag = e1save; /* restore */ - prim.v[2]->edgeflag = 0; /* disable edge after 1st tri */ - } - prim.v[2]->edgeflag = e2save; - } - break; +static void do_triangle( struct draw_context *draw, + GLuint i0, + GLuint i1, + GLuint i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + + prim->reset_line_stipple = 1; + prim->edgeflags = ~0; + prim->pad = 0; + prim->v[0] = draw->get_vertex( draw, i0 ); + prim->v[1] = draw->get_vertex( draw, i1 ); + prim->v[2] = draw->get_vertex( draw, i2 ); +} + +static void do_ef_triangle( struct draw_context *draw, + GLboolean reset_stipple, + GLuint ef_mask, + GLuint i0, + GLuint i1, + GLuint i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + struct vertex_header *v0 = draw->get_vertex( draw, i0 ); + struct vertex_header *v1 = draw->get_vertex( draw, i1 ); + struct vertex_header *v2 = draw->get_vertex( draw, i2 ); + + prim->reset_line_stipple = reset_stipple; + + prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | + (v1->edgeflag << 1) | + (v2->edgeflag << 2)); + prim->pad = 0; + prim->v[0] = v0; + prim->v[1] = v1; + prim->v[2] = v2; +} - default: - assert(0); - break; - } + +static void do_quad( struct draw_context *draw, + unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3 ) +{ + do_ef_triangle( draw, 1, ~(1<<0), v0, v1, v3 ); + do_ef_triangle( draw, 0, ~(1<<1), v1, v2, v3 ); } + static void draw_prim( struct draw_context *draw, GLuint start, GLuint count ) { - struct draw_stage * const first = draw->pipeline.first; - struct prim_header prim; GLuint i; // _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - prim.det = 0; /* valid from cull stage onwards */ - prim.v[0] = 0; - prim.v[1] = 0; - prim.v[2] = 0; - switch (draw->prim) { case GL_POINTS: for (i = 0; i < count; i ++) { - prim.v[0] = get_vertex( draw, start + i ); - first->point( first, &prim ); + do_point( draw, + start + i ); } break; case GL_LINES: for (i = 0; i+1 < count; i += 2) { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - - first->reset_stipple_counter( first ); - first->line( first, &prim ); + do_line( draw, + TRUE, + start + i + 0, + start + i + 1); } break; case GL_LINE_LOOP: if (count >= 2) { - first->reset_stipple_counter( first ); for (i = 1; i < count; i++) { - prim.v[0] = get_vertex( draw, start + i - 1 ); - prim.v[1] = get_vertex( draw, start + i ); - first->line( first, &prim ); + do_line( draw, + i == 1, /* XXX: only if vb not split */ + start + i - 1, + start + i ); } - prim.v[0] = get_vertex( draw, start + count - 1 ); - prim.v[1] = get_vertex( draw, start + 0 ); - first->line( first, &prim ); + do_line( draw, + 0, + start + count - 1, + start + 0 ); } break; case GL_LINE_STRIP: if (count >= 2) { - first->reset_stipple_counter( first ); - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, start + 0 ); - for (i = 1; i < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, start + i ); - - first->line( first, &prim ); + do_line( draw, + i == 1, + start + i - 1, + start + i ); } } break; case GL_TRIANGLES: for (i = 0; i+2 < count; i += 3) { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); + do_ef_triangle( draw, + 1, + ~0, + start + i + 0, + start + i + 1, + start + i + 2 ); } break; case GL_TRIANGLE_STRIP: for (i = 0; i+2 < count; i++) { if (i & 1) { - prim.v[0] = get_vertex( draw, start + i + 1 ); - prim.v[1] = get_vertex( draw, start + i + 0 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); + do_triangle( draw, + start + i + 1, + start + i + 0, + start + i + 2 ); } else { - prim.v[0] = get_vertex( draw, start + i + 0 ); - prim.v[1] = get_vertex( draw, start + i + 1 ); - prim.v[2] = get_vertex( draw, start + i + 2 ); + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); } - - first->tri( first, &prim ); } break; case GL_TRIANGLE_FAN: if (count >= 3) { - prim.v[0] = get_vertex( draw, start + 0 ); - prim.v[1] = 0; - prim.v[2] = get_vertex( draw, start + 1 ); - for (i = 0; i+2 < count; i++) { - prim.v[1] = prim.v[2]; - prim.v[2] = get_vertex( draw, start + i + 2 ); - - first->tri( first, &prim ); + do_triangle( draw, + start + 0, + start + i + 1, + start + i + 2 ); } } break; @@ -397,48 +418,42 @@ static void draw_prim( struct draw_context *draw, case GL_QUADS: for (i = 0; i+3 < count; i += 4) { - do_quad( first, - get_vertex( draw, start + i + 0 ), - get_vertex( draw, start + i + 1 ), - get_vertex( draw, start + i + 2 ), - get_vertex( draw, start + i + 3 )); + do_quad( draw, + start + i + 0, + start + i + 1, + start + i + 2, + start + i + 3); } break; case GL_QUAD_STRIP: for (i = 0; i+3 < count; i += 2) { - do_quad( first, - get_vertex( draw, start + i + 2 ), - get_vertex( draw, start + i + 0 ), - get_vertex( draw, start + i + 1 ), - get_vertex( draw, start + i + 3 )); + do_quad( draw, + start + i + 2, + start + i + 0, + start + i + 1, + start + i + 3); } break; case GL_POLYGON: if (count >= 3) { - int e1save, e2save; - prim.v[0] = 0; - prim.v[1] = get_vertex( draw, start + 1 ); - prim.v[2] = get_vertex( draw, start + 0 ); - e2save = prim.v[2]->edgeflag; + unsigned ef_mask = (1<<2) | (1<<0); for (i = 0; i+2 < count; i++) { - prim.v[0] = prim.v[1]; - prim.v[1] = get_vertex( draw, start + i + 2 ); - /* save v1 edge flag, and clear if not last triangle */ - e1save = prim.v[1]->edgeflag; - if (i + 3 < count) - prim.v[1]->edgeflag = 0; + if (i + 3 >= count) + ef_mask |= (1<<1); - /* draw */ - first->tri( first, &prim ); + do_ef_triangle( draw, + i == 0, + ef_mask, + start + i + 1, + start + i + 2, + start + i + 0); - prim.v[1]->edgeflag = e1save; /* restore */ - prim.v[2]->edgeflag = 0; /* disable edge after 1st tri */ + ef_mask &= ~(1<<2); } - prim.v[2]->edgeflag = e2save; } break; @@ -449,10 +464,18 @@ static void draw_prim( struct draw_context *draw, } -static void draw_release_vertices( struct draw_context *draw ) +static void draw_allocate_vertices( struct draw_context *draw, + GLuint nr_vertices ) { - draw->pipeline.first->end( draw->pipeline.first ); + draw->nr_vertices = nr_vertices; + draw->verts = (GLubyte *) malloc( nr_vertices * draw->vertex_size ); + draw_invalidate_vcache( draw ); +} + + +static void draw_release_vertices( struct draw_context *draw ) +{ free(draw->verts); draw->verts = NULL; } @@ -598,6 +621,8 @@ void draw_vb(struct draw_context *draw, draw->in_vb = 1; + draw->pipeline.first->begin( draw->pipeline.first ); + /* Allocate the vertices: */ draw_allocate_vertices( draw, VB->Count ); @@ -605,11 +630,14 @@ void draw_vb(struct draw_context *draw, /* Bind the vb outputs: */ vf_set_sources( draw->vf, VB->AttribPtr, 0 ); - - /* Build the hardware or prim-pipe vertices: - */ vf_emit_vertices( draw->vf, VB->Count, draw->verts ); + draw->elts = VB->Elts; + + if (VB->Elts) + draw->get_vertex = get_uint_elt_vertex; + else + draw->get_vertex = get_vertex; for (i = 0; i < VB->PrimitiveCount; i++) { @@ -628,21 +656,15 @@ void draw_vb(struct draw_context *draw, if (draw->prim != mode) draw_set_prim( draw, mode ); - if (VB->Elts) { - draw_indexed_prim( draw, - VB->Elts + start, - length ); - } - else { - draw_prim( draw, - start, - length ); - } + draw_prim( draw, start, length ); } + draw_flush(draw); + draw->pipeline.first->end( draw->pipeline.first ); draw_release_vertices( draw ); draw->verts = NULL; draw->in_vb = 0; + draw->elts = NULL; } @@ -661,11 +683,17 @@ draw_vertices(struct draw_context *draw, assert(mode <= GL_POLYGON); + draw->get_vertex = get_vertex; draw->vertex_size = sizeof(struct vertex_header) + numAttrs * 4 * sizeof(GLfloat); + + + /*draw_prim_info(mode, &first, &incr);*/ draw_allocate_vertices( draw, numVerts ); + draw->pipeline.first->begin( draw->pipeline.first ); + if (draw->prim != mode) draw_set_prim( draw, mode ); @@ -694,6 +722,9 @@ draw_vertices(struct draw_context *draw, /* draw */ draw_prim(draw, 0, numVerts); + draw_flush(draw); + draw->pipeline.first->end( draw->pipeline.first ); + /* clean up */ draw_release_vertices( draw ); @@ -759,8 +790,6 @@ void draw_set_vertex_attributes( struct draw_context *draw, } -#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) - void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ) { stage->nr_tmps = nr; -- cgit v1.2.3 From b4480285ed5098f1c862690ee105dd46f5e6cd1e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 14 Aug 2007 11:00:35 -0600 Subject: tgsi_sampler->get_samples() now operates on a quad (returns 4 colors). Lambda/level-of-detail is also computed in get_samples() now. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/pipe/softpipe/sp_tex_sample.c | 405 +++++++++++++++++++++------------ src/mesa/pipe/softpipe/sp_tex_sample.h | 7 +- src/mesa/pipe/tgsi/core/tgsi_exec.c | 87 +------ src/mesa/pipe/tgsi/core/tgsi_exec.h | 13 +- 5 files changed, 286 insertions(+), 228 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 70a80aff88..096385d7a1 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -322,7 +322,7 @@ static void shade_begin(struct quad_stage *qs) for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { qss->samplers[i].state = &softpipe->sampler[i]; qss->samplers[i].texture = softpipe->texture[i]; - qss->samplers[i].get_sample = sp_get_sample; + qss->samplers[i].get_samples = sp_get_samples; qss->samplers[i].pipe = &softpipe->pipe; /* init cache info here */ qss->samplers[i].cache_x = diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index fb4f7930ac..2005a3dbcd 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -66,12 +66,12 @@ * optimization! If we find that's not true on some systems, convert * to a macro. */ -static INLINE GLfloat -lerp_2d(GLfloat a, GLfloat b, - GLfloat v00, GLfloat v10, GLfloat v01, GLfloat v11) +static INLINE float +lerp_2d(float a, float b, + float v00, float v10, float v01, float v11) { - const GLfloat temp0 = LERP(a, v00, v10); - const GLfloat temp1 = LERP(a, v01, v11); + const float temp0 = LERP(a, v00, v10); + const float temp1 = LERP(a, v01, v11); return LERP(b, temp0, temp1); } @@ -98,7 +98,7 @@ repeat_remainder(GLint a, GLint b) * \return integer texture index */ static INLINE GLint -nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint size) +nearest_texcoord(GLuint wrapMode, float s, GLuint size) { GLint i; switch (wrapMode) { @@ -122,8 +122,8 @@ nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint size) { /* s limited to [min,max] */ /* i limited to [0, size-1] */ - const GLfloat min = 1.0F / (2.0F * size); - const GLfloat max = 1.0F - min; + const float min = 1.0F / (2.0F * size); + const float max = 1.0F - min; if (s < min) i = 0; else if (s > max) @@ -136,8 +136,8 @@ nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint size) { /* s limited to [min,max] */ /* i limited to [-1, size] */ - const GLfloat min = -1.0F / (2.0F * size); - const GLfloat max = 1.0F - min; + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; if (s <= min) i = -1; else if (s >= max) @@ -148,14 +148,14 @@ nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint size) return i; case PIPE_TEX_WRAP_MIRROR_REPEAT: { - const GLfloat min = 1.0F / (2.0F * size); - const GLfloat max = 1.0F - min; + const float min = 1.0F / (2.0F * size); + const float max = 1.0F - min; const GLint flr = IFLOOR(s); - GLfloat u; + float u; if (flr & 1) - u = 1.0F - (s - (GLfloat) flr); + u = 1.0F - (s - (float) flr); else - u = s - (GLfloat) flr; + u = s - (float) flr; if (u < min) i = 0; else if (u > max) @@ -168,7 +168,7 @@ nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint size) { /* s limited to [0,1] */ /* i limited to [0,size-1] */ - const GLfloat u = FABSF(s); + const float u = FABSF(s); if (u <= 0.0F) i = 0; else if (u >= 1.0F) @@ -181,9 +181,9 @@ nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint size) { /* s limited to [min,max] */ /* i limited to [0, size-1] */ - const GLfloat min = 1.0F / (2.0F * size); - const GLfloat max = 1.0F - min; - const GLfloat u = FABSF(s); + const float min = 1.0F / (2.0F * size); + const float max = 1.0F - min; + const float u = FABSF(s); if (u < min) i = 0; else if (u > max) @@ -196,9 +196,9 @@ nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint size) { /* s limited to [min,max] */ /* i limited to [0, size-1] */ - const GLfloat min = -1.0F / (2.0F * size); - const GLfloat max = 1.0F - min; - const GLfloat u = FABSF(s); + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; + const float u = FABSF(s); if (u < min) i = -1; else if (u > max) @@ -224,10 +224,10 @@ nearest_texcoord(GLuint wrapMode, GLfloat s, GLuint size) * \param a returns blend factor/weight between texture indexes */ static INLINE void -linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, - GLint *i0, GLint *i1, GLfloat *a) +linear_texcoord(GLuint wrapMode, float s, GLuint size, + GLint *i0, GLint *i1, float *a) { - GLfloat u; + float u; switch (wrapMode) { case PIPE_TEX_WRAP_REPEAT: u = s * size - 0.5F; @@ -238,7 +238,7 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, if (s <= 0.0F) u = 0.0F; else if (s >= 1.0F) - u = (GLfloat) size; + u = (float) size; else u = s * size; u -= 0.5F; @@ -249,7 +249,7 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, if (s <= 0.0F) u = 0.0F; else if (s >= 1.0F) - u = (GLfloat) size; + u = (float) size; else u = s * size; u -= 0.5F; @@ -262,8 +262,8 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, break; case PIPE_TEX_WRAP_CLAMP_TO_BORDER: { - const GLfloat min = -1.0F / (2.0F * size); - const GLfloat max = 1.0F - min; + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; if (s <= min) u = min * size; else if (s >= max) @@ -279,9 +279,9 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, { const GLint flr = IFLOOR(s); if (flr & 1) - u = 1.0F - (s - (GLfloat) flr); + u = 1.0F - (s - (float) flr); else - u = s - (GLfloat) flr; + u = s - (float) flr; u = (u * size) - 0.5F; *i0 = IFLOOR(u); *i1 = *i0 + 1; @@ -294,7 +294,7 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, case PIPE_TEX_WRAP_MIRROR_CLAMP: u = FABSF(s); if (u >= 1.0F) - u = (GLfloat) size; + u = (float) size; else u *= size; u -= 0.5F; @@ -304,7 +304,7 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: u = FABSF(s); if (u >= 1.0F) - u = (GLfloat) size; + u = (float) size; else u *= size; u -= 0.5F; @@ -317,8 +317,8 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, break; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: { - const GLfloat min = -1.0F / (2.0F * size); - const GLfloat max = 1.0F - min; + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; u = FABSF(s); if (u <= min) u = min * size; @@ -339,7 +339,8 @@ linear_texcoord(GLuint wrapMode, GLfloat s, GLuint size, static GLuint -choose_cube_face(const GLfloat texcoord[4], GLfloat newCoord[4]) +choose_cube_face(float rx, float ry, float rz, + float newCoord[4]) { /* major axis @@ -352,12 +353,9 @@ choose_cube_face(const GLfloat texcoord[4], GLfloat newCoord[4]) +rz TEXTURE_CUBE_MAP_POSITIVE_Z_EXT +rx -ry rz -rz TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT -rx -ry rz */ - const GLfloat rx = texcoord[0]; - const GLfloat ry = texcoord[1]; - const GLfloat rz = texcoord[2]; - const GLfloat arx = FABSF(rx), ary = FABSF(ry), arz = FABSF(rz); + const float arx = FABSF(rx), ary = FABSF(ry), arz = FABSF(rz); GLuint face; - GLfloat sc, tc, ma; + float sc, tc, ma; if (arx > ary && arx > arz) { if (rx >= 0.0F) { @@ -409,9 +407,62 @@ choose_cube_face(const GLfloat texcoord[4], GLfloat newCoord[4]) } +/** + * Examine the quad's texture coordinates to compute the partial + * derivatives w.r.t X and Y, then compute lambda (level of detail). + * + * This is only done for fragment shaders, not vertex shaders. + */ +static float +compute_lambda(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE]) +{ + float rho, lambda; + + assert(s); + { + float dsdx = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]; + float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; + dsdx = FABSF(dsdx); + dsdy = FABSF(dsdy); + rho = MAX2(dsdx, dsdy) * sampler->texture->width0; + } + if (t) { + float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; + float dtdy = t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]; + float max; + dtdx = FABSF(dtdx); + dtdy = FABSF(dtdy); + max = MAX2(dtdx, dtdy) * sampler->texture->height0; + rho = MAX2(rho, max); + } + if (p) { + float dpdx = p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT]; + float dpdy = p[QUAD_TOP_LEFT] - p[QUAD_BOTTOM_LEFT]; + float max; + dpdx = FABSF(dpdx); + dpdy = FABSF(dpdy); + max = MAX2(dpdx, dpdy) * sampler->texture->depth0; + rho = MAX2(rho, max); + } + + lambda = LOG2(rho); + + lambda += sampler->state->lod_bias; + lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); + + return lambda; +} + + static void -sp_get_sample_1d(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]) +sp_get_samples_1d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float rgba[NUM_CHANNELS][QUAD_SIZE]) { struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; struct pipe_surface *ps @@ -420,26 +471,40 @@ sp_get_sample_1d(struct tgsi_sampler *sampler, switch (sampler->state->min_img_filter) { case PIPE_TEX_FILTER_NEAREST: { - GLint x; - x = nearest_texcoord(sampler->state->wrap_s, strq[0], - sampler->texture->width0); - ps->get_tile(ps, x, 0, 1, 1, rgba); + GLuint j; + for (j = 0; j < QUAD_SIZE; j++) { + GLint x = nearest_texcoord(sampler->state->wrap_s, s[j], + sampler->texture->width0); + float texel[4]; + ps->get_tile(ps, x, 0, 1, 1, texel); + rgba[0][j] = texel[0]; + rgba[1][j] = texel[1]; + rgba[2][j] = texel[2]; + rgba[3][j] = texel[3]; + } } break; case PIPE_TEX_FILTER_LINEAR: { - GLfloat t0[4], t1[4]; - GLint x0, x1; - GLfloat a; - linear_texcoord(sampler->state->wrap_s, strq[0], - sampler->texture->width0, &x0, &x1, &a); - ps->get_tile(ps, x0, 0, 1, 1, t0); - ps->get_tile(ps, x1, 0, 1, 1, t1); - - rgba[0] = LERP(a, t0[0], t1[0]); - rgba[1] = LERP(a, t0[1], t1[1]); - rgba[2] = LERP(a, t0[2], t1[2]); - rgba[3] = LERP(a, t0[3], t1[3]); + GLuint j; + for (j = 0; j < QUAD_SIZE; j++) { + float t0[4], t1[4], texel[4]; + GLint x0, x1; + float a; + linear_texcoord(sampler->state->wrap_s, s[j], + sampler->texture->width0, &x0, &x1, &a); + ps->get_tile(ps, x0, 0, 1, 1, t0); + ps->get_tile(ps, x1, 0, 1, 1, t1); + + texel[0] = LERP(a, t0[0], t1[0]); + texel[1] = LERP(a, t0[1], t1[1]); + texel[2] = LERP(a, t0[2], t1[2]); + texel[3] = LERP(a, t0[3], t1[3]); + rgba[0][j] = texel[0]; + rgba[1][j] = texel[1]; + rgba[2][j] = texel[2]; + rgba[3][j] = texel[3]; + } } break; default: @@ -448,7 +513,7 @@ sp_get_sample_1d(struct tgsi_sampler *sampler, } static GLuint -choose_mipmap_level(struct tgsi_sampler *sampler, GLfloat lambda) +choose_mipmap_level(struct tgsi_sampler *sampler, float lambda) { if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { return 0; @@ -463,7 +528,68 @@ choose_mipmap_level(struct tgsi_sampler *sampler, GLfloat lambda) /** - * Called via tgsi_sampler::get_sample() + * Load the texture cache with a new texture tile. + */ +static void +cache_tex_tile(struct tgsi_sampler *sampler, + unsigned face, unsigned level, unsigned zslice, int cx, int cy) +{ + struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; + struct pipe_surface *ps + = pipe->get_tex_surface(pipe, sampler->texture, face, level, zslice); + assert(ps->width == sampler->texture->level[level].width); + assert(ps->height == sampler->texture->level[level].height); + sampler->cache_level = level; + sampler->cache_x = cx; + sampler->cache_y = cy; + ps->get_tile(ps, + cx * SAMPLER_CACHE_SIZE, + cy * SAMPLER_CACHE_SIZE, + SAMPLER_CACHE_SIZE, SAMPLER_CACHE_SIZE, + (float *) sampler->cache); +} + + +/** + * Get a texel from a texture. + * \param face the cube face in 0..5 + * \param level the mipmap level + * \param zslize which slice of a 3D texture + * \param x the x coord of texel within 2D image + * \param y the y coord of texel within 2D image + * \param rgba the quad to put the texel/color into + * \param j which element of the rgba quad to write to + */ +static void +get_texel(struct tgsi_sampler *sampler, + unsigned face, unsigned level, unsigned zslice, int x, int y, + float rgba[NUM_CHANNELS][QUAD_SIZE], GLuint j) +{ + int cx = x / SAMPLER_CACHE_SIZE; + int cy = y / SAMPLER_CACHE_SIZE; + + if (cx != sampler->cache_x || cy != sampler->cache_y || + level != sampler->cache_level) { + cache_tex_tile(sampler, face, level, zslice, cx, cy); + printf("cache miss (%d, %d)\n", x, y); + } + else { + printf("cache hit (%d, %d)\n", x, y); + } + + /* get texel from cache */ + cx = x % SAMPLER_CACHE_SIZE; + cy = y % SAMPLER_CACHE_SIZE; + rgba[0][j] = sampler->cache[cy][cx][0]; + rgba[1][j] = sampler->cache[cy][cx][1]; + rgba[2][j] = sampler->cache[cy][cx][2]; + rgba[3][j] = sampler->cache[cy][cx][3]; +} + + + +/** + * Called via tgsi_sampler::get_samples() * Use the sampler's state setting to get a filtered RGBA value * from the sampler's texture (mipmap tree). * @@ -475,87 +601,62 @@ choose_mipmap_level(struct tgsi_sampler *sampler, GLfloat lambda) * a new tgsi_sampler object for each state combo it finds.... */ static void -sp_get_sample_2d(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]) +sp_get_samples_2d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float rgba[NUM_CHANNELS][QUAD_SIZE]) { - struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; - GLuint filter; - GLint level0; - - if (lambda < 0.0) - filter = sampler->state->mag_img_filter; - else - filter = sampler->state->min_img_filter; + GLuint j, imgFilter; + int level0, width, height; + + /* compute level0 and imgFilter */ + switch (sampler->state->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NONE: + imgFilter = sampler->state->mag_img_filter; + level0 = 0; + assert(sampler->state->min_img_filter == + sampler->state->mag_img_filter); + break; + default: + { + float lambda = compute_lambda(sampler, s, t, p); + if (lambda < 0.0) + imgFilter = sampler->state->mag_img_filter; + else + imgFilter = sampler->state->min_img_filter; + level0 = choose_mipmap_level(sampler, lambda); + } + } - level0 = choose_mipmap_level(sampler, lambda); + width = sampler->texture->level[level0].width; + height = sampler->texture->level[level0].height; - assert(sampler->texture->level[level0].width); + assert(width > 0); - switch (filter) { + switch (imgFilter) { case PIPE_TEX_FILTER_NEAREST: - { - GLint x = nearest_texcoord(sampler->state->wrap_s, strq[0], - sampler->texture->level[level0].width); - GLint y = nearest_texcoord(sampler->state->wrap_t, strq[1], - sampler->texture->level[level0].height); - GLint cx = x / SAMPLER_CACHE_SIZE; - GLint cy = y / SAMPLER_CACHE_SIZE; - if (cx != sampler->cache_x || cy != sampler->cache_y || - level0 != sampler->cache_level) { - /* cache miss, replace cache with new tile */ - struct pipe_surface *ps - = pipe->get_tex_surface(pipe, sampler->texture, 0, level0, 0); - assert(ps->width == sampler->texture->level[level0].width); - assert(ps->height == sampler->texture->level[level0].height); - sampler->cache_level = level0; - sampler->cache_x = cx; - sampler->cache_y = cy; - ps->get_tile(ps, - cx * SAMPLER_CACHE_SIZE, - cy * SAMPLER_CACHE_SIZE, - SAMPLER_CACHE_SIZE, SAMPLER_CACHE_SIZE, - (GLfloat *) sampler->cache); - /*printf("cache miss (%d, %d)\n", x, y);*/ - } - else { - /*printf("cache hit (%d, %d)\n", x, y);*/ - } - /* get texel from cache */ - cx = x % SAMPLER_CACHE_SIZE; - cy = y % SAMPLER_CACHE_SIZE; - COPY_4V(rgba, sampler->cache[cy][cx]); + for (j = 0; j < QUAD_SIZE; j++) { + int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); + int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); + get_texel(sampler, 0, level0, 0, x, y, rgba, j); } break; case PIPE_TEX_FILTER_LINEAR: - { - GLfloat t00[4], t01[4], t10[4], t11[4]; - GLint x0, y0, x1, y1; - GLfloat a, b; - struct pipe_surface *ps - = pipe->get_tex_surface(pipe, sampler->texture, 0, level0, 0); - - linear_texcoord(sampler->state->wrap_s, strq[0], - sampler->texture->width0, &x0, &x1, &a); - linear_texcoord(sampler->state->wrap_t, strq[1], - sampler->texture->height0, &y0, &y1, &b); - ps->get_tile(ps, x0, y0, 1, 1, t00); - ps->get_tile(ps, x1, y0, 1, 1, t10); - ps->get_tile(ps, x0, y1, 1, 1, t01); - ps->get_tile(ps, x1, y1, 1, 1, t11); - - rgba[0] = lerp_2d(a, b, t00[0], t10[0], t01[0], t11[0]); - rgba[1] = lerp_2d(a, b, t00[1], t10[1], t01[1], t11[1]); - rgba[2] = lerp_2d(a, b, t00[2], t10[2], t01[2], t11[2]); - rgba[3] = lerp_2d(a, b, t00[3], t10[3], t01[3], t11[3]); - } - break; - /* - { - GLuint level0, level1; - level0 = choose_mipmap_level(sampler, lambda); + for (j = 0; j < QUAD_SIZE; j++) { + float tx[4][4], a, b; + int x0, y0, x1, y1, c; + linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &a); + linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &b); + get_texel(sampler, 0, level0, 0, x0, y0, tx, 0); + get_texel(sampler, 0, level0, 0, x1, y0, tx, 1); + get_texel(sampler, 0, level0, 0, x0, y1, tx, 2); + get_texel(sampler, 0, level0, 0, x1, y1, tx, 3); + for (c = 0; c < 4; c++) { + rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + } } break; - */ default: assert(0); } @@ -563,40 +664,52 @@ sp_get_sample_2d(struct tgsi_sampler *sampler, static void -sp_get_sample_3d(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat lamba, GLfloat rgba[4]) +sp_get_samples_3d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float rgba[NUM_CHANNELS][QUAD_SIZE]) { /* get/map pipe_surfaces corresponding to 3D tex slices */ } static void -sp_get_sample_cube(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]) +sp_get_samples_cube(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float rgba[NUM_CHANNELS][QUAD_SIZE]) { - GLfloat st[4]; - GLuint face = choose_cube_face(strq, st); - - /* get/map surface corresponding to the face */ + GLuint j; + for (j = 0; j < QUAD_SIZE; j++) { + float st[4]; + GLuint face = choose_cube_face(s[j], t[j], p[j], st); + (void) face; + /* get/map surface corresponding to the face */ + } } void -sp_get_sample(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]) +sp_get_samples(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float rgba[NUM_CHANNELS][QUAD_SIZE]) { switch (sampler->texture->target) { case GL_TEXTURE_1D: - sp_get_sample_1d(sampler, strq, lambda, rgba); + sp_get_samples_1d(sampler, s, t, p, rgba); break; case GL_TEXTURE_2D: - sp_get_sample_2d(sampler, strq, lambda, rgba); + sp_get_samples_2d(sampler, s, t, p, rgba); break; case GL_TEXTURE_3D: - sp_get_sample_3d(sampler, strq, lambda, rgba); + sp_get_samples_3d(sampler, s, t, p, rgba); break; case GL_TEXTURE_CUBE_MAP: - sp_get_sample_cube(sampler, strq, lambda, rgba); + sp_get_samples_cube(sampler, s, t, p, rgba); break; default: assert(0); diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.h b/src/mesa/pipe/softpipe/sp_tex_sample.h index 55ae49094b..417752d2d5 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.h +++ b/src/mesa/pipe/softpipe/sp_tex_sample.h @@ -6,8 +6,11 @@ struct tgsi_sampler; extern void -sp_get_sample(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]); +sp_get_samples(struct tgsi_sampler *sampler, + const GLfloat s[QUAD_SIZE], + const GLfloat t[QUAD_SIZE], + const GLfloat p[QUAD_SIZE], + GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]); #endif /* SP_TEX_SAMPLE_H */ diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 64d413b1b7..03b3b49e2b 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -1034,44 +1034,6 @@ exec_kil (struct tgsi_exec_machine *mach, } -static GLfloat -compute_lambda(struct tgsi_sampler *sampler, - const union tgsi_exec_channel *s, - const union tgsi_exec_channel *t, - const union tgsi_exec_channel *p) -{ - GLfloat rho, lambda; - - assert(s); - { - GLfloat dsdx = s->f[TILE_BOTTOM_RIGHT] - s->f[TILE_BOTTOM_LEFT]; - GLfloat dsdy = s->f[TILE_TOP_LEFT] - s->f[TILE_BOTTOM_LEFT]; - dsdx = FABSF(dsdx); - dsdy = FABSF(dsdy); - rho = MAX2(dsdx, dsdy) * sampler->texture->width0; - } - if (t) { - GLfloat dtdx = t->f[TILE_BOTTOM_RIGHT] - t->f[TILE_BOTTOM_LEFT]; - GLfloat dtdy = t->f[TILE_TOP_LEFT] - t->f[TILE_BOTTOM_LEFT]; - GLfloat max; - dtdx = FABSF(dtdx); - dtdy = FABSF(dtdy); - max = MAX2(dtdx, dtdy) * sampler->texture->height0; - rho = MAX2(rho, max); - } - if (p) { - GLfloat dpdx = p->f[TILE_BOTTOM_RIGHT] - p->f[TILE_BOTTOM_LEFT]; - GLfloat dpdy = p->f[TILE_TOP_LEFT] - p->f[TILE_BOTTOM_LEFT]; - GLfloat max; - dpdx = FABSF(dpdx); - dpdy = FABSF(dpdy); - max = MAX2(dpdx, dpdy) * sampler->texture->depth0; - rho = MAX2(rho, max); - } - - lambda = LOG2(rho); - return lambda; -} /* @@ -1079,44 +1041,24 @@ compute_lambda(struct tgsi_sampler *sampler, */ static void fetch_texel( struct tgsi_sampler *sampler, - GLfloat lambda, const union tgsi_exec_channel *s, const union tgsi_exec_channel *t, const union tgsi_exec_channel *p, - GLuint unit, union tgsi_exec_channel *r, union tgsi_exec_channel *g, union tgsi_exec_channel *b, union tgsi_exec_channel *a ) { - GLuint fragment_index; - GLfloat stpq[4][4]; - - for (fragment_index = 0; fragment_index < 4; fragment_index++) { - stpq[fragment_index][0] = s->f[fragment_index]; - if (t) - stpq[fragment_index][1] = t->f[fragment_index]; - if (p) - stpq[fragment_index][2] = p->f[fragment_index]; - } + GLuint j; + GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]; - lambda += sampler->state->lod_bias; - lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); - - /* XXX: Use the same lambda value throughout the tile. Could - * end up with four unique values by recalculating partial - * derivs in the other row and column, and calculating lambda - * using the dx and dy values appropriate for each fragment in - * the tile. - */ - - for (fragment_index = 0; fragment_index < 4; fragment_index++) { - GLfloat rgba[4]; - sampler->get_sample(sampler, stpq[fragment_index], lambda, rgba); - r->f[fragment_index] = rgba[0]; - g->f[fragment_index] = rgba[1]; - b->f[fragment_index] = rgba[2]; - a->f[fragment_index] = rgba[3]; + sampler->get_samples(sampler, s->f, t->f, p->f, rgba); + + for (j = 0; j < 4; j++) { + r->f[j] = rgba[0][j]; + g->f[j] = rgba[1][j]; + b->f[j] = rgba[2][j]; + a->f[j] = rgba[3][j]; } } @@ -1647,7 +1589,6 @@ exec_instruction( case TGSI_OPCODE_TEX: { const GLuint unit = inst->FullSrcRegisters[1].SrcRegister.Index; - GLfloat lambda; switch (inst->InstructionExtTexture.Texture) { case TGSI_TEXTURE_1D: @@ -1666,10 +1607,8 @@ exec_instruction( assert (0); } - lambda = compute_lambda(&mach->Samplers[unit], &r[0], NULL, NULL); - fetch_texel(&mach->Samplers[unit], lambda, + fetch_texel(&mach->Samplers[unit], &r[0], NULL, NULL, - inst->FullSrcRegisters[1].SrcRegister.Index, &r[0], &r[1], &r[2], &r[3]); break; @@ -1693,11 +1632,8 @@ exec_instruction( assert (0); } - lambda = compute_lambda(&mach->Samplers[unit], &r[0], &r[1], NULL); fetch_texel(&mach->Samplers[unit], - lambda, &r[0], &r[1], NULL, - inst->FullSrcRegisters[1].SrcRegister.Index, &r[0], &r[1], &r[2], &r[3]); break; @@ -1723,11 +1659,8 @@ exec_instruction( assert (0); } - lambda = compute_lambda(&mach->Samplers[unit], &r[0], &r[1], &r[2]); fetch_texel(&mach->Samplers[unit], - lambda, &r[0], &r[1], &r[2], - inst->FullSrcRegisters[1].SrcRegister.Index, &r[0], &r[1], &r[2], &r[3]); break; diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index 01dd7f11b8..4e6fef5724 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -23,12 +23,21 @@ struct tgsi_exec_vector #define SAMPLER_CACHE_SIZE 8 +#define NUM_CHANNELS 4 /* R,G,B,A */ +#ifndef QUAD_SIZE +#define QUAD_SIZE 4 /* 4 pixel/quad */ +#endif + struct tgsi_sampler { const struct pipe_sampler_state *state; struct pipe_mipmap_tree *texture; - void (*get_sample)(struct tgsi_sampler *sampler, - const GLfloat strq[4], GLfloat lambda, GLfloat rgba[4]); + /** Get samples for four fragments in a quad */ + void (*get_samples)(struct tgsi_sampler *sampler, + const GLfloat s[QUAD_SIZE], + const GLfloat t[QUAD_SIZE], + const GLfloat p[QUAD_SIZE], + GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]); void *pipe; /*XXX temporary*/ GLint cache_x, cache_y, cache_level; -- cgit v1.2.3 From f9e331a574cc4eba60e0de5a29a4aed4bb40520c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 14 Aug 2007 13:04:07 -0600 Subject: Implement remaining mipmap modes for 2D textures. Also, add lodbias param to get_samples() --- src/mesa/pipe/softpipe/sp_tex_sample.c | 129 +++++++++++++++++++++++++++------ src/mesa/pipe/softpipe/sp_tex_sample.h | 1 + src/mesa/pipe/tgsi/core/tgsi_exec.c | 9 ++- src/mesa/pipe/tgsi/core/tgsi_exec.h | 1 + 4 files changed, 113 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 2005a3dbcd..ab26b184c1 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -417,7 +417,8 @@ static float compute_lambda(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], - const float p[QUAD_SIZE]) + const float p[QUAD_SIZE], + float lodbias) { float rho, lambda; @@ -450,7 +451,7 @@ compute_lambda(struct tgsi_sampler *sampler, lambda = LOG2(rho); - lambda += sampler->state->lod_bias; + lambda += lodbias + sampler->state->lod_bias; lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); return lambda; @@ -462,6 +463,7 @@ sp_get_samples_1d(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; @@ -512,17 +514,43 @@ sp_get_samples_1d(struct tgsi_sampler *sampler, } } -static GLuint -choose_mipmap_level(struct tgsi_sampler *sampler, float lambda) + +/** + * From lambda and the min_mip_filter setting, choose the mipmap level(s) + * that are to be sampled from. + */ +static void +choose_mipmap_levels(struct tgsi_sampler *sampler, float lambda, + unsigned *level0, unsigned *level1) { - if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { - return 0; - } - else { - GLint level = (int) lambda; - level = CLAMP(level, sampler->texture->first_level, - sampler->texture->last_level); - return level; + switch (sampler->state->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NONE: + *level0 = *level1 = 0; + return; + case PIPE_TEX_MIPFILTER_NEAREST: + { + const int lvl = (int) (lambda + 0.5); + *level0 = + *level1 = CLAMP(lvl, + (int) sampler->texture->first_level, + (int) sampler->texture->last_level); + } + return; + case PIPE_TEX_MIPFILTER_LINEAR: + { + const int lvl = (int) lambda; + *level0 = CLAMP(lvl, + (int) sampler->texture->first_level, + (int) sampler->texture->last_level); + *level1 = CLAMP(lvl + 1, + (int) sampler->texture->first_level, + (int) sampler->texture->last_level); + assert(*level0 < 100); + assert(*level1 < 100); + } + return; + default: + assert(0); } } @@ -571,10 +599,14 @@ get_texel(struct tgsi_sampler *sampler, if (cx != sampler->cache_x || cy != sampler->cache_y || level != sampler->cache_level) { cache_tex_tile(sampler, face, level, zslice, cx, cy); + /* printf("cache miss (%d, %d)\n", x, y); + */ } else { + /* printf("cache hit (%d, %d)\n", x, y); + */ } /* get texel from cache */ @@ -605,27 +637,41 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { - GLuint j, imgFilter; - int level0, width, height; + unsigned level0, level1, j, imgFilter; + int width, height; + float mipBlend; /* compute level0 and imgFilter */ switch (sampler->state->min_mip_filter) { case PIPE_TEX_MIPFILTER_NONE: imgFilter = sampler->state->mag_img_filter; - level0 = 0; + level0 = level1 = 0; assert(sampler->state->min_img_filter == sampler->state->mag_img_filter); break; default: { - float lambda = compute_lambda(sampler, s, t, p); - if (lambda < 0.0) - imgFilter = sampler->state->mag_img_filter; + float lambda; + int fragment = 1; + if (fragment) + lambda = compute_lambda(sampler, s, t, p, lodbias); else + lambda = lodbias; /* not really a bias, but absolute LOD */ + + if (lambda < 0.0) { + /* magnifying */ + imgFilter = sampler->state->mag_img_filter; + level0 = level1 = 0; + } + else { + /* minifying */ imgFilter = sampler->state->min_img_filter; - level0 = choose_mipmap_level(sampler, lambda); + choose_mipmap_levels(sampler, lambda, &level0, &level1); + mipBlend = FRAC(lambda); /* blending weight between levels */ + } } } @@ -640,6 +686,18 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); get_texel(sampler, 0, level0, 0, x, y, rgba, j); + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + unsigned c; + x = x / 2; + y = y / 2; + get_texel(sampler, 0, level1, 0, x, y, rgba2, j); + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = LERP(mipBlend, rgba2[c][j], rgba[c][j]); + } + } } break; case PIPE_TEX_FILTER_LINEAR: @@ -655,6 +713,28 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, for (c = 0; c < 4; c++) { rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); } + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + unsigned c; + x0 = x0 / 2; + y0 = y0 / 2; + x1 = x1 / 2; + y1 = y1 / 2; + get_texel(sampler, 0, level1, 0, x0, y0, tx, 0); + get_texel(sampler, 0, level1, 0, x1, y0, tx, 1); + get_texel(sampler, 0, level1, 0, x0, y1, tx, 2); + get_texel(sampler, 0, level1, 0, x1, y1, tx, 3); + for (c = 0; c < 4; c++) { + rgba2[c][j] = lerp_2d(a, b, + tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + } + + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = LERP(mipBlend, rgba[c][j], rgba2[c][j]); + } + } } break; default: @@ -668,6 +748,7 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { /* get/map pipe_surfaces corresponding to 3D tex slices */ @@ -679,6 +760,7 @@ sp_get_samples_cube(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { GLuint j; @@ -696,20 +778,21 @@ sp_get_samples(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], const float t[QUAD_SIZE], const float p[QUAD_SIZE], + float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { switch (sampler->texture->target) { case GL_TEXTURE_1D: - sp_get_samples_1d(sampler, s, t, p, rgba); + sp_get_samples_1d(sampler, s, t, p, lodbias, rgba); break; case GL_TEXTURE_2D: - sp_get_samples_2d(sampler, s, t, p, rgba); + sp_get_samples_2d(sampler, s, t, p, lodbias, rgba); break; case GL_TEXTURE_3D: - sp_get_samples_3d(sampler, s, t, p, rgba); + sp_get_samples_3d(sampler, s, t, p, lodbias, rgba); break; case GL_TEXTURE_CUBE_MAP: - sp_get_samples_cube(sampler, s, t, p, rgba); + sp_get_samples_cube(sampler, s, t, p, lodbias, rgba); break; default: assert(0); diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.h b/src/mesa/pipe/softpipe/sp_tex_sample.h index 417752d2d5..43aae87283 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.h +++ b/src/mesa/pipe/softpipe/sp_tex_sample.h @@ -10,6 +10,7 @@ sp_get_samples(struct tgsi_sampler *sampler, const GLfloat s[QUAD_SIZE], const GLfloat t[QUAD_SIZE], const GLfloat p[QUAD_SIZE], + float lodbias, GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]); diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 03b3b49e2b..0a34754c40 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -1044,6 +1044,7 @@ fetch_texel( struct tgsi_sampler *sampler, const union tgsi_exec_channel *s, const union tgsi_exec_channel *t, const union tgsi_exec_channel *p, + float lodbias, union tgsi_exec_channel *r, union tgsi_exec_channel *g, union tgsi_exec_channel *b, @@ -1052,7 +1053,7 @@ fetch_texel( struct tgsi_sampler *sampler, GLuint j; GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]; - sampler->get_samples(sampler, s->f, t->f, p->f, rgba); + sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); for (j = 0; j < 4; j++) { r->f[j] = rgba[0][j]; @@ -1608,7 +1609,7 @@ exec_instruction( } fetch_texel(&mach->Samplers[unit], - &r[0], NULL, NULL, + &r[0], NULL, NULL, 0.0, &r[0], &r[1], &r[2], &r[3]); break; @@ -1633,7 +1634,7 @@ exec_instruction( } fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], NULL, + &r[0], &r[1], NULL, 0.0, &r[0], &r[1], &r[2], &r[3]); break; @@ -1660,7 +1661,7 @@ exec_instruction( } fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], + &r[0], &r[1], &r[2], 0.0, &r[0], &r[1], &r[2], &r[3]); break; diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index 4e6fef5724..b3ed124d3f 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -37,6 +37,7 @@ struct tgsi_sampler const GLfloat s[QUAD_SIZE], const GLfloat t[QUAD_SIZE], const GLfloat p[QUAD_SIZE], + GLfloat lodbias, GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]); void *pipe; /*XXX temporary*/ -- cgit v1.2.3 From c7722edcfdf36e0d0bfdc51013ecb199fc7fa9f6 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 14 Aug 2007 13:32:19 -0600 Subject: Consolidate filter and mipmap level selection in choose_mipmap_levels(). --- src/mesa/pipe/softpipe/sp_tex_sample.c | 193 +++++++++++++-------------------- 1 file changed, 75 insertions(+), 118 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index ab26b184c1..f62f0abf87 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -428,6 +428,7 @@ compute_lambda(struct tgsi_sampler *sampler, float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; dsdx = FABSF(dsdx); dsdy = FABSF(dsdy); + /* XXX only multiply by width for NORMALIZEd texcoords */ rho = MAX2(dsdx, dsdy) * sampler->texture->width0; } if (t) { @@ -436,6 +437,7 @@ compute_lambda(struct tgsi_sampler *sampler, float max; dtdx = FABSF(dtdx); dtdy = FABSF(dtdy); + /* XXX only multiply by height for NORMALIZEd texcoords */ max = MAX2(dtdx, dtdy) * sampler->texture->height0; rho = MAX2(rho, max); } @@ -445,6 +447,7 @@ compute_lambda(struct tgsi_sampler *sampler, float max; dpdx = FABSF(dpdx); dpdy = FABSF(dpdy); + /* XXX only multiply by depth for NORMALIZEd texcoords */ max = MAX2(dpdx, dpdy) * sampler->texture->depth0; rho = MAX2(rho, max); } @@ -458,99 +461,70 @@ compute_lambda(struct tgsi_sampler *sampler, } -static void -sp_get_samples_1d(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]) -{ - struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; - struct pipe_surface *ps - = pipe->get_tex_surface(pipe, sampler->texture, 0, 0, 0); - - switch (sampler->state->min_img_filter) { - case PIPE_TEX_FILTER_NEAREST: - { - GLuint j; - for (j = 0; j < QUAD_SIZE; j++) { - GLint x = nearest_texcoord(sampler->state->wrap_s, s[j], - sampler->texture->width0); - float texel[4]; - ps->get_tile(ps, x, 0, 1, 1, texel); - rgba[0][j] = texel[0]; - rgba[1][j] = texel[1]; - rgba[2][j] = texel[2]; - rgba[3][j] = texel[3]; - } - } - break; - case PIPE_TEX_FILTER_LINEAR: - { - GLuint j; - for (j = 0; j < QUAD_SIZE; j++) { - float t0[4], t1[4], texel[4]; - GLint x0, x1; - float a; - linear_texcoord(sampler->state->wrap_s, s[j], - sampler->texture->width0, &x0, &x1, &a); - ps->get_tile(ps, x0, 0, 1, 1, t0); - ps->get_tile(ps, x1, 0, 1, 1, t1); - - texel[0] = LERP(a, t0[0], t1[0]); - texel[1] = LERP(a, t0[1], t1[1]); - texel[2] = LERP(a, t0[2], t1[2]); - texel[3] = LERP(a, t0[3], t1[3]); - rgba[0][j] = texel[0]; - rgba[1][j] = texel[1]; - rgba[2][j] = texel[2]; - rgba[3][j] = texel[3]; - } - } - break; - default: - assert(0); - } -} - /** - * From lambda and the min_mip_filter setting, choose the mipmap level(s) - * that are to be sampled from. + * Do several things here: + * 1. Compute lambda from the texcoords, if needed + * 2. Determine if we're minifying or magnifying + * 3. If minifying, choose mipmap levels + * 4. Return image filter to use within mipmap images */ static void -choose_mipmap_levels(struct tgsi_sampler *sampler, float lambda, - unsigned *level0, unsigned *level1) +choose_mipmap_levels(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + unsigned *level0, unsigned *level1, float *levelBlend, + unsigned *imgFilter) { - switch (sampler->state->min_mip_filter) { - case PIPE_TEX_MIPFILTER_NONE: + if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { + /* no mipmap selection needed */ + assert(sampler->state->min_img_filter == + sampler->state->mag_img_filter); + *imgFilter = sampler->state->mag_img_filter; *level0 = *level1 = 0; - return; - case PIPE_TEX_MIPFILTER_NEAREST: - { - const int lvl = (int) (lambda + 0.5); - *level0 = - *level1 = CLAMP(lvl, - (int) sampler->texture->first_level, - (int) sampler->texture->last_level); + } + else { + float lambda; + + if (1) + /* fragment shader */ + lambda = compute_lambda(sampler, s, t, p, lodbias); + else + /* vertex shader */ + lambda = lodbias; /* not really a bias, but absolute LOD */ + + if (lambda < 0.0) { /* XXX threshold depends on the filter */ + /* magnifying */ + *imgFilter = sampler->state->mag_img_filter; + *level0 = *level1 = 0; } - return; - case PIPE_TEX_MIPFILTER_LINEAR: - { - const int lvl = (int) lambda; - *level0 = CLAMP(lvl, - (int) sampler->texture->first_level, - (int) sampler->texture->last_level); - *level1 = CLAMP(lvl + 1, - (int) sampler->texture->first_level, - (int) sampler->texture->last_level); - assert(*level0 < 100); - assert(*level1 < 100); + else { + /* minifying */ + *imgFilter = sampler->state->min_img_filter; + + /* choose mipmap level(s) and compute the blend factor between them */ + if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NEAREST) { + /* Nearest mipmap level */ + const int lvl = (int) (lambda + 0.5); + *level0 = + *level1 = CLAMP(lvl, + (int) sampler->texture->first_level, + (int) sampler->texture->last_level); + } + else { + /* Linear interpolation between mipmap levels */ + const int lvl = (int) lambda; + *level0 = CLAMP(lvl, + (int) sampler->texture->first_level, + (int) sampler->texture->last_level); + *level1 = CLAMP(lvl + 1, + (int) sampler->texture->first_level, + (int) sampler->texture->last_level); + *levelBlend = FRAC(lambda); /* blending weight between levels */ + } } - return; - default: - assert(0); } } @@ -620,6 +594,17 @@ get_texel(struct tgsi_sampler *sampler, +static void +sp_get_samples_1d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ +} + + /** * Called via tgsi_sampler::get_samples() * Use the sampler's state setting to get a filtered RGBA value @@ -642,38 +627,10 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, { unsigned level0, level1, j, imgFilter; int width, height; - float mipBlend; + float levelBlend; - /* compute level0 and imgFilter */ - switch (sampler->state->min_mip_filter) { - case PIPE_TEX_MIPFILTER_NONE: - imgFilter = sampler->state->mag_img_filter; - level0 = level1 = 0; - assert(sampler->state->min_img_filter == - sampler->state->mag_img_filter); - break; - default: - { - float lambda; - int fragment = 1; - if (fragment) - lambda = compute_lambda(sampler, s, t, p, lodbias); - else - lambda = lodbias; /* not really a bias, but absolute LOD */ - - if (lambda < 0.0) { - /* magnifying */ - imgFilter = sampler->state->mag_img_filter; - level0 = level1 = 0; - } - else { - /* minifying */ - imgFilter = sampler->state->min_img_filter; - choose_mipmap_levels(sampler, lambda, &level0, &level1); - mipBlend = FRAC(lambda); /* blending weight between levels */ - } - } - } + choose_mipmap_levels(sampler, s, t, p, lodbias, + &level0, &level1, &levelBlend, &imgFilter); width = sampler->texture->level[level0].width; height = sampler->texture->level[level0].height; @@ -695,7 +652,7 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, y = y / 2; get_texel(sampler, 0, level1, 0, x, y, rgba2, j); for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(mipBlend, rgba2[c][j], rgba[c][j]); + rgba[c][j] = LERP(levelBlend, rgba2[c][j], rgba[c][j]); } } } @@ -732,7 +689,7 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, } for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(mipBlend, rgba[c][j], rgba2[c][j]); + rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); } } } -- cgit v1.2.3 From a13de2464dd034ff117f9314df5757d068cae8e5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 14 Aug 2007 14:52:38 -0600 Subject: Implement texture cache with multiple, direct-mapped entries. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 13 ++++-- src/mesa/pipe/softpipe/sp_tex_sample.c | 81 ++++++++++++++++++++-------------- src/mesa/pipe/tgsi/core/tgsi_exec.h | 14 ++++-- 3 files changed, 67 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 096385d7a1..e767d0a470 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -318,16 +318,21 @@ static void shade_begin(struct quad_stage *qs) { struct quad_shade_stage *qss = quad_shade_stage(qs); struct softpipe_context *softpipe = qs->softpipe; - GLuint i; + GLuint i, entry; + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { qss->samplers[i].state = &softpipe->sampler[i]; qss->samplers[i].texture = softpipe->texture[i]; qss->samplers[i].get_samples = sp_get_samples; qss->samplers[i].pipe = &softpipe->pipe; /* init cache info here */ - qss->samplers[i].cache_x = - qss->samplers[i].cache_y = -1; - qss->samplers[i].cache_level = -1; + for (entry = 0; entry < TEX_CACHE_NUM_ENTRIES; entry++) { + qss->samplers[i].cache[entry].x = -1; + qss->samplers[i].cache[entry].y = -1; + qss->samplers[i].cache[entry].level = -1; + qss->samplers[i].cache[entry].face = -1; + qss->samplers[i].cache[entry].zslice = -1; + } } if (qs->next) diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index f62f0abf87..de14586b67 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -529,31 +529,27 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, } + /** - * Load the texture cache with a new texture tile. + * Given the texture face, level, zslice, x and y values, compute + * the cache entry position/index where we'd hope to find the + * cached texture tile. + * This is basically a direct-map cache. + * XXX There's probably lots of ways in which we can improve + * texture caching.... */ -static void -cache_tex_tile(struct tgsi_sampler *sampler, - unsigned face, unsigned level, unsigned zslice, int cx, int cy) +static unsigned +compute_cache_pos(unsigned face, unsigned level, unsigned zslice, + int tx, int ty) { - struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; - struct pipe_surface *ps - = pipe->get_tex_surface(pipe, sampler->texture, face, level, zslice); - assert(ps->width == sampler->texture->level[level].width); - assert(ps->height == sampler->texture->level[level].height); - sampler->cache_level = level; - sampler->cache_x = cx; - sampler->cache_y = cy; - ps->get_tile(ps, - cx * SAMPLER_CACHE_SIZE, - cy * SAMPLER_CACHE_SIZE, - SAMPLER_CACHE_SIZE, SAMPLER_CACHE_SIZE, - (float *) sampler->cache); + unsigned entry = tx + ty * 2 + zslice *4 + level + face; + return entry % TEX_CACHE_NUM_ENTRIES; } /** - * Get a texel from a texture. + * Get a texel from a texture, using the texture tile cache. + * * \param face the cube face in 0..5 * \param level the mipmap level * \param zslize which slice of a 3D texture @@ -565,17 +561,36 @@ cache_tex_tile(struct tgsi_sampler *sampler, static void get_texel(struct tgsi_sampler *sampler, unsigned face, unsigned level, unsigned zslice, int x, int y, - float rgba[NUM_CHANNELS][QUAD_SIZE], GLuint j) + float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j) { - int cx = x / SAMPLER_CACHE_SIZE; - int cy = y / SAMPLER_CACHE_SIZE; + int tx = x / TEX_CACHE_TILE_SIZE; + int ty = y / TEX_CACHE_TILE_SIZE; + unsigned entry = compute_cache_pos(face, level, zslice, tx, ty); + + if (tx != sampler->cache[entry].x || + ty != sampler->cache[entry].y || + face != sampler->cache[entry].face || + level != sampler->cache[entry].level || + zslice != sampler->cache[entry].zslice) { + /* entry is not what's expected */ + struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; + struct pipe_surface *ps + = pipe->get_tex_surface(pipe, sampler->texture, face, level, zslice); - if (cx != sampler->cache_x || cy != sampler->cache_y || - level != sampler->cache_level) { - cache_tex_tile(sampler, face, level, zslice, cx, cy); - /* printf("cache miss (%d, %d)\n", x, y); - */ + + assert(ps->width == sampler->texture->level[level].width); + assert(ps->height == sampler->texture->level[level].height); + sampler->cache[entry].x = tx; + sampler->cache[entry].y = ty; + sampler->cache[entry].level = level; + sampler->cache[entry].face = face; + sampler->cache[entry].zslice = zslice; + ps->get_tile(ps, + tx * TEX_CACHE_TILE_SIZE, + ty * TEX_CACHE_TILE_SIZE, + TEX_CACHE_TILE_SIZE, TEX_CACHE_TILE_SIZE, + (float *) sampler->cache[entry].data); } else { /* @@ -583,13 +598,13 @@ get_texel(struct tgsi_sampler *sampler, */ } - /* get texel from cache */ - cx = x % SAMPLER_CACHE_SIZE; - cy = y % SAMPLER_CACHE_SIZE; - rgba[0][j] = sampler->cache[cy][cx][0]; - rgba[1][j] = sampler->cache[cy][cx][1]; - rgba[2][j] = sampler->cache[cy][cx][2]; - rgba[3][j] = sampler->cache[cy][cx][3]; + /* get the texel from cache entry */ + tx = x % TEX_CACHE_TILE_SIZE; + ty = y % TEX_CACHE_TILE_SIZE; + rgba[0][j] = sampler->cache[entry].data[ty][tx][0]; + rgba[1][j] = sampler->cache[entry].data[ty][tx][1]; + rgba[2][j] = sampler->cache[entry].data[ty][tx][2]; + rgba[3][j] = sampler->cache[entry].data[ty][tx][3]; } diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index b3ed124d3f..5e07e18a31 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -21,13 +21,21 @@ struct tgsi_exec_vector union tgsi_exec_channel xyzw[4]; }; -#define SAMPLER_CACHE_SIZE 8 #define NUM_CHANNELS 4 /* R,G,B,A */ #ifndef QUAD_SIZE #define QUAD_SIZE 4 /* 4 pixel/quad */ #endif +#define TEX_CACHE_TILE_SIZE 8 +#define TEX_CACHE_NUM_ENTRIES 8 + +struct tgsi_texture_cache_entry +{ + int x, y, face, level, zslice; + GLfloat data[TEX_CACHE_TILE_SIZE][TEX_CACHE_TILE_SIZE][4]; +}; + struct tgsi_sampler { const struct pipe_sampler_state *state; @@ -40,9 +48,7 @@ struct tgsi_sampler GLfloat lodbias, GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]); void *pipe; /*XXX temporary*/ - - GLint cache_x, cache_y, cache_level; - GLfloat cache[SAMPLER_CACHE_SIZE][SAMPLER_CACHE_SIZE][4]; + struct tgsi_texture_cache_entry cache[TEX_CACHE_NUM_ENTRIES]; }; struct tgsi_exec_labels -- cgit v1.2.3 From a34b8594b7b2d00404bb639621ec1ce918ba0786 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 14 Aug 2007 15:12:04 -0600 Subject: refactor some code - cube map textures work now --- src/mesa/pipe/softpipe/sp_tex_sample.c | 122 ++++++++++++++++++++------------- 1 file changed, 73 insertions(+), 49 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index de14586b67..72000f9510 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -339,8 +339,7 @@ linear_texcoord(GLuint wrapMode, float s, GLuint size, static GLuint -choose_cube_face(float rx, float ry, float rz, - float newCoord[4]) +choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) { /* major axis @@ -400,8 +399,8 @@ choose_cube_face(float rx, float ry, float rz, } } - newCoord[0] = ( sc / ma + 1.0F ) * 0.5F; - newCoord[1] = ( tc / ma + 1.0F ) * 0.5F; + *newS = ( sc / ma + 1.0F ) * 0.5F; + *newT = ( tc / ma + 1.0F ) * 0.5F; return face; } @@ -542,8 +541,12 @@ static unsigned compute_cache_pos(unsigned face, unsigned level, unsigned zslice, int tx, int ty) { - unsigned entry = tx + ty * 2 + zslice *4 + level + face; +#if 01 + unsigned entry = tx + ty * 2 + zslice * 4 + face + level; return entry % TEX_CACHE_NUM_ENTRIES; +#else + return 0; +#endif } @@ -577,7 +580,9 @@ get_texel(struct tgsi_sampler *sampler, struct pipe_surface *ps = pipe->get_tex_surface(pipe, sampler->texture, face, level, zslice); - printf("cache miss (%d, %d)\n", x, y); + /* + printf("cache miss (%d, %d) face %u\n", tx, ty, face); + */ assert(ps->width == sampler->texture->level[level].width); assert(ps->height == sampler->texture->level[level].height); @@ -609,41 +614,23 @@ get_texel(struct tgsi_sampler *sampler, -static void -sp_get_samples_1d(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]) -{ -} - - /** - * Called via tgsi_sampler::get_samples() - * Use the sampler's state setting to get a filtered RGBA value - * from the sampler's texture (mipmap tree). - * - * XXX we can implement many versions of this function, each - * tightly coded for a specific combination of sampler state - * (nearest + repeat), (bilinear mipmap + clamp), etc. - * - * The update_samplers() function in st_atom_sampler.c could create - * a new tgsi_sampler object for each state combo it finds.... + * Common code for sampling 1D/2D/cube textures. + * Could probably extend for 3D... */ static void -sp_get_samples_2d(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]) +sp_get_samples_2d_common(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE], + const unsigned faces[4]) { unsigned level0, level1, j, imgFilter; int width, height; float levelBlend; - + choose_mipmap_levels(sampler, s, t, p, lodbias, &level0, &level1, &levelBlend, &imgFilter); @@ -657,7 +644,7 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, for (j = 0; j < QUAD_SIZE; j++) { int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); - get_texel(sampler, 0, level0, 0, x, y, rgba, j); + get_texel(sampler, faces[j], level0, 0, x, y, rgba, j); if (level0 != level1) { /* get texels from second mipmap level and blend */ @@ -665,7 +652,7 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, unsigned c; x = x / 2; y = y / 2; - get_texel(sampler, 0, level1, 0, x, y, rgba2, j); + get_texel(sampler, faces[j], level1, 0, x, y, rgba2, j); for (c = 0; c < NUM_CHANNELS; c++) { rgba[c][j] = LERP(levelBlend, rgba2[c][j], rgba[c][j]); } @@ -678,10 +665,10 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, int x0, y0, x1, y1, c; linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &a); linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &b); - get_texel(sampler, 0, level0, 0, x0, y0, tx, 0); - get_texel(sampler, 0, level0, 0, x1, y0, tx, 1); - get_texel(sampler, 0, level0, 0, x0, y1, tx, 2); - get_texel(sampler, 0, level0, 0, x1, y1, tx, 3); + get_texel(sampler, faces[j], level0, 0, x0, y0, tx, 0); + get_texel(sampler, faces[j], level0, 0, x1, y0, tx, 1); + get_texel(sampler, faces[j], level0, 0, x0, y1, tx, 2); + get_texel(sampler, faces[j], level0, 0, x1, y1, tx, 3); for (c = 0; c < 4; c++) { rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); } @@ -694,10 +681,10 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, y0 = y0 / 2; x1 = x1 / 2; y1 = y1 / 2; - get_texel(sampler, 0, level1, 0, x0, y0, tx, 0); - get_texel(sampler, 0, level1, 0, x1, y0, tx, 1); - get_texel(sampler, 0, level1, 0, x0, y1, tx, 2); - get_texel(sampler, 0, level1, 0, x1, y1, tx, 3); + get_texel(sampler, faces[j], level1, 0, x0, y0, tx, 0); + get_texel(sampler, faces[j], level1, 0, x1, y0, tx, 1); + get_texel(sampler, faces[j], level1, 0, x0, y1, tx, 2); + get_texel(sampler, faces[j], level1, 0, x1, y1, tx, 3); for (c = 0; c < 4; c++) { rgba2[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); @@ -715,6 +702,32 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, } +static void +sp_get_samples_1d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + static const unsigned faces[4] = {0, 0, 0, 0}; + sp_get_samples_2d_common(sampler, s, NULL, NULL, lodbias, rgba, faces); +} + + +static void +sp_get_samples_2d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + static const unsigned faces[4] = {0, 0, 0, 0}; + sp_get_samples_2d_common(sampler, s, t, NULL, lodbias, rgba, faces); +} + + static void sp_get_samples_3d(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], @@ -735,16 +748,27 @@ sp_get_samples_cube(struct tgsi_sampler *sampler, float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { - GLuint j; + unsigned faces[QUAD_SIZE], j; + float ssss[4], tttt[4]; for (j = 0; j < QUAD_SIZE; j++) { - float st[4]; - GLuint face = choose_cube_face(s[j], t[j], p[j], st); - (void) face; - /* get/map surface corresponding to the face */ + faces[j] = choose_cube_face(s[j], t[j], p[j], ssss + j, tttt + j); } + sp_get_samples_2d_common(sampler, ssss, tttt, NULL, lodbias, rgba, faces); } +/** + * Called via tgsi_sampler::get_samples() + * Use the sampler's state setting to get a filtered RGBA value + * from the sampler's texture (mipmap tree). + * + * XXX we can implement many versions of this function, each + * tightly coded for a specific combination of sampler state + * (nearest + repeat), (bilinear mipmap + clamp), etc. + * + * The update_samplers() function in st_atom_sampler.c could create + * a new tgsi_sampler object for each state combo it finds.... + */ void sp_get_samples(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], -- cgit v1.2.3 From d16b4bc32a731cb6ae320e8c187af3bc751d4138 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 14 Aug 2007 15:20:00 -0600 Subject: minor hack to allow 1D textures to work --- src/mesa/pipe/softpipe/sp_tex_sample.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 72000f9510..03fb539e7d 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -711,7 +711,8 @@ sp_get_samples_1d(struct tgsi_sampler *sampler, float rgba[NUM_CHANNELS][QUAD_SIZE]) { static const unsigned faces[4] = {0, 0, 0, 0}; - sp_get_samples_2d_common(sampler, s, NULL, NULL, lodbias, rgba, faces); + static const float tzero[4] = {0, 0, 0, 0}; + sp_get_samples_2d_common(sampler, s, tzero, NULL, lodbias, rgba, faces); } -- cgit v1.2.3 From b7316e4732e6930c118c7034166b47bd096a0046 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 14 Aug 2007 15:37:45 -0600 Subject: fix typo --- src/mesa/pipe/softpipe/sp_tex_sample.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 03fb539e7d..35993b946f 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -555,7 +555,7 @@ compute_cache_pos(unsigned face, unsigned level, unsigned zslice, * * \param face the cube face in 0..5 * \param level the mipmap level - * \param zslize which slice of a 3D texture + * \param zslice which slice of a 3D texture * \param x the x coord of texel within 2D image * \param y the y coord of texel within 2D image * \param rgba the quad to put the texel/color into -- cgit v1.2.3 From 267052c735f25dda7b49ded7b46cc59cb84ecaaa Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 14 Aug 2007 15:42:40 -0600 Subject: added glCopyPixels stub --- src/mesa/state_tracker/st_cb_drawpixels.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 7a89e853ee..df0b14463b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -485,9 +485,23 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } +static void +st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, + GLsizei width, GLsizei height, + GLint dstx, GLint dsty, GLenum type) +{ + struct st_context *st = ctx->st; + + st_validate_state(st); + + /* XXX to do */ +} + + void st_init_drawpixels_functions(struct dd_function_table *functions) { functions->DrawPixels = st_DrawPixels; + functions->CopyPixels = st_CopyPixels; functions->Bitmap = st_Bitmap; } -- cgit v1.2.3 From 681b1ebd5e1a94961a1d38b3018af97a10d2d5b0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 14 Aug 2007 17:20:40 -0600 Subject: use regular malloc/free instead of macros --- src/mesa/pipe/draw/draw_context.c | 4 ++-- src/mesa/pipe/draw/draw_vb.c | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 4335b47e09..0707c03e5f 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -80,8 +80,8 @@ void draw_destroy( struct draw_context *draw ) vf_destroy( draw->vf ); - FREE( draw->vcache.vertex[0] ); /* Frees all the vertices. */ - FREE( draw ); + free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ + free( draw ); } diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index ef32f02ec1..c0600314d6 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -795,7 +795,7 @@ void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ) stage->nr_tmps = nr; if (nr) { - GLubyte *store = MALLOC(MAX_VERTEX_SIZE * nr); + GLubyte *store = (GLubyte *) malloc(MAX_VERTEX_SIZE * nr); GLuint i; stage->tmp = MALLOC(sizeof(struct vertex_header *) * nr); @@ -808,7 +808,7 @@ void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ) void draw_free_tmps( struct draw_stage *stage ) { if (stage->tmp) { - FREE(stage->tmp[0]); - FREE(stage->tmp); + free(stage->tmp[0]); + free(stage->tmp); } } -- cgit v1.2.3 From 8038d5b68ca06e8ae4db4c999d7194593426d3bb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 11:42:37 -0600 Subject: comments, minor clean-ups --- src/mesa/pipe/draw/draw_vb.c | 30 ++++++++++++++++++++++-------- 1 file changed, 22 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index c0600314d6..dfd1f22eb3 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -464,9 +464,13 @@ static void draw_prim( struct draw_context *draw, } -static void draw_allocate_vertices( struct draw_context *draw, - GLuint nr_vertices ) +/** + * Allocate storage for post-transformation vertices. + */ +static void +draw_allocate_vertices( struct draw_context *draw, GLuint nr_vertices ) { + assert(draw->vertex_size > 0); draw->nr_vertices = nr_vertices; draw->verts = (GLubyte *) malloc( nr_vertices * draw->vertex_size ); draw_invalidate_vcache( draw ); @@ -474,7 +478,11 @@ static void draw_allocate_vertices( struct draw_context *draw, -static void draw_release_vertices( struct draw_context *draw ) +/** + * Free storage which was allocated by draw_allocate_vertices() + */ +static void +draw_release_vertices( struct draw_context *draw ) { free(draw->verts); draw->verts = NULL; @@ -603,7 +611,8 @@ static GLuint trim( GLuint count, GLuint first, GLuint incr ) } -/* This is a hack & will all go away. +/** + * This is a hack & will all go away. */ void draw_vb(struct draw_context *draw, struct vertex_buffer *VB ) @@ -621,9 +630,10 @@ void draw_vb(struct draw_context *draw, draw->in_vb = 1; + /* tell drawing pipeline we're beginning drawing */ draw->pipeline.first->begin( draw->pipeline.first ); - /* Allocate the vertices: + /* Allocate storage for the post-transform vertices: */ draw_allocate_vertices( draw, VB->Count ); @@ -640,9 +650,8 @@ void draw_vb(struct draw_context *draw, draw->get_vertex = get_vertex; for (i = 0; i < VB->PrimitiveCount; i++) { - - GLenum mode = VB->Primitive[i].mode; - GLuint start = VB->Primitive[i].start; + const GLenum mode = VB->Primitive[i].mode; + const GLuint start = VB->Primitive[i].start; GLuint length, first, incr; /* Trim the primitive down to a legal size. @@ -659,8 +668,13 @@ void draw_vb(struct draw_context *draw, draw_prim( draw, start, length ); } + /* draw any left-over buffered prims */ draw_flush(draw); + + /* tell drawing pipeline we're done drawing */ draw->pipeline.first->end( draw->pipeline.first ); + + /* free the post-transformed vertices */ draw_release_vertices( draw ); draw->verts = NULL; draw->in_vb = 0; -- cgit v1.2.3 From c431a22f520c9685e9da37bfdd6b6b2b0af03800 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 11:42:53 -0600 Subject: added some const qualifiers --- src/mesa/pipe/p_context.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index c0685fa8cb..ae2cc25fe3 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -138,11 +138,11 @@ struct pipe_context { void (*set_vertex_buffer)( struct pipe_context *, unsigned index, - struct pipe_vertex_buffer * ); + const struct pipe_vertex_buffer * ); void (*set_vertex_element)( struct pipe_context *, unsigned index, - struct pipe_vertex_element * ); + const struct pipe_vertex_element * ); /* -- cgit v1.2.3 From 6e02b4c7c3c6efc05ead2ffd46b0928f7c9316bf Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 11:43:14 -0600 Subject: comments --- src/mesa/pipe/p_state.h | 30 +++++++++++++++++++++--------- 1 file changed, 21 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 5051ebbfc0..4878cb41bd 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -320,23 +320,35 @@ struct pipe_mipmap_tree }; - +/** + * A vertex buffer. Typically, all the vertex data/attributes for + * drawing something will be in one buffer. But it's also possible, for + * example, to put colors in one buffer and texcoords in another. + */ struct pipe_vertex_buffer { - unsigned pitch:11; - unsigned max_index; - struct pipe_buffer_handle *buffer; - unsigned buffer_offset; + unsigned pitch:11; /**< stride to same attrib in next vertex, in bytes */ + unsigned max_index; /**< number of vertices in this buffer */ + unsigned buffer_offset; /**< offset to start of data in buffer, in bytes */ + struct pipe_buffer_handle *buffer; /**< the actual buffer */ }; - +/** + * Information to describe a vertex attribute (position, color, etc) + */ struct pipe_vertex_element { - unsigned src_offset:11; - unsigned vertex_buffer_index:5; + /** Offset of this attribute, in bytes, from the start of the vertex */ + unsigned src_offset:11; + + /** Which vertex_buffer (as given to pipe->set_vertex_buffer()) does + * this attribute live in? + */ + unsigned vertex_buffer_index:5; + unsigned dst_offset:8; - unsigned src_format:8; /* PIPE_FORMAT_* */ + unsigned src_format:8; /**< PIPE_FORMAT_* */ }; -- cgit v1.2.3 From 3f70b504386e5d0b67cbd719769e44f33db0ae87 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 11:44:33 -0600 Subject: softpipe vertex element/buffer state funcs --- src/mesa/pipe/softpipe/sp_state_vertex.c | 56 ++++++++++++++++++++++++++++++++ 1 file changed, 56 insertions(+) create mode 100644 src/mesa/pipe/softpipe/sp_state_vertex.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_vertex.c b/src/mesa/pipe/softpipe/sp_state_vertex.c new file mode 100644 index 0000000000..d985b20c36 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_vertex.c @@ -0,0 +1,56 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "sp_context.h" +#include "sp_state.h" +#include "sp_surface.h" + + +void +softpipe_set_vertex_element(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_element *attrib) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + assert(index < PIPE_ATTRIB_MAX); + softpipe->vertex_element[index] = *attrib; /* struct copy */ + softpipe->dirty |= SP_NEW_VERTEX; +} + + +void +softpipe_set_vertex_buffer(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_buffer *buffer) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + assert(index < PIPE_ATTRIB_MAX); + softpipe->vertex_buffer[index] = *buffer; /* struct copy */ + softpipe->dirty |= SP_NEW_VERTEX; +} -- cgit v1.2.3 From f30093418928dce5872ea179ed5d21d3b24b7c82 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 11:45:12 -0600 Subject: added state funcs for vertex buffer/element info --- src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_context.h | 3 +++ src/mesa/pipe/softpipe/sp_state.h | 8 ++++++++ 3 files changed, 12 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 615c612e9c..9c6d78c290 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -29,6 +29,7 @@ DRIVER_SOURCES = \ sp_state_sampler.c \ sp_state_setup.c \ sp_state_surface.c \ + sp_state_vertex.c \ sp_tex_layout.c \ sp_tex_sample.c \ sp_surface.c diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 8887c576c1..8b0e970ea2 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -64,6 +64,7 @@ enum interp_mode { #define SP_NEW_SAMPLER 0x400 #define SP_NEW_TEXTURE 0x800 #define SP_NEW_STENCIL 0x1000 +#define SP_NEW_VERTEX 0x2000 struct softpipe_context { @@ -88,6 +89,8 @@ struct softpipe_context { struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; unsigned dirty; /* Setup derived state. TODO: this should be passed in the program diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 5adf5930b8..9e5ed96901 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -81,6 +81,14 @@ void softpipe_set_texture_state( struct pipe_context *, void softpipe_set_viewport_state( struct pipe_context *, const struct pipe_viewport_state * ); +void softpipe_set_vertex_element(struct pipe_context *, + unsigned index, + const struct pipe_vertex_element *); + +void softpipe_set_vertex_buffer(struct pipe_context *, + unsigned index, + const struct pipe_vertex_buffer *); + void softpipe_update_derived( struct softpipe_context *softpipe ); #endif -- cgit v1.2.3 From bff0411c5a1a9fdfff7f8a5128af4d496b89fa0d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 11:45:54 -0600 Subject: sketch out vbo drawing function --- src/mesa/state_tracker/st_draw.c | 58 +++++++++++++++++++++++++++++++++++++++- 1 file changed, 57 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 95fa43df4d..002565a658 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -30,7 +30,9 @@ * Keith Whitwell */ -#include "imports.h" +#include "main/imports.h" + +#include "vbo/vbo.h" #include "tnl/t_context.h" #include "tnl/t_pipeline.h" @@ -38,7 +40,10 @@ #include "st_context.h" #include "st_atom.h" #include "st_draw.h" +#include "st_cb_bufferobjects.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" + /* * TNL stage which feeds into the above. @@ -88,6 +93,57 @@ static const struct tnl_pipeline_stage *st_pipeline[] = { 0, }; + + +/** + * This function gets plugged into the VBO module and is called when + * we have something to render. + * Basically, translate the information into the format expected by pipe. + */ +static void +draw_vbo(GLcontext *ctx, + const struct gl_client_array **arrays, + const struct _mesa_prim *prims, + GLuint nr_prims, + const struct _mesa_index_buffer *ib, + GLuint min_index, + GLuint max_index) +{ + struct pipe_context *pipe = ctx->st->pipe; + GLuint attr; + + /* tell pipe about the vertex array element/attributes */ + for (attr = 0; attr < 16; attr++) { + struct gl_buffer_object *bufobj = arrays[attr]->BufferObj; + if (bufobj && bufobj->Name) { + struct st_buffer_object *stobj = st_buffer_object(bufobj); + struct pipe_buffer_handle *buffer = stobj->buffer; + GLenum type = arrays[attr]->Type; + GLint size = arrays[attr]->Size; + struct pipe_vertex_buffer vbuffer; + struct pipe_vertex_element velement; + + vbuffer.pitch = 0; + vbuffer.max_index = 0; + vbuffer.buffer = NULL; + vbuffer.buffer_offset = 0; + + velement.src_offset = 0; + velement.vertex_buffer_index = attr; + velement.dst_offset = 0; + velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + + pipe->set_vertex_buffer(pipe, attr, &vbuffer); + pipe->set_vertex_element(pipe, attr, &velement); + } + } + + /* do actual drawing */ +} + + + + /* This is all a hack to keep using tnl until we have vertex programs * up and running. */ -- cgit v1.2.3 From 477ab9c1969e9d4cb42215c7e5a1eb4f93bbbddd Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 15 Aug 2007 12:19:09 +0100 Subject: Cleanup, do not use fp_attr_to_slot[] mapping array. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 151 +++++++++++++----------------------- 1 file changed, 52 insertions(+), 99 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index edb3245cba..bffb03a6bc 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -42,14 +42,13 @@ #include "main/mtypes.h" -#if 0 + #if defined __GNUC__ -#define ALIGNED_ATTRIBS 1 -#else -#define ALIGNED_ATTRIBS 0 -#endif +#define USE_ALIGNED_ATTRIBS 1 +#define ALIGN16_SUFFIX __attribute__(( aligned( 16 ) )) #else -#define ALIGNED_ATTRIBS 0 +#define USE_ALIGNED_ATTRIBS 0 +#define ALIGN16_SUFFIX #endif @@ -71,12 +70,7 @@ quad_shade_stage(struct quad_stage *qs) struct exec_machine { const struct setup_coefficient *coef; /**< will point to quad->coef */ - -#if ALIGNED_ATTRIBS - float attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE] __attribute__(( aligned( 16 ) )); -#else - float attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE]; -#endif + float attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE] ALIGN16_SUFFIX; }; @@ -161,6 +155,14 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) const float fx = quad->x0; const float fy = quad->y0; unsigned attr, i; + struct tgsi_exec_machine machine; + +#if USE_ALIGNED_ATTRIBS + struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX] ALIGN16_SUFFIX; +#else + struct tgsi_exec_vector inputs[FRAG_ATTRIB_MAX + 1]; + struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; +#endif exec.coef = quad->coef; @@ -209,105 +211,56 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) } } -#if 1 - /*softpipe->run_fs( tri->fp, quad, &tri->outputs );*/ - - { - struct tgsi_exec_machine machine; - struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; - struct tgsi_exec_vector *aoutputs; - unsigned i; - -#if !ALIGNED_ATTRIBS - struct tgsi_exec_vector inputs[FRAG_ATTRIB_MAX + 1]; - struct tgsi_exec_vector *ainputs; -#endif - #ifdef DEBUG - memset(&machine, 0, sizeof(machine)); + memset( &machine, 0, sizeof( machine ) ); #endif - /* init machine state */ - tgsi_exec_machine_init( - &machine, - softpipe->fs.tokens, - PIPE_MAX_SAMPLERS, qss->samplers); + assert( sizeof( struct tgsi_exec_vector ) == sizeof( exec.attr[0] ) ); - /* Consts does not require 16 byte alignment. */ - machine.Consts = softpipe->fs.constants->constant; + /* init machine state */ + tgsi_exec_machine_init( + &machine, + softpipe->fs.tokens, + PIPE_MAX_SAMPLERS, + qss->samplers ); - aoutputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs ); - machine.Outputs = aoutputs; + /* Consts does not require 16 byte alignment. */ + machine.Consts = softpipe->fs.constants->constant; - assert( sizeof( struct tgsi_exec_vector ) == sizeof( exec.attr[0] ) ); - -#if ALIGNED_ATTRIBS - machine.Inputs = (struct tgsi_exec_vector *) exec.attr; - - for (i = 0; i < softpipe->nr_attrs; i++) { - /* Make sure fp_attr_to_slot[] is an identity transform. */ - assert( softpipe->fp_attr_to_slot[i] == i ); - } -#else - ainputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); - machine.Inputs = ainputs; - - /* load input registers */ - for (i = 0; i < softpipe->nr_attrs; i++) { -#if 01 - /* Make sure fp_attr_to_slot[] is an identity transform. */ - /* - assert( softpipe->fp_attr_to_slot[i] == i ); - */ - memcpy( - &ainputs[i], - exec.attr[i], - sizeof( ainputs[0] ) ); +#if USE_ALIGNED_ATTRIBS + machine.Inputs = (struct tgsi_exec_vector *) exec.attr; + machine.Outputs = outputs; #else - memcpy( - &ainputs[i], - exec.attr[softpipe->fp_attr_to_slot[i]], - sizeof( ainputs[0] ) ); -#endif - } + machine.Inputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); + machine.Outputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs ); + + memcpy( + machine.Inputs, + exec.attr, + softpipe->nr_attrs * sizeof( struct tgsi_exec_vector ) ); #endif - /* run shader */ - tgsi_exec_machine_run( &machine ); - - /* store result color */ - memcpy(quad->outputs.color, - &aoutputs[FRAG_ATTRIB_COL0].xyzw[0].f[0], - sizeof(quad->outputs.color)); - if (softpipe->need_z) { - /* XXX temporary */ - quad->outputs.depth[0] = exec.attr[0][2][0]; - quad->outputs.depth[1] = exec.attr[0][2][1]; - quad->outputs.depth[2] = exec.attr[0][2][2]; - quad->outputs.depth[3] = exec.attr[0][2][3]; - } - } -#else - { - unsigned attr = softpipe->fp_attr_to_slot[FRAG_ATTRIB_COL0]; - assert(attr); - - memcpy(quad->outputs.color, - exec.attr[attr], - sizeof(quad->outputs.color)); - - if (softpipe->need_z) { - quad->outputs.depth[0] = exec.attr[0][2][0]; - quad->outputs.depth[1] = exec.attr[0][2][1]; - quad->outputs.depth[2] = exec.attr[0][2][2]; - quad->outputs.depth[3] = exec.attr[0][2][3]; - } + /* run shader */ + tgsi_exec_machine_run( &machine ); + + /* store result color */ + memcpy( + quad->outputs.color, + &machine.Outputs[FRAG_ATTRIB_COL0].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); + + if( softpipe->need_z ) { + /* XXX temporary */ + quad->outputs.depth[0] = exec.attr[0][2][0]; + quad->outputs.depth[1] = exec.attr[0][2][1]; + quad->outputs.depth[2] = exec.attr[0][2][2]; + quad->outputs.depth[3] = exec.attr[0][2][3]; } -#endif /* shader may cull fragments */ - if (quad->mask) - qs->next->run(qs->next, quad); + if( quad->mask ) { + qs->next->run( qs->next, quad ); + } } -- cgit v1.2.3 From 6504dc31782ae8087a21607a84283ebe93c1c344 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 15 Aug 2007 13:08:40 +0100 Subject: Declare fragment shader output registers. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 104 ++++++++++++++++++++++++--------- 1 file changed, 77 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index c3201ece2c..690339bc66 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -1,7 +1,7 @@ #include "tgsi_platform.h" #include "tgsi_mesa.h" -#define TGSI_DEBUG 0 +#define TGSI_DEBUG 1 /* * Map mesa register file to SBIR register file. @@ -463,13 +463,37 @@ compile_instruction( return GL_FALSE; } +static struct tgsi_full_declaration +make_declaration( + GLuint file, + GLuint first, + GLuint last, + GLboolean do_interpolate, + GLuint interpolate ) +{ + struct tgsi_full_declaration decl; + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = file; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.u.DeclarationRange.First = first; + decl.u.DeclarationRange.Last = last; + + if( do_interpolate ) { + decl.Declaration.Interpolate = 1; + decl.Interpolation.Interpolate = interpolate; + } + + return decl; +} + GLboolean tgsi_mesa_compile_fp_program( const struct gl_fragment_program *program, struct tgsi_token *tokens, GLuint maxTokens ) { - GLuint i, ti; + GLuint i, ti, count; struct tgsi_header *header; struct tgsi_full_declaration fulldecl; struct tgsi_full_instruction fullinst; @@ -486,53 +510,79 @@ tgsi_mesa_compile_fp_program( ti = 2; /* - * Input 0 is always read, at least implicitly by the instruction generated - * above, so mark it as used. + * Input 0 is always read, at least implicitly by the MOV instruction generated + * below, so mark it as used. */ inputs_read = program->Base.InputsRead | 1; /* - * Declare input attributes. + * Declare input attributes. Note that we do not interpolate fragment position. */ - fulldecl = tgsi_default_full_declaration(); - - fulldecl.Declaration.File = TGSI_FILE_INPUT; - fulldecl.Declaration.Declare = TGSI_DECLARE_RANGE; - fulldecl.Declaration.Interpolate = 1; - - /* - * Do not interpolate fragment position. - */ - fulldecl.u.DeclarationRange.First = 0; - fulldecl.u.DeclarationRange.Last = 0; - - fulldecl.Interpolation.Interpolate = TGSI_INTERPOLATE_CONSTANT; - + fulldecl = make_declaration( + TGSI_FILE_INPUT, + 0, + 0, + GL_TRUE, + TGSI_INTERPOLATE_CONSTANT ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], header, maxTokens - ti ); - /* - * Interpolate generic attributes. - */ - fulldecl.u.DeclarationRange.First = 1; - fulldecl.u.DeclarationRange.Last = 1; + count = 0; for( i = 1; i < 32; i++ ) { if( inputs_read & (1 << i) ) { - fulldecl.u.DeclarationRange.Last++; + count++; } } + if( count > 0 ) { + fulldecl = make_declaration( + TGSI_FILE_INPUT, + 1, + count + 1, + GL_TRUE, + TGSI_INTERPOLATE_LINEAR ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } - fulldecl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR; - + /* + * Declare output attributes. + */ + assert( + program->Base.OutputsWritten == + (program->Base.OutputsWritten & ((1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR))) ); + + fulldecl = make_declaration( + TGSI_FILE_OUTPUT, + 0, + 0, + GL_FALSE, + 0 ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], header, maxTokens - ti ); + if( program->Base.OutputsWritten & (1 << FRAG_RESULT_COLR) ) { + fulldecl = make_declaration( + TGSI_FILE_OUTPUT, + 1, + 1, + GL_FALSE, + 0 ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + /* * Copy input fragment xyz to output xyz. * If the shader writes depth, do not copy the z component. -- cgit v1.2.3 From b9eeb8dccff3b440a299f19a0868a3ff1cda1e09 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 15 Aug 2007 13:55:22 +0100 Subject: Add declaration_semantic token to TGSI. Cosmetic changes, GLuint -> unsigned. Preserve mesa gl_fragment_program DEPTH and COLOR semantics. --- src/mesa/pipe/tgsi/core/tgsi_build.c | 103 +++++++++++++++++++++++++-------- src/mesa/pipe/tgsi/core/tgsi_build.h | 11 ++++ src/mesa/pipe/tgsi/core/tgsi_dump.c | 36 ++++++++++++ src/mesa/pipe/tgsi/core/tgsi_parse.c | 14 +++-- src/mesa/pipe/tgsi/core/tgsi_parse.h | 3 +- src/mesa/pipe/tgsi/core/tgsi_token.h | 69 ++++++++++++---------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 54 +++++++++-------- 7 files changed, 205 insertions(+), 85 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.c b/src/mesa/pipe/tgsi/core/tgsi_build.c index bbf4a89a5e..d9c9a45532 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_build.c +++ b/src/mesa/pipe/tgsi/core/tgsi_build.c @@ -35,8 +35,8 @@ tgsi_build_header( void ) static void header_headersize_grow( struct tgsi_header *header ) { - assert (header->HeaderSize < 0xFF); - assert (header->BodySize == 0); + assert( header->HeaderSize < 0xFF ); + assert( header->BodySize == 0 ); header->HeaderSize++; } @@ -44,7 +44,7 @@ header_headersize_grow( struct tgsi_header *header ) static void header_bodysize_grow( struct tgsi_header *header ) { - assert (header->BodySize < 0xFFFFFF); + assert( header->BodySize < 0xFFFFFF ); header->BodySize++; } @@ -62,7 +62,7 @@ tgsi_default_processor( void ) struct tgsi_processor tgsi_build_processor( - GLuint type, + unsigned type, struct tgsi_header *header ) { struct tgsi_processor processor; @@ -89,6 +89,7 @@ tgsi_default_declaration( void ) declaration.File = TGSI_FILE_NULL; declaration.Declare = TGSI_DECLARE_RANGE; declaration.Interpolate = 0; + declaration.Semantic = 0; declaration.Padding = 0; declaration.Extended = 0; @@ -97,20 +98,22 @@ tgsi_default_declaration( void ) struct tgsi_declaration tgsi_build_declaration( - GLuint file, - GLuint declare, - GLuint interpolate, + unsigned file, + unsigned declare, + unsigned interpolate, + unsigned semantic, struct tgsi_header *header ) { struct tgsi_declaration declaration; - assert (file <= TGSI_FILE_IMMEDIATE); - assert (declare <= TGSI_DECLARE_MASK); + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( declare <= TGSI_DECLARE_MASK ); declaration = tgsi_default_declaration(); declaration.File = file; declaration.Declare = declare; declaration.Interpolate = interpolate; + declaration.Semantic = semantic; header_bodysize_grow( header ); @@ -122,7 +125,7 @@ declaration_grow( struct tgsi_declaration *declaration, struct tgsi_header *header ) { - assert (declaration->Size < 0xFF); + assert( declaration->Size < 0xFF ); declaration->Size++; @@ -136,18 +139,19 @@ tgsi_default_full_declaration( void ) full_declaration.Declaration = tgsi_default_declaration(); full_declaration.Interpolation = tgsi_default_declaration_interpolation(); + full_declaration.Semantic = tgsi_default_declaration_semantic(); return full_declaration; } -GLuint +unsigned tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, + const struct tgsi_full_declaration *full_decl, struct tgsi_token *tokens, struct tgsi_header *header, - GLuint maxsize ) + unsigned maxsize ) { - GLuint size = 0; + unsigned size = 0; struct tgsi_declaration *declaration; if( maxsize <= size ) @@ -159,14 +163,15 @@ tgsi_build_full_declaration( full_decl->Declaration.File, full_decl->Declaration.Declare, full_decl->Declaration.Interpolate, + full_decl->Declaration.Semantic, header ); switch( full_decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: + case TGSI_DECLARE_RANGE: { struct tgsi_declaration_range *dr; - if( maxsize <= size ) + if( maxsize <= size ) return 0; dr = (struct tgsi_declaration_range *) &tokens[size]; size++; @@ -179,11 +184,11 @@ tgsi_build_full_declaration( break; } - case TGSI_DECLARE_MASK: + case TGSI_DECLARE_MASK: { - struct tgsi_declaration_mask *dm; + struct tgsi_declaration_mask *dm; - if( maxsize <= size ) + if( maxsize <= size ) return 0; dm = (struct tgsi_declaration_mask *) &tokens[size]; size++; @@ -213,20 +218,35 @@ tgsi_build_full_declaration( header ); } + if( full_decl->Declaration.Semantic ) { + struct tgsi_declaration_semantic *ds; + + if( maxsize <= size ) + return 0; + ds = (struct tgsi_declaration_semantic *) &tokens[size]; + size++; + + *ds = tgsi_build_declaration_semantic( + full_decl->Semantic.SemanticName, + full_decl->Semantic.SemanticIndex, + declaration, + header ); + } + return size; } struct tgsi_declaration_range tgsi_build_declaration_range( - GLuint first, - GLuint last, + unsigned first, + unsigned last, struct tgsi_declaration *declaration, struct tgsi_header *header ) { struct tgsi_declaration_range declaration_range; - assert (last >= first); - assert (last <= 0xFFFF); + assert( last >= first ); + assert( last <= 0xFFFF ); declaration_range.First = first; declaration_range.Last = last; @@ -238,7 +258,7 @@ tgsi_build_declaration_range( struct tgsi_declaration_mask tgsi_build_declaration_mask( - GLuint mask, + unsigned mask, struct tgsi_declaration *declaration, struct tgsi_header *header ) { @@ -264,7 +284,7 @@ tgsi_default_declaration_interpolation( void ) struct tgsi_declaration_interpolation tgsi_build_declaration_interpolation( - GLuint interpolate, + unsigned interpolate, struct tgsi_declaration *declaration, struct tgsi_header *header ) { @@ -280,6 +300,39 @@ tgsi_build_declaration_interpolation( return di; } +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ) +{ + struct tgsi_declaration_semantic ds; + + ds.SemanticName = TGSI_SEMANTIC_DEPTH; + ds.SemanticIndex = 0; + ds.Padding = 0; + + return ds; +} + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_semantic ds; + + assert( semantic_name <= TGSI_SEMANTIC_COLOR ); + assert( semantic_index <= 0xFFFF ); + + ds = tgsi_default_declaration_semantic(); + ds.SemanticName = semantic_name; + ds.SemanticIndex = semantic_index; + + declaration_grow( declaration, header ); + + return ds; +} + /* * immediate */ diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.h b/src/mesa/pipe/tgsi/core/tgsi_build.h index 8b5db5662c..b3eb0715ee 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_build.h +++ b/src/mesa/pipe/tgsi/core/tgsi_build.h @@ -39,6 +39,7 @@ tgsi_build_declaration( unsigned file, unsigned declare, unsigned interpolate, + unsigned semantic, struct tgsi_header *header ); struct tgsi_full_declaration @@ -73,6 +74,16 @@ tgsi_build_declaration_interpolation( struct tgsi_declaration *declaration, struct tgsi_header *header ); +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ); + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + /* * immediate */ diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index 51d3bb703d..5a9f92004e 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -199,6 +199,18 @@ static const char *TGSI_INTERPOLATES_SHORT[] = "PERSPECTIVE" }; +static const char *TGSI_SEMANTICS[] = +{ + "SEMANTIC_DEPTH", + "SEMANTIC_COLOR" +}; + +static const char *TGSI_SEMANTICS_SHORT[] = +{ + "DEPTH", + "COLOR" +}; + static const char *TGSI_IMMS[] = { "IMM_FLOAT32" @@ -625,6 +637,14 @@ dump_declaration_short( TXT( ", " ); ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); } + + if( decl->Declaration.Semantic ) { + TXT( ", " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); + CHR( '[' ); + UID( decl->Semantic.SemanticIndex ); + CHR( ']' ); + } } static void @@ -643,6 +663,10 @@ dump_declaration_verbose( TXT( "\nInterpolate: " ); UID( decl->Declaration.Interpolate ); } + if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { + TXT( "\nSemantic : " ); + UID( decl->Declaration.Semantic ); + } if( ignored ) { TXT( "\nPadding : " ); UIX( decl->Declaration.Padding ); @@ -675,6 +699,18 @@ dump_declaration_verbose( UIX( decl->Interpolation.Padding ); } } + + if( decl->Declaration.Semantic ) { + CHR( '\n' ); + TXT( "\nSemanticName : " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); + TXT( "\nSemanticIndex: " ); + UID( decl->Semantic.SemanticIndex ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Semantic.Padding ); + } + } } static void diff --git a/src/mesa/pipe/tgsi/core/tgsi_parse.c b/src/mesa/pipe/tgsi/core/tgsi_parse.c index df34fa7d7e..a6724d762c 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_parse.c +++ b/src/mesa/pipe/tgsi/core/tgsi_parse.c @@ -16,7 +16,7 @@ tgsi_full_token_free( free( full_token->FullImmediate.u.Pointer ); } -GLuint +unsigned tgsi_parse_init( struct tgsi_parse_context *ctx, const struct tgsi_token *tokens ) @@ -72,7 +72,7 @@ tgsi_parse_token( struct tgsi_parse_context *ctx ) { struct tgsi_token token; - GLuint i; + unsigned i; tgsi_full_token_free( &ctx->FullToken ); tgsi_full_token_init( &ctx->FullToken ); @@ -104,6 +104,10 @@ tgsi_parse_token( next_token( ctx, &decl->Interpolation ); } + if( decl->Declaration.Semantic ) { + next_token( ctx, &decl->Semantic ); + } + break; } @@ -135,7 +139,7 @@ tgsi_parse_token( case TGSI_TOKEN_TYPE_INSTRUCTION: { struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; - GLuint extended; + unsigned extended; *inst = tgsi_default_full_instruction(); inst->Instruction = *(struct tgsi_instruction *) &token; @@ -173,7 +177,7 @@ tgsi_parse_token( assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - GLuint extended; + unsigned extended; next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); @@ -212,7 +216,7 @@ tgsi_parse_token( assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - GLuint extended; + unsigned extended; next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); diff --git a/src/mesa/pipe/tgsi/core/tgsi_parse.h b/src/mesa/pipe/tgsi/core/tgsi_parse.h index bba01431fa..56a8c254c2 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_parse.h +++ b/src/mesa/pipe/tgsi/core/tgsi_parse.h @@ -42,6 +42,7 @@ struct tgsi_full_declaration struct tgsi_declaration_mask DeclarationMask; } u; struct tgsi_declaration_interpolation Interpolation; + struct tgsi_declaration_semantic Semantic; }; struct tgsi_full_immediate @@ -86,7 +87,7 @@ tgsi_full_token_free( struct tgsi_parse_context { const struct tgsi_token *Tokens; - unsigned Position; + unsigned Position; struct tgsi_full_version FullVersion; struct tgsi_full_header FullHeader; union tgsi_full_token FullToken; diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index dc9301ed37..f72c7dc4be 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -54,13 +54,14 @@ struct tgsi_token struct tgsi_declaration { - unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ - unsigned Size : 8; /* UINT */ - unsigned File : 4; /* TGSI_FILE_ */ - unsigned Declare : 4; /* TGSI_DECLARE_ */ - unsigned Interpolate : 1; /* BOOL */ - unsigned Padding : 10; - unsigned Extended : 1; /* BOOL */ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ + unsigned Size : 8; /* UINT */ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned Declare : 4; /* TGSI_DECLARE_ */ + unsigned Interpolate : 1; /* BOOL */ + unsigned Semantic : 1; /* BOOL */ + unsigned Padding : 9; + unsigned Extended : 1; /* BOOL */ }; struct tgsi_declaration_range @@ -84,6 +85,16 @@ struct tgsi_declaration_interpolation unsigned Padding : 28; }; +#define TGSI_SEMANTIC_DEPTH 0 +#define TGSI_SEMANTIC_COLOR 1 + +struct tgsi_declaration_semantic +{ + unsigned SemanticName : 8; /* TGSI_SEMANTIC_ */ + unsigned SemanticIndex : 16; /* UINT */ + unsigned Padding : 8; +}; + #define TGSI_IMM_FLOAT32 0 struct tgsi_immediate @@ -1261,16 +1272,16 @@ struct tgsi_instruction_ext_predicate struct tgsi_src_register { - unsigned File : 4; /* TGSI_FILE_ */ - unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned Negate : 1; /* BOOL */ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - int Index : 16; /* SINT */ - unsigned Extended : 1; /* BOOL */ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned Negate : 1; /* BOOL */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ }; /* @@ -1364,22 +1375,22 @@ struct tgsi_src_register_ext_mod struct tgsi_dimension { - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - unsigned Padding : 13; - int Index : 16; /* SINT */ - unsigned Extended : 1; /* BOOL */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + unsigned Padding : 13; + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ }; struct tgsi_dst_register { - unsigned File : 4; /* TGSI_FILE_ */ - unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - int Index : 16; /* SINT */ - unsigned Padding : 5; - unsigned Extended : 1; /* BOOL */ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Padding : 5; + unsigned Extended : 1; /* BOOL */ }; /* diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 690339bc66..609d329242 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -464,25 +464,39 @@ compile_instruction( } static struct tgsi_full_declaration -make_declaration( - GLuint file, +make_frag_input_decl( GLuint first, GLuint last, - GLboolean do_interpolate, GLuint interpolate ) { struct tgsi_full_declaration decl; decl = tgsi_default_full_declaration(); - decl.Declaration.File = file; + decl.Declaration.File = TGSI_FILE_INPUT; decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.Declaration.Interpolate = 1; decl.u.DeclarationRange.First = first; decl.u.DeclarationRange.Last = last; + decl.Interpolation.Interpolate = interpolate; - if( do_interpolate ) { - decl.Declaration.Interpolate = 1; - decl.Interpolation.Interpolate = interpolate; - } + return decl; +} + +static struct tgsi_full_declaration +make_frag_output_decl( + GLuint index, + GLuint semantic_name ) +{ + struct tgsi_full_declaration decl; + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.Declaration.Semantic = 1; + decl.u.DeclarationRange.First = index; + decl.u.DeclarationRange.Last = index; + decl.Semantic.SemanticName = semantic_name; + decl.Semantic.SemanticIndex = 0; return decl; } @@ -518,11 +532,9 @@ tgsi_mesa_compile_fp_program( /* * Declare input attributes. Note that we do not interpolate fragment position. */ - fulldecl = make_declaration( - TGSI_FILE_INPUT, + fulldecl = make_frag_input_decl( 0, 0, - GL_TRUE, TGSI_INTERPOLATE_CONSTANT ); ti += tgsi_build_full_declaration( &fulldecl, @@ -537,11 +549,9 @@ tgsi_mesa_compile_fp_program( } } if( count > 0 ) { - fulldecl = make_declaration( - TGSI_FILE_INPUT, + fulldecl = make_frag_input_decl( 1, - count + 1, - GL_TRUE, + 1 + count - 1, TGSI_INTERPOLATE_LINEAR ); ti += tgsi_build_full_declaration( &fulldecl, @@ -557,12 +567,9 @@ tgsi_mesa_compile_fp_program( program->Base.OutputsWritten == (program->Base.OutputsWritten & ((1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR))) ); - fulldecl = make_declaration( - TGSI_FILE_OUTPUT, - 0, + fulldecl = make_frag_output_decl( 0, - GL_FALSE, - 0 ); + TGSI_SEMANTIC_DEPTH ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -570,12 +577,9 @@ tgsi_mesa_compile_fp_program( maxTokens - ti ); if( program->Base.OutputsWritten & (1 << FRAG_RESULT_COLR) ) { - fulldecl = make_declaration( - TGSI_FILE_OUTPUT, - 1, + fulldecl = make_frag_output_decl( 1, - GL_FALSE, - 0 ); + TGSI_SEMANTIC_COLOR ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], -- cgit v1.2.3 From 058b978a5ae2a56e09fed6335b686c654444f4ac Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 15 Aug 2007 18:16:11 +0100 Subject: Add UsageMask to DECLARATION in TGSI. Interpolate FS attributes in the shader. Do not copy WPOS in FS. --- src/mesa/pipe/softpipe/sp_headers.h | 21 +--- src/mesa/pipe/softpipe/sp_prim_setup.c | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 172 +++++------------------------- src/mesa/pipe/softpipe/sp_state_derived.c | 4 +- src/mesa/pipe/tgsi/core/tgsi_build.c | 4 + src/mesa/pipe/tgsi/core/tgsi_build.h | 1 + src/mesa/pipe/tgsi/core/tgsi_dump.c | 31 ++++++ src/mesa/pipe/tgsi/core/tgsi_exec.c | 127 +++++++++++++++++++++- src/mesa/pipe/tgsi/core/tgsi_exec.h | 52 +++++---- src/mesa/pipe/tgsi/core/tgsi_token.h | 37 +++---- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 76 ++++++------- 11 files changed, 276 insertions(+), 251 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index b7f46cb658..cc8294b18e 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -31,6 +31,7 @@ #ifndef SP_HEADERS_H #define SP_HEADERS_H +#include "../tgsi/core/tgsi_core.h" #define PRIM_POINT 1 #define PRIM_LINE 2 @@ -44,7 +45,6 @@ #define QUAD_BOTTOM_RIGHT 1 #define QUAD_TOP_LEFT 2 #define QUAD_TOP_RIGHT 3 -#define QUAD_SIZE (2*2) #define MASK_BOTTOM_LEFT 0x1 #define MASK_BOTTOM_RIGHT 0x2 @@ -53,17 +53,6 @@ #define MASK_ALL 0xf -#define NUM_CHANNELS 4 /* avoid confusion between 4 pixels and 4 channels */ - - -struct setup_coefficient { - float a0[NUM_CHANNELS]; /* in an xyzw layout */ - float dadx[NUM_CHANNELS]; - float dady[NUM_CHANNELS]; -}; - - - /** * Encodes everything we need to know about a 2x2 pixel block. Uses * "Channel-Serial" or "SoA" layout. @@ -76,17 +65,13 @@ struct quad_header { unsigned prim:2; /**< PRIM_POINT, LINE, TRI */ struct { - float color[4][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ + float color[NUM_CHANNELS][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ float depth[QUAD_SIZE]; } outputs; float coverage[QUAD_SIZE]; /** fragment coverage for antialiasing */ - const struct setup_coefficient *coef; - - const enum interp_mode *interp; /* XXX: this information should be - * encoded in fragment program DECL - * statements. */ + const struct tgsi_interp_coef *coef; unsigned nr_attrs; }; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 45d09860c3..71ef798cbb 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -80,7 +80,7 @@ struct setup_stage { float oneoverarea; - struct setup_coefficient coef[FRAG_ATTRIB_MAX]; + struct tgsi_interp_coef coef[FRAG_ATTRIB_MAX]; struct quad_header quad; struct { diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index bffb03a6bc..0a345bbf11 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -33,15 +33,12 @@ */ #include "pipe/p_util.h" -#include "tgsi/core/tgsi_core.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" #include "sp_tex_sample.h" -#include "main/mtypes.h" - #if defined __GNUC__ #define USE_ALIGNED_ATTRIBS 1 @@ -66,157 +63,33 @@ quad_shade_stage(struct quad_stage *qs) return (struct quad_shade_stage *) qs; } - - -struct exec_machine { - const struct setup_coefficient *coef; /**< will point to quad->coef */ - float attr[PIPE_ATTRIB_MAX][NUM_CHANNELS][QUAD_SIZE] ALIGN16_SUFFIX; -}; - - -/** - * Compute quad's attributes values, as constants (GL_FLAT shading). - */ -static INLINE void cinterp( struct exec_machine *exec, - unsigned attrib, - unsigned i ) -{ - unsigned j; - - for (j = 0; j < QUAD_SIZE; j++) { - exec->attr[attrib][i][j] = exec->coef[attrib].a0[i]; - } -} - - -/** - * Compute quad's attribute values by linear interpolation. - * - * Push into the fp: - * - * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx - * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy - */ -static INLINE void linterp( struct exec_machine *exec, - unsigned attrib, - unsigned i ) -{ - unsigned j; - - for (j = 0; j < QUAD_SIZE; j++) { - const float x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const float y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; - exec->attr[attrib][i][j] = (exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * x + - exec->coef[attrib].dady[i] * y); - } -} - - -/** - * Compute quad's attribute values by linear interpolation with - * perspective correction. - * - * Push into the fp: - * - * INPUT[attr] = MAD COEF_DADX[attr], INPUT_WPOS.xxxx, COEF_A0[attr] - * INPUT[attr] = MAD COEF_DADY[attr], INPUT_WPOS.yyyy, INPUT[attr] - * TMP = RCP INPUT_WPOS.w - * INPUT[attr] = MUL INPUT[attr], TMP.xxxx - * - */ -static INLINE void pinterp( struct exec_machine *exec, - unsigned attrib, - unsigned i ) -{ - unsigned j; - - for (j = 0; j < QUAD_SIZE; j++) { - const float x = exec->attr[FRAG_ATTRIB_WPOS][0][j]; - const float y = exec->attr[FRAG_ATTRIB_WPOS][1][j]; - /* FRAG_ATTRIB_WPOS.w here is really 1/w */ - const float w = 1.0 / exec->attr[FRAG_ATTRIB_WPOS][3][j]; - exec->attr[attrib][i][j] = ((exec->coef[attrib].a0[i] + - exec->coef[attrib].dadx[i] * x + - exec->coef[attrib].dady[i] * y) * w); - } -} - - /* This should be done by the fragment shader execution unit (code * generated from the decl instructions). Do it here for now. */ static void -shade_quad( struct quad_stage *qs, struct quad_header *quad ) +shade_quad( + struct quad_stage *qs, + struct quad_header *quad ) { - struct quad_shade_stage *qss = quad_shade_stage(qs); + struct quad_shade_stage *qss = quad_shade_stage( qs ); struct softpipe_context *softpipe = qs->softpipe; - struct exec_machine exec; const float fx = quad->x0; const float fy = quad->y0; unsigned attr, i; struct tgsi_exec_machine machine; #if USE_ALIGNED_ATTRIBS - struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX] ALIGN16_SUFFIX; + struct tgsi_exec_vector inputs[PIPE_ATTRIB_MAX] ALIGN16_SUFFIX; + struct tgsi_exec_vector outputs[PIPE_ATTRIB_MAX] ALIGN16_SUFFIX; #else - struct tgsi_exec_vector inputs[FRAG_ATTRIB_MAX + 1]; - struct tgsi_exec_vector outputs[FRAG_ATTRIB_MAX + 1]; + struct tgsi_exec_vector inputs[PIPE_ATTRIB_MAX + 1]; + struct tgsi_exec_vector outputs[PIPE_ATTRIB_MAX + 1]; #endif - exec.coef = quad->coef; - - /* Position: - */ - exec.attr[FRAG_ATTRIB_WPOS][0][0] = fx; - exec.attr[FRAG_ATTRIB_WPOS][0][1] = fx + 1.0; - exec.attr[FRAG_ATTRIB_WPOS][0][2] = fx; - exec.attr[FRAG_ATTRIB_WPOS][0][3] = fx + 1.0; - - exec.attr[FRAG_ATTRIB_WPOS][1][0] = fy; - exec.attr[FRAG_ATTRIB_WPOS][1][1] = fy; - exec.attr[FRAG_ATTRIB_WPOS][1][2] = fy + 1.0; - exec.attr[FRAG_ATTRIB_WPOS][1][3] = fy + 1.0; - - /* Z and W are done by linear interpolation */ - if (softpipe->need_z) { - linterp(&exec, 0, 2); /* attr[0].z */ - } - - if (softpipe->need_w) { - linterp(&exec, 0, 3); /* attr[0].w */ - /*invert(&exec, 0, 3);*/ - } - - /* Interpolate all the remaining attributes. This will get pushed - * into the fragment program's responsibilities at some point. - * Start at 1 to skip fragment position attribute (computed above). - */ - for (attr = 1; attr < quad->nr_attrs; attr++) { - switch (softpipe->interp[attr]) { - case INTERP_CONSTANT: - for (i = 0; i < NUM_CHANNELS; i++) - cinterp(&exec, attr, i); - break; - - case INTERP_LINEAR: - for (i = 0; i < NUM_CHANNELS; i++) - linterp(&exec, attr, i); - break; - - case INTERP_PERSPECTIVE: - for (i = 0; i < NUM_CHANNELS; i++) - pinterp(&exec, attr, i); - break; - } - } - #ifdef DEBUG memset( &machine, 0, sizeof( machine ) ); #endif - assert( sizeof( struct tgsi_exec_vector ) == sizeof( exec.attr[0] ) ); - /* init machine state */ tgsi_exec_machine_init( &machine, @@ -228,33 +101,40 @@ shade_quad( struct quad_stage *qs, struct quad_header *quad ) machine.Consts = softpipe->fs.constants->constant; #if USE_ALIGNED_ATTRIBS - machine.Inputs = (struct tgsi_exec_vector *) exec.attr; + machine.Inputs = inputs; machine.Outputs = outputs; #else machine.Inputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); machine.Outputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs ); - - memcpy( - machine.Inputs, - exec.attr, - softpipe->nr_attrs * sizeof( struct tgsi_exec_vector ) ); #endif + machine.InterpCoefs = quad->coef; + + machine.Inputs[0].xyzw[0].f[0] = fx; + machine.Inputs[0].xyzw[0].f[1] = fx + 1.0; + machine.Inputs[0].xyzw[0].f[2] = fx; + machine.Inputs[0].xyzw[0].f[3] = fx + 1.0; + + machine.Inputs[0].xyzw[1].f[0] = fy; + machine.Inputs[0].xyzw[1].f[1] = fy; + machine.Inputs[0].xyzw[1].f[2] = fy + 1.0; + machine.Inputs[0].xyzw[1].f[3] = fy + 1.0; + /* run shader */ tgsi_exec_machine_run( &machine ); /* store result color */ memcpy( quad->outputs.color, - &machine.Outputs[FRAG_ATTRIB_COL0].xyzw[0].f[0], + &machine.Outputs[1].xyzw[0].f[0], sizeof( quad->outputs.color ) ); if( softpipe->need_z ) { /* XXX temporary */ - quad->outputs.depth[0] = exec.attr[0][2][0]; - quad->outputs.depth[1] = exec.attr[0][2][1]; - quad->outputs.depth[2] = exec.attr[0][2][2]; - quad->outputs.depth[3] = exec.attr[0][2][3]; + memcpy( + quad->outputs.depth, + &machine.Outputs[0].xyzw[2], + sizeof( quad->outputs.depth ) ); } /* shader may cull fragments */ diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 0a2cfbb7d1..cd67d1c46f 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -87,7 +87,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) * fragment position (XYZW). */ if (softpipe->depth_test.enabled || - (inputsRead & FRAG_ATTRIB_WPOS)) + (inputsRead & (1 << FRAG_ATTRIB_WPOS))) softpipe->need_z = TRUE; else softpipe->need_z = FALSE; @@ -95,7 +95,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* Need W if we do any perspective-corrected interpolation or the * fragment program uses the fragment position. */ - if (inputsRead & FRAG_ATTRIB_WPOS) + if (inputsRead & (1 << FRAG_ATTRIB_WPOS)) softpipe->need_w = TRUE; else softpipe->need_w = FALSE; diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.c b/src/mesa/pipe/tgsi/core/tgsi_build.c index d9c9a45532..d6f8af656a 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_build.c +++ b/src/mesa/pipe/tgsi/core/tgsi_build.c @@ -88,6 +88,7 @@ tgsi_default_declaration( void ) declaration.Size = 1; declaration.File = TGSI_FILE_NULL; declaration.Declare = TGSI_DECLARE_RANGE; + declaration.UsageMask = TGSI_WRITEMASK_XYZW; declaration.Interpolate = 0; declaration.Semantic = 0; declaration.Padding = 0; @@ -100,6 +101,7 @@ struct tgsi_declaration tgsi_build_declaration( unsigned file, unsigned declare, + unsigned usage_mask, unsigned interpolate, unsigned semantic, struct tgsi_header *header ) @@ -112,6 +114,7 @@ tgsi_build_declaration( declaration = tgsi_default_declaration(); declaration.File = file; declaration.Declare = declare; + declaration.UsageMask = usage_mask; declaration.Interpolate = interpolate; declaration.Semantic = semantic; @@ -162,6 +165,7 @@ tgsi_build_full_declaration( *declaration = tgsi_build_declaration( full_decl->Declaration.File, full_decl->Declaration.Declare, + full_decl->Declaration.UsageMask, full_decl->Declaration.Interpolate, full_decl->Declaration.Semantic, header ); diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.h b/src/mesa/pipe/tgsi/core/tgsi_build.h index b3eb0715ee..116c78abf3 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_build.h +++ b/src/mesa/pipe/tgsi/core/tgsi_build.h @@ -38,6 +38,7 @@ struct tgsi_declaration tgsi_build_declaration( unsigned file, unsigned declare, + unsigned usage_mask, unsigned interpolate, unsigned semantic, struct tgsi_header *header ); diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c index 5a9f92004e..622d617e29 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/core/tgsi_dump.c @@ -633,6 +633,22 @@ dump_declaration_short( assert( 0 ); } + if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + if( decl->Declaration.Interpolate ) { TXT( ", " ); ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); @@ -659,6 +675,21 @@ dump_declaration_verbose( ENM( decl->Declaration.File, TGSI_FILES ); TXT( "\nDeclare : " ); ENM( decl->Declaration.Declare, TGSI_DECLARES ); + if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { + TXT( "\nUsageMask : " ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'X' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'Y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'Z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'W' ); + } + } if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { TXT( "\nInterpolate: " ); UID( decl->Declaration.Interpolate ); diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 0a34754c40..0ffed20066 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -62,7 +62,7 @@ void tgsi_exec_machine_init( struct tgsi_exec_machine *mach, - struct tgsi_token *tokens, + const struct tgsi_token *tokens, GLuint numSamplers, struct tgsi_sampler *samplers) { @@ -1063,7 +1063,131 @@ fetch_texel( struct tgsi_sampler *sampler, } } +static void +constant_interpolation( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; + } +} + +static void +linear_interpolation( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = mach->Inputs[0].xyzw[0].f[i]; + const float y = mach->Inputs[0].xyzw[1].f[i]; + + mach->Inputs[attrib].xyzw[chan].f[i] = + mach->InterpCoefs[attrib].a0[chan] + + mach->InterpCoefs[attrib].dadx[chan] * x + + mach->InterpCoefs[attrib].dady[chan] * y; + } +} + +static void +perspective_interpolation( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = mach->Inputs[0].xyzw[0].f[i]; + const float y = mach->Inputs[0].xyzw[1].f[i]; + // WPOS.w here is really 1/w + const float w = 1.0f / mach->Inputs[0].xyzw[3].f[i]; + + mach->Inputs[attrib].xyzw[chan].f[i] = + (mach->InterpCoefs[attrib].a0[chan] + + mach->InterpCoefs[attrib].dadx[chan] * x + + mach->InterpCoefs[attrib].dady[chan] * y) * w; + } +} + +typedef void (* interpolation_func)( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ); + +static void +exec_declaration( + struct tgsi_exec_machine *mach, + const struct tgsi_full_declaration *decl ) +{ + if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask, i, j; + interpolation_func interp; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if( first == 0 ) { + mask &= ~TGSI_WRITEMASK_XY; + if( mask == TGSI_WRITEMASK_NONE ) { + first++; + if( first > last ) { + return; + } + } + } + + switch( decl->Interpolation.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + interp = constant_interpolation; + break; + + case TGSI_INTERPOLATE_LINEAR: + interp = linear_interpolation; + break; + case TGSI_INTERPOLATE_PERSPECTIVE: + interp = perspective_interpolation; + break; + + default: + assert( 0 ); + } + + if( mask == TGSI_WRITEMASK_XYZW ) { + unsigned i, j; + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + interp( mach, i, j ); + } + } + } + else { + unsigned i, j; + + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + interp( mach, i, j ); + } + } + } + } + } + } +} static void exec_instruction( @@ -2161,6 +2285,7 @@ tgsi_exec_machine_run2( tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: + exec_declaration( mach, &parse.FullToken.FullDeclaration ); break; case TGSI_TOKEN_TYPE_IMMEDIATE: break; diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h index eed2207d7d..e5e8c3608e 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.h @@ -11,23 +11,27 @@ extern "C" { #endif // defined __cplusplus +#define NUM_CHANNELS 4 /* R,G,B,A */ +#define QUAD_SIZE 4 /* 4 pixel/quad */ + union tgsi_exec_channel { - float f[4]; - int i[4]; - unsigned u[4]; + float f[QUAD_SIZE]; + int i[QUAD_SIZE]; + unsigned u[QUAD_SIZE]; }; struct tgsi_exec_vector { - union tgsi_exec_channel xyzw[4]; + union tgsi_exec_channel xyzw[NUM_CHANNELS]; }; - -#define NUM_CHANNELS 4 /* R,G,B,A */ -#ifndef QUAD_SIZE -#define QUAD_SIZE 4 /* 4 pixel/quad */ -#endif +struct tgsi_interp_coef +{ + float a0[NUM_CHANNELS]; /* in an xyzw layout */ + float dadx[NUM_CHANNELS]; + float dady[NUM_CHANNELS]; +}; #define TEX_CACHE_TILE_SIZE 8 #define TEX_CACHE_NUM_ENTRIES 8 @@ -55,8 +59,8 @@ struct tgsi_sampler struct tgsi_exec_labels { - unsigned labels[128][2]; - unsigned count; + unsigned labels[128][2]; + unsigned count; }; #define TGSI_EXEC_TEMP_00000000_I 32 @@ -109,15 +113,15 @@ struct tgsi_exec_cond_state { struct tgsi_exec_cond_regs IfPortion; struct tgsi_exec_cond_regs ElsePortion; - unsigned Condition; - boolean WasElse; + unsigned Condition; + boolean WasElse; }; /* XXX: This is temporary */ struct tgsi_exec_cond_stack { struct tgsi_exec_cond_state States[8]; - unsigned Index; /* into States[] */ + unsigned Index; /* into States[] */ }; struct tgsi_exec_machine @@ -138,15 +142,19 @@ struct tgsi_exec_machine struct tgsi_sampler *Samplers; - float Imms[256][4]; - unsigned ImmLimit; - float (*Consts)[4]; - const struct tgsi_exec_vector *Inputs; + float Imms[256][4]; + unsigned ImmLimit; + float (*Consts)[4]; + struct tgsi_exec_vector *Inputs; struct tgsi_exec_vector *Outputs; - struct tgsi_token *Tokens; - unsigned Processor; + const struct tgsi_token *Tokens; + unsigned Processor; + + /* GEOMETRY processor only. */ + unsigned *Primitives; - unsigned *Primitives; + /* FRAGMENT processor only. */ + const struct tgsi_interp_coef *InterpCoefs; struct tgsi_exec_cond_stack CondStack; #if XXX_SSE @@ -157,7 +165,7 @@ struct tgsi_exec_machine void tgsi_exec_machine_init( struct tgsi_exec_machine *mach, - struct tgsi_token *tokens, + const struct tgsi_token *tokens, unsigned numSamplers, struct tgsi_sampler *samplers); diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index f72c7dc4be..3ed341fb7b 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -52,15 +52,33 @@ struct tgsi_token #define TGSI_DECLARE_RANGE 0 #define TGSI_DECLARE_MASK 1 +#define TGSI_WRITEMASK_NONE 0x00 +#define TGSI_WRITEMASK_X 0x01 +#define TGSI_WRITEMASK_Y 0x02 +#define TGSI_WRITEMASK_XY 0x03 +#define TGSI_WRITEMASK_Z 0x04 +#define TGSI_WRITEMASK_XZ 0x05 +#define TGSI_WRITEMASK_YZ 0x06 +#define TGSI_WRITEMASK_XYZ 0x07 +#define TGSI_WRITEMASK_W 0x08 +#define TGSI_WRITEMASK_XW 0x09 +#define TGSI_WRITEMASK_YW 0x0A +#define TGSI_WRITEMASK_XYW 0x0B +#define TGSI_WRITEMASK_ZW 0x0C +#define TGSI_WRITEMASK_XZW 0x0D +#define TGSI_WRITEMASK_YZW 0x0E +#define TGSI_WRITEMASK_XYZW 0x0F + struct tgsi_declaration { unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ unsigned Size : 8; /* UINT */ unsigned File : 4; /* TGSI_FILE_ */ unsigned Declare : 4; /* TGSI_DECLARE_ */ + unsigned UsageMask : 4; /* TGSI_WRITEMASK_ */ unsigned Interpolate : 1; /* BOOL */ unsigned Semantic : 1; /* BOOL */ - unsigned Padding : 9; + unsigned Padding : 5; unsigned Extended : 1; /* BOOL */ }; @@ -1226,23 +1244,6 @@ struct tgsi_instruction_ext_texture unsigned Extended : 1; /* BOOL */ }; -#define TGSI_WRITEMASK_NONE 0x00 -#define TGSI_WRITEMASK_X 0x01 -#define TGSI_WRITEMASK_Y 0x02 -#define TGSI_WRITEMASK_XY 0x03 -#define TGSI_WRITEMASK_Z 0x04 -#define TGSI_WRITEMASK_XZ 0x05 -#define TGSI_WRITEMASK_YZ 0x06 -#define TGSI_WRITEMASK_XYZ 0x07 -#define TGSI_WRITEMASK_W 0x08 -#define TGSI_WRITEMASK_XW 0x09 -#define TGSI_WRITEMASK_YW 0x0A -#define TGSI_WRITEMASK_XYW 0x0B -#define TGSI_WRITEMASK_ZW 0x0C -#define TGSI_WRITEMASK_XZW 0x0D -#define TGSI_WRITEMASK_YZW 0x0E -#define TGSI_WRITEMASK_XYZW 0x0F - struct tgsi_instruction_ext_predicate { unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 609d329242..2ac04e8d56 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -467,13 +467,15 @@ static struct tgsi_full_declaration make_frag_input_decl( GLuint first, GLuint last, - GLuint interpolate ) + GLuint interpolate, + GLuint usage_mask ) { struct tgsi_full_declaration decl; decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_INPUT; decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.Declaration.UsageMask = usage_mask; decl.Declaration.Interpolate = 1; decl.u.DeclarationRange.First = first; decl.u.DeclarationRange.Last = last; @@ -485,13 +487,15 @@ make_frag_input_decl( static struct tgsi_full_declaration make_frag_output_decl( GLuint index, - GLuint semantic_name ) + GLuint semantic_name, + GLuint usage_mask ) { struct tgsi_full_declaration decl; decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_OUTPUT; decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.Declaration.UsageMask = usage_mask; decl.Declaration.Semantic = 1; decl.u.DeclarationRange.First = index; decl.u.DeclarationRange.Last = index; @@ -514,6 +518,7 @@ tgsi_mesa_compile_fp_program( struct tgsi_full_dst_register *fulldst; struct tgsi_full_src_register *fullsrc; GLuint inputs_read; + GLboolean reads_wpos; GLuint preamble_size = 0; *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -523,19 +528,33 @@ tgsi_mesa_compile_fp_program( ti = 2; - /* - * Input 0 is always read, at least implicitly by the MOV instruction generated - * below, so mark it as used. - */ - inputs_read = program->Base.InputsRead | 1; + reads_wpos = program->Base.InputsRead & (1 << FRAG_ATTRIB_WPOS); + inputs_read = program->Base.InputsRead | (1 << FRAG_ATTRIB_WPOS); /* * Declare input attributes. Note that we do not interpolate fragment position. */ + + /* Fragment position. */ + if( reads_wpos ) { + fulldecl = make_frag_input_decl( + 0, + 0, + TGSI_INTERPOLATE_CONSTANT, + TGSI_WRITEMASK_XY ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + + /* Fragment zw. */ fulldecl = make_frag_input_decl( 0, 0, - TGSI_INTERPOLATE_CONSTANT ); + TGSI_INTERPOLATE_LINEAR, + reads_wpos ? TGSI_WRITEMASK_ZW : TGSI_WRITEMASK_Z ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -552,7 +571,8 @@ tgsi_mesa_compile_fp_program( fulldecl = make_frag_input_decl( 1, 1 + count - 1, - TGSI_INTERPOLATE_LINEAR ); + TGSI_INTERPOLATE_LINEAR, + TGSI_WRITEMASK_XYZW ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -569,7 +589,8 @@ tgsi_mesa_compile_fp_program( fulldecl = make_frag_output_decl( 0, - TGSI_SEMANTIC_DEPTH ); + TGSI_SEMANTIC_DEPTH, + TGSI_WRITEMASK_Z ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -579,7 +600,8 @@ tgsi_mesa_compile_fp_program( if( program->Base.OutputsWritten & (1 << FRAG_RESULT_COLR) ) { fulldecl = make_frag_output_decl( 1, - TGSI_SEMANTIC_COLOR ); + TGSI_SEMANTIC_COLOR, + TGSI_WRITEMASK_XYZW ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -587,38 +609,6 @@ tgsi_mesa_compile_fp_program( maxTokens - ti ); } - /* - * Copy input fragment xyz to output xyz. - * If the shader writes depth, do not copy the z component. - */ - - fullinst = tgsi_default_full_instruction(); - - fullinst.Instruction.Opcode = TGSI_OPCODE_MOV; - fullinst.Instruction.NumDstRegs = 1; - fullinst.Instruction.NumSrcRegs = 1; - - fulldst = &fullinst.FullDstRegisters[0]; - fulldst->DstRegister.File = TGSI_FILE_OUTPUT; - fulldst->DstRegister.Index = 0; - if( program->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR) ) { - fulldst->DstRegister.WriteMask = TGSI_WRITEMASK_XY; - } - else { - fulldst->DstRegister.WriteMask = TGSI_WRITEMASK_XYZ; - } - - fullsrc = &fullinst.FullSrcRegisters[0]; - fullsrc->SrcRegister.File = TGSI_FILE_INPUT; - fullsrc->SrcRegister.Index = 0; - - ti += tgsi_build_full_instruction( - &fullinst, - &tokens[ti], - header, - maxTokens - ti ); - preamble_size++; - for( i = 0; i < program->Base.NumInstructions; i++ ) { if( compile_instruction( &program->Base.Instructions[i], -- cgit v1.2.3 From 24083f4e4702f4e9b8cc1679139d45ee771a0f16 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 18:53:00 -0600 Subject: comment --- src/mesa/pipe/draw/draw_vb.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index dfd1f22eb3..23a5e3fc3b 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -489,6 +489,9 @@ draw_release_vertices( struct draw_context *draw ) } +/** + * Note: this must match struct vertex_header's layout (I think). + */ struct header_dword { GLuint clipmask:12; GLuint edgeflag:1; @@ -802,7 +805,7 @@ void draw_set_vertex_attributes( struct draw_context *draw, draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); } - + void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ) { -- cgit v1.2.3 From 293cc252616543d36aeb8a432441006bf65e3330 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 18:53:37 -0600 Subject: added draw_set_vertex_array_info() --- src/mesa/pipe/draw/draw_context.c | 10 ++++++++++ src/mesa/pipe/draw/draw_context.h | 10 ++++++++++ 2 files changed, 20 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 0707c03e5f..f2ee6c52e0 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -192,3 +192,13 @@ void draw_set_viewport_state( struct draw_context *draw, * Full pipe will have vertex shader, vertex fetch of its own. */ } + + +void draw_set_vertex_array_info(struct draw_context *draw, + const struct pipe_vertex_buffer *buffers, + const struct pipe_vertex_element *elements) +{ + draw->vertex_buffer = buffers; + draw->vertex_element = elements; +} + diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index be0a18d6d7..b9cac7c6da 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -67,6 +67,16 @@ void draw_set_vertex_attributes( struct draw_context *draw, const GLuint *attrs, GLuint nr_attrs ); +/* XXX temporary */ +void draw_set_vertex_attributes2( struct draw_context *draw, + const GLuint *attrs, + GLuint nr_attrs ); + +void draw_set_vertex_array_info(struct draw_context *draw, + const struct pipe_vertex_buffer *buffers, + const struct pipe_vertex_element *elements); + +/* XXX temporary */ void draw_vb(struct draw_context *draw, struct vertex_buffer *VB ); -- cgit v1.2.3 From 3e2be1ce4ab0a2d61519f47b1eb78626c710b378 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 18:54:05 -0600 Subject: added vertex array info fields --- src/mesa/pipe/draw/draw_private.h | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 597393afdc..76ca80dab0 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -137,6 +137,10 @@ struct draw_context struct pipe_setup_state setup; struct pipe_viewport_state viewport; + const struct pipe_vertex_buffer *vertex_buffer; /**< note: pointer */ + const struct pipe_vertex_element *vertex_element; /**< note: pointer */ + + /* Clip derived state: */ GLfloat plane[12][4]; @@ -196,6 +200,9 @@ struct draw_context /* Helper for tnl: */ GLvector4f header; + + /* helper for sp_draw_arrays.c - temporary? */ + void *mapped_vbuffer; }; -- cgit v1.2.3 From 181cf1d52a9fe0ad8be9d48aa69775f575336d3d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 19:00:06 -0600 Subject: convert vbo->draw_prims() to pipe->draw_arrays(): works in very limited cases, disabled for now --- src/mesa/state_tracker/st_draw.c | 81 ++++++++++++++++++++++++++++++++++------ 1 file changed, 69 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 002565a658..0b13432256 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -44,6 +44,15 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "vbo/vbo_context.h" + +/* + * Enabling this causes the VBO module to call draw_vbo() below, + * bypassing the T&L module. This only works with VBO-based demos, + * such as progs/test/bufferobj.c + */ +#define USE_NEW_DRAW 0 + /* * TNL stage which feeds into the above. @@ -95,6 +104,31 @@ static const struct tnl_pipeline_stage *st_pipeline[] = { +static GLuint +pipe_vertex_format(GLenum format, GLuint size) +{ + static const GLuint float_fmts[4] = { + PIPE_FORMAT_R32_FLOAT, + PIPE_FORMAT_R32G32_FLOAT, + PIPE_FORMAT_R32G32B32_FLOAT, + PIPE_FORMAT_R32G32B32A32_FLOAT, + }; + + assert(format >= GL_BYTE); + assert(format <= GL_DOUBLE); + assert(size >= 1); + assert(size <= 4); + + switch (format) { + case GL_FLOAT: + return float_fmts[size - 1]; + default: + assert(0); + } +} + + + /** * This function gets plugged into the VBO module and is called when * we have something to render. @@ -110,35 +144,48 @@ draw_vbo(GLcontext *ctx, GLuint max_index) { struct pipe_context *pipe = ctx->st->pipe; - GLuint attr; + GLuint attr, i; /* tell pipe about the vertex array element/attributes */ for (attr = 0; attr < 16; attr++) { struct gl_buffer_object *bufobj = arrays[attr]->BufferObj; + struct pipe_vertex_buffer vbuffer; + struct pipe_vertex_element velement; + if (bufobj && bufobj->Name) { struct st_buffer_object *stobj = st_buffer_object(bufobj); - struct pipe_buffer_handle *buffer = stobj->buffer; - GLenum type = arrays[attr]->Type; - GLint size = arrays[attr]->Size; - struct pipe_vertex_buffer vbuffer; - struct pipe_vertex_element velement; - vbuffer.pitch = 0; + assert(stobj->buffer); + + vbuffer.pitch = arrays[attr]->StrideB; /* in bytes */ vbuffer.max_index = 0; - vbuffer.buffer = NULL; + vbuffer.buffer = stobj->buffer; vbuffer.buffer_offset = 0; - velement.src_offset = 0; + /* Recall that for VBOs, the gl_client_array->Ptr field is + * really an offset from the start of the VBO, not a pointer. + */ + velement.src_offset = (unsigned) arrays[attr]->Ptr; velement.vertex_buffer_index = attr; velement.dst_offset = 0; - velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + velement.src_format = pipe_vertex_format(arrays[attr]->Type, + arrays[attr]->Size); + } + else { + /* vertex attribute data is not an a real buffer! */ + /* XXX we'll want to handle that someday... */ - pipe->set_vertex_buffer(pipe, attr, &vbuffer); - pipe->set_vertex_element(pipe, attr, &velement); + vbuffer.buffer = NULL; } + + pipe->set_vertex_buffer(pipe, attr, &vbuffer); + pipe->set_vertex_element(pipe, attr, &velement); } /* do actual drawing */ + for (i = 0; i < nr_prims; i++) { + pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count); + } } @@ -151,8 +198,18 @@ void st_init_draw( struct st_context *st ) { GLcontext *ctx = st->ctx; +#if USE_NEW_DRAW + struct vbo_context *vbo = (struct vbo_context *) ctx->swtnl_im; + + assert(vbo); + assert(vbo->draw_prims); + vbo->draw_prims = draw_vbo; + +#else _tnl_destroy_pipeline( ctx ); _tnl_install_pipeline( ctx, st_pipeline ); +#endif + } -- cgit v1.2.3 From ab0b040b6023f2e6c0eb7c12f085b4f24a847851 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 19:00:32 -0600 Subject: added new draw_arrays() --- src/mesa/pipe/p_context.h | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index ae2cc25fe3..6679aae089 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -68,14 +68,20 @@ struct pipe_context { /* * Drawing */ + /* XXX this is temporary */ void (*draw_vb)( struct pipe_context *pipe, struct vertex_buffer *VB ); + /* XXX this is temporary */ void (*draw_vertices)( struct pipe_context *pipe, unsigned mode, unsigned numVertex, const float *verts, unsigned numAttribs, const unsigned attribs[]); + /** this is basically what we want */ + void (*draw_arrays)( struct pipe_context *pipe, + unsigned mode, unsigned start, unsigned count); + /** Clear a surface to given value (no scissor; clear whole surface) */ void (*clear)(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue); -- cgit v1.2.3 From 2de99586246d758da14e281718d85c499d1969d6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 19:02:37 -0600 Subject: softpipe_draw_arrays() function Also includes: Temporary stand-in for vertex program execution. draw_prim() code post-transform vertex cache --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 663 ++++++++++++++++++++++++++++++++ 1 file changed, 663 insertions(+) create mode 100644 src/mesa/pipe/softpipe/sp_draw_arrays.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c new file mode 100644 index 0000000000..c466884fdd --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -0,0 +1,663 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + * Keith Whitwell + */ + + +#include "main/mtypes.h" +#include "main/context.h" + +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +#include "sp_context.h" +#include "sp_state.h" + +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_context.h" + + + +#define RP_NONE 0 +#define RP_POINT 1 +#define RP_LINE 2 +#define RP_TRI 3 + +static unsigned reduced_prim[GL_POLYGON + 1] = { + RP_POINT, + RP_LINE, + RP_LINE, + RP_LINE, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI +}; + + +/** + * Stand-in for actual vertex program execution + * XXX this will probably live in a new file, like "sp_vs.c" + * \param draw the drawing context + * \param vbuffer the mapped vertex buffer pointer + * \param elem which element of the vertex buffer to use as input + * \param vOut the output vertex + */ +static void +run_vertex_program(struct draw_context *draw, + const void *vbuffer, GLuint elem, + struct vertex_header *vOut) +{ + const float *vIn, *cIn; + const GLfloat *scale = draw->viewport.scale; + const GLfloat *trans = draw->viewport.translate; + const void *mapped = vbuffer; + + GET_CURRENT_CONTEXT(ctx); + const GLfloat *m = ctx->_ModelProjectMatrix.m; + + vIn = (const float *) ((const GLubyte *) mapped + + draw->vertex_buffer[0].buffer_offset + + draw->vertex_element[0].src_offset + + elem * draw->vertex_buffer[0].pitch); + + cIn = (const float *) ((const GLubyte *) mapped + + draw->vertex_buffer[3].buffer_offset + + draw->vertex_element[3].src_offset + + elem * draw->vertex_buffer[3].pitch); + + { + float x = vIn[0]; + float y = vIn[1]; + float z = vIn[2]; + float w = 1.0; + + vOut->clipmask = 0x0; + vOut->edgeflag = 0; + /* MVP */ + vOut->clip[0] = m[0] * x + m[4] * y + m[ 8] * z + m[12] * w; + vOut->clip[1] = m[1] * x + m[5] * y + m[ 9] * z + m[13] * w; + vOut->clip[2] = m[2] * x + m[6] * y + m[10] * z + m[14] * w; + vOut->clip[3] = m[3] * x + m[7] * y + m[11] * z + m[15] * w; + + /* divide by w */ + x = vOut->clip[0] / vOut->clip[3]; + y = vOut->clip[1] / vOut->clip[3]; + z = vOut->clip[2] / vOut->clip[3]; + w = 1.0 / vOut->clip[3]; + + /* Viewport */ + vOut->data[0][0] = scale[0] * x + trans[0]; + vOut->data[0][1] = scale[1] * y + trans[1]; + vOut->data[0][2] = scale[2] * z + trans[2]; + vOut->data[0][3] = w; + + /* color */ + vOut->data[1][0] = cIn[0]; + vOut->data[1][1] = cIn[1]; + vOut->data[1][2] = cIn[2]; + vOut->data[1][3] = 1.0; + } +} + + +static void vs_flush( struct draw_context *draw ) +{ + unsigned i; + + /* We're not really running a vertex shader yet, so flushing the vs + * queue is just a matter of building the vertices and returning. + */ + /* Actually, I'm cheating even more and pre-building them still + * with the mesa/vf module. So it's very easy... + */ + for (i = 0; i < draw->vs.queue_nr; i++) { + /* Would do the following steps here: + * + * 1) Loop over vertex element descriptors, fetch data from each + * to build the pre-tnl vertex. This might require a new struct + * to represent the pre-tnl vertex. + * + * 2) Bundle groups of upto 4 pre-tnl vertices together and pass + * to vertex shader. + * + * 3) Do any necessary unswizzling, make sure vertex headers are + * correctly populated, store resulting post-transformed + * vertices in vcache. + * + * In this version, just do the last step: + */ + const unsigned elt = draw->vs.queue[i].elt; + struct vertex_header *dest = draw->vs.queue[i].dest; + + run_vertex_program(draw, draw->mapped_vbuffer, elt, dest); + } + draw->vs.queue_nr = 0; +} + + +static void draw_flush( struct draw_context *draw ) +{ + struct draw_stage *first = draw->pipeline.first; + unsigned i; + + /* Make sure all vertices are available: + */ + vs_flush( draw ); + + + switch (draw->reduced_prim) { + case RP_TRI: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + first->reset_stipple_counter( first ); + + first->tri( first, &draw->pq.queue[i] ); + } + break; + case RP_LINE: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + first->reset_stipple_counter( first ); + + first->line( first, &draw->pq.queue[i] ); + } + break; + case RP_POINT: + first->reset_stipple_counter( first ); + for (i = 0; i < draw->pq.queue_nr; i++) + first->point( first, &draw->pq.queue[i] ); + break; + } + + draw->pq.queue_nr = 0; + draw->vcache.referenced = 0; + draw->vcache.overflow = 0; +} + + +static void draw_invalidate_vcache( struct draw_context *draw ) +{ + unsigned i; + + assert(draw->pq.queue_nr == 0); + assert(draw->vs.queue_nr == 0); + assert(draw->vcache.referenced == 0); + + for (i = 0; i < Elements( draw->vcache.idx ); i++) + draw->vcache.idx[i] = ~0; +} + + +/* Return a pointer to a freshly queued primitive header. Ensure that + * there is room in the vertex cache for a maximum of "nr_verts" new + * vertices. Flush primitive and/or vertex queues if necessary to + * make space. + */ +static struct prim_header *get_queued_prim( struct draw_context *draw, + GLuint nr_verts ) +{ + if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH || + draw->vcache.overflow + nr_verts >= VCACHE_OVERFLOW) + draw_flush( draw ); + + /* The vs queue is sized so that this can never happen: + */ + assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); + + return &draw->pq.queue[draw->pq.queue_nr++]; +} + + +/* Check if vertex is in cache, otherwise add it. It won't go through + * VS yet, not until there is a flush operation or the VS queue fills up. + */ +static struct vertex_header *get_vertex( struct draw_context *draw, + GLuint i ) +{ + unsigned slot = (i + (i>>5)) & 31; + + /* Cache miss? + */ + if (draw->vcache.idx[slot] != i) { + + /* If slot is in use, use the overflow area: + */ + if (draw->vcache.referenced & (1<vcache.overflow++; + + draw->vcache.idx[slot] = i; + + /* Add to vertex shader queue: + */ + draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; + draw->vs.queue[draw->vs.queue_nr].elt = i; + draw->vs.queue_nr++; + } + + /* Mark slot as in-use: + */ + draw->vcache.referenced |= (1<vcache.vertex[slot]; +} + + + +static void draw_set_prim( struct draw_context *draw, + GLenum prim ) +{ + if (reduced_prim[prim] != draw->reduced_prim) { + draw_flush( draw ); + draw->reduced_prim = reduced_prim[prim]; + } + + draw->prim = prim; +} + + +static void do_point( struct draw_context *draw, + GLuint i0 ) +{ + struct prim_header *prim = get_queued_prim( draw, 1 ); + + prim->reset_line_stipple = 0; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->get_vertex( draw, i0 ); +} + + +static void do_line( struct draw_context *draw, + GLboolean reset_stipple, + GLuint i0, + GLuint i1 ) +{ + struct prim_header *prim = get_queued_prim( draw, 2 ); + + prim->reset_line_stipple = reset_stipple; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->get_vertex( draw, i0 ); + prim->v[1] = draw->get_vertex( draw, i1 ); +} + +static void do_triangle( struct draw_context *draw, + GLuint i0, + GLuint i1, + GLuint i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + + prim->reset_line_stipple = 1; + prim->edgeflags = ~0; + prim->pad = 0; + prim->v[0] = draw->get_vertex( draw, i0 ); + prim->v[1] = draw->get_vertex( draw, i1 ); + prim->v[2] = draw->get_vertex( draw, i2 ); +} + +static void do_ef_triangle( struct draw_context *draw, + GLboolean reset_stipple, + GLuint ef_mask, + GLuint i0, + GLuint i1, + GLuint i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + struct vertex_header *v0 = draw->get_vertex( draw, i0 ); + struct vertex_header *v1 = draw->get_vertex( draw, i1 ); + struct vertex_header *v2 = draw->get_vertex( draw, i2 ); + + prim->reset_line_stipple = reset_stipple; + + prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | + (v1->edgeflag << 1) | + (v2->edgeflag << 2)); + prim->pad = 0; + prim->v[0] = v0; + prim->v[1] = v1; + prim->v[2] = v2; +} + + +static void do_quad( struct draw_context *draw, + unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3 ) +{ + do_ef_triangle( draw, 1, ~(1<<0), v0, v1, v3 ); + do_ef_triangle( draw, 0, ~(1<<1), v1, v2, v3 ); +} + + +static void draw_prim( struct draw_context *draw, + GLuint start, + GLuint count ) +{ + GLuint i; + +// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); + + switch (draw->prim) { + case GL_POINTS: + for (i = 0; i < count; i ++) { + do_point( draw, + start + i ); + } + break; + + case GL_LINES: + for (i = 0; i+1 < count; i += 2) { + do_line( draw, + TRUE, + start + i + 0, + start + i + 1); + } + break; + + case GL_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, /* XXX: only if vb not split */ + start + i - 1, + start + i ); + } + + do_line( draw, + 0, + start + count - 1, + start + 0 ); + } + break; + + case GL_LINE_STRIP: + if (count >= 2) { + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, + start + i - 1, + start + i ); + } + } + break; + + case GL_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + do_ef_triangle( draw, + 1, + ~0, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + break; + + case GL_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + do_triangle( draw, + start + i + 1, + start + i + 0, + start + i + 2 ); + } + else { + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + case GL_TRIANGLE_FAN: + if (count >= 3) { + for (i = 0; i+2 < count; i++) { + do_triangle( draw, + start + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + + case GL_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( draw, + start + i + 0, + start + i + 1, + start + i + 2, + start + i + 3); + } + break; + + case GL_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( draw, + start + i + 2, + start + i + 0, + start + i + 1, + start + i + 3); + } + break; + + case GL_POLYGON: + if (count >= 3) { + unsigned ef_mask = (1<<2) | (1<<0); + + for (i = 0; i+2 < count; i++) { + + if (i + 3 >= count) + ef_mask |= (1<<1); + + do_ef_triangle( draw, + i == 0, + ef_mask, + start + i + 1, + start + i + 2, + start + i + 0); + + ef_mask &= ~(1<<2); + } + } + break; + + default: + assert(0); + break; + } +} + + + +static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) +{ + switch (mode) { + case GL_POINTS: + *first = 1; + *incr = 1; + return 0; + case GL_LINES: + *first = 2; + *incr = 2; + return 0; + case GL_LINE_STRIP: + *first = 2; + *incr = 1; + return 0; + case GL_LINE_LOOP: + *first = 2; + *incr = 1; + return 1; + case GL_TRIANGLES: + *first = 3; + *incr = 3; + return 0; + case GL_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + return 0; + case GL_TRIANGLE_FAN: + case GL_POLYGON: + *first = 3; + *incr = 1; + return 1; + case GL_QUADS: + *first = 4; + *incr = 4; + return 0; + case GL_QUAD_STRIP: + *first = 4; + *incr = 2; + return 0; + default: + assert(0); + *first = 1; + *incr = 1; + return 0; + } +} + + +static GLuint trim( GLuint count, GLuint first, GLuint incr ) +{ + if (count < first) + return 0; + else + return count - (count - first) % incr; +} + + + +void +softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count) +{ + struct softpipe_context *sp = softpipe_context(pipe); + struct draw_context *draw = sp->draw; + struct pipe_buffer_handle *buf; + + softpipe_map_surfaces(sp); + + /* + * Map vertex buffers + */ + buf = sp->vertex_buffer[0].buffer; + draw->mapped_vbuffer + = pipe->winsys->buffer_map(pipe->winsys, buf, PIPE_BUFFER_FLAG_READ); + + + /* tell drawing pipeline we're beginning drawing */ + draw->pipeline.first->begin( draw->pipeline.first ); + + draw_invalidate_vcache( draw ); + +#if 0 + if (VB->Elts) + draw->get_vertex = get_uint_elt_vertex; + else +#endif + draw->get_vertex = get_vertex; + + draw_set_prim( draw, mode ); + + /* XXX draw_prim_info() and TRIM here */ + draw_prim(draw, start, count); + + /* draw any left-over buffered prims */ + draw_flush(draw); + + /* tell drawing pipeline we're done drawing */ + draw->pipeline.first->end( draw->pipeline.first ); + +#if 0 + draw->verts = NULL; + draw->in_vb = 0; + draw->elts = NULL; +#endif + + pipe->winsys->buffer_unmap(pipe->winsys, buf); + + softpipe_unmap_surfaces(sp); +} + + + +#define EMIT_ATTR( VF_ATTR, STYLE, SIZE ) \ +do { \ + if (draw->nr_attrs >= 2) \ + draw->vf_attr_to_slot[VF_ATTR] = draw->nr_attrs - 2; \ + draw->attrs[draw->nr_attrs].attrib = VF_ATTR; \ + draw->attrs[draw->nr_attrs].format = STYLE; \ + draw->nr_attrs++; \ + draw->vertex_size += SIZE; \ +} while (0) + + +void draw_set_vertex_attributes2( struct draw_context *draw, + const GLuint *slot_to_vf_attr, + GLuint nr_attrs ) +{ + GLuint i; + + memset(draw->vf_attr_to_slot, 0, sizeof(draw->vf_attr_to_slot)); + draw->nr_attrs = 0; + draw->vertex_size = 0; + + /* + * First three attribs are always the same: header, clip pos, winpos + */ + EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F, 1); + EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F, 4); + + assert(slot_to_vf_attr[0] == VF_ATTRIB_POS); + EMIT_ATTR(slot_to_vf_attr[0], EMIT_4F_VIEWPORT, 4); + + /* + * Remaining attribs (color, texcoords, etc) + */ + for (i = 1; i < nr_attrs; i++) + EMIT_ATTR(slot_to_vf_attr[i], EMIT_4F, 4); + +#if 0 + /* tell the vertex format module how to construct vertices for us */ + draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, + draw->nr_attrs, 0 ); +#endif + + draw->vertex_size *= 4; /* floats to bytes */ +} + + -- cgit v1.2.3 From d9a230f30a98eb677bba869a6f7bb9a840e36354 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 19:03:20 -0600 Subject: Hook in new draw_arrays() code, disabled for now. --- src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_context.c | 23 +++++++++++++++++------ src/mesa/pipe/softpipe/sp_state.h | 11 +++++++++++ src/mesa/pipe/softpipe/sp_state_derived.c | 7 +++++++ 4 files changed, 36 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 9c6d78c290..ffa31102d5 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -8,6 +8,7 @@ DRIVER_SOURCES = \ sp_clear.c \ sp_flush.c \ sp_context.c \ + sp_draw_arrays.c \ sp_prim_setup.c \ sp_quad.c \ sp_quad_alpha_test.c \ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 1ea6bbbf7f..4c0dacd458 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -108,7 +108,8 @@ softpipe_max_texture_size(struct pipe_context *pipe, unsigned textureType, } -static void map_surfaces(struct softpipe_context *sp) +void +softpipe_map_surfaces(struct softpipe_context *sp) { struct pipe_context *pipe = &sp->pipe; unsigned i; @@ -143,7 +144,8 @@ static void map_surfaces(struct softpipe_context *sp) } -static void unmap_surfaces(struct softpipe_context *sp) +void +softpipe_unmap_surfaces(struct softpipe_context *sp) { struct pipe_context *pipe = &sp->pipe; unsigned i; @@ -197,9 +199,9 @@ static void softpipe_draw_vb( struct pipe_context *pipe, softpipe_update_derived( softpipe ); /* XXX move mapping/unmapping to higher/coarser level? */ - map_surfaces(softpipe); + softpipe_map_surfaces(softpipe); draw_vb( softpipe->draw, VB ); - unmap_surfaces(softpipe); + softpipe_unmap_surfaces(softpipe); } @@ -215,9 +217,9 @@ softpipe_draw_vertices(struct pipe_context *pipe, softpipe_update_derived( softpipe ); /* XXX move mapping/unmapping to higher/coarser level? */ - map_surfaces(softpipe); + softpipe_map_surfaces(softpipe); draw_vertices(softpipe->draw, mode, numVertex, verts, numAttribs, attribs); - unmap_surfaces(softpipe); + softpipe_unmap_surfaces(softpipe); } @@ -274,8 +276,13 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_stencil_state = softpipe_set_stencil_state; softpipe->pipe.set_texture_state = softpipe_set_texture_state; softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; + softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; + softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; + softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.draw_vertices = softpipe_draw_vertices; + softpipe->pipe.draw_arrays = softpipe_draw_arrays; + softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; @@ -309,6 +316,10 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, assert(softpipe->draw); draw_set_setup_stage(softpipe->draw, sp_draw_render_stage(softpipe)); + draw_set_vertex_array_info(softpipe->draw, + softpipe->vertex_buffer, + softpipe->vertex_element); + sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 9e5ed96901..d754acaa38 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -91,4 +91,15 @@ void softpipe_set_vertex_buffer(struct pipe_context *, void softpipe_update_derived( struct softpipe_context *softpipe ); + +void softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count); + + +void +softpipe_map_surfaces(struct softpipe_context *sp); + +void +softpipe_unmap_surfaces(struct softpipe_context *sp); + #endif diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index cd67d1c46f..44beb9c3eb 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -156,9 +156,16 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) if (attr_mask != softpipe->attr_mask) { softpipe->attr_mask = attr_mask; +#define USE_NEW_DRAW 0 +#if USE_NEW_DRAW + draw_set_vertex_attributes2( softpipe->draw, + slot_to_vf_attr, + softpipe->nr_attrs ); +#else draw_set_vertex_attributes( softpipe->draw, slot_to_vf_attr, softpipe->nr_attrs ); +#endif } } -- cgit v1.2.3 From 94a4910c9a1ef48470f45c01c379254cb033119f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 19:13:03 -0600 Subject: added PIPE_PRIM_x tokens (matching GL), use in a few places --- src/mesa/pipe/p_defines.h | 14 +++++ src/mesa/pipe/softpipe/sp_draw_arrays.c | 90 +++++++++++++++---------------- src/mesa/state_tracker/st_cb_clear.c | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- 4 files changed, 61 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index addbb026bf..43d1c438ae 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -276,5 +276,19 @@ #define PIPE_FLUSH_TEXTURE_CACHE 0x2 +/** + * Primitive types: + */ +#define PIPE_PRIM_POINTS 0 +#define PIPE_PRIM_LINES 1 +#define PIPE_PRIM_LINE_LOOP 2 +#define PIPE_PRIM_LINE_STRIP 3 +#define PIPE_PRIM_TRIANGLES 4 +#define PIPE_PRIM_TRIANGLE_STRIP 5 +#define PIPE_PRIM_TRIANGLE_FAN 6 +#define PIPE_PRIM_QUADS 7 +#define PIPE_PRIM_QUAD_STRIP 8 +#define PIPE_PRIM_POLYGON 9 + #endif diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index c466884fdd..43a53f108c 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -75,7 +75,7 @@ static unsigned reduced_prim[GL_POLYGON + 1] = { */ static void run_vertex_program(struct draw_context *draw, - const void *vbuffer, GLuint elem, + const void *vbuffer, unsigned elem, struct vertex_header *vOut) { const float *vIn, *cIn; @@ -224,8 +224,8 @@ static void draw_invalidate_vcache( struct draw_context *draw ) * vertices. Flush primitive and/or vertex queues if necessary to * make space. */ -static struct prim_header *get_queued_prim( struct draw_context *draw, - GLuint nr_verts ) +static struct prim_header * +get_queued_prim( struct draw_context *draw, unsigned nr_verts ) { if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH || draw->vcache.overflow + nr_verts >= VCACHE_OVERFLOW) @@ -242,8 +242,8 @@ static struct prim_header *get_queued_prim( struct draw_context *draw, /* Check if vertex is in cache, otherwise add it. It won't go through * VS yet, not until there is a flush operation or the VS queue fills up. */ -static struct vertex_header *get_vertex( struct draw_context *draw, - GLuint i ) +static struct vertex_header * +get_vertex( struct draw_context *draw, unsigned i ) { unsigned slot = (i + (i>>5)) & 31; @@ -273,8 +273,8 @@ static struct vertex_header *get_vertex( struct draw_context *draw, -static void draw_set_prim( struct draw_context *draw, - GLenum prim ) +static void +draw_set_prim( struct draw_context *draw, unsigned prim ) { if (reduced_prim[prim] != draw->reduced_prim) { draw_flush( draw ); @@ -286,7 +286,7 @@ static void draw_set_prim( struct draw_context *draw, static void do_point( struct draw_context *draw, - GLuint i0 ) + unsigned i0 ) { struct prim_header *prim = get_queued_prim( draw, 1 ); @@ -299,8 +299,8 @@ static void do_point( struct draw_context *draw, static void do_line( struct draw_context *draw, GLboolean reset_stipple, - GLuint i0, - GLuint i1 ) + unsigned i0, + unsigned i1 ) { struct prim_header *prim = get_queued_prim( draw, 2 ); @@ -312,9 +312,9 @@ static void do_line( struct draw_context *draw, } static void do_triangle( struct draw_context *draw, - GLuint i0, - GLuint i1, - GLuint i2 ) + unsigned i0, + unsigned i1, + unsigned i2 ) { struct prim_header *prim = get_queued_prim( draw, 3 ); @@ -328,10 +328,10 @@ static void do_triangle( struct draw_context *draw, static void do_ef_triangle( struct draw_context *draw, GLboolean reset_stipple, - GLuint ef_mask, - GLuint i0, - GLuint i1, - GLuint i2 ) + unsigned ef_mask, + unsigned i0, + unsigned i1, + unsigned i2 ) { struct prim_header *prim = get_queued_prim( draw, 3 ); struct vertex_header *v0 = draw->get_vertex( draw, i0 ); @@ -362,22 +362,22 @@ static void do_quad( struct draw_context *draw, static void draw_prim( struct draw_context *draw, - GLuint start, - GLuint count ) + unsigned start, + unsigned count ) { - GLuint i; + unsigned i; // _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); switch (draw->prim) { - case GL_POINTS: + case PIPE_PRIM_POINTS: for (i = 0; i < count; i ++) { do_point( draw, start + i ); } break; - case GL_LINES: + case PIPE_PRIM_LINES: for (i = 0; i+1 < count; i += 2) { do_line( draw, TRUE, @@ -386,7 +386,7 @@ static void draw_prim( struct draw_context *draw, } break; - case GL_LINE_LOOP: + case PIPE_PRIM_LINE_LOOP: if (count >= 2) { for (i = 1; i < count; i++) { do_line( draw, @@ -402,7 +402,7 @@ static void draw_prim( struct draw_context *draw, } break; - case GL_LINE_STRIP: + case PIPE_PRIM_LINE_STRIP: if (count >= 2) { for (i = 1; i < count; i++) { do_line( draw, @@ -413,7 +413,7 @@ static void draw_prim( struct draw_context *draw, } break; - case GL_TRIANGLES: + case PIPE_PRIM_TRIANGLES: for (i = 0; i+2 < count; i += 3) { do_ef_triangle( draw, 1, @@ -424,7 +424,7 @@ static void draw_prim( struct draw_context *draw, } break; - case GL_TRIANGLE_STRIP: + case PIPE_PRIM_TRIANGLE_STRIP: for (i = 0; i+2 < count; i++) { if (i & 1) { do_triangle( draw, @@ -441,7 +441,7 @@ static void draw_prim( struct draw_context *draw, } break; - case GL_TRIANGLE_FAN: + case PIPE_PRIM_TRIANGLE_FAN: if (count >= 3) { for (i = 0; i+2 < count; i++) { do_triangle( draw, @@ -453,7 +453,7 @@ static void draw_prim( struct draw_context *draw, break; - case GL_QUADS: + case PIPE_PRIM_QUADS: for (i = 0; i+3 < count; i += 4) { do_quad( draw, start + i + 0, @@ -463,7 +463,7 @@ static void draw_prim( struct draw_context *draw, } break; - case GL_QUAD_STRIP: + case PIPE_PRIM_QUAD_STRIP: for (i = 0; i+3 < count; i += 2) { do_quad( draw, start + i + 2, @@ -473,7 +473,7 @@ static void draw_prim( struct draw_context *draw, } break; - case GL_POLYGON: + case PIPE_PRIM_POLYGON: if (count >= 3) { unsigned ef_mask = (1<<2) | (1<<0); @@ -502,43 +502,43 @@ static void draw_prim( struct draw_context *draw, -static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) +static unsigned draw_prim_info(unsigned mode, unsigned *first, unsigned *incr) { switch (mode) { - case GL_POINTS: + case PIPE_PRIM_POINTS: *first = 1; *incr = 1; return 0; - case GL_LINES: + case PIPE_PRIM_LINES: *first = 2; *incr = 2; return 0; - case GL_LINE_STRIP: + case PIPE_PRIM_LINE_STRIP: *first = 2; *incr = 1; return 0; - case GL_LINE_LOOP: + case PIPE_PRIM_LINE_LOOP: *first = 2; *incr = 1; return 1; - case GL_TRIANGLES: + case PIPE_PRIM_TRIANGLES: *first = 3; *incr = 3; return 0; - case GL_TRIANGLE_STRIP: + case PIPE_PRIM_TRIANGLE_STRIP: *first = 3; *incr = 1; return 0; - case GL_TRIANGLE_FAN: - case GL_POLYGON: + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: *first = 3; *incr = 1; return 1; - case GL_QUADS: + case PIPE_PRIM_QUADS: *first = 4; *incr = 4; return 0; - case GL_QUAD_STRIP: + case PIPE_PRIM_QUAD_STRIP: *first = 4; *incr = 2; return 0; @@ -551,7 +551,7 @@ static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) } -static GLuint trim( GLuint count, GLuint first, GLuint incr ) +static unsigned trim( unsigned count, unsigned first, unsigned incr ) { if (count < first) return 0; @@ -627,10 +627,10 @@ do { \ void draw_set_vertex_attributes2( struct draw_context *draw, - const GLuint *slot_to_vf_attr, - GLuint nr_attrs ) + const unsigned *slot_to_vf_attr, + unsigned nr_attrs ) { - GLuint i; + unsigned i; memset(draw->vf_attr_to_slot, 0, sizeof(draw->vf_attr_to_slot)); draw->nr_attrs = 0; diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 80f136e0a6..2f7ade73e4 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -190,7 +190,7 @@ draw_quad(GLcontext *ctx, verts[i][1][3] = color[3]; } - ctx->st->pipe->draw_vertices(ctx->st->pipe, GL_QUADS, + ctx->st->pipe->draw_vertices(ctx->st->pipe, PIPE_PRIM_QUADS, 4, (GLfloat *) verts, 2, attribs); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index df0b14463b..dd27760a58 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -235,7 +235,7 @@ draw_quad(struct st_context *st, GLfloat x0, GLfloat y0, GLfloat z, verts[i][1][3] = 1.0; /*Q*/ } - st->pipe->draw_vertices(st->pipe, GL_QUADS, + st->pipe->draw_vertices(st->pipe, PIPE_PRIM_QUADS, 4, (GLfloat *) verts, 2, attribs); } -- cgit v1.2.3 From 5390a43ce06b27f6d54bc5f237aa305b6948f2af Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 15 Aug 2007 19:23:58 -0600 Subject: translate GL texture targets to PIPE_TEXTURE_x values --- src/mesa/state_tracker/st_cb_texture.c | 48 ++++++++++++++++++++++++--------- src/mesa/state_tracker/st_mipmap_tree.c | 9 +++++-- 2 files changed, 43 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index f633bd328f..770c871434 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -114,6 +114,30 @@ st_get_texobj_mipmap_tree(struct gl_texture_object *texObj) } +static unsigned +gl_target_to_pipe(GLenum target) +{ + switch (target) { + case GL_TEXTURE_1D: + return PIPE_TEXTURE_1D; + + case GL_TEXTURE_2D: + case GL_TEXTURE_RECTANGLE_NV: + return PIPE_TEXTURE_2D; + + case GL_TEXTURE_3D: + return PIPE_TEXTURE_3D; + + case GL_TEXTURE_CUBE_MAP_ARB: + return PIPE_TEXTURE_CUBE; + + default: + assert(0); + return 0; + } +} + + static int compressed_num_bytes(GLuint mesaFormat) { @@ -360,15 +384,15 @@ guess_and_alloc_mipmap_tree(struct pipe_context *pipe, if (stImage->base.IsCompressed) comp_byte = compressed_num_bytes(stImage->base.TexFormat->MesaFormat); stObj->mt = st_miptree_create(pipe, - stObj->base.Target, - stImage->base.InternalFormat, - firstLevel, - lastLevel, - width, - height, - depth, - stImage->base.TexFormat->TexelBytes, - comp_byte); + gl_target_to_pipe(stObj->base.Target), + stImage->base.InternalFormat, + firstLevel, + lastLevel, + width, + height, + depth, + stImage->base.TexFormat->TexelBytes, + comp_byte); stObj->mt->format = st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat); @@ -587,7 +611,7 @@ st_TexImage(GLcontext * ctx, if (stObj->mt && stObj->mt->first_level == level && stObj->mt->last_level == level && - stObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && + stObj->mt->target != PIPE_TEXTURE_CUBE && !st_miptree_match_image(stObj->mt, &stImage->base, stImage->face, stImage->level)) { @@ -1476,7 +1500,7 @@ st_finalize_mipmap_tree(GLcontext *ctx, * leaving the tree alone. */ if (stObj->mt && - (stObj->mt->target != stObj->base.Target || + (stObj->mt->target != gl_target_to_pipe(stObj->base.Target) || stObj->mt->internal_format != firstImage->base.InternalFormat || stObj->mt->first_level != stObj->firstLevel || stObj->mt->last_level != stObj->lastLevel || @@ -1493,7 +1517,7 @@ st_finalize_mipmap_tree(GLcontext *ctx, */ if (!stObj->mt) { stObj->mt = st_miptree_create(pipe, - stObj->base.Target, + gl_target_to_pipe(stObj->base.Target), firstImage->base.InternalFormat, stObj->firstLevel, stObj->lastLevel, diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index 3cbe697ab3..535004c892 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -30,10 +30,12 @@ #include "pipe/p_state.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" #define DBG if(0) printf +#if 0 static GLenum target_to_target(GLenum target) { @@ -49,10 +51,11 @@ target_to_target(GLenum target) return target; } } +#endif struct pipe_mipmap_tree * st_miptree_create(struct pipe_context *pipe, - GLenum target, + unsigned target, GLenum internal_format, GLuint first_level, GLuint last_level, @@ -64,11 +67,13 @@ st_miptree_create(struct pipe_context *pipe, struct pipe_mipmap_tree *mt = calloc(sizeof(*mt), 1); GLbitfield flags = 0x0; + assert(target <= PIPE_TEXTURE_CUBE); + DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); - mt->target = target_to_target(target); + mt->target = target; mt->internal_format = internal_format; mt->first_level = first_level; mt->last_level = last_level; -- cgit v1.2.3 From b0b871429daabfb6a8788444283021fab79e2d0f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 15 Aug 2007 15:12:03 +0100 Subject: Remove mesa include directories, be stricter about include paths. --- src/mesa/math/m_vector.h | 4 ++-- src/mesa/pipe/Makefile.template | 3 --- src/mesa/pipe/draw/draw_context.h | 2 +- src/mesa/pipe/softpipe/sp_clear.c | 1 - src/mesa/pipe/softpipe/sp_context.h | 2 -- src/mesa/pipe/softpipe/sp_prim_setup.c | 3 --- src/mesa/vf/vf.h | 2 +- 7 files changed, 4 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/math/m_vector.h b/src/mesa/math/m_vector.h index 222b47f6e1..b1a65b3356 100644 --- a/src/mesa/math/m_vector.h +++ b/src/mesa/math/m_vector.h @@ -31,8 +31,8 @@ #ifndef _M_VECTOR_H_ #define _M_VECTOR_H_ -#include "glheader.h" -#include "mtypes.h" /* hack for GLchan */ +#include "main/glheader.h" +#include "main/mtypes.h" /* hack for GLchan */ #define VEC_DIRTY_0 0x1 diff --git a/src/mesa/pipe/Makefile.template b/src/mesa/pipe/Makefile.template index f2a9591de5..fe0a81af6a 100644 --- a/src/mesa/pipe/Makefile.template +++ b/src/mesa/pipe/Makefile.template @@ -14,10 +14,7 @@ OBJECTS = $(C_SOURCES:.c=.o) \ ### Include directories INCLUDES = \ -I. \ - -I$(TOP)/src/mesa/drivers/dri/common \ - -I$(TOP)/src/mesa/main \ -I$(TOP)/src/mesa/pipe \ - -I$(TOP)/include \ -I$(TOP)/src/mesa diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index b9cac7c6da..beba9429c9 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -38,7 +38,7 @@ #define DRAW_CONTEXT_H -#include "glheader.h" +#include "main/glheader.h" #include "pipe/p_state.h" diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 2774413720..6e5e510637 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -35,7 +35,6 @@ #include "sp_context.h" #include "sp_surface.h" #include "sp_state.h" -//#include "colormac.h" /** diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 8b0e970ea2..f45cb6fe2c 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -31,8 +31,6 @@ #ifndef SP_CONTEXT_H #define SP_CONTEXT_H -//#include "glheader.h" - #include "pipe/p_state.h" #include "pipe/p_context.h" diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 71ef798cbb..9b91628aaf 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -33,9 +33,6 @@ */ -//#include "imports.h" -//#include "macros.h" - #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" diff --git a/src/mesa/vf/vf.h b/src/mesa/vf/vf.h index 8ca3ab6d4a..83d7547619 100644 --- a/src/mesa/vf/vf.h +++ b/src/mesa/vf/vf.h @@ -28,7 +28,7 @@ #ifndef VF_VERTEX_H #define VF_VERTEX_H -#include "mtypes.h" +#include "main/mtypes.h" #include "math/m_vector.h" enum { -- cgit v1.2.3 From 9982044a8de79d927a2921126c1684e04f18c170 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 12:12:14 +0100 Subject: Use 'f' for float suffix. --- src/mesa/pipe/p_util.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) mode change 100644 => 100755 src/mesa/pipe/p_util.h (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h old mode 100644 new mode 100755 index 50d57a2ed0..072dad0573 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -60,7 +60,7 @@ union fi { else if (__tmp.i >= IEEE_0996) \ UB = (ubyte) 255; \ else { \ - __tmp.f = __tmp.f * (255.0F/256.0F) + 32768.0F; \ + __tmp.f = __tmp.f * (255.0f/256.0f) + 32768.0f; \ UB = (ubyte) __tmp.i; \ } \ } while (0) -- cgit v1.2.3 From 830169414e2886c2224d62a1a601cc951dcd4b08 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 12:13:37 +0100 Subject: Silence warnings. --- src/mesa/pipe/softpipe/sp_quad_blend.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) mode change 100644 => 100755 src/mesa/pipe/softpipe/sp_quad_blend.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c old mode 100644 new mode 100755 index 7e8851d214..e5335f3b19 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -221,20 +221,20 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) { float inv_comp[4]; /* R */ - VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[0]); + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* G */ - VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[1]); + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); /* B */ - VEC4_SCALAR(inv_comp, 1.0 - softpipe->blend_color.color[2]); + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); } break; case PIPE_BLENDFACTOR_INV_CONST_ALPHA: { - float alpha[4], inv_alpha[4]; - VEC4_SCALAR(alpha, 1.0 - softpipe->blend_color.color[3]); + float inv_alpha[4]; + VEC4_SCALAR(inv_alpha, 1.0f - softpipe->blend_color.color[3]); VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ -- cgit v1.2.3 From e36b27cf86fb767a4a9754480abe9e299c746cf7 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 12:16:27 +0100 Subject: Silence warnings. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) mode change 100644 => 100755 src/mesa/pipe/softpipe/sp_quad_fs.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c old mode 100644 new mode 100755 index 0a345bbf11..c2decfecf8 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -73,9 +73,8 @@ shade_quad( { struct quad_shade_stage *qss = quad_shade_stage( qs ); struct softpipe_context *softpipe = qs->softpipe; - const float fx = quad->x0; - const float fy = quad->y0; - unsigned attr, i; + const float fx = (float) quad->x0; + const float fy = (float) quad->y0; struct tgsi_exec_machine machine; #if USE_ALIGNED_ATTRIBS @@ -111,14 +110,14 @@ shade_quad( machine.InterpCoefs = quad->coef; machine.Inputs[0].xyzw[0].f[0] = fx; - machine.Inputs[0].xyzw[0].f[1] = fx + 1.0; + machine.Inputs[0].xyzw[0].f[1] = fx + 1.0f; machine.Inputs[0].xyzw[0].f[2] = fx; - machine.Inputs[0].xyzw[0].f[3] = fx + 1.0; + machine.Inputs[0].xyzw[0].f[3] = fx + 1.0f; machine.Inputs[0].xyzw[1].f[0] = fy; machine.Inputs[0].xyzw[1].f[1] = fy; - machine.Inputs[0].xyzw[1].f[2] = fy + 1.0; - machine.Inputs[0].xyzw[1].f[3] = fy + 1.0; + machine.Inputs[0].xyzw[1].f[2] = fy + 1.0f; + machine.Inputs[0].xyzw[1].f[3] = fy + 1.0f; /* run shader */ tgsi_exec_machine_run( &machine ); -- cgit v1.2.3 From 4fcfab33477a0b01ef2f5b1fb2658f244b91e03b Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 12:16:57 +0100 Subject: Silence warnings. --- src/mesa/pipe/softpipe/sp_surface.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) mode change 100644 => 100755 src/mesa/pipe/softpipe/sp_surface.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c old mode 100644 new mode 100755 index 4152e3510d..84d02f91e7 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -56,8 +56,8 @@ a8r8g8b8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, unsigned i, j; assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - assert(x < sps->surface.width - 1); - assert(y < sps->surface.height - 1); + assert(x < (int) sps->surface.width - 1); + assert(y < (int) sps->surface.height - 1); for (i = 0; i < 2; i++) { /* loop over pixel row */ for (j = 0; j < 2; j++) { /* loop over pixel column */ @@ -149,10 +149,10 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { const ushort pixel = src[j]; - p[0] = ((pixel >> 10) & 0x1f) * (1.0 / 31); - p[1] = ((pixel >> 5) & 0x1f) * (1.0 / 31); - p[2] = ((pixel ) & 0x1f) * (1.0 / 31); - p[3] = ((pixel >> 15) ); + p[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); + p[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); + p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + p[3] = ((pixel >> 15) ) * 1.0f; p += 4; } src += ps->region->pitch; -- cgit v1.2.3 From 76f9f3ee832a5e47a321afb01ccac38fe6696196 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 12:17:30 +0100 Subject: Silence warnings. --- src/mesa/pipe/softpipe/sp_tex_layout.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) mode change 100644 => 100755 src/mesa/pipe/softpipe/sp_tex_layout.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c old mode 100644 new mode 100755 index 0517c01b5d..15447d6600 --- a/src/mesa/pipe/softpipe/sp_tex_layout.c +++ b/src/mesa/pipe/softpipe/sp_tex_layout.c @@ -197,7 +197,7 @@ static const int step_offsets[6][2] = { boolean softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) { - int level; + unsigned level; switch (mt->target) { case PIPE_TEXTURE_CUBE:{ @@ -305,7 +305,7 @@ softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * unsigned nr_images = mt->target == PIPE_TEXTURE_3D ? depth : 6; int x = 0; int y = 0; - int q, j; + unsigned q, j; sp_miptree_set_level_info(mt, level, nr_images, 0, mt->total_height, -- cgit v1.2.3 From 926323df8692f48134a65322d5790d3f950b67dd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 12:35:13 -0600 Subject: point/line/polygon drawing (factored out of draw_vb.c) --- src/mesa/pipe/draw/draw_prim.c | 519 +++++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_prim.h | 23 ++ 2 files changed, 542 insertions(+) create mode 100644 src/mesa/pipe/draw/draw_prim.c create mode 100644 src/mesa/pipe/draw/draw_prim.h (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c new file mode 100644 index 0000000000..b7d30c567a --- /dev/null +++ b/src/mesa/pipe/draw/draw_prim.c @@ -0,0 +1,519 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "draw_private.h" +#include "draw_context.h" +#include "draw_prim.h" + + +#define RP_NONE 0 +#define RP_POINT 1 +#define RP_LINE 2 +#define RP_TRI 3 + + +static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { + RP_POINT, + RP_LINE, + RP_LINE, + RP_LINE, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI +}; + + +void draw_flush( struct draw_context *draw ) +{ + struct draw_stage *first = draw->pipeline.first; + unsigned i; + + /* Make sure all vertices are available: + */ + assert(draw->vs_flush); + draw->vs_flush( draw ); + + + switch (draw->reduced_prim) { + case RP_TRI: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + first->reset_stipple_counter( first ); + + first->tri( first, &draw->pq.queue[i] ); + } + break; + case RP_LINE: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + first->reset_stipple_counter( first ); + + first->line( first, &draw->pq.queue[i] ); + } + break; + case RP_POINT: + first->reset_stipple_counter( first ); + for (i = 0; i < draw->pq.queue_nr; i++) + first->point( first, &draw->pq.queue[i] ); + break; + } + + draw->pq.queue_nr = 0; + draw->vcache.referenced = 0; + draw->vcache.overflow = 0; +} + + +void draw_invalidate_vcache( struct draw_context *draw ) +{ + unsigned i; + + assert(draw->pq.queue_nr == 0); + assert(draw->vs.queue_nr == 0); + assert(draw->vcache.referenced == 0); + + for (i = 0; i < Elements( draw->vcache.idx ); i++) + draw->vcache.idx[i] = ~0; +} + + +/* Return a pointer to a freshly queued primitive header. Ensure that + * there is room in the vertex cache for a maximum of "nr_verts" new + * vertices. Flush primitive and/or vertex queues if necessary to + * make space. + */ +static struct prim_header *get_queued_prim( struct draw_context *draw, + unsigned nr_verts ) +{ + if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH || + draw->vcache.overflow + nr_verts >= VCACHE_OVERFLOW) + draw_flush( draw ); + + /* The vs queue is sized so that this can never happen: + */ + assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); + + return &draw->pq.queue[draw->pq.queue_nr++]; +} + + +/* Check if vertex is in cache, otherwise add it. It won't go through + * VS yet, not until there is a flush operation or the VS queue fills up. + */ +static struct vertex_header *get_vertex( struct draw_context *draw, + unsigned i ) +{ + unsigned slot = (i + (i>>5)) & 31; + + /* Cache miss? + */ + if (draw->vcache.idx[slot] != i) { + + /* If slot is in use, use the overflow area: + */ + if (draw->vcache.referenced & (1<vcache.overflow++; + + draw->vcache.idx[slot] = i; + + /* Add to vertex shader queue: + */ + draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; + draw->vs.queue[draw->vs.queue_nr].elt = i; + draw->vs.queue_nr++; + } + + /* Mark slot as in-use: + */ + draw->vcache.referenced |= (1<vcache.vertex[slot]; +} + + +static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const unsigned *elts = (const unsigned *)draw->elts; + return get_vertex( draw, elts[i] ); +} + + +static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const GLushort *elts = (const GLushort *)draw->elts; + return get_vertex( draw, elts[i] ); +} + + +static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const GLubyte *elts = (const GLubyte *)draw->elts; + return get_vertex( draw, elts[i] ); +} + + +static void do_point( struct draw_context *draw, + unsigned i0 ) +{ + struct prim_header *prim = get_queued_prim( draw, 1 ); + + prim->reset_line_stipple = 0; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->get_vertex( draw, i0 ); +} + + +static void do_line( struct draw_context *draw, + boolean reset_stipple, + unsigned i0, + unsigned i1 ) +{ + struct prim_header *prim = get_queued_prim( draw, 2 ); + + prim->reset_line_stipple = reset_stipple; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->get_vertex( draw, i0 ); + prim->v[1] = draw->get_vertex( draw, i1 ); +} + +static void do_triangle( struct draw_context *draw, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + + prim->reset_line_stipple = 1; + prim->edgeflags = ~0; + prim->pad = 0; + prim->v[0] = draw->get_vertex( draw, i0 ); + prim->v[1] = draw->get_vertex( draw, i1 ); + prim->v[2] = draw->get_vertex( draw, i2 ); +} + +static void do_ef_triangle( struct draw_context *draw, + boolean reset_stipple, + unsigned ef_mask, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + struct vertex_header *v0 = draw->get_vertex( draw, i0 ); + struct vertex_header *v1 = draw->get_vertex( draw, i1 ); + struct vertex_header *v2 = draw->get_vertex( draw, i2 ); + + prim->reset_line_stipple = reset_stipple; + + prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | + (v1->edgeflag << 1) | + (v2->edgeflag << 2)); + prim->pad = 0; + prim->v[0] = v0; + prim->v[1] = v1; + prim->v[2] = v2; +} + + +static void do_quad( struct draw_context *draw, + unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3 ) +{ + do_ef_triangle( draw, 1, ~(1<<0), v0, v1, v3 ); + do_ef_triangle( draw, 0, ~(1<<1), v1, v2, v3 ); +} + + +/** + * Main entrypoint to draw some number of points/lines/triangles + */ +void +draw_prim( struct draw_context *draw, unsigned start, unsigned count ) +{ + unsigned i; + +// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); + + switch (draw->prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i ++) { + do_point( draw, + start + i ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) { + do_line( draw, + TRUE, + start + i + 0, + start + i + 1); + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, /* XXX: only if vb not split */ + start + i - 1, + start + i ); + } + + do_line( draw, + 0, + start + count - 1, + start + 0 ); + } + break; + + case PIPE_PRIM_LINE_STRIP: + if (count >= 2) { + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, + start + i - 1, + start + i ); + } + } + break; + + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + do_ef_triangle( draw, + 1, + ~0, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + do_triangle( draw, + start + i + 1, + start + i + 0, + start + i + 2 ); + } + else { + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + if (count >= 3) { + for (i = 0; i+2 < count; i++) { + do_triangle( draw, + start + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + + case PIPE_PRIM_QUADS: + for (i = 0; i+3 < count; i += 4) { + do_quad( draw, + start + i + 0, + start + i + 1, + start + i + 2, + start + i + 3); + } + break; + + case PIPE_PRIM_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + do_quad( draw, + start + i + 2, + start + i + 0, + start + i + 1, + start + i + 3); + } + break; + + case PIPE_PRIM_POLYGON: + if (count >= 3) { + unsigned ef_mask = (1<<2) | (1<<0); + + for (i = 0; i+2 < count; i++) { + + if (i + 3 >= count) + ef_mask |= (1<<1); + + do_ef_triangle( draw, + i == 0, + ef_mask, + start + i + 1, + start + i + 2, + start + i + 0); + + ef_mask &= ~(1<<2); + } + } + break; + + default: + assert(0); + break; + } +} + + +void +draw_set_prim( struct draw_context *draw, unsigned prim ) +{ + assert(prim >= PIPE_PRIM_POINTS); + assert(prim <= PIPE_PRIM_POLYGON); + + if (reduced_prim[prim] != draw->reduced_prim) { + draw_flush( draw ); + draw->reduced_prim = reduced_prim[prim]; + } + + draw->prim = prim; +} + + +/** + * Tell the drawing context about the index/element buffer to use + * (ala glDrawElements) + * If no element buffer is to be used (i.e. glDrawArrays) then this + * should be called with eltSize=0 and elements=NULL. + * + * \param draw the drawing context + * \param eltSize size of each element (1, 2 or 4 bytes) + * \param elements the element buffer ptr + */ +void +draw_set_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ) +{ + /* choose the get_vertex() function to use */ + switch (eltSize) { + case 0: + draw->get_vertex = get_vertex; + break; + case 1: + draw->get_vertex = get_ubyte_elt_vertex; + break; + case 2: + draw->get_vertex = get_ushort_elt_vertex; + break; + case 4: + draw->get_vertex = get_uint_elt_vertex; + break; + default: + assert(0); + } + draw->elts = elements; + draw->eltSize = eltSize; + + +} + +unsigned +draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) +{ + assert(prim >= PIPE_PRIM_POINTS); + assert(prim <= PIPE_PRIM_POLYGON); + + switch (prim) { + case PIPE_PRIM_POINTS: + *first = 1; + *incr = 1; + return 0; + case PIPE_PRIM_LINES: + *first = 2; + *incr = 2; + return 0; + case PIPE_PRIM_LINE_STRIP: + *first = 2; + *incr = 1; + return 0; + case PIPE_PRIM_LINE_LOOP: + *first = 2; + *incr = 1; + return 1; + case PIPE_PRIM_TRIANGLES: + *first = 3; + *incr = 3; + return 0; + case PIPE_PRIM_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + return 0; + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + *first = 3; + *incr = 1; + return 1; + case PIPE_PRIM_QUADS: + *first = 4; + *incr = 4; + return 0; + case PIPE_PRIM_QUAD_STRIP: + *first = 4; + *incr = 2; + return 0; + default: + assert(0); + *first = 1; + *incr = 1; + return 0; + } +} + + +unsigned +draw_trim( unsigned count, unsigned first, unsigned incr ) +{ + if (count < first) + return 0; + else + return count - (count - first) % incr; +} diff --git a/src/mesa/pipe/draw/draw_prim.h b/src/mesa/pipe/draw/draw_prim.h new file mode 100644 index 0000000000..3224989bef --- /dev/null +++ b/src/mesa/pipe/draw/draw_prim.h @@ -0,0 +1,23 @@ + + +#ifndef DRAW_PRIM_H +#define DRAW_PRIM_H + + +void draw_invalidate_vcache( struct draw_context *draw ); + +void draw_set_prim( struct draw_context *draw, unsigned prim ); + +void draw_set_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ); + +void draw_prim( struct draw_context *draw, unsigned start, unsigned count ); + +void draw_flush( struct draw_context *draw ); + +unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); + +unsigned draw_trim( unsigned count, unsigned first, unsigned incr ); + + +#endif /* DRAW_PRIM_H */ -- cgit v1.2.3 From 59bd1e260bf40e4d2b1662cc4e68eff8235739e4 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 12:36:17 -0600 Subject: Drawing code refactoring. Move code duplicated between draw_vb.c and sp_draw_arrays.c into draw_prim.c draw_vb.c will eventually go away, but this seems like a good step anyway. --- src/mesa/pipe/draw/draw_private.h | 4 + src/mesa/pipe/draw/draw_vb.c | 454 +------------------------------ src/mesa/pipe/softpipe/sp_draw_arrays.c | 465 ++------------------------------ src/mesa/sources | 1 + 4 files changed, 39 insertions(+), 885 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 76ca80dab0..c1ebd34121 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -157,6 +157,8 @@ struct draw_context GLuint nr_vertices; GLboolean in_vb; + /** Pointer to vertex element/index buffer */ + unsigned eltSize; /**< bytes per index (0, 1, 2 or 4) */ void *elts; struct vertex_header *(*get_vertex)( struct draw_context *draw, @@ -197,6 +199,8 @@ struct draw_context GLenum prim; /**< GL_POINTS, GL_LINE_STRIP, GL_QUADS, etc */ unsigned reduced_prim; + void (*vs_flush)( struct draw_context *draw ); + /* Helper for tnl: */ GLvector4f header; diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index 23a5e3fc3b..a88babc44e 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -38,25 +38,7 @@ #include "draw_private.h" #include "draw_context.h" - - -#define RP_NONE 0 -#define RP_POINT 1 -#define RP_LINE 2 -#define RP_TRI 3 - -static unsigned reduced_prim[GL_POLYGON + 1] = { - RP_POINT, - RP_LINE, - RP_LINE, - RP_LINE, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI -}; +#include "draw_prim.h" @@ -105,364 +87,6 @@ static void vs_flush( struct draw_context *draw ) } -static void draw_flush( struct draw_context *draw ) -{ - struct draw_stage *first = draw->pipeline.first; - unsigned i; - - /* Make sure all vertices are available: - */ - vs_flush( draw ); - - - switch (draw->reduced_prim) { - case RP_TRI: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - first->reset_stipple_counter( first ); - - first->tri( first, &draw->pq.queue[i] ); - } - break; - case RP_LINE: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - first->reset_stipple_counter( first ); - - first->line( first, &draw->pq.queue[i] ); - } - break; - case RP_POINT: - first->reset_stipple_counter( first ); - for (i = 0; i < draw->pq.queue_nr; i++) - first->point( first, &draw->pq.queue[i] ); - break; - } - - draw->pq.queue_nr = 0; - draw->vcache.referenced = 0; - draw->vcache.overflow = 0; -} - -static void draw_invalidate_vcache( struct draw_context *draw ) -{ - unsigned i; - - assert(draw->pq.queue_nr == 0); - assert(draw->vs.queue_nr == 0); - assert(draw->vcache.referenced == 0); - - for (i = 0; i < Elements( draw->vcache.idx ); i++) - draw->vcache.idx[i] = ~0; -} - - -/* Return a pointer to a freshly queued primitive header. Ensure that - * there is room in the vertex cache for a maximum of "nr_verts" new - * vertices. Flush primitive and/or vertex queues if necessary to - * make space. - */ -static struct prim_header *get_queued_prim( struct draw_context *draw, - GLuint nr_verts ) -{ - if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH || - draw->vcache.overflow + nr_verts >= VCACHE_OVERFLOW) - draw_flush( draw ); - - /* The vs queue is sized so that this can never happen: - */ - assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); - - return &draw->pq.queue[draw->pq.queue_nr++]; -} - - -/* Check if vertex is in cache, otherwise add it. It won't go through - * VS yet, not until there is a flush operation or the VS queue fills up. - */ -static struct vertex_header *get_vertex( struct draw_context *draw, - GLuint i ) -{ - unsigned slot = (i + (i>>5)) & 31; - - /* Cache miss? - */ - if (draw->vcache.idx[slot] != i) { - - /* If slot is in use, use the overflow area: - */ - if (draw->vcache.referenced & (1<vcache.overflow++; - - draw->vcache.idx[slot] = i; - - /* Add to vertex shader queue: - */ - draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; - draw->vs.queue[draw->vs.queue_nr].elt = i; - draw->vs.queue_nr++; - } - - /* Mark slot as in-use: - */ - draw->vcache.referenced |= (1<vcache.vertex[slot]; -} - - -static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, - GLuint i ) -{ - const GLuint *elts = (const GLuint *)draw->elts; - return get_vertex( draw, elts[i] ); -} - -#if 0 -static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, - const void *elts, - GLuint i ) -{ - const GLushort *elts = (const GLushort *)draw->elts; - return get_vertex( draw, elts[i] ); -} - -static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, - const void *elts, - GLuint i ) -{ - const GLubyte *elts = (const GLubyte *)draw->elts; - return get_vertex( draw, elts[i] ); -} -#endif - - -static void draw_set_prim( struct draw_context *draw, - GLenum prim ) -{ - if (reduced_prim[prim] != draw->reduced_prim) { - draw_flush( draw ); - draw->reduced_prim = reduced_prim[prim]; - } - - draw->prim = prim; -} - - -static void do_point( struct draw_context *draw, - GLuint i0 ) -{ - struct prim_header *prim = get_queued_prim( draw, 1 ); - - prim->reset_line_stipple = 0; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->get_vertex( draw, i0 ); -} - - -static void do_line( struct draw_context *draw, - GLboolean reset_stipple, - GLuint i0, - GLuint i1 ) -{ - struct prim_header *prim = get_queued_prim( draw, 2 ); - - prim->reset_line_stipple = reset_stipple; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->get_vertex( draw, i0 ); - prim->v[1] = draw->get_vertex( draw, i1 ); -} - -static void do_triangle( struct draw_context *draw, - GLuint i0, - GLuint i1, - GLuint i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - - prim->reset_line_stipple = 1; - prim->edgeflags = ~0; - prim->pad = 0; - prim->v[0] = draw->get_vertex( draw, i0 ); - prim->v[1] = draw->get_vertex( draw, i1 ); - prim->v[2] = draw->get_vertex( draw, i2 ); -} - -static void do_ef_triangle( struct draw_context *draw, - GLboolean reset_stipple, - GLuint ef_mask, - GLuint i0, - GLuint i1, - GLuint i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - struct vertex_header *v0 = draw->get_vertex( draw, i0 ); - struct vertex_header *v1 = draw->get_vertex( draw, i1 ); - struct vertex_header *v2 = draw->get_vertex( draw, i2 ); - - prim->reset_line_stipple = reset_stipple; - - prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | - (v1->edgeflag << 1) | - (v2->edgeflag << 2)); - prim->pad = 0; - prim->v[0] = v0; - prim->v[1] = v1; - prim->v[2] = v2; -} - - -static void do_quad( struct draw_context *draw, - unsigned v0, - unsigned v1, - unsigned v2, - unsigned v3 ) -{ - do_ef_triangle( draw, 1, ~(1<<0), v0, v1, v3 ); - do_ef_triangle( draw, 0, ~(1<<1), v1, v2, v3 ); -} - - -static void draw_prim( struct draw_context *draw, - GLuint start, - GLuint count ) -{ - GLuint i; - -// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - - switch (draw->prim) { - case GL_POINTS: - for (i = 0; i < count; i ++) { - do_point( draw, - start + i ); - } - break; - - case GL_LINES: - for (i = 0; i+1 < count; i += 2) { - do_line( draw, - TRUE, - start + i + 0, - start + i + 1); - } - break; - - case GL_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, /* XXX: only if vb not split */ - start + i - 1, - start + i ); - } - - do_line( draw, - 0, - start + count - 1, - start + 0 ); - } - break; - - case GL_LINE_STRIP: - if (count >= 2) { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, - start + i - 1, - start + i ); - } - } - break; - - case GL_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - do_ef_triangle( draw, - 1, - ~0, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - break; - - case GL_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - do_triangle( draw, - start + i + 1, - start + i + 0, - start + i + 2 ); - } - else { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - case GL_TRIANGLE_FAN: - if (count >= 3) { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - - case GL_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( draw, - start + i + 0, - start + i + 1, - start + i + 2, - start + i + 3); - } - break; - - case GL_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( draw, - start + i + 2, - start + i + 0, - start + i + 1, - start + i + 3); - } - break; - - case GL_POLYGON: - if (count >= 3) { - unsigned ef_mask = (1<<2) | (1<<0); - - for (i = 0; i+2 < count; i++) { - - if (i + 3 >= count) - ef_mask |= (1<<1); - - do_ef_triangle( draw, - i == 0, - ef_mask, - start + i + 1, - start + i + 2, - start + i + 0); - - ef_mask &= ~(1<<2); - } - } - break; - - default: - assert(0); - break; - } -} - /** * Allocate storage for post-transformation vertices. @@ -554,66 +178,6 @@ build_vertex_headers( struct draw_context *draw, - - -static GLuint draw_prim_info(GLenum mode, GLuint *first, GLuint *incr) -{ - switch (mode) { - case GL_POINTS: - *first = 1; - *incr = 1; - return 0; - case GL_LINES: - *first = 2; - *incr = 2; - return 0; - case GL_LINE_STRIP: - *first = 2; - *incr = 1; - return 0; - case GL_LINE_LOOP: - *first = 2; - *incr = 1; - return 1; - case GL_TRIANGLES: - *first = 3; - *incr = 3; - return 0; - case GL_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - return 0; - case GL_TRIANGLE_FAN: - case GL_POLYGON: - *first = 3; - *incr = 1; - return 1; - case GL_QUADS: - *first = 4; - *incr = 4; - return 0; - case GL_QUAD_STRIP: - *first = 4; - *incr = 2; - return 0; - default: - assert(0); - *first = 1; - *incr = 1; - return 0; - } -} - - -static GLuint trim( GLuint count, GLuint first, GLuint incr ) -{ - if (count < first) - return 0; - else - return count - (count - first) % incr; -} - - /** * This is a hack & will all go away. */ @@ -631,6 +195,8 @@ void draw_vb(struct draw_context *draw, */ build_vertex_headers( draw, VB ); + draw->vs_flush = vs_flush; + draw->in_vb = 1; /* tell drawing pipeline we're beginning drawing */ @@ -645,12 +211,10 @@ void draw_vb(struct draw_context *draw, vf_set_sources( draw->vf, VB->AttribPtr, 0 ); vf_emit_vertices( draw->vf, VB->Count, draw->verts ); - draw->elts = VB->Elts; - if (VB->Elts) - draw->get_vertex = get_uint_elt_vertex; + draw_set_element_buffer(draw, sizeof(GLuint), VB->Elts); else - draw->get_vertex = get_vertex; + draw_set_element_buffer(draw, 0, NULL); for (i = 0; i < VB->PrimitiveCount; i++) { const GLenum mode = VB->Primitive[i].mode; @@ -660,7 +224,7 @@ void draw_vb(struct draw_context *draw, /* Trim the primitive down to a legal size. */ draw_prim_info( mode, &first, &incr ); - length = trim( VB->Primitive[i].count, first, incr ); + length = draw_trim( VB->Primitive[i].count, first, incr ); if (!length) continue; @@ -700,12 +264,14 @@ draw_vertices(struct draw_context *draw, assert(mode <= GL_POLYGON); - draw->get_vertex = get_vertex; + draw->vs_flush = vs_flush; + draw->vertex_size = sizeof(struct vertex_header) + numAttrs * 4 * sizeof(GLfloat); - + /* no element/index buffer */ + draw_set_element_buffer(draw, 0, NULL); /*draw_prim_info(mode, &first, &incr);*/ draw_allocate_vertices( draw, numVerts ); diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 43a53f108c..fbe2064bac 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -31,7 +31,6 @@ */ -#include "main/mtypes.h" #include "main/context.h" #include "pipe/p_defines.h" @@ -43,28 +42,10 @@ #include "pipe/draw/draw_private.h" #include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_prim.h" -#define RP_NONE 0 -#define RP_POINT 1 -#define RP_LINE 2 -#define RP_TRI 3 - -static unsigned reduced_prim[GL_POLYGON + 1] = { - RP_POINT, - RP_LINE, - RP_LINE, - RP_LINE, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI -}; - - /** * Stand-in for actual vertex program execution * XXX this will probably live in a new file, like "sp_vs.c" @@ -79,19 +60,20 @@ run_vertex_program(struct draw_context *draw, struct vertex_header *vOut) { const float *vIn, *cIn; - const GLfloat *scale = draw->viewport.scale; - const GLfloat *trans = draw->viewport.translate; + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; const void *mapped = vbuffer; + /* XXX temporary hack: */ GET_CURRENT_CONTEXT(ctx); - const GLfloat *m = ctx->_ModelProjectMatrix.m; + const float *m = ctx->_ModelProjectMatrix.m; - vIn = (const float *) ((const GLubyte *) mapped + vIn = (const float *) ((const ubyte *) mapped + draw->vertex_buffer[0].buffer_offset + draw->vertex_element[0].src_offset + elem * draw->vertex_buffer[0].pitch); - cIn = (const float *) ((const GLubyte *) mapped + cIn = (const float *) ((const ubyte *) mapped + draw->vertex_buffer[3].buffer_offset + draw->vertex_element[3].src_offset + elem * draw->vertex_buffer[3].pitch); @@ -131,6 +113,10 @@ run_vertex_program(struct draw_context *draw, } +/** + * Called by the draw module when the vertx cache needs to be flushed. + * This involves running the vertex shader. + */ static void vs_flush( struct draw_context *draw ) { unsigned i; @@ -166,400 +152,6 @@ static void vs_flush( struct draw_context *draw ) } -static void draw_flush( struct draw_context *draw ) -{ - struct draw_stage *first = draw->pipeline.first; - unsigned i; - - /* Make sure all vertices are available: - */ - vs_flush( draw ); - - - switch (draw->reduced_prim) { - case RP_TRI: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - first->reset_stipple_counter( first ); - - first->tri( first, &draw->pq.queue[i] ); - } - break; - case RP_LINE: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - first->reset_stipple_counter( first ); - - first->line( first, &draw->pq.queue[i] ); - } - break; - case RP_POINT: - first->reset_stipple_counter( first ); - for (i = 0; i < draw->pq.queue_nr; i++) - first->point( first, &draw->pq.queue[i] ); - break; - } - - draw->pq.queue_nr = 0; - draw->vcache.referenced = 0; - draw->vcache.overflow = 0; -} - - -static void draw_invalidate_vcache( struct draw_context *draw ) -{ - unsigned i; - - assert(draw->pq.queue_nr == 0); - assert(draw->vs.queue_nr == 0); - assert(draw->vcache.referenced == 0); - - for (i = 0; i < Elements( draw->vcache.idx ); i++) - draw->vcache.idx[i] = ~0; -} - - -/* Return a pointer to a freshly queued primitive header. Ensure that - * there is room in the vertex cache for a maximum of "nr_verts" new - * vertices. Flush primitive and/or vertex queues if necessary to - * make space. - */ -static struct prim_header * -get_queued_prim( struct draw_context *draw, unsigned nr_verts ) -{ - if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH || - draw->vcache.overflow + nr_verts >= VCACHE_OVERFLOW) - draw_flush( draw ); - - /* The vs queue is sized so that this can never happen: - */ - assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); - - return &draw->pq.queue[draw->pq.queue_nr++]; -} - - -/* Check if vertex is in cache, otherwise add it. It won't go through - * VS yet, not until there is a flush operation or the VS queue fills up. - */ -static struct vertex_header * -get_vertex( struct draw_context *draw, unsigned i ) -{ - unsigned slot = (i + (i>>5)) & 31; - - /* Cache miss? - */ - if (draw->vcache.idx[slot] != i) { - - /* If slot is in use, use the overflow area: - */ - if (draw->vcache.referenced & (1<vcache.overflow++; - - draw->vcache.idx[slot] = i; - - /* Add to vertex shader queue: - */ - draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; - draw->vs.queue[draw->vs.queue_nr].elt = i; - draw->vs.queue_nr++; - } - - /* Mark slot as in-use: - */ - draw->vcache.referenced |= (1<vcache.vertex[slot]; -} - - - -static void -draw_set_prim( struct draw_context *draw, unsigned prim ) -{ - if (reduced_prim[prim] != draw->reduced_prim) { - draw_flush( draw ); - draw->reduced_prim = reduced_prim[prim]; - } - - draw->prim = prim; -} - - -static void do_point( struct draw_context *draw, - unsigned i0 ) -{ - struct prim_header *prim = get_queued_prim( draw, 1 ); - - prim->reset_line_stipple = 0; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->get_vertex( draw, i0 ); -} - - -static void do_line( struct draw_context *draw, - GLboolean reset_stipple, - unsigned i0, - unsigned i1 ) -{ - struct prim_header *prim = get_queued_prim( draw, 2 ); - - prim->reset_line_stipple = reset_stipple; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->get_vertex( draw, i0 ); - prim->v[1] = draw->get_vertex( draw, i1 ); -} - -static void do_triangle( struct draw_context *draw, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - - prim->reset_line_stipple = 1; - prim->edgeflags = ~0; - prim->pad = 0; - prim->v[0] = draw->get_vertex( draw, i0 ); - prim->v[1] = draw->get_vertex( draw, i1 ); - prim->v[2] = draw->get_vertex( draw, i2 ); -} - -static void do_ef_triangle( struct draw_context *draw, - GLboolean reset_stipple, - unsigned ef_mask, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - struct vertex_header *v0 = draw->get_vertex( draw, i0 ); - struct vertex_header *v1 = draw->get_vertex( draw, i1 ); - struct vertex_header *v2 = draw->get_vertex( draw, i2 ); - - prim->reset_line_stipple = reset_stipple; - - prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | - (v1->edgeflag << 1) | - (v2->edgeflag << 2)); - prim->pad = 0; - prim->v[0] = v0; - prim->v[1] = v1; - prim->v[2] = v2; -} - - -static void do_quad( struct draw_context *draw, - unsigned v0, - unsigned v1, - unsigned v2, - unsigned v3 ) -{ - do_ef_triangle( draw, 1, ~(1<<0), v0, v1, v3 ); - do_ef_triangle( draw, 0, ~(1<<1), v1, v2, v3 ); -} - - -static void draw_prim( struct draw_context *draw, - unsigned start, - unsigned count ) -{ - unsigned i; - -// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - - switch (draw->prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i ++) { - do_point( draw, - start + i ); - } - break; - - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) { - do_line( draw, - TRUE, - start + i + 0, - start + i + 1); - } - break; - - case PIPE_PRIM_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, /* XXX: only if vb not split */ - start + i - 1, - start + i ); - } - - do_line( draw, - 0, - start + count - 1, - start + 0 ); - } - break; - - case PIPE_PRIM_LINE_STRIP: - if (count >= 2) { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, - start + i - 1, - start + i ); - } - } - break; - - case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - do_ef_triangle( draw, - 1, - ~0, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - break; - - case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - do_triangle( draw, - start + i + 1, - start + i + 0, - start + i + 2 ); - } - else { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - case PIPE_PRIM_TRIANGLE_FAN: - if (count >= 3) { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - - case PIPE_PRIM_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( draw, - start + i + 0, - start + i + 1, - start + i + 2, - start + i + 3); - } - break; - - case PIPE_PRIM_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( draw, - start + i + 2, - start + i + 0, - start + i + 1, - start + i + 3); - } - break; - - case PIPE_PRIM_POLYGON: - if (count >= 3) { - unsigned ef_mask = (1<<2) | (1<<0); - - for (i = 0; i+2 < count; i++) { - - if (i + 3 >= count) - ef_mask |= (1<<1); - - do_ef_triangle( draw, - i == 0, - ef_mask, - start + i + 1, - start + i + 2, - start + i + 0); - - ef_mask &= ~(1<<2); - } - } - break; - - default: - assert(0); - break; - } -} - - - -static unsigned draw_prim_info(unsigned mode, unsigned *first, unsigned *incr) -{ - switch (mode) { - case PIPE_PRIM_POINTS: - *first = 1; - *incr = 1; - return 0; - case PIPE_PRIM_LINES: - *first = 2; - *incr = 2; - return 0; - case PIPE_PRIM_LINE_STRIP: - *first = 2; - *incr = 1; - return 0; - case PIPE_PRIM_LINE_LOOP: - *first = 2; - *incr = 1; - return 1; - case PIPE_PRIM_TRIANGLES: - *first = 3; - *incr = 3; - return 0; - case PIPE_PRIM_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - return 0; - case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: - *first = 3; - *incr = 1; - return 1; - case PIPE_PRIM_QUADS: - *first = 4; - *incr = 4; - return 0; - case PIPE_PRIM_QUAD_STRIP: - *first = 4; - *incr = 2; - return 0; - default: - assert(0); - *first = 1; - *incr = 1; - return 0; - } -} - - -static unsigned trim( unsigned count, unsigned first, unsigned incr ) -{ - if (count < first) - return 0; - else - return count - (count - first) % incr; -} - - void softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, @@ -582,15 +174,11 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, /* tell drawing pipeline we're beginning drawing */ draw->pipeline.first->begin( draw->pipeline.first ); - draw_invalidate_vcache( draw ); + draw->vs_flush = vs_flush; -#if 0 - if (VB->Elts) - draw->get_vertex = get_uint_elt_vertex; - else -#endif - draw->get_vertex = get_vertex; + draw_invalidate_vcache( draw ); + draw_set_element_buffer(draw, 0, NULL); /* no index/element buffer */ draw_set_prim( draw, mode ); /* XXX draw_prim_info() and TRIM here */ @@ -602,12 +190,9 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, /* tell drawing pipeline we're done drawing */ draw->pipeline.first->end( draw->pipeline.first ); -#if 0 - draw->verts = NULL; - draw->in_vb = 0; - draw->elts = NULL; -#endif - + /* + * unmap vertex buffer + */ pipe->winsys->buffer_unmap(pipe->winsys, buf); softpipe_unmap_surfaces(sp); @@ -626,9 +211,13 @@ do { \ } while (0) -void draw_set_vertex_attributes2( struct draw_context *draw, - const unsigned *slot_to_vf_attr, - unsigned nr_attrs ) +/** + * XXX very similar to same func in draw_vb.c (which will go away) + */ +void +draw_set_vertex_attributes2( struct draw_context *draw, + const unsigned *slot_to_vf_attr, + unsigned nr_attrs ) { unsigned i; @@ -651,12 +240,6 @@ void draw_set_vertex_attributes2( struct draw_context *draw, for (i = 1; i < nr_attrs; i++) EMIT_ATTR(slot_to_vf_attr[i], EMIT_4F, 4); -#if 0 - /* tell the vertex format module how to construct vertices for us */ - draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, - draw->nr_attrs, 0 ); -#endif - draw->vertex_size *= 4; /* floats to bytes */ } diff --git a/src/mesa/sources b/src/mesa/sources index a3b8ac34b1..d97a34f534 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -161,6 +161,7 @@ DRAW_SOURCES = \ pipe/draw/draw_cull.c \ pipe/draw/draw_flatshade.c \ pipe/draw/draw_offset.c \ + pipe/draw/draw_prim.c \ pipe/draw/draw_twoside.c \ pipe/draw/draw_unfilled.c \ pipe/draw/draw_vb.c -- cgit v1.2.3 From cbef86a6ea7fa2901d2a123544591932114ae603 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 12:40:46 +0100 Subject: Copy fragment z if the fragment shader does not write depth on its own.` --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 33 ++++++++++++++++++++++++++++++--- 1 file changed, 30 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 2ac04e8d56..2c8deac034 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -4,7 +4,7 @@ #define TGSI_DEBUG 1 /* - * Map mesa register file to SBIR register file. + * Map mesa register file to TGSI register file. */ static GLuint map_register_file( @@ -35,7 +35,7 @@ map_register_file( } /** - * Map mesa register file index to SBIR index. + * Map mesa register file index to TGSI index. * Take special care when processing input and output indices. * \param processor either TGSI_PROCESSOR_FRAGMENT or TGSI_PROCESSOR_VERTEX * \param file one of TGSI_FILE_x @@ -101,7 +101,7 @@ map_register_file_index( } /* - * Map mesa texture target to SBIR texture target. + * Map mesa texture target to TGSI texture target. */ static GLuint map_texture_target( @@ -609,6 +609,33 @@ tgsi_mesa_compile_fp_program( maxTokens - ti ); } + /* + * Copy fragment z if the shader does not write it. + */ + if( !(program->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR)) ) { + fullinst = tgsi_default_full_instruction(); + + fullinst.Instruction.Opcode = TGSI_OPCODE_MOV; + fullinst.Instruction.NumDstRegs = 1; + fullinst.Instruction.NumSrcRegs = 1; + + fulldst = &fullinst.FullDstRegisters[0]; + fulldst->DstRegister.File = TGSI_FILE_OUTPUT; + fulldst->DstRegister.Index = 0; + fulldst->DstRegister.WriteMask = TGSI_WRITEMASK_Z; + + fullsrc = &fullinst.FullSrcRegisters[0]; + fullsrc->SrcRegister.File = TGSI_FILE_INPUT; + fullsrc->SrcRegister.Index = 0; + + ti += tgsi_build_full_instruction( + &fullinst, + &tokens[ti], + header, + maxTokens - ti ); + preamble_size++; + } + for( i = 0; i < program->Base.NumInstructions; i++ ) { if( compile_instruction( &program->Base.Instructions[i], -- cgit v1.2.3 From 4fb2f09fdd61cc15db097bc7cbb6a410df075383 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 12:50:51 +0100 Subject: Add missing inlcude directory. --- src/mesa/pipe/Makefile.template | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/Makefile.template b/src/mesa/pipe/Makefile.template index fe0a81af6a..495bf44c07 100644 --- a/src/mesa/pipe/Makefile.template +++ b/src/mesa/pipe/Makefile.template @@ -15,7 +15,8 @@ OBJECTS = $(C_SOURCES:.c=.o) \ INCLUDES = \ -I. \ -I$(TOP)/src/mesa/pipe \ - -I$(TOP)/src/mesa + -I$(TOP)/src/mesa \ + -I$(TOP)/include ##### RULES ##### -- cgit v1.2.3 From fd0a6d6b4774c8cfbfccd8baa6c0ccd99a3214b9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 12:52:20 -0600 Subject: Remove GL/mesa types, rearrange things into more logical groups --- src/mesa/pipe/draw/draw_private.h | 81 ++++++++++++++++++++------------------- 1 file changed, 41 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index c1ebd34121..b6ef32855a 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -52,16 +52,16 @@ * Carry some useful information around with the vertices in the prim pipe. */ struct vertex_header { - GLuint clipmask:12; - GLuint edgeflag:1; - GLuint pad:19; + unsigned clipmask:12; + unsigned edgeflag:1; + unsigned pad:19; - GLfloat clip[4]; + float clip[4]; - GLfloat data[][4]; /* Note variable size */ + float data[][4]; /* Note variable size */ }; -#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(GLfloat)) +#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(float)) @@ -69,10 +69,10 @@ struct vertex_header { * Basic info for a point/line/triangle primitive. */ struct prim_header { - GLfloat det; /**< front/back face determinant */ - GLuint reset_line_stipple:1; - GLuint edgeflags:3; - GLuint pad:28; + float det; /**< front/back face determinant */ + unsigned reset_line_stipple:1; + unsigned edgeflags:3; + unsigned pad:28; struct vertex_header *v[3]; /**< 1 to 3 vertex pointers */ }; @@ -90,7 +90,7 @@ struct draw_stage struct draw_stage *next; /**< next stage in pipeline */ struct vertex_header **tmp; - GLuint nr_tmps; + unsigned nr_tmps; void (*begin)( struct draw_stage * ); @@ -120,6 +120,7 @@ struct draw_stage */ struct draw_context { + /** Drawing/primitive pipeline stages */ struct { struct draw_stage *first; /**< one of the following */ @@ -136,51 +137,54 @@ struct draw_context /* pipe state that we need: */ struct pipe_setup_state setup; struct pipe_viewport_state viewport; - const struct pipe_vertex_buffer *vertex_buffer; /**< note: pointer */ const struct pipe_vertex_element *vertex_element; /**< note: pointer */ /* Clip derived state: */ - GLfloat plane[12][4]; - GLuint nr_planes; - - GLuint vf_attr_to_slot[PIPE_ATTRIB_MAX]; + float plane[12][4]; + unsigned nr_planes; + /* + * Vertex attribute info + */ + unsigned vf_attr_to_slot[PIPE_ATTRIB_MAX]; struct vf_attr_map attrs[VF_ATTRIB_MAX]; - GLuint nr_attrs; - GLuint vertex_size; /**< in bytes */ - struct vertex_fetch *vf; + unsigned nr_attrs; - GLubyte *verts; - GLuint nr_vertices; - GLboolean in_vb; + unsigned vertex_size; /**< in bytes */ + unsigned nr_vertices; /** Pointer to vertex element/index buffer */ unsigned eltSize; /**< bytes per index (0, 1, 2 or 4) */ void *elts; + unsigned prim; /**< current prim type: PIPE_PRIM_x */ + unsigned reduced_prim; + + void (*vs_flush)( struct draw_context *draw ); + struct vertex_header *(*get_vertex)( struct draw_context *draw, - GLuint i ); + unsigned i ); /* Post-tnl vertex cache: */ struct { - GLuint referenced; - GLuint idx[VCACHE_SIZE + VCACHE_OVERFLOW]; + unsigned referenced; + unsigned idx[VCACHE_SIZE + VCACHE_OVERFLOW]; struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; - GLuint overflow; + unsigned overflow; } vcache; /* Vertex shader queue: */ struct { struct { - GLuint elt; + unsigned elt; struct vertex_header *dest; } queue[VS_QUEUE_LENGTH]; - GLuint queue_nr; + unsigned queue_nr; } vs; /* Prim pipeline queue: @@ -191,21 +195,18 @@ struct draw_context * transformed by a vs queue flush. */ struct prim_header queue[PRIM_QUEUE_LENGTH]; - GLuint queue_nr; + unsigned queue_nr; } pq; - - GLenum prim; /**< GL_POINTS, GL_LINE_STRIP, GL_QUADS, etc */ - unsigned reduced_prim; - - void (*vs_flush)( struct draw_context *draw ); - - /* Helper for tnl: + /* Misc for draw_vb.c (XXX temporary) */ - GLvector4f header; + GLvector4f header; + ubyte *verts; + boolean in_vb; + struct vertex_fetch *vf; - /* helper for sp_draw_arrays.c - temporary? */ + /* Misc for sp_draw_arrays.c (temporary?) */ void *mapped_vbuffer; }; @@ -220,7 +221,7 @@ extern struct draw_stage *draw_cull_stage( struct draw_context *context ); extern void draw_free_tmps( struct draw_stage *stage ); -extern void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ); +extern void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ); @@ -234,7 +235,7 @@ extern void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ); static INLINE struct vertex_header * dup_vert( struct draw_stage *stage, const struct vertex_header *vert, - GLuint idx ) + unsigned idx ) { struct vertex_header *tmp = stage->tmp[idx]; memcpy(tmp, vert, stage->draw->vertex_size ); -- cgit v1.2.3 From 717cc0f214b741bc9ef679dd372654d8e2192f25 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 12:55:47 -0600 Subject: move the draw_alloc/free_tmps() functions to draw_prim.c --- src/mesa/pipe/draw/draw_prim.c | 27 +++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_vb.c | 24 ------------------------ 2 files changed, 27 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index b7d30c567a..931404caff 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -517,3 +517,30 @@ draw_trim( unsigned count, unsigned first, unsigned incr ) else return count - (count - first) % incr; } + + +/** + * Allocate space for temporary post-transform vertices, such as for clipping. + */ +void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ) +{ + stage->nr_tmps = nr; + + if (nr) { + GLubyte *store = (GLubyte *) malloc(MAX_VERTEX_SIZE * nr); + GLuint i; + + stage->tmp = (struct vertex_header **) malloc(sizeof(struct vertex_header *) * nr); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } +} + +void draw_free_tmps( struct draw_stage *stage ) +{ + if (stage->tmp) { + free(stage->tmp[0]); + free(stage->tmp); + } +} diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index a88babc44e..db7b217c2f 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -371,27 +371,3 @@ void draw_set_vertex_attributes( struct draw_context *draw, draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); } - - -void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ) -{ - stage->nr_tmps = nr; - - if (nr) { - GLubyte *store = (GLubyte *) malloc(MAX_VERTEX_SIZE * nr); - GLuint i; - - stage->tmp = MALLOC(sizeof(struct vertex_header *) * nr); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); - } -} - -void draw_free_tmps( struct draw_stage *stage ) -{ - if (stage->tmp) { - free(stage->tmp[0]); - free(stage->tmp); - } -} -- cgit v1.2.3 From ebe510766927315ec6dcd846a980cb851b5331d4 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 13:12:24 -0600 Subject: disable Z value scaling - it's done in sp_quad_depth_test.c --- src/mesa/pipe/tgsi/core/tgsi_exec.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c index 0ffed20066..a1f46712b4 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/core/tgsi_exec.c @@ -1128,7 +1128,7 @@ exec_declaration( { if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask, i, j; + unsigned first, last, mask; interpolation_func interp; assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); @@ -2299,7 +2299,8 @@ tgsi_exec_machine_run2( tgsi_parse_free (&parse); #endif -#if MESA +#if 0 + /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { /* * Scale back depth component. -- cgit v1.2.3 From de653b4c9bddcec46f3ddf411ec082dd178d7b38 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 13:33:43 -0600 Subject: Begin added vertex shader state/support. Renamed pipe_fs_state to pipe_shader_state since it can be used for both vertex and fragment shader info. --- src/mesa/pipe/i915simple/i915_context.h | 2 +- src/mesa/pipe/i915simple/i915_state.c | 2 +- src/mesa/pipe/p_context.h | 5 +- src/mesa/pipe/p_state.h | 4 +- src/mesa/pipe/softpipe/sp_context.h | 4 +- src/mesa/pipe/softpipe/sp_state.h | 5 +- src/mesa/pipe/softpipe/sp_state_fs.c | 14 +++--- src/mesa/state_tracker/st_atom_fs.c | 83 ++++++++++++++++++++++++++++++- src/mesa/state_tracker/st_cb_clear.c | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_context.h | 3 +- src/mesa/state_tracker/st_program.h | 2 + 12 files changed, 111 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index d69657daa9..1e48485c56 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -105,7 +105,7 @@ struct i915_context struct pipe_clip_state clip; struct pipe_depth_state depth_test; struct pipe_framebuffer_state framebuffer; - struct pipe_fs_state fs; + struct pipe_shader_state fs; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index f874b21016..06fa716c4f 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -111,7 +111,7 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe, static void i915_set_fs_state( struct pipe_context *pipe, - const struct pipe_fs_state *fs ) + const struct pipe_shader_state *fs ) { struct i915_context *i915 = i915_context(pipe); diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 6679aae089..320b4877c3 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -117,7 +117,10 @@ struct pipe_context { const struct pipe_framebuffer_state * ); void (*set_fs_state)( struct pipe_context *, - const struct pipe_fs_state * ); + const struct pipe_shader_state * ); + + void (*set_vs_state)( struct pipe_context *, + const struct pipe_shader_state * ); void (*set_polygon_stipple)( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 4878cb41bd..c1972c56b3 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -121,8 +121,8 @@ struct pipe_constant_buffer { }; -struct pipe_fs_state { - unsigned inputs_read; /* FRAG_ATTRIB_* */ +struct pipe_shader_state { + unsigned inputs_read; /* FRAG/VERT_ATTRIB_* */ const struct tgsi_token *tokens; struct pipe_constant_buffer *constants; /* XXX temporary? */ }; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index f45cb6fe2c..8f184c644a 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -63,6 +63,7 @@ enum interp_mode { #define SP_NEW_TEXTURE 0x800 #define SP_NEW_STENCIL 0x1000 #define SP_NEW_VERTEX 0x2000 +#define SP_NEW_VS 0x4000 struct softpipe_context { @@ -79,7 +80,8 @@ struct softpipe_context { struct pipe_clip_state clip; struct pipe_depth_state depth_test; struct pipe_framebuffer_state framebuffer; - struct pipe_fs_state fs; + struct pipe_shader_state fs; + struct pipe_shader_state vs; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index d754acaa38..f25a20a5e3 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -56,7 +56,10 @@ void softpipe_set_depth_test_state( struct pipe_context *, const struct pipe_depth_state * ); void softpipe_set_fs_state( struct pipe_context *, - const struct pipe_fs_state * ); + const struct pipe_shader_state * ); + +void softpipe_set_vs_state( struct pipe_context *, + const struct pipe_shader_state * ); void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index c7ef1f1cfc..3505c2f1fb 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -31,7 +31,7 @@ void softpipe_set_fs_state( struct pipe_context *pipe, - const struct pipe_fs_state *fs ) + const struct pipe_shader_state *fs ) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -41,10 +41,12 @@ void softpipe_set_fs_state( struct pipe_context *pipe, } +void softpipe_set_vs_state( struct pipe_context *pipe, + const struct pipe_shader_state *vs ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + memcpy(&softpipe->vs, vs, sizeof(*vs)); - - - - - + softpipe->dirty |= SP_NEW_VS; +} diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 58875de3f9..9731ab6cee 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -39,6 +39,11 @@ #define TGSI_DEBUG 0 + +/** + ** Fragment programs + **/ + static void compile_fs( struct st_context *st, struct st_fragment_program *fs ) { @@ -53,7 +58,7 @@ static void compile_fs( struct st_context *st, static void update_fs( struct st_context *st ) { - struct pipe_fs_state fs; + struct pipe_shader_state fs; struct st_fragment_program *fp = NULL; struct gl_program_parameter_list *params = NULL; @@ -103,3 +108,79 @@ const struct st_tracked_state st_update_fs = { }, .update = update_fs }; + + + +/** + ** Vertex programs + **/ + + +static void compile_vs( struct st_context *st, + struct st_vertex_program *vs ) +{ + /* XXX: fix static allocation of tokens: + */ + tgsi_mesa_compile_vp_program( &vs->Base, vs->tokens, ST_FP_MAX_TOKENS ); + + if (TGSI_DEBUG) + tgsi_dump( vs->tokens, TGSI_DUMP_VERBOSE ); +} + + +static void update_vs( struct st_context *st ) +{ + struct pipe_shader_state vs; + struct st_vertex_program *vp = NULL; + struct gl_program_parameter_list *params = NULL; + + if (st->ctx->Shader.CurrentProgram && + st->ctx->Shader.CurrentProgram->LinkStatus && + st->ctx->Shader.CurrentProgram->VertexProgram) { + struct gl_vertex_program *f + = st->ctx->Shader.CurrentProgram->VertexProgram; + vp = st_vertex_program(f); + params = f->Base.Parameters; + } + else if (st->ctx->VertexProgram._Current) { + vp = st_vertex_program(st->ctx->VertexProgram._Current); + params = st->ctx->VertexProgram._Current->Base.Parameters; + } + + /* XXXX temp */ + if (!vp) + return; + + if (vp && params) { + /* load program's constants array */ + vp->constants.nr_constants = params->NumParameters; + memcpy(vp->constants.constant, + params->ParameterValues, + params->NumParameters * sizeof(GLfloat) * 4); + } + + if (vp->dirty) + compile_vs( st, vp ); + + memset( &vs, 0, sizeof(vs) ); + vs.inputs_read = vp->Base.Base.InputsRead; + vs.tokens = &vp->tokens[0]; + vs.constants = &vp->constants; + + if (memcmp(&vs, &st->state.vs, sizeof(vs)) != 0 || + vp->dirty) + { + vp->dirty = 0; + st->state.vs = vs; + st->pipe->set_vs_state(st->pipe, &vs); + } +} + + +const struct st_tracked_state st_update_vs = { + .dirty = { + .mesa = _NEW_PROGRAM, + .st = ST_NEW_VERTEX_PROGRAM, + }, + .update = update_vs +}; diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 2f7ade73e4..3777c53d48 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -278,7 +278,7 @@ clear_with_quad(GLcontext *ctx, /* fragment shader state: color pass-through program */ { static struct st_fragment_program *stfp = NULL; - struct pipe_fs_state fs; + struct pipe_shader_state fs; if (!stfp) { stfp = make_color_shader(st); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index dd27760a58..11261f1d99 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -272,7 +272,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* fragment shader state: color pass-through program */ { static struct st_fragment_program *stfp = NULL; - struct pipe_fs_state fs; + struct pipe_shader_state fs; if (!stfp) { stfp = make_drawpixels_shader(ctx->st); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 13843d9b7f..25ed20a5f8 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -73,7 +73,8 @@ struct st_context struct pipe_clip_state clip; struct pipe_depth_state depth; struct pipe_framebuffer_state framebuffer; - struct pipe_fs_state fs; + struct pipe_shader_state fs; + struct pipe_shader_state vs; struct pipe_poly_stipple poly_stipple; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_scissor_state scissor; diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 8dcb2ceb48..b077fdf069 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -83,7 +83,9 @@ struct st_vertex_program * ProgramStringNotify changes. */ + struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; + struct pipe_constant_buffer constants; GLuint param_state; }; -- cgit v1.2.3 From 9ba4311e71f510797e1b446ab7dc6b6d4dc55dd5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 17:30:38 -0600 Subject: check for null swrast (pipe work-around) --- src/mesa/swrast/s_context.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/swrast/s_context.c b/src/mesa/swrast/s_context.c index 3956925651..f43fa61f8a 100644 --- a/src/mesa/swrast/s_context.c +++ b/src/mesa/swrast/s_context.c @@ -524,6 +524,9 @@ _swrast_update_texture_samplers(GLcontext *ctx) SWcontext *swrast = SWRAST_CONTEXT(ctx); GLuint u; + if (!swrast) + return; /* pipe hack */ + for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++) { const struct gl_texture_object *tObj = ctx->Texture.Unit[u]._Current; /* Note: If tObj is NULL, the sample function will be a simple -- cgit v1.2.3 From 78d6d5e8fe7281f67b7ce8d9911e4f280435b1f5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 17:32:24 -0600 Subject: make pass-through vertex shader --- src/mesa/state_tracker/st_cb_clear.c | 54 ++++++++++++++++++++++++++++++++++++ 1 file changed, 54 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 3777c53d48..492073e840 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -151,9 +151,63 @@ make_color_shader(struct st_context *st) } +/** + * Create a simple vertex shader that just passes through the + * vertex position and color. + */ +static struct st_vertex_program * +make_vertex_shader(struct st_context *st) +{ + GLcontext *ctx = st->ctx; + struct st_vertex_program *stvp; + struct gl_program *p; + GLboolean b; + + p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); + if (!p) + return NULL; + + p->NumInstructions = 3; + p->Instructions = _mesa_alloc_instructions(3); + if (!p->Instructions) { + ctx->Driver.DeleteProgram(ctx, p); + return NULL; + } + _mesa_init_instructions(p->Instructions, 3); + /* MOV result.pos, vertex.pos; */ + p->Instructions[0].Opcode = OPCODE_MOV; + p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[0].DstReg.Index = VERT_RESULT_HPOS; + p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[0].SrcReg[0].Index = VERT_ATTRIB_POS; + /* MOV result.color, vertex.color; */ + p->Instructions[1].Opcode = OPCODE_MOV; + p->Instructions[1].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[1].DstReg.Index = VERT_RESULT_COL0; + p->Instructions[1].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[1].SrcReg[0].Index = VERT_ATTRIB_COLOR0; + /* END; */ + p->Instructions[2].Opcode = OPCODE_END; + + p->InputsRead = VERT_BIT_POS | VERT_BIT_COLOR0; + p->OutputsWritten = ((1 << VERT_RESULT_COL0) | + (1 << VERT_RESULT_HPOS)); + + stvp = (struct st_vertex_program *) p; + /* compile into tgsi format */ + b = tgsi_mesa_compile_vp_program(&stvp->Base, + stvp->tokens, ST_FP_MAX_TOKENS); + assert(b); + + return stvp; +} + + + /** * Draw a screen-aligned quadrilateral. * Coords are window coords. + * XXX need to emit clip coords when using a vertex program! */ static void draw_quad(GLcontext *ctx, -- cgit v1.2.3 From bf55f32c6fd74b8eb0b31dce0ebd8f4c52f57743 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 17:33:05 -0600 Subject: run real vertex shaders, a total hack for now --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 163 ++++++++++++++++++++++++++++++++ 1 file changed, 163 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index fbe2064bac..c8e161cfb1 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -44,6 +44,163 @@ #include "pipe/draw/draw_context.h" #include "pipe/draw/draw_prim.h" +#include "pipe/tgsi/core/tgsi_exec.h" +#include "pipe/tgsi/core/tgsi_build.h" +#include "pipe/tgsi/core/tgsi_util.h" + + +#if defined __GNUC__ +#define USE_ALIGNED_ATTRIBS 1 +#define ALIGN16_SUFFIX __attribute__(( aligned( 16 ) )) +#else +#define USE_ALIGNED_ATTRIBS 0 +#define ALIGN16_SUFFIX +#endif + + +static struct softpipe_context *sp_global = NULL; + + +static void +run_vertex_program2(struct draw_context *draw, + const void *vbuffer, unsigned elem, + struct vertex_header *vOut) +{ +#if 1 + struct softpipe_context *sp = sp_global; +#endif + struct tgsi_exec_machine machine; + int i; + +#if USE_ALIGNED_ATTRIBS + struct tgsi_exec_vector inputs[PIPE_ATTRIB_MAX] ALIGN16_SUFFIX; + struct tgsi_exec_vector outputs[PIPE_ATTRIB_MAX] ALIGN16_SUFFIX; +#else + struct tgsi_exec_vector inputs[PIPE_ATTRIB_MAX + 1]; + struct tgsi_exec_vector outputs[PIPE_ATTRIB_MAX + 1]; +#endif + +#ifdef DEBUG + memset( &machine, 0, sizeof( machine ) ); +#endif + + /* init machine state */ + tgsi_exec_machine_init( + &machine, + sp->vs.tokens, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/ ); + + /* Consts does not require 16 byte alignment. */ + machine.Consts = sp->vs.constants->constant; + +#if USE_ALIGNED_ATTRIBS + machine.Inputs = inputs; + machine.Outputs = outputs; +#else + machine.Inputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); + machine.Outputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs ); +#endif + + { + const void *mapped = vbuffer; + const float *vIn, *cIn; + vIn = (const float *) ((const ubyte *) mapped + + draw->vertex_buffer[0].buffer_offset + + draw->vertex_element[0].src_offset + + elem * draw->vertex_buffer[0].pitch); + + cIn = (const float *) ((const ubyte *) mapped + + draw->vertex_buffer[3].buffer_offset + + draw->vertex_element[3].src_offset + + elem * draw->vertex_buffer[3].pitch); + /*X*/ + machine.Inputs[0].xyzw[0].f[0] = vIn[0]; + machine.Inputs[0].xyzw[0].f[1] = vIn[0]; + machine.Inputs[0].xyzw[0].f[2] = vIn[0]; + machine.Inputs[0].xyzw[0].f[3] = vIn[0]; + + /*Y*/ + machine.Inputs[0].xyzw[1].f[0] = vIn[1]; + machine.Inputs[0].xyzw[1].f[1] = vIn[1]; + machine.Inputs[0].xyzw[1].f[2] = vIn[1]; + machine.Inputs[0].xyzw[1].f[3] = vIn[1]; + + /*Z*/ + machine.Inputs[0].xyzw[2].f[0] = vIn[2]; + machine.Inputs[0].xyzw[2].f[1] = vIn[2]; + machine.Inputs[0].xyzw[2].f[2] = vIn[2]; + machine.Inputs[0].xyzw[2].f[3] = vIn[2]; + + /*W*/ + machine.Inputs[0].xyzw[3].f[0] = 1.0; + machine.Inputs[0].xyzw[3].f[1] = 1.0; + machine.Inputs[0].xyzw[3].f[2] = 1.0; + machine.Inputs[0].xyzw[3].f[3] = 1.0; + + printf("VS Input: %f %f %f %f\n", + vIn[0], vIn[1], vIn[2], 1.0); + } + + printf("Consts:\n"); + for (i = 0; i < 4; i++) { + printf(" %d: %f %f %f %f\n", i, + machine.Consts[i][0], + machine.Consts[i][1], + machine.Consts[i][2], + machine.Consts[i][3]); + } + + + /* run shader */ + tgsi_exec_machine_run( &machine ); + + /* store result pos */ + printf("VS result: %f %f %f %f\n", + outputs[0].xyzw[0].f[0], + outputs[0].xyzw[1].f[0], + outputs[0].xyzw[2].f[0], + outputs[0].xyzw[3].f[0]); + { + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + float x, y, z, w; + + x = outputs[0].xyzw[0].f[0]; + y = outputs[0].xyzw[1].f[0]; + z = outputs[0].xyzw[2].f[0]; + w = outputs[0].xyzw[3].f[0]; + + /* divide by w */ + x /= w; + y /= w; + z /= w; + w = 1.0 / w; + + /* Viewport */ + vOut->data[0][0] = scale[0] * x + trans[0]; + vOut->data[0][1] = scale[1] * y + trans[1]; + vOut->data[0][2] = scale[2] * z + trans[2]; + vOut->data[0][3] = w; + printf("wincoord: %f %f %f\n", + vOut->data[0][0], + vOut->data[0][1], + vOut->data[0][2]); + + vOut->data[1][0] = 1.0; + vOut->data[1][1] = 1.0; + vOut->data[1][2] = 1.0; + vOut->data[1][3] = 1.0; + + } + +#if 0 + memcpy( + quad->outputs.color, + &machine.Outputs[1].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); +#endif +} /** @@ -59,6 +216,9 @@ run_vertex_program(struct draw_context *draw, const void *vbuffer, unsigned elem, struct vertex_header *vOut) { + run_vertex_program2(draw, vbuffer, elem, vOut); + +#if 0 const float *vIn, *cIn; const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; @@ -110,6 +270,7 @@ run_vertex_program(struct draw_context *draw, vOut->data[1][2] = cIn[2]; vOut->data[1][3] = 1.0; } +#endif } @@ -161,6 +322,8 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, struct draw_context *draw = sp->draw; struct pipe_buffer_handle *buf; + sp_global = sp; + softpipe_map_surfaces(sp); /* -- cgit v1.2.3 From e3bdd66bf69fb12fa02d161989755944077b280d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 17:33:49 -0600 Subject: misc changes to support vertex shaders (disabled by default) --- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/state_tracker/st_atom_fs.c | 11 +++++++++-- src/mesa/state_tracker/st_context.c | 3 +++ src/mesa/state_tracker/st_draw.c | 13 +++++++++++-- 4 files changed, 24 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 4c0dacd458..3f7044825a 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -269,6 +269,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_depth_state = softpipe_set_depth_test_state; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_fs_state = softpipe_set_fs_state; + softpipe->pipe.set_vs_state = softpipe_set_vs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.set_sampler_state = softpipe_set_sampler_state; softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 9731ab6cee..9dddb18498 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -134,6 +134,9 @@ static void update_vs( struct st_context *st ) struct st_vertex_program *vp = NULL; struct gl_program_parameter_list *params = NULL; + if (st->ctx->VertexProgram._MaintainTnlProgram) + _tnl_UpdateFixedFunctionProgram( st->ctx ); + if (st->ctx->Shader.CurrentProgram && st->ctx->Shader.CurrentProgram->LinkStatus && st->ctx->Shader.CurrentProgram->VertexProgram) { @@ -148,11 +151,15 @@ static void update_vs( struct st_context *st ) } /* XXXX temp */ +#if 1 if (!vp) return; - +#endif if (vp && params) { /* load program's constants array */ + + _mesa_load_state_parameters(st->ctx, params); + vp->constants.nr_constants = params->NumParameters; memcpy(vp->constants.constant, params->ParameterValues, @@ -179,7 +186,7 @@ static void update_vs( struct st_context *st ) const struct st_tracked_state st_update_vs = { .dirty = { - .mesa = _NEW_PROGRAM, + .mesa = _NEW_PROGRAM | _NEW_MODELVIEW, .st = ST_NEW_VERTEX_PROGRAM, }, .update = update_vs diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 2a85ce6baa..35a262ab5b 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -73,6 +73,9 @@ struct st_context *st_create_context( GLcontext *ctx, st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; st->ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; + st->ctx->VertexProgram._MaintainTnlProgram = GL_TRUE; + + #if 0 st_init_cb_clear( st ); st_init_cb_program( st ); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 0b13432256..d66a0ed851 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -36,6 +36,7 @@ #include "tnl/t_context.h" #include "tnl/t_pipeline.h" +#include "tnl/t_vp_build.h" /* USE_NEW_DRAW */ #include "st_context.h" #include "st_atom.h" @@ -89,6 +90,9 @@ const struct tnl_pipeline_stage st_draw = { }; static const struct tnl_pipeline_stage *st_pipeline[] = { +#if USE_NEW_DRAW + &_tnl_vertex_program_stage, +#else &_tnl_vertex_transform_stage, &_tnl_vertex_cull_stage, &_tnl_normal_transform_stage, @@ -98,6 +102,7 @@ static const struct tnl_pipeline_stage *st_pipeline[] = { &_tnl_texture_transform_stage, &_tnl_point_attenuation_stage, &_tnl_vertex_program_stage, +#endif &st_draw, /* ADD: escape to pipe */ 0, }; @@ -146,6 +151,9 @@ draw_vbo(GLcontext *ctx, struct pipe_context *pipe = ctx->st->pipe; GLuint attr, i; + st_validate_state(ctx->st); + + /* tell pipe about the vertex array element/attributes */ for (attr = 0; attr < 16; attr++) { struct gl_buffer_object *bufobj = arrays[attr]->BufferObj; @@ -205,11 +213,12 @@ void st_init_draw( struct st_context *st ) assert(vbo->draw_prims); vbo->draw_prims = draw_vbo; -#else +#endif _tnl_destroy_pipeline( ctx ); _tnl_install_pipeline( ctx, st_pipeline ); -#endif + /* USE_NEW_DRAW */ + _tnl_ProgramCacheInit( ctx ); } -- cgit v1.2.3 From 3fc926f3740da9ec27853d158243055f3cb43d43 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 16 Aug 2007 18:11:55 -0600 Subject: Remove many dependencies on mesa headers. To build with mesa, need -DMESA in makefile/config file. --- src/mesa/pipe/draw/draw_clip.c | 110 ++++++++++++++++++--------------- src/mesa/pipe/draw/draw_context.c | 17 +++-- src/mesa/pipe/draw/draw_context.h | 15 +++-- src/mesa/pipe/draw/draw_cull.c | 20 +++--- src/mesa/pipe/draw/draw_flatshade.c | 8 +-- src/mesa/pipe/draw/draw_offset.c | 39 ++++++------ src/mesa/pipe/draw/draw_prim.c | 11 ++-- src/mesa/pipe/draw/draw_private.h | 25 +++++++- src/mesa/pipe/draw/draw_twoside.c | 12 ++-- src/mesa/pipe/draw/draw_unfilled.c | 6 +- src/mesa/pipe/draw/draw_vb.c | 66 ++++++++++++-------- src/mesa/pipe/p_util.h | 12 ++++ src/mesa/pipe/softpipe/sp_context.c | 5 ++ src/mesa/pipe/softpipe/sp_prim_setup.c | 5 ++ src/mesa/state_tracker/st_draw.c | 2 +- 15 files changed, 215 insertions(+), 138 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index f8bacf8a81..53a7b10f7d 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -32,15 +32,27 @@ */ -#include "main/macros.h" +#include "pipe/p_util.h" #include "draw_private.h" +#ifndef IS_NEGATIVE +#define IS_NEGATIVE(X) ((X) < 0.0) +#endif + +#ifndef DIFFERENT_SIGNS +#define DIFFERENT_SIGNS(x, y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F) +#endif + +#ifndef MAX_CLIPPED_VERTICES +#define MAX_CLIPPED_VERTICES ((2 * (6 + PIPE_MAX_CLIP_PLANES))+1) +#endif + struct clipper { struct draw_stage stage; /**< base class */ - GLuint active_user_planes; - GLfloat (*plane)[4]; + unsigned active_user_planes; + float (*plane)[4]; }; @@ -57,10 +69,10 @@ static INLINE struct clipper *clipper_stage( struct draw_stage *stage ) /* All attributes are float[4], so this is easy: */ -static void interp_attr( GLfloat *fdst, - GLfloat t, - const GLfloat *fin, - const GLfloat *fout ) +static void interp_attr( float *fdst, + float t, + const float *fin, + const float *fout ) { fdst[0] = LINTERP( t, fout[0], fin[0] ); fdst[1] = LINTERP( t, fout[1], fin[1] ); @@ -75,12 +87,12 @@ static void interp_attr( GLfloat *fdst, */ static void interp( const struct clipper *clip, struct vertex_header *dst, - GLfloat t, + float t, const struct vertex_header *out, const struct vertex_header *in ) { - const GLuint nr_attrs = clip->stage.draw->nr_attrs; - GLuint j; + const unsigned nr_attrs = clip->stage.draw->nr_attrs; + unsigned j; /* Vertex header. */ @@ -99,10 +111,10 @@ static void interp( const struct clipper *clip, /* Do the projective divide and insert window coordinates: */ { - const GLfloat *pos = dst->clip; - const GLfloat *scale = clip->stage.draw->viewport.scale; - const GLfloat *trans = clip->stage.draw->viewport.translate; - const GLfloat oow = 1.0 / pos[3]; + const float *pos = dst->clip; + const float *scale = clip->stage.draw->viewport.scale; + const float *trans = clip->stage.draw->viewport.translate; + const float oow = 1.0 / pos[3]; dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; @@ -128,10 +140,10 @@ static void interp( const struct clipper *clip, #define CLIP_CULL_BIT 0x80 -static INLINE GLfloat dot4( const GLfloat *a, - const GLfloat *b ) +static INLINE float dot4( const float *a, + const float *b ) { - GLfloat result = (a[0]*b[0] + + float result = (a[0]*b[0] + a[1]*b[1] + a[2]*b[2] + a[3]*b[3]); @@ -144,9 +156,9 @@ static INLINE GLfloat dot4( const GLfloat *a, static INLINE void do_tri( struct draw_stage *next, struct prim_header *header ) { - GLuint i; + unsigned i; for (i = 0; i < 3; i++) { - GLfloat *ndc = header->v[i]->data[0]; + float *ndc = header->v[i]->data[0]; _mesa_printf("ndc %f %f %f\n", ndc[0], ndc[1], ndc[2]); assert(ndc[0] >= -1 && ndc[0] <= 641); assert(ndc[1] >= 30 && ndc[1] <= 481); @@ -159,10 +171,10 @@ static INLINE void do_tri( struct draw_stage *next, static void emit_poly( struct draw_stage *stage, struct vertex_header **inlist, - GLuint n ) + unsigned n ) { struct prim_header header; - GLuint i; + unsigned i; for (i = 2; i < n; i++) { header.v[0] = inlist[0]; @@ -170,8 +182,8 @@ static void emit_poly( struct draw_stage *stage, header.v[2] = inlist[i]; { - GLuint tmp0 = header.v[0]->edgeflag; - GLuint tmp2 = header.v[2]->edgeflag; + unsigned tmp0 = header.v[0]->edgeflag; + unsigned tmp2 = header.v[2]->edgeflag; if (i != 2) header.v[0]->edgeflag = 0; if (i != n-1) header.v[2]->edgeflag = 0; @@ -188,7 +200,7 @@ static void emit_poly( struct draw_stage *stage, #if 0 static void emit_poly( struct draw_stage *stage ) { - GLuint i; + unsigned i; for (i = 2; i < n; i++) { header->v[0] = inlist[0]; @@ -206,16 +218,16 @@ static void emit_poly( struct draw_stage *stage ) static void do_clip_tri( struct draw_stage *stage, struct prim_header *header, - GLuint clipmask ) + unsigned clipmask ) { struct clipper *clipper = clipper_stage( stage ); struct vertex_header *a[MAX_CLIPPED_VERTICES]; struct vertex_header *b[MAX_CLIPPED_VERTICES]; struct vertex_header **inlist = a; struct vertex_header **outlist = b; - GLuint tmpnr = 0; - GLuint n = 3; - GLuint i; + unsigned tmpnr = 0; + unsigned n = 3; + unsigned i; inlist[0] = header->v[0]; inlist[1] = header->v[1]; @@ -231,11 +243,11 @@ do_clip_tri( struct draw_stage *stage, } while (clipmask && n >= 3) { - GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->plane[plane_idx]; + unsigned plane_idx = ffs(clipmask)-1; + const float *plane = clipper->plane[plane_idx]; struct vertex_header *vert_prev = inlist[0]; - GLfloat dp_prev = dot4( vert_prev->clip, plane ); - GLuint outcount = 0; + float dp_prev = dot4( vert_prev->clip, plane ); + unsigned outcount = 0; clipmask &= ~(1<clip, plane ); + float dp = dot4( vert->clip, plane ); if (!IS_NEGATIVE(dp_prev)) { outlist[outcount++] = vert_prev; @@ -258,7 +270,7 @@ do_clip_tri( struct draw_stage *stage, /* Going out of bounds. Avoid division by zero as we * know dp != dp_prev from DIFFERENT_SIGNS, above. */ - GLfloat t = dp / (dp - dp_prev); + float t = dp / (dp - dp_prev); interp( clipper, new_vert, t, vert, vert_prev ); /* Force edgeflag true in this case: @@ -267,7 +279,7 @@ do_clip_tri( struct draw_stage *stage, } else { /* Coming back in. */ - GLfloat t = dp_prev / (dp_prev - dp); + float t = dp_prev / (dp_prev - dp); interp( clipper, new_vert, t, vert_prev, vert ); /* Copy starting vert's edgeflag: @@ -300,15 +312,15 @@ do_clip_tri( struct draw_stage *stage, static void do_clip_line( struct draw_stage *stage, struct prim_header *header, - GLuint clipmask ) + unsigned clipmask ) { const struct clipper *clipper = clipper_stage( stage ); struct vertex_header *v0 = header->v[0]; struct vertex_header *v1 = header->v[1]; - const GLfloat *pos0 = v0->clip; - const GLfloat *pos1 = v1->clip; - GLfloat t0 = 0; - GLfloat t1 = 0; + const float *pos0 = v0->clip; + const float *pos1 = v1->clip; + float t0 = 0; + float t1 = 0; struct prim_header newprim; /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove @@ -321,18 +333,18 @@ do_clip_line( struct draw_stage *stage, } while (clipmask) { - const GLuint plane_idx = ffs(clipmask)-1; - const GLfloat *plane = clipper->plane[plane_idx]; - const GLfloat dp0 = dot4( pos0, plane ); - const GLfloat dp1 = dot4( pos1, plane ); + const unsigned plane_idx = ffs(clipmask)-1; + const float *plane = clipper->plane[plane_idx]; + const float dp0 = dot4( pos0, plane ); + const float dp1 = dot4( pos1, plane ); if (dp1 < 0) { - GLfloat t = dp1 / (dp1 - dp0); + float t = dp1 / (dp1 - dp0); t1 = MAX2(t1, t); } if (dp0 < 0) { - GLfloat t = dp0 / (dp0 - dp1); + float t = dp0 / (dp0 - dp1); t0 = MAX2(t0, t); } @@ -365,7 +377,7 @@ do_clip_line( struct draw_stage *stage, static void clip_begin( struct draw_stage *stage ) { struct clipper *clipper = clipper_stage(stage); - GLuint nr = stage->draw->nr_planes; + unsigned nr = stage->draw->nr_planes; /* sanity checks. If these fail, review the clip/interp code! */ assert(stage->draw->nr_attrs >= 3); @@ -393,7 +405,7 @@ static void clip_line( struct draw_stage *stage, struct prim_header *header ) { - GLuint clipmask = (header->v[0]->clipmask | + unsigned clipmask = (header->v[0]->clipmask | header->v[1]->clipmask); if (clipmask == 0) { @@ -411,7 +423,7 @@ static void clip_tri( struct draw_stage *stage, struct prim_header *header ) { - GLuint clipmask = (header->v[0]->clipmask | + unsigned clipmask = (header->v[0]->clipmask | header->v[1]->clipmask | header->v[2]->clipmask); diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index f2ee6c52e0..7e6a95127a 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -30,13 +30,15 @@ * Keith Whitwell */ -#include "imports.h" -#include "macros.h" +#ifdef MESA +#include "main/macros.h" +#else +#include "pipe/p_util.h" +#endif #include "draw_context.h" #include "draw_private.h" - struct draw_context *draw_create( void ) { struct draw_context *draw = CALLOC_STRUCT( draw_context ); @@ -57,7 +59,9 @@ struct draw_context *draw_create( void ) ASSIGN_4V( draw->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ draw->nr_planes = 6; +#ifdef MESA draw->vf = vf_create( GL_TRUE ); +#endif /* Statically allocate maximum sized vertices for the cache - could be cleverer... */ @@ -75,10 +79,13 @@ struct draw_context *draw_create( void ) void draw_destroy( struct draw_context *draw ) { - if (draw->header.storage) +#ifdef MESA + if (draw->header.storage) { ALIGN_FREE( draw->header.storage ); + } vf_destroy( draw->vf ); +#endif free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ free( draw ); @@ -186,7 +193,9 @@ void draw_set_viewport_state( struct draw_context *draw, { draw->viewport = *viewport; /* struct copy */ +#ifdef MESA vf_set_vp_scale_translate( draw->vf, viewport->scale, viewport->translate ); +#endif /* Using tnl/ and vf/ modules is temporary while getting started. * Full pipe will have vertex shader, vertex fetch of its own. diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index beba9429c9..ac994d6419 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -38,7 +38,6 @@ #define DRAW_CONTEXT_H -#include "main/glheader.h" #include "pipe/p_state.h" @@ -64,13 +63,13 @@ void draw_set_setup_stage( struct draw_context *draw, struct draw_stage *stage ); void draw_set_vertex_attributes( struct draw_context *draw, - const GLuint *attrs, - GLuint nr_attrs ); + const unsigned *attrs, + unsigned nr_attrs ); /* XXX temporary */ void draw_set_vertex_attributes2( struct draw_context *draw, - const GLuint *attrs, - GLuint nr_attrs ); + const unsigned *attrs, + unsigned nr_attrs ); void draw_set_vertex_array_info(struct draw_context *draw, const struct pipe_vertex_buffer *buffers, @@ -81,9 +80,9 @@ void draw_vb(struct draw_context *draw, struct vertex_buffer *VB ); void draw_vertices(struct draw_context *draw, - GLuint mode, - GLuint numVertex, const GLfloat *verts, - GLuint numAttribs, const GLuint attribs[]); + unsigned mode, + unsigned numVertex, const float *verts, + unsigned numAttribs, const unsigned attribs[]); #endif /* DRAW_CONTEXT_H */ diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index 48a7f5dab8..f3d56ad719 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -33,14 +33,14 @@ */ -#include "main/imports.h" +#include "pipe/p_util.h" #include "pipe/p_defines.h" #include "draw_private.h" struct cull_stage { struct draw_stage stage; - GLuint winding; /**< which winding(s) to cull (one of PIPE_WINDING_x) */ + unsigned winding; /**< which winding(s) to cull (one of PIPE_WINDING_x) */ }; @@ -64,15 +64,15 @@ static void cull_tri( struct draw_stage *stage, struct prim_header *header ) { /* Window coords: */ - const GLfloat *v0 = header->v[0]->data[0]; - const GLfloat *v1 = header->v[1]->data[0]; - const GLfloat *v2 = header->v[2]->data[0]; + const float *v0 = header->v[0]->data[0]; + const float *v1 = header->v[1]->data[0]; + const float *v2 = header->v[2]->data[0]; /* edge vectors e = v0 - v2, f = v1 - v2 */ - GLfloat ex = v0[0] - v2[0]; - GLfloat ey = v0[1] - v2[1]; - GLfloat fx = v1[0] - v2[0]; - GLfloat fy = v1[1] - v2[1]; + const float ex = v0[0] - v2[0]; + const float ey = v0[1] - v2[1]; + const float fx = v1[0] - v2[0]; + const float fy = v1[1] - v2[1]; /* det = cross(e,f).z */ header->det = ex * fy - ey * fx; @@ -81,7 +81,7 @@ static void cull_tri( struct draw_stage *stage, /* if (det < 0 then Z points toward camera and triangle is * counter-clockwise winding. */ - GLuint winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; + unsigned winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; if ((winding & cull_stage(stage)->winding) == 0) { /* triangle is not culled, pass to next stage */ diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index cf5e762079..128acc5b2b 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -28,14 +28,14 @@ /* Authors: Keith Whitwell */ -#include "main/imports.h" +#include "pipe/p_util.h" #include "draw_private.h" struct flatshade_stage { struct draw_stage stage; - const GLuint *lookup; + const unsigned *lookup; }; @@ -53,7 +53,7 @@ static void flatshade_begin( struct draw_stage *stage ) -static INLINE void copy_attr( GLuint attr, +static INLINE void copy_attr( unsigned attr, struct vertex_header *dst, const struct vertex_header *src ) { @@ -70,7 +70,7 @@ static INLINE void copy_colors( struct draw_stage *stage, const struct vertex_header *src ) { const struct flatshade_stage *flatshade = flatshade_stage(stage); - const GLuint *lookup = flatshade->lookup; + const unsigned *lookup = flatshade->lookup; copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); copy_attr( lookup[VF_ATTRIB_COLOR1], dst, src ); diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 9f66566353..9b00b09032 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -32,8 +32,7 @@ * \author Brian Paul */ -#include "main/imports.h" -#include "main/macros.h" +#include "pipe/p_util.h" #include "draw_private.h" @@ -41,8 +40,8 @@ struct offset_stage { struct draw_stage stage; - GLfloat scale; - GLfloat units; + float scale; + float units; }; @@ -56,7 +55,7 @@ static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) static void offset_begin( struct draw_stage *stage ) { struct offset_stage *offset = offset_stage(stage); - GLfloat mrd = 1.0 / 65535.0; /* XXX this depends on depthbuffer bits! */ + float mrd = 1.0 / 65535.0; /* XXX this depends on depthbuffer bits! */ offset->units = stage->draw->setup.offset_units * mrd; offset->scale = stage->draw->setup.offset_scale; @@ -73,30 +72,30 @@ static void do_offset_tri( struct draw_stage *stage, struct prim_header *header ) { struct offset_stage *offset = offset_stage(stage); - GLfloat inv_det = 1.0 / header->det; + float inv_det = 1.0 / header->det; /* Window coords: */ - GLfloat *v0 = header->v[0]->data[0]; - GLfloat *v1 = header->v[1]->data[0]; - GLfloat *v2 = header->v[2]->data[0]; + float *v0 = header->v[0]->data[0]; + float *v1 = header->v[1]->data[0]; + float *v2 = header->v[2]->data[0]; /* edge vectors e = v0 - v2, f = v1 - v2 */ - GLfloat ex = v0[0] - v2[0]; - GLfloat ey = v0[1] - v2[1]; - GLfloat ez = v0[2] - v2[2]; - GLfloat fx = v1[0] - v2[0]; - GLfloat fy = v1[1] - v2[1]; - GLfloat fz = v1[2] - v2[2]; + float ex = v0[0] - v2[0]; + float ey = v0[1] - v2[1]; + float ez = v0[2] - v2[2]; + float fx = v1[0] - v2[0]; + float fy = v1[1] - v2[1]; + float fz = v1[2] - v2[2]; /* (a,b) = cross(e,f).xy */ - GLfloat a = ey*fz - ez*fy; - GLfloat b = ez*fx - ex*fz; + float a = ey*fz - ez*fy; + float b = ez*fx - ex*fz; - GLfloat dzdx = FABSF(a * inv_det); - GLfloat dzdy = FABSF(b * inv_det); + float dzdx = FABSF(a * inv_det); + float dzdy = FABSF(b * inv_det); - GLfloat zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; + float zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; /* * Note: we're applying the offset and clamping per-vertex. diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 931404caff..7757dacf34 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -30,6 +30,7 @@ * Keith Whitwell */ +#include "pipe/p_util.h" #include "draw_private.h" #include "draw_context.h" #include "draw_prim.h" @@ -173,7 +174,7 @@ static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, unsigned i ) { - const GLushort *elts = (const GLushort *)draw->elts; + const ushort *elts = (const ushort *)draw->elts; return get_vertex( draw, elts[i] ); } @@ -181,7 +182,7 @@ static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, unsigned i ) { - const GLubyte *elts = (const GLubyte *)draw->elts; + const ubyte *elts = (const ubyte *)draw->elts; return get_vertex( draw, elts[i] ); } @@ -522,13 +523,13 @@ draw_trim( unsigned count, unsigned first, unsigned incr ) /** * Allocate space for temporary post-transform vertices, such as for clipping. */ -void draw_alloc_tmps( struct draw_stage *stage, GLuint nr ) +void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) { stage->nr_tmps = nr; if (nr) { - GLubyte *store = (GLubyte *) malloc(MAX_VERTEX_SIZE * nr); - GLuint i; + ubyte *store = (ubyte *) malloc(MAX_VERTEX_SIZE * nr); + unsigned i; stage->tmp = (struct vertex_header **) malloc(sizeof(struct vertex_header *) * nr); diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index b6ef32855a..2748772fdb 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -41,10 +41,31 @@ #define DRAW_PRIVATE_H -#include "main/glheader.h" #include "pipe/p_state.h" #include "pipe/p_defines.h" +#ifdef MESA #include "vf/vf.h" +#else +/* XXX these are temporary */ +struct vf_attr_map { + unsigned attrib; + unsigned format; + unsigned offset; +}; +#define VF_ATTRIB_POS 0 +#define VF_ATTRIB_COLOR0 3 +#define VF_ATTRIB_COLOR1 4 +#define VF_ATTRIB_BFC0 25 +#define VF_ATTRIB_BFC1 26 + +#define VF_ATTRIB_CLIP_POS 27 +#define VF_ATTRIB_VERTEX_HEADER 28 +#define VF_ATTRIB_MAX 29 +#define EMIT_1F 0 +#define EMIT_4F 3 +#define EMIT_4F_VIEWPORT 6 +#define FRAG_ATTRIB_MAX 13 +#endif /** @@ -201,7 +222,9 @@ struct draw_context /* Misc for draw_vb.c (XXX temporary) */ +#ifdef MESA GLvector4f header; +#endif ubyte *verts; boolean in_vb; struct vertex_fetch *vf; diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 20b2b4a8aa..045109dae0 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -28,15 +28,15 @@ /* Authors: Keith Whitwell */ -#include "main/imports.h" +#include "pipe/p_util.h" #include "pipe/p_defines.h" #include "draw_private.h" struct twoside_stage { struct draw_stage stage; - GLfloat sign; /**< +1 or -1 */ - const GLuint *lookup; + float sign; /**< +1 or -1 */ + const unsigned *lookup; }; @@ -61,8 +61,8 @@ static void twoside_begin( struct draw_stage *stage ) } -static INLINE void copy_color( GLuint attr_dst, - GLuint attr_src, +static INLINE void copy_color( unsigned attr_dst, + unsigned attr_src, struct vertex_header *v ) { if (attr_dst && attr_src) { @@ -75,7 +75,7 @@ static INLINE void copy_color( GLuint attr_dst, static struct vertex_header *copy_bfc( struct twoside_stage *twoside, const struct vertex_header *v, - GLuint idx ) + unsigned idx ) { struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index e0e486ebe6..4215b8b228 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -33,7 +33,7 @@ /* Authors: Keith Whitwell */ -#include "main/imports.h" +#include "pipe/p_util.h" #include "pipe/p_defines.h" #include "draw_private.h" @@ -45,7 +45,7 @@ struct unfilled_stage { * legal values: PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE, * and PIPE_POLYGON_MODE_POINT, */ - GLuint mode[2]; + unsigned mode[2]; }; @@ -119,7 +119,7 @@ static void unfilled_tri( struct draw_stage *stage, struct prim_header *header ) { struct unfilled_stage *unfilled = unfilled_stage(stage); - GLuint mode = unfilled->mode[header->det < 0.0]; + unsigned mode = unfilled->mode[header->det < 0.0]; switch (mode) { case PIPE_POLYGON_MODE_FILL: diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index db7b217c2f..18e48c0a68 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -30,11 +30,17 @@ * Keith Whitwell */ -#include "imports.h" -#include "macros.h" - +#ifdef MESA +#include "main/imports.h" +#include "main/macros.h" #include "tnl/t_context.h" #include "vf/vf.h" +#else +#define VF_ +#include "pipe/p_util.h" + +#endif /*MESA*/ + #include "draw_private.h" #include "draw_context.h" @@ -92,11 +98,11 @@ static void vs_flush( struct draw_context *draw ) * Allocate storage for post-transformation vertices. */ static void -draw_allocate_vertices( struct draw_context *draw, GLuint nr_vertices ) +draw_allocate_vertices( struct draw_context *draw, unsigned nr_vertices ) { assert(draw->vertex_size > 0); draw->nr_vertices = nr_vertices; - draw->verts = (GLubyte *) malloc( nr_vertices * draw->vertex_size ); + draw->verts = (ubyte *) malloc( nr_vertices * draw->vertex_size ); draw_invalidate_vcache( draw ); } @@ -117,20 +123,21 @@ draw_release_vertices( struct draw_context *draw ) * Note: this must match struct vertex_header's layout (I think). */ struct header_dword { - GLuint clipmask:12; - GLuint edgeflag:1; - GLuint pad:19; + unsigned clipmask:12; + unsigned edgeflag:1; + unsigned pad:19; }; +#ifdef MESA static void build_vertex_headers( struct draw_context *draw, struct vertex_buffer *VB ) { if (draw->header.storage == NULL) { - draw->header.stride = sizeof(GLfloat); + draw->header.stride = sizeof(float); draw->header.size = 1; - draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(GLfloat), 32 ); + draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(float), 32 ); draw->header.data = draw->header.storage; draw->header.count = 0; draw->header.flags = VEC_SIZE_1 | VEC_MALLOC; @@ -141,12 +148,12 @@ build_vertex_headers( struct draw_context *draw, */ { - GLuint i; + unsigned i; struct header_dword *header = (struct header_dword *)draw->header.storage; /* yes its a hack */ - assert(sizeof(*header) == sizeof(GLfloat)); + assert(sizeof(*header) == sizeof(float)); draw->header.count = VB->Count; @@ -175,16 +182,18 @@ build_vertex_headers( struct draw_context *draw, VB->AttribPtr[VF_ATTRIB_VERTEX_HEADER] = &draw->header; } +#endif /*MESA*/ +#ifdef MESA /** * This is a hack & will all go away. */ void draw_vb(struct draw_context *draw, struct vertex_buffer *VB ) { - GLuint i; + unsigned i; VB->AttribPtr[VF_ATTRIB_POS] = VB->NdcPtr; VB->AttribPtr[VF_ATTRIB_BFC0] = VB->ColorPtr[1]; @@ -212,14 +221,14 @@ void draw_vb(struct draw_context *draw, vf_emit_vertices( draw->vf, VB->Count, draw->verts ); if (VB->Elts) - draw_set_element_buffer(draw, sizeof(GLuint), VB->Elts); + draw_set_element_buffer(draw, sizeof(unsigned), VB->Elts); else draw_set_element_buffer(draw, 0, NULL); for (i = 0; i < VB->PrimitiveCount; i++) { const GLenum mode = VB->Primitive[i].mode; - const GLuint start = VB->Primitive[i].start; - GLuint length, first, incr; + const unsigned start = VB->Primitive[i].start; + unsigned length, first, incr; /* Trim the primitive down to a legal size. */ @@ -247,27 +256,28 @@ void draw_vb(struct draw_context *draw, draw->in_vb = 0; draw->elts = NULL; } +#endif /*MESA*/ /** * XXX Temporary mechanism to draw simple vertex arrays. - * All attribs are GLfloat[4]. Arrays are interleaved, in GL-speak. + * All attribs are float[4]. Arrays are interleaved, in GL-speak. */ void draw_vertices(struct draw_context *draw, - GLuint mode, - GLuint numVerts, const GLfloat *vertices, - GLuint numAttrs, const GLuint attribs[]) + unsigned mode, + unsigned numVerts, const float *vertices, + unsigned numAttrs, const unsigned attribs[]) { - /*GLuint first, incr;*/ - GLuint i, j; + /*unsigned first, incr;*/ + unsigned i, j; - assert(mode <= GL_POLYGON); + assert(mode <= PIPE_PRIM_POLYGON); draw->vs_flush = vs_flush; draw->vertex_size - = sizeof(struct vertex_header) + numAttrs * 4 * sizeof(GLfloat); + = sizeof(struct vertex_header) + numAttrs * 4 * sizeof(float); /* no element/index buffer */ @@ -344,10 +354,10 @@ do { \ * (and number of attributes) */ void draw_set_vertex_attributes( struct draw_context *draw, - const GLuint *slot_to_vf_attr, - GLuint nr_attrs ) + const unsigned *slot_to_vf_attr, + unsigned nr_attrs ) { - GLuint i; + unsigned i; memset(draw->vf_attr_to_slot, 0, sizeof(draw->vf_attr_to_slot)); draw->nr_attrs = 0; @@ -368,6 +378,8 @@ void draw_set_vertex_attributes( struct draw_context *draw, EMIT_ATTR(slot_to_vf_attr[i], EMIT_4F); /* tell the vertex format module how to construct vertices for us */ +#if MESA draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, draw->nr_attrs, 0 ); +#endif } diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 072dad0573..f4af16a2de 100755 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -112,6 +112,18 @@ do { \ } while (0) +#define COPY_4FV( DST, SRC ) COPY_4V(DST, SRC) + + +#define ASSIGN_4V( DST, V0, V1, V2, V3 ) \ +do { \ + (DST)[0] = (V0); \ + (DST)[1] = (V1); \ + (DST)[2] = (V2); \ + (DST)[3] = (V3); \ +} while (0) + + static INLINE int ifloor(float f) { int ai, bi; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 3f7044825a..e63ce208d4 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -80,6 +80,11 @@ softpipe_supported_formats(struct pipe_context *pipe, unsigned *numFormats) } +/** XXX remove these? */ +#define MAX_TEXTURE_LEVELS 11 +#define MAX_TEXTURE_RECT_SIZE 2048 +#define MAX_3D_TEXTURE_LEVELS 8 + static void softpipe_max_texture_size(struct pipe_context *pipe, unsigned textureType, unsigned *maxWidth, unsigned *maxHeight, diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 9b91628aaf..21c1032b13 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -40,6 +40,11 @@ #include "pipe/draw/draw_private.h" #include "pipe/p_util.h" +#ifndef MESA +#define FRAG_ATTRIB_WPOS 0 +#define FRAG_ATTRIB_MAX 13 +#endif + /** * Triangle edge info diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index d66a0ed851..425076d653 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -74,7 +74,7 @@ static GLboolean draw( GLcontext * ctx, struct tnl_pipeline_stage *stage ) /* Call into the new draw code to handle the VB: */ st->pipe->draw_vb( st->pipe, VB ); - + /* Finished */ return GL_FALSE; -- cgit v1.2.3 From a279b221bf9e64349b722449443d25db45f5a6df Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 18:16:32 +0100 Subject: Cosmetic changes. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 2c8deac034..1de80eb734 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -513,6 +513,7 @@ tgsi_mesa_compile_fp_program( { GLuint i, ti, count; struct tgsi_header *header; + struct tgsi_processor *processor; struct tgsi_full_declaration fulldecl; struct tgsi_full_instruction fullinst; struct tgsi_full_dst_register *fulldst; @@ -526,7 +527,10 @@ tgsi_mesa_compile_fp_program( header = (struct tgsi_header *) &tokens[1]; *header = tgsi_build_header(); - ti = 2; + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor( TGSI_PROCESSOR_FRAGMENT, header ); + + ti = 3; reads_wpos = program->Base.InputsRead & (1 << FRAG_ATTRIB_WPOS); inputs_read = program->Base.InputsRead | (1 << FRAG_ATTRIB_WPOS); @@ -645,10 +649,9 @@ tgsi_mesa_compile_fp_program( TGSI_PROCESSOR_FRAGMENT ) ) { assert( i == program->Base.NumInstructions - 1 ); - if (TGSI_DEBUG) - tgsi_dump( - tokens, - 0/*TGSI_DUMP_VERBOSE | TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT*/ ); + if( TGSI_DEBUG ) { + tgsi_dump( tokens, 0 ); + } break; } @@ -668,7 +671,7 @@ tgsi_mesa_compile_vp_program( struct tgsi_token *tokens, GLuint maxTokens ) { - GLuint ii, ti; + GLuint i, ti; struct tgsi_header *header; struct tgsi_processor *processor; struct tgsi_full_instruction fullinst; @@ -684,16 +687,18 @@ tgsi_mesa_compile_vp_program( ti = 3; - for( ii = 0; ii < program->Base.NumInstructions; ii++ ) { + for( i = 0; i < program->Base.NumInstructions; i++ ) { if( compile_instruction( - &program->Base.Instructions[ii], + &program->Base.Instructions[i], &fullinst, inputs_read, 0, TGSI_PROCESSOR_VERTEX ) ) { - assert( ii == program->Base.NumInstructions - 1 ); - if (TGSI_DEBUG) - tgsi_dump( tokens, TGSI_DUMP_NO_IGNORED | TGSI_DUMP_NO_DEFAULT ); + assert( i == program->Base.NumInstructions - 1 ); + + if( TGSI_DEBUG ) { + tgsi_dump( tokens, 0 ); + } break; } -- cgit v1.2.3 From 00b7a915084067cef85c25cedcec257ce4eb25f0 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 18:29:45 +0100 Subject: Move VS code from st_atom_fs.c to st_atom_vs.c. --- src/mesa/state_tracker/st_atom_fs.c | 100 +++--------------------------------- src/mesa/state_tracker/st_atom_vs.c | 77 +++++++++++++++++++++++++-- 2 files changed, 80 insertions(+), 97 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 9dddb18498..7e12441784 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -28,21 +28,18 @@ * Authors: * Keith Whitwell */ - + #include "shader/prog_parameter.h" -#include "st_context.h" + #include "pipe/p_context.h" -#include "st_atom.h" -#include "st_program.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/core/tgsi_dump.h" -#define TGSI_DEBUG 0 - +#include "st_context.h" +#include "st_atom.h" +#include "st_program.h" -/** - ** Fragment programs - **/ +#define TGSI_DEBUG 0 static void compile_fs( struct st_context *st, struct st_fragment_program *fs ) @@ -90,7 +87,7 @@ static void update_fs( struct st_context *st ) fs.inputs_read = fp->Base.Base.InputsRead; fs.tokens = &fp->tokens[0]; fs.constants = &fp->constants; - + if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0 || fp->dirty) { @@ -108,86 +105,3 @@ const struct st_tracked_state st_update_fs = { }, .update = update_fs }; - - - -/** - ** Vertex programs - **/ - - -static void compile_vs( struct st_context *st, - struct st_vertex_program *vs ) -{ - /* XXX: fix static allocation of tokens: - */ - tgsi_mesa_compile_vp_program( &vs->Base, vs->tokens, ST_FP_MAX_TOKENS ); - - if (TGSI_DEBUG) - tgsi_dump( vs->tokens, TGSI_DUMP_VERBOSE ); -} - - -static void update_vs( struct st_context *st ) -{ - struct pipe_shader_state vs; - struct st_vertex_program *vp = NULL; - struct gl_program_parameter_list *params = NULL; - - if (st->ctx->VertexProgram._MaintainTnlProgram) - _tnl_UpdateFixedFunctionProgram( st->ctx ); - - if (st->ctx->Shader.CurrentProgram && - st->ctx->Shader.CurrentProgram->LinkStatus && - st->ctx->Shader.CurrentProgram->VertexProgram) { - struct gl_vertex_program *f - = st->ctx->Shader.CurrentProgram->VertexProgram; - vp = st_vertex_program(f); - params = f->Base.Parameters; - } - else if (st->ctx->VertexProgram._Current) { - vp = st_vertex_program(st->ctx->VertexProgram._Current); - params = st->ctx->VertexProgram._Current->Base.Parameters; - } - - /* XXXX temp */ -#if 1 - if (!vp) - return; -#endif - if (vp && params) { - /* load program's constants array */ - - _mesa_load_state_parameters(st->ctx, params); - - vp->constants.nr_constants = params->NumParameters; - memcpy(vp->constants.constant, - params->ParameterValues, - params->NumParameters * sizeof(GLfloat) * 4); - } - - if (vp->dirty) - compile_vs( st, vp ); - - memset( &vs, 0, sizeof(vs) ); - vs.inputs_read = vp->Base.Base.InputsRead; - vs.tokens = &vp->tokens[0]; - vs.constants = &vp->constants; - - if (memcmp(&vs, &st->state.vs, sizeof(vs)) != 0 || - vp->dirty) - { - vp->dirty = 0; - st->state.vs = vs; - st->pipe->set_vs_state(st->pipe, &vs); - } -} - - -const struct st_tracked_state st_update_vs = { - .dirty = { - .mesa = _NEW_PROGRAM | _NEW_MODELVIEW, - .st = ST_NEW_VERTEX_PROGRAM, - }, - .update = update_vs -}; diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 6a26bfdd19..d94c560bef 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -28,21 +28,90 @@ * Authors: * Keith Whitwell */ - -#include "st_context.h" + +#include "shader/prog_parameter.h" + #include "pipe/p_context.h" +#include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "pipe/tgsi/core/tgsi_dump.h" + +#include "st_context.h" #include "st_atom.h" +#include "st_program.h" + +#define TGSI_DEBUG 0 + +static void compile_vs( struct st_context *st, + struct st_vertex_program *vs ) +{ + /* XXX: fix static allocation of tokens: + */ + tgsi_mesa_compile_vp_program( &vs->Base, vs->tokens, ST_FP_MAX_TOKENS ); + + if (TGSI_DEBUG) + tgsi_dump( vs->tokens, TGSI_DUMP_VERBOSE ); +} - static void update_vs( struct st_context *st ) { + struct pipe_shader_state vs; + struct st_vertex_program *vp = NULL; + struct gl_program_parameter_list *params = NULL; + + if (st->ctx->VertexProgram._MaintainTnlProgram) + _tnl_UpdateFixedFunctionProgram( st->ctx ); + + if (st->ctx->Shader.CurrentProgram && + st->ctx->Shader.CurrentProgram->LinkStatus && + st->ctx->Shader.CurrentProgram->VertexProgram) { + struct gl_vertex_program *f + = st->ctx->Shader.CurrentProgram->VertexProgram; + vp = st_vertex_program(f); + params = f->Base.Parameters; + } + else if (st->ctx->VertexProgram._Current) { + vp = st_vertex_program(st->ctx->VertexProgram._Current); + params = st->ctx->VertexProgram._Current->Base.Parameters; + } + + /* XXXX temp */ +#if 1 + if (!vp) + return; +#endif + if (vp && params) { + /* load program's constants array */ + + _mesa_load_state_parameters(st->ctx, params); + + vp->constants.nr_constants = params->NumParameters; + memcpy(vp->constants.constant, + params->ParameterValues, + params->NumParameters * sizeof(GLfloat) * 4); + } + + if (vp->dirty) + compile_vs( st, vp ); + + memset( &vs, 0, sizeof(vs) ); + vs.inputs_read = vp->Base.Base.InputsRead; + vs.tokens = &vp->tokens[0]; + vs.constants = &vp->constants; + + if (memcmp(&vs, &st->state.vs, sizeof(vs)) != 0 || + vp->dirty) + { + vp->dirty = 0; + st->state.vs = vs; + st->pipe->set_vs_state(st->pipe, &vs); + } } const struct st_tracked_state st_update_vs = { .dirty = { - .mesa = 0, + .mesa = _NEW_PROGRAM | _NEW_MODELVIEW, .st = ST_NEW_VERTEX_PROGRAM, }, .update = update_vs -- cgit v1.2.3 From bef9b40c3904fdffea6cc426c5c82ed3ab70a902 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 19:42:28 +0100 Subject: Define __WIN32__. --- src/mesa/pipe/p_compiler.h | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h index 51a78bf4f5..4f968404eb 100644 --- a/src/mesa/pipe/p_compiler.h +++ b/src/mesa/pipe/p_compiler.h @@ -33,6 +33,12 @@ #include #include + +#if defined(_WIN32) && !defined(__WIN32__) +#define __WIN32__ +#endif + + typedef unsigned int uint; typedef unsigned char ubyte; typedef unsigned char boolean; -- cgit v1.2.3 From 546f1fa01920aeb777a03c4d2c72405d038dd96e Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 19:44:53 +0100 Subject: Define ffs() for __WIN32__ & __MSC__. Break mesa dependencies. --- src/mesa/pipe/p_util.h | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) mode change 100755 => 100644 src/mesa/pipe/p_util.h (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h old mode 100755 new mode 100644 index f4af16a2de..69b69a3c67 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -38,6 +38,18 @@ #define Elements(x) sizeof(x)/sizeof(*(x)) +#if defined(__MSC__) && defined(__WIN32__) +static INLINE unsigned ffs( unsigned u ) +{ + unsigned i; + + __asm bsf eax, [u] + __asm mov [i], eax + + return i; +} +#endif + union fi { float f; int i; @@ -165,7 +177,7 @@ static INLINE int iround(float f) #if defined(__GNUC__) #define FABSF(x) fabsf(x) #else -#define FABSF(x) ((GLfloat) fabs(x)) +#define FABSF(x) ((float) fabs(x)) #endif /* Pretty fast, and accurate. -- cgit v1.2.3 From 2f559606dc3a27104199226ccf6568a222797055 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 19:45:55 +0100 Subject: Silence compiler warnings. --- src/mesa/pipe/draw/draw_offset.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 9b00b09032..83c9b163eb 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -55,7 +55,7 @@ static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) static void offset_begin( struct draw_stage *stage ) { struct offset_stage *offset = offset_stage(stage); - float mrd = 1.0 / 65535.0; /* XXX this depends on depthbuffer bits! */ + float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ offset->units = stage->draw->setup.offset_units * mrd; offset->scale = stage->draw->setup.offset_scale; @@ -72,7 +72,7 @@ static void do_offset_tri( struct draw_stage *stage, struct prim_header *header ) { struct offset_stage *offset = offset_stage(stage); - float inv_det = 1.0 / header->det; + float inv_det = 1.0f / header->det; /* Window coords: */ @@ -101,9 +101,9 @@ static void do_offset_tri( struct draw_stage *stage, * Note: we're applying the offset and clamping per-vertex. * Ideally, the offset is applied per-fragment prior to fragment shading. */ - v0[2] = CLAMP(v0[2] + zoffset, 0.0, 1.0); - v1[2] = CLAMP(v1[2] + zoffset, 0.0, 1.0); - v2[2] = CLAMP(v2[2] + zoffset, 0.0, 1.0); + v0[2] = CLAMP(v0[2] + zoffset, 0.0f, 1.0f); + v1[2] = CLAMP(v1[2] + zoffset, 0.0f, 1.0f); + v2[2] = CLAMP(v2[2] + zoffset, 0.0f, 1.0f); stage->next->tri( stage->next, header ); } -- cgit v1.2.3 From 1b19aa522abc3dc4e38666b4a98337a33a82a946 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 19:47:18 +0100 Subject: Break mesa dependencies. Silence compiler warnings. --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index c8e161cfb1..69fd6908bb 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -31,8 +31,6 @@ */ -#include "main/context.h" - #include "pipe/p_defines.h" #include "pipe/p_context.h" #include "pipe/p_winsys.h" @@ -175,7 +173,7 @@ run_vertex_program2(struct draw_context *draw, x /= w; y /= w; z /= w; - w = 1.0 / w; + w = 1.0f / w; /* Viewport */ vOut->data[0][0] = scale[0] * x + trans[0]; -- cgit v1.2.3 From f692cdb544bf42eb36154f6a132d9e144dbae858 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 19:49:04 +0100 Subject: Silence compiler warnings. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 118 ++++++++++++++++++--------------- 1 file changed, 63 insertions(+), 55 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 21c1032b13..43ab0e97db 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -112,21 +112,26 @@ static INLINE void quad_clip(struct setup_stage *setup) { const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - if (setup->quad.x0 >= cliprect->maxx || - setup->quad.y0 >= cliprect->maxy || - setup->quad.x0 + 1 < cliprect->minx || - setup->quad.y0 + 1 < cliprect->miny) { + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + + if (setup->quad.x0 >= maxx || + setup->quad.y0 >= maxy || + setup->quad.x0 + 1 < minx || + setup->quad.y0 + 1 < miny) { /* totally clipped */ setup->quad.mask = 0x0; return; } - if (setup->quad.x0 < cliprect->minx) + if (setup->quad.x0 < minx) setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup->quad.y0 < cliprect->miny) + if (setup->quad.y0 < miny) setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); - if (setup->quad.x0 == cliprect->maxx - 1) + if (setup->quad.x0 == maxx - 1) setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup->quad.y0 == cliprect->maxy - 1) + if (setup->quad.y0 == maxy - 1) setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); } @@ -336,7 +341,7 @@ static boolean setup_sort_vertices( struct setup_stage *setup, const float area = (setup->emaj.dx * setup->ebot.dy - setup->ebot.dx * setup->emaj.dy); - setup->oneoverarea = 1.0 / area; + setup->oneoverarea = 1.0f / area; /* _mesa_printf("%s one-over-area %f area %f det %f\n", __FUNCTION__, setup->oneoverarea, area, prim->det ); @@ -408,8 +413,8 @@ static void tri_linear_coeff( struct setup_stage *setup, * instead - i'll switch to this later. */ setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); /* _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", @@ -446,8 +451,8 @@ static void tri_persp_coeff( struct setup_stage *setup, setup->coef[slot].dadx[i] = a * setup->oneoverarea; setup->coef[slot].dady[i] = b * setup->oneoverarea; setup->coef[slot].a0[i] = (mina - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5))); + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); } @@ -491,12 +496,12 @@ static void setup_tri_coefficients( struct setup_stage *setup ) static void setup_tri_edges( struct setup_stage *setup ) { - float vmin_x = setup->vmin->data[0][0] + 0.5; - float vmid_x = setup->vmid->data[0][0] + 0.5; + float vmin_x = setup->vmin->data[0][0] + 0.5f; + float vmid_x = setup->vmid->data[0][0] + 0.5f; - float vmin_y = setup->vmin->data[0][1] - 0.5; - float vmid_y = setup->vmid->data[0][1] - 0.5; - float vmax_y = setup->vmax->data[0][1] - 0.5; + float vmin_y = setup->vmin->data[0][1] - 0.5f; + float vmid_y = setup->vmid->data[0][1] - 0.5f; + float vmax_y = setup->vmax->data[0][1] - 0.5f; setup->emaj.sy = ceilf(vmin_y); setup->emaj.lines = (int) ceilf(vmax_y - setup->emaj.sy); @@ -525,6 +530,10 @@ static void subtriangle( struct setup_stage *setup, unsigned lines ) { const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; int y, start_y, finish_y; int sy = (int)eleft->sy; @@ -534,11 +543,11 @@ static void subtriangle( struct setup_stage *setup, start_y = sy; finish_y = sy + lines; - if (start_y < cliprect->miny) - start_y = cliprect->miny; + if (start_y < miny) + start_y = miny; - if (finish_y > cliprect->maxy) - finish_y = cliprect->maxy; + if (finish_y > maxy) + finish_y = maxy; start_y -= sy; finish_y -= sy; @@ -559,23 +568,22 @@ static void subtriangle( struct setup_stage *setup, int right = (int)(eright->sx + y * eright->dxdy); /* clip left/right */ - if (left < cliprect->minx) - left = cliprect->minx; - if (right > cliprect->maxx) - right = cliprect->maxx; + if (left < minx) + left = minx; + if (right > maxx) + right = maxx; if (left < right) { - int _y = sy+y; - if (block(_y) != setup->span.y) { - flush_spans(setup); - setup->span.y = block(_y); - } - - setup->span.left[_y&1] = left; - setup->span.right[_y&1] = right; - setup->span.y_flags |= 1<<(_y&1); + int _y = sy + y; + if (block(_y) != setup->span.y) { + flush_spans(setup); + setup->span.y = block(_y); + } + + setup->span.left[_y&1] = left;setup->span.right[_y&1] = right; + setup->span.y_flags |= 1<<(_y&1); } - } + } /* save the values so that emaj can be restarted: @@ -645,8 +653,8 @@ line_linear_coeff(struct setup_stage *setup, unsigned slot, unsigned i) setup->coef[slot].dady[i] = dady; setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5) + - dady * (setup->vmin->data[0][1] - 0.5))); + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); } @@ -680,8 +688,8 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; /* NOTE: this is not really 1/area */ - setup->oneoverarea = 1.0 / (setup->emaj.dx * setup->emaj.dx + - setup->emaj.dy * setup->emaj.dy); + setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx + + setup->emaj.dy * setup->emaj.dy); /* z and w are done by linear interpolation: */ @@ -880,7 +888,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); /*XXX this should be a vertex attrib! */ - const float halfSize = 0.5 * setup->softpipe->setup.point_size; + const float halfSize = 0.5f * setup->softpipe->setup.point_size; const boolean round = setup->softpipe->setup.point_smooth; const struct vertex_header *v0 = prim->v[0]; const float x = v0->data[FRAG_ATTRIB_WPOS][0]; @@ -917,8 +925,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) /* special case for 1-pixel points */ const int ix = ((int) x) & 1; const int iy = ((int) y) & 1; - setup->quad.x0 = x - ix; - setup->quad.y0 = y - iy; + setup->quad.x0 = (int) x - ix; + setup->quad.y0 = (int) y - iy; setup->quad.mask = (1 << ix) << (2 * iy); clip_emit_quad(setup); } @@ -943,39 +951,39 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) setup->quad.mask = 0x0; - dx = (ix + 0.5) - x; - dy = (iy + 0.5) - y; + dx = (ix + 0.5f) - x; + dy = (iy + 0.5f) - y; dist2 = dx * dx + dy * dy; if (dist2 <= rmax2) { cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0); + setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f); setup->quad.mask |= MASK_BOTTOM_LEFT; } - dx = (ix + 1.5) - x; - dy = (iy + 0.5) - y; + dx = (ix + 1.5f) - x; + dy = (iy + 0.5f) - y; dist2 = dx * dx + dy * dy; if (dist2 <= rmax2) { cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0); + setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f); setup->quad.mask |= MASK_BOTTOM_RIGHT; } - dx = (ix + 0.5) - x; - dy = (iy + 1.5) - y; + dx = (ix + 0.5f) - x; + dy = (iy + 1.5f) - y; dist2 = dx * dx + dy * dy; if (dist2 <= rmax2) { cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0); + setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f); setup->quad.mask |= MASK_TOP_LEFT; } - dx = (ix + 1.5) - x; - dy = (iy + 1.5) - y; + dx = (ix + 1.5f) - x; + dy = (iy + 1.5f) - y; dist2 = dx * dx + dy * dy; if (dist2 <= rmax2) { cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0); + setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f); setup->quad.mask |= MASK_TOP_RIGHT; } -- cgit v1.2.3 From 24b2f3041090820b926b5902f0c6c65922874d68 Mon Sep 17 00:00:00 2001 From: michal Date: Thu, 16 Aug 2007 19:50:08 +0100 Subject: Break mesa dependencies. Silence compiler warnings. --- src/mesa/pipe/softpipe/sp_state_derived.c | 72 ++++++++++++++++++++++++++++--- 1 file changed, 66 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 44beb9c3eb..35f1a971ca 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -25,16 +25,76 @@ * **************************************************************************/ -#include "main/glheader.h" -#include "main/macros.h" -#include "main/enums.h" +#include "pipe/p_util.h" -#include "vf/vf.h" #include "pipe/draw/draw_context.h" + #include "sp_context.h" #include "sp_state.h" +/* XXX: copied from vf.h */ +enum { + VF_ATTRIB_POS = 0, + VF_ATTRIB_WEIGHT = 1, + VF_ATTRIB_NORMAL = 2, + VF_ATTRIB_COLOR0 = 3, + VF_ATTRIB_COLOR1 = 4, + VF_ATTRIB_FOG = 5, + VF_ATTRIB_COLOR_INDEX = 6, + VF_ATTRIB_EDGEFLAG = 7, + VF_ATTRIB_TEX0 = 8, + VF_ATTRIB_TEX1 = 9, + VF_ATTRIB_TEX2 = 10, + VF_ATTRIB_TEX3 = 11, + VF_ATTRIB_TEX4 = 12, + VF_ATTRIB_TEX5 = 13, + VF_ATTRIB_TEX6 = 14, + VF_ATTRIB_TEX7 = 15, + VF_ATTRIB_VAR0 = 16, + VF_ATTRIB_VAR1 = 17, + VF_ATTRIB_VAR2 = 18, + VF_ATTRIB_VAR3 = 19, + VF_ATTRIB_VAR4 = 20, + VF_ATTRIB_VAR5 = 21, + VF_ATTRIB_VAR6 = 22, + VF_ATTRIB_VAR7 = 23, + VF_ATTRIB_POINTSIZE = 24, + VF_ATTRIB_BFC0 = 25, + VF_ATTRIB_BFC1 = 26, + VF_ATTRIB_CLIP_POS = 27, + VF_ATTRIB_VERTEX_HEADER = 28, + VF_ATTRIB_MAX = 29 +}; + +/* XXX: copied from config.h */ +#define MAX_VARYING 8 + +/* XXX: copied from mtypes.h */ +enum +{ + FRAG_ATTRIB_WPOS = 0, + FRAG_ATTRIB_COL0 = 1, + FRAG_ATTRIB_COL1 = 2, + FRAG_ATTRIB_FOGC = 3, + FRAG_ATTRIB_TEX0 = 4, + FRAG_ATTRIB_TEX1 = 5, + FRAG_ATTRIB_TEX2 = 6, + FRAG_ATTRIB_TEX3 = 7, + FRAG_ATTRIB_TEX4 = 8, + FRAG_ATTRIB_TEX5 = 9, + FRAG_ATTRIB_TEX6 = 10, + FRAG_ATTRIB_TEX7 = 11, + FRAG_ATTRIB_VAR0 = 12, /**< shader varying */ + FRAG_ATTRIB_MAX = (FRAG_ATTRIB_VAR0 + MAX_VARYING) +}; + +#define FRAG_BIT_COL0 (1 << FRAG_ATTRIB_COL0) +#define FRAG_BIT_COL1 (1 << FRAG_ATTRIB_COL1) + + + + #define EMIT_ATTR( VF_ATTR, FRAG_ATTR, INTERP ) \ do { \ slot_to_vf_attr[softpipe->nr_attrs] = VF_ATTR; \ @@ -46,7 +106,7 @@ do { \ } while (0) -static const unsigned frag_to_vf[FRAG_ATTRIB_MAX] = +static const unsigned frag_to_vf[PIPE_ATTRIB_MAX] = { VF_ATTRIB_POS, VF_ATTRIB_COLOR0, @@ -176,7 +236,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) static void compute_cliprect(struct softpipe_context *sp) { - int surfWidth, surfHeight; + unsigned surfWidth, surfHeight; if (sp->framebuffer.num_cbufs > 0) { surfWidth = sp->framebuffer.cbufs[0]->width; -- cgit v1.2.3 From 04f2078860d0649b016e9281f86725cb42e15b42 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 10:27:37 +0100 Subject: added assertions, disable debug output --- src/mesa/state_tracker/st_cb_clear.c | 2 ++ src/mesa/state_tracker/st_cb_fbo.c | 5 +++++ 2 files changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 492073e840..d584f0cafc 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -409,6 +409,8 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { struct st_renderbuffer *strb = st_renderbuffer(rb); + assert(strb->surface->format); + if (!ctx->Scissor.Enabled && !is_depth_stencil_format(strb->surface->format)) { /* clear whole depth buffer w/out masking */ diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 7336327330..2b9aa3e9d2 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -104,6 +104,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, return GL_FALSE; /* out of memory, try s/w buffer? */ ASSERT(strb->surface->region->buffer); + ASSERT(strb->surface->format); strb->Base.Width = strb->surface->width = width; strb->Base.Height = strb->surface->height = height; @@ -319,8 +320,10 @@ st_render_texture(GLcontext *ctx, att->Zoffset); assert(strb->surface); + /* printf("RENDER TO TEXTURE obj=%p mt=%p surf=%p %d x %d\n", att->Texture, mt, strb->surface, rb->Width, rb->Height); + */ /* Invalidate buffer state so that the pipe's framebuffer state * gets updated. @@ -344,7 +347,9 @@ st_finish_render_texture(GLcontext *ctx, ctx->st->pipe->flush(ctx->st->pipe, 0x0); + /* printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface); + */ pipe_surface_unreference(&strb->surface); -- cgit v1.2.3 From 93efcf50fae7bdd9ee9468f4f4049ecb4c06e2df Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 10:27:53 +0100 Subject: assertions --- src/mesa/pipe/softpipe/sp_clear.c | 3 +++ src/mesa/pipe/softpipe/sp_region.c | 8 +++++++- 2 files changed, 10 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 6e5e510637..caa6ac1c84 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -55,5 +55,8 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, w = softpipe->framebuffer.cbufs[0]->width; h = softpipe->framebuffer.cbufs[0]->height; + assert(w <= ps->region->pitch); + assert(h <= ps->region->height); + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue); } diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 231ea1ea22..25d0a419aa 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -33,6 +33,7 @@ #include "sp_context.h" #include "sp_region.h" +#include "pipe/p_util.h" #include "pipe/p_winsys.h" #include "pipe/p_defines.h" @@ -86,7 +87,7 @@ sp_region_alloc(struct pipe_context *pipe, unsigned cpp, unsigned width, unsigned height, unsigned flags) { struct softpipe_context *sp = softpipe_context( pipe ); - struct pipe_region *region = calloc(sizeof(*region), 1); + struct pipe_region *region = CALLOC_STRUCT(pipe_region); const unsigned alignment = 64; region->cpp = cpp; @@ -94,6 +95,8 @@ sp_region_alloc(struct pipe_context *pipe, region->height = height; region->refcount = 1; + assert(region->pitch > 0); + region->buffer = sp->pipe.winsys->buffer_create( sp->pipe.winsys, alignment ); /* NULL data --> just allocate the space */ @@ -233,6 +236,9 @@ sp_region_fill(struct pipe_context *pipe, { unsigned i, j; + assert(dst->pitch > 0); + assert(width <= dst->pitch); + (void)pipe->region_map(pipe, dst); switch (dst->cpp) { -- cgit v1.2.3 From ae64d5c173a4525800383668cc91c176a1bd7a15 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 10:28:20 +0100 Subject: fix buffer clearing problems --- src/mesa/drivers/x11/xm_api.c | 6 +----- src/mesa/drivers/x11/xm_buffer.c | 3 ++- src/mesa/drivers/x11/xm_dd.c | 35 ----------------------------------- src/mesa/drivers/x11/xm_softpipe.c | 2 ++ src/mesa/drivers/x11/xm_surface.c | 28 +++++++++++++++++++++++++++- 5 files changed, 32 insertions(+), 42 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index cea790dc60..69b7277c1b 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -1591,12 +1591,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) mesaCtx->st->pipe->surface_alloc = xmesa_surface_alloc; mesaCtx->st->pipe->supported_formats = xmesa_supported_formats; -#if 1 - mesaCtx->Driver.Clear = xmesa_clear_buffers; -#endif -#if 0 + /* special pipe->clear function */ mesaCtx->st->pipe->clear = xmesa_clear; -#endif return c; } diff --git a/src/mesa/drivers/x11/xm_buffer.c b/src/mesa/drivers/x11/xm_buffer.c index 52629aca18..b8d3df1379 100644 --- a/src/mesa/drivers/x11/xm_buffer.c +++ b/src/mesa/drivers/x11/xm_buffer.c @@ -253,7 +253,8 @@ finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) { struct pipe_context *pipe = ctx->st->pipe; if (!xrb->St.surface->region) { - xrb->St.surface->region = pipe->region_alloc(pipe, 1, 0, 0, 0x0); + int w = 1, h = 1; + xrb->St.surface->region = pipe->region_alloc(pipe, 1, w, h, 0x0); } } diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index 22eae51ca9..d7fb3e6e03 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -425,41 +425,6 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers) } -#if 0 -void -xmesa_clear(struct pipe_context *pipe, GLboolean color, GLboolean depth, - GLboolean stencil, GLboolean accum) -{ - struct softpipe_context *sp = (struct softpipe_context *) pipe; - - /* Clear non-color buffers first. This will cause softpipe to - * re-validate the scissor/surface bounds. - */ - softpipe_clear(pipe, GL_FALSE, depth, stencil, accum); - - if (color) { - GET_CURRENT_CONTEXT(ctx); - GLuint i; - for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - struct pipe_surface *ps = sp->framebuffer.cbufs[i]; - struct xmesa_renderbuffer *xrb = (struct xmesa_renderbuffer *) ps->rb; - const GLint x = sp->cliprect.minx; - const GLint y = sp->cliprect.miny; - const GLint w = sp->cliprect.maxx - x; - const GLint h = sp->cliprect.maxy - y; - xrb->clearFunc(ctx, xrb, x, y, w, h); - } - } -} -#endif - -void -xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) -{ - -} - - #ifndef XFree86Server /* XXX this was never tested in the Xserver environment */ diff --git a/src/mesa/drivers/x11/xm_softpipe.c b/src/mesa/drivers/x11/xm_softpipe.c index a08673444f..09edea6265 100644 --- a/src/mesa/drivers/x11/xm_softpipe.c +++ b/src/mesa/drivers/x11/xm_softpipe.c @@ -113,6 +113,7 @@ xm_buffer_unreference(struct pipe_winsys *pws, struct pipe_buffer_handle **buf) { struct xm_buffer *xm_buf = xm_bo(*buf); xm_buf->refcount--; + assert(xm_buf->refcount >= 0); if (xm_buf->refcount == 0) { if (xm_buf->data) { free(xm_buf->data); @@ -194,6 +195,7 @@ static struct pipe_buffer_handle * xm_buffer_create(struct pipe_winsys *pws, unsigned alignment) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); + buffer->refcount = 1; return pipe_bo(buffer); } diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 6f6c549c07..a751b0868c 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -45,6 +45,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/softpipe/sp_context.h" +#include "pipe/softpipe/sp_clear.h" #include "state_tracker/st_context.h" @@ -265,7 +266,7 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) softpipe_init_surface_funcs(&xms->surface); assert(pipe); - xms->surface.surface.region = pipe->region_alloc(pipe, 1, 0, 0, 0x0); + xms->surface.surface.region = pipe->region_alloc(pipe, 1, 1, 1, 0x0); return &xms->surface.surface; } @@ -284,3 +285,28 @@ xmesa_supported_formats(struct pipe_context *pipe, GLuint *numFormats) return formats; } + +/** + * Called via pipe->clear() + */ +void +xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) +{ + struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); + assert(xrb); + if (xrb->ximage) { + /* clearing back color buffer */ + GET_CURRENT_CONTEXT(ctx); + xmesa_clear_buffers(ctx, BUFFER_BIT_BACK_LEFT); + } + else if (xrb->pixmap) { + /* clearing front color buffer */ + GET_CURRENT_CONTEXT(ctx); + xmesa_clear_buffers(ctx, BUFFER_BIT_FRONT_LEFT); + } + else { + /* clearing other buffer */ + softpipe_clear(pipe, ps, value); + } +} + -- cgit v1.2.3 From 5b59e6116db5241dc9a08b98e5eb8f0c770c05ea Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 10:49:36 +0100 Subject: Use simple st->pipe->clear() in more cases for depth/stencil clears. --- src/mesa/state_tracker/st_cb_clear.c | 31 +++++++++++++++++-------------- 1 file changed, 17 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index d584f0cafc..f6c65ff466 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -408,18 +408,19 @@ static void clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { struct st_renderbuffer *strb = st_renderbuffer(rb); + const GLboolean isDS = is_depth_stencil_format(strb->surface->format); assert(strb->surface->format); - if (!ctx->Scissor.Enabled && - !is_depth_stencil_format(strb->surface->format)) { - /* clear whole depth buffer w/out masking */ - GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); - ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); + if (ctx->Scissor.Enabled || + (isDS && ctx->DrawBuffer->Visual.stencilBits > 0)) { + /* scissoring or we have a combined depth/stencil buffer */ + clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_FALSE); } else { - /* masking or scissoring or combined z/stencil buffer */ - clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_FALSE); + /* simple clear of whole buffer */ + GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); + ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } } @@ -428,18 +429,20 @@ static void clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { struct st_renderbuffer *strb = st_renderbuffer(rb); + const GLboolean isDS = is_depth_stencil_format(strb->surface->format); const GLuint stencilMax = (1 << rb->StencilBits) - 1; GLboolean maskStencil = ctx->Stencil.WriteMask[0] != stencilMax; - if (!maskStencil && !ctx->Scissor.Enabled && - !is_depth_stencil_format(strb->surface->format)) { - /* clear whole stencil buffer w/out masking */ - GLuint clearValue = ctx->Stencil.Clear; - ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); + if (maskStencil || + ctx->Scissor.Enabled || + (isDS && ctx->DrawBuffer->Visual.depthBits > 0)) { + /* masking or scissoring or combined depth/stencil buffer */ + clear_with_quad(ctx, GL_FALSE, GL_FALSE, GL_TRUE); } else { - /* masking or scissoring */ - clear_with_quad(ctx, GL_FALSE, GL_FALSE, GL_TRUE); + /* simple clear of whole buffer */ + GLuint clearValue = ctx->Stencil.Clear; + ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } } -- cgit v1.2.3 From 89cda7bdc85b10526987321214965f21b148735f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 10:50:06 +0100 Subject: replace assertion w/ conditionals in xmesa_clear() --- src/mesa/drivers/x11/xm_surface.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index a751b0868c..d40d9eda84 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -293,13 +293,13 @@ void xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) { struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); - assert(xrb); - if (xrb->ximage) { + + if (xrb && xrb->ximage) { /* clearing back color buffer */ GET_CURRENT_CONTEXT(ctx); xmesa_clear_buffers(ctx, BUFFER_BIT_BACK_LEFT); } - else if (xrb->pixmap) { + else if (xrb && xrb->pixmap) { /* clearing front color buffer */ GET_CURRENT_CONTEXT(ctx); xmesa_clear_buffers(ctx, BUFFER_BIT_FRONT_LEFT); -- cgit v1.2.3 From 5568a7d30120d830c93494a7b3382bfa8b4d2800 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 15:26:33 +0100 Subject: added vbo_use_buffer_objects() to specify that immediate mode data should be put into bufferobjects --- src/mesa/vbo/vbo.h | 3 +++ src/mesa/vbo/vbo_exec_api.c | 35 +++++++++++++++++++++++++++++++++++ src/mesa/vbo/vbo_exec_draw.c | 17 ++++++++++++++--- 3 files changed, 52 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h index 04c59c05b2..4c51b44cda 100644 --- a/src/mesa/vbo/vbo.h +++ b/src/mesa/vbo/vbo.h @@ -114,4 +114,7 @@ void vbo_rebase_prims( GLcontext *ctx, vbo_draw_func draw ); +void vbo_use_buffer_objects(GLcontext *ctx); + + #endif diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index 7f56b3b629..24e2eba472 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -631,6 +631,41 @@ static void vbo_exec_vtxfmt_init( struct vbo_exec_context *exec ) } +/** + * Tell the VBO module to use a real OpenGL vertex buffer object to + * store accumulated immediate-mode vertex data. + * This replaces the malloced buffer which was created in + * vb_exec_vtx_init() below. + */ +void vbo_use_buffer_objects(GLcontext *ctx) +{ + struct vbo_exec_context *exec = &vbo_context(ctx)->exec; + /* Any buffer name but 0 can be used here since this bufferobj won't + * go into the bufferobj hashtable. + */ + GLuint bufName = 0xaabbccdd; + GLenum target = GL_ARRAY_BUFFER_ARB; + GLenum access = GL_READ_WRITE_ARB; + GLenum usage = GL_STREAM_DRAW_ARB; + GLsizei size = VBO_VERT_BUFFER_SIZE * sizeof(GLfloat); + + /* Make sure this func is only used once */ + assert(exec->vtx.bufferobj == ctx->Array.NullBufferObj); + if (exec->vtx.buffer_map) { + _mesa_align_free(exec->vtx.buffer_map); + } + + /* Allocate a real buffer object now */ + exec->vtx.bufferobj = ctx->Driver.NewBufferObject(ctx, bufName, target); + ctx->Driver.BufferData(ctx, target, size, NULL, usage, exec->vtx.bufferobj); + + /* and map it */ + exec->vtx.buffer_map + = ctx->Driver.MapBuffer(ctx, target, access, exec->vtx.bufferobj); +} + + + void vbo_exec_vtx_init( struct vbo_exec_context *exec ) { GLcontext *ctx = exec->ctx; diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c index 0ef26cdfe3..68aba1df66 100644 --- a/src/mesa/vbo/vbo_exec_draw.c +++ b/src/mesa/vbo/vbo_exec_draw.c @@ -175,16 +175,27 @@ static void vbo_exec_bind_arrays( GLcontext *ctx ) * arrays of floats. */ for (attr = 0; attr < VERT_ATTRIB_MAX ; attr++) { - GLuint src = map[attr]; + const GLuint src = map[attr]; if (exec->vtx.attrsz[src]) { - arrays[attr].Ptr = (void *)data; + if (exec->vtx.bufferobj->Name) { + /* a real buffer obj: Ptr is an offset, not a pointer*/ + int offset; + assert(exec->vtx.bufferobj->Pointer); /* buf should be mapped */ + offset = (GLbyte *) data - (GLbyte *) exec->vtx.bufferobj->Pointer; + assert(offset >= 0); + arrays[attr].Ptr = (void *) offset; + } + else { + /* Ptr into ordinary app memory */ + arrays[attr].Ptr = (void *) data; + } arrays[attr].Size = exec->vtx.attrsz[src]; arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat); arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat); arrays[attr].Type = GL_FLOAT; arrays[attr].Enabled = 1; - arrays[attr].BufferObj = exec->vtx.bufferobj; /* NullBufferObj */ + arrays[attr].BufferObj = exec->vtx.bufferobj; arrays[attr]._MaxElement = count; /* ??? */ data += exec->vtx.attrsz[attr] * sizeof(GLfloat); -- cgit v1.2.3 From 305b83a4153442e53f54cd5b817238c95161f7db Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 15:27:00 +0100 Subject: added call to _mesa_load_state_parameters() --- src/mesa/state_tracker/st_atom_fs.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 41ad49b317..9aba9aaa56 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -74,6 +74,9 @@ static void update_fs( struct st_context *st ) if (fp && params) { /* load program's constants array */ + + _mesa_load_state_parameters(st->ctx, params); + fp->constants.nr_constants = params->NumParameters; memcpy(fp->constants.constant, params->ParameterValues, -- cgit v1.2.3 From 1e6d1ab6fc85ff928c629627991a60f515b73857 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 15:27:18 +0100 Subject: tell vbo module to use bufferobjs --- src/mesa/state_tracker/st_context.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 35a262ab5b..9e89ece52e 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -25,7 +25,8 @@ * **************************************************************************/ -#include "imports.h" +#include "main/imports.h" +#include "vbo/vbo.h" #include "st_public.h" #include "st_context.h" #include "st_cb_bufferobjects.h" @@ -68,6 +69,9 @@ struct st_context *st_create_context( GLcontext *ctx, st_init_atoms( st ); st_init_draw( st ); + /* we want all vertex data to be placed in buffer objects */ + vbo_use_buffer_objects(ctx); + /* Need these flags: */ st->ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; -- cgit v1.2.3 From c5004c7db6991a11ffbc76cd4a7a0ce940f8b54a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 15:28:17 +0100 Subject: pack output vertex attributes in sequential slots --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 8f684ba00a..18ed3c9d5a 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -144,6 +144,12 @@ run_vertex_program(struct draw_context *draw, machine.Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ machine.Inputs[attr].xyzw[3].f[j] = 1.0; /*W*/ +#if 0 + if (attr == 0) { + printf("Input vertex %d: %f %f %f\n", + j, p[0], p[1], p[2]); + } +#endif } } } @@ -173,7 +179,7 @@ run_vertex_program(struct draw_context *draw, /* store machine results */ assert(sp->vs.outputs_written & (1 << VERT_RESULT_HPOS)); for (j = 0; j < count; j++) { - unsigned attr; + unsigned attr, slot; float x, y, z, w; /* Handle attr[0] (position) specially: */ @@ -183,7 +189,7 @@ run_vertex_program(struct draw_context *draw, w = vOut[j]->clip[3] = outputs[0].xyzw[3].f[j]; vOut[j]->clipmask = compute_clipmask(x, y, z, w); - vOut[j]->edgeflag = 0; + vOut[j]->edgeflag = 1; /* divide by w */ w = 1.0 / w; @@ -204,12 +210,16 @@ run_vertex_program(struct draw_context *draw, #endif /* remaining attributes: */ + /* pack into sequential post-transform attrib slots */ + slot = 1; for (attr = 1; attr < VERT_RESULT_MAX; attr++) { if (sp->vs.outputs_written & (1 << attr)) { - vOut[j]->data[attr][0] = outputs[attr].xyzw[0].f[j]; - vOut[j]->data[attr][1] = outputs[attr].xyzw[1].f[j]; - vOut[j]->data[attr][2] = outputs[attr].xyzw[2].f[j]; - vOut[j]->data[attr][3] = outputs[attr].xyzw[3].f[j]; + assert(slot < draw->nr_attrs - 2); + vOut[j]->data[slot][0] = outputs[attr].xyzw[0].f[j]; + vOut[j]->data[slot][1] = outputs[attr].xyzw[1].f[j]; + vOut[j]->data[slot][2] = outputs[attr].xyzw[2].f[j]; + vOut[j]->data[slot][3] = outputs[attr].xyzw[3].f[j]; + slot++; } } } -- cgit v1.2.3 From bc3f87683a25803fde6e69eddccc7e0a1c59df17 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 15:37:22 +0100 Subject: fix edgeflag bug --- src/mesa/pipe/draw/draw_prim.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 7757dacf34..95e71efd6d 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -258,8 +258,10 @@ static void do_quad( struct draw_context *draw, unsigned v2, unsigned v3 ) { - do_ef_triangle( draw, 1, ~(1<<0), v0, v1, v3 ); - do_ef_triangle( draw, 0, ~(1<<1), v1, v2, v3 ); + const unsigned omitEdge2 = ~(1 << 1); + const unsigned omitEdge3 = ~(1 << 2); + do_ef_triangle( draw, 1, omitEdge2, v0, v1, v3 ); + do_ef_triangle( draw, 0, omitEdge3, v1, v2, v3 ); } -- cgit v1.2.3 From 8e4ed6323dff7c021b0f7d7711a89379e26ddb25 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 15:38:29 +0100 Subject: added null VB->AttribPtr[i] check --- src/mesa/tnl/t_vp_build.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index ee1a2498b3..eb5e176895 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -146,7 +146,7 @@ static struct state_key *make_state_key( GLcontext *ctx ) } for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) - if (VB->AttribPtr[i]->stride) + if (VB->AttribPtr[i] && VB->AttribPtr[i]->stride) key->light_material_mask |= 1<<(i-_TNL_ATTRIB_MAT_FRONT_AMBIENT); for (i = 0; i < MAX_LIGHTS; i++) { -- cgit v1.2.3 From 2fc54f5bb56a7714c843fc5a6a3bb6895af9babc Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 11:39:45 -0600 Subject: added uint64 type --- src/mesa/pipe/p_compiler.h | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h index 4f968404eb..d139cb1844 100644 --- a/src/mesa/pipe/p_compiler.h +++ b/src/mesa/pipe/p_compiler.h @@ -39,15 +39,17 @@ #endif -typedef unsigned int uint; -typedef unsigned char ubyte; -typedef unsigned char boolean; -typedef unsigned short ushort; +typedef unsigned int uint; +typedef unsigned char ubyte; +typedef unsigned char boolean; +typedef unsigned short ushort; +typedef unsigned long long uint64; #define TRUE 1 #define FALSE 0 + /* Function inlining */ #if defined(__GNUC__) # define INLINE __inline__ -- cgit v1.2.3 From cd3162f57831fd0b8d9af6bfcd08a521ab352386 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 11:40:06 -0600 Subject: added align16() function --- src/mesa/pipe/p_util.h | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 69b69a3c67..3eff3328c7 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -38,6 +38,22 @@ #define Elements(x) sizeof(x)/sizeof(*(x)) +/** + * Return pointer aligned to next multiple of 16 bytes. + */ +static INLINE void * +align16(void *unaligned) +{ + union { + void *p; + uint64 u; + } pu; + pu.p = unaligned; + pu.u = (pu.u + 15) & ~15; + return pu.p; +} + + #if defined(__MSC__) && defined(__WIN32__) static INLINE unsigned ffs( unsigned u ) { -- cgit v1.2.3 From 83f428e799d598732494f60601c5984e09829a81 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 11:40:53 -0600 Subject: shade four vertices at a time --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 171 +++++++++++++++++--------------- 1 file changed, 92 insertions(+), 79 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 69fd6908bb..bb32553cf4 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -31,10 +31,15 @@ */ +/** TEMP */ +#include "main/context.h" +#include "main/macros.h" + #include "pipe/p_defines.h" #include "pipe/p_context.h" #include "pipe/p_winsys.h" + #include "sp_context.h" #include "sp_state.h" @@ -48,35 +53,43 @@ #if defined __GNUC__ -#define USE_ALIGNED_ATTRIBS 1 -#define ALIGN16_SUFFIX __attribute__(( aligned( 16 ) )) +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE] __attribute__(( aligned( 16 ) )) +#define ALIGN16_ASSIGN(P) P #else -#define USE_ALIGNED_ATTRIBS 0 -#define ALIGN16_SUFFIX +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE + 1] +#define ALIGN16_ASSIGN(P) align16(P) #endif static struct softpipe_context *sp_global = NULL; + +/** + * Transform vertices with the current vertex program/shader + * Up to four vertices can be shaded at a time. + * \param vbuffer the input vertex data + * \param elts indexes of four input vertices + * \param count number of vertices to shade [1..4] + * \param vOut array of pointers to four output vertices + */ static void -run_vertex_program2(struct draw_context *draw, - const void *vbuffer, unsigned elem, - struct vertex_header *vOut) +run_vertex_program(struct draw_context *draw, + const void *vbuffer, unsigned elts[4], unsigned count, + struct vertex_header *vOut[]) { #if 1 struct softpipe_context *sp = sp_global; #endif struct tgsi_exec_machine machine; - int i; + unsigned int j; -#if USE_ALIGNED_ATTRIBS - struct tgsi_exec_vector inputs[PIPE_ATTRIB_MAX] ALIGN16_SUFFIX; - struct tgsi_exec_vector outputs[PIPE_ATTRIB_MAX] ALIGN16_SUFFIX; -#else - struct tgsi_exec_vector inputs[PIPE_ATTRIB_MAX + 1]; - struct tgsi_exec_vector outputs[PIPE_ATTRIB_MAX + 1]; -#endif + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); #ifdef DEBUG memset( &machine, 0, sizeof( machine ) ); @@ -92,54 +105,34 @@ run_vertex_program2(struct draw_context *draw, /* Consts does not require 16 byte alignment. */ machine.Consts = sp->vs.constants->constant; -#if USE_ALIGNED_ATTRIBS - machine.Inputs = inputs; - machine.Outputs = outputs; -#else - machine.Inputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); - machine.Outputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs ); -#endif + machine.Inputs = ALIGN16_ASSIGN(inputs); + machine.Outputs = ALIGN16_ASSIGN(outputs); - { + /* load machine inputs */ + for (j = 0; j < count; j++) { const void *mapped = vbuffer; const float *vIn, *cIn; vIn = (const float *) ((const ubyte *) mapped + draw->vertex_buffer[0].buffer_offset + draw->vertex_element[0].src_offset - + elem * draw->vertex_buffer[0].pitch); + + elts[j] * draw->vertex_buffer[0].pitch); cIn = (const float *) ((const ubyte *) mapped + draw->vertex_buffer[3].buffer_offset + draw->vertex_element[3].src_offset - + elem * draw->vertex_buffer[3].pitch); - /*X*/ - machine.Inputs[0].xyzw[0].f[0] = vIn[0]; - machine.Inputs[0].xyzw[0].f[1] = vIn[0]; - machine.Inputs[0].xyzw[0].f[2] = vIn[0]; - machine.Inputs[0].xyzw[0].f[3] = vIn[0]; - - /*Y*/ - machine.Inputs[0].xyzw[1].f[0] = vIn[1]; - machine.Inputs[0].xyzw[1].f[1] = vIn[1]; - machine.Inputs[0].xyzw[1].f[2] = vIn[1]; - machine.Inputs[0].xyzw[1].f[3] = vIn[1]; - - /*Z*/ - machine.Inputs[0].xyzw[2].f[0] = vIn[2]; - machine.Inputs[0].xyzw[2].f[1] = vIn[2]; - machine.Inputs[0].xyzw[2].f[2] = vIn[2]; - machine.Inputs[0].xyzw[2].f[3] = vIn[2]; - - /*W*/ - machine.Inputs[0].xyzw[3].f[0] = 1.0; - machine.Inputs[0].xyzw[3].f[1] = 1.0; - machine.Inputs[0].xyzw[3].f[2] = 1.0; - machine.Inputs[0].xyzw[3].f[3] = 1.0; - - printf("VS Input: %f %f %f %f\n", - vIn[0], vIn[1], vIn[2], 1.0); + + elts[j] * draw->vertex_buffer[3].pitch); + + machine.Inputs[0].xyzw[0].f[j] = vIn[0]; /*X*/ + machine.Inputs[0].xyzw[1].f[j] = vIn[1]; /*Y*/ + machine.Inputs[0].xyzw[2].f[j] = vIn[2]; /*Z*/ + machine.Inputs[0].xyzw[3].f[j] = 1.0; /*W*/ +#if 0 + printf("VS Input %d: %f %f %f %f\n", + j, vIn[0], vIn[1], vIn[2], 1.0); +#endif } +#if 0 printf("Consts:\n"); for (i = 0; i < 4; i++) { printf(" %d: %f %f %f %f\n", i, @@ -148,48 +141,49 @@ run_vertex_program2(struct draw_context *draw, machine.Consts[i][2], machine.Consts[i][3]); } - +#endif /* run shader */ tgsi_exec_machine_run( &machine ); - /* store result pos */ +#if 0 printf("VS result: %f %f %f %f\n", outputs[0].xyzw[0].f[0], outputs[0].xyzw[1].f[0], outputs[0].xyzw[2].f[0], outputs[0].xyzw[3].f[0]); - { - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; +#endif + + /* store machine results */ + for (j = 0; j < count; j++) { float x, y, z, w; - x = outputs[0].xyzw[0].f[0]; - y = outputs[0].xyzw[1].f[0]; - z = outputs[0].xyzw[2].f[0]; - w = outputs[0].xyzw[3].f[0]; + x = outputs[0].xyzw[0].f[j]; + y = outputs[0].xyzw[1].f[j]; + z = outputs[0].xyzw[2].f[j]; + w = outputs[0].xyzw[3].f[j]; /* divide by w */ x /= w; y /= w; z /= w; - w = 1.0f / w; + w = 1.0 / w; /* Viewport */ - vOut->data[0][0] = scale[0] * x + trans[0]; - vOut->data[0][1] = scale[1] * y + trans[1]; - vOut->data[0][2] = scale[2] * z + trans[2]; - vOut->data[0][3] = w; + vOut[j]->data[0][0] = scale[0] * x + trans[0]; + vOut[j]->data[0][1] = scale[1] * y + trans[1]; + vOut[j]->data[0][2] = scale[2] * z + trans[2]; + vOut[j]->data[0][3] = w; +#if 0 printf("wincoord: %f %f %f\n", - vOut->data[0][0], - vOut->data[0][1], - vOut->data[0][2]); - - vOut->data[1][0] = 1.0; - vOut->data[1][1] = 1.0; - vOut->data[1][2] = 1.0; - vOut->data[1][3] = 1.0; - + vOut[j]->data[0][0], + vOut[j]->data[0][1], + vOut[j]->data[0][2]); +#endif + vOut[j]->data[1][0] = 1.0; + vOut[j]->data[1][1] = 1.0; + vOut[j]->data[1][2] = 1.0; + vOut[j]->data[1][3] = 1.0; } #if 0 @@ -209,14 +203,12 @@ run_vertex_program2(struct draw_context *draw, * \param elem which element of the vertex buffer to use as input * \param vOut the output vertex */ +#if 0 static void run_vertex_program(struct draw_context *draw, const void *vbuffer, unsigned elem, struct vertex_header *vOut) { - run_vertex_program2(draw, vbuffer, elem, vOut); - -#if 0 const float *vIn, *cIn; const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; @@ -268,8 +260,8 @@ run_vertex_program(struct draw_context *draw, vOut->data[1][2] = cIn[2]; vOut->data[1][3] = 1.0; } -#endif } +#endif /** @@ -278,7 +270,7 @@ run_vertex_program(struct draw_context *draw, */ static void vs_flush( struct draw_context *draw ) { - unsigned i; + unsigned i, j; /* We're not really running a vertex shader yet, so flushing the vs * queue is just a matter of building the vertices and returning. @@ -286,7 +278,11 @@ static void vs_flush( struct draw_context *draw ) /* Actually, I'm cheating even more and pre-building them still * with the mesa/vf module. So it's very easy... */ +#if 0 for (i = 0; i < draw->vs.queue_nr; i++) { +#else + for (i = 0; i < draw->vs.queue_nr; i+=4) { +#endif /* Would do the following steps here: * * 1) Loop over vertex element descriptors, fetch data from each @@ -302,10 +298,27 @@ static void vs_flush( struct draw_context *draw ) * * In this version, just do the last step: */ +#if 0 const unsigned elt = draw->vs.queue[i].elt; struct vertex_header *dest = draw->vs.queue[i].dest; run_vertex_program(draw, draw->mapped_vbuffer, elt, dest); +#else + struct vertex_header *dests[4]; + unsigned elts[4]; + int n; + + for (j = 0; j < 4; j++) { + elts[j] = draw->vs.queue[i + j].elt; + dests[j] = draw->vs.queue[i + j].dest; + } + + n = MIN2(4, draw->vs.queue_nr - i); + assert(n > 0); + assert(n <= 4); + + run_vertex_program(draw, draw->mapped_vbuffer, elts, n, dests); +#endif } draw->vs.queue_nr = 0; } -- cgit v1.2.3 From 7668e53c8c64570d66a626c96302a953164f319e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 12:09:31 -0600 Subject: define CLIP_LEFT/RIGHT etc --- src/mesa/pipe/draw/draw_clip.c | 5 ++--- src/mesa/pipe/draw/draw_context.h | 11 +++++++++++ 2 files changed, 13 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 53a7b10f7d..90bfb61568 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -33,8 +33,10 @@ #include "pipe/p_util.h" +#include "draw_context.h" #include "draw_private.h" + #ifndef IS_NEGATIVE #define IS_NEGATIVE(X) ((X) < 0.0) #endif @@ -136,9 +138,6 @@ static void interp( const struct clipper *clip, } -#define CLIP_USER_BIT 0x40 -#define CLIP_CULL_BIT 0x80 - static INLINE float dot4( const float *a, const float *b ) diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index ac994d6419..afa3a04165 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -46,6 +46,17 @@ struct draw_context; struct draw_stage; +#define CLIP_RIGHT_BIT 0x01 +#define CLIP_LEFT_BIT 0x02 +#define CLIP_TOP_BIT 0x04 +#define CLIP_BOTTOM_BIT 0x08 +#define CLIP_NEAR_BIT 0x10 +#define CLIP_FAR_BIT 0x20 +#define CLIP_USER_BIT 0x40 +#define CLIP_CULL_BIT 0x80 + + + struct draw_context *draw_create( void ); void draw_destroy( struct draw_context *draw ); -- cgit v1.2.3 From e3647578d159dcb93799023eacba0c2d3c7fbd18 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 12:09:50 -0600 Subject: compute clipmask --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 67 +++++++++++++++++++++++++-------- 1 file changed, 51 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index bb32553cf4..52a54a2c96 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -65,6 +65,32 @@ static struct softpipe_context *sp_global = NULL; +static INLINE unsigned +compute_clipmask(float cx, float cy, float cz, float cw) +{ + unsigned mask; +#if defined(macintosh) || defined(__powerpc__) + /* on powerpc cliptest is 17% faster in this way. */ + mask = (((cw < cx) << CLIP_RIGHT_SHIFT)); + mask |= (((cw < -cx) << CLIP_LEFT_SHIFT)); + mask |= (((cw < cy) << CLIP_TOP_SHIFT)); + mask |= (((cw < -cy) << CLIP_BOTTOM_SHIFT)); + mask |= (((cw < cz) << CLIP_FAR_SHIFT)); + mask |= (((cw < -cz) << CLIP_NEAR_SHIFT)); +#else /* !defined(macintosh)) */ + mask = 0x0; + if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT; + if ( cx + cw < 0) mask |= CLIP_LEFT_BIT; + if (-cy + cw < 0) mask |= CLIP_TOP_BIT; + if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT; + if (-cz + cw < 0) mask |= CLIP_FAR_BIT; + if ( cz + cw < 0) mask |= CLIP_NEAR_BIT; +#endif /* defined(macintosh) */ + return mask; +} + + + /** * Transform vertices with the current vertex program/shader * Up to four vertices can be shaded at a time. @@ -126,6 +152,12 @@ run_vertex_program(struct draw_context *draw, machine.Inputs[0].xyzw[1].f[j] = vIn[1]; /*Y*/ machine.Inputs[0].xyzw[2].f[j] = vIn[2]; /*Z*/ machine.Inputs[0].xyzw[3].f[j] = 1.0; /*W*/ + + machine.Inputs[3].xyzw[0].f[j] = cIn[0]; + machine.Inputs[3].xyzw[1].f[j] = cIn[1]; + machine.Inputs[3].xyzw[2].f[j] = cIn[2]; + machine.Inputs[3].xyzw[3].f[j] = 1.0; + #if 0 printf("VS Input %d: %f %f %f %f\n", j, vIn[0], vIn[1], vIn[2], 1.0); @@ -158,21 +190,24 @@ run_vertex_program(struct draw_context *draw, for (j = 0; j < count; j++) { float x, y, z, w; - x = outputs[0].xyzw[0].f[j]; - y = outputs[0].xyzw[1].f[j]; - z = outputs[0].xyzw[2].f[j]; - w = outputs[0].xyzw[3].f[j]; + x = vOut[j]->clip[0] = outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(x, y, z, w); + vOut[j]->edgeflag = 0; /* divide by w */ - x /= w; - y /= w; - z /= w; w = 1.0 / w; - - /* Viewport */ - vOut[j]->data[0][0] = scale[0] * x + trans[0]; - vOut[j]->data[0][1] = scale[1] * y + trans[1]; - vOut[j]->data[0][2] = scale[2] * z + trans[2]; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; vOut[j]->data[0][3] = w; #if 0 printf("wincoord: %f %f %f\n", @@ -180,10 +215,10 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][1], vOut[j]->data[0][2]); #endif - vOut[j]->data[1][0] = 1.0; - vOut[j]->data[1][1] = 1.0; - vOut[j]->data[1][2] = 1.0; - vOut[j]->data[1][3] = 1.0; + vOut[j]->data[1][0] = outputs[1].xyzw[0].f[j]; + vOut[j]->data[1][1] = outputs[1].xyzw[1].f[j]; + vOut[j]->data[1][2] = outputs[1].xyzw[2].f[j]; + vOut[j]->data[1][3] = outputs[1].xyzw[3].f[j]; } #if 0 -- cgit v1.2.3 From a30939714cdc73cb633672cad7d86b425781526d Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 12:19:40 -0600 Subject: added pipe ptr to draw_context --- src/mesa/pipe/draw/draw_private.h | 2 ++ src/mesa/pipe/softpipe/sp_draw_arrays.c | 10 ++-------- 2 files changed, 4 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 2748772fdb..278950ec48 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -161,6 +161,8 @@ struct draw_context const struct pipe_vertex_buffer *vertex_buffer; /**< note: pointer */ const struct pipe_vertex_element *vertex_element; /**< note: pointer */ + /** need to know the pipe for vertex flushing/transformation: */ + struct pipe_context *pipe; /* Clip derived state: */ diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 52a54a2c96..ac7ae9a887 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -61,9 +61,6 @@ #endif -static struct softpipe_context *sp_global = NULL; - - static INLINE unsigned compute_clipmask(float cx, float cy, float cz, float cw) @@ -104,9 +101,7 @@ run_vertex_program(struct draw_context *draw, const void *vbuffer, unsigned elts[4], unsigned count, struct vertex_header *vOut[]) { -#if 1 - struct softpipe_context *sp = sp_global; -#endif + struct softpipe_context *sp = softpipe_context(draw->pipe); struct tgsi_exec_machine machine; unsigned int j; @@ -368,8 +363,6 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, struct draw_context *draw = sp->draw; struct pipe_buffer_handle *buf; - sp_global = sp; - softpipe_map_surfaces(sp); /* @@ -384,6 +377,7 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, draw->pipeline.first->begin( draw->pipeline.first ); draw->vs_flush = vs_flush; + draw->pipe = pipe; /* XXX pass pipe to draw_create() */ draw_invalidate_vcache( draw ); -- cgit v1.2.3 From 4973136e2062fe6930cee1a44f97aefa6b423295 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 12:23:26 -0600 Subject: load machine.Inputs based on vs.inputs_read --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 41 ++++++++++++--------------------- 1 file changed, 15 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index ac7ae9a887..be03ae8657 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -131,32 +131,21 @@ run_vertex_program(struct draw_context *draw, /* load machine inputs */ for (j = 0; j < count; j++) { - const void *mapped = vbuffer; - const float *vIn, *cIn; - vIn = (const float *) ((const ubyte *) mapped - + draw->vertex_buffer[0].buffer_offset - + draw->vertex_element[0].src_offset - + elts[j] * draw->vertex_buffer[0].pitch); - - cIn = (const float *) ((const ubyte *) mapped - + draw->vertex_buffer[3].buffer_offset - + draw->vertex_element[3].src_offset - + elts[j] * draw->vertex_buffer[3].pitch); - - machine.Inputs[0].xyzw[0].f[j] = vIn[0]; /*X*/ - machine.Inputs[0].xyzw[1].f[j] = vIn[1]; /*Y*/ - machine.Inputs[0].xyzw[2].f[j] = vIn[2]; /*Z*/ - machine.Inputs[0].xyzw[3].f[j] = 1.0; /*W*/ - - machine.Inputs[3].xyzw[0].f[j] = cIn[0]; - machine.Inputs[3].xyzw[1].f[j] = cIn[1]; - machine.Inputs[3].xyzw[2].f[j] = cIn[2]; - machine.Inputs[3].xyzw[3].f[j] = 1.0; - -#if 0 - printf("VS Input %d: %f %f %f %f\n", - j, vIn[0], vIn[1], vIn[2], 1.0); -#endif + unsigned attr; + for (attr = 0; attr < 16; attr++) { + if (sp->vs.inputs_read & (1 << attr)) { + const float *p + = (const float *) ((const ubyte *) vbuffer + + draw->vertex_buffer[attr].buffer_offset + + draw->vertex_element[attr].src_offset + + elts[j] * draw->vertex_buffer[attr].pitch); + + machine.Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ + machine.Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ + machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ + machine.Inputs[attr].xyzw[3].f[j] = 1.0; /*W*/ + } + } } #if 0 -- cgit v1.2.3 From a2805a81658fd4a4b9732167f24f334630549fa6 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 12:30:30 -0600 Subject: added outputs_written field to pipe_shader_state --- src/mesa/pipe/p_state.h | 3 ++- src/mesa/state_tracker/st_atom_fs.c | 1 + src/mesa/state_tracker/st_atom_vs.c | 1 + 3 files changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index c1972c56b3..42bf50a617 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -122,7 +122,8 @@ struct pipe_constant_buffer { struct pipe_shader_state { - unsigned inputs_read; /* FRAG/VERT_ATTRIB_* */ + unsigned inputs_read; /**< FRAG/VERT_ATTRIB_x */ + unsigned outputs_written; /**< FRAG/VERT_RESULT_x */ const struct tgsi_token *tokens; struct pipe_constant_buffer *constants; /* XXX temporary? */ }; diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 7e12441784..41ad49b317 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -85,6 +85,7 @@ static void update_fs( struct st_context *st ) memset( &fs, 0, sizeof(fs) ); fs.inputs_read = fp->Base.Base.InputsRead; + fs.outputs_written = fp->Base.Base.OutputsWritten; fs.tokens = &fp->tokens[0]; fs.constants = &fp->constants; diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index d94c560bef..cef6e38d56 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -95,6 +95,7 @@ static void update_vs( struct st_context *st ) compile_vs( st, vp ); memset( &vs, 0, sizeof(vs) ); + vs.outputs_written = vp->Base.Base.OutputsWritten; vs.inputs_read = vp->Base.Base.InputsRead; vs.tokens = &vp->tokens[0]; vs.constants = &vp->constants; -- cgit v1.2.3 From 48b09322ee1b701558e1f223320cd2a9259bb37f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 12:31:13 -0600 Subject: write outputs according to vs.outputs_written --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index be03ae8657..8f684ba00a 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -171,9 +171,12 @@ run_vertex_program(struct draw_context *draw, #endif /* store machine results */ + assert(sp->vs.outputs_written & (1 << VERT_RESULT_HPOS)); for (j = 0; j < count; j++) { + unsigned attr; float x, y, z, w; + /* Handle attr[0] (position) specially: */ x = vOut[j]->clip[0] = outputs[0].xyzw[0].f[j]; y = vOut[j]->clip[1] = outputs[0].xyzw[1].f[j]; z = vOut[j]->clip[2] = outputs[0].xyzw[2].f[j]; @@ -199,10 +202,16 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][1], vOut[j]->data[0][2]); #endif - vOut[j]->data[1][0] = outputs[1].xyzw[0].f[j]; - vOut[j]->data[1][1] = outputs[1].xyzw[1].f[j]; - vOut[j]->data[1][2] = outputs[1].xyzw[2].f[j]; - vOut[j]->data[1][3] = outputs[1].xyzw[3].f[j]; + + /* remaining attributes: */ + for (attr = 1; attr < VERT_RESULT_MAX; attr++) { + if (sp->vs.outputs_written & (1 << attr)) { + vOut[j]->data[attr][0] = outputs[attr].xyzw[0].f[j]; + vOut[j]->data[attr][1] = outputs[attr].xyzw[1].f[j]; + vOut[j]->data[attr][2] = outputs[attr].xyzw[2].f[j]; + vOut[j]->data[attr][3] = outputs[attr].xyzw[3].f[j]; + } + } } #if 0 -- cgit v1.2.3 From 74e58c5b82a15d250f42c9695e3ed61f64faedaf Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 12:34:40 -0600 Subject: added CLIP_x_SHIFT vals --- src/mesa/pipe/draw/draw_context.h | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index afa3a04165..19fddf7104 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -46,6 +46,10 @@ struct draw_context; struct draw_stage; +/** + * Clipmask flags + */ +/*@{*/ #define CLIP_RIGHT_BIT 0x01 #define CLIP_LEFT_BIT 0x02 #define CLIP_TOP_BIT 0x04 @@ -54,7 +58,19 @@ struct draw_stage; #define CLIP_FAR_BIT 0x20 #define CLIP_USER_BIT 0x40 #define CLIP_CULL_BIT 0x80 +/*@}*/ +/** + * Bitshift for each clip flag + */ +/*@{*/ +#define CLIP_RIGHT_SHIFT 0 +#define CLIP_LEFT_SHIFT 1 +#define CLIP_TOP_SHIFT 2 +#define CLIP_BOTTOM_SHIFT 3 +#define CLIP_NEAR_SHIFT 4 +#define CLIP_FAR_SHIFT 5 +/*@}*/ struct draw_context *draw_create( void ); -- cgit v1.2.3 From 8339ca7d1e9a3fe90f46e6e81f7ec8574d121072 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 17 Aug 2007 12:40:35 -0600 Subject: replace memcpy() with assignment, better type safety --- src/mesa/pipe/draw/draw_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 7e6a95127a..d5d7d408de 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -158,7 +158,7 @@ static void validate_pipeline( struct draw_context *draw ) void draw_set_setup_state( struct draw_context *draw, const struct pipe_setup_state *setup ) { - memcpy( &draw->setup, setup, sizeof(*setup) ); + draw->setup = *setup; /* struct copy */ validate_pipeline( draw ); } -- cgit v1.2.3 From 7274e14d4ada04d150f1115252cebd7a3ada797a Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 18 Aug 2007 09:22:15 +0100 Subject: fix xm_buffer_data --- src/mesa/drivers/x11/xm_softpipe.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_softpipe.c b/src/mesa/drivers/x11/xm_softpipe.c index 09edea6265..432f306e94 100644 --- a/src/mesa/drivers/x11/xm_softpipe.c +++ b/src/mesa/drivers/x11/xm_softpipe.c @@ -51,6 +51,7 @@ struct xm_softpipe_winsys struct xm_buffer { int refcount; + int size; void *data; void *mapped; }; @@ -129,8 +130,12 @@ xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, unsigned size, const void *data ) { struct xm_buffer *xm_buf = xm_bo(buf); - assert(!xm_buf->data); - xm_buf->data = malloc(size); + if (xm_buf->size != size) { + if (xm_buf->data) + free(xm_buf->data); + xm_buf->data = malloc(size); + xm_buf->size = size; + } if (data) memcpy(xm_buf->data, data, size); } -- cgit v1.2.3 From b3f067e8593b9fda0474041f32479b20cd64e728 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 18 Aug 2007 16:04:49 +0100 Subject: added tnl state atom --- src/mesa/state_tracker/st_atom.c | 1 + src/mesa/state_tracker/st_atom.h | 1 + src/mesa/state_tracker/st_atom_vs.c | 31 ++++++++++++++++++++++++++++++- 3 files changed, 32 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 32b8b8f277..d67291e50b 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -46,6 +46,7 @@ static const struct st_tracked_state *atoms[] = &st_update_clear_color, &st_update_depth, &st_update_clip, + &st_update_tnl, &st_update_vs, &st_update_fs, &st_update_setup, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 2f628206ca..51da489f6d 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -48,6 +48,7 @@ const struct st_tracked_state st_update_framebuffer; const struct st_tracked_state st_update_clip; const struct st_tracked_state st_update_clear_color; const struct st_tracked_state st_update_depth; +const struct st_tracked_state st_update_tnl; const struct st_tracked_state st_update_fs; const struct st_tracked_state st_update_vs; const struct st_tracked_state st_update_setup; diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index cef6e38d56..c8bd805e02 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -27,9 +27,11 @@ /* * Authors: * Keith Whitwell + * Brian Paul */ #include "shader/prog_parameter.h" +#include "tnl/t_vp_build.h" #include "pipe/p_context.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" @@ -39,6 +41,7 @@ #include "st_atom.h" #include "st_program.h" + #define TGSI_DEBUG 0 static void compile_vs( struct st_context *st, @@ -59,8 +62,10 @@ static void update_vs( struct st_context *st ) struct st_vertex_program *vp = NULL; struct gl_program_parameter_list *params = NULL; +#if 0 if (st->ctx->VertexProgram._MaintainTnlProgram) _tnl_UpdateFixedFunctionProgram( st->ctx ); +#endif if (st->ctx->Shader.CurrentProgram && st->ctx->Shader.CurrentProgram->LinkStatus && @@ -83,6 +88,7 @@ static void update_vs( struct st_context *st ) if (vp && params) { /* load program's constants array */ + /* XXX this should probably be done elsewhere/separately */ _mesa_load_state_parameters(st->ctx, params); vp->constants.nr_constants = params->NumParameters; @@ -117,3 +123,26 @@ const struct st_tracked_state st_update_vs = { }, .update = update_vs }; + + + + + +/** + * When TnL state has changed, need to generate new vertex program. + * This should be done before updating the vertes shader (vs) state. + */ +static void update_tnl( struct st_context *st ) +{ + if (st->ctx->VertexProgram._MaintainTnlProgram) + _tnl_UpdateFixedFunctionProgram( st->ctx ); +} + + +const struct st_tracked_state st_update_tnl = { + .dirty = { + .mesa = _NEW_PROGRAM | _NEW_LIGHT | _NEW_TEXTURE, /* XXX more? */ + .st = ST_NEW_MESA, /* XXX correct? */ + }, + .update = update_tnl +}; -- cgit v1.2.3 From 1f026d98dd77b6d26cc76946d92f69a9e6091b8e Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 18 Aug 2007 16:05:49 +0100 Subject: Create a default vertex attribute buffer which mirrors ctx->Current.Attrrib[] Used when the vertex shader references attributes which aren't present in VBOs. --- src/mesa/state_tracker/st_context.h | 2 + src/mesa/state_tracker/st_draw.c | 113 +++++++++++++++++++++++++++--------- 2 files changed, 88 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 25ed20a5f8..35774a790e 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -104,6 +104,8 @@ struct st_context struct st_state_flags dirty; GLfloat polygon_offset_scale; /* ?? */ + + struct pipe_buffer_handle *default_attrib_buffer; }; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 425076d653..db24b24fa7 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -44,6 +44,7 @@ #include "st_cb_bufferobjects.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #include "vbo/vbo_context.h" @@ -52,7 +53,7 @@ * bypassing the T&L module. This only works with VBO-based demos, * such as progs/test/bufferobj.c */ -#define USE_NEW_DRAW 0 +#define USE_NEW_DRAW 01 /* @@ -134,6 +135,38 @@ pipe_vertex_format(GLenum format, GLuint size) +/** + * The default attribute buffer is basically a copy of the + * ctx->Current.Attrib[] array. It's used when the vertex program + * references an attribute for which we don't have a VBO/array. + */ +static void +create_default_attribs_buffer(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + st->default_attrib_buffer = pipe->winsys->buffer_create( pipe->winsys, 32 ); +} + + +static void +destroy_default_attribs_buffer(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + pipe->winsys->buffer_unreference(pipe->winsys, &st->default_attrib_buffer); +} + + +static void +update_default_attribs_buffer(GLcontext *ctx) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct pipe_buffer_handle *buf = ctx->st->default_attrib_buffer; + const unsigned size = sizeof(ctx->Current.Attrib); + const void *data = ctx->Current.Attrib; + pipe->winsys->buffer_data(pipe->winsys, buf, size, data); +} + + /** * This function gets plugged into the VBO module and is called when * we have something to render. @@ -150,41 +183,66 @@ draw_vbo(GLcontext *ctx, { struct pipe_context *pipe = ctx->st->pipe; GLuint attr, i; + GLbitfield attrsNeeded; + const unsigned attr0_offset = (unsigned) arrays[0]->Ptr; st_validate_state(ctx->st); + update_default_attribs_buffer(ctx); + /* this must be after state validation */ + attrsNeeded = ctx->st->state.vs.inputs_read; /* tell pipe about the vertex array element/attributes */ for (attr = 0; attr < 16; attr++) { - struct gl_buffer_object *bufobj = arrays[attr]->BufferObj; struct pipe_vertex_buffer vbuffer; struct pipe_vertex_element velement; - if (bufobj && bufobj->Name) { - struct st_buffer_object *stobj = st_buffer_object(bufobj); - - assert(stobj->buffer); - - vbuffer.pitch = arrays[attr]->StrideB; /* in bytes */ - vbuffer.max_index = 0; - vbuffer.buffer = stobj->buffer; - vbuffer.buffer_offset = 0; - - /* Recall that for VBOs, the gl_client_array->Ptr field is - * really an offset from the start of the VBO, not a pointer. - */ - velement.src_offset = (unsigned) arrays[attr]->Ptr; - velement.vertex_buffer_index = attr; - velement.dst_offset = 0; - velement.src_format = pipe_vertex_format(arrays[attr]->Type, - arrays[attr]->Size); + vbuffer.buffer = NULL; + vbuffer.pitch = 0; + velement.src_offset = 0; + velement.vertex_buffer_index = 0; + velement.src_format = 0; + + if (attrsNeeded & (1 << attr)) { + struct gl_buffer_object *bufobj = arrays[attr]->BufferObj; + + if (bufobj && bufobj->Name) { + struct st_buffer_object *stobj = st_buffer_object(bufobj); + /* Recall that for VBOs, the gl_client_array->Ptr field is + * really an offset from the start of the VBO, not a pointer. + */ + unsigned offset = (unsigned) arrays[attr]->Ptr; + + assert(stobj->buffer); + + vbuffer.buffer = stobj->buffer; + vbuffer.buffer_offset = attr0_offset; /* in bytes */ + vbuffer.pitch = arrays[attr]->StrideB; /* in bytes */ + vbuffer.max_index = 0; /* need this? */ + + velement.src_offset = offset - attr0_offset; /* bytes */ + velement.vertex_buffer_index = attr; + velement.dst_offset = 0; /* need this? */ + velement.src_format = pipe_vertex_format(arrays[attr]->Type, + arrays[attr]->Size); + assert(velement.src_format); + } + else { + /* use the default attribute buffer */ + vbuffer.buffer = ctx->st->default_attrib_buffer; + vbuffer.buffer_offset = 0; + vbuffer.pitch = 0; /* must be zero! */ + vbuffer.max_index = 1; + + velement.src_offset = attr * 4 * sizeof(GLfloat); + velement.vertex_buffer_index = attr; + velement.dst_offset = 0; + velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + } } - else { - /* vertex attribute data is not an a real buffer! */ - /* XXX we'll want to handle that someday... */ - vbuffer.buffer = NULL; - } + if (attr == 0) + assert(vbuffer.buffer); pipe->set_vertex_buffer(pipe, attr, &vbuffer); pipe->set_vertex_element(pipe, attr, &velement); @@ -209,6 +267,8 @@ void st_init_draw( struct st_context *st ) #if USE_NEW_DRAW struct vbo_context *vbo = (struct vbo_context *) ctx->swtnl_im; + create_default_attribs_buffer(st); + assert(vbo); assert(vbo->draw_prims); vbo->draw_prims = draw_vbo; @@ -224,8 +284,7 @@ void st_init_draw( struct st_context *st ) void st_destroy_draw( struct st_context *st ) { - /* Nothing to do. - */ + destroy_default_attribs_buffer(st); } -- cgit v1.2.3 From 0453760a0bf139f113792d22e65ce6904f06f28b Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 18 Aug 2007 16:06:49 +0100 Subject: more work on vertex shaders, now using vertex shaders instead of TnL module by default --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 64 ++++++++++++++++++++++++++++--- src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- 2 files changed, 59 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 18ed3c9d5a..e7dec31921 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -87,6 +87,36 @@ compute_clipmask(float cx, float cy, float cz, float cw) } +/** + * Fetch a float[4] vertex attribute from memory, doing format/type + * conversion as needed. + * XXX this might be a temporary thing. + */ +static void +fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) +{ + /* defaults */ + attrib[1] = 0.0; + attrib[2] = 0.0; + attrib[3] = 1.0; + switch (format) { + case PIPE_FORMAT_R32G32B32A32_FLOAT: + attrib[3] = ((float *) ptr)[3]; + /* fall-through */ + case PIPE_FORMAT_R32G32B32_FLOAT: + attrib[2] = ((float *) ptr)[2]; + /* fall-through */ + case PIPE_FORMAT_R32G32_FLOAT: + attrib[1] = ((float *) ptr)[1]; + /* fall-through */ + case PIPE_FORMAT_R32_FLOAT: + attrib[0] = ((float *) ptr)[0]; + break; + default: + assert(0); + } +} + /** * Transform vertices with the current vertex program/shader @@ -129,21 +159,39 @@ run_vertex_program(struct draw_context *draw, machine.Inputs = ALIGN16_ASSIGN(inputs); machine.Outputs = ALIGN16_ASSIGN(outputs); + + if (0) + { + unsigned attr; + for (attr = 0; attr < 16; attr++) { + if (sp->vs.inputs_read & (1 << attr)) { + printf("attr %d: buf_off %d src_off %d pitch %d\n", + attr, + draw->vertex_buffer[attr].buffer_offset, + draw->vertex_element[attr].src_offset, + draw->vertex_buffer[attr].pitch); + } + } + } + /* load machine inputs */ for (j = 0; j < count; j++) { unsigned attr; for (attr = 0; attr < 16; attr++) { if (sp->vs.inputs_read & (1 << attr)) { - const float *p - = (const float *) ((const ubyte *) vbuffer - + draw->vertex_buffer[attr].buffer_offset - + draw->vertex_element[attr].src_offset - + elts[j] * draw->vertex_buffer[attr].pitch); + const void *src + = (const void *) ((const ubyte *) vbuffer + + draw->vertex_buffer[attr].buffer_offset + + draw->vertex_element[attr].src_offset + + elts[j] * draw->vertex_buffer[attr].pitch); + float p[4]; + + fetch_attrib4(src, draw->vertex_element[attr].src_format, p); machine.Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ machine.Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ - machine.Inputs[attr].xyzw[3].f[j] = 1.0; /*W*/ + machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ #if 0 if (attr == 0) { printf("Input vertex %d: %f %f %f\n", @@ -371,12 +419,16 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, struct draw_context *draw = sp->draw; struct pipe_buffer_handle *buf; + if (sp->dirty) + softpipe_update_derived( sp ); + softpipe_map_surfaces(sp); /* * Map vertex buffers */ buf = sp->vertex_buffer[0].buffer; + assert(buf); draw->mapped_vbuffer = pipe->winsys->buffer_map(pipe->winsys, buf, PIPE_BUFFER_FLAG_READ); diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 35f1a971ca..95b0cfee03 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -216,7 +216,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) if (attr_mask != softpipe->attr_mask) { softpipe->attr_mask = attr_mask; -#define USE_NEW_DRAW 0 +#define USE_NEW_DRAW 01 #if USE_NEW_DRAW draw_set_vertex_attributes2( softpipe->draw, slot_to_vf_attr, -- cgit v1.2.3 From 14327705fd53e984b74f8e9adb053df03fba7aff Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 19 Aug 2007 17:31:21 +0100 Subject: Fix some draw_arrays issues. We weren't mapping all the needed vertex array buffers. Move array state that was temporarily in draw_context to softpipe_context. Remove a bunch of dead code. --- src/mesa/pipe/draw/draw_context.c | 10 -- src/mesa/pipe/draw/draw_context.h | 4 - src/mesa/pipe/draw/draw_private.h | 5 - src/mesa/pipe/softpipe/sp_context.c | 4 - src/mesa/pipe/softpipe/sp_context.h | 5 + src/mesa/pipe/softpipe/sp_draw_arrays.c | 158 +++++++------------------------- 6 files changed, 38 insertions(+), 148 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 7e6a95127a..532c9c3682 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -201,13 +201,3 @@ void draw_set_viewport_state( struct draw_context *draw, * Full pipe will have vertex shader, vertex fetch of its own. */ } - - -void draw_set_vertex_array_info(struct draw_context *draw, - const struct pipe_vertex_buffer *buffers, - const struct pipe_vertex_element *elements) -{ - draw->vertex_buffer = buffers; - draw->vertex_element = elements; -} - diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index afa3a04165..e1b4848ab3 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -82,10 +82,6 @@ void draw_set_vertex_attributes2( struct draw_context *draw, const unsigned *attrs, unsigned nr_attrs ); -void draw_set_vertex_array_info(struct draw_context *draw, - const struct pipe_vertex_buffer *buffers, - const struct pipe_vertex_element *elements); - /* XXX temporary */ void draw_vb(struct draw_context *draw, struct vertex_buffer *VB ); diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 278950ec48..798fa5c3a6 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -158,8 +158,6 @@ struct draw_context /* pipe state that we need: */ struct pipe_setup_state setup; struct pipe_viewport_state viewport; - const struct pipe_vertex_buffer *vertex_buffer; /**< note: pointer */ - const struct pipe_vertex_element *vertex_element; /**< note: pointer */ /** need to know the pipe for vertex flushing/transformation: */ struct pipe_context *pipe; @@ -230,9 +228,6 @@ struct draw_context ubyte *verts; boolean in_vb; struct vertex_fetch *vf; - - /* Misc for sp_draw_arrays.c (temporary?) */ - void *mapped_vbuffer; }; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index e63ce208d4..ea2f0ec04f 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -322,10 +322,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, assert(softpipe->draw); draw_set_setup_stage(softpipe->draw, sp_draw_render_stage(softpipe)); - draw_set_vertex_array_info(softpipe->draw, - softpipe->vertex_buffer, - softpipe->vertex_element); - sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 8f184c644a..14ae9f2105 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -101,6 +101,11 @@ struct softpipe_context { */ enum interp_mode interp[PIPE_ATTRIB_MAX]; + /* + * Mapped vertex buffers + */ + ubyte *mapped_vbuffer[PIPE_ATTRIB_MAX]; + /* FS + setup derived state: */ diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index e7dec31921..a7fb39ca09 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -127,18 +127,17 @@ fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) * \param vOut array of pointers to four output vertices */ static void -run_vertex_program(struct draw_context *draw, - const void *vbuffer, unsigned elts[4], unsigned count, +run_vertex_program(struct softpipe_context *sp, + unsigned elts[4], unsigned count, struct vertex_header *vOut[]) { - struct softpipe_context *sp = softpipe_context(draw->pipe); struct tgsi_exec_machine machine; unsigned int j; ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; + const float *scale = sp->viewport.scale; + const float *trans = sp->viewport.translate; assert(count <= 4); @@ -167,9 +166,9 @@ run_vertex_program(struct draw_context *draw, if (sp->vs.inputs_read & (1 << attr)) { printf("attr %d: buf_off %d src_off %d pitch %d\n", attr, - draw->vertex_buffer[attr].buffer_offset, - draw->vertex_element[attr].src_offset, - draw->vertex_buffer[attr].pitch); + sp->vertex_buffer[attr].buffer_offset, + sp->vertex_element[attr].src_offset, + sp->vertex_buffer[attr].pitch); } } } @@ -180,13 +179,13 @@ run_vertex_program(struct draw_context *draw, for (attr = 0; attr < 16; attr++) { if (sp->vs.inputs_read & (1 << attr)) { const void *src - = (const void *) ((const ubyte *) vbuffer - + draw->vertex_buffer[attr].buffer_offset - + draw->vertex_element[attr].src_offset - + elts[j] * draw->vertex_buffer[attr].pitch); + = (const void *) ((const ubyte *) sp->mapped_vbuffer[attr] + + sp->vertex_buffer[attr].buffer_offset + + sp->vertex_element[attr].src_offset + + elts[j] * sp->vertex_buffer[attr].pitch); float p[4]; - fetch_attrib4(src, draw->vertex_element[attr].src_format, p); + fetch_attrib4(src, sp->vertex_element[attr].src_format, p); machine.Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ machine.Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ @@ -262,7 +261,7 @@ run_vertex_program(struct draw_context *draw, slot = 1; for (attr = 1; attr < VERT_RESULT_MAX; attr++) { if (sp->vs.outputs_written & (1 << attr)) { - assert(slot < draw->nr_attrs - 2); + assert(slot < sp->nr_attrs); vOut[j]->data[slot][0] = outputs[attr].xyzw[0].f[j]; vOut[j]->data[slot][1] = outputs[attr].xyzw[1].f[j]; vOut[j]->data[slot][2] = outputs[attr].xyzw[2].f[j]; @@ -281,115 +280,17 @@ run_vertex_program(struct draw_context *draw, } -/** - * Stand-in for actual vertex program execution - * XXX this will probably live in a new file, like "sp_vs.c" - * \param draw the drawing context - * \param vbuffer the mapped vertex buffer pointer - * \param elem which element of the vertex buffer to use as input - * \param vOut the output vertex - */ -#if 0 -static void -run_vertex_program(struct draw_context *draw, - const void *vbuffer, unsigned elem, - struct vertex_header *vOut) -{ - const float *vIn, *cIn; - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - const void *mapped = vbuffer; - - /* XXX temporary hack: */ - GET_CURRENT_CONTEXT(ctx); - const float *m = ctx->_ModelProjectMatrix.m; - - vIn = (const float *) ((const ubyte *) mapped - + draw->vertex_buffer[0].buffer_offset - + draw->vertex_element[0].src_offset - + elem * draw->vertex_buffer[0].pitch); - - cIn = (const float *) ((const ubyte *) mapped - + draw->vertex_buffer[3].buffer_offset - + draw->vertex_element[3].src_offset - + elem * draw->vertex_buffer[3].pitch); - - { - float x = vIn[0]; - float y = vIn[1]; - float z = vIn[2]; - float w = 1.0; - - vOut->clipmask = 0x0; - vOut->edgeflag = 0; - /* MVP */ - vOut->clip[0] = m[0] * x + m[4] * y + m[ 8] * z + m[12] * w; - vOut->clip[1] = m[1] * x + m[5] * y + m[ 9] * z + m[13] * w; - vOut->clip[2] = m[2] * x + m[6] * y + m[10] * z + m[14] * w; - vOut->clip[3] = m[3] * x + m[7] * y + m[11] * z + m[15] * w; - - /* divide by w */ - x = vOut->clip[0] / vOut->clip[3]; - y = vOut->clip[1] / vOut->clip[3]; - z = vOut->clip[2] / vOut->clip[3]; - w = 1.0 / vOut->clip[3]; - - /* Viewport */ - vOut->data[0][0] = scale[0] * x + trans[0]; - vOut->data[0][1] = scale[1] * y + trans[1]; - vOut->data[0][2] = scale[2] * z + trans[2]; - vOut->data[0][3] = w; - - /* color */ - vOut->data[1][0] = cIn[0]; - vOut->data[1][1] = cIn[1]; - vOut->data[1][2] = cIn[2]; - vOut->data[1][3] = 1.0; - } -} -#endif - - /** * Called by the draw module when the vertx cache needs to be flushed. * This involves running the vertex shader. */ static void vs_flush( struct draw_context *draw ) { + struct softpipe_context *sp = (struct softpipe_context *) draw->pipe; unsigned i, j; - /* We're not really running a vertex shader yet, so flushing the vs - * queue is just a matter of building the vertices and returning. - */ - /* Actually, I'm cheating even more and pre-building them still - * with the mesa/vf module. So it's very easy... - */ -#if 0 - for (i = 0; i < draw->vs.queue_nr; i++) { -#else - for (i = 0; i < draw->vs.queue_nr; i+=4) { -#endif - /* Would do the following steps here: - * - * 1) Loop over vertex element descriptors, fetch data from each - * to build the pre-tnl vertex. This might require a new struct - * to represent the pre-tnl vertex. - * - * 2) Bundle groups of upto 4 pre-tnl vertices together and pass - * to vertex shader. - * - * 3) Do any necessary unswizzling, make sure vertex headers are - * correctly populated, store resulting post-transformed - * vertices in vcache. - * - * In this version, just do the last step: - */ -#if 0 - const unsigned elt = draw->vs.queue[i].elt; - struct vertex_header *dest = draw->vs.queue[i].dest; - - run_vertex_program(draw, draw->mapped_vbuffer, elt, dest); -#else + /* run vertex shader on vertex cache entries, four per invokation */ + for (i = 0; i < draw->vs.queue_nr; i += 4) { struct vertex_header *dests[4]; unsigned elts[4]; int n; @@ -403,9 +304,9 @@ static void vs_flush( struct draw_context *draw ) assert(n > 0); assert(n <= 4); - run_vertex_program(draw, draw->mapped_vbuffer, elts, n, dests); -#endif + run_vertex_program(sp, elts, n, dests); } + draw->vs.queue_nr = 0; } @@ -417,7 +318,7 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, { struct softpipe_context *sp = softpipe_context(pipe); struct draw_context *draw = sp->draw; - struct pipe_buffer_handle *buf; + unsigned int i; if (sp->dirty) softpipe_update_derived( sp ); @@ -427,11 +328,14 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, /* * Map vertex buffers */ - buf = sp->vertex_buffer[0].buffer; - assert(buf); - draw->mapped_vbuffer - = pipe->winsys->buffer_map(pipe->winsys, buf, PIPE_BUFFER_FLAG_READ); - + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (sp->vertex_buffer[i].buffer) { + sp->mapped_vbuffer[i] + = pipe->winsys->buffer_map(pipe->winsys, + sp->vertex_buffer[i].buffer, + PIPE_BUFFER_FLAG_READ); + } + } /* tell drawing pipeline we're beginning drawing */ draw->pipeline.first->begin( draw->pipeline.first ); @@ -454,9 +358,13 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, draw->pipeline.first->end( draw->pipeline.first ); /* - * unmap vertex buffer + * unmap vertex buffers */ - pipe->winsys->buffer_unmap(pipe->winsys, buf); + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (sp->vertex_buffer[i].buffer) { + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); + } + } softpipe_unmap_surfaces(sp); } -- cgit v1.2.3 From f252974121febc6a1a59793d932b32b798f90fc6 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 11:05:35 -0600 Subject: fix VBO clean-up in vbo_exec_vtx_destroy() --- src/mesa/vbo/vbo_exec_api.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index 24e2eba472..b7f4d8a307 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -673,7 +673,8 @@ void vbo_exec_vtx_init( struct vbo_exec_context *exec ) GLuint i; /* Allocate a buffer object. Will just reuse this object - * continuously. + * continuously, unless vbo_use_buffer_objects() is called to enable + * use of real VBOs. */ exec->vtx.bufferobj = ctx->Array.NullBufferObj; exec->vtx.buffer_map = ALIGN_MALLOC(VBO_VERT_BUFFER_SIZE * sizeof(GLfloat), 64); @@ -702,9 +703,17 @@ void vbo_exec_vtx_init( struct vbo_exec_context *exec ) void vbo_exec_vtx_destroy( struct vbo_exec_context *exec ) { - if (exec->vtx.buffer_map) { - ALIGN_FREE(exec->vtx.buffer_map); - exec->vtx.buffer_map = NULL; + GLcontext *ctx = exec->ctx; + if (exec->vtx.bufferobj->Name) { + ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, exec->vtx.bufferobj); + ctx->Driver.DeleteBuffer(ctx, exec->vtx.bufferobj); + exec->vtx.bufferobj = NULL; + } + else { + if (exec->vtx.buffer_map) { + ALIGN_FREE(exec->vtx.buffer_map); + exec->vtx.buffer_map = NULL; + } } } -- cgit v1.2.3 From 904163e4e94dbf3ea9a24738077e9916e97191f8 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 11:41:23 -0600 Subject: fix cache overflow bug in get_vertex() --- src/mesa/pipe/draw/draw_prim.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 95e71efd6d..fb8bc0f36d 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -144,8 +144,8 @@ static struct vertex_header *get_vertex( struct draw_context *draw, /* If slot is in use, use the overflow area: */ - if (draw->vcache.referenced & (1<vcache.overflow++; + if (draw->vcache.referenced & (1 << slot)) + slot = VCACHE_SIZE + draw->vcache.overflow++; draw->vcache.idx[slot] = i; @@ -158,7 +158,8 @@ static struct vertex_header *get_vertex( struct draw_context *draw, /* Mark slot as in-use: */ - draw->vcache.referenced |= (1<vcache.referenced |= (1 << slot); return draw->vcache.vertex[slot]; } -- cgit v1.2.3 From 98d30cb2d4e17c288ef5f0fd41c09b372832d783 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 11:49:16 -0600 Subject: call draw_trim() --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index a7fb39ca09..5371ef58e6 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -318,8 +318,9 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, { struct softpipe_context *sp = softpipe_context(pipe); struct draw_context *draw = sp->draw; - unsigned int i; + unsigned length, first, incr, i; + printf("Draw arrays start %u count %u\n", start, count); if (sp->dirty) softpipe_update_derived( sp ); @@ -346,10 +347,13 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, draw_invalidate_vcache( draw ); draw_set_element_buffer(draw, 0, NULL); /* no index/element buffer */ - draw_set_prim( draw, mode ); - /* XXX draw_prim_info() and TRIM here */ - draw_prim(draw, start, count); + draw_prim_info( mode, &first, &incr ); + length = draw_trim( count, first, incr ); + if (length) { + draw_set_prim( draw, mode ); + draw_prim(draw, start, count); + } /* draw any left-over buffered prims */ draw_flush(draw); -- cgit v1.2.3 From 7d14c3e538f9582e2f5fb0ff827af2ecce9b0f40 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 11:53:36 -0600 Subject: remove debug printf --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 5371ef58e6..ce5d4e0cbc 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -320,7 +320,6 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, struct draw_context *draw = sp->draw; unsigned length, first, incr, i; - printf("Draw arrays start %u count %u\n", start, count); if (sp->dirty) softpipe_update_derived( sp ); -- cgit v1.2.3 From 730df7662f57a46dee892733afc9a55f37d2ab03 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 12:50:34 -0600 Subject: don't map element buffer in _mesa_validate_DrawElements() unless necessary --- src/mesa/main/api_validate.c | 31 +++++++++++++++++++------------ 1 file changed, 19 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c index 841c6a5302..64ab324af2 100644 --- a/src/mesa/main/api_validate.c +++ b/src/mesa/main/api_validate.c @@ -69,8 +69,9 @@ _mesa_validate_DrawElements(GLcontext *ctx, GLuint indexBytes; /* use indices in the buffer object */ - if (!ctx->Array.ElementArrayBufferObj->Data) { - _mesa_warning(ctx, "DrawElements with empty vertex elements buffer!"); + if (!ctx->Array.ElementArrayBufferObj->Size) { + _mesa_warning(ctx, + "glDrawElements called with empty array elements buffer"); return GL_FALSE; } @@ -86,19 +87,10 @@ _mesa_validate_DrawElements(GLcontext *ctx, indexBytes = count * sizeof(GLushort); } - if ((GLubyte *) indices + indexBytes > - ctx->Array.ElementArrayBufferObj->Data + - ctx->Array.ElementArrayBufferObj->Size) { + if (indexBytes > ctx->Array.ElementArrayBufferObj->Size) { _mesa_warning(ctx, "glDrawElements index out of buffer bounds"); return GL_FALSE; } - - /* Actual address is the sum of pointers. Indices may be used below. */ - if (ctx->Const.CheckArrayBounds) { - indices = (const GLvoid *) - ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data, - (const GLubyte *) indices); - } } else { /* not using a VBO */ @@ -108,8 +100,18 @@ _mesa_validate_DrawElements(GLcontext *ctx, if (ctx->Const.CheckArrayBounds) { /* find max array index */ + const GLubyte *map; GLuint max = 0; GLint i; + + map = ctx->Driver.MapBuffer(ctx, + GL_ELEMENT_ARRAY_BUFFER_ARB, + GL_READ_ONLY, + ctx->Array.ElementArrayBufferObj); + + /* Actual address is the sum of pointers */ + indices = (const GLvoid *) ADD_POINTERS(map, (const GLubyte *) indices); + if (type == GL_UNSIGNED_INT) { for (i = 0; i < count; i++) if (((GLuint *) indices)[i] > max) @@ -126,6 +128,11 @@ _mesa_validate_DrawElements(GLcontext *ctx, if (((GLubyte *) indices)[i] > max) max = ((GLubyte *) indices)[i]; } + + ctx->Driver.UnmapBuffer(ctx, + GL_ELEMENT_ARRAY_BUFFER_ARB, + ctx->Array.ElementArrayBufferObj); + if (max >= ctx->Array._MaxElement) { /* the max element is out of bounds of one or more enabled arrays */ return GL_FALSE; -- cgit v1.2.3 From 1dd55a77c87bd4e057d689163efd070a2dfe3454 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 12:52:00 -0600 Subject: implement draw_elements() --- src/mesa/pipe/p_context.h | 5 ++++ src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_draw_arrays.c | 48 +++++++++++++++++++++++++++------ src/mesa/pipe/softpipe/sp_state.h | 5 ++++ 4 files changed, 51 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 320b4877c3..a3664a3b80 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -82,6 +82,11 @@ struct pipe_context { void (*draw_arrays)( struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count); + void (*draw_elements)( struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); + /** Clear a surface to given value (no scissor; clear whole surface) */ void (*clear)(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index ea2f0ec04f..e7694bbe86 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -288,6 +288,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.draw_vertices = softpipe_draw_vertices; softpipe->pipe.draw_arrays = softpipe_draw_arrays; + softpipe->pipe.draw_elements = softpipe_draw_elements; softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index ce5d4e0cbc..6213286f47 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -316,9 +316,32 @@ void softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { + softpipe_draw_elements(pipe, NULL, 0, mode, start, count); +} + + + +/** + * XXX should the element buffer be specified/bound with a separate function? + */ +void +softpipe_draw_elements(struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + struct softpipe_context *sp = softpipe_context(pipe); struct draw_context *draw = sp->draw; unsigned length, first, incr, i; + void *mapped_indexes = NULL; + + /* first, check that the primitive is not malformed */ + draw_prim_info( mode, &first, &incr ); + length = draw_trim( count, first, incr ); + if (!length) + return; + if (sp->dirty) softpipe_update_derived( sp ); @@ -336,6 +359,16 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, PIPE_BUFFER_FLAG_READ); } } + /* Map index buffer, if present */ + if (indexBuffer) { + mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, + indexBuffer, + PIPE_BUFFER_FLAG_READ); + draw_set_element_buffer(draw, indexSize, mapped_indexes); + } + else { + draw_set_element_buffer(draw, 0, NULL); /* no index/element buffer */ + } /* tell drawing pipeline we're beginning drawing */ draw->pipeline.first->begin( draw->pipeline.first ); @@ -345,14 +378,10 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, draw_invalidate_vcache( draw ); - draw_set_element_buffer(draw, 0, NULL); /* no index/element buffer */ + draw_set_prim( draw, mode ); - draw_prim_info( mode, &first, &incr ); - length = draw_trim( count, first, incr ); - if (length) { - draw_set_prim( draw, mode ); - draw_prim(draw, start, count); - } + /* drawing done here: */ + draw_prim(draw, start, count); /* draw any left-over buffered prims */ draw_flush(draw); @@ -361,13 +390,16 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, draw->pipeline.first->end( draw->pipeline.first ); /* - * unmap vertex buffers + * unmap vertex/index buffers */ for (i = 0; i < PIPE_ATTRIB_MAX; i++) { if (sp->vertex_buffer[i].buffer) { pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); } } + if (indexBuffer) { + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + } softpipe_unmap_surfaces(sp); } diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index f25a20a5e3..49fef0ddce 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -98,6 +98,11 @@ void softpipe_update_derived( struct softpipe_context *softpipe ); void softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count); +void softpipe_draw_elements(struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); + void softpipe_map_surfaces(struct softpipe_context *sp); -- cgit v1.2.3 From a83b72a67263faf21bf16ff879c9718660684aed Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 12:52:33 -0600 Subject: implement DrawElements, still some rough spots --- src/mesa/state_tracker/st_draw.c | 37 +++++++++++++++++++++++++++++++++++-- 1 file changed, 35 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index db24b24fa7..e82f7cff2b 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -249,8 +249,41 @@ draw_vbo(GLcontext *ctx, } /* do actual drawing */ - for (i = 0; i < nr_prims; i++) { - pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count); + if (ib) { + /* indexed primitive */ + struct gl_buffer_object *bufobj = ib->obj; + struct pipe_buffer_handle *bh = NULL; + unsigned indexSize; + + if (bufobj && bufobj->Name) { + /* elements/indexes are in a real VBO */ + struct st_buffer_object *stobj = st_buffer_object(bufobj); + bh = stobj->buffer; + switch (ib->type) { + case GL_UNSIGNED_INT: + indexSize = 4; + break; + case GL_UNSIGNED_SHORT: + indexSize = 2; + break; + default: + assert(0); + } + } + else { + assert(0); + } + + for (i = 0; i < nr_prims; i++) { + pipe->draw_elements(pipe, bh, indexSize, + prims[i].mode, prims[i].start, prims[i].count); + } + } + else { + /* non-indexed */ + for (i = 0; i < nr_prims; i++) { + pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count); + } } } -- cgit v1.2.3 From 0a262998ef2813d19e9fee01d3e5808416e9cb04 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 15:11:11 -0600 Subject: Move guts of vertex array drawing into the 'draw' module. --- src/mesa/pipe/draw/draw_arrays.c | 386 ++++++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_context.c | 28 +++ src/mesa/pipe/draw/draw_context.h | 36 ++- src/mesa/pipe/draw/draw_prim.c | 7 +- src/mesa/pipe/draw/draw_prim.h | 7 - src/mesa/pipe/draw/draw_private.h | 6 +- src/mesa/pipe/draw/draw_vb.c | 9 +- src/mesa/pipe/softpipe/sp_draw_arrays.c | 349 +-------------------------- src/mesa/pipe/softpipe/sp_state_derived.c | 15 +- src/mesa/pipe/softpipe/sp_state_fs.c | 2 + src/mesa/pipe/softpipe/sp_state_vertex.c | 4 + src/mesa/sources | 1 + 12 files changed, 478 insertions(+), 372 deletions(-) create mode 100644 src/mesa/pipe/draw/draw_arrays.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c new file mode 100644 index 0000000000..59098fe242 --- /dev/null +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -0,0 +1,386 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" + +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_prim.h" + +#include "pipe/tgsi/core/tgsi_exec.h" +#include "pipe/tgsi/core/tgsi_build.h" +#include "pipe/tgsi/core/tgsi_util.h" + + +#if defined __GNUC__ +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE] __attribute__(( aligned( 16 ) )) +#define ALIGN16_ASSIGN(P) P +#else +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE + 1] +#define ALIGN16_ASSIGN(P) align16(P) +#endif + + + +static INLINE unsigned +compute_clipmask(float cx, float cy, float cz, float cw) +{ + unsigned mask; +#if defined(macintosh) || defined(__powerpc__) + /* on powerpc cliptest is 17% faster in this way. */ + mask = (((cw < cx) << CLIP_RIGHT_SHIFT)); + mask |= (((cw < -cx) << CLIP_LEFT_SHIFT)); + mask |= (((cw < cy) << CLIP_TOP_SHIFT)); + mask |= (((cw < -cy) << CLIP_BOTTOM_SHIFT)); + mask |= (((cw < cz) << CLIP_FAR_SHIFT)); + mask |= (((cw < -cz) << CLIP_NEAR_SHIFT)); +#else /* !defined(macintosh)) */ + mask = 0x0; + if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT; + if ( cx + cw < 0) mask |= CLIP_LEFT_BIT; + if (-cy + cw < 0) mask |= CLIP_TOP_BIT; + if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT; + if (-cz + cw < 0) mask |= CLIP_FAR_BIT; + if ( cz + cw < 0) mask |= CLIP_NEAR_BIT; +#endif /* defined(macintosh) */ + return mask; +} + + +/** + * Fetch a float[4] vertex attribute from memory, doing format/type + * conversion as needed. + * XXX this might be a temporary thing. + */ +static void +fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) +{ + /* defaults */ + attrib[1] = 0.0; + attrib[2] = 0.0; + attrib[3] = 1.0; + switch (format) { + case PIPE_FORMAT_R32G32B32A32_FLOAT: + attrib[3] = ((float *) ptr)[3]; + /* fall-through */ + case PIPE_FORMAT_R32G32B32_FLOAT: + attrib[2] = ((float *) ptr)[2]; + /* fall-through */ + case PIPE_FORMAT_R32G32_FLOAT: + attrib[1] = ((float *) ptr)[1]; + /* fall-through */ + case PIPE_FORMAT_R32_FLOAT: + attrib[0] = ((float *) ptr)[0]; + break; + default: + assert(0); + } +} + + +/** + * Transform vertices with the current vertex program/shader + * Up to four vertices can be shaded at a time. + * \param vbuffer the input vertex data + * \param elts indexes of four input vertices + * \param count number of vertices to shade [1..4] + * \param vOut array of pointers to four output vertices + */ +static void +run_vertex_program(struct draw_context *draw, + unsigned elts[4], unsigned count, + struct vertex_header *vOut[]) +{ + struct tgsi_exec_machine machine; + unsigned int j; + + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); + +#ifdef DEBUG + memset( &machine, 0, sizeof( machine ) ); +#endif + + /* init machine state */ + tgsi_exec_machine_init(&machine, + draw->vertex_shader.tokens, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/ ); + + /* Consts does not require 16 byte alignment. */ + machine.Consts = draw->vertex_shader.constants->constant; + + machine.Inputs = ALIGN16_ASSIGN(inputs); + machine.Outputs = ALIGN16_ASSIGN(outputs); + + + if (0) + { + unsigned attr; + for (attr = 0; attr < 16; attr++) { + if (draw->vertex_shader.inputs_read & (1 << attr)) { + printf("attr %d: buf_off %d src_off %d pitch %d\n", + attr, + draw->vertex_buffer[attr].buffer_offset, + draw->vertex_element[attr].src_offset, + draw->vertex_buffer[attr].pitch); + } + } + } + + /* load machine inputs */ + for (j = 0; j < count; j++) { + unsigned attr; + for (attr = 0; attr < 16; attr++) { + if (draw->vertex_shader.inputs_read & (1 << attr)) { + const void *src + = (const void *) ((const ubyte *) draw->mapped_vbuffer[attr] + + draw->vertex_buffer[attr].buffer_offset + + draw->vertex_element[attr].src_offset + + elts[j] * draw->vertex_buffer[attr].pitch); + float p[4]; + + fetch_attrib4(src, draw->vertex_element[attr].src_format, p); + + machine.Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ + machine.Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ + machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ + machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ +#if 0 + if (attr == 0) { + printf("Input vertex %d: %f %f %f\n", + j, p[0], p[1], p[2]); + } +#endif + } + } + } + +#if 0 + printf("Consts:\n"); + for (i = 0; i < 4; i++) { + printf(" %d: %f %f %f %f\n", i, + machine.Consts[i][0], + machine.Consts[i][1], + machine.Consts[i][2], + machine.Consts[i][3]); + } +#endif + + /* run shader */ + tgsi_exec_machine_run( &machine ); + +#if 0 + printf("VS result: %f %f %f %f\n", + outputs[0].xyzw[0].f[0], + outputs[0].xyzw[1].f[0], + outputs[0].xyzw[2].f[0], + outputs[0].xyzw[3].f[0]); +#endif + + /* store machine results */ + assert(draw->vertex_shader.outputs_written & (1 << VERT_RESULT_HPOS)); + for (j = 0; j < count; j++) { + unsigned attr, slot; + float x, y, z, w; + + /* Handle attr[0] (position) specially: */ + x = vOut[j]->clip[0] = outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(x, y, z, w); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0 / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; +#if 0 + printf("wincoord: %f %f %f\n", + vOut[j]->data[0][0], + vOut[j]->data[0][1], + vOut[j]->data[0][2]); +#endif + + /* remaining attributes: */ + /* pack into sequential post-transform attrib slots */ + slot = 1; + for (attr = 1; attr < VERT_RESULT_MAX; attr++) { + if (draw->vertex_shader.outputs_written & (1 << attr)) { + assert(slot < draw->nr_attrs); + vOut[j]->data[slot][0] = outputs[attr].xyzw[0].f[j]; + vOut[j]->data[slot][1] = outputs[attr].xyzw[1].f[j]; + vOut[j]->data[slot][2] = outputs[attr].xyzw[2].f[j]; + vOut[j]->data[slot][3] = outputs[attr].xyzw[3].f[j]; + slot++; + } + } + } + +#if 0 + memcpy( + quad->outputs.color, + &machine.Outputs[1].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); +#endif +} + + +/** + * Called by the draw module when the vertx cache needs to be flushed. + * This involves running the vertex shader. + */ +static void vs_flush( struct draw_context *draw ) +{ + unsigned i, j; + + /* run vertex shader on vertex cache entries, four per invokation */ + for (i = 0; i < draw->vs.queue_nr; i += 4) { + struct vertex_header *dests[4]; + unsigned elts[4]; + int n; + + for (j = 0; j < 4; j++) { + elts[j] = draw->vs.queue[i + j].elt; + dests[j] = draw->vs.queue[i + j].dest; + } + + n = MIN2(4, draw->vs.queue_nr - i); + assert(n > 0); + assert(n <= 4); + + run_vertex_program(draw, elts, n, dests); + } + + draw->vs.queue_nr = 0; +} + + +void draw_set_mapped_vertex_buffer(struct draw_context *draw, + unsigned attr, const void *buffer) +{ + draw->mapped_vbuffer[attr] = buffer; +} + + +/** + * Draw vertex arrays + * This is the main entrypoint into the drawing module. + * \param prim one of PIPE_PRIM_x + * \param start index of first vertex to draw + * \param count number of vertices to draw + */ +void +draw_arrays(struct draw_context *draw, unsigned prim, + unsigned start, unsigned count) +{ + /* tell drawing pipeline we're beginning drawing */ + draw->pipeline.first->begin( draw->pipeline.first ); + + draw->vs_flush = vs_flush; + + draw_invalidate_vcache( draw ); + + draw_set_prim( draw, prim ); + + /* drawing done here: */ + draw_prim(draw, start, count); + + /* draw any left-over buffered prims */ + draw_flush(draw); + + /* tell drawing pipeline we're done drawing */ + draw->pipeline.first->end( draw->pipeline.first ); +} + + +/* XXX move this into draw_context.c? */ + +#define EMIT_ATTR( VF_ATTR, STYLE, SIZE ) \ +do { \ + if (draw->nr_attrs >= 2) \ + draw->vf_attr_to_slot[VF_ATTR] = draw->nr_attrs - 2; \ + draw->attrs[draw->nr_attrs].attrib = VF_ATTR; \ + draw->attrs[draw->nr_attrs].format = STYLE; \ + draw->nr_attrs++; \ + draw->vertex_size += SIZE; \ +} while (0) + + +/** + * XXX very similar to same func in draw_vb.c (which will go away) + */ +void +draw_set_vertex_attributes( struct draw_context *draw, + const unsigned *slot_to_vf_attr, + unsigned nr_attrs ) +{ + unsigned i; + + memset(draw->vf_attr_to_slot, 0, sizeof(draw->vf_attr_to_slot)); + draw->nr_attrs = 0; + draw->vertex_size = 0; + + /* + * First three attribs are always the same: header, clip pos, winpos + */ + EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F, 1); + EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F, 4); + + assert(slot_to_vf_attr[0] == VF_ATTRIB_POS); + EMIT_ATTR(slot_to_vf_attr[0], EMIT_4F_VIEWPORT, 4); + + /* + * Remaining attribs (color, texcoords, etc) + */ + for (i = 1; i < nr_attrs; i++) + EMIT_ATTR(slot_to_vf_attr[i], EMIT_4F, 4); + + draw->vertex_size *= 4; /* floats to bytes */ +} diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index a808fb7777..cc00576c53 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -201,3 +201,31 @@ void draw_set_viewport_state( struct draw_context *draw, * Full pipe will have vertex shader, vertex fetch of its own. */ } + + +void +draw_set_vertex_buffer(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_buffer *buffer) +{ + assert(attr < PIPE_ATTRIB_MAX); + draw->vertex_buffer[attr] = *buffer; +} + + +void +draw_set_vertex_element(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_element *element) +{ + assert(attr < PIPE_ATTRIB_MAX); + draw->vertex_element[attr] = *element; +} + + +void +draw_set_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader) +{ + draw->vertex_shader = *shader; +} diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 74fdd46262..2fce1322c5 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -93,14 +93,36 @@ void draw_set_vertex_attributes( struct draw_context *draw, const unsigned *attrs, unsigned nr_attrs ); -/* XXX temporary */ -void draw_set_vertex_attributes2( struct draw_context *draw, - const unsigned *attrs, - unsigned nr_attrs ); +unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); + +unsigned draw_trim( unsigned count, unsigned first, unsigned incr ); + +void draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ); + +void draw_set_mapped_vertex_buffer(struct draw_context *draw, + unsigned attr, const void *buffer); + + +void +draw_set_vertex_buffer(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_buffer *buffer); + +void +draw_set_vertex_element(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_element *element); + +void +draw_set_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader); + + +void +draw_arrays(struct draw_context *draw, unsigned prim, + unsigned start, unsigned count); -/* XXX temporary */ -void draw_vb(struct draw_context *draw, - struct vertex_buffer *VB ); void draw_vertices(struct draw_context *draw, unsigned mode, diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index fb8bc0f36d..fbd0672875 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -434,8 +434,8 @@ draw_set_prim( struct draw_context *draw, unsigned prim ) * \param elements the element buffer ptr */ void -draw_set_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ) +draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ) { /* choose the get_vertex() function to use */ switch (eltSize) { @@ -456,10 +456,9 @@ draw_set_element_buffer( struct draw_context *draw, } draw->elts = elements; draw->eltSize = eltSize; - - } + unsigned draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) { diff --git a/src/mesa/pipe/draw/draw_prim.h b/src/mesa/pipe/draw/draw_prim.h index 3224989bef..4c55b02978 100644 --- a/src/mesa/pipe/draw/draw_prim.h +++ b/src/mesa/pipe/draw/draw_prim.h @@ -8,16 +8,9 @@ void draw_invalidate_vcache( struct draw_context *draw ); void draw_set_prim( struct draw_context *draw, unsigned prim ); -void draw_set_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ); - void draw_prim( struct draw_context *draw, unsigned start, unsigned count ); void draw_flush( struct draw_context *draw ); -unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); - -unsigned draw_trim( unsigned count, unsigned first, unsigned incr ); - #endif /* DRAW_PRIM_H */ diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 798fa5c3a6..5c3efb80e9 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -158,9 +158,11 @@ struct draw_context /* pipe state that we need: */ struct pipe_setup_state setup; struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + struct pipe_shader_state vertex_shader; - /** need to know the pipe for vertex flushing/transformation: */ - struct pipe_context *pipe; + const void *mapped_vbuffer[PIPE_ATTRIB_MAX]; /* Clip derived state: */ diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index 18e48c0a68..0eefb0b250 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -221,9 +221,9 @@ void draw_vb(struct draw_context *draw, vf_emit_vertices( draw->vf, VB->Count, draw->verts ); if (VB->Elts) - draw_set_element_buffer(draw, sizeof(unsigned), VB->Elts); + draw_set_mapped_element_buffer(draw, sizeof(unsigned), VB->Elts); else - draw_set_element_buffer(draw, 0, NULL); + draw_set_mapped_element_buffer(draw, 0, NULL); for (i = 0; i < VB->PrimitiveCount; i++) { const GLenum mode = VB->Primitive[i].mode; @@ -281,7 +281,7 @@ draw_vertices(struct draw_context *draw, /* no element/index buffer */ - draw_set_element_buffer(draw, 0, NULL); + draw_set_mapped_element_buffer(draw, 0, NULL); /*draw_prim_info(mode, &first, &incr);*/ draw_allocate_vertices( draw, numVerts ); @@ -325,7 +325,7 @@ draw_vertices(struct draw_context *draw, draw->in_vb = 0; } - +#if 000 /** * Accumulate another attribute's info. @@ -383,3 +383,4 @@ void draw_set_vertex_attributes( struct draw_context *draw, draw->nr_attrs, 0 ); #endif } +#endif diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 6213286f47..2df07eb7d6 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -31,284 +31,15 @@ */ -/** TEMP */ -#include "main/context.h" -#include "main/macros.h" - #include "pipe/p_defines.h" #include "pipe/p_context.h" #include "pipe/p_winsys.h" - +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_state.h" -#include "pipe/draw/draw_private.h" #include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_prim.h" - -#include "pipe/tgsi/core/tgsi_exec.h" -#include "pipe/tgsi/core/tgsi_build.h" -#include "pipe/tgsi/core/tgsi_util.h" - - -#if defined __GNUC__ -#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE] __attribute__(( aligned( 16 ) )) -#define ALIGN16_ASSIGN(P) P -#else -#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE + 1] -#define ALIGN16_ASSIGN(P) align16(P) -#endif - - - -static INLINE unsigned -compute_clipmask(float cx, float cy, float cz, float cw) -{ - unsigned mask; -#if defined(macintosh) || defined(__powerpc__) - /* on powerpc cliptest is 17% faster in this way. */ - mask = (((cw < cx) << CLIP_RIGHT_SHIFT)); - mask |= (((cw < -cx) << CLIP_LEFT_SHIFT)); - mask |= (((cw < cy) << CLIP_TOP_SHIFT)); - mask |= (((cw < -cy) << CLIP_BOTTOM_SHIFT)); - mask |= (((cw < cz) << CLIP_FAR_SHIFT)); - mask |= (((cw < -cz) << CLIP_NEAR_SHIFT)); -#else /* !defined(macintosh)) */ - mask = 0x0; - if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT; - if ( cx + cw < 0) mask |= CLIP_LEFT_BIT; - if (-cy + cw < 0) mask |= CLIP_TOP_BIT; - if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT; - if (-cz + cw < 0) mask |= CLIP_FAR_BIT; - if ( cz + cw < 0) mask |= CLIP_NEAR_BIT; -#endif /* defined(macintosh) */ - return mask; -} - - -/** - * Fetch a float[4] vertex attribute from memory, doing format/type - * conversion as needed. - * XXX this might be a temporary thing. - */ -static void -fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) -{ - /* defaults */ - attrib[1] = 0.0; - attrib[2] = 0.0; - attrib[3] = 1.0; - switch (format) { - case PIPE_FORMAT_R32G32B32A32_FLOAT: - attrib[3] = ((float *) ptr)[3]; - /* fall-through */ - case PIPE_FORMAT_R32G32B32_FLOAT: - attrib[2] = ((float *) ptr)[2]; - /* fall-through */ - case PIPE_FORMAT_R32G32_FLOAT: - attrib[1] = ((float *) ptr)[1]; - /* fall-through */ - case PIPE_FORMAT_R32_FLOAT: - attrib[0] = ((float *) ptr)[0]; - break; - default: - assert(0); - } -} - - -/** - * Transform vertices with the current vertex program/shader - * Up to four vertices can be shaded at a time. - * \param vbuffer the input vertex data - * \param elts indexes of four input vertices - * \param count number of vertices to shade [1..4] - * \param vOut array of pointers to four output vertices - */ -static void -run_vertex_program(struct softpipe_context *sp, - unsigned elts[4], unsigned count, - struct vertex_header *vOut[]) -{ - struct tgsi_exec_machine machine; - unsigned int j; - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); - const float *scale = sp->viewport.scale; - const float *trans = sp->viewport.translate; - - assert(count <= 4); - -#ifdef DEBUG - memset( &machine, 0, sizeof( machine ) ); -#endif - - /* init machine state */ - tgsi_exec_machine_init( - &machine, - sp->vs.tokens, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/ ); - - /* Consts does not require 16 byte alignment. */ - machine.Consts = sp->vs.constants->constant; - - machine.Inputs = ALIGN16_ASSIGN(inputs); - machine.Outputs = ALIGN16_ASSIGN(outputs); - - - if (0) - { - unsigned attr; - for (attr = 0; attr < 16; attr++) { - if (sp->vs.inputs_read & (1 << attr)) { - printf("attr %d: buf_off %d src_off %d pitch %d\n", - attr, - sp->vertex_buffer[attr].buffer_offset, - sp->vertex_element[attr].src_offset, - sp->vertex_buffer[attr].pitch); - } - } - } - - /* load machine inputs */ - for (j = 0; j < count; j++) { - unsigned attr; - for (attr = 0; attr < 16; attr++) { - if (sp->vs.inputs_read & (1 << attr)) { - const void *src - = (const void *) ((const ubyte *) sp->mapped_vbuffer[attr] - + sp->vertex_buffer[attr].buffer_offset - + sp->vertex_element[attr].src_offset - + elts[j] * sp->vertex_buffer[attr].pitch); - float p[4]; - - fetch_attrib4(src, sp->vertex_element[attr].src_format, p); - - machine.Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ - machine.Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ - machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ - machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ -#if 0 - if (attr == 0) { - printf("Input vertex %d: %f %f %f\n", - j, p[0], p[1], p[2]); - } -#endif - } - } - } - -#if 0 - printf("Consts:\n"); - for (i = 0; i < 4; i++) { - printf(" %d: %f %f %f %f\n", i, - machine.Consts[i][0], - machine.Consts[i][1], - machine.Consts[i][2], - machine.Consts[i][3]); - } -#endif - - /* run shader */ - tgsi_exec_machine_run( &machine ); - -#if 0 - printf("VS result: %f %f %f %f\n", - outputs[0].xyzw[0].f[0], - outputs[0].xyzw[1].f[0], - outputs[0].xyzw[2].f[0], - outputs[0].xyzw[3].f[0]); -#endif - - /* store machine results */ - assert(sp->vs.outputs_written & (1 << VERT_RESULT_HPOS)); - for (j = 0; j < count; j++) { - unsigned attr, slot; - float x, y, z, w; - - /* Handle attr[0] (position) specially: */ - x = vOut[j]->clip[0] = outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = outputs[0].xyzw[3].f[j]; - - vOut[j]->clipmask = compute_clipmask(x, y, z, w); - vOut[j]->edgeflag = 1; - - /* divide by w */ - w = 1.0 / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; -#if 0 - printf("wincoord: %f %f %f\n", - vOut[j]->data[0][0], - vOut[j]->data[0][1], - vOut[j]->data[0][2]); -#endif - - /* remaining attributes: */ - /* pack into sequential post-transform attrib slots */ - slot = 1; - for (attr = 1; attr < VERT_RESULT_MAX; attr++) { - if (sp->vs.outputs_written & (1 << attr)) { - assert(slot < sp->nr_attrs); - vOut[j]->data[slot][0] = outputs[attr].xyzw[0].f[j]; - vOut[j]->data[slot][1] = outputs[attr].xyzw[1].f[j]; - vOut[j]->data[slot][2] = outputs[attr].xyzw[2].f[j]; - vOut[j]->data[slot][3] = outputs[attr].xyzw[3].f[j]; - slot++; - } - } - } - -#if 0 - memcpy( - quad->outputs.color, - &machine.Outputs[1].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); -#endif -} - - -/** - * Called by the draw module when the vertx cache needs to be flushed. - * This involves running the vertex shader. - */ -static void vs_flush( struct draw_context *draw ) -{ - struct softpipe_context *sp = (struct softpipe_context *) draw->pipe; - unsigned i, j; - - /* run vertex shader on vertex cache entries, four per invokation */ - for (i = 0; i < draw->vs.queue_nr; i += 4) { - struct vertex_header *dests[4]; - unsigned elts[4]; - int n; - - for (j = 0; j < 4; j++) { - elts[j] = draw->vs.queue[i + j].elt; - dests[j] = draw->vs.queue[i + j].dest; - } - - n = MIN2(4, draw->vs.queue_nr - i); - assert(n > 0); - assert(n <= 4); - - run_vertex_program(sp, elts, n, dests); - } - - draw->vs.queue_nr = 0; -} @@ -322,6 +53,10 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, /** + * Draw vertex arrays, with optional indexing. + * Basically, map the vertex buffers (and drawing surfaces), then hand off + * the drawing to the 'draw' module. + * * XXX should the element buffer be specified/bound with a separate function? */ void @@ -330,7 +65,6 @@ softpipe_draw_elements(struct pipe_context *pipe, unsigned indexSize, unsigned mode, unsigned start, unsigned count) { - struct softpipe_context *sp = softpipe_context(pipe); struct draw_context *draw = sp->draw; unsigned length, first, incr, i; @@ -353,10 +87,11 @@ softpipe_draw_elements(struct pipe_context *pipe, */ for (i = 0; i < PIPE_ATTRIB_MAX; i++) { if (sp->vertex_buffer[i].buffer) { - sp->mapped_vbuffer[i] + void *buf = pipe->winsys->buffer_map(pipe->winsys, sp->vertex_buffer[i].buffer, PIPE_BUFFER_FLAG_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); } } /* Map index buffer, if present */ @@ -364,30 +99,15 @@ softpipe_draw_elements(struct pipe_context *pipe, mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, PIPE_BUFFER_FLAG_READ); - draw_set_element_buffer(draw, indexSize, mapped_indexes); + draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { - draw_set_element_buffer(draw, 0, NULL); /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); /* no index/element buffer */ } - /* tell drawing pipeline we're beginning drawing */ - draw->pipeline.first->begin( draw->pipeline.first ); - - draw->vs_flush = vs_flush; - draw->pipe = pipe; /* XXX pass pipe to draw_create() */ - - draw_invalidate_vcache( draw ); - draw_set_prim( draw, mode ); + draw_arrays(draw, mode, start, count); - /* drawing done here: */ - draw_prim(draw, start, count); - - /* draw any left-over buffered prims */ - draw_flush(draw); - - /* tell drawing pipeline we're done drawing */ - draw->pipeline.first->end( draw->pipeline.first ); /* * unmap vertex/index buffers @@ -395,58 +115,13 @@ softpipe_draw_elements(struct pipe_context *pipe, for (i = 0; i < PIPE_ATTRIB_MAX; i++) { if (sp->vertex_buffer[i].buffer) { pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); } } if (indexBuffer) { pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + draw_set_mapped_element_buffer(draw, 0, NULL); } softpipe_unmap_surfaces(sp); } - - - -#define EMIT_ATTR( VF_ATTR, STYLE, SIZE ) \ -do { \ - if (draw->nr_attrs >= 2) \ - draw->vf_attr_to_slot[VF_ATTR] = draw->nr_attrs - 2; \ - draw->attrs[draw->nr_attrs].attrib = VF_ATTR; \ - draw->attrs[draw->nr_attrs].format = STYLE; \ - draw->nr_attrs++; \ - draw->vertex_size += SIZE; \ -} while (0) - - -/** - * XXX very similar to same func in draw_vb.c (which will go away) - */ -void -draw_set_vertex_attributes2( struct draw_context *draw, - const unsigned *slot_to_vf_attr, - unsigned nr_attrs ) -{ - unsigned i; - - memset(draw->vf_attr_to_slot, 0, sizeof(draw->vf_attr_to_slot)); - draw->nr_attrs = 0; - draw->vertex_size = 0; - - /* - * First three attribs are always the same: header, clip pos, winpos - */ - EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F, 1); - EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F, 4); - - assert(slot_to_vf_attr[0] == VF_ATTRIB_POS); - EMIT_ATTR(slot_to_vf_attr[0], EMIT_4F_VIEWPORT, 4); - - /* - * Remaining attribs (color, texcoords, etc) - */ - for (i = 1; i < nr_attrs; i++) - EMIT_ATTR(slot_to_vf_attr[i], EMIT_4F, 4); - - draw->vertex_size *= 4; /* floats to bytes */ -} - - diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 95b0cfee03..eba789e9c4 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -210,22 +210,15 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) } } - /* If the attributes have changed, tell the draw module (which in turn - * tells the vf module) about the new vertex layout. + /* If the attributes have changed, tell the draw module about + * the new vertex layout. */ if (attr_mask != softpipe->attr_mask) { softpipe->attr_mask = attr_mask; -#define USE_NEW_DRAW 01 -#if USE_NEW_DRAW - draw_set_vertex_attributes2( softpipe->draw, - slot_to_vf_attr, - softpipe->nr_attrs ); -#else draw_set_vertex_attributes( softpipe->draw, - slot_to_vf_attr, - softpipe->nr_attrs ); -#endif + slot_to_vf_attr, + softpipe->nr_attrs ); } } diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 3505c2f1fb..98d596d4c5 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -49,4 +49,6 @@ void softpipe_set_vs_state( struct pipe_context *pipe, memcpy(&softpipe->vs, vs, sizeof(*vs)); softpipe->dirty |= SP_NEW_VS; + + draw_set_vertex_shader(softpipe->draw, vs); } diff --git a/src/mesa/pipe/softpipe/sp_state_vertex.c b/src/mesa/pipe/softpipe/sp_state_vertex.c index d985b20c36..18852552eb 100644 --- a/src/mesa/pipe/softpipe/sp_state_vertex.c +++ b/src/mesa/pipe/softpipe/sp_state_vertex.c @@ -41,6 +41,8 @@ softpipe_set_vertex_element(struct pipe_context *pipe, assert(index < PIPE_ATTRIB_MAX); softpipe->vertex_element[index] = *attrib; /* struct copy */ softpipe->dirty |= SP_NEW_VERTEX; + + draw_set_vertex_element(softpipe->draw, index, attrib); } @@ -53,4 +55,6 @@ softpipe_set_vertex_buffer(struct pipe_context *pipe, assert(index < PIPE_ATTRIB_MAX); softpipe->vertex_buffer[index] = *buffer; /* struct copy */ softpipe->dirty |= SP_NEW_VERTEX; + + draw_set_vertex_buffer(softpipe->draw, index, buffer); } diff --git a/src/mesa/sources b/src/mesa/sources index d97a34f534..0731e64d8a 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -156,6 +156,7 @@ VF_SOURCES = \ DRAW_SOURCES = \ + pipe/draw/draw_arrays.c \ pipe/draw/draw_clip.c \ pipe/draw/draw_context.c\ pipe/draw/draw_cull.c \ -- cgit v1.2.3 From 51da8ee85eccf0df3721cbd863cd174382d1ddfd Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 16:21:08 -0600 Subject: Start to remove the temporary draw_vb() and draw_vertices() code. new st_draw_vertices() utility used by glClear and glDrawPixels --- src/mesa/pipe/draw/draw_arrays.c | 7 +-- src/mesa/pipe/draw/draw_vb.c | 66 ---------------------------- src/mesa/pipe/i915simple/i915_context.c | 5 ++- src/mesa/pipe/softpipe/sp_context.c | 6 +++ src/mesa/sources | 3 +- src/mesa/state_tracker/st_cb_clear.c | 29 +++++++++++-- src/mesa/state_tracker/st_cb_drawpixels.c | 12 +++--- src/mesa/state_tracker/st_draw.c | 72 +++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_draw.h | 12 ++++-- 9 files changed, 126 insertions(+), 86 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index 59098fe242..fe9ec44425 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -170,11 +170,12 @@ run_vertex_program(struct draw_context *draw, unsigned attr; for (attr = 0; attr < 16; attr++) { if (draw->vertex_shader.inputs_read & (1 << attr)) { + unsigned buf = draw->vertex_element[attr].vertex_buffer_index; const void *src - = (const void *) ((const ubyte *) draw->mapped_vbuffer[attr] - + draw->vertex_buffer[attr].buffer_offset + = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + draw->vertex_element[attr].src_offset - + elts[j] * draw->vertex_buffer[attr].pitch); + + elts[j] * draw->vertex_buffer[buf].pitch); float p[4]; fetch_attrib4(src, draw->vertex_element[attr].src_format, p); diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c index 0eefb0b250..f83f8b8a0e 100644 --- a/src/mesa/pipe/draw/draw_vb.c +++ b/src/mesa/pipe/draw/draw_vb.c @@ -259,72 +259,6 @@ void draw_vb(struct draw_context *draw, #endif /*MESA*/ -/** - * XXX Temporary mechanism to draw simple vertex arrays. - * All attribs are float[4]. Arrays are interleaved, in GL-speak. - */ -void -draw_vertices(struct draw_context *draw, - unsigned mode, - unsigned numVerts, const float *vertices, - unsigned numAttrs, const unsigned attribs[]) -{ - /*unsigned first, incr;*/ - unsigned i, j; - - assert(mode <= PIPE_PRIM_POLYGON); - - draw->vs_flush = vs_flush; - - draw->vertex_size - = sizeof(struct vertex_header) + numAttrs * 4 * sizeof(float); - - - /* no element/index buffer */ - draw_set_mapped_element_buffer(draw, 0, NULL); - - /*draw_prim_info(mode, &first, &incr);*/ - draw_allocate_vertices( draw, numVerts ); - draw->pipeline.first->begin( draw->pipeline.first ); - - if (draw->prim != mode) - draw_set_prim( draw, mode ); - - /* setup attr info */ - draw->nr_attrs = numAttrs + 2; - draw->attrs[0].attrib = VF_ATTRIB_VERTEX_HEADER; - draw->attrs[0].format = EMIT_1F; - draw->attrs[1].attrib = VF_ATTRIB_CLIP_POS; - draw->attrs[1].format = EMIT_4F; - for (j = 0; j < numAttrs; j++) { - draw->vf_attr_to_slot[attribs[j]] = 2+j; - draw->attrs[2+j].attrib = attribs[j]; - draw->attrs[2+j].format = EMIT_4F; - } - - /* build vertices */ - for (i = 0; i < numVerts; i++) { - struct vertex_header *v - = (struct vertex_header *) (draw->verts + i * draw->vertex_size); - v->clipmask = 0x0; - v->edgeflag = 0; - for (j = 0; j < numAttrs; j++) { - COPY_4FV(v->data[j], vertices + (i * numAttrs + j) * 4); - } - } - - /* draw */ - draw_prim(draw, 0, numVerts); - draw_flush(draw); - draw->pipeline.first->end( draw->pipeline.first ); - - - /* clean up */ - draw_release_vertices( draw ); - draw->verts = NULL; - draw->in_vb = 0; -} - #if 000 /** diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 8a8582e0b2..c7d469583c 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -155,7 +155,9 @@ static void i915_draw_vb( struct pipe_context *pipe, if (i915->dirty) i915_update_derived( i915 ); +#if 0 draw_vb( i915->draw, VB ); +#endif } @@ -169,8 +171,9 @@ i915_draw_vertices(struct pipe_context *pipe, if (i915->dirty) i915_update_derived( i915 ); - +#if 0 draw_vertices(i915->draw, mode, numVertex, verts, numAttribs, attribs); +#endif } diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index e7694bbe86..91ede05c05 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -195,6 +195,7 @@ static void softpipe_destroy( struct pipe_context *pipe ) } +#if 0 static void softpipe_draw_vb( struct pipe_context *pipe, struct vertex_buffer *VB ) { @@ -208,6 +209,7 @@ static void softpipe_draw_vb( struct pipe_context *pipe, draw_vb( softpipe->draw, VB ); softpipe_unmap_surfaces(softpipe); } +#endif static void @@ -223,7 +225,9 @@ softpipe_draw_vertices(struct pipe_context *pipe, /* XXX move mapping/unmapping to higher/coarser level? */ softpipe_map_surfaces(softpipe); +#if 0 draw_vertices(softpipe->draw, mode, numVertex, verts, numAttribs, attribs); +#endif softpipe_unmap_surfaces(softpipe); } @@ -285,8 +289,10 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; +#if 0 softpipe->pipe.draw_vb = softpipe_draw_vb; softpipe->pipe.draw_vertices = softpipe_draw_vertices; +#endif softpipe->pipe.draw_arrays = softpipe_draw_arrays; softpipe->pipe.draw_elements = softpipe_draw_elements; diff --git a/src/mesa/sources b/src/mesa/sources index 0731e64d8a..0fe97e0287 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -164,8 +164,7 @@ DRAW_SOURCES = \ pipe/draw/draw_offset.c \ pipe/draw/draw_prim.c \ pipe/draw/draw_twoside.c \ - pipe/draw/draw_unfilled.c \ - pipe/draw/draw_vb.c + pipe/draw/draw_unfilled.c TGSICORE_SOURCES = \ pipe/tgsi/core/tgsi_build.c \ diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index f6c65ff466..bd050ca2cd 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -38,11 +38,17 @@ #include "st_context.h" #include "st_cb_clear.h" #include "st_cb_fbo.h" +#include "st_draw.h" #include "st_program.h" #include "st_public.h" + #include "pipe/p_context.h" +#include "pipe/p_state.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" + #include "pipe/tgsi/mesa/mesa_to_tgsi.h" + #include "vf/vf.h" @@ -215,8 +221,8 @@ draw_quad(GLcontext *ctx, const GLfloat color[4]) { static const GLuint attribs[2] = { - VF_ATTRIB_POS, - VF_ATTRIB_COLOR0 + 0, /* pos */ + 3 /* color */ }; GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ GLuint i; @@ -244,8 +250,7 @@ draw_quad(GLcontext *ctx, verts[i][1][3] = color[3]; } - ctx->st->pipe->draw_vertices(ctx->st->pipe, PIPE_PRIM_QUADS, - 4, (GLfloat *) verts, 2, attribs); + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, attribs); } @@ -343,6 +348,21 @@ clear_with_quad(GLcontext *ctx, st->pipe->set_fs_state(st->pipe, &fs); } + /* vertex shader state: color/position pass-through */ + { + static struct st_vertex_program *stvp = NULL; + struct pipe_shader_state vs; + if (!stvp) { + stvp = make_vertex_shader(st); + } + memset(&vs, 0, sizeof(vs)); + vs.inputs_read = stvp->Base.Base.InputsRead; + vs.outputs_written = stvp->Base.Base.OutputsWritten; + vs.tokens = &stvp->tokens[0]; + vs.constants = NULL; + st->pipe->set_vs_state(st->pipe, &vs); + } + /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); @@ -351,6 +371,7 @@ clear_with_quad(GLcontext *ctx, st->pipe->set_blend_state(st->pipe, &st->state.blend); st->pipe->set_depth_state(st->pipe, &st->state.depth); st->pipe->set_fs_state(st->pipe, &st->state.fs); + st->pipe->set_vs_state(st->pipe, &st->state.vs); st->pipe->set_setup_state(st->pipe, &st->state.setup); st->pipe->set_stencil_state(st->pipe, &st->state.stencil); /* OR: diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 11261f1d99..1aa56fcf8b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -38,6 +38,7 @@ #include "st_program.h" #include "st_cb_drawpixels.h" #include "st_cb_texture.h" +#include "st_draw.h" #include "st_format.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -193,12 +194,12 @@ free_mipmap_tree(struct pipe_context *pipe, struct pipe_mipmap_tree *mt) * Y=0=top */ static void -draw_quad(struct st_context *st, GLfloat x0, GLfloat y0, GLfloat z, +draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, GLfloat x1, GLfloat y1) { static const GLuint attribs[2] = { - VF_ATTRIB_POS, - VF_ATTRIB_TEX0 + 0, /* pos */ + 8 /* tex0 */ }; GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ GLuint i; @@ -235,8 +236,7 @@ draw_quad(struct st_context *st, GLfloat x0, GLfloat y0, GLfloat z, verts[i][1][3] = 1.0; /*Q*/ } - st->pipe->draw_vertices(st->pipe, PIPE_PRIM_QUADS, - 4, (GLfloat *) verts, 2, attribs); + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (GLfloat *) verts, 2, attribs); } @@ -310,7 +310,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, y1 = ctx->DrawBuffer->Height - 1 - (y + height * ctx->Pixel.ZoomY); /* draw textured quad */ - draw_quad(ctx->st, x0, y0, z, x1, y1); + draw_quad(ctx, x0, y0, z, x1, y1); /* restore GL state */ pipe->set_setup_state(pipe, &ctx->st->state.setup); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index e82f7cff2b..a409dd9764 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -66,7 +66,9 @@ static GLboolean draw( GLcontext * ctx, struct tnl_pipeline_stage *stage ) { struct st_context *st = st_context(ctx); +#if 0 struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; +#endif /* Validate driver and pipe state: */ @@ -74,7 +76,9 @@ static GLboolean draw( GLcontext * ctx, struct tnl_pipeline_stage *stage ) /* Call into the new draw code to handle the VB: */ +#if 0 st->pipe->draw_vb( st->pipe, VB ); +#endif /* Finished */ @@ -289,6 +293,72 @@ draw_vbo(GLcontext *ctx, +/** + * Utility function for drawing simple primitives (such as quads for + * glClear and glDrawPixels). Coordinates are in screen space. + * \param mode one of PIPE_PRIM_x + * \param numVertex number of vertices + * \param verts vertex data (all attributes are float[4]) + * \param numAttribs number of attributes per vertex + * \param attribs index of each attribute (0=pos, 3=color, etc) + */ +void +st_draw_vertices(GLcontext *ctx, unsigned prim, + unsigned numVertex, float *verts, + unsigned numAttribs, const unsigned attribs[]) +{ + const float width = ctx->DrawBuffer->Width; + const float height = ctx->DrawBuffer->Height; + const unsigned vertex_bytes = numVertex * numAttribs * 4 * sizeof(float); + struct pipe_context *pipe = ctx->st->pipe; + struct pipe_buffer_handle *vbuf; + struct pipe_vertex_buffer vbuffer; + struct pipe_vertex_element velement; + unsigned i; + + assert(numAttribs > 0); + assert(attribs[0] == 0); /* position */ + + /* convert to clip coords */ + for (i = 0; i < numVertex; i++) { + float x = verts[i * numAttribs * 4 + 0]; + float y = verts[i * numAttribs * 4 + 1]; + x = x / width * 2.0 - 1.0; + y = y / height * 2.0 - 1.0; + verts[i * numAttribs * 4 + 0] = x; + verts[i * numAttribs * 4 + 1] = y; + } + + /* XXX create one-time */ + vbuf = pipe->winsys->buffer_create(pipe->winsys, 32); + pipe->winsys->buffer_data(pipe->winsys, vbuf, vertex_bytes, verts); + + /* tell pipe about the vertex buffer */ + vbuffer.buffer = vbuf; + vbuffer.pitch = numAttribs * 4 * sizeof(float); /* vertex size */ + vbuffer.buffer_offset = 0; + pipe->set_vertex_buffer(pipe, 0, &vbuffer); + + /* tell pipe about the vertex attributes */ + for (i = 0; i < numAttribs; i++) { + velement.src_offset = i * 4 * sizeof(GLfloat); + velement.vertex_buffer_index = 0; + velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + velement.dst_offset = 0; + pipe->set_vertex_element(pipe, attribs[i], &velement); + } + + /* draw */ + pipe->draw_arrays(pipe, prim, 0, numVertex); + + /* XXX: do one-time */ + pipe->winsys->buffer_unreference(pipe->winsys, &vbuf); +} + + + + + /* This is all a hack to keep using tnl until we have vertex programs * up and running. @@ -307,8 +377,10 @@ void st_init_draw( struct st_context *st ) vbo->draw_prims = draw_vbo; #endif +#if 0 _tnl_destroy_pipeline( ctx ); _tnl_install_pipeline( ctx, st_pipeline ); +#endif /* USE_NEW_DRAW */ _tnl_ProgramCacheInit( ctx ); diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h index 0afadab577..0005fbc51f 100644 --- a/src/mesa/state_tracker/st_draw.h +++ b/src/mesa/state_tracker/st_draw.h @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2004 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -34,11 +34,15 @@ #ifndef ST_DRAW_H #define ST_DRAW_H + void st_init_draw( struct st_context *st ); + void st_destroy_draw( struct st_context *st ); -/** XXX temporary here */ -void st_clear(struct st_context *st, GLboolean color, GLboolean depth, - GLboolean stencil); +void +st_draw_vertices(GLcontext *ctx, unsigned prim, + unsigned numVertex, float *verts, + unsigned numAttribs, const unsigned attribs[]); + #endif -- cgit v1.2.3 From 8175eaa3b49f3b62f7ab251c4e0fbd14dd9f7e2b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 16:31:12 -0600 Subject: Checkpoint: remove more of the old draw_vb() code. --- src/mesa/pipe/draw/draw_arrays.c | 3 -- src/mesa/pipe/draw/draw_context.c | 6 ++- src/mesa/pipe/draw/draw_private.h | 7 +-- src/mesa/pipe/i915simple/i915_context.c | 28 ++--------- src/mesa/pipe/p_context.h | 17 ------- src/mesa/pipe/softpipe/sp_context.c | 42 ----------------- src/mesa/pipe/softpipe/sp_state_vertex.c | 4 ++ src/mesa/state_tracker/st_draw.c | 80 +------------------------------- 8 files changed, 18 insertions(+), 169 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index fe9ec44425..8cb34e8f8e 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -354,9 +354,6 @@ do { \ } while (0) -/** - * XXX very similar to same func in draw_vb.c (which will go away) - */ void draw_set_vertex_attributes( struct draw_context *draw, const unsigned *slot_to_vf_attr, diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index cc00576c53..0a3df37b7f 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -60,7 +60,9 @@ struct draw_context *draw_create( void ) draw->nr_planes = 6; #ifdef MESA +#if 0 draw->vf = vf_create( GL_TRUE ); +#endif #endif /* Statically allocate maximum sized vertices for the cache - could be cleverer... @@ -79,7 +81,7 @@ struct draw_context *draw_create( void ) void draw_destroy( struct draw_context *draw ) { -#ifdef MESA +#if 0/*def MESA*/ if (draw->header.storage) { ALIGN_FREE( draw->header.storage ); } @@ -194,7 +196,9 @@ void draw_set_viewport_state( struct draw_context *draw, draw->viewport = *viewport; /* struct copy */ #ifdef MESA +#if 0 vf_set_vp_scale_translate( draw->vf, viewport->scale, viewport->translate ); +#endif #endif /* Using tnl/ and vf/ modules is temporary while getting started. diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 5c3efb80e9..a9825e646b 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -222,14 +222,11 @@ struct draw_context } pq; - /* Misc for draw_vb.c (XXX temporary) - */ -#ifdef MESA - GLvector4f header; -#endif +#if 0 ubyte *verts; boolean in_vb; struct vertex_fetch *vf; +#endif }; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index c7d469583c..b34899a867 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -147,33 +147,16 @@ static void i915_destroy( struct pipe_context *pipe ) free( i915 ); } -static void i915_draw_vb( struct pipe_context *pipe, - struct vertex_buffer *VB ) -{ - struct i915_context *i915 = i915_context( pipe ); - - if (i915->dirty) - i915_update_derived( i915 ); - -#if 0 - draw_vb( i915->draw, VB ); -#endif -} - -static void -i915_draw_vertices(struct pipe_context *pipe, - unsigned mode, - unsigned numVertex, const float *verts, - unsigned numAttribs, const unsigned attribs[]) +static void i915_draw_arrays( struct pipe_context *pipe, + unsigned mode, unsigned start, unsigned count) { struct i915_context *i915 = i915_context( pipe ); if (i915->dirty) i915_update_derived( i915 ); -#if 0 - draw_vertices(i915->draw, mode, numVertex, verts, numAttribs, attribs); -#endif + + draw_arrays(i915->draw, mode, start, count); } @@ -217,12 +200,11 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.destroy = i915_destroy; i915->pipe.supported_formats = i915_supported_formats; i915->pipe.max_texture_size = i915_max_texture_size; - i915->pipe.draw_vb = i915_draw_vb; - i915->pipe.draw_vertices = i915_draw_vertices; i915->pipe.clear = i915_clear; i915->pipe.reset_occlusion_counter = NULL; /* no support */ i915->pipe.get_occlusion_counter = NULL; + i915->pipe.draw_arrays = i915_draw_arrays; /* * Create drawing context and plug our rendering stage into it. diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index a3664a3b80..c4496cda50 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -28,16 +28,10 @@ #ifndef PIPE_CONTEXT_H #define PIPE_CONTEXT_H -//#include "main/mtypes.h" #include "p_state.h" #include "p_compiler.h" -/* Drawing currently kludged up via the existing tnl/ module. - */ -struct vertex_buffer; - - /** * Software pipeline rendering context. Basically a collection of * state setting functions, plus VBO drawing entrypoint. @@ -68,17 +62,6 @@ struct pipe_context { /* * Drawing */ - /* XXX this is temporary */ - void (*draw_vb)( struct pipe_context *pipe, - struct vertex_buffer *VB ); - - /* XXX this is temporary */ - void (*draw_vertices)( struct pipe_context *pipe, - unsigned mode, - unsigned numVertex, const float *verts, - unsigned numAttribs, const unsigned attribs[]); - - /** this is basically what we want */ void (*draw_arrays)( struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 91ede05c05..ab9becc99e 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -195,44 +195,6 @@ static void softpipe_destroy( struct pipe_context *pipe ) } -#if 0 -static void softpipe_draw_vb( struct pipe_context *pipe, - struct vertex_buffer *VB ) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - - if (softpipe->dirty) - softpipe_update_derived( softpipe ); - - /* XXX move mapping/unmapping to higher/coarser level? */ - softpipe_map_surfaces(softpipe); - draw_vb( softpipe->draw, VB ); - softpipe_unmap_surfaces(softpipe); -} -#endif - - -static void -softpipe_draw_vertices(struct pipe_context *pipe, - unsigned mode, - unsigned numVertex, const float *verts, - unsigned numAttribs, const unsigned attribs[]) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - - if (softpipe->dirty) - softpipe_update_derived( softpipe ); - - /* XXX move mapping/unmapping to higher/coarser level? */ - softpipe_map_surfaces(softpipe); -#if 0 - draw_vertices(softpipe->draw, mode, numVertex, verts, numAttribs, attribs); -#endif - softpipe_unmap_surfaces(softpipe); -} - - - static void softpipe_reset_occlusion_counter(struct pipe_context *pipe) { struct softpipe_context *softpipe = softpipe_context( pipe ); @@ -289,10 +251,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; -#if 0 - softpipe->pipe.draw_vb = softpipe_draw_vb; - softpipe->pipe.draw_vertices = softpipe_draw_vertices; -#endif softpipe->pipe.draw_arrays = softpipe_draw_arrays; softpipe->pipe.draw_elements = softpipe_draw_elements; diff --git a/src/mesa/pipe/softpipe/sp_state_vertex.c b/src/mesa/pipe/softpipe/sp_state_vertex.c index 18852552eb..09ff540ccf 100644 --- a/src/mesa/pipe/softpipe/sp_state_vertex.c +++ b/src/mesa/pipe/softpipe/sp_state_vertex.c @@ -27,10 +27,14 @@ /* Authors: Keith Whitwell */ + + #include "sp_context.h" #include "sp_state.h" #include "sp_surface.h" +#include "pipe/draw/draw_context.h" + void softpipe_set_vertex_element(struct pipe_context *pipe, diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index a409dd9764..5a13fdd2bf 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -33,10 +33,9 @@ #include "main/imports.h" #include "vbo/vbo.h" +#include "vbo/vbo_context.h" -#include "tnl/t_context.h" -#include "tnl/t_pipeline.h" -#include "tnl/t_vp_build.h" /* USE_NEW_DRAW */ +#include "tnl/t_vp_build.h" #include "st_context.h" #include "st_atom.h" @@ -46,72 +45,6 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "vbo/vbo_context.h" - -/* - * Enabling this causes the VBO module to call draw_vbo() below, - * bypassing the T&L module. This only works with VBO-based demos, - * such as progs/test/bufferobj.c - */ -#define USE_NEW_DRAW 01 - - -/* - * TNL stage which feeds into the above. - * - * XXX: this needs to go into each driver using this code, because we - * cannot make the leap from ctx->draw_context in this file. The - * driver needs to customize tnl anyway, so this isn't a big deal. - */ -static GLboolean draw( GLcontext * ctx, struct tnl_pipeline_stage *stage ) -{ - struct st_context *st = st_context(ctx); -#if 0 - struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; -#endif - - /* Validate driver and pipe state: - */ - st_validate_state( st ); - - /* Call into the new draw code to handle the VB: - */ -#if 0 - st->pipe->draw_vb( st->pipe, VB ); -#endif - - /* Finished - */ - return GL_FALSE; -} - -const struct tnl_pipeline_stage st_draw = { - "check state and draw", - NULL, - NULL, - NULL, - NULL, - draw -}; - -static const struct tnl_pipeline_stage *st_pipeline[] = { -#if USE_NEW_DRAW - &_tnl_vertex_program_stage, -#else - &_tnl_vertex_transform_stage, - &_tnl_vertex_cull_stage, - &_tnl_normal_transform_stage, - &_tnl_lighting_stage, - &_tnl_fog_coordinate_stage, - &_tnl_texgen_stage, - &_tnl_texture_transform_stage, - &_tnl_point_attenuation_stage, - &_tnl_vertex_program_stage, -#endif - &st_draw, /* ADD: escape to pipe */ - 0, -}; - static GLuint @@ -366,8 +299,6 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, void st_init_draw( struct st_context *st ) { GLcontext *ctx = st->ctx; - -#if USE_NEW_DRAW struct vbo_context *vbo = (struct vbo_context *) ctx->swtnl_im; create_default_attribs_buffer(st); @@ -376,13 +307,6 @@ void st_init_draw( struct st_context *st ) assert(vbo->draw_prims); vbo->draw_prims = draw_vbo; -#endif -#if 0 - _tnl_destroy_pipeline( ctx ); - _tnl_install_pipeline( ctx, st_pipeline ); -#endif - - /* USE_NEW_DRAW */ _tnl_ProgramCacheInit( ctx ); } -- cgit v1.2.3 From 9d9e838974d1b9ab35235387bdfda5d99cb19d38 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 16:43:58 -0600 Subject: remove dead code related to vf module --- src/mesa/pipe/draw/draw_context.c | 30 ++---------------------------- src/mesa/pipe/draw/draw_private.h | 7 ------- 2 files changed, 2 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 0a3df37b7f..760756ce89 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -31,14 +31,12 @@ */ -#ifdef MESA -#include "main/macros.h" -#else #include "pipe/p_util.h" -#endif #include "draw_context.h" #include "draw_private.h" + + struct draw_context *draw_create( void ) { struct draw_context *draw = CALLOC_STRUCT( draw_context ); @@ -59,12 +57,6 @@ struct draw_context *draw_create( void ) ASSIGN_4V( draw->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ draw->nr_planes = 6; -#ifdef MESA -#if 0 - draw->vf = vf_create( GL_TRUE ); -#endif -#endif - /* Statically allocate maximum sized vertices for the cache - could be cleverer... */ { @@ -81,14 +73,6 @@ struct draw_context *draw_create( void ) void draw_destroy( struct draw_context *draw ) { -#if 0/*def MESA*/ - if (draw->header.storage) { - ALIGN_FREE( draw->header.storage ); - } - - vf_destroy( draw->vf ); -#endif - free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ free( draw ); } @@ -194,16 +178,6 @@ void draw_set_viewport_state( struct draw_context *draw, const struct pipe_viewport_state *viewport ) { draw->viewport = *viewport; /* struct copy */ - -#ifdef MESA -#if 0 - vf_set_vp_scale_translate( draw->vf, viewport->scale, viewport->translate ); -#endif -#endif - - /* Using tnl/ and vf/ modules is temporary while getting started. - * Full pipe will have vertex shader, vertex fetch of its own. - */ } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index a9825e646b..5983bc1174 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -220,13 +220,6 @@ struct draw_context struct prim_header queue[PRIM_QUEUE_LENGTH]; unsigned queue_nr; } pq; - - -#if 0 - ubyte *verts; - boolean in_vb; - struct vertex_fetch *vf; -#endif }; -- cgit v1.2.3 From ddf8ba2d1b4f85498613f7ace1ffda70bb9d9e92 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 16:48:55 -0600 Subject: Obsolete --- src/mesa/pipe/draw/draw_vb.c | 320 ------------------------------------------- 1 file changed, 320 deletions(-) delete mode 100644 src/mesa/pipe/draw/draw_vb.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vb.c b/src/mesa/pipe/draw/draw_vb.c deleted file mode 100644 index f83f8b8a0e..0000000000 --- a/src/mesa/pipe/draw/draw_vb.c +++ /dev/null @@ -1,320 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#ifdef MESA -#include "main/imports.h" -#include "main/macros.h" -#include "tnl/t_context.h" -#include "vf/vf.h" -#else -#define VF_ -#include "pipe/p_util.h" - -#endif /*MESA*/ - - -#include "draw_private.h" -#include "draw_context.h" -#include "draw_prim.h" - - - - -/* This file is a temporary set of hooks to allow us to use the tnl/ - * and vf/ modules until we have replacements in pipe. - */ - -static void vs_flush( struct draw_context *draw ) -{ - unsigned i; - - /* We're not really running a vertex shader yet, so flushing the vs - * queue is just a matter of building the vertices and returning. - */ - /* Actually, I'm cheating even more and pre-building them still - * with the mesa/vf module. So it's very easy... - */ - for (i = 0; i < draw->vs.queue_nr; i++) { - /* Would do the following steps here: - * - * 1) Loop over vertex element descriptors, fetch data from each - * to build the pre-tnl vertex. This might require a new struct - * to represent the pre-tnl vertex. - * - * 2) Bundle groups of upto 4 pre-tnl vertices together and pass - * to vertex shader. - * - * 3) Do any necessary unswizzling, make sure vertex headers are - * correctly populated, store resulting post-transformed - * vertices in vcache. - * - * In this version, just do the last step: - */ - unsigned elt = draw->vs.queue[i].elt; - struct vertex_header *dest = draw->vs.queue[i].dest; - - /* Magic: - */ - memcpy(dest, - draw->verts + elt * draw->vertex_size, - draw->vertex_size); - - } - draw->vs.queue_nr = 0; -} - - - -/** - * Allocate storage for post-transformation vertices. - */ -static void -draw_allocate_vertices( struct draw_context *draw, unsigned nr_vertices ) -{ - assert(draw->vertex_size > 0); - draw->nr_vertices = nr_vertices; - draw->verts = (ubyte *) malloc( nr_vertices * draw->vertex_size ); - draw_invalidate_vcache( draw ); -} - - - -/** - * Free storage which was allocated by draw_allocate_vertices() - */ -static void -draw_release_vertices( struct draw_context *draw ) -{ - free(draw->verts); - draw->verts = NULL; -} - - -/** - * Note: this must match struct vertex_header's layout (I think). - */ -struct header_dword { - unsigned clipmask:12; - unsigned edgeflag:1; - unsigned pad:19; -}; - - -#ifdef MESA -static void -build_vertex_headers( struct draw_context *draw, - struct vertex_buffer *VB ) -{ - if (draw->header.storage == NULL) { - draw->header.stride = sizeof(float); - draw->header.size = 1; - draw->header.storage = ALIGN_MALLOC( VB->Size * sizeof(float), 32 ); - draw->header.data = draw->header.storage; - draw->header.count = 0; - draw->header.flags = VEC_SIZE_1 | VEC_MALLOC; - } - - /* Build vertex header attribute. - * - */ - - { - unsigned i; - struct header_dword *header = (struct header_dword *)draw->header.storage; - - /* yes its a hack - */ - assert(sizeof(*header) == sizeof(float)); - - draw->header.count = VB->Count; - - if (VB->EdgeFlag) { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = VB->ClipMask[i]; - header[i].edgeflag = VB->EdgeFlag[i]; - header[i].pad = 0; - } - } - else if (VB->ClipOrMask) { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = VB->ClipMask[i]; - header[i].edgeflag = 0; - header[i].pad = 0; - } - } - else { - for (i = 0; i < VB->Count; i++) { - header[i].clipmask = 0; - header[i].edgeflag = 0; - header[i].pad = 0; - } - } - } - - VB->AttribPtr[VF_ATTRIB_VERTEX_HEADER] = &draw->header; -} -#endif /*MESA*/ - - - -#ifdef MESA -/** - * This is a hack & will all go away. - */ -void draw_vb(struct draw_context *draw, - struct vertex_buffer *VB ) -{ - unsigned i; - - VB->AttribPtr[VF_ATTRIB_POS] = VB->NdcPtr; - VB->AttribPtr[VF_ATTRIB_BFC0] = VB->ColorPtr[1]; - VB->AttribPtr[VF_ATTRIB_BFC1] = VB->SecondaryColorPtr[1]; - VB->AttribPtr[VF_ATTRIB_CLIP_POS] = VB->ClipPtr; - - /* Build vertex headers: - */ - build_vertex_headers( draw, VB ); - - draw->vs_flush = vs_flush; - - draw->in_vb = 1; - - /* tell drawing pipeline we're beginning drawing */ - draw->pipeline.first->begin( draw->pipeline.first ); - - /* Allocate storage for the post-transform vertices: - */ - draw_allocate_vertices( draw, VB->Count ); - - /* Bind the vb outputs: - */ - vf_set_sources( draw->vf, VB->AttribPtr, 0 ); - vf_emit_vertices( draw->vf, VB->Count, draw->verts ); - - if (VB->Elts) - draw_set_mapped_element_buffer(draw, sizeof(unsigned), VB->Elts); - else - draw_set_mapped_element_buffer(draw, 0, NULL); - - for (i = 0; i < VB->PrimitiveCount; i++) { - const GLenum mode = VB->Primitive[i].mode; - const unsigned start = VB->Primitive[i].start; - unsigned length, first, incr; - - /* Trim the primitive down to a legal size. - */ - draw_prim_info( mode, &first, &incr ); - length = draw_trim( VB->Primitive[i].count, first, incr ); - - if (!length) - continue; - - if (draw->prim != mode) - draw_set_prim( draw, mode ); - - draw_prim( draw, start, length ); - } - - /* draw any left-over buffered prims */ - draw_flush(draw); - - /* tell drawing pipeline we're done drawing */ - draw->pipeline.first->end( draw->pipeline.first ); - - /* free the post-transformed vertices */ - draw_release_vertices( draw ); - draw->verts = NULL; - draw->in_vb = 0; - draw->elts = NULL; -} -#endif /*MESA*/ - - -#if 000 - -/** - * Accumulate another attribute's info. - * Note the "- 2" factor here. We need this because the vertex->data[] - * array does not include the first two attributes we emit (VERTEX_HEADER - * and CLIP_POS). So, the 3rd attribute actually winds up in the 1st - * position of the data[] array. - */ -#define EMIT_ATTR( VF_ATTR, STYLE ) \ -do { \ - if (draw->nr_attrs >= 2) \ - draw->vf_attr_to_slot[VF_ATTR] = draw->nr_attrs - 2; \ - draw->attrs[draw->nr_attrs].attrib = VF_ATTR; \ - draw->attrs[draw->nr_attrs].format = STYLE; \ - draw->nr_attrs++; \ -} while (0) - - -/** - * Tell the draw module about the layout of attributes in the vertex. - * We need this in order to know which vertex slot has color0, etc. - * - * \param slot_to_vf_attr an array which maps slot indexes to vertex - * format tokens (VF_*). - * \param nr_attrs the size of the slot_to_vf_attr array - * (and number of attributes) - */ -void draw_set_vertex_attributes( struct draw_context *draw, - const unsigned *slot_to_vf_attr, - unsigned nr_attrs ) -{ - unsigned i; - - memset(draw->vf_attr_to_slot, 0, sizeof(draw->vf_attr_to_slot)); - draw->nr_attrs = 0; - - /* - * First three attribs are always the same: header, clip pos, winpos - */ - EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F); - EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F); - - assert(slot_to_vf_attr[0] == VF_ATTRIB_POS); - EMIT_ATTR(slot_to_vf_attr[0], EMIT_4F_VIEWPORT); - - /* - * Remaining attribs (color, texcoords, etc) - */ - for (i = 1; i < nr_attrs; i++) - EMIT_ATTR(slot_to_vf_attr[i], EMIT_4F); - - /* tell the vertex format module how to construct vertices for us */ -#if MESA - draw->vertex_size = vf_set_vertex_attributes( draw->vf, draw->attrs, - draw->nr_attrs, 0 ); -#endif -} -#endif -- cgit v1.2.3 From 766fa51537dabd978eb04fb4c3f29b5dfeacd9fb Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 17:02:07 -0600 Subject: remove some of the #ifndef MESA stuff --- src/mesa/pipe/draw/draw_arrays.c | 7 ++++++- src/mesa/pipe/draw/draw_prim.c | 8 ++++---- src/mesa/pipe/draw/draw_private.h | 16 ++++++++-------- src/mesa/pipe/softpipe/sp_prim_setup.c | 4 ++-- 4 files changed, 20 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index 8cb34e8f8e..17ed255f85 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -45,6 +45,11 @@ #include "pipe/tgsi/core/tgsi_util.h" +/** XXX remove */ +#define VERT_RESULT_HPOS 0 +#define VERT_RESULT_MAX 24 + + #if defined __GNUC__ #define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE] __attribute__(( aligned( 16 ) )) #define ALIGN16_ASSIGN(P) P @@ -348,7 +353,7 @@ do { \ if (draw->nr_attrs >= 2) \ draw->vf_attr_to_slot[VF_ATTR] = draw->nr_attrs - 2; \ draw->attrs[draw->nr_attrs].attrib = VF_ATTR; \ - draw->attrs[draw->nr_attrs].format = STYLE; \ + /*draw->attrs[draw->nr_attrs].format = STYLE*/; \ draw->nr_attrs++; \ draw->vertex_size += SIZE; \ } while (0) diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index fbd0672875..d296780fd7 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -167,7 +167,7 @@ static struct vertex_header *get_vertex( struct draw_context *draw, static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, unsigned i ) { - const unsigned *elts = (const unsigned *)draw->elts; + const unsigned *elts = (const unsigned *) draw->mapped_elts; return get_vertex( draw, elts[i] ); } @@ -175,7 +175,7 @@ static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, unsigned i ) { - const ushort *elts = (const ushort *)draw->elts; + const ushort *elts = (const ushort *) draw->mapped_elts; return get_vertex( draw, elts[i] ); } @@ -183,7 +183,7 @@ static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, unsigned i ) { - const ubyte *elts = (const ubyte *)draw->elts; + const ubyte *elts = (const ubyte *) draw->mapped_elts; return get_vertex( draw, elts[i] ); } @@ -454,7 +454,7 @@ draw_set_mapped_element_buffer( struct draw_context *draw, default: assert(0); } - draw->elts = elements; + draw->mapped_elts = elements; draw->eltSize = eltSize; } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 5983bc1174..1912607b66 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -43,13 +43,14 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#ifdef MESA -#include "vf/vf.h" -#else + + /* XXX these are temporary */ struct vf_attr_map { unsigned attrib; + /* unsigned format; + */ unsigned offset; }; #define VF_ATTRIB_POS 0 @@ -65,7 +66,6 @@ struct vf_attr_map { #define EMIT_4F 3 #define EMIT_4F_VIEWPORT 6 #define FRAG_ATTRIB_MAX 13 -#endif /** @@ -162,6 +162,10 @@ struct draw_context struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; struct pipe_shader_state vertex_shader; + /** The mapped vertex element/index buffer */ + const void *mapped_elts; + unsigned eltSize; /**< bytes per index (0, 1, 2 or 4) */ + /** The mapped vertex arrays */ const void *mapped_vbuffer[PIPE_ATTRIB_MAX]; /* Clip derived state: @@ -179,10 +183,6 @@ struct draw_context unsigned vertex_size; /**< in bytes */ unsigned nr_vertices; - /** Pointer to vertex element/index buffer */ - unsigned eltSize; /**< bytes per index (0, 1, 2 or 4) */ - void *elts; - unsigned prim; /**< current prim type: PIPE_PRIM_x */ unsigned reduced_prim; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 43ab0e97db..f779deac28 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -40,10 +40,10 @@ #include "pipe/draw/draw_private.h" #include "pipe/p_util.h" -#ifndef MESA + +/** XXX remove */ #define FRAG_ATTRIB_WPOS 0 #define FRAG_ATTRIB_MAX 13 -#endif /** -- cgit v1.2.3 From 3239532795a027ddc578261f556e13e2c80f9676 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 17:25:38 -0600 Subject: Use new draw_arrays() code. The i915 driver now uses the software-based vertex shader interpreter and draws everything through pipe->draw_arrays(). --- src/mesa/pipe/i915simple/i915_context.c | 58 +++++++++++++++++++++++++++++++-- src/mesa/pipe/i915simple/i915_context.h | 2 ++ src/mesa/pipe/i915simple/i915_state.c | 35 ++++++++++++++++++++ 3 files changed, 92 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index b34899a867..fdb1a7422d 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -148,18 +148,69 @@ static void i915_destroy( struct pipe_context *pipe ) } -static void i915_draw_arrays( struct pipe_context *pipe, - unsigned mode, unsigned start, unsigned count) + +static void i915_draw_elements( struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) { struct i915_context *i915 = i915_context( pipe ); + struct draw_context *draw = i915->draw; + unsigned i; if (i915->dirty) i915_update_derived( i915 ); - draw_arrays(i915->draw, mode, start, count); + + /* + * Map vertex buffers + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (i915->vertex_buffer[i].buffer) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + i915->vertex_buffer[i].buffer, + PIPE_BUFFER_FLAG_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); + } + } + /* Map index buffer, if present */ + if (indexBuffer) { + void *mapped_indexes + = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, + PIPE_BUFFER_FLAG_READ); + draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + } + else { + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + /* draw! */ + draw_arrays(i915->draw, prim, start, count); + + /* + * unmap vertex/index buffers + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (i915->vertex_buffer[i].buffer) { + pipe->winsys->buffer_unmap(pipe->winsys, i915->vertex_buffer[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); + } + } + if (indexBuffer) { + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + draw_set_mapped_element_buffer(draw, 0, NULL); + } } +static void i915_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) +{ + i915_draw_elements(pipe, NULL, 0, prim, start, count); +} + struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, @@ -205,6 +256,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.get_occlusion_counter = NULL; i915->pipe.draw_arrays = i915_draw_arrays; + i915->pipe.draw_elements = i915_draw_elements; /* * Create drawing context and plug our rendering stage into it. diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 1e48485c56..a3927bf8b8 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -113,6 +113,8 @@ struct i915_context struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + unsigned dirty; unsigned *batch_start; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 06fa716c4f..1de6fea2e9 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -121,6 +121,16 @@ static void i915_set_fs_state( struct pipe_context *pipe, } +static void i915_set_vs_state( struct pipe_context *pipe, + const struct pipe_shader_state *vs ) +{ + struct i915_context *i915 = i915_context(pipe); + + /* just pass-through to draw module */ + draw_set_vertex_shader(i915->draw, vs); +} + + static void i915_set_sampler_state(struct pipe_context *pipe, unsigned unit, const struct pipe_sampler_state *sampler) @@ -216,6 +226,28 @@ static void i915_set_setup_state( struct pipe_context *pipe, } +static void i915_set_vertex_buffer( struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_buffer *buffer ) +{ + struct i915_context *i915 = i915_context(pipe); + i915->vertex_buffer[index] = *buffer; + /* pass-through to draw module */ + draw_set_vertex_buffer(i915->draw, index, buffer); +} + + +static void i915_set_vertex_element( struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_element *element) +{ + struct i915_context *i915 = i915_context(pipe); + /* pass-through to draw module */ + draw_set_vertex_element(i915->draw, index, element); +} + + + void i915_init_state_functions( struct i915_context *i915 ) { @@ -227,6 +259,7 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_depth_state = i915_set_depth_test_state; i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; i915->pipe.set_fs_state = i915_set_fs_state; + i915->pipe.set_vs_state = i915_set_vs_state; i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; i915->pipe.set_sampler_state = i915_set_sampler_state; i915->pipe.set_scissor_state = i915_set_scissor_state; @@ -234,4 +267,6 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_stencil_state = i915_set_stencil_state; i915->pipe.set_texture_state = i915_set_texture_state; i915->pipe.set_viewport_state = i915_set_viewport_state; + i915->pipe.set_vertex_buffer = i915_set_vertex_buffer; + i915->pipe.set_vertex_element = i915_set_vertex_element; } -- cgit v1.2.3 From c1fbd72adcbe1aa3a627e8f9a7a0803740c0a008 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 17:26:00 -0600 Subject: minor clean-ups --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 2df07eb7d6..9ced89cd7c 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -68,7 +68,6 @@ softpipe_draw_elements(struct pipe_context *pipe, struct softpipe_context *sp = softpipe_context(pipe); struct draw_context *draw = sp->draw; unsigned length, first, incr, i; - void *mapped_indexes = NULL; /* first, check that the primitive is not malformed */ draw_prim_info( mode, &first, &incr ); @@ -96,19 +95,19 @@ softpipe_draw_elements(struct pipe_context *pipe, } /* Map index buffer, if present */ if (indexBuffer) { - mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, - indexBuffer, - PIPE_BUFFER_FLAG_READ); + void *mapped_indexes + = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, + PIPE_BUFFER_FLAG_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { - draw_set_mapped_element_buffer(draw, 0, NULL); /* no index/element buffer */ + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); } - + /* draw! */ draw_arrays(draw, mode, start, count); - /* * unmap vertex/index buffers */ -- cgit v1.2.3 From d054331c4788b0a5f13f4d7a9b497d274d267acb Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 18:24:02 -0600 Subject: Update glClear and glDrawPixels code. Since they're implemented by drawing quadrilaterals (and go through vertex transformation and viewport mapping) we don't have to invert Y coords. --- src/mesa/pipe/draw/draw_arrays.c | 5 +- src/mesa/state_tracker/st_cb_clear.c | 13 +++-- src/mesa/state_tracker/st_cb_drawpixels.c | 94 +++++++++++++++++++++++++++---- 3 files changed, 95 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index 17ed255f85..34a031ddf0 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -248,10 +248,11 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][2] = z * scale[2] + trans[2]; vOut[j]->data[0][3] = w; #if 0 - printf("wincoord: %f %f %f\n", + printf("wincoord: %f %f %f %f\n", vOut[j]->data[0][0], vOut[j]->data[0][1], - vOut[j]->data[0][2]); + vOut[j]->data[0][2], + vOut[j]->data[0][3]); #endif /* remaining attributes: */ diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index bd050ca2cd..1ed48facd1 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -212,8 +212,8 @@ make_vertex_shader(struct st_context *st) /** * Draw a screen-aligned quadrilateral. - * Coords are window coords. - * XXX need to emit clip coords when using a vertex program! + * Coords are window coords with y=0=bottom. These coords will be transformed + * by the vertex shader and viewport transform (which will flip Y if needed). */ static void draw_quad(GLcontext *ctx, @@ -266,9 +266,9 @@ clear_with_quad(GLcontext *ctx, { struct st_context *st = ctx->st; const GLfloat x0 = ctx->DrawBuffer->_Xmin; - const GLfloat y0 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax; + const GLfloat y0 = ctx->DrawBuffer->_Ymin; const GLfloat x1 = ctx->DrawBuffer->_Xmax; - const GLfloat y1 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin; + const GLfloat y1 = ctx->DrawBuffer->_Ymax; /* alpha state: disabled */ { @@ -312,8 +312,13 @@ clear_with_quad(GLcontext *ctx, { struct pipe_setup_state setup; memset(&setup, 0, sizeof(setup)); +#if 0 + /* don't do per-pixel scissor; we'll just draw a PIPE_PRIM_QUAD + * that matches the scissor bounds. + */ if (ctx->Scissor.Enabled) setup.scissor = 1; +#endif st->pipe->set_setup_state(st->pipe, &setup); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 1aa56fcf8b..108304f5e2 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -49,11 +49,11 @@ /** - * Create a simple fragment shader that passes the texture color through - * to the fragment color. + * Create a simple fragment shader that does a TEX() instruction to get + * the fragment color. */ static struct st_fragment_program * -make_drawpixels_shader(struct st_context *st) +make_fragment_shader(struct st_context *st) { GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; @@ -95,6 +95,59 @@ make_drawpixels_shader(struct st_context *st) } +/** + * Create a simple vertex shader that just passes through the + * vertex position and color. + */ +static struct st_vertex_program * +make_vertex_shader(struct st_context *st) +{ + GLcontext *ctx = st->ctx; + struct st_vertex_program *stvp; + struct gl_program *p; + GLboolean b; + + p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); + if (!p) + return NULL; + + p->NumInstructions = 3; + p->Instructions = _mesa_alloc_instructions(3); + if (!p->Instructions) { + ctx->Driver.DeleteProgram(ctx, p); + return NULL; + } + _mesa_init_instructions(p->Instructions, 3); + /* MOV result.pos, vertex.pos; */ + p->Instructions[0].Opcode = OPCODE_MOV; + p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[0].DstReg.Index = VERT_RESULT_HPOS; + p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[0].SrcReg[0].Index = VERT_ATTRIB_POS; + /* MOV result.color, vertex.color; */ + p->Instructions[1].Opcode = OPCODE_MOV; + p->Instructions[1].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[1].DstReg.Index = VERT_RESULT_TEX0; + p->Instructions[1].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[1].SrcReg[0].Index = VERT_ATTRIB_TEX0; + /* END; */ + p->Instructions[2].Opcode = OPCODE_END; + + p->InputsRead = VERT_BIT_POS | VERT_BIT_TEX0; + p->OutputsWritten = ((1 << VERT_RESULT_TEX0) | + (1 << VERT_RESULT_HPOS)); + + stvp = (struct st_vertex_program *) p; + /* compile into tgsi format */ + b = tgsi_mesa_compile_vp_program(&stvp->Base, + stvp->tokens, ST_FP_MAX_TOKENS); + assert(b); + + return stvp; +} + + + static struct pipe_mipmap_tree * make_mipmap_tree(struct st_context *st, GLsizei width, GLsizei height, GLenum format, GLenum type, @@ -156,7 +209,7 @@ make_mipmap_tree(struct st_context *st, assert(success); } - mt->target = GL_TEXTURE_2D; + mt->target = PIPE_TEXTURE_2D; mt->internal_format = GL_RGBA; mt->format = pipeFormat; mt->first_level = 0; @@ -191,7 +244,7 @@ free_mipmap_tree(struct pipe_context *pipe, struct pipe_mipmap_tree *mt) /** * Draw textured quad. - * Y=0=top + * Coords are window coords with y=0=bottom. */ static void draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, @@ -236,7 +289,7 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, verts[i][1][3] = 1.0; /*Q*/ } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (GLfloat *) verts, 2, attribs); + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, attribs); } @@ -269,12 +322,12 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->set_setup_state(pipe, &setup); } - /* fragment shader state: color pass-through program */ + /* fragment shader state: TEX lookup program */ { static struct st_fragment_program *stfp = NULL; struct pipe_shader_state fs; if (!stfp) { - stfp = make_drawpixels_shader(ctx->st); + stfp = make_fragment_shader(ctx->st); } memset(&fs, 0, sizeof(fs)); fs.inputs_read = stfp->Base.Base.InputsRead; @@ -283,6 +336,21 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->set_fs_state(pipe, &fs); } + /* vertex shader state: position + texcoord pass-through */ + { + static struct st_vertex_program *stvp = NULL; + struct pipe_shader_state vs; + if (!stvp) { + stvp = make_vertex_shader(ctx->st); + } + memset(&vs, 0, sizeof(vs)); + vs.inputs_read = stvp->Base.Base.InputsRead; + vs.outputs_written = stvp->Base.Base.OutputsWritten; + vs.tokens = &stvp->tokens[0]; + vs.constants = NULL; + pipe->set_vs_state(pipe, &vs); + } + /* texture sampling state: */ { struct pipe_sampler_state sampler; @@ -303,11 +371,14 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->set_texture_state(pipe, unit, mt); } - /* compute window coords (y=0=top) with pixel zoom */ + /* Compute window coords (y=0=bottom) with pixel zoom. + * Recall that these coords are transformed by the current + * vertex shader and viewport transformation. + */ x0 = x; - y0 = ctx->DrawBuffer->Height - 1 - y; x1 = x + width * ctx->Pixel.ZoomX; - y1 = ctx->DrawBuffer->Height - 1 - (y + height * ctx->Pixel.ZoomY); + y0 = y; + y1 = y + height * ctx->Pixel.ZoomY; /* draw textured quad */ draw_quad(ctx, x0, y0, z, x1, y1); @@ -315,6 +386,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* restore GL state */ pipe->set_setup_state(pipe, &ctx->st->state.setup); pipe->set_fs_state(pipe, &ctx->st->state.fs); + pipe->set_vs_state(pipe, &ctx->st->state.vs); pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); pipe->set_sampler_state(pipe, unit, &ctx->st->state.sampler[unit]); -- cgit v1.2.3 From 79b8f97457a9634df28ba0def4623dbe2b142f09 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 18:44:53 -0600 Subject: include draw_context.h --- src/mesa/pipe/softpipe/sp_state_fs.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 98d596d4c5..bab407f047 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -28,6 +28,7 @@ #include "sp_context.h" #include "sp_state.h" +#include "pipe/draw/draw_context.h" void softpipe_set_fs_state( struct pipe_context *pipe, -- cgit v1.2.3 From 3772441c498b997947f68e98830f72649bbc8442 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 18:45:14 -0600 Subject: move ALIGN16-related macros into p_compiler.h --- src/mesa/pipe/draw/draw_arrays.c | 10 ---------- src/mesa/pipe/p_compiler.h | 10 ++++++++++ src/mesa/pipe/softpipe/sp_quad_fs.c | 27 ++++----------------------- 3 files changed, 14 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index 34a031ddf0..9b1426d1ec 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -50,16 +50,6 @@ #define VERT_RESULT_MAX 24 -#if defined __GNUC__ -#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE] __attribute__(( aligned( 16 ) )) -#define ALIGN16_ASSIGN(P) P -#else -#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE + 1] -#define ALIGN16_ASSIGN(P) align16(P) -#endif - - - static INLINE unsigned compute_clipmask(float cx, float cy, float cz, float cw) { diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h index d139cb1844..a4d05f74cd 100644 --- a/src/mesa/pipe/p_compiler.h +++ b/src/mesa/pipe/p_compiler.h @@ -68,4 +68,14 @@ typedef unsigned long long uint64; #endif +#if defined __GNUC__ +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE] __attribute__(( aligned( 16 ) )) +#define ALIGN16_ASSIGN(P) P +#else +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE + 1] +#define ALIGN16_ASSIGN(P) align16(P) +#endif + + + #endif /* P_COMPILER_H */ diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index c2decfecf8..c20f09d309 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -40,15 +40,6 @@ #include "sp_tex_sample.h" -#if defined __GNUC__ -#define USE_ALIGNED_ATTRIBS 1 -#define ALIGN16_SUFFIX __attribute__(( aligned( 16 ) )) -#else -#define USE_ALIGNED_ATTRIBS 0 -#define ALIGN16_SUFFIX -#endif - - struct quad_shade_stage { struct quad_stage stage; @@ -77,13 +68,8 @@ shade_quad( const float fy = (float) quad->y0; struct tgsi_exec_machine machine; -#if USE_ALIGNED_ATTRIBS - struct tgsi_exec_vector inputs[PIPE_ATTRIB_MAX] ALIGN16_SUFFIX; - struct tgsi_exec_vector outputs[PIPE_ATTRIB_MAX] ALIGN16_SUFFIX; -#else - struct tgsi_exec_vector inputs[PIPE_ATTRIB_MAX + 1]; - struct tgsi_exec_vector outputs[PIPE_ATTRIB_MAX + 1]; -#endif + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); #ifdef DEBUG memset( &machine, 0, sizeof( machine ) ); @@ -99,13 +85,8 @@ shade_quad( /* Consts does not require 16 byte alignment. */ machine.Consts = softpipe->fs.constants->constant; -#if USE_ALIGNED_ATTRIBS - machine.Inputs = inputs; - machine.Outputs = outputs; -#else - machine.Inputs = (struct tgsi_exec_vector *) tgsi_align_128bit( inputs ); - machine.Outputs = (struct tgsi_exec_vector *) tgsi_align_128bit( outputs ); -#endif + machine.Inputs = ALIGN16_ASSIGN(inputs); + machine.Outputs = ALIGN16_ASSIGN(outputs); machine.InterpCoefs = quad->coef; -- cgit v1.2.3 From 3cb6fc3f9aaa94236be611bad04125c17e01ad6c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 18:53:41 -0600 Subject: Move vertex transformation/shader code into draw_prim.c to avoid need for vs_flush() function pointer. --- src/mesa/pipe/draw/draw_arrays.c | 267 ------------------------------------- src/mesa/pipe/draw/draw_prim.c | 273 +++++++++++++++++++++++++++++++++++++- src/mesa/pipe/draw/draw_private.h | 2 - 3 files changed, 270 insertions(+), 272 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index 9b1426d1ec..738fa144b2 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -40,271 +40,6 @@ #include "pipe/draw/draw_context.h" #include "pipe/draw/draw_prim.h" -#include "pipe/tgsi/core/tgsi_exec.h" -#include "pipe/tgsi/core/tgsi_build.h" -#include "pipe/tgsi/core/tgsi_util.h" - - -/** XXX remove */ -#define VERT_RESULT_HPOS 0 -#define VERT_RESULT_MAX 24 - - -static INLINE unsigned -compute_clipmask(float cx, float cy, float cz, float cw) -{ - unsigned mask; -#if defined(macintosh) || defined(__powerpc__) - /* on powerpc cliptest is 17% faster in this way. */ - mask = (((cw < cx) << CLIP_RIGHT_SHIFT)); - mask |= (((cw < -cx) << CLIP_LEFT_SHIFT)); - mask |= (((cw < cy) << CLIP_TOP_SHIFT)); - mask |= (((cw < -cy) << CLIP_BOTTOM_SHIFT)); - mask |= (((cw < cz) << CLIP_FAR_SHIFT)); - mask |= (((cw < -cz) << CLIP_NEAR_SHIFT)); -#else /* !defined(macintosh)) */ - mask = 0x0; - if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT; - if ( cx + cw < 0) mask |= CLIP_LEFT_BIT; - if (-cy + cw < 0) mask |= CLIP_TOP_BIT; - if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT; - if (-cz + cw < 0) mask |= CLIP_FAR_BIT; - if ( cz + cw < 0) mask |= CLIP_NEAR_BIT; -#endif /* defined(macintosh) */ - return mask; -} - - -/** - * Fetch a float[4] vertex attribute from memory, doing format/type - * conversion as needed. - * XXX this might be a temporary thing. - */ -static void -fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) -{ - /* defaults */ - attrib[1] = 0.0; - attrib[2] = 0.0; - attrib[3] = 1.0; - switch (format) { - case PIPE_FORMAT_R32G32B32A32_FLOAT: - attrib[3] = ((float *) ptr)[3]; - /* fall-through */ - case PIPE_FORMAT_R32G32B32_FLOAT: - attrib[2] = ((float *) ptr)[2]; - /* fall-through */ - case PIPE_FORMAT_R32G32_FLOAT: - attrib[1] = ((float *) ptr)[1]; - /* fall-through */ - case PIPE_FORMAT_R32_FLOAT: - attrib[0] = ((float *) ptr)[0]; - break; - default: - assert(0); - } -} - - -/** - * Transform vertices with the current vertex program/shader - * Up to four vertices can be shaded at a time. - * \param vbuffer the input vertex data - * \param elts indexes of four input vertices - * \param count number of vertices to shade [1..4] - * \param vOut array of pointers to four output vertices - */ -static void -run_vertex_program(struct draw_context *draw, - unsigned elts[4], unsigned count, - struct vertex_header *vOut[]) -{ - struct tgsi_exec_machine machine; - unsigned int j; - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - - assert(count <= 4); - -#ifdef DEBUG - memset( &machine, 0, sizeof( machine ) ); -#endif - - /* init machine state */ - tgsi_exec_machine_init(&machine, - draw->vertex_shader.tokens, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/ ); - - /* Consts does not require 16 byte alignment. */ - machine.Consts = draw->vertex_shader.constants->constant; - - machine.Inputs = ALIGN16_ASSIGN(inputs); - machine.Outputs = ALIGN16_ASSIGN(outputs); - - - if (0) - { - unsigned attr; - for (attr = 0; attr < 16; attr++) { - if (draw->vertex_shader.inputs_read & (1 << attr)) { - printf("attr %d: buf_off %d src_off %d pitch %d\n", - attr, - draw->vertex_buffer[attr].buffer_offset, - draw->vertex_element[attr].src_offset, - draw->vertex_buffer[attr].pitch); - } - } - } - - /* load machine inputs */ - for (j = 0; j < count; j++) { - unsigned attr; - for (attr = 0; attr < 16; attr++) { - if (draw->vertex_shader.inputs_read & (1 << attr)) { - unsigned buf = draw->vertex_element[attr].vertex_buffer_index; - const void *src - = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] - + draw->vertex_buffer[buf].buffer_offset - + draw->vertex_element[attr].src_offset - + elts[j] * draw->vertex_buffer[buf].pitch); - float p[4]; - - fetch_attrib4(src, draw->vertex_element[attr].src_format, p); - - machine.Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ - machine.Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ - machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ - machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ -#if 0 - if (attr == 0) { - printf("Input vertex %d: %f %f %f\n", - j, p[0], p[1], p[2]); - } -#endif - } - } - } - -#if 0 - printf("Consts:\n"); - for (i = 0; i < 4; i++) { - printf(" %d: %f %f %f %f\n", i, - machine.Consts[i][0], - machine.Consts[i][1], - machine.Consts[i][2], - machine.Consts[i][3]); - } -#endif - - /* run shader */ - tgsi_exec_machine_run( &machine ); - -#if 0 - printf("VS result: %f %f %f %f\n", - outputs[0].xyzw[0].f[0], - outputs[0].xyzw[1].f[0], - outputs[0].xyzw[2].f[0], - outputs[0].xyzw[3].f[0]); -#endif - - /* store machine results */ - assert(draw->vertex_shader.outputs_written & (1 << VERT_RESULT_HPOS)); - for (j = 0; j < count; j++) { - unsigned attr, slot; - float x, y, z, w; - - /* Handle attr[0] (position) specially: */ - x = vOut[j]->clip[0] = outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = outputs[0].xyzw[3].f[j]; - - vOut[j]->clipmask = compute_clipmask(x, y, z, w); - vOut[j]->edgeflag = 1; - - /* divide by w */ - w = 1.0 / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; -#if 0 - printf("wincoord: %f %f %f %f\n", - vOut[j]->data[0][0], - vOut[j]->data[0][1], - vOut[j]->data[0][2], - vOut[j]->data[0][3]); -#endif - - /* remaining attributes: */ - /* pack into sequential post-transform attrib slots */ - slot = 1; - for (attr = 1; attr < VERT_RESULT_MAX; attr++) { - if (draw->vertex_shader.outputs_written & (1 << attr)) { - assert(slot < draw->nr_attrs); - vOut[j]->data[slot][0] = outputs[attr].xyzw[0].f[j]; - vOut[j]->data[slot][1] = outputs[attr].xyzw[1].f[j]; - vOut[j]->data[slot][2] = outputs[attr].xyzw[2].f[j]; - vOut[j]->data[slot][3] = outputs[attr].xyzw[3].f[j]; - slot++; - } - } - } - -#if 0 - memcpy( - quad->outputs.color, - &machine.Outputs[1].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); -#endif -} - - -/** - * Called by the draw module when the vertx cache needs to be flushed. - * This involves running the vertex shader. - */ -static void vs_flush( struct draw_context *draw ) -{ - unsigned i, j; - - /* run vertex shader on vertex cache entries, four per invokation */ - for (i = 0; i < draw->vs.queue_nr; i += 4) { - struct vertex_header *dests[4]; - unsigned elts[4]; - int n; - - for (j = 0; j < 4; j++) { - elts[j] = draw->vs.queue[i + j].elt; - dests[j] = draw->vs.queue[i + j].dest; - } - - n = MIN2(4, draw->vs.queue_nr - i); - assert(n > 0); - assert(n <= 4); - - run_vertex_program(draw, elts, n, dests); - } - - draw->vs.queue_nr = 0; -} - - -void draw_set_mapped_vertex_buffer(struct draw_context *draw, - unsigned attr, const void *buffer) -{ - draw->mapped_vbuffer[attr] = buffer; -} - /** * Draw vertex arrays @@ -320,8 +55,6 @@ draw_arrays(struct draw_context *draw, unsigned prim, /* tell drawing pipeline we're beginning drawing */ draw->pipeline.first->begin( draw->pipeline.first ); - draw->vs_flush = vs_flush; - draw_invalidate_vcache( draw ); draw_set_prim( draw, prim ); diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index d296780fd7..fd8a39106d 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -35,6 +35,10 @@ #include "draw_context.h" #include "draw_prim.h" +#include "pipe/tgsi/core/tgsi_exec.h" +#include "pipe/tgsi/core/tgsi_build.h" +#include "pipe/tgsi/core/tgsi_util.h" + #define RP_NONE 0 #define RP_POINT 1 @@ -56,6 +60,261 @@ static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { }; + +/** XXX remove */ +#define VERT_RESULT_HPOS 0 +#define VERT_RESULT_MAX 24 + + +static INLINE unsigned +compute_clipmask(float cx, float cy, float cz, float cw) +{ + unsigned mask; +#if defined(macintosh) || defined(__powerpc__) + /* on powerpc cliptest is 17% faster in this way. */ + mask = (((cw < cx) << CLIP_RIGHT_SHIFT)); + mask |= (((cw < -cx) << CLIP_LEFT_SHIFT)); + mask |= (((cw < cy) << CLIP_TOP_SHIFT)); + mask |= (((cw < -cy) << CLIP_BOTTOM_SHIFT)); + mask |= (((cw < cz) << CLIP_FAR_SHIFT)); + mask |= (((cw < -cz) << CLIP_NEAR_SHIFT)); +#else /* !defined(macintosh)) */ + mask = 0x0; + if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT; + if ( cx + cw < 0) mask |= CLIP_LEFT_BIT; + if (-cy + cw < 0) mask |= CLIP_TOP_BIT; + if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT; + if (-cz + cw < 0) mask |= CLIP_FAR_BIT; + if ( cz + cw < 0) mask |= CLIP_NEAR_BIT; +#endif /* defined(macintosh) */ + return mask; +} + + +/** + * Fetch a float[4] vertex attribute from memory, doing format/type + * conversion as needed. + * XXX this might be a temporary thing. + */ +static void +fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) +{ + /* defaults */ + attrib[1] = 0.0; + attrib[2] = 0.0; + attrib[3] = 1.0; + switch (format) { + case PIPE_FORMAT_R32G32B32A32_FLOAT: + attrib[3] = ((float *) ptr)[3]; + /* fall-through */ + case PIPE_FORMAT_R32G32B32_FLOAT: + attrib[2] = ((float *) ptr)[2]; + /* fall-through */ + case PIPE_FORMAT_R32G32_FLOAT: + attrib[1] = ((float *) ptr)[1]; + /* fall-through */ + case PIPE_FORMAT_R32_FLOAT: + attrib[0] = ((float *) ptr)[0]; + break; + default: + assert(0); + } +} + + +/** + * Transform vertices with the current vertex program/shader + * Up to four vertices can be shaded at a time. + * \param vbuffer the input vertex data + * \param elts indexes of four input vertices + * \param count number of vertices to shade [1..4] + * \param vOut array of pointers to four output vertices + */ +static void +run_vertex_program(struct draw_context *draw, + unsigned elts[4], unsigned count, + struct vertex_header *vOut[]) +{ + struct tgsi_exec_machine machine; + unsigned int j; + + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); + +#ifdef DEBUG + memset( &machine, 0, sizeof( machine ) ); +#endif + + /* init machine state */ + tgsi_exec_machine_init(&machine, + draw->vertex_shader.tokens, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/ ); + + /* Consts does not require 16 byte alignment. */ + machine.Consts = draw->vertex_shader.constants->constant; + + machine.Inputs = ALIGN16_ASSIGN(inputs); + machine.Outputs = ALIGN16_ASSIGN(outputs); + + + if (0) + { + unsigned attr; + for (attr = 0; attr < 16; attr++) { + if (draw->vertex_shader.inputs_read & (1 << attr)) { + printf("attr %d: buf_off %d src_off %d pitch %d\n", + attr, + draw->vertex_buffer[attr].buffer_offset, + draw->vertex_element[attr].src_offset, + draw->vertex_buffer[attr].pitch); + } + } + } + + /* load machine inputs */ + for (j = 0; j < count; j++) { + unsigned attr; + for (attr = 0; attr < 16; attr++) { + if (draw->vertex_shader.inputs_read & (1 << attr)) { + unsigned buf = draw->vertex_element[attr].vertex_buffer_index; + const void *src + = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[attr].src_offset + + elts[j] * draw->vertex_buffer[buf].pitch); + float p[4]; + + fetch_attrib4(src, draw->vertex_element[attr].src_format, p); + + machine.Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ + machine.Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ + machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ + machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ +#if 0 + if (attr == 0) { + printf("Input vertex %d: %f %f %f\n", + j, p[0], p[1], p[2]); + } +#endif + } + } + } + +#if 0 + printf("Consts:\n"); + for (i = 0; i < 4; i++) { + printf(" %d: %f %f %f %f\n", i, + machine.Consts[i][0], + machine.Consts[i][1], + machine.Consts[i][2], + machine.Consts[i][3]); + } +#endif + + /* run shader */ + tgsi_exec_machine_run( &machine ); + +#if 0 + printf("VS result: %f %f %f %f\n", + outputs[0].xyzw[0].f[0], + outputs[0].xyzw[1].f[0], + outputs[0].xyzw[2].f[0], + outputs[0].xyzw[3].f[0]); +#endif + + /* store machine results */ + assert(draw->vertex_shader.outputs_written & (1 << VERT_RESULT_HPOS)); + for (j = 0; j < count; j++) { + unsigned attr, slot; + float x, y, z, w; + + /* Handle attr[0] (position) specially: */ + x = vOut[j]->clip[0] = outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(x, y, z, w); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0 / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; +#if 0 + printf("wincoord: %f %f %f %f\n", + vOut[j]->data[0][0], + vOut[j]->data[0][1], + vOut[j]->data[0][2], + vOut[j]->data[0][3]); +#endif + + /* remaining attributes: */ + /* pack into sequential post-transform attrib slots */ + slot = 1; + for (attr = 1; attr < VERT_RESULT_MAX; attr++) { + if (draw->vertex_shader.outputs_written & (1 << attr)) { + assert(slot < draw->nr_attrs); + vOut[j]->data[slot][0] = outputs[attr].xyzw[0].f[j]; + vOut[j]->data[slot][1] = outputs[attr].xyzw[1].f[j]; + vOut[j]->data[slot][2] = outputs[attr].xyzw[2].f[j]; + vOut[j]->data[slot][3] = outputs[attr].xyzw[3].f[j]; + slot++; + } + } + } + +#if 0 + memcpy( + quad->outputs.color, + &machine.Outputs[1].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); +#endif +} + + +/** + * Called by the draw module when the vertx cache needs to be flushed. + * This involves running the vertex shader. + */ +static void transform_vertices( struct draw_context *draw ) +{ + unsigned i, j; + + /* run vertex shader on vertex cache entries, four per invokation */ + for (i = 0; i < draw->vs.queue_nr; i += 4) { + struct vertex_header *dests[4]; + unsigned elts[4]; + int n; + + for (j = 0; j < 4; j++) { + elts[j] = draw->vs.queue[i + j].elt; + dests[j] = draw->vs.queue[i + j].dest; + } + + n = MIN2(4, draw->vs.queue_nr - i); + assert(n > 0); + assert(n <= 4); + + run_vertex_program(draw, elts, n, dests); + } + + draw->vs.queue_nr = 0; +} + + void draw_flush( struct draw_context *draw ) { struct draw_stage *first = draw->pipeline.first; @@ -63,9 +322,7 @@ void draw_flush( struct draw_context *draw ) /* Make sure all vertices are available: */ - assert(draw->vs_flush); - draw->vs_flush( draw ); - + transform_vertices(draw); switch (draw->reduced_prim) { case RP_TRI: @@ -459,6 +716,16 @@ draw_set_mapped_element_buffer( struct draw_context *draw, } +/** + * Tell drawing context where to find mapped vertex buffers. + */ +void draw_set_mapped_vertex_buffer(struct draw_context *draw, + unsigned attr, const void *buffer) +{ + draw->mapped_vbuffer[attr] = buffer; +} + + unsigned draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) { diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 1912607b66..85feb14688 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -186,8 +186,6 @@ struct draw_context unsigned prim; /**< current prim type: PIPE_PRIM_x */ unsigned reduced_prim; - void (*vs_flush)( struct draw_context *draw ); - struct vertex_header *(*get_vertex)( struct draw_context *draw, unsigned i ); -- cgit v1.2.3 From d640198b2d52c104c707522e79d53a36f708ccd0 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 20 Aug 2007 21:45:14 +0100 Subject: Set viewport state so viewport matches window dims. trivial/readpixels.c works again. --- src/mesa/state_tracker/st_cb_clear.c | 46 +++++++++++++++++++++---------- src/mesa/state_tracker/st_cb_drawpixels.c | 17 ++++++++++++ 2 files changed, 49 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 1ed48facd1..2f2d11c014 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -265,6 +265,7 @@ clear_with_quad(GLcontext *ctx, GLboolean color, GLboolean depth, GLboolean stencil) { struct st_context *st = ctx->st; + struct pipe_context *pipe = ctx->st->pipe; const GLfloat x0 = ctx->DrawBuffer->_Xmin; const GLfloat y0 = ctx->DrawBuffer->_Ymin; const GLfloat x1 = ctx->DrawBuffer->_Xmax; @@ -274,7 +275,7 @@ clear_with_quad(GLcontext *ctx, { struct pipe_alpha_test_state alpha_test; memset(&alpha_test, 0, sizeof(alpha_test)); - st->pipe->set_alpha_test_state(st->pipe, &alpha_test); + pipe->set_alpha_test_state(pipe, &alpha_test); } /* blend state: RGBA masking */ @@ -293,7 +294,7 @@ clear_with_quad(GLcontext *ctx, if (st->ctx->Color.DitherFlag) blend.dither = 1; } - st->pipe->set_blend_state(st->pipe, &blend); + pipe->set_blend_state(pipe, &blend); } /* depth state: always pass */ @@ -305,7 +306,7 @@ clear_with_quad(GLcontext *ctx, depth_test.writemask = 1; depth_test.func = PIPE_FUNC_ALWAYS; } - st->pipe->set_depth_state(st->pipe, &depth_test); + pipe->set_depth_state(pipe, &depth_test); } /* setup state: nothing */ @@ -319,7 +320,7 @@ clear_with_quad(GLcontext *ctx, if (ctx->Scissor.Enabled) setup.scissor = 1; #endif - st->pipe->set_setup_state(st->pipe, &setup); + pipe->set_setup_state(pipe, &setup); } /* stencil state: always set to ref value */ @@ -336,7 +337,7 @@ clear_with_quad(GLcontext *ctx, stencil_test.value_mask[0] = 0xff; stencil_test.write_mask[0] = ctx->Stencil.WriteMask[0] & 0xff; } - st->pipe->set_stencil_state(st->pipe, &stencil_test); + pipe->set_stencil_state(pipe, &stencil_test); } /* fragment shader state: color pass-through program */ @@ -350,7 +351,7 @@ clear_with_quad(GLcontext *ctx, fs.inputs_read = stfp->Base.Base.InputsRead; fs.tokens = &stfp->tokens[0]; fs.constants = NULL; - st->pipe->set_fs_state(st->pipe, &fs); + pipe->set_fs_state(pipe, &fs); } /* vertex shader state: color/position pass-through */ @@ -365,20 +366,37 @@ clear_with_quad(GLcontext *ctx, vs.outputs_written = stvp->Base.Base.OutputsWritten; vs.tokens = &stvp->tokens[0]; vs.constants = NULL; - st->pipe->set_vs_state(st->pipe, &vs); + pipe->set_vs_state(pipe, &vs); + } + + /* viewport state: viewport matching window dims */ + { + const float width = ctx->DrawBuffer->Width; + const float height = ctx->DrawBuffer->Height; + struct pipe_viewport_state vp; + vp.scale[0] = 0.5 * width; + vp.scale[1] = -0.5 * height; + vp.scale[2] = 0.5; + vp.scale[3] = 1.0; + vp.translate[0] = 0.5 * width; + vp.translate[1] = 0.5 * height; + vp.translate[2] = 0.5; + vp.translate[3] = 0.0; + pipe->set_viewport_state(pipe, &vp); } /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); /* Restore pipe state */ - st->pipe->set_alpha_test_state(st->pipe, &st->state.alpha_test); - st->pipe->set_blend_state(st->pipe, &st->state.blend); - st->pipe->set_depth_state(st->pipe, &st->state.depth); - st->pipe->set_fs_state(st->pipe, &st->state.fs); - st->pipe->set_vs_state(st->pipe, &st->state.vs); - st->pipe->set_setup_state(st->pipe, &st->state.setup); - st->pipe->set_stencil_state(st->pipe, &st->state.stencil); + pipe->set_alpha_test_state(pipe, &st->state.alpha_test); + pipe->set_blend_state(pipe, &st->state.blend); + pipe->set_depth_state(pipe, &st->state.depth); + pipe->set_fs_state(pipe, &st->state.fs); + pipe->set_vs_state(pipe, &st->state.vs); + pipe->set_setup_state(pipe, &st->state.setup); + pipe->set_stencil_state(pipe, &st->state.stencil); + pipe->set_viewport_state(pipe, &ctx->st->state.viewport); /* OR: st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); */ diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 108304f5e2..d3f060e691 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -364,6 +364,22 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->set_sampler_state(pipe, unit, &sampler); } + /* viewport state: viewport matching window dims */ + { + const float width = ctx->DrawBuffer->Width; + const float height = ctx->DrawBuffer->Height; + struct pipe_viewport_state vp; + vp.scale[0] = 0.5 * width; + vp.scale[1] = -0.5 * height; + vp.scale[2] = 0.5; + vp.scale[3] = 1.0; + vp.translate[0] = 0.5 * width; + vp.translate[1] = 0.5 * height; + vp.translate[2] = 0.5; + vp.translate[3] = 0.0; + pipe->set_viewport_state(pipe, &vp); + } + /* mipmap tree state: */ { mt = make_mipmap_tree(ctx->st, width, height, format, type, @@ -389,6 +405,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->set_vs_state(pipe, &ctx->st->state.vs); pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); pipe->set_sampler_state(pipe, unit, &ctx->st->state.sampler[unit]); + pipe->set_viewport_state(pipe, &ctx->st->state.viewport); free_mipmap_tree(pipe, mt); } -- cgit v1.2.3 From 3af1f3b9220733f5e3a76fe38fbc397974678234 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 21 Aug 2007 16:24:38 -0600 Subject: Initial check-in of i915 fragment program translation (from tgsi). --- src/mesa/pipe/i915simple/i915_fpc.c | 183 ++++++ src/mesa/pipe/i915simple/i915_fpc.h | 339 +++++++++++ src/mesa/pipe/i915simple/i915_fpc_debug.c | 346 +++++++++++ src/mesa/pipe/i915simple/i915_fpc_emit.c | 430 +++++++++++++ src/mesa/pipe/i915simple/i915_fpc_translate.c | 838 ++++++++++++++++++++++++++ 5 files changed, 2136 insertions(+) create mode 100644 src/mesa/pipe/i915simple/i915_fpc.c create mode 100644 src/mesa/pipe/i915simple/i915_fpc.h create mode 100644 src/mesa/pipe/i915simple/i915_fpc_debug.c create mode 100644 src/mesa/pipe/i915simple/i915_fpc_emit.c create mode 100644 src/mesa/pipe/i915simple/i915_fpc_translate.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc.c b/src/mesa/pipe/i915simple/i915_fpc.c new file mode 100644 index 0000000000..fd0bbbc482 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_fpc.c @@ -0,0 +1,183 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#if 0 +#include + +#include "glheader.h" +#include "macros.h" +#include "enums.h" +#endif + +#include "i915_fpc.h" + + + +void +i915_program_error(struct i915_fp_compile *p, const char *msg) +{ + fprintf(stderr, "i915_program_error: %s", msg); + p->fp->error = 1; +} + + +static struct i915_fp_compile * +i915_init_compile(struct i915_context *i915, struct i915_fragment_program *fp) +{ + struct i915_fp_compile *p = CALLOC_STRUCT(i915_fp_compile); + + p->fp = fp; +#if 0 + p->env_param = NULL; /*i915->intel.ctx.FragmentProgram.Parameters;*/ +#endif + p->constants = i915->fs.constants; + p->nr_tex_indirect = 1; /* correct? */ + p->nr_tex_insn = 0; + p->nr_alu_insn = 0; + p->nr_decl_insn = 0; + + memset(p->constant_flags, 0, sizeof(p->constant_flags)); + + p->csr = p->program; + p->decl = p->declarations; + p->decl_s = 0; + p->decl_t = 0; + p->temp_flag = 0xffff000; + p->utemp_flag = ~0x7; + +#if 0 + p->fp->translated = 0; + p->fp->error = 0; + p->fp->nr_constants = 0; +#endif + p->fp->wpos_tex = -1; + p->fp->nr_params = 0; + + *(p->decl++) = _3DSTATE_PIXEL_SHADER_PROGRAM; + + return p; +} + +/* Copy compile results to the fragment program struct and destroy the + * compilation context. + */ +static void +i915_fini_compile(struct i915_fp_compile *p) +{ + uint program_size = p->csr - p->program; + uint decl_size = p->decl - p->declarations; + + if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT) + i915_program_error(p, "Exceeded max nr indirect texture lookups"); + + if (p->nr_tex_insn > I915_MAX_TEX_INSN) + i915_program_error(p, "Exceeded max TEX instructions"); + + if (p->nr_alu_insn > I915_MAX_ALU_INSN) + i915_program_error(p, "Exceeded max ALU instructions"); + + if (p->nr_decl_insn > I915_MAX_DECL_INSN) + i915_program_error(p, "Exceeded max DECL instructions"); + + if (p->fp->error) { + p->fp->NumNativeInstructions = 0; + p->fp->NumNativeAluInstructions = 0; + p->fp->NumNativeTexInstructions = 0; + p->fp->NumNativeTexIndirections = 0; + return; + } + else { + p->fp->NumNativeInstructions = (p->nr_alu_insn + + p->nr_tex_insn + + p->nr_decl_insn); + p->fp->NumNativeAluInstructions = p->nr_alu_insn; + p->fp->NumNativeTexInstructions = p->nr_tex_insn; + p->fp->NumNativeTexIndirections = p->nr_tex_indirect; + } + + p->declarations[0] |= program_size + decl_size - 2; + + /* Copy compilation results to fragment program struct: + */ + memcpy(p->fp->program, + p->declarations, + decl_size * sizeof(uint)); + + memcpy(p->fp->program + decl_size, + p->program, + program_size * sizeof(uint)); + + p->fp->program_size = program_size + decl_size; + + /* Release the compilation struct: + */ + free(p); +} + + +/** + * Find an unused texture coordinate slot to use for fragment WPOS. + * Update p->fp->wpos_tex with the result (-1 if no used texcoord slot is found). + */ +static void +find_wpos_space(struct i915_fp_compile *p) +{ + const uint inputs = p->shader->inputs_read; + uint i; + + p->fp->wpos_tex = -1; + + if (inputs & FRAG_BIT_WPOS) { + for (i = 0; i < I915_TEX_UNITS; i++) { + if ((inputs & (FRAG_BIT_TEX0 << i)) == 0) { + p->fp->wpos_tex = i; + return; + } + } + + i915_program_error(p, "No free texcoord for wpos value"); + } +} + + + +void i915_compile_fragment_program( struct i915_context *i915, + struct i915_fragment_program *fp ) +{ + struct i915_fp_compile *p = i915_init_compile(i915, fp); + struct tgsi_token *tokens = i915->fs.tokens; + + find_wpos_space(p); + + i915_translate_program(p, tokens); + i915_fixup_depth_write(p); + + i915_fini_compile(p); +#if 0 + fp->translated = 1; +#endif +} diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h new file mode 100644 index 0000000000..0a8bffcd9a --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -0,0 +1,339 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef I915_FPC_H +#define I915_FPC_H + +#include "pipe/p_util.h" + +#include "i915_context.h" +#include "i915_reg.h" + + + +#define I915_PROGRAM_SIZE 192 +#define I915_MAX_CONSTANT 32 + +#define MAX_VARYING 8 + +enum +{ + FRAG_ATTRIB_WPOS = 0, + FRAG_ATTRIB_COL0 = 1, + FRAG_ATTRIB_COL1 = 2, + FRAG_ATTRIB_FOGC = 3, + FRAG_ATTRIB_TEX0 = 4, + FRAG_ATTRIB_TEX1 = 5, + FRAG_ATTRIB_TEX2 = 6, + FRAG_ATTRIB_TEX3 = 7, + FRAG_ATTRIB_TEX4 = 8, + FRAG_ATTRIB_TEX5 = 9, + FRAG_ATTRIB_TEX6 = 10, + FRAG_ATTRIB_TEX7 = 11, + FRAG_ATTRIB_VAR0 = 12, /**< shader varying */ + FRAG_ATTRIB_MAX = (FRAG_ATTRIB_VAR0 + MAX_VARYING) +}; + +/** + * Bitflags for fragment program input attributes. + */ +/*@{*/ +#define FRAG_BIT_WPOS (1 << FRAG_ATTRIB_WPOS) +#define FRAG_BIT_COL0 (1 << FRAG_ATTRIB_COL0) +#define FRAG_BIT_COL1 (1 << FRAG_ATTRIB_COL1) +#define FRAG_BIT_FOGC (1 << FRAG_ATTRIB_FOGC) +#define FRAG_BIT_TEX0 (1 << FRAG_ATTRIB_TEX0) +#define FRAG_BIT_TEX1 (1 << FRAG_ATTRIB_TEX1) +#define FRAG_BIT_TEX2 (1 << FRAG_ATTRIB_TEX2) +#define FRAG_BIT_TEX3 (1 << FRAG_ATTRIB_TEX3) +#define FRAG_BIT_TEX4 (1 << FRAG_ATTRIB_TEX4) +#define FRAG_BIT_TEX5 (1 << FRAG_ATTRIB_TEX5) +#define FRAG_BIT_TEX6 (1 << FRAG_ATTRIB_TEX6) +#define FRAG_BIT_TEX7 (1 << FRAG_ATTRIB_TEX7) +#define FRAG_BIT_VAR0 (1 << FRAG_ATTRIB_VAR0) + +#define MAX_DRAW_BUFFERS 4 + +enum +{ + FRAG_RESULT_COLR = 0, + FRAG_RESULT_COLH = 1, + FRAG_RESULT_DEPR = 2, + FRAG_RESULT_DATA0 = 3, + FRAG_RESULT_MAX = (FRAG_RESULT_DATA0 + MAX_DRAW_BUFFERS) +}; + + + +#if 1 /*XXX temp */ +/* Hardware version of a parsed fragment program. "Derived" from the + * mesa fragment_program struct. + */ +struct i915_fragment_program +{ +#if 0 + struct gl_fragment_program Base; +#else + uint NumNativeInstructions; + uint NumNativeAluInstructions; + uint NumNativeTexInstructions; + uint NumNativeTexIndirections; +#endif + + boolean error; /**< Set if i915_program_error() is called */ +#if 0 + uint id; /**< String id */ + boolean translated; +#endif + + /* Decls + instructions: + */ + uint program[I915_PROGRAM_SIZE]; + uint program_size; + +#if 0 + /* Constant buffer: + */ + float constant[I915_MAX_CONSTANT][4]; + uint nr_constants; +#endif + + /* Some of which are parameters: + */ + struct + { + uint reg; /* Hardware constant idx */ + const float *values; /* Pointer to tracked values */ + } param[I915_MAX_CONSTANT]; + uint nr_params; + +#if 0 + uint param_state; +#endif + uint wpos_tex; +}; +#endif + + +/*********************************************************************** + * Public interface for the compiler + */ + +void i915_compile_fragment_program( struct i915_context *i915, + struct i915_fragment_program *fp ); + + +/*********************************************************************** + * Private details of the compiler + */ + +struct i915_fp_compile { + struct i915_fragment_program *fp; + + struct pipe_shader_state *shader; + + uint declarations[I915_PROGRAM_SIZE]; + uint program[I915_PROGRAM_SIZE]; + + uint constant_flags[I915_MAX_CONSTANT]; + + struct pipe_constant_buffer *constants; + + uint *csr; /* Cursor, points into program. + */ + + uint *decl; /* Cursor, points into declarations. + */ + + uint decl_s; /* flags for which s regs need to be decl'd */ + uint decl_t; /* flags for which t regs need to be decl'd */ + + uint temp_flag; /* Tracks temporary regs which are in + * use. + */ + + uint utemp_flag; /* Tracks TYPE_U temporary regs which are in + * use. + */ + + uint nr_tex_indirect; + uint nr_tex_insn; + uint nr_alu_insn; + uint nr_decl_insn; + +#if 0 + float (*env_param)[4]; +#endif +}; + + +/* Having zero and one in here makes the definition of swizzle a lot + * easier. + */ +#define UREG_TYPE_SHIFT 29 +#define UREG_NR_SHIFT 24 +#define UREG_CHANNEL_X_NEGATE_SHIFT 23 +#define UREG_CHANNEL_X_SHIFT 20 +#define UREG_CHANNEL_Y_NEGATE_SHIFT 19 +#define UREG_CHANNEL_Y_SHIFT 16 +#define UREG_CHANNEL_Z_NEGATE_SHIFT 15 +#define UREG_CHANNEL_Z_SHIFT 12 +#define UREG_CHANNEL_W_NEGATE_SHIFT 11 +#define UREG_CHANNEL_W_SHIFT 8 +#define UREG_CHANNEL_ZERO_NEGATE_MBZ 5 +#define UREG_CHANNEL_ZERO_SHIFT 4 +#define UREG_CHANNEL_ONE_NEGATE_MBZ 1 +#define UREG_CHANNEL_ONE_SHIFT 0 + +#define UREG_BAD 0xffffffff /* not a valid ureg */ + +#define X SRC_X +#define Y SRC_Y +#define Z SRC_Z +#define W SRC_W +#define ZERO SRC_ZERO +#define ONE SRC_ONE + +/* Construct a ureg: + */ +#define UREG( type, nr ) (((type)<< UREG_TYPE_SHIFT) | \ + ((nr) << UREG_NR_SHIFT) | \ + (X << UREG_CHANNEL_X_SHIFT) | \ + (Y << UREG_CHANNEL_Y_SHIFT) | \ + (Z << UREG_CHANNEL_Z_SHIFT) | \ + (W << UREG_CHANNEL_W_SHIFT) | \ + (ZERO << UREG_CHANNEL_ZERO_SHIFT) | \ + (ONE << UREG_CHANNEL_ONE_SHIFT)) + +#define GET_CHANNEL_SRC( reg, channel ) ((reg<<(channel*4)) & (0xf<<20)) +#define CHANNEL_SRC( src, channel ) (src>>(channel*4)) + +#define GET_UREG_TYPE(reg) (((reg)>>UREG_TYPE_SHIFT)®_TYPE_MASK) +#define GET_UREG_NR(reg) (((reg)>>UREG_NR_SHIFT)®_NR_MASK) + + + +#define UREG_XYZW_CHANNEL_MASK 0x00ffff00 + +/* One neat thing about the UREG representation: + */ +static INLINE int +swizzle(int reg, int x, int y, int z, int w) +{ + assert(x < 4); + assert(y < 4); + assert(z < 4); + assert(w < 4); + return ((reg & ~UREG_XYZW_CHANNEL_MASK) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, x), 0) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, y), 1) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, z), 2) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, w), 3)); +} + +/* Another neat thing about the UREG representation: + */ +static INLINE int +negate(int reg, int x, int y, int z, int w) +{ + return reg ^ (((x & 1) << UREG_CHANNEL_X_NEGATE_SHIFT) | + ((y & 1) << UREG_CHANNEL_Y_NEGATE_SHIFT) | + ((z & 1) << UREG_CHANNEL_Z_NEGATE_SHIFT) | + ((w & 1) << UREG_CHANNEL_W_NEGATE_SHIFT)); +} + + +extern uint i915_get_temp(struct i915_fp_compile *p); +extern uint i915_get_utemp(struct i915_fp_compile *p); +extern void i915_release_utemps(struct i915_fp_compile *p); + + +extern uint i915_emit_texld(struct i915_fp_compile *p, + uint dest, + uint destmask, + uint sampler, uint coord, uint op); + +extern uint i915_emit_arith(struct i915_fp_compile *p, + uint op, + uint dest, + uint mask, + uint saturate, + uint src0, uint src1, uint src2); + +extern uint i915_emit_decl(struct i915_fp_compile *p, + uint type, uint nr, uint d0_flags); + + +extern uint i915_emit_const1f(struct i915_fp_compile *p, float c0); + +extern uint i915_emit_const2f(struct i915_fp_compile *p, + float c0, float c1); + +extern uint i915_emit_const4fv(struct i915_fp_compile *p, + const float * c); + +extern uint i915_emit_const4f(struct i915_fp_compile *p, + float c0, float c1, + float c2, float c3); + + +#if 0 +extern uint i915_emit_param4fv(struct i915_fp_compile *p, + const float * values); +#endif + + + +/*====================================================================== + * i915_fpc_debug.c + */ +extern void i915_program_error(struct i915_fp_compile *p, + const char *msg); + + +/*====================================================================== + * i915_fpc_debug.c + */ +extern void i915_disassemble_program(const uint * program, uint sz); + +#if 0 +extern void i915_print_mesa_instructions( const struct prog_instruction *insn, + uint nr ); +#endif + +/*====================================================================== + * i915_fpc_translate.c + */ +void i915_fixup_depth_write(struct i915_fp_compile *p); + +extern void +i915_translate_program(struct i915_fp_compile *p, const struct tgsi_token *token); + + + +#endif diff --git a/src/mesa/pipe/i915simple/i915_fpc_debug.c b/src/mesa/pipe/i915simple/i915_fpc_debug.c new file mode 100644 index 0000000000..77deab38bb --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_fpc_debug.c @@ -0,0 +1,346 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#if 0 +#include +#endif + +#include "i915_reg.h" +#include "i915_fpc.h" + +#if 0 +#include "shader/program.h" +#include "shader/prog_instruction.h" +#include "shader/prog_print.h" +#endif + +static const char *opcodes[0x20] = { + "NOP", + "ADD", + "MOV", + "MUL", + "MAD", + "DP2ADD", + "DP3", + "DP4", + "FRC", + "RCP", + "RSQ", + "EXP", + "LOG", + "CMP", + "MIN", + "MAX", + "FLR", + "MOD", + "TRC", + "SGE", + "SLT", + "TEXLD", + "TEXLDP", + "TEXLDB", + "TEXKILL", + "DCL", + "0x1a", + "0x1b", + "0x1c", + "0x1d", + "0x1e", + "0x1f", +}; + + +static const int args[0x20] = { + 0, /* 0 nop */ + 2, /* 1 add */ + 1, /* 2 mov */ + 2, /* 3 m ul */ + 3, /* 4 mad */ + 3, /* 5 dp2add */ + 2, /* 6 dp3 */ + 2, /* 7 dp4 */ + 1, /* 8 frc */ + 1, /* 9 rcp */ + 1, /* a rsq */ + 1, /* b exp */ + 1, /* c log */ + 3, /* d cmp */ + 2, /* e min */ + 2, /* f max */ + 1, /* 10 flr */ + 1, /* 11 mod */ + 1, /* 12 trc */ + 2, /* 13 sge */ + 2, /* 14 slt */ + 1, + 1, + 1, + 1, + 0, + 0, + 0, + 0, + 0, + 0, + 0, +}; + + +static const char *regname[0x8] = { + "R", + "T", + "CONST", + "S", + "OC", + "OD", + "U", + "UNKNOWN", +}; + +static void +print_reg_type_nr(uint type, uint nr) +{ + switch (type) { + case REG_TYPE_T: + switch (nr) { + case T_DIFFUSE: + printf("T_DIFFUSE"); + return; + case T_SPECULAR: + printf("T_SPECULAR"); + return; + case T_FOG_W: + printf("T_FOG_W"); + return; + default: + printf("T_TEX%d", nr); + return; + } + case REG_TYPE_OC: + if (nr == 0) { + printf("oC"); + return; + } + break; + case REG_TYPE_OD: + if (nr == 0) { + printf("oD"); + return; + } + break; + default: + break; + } + + printf("%s[%d]", regname[type], nr); +} + +#define REG_SWIZZLE_MASK 0x7777 +#define REG_NEGATE_MASK 0x8888 + +#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ + (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ + (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ + (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) + + +static void +print_reg_neg_swizzle(uint reg) +{ + int i; + + if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && + (reg & REG_NEGATE_MASK) == 0) + return; + + printf("."); + + for (i = 3; i >= 0; i--) { + if (reg & (1 << ((i * 4) + 3))) + printf("-"); + + switch ((reg >> (i * 4)) & 0x7) { + case 0: + printf("x"); + break; + case 1: + printf("y"); + break; + case 2: + printf("z"); + break; + case 3: + printf("w"); + break; + case 4: + printf("0"); + break; + case 5: + printf("1"); + break; + default: + printf("?"); + break; + } + } +} + + +static void +print_src_reg(uint dword) +{ + uint nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; + uint type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(type, nr); + print_reg_neg_swizzle(dword); +} + + +static void +print_dest_reg(uint dword) +{ + uint nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; + uint type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(type, nr); + if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) + return; + printf("."); + if (dword & A0_DEST_CHANNEL_X) + printf("x"); + if (dword & A0_DEST_CHANNEL_Y) + printf("y"); + if (dword & A0_DEST_CHANNEL_Z) + printf("z"); + if (dword & A0_DEST_CHANNEL_W) + printf("w"); +} + + +#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) +#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) +#define GET_SRC2_REG(r) (r) + + +static void +print_arith_op(uint opcode, const uint * program) +{ + if (opcode != A0_NOP) { + print_dest_reg(program[0]); + if (program[0] & A0_DEST_SATURATE) + printf(" = SATURATE "); + else + printf(" = "); + } + + printf("%s ", opcodes[opcode]); + + print_src_reg(GET_SRC0_REG(program[0], program[1])); + if (args[opcode] == 1) { + printf("\n"); + return; + } + + printf(", "); + print_src_reg(GET_SRC1_REG(program[1], program[2])); + if (args[opcode] == 2) { + printf("\n"); + return; + } + + printf(", "); + print_src_reg(GET_SRC2_REG(program[2])); + printf("\n"); + return; +} + + +static void +print_tex_op(uint opcode, const uint * program) +{ + print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); + printf(" = "); + + printf("%s ", opcodes[opcode]); + + printf("S[%d],", program[0] & T0_SAMPLER_NR_MASK); + + print_reg_type_nr((program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & + REG_TYPE_MASK, + (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); + printf("\n"); +} + +static void +print_dcl_op(uint opcode, const uint * program) +{ + printf("%s ", opcodes[opcode]); + print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); + printf("\n"); +} + + +void +i915_disassemble_program(const uint * program, uint sz) +{ + uint size = program[0] & 0x1ff; + int i; + + printf("\t\tBEGIN\n"); + + assert(size + 2 == sz); + + program++; + for (i = 1; i < sz; i += 3, program += 3) { + uint opcode = program[0] & (0x1f << 24); + + printf("\t\t"); + + if ((int) opcode >= A0_NOP && opcode <= A0_SLT) + print_arith_op(opcode >> 24, program); + else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) + print_tex_op(opcode >> 24, program); + else if (opcode == D0_DCL) + print_dcl_op(opcode >> 24, program); + else + printf("Unknown opcode 0x%x\n", opcode); + } + + printf("\t\tEND\n\n"); +} + + +#if 0 +void i915_print_mesa_instructions( const struct prog_instruction *insn, + uint nr ) +{ + uint i; + for (i = 0; i < nr; i++, insn++) { + printf("%3d: ", i); + print_instruction(insn); + } +} +#endif diff --git a/src/mesa/pipe/i915simple/i915_fpc_emit.c b/src/mesa/pipe/i915simple/i915_fpc_emit.c new file mode 100644 index 0000000000..7259bb503d --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_fpc_emit.c @@ -0,0 +1,430 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#if 0 +#include +#include "glheader.h" +#include "macros.h" +#include "enums.h" +#endif + +#include "i915_reg.h" +#include "i915_context.h" +#include "i915_fpc.h" + + +#define A0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) +#define D0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) +#define T0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) +#define A0_SRC0( reg ) (((reg)&UREG_MASK)>>UREG_A0_SRC0_SHIFT_LEFT) +#define A1_SRC0( reg ) (((reg)&UREG_MASK)<>UREG_A1_SRC1_SHIFT_LEFT) +#define A2_SRC1( reg ) (((reg)&UREG_MASK)<>UREG_A2_SRC2_SHIFT_LEFT) + +/* These are special, and don't have swizzle/negate bits. + */ +#define T0_SAMPLER( reg ) (GET_UREG_NR(reg)<temp_flag); + if (!bit) { + i915_program_error(p, "i915_get_temp: out of temporaries\n"); + return 0; + } + + p->temp_flag |= 1 << (bit - 1); + return UREG(REG_TYPE_R, (bit - 1)); +} + + +uint +i915_get_utemp(struct i915_fp_compile * p) +{ + int bit = ffs(~p->utemp_flag); + if (!bit) { + i915_program_error(p, "i915_get_utemp: out of temporaries\n"); + return 0; + } + + p->utemp_flag |= 1 << (bit - 1); + return UREG(REG_TYPE_U, (bit - 1)); +} + +void +i915_release_utemps(struct i915_fp_compile *p) +{ + p->utemp_flag = ~0x7; +} + + +uint +i915_emit_decl(struct i915_fp_compile *p, + uint type, uint nr, uint d0_flags) +{ + uint reg = UREG(type, nr); + + if (type == REG_TYPE_T) { + if (p->decl_t & (1 << nr)) + return reg; + + p->decl_t |= (1 << nr); + } + else if (type == REG_TYPE_S) { + if (p->decl_s & (1 << nr)) + return reg; + + p->decl_s |= (1 << nr); + } + else + return reg; + + *(p->decl++) = (D0_DCL | D0_DEST(reg) | d0_flags); + *(p->decl++) = D1_MBZ; + *(p->decl++) = D2_MBZ; + + p->nr_decl_insn++; + return reg; +} + +uint +i915_emit_arith(struct i915_fp_compile * p, + uint op, + uint dest, + uint mask, + uint saturate, uint src0, uint src1, uint src2) +{ + uint c[3]; + uint nr_const = 0; + + assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); + dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest)); + assert(dest); + + if (GET_UREG_TYPE(src0) == REG_TYPE_CONST) + c[nr_const++] = 0; + if (GET_UREG_TYPE(src1) == REG_TYPE_CONST) + c[nr_const++] = 1; + if (GET_UREG_TYPE(src2) == REG_TYPE_CONST) + c[nr_const++] = 2; + + /* Recursively call this function to MOV additional const values + * into temporary registers. Use utemp registers for this - + * currently shouldn't be possible to run out, but keep an eye on + * this. + */ + if (nr_const > 1) { + uint s[3], first, i, old_utemp_flag; + + s[0] = src0; + s[1] = src1; + s[2] = src2; + old_utemp_flag = p->utemp_flag; + + first = GET_UREG_NR(s[c[0]]); + for (i = 1; i < nr_const; i++) { + if (GET_UREG_NR(s[c[i]]) != first) { + uint tmp = i915_get_utemp(p); + + i915_emit_arith(p, A0_MOV, tmp, A0_DEST_CHANNEL_ALL, 0, + s[c[i]], 0, 0); + s[c[i]] = tmp; + } + } + + src0 = s[0]; + src1 = s[1]; + src2 = s[2]; + p->utemp_flag = old_utemp_flag; /* restore */ + } + + *(p->csr++) = (op | A0_DEST(dest) | mask | saturate | A0_SRC0(src0)); + *(p->csr++) = (A1_SRC0(src0) | A1_SRC1(src1)); + *(p->csr++) = (A2_SRC1(src1) | A2_SRC2(src2)); + + p->nr_alu_insn++; + return dest; +} + +uint i915_emit_texld( struct i915_fp_compile *p, + uint dest, + uint destmask, + uint sampler, + uint coord, + uint op ) +{ + if (coord != UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord))) { + /* No real way to work around this in the general case - need to + * allocate and declare a new temporary register (a utemp won't + * do). Will fallback for now. + */ + i915_program_error(p, "Can't (yet) swizzle TEX arguments"); + return 0; + } + + /* Don't worry about saturate as we only support + */ + if (destmask != A0_DEST_CHANNEL_ALL) { + uint tmp = i915_get_utemp(p); + i915_emit_texld( p, tmp, A0_DEST_CHANNEL_ALL, sampler, coord, op ); + i915_emit_arith( p, A0_MOV, dest, destmask, 0, tmp, 0, 0 ); + return dest; + } + else { + assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); + assert(dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest))); + + if (GET_UREG_TYPE(coord) != REG_TYPE_T) { + p->nr_tex_indirect++; + } + + *(p->csr++) = (op | + T0_DEST( dest ) | + T0_SAMPLER( sampler )); + + *(p->csr++) = T1_ADDRESS_REG( coord ); + *(p->csr++) = T2_MBZ; + + p->nr_tex_insn++; + return dest; + } +} + + +uint +i915_emit_const1f(struct i915_fp_compile * p, float c0) +{ + int reg, idx; + + if (c0 == 0.0) + return swizzle(UREG(REG_TYPE_R, 0), ZERO, ZERO, ZERO, ZERO); + if (c0 == 1.0) + return swizzle(UREG(REG_TYPE_R, 0), ONE, ONE, ONE, ONE); + + for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { + if (p->constant_flags[reg] == I915_CONSTFLAG_PARAM) + continue; + for (idx = 0; idx < 4; idx++) { +#if 0 + if (!(p->constant_flags[reg] & (1 << idx)) || + p->fp->constant[reg][idx] == c0) { + p->fp->constant[reg][idx] = c0; + p->constant_flags[reg] |= 1 << idx; + if (reg + 1 > p->fp->nr_constants) + p->fp->nr_constants = reg + 1; + return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE); + } +#else + if (!(p->constant_flags[reg] & (1 << idx)) || + p->constants->constant[reg][idx] == c0) { + p->constants->constant[reg][idx] = c0; + p->constant_flags[reg] |= 1 << idx; + if (reg + 1 > p->constants->nr_constants) + p->constants->nr_constants = reg + 1; + return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE); + } +#endif + } + } + + i915_program_error(p, "i915_emit_const1f: out of constants\n"); + return 0; +} + +uint +i915_emit_const2f(struct i915_fp_compile * p, float c0, float c1) +{ + int reg, idx; + + if (c0 == 0.0) + return swizzle(i915_emit_const1f(p, c1), ZERO, X, Z, W); + if (c0 == 1.0) + return swizzle(i915_emit_const1f(p, c1), ONE, X, Z, W); + + if (c1 == 0.0) + return swizzle(i915_emit_const1f(p, c0), X, ZERO, Z, W); + if (c1 == 1.0) + return swizzle(i915_emit_const1f(p, c0), X, ONE, Z, W); + + for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { + if (p->constant_flags[reg] == 0xf || + p->constant_flags[reg] == I915_CONSTFLAG_PARAM) + continue; + for (idx = 0; idx < 3; idx++) { + if (!(p->constant_flags[reg] & (3 << idx))) { +#if 0 + p->fp->constant[reg][idx] = c0; + p->fp->constant[reg][idx + 1] = c1; + p->constant_flags[reg] |= 3 << idx; + if (reg + 1 > p->fp->nr_constants) + p->fp->nr_constants = reg + 1; + return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, + ONE); +#else + p->constants->constant[reg][idx + 0] = c0; + p->constants->constant[reg][idx + 1] = c1; + p->constant_flags[reg] |= 3 << idx; + if (reg + 1 > p->constants->nr_constants) + p->constants->nr_constants = reg + 1; + return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, + ONE); +#endif + } + } + } + + i915_program_error(p, "i915_emit_const2f: out of constants\n"); + return 0; +} + + + +uint +i915_emit_const4f(struct i915_fp_compile * p, + float c0, float c1, float c2, float c3) +{ + int reg; + + for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { + if (p->constant_flags[reg] == 0xf && +#if 0 + p->fp->constant[reg][0] == c0 && + p->fp->constant[reg][1] == c1 && + p->fp->constant[reg][2] == c2 && + p->fp->constant[reg][3] == c3 +#else + p->constants->constant[reg][0] == c0 && + p->constants->constant[reg][1] == c1 && + p->constants->constant[reg][2] == c2 && + p->constants->constant[reg][3] == c3 +#endif + ) { + return UREG(REG_TYPE_CONST, reg); + } + else if (p->constant_flags[reg] == 0) { +#if 0 + p->fp->constant[reg][0] = c0; + p->fp->constant[reg][1] = c1; + p->fp->constant[reg][2] = c2; + p->fp->constant[reg][3] = c3; +#else + p->constants->constant[reg][0] = c0; + p->constants->constant[reg][1] = c1; + p->constants->constant[reg][2] = c2; + p->constants->constant[reg][3] = c3; +#endif + p->constant_flags[reg] = 0xf; +#if 0 + if (reg + 1 > p->fp->nr_constants) + p->fp->nr_constants = reg + 1; +#else + if (reg + 1 > p->constants->nr_constants) + p->constants->nr_constants = reg + 1; +#endif + return UREG(REG_TYPE_CONST, reg); + } + } + + i915_program_error(p, "i915_emit_const4f: out of constants\n"); + return 0; +} + + +uint +i915_emit_const4fv(struct i915_fp_compile * p, const float * c) +{ + return i915_emit_const4f(p, c[0], c[1], c[2], c[3]); +} + + +#if 00000/*UNUSED*/ +/* Reserve a slot in the constant file for a Mesa state parameter. + * These will later need to be tracked on statechanges, but that is + * done elsewhere. + */ +uint +i915_emit_param4fv(struct i915_fp_compile * p, const float * values) +{ + struct i915_fragment_program *fp = p->fp; + int i; + + for (i = 0; i < fp->nr_params; i++) { + if (fp->param[i].values == values) + return UREG(REG_TYPE_CONST, fp->param[i].reg); + } + +#if 0 + if (fp->nr_constants == I915_MAX_CONSTANT || + fp->nr_params == I915_MAX_CONSTANT) { +#else + if (p->constants->nr_constants == I915_MAX_CONSTANT || + fp->nr_params == I915_MAX_CONSTANT) { +#endif + i915_program_error(p, "i915_emit_param4fv: out of constants\n"); + return 0; + } + + { +#if 0 + int reg = fp->nr_constants++; +#else + int reg = p->constants->nr_constants++; +#endif + int i = fp->nr_params++; + + assert (p->constant_flags[reg] == 0); + p->constant_flags[reg] = I915_CONSTFLAG_PARAM; + + fp->param[i].values = values; + fp->param[i].reg = reg; + + return UREG(REG_TYPE_CONST, reg); + } +} +#endif diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c new file mode 100644 index 0000000000..a034e734c3 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -0,0 +1,838 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_reg.h" +#include "i915_context.h" +#include "i915_fpc.h" + +#include "pipe/tgsi/core/tgsi_parse.h" + + +/* 1, -1/3!, 1/5!, -1/7! */ +static const float sin_constants[4] = { 1.0, + -1.0 / (3 * 2 * 1), + 1.0 / (5 * 4 * 3 * 2 * 1), + -1.0 / (7 * 6 * 5 * 4 * 3 * 2 * 1) +}; + +/* 1, -1/2!, 1/4!, -1/6! */ +static const float cos_constants[4] = { 1.0, + -1.0 / (2 * 1), + 1.0 / (4 * 3 * 2 * 1), + -1.0 / (6 * 5 * 4 * 3 * 2 * 1) +}; + + +/** + * Construct a ureg for the given source register. Will emit + * constants, apply swizzling and negation as needed. + */ +static uint +src_vector(struct i915_fp_compile *p, + const struct tgsi_full_src_register *source) +{ + const uint index = source->SrcRegister.Index; + uint src; + + switch (source->SrcRegisterInd.File) { + case TGSI_FILE_TEMPORARY: + if (source->SrcRegister.Index >= I915_MAX_TEMPORARY) { + i915_program_error(p, "Exceeded max temporary reg"); + return 0; + } + src = UREG(REG_TYPE_R, index); + break; + case TGSI_FILE_INPUT: + /* XXX: Packing COL1, FOGC into a single attribute works for + * texenv programs, but will fail for real fragment programs + * that use these attributes and expect them to be a full 4 + * components wide. Could use a texcoord to pass these + * attributes if necessary, but that won't work in the general + * case. + * + * We also use a texture coordinate to pass wpos when possible. + */ + switch (index) { + case FRAG_ATTRIB_WPOS: + src = i915_emit_decl(p, REG_TYPE_T, p->fp->wpos_tex, D0_CHANNEL_ALL); + break; + case FRAG_ATTRIB_COL0: + src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); + break; + case FRAG_ATTRIB_COL1: + src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); + src = swizzle(src, X, Y, Z, ONE); + break; + case FRAG_ATTRIB_FOGC: + src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); + src = swizzle(src, W, W, W, W); + break; + case FRAG_ATTRIB_TEX0: + case FRAG_ATTRIB_TEX1: + case FRAG_ATTRIB_TEX2: + case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX4: + case FRAG_ATTRIB_TEX5: + case FRAG_ATTRIB_TEX6: + case FRAG_ATTRIB_TEX7: + src = i915_emit_decl(p, REG_TYPE_T, + T_TEX0 + (index - FRAG_ATTRIB_TEX0), + D0_CHANNEL_ALL); + break; + + default: + i915_program_error(p, "Bad source->Index"); + return 0; + } + break; + + /* Various parameters and env values. All emitted to + * hardware as program constants. + */ +#if 0 + case PROGRAM_LOCAL_PARAM: + src = i915_emit_param4fv(p, program->Base.LocalParams[index]); + break; + case PROGRAM_ENV_PARAM: + src = i915_emit_param4fv(p, p->env_param[index]); + break; + case PROGRAM_CONSTANT: + case PROGRAM_STATE_VAR: + case PROGRAM_NAMED_PARAM: + src = i915_emit_param4fv( + p, program->Base.Parameters->ParameterValues[index]); + break; +#else + case TGSI_FILE_CONSTANT: + src = UREG(REG_TYPE_CONST, index); + break; +#endif + + default: + i915_program_error(p, "Bad source->File"); + return 0; + } + + src = swizzle(src, + source->SrcRegister.SwizzleX, + source->SrcRegister.SwizzleY, + source->SrcRegister.SwizzleZ, + source->SrcRegister.SwizzleW); + + assert(!source->SrcRegister.Negate); + assert(!source->SrcRegisterExtSwz.NegateX); + assert(!source->SrcRegisterExtSwz.NegateY); + assert(!source->SrcRegisterExtSwz.NegateZ); + assert(!source->SrcRegisterExtSwz.NegateW); + assert(!source->SrcRegisterExtMod.Absolute); + assert(!source->SrcRegisterExtMod.Negate); +#if 0 + if (source->SrcRegister.Negate) + negate all + + if (extended source swiz per component) + src = negate(src, + source->SrcRegisterExtSwz.NegateX, + source->SrcRegisterExtSwz.NegateY, + source->SrcRegisterExtSwz.NegateZ, + source->SrcRegisterExtSwz.NegateW); + if (mod.abs) + absolute value + + if (mod.negate) + another negate; +#endif + return src; +} + + +static uint +get_result_vector(struct i915_fp_compile *p, + const struct tgsi_full_dst_register *dest) +{ + switch (dest->DstRegister.File) { + case TGSI_FILE_OUTPUT: + switch (dest->DstRegister.Index) { + case FRAG_RESULT_COLR: + return UREG(REG_TYPE_OC, 0); + case FRAG_RESULT_DEPR: + return UREG(REG_TYPE_OD, 0); + default: + i915_program_error(p, "Bad inst->DstReg.Index"); + return 0; + } + case TGSI_FILE_TEMPORARY: + return UREG(REG_TYPE_R, dest->DstRegister.Index); + default: + i915_program_error(p, "Bad inst->DstReg.File"); + return 0; + } +} + + +/** + * Compute flags for saturation and writemask. + */ +static uint +get_result_flags(const struct tgsi_full_instruction *inst) +{ + const uint writeMask + = inst->FullDstRegisters[0].DstRegister.WriteMask; + uint flags = 0x0; + + if (inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE) + flags |= A0_DEST_SATURATE; + + if (writeMask & TGSI_WRITEMASK_X) + flags |= A0_DEST_CHANNEL_X; + if (writeMask & TGSI_WRITEMASK_Y) + flags |= A0_DEST_CHANNEL_Y; + if (writeMask & TGSI_WRITEMASK_Z) + flags |= A0_DEST_CHANNEL_Z; + if (writeMask & TGSI_WRITEMASK_W) + flags |= A0_DEST_CHANNEL_W; + + return flags; +} + + +/** + * Convert TGSI_TEXTURE_x token to DO_SAMPLE_TYPE_x token + */ +static uint +translate_tex_src_target(struct i915_fp_compile *p, uint tex) +{ + switch (tex) { + case TGSI_TEXTURE_1D: + return D0_SAMPLE_TYPE_2D; + case TGSI_TEXTURE_2D: + return D0_SAMPLE_TYPE_2D; + case TGSI_TEXTURE_RECT: + return D0_SAMPLE_TYPE_2D; + case TGSI_TEXTURE_3D: + return D0_SAMPLE_TYPE_VOLUME; + case TGSI_TEXTURE_CUBE: + return D0_SAMPLE_TYPE_CUBE; + default: + i915_program_error(p, "TexSrc type"); + return 0; + } +} + + +/** + * Generate texel lookup instruction. + */ +static void +emit_tex(struct i915_fp_compile *p, + const struct tgsi_full_instruction *inst, + uint opcode) +{ + uint texture = inst->InstructionExtTexture.Texture; + uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + uint tex = translate_tex_src_target( p, texture ); + uint sampler = i915_emit_decl(p, REG_TYPE_S, unit, tex); + uint coord = src_vector( p, &inst->FullSrcRegisters[0]); + + i915_emit_texld( p, + get_result_vector( p, &inst->FullDstRegisters[0] ), + get_result_flags( inst ), + sampler, + coord, + opcode); +} + + +/** + * Generate a simple arithmetic instruction + * \param opcode the i915 opcode + * \param numArgs the number of input/src arguments + */ +static void +emit_simple_arith(struct i915_fp_compile *p, + const struct tgsi_full_instruction *inst, + uint opcode, uint numArgs) +{ + uint arg1, arg2, arg3; + + assert(numArgs <= 3); + + arg1 = (numArgs < 1) ? 0 : src_vector( p, &inst->FullSrcRegisters[0] ); + arg2 = (numArgs < 2) ? 0 : src_vector( p, &inst->FullSrcRegisters[1] ); + arg3 = (numArgs < 3) ? 0 : src_vector( p, &inst->FullSrcRegisters[2] ); + + i915_emit_arith( p, + opcode, + get_result_vector( p, &inst->FullDstRegisters[0]), + get_result_flags( inst ), 0, + arg1, + arg2, + arg3 ); +} + + +#define EMIT_1ARG_ARITH( OP ) emit_simple_arith(p, inst, OP, 1) +#define EMIT_2ARG_ARITH( OP ) emit_simple_arith(p, inst, OP, 2) +#define EMIT_3ARG_ARITH( OP ) emit_simple_arith(p, inst, OP, 3) + + + +static void +i915_translate_instruction(struct i915_fp_compile *p, + const struct tgsi_full_instruction *inst) +{ + uint writemask; + uint src0, src1, src2, flags; + uint tmp = 0; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + i915_emit_arith(p, + A0_MAX, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + src0, negate(src0, 1, 1, 1, 1), 0); + break; + + case TGSI_OPCODE_ADD: + EMIT_2ARG_ARITH(A0_ADD); + break; + + case TGSI_OPCODE_CMP: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + src2 = src_vector(p, &inst->FullSrcRegisters[2]); + i915_emit_arith(p, A0_CMP, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), + 0, src0, src2, src1); /* NOTE: order of src2, src1 */ + break; + + case TGSI_OPCODE_COS: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + src0, i915_emit_const1f(p, 1.0 / (M_PI * 2)), 0); + + i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); + + /* By choosing different taylor constants, could get rid of this mul: + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + tmp, i915_emit_const1f(p, (M_PI * 2)), 0); + + /* + * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 + * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, 1 + * t0 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 + * result = DP4 t0, cos_constants + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XY, 0, + swizzle(tmp, X, X, ONE, ONE), + swizzle(tmp, X, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XYZ, 0, + swizzle(tmp, X, Y, X, ONE), + swizzle(tmp, X, X, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XYZ, 0, + swizzle(tmp, X, X, Z, ONE), + swizzle(tmp, Z, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(tmp, ONE, Z, Y, X), + i915_emit_const4fv(p, cos_constants), 0); + break; + + case TGSI_OPCODE_DP3: + EMIT_2ARG_ARITH(A0_DP3); + break; + + case TGSI_OPCODE_DP4: + EMIT_2ARG_ARITH(A0_DP4); + break; + + case TGSI_OPCODE_DPH: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, Y, Z, ONE), src1, 0); + break; + + case TGSI_OPCODE_DST: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + + /* result[0] = 1 * 1; + * result[1] = a[1] * b[1]; + * result[2] = a[2] * 1; + * result[3] = 1 * b[3]; + */ + i915_emit_arith(p, + A0_MUL, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, ONE, Y, Z, ONE), + swizzle(src1, ONE, Y, ONE, W), 0); + break; + + case TGSI_OPCODE_EX2: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + + i915_emit_arith(p, + A0_EXP, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_FLR: + EMIT_1ARG_ARITH(A0_FLR); + break; + + case TGSI_OPCODE_FRC: + EMIT_1ARG_ARITH(A0_FRC); + break; + + case TGSI_OPCODE_KIL: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ + 0, src0, T0_TEXKILL); + break; + + case TGSI_OPCODE_LG2: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + + i915_emit_arith(p, + A0_LOG, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_LIT: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + /* tmp = max( a.xyzw, a.00zw ) + * XXX: Clamp tmp.w to -128..128 + * tmp.y = log(tmp.y) + * tmp.y = tmp.w * tmp.y + * tmp.y = exp(tmp.y) + * result = cmp (a.11-x1, a.1x01, a.1xy1 ) + */ + i915_emit_arith(p, A0_MAX, tmp, A0_DEST_CHANNEL_ALL, 0, + src0, swizzle(src0, ZERO, ZERO, Z, W), 0); + + i915_emit_arith(p, A0_LOG, tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, Y, Y, Y, Y), 0, 0); + + i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, Y, ZERO, ZERO), + swizzle(tmp, ZERO, W, ZERO, ZERO), 0); + + i915_emit_arith(p, A0_EXP, tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, Y, Y, Y, Y), 0, 0); + + i915_emit_arith(p, A0_CMP, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + negate(swizzle(tmp, ONE, ONE, X, ONE), 0, 0, 1, 0), + swizzle(tmp, ONE, X, ZERO, ONE), + swizzle(tmp, ONE, X, Y, ONE)); + + break; + + case TGSI_OPCODE_LRP: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + src2 = src_vector(p, &inst->FullSrcRegisters[2]); + flags = get_result_flags(inst); + tmp = i915_get_utemp(p); + + /* b*a + c*(1-a) + * + * b*a + c - ca + * + * tmp = b*a + c, + * result = (-c)*a + tmp + */ + i915_emit_arith(p, A0_MAD, tmp, + flags & A0_DEST_CHANNEL_ALL, 0, src1, src0, src2); + + i915_emit_arith(p, A0_MAD, + get_result_vector(p, &inst->FullDstRegisters[0]), + flags, 0, negate(src2, 1, 1, 1, 1), src0, tmp); + break; + + case TGSI_OPCODE_MAD: + EMIT_3ARG_ARITH(A0_MAD); + break; + + case TGSI_OPCODE_MAX: + EMIT_2ARG_ARITH(A0_MAX); + break; + + case TGSI_OPCODE_MIN: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + tmp = i915_get_utemp(p); + flags = get_result_flags(inst); + + i915_emit_arith(p, + A0_MAX, + tmp, flags & A0_DEST_CHANNEL_ALL, 0, + negate(src0, 1, 1, 1, 1), + negate(src1, 1, 1, 1, 1), 0); + + i915_emit_arith(p, + A0_MOV, + get_result_vector(p, &inst->FullDstRegisters[0]), + flags, 0, negate(tmp, 1, 1, 1, 1), 0, 0); + break; + + case TGSI_OPCODE_MOV: + /* aka TGSI_OPCODE_SWZ */ + EMIT_1ARG_ARITH(A0_MOV); + break; + + case TGSI_OPCODE_MUL: + EMIT_2ARG_ARITH(A0_MUL); + break; + + case TGSI_OPCODE_POW: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + tmp = i915_get_utemp(p); + flags = get_result_flags(inst); + + /* XXX: masking on intermediate values, here and elsewhere. + */ + i915_emit_arith(p, + A0_LOG, + tmp, A0_DEST_CHANNEL_X, 0, + swizzle(src0, X, X, X, X), 0, 0); + + i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, tmp, src1, 0); + + i915_emit_arith(p, + A0_EXP, + get_result_vector(p, &inst->FullDstRegisters[0]), + flags, 0, swizzle(tmp, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_RCP: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + + i915_emit_arith(p, + A0_RCP, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_RSQ: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + + i915_emit_arith(p, + A0_RSQ, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_SCS: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + /* + * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 + * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x + * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x + * scs.x = DP4 t1, sin_constants + * t1 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 + * scs.y = DP4 t1, cos_constants + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XY, 0, + swizzle(src0, X, X, ONE, ONE), + swizzle(src0, X, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, X, Y), + swizzle(tmp, X, X, ONE, ONE), 0); + + writemask = inst->FullDstRegisters[0].DstRegister.WriteMask; + + if (writemask & TGSI_WRITEMASK_Y) { + uint tmp1; + + if (writemask & TGSI_WRITEMASK_X) + tmp1 = i915_get_utemp(p); + else + tmp1 = tmp; + + i915_emit_arith(p, + A0_MUL, + tmp1, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, Y, W), + swizzle(tmp, X, Z, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + A0_DEST_CHANNEL_Y, 0, + swizzle(tmp1, W, Z, Y, X), + i915_emit_const4fv(p, sin_constants), 0); + } + + if (writemask & TGSI_WRITEMASK_X) { + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XYZ, 0, + swizzle(tmp, X, X, Z, ONE), + swizzle(tmp, Z, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + A0_DEST_CHANNEL_X, 0, + swizzle(tmp, ONE, Z, Y, X), + i915_emit_const4fv(p, cos_constants), 0); + } + break; + + case TGSI_OPCODE_SGE: + EMIT_2ARG_ARITH(A0_SGE); + break; + + case TGSI_OPCODE_SIN: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + src0, i915_emit_const1f(p, 1.0 / (M_PI * 2)), 0); + + i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); + + /* By choosing different taylor constants, could get rid of this mul: + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + tmp, i915_emit_const1f(p, (M_PI * 2)), 0); + + /* + * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 + * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x + * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x + * result = DP4 t1.wzyx, sin_constants + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XY, 0, + swizzle(tmp, X, X, ONE, ONE), + swizzle(tmp, X, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, X, Y), + swizzle(tmp, X, X, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, Y, W), + swizzle(tmp, X, Z, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(tmp, W, Z, Y, X), + i915_emit_const4fv(p, sin_constants), 0); + break; + + case TGSI_OPCODE_SLT: + EMIT_2ARG_ARITH(A0_SLT); + break; + + case TGSI_OPCODE_SUB: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + + i915_emit_arith(p, + A0_ADD, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + src0, negate(src1, 1, 1, 1, 1), 0); + break; + + case TGSI_OPCODE_TEX: + emit_tex(p, inst, T0_TEXLD); + break; + + case TGSI_OPCODE_TXB: + emit_tex(p, inst, T0_TEXLDB); + break; + + case TGSI_OPCODE_TXP: + emit_tex(p, inst, T0_TEXLDP); + break; + + case TGSI_OPCODE_XPD: + /* Cross product: + * result.x = src0.y * src1.z - src0.z * src1.y; + * result.y = src0.z * src1.x - src0.x * src1.z; + * result.z = src0.x * src1.y - src0.y * src1.x; + * result.w = undef; + */ + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + tmp = i915_get_utemp(p); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(src0, Z, X, Y, ONE), + swizzle(src1, Y, Z, X, ONE), 0); + + i915_emit_arith(p, + A0_MAD, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, Y, Z, X, ONE), + swizzle(src1, Z, X, Y, ONE), + negate(tmp, 1, 1, 1, 0)); + break; + + default: + i915_program_error(p, "bad opcode"); + return; + } + + i915_release_utemps(p); +} + + +/** + * Translate TGSI fragment shader into i915 hardware instructions. + * + * Possible concerns: + * + * SIN, COS -- could use another taylor step? + * LIT -- results seem a little different to sw mesa + * LOG -- different to mesa on negative numbers, but this is conformant. + * + * Parse failures -- Mesa doesn't currently give a good indication + * internally whether a particular program string parsed or not. This + * can lead to confusion -- hopefully we cope with it ok now. + */ +void +i915_translate_program(struct i915_fp_compile *p, + const struct tgsi_token *tokens) +{ + struct tgsi_parse_context parse; + + tgsi_parse_init( &parse, tokens ); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + assert(0); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + assert(0); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + i915_translate_instruction(p, &parse.FullToken.FullInstruction); + break; + + default: + assert( 0 ); + } + + } /* while */ + + tgsi_parse_free (&parse); +} + + + + +/* Rather than trying to intercept and jiggle depth writes during + * emit, just move the value into its correct position at the end of + * the program: + */ +void +i915_fixup_depth_write(struct i915_fp_compile *p) +{ + if (p->shader->outputs_written & (1< Date: Tue, 21 Aug 2007 16:24:48 -0600 Subject: added fpc sources --- src/mesa/pipe/i915simple/Makefile | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 9a0d40a8d4..fc825c7302 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -22,6 +22,10 @@ DRIVER_SOURCES = \ i915_strings.c \ i915_prim_emit.c \ i915_tex_layout.c \ + i915_fpc.c \ + i915_fpc_debug.c \ + i915_fpc_emit.c \ + i915_fpc_translate.c \ i915_surface.c C_SOURCES = \ -- cgit v1.2.3 From 93c7173ed2fbe2bc47ee6879af89366cafd90ba5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 16 Aug 2007 12:41:51 +0100 Subject: Add missing #include --- src/mesa/pipe/i915simple/i915_debug.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index 8b7e3628f3..79060e9780 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -31,6 +31,7 @@ #include "i915_context.h" #include "i915_winsys.h" #include "i915_debug.h" +#include "pipe/p_winsys.h" #define PRINTF( ... ) (stream)->winsys->printf( (stream)->winsys, __VA_ARGS__ ) -- cgit v1.2.3 From ce4659e92acfbb0ff8a93a6e619c44b839505ca9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 21 Aug 2007 19:46:48 +0100 Subject: First pass at a fallback concept for pipe devices. Creates a new pipe driver that feeds commands to either a hardware or software pipe depending on fallback state. Untested concept checkpoint. At this point it compiles. --- src/mesa/pipe/Makefile | 1 + src/mesa/pipe/failover/Makefile | 21 ++ src/mesa/pipe/failover/fo_context.c | 155 ++++++++++++++ src/mesa/pipe/failover/fo_context.h | 113 ++++++++++ src/mesa/pipe/failover/fo_state.c | 287 ++++++++++++++++++++++++++ src/mesa/pipe/failover/fo_state_emit.c | 131 ++++++++++++ src/mesa/pipe/failover/fo_winsys.h | 45 ++++ src/mesa/pipe/i915simple/i915_context.c | 10 +- src/mesa/pipe/i915simple/i915_fpc_translate.c | 1 + src/mesa/pipe/p_context.h | 15 +- src/mesa/pipe/softpipe/sp_draw_arrays.c | 10 +- src/mesa/pipe/softpipe/sp_state.h | 12 +- 12 files changed, 780 insertions(+), 21 deletions(-) create mode 100644 src/mesa/pipe/failover/Makefile create mode 100644 src/mesa/pipe/failover/fo_context.c create mode 100644 src/mesa/pipe/failover/fo_context.h create mode 100644 src/mesa/pipe/failover/fo_state.c create mode 100644 src/mesa/pipe/failover/fo_state_emit.c create mode 100644 src/mesa/pipe/failover/fo_winsys.h (limited to 'src') diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile index f5e884e3a0..d448f3f5a6 100644 --- a/src/mesa/pipe/Makefile +++ b/src/mesa/pipe/Makefile @@ -1,6 +1,7 @@ default: cd softpipe ; make cd i915simple ; make + cd failover ; make clean: rm -f `find . -name \*.[oa]` \ No newline at end of file diff --git a/src/mesa/pipe/failover/Makefile b/src/mesa/pipe/failover/Makefile new file mode 100644 index 0000000000..72d0895c74 --- /dev/null +++ b/src/mesa/pipe/failover/Makefile @@ -0,0 +1,21 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = failover + +DRIVER_SOURCES = \ + fo_state.c \ + fo_state_emit.c \ + fo_context.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../Makefile.template + +symlinks: + diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c new file mode 100644 index 0000000000..b88f1b466c --- /dev/null +++ b/src/mesa/pipe/failover/fo_context.c @@ -0,0 +1,155 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_context.h" + +#include "fo_context.h" +#include "fo_winsys.h" + + + +static void failover_destroy( struct pipe_context *pipe ) +{ + struct failover_context *failover = failover_context( pipe ); + + free( failover ); +} + + + +static boolean failover_draw_elements( struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) +{ + struct failover_context *failover = failover_context( pipe ); + + /* If there has been any statechange since last time, try hardware + * rendering again: + */ + if (failover->dirty) { + failover->mode = FO_HW; + } + + /* Try hardware: + */ + if (failover->mode == FO_HW) { + if (!failover->hw->draw_elements( failover->hw, + indexBuffer, + indexSize, + prim, + start, + count )) { + + failover->hw->flush( failover->hw, ~0 ); + failover->mode = FO_SW; + } + } + + /* Possibly try software: + */ + if (failover->mode == FO_SW) { + + if (failover->dirty) + failover_state_emit( failover ); + + failover->sw->draw_elements( failover->sw, + indexBuffer, + indexSize, + prim, + start, + count ); + + /* Be ready to switch back to hardware rendering without an + * intervening flush. Unlikely to be much performance impact to + * this: + */ + failover->sw->flush( failover->sw, ~0 ); + } + + return TRUE; +} + + +static boolean failover_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) +{ + return failover_draw_elements(pipe, NULL, 0, prim, start, count); +} + + + +struct pipe_context *failover_create( struct pipe_context *hw, + struct pipe_context *sw ) +{ + struct failover_context *failover = CALLOC_STRUCT(failover_context); + if (failover == NULL) + return NULL; + + failover->pipe.winsys = hw->winsys; + failover->pipe.destroy = failover_destroy; + failover->pipe.supported_formats = hw->supported_formats; + failover->pipe.max_texture_size = hw->max_texture_size; + failover->pipe.get_name = hw->get_name; + failover->pipe.get_vendor = hw->get_vendor; + + failover->pipe.draw_arrays = failover_draw_arrays; + failover->pipe.draw_elements = failover_draw_elements; + failover->pipe.clear = hw->clear; + + /* No software occlusion fallback (or other optional functionality) + * at this point - if the hardware doesn't support it, don't + * advertise it to the application. + */ + failover->pipe.reset_occlusion_counter = hw->reset_occlusion_counter; + failover->pipe.get_occlusion_counter = hw->get_occlusion_counter; + + failover_init_state_functions( failover ); + + failover->pipe.surface_alloc = hw->surface_alloc; + failover->pipe.get_tex_surface = hw->get_tex_surface; + + failover->pipe.region_alloc = hw->region_alloc; + failover->pipe.region_release = hw->region_release; + failover->pipe.region_idle = hw->region_idle; + failover->pipe.region_map = hw->region_map; + failover->pipe.region_unmap = hw->region_unmap; + failover->pipe.region_data = hw->region_data; + failover->pipe.region_copy = hw->region_copy; + failover->pipe.region_fill = hw->region_fill; + failover->pipe.mipmap_tree_layout = hw->mipmap_tree_layout; + failover->pipe.flush = hw->flush; + + failover->dirty = 0; + + return &failover->pipe; +} + diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h new file mode 100644 index 0000000000..5666d4e830 --- /dev/null +++ b/src/mesa/pipe/failover/fo_context.h @@ -0,0 +1,113 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef FO_CONTEXT_H +#define FO_CONTEXT_H + +#include "pipe/p_state.h" +#include "pipe/p_context.h" + + + +#define FO_NEW_VIEWPORT 0x1 +#define FO_NEW_SETUP 0x2 +#define FO_NEW_FRAGMENT_SHADER 0x4 +#define FO_NEW_BLEND 0x8 +#define FO_NEW_CLIP 0x10 +#define FO_NEW_SCISSOR 0x20 +#define FO_NEW_STIPPLE 0x40 +#define FO_NEW_FRAMEBUFFER 0x80 +#define FO_NEW_ALPHA_TEST 0x100 +#define FO_NEW_DEPTH_TEST 0x200 +#define FO_NEW_SAMPLER 0x400 +#define FO_NEW_TEXTURE 0x800 +#define FO_NEW_STENCIL 0x1000 +#define FO_NEW_VERTEX 0x2000 +#define FO_NEW_VERTEX_SHADER 0x4000 +#define FO_NEW_BLEND_COLOR 0x8000 +#define FO_NEW_CLEAR_COLOR 0x10000 +#define FO_NEW_VERTEX_BUFFER 0x20000 +#define FO_NEW_VERTEX_ELEMENT 0x40000 + + + +#define FO_HW 0 +#define FO_SW 1 + +struct failover_context { + struct pipe_context pipe; /**< base class */ + + + /* The most recent drawing state as set by the driver: + */ + struct pipe_alpha_test_state alpha_test; + struct pipe_blend_state blend; + struct pipe_blend_color blend_color; + struct pipe_clear_color_state clear_color; + struct pipe_clip_state clip; + struct pipe_depth_state depth_test; + struct pipe_framebuffer_state framebuffer; + struct pipe_shader_state fragment_shader; + struct pipe_shader_state vertex_shader; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; + struct pipe_setup_state setup; + struct pipe_stencil_state stencil; + struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + + unsigned dirty; + unsigned dirty_sampler; + unsigned dirty_texture; + unsigned dirty_vertex_buffer; + unsigned dirty_vertex_element; + + + unsigned mode; + struct pipe_context *hw; + struct pipe_context *sw; +}; + + + +void failover_init_state_functions( struct failover_context *failover ); +void failover_state_emit( struct failover_context *failover ); + +static INLINE struct failover_context * +failover_context( struct pipe_context *pipe ) +{ + return (struct failover_context *)pipe; +} + + +#endif /* FO_CONTEXT_H */ diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c new file mode 100644 index 0000000000..097acf7d57 --- /dev/null +++ b/src/mesa/pipe/failover/fo_state.c @@ -0,0 +1,287 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "fo_context.h" + + +/* This looks like a lot of work at the moment - we're keeping a + * duplicate copy of the state up-to-date. + * + * This can change in two ways: + * - With constant state objects we would only need to save a pointer, + * not the whole object. + * - By adding a callback in the state tracker to re-emit state. The + * state tracker knows the current state already and can re-emit it + * without additional complexity. + * + * This works as a proof-of-concept, but a final version will have + * lower overheads. + */ + +static void +failover_set_alpha_test_state(struct pipe_context *pipe, + const struct pipe_alpha_test_state *alpha) +{ + struct failover_context *failover = failover_context(pipe); + + failover->alpha_test = *alpha; + failover->dirty |= FO_NEW_ALPHA_TEST; + failover->hw->set_alpha_test_state( failover->hw, alpha ); +} + + +static void +failover_set_blend_state( struct pipe_context *pipe, + const struct pipe_blend_state *blend ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->blend = *blend; + failover->dirty |= FO_NEW_BLEND; + failover->hw->set_blend_state( failover->hw, blend ); +} + + +static void +failover_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->blend_color = *blend_color; + failover->dirty |= FO_NEW_BLEND_COLOR; + failover->hw->set_blend_color( failover->hw, blend_color ); +} + +static void +failover_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->clip = *clip; + failover->dirty |= FO_NEW_CLIP; + failover->hw->set_clip_state( failover->hw, clip ); +} + +static void +failover_set_clear_color_state( struct pipe_context *pipe, + const struct pipe_clear_color_state *clear_color ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->clear_color = *clear_color; + failover->dirty |= FO_NEW_CLEAR_COLOR; + failover->hw->set_clear_color_state( failover->hw, clear_color ); +} + +static void +failover_set_depth_test_state(struct pipe_context *pipe, + const struct pipe_depth_state *depth) +{ + struct failover_context *failover = failover_context(pipe); + + failover->depth_test = *depth; + failover->dirty |= FO_NEW_DEPTH_TEST; + failover->hw->set_depth_state( failover->hw, depth ); +} + +static void +failover_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *framebuffer) +{ + struct failover_context *failover = failover_context(pipe); + + failover->framebuffer = *framebuffer; + failover->dirty |= FO_NEW_FRAMEBUFFER; + failover->hw->set_framebuffer_state( failover->hw, framebuffer ); +} + +static void +failover_set_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *fs) +{ + struct failover_context *failover = failover_context(pipe); + + failover->fragment_shader = *fs; + failover->dirty |= FO_NEW_FRAGMENT_SHADER; + failover->hw->set_fs_state( failover->hw, fs ); +} + +static void +failover_set_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *vs) +{ + struct failover_context *failover = failover_context(pipe); + + failover->vertex_shader = *vs; + failover->dirty |= FO_NEW_VERTEX_SHADER; + failover->hw->set_vs_state( failover->hw, vs ); +} + + +static void +failover_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->poly_stipple = *stipple; + failover->dirty |= FO_NEW_STIPPLE; + failover->hw->set_polygon_stipple( failover->hw, stipple ); +} + + + +static void +failover_set_setup_state( struct pipe_context *pipe, + const struct pipe_setup_state *setup ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->setup = *setup; + failover->dirty |= FO_NEW_SETUP; + failover->hw->set_setup_state( failover->hw, setup ); +} + + +static void +failover_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->scissor = *scissor; + failover->dirty |= FO_NEW_SCISSOR; + failover->hw->set_scissor_state( failover->hw, scissor ); +} + +static void +failover_set_stencil_state(struct pipe_context *pipe, + const struct pipe_stencil_state *stencil) +{ + struct failover_context *failover = failover_context(pipe); + + failover->stencil = *stencil; + failover->dirty |= FO_NEW_STENCIL; + failover->hw->set_stencil_state( failover->hw, stencil ); +} + + +static void +failover_set_sampler_state(struct pipe_context *pipe, + unsigned unit, + const struct pipe_sampler_state *sampler) +{ + struct failover_context *failover = failover_context(pipe); + + failover->sampler[unit] = *sampler; + failover->dirty |= FO_NEW_SAMPLER; + failover->dirty_sampler |= (1<hw->set_sampler_state( failover->hw, unit, sampler ); +} + + +static void +failover_set_texture_state(struct pipe_context *pipe, + unsigned unit, + struct pipe_mipmap_tree *texture) +{ + struct failover_context *failover = failover_context(pipe); + + failover->texture[unit] = texture; + failover->dirty |= FO_NEW_TEXTURE; + failover->dirty_texture |= (1<hw->set_texture_state( failover->hw, unit, texture ); +} + + +static void +failover_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->viewport = *viewport; + failover->dirty |= FO_NEW_VIEWPORT; + failover->hw->set_viewport_state( failover->hw, viewport ); +} + + +static void +failover_set_vertex_buffer(struct pipe_context *pipe, + unsigned unit, + const struct pipe_vertex_buffer *vertex_buffer) +{ + struct failover_context *failover = failover_context(pipe); + + failover->vertex_buffer[unit] = *vertex_buffer; + failover->dirty |= FO_NEW_VERTEX_BUFFER; + failover->dirty_vertex_buffer |= (1<hw->set_vertex_buffer( failover->hw, unit, vertex_buffer ); +} + + +static void +failover_set_vertex_element(struct pipe_context *pipe, + unsigned unit, + const struct pipe_vertex_element *vertex_element) +{ + struct failover_context *failover = failover_context(pipe); + + failover->vertex_element[unit] = *vertex_element; + failover->dirty |= FO_NEW_VERTEX_ELEMENT; + failover->dirty_vertex_element |= (1<hw->set_vertex_element( failover->hw, unit, vertex_element ); +} + + +void +failover_init_state_functions( struct failover_context *failover ) +{ + failover->pipe.set_alpha_test_state = failover_set_alpha_test_state; + failover->pipe.set_blend_color = failover_set_blend_color; + failover->pipe.set_blend_state = failover_set_blend_state; + failover->pipe.set_clip_state = failover_set_clip_state; + failover->pipe.set_clear_color_state = failover_set_clear_color_state; + failover->pipe.set_depth_state = failover_set_depth_test_state; + failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; + failover->pipe.set_fs_state = failover_set_fs_state; + failover->pipe.set_vs_state = failover_set_vs_state; + failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; + failover->pipe.set_sampler_state = failover_set_sampler_state; + failover->pipe.set_scissor_state = failover_set_scissor_state; + failover->pipe.set_setup_state = failover_set_setup_state; + failover->pipe.set_stencil_state = failover_set_stencil_state; + failover->pipe.set_texture_state = failover_set_texture_state; + failover->pipe.set_viewport_state = failover_set_viewport_state; + failover->pipe.set_vertex_buffer = failover_set_vertex_buffer; + failover->pipe.set_vertex_element = failover_set_vertex_element; +} diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c new file mode 100644 index 0000000000..6614288297 --- /dev/null +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -0,0 +1,131 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "fo_context.h" + +/* This looks like a lot of work at the moment - we're keeping a + * duplicate copy of the state up-to-date. + * + * This can change in two ways: + * - With constant state objects we would only need to save a pointer, + * not the whole object. + * - By adding a callback in the state tracker to re-emit state. The + * state tracker knows the current state already and can re-emit it + * without additional complexity. + * + * This works as a proof-of-concept, but a final version will have + * lower overheads. + */ + +void +failover_state_emit( struct failover_context *failover ) +{ + unsigned i; + + if (failover->dirty & FO_NEW_ALPHA_TEST) + failover->hw->set_alpha_test_state( failover->hw, &failover->alpha_test ); + + if (failover->dirty & FO_NEW_BLEND) + failover->hw->set_blend_state( failover->hw, &failover->blend ); + + if (failover->dirty & FO_NEW_BLEND_COLOR) + failover->hw->set_blend_color( failover->hw, &failover->blend_color ); + + if (failover->dirty & FO_NEW_CLIP) + failover->hw->set_clip_state( failover->hw, &failover->clip ); + + if (failover->dirty & FO_NEW_CLEAR_COLOR) + failover->hw->set_clear_color_state( failover->hw, &failover->clear_color ); + + if (failover->dirty & FO_NEW_DEPTH_TEST) + failover->hw->set_depth_state( failover->hw, &failover->depth_test ); + + if (failover->dirty & FO_NEW_FRAMEBUFFER) + failover->hw->set_framebuffer_state( failover->hw, &failover->framebuffer ); + + if (failover->dirty & FO_NEW_FRAGMENT_SHADER) + failover->hw->set_fs_state( failover->hw, &failover->fragment_shader ); + + if (failover->dirty & FO_NEW_VERTEX_SHADER) + failover->hw->set_vs_state( failover->hw, &failover->vertex_shader ); + + if (failover->dirty & FO_NEW_STIPPLE) + failover->hw->set_polygon_stipple( failover->hw, &failover->poly_stipple ); + + if (failover->dirty & FO_NEW_SETUP) + failover->hw->set_setup_state( failover->hw, &failover->setup ); + + if (failover->dirty & FO_NEW_SCISSOR) + failover->hw->set_scissor_state( failover->hw, &failover->scissor ); + + if (failover->dirty & FO_NEW_STENCIL) + failover->hw->set_stencil_state( failover->hw, &failover->stencil ); + + if (failover->dirty & FO_NEW_VIEWPORT) + failover->hw->set_viewport_state( failover->hw, &failover->viewport ); + + if (failover->dirty & FO_NEW_SAMPLER) { + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + if (failover->dirty_sampler & (1<hw->set_sampler_state( failover->hw, i, + &failover->sampler[i] ); + } + } + } + + if (failover->dirty & FO_NEW_TEXTURE) { + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + if (failover->dirty_texture & (1<hw->set_texture_state( failover->hw, i, + failover->texture[i] ); + } + } + } + + if (failover->dirty & FO_NEW_VERTEX_BUFFER) { + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (failover->dirty_vertex_buffer & (1<hw->set_vertex_buffer( failover->hw, i, + &failover->vertex_buffer[i] ); + } + } + } + + if (failover->dirty & FO_NEW_VERTEX_ELEMENT) { + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (failover->dirty_vertex_element & (1<hw->set_vertex_element( failover->hw, i, + &failover->vertex_element[i] ); + } + } + } + + failover->dirty = 0; +} diff --git a/src/mesa/pipe/failover/fo_winsys.h b/src/mesa/pipe/failover/fo_winsys.h new file mode 100644 index 0000000000..a8ce997a1f --- /dev/null +++ b/src/mesa/pipe/failover/fo_winsys.h @@ -0,0 +1,45 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef FO_WINSYS_H +#define FO_WINSYS_H + + +/* This is the interface that failover requires any window system + * hosting it to implement. This is the only include file in failover + * which is public. + */ + + +struct pipe_context; + + +struct pipe_context *failover_create( struct pipe_context *hw, + struct pipe_context *sw ); + + +#endif /* FO_WINSYS_H */ diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index fdb1a7422d..9856c7c10c 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -149,7 +149,7 @@ static void i915_destroy( struct pipe_context *pipe ) -static void i915_draw_elements( struct pipe_context *pipe, +static boolean i915_draw_elements( struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count) @@ -202,13 +202,15 @@ static void i915_draw_elements( struct pipe_context *pipe, pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); draw_set_mapped_element_buffer(draw, 0, NULL); } + + return TRUE; } -static void i915_draw_arrays( struct pipe_context *pipe, - unsigned prim, unsigned start, unsigned count) +static boolean i915_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) { - i915_draw_elements(pipe, NULL, 0, prim, start, count); + return i915_draw_elements(pipe, NULL, 0, prim, start, count); } diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index a034e734c3..db6b92ad7f 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -29,6 +29,7 @@ #include "i915_context.h" #include "i915_fpc.h" +#include "pipe/tgsi/core/tgsi_token.h" #include "pipe/tgsi/core/tgsi_parse.h" diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index c4496cda50..3a656b272e 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -60,15 +60,16 @@ struct pipe_context { /* - * Drawing + * Drawing. + * Return false on fallbacks (temporary??) */ - void (*draw_arrays)( struct pipe_context *pipe, - unsigned mode, unsigned start, unsigned count); + boolean (*draw_arrays)( struct pipe_context *pipe, + unsigned mode, unsigned start, unsigned count); - void (*draw_elements)( struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count); + boolean (*draw_elements)( struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); /** Clear a surface to given value (no scissor; clear whole surface) */ void (*clear)(struct pipe_context *pipe, struct pipe_surface *ps, diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 9ced89cd7c..0392b6b64e 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -43,11 +43,11 @@ -void +boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { - softpipe_draw_elements(pipe, NULL, 0, mode, start, count); + return softpipe_draw_elements(pipe, NULL, 0, mode, start, count); } @@ -59,7 +59,7 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, * * XXX should the element buffer be specified/bound with a separate function? */ -void +boolean softpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, unsigned indexSize, @@ -73,7 +73,7 @@ softpipe_draw_elements(struct pipe_context *pipe, draw_prim_info( mode, &first, &incr ); length = draw_trim( count, first, incr ); if (!length) - return; + return TRUE; if (sp->dirty) @@ -123,4 +123,6 @@ softpipe_draw_elements(struct pipe_context *pipe, } softpipe_unmap_surfaces(sp); + + return TRUE; } diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 49fef0ddce..5e1ecd94e0 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -95,13 +95,13 @@ void softpipe_set_vertex_buffer(struct pipe_context *, void softpipe_update_derived( struct softpipe_context *softpipe ); -void softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, - unsigned start, unsigned count); +boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count); -void softpipe_draw_elements(struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count); +boolean softpipe_draw_elements(struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); void -- cgit v1.2.3 From d70d41eefc1045bd2f8ddf2cb701d7789c671012 Mon Sep 17 00:00:00 2001 From: keithw Date: Tue, 21 Aug 2007 19:49:46 +0100 Subject: fix cut and paste --- src/mesa/pipe/failover/fo_state_emit.c | 42 +++++++++++++++++++--------------- 1 file changed, 24 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 6614288297..7760419394 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -44,57 +44,63 @@ * lower overheads. */ + +/* Bring the software pipe uptodate with current state. + * + * With constant state objects we would probably just send all state + * to both rasterizers all the time??? + */ void failover_state_emit( struct failover_context *failover ) { unsigned i; if (failover->dirty & FO_NEW_ALPHA_TEST) - failover->hw->set_alpha_test_state( failover->hw, &failover->alpha_test ); + failover->sw->set_alpha_test_state( failover->sw, &failover->alpha_test ); if (failover->dirty & FO_NEW_BLEND) - failover->hw->set_blend_state( failover->hw, &failover->blend ); + failover->sw->set_blend_state( failover->sw, &failover->blend ); if (failover->dirty & FO_NEW_BLEND_COLOR) - failover->hw->set_blend_color( failover->hw, &failover->blend_color ); + failover->sw->set_blend_color( failover->sw, &failover->blend_color ); if (failover->dirty & FO_NEW_CLIP) - failover->hw->set_clip_state( failover->hw, &failover->clip ); + failover->sw->set_clip_state( failover->sw, &failover->clip ); if (failover->dirty & FO_NEW_CLEAR_COLOR) - failover->hw->set_clear_color_state( failover->hw, &failover->clear_color ); + failover->sw->set_clear_color_state( failover->sw, &failover->clear_color ); if (failover->dirty & FO_NEW_DEPTH_TEST) - failover->hw->set_depth_state( failover->hw, &failover->depth_test ); + failover->sw->set_depth_state( failover->sw, &failover->depth_test ); if (failover->dirty & FO_NEW_FRAMEBUFFER) - failover->hw->set_framebuffer_state( failover->hw, &failover->framebuffer ); + failover->sw->set_framebuffer_state( failover->sw, &failover->framebuffer ); if (failover->dirty & FO_NEW_FRAGMENT_SHADER) - failover->hw->set_fs_state( failover->hw, &failover->fragment_shader ); + failover->sw->set_fs_state( failover->sw, &failover->fragment_shader ); if (failover->dirty & FO_NEW_VERTEX_SHADER) - failover->hw->set_vs_state( failover->hw, &failover->vertex_shader ); + failover->sw->set_vs_state( failover->sw, &failover->vertex_shader ); if (failover->dirty & FO_NEW_STIPPLE) - failover->hw->set_polygon_stipple( failover->hw, &failover->poly_stipple ); + failover->sw->set_polygon_stipple( failover->sw, &failover->poly_stipple ); if (failover->dirty & FO_NEW_SETUP) - failover->hw->set_setup_state( failover->hw, &failover->setup ); + failover->sw->set_setup_state( failover->sw, &failover->setup ); if (failover->dirty & FO_NEW_SCISSOR) - failover->hw->set_scissor_state( failover->hw, &failover->scissor ); + failover->sw->set_scissor_state( failover->sw, &failover->scissor ); if (failover->dirty & FO_NEW_STENCIL) - failover->hw->set_stencil_state( failover->hw, &failover->stencil ); + failover->sw->set_stencil_state( failover->sw, &failover->stencil ); if (failover->dirty & FO_NEW_VIEWPORT) - failover->hw->set_viewport_state( failover->hw, &failover->viewport ); + failover->sw->set_viewport_state( failover->sw, &failover->viewport ); if (failover->dirty & FO_NEW_SAMPLER) { for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { if (failover->dirty_sampler & (1<hw->set_sampler_state( failover->hw, i, + failover->sw->set_sampler_state( failover->sw, i, &failover->sampler[i] ); } } @@ -103,7 +109,7 @@ failover_state_emit( struct failover_context *failover ) if (failover->dirty & FO_NEW_TEXTURE) { for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { if (failover->dirty_texture & (1<hw->set_texture_state( failover->hw, i, + failover->sw->set_texture_state( failover->sw, i, failover->texture[i] ); } } @@ -112,7 +118,7 @@ failover_state_emit( struct failover_context *failover ) if (failover->dirty & FO_NEW_VERTEX_BUFFER) { for (i = 0; i < PIPE_ATTRIB_MAX; i++) { if (failover->dirty_vertex_buffer & (1<hw->set_vertex_buffer( failover->hw, i, + failover->sw->set_vertex_buffer( failover->sw, i, &failover->vertex_buffer[i] ); } } @@ -121,7 +127,7 @@ failover_state_emit( struct failover_context *failover ) if (failover->dirty & FO_NEW_VERTEX_ELEMENT) { for (i = 0; i < PIPE_ATTRIB_MAX; i++) { if (failover->dirty_vertex_element & (1<hw->set_vertex_element( failover->hw, i, + failover->sw->set_vertex_element( failover->sw, i, &failover->vertex_element[i] ); } } -- cgit v1.2.3 From 0d9bcdbeb11ad1ce7e5257f652ccf9ebf1aa59a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 21 Aug 2007 20:13:31 -0600 Subject: Fragment shader translation seems to basically work now. More testing needed. --- src/mesa/pipe/i915simple/Makefile | 1 - src/mesa/pipe/i915simple/i915_context.h | 7 + src/mesa/pipe/i915simple/i915_fpc.c | 146 ------------- src/mesa/pipe/i915simple/i915_fpc.h | 105 ++------- src/mesa/pipe/i915simple/i915_fpc_translate.c | 297 ++++++++++++++++++++------ src/mesa/pipe/i915simple/i915_state.c | 2 +- src/mesa/pipe/i915simple/i915_state_derived.c | 5 + src/mesa/pipe/i915simple/i915_state_emit.c | 43 ++-- 8 files changed, 284 insertions(+), 322 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index fc825c7302..2353d0df01 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -22,7 +22,6 @@ DRIVER_SOURCES = \ i915_strings.c \ i915_prim_emit.c \ i915_tex_layout.c \ - i915_fpc.c \ i915_fpc_debug.c \ i915_fpc_emit.c \ i915_fpc_translate.c \ diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index a3927bf8b8..bc00d61d4b 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -84,6 +84,11 @@ struct i915_state { unsigned immediate[I915_MAX_IMMEDIATE]; unsigned dynamic[I915_MAX_DYNAMIC]; + + uint *program; + uint program_len; + uint *constants; + uint num_constants; unsigned id; /* track lost context events */ }; @@ -115,6 +120,8 @@ struct i915_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_constant_buffer temp_constants; /*XXX temporary*/ + unsigned dirty; unsigned *batch_start; diff --git a/src/mesa/pipe/i915simple/i915_fpc.c b/src/mesa/pipe/i915simple/i915_fpc.c index fd0bbbc482..3bf2231589 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.c +++ b/src/mesa/pipe/i915simple/i915_fpc.c @@ -35,149 +35,3 @@ #include "i915_fpc.h" - - -void -i915_program_error(struct i915_fp_compile *p, const char *msg) -{ - fprintf(stderr, "i915_program_error: %s", msg); - p->fp->error = 1; -} - - -static struct i915_fp_compile * -i915_init_compile(struct i915_context *i915, struct i915_fragment_program *fp) -{ - struct i915_fp_compile *p = CALLOC_STRUCT(i915_fp_compile); - - p->fp = fp; -#if 0 - p->env_param = NULL; /*i915->intel.ctx.FragmentProgram.Parameters;*/ -#endif - p->constants = i915->fs.constants; - p->nr_tex_indirect = 1; /* correct? */ - p->nr_tex_insn = 0; - p->nr_alu_insn = 0; - p->nr_decl_insn = 0; - - memset(p->constant_flags, 0, sizeof(p->constant_flags)); - - p->csr = p->program; - p->decl = p->declarations; - p->decl_s = 0; - p->decl_t = 0; - p->temp_flag = 0xffff000; - p->utemp_flag = ~0x7; - -#if 0 - p->fp->translated = 0; - p->fp->error = 0; - p->fp->nr_constants = 0; -#endif - p->fp->wpos_tex = -1; - p->fp->nr_params = 0; - - *(p->decl++) = _3DSTATE_PIXEL_SHADER_PROGRAM; - - return p; -} - -/* Copy compile results to the fragment program struct and destroy the - * compilation context. - */ -static void -i915_fini_compile(struct i915_fp_compile *p) -{ - uint program_size = p->csr - p->program; - uint decl_size = p->decl - p->declarations; - - if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT) - i915_program_error(p, "Exceeded max nr indirect texture lookups"); - - if (p->nr_tex_insn > I915_MAX_TEX_INSN) - i915_program_error(p, "Exceeded max TEX instructions"); - - if (p->nr_alu_insn > I915_MAX_ALU_INSN) - i915_program_error(p, "Exceeded max ALU instructions"); - - if (p->nr_decl_insn > I915_MAX_DECL_INSN) - i915_program_error(p, "Exceeded max DECL instructions"); - - if (p->fp->error) { - p->fp->NumNativeInstructions = 0; - p->fp->NumNativeAluInstructions = 0; - p->fp->NumNativeTexInstructions = 0; - p->fp->NumNativeTexIndirections = 0; - return; - } - else { - p->fp->NumNativeInstructions = (p->nr_alu_insn + - p->nr_tex_insn + - p->nr_decl_insn); - p->fp->NumNativeAluInstructions = p->nr_alu_insn; - p->fp->NumNativeTexInstructions = p->nr_tex_insn; - p->fp->NumNativeTexIndirections = p->nr_tex_indirect; - } - - p->declarations[0] |= program_size + decl_size - 2; - - /* Copy compilation results to fragment program struct: - */ - memcpy(p->fp->program, - p->declarations, - decl_size * sizeof(uint)); - - memcpy(p->fp->program + decl_size, - p->program, - program_size * sizeof(uint)); - - p->fp->program_size = program_size + decl_size; - - /* Release the compilation struct: - */ - free(p); -} - - -/** - * Find an unused texture coordinate slot to use for fragment WPOS. - * Update p->fp->wpos_tex with the result (-1 if no used texcoord slot is found). - */ -static void -find_wpos_space(struct i915_fp_compile *p) -{ - const uint inputs = p->shader->inputs_read; - uint i; - - p->fp->wpos_tex = -1; - - if (inputs & FRAG_BIT_WPOS) { - for (i = 0; i < I915_TEX_UNITS; i++) { - if ((inputs & (FRAG_BIT_TEX0 << i)) == 0) { - p->fp->wpos_tex = i; - return; - } - } - - i915_program_error(p, "No free texcoord for wpos value"); - } -} - - - -void i915_compile_fragment_program( struct i915_context *i915, - struct i915_fragment_program *fp ) -{ - struct i915_fp_compile *p = i915_init_compile(i915, fp); - struct tgsi_token *tokens = i915->fs.tokens; - - find_wpos_space(p); - - i915_translate_program(p, tokens); - i915_fixup_depth_write(p); - - i915_fini_compile(p); -#if 0 - fp->translated = 1; -#endif -} diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index 0a8bffcd9a..1acb69e8ed 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -90,71 +90,10 @@ enum -#if 1 /*XXX temp */ -/* Hardware version of a parsed fragment program. "Derived" from the - * mesa fragment_program struct. - */ -struct i915_fragment_program -{ -#if 0 - struct gl_fragment_program Base; -#else - uint NumNativeInstructions; - uint NumNativeAluInstructions; - uint NumNativeTexInstructions; - uint NumNativeTexIndirections; -#endif - - boolean error; /**< Set if i915_program_error() is called */ -#if 0 - uint id; /**< String id */ - boolean translated; -#endif - - /* Decls + instructions: - */ - uint program[I915_PROGRAM_SIZE]; - uint program_size; - -#if 0 - /* Constant buffer: - */ - float constant[I915_MAX_CONSTANT][4]; - uint nr_constants; -#endif - - /* Some of which are parameters: - */ - struct - { - uint reg; /* Hardware constant idx */ - const float *values; /* Pointer to tracked values */ - } param[I915_MAX_CONSTANT]; - uint nr_params; - -#if 0 - uint param_state; -#endif - uint wpos_tex; -}; -#endif - - -/*********************************************************************** - * Public interface for the compiler - */ - -void i915_compile_fragment_program( struct i915_context *i915, - struct i915_fragment_program *fp ); - - -/*********************************************************************** - * Private details of the compiler +/** + * Program translation state */ - struct i915_fp_compile { - struct i915_fragment_program *fp; - struct pipe_shader_state *shader; uint declarations[I915_PROGRAM_SIZE]; @@ -186,9 +125,12 @@ struct i915_fp_compile { uint nr_alu_insn; uint nr_decl_insn; -#if 0 - float (*env_param)[4]; -#endif + boolean error; /**< Set if i915_program_error() is called */ + uint wpos_tex; + uint NumNativeInstructions; + uint NumNativeAluInstructions; + uint NumNativeTexInstructions; + uint NumNativeTexIndirections; }; @@ -268,6 +210,14 @@ negate(int reg, int x, int y, int z, int w) } + +/*********************************************************************** + * Public interface for the compiler + */ +extern void i915_translate_fragment_program( struct i915_context *i915 ); + + + extern uint i915_get_temp(struct i915_fp_compile *p); extern uint i915_get_utemp(struct i915_fp_compile *p); extern void i915_release_utemps(struct i915_fp_compile *p); @@ -302,38 +252,21 @@ extern uint i915_emit_const4f(struct i915_fp_compile *p, float c2, float c3); -#if 0 -extern uint i915_emit_param4fv(struct i915_fp_compile *p, - const float * values); -#endif - - - -/*====================================================================== - * i915_fpc_debug.c - */ -extern void i915_program_error(struct i915_fp_compile *p, - const char *msg); - - /*====================================================================== * i915_fpc_debug.c */ extern void i915_disassemble_program(const uint * program, uint sz); -#if 0 -extern void i915_print_mesa_instructions( const struct prog_instruction *insn, - uint nr ); -#endif /*====================================================================== * i915_fpc_translate.c */ -void i915_fixup_depth_write(struct i915_fp_compile *p); extern void -i915_translate_program(struct i915_fp_compile *p, const struct tgsi_token *token); +i915_program_error(struct i915_fp_compile *p, const char *msg); +extern void +i915_translate_fragment_program(struct i915_context *i915); #endif diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index db6b92ad7f..cf6a142075 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -32,6 +32,34 @@ #include "pipe/tgsi/core/tgsi_token.h" #include "pipe/tgsi/core/tgsi_parse.h" +/** + * Simple pass-through fragment shader to use when we don't have + * a real shader (or it fails to compile for some reason). + */ +static unsigned passthrough[] = +{ + _3DSTATE_PIXEL_SHADER_PROGRAM | ((2*3)-1), + + /* declare input color: + */ + (D0_DCL | + (REG_TYPE_T << D0_TYPE_SHIFT) | + (T_DIFFUSE << D0_NR_SHIFT) | + D0_CHANNEL_ALL), + 0, + 0, + + /* move to output color: + */ + (A0_MOV | + (REG_TYPE_OC << A0_DEST_TYPE_SHIFT) | + A0_DEST_CHANNEL_ALL | + (REG_TYPE_T << A0_SRC0_TYPE_SHIFT) | + (T_DIFFUSE << A0_SRC0_NR_SHIFT)), + 0x01230000, /* .xyzw */ + 0 +}; + /* 1, -1/3!, 1/5!, -1/7! */ static const float sin_constants[4] = { 1.0, @@ -48,6 +76,30 @@ static const float cos_constants[4] = { 1.0, }; + +static void +i915_use_passthrough_shader(struct i915_context *i915) +{ + fprintf(stderr, "**** Using i915 pass-through fragment shader\n"); + + i915->current.program = (uint *) malloc(sizeof(passthrough)); + memcpy(i915->current.program, passthrough, sizeof(passthrough)); + i915->current.program_len = Elements(passthrough); + + i915->current.constants = NULL; + i915->current.num_constants = 0; +} + + +void +i915_program_error(struct i915_fp_compile *p, const char *msg) +{ + fprintf(stderr, "i915_program_error: %s\n", msg); + p->error = 1; +} + + + /** * Construct a ureg for the given source register. Will emit * constants, apply swizzling and negation as needed. @@ -59,7 +111,7 @@ src_vector(struct i915_fp_compile *p, const uint index = source->SrcRegister.Index; uint src; - switch (source->SrcRegisterInd.File) { + switch (source->SrcRegister.File) { case TGSI_FILE_TEMPORARY: if (source->SrcRegister.Index >= I915_MAX_TEMPORARY) { i915_program_error(p, "Exceeded max temporary reg"); @@ -79,7 +131,7 @@ src_vector(struct i915_fp_compile *p, */ switch (index) { case FRAG_ATTRIB_WPOS: - src = i915_emit_decl(p, REG_TYPE_T, p->fp->wpos_tex, D0_CHANNEL_ALL); + src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); break; case FRAG_ATTRIB_COL0: src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); @@ -111,27 +163,9 @@ src_vector(struct i915_fp_compile *p, } break; - /* Various parameters and env values. All emitted to - * hardware as program constants. - */ -#if 0 - case PROGRAM_LOCAL_PARAM: - src = i915_emit_param4fv(p, program->Base.LocalParams[index]); - break; - case PROGRAM_ENV_PARAM: - src = i915_emit_param4fv(p, p->env_param[index]); - break; - case PROGRAM_CONSTANT: - case PROGRAM_STATE_VAR: - case PROGRAM_NAMED_PARAM: - src = i915_emit_param4fv( - p, program->Base.Parameters->ParameterValues[index]); - break; -#else case TGSI_FILE_CONSTANT: src = UREG(REG_TYPE_CONST, index); break; -#endif default: i915_program_error(p, "Bad source->File"); @@ -151,26 +185,14 @@ src_vector(struct i915_fp_compile *p, assert(!source->SrcRegisterExtSwz.NegateW); assert(!source->SrcRegisterExtMod.Absolute); assert(!source->SrcRegisterExtMod.Negate); -#if 0 - if (source->SrcRegister.Negate) - negate all - - if (extended source swiz per component) - src = negate(src, - source->SrcRegisterExtSwz.NegateX, - source->SrcRegisterExtSwz.NegateY, - source->SrcRegisterExtSwz.NegateZ, - source->SrcRegisterExtSwz.NegateW); - if (mod.abs) - absolute value - - if (mod.negate) - another negate; -#endif + return src; } +/** + * Construct a ureg for a destination register. + */ static uint get_result_vector(struct i915_fp_compile *p, const struct tgsi_full_dst_register *dest) @@ -178,9 +200,9 @@ get_result_vector(struct i915_fp_compile *p, switch (dest->DstRegister.File) { case TGSI_FILE_OUTPUT: switch (dest->DstRegister.Index) { - case FRAG_RESULT_COLR: + case 1: /*COLOR*/ /*FRAG_RESULT_COLR:*/ return UREG(REG_TYPE_OC, 0); - case FRAG_RESULT_DEPR: + case 0: /*DEPTH*/ /*FRAG_RESULT_DEPR:*/ return UREG(REG_TYPE_OD, 0); default: i915_program_error(p, "Bad inst->DstReg.Index"); @@ -296,12 +318,15 @@ emit_simple_arith(struct i915_fp_compile *p, } -#define EMIT_1ARG_ARITH( OP ) emit_simple_arith(p, inst, OP, 1) -#define EMIT_2ARG_ARITH( OP ) emit_simple_arith(p, inst, OP, 2) -#define EMIT_3ARG_ARITH( OP ) emit_simple_arith(p, inst, OP, 3) - - - +/* + * Translate TGSI instruction to i915 instruction. + * + * Possible concerns: + * + * SIN, COS -- could use another taylor step? + * LIT -- results seem a little different to sw mesa + * LOG -- different to mesa on negative numbers, but this is conformant. + */ static void i915_translate_instruction(struct i915_fp_compile *p, const struct tgsi_full_instruction *inst) @@ -321,7 +346,7 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_ADD: - EMIT_2ARG_ARITH(A0_ADD); + emit_simple_arith(p, inst, A0_ADD, 2); break; case TGSI_OPCODE_CMP: @@ -385,11 +410,11 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_DP3: - EMIT_2ARG_ARITH(A0_DP3); + emit_simple_arith(p, inst, A0_DP3, 2); break; case TGSI_OPCODE_DP4: - EMIT_2ARG_ARITH(A0_DP4); + emit_simple_arith(p, inst, A0_DP4, 2); break; case TGSI_OPCODE_DPH: @@ -431,11 +456,11 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_FLR: - EMIT_1ARG_ARITH(A0_FLR); + emit_simple_arith(p, inst, A0_FLR, 1); break; case TGSI_OPCODE_FRC: - EMIT_1ARG_ARITH(A0_FRC); + emit_simple_arith(p, inst, A0_FRC, 1); break; case TGSI_OPCODE_KIL: @@ -512,11 +537,11 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_MAD: - EMIT_3ARG_ARITH(A0_MAD); + emit_simple_arith(p, inst, A0_MAD, 3); break; case TGSI_OPCODE_MAX: - EMIT_2ARG_ARITH(A0_MAX); + emit_simple_arith(p, inst, A0_MAX, 2); break; case TGSI_OPCODE_MIN: @@ -539,11 +564,11 @@ i915_translate_instruction(struct i915_fp_compile *p, case TGSI_OPCODE_MOV: /* aka TGSI_OPCODE_SWZ */ - EMIT_1ARG_ARITH(A0_MOV); + emit_simple_arith(p, inst, A0_MOV, 1); break; case TGSI_OPCODE_MUL: - EMIT_2ARG_ARITH(A0_MUL); + emit_simple_arith(p, inst, A0_MUL, 2); break; case TGSI_OPCODE_POW: @@ -652,7 +677,7 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_SGE: - EMIT_2ARG_ARITH(A0_SGE); + emit_simple_arith(p, inst, A0_SGE, 2); break; case TGSI_OPCODE_SIN: @@ -706,7 +731,7 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_SLT: - EMIT_2ARG_ARITH(A0_SLT); + emit_simple_arith(p, inst, A0_SLT, 2); break; case TGSI_OPCODE_SUB: @@ -769,20 +794,12 @@ i915_translate_instruction(struct i915_fp_compile *p, /** * Translate TGSI fragment shader into i915 hardware instructions. - * - * Possible concerns: - * - * SIN, COS -- could use another taylor step? - * LIT -- results seem a little different to sw mesa - * LOG -- different to mesa on negative numbers, but this is conformant. - * - * Parse failures -- Mesa doesn't currently give a good indication - * internally whether a particular program string parsed or not. This - * can lead to confusion -- hopefully we cope with it ok now. + * \param p the translation state + * \param tokens the TGSI token array */ -void -i915_translate_program(struct i915_fp_compile *p, - const struct tgsi_token *tokens) +static void +i915_translate_instructions(struct i915_fp_compile *p, + const struct tgsi_token *tokens) { struct tgsi_parse_context parse; @@ -794,10 +811,11 @@ i915_translate_program(struct i915_fp_compile *p, switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: - assert(0); + /* XXX no-op? */ break; case TGSI_TOKEN_TYPE_IMMEDIATE: + /* XXX no-op? */ assert(0); break; @@ -815,13 +833,139 @@ i915_translate_program(struct i915_fp_compile *p, } +static struct i915_fp_compile * +i915_init_compile(struct i915_context *i915, + struct pipe_shader_state *fs) +{ + struct i915_fp_compile *p = CALLOC_STRUCT(i915_fp_compile); + + p->shader = &i915->fs; + + /* a bit of a hack, need to improve constant buffer infrastructure */ + if (i915->fs.constants) + p->constants = i915->fs.constants; + else + p->constants = &i915->temp_constants; + + p->nr_tex_indirect = 1; /* correct? */ + p->nr_tex_insn = 0; + p->nr_alu_insn = 0; + p->nr_decl_insn = 0; + + memset(p->constant_flags, 0, sizeof(p->constant_flags)); + + p->csr = p->program; + p->decl = p->declarations; + p->decl_s = 0; + p->decl_t = 0; + p->temp_flag = 0xffff000; + p->utemp_flag = ~0x7; + + p->wpos_tex = -1; + + /* initialize the first program word */ + *(p->decl++) = _3DSTATE_PIXEL_SHADER_PROGRAM; + + return p; +} + +/* Copy compile results to the fragment program struct and destroy the + * compilation context. + */ +static void +i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) +{ + uint program_size = p->csr - p->program; + uint decl_size = p->decl - p->declarations; + + if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT) + i915_program_error(p, "Exceeded max nr indirect texture lookups"); + + if (p->nr_tex_insn > I915_MAX_TEX_INSN) + i915_program_error(p, "Exceeded max TEX instructions"); + + if (p->nr_alu_insn > I915_MAX_ALU_INSN) + i915_program_error(p, "Exceeded max ALU instructions"); + + if (p->nr_decl_insn > I915_MAX_DECL_INSN) + i915_program_error(p, "Exceeded max DECL instructions"); + + if (p->error) { + p->NumNativeInstructions = 0; + p->NumNativeAluInstructions = 0; + p->NumNativeTexInstructions = 0; + p->NumNativeTexIndirections = 0; + + i915_use_passthrough_shader(i915); + } + else { + p->NumNativeInstructions = (p->nr_alu_insn + + p->nr_tex_insn + + p->nr_decl_insn); + p->NumNativeAluInstructions = p->nr_alu_insn; + p->NumNativeTexInstructions = p->nr_tex_insn; + p->NumNativeTexIndirections = p->nr_tex_indirect; + + /* patch in the program length */ + p->declarations[0] |= program_size + decl_size - 2; + + /* Copy compilation results to fragment program struct: + */ + i915->current.program + = (uint *) malloc((program_size + decl_size) * sizeof(uint)); + i915->current.program_len = program_size + decl_size; + + memcpy(i915->current.program, + p->declarations, + decl_size * sizeof(uint)); + memcpy(i915->current.program + decl_size, + p->program, + program_size * sizeof(uint)); -/* Rather than trying to intercept and jiggle depth writes during + i915->current.constants = (uint *) p->constants->constant; + i915->current.num_constants = p->constants->nr_constants; + } + + /* Release the compilation struct: + */ + free(p); +} + + +/** + * Find an unused texture coordinate slot to use for fragment WPOS. + * Update p->fp->wpos_tex with the result (-1 if no used texcoord slot is found). + */ +static void +i915_find_wpos_space(struct i915_fp_compile *p) +{ + const uint inputs = p->shader->inputs_read; + uint i; + + p->wpos_tex = -1; + + if (inputs & FRAG_BIT_WPOS) { + for (i = 0; i < I915_TEX_UNITS; i++) { + if ((inputs & (FRAG_BIT_TEX0 << i)) == 0) { + p->wpos_tex = i; + return; + } + } + + i915_program_error(p, "No free texcoord for wpos value"); + } +} + + + + +/** + * Rather than trying to intercept and jiggle depth writes during * emit, just move the value into its correct position at the end of * the program: */ -void +static void i915_fixup_depth_write(struct i915_fp_compile *p) { if (p->shader->outputs_written & (1<fs); + const struct tgsi_token *tokens = i915->fs.tokens; + i915_find_wpos_space(p); + i915_translate_instructions(p, tokens); + i915_fixup_depth_write(p); + + i915_fini_compile(i915, p); +} diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 1de6fea2e9..e8ffd1fd7b 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -27,7 +27,7 @@ /* Authors: Keith Whitwell */ -//#include "imports.h" + #include "pipe/draw/draw_context.h" diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index a01b193676..b4ec480f2d 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -149,6 +149,11 @@ void i915_update_derived( struct i915_context *i915 ) if (i915->dirty) i915_update_dynamic( i915 ); + if (i915->dirty & I915_NEW_FS) { + i915_translate_fragment_program(i915); + i915->hardware_dirty |= I915_HW_PROGRAM; /* XXX right? */ + } + /* HW emit currently references framebuffer state directly: */ if (i915->dirty & I915_NEW_FRAMEBUFFER) diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index a7ae92d93c..da9d541598 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -26,10 +26,6 @@ **************************************************************************/ - -//#include "glheader.h" -//#include "mtypes.h" - #include "i915_reg.h" #include "i915_context.h" #include "i915_winsys.h" @@ -116,9 +112,6 @@ i915_emit_hardware_state(struct i915_context *i915 ) } - - - if (i915->hardware_dirty & I915_HW_IMMEDIATE) { OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | @@ -143,8 +136,6 @@ i915_emit_hardware_state(struct i915_context *i915 ) } } - - if (i915->hardware_dirty & I915_HW_STATIC) { if (i915->framebuffer.cbufs[0]) { @@ -180,7 +171,6 @@ i915_emit_hardware_state(struct i915_context *i915 ) I915_BUFFER_ACCESS_WRITE, 0); } - { unsigned cformat = translate_format( i915->framebuffer.cbufs[0]->format ); @@ -198,19 +188,38 @@ i915_emit_hardware_state(struct i915_context *i915 ) cformat | zformat ); } - } + /* constants */ + if (i915->hardware_dirty & I915_HW_PROGRAM) + { + const uint nr = i915->current.num_constants; + if (nr > 0) { + const uint *c = (const uint *) i915->current.constants; + uint i; + OUT_BATCH( _3DSTATE_PIXEL_SHADER_CONSTANTS | (nr * 4) ); + OUT_BATCH( (1 << (nr - 1)) | ((1 << (nr - 1)) - 1) ); + for (i = 0; i < nr; i++) { + OUT_BATCH(*c++); + OUT_BATCH(*c++); + OUT_BATCH(*c++); + OUT_BATCH(*c++); + } + } + } - + /* Fragment program */ + if (i915->hardware_dirty & I915_HW_PROGRAM) { - unsigned i, dwords; - unsigned *prog = i915_passthrough_program( &dwords ); - - for (i = 0; i < dwords; i++) - OUT_BATCH( prog[i] ); + uint i; + /* we should always have, at least, a pass-through program */ + assert(i915->current.program_len > 0); + for (i = 0; i < i915->current.program_len; i++) { + OUT_BATCH(i915->current.program[i]); + } } + /* drawing surface size */ { int w = i915->framebuffer.cbufs[0]->width; int h = i915->framebuffer.cbufs[0]->height; -- cgit v1.2.3 From 88e12872fa3e1d1d2aa33653b3bf2b6982b2fafd Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 21 Aug 2007 20:13:53 -0600 Subject: comments --- src/mesa/pipe/tgsi/core/tgsi_token.h | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index 3ed341fb7b..3e66d55b7d 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -1349,7 +1349,9 @@ struct tgsi_src_register_ext_swz unsigned Extended : 1; /* BOOL */ }; -/* +/** + * Extra src register modifiers + * * If Complement is TRUE, the source register is modified by subtracting it * from 1.0. * @@ -1411,7 +1413,9 @@ struct tgsi_dst_register_ext unsigned Extended : 1; /* BOOL */ }; -/* +/** + * Extra destination register modifiers + * * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_CONDCODE, * it should be cast to tgsi_dst_register_ext_condcode. * @@ -1424,7 +1428,6 @@ struct tgsi_dst_register_ext * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext * follows. */ - struct tgsi_dst_register_ext_concode { unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ -- cgit v1.2.3 From af3d6c83d245c3a5b2af3bddfc261c3678afb7d1 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 21 Aug 2007 20:15:00 -0600 Subject: temporarily call _mesa_enable_sw_extensions() so we can run more test progs without failing the extension checks --- src/mesa/state_tracker/st_context.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 9e89ece52e..42263cab64 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -87,6 +87,9 @@ struct st_context *st_create_context( GLcontext *ctx, st_init_cb_texture( st ); #endif + /* XXXX This is temporary! */ + _mesa_enable_sw_extensions(ctx); + return st; } -- cgit v1.2.3 From 044c19e6b347d13df70dbc9980cec8a464afe214 Mon Sep 17 00:00:00 2001 From: michal Date: Tue, 21 Aug 2007 20:42:33 +0100 Subject: Define __MSC__. Make ALIGN macros more error-proof. --- src/mesa/pipe/p_compiler.h | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h index a4d05f74cd..3dd6bc5c5a 100644 --- a/src/mesa/pipe/p_compiler.h +++ b/src/mesa/pipe/p_compiler.h @@ -38,6 +38,10 @@ #define __WIN32__ #endif +#if defined(_MSC_VER) && !defined(__MSC__) +#define __MSC__ +#endif + typedef unsigned int uint; typedef unsigned char ubyte; @@ -55,8 +59,6 @@ typedef unsigned long long uint64; # define INLINE __inline__ #elif defined(__MSC__) # define INLINE __inline -#elif defined(_MSC_VER) -# define INLINE __inline #elif defined(__ICL) # define INLINE __inline #elif defined(__INTEL_COMPILER) @@ -69,11 +71,11 @@ typedef unsigned long long uint64; #if defined __GNUC__ -#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE] __attribute__(( aligned( 16 ) )) -#define ALIGN16_ASSIGN(P) P +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___aligned[SIZE] __attribute__(( aligned( 16 ) )) +#define ALIGN16_ASSIGN(NAME) NAME##___aligned #else -#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME[SIZE + 1] -#define ALIGN16_ASSIGN(P) align16(P) +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___unaligned[SIZE + 1] +#define ALIGN16_ASSIGN(NAME) align16(NAME##___unaligned) #endif -- cgit v1.2.3 From eb51761b825018bf89080855d0fa3fcb84b9c215 Mon Sep 17 00:00:00 2001 From: michal Date: Tue, 21 Aug 2007 20:45:09 +0100 Subject: Silence compiler warnings. --- src/mesa/pipe/draw/draw_clip.c | 2 +- src/mesa/pipe/draw/draw_twoside.c | 2 +- src/mesa/pipe/softpipe/sp_clear.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 90bfb61568..7e5ceacdfe 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -116,7 +116,7 @@ static void interp( const struct clipper *clip, const float *pos = dst->clip; const float *scale = clip->stage.draw->viewport.scale; const float *trans = clip->stage.draw->viewport.translate; - const float oow = 1.0 / pos[3]; + const float oow = 1.0f / pos[3]; dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 045109dae0..d7d993841b 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -55,7 +55,7 @@ static void twoside_begin( struct draw_stage *stage ) * if the triangle is back-facing (negative). * sign = -1 for CCW, +1 for CW */ - twoside->sign = (stage->draw->setup.front_winding == PIPE_WINDING_CCW) ? -1 : 1; + twoside->sign = (stage->draw->setup.front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index caa6ac1c84..2d5344f424 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -46,7 +46,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); - int x, y, w, h; + unsigned x, y, w, h; softpipe_update_derived(softpipe); /* not needed?? */ -- cgit v1.2.3 From 5c086518f15c2587a4e81c889edb37363015dd0c Mon Sep 17 00:00:00 2001 From: michal Date: Tue, 21 Aug 2007 20:47:31 +0100 Subject: Headers. Fix alignment issues. Silence compiler warnings. More debugging code. --- src/mesa/pipe/draw/draw_prim.c | 62 ++++++++++++++++++++++++++---------------- 1 file changed, 39 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index fd8a39106d..97484d5fb9 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -35,9 +35,7 @@ #include "draw_context.h" #include "draw_prim.h" -#include "pipe/tgsi/core/tgsi_exec.h" -#include "pipe/tgsi/core/tgsi_build.h" -#include "pipe/tgsi/core/tgsi_util.h" +#include "pipe/tgsi/core/tgsi_core.h" #define RP_NONE 0 @@ -138,6 +136,11 @@ run_vertex_program(struct draw_context *draw, struct tgsi_exec_machine machine; unsigned int j; +#if 0 + static FILE *file = NULL; + unsigned i; +#endif + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); const float *scale = draw->viewport.scale; @@ -145,6 +148,12 @@ run_vertex_program(struct draw_context *draw, assert(count <= 4); +#if 0 + if( file == NULL ) { + file = fopen( "vs-exec.txt", "wt" ); + } +#endif + #ifdef DEBUG memset( &machine, 0, sizeof( machine ) ); #endif @@ -162,19 +171,21 @@ run_vertex_program(struct draw_context *draw, machine.Outputs = ALIGN16_ASSIGN(outputs); - if (0) +#if 0 { unsigned attr; for (attr = 0; attr < 16; attr++) { if (draw->vertex_shader.inputs_read & (1 << attr)) { - printf("attr %d: buf_off %d src_off %d pitch %d\n", + unsigned buf = draw->vertex_element[attr].vertex_buffer_index; + fprintf(file, "attr %d: buf_off %d src_off %d pitch %d\n", attr, - draw->vertex_buffer[attr].buffer_offset, + draw->vertex_buffer[buf].buffer_offset, draw->vertex_element[attr].src_offset, - draw->vertex_buffer[attr].pitch); + draw->vertex_buffer[buf].pitch); } } } +#endif /* load machine inputs */ for (j = 0; j < count; j++) { @@ -197,8 +208,9 @@ run_vertex_program(struct draw_context *draw, machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ #if 0 if (attr == 0) { - printf("Input vertex %d: %f %f %f\n", + fprintf(file, "Input vertex %d: %f %f %f\n", j, p[0], p[1], p[2]); + fflush( file ); } #endif } @@ -220,11 +232,14 @@ run_vertex_program(struct draw_context *draw, tgsi_exec_machine_run( &machine ); #if 0 - printf("VS result: %f %f %f %f\n", - outputs[0].xyzw[0].f[0], - outputs[0].xyzw[1].f[0], - outputs[0].xyzw[2].f[0], - outputs[0].xyzw[3].f[0]); + for (i = 0; i < 4; i++) { + fprintf(file, "VS result: %f %f %f %f\n", + machine.Outputs[0].xyzw[0].f[i], + machine.Outputs[0].xyzw[1].f[i], + machine.Outputs[0].xyzw[2].f[i], + machine.Outputs[0].xyzw[3].f[i]); + } + fflush( file ); #endif /* store machine results */ @@ -234,16 +249,16 @@ run_vertex_program(struct draw_context *draw, float x, y, z, w; /* Handle attr[0] (position) specially: */ - x = vOut[j]->clip[0] = outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = outputs[0].xyzw[3].f[j]; + x = vOut[j]->clip[0] = machine.Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine.Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine.Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine.Outputs[0].xyzw[3].f[j]; vOut[j]->clipmask = compute_clipmask(x, y, z, w); vOut[j]->edgeflag = 1; /* divide by w */ - w = 1.0 / w; + w = 1.0f / w; x *= w; y *= w; z *= w; @@ -254,11 +269,12 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][2] = z * scale[2] + trans[2]; vOut[j]->data[0][3] = w; #if 0 - printf("wincoord: %f %f %f %f\n", + fprintf(file, "wincoord: %f %f %f %f\n", vOut[j]->data[0][0], vOut[j]->data[0][1], vOut[j]->data[0][2], vOut[j]->data[0][3]); + fflush( file ); #endif /* remaining attributes: */ @@ -267,10 +283,10 @@ run_vertex_program(struct draw_context *draw, for (attr = 1; attr < VERT_RESULT_MAX; attr++) { if (draw->vertex_shader.outputs_written & (1 << attr)) { assert(slot < draw->nr_attrs); - vOut[j]->data[slot][0] = outputs[attr].xyzw[0].f[j]; - vOut[j]->data[slot][1] = outputs[attr].xyzw[1].f[j]; - vOut[j]->data[slot][2] = outputs[attr].xyzw[2].f[j]; - vOut[j]->data[slot][3] = outputs[attr].xyzw[3].f[j]; + vOut[j]->data[slot][0] = machine.Outputs[attr].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine.Outputs[attr].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine.Outputs[attr].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine.Outputs[attr].xyzw[3].f[j]; slot++; } } -- cgit v1.2.3 From 9606cb15e7d5821b60fd4413e3c79afb745c0e16 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 22 Aug 2007 08:59:51 +0100 Subject: Remove duplicate i915 shader disassembler --- src/mesa/pipe/i915simple/Makefile | 2 +- src/mesa/pipe/i915simple/i915_fpc_debug.c | 346 ------------------------------ 2 files changed, 1 insertion(+), 347 deletions(-) delete mode 100644 src/mesa/pipe/i915simple/i915_fpc_debug.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 2353d0df01..9910a732ba 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -22,7 +22,7 @@ DRIVER_SOURCES = \ i915_strings.c \ i915_prim_emit.c \ i915_tex_layout.c \ - i915_fpc_debug.c \ + i915_fpc.c \ i915_fpc_emit.c \ i915_fpc_translate.c \ i915_surface.c diff --git a/src/mesa/pipe/i915simple/i915_fpc_debug.c b/src/mesa/pipe/i915simple/i915_fpc_debug.c deleted file mode 100644 index 77deab38bb..0000000000 --- a/src/mesa/pipe/i915simple/i915_fpc_debug.c +++ /dev/null @@ -1,346 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#if 0 -#include -#endif - -#include "i915_reg.h" -#include "i915_fpc.h" - -#if 0 -#include "shader/program.h" -#include "shader/prog_instruction.h" -#include "shader/prog_print.h" -#endif - -static const char *opcodes[0x20] = { - "NOP", - "ADD", - "MOV", - "MUL", - "MAD", - "DP2ADD", - "DP3", - "DP4", - "FRC", - "RCP", - "RSQ", - "EXP", - "LOG", - "CMP", - "MIN", - "MAX", - "FLR", - "MOD", - "TRC", - "SGE", - "SLT", - "TEXLD", - "TEXLDP", - "TEXLDB", - "TEXKILL", - "DCL", - "0x1a", - "0x1b", - "0x1c", - "0x1d", - "0x1e", - "0x1f", -}; - - -static const int args[0x20] = { - 0, /* 0 nop */ - 2, /* 1 add */ - 1, /* 2 mov */ - 2, /* 3 m ul */ - 3, /* 4 mad */ - 3, /* 5 dp2add */ - 2, /* 6 dp3 */ - 2, /* 7 dp4 */ - 1, /* 8 frc */ - 1, /* 9 rcp */ - 1, /* a rsq */ - 1, /* b exp */ - 1, /* c log */ - 3, /* d cmp */ - 2, /* e min */ - 2, /* f max */ - 1, /* 10 flr */ - 1, /* 11 mod */ - 1, /* 12 trc */ - 2, /* 13 sge */ - 2, /* 14 slt */ - 1, - 1, - 1, - 1, - 0, - 0, - 0, - 0, - 0, - 0, - 0, -}; - - -static const char *regname[0x8] = { - "R", - "T", - "CONST", - "S", - "OC", - "OD", - "U", - "UNKNOWN", -}; - -static void -print_reg_type_nr(uint type, uint nr) -{ - switch (type) { - case REG_TYPE_T: - switch (nr) { - case T_DIFFUSE: - printf("T_DIFFUSE"); - return; - case T_SPECULAR: - printf("T_SPECULAR"); - return; - case T_FOG_W: - printf("T_FOG_W"); - return; - default: - printf("T_TEX%d", nr); - return; - } - case REG_TYPE_OC: - if (nr == 0) { - printf("oC"); - return; - } - break; - case REG_TYPE_OD: - if (nr == 0) { - printf("oD"); - return; - } - break; - default: - break; - } - - printf("%s[%d]", regname[type], nr); -} - -#define REG_SWIZZLE_MASK 0x7777 -#define REG_NEGATE_MASK 0x8888 - -#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ - (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ - (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ - (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) - - -static void -print_reg_neg_swizzle(uint reg) -{ - int i; - - if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && - (reg & REG_NEGATE_MASK) == 0) - return; - - printf("."); - - for (i = 3; i >= 0; i--) { - if (reg & (1 << ((i * 4) + 3))) - printf("-"); - - switch ((reg >> (i * 4)) & 0x7) { - case 0: - printf("x"); - break; - case 1: - printf("y"); - break; - case 2: - printf("z"); - break; - case 3: - printf("w"); - break; - case 4: - printf("0"); - break; - case 5: - printf("1"); - break; - default: - printf("?"); - break; - } - } -} - - -static void -print_src_reg(uint dword) -{ - uint nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; - uint type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(type, nr); - print_reg_neg_swizzle(dword); -} - - -static void -print_dest_reg(uint dword) -{ - uint nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; - uint type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(type, nr); - if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) - return; - printf("."); - if (dword & A0_DEST_CHANNEL_X) - printf("x"); - if (dword & A0_DEST_CHANNEL_Y) - printf("y"); - if (dword & A0_DEST_CHANNEL_Z) - printf("z"); - if (dword & A0_DEST_CHANNEL_W) - printf("w"); -} - - -#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) -#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) -#define GET_SRC2_REG(r) (r) - - -static void -print_arith_op(uint opcode, const uint * program) -{ - if (opcode != A0_NOP) { - print_dest_reg(program[0]); - if (program[0] & A0_DEST_SATURATE) - printf(" = SATURATE "); - else - printf(" = "); - } - - printf("%s ", opcodes[opcode]); - - print_src_reg(GET_SRC0_REG(program[0], program[1])); - if (args[opcode] == 1) { - printf("\n"); - return; - } - - printf(", "); - print_src_reg(GET_SRC1_REG(program[1], program[2])); - if (args[opcode] == 2) { - printf("\n"); - return; - } - - printf(", "); - print_src_reg(GET_SRC2_REG(program[2])); - printf("\n"); - return; -} - - -static void -print_tex_op(uint opcode, const uint * program) -{ - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - printf(" = "); - - printf("%s ", opcodes[opcode]); - - printf("S[%d],", program[0] & T0_SAMPLER_NR_MASK); - - print_reg_type_nr((program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & - REG_TYPE_MASK, - (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); - printf("\n"); -} - -static void -print_dcl_op(uint opcode, const uint * program) -{ - printf("%s ", opcodes[opcode]); - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - printf("\n"); -} - - -void -i915_disassemble_program(const uint * program, uint sz) -{ - uint size = program[0] & 0x1ff; - int i; - - printf("\t\tBEGIN\n"); - - assert(size + 2 == sz); - - program++; - for (i = 1; i < sz; i += 3, program += 3) { - uint opcode = program[0] & (0x1f << 24); - - printf("\t\t"); - - if ((int) opcode >= A0_NOP && opcode <= A0_SLT) - print_arith_op(opcode >> 24, program); - else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) - print_tex_op(opcode >> 24, program); - else if (opcode == D0_DCL) - print_dcl_op(opcode >> 24, program); - else - printf("Unknown opcode 0x%x\n", opcode); - } - - printf("\t\tEND\n\n"); -} - - -#if 0 -void i915_print_mesa_instructions( const struct prog_instruction *insn, - uint nr ) -{ - uint i; - for (i = 0; i < nr; i++, insn++) { - printf("%3d: ", i); - print_instruction(insn); - } -} -#endif -- cgit v1.2.3 From d021381d4e17d0c802b6e6b1c11de04563274dea Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 22 Aug 2007 09:00:51 +0100 Subject: Clear all the dirtyflag words --- src/mesa/pipe/failover/fo_state_emit.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 7760419394..52fcf5dbc9 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -134,4 +134,8 @@ failover_state_emit( struct failover_context *failover ) } failover->dirty = 0; + failover->dirty_vertex_element = 0; + failover->dirty_vertex_buffer = 0; + failover->dirty_texture = 0; + failover->dirty_sampler = 0; } -- cgit v1.2.3 From 2154d1c5ac0bb37614ef00bed766e49635b519e8 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 09:41:20 -0600 Subject: implement negation for src operands --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 23 ++++++++++++++++++----- 1 file changed, 18 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index cf6a142075..6583cbfa28 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -178,11 +178,24 @@ src_vector(struct i915_fp_compile *p, source->SrcRegister.SwizzleZ, source->SrcRegister.SwizzleW); - assert(!source->SrcRegister.Negate); - assert(!source->SrcRegisterExtSwz.NegateX); - assert(!source->SrcRegisterExtSwz.NegateY); - assert(!source->SrcRegisterExtSwz.NegateZ); - assert(!source->SrcRegisterExtSwz.NegateW); + /* There's both negate-all-components and per-component negation. + * Try to handle both here. + */ + { + int nx = source->SrcRegisterExtSwz.NegateX; + int ny = source->SrcRegisterExtSwz.NegateY; + int nz = source->SrcRegisterExtSwz.NegateZ; + int nw = source->SrcRegisterExtSwz.NegateW; + if (source->SrcRegister.Negate) { + nx = !nx; + ny = !ny; + nz = !nz; + nw = !nw; + } + src = negate(src, nx, ny, nz, nw); + } + + /* no abs() or post-abs negation */ assert(!source->SrcRegisterExtMod.Absolute); assert(!source->SrcRegisterExtMod.Negate); -- cgit v1.2.3 From c8c616d33c2d0b9076a625fdee7f257a042df62f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 09:41:32 -0600 Subject: don't need i915_fpc.c --- src/mesa/pipe/i915simple/Makefile | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 9910a732ba..bb1e6104d8 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -22,7 +22,6 @@ DRIVER_SOURCES = \ i915_strings.c \ i915_prim_emit.c \ i915_tex_layout.c \ - i915_fpc.c \ i915_fpc_emit.c \ i915_fpc_translate.c \ i915_surface.c -- cgit v1.2.3 From 0e8761725536388603b15b093c0293892bd72635 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 09:41:51 -0600 Subject: Obsolete. --- src/mesa/pipe/i915simple/i915_fpc.c | 37 ------------------------------------- 1 file changed, 37 deletions(-) delete mode 100644 src/mesa/pipe/i915simple/i915_fpc.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc.c b/src/mesa/pipe/i915simple/i915_fpc.c deleted file mode 100644 index 3bf2231589..0000000000 --- a/src/mesa/pipe/i915simple/i915_fpc.c +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#if 0 -#include - -#include "glheader.h" -#include "macros.h" -#include "enums.h" -#endif - -#include "i915_fpc.h" - -- cgit v1.2.3 From 4ed87bf2291e4873e94cefe44ddf2b590b09cc42 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 10:04:16 -0600 Subject: remove unneeded includes --- src/mesa/pipe/i915simple/i915_blit.c | 4 ---- src/mesa/pipe/i915simple/i915_fpc_emit.c | 7 ------- src/mesa/pipe/i915simple/i915_prim_emit.c | 3 --- src/mesa/pipe/i915simple/i915_state_derived.c | 14 +++++++------- src/mesa/pipe/i915simple/i915_state_dynamic.c | 5 ----- src/mesa/pipe/i915simple/i915_tex_layout.c | 1 - src/mesa/pipe/i915simple/i915_winsys.h | 1 - 7 files changed, 7 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/mesa/pipe/i915simple/i915_blit.c index 5044cf17f7..0a8a1542d7 100644 --- a/src/mesa/pipe/i915simple/i915_blit.c +++ b/src/mesa/pipe/i915simple/i915_blit.c @@ -26,10 +26,6 @@ **************************************************************************/ -#include - -//#include "mtypes.h" - #include "i915_context.h" #include "i915_winsys.h" #include "i915_blit.h" diff --git a/src/mesa/pipe/i915simple/i915_fpc_emit.c b/src/mesa/pipe/i915simple/i915_fpc_emit.c index 7259bb503d..26a4f36e71 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_emit.c +++ b/src/mesa/pipe/i915simple/i915_fpc_emit.c @@ -25,13 +25,6 @@ * **************************************************************************/ -#if 0 -#include -#include "glheader.h" -#include "macros.h" -#include "enums.h" -#endif - #include "i915_reg.h" #include "i915_context.h" #include "i915_fpc.h" diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index 5a18d03e50..2c5f11c225 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -26,9 +26,6 @@ **************************************************************************/ -//#include "imports.h" -//#include "macros.h" - #include "pipe/draw/draw_private.h" #include "pipe/p_util.h" diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index b4ec480f2d..5a493cc9cf 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -25,21 +25,21 @@ * **************************************************************************/ -#include "main/glheader.h" -#include "main/macros.h" -#include "main/enums.h" - #include "vf/vf.h" #include "pipe/draw/draw_context.h" #include "i915_context.h" #include "i915_state.h" +/* XXX should include i915_fpc.h but that causes some trouble atm */ +extern void i915_translate_fragment_program( struct i915_context *i915 ); + + -#define EMIT_ATTR( VF_ATTR, FRAG_ATTR, INTERP ) \ -do { \ +#define EMIT_ATTR( VF_ATTR, FRAG_ATTR, INTERP ) \ +do { \ slot_to_vf_attr[nr_attrs] = VF_ATTR; \ nr_attrs++; \ - attr_mask |= (1 << (VF_ATTR)); \ + attr_mask |= (1 << (VF_ATTR)); \ } while (0) diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index 84bff17b2f..e648357754 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -25,11 +25,6 @@ * **************************************************************************/ -//#include "glheader.h" -//#include "context.h" -//#include "macros.h" -//#include "enums.h" - #include "i915_batch.h" #include "i915_state_inlines.h" #include "i915_context.h" diff --git a/src/mesa/pipe/i915simple/i915_tex_layout.c b/src/mesa/pipe/i915simple/i915_tex_layout.c index ec1fb9532d..39c3cf2f71 100644 --- a/src/mesa/pipe/i915simple/i915_tex_layout.c +++ b/src/mesa/pipe/i915simple/i915_tex_layout.c @@ -30,7 +30,6 @@ * Michel Dänzer */ -//#include "macros.h" #include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 10a8283035..a95927d803 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -28,7 +28,6 @@ #ifndef I915_WINSYS_H #define I915_WINSYS_H -//#include "main/mtypes.h" /* This is the interface that softpipe requires any window system * hosting it to implement. This is the only include file in softpipe -- cgit v1.2.3 From e7f814e0340454346933f4c8450c07a6c784d87a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 10:04:36 -0600 Subject: remove unneeded include --- src/mesa/pipe/i915simple/i915_state_immediate.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index d06fac2e4c..1cd1866499 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -29,8 +29,6 @@ * Keith Whitwell */ -//#include "macros.h" - #include "i915_state_inlines.h" #include "i915_context.h" #include "i915_state.h" -- cgit v1.2.3 From 5ac22877c98f4934cced4d42b9ec87245f266967 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 10:04:45 -0600 Subject: code movement, null ptr checks, etc --- src/mesa/pipe/i915simple/i915_fpc.h | 11 ------ src/mesa/pipe/i915simple/i915_fpc_translate.c | 48 ++++++++++++++++++++------- 2 files changed, 36 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index 1acb69e8ed..30bc290ad8 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -198,17 +198,6 @@ swizzle(int reg, int x, int y, int z, int w) CHANNEL_SRC(GET_CHANNEL_SRC(reg, w), 3)); } -/* Another neat thing about the UREG representation: - */ -static INLINE int -negate(int reg, int x, int y, int z, int w) -{ - return reg ^ (((x & 1) << UREG_CHANNEL_X_NEGATE_SHIFT) | - ((y & 1) << UREG_CHANNEL_Y_NEGATE_SHIFT) | - ((z & 1) << UREG_CHANNEL_Z_NEGATE_SHIFT) | - ((w & 1) << UREG_CHANNEL_W_NEGATE_SHIFT)); -} - /*********************************************************************** diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 6583cbfa28..db2691ebe1 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -77,14 +77,30 @@ static const float cos_constants[4] = { 1.0, +/** + * component-wise negation of ureg + */ +static INLINE int +negate(int reg, int x, int y, int z, int w) +{ + /* Another neat thing about the UREG representation */ + return reg ^ (((x & 1) << UREG_CHANNEL_X_NEGATE_SHIFT) | + ((y & 1) << UREG_CHANNEL_Y_NEGATE_SHIFT) | + ((z & 1) << UREG_CHANNEL_Z_NEGATE_SHIFT) | + ((w & 1) << UREG_CHANNEL_W_NEGATE_SHIFT)); +} + + static void i915_use_passthrough_shader(struct i915_context *i915) { fprintf(stderr, "**** Using i915 pass-through fragment shader\n"); i915->current.program = (uint *) malloc(sizeof(passthrough)); - memcpy(i915->current.program, passthrough, sizeof(passthrough)); - i915->current.program_len = Elements(passthrough); + if (i915->current.program) { + memcpy(i915->current.program, passthrough, sizeof(passthrough)); + i915->current.program_len = Elements(passthrough); + } i915->current.constants = NULL; i915->current.num_constants = 0; @@ -882,6 +898,7 @@ i915_init_compile(struct i915_context *i915, return p; } + /* Copy compile results to the fragment program struct and destroy the * compilation context. */ @@ -903,6 +920,12 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) if (p->nr_decl_insn > I915_MAX_DECL_INSN) i915_program_error(p, "Exceeded max DECL instructions"); + /* free old program, if present */ + if (i915->current.program) { + free(i915->current.program); + i915->current.program_len = 0; + } + if (p->error) { p->NumNativeInstructions = 0; p->NumNativeAluInstructions = 0; @@ -912,9 +935,8 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) i915_use_passthrough_shader(i915); } else { - p->NumNativeInstructions = (p->nr_alu_insn + - p->nr_tex_insn + - p->nr_decl_insn); + p->NumNativeInstructions + = p->nr_alu_insn + p->nr_tex_insn + p->nr_decl_insn; p->NumNativeAluInstructions = p->nr_alu_insn; p->NumNativeTexInstructions = p->nr_tex_insn; p->NumNativeTexIndirections = p->nr_tex_indirect; @@ -926,15 +948,17 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) */ i915->current.program = (uint *) malloc((program_size + decl_size) * sizeof(uint)); - i915->current.program_len = program_size + decl_size; + if (i915->current.program) { + i915->current.program_len = program_size + decl_size; - memcpy(i915->current.program, - p->declarations, - decl_size * sizeof(uint)); + memcpy(i915->current.program, + p->declarations, + decl_size * sizeof(uint)); - memcpy(i915->current.program + decl_size, - p->program, - program_size * sizeof(uint)); + memcpy(i915->current.program + decl_size, + p->program, + program_size * sizeof(uint)); + } i915->current.constants = (uint *) p->constants->constant; i915->current.num_constants = p->constants->nr_constants; -- cgit v1.2.3 From d1fbf621dc48a488c0f860c5851332d269e6d637 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 22 Aug 2007 10:33:55 +0100 Subject: Import some sampler state code. Doesn't compile, isn't built. --- src/mesa/pipe/i915simple/i915_context.h | 4 + src/mesa/pipe/i915simple/i915_state_emit.c | 23 +++ src/mesa/pipe/i915simple/i915_state_sampler.c | 222 ++++++++++++++++++++++++++ 3 files changed, 249 insertions(+) create mode 100644 src/mesa/pipe/i915simple/i915_state_sampler.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index bc00d61d4b..b40dfa695b 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -89,6 +89,10 @@ struct i915_state uint program_len; uint *constants; uint num_constants; + + unsigned sampler[I915_TEX_UNITS][3]; + unsigned sampler_enable_flags; + unsigned sampler_enable_nr; unsigned id; /* track lost context events */ }; diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index da9d541598..dcf02abb14 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -190,6 +190,29 @@ i915_emit_hardware_state(struct i915_context *i915 ) } } + /* samplers */ + if (i915->hardware_dirty & I915_HW_SAMPLER) + { + if (i915->current.sampler_enable_nr) { + int i; + + OUT_BATCH( _3DSTATE_SAMPLER_STATE | + (3 * i915->current.sampler_enable_nr) ); + + OUT_BATCH( i915->current.sampler_enable_flags ); + + for (i = 0; i < I915_TEX_UNITS; i++) { + if (i915->current.sampler_enable_flags & (1<current.sampler[i][0] ); + OUT_BATCH( i915->current.sampler[i][1] ); + OUT_BATCH( i915->current.sampler[i][2] ); + } + } + } + } + + + /* constants */ if (i915->hardware_dirty & I915_HW_PROGRAM) { diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c new file mode 100644 index 0000000000..3e44bba543 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -0,0 +1,222 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +//#include "mtypes.h" +//#include "enums.h" +//#include "texformat.h" +//#include "macros.h" +//#include "dri_bufmgr.h" + + +#include "pipe/p_context.h" +#include "pipe/p_state.h" + +#include "i915_state_inlines.h" +#include "i915_context.h" +#include "i915_reg.h" +//#include "i915_cache.h" + + + + + + +/* The i915 (and related graphics cores) do not support GL_CLAMP. The + * Intel drivers for "other operating systems" implement GL_CLAMP as + * GL_CLAMP_TO_EDGE, so the same is done here. + */ +static unsigned +translate_wrap_mode(unsigned wrap) +{ + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + return TEXCOORDMODE_WRAP; + case PIPE_TEX_WRAP_CLAMP: + return TEXCOORDMODE_CLAMP_EDGE; /* not quite correct */ + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return TEXCOORDMODE_CLAMP_EDGE; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return TEXCOORDMODE_CLAMP_BORDER; +// case PIPE_TEX_WRAP_MIRRORED_REPEAT: +// return TEXCOORDMODE_MIRROR; + default: + return TEXCOORDMODE_WRAP; + } +} + +static unsigned translate_img_filter( unsigned filter ) +{ + switch (filter) { + case PIPE_TEX_FILTER_NEAREST: + return FILTER_NEAREST; + case PIPE_TEX_FILTER_LINEAR: + return FILTER_LINEAR; + default: + assert(0); + return FILTER_NEAREST; + } +} + +static unsigned translate_mip_filter( unsigned filter ) +{ + switch (filter) { + case PIPE_TEX_MIPFILTER_NONE: + return MIPFILTER_NONE; + case PIPE_TEX_MIPFILTER_NEAREST: + return MIPFILTER_NEAREST; + case PIPE_TEX_FILTER_LINEAR: + return MIPFILTER_LINEAR; + default: + assert(0); + return MIPFILTER_NONE; + } +} + + +/* Recalculate all state from scratch. Perhaps not the most + * efficient, but this has gotten complex enough that we need + * something which is understandable and reliable. + */ +static void update_sampler(struct i915_context *i915, + const struct pipe_sampler_state *sampler, + const struct pipe_surface_state *surface, + unsigned surface_index, + unsigned target, + unsigned *state ) +{ + + /* Need to do this after updating the maps, which call the + * intel_finalize_mipmap_tree and hence can update firstLevel: + */ + unsigned minFilt, magFilt; + unsigned mipFilt; + + state[0] = state[1] = state[2] = 0; + + mipFilt = translate_mip_filter( sampler->min_mip_filter ); + + if (tObj->MaxAnisotropy > 1.0) { + minFilt = FILTER_ANISOTROPIC; + magFilt = FILTER_ANISOTROPIC; + } + else { + minFilt = translate_img_filter( sampler->min_img_filter ); + magFilt = translate_img_filter( sampler->mag_img_filter ); + } + + { + int b = sampler->lod_bias * 16.0; + b = CLAMP(b, -256, 255); + state[0] |= ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); + } + + if (surface->format == MESA_FORMAT_YCBCR || + surface->format == MESA_FORMAT_YCBCR_REV) + state[0] |= SS2_COLORSPACE_CONVERSION; + + + /* Shadow: + */ + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) + { + state[0] |= (SS2_SHADOW_ENABLE | + translate_compare_func(sampler->compare_func)); + + minFilt = FILTER_4X4_FLAT; + magFilt = FILTER_4X4_FLAT; + } + + state[0] |= ((minFilt << SS2_MIN_FILTER_SHIFT) | + (mipFilt << SS2_MIP_FILTER_SHIFT) | + (magFilt << SS2_MAG_FILTER_SHIFT)); + + + if (0) + { + unsigned ws = tObj->WrapS; + unsigned wt = tObj->WrapT; + unsigned wr = tObj->WrapR; + + /* 3D textures don't seem to respect the border color. + * Fallback if there's ever a danger that they might refer to + * it. + * + * Effectively this means fallback on 3D clamp or + * clamp_to_border. + * + * XXX: Check if this is true on i945. + * XXX: Check if this bug got fixed in release silicon. + */ + if (target == PIPE_TEXTURE_3D && + (sampler->min_img_filter != PIPE_TEX_FILTER_NEAREST || + sampler->mag_img_filter != PIPE_TEX_FILTER_NEAREST) && + (ws == GL_CLAMP || + wt == GL_CLAMP || + wr == GL_CLAMP || + ws == GL_CLAMP_TO_BORDER || + wt == GL_CLAMP_TO_BORDER || + wr == GL_CLAMP_TO_BORDER)) { + + if (intel->strict_conformance) { + assert(0); +/* sampler->fallback = true; */ + /* TODO */ + } + } + + + state[1] = + ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT) | + (unit << SS3_TEXTUREMAP_INDEX_SHIFT)); + + if (target != TEXTURE_RECT_BIT) + state[1] |= SS3_NORMALIZED_COORDS; + + state[2] = INTEL_PACKCOLOR8888(tObj->_BorderChan[0], + tObj->_BorderChan[1], + tObj->_BorderChan[2], + tObj->_BorderChan[3]); +} + + + +void i915_update_samplers( struct i915_context *i915 ) +{ + int i, dirty = 0, nr = 0; + + for (i = 0; i < I915_TEX_UNITS; i++) { + if (i915->sampler[i].enabled) { /* ??? */ + update_sampler( i915, i, + i915->current.sampler[i] ); + nr++; + dirty |= (1< Date: Wed, 22 Aug 2007 12:24:51 -0600 Subject: Rework of shader constant buffers. They're now totally independent of the actual shaders. Also, implemented in terms of pipe_buffer_handles/objects. --- src/mesa/pipe/draw/draw_context.h | 2 + src/mesa/pipe/draw/draw_prim.c | 9 ++- src/mesa/pipe/draw/draw_private.h | 2 + src/mesa/pipe/i915simple/i915_context.c | 3 + src/mesa/pipe/i915simple/i915_context.h | 16 ++++-- src/mesa/pipe/i915simple/i915_fpc.h | 8 +-- src/mesa/pipe/i915simple/i915_fpc_emit.c | 81 +++++++-------------------- src/mesa/pipe/i915simple/i915_fpc_translate.c | 20 ++++--- src/mesa/pipe/i915simple/i915_state.c | 40 +++++++++++++ src/mesa/pipe/i915simple/i915_state_emit.c | 6 +- src/mesa/pipe/p_context.h | 6 +- src/mesa/pipe/p_defines.h | 8 +++ src/mesa/pipe/p_state.h | 8 ++- src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_context.h | 7 ++- src/mesa/pipe/softpipe/sp_draw_arrays.c | 31 ++++++++++ src/mesa/pipe/softpipe/sp_quad_fs.c | 3 +- src/mesa/pipe/softpipe/sp_state.h | 4 ++ src/mesa/pipe/softpipe/sp_state_fs.c | 25 ++++++++- src/mesa/state_tracker/st_atom_fs.c | 38 ++++++++++--- src/mesa/state_tracker/st_atom_vs.c | 49 +++++++++------- src/mesa/state_tracker/st_cb_clear.c | 2 - src/mesa/state_tracker/st_cb_drawpixels.c | 2 - src/mesa/state_tracker/st_context.h | 7 ++- src/mesa/state_tracker/st_program.h | 4 +- 25 files changed, 257 insertions(+), 125 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 2fce1322c5..893db1e0c6 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -103,6 +103,8 @@ void draw_set_mapped_element_buffer( struct draw_context *draw, void draw_set_mapped_vertex_buffer(struct draw_context *draw, unsigned attr, const void *buffer); +void draw_set_mapped_constant_buffer(struct draw_context *draw, + const void *buffer); void draw_set_vertex_buffer(struct draw_context *draw, diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 97484d5fb9..94cedb02a9 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -165,7 +165,7 @@ run_vertex_program(struct draw_context *draw, NULL /*samplers*/ ); /* Consts does not require 16 byte alignment. */ - machine.Consts = draw->vertex_shader.constants->constant; + machine.Consts = (float (*)[4]) draw->mapped_constants; machine.Inputs = ALIGN16_ASSIGN(inputs); machine.Outputs = ALIGN16_ASSIGN(outputs); @@ -742,6 +742,13 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw, } +void draw_set_mapped_constant_buffer(struct draw_context *draw, + const void *buffer) +{ + draw->mapped_constants = buffer; +} + + unsigned draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) { diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 85feb14688..75f7c5d84e 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -167,6 +167,8 @@ struct draw_context unsigned eltSize; /**< bytes per index (0, 1, 2 or 4) */ /** The mapped vertex arrays */ const void *mapped_vbuffer[PIPE_ATTRIB_MAX]; + /** The mapped constant buffers (for vertex shader) */ + const void *mapped_constants; /* Clip derived state: */ diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 9856c7c10c..f4121419f7 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -186,6 +186,9 @@ static boolean i915_draw_elements( struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } + draw_set_mapped_constant_buffer(draw, + i915->current.constants[PIPE_SHADER_VERTEX]); + /* draw! */ draw_arrays(i915->draw, prim, start, count); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index b40dfa695b..a037b20289 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -30,10 +30,10 @@ #include "pipe/p_context.h" +#include "pipe/p_defines.h" #include "pipe/p_state.h" - #define I915_TEX_UNITS 8 #define I915_DYNAMIC_MODES4 0 @@ -74,6 +74,7 @@ #define I915_CACHE_CONSTANTS 5 #define I915_MAX_CACHE 6 +#define I915_MAX_CONSTANT 32 struct i915_cache_context; @@ -85,10 +86,14 @@ struct i915_state unsigned immediate[I915_MAX_IMMEDIATE]; unsigned dynamic[I915_MAX_DYNAMIC]; + float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; + /** number of constants passed in through a constant buffer */ + uint num_user_constants[PIPE_SHADER_TYPES]; + /** user constants, plus extra constants from shader translation */ + uint num_constants[PIPE_SHADER_TYPES]; + uint *program; uint program_len; - uint *constants; - uint num_constants; unsigned sampler[I915_TEX_UNITS][3]; unsigned sampler_enable_flags; @@ -112,6 +117,7 @@ struct i915_context struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; + struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; struct pipe_depth_state depth_test; struct pipe_framebuffer_state framebuffer; struct pipe_shader_state fs; @@ -124,8 +130,6 @@ struct i915_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_constant_buffer temp_constants; /*XXX temporary*/ - unsigned dirty; unsigned *batch_start; @@ -156,6 +160,8 @@ struct i915_context #define I915_NEW_SAMPLER 0x400 #define I915_NEW_TEXTURE 0x800 #define I915_NEW_STENCIL 0x1000 +#define I915_NEW_CONSTANTS 0x2000 + /* Driver's internally generated state flags: */ diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index 30bc290ad8..afef706418 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -37,7 +37,6 @@ #define I915_PROGRAM_SIZE 192 -#define I915_MAX_CONSTANT 32 #define MAX_VARYING 8 @@ -99,9 +98,10 @@ struct i915_fp_compile { uint declarations[I915_PROGRAM_SIZE]; uint program[I915_PROGRAM_SIZE]; - uint constant_flags[I915_MAX_CONSTANT]; - - struct pipe_constant_buffer *constants; + /** points into the i915->current.constants array: */ + float (*constants)[4]; + uint num_constants; + uint constant_flags[I915_MAX_CONSTANT]; /**< status of each constant */ uint *csr; /* Cursor, points into program. */ diff --git a/src/mesa/pipe/i915simple/i915_fpc_emit.c b/src/mesa/pipe/i915simple/i915_fpc_emit.c index 26a4f36e71..dfbc5f180c 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_emit.c +++ b/src/mesa/pipe/i915simple/i915_fpc_emit.c @@ -244,25 +244,14 @@ i915_emit_const1f(struct i915_fp_compile * p, float c0) if (p->constant_flags[reg] == I915_CONSTFLAG_PARAM) continue; for (idx = 0; idx < 4; idx++) { -#if 0 - if (!(p->constant_flags[reg] & (1 << idx)) || - p->fp->constant[reg][idx] == c0) { - p->fp->constant[reg][idx] = c0; - p->constant_flags[reg] |= 1 << idx; - if (reg + 1 > p->fp->nr_constants) - p->fp->nr_constants = reg + 1; - return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE); - } -#else if (!(p->constant_flags[reg] & (1 << idx)) || - p->constants->constant[reg][idx] == c0) { - p->constants->constant[reg][idx] = c0; + p->constants[reg][idx] == c0) { + p->constants[reg][idx] = c0; p->constant_flags[reg] |= 1 << idx; - if (reg + 1 > p->constants->nr_constants) - p->constants->nr_constants = reg + 1; + if (reg + 1 > p->num_constants) + p->num_constants = reg + 1; return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE); } -#endif } } @@ -291,23 +280,12 @@ i915_emit_const2f(struct i915_fp_compile * p, float c0, float c1) continue; for (idx = 0; idx < 3; idx++) { if (!(p->constant_flags[reg] & (3 << idx))) { -#if 0 - p->fp->constant[reg][idx] = c0; - p->fp->constant[reg][idx + 1] = c1; + p->constants[reg][idx + 0] = c0; + p->constants[reg][idx + 1] = c1; p->constant_flags[reg] |= 3 << idx; - if (reg + 1 > p->fp->nr_constants) - p->fp->nr_constants = reg + 1; - return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, - ONE); -#else - p->constants->constant[reg][idx + 0] = c0; - p->constants->constant[reg][idx + 1] = c1; - p->constant_flags[reg] |= 3 << idx; - if (reg + 1 > p->constants->nr_constants) - p->constants->nr_constants = reg + 1; - return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, - ONE); -#endif + if (reg + 1 > p->num_constants) + p->num_constants = reg + 1; + return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, ONE); } } } @@ -326,40 +304,21 @@ i915_emit_const4f(struct i915_fp_compile * p, for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { if (p->constant_flags[reg] == 0xf && -#if 0 - p->fp->constant[reg][0] == c0 && - p->fp->constant[reg][1] == c1 && - p->fp->constant[reg][2] == c2 && - p->fp->constant[reg][3] == c3 -#else - p->constants->constant[reg][0] == c0 && - p->constants->constant[reg][1] == c1 && - p->constants->constant[reg][2] == c2 && - p->constants->constant[reg][3] == c3 -#endif - ) { + p->constants[reg][0] == c0 && + p->constants[reg][1] == c1 && + p->constants[reg][2] == c2 && + p->constants[reg][3] == c3) { return UREG(REG_TYPE_CONST, reg); } else if (p->constant_flags[reg] == 0) { -#if 0 - p->fp->constant[reg][0] = c0; - p->fp->constant[reg][1] = c1; - p->fp->constant[reg][2] = c2; - p->fp->constant[reg][3] = c3; -#else - p->constants->constant[reg][0] = c0; - p->constants->constant[reg][1] = c1; - p->constants->constant[reg][2] = c2; - p->constants->constant[reg][3] = c3; -#endif + + p->constants[reg][0] = c0; + p->constants[reg][1] = c1; + p->constants[reg][2] = c2; + p->constants[reg][3] = c3; p->constant_flags[reg] = 0xf; -#if 0 - if (reg + 1 > p->fp->nr_constants) - p->fp->nr_constants = reg + 1; -#else - if (reg + 1 > p->constants->nr_constants) - p->constants->nr_constants = reg + 1; -#endif + if (reg + 1 > p->num_constants) + p->num_constants = reg + 1; return UREG(REG_TYPE_CONST, reg); } } diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index db2691ebe1..dcf0d18f4e 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -102,8 +102,8 @@ i915_use_passthrough_shader(struct i915_context *i915) i915->current.program_len = Elements(passthrough); } - i915->current.constants = NULL; - i915->current.num_constants = 0; + i915->current.num_constants[PIPE_SHADER_FRAGMENT] = 0; + i915->current.num_user_constants[PIPE_SHADER_FRAGMENT] = 0; } @@ -870,11 +870,11 @@ i915_init_compile(struct i915_context *i915, p->shader = &i915->fs; - /* a bit of a hack, need to improve constant buffer infrastructure */ - if (i915->fs.constants) - p->constants = i915->fs.constants; - else - p->constants = &i915->temp_constants; + /* new constants found during translation get appended after the + * user-provided constants. + */ + p->constants = i915->current.constants[PIPE_SHADER_FRAGMENT]; + p->num_constants = i915->current.num_user_constants[PIPE_SHADER_FRAGMENT]; p->nr_tex_indirect = 1; /* correct? */ p->nr_tex_insn = 0; @@ -960,8 +960,10 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) program_size * sizeof(uint)); } - i915->current.constants = (uint *) p->constants->constant; - i915->current.num_constants = p->constants->nr_constants; + /* update number of constants */ + i915->current.num_constants[PIPE_SHADER_FRAGMENT] = p->num_constants; + assert(i915->current.num_constants[PIPE_SHADER_FRAGMENT] + >= i915->current.num_user_constants[PIPE_SHADER_FRAGMENT]); } /* Release the compilation struct: diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index e8ffd1fd7b..8f8b13253d 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -30,6 +30,7 @@ #include "pipe/draw/draw_context.h" +#include "pipe/p_winsys.h" #include "i915_context.h" #include "i915_state.h" @@ -131,6 +132,44 @@ static void i915_set_vs_state( struct pipe_context *pipe, } +static void i915_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct i915_context *i915 = i915_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + /* Make a copy of shader constants. + * During fragment program translation we may add additional + * constants to the array. + * + * We want to consider the situation where some user constants + * (ex: a material color) may change frequently but the shader program + * stays the same. In that case we should only be updating the first + * N constants, leaving any extras from shader translation alone. + */ + { + void *mapped; + if (buf->size && + (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_FLAG_READ))) { + memcpy(i915->current.constants[shader], mapped, buf->size); + fprintf(stderr, "i915 problem: map of constant buffer failed\n"); + ws->buffer_unmap(ws, buf->buffer); + i915->current.num_user_constants[shader] + = buf->size / (4 * sizeof(float)); + } + else { + i915->current.num_user_constants[shader] = 0; + } + } + + i915->dirty |= I915_NEW_CONSTANTS; +} + + static void i915_set_sampler_state(struct pipe_context *pipe, unsigned unit, const struct pipe_sampler_state *sampler) @@ -256,6 +295,7 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_blend_state = i915_set_blend_state; i915->pipe.set_clip_state = i915_set_clip_state; i915->pipe.set_clear_color_state = i915_set_clear_color_state; + i915->pipe.set_constant_buffer = i915_set_constant_buffer; i915->pipe.set_depth_state = i915_set_depth_test_state; i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; i915->pipe.set_fs_state = i915_set_fs_state; diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index dcf02abb14..dff9d40a83 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -216,9 +216,11 @@ i915_emit_hardware_state(struct i915_context *i915 ) /* constants */ if (i915->hardware_dirty & I915_HW_PROGRAM) { - const uint nr = i915->current.num_constants; + const uint nr = i915->current.num_constants[PIPE_SHADER_FRAGMENT]; + assert(nr <= I915_MAX_CONSTANT); if (nr > 0) { - const uint *c = (const uint *) i915->current.constants; + const uint *c + = (const uint *) i915->current.constants[PIPE_SHADER_FRAGMENT]; uint i; OUT_BATCH( _3DSTATE_PIXEL_SHADER_CONSTANTS | (nr * 4) ); OUT_BATCH( (1 << (nr - 1)) | ((1 << (nr - 1)) - 1) ); diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 3a656b272e..0d90a967cc 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -99,8 +99,12 @@ struct pipe_context { void (*set_clear_color_state)( struct pipe_context *, const struct pipe_clear_color_state * ); + void (*set_constant_buffer)( struct pipe_context *, + uint shader, uint index, + const struct pipe_constant_buffer *buf ); + void (*set_depth_state)( struct pipe_context *, - const struct pipe_depth_state * ); + const struct pipe_depth_state * ); void (*set_framebuffer_state)( struct pipe_context *, const struct pipe_framebuffer_state * ); diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 43d1c438ae..636711938c 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -276,6 +276,14 @@ #define PIPE_FLUSH_TEXTURE_CACHE 0x2 +/** + * Shaders + */ +#define PIPE_SHADER_VERTEX 0 +#define PIPE_SHADER_FRAGMENT 1 +#define PIPE_SHADER_TYPES 2 + + /** * Primitive types: */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 42bf50a617..317e8e5634 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -115,9 +115,12 @@ struct pipe_clip_state { }; +/** + * Constants for vertex/fragment shaders + */ struct pipe_constant_buffer { - float constant[PIPE_MAX_CONSTANT][4]; - unsigned nr_constants; + struct pipe_buffer_handle *buffer; + unsigned size; /** in bytes */ }; @@ -125,7 +128,6 @@ struct pipe_shader_state { unsigned inputs_read; /**< FRAG/VERT_ATTRIB_x */ unsigned outputs_written; /**< FRAG/VERT_RESULT_x */ const struct tgsi_token *tokens; - struct pipe_constant_buffer *constants; /* XXX temporary? */ }; struct pipe_depth_state diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index ab9becc99e..5404b7f790 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -237,6 +237,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_blend_state = softpipe_set_blend_state; softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; + softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; softpipe->pipe.set_depth_state = softpipe_set_depth_test_state; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_fs_state = softpipe_set_fs_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 14ae9f2105..6458573917 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -33,6 +33,7 @@ #include "pipe/p_state.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" #include "sp_quad.h" @@ -64,6 +65,7 @@ enum interp_mode { #define SP_NEW_STENCIL 0x1000 #define SP_NEW_VERTEX 0x2000 #define SP_NEW_VS 0x4000 +#define SP_NEW_CONSTANTS 0x8000 struct softpipe_context { @@ -78,6 +80,7 @@ struct softpipe_context { struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; + struct pipe_constant_buffer constants[2]; struct pipe_depth_state depth_test; struct pipe_framebuffer_state framebuffer; struct pipe_shader_state fs; @@ -105,7 +108,9 @@ struct softpipe_context { * Mapped vertex buffers */ ubyte *mapped_vbuffer[PIPE_ATTRIB_MAX]; - + + /** Mapped constant buffers */ + void *mapped_constants[PIPE_SHADER_TYPES]; /* FS + setup derived state: */ diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 0392b6b64e..5198a493da 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -43,6 +43,34 @@ +static void +softpipe_map_constant_buffers(struct softpipe_context *sp) +{ + struct pipe_winsys *ws = sp->pipe.winsys; + uint i; + for (i = 0; i < 2; i++) { + if (sp->constants[i].size) + sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, + PIPE_BUFFER_FLAG_READ); + } + + draw_set_mapped_constant_buffer(sp->draw, + sp->mapped_constants[PIPE_SHADER_VERTEX]); +} + +static void +softpipe_unmap_constant_buffers(struct softpipe_context *sp) +{ + struct pipe_winsys *ws = sp->pipe.winsys; + uint i; + for (i = 0; i < 2; i++) { + if (sp->constants[i].size) + ws->buffer_unmap(ws, sp->constants[i].buffer); + sp->mapped_constants[i] = NULL; + } +} + + boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) @@ -81,6 +109,8 @@ softpipe_draw_elements(struct pipe_context *pipe, softpipe_map_surfaces(sp); + softpipe_map_constant_buffers(sp); + /* * Map vertex buffers */ @@ -123,6 +153,7 @@ softpipe_draw_elements(struct pipe_context *pipe, } softpipe_unmap_surfaces(sp); + softpipe_unmap_constant_buffers(sp); return TRUE; } diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index c20f09d309..ceba94aa94 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -33,6 +33,7 @@ */ #include "pipe/p_util.h" +#include "pipe/p_defines.h" #include "sp_context.h" #include "sp_headers.h" @@ -83,7 +84,7 @@ shade_quad( qss->samplers ); /* Consts does not require 16 byte alignment. */ - machine.Consts = softpipe->fs.constants->constant; + machine.Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; machine.Inputs = ALIGN16_ASSIGN(inputs); machine.Outputs = ALIGN16_ASSIGN(outputs); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 5e1ecd94e0..354580b2a5 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -52,6 +52,10 @@ void softpipe_set_clear_color_state( struct pipe_context *, void softpipe_set_clip_state( struct pipe_context *, const struct pipe_clip_state * ); +void softpipe_set_constant_buffer(struct pipe_context *, + uint shader, uint index, + const struct pipe_constant_buffer *buf); + void softpipe_set_depth_test_state( struct pipe_context *, const struct pipe_depth_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index bab407f047..9e3ff6d35c 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -28,6 +28,8 @@ #include "sp_context.h" #include "sp_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #include "pipe/draw/draw_context.h" @@ -53,3 +55,24 @@ void softpipe_set_vs_state( struct pipe_context *pipe, draw_set_vertex_shader(softpipe->draw, vs); } + + +void softpipe_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + /* note: reference counting */ + ws->buffer_unreference(ws, &softpipe->constants[shader].buffer); + softpipe->constants[shader].buffer = ws->buffer_reference(ws, buf->buffer); + softpipe->constants[shader].size = buf->size; + + softpipe->dirty |= SP_NEW_CONSTANTS; +} + + diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 9aba9aaa56..019b6457e0 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -32,6 +32,8 @@ #include "shader/prog_parameter.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/core/tgsi_dump.h" @@ -53,12 +55,36 @@ static void compile_fs( struct st_context *st, } +static void +update_fs_constants(struct st_context *st, + struct gl_program_parameter_list *params) + +{ + const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4; + struct pipe_winsys *ws = st->pipe->winsys; + struct pipe_constant_buffer *cbuf + = &st->state.constants[PIPE_SHADER_FRAGMENT]; + + if (!cbuf->buffer) + cbuf->buffer = ws->buffer_create(ws, 1); + + /* load Mesa constants into the constant buffer */ + if (paramBytes) + ws->buffer_data(ws, cbuf->buffer, paramBytes, params->ParameterValues); + + cbuf->size = paramBytes; + + st->pipe->set_constant_buffer(st->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf); +} + + static void update_fs( struct st_context *st ) { struct pipe_shader_state fs; struct st_fragment_program *fp = NULL; struct gl_program_parameter_list *params = NULL; + /* find active shader and params */ if (st->ctx->Shader.CurrentProgram && st->ctx->Shader.CurrentProgram->LinkStatus && st->ctx->Shader.CurrentProgram->FragmentProgram) { @@ -72,25 +98,21 @@ static void update_fs( struct st_context *st ) params = st->ctx->FragmentProgram._Current->Base.Parameters; } + /* update constants */ if (fp && params) { - /* load program's constants array */ - _mesa_load_state_parameters(st->ctx, params); - - fp->constants.nr_constants = params->NumParameters; - memcpy(fp->constants.constant, - params->ParameterValues, - params->NumParameters * sizeof(GLfloat) * 4); + update_fs_constants(st, params); } + /* translate shader to TGSI format */ if (fp->dirty) compile_fs( st, fp ); + /* update pipe state */ memset( &fs, 0, sizeof(fs) ); fs.inputs_read = fp->Base.Base.InputsRead; fs.outputs_written = fp->Base.Base.OutputsWritten; fs.tokens = &fp->tokens[0]; - fs.constants = &fp->constants; if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0 || fp->dirty) diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index c8bd805e02..8a38020afd 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -34,6 +34,8 @@ #include "tnl/t_vp_build.h" #include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/core/tgsi_dump.h" @@ -56,17 +58,36 @@ static void compile_vs( struct st_context *st, } +static void +update_vs_constants(struct st_context *st, + struct gl_program_parameter_list *params) + +{ + const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4; + struct pipe_winsys *ws = st->pipe->winsys; + struct pipe_constant_buffer *cbuf + = &st->state.constants[PIPE_SHADER_VERTEX]; + + if (!cbuf->buffer) + cbuf->buffer = ws->buffer_create(ws, 1); + + /* load Mesa constants into the constant buffer */ + if (paramBytes) + ws->buffer_data(ws, cbuf->buffer, paramBytes, params->ParameterValues); + + cbuf->size = paramBytes; + + st->pipe->set_constant_buffer(st->pipe, PIPE_SHADER_VERTEX, 0, cbuf); +} + + static void update_vs( struct st_context *st ) { struct pipe_shader_state vs; struct st_vertex_program *vp = NULL; struct gl_program_parameter_list *params = NULL; -#if 0 - if (st->ctx->VertexProgram._MaintainTnlProgram) - _tnl_UpdateFixedFunctionProgram( st->ctx ); -#endif - + /* find active shader and params */ if (st->ctx->Shader.CurrentProgram && st->ctx->Shader.CurrentProgram->LinkStatus && st->ctx->Shader.CurrentProgram->VertexProgram) { @@ -80,31 +101,21 @@ static void update_vs( struct st_context *st ) params = st->ctx->VertexProgram._Current->Base.Parameters; } - /* XXXX temp */ -#if 1 - if (!vp) - return; -#endif + /* update constants */ if (vp && params) { - /* load program's constants array */ - - /* XXX this should probably be done elsewhere/separately */ _mesa_load_state_parameters(st->ctx, params); - - vp->constants.nr_constants = params->NumParameters; - memcpy(vp->constants.constant, - params->ParameterValues, - params->NumParameters * sizeof(GLfloat) * 4); + update_vs_constants(st, params); } + /* translate shader to TGSI format */ if (vp->dirty) compile_vs( st, vp ); + /* update pipe state */ memset( &vs, 0, sizeof(vs) ); vs.outputs_written = vp->Base.Base.OutputsWritten; vs.inputs_read = vp->Base.Base.InputsRead; vs.tokens = &vp->tokens[0]; - vs.constants = &vp->constants; if (memcmp(&vs, &st->state.vs, sizeof(vs)) != 0 || vp->dirty) diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 2f2d11c014..e3690deb5a 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -350,7 +350,6 @@ clear_with_quad(GLcontext *ctx, memset(&fs, 0, sizeof(fs)); fs.inputs_read = stfp->Base.Base.InputsRead; fs.tokens = &stfp->tokens[0]; - fs.constants = NULL; pipe->set_fs_state(pipe, &fs); } @@ -365,7 +364,6 @@ clear_with_quad(GLcontext *ctx, vs.inputs_read = stvp->Base.Base.InputsRead; vs.outputs_written = stvp->Base.Base.OutputsWritten; vs.tokens = &stvp->tokens[0]; - vs.constants = NULL; pipe->set_vs_state(pipe, &vs); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index d3f060e691..a45700e1db 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -332,7 +332,6 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, memset(&fs, 0, sizeof(fs)); fs.inputs_read = stfp->Base.Base.InputsRead; fs.tokens = &stfp->tokens[0]; - fs.constants = NULL; pipe->set_fs_state(pipe, &fs); } @@ -347,7 +346,6 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, vs.inputs_read = stvp->Base.Base.InputsRead; vs.outputs_written = stvp->Base.Base.OutputsWritten; vs.tokens = &stvp->tokens[0]; - vs.constants = NULL; pipe->set_vs_state(pipe, &vs); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 35774a790e..38d6a3ed86 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -71,16 +71,17 @@ struct st_context struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; + struct pipe_constant_buffer constants[2]; struct pipe_depth_state depth; struct pipe_framebuffer_state framebuffer; - struct pipe_shader_state fs; - struct pipe_shader_state vs; + struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_scissor_state scissor; struct pipe_setup_state setup; + struct pipe_shader_state fs; + struct pipe_shader_state vs; struct pipe_stencil_state stencil; - struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; } state; diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index b077fdf069..3ff4f4e9c7 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -53,8 +53,6 @@ struct st_fragment_program struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; - struct pipe_constant_buffer constants; - #if 0 GLfloat (*cbuffer)[4]; GLuint nr_constants; @@ -85,7 +83,9 @@ struct st_vertex_program struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; +#if 0 struct pipe_constant_buffer constants; +#endif GLuint param_state; }; -- cgit v1.2.3 From 8a868919b50bdca4dc697d61e220a8fb50764f8e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 12:41:59 -0600 Subject: Improved pipe_region/surface_reference() functions Now dereferences the old object first. Target object may be NULL to clear the pointer. --- src/mesa/pipe/p_context.h | 66 +++++++++++++++++++++++++------------- src/mesa/state_tracker/st_cb_fbo.c | 2 +- 2 files changed, 45 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 0d90a967cc..ec9973383c 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -213,38 +213,60 @@ struct pipe_context { }; - +/** + * Set 'ptr' to point to 'region' and update reference counting. + * The old thing pointed to, if any, will be unreferenced first. + * 'region' may be NULL. + */ static INLINE void -pipe_region_reference(struct pipe_region **dst, struct pipe_region *src) +pipe_region_reference(struct pipe_region **ptr, struct pipe_region *region) { - assert(*dst == NULL); - if (src) { - src->refcount++; - *dst = src; + assert(ptr); + if (*ptr) { + /* unreference the old thing */ + struct pipe_region *oldReg = *ptr; + oldReg->refcount--; + assert(oldReg->refcount >= 0); + if (oldReg->refcount == 0) { + /* free the old region */ + assert(oldReg->map_refcount == 0); + /* XXX dereference the region->buffer */ + free(oldReg); + } + *ptr = NULL; + } + if (region) { + /* reference the new thing */ + region->refcount++; + *ptr = region; } } +/** + * \sa pipe_region_reference + */ static INLINE void -pipe_surface_reference(struct pipe_surface **dst, struct pipe_surface *src) +pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) { - assert(*dst == NULL); - if (src) { - src->refcount++; - *dst = src; + assert(ptr); + if (*ptr) { + /* unreference the old thing */ + struct pipe_surface *oldSurf = *ptr; + oldSurf->refcount--; + assert(oldSurf->refcount >= 0); + if (oldSurf->refcount == 0) { + /* free the old region */ + pipe_region_reference(&oldSurf->region, NULL); + free(oldSurf); + } + *ptr = NULL; } -} - -static INLINE void -pipe_surface_unreference(struct pipe_surface **ps) -{ - assert(*ps); - (*ps)->refcount--; - if ((*ps)->refcount <= 0) { - /* XXX need a proper surface->free method */ - free(*ps); + if (surf) { + /* reference the new thing */ + surf->refcount++; + *ptr = surf; } - *ps = NULL; } diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 2b9aa3e9d2..430ac715e5 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -351,7 +351,7 @@ st_finish_render_texture(GLcontext *ctx, printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface); */ - pipe_surface_unreference(&strb->surface); + pipe_surface_reference(&strb->surface, NULL); _mesa_reference_renderbuffer(&att->Renderbuffer, NULL); -- cgit v1.2.3 From d45dd5672ae60f1f49648725974693bf4ebf4bf8 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 13:30:26 -0600 Subject: added border_color to sampler state --- src/mesa/pipe/p_state.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 317e8e5634..d29b85117b 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -222,8 +222,8 @@ struct pipe_sampler_state #if 0 /* need these? */ int BaseLevel; /**< min mipmap level, OpenGL 1.2 */ int MaxLevel; /**< max mipmap level, OpenGL 1.2 */ - float border_color[4]; #endif + float border_color[4]; float max_anisotropy; }; -- cgit v1.2.3 From be673e9719593ee22c16e289fe4d154db3fb7805 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 13:33:19 -0600 Subject: Sampler state code now compiles and is built. Not tested yet. --- src/mesa/pipe/i915simple/Makefile | 1 + src/mesa/pipe/i915simple/i915_state.h | 1 + src/mesa/pipe/i915simple/i915_state_sampler.c | 185 +++++++++++++++++--------- 3 files changed, 126 insertions(+), 61 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index bb1e6104d8..7e4e3d413e 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -19,6 +19,7 @@ DRIVER_SOURCES = \ i915_state_derived.c \ i915_state_emit.c \ i915_state_fragprog.c \ + i915_state_sampler.c \ i915_strings.c \ i915_prim_emit.c \ i915_tex_layout.c \ diff --git a/src/mesa/pipe/i915simple/i915_state.h b/src/mesa/pipe/i915simple/i915_state.h index 794a251ad8..43058451f2 100644 --- a/src/mesa/pipe/i915simple/i915_state.h +++ b/src/mesa/pipe/i915simple/i915_state.h @@ -42,6 +42,7 @@ struct i915_tracked_state { void i915_update_immediate( struct i915_context *i915 ); void i915_update_dynamic( struct i915_context *i915 ); void i915_update_derived( struct i915_context *i915 ); +void i915_update_samplers( struct i915_context *i915 ); void i915_emit_hardware_state( struct i915_context *i915 ); diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 3e44bba543..591e3684da 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -34,6 +34,7 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "pipe/p_util.h" #include "i915_state_inlines.h" #include "i915_context.h" @@ -96,18 +97,69 @@ static unsigned translate_mip_filter( unsigned filter ) } } +static unsigned translate_compare_func(unsigned func) +{ + switch (func) { + case PIPE_FUNC_NEVER: + case PIPE_FUNC_LESS: + case PIPE_FUNC_EQUAL: + case PIPE_FUNC_LEQUAL: + case PIPE_FUNC_GREATER: + case PIPE_FUNC_NOTEQUAL: + case PIPE_FUNC_GEQUAL: + case PIPE_FUNC_ALWAYS: + return 0; + default: + assert(0); + return 0; + } +} + -/* Recalculate all state from scratch. Perhaps not the most +static uint +bitcount(uint k) +{ + uint count = 0; + while (k) { + if (k & 1) + count++; + k = k >> 1; + } + return count; +} + + +static boolean +is_power_of_two_texture(const struct pipe_mipmap_tree *mt) +{ + if (bitcount(mt->width0) == 1 && + bitcount(mt->height0) == 1 && + bitcount(mt->depth0) == 1) { + return 1; + } + else + return 0; +} + + +/** + * Compute i915 texture sampling state. + * + * Recalculate all state from scratch. Perhaps not the most * efficient, but this has gotten complex enough that we need * something which is understandable and reliable. + * \param state returns the 3 words of compute state */ -static void update_sampler(struct i915_context *i915, +static void update_sampler(struct i915_context *i915, + uint unit, const struct pipe_sampler_state *sampler, - const struct pipe_surface_state *surface, - unsigned surface_index, - unsigned target, - unsigned *state ) + const struct pipe_mipmap_tree *mt, + const struct pipe_surface *surface, + unsigned state[3] ) { + const unsigned ws = sampler->wrap_s; + const unsigned wt = sampler->wrap_t; + const unsigned wr = sampler->wrap_r; /* Need to do this after updating the maps, which call the * intel_finalize_mipmap_tree and hence can update firstLevel: @@ -119,7 +171,7 @@ static void update_sampler(struct i915_context *i915, mipFilt = translate_mip_filter( sampler->min_mip_filter ); - if (tObj->MaxAnisotropy > 1.0) { + if (sampler->max_anisotropy > 1.0) { minFilt = FILTER_ANISOTROPIC; magFilt = FILTER_ANISOTROPIC; } @@ -134,8 +186,8 @@ static void update_sampler(struct i915_context *i915, state[0] |= ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); } - if (surface->format == MESA_FORMAT_YCBCR || - surface->format == MESA_FORMAT_YCBCR_REV) + if (surface->format == PIPE_FORMAT_YCBCR || + surface->format == PIPE_FORMAT_YCBCR_REV) state[0] |= SS2_COLORSPACE_CONVERSION; @@ -155,68 +207,79 @@ static void update_sampler(struct i915_context *i915, (magFilt << SS2_MAG_FILTER_SHIFT)); - if (0) - { - unsigned ws = tObj->WrapS; - unsigned wt = tObj->WrapT; - unsigned wr = tObj->WrapR; - - /* 3D textures don't seem to respect the border color. - * Fallback if there's ever a danger that they might refer to - * it. - * - * Effectively this means fallback on 3D clamp or - * clamp_to_border. - * - * XXX: Check if this is true on i945. - * XXX: Check if this bug got fixed in release silicon. - */ - if (target == PIPE_TEXTURE_3D && - (sampler->min_img_filter != PIPE_TEX_FILTER_NEAREST || - sampler->mag_img_filter != PIPE_TEX_FILTER_NEAREST) && - (ws == GL_CLAMP || - wt == GL_CLAMP || - wr == GL_CLAMP || - ws == GL_CLAMP_TO_BORDER || - wt == GL_CLAMP_TO_BORDER || - wr == GL_CLAMP_TO_BORDER)) { - - if (intel->strict_conformance) { - assert(0); -/* sampler->fallback = true; */ - /* TODO */ - } + /* 3D textures don't seem to respect the border color. + * Fallback if there's ever a danger that they might refer to + * it. + * + * Effectively this means fallback on 3D clamp or + * clamp_to_border. + * + * XXX: Check if this is true on i945. + * XXX: Check if this bug got fixed in release silicon. + */ + if (mt->target == PIPE_TEXTURE_3D && + (sampler->min_img_filter != PIPE_TEX_FILTER_NEAREST || + sampler->mag_img_filter != PIPE_TEX_FILTER_NEAREST) && + (ws == PIPE_TEX_WRAP_CLAMP || + wt == PIPE_TEX_WRAP_CLAMP || + wr == PIPE_TEX_WRAP_CLAMP || + ws == PIPE_TEX_WRAP_CLAMP_TO_BORDER || + wt == PIPE_TEX_WRAP_CLAMP_TO_BORDER || + wr == PIPE_TEX_WRAP_CLAMP_TO_BORDER)) { +#if 0 + if (i915->strict_conformance) { + assert(0); + /* sampler->fallback = true; */ + /* TODO */ } +#endif + } + state[1] = + ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT) | + (unit << SS3_TEXTUREMAP_INDEX_SHIFT)); - state[1] = - ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) | - (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | - (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT) | - (unit << SS3_TEXTUREMAP_INDEX_SHIFT)); - - if (target != TEXTURE_RECT_BIT) - state[1] |= SS3_NORMALIZED_COORDS; + if (is_power_of_two_texture(mt)) { + state[1] |= SS3_NORMALIZED_COORDS; + } - state[2] = INTEL_PACKCOLOR8888(tObj->_BorderChan[0], - tObj->_BorderChan[1], - tObj->_BorderChan[2], - tObj->_BorderChan[3]); + { + ubyte r = float_to_ubyte(sampler->border_color[0]); + ubyte g = float_to_ubyte(sampler->border_color[1]); + ubyte b = float_to_ubyte(sampler->border_color[2]); + ubyte a = float_to_ubyte(sampler->border_color[3]); + state[2] = I915PACKCOLOR8888(r, g, b, a); + } } void i915_update_samplers( struct i915_context *i915 ) { - int i, dirty = 0, nr = 0; - - for (i = 0; i < I915_TEX_UNITS; i++) { - if (i915->sampler[i].enabled) { /* ??? */ - update_sampler( i915, i, - i915->current.sampler[i] ); - nr++; - dirty |= (1<framebuffer.cbufs[0]; + uint unit; + + i915->current.sampler_enable_nr = 0; + i915->current.sampler_enable_flags = 0x0; + + for (unit = 0; unit < I915_TEX_UNITS; unit++) { + /* determine unit enable/disable by looking for a bound mipmap tree */ + /* could also examine the fragment program? */ + if (i915->texture[unit]) { + update_sampler( i915, + unit, + i915->sampler + unit, /* sampler state */ + i915->texture[unit], /* mipmap tree */ + surface, /* cbuffer info */ + i915->current.sampler[unit] /* the result */ + ); + + i915->current.sampler_enable_nr++; + i915->current.sampler_enable_flags |= (1 << unit); } } -} + i915->hardware_dirty |= I915_HW_SAMPLER; +} -- cgit v1.2.3 From d866e648aee69f06c59ecd65e101dd3d618bbd58 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 22 Aug 2007 16:17:41 +0100 Subject: New file. --- src/mesa/pipe/tgsi/deco/tgsi_deco.h | 0 1 file changed, 0 insertions(+), 0 deletions(-) create mode 100644 src/mesa/pipe/tgsi/deco/tgsi_deco.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/deco/tgsi_deco.h b/src/mesa/pipe/tgsi/deco/tgsi_deco.h new file mode 100644 index 0000000000..e69de29bb2 -- cgit v1.2.3 From fa8c74fcc93717da7a2e409d7e1a4ff593a35d53 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 22 Aug 2007 16:24:35 +0100 Subject: New file. --- src/mesa/pipe/tgsi/deco/Makefile | 3 +++ 1 file changed, 3 insertions(+) create mode 100644 src/mesa/pipe/tgsi/deco/Makefile (limited to 'src') diff --git a/src/mesa/pipe/tgsi/deco/Makefile b/src/mesa/pipe/tgsi/deco/Makefile new file mode 100644 index 0000000000..eb8b14e0e8 --- /dev/null +++ b/src/mesa/pipe/tgsi/deco/Makefile @@ -0,0 +1,3 @@ +default: + cd ../../.. ; make + -- cgit v1.2.3 From 6b6cc4c0fe04884b22f6f0010be9dcee2b4efa9f Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 22 Aug 2007 16:38:00 +0100 Subject: Include headers. --- src/mesa/pipe/tgsi/deco/deco_caps.h | 15 +++++++++++++++ src/mesa/pipe/tgsi/deco/tgsi_deco.h | 8 ++++++++ 2 files changed, 23 insertions(+) create mode 100644 src/mesa/pipe/tgsi/deco/deco_caps.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/deco/deco_caps.h b/src/mesa/pipe/tgsi/deco/deco_caps.h new file mode 100644 index 0000000000..b3a40b03d8 --- /dev/null +++ b/src/mesa/pipe/tgsi/deco/deco_caps.h @@ -0,0 +1,15 @@ +#if !defined DECO_CAPS_H +#define DECO_CAPS_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + + + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined DECO_CAPS_H + diff --git a/src/mesa/pipe/tgsi/deco/tgsi_deco.h b/src/mesa/pipe/tgsi/deco/tgsi_deco.h index e69de29bb2..06a2654fb6 100644 --- a/src/mesa/pipe/tgsi/deco/tgsi_deco.h +++ b/src/mesa/pipe/tgsi/deco/tgsi_deco.h @@ -0,0 +1,8 @@ +#if !defined TGSI_DECO_H +#define TGSI_DECO_H + +#include "../core/tgsi_core.h" +#include "deco_caps.h" + +#endif // !defined TGSI_DECO_H + -- cgit v1.2.3 From 89c0d32b458aa467987a42f2c634900729b4df36 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 22 Aug 2007 17:21:02 +0100 Subject: First pass on tgsi deco. --- src/mesa/pipe/tgsi/deco/deco_caps.c | 10 ++++ src/mesa/pipe/tgsi/deco/deco_caps.h | 97 +++++++++++++++++++++++++++++++++++++ 2 files changed, 107 insertions(+) create mode 100644 src/mesa/pipe/tgsi/deco/deco_caps.c (limited to 'src') diff --git a/src/mesa/pipe/tgsi/deco/deco_caps.c b/src/mesa/pipe/tgsi/deco/deco_caps.c new file mode 100644 index 0000000000..66df05a14b --- /dev/null +++ b/src/mesa/pipe/tgsi/deco/deco_caps.c @@ -0,0 +1,10 @@ +#include "tgsi_platform.h" +#include "tgsi_deco.h" + +void +tgsi_deco_caps_init( + struct tgsi_deco_caps *caps ) +{ + memset( caps, 0, sizeof( *caps ) ); +} + diff --git a/src/mesa/pipe/tgsi/deco/deco_caps.h b/src/mesa/pipe/tgsi/deco/deco_caps.h index b3a40b03d8..40ca1c96f3 100644 --- a/src/mesa/pipe/tgsi/deco/deco_caps.h +++ b/src/mesa/pipe/tgsi/deco/deco_caps.h @@ -5,7 +5,104 @@ extern "C" { #endif // defined __cplusplus +struct tgsi_deco_caps +{ + /* + * Predicates (D3D9-specific). + * + * Constraints: + * 1. Token tgsi_dst_register_ext_predicate must not be used. + * 2. Token tgsi_instruction_ext_predicate must not be used. + */ + unsigned Predicates : 1; + /* + * Destination register post-modulate. + * + * Constraints: + * 1. Field tgsi_dst_register_ext_modulate::Modulate + * must be set to TGSI_MODULATE_1X. + */ + unsigned DstModulate : 1; + + /* + * Condition codes (NVIDIA-specific). + * + * Constraints: + * 1. Token tgsi_dst_register_ext_concode must not be used. + * 2. Field tgsi_instruction_ext_nv::CondDstUpdate must be set to FALSE. + * 3. Field tgsi_instruction_ext_nv::CondFlowEnable must be set to FALSE. + */ + unsigned ConCodes : 1; + + /* + * Source register invert. + * + * Constraints: + * 1. Field tgsi_src_register_ext_mod::Complement must be set to FALSE. + */ + unsigned SrcInvert : 1; + + /* + * Source register bias. + * + * Constraints: + * 1. Field tgsi_src_register_ext_mod::Bias must be set to FALSE. + */ + unsigned SrcBias : 1; + + /* + * Source register scale by 2. + * + * Constraints: + * 1. Field tgsi_src_register_ext_mod::Scale2X must be set to FALSE. + */ + unsigned SrcScale : 1; + + /* + * Source register absolute. + * + * Constraints: + * 1. Field tgsi_src_register_ext_mod::Absolute must be set to FALSE. + */ + unsigned SrcAbsolute : 1; + + /* + * Source register force sign. + * + * Constraints: + * 1. Fields tgsi_src_register_ext_mod::Absolute and + * tgsi_src_register_ext_mod::Negate must not be both set to TRUE + * at the same time. + */ + unsigned SrcForceSign : 1; + + /* + * Source register divide. + * + * Constraints: + * 1. Field tgsi_src_register_ext_swz::ExtDivide + * must be set to TGSI_EXTSWIZZLE_ONE. + */ + unsigned SrcDivide : 1; + + /* + * Source register extended swizzle. + * + * Constraints: + * 1. Field tgsi_src_register_ext_swz::ExtSwizzleX/Y/Z/W + * must be set to TGSI_EXTSWIZZLE_X/Y/Z/W. + * 2. Fields tgsi_src_register_ext_swz::NegateX/Y/Z/W + * must all be set to the same value. + */ + unsigned SrcExtSwizzle : 1; + + unsigned Padding : 22; +}; + +void +tgsi_deco_caps_init( + struct tgsi_deco_caps *caps ); #if defined __cplusplus } // extern "C" -- cgit v1.2.3 From e94ea99254a5f160d115a53abace7a6f159a0749 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 22 Aug 2007 17:37:32 +0100 Subject: Add deco_caps.c. --- src/mesa/sources | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 0fe97e0287..e78db591ab 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -173,6 +173,9 @@ TGSICORE_SOURCES = \ pipe/tgsi/core/tgsi_parse.c \ pipe/tgsi/core/tgsi_util.c +TGSIDECO_SOURCES = \ + pipe/tgsi/deco/deco_caps.c + TGSIMESA_SOURCES = \ pipe/tgsi/mesa/mesa_to_tgsi.c @@ -357,6 +360,7 @@ SOLO_SOURCES = \ $(VF_SOURCES) \ $(DRAW_SOURCES) \ $(TGSICORE_SOURCES) \ + $(TGSIDECO_SOURCES) \ $(TGSIMESA_SOURCES) \ $(STATETRACKER_SOURCES) \ $(TNL_SOURCES) \ -- cgit v1.2.3 From e9190ec534abb817d4f5d1660c84a0ace3376c84 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 18:17:49 -0600 Subject: remove old draw_vertices() --- src/mesa/pipe/draw/draw_context.h | 6 ------ 1 file changed, 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 893db1e0c6..05a3afd5c6 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -126,10 +126,4 @@ draw_arrays(struct draw_context *draw, unsigned prim, unsigned start, unsigned count); -void draw_vertices(struct draw_context *draw, - unsigned mode, - unsigned numVertex, const float *verts, - unsigned numAttribs, const unsigned attribs[]); - - #endif /* DRAW_CONTEXT_H */ -- cgit v1.2.3 From fa8cbc45e883762f3c3f1f11497a035c217f8d65 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 18:35:50 -0600 Subject: Need to store vendor and renderer strings in the context. As it was, we always returned the same pointer. So glxinfo, which calls glGetString() before printing anything, was printing the same string for both vendor and renderer. --- src/mesa/state_tracker/st_cb_strings.c | 14 ++++++-------- src/mesa/state_tracker/st_context.h | 3 +++ 2 files changed, 9 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_strings.c b/src/mesa/state_tracker/st_cb_strings.c index ac1a8b1422..776f518e2b 100644 --- a/src/mesa/state_tracker/st_cb_strings.c +++ b/src/mesa/state_tracker/st_cb_strings.c @@ -46,8 +46,6 @@ st_get_string(GLcontext * ctx, GLenum name) { struct st_context *st = st_context(ctx); struct pipe_context *pipe = st->pipe; - static char buffer[128]; - switch (name) { case GL_VENDOR: { @@ -59,20 +57,20 @@ st_get_string(GLcontext * ctx, GLenum name) * additional string allows "and XyzCorp" to reflect this. */ if (vendor && strcmp(vendor, tungsten) != 0) - snprintf(buffer, sizeof(buffer), "%s and %s", tungsten, vendor); + snprintf(st->vendor, sizeof(st->vendor), + "%s and %s", tungsten, vendor); else - snprintf(buffer, sizeof(buffer), "%s", tungsten); + snprintf(st->vendor, sizeof(st->vendor), "%s", tungsten); - return (GLubyte *) buffer; - break; + return (GLubyte *) st->vendor; } case GL_RENDERER: - snprintf(buffer, sizeof(buffer), "TG3D, %s on %s", + snprintf(st->renderer, sizeof(st->renderer), "TG3D, %s on %s", pipe->get_name( pipe ), pipe->winsys->get_name( pipe->winsys )); - return (GLubyte *) buffer; + return (GLubyte *) st->renderer; default: return NULL; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 38d6a3ed86..13ea28237c 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -97,6 +97,9 @@ struct st_context GLuint frontbuffer_dirty:1; } flags; + char vendor[100]; + char renderer[100]; + /* State to be validated: */ struct st_tracked_state **atoms; -- cgit v1.2.3 From 69ab4b23f5a67451e262b75f4a1a40949fc5868c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 18:47:19 -0600 Subject: more debug output --- src/mesa/pipe/draw/draw_prim.c | 28 ++++++++++++++++++---------- 1 file changed, 18 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 94cedb02a9..8f409424ca 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -137,8 +137,7 @@ run_vertex_program(struct draw_context *draw, unsigned int j; #if 0 - static FILE *file = NULL; - unsigned i; + FILE *file = stdout; #endif ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); @@ -207,9 +206,9 @@ run_vertex_program(struct draw_context *draw, machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ #if 0 - if (attr == 0) { - fprintf(file, "Input vertex %d: %f %f %f\n", - j, p[0], p[1], p[2]); + if (1/*attr == 0*/) { + fprintf(file, "Input vertex %d: attr %d: %f %f %f %f\n", + j, attr, p[0], p[1], p[2], p[3]); fflush( file ); } #endif @@ -269,11 +268,12 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][2] = z * scale[2] + trans[2]; vOut[j]->data[0][3] = w; #if 0 - fprintf(file, "wincoord: %f %f %f %f\n", - vOut[j]->data[0][0], - vOut[j]->data[0][1], - vOut[j]->data[0][2], - vOut[j]->data[0][3]); + fprintf(file, "Vert %d: wincoord: %f %f %f %f\n", + j, + vOut[j]->data[0][0], + vOut[j]->data[0][1], + vOut[j]->data[0][2], + vOut[j]->data[0][3]); fflush( file ); #endif @@ -287,6 +287,14 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[slot][1] = machine.Outputs[attr].xyzw[1].f[j]; vOut[j]->data[slot][2] = machine.Outputs[attr].xyzw[2].f[j]; vOut[j]->data[slot][3] = machine.Outputs[attr].xyzw[3].f[j]; +#if 0 + fprintf(file, "output attrib %d slot %d: %f %f %f %f\n", + attr, slot, + vOut[j]->data[slot][0], + vOut[j]->data[slot][1], + vOut[j]->data[slot][2], + vOut[j]->data[slot][3]); +#endif slot++; } } -- cgit v1.2.3 From ace2b98dd3511ea6959980a2ae2ee90fa7c28748 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 18:48:28 -0600 Subject: fix assertions in swizzle() --- src/mesa/pipe/i915simple/i915_fpc.h | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index afef706418..af347c7c8c 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -185,12 +185,12 @@ struct i915_fp_compile { /* One neat thing about the UREG representation: */ static INLINE int -swizzle(int reg, int x, int y, int z, int w) +swizzle(int reg, uint x, uint y, uint z, uint w) { - assert(x < 4); - assert(y < 4); - assert(z < 4); - assert(w < 4); + assert(x <= SRC_ONE); + assert(y <= SRC_ONE); + assert(z <= SRC_ONE); + assert(w <= SRC_ONE); return ((reg & ~UREG_XYZW_CHANNEL_MASK) | CHANNEL_SRC(GET_CHANNEL_SRC(reg, x), 0) | CHANNEL_SRC(GET_CHANNEL_SRC(reg, y), 1) | -- cgit v1.2.3 From c990d0fd57a05301429b3af75b7fed0337621941 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 22 Aug 2007 18:51:39 -0600 Subject: Checkpoint: texture image and sampler state handling, plus better vertex format code. Texture image/sampler state code should be working, but is disabled for now. Need to fix outstanding issues with vertex formats and texcoords first... --- src/mesa/pipe/i915simple/i915_context.h | 35 ++++++- src/mesa/pipe/i915simple/i915_fpc_emit.c | 3 +- src/mesa/pipe/i915simple/i915_fpc_translate.c | 9 +- src/mesa/pipe/i915simple/i915_prim_emit.c | 51 ++++++++- src/mesa/pipe/i915simple/i915_state.h | 1 + src/mesa/pipe/i915simple/i915_state_derived.c | 82 ++++++++++----- src/mesa/pipe/i915simple/i915_state_emit.c | 36 ++++++- src/mesa/pipe/i915simple/i915_state_immediate.c | 8 ++ src/mesa/pipe/i915simple/i915_state_sampler.c | 134 ++++++++++++++++++++++++ 9 files changed, 321 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index a037b20289..849c3b0efc 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -76,6 +76,32 @@ #define I915_MAX_CONSTANT 32 + +/** + * New vertex format stuff... + */ +#define MAX_VERT_ATTRIBS 12 /* OK? */ + +#define FORMAT_OMIT 0 +#define FORMAT_1F 1 +#define FORMAT_2F 2 +#define FORMAT_3F 3 +#define FORMAT_4F 4 +#define FORMAT_4UB 5 + +struct vertex_info +{ + uint num_attribs; + uint hwfmt[2]; /** hardware format info for this format */ + uint attr_mask; /** mask of VF_ATTR_ bits */ + uint slot_to_attrib[MAX_VERT_ATTRIBS]; + uint interp_mode[MAX_VERT_ATTRIBS]; + uint format[MAX_VERT_ATTRIBS]; /**< FORMAT_x */ +}; + + + + struct i915_cache_context; /* Use to calculate differences between state emitted to hardware and @@ -95,10 +121,17 @@ struct i915_state uint *program; uint program_len; + /* texture sampler state */ unsigned sampler[I915_TEX_UNITS][3]; unsigned sampler_enable_flags; unsigned sampler_enable_nr; - + + /* texture image buffers */ + unsigned texbuffer[I915_TEX_UNITS][2]; + + /* vertex format registers */ + struct vertex_info vertex_info; + unsigned id; /* track lost context events */ }; diff --git a/src/mesa/pipe/i915simple/i915_fpc_emit.c b/src/mesa/pipe/i915simple/i915_fpc_emit.c index dfbc5f180c..235938ac04 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_emit.c +++ b/src/mesa/pipe/i915simple/i915_fpc_emit.c @@ -192,7 +192,8 @@ uint i915_emit_texld( struct i915_fp_compile *p, uint coord, uint op ) { - if (coord != UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord))) { + uint k = UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord)); + if (coord != k) { /* No real way to work around this in the general case - need to * allocate and declare a new temporary register (a utemp won't * do). Will fallback for now. diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index dcf0d18f4e..cd2faa8b97 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -153,8 +153,15 @@ src_vector(struct i915_fp_compile *p, src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); break; case FRAG_ATTRIB_COL1: +#if 1 src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); src = swizzle(src, X, Y, Z, ONE); +#else + /* total hack to force texture mapping */ + src = i915_emit_decl(p, REG_TYPE_T, + T_TEX0/* + (index - FRAG_ATTRIB_TEX0)*/, + D0_CHANNEL_ALL); +#endif break; case FRAG_ATTRIB_FOGC: src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); @@ -979,7 +986,7 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) static void i915_find_wpos_space(struct i915_fp_compile *p) { - const uint inputs = p->shader->inputs_read; + const uint inputs = p->shader->inputs_read | FRAG_BIT_WPOS; /*XXX hack*/ uint i; p->wpos_tex = -1; diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index 2c5f11c225..ff57d26d38 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -71,13 +71,56 @@ emit_hw_vertex( struct i915_context *i915, /* colors are ARGB (MSB to LSB) */ OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[1][2] ), - float_to_ubyte( vertex->data[1][1] ), - float_to_ubyte( vertex->data[1][0] ), - float_to_ubyte( vertex->data[1][3] )) ); + float_to_ubyte( vertex->data[1][1] ), + float_to_ubyte( vertex->data[1][0] ), + float_to_ubyte( vertex->data[1][3] )) ); } +static INLINE void +emit_hw_vertex2( struct i915_context *i915, + const struct vertex_header *vertex) +{ + const struct vertex_info *vinfo = &i915->current.vertex_info; + uint i; + + for (i = 0; i < vinfo->num_attribs; i++) { + switch (vinfo->format[i]) { + case FORMAT_OMIT: + /* no-op */ + break; + case FORMAT_1F: + OUT_BATCH( fui(vertex->data[i][0]) ); + break; + case FORMAT_2F: + OUT_BATCH( fui(vertex->data[i][0]) ); + OUT_BATCH( fui(vertex->data[i][1]) ); + break; + case FORMAT_3F: + OUT_BATCH( fui(vertex->data[i][0]) ); + OUT_BATCH( fui(vertex->data[i][1]) ); + OUT_BATCH( fui(vertex->data[i][2]) ); + break; + case FORMAT_4F: + OUT_BATCH( fui(vertex->data[i][0]) ); + OUT_BATCH( fui(vertex->data[i][1]) ); + OUT_BATCH( fui(vertex->data[i][2]) ); + OUT_BATCH( fui(vertex->data[i][3]) ); + break; + case FORMAT_4UB: + OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[i][2] ), + float_to_ubyte( vertex->data[i][1] ), + float_to_ubyte( vertex->data[i][0] ), + float_to_ubyte( vertex->data[i][3] )) ); + break; + default: + assert(0); + } + } +} + + static INLINE void emit_prim( struct draw_stage *stage, @@ -120,7 +163,7 @@ emit_prim( struct draw_stage *stage, ((4 + vertex_size * nr)/4 - 2)); for (i = 0; i < nr; i++) { - emit_hw_vertex(i915, prim->v[i]); + emit_hw_vertex2(i915, prim->v[i]); ptr += vertex_size / sizeof(int); } } diff --git a/src/mesa/pipe/i915simple/i915_state.h b/src/mesa/pipe/i915simple/i915_state.h index 43058451f2..86c6b0027d 100644 --- a/src/mesa/pipe/i915simple/i915_state.h +++ b/src/mesa/pipe/i915simple/i915_state.h @@ -43,6 +43,7 @@ void i915_update_immediate( struct i915_context *i915 ); void i915_update_dynamic( struct i915_context *i915 ); void i915_update_derived( struct i915_context *i915 ); void i915_update_samplers( struct i915_context *i915 ); +void i915_update_textures(struct i915_context *i915); void i915_emit_hardware_state( struct i915_context *i915 ); diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 5a493cc9cf..bd952cc567 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -29,20 +29,13 @@ #include "pipe/draw/draw_context.h" #include "i915_context.h" #include "i915_state.h" +#include "i915_reg.h" /* XXX should include i915_fpc.h but that causes some trouble atm */ extern void i915_translate_fragment_program( struct i915_context *i915 ); -#define EMIT_ATTR( VF_ATTR, FRAG_ATTR, INTERP ) \ -do { \ - slot_to_vf_attr[nr_attrs] = VF_ATTR; \ - nr_attrs++; \ - attr_mask |= (1 << (VF_ATTR)); \ -} while (0) - - static const unsigned frag_to_vf[FRAG_ATTRIB_MAX] = { VF_ATTRIB_POS, @@ -68,6 +61,19 @@ static const unsigned frag_to_vf[FRAG_ATTRIB_MAX] = }; +static INLINE void +emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) +{ + const uint n = vinfo->num_attribs; + vinfo->attr_mask |= (1 << vfAttr); + vinfo->slot_to_attrib[n] = vfAttr; + /*vinfo->interp_mode[n] = interpMode;*/ + vinfo->format[n] = format; + vinfo->num_attribs++; +} + + + /** * Determine which post-transform / pre-rasterization vertex attributes * we need. @@ -75,19 +81,17 @@ static const unsigned frag_to_vf[FRAG_ATTRIB_MAX] = */ static void calculate_vertex_layout( struct i915_context *i915 ) { -// const unsigned inputsRead = i915->fs.inputs_read; - const unsigned inputsRead = (FRAG_BIT_WPOS | FRAG_BIT_COL0); - unsigned slot_to_vf_attr[VF_ATTRIB_MAX]; - unsigned attr_mask = 0x0; - unsigned nr_attrs = 0; + const unsigned inputsRead = i915->fs.inputs_read; +// const unsigned inputsRead = (FRAG_BIT_WPOS | FRAG_BIT_COL0); unsigned i; + struct vertex_info *vinfo = &i915->current.vertex_info; - memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); - + memset(vinfo, 0, sizeof(*vinfo)); /* TODO - Figure out if we need to do perspective divide, etc. */ - EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); + emit_vertex_attr(vinfo, VF_ATTRIB_POS, FORMAT_3F); + vinfo->hwfmt[0] |= S4_VFMT_XYZ; /* Pull in the rest of the attributes. They are all in float4 * format. Future optimizations could be to keep some attributes @@ -95,20 +99,29 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { if (inputsRead & (1 << i)) { - assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); + assert(i < Elements(frag_to_vf)); if (i915->setup.flatshade - && (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1)) - EMIT_ATTR(frag_to_vf[i], i, INTERP_CONSTANT); - else - EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); + && (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1)) { + emit_vertex_attr(vinfo, frag_to_vf[i], FORMAT_4UB); + } + else { + emit_vertex_attr(vinfo, frag_to_vf[i], FORMAT_4UB); + } + vinfo->hwfmt[0] |= S4_VFMT_COLOR; } } for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { + uint hwtc; if (inputsRead & (1 << i)) { + hwtc = TEXCOORDFMT_4D; assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); - EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); + emit_vertex_attr(vinfo, frag_to_vf[i], FORMAT_4F); + } + else { + hwtc = TEXCOORDFMT_NOT_PRESENT; } + vinfo->hwfmt[1] |= hwtc << ((i - FRAG_ATTRIB_TEX0) * 4); } /* Additional attributes required for setup: Just twosided @@ -117,24 +130,31 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ if (i915->setup.light_twoside) { if (inputsRead & FRAG_BIT_COL0) { - EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ + /* XXX: mark as discarded after setup */ + emit_vertex_attr(vinfo, VF_ATTRIB_BFC0, FORMAT_OMIT); } if (inputsRead & FRAG_BIT_COL1) { - EMIT_ATTR(VF_ATTRIB_BFC1, FRAG_ATTRIB_MAX, 0); /* XXX: discard after setup */ + /* XXX: discard after setup */ + emit_vertex_attr(vinfo, VF_ATTRIB_BFC1, FORMAT_OMIT); } } - /* If the attributes have changed, tell the draw module (which in turn - * tells the vf module) about the new vertex layout. + /* If the attributes have changed, tell the draw module about the new + * vertex layout. We'll also update the hardware vertex format info. */ draw_set_vertex_attributes( i915->draw, - slot_to_vf_attr, - nr_attrs ); + vinfo->slot_to_attrib, + vinfo->num_attribs); +#if 0 + printf("VERTEX_FORMAT LIS2: 0x%x LIS4: 0x%x\n", + vinfo->hwfmt[1], vinfo->hwfmt[0]); +#endif } + /* Hopefully this will remain quite simple, otherwise need to pull in * something like the state tracker mechanism. */ @@ -143,6 +163,12 @@ void i915_update_derived( struct i915_context *i915 ) if (i915->dirty & (I915_NEW_SETUP | I915_NEW_FS)) calculate_vertex_layout( i915 ); + if (i915->dirty & I915_NEW_SAMPLER) + i915_update_samplers(i915); + + if (i915->dirty & I915_NEW_TEXTURE) + i915_update_textures(i915); + if (i915->dirty) i915_update_immediate( i915 ); diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index dff9d40a83..ceb96289b5 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -180,7 +180,6 @@ i915_emit_hardware_state(struct i915_context *i915 ) zformat = translate_depth_format( i915->framebuffer.zbuf->format ); OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); - OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */ DSTORG_VERT_BIAS(0x8) | /* .5 */ LOD_PRECLAMP_OGL | @@ -190,6 +189,38 @@ i915_emit_hardware_state(struct i915_context *i915 ) } } + +#if 0 + /* texture images */ + if (i915->hardware_dirty & I915_HW_MAP) + { + const uint nr = i915->current.sampler_enable_nr; + const uint enabled = i915->current.sampler_enable_flags; + uint unit; + uint count = 0; + OUT_BATCH(_3DSTATE_MAP_STATE | (3 * nr)); + OUT_BATCH(enabled); + for (unit = 0; unit < I915_TEX_UNITS; unit++) { + if (enabled & (1 << unit)) { + struct pipe_buffer_handle *buf = + i915->texture[unit]->region->buffer; + uint offset = 0; + assert(buf); + + count++; + + OUT_RELOC(buf, + I915_BUFFER_ACCESS_READ, + offset); + OUT_BATCH(i915->current.texbuffer[unit][0]); /* MS3 */ + OUT_BATCH(i915->current.texbuffer[unit][1]); /* MS4 */ + } + } + assert(count == nr); + } +#endif + +#if 0 /* samplers */ if (i915->hardware_dirty & I915_HW_SAMPLER) { @@ -210,8 +241,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) } } } - - +#endif /* constants */ if (i915->hardware_dirty & I915_HW_PROGRAM) diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 1cd1866499..c6f1f59f88 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -53,8 +53,16 @@ static void upload_S2S4(struct i915_context *i915) unsigned LIS2, LIS4; /* I915_NEW_VERTEX_FORMAT */ +#if 01 LIS2 = 0xffffffff; LIS4 = (S4_VFMT_XYZ | S4_VFMT_COLOR); +#else + assert(LIS2 == i915->current.vertex_info.hwfmt[1]); + assert(LIS4 == i915->current.vertex_info.hwfmt[0]); + LIS2 = i915->current.vertex_info.hwfmt[1]; + LIS4 = i915->current.vertex_info.hwfmt[0]; + printf("UPLOAD FORMAT LIS2: 0x%x LIS4: 0x%x\n", LIS2, LIS4); +#endif /* I915_NEW_SETUP */ diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 591e3684da..e23e4bdcf7 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -39,6 +39,7 @@ #include "i915_state_inlines.h" #include "i915_context.h" #include "i915_reg.h" +#include "i915_state.h" //#include "i915_cache.h" @@ -283,3 +284,136 @@ void i915_update_samplers( struct i915_context *i915 ) i915->hardware_dirty |= I915_HW_SAMPLER; } + + + + +static uint +translate_texture_format(uint pipeFormat) +{ + switch (pipeFormat) { + case PIPE_FORMAT_U_L8: + return MAPSURF_8BIT | MT_8BIT_L8; + case PIPE_FORMAT_U_I8: + return MAPSURF_8BIT | MT_8BIT_I8; + case PIPE_FORMAT_U_A8: + return MAPSURF_8BIT | MT_8BIT_A8; + case PIPE_FORMAT_U_L8_A8: + return MAPSURF_16BIT | MT_16BIT_AY88; + case PIPE_FORMAT_U_R5_G6_B5: + return MAPSURF_16BIT | MT_16BIT_RGB565; + case PIPE_FORMAT_U_A1_R5_G5_B5: + return MAPSURF_16BIT | MT_16BIT_ARGB1555; + case PIPE_FORMAT_U_A4_R4_G4_B4: + return MAPSURF_16BIT | MT_16BIT_ARGB4444; + case PIPE_FORMAT_U_A8_R8_G8_B8: + return MAPSURF_32BIT | MT_32BIT_ARGB8888; + case PIPE_FORMAT_YCBCR_REV: + return (MAPSURF_422 | MT_422_YCRCB_NORMAL); + case PIPE_FORMAT_YCBCR: + return (MAPSURF_422 | MT_422_YCRCB_SWAPY); +#if 0 + case PIPE_FORMAT_RGB_FXT1: + case PIPE_FORMAT_RGBA_FXT1: + return (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1); +#endif + case PIPE_FORMAT_U_Z16: + return (MAPSURF_16BIT | MT_16BIT_L16); +#if 0 + case PIPE_FORMAT_RGBA_DXT1: + case PIPE_FORMAT_RGB_DXT1: + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1); + case PIPE_FORMAT_RGBA_DXT3: + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3); + case PIPE_FORMAT_RGBA_DXT5: + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); +#endif + case PIPE_FORMAT_S8_Z24: + return (MAPSURF_32BIT | MT_32BIT_xL824); + default: + fprintf(stderr, "i915: translate_texture_format() bad image format %x\n", + pipeFormat); + assert(0); + return 0; + } +} + + +#define I915_TEXREG_MS3 1 +#define I915_TEXREG_MS4 2 + + +static void +i915_update_texture(struct i915_context *i915, uint unit, + uint state[6]) +{ + const struct pipe_mipmap_tree *mt = i915->texture[unit]; + uint format, pitch; + const uint width = mt->width0, height = mt->height0, depth = mt->depth0; + const uint num_levels = mt->last_level - mt->first_level; + + assert(mt); + assert(width); + assert(height); + assert(depth); + +#if 0 + if (i915->state.tex_buffer[unit] != NULL) { + driBOUnReference(i915->state.tex_buffer[unit]); + i915->state.tex_buffer[unit] = NULL; + } +#endif + + + { + struct pipe_buffer_handle *p = driBOReference(mt->region->buffer); + } + +#if 0 + i915->state.tex_buffer[unit] = driBOReference(intelObj->mt->region-> + buffer); + i915->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt, + 0, intelObj-> + firstLevel); +#endif + + format = translate_texture_format(mt->format); + pitch = mt->pitch * mt->cpp; + + assert(format); + assert(pitch); + + printf("texture format = 0x%x\n", format); + + /* MS3 state */ + state[0] = + (((height - 1) << MS3_HEIGHT_SHIFT) + | ((width - 1) << MS3_WIDTH_SHIFT) + | format + | MS3_USE_FENCE_REGS); + + /* MS4 state */ + state[1] = + ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) + | MS4_CUBE_FACE_ENA_MASK + | ((num_levels * 4) << MS4_MAX_LOD_SHIFT) + | ((depth - 1) << MS4_VOLUME_DEPTH_SHIFT)); +} + + + +void +i915_update_textures(struct i915_context *i915) +{ + uint unit; + + for (unit = 0; unit < I915_TEX_UNITS; unit++) { + /* determine unit enable/disable by looking for a bound mipmap tree */ + /* could also examine the fragment program? */ + if (i915->texture[unit]) { + i915_update_texture(i915, unit, i915->current.texbuffer[unit]); + } + } + + i915->hardware_dirty |= I915_HW_MAP; +} -- cgit v1.2.3 From ef7c25090074cb02f18acc905bca5f25a56bd021 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 13:24:06 -0600 Subject: For the time being, interpolate Z in shade_quad() rather in the shader. This was causing trouble for the i915 driver. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 43 ++++++++++++++++++++++++++++++++++ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 5 ++++ 2 files changed, 48 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index ceba94aa94..8a419c9ac7 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -55,6 +55,34 @@ quad_shade_stage(struct quad_stage *qs) return (struct quad_shade_stage *) qs; } + + + +/** + * Compute quad's attribute values by linear interpolation. + * + * Push into the fp: + * + * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx + * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy + */ +static INLINE void +linterp_z(const struct tgsi_interp_coef *coef, + struct tgsi_exec_vector *pos) +{ + uint ch = 2; + uint j; + for (j = 0; j < QUAD_SIZE; j++) { + const float x = pos->xyzw[0].f[j]; + const float y = pos->xyzw[1].f[j]; + pos->xyzw[ch].f[j] = (coef->a0[ch] + + coef->dadx[ch] * x + + coef->dady[ch] * y); + } +} + + + /* This should be done by the fragment shader execution unit (code * generated from the decl instructions). Do it here for now. */ @@ -101,6 +129,9 @@ shade_quad( machine.Inputs[0].xyzw[1].f[2] = fy + 1.0f; machine.Inputs[0].xyzw[1].f[3] = fy + 1.0f; + /* interp Z */ + linterp_z(&quad->coef[0], &machine.Inputs[0]); + /* run shader */ tgsi_exec_machine_run( &machine ); @@ -110,6 +141,7 @@ shade_quad( &machine.Outputs[1].xyzw[0].f[0], sizeof( quad->outputs.color ) ); +#if 0 if( softpipe->need_z ) { /* XXX temporary */ memcpy( @@ -117,6 +149,17 @@ shade_quad( &machine.Outputs[0].xyzw[2], sizeof( quad->outputs.depth ) ); } +#else + { + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine.Inputs[0].xyzw[2].f[i]; +#if 0 + printf("output z %f\n", quad->outputs.depth[i]); +#endif + } + } +#endif /* shader may cull fragments */ if( quad->mask ) { diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 1de80eb734..eeaed844d5 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -52,6 +52,9 @@ map_register_file_index( GLuint mapped_index; GLuint i; + assert(processor == TGSI_PROCESSOR_FRAGMENT + || processor == TGSI_PROCESSOR_VERTEX); + switch( file ) { case TGSI_FILE_INPUT: /* @@ -616,6 +619,7 @@ tgsi_mesa_compile_fp_program( /* * Copy fragment z if the shader does not write it. */ +#if 0 if( !(program->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR)) ) { fullinst = tgsi_default_full_instruction(); @@ -639,6 +643,7 @@ tgsi_mesa_compile_fp_program( maxTokens - ti ); preamble_size++; } +#endif for( i = 0; i < program->Base.NumInstructions; i++ ) { if( compile_instruction( -- cgit v1.2.3 From 690a9de40b20092ae9027dc52d7b26a48995bbff Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 13:25:46 -0600 Subject: Define a re-usable vertex_info structure to describe vertex layouts. This also includes legacy #defines for VF_ATTRIB_*, FRAG_ATTRIB_*, etc tokens. Those will go away eventually, but at least we no longer need vf/vf.h --- src/mesa/pipe/draw/draw_vertex.h | 141 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 141 insertions(+) create mode 100644 src/mesa/pipe/draw/draw_vertex.h (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h new file mode 100644 index 0000000000..3262431487 --- /dev/null +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -0,0 +1,141 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + */ + + +#ifndef DRAW_VERTEX_H +#define DRAW_VERTEX_H + + +/*** + *** XXX There's a lot of legacy tokens here that'll eventually go away. + *** (at least we don't include vf/vf.h anymore) + ***/ + + +enum { + VF_ATTRIB_POS = 0, + VF_ATTRIB_WEIGHT = 1, + VF_ATTRIB_NORMAL = 2, + VF_ATTRIB_COLOR0 = 3, + VF_ATTRIB_COLOR1 = 4, + VF_ATTRIB_FOG = 5, + VF_ATTRIB_COLOR_INDEX = 6, + VF_ATTRIB_EDGEFLAG = 7, + VF_ATTRIB_TEX0 = 8, + VF_ATTRIB_TEX1 = 9, + VF_ATTRIB_TEX2 = 10, + VF_ATTRIB_TEX3 = 11, + VF_ATTRIB_TEX4 = 12, + VF_ATTRIB_TEX5 = 13, + VF_ATTRIB_TEX6 = 14, + VF_ATTRIB_TEX7 = 15, + VF_ATTRIB_VAR0 = 16, + VF_ATTRIB_VAR1 = 17, + VF_ATTRIB_VAR2 = 18, + VF_ATTRIB_VAR3 = 19, + VF_ATTRIB_VAR4 = 20, + VF_ATTRIB_VAR5 = 21, + VF_ATTRIB_VAR6 = 22, + VF_ATTRIB_VAR7 = 23, + VF_ATTRIB_POINTSIZE = 24, + VF_ATTRIB_BFC0 = 25, + VF_ATTRIB_BFC1 = 26, + VF_ATTRIB_CLIP_POS = 27, + VF_ATTRIB_VERTEX_HEADER = 28, + VF_ATTRIB_MAX = 29 +}; + +#define MAX_VARYING 8 +enum +{ + FRAG_ATTRIB_WPOS = 0, + FRAG_ATTRIB_COL0 = 1, + FRAG_ATTRIB_COL1 = 2, + FRAG_ATTRIB_FOGC = 3, + FRAG_ATTRIB_TEX0 = 4, + FRAG_ATTRIB_TEX1 = 5, + FRAG_ATTRIB_TEX2 = 6, + FRAG_ATTRIB_TEX3 = 7, + FRAG_ATTRIB_TEX4 = 8, + FRAG_ATTRIB_TEX5 = 9, + FRAG_ATTRIB_TEX6 = 10, + FRAG_ATTRIB_TEX7 = 11, + FRAG_ATTRIB_VAR0 = 12, /**< shader varying */ + FRAG_ATTRIB_MAX = (FRAG_ATTRIB_VAR0 + MAX_VARYING) +}; + +#define FRAG_BIT_WPOS (1 << FRAG_ATTRIB_WPOS) +#define FRAG_BIT_COL0 (1 << FRAG_ATTRIB_COL0) +#define FRAG_BIT_COL1 (1 << FRAG_ATTRIB_COL1) +#define FRAG_BIT_FOGC (1 << FRAG_ATTRIB_FOGC) +#define FRAG_BIT_TEX0 (1 << FRAG_ATTRIB_TEX0) + + + +#define MAX_DRAW_BUFFERS 4 + +enum +{ + FRAG_RESULT_COLR = 0, + FRAG_RESULT_COLH = 1, + FRAG_RESULT_DEPR = 2, + FRAG_RESULT_DATA0 = 3, + FRAG_RESULT_MAX = (FRAG_RESULT_DATA0 + MAX_DRAW_BUFFERS) +}; + + + +#define MAX_VERT_ATTRIBS 12 /* OK? */ + +#define FORMAT_OMIT 0 +#define FORMAT_1F 1 +#define FORMAT_2F 2 +#define FORMAT_3F 3 +#define FORMAT_4F 4 +#define FORMAT_4F_VIEWPORT 4 +#define FORMAT_4UB 5 + + +struct vertex_info +{ + uint num_attribs; + uint hwfmt[2]; /**< hardware format info for this format */ + uint attr_mask; /**< mask of VF_ATTR_ bits */ + uint slot_to_attrib[MAX_VERT_ATTRIBS]; + uint attrib_to_slot[VF_ATTRIB_MAX]; + uint interp_mode[MAX_VERT_ATTRIBS]; + uint format[MAX_VERT_ATTRIBS]; /**< FORMAT_x */ + uint size; /**< total vertex size in dwords */ +}; + + + +#endif /* DRAW_VERTEX_H */ -- cgit v1.2.3 From d3eb25c575464bed7dbfc8be4717d85cb2928ec1 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 13:27:18 -0600 Subject: Checkpoint commit: i915 texture works, use new vertex_info struct Basic i915 2D texturing seems to work now. The vertex format is determined from the current fragment shader. --- src/mesa/pipe/draw/draw_arrays.c | 78 ++++++++++++++++++------- src/mesa/pipe/draw/draw_clip.c | 8 +-- src/mesa/pipe/draw/draw_flatshade.c | 2 +- src/mesa/pipe/draw/draw_prim.c | 2 +- src/mesa/pipe/draw/draw_private.h | 34 ++--------- src/mesa/pipe/draw/draw_twoside.c | 2 +- src/mesa/pipe/i915simple/i915_context.h | 28 +-------- src/mesa/pipe/i915simple/i915_debug.c | 6 ++ src/mesa/pipe/i915simple/i915_fpc.h | 51 +--------------- src/mesa/pipe/i915simple/i915_fpc_emit.c | 1 + src/mesa/pipe/i915simple/i915_fpc_translate.c | 62 +++++++++++++++++--- src/mesa/pipe/i915simple/i915_prim_emit.c | 25 +++++++- src/mesa/pipe/i915simple/i915_state.c | 1 - src/mesa/pipe/i915simple/i915_state_derived.c | 43 +++++++++++++- src/mesa/pipe/i915simple/i915_state_emit.c | 46 ++++++++------- src/mesa/pipe/i915simple/i915_state_immediate.c | 11 +++- src/mesa/pipe/softpipe/sp_state_derived.c | 62 +------------------- 17 files changed, 230 insertions(+), 232 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index 738fa144b2..430dc15ce2 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -70,17 +70,53 @@ draw_arrays(struct draw_context *draw, unsigned prim, } -/* XXX move this into draw_context.c? */ +static INLINE void +emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) +{ + const uint n = vinfo->num_attribs; + vinfo->attr_mask |= (1 << vfAttr); + vinfo->slot_to_attrib[n] = vfAttr; + if (n >= 2) { + /* the first two slots are the vertex header & clippos */ + vinfo->attrib_to_slot[vfAttr] = n - 2; + } + /*printf("Vertex slot %d = vfattrib %d\n", n, vfAttr);*/ + /*vinfo->interp_mode[n] = interpMode;*/ + vinfo->format[n] = format; + vinfo->num_attribs++; + +} -#define EMIT_ATTR( VF_ATTR, STYLE, SIZE ) \ -do { \ - if (draw->nr_attrs >= 2) \ - draw->vf_attr_to_slot[VF_ATTR] = draw->nr_attrs - 2; \ - draw->attrs[draw->nr_attrs].attrib = VF_ATTR; \ - /*draw->attrs[draw->nr_attrs].format = STYLE*/; \ - draw->nr_attrs++; \ - draw->vertex_size += SIZE; \ -} while (0) + +static void +compute_vertex_size(struct vertex_info *vinfo) +{ + uint i; + + vinfo->size = 0; + for (i = 0; i < vinfo->num_attribs; i++) { + switch (vinfo->format[i]) { + case FORMAT_OMIT: + break; + case FORMAT_4UB: + /* fall-through */ + case FORMAT_1F: + vinfo->size += 1; + break; + case FORMAT_2F: + vinfo->size += 2; + break; + case FORMAT_3F: + vinfo->size += 3; + break; + case FORMAT_4F: + vinfo->size += 4; + break; + default: + assert(0); + } + } +} void @@ -88,26 +124,26 @@ draw_set_vertex_attributes( struct draw_context *draw, const unsigned *slot_to_vf_attr, unsigned nr_attrs ) { + struct vertex_info *vinfo = &draw->vertex_info; unsigned i; - memset(draw->vf_attr_to_slot, 0, sizeof(draw->vf_attr_to_slot)); - draw->nr_attrs = 0; - draw->vertex_size = 0; + assert(slot_to_vf_attr[0] == VF_ATTRIB_POS); + + memset(vinfo, 0, sizeof(*vinfo)); /* * First three attribs are always the same: header, clip pos, winpos */ - EMIT_ATTR(VF_ATTRIB_VERTEX_HEADER, EMIT_1F, 1); - EMIT_ATTR(VF_ATTRIB_CLIP_POS, EMIT_4F, 4); - - assert(slot_to_vf_attr[0] == VF_ATTRIB_POS); - EMIT_ATTR(slot_to_vf_attr[0], EMIT_4F_VIEWPORT, 4); + emit_vertex_attr(vinfo, VF_ATTRIB_VERTEX_HEADER, FORMAT_1F); + emit_vertex_attr(vinfo, VF_ATTRIB_CLIP_POS, FORMAT_4F); + emit_vertex_attr(vinfo, VF_ATTRIB_POS, FORMAT_4F_VIEWPORT); /* * Remaining attribs (color, texcoords, etc) */ - for (i = 1; i < nr_attrs; i++) - EMIT_ATTR(slot_to_vf_attr[i], EMIT_4F, 4); + for (i = 1; i < nr_attrs; i++) { + emit_vertex_attr(vinfo, slot_to_vf_attr[i], FORMAT_4F); + } - draw->vertex_size *= 4; /* floats to bytes */ + compute_vertex_size(vinfo); } diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 7e5ceacdfe..1c2491d2c6 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -93,7 +93,7 @@ static void interp( const struct clipper *clip, const struct vertex_header *out, const struct vertex_header *in ) { - const unsigned nr_attrs = clip->stage.draw->nr_attrs; + const unsigned nr_attrs = clip->stage.draw->vertex_info.num_attribs; unsigned j; /* Vertex header. @@ -379,9 +379,9 @@ static void clip_begin( struct draw_stage *stage ) unsigned nr = stage->draw->nr_planes; /* sanity checks. If these fail, review the clip/interp code! */ - assert(stage->draw->nr_attrs >= 3); - assert(stage->draw->attrs[0].attrib == VF_ATTRIB_VERTEX_HEADER); - assert(stage->draw->attrs[1].attrib == VF_ATTRIB_CLIP_POS); + assert(stage->draw->vertex_info.num_attribs >= 3); + assert(stage->draw->vertex_info.slot_to_attrib[0] == VF_ATTRIB_VERTEX_HEADER); + assert(stage->draw->vertex_info.slot_to_attrib[1] == VF_ATTRIB_CLIP_POS); /* Hacky bitmask to use when we hit CLIP_USER_BIT: */ diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 128acc5b2b..34588c83b6 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -154,7 +154,7 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) flatshade->stage.end = flatshade_end; flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; - flatshade->lookup = draw->vf_attr_to_slot; + flatshade->lookup = draw->vertex_info.attrib_to_slot; return &flatshade->stage; } diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 8f409424ca..23eb578221 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -282,7 +282,7 @@ run_vertex_program(struct draw_context *draw, slot = 1; for (attr = 1; attr < VERT_RESULT_MAX; attr++) { if (draw->vertex_shader.outputs_written & (1 << attr)) { - assert(slot < draw->nr_attrs); + assert(slot < draw->vertex_info.num_attribs); vOut[j]->data[slot][0] = machine.Outputs[attr].xyzw[0].f[j]; vOut[j]->data[slot][1] = machine.Outputs[attr].xyzw[1].f[j]; vOut[j]->data[slot][2] = machine.Outputs[attr].xyzw[2].f[j]; diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 75f7c5d84e..531bb2e254 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -44,28 +44,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" - -/* XXX these are temporary */ -struct vf_attr_map { - unsigned attrib; - /* - unsigned format; - */ - unsigned offset; -}; -#define VF_ATTRIB_POS 0 -#define VF_ATTRIB_COLOR0 3 -#define VF_ATTRIB_COLOR1 4 -#define VF_ATTRIB_BFC0 25 -#define VF_ATTRIB_BFC1 26 - -#define VF_ATTRIB_CLIP_POS 27 -#define VF_ATTRIB_VERTEX_HEADER 28 -#define VF_ATTRIB_MAX 29 -#define EMIT_1F 0 -#define EMIT_4F 3 -#define EMIT_4F_VIEWPORT 6 -#define FRAG_ATTRIB_MAX 13 +#include "draw_vertex.h" /** @@ -175,14 +154,9 @@ struct draw_context float plane[12][4]; unsigned nr_planes; - /* - * Vertex attribute info - */ - unsigned vf_attr_to_slot[PIPE_ATTRIB_MAX]; - struct vf_attr_map attrs[VF_ATTRIB_MAX]; - unsigned nr_attrs; + /** Describes the layout of post-transformation vertices */ + struct vertex_info vertex_info; - unsigned vertex_size; /**< in bytes */ unsigned nr_vertices; unsigned prim; /**< current prim type: PIPE_PRIM_x */ @@ -250,7 +224,7 @@ dup_vert( struct draw_stage *stage, unsigned idx ) { struct vertex_header *tmp = stage->tmp[idx]; - memcpy(tmp, vert, stage->draw->vertex_size ); + memcpy(tmp, vert, stage->draw->vertex_info.size * sizeof(float) ); return tmp; } diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index d7d993841b..822cadd61f 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -163,7 +163,7 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) twoside->stage.end = twoside_end; twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; - twoside->lookup = draw->vf_attr_to_slot; + twoside->lookup = draw->vertex_info.attrib_to_slot; return &twoside->stage; } diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 849c3b0efc..5de6afa9e9 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -33,6 +33,8 @@ #include "pipe/p_defines.h" #include "pipe/p_state.h" +#include "pipe/draw/draw_vertex.h" + #define I915_TEX_UNITS 8 @@ -77,30 +79,6 @@ #define I915_MAX_CONSTANT 32 -/** - * New vertex format stuff... - */ -#define MAX_VERT_ATTRIBS 12 /* OK? */ - -#define FORMAT_OMIT 0 -#define FORMAT_1F 1 -#define FORMAT_2F 2 -#define FORMAT_3F 3 -#define FORMAT_4F 4 -#define FORMAT_4UB 5 - -struct vertex_info -{ - uint num_attribs; - uint hwfmt[2]; /** hardware format info for this format */ - uint attr_mask; /** mask of VF_ATTR_ bits */ - uint slot_to_attrib[MAX_VERT_ATTRIBS]; - uint interp_mode[MAX_VERT_ATTRIBS]; - uint format[MAX_VERT_ATTRIBS]; /**< FORMAT_x */ -}; - - - struct i915_cache_context; @@ -129,7 +107,7 @@ struct i915_state /* texture image buffers */ unsigned texbuffer[I915_TEX_UNITS][2]; - /* vertex format registers */ + /** Describes the current hardware vertex layout */ struct vertex_info vertex_info; unsigned id; /* track lost context events */ diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index 79060e9780..d142194d84 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -692,10 +692,14 @@ static boolean i915_debug_packet( struct debug_stream *stream ) case 0x31: return debug_chain(stream, "MI_BATCH_BUFFER_START", 2); default: + (void)debug(stream, "UNKNOWN 0x0 case!", 1); + assert(0); break; } break; case 0x1: + (void) debug(stream, "UNKNOWN 0x1 case!", 1); + assert(0); break; case 0x2: switch ((cmd >> 22) & 0xff) { @@ -735,6 +739,8 @@ static boolean i915_debug_packet( struct debug_stream *stream ) case 0x11: return debug(stream, "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE", 1); default: + (void) debug(stream, "UNKNOWN 0x1c case!", 1); + assert(0); break; } break; diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index af347c7c8c..7bbfc69d29 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -38,55 +38,6 @@ #define I915_PROGRAM_SIZE 192 -#define MAX_VARYING 8 - -enum -{ - FRAG_ATTRIB_WPOS = 0, - FRAG_ATTRIB_COL0 = 1, - FRAG_ATTRIB_COL1 = 2, - FRAG_ATTRIB_FOGC = 3, - FRAG_ATTRIB_TEX0 = 4, - FRAG_ATTRIB_TEX1 = 5, - FRAG_ATTRIB_TEX2 = 6, - FRAG_ATTRIB_TEX3 = 7, - FRAG_ATTRIB_TEX4 = 8, - FRAG_ATTRIB_TEX5 = 9, - FRAG_ATTRIB_TEX6 = 10, - FRAG_ATTRIB_TEX7 = 11, - FRAG_ATTRIB_VAR0 = 12, /**< shader varying */ - FRAG_ATTRIB_MAX = (FRAG_ATTRIB_VAR0 + MAX_VARYING) -}; - -/** - * Bitflags for fragment program input attributes. - */ -/*@{*/ -#define FRAG_BIT_WPOS (1 << FRAG_ATTRIB_WPOS) -#define FRAG_BIT_COL0 (1 << FRAG_ATTRIB_COL0) -#define FRAG_BIT_COL1 (1 << FRAG_ATTRIB_COL1) -#define FRAG_BIT_FOGC (1 << FRAG_ATTRIB_FOGC) -#define FRAG_BIT_TEX0 (1 << FRAG_ATTRIB_TEX0) -#define FRAG_BIT_TEX1 (1 << FRAG_ATTRIB_TEX1) -#define FRAG_BIT_TEX2 (1 << FRAG_ATTRIB_TEX2) -#define FRAG_BIT_TEX3 (1 << FRAG_ATTRIB_TEX3) -#define FRAG_BIT_TEX4 (1 << FRAG_ATTRIB_TEX4) -#define FRAG_BIT_TEX5 (1 << FRAG_ATTRIB_TEX5) -#define FRAG_BIT_TEX6 (1 << FRAG_ATTRIB_TEX6) -#define FRAG_BIT_TEX7 (1 << FRAG_ATTRIB_TEX7) -#define FRAG_BIT_VAR0 (1 << FRAG_ATTRIB_VAR0) - -#define MAX_DRAW_BUFFERS 4 - -enum -{ - FRAG_RESULT_COLR = 0, - FRAG_RESULT_COLH = 1, - FRAG_RESULT_DEPR = 2, - FRAG_RESULT_DATA0 = 3, - FRAG_RESULT_MAX = (FRAG_RESULT_DATA0 + MAX_DRAW_BUFFERS) -}; - /** @@ -95,6 +46,8 @@ enum struct i915_fp_compile { struct pipe_shader_state *shader; + struct vertex_info *vertex_info; + uint declarations[I915_PROGRAM_SIZE]; uint program[I915_PROGRAM_SIZE]; diff --git a/src/mesa/pipe/i915simple/i915_fpc_emit.c b/src/mesa/pipe/i915simple/i915_fpc_emit.c index 235938ac04..f062885f8a 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_emit.c +++ b/src/mesa/pipe/i915simple/i915_fpc_emit.c @@ -199,6 +199,7 @@ uint i915_emit_texld( struct i915_fp_compile *p, * do). Will fallback for now. */ i915_program_error(p, "Can't (yet) swizzle TEX arguments"); + assert(0); return 0; } diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index cd2faa8b97..44a869abaf 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -32,6 +32,9 @@ #include "pipe/tgsi/core/tgsi_token.h" #include "pipe/tgsi/core/tgsi_parse.h" +#include "pipe/draw/draw_vertex.h" + + /** * Simple pass-through fragment shader to use when we don't have * a real shader (or it fails to compile for some reason). @@ -124,7 +127,7 @@ static uint src_vector(struct i915_fp_compile *p, const struct tgsi_full_src_register *source) { - const uint index = source->SrcRegister.Index; + uint index = source->SrcRegister.Index; uint src; switch (source->SrcRegister.File) { @@ -145,23 +148,62 @@ src_vector(struct i915_fp_compile *p, * * We also use a texture coordinate to pass wpos when possible. */ +#if 1 + /* use vertex format info to remap input regs */ + assert(index < p->vertex_info->num_attribs); + printf("%s map index %d to %d\n", + __FUNCTION__, + index, + p->vertex_info->slot_to_attrib[index]); + + index = p->vertex_info->slot_to_attrib[index]; + + switch (index) { + case VF_ATTRIB_POS: + assert(p->wpos_tex != -1); + src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); + break; + case VF_ATTRIB_COLOR0: + src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); + break; + case VF_ATTRIB_COLOR1: + src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); + src = swizzle(src, X, Y, Z, ONE); + break; + case VF_ATTRIB_FOG: + src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); + src = swizzle(src, W, W, W, W); + break; + case VF_ATTRIB_TEX0: + case VF_ATTRIB_TEX1: + case VF_ATTRIB_TEX2: + case VF_ATTRIB_TEX3: + case VF_ATTRIB_TEX4: + case VF_ATTRIB_TEX5: + case VF_ATTRIB_TEX6: + case VF_ATTRIB_TEX7: + src = i915_emit_decl(p, REG_TYPE_T, + T_TEX0 + (index - VF_ATTRIB_TEX0), + D0_CHANNEL_ALL); + break; + + default: + i915_program_error(p, "Bad source->Index"); + return 0; + } + +#else switch (index) { case FRAG_ATTRIB_WPOS: + assert(p->wpos_tex != -1); src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); break; case FRAG_ATTRIB_COL0: src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); break; case FRAG_ATTRIB_COL1: -#if 1 src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); src = swizzle(src, X, Y, Z, ONE); -#else - /* total hack to force texture mapping */ - src = i915_emit_decl(p, REG_TYPE_T, - T_TEX0/* + (index - FRAG_ATTRIB_TEX0)*/, - D0_CHANNEL_ALL); -#endif break; case FRAG_ATTRIB_FOGC: src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); @@ -184,6 +226,8 @@ src_vector(struct i915_fp_compile *p, i915_program_error(p, "Bad source->Index"); return 0; } +#endif + break; case TGSI_FILE_CONSTANT: @@ -877,6 +921,8 @@ i915_init_compile(struct i915_context *i915, p->shader = &i915->fs; + p->vertex_info = &i915->current.vertex_info; + /* new constants found during translation get appended after the * user-provided constants. */ diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index ff57d26d38..caac6b1f2a 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -58,7 +58,7 @@ static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) } - +#if 0 /* Hardcoded vertex format: xyz/rgba */ static INLINE void @@ -75,15 +75,22 @@ emit_hw_vertex( struct i915_context *i915, float_to_ubyte( vertex->data[1][0] ), float_to_ubyte( vertex->data[1][3] )) ); } - - +#endif + +/** + * Extract the needed fields from vertex_header and emit i915 dwords. + * Recall that the vertices are constructed by the 'draw' module and + * have a couple of slots at the beginning (1-dword header, 4-dword + * clip pos) that we ignore here. + */ static INLINE void emit_hw_vertex2( struct i915_context *i915, const struct vertex_header *vertex) { const struct vertex_info *vinfo = &i915->current.vertex_info; uint i; + uint count = 0; /* for debug/sanity */ for (i = 0; i < vinfo->num_attribs; i++) { switch (vinfo->format[i]) { @@ -92,32 +99,38 @@ emit_hw_vertex2( struct i915_context *i915, break; case FORMAT_1F: OUT_BATCH( fui(vertex->data[i][0]) ); + count++; break; case FORMAT_2F: OUT_BATCH( fui(vertex->data[i][0]) ); OUT_BATCH( fui(vertex->data[i][1]) ); + count += 2; break; case FORMAT_3F: OUT_BATCH( fui(vertex->data[i][0]) ); OUT_BATCH( fui(vertex->data[i][1]) ); OUT_BATCH( fui(vertex->data[i][2]) ); + count += 3; break; case FORMAT_4F: OUT_BATCH( fui(vertex->data[i][0]) ); OUT_BATCH( fui(vertex->data[i][1]) ); OUT_BATCH( fui(vertex->data[i][2]) ); OUT_BATCH( fui(vertex->data[i][3]) ); + count += 4; break; case FORMAT_4UB: OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[i][2] ), float_to_ubyte( vertex->data[i][1] ), float_to_ubyte( vertex->data[i][0] ), float_to_ubyte( vertex->data[i][3] )) ); + count += 1; break; default: assert(0); } } + assert(count == vinfo->size); } @@ -129,10 +142,16 @@ emit_prim( struct draw_stage *stage, unsigned nr ) { struct i915_context *i915 = setup_stage(stage)->i915; +#if 0 unsigned vertex_size = 4 * sizeof(int); +#else + unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ +#endif unsigned *ptr; unsigned i; + assert(vertex_size >= 12); /* never smaller than 12 bytes */ + if (i915->dirty) i915_update_derived( i915 ); diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 8f8b13253d..f5ea721cc8 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -156,7 +156,6 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, if (buf->size && (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_FLAG_READ))) { memcpy(i915->current.constants[shader], mapped, buf->size); - fprintf(stderr, "i915 problem: map of constant buffer failed\n"); ws->buffer_unmap(ws, buf->buffer); i915->current.num_user_constants[shader] = buf->size / (4 * sizeof(float)); diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index bd952cc567..f8c56775ea 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -25,8 +25,10 @@ * **************************************************************************/ -#include "vf/vf.h" + +#include "pipe/p_util.h" #include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_vertex.h" #include "i915_context.h" #include "i915_state.h" #include "i915_reg.h" @@ -67,12 +69,47 @@ emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) const uint n = vinfo->num_attribs; vinfo->attr_mask |= (1 << vfAttr); vinfo->slot_to_attrib[n] = vfAttr; + printf("Vertex slot %d = vfattrib %d\n", n, vfAttr); /*vinfo->interp_mode[n] = interpMode;*/ vinfo->format[n] = format; vinfo->num_attribs++; } +/** + * Recompute the vinfo->size field. + */ +static void +compute_vertex_size(struct vertex_info *vinfo) +{ + uint i; + + vinfo->size = 0; + for (i = 0; i < vinfo->num_attribs; i++) { + switch (vinfo->format[i]) { + case FORMAT_OMIT: + break; + case FORMAT_4UB: + /* fall-through */ + case FORMAT_1F: + vinfo->size += 1; + break; + case FORMAT_2F: + vinfo->size += 2; + break; + case FORMAT_3F: + vinfo->size += 3; + break; + case FORMAT_4F: + vinfo->size += 4; + break; + default: + assert(0); + } + } +} + + /** * Determine which post-transform / pre-rasterization vertex attributes @@ -111,7 +148,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) } } - for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { + for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7/*MAX*/; i++) { uint hwtc; if (inputsRead & (1 << i)) { hwtc = TEXCOORDFMT_4D; @@ -140,6 +177,8 @@ static void calculate_vertex_layout( struct i915_context *i915 ) } } + compute_vertex_size(vinfo); + /* If the attributes have changed, tell the draw module about the new * vertex layout. We'll also update the hardware vertex format info. */ diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index ceb96289b5..48eb86091d 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -190,37 +190,39 @@ i915_emit_hardware_state(struct i915_context *i915 ) } -#if 0 +#if 01 /* texture images */ if (i915->hardware_dirty & I915_HW_MAP) { const uint nr = i915->current.sampler_enable_nr; - const uint enabled = i915->current.sampler_enable_flags; - uint unit; - uint count = 0; - OUT_BATCH(_3DSTATE_MAP_STATE | (3 * nr)); - OUT_BATCH(enabled); - for (unit = 0; unit < I915_TEX_UNITS; unit++) { - if (enabled & (1 << unit)) { - struct pipe_buffer_handle *buf = - i915->texture[unit]->region->buffer; - uint offset = 0; - assert(buf); - - count++; - - OUT_RELOC(buf, - I915_BUFFER_ACCESS_READ, - offset); - OUT_BATCH(i915->current.texbuffer[unit][0]); /* MS3 */ - OUT_BATCH(i915->current.texbuffer[unit][1]); /* MS4 */ + if (nr) { + const uint enabled = i915->current.sampler_enable_flags; + uint unit; + uint count = 0; + OUT_BATCH(_3DSTATE_MAP_STATE | (3 * nr)); + OUT_BATCH(enabled); + for (unit = 0; unit < I915_TEX_UNITS; unit++) { + if (enabled & (1 << unit)) { + struct pipe_buffer_handle *buf = + i915->texture[unit]->region->buffer; + uint offset = 0; + assert(buf); + + count++; + + OUT_RELOC(buf, + I915_BUFFER_ACCESS_READ, + offset); + OUT_BATCH(i915->current.texbuffer[unit][0]); /* MS3 */ + OUT_BATCH(i915->current.texbuffer[unit][1]); /* MS4 */ + } } + assert(count == nr); } - assert(count == nr); } #endif -#if 0 +#if 01 /* samplers */ if (i915->hardware_dirty & I915_HW_SAMPLER) { diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index c6f1f59f88..81f2051be1 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -53,17 +53,22 @@ static void upload_S2S4(struct i915_context *i915) unsigned LIS2, LIS4; /* I915_NEW_VERTEX_FORMAT */ -#if 01 +#if 0 LIS2 = 0xffffffff; LIS4 = (S4_VFMT_XYZ | S4_VFMT_COLOR); #else + /* assert(LIS2 == i915->current.vertex_info.hwfmt[1]); assert(LIS4 == i915->current.vertex_info.hwfmt[0]); + */ LIS2 = i915->current.vertex_info.hwfmt[1]; LIS4 = i915->current.vertex_info.hwfmt[0]; - printf("UPLOAD FORMAT LIS2: 0x%x LIS4: 0x%x\n", LIS2, LIS4); #endif - + printf("DEFAULT FORMT LIS2: 0x%x LIS4: 0x%x\n", ~0, (S4_VFMT_XYZ | S4_VFMT_COLOR)); + printf("UPLOAD FORMAT LIS2: 0x%x LIS4: 0x%x\n", LIS2, LIS4); + printf("VF FORMAT LIS2: 0x%x LIS4: 0x%x\n", + i915->current.vertex_info.hwfmt[1], + i915->current.vertex_info.hwfmt[0]); /* I915_NEW_SETUP */ switch (i915->setup.cull_mode) { diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index eba789e9c4..af230111dc 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -28,72 +28,12 @@ #include "pipe/p_util.h" #include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_vertex.h" #include "sp_context.h" #include "sp_state.h" -/* XXX: copied from vf.h */ -enum { - VF_ATTRIB_POS = 0, - VF_ATTRIB_WEIGHT = 1, - VF_ATTRIB_NORMAL = 2, - VF_ATTRIB_COLOR0 = 3, - VF_ATTRIB_COLOR1 = 4, - VF_ATTRIB_FOG = 5, - VF_ATTRIB_COLOR_INDEX = 6, - VF_ATTRIB_EDGEFLAG = 7, - VF_ATTRIB_TEX0 = 8, - VF_ATTRIB_TEX1 = 9, - VF_ATTRIB_TEX2 = 10, - VF_ATTRIB_TEX3 = 11, - VF_ATTRIB_TEX4 = 12, - VF_ATTRIB_TEX5 = 13, - VF_ATTRIB_TEX6 = 14, - VF_ATTRIB_TEX7 = 15, - VF_ATTRIB_VAR0 = 16, - VF_ATTRIB_VAR1 = 17, - VF_ATTRIB_VAR2 = 18, - VF_ATTRIB_VAR3 = 19, - VF_ATTRIB_VAR4 = 20, - VF_ATTRIB_VAR5 = 21, - VF_ATTRIB_VAR6 = 22, - VF_ATTRIB_VAR7 = 23, - VF_ATTRIB_POINTSIZE = 24, - VF_ATTRIB_BFC0 = 25, - VF_ATTRIB_BFC1 = 26, - VF_ATTRIB_CLIP_POS = 27, - VF_ATTRIB_VERTEX_HEADER = 28, - VF_ATTRIB_MAX = 29 -}; - -/* XXX: copied from config.h */ -#define MAX_VARYING 8 - -/* XXX: copied from mtypes.h */ -enum -{ - FRAG_ATTRIB_WPOS = 0, - FRAG_ATTRIB_COL0 = 1, - FRAG_ATTRIB_COL1 = 2, - FRAG_ATTRIB_FOGC = 3, - FRAG_ATTRIB_TEX0 = 4, - FRAG_ATTRIB_TEX1 = 5, - FRAG_ATTRIB_TEX2 = 6, - FRAG_ATTRIB_TEX3 = 7, - FRAG_ATTRIB_TEX4 = 8, - FRAG_ATTRIB_TEX5 = 9, - FRAG_ATTRIB_TEX6 = 10, - FRAG_ATTRIB_TEX7 = 11, - FRAG_ATTRIB_VAR0 = 12, /**< shader varying */ - FRAG_ATTRIB_MAX = (FRAG_ATTRIB_VAR0 + MAX_VARYING) -}; - -#define FRAG_BIT_COL0 (1 << FRAG_ATTRIB_COL0) -#define FRAG_BIT_COL1 (1 << FRAG_ATTRIB_COL1) - - - #define EMIT_ATTR( VF_ATTR, FRAG_ATTR, INTERP ) \ do { \ -- cgit v1.2.3 From 6d88515190ed7500e201706cc14d30072ba42d6f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 13:41:23 -0600 Subject: remove dead/debug code, misc clean-up --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 47 ++----------------------- src/mesa/pipe/i915simple/i915_prim_emit.c | 30 ++-------------- src/mesa/pipe/i915simple/i915_state_derived.c | 13 ++----- src/mesa/pipe/i915simple/i915_state_immediate.c | 26 +++++--------- 4 files changed, 16 insertions(+), 100 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 44a869abaf..be3de0f3f4 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -148,14 +148,9 @@ src_vector(struct i915_fp_compile *p, * * We also use a texture coordinate to pass wpos when possible. */ -#if 1 - /* use vertex format info to remap input regs */ - assert(index < p->vertex_info->num_attribs); - printf("%s map index %d to %d\n", - __FUNCTION__, - index, - p->vertex_info->slot_to_attrib[index]); + /* use vertex format info to map a slot number to a VF attrib */ + assert(index < p->vertex_info->num_attribs); index = p->vertex_info->slot_to_attrib[index]; switch (index) { @@ -186,48 +181,10 @@ src_vector(struct i915_fp_compile *p, T_TEX0 + (index - VF_ATTRIB_TEX0), D0_CHANNEL_ALL); break; - default: i915_program_error(p, "Bad source->Index"); return 0; } - -#else - switch (index) { - case FRAG_ATTRIB_WPOS: - assert(p->wpos_tex != -1); - src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); - break; - case FRAG_ATTRIB_COL0: - src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); - break; - case FRAG_ATTRIB_COL1: - src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); - src = swizzle(src, X, Y, Z, ONE); - break; - case FRAG_ATTRIB_FOGC: - src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); - src = swizzle(src, W, W, W, W); - break; - case FRAG_ATTRIB_TEX0: - case FRAG_ATTRIB_TEX1: - case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: - case FRAG_ATTRIB_TEX4: - case FRAG_ATTRIB_TEX5: - case FRAG_ATTRIB_TEX6: - case FRAG_ATTRIB_TEX7: - src = i915_emit_decl(p, REG_TYPE_T, - T_TEX0 + (index - FRAG_ATTRIB_TEX0), - D0_CHANNEL_ALL); - break; - - default: - i915_program_error(p, "Bad source->Index"); - return 0; - } -#endif - break; case TGSI_FILE_CONSTANT: diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index caac6b1f2a..fc26d26627 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -58,26 +58,6 @@ static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) } -#if 0 -/* Hardcoded vertex format: xyz/rgba - */ -static INLINE void -emit_hw_vertex( struct i915_context *i915, - struct vertex_header *vertex ) -{ - OUT_BATCH( fui(vertex->data[0][0]) ); - OUT_BATCH( fui(vertex->data[0][1]) ); - OUT_BATCH( fui(vertex->data[0][2]) ); - - /* colors are ARGB (MSB to LSB) */ - OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[1][2] ), - float_to_ubyte( vertex->data[1][1] ), - float_to_ubyte( vertex->data[1][0] ), - float_to_ubyte( vertex->data[1][3] )) ); -} -#endif - - /** * Extract the needed fields from vertex_header and emit i915 dwords. * Recall that the vertices are constructed by the 'draw' module and @@ -85,8 +65,8 @@ emit_hw_vertex( struct i915_context *i915, * clip pos) that we ignore here. */ static INLINE void -emit_hw_vertex2( struct i915_context *i915, - const struct vertex_header *vertex) +emit_hw_vertex( struct i915_context *i915, + const struct vertex_header *vertex) { const struct vertex_info *vinfo = &i915->current.vertex_info; uint i; @@ -142,11 +122,7 @@ emit_prim( struct draw_stage *stage, unsigned nr ) { struct i915_context *i915 = setup_stage(stage)->i915; -#if 0 - unsigned vertex_size = 4 * sizeof(int); -#else unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ -#endif unsigned *ptr; unsigned i; @@ -182,7 +158,7 @@ emit_prim( struct draw_stage *stage, ((4 + vertex_size * nr)/4 - 2)); for (i = 0; i < nr; i++) { - emit_hw_vertex2(i915, prim->v[i]); + emit_hw_vertex(i915, prim->v[i]); ptr += vertex_size / sizeof(int); } } diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index f8c56775ea..e8c7dfba6f 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -32,9 +32,7 @@ #include "i915_context.h" #include "i915_state.h" #include "i915_reg.h" - -/* XXX should include i915_fpc.h but that causes some trouble atm */ -extern void i915_translate_fragment_program( struct i915_context *i915 ); +#include "i915_fpc.h" @@ -119,9 +117,8 @@ compute_vertex_size(struct vertex_info *vinfo) static void calculate_vertex_layout( struct i915_context *i915 ) { const unsigned inputsRead = i915->fs.inputs_read; -// const unsigned inputsRead = (FRAG_BIT_WPOS | FRAG_BIT_COL0); - unsigned i; struct vertex_info *vinfo = &i915->current.vertex_info; + uint i; memset(vinfo, 0, sizeof(*vinfo)); @@ -167,12 +164,10 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ if (i915->setup.light_twoside) { if (inputsRead & FRAG_BIT_COL0) { - /* XXX: mark as discarded after setup */ emit_vertex_attr(vinfo, VF_ATTRIB_BFC0, FORMAT_OMIT); } if (inputsRead & FRAG_BIT_COL1) { - /* XXX: discard after setup */ emit_vertex_attr(vinfo, VF_ATTRIB_BFC1, FORMAT_OMIT); } } @@ -185,10 +180,6 @@ static void calculate_vertex_layout( struct i915_context *i915 ) draw_set_vertex_attributes( i915->draw, vinfo->slot_to_attrib, vinfo->num_attribs); -#if 0 - printf("VERTEX_FORMAT LIS2: 0x%x LIS4: 0x%x\n", - vinfo->hwfmt[1], vinfo->hwfmt[0]); -#endif } diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 81f2051be1..38a24733e1 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -51,24 +51,16 @@ static void upload_S2S4(struct i915_context *i915) { unsigned LIS2, LIS4; - + /* I915_NEW_VERTEX_FORMAT */ -#if 0 - LIS2 = 0xffffffff; - LIS4 = (S4_VFMT_XYZ | S4_VFMT_COLOR); -#else - /* - assert(LIS2 == i915->current.vertex_info.hwfmt[1]); - assert(LIS4 == i915->current.vertex_info.hwfmt[0]); - */ - LIS2 = i915->current.vertex_info.hwfmt[1]; - LIS4 = i915->current.vertex_info.hwfmt[0]; -#endif - printf("DEFAULT FORMT LIS2: 0x%x LIS4: 0x%x\n", ~0, (S4_VFMT_XYZ | S4_VFMT_COLOR)); - printf("UPLOAD FORMAT LIS2: 0x%x LIS4: 0x%x\n", LIS2, LIS4); - printf("VF FORMAT LIS2: 0x%x LIS4: 0x%x\n", - i915->current.vertex_info.hwfmt[1], - i915->current.vertex_info.hwfmt[0]); + { + LIS2 = i915->current.vertex_info.hwfmt[1]; + LIS4 = i915->current.vertex_info.hwfmt[0]; + /* + printf("LIS2: 0x%x LIS4: 0x%x\n", LIS2, LIS4); + */ + assert(LIS4); /* should never be zero? */ + } /* I915_NEW_SETUP */ switch (i915->setup.cull_mode) { -- cgit v1.2.3 From 399077d760b4c98ab38d48d17f2480114e70dfc6 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 14:55:43 -0600 Subject: clean-up comments --- src/mesa/pipe/i915simple/i915_fpc.h | 19 ++++++------------- src/mesa/pipe/i915simple/i915_prim_emit.c | 3 ++- 2 files changed, 8 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index 7bbfc69d29..59fcbf40d9 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -56,22 +56,15 @@ struct i915_fp_compile { uint num_constants; uint constant_flags[I915_MAX_CONSTANT]; /**< status of each constant */ - uint *csr; /* Cursor, points into program. - */ + uint *csr; /**< Cursor, points into program. */ - uint *decl; /* Cursor, points into declarations. - */ + uint *decl; /**< Cursor, points into declarations. */ - uint decl_s; /* flags for which s regs need to be decl'd */ - uint decl_t; /* flags for which t regs need to be decl'd */ + uint decl_s; /**< flags for which s regs need to be decl'd */ + uint decl_t; /**< flags for which t regs need to be decl'd */ - uint temp_flag; /* Tracks temporary regs which are in - * use. - */ - - uint utemp_flag; /* Tracks TYPE_U temporary regs which are in - * use. - */ + uint temp_flag; /**< Tracks temporary regs which are in use */ + uint utemp_flag; /**< Tracks TYPE_U temporary regs which are in use */ uint nr_tex_indirect; uint nr_tex_insn; diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index fc26d26627..fcf3c99cac 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -201,7 +201,8 @@ static void reset_stipple_counter( struct draw_stage *stage ) /** - * Create a new primitive setup/render stage. + * Create a new primitive setup/render stage. This gets plugged into + * the 'draw' module's pipeline. */ struct draw_stage *i915_draw_render_stage( struct i915_context *i915 ) { -- cgit v1.2.3 From ba50b6958ab69e5e3ee190126bb08aad6118c607 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 14:56:00 -0600 Subject: comment about code duplication --- src/mesa/pipe/draw/draw_arrays.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index 430dc15ce2..784eb3f2e6 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -88,6 +88,7 @@ emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) } +/** XXX this is duplicated in the i915 driver... */ static void compute_vertex_size(struct vertex_info *vinfo) { -- cgit v1.2.3 From 074f07176a3c916573d9e2b4cb20316591c1e99e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 14:58:26 -0600 Subject: define new TGSI_ATTRIB_ tokens for program inputs/outputs, plus translation functions --- src/mesa/pipe/tgsi/core/tgsi_token.h | 42 +++++++ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 207 +++++++++++++++++++++++++++++++++ 2 files changed, 249 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index 3e66d55b7d..a12a2d7370 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -1477,6 +1477,48 @@ struct tgsi_dst_register_ext_predicate unsigned Extended : 1; /* BOOL */ }; + +/** + * The specific values here are not important. + */ +enum { + TGSI_ATTRIB_POS = 0, + TGSI_ATTRIB_WEIGHT = 1, + TGSI_ATTRIB_NORMAL = 2, + TGSI_ATTRIB_COLOR0 = 3, + TGSI_ATTRIB_COLOR1 = 4, + TGSI_ATTRIB_FOG = 5, + TGSI_ATTRIB_COLOR_INDEX = 6, /* XXX omit? */ + TGSI_ATTRIB_EDGEFLAG = 7, + TGSI_ATTRIB_TEX0 = 8, + TGSI_ATTRIB_TEX1 = 9, + TGSI_ATTRIB_TEX2 = 10, + TGSI_ATTRIB_TEX3 = 11, + TGSI_ATTRIB_TEX4 = 12, + TGSI_ATTRIB_TEX5 = 13, + TGSI_ATTRIB_TEX6 = 14, + TGSI_ATTRIB_TEX7 = 15, + TGSI_ATTRIB_VAR0 = 16, + TGSI_ATTRIB_VAR1 = 17, + TGSI_ATTRIB_VAR2 = 18, + TGSI_ATTRIB_VAR3 = 19, + TGSI_ATTRIB_VAR4 = 20, + TGSI_ATTRIB_VAR5 = 21, + TGSI_ATTRIB_VAR6 = 22, + TGSI_ATTRIB_VAR7 = 23, + TGSI_ATTRIB_POINTSIZE = 24, + TGSI_ATTRIB_BFC0 = 25, + TGSI_ATTRIB_BFC1 = 26, + TGSI_ATTRIB_CLIP_POS = 27, + TGSI_ATTRIB_VERTEX_HEADER = 28, + TGSI_ATTRIB_MAX = 29 +}; + + +#define TGSI_MAX_TEXTURE 8 +#define TGSI_MAX_VARYING 8 + + #if defined __cplusplus } // extern "C" #endif // defined __cplusplus diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index eeaed844d5..dfb263ebdc 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -3,6 +3,213 @@ #define TGSI_DEBUG 1 + +/** + * Convert a VERT_ATTRIB_x to a TGSI_ATTRIB_y + */ +static GLuint +translate_vertex_input(GLuint attrib) +{ + /* XXX these could be implemented with array lookups too.... */ + switch (attrib) { + case VERT_ATTRIB_POS: + return TGSI_ATTRIB_POS; + case VERT_ATTRIB_WEIGHT: + return TGSI_ATTRIB_WEIGHT; + case VERT_ATTRIB_NORMAL: + return TGSI_ATTRIB_NORMAL; + case VERT_ATTRIB_COLOR0: + return TGSI_ATTRIB_COLOR0; + case VERT_ATTRIB_COLOR1: + return TGSI_ATTRIB_COLOR1; + case VERT_ATTRIB_FOG: + return TGSI_ATTRIB_FOG; + case VERT_ATTRIB_COLOR_INDEX: + return TGSI_ATTRIB_COLOR_INDEX; + case VERT_ATTRIB_EDGEFLAG: + return TGSI_ATTRIB_EDGEFLAG; + case VERT_ATTRIB_TEX0: + return TGSI_ATTRIB_TEX0; + case VERT_ATTRIB_TEX1: + return TGSI_ATTRIB_TEX1; + case VERT_ATTRIB_TEX2: + return TGSI_ATTRIB_TEX2; + case VERT_ATTRIB_TEX3: + return TGSI_ATTRIB_TEX3; + case VERT_ATTRIB_TEX4: + return TGSI_ATTRIB_TEX4; + case VERT_ATTRIB_TEX5: + return TGSI_ATTRIB_TEX5; + case VERT_ATTRIB_TEX6: + return TGSI_ATTRIB_TEX6; + case VERT_ATTRIB_TEX7: + return TGSI_ATTRIB_TEX7; + case VERT_ATTRIB_GENERIC0: + return TGSI_ATTRIB_VAR0; + case VERT_ATTRIB_GENERIC1: + return TGSI_ATTRIB_VAR1; + case VERT_ATTRIB_GENERIC2: + return TGSI_ATTRIB_VAR2; + case VERT_ATTRIB_GENERIC3: + return TGSI_ATTRIB_VAR3; + case VERT_ATTRIB_GENERIC4: + return TGSI_ATTRIB_VAR4; + case VERT_ATTRIB_GENERIC5: + return TGSI_ATTRIB_VAR5; + case VERT_ATTRIB_GENERIC6: + return TGSI_ATTRIB_VAR6; + case VERT_ATTRIB_GENERIC7: + return TGSI_ATTRIB_VAR7; + default: + assert(0); + return 0; + } +} + + +/** + * Convert VERT_RESULT_x to TGSI_ATTRIB_y + */ +static GLuint +translate_vertex_ouput(GLuint attrib) +{ + switch (attrib) { + case VERT_RESULT_HPOS: + return TGSI_ATTRIB_POS; + case VERT_RESULT_COL0: + return TGSI_ATTRIB_COLOR0; + case VERT_RESULT_COL1: + return TGSI_ATTRIB_COLOR1; + case VERT_RESULT_FOGC: + return TGSI_ATTRIB_FOG; + case VERT_RESULT_TEX0: + return TGSI_ATTRIB_TEX0; + case VERT_RESULT_TEX1: + return TGSI_ATTRIB_TEX1; + case VERT_RESULT_TEX2: + return TGSI_ATTRIB_TEX2; + case VERT_RESULT_TEX3: + return TGSI_ATTRIB_TEX3; + case VERT_RESULT_TEX4: + return TGSI_ATTRIB_TEX4; + case VERT_RESULT_TEX5: + return TGSI_ATTRIB_TEX5; + case VERT_RESULT_TEX6: + return TGSI_ATTRIB_TEX6; + case VERT_RESULT_TEX7: + return TGSI_ATTRIB_TEX7; + case VERT_RESULT_PSIZ: + return TGSI_ATTRIB_POINTSIZE; + case VERT_RESULT_BFC0: + return TGSI_ATTRIB_BFC0; + case VERT_RESULT_BFC1: + return TGSI_ATTRIB_BFC1; + case VERT_RESULT_VAR0: + return TGSI_ATTRIB_VAR0; + case VERT_RESULT_VAR0 + 1: + return TGSI_ATTRIB_VAR1; + case VERT_RESULT_VAR0 + 2: + return TGSI_ATTRIB_VAR2; + case VERT_RESULT_VAR0 + 3: + return TGSI_ATTRIB_VAR3; + case VERT_RESULT_VAR0 + 4: + return TGSI_ATTRIB_VAR4; + case VERT_RESULT_VAR0 + 5: + return TGSI_ATTRIB_VAR5; + case VERT_RESULT_VAR0 + 6: + return TGSI_ATTRIB_VAR6; + case VERT_RESULT_VAR0 + 7: + return TGSI_ATTRIB_VAR7; + default: + assert(0); + return 0; + } +} + + +/** + * Convert a FRAG_ATTRIB_x to a TGSI_ATTRIB_y + */ +static GLuint +translate_fragment_input(GLuint attrib) +{ + switch (attrib) { + case FRAG_ATTRIB_WPOS: + return TGSI_ATTRIB_POS; + case FRAG_ATTRIB_COL0: + return TGSI_ATTRIB_COLOR0; + case FRAG_ATTRIB_COL1: + return TGSI_ATTRIB_COLOR1; + case FRAG_ATTRIB_FOGC: + return TGSI_ATTRIB_FOG; + case FRAG_ATTRIB_TEX0: + return TGSI_ATTRIB_TEX0; + case FRAG_ATTRIB_TEX1: + return TGSI_ATTRIB_TEX1; + case FRAG_ATTRIB_TEX2: + return TGSI_ATTRIB_TEX2; + case FRAG_ATTRIB_TEX3: + return TGSI_ATTRIB_TEX3; + case FRAG_ATTRIB_TEX4: + return TGSI_ATTRIB_TEX4; + case FRAG_ATTRIB_TEX5: + return TGSI_ATTRIB_TEX5; + case FRAG_ATTRIB_TEX6: + return TGSI_ATTRIB_TEX6; + case FRAG_ATTRIB_TEX7: + return TGSI_ATTRIB_TEX7; + case FRAG_ATTRIB_VAR0: + return TGSI_ATTRIB_VAR0; + case FRAG_ATTRIB_VAR0 + 1: + return TGSI_ATTRIB_VAR1; + case FRAG_ATTRIB_VAR0 + 2: + return TGSI_ATTRIB_VAR2; + case FRAG_ATTRIB_VAR0 + 3: + return TGSI_ATTRIB_VAR3; + case FRAG_ATTRIB_VAR0 + 4: + return TGSI_ATTRIB_VAR4; + case FRAG_ATTRIB_VAR0 + 5: + return TGSI_ATTRIB_VAR5; + case FRAG_ATTRIB_VAR0 + 6: + return TGSI_ATTRIB_VAR6; + case FRAG_ATTRIB_VAR0 + 7: + return TGSI_ATTRIB_VAR7; + default: + assert(0); + return 0; + } +} + + +/** + * Convert FRAG_RESULT_x to TGSI_ATTRIB_y + */ +static GLuint +translate_fragment_output(GLuint attrib) +{ + switch (attrib) { + case FRAG_RESULT_DEPR: + return TGSI_ATTRIB_POS; + case FRAG_RESULT_COLR: + /* fall-through */ + case FRAG_RESULT_COLH: + /* fall-through */ + case FRAG_RESULT_DATA0: + return TGSI_ATTRIB_COLOR0; + case FRAG_RESULT_DATA0 + 1: + return TGSI_ATTRIB_COLOR0 + 1; + case FRAG_RESULT_DATA0 + 2: + return TGSI_ATTRIB_COLOR0 + 2; + case FRAG_RESULT_DATA0 + 3: + return TGSI_ATTRIB_COLOR0 + 3; + default: + assert(0); + return 0; + } +} + + + /* * Map mesa register file to TGSI register file. */ -- cgit v1.2.3 From 83547d1dca7281ffe03424d12316b26bb07b89c9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 16:59:08 -0600 Subject: TGSI_ATTRIB_ tokens --- src/mesa/pipe/tgsi/core/tgsi_attribs.h | 47 ++++++++++++++++++++++++++++++++++ 1 file changed, 47 insertions(+) create mode 100644 src/mesa/pipe/tgsi/core/tgsi_attribs.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/tgsi_attribs.h b/src/mesa/pipe/tgsi/core/tgsi_attribs.h new file mode 100644 index 0000000000..b8f6ba7da8 --- /dev/null +++ b/src/mesa/pipe/tgsi/core/tgsi_attribs.h @@ -0,0 +1,47 @@ +#ifndef TGSI_ATTRIBS_H +#define TGSI_ATTRIBS_H + + +/** + * The specific values here are not important (could remove them). + */ +enum { + TGSI_ATTRIB_POS = 0, + TGSI_ATTRIB_WEIGHT = 1, + TGSI_ATTRIB_NORMAL = 2, + TGSI_ATTRIB_COLOR0 = 3, + TGSI_ATTRIB_COLOR1 = 4, + TGSI_ATTRIB_FOG = 5, + TGSI_ATTRIB_COLOR_INDEX = 6, /* XXX omit? */ + TGSI_ATTRIB_EDGEFLAG = 7, + TGSI_ATTRIB_TEX0 = 8, + TGSI_ATTRIB_TEX1 = 9, + TGSI_ATTRIB_TEX2 = 10, + TGSI_ATTRIB_TEX3 = 11, + TGSI_ATTRIB_TEX4 = 12, + TGSI_ATTRIB_TEX5 = 13, + TGSI_ATTRIB_TEX6 = 14, + TGSI_ATTRIB_TEX7 = 15, + TGSI_ATTRIB_VAR0 = 16, + TGSI_ATTRIB_VAR1 = 17, + TGSI_ATTRIB_VAR2 = 18, + TGSI_ATTRIB_VAR3 = 19, + TGSI_ATTRIB_VAR4 = 20, + TGSI_ATTRIB_VAR5 = 21, + TGSI_ATTRIB_VAR6 = 22, + TGSI_ATTRIB_VAR7 = 23, + TGSI_ATTRIB_POINTSIZE = 24, + TGSI_ATTRIB_BFC0 = 25, + TGSI_ATTRIB_BFC1 = 26, + TGSI_ATTRIB_CLIP_POS = 27, + TGSI_ATTRIB_VERTEX_HEADER = 28, + TGSI_ATTRIB_MAX = 29 +}; + + +#define TGSI_MAX_TEXTURE 8 +#define TGSI_MAX_VARYING 8 + + + +#endif /* TGSI_ATTRIBS_H */ -- cgit v1.2.3 From d8b16d416de95daa4f0ede9b839bdbf0fa6bf1b1 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 17:00:47 -0600 Subject: Checkpoint: new vertex/fragment attribute naming Replace VF_ATTRIB_x with TGSI_ATTRIB_x When converting mesa programs to TGSI programs, also convert the InputsRead and OutputsWritten to a mask of TGSI_ATTRIB_ bits. Still need to do conversion for vertex programs... --- src/mesa/pipe/draw/draw_arrays.c | 9 +-- src/mesa/pipe/draw/draw_clip.c | 4 +- src/mesa/pipe/draw/draw_flatshade.c | 8 +-- src/mesa/pipe/draw/draw_private.h | 3 +- src/mesa/pipe/draw/draw_twoside.c | 8 +-- src/mesa/pipe/draw/draw_vertex.h | 83 +---------------------- src/mesa/pipe/i915simple/i915_fpc_emit.c | 9 --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 39 +++++------ src/mesa/pipe/i915simple/i915_state_derived.c | 96 ++++++++++----------------- src/mesa/pipe/p_state.h | 4 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 17 ++--- src/mesa/pipe/softpipe/sp_quad_fs.c | 3 +- src/mesa/pipe/softpipe/sp_state_derived.c | 72 ++++++++++---------- src/mesa/pipe/tgsi/core/tgsi_token.h | 41 ------------ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 84 ++++++++++++++++++++--- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 25 +++++++ src/mesa/state_tracker/st_atom_fs.c | 6 +- src/mesa/state_tracker/st_cb_clear.c | 3 +- 18 files changed, 228 insertions(+), 286 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index 784eb3f2e6..f6bf174dc0 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -78,6 +78,7 @@ emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) vinfo->slot_to_attrib[n] = vfAttr; if (n >= 2) { /* the first two slots are the vertex header & clippos */ + assert(vfAttr < Elements(vinfo->attrib_to_slot)); vinfo->attrib_to_slot[vfAttr] = n - 2; } /*printf("Vertex slot %d = vfattrib %d\n", n, vfAttr);*/ @@ -128,16 +129,16 @@ draw_set_vertex_attributes( struct draw_context *draw, struct vertex_info *vinfo = &draw->vertex_info; unsigned i; - assert(slot_to_vf_attr[0] == VF_ATTRIB_POS); + assert(slot_to_vf_attr[0] == TGSI_ATTRIB_POS); memset(vinfo, 0, sizeof(*vinfo)); /* * First three attribs are always the same: header, clip pos, winpos */ - emit_vertex_attr(vinfo, VF_ATTRIB_VERTEX_HEADER, FORMAT_1F); - emit_vertex_attr(vinfo, VF_ATTRIB_CLIP_POS, FORMAT_4F); - emit_vertex_attr(vinfo, VF_ATTRIB_POS, FORMAT_4F_VIEWPORT); + emit_vertex_attr(vinfo, TGSI_ATTRIB_VERTEX_HEADER, FORMAT_1F); + emit_vertex_attr(vinfo, TGSI_ATTRIB_CLIP_POS, FORMAT_4F); + emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F_VIEWPORT); /* * Remaining attribs (color, texcoords, etc) diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 1c2491d2c6..f46f4e3de6 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -380,8 +380,8 @@ static void clip_begin( struct draw_stage *stage ) /* sanity checks. If these fail, review the clip/interp code! */ assert(stage->draw->vertex_info.num_attribs >= 3); - assert(stage->draw->vertex_info.slot_to_attrib[0] == VF_ATTRIB_VERTEX_HEADER); - assert(stage->draw->vertex_info.slot_to_attrib[1] == VF_ATTRIB_CLIP_POS); + assert(stage->draw->vertex_info.slot_to_attrib[0] == TGSI_ATTRIB_VERTEX_HEADER); + assert(stage->draw->vertex_info.slot_to_attrib[1] == TGSI_ATTRIB_CLIP_POS); /* Hacky bitmask to use when we hit CLIP_USER_BIT: */ diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 34588c83b6..d8db1f748c 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -72,10 +72,10 @@ static INLINE void copy_colors( struct draw_stage *stage, const struct flatshade_stage *flatshade = flatshade_stage(stage); const unsigned *lookup = flatshade->lookup; - copy_attr( lookup[VF_ATTRIB_COLOR0], dst, src ); - copy_attr( lookup[VF_ATTRIB_COLOR1], dst, src ); - copy_attr( lookup[VF_ATTRIB_BFC0], dst, src ); - copy_attr( lookup[VF_ATTRIB_BFC1], dst, src ); + copy_attr( lookup[TGSI_ATTRIB_COLOR0], dst, src ); + copy_attr( lookup[TGSI_ATTRIB_COLOR1], dst, src ); + copy_attr( lookup[TGSI_ATTRIB_BFC0], dst, src ); + copy_attr( lookup[TGSI_ATTRIB_BFC1], dst, src ); } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 531bb2e254..c5d4c62f71 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -61,7 +61,8 @@ struct vertex_header { float data[][4]; /* Note variable size */ }; -#define MAX_VERTEX_SIZE ((2 + FRAG_ATTRIB_MAX) * 4 * sizeof(float)) +/* XXX This is too large */ +#define MAX_VERTEX_SIZE ((2 + TGSI_ATTRIB_MAX) * 4 * sizeof(float)) diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 822cadd61f..4fd87786f8 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -79,12 +79,12 @@ static struct vertex_header *copy_bfc( struct twoside_stage *twoside, { struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); - copy_color( twoside->lookup[VF_ATTRIB_COLOR0], - twoside->lookup[VF_ATTRIB_BFC0], + copy_color( twoside->lookup[TGSI_ATTRIB_COLOR0], + twoside->lookup[TGSI_ATTRIB_BFC0], tmp ); - copy_color( twoside->lookup[VF_ATTRIB_COLOR1], - twoside->lookup[VF_ATTRIB_BFC1], + copy_color( twoside->lookup[TGSI_ATTRIB_COLOR1], + twoside->lookup[TGSI_ATTRIB_BFC1], tmp ); return tmp; diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 3262431487..0e15ea89a2 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -33,84 +33,7 @@ #ifndef DRAW_VERTEX_H #define DRAW_VERTEX_H - -/*** - *** XXX There's a lot of legacy tokens here that'll eventually go away. - *** (at least we don't include vf/vf.h anymore) - ***/ - - -enum { - VF_ATTRIB_POS = 0, - VF_ATTRIB_WEIGHT = 1, - VF_ATTRIB_NORMAL = 2, - VF_ATTRIB_COLOR0 = 3, - VF_ATTRIB_COLOR1 = 4, - VF_ATTRIB_FOG = 5, - VF_ATTRIB_COLOR_INDEX = 6, - VF_ATTRIB_EDGEFLAG = 7, - VF_ATTRIB_TEX0 = 8, - VF_ATTRIB_TEX1 = 9, - VF_ATTRIB_TEX2 = 10, - VF_ATTRIB_TEX3 = 11, - VF_ATTRIB_TEX4 = 12, - VF_ATTRIB_TEX5 = 13, - VF_ATTRIB_TEX6 = 14, - VF_ATTRIB_TEX7 = 15, - VF_ATTRIB_VAR0 = 16, - VF_ATTRIB_VAR1 = 17, - VF_ATTRIB_VAR2 = 18, - VF_ATTRIB_VAR3 = 19, - VF_ATTRIB_VAR4 = 20, - VF_ATTRIB_VAR5 = 21, - VF_ATTRIB_VAR6 = 22, - VF_ATTRIB_VAR7 = 23, - VF_ATTRIB_POINTSIZE = 24, - VF_ATTRIB_BFC0 = 25, - VF_ATTRIB_BFC1 = 26, - VF_ATTRIB_CLIP_POS = 27, - VF_ATTRIB_VERTEX_HEADER = 28, - VF_ATTRIB_MAX = 29 -}; - -#define MAX_VARYING 8 -enum -{ - FRAG_ATTRIB_WPOS = 0, - FRAG_ATTRIB_COL0 = 1, - FRAG_ATTRIB_COL1 = 2, - FRAG_ATTRIB_FOGC = 3, - FRAG_ATTRIB_TEX0 = 4, - FRAG_ATTRIB_TEX1 = 5, - FRAG_ATTRIB_TEX2 = 6, - FRAG_ATTRIB_TEX3 = 7, - FRAG_ATTRIB_TEX4 = 8, - FRAG_ATTRIB_TEX5 = 9, - FRAG_ATTRIB_TEX6 = 10, - FRAG_ATTRIB_TEX7 = 11, - FRAG_ATTRIB_VAR0 = 12, /**< shader varying */ - FRAG_ATTRIB_MAX = (FRAG_ATTRIB_VAR0 + MAX_VARYING) -}; - -#define FRAG_BIT_WPOS (1 << FRAG_ATTRIB_WPOS) -#define FRAG_BIT_COL0 (1 << FRAG_ATTRIB_COL0) -#define FRAG_BIT_COL1 (1 << FRAG_ATTRIB_COL1) -#define FRAG_BIT_FOGC (1 << FRAG_ATTRIB_FOGC) -#define FRAG_BIT_TEX0 (1 << FRAG_ATTRIB_TEX0) - - - -#define MAX_DRAW_BUFFERS 4 - -enum -{ - FRAG_RESULT_COLR = 0, - FRAG_RESULT_COLH = 1, - FRAG_RESULT_DEPR = 2, - FRAG_RESULT_DATA0 = 3, - FRAG_RESULT_MAX = (FRAG_RESULT_DATA0 + MAX_DRAW_BUFFERS) -}; - +#include "pipe/tgsi/core/tgsi_attribs.h" #define MAX_VERT_ATTRIBS 12 /* OK? */ @@ -127,10 +50,10 @@ enum struct vertex_info { uint num_attribs; - uint hwfmt[2]; /**< hardware format info for this format */ + uint hwfmt[4]; /**< hardware format info for this format */ uint attr_mask; /**< mask of VF_ATTR_ bits */ uint slot_to_attrib[MAX_VERT_ATTRIBS]; - uint attrib_to_slot[VF_ATTRIB_MAX]; + uint attrib_to_slot[TGSI_ATTRIB_MAX]; uint interp_mode[MAX_VERT_ATTRIBS]; uint format[MAX_VERT_ATTRIBS]; /**< FORMAT_x */ uint size; /**< total vertex size in dwords */ diff --git a/src/mesa/pipe/i915simple/i915_fpc_emit.c b/src/mesa/pipe/i915simple/i915_fpc_emit.c index f062885f8a..c8d36435d9 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_emit.c +++ b/src/mesa/pipe/i915simple/i915_fpc_emit.c @@ -353,23 +353,14 @@ i915_emit_param4fv(struct i915_fp_compile * p, const float * values) return UREG(REG_TYPE_CONST, fp->param[i].reg); } -#if 0 - if (fp->nr_constants == I915_MAX_CONSTANT || - fp->nr_params == I915_MAX_CONSTANT) { -#else if (p->constants->nr_constants == I915_MAX_CONSTANT || fp->nr_params == I915_MAX_CONSTANT) { -#endif i915_program_error(p, "i915_emit_param4fv: out of constants\n"); return 0; } { -#if 0 - int reg = fp->nr_constants++; -#else int reg = p->constants->nr_constants++; -#endif int i = fp->nr_params++; assert (p->constant_flags[reg] == 0); diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index be3de0f3f4..e7315d2263 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -154,31 +154,31 @@ src_vector(struct i915_fp_compile *p, index = p->vertex_info->slot_to_attrib[index]; switch (index) { - case VF_ATTRIB_POS: + case TGSI_ATTRIB_POS: assert(p->wpos_tex != -1); src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); break; - case VF_ATTRIB_COLOR0: + case TGSI_ATTRIB_COLOR0: src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); break; - case VF_ATTRIB_COLOR1: + case TGSI_ATTRIB_COLOR1: src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); src = swizzle(src, X, Y, Z, ONE); break; - case VF_ATTRIB_FOG: + case TGSI_ATTRIB_FOG: src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); src = swizzle(src, W, W, W, W); break; - case VF_ATTRIB_TEX0: - case VF_ATTRIB_TEX1: - case VF_ATTRIB_TEX2: - case VF_ATTRIB_TEX3: - case VF_ATTRIB_TEX4: - case VF_ATTRIB_TEX5: - case VF_ATTRIB_TEX6: - case VF_ATTRIB_TEX7: + case TGSI_ATTRIB_TEX0: + case TGSI_ATTRIB_TEX1: + case TGSI_ATTRIB_TEX2: + case TGSI_ATTRIB_TEX3: + case TGSI_ATTRIB_TEX4: + case TGSI_ATTRIB_TEX5: + case TGSI_ATTRIB_TEX6: + case TGSI_ATTRIB_TEX7: src = i915_emit_decl(p, REG_TYPE_T, - T_TEX0 + (index - VF_ATTRIB_TEX0), + T_TEX0 + (index - TGSI_ATTRIB_TEX0), D0_CHANNEL_ALL); break; default: @@ -237,9 +237,9 @@ get_result_vector(struct i915_fp_compile *p, switch (dest->DstRegister.File) { case TGSI_FILE_OUTPUT: switch (dest->DstRegister.Index) { - case 1: /*COLOR*/ /*FRAG_RESULT_COLR:*/ + case TGSI_ATTRIB_COLOR0: return UREG(REG_TYPE_OC, 0); - case 0: /*DEPTH*/ /*FRAG_RESULT_DEPR:*/ + case TGSI_ATTRIB_POS: return UREG(REG_TYPE_OD, 0); default: i915_program_error(p, "Bad inst->DstReg.Index"); @@ -989,14 +989,15 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) static void i915_find_wpos_space(struct i915_fp_compile *p) { - const uint inputs = p->shader->inputs_read | FRAG_BIT_WPOS; /*XXX hack*/ + const uint inputs + = p->shader->inputs_read | (1 << TGSI_ATTRIB_POS); /*XXX hack*/ uint i; p->wpos_tex = -1; - if (inputs & FRAG_BIT_WPOS) { + if (inputs & (1 << TGSI_ATTRIB_POS)) { for (i = 0; i < I915_TEX_UNITS; i++) { - if ((inputs & (FRAG_BIT_TEX0 << i)) == 0) { + if ((inputs & (1 << (TGSI_ATTRIB_TEX0 + i))) == 0) { p->wpos_tex = i; return; } @@ -1017,7 +1018,7 @@ i915_find_wpos_space(struct i915_fp_compile *p) static void i915_fixup_depth_write(struct i915_fp_compile *p) { - if (p->shader->outputs_written & (1<shader->outputs_written & (1 << TGSI_ATTRIB_POS)) { uint depth = UREG(REG_TYPE_OD, 0); i915_emit_arith(p, diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index e8c7dfba6f..426d43f288 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -35,32 +35,6 @@ #include "i915_fpc.h" - -static const unsigned frag_to_vf[FRAG_ATTRIB_MAX] = -{ - VF_ATTRIB_POS, - VF_ATTRIB_COLOR0, - VF_ATTRIB_COLOR1, - VF_ATTRIB_FOG, - VF_ATTRIB_TEX0, - VF_ATTRIB_TEX1, - VF_ATTRIB_TEX2, - VF_ATTRIB_TEX3, - VF_ATTRIB_TEX4, - VF_ATTRIB_TEX5, - VF_ATTRIB_TEX6, - VF_ATTRIB_TEX7, - VF_ATTRIB_VAR0, - VF_ATTRIB_VAR1, - VF_ATTRIB_VAR2, - VF_ATTRIB_VAR3, - VF_ATTRIB_VAR4, - VF_ATTRIB_VAR5, - VF_ATTRIB_VAR6, - VF_ATTRIB_VAR7, -}; - - static INLINE void emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) { @@ -118,44 +92,45 @@ static void calculate_vertex_layout( struct i915_context *i915 ) { const unsigned inputsRead = i915->fs.inputs_read; struct vertex_info *vinfo = &i915->current.vertex_info; - uint i; memset(vinfo, 0, sizeof(*vinfo)); /* TODO - Figure out if we need to do perspective divide, etc. */ - emit_vertex_attr(vinfo, VF_ATTRIB_POS, FORMAT_3F); + + /* pos */ + emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_3F); vinfo->hwfmt[0] |= S4_VFMT_XYZ; - - /* Pull in the rest of the attributes. They are all in float4 - * format. Future optimizations could be to keep some attributes - * as fixed point or ubyte format. - */ - for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { - if (inputsRead & (1 << i)) { - assert(i < Elements(frag_to_vf)); - if (i915->setup.flatshade - && (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1)) { - emit_vertex_attr(vinfo, frag_to_vf[i], FORMAT_4UB); - } - else { - emit_vertex_attr(vinfo, frag_to_vf[i], FORMAT_4UB); - } - vinfo->hwfmt[0] |= S4_VFMT_COLOR; - } + + /* color0 */ + if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { + emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, FORMAT_4UB); + vinfo->hwfmt[0] |= S4_VFMT_COLOR; } - for (i = FRAG_ATTRIB_TEX0; i <= FRAG_ATTRIB_TEX7/*MAX*/; i++) { - uint hwtc; - if (inputsRead & (1 << i)) { - hwtc = TEXCOORDFMT_4D; - assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); - emit_vertex_attr(vinfo, frag_to_vf[i], FORMAT_4F); - } - else { - hwtc = TEXCOORDFMT_NOT_PRESENT; + /* color 1 */ + if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { + assert(0); /* untested */ + emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, FORMAT_4UB); + vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; + } + + /* XXX fog? */ + + /* texcoords */ + { + uint i; + for (i = TGSI_ATTRIB_TEX0; i <= TGSI_ATTRIB_TEX7; i++) { + uint hwtc; + if (inputsRead & (1 << i)) { + emit_vertex_attr(vinfo, i, FORMAT_4F); + hwtc = TEXCOORDFMT_4D; + } + else { + hwtc = TEXCOORDFMT_NOT_PRESENT; + } + vinfo->hwfmt[1] |= hwtc << ((i - TGSI_ATTRIB_TEX0) * 4); } - vinfo->hwfmt[1] |= hwtc << ((i - FRAG_ATTRIB_TEX0) * 4); } /* Additional attributes required for setup: Just twosided @@ -163,12 +138,11 @@ static void calculate_vertex_layout( struct i915_context *i915 ) * the vertex header. */ if (i915->setup.light_twoside) { - if (inputsRead & FRAG_BIT_COL0) { - emit_vertex_attr(vinfo, VF_ATTRIB_BFC0, FORMAT_OMIT); - } - - if (inputsRead & FRAG_BIT_COL1) { - emit_vertex_attr(vinfo, VF_ATTRIB_BFC1, FORMAT_OMIT); + if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { + emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT); + } + if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { + emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, FORMAT_OMIT); } } diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index d29b85117b..e562a8d058 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -125,8 +125,8 @@ struct pipe_constant_buffer { struct pipe_shader_state { - unsigned inputs_read; /**< FRAG/VERT_ATTRIB_x */ - unsigned outputs_written; /**< FRAG/VERT_RESULT_x */ + unsigned inputs_read; /**< TGSI_ATTRIB_ bits */ + unsigned outputs_written; /**< TGSI_ATTRIB_ bits */ const struct tgsi_token *tokens; }; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index f779deac28..e8ed548cf0 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -40,10 +40,7 @@ #include "pipe/draw/draw_private.h" #include "pipe/p_util.h" - -/** XXX remove */ -#define FRAG_ATTRIB_WPOS 0 -#define FRAG_ATTRIB_MAX 13 +#include "pipe/draw/draw_vertex.h" /** @@ -82,7 +79,7 @@ struct setup_stage { float oneoverarea; - struct tgsi_interp_coef coef[FRAG_ATTRIB_MAX]; + struct tgsi_interp_coef coef[TGSI_ATTRIB_MAX]; struct quad_header quad; struct { @@ -369,7 +366,7 @@ static void const_coeff( struct setup_stage *setup, unsigned slot, unsigned i ) { - assert(slot < FRAG_ATTRIB_MAX); + assert(slot < TGSI_ATTRIB_MAX); assert(i <= 3); setup->coef[slot].dadx[i] = 0; @@ -394,7 +391,7 @@ static void tri_linear_coeff( struct setup_stage *setup, float a = setup->ebot.dy * majda - botda * setup->emaj.dy; float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - assert(slot < FRAG_ATTRIB_MAX); + assert(slot < TGSI_ATTRIB_MAX); assert(i <= 3); setup->coef[slot].dadx[i] = a * setup->oneoverarea; @@ -445,7 +442,7 @@ static void tri_persp_coeff( struct setup_stage *setup, float a = setup->ebot.dy * majda - botda * setup->emaj.dy; float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - assert(slot < FRAG_ATTRIB_MAX); + assert(slot < TGSI_ATTRIB_MAX); assert(i <= 3); setup->coef[slot].dadx[i] = a * setup->oneoverarea; @@ -891,8 +888,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const float halfSize = 0.5f * setup->softpipe->setup.point_size; const boolean round = setup->softpipe->setup.point_smooth; const struct vertex_header *v0 = prim->v[0]; - const float x = v0->data[FRAG_ATTRIB_WPOS][0]; - const float y = v0->data[FRAG_ATTRIB_WPOS][1]; + const float x = v0->data[TGSI_ATTRIB_POS][0]; + const float y = v0->data[TGSI_ATTRIB_POS][1]; unsigned slot, j; /* For points, all interpolants are constant-valued. diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 8a419c9ac7..8d41e09465 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -34,6 +34,7 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" +#include "pipe/tgsi/core/tgsi_attribs.h" #include "sp_context.h" #include "sp_headers.h" @@ -138,7 +139,7 @@ shade_quad( /* store result color */ memcpy( quad->outputs.color, - &machine.Outputs[1].xyzw[0].f[0], + &machine.Outputs[TGSI_ATTRIB_COLOR0].xyzw[0].f[0], sizeof( quad->outputs.color ) ); #if 0 diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index af230111dc..5c119ec8cd 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -33,41 +33,43 @@ #include "sp_context.h" #include "sp_state.h" +#include "pipe/tgsi/core/tgsi_attribs.h" -#define EMIT_ATTR( VF_ATTR, FRAG_ATTR, INTERP ) \ + +#define EMIT_ATTR( ATTR, FRAG_ATTR, INTERP ) \ do { \ - slot_to_vf_attr[softpipe->nr_attrs] = VF_ATTR; \ - softpipe->vf_attr_to_slot[VF_ATTR] = softpipe->nr_attrs; \ + slot_to_vf_attr[softpipe->nr_attrs] = ATTR; \ + softpipe->vf_attr_to_slot[ATTR] = softpipe->nr_attrs; \ softpipe->fp_attr_to_slot[FRAG_ATTR] = softpipe->nr_attrs; \ softpipe->interp[softpipe->nr_attrs] = INTERP; \ softpipe->nr_attrs++; \ - attr_mask |= (1 << (VF_ATTR)); \ + attr_mask |= (1 << (ATTR)); \ } while (0) static const unsigned frag_to_vf[PIPE_ATTRIB_MAX] = { - VF_ATTRIB_POS, - VF_ATTRIB_COLOR0, - VF_ATTRIB_COLOR1, - VF_ATTRIB_FOG, - VF_ATTRIB_TEX0, - VF_ATTRIB_TEX1, - VF_ATTRIB_TEX2, - VF_ATTRIB_TEX3, - VF_ATTRIB_TEX4, - VF_ATTRIB_TEX5, - VF_ATTRIB_TEX6, - VF_ATTRIB_TEX7, - VF_ATTRIB_VAR0, - VF_ATTRIB_VAR1, - VF_ATTRIB_VAR2, - VF_ATTRIB_VAR3, - VF_ATTRIB_VAR4, - VF_ATTRIB_VAR5, - VF_ATTRIB_VAR6, - VF_ATTRIB_VAR7, + TGSI_ATTRIB_POS, + TGSI_ATTRIB_COLOR0, + TGSI_ATTRIB_COLOR1, + TGSI_ATTRIB_FOG, + TGSI_ATTRIB_TEX0, + TGSI_ATTRIB_TEX1, + TGSI_ATTRIB_TEX2, + TGSI_ATTRIB_TEX3, + TGSI_ATTRIB_TEX4, + TGSI_ATTRIB_TEX5, + TGSI_ATTRIB_TEX6, + TGSI_ATTRIB_TEX7, + TGSI_ATTRIB_VAR0, + TGSI_ATTRIB_VAR1, + TGSI_ATTRIB_VAR2, + TGSI_ATTRIB_VAR3, + TGSI_ATTRIB_VAR4, + TGSI_ATTRIB_VAR5, + TGSI_ATTRIB_VAR6, + TGSI_ATTRIB_VAR7, }; @@ -79,7 +81,7 @@ static const unsigned frag_to_vf[PIPE_ATTRIB_MAX] = static void calculate_vertex_layout( struct softpipe_context *softpipe ) { const unsigned inputsRead = softpipe->fs.inputs_read; - unsigned slot_to_vf_attr[VF_ATTRIB_MAX]; + unsigned slot_to_vf_attr[TGSI_ATTRIB_MAX]; unsigned attr_mask = 0x0; unsigned i; @@ -87,7 +89,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) * fragment position (XYZW). */ if (softpipe->depth_test.enabled || - (inputsRead & (1 << FRAG_ATTRIB_WPOS))) + (inputsRead & (1 << TGSI_ATTRIB_POS))) softpipe->need_z = TRUE; else softpipe->need_z = FALSE; @@ -95,7 +97,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* Need W if we do any perspective-corrected interpolation or the * fragment program uses the fragment position. */ - if (inputsRead & (1 << FRAG_ATTRIB_WPOS)) + if (inputsRead & (1 << TGSI_ATTRIB_POS)) softpipe->need_w = TRUE; else softpipe->need_w = FALSE; @@ -109,24 +111,24 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* TODO - Figure out if we need to do perspective divide, etc. */ - EMIT_ATTR(VF_ATTRIB_POS, FRAG_ATTRIB_WPOS, INTERP_LINEAR); + EMIT_ATTR(TGSI_ATTRIB_POS, TGSI_ATTRIB_POS, INTERP_LINEAR); /* Pull in the rest of the attributes. They are all in float4 * format. Future optimizations could be to keep some attributes * as fixed point or ubyte format. */ - for (i = 1; i < FRAG_ATTRIB_TEX0; i++) { + for (i = 1; i < TGSI_ATTRIB_TEX0; i++) { if (inputsRead & (1 << i)) { assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); if (softpipe->setup.flatshade - && (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1)) + && (i == TGSI_ATTRIB_COLOR0 || i == TGSI_ATTRIB_COLOR1)) EMIT_ATTR(frag_to_vf[i], i, INTERP_CONSTANT); else EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); } } - for (i = FRAG_ATTRIB_TEX0; i < FRAG_ATTRIB_MAX; i++) { + for (i = TGSI_ATTRIB_TEX0; i < TGSI_ATTRIB_MAX; i++) { if (inputsRead & (1 << i)) { assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); @@ -141,12 +143,12 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) * the vertex header. */ if (softpipe->setup.light_twoside) { - if (inputsRead & FRAG_BIT_COL0) { - EMIT_ATTR(VF_ATTRIB_BFC0, FRAG_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ + if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { + EMIT_ATTR(TGSI_ATTRIB_BFC0, TGSI_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ } - if (inputsRead & FRAG_BIT_COL1) { - EMIT_ATTR(VF_ATTRIB_BFC1, FRAG_ATTRIB_MAX, 0); /* XXX: discard after setup */ + if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { + EMIT_ATTR(TGSI_ATTRIB_BFC1, TGSI_ATTRIB_MAX, 0); /* XXX: discard after setup */ } } diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h index a12a2d7370..ca53071a60 100644 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ b/src/mesa/pipe/tgsi/core/tgsi_token.h @@ -1478,47 +1478,6 @@ struct tgsi_dst_register_ext_predicate }; -/** - * The specific values here are not important. - */ -enum { - TGSI_ATTRIB_POS = 0, - TGSI_ATTRIB_WEIGHT = 1, - TGSI_ATTRIB_NORMAL = 2, - TGSI_ATTRIB_COLOR0 = 3, - TGSI_ATTRIB_COLOR1 = 4, - TGSI_ATTRIB_FOG = 5, - TGSI_ATTRIB_COLOR_INDEX = 6, /* XXX omit? */ - TGSI_ATTRIB_EDGEFLAG = 7, - TGSI_ATTRIB_TEX0 = 8, - TGSI_ATTRIB_TEX1 = 9, - TGSI_ATTRIB_TEX2 = 10, - TGSI_ATTRIB_TEX3 = 11, - TGSI_ATTRIB_TEX4 = 12, - TGSI_ATTRIB_TEX5 = 13, - TGSI_ATTRIB_TEX6 = 14, - TGSI_ATTRIB_TEX7 = 15, - TGSI_ATTRIB_VAR0 = 16, - TGSI_ATTRIB_VAR1 = 17, - TGSI_ATTRIB_VAR2 = 18, - TGSI_ATTRIB_VAR3 = 19, - TGSI_ATTRIB_VAR4 = 20, - TGSI_ATTRIB_VAR5 = 21, - TGSI_ATTRIB_VAR6 = 22, - TGSI_ATTRIB_VAR7 = 23, - TGSI_ATTRIB_POINTSIZE = 24, - TGSI_ATTRIB_BFC0 = 25, - TGSI_ATTRIB_BFC1 = 26, - TGSI_ATTRIB_CLIP_POS = 27, - TGSI_ATTRIB_VERTEX_HEADER = 28, - TGSI_ATTRIB_MAX = 29 -}; - - -#define TGSI_MAX_TEXTURE 8 -#define TGSI_MAX_VARYING 8 - - #if defined __cplusplus } // extern "C" #endif // defined __cplusplus diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index dfb263ebdc..993d220c50 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -1,5 +1,7 @@ #include "tgsi_platform.h" #include "tgsi_mesa.h" +#include "pipe/tgsi/core/tgsi_attribs.h" +#include "pipe/tgsi/mesa/mesa_to_tgsi.h" #define TGSI_DEBUG 1 @@ -7,8 +9,8 @@ /** * Convert a VERT_ATTRIB_x to a TGSI_ATTRIB_y */ -static GLuint -translate_vertex_input(GLuint attrib) +uint +tgsi_mesa_translate_vertex_input(GLuint attrib) { /* XXX these could be implemented with array lookups too.... */ switch (attrib) { @@ -70,8 +72,8 @@ translate_vertex_input(GLuint attrib) /** * Convert VERT_RESULT_x to TGSI_ATTRIB_y */ -static GLuint -translate_vertex_ouput(GLuint attrib) +uint +tgsi_mesa_translate_vertex_output(GLuint attrib) { switch (attrib) { case VERT_RESULT_HPOS: @@ -130,8 +132,8 @@ translate_vertex_ouput(GLuint attrib) /** * Convert a FRAG_ATTRIB_x to a TGSI_ATTRIB_y */ -static GLuint -translate_fragment_input(GLuint attrib) +uint +tgsi_mesa_translate_fragment_input(GLuint attrib) { switch (attrib) { case FRAG_ATTRIB_WPOS: @@ -184,8 +186,8 @@ translate_fragment_input(GLuint attrib) /** * Convert FRAG_RESULT_x to TGSI_ATTRIB_y */ -static GLuint -translate_fragment_output(GLuint attrib) +uint +tgsi_mesa_translate_fragment_output(GLuint attrib) { switch (attrib) { case FRAG_RESULT_DEPR: @@ -209,6 +211,68 @@ translate_fragment_output(GLuint attrib) } +uint +tgsi_mesa_translate_vertex_input_mask(GLbitfield mask) +{ + uint tgsiMask = 0x0; + uint i; + for (i = 0; i < VERT_ATTRIB_MAX && mask; i++) { + if (mask & (1 << i)) { + tgsiMask |= 1 << tgsi_mesa_translate_vertex_input(i); + } + mask &= ~(1 << i); + } + return tgsiMask; +} + + +uint +tgsi_mesa_translate_vertex_output_mask(GLbitfield mask) +{ + uint tgsiMask = 0x0; + uint i; + for (i = 0; i < VERT_RESULT_MAX && mask; i++) { + if (mask & (1 << i)) { + tgsiMask |= 1 << tgsi_mesa_translate_vertex_output(i); + } + mask &= ~(1 << i); + } + return tgsiMask; +} + +uint +tgsi_mesa_translate_fragment_input_mask(GLbitfield mask) +{ + uint tgsiMask = 0x0; + uint i; + for (i = 0; i < FRAG_ATTRIB_MAX && mask; i++) { + if (mask & (1 << i)) { + tgsiMask |= 1 << tgsi_mesa_translate_fragment_input(i); + } + mask &= ~(1 << i); + } + return tgsiMask; +} + + +uint +tgsi_mesa_translate_fragment_output_mask(GLbitfield mask) +{ + uint tgsiMask = 0x0; + uint i; + for (i = 0; i < FRAG_RESULT_MAX && mask; i++) { + if (mask & (1 << i)) { + tgsiMask |= 1 << tgsi_mesa_translate_fragment_output(i); + } + mask &= ~(1 << i); + } + return tgsiMask; +} + + + + + /* * Map mesa register file to TGSI register file. @@ -290,11 +354,11 @@ map_register_file_index( * color results -> index 1, 2, ... */ if( index == FRAG_RESULT_DEPR ) { - mapped_index = 0; + mapped_index = TGSI_ATTRIB_POS; } else { assert( index == FRAG_RESULT_COLR ); - mapped_index = index + 1; + mapped_index = TGSI_ATTRIB_COLOR0; } } else { diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index 9256318997..fda3fa397f 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -19,6 +19,31 @@ tgsi_mesa_compile_vp_program( struct tgsi_token *tokens, GLuint maxTokens ); +uint +tgsi_mesa_translate_vertex_input(GLuint attrib); + +uint +tgsi_mesa_translate_vertex_output(GLuint attrib); + +uint +tgsi_mesa_translate_fragment_input(GLuint attrib); + +uint +tgsi_mesa_translate_fragment_output(GLuint attrib); + +uint +tgsi_mesa_translate_vertex_input_mask(GLbitfield mask); + +uint +tgsi_mesa_translate_vertex_output_mask(GLbitfield mask); + +uint +tgsi_mesa_translate_fragment_input_mask(GLbitfield mask); + +uint +tgsi_mesa_translate_fragment_output_mask(GLbitfield mask); + + #if defined __cplusplus } // extern "C" #endif // defined __cplusplus diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 019b6457e0..a34a227ac0 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -110,8 +110,10 @@ static void update_fs( struct st_context *st ) /* update pipe state */ memset( &fs, 0, sizeof(fs) ); - fs.inputs_read = fp->Base.Base.InputsRead; - fs.outputs_written = fp->Base.Base.OutputsWritten; + fs.inputs_read + = tgsi_mesa_translate_fragment_input_mask(fp->Base.Base.InputsRead); + fs.outputs_written + = tgsi_mesa_translate_fragment_output_mask(fp->Base.Base.OutputsWritten); fs.tokens = &fp->tokens[0]; if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0 || diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index e3690deb5a..69b985e405 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -348,7 +348,8 @@ clear_with_quad(GLcontext *ctx, stfp = make_color_shader(st); } memset(&fs, 0, sizeof(fs)); - fs.inputs_read = stfp->Base.Base.InputsRead; + fs.inputs_read = tgsi_mesa_translate_fragment_input_mask(stfp->Base.Base.InputsRead); + fs.outputs_written = tgsi_mesa_translate_fragment_output_mask(stfp->Base.Base.OutputsWritten); fs.tokens = &stfp->tokens[0]; pipe->set_fs_state(pipe, &fs); } -- cgit v1.2.3 From 7d5e7f03191673a02f8d3b9d5e4b95c074c5acdd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 17:39:52 -0600 Subject: map Mesa vertex attribs/arrays to TGSI attribs --- src/mesa/state_tracker/st_draw.c | 74 +++++++++++++++++++++++++++++++++++++--- 1 file changed, 69 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 5a13fdd2bf..a88e2106d7 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -44,6 +44,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" +#include "pipe/tgsi/core/tgsi_attribs.h" @@ -71,6 +72,68 @@ pipe_vertex_format(GLenum format, GLuint size) } +/** + * Convert a mesa vertex attribute to a TGSI attribute + */ +static GLuint +tgsi_attrib_to_mesa_attrib(GLuint attr) +{ + switch (attr) { + case TGSI_ATTRIB_POS: + return VERT_ATTRIB_POS; + case TGSI_ATTRIB_WEIGHT: + return VERT_ATTRIB_WEIGHT; + case TGSI_ATTRIB_NORMAL: + return VERT_ATTRIB_NORMAL; + case TGSI_ATTRIB_COLOR0: + return VERT_ATTRIB_COLOR0; + case TGSI_ATTRIB_COLOR1: + return VERT_ATTRIB_COLOR1; + case TGSI_ATTRIB_FOG: + return VERT_ATTRIB_FOG; + case TGSI_ATTRIB_COLOR_INDEX: + return VERT_ATTRIB_COLOR_INDEX; + case TGSI_ATTRIB_EDGEFLAG: + return VERT_ATTRIB_EDGEFLAG; + case TGSI_ATTRIB_TEX0: + return VERT_ATTRIB_TEX0; + case TGSI_ATTRIB_TEX1: + return VERT_ATTRIB_TEX1; + case TGSI_ATTRIB_TEX2: + return VERT_ATTRIB_TEX2; + case TGSI_ATTRIB_TEX3: + return VERT_ATTRIB_TEX3; + case TGSI_ATTRIB_TEX4: + return VERT_ATTRIB_TEX4; + case TGSI_ATTRIB_TEX5: + return VERT_ATTRIB_TEX5; + case TGSI_ATTRIB_TEX6: + return VERT_ATTRIB_TEX6; + case TGSI_ATTRIB_TEX7: + return VERT_ATTRIB_TEX7; + case TGSI_ATTRIB_VAR0: + return VERT_ATTRIB_GENERIC0; + case TGSI_ATTRIB_VAR1: + return VERT_ATTRIB_GENERIC1; + case TGSI_ATTRIB_VAR2: + return VERT_ATTRIB_GENERIC2; + case TGSI_ATTRIB_VAR3: + return VERT_ATTRIB_GENERIC3; + case TGSI_ATTRIB_VAR4: + return VERT_ATTRIB_GENERIC4; + case TGSI_ATTRIB_VAR5: + return VERT_ATTRIB_GENERIC5; + case TGSI_ATTRIB_VAR6: + return VERT_ATTRIB_GENERIC6; + case TGSI_ATTRIB_VAR7: + return VERT_ATTRIB_GENERIC7; + default: + assert(0); + return 0; + } +} + + /** * The default attribute buffer is basically a copy of the @@ -141,27 +204,28 @@ draw_vbo(GLcontext *ctx, velement.src_format = 0; if (attrsNeeded & (1 << attr)) { - struct gl_buffer_object *bufobj = arrays[attr]->BufferObj; + const GLuint mesaAttr = tgsi_attrib_to_mesa_attrib(attr); + struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; if (bufobj && bufobj->Name) { struct st_buffer_object *stobj = st_buffer_object(bufobj); /* Recall that for VBOs, the gl_client_array->Ptr field is * really an offset from the start of the VBO, not a pointer. */ - unsigned offset = (unsigned) arrays[attr]->Ptr; + unsigned offset = (unsigned) arrays[mesaAttr]->Ptr; assert(stobj->buffer); vbuffer.buffer = stobj->buffer; vbuffer.buffer_offset = attr0_offset; /* in bytes */ - vbuffer.pitch = arrays[attr]->StrideB; /* in bytes */ + vbuffer.pitch = arrays[mesaAttr]->StrideB; /* in bytes */ vbuffer.max_index = 0; /* need this? */ velement.src_offset = offset - attr0_offset; /* bytes */ velement.vertex_buffer_index = attr; velement.dst_offset = 0; /* need this? */ - velement.src_format = pipe_vertex_format(arrays[attr]->Type, - arrays[attr]->Size); + velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, + arrays[mesaAttr]->Size); assert(velement.src_format); } else { -- cgit v1.2.3 From d6d995d20c7a5d8ce03829fe8f5edfaf7a6ea2b3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 17:40:19 -0600 Subject: get rid of VERT_RESULT_ usage --- src/mesa/pipe/draw/draw_prim.c | 10 ++-------- 1 file changed, 2 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 23eb578221..247a86bdbf 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -58,12 +58,6 @@ static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { }; - -/** XXX remove */ -#define VERT_RESULT_HPOS 0 -#define VERT_RESULT_MAX 24 - - static INLINE unsigned compute_clipmask(float cx, float cy, float cz, float cw) { @@ -242,7 +236,7 @@ run_vertex_program(struct draw_context *draw, #endif /* store machine results */ - assert(draw->vertex_shader.outputs_written & (1 << VERT_RESULT_HPOS)); + assert(draw->vertex_shader.outputs_written & (1 << TGSI_ATTRIB_POS)); for (j = 0; j < count; j++) { unsigned attr, slot; float x, y, z, w; @@ -280,7 +274,7 @@ run_vertex_program(struct draw_context *draw, /* remaining attributes: */ /* pack into sequential post-transform attrib slots */ slot = 1; - for (attr = 1; attr < VERT_RESULT_MAX; attr++) { + for (attr = 1; attr < TGSI_ATTRIB_MAX; attr++) { if (draw->vertex_shader.outputs_written & (1 << attr)) { assert(slot < draw->vertex_info.num_attribs); vOut[j]->data[slot][0] = machine.Outputs[attr].xyzw[0].f[j]; -- cgit v1.2.3 From eef2e35981b2e4c6f2be6fc6868672888719d4b6 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 17:41:03 -0600 Subject: fix incorrect vertex index for POLYGON drawing --- src/mesa/pipe/draw/draw_prim.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 247a86bdbf..4bb81281b5 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -669,7 +669,7 @@ draw_prim( struct draw_context *draw, unsigned start, unsigned count ) ef_mask, start + i + 1, start + i + 2, - start + i + 0); + start + 0); ef_mask &= ~(1<<2); } -- cgit v1.2.3 From 38091ef42a647b0be20e0462b065cb9d88f0784b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 17:57:08 -0600 Subject: fix an edgeflag logic error, see comments --- src/mesa/pipe/draw/draw_prim.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 4bb81281b5..509ab376aa 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -429,6 +429,11 @@ static struct vertex_header *get_vertex( struct draw_context *draw, draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; draw->vs.queue[draw->vs.queue_nr].elt = i; draw->vs.queue_nr++; + + /* Need to set the vertex's edge flag here. If we're being called + * by do_ef_triangle(), that function needs edge flag info! + */ + draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ } /* Mark slot as in-use: -- cgit v1.2.3 From c998ce1df09f2d9dc834ebdf9484673d8084db34 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 17:59:17 -0600 Subject: minor improvement to get_vertex() in-use flag setting --- src/mesa/pipe/draw/draw_prim.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 509ab376aa..2ba3cb3eb5 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -421,6 +421,8 @@ static struct vertex_header *get_vertex( struct draw_context *draw, */ if (draw->vcache.referenced & (1 << slot)) slot = VCACHE_SIZE + draw->vcache.overflow++; + else + draw->vcache.referenced |= (1 << slot); /* slot now in use */ draw->vcache.idx[slot] = i; @@ -436,10 +438,6 @@ static struct vertex_header *get_vertex( struct draw_context *draw, draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ } - /* Mark slot as in-use: - */ - if (slot < VCACHE_SIZE) - draw->vcache.referenced |= (1 << slot); return draw->vcache.vertex[slot]; } -- cgit v1.2.3 From 30236573dadd83714220b72b0c04f1bbce69fbd6 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 18:54:51 -0600 Subject: Need to copy header->edgeflags to tmp when creating new prim_header. Edge flags were random/undefined otherwise. --- src/mesa/pipe/draw/draw_flatshade.c | 1 + src/mesa/pipe/draw/draw_offset.c | 1 + src/mesa/pipe/draw/draw_twoside.c | 1 + 3 files changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index d8db1f748c..5a23e10c03 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -89,6 +89,7 @@ static void flatshade_tri( struct draw_stage *stage, struct prim_header tmp; tmp.det = header->det; + tmp.edgeflags = header->edgeflags; tmp.v[0] = dup_vert(stage, header->v[0], 0); tmp.v[1] = dup_vert(stage, header->v[1], 1); tmp.v[2] = header->v[2]; diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 83c9b163eb..4f653e8c54 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -115,6 +115,7 @@ static void offset_tri( struct draw_stage *stage, struct prim_header tmp; tmp.det = header->det; + tmp.edgeflags = header->edgeflags; tmp.v[0] = dup_vert(stage, header->v[0], 0); tmp.v[1] = dup_vert(stage, header->v[1], 1); tmp.v[2] = dup_vert(stage, header->v[2], 2); diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 4fd87786f8..5a7697d5f8 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -103,6 +103,7 @@ static void twoside_tri( struct draw_stage *stage, struct prim_header tmp; tmp.det = header->det; + tmp.edgeflags = header->edgeflags; /* copy back colors to front color slots */ tmp.v[0] = copy_bfc(twoside, header->v[0], 0); tmp.v[1] = copy_bfc(twoside, header->v[1], 1); -- cgit v1.2.3 From e967b246ecdc665d9f974d1e35380cb1cd11d61e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 19:03:42 -0600 Subject: Vertex shader outputs are now tightly packed into the output slots. Fix softpipe vertex attribute setup. Also, update vs constants when the projection matrix changes (fixes samples/prim.c) --- src/mesa/pipe/draw/draw_prim.c | 57 ++++++++++++++++++++----------- src/mesa/pipe/softpipe/sp_state_derived.c | 33 ++---------------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 19 +++++++++-- src/mesa/state_tracker/st_atom_vs.c | 8 +++-- 4 files changed, 62 insertions(+), 55 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 2ba3cb3eb5..76c2a821b7 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -200,7 +200,7 @@ run_vertex_program(struct draw_context *draw, machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ #if 0 - if (1/*attr == 0*/) { + if (1) { fprintf(file, "Input vertex %d: attr %d: %f %f %f %f\n", j, attr, p[0], p[1], p[2], p[3]); fflush( file ); @@ -212,12 +212,15 @@ run_vertex_program(struct draw_context *draw, #if 0 printf("Consts:\n"); - for (i = 0; i < 4; i++) { - printf(" %d: %f %f %f %f\n", i, - machine.Consts[i][0], - machine.Consts[i][1], - machine.Consts[i][2], - machine.Consts[i][3]); + { + int i; + for (i = 0; i < 4; i++) { + printf(" %d: %f %f %f %f\n", i, + machine.Consts[i][0], + machine.Consts[i][1], + machine.Consts[i][2], + machine.Consts[i][3]); + } } #endif @@ -238,7 +241,7 @@ run_vertex_program(struct draw_context *draw, /* store machine results */ assert(draw->vertex_shader.outputs_written & (1 << TGSI_ATTRIB_POS)); for (j = 0; j < count; j++) { - unsigned attr, slot; + unsigned /**attr,**/ slot; float x, y, z, w; /* Handle attr[0] (position) specially: */ @@ -247,7 +250,7 @@ run_vertex_program(struct draw_context *draw, z = vOut[j]->clip[2] = machine.Outputs[0].xyzw[2].f[j]; w = vOut[j]->clip[3] = machine.Outputs[0].xyzw[3].f[j]; - vOut[j]->clipmask = compute_clipmask(x, y, z, w); + vOut[j]->clipmask = 0;/*compute_clipmask(x, y, z, w);*/ vOut[j]->edgeflag = 1; /* divide by w */ @@ -273,33 +276,47 @@ run_vertex_program(struct draw_context *draw, /* remaining attributes: */ /* pack into sequential post-transform attrib slots */ +#if 0 slot = 1; for (attr = 1; attr < TGSI_ATTRIB_MAX; attr++) { if (draw->vertex_shader.outputs_written & (1 << attr)) { assert(slot < draw->vertex_info.num_attribs); - vOut[j]->data[slot][0] = machine.Outputs[attr].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine.Outputs[attr].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine.Outputs[attr].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine.Outputs[attr].xyzw[3].f[j]; + vOut[j]->data[slot][0] = machine.Outputs[/*attr*/slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine.Outputs[/*attr*/slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine.Outputs[/*attr*/slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine.Outputs[/*attr*/slot].xyzw[3].f[j]; #if 0 - fprintf(file, "output attrib %d slot %d: %f %f %f %f\n", + fprintf(file, "output attrib %d slot %d: %f %f %f %f vert %p\n", attr, slot, vOut[j]->data[slot][0], vOut[j]->data[slot][1], vOut[j]->data[slot][2], - vOut[j]->data[slot][3]); + vOut[j]->data[slot][3], vOut[j]); #endif slot++; } } - } +#else + + for (slot = 1; slot < draw->vertex_info.num_attribs; slot++) { + vOut[j]->data[slot][0] = machine.Outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine.Outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine.Outputs[slot].xyzw[3].f[j]; #if 0 - memcpy( - quad->outputs.color, - &machine.Outputs[1].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); + fprintf(file, "output attrib slot %d: %f %f %f %f vert %p\n", + slot, + vOut[j]->data[slot][0], + vOut[j]->data[slot][1], + vOut[j]->data[slot][2], + vOut[j]->data[slot][3], vOut[j]); #endif + } + +#endif + + } /* loop over vertices */ } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 5c119ec8cd..a880e23abc 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -48,31 +48,6 @@ do { \ } while (0) -static const unsigned frag_to_vf[PIPE_ATTRIB_MAX] = -{ - TGSI_ATTRIB_POS, - TGSI_ATTRIB_COLOR0, - TGSI_ATTRIB_COLOR1, - TGSI_ATTRIB_FOG, - TGSI_ATTRIB_TEX0, - TGSI_ATTRIB_TEX1, - TGSI_ATTRIB_TEX2, - TGSI_ATTRIB_TEX3, - TGSI_ATTRIB_TEX4, - TGSI_ATTRIB_TEX5, - TGSI_ATTRIB_TEX6, - TGSI_ATTRIB_TEX7, - TGSI_ATTRIB_VAR0, - TGSI_ATTRIB_VAR1, - TGSI_ATTRIB_VAR2, - TGSI_ATTRIB_VAR3, - TGSI_ATTRIB_VAR4, - TGSI_ATTRIB_VAR5, - TGSI_ATTRIB_VAR6, - TGSI_ATTRIB_VAR7, -}; - - /** * Determine which post-transform / pre-rasterization vertex attributes * we need. @@ -119,19 +94,17 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ for (i = 1; i < TGSI_ATTRIB_TEX0; i++) { if (inputsRead & (1 << i)) { - assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); if (softpipe->setup.flatshade && (i == TGSI_ATTRIB_COLOR0 || i == TGSI_ATTRIB_COLOR1)) - EMIT_ATTR(frag_to_vf[i], i, INTERP_CONSTANT); + EMIT_ATTR(i, i, INTERP_CONSTANT); else - EMIT_ATTR(frag_to_vf[i], i, INTERP_LINEAR); + EMIT_ATTR(i, i, INTERP_LINEAR); } } for (i = TGSI_ATTRIB_TEX0; i < TGSI_ATTRIB_MAX; i++) { if (inputsRead & (1 << i)) { - assert(i < sizeof(frag_to_vf) / sizeof(frag_to_vf[0])); - EMIT_ATTR(frag_to_vf[i], i, INTERP_PERSPECTIVE); + EMIT_ATTR(i, i, INTERP_PERSPECTIVE); softpipe->need_w = TRUE; } } diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 993d220c50..f33a702958 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -348,7 +348,9 @@ map_register_file_index( break; case TGSI_FILE_OUTPUT: + /* assert( usage_bitmask == 0x0 ); + */ if( processor == TGSI_PROCESSOR_FRAGMENT ) { /* depth result -> index 0 * color results -> index 1, 2, ... @@ -362,8 +364,14 @@ map_register_file_index( } } else { + /* vertex output slots are tightly packed, find mapped pos */ /* mapped_index = VERT_RESULT_x */ - mapped_index = index; + mapped_index = 0; + for( i = 0; i < index; i++ ) { + if( usage_bitmask & (1 << i) ) { + mapped_index++; + } + } } break; @@ -434,6 +442,7 @@ compile_instruction( const struct prog_instruction *inst, struct tgsi_full_instruction *fullinst, GLuint inputs_read, + GLuint outputs_written, GLuint preamble_size, GLuint processor ) { @@ -453,7 +462,8 @@ compile_instruction( processor, fulldst->DstRegister.File, inst->DstReg.Index, - 0x0 ); + outputs_written + ); fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); for( i = 0; i < fullinst->Instruction.NumSrcRegs; i++ ) { @@ -921,6 +931,7 @@ tgsi_mesa_compile_fp_program( &program->Base.Instructions[i], &fullinst, inputs_read, + ~0, /*outputs_written*/ preamble_size, TGSI_PROCESSOR_FRAGMENT ) ) { assert( i == program->Base.NumInstructions - 1 ); @@ -952,6 +963,9 @@ tgsi_mesa_compile_vp_program( struct tgsi_processor *processor; struct tgsi_full_instruction fullinst; GLuint inputs_read = ~0; + GLuint outputs_written; + + outputs_written = program->Base.OutputsWritten; *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -968,6 +982,7 @@ tgsi_mesa_compile_vp_program( &program->Base.Instructions[i], &fullinst, inputs_read, + outputs_written, 0, TGSI_PROCESSOR_VERTEX ) ) { assert( i == program->Base.NumInstructions - 1 ); diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 8a38020afd..f16fe2ae94 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -113,8 +113,10 @@ static void update_vs( struct st_context *st ) /* update pipe state */ memset( &vs, 0, sizeof(vs) ); - vs.outputs_written = vp->Base.Base.OutputsWritten; - vs.inputs_read = vp->Base.Base.InputsRead; + vs.inputs_read + = tgsi_mesa_translate_vertex_input_mask(vp->Base.Base.InputsRead); + vs.outputs_written + = tgsi_mesa_translate_vertex_output_mask(vp->Base.Base.OutputsWritten); vs.tokens = &vp->tokens[0]; if (memcmp(&vs, &st->state.vs, sizeof(vs)) != 0 || @@ -129,7 +131,7 @@ static void update_vs( struct st_context *st ) const struct st_tracked_state st_update_vs = { .dirty = { - .mesa = _NEW_PROGRAM | _NEW_MODELVIEW, + .mesa = _NEW_PROGRAM | _NEW_MODELVIEW | _NEW_PROJECTION, /*XXX MORE*/ .st = ST_NEW_VERTEX_PROGRAM, }, .update = update_vs -- cgit v1.2.3 From d4883ee24c52b7af0193561b51fbb8023d5eb5e3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 19:05:38 -0600 Subject: remove dead code --- src/mesa/pipe/draw/draw_prim.c | 48 +++++++++--------------------------------- 1 file changed, 10 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 76c2a821b7..18a5d39077 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -140,6 +140,7 @@ run_vertex_program(struct draw_context *draw, const float *trans = draw->viewport.translate; assert(count <= 4); + assert(draw->vertex_shader.outputs_written & (1 << TGSI_ATTRIB_POS)); #if 0 if( file == NULL ) { @@ -200,18 +201,16 @@ run_vertex_program(struct draw_context *draw, machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ #if 0 - if (1) { - fprintf(file, "Input vertex %d: attr %d: %f %f %f %f\n", - j, attr, p[0], p[1], p[2], p[3]); - fflush( file ); - } + fprintf(file, "Input vertex %d: attr %d: %f %f %f %f\n", + j, attr, p[0], p[1], p[2], p[3]); + fflush( file ); #endif } } } #if 0 - printf("Consts:\n"); + printf("Vertex shader Constants:\n"); { int i; for (i = 0; i < 4; i++) { @@ -239,9 +238,8 @@ run_vertex_program(struct draw_context *draw, #endif /* store machine results */ - assert(draw->vertex_shader.outputs_written & (1 << TGSI_ATTRIB_POS)); for (j = 0; j < count; j++) { - unsigned /**attr,**/ slot; + unsigned slot; float x, y, z, w; /* Handle attr[0] (position) specially: */ @@ -265,8 +263,7 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][2] = z * scale[2] + trans[2]; vOut[j]->data[0][3] = w; #if 0 - fprintf(file, "Vert %d: wincoord: %f %f %f %f\n", - j, + fprintf(file, "Vert %d: wincoord: %f %f %f %f\n", j, vOut[j]->data[0][0], vOut[j]->data[0][1], vOut[j]->data[0][2], @@ -274,31 +271,9 @@ run_vertex_program(struct draw_context *draw, fflush( file ); #endif - /* remaining attributes: */ - /* pack into sequential post-transform attrib slots */ -#if 0 - slot = 1; - for (attr = 1; attr < TGSI_ATTRIB_MAX; attr++) { - if (draw->vertex_shader.outputs_written & (1 << attr)) { - assert(slot < draw->vertex_info.num_attribs); - vOut[j]->data[slot][0] = machine.Outputs[/*attr*/slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine.Outputs[/*attr*/slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine.Outputs[/*attr*/slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine.Outputs[/*attr*/slot].xyzw[3].f[j]; -#if 0 - fprintf(file, "output attrib %d slot %d: %f %f %f %f vert %p\n", - attr, slot, - vOut[j]->data[slot][0], - vOut[j]->data[slot][1], - vOut[j]->data[slot][2], - vOut[j]->data[slot][3], vOut[j]); -#endif - slot++; - } - } - -#else - + /* remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ for (slot = 1; slot < draw->vertex_info.num_attribs; slot++) { vOut[j]->data[slot][0] = machine.Outputs[slot].xyzw[0].f[j]; vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j]; @@ -313,9 +288,6 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[slot][3], vOut[j]); #endif } - -#endif - } /* loop over vertices */ } -- cgit v1.2.3 From 440e794b05511c1854c22956e6e306d1b1d8682d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 19:28:59 -0600 Subject: Use vertex_info struct for softpipe - lots of clean-up. --- src/mesa/pipe/softpipe/sp_context.h | 24 +++------ src/mesa/pipe/softpipe/sp_prim_setup.c | 4 +- src/mesa/pipe/softpipe/sp_state_derived.c | 85 ++++++++++++++++--------------- 3 files changed, 51 insertions(+), 62 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 6458573917..ddd06ab0b7 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -35,8 +35,11 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/draw/draw_vertex.h" + #include "sp_quad.h" + struct softpipe_surface; struct softpipe_winsys; struct draw_context; @@ -96,14 +99,6 @@ struct softpipe_context { struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; unsigned dirty; - /* Setup derived state. TODO: this should be passed in the program - * tokens as parameters to DECL instructions. - * - * For now we just set colors to CONST on flatshade, textures to - * perspective always and everything else to linear. - */ - enum interp_mode interp[PIPE_ATTRIB_MAX]; - /* * Mapped vertex buffers */ @@ -112,17 +107,10 @@ struct softpipe_context { /** Mapped constant buffers */ void *mapped_constants[PIPE_SHADER_TYPES]; - /* FS + setup derived state: - */ - - /** Map fragment program attribute to quad/coef array slot */ - unsigned fp_attr_to_slot[PIPE_ATTRIB_MAX]; - /** Map vertex format attribute to a vertex attribute slot */ - unsigned vf_attr_to_slot[PIPE_ATTRIB_MAX]; - unsigned nr_attrs; + /** Vertex format */ + struct vertex_info vertex_info; + unsigned attr_mask; unsigned nr_frag_attrs; /**< number of active fragment attribs */ - unsigned attr_mask; /**< bitfield of VF_ATTRIB_ indexes/bits */ - boolean need_z; /**< produce quad/fragment Z values? */ boolean need_w; /**< produce quad/fragment W values? */ diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index e8ed548cf0..c2a5a7907d 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -459,7 +459,7 @@ static void tri_persp_coeff( struct setup_stage *setup, */ static void setup_tri_coefficients( struct setup_stage *setup ) { - const enum interp_mode *interp = setup->softpipe->interp; + const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; unsigned slot, j; /* z and w are done by linear interpolation: @@ -674,7 +674,7 @@ line_persp_coeff(struct setup_stage *setup, unsigned slot, unsigned i) static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { - const enum interp_mode *interp = setup->softpipe->interp; + const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; unsigned slot, j; /* use setup->vmin, vmax to point to vertices */ diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index a880e23abc..76200857f0 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -37,15 +37,18 @@ -#define EMIT_ATTR( ATTR, FRAG_ATTR, INTERP ) \ -do { \ - slot_to_vf_attr[softpipe->nr_attrs] = ATTR; \ - softpipe->vf_attr_to_slot[ATTR] = softpipe->nr_attrs; \ - softpipe->fp_attr_to_slot[FRAG_ATTR] = softpipe->nr_attrs; \ - softpipe->interp[softpipe->nr_attrs] = INTERP; \ - softpipe->nr_attrs++; \ - attr_mask |= (1 << (ATTR)); \ -} while (0) +static void +emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, + uint interp) +{ + const uint n = vinfo->num_attribs; + vinfo->attr_mask |= (1 << vfAttr); + vinfo->slot_to_attrib[n] = vfAttr; + vinfo->interp_mode[n] = interp; + vinfo->format[n] = format; + vinfo->num_attribs++; +} + /** @@ -56,9 +59,12 @@ do { \ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { const unsigned inputsRead = softpipe->fs.inputs_read; - unsigned slot_to_vf_attr[TGSI_ATTRIB_MAX]; - unsigned attr_mask = 0x0; - unsigned i; + const uint colorInterp + = softpipe->setup.flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + struct vertex_info *vinfo = &softpipe->vertex_info; + uint i; + + memset(vinfo, 0, sizeof(*vinfo)); /* Need Z if depth test is enabled or the fragment program uses the * fragment position (XYZW). @@ -77,39 +83,34 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) else softpipe->need_w = FALSE; + /* position */ + /* TODO - Figure out if we need to do perspective divide, etc. */ + emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F, INTERP_LINEAR); + + /* color0 */ + if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { + emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, FORMAT_4F, colorInterp); + } - softpipe->nr_attrs = 0; - memset(slot_to_vf_attr, 0, sizeof(slot_to_vf_attr)); - - memset(softpipe->fp_attr_to_slot, 0, sizeof(softpipe->fp_attr_to_slot)); - memset(softpipe->vf_attr_to_slot, 0, sizeof(softpipe->vf_attr_to_slot)); + /* color1 */ + if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { + emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, FORMAT_4F, colorInterp); + } - /* TODO - Figure out if we need to do perspective divide, etc. - */ - EMIT_ATTR(TGSI_ATTRIB_POS, TGSI_ATTRIB_POS, INTERP_LINEAR); - - /* Pull in the rest of the attributes. They are all in float4 - * format. Future optimizations could be to keep some attributes - * as fixed point or ubyte format. - */ - for (i = 1; i < TGSI_ATTRIB_TEX0; i++) { - if (inputsRead & (1 << i)) { - if (softpipe->setup.flatshade - && (i == TGSI_ATTRIB_COLOR0 || i == TGSI_ATTRIB_COLOR1)) - EMIT_ATTR(i, i, INTERP_CONSTANT); - else - EMIT_ATTR(i, i, INTERP_LINEAR); - } + /* fog */ + if (inputsRead & (1 << TGSI_ATTRIB_FOG)) { + emit_vertex_attr(vinfo, TGSI_ATTRIB_FOG, FORMAT_1F, INTERP_PERSPECTIVE); } - for (i = TGSI_ATTRIB_TEX0; i < TGSI_ATTRIB_MAX; i++) { + /* texcoords and varying vars */ + for (i = TGSI_ATTRIB_TEX0; i < TGSI_ATTRIB_VAR7; i++) { if (inputsRead & (1 << i)) { - EMIT_ATTR(i, i, INTERP_PERSPECTIVE); + emit_vertex_attr(vinfo, i, FORMAT_4F, INTERP_PERSPECTIVE); softpipe->need_w = TRUE; } } - softpipe->nr_frag_attrs = softpipe->nr_attrs; + softpipe->nr_frag_attrs = vinfo->num_attribs; /* Additional attributes required for setup: Just twosided * lighting. Edgeflag is dealt with specially by setting bits in @@ -117,23 +118,23 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ if (softpipe->setup.light_twoside) { if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - EMIT_ATTR(TGSI_ATTRIB_BFC0, TGSI_ATTRIB_MAX, 0); /* XXX: mark as discarded after setup */ + emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT, INTERP_LINEAR); } if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - EMIT_ATTR(TGSI_ATTRIB_BFC1, TGSI_ATTRIB_MAX, 0); /* XXX: discard after setup */ + emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, FORMAT_OMIT, INTERP_LINEAR); } } /* If the attributes have changed, tell the draw module about * the new vertex layout. */ - if (attr_mask != softpipe->attr_mask) { - softpipe->attr_mask = attr_mask; + if (vinfo->attr_mask != softpipe->attr_mask) { + softpipe->attr_mask = vinfo->attr_mask; draw_set_vertex_attributes( softpipe->draw, - slot_to_vf_attr, - softpipe->nr_attrs ); + vinfo->slot_to_attrib, + vinfo->num_attribs); } } -- cgit v1.2.3 From ac8249d0ed382c547a2219449fa6dea328b19193 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 21:39:51 +0100 Subject: check for null ptr in xm_buffer_unreference() --- src/mesa/drivers/x11/xm_softpipe.c | 20 +++++++++++--------- 1 file changed, 11 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_softpipe.c b/src/mesa/drivers/x11/xm_softpipe.c index 432f306e94..ff33ad7850 100644 --- a/src/mesa/drivers/x11/xm_softpipe.c +++ b/src/mesa/drivers/x11/xm_softpipe.c @@ -112,17 +112,19 @@ xm_buffer_reference(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) static void xm_buffer_unreference(struct pipe_winsys *pws, struct pipe_buffer_handle **buf) { - struct xm_buffer *xm_buf = xm_bo(*buf); - xm_buf->refcount--; - assert(xm_buf->refcount >= 0); - if (xm_buf->refcount == 0) { - if (xm_buf->data) { - free(xm_buf->data); - xm_buf->data = NULL; + if (*buf) { + struct xm_buffer *xm_buf = xm_bo(*buf); + xm_buf->refcount--; + assert(xm_buf->refcount >= 0); + if (xm_buf->refcount == 0) { + if (xm_buf->data) { + free(xm_buf->data); + xm_buf->data = NULL; + } + free(xm_buf); } - free(xm_buf); + *buf = NULL; } - *buf = NULL; } static void -- cgit v1.2.3 From 17598ad591a1eebe66cdefa332600f2952de92a0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 21:59:06 +0100 Subject: fix-up stencil buffers --- src/mesa/drivers/x11/xm_api.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 69b7277c1b..254d6863ba 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -436,6 +436,12 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_DEPTH, rb); } + if (vis->mesa_visual.stencilBits > 0) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_STENCIL, rb); + } + /* * Other renderbuffer (depth, stencil, etc) @@ -443,7 +449,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, _mesa_add_soft_renderbuffers(&b->mesa_buffer, GL_FALSE, /* color */ GL_FALSE,/*vis->mesa_visual.haveDepthBuffer,*/ - vis->mesa_visual.haveStencilBuffer, + GL_FALSE, vis->mesa_visual.haveAccumBuffer, b->swAlpha, vis->mesa_visual.numAuxBuffers > 0 ); -- cgit v1.2.3 From 7fe09341489ff61dc1a3a771fd3e75b3f866d6a9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 21:59:27 +0100 Subject: added GL_STENCIL_INDEX8_EXT --- src/mesa/state_tracker/st_cb_fbo.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 430ac715e5..563b5add38 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -233,6 +233,9 @@ st_new_renderbuffer_fb(GLenum intFormat) case GL_DEPTH24_STENCIL8_EXT: strb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; break; + case GL_STENCIL_INDEX8_EXT: + strb->Base._BaseFormat = GL_STENCIL_INDEX; + break; default: _mesa_problem(NULL, "Unexpected intFormat in st_new_renderbuffer"); -- cgit v1.2.3 From c50e054bd43882778aa8972d2444e5c3178f4c74 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 23 Aug 2007 21:59:48 +0100 Subject: track more vertex-related state changes --- src/mesa/state_tracker/st_atom_vs.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index f16fe2ae94..8c9365d9d8 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -154,7 +154,11 @@ static void update_tnl( struct st_context *st ) const struct st_tracked_state st_update_tnl = { .dirty = { - .mesa = _NEW_PROGRAM | _NEW_LIGHT | _NEW_TEXTURE, /* XXX more? */ + .mesa = (_NEW_PROGRAM | + _NEW_LIGHT | + _NEW_TEXTURE | + _NEW_TRANSFORM | + _NEW_LIGHT), /* XXX more? */ .st = ST_NEW_MESA, /* XXX correct? */ }, .update = update_tnl -- cgit v1.2.3 From f93a360319fb35f9bac571931fc81afb9e0a8ab8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 24 Aug 2007 10:01:03 +0100 Subject: Rename tgsi "core" directory --- src/mesa/pipe/tgsi/core/Makefile | 3 - src/mesa/pipe/tgsi/core/tgsi_attribs.h | 47 - src/mesa/pipe/tgsi/core/tgsi_build.c | 1368 ------------------- src/mesa/pipe/tgsi/core/tgsi_build.h | 320 ----- src/mesa/pipe/tgsi/core/tgsi_core.h | 12 - src/mesa/pipe/tgsi/core/tgsi_dump.c | 1402 ------------------- src/mesa/pipe/tgsi/core/tgsi_dump.h | 22 - src/mesa/pipe/tgsi/core/tgsi_exec.c | 2313 -------------------------------- src/mesa/pipe/tgsi/core/tgsi_exec.h | 191 --- src/mesa/pipe/tgsi/core/tgsi_parse.c | 288 ---- src/mesa/pipe/tgsi/core/tgsi_parse.h | 121 -- src/mesa/pipe/tgsi/core/tgsi_token.h | 1486 -------------------- src/mesa/pipe/tgsi/core/tgsi_util.c | 270 ---- src/mesa/pipe/tgsi/core/tgsi_util.h | 70 - src/mesa/pipe/tgsi/exec/Makefile | 3 + src/mesa/pipe/tgsi/exec/tgsi_attribs.h | 47 + src/mesa/pipe/tgsi/exec/tgsi_build.c | 1368 +++++++++++++++++++ src/mesa/pipe/tgsi/exec/tgsi_build.h | 320 +++++ src/mesa/pipe/tgsi/exec/tgsi_core.h | 12 + src/mesa/pipe/tgsi/exec/tgsi_dump.c | 1402 +++++++++++++++++++ src/mesa/pipe/tgsi/exec/tgsi_dump.h | 22 + src/mesa/pipe/tgsi/exec/tgsi_exec.c | 2313 ++++++++++++++++++++++++++++++++ src/mesa/pipe/tgsi/exec/tgsi_exec.h | 191 +++ src/mesa/pipe/tgsi/exec/tgsi_parse.c | 288 ++++ src/mesa/pipe/tgsi/exec/tgsi_parse.h | 121 ++ src/mesa/pipe/tgsi/exec/tgsi_token.h | 1486 ++++++++++++++++++++ src/mesa/pipe/tgsi/exec/tgsi_util.c | 270 ++++ src/mesa/pipe/tgsi/exec/tgsi_util.h | 70 + 28 files changed, 7913 insertions(+), 7913 deletions(-) delete mode 100644 src/mesa/pipe/tgsi/core/Makefile delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_attribs.h delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_build.c delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_build.h delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_core.h delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_dump.c delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_dump.h delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_exec.c delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_exec.h delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_parse.c delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_parse.h delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_token.h delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_util.c delete mode 100644 src/mesa/pipe/tgsi/core/tgsi_util.h create mode 100644 src/mesa/pipe/tgsi/exec/Makefile create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_attribs.h create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_build.c create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_build.h create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_core.h create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_dump.c create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_dump.h create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_exec.c create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_exec.h create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_parse.c create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_parse.h create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_token.h create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_util.c create mode 100644 src/mesa/pipe/tgsi/exec/tgsi_util.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/core/Makefile b/src/mesa/pipe/tgsi/core/Makefile deleted file mode 100644 index eb8b14e0e8..0000000000 --- a/src/mesa/pipe/tgsi/core/Makefile +++ /dev/null @@ -1,3 +0,0 @@ -default: - cd ../../.. ; make - diff --git a/src/mesa/pipe/tgsi/core/tgsi_attribs.h b/src/mesa/pipe/tgsi/core/tgsi_attribs.h deleted file mode 100644 index b8f6ba7da8..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_attribs.h +++ /dev/null @@ -1,47 +0,0 @@ -#ifndef TGSI_ATTRIBS_H -#define TGSI_ATTRIBS_H - - -/** - * The specific values here are not important (could remove them). - */ -enum { - TGSI_ATTRIB_POS = 0, - TGSI_ATTRIB_WEIGHT = 1, - TGSI_ATTRIB_NORMAL = 2, - TGSI_ATTRIB_COLOR0 = 3, - TGSI_ATTRIB_COLOR1 = 4, - TGSI_ATTRIB_FOG = 5, - TGSI_ATTRIB_COLOR_INDEX = 6, /* XXX omit? */ - TGSI_ATTRIB_EDGEFLAG = 7, - TGSI_ATTRIB_TEX0 = 8, - TGSI_ATTRIB_TEX1 = 9, - TGSI_ATTRIB_TEX2 = 10, - TGSI_ATTRIB_TEX3 = 11, - TGSI_ATTRIB_TEX4 = 12, - TGSI_ATTRIB_TEX5 = 13, - TGSI_ATTRIB_TEX6 = 14, - TGSI_ATTRIB_TEX7 = 15, - TGSI_ATTRIB_VAR0 = 16, - TGSI_ATTRIB_VAR1 = 17, - TGSI_ATTRIB_VAR2 = 18, - TGSI_ATTRIB_VAR3 = 19, - TGSI_ATTRIB_VAR4 = 20, - TGSI_ATTRIB_VAR5 = 21, - TGSI_ATTRIB_VAR6 = 22, - TGSI_ATTRIB_VAR7 = 23, - TGSI_ATTRIB_POINTSIZE = 24, - TGSI_ATTRIB_BFC0 = 25, - TGSI_ATTRIB_BFC1 = 26, - TGSI_ATTRIB_CLIP_POS = 27, - TGSI_ATTRIB_VERTEX_HEADER = 28, - TGSI_ATTRIB_MAX = 29 -}; - - -#define TGSI_MAX_TEXTURE 8 -#define TGSI_MAX_VARYING 8 - - - -#endif /* TGSI_ATTRIBS_H */ diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.c b/src/mesa/pipe/tgsi/core/tgsi_build.c deleted file mode 100644 index d6f8af656a..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_build.c +++ /dev/null @@ -1,1368 +0,0 @@ -#include "tgsi_platform.h" -#include "tgsi_core.h" - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ) -{ - struct tgsi_version version; - - version.MajorVersion = 1; - version.MinorVersion = 1; - version.Padding = 0; - - return version; -} - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ) -{ - struct tgsi_header header; - - header.HeaderSize = 1; - header.BodySize = 0; - - return header; -} - -static void -header_headersize_grow( struct tgsi_header *header ) -{ - assert( header->HeaderSize < 0xFF ); - assert( header->BodySize == 0 ); - - header->HeaderSize++; -} - -static void -header_bodysize_grow( struct tgsi_header *header ) -{ - assert( header->BodySize < 0xFFFFFF ); - - header->BodySize++; -} - -struct tgsi_processor -tgsi_default_processor( void ) -{ - struct tgsi_processor processor; - - processor.Processor = TGSI_PROCESSOR_FRAGMENT; - processor.Padding = 0; - - return processor; -} - -struct tgsi_processor -tgsi_build_processor( - unsigned type, - struct tgsi_header *header ) -{ - struct tgsi_processor processor; - - processor = tgsi_default_processor(); - processor.Processor = type; - - header_headersize_grow( header ); - - return processor; -} - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ) -{ - struct tgsi_declaration declaration; - - declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; - declaration.Size = 1; - declaration.File = TGSI_FILE_NULL; - declaration.Declare = TGSI_DECLARE_RANGE; - declaration.UsageMask = TGSI_WRITEMASK_XYZW; - declaration.Interpolate = 0; - declaration.Semantic = 0; - declaration.Padding = 0; - declaration.Extended = 0; - - return declaration; -} - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned declare, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ) -{ - struct tgsi_declaration declaration; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( declare <= TGSI_DECLARE_MASK ); - - declaration = tgsi_default_declaration(); - declaration.File = file; - declaration.Declare = declare; - declaration.UsageMask = usage_mask; - declaration.Interpolate = interpolate; - declaration.Semantic = semantic; - - header_bodysize_grow( header ); - - return declaration; -} - -static void -declaration_grow( - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - assert( declaration->Size < 0xFF ); - - declaration->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ) -{ - struct tgsi_full_declaration full_declaration; - - full_declaration.Declaration = tgsi_default_declaration(); - full_declaration.Interpolation = tgsi_default_declaration_interpolation(); - full_declaration.Semantic = tgsi_default_declaration_semantic(); - - return full_declaration; -} - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0; - struct tgsi_declaration *declaration; - - if( maxsize <= size ) - return 0; - declaration = (struct tgsi_declaration *) &tokens[size]; - size++; - - *declaration = tgsi_build_declaration( - full_decl->Declaration.File, - full_decl->Declaration.Declare, - full_decl->Declaration.UsageMask, - full_decl->Declaration.Interpolate, - full_decl->Declaration.Semantic, - header ); - - switch( full_decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - { - struct tgsi_declaration_range *dr; - - if( maxsize <= size ) - return 0; - dr = (struct tgsi_declaration_range *) &tokens[size]; - size++; - - *dr = tgsi_build_declaration_range( - full_decl->u.DeclarationRange.First, - full_decl->u.DeclarationRange.Last, - declaration, - header ); - break; - } - - case TGSI_DECLARE_MASK: - { - struct tgsi_declaration_mask *dm; - - if( maxsize <= size ) - return 0; - dm = (struct tgsi_declaration_mask *) &tokens[size]; - size++; - - *dm = tgsi_build_declaration_mask( - full_decl->u.DeclarationMask.Mask, - declaration, - header ); - break; - } - - default: - assert( 0 ); - } - - if( full_decl->Declaration.Interpolate ) { - struct tgsi_declaration_interpolation *di; - - if( maxsize <= size ) - return 0; - di = (struct tgsi_declaration_interpolation *) &tokens[size]; - size++; - - *di = tgsi_build_declaration_interpolation( - full_decl->Interpolation.Interpolate, - declaration, - header ); - } - - if( full_decl->Declaration.Semantic ) { - struct tgsi_declaration_semantic *ds; - - if( maxsize <= size ) - return 0; - ds = (struct tgsi_declaration_semantic *) &tokens[size]; - size++; - - *ds = tgsi_build_declaration_semantic( - full_decl->Semantic.SemanticName, - full_decl->Semantic.SemanticIndex, - declaration, - header ); - } - - return size; -} - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_range declaration_range; - - assert( last >= first ); - assert( last <= 0xFFFF ); - - declaration_range.First = first; - declaration_range.Last = last; - - declaration_grow( declaration, header ); - - return declaration_range; -} - -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_mask declaration_mask; - - declaration_mask.Mask = mask; - - declaration_grow( declaration, header ); - - return declaration_mask; -} - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ) -{ - struct tgsi_declaration_interpolation di; - - di.Interpolate = TGSI_INTERPOLATE_CONSTANT; - di.Padding = 0; - - return di; -} - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_interpolation di; - - assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); - - di = tgsi_default_declaration_interpolation(); - di.Interpolate = interpolate; - - declaration_grow( declaration, header ); - - return di; -} - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ) -{ - struct tgsi_declaration_semantic ds; - - ds.SemanticName = TGSI_SEMANTIC_DEPTH; - ds.SemanticIndex = 0; - ds.Padding = 0; - - return ds; -} - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_semantic ds; - - assert( semantic_name <= TGSI_SEMANTIC_COLOR ); - assert( semantic_index <= 0xFFFF ); - - ds = tgsi_default_declaration_semantic(); - ds.SemanticName = semantic_name; - ds.SemanticIndex = semantic_index; - - declaration_grow( declaration, header ); - - return ds; -} - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ) -{ - struct tgsi_immediate immediate; - - immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; - immediate.Size = 1; - immediate.DataType = TGSI_IMM_FLOAT32; - immediate.Padding = 0; - immediate.Extended = 0; - - return immediate; -} - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ) -{ - struct tgsi_immediate immediate; - - immediate = tgsi_default_immediate(); - - header_bodysize_grow( header ); - - return immediate; -} - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ) -{ - struct tgsi_full_immediate fullimm; - - fullimm.Immediate = tgsi_default_immediate(); - fullimm.u.Pointer = (void *) 0; - - return fullimm; -} - -static void -immediate_grow( - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - assert( immediate->Size < 0xFF ); - - immediate->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - GLfloat value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - struct tgsi_immediate_float32 immediate_float32; - - immediate_float32.Float = value; - - immediate_grow( immediate, header ); - - return immediate_float32; -} - -GLuint -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - GLuint maxsize ) -{ - GLuint size = 0, i; - struct tgsi_immediate *immediate; - - if( maxsize <= size ) - return 0; - immediate = (struct tgsi_immediate *) &tokens[size]; - size++; - - *immediate = tgsi_build_immediate( header ); - - for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { - struct tgsi_immediate_float32 *if32; - - if( maxsize <= size ) - return 0; - if32 = (struct tgsi_immediate_float32 *) &tokens[size]; - size++; - - *if32 = tgsi_build_immediate_float32( - full_imm->u.ImmediateFloat32[i].Float, - immediate, - header ); - } - - return size; -} - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ) -{ - struct tgsi_instruction instruction; - - instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; - instruction.Size = 1; - instruction.Opcode = TGSI_OPCODE_MOV; - instruction.Saturate = TGSI_SAT_NONE; - instruction.NumDstRegs = 1; - instruction.NumSrcRegs = 1; - instruction.Padding = 0; - instruction.Extended = 0; - - return instruction; -} - -struct tgsi_instruction -tgsi_build_instruction( - GLuint opcode, - GLuint saturate, - GLuint num_dst_regs, - GLuint num_src_regs, - struct tgsi_header *header ) -{ - struct tgsi_instruction instruction; - - assert (opcode <= TGSI_OPCODE_LAST); - assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); - assert (num_dst_regs <= 3); - assert (num_src_regs <= 15); - - instruction = tgsi_default_instruction(); - instruction.Opcode = opcode; - instruction.Saturate = saturate; - instruction.NumDstRegs = num_dst_regs; - instruction.NumSrcRegs = num_src_regs; - - header_bodysize_grow( header ); - - return instruction; -} - -static void -instruction_grow( - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - assert (instruction->Size < 0xFF); - - instruction->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ) -{ - struct tgsi_full_instruction full_instruction; - GLuint i; - - full_instruction.Instruction = tgsi_default_instruction(); - full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); - full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); - full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); - for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { - full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); - } - for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { - full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); - } - - return full_instruction; -} - -GLuint -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - GLuint maxsize ) -{ - GLuint size = 0; - GLuint i; - struct tgsi_instruction *instruction; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - instruction = (struct tgsi_instruction *) &tokens[size]; - size++; - - *instruction = tgsi_build_instruction( - full_inst->Instruction.Opcode, - full_inst->Instruction.Saturate, - full_inst->Instruction.NumDstRegs, - full_inst->Instruction.NumSrcRegs, - header ); - prev_token = (struct tgsi_token *) instruction; - - if( tgsi_compare_instruction_ext_nv( - full_inst->InstructionExtNv, - tgsi_default_instruction_ext_nv() ) ) { - struct tgsi_instruction_ext_nv *instruction_ext_nv; - - if( maxsize <= size ) - return 0; - instruction_ext_nv = - (struct tgsi_instruction_ext_nv *) &tokens[size]; - size++; - - *instruction_ext_nv = tgsi_build_instruction_ext_nv( - full_inst->InstructionExtNv.Precision, - full_inst->InstructionExtNv.CondDstIndex, - full_inst->InstructionExtNv.CondFlowIndex, - full_inst->InstructionExtNv.CondMask, - full_inst->InstructionExtNv.CondSwizzleX, - full_inst->InstructionExtNv.CondSwizzleY, - full_inst->InstructionExtNv.CondSwizzleZ, - full_inst->InstructionExtNv.CondSwizzleW, - full_inst->InstructionExtNv.CondDstUpdate, - full_inst->InstructionExtNv.CondFlowEnable, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_nv; - } - - if( tgsi_compare_instruction_ext_label( - full_inst->InstructionExtLabel, - tgsi_default_instruction_ext_label() ) ) { - struct tgsi_instruction_ext_label *instruction_ext_label; - - if( maxsize <= size ) - return 0; - instruction_ext_label = - (struct tgsi_instruction_ext_label *) &tokens[size]; - size++; - - *instruction_ext_label = tgsi_build_instruction_ext_label( - full_inst->InstructionExtLabel.Label, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_label; - } - - if( tgsi_compare_instruction_ext_texture( - full_inst->InstructionExtTexture, - tgsi_default_instruction_ext_texture() ) ) { - struct tgsi_instruction_ext_texture *instruction_ext_texture; - - if( maxsize <= size ) - return 0; - instruction_ext_texture = - (struct tgsi_instruction_ext_texture *) &tokens[size]; - size++; - - *instruction_ext_texture = tgsi_build_instruction_ext_texture( - full_inst->InstructionExtTexture.Texture, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_texture; - } - - for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { - const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; - struct tgsi_dst_register *dst_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - dst_register = (struct tgsi_dst_register *) &tokens[size]; - size++; - - *dst_register = tgsi_build_dst_register( - reg->DstRegister.File, - reg->DstRegister.WriteMask, - reg->DstRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register; - - if( tgsi_compare_dst_register_ext_concode( - reg->DstRegisterExtConcode, - tgsi_default_dst_register_ext_concode() ) ) { - struct tgsi_dst_register_ext_concode *dst_register_ext_concode; - - if( maxsize <= size ) - return 0; - dst_register_ext_concode = - (struct tgsi_dst_register_ext_concode *) &tokens[size]; - size++; - - *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( - reg->DstRegisterExtConcode.CondMask, - reg->DstRegisterExtConcode.CondSwizzleX, - reg->DstRegisterExtConcode.CondSwizzleY, - reg->DstRegisterExtConcode.CondSwizzleZ, - reg->DstRegisterExtConcode.CondSwizzleW, - reg->DstRegisterExtConcode.CondSrcIndex, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_concode; - } - - if( tgsi_compare_dst_register_ext_modulate( - reg->DstRegisterExtModulate, - tgsi_default_dst_register_ext_modulate() ) ) { - struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; - - if( maxsize <= size ) - return 0; - dst_register_ext_modulate = - (struct tgsi_dst_register_ext_modulate *) &tokens[size]; - size++; - - *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( - reg->DstRegisterExtModulate.Modulate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_modulate; - } - } - - for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { - const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; - struct tgsi_src_register *src_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - src_register = (struct tgsi_src_register *) &tokens[size]; - size++; - - *src_register = tgsi_build_src_register( - reg->SrcRegister.File, - reg->SrcRegister.SwizzleX, - reg->SrcRegister.SwizzleY, - reg->SrcRegister.SwizzleZ, - reg->SrcRegister.SwizzleW, - reg->SrcRegister.Negate, - reg->SrcRegister.Indirect, - reg->SrcRegister.Dimension, - reg->SrcRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register; - - if( tgsi_compare_src_register_ext_swz( - reg->SrcRegisterExtSwz, - tgsi_default_src_register_ext_swz() ) ) { - struct tgsi_src_register_ext_swz *src_register_ext_swz; - - if( maxsize <= size ) - return 0; - src_register_ext_swz = - (struct tgsi_src_register_ext_swz *) &tokens[size]; - size++; - - *src_register_ext_swz = tgsi_build_src_register_ext_swz( - reg->SrcRegisterExtSwz.ExtSwizzleX, - reg->SrcRegisterExtSwz.ExtSwizzleY, - reg->SrcRegisterExtSwz.ExtSwizzleZ, - reg->SrcRegisterExtSwz.ExtSwizzleW, - reg->SrcRegisterExtSwz.NegateX, - reg->SrcRegisterExtSwz.NegateY, - reg->SrcRegisterExtSwz.NegateZ, - reg->SrcRegisterExtSwz.NegateW, - reg->SrcRegisterExtSwz.ExtDivide, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_swz; - } - - if( tgsi_compare_src_register_ext_mod( - reg->SrcRegisterExtMod, - tgsi_default_src_register_ext_mod() ) ) { - struct tgsi_src_register_ext_mod *src_register_ext_mod; - - if( maxsize <= size ) - return 0; - src_register_ext_mod = - (struct tgsi_src_register_ext_mod *) &tokens[size]; - size++; - - *src_register_ext_mod = tgsi_build_src_register_ext_mod( - reg->SrcRegisterExtMod.Complement, - reg->SrcRegisterExtMod.Bias, - reg->SrcRegisterExtMod.Scale2X, - reg->SrcRegisterExtMod.Absolute, - reg->SrcRegisterExtMod.Negate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_mod; - } - - if( reg->SrcRegister.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterInd.File, - reg->SrcRegisterInd.SwizzleX, - reg->SrcRegisterInd.SwizzleY, - reg->SrcRegisterInd.SwizzleZ, - reg->SrcRegisterInd.SwizzleW, - reg->SrcRegisterInd.Negate, - reg->SrcRegisterInd.Indirect, - reg->SrcRegisterInd.Dimension, - reg->SrcRegisterInd.Index, - instruction, - header ); - } - - if( reg->SrcRegister.Dimension ) { - struct tgsi_dimension *dim; - - assert( !reg->SrcRegisterDim.Dimension ); - - if( maxsize <= size ) - return 0; - dim = (struct tgsi_dimension *) &tokens[size]; - size++; - - *dim = tgsi_build_dimension( - reg->SrcRegisterDim.Indirect, - reg->SrcRegisterDim.Index, - instruction, - header ); - - if( reg->SrcRegisterDim.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterDimInd.File, - reg->SrcRegisterDimInd.SwizzleX, - reg->SrcRegisterDimInd.SwizzleY, - reg->SrcRegisterDimInd.SwizzleZ, - reg->SrcRegisterDimInd.SwizzleW, - reg->SrcRegisterDimInd.Negate, - reg->SrcRegisterDimInd.Indirect, - reg->SrcRegisterDimInd.Dimension, - reg->SrcRegisterDimInd.Index, - instruction, - header ); - } - } - } - - return size; -} - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; - instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; - instruction_ext_nv.CondDstIndex = 0; - instruction_ext_nv.CondFlowIndex = 0; - instruction_ext_nv.CondMask = TGSI_CC_TR; - instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; - instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; - instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; - instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; - instruction_ext_nv.CondDstUpdate = 0; - instruction_ext_nv.CondFlowEnable = 0; - instruction_ext_nv.Padding = 0; - instruction_ext_nv.Extended = 0; - - return instruction_ext_nv; -} - -union token_u32 -{ - GLuint u32; -}; - -GLuint -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - GLuint precision, - GLuint cond_dst_index, - GLuint cond_flow_index, - GLuint cond_mask, - GLuint cond_swizzle_x, - GLuint cond_swizzle_y, - GLuint cond_swizzle_z, - GLuint cond_swizzle_w, - GLuint cond_dst_update, - GLuint cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv = tgsi_default_instruction_ext_nv(); - instruction_ext_nv.Precision = precision; - instruction_ext_nv.CondDstIndex = cond_dst_index; - instruction_ext_nv.CondFlowIndex = cond_flow_index; - instruction_ext_nv.CondMask = cond_mask; - instruction_ext_nv.CondSwizzleX = cond_swizzle_x; - instruction_ext_nv.CondSwizzleY = cond_swizzle_y; - instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; - instruction_ext_nv.CondSwizzleW = cond_swizzle_w; - instruction_ext_nv.CondDstUpdate = cond_dst_update; - instruction_ext_nv.CondFlowEnable = cond_flow_update; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_nv; -} - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; - instruction_ext_label.Label = 0; - instruction_ext_label.Padding = 0; - instruction_ext_label.Extended = 0; - - return instruction_ext_label; -} - -GLuint -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; -} - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - GLuint label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label = tgsi_default_instruction_ext_label(); - instruction_ext_label.Label = label; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_label; -} - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; - instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; - instruction_ext_texture.Padding = 0; - instruction_ext_texture.Extended = 0; - - return instruction_ext_texture; -} - -GLuint -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; -} - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - GLuint texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture = tgsi_default_instruction_ext_texture(); - instruction_ext_texture.Texture = texture; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_texture; -} - -struct tgsi_src_register -tgsi_default_src_register( void ) -{ - struct tgsi_src_register src_register; - - src_register.File = TGSI_FILE_NULL; - src_register.SwizzleX = TGSI_SWIZZLE_X; - src_register.SwizzleY = TGSI_SWIZZLE_Y; - src_register.SwizzleZ = TGSI_SWIZZLE_Z; - src_register.SwizzleW = TGSI_SWIZZLE_W; - src_register.Negate = 0; - src_register.Indirect = 0; - src_register.Dimension = 0; - src_register.Index = 0; - src_register.Extended = 0; - - return src_register; -} - -struct tgsi_src_register -tgsi_build_src_register( - GLuint file, - GLuint swizzle_x, - GLuint swizzle_y, - GLuint swizzle_z, - GLuint swizzle_w, - GLuint negate, - GLuint indirect, - GLuint dimension, - GLint index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register src_register; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( swizzle_x <= TGSI_SWIZZLE_W ); - assert( swizzle_y <= TGSI_SWIZZLE_W ); - assert( swizzle_z <= TGSI_SWIZZLE_W ); - assert( swizzle_w <= TGSI_SWIZZLE_W ); - assert( negate <= 1 ); - assert( index >= -0x8000 && index <= 0x7FFF ); - - src_register = tgsi_default_src_register(); - src_register.File = file; - src_register.SwizzleX = swizzle_x; - src_register.SwizzleY = swizzle_y; - src_register.SwizzleZ = swizzle_z; - src_register.SwizzleW = swizzle_w; - src_register.Negate = negate; - src_register.Indirect = indirect; - src_register.Dimension = dimension; - src_register.Index = index; - - instruction_grow( instruction, header ); - - return src_register; -} - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ) -{ - struct tgsi_full_src_register full_src_register; - - full_src_register.SrcRegister = tgsi_default_src_register(); - full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); - full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); - full_src_register.SrcRegisterInd = tgsi_default_src_register(); - full_src_register.SrcRegisterDim = tgsi_default_dimension(); - full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); - - return full_src_register; -} - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; - src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; - src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; - src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; - src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; - src_register_ext_swz.NegateX = 0; - src_register_ext_swz.NegateY = 0; - src_register_ext_swz.NegateZ = 0; - src_register_ext_swz.NegateW = 0; - src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; - src_register_ext_swz.Padding = 0; - src_register_ext_swz.Extended = 0; - - return src_register_ext_swz; -} - -GLuint -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; -} - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - GLuint ext_swizzle_x, - GLuint ext_swizzle_y, - GLuint ext_swizzle_z, - GLuint ext_swizzle_w, - GLuint negate_x, - GLuint negate_y, - GLuint negate_z, - GLuint negate_w, - GLuint ext_divide, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - assert (ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE); - assert (ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE); - assert (ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE); - assert (ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE); - assert (negate_x <= 1); - assert (negate_y <= 1); - assert (negate_z <= 1); - assert (negate_w <= 1); - assert (ext_divide <= TGSI_EXTSWIZZLE_ONE); - - src_register_ext_swz = tgsi_default_src_register_ext_swz(); - src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; - src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; - src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; - src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; - src_register_ext_swz.NegateX = negate_x; - src_register_ext_swz.NegateY = negate_y; - src_register_ext_swz.NegateZ = negate_z; - src_register_ext_swz.NegateW = negate_w; - src_register_ext_swz.ExtDivide = ext_divide; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_swz; -} - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; - src_register_ext_mod.Complement = 0; - src_register_ext_mod.Bias = 0; - src_register_ext_mod.Scale2X = 0; - src_register_ext_mod.Absolute = 0; - src_register_ext_mod.Negate = 0; - src_register_ext_mod.Padding = 0; - src_register_ext_mod.Extended = 0; - - return src_register_ext_mod; -} - -GLuint -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; -} - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - GLuint complement, - GLuint bias, - GLuint scale_2x, - GLuint absolute, - GLuint negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - assert (complement <= 1); - assert (bias <= 1); - assert (scale_2x <= 1); - assert (absolute <= 1); - assert (negate <= 1); - - src_register_ext_mod = tgsi_default_src_register_ext_mod(); - src_register_ext_mod.Complement = complement; - src_register_ext_mod.Bias = bias; - src_register_ext_mod.Scale2X = scale_2x; - src_register_ext_mod.Absolute = absolute; - src_register_ext_mod.Negate = negate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_mod; -} - -struct tgsi_dimension -tgsi_default_dimension( void ) -{ - struct tgsi_dimension dimension; - - dimension.Indirect = 0; - dimension.Dimension = 0; - dimension.Padding = 0; - dimension.Index = 0; - dimension.Extended = 0; - - return dimension; -} - -struct tgsi_dimension -tgsi_build_dimension( - GLuint indirect, - GLuint index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dimension dimension; - - dimension = tgsi_default_dimension(); - dimension.Indirect = indirect; - dimension.Index = index; - - instruction_grow( instruction, header ); - - return dimension; -} - -struct tgsi_dst_register -tgsi_default_dst_register( void ) -{ - struct tgsi_dst_register dst_register; - - dst_register.File = TGSI_FILE_NULL; - dst_register.WriteMask = TGSI_WRITEMASK_XYZW; - dst_register.Indirect = 0; - dst_register.Dimension = 0; - dst_register.Index = 0; - dst_register.Padding = 0; - dst_register.Extended = 0; - - return dst_register; -} - -struct tgsi_dst_register -tgsi_build_dst_register( - GLuint file, - GLuint mask, - GLint index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register dst_register; - - assert (file <= TGSI_FILE_IMMEDIATE); - assert (mask <= TGSI_WRITEMASK_XYZW); - assert (index >= -32768 && index <= 32767); - - dst_register = tgsi_default_dst_register(); - dst_register.File = file; - dst_register.WriteMask = mask; - dst_register.Index = index; - - instruction_grow( instruction, header ); - - return dst_register; -} - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ) -{ - struct tgsi_full_dst_register full_dst_register; - - full_dst_register.DstRegister = tgsi_default_dst_register(); - full_dst_register.DstRegisterExtConcode = - tgsi_default_dst_register_ext_concode(); - full_dst_register.DstRegisterExtModulate = - tgsi_default_dst_register_ext_modulate(); - - return full_dst_register; -} - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; - dst_register_ext_concode.CondMask = TGSI_CC_TR; - dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; - dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; - dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; - dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; - dst_register_ext_concode.CondSrcIndex = 0; - dst_register_ext_concode.Padding = 0; - dst_register_ext_concode.Extended = 0; - - return dst_register_ext_concode; -} - -GLuint -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; -} - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - GLuint cc, - GLuint swizzle_x, - GLuint swizzle_y, - GLuint swizzle_z, - GLuint swizzle_w, - GLint index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - assert (cc <= TGSI_CC_FL); - assert (swizzle_x <= TGSI_SWIZZLE_W); - assert (swizzle_y <= TGSI_SWIZZLE_W); - assert (swizzle_z <= TGSI_SWIZZLE_W); - assert (swizzle_w <= TGSI_SWIZZLE_W); - assert (index >= -32768 && index <= 32767); - - dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); - dst_register_ext_concode.CondMask = cc; - dst_register_ext_concode.CondSwizzleX = swizzle_x; - dst_register_ext_concode.CondSwizzleY = swizzle_y; - dst_register_ext_concode.CondSwizzleZ = swizzle_z; - dst_register_ext_concode.CondSwizzleW = swizzle_w; - dst_register_ext_concode.CondSrcIndex = index; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_concode; -} - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; - dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; - dst_register_ext_modulate.Padding = 0; - dst_register_ext_modulate.Extended = 0; - - return dst_register_ext_modulate; -} - -GLuint -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; -} - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - GLuint modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - assert (modulate <= TGSI_MODULATE_EIGHTH); - - dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); - dst_register_ext_modulate.Modulate = modulate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_modulate; -} - diff --git a/src/mesa/pipe/tgsi/core/tgsi_build.h b/src/mesa/pipe/tgsi/core/tgsi_build.h deleted file mode 100644 index 116c78abf3..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_build.h +++ /dev/null @@ -1,320 +0,0 @@ -#if !defined TGSI_BUILD_H -#define TGSI_BUILD_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ); - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ); - -struct tgsi_processor -tgsi_default_processor( void ); - -struct tgsi_processor -tgsi_build_processor( - unsigned processor, - struct tgsi_header *header ); - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ); - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned declare, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ); - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ); - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ); - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ); - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ); - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ); - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ); - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - float value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ); - -unsigned -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ); - -struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ); - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ); - -unsigned -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ); - -unsigned -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ); - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - unsigned precision, - unsigned cond_dst_index, - unsigned cond_flow_index, - unsigned cond_mask, - unsigned cond_swizzle_x, - unsigned cond_swizzle_y, - unsigned cond_swizzle_z, - unsigned cond_swizzle_w, - unsigned cond_dst_update, - unsigned cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ); - -unsigned -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ); - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - unsigned label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ); - -unsigned -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ); - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - unsigned texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register -tgsi_default_src_register( void ); - -struct tgsi_src_register -tgsi_build_src_register( - unsigned file, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - unsigned negate, - unsigned indirect, - unsigned dimension, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ); - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ); - -unsigned -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ); - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - unsigned ext_swizzle_x, - unsigned ext_swizzle_y, - unsigned ext_swizzle_z, - unsigned ext_swizzle_w, - unsigned negate_x, - unsigned negate_y, - unsigned negate_z, - unsigned negate_w, - unsigned ext_divide, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ); - -unsigned -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ); - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - unsigned complement, - unsigned bias, - unsigned scale_2x, - unsigned absolute, - unsigned negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dimension -tgsi_default_dimension( void ); - -struct tgsi_dimension -tgsi_build_dimension( - unsigned indirect, - unsigned index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register -tgsi_default_dst_register( void ); - -struct tgsi_dst_register -tgsi_build_dst_register( - unsigned file, - unsigned mask, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ); - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ); - -unsigned -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ); - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - unsigned cc, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - int index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ); - -unsigned -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ); - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - unsigned modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_BUILD_H - diff --git a/src/mesa/pipe/tgsi/core/tgsi_core.h b/src/mesa/pipe/tgsi/core/tgsi_core.h deleted file mode 100644 index 1f5f00a38e..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_core.h +++ /dev/null @@ -1,12 +0,0 @@ -#if !defined TGSI_CORE_H -#define TGSI_CORE_H - -#include "tgsi_token.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" -#include "tgsi_exec.h" -#include "tgsi_dump.h" -#include "tgsi_util.h" - -#endif // !defined TGSI_CORE_H - diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.c b/src/mesa/pipe/tgsi/core/tgsi_dump.c deleted file mode 100644 index 622d617e29..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.c +++ /dev/null @@ -1,1402 +0,0 @@ -#include "tgsi_platform.h" -#include "tgsi_core.h" - -struct text_dump -{ - FILE *file; - GLuint tabs; -}; - -static void -text_dump_write( - struct text_dump *dump, - const void *buffer, - GLuint size ) -{ - fwrite( buffer, size, 1, dump->file ); -} - -static void -text_dump_str( - struct text_dump *dump, - const char *str ) -{ - GLuint i; - GLuint len = strlen( str ); - - for( i = 0; i < len; i++ ) { - text_dump_write( dump, &str[i], 1 ); - - if( str[i] == '\n' ) { - GLuint i; - - for( i = 0; i < dump->tabs; i++ ) { - text_dump_write( dump, " ", 4 ); - } - } - } -} - -static void -text_dump_chr( - struct text_dump *dump, - const char chr ) -{ - char str[2]; - - str[0] = chr; - str[1] = '\0'; - text_dump_str( dump, str ); -} - -static void -text_dump_uix( - struct text_dump *dump, - const GLuint ui) -{ - char str[36]; - - sprintf( str, "0x%x", ui ); - text_dump_str( dump, str ); -} - -static void -text_dump_uid( - struct text_dump *dump, - const GLuint ui ) -{ - char str[16]; - - sprintf( str, "%u", ui ); - text_dump_str( dump, str ); -} - -static void -text_dump_sid( - struct text_dump *dump, - const GLint si ) -{ - char str[16]; - - sprintf( str, "%d", si ); - text_dump_str( dump, str ); -} - -static void -text_dump_flt( - struct text_dump *dump, - const GLfloat f ) -{ - char str[48]; - - sprintf( str, "%40.6f", f ); - text_dump_str( dump, str ); -} - -static void -text_dump_enum( - struct text_dump *dump, - const GLuint e, - const char **enums, - const GLuint enums_count ) -{ - if( e >= enums_count ) { - text_dump_uid( dump, e ); - } - else { - text_dump_str( dump, enums[e] ); - } -} - -static void -text_dump_tab( - struct text_dump *dump ) -{ - dump->tabs++; -} - -static void -text_dump_untab( - struct text_dump *dump ) -{ - assert( dump->tabs > 0 ); - - --dump->tabs; -} - -#define TXT(S) text_dump_str( dump, S ) -#define CHR(C) text_dump_chr( dump, C ) -#define UIX(I) text_dump_uix( dump, I ) -#define UID(I) text_dump_uid( dump, I ) -#define SID(I) text_dump_sid( dump, I ) -#define FLT(F) text_dump_flt( dump, F ) -#define TAB() text_dump_tab( dump ) -#define UNT() text_dump_untab( dump ) -#define ENM(E,ENUMS) text_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) - -static const char *TGSI_PROCESSOR_TYPES[] = -{ - "PROCESSOR_FRAGMENT", - "PROCESSOR_VERTEX", - "PROCESSOR_GEOMETRY" -}; - -static const char *TGSI_PROCESSOR_TYPES_SHORT[] = -{ - "FRAG", - "VERT", - "GEOM" -}; - -static const char *TGSI_TOKEN_TYPES[] = -{ - "TOKEN_TYPE_DECLARATION", - "TOKEN_TYPE_IMMEDIATE", - "TOKEN_TYPE_INSTRUCTION" -}; - -static const char *TGSI_FILES[] = -{ - "FILE_NULL", - "FILE_CONSTANT", - "FILE_INPUT", - "FILE_OUTPUT", - "FILE_TEMPORARY", - "FILE_SAMPLER", - "FILE_ADDRESS", - "FILE_IMMEDIATE" -}; - -static const char *TGSI_FILES_SHORT[] = -{ - "NULL", - "CONST", - "IN", - "OUT", - "TEMP", - "SAMP", - "ADDR", - "IMM" -}; - -static const char *TGSI_DECLARES[] = -{ - "DECLARE_RANGE", - "DECLARE_MASK" -}; - -static const char *TGSI_INTERPOLATES[] = -{ - "INTERPOLATE_CONSTANT", - "INTERPOLATE_LINEAR", - "INTERPOLATE_PERSPECTIVE" -}; - -static const char *TGSI_INTERPOLATES_SHORT[] = -{ - "CONSTANT", - "LINEAR", - "PERSPECTIVE" -}; - -static const char *TGSI_SEMANTICS[] = -{ - "SEMANTIC_DEPTH", - "SEMANTIC_COLOR" -}; - -static const char *TGSI_SEMANTICS_SHORT[] = -{ - "DEPTH", - "COLOR" -}; - -static const char *TGSI_IMMS[] = -{ - "IMM_FLOAT32" -}; - -static const char *TGSI_IMMS_SHORT[] = -{ - "FLT32" -}; - -static const char *TGSI_OPCODES[] = -{ - "OPCODE_ARL", - "OPCODE_MOV", - "OPCODE_LIT", - "OPCODE_RCP", - "OPCODE_RSQ", - "OPCODE_EXP", - "OPCODE_LOG", - "OPCODE_MUL", - "OPCODE_ADD", - "OPCODE_DP3", - "OPCODE_DP4", - "OPCODE_DST", - "OPCODE_MIN", - "OPCODE_MAX", - "OPCODE_SLT", - "OPCODE_SGE", - "OPCODE_MAD", - "OPCODE_SUB", - "OPCODE_LERP", - "OPCODE_CND", - "OPCODE_CND0", - "OPCODE_DOT2ADD", - "OPCODE_INDEX", - "OPCODE_NEGATE", - "OPCODE_FRAC", - "OPCODE_CLAMP", - "OPCODE_FLOOR", - "OPCODE_ROUND", - "OPCODE_EXPBASE2", - "OPCODE_LOGBASE2", - "OPCODE_POWER", - "OPCODE_CROSSPRODUCT", - "OPCODE_MULTIPLYMATRIX", - "OPCODE_ABS", - "OPCODE_RCC", - "OPCODE_DPH", - "OPCODE_COS", - "OPCODE_DDX", - "OPCODE_DDY", - "OPCODE_KIL", - "OPCODE_PK2H", - "OPCODE_PK2US", - "OPCODE_PK4B", - "OPCODE_PK4UB", - "OPCODE_RFL", - "OPCODE_SEQ", - "OPCODE_SFL", - "OPCODE_SGT", - "OPCODE_SIN", - "OPCODE_SLE", - "OPCODE_SNE", - "OPCODE_STR", - "OPCODE_TEX", - "OPCODE_TXD", - "OPCODE_UP2H", - "OPCODE_UP2US", - "OPCODE_UP4B", - "OPCODE_UP4UB", - "OPCODE_X2D", - "OPCODE_ARA", - "OPCODE_ARR", - "OPCODE_BRA", - "OPCODE_CAL", - "OPCODE_RET", - "OPCODE_SSG", - "OPCODE_CMP", - "OPCODE_SCS", - "OPCODE_TXB", - "OPCODE_NRM", - "OPCODE_DIV", - "OPCODE_DP2", - "OPCODE_TXL", - "OPCODE_BRK", - "OPCODE_IF", - "OPCODE_LOOP", - "OPCODE_REP", - "OPCODE_ELSE", - "OPCODE_ENDIF", - "OPCODE_ENDLOOP", - "OPCODE_ENDREP", - "OPCODE_PUSHA", - "OPCODE_POPA", - "OPCODE_CEIL", - "OPCODE_I2F", - "OPCODE_NOT", - "OPCODE_TRUNC", - "OPCODE_SHL", - "OPCODE_SHR", - "OPCODE_AND", - "OPCODE_OR", - "OPCODE_MOD", - "OPCODE_XOR", - "OPCODE_SAD", - "OPCODE_TXF", - "OPCODE_TXQ", - "OPCODE_CONT", - "OPCODE_EMIT", - "OPCODE_ENDPRIM", - "OPCODE_BGNLOOP2", - "OPCODE_BGNSUB", - "OPCODE_ENDLOOP2", - "OPCODE_ENDSUB", - "OPCODE_NOISE1", - "OPCODE_NOISE2", - "OPCODE_NOISE3", - "OPCODE_NOISE4", - "OPCODE_NOP", - "OPCODE_TEXBEM", - "OPCODE_TEXBEML", - "OPCODE_TEXREG2AR", - "OPCODE_TEXM3X2PAD", - "OPCODE_TEXM3X2TEX", - "OPCODE_TEXM3X3PAD", - "OPCODE_TEXM3X3TEX", - "OPCODE_TEXM3X3SPEC", - "OPCODE_TEXM3X3VSPEC", - "OPCODE_TEXREG2GB", - "OPCODE_TEXREG2RGB", - "OPCODE_TEXDP3TEX", - "OPCODE_TEXDP3", - "OPCODE_TEXM3X3", - "OPCODE_TEXM3X2DEPTH", - "OPCODE_TEXDEPTH", - "OPCODE_BEM", - "OPCODE_M4X3", - "OPCODE_M3X4", - "OPCODE_M3X3", - "OPCODE_M3X2", - "OPCODE_NRM4", - "OPCODE_CALLNZ", - "OPCODE_IFC", - "OPCODE_BREAKC", - "OPCODE_TXP" -}; - -static const char *TGSI_OPCODES_SHORT[] = -{ - "ARL", - "MOV", - "LIT", - "RCP", - "RSQ", - "EXP", - "LOG", - "MUL", - "ADD", - "DP3", - "DP4", - "DST", - "MIN", - "MAX", - "SLT", - "SGE", - "MAD", - "SUB", - "LERP", - "CND", - "CND0", - "DOT2ADD", - "INDEX", - "NEGATE", - "FRAC", - "CLAMP", - "FLOOR", - "ROUND", - "EXPBASE2", - "LOGBASE2", - "POWER", - "CROSSPRODUCT", - "MULTIPLYMATRIX", - "ABS", - "RCC", - "DPH", - "COS", - "DDX", - "DDY", - "KIL", - "PK2H", - "PK2US", - "PK4B", - "PK4UB", - "RFL", - "SEQ", - "SFL", - "SGT", - "SIN", - "SLE", - "SNE", - "STR", - "TEX", - "TXD", - "UP2H", - "UP2US", - "UP4B", - "UP4UB", - "X2D", - "ARA", - "ARR", - "BRA", - "CAL", - "RET", - "SSG", - "CMP", - "SCS", - "TXB", - "NRM", - "DIV", - "DP2", - "TXL", - "BRK", - "IF", - "LOOP", - "REP", - "ELSE", - "ENDIF", - "ENDLOOP", - "ENDREP", - "PUSHA", - "POPA", - "CEIL", - "I2F", - "NOT", - "TRUNC", - "SHL", - "SHR", - "AND", - "OR", - "MOD", - "XOR", - "SAD", - "TXF", - "TXQ", - "CONT", - "EMIT", - "ENDPRIM", - "BGNLOOP2", - "BGNSUB", - "ENDLOOP2", - "ENDSUB", - "NOISE1", - "NOISE2", - "NOISE3", - "NOISE4", - "NOP", - "TEXBEM", - "TEXBEML", - "TEXREG2AR", - "TEXM3X2PAD", - "TEXM3X2TEX", - "TEXM3X3PAD", - "TEXM3X3TEX", - "TEXM3X3SPEC", - "TEXM3X3VSPEC", - "TEXREG2GB", - "TEXREG2RGB", - "TEXDP3TEX", - "TEXDP3", - "TEXM3X3", - "TEXM3X2DEPTH", - "TEXDEPTH", - "BEM", - "M4X3", - "M3X4", - "M3X3", - "M3X2", - "NRM4", - "CALLNZ", - "IFC", - "BREAKC", -}; - -static const char *TGSI_SATS[] = -{ - "SAT_NONE", - "SAT_ZERO_ONE", - "SAT_MINUS_PLUS_ONE" -}; - -static const char *TGSI_INSTRUCTION_EXTS[] = -{ - "INSTRUCTION_EXT_TYPE_NV", - "INSTRUCTION_EXT_TYPE_LABEL", - "INSTRUCTION_EXT_TYPE_TEXTURE" -}; - -static const char *TGSI_PRECISIONS[] = -{ - "PRECISION_DEFAULT", - "TGSI_PRECISION_FLOAT32", - "TGSI_PRECISION_FLOAT16", - "TGSI_PRECISION_FIXED12" -}; - -static const char *TGSI_CCS[] = -{ - "CC_GT", - "CC_EQ", - "CC_LT", - "CC_UN", - "CC_GE", - "CC_LE", - "CC_NE", - "CC_TR", - "CC_FL" -}; - -static const char *TGSI_SWIZZLES[] = -{ - "SWIZZLE_X", - "SWIZZLE_Y", - "SWIZZLE_Z", - "SWIZZLE_W" -}; - -static const char *TGSI_SWIZZLES_SHORT[] = -{ - "x", - "y", - "z", - "w" -}; - -static const char *TGSI_TEXTURES[] = -{ - "TEXTURE_UNKNOWN", - "TEXTURE_1D", - "TEXTURE_2D", - "TEXTURE_3D", - "TEXTURE_CUBE", - "TEXTURE_RECT", - "TEXTURE_SHADOW1D", - "TEXTURE_SHADOW2D", - "TEXTURE_SHADOWRECT" -}; - -static const char *TGSI_SRC_REGISTER_EXTS[] = -{ - "SRC_REGISTER_EXT_TYPE_SWZ", - "SRC_REGISTER_EXT_TYPE_MOD" -}; - -static const char *TGSI_EXTSWIZZLES[] = -{ - "EXTSWIZZLE_X", - "EXTSWIZZLE_Y", - "EXTSWIZZLE_Z", - "EXTSWIZZLE_W", - "EXTSWIZZLE_ZERO", - "EXTSWIZZLE_ONE" -}; - -static const char *TGSI_WRITEMASKS[] = -{ - "0", - "WRITEMASK_X", - "WRITEMASK_Y", - "WRITEMASK_XY", - "WRITEMASK_Z", - "WRITEMASK_XZ", - "WRITEMASK_YZ", - "WRITEMASK_XYZ", - "WRITEMASK_W", - "WRITEMASK_XW", - "WRITEMASK_YW", - "WRITEMASK_XYW", - "WRITEMASK_ZW", - "WRITEMASK_XZW", - "WRITEMASK_YZW", - "WRITEMASK_XYZW" -}; - -static const char *TGSI_DST_REGISTER_EXTS[] = -{ - "DST_REGISTER_EXT_TYPE_CONDCODE", - "DST_REGISTER_EXT_TYPE_MODULATE" -}; - -static const char *TGSI_MODULATES[] = -{ - "MODULATE_1X", - "MODULATE_2X", - "MODULATE_4X", - "MODULATE_8X", - "MODULATE_HALF", - "MODULATE_QUARTER", - "MODULATE_EIGHTH" -}; - -static void -dump_declaration_short( - struct text_dump *dump, - struct tgsi_full_declaration *decl ) -{ - TXT( "\nDCL " ); - ENM( decl->Declaration.File, TGSI_FILES_SHORT ); - - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - CHR( '[' ); - UID( decl->u.DeclarationRange.First ); - if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { - TXT( ".." ); - UID( decl->u.DeclarationRange.Last ); - } - CHR( ']' ); - break; - default: - assert( 0 ); - } - - if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { - CHR( '.' ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'x' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'w' ); - } - } - - if( decl->Declaration.Interpolate ) { - TXT( ", " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); - } - - if( decl->Declaration.Semantic ) { - TXT( ", " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); - CHR( '[' ); - UID( decl->Semantic.SemanticIndex ); - CHR( ']' ); - } -} - -static void -dump_declaration_verbose( - struct text_dump *dump, - struct tgsi_full_declaration *decl, - GLuint ignored, - GLuint deflt, - struct tgsi_full_declaration *fd ) -{ - TXT( "\nFile : " ); - ENM( decl->Declaration.File, TGSI_FILES ); - TXT( "\nDeclare : " ); - ENM( decl->Declaration.Declare, TGSI_DECLARES ); - if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { - TXT( "\nUsageMask : " ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'X' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'Y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'Z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'W' ); - } - } - if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { - TXT( "\nInterpolate: " ); - UID( decl->Declaration.Interpolate ); - } - if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { - TXT( "\nSemantic : " ); - UID( decl->Declaration.Semantic ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Declaration.Padding ); - } - - CHR( '\n' ); - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - TXT( "\nFirst: " ); - UID( decl->u.DeclarationRange.First ); - TXT( "\nLast : " ); - UID( decl->u.DeclarationRange.Last ); - break; - - case TGSI_DECLARE_MASK: - TXT( "\nMask: " ); - UIX( decl->u.DeclarationMask.Mask ); - break; - - default: - assert( 0 ); - } - - if( decl->Declaration.Interpolate ) { - CHR( '\n' ); - TXT( "\nInterpolate: " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Interpolation.Padding ); - } - } - - if( decl->Declaration.Semantic ) { - CHR( '\n' ); - TXT( "\nSemanticName : " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); - TXT( "\nSemanticIndex: " ); - UID( decl->Semantic.SemanticIndex ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Semantic.Padding ); - } - } -} - -static void -dump_immediate_short( - struct text_dump *dump, - struct tgsi_full_immediate *imm ) -{ - GLuint i; - - TXT( "\nIMM " ); - ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); - - TXT( " { " ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - - if( i < imm->Immediate.Size - 2 ) { - TXT( ", " ); - } - } - TXT( " }" ); -} - -static void -dump_immediate_verbose( - struct text_dump *dump, - struct tgsi_full_immediate *imm, - GLuint ignored ) -{ - GLuint i; - - TXT( "\nDataType : " ); - ENM( imm->Immediate.DataType, TGSI_IMMS ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( imm->Immediate.Padding ); - } - - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - CHR( '\n' ); - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - TXT( "\nFloat: " ); - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - } -} - -static void -dump_instruction_short( - struct text_dump *dump, - struct tgsi_full_instruction *inst, - GLuint instno ) -{ - GLuint i; - GLboolean first_reg = GL_TRUE; - - CHR( '\n' ); - UID( instno ); - CHR( ':' ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); - - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - TXT( "_SAT" ); - break; - case TGSI_SAT_MINUS_PLUS_ONE: - TXT( "_SAT[-1,1]" ); - break; - default: - assert( 0 ); - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if( !first_reg ) { - CHR( ',' ); - } - CHR( ' ' ); - - ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); - SID( dst->DstRegister.Index ); - CHR( ']' ); - - if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { - CHR( '.' ); - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { - CHR( 'x' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { - CHR( 'y' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { - CHR( 'z' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { - CHR( 'w' ); - } - } - - first_reg = GL_FALSE; - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - - if( !first_reg ) { - CHR( ',' ); - } - CHR( ' ' ); - - if( src->SrcRegisterExtMod.Negate ) { - CHR( '-' ); - } - if( src->SrcRegisterExtMod.Absolute ) { - CHR( '|' ); - } - if( src->SrcRegister.Negate ) { - CHR( '-' ); - } - - ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); - SID( src->SrcRegister.Index ); - CHR( ']' ); - - if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || - src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || - src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || - src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { - CHR( '.' ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); - } - - if( src->SrcRegisterExtMod.Absolute ) { - CHR( '|' ); - } - - first_reg = GL_FALSE; - } - - switch( inst->Instruction.Opcode ) { - case TGSI_OPCODE_IF: - case TGSI_OPCODE_ELSE: - TXT( " :" ); - UID( inst->InstructionExtLabel.Label ); - break; - } -} - -static void -dump_instruction_verbose( - struct text_dump *dump, - struct tgsi_full_instruction *inst, - GLuint ignored, - GLuint deflt, - struct tgsi_full_instruction *fi ) -{ - GLuint i; - - TXT( "\nOpcode : " ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES ); - if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { - TXT( "\nSaturate : " ); - ENM( inst->Instruction.Saturate, TGSI_SATS ); - } - if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { - TXT( "\nNumDstRegs : " ); - UID( inst->Instruction.NumDstRegs ); - } - if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { - TXT( "\nNumSrcRegs : " ); - UID( inst->Instruction.NumSrcRegs ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->Instruction.Padding ); - } - - if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { - TXT( "\nPrecision : " ); - ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); - } - if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { - TXT( "\nCondDstIndex : " ); - UID( inst->InstructionExtNv.CondDstIndex ); - } - if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { - TXT( "\nCondFlowIndex : " ); - UID( inst->InstructionExtNv.CondFlowIndex ); - } - if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { - TXT( "\nCondMask : " ); - ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { - TXT( "\nCondSwizzleX : " ); - ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { - TXT( "\nCondSwizzleY : " ); - ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ : " ); - ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { - TXT( "\nCondSwizzleW : " ); - ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { - TXT( "\nCondDstUpdate : " ); - UID( inst->InstructionExtNv.CondDstUpdate ); - } - if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { - TXT( "\nCondFlowEnable: " ); - UID( inst->InstructionExtNv.CondFlowEnable ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtNv.Padding ); - if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { - TXT( "\nExtended : " ); - UID( inst->InstructionExtNv.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { - TXT( "\nLabel : " ); - UID( inst->InstructionExtLabel.Label ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtLabel.Padding ); - if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtLabel.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { - TXT( "\nTexture : " ); - ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtTexture.Padding ); - if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtTexture.Extended ); - } - } - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : " ); - ENM( dst->DstRegister.File, TGSI_FILES ); - if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { - TXT( "\nWriteMask: " ); - ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); - } - if( ignored ) { - if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( dst->DstRegister.Indirect ); - } - if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( dst->DstRegister.Dimension ); - } - } - if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { - TXT( "\nIndex : " ); - SID( dst->DstRegister.Index ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegister.Padding ); - if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegister.Extended ); - } - } - - if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { - TXT( "\nCondMask : " ); - ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { - TXT( "\nCondSwizzleX: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { - TXT( "\nCondSwizzleY: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { - TXT( "\nCondSwizzleW: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { - TXT( "\nCondSrcIndex: " ); - UID( dst->DstRegisterExtConcode.CondSrcIndex ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtConcode.Padding ); - if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegisterExtConcode.Extended ); - } - } - } - - if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { - TXT( "\nModulate: " ); - ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtModulate.Padding ); - if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { - TXT( "\nExtended: " ); - UID( dst->DstRegisterExtModulate.Extended ); - } - } - } - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : "); - ENM( src->SrcRegister.File, TGSI_FILES ); - if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { - TXT( "\nSwizzleX : " ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { - TXT( "\nSwizzleY : " ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { - TXT( "\nSwizzleZ : " ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { - TXT( "\nSwizzleW : " ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegister.Negate ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( src->SrcRegister.Indirect ); - } - if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( src->SrcRegister.Dimension ); - } - } - if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { - TXT( "\nIndex : " ); - SID( src->SrcRegister.Index ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegister.Extended ); - } - } - - if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { - TXT( "\nExtSwizzleX: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { - TXT( "\nExtSwizzleY: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { - TXT( "\nExtSwizzleZ: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { - TXT( "\nExtSwizzleW: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { - TXT( "\nNegateX : " ); - UID( src->SrcRegisterExtSwz.NegateX ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { - TXT( "\nNegateY : " ); - UID( src->SrcRegisterExtSwz.NegateY ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { - TXT( "\nNegateZ : " ); - UID( src->SrcRegisterExtSwz.NegateZ ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { - TXT( "\nNegateW : " ); - UID( src->SrcRegisterExtSwz.NegateW ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { - TXT( "\nExtDivide : " ); - ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtSwz.Padding ); - if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtSwz.Extended ); - } - } - } - - if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { - TXT( "\nComplement: " ); - UID( src->SrcRegisterExtMod.Complement ); - } - if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { - TXT( "\nBias : " ); - UID( src->SrcRegisterExtMod.Bias ); - } - if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { - TXT( "\nScale2X : " ); - UID( src->SrcRegisterExtMod.Scale2X ); - } - if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { - TXT( "\nAbsolute : " ); - UID( src->SrcRegisterExtMod.Absolute ); - } - if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegisterExtMod.Negate ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtMod.Padding ); - if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtMod.Extended ); - } - } - } - } -} - -void -tgsi_dump( - const struct tgsi_token *tokens, - GLuint flags ) -{ - struct text_dump _dump; - struct text_dump *dump = &_dump; - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - GLuint verbose = flags & TGSI_DUMP_VERBOSE; - GLuint ignored = !(flags & TGSI_DUMP_NO_IGNORED); - GLuint deflt = !(flags & TGSI_DUMP_NO_DEFAULT); - GLuint instno = 0; - - { -#if 0 - static GLuint counter = 0; - char buffer[64]; - - sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); - dump->file = fopen( buffer, "wt" ); -#else - dump->file = stderr; - dump->tabs = 0; -#endif - } - - /* sanity check */ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); - - tgsi_parse_init( &parse, tokens ); - - TXT( "tgsi-dump begin -----------------" ); - - CHR( '\n' ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); - CHR( ' ' ); - UID( parse.FullVersion.Version.MajorVersion ); - CHR( '.' ); - UID( parse.FullVersion.Version.MinorVersion ); - - if( verbose ) { - TXT( "\nMajorVersion: " ); - UID( parse.FullVersion.Version.MajorVersion ); - TXT( "\nMinorVersion: " ); - UID( parse.FullVersion.Version.MinorVersion ); - CHR( '\n' ); - - TXT( "\nHeaderSize: " ); - UID( parse.FullHeader.Header.HeaderSize ); - TXT( "\nBodySize : " ); - UID( parse.FullHeader.Header.BodySize ); - TXT( "\nProcessor : " ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); - CHR( '\n' ); - } - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_short( - dump, - &parse.FullToken.FullDeclaration ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_short( - dump, - &parse.FullToken.FullImmediate ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_short( - dump, - &parse.FullToken.FullInstruction, - instno ); - instno++; - break; - - default: - assert( 0 ); - } - - if( verbose ) { - TXT( "\nType : " ); - ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); - if( ignored ) { - TXT( "\nSize : " ); - UID( parse.FullToken.Token.Size ); - if( deflt || parse.FullToken.Token.Extended ) { - TXT( "\nExtended : " ); - UID( parse.FullToken.Token.Extended ); - } - } - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_verbose( - dump, - &parse.FullToken.FullDeclaration, - ignored, - deflt, - &fd ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_verbose( - dump, - &parse.FullToken.FullImmediate, - ignored ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_verbose( - dump, - &parse.FullToken.FullInstruction, - ignored, - deflt, - &fi ); - break; - - default: - assert( 0 ); - } - - CHR( '\n' ); - } - } - - TXT( "\ntgsi-dump end -------------------\n" ); - - tgsi_parse_free( &parse ); - - if (dump->file != stderr && - dump->file != stdout) - fclose( dump->file ); -} - diff --git a/src/mesa/pipe/tgsi/core/tgsi_dump.h b/src/mesa/pipe/tgsi/core/tgsi_dump.h deleted file mode 100644 index 70860c0885..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_dump.h +++ /dev/null @@ -1,22 +0,0 @@ -#if !defined TGSI_DUMP_H -#define TGSI_DUMP_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -#define TGSI_DUMP_VERBOSE 1 -#define TGSI_DUMP_NO_IGNORED 2 -#define TGSI_DUMP_NO_DEFAULT 4 - -void -tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_DUMP_H - diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.c b/src/mesa/pipe/tgsi/core/tgsi_exec.c deleted file mode 100644 index a1f46712b4..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.c +++ /dev/null @@ -1,2313 +0,0 @@ -#include "tgsi_platform.h" -#include "tgsi_core.h" -#include "pipe/p_state.h" - -#define MESA 1 -#if MESA -#include "main/context.h" -#include "main/macros.h" -#endif - -#define TILE_BOTTOM_LEFT 0 -#define TILE_BOTTOM_RIGHT 1 -#define TILE_TOP_LEFT 2 -#define TILE_TOP_RIGHT 3 - -#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I -#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C -#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I -#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C -#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I -#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C -#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I -#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C -#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I -#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C -#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I -#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C -#define TEMP_128_I TGSI_EXEC_TEMP_128_I -#define TEMP_128_C TGSI_EXEC_TEMP_128_C -#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I -#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C -#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I -#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C -#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I -#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C -#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I -#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C -#define TEMP_R0 TGSI_EXEC_TEMP_R0 - -#define FOR_EACH_CHANNEL(CHAN)\ - for (CHAN = 0; CHAN < 4; CHAN++) - -#define IS_CHANNEL_ENABLED(INST, CHAN)\ - ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) - -#define IS_CHANNEL_ENABLED2(INST, CHAN)\ - ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) - -#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED( INST, CHAN )) - -#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED2( INST, CHAN )) - -#define CHAN_X 0 -#define CHAN_Y 1 -#define CHAN_Z 2 -#define CHAN_W 3 - -void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach, - const struct tgsi_token *tokens, - GLuint numSamplers, - struct tgsi_sampler *samplers) -{ - GLuint i, k; - struct tgsi_parse_context parse; - - mach->Tokens = tokens; - - mach->Samplers = samplers; - - k = tgsi_parse_init (&parse, mach->Tokens); - if (k != TGSI_PARSE_OK) { - printf("Problem parsing!\n"); - return; - } - - mach->Processor = parse.FullHeader.Processor.Processor; - tgsi_parse_free (&parse); - - mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); - mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; - -#if XXX_SSE - tgsi_emit_sse (tokens, - &mach->Function); -#endif - - /* Setup constants. */ - for( i = 0; i < 4; i++ ) { - mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; - mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; - mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; - mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; - mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; - mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; - mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; - mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; - } -} - -void -tgsi_exec_prepare( - struct tgsi_exec_machine *mach, - struct tgsi_exec_labels *labels ) -{ - struct tgsi_parse_context parse; - GLuint k; - GLuint instno = 0; - - mach->ImmLimit = 0; - labels->count = 0; - - k = tgsi_parse_init( &parse, mach->Tokens ); - if (k != TGSI_PARSE_OK) { - printf("Problem parsing!\n"); - return; - } - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - GLuint pointer = parse.Position; - GLuint i; - - tgsi_parse_token( &parse ); - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - assert( (parse.FullToken.FullImmediate.Immediate.Size - 1) % 4 == 0 ); - assert( mach->ImmLimit + (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4 <= 256 ); - - for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size - 1; i++ ) { - mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; - } - mach->ImmLimit += (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4; - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - assert( labels->count < 128 ); - - labels->labels[labels->count][0] = instno; - labels->labels[labels->count][1] = pointer; - labels->count++; - break; - - default: - assert( 0 ); - } - } - tgsi_parse_free (&parse); -} - -void -tgsi_exec_machine_run( - struct tgsi_exec_machine *mach ) -{ - struct tgsi_exec_labels labels; - - tgsi_exec_prepare( mach, &labels ); - tgsi_exec_machine_run2( mach, &labels ); -} - -static void -micro_abs( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) fabs( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) fabs( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) fabs( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) fabs( (GLdouble) src->f[3] ); -} - -static void -micro_add( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] + src1->f[0]; - dst->f[1] = src0->f[1] + src1->f[1]; - dst->f[2] = src0->f[2] + src1->f[2]; - dst->f[3] = src0->f[3] + src1->f[3]; -} - -static void -micro_iadd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] + src1->i[0]; - dst->i[1] = src0->i[1] + src1->i[1]; - dst->i[2] = src0->i[2] + src1->i[2]; - dst->i[3] = src0->i[3] + src1->i[3]; -} - -static void -micro_and( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] & src1->u[0]; - dst->u[1] = src0->u[1] & src1->u[1]; - dst->u[2] = src0->u[2] & src1->u[2]; - dst->u[3] = src0->u[3] & src1->u[3]; -} - -static void -micro_ceil( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) ceil( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) ceil( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) ceil( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) ceil( (GLdouble) src->f[3] ); -} - -static void -micro_cos( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) cos( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) cos( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) cos( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) cos( (GLdouble) src->f[3] ); -} - -static void -micro_ddx( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_ddy( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_div( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] / src1->f[0]; - dst->f[1] = src0->f[1] / src1->f[1]; - dst->f[2] = src0->f[2] / src1->f[2]; - dst->f[3] = src0->f[3] / src1->f[3]; -} - -static void -micro_udiv( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] / src1->u[0]; - dst->u[1] = src0->u[1] / src1->u[1]; - dst->u[2] = src0->u[2] / src1->u[2]; - dst->u[3] = src0->u[3] / src1->u[3]; -} - -static void -micro_eq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_ieq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; -} - -static void -micro_exp2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) pow( 2.0, (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) pow( 2.0, (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) pow( 2.0, (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) pow( 2.0, (GLdouble) src->f[3] ); -} - -static void -micro_f2it( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->i[0] = (GLint) src->f[0]; - dst->i[1] = (GLint) src->f[1]; - dst->i[2] = (GLint) src->f[2]; - dst->i[3] = (GLint) src->f[3]; -} - -static void -micro_f2ut( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = (GLuint) src->f[0]; - dst->u[1] = (GLuint) src->f[1]; - dst->u[2] = (GLuint) src->f[2]; - dst->u[3] = (GLuint) src->f[3]; -} - -static void -micro_flr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) floor( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) floor( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) floor( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) floor( (GLdouble) src->f[3] ); -} - -static void -micro_frc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = src->f[0] - (GLfloat) floor( (GLdouble) src->f[0] ); - dst->f[1] = src->f[1] - (GLfloat) floor( (GLdouble) src->f[1] ); - dst->f[2] = src->f[2] - (GLfloat) floor( (GLdouble) src->f[2] ); - dst->f[3] = src->f[3] - (GLfloat) floor( (GLdouble) src->f[3] ); -} - -static void -micro_ge( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_i2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) src->i[0]; - dst->f[1] = (GLfloat) src->i[1]; - dst->f[2] = (GLfloat) src->i[2]; - dst->f[3] = (GLfloat) src->i[3]; -} - -static void -micro_lg2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) log( (GLdouble) src->f[0] ) * 1.442695f; - dst->f[1] = (GLfloat) log( (GLdouble) src->f[1] ) * 1.442695f; - dst->f[2] = (GLfloat) log( (GLdouble) src->f[2] ) * 1.442695f; - dst->f[3] = (GLfloat) log( (GLdouble) src->f[3] ) * 1.442695f; -} - -static void -micro_lt( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_ilt( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; -} - -static void -micro_ult( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; -} - -static void -micro_max( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; -} - -static void -micro_imax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; -} - -static void -micro_umax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; -} - -static void -micro_min( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; -} - -static void -micro_imin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; -} - -static void -micro_umin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; -} - -static void -micro_umod( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] % src1->u[0]; - dst->u[1] = src0->u[1] % src1->u[1]; - dst->u[2] = src0->u[2] % src1->u[2]; - dst->u[3] = src0->u[3] % src1->u[3]; -} - -static void -micro_mul( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] * src1->f[0]; - dst->f[1] = src0->f[1] * src1->f[1]; - dst->f[2] = src0->f[2] * src1->f[2]; - dst->f[3] = src0->f[3] * src1->f[3]; -} - -static void -micro_imul( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] * src1->i[0]; - dst->i[1] = src0->i[1] * src1->i[1]; - dst->i[2] = src0->i[2] * src1->i[2]; - dst->i[3] = src0->i[3] * src1->i[3]; -} - -static void -micro_imul64( - union tgsi_exec_channel *dst0, - union tgsi_exec_channel *dst1, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst1->i[0] = src0->i[0] * src1->i[0]; - dst1->i[1] = src0->i[1] * src1->i[1]; - dst1->i[2] = src0->i[2] * src1->i[2]; - dst1->i[3] = src0->i[3] * src1->i[3]; - dst0->i[0] = 0; - dst0->i[1] = 0; - dst0->i[2] = 0; - dst0->i[3] = 0; -} - -static void -micro_umul64( - union tgsi_exec_channel *dst0, - union tgsi_exec_channel *dst1, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst1->u[0] = src0->u[0] * src1->u[0]; - dst1->u[1] = src0->u[1] * src1->u[1]; - dst1->u[2] = src0->u[2] * src1->u[2]; - dst1->u[3] = src0->u[3] * src1->u[3]; - dst0->u[0] = 0; - dst0->u[1] = 0; - dst0->u[2] = 0; - dst0->u[3] = 0; -} - -static void -micro_movc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2 ) -{ - dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; - dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; - dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; - dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; -} - -static void -micro_neg( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = -src->f[0]; - dst->f[1] = -src->f[1]; - dst->f[2] = -src->f[2]; - dst->f[3] = -src->f[3]; -} - -static void -micro_ineg( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->i[0] = -src->i[0]; - dst->i[1] = -src->i[1]; - dst->i[2] = -src->i[2]; - dst->i[3] = -src->i[3]; -} - -static void -micro_not( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = ~src->u[0]; - dst->u[1] = ~src->u[1]; - dst->u[2] = ~src->u[2]; - dst->u[3] = ~src->u[3]; -} - -static void -micro_or( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] | src1->u[0]; - dst->u[1] = src0->u[1] | src1->u[1]; - dst->u[2] = src0->u[2] | src1->u[2]; - dst->u[3] = src0->u[3] | src1->u[3]; -} - -static void -micro_pow( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = (GLfloat) pow( (GLdouble) src0->f[0], (GLdouble) src1->f[0] ); - dst->f[1] = (GLfloat) pow( (GLdouble) src0->f[1], (GLdouble) src1->f[1] ); - dst->f[2] = (GLfloat) pow( (GLdouble) src0->f[2], (GLdouble) src1->f[2] ); - dst->f[3] = (GLfloat) pow( (GLdouble) src0->f[3], (GLdouble) src1->f[3] ); -} - -static void -micro_rnd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) floor( (GLdouble) (src->f[0] + 0.5f) ); - dst->f[1] = (GLfloat) floor( (GLdouble) (src->f[1] + 0.5f) ); - dst->f[2] = (GLfloat) floor( (GLdouble) (src->f[2] + 0.5f) ); - dst->f[3] = (GLfloat) floor( (GLdouble) (src->f[3] + 0.5f) ); -} - -static void -micro_shl( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] << src1->i[0]; - dst->i[1] = src0->i[1] << src1->i[1]; - dst->i[2] = src0->i[2] << src1->i[2]; - dst->i[3] = src0->i[3] << src1->i[3]; -} - -static void -micro_ishr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] >> src1->i[0]; - dst->i[1] = src0->i[1] >> src1->i[1]; - dst->i[2] = src0->i[2] >> src1->i[2]; - dst->i[3] = src0->i[3] >> src1->i[3]; -} - -static void -micro_ushr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] >> src1->u[0]; - dst->u[1] = src0->u[1] >> src1->u[1]; - dst->u[2] = src0->u[2] >> src1->u[2]; - dst->u[3] = src0->u[3] >> src1->u[3]; -} - -static void -micro_sin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) sin( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) sin( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) sin( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) sin( (GLdouble) src->f[3] ); -} - -static void -micro_sqrt( union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) sqrt( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) sqrt( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) sqrt( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) sqrt( (GLdouble) src->f[3] ); -} - -static void -micro_sub( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] - src1->f[0]; - dst->f[1] = src0->f[1] - src1->f[1]; - dst->f[2] = src0->f[2] - src1->f[2]; - dst->f[3] = src0->f[3] - src1->f[3]; -} - -static void -micro_u2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (GLfloat) src->u[0]; - dst->f[1] = (GLfloat) src->u[1]; - dst->f[2] = (GLfloat) src->u[2]; - dst->f[3] = (GLfloat) src->u[3]; -} - -static void -micro_xor( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] ^ src1->u[0]; - dst->u[1] = src0->u[1] ^ src1->u[1]; - dst->u[2] = src0->u[2] ^ src1->u[2]; - dst->u[3] = src0->u[3] ^ src1->u[3]; -} - -static void -fetch_src_file_channel( - const struct tgsi_exec_machine *mach, - const GLuint file, - const GLuint swizzle, - const union tgsi_exec_channel *index, - union tgsi_exec_channel *chan ) -{ - switch( swizzle ) { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - switch( file ) { - case TGSI_FILE_CONSTANT: - chan->f[0] = mach->Consts[index->i[0]][swizzle]; - chan->f[1] = mach->Consts[index->i[1]][swizzle]; - chan->f[2] = mach->Consts[index->i[2]][swizzle]; - chan->f[3] = mach->Consts[index->i[3]][swizzle]; - break; - - case TGSI_FILE_INPUT: - chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_TEMPORARY: - chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_IMMEDIATE: - assert( index->i[0] < (GLint) mach->ImmLimit ); - chan->f[0] = mach->Imms[index->i[0]][swizzle]; - assert( index->i[1] < (GLint) mach->ImmLimit ); - chan->f[1] = mach->Imms[index->i[1]][swizzle]; - assert( index->i[2] < (GLint) mach->ImmLimit ); - chan->f[2] = mach->Imms[index->i[2]][swizzle]; - assert( index->i[3] < (GLint) mach->ImmLimit ); - chan->f[3] = mach->Imms[index->i[3]][swizzle]; - break; - - case TGSI_FILE_ADDRESS: - chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; - break; - - default: - assert( 0 ); - } - break; - - case TGSI_EXTSWIZZLE_ZERO: - *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; - break; - - case TGSI_EXTSWIZZLE_ONE: - *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; - break; - - default: - assert( 0 ); - } -} - -static void -fetch_source( - const struct tgsi_exec_machine *mach, - union tgsi_exec_channel *chan, - const struct tgsi_full_src_register *reg, - const GLuint chan_index ) -{ - union tgsi_exec_channel index; - GLuint swizzle; - - index.i[0] = - index.i[1] = - index.i[2] = - index.i[3] = reg->SrcRegister.Index; - - if (reg->SrcRegister.Indirect) { - union tgsi_exec_channel index2; - union tgsi_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterInd, CHAN_X ); - fetch_src_file_channel( - mach, - reg->SrcRegisterInd.File, - swizzle, - &index2, - &indir_index ); - - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; - } - - if( reg->SrcRegister.Dimension ) { - switch( reg->SrcRegister.File ) { - case TGSI_FILE_INPUT: - index.i[0] *= 17; - index.i[1] *= 17; - index.i[2] *= 17; - index.i[3] *= 17; - break; - case TGSI_FILE_CONSTANT: - index.i[0] *= 4096; - index.i[1] *= 4096; - index.i[2] *= 4096; - index.i[3] *= 4096; - break; - default: - assert( 0 ); - } - - index.i[0] += reg->SrcRegisterDim.Index; - index.i[1] += reg->SrcRegisterDim.Index; - index.i[2] += reg->SrcRegisterDim.Index; - index.i[3] += reg->SrcRegisterDim.Index; - - if (reg->SrcRegisterDim.Indirect) { - union tgsi_exec_channel index2; - union tgsi_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterDimInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); - fetch_src_file_channel( - mach, - reg->SrcRegisterDimInd.File, - swizzle, - &index2, - &indir_index ); - - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; - } - } - - swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); - fetch_src_file_channel( - mach, - reg->SrcRegister.File, - swizzle, - &index, - chan ); - - switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { - case TGSI_UTIL_SIGN_CLEAR: - micro_abs( chan, chan ); - break; - - case TGSI_UTIL_SIGN_SET: - micro_abs( chan, chan ); - micro_neg( chan, chan ); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - micro_neg( chan, chan ); - break; - - case TGSI_UTIL_SIGN_KEEP: - break; - } -} - -static void -store_dest( - struct tgsi_exec_machine *mach, - const union tgsi_exec_channel *chan, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - GLuint chan_index ) -{ - union tgsi_exec_channel *dst; - - switch( reg->DstRegister.File ) { - case TGSI_FILE_NULL: - return; - - case TGSI_FILE_OUTPUT: - dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_TEMPORARY: - dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_ADDRESS: - dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; - break; - - default: - assert( 0 ); - } - - switch (inst->Instruction.Saturate) - { - case TGSI_SAT_NONE: - *dst = *chan; - break; - - case TGSI_SAT_ZERO_ONE: - micro_lt( dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], chan ); - micro_lt( dst, chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - break; - - case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); - break; - - default: - assert( 0 ); - } -} - -#define FETCH(VAL,INDEX,CHAN)\ - fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) - -#define STORE(VAL,INDEX,CHAN)\ - store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN) - -static void -exec_kil (struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst) -{ - GLuint uniquemask; - GLuint chan_index; - GLuint kilmask = 0; - union tgsi_exec_channel r[1]; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - for (chan_index = 0; chan_index < 4; chan_index++) - { - GLuint swizzle; - GLuint i; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle ( - &inst->FullSrcRegisters[0], - chan_index); - - /* check if the component has not been already tested */ - if (uniquemask & (1 << swizzle)) - continue; - uniquemask |= 1 << swizzle; - - FETCH(&r[0], 0, chan_index); - for (i = 0; i < 4; i++) - if (r[0].f[i] < 0.0f) - kilmask |= 1 << (i * 4); - } - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; -} - - - - -/* - * Fetch a texel using STR texture coordinates. - */ -static void -fetch_texel( struct tgsi_sampler *sampler, - const union tgsi_exec_channel *s, - const union tgsi_exec_channel *t, - const union tgsi_exec_channel *p, - float lodbias, - union tgsi_exec_channel *r, - union tgsi_exec_channel *g, - union tgsi_exec_channel *b, - union tgsi_exec_channel *a ) -{ - GLuint j; - GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]; - - sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); - - for (j = 0; j < 4; j++) { - r->f[j] = rgba[0][j]; - g->f[j] = rgba[1][j]; - b->f[j] = rgba[2][j]; - a->f[j] = rgba[3][j]; - } -} - -static void -constant_interpolation( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; - } -} - -static void -linear_interpolation( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = mach->Inputs[0].xyzw[0].f[i]; - const float y = mach->Inputs[0].xyzw[1].f[i]; - - mach->Inputs[attrib].xyzw[chan].f[i] = - mach->InterpCoefs[attrib].a0[chan] + - mach->InterpCoefs[attrib].dadx[chan] * x + - mach->InterpCoefs[attrib].dady[chan] * y; - } -} - -static void -perspective_interpolation( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = mach->Inputs[0].xyzw[0].f[i]; - const float y = mach->Inputs[0].xyzw[1].f[i]; - // WPOS.w here is really 1/w - const float w = 1.0f / mach->Inputs[0].xyzw[3].f[i]; - - mach->Inputs[attrib].xyzw[chan].f[i] = - (mach->InterpCoefs[attrib].a0[chan] + - mach->InterpCoefs[attrib].dadx[chan] * x + - mach->InterpCoefs[attrib].dady[chan] * y) * w; - } -} - -typedef void (* interpolation_func)( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ); - -static void -exec_declaration( - struct tgsi_exec_machine *mach, - const struct tgsi_full_declaration *decl ) -{ - if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { - if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask; - interpolation_func interp; - - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - /* Do not touch WPOS.xy */ - if( first == 0 ) { - mask &= ~TGSI_WRITEMASK_XY; - if( mask == TGSI_WRITEMASK_NONE ) { - first++; - if( first > last ) { - return; - } - } - } - - switch( decl->Interpolation.Interpolate ) { - case TGSI_INTERPOLATE_CONSTANT: - interp = constant_interpolation; - break; - - case TGSI_INTERPOLATE_LINEAR: - interp = linear_interpolation; - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - interp = perspective_interpolation; - break; - - default: - assert( 0 ); - } - - if( mask == TGSI_WRITEMASK_XYZW ) { - unsigned i, j; - - for( i = first; i <= last; i++ ) { - for( j = 0; j < NUM_CHANNELS; j++ ) { - interp( mach, i, j ); - } - } - } - else { - unsigned i, j; - - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - interp( mach, i, j ); - } - } - } - } - } - } -} - -static void -exec_instruction( - struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst, - struct tgsi_exec_labels *labels, - GLuint *programCounter ) -{ - GLuint chan_index; - union tgsi_exec_channel r[8]; - - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_f2it( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LIT: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_X ); - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[1], 0, CHAN_Y ); - micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - - FETCH( &r[2], 0, CHAN_W ); - micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); - micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); - micro_pow( &r[1], &r[1], &r[2] ); - micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Z ); - } - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH( &r[0], 0, CHAN_X ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH( &r[0], 0, CHAN_X ); - micro_sqrt( &r[0], &r[0] ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXP: - assert (0); - break; - - case TGSI_OPCODE_LOG: - assert (0); - break; - - case TGSI_OPCODE_MUL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_mul( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_ADD: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Z ); - FETCH( &r[2], 1, CHAN_Z ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_W); - FETCH(&r[2], 1, CHAN_W); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DST: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - FETCH( &r[0], 0, CHAN_Y ); - FETCH( &r[1], 1, CHAN_Y); - micro_mul( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_Z ); - STORE( &r[0], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - FETCH( &r[0], 1, CHAN_W ); - STORE( &r[0], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MIN: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_MAX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_mul( &r[0], &r[0], &r[1] ); - FETCH( &r[1], 2, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SUB: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_sub( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - micro_sub( &r[1], &r[1], &r[2] ); - micro_mul( &r[0], &r[0], &r[1] ); - micro_add( &r[0], &r[0], &r[2] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_CND: - assert (0); - break; - - case TGSI_OPCODE_CND0: - assert (0); - break; - - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - assert (0); - break; - - case TGSI_OPCODE_INDEX: - assert (0); - break; - - case TGSI_OPCODE_NEGATE: - assert (0); - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_frc( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CLAMP: - assert (0); - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_flr( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_ROUND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_rnd( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ - FETCH(&r[0], 0, CHAN_X); - - micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ - FETCH( &r[0], 0, CHAN_X ); - micro_lg2( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_pow( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - FETCH(&r[0], 0, CHAN_Y); - FETCH(&r[1], 1, CHAN_Z); - - micro_mul( &r[2], &r[0], &r[1] ); - - FETCH(&r[3], 0, CHAN_Z); - FETCH(&r[4], 1, CHAN_Y); - - micro_mul( &r[5], &r[3], &r[4] ); - micro_sub( &r[2], &r[2], &r[5] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &r[2], 0, CHAN_X ); - } - - FETCH(&r[2], 1, CHAN_X); - - micro_mul( &r[3], &r[3], &r[2] ); - - FETCH(&r[5], 0, CHAN_X); - - micro_mul( &r[1], &r[1], &r[5] ); - micro_sub( &r[3], &r[3], &r[1] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - STORE( &r[3], 0, CHAN_Y ); - } - - micro_mul( &r[5], &r[5], &r[4] ); - micro_mul( &r[0], &r[0], &r[2] ); - micro_sub( &r[5], &r[5], &r[0] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - STORE( &r[5], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MULTIPLYMATRIX: - assert (0); - break; - - case TGSI_OPCODE_ABS: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - - micro_abs( &r[0], &r[0] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_RCC: - assert (0); - break; - - case TGSI_OPCODE_DPH: - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 1, CHAN_W); - - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_COS: - FETCH(&r[0], 0, CHAN_X); - - micro_cos( &r[0], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddx( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDY: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddy( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_KIL: - exec_kil (mach, inst); - break; - - case TGSI_OPCODE_PK2H: - assert (0); - break; - - case TGSI_OPCODE_PK2US: - assert (0); - break; - - case TGSI_OPCODE_PK4B: - assert (0); - break; - - case TGSI_OPCODE_PK4UB: - assert (0); - break; - - case TGSI_OPCODE_RFL: - assert (0); - break; - - case TGSI_OPCODE_SEQ: - assert (0); - break; - - case TGSI_OPCODE_SFL: - assert (0); - break; - - case TGSI_OPCODE_SGT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SIN: - FETCH( &r[0], 0, CHAN_X ); - micro_sin( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SNE: - assert (0); - break; - - case TGSI_OPCODE_STR: - assert (0); - break; - - case TGSI_OPCODE_TEX: - { - const GLuint unit = inst->FullSrcRegisters[1].SrcRegister.Index; - switch (inst->InstructionExtTexture.Texture) { - case TGSI_TEXTURE_1D: - - FETCH(&r[0], 0, CHAN_X); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[1], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[1] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - fetch_texel(&mach->Samplers[unit], - &r[0], NULL, NULL, 0.0, - &r[0], &r[1], &r[2], &r[3]); - break; - - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[2], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[2] ); - micro_div( &r[1], &r[1], &r[2] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], NULL, 0.0, - &r[0], &r[1], &r[2], &r[3]); - break; - - case TGSI_TEXTURE_3D: - case TGSI_TEXTURE_CUBE: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], 0.0, - &r[0], &r[1], &r[2], &r[3]); - break; - - default: - assert (0); - } - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[chan_index], 0, chan_index ); - } - } - break; - - case TGSI_OPCODE_TXD: - assert (0); - break; - - case TGSI_OPCODE_UP2H: - assert (0); - break; - - case TGSI_OPCODE_UP2US: - assert (0); - break; - - case TGSI_OPCODE_UP4B: - assert (0); - break; - - case TGSI_OPCODE_UP4UB: - assert (0); - break; - - case TGSI_OPCODE_X2D: - assert (0); - break; - - case TGSI_OPCODE_ARA: - assert (0); - break; - - case TGSI_OPCODE_ARR: - assert (0); - break; - - case TGSI_OPCODE_BRA: - assert (0); - break; - - case TGSI_OPCODE_CAL: - assert (0); - break; - - case TGSI_OPCODE_RET: - /* XXX: end of shader! */ - /*assert (0);*/ - break; - - case TGSI_OPCODE_SSG: - assert (0); - break; - - case TGSI_OPCODE_CMP: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_SCS: - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( &r[0], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - micro_cos( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - micro_sin( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_Y ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_TXB: - assert (0); - break; - - case TGSI_OPCODE_NRM: - assert (0); - break; - - case TGSI_OPCODE_DIV: - assert( 0 ); - break; - - case TGSI_OPCODE_DP2: - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_TXL: - assert (0); - break; - - case TGSI_OPCODE_BRK: - assert (0); - break; - - case TGSI_OPCODE_IF: - { - GLuint cond = 0; - struct tgsi_exec_cond_state *state; - - /* Allocate condition state. */ - assert( mach->CondStack.Index > 0 ); - mach->CondStack.Index--; - - /* Evaluate the condition mask. */ - FETCH( &r[0], 0, CHAN_X ); - if( r[0].u[0] ) { - cond |= 1; - } - if( r[0].u[1] ) { - cond |= 2; - } - if( r[0].u[2] ) { - cond |= 4; - } - if( r[0].u[3] ) { - cond |= 8; - } - - state = &mach->CondStack.States[mach->CondStack.Index]; - - /* Initialize the If portion of condition state. */ - memcpy( - state->IfPortion.TempsAddrs, - mach->Temps, - sizeof( state->IfPortion.TempsAddrs ) ); - memcpy( - state->IfPortion.Outputs, - mach->Outputs, - sizeof( state->IfPortion.Outputs ) ); - state->Condition = cond; - state->WasElse = GL_FALSE; - } - break; - - case TGSI_OPCODE_LOOP: - assert (0); - break; - - case TGSI_OPCODE_REP: - assert (0); - break; - - case TGSI_OPCODE_ELSE: - { - struct tgsi_exec_cond_state *state; - struct tgsi_exec_cond_regs temp; - - state = &mach->CondStack.States[mach->CondStack.Index]; - - /* Copy the results of the If portion to temporary storage. */ - memcpy( - temp.TempsAddrs, - mach->Temps, - sizeof( temp.TempsAddrs ) ); - memcpy( - temp.Outputs, - mach->Outputs, - sizeof( temp.Outputs ) ); - - /* Restore the state of registers from before the If statement. */ - memcpy( - mach->Temps, - state->IfPortion.TempsAddrs, - sizeof( state->IfPortion.TempsAddrs ) ); - memcpy( - mach->Outputs, - state->IfPortion.Outputs, - sizeof( state->IfPortion.Outputs ) ); - - /* Save the results of If portion. */ - memcpy( - &state->IfPortion, - &temp, - sizeof( state->IfPortion ) ); - state->WasElse = GL_TRUE; - } - break; - - case TGSI_OPCODE_ENDIF: - { - struct tgsi_exec_cond_state *state; - GLuint i; - - state = &mach->CondStack.States[mach->CondStack.Index]; - - if( state->WasElse ) { - /* Save the results of Else portion. */ - memcpy( - state->ElsePortion.TempsAddrs, - mach->Temps, - sizeof( state->ElsePortion.TempsAddrs ) ); - memcpy( - state->ElsePortion.Outputs, - mach->Outputs, - sizeof( state->ElsePortion.Outputs ) ); - } - else { - /* Copy the state of registers from before the If statement to Else portion. */ - memcpy( - &state->ElsePortion, - &state->IfPortion, - sizeof( state->ElsePortion ) ); - - /* Save the results of the If portion. */ - memcpy( - state->IfPortion.TempsAddrs, - mach->Temps, - sizeof( state->IfPortion.TempsAddrs ) ); - memcpy( - state->IfPortion.Outputs, - mach->Outputs, - sizeof( state->IfPortion.Outputs ) ); - } - - /* Mix the If and Else portions based on condition mask. */ - for( i = 0; i < 4; i++ ) { - struct tgsi_exec_cond_regs *regs; - GLuint j; - - if( state->Condition & (1 << i) ) { - regs = &state->IfPortion; - } - else { - regs = &state->ElsePortion; - } - - for( j = 0; j < TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS; j++ ) { - mach->Temps[j].xyzw[0].u[i] = regs->TempsAddrs[j].xyzw[0].u[i]; - mach->Temps[j].xyzw[1].u[i] = regs->TempsAddrs[j].xyzw[1].u[i]; - mach->Temps[j].xyzw[2].u[i] = regs->TempsAddrs[j].xyzw[2].u[i]; - mach->Temps[j].xyzw[3].u[i] = regs->TempsAddrs[j].xyzw[3].u[i]; - } - for( j = 0; j < 2; j++ ) { - mach->Outputs[j].xyzw[0].u[i] = regs->Outputs[j].xyzw[0].u[i]; - mach->Outputs[j].xyzw[1].u[i] = regs->Outputs[j].xyzw[1].u[i]; - mach->Outputs[j].xyzw[2].u[i] = regs->Outputs[j].xyzw[2].u[i]; - mach->Outputs[j].xyzw[3].u[i] = regs->Outputs[j].xyzw[3].u[i]; - } - } - - /* Release condition state. */ - mach->CondStack.Index++; - } - break; - - case TGSI_OPCODE_ENDLOOP: - assert (0); - break; - - case TGSI_OPCODE_ENDREP: - assert (0); - break; - - case TGSI_OPCODE_PUSHA: - assert (0); - break; - - case TGSI_OPCODE_POPA: - assert (0); - break; - - case TGSI_OPCODE_CEIL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ceil( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_I2F: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_i2f( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_NOT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_not( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_TRUNC: - /* TGSI_OPCODE_INT */ - assert (0); - break; - - case TGSI_OPCODE_SHL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_shl( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SHR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ishr( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_AND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_and( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_OR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_or( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOD: - assert (0); - break; - - case TGSI_OPCODE_XOR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_xor( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SAD: - assert (0); - break; - - case TGSI_OPCODE_TXF: - assert (0); - break; - - case TGSI_OPCODE_TXQ: - assert (0); - break; - - case TGSI_OPCODE_CONT: - assert (0); - break; - - case TGSI_OPCODE_EMIT: - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; - break; - - case TGSI_OPCODE_ENDPRIM: - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; - break; - - case TGSI_OPCODE_BGNLOOP2: - assert( 0 ); - break; - - case TGSI_OPCODE_BGNSUB: - assert( 0 ); - break; - - case TGSI_OPCODE_ENDLOOP2: - assert( 0 ); - break; - - case TGSI_OPCODE_ENDSUB: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE1: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE2: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE3: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE4: - assert( 0 ); - break; - - case TGSI_OPCODE_NOP: - break; - - default: - assert( 0 ); - } -} - - -#if !defined(XSTDCALL) -#if defined(WIN32) -#define XSTDCALL __stdcall -#else -#define XSTDCALL -#endif -#endif - -typedef void (XSTDCALL *fp_function) (const struct tgsi_exec_vector *input, - struct tgsi_exec_vector *output, - GLfloat (*constant)[4], - struct tgsi_exec_vector *temporary); - -void -tgsi_exec_machine_run2( - struct tgsi_exec_machine *mach, - struct tgsi_exec_labels *labels ) -{ -#if MESA - GET_CURRENT_CONTEXT(ctx); - GLuint i; -#endif - -#if XXX_SSE - fp_function function; - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; - - function = (fp_function) x86_get_func (&mach->Function); - - function (mach->Inputs, - mach->Outputs, - mach->Consts, - mach->Temps); -#else - struct tgsi_parse_context parse; - GLuint k; - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; - - if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; - mach->Primitives[0] = 0; - } - - mach->CondStack.Index = 8; - - k = tgsi_parse_init( &parse, mach->Tokens ); - if (k != TGSI_PARSE_OK) { - printf("Problem parsing!\n"); - return; - } - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - exec_declaration( mach, &parse.FullToken.FullDeclaration ); - break; - case TGSI_TOKEN_TYPE_IMMEDIATE: - break; - case TGSI_TOKEN_TYPE_INSTRUCTION: - exec_instruction( mach, &parse.FullToken.FullInstruction, labels, &parse.Position ); - break; - default: - assert( 0 ); - } - } - tgsi_parse_free (&parse); -#endif - -#if 0 - /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ - if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { - /* - * Scale back depth component. - */ - for (i = 0; i < 4; i++) - mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; - } -#endif -} - diff --git a/src/mesa/pipe/tgsi/core/tgsi_exec.h b/src/mesa/pipe/tgsi/core/tgsi_exec.h deleted file mode 100644 index e5e8c3608e..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_exec.h +++ /dev/null @@ -1,191 +0,0 @@ -#if !defined TGSI_EXEC_H -#define TGSI_EXEC_H - -#include "pipe/p_compiler.h" - -#if 0 -#include "x86/rtasm/x86sse.h" -#endif - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -#define NUM_CHANNELS 4 /* R,G,B,A */ -#define QUAD_SIZE 4 /* 4 pixel/quad */ - -union tgsi_exec_channel -{ - float f[QUAD_SIZE]; - int i[QUAD_SIZE]; - unsigned u[QUAD_SIZE]; -}; - -struct tgsi_exec_vector -{ - union tgsi_exec_channel xyzw[NUM_CHANNELS]; -}; - -struct tgsi_interp_coef -{ - float a0[NUM_CHANNELS]; /* in an xyzw layout */ - float dadx[NUM_CHANNELS]; - float dady[NUM_CHANNELS]; -}; - -#define TEX_CACHE_TILE_SIZE 8 -#define TEX_CACHE_NUM_ENTRIES 8 - -struct tgsi_texture_cache_entry -{ - int x, y, face, level, zslice; - float data[TEX_CACHE_TILE_SIZE][TEX_CACHE_TILE_SIZE][4]; -}; - -struct tgsi_sampler -{ - const struct pipe_sampler_state *state; - struct pipe_mipmap_tree *texture; - /** Get samples for four fragments in a quad */ - void (*get_samples)(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]); - void *pipe; /*XXX temporary*/ - struct tgsi_texture_cache_entry cache[TEX_CACHE_NUM_ENTRIES]; -}; - -struct tgsi_exec_labels -{ - unsigned labels[128][2]; - unsigned count; -}; - -#define TGSI_EXEC_TEMP_00000000_I 32 -#define TGSI_EXEC_TEMP_00000000_C 0 - -#define TGSI_EXEC_TEMP_7FFFFFFF_I 32 -#define TGSI_EXEC_TEMP_7FFFFFFF_C 1 - -#define TGSI_EXEC_TEMP_80000000_I 32 -#define TGSI_EXEC_TEMP_80000000_C 2 - -#define TGSI_EXEC_TEMP_FFFFFFFF_I 32 -#define TGSI_EXEC_TEMP_FFFFFFFF_C 3 - -#define TGSI_EXEC_TEMP_ONE_I 33 -#define TGSI_EXEC_TEMP_ONE_C 0 - -#define TGSI_EXEC_TEMP_TWO_I 33 -#define TGSI_EXEC_TEMP_TWO_C 1 - -#define TGSI_EXEC_TEMP_128_I 33 -#define TGSI_EXEC_TEMP_128_C 2 - -#define TGSI_EXEC_TEMP_MINUS_128_I 33 -#define TGSI_EXEC_TEMP_MINUS_128_C 3 - -#define TGSI_EXEC_TEMP_KILMASK_I 34 -#define TGSI_EXEC_TEMP_KILMASK_C 0 - -#define TGSI_EXEC_TEMP_OUTPUT_I 34 -#define TGSI_EXEC_TEMP_OUTPUT_C 1 - -#define TGSI_EXEC_TEMP_PRIMITIVE_I 34 -#define TGSI_EXEC_TEMP_PRIMITIVE_C 2 - -#define TGSI_EXEC_TEMP_R0 35 - -#define TGSI_EXEC_NUM_TEMPS (32 + 4) -#define TGSI_EXEC_NUM_ADDRS 1 - -/* XXX: This is temporary */ -struct tgsi_exec_cond_regs -{ - struct tgsi_exec_vector TempsAddrs[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS]; - struct tgsi_exec_vector Outputs[2]; /* XXX: That's just enough for fragment shader only! */ -}; - -/* XXX: This is temporary */ -struct tgsi_exec_cond_state -{ - struct tgsi_exec_cond_regs IfPortion; - struct tgsi_exec_cond_regs ElsePortion; - unsigned Condition; - boolean WasElse; -}; - -/* XXX: This is temporary */ -struct tgsi_exec_cond_stack -{ - struct tgsi_exec_cond_state States[8]; - unsigned Index; /* into States[] */ -}; - -struct tgsi_exec_machine -{ - /* - * 32 program temporaries - * 4 internal temporaries - * 1 address - * 1 temporary of padding to align to 16 bytes - */ - struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS + 1]; - - /* - * This will point to _Temps after aligning to 16B boundary. - */ - struct tgsi_exec_vector *Temps; - struct tgsi_exec_vector *Addrs; - - struct tgsi_sampler *Samplers; - - float Imms[256][4]; - unsigned ImmLimit; - float (*Consts)[4]; - struct tgsi_exec_vector *Inputs; - struct tgsi_exec_vector *Outputs; - const struct tgsi_token *Tokens; - unsigned Processor; - - /* GEOMETRY processor only. */ - unsigned *Primitives; - - /* FRAGMENT processor only. */ - const struct tgsi_interp_coef *InterpCoefs; - - struct tgsi_exec_cond_stack CondStack; -#if XXX_SSE - struct x86_function Function; -#endif -}; - -void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach, - const struct tgsi_token *tokens, - unsigned numSamplers, - struct tgsi_sampler *samplers); - -void -tgsi_exec_prepare( - struct tgsi_exec_machine *mach, - struct tgsi_exec_labels *labels ); - -void -tgsi_exec_machine_run( - struct tgsi_exec_machine *mach ); - -void -tgsi_exec_machine_run2( - struct tgsi_exec_machine *mach, - struct tgsi_exec_labels *labels ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_EXEC_H - diff --git a/src/mesa/pipe/tgsi/core/tgsi_parse.c b/src/mesa/pipe/tgsi/core/tgsi_parse.c deleted file mode 100644 index a6724d762c..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_parse.c +++ /dev/null @@ -1,288 +0,0 @@ -#include "tgsi_platform.h" -#include "tgsi_core.h" - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ) -{ - full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; -} - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ) -{ - if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) - free( full_token->FullImmediate.u.Pointer ); -} - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ) -{ - ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; - if( ctx->FullVersion.Version.MajorVersion > 1 ) { - return TGSI_PARSE_ERROR; - } - - ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; - if( ctx->FullHeader.Header.HeaderSize >= 2 ) { - ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; - } - else { - ctx->FullHeader.Processor = tgsi_default_processor(); - } - - ctx->Tokens = tokens; - ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; - - tgsi_full_token_init( &ctx->FullToken ); - - return TGSI_PARSE_OK; -} - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ) -{ - tgsi_full_token_free( &ctx->FullToken ); -} - -GLuint -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ) -{ - return ctx->Position >= - 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; -} - -static void -next_token( - struct tgsi_parse_context *ctx, - void *token ) -{ - assert( !tgsi_parse_end_of_tokens( ctx ) ); - - *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; -} - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ) -{ - struct tgsi_token token; - unsigned i; - - tgsi_full_token_free( &ctx->FullToken ); - tgsi_full_token_init( &ctx->FullToken ); - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; - - *decl = tgsi_default_full_declaration(); - decl->Declaration = *(struct tgsi_declaration *) &token; - - switch( decl->Declaration.Type ) { - case TGSI_DECLARE_RANGE: - next_token( ctx, &decl->u.DeclarationRange ); - break; - - case TGSI_DECLARE_MASK: - next_token( ctx, &decl->u.DeclarationMask ); - break; - - default: - assert (0); - } - - if( decl->Declaration.Interpolate ) { - next_token( ctx, &decl->Interpolation ); - } - - if( decl->Declaration.Semantic ) { - next_token( ctx, &decl->Semantic ); - } - - break; - } - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; - - *imm = tgsi_default_full_immediate(); - imm->Immediate = *(struct tgsi_immediate *) &token; - - assert( !imm->Immediate.Extended ); - - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - imm->u.Pointer = malloc( - sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - next_token( ctx, &imm->u.ImmediateFloat32[i] ); - } - break; - - default: - assert( 0 ); - } - - break; - } - - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; - unsigned extended; - - *inst = tgsi_default_full_instruction(); - inst->Instruction = *(struct tgsi_instruction *) &token; - - extended = inst->Instruction.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_INSTRUCTION_EXT_TYPE_NV: - inst->InstructionExtNv = - *(struct tgsi_instruction_ext_nv *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_LABEL: - inst->InstructionExtLabel = - *(struct tgsi_instruction_ext_label *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: - inst->InstructionExtTexture = - *(struct tgsi_instruction_ext_texture *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); - assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); - - extended = inst->FullDstRegisters[i].DstRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: - inst->FullDstRegisters[i].DstRegisterExtConcode = - *(struct tgsi_dst_register_ext_concode *) &token; - break; - - case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: - inst->FullDstRegisters[i].DstRegisterExtModulate = - *(struct tgsi_dst_register_ext_modulate *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - } - - assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); - - extended = inst->FullSrcRegisters[i].SrcRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: - inst->FullSrcRegisters[i].SrcRegisterExtSwz = - *(struct tgsi_src_register_ext_swz *) &token; - break; - - case TGSI_SRC_REGISTER_EXT_TYPE_MOD: - inst->FullSrcRegisters[i].SrcRegisterExtMod = - *(struct tgsi_src_register_ext_mod *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - - if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); - - /* - * No support for multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); - - if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - } - } - - break; - } - - default: - assert( 0 ); - } -} - diff --git a/src/mesa/pipe/tgsi/core/tgsi_parse.h b/src/mesa/pipe/tgsi/core/tgsi_parse.h deleted file mode 100644 index 56a8c254c2..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_parse.h +++ /dev/null @@ -1,121 +0,0 @@ -#if !defined TGSI_PARSE_H -#define TGSI_PARSE_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_full_version -{ - struct tgsi_version Version; -}; - -struct tgsi_full_header -{ - struct tgsi_header Header; - struct tgsi_processor Processor; -}; - -struct tgsi_full_dst_register -{ - struct tgsi_dst_register DstRegister; - struct tgsi_dst_register_ext_concode DstRegisterExtConcode; - struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; -}; - -struct tgsi_full_src_register -{ - struct tgsi_src_register SrcRegister; - struct tgsi_src_register_ext_swz SrcRegisterExtSwz; - struct tgsi_src_register_ext_mod SrcRegisterExtMod; - struct tgsi_src_register SrcRegisterInd; - struct tgsi_dimension SrcRegisterDim; - struct tgsi_src_register SrcRegisterDimInd; -}; - -struct tgsi_full_declaration -{ - struct tgsi_declaration Declaration; - union - { - struct tgsi_declaration_range DeclarationRange; - struct tgsi_declaration_mask DeclarationMask; - } u; - struct tgsi_declaration_interpolation Interpolation; - struct tgsi_declaration_semantic Semantic; -}; - -struct tgsi_full_immediate -{ - struct tgsi_immediate Immediate; - union - { - void *Pointer; - struct tgsi_immediate_float32 *ImmediateFloat32; - } u; -}; - -#define TGSI_FULL_MAX_DST_REGISTERS 2 -#define TGSI_FULL_MAX_SRC_REGISTERS 3 - -struct tgsi_full_instruction -{ - struct tgsi_instruction Instruction; - struct tgsi_instruction_ext_nv InstructionExtNv; - struct tgsi_instruction_ext_label InstructionExtLabel; - struct tgsi_instruction_ext_texture InstructionExtTexture; - struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; - struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; -}; - -union tgsi_full_token -{ - struct tgsi_token Token; - struct tgsi_full_declaration FullDeclaration; - struct tgsi_full_immediate FullImmediate; - struct tgsi_full_instruction FullInstruction; -}; - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ); - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ); - -struct tgsi_parse_context -{ - const struct tgsi_token *Tokens; - unsigned Position; - struct tgsi_full_version FullVersion; - struct tgsi_full_header FullHeader; - union tgsi_full_token FullToken; -}; - -#define TGSI_PARSE_OK 0 -#define TGSI_PARSE_ERROR 1 - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ); - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ); - -unsigned -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ); - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_PARSE_H - diff --git a/src/mesa/pipe/tgsi/core/tgsi_token.h b/src/mesa/pipe/tgsi/core/tgsi_token.h deleted file mode 100644 index ca53071a60..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_token.h +++ /dev/null @@ -1,1486 +0,0 @@ -#if !defined TGSI_TOKEN_H -#define TGSI_TOKEN_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_version -{ - unsigned MajorVersion : 8; - unsigned MinorVersion : 8; - unsigned Padding : 16; -}; - -struct tgsi_header -{ - unsigned HeaderSize : 8; - unsigned BodySize : 24; -}; - -#define TGSI_PROCESSOR_FRAGMENT 0 -#define TGSI_PROCESSOR_VERTEX 1 -#define TGSI_PROCESSOR_GEOMETRY 2 - -struct tgsi_processor -{ - unsigned Processor : 4; /* TGSI_PROCESSOR_ */ - unsigned Padding : 28; -}; - -#define TGSI_TOKEN_TYPE_DECLARATION 0 -#define TGSI_TOKEN_TYPE_IMMEDIATE 1 -#define TGSI_TOKEN_TYPE_INSTRUCTION 2 - -struct tgsi_token -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_ */ - unsigned Size : 8; /* UINT */ - unsigned Padding : 19; - unsigned Extended : 1; /* BOOL */ -}; - -#define TGSI_FILE_NULL 0 -#define TGSI_FILE_CONSTANT 1 -#define TGSI_FILE_INPUT 2 -#define TGSI_FILE_OUTPUT 3 -#define TGSI_FILE_TEMPORARY 4 -#define TGSI_FILE_SAMPLER 5 -#define TGSI_FILE_ADDRESS 6 -#define TGSI_FILE_IMMEDIATE 7 - -#define TGSI_DECLARE_RANGE 0 -#define TGSI_DECLARE_MASK 1 - -#define TGSI_WRITEMASK_NONE 0x00 -#define TGSI_WRITEMASK_X 0x01 -#define TGSI_WRITEMASK_Y 0x02 -#define TGSI_WRITEMASK_XY 0x03 -#define TGSI_WRITEMASK_Z 0x04 -#define TGSI_WRITEMASK_XZ 0x05 -#define TGSI_WRITEMASK_YZ 0x06 -#define TGSI_WRITEMASK_XYZ 0x07 -#define TGSI_WRITEMASK_W 0x08 -#define TGSI_WRITEMASK_XW 0x09 -#define TGSI_WRITEMASK_YW 0x0A -#define TGSI_WRITEMASK_XYW 0x0B -#define TGSI_WRITEMASK_ZW 0x0C -#define TGSI_WRITEMASK_XZW 0x0D -#define TGSI_WRITEMASK_YZW 0x0E -#define TGSI_WRITEMASK_XYZW 0x0F - -struct tgsi_declaration -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ - unsigned Size : 8; /* UINT */ - unsigned File : 4; /* TGSI_FILE_ */ - unsigned Declare : 4; /* TGSI_DECLARE_ */ - unsigned UsageMask : 4; /* TGSI_WRITEMASK_ */ - unsigned Interpolate : 1; /* BOOL */ - unsigned Semantic : 1; /* BOOL */ - unsigned Padding : 5; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_declaration_range -{ - unsigned First : 16; /* UINT */ - unsigned Last : 16; /* UINT */ -}; - -struct tgsi_declaration_mask -{ - unsigned Mask : 32; /* UINT */ -}; - -#define TGSI_INTERPOLATE_CONSTANT 0 -#define TGSI_INTERPOLATE_LINEAR 1 -#define TGSI_INTERPOLATE_PERSPECTIVE 2 - -struct tgsi_declaration_interpolation -{ - unsigned Interpolate : 4; /* TGSI_INTERPOLATE_ */ - unsigned Padding : 28; -}; - -#define TGSI_SEMANTIC_DEPTH 0 -#define TGSI_SEMANTIC_COLOR 1 - -struct tgsi_declaration_semantic -{ - unsigned SemanticName : 8; /* TGSI_SEMANTIC_ */ - unsigned SemanticIndex : 16; /* UINT */ - unsigned Padding : 8; -}; - -#define TGSI_IMM_FLOAT32 0 - -struct tgsi_immediate -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_IMMEDIATE */ - unsigned Size : 8; /* UINT */ - unsigned DataType : 4; /* TGSI_IMM_ */ - unsigned Padding : 15; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_immediate_float32 -{ - float Float; -}; - -/* - * GL_NV_vertex_program - */ -#define TGSI_OPCODE_ARL 0 -#define TGSI_OPCODE_MOV 1 -#define TGSI_OPCODE_LIT 2 -#define TGSI_OPCODE_RCP 3 -#define TGSI_OPCODE_RSQ 4 -#define TGSI_OPCODE_EXP 5 -#define TGSI_OPCODE_LOG 6 -#define TGSI_OPCODE_MUL 7 -#define TGSI_OPCODE_ADD 8 -#define TGSI_OPCODE_DP3 9 -#define TGSI_OPCODE_DP4 10 -#define TGSI_OPCODE_DST 11 -#define TGSI_OPCODE_MIN 12 -#define TGSI_OPCODE_MAX 13 -#define TGSI_OPCODE_SLT 14 -#define TGSI_OPCODE_SGE 15 -#define TGSI_OPCODE_MAD 16 - -/* - * GL_ATI_fragment_shader - */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_MUL */ -#define TGSI_OPCODE_SUB 17 -#define TGSI_OPCODE_DOT3 TGSI_OPCODE_DP3 -#define TGSI_OPCODE_DOT4 TGSI_OPCODE_DP4 -/* TGSI_OPCODE_MAD */ -#define TGSI_OPCODE_LERP 18 -#define TGSI_OPCODE_CND 19 -#define TGSI_OPCODE_CND0 20 -#define TGSI_OPCODE_DOT2ADD 21 - -/* - * GL_EXT_vertex_shader - */ -#define TGSI_OPCODE_INDEX 22 -#define TGSI_OPCODE_NEGATE 23 -/* TGSI_OPCODE_DOT3 */ -/* TGSI_OPCODE_DOT4 */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_MADD TGSI_OPCODE_MAD -#define TGSI_OPCODE_FRAC 24 -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -#define TGSI_OPCODE_SETGE TGSI_OPCODE_SGE -#define TGSI_OPCODE_SETLT TGSI_OPCODE_SLT -#define TGSI_OPCODE_CLAMP 25 -#define TGSI_OPCODE_FLOOR 26 -#define TGSI_OPCODE_ROUND 27 -#define TGSI_OPCODE_EXPBASE2 28 -#define TGSI_OPCODE_LOGBASE2 29 -#define TGSI_OPCODE_POWER 30 -#define TGSI_OPCODE_RECIP TGSI_OPCODE_RCP -#define TGSI_OPCODE_RECIPSQRT TGSI_OPCODE_RSQ -/* TGSI_OPCODE_SUB */ -#define TGSI_OPCODE_CROSSPRODUCT 31 -#define TGSI_OPCODE_MULTIPLYMATRIX 32 -/* TGSI_OPCODE_MOV */ - -/* - * GL_NV_vertex_program1_1 - */ -/* TGSI_OPCODE_ARL */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_LIT */ -#define TGSI_OPCODE_ABS 33 -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_LOG */ -#define TGSI_OPCODE_RCC 34 -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -#define TGSI_OPCODE_DPH 35 -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ - -/* - * GL_NV_fragment_program - */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_COS 36 -#define TGSI_OPCODE_DDX 37 -#define TGSI_OPCODE_DDY 38 -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DST */ -#define TGSI_OPCODE_EX2 TGSI_OPCODE_EXPBASE2 -#define TGSI_OPCODE_FLR TGSI_OPCODE_FLOOR -#define TGSI_OPCODE_FRC TGSI_OPCODE_FRAC -#define TGSI_OPCODE_KIL 39 -#define TGSI_OPCODE_LG2 TGSI_OPCODE_LOGBASE2 -/* TGSI_OPCODE_LIT */ -#define TGSI_OPCODE_LRP TGSI_OPCODE_LERP -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -#define TGSI_OPCODE_PK2H 40 -#define TGSI_OPCODE_PK2US 41 -#define TGSI_OPCODE_PK4B 42 -#define TGSI_OPCODE_PK4UB 43 -#define TGSI_OPCODE_POW TGSI_OPCODE_POWER -/* TGSI_OPCODE_RCP */ -#define TGSI_OPCODE_RFL 44 -/* TGSI_OPCODE_RSQ */ -#define TGSI_OPCODE_SEQ 45 -#define TGSI_OPCODE_SFL 46 -/* TGSI_OPCODE_SGE */ -#define TGSI_OPCODE_SGT 47 -#define TGSI_OPCODE_SIN 48 -#define TGSI_OPCODE_SLE 49 -/* TGSI_OPCODE_SLT */ -#define TGSI_OPCODE_SNE 50 -#define TGSI_OPCODE_STR 51 -/* TGSI_OPCODE_SUB */ -#define TGSI_OPCODE_TEX 52 -#define TGSI_OPCODE_TXD 53 -#define TGSI_OPCODE_TXP 132 -#define TGSI_OPCODE_UP2H 54 -#define TGSI_OPCODE_UP2US 55 -#define TGSI_OPCODE_UP4B 56 -#define TGSI_OPCODE_UP4UB 57 -#define TGSI_OPCODE_X2D 58 - -/* - * GL_NV_vertex_program2 - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_ARA 59 -/* TGSI_OPCODE_ARL */ -#define TGSI_OPCODE_ARR 60 -#define TGSI_OPCODE_BRA 61 -#define TGSI_OPCODE_CAL 62 -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCC */ -/* TGSI_OPCODE_RCP */ -#define TGSI_OPCODE_RET 63 -/* TGSI_OPCODE_RSQNV - use TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SNE */ -#define TGSI_OPCODE_SSG 64 -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_SUB */ - -/* - * GL_ARB_vertex_program - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_ARL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -#define TGSI_OPCODE_SWZ TGSI_OPCODE_MOV -#define TGSI_OPCODE_XPD TGSI_OPCODE_CROSSPRODUCT - -/* - * GL_ARB_fragment_program - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_CMP 65 -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -#define TGSI_OPCODE_SCS 66 -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXP */ -#define TGSI_OPCODE_TXB 67 -/* TGSI_OPCODE_KIL */ - -/* - * GL_NV_fragment_program_option - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_DDX */ -/* TGSI_OPCODE_DDY */ -/* TGSI_OPCODE_PK2H */ -/* TGSI_OPCODE_PK2US */ -/* TGSI_OPCODE_PK4B */ -/* TGSI_OPCODE_PK4UB */ -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SCS */ -/* TGSI_OPCODE_UP2H */ -/* TGSI_OPCODE_UP2US */ -/* TGSI_OPCODE_UP4B */ -/* TGSI_OPCODE_UP4UB */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_RFL */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_CMP */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_X2D */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXP */ -/* TGSI_OPCODE_TXB */ -/* TGSI_OPCODE_KIL */ -/* TGSI_OPCODE_TXD */ - -/* - * GL_NV_fragment_program2 - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_DDX */ -/* TGSI_OPCODE_DDY */ -/* TGSI_OPCODE_PK2H */ -/* TGSI_OPCODE_PK2US */ -/* TGSI_OPCODE_PK4B */ -/* TGSI_OPCODE_PK4UB */ -#define TGSI_OPCODE_NRM 68 -#define TGSI_OPCODE_DIV 69 -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SCS */ -/* TGSI_OPCODE_UP2H */ -/* TGSI_OPCODE_UP2US */ -/* TGSI_OPCODE_UP4B */ -/* TGSI_OPCODE_UP4UB */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_RFL */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -#define TGSI_OPCODE_DP2 70 -/* TGSI_OPCODE_CMP */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_X2D */ -#define TGSI_OPCODE_DP2A TGSI_OPCODE_DOT2ADD -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXP */ -/* TGSI_OPCODE_TXB */ -#define TGSI_OPCODE_TXL 71 -/* TGSI_OPCODE_KIL */ -/* TGSI_OPCODE_TXD */ -/* TGSI_OPCODE_CAL */ -/* TGSI_OPCODE_RET */ -#define TGSI_OPCODE_BRK 72 -#define TGSI_OPCODE_IF 73 -#define TGSI_OPCODE_LOOP 74 -#define TGSI_OPCODE_REP 75 -#define TGSI_OPCODE_ELSE 76 -#define TGSI_OPCODE_ENDIF 77 -#define TGSI_OPCODE_ENDLOOP 78 -#define TGSI_OPCODE_ENDREP 79 - -/* - * GL_NV_vertex_program2_option - */ -/* TGSI_OPCODE_ARL */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_SSG */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_RCC */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_ARR */ -/* TGSI_OPCODE_ARA */ -/* TGSI_OPCODE_BRA */ -/* TGSI_OPCODE_CAL */ -/* TGSI_OPCODE_RET */ - -/* - * GL_NV_vertex_program3 - */ -/* TGSI_OPCODE_ARL */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_SSG */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_RCC */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_ARR */ -/* TGSI_OPCODE_ARA */ -/* TGSI_OPCODE_BRA */ -/* TGSI_OPCODE_CAL */ -/* TGSI_OPCODE_RET */ -#define TGSI_OPCODE_PUSHA 80 -#define TGSI_OPCODE_POPA 81 -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXP */ -/* TGSI_OPCODE_TXB */ -/* TGSI_OPCODE_TXL */ - -/* - * GL_NV_gpu_program4 - */ -/* TGSI_OPCODE_ABS */ -#define TGSI_OPCODE_CEIL 82 -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -#define TGSI_OPCODE_I2F 83 -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -#define TGSI_OPCODE_NOT 84 -/* TGSI_OPCODE_NRM */ -/* TGSI_OPCODE_PK2H */ -/* TGSI_OPCODE_PK2US */ -/* TGSI_OPCODE_PK4B */ -/* TGSI_OPCODE_PK4UB */ -/* TGSI_OPCODE_ROUND */ -/* TGSI_OPCODE_SSG */ -#define TGSI_OPCODE_TRUNC 85 -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_RCC */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SCS */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_UP2H */ -/* TGSI_OPCODE_UP2US */ -/* TGSI_OPCODE_UP4B */ -/* TGSI_OPCODE_UP4UB */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_DIV */ -#define TGSI_OPCODE_SHL 86 -#define TGSI_OPCODE_SHR 87 -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_AND 88 -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -#define TGSI_OPCODE_OR 89 -/* TGSI_OPCODE_RFL */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_DP2 */ -#define TGSI_OPCODE_MOD 90 -#define TGSI_OPCODE_XOR 91 -/* TGSI_OPCODE_CMP */ -/* TGSI_OPCODE_DP2A */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_MAD */ -#define TGSI_OPCODE_SAD 92 -/* TGSI_OPCODE_X2D */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXB */ -#define TGSI_OPCODE_TXF 93 -/* TGSI_OPCODE_TXL */ -/* TGSI_OPCODE_TXP */ -#define TGSI_OPCODE_TXQ 94 -/* TGSI_OPCODE_TXD */ -/* TGSI_OPCODE_CAL */ -/* TGSI_OPCODE_RET */ -/* TGSI_OPCODE_BRK */ -#define TGSI_OPCODE_CONT 95 -/* TGSI_OPCODE_IF */ -/* TGSI_OPCODE_REP */ -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -/* TGSI_OPCODE_ENDREP */ - -/* - * GL_NV_vertex_program4 - */ -/* Same as GL_NV_gpu_program4 */ - -/* - * GL_NV_fragment_program4 - */ -/* Same as GL_NV_gpu_program4 */ -/* TGSI_OPCODE_KIL */ -/* TGSI_OPCODE_DDX */ -/* TGSI_OPCODE_DDY */ - -/* - * GL_NV_geometry_program4 - */ -/* Same as GL_NV_gpu_program4 */ -#define TGSI_OPCODE_EMIT 96 -#define TGSI_OPCODE_ENDPRIM 97 - -/* - * GLSL - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_BGNLOOP2 98 -#define TGSI_OPCODE_BGNSUB 99 -/* TGSI_OPCODE_BRA */ -/* TGSI_OPCODE_BRK */ -/* TGSI_OPCODE_CONT */ -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_DDX */ -/* TGSI_OPCODE_DDY */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -#define TGSI_OPCODE_ENDLOOP2 100 -#define TGSI_OPCODE_ENDSUB 101 -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_IF */ -#define TGSI_OPCODE_INT TGSI_OPCODE_TRUNC -/* TGSI_OPCODE_KIL */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -#define TGSI_OPCODE_NOISE1 102 -#define TGSI_OPCODE_NOISE2 103 -#define TGSI_OPCODE_NOISE3 104 -#define TGSI_OPCODE_NOISE4 105 -#define TGSI_OPCODE_NOP 106 -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXB */ -/* TGSI_OPCODE_TXD */ -/* TGSI_OPCODE_TXL */ -/* TGSI_OPCODE_TXP */ -/* TGSI_OPCODE_XPD */ - -/* - * ps_1_1 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_LRP */ -#define TGSI_OPCODE_TEXCOORD TGSI_OPCODE_NOP -#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL -/* TGSI_OPCODE_TEX */ -#define TGSI_OPCODE_TEXBEM 107 -#define TGSI_OPCODE_TEXBEML 108 -#define TGSI_OPCODE_TEXREG2AR 109 -#define TGSI_OPCODE_TEXM3X2PAD 110 -#define TGSI_OPCODE_TEXM3X2TEX 111 -#define TGSI_OPCODE_TEXM3X3PAD 112 -#define TGSI_OPCODE_TEXM3X3TEX 113 -#define TGSI_OPCODE_TEXM3X3SPEC 114 -#define TGSI_OPCODE_TEXM3X3VSPEC 115 -/* TGSI_OPCODE_CND */ - -/* - * ps_1_2 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_TEXCOORD */ -/* TGSI_OPCODE_TEXKILL */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TEXBEM */ -/* TGSI_OPCODE_TEXBEML */ -/* TGSI_OPCODE_TEXREG2AR */ -#define TGSI_OPCODE_TEXREG2GB 116 -/* TGSI_OPCODE_TEXM3X2PAD */ -/* TGSI_OPCODE_TEXM3X2TEX */ -/* TGSI_OPCODE_TEXM3X3PAD */ -/* TGSI_OPCODE_TEXM3X3TEX */ -/* TGSI_OPCODE_TEXM3X3SPEC */ -/* TGSI_OPCODE_TEXM3X3VSPEC */ -/* TGSI_OPCODE_CND */ -#define TGSI_OPCODE_TEXREG2RGB 117 -#define TGSI_OPCODE_TEXDP3TEX 118 -#define TGSI_OPCODE_TEXDP3 119 -#define TGSI_OPCODE_TEXM3X3 120 -/* CMP - use TGSI_OPCODE_CND0 */ - -/* - * ps_1_3 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_TEXCOORD */ -/* TGSI_OPCODE_TEXKILL */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TEXBEM */ -/* TGSI_OPCODE_TEXBEML */ -/* TGSI_OPCODE_TEXREG2AR */ -/* TGSI_OPCODE_TEXREG2GB */ -/* TGSI_OPCODE_TEXM3X2PAD */ -/* TGSI_OPCODE_TEXM3X2TEX */ -/* TGSI_OPCODE_TEXM3X3PAD */ -/* TGSI_OPCODE_TEXM3X3TEX */ -/* TGSI_OPCODE_TEXM3X3SPEC */ -/* TGSI_OPCODE_TEXM3X3VSPEC */ -/* TGSI_OPCODE_CND */ -/* TGSI_OPCODE_TEXREG2RGB */ -/* TGSI_OPCODE_TEXDP3TEX */ -#define TGSI_OPCODE_TEXM3X2DEPTH 121 -/* TGSI_OPCODE_TEXDP3 */ -/* TGSI_OPCODE_TEXM3X3 */ -/* CMP - use TGSI_OPCODE_CND0 */ - -/* - * ps_1_4 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_LRP */ -#define TGSI_OPCODE_TEXCRD TGSI_OPCODE_TEXCOORD -/* TGSI_OPCODE_TEXKILL */ -#define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX -/* TGSI_OPCODE_CND */ -#define TGSI_OPCODE_TEXDEPTH 122 -/* CMP - use TGSI_OPCODE_CND0 */ -#define TGSI_OPCODE_BEM 123 - -/* - * ps_2_0 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_FRC */ -#define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX -#define TGSI_OPCODE_M4X3 124 -#define TGSI_OPCODE_M3X4 125 -#define TGSI_OPCODE_M3X3 126 -#define TGSI_OPCODE_M3X2 127 -/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ -#define TGSI_OPCODE_CRS TGSI_OPCODE_XPD -/* TGSI_OPCODE_ABS */ -#define TGSI_OPCODE_NRM4 128 -#define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS -/* TGSI_OPCODE_TEXKILL */ -/* TGSI_OPCODE_TEXLD */ -#define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB -#define TGSI_OPCODE_TEXLDP TGSI_OPCODE_TXP -/* CMP - use TGSI_OPCODE_CND0 */ -#define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A - -/* - * ps_2_x - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SNE */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_M4X4 */ -/* TGSI_OPCODE_M4X3 */ -/* TGSI_OPCODE_M3X4 */ -/* TGSI_OPCODE_M3X3 */ -/* TGSI_OPCODE_M3X2 */ -#define TGSI_OPCODE_CALL TGSI_OPCODE_CAL -#define TGSI_OPCODE_CALLNZ 129 -/* TGSI_OPCODE_RET */ -/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ -/* TGSI_OPCODE_CRS */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_NRM4 */ -/* TGSI_OPCODE_SINCOS */ -/* TGSI_OPCODE_REP */ -/* TGSI_OPCODE_ENDREP */ -/* TGSI_OPCODE_IF */ -#define TGSI_OPCODE_IFC 130 -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -#define TGSI_OPCODE_BREAK TGSI_OPCODE_BRK -#define TGSI_OPCODE_BREAKC 131 -/* TGSI_OPCODE_TEXKILL */ -/* TGSI_OPCODE_TEXLD */ -/* TGSI_OPCODE_TEXLDB */ -/* CMP - use TGSI_OPCODE_CND0 */ -/* TGSI_OPCODE_DP2ADD */ -#define TGSI_OPCODE_DSX TGSI_OPCODE_DDX -#define TGSI_OPCODE_DSY TGSI_OPCODE_DDY -#define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD -/* TGSI_OPCODE_TEXLDP */ - -/* - * vs_1_1 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_M4X4 */ -/* TGSI_OPCODE_M4X3 */ -/* TGSI_OPCODE_M3X4 */ -/* TGSI_OPCODE_M3X3 */ -/* TGSI_OPCODE_M3X2 */ -#define TGSI_OPCODE_EXPP TGSI_OPCODE_EXP -#define TGSI_OPCODE_LOGP TGSI_OPCODE_LG2 - -/* - * vs_2_0 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_M4X4 */ -/* TGSI_OPCODE_M4X3 */ -/* TGSI_OPCODE_M3X4 */ -/* TGSI_OPCODE_M3X3 */ -/* TGSI_OPCODE_M3X2 */ -/* TGSI_OPCODE_CALL */ -/* TGSI_OPCODE_CALLNZ */ -/* TGSI_OPCODE_LOOP */ -/* TGSI_OPCODE_RET */ -/* TGSI_OPCODE_ENDLOOP */ -/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ -/* TGSI_OPCODE_CRS */ -#define TGSI_OPCODE_SGN TGSI_OPCODE_SSG -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_NRM4 */ -/* TGSI_OPCODE_SINCOS */ -/* TGSI_OPCODE_REP */ -/* TGSI_OPCODE_ENDREP */ -/* TGSI_OPCODE_IF */ -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -#define TGSI_OPCODE_MOVA TGSI_OPCODE_ARR -/* TGSI_OPCODE_LOGP */ - -/* - * vs_2_x - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SNE */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_M4X4 */ -/* TGSI_OPCODE_M4X3 */ -/* TGSI_OPCODE_M3X4 */ -/* TGSI_OPCODE_M3X3 */ -/* TGSI_OPCODE_M3X2 */ -/* TGSI_OPCODE_CALL */ -/* TGSI_OPCODE_CALLNZ */ -/* TGSI_OPCODE_LOOP */ -/* TGSI_OPCODE_RET */ -/* TGSI_OPCODE_ENDLOOP */ -/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ -/* TGSI_OPCODE_CRS */ -/* TGSI_OPCODE_SGN */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_NRM4 */ -/* TGSI_OPCODE_SINCOS */ -/* TGSI_OPCODE_REP */ -/* TGSI_OPCODE_ENDREP */ -/* TGSI_OPCODE_IF */ -/* TGSI_OPCODE_IFC */ -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -/* TGSI_OPCODE_BREAK */ -/* TGSI_OPCODE_BREAKC */ -/* TGSI_OPCODE_MOVA */ -/* TGSI_OPCODE_LOGP */ - -#define TGSI_OPCODE_LAST 133 - -#define TGSI_SAT_NONE 0 /* do not saturate */ -#define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ -#define TGSI_SAT_MINUS_PLUS_ONE 2 /* clamp to [-1,1] */ - -/* - * Opcode is the operation code to execute. A given operation defines the - * semantics how the source registers (if any) are interpreted and what is - * written to the destination registers (if any) as a result of execution. - * - * NumDstRegs and NumSrcRegs is the number of destination and source registers, - * respectively. For a given operation code, those numbers are fixed and are - * present here only for convenience. - * - * If Extended is TRUE, it is now executed. - * - * Saturate controls how are final results in destination registers modified. - */ - -struct tgsi_instruction -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_INSTRUCTION */ - unsigned Size : 8; /* UINT */ - unsigned Opcode : 8; /* TGSI_OPCODE_ */ - unsigned Saturate : 2; /* TGSI_SAT_ */ - unsigned NumDstRegs : 2; /* UINT */ - unsigned NumSrcRegs : 4; /* UINT */ - unsigned Padding : 3; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows. - * - * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow. - * - * Then, tgsi_instruction::NumSrcRegs of tgsi_src_register follow. - * - * tgsi_instruction::Size contains the total number of words that make the - * instruction, including the instruction word. - */ - -#define TGSI_INSTRUCTION_EXT_TYPE_NV 0 -#define TGSI_INSTRUCTION_EXT_TYPE_LABEL 1 -#define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE 2 -#define TGSI_INSTRUCTION_EXT_TYPE_PREDICATE 3 - -struct tgsi_instruction_ext -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_ */ - unsigned Padding : 27; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_NV, it should - * be cast to tgsi_instruction_ext_nv. - * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_LABEL, it - * should be cast to tgsi_instruction_ext_label. - * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it - * should be cast to tgsi_instruction_ext_texture. - * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_PREDICATE, it - * should be cast to tgsi_instruction_ext_predicate. - * - * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext - * follows. - */ - -#define TGSI_PRECISION_DEFAULT 0 -#define TGSI_PRECISION_FLOAT32 1 -#define TGSI_PRECISION_FLOAT16 2 -#define TGSI_PRECISION_FIXED12 3 - -#define TGSI_CC_GT 0 -#define TGSI_CC_EQ 1 -#define TGSI_CC_LT 2 -#define TGSI_CC_UN 3 -#define TGSI_CC_GE 4 -#define TGSI_CC_LE 5 -#define TGSI_CC_NE 6 -#define TGSI_CC_TR 7 -#define TGSI_CC_FL 8 - -#define TGSI_SWIZZLE_X 0 -#define TGSI_SWIZZLE_Y 1 -#define TGSI_SWIZZLE_Z 2 -#define TGSI_SWIZZLE_W 3 - -/* - * Precision controls the precision at which the operation should be executed. - * - * CondDstUpdate enables condition code register writes. When this field is - * TRUE, CondDstIndex specifies the index of the condition code register to - * update. - * - * CondFlowEnable enables conditional execution of the operation. When this - * field is TRUE, CondFlowIndex specifies the index of the condition code - * register to test against CondMask with component swizzle controled by - * CondSwizzleX, CondSwizzleY, CondSwizzleZ and CondSwizzleW. If the test fails, - * the operation is not executed. - */ - -struct tgsi_instruction_ext_nv -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_NV */ - unsigned Precision : 4; /* TGSI_PRECISION_ */ - unsigned CondDstIndex : 4; /* UINT */ - unsigned CondFlowIndex : 4; /* UINT */ - unsigned CondMask : 4; /* TGSI_CC_ */ - unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned CondDstUpdate : 1; /* BOOL */ - unsigned CondFlowEnable : 1; /* BOOL */ - unsigned Padding : 1; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_instruction_ext_label -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ - unsigned Label : 24; /* UINT */ - unsigned Padding : 3; - unsigned Extended : 1; /* BOOL */ -}; - -#define TGSI_TEXTURE_UNKNOWN 0 -#define TGSI_TEXTURE_1D 1 -#define TGSI_TEXTURE_2D 2 -#define TGSI_TEXTURE_3D 3 -#define TGSI_TEXTURE_CUBE 4 -#define TGSI_TEXTURE_RECT 5 -#define TGSI_TEXTURE_SHADOW1D 6 -#define TGSI_TEXTURE_SHADOW2D 7 -#define TGSI_TEXTURE_SHADOWRECT 8 - -struct tgsi_instruction_ext_texture -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */ - unsigned Texture : 8; /* TGSI_TEXTURE_ */ - unsigned Padding : 19; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_instruction_ext_predicate -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ - unsigned PredDstIndex : 4; /* UINT */ - unsigned PredWriteMask : 4; /* TGSI_WRITEMASK_ */ - unsigned Padding : 19; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * File specifies the register array to access. - * - * Index specifies the element number of a register in the register file. - * - * If Indirect is TRUE, Index should be offset by the X component of a source - * register that follows. The register can be now fetched into local storage - * for further processing. - * - * If Negate is TRUE, all components of the fetched register are negated. - * - * The fetched register components are swizzled according to SwizzleX, SwizzleY, - * SwizzleZ and SwizzleW. - * - * If Extended is TRUE, any further modifications to the source register are - * made to this temporary storage. - */ - -struct tgsi_src_register -{ - unsigned File : 4; /* TGSI_FILE_ */ - unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned Negate : 1; /* BOOL */ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - int Index : 16; /* SINT */ - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_src_register::Extended is TRUE, tgsi_src_register_ext follows. - * - * Then, if tgsi_src_register::Indirect is TRUE, another tgsi_src_register - * follows. - * - * Then, if tgsi_src_register::Dimension is TRUE, tgsi_dimension follows. - */ - -#define TGSI_SRC_REGISTER_EXT_TYPE_SWZ 0 -#define TGSI_SRC_REGISTER_EXT_TYPE_MOD 1 - -struct tgsi_src_register_ext -{ - unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_ */ - unsigned Padding : 27; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_SWZ, - * it should be cast to tgsi_src_register_ext_extswz. - * - * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_MOD, - * it should be cast to tgsi_src_register_ext_mod. - * - * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext - * follows. - */ - -#define TGSI_EXTSWIZZLE_X TGSI_SWIZZLE_X -#define TGSI_EXTSWIZZLE_Y TGSI_SWIZZLE_Y -#define TGSI_EXTSWIZZLE_Z TGSI_SWIZZLE_Z -#define TGSI_EXTSWIZZLE_W TGSI_SWIZZLE_W -#define TGSI_EXTSWIZZLE_ZERO 4 -#define TGSI_EXTSWIZZLE_ONE 5 - -/* - * ExtSwizzleX, ExtSwizzleY, ExtSwizzleZ and ExtSwizzleW swizzle the source - * register in an extended manner. - * - * NegateX, NegateY, NegateZ and NegateW negate individual components of the - * source register. - * - * ExtDivide specifies which component is used to divide all components of the - * source register. - */ - -struct tgsi_src_register_ext_swz -{ - unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */ - unsigned ExtSwizzleX : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned ExtSwizzleY : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned ExtSwizzleZ : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned ExtSwizzleW : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned NegateX : 1; /* BOOL */ - unsigned NegateY : 1; /* BOOL */ - unsigned NegateZ : 1; /* BOOL */ - unsigned NegateW : 1; /* BOOL */ - unsigned ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned Padding : 3; - unsigned Extended : 1; /* BOOL */ -}; - -/** - * Extra src register modifiers - * - * If Complement is TRUE, the source register is modified by subtracting it - * from 1.0. - * - * If Bias is TRUE, the source register is modified by subtracting 0.5 from it. - * - * If Scale2X is TRUE, the source register is modified by multiplying it by 2.0. - * - * If Absolute is TRUE, the source register is modified by removing the sign. - * - * If Negate is TRUE, the source register is modified by negating it. - */ - -struct tgsi_src_register_ext_mod -{ - unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */ - unsigned Complement : 1; /* BOOL */ - unsigned Bias : 1; /* BOOL */ - unsigned Scale2X : 1; /* BOOL */ - unsigned Absolute : 1; /* BOOL */ - unsigned Negate : 1; /* BOOL */ - unsigned Padding : 22; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_dimension -{ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - unsigned Padding : 13; - int Index : 16; /* SINT */ - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_dst_register -{ - unsigned File : 4; /* TGSI_FILE_ */ - unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - int Index : 16; /* SINT */ - unsigned Padding : 5; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_dst_register::Extended is TRUE, tgsi_dst_register_ext follows. - * - * Then, if tgsi_dst_register::Indirect is TRUE, tgsi_src_register follows. - */ - -#define TGSI_DST_REGISTER_EXT_TYPE_CONDCODE 0 -#define TGSI_DST_REGISTER_EXT_TYPE_MODULATE 1 -#define TGSI_DST_REGISTER_EXT_TYPE_PREDICATE 2 - -struct tgsi_dst_register_ext -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_ */ - unsigned Padding : 27; - unsigned Extended : 1; /* BOOL */ -}; - -/** - * Extra destination register modifiers - * - * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_CONDCODE, - * it should be cast to tgsi_dst_register_ext_condcode. - * - * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_MODULATE, - * it should be cast to tgsi_dst_register_ext_modulate. - * - * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_PREDICATE, - * it should be cast to tgsi_dst_register_ext_predicate. - * - * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext - * follows. - */ -struct tgsi_dst_register_ext_concode -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ - unsigned CondMask : 4; /* TGSI_CC_ */ - unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSrcIndex : 4; /* UINT */ - unsigned Padding : 11; - unsigned Extended : 1; /* BOOL */ -}; - -#define TGSI_MODULATE_1X 0 -#define TGSI_MODULATE_2X 1 -#define TGSI_MODULATE_4X 2 -#define TGSI_MODULATE_8X 3 -#define TGSI_MODULATE_HALF 4 -#define TGSI_MODULATE_QUARTER 5 -#define TGSI_MODULATE_EIGHTH 6 - -struct tgsi_dst_register_ext_modulate -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */ - unsigned Modulate : 4; /* TGSI_MODULATE_ */ - unsigned Padding : 23; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * Currently, the following constraints apply. - * - * - PredSwizzleXYZW is either set to identity or replicate. - * - PredSrcIndex is 0. - */ - -struct tgsi_dst_register_ext_predicate -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */ - unsigned PredSwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSrcIndex : 4; /* UINT */ - unsigned Negate : 1; /* BOOL */ - unsigned Padding : 14; - unsigned Extended : 1; /* BOOL */ -}; - - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_TOKEN_H - diff --git a/src/mesa/pipe/tgsi/core/tgsi_util.c b/src/mesa/pipe/tgsi/core/tgsi_util.c deleted file mode 100644 index 38d6d6e6bc..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_util.c +++ /dev/null @@ -1,270 +0,0 @@ -#include "tgsi_platform.h" -#include "tgsi_core.h" - -union pointer_hack -{ - void *pointer; - unsigned long long uint64; -}; - -void * -tgsi_align_128bit( - void *unaligned ) -{ - union pointer_hack ph; - - ph.uint64 = 0; - ph.pointer = unaligned; - ph.uint64 = (ph.uint64 + 15) & ~15; - return ph.pointer; -} - -GLuint -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - GLuint component ) -{ - switch( component ) { - case 0: - return reg->SwizzleX; - case 1: - return reg->SwizzleY; - case 2: - return reg->SwizzleZ; - case 3: - return reg->SwizzleW; - default: - assert( 0 ); - } - return 0; -} - -GLuint -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - GLuint component ) -{ - switch( component ) { - case 0: - return reg->ExtSwizzleX; - case 1: - return reg->ExtSwizzleY; - case 2: - return reg->ExtSwizzleZ; - case 3: - return reg->ExtSwizzleW; - default: - assert( 0 ); - } - return 0; -} - -GLuint -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - GLuint component ) -{ - GLuint swizzle; - - /* - * First, calculate the extended swizzle for a given channel. This will give - * us either a channel index into the simple swizzle or a constant 1 or 0. - */ - swizzle = tgsi_util_get_src_register_extswizzle( - ®->SrcRegisterExtSwz, - component ); - - assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); - assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); - assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); - assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); - - /* - * Second, calculate the simple swizzle for the unswizzled channel index. - * Leave the constants intact, they are not affected by the simple swizzle. - */ - if( swizzle <= TGSI_SWIZZLE_W ) { - swizzle = tgsi_util_get_src_register_swizzle( - ®->SrcRegister, - swizzle ); - } - - return swizzle; -} - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - GLuint swizzle, - GLuint component ) -{ - switch( component ) { - case 0: - reg->SwizzleX = swizzle; - break; - case 1: - reg->SwizzleY = swizzle; - break; - case 2: - reg->SwizzleZ = swizzle; - break; - case 3: - reg->SwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - GLuint swizzle, - GLuint component ) -{ - switch( component ) { - case 0: - reg->ExtSwizzleX = swizzle; - break; - case 1: - reg->ExtSwizzleY = swizzle; - break; - case 2: - reg->ExtSwizzleZ = swizzle; - break; - case 3: - reg->ExtSwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -GLuint -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - GLuint component ) -{ - switch( component ) { - case 0: - return reg->NegateX; - case 1: - return reg->NegateY; - case 2: - return reg->NegateZ; - case 3: - return reg->NegateW; - default: - assert( 0 ); - } - return 0; -} - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - GLuint negate, - GLuint component ) -{ - switch( component ) { - case 0: - reg->NegateX = negate; - break; - case 1: - reg->NegateY = negate; - break; - case 2: - reg->NegateZ = negate; - break; - case 3: - reg->NegateW = negate; - break; - default: - assert( 0 ); - } -} - -GLuint -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - GLuint component ) -{ - GLuint sign_mode; - - if( reg->SrcRegisterExtMod.Absolute ) { - /* Consider only the post-abs negation. */ - - if( reg->SrcRegisterExtMod.Negate ) { - sign_mode = TGSI_UTIL_SIGN_SET; - } - else { - sign_mode = TGSI_UTIL_SIGN_CLEAR; - } - } - else { - /* Accumulate the three negations. */ - - GLuint negate; - - negate = reg->SrcRegister.Negate; - if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { - negate = !negate; - } - if( reg->SrcRegisterExtMod.Negate ) { - negate = !negate; - } - - if( negate ) { - sign_mode = TGSI_UTIL_SIGN_TOGGLE; - } - else { - sign_mode = TGSI_UTIL_SIGN_KEEP; - } - } - - return sign_mode; -} - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - GLuint sign_mode ) -{ - reg->SrcRegisterExtSwz.NegateX = 0; - reg->SrcRegisterExtSwz.NegateY = 0; - reg->SrcRegisterExtSwz.NegateZ = 0; - reg->SrcRegisterExtSwz.NegateW = 0; - - switch (sign_mode) - { - case TGSI_UTIL_SIGN_CLEAR: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_SET: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 1; - break; - - case TGSI_UTIL_SIGN_TOGGLE: - reg->SrcRegister.Negate = 1; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_KEEP: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - default: - assert( 0 ); - } -} - diff --git a/src/mesa/pipe/tgsi/core/tgsi_util.h b/src/mesa/pipe/tgsi/core/tgsi_util.h deleted file mode 100644 index ef14446f0e..0000000000 --- a/src/mesa/pipe/tgsi/core/tgsi_util.h +++ /dev/null @@ -1,70 +0,0 @@ -#if !defined TGSI_UTIL_H -#define TGSI_UTIL_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -void * -tgsi_align_128bit( - void *unaligned ); - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component); - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - unsigned swizzle, - unsigned component ); - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - unsigned swizzle, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ); - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ); - -#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ -#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ -#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ -#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - unsigned sign_mode ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_UTIL_H - diff --git a/src/mesa/pipe/tgsi/exec/Makefile b/src/mesa/pipe/tgsi/exec/Makefile new file mode 100644 index 0000000000..eb8b14e0e8 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/Makefile @@ -0,0 +1,3 @@ +default: + cd ../../.. ; make + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_attribs.h b/src/mesa/pipe/tgsi/exec/tgsi_attribs.h new file mode 100644 index 0000000000..b8f6ba7da8 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_attribs.h @@ -0,0 +1,47 @@ +#ifndef TGSI_ATTRIBS_H +#define TGSI_ATTRIBS_H + + +/** + * The specific values here are not important (could remove them). + */ +enum { + TGSI_ATTRIB_POS = 0, + TGSI_ATTRIB_WEIGHT = 1, + TGSI_ATTRIB_NORMAL = 2, + TGSI_ATTRIB_COLOR0 = 3, + TGSI_ATTRIB_COLOR1 = 4, + TGSI_ATTRIB_FOG = 5, + TGSI_ATTRIB_COLOR_INDEX = 6, /* XXX omit? */ + TGSI_ATTRIB_EDGEFLAG = 7, + TGSI_ATTRIB_TEX0 = 8, + TGSI_ATTRIB_TEX1 = 9, + TGSI_ATTRIB_TEX2 = 10, + TGSI_ATTRIB_TEX3 = 11, + TGSI_ATTRIB_TEX4 = 12, + TGSI_ATTRIB_TEX5 = 13, + TGSI_ATTRIB_TEX6 = 14, + TGSI_ATTRIB_TEX7 = 15, + TGSI_ATTRIB_VAR0 = 16, + TGSI_ATTRIB_VAR1 = 17, + TGSI_ATTRIB_VAR2 = 18, + TGSI_ATTRIB_VAR3 = 19, + TGSI_ATTRIB_VAR4 = 20, + TGSI_ATTRIB_VAR5 = 21, + TGSI_ATTRIB_VAR6 = 22, + TGSI_ATTRIB_VAR7 = 23, + TGSI_ATTRIB_POINTSIZE = 24, + TGSI_ATTRIB_BFC0 = 25, + TGSI_ATTRIB_BFC1 = 26, + TGSI_ATTRIB_CLIP_POS = 27, + TGSI_ATTRIB_VERTEX_HEADER = 28, + TGSI_ATTRIB_MAX = 29 +}; + + +#define TGSI_MAX_TEXTURE 8 +#define TGSI_MAX_VARYING 8 + + + +#endif /* TGSI_ATTRIBS_H */ diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.c b/src/mesa/pipe/tgsi/exec/tgsi_build.c new file mode 100644 index 0000000000..d6f8af656a --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_build.c @@ -0,0 +1,1368 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ) +{ + struct tgsi_version version; + + version.MajorVersion = 1; + version.MinorVersion = 1; + version.Padding = 0; + + return version; +} + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ) +{ + struct tgsi_header header; + + header.HeaderSize = 1; + header.BodySize = 0; + + return header; +} + +static void +header_headersize_grow( struct tgsi_header *header ) +{ + assert( header->HeaderSize < 0xFF ); + assert( header->BodySize == 0 ); + + header->HeaderSize++; +} + +static void +header_bodysize_grow( struct tgsi_header *header ) +{ + assert( header->BodySize < 0xFFFFFF ); + + header->BodySize++; +} + +struct tgsi_processor +tgsi_default_processor( void ) +{ + struct tgsi_processor processor; + + processor.Processor = TGSI_PROCESSOR_FRAGMENT; + processor.Padding = 0; + + return processor; +} + +struct tgsi_processor +tgsi_build_processor( + unsigned type, + struct tgsi_header *header ) +{ + struct tgsi_processor processor; + + processor = tgsi_default_processor(); + processor.Processor = type; + + header_headersize_grow( header ); + + return processor; +} + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ) +{ + struct tgsi_declaration declaration; + + declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; + declaration.Size = 1; + declaration.File = TGSI_FILE_NULL; + declaration.Declare = TGSI_DECLARE_RANGE; + declaration.UsageMask = TGSI_WRITEMASK_XYZW; + declaration.Interpolate = 0; + declaration.Semantic = 0; + declaration.Padding = 0; + declaration.Extended = 0; + + return declaration; +} + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned declare, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ) +{ + struct tgsi_declaration declaration; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( declare <= TGSI_DECLARE_MASK ); + + declaration = tgsi_default_declaration(); + declaration.File = file; + declaration.Declare = declare; + declaration.UsageMask = usage_mask; + declaration.Interpolate = interpolate; + declaration.Semantic = semantic; + + header_bodysize_grow( header ); + + return declaration; +} + +static void +declaration_grow( + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + assert( declaration->Size < 0xFF ); + + declaration->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ) +{ + struct tgsi_full_declaration full_declaration; + + full_declaration.Declaration = tgsi_default_declaration(); + full_declaration.Interpolation = tgsi_default_declaration_interpolation(); + full_declaration.Semantic = tgsi_default_declaration_semantic(); + + return full_declaration; +} + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0; + struct tgsi_declaration *declaration; + + if( maxsize <= size ) + return 0; + declaration = (struct tgsi_declaration *) &tokens[size]; + size++; + + *declaration = tgsi_build_declaration( + full_decl->Declaration.File, + full_decl->Declaration.Declare, + full_decl->Declaration.UsageMask, + full_decl->Declaration.Interpolate, + full_decl->Declaration.Semantic, + header ); + + switch( full_decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + { + struct tgsi_declaration_range *dr; + + if( maxsize <= size ) + return 0; + dr = (struct tgsi_declaration_range *) &tokens[size]; + size++; + + *dr = tgsi_build_declaration_range( + full_decl->u.DeclarationRange.First, + full_decl->u.DeclarationRange.Last, + declaration, + header ); + break; + } + + case TGSI_DECLARE_MASK: + { + struct tgsi_declaration_mask *dm; + + if( maxsize <= size ) + return 0; + dm = (struct tgsi_declaration_mask *) &tokens[size]; + size++; + + *dm = tgsi_build_declaration_mask( + full_decl->u.DeclarationMask.Mask, + declaration, + header ); + break; + } + + default: + assert( 0 ); + } + + if( full_decl->Declaration.Interpolate ) { + struct tgsi_declaration_interpolation *di; + + if( maxsize <= size ) + return 0; + di = (struct tgsi_declaration_interpolation *) &tokens[size]; + size++; + + *di = tgsi_build_declaration_interpolation( + full_decl->Interpolation.Interpolate, + declaration, + header ); + } + + if( full_decl->Declaration.Semantic ) { + struct tgsi_declaration_semantic *ds; + + if( maxsize <= size ) + return 0; + ds = (struct tgsi_declaration_semantic *) &tokens[size]; + size++; + + *ds = tgsi_build_declaration_semantic( + full_decl->Semantic.SemanticName, + full_decl->Semantic.SemanticIndex, + declaration, + header ); + } + + return size; +} + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_range declaration_range; + + assert( last >= first ); + assert( last <= 0xFFFF ); + + declaration_range.First = first; + declaration_range.Last = last; + + declaration_grow( declaration, header ); + + return declaration_range; +} + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + unsigned mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_mask declaration_mask; + + declaration_mask.Mask = mask; + + declaration_grow( declaration, header ); + + return declaration_mask; +} + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ) +{ + struct tgsi_declaration_interpolation di; + + di.Interpolate = TGSI_INTERPOLATE_CONSTANT; + di.Padding = 0; + + return di; +} + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + unsigned interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_interpolation di; + + assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); + + di = tgsi_default_declaration_interpolation(); + di.Interpolate = interpolate; + + declaration_grow( declaration, header ); + + return di; +} + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ) +{ + struct tgsi_declaration_semantic ds; + + ds.SemanticName = TGSI_SEMANTIC_DEPTH; + ds.SemanticIndex = 0; + ds.Padding = 0; + + return ds; +} + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_semantic ds; + + assert( semantic_name <= TGSI_SEMANTIC_COLOR ); + assert( semantic_index <= 0xFFFF ); + + ds = tgsi_default_declaration_semantic(); + ds.SemanticName = semantic_name; + ds.SemanticIndex = semantic_index; + + declaration_grow( declaration, header ); + + return ds; +} + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ) +{ + struct tgsi_immediate immediate; + + immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + immediate.Size = 1; + immediate.DataType = TGSI_IMM_FLOAT32; + immediate.Padding = 0; + immediate.Extended = 0; + + return immediate; +} + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ) +{ + struct tgsi_immediate immediate; + + immediate = tgsi_default_immediate(); + + header_bodysize_grow( header ); + + return immediate; +} + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ) +{ + struct tgsi_full_immediate fullimm; + + fullimm.Immediate = tgsi_default_immediate(); + fullimm.u.Pointer = (void *) 0; + + return fullimm; +} + +static void +immediate_grow( + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + assert( immediate->Size < 0xFF ); + + immediate->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + GLfloat value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + struct tgsi_immediate_float32 immediate_float32; + + immediate_float32.Float = value; + + immediate_grow( immediate, header ); + + return immediate_float32; +} + +GLuint +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + GLuint maxsize ) +{ + GLuint size = 0, i; + struct tgsi_immediate *immediate; + + if( maxsize <= size ) + return 0; + immediate = (struct tgsi_immediate *) &tokens[size]; + size++; + + *immediate = tgsi_build_immediate( header ); + + for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { + struct tgsi_immediate_float32 *if32; + + if( maxsize <= size ) + return 0; + if32 = (struct tgsi_immediate_float32 *) &tokens[size]; + size++; + + *if32 = tgsi_build_immediate_float32( + full_imm->u.ImmediateFloat32[i].Float, + immediate, + header ); + } + + return size; +} + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ) +{ + struct tgsi_instruction instruction; + + instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; + instruction.Size = 1; + instruction.Opcode = TGSI_OPCODE_MOV; + instruction.Saturate = TGSI_SAT_NONE; + instruction.NumDstRegs = 1; + instruction.NumSrcRegs = 1; + instruction.Padding = 0; + instruction.Extended = 0; + + return instruction; +} + +struct tgsi_instruction +tgsi_build_instruction( + GLuint opcode, + GLuint saturate, + GLuint num_dst_regs, + GLuint num_src_regs, + struct tgsi_header *header ) +{ + struct tgsi_instruction instruction; + + assert (opcode <= TGSI_OPCODE_LAST); + assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); + assert (num_dst_regs <= 3); + assert (num_src_regs <= 15); + + instruction = tgsi_default_instruction(); + instruction.Opcode = opcode; + instruction.Saturate = saturate; + instruction.NumDstRegs = num_dst_regs; + instruction.NumSrcRegs = num_src_regs; + + header_bodysize_grow( header ); + + return instruction; +} + +static void +instruction_grow( + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + assert (instruction->Size < 0xFF); + + instruction->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ) +{ + struct tgsi_full_instruction full_instruction; + GLuint i; + + full_instruction.Instruction = tgsi_default_instruction(); + full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); + full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); + full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); + for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { + full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); + } + for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { + full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); + } + + return full_instruction; +} + +GLuint +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + GLuint maxsize ) +{ + GLuint size = 0; + GLuint i; + struct tgsi_instruction *instruction; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + instruction = (struct tgsi_instruction *) &tokens[size]; + size++; + + *instruction = tgsi_build_instruction( + full_inst->Instruction.Opcode, + full_inst->Instruction.Saturate, + full_inst->Instruction.NumDstRegs, + full_inst->Instruction.NumSrcRegs, + header ); + prev_token = (struct tgsi_token *) instruction; + + if( tgsi_compare_instruction_ext_nv( + full_inst->InstructionExtNv, + tgsi_default_instruction_ext_nv() ) ) { + struct tgsi_instruction_ext_nv *instruction_ext_nv; + + if( maxsize <= size ) + return 0; + instruction_ext_nv = + (struct tgsi_instruction_ext_nv *) &tokens[size]; + size++; + + *instruction_ext_nv = tgsi_build_instruction_ext_nv( + full_inst->InstructionExtNv.Precision, + full_inst->InstructionExtNv.CondDstIndex, + full_inst->InstructionExtNv.CondFlowIndex, + full_inst->InstructionExtNv.CondMask, + full_inst->InstructionExtNv.CondSwizzleX, + full_inst->InstructionExtNv.CondSwizzleY, + full_inst->InstructionExtNv.CondSwizzleZ, + full_inst->InstructionExtNv.CondSwizzleW, + full_inst->InstructionExtNv.CondDstUpdate, + full_inst->InstructionExtNv.CondFlowEnable, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_nv; + } + + if( tgsi_compare_instruction_ext_label( + full_inst->InstructionExtLabel, + tgsi_default_instruction_ext_label() ) ) { + struct tgsi_instruction_ext_label *instruction_ext_label; + + if( maxsize <= size ) + return 0; + instruction_ext_label = + (struct tgsi_instruction_ext_label *) &tokens[size]; + size++; + + *instruction_ext_label = tgsi_build_instruction_ext_label( + full_inst->InstructionExtLabel.Label, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_label; + } + + if( tgsi_compare_instruction_ext_texture( + full_inst->InstructionExtTexture, + tgsi_default_instruction_ext_texture() ) ) { + struct tgsi_instruction_ext_texture *instruction_ext_texture; + + if( maxsize <= size ) + return 0; + instruction_ext_texture = + (struct tgsi_instruction_ext_texture *) &tokens[size]; + size++; + + *instruction_ext_texture = tgsi_build_instruction_ext_texture( + full_inst->InstructionExtTexture.Texture, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_texture; + } + + for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { + const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; + struct tgsi_dst_register *dst_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + dst_register = (struct tgsi_dst_register *) &tokens[size]; + size++; + + *dst_register = tgsi_build_dst_register( + reg->DstRegister.File, + reg->DstRegister.WriteMask, + reg->DstRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register; + + if( tgsi_compare_dst_register_ext_concode( + reg->DstRegisterExtConcode, + tgsi_default_dst_register_ext_concode() ) ) { + struct tgsi_dst_register_ext_concode *dst_register_ext_concode; + + if( maxsize <= size ) + return 0; + dst_register_ext_concode = + (struct tgsi_dst_register_ext_concode *) &tokens[size]; + size++; + + *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( + reg->DstRegisterExtConcode.CondMask, + reg->DstRegisterExtConcode.CondSwizzleX, + reg->DstRegisterExtConcode.CondSwizzleY, + reg->DstRegisterExtConcode.CondSwizzleZ, + reg->DstRegisterExtConcode.CondSwizzleW, + reg->DstRegisterExtConcode.CondSrcIndex, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_concode; + } + + if( tgsi_compare_dst_register_ext_modulate( + reg->DstRegisterExtModulate, + tgsi_default_dst_register_ext_modulate() ) ) { + struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; + + if( maxsize <= size ) + return 0; + dst_register_ext_modulate = + (struct tgsi_dst_register_ext_modulate *) &tokens[size]; + size++; + + *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( + reg->DstRegisterExtModulate.Modulate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_modulate; + } + } + + for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { + const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; + struct tgsi_src_register *src_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + src_register = (struct tgsi_src_register *) &tokens[size]; + size++; + + *src_register = tgsi_build_src_register( + reg->SrcRegister.File, + reg->SrcRegister.SwizzleX, + reg->SrcRegister.SwizzleY, + reg->SrcRegister.SwizzleZ, + reg->SrcRegister.SwizzleW, + reg->SrcRegister.Negate, + reg->SrcRegister.Indirect, + reg->SrcRegister.Dimension, + reg->SrcRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register; + + if( tgsi_compare_src_register_ext_swz( + reg->SrcRegisterExtSwz, + tgsi_default_src_register_ext_swz() ) ) { + struct tgsi_src_register_ext_swz *src_register_ext_swz; + + if( maxsize <= size ) + return 0; + src_register_ext_swz = + (struct tgsi_src_register_ext_swz *) &tokens[size]; + size++; + + *src_register_ext_swz = tgsi_build_src_register_ext_swz( + reg->SrcRegisterExtSwz.ExtSwizzleX, + reg->SrcRegisterExtSwz.ExtSwizzleY, + reg->SrcRegisterExtSwz.ExtSwizzleZ, + reg->SrcRegisterExtSwz.ExtSwizzleW, + reg->SrcRegisterExtSwz.NegateX, + reg->SrcRegisterExtSwz.NegateY, + reg->SrcRegisterExtSwz.NegateZ, + reg->SrcRegisterExtSwz.NegateW, + reg->SrcRegisterExtSwz.ExtDivide, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_swz; + } + + if( tgsi_compare_src_register_ext_mod( + reg->SrcRegisterExtMod, + tgsi_default_src_register_ext_mod() ) ) { + struct tgsi_src_register_ext_mod *src_register_ext_mod; + + if( maxsize <= size ) + return 0; + src_register_ext_mod = + (struct tgsi_src_register_ext_mod *) &tokens[size]; + size++; + + *src_register_ext_mod = tgsi_build_src_register_ext_mod( + reg->SrcRegisterExtMod.Complement, + reg->SrcRegisterExtMod.Bias, + reg->SrcRegisterExtMod.Scale2X, + reg->SrcRegisterExtMod.Absolute, + reg->SrcRegisterExtMod.Negate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_mod; + } + + if( reg->SrcRegister.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterInd.File, + reg->SrcRegisterInd.SwizzleX, + reg->SrcRegisterInd.SwizzleY, + reg->SrcRegisterInd.SwizzleZ, + reg->SrcRegisterInd.SwizzleW, + reg->SrcRegisterInd.Negate, + reg->SrcRegisterInd.Indirect, + reg->SrcRegisterInd.Dimension, + reg->SrcRegisterInd.Index, + instruction, + header ); + } + + if( reg->SrcRegister.Dimension ) { + struct tgsi_dimension *dim; + + assert( !reg->SrcRegisterDim.Dimension ); + + if( maxsize <= size ) + return 0; + dim = (struct tgsi_dimension *) &tokens[size]; + size++; + + *dim = tgsi_build_dimension( + reg->SrcRegisterDim.Indirect, + reg->SrcRegisterDim.Index, + instruction, + header ); + + if( reg->SrcRegisterDim.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterDimInd.File, + reg->SrcRegisterDimInd.SwizzleX, + reg->SrcRegisterDimInd.SwizzleY, + reg->SrcRegisterDimInd.SwizzleZ, + reg->SrcRegisterDimInd.SwizzleW, + reg->SrcRegisterDimInd.Negate, + reg->SrcRegisterDimInd.Indirect, + reg->SrcRegisterDimInd.Dimension, + reg->SrcRegisterDimInd.Index, + instruction, + header ); + } + } + } + + return size; +} + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; + instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; + instruction_ext_nv.CondDstIndex = 0; + instruction_ext_nv.CondFlowIndex = 0; + instruction_ext_nv.CondMask = TGSI_CC_TR; + instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; + instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; + instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; + instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; + instruction_ext_nv.CondDstUpdate = 0; + instruction_ext_nv.CondFlowEnable = 0; + instruction_ext_nv.Padding = 0; + instruction_ext_nv.Extended = 0; + + return instruction_ext_nv; +} + +union token_u32 +{ + GLuint u32; +}; + +GLuint +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + GLuint precision, + GLuint cond_dst_index, + GLuint cond_flow_index, + GLuint cond_mask, + GLuint cond_swizzle_x, + GLuint cond_swizzle_y, + GLuint cond_swizzle_z, + GLuint cond_swizzle_w, + GLuint cond_dst_update, + GLuint cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv = tgsi_default_instruction_ext_nv(); + instruction_ext_nv.Precision = precision; + instruction_ext_nv.CondDstIndex = cond_dst_index; + instruction_ext_nv.CondFlowIndex = cond_flow_index; + instruction_ext_nv.CondMask = cond_mask; + instruction_ext_nv.CondSwizzleX = cond_swizzle_x; + instruction_ext_nv.CondSwizzleY = cond_swizzle_y; + instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; + instruction_ext_nv.CondSwizzleW = cond_swizzle_w; + instruction_ext_nv.CondDstUpdate = cond_dst_update; + instruction_ext_nv.CondFlowEnable = cond_flow_update; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_nv; +} + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; + instruction_ext_label.Label = 0; + instruction_ext_label.Padding = 0; + instruction_ext_label.Extended = 0; + + return instruction_ext_label; +} + +GLuint +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + GLuint label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label = tgsi_default_instruction_ext_label(); + instruction_ext_label.Label = label; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_label; +} + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; + instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; + instruction_ext_texture.Padding = 0; + instruction_ext_texture.Extended = 0; + + return instruction_ext_texture; +} + +GLuint +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + GLuint texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture = tgsi_default_instruction_ext_texture(); + instruction_ext_texture.Texture = texture; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_texture; +} + +struct tgsi_src_register +tgsi_default_src_register( void ) +{ + struct tgsi_src_register src_register; + + src_register.File = TGSI_FILE_NULL; + src_register.SwizzleX = TGSI_SWIZZLE_X; + src_register.SwizzleY = TGSI_SWIZZLE_Y; + src_register.SwizzleZ = TGSI_SWIZZLE_Z; + src_register.SwizzleW = TGSI_SWIZZLE_W; + src_register.Negate = 0; + src_register.Indirect = 0; + src_register.Dimension = 0; + src_register.Index = 0; + src_register.Extended = 0; + + return src_register; +} + +struct tgsi_src_register +tgsi_build_src_register( + GLuint file, + GLuint swizzle_x, + GLuint swizzle_y, + GLuint swizzle_z, + GLuint swizzle_w, + GLuint negate, + GLuint indirect, + GLuint dimension, + GLint index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register src_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( negate <= 1 ); + assert( index >= -0x8000 && index <= 0x7FFF ); + + src_register = tgsi_default_src_register(); + src_register.File = file; + src_register.SwizzleX = swizzle_x; + src_register.SwizzleY = swizzle_y; + src_register.SwizzleZ = swizzle_z; + src_register.SwizzleW = swizzle_w; + src_register.Negate = negate; + src_register.Indirect = indirect; + src_register.Dimension = dimension; + src_register.Index = index; + + instruction_grow( instruction, header ); + + return src_register; +} + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ) +{ + struct tgsi_full_src_register full_src_register; + + full_src_register.SrcRegister = tgsi_default_src_register(); + full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); + full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); + full_src_register.SrcRegisterInd = tgsi_default_src_register(); + full_src_register.SrcRegisterDim = tgsi_default_dimension(); + full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); + + return full_src_register; +} + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; + src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; + src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; + src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; + src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; + src_register_ext_swz.NegateX = 0; + src_register_ext_swz.NegateY = 0; + src_register_ext_swz.NegateZ = 0; + src_register_ext_swz.NegateW = 0; + src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; + src_register_ext_swz.Padding = 0; + src_register_ext_swz.Extended = 0; + + return src_register_ext_swz; +} + +GLuint +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + GLuint ext_swizzle_x, + GLuint ext_swizzle_y, + GLuint ext_swizzle_z, + GLuint ext_swizzle_w, + GLuint negate_x, + GLuint negate_y, + GLuint negate_z, + GLuint negate_w, + GLuint ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + assert (ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE); + assert (ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE); + assert (ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE); + assert (ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE); + assert (negate_x <= 1); + assert (negate_y <= 1); + assert (negate_z <= 1); + assert (negate_w <= 1); + assert (ext_divide <= TGSI_EXTSWIZZLE_ONE); + + src_register_ext_swz = tgsi_default_src_register_ext_swz(); + src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; + src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; + src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; + src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; + src_register_ext_swz.NegateX = negate_x; + src_register_ext_swz.NegateY = negate_y; + src_register_ext_swz.NegateZ = negate_z; + src_register_ext_swz.NegateW = negate_w; + src_register_ext_swz.ExtDivide = ext_divide; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_swz; +} + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; + src_register_ext_mod.Complement = 0; + src_register_ext_mod.Bias = 0; + src_register_ext_mod.Scale2X = 0; + src_register_ext_mod.Absolute = 0; + src_register_ext_mod.Negate = 0; + src_register_ext_mod.Padding = 0; + src_register_ext_mod.Extended = 0; + + return src_register_ext_mod; +} + +GLuint +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + GLuint complement, + GLuint bias, + GLuint scale_2x, + GLuint absolute, + GLuint negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + assert (complement <= 1); + assert (bias <= 1); + assert (scale_2x <= 1); + assert (absolute <= 1); + assert (negate <= 1); + + src_register_ext_mod = tgsi_default_src_register_ext_mod(); + src_register_ext_mod.Complement = complement; + src_register_ext_mod.Bias = bias; + src_register_ext_mod.Scale2X = scale_2x; + src_register_ext_mod.Absolute = absolute; + src_register_ext_mod.Negate = negate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_mod; +} + +struct tgsi_dimension +tgsi_default_dimension( void ) +{ + struct tgsi_dimension dimension; + + dimension.Indirect = 0; + dimension.Dimension = 0; + dimension.Padding = 0; + dimension.Index = 0; + dimension.Extended = 0; + + return dimension; +} + +struct tgsi_dimension +tgsi_build_dimension( + GLuint indirect, + GLuint index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dimension dimension; + + dimension = tgsi_default_dimension(); + dimension.Indirect = indirect; + dimension.Index = index; + + instruction_grow( instruction, header ); + + return dimension; +} + +struct tgsi_dst_register +tgsi_default_dst_register( void ) +{ + struct tgsi_dst_register dst_register; + + dst_register.File = TGSI_FILE_NULL; + dst_register.WriteMask = TGSI_WRITEMASK_XYZW; + dst_register.Indirect = 0; + dst_register.Dimension = 0; + dst_register.Index = 0; + dst_register.Padding = 0; + dst_register.Extended = 0; + + return dst_register; +} + +struct tgsi_dst_register +tgsi_build_dst_register( + GLuint file, + GLuint mask, + GLint index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register dst_register; + + assert (file <= TGSI_FILE_IMMEDIATE); + assert (mask <= TGSI_WRITEMASK_XYZW); + assert (index >= -32768 && index <= 32767); + + dst_register = tgsi_default_dst_register(); + dst_register.File = file; + dst_register.WriteMask = mask; + dst_register.Index = index; + + instruction_grow( instruction, header ); + + return dst_register; +} + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ) +{ + struct tgsi_full_dst_register full_dst_register; + + full_dst_register.DstRegister = tgsi_default_dst_register(); + full_dst_register.DstRegisterExtConcode = + tgsi_default_dst_register_ext_concode(); + full_dst_register.DstRegisterExtModulate = + tgsi_default_dst_register_ext_modulate(); + + return full_dst_register; +} + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; + dst_register_ext_concode.CondMask = TGSI_CC_TR; + dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; + dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; + dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; + dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; + dst_register_ext_concode.CondSrcIndex = 0; + dst_register_ext_concode.Padding = 0; + dst_register_ext_concode.Extended = 0; + + return dst_register_ext_concode; +} + +GLuint +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + GLuint cc, + GLuint swizzle_x, + GLuint swizzle_y, + GLuint swizzle_z, + GLuint swizzle_w, + GLint index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + assert (cc <= TGSI_CC_FL); + assert (swizzle_x <= TGSI_SWIZZLE_W); + assert (swizzle_y <= TGSI_SWIZZLE_W); + assert (swizzle_z <= TGSI_SWIZZLE_W); + assert (swizzle_w <= TGSI_SWIZZLE_W); + assert (index >= -32768 && index <= 32767); + + dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); + dst_register_ext_concode.CondMask = cc; + dst_register_ext_concode.CondSwizzleX = swizzle_x; + dst_register_ext_concode.CondSwizzleY = swizzle_y; + dst_register_ext_concode.CondSwizzleZ = swizzle_z; + dst_register_ext_concode.CondSwizzleW = swizzle_w; + dst_register_ext_concode.CondSrcIndex = index; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_concode; +} + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; + dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; + dst_register_ext_modulate.Padding = 0; + dst_register_ext_modulate.Extended = 0; + + return dst_register_ext_modulate; +} + +GLuint +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return *(GLuint *) &a != *(GLuint *) &b; +} + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + GLuint modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + assert (modulate <= TGSI_MODULATE_EIGHTH); + + dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); + dst_register_ext_modulate.Modulate = modulate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_modulate; +} + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.h b/src/mesa/pipe/tgsi/exec/tgsi_build.h new file mode 100644 index 0000000000..116c78abf3 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_build.h @@ -0,0 +1,320 @@ +#if !defined TGSI_BUILD_H +#define TGSI_BUILD_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ); + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ); + +struct tgsi_processor +tgsi_default_processor( void ); + +struct tgsi_processor +tgsi_build_processor( + unsigned processor, + struct tgsi_header *header ); + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ); + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned declare, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ); + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ); + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + unsigned mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ); + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + unsigned interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ); + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ); + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ); + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ); + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + float value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ); + +unsigned +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ); + +struct tgsi_instruction +tgsi_build_instruction( + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header ); + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ); + +unsigned +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ); + +unsigned +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ); + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ); + +unsigned +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ); + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + unsigned label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ); + +unsigned +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ); + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + unsigned texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register +tgsi_default_src_register( void ); + +struct tgsi_src_register +tgsi_build_src_register( + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ); + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ); + +unsigned +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ); + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + unsigned ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ); + +unsigned +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ); + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dimension +tgsi_default_dimension( void ); + +struct tgsi_dimension +tgsi_build_dimension( + unsigned indirect, + unsigned index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register +tgsi_default_dst_register( void ); + +struct tgsi_dst_register +tgsi_build_dst_register( + unsigned file, + unsigned mask, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ); + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ); + +unsigned +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ); + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ); + +unsigned +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ); + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + unsigned modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_BUILD_H + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_core.h b/src/mesa/pipe/tgsi/exec/tgsi_core.h new file mode 100644 index 0000000000..1f5f00a38e --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_core.h @@ -0,0 +1,12 @@ +#if !defined TGSI_CORE_H +#define TGSI_CORE_H + +#include "tgsi_token.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" +#include "tgsi_exec.h" +#include "tgsi_dump.h" +#include "tgsi_util.h" + +#endif // !defined TGSI_CORE_H + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c new file mode 100644 index 0000000000..622d617e29 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -0,0 +1,1402 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +struct text_dump +{ + FILE *file; + GLuint tabs; +}; + +static void +text_dump_write( + struct text_dump *dump, + const void *buffer, + GLuint size ) +{ + fwrite( buffer, size, 1, dump->file ); +} + +static void +text_dump_str( + struct text_dump *dump, + const char *str ) +{ + GLuint i; + GLuint len = strlen( str ); + + for( i = 0; i < len; i++ ) { + text_dump_write( dump, &str[i], 1 ); + + if( str[i] == '\n' ) { + GLuint i; + + for( i = 0; i < dump->tabs; i++ ) { + text_dump_write( dump, " ", 4 ); + } + } + } +} + +static void +text_dump_chr( + struct text_dump *dump, + const char chr ) +{ + char str[2]; + + str[0] = chr; + str[1] = '\0'; + text_dump_str( dump, str ); +} + +static void +text_dump_uix( + struct text_dump *dump, + const GLuint ui) +{ + char str[36]; + + sprintf( str, "0x%x", ui ); + text_dump_str( dump, str ); +} + +static void +text_dump_uid( + struct text_dump *dump, + const GLuint ui ) +{ + char str[16]; + + sprintf( str, "%u", ui ); + text_dump_str( dump, str ); +} + +static void +text_dump_sid( + struct text_dump *dump, + const GLint si ) +{ + char str[16]; + + sprintf( str, "%d", si ); + text_dump_str( dump, str ); +} + +static void +text_dump_flt( + struct text_dump *dump, + const GLfloat f ) +{ + char str[48]; + + sprintf( str, "%40.6f", f ); + text_dump_str( dump, str ); +} + +static void +text_dump_enum( + struct text_dump *dump, + const GLuint e, + const char **enums, + const GLuint enums_count ) +{ + if( e >= enums_count ) { + text_dump_uid( dump, e ); + } + else { + text_dump_str( dump, enums[e] ); + } +} + +static void +text_dump_tab( + struct text_dump *dump ) +{ + dump->tabs++; +} + +static void +text_dump_untab( + struct text_dump *dump ) +{ + assert( dump->tabs > 0 ); + + --dump->tabs; +} + +#define TXT(S) text_dump_str( dump, S ) +#define CHR(C) text_dump_chr( dump, C ) +#define UIX(I) text_dump_uix( dump, I ) +#define UID(I) text_dump_uid( dump, I ) +#define SID(I) text_dump_sid( dump, I ) +#define FLT(F) text_dump_flt( dump, F ) +#define TAB() text_dump_tab( dump ) +#define UNT() text_dump_untab( dump ) +#define ENM(E,ENUMS) text_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) + +static const char *TGSI_PROCESSOR_TYPES[] = +{ + "PROCESSOR_FRAGMENT", + "PROCESSOR_VERTEX", + "PROCESSOR_GEOMETRY" +}; + +static const char *TGSI_PROCESSOR_TYPES_SHORT[] = +{ + "FRAG", + "VERT", + "GEOM" +}; + +static const char *TGSI_TOKEN_TYPES[] = +{ + "TOKEN_TYPE_DECLARATION", + "TOKEN_TYPE_IMMEDIATE", + "TOKEN_TYPE_INSTRUCTION" +}; + +static const char *TGSI_FILES[] = +{ + "FILE_NULL", + "FILE_CONSTANT", + "FILE_INPUT", + "FILE_OUTPUT", + "FILE_TEMPORARY", + "FILE_SAMPLER", + "FILE_ADDRESS", + "FILE_IMMEDIATE" +}; + +static const char *TGSI_FILES_SHORT[] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + +static const char *TGSI_DECLARES[] = +{ + "DECLARE_RANGE", + "DECLARE_MASK" +}; + +static const char *TGSI_INTERPOLATES[] = +{ + "INTERPOLATE_CONSTANT", + "INTERPOLATE_LINEAR", + "INTERPOLATE_PERSPECTIVE" +}; + +static const char *TGSI_INTERPOLATES_SHORT[] = +{ + "CONSTANT", + "LINEAR", + "PERSPECTIVE" +}; + +static const char *TGSI_SEMANTICS[] = +{ + "SEMANTIC_DEPTH", + "SEMANTIC_COLOR" +}; + +static const char *TGSI_SEMANTICS_SHORT[] = +{ + "DEPTH", + "COLOR" +}; + +static const char *TGSI_IMMS[] = +{ + "IMM_FLOAT32" +}; + +static const char *TGSI_IMMS_SHORT[] = +{ + "FLT32" +}; + +static const char *TGSI_OPCODES[] = +{ + "OPCODE_ARL", + "OPCODE_MOV", + "OPCODE_LIT", + "OPCODE_RCP", + "OPCODE_RSQ", + "OPCODE_EXP", + "OPCODE_LOG", + "OPCODE_MUL", + "OPCODE_ADD", + "OPCODE_DP3", + "OPCODE_DP4", + "OPCODE_DST", + "OPCODE_MIN", + "OPCODE_MAX", + "OPCODE_SLT", + "OPCODE_SGE", + "OPCODE_MAD", + "OPCODE_SUB", + "OPCODE_LERP", + "OPCODE_CND", + "OPCODE_CND0", + "OPCODE_DOT2ADD", + "OPCODE_INDEX", + "OPCODE_NEGATE", + "OPCODE_FRAC", + "OPCODE_CLAMP", + "OPCODE_FLOOR", + "OPCODE_ROUND", + "OPCODE_EXPBASE2", + "OPCODE_LOGBASE2", + "OPCODE_POWER", + "OPCODE_CROSSPRODUCT", + "OPCODE_MULTIPLYMATRIX", + "OPCODE_ABS", + "OPCODE_RCC", + "OPCODE_DPH", + "OPCODE_COS", + "OPCODE_DDX", + "OPCODE_DDY", + "OPCODE_KIL", + "OPCODE_PK2H", + "OPCODE_PK2US", + "OPCODE_PK4B", + "OPCODE_PK4UB", + "OPCODE_RFL", + "OPCODE_SEQ", + "OPCODE_SFL", + "OPCODE_SGT", + "OPCODE_SIN", + "OPCODE_SLE", + "OPCODE_SNE", + "OPCODE_STR", + "OPCODE_TEX", + "OPCODE_TXD", + "OPCODE_UP2H", + "OPCODE_UP2US", + "OPCODE_UP4B", + "OPCODE_UP4UB", + "OPCODE_X2D", + "OPCODE_ARA", + "OPCODE_ARR", + "OPCODE_BRA", + "OPCODE_CAL", + "OPCODE_RET", + "OPCODE_SSG", + "OPCODE_CMP", + "OPCODE_SCS", + "OPCODE_TXB", + "OPCODE_NRM", + "OPCODE_DIV", + "OPCODE_DP2", + "OPCODE_TXL", + "OPCODE_BRK", + "OPCODE_IF", + "OPCODE_LOOP", + "OPCODE_REP", + "OPCODE_ELSE", + "OPCODE_ENDIF", + "OPCODE_ENDLOOP", + "OPCODE_ENDREP", + "OPCODE_PUSHA", + "OPCODE_POPA", + "OPCODE_CEIL", + "OPCODE_I2F", + "OPCODE_NOT", + "OPCODE_TRUNC", + "OPCODE_SHL", + "OPCODE_SHR", + "OPCODE_AND", + "OPCODE_OR", + "OPCODE_MOD", + "OPCODE_XOR", + "OPCODE_SAD", + "OPCODE_TXF", + "OPCODE_TXQ", + "OPCODE_CONT", + "OPCODE_EMIT", + "OPCODE_ENDPRIM", + "OPCODE_BGNLOOP2", + "OPCODE_BGNSUB", + "OPCODE_ENDLOOP2", + "OPCODE_ENDSUB", + "OPCODE_NOISE1", + "OPCODE_NOISE2", + "OPCODE_NOISE3", + "OPCODE_NOISE4", + "OPCODE_NOP", + "OPCODE_TEXBEM", + "OPCODE_TEXBEML", + "OPCODE_TEXREG2AR", + "OPCODE_TEXM3X2PAD", + "OPCODE_TEXM3X2TEX", + "OPCODE_TEXM3X3PAD", + "OPCODE_TEXM3X3TEX", + "OPCODE_TEXM3X3SPEC", + "OPCODE_TEXM3X3VSPEC", + "OPCODE_TEXREG2GB", + "OPCODE_TEXREG2RGB", + "OPCODE_TEXDP3TEX", + "OPCODE_TEXDP3", + "OPCODE_TEXM3X3", + "OPCODE_TEXM3X2DEPTH", + "OPCODE_TEXDEPTH", + "OPCODE_BEM", + "OPCODE_M4X3", + "OPCODE_M3X4", + "OPCODE_M3X3", + "OPCODE_M3X2", + "OPCODE_NRM4", + "OPCODE_CALLNZ", + "OPCODE_IFC", + "OPCODE_BREAKC", + "OPCODE_TXP" +}; + +static const char *TGSI_OPCODES_SHORT[] = +{ + "ARL", + "MOV", + "LIT", + "RCP", + "RSQ", + "EXP", + "LOG", + "MUL", + "ADD", + "DP3", + "DP4", + "DST", + "MIN", + "MAX", + "SLT", + "SGE", + "MAD", + "SUB", + "LERP", + "CND", + "CND0", + "DOT2ADD", + "INDEX", + "NEGATE", + "FRAC", + "CLAMP", + "FLOOR", + "ROUND", + "EXPBASE2", + "LOGBASE2", + "POWER", + "CROSSPRODUCT", + "MULTIPLYMATRIX", + "ABS", + "RCC", + "DPH", + "COS", + "DDX", + "DDY", + "KIL", + "PK2H", + "PK2US", + "PK4B", + "PK4UB", + "RFL", + "SEQ", + "SFL", + "SGT", + "SIN", + "SLE", + "SNE", + "STR", + "TEX", + "TXD", + "UP2H", + "UP2US", + "UP4B", + "UP4UB", + "X2D", + "ARA", + "ARR", + "BRA", + "CAL", + "RET", + "SSG", + "CMP", + "SCS", + "TXB", + "NRM", + "DIV", + "DP2", + "TXL", + "BRK", + "IF", + "LOOP", + "REP", + "ELSE", + "ENDIF", + "ENDLOOP", + "ENDREP", + "PUSHA", + "POPA", + "CEIL", + "I2F", + "NOT", + "TRUNC", + "SHL", + "SHR", + "AND", + "OR", + "MOD", + "XOR", + "SAD", + "TXF", + "TXQ", + "CONT", + "EMIT", + "ENDPRIM", + "BGNLOOP2", + "BGNSUB", + "ENDLOOP2", + "ENDSUB", + "NOISE1", + "NOISE2", + "NOISE3", + "NOISE4", + "NOP", + "TEXBEM", + "TEXBEML", + "TEXREG2AR", + "TEXM3X2PAD", + "TEXM3X2TEX", + "TEXM3X3PAD", + "TEXM3X3TEX", + "TEXM3X3SPEC", + "TEXM3X3VSPEC", + "TEXREG2GB", + "TEXREG2RGB", + "TEXDP3TEX", + "TEXDP3", + "TEXM3X3", + "TEXM3X2DEPTH", + "TEXDEPTH", + "BEM", + "M4X3", + "M3X4", + "M3X3", + "M3X2", + "NRM4", + "CALLNZ", + "IFC", + "BREAKC", +}; + +static const char *TGSI_SATS[] = +{ + "SAT_NONE", + "SAT_ZERO_ONE", + "SAT_MINUS_PLUS_ONE" +}; + +static const char *TGSI_INSTRUCTION_EXTS[] = +{ + "INSTRUCTION_EXT_TYPE_NV", + "INSTRUCTION_EXT_TYPE_LABEL", + "INSTRUCTION_EXT_TYPE_TEXTURE" +}; + +static const char *TGSI_PRECISIONS[] = +{ + "PRECISION_DEFAULT", + "TGSI_PRECISION_FLOAT32", + "TGSI_PRECISION_FLOAT16", + "TGSI_PRECISION_FIXED12" +}; + +static const char *TGSI_CCS[] = +{ + "CC_GT", + "CC_EQ", + "CC_LT", + "CC_UN", + "CC_GE", + "CC_LE", + "CC_NE", + "CC_TR", + "CC_FL" +}; + +static const char *TGSI_SWIZZLES[] = +{ + "SWIZZLE_X", + "SWIZZLE_Y", + "SWIZZLE_Z", + "SWIZZLE_W" +}; + +static const char *TGSI_SWIZZLES_SHORT[] = +{ + "x", + "y", + "z", + "w" +}; + +static const char *TGSI_TEXTURES[] = +{ + "TEXTURE_UNKNOWN", + "TEXTURE_1D", + "TEXTURE_2D", + "TEXTURE_3D", + "TEXTURE_CUBE", + "TEXTURE_RECT", + "TEXTURE_SHADOW1D", + "TEXTURE_SHADOW2D", + "TEXTURE_SHADOWRECT" +}; + +static const char *TGSI_SRC_REGISTER_EXTS[] = +{ + "SRC_REGISTER_EXT_TYPE_SWZ", + "SRC_REGISTER_EXT_TYPE_MOD" +}; + +static const char *TGSI_EXTSWIZZLES[] = +{ + "EXTSWIZZLE_X", + "EXTSWIZZLE_Y", + "EXTSWIZZLE_Z", + "EXTSWIZZLE_W", + "EXTSWIZZLE_ZERO", + "EXTSWIZZLE_ONE" +}; + +static const char *TGSI_WRITEMASKS[] = +{ + "0", + "WRITEMASK_X", + "WRITEMASK_Y", + "WRITEMASK_XY", + "WRITEMASK_Z", + "WRITEMASK_XZ", + "WRITEMASK_YZ", + "WRITEMASK_XYZ", + "WRITEMASK_W", + "WRITEMASK_XW", + "WRITEMASK_YW", + "WRITEMASK_XYW", + "WRITEMASK_ZW", + "WRITEMASK_XZW", + "WRITEMASK_YZW", + "WRITEMASK_XYZW" +}; + +static const char *TGSI_DST_REGISTER_EXTS[] = +{ + "DST_REGISTER_EXT_TYPE_CONDCODE", + "DST_REGISTER_EXT_TYPE_MODULATE" +}; + +static const char *TGSI_MODULATES[] = +{ + "MODULATE_1X", + "MODULATE_2X", + "MODULATE_4X", + "MODULATE_8X", + "MODULATE_HALF", + "MODULATE_QUARTER", + "MODULATE_EIGHTH" +}; + +static void +dump_declaration_short( + struct text_dump *dump, + struct tgsi_full_declaration *decl ) +{ + TXT( "\nDCL " ); + ENM( decl->Declaration.File, TGSI_FILES_SHORT ); + + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + CHR( '[' ); + UID( decl->u.DeclarationRange.First ); + if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { + TXT( ".." ); + UID( decl->u.DeclarationRange.Last ); + } + CHR( ']' ); + break; + default: + assert( 0 ); + } + + if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + + if( decl->Declaration.Interpolate ) { + TXT( ", " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); + } + + if( decl->Declaration.Semantic ) { + TXT( ", " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); + CHR( '[' ); + UID( decl->Semantic.SemanticIndex ); + CHR( ']' ); + } +} + +static void +dump_declaration_verbose( + struct text_dump *dump, + struct tgsi_full_declaration *decl, + GLuint ignored, + GLuint deflt, + struct tgsi_full_declaration *fd ) +{ + TXT( "\nFile : " ); + ENM( decl->Declaration.File, TGSI_FILES ); + TXT( "\nDeclare : " ); + ENM( decl->Declaration.Declare, TGSI_DECLARES ); + if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { + TXT( "\nUsageMask : " ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'X' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'Y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'Z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'W' ); + } + } + if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { + TXT( "\nInterpolate: " ); + UID( decl->Declaration.Interpolate ); + } + if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { + TXT( "\nSemantic : " ); + UID( decl->Declaration.Semantic ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Declaration.Padding ); + } + + CHR( '\n' ); + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + TXT( "\nFirst: " ); + UID( decl->u.DeclarationRange.First ); + TXT( "\nLast : " ); + UID( decl->u.DeclarationRange.Last ); + break; + + case TGSI_DECLARE_MASK: + TXT( "\nMask: " ); + UIX( decl->u.DeclarationMask.Mask ); + break; + + default: + assert( 0 ); + } + + if( decl->Declaration.Interpolate ) { + CHR( '\n' ); + TXT( "\nInterpolate: " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Interpolation.Padding ); + } + } + + if( decl->Declaration.Semantic ) { + CHR( '\n' ); + TXT( "\nSemanticName : " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); + TXT( "\nSemanticIndex: " ); + UID( decl->Semantic.SemanticIndex ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Semantic.Padding ); + } + } +} + +static void +dump_immediate_short( + struct text_dump *dump, + struct tgsi_full_immediate *imm ) +{ + GLuint i; + + TXT( "\nIMM " ); + ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); + + TXT( " { " ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + + if( i < imm->Immediate.Size - 2 ) { + TXT( ", " ); + } + } + TXT( " }" ); +} + +static void +dump_immediate_verbose( + struct text_dump *dump, + struct tgsi_full_immediate *imm, + GLuint ignored ) +{ + GLuint i; + + TXT( "\nDataType : " ); + ENM( imm->Immediate.DataType, TGSI_IMMS ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( imm->Immediate.Padding ); + } + + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + CHR( '\n' ); + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + TXT( "\nFloat: " ); + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + } +} + +static void +dump_instruction_short( + struct text_dump *dump, + struct tgsi_full_instruction *inst, + GLuint instno ) +{ + GLuint i; + GLboolean first_reg = GL_TRUE; + + CHR( '\n' ); + UID( instno ); + CHR( ':' ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); + + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + TXT( "_SAT" ); + break; + case TGSI_SAT_MINUS_PLUS_ONE: + TXT( "_SAT[-1,1]" ); + break; + default: + assert( 0 ); + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( dst->DstRegister.Index ); + CHR( ']' ); + + if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + + first_reg = GL_FALSE; + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + if( src->SrcRegisterExtMod.Negate ) { + CHR( '-' ); + } + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + if( src->SrcRegister.Negate ) { + CHR( '-' ); + } + + ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( src->SrcRegister.Index ); + CHR( ']' ); + + if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { + CHR( '.' ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); + } + + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + + first_reg = GL_FALSE; + } + + switch( inst->Instruction.Opcode ) { + case TGSI_OPCODE_IF: + case TGSI_OPCODE_ELSE: + TXT( " :" ); + UID( inst->InstructionExtLabel.Label ); + break; + } +} + +static void +dump_instruction_verbose( + struct text_dump *dump, + struct tgsi_full_instruction *inst, + GLuint ignored, + GLuint deflt, + struct tgsi_full_instruction *fi ) +{ + GLuint i; + + TXT( "\nOpcode : " ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES ); + if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { + TXT( "\nSaturate : " ); + ENM( inst->Instruction.Saturate, TGSI_SATS ); + } + if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { + TXT( "\nNumDstRegs : " ); + UID( inst->Instruction.NumDstRegs ); + } + if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { + TXT( "\nNumSrcRegs : " ); + UID( inst->Instruction.NumSrcRegs ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->Instruction.Padding ); + } + + if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { + TXT( "\nPrecision : " ); + ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); + } + if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { + TXT( "\nCondDstIndex : " ); + UID( inst->InstructionExtNv.CondDstIndex ); + } + if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { + TXT( "\nCondFlowIndex : " ); + UID( inst->InstructionExtNv.CondFlowIndex ); + } + if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { + TXT( "\nCondMask : " ); + ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { + TXT( "\nCondSwizzleX : " ); + ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { + TXT( "\nCondSwizzleY : " ); + ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ : " ); + ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { + TXT( "\nCondSwizzleW : " ); + ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { + TXT( "\nCondDstUpdate : " ); + UID( inst->InstructionExtNv.CondDstUpdate ); + } + if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { + TXT( "\nCondFlowEnable: " ); + UID( inst->InstructionExtNv.CondFlowEnable ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtNv.Padding ); + if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { + TXT( "\nExtended : " ); + UID( inst->InstructionExtNv.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { + TXT( "\nLabel : " ); + UID( inst->InstructionExtLabel.Label ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtLabel.Padding ); + if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtLabel.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { + TXT( "\nTexture : " ); + ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtTexture.Padding ); + if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtTexture.Extended ); + } + } + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : " ); + ENM( dst->DstRegister.File, TGSI_FILES ); + if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { + TXT( "\nWriteMask: " ); + ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); + } + if( ignored ) { + if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( dst->DstRegister.Indirect ); + } + if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( dst->DstRegister.Dimension ); + } + } + if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { + TXT( "\nIndex : " ); + SID( dst->DstRegister.Index ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegister.Padding ); + if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegister.Extended ); + } + } + + if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { + TXT( "\nCondMask : " ); + ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { + TXT( "\nCondSwizzleX: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { + TXT( "\nCondSwizzleY: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { + TXT( "\nCondSwizzleW: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { + TXT( "\nCondSrcIndex: " ); + UID( dst->DstRegisterExtConcode.CondSrcIndex ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtConcode.Padding ); + if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegisterExtConcode.Extended ); + } + } + } + + if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { + TXT( "\nModulate: " ); + ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtModulate.Padding ); + if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { + TXT( "\nExtended: " ); + UID( dst->DstRegisterExtModulate.Extended ); + } + } + } + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : "); + ENM( src->SrcRegister.File, TGSI_FILES ); + if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { + TXT( "\nSwizzleX : " ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { + TXT( "\nSwizzleY : " ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { + TXT( "\nSwizzleZ : " ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { + TXT( "\nSwizzleW : " ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegister.Negate ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( src->SrcRegister.Indirect ); + } + if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( src->SrcRegister.Dimension ); + } + } + if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { + TXT( "\nIndex : " ); + SID( src->SrcRegister.Index ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegister.Extended ); + } + } + + if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { + TXT( "\nExtSwizzleX: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { + TXT( "\nExtSwizzleY: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { + TXT( "\nExtSwizzleZ: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { + TXT( "\nExtSwizzleW: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { + TXT( "\nNegateX : " ); + UID( src->SrcRegisterExtSwz.NegateX ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { + TXT( "\nNegateY : " ); + UID( src->SrcRegisterExtSwz.NegateY ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { + TXT( "\nNegateZ : " ); + UID( src->SrcRegisterExtSwz.NegateZ ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { + TXT( "\nNegateW : " ); + UID( src->SrcRegisterExtSwz.NegateW ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { + TXT( "\nExtDivide : " ); + ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtSwz.Padding ); + if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtSwz.Extended ); + } + } + } + + if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { + TXT( "\nComplement: " ); + UID( src->SrcRegisterExtMod.Complement ); + } + if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { + TXT( "\nBias : " ); + UID( src->SrcRegisterExtMod.Bias ); + } + if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { + TXT( "\nScale2X : " ); + UID( src->SrcRegisterExtMod.Scale2X ); + } + if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { + TXT( "\nAbsolute : " ); + UID( src->SrcRegisterExtMod.Absolute ); + } + if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegisterExtMod.Negate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtMod.Padding ); + if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtMod.Extended ); + } + } + } + } +} + +void +tgsi_dump( + const struct tgsi_token *tokens, + GLuint flags ) +{ + struct text_dump _dump; + struct text_dump *dump = &_dump; + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + GLuint verbose = flags & TGSI_DUMP_VERBOSE; + GLuint ignored = !(flags & TGSI_DUMP_NO_IGNORED); + GLuint deflt = !(flags & TGSI_DUMP_NO_DEFAULT); + GLuint instno = 0; + + { +#if 0 + static GLuint counter = 0; + char buffer[64]; + + sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); + dump->file = fopen( buffer, "wt" ); +#else + dump->file = stderr; + dump->tabs = 0; +#endif + } + + /* sanity check */ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); + + tgsi_parse_init( &parse, tokens ); + + TXT( "tgsi-dump begin -----------------" ); + + CHR( '\n' ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); + CHR( ' ' ); + UID( parse.FullVersion.Version.MajorVersion ); + CHR( '.' ); + UID( parse.FullVersion.Version.MinorVersion ); + + if( verbose ) { + TXT( "\nMajorVersion: " ); + UID( parse.FullVersion.Version.MajorVersion ); + TXT( "\nMinorVersion: " ); + UID( parse.FullVersion.Version.MinorVersion ); + CHR( '\n' ); + + TXT( "\nHeaderSize: " ); + UID( parse.FullHeader.Header.HeaderSize ); + TXT( "\nBodySize : " ); + UID( parse.FullHeader.Header.BodySize ); + TXT( "\nProcessor : " ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + CHR( '\n' ); + } + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_short( + dump, + &parse.FullToken.FullDeclaration ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_short( + dump, + &parse.FullToken.FullImmediate ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_short( + dump, + &parse.FullToken.FullInstruction, + instno ); + instno++; + break; + + default: + assert( 0 ); + } + + if( verbose ) { + TXT( "\nType : " ); + ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); + if( ignored ) { + TXT( "\nSize : " ); + UID( parse.FullToken.Token.Size ); + if( deflt || parse.FullToken.Token.Extended ) { + TXT( "\nExtended : " ); + UID( parse.FullToken.Token.Extended ); + } + } + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_verbose( + dump, + &parse.FullToken.FullDeclaration, + ignored, + deflt, + &fd ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_verbose( + dump, + &parse.FullToken.FullImmediate, + ignored ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_verbose( + dump, + &parse.FullToken.FullInstruction, + ignored, + deflt, + &fi ); + break; + + default: + assert( 0 ); + } + + CHR( '\n' ); + } + } + + TXT( "\ntgsi-dump end -------------------\n" ); + + tgsi_parse_free( &parse ); + + if (dump->file != stderr && + dump->file != stdout) + fclose( dump->file ); +} + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.h b/src/mesa/pipe/tgsi/exec/tgsi_dump.h new file mode 100644 index 0000000000..70860c0885 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.h @@ -0,0 +1,22 @@ +#if !defined TGSI_DUMP_H +#define TGSI_DUMP_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +#define TGSI_DUMP_VERBOSE 1 +#define TGSI_DUMP_NO_IGNORED 2 +#define TGSI_DUMP_NO_DEFAULT 4 + +void +tgsi_dump( + const struct tgsi_token *tokens, + unsigned flags ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_DUMP_H + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c new file mode 100644 index 0000000000..a1f46712b4 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -0,0 +1,2313 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" +#include "pipe/p_state.h" + +#define MESA 1 +#if MESA +#include "main/context.h" +#include "main/macros.h" +#endif + +#define TILE_BOTTOM_LEFT 0 +#define TILE_BOTTOM_RIGHT 1 +#define TILE_TOP_LEFT 2 +#define TILE_TOP_RIGHT 3 + +#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I +#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C +#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I +#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C +#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I +#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C +#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I +#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C +#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I +#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C +#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I +#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C +#define TEMP_128_I TGSI_EXEC_TEMP_128_I +#define TEMP_128_C TGSI_EXEC_TEMP_128_C +#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I +#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C +#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I +#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C +#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I +#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C +#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I +#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +#define FOR_EACH_CHANNEL(CHAN)\ + for (CHAN = 0; CHAN < 4; CHAN++) + +#define IS_CHANNEL_ENABLED(INST, CHAN)\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IS_CHANNEL_ENABLED2(INST, CHAN)\ + ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) + +#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED2( INST, CHAN )) + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach, + const struct tgsi_token *tokens, + GLuint numSamplers, + struct tgsi_sampler *samplers) +{ + GLuint i, k; + struct tgsi_parse_context parse; + + mach->Tokens = tokens; + + mach->Samplers = samplers; + + k = tgsi_parse_init (&parse, mach->Tokens); + if (k != TGSI_PARSE_OK) { + printf("Problem parsing!\n"); + return; + } + + mach->Processor = parse.FullHeader.Processor.Processor; + tgsi_parse_free (&parse); + + mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); + mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; + +#if XXX_SSE + tgsi_emit_sse (tokens, + &mach->Function); +#endif + + /* Setup constants. */ + for( i = 0; i < 4; i++ ) { + mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; + mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; + mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; + mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; + mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; + mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; + mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; + mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; + } +} + +void +tgsi_exec_prepare( + struct tgsi_exec_machine *mach, + struct tgsi_exec_labels *labels ) +{ + struct tgsi_parse_context parse; + GLuint k; + GLuint instno = 0; + + mach->ImmLimit = 0; + labels->count = 0; + + k = tgsi_parse_init( &parse, mach->Tokens ); + if (k != TGSI_PARSE_OK) { + printf("Problem parsing!\n"); + return; + } + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + GLuint pointer = parse.Position; + GLuint i; + + tgsi_parse_token( &parse ); + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + assert( (parse.FullToken.FullImmediate.Immediate.Size - 1) % 4 == 0 ); + assert( mach->ImmLimit + (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4 <= 256 ); + + for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size - 1; i++ ) { + mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } + mach->ImmLimit += (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4; + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + assert( labels->count < 128 ); + + labels->labels[labels->count][0] = instno; + labels->labels[labels->count][1] = pointer; + labels->count++; + break; + + default: + assert( 0 ); + } + } + tgsi_parse_free (&parse); +} + +void +tgsi_exec_machine_run( + struct tgsi_exec_machine *mach ) +{ + struct tgsi_exec_labels labels; + + tgsi_exec_prepare( mach, &labels ); + tgsi_exec_machine_run2( mach, &labels ); +} + +static void +micro_abs( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) fabs( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) fabs( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) fabs( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) fabs( (GLdouble) src->f[3] ); +} + +static void +micro_add( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] + src1->f[0]; + dst->f[1] = src0->f[1] + src1->f[1]; + dst->f[2] = src0->f[2] + src1->f[2]; + dst->f[3] = src0->f[3] + src1->f[3]; +} + +static void +micro_iadd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] + src1->i[0]; + dst->i[1] = src0->i[1] + src1->i[1]; + dst->i[2] = src0->i[2] + src1->i[2]; + dst->i[3] = src0->i[3] + src1->i[3]; +} + +static void +micro_and( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] & src1->u[0]; + dst->u[1] = src0->u[1] & src1->u[1]; + dst->u[2] = src0->u[2] & src1->u[2]; + dst->u[3] = src0->u[3] & src1->u[3]; +} + +static void +micro_ceil( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) ceil( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) ceil( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) ceil( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) ceil( (GLdouble) src->f[3] ); +} + +static void +micro_cos( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) cos( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) cos( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) cos( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) cos( (GLdouble) src->f[3] ); +} + +static void +micro_ddx( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_ddy( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_div( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] / src1->f[0]; + dst->f[1] = src0->f[1] / src1->f[1]; + dst->f[2] = src0->f[2] / src1->f[2]; + dst->f[3] = src0->f[3] / src1->f[3]; +} + +static void +micro_udiv( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] / src1->u[0]; + dst->u[1] = src0->u[1] / src1->u[1]; + dst->u[2] = src0->u[2] / src1->u[2]; + dst->u[3] = src0->u[3] / src1->u[3]; +} + +static void +micro_eq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ieq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_exp2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) pow( 2.0, (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) pow( 2.0, (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) pow( 2.0, (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) pow( 2.0, (GLdouble) src->f[3] ); +} + +static void +micro_f2it( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = (GLint) src->f[0]; + dst->i[1] = (GLint) src->f[1]; + dst->i[2] = (GLint) src->f[2]; + dst->i[3] = (GLint) src->f[3]; +} + +static void +micro_f2ut( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = (GLuint) src->f[0]; + dst->u[1] = (GLuint) src->f[1]; + dst->u[2] = (GLuint) src->f[2]; + dst->u[3] = (GLuint) src->f[3]; +} + +static void +micro_flr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) floor( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) floor( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) floor( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) floor( (GLdouble) src->f[3] ); +} + +static void +micro_frc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = src->f[0] - (GLfloat) floor( (GLdouble) src->f[0] ); + dst->f[1] = src->f[1] - (GLfloat) floor( (GLdouble) src->f[1] ); + dst->f[2] = src->f[2] - (GLfloat) floor( (GLdouble) src->f[2] ); + dst->f[3] = src->f[3] - (GLfloat) floor( (GLdouble) src->f[3] ); +} + +static void +micro_ge( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_i2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) src->i[0]; + dst->f[1] = (GLfloat) src->i[1]; + dst->f[2] = (GLfloat) src->i[2]; + dst->f[3] = (GLfloat) src->i[3]; +} + +static void +micro_lg2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) log( (GLdouble) src->f[0] ) * 1.442695f; + dst->f[1] = (GLfloat) log( (GLdouble) src->f[1] ) * 1.442695f; + dst->f[2] = (GLfloat) log( (GLdouble) src->f[2] ) * 1.442695f; + dst->f[3] = (GLfloat) log( (GLdouble) src->f[3] ) * 1.442695f; +} + +static void +micro_lt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ilt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_ult( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; +} + +static void +micro_max( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_min( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_umod( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] % src1->u[0]; + dst->u[1] = src0->u[1] % src1->u[1]; + dst->u[2] = src0->u[2] % src1->u[2]; + dst->u[3] = src0->u[3] % src1->u[3]; +} + +static void +micro_mul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] * src1->f[0]; + dst->f[1] = src0->f[1] * src1->f[1]; + dst->f[2] = src0->f[2] * src1->f[2]; + dst->f[3] = src0->f[3] * src1->f[3]; +} + +static void +micro_imul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] * src1->i[0]; + dst->i[1] = src0->i[1] * src1->i[1]; + dst->i[2] = src0->i[2] * src1->i[2]; + dst->i[3] = src0->i[3] * src1->i[3]; +} + +static void +micro_imul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->i[0] = src0->i[0] * src1->i[0]; + dst1->i[1] = src0->i[1] * src1->i[1]; + dst1->i[2] = src0->i[2] * src1->i[2]; + dst1->i[3] = src0->i[3] * src1->i[3]; + dst0->i[0] = 0; + dst0->i[1] = 0; + dst0->i[2] = 0; + dst0->i[3] = 0; +} + +static void +micro_umul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->u[0] = src0->u[0] * src1->u[0]; + dst1->u[1] = src0->u[1] * src1->u[1]; + dst1->u[2] = src0->u[2] * src1->u[2]; + dst1->u[3] = src0->u[3] * src1->u[3]; + dst0->u[0] = 0; + dst0->u[1] = 0; + dst0->u[2] = 0; + dst0->u[3] = 0; +} + +static void +micro_movc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2 ) +{ + dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; + dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; + dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; + dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; +} + +static void +micro_neg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = -src->f[0]; + dst->f[1] = -src->f[1]; + dst->f[2] = -src->f[2]; + dst->f[3] = -src->f[3]; +} + +static void +micro_ineg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = -src->i[0]; + dst->i[1] = -src->i[1]; + dst->i[2] = -src->i[2]; + dst->i[3] = -src->i[3]; +} + +static void +micro_not( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = ~src->u[0]; + dst->u[1] = ~src->u[1]; + dst->u[2] = ~src->u[2]; + dst->u[3] = ~src->u[3]; +} + +static void +micro_or( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] | src1->u[0]; + dst->u[1] = src0->u[1] | src1->u[1]; + dst->u[2] = src0->u[2] | src1->u[2]; + dst->u[3] = src0->u[3] | src1->u[3]; +} + +static void +micro_pow( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = (GLfloat) pow( (GLdouble) src0->f[0], (GLdouble) src1->f[0] ); + dst->f[1] = (GLfloat) pow( (GLdouble) src0->f[1], (GLdouble) src1->f[1] ); + dst->f[2] = (GLfloat) pow( (GLdouble) src0->f[2], (GLdouble) src1->f[2] ); + dst->f[3] = (GLfloat) pow( (GLdouble) src0->f[3], (GLdouble) src1->f[3] ); +} + +static void +micro_rnd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) floor( (GLdouble) (src->f[0] + 0.5f) ); + dst->f[1] = (GLfloat) floor( (GLdouble) (src->f[1] + 0.5f) ); + dst->f[2] = (GLfloat) floor( (GLdouble) (src->f[2] + 0.5f) ); + dst->f[3] = (GLfloat) floor( (GLdouble) (src->f[3] + 0.5f) ); +} + +static void +micro_shl( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] << src1->i[0]; + dst->i[1] = src0->i[1] << src1->i[1]; + dst->i[2] = src0->i[2] << src1->i[2]; + dst->i[3] = src0->i[3] << src1->i[3]; +} + +static void +micro_ishr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] >> src1->i[0]; + dst->i[1] = src0->i[1] >> src1->i[1]; + dst->i[2] = src0->i[2] >> src1->i[2]; + dst->i[3] = src0->i[3] >> src1->i[3]; +} + +static void +micro_ushr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] >> src1->u[0]; + dst->u[1] = src0->u[1] >> src1->u[1]; + dst->u[2] = src0->u[2] >> src1->u[2]; + dst->u[3] = src0->u[3] >> src1->u[3]; +} + +static void +micro_sin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) sin( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) sin( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) sin( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) sin( (GLdouble) src->f[3] ); +} + +static void +micro_sqrt( union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) sqrt( (GLdouble) src->f[0] ); + dst->f[1] = (GLfloat) sqrt( (GLdouble) src->f[1] ); + dst->f[2] = (GLfloat) sqrt( (GLdouble) src->f[2] ); + dst->f[3] = (GLfloat) sqrt( (GLdouble) src->f[3] ); +} + +static void +micro_sub( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] - src1->f[0]; + dst->f[1] = src0->f[1] - src1->f[1]; + dst->f[2] = src0->f[2] - src1->f[2]; + dst->f[3] = src0->f[3] - src1->f[3]; +} + +static void +micro_u2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (GLfloat) src->u[0]; + dst->f[1] = (GLfloat) src->u[1]; + dst->f[2] = (GLfloat) src->u[2]; + dst->f[3] = (GLfloat) src->u[3]; +} + +static void +micro_xor( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] ^ src1->u[0]; + dst->u[1] = src0->u[1] ^ src1->u[1]; + dst->u[2] = src0->u[2] ^ src1->u[2]; + dst->u[3] = src0->u[3] ^ src1->u[3]; +} + +static void +fetch_src_file_channel( + const struct tgsi_exec_machine *mach, + const GLuint file, + const GLuint swizzle, + const union tgsi_exec_channel *index, + union tgsi_exec_channel *chan ) +{ + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( file ) { + case TGSI_FILE_CONSTANT: + chan->f[0] = mach->Consts[index->i[0]][swizzle]; + chan->f[1] = mach->Consts[index->i[1]][swizzle]; + chan->f[2] = mach->Consts[index->i[2]][swizzle]; + chan->f[3] = mach->Consts[index->i[3]][swizzle]; + break; + + case TGSI_FILE_INPUT: + chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_TEMPORARY: + chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_IMMEDIATE: + assert( index->i[0] < (GLint) mach->ImmLimit ); + chan->f[0] = mach->Imms[index->i[0]][swizzle]; + assert( index->i[1] < (GLint) mach->ImmLimit ); + chan->f[1] = mach->Imms[index->i[1]][swizzle]; + assert( index->i[2] < (GLint) mach->ImmLimit ); + chan->f[2] = mach->Imms[index->i[2]][swizzle]; + assert( index->i[3] < (GLint) mach->ImmLimit ); + chan->f[3] = mach->Imms[index->i[3]][swizzle]; + break; + + case TGSI_FILE_ADDRESS: + chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; + break; + + case TGSI_EXTSWIZZLE_ONE: + *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; + break; + + default: + assert( 0 ); + } +} + +static void +fetch_source( + const struct tgsi_exec_machine *mach, + union tgsi_exec_channel *chan, + const struct tgsi_full_src_register *reg, + const GLuint chan_index ) +{ + union tgsi_exec_channel index; + GLuint swizzle; + + index.i[0] = + index.i[1] = + index.i[2] = + index.i[3] = reg->SrcRegister.Index; + + if (reg->SrcRegister.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + + if( reg->SrcRegister.Dimension ) { + switch( reg->SrcRegister.File ) { + case TGSI_FILE_INPUT: + index.i[0] *= 17; + index.i[1] *= 17; + index.i[2] *= 17; + index.i[3] *= 17; + break; + case TGSI_FILE_CONSTANT: + index.i[0] *= 4096; + index.i[1] *= 4096; + index.i[2] *= 4096; + index.i[3] *= 4096; + break; + default: + assert( 0 ); + } + + index.i[0] += reg->SrcRegisterDim.Index; + index.i[1] += reg->SrcRegisterDim.Index; + index.i[2] += reg->SrcRegisterDim.Index; + index.i[3] += reg->SrcRegisterDim.Index; + + if (reg->SrcRegisterDim.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterDimInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterDimInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + } + + swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + fetch_src_file_channel( + mach, + reg->SrcRegister.File, + swizzle, + &index, + chan ); + + switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { + case TGSI_UTIL_SIGN_CLEAR: + micro_abs( chan, chan ); + break; + + case TGSI_UTIL_SIGN_SET: + micro_abs( chan, chan ); + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } +} + +static void +store_dest( + struct tgsi_exec_machine *mach, + const union tgsi_exec_channel *chan, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + GLuint chan_index ) +{ + union tgsi_exec_channel *dst; + + switch( reg->DstRegister.File ) { + case TGSI_FILE_NULL: + return; + + case TGSI_FILE_OUTPUT: + dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_TEMPORARY: + dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_ADDRESS: + dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; + break; + + default: + assert( 0 ); + } + + switch (inst->Instruction.Saturate) + { + case TGSI_SAT_NONE: + *dst = *chan; + break; + + case TGSI_SAT_ZERO_ONE: + micro_lt( dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], chan ); + micro_lt( dst, chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + + default: + assert( 0 ); + } +} + +#define FETCH(VAL,INDEX,CHAN)\ + fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) + +#define STORE(VAL,INDEX,CHAN)\ + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN) + +static void +exec_kil (struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + GLuint uniquemask; + GLuint chan_index; + GLuint kilmask = 0; + union tgsi_exec_channel r[1]; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + GLuint swizzle; + GLuint i; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle ( + &inst->FullSrcRegisters[0], + chan_index); + + /* check if the component has not been already tested */ + if (uniquemask & (1 << swizzle)) + continue; + uniquemask |= 1 << swizzle; + + FETCH(&r[0], 0, chan_index); + for (i = 0; i < 4; i++) + if (r[0].f[i] < 0.0f) + kilmask |= 1 << (i * 4); + } + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; +} + + + + +/* + * Fetch a texel using STR texture coordinates. + */ +static void +fetch_texel( struct tgsi_sampler *sampler, + const union tgsi_exec_channel *s, + const union tgsi_exec_channel *t, + const union tgsi_exec_channel *p, + float lodbias, + union tgsi_exec_channel *r, + union tgsi_exec_channel *g, + union tgsi_exec_channel *b, + union tgsi_exec_channel *a ) +{ + GLuint j; + GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]; + + sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); + + for (j = 0; j < 4; j++) { + r->f[j] = rgba[0][j]; + g->f[j] = rgba[1][j]; + b->f[j] = rgba[2][j]; + a->f[j] = rgba[3][j]; + } +} + +static void +constant_interpolation( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; + } +} + +static void +linear_interpolation( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = mach->Inputs[0].xyzw[0].f[i]; + const float y = mach->Inputs[0].xyzw[1].f[i]; + + mach->Inputs[attrib].xyzw[chan].f[i] = + mach->InterpCoefs[attrib].a0[chan] + + mach->InterpCoefs[attrib].dadx[chan] * x + + mach->InterpCoefs[attrib].dady[chan] * y; + } +} + +static void +perspective_interpolation( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = mach->Inputs[0].xyzw[0].f[i]; + const float y = mach->Inputs[0].xyzw[1].f[i]; + // WPOS.w here is really 1/w + const float w = 1.0f / mach->Inputs[0].xyzw[3].f[i]; + + mach->Inputs[attrib].xyzw[chan].f[i] = + (mach->InterpCoefs[attrib].a0[chan] + + mach->InterpCoefs[attrib].dadx[chan] * x + + mach->InterpCoefs[attrib].dady[chan] * y) * w; + } +} + +typedef void (* interpolation_func)( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ); + +static void +exec_declaration( + struct tgsi_exec_machine *mach, + const struct tgsi_full_declaration *decl ) +{ + if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + interpolation_func interp; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if( first == 0 ) { + mask &= ~TGSI_WRITEMASK_XY; + if( mask == TGSI_WRITEMASK_NONE ) { + first++; + if( first > last ) { + return; + } + } + } + + switch( decl->Interpolation.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + interp = constant_interpolation; + break; + + case TGSI_INTERPOLATE_LINEAR: + interp = linear_interpolation; + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + interp = perspective_interpolation; + break; + + default: + assert( 0 ); + } + + if( mask == TGSI_WRITEMASK_XYZW ) { + unsigned i, j; + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + interp( mach, i, j ); + } + } + } + else { + unsigned i, j; + + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + interp( mach, i, j ); + } + } + } + } + } + } +} + +static void +exec_instruction( + struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + struct tgsi_exec_labels *labels, + GLuint *programCounter ) +{ + GLuint chan_index; + union tgsi_exec_channel r[8]; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_f2it( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_X ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[1], 0, CHAN_Y ); + micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + + FETCH( &r[2], 0, CHAN_W ); + micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); + micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); + micro_pow( &r[1], &r[1], &r[2] ); + micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Z ); + } + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( &r[0], 0, CHAN_X ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( &r[0], 0, CHAN_X ); + micro_sqrt( &r[0], &r[0] ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + assert (0); + break; + + case TGSI_OPCODE_LOG: + assert (0); + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_mul( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Z ); + FETCH( &r[2], 1, CHAN_Z ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_W); + FETCH(&r[2], 1, CHAN_W); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + FETCH( &r[0], 0, CHAN_Y ); + FETCH( &r[1], 1, CHAN_Y); + micro_mul( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_Z ); + STORE( &r[0], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + FETCH( &r[0], 1, CHAN_W ); + STORE( &r[0], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_mul( &r[0], &r[0], &r[1] ); + FETCH( &r[1], 2, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_sub( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_sub( &r[1], &r[1], &r[2] ); + micro_mul( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_CND: + assert (0); + break; + + case TGSI_OPCODE_CND0: + assert (0); + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert (0); + break; + + case TGSI_OPCODE_INDEX: + assert (0); + break; + + case TGSI_OPCODE_NEGATE: + assert (0); + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_frc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + assert (0); + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_flr( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_rnd( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH(&r[0], 0, CHAN_X); + + micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( &r[0], 0, CHAN_X ); + micro_lg2( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_pow( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + FETCH(&r[0], 0, CHAN_Y); + FETCH(&r[1], 1, CHAN_Z); + + micro_mul( &r[2], &r[0], &r[1] ); + + FETCH(&r[3], 0, CHAN_Z); + FETCH(&r[4], 1, CHAN_Y); + + micro_mul( &r[5], &r[3], &r[4] ); + micro_sub( &r[2], &r[2], &r[5] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &r[2], 0, CHAN_X ); + } + + FETCH(&r[2], 1, CHAN_X); + + micro_mul( &r[3], &r[3], &r[2] ); + + FETCH(&r[5], 0, CHAN_X); + + micro_mul( &r[1], &r[1], &r[5] ); + micro_sub( &r[3], &r[3], &r[1] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + STORE( &r[3], 0, CHAN_Y ); + } + + micro_mul( &r[5], &r[5], &r[4] ); + micro_mul( &r[0], &r[0], &r[2] ); + micro_sub( &r[5], &r[5], &r[0] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( &r[5], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + assert (0); + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + + micro_abs( &r[0], &r[0] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_RCC: + assert (0); + break; + + case TGSI_OPCODE_DPH: + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 1, CHAN_W); + + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH(&r[0], 0, CHAN_X); + + micro_cos( &r[0], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddx( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDY: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddy( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_KIL: + exec_kil (mach, inst); + break; + + case TGSI_OPCODE_PK2H: + assert (0); + break; + + case TGSI_OPCODE_PK2US: + assert (0); + break; + + case TGSI_OPCODE_PK4B: + assert (0); + break; + + case TGSI_OPCODE_PK4UB: + assert (0); + break; + + case TGSI_OPCODE_RFL: + assert (0); + break; + + case TGSI_OPCODE_SEQ: + assert (0); + break; + + case TGSI_OPCODE_SFL: + assert (0); + break; + + case TGSI_OPCODE_SGT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SIN: + FETCH( &r[0], 0, CHAN_X ); + micro_sin( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SNE: + assert (0); + break; + + case TGSI_OPCODE_STR: + assert (0); + break; + + case TGSI_OPCODE_TEX: + { + const GLuint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: + + FETCH(&r[0], 0, CHAN_X); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[1], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[1] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + fetch_texel(&mach->Samplers[unit], + &r[0], NULL, NULL, 0.0, + &r[0], &r[1], &r[2], &r[3]); + break; + + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[2], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[2] ); + micro_div( &r[1], &r[1], &r[2] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], NULL, 0.0, + &r[0], &r[1], &r[2], &r[3]); + break; + + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], 0.0, + &r[0], &r[1], &r[2], &r[3]); + break; + + default: + assert (0); + } + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } + } + break; + + case TGSI_OPCODE_TXD: + assert (0); + break; + + case TGSI_OPCODE_UP2H: + assert (0); + break; + + case TGSI_OPCODE_UP2US: + assert (0); + break; + + case TGSI_OPCODE_UP4B: + assert (0); + break; + + case TGSI_OPCODE_UP4UB: + assert (0); + break; + + case TGSI_OPCODE_X2D: + assert (0); + break; + + case TGSI_OPCODE_ARA: + assert (0); + break; + + case TGSI_OPCODE_ARR: + assert (0); + break; + + case TGSI_OPCODE_BRA: + assert (0); + break; + + case TGSI_OPCODE_CAL: + assert (0); + break; + + case TGSI_OPCODE_RET: + /* XXX: end of shader! */ + /*assert (0);*/ + break; + + case TGSI_OPCODE_SSG: + assert (0); + break; + + case TGSI_OPCODE_CMP: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_SCS: + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( &r[0], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + micro_cos( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + micro_sin( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_Y ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_TXB: + assert (0); + break; + + case TGSI_OPCODE_NRM: + assert (0); + break; + + case TGSI_OPCODE_DIV: + assert( 0 ); + break; + + case TGSI_OPCODE_DP2: + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TXL: + assert (0); + break; + + case TGSI_OPCODE_BRK: + assert (0); + break; + + case TGSI_OPCODE_IF: + { + GLuint cond = 0; + struct tgsi_exec_cond_state *state; + + /* Allocate condition state. */ + assert( mach->CondStack.Index > 0 ); + mach->CondStack.Index--; + + /* Evaluate the condition mask. */ + FETCH( &r[0], 0, CHAN_X ); + if( r[0].u[0] ) { + cond |= 1; + } + if( r[0].u[1] ) { + cond |= 2; + } + if( r[0].u[2] ) { + cond |= 4; + } + if( r[0].u[3] ) { + cond |= 8; + } + + state = &mach->CondStack.States[mach->CondStack.Index]; + + /* Initialize the If portion of condition state. */ + memcpy( + state->IfPortion.TempsAddrs, + mach->Temps, + sizeof( state->IfPortion.TempsAddrs ) ); + memcpy( + state->IfPortion.Outputs, + mach->Outputs, + sizeof( state->IfPortion.Outputs ) ); + state->Condition = cond; + state->WasElse = GL_FALSE; + } + break; + + case TGSI_OPCODE_LOOP: + assert (0); + break; + + case TGSI_OPCODE_REP: + assert (0); + break; + + case TGSI_OPCODE_ELSE: + { + struct tgsi_exec_cond_state *state; + struct tgsi_exec_cond_regs temp; + + state = &mach->CondStack.States[mach->CondStack.Index]; + + /* Copy the results of the If portion to temporary storage. */ + memcpy( + temp.TempsAddrs, + mach->Temps, + sizeof( temp.TempsAddrs ) ); + memcpy( + temp.Outputs, + mach->Outputs, + sizeof( temp.Outputs ) ); + + /* Restore the state of registers from before the If statement. */ + memcpy( + mach->Temps, + state->IfPortion.TempsAddrs, + sizeof( state->IfPortion.TempsAddrs ) ); + memcpy( + mach->Outputs, + state->IfPortion.Outputs, + sizeof( state->IfPortion.Outputs ) ); + + /* Save the results of If portion. */ + memcpy( + &state->IfPortion, + &temp, + sizeof( state->IfPortion ) ); + state->WasElse = GL_TRUE; + } + break; + + case TGSI_OPCODE_ENDIF: + { + struct tgsi_exec_cond_state *state; + GLuint i; + + state = &mach->CondStack.States[mach->CondStack.Index]; + + if( state->WasElse ) { + /* Save the results of Else portion. */ + memcpy( + state->ElsePortion.TempsAddrs, + mach->Temps, + sizeof( state->ElsePortion.TempsAddrs ) ); + memcpy( + state->ElsePortion.Outputs, + mach->Outputs, + sizeof( state->ElsePortion.Outputs ) ); + } + else { + /* Copy the state of registers from before the If statement to Else portion. */ + memcpy( + &state->ElsePortion, + &state->IfPortion, + sizeof( state->ElsePortion ) ); + + /* Save the results of the If portion. */ + memcpy( + state->IfPortion.TempsAddrs, + mach->Temps, + sizeof( state->IfPortion.TempsAddrs ) ); + memcpy( + state->IfPortion.Outputs, + mach->Outputs, + sizeof( state->IfPortion.Outputs ) ); + } + + /* Mix the If and Else portions based on condition mask. */ + for( i = 0; i < 4; i++ ) { + struct tgsi_exec_cond_regs *regs; + GLuint j; + + if( state->Condition & (1 << i) ) { + regs = &state->IfPortion; + } + else { + regs = &state->ElsePortion; + } + + for( j = 0; j < TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS; j++ ) { + mach->Temps[j].xyzw[0].u[i] = regs->TempsAddrs[j].xyzw[0].u[i]; + mach->Temps[j].xyzw[1].u[i] = regs->TempsAddrs[j].xyzw[1].u[i]; + mach->Temps[j].xyzw[2].u[i] = regs->TempsAddrs[j].xyzw[2].u[i]; + mach->Temps[j].xyzw[3].u[i] = regs->TempsAddrs[j].xyzw[3].u[i]; + } + for( j = 0; j < 2; j++ ) { + mach->Outputs[j].xyzw[0].u[i] = regs->Outputs[j].xyzw[0].u[i]; + mach->Outputs[j].xyzw[1].u[i] = regs->Outputs[j].xyzw[1].u[i]; + mach->Outputs[j].xyzw[2].u[i] = regs->Outputs[j].xyzw[2].u[i]; + mach->Outputs[j].xyzw[3].u[i] = regs->Outputs[j].xyzw[3].u[i]; + } + } + + /* Release condition state. */ + mach->CondStack.Index++; + } + break; + + case TGSI_OPCODE_ENDLOOP: + assert (0); + break; + + case TGSI_OPCODE_ENDREP: + assert (0); + break; + + case TGSI_OPCODE_PUSHA: + assert (0); + break; + + case TGSI_OPCODE_POPA: + assert (0); + break; + + case TGSI_OPCODE_CEIL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ceil( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_I2F: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_i2f( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_NOT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_not( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TRUNC: + /* TGSI_OPCODE_INT */ + assert (0); + break; + + case TGSI_OPCODE_SHL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_shl( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ishr( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_AND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_and( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_OR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_or( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOD: + assert (0); + break; + + case TGSI_OPCODE_XOR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_xor( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SAD: + assert (0); + break; + + case TGSI_OPCODE_TXF: + assert (0); + break; + + case TGSI_OPCODE_TXQ: + assert (0); + break; + + case TGSI_OPCODE_CONT: + assert (0); + break; + + case TGSI_OPCODE_EMIT: + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; + break; + + case TGSI_OPCODE_ENDPRIM: + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; + break; + + case TGSI_OPCODE_BGNLOOP2: + assert( 0 ); + break; + + case TGSI_OPCODE_BGNSUB: + assert( 0 ); + break; + + case TGSI_OPCODE_ENDLOOP2: + assert( 0 ); + break; + + case TGSI_OPCODE_ENDSUB: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE1: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE2: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE3: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE4: + assert( 0 ); + break; + + case TGSI_OPCODE_NOP: + break; + + default: + assert( 0 ); + } +} + + +#if !defined(XSTDCALL) +#if defined(WIN32) +#define XSTDCALL __stdcall +#else +#define XSTDCALL +#endif +#endif + +typedef void (XSTDCALL *fp_function) (const struct tgsi_exec_vector *input, + struct tgsi_exec_vector *output, + GLfloat (*constant)[4], + struct tgsi_exec_vector *temporary); + +void +tgsi_exec_machine_run2( + struct tgsi_exec_machine *mach, + struct tgsi_exec_labels *labels ) +{ +#if MESA + GET_CURRENT_CONTEXT(ctx); + GLuint i; +#endif + +#if XXX_SSE + fp_function function; + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + + function = (fp_function) x86_get_func (&mach->Function); + + function (mach->Inputs, + mach->Outputs, + mach->Consts, + mach->Temps); +#else + struct tgsi_parse_context parse; + GLuint k; + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; + + if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; + mach->Primitives[0] = 0; + } + + mach->CondStack.Index = 8; + + k = tgsi_parse_init( &parse, mach->Tokens ); + if (k != TGSI_PARSE_OK) { + printf("Problem parsing!\n"); + return; + } + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + exec_declaration( mach, &parse.FullToken.FullDeclaration ); + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + exec_instruction( mach, &parse.FullToken.FullInstruction, labels, &parse.Position ); + break; + default: + assert( 0 ); + } + } + tgsi_parse_free (&parse); +#endif + +#if 0 + /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ + if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { + /* + * Scale back depth component. + */ + for (i = 0; i < 4; i++) + mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; + } +#endif +} + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h new file mode 100644 index 0000000000..e5e8c3608e --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -0,0 +1,191 @@ +#if !defined TGSI_EXEC_H +#define TGSI_EXEC_H + +#include "pipe/p_compiler.h" + +#if 0 +#include "x86/rtasm/x86sse.h" +#endif + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +#define NUM_CHANNELS 4 /* R,G,B,A */ +#define QUAD_SIZE 4 /* 4 pixel/quad */ + +union tgsi_exec_channel +{ + float f[QUAD_SIZE]; + int i[QUAD_SIZE]; + unsigned u[QUAD_SIZE]; +}; + +struct tgsi_exec_vector +{ + union tgsi_exec_channel xyzw[NUM_CHANNELS]; +}; + +struct tgsi_interp_coef +{ + float a0[NUM_CHANNELS]; /* in an xyzw layout */ + float dadx[NUM_CHANNELS]; + float dady[NUM_CHANNELS]; +}; + +#define TEX_CACHE_TILE_SIZE 8 +#define TEX_CACHE_NUM_ENTRIES 8 + +struct tgsi_texture_cache_entry +{ + int x, y, face, level, zslice; + float data[TEX_CACHE_TILE_SIZE][TEX_CACHE_TILE_SIZE][4]; +}; + +struct tgsi_sampler +{ + const struct pipe_sampler_state *state; + struct pipe_mipmap_tree *texture; + /** Get samples for four fragments in a quad */ + void (*get_samples)(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); + void *pipe; /*XXX temporary*/ + struct tgsi_texture_cache_entry cache[TEX_CACHE_NUM_ENTRIES]; +}; + +struct tgsi_exec_labels +{ + unsigned labels[128][2]; + unsigned count; +}; + +#define TGSI_EXEC_TEMP_00000000_I 32 +#define TGSI_EXEC_TEMP_00000000_C 0 + +#define TGSI_EXEC_TEMP_7FFFFFFF_I 32 +#define TGSI_EXEC_TEMP_7FFFFFFF_C 1 + +#define TGSI_EXEC_TEMP_80000000_I 32 +#define TGSI_EXEC_TEMP_80000000_C 2 + +#define TGSI_EXEC_TEMP_FFFFFFFF_I 32 +#define TGSI_EXEC_TEMP_FFFFFFFF_C 3 + +#define TGSI_EXEC_TEMP_ONE_I 33 +#define TGSI_EXEC_TEMP_ONE_C 0 + +#define TGSI_EXEC_TEMP_TWO_I 33 +#define TGSI_EXEC_TEMP_TWO_C 1 + +#define TGSI_EXEC_TEMP_128_I 33 +#define TGSI_EXEC_TEMP_128_C 2 + +#define TGSI_EXEC_TEMP_MINUS_128_I 33 +#define TGSI_EXEC_TEMP_MINUS_128_C 3 + +#define TGSI_EXEC_TEMP_KILMASK_I 34 +#define TGSI_EXEC_TEMP_KILMASK_C 0 + +#define TGSI_EXEC_TEMP_OUTPUT_I 34 +#define TGSI_EXEC_TEMP_OUTPUT_C 1 + +#define TGSI_EXEC_TEMP_PRIMITIVE_I 34 +#define TGSI_EXEC_TEMP_PRIMITIVE_C 2 + +#define TGSI_EXEC_TEMP_R0 35 + +#define TGSI_EXEC_NUM_TEMPS (32 + 4) +#define TGSI_EXEC_NUM_ADDRS 1 + +/* XXX: This is temporary */ +struct tgsi_exec_cond_regs +{ + struct tgsi_exec_vector TempsAddrs[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS]; + struct tgsi_exec_vector Outputs[2]; /* XXX: That's just enough for fragment shader only! */ +}; + +/* XXX: This is temporary */ +struct tgsi_exec_cond_state +{ + struct tgsi_exec_cond_regs IfPortion; + struct tgsi_exec_cond_regs ElsePortion; + unsigned Condition; + boolean WasElse; +}; + +/* XXX: This is temporary */ +struct tgsi_exec_cond_stack +{ + struct tgsi_exec_cond_state States[8]; + unsigned Index; /* into States[] */ +}; + +struct tgsi_exec_machine +{ + /* + * 32 program temporaries + * 4 internal temporaries + * 1 address + * 1 temporary of padding to align to 16 bytes + */ + struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS + 1]; + + /* + * This will point to _Temps after aligning to 16B boundary. + */ + struct tgsi_exec_vector *Temps; + struct tgsi_exec_vector *Addrs; + + struct tgsi_sampler *Samplers; + + float Imms[256][4]; + unsigned ImmLimit; + float (*Consts)[4]; + struct tgsi_exec_vector *Inputs; + struct tgsi_exec_vector *Outputs; + const struct tgsi_token *Tokens; + unsigned Processor; + + /* GEOMETRY processor only. */ + unsigned *Primitives; + + /* FRAGMENT processor only. */ + const struct tgsi_interp_coef *InterpCoefs; + + struct tgsi_exec_cond_stack CondStack; +#if XXX_SSE + struct x86_function Function; +#endif +}; + +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach, + const struct tgsi_token *tokens, + unsigned numSamplers, + struct tgsi_sampler *samplers); + +void +tgsi_exec_prepare( + struct tgsi_exec_machine *mach, + struct tgsi_exec_labels *labels ); + +void +tgsi_exec_machine_run( + struct tgsi_exec_machine *mach ); + +void +tgsi_exec_machine_run2( + struct tgsi_exec_machine *mach, + struct tgsi_exec_labels *labels ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_EXEC_H + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_parse.c b/src/mesa/pipe/tgsi/exec/tgsi_parse.c new file mode 100644 index 0000000000..a6724d762c --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_parse.c @@ -0,0 +1,288 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ) +{ + full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; +} + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ) +{ + if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) + free( full_token->FullImmediate.u.Pointer ); +} + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ) +{ + ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; + if( ctx->FullVersion.Version.MajorVersion > 1 ) { + return TGSI_PARSE_ERROR; + } + + ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; + if( ctx->FullHeader.Header.HeaderSize >= 2 ) { + ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; + } + else { + ctx->FullHeader.Processor = tgsi_default_processor(); + } + + ctx->Tokens = tokens; + ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; + + tgsi_full_token_init( &ctx->FullToken ); + + return TGSI_PARSE_OK; +} + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ) +{ + tgsi_full_token_free( &ctx->FullToken ); +} + +GLuint +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ) +{ + return ctx->Position >= + 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; +} + +static void +next_token( + struct tgsi_parse_context *ctx, + void *token ) +{ + assert( !tgsi_parse_end_of_tokens( ctx ) ); + + *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; +} + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ) +{ + struct tgsi_token token; + unsigned i; + + tgsi_full_token_free( &ctx->FullToken ); + tgsi_full_token_init( &ctx->FullToken ); + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; + + *decl = tgsi_default_full_declaration(); + decl->Declaration = *(struct tgsi_declaration *) &token; + + switch( decl->Declaration.Type ) { + case TGSI_DECLARE_RANGE: + next_token( ctx, &decl->u.DeclarationRange ); + break; + + case TGSI_DECLARE_MASK: + next_token( ctx, &decl->u.DeclarationMask ); + break; + + default: + assert (0); + } + + if( decl->Declaration.Interpolate ) { + next_token( ctx, &decl->Interpolation ); + } + + if( decl->Declaration.Semantic ) { + next_token( ctx, &decl->Semantic ); + } + + break; + } + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; + + *imm = tgsi_default_full_immediate(); + imm->Immediate = *(struct tgsi_immediate *) &token; + + assert( !imm->Immediate.Extended ); + + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + imm->u.Pointer = malloc( + sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + next_token( ctx, &imm->u.ImmediateFloat32[i] ); + } + break; + + default: + assert( 0 ); + } + + break; + } + + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; + unsigned extended; + + *inst = tgsi_default_full_instruction(); + inst->Instruction = *(struct tgsi_instruction *) &token; + + extended = inst->Instruction.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_INSTRUCTION_EXT_TYPE_NV: + inst->InstructionExtNv = + *(struct tgsi_instruction_ext_nv *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_LABEL: + inst->InstructionExtLabel = + *(struct tgsi_instruction_ext_label *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: + inst->InstructionExtTexture = + *(struct tgsi_instruction_ext_texture *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); + assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); + + extended = inst->FullDstRegisters[i].DstRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: + inst->FullDstRegisters[i].DstRegisterExtConcode = + *(struct tgsi_dst_register_ext_concode *) &token; + break; + + case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: + inst->FullDstRegisters[i].DstRegisterExtModulate = + *(struct tgsi_dst_register_ext_modulate *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + } + + assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); + + extended = inst->FullSrcRegisters[i].SrcRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: + inst->FullSrcRegisters[i].SrcRegisterExtSwz = + *(struct tgsi_src_register_ext_swz *) &token; + break; + + case TGSI_SRC_REGISTER_EXT_TYPE_MOD: + inst->FullSrcRegisters[i].SrcRegisterExtMod = + *(struct tgsi_src_register_ext_mod *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + + if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); + + /* + * No support for multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); + + if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + } + } + + break; + } + + default: + assert( 0 ); + } +} + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_parse.h b/src/mesa/pipe/tgsi/exec/tgsi_parse.h new file mode 100644 index 0000000000..56a8c254c2 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_parse.h @@ -0,0 +1,121 @@ +#if !defined TGSI_PARSE_H +#define TGSI_PARSE_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_full_version +{ + struct tgsi_version Version; +}; + +struct tgsi_full_header +{ + struct tgsi_header Header; + struct tgsi_processor Processor; +}; + +struct tgsi_full_dst_register +{ + struct tgsi_dst_register DstRegister; + struct tgsi_dst_register_ext_concode DstRegisterExtConcode; + struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; +}; + +struct tgsi_full_src_register +{ + struct tgsi_src_register SrcRegister; + struct tgsi_src_register_ext_swz SrcRegisterExtSwz; + struct tgsi_src_register_ext_mod SrcRegisterExtMod; + struct tgsi_src_register SrcRegisterInd; + struct tgsi_dimension SrcRegisterDim; + struct tgsi_src_register SrcRegisterDimInd; +}; + +struct tgsi_full_declaration +{ + struct tgsi_declaration Declaration; + union + { + struct tgsi_declaration_range DeclarationRange; + struct tgsi_declaration_mask DeclarationMask; + } u; + struct tgsi_declaration_interpolation Interpolation; + struct tgsi_declaration_semantic Semantic; +}; + +struct tgsi_full_immediate +{ + struct tgsi_immediate Immediate; + union + { + void *Pointer; + struct tgsi_immediate_float32 *ImmediateFloat32; + } u; +}; + +#define TGSI_FULL_MAX_DST_REGISTERS 2 +#define TGSI_FULL_MAX_SRC_REGISTERS 3 + +struct tgsi_full_instruction +{ + struct tgsi_instruction Instruction; + struct tgsi_instruction_ext_nv InstructionExtNv; + struct tgsi_instruction_ext_label InstructionExtLabel; + struct tgsi_instruction_ext_texture InstructionExtTexture; + struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; + struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; +}; + +union tgsi_full_token +{ + struct tgsi_token Token; + struct tgsi_full_declaration FullDeclaration; + struct tgsi_full_immediate FullImmediate; + struct tgsi_full_instruction FullInstruction; +}; + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ); + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ); + +struct tgsi_parse_context +{ + const struct tgsi_token *Tokens; + unsigned Position; + struct tgsi_full_version FullVersion; + struct tgsi_full_header FullHeader; + union tgsi_full_token FullToken; +}; + +#define TGSI_PARSE_OK 0 +#define TGSI_PARSE_ERROR 1 + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ); + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ); + +unsigned +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ); + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_PARSE_H + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h new file mode 100644 index 0000000000..ca53071a60 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -0,0 +1,1486 @@ +#if !defined TGSI_TOKEN_H +#define TGSI_TOKEN_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_version +{ + unsigned MajorVersion : 8; + unsigned MinorVersion : 8; + unsigned Padding : 16; +}; + +struct tgsi_header +{ + unsigned HeaderSize : 8; + unsigned BodySize : 24; +}; + +#define TGSI_PROCESSOR_FRAGMENT 0 +#define TGSI_PROCESSOR_VERTEX 1 +#define TGSI_PROCESSOR_GEOMETRY 2 + +struct tgsi_processor +{ + unsigned Processor : 4; /* TGSI_PROCESSOR_ */ + unsigned Padding : 28; +}; + +#define TGSI_TOKEN_TYPE_DECLARATION 0 +#define TGSI_TOKEN_TYPE_IMMEDIATE 1 +#define TGSI_TOKEN_TYPE_INSTRUCTION 2 + +struct tgsi_token +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_ */ + unsigned Size : 8; /* UINT */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ +}; + +#define TGSI_FILE_NULL 0 +#define TGSI_FILE_CONSTANT 1 +#define TGSI_FILE_INPUT 2 +#define TGSI_FILE_OUTPUT 3 +#define TGSI_FILE_TEMPORARY 4 +#define TGSI_FILE_SAMPLER 5 +#define TGSI_FILE_ADDRESS 6 +#define TGSI_FILE_IMMEDIATE 7 + +#define TGSI_DECLARE_RANGE 0 +#define TGSI_DECLARE_MASK 1 + +#define TGSI_WRITEMASK_NONE 0x00 +#define TGSI_WRITEMASK_X 0x01 +#define TGSI_WRITEMASK_Y 0x02 +#define TGSI_WRITEMASK_XY 0x03 +#define TGSI_WRITEMASK_Z 0x04 +#define TGSI_WRITEMASK_XZ 0x05 +#define TGSI_WRITEMASK_YZ 0x06 +#define TGSI_WRITEMASK_XYZ 0x07 +#define TGSI_WRITEMASK_W 0x08 +#define TGSI_WRITEMASK_XW 0x09 +#define TGSI_WRITEMASK_YW 0x0A +#define TGSI_WRITEMASK_XYW 0x0B +#define TGSI_WRITEMASK_ZW 0x0C +#define TGSI_WRITEMASK_XZW 0x0D +#define TGSI_WRITEMASK_YZW 0x0E +#define TGSI_WRITEMASK_XYZW 0x0F + +struct tgsi_declaration +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ + unsigned Size : 8; /* UINT */ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned Declare : 4; /* TGSI_DECLARE_ */ + unsigned UsageMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Interpolate : 1; /* BOOL */ + unsigned Semantic : 1; /* BOOL */ + unsigned Padding : 5; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_declaration_range +{ + unsigned First : 16; /* UINT */ + unsigned Last : 16; /* UINT */ +}; + +struct tgsi_declaration_mask +{ + unsigned Mask : 32; /* UINT */ +}; + +#define TGSI_INTERPOLATE_CONSTANT 0 +#define TGSI_INTERPOLATE_LINEAR 1 +#define TGSI_INTERPOLATE_PERSPECTIVE 2 + +struct tgsi_declaration_interpolation +{ + unsigned Interpolate : 4; /* TGSI_INTERPOLATE_ */ + unsigned Padding : 28; +}; + +#define TGSI_SEMANTIC_DEPTH 0 +#define TGSI_SEMANTIC_COLOR 1 + +struct tgsi_declaration_semantic +{ + unsigned SemanticName : 8; /* TGSI_SEMANTIC_ */ + unsigned SemanticIndex : 16; /* UINT */ + unsigned Padding : 8; +}; + +#define TGSI_IMM_FLOAT32 0 + +struct tgsi_immediate +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_IMMEDIATE */ + unsigned Size : 8; /* UINT */ + unsigned DataType : 4; /* TGSI_IMM_ */ + unsigned Padding : 15; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_immediate_float32 +{ + float Float; +}; + +/* + * GL_NV_vertex_program + */ +#define TGSI_OPCODE_ARL 0 +#define TGSI_OPCODE_MOV 1 +#define TGSI_OPCODE_LIT 2 +#define TGSI_OPCODE_RCP 3 +#define TGSI_OPCODE_RSQ 4 +#define TGSI_OPCODE_EXP 5 +#define TGSI_OPCODE_LOG 6 +#define TGSI_OPCODE_MUL 7 +#define TGSI_OPCODE_ADD 8 +#define TGSI_OPCODE_DP3 9 +#define TGSI_OPCODE_DP4 10 +#define TGSI_OPCODE_DST 11 +#define TGSI_OPCODE_MIN 12 +#define TGSI_OPCODE_MAX 13 +#define TGSI_OPCODE_SLT 14 +#define TGSI_OPCODE_SGE 15 +#define TGSI_OPCODE_MAD 16 + +/* + * GL_ATI_fragment_shader + */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_MUL */ +#define TGSI_OPCODE_SUB 17 +#define TGSI_OPCODE_DOT3 TGSI_OPCODE_DP3 +#define TGSI_OPCODE_DOT4 TGSI_OPCODE_DP4 +/* TGSI_OPCODE_MAD */ +#define TGSI_OPCODE_LERP 18 +#define TGSI_OPCODE_CND 19 +#define TGSI_OPCODE_CND0 20 +#define TGSI_OPCODE_DOT2ADD 21 + +/* + * GL_EXT_vertex_shader + */ +#define TGSI_OPCODE_INDEX 22 +#define TGSI_OPCODE_NEGATE 23 +/* TGSI_OPCODE_DOT3 */ +/* TGSI_OPCODE_DOT4 */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_MADD TGSI_OPCODE_MAD +#define TGSI_OPCODE_FRAC 24 +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +#define TGSI_OPCODE_SETGE TGSI_OPCODE_SGE +#define TGSI_OPCODE_SETLT TGSI_OPCODE_SLT +#define TGSI_OPCODE_CLAMP 25 +#define TGSI_OPCODE_FLOOR 26 +#define TGSI_OPCODE_ROUND 27 +#define TGSI_OPCODE_EXPBASE2 28 +#define TGSI_OPCODE_LOGBASE2 29 +#define TGSI_OPCODE_POWER 30 +#define TGSI_OPCODE_RECIP TGSI_OPCODE_RCP +#define TGSI_OPCODE_RECIPSQRT TGSI_OPCODE_RSQ +/* TGSI_OPCODE_SUB */ +#define TGSI_OPCODE_CROSSPRODUCT 31 +#define TGSI_OPCODE_MULTIPLYMATRIX 32 +/* TGSI_OPCODE_MOV */ + +/* + * GL_NV_vertex_program1_1 + */ +/* TGSI_OPCODE_ARL */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_LIT */ +#define TGSI_OPCODE_ABS 33 +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_LOG */ +#define TGSI_OPCODE_RCC 34 +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +#define TGSI_OPCODE_DPH 35 +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ + +/* + * GL_NV_fragment_program + */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_COS 36 +#define TGSI_OPCODE_DDX 37 +#define TGSI_OPCODE_DDY 38 +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DST */ +#define TGSI_OPCODE_EX2 TGSI_OPCODE_EXPBASE2 +#define TGSI_OPCODE_FLR TGSI_OPCODE_FLOOR +#define TGSI_OPCODE_FRC TGSI_OPCODE_FRAC +#define TGSI_OPCODE_KIL 39 +#define TGSI_OPCODE_LG2 TGSI_OPCODE_LOGBASE2 +/* TGSI_OPCODE_LIT */ +#define TGSI_OPCODE_LRP TGSI_OPCODE_LERP +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +#define TGSI_OPCODE_PK2H 40 +#define TGSI_OPCODE_PK2US 41 +#define TGSI_OPCODE_PK4B 42 +#define TGSI_OPCODE_PK4UB 43 +#define TGSI_OPCODE_POW TGSI_OPCODE_POWER +/* TGSI_OPCODE_RCP */ +#define TGSI_OPCODE_RFL 44 +/* TGSI_OPCODE_RSQ */ +#define TGSI_OPCODE_SEQ 45 +#define TGSI_OPCODE_SFL 46 +/* TGSI_OPCODE_SGE */ +#define TGSI_OPCODE_SGT 47 +#define TGSI_OPCODE_SIN 48 +#define TGSI_OPCODE_SLE 49 +/* TGSI_OPCODE_SLT */ +#define TGSI_OPCODE_SNE 50 +#define TGSI_OPCODE_STR 51 +/* TGSI_OPCODE_SUB */ +#define TGSI_OPCODE_TEX 52 +#define TGSI_OPCODE_TXD 53 +#define TGSI_OPCODE_TXP 132 +#define TGSI_OPCODE_UP2H 54 +#define TGSI_OPCODE_UP2US 55 +#define TGSI_OPCODE_UP4B 56 +#define TGSI_OPCODE_UP4UB 57 +#define TGSI_OPCODE_X2D 58 + +/* + * GL_NV_vertex_program2 + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_ARA 59 +/* TGSI_OPCODE_ARL */ +#define TGSI_OPCODE_ARR 60 +#define TGSI_OPCODE_BRA 61 +#define TGSI_OPCODE_CAL 62 +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCC */ +/* TGSI_OPCODE_RCP */ +#define TGSI_OPCODE_RET 63 +/* TGSI_OPCODE_RSQNV - use TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SNE */ +#define TGSI_OPCODE_SSG 64 +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_SUB */ + +/* + * GL_ARB_vertex_program + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_ARL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +#define TGSI_OPCODE_SWZ TGSI_OPCODE_MOV +#define TGSI_OPCODE_XPD TGSI_OPCODE_CROSSPRODUCT + +/* + * GL_ARB_fragment_program + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_CMP 65 +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +#define TGSI_OPCODE_SCS 66 +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXP */ +#define TGSI_OPCODE_TXB 67 +/* TGSI_OPCODE_KIL */ + +/* + * GL_NV_fragment_program_option + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_DDX */ +/* TGSI_OPCODE_DDY */ +/* TGSI_OPCODE_PK2H */ +/* TGSI_OPCODE_PK2US */ +/* TGSI_OPCODE_PK4B */ +/* TGSI_OPCODE_PK4UB */ +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SCS */ +/* TGSI_OPCODE_UP2H */ +/* TGSI_OPCODE_UP2US */ +/* TGSI_OPCODE_UP4B */ +/* TGSI_OPCODE_UP4UB */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_RFL */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_CMP */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_X2D */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXP */ +/* TGSI_OPCODE_TXB */ +/* TGSI_OPCODE_KIL */ +/* TGSI_OPCODE_TXD */ + +/* + * GL_NV_fragment_program2 + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_DDX */ +/* TGSI_OPCODE_DDY */ +/* TGSI_OPCODE_PK2H */ +/* TGSI_OPCODE_PK2US */ +/* TGSI_OPCODE_PK4B */ +/* TGSI_OPCODE_PK4UB */ +#define TGSI_OPCODE_NRM 68 +#define TGSI_OPCODE_DIV 69 +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SCS */ +/* TGSI_OPCODE_UP2H */ +/* TGSI_OPCODE_UP2US */ +/* TGSI_OPCODE_UP4B */ +/* TGSI_OPCODE_UP4UB */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_RFL */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +#define TGSI_OPCODE_DP2 70 +/* TGSI_OPCODE_CMP */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_X2D */ +#define TGSI_OPCODE_DP2A TGSI_OPCODE_DOT2ADD +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXP */ +/* TGSI_OPCODE_TXB */ +#define TGSI_OPCODE_TXL 71 +/* TGSI_OPCODE_KIL */ +/* TGSI_OPCODE_TXD */ +/* TGSI_OPCODE_CAL */ +/* TGSI_OPCODE_RET */ +#define TGSI_OPCODE_BRK 72 +#define TGSI_OPCODE_IF 73 +#define TGSI_OPCODE_LOOP 74 +#define TGSI_OPCODE_REP 75 +#define TGSI_OPCODE_ELSE 76 +#define TGSI_OPCODE_ENDIF 77 +#define TGSI_OPCODE_ENDLOOP 78 +#define TGSI_OPCODE_ENDREP 79 + +/* + * GL_NV_vertex_program2_option + */ +/* TGSI_OPCODE_ARL */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_SSG */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_RCC */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_ARR */ +/* TGSI_OPCODE_ARA */ +/* TGSI_OPCODE_BRA */ +/* TGSI_OPCODE_CAL */ +/* TGSI_OPCODE_RET */ + +/* + * GL_NV_vertex_program3 + */ +/* TGSI_OPCODE_ARL */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_SSG */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_RCC */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_ARR */ +/* TGSI_OPCODE_ARA */ +/* TGSI_OPCODE_BRA */ +/* TGSI_OPCODE_CAL */ +/* TGSI_OPCODE_RET */ +#define TGSI_OPCODE_PUSHA 80 +#define TGSI_OPCODE_POPA 81 +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXP */ +/* TGSI_OPCODE_TXB */ +/* TGSI_OPCODE_TXL */ + +/* + * GL_NV_gpu_program4 + */ +/* TGSI_OPCODE_ABS */ +#define TGSI_OPCODE_CEIL 82 +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +#define TGSI_OPCODE_I2F 83 +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_MOV */ +#define TGSI_OPCODE_NOT 84 +/* TGSI_OPCODE_NRM */ +/* TGSI_OPCODE_PK2H */ +/* TGSI_OPCODE_PK2US */ +/* TGSI_OPCODE_PK4B */ +/* TGSI_OPCODE_PK4UB */ +/* TGSI_OPCODE_ROUND */ +/* TGSI_OPCODE_SSG */ +#define TGSI_OPCODE_TRUNC 85 +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_RCC */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SCS */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_UP2H */ +/* TGSI_OPCODE_UP2US */ +/* TGSI_OPCODE_UP4B */ +/* TGSI_OPCODE_UP4UB */ +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_DIV */ +#define TGSI_OPCODE_SHL 86 +#define TGSI_OPCODE_SHR 87 +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_AND 88 +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_DPH */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MUL */ +#define TGSI_OPCODE_OR 89 +/* TGSI_OPCODE_RFL */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SFL */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_STR */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_XPD */ +/* TGSI_OPCODE_DP2 */ +#define TGSI_OPCODE_MOD 90 +#define TGSI_OPCODE_XOR 91 +/* TGSI_OPCODE_CMP */ +/* TGSI_OPCODE_DP2A */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_MAD */ +#define TGSI_OPCODE_SAD 92 +/* TGSI_OPCODE_X2D */ +/* TGSI_OPCODE_SWZ */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXB */ +#define TGSI_OPCODE_TXF 93 +/* TGSI_OPCODE_TXL */ +/* TGSI_OPCODE_TXP */ +#define TGSI_OPCODE_TXQ 94 +/* TGSI_OPCODE_TXD */ +/* TGSI_OPCODE_CAL */ +/* TGSI_OPCODE_RET */ +/* TGSI_OPCODE_BRK */ +#define TGSI_OPCODE_CONT 95 +/* TGSI_OPCODE_IF */ +/* TGSI_OPCODE_REP */ +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +/* TGSI_OPCODE_ENDREP */ + +/* + * GL_NV_vertex_program4 + */ +/* Same as GL_NV_gpu_program4 */ + +/* + * GL_NV_fragment_program4 + */ +/* Same as GL_NV_gpu_program4 */ +/* TGSI_OPCODE_KIL */ +/* TGSI_OPCODE_DDX */ +/* TGSI_OPCODE_DDY */ + +/* + * GL_NV_geometry_program4 + */ +/* Same as GL_NV_gpu_program4 */ +#define TGSI_OPCODE_EMIT 96 +#define TGSI_OPCODE_ENDPRIM 97 + +/* + * GLSL + */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_ADD */ +#define TGSI_OPCODE_BGNLOOP2 98 +#define TGSI_OPCODE_BGNSUB 99 +/* TGSI_OPCODE_BRA */ +/* TGSI_OPCODE_BRK */ +/* TGSI_OPCODE_CONT */ +/* TGSI_OPCODE_COS */ +/* TGSI_OPCODE_DDX */ +/* TGSI_OPCODE_DDY */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +#define TGSI_OPCODE_ENDLOOP2 100 +#define TGSI_OPCODE_ENDSUB 101 +/* TGSI_OPCODE_EX2 */ +/* TGSI_OPCODE_EXP */ +/* TGSI_OPCODE_FLR */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_IF */ +#define TGSI_OPCODE_INT TGSI_OPCODE_TRUNC +/* TGSI_OPCODE_KIL */ +/* TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LOG */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_MUL */ +#define TGSI_OPCODE_NOISE1 102 +#define TGSI_OPCODE_NOISE2 103 +#define TGSI_OPCODE_NOISE3 104 +#define TGSI_OPCODE_NOISE4 105 +#define TGSI_OPCODE_NOP 106 +/* TGSI_OPCODE_POW */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SIN */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SNE */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TXB */ +/* TGSI_OPCODE_TXD */ +/* TGSI_OPCODE_TXL */ +/* TGSI_OPCODE_TXP */ +/* TGSI_OPCODE_XPD */ + +/* + * ps_1_1 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_LRP */ +#define TGSI_OPCODE_TEXCOORD TGSI_OPCODE_NOP +#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL +/* TGSI_OPCODE_TEX */ +#define TGSI_OPCODE_TEXBEM 107 +#define TGSI_OPCODE_TEXBEML 108 +#define TGSI_OPCODE_TEXREG2AR 109 +#define TGSI_OPCODE_TEXM3X2PAD 110 +#define TGSI_OPCODE_TEXM3X2TEX 111 +#define TGSI_OPCODE_TEXM3X3PAD 112 +#define TGSI_OPCODE_TEXM3X3TEX 113 +#define TGSI_OPCODE_TEXM3X3SPEC 114 +#define TGSI_OPCODE_TEXM3X3VSPEC 115 +/* TGSI_OPCODE_CND */ + +/* + * ps_1_2 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_TEXCOORD */ +/* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TEXBEM */ +/* TGSI_OPCODE_TEXBEML */ +/* TGSI_OPCODE_TEXREG2AR */ +#define TGSI_OPCODE_TEXREG2GB 116 +/* TGSI_OPCODE_TEXM3X2PAD */ +/* TGSI_OPCODE_TEXM3X2TEX */ +/* TGSI_OPCODE_TEXM3X3PAD */ +/* TGSI_OPCODE_TEXM3X3TEX */ +/* TGSI_OPCODE_TEXM3X3SPEC */ +/* TGSI_OPCODE_TEXM3X3VSPEC */ +/* TGSI_OPCODE_CND */ +#define TGSI_OPCODE_TEXREG2RGB 117 +#define TGSI_OPCODE_TEXDP3TEX 118 +#define TGSI_OPCODE_TEXDP3 119 +#define TGSI_OPCODE_TEXM3X3 120 +/* CMP - use TGSI_OPCODE_CND0 */ + +/* + * ps_1_3 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_TEXCOORD */ +/* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_TEX */ +/* TGSI_OPCODE_TEXBEM */ +/* TGSI_OPCODE_TEXBEML */ +/* TGSI_OPCODE_TEXREG2AR */ +/* TGSI_OPCODE_TEXREG2GB */ +/* TGSI_OPCODE_TEXM3X2PAD */ +/* TGSI_OPCODE_TEXM3X2TEX */ +/* TGSI_OPCODE_TEXM3X3PAD */ +/* TGSI_OPCODE_TEXM3X3TEX */ +/* TGSI_OPCODE_TEXM3X3SPEC */ +/* TGSI_OPCODE_TEXM3X3VSPEC */ +/* TGSI_OPCODE_CND */ +/* TGSI_OPCODE_TEXREG2RGB */ +/* TGSI_OPCODE_TEXDP3TEX */ +#define TGSI_OPCODE_TEXM3X2DEPTH 121 +/* TGSI_OPCODE_TEXDP3 */ +/* TGSI_OPCODE_TEXM3X3 */ +/* CMP - use TGSI_OPCODE_CND0 */ + +/* + * ps_1_4 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_LRP */ +#define TGSI_OPCODE_TEXCRD TGSI_OPCODE_TEXCOORD +/* TGSI_OPCODE_TEXKILL */ +#define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX +/* TGSI_OPCODE_CND */ +#define TGSI_OPCODE_TEXDEPTH 122 +/* CMP - use TGSI_OPCODE_CND0 */ +#define TGSI_OPCODE_BEM 123 + +/* + * ps_2_0 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_FRC */ +#define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX +#define TGSI_OPCODE_M4X3 124 +#define TGSI_OPCODE_M3X4 125 +#define TGSI_OPCODE_M3X3 126 +#define TGSI_OPCODE_M3X2 127 +/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ +#define TGSI_OPCODE_CRS TGSI_OPCODE_XPD +/* TGSI_OPCODE_ABS */ +#define TGSI_OPCODE_NRM4 128 +#define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS +/* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_TEXLD */ +#define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB +#define TGSI_OPCODE_TEXLDP TGSI_OPCODE_TXP +/* CMP - use TGSI_OPCODE_CND0 */ +#define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A + +/* + * ps_2_x + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SNE */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_M4X4 */ +/* TGSI_OPCODE_M4X3 */ +/* TGSI_OPCODE_M3X4 */ +/* TGSI_OPCODE_M3X3 */ +/* TGSI_OPCODE_M3X2 */ +#define TGSI_OPCODE_CALL TGSI_OPCODE_CAL +#define TGSI_OPCODE_CALLNZ 129 +/* TGSI_OPCODE_RET */ +/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ +/* TGSI_OPCODE_CRS */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_NRM4 */ +/* TGSI_OPCODE_SINCOS */ +/* TGSI_OPCODE_REP */ +/* TGSI_OPCODE_ENDREP */ +/* TGSI_OPCODE_IF */ +#define TGSI_OPCODE_IFC 130 +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +#define TGSI_OPCODE_BREAK TGSI_OPCODE_BRK +#define TGSI_OPCODE_BREAKC 131 +/* TGSI_OPCODE_TEXKILL */ +/* TGSI_OPCODE_TEXLD */ +/* TGSI_OPCODE_TEXLDB */ +/* CMP - use TGSI_OPCODE_CND0 */ +/* TGSI_OPCODE_DP2ADD */ +#define TGSI_OPCODE_DSX TGSI_OPCODE_DDX +#define TGSI_OPCODE_DSY TGSI_OPCODE_DDY +#define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD +/* TGSI_OPCODE_TEXLDP */ + +/* + * vs_1_1 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_M4X4 */ +/* TGSI_OPCODE_M4X3 */ +/* TGSI_OPCODE_M3X4 */ +/* TGSI_OPCODE_M3X3 */ +/* TGSI_OPCODE_M3X2 */ +#define TGSI_OPCODE_EXPP TGSI_OPCODE_EXP +#define TGSI_OPCODE_LOGP TGSI_OPCODE_LG2 + +/* + * vs_2_0 + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_M4X4 */ +/* TGSI_OPCODE_M4X3 */ +/* TGSI_OPCODE_M3X4 */ +/* TGSI_OPCODE_M3X3 */ +/* TGSI_OPCODE_M3X2 */ +/* TGSI_OPCODE_CALL */ +/* TGSI_OPCODE_CALLNZ */ +/* TGSI_OPCODE_LOOP */ +/* TGSI_OPCODE_RET */ +/* TGSI_OPCODE_ENDLOOP */ +/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ +/* TGSI_OPCODE_CRS */ +#define TGSI_OPCODE_SGN TGSI_OPCODE_SSG +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_NRM4 */ +/* TGSI_OPCODE_SINCOS */ +/* TGSI_OPCODE_REP */ +/* TGSI_OPCODE_ENDREP */ +/* TGSI_OPCODE_IF */ +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +#define TGSI_OPCODE_MOVA TGSI_OPCODE_ARR +/* TGSI_OPCODE_LOGP */ + +/* + * vs_2_x + */ +/* TGSI_OPCODE_NOP */ +/* TGSI_OPCODE_MOV */ +/* TGSI_OPCODE_ADD */ +/* TGSI_OPCODE_SUB */ +/* TGSI_OPCODE_MAD */ +/* TGSI_OPCODE_MUL */ +/* TGSI_OPCODE_RCP */ +/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ +/* TGSI_OPCODE_DP3 */ +/* TGSI_OPCODE_DP4 */ +/* TGSI_OPCODE_MIN */ +/* TGSI_OPCODE_MAX */ +/* TGSI_OPCODE_SLT */ +/* TGSI_OPCODE_SGE */ +/* TGSI_OPCODE_SGT */ +/* TGSI_OPCODE_SLE */ +/* TGSI_OPCODE_SEQ */ +/* TGSI_OPCODE_SNE */ +/* EXP - use TGSI_OPCODE_EX2 */ +/* LOG - use TGSI_OPCODE_LG2 */ +/* TGSI_OPCODE_LIT */ +/* TGSI_OPCODE_DST */ +/* TGSI_OPCODE_LRP */ +/* TGSI_OPCODE_FRC */ +/* TGSI_OPCODE_M4X4 */ +/* TGSI_OPCODE_M4X3 */ +/* TGSI_OPCODE_M3X4 */ +/* TGSI_OPCODE_M3X3 */ +/* TGSI_OPCODE_M3X2 */ +/* TGSI_OPCODE_CALL */ +/* TGSI_OPCODE_CALLNZ */ +/* TGSI_OPCODE_LOOP */ +/* TGSI_OPCODE_RET */ +/* TGSI_OPCODE_ENDLOOP */ +/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ +/* TGSI_OPCODE_CRS */ +/* TGSI_OPCODE_SGN */ +/* TGSI_OPCODE_ABS */ +/* TGSI_OPCODE_NRM4 */ +/* TGSI_OPCODE_SINCOS */ +/* TGSI_OPCODE_REP */ +/* TGSI_OPCODE_ENDREP */ +/* TGSI_OPCODE_IF */ +/* TGSI_OPCODE_IFC */ +/* TGSI_OPCODE_ELSE */ +/* TGSI_OPCODE_ENDIF */ +/* TGSI_OPCODE_BREAK */ +/* TGSI_OPCODE_BREAKC */ +/* TGSI_OPCODE_MOVA */ +/* TGSI_OPCODE_LOGP */ + +#define TGSI_OPCODE_LAST 133 + +#define TGSI_SAT_NONE 0 /* do not saturate */ +#define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ +#define TGSI_SAT_MINUS_PLUS_ONE 2 /* clamp to [-1,1] */ + +/* + * Opcode is the operation code to execute. A given operation defines the + * semantics how the source registers (if any) are interpreted and what is + * written to the destination registers (if any) as a result of execution. + * + * NumDstRegs and NumSrcRegs is the number of destination and source registers, + * respectively. For a given operation code, those numbers are fixed and are + * present here only for convenience. + * + * If Extended is TRUE, it is now executed. + * + * Saturate controls how are final results in destination registers modified. + */ + +struct tgsi_instruction +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_INSTRUCTION */ + unsigned Size : 8; /* UINT */ + unsigned Opcode : 8; /* TGSI_OPCODE_ */ + unsigned Saturate : 2; /* TGSI_SAT_ */ + unsigned NumDstRegs : 2; /* UINT */ + unsigned NumSrcRegs : 4; /* UINT */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows. + * + * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow. + * + * Then, tgsi_instruction::NumSrcRegs of tgsi_src_register follow. + * + * tgsi_instruction::Size contains the total number of words that make the + * instruction, including the instruction word. + */ + +#define TGSI_INSTRUCTION_EXT_TYPE_NV 0 +#define TGSI_INSTRUCTION_EXT_TYPE_LABEL 1 +#define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE 2 +#define TGSI_INSTRUCTION_EXT_TYPE_PREDICATE 3 + +struct tgsi_instruction_ext +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_NV, it should + * be cast to tgsi_instruction_ext_nv. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_LABEL, it + * should be cast to tgsi_instruction_ext_label. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it + * should be cast to tgsi_instruction_ext_texture. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_PREDICATE, it + * should be cast to tgsi_instruction_ext_predicate. + * + * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext + * follows. + */ + +#define TGSI_PRECISION_DEFAULT 0 +#define TGSI_PRECISION_FLOAT32 1 +#define TGSI_PRECISION_FLOAT16 2 +#define TGSI_PRECISION_FIXED12 3 + +#define TGSI_CC_GT 0 +#define TGSI_CC_EQ 1 +#define TGSI_CC_LT 2 +#define TGSI_CC_UN 3 +#define TGSI_CC_GE 4 +#define TGSI_CC_LE 5 +#define TGSI_CC_NE 6 +#define TGSI_CC_TR 7 +#define TGSI_CC_FL 8 + +#define TGSI_SWIZZLE_X 0 +#define TGSI_SWIZZLE_Y 1 +#define TGSI_SWIZZLE_Z 2 +#define TGSI_SWIZZLE_W 3 + +/* + * Precision controls the precision at which the operation should be executed. + * + * CondDstUpdate enables condition code register writes. When this field is + * TRUE, CondDstIndex specifies the index of the condition code register to + * update. + * + * CondFlowEnable enables conditional execution of the operation. When this + * field is TRUE, CondFlowIndex specifies the index of the condition code + * register to test against CondMask with component swizzle controled by + * CondSwizzleX, CondSwizzleY, CondSwizzleZ and CondSwizzleW. If the test fails, + * the operation is not executed. + */ + +struct tgsi_instruction_ext_nv +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_NV */ + unsigned Precision : 4; /* TGSI_PRECISION_ */ + unsigned CondDstIndex : 4; /* UINT */ + unsigned CondFlowIndex : 4; /* UINT */ + unsigned CondMask : 4; /* TGSI_CC_ */ + unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned CondDstUpdate : 1; /* BOOL */ + unsigned CondFlowEnable : 1; /* BOOL */ + unsigned Padding : 1; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_instruction_ext_label +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ + unsigned Label : 24; /* UINT */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ +}; + +#define TGSI_TEXTURE_UNKNOWN 0 +#define TGSI_TEXTURE_1D 1 +#define TGSI_TEXTURE_2D 2 +#define TGSI_TEXTURE_3D 3 +#define TGSI_TEXTURE_CUBE 4 +#define TGSI_TEXTURE_RECT 5 +#define TGSI_TEXTURE_SHADOW1D 6 +#define TGSI_TEXTURE_SHADOW2D 7 +#define TGSI_TEXTURE_SHADOWRECT 8 + +struct tgsi_instruction_ext_texture +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */ + unsigned Texture : 8; /* TGSI_TEXTURE_ */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_instruction_ext_predicate +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ + unsigned PredDstIndex : 4; /* UINT */ + unsigned PredWriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * File specifies the register array to access. + * + * Index specifies the element number of a register in the register file. + * + * If Indirect is TRUE, Index should be offset by the X component of a source + * register that follows. The register can be now fetched into local storage + * for further processing. + * + * If Negate is TRUE, all components of the fetched register are negated. + * + * The fetched register components are swizzled according to SwizzleX, SwizzleY, + * SwizzleZ and SwizzleW. + * + * If Extended is TRUE, any further modifications to the source register are + * made to this temporary storage. + */ + +struct tgsi_src_register +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned Negate : 1; /* BOOL */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_src_register::Extended is TRUE, tgsi_src_register_ext follows. + * + * Then, if tgsi_src_register::Indirect is TRUE, another tgsi_src_register + * follows. + * + * Then, if tgsi_src_register::Dimension is TRUE, tgsi_dimension follows. + */ + +#define TGSI_SRC_REGISTER_EXT_TYPE_SWZ 0 +#define TGSI_SRC_REGISTER_EXT_TYPE_MOD 1 + +struct tgsi_src_register_ext +{ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_SWZ, + * it should be cast to tgsi_src_register_ext_extswz. + * + * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_MOD, + * it should be cast to tgsi_src_register_ext_mod. + * + * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext + * follows. + */ + +#define TGSI_EXTSWIZZLE_X TGSI_SWIZZLE_X +#define TGSI_EXTSWIZZLE_Y TGSI_SWIZZLE_Y +#define TGSI_EXTSWIZZLE_Z TGSI_SWIZZLE_Z +#define TGSI_EXTSWIZZLE_W TGSI_SWIZZLE_W +#define TGSI_EXTSWIZZLE_ZERO 4 +#define TGSI_EXTSWIZZLE_ONE 5 + +/* + * ExtSwizzleX, ExtSwizzleY, ExtSwizzleZ and ExtSwizzleW swizzle the source + * register in an extended manner. + * + * NegateX, NegateY, NegateZ and NegateW negate individual components of the + * source register. + * + * ExtDivide specifies which component is used to divide all components of the + * source register. + */ + +struct tgsi_src_register_ext_swz +{ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */ + unsigned ExtSwizzleX : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleY : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleZ : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleW : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned NegateX : 1; /* BOOL */ + unsigned NegateY : 1; /* BOOL */ + unsigned NegateZ : 1; /* BOOL */ + unsigned NegateW : 1; /* BOOL */ + unsigned ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ +}; + +/** + * Extra src register modifiers + * + * If Complement is TRUE, the source register is modified by subtracting it + * from 1.0. + * + * If Bias is TRUE, the source register is modified by subtracting 0.5 from it. + * + * If Scale2X is TRUE, the source register is modified by multiplying it by 2.0. + * + * If Absolute is TRUE, the source register is modified by removing the sign. + * + * If Negate is TRUE, the source register is modified by negating it. + */ + +struct tgsi_src_register_ext_mod +{ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */ + unsigned Complement : 1; /* BOOL */ + unsigned Bias : 1; /* BOOL */ + unsigned Scale2X : 1; /* BOOL */ + unsigned Absolute : 1; /* BOOL */ + unsigned Negate : 1; /* BOOL */ + unsigned Padding : 22; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_dimension +{ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + unsigned Padding : 13; + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_dst_register +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Padding : 5; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_dst_register::Extended is TRUE, tgsi_dst_register_ext follows. + * + * Then, if tgsi_dst_register::Indirect is TRUE, tgsi_src_register follows. + */ + +#define TGSI_DST_REGISTER_EXT_TYPE_CONDCODE 0 +#define TGSI_DST_REGISTER_EXT_TYPE_MODULATE 1 +#define TGSI_DST_REGISTER_EXT_TYPE_PREDICATE 2 + +struct tgsi_dst_register_ext +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ +}; + +/** + * Extra destination register modifiers + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_CONDCODE, + * it should be cast to tgsi_dst_register_ext_condcode. + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_MODULATE, + * it should be cast to tgsi_dst_register_ext_modulate. + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_PREDICATE, + * it should be cast to tgsi_dst_register_ext_predicate. + * + * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext + * follows. + */ +struct tgsi_dst_register_ext_concode +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ + unsigned CondMask : 4; /* TGSI_CC_ */ + unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSrcIndex : 4; /* UINT */ + unsigned Padding : 11; + unsigned Extended : 1; /* BOOL */ +}; + +#define TGSI_MODULATE_1X 0 +#define TGSI_MODULATE_2X 1 +#define TGSI_MODULATE_4X 2 +#define TGSI_MODULATE_8X 3 +#define TGSI_MODULATE_HALF 4 +#define TGSI_MODULATE_QUARTER 5 +#define TGSI_MODULATE_EIGHTH 6 + +struct tgsi_dst_register_ext_modulate +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */ + unsigned Modulate : 4; /* TGSI_MODULATE_ */ + unsigned Padding : 23; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * Currently, the following constraints apply. + * + * - PredSwizzleXYZW is either set to identity or replicate. + * - PredSrcIndex is 0. + */ + +struct tgsi_dst_register_ext_predicate +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */ + unsigned PredSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSrcIndex : 4; /* UINT */ + unsigned Negate : 1; /* BOOL */ + unsigned Padding : 14; + unsigned Extended : 1; /* BOOL */ +}; + + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_TOKEN_H + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_util.c b/src/mesa/pipe/tgsi/exec/tgsi_util.c new file mode 100644 index 0000000000..38d6d6e6bc --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_util.c @@ -0,0 +1,270 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +union pointer_hack +{ + void *pointer; + unsigned long long uint64; +}; + +void * +tgsi_align_128bit( + void *unaligned ) +{ + union pointer_hack ph; + + ph.uint64 = 0; + ph.pointer = unaligned; + ph.uint64 = (ph.uint64 + 15) & ~15; + return ph.pointer; +} + +GLuint +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + GLuint component ) +{ + switch( component ) { + case 0: + return reg->SwizzleX; + case 1: + return reg->SwizzleY; + case 2: + return reg->SwizzleZ; + case 3: + return reg->SwizzleW; + default: + assert( 0 ); + } + return 0; +} + +GLuint +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + GLuint component ) +{ + switch( component ) { + case 0: + return reg->ExtSwizzleX; + case 1: + return reg->ExtSwizzleY; + case 2: + return reg->ExtSwizzleZ; + case 3: + return reg->ExtSwizzleW; + default: + assert( 0 ); + } + return 0; +} + +GLuint +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + GLuint component ) +{ + GLuint swizzle; + + /* + * First, calculate the extended swizzle for a given channel. This will give + * us either a channel index into the simple swizzle or a constant 1 or 0. + */ + swizzle = tgsi_util_get_src_register_extswizzle( + ®->SrcRegisterExtSwz, + component ); + + assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); + assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); + assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); + assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); + + /* + * Second, calculate the simple swizzle for the unswizzled channel index. + * Leave the constants intact, they are not affected by the simple swizzle. + */ + if( swizzle <= TGSI_SWIZZLE_W ) { + swizzle = tgsi_util_get_src_register_swizzle( + ®->SrcRegister, + swizzle ); + } + + return swizzle; +} + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + GLuint swizzle, + GLuint component ) +{ + switch( component ) { + case 0: + reg->SwizzleX = swizzle; + break; + case 1: + reg->SwizzleY = swizzle; + break; + case 2: + reg->SwizzleZ = swizzle; + break; + case 3: + reg->SwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + GLuint swizzle, + GLuint component ) +{ + switch( component ) { + case 0: + reg->ExtSwizzleX = swizzle; + break; + case 1: + reg->ExtSwizzleY = swizzle; + break; + case 2: + reg->ExtSwizzleZ = swizzle; + break; + case 3: + reg->ExtSwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +GLuint +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + GLuint component ) +{ + switch( component ) { + case 0: + return reg->NegateX; + case 1: + return reg->NegateY; + case 2: + return reg->NegateZ; + case 3: + return reg->NegateW; + default: + assert( 0 ); + } + return 0; +} + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + GLuint negate, + GLuint component ) +{ + switch( component ) { + case 0: + reg->NegateX = negate; + break; + case 1: + reg->NegateY = negate; + break; + case 2: + reg->NegateZ = negate; + break; + case 3: + reg->NegateW = negate; + break; + default: + assert( 0 ); + } +} + +GLuint +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + GLuint component ) +{ + GLuint sign_mode; + + if( reg->SrcRegisterExtMod.Absolute ) { + /* Consider only the post-abs negation. */ + + if( reg->SrcRegisterExtMod.Negate ) { + sign_mode = TGSI_UTIL_SIGN_SET; + } + else { + sign_mode = TGSI_UTIL_SIGN_CLEAR; + } + } + else { + /* Accumulate the three negations. */ + + GLuint negate; + + negate = reg->SrcRegister.Negate; + if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { + negate = !negate; + } + if( reg->SrcRegisterExtMod.Negate ) { + negate = !negate; + } + + if( negate ) { + sign_mode = TGSI_UTIL_SIGN_TOGGLE; + } + else { + sign_mode = TGSI_UTIL_SIGN_KEEP; + } + } + + return sign_mode; +} + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + GLuint sign_mode ) +{ + reg->SrcRegisterExtSwz.NegateX = 0; + reg->SrcRegisterExtSwz.NegateY = 0; + reg->SrcRegisterExtSwz.NegateZ = 0; + reg->SrcRegisterExtSwz.NegateW = 0; + + switch (sign_mode) + { + case TGSI_UTIL_SIGN_CLEAR: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_SET: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 1; + break; + + case TGSI_UTIL_SIGN_TOGGLE: + reg->SrcRegister.Negate = 1; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_KEEP: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + default: + assert( 0 ); + } +} + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_util.h b/src/mesa/pipe/tgsi/exec/tgsi_util.h new file mode 100644 index 0000000000..ef14446f0e --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_util.h @@ -0,0 +1,70 @@ +#if !defined TGSI_UTIL_H +#define TGSI_UTIL_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +void * +tgsi_align_128bit( + void *unaligned ); + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component); + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + unsigned swizzle, + unsigned component ); + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + unsigned swizzle, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ); + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ); + +#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ +#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ +#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ +#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + unsigned sign_mode ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_UTIL_H + -- cgit v1.2.3 From c89502bb4e6d80182e4977ab084639850e1eba68 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 24 Aug 2007 10:01:20 +0100 Subject: Fixup include paths after rename --- src/mesa/pipe/draw/draw_prim.c | 2 +- src/mesa/pipe/draw/draw_vertex.h | 2 +- src/mesa/pipe/i915simple/i915_fpc_translate.c | 4 ++-- src/mesa/pipe/softpipe/sp_headers.h | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- src/mesa/pipe/softpipe/sp_tex_sample.c | 2 +- src/mesa/pipe/tgsi/deco/tgsi_deco.h | 2 +- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 +- src/mesa/pipe/tgsi/mesa/tgsi_mesa.h | 2 +- src/mesa/sources | 10 +++++----- src/mesa/state_tracker/st_atom_fs.c | 2 +- src/mesa/state_tracker/st_atom_vs.c | 2 +- src/mesa/state_tracker/st_draw.c | 2 +- src/mesa/state_tracker/st_program.h | 2 +- 15 files changed, 20 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 18a5d39077..c98d6ff298 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -35,7 +35,7 @@ #include "draw_context.h" #include "draw_prim.h" -#include "pipe/tgsi/core/tgsi_core.h" +#include "pipe/tgsi/exec/tgsi_core.h" #define RP_NONE 0 diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 0e15ea89a2..c104aa1143 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -33,7 +33,7 @@ #ifndef DRAW_VERTEX_H #define DRAW_VERTEX_H -#include "pipe/tgsi/core/tgsi_attribs.h" +#include "pipe/tgsi/exec/tgsi_attribs.h" #define MAX_VERT_ATTRIBS 12 /* OK? */ diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index e7315d2263..6fcba719ae 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -29,8 +29,8 @@ #include "i915_context.h" #include "i915_fpc.h" -#include "pipe/tgsi/core/tgsi_token.h" -#include "pipe/tgsi/core/tgsi_parse.h" +#include "pipe/tgsi/exec/tgsi_token.h" +#include "pipe/tgsi/exec/tgsi_parse.h" #include "pipe/draw/draw_vertex.h" diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index cc8294b18e..b377c34712 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -31,7 +31,7 @@ #ifndef SP_HEADERS_H #define SP_HEADERS_H -#include "../tgsi/core/tgsi_core.h" +#include "../tgsi/exec/tgsi_core.h" #define PRIM_POINT 1 #define PRIM_LINE 2 diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 8d41e09465..bfc31f94a6 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -34,7 +34,7 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" -#include "pipe/tgsi/core/tgsi_attribs.h" +#include "pipe/tgsi/exec/tgsi_attribs.h" #include "sp_context.h" #include "sp_headers.h" diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 76200857f0..b0d79eedda 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -33,7 +33,7 @@ #include "sp_context.h" #include "sp_state.h" -#include "pipe/tgsi/core/tgsi_attribs.h" +#include "pipe/tgsi/exec/tgsi_attribs.h" diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 23a0306bba..40d0cf40ae 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -39,7 +39,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" -#include "pipe/tgsi/core/tgsi_exec.h" +#include "pipe/tgsi/exec/tgsi_exec.h" /* diff --git a/src/mesa/pipe/tgsi/deco/tgsi_deco.h b/src/mesa/pipe/tgsi/deco/tgsi_deco.h index 06a2654fb6..3560e455f3 100644 --- a/src/mesa/pipe/tgsi/deco/tgsi_deco.h +++ b/src/mesa/pipe/tgsi/deco/tgsi_deco.h @@ -1,7 +1,7 @@ #if !defined TGSI_DECO_H #define TGSI_DECO_H -#include "../core/tgsi_core.h" +#include "../exec/tgsi_core.h" #include "deco_caps.h" #endif // !defined TGSI_DECO_H diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index f33a702958..78390767df 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -1,6 +1,6 @@ #include "tgsi_platform.h" #include "tgsi_mesa.h" -#include "pipe/tgsi/core/tgsi_attribs.h" +#include "pipe/tgsi/exec/tgsi_attribs.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #define TGSI_DEBUG 1 diff --git a/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h b/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h index 0053748fde..a4e16e708e 100644 --- a/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h +++ b/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h @@ -1,7 +1,7 @@ #if !defined TGSI_MESA_H #define TGSI_MESA_H -#include "../core/tgsi_core.h" +#include "../exec/tgsi_core.h" #include "mesa_to_tgsi.h" #endif // !defined TGSI_MESA_H diff --git a/src/mesa/sources b/src/mesa/sources index e78db591ab..ec30a889f6 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -167,11 +167,11 @@ DRAW_SOURCES = \ pipe/draw/draw_unfilled.c TGSICORE_SOURCES = \ - pipe/tgsi/core/tgsi_build.c \ - pipe/tgsi/core/tgsi_dump.c \ - pipe/tgsi/core/tgsi_exec.c \ - pipe/tgsi/core/tgsi_parse.c \ - pipe/tgsi/core/tgsi_util.c + pipe/tgsi/exec/tgsi_build.c \ + pipe/tgsi/exec/tgsi_dump.c \ + pipe/tgsi/exec/tgsi_exec.c \ + pipe/tgsi/exec/tgsi_parse.c \ + pipe/tgsi/exec/tgsi_util.c TGSIDECO_SOURCES = \ pipe/tgsi/deco/deco_caps.c diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index a34a227ac0..a164ded184 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -35,7 +35,7 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" -#include "pipe/tgsi/core/tgsi_dump.h" +#include "pipe/tgsi/exec/tgsi_dump.h" #include "st_context.h" #include "st_atom.h" diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 8c9365d9d8..4c83a8c5c1 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -37,7 +37,7 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" -#include "pipe/tgsi/core/tgsi_dump.h" +#include "pipe/tgsi/exec/tgsi_dump.h" #include "st_context.h" #include "st_atom.h" diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index a88e2106d7..afb6f96a26 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -44,7 +44,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/tgsi/core/tgsi_attribs.h" +#include "pipe/tgsi/exec/tgsi_attribs.h" diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 3ff4f4e9c7..e7eb7a8e3f 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -35,7 +35,7 @@ #define ST_PROGRAM_H #include "mtypes.h" -#include "pipe/tgsi/core/tgsi_token.h" +#include "pipe/tgsi/exec/tgsi_token.h" #define ST_FP_MAX_TOKENS 1024 -- cgit v1.2.3 From 6998402016e146d6afe06549ce2fb5941d6b03c7 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 09:52:25 -0600 Subject: emit S4_VFMT_XYZ vs. S4_VFMT_XYZW as needed --- src/mesa/pipe/i915simple/i915_state_derived.c | 19 +++++++++++++------ 1 file changed, 13 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 426d43f288..c24f54580e 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -41,8 +41,6 @@ emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) const uint n = vinfo->num_attribs; vinfo->attr_mask |= (1 << vfAttr); vinfo->slot_to_attrib[n] = vfAttr; - printf("Vertex slot %d = vfattrib %d\n", n, vfAttr); - /*vinfo->interp_mode[n] = interpMode;*/ vinfo->format[n] = format; vinfo->num_attribs++; } @@ -92,15 +90,13 @@ static void calculate_vertex_layout( struct i915_context *i915 ) { const unsigned inputsRead = i915->fs.inputs_read; struct vertex_info *vinfo = &i915->current.vertex_info; + boolean needW = 0; memset(vinfo, 0, sizeof(*vinfo)); - /* TODO - Figure out if we need to do perspective divide, etc. - */ - /* pos */ emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_3F); - vinfo->hwfmt[0] |= S4_VFMT_XYZ; + /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ /* color0 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { @@ -125,6 +121,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) if (inputsRead & (1 << i)) { emit_vertex_attr(vinfo, i, FORMAT_4F); hwtc = TEXCOORDFMT_4D; + needW = TRUE; } else { hwtc = TEXCOORDFMT_NOT_PRESENT; @@ -133,6 +130,16 @@ static void calculate_vertex_layout( struct i915_context *i915 ) } } + /* go back and fill in the vertex position info now that we have needW */ + if (needW) { + vinfo->hwfmt[0] |= S4_VFMT_XYZW; + vinfo->format[0] = FORMAT_4F; + } + else { + vinfo->hwfmt[0] |= S4_VFMT_XYZ; + vinfo->format[0] == FORMAT_3F; + } + /* Additional attributes required for setup: Just twosided * lighting. Edgeflag is dealt with specially by setting bits in * the vertex header. -- cgit v1.2.3 From 6a13caec093dc86fbce80c9f78426d39e9b4276b Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 09:56:47 -0600 Subject: re-enable compute_clipmask() call --- src/mesa/pipe/draw/draw_prim.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index c98d6ff298..309017cc60 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -248,7 +248,7 @@ run_vertex_program(struct draw_context *draw, z = vOut[j]->clip[2] = machine.Outputs[0].xyzw[2].f[j]; w = vOut[j]->clip[3] = machine.Outputs[0].xyzw[3].f[j]; - vOut[j]->clipmask = 0;/*compute_clipmask(x, y, z, w);*/ + vOut[j]->clipmask = compute_clipmask(x, y, z, w); vOut[j]->edgeflag = 1; /* divide by w */ -- cgit v1.2.3 From f9a77a3080598d03c484fa5d04c213b8a06d43d3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 11:05:31 -0600 Subject: Fix user-defined clip planes. They seem to work properly now. The bits for the N user-defined planes are now set in the vertex_header->clipmask. See some detailed comments about this in the clip_line() function. CLIP_USER_BIT no longer exists. --- src/mesa/pipe/draw/draw_clip.c | 94 ++++++++++++++++++++++++--------------- src/mesa/pipe/draw/draw_context.c | 3 ++ src/mesa/pipe/draw/draw_context.h | 3 +- src/mesa/pipe/draw/draw_prim.c | 2 +- src/mesa/pipe/draw/draw_private.h | 1 + 5 files changed, 63 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index f46f4e3de6..a23b7a2810 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -50,10 +50,17 @@ #endif +/** bitmask for the first view-frustum clip planes */ +#define VIEWPLANE_MASK ((1 << 6) - 1) + +/** bitmask for the user-defined clip planes */ +#define USERPLANE_MASK (((1 << PIPE_MAX_CLIP_PLANES) - 1) << 6) + + + struct clipper { struct draw_stage stage; /**< base class */ - unsigned active_user_planes; float (*plane)[4]; }; @@ -232,17 +239,10 @@ do_clip_tri( struct draw_stage *stage, inlist[1] = header->v[1]; inlist[2] = header->v[2]; - /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove - * this once we correctly use 16bit masks for userclip planes. - */ clipmask &= ~CLIP_CULL_BIT; - if (clipmask & CLIP_USER_BIT) { - clipmask &= ~CLIP_USER_BIT; - clipmask |= clipper->active_user_planes; - } while (clipmask && n >= 3) { - unsigned plane_idx = ffs(clipmask)-1; + const unsigned plane_idx = ffs(clipmask)-1; const float *plane = clipper->plane[plane_idx]; struct vertex_header *vert_prev = inlist[0]; float dp_prev = dot4( vert_prev->clip, plane ); @@ -318,18 +318,11 @@ do_clip_line( struct draw_stage *stage, struct vertex_header *v1 = header->v[1]; const float *pos0 = v0->clip; const float *pos1 = v1->clip; - float t0 = 0; - float t1 = 0; + float t0 = 0.0F; + float t1 = 0.0F; struct prim_header newprim; - /* XXX: Note stupid hack to deal with tnl's 8-bit clipmask. Remove - * this once we correctly use 16bit masks for userclip planes. - */ clipmask &= ~CLIP_CULL_BIT; - if (clipmask & CLIP_USER_BIT) { - clipmask &= ~CLIP_USER_BIT; - clipmask |= clipper->active_user_planes; - } while (clipmask) { const unsigned plane_idx = ffs(clipmask)-1; @@ -337,17 +330,21 @@ do_clip_line( struct draw_stage *stage, const float dp0 = dot4( pos0, plane ); const float dp1 = dot4( pos1, plane ); - if (dp1 < 0) { + /* need this to handle user-clip planes properly */ + if (dp0 < 0.0F && dp1 < 0.0F) + return; + + if (dp1 < 0.0F) { float t = dp1 / (dp1 - dp0); t1 = MAX2(t1, t); } - if (dp0 < 0) { + if (dp0 < 0.0F) { float t = dp0 / (dp0 - dp1); t0 = MAX2(t0, t); } - if (t0 + t1 >= 1.0) + if (t0 + t1 >= 1.0F) return; /* discard */ clipmask &= ~(1 << plane_idx); /* turn off this plane's bit */ @@ -375,18 +372,11 @@ do_clip_line( struct draw_stage *stage, static void clip_begin( struct draw_stage *stage ) { - struct clipper *clipper = clipper_stage(stage); - unsigned nr = stage->draw->nr_planes; - /* sanity checks. If these fail, review the clip/interp code! */ assert(stage->draw->vertex_info.num_attribs >= 3); assert(stage->draw->vertex_info.slot_to_attrib[0] == TGSI_ATTRIB_VERTEX_HEADER); assert(stage->draw->vertex_info.slot_to_attrib[1] == TGSI_ATTRIB_CLIP_POS); - /* Hacky bitmask to use when we hit CLIP_USER_BIT: - */ - clipper->active_user_planes = ((1<next->begin( stage->next ); } @@ -395,8 +385,21 @@ static void clip_point( struct draw_stage *stage, struct prim_header *header ) { - if (header->v[0]->clipmask == 0) + if (header->v[0]->clipmask == 0) { stage->next->point( stage->next, header ); + } + else if (header->v[0]->clipmask & USERPLANE_MASK) { + /* test against user clip planes now */ + const struct clipper *clipper = clipper_stage( stage ); + uint i; + for (i = 6; i < stage->draw->nr_planes; i++) { + float dot = dot4(clipper->plane[i], header->v[0]->clip); + if (dot < 0.0F) + return; /* clipped! */ + } + /* not clipped */ + stage->next->point( stage->next, header ); + } } @@ -405,16 +408,31 @@ clip_line( struct draw_stage *stage, struct prim_header *header ) { unsigned clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask); + header->v[1]->clipmask); if (clipmask == 0) { /* no clipping needed */ stage->next->line( stage->next, header ); } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask) == 0) { + else if (((header->v[0]->clipmask & + header->v[1]->clipmask & + VIEWPLANE_MASK) == 0) || + (clipmask & USERPLANE_MASK)) { + /* About the above predicate: the clipmask bits for the view volume + * exactly indicate whether the coordinate is inside or outside each + * frustum plane. However, the bits for user-defined planes are set + * if the plane is enabled, and does not really indicate if the + * coordinate is inside or outside the user-defined plane. + * + * To change this (so that the user-plane bits really indicate + * inside/outside) we'd have to compute the dot products earlier + * in the draw_prim.c code (see compute_clipmask()). + * We will probably do that when we have support for user clip coord + * in vertex shaders... + */ do_clip_line(stage, header, clipmask); } + /* else, totally clipped */ } @@ -423,16 +441,18 @@ clip_tri( struct draw_stage *stage, struct prim_header *header ) { unsigned clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask | - header->v[2]->clipmask); + header->v[1]->clipmask | + header->v[2]->clipmask); if (clipmask == 0) { /* no clipping needed */ stage->next->tri( stage->next, header ); } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask & - header->v[2]->clipmask) == 0) { + else if (((header->v[0]->clipmask & + header->v[1]->clipmask & + header->v[2]->clipmask & + VIEWPLANE_MASK) == 0) || + (clipmask & USERPLANE_MASK)) { do_clip_tri(stage, header, clipmask); } } diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 760756ce89..b14de34141 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -166,8 +166,11 @@ void draw_set_setup_stage( struct draw_context *draw, void draw_set_clip_state( struct draw_context *draw, const struct pipe_clip_state *clip ) { + assert(clip->nr <= PIPE_MAX_CLIP_PLANES); memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); draw->nr_planes = 6 + clip->nr; + /* bitmask of the enabled user-defined clip planes */ + draw->user_clipmask = ((1 << clip->nr) - 1) << 6; } diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 05a3afd5c6..1bd0e624fe 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -56,8 +56,7 @@ struct draw_stage; #define CLIP_BOTTOM_BIT 0x08 #define CLIP_NEAR_BIT 0x10 #define CLIP_FAR_BIT 0x20 -#define CLIP_USER_BIT 0x40 -#define CLIP_CULL_BIT 0x80 +#define CLIP_CULL_BIT (1 << (6 + PIPE_MAX_CLIP_PLANES)) /*unused? */ /*@}*/ /** diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 309017cc60..2d613754e8 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -248,7 +248,7 @@ run_vertex_program(struct draw_context *draw, z = vOut[j]->clip[2] = machine.Outputs[0].xyzw[2].f[j]; w = vOut[j]->clip[3] = machine.Outputs[0].xyzw[3].f[j]; - vOut[j]->clipmask = compute_clipmask(x, y, z, w); + vOut[j]->clipmask = compute_clipmask(x, y, z, w) | draw->user_clipmask; vOut[j]->edgeflag = 1; /* divide by w */ diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index c5d4c62f71..e61f228d12 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -154,6 +154,7 @@ struct draw_context */ float plane[12][4]; unsigned nr_planes; + unsigned user_clipmask; /** Describes the layout of post-transformation vertices */ struct vertex_info vertex_info; -- cgit v1.2.3 From b4d8be235b018594a79258d36e70880ab9019b72 Mon Sep 17 00:00:00 2001 From: keithw Date: Fri, 24 Aug 2007 10:09:34 +0100 Subject: sampler state tweaks --- src/mesa/pipe/i915simple/i915_state_derived.c | 2 +- src/mesa/pipe/i915simple/i915_state_sampler.c | 7 ++----- 2 files changed, 3 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index c24f54580e..bab3015fde 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -174,7 +174,7 @@ void i915_update_derived( struct i915_context *i915 ) if (i915->dirty & (I915_NEW_SETUP | I915_NEW_FS)) calculate_vertex_layout( i915 ); - if (i915->dirty & I915_NEW_SAMPLER) + if (i915->dirty & (I915_NEW_SAMPLER | I915_NEW_TEXTURE)) i915_update_samplers(i915); if (i915->dirty & I915_NEW_TEXTURE) diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index e23e4bdcf7..e4d7fec27b 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -155,7 +155,6 @@ static void update_sampler(struct i915_context *i915, uint unit, const struct pipe_sampler_state *sampler, const struct pipe_mipmap_tree *mt, - const struct pipe_surface *surface, unsigned state[3] ) { const unsigned ws = sampler->wrap_s; @@ -187,8 +186,8 @@ static void update_sampler(struct i915_context *i915, state[0] |= ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); } - if (surface->format == PIPE_FORMAT_YCBCR || - surface->format == PIPE_FORMAT_YCBCR_REV) + if (mt->format == PIPE_FORMAT_YCBCR || + mt->format == PIPE_FORMAT_YCBCR_REV) state[0] |= SS2_COLORSPACE_CONVERSION; @@ -259,7 +258,6 @@ static void update_sampler(struct i915_context *i915, void i915_update_samplers( struct i915_context *i915 ) { - const struct pipe_surface *surface = i915->framebuffer.cbufs[0]; uint unit; i915->current.sampler_enable_nr = 0; @@ -273,7 +271,6 @@ void i915_update_samplers( struct i915_context *i915 ) unit, i915->sampler + unit, /* sampler state */ i915->texture[unit], /* mipmap tree */ - surface, /* cbuffer info */ i915->current.sampler[unit] /* the result */ ); -- cgit v1.2.3 From 7a1027cd4098c25dc59be889f4d00cdb47550b44 Mon Sep 17 00:00:00 2001 From: keithw Date: Fri, 24 Aug 2007 10:15:06 +0100 Subject: have to check I915_HW_SAMPLER when emitting MAP state, because of sampler_enable_flags. --- src/mesa/pipe/i915simple/i915_state_emit.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 48eb86091d..4804b0caee 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -192,8 +192,14 @@ i915_emit_hardware_state(struct i915_context *i915 ) #if 01 /* texture images */ - if (i915->hardware_dirty & I915_HW_MAP) + if (i915->hardware_dirty & (I915_HW_MAP | I915_HW_SAMPLER)) { + /* XXX: we were refering to sampler state + * (current.sampler_enable_nr) below, but only checking + * I915_HW_MAP above. Should probably calculate the enabled + * flags separately - but there will be further rework of + * state so perhaps not necessary yet. + */ const uint nr = i915->current.sampler_enable_nr; if (nr) { const uint enabled = i915->current.sampler_enable_flags; -- cgit v1.2.3 From d36e01aa940cfd48fcdbc7d57d356c4d4c7cfc9c Mon Sep 17 00:00:00 2001 From: keithw Date: Fri, 24 Aug 2007 10:22:37 +0100 Subject: remove dead file --- src/mesa/pipe/i915simple/i915_context.h | 1 - src/mesa/pipe/i915simple/i915_state_fragprog.c | 65 -------------------------- 2 files changed, 66 deletions(-) delete mode 100644 src/mesa/pipe/i915simple/i915_state_fragprog.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 5de6afa9e9..c30c79d83c 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -201,7 +201,6 @@ struct draw_stage *i915_draw_render_stage( struct i915_context *i915 ); * i915_state_emit.c: */ void i915_emit_hardware_state(struct i915_context *i915 ); -unsigned *i915_passthrough_program( unsigned *dwords ); diff --git a/src/mesa/pipe/i915simple/i915_state_fragprog.c b/src/mesa/pipe/i915simple/i915_state_fragprog.c deleted file mode 100644 index 51df86e0b1..0000000000 --- a/src/mesa/pipe/i915simple/i915_state_fragprog.c +++ /dev/null @@ -1,65 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915_reg.h" -#include "i915_context.h" -#include "i915_debug.h" -#include "pipe/p_util.h" - -static unsigned passthrough[] = -{ - _3DSTATE_PIXEL_SHADER_PROGRAM | ((2*3)-1), - - /* declare input color: - */ - (D0_DCL | - (REG_TYPE_T << D0_TYPE_SHIFT) | - (T_DIFFUSE << D0_NR_SHIFT) | - D0_CHANNEL_ALL), - 0, - 0, - - /* move to output color: - */ - (A0_MOV | - (REG_TYPE_OC << A0_DEST_TYPE_SHIFT) | - A0_DEST_CHANNEL_ALL | - (REG_TYPE_T << A0_SRC0_TYPE_SHIFT) | - (T_DIFFUSE << A0_SRC0_NR_SHIFT)), - 0x01230000, /* .xyzw */ - 0 -}; - -unsigned *i915_passthrough_program( unsigned *dwords ) -{ - *dwords = Elements(passthrough); - return passthrough; -} - - - - -- cgit v1.2.3 From e648a86601a7cf7a6ddfa7afa69e8c141efc45ad Mon Sep 17 00:00:00 2001 From: keithw Date: Fri, 24 Aug 2007 11:23:55 +0100 Subject: disable debug --- src/mesa/pipe/i915simple/i915_batch.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h index b11bbeedbd..d0b0946cdf 100644 --- a/src/mesa/pipe/i915simple/i915_batch.h +++ b/src/mesa/pipe/i915simple/i915_batch.h @@ -45,7 +45,7 @@ #define ADVANCE_BATCH() #define FLUSH_BATCH() do { \ - i915_dump_batchbuffer( i915, i915->batch_start, BEGIN_BATCH(0, 0) ); \ + if (0) i915_dump_batchbuffer( i915, i915->batch_start, BEGIN_BATCH(0, 0) ); \ i915->winsys->batch_flush( i915->winsys ); \ i915->batch_start = BEGIN_BATCH(0, 0); \ i915->hardware_dirty = ~0; \ -- cgit v1.2.3 From 6ef1bdbdaaf1d30d891475f3dbf27ec791ff40eb Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 12:08:11 -0600 Subject: add some warnings about bitmaps/copypixels --- src/mesa/state_tracker/st_cb_drawpixels.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index a45700e1db..e2280cdafa 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -568,6 +568,7 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, st_validate_state(st); + fprintf(stderr, "st_Bitmap not implemented yet\n"); /* XXX to do */ } @@ -581,6 +582,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, st_validate_state(st); + fprintf(stderr, "st_CopyPixels not implemented yet\n"); /* XXX to do */ } -- cgit v1.2.3 From 57394f344653f4eaec2de521a9a300a69e3bc71d Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 12:08:38 -0600 Subject: fix a tracking of changed lighting state --- src/mesa/state_tracker/st_atom_vs.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 4c83a8c5c1..7420a49520 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -31,6 +31,7 @@ */ #include "shader/prog_parameter.h" +#include "shader/prog_print.h" #include "tnl/t_vp_build.h" #include "pipe/p_context.h" @@ -104,6 +105,7 @@ static void update_vs( struct st_context *st ) /* update constants */ if (vp && params) { _mesa_load_state_parameters(st->ctx, params); + /*_mesa_print_parameter_list(params);*/ update_vs_constants(st, params); } @@ -131,7 +133,10 @@ static void update_vs( struct st_context *st ) const struct st_tracked_state st_update_vs = { .dirty = { - .mesa = _NEW_PROGRAM | _NEW_MODELVIEW | _NEW_PROJECTION, /*XXX MORE*/ + .mesa = (_NEW_PROGRAM | + _NEW_MODELVIEW | + _NEW_PROJECTION | + _NEW_LIGHT), /*XXX MORE?*/ .st = ST_NEW_VERTEX_PROGRAM, }, .update = update_vs -- cgit v1.2.3 From cd9368253011e00c2b5838e0f1dbb440f04b7f3a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 12:38:31 -0600 Subject: remove deleted i915_state_fragprog.c --- src/mesa/pipe/i915simple/Makefile | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 7e4e3d413e..670350c0a5 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -18,7 +18,6 @@ DRIVER_SOURCES = \ i915_state_dynamic.c \ i915_state_derived.c \ i915_state_emit.c \ - i915_state_fragprog.c \ i915_state_sampler.c \ i915_strings.c \ i915_prim_emit.c \ -- cgit v1.2.3 From b5c14b36addc46d8cc02e1e7d1ac85ea3ce82e8e Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 24 Aug 2007 12:52:47 +0100 Subject: Add tgsi_sse2. --- src/mesa/pipe/tgsi/exec/tgsi_core.h | 1 + src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 1730 +++++++++++++++++++++++++++++++++++ src/mesa/pipe/tgsi/exec/tgsi_sse2.h | 21 + src/mesa/sources | 5 +- 4 files changed, 1755 insertions(+), 2 deletions(-) create mode 100755 src/mesa/pipe/tgsi/exec/tgsi_sse2.c create mode 100755 src/mesa/pipe/tgsi/exec/tgsi_sse2.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_core.h b/src/mesa/pipe/tgsi/exec/tgsi_core.h index 1f5f00a38e..30ad801b6e 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_core.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_core.h @@ -7,6 +7,7 @@ #include "tgsi_exec.h" #include "tgsi_dump.h" #include "tgsi_util.h" +#include "tgsi_sse2.h" #endif // !defined TGSI_CORE_H diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c new file mode 100755 index 0000000000..36d9b86f75 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -0,0 +1,1730 @@ +#include "tgsi_platform.h" +#include "tgsi_core.h" + +#if defined USE_X86_ASM + +#define FOR_EACH_CHANNEL( CHAN )\ + for( CHAN = 0; CHAN < 4; CHAN++ ) + +#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ + if( IS_DST0_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\ + FOR_EACH_CHANNEL( CHAN )\ + IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN ) + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + +#define ALIGN16( ADDR ) (((ADDR) + 15) & ~15) + +static struct x86_reg +get_argument( + unsigned index ) +{ + return x86_make_disp( + x86_make_reg( file_REG32, reg_SP ), + (index + 1) * 4 ); +} + +static struct x86_reg +make_xmm( + unsigned xmm ) +{ + return x86_make_reg( + file_XMM, + (enum x86_reg_name) xmm ); +} + +static struct x86_reg +get_const_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_CX ); +} + +static struct x86_reg +get_const( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_const_base(), + (vec * 4 + chan) * 4 ); +} + +static struct x86_reg +get_input_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_AX ); +} + +static struct x86_reg +get_input( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_input_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_output_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_DX ); +} + +static struct x86_reg +get_output( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_output_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_temp_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_BX ); +} + +static struct x86_reg +get_temp( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_temp_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_addr( + unsigned vec, + unsigned chan ) +{ + return get_temp( + vec + TGSI_EXEC_NUM_TEMPS, + chan ); +} + +static void +emit_const( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movss( + func, + make_xmm( xmm ), + get_const( vec, chan ) ); + sse_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + +static void +emit_input( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movups( + func, + make_xmm( xmm ), + get_input( vec, chan ) ); +} + +static void +emit_output( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movups( + func, + get_output( vec, chan ), + make_xmm( xmm ) ); +} + +static void +emit_tempf( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movaps( + func, + make_xmm( xmm ), + get_temp( vec, chan ) ); +} + +static void +emit_temps (struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan) +{ + sse_movaps (func, + get_temp (vec, chan), + make_xmm (xmm)); +} + +static void +emit_addrf( struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_tempf( func, xmm, vec + TGSI_EXEC_NUM_TEMPS, chan ); +} + +static void +emit_addrs( struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_temps( func, xmm, vec + TGSI_EXEC_NUM_TEMPS, chan ); +} + +static void +emit_abs (struct x86_function *func, + unsigned xmm) +{ + sse_andps (func, + make_xmm (xmm), + get_temp (TGSI_EXEC_TEMP_7FFFFFFF_I, + TGSI_EXEC_TEMP_7FFFFFFF_C)); +} + +static void +emit_neg (struct x86_function *func, + unsigned xmm) +{ + sse_xorps (func, + make_xmm (xmm), + get_temp (TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C)); +} + +static void +emit_setsign (struct x86_function *func, + unsigned xmm) +{ + sse_orps (func, + make_xmm (xmm), + get_temp (TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C)); +} + +static void +emit_add (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + sse_addps (func, + make_xmm (xmm_dst), + make_xmm (xmm_src)); +} + +static GLfloat g_cos_storage[4 + 3]; + +static void +cos4f (void) +{ + GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_cos_storage); + +#ifdef WIN32 + store[0] = (GLfloat) cos ((GLdouble) store[0]); + store[1] = (GLfloat) cos ((GLdouble) store[1]); + store[2] = (GLfloat) cos ((GLdouble) store[2]); + store[3] = (GLfloat) cos ((GLdouble) store[3]); +#else + store[0] = cosf (store[0]); + store[1] = cosf (store[1]); + store[2] = cosf (store[2]); + store[3] = cosf (store[3]); +#endif +} + +static void +emit_cos (struct x86_function *func, + unsigned xmm_dst) +{ + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_mov_reg_imm (func, + x86_make_reg (file_REG32, reg_AX), + ALIGN16((GLint) g_cos_storage)); + sse_movaps (func, + x86_deref (x86_make_reg (file_REG32, reg_AX)), + make_xmm (xmm_dst)); + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_push (func, + x86_make_reg (file_REG32, reg_BX)); + x86_push (func, + x86_make_reg (file_REG32, reg_CX)); + x86_push (func, + x86_make_reg (file_REG32, reg_DX)); + x86_call (func, + (GLubyte *) cos4f); + x86_pop (func, + x86_make_reg (file_REG32, reg_DX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_CX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_BX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); + sse_movaps (func, + make_xmm (xmm_dst), + x86_deref (x86_make_reg (file_REG32, reg_AX))); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); +} + +static GLfloat g_sin_storage[4 + 3]; + +static void +sin4f (void) +{ + GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_sin_storage); + +#ifdef WIN32 + store[0] = (GLfloat) sin ((GLdouble) store[0]); + store[1] = (GLfloat) sin ((GLdouble) store[1]); + store[2] = (GLfloat) sin ((GLdouble) store[2]); + store[3] = (GLfloat) sin ((GLdouble) store[3]); +#else + store[0] = sin (store[0]); + store[1] = sin (store[1]); + store[2] = sin (store[2]); + store[3] = sin (store[3]); +#endif +} + +static void +emit_sin (struct x86_function *func, + unsigned xmm_dst) +{ + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_mov_reg_imm (func, + x86_make_reg (file_REG32, reg_AX), + ALIGN16((GLint) g_sin_storage)); + sse_movaps (func, + x86_deref (x86_make_reg (file_REG32, reg_AX)), + make_xmm (xmm_dst)); + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_push (func, + x86_make_reg (file_REG32, reg_BX)); + x86_push (func, + x86_make_reg (file_REG32, reg_CX)); + x86_push (func, + x86_make_reg (file_REG32, reg_DX)); + x86_call (func, + (GLubyte *) sin4f); + x86_pop (func, + x86_make_reg (file_REG32, reg_DX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_CX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_BX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); + sse_movaps (func, + make_xmm (xmm_dst), + x86_deref (x86_make_reg (file_REG32, reg_AX))); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); +} + +static void +emit_mov (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + sse_movups (func, + make_xmm (xmm_dst), + make_xmm (xmm_src)); +} + +static void +emit_mul (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + sse_mulps (func, + make_xmm (xmm_dst), + make_xmm (xmm_src)); +} + +static GLfloat g_pow_storage[4 + 4 + 3]; + +static void +pow4f (void) +{ + GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_pow_storage); + +#ifdef WIN32 + store[0] = (GLfloat) pow ((GLdouble) store[0], (GLdouble) store[4]); + store[1] = (GLfloat) pow ((GLdouble) store[1], (GLdouble) store[5]); + store[2] = (GLfloat) pow ((GLdouble) store[2], (GLdouble) store[6]); + store[3] = (GLfloat) pow ((GLdouble) store[3], (GLdouble) store[7]); +#else + store[0] = powf (store[0], store[4]); + store[1] = powf (store[1], store[5]); + store[2] = powf (store[2], store[6]); + store[3] = powf (store[3], store[7]); +#endif +} + +static void +emit_pow (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_mov_reg_imm (func, + x86_make_reg (file_REG32, reg_AX), + ALIGN16((GLint) g_pow_storage)); + sse_movaps (func, + x86_make_disp (x86_make_reg (file_REG32, reg_AX), 0), + make_xmm (xmm_dst)); + sse_movaps (func, + x86_make_disp (x86_make_reg (file_REG32, reg_AX), 16), + make_xmm (xmm_src)); + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_push (func, + x86_make_reg (file_REG32, reg_BX)); + x86_push (func, + x86_make_reg (file_REG32, reg_CX)); + x86_push (func, + x86_make_reg (file_REG32, reg_DX)); + x86_call (func, + (GLubyte *) pow4f); + x86_pop (func, + x86_make_reg (file_REG32, reg_DX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_CX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_BX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); + sse_movaps (func, + make_xmm (xmm_dst), + x86_deref (x86_make_reg (file_REG32, reg_AX))); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); +} + +static GLfloat g_ex2_storage[4 + 3]; + +static void +ex24f (void) +{ + GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_ex2_storage); + + store[0] = (GLfloat) pow (2.0, (GLdouble) store[0]); + store[1] = (GLfloat) pow (2.0, (GLdouble) store[1]); + store[2] = (GLfloat) pow (2.0, (GLdouble) store[2]); + store[3] = (GLfloat) pow (2.0, (GLdouble) store[3]); +} + +static void +emit_ex2 (struct x86_function *func, + unsigned xmm_dst) +{ + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_mov_reg_imm (func, + x86_make_reg (file_REG32, reg_AX), + ALIGN16((GLint) g_ex2_storage)); + sse_movaps (func, + x86_deref (x86_make_reg (file_REG32, reg_AX)), + make_xmm (xmm_dst)); + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_push (func, + x86_make_reg (file_REG32, reg_BX)); + x86_push (func, + x86_make_reg (file_REG32, reg_CX)); + x86_push (func, + x86_make_reg (file_REG32, reg_DX)); + x86_call (func, + (GLubyte *) ex24f); + x86_pop (func, + x86_make_reg (file_REG32, reg_DX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_CX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_BX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); + sse_movaps (func, + make_xmm (xmm_dst), + x86_deref (x86_make_reg (file_REG32, reg_AX))); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); +} + +static GLfloat g_lg2_storage[4 + 3]; + +static void +lg24f (void) +{ + GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_lg2_storage); + + store[0] = LOG2 (store[0]); + store[1] = LOG2 (store[1]); + store[2] = LOG2 (store[2]); + store[3] = LOG2 (store[3]); +} + +static void +emit_lg2 (struct x86_function *func, + unsigned xmm_dst) +{ + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_mov_reg_imm (func, + x86_make_reg (file_REG32, reg_AX), + ALIGN16((GLint) g_lg2_storage)); + sse_movaps (func, + x86_deref (x86_make_reg (file_REG32, reg_AX)), + make_xmm (xmm_dst)); + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_push (func, + x86_make_reg (file_REG32, reg_BX)); + x86_push (func, + x86_make_reg (file_REG32, reg_CX)); + x86_push (func, + x86_make_reg (file_REG32, reg_DX)); + x86_call (func, + (GLubyte *) lg24f); + x86_pop (func, + x86_make_reg (file_REG32, reg_DX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_CX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_BX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); + sse_movaps (func, + make_xmm (xmm_dst), + x86_deref (x86_make_reg (file_REG32, reg_AX))); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); +} + +static GLfloat g_flr_storage[4 + 3]; + +static void +flr4f (void) +{ + GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_flr_storage); + + store[0] = (GLfloat) floor ((GLdouble) store[0]); + store[1] = (GLfloat) floor ((GLdouble) store[1]); + store[2] = (GLfloat) floor ((GLdouble) store[2]); + store[3] = (GLfloat) floor ((GLdouble) store[3]); +} + +static void +emit_flr (struct x86_function *func, + unsigned xmm_dst) +{ + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_mov_reg_imm (func, + x86_make_reg (file_REG32, reg_AX), + ALIGN16((GLint) g_flr_storage)); + sse_movaps (func, + x86_deref (x86_make_reg (file_REG32, reg_AX)), + make_xmm (xmm_dst)); + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_push (func, + x86_make_reg (file_REG32, reg_BX)); + x86_push (func, + x86_make_reg (file_REG32, reg_CX)); + x86_push (func, + x86_make_reg (file_REG32, reg_DX)); + x86_call (func, + (GLubyte *) flr4f); + x86_pop (func, + x86_make_reg (file_REG32, reg_DX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_CX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_BX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); + sse_movaps (func, + make_xmm (xmm_dst), + x86_deref (x86_make_reg (file_REG32, reg_AX))); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); +} + +static GLfloat g_frc_storage[4 + 3]; + +static void +frc4f (void) +{ + GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_frc_storage); + + store[0] -= (GLfloat) floor ((GLdouble) store[0]); + store[1] -= (GLfloat) floor ((GLdouble) store[1]); + store[2] -= (GLfloat) floor ((GLdouble) store[2]); + store[3] -= (GLfloat) floor ((GLdouble) store[3]); +} + +static void +emit_frc (struct x86_function *func, + unsigned xmm_dst) +{ + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_mov_reg_imm (func, + x86_make_reg (file_REG32, reg_AX), + ALIGN16((GLint) g_frc_storage)); + sse_movaps (func, + x86_deref (x86_make_reg (file_REG32, reg_AX)), + make_xmm (xmm_dst)); + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_push (func, + x86_make_reg (file_REG32, reg_BX)); + x86_push (func, + x86_make_reg (file_REG32, reg_CX)); + x86_push (func, + x86_make_reg (file_REG32, reg_DX)); + x86_call (func, + (GLubyte *) frc4f); + x86_pop (func, + x86_make_reg (file_REG32, reg_DX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_CX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_BX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); + sse_movaps (func, + make_xmm (xmm_dst), + x86_deref (x86_make_reg (file_REG32, reg_AX))); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); +} + +static void +emit_rcp (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + sse2_rcpps (func, + make_xmm (xmm_dst), + make_xmm (xmm_src)); +} + +static void +emit_rsqrt (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + sse_rsqrtps (func, + make_xmm (xmm_dst), + make_xmm (xmm_src)); +} + +static void +emit_sub (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + sse_subps (func, + make_xmm (xmm_dst), + make_xmm (xmm_src)); +} + +static void +emit_fetch (struct x86_function *func, + unsigned xmm, + const struct tgsi_full_src_register *reg, + const unsigned chan_index) +{ + unsigned swizzle = tgsi_util_get_full_src_register_extswizzle (reg, chan_index); + + switch (swizzle) + { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch (reg->SrcRegister.File) + { + case TGSI_FILE_CONSTANT: + emit_const (func, xmm, reg->SrcRegister.Index, swizzle); + break; + + case TGSI_FILE_INPUT: + emit_input (func, xmm, reg->SrcRegister.Index, swizzle); + break; + + case TGSI_FILE_TEMPORARY: + emit_tempf (func, xmm, reg->SrcRegister.Index, swizzle); + break; + + default: + assert (0); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + emit_tempf (func, + xmm, + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C); + break; + + case TGSI_EXTSWIZZLE_ONE: + emit_tempf (func, + xmm, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C); + break; + + default: + assert (0); + } + + switch (tgsi_util_get_full_src_register_sign_mode (reg, chan_index)) + { + case TGSI_UTIL_SIGN_CLEAR: + emit_abs (func, xmm); + break; + + case TGSI_UTIL_SIGN_SET: + emit_setsign (func, xmm); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + emit_neg (func, xmm); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } +} + +static void +emit_store (struct x86_function *func, + unsigned xmm, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + unsigned chan_index) +{ + switch (reg->DstRegister.File) + { + case TGSI_FILE_OUTPUT: + emit_output (func, xmm, reg->DstRegister.Index, chan_index); + break; + + case TGSI_FILE_TEMPORARY: + emit_temps (func, xmm, reg->DstRegister.Index, chan_index); + break; + + case TGSI_FILE_ADDRESS: + emit_addrs( func, xmm, reg->DstRegister.Index, chan_index ); + break; + + default: + assert (0); + } + + switch (inst->Instruction.Saturate) + { + case TGSI_SAT_NONE: + break; + + case TGSI_SAT_ZERO_ONE: +// assert (0); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert (0); + break; + } +} + +#define FETCH(XMM,INDEX,CHAN)\ + emit_fetch (func, XMM, &inst->FullSrcRegisters[INDEX], CHAN) + +#define STORE(XMM,INDEX,CHAN)\ + emit_store (func, XMM, &inst->FullDstRegisters[INDEX], inst, CHAN) + +static void +emit_kil (struct x86_function *func, + const struct tgsi_full_src_register *reg) +{ + unsigned uniquemask; + unsigned registers[4]; + unsigned nextregister = 0; + unsigned firstchan = ~0; + unsigned chan_index; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + unsigned swizzle; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle (reg, chan_index); + + /* check if the component has not been already tested */ + if (!(uniquemask & (1 << swizzle))) + { + uniquemask |= 1 << swizzle; + + /* allocate register */ + registers[chan_index] = nextregister; + emit_fetch (func, nextregister, reg, chan_index); + nextregister++; + + /* mark the first channel used */ + if (firstchan == ~0) + firstchan = chan_index; + } + } + + x86_push (func, + x86_make_reg (file_REG32, reg_AX)); + x86_push (func, + x86_make_reg (file_REG32, reg_DX)); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + if (uniquemask & (1 << chan_index)) + { + sse_cmpps (func, + make_xmm (registers[chan_index]), + get_temp (TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C), + cc_LessThan); + + if (chan_index == firstchan) + { + sse_pmovmskb (func, + x86_make_reg (file_REG32, reg_AX), + make_xmm (registers[chan_index])); + } + else + { + sse_pmovmskb (func, + x86_make_reg (file_REG32, reg_DX), + make_xmm (registers[chan_index])); + x86_or (func, + x86_make_reg (file_REG32, reg_AX), + x86_make_reg (file_REG32, reg_DX)); + } + } + } + + x86_or (func, + get_temp(TGSI_EXEC_TEMP_KILMASK_I, + TGSI_EXEC_TEMP_KILMASK_C), + x86_make_reg (file_REG32, reg_AX)); + + x86_pop (func, + x86_make_reg (file_REG32, reg_DX)); + x86_pop (func, + x86_make_reg (file_REG32, reg_AX)); +} + +static void +emit_setcc (struct x86_function *func, + struct tgsi_full_instruction *inst, + enum sse_cc cc) +{ + unsigned chan_index; + + FOR_EACH_ENABLED_CHANNEL + { + emit_fetch (func, 0, &inst->FullSrcRegisters[0], chan_index); + emit_fetch (func, 1, &inst->FullSrcRegisters[1], chan_index); + + sse_cmpps (func, + make_xmm (0), + make_xmm (1), + cc); + + sse_andps (func, + make_xmm (0), + get_temp (TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C)); + + emit_store (func, 0, &inst->FullDstRegisters[0], inst, chan_index); + } +} + +static void +emit_cmp (struct x86_function *func, + struct tgsi_full_instruction *inst) +{ + unsigned chan_index; + + FOR_EACH_ENABLED_CHANNEL + { + emit_fetch (func, 0, &inst->FullSrcRegisters[0], chan_index); + emit_fetch (func, 1, &inst->FullSrcRegisters[1], chan_index); + emit_fetch (func, 2, &inst->FullSrcRegisters[2], chan_index); + + sse_cmpps (func, + make_xmm (0), + get_temp (TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C), + cc_LessThan); + + sse_andps (func, + make_xmm (1), + make_xmm (0)); + sse_andnps (func, + make_xmm (0), + make_xmm (2)); + sse_orps (func, + make_xmm (0), + make_xmm (1)); + + emit_store (func, 0, &inst->FullDstRegisters[0], inst, chan_index); + } +} + +static void +emit_f2it( struct x86_function *func, + unsigned xmm ) +{ + sse2_cvttps2dq( func, make_xmm( xmm ), make_xmm( xmm ) ); +} + +static void +emit_instruction (struct x86_function *func, + struct tgsi_full_instruction *inst) +{ + unsigned chan_index; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: + FOR_EACH_ENABLED_CHANNEL { + FETCH( 0, 0, chan_index ); + emit_f2it( func, 0 ); + STORE( 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_LIT: + if (IS_CHANNEL_ENABLED(CHAN_X) || IS_CHANNEL_ENABLED(CHAN_W)) + { + emit_tempf (func, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C); + if (IS_CHANNEL_ENABLED(CHAN_X)) + STORE(0, 0, CHAN_X); + if (IS_CHANNEL_ENABLED(CHAN_W)) + STORE(0, 0, CHAN_W); + } + if (IS_CHANNEL_ENABLED(CHAN_Y) || IS_CHANNEL_ENABLED(CHAN_Z)) + { + if (IS_CHANNEL_ENABLED(CHAN_Y)) + { + FETCH(0, 0, CHAN_X); + sse_maxps (func, + make_xmm (0), + get_temp (TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C)); + STORE(0, 0, CHAN_Y); + } + if (IS_CHANNEL_ENABLED(CHAN_Z)) + { + FETCH(1, 0, CHAN_Y); + sse_maxps (func, + make_xmm (1), + get_temp (TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C)); + + FETCH(2, 0, CHAN_W); + sse_minps (func, + make_xmm (2), + get_temp (TGSI_EXEC_TEMP_128_I, + TGSI_EXEC_TEMP_128_C)); + sse_maxps (func, + make_xmm (2), + get_temp (TGSI_EXEC_TEMP_MINUS_128_I, + TGSI_EXEC_TEMP_MINUS_128_C)); + + emit_pow (func, 1, 2); + + FETCH(0, 0, CHAN_X); + sse_xorps (func, make_xmm (2), make_xmm (2)); + sse_cmpps (func, + make_xmm (2), + make_xmm (0), + cc_LessThanEqual); + sse_andps (func, + make_xmm (2), + make_xmm (1)); + + emit_store (func, 2, &inst->FullDstRegisters[0], inst, CHAN_Z); + } + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH(0, 0, CHAN_X); + emit_rcp (func, 0, 0); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH(0, 0, CHAN_X); + emit_rsqrt (func, 0, 0); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_EXP: + assert (0); + break; + + case TGSI_OPCODE_LOG: + assert (0); + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + FETCH(1, 1, chan_index); + emit_mul (func, 0, 1); + + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + FETCH(1, 1, chan_index); + emit_add (func, 0, 1); + + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH(0, 0, CHAN_X); + FETCH(1, 1, CHAN_X); + emit_mul (func, 0, 1); + + FETCH(1, 0, CHAN_Y); + FETCH(2, 1, CHAN_Y); + emit_mul (func, 1, 2); + emit_add (func, 0, 1); + + FETCH(1, 0, CHAN_Z); + FETCH(2, 1, CHAN_Z); + emit_mul (func, 1, 2); + emit_add (func, 0, 1); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH(0, 0, CHAN_X); + FETCH(1, 1, CHAN_X); + emit_mul (func, 0, 1); + + FETCH(1, 0, CHAN_Y); + FETCH(2, 1, CHAN_Y); + emit_mul (func, 1, 2); + emit_add (func, 0, 1); + + FETCH(1, 0, CHAN_Z); + FETCH(2, 1, CHAN_Z); + emit_mul (func, 1, 2); + emit_add (func, 0, 1); + + FETCH(1, 0, CHAN_W); + FETCH(2, 1, CHAN_W); + emit_mul (func, 1, 2); + emit_add (func, 0, 1); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_DST: + IF_IS_CHANNEL_ENABLED(CHAN_X) + { + emit_tempf (func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C); + STORE(0, 0, CHAN_X); + } + IF_IS_CHANNEL_ENABLED(CHAN_Y) + { + FETCH(0, 0, CHAN_Y); + FETCH(1, 1, CHAN_Y); + emit_mul (func, 0, 1); + STORE(0, 0, CHAN_Y); + } + IF_IS_CHANNEL_ENABLED(CHAN_Z) + { + FETCH(0, 0, CHAN_Z); + STORE(0, 0, CHAN_Z); + } + IF_IS_CHANNEL_ENABLED(CHAN_W) + { + FETCH(0, 1, CHAN_W); + STORE(0, 0, CHAN_W); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + FETCH(1, 1, chan_index); + sse_minps (func, + make_xmm (0), + make_xmm (1)); + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + FETCH(1, 1, chan_index); + sse_maxps (func, + make_xmm (0), + make_xmm (1)); + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + emit_setcc (func, inst, cc_LessThan); + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + emit_setcc (func, inst, cc_NotLessThan); + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + FETCH(1, 1, chan_index); + FETCH(2, 2, chan_index); + emit_mul (func, 0, 1); + emit_add (func, 0, 2); + + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + FETCH(1, 1, chan_index); + emit_sub (func, 0, 1); + + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + FETCH(1, 1, chan_index); + FETCH(2, 2, chan_index); + emit_sub (func, 1, 2); + emit_mul (func, 0, 1); + emit_add (func, 0, 2); + + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_CND: + assert (0); + break; + + case TGSI_OPCODE_CND0: + assert (0); + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert (0); + break; + + case TGSI_OPCODE_INDEX: + assert (0); + break; + + case TGSI_OPCODE_NEGATE: + assert (0); + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + emit_frc (func, 0); + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_CLAMP: + assert (0); + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + emit_flr (func, 0); + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_ROUND: + assert (0); + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH(0, 0, CHAN_X); + emit_ex2 (func, 0); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH(0, 0, CHAN_X); + emit_lg2 (func, 0); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH(0, 0, CHAN_X); + FETCH(1, 1, CHAN_X); + emit_pow (func, 0, 1); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + if (IS_CHANNEL_ENABLED(CHAN_X) || IS_CHANNEL_ENABLED(CHAN_Y)) + { + FETCH(1, 1, CHAN_Z); + FETCH(3, 0, CHAN_Z); + } + if (IS_CHANNEL_ENABLED(CHAN_X) || IS_CHANNEL_ENABLED(CHAN_Z)) + { + FETCH(0, 0, CHAN_Y); + FETCH(4, 1, CHAN_Y); + } + IF_IS_CHANNEL_ENABLED(CHAN_X) + { + emit_mov (func, 2, 0); + emit_mul (func, 2, 1); + emit_mov (func, 5, 3); + emit_mul (func, 5, 4); + emit_sub (func, 2, 5); + STORE(2, 0, CHAN_X); + } + + if (IS_CHANNEL_ENABLED(CHAN_Y) || IS_CHANNEL_ENABLED(CHAN_Z)) + { + FETCH(2, 1, CHAN_X); + FETCH(5, 0, CHAN_X); + } + IF_IS_CHANNEL_ENABLED(CHAN_Y) + { + emit_mul (func, 3, 2); + emit_mul (func, 1, 5); + emit_sub (func, 3, 1); + STORE(3, 0, CHAN_Y); + } + + IF_IS_CHANNEL_ENABLED(CHAN_Z) + { + emit_mul (func, 5, 4); + emit_mul (func, 0, 2); + emit_sub (func, 5, 0); + STORE(5, 0, CHAN_Z); + } + + IF_IS_CHANNEL_ENABLED(CHAN_W) + { + FETCH(0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C); + STORE(0, 0, CHAN_W); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + assert (0); + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_ENABLED_CHANNEL + { + FETCH(0, 0, chan_index); + emit_abs (func, 0); + + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_RCC: + assert (0); + break; + + case TGSI_OPCODE_DPH: + FETCH(0, 0, CHAN_X); + FETCH(1, 1, CHAN_X); + emit_mul (func, 0, 1); + + FETCH(1, 0, CHAN_Y); + FETCH(2, 1, CHAN_Y); + emit_mul (func, 1, 2); + emit_add (func, 0, 1); + + FETCH(1, 0, CHAN_Z); + FETCH(2, 1, CHAN_Z); + emit_mul (func, 1, 2); + emit_add (func, 0, 1); + + FETCH(1, 1, CHAN_W); + emit_add (func, 0, 1); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_COS: + FETCH(0, 0, CHAN_X); + emit_cos (func, 0); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_DDX: + assert (0); + break; + + case TGSI_OPCODE_DDY: + assert (0); + break; + + case TGSI_OPCODE_KIL: + emit_kil (func, &inst->FullSrcRegisters[0]); + break; + + case TGSI_OPCODE_PK2H: + assert (0); + break; + + case TGSI_OPCODE_PK2US: + assert (0); + break; + + case TGSI_OPCODE_PK4B: + assert (0); + break; + + case TGSI_OPCODE_PK4UB: + assert (0); + break; + + case TGSI_OPCODE_RFL: + assert (0); + break; + + case TGSI_OPCODE_SEQ: + assert (0); + break; + + case TGSI_OPCODE_SFL: + assert (0); + break; + + case TGSI_OPCODE_SGT: + assert (0); + break; + + case TGSI_OPCODE_SIN: + FETCH(0, 0, CHAN_X); + emit_sin (func, 0); + + FOR_EACH_ENABLED_CHANNEL + { + STORE(0, 0, chan_index); + } + break; + + case TGSI_OPCODE_SLE: + assert (0); + break; + + case TGSI_OPCODE_SNE: + assert (0); + break; + + case TGSI_OPCODE_STR: + assert (0); + break; + + case TGSI_OPCODE_TEX: + assert (0); + break; + + case TGSI_OPCODE_TXD: + assert (0); + break; + + case TGSI_OPCODE_UP2H: + assert (0); + break; + + case TGSI_OPCODE_UP2US: + assert (0); + break; + + case TGSI_OPCODE_UP4B: + assert (0); + break; + + case TGSI_OPCODE_UP4UB: + assert (0); + break; + + case TGSI_OPCODE_X2D: + assert (0); + break; + + case TGSI_OPCODE_ARA: + assert (0); + break; + + case TGSI_OPCODE_ARR: + assert (0); + break; + + case TGSI_OPCODE_BRA: + assert (0); + break; + + case TGSI_OPCODE_CAL: + assert (0); + break; + + case TGSI_OPCODE_RET: + assert (0); + break; + + case TGSI_OPCODE_SSG: + assert (0); + break; + + case TGSI_OPCODE_CMP: + emit_cmp (func, inst); + break; + + case TGSI_OPCODE_SCS: + IF_IS_CHANNEL_ENABLED(CHAN_X) + { + FETCH(0, 0, CHAN_X); + emit_cos (func, 0); + STORE(0, 0, CHAN_X); + } + + IF_IS_CHANNEL_ENABLED(CHAN_Y) + { + FETCH(0, 0, CHAN_Y); + emit_sin (func, 0); + STORE(0, 0, CHAN_Y); + } + + IF_IS_CHANNEL_ENABLED(CHAN_Z) + { + FETCH(0, TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C); + STORE(0, 0, CHAN_Z); + } + + IF_IS_CHANNEL_ENABLED(CHAN_W) + { + FETCH(0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C); + STORE(0, 0, CHAN_W); + } + break; + + case TGSI_OPCODE_TXB: + assert (0); + break; + + case TGSI_OPCODE_NRM: + assert (0); + break; + + case TGSI_OPCODE_DIV: + assert (0); + break; + + case TGSI_OPCODE_DP2: + assert (0); + break; + + case TGSI_OPCODE_TXL: + assert (0); + break; + + case TGSI_OPCODE_BRK: + assert (0); + break; + + case TGSI_OPCODE_IF: + assert (0); + break; + + case TGSI_OPCODE_LOOP: + assert (0); + break; + + case TGSI_OPCODE_REP: + assert (0); + break; + + case TGSI_OPCODE_ELSE: + assert (0); + break; + + case TGSI_OPCODE_ENDIF: + assert (0); + break; + + case TGSI_OPCODE_ENDLOOP: + assert (0); + break; + + case TGSI_OPCODE_ENDREP: + assert (0); + break; + + case TGSI_OPCODE_PUSHA: + assert (0); + break; + + case TGSI_OPCODE_POPA: + assert (0); + break; + + case TGSI_OPCODE_CEIL: + assert (0); + break; + + case TGSI_OPCODE_I2F: + assert (0); + break; + + case TGSI_OPCODE_NOT: + assert (0); + break; + + case TGSI_OPCODE_TRUNC: + assert (0); + break; + + case TGSI_OPCODE_SHL: + assert (0); + break; + + case TGSI_OPCODE_SHR: + assert (0); + break; + + case TGSI_OPCODE_AND: + assert (0); + break; + + case TGSI_OPCODE_OR: + assert (0); + break; + + case TGSI_OPCODE_MOD: + assert (0); + break; + + case TGSI_OPCODE_XOR: + assert (0); + break; + + case TGSI_OPCODE_SAD: + assert (0); + break; + + case TGSI_OPCODE_TXF: + assert (0); + break; + + case TGSI_OPCODE_TXQ: + assert (0); + break; + + case TGSI_OPCODE_CONT: + assert (0); + break; + + case TGSI_OPCODE_EMIT: + assert (0); + break; + + case TGSI_OPCODE_ENDPRIM: + assert (0); + break; + + default: + assert (0); + } +} + +GLboolean +tgsi_emit_sse (struct tgsi_token *tokens, + struct x86_function *function) +{ + struct tgsi_parse_context parse; + + x86_init_func (function); + + x86_mov (function, get_input_base (), get_argument (0)); + x86_mov (function, get_output_base (), get_argument (1)); + x86_mov (function, get_const_base (), get_argument (2)); + x86_mov (function, get_temp_base (), get_argument (3)); + + tgsi_parse_init (&parse, tokens); + + while (!tgsi_parse_end_of_tokens (&parse)) + { + tgsi_parse_token (&parse); + + switch (parse.FullToken.Token.Type) + { + case TGSI_TOKEN_TYPE_DECLARATION: + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + emit_instruction (function, &parse.FullToken.FullInstruction); + break; + + default: + assert (0); + } + } + + tgsi_parse_free (&parse); + +#ifdef WIN32 + x86_retw (function, 16); +#else + x86_ret (function); +#endif + + return GL_FALSE; +} + +#endif diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h new file mode 100755 index 0000000000..4a39658484 --- /dev/null +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h @@ -0,0 +1,21 @@ +#if !defined TGSI_SSE2_H +#define TGSI_SSE2_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_token; +struct x86_function; + +unsigned +tgsi_emit_sse2( + struct tgsi_token *tokens, + struct x86_function *function ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_SSE2_H + diff --git a/src/mesa/sources b/src/mesa/sources index ec30a889f6..61a8d580ad 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -166,11 +166,12 @@ DRAW_SOURCES = \ pipe/draw/draw_twoside.c \ pipe/draw/draw_unfilled.c -TGSICORE_SOURCES = \ +TGSIEXEC_SOURCES = \ pipe/tgsi/exec/tgsi_build.c \ pipe/tgsi/exec/tgsi_dump.c \ pipe/tgsi/exec/tgsi_exec.c \ pipe/tgsi/exec/tgsi_parse.c \ + pipe/tgsi/exec/tgsi_sse2.c \ pipe/tgsi/exec/tgsi_util.c TGSIDECO_SOURCES = \ @@ -359,7 +360,7 @@ SOLO_SOURCES = \ $(VBO_SOURCES) \ $(VF_SOURCES) \ $(DRAW_SOURCES) \ - $(TGSICORE_SOURCES) \ + $(TGSIEXEC_SOURCES) \ $(TGSIDECO_SOURCES) \ $(TGSIMESA_SOURCES) \ $(STATETRACKER_SOURCES) \ -- cgit v1.2.3 From 2c7b74725bfbd0a022ebfada4736d6cb8ac28047 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 24 Aug 2007 16:23:02 +0100 Subject: Indent. Remove GL dependencies. Simplify. Add rtasm instructions. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 2690 +++++++++++++++++------------------ src/mesa/x86/rtasm/x86sse.c | 54 + src/mesa/x86/rtasm/x86sse.h | 7 + 3 files changed, 1400 insertions(+), 1351 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 36d9b86f75..359775fdfb 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -1,7 +1,8 @@ #include "tgsi_platform.h" #include "tgsi_core.h" +#include "x86/rtasm/x86sse.h" -#if defined USE_X86_ASM +#if defined(USE_X86_ASM) || defined(SLANG_X86) #define FOR_EACH_CHANNEL( CHAN )\ for( CHAN = 0; CHAN < 4; CHAN++ ) @@ -181,148 +182,257 @@ emit_tempf( } static void -emit_temps (struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan) +emit_temps ( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) { - sse_movaps (func, - get_temp (vec, chan), - make_xmm (xmm)); + sse_movaps( + func, + get_temp( vec, chan ), + make_xmm( xmm ) ); } static void -emit_addrf( struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) +emit_addrf( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) { - emit_tempf( func, xmm, vec + TGSI_EXEC_NUM_TEMPS, chan ); + emit_tempf( + func, + xmm, + vec + TGSI_EXEC_NUM_TEMPS, + chan ); } static void -emit_addrs( struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) +emit_addrs( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) { - emit_temps( func, xmm, vec + TGSI_EXEC_NUM_TEMPS, chan ); + emit_temps( + func, + xmm, + vec + TGSI_EXEC_NUM_TEMPS, + chan ); } static void -emit_abs (struct x86_function *func, - unsigned xmm) +emit_abs( + struct x86_function *func, + unsigned xmm ) { - sse_andps (func, - make_xmm (xmm), - get_temp (TGSI_EXEC_TEMP_7FFFFFFF_I, - TGSI_EXEC_TEMP_7FFFFFFF_C)); + sse_andps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_7FFFFFFF_I, + TGSI_EXEC_TEMP_7FFFFFFF_C ) ); } static void -emit_neg (struct x86_function *func, - unsigned xmm) +emit_neg( + struct x86_function *func, + unsigned xmm ) { - sse_xorps (func, - make_xmm (xmm), - get_temp (TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C)); + sse_xorps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); } static void -emit_setsign (struct x86_function *func, - unsigned xmm) +emit_setsign( + struct x86_function *func, + unsigned xmm ) { - sse_orps (func, - make_xmm (xmm), - get_temp (TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C)); + sse_orps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); } static void -emit_add (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) +emit_add( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) { - sse_addps (func, - make_xmm (xmm_dst), - make_xmm (xmm_src)); + sse_addps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); } -static GLfloat g_cos_storage[4 + 3]; +static void +emit_push_abcd( + struct x86_function *func ) +{ + x86_push( + func, + x86_make_reg( file_REG32, reg_AX ) ); + x86_push( + func, + x86_make_reg( file_REG32, reg_BX ) ); + x86_push( + func, + x86_make_reg( file_REG32, reg_CX ) ); + x86_push( + func, + x86_make_reg( file_REG32, reg_DX ) ); +} static void -cos4f (void) +emit_pop_dcba( + struct x86_function *func ) { - GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_cos_storage); + x86_pop( + func, + x86_make_reg( file_REG32, reg_DX ) ); + x86_pop( + func, + x86_make_reg( file_REG32, reg_CX ) ); + x86_pop( + func, + x86_make_reg( file_REG32, reg_BX ) ); + x86_pop( + func, + x86_make_reg( file_REG32, reg_AX ) ); +} + +static void +emit_func_call1( + struct x86_function *func, + unsigned xmm_dst, + unsigned storage, + unsigned char *code ) +{ + x86_push( + func, + x86_make_reg( file_REG32, reg_AX ) ); + x86_mov_reg_imm( + func, + x86_make_reg( file_REG32, reg_AX ), + storage ); + sse_movaps( + func, + x86_deref( x86_make_reg( file_REG32, reg_AX ) ), + make_xmm( xmm_dst ) ); + emit_push_abcd( + func ); + x86_call( + func, + code ); + emit_pop_dcba( + func ); + sse_movaps( + func, + make_xmm( xmm_dst ), + x86_deref( x86_make_reg( file_REG32, reg_AX ) ) ); + x86_pop( + func, + x86_make_reg( file_REG32, reg_AX ) ); +} + +static void +emit_func_call2( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src, + unsigned storage, + unsigned char *code ) +{ + x86_push( + func, + x86_make_reg( file_REG32, reg_AX ) ); + x86_mov_reg_imm( + func, + x86_make_reg( file_REG32, reg_AX ), + storage ); + sse_movaps( + func, + x86_deref( x86_make_reg( file_REG32, reg_AX ) ), + make_xmm( xmm_dst ) ); + sse_movaps( + func, + x86_make_disp( x86_make_reg( file_REG32, reg_AX ), 16 ), + make_xmm( xmm_src ) ); + emit_push_abcd( + func ); + x86_call( + func, + code ); + emit_pop_dcba( + func ); + sse_movaps( + func, + make_xmm( xmm_dst ), + x86_deref( x86_make_reg( file_REG32, reg_AX ) ) ); + x86_pop( + func, + x86_make_reg( file_REG32, reg_AX ) ); +} + +/* XXX: move into machine context */ +static float g_cos_storage[4 + 3]; + +static void +cos4f( void ) +{ + float *store = (float *) ALIGN16( (unsigned) g_cos_storage ); #ifdef WIN32 - store[0] = (GLfloat) cos ((GLdouble) store[0]); - store[1] = (GLfloat) cos ((GLdouble) store[1]); - store[2] = (GLfloat) cos ((GLdouble) store[2]); - store[3] = (GLfloat) cos ((GLdouble) store[3]); + store[0] = (float) cos( (double) store[0] ); + store[1] = (float) cos( (double) store[1] ); + store[2] = (float) cos( (double) store[2] ); + store[3] = (float) cos( (double) store[3] ); #else - store[0] = cosf (store[0]); - store[1] = cosf (store[1]); - store[2] = cosf (store[2]); - store[3] = cosf (store[3]); + store[0] = cosf( store[0] ); + store[1] = cosf( store[1] ); + store[2] = cosf( store[2] ); + store[3] = cosf( store[3] ); #endif } static void -emit_cos (struct x86_function *func, - unsigned xmm_dst) +emit_cos( + struct x86_function *func, + unsigned xmm_dst ) { - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_mov_reg_imm (func, - x86_make_reg (file_REG32, reg_AX), - ALIGN16((GLint) g_cos_storage)); - sse_movaps (func, - x86_deref (x86_make_reg (file_REG32, reg_AX)), - make_xmm (xmm_dst)); - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_push (func, - x86_make_reg (file_REG32, reg_BX)); - x86_push (func, - x86_make_reg (file_REG32, reg_CX)); - x86_push (func, - x86_make_reg (file_REG32, reg_DX)); - x86_call (func, - (GLubyte *) cos4f); - x86_pop (func, - x86_make_reg (file_REG32, reg_DX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_CX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_BX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); - sse_movaps (func, - make_xmm (xmm_dst), - x86_deref (x86_make_reg (file_REG32, reg_AX))); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); + emit_func_call1( + func, + xmm_dst, + ALIGN16( (unsigned) g_cos_storage ), + (unsigned char *) cos4f ); } -static GLfloat g_sin_storage[4 + 3]; +/* XXX: move into machine context */ +static float g_sin_storage[4 + 3]; static void -sin4f (void) +sin4f( void ) { - GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_sin_storage); + float *store = (float *) ALIGN16( (unsigned) g_sin_storage ); #ifdef WIN32 - store[0] = (GLfloat) sin ((GLdouble) store[0]); - store[1] = (GLfloat) sin ((GLdouble) store[1]); - store[2] = (GLfloat) sin ((GLdouble) store[2]); - store[3] = (GLfloat) sin ((GLdouble) store[3]); + store[0] = (float) sin( (double) store[0] ); + store[1] = (float) sin( (double) store[1] ); + store[2] = (float) sin( (double) store[2] ); + store[3] = (float) sin( (double) store[3] ); #else - store[0] = sin (store[0]); - store[1] = sin (store[1]); - store[2] = sin (store[2]); - store[3] = sin (store[3]); + store[0] = sinf( store[0] ); + store[1] = sinf( store[1] ); + store[2] = sinf( store[2] ); + store[3] = sinf( store[3] ); #endif } @@ -330,47 +440,23 @@ static void emit_sin (struct x86_function *func, unsigned xmm_dst) { - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_mov_reg_imm (func, - x86_make_reg (file_REG32, reg_AX), - ALIGN16((GLint) g_sin_storage)); - sse_movaps (func, - x86_deref (x86_make_reg (file_REG32, reg_AX)), - make_xmm (xmm_dst)); - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_push (func, - x86_make_reg (file_REG32, reg_BX)); - x86_push (func, - x86_make_reg (file_REG32, reg_CX)); - x86_push (func, - x86_make_reg (file_REG32, reg_DX)); - x86_call (func, - (GLubyte *) sin4f); - x86_pop (func, - x86_make_reg (file_REG32, reg_DX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_CX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_BX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); - sse_movaps (func, - make_xmm (xmm_dst), - x86_deref (x86_make_reg (file_REG32, reg_AX))); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); + emit_func_call1( + func, + xmm_dst, + ALIGN16( (unsigned) g_sin_storage ), + (unsigned char *) sin4f ); } static void -emit_mov (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) +emit_mov( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) { - sse_movups (func, - make_xmm (xmm_dst), - make_xmm (xmm_src)); + sse_movups( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); } static void @@ -378,1353 +464,1255 @@ emit_mul (struct x86_function *func, unsigned xmm_dst, unsigned xmm_src) { - sse_mulps (func, - make_xmm (xmm_dst), - make_xmm (xmm_src)); + sse_mulps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); } -static GLfloat g_pow_storage[4 + 4 + 3]; +/* XXX: move into machine context */ +static float g_pow_storage[4 + 4 + 3]; static void -pow4f (void) +pow4f( void ) { - GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_pow_storage); + float *store = (float *) ALIGN16( (unsigned) g_pow_storage ); #ifdef WIN32 - store[0] = (GLfloat) pow ((GLdouble) store[0], (GLdouble) store[4]); - store[1] = (GLfloat) pow ((GLdouble) store[1], (GLdouble) store[5]); - store[2] = (GLfloat) pow ((GLdouble) store[2], (GLdouble) store[6]); - store[3] = (GLfloat) pow ((GLdouble) store[3], (GLdouble) store[7]); + store[0] = (float) pow( (double) store[0], (double) store[4] ); + store[1] = (float) pow( (double) store[1], (double) store[5] ); + store[2] = (float) pow( (double) store[2], (double) store[6] ); + store[3] = (float) pow( (double) store[3], (double) store[7] ); #else - store[0] = powf (store[0], store[4]); - store[1] = powf (store[1], store[5]); - store[2] = powf (store[2], store[6]); - store[3] = powf (store[3], store[7]); + store[0] = powf( store[0], store[4] ); + store[1] = powf( store[1], store[5] ); + store[2] = powf( store[2], store[6] ); + store[3] = powf( store[3], store[7] ); #endif } static void -emit_pow (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) +emit_pow( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) { - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_mov_reg_imm (func, - x86_make_reg (file_REG32, reg_AX), - ALIGN16((GLint) g_pow_storage)); - sse_movaps (func, - x86_make_disp (x86_make_reg (file_REG32, reg_AX), 0), - make_xmm (xmm_dst)); - sse_movaps (func, - x86_make_disp (x86_make_reg (file_REG32, reg_AX), 16), - make_xmm (xmm_src)); - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_push (func, - x86_make_reg (file_REG32, reg_BX)); - x86_push (func, - x86_make_reg (file_REG32, reg_CX)); - x86_push (func, - x86_make_reg (file_REG32, reg_DX)); - x86_call (func, - (GLubyte *) pow4f); - x86_pop (func, - x86_make_reg (file_REG32, reg_DX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_CX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_BX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); - sse_movaps (func, - make_xmm (xmm_dst), - x86_deref (x86_make_reg (file_REG32, reg_AX))); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); + emit_func_call2( + func, + xmm_dst, + xmm_src, + ALIGN16( (unsigned) g_pow_storage ), + (unsigned char *) pow4f ); } -static GLfloat g_ex2_storage[4 + 3]; +/* XXX: move into machine context */ +static float g_ex2_storage[4 + 3]; static void -ex24f (void) +ex24f( void ) { - GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_ex2_storage); + float *store = (float *) ALIGN16( (unsigned) g_ex2_storage ); - store[0] = (GLfloat) pow (2.0, (GLdouble) store[0]); - store[1] = (GLfloat) pow (2.0, (GLdouble) store[1]); - store[2] = (GLfloat) pow (2.0, (GLdouble) store[2]); - store[3] = (GLfloat) pow (2.0, (GLdouble) store[3]); +#ifdef WIN32 + store[0] = (float) pow( 2.0, (double) store[0] ); + store[1] = (float) pow( 2.0, (double) store[1] ); + store[2] = (float) pow( 2.0, (double) store[2] ); + store[3] = (float) pow( 2.0, (double) store[3] ); +#else + store[0] = powf( 2.0f, store[0] ); + store[1] = powf( 2.0f, store[1] ); + store[2] = powf( 2.0f, store[2] ); + store[3] = powf( 2.0f, store[3] ); +#endif } static void -emit_ex2 (struct x86_function *func, - unsigned xmm_dst) +emit_ex2( + struct x86_function *func, + unsigned xmm_dst ) { - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_mov_reg_imm (func, - x86_make_reg (file_REG32, reg_AX), - ALIGN16((GLint) g_ex2_storage)); - sse_movaps (func, - x86_deref (x86_make_reg (file_REG32, reg_AX)), - make_xmm (xmm_dst)); - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_push (func, - x86_make_reg (file_REG32, reg_BX)); - x86_push (func, - x86_make_reg (file_REG32, reg_CX)); - x86_push (func, - x86_make_reg (file_REG32, reg_DX)); - x86_call (func, - (GLubyte *) ex24f); - x86_pop (func, - x86_make_reg (file_REG32, reg_DX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_CX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_BX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); - sse_movaps (func, - make_xmm (xmm_dst), - x86_deref (x86_make_reg (file_REG32, reg_AX))); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); + emit_func_call1( + func, + xmm_dst, + ALIGN16( (unsigned) g_ex2_storage ), + (unsigned char *) ex24f ); } -static GLfloat g_lg2_storage[4 + 3]; +/* XXX: move into machine context */ +static float g_lg2_storage[4 + 3]; static void -lg24f (void) +lg24f( void ) { - GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_lg2_storage); + float *store = (float *) ALIGN16( (unsigned) g_lg2_storage ); - store[0] = LOG2 (store[0]); - store[1] = LOG2 (store[1]); - store[2] = LOG2 (store[2]); - store[3] = LOG2 (store[3]); + store[0] = LOG2( store[0] ); + store[1] = LOG2( store[1] ); + store[2] = LOG2( store[2] ); + store[3] = LOG2( store[3] ); } static void -emit_lg2 (struct x86_function *func, - unsigned xmm_dst) +emit_lg2( + struct x86_function *func, + unsigned xmm_dst ) { - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_mov_reg_imm (func, - x86_make_reg (file_REG32, reg_AX), - ALIGN16((GLint) g_lg2_storage)); - sse_movaps (func, - x86_deref (x86_make_reg (file_REG32, reg_AX)), - make_xmm (xmm_dst)); - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_push (func, - x86_make_reg (file_REG32, reg_BX)); - x86_push (func, - x86_make_reg (file_REG32, reg_CX)); - x86_push (func, - x86_make_reg (file_REG32, reg_DX)); - x86_call (func, - (GLubyte *) lg24f); - x86_pop (func, - x86_make_reg (file_REG32, reg_DX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_CX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_BX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); - sse_movaps (func, - make_xmm (xmm_dst), - x86_deref (x86_make_reg (file_REG32, reg_AX))); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); + emit_func_call1( + func, + xmm_dst, + ALIGN16( (unsigned) g_lg2_storage ), + (unsigned char *) lg24f ); } -static GLfloat g_flr_storage[4 + 3]; +/* XXX: move into machine context */ +static float g_flr_storage[4 + 3]; static void -flr4f (void) +flr4f( void ) { - GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_flr_storage); + float *store = (float *) ALIGN16( (unsigned) g_flr_storage ); - store[0] = (GLfloat) floor ((GLdouble) store[0]); - store[1] = (GLfloat) floor ((GLdouble) store[1]); - store[2] = (GLfloat) floor ((GLdouble) store[2]); - store[3] = (GLfloat) floor ((GLdouble) store[3]); + store[0] = (float) floor( (double) store[0] ); + store[1] = (float) floor( (double) store[1] ); + store[2] = (float) floor( (double) store[2] ); + store[3] = (float) floor( (double) store[3] ); } static void -emit_flr (struct x86_function *func, - unsigned xmm_dst) +emit_flr( + struct x86_function *func, + unsigned xmm_dst ) { - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_mov_reg_imm (func, - x86_make_reg (file_REG32, reg_AX), - ALIGN16((GLint) g_flr_storage)); - sse_movaps (func, - x86_deref (x86_make_reg (file_REG32, reg_AX)), - make_xmm (xmm_dst)); - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_push (func, - x86_make_reg (file_REG32, reg_BX)); - x86_push (func, - x86_make_reg (file_REG32, reg_CX)); - x86_push (func, - x86_make_reg (file_REG32, reg_DX)); - x86_call (func, - (GLubyte *) flr4f); - x86_pop (func, - x86_make_reg (file_REG32, reg_DX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_CX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_BX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); - sse_movaps (func, - make_xmm (xmm_dst), - x86_deref (x86_make_reg (file_REG32, reg_AX))); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); + emit_func_call1( + func, + xmm_dst, + ALIGN16( (unsigned) g_flr_storage ), + (unsigned char *) flr4f ); } -static GLfloat g_frc_storage[4 + 3]; +/* XXX: move into machine context */ +static float g_frc_storage[4 + 3]; static void -frc4f (void) +frc4f( void ) { - GLfloat *store = (GLfloat *) ALIGN16((unsigned) g_frc_storage); + float *store = (float *) ALIGN16( (unsigned) g_frc_storage ); - store[0] -= (GLfloat) floor ((GLdouble) store[0]); - store[1] -= (GLfloat) floor ((GLdouble) store[1]); - store[2] -= (GLfloat) floor ((GLdouble) store[2]); - store[3] -= (GLfloat) floor ((GLdouble) store[3]); + store[0] -= (float) floor( (double) store[0] ); + store[1] -= (float) floor( (double) store[1] ); + store[2] -= (float) floor( (double) store[2] ); + store[3] -= (float) floor( (double) store[3] ); } static void -emit_frc (struct x86_function *func, - unsigned xmm_dst) +emit_frc( + struct x86_function *func, + unsigned xmm_dst ) { - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_mov_reg_imm (func, - x86_make_reg (file_REG32, reg_AX), - ALIGN16((GLint) g_frc_storage)); - sse_movaps (func, - x86_deref (x86_make_reg (file_REG32, reg_AX)), - make_xmm (xmm_dst)); - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_push (func, - x86_make_reg (file_REG32, reg_BX)); - x86_push (func, - x86_make_reg (file_REG32, reg_CX)); - x86_push (func, - x86_make_reg (file_REG32, reg_DX)); - x86_call (func, - (GLubyte *) frc4f); - x86_pop (func, - x86_make_reg (file_REG32, reg_DX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_CX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_BX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); - sse_movaps (func, - make_xmm (xmm_dst), - x86_deref (x86_make_reg (file_REG32, reg_AX))); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); + emit_func_call1( + func, + xmm_dst, + ALIGN16( (unsigned) g_frc_storage ), + (unsigned char *) frc4f ); } static void -emit_rcp (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) +emit_rcp ( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) { - sse2_rcpps (func, - make_xmm (xmm_dst), - make_xmm (xmm_src)); + sse2_rcpps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); } static void -emit_rsqrt (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) +emit_rsqrt( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) { - sse_rsqrtps (func, - make_xmm (xmm_dst), - make_xmm (xmm_src)); + sse_rsqrtps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); } static void -emit_sub (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) +emit_sub( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) { - sse_subps (func, - make_xmm (xmm_dst), - make_xmm (xmm_src)); + sse_subps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); } static void -emit_fetch (struct x86_function *func, - unsigned xmm, - const struct tgsi_full_src_register *reg, - const unsigned chan_index) +emit_fetch( + struct x86_function *func, + unsigned xmm, + const struct tgsi_full_src_register *reg, + const unsigned chan_index ) { - unsigned swizzle = tgsi_util_get_full_src_register_extswizzle (reg, chan_index); - - switch (swizzle) - { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - switch (reg->SrcRegister.File) - { - case TGSI_FILE_CONSTANT: - emit_const (func, xmm, reg->SrcRegister.Index, swizzle); - break; - - case TGSI_FILE_INPUT: - emit_input (func, xmm, reg->SrcRegister.Index, swizzle); - break; - - case TGSI_FILE_TEMPORARY: - emit_tempf (func, xmm, reg->SrcRegister.Index, swizzle); - break; - - default: - assert (0); - } - break; - - case TGSI_EXTSWIZZLE_ZERO: - emit_tempf (func, - xmm, - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C); - break; - - case TGSI_EXTSWIZZLE_ONE: - emit_tempf (func, - xmm, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C); - break; - - default: - assert (0); - } - - switch (tgsi_util_get_full_src_register_sign_mode (reg, chan_index)) - { - case TGSI_UTIL_SIGN_CLEAR: - emit_abs (func, xmm); - break; - - case TGSI_UTIL_SIGN_SET: - emit_setsign (func, xmm); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - emit_neg (func, xmm); - break; - - case TGSI_UTIL_SIGN_KEEP: - break; - } + unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( reg->SrcRegister.File ) { + case TGSI_FILE_CONSTANT: + emit_const( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + case TGSI_FILE_INPUT: + emit_input( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + case TGSI_FILE_TEMPORARY: + emit_tempf( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + emit_tempf( + func, + xmm, + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ); + break; + + case TGSI_EXTSWIZZLE_ONE: + emit_tempf( + func, + xmm, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + break; + + default: + assert( 0 ); + } + + switch( tgsi_util_get_full_src_register_sign_mode( reg, chan_index ) ) { + case TGSI_UTIL_SIGN_CLEAR: + emit_abs( func, xmm ); + break; + + case TGSI_UTIL_SIGN_SET: + emit_setsign( func, xmm ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + emit_neg( func, xmm ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } } static void -emit_store (struct x86_function *func, - unsigned xmm, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - unsigned chan_index) +emit_store( + struct x86_function *func, + unsigned xmm, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + unsigned chan_index ) { - switch (reg->DstRegister.File) - { - case TGSI_FILE_OUTPUT: - emit_output (func, xmm, reg->DstRegister.Index, chan_index); - break; + switch( reg->DstRegister.File ) { + case TGSI_FILE_OUTPUT: + emit_output( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; - case TGSI_FILE_TEMPORARY: - emit_temps (func, xmm, reg->DstRegister.Index, chan_index); - break; + case TGSI_FILE_TEMPORARY: + emit_temps( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; case TGSI_FILE_ADDRESS: - emit_addrs( func, xmm, reg->DstRegister.Index, chan_index ); + emit_addrs( + func, + xmm, + reg->DstRegister.Index, + chan_index ); break; - default: - assert (0); - } + default: + assert( 0 ); + } - switch (inst->Instruction.Saturate) - { - case TGSI_SAT_NONE: - break; + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; - case TGSI_SAT_ZERO_ONE: -// assert (0); - break; + case TGSI_SAT_ZERO_ONE: +// assert( 0 ); + break; - case TGSI_SAT_MINUS_PLUS_ONE: - assert (0); - break; - } + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + } } -#define FETCH(XMM,INDEX,CHAN)\ - emit_fetch (func, XMM, &inst->FullSrcRegisters[INDEX], CHAN) - -#define STORE(XMM,INDEX,CHAN)\ - emit_store (func, XMM, &inst->FullDstRegisters[INDEX], inst, CHAN) - static void -emit_kil (struct x86_function *func, - const struct tgsi_full_src_register *reg) +emit_kil( + struct x86_function *func, + const struct tgsi_full_src_register *reg ) { - unsigned uniquemask; - unsigned registers[4]; - unsigned nextregister = 0; - unsigned firstchan = ~0; - unsigned chan_index; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - for (chan_index = 0; chan_index < 4; chan_index++) - { - unsigned swizzle; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle (reg, chan_index); - - /* check if the component has not been already tested */ - if (!(uniquemask & (1 << swizzle))) - { - uniquemask |= 1 << swizzle; - - /* allocate register */ - registers[chan_index] = nextregister; - emit_fetch (func, nextregister, reg, chan_index); - nextregister++; - - /* mark the first channel used */ - if (firstchan == ~0) - firstchan = chan_index; - } - } - - x86_push (func, - x86_make_reg (file_REG32, reg_AX)); - x86_push (func, - x86_make_reg (file_REG32, reg_DX)); - - for (chan_index = 0; chan_index < 4; chan_index++) - { - if (uniquemask & (1 << chan_index)) - { - sse_cmpps (func, - make_xmm (registers[chan_index]), - get_temp (TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C), - cc_LessThan); - - if (chan_index == firstchan) - { - sse_pmovmskb (func, - x86_make_reg (file_REG32, reg_AX), - make_xmm (registers[chan_index])); - } - else - { - sse_pmovmskb (func, - x86_make_reg (file_REG32, reg_DX), - make_xmm (registers[chan_index])); - x86_or (func, - x86_make_reg (file_REG32, reg_AX), - x86_make_reg (file_REG32, reg_DX)); - } - } - } - - x86_or (func, - get_temp(TGSI_EXEC_TEMP_KILMASK_I, - TGSI_EXEC_TEMP_KILMASK_C), - x86_make_reg (file_REG32, reg_AX)); - - x86_pop (func, - x86_make_reg (file_REG32, reg_DX)); - x86_pop (func, - x86_make_reg (file_REG32, reg_AX)); -} + unsigned uniquemask; + unsigned registers[4]; + unsigned nextregister = 0; + unsigned firstchan = ~0; + unsigned chan_index; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + FOR_EACH_CHANNEL( chan_index ) { + unsigned swizzle; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle( + reg, + chan_index ); + + /* check if the component has not been already tested */ + if( !(uniquemask & (1 << swizzle)) ) { + uniquemask |= 1 << swizzle; + + /* allocate register */ + registers[chan_index] = nextregister; + emit_fetch( + func, + nextregister, + reg, + chan_index ); + nextregister++; + + /* mark the first channel used */ + if( firstchan == ~0 ) { + firstchan = chan_index; + } + } + } -static void -emit_setcc (struct x86_function *func, - struct tgsi_full_instruction *inst, - enum sse_cc cc) -{ - unsigned chan_index; + x86_push( + func, + x86_make_reg( file_REG32, reg_AX ) ); + x86_push( + func, + x86_make_reg( file_REG32, reg_DX ) ); + + FOR_EACH_CHANNEL( chan_index ) { + if( uniquemask & (1 << chan_index) ) { + sse_cmpps( + func, + make_xmm( registers[chan_index] ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ), + cc_LessThan ); + + if( chan_index == firstchan ) { + sse_pmovmskb( + func, + x86_make_reg( file_REG32, reg_AX ), + make_xmm( registers[chan_index] ) ); + } + else { + sse_pmovmskb( + func, + x86_make_reg( file_REG32, reg_DX ), + make_xmm( registers[chan_index] ) ); + x86_or( + func, + x86_make_reg( file_REG32, reg_AX ), + x86_make_reg( file_REG32, reg_DX ) ); + } + } + } + + x86_or( + func, + get_temp( + TGSI_EXEC_TEMP_KILMASK_I, + TGSI_EXEC_TEMP_KILMASK_C ), + x86_make_reg( file_REG32, reg_AX ) ); - FOR_EACH_ENABLED_CHANNEL - { - emit_fetch (func, 0, &inst->FullSrcRegisters[0], chan_index); - emit_fetch (func, 1, &inst->FullSrcRegisters[1], chan_index); + x86_pop( + func, + x86_make_reg( file_REG32, reg_DX ) ); + x86_pop( + func, + x86_make_reg( file_REG32, reg_AX ) ); +} - sse_cmpps (func, - make_xmm (0), - make_xmm (1), - cc); +#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\ + emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN ) - sse_andps (func, - make_xmm (0), - get_temp (TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C)); +#define STORE( FUNC, INST, XMM, INDEX, CHAN )\ + emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) - emit_store (func, 0, &inst->FullDstRegisters[0], inst, chan_index); - } +static void +emit_setcc( + struct x86_function *func, + struct tgsi_full_instruction *inst, + enum sse_cc cc ) +{ + unsigned chan_index; + + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + sse_cmpps( + func, + make_xmm( 0 ), + make_xmm( 1 ), + cc ); + sse_andps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ) ); + STORE( func, *inst, 0, 0, chan_index ); + } } static void -emit_cmp (struct x86_function *func, - struct tgsi_full_instruction *inst) +emit_cmp( + struct x86_function *func, + struct tgsi_full_instruction *inst ) { - unsigned chan_index; - - FOR_EACH_ENABLED_CHANNEL - { - emit_fetch (func, 0, &inst->FullSrcRegisters[0], chan_index); - emit_fetch (func, 1, &inst->FullSrcRegisters[1], chan_index); - emit_fetch (func, 2, &inst->FullSrcRegisters[2], chan_index); - - sse_cmpps (func, - make_xmm (0), - get_temp (TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C), - cc_LessThan); - - sse_andps (func, - make_xmm (1), - make_xmm (0)); - sse_andnps (func, - make_xmm (0), - make_xmm (2)); - sse_orps (func, - make_xmm (0), - make_xmm (1)); - - emit_store (func, 0, &inst->FullDstRegisters[0], inst, chan_index); - } + unsigned chan_index; + + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + sse_cmpps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ), + cc_LessThan ); + sse_andps( + func, + make_xmm( 1 ), + make_xmm( 0 ) ); + sse_andnps( + func, + make_xmm( 0 ), + make_xmm( 2 ) ); + sse_orps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } } static void -emit_f2it( struct x86_function *func, - unsigned xmm ) +emit_f2it( + struct x86_function *func, + unsigned xmm ) { - sse2_cvttps2dq( func, make_xmm( xmm ), make_xmm( xmm ) ); + sse2_cvttps2dq( + func, + make_xmm( xmm ), + make_xmm( xmm ) ); } static void -emit_instruction (struct x86_function *func, - struct tgsi_full_instruction *inst) +emit_instruction( + struct x86_function *func, + struct tgsi_full_instruction *inst ) { - unsigned chan_index; + unsigned chan_index; - switch (inst->Instruction.Opcode) { + switch( inst->Instruction.Opcode ) { case TGSI_OPCODE_ARL: - FOR_EACH_ENABLED_CHANNEL { - FETCH( 0, 0, chan_index ); - emit_f2it( func, 0 ); - STORE( 0, 0, chan_index ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_f2it( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C); + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + STORE( func, *inst, 0, 0, CHAN_X ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( func, *inst, 0, 0, CHAN_W ); + } + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( func, *inst, 0, 0, CHAN_X ); + sse_maxps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ) ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 1, 0, CHAN_Y ); + sse_maxps( + func, + make_xmm( 1 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ) ); + FETCH( func, *inst, 2, 0, CHAN_W ); + sse_minps( + func, + make_xmm( 2 ), + get_temp( + TGSI_EXEC_TEMP_128_I, + TGSI_EXEC_TEMP_128_C ) ); + sse_maxps( + func, + make_xmm( 2 ), + get_temp( + TGSI_EXEC_TEMP_MINUS_128_I, + TGSI_EXEC_TEMP_MINUS_128_C ) ); + emit_pow( func, 1, 2 ); + FETCH( func, *inst, 0, 0, CHAN_X ); + sse_xorps( + func, + make_xmm( 2 ), + make_xmm( 2 ) ); + sse_cmpps( + func, + make_xmm( 2 ), + make_xmm( 0 ), + cc_LessThanEqual ); + sse_andps( + func, + make_xmm( 2 ), + make_xmm( 1 ) ); + STORE( func, *inst, 2, 0, CHAN_Z ); + } + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_rcp( func, 0, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_rsqrt( func, 0, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + assert( 0 ); + break; + + case TGSI_OPCODE_LOG: + assert( 0 ); + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_mul( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_add( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul(func, 1, 2 ); + emit_add(func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_W ); + FETCH( func, *inst, 2, 1, CHAN_W ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + FETCH( func, *inst, 1, 1, CHAN_Y ); + emit_mul( func, 0, 1 ); + STORE( func, *inst, 0, 0, CHAN_Y ); } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + FETCH( func, *inst, 0, 0, CHAN_Z ); + STORE( func, *inst, 0, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, 1, CHAN_W ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + sse_minps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + sse_maxps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + emit_setcc( func, inst, cc_LessThan ); + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + emit_setcc( func, inst, cc_NotLessThan ); + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_mul( func, 0, 1 ); + emit_add( func, 0, 2 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_sub( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_sub( func, 1, 2 ); + emit_mul( func, 0, 1 ); + emit_add( func, 0, 2 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CND: + assert( 0 ); + break; + + case TGSI_OPCODE_CND0: + assert( 0 ); + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert( 0 ); + break; + + case TGSI_OPCODE_INDEX: + assert( 0 ); + break; + + case TGSI_OPCODE_NEGATE: + assert( 0 ); + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_frc( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + assert( 0 ); + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_flr( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + assert( 0 ); + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_ex2( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_lg2( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_pow( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( func, *inst, 1, 1, CHAN_Z ); + FETCH( func, *inst, 3, 0, CHAN_Z ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + FETCH( func, *inst, 4, 1, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + emit_mov( func, 2, 0 ); + emit_mul( func, 2, 1 ); + emit_mov( func, 5, 3 ); + emit_mul( func, 5, 4 ); + emit_sub( func, 2, 5 ); + STORE( func, *inst, 2, 0, CHAN_X ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 2, 1, CHAN_X ); + FETCH( func, *inst, 5, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + emit_mul( func, 3, 2 ); + emit_mul( func, 1, 5 ); + emit_sub( func, 3, 1 ); + STORE( func, *inst, 3, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + emit_mul( func, 5, 4 ); + emit_mul( func, 0, 2 ); + emit_sub( func, 5, 0 ); + STORE( func, *inst, 5, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + assert( 0 ); + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_abs( func, 0) ; + + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_RCC: + assert( 0 ); + break; + + case TGSI_OPCODE_DPH: + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 1, CHAN_W ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_cos( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + assert( 0 ); + break; + + case TGSI_OPCODE_DDY: + assert( 0 ); + break; + + case TGSI_OPCODE_KIL: + emit_kil( func, &inst->FullSrcRegisters[0] ); + break; + + case TGSI_OPCODE_PK2H: + assert( 0 ); + break; + + case TGSI_OPCODE_PK2US: + assert( 0 ); + break; + + case TGSI_OPCODE_PK4B: + assert( 0 ); + break; + + case TGSI_OPCODE_PK4UB: + assert( 0 ); + break; + + case TGSI_OPCODE_RFL: + assert( 0 ); + break; + + case TGSI_OPCODE_SEQ: + assert( 0 ); + break; + + case TGSI_OPCODE_SFL: + assert( 0 ); + break; + + case TGSI_OPCODE_SGT: + assert( 0 ); + break; + + case TGSI_OPCODE_SIN: + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_sin( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + assert( 0 ); + break; + + case TGSI_OPCODE_SNE: + assert( 0 ); + break; + + case TGSI_OPCODE_STR: + assert( 0 ); + break; + + case TGSI_OPCODE_TEX: + assert( 0 ); + break; + + case TGSI_OPCODE_TXD: + assert( 0 ); + break; + + case TGSI_OPCODE_UP2H: + assert( 0 ); + break; + + case TGSI_OPCODE_UP2US: + assert( 0 ); + break; + + case TGSI_OPCODE_UP4B: + assert( 0 ); break; - case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_LIT: - if (IS_CHANNEL_ENABLED(CHAN_X) || IS_CHANNEL_ENABLED(CHAN_W)) - { - emit_tempf (func, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C); - if (IS_CHANNEL_ENABLED(CHAN_X)) - STORE(0, 0, CHAN_X); - if (IS_CHANNEL_ENABLED(CHAN_W)) - STORE(0, 0, CHAN_W); - } - if (IS_CHANNEL_ENABLED(CHAN_Y) || IS_CHANNEL_ENABLED(CHAN_Z)) - { - if (IS_CHANNEL_ENABLED(CHAN_Y)) - { - FETCH(0, 0, CHAN_X); - sse_maxps (func, - make_xmm (0), - get_temp (TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C)); - STORE(0, 0, CHAN_Y); - } - if (IS_CHANNEL_ENABLED(CHAN_Z)) - { - FETCH(1, 0, CHAN_Y); - sse_maxps (func, - make_xmm (1), - get_temp (TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C)); - - FETCH(2, 0, CHAN_W); - sse_minps (func, - make_xmm (2), - get_temp (TGSI_EXEC_TEMP_128_I, - TGSI_EXEC_TEMP_128_C)); - sse_maxps (func, - make_xmm (2), - get_temp (TGSI_EXEC_TEMP_MINUS_128_I, - TGSI_EXEC_TEMP_MINUS_128_C)); - - emit_pow (func, 1, 2); - - FETCH(0, 0, CHAN_X); - sse_xorps (func, make_xmm (2), make_xmm (2)); - sse_cmpps (func, - make_xmm (2), - make_xmm (0), - cc_LessThanEqual); - sse_andps (func, - make_xmm (2), - make_xmm (1)); - - emit_store (func, 2, &inst->FullDstRegisters[0], inst, CHAN_Z); - } - } - break; - - case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH(0, 0, CHAN_X); - emit_rcp (func, 0, 0); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH(0, 0, CHAN_X); - emit_rsqrt (func, 0, 0); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_EXP: - assert (0); - break; - - case TGSI_OPCODE_LOG: - assert (0); - break; - - case TGSI_OPCODE_MUL: - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - FETCH(1, 1, chan_index); - emit_mul (func, 0, 1); - - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_ADD: - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - FETCH(1, 1, chan_index); - emit_add (func, 0, 1); - - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH(0, 0, CHAN_X); - FETCH(1, 1, CHAN_X); - emit_mul (func, 0, 1); - - FETCH(1, 0, CHAN_Y); - FETCH(2, 1, CHAN_Y); - emit_mul (func, 1, 2); - emit_add (func, 0, 1); - - FETCH(1, 0, CHAN_Z); - FETCH(2, 1, CHAN_Z); - emit_mul (func, 1, 2); - emit_add (func, 0, 1); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH(0, 0, CHAN_X); - FETCH(1, 1, CHAN_X); - emit_mul (func, 0, 1); - - FETCH(1, 0, CHAN_Y); - FETCH(2, 1, CHAN_Y); - emit_mul (func, 1, 2); - emit_add (func, 0, 1); - - FETCH(1, 0, CHAN_Z); - FETCH(2, 1, CHAN_Z); - emit_mul (func, 1, 2); - emit_add (func, 0, 1); - - FETCH(1, 0, CHAN_W); - FETCH(2, 1, CHAN_W); - emit_mul (func, 1, 2); - emit_add (func, 0, 1); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_DST: - IF_IS_CHANNEL_ENABLED(CHAN_X) - { - emit_tempf (func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C); - STORE(0, 0, CHAN_X); - } - IF_IS_CHANNEL_ENABLED(CHAN_Y) - { - FETCH(0, 0, CHAN_Y); - FETCH(1, 1, CHAN_Y); - emit_mul (func, 0, 1); - STORE(0, 0, CHAN_Y); - } - IF_IS_CHANNEL_ENABLED(CHAN_Z) - { - FETCH(0, 0, CHAN_Z); - STORE(0, 0, CHAN_Z); - } - IF_IS_CHANNEL_ENABLED(CHAN_W) - { - FETCH(0, 1, CHAN_W); - STORE(0, 0, CHAN_W); - } - break; - - case TGSI_OPCODE_MIN: - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - FETCH(1, 1, chan_index); - sse_minps (func, - make_xmm (0), - make_xmm (1)); - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_MAX: - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - FETCH(1, 1, chan_index); - sse_maxps (func, - make_xmm (0), - make_xmm (1)); - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - emit_setcc (func, inst, cc_LessThan); - break; - - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - emit_setcc (func, inst, cc_NotLessThan); - break; - - case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - FETCH(1, 1, chan_index); - FETCH(2, 2, chan_index); - emit_mul (func, 0, 1); - emit_add (func, 0, 2); - - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_SUB: - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - FETCH(1, 1, chan_index); - emit_sub (func, 0, 1); - - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - FETCH(1, 1, chan_index); - FETCH(2, 2, chan_index); - emit_sub (func, 1, 2); - emit_mul (func, 0, 1); - emit_add (func, 0, 2); - - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_CND: - assert (0); - break; - - case TGSI_OPCODE_CND0: - assert (0); - break; - - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - assert (0); - break; - - case TGSI_OPCODE_INDEX: - assert (0); - break; - - case TGSI_OPCODE_NEGATE: - assert (0); - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - emit_frc (func, 0); - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_CLAMP: - assert (0); - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - emit_flr (func, 0); - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_ROUND: - assert (0); - break; - - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ - FETCH(0, 0, CHAN_X); - emit_ex2 (func, 0); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ - FETCH(0, 0, CHAN_X); - emit_lg2 (func, 0); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH(0, 0, CHAN_X); - FETCH(1, 1, CHAN_X); - emit_pow (func, 0, 1); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - if (IS_CHANNEL_ENABLED(CHAN_X) || IS_CHANNEL_ENABLED(CHAN_Y)) - { - FETCH(1, 1, CHAN_Z); - FETCH(3, 0, CHAN_Z); - } - if (IS_CHANNEL_ENABLED(CHAN_X) || IS_CHANNEL_ENABLED(CHAN_Z)) - { - FETCH(0, 0, CHAN_Y); - FETCH(4, 1, CHAN_Y); - } - IF_IS_CHANNEL_ENABLED(CHAN_X) - { - emit_mov (func, 2, 0); - emit_mul (func, 2, 1); - emit_mov (func, 5, 3); - emit_mul (func, 5, 4); - emit_sub (func, 2, 5); - STORE(2, 0, CHAN_X); - } - - if (IS_CHANNEL_ENABLED(CHAN_Y) || IS_CHANNEL_ENABLED(CHAN_Z)) - { - FETCH(2, 1, CHAN_X); - FETCH(5, 0, CHAN_X); - } - IF_IS_CHANNEL_ENABLED(CHAN_Y) - { - emit_mul (func, 3, 2); - emit_mul (func, 1, 5); - emit_sub (func, 3, 1); - STORE(3, 0, CHAN_Y); - } - - IF_IS_CHANNEL_ENABLED(CHAN_Z) - { - emit_mul (func, 5, 4); - emit_mul (func, 0, 2); - emit_sub (func, 5, 0); - STORE(5, 0, CHAN_Z); - } - - IF_IS_CHANNEL_ENABLED(CHAN_W) - { - FETCH(0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C); - STORE(0, 0, CHAN_W); - } - break; - - case TGSI_OPCODE_MULTIPLYMATRIX: - assert (0); - break; - - case TGSI_OPCODE_ABS: - FOR_EACH_ENABLED_CHANNEL - { - FETCH(0, 0, chan_index); - emit_abs (func, 0); - - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_RCC: - assert (0); - break; - - case TGSI_OPCODE_DPH: - FETCH(0, 0, CHAN_X); - FETCH(1, 1, CHAN_X); - emit_mul (func, 0, 1); - - FETCH(1, 0, CHAN_Y); - FETCH(2, 1, CHAN_Y); - emit_mul (func, 1, 2); - emit_add (func, 0, 1); - - FETCH(1, 0, CHAN_Z); - FETCH(2, 1, CHAN_Z); - emit_mul (func, 1, 2); - emit_add (func, 0, 1); - - FETCH(1, 1, CHAN_W); - emit_add (func, 0, 1); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_COS: - FETCH(0, 0, CHAN_X); - emit_cos (func, 0); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_DDX: - assert (0); - break; - - case TGSI_OPCODE_DDY: - assert (0); - break; - - case TGSI_OPCODE_KIL: - emit_kil (func, &inst->FullSrcRegisters[0]); - break; - - case TGSI_OPCODE_PK2H: - assert (0); - break; - - case TGSI_OPCODE_PK2US: - assert (0); - break; - - case TGSI_OPCODE_PK4B: - assert (0); - break; - - case TGSI_OPCODE_PK4UB: - assert (0); - break; - - case TGSI_OPCODE_RFL: - assert (0); - break; - - case TGSI_OPCODE_SEQ: - assert (0); - break; - - case TGSI_OPCODE_SFL: - assert (0); - break; - - case TGSI_OPCODE_SGT: - assert (0); - break; - - case TGSI_OPCODE_SIN: - FETCH(0, 0, CHAN_X); - emit_sin (func, 0); - - FOR_EACH_ENABLED_CHANNEL - { - STORE(0, 0, chan_index); - } - break; - - case TGSI_OPCODE_SLE: - assert (0); - break; - - case TGSI_OPCODE_SNE: - assert (0); - break; - - case TGSI_OPCODE_STR: - assert (0); - break; - - case TGSI_OPCODE_TEX: - assert (0); - break; - - case TGSI_OPCODE_TXD: - assert (0); - break; - - case TGSI_OPCODE_UP2H: - assert (0); - break; - - case TGSI_OPCODE_UP2US: - assert (0); - break; - - case TGSI_OPCODE_UP4B: - assert (0); - break; - - case TGSI_OPCODE_UP4UB: - assert (0); - break; - - case TGSI_OPCODE_X2D: - assert (0); - break; - - case TGSI_OPCODE_ARA: - assert (0); - break; - - case TGSI_OPCODE_ARR: - assert (0); - break; - - case TGSI_OPCODE_BRA: - assert (0); - break; - - case TGSI_OPCODE_CAL: - assert (0); - break; - - case TGSI_OPCODE_RET: - assert (0); - break; - - case TGSI_OPCODE_SSG: - assert (0); - break; - - case TGSI_OPCODE_CMP: - emit_cmp (func, inst); - break; - - case TGSI_OPCODE_SCS: - IF_IS_CHANNEL_ENABLED(CHAN_X) - { - FETCH(0, 0, CHAN_X); - emit_cos (func, 0); - STORE(0, 0, CHAN_X); - } - - IF_IS_CHANNEL_ENABLED(CHAN_Y) - { - FETCH(0, 0, CHAN_Y); - emit_sin (func, 0); - STORE(0, 0, CHAN_Y); - } - - IF_IS_CHANNEL_ENABLED(CHAN_Z) - { - FETCH(0, TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C); - STORE(0, 0, CHAN_Z); - } - - IF_IS_CHANNEL_ENABLED(CHAN_W) - { - FETCH(0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C); - STORE(0, 0, CHAN_W); - } - break; - - case TGSI_OPCODE_TXB: - assert (0); - break; - - case TGSI_OPCODE_NRM: - assert (0); - break; - - case TGSI_OPCODE_DIV: - assert (0); - break; - - case TGSI_OPCODE_DP2: - assert (0); - break; - - case TGSI_OPCODE_TXL: - assert (0); - break; - - case TGSI_OPCODE_BRK: - assert (0); - break; - - case TGSI_OPCODE_IF: - assert (0); - break; - - case TGSI_OPCODE_LOOP: - assert (0); - break; - - case TGSI_OPCODE_REP: - assert (0); - break; - - case TGSI_OPCODE_ELSE: - assert (0); - break; - - case TGSI_OPCODE_ENDIF: - assert (0); - break; - - case TGSI_OPCODE_ENDLOOP: - assert (0); - break; - - case TGSI_OPCODE_ENDREP: - assert (0); - break; - - case TGSI_OPCODE_PUSHA: - assert (0); - break; - - case TGSI_OPCODE_POPA: - assert (0); - break; - - case TGSI_OPCODE_CEIL: - assert (0); - break; - - case TGSI_OPCODE_I2F: - assert (0); - break; - - case TGSI_OPCODE_NOT: - assert (0); - break; - - case TGSI_OPCODE_TRUNC: - assert (0); - break; - - case TGSI_OPCODE_SHL: - assert (0); - break; - - case TGSI_OPCODE_SHR: - assert (0); - break; - - case TGSI_OPCODE_AND: - assert (0); - break; - - case TGSI_OPCODE_OR: - assert (0); - break; - - case TGSI_OPCODE_MOD: - assert (0); - break; - - case TGSI_OPCODE_XOR: - assert (0); - break; + case TGSI_OPCODE_UP4UB: + assert( 0 ); + break; - case TGSI_OPCODE_SAD: - assert (0); - break; + case TGSI_OPCODE_X2D: + assert( 0 ); + break; - case TGSI_OPCODE_TXF: - assert (0); - break; + case TGSI_OPCODE_ARA: + assert( 0 ); + break; - case TGSI_OPCODE_TXQ: - assert (0); - break; + case TGSI_OPCODE_ARR: + assert( 0 ); + break; - case TGSI_OPCODE_CONT: - assert (0); - break; + case TGSI_OPCODE_BRA: + assert( 0 ); + break; - case TGSI_OPCODE_EMIT: - assert (0); - break; + case TGSI_OPCODE_CAL: + assert( 0 ); + break; - case TGSI_OPCODE_ENDPRIM: - assert (0); - break; + case TGSI_OPCODE_RET: + assert( 0 ); + break; - default: - assert (0); - } + case TGSI_OPCODE_SSG: + assert( 0 ); + break; + + case TGSI_OPCODE_CMP: + emit_cmp (func, inst); + break; + + case TGSI_OPCODE_SCS: + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_cos( func, 0 ); + STORE( func, *inst, 0, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + emit_sin( func, 0 ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + FETCH( func, *inst, 0, TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ); + STORE( func, *inst, 0, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_TXB: + assert( 0 ); + break; + + case TGSI_OPCODE_NRM: + assert( 0 ); + break; + + case TGSI_OPCODE_DIV: + assert( 0 ); + break; + + case TGSI_OPCODE_DP2: + assert( 0 ); + break; + + case TGSI_OPCODE_TXL: + assert( 0 ); + break; + + case TGSI_OPCODE_BRK: + assert( 0 ); + break; + + case TGSI_OPCODE_IF: + assert( 0 ); + break; + + case TGSI_OPCODE_LOOP: + assert( 0 ); + break; + + case TGSI_OPCODE_REP: + assert( 0 ); + break; + + case TGSI_OPCODE_ELSE: + assert( 0 ); + break; + + case TGSI_OPCODE_ENDIF: + assert( 0 ); + break; + + case TGSI_OPCODE_ENDLOOP: + assert( 0 ); + break; + + case TGSI_OPCODE_ENDREP: + assert( 0 ); + break; + + case TGSI_OPCODE_PUSHA: + assert( 0 ); + break; + + case TGSI_OPCODE_POPA: + assert( 0 ); + break; + + case TGSI_OPCODE_CEIL: + assert( 0 ); + break; + + case TGSI_OPCODE_I2F: + assert( 0 ); + break; + + case TGSI_OPCODE_NOT: + assert( 0 ); + break; + + case TGSI_OPCODE_TRUNC: + assert( 0 ); + break; + + case TGSI_OPCODE_SHL: + assert( 0 ); + break; + + case TGSI_OPCODE_SHR: + assert( 0 ); + break; + + case TGSI_OPCODE_AND: + assert( 0 ); + break; + + case TGSI_OPCODE_OR: + assert( 0 ); + break; + + case TGSI_OPCODE_MOD: + assert( 0 ); + break; + + case TGSI_OPCODE_XOR: + assert( 0 ); + break; + + case TGSI_OPCODE_SAD: + assert( 0 ); + break; + + case TGSI_OPCODE_TXF: + assert( 0 ); + break; + + case TGSI_OPCODE_TXQ: + assert( 0 ); + break; + + case TGSI_OPCODE_CONT: + assert( 0 ); + break; + + case TGSI_OPCODE_EMIT: + assert( 0 ); + break; + + case TGSI_OPCODE_ENDPRIM: + assert( 0 ); + break; + + default: + assert( 0 ); + } } -GLboolean -tgsi_emit_sse (struct tgsi_token *tokens, - struct x86_function *function) +unsigned +tgsi_emit_sse( + struct tgsi_token *tokens, + struct x86_function *func ) { - struct tgsi_parse_context parse; + struct tgsi_parse_context parse; - x86_init_func (function); + x86_init_func( func ); - x86_mov (function, get_input_base (), get_argument (0)); - x86_mov (function, get_output_base (), get_argument (1)); - x86_mov (function, get_const_base (), get_argument (2)); - x86_mov (function, get_temp_base (), get_argument (3)); + x86_mov( + func, + get_input_base(), + get_argument( 0 ) ); + x86_mov( + func, + get_output_base(), + get_argument( 1 ) ); + x86_mov( + func, + get_const_base(), + get_argument( 2 ) ); + x86_mov( + func, + get_temp_base(), + get_argument( 3 ) ); - tgsi_parse_init (&parse, tokens); + tgsi_parse_init( &parse, tokens ); - while (!tgsi_parse_end_of_tokens (&parse)) - { - tgsi_parse_token (&parse); + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); - switch (parse.FullToken.Token.Type) - { - case TGSI_TOKEN_TYPE_DECLARATION: - break; + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + break; - case TGSI_TOKEN_TYPE_INSTRUCTION: - emit_instruction (function, &parse.FullToken.FullInstruction); - break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + emit_instruction( + func, + &parse.FullToken.FullInstruction ); + break; - default: - assert (0); - } - } + default: + assert( 0 ); + } + } - tgsi_parse_free (&parse); + tgsi_parse_free( &parse ); #ifdef WIN32 - x86_retw (function, 16); + x86_retw( func, 16 ); #else - x86_ret (function); + x86_ret( func ); #endif - return GL_FALSE; + return 1; } #endif diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index 3ea37bb5e7..f5b0ccdb9b 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -502,6 +502,14 @@ void sse_addss( struct x86_function *p, emit_modrm( p, dst, src ); } +void sse_andnps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x55); + emit_modrm( p, dst, src ); +} + void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) @@ -510,6 +518,13 @@ void sse_andps( struct x86_function *p, emit_modrm( p, dst, src ); } +void sse_rsqrtps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x52); + emit_modrm( p, dst, src ); +} void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, @@ -538,6 +553,21 @@ void sse_movlhps( struct x86_function *p, emit_modrm( p, dst, src ); } +void sse_orps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x56); + emit_modrm( p, dst, src ); +} + +void sse_xorps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x57); + emit_modrm( p, dst, src ); +} void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, @@ -576,6 +606,14 @@ void sse_cmpps( struct x86_function *p, emit_1ub(p, cc); } +void sse_pmovmskb( struct x86_function *p, + struct x86_reg dest, + struct x86_reg src) +{ + emit_3ub(p, 0x66, X86_TWOB, 0xD7); + emit_modrm(p, dest, src); +} + /*********************************************************************** * SSE2 instructions */ @@ -593,6 +631,14 @@ void sse2_pshufd( struct x86_function *p, emit_1ub(p, shuf); } +void sse2_cvttps2dq( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub( p, 0xF3, X86_TWOB, 0x5B ); + emit_modrm( p, dst, src ); +} + void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) @@ -625,6 +671,14 @@ void sse2_packuswb( struct x86_function *p, emit_modrm( p, dst, src ); } +void sse2_rcpps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x53); + emit_modrm( p, dst, src ); +} + void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h index 66fb852ac9..c6236395b2 100644 --- a/src/mesa/x86/rtasm/x86sse.h +++ b/src/mesa/x86/rtasm/x86sse.h @@ -142,17 +142,20 @@ void mmx_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg sr void mmx_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_cvttps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf ); +void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_addps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_addss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, GLubyte cc ); void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -167,9 +170,13 @@ void sse_movss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void sse_movups( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_mulps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_mulss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_orps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf ); +void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -- cgit v1.2.3 From 4fd7bc00f06a98e1db2ad886a13566f19895e3c0 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 24 Aug 2007 17:51:12 +0100 Subject: Hook-up SSE2 to VS. --- src/mesa/pipe/draw/draw_prim.c | 31 ++++++++++++++++++++++++++++++- src/mesa/pipe/p_state.h | 1 + src/mesa/pipe/tgsi/exec/tgsi_exec.c | 29 +---------------------------- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 7 ------- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 4 ++-- src/mesa/state_tracker/st_atom_vs.c | 12 +++++++++++- src/mesa/state_tracker/st_cb_program.c | 14 ++++++++++++++ src/mesa/state_tracker/st_program.h | 6 ++++++ 8 files changed, 65 insertions(+), 39 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 2d613754e8..09616cf315 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -113,6 +113,21 @@ fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) } } +#if !defined(XSTDCALL) +#if defined(WIN32) +#define XSTDCALL __stdcall +#else +#define XSTDCALL +#endif +#endif + +#if defined(USE_X86_ASM) || defined(SLANG_X86) +typedef void (XSTDCALL *sse2_function)( + const struct tgsi_exec_vector *input, + struct tgsi_exec_vector *output, + float (*constant)[4], + struct tgsi_exec_vector *temporary ); +#endif /** * Transform vertices with the current vertex program/shader @@ -224,7 +239,21 @@ run_vertex_program(struct draw_context *draw, #endif /* run shader */ - tgsi_exec_machine_run( &machine ); + if( draw->vertex_shader.executable != NULL ) { +#if defined(USE_X86_ASM) || defined(SLANG_X86) + sse2_function func = (sse2_function) draw->vertex_shader.executable; + func( + machine.Inputs, + machine.Outputs, + machine.Consts, + machine.Temps ); +#else + assert( 0 ); +#endif + } + else { + tgsi_exec_machine_run( &machine ); + } #if 0 for (i = 0; i < 4; i++) { diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index e562a8d058..25e5861e1e 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -128,6 +128,7 @@ struct pipe_shader_state { unsigned inputs_read; /**< TGSI_ATTRIB_ bits */ unsigned outputs_written; /**< TGSI_ATTRIB_ bits */ const struct tgsi_token *tokens; + void *executable; }; struct pipe_depth_state diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index a1f46712b4..793f8bc0f6 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -85,11 +85,6 @@ tgsi_exec_machine_init( mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; -#if XXX_SSE - tgsi_emit_sse (tokens, - &mach->Function); -#endif - /* Setup constants. */ for( i = 0; i < 4; i++ ) { mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; @@ -967,6 +962,7 @@ store_dest( default: assert( 0 ); + return; } switch (inst->Instruction.Saturate) @@ -2226,20 +2222,6 @@ exec_instruction( } } - -#if !defined(XSTDCALL) -#if defined(WIN32) -#define XSTDCALL __stdcall -#else -#define XSTDCALL -#endif -#endif - -typedef void (XSTDCALL *fp_function) (const struct tgsi_exec_vector *input, - struct tgsi_exec_vector *output, - GLfloat (*constant)[4], - struct tgsi_exec_vector *temporary); - void tgsi_exec_machine_run2( struct tgsi_exec_machine *mach, @@ -2251,16 +2233,7 @@ tgsi_exec_machine_run2( #endif #if XXX_SSE - fp_function function; - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; - - function = (fp_function) x86_get_func (&mach->Function); - - function (mach->Inputs, - mach->Outputs, - mach->Consts, - mach->Temps); #else struct tgsi_parse_context parse; GLuint k; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index e5e8c3608e..d50f786029 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -3,10 +3,6 @@ #include "pipe/p_compiler.h" -#if 0 -#include "x86/rtasm/x86sse.h" -#endif - #if defined __cplusplus extern "C" { #endif // defined __cplusplus @@ -157,9 +153,6 @@ struct tgsi_exec_machine const struct tgsi_interp_coef *InterpCoefs; struct tgsi_exec_cond_stack CondStack; -#if XXX_SSE - struct x86_function Function; -#endif }; void diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 359775fdfb..d89bb19970 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -1659,13 +1659,13 @@ emit_instruction( } unsigned -tgsi_emit_sse( +tgsi_emit_sse2( struct tgsi_token *tokens, struct x86_function *func ) { struct tgsi_parse_context parse; - x86_init_func( func ); + func->csr = func->store; x86_mov( func, diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 7420a49520..a326c12608 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -38,7 +38,7 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" -#include "pipe/tgsi/exec/tgsi_dump.h" +#include "pipe/tgsi/exec/tgsi_core.h" #include "st_context.h" #include "st_atom.h" @@ -56,6 +56,12 @@ static void compile_vs( struct st_context *st, if (TGSI_DEBUG) tgsi_dump( vs->tokens, TGSI_DUMP_VERBOSE ); + +#if defined(USE_X86_ASM) || defined(SLANG_X86) + tgsi_emit_sse2( + vs->tokens, + &vs->sse2_program ); +#endif } @@ -121,6 +127,10 @@ static void update_vs( struct st_context *st ) = tgsi_mesa_translate_vertex_output_mask(vp->Base.Base.OutputsWritten); vs.tokens = &vp->tokens[0]; +#if defined(USE_X86_ASM) || defined(SLANG_X86) + vs.executable = (void *) x86_get_func( &vp->sse2_program ); +#endif + if (memcmp(&vs, &st->state.vs, sizeof(vs)) != 0 || vp->dirty) { diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index d0344efa0c..25da720d86 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -77,6 +77,10 @@ static struct gl_program *st_new_program( GLcontext *ctx, prog->id = program_id++; prog->dirty = 1; +#if defined(USE_X86_ASM) || defined(SLANG_X86) + x86_init_func( &prog->sse2_program ); +#endif + return _mesa_init_vertex_program( ctx, &prog->Base, target, @@ -105,6 +109,16 @@ static struct gl_program *st_new_program( GLcontext *ctx, static void st_delete_program( GLcontext *ctx, struct gl_program *prog ) { + switch( prog->Target ) { + case GL_VERTEX_PROGRAM_ARB: + { + struct st_vertex_program *p = (struct st_vertex_program *) prog; + + x86_release_func( &p->sse2_program ); + break; + } + + } _mesa_delete_program( ctx, prog ); } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index e7eb7a8e3f..883953399c 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -36,6 +36,7 @@ #include "mtypes.h" #include "pipe/tgsi/exec/tgsi_token.h" +#include "x86/rtasm/x86sse.h" #define ST_FP_MAX_TOKENS 1024 @@ -83,6 +84,11 @@ struct st_vertex_program struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; + +#if defined(USE_X86_ASM) || defined(SLANG_X86) + struct x86_function sse2_program; +#endif + #if 0 struct pipe_constant_buffer constants; #endif -- cgit v1.2.3 From 59e7bfa1de77ab03ca174bdf82fe33ab36050ce7 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 13:17:08 -0600 Subject: Add support for more surface types in sp_surface.c replace PIPE_FORMAT_U_L8_A8 with PIPE_FORMAT_U_A8_L8 --- src/mesa/pipe/i915simple/i915_state_sampler.c | 2 +- src/mesa/pipe/p_defines.h | 2 +- src/mesa/pipe/softpipe/sp_surface.c | 364 ++++++++++++++++++++++++++ src/mesa/state_tracker/st_atom_vs.c | 2 + src/mesa/state_tracker/st_format.c | 12 +- 5 files changed, 378 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index e4d7fec27b..7a595d1022 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -295,7 +295,7 @@ translate_texture_format(uint pipeFormat) return MAPSURF_8BIT | MT_8BIT_I8; case PIPE_FORMAT_U_A8: return MAPSURF_8BIT | MT_8BIT_A8; - case PIPE_FORMAT_U_L8_A8: + case PIPE_FORMAT_U_A8_L8: return MAPSURF_16BIT | MT_16BIT_AY88; case PIPE_FORMAT_U_R5_G6_B5: return MAPSURF_16BIT | MT_16BIT_RGB565; diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 636711938c..c1164c5c08 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -178,7 +178,7 @@ #define PIPE_FORMAT_U_L8 6 /**< ubyte luminance */ #define PIPE_FORMAT_U_A8 7 /**< ubyte alpha */ #define PIPE_FORMAT_U_I8 8 /**< ubyte intensity */ -#define PIPE_FORMAT_U_L8_A8 9 /**< ubyte luminance, alpha */ +#define PIPE_FORMAT_U_A8_L8 9 /**< ubyte alpha, luminance */ #define PIPE_FORMAT_S_R16_G16_B16_A16 10 /**< signed 16-bit RGBA (accum) */ #define PIPE_FORMAT_YCBCR 11 #define PIPE_FORMAT_YCBCR_REV 12 diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 84d02f91e7..75e36e3e3f 100755 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -200,6 +200,348 @@ z16_write_quad_z(struct softpipe_surface *sps, } +/*** PIPE_FORMAT_U_L8 ***/ + +static void +l8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) +{ + const ubyte *src + = ((const ubyte *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + unsigned i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_L8); + assert(x < (int) sps->surface.width - 1); + assert(y < (int) sps->surface.height - 1); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + rrrr[0][i * 2 + j] = + rrrr[1][i * 2 + j] = + rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT(src[j]); + rrrr[3][i * 2 + j] = 1.0F; + } + src += sps->surface.region->pitch; + } +} + +static void +l8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) +{ + ubyte *dst + = ((ubyte *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + unsigned i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_L8); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + ubyte r; + UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/ + dst[j] = r; + } + dst += sps->surface.region->pitch; + } +} + +static void +l8_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const ubyte *src + = ((const ubyte *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_L8); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp clipping hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(src[j]); + pRow[3] = 1.0; + pRow += 4; + } + src += ps->region->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_U_A8 ***/ + +static void +a8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) +{ + const ubyte *src + = ((const ubyte *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + unsigned i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8); + assert(x < (int) sps->surface.width - 1); + assert(y < (int) sps->surface.height - 1); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + rrrr[0][i * 2 + j] = + rrrr[1][i * 2 + j] = + rrrr[2][i * 2 + j] = 0.0F; + rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT(src[j]); + } + src += sps->surface.region->pitch; + } +} + +static void +a8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) +{ + ubyte *dst + = ((ubyte *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + unsigned i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + ubyte r; + UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[3][i * 2 + j]); /*A*/ + dst[j] = r; + } + dst += sps->surface.region->pitch; + } +} + +static void +a8_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const ubyte *src + = ((const ubyte *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_A8); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp clipping hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[0] = + pRow[1] = + pRow[2] = 0.0; + pRow[3] = UBYTE_TO_FLOAT(src[j]); + pRow += 4; + } + src += ps->region->pitch; + p += w0 * 4; + } +} + + + +/*** PIPE_FORMAT_U_I8 ***/ + +static void +i8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) +{ + const ubyte *src + = ((const ubyte *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + unsigned i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_I8); + assert(x < (int) sps->surface.width - 1); + assert(y < (int) sps->surface.height - 1); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + rrrr[0][i * 2 + j] = + rrrr[1][i * 2 + j] = + rrrr[2][i * 2 + j] = + rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT(src[j]); + } + src += sps->surface.region->pitch; + } +} + +static void +i8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) +{ + ubyte *dst + = ((ubyte *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + unsigned i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_I8); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + ubyte r; + UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/ + dst[j] = r; + } + dst += sps->surface.region->pitch; + } +} + +static void +i8_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const ubyte *src + = ((const ubyte *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_I8); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp clipping hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = UBYTE_TO_FLOAT(src[j]); + pRow += 4; + } + src += ps->region->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_U_A8_L8 ***/ + +static void +a8_l8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) +{ + const ushort *src + = ((const ushort *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + unsigned i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_L8); + assert(x < (int) sps->surface.width - 1); + assert(y < (int) sps->surface.height - 1); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + const ushort p = src[j]; + rrrr[0][i * 2 + j] = + rrrr[1][i * 2 + j] = + rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT(p >> 8); + rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT(p & 0xff); + } + src += sps->surface.region->pitch; + } +} + +static void +a8_l8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, + float (*rrrr)[QUAD_SIZE]) +{ + ushort *dst + = ((ushort *) (sps->surface.region->map + sps->surface.offset)) + + y * sps->surface.region->pitch + x; + unsigned i, j; + + assert(sps->surface.format == PIPE_FORMAT_U_A8_L8); + + for (i = 0; i < 2; i++) { /* loop over pixel row */ + for (j = 0; j < 2; j++) { /* loop over pixel column */ + ubyte l, a; + UNCLAMPED_FLOAT_TO_UBYTE(l, rrrr[0][i * 2 + j]); /*R*/ + UNCLAMPED_FLOAT_TO_UBYTE(a, rrrr[3][i * 2 + j]); /*A*/ + dst[j] = (l << 8) | a; + } + dst += sps->surface.region->pitch; + } +} + +static void +a8_l8_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const ushort *src + = ((const ushort *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_A8_L8); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp clipping hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const ushort p = src[j]; + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(p & 0xff); + pRow[3] = UBYTE_TO_FLOAT(p >> 8); + pRow += 4; + } + src += ps->region->pitch; + p += w0 * 4; + } +} + + + + /*** PIPE_FORMAT_U_Z32 ***/ static void @@ -367,6 +709,27 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) case PIPE_FORMAT_U_A1_R5_G5_B5: sps->surface.get_tile = a1r5g5b5_get_tile; break; + case PIPE_FORMAT_U_L8: + sps->read_quad_f_swz = l8_read_quad_f_swz; + sps->write_quad_f_swz = l8_write_quad_f_swz; + sps->surface.get_tile = l8_get_tile; + break; + case PIPE_FORMAT_U_A8: + sps->read_quad_f_swz = a8_read_quad_f_swz; + sps->write_quad_f_swz = a8_write_quad_f_swz; + sps->surface.get_tile = a8_get_tile; + break; + case PIPE_FORMAT_U_I8: + sps->read_quad_f_swz = i8_read_quad_f_swz; + sps->write_quad_f_swz = i8_write_quad_f_swz; + sps->surface.get_tile = i8_get_tile; + break; + case PIPE_FORMAT_U_A8_L8: + sps->read_quad_f_swz = a8_l8_read_quad_f_swz; + sps->write_quad_f_swz = a8_l8_write_quad_f_swz; + sps->surface.get_tile = a8_l8_get_tile; + break; + case PIPE_FORMAT_U_Z16: sps->read_quad_z = z16_read_quad_z; sps->write_quad_z = z16_write_quad_z; @@ -381,6 +744,7 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) sps->read_quad_stencil = s8z24_read_quad_stencil; sps->write_quad_stencil = s8z24_write_quad_stencil; break; + case PIPE_FORMAT_U_S8: sps->read_quad_stencil = s8_read_quad_stencil; sps->write_quad_stencil = s8_write_quad_stencil; diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index a326c12608..7f11533f74 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -162,8 +162,10 @@ const struct st_tracked_state st_update_vs = { */ static void update_tnl( struct st_context *st ) { + uint before = st->ctx->NewState; if (st->ctx->VertexProgram._MaintainTnlProgram) _tnl_UpdateFixedFunctionProgram( st->ctx ); + assert(before == st->ctx->NewState); } diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index a89dfb10fb..84e6a2d5fd 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -115,6 +115,14 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) case MESA_FORMAT_ARGB8888_REV: case MESA_FORMAT_ARGB8888: return PIPE_FORMAT_U_A8_R8_G8_B8; + case MESA_FORMAT_AL88: + return PIPE_FORMAT_U_A8_L8; + case MESA_FORMAT_A8: + return PIPE_FORMAT_U_A8; + case MESA_FORMAT_L8: + return PIPE_FORMAT_U_L8; + case MESA_FORMAT_I8: + return PIPE_FORMAT_U_I8; default: assert(0); return 0; @@ -277,8 +285,8 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - if (allow[PIPE_FORMAT_U_L8_A8]) - return PIPE_FORMAT_U_L8_A8; + if (allow[PIPE_FORMAT_U_A8_L8]) + return PIPE_FORMAT_U_A8_L8; return default_rgba_format(supported, n); case GL_INTENSITY: -- cgit v1.2.3 From b7de64c46cc76abea99ec3fbe80d6b29a0645b56 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 18:06:11 -0600 Subject: set I915_NEW_VERTEX_FORMAT in calculate_vertex_layout(), fixes demos/texenv --- src/mesa/pipe/i915simple/i915_state_derived.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index bab3015fde..b1ea4f86c6 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -161,6 +161,12 @@ static void calculate_vertex_layout( struct i915_context *i915 ) draw_set_vertex_attributes( i915->draw, vinfo->slot_to_attrib, vinfo->num_attribs); + + /* Need to set this flag so that the LIS2/4 registers get set. + * It also means the i915_update_immediate() function must be called + * after this one, in i915_update_derived(). + */ + i915->dirty |= I915_NEW_VERTEX_FORMAT; } -- cgit v1.2.3 From 29dc25bcfebfcb1764382bee960851ab051e6bbe Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 18:19:50 -0600 Subject: check for USE_X86_ASM or SLANG_X86 in st_delete_program() --- src/mesa/state_tracker/st_cb_program.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 25da720d86..b84f552404 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -112,9 +112,11 @@ static void st_delete_program( GLcontext *ctx, switch( prog->Target ) { case GL_VERTEX_PROGRAM_ARB: { +#if defined(USE_X86_ASM) || defined(SLANG_X86) struct st_vertex_program *p = (struct st_vertex_program *) prog; x86_release_func( &p->sse2_program ); +#endif break; } -- cgit v1.2.3 From 703140bbd58931046c00b43e52a7c4e9235875ca Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 18:26:42 -0600 Subject: Fix fragment program color output mapping (0=depth, 1=color) --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_fs.c | 4 ++-- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 4 ++-- 3 files changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 6fcba719ae..c2ad80c5d0 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -237,10 +237,10 @@ get_result_vector(struct i915_fp_compile *p, switch (dest->DstRegister.File) { case TGSI_FILE_OUTPUT: switch (dest->DstRegister.Index) { - case TGSI_ATTRIB_COLOR0: - return UREG(REG_TYPE_OC, 0); - case TGSI_ATTRIB_POS: + case 0: /**TGSI_ATTRIB_POS:**/ return UREG(REG_TYPE_OD, 0); + case 1: /**TGSI_ATTRIB_COLOR0:**/ + return UREG(REG_TYPE_OC, 0); default: i915_program_error(p, "Bad inst->DstReg.Index"); return 0; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index bfc31f94a6..cb0b6d8a77 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -136,10 +136,10 @@ shade_quad( /* run shader */ tgsi_exec_machine_run( &machine ); - /* store result color */ + /* store result color (always in output[1]) */ memcpy( quad->outputs.color, - &machine.Outputs[TGSI_ATTRIB_COLOR0].xyzw[0].f[0], + &machine.Outputs[1].xyzw[0].f[0], sizeof( quad->outputs.color ) ); #if 0 diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 78390767df..182a7eff05 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -356,11 +356,11 @@ map_register_file_index( * color results -> index 1, 2, ... */ if( index == FRAG_RESULT_DEPR ) { - mapped_index = TGSI_ATTRIB_POS; + mapped_index = 0; /**TGSI_ATTRIB_POS;**/ } else { assert( index == FRAG_RESULT_COLR ); - mapped_index = TGSI_ATTRIB_COLOR0; + mapped_index = 1; /**TGSI_ATTRIB_COLOR0;**/ } } else { -- cgit v1.2.3 From b4a8175da08f950196047a22d0c8428498ecbe97 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 18:50:56 -0600 Subject: s/==/=/ --- src/mesa/pipe/i915simple/i915_state_derived.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index b1ea4f86c6..9bd0af0f53 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -137,7 +137,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) } else { vinfo->hwfmt[0] |= S4_VFMT_XYZ; - vinfo->format[0] == FORMAT_3F; + vinfo->format[0] = FORMAT_3F; } /* Additional attributes required for setup: Just twosided -- cgit v1.2.3 From def8bb784cf403474eea3b4d25a3262155dc38f2 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 24 Aug 2007 19:36:23 -0600 Subject: code re-org, minor improvements --- src/mesa/state_tracker/st_cb_clear.c | 64 ++++++++++++++++++++++++++++++++---- 1 file changed, 58 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 69b985e405..fa222df2a4 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -93,6 +93,7 @@ depth_value(GLuint pipeFormat, GLfloat value) val = (GLuint) (value * 0xffffff); break; default: + val = 0; assert(0); } return val; @@ -402,6 +403,52 @@ clear_with_quad(GLcontext *ctx, } +/** + * Determine if we need to clear the depth buffer by drawing a quad. + */ +static INLINE GLboolean +check_clear_color_with_quad(GLcontext *ctx) +{ + return !(ctx->Color.ColorMask[0] && + ctx->Color.ColorMask[1] && + ctx->Color.ColorMask[2] && + ctx->Color.ColorMask[3] && + !ctx->Scissor.Enabled); +} + + +/** + * Determine if we need to clear the depth buffer by drawing a quad. + */ +static INLINE GLboolean +check_clear_depth_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + const struct st_renderbuffer *strb = st_renderbuffer(rb); + const GLboolean isDS = is_depth_stencil_format(strb->surface->format); + return ctx->Scissor.Enabled + || (isDS && ctx->DrawBuffer->Visual.stencilBits > 0); +} + + +/** + * Determine if we need to clear the stencil buffer by drawing a quad. + */ +static INLINE GLboolean +check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + const struct st_renderbuffer *strb = st_renderbuffer(rb); + const GLboolean isDS = is_depth_stencil_format(strb->surface->format); + const GLuint stencilMax = (1 << rb->StencilBits) - 1; + const GLboolean maskStencil + = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax; + return maskStencil + || ctx->Scissor.Enabled + || (isDS && ctx->DrawBuffer->Visual.depthBits > 0); +} + + + + static void clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { @@ -474,7 +521,8 @@ clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) struct st_renderbuffer *strb = st_renderbuffer(rb); const GLboolean isDS = is_depth_stencil_format(strb->surface->format); const GLuint stencilMax = (1 << rb->StencilBits) - 1; - GLboolean maskStencil = ctx->Stencil.WriteMask[0] != stencilMax; + GLboolean maskStencil + = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax; if (maskStencil || ctx->Scissor.Enabled || @@ -494,8 +542,9 @@ static void clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { struct st_renderbuffer *strb = st_renderbuffer(rb); - const GLuint stencilMax = 1 << rb->StencilBits; - GLboolean maskStencil = ctx->Stencil.WriteMask[0] != stencilMax; + const GLuint stencilMax = (1 << rb->StencilBits) - 1; + GLboolean maskStencil + = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax; assert(is_depth_stencil_format(strb->surface->format)); @@ -540,6 +589,7 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) = ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer; struct gl_renderbuffer *stencilRb = ctx->DrawBuffer->Attachment[BUFFER_STENCIL].Renderbuffer; + GLbitfield cmask = mask & BUFFER_BITS_COLOR; /* This makes sure the softpipe has the latest scissor, etc values */ st_validate_state( st ); @@ -552,15 +602,17 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) * color/depth/stencil individually... */ - if (mask & BUFFER_BITS_COLOR) { + if (cmask) { GLuint b; - for (b = 0; b < BUFFER_COUNT; b++) { - if (BUFFER_BITS_COLOR & mask & (1 << b)) { + for (b = 0; cmask; b++) { + if (cmask & (1 << b)) { struct gl_renderbuffer *rb = ctx->DrawBuffer->Attachment[b].Renderbuffer; assert(rb); clear_color_buffer(ctx, rb); + cmask &= ~(1 << b); /* turn off bit */ } + assert(b < BUFFER_COUNT); } } -- cgit v1.2.3 From 599ba515f979a343febbaf684a744ff525554727 Mon Sep 17 00:00:00 2001 From: keithw Date: Fri, 24 Aug 2007 12:30:28 +0100 Subject: remove unneeded dirty flag --- src/mesa/state_tracker/st_atom_vs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 7f11533f74..b2d25fa993 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -176,7 +176,7 @@ const struct st_tracked_state st_update_tnl = { _NEW_TEXTURE | _NEW_TRANSFORM | _NEW_LIGHT), /* XXX more? */ - .st = ST_NEW_MESA, /* XXX correct? */ + .st = 0 }, .update = update_tnl }; -- cgit v1.2.3 From 07d97e80e616d9fdc437d3b41055c347d5c54932 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 24 Aug 2007 18:30:00 +0100 Subject: checkpoint in constant tracking rework --- src/mesa/sources | 6 +- src/mesa/state_tracker/st_atom.c | 51 ++++++++++---- src/mesa/state_tracker/st_atom.h | 3 +- src/mesa/state_tracker/st_atom_cbuf.c | 72 ------------------- src/mesa/state_tracker/st_atom_fs.c | 84 ++++++++-------------- src/mesa/state_tracker/st_atom_vs.c | 124 ++++++++++----------------------- src/mesa/state_tracker/st_cb_program.c | 16 ++++- src/mesa/state_tracker/st_context.h | 5 +- src/mesa/state_tracker/st_program.h | 20 +----- src/mesa/tnl/t_vp_build.h | 7 ++ 10 files changed, 134 insertions(+), 254 deletions(-) delete mode 100644 src/mesa/state_tracker/st_atom_cbuf.c (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 61a8d580ad..53969774f9 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -186,10 +186,11 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_blend.c \ state_tracker/st_atom_clear_color.c \ state_tracker/st_atom_clip.c \ + state_tracker/st_atom_constbuf.c \ state_tracker/st_atom_depth.c \ - state_tracker/st_atom_fs.c \ - state_tracker/st_atom_vs.c \ + state_tracker/st_atom_fixedfunction.c \ state_tracker/st_atom_framebuffer.c \ + state_tracker/st_atom_fs.c \ state_tracker/st_atom_sampler.c \ state_tracker/st_atom_scissor.c \ state_tracker/st_atom_setup.c \ @@ -197,6 +198,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_stipple.c \ state_tracker/st_atom_texture.c \ state_tracker/st_atom_viewport.c \ + state_tracker/st_atom_vs.c \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_flush.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index d67291e50b..99fcbdfda7 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -29,8 +29,10 @@ #include "glheader.h" #include "context.h" +#include "pipe/p_defines.h" #include "st_context.h" #include "st_atom.h" +#include "st_program.h" @@ -46,9 +48,11 @@ static const struct st_tracked_state *atoms[] = &st_update_clear_color, &st_update_depth, &st_update_clip, + &st_update_tnl, &st_update_vs, &st_update_fs, + &st_update_setup, &st_update_polygon_stipple, &st_update_viewport, @@ -57,8 +61,8 @@ static const struct st_tracked_state *atoms[] = &st_update_stencil, &st_update_sampler, &st_update_texture, - /* will be patched out at runtime */ -/* &st_update_constants */ + &st_update_vs_constants, + &st_update_fs_constants, }; @@ -72,13 +76,17 @@ void st_init_atoms( struct st_context *st ) /* Patch in a pointer to the dynamic state atom: */ - for (i = 0; i < st->nr_atoms; i++) - if (st->atoms[i] == &st_update_constants) - st->atoms[i] = &st->constants.tracked_state; + for (i = 0; i < st->nr_atoms; i++) { + if (st->atoms[i] == &st_update_vs_constants) { + st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_VERTEX]; + st->atoms[i][0] = st_update_vs_constants; + } - memcpy(&st->constants.tracked_state, - &st_update_constants, - sizeof(st_update_constants)); + if (st->atoms[i] == &st_update_fs_constants) { + st->atoms[i] = &st->constants.tracked_state[PIPE_SHADER_FRAGMENT]; + st->atoms[i][0] = st_update_fs_constants; + } + } } @@ -118,6 +126,21 @@ static void xor_states( struct st_state_flags *result, } +/* Too complex to figure out, just check every time: + */ +static void check_program_state( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + + if (ctx->VertexProgram._Current != &st->vp->Base) + st->dirty.st |= ST_NEW_VERTEX_PROGRAM; + + if (ctx->FragmentProgram._Current != &st->fp->Base) + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; + +} + + /*********************************************************************** * Update all derived state: */ @@ -127,6 +150,8 @@ void st_validate_state( struct st_context *st ) struct st_state_flags *state = &st->dirty; GLuint i; + check_program_state( st ); + if (state->st == 0) return; @@ -142,10 +167,12 @@ void st_validate_state( struct st_context *st ) for (i = 0; i < st->nr_atoms; i++) { const struct st_tracked_state *atom = st->atoms[i]; struct st_state_flags generated; - - assert(atom->dirty.mesa || - atom->dirty.st); - assert(atom->update); + + if (!(atom->dirty.mesa || atom->dirty.st) || + !atom->update) { + _mesa_printf("malformed atom %d\n", i); + assert(0); + } if (check_state(state, &atom->dirty)) { st->atoms[i]->update( st ); diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 51da489f6d..447430bfef 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -54,12 +54,13 @@ const struct st_tracked_state st_update_vs; const struct st_tracked_state st_update_setup; const struct st_tracked_state st_update_polygon_stipple; const struct st_tracked_state st_update_viewport; -const struct st_tracked_state st_update_constants; const struct st_tracked_state st_update_scissor; const struct st_tracked_state st_update_blend; const struct st_tracked_state st_update_stencil; const struct st_tracked_state st_update_sampler; const struct st_tracked_state st_update_texture; +const struct st_tracked_state st_update_fs_constants; +const struct st_tracked_state st_update_vs_constants; #endif diff --git a/src/mesa/state_tracker/st_atom_cbuf.c b/src/mesa/state_tracker/st_atom_cbuf.c deleted file mode 100644 index 0f90aa7646..0000000000 --- a/src/mesa/state_tracker/st_atom_cbuf.c +++ /dev/null @@ -1,72 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "st_context.h" -#include "pipe/p_context.h" -#include "st_atom.h" - -extern GLboolean xmesa_get_cbuf_details( GLcontext *ctx, - void **ptr, - GLuint *cpp, - GLint *stride, - GLuint *format ); - - -/* This is a hack to work with the X11 driver as a test harness - */ -static void update_cbuf_state( struct st_context *st ) -{ - struct pipe_surface cbuf; - GLboolean ok; - - ok = xmesa_get_cbuf_details( st->ctx, - (void **)&cbuf.ptr, - &cbuf.cpp, - &cbuf.stride, - &cbuf.format ); - - assert(ok); - - if (memcmp(&cbuf, &st->state.cbuf, sizeof(cbuf)) != 0) { - st->state.cbuf = cbuf; - st->pipe->set_cbuf_state( st->pipe, &cbuf ); - } -} - -const struct st_tracked_state st_update_cbuf = { - .dirty = { - .mesa = _NEW_BUFFERS, - .st = 0, - }, - .update = update_cbuf_state -}; - diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index a164ded184..f706761198 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -43,92 +43,64 @@ #define TGSI_DEBUG 0 -static void compile_fs( struct st_context *st, - struct st_fragment_program *fs ) +static void compile_fs( struct st_context *st ) { + struct st_fragment_program *fp = st->fp; + /* XXX: fix static allocation of tokens: */ - tgsi_mesa_compile_fp_program( &fs->Base, fs->tokens, ST_FP_MAX_TOKENS ); - - if (TGSI_DEBUG) - tgsi_dump( fs->tokens, TGSI_DUMP_VERBOSE ); -} - - -static void -update_fs_constants(struct st_context *st, - struct gl_program_parameter_list *params) - -{ - const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4; - struct pipe_winsys *ws = st->pipe->winsys; - struct pipe_constant_buffer *cbuf - = &st->state.constants[PIPE_SHADER_FRAGMENT]; + tgsi_mesa_compile_fp_program( &fp->Base, fp->tokens, ST_FP_MAX_TOKENS ); - if (!cbuf->buffer) - cbuf->buffer = ws->buffer_create(ws, 1); + fp->fs.inputs_read + = tgsi_mesa_translate_vertex_input_mask(fp->Base.Base.InputsRead); + fp->fs.outputs_written + = tgsi_mesa_translate_vertex_output_mask(fp->Base.Base.OutputsWritten); + fp->fs.tokens = &fp->tokens[0]; - /* load Mesa constants into the constant buffer */ - if (paramBytes) - ws->buffer_data(ws, cbuf->buffer, paramBytes, params->ParameterValues); - - cbuf->size = paramBytes; + if (TGSI_DEBUG) + tgsi_dump( fp->tokens, TGSI_DUMP_VERBOSE ); - st->pipe->set_constant_buffer(st->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf); + fp->dirty = 0; } + static void update_fs( struct st_context *st ) { - struct pipe_shader_state fs; struct st_fragment_program *fp = NULL; - struct gl_program_parameter_list *params = NULL; - /* find active shader and params */ + /* find active shader and params. Changes to this Mesa state + * should be covered by ST_NEW_FRAGMENT_PROGRAM, thanks to the + * logic in st_cb_program.c + */ if (st->ctx->Shader.CurrentProgram && st->ctx->Shader.CurrentProgram->LinkStatus && st->ctx->Shader.CurrentProgram->FragmentProgram) { struct gl_fragment_program *f = st->ctx->Shader.CurrentProgram->FragmentProgram; fp = st_fragment_program(f); - params = f->Base.Parameters; } - else if (st->ctx->FragmentProgram._Current) { + else { + assert(st->ctx->FragmentProgram._Current); fp = st_fragment_program(st->ctx->FragmentProgram._Current); - params = st->ctx->FragmentProgram._Current->Base.Parameters; - } - - /* update constants */ - if (fp && params) { - _mesa_load_state_parameters(st->ctx, params); - update_fs_constants(st, params); } /* translate shader to TGSI format */ - if (fp->dirty) - compile_fs( st, fp ); - - /* update pipe state */ - memset( &fs, 0, sizeof(fs) ); - fs.inputs_read - = tgsi_mesa_translate_fragment_input_mask(fp->Base.Base.InputsRead); - fs.outputs_written - = tgsi_mesa_translate_fragment_output_mask(fp->Base.Base.OutputsWritten); - fs.tokens = &fp->tokens[0]; - - if (memcmp(&fs, &st->state.fs, sizeof(fs)) != 0 || - fp->dirty) - { - fp->dirty = 0; - st->state.fs = fs; - st->pipe->set_fs_state(st->pipe, &fs); + if (st->fp != fp || fp->dirty) { + st->fp = fp; + + if (fp->dirty) + compile_fs( st ); + + st->state.fs = fp->fs; + st->pipe->set_fs_state(st->pipe, &st->state.fs); } } const struct st_tracked_state st_update_fs = { .dirty = { - .mesa = _NEW_PROGRAM, + .mesa = 0, .st = ST_NEW_FRAGMENT_PROGRAM, }, .update = update_fs diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index b2d25fa993..ab7e2ae4be 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -45,108 +45,79 @@ #include "st_program.h" -#define TGSI_DEBUG 0 +#define TGSI_DEBUG 1 -static void compile_vs( struct st_context *st, - struct st_vertex_program *vs ) + + + +/* translate shader to TGSI format +*/ +static void compile_vs( struct st_context *st ) { + struct st_vertex_program *vp = st->vp; + /* XXX: fix static allocation of tokens: */ - tgsi_mesa_compile_vp_program( &vs->Base, vs->tokens, ST_FP_MAX_TOKENS ); + tgsi_mesa_compile_vp_program( &vp->Base, vp->tokens, ST_FP_MAX_TOKENS ); + + vp->vs.inputs_read + = tgsi_mesa_translate_vertex_input_mask(vp->Base.Base.InputsRead); + vp->vs.outputs_written + = tgsi_mesa_translate_vertex_output_mask(vp->Base.Base.OutputsWritten); + vp->vs.tokens = &vp->tokens[0]; if (TGSI_DEBUG) - tgsi_dump( vs->tokens, TGSI_DUMP_VERBOSE ); + tgsi_dump( vp->tokens, 0 ); #if defined(USE_X86_ASM) || defined(SLANG_X86) tgsi_emit_sse2( - vs->tokens, - &vs->sse2_program ); + vp->vs.tokens, + &vp->vs.sse2_program ); #endif -} - - -static void -update_vs_constants(struct st_context *st, - struct gl_program_parameter_list *params) - -{ - const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4; - struct pipe_winsys *ws = st->pipe->winsys; - struct pipe_constant_buffer *cbuf - = &st->state.constants[PIPE_SHADER_VERTEX]; - - if (!cbuf->buffer) - cbuf->buffer = ws->buffer_create(ws, 1); - - /* load Mesa constants into the constant buffer */ - if (paramBytes) - ws->buffer_data(ws, cbuf->buffer, paramBytes, params->ParameterValues); - - cbuf->size = paramBytes; - st->pipe->set_constant_buffer(st->pipe, PIPE_SHADER_VERTEX, 0, cbuf); + vp->dirty = 0; } + static void update_vs( struct st_context *st ) { - struct pipe_shader_state vs; - struct st_vertex_program *vp = NULL; - struct gl_program_parameter_list *params = NULL; + struct st_vertex_program *vp; - /* find active shader and params */ + /* find active shader and params -- Should be covered by + * ST_NEW_VERTEX_PROGRAM + */ if (st->ctx->Shader.CurrentProgram && st->ctx->Shader.CurrentProgram->LinkStatus && st->ctx->Shader.CurrentProgram->VertexProgram) { struct gl_vertex_program *f = st->ctx->Shader.CurrentProgram->VertexProgram; vp = st_vertex_program(f); - params = f->Base.Parameters; } - else if (st->ctx->VertexProgram._Current) { + else { + assert(st->ctx->VertexProgram._Current); vp = st_vertex_program(st->ctx->VertexProgram._Current); - params = st->ctx->VertexProgram._Current->Base.Parameters; } - /* update constants */ - if (vp && params) { - _mesa_load_state_parameters(st->ctx, params); - /*_mesa_print_parameter_list(params);*/ - update_vs_constants(st, params); - } + if (st->vp != vp || vp->dirty) { + st->vp = vp; - /* translate shader to TGSI format */ - if (vp->dirty) - compile_vs( st, vp ); - - /* update pipe state */ - memset( &vs, 0, sizeof(vs) ); - vs.inputs_read - = tgsi_mesa_translate_vertex_input_mask(vp->Base.Base.InputsRead); - vs.outputs_written - = tgsi_mesa_translate_vertex_output_mask(vp->Base.Base.OutputsWritten); - vs.tokens = &vp->tokens[0]; + if (vp->dirty) + compile_vs( st ); #if defined(USE_X86_ASM) || defined(SLANG_X86) - vs.executable = (void *) x86_get_func( &vp->sse2_program ); + vs.executable = (void *) x86_get_func( &vp->sse2_program ); #endif - if (memcmp(&vs, &st->state.vs, sizeof(vs)) != 0 || - vp->dirty) - { - vp->dirty = 0; - st->state.vs = vs; - st->pipe->set_vs_state(st->pipe, &vs); + st->state.vs = st->vp->vs; + st->pipe->set_vs_state(st->pipe, &st->state.vs); } } const struct st_tracked_state st_update_vs = { .dirty = { - .mesa = (_NEW_PROGRAM | - _NEW_MODELVIEW | - _NEW_PROJECTION | - _NEW_LIGHT), /*XXX MORE?*/ + .mesa = 0, .st = ST_NEW_VERTEX_PROGRAM, }, .update = update_vs @@ -155,28 +126,3 @@ const struct st_tracked_state st_update_vs = { - -/** - * When TnL state has changed, need to generate new vertex program. - * This should be done before updating the vertes shader (vs) state. - */ -static void update_tnl( struct st_context *st ) -{ - uint before = st->ctx->NewState; - if (st->ctx->VertexProgram._MaintainTnlProgram) - _tnl_UpdateFixedFunctionProgram( st->ctx ); - assert(before == st->ctx->NewState); -} - - -const struct st_tracked_state st_update_tnl = { - .dirty = { - .mesa = (_NEW_PROGRAM | - _NEW_LIGHT | - _NEW_TEXTURE | - _NEW_TRANSFORM | - _NEW_LIGHT), /* XXX more? */ - .st = 0 - }, - .update = update_tnl -}; diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index b84f552404..b9c19bdd3e 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -64,6 +64,17 @@ static void st_bind_program( GLcontext *ctx, } } +static void st_use_program( GLcontext *ctx, + GLuint program ) +{ + struct st_context *st = st_context(ctx); + + st->dirty.st |= ST_NEW_VERTEX_PROGRAM; + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; +} + + + static struct gl_program *st_new_program( GLcontext *ctx, GLenum target, GLuint id ) @@ -132,6 +143,7 @@ static GLboolean st_is_program_native( GLcontext *ctx, return GL_TRUE; } + static void st_program_string_notify( GLcontext *ctx, GLenum target, struct gl_program *prog ) @@ -166,10 +178,8 @@ static void st_program_string_notify( GLcontext *ctx, void st_init_program_functions(struct dd_function_table *functions) { -#if 0 - assert(functions->ProgramStringNotify == _tnl_program_string); -#endif functions->BindProgram = st_bind_program; + functions->UseProgram = st_use_program; functions->NewProgram = st_new_program; functions->DeleteProgram = st_delete_program; functions->IsProgramNative = st_is_program_native; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 13ea28237c..9e8015d4c7 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -86,7 +86,7 @@ struct st_context } state; struct { - struct st_tracked_state tracked_state; + struct st_tracked_state tracked_state[2]; } constants; struct { @@ -109,6 +109,9 @@ struct st_context GLfloat polygon_offset_scale; /* ?? */ + struct st_vertex_program *vp; + struct st_fragment_program *fp; + struct pipe_buffer_handle *default_attrib_buffer; }; diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 883953399c..7a91983ce9 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -54,21 +54,7 @@ struct st_fragment_program struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; -#if 0 - GLfloat (*cbuffer)[4]; - GLuint nr_constants; - - /* Translate all the parameters, etc, into a constant buffer which - * we update on state changes. - */ - struct - { - GLuint reg; /* Constant idx */ - const GLfloat *values; /* Pointer to tracked values */ - } *param; - GLuint nr_params; -#endif - + struct pipe_shader_state fs; GLuint param_state; }; @@ -89,9 +75,7 @@ struct st_vertex_program struct x86_function sse2_program; #endif -#if 0 - struct pipe_constant_buffer constants; -#endif + struct pipe_shader_state vs; GLuint param_state; }; diff --git a/src/mesa/tnl/t_vp_build.h b/src/mesa/tnl/t_vp_build.h index 4a98fff026..5e22fcf8c4 100644 --- a/src/mesa/tnl/t_vp_build.h +++ b/src/mesa/tnl/t_vp_build.h @@ -28,6 +28,13 @@ #include "mtypes.h" +#define TNL_FIXED_FUNCTION_STATE_FLAGS (_NEW_PROGRAM | \ + _NEW_LIGHT | \ + _NEW_TEXTURE | \ + _NEW_TRANSFORM | \ + _NEW_FOG | \ + _NEW_POINT) + extern void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ); extern void _tnl_ProgramCacheInit( GLcontext *ctx ); -- cgit v1.2.3 From 4185da4681405f3cc4d0cc601d428f2f44d0dda8 Mon Sep 17 00:00:00 2001 From: keithw Date: Sat, 25 Aug 2007 22:00:36 +0100 Subject: add names to tracked state atoms to improve debug --- src/mesa/state_tracker/st_atom.c | 9 ++++++++- src/mesa/state_tracker/st_atom_alphatest.c | 1 + src/mesa/state_tracker/st_atom_blend.c | 1 + src/mesa/state_tracker/st_atom_clear_color.c | 1 + src/mesa/state_tracker/st_atom_clip.c | 1 + src/mesa/state_tracker/st_atom_depth.c | 1 + src/mesa/state_tracker/st_atom_framebuffer.c | 1 + src/mesa/state_tracker/st_atom_fs.c | 1 + src/mesa/state_tracker/st_atom_sampler.c | 1 + src/mesa/state_tracker/st_atom_scissor.c | 1 + src/mesa/state_tracker/st_atom_setup.c | 1 + src/mesa/state_tracker/st_atom_stencil.c | 1 + src/mesa/state_tracker/st_atom_stipple.c | 1 + src/mesa/state_tracker/st_atom_texture.c | 1 + src/mesa/state_tracker/st_atom_viewport.c | 1 + src/mesa/state_tracker/st_atom_vs.c | 1 + src/mesa/state_tracker/st_context.h | 1 + 17 files changed, 24 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 99fcbdfda7..66ab5d7c3a 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -155,6 +155,8 @@ void st_validate_state( struct st_context *st ) if (state->st == 0) return; +// _mesa_printf("%s %x/%x\n", __FUNCTION__, state->mesa, state->st); + if (1) { /* Debug version which enforces various sanity checks on the * state flags which are generated and checked to help ensure @@ -168,14 +170,17 @@ void st_validate_state( struct st_context *st ) const struct st_tracked_state *atom = st->atoms[i]; struct st_state_flags generated; +// _mesa_printf("atom %s %x/%x\n", atom->name, atom->dirty.mesa, atom->dirty.st); + if (!(atom->dirty.mesa || atom->dirty.st) || !atom->update) { - _mesa_printf("malformed atom %d\n", i); + _mesa_printf("malformed atom %s\n", atom->name); assert(0); } if (check_state(state, &atom->dirty)) { st->atoms[i]->update( st ); +// _mesa_printf("after: %x\n", atom->dirty.mesa); } accumulate_state(&examined, &atom->dirty); @@ -188,6 +193,8 @@ void st_validate_state( struct st_context *st ) assert(!check_state(&examined, &generated)); prev = *state; } +// _mesa_printf("\n"); + } else { const GLuint nr = st->nr_atoms; diff --git a/src/mesa/state_tracker/st_atom_alphatest.c b/src/mesa/state_tracker/st_atom_alphatest.c index 1e2e449795..4378154053 100644 --- a/src/mesa/state_tracker/st_atom_alphatest.c +++ b/src/mesa/state_tracker/st_atom_alphatest.c @@ -81,6 +81,7 @@ update_alpha_test( struct st_context *st ) const struct st_tracked_state st_update_alpha_test = { + .name = "st_update_alpha_test", .dirty = { .mesa = (_NEW_COLOR), .st = 0, diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index 256f13471a..afd21a6141 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -227,6 +227,7 @@ update_blend( struct st_context *st ) const struct st_tracked_state st_update_blend = { + .name = "st_update_blend", .dirty = { .mesa = (_NEW_COLOR), /* XXX _NEW_BLEND someday? */ .st = 0, diff --git a/src/mesa/state_tracker/st_atom_clear_color.c b/src/mesa/state_tracker/st_atom_clear_color.c index adf730cd8c..ce3431c5d3 100644 --- a/src/mesa/state_tracker/st_atom_clear_color.c +++ b/src/mesa/state_tracker/st_atom_clear_color.c @@ -54,6 +54,7 @@ update_clear_color_state( struct st_context *st ) const struct st_tracked_state st_update_clear_color = { + .name = "st_update_clear_color", .dirty = { .mesa = _NEW_COLOR, .st = 0, diff --git a/src/mesa/state_tracker/st_atom_clip.c b/src/mesa/state_tracker/st_atom_clip.c index 8ccad637d5..a6f0568660 100644 --- a/src/mesa/state_tracker/st_atom_clip.c +++ b/src/mesa/state_tracker/st_atom_clip.c @@ -62,6 +62,7 @@ static void update_clip( struct st_context *st ) const struct st_tracked_state st_update_clip = { + .name = "st_update_clip", .dirty = { .mesa = (_NEW_TRANSFORM), .st = 0, diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index 7fc51953dc..df05c79e36 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -84,6 +84,7 @@ update_depth( struct st_context *st ) const struct st_tracked_state st_update_depth = { + .name = "st_update_depth", .dirty = { .mesa = (_NEW_DEPTH), .st = 0, diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index f054eb8f21..3c4b37e7c5 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -82,6 +82,7 @@ update_framebuffer_state( struct st_context *st ) const struct st_tracked_state st_update_framebuffer = { + .name = "st_update_framebuffer", .dirty = { .mesa = _NEW_BUFFERS, .st = 0, diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index f706761198..767816bf23 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -99,6 +99,7 @@ static void update_fs( struct st_context *st ) const struct st_tracked_state st_update_fs = { + .name = "st_update_fs", .dirty = { .mesa = 0, .st = ST_NEW_FRAGMENT_PROGRAM, diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 7d568baf9e..d65565f991 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -152,6 +152,7 @@ update_samplers(struct st_context *st) const struct st_tracked_state st_update_sampler = { + .name = "st_update_sampler", .dirty = { .mesa = _NEW_TEXTURE, .st = 0, diff --git a/src/mesa/state_tracker/st_atom_scissor.c b/src/mesa/state_tracker/st_atom_scissor.c index 2bf633828e..59601e91a1 100644 --- a/src/mesa/state_tracker/st_atom_scissor.c +++ b/src/mesa/state_tracker/st_atom_scissor.c @@ -83,6 +83,7 @@ update_scissor( struct st_context *st ) const struct st_tracked_state st_update_scissor = { + .name = "st_update_scissor", .dirty = { .mesa = (_NEW_SCISSOR | _NEW_BUFFERS), .st = 0, diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index 3eac2588df..09d921560d 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -211,6 +211,7 @@ static void update_setup_state( struct st_context *st ) } const struct st_tracked_state st_update_setup = { + .name = "st_update_setup", .dirty = { .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | _NEW_SCISSOR | _NEW_POINT | _NEW_BUFFERS | _NEW_MULTISAMPLE), diff --git a/src/mesa/state_tracker/st_atom_stencil.c b/src/mesa/state_tracker/st_atom_stencil.c index d037335e9e..b8aec0b3b6 100644 --- a/src/mesa/state_tracker/st_atom_stencil.c +++ b/src/mesa/state_tracker/st_atom_stencil.c @@ -127,6 +127,7 @@ update_stencil( struct st_context *st ) const struct st_tracked_state st_update_stencil = { + .name = "st_update_stencil", .dirty = { .mesa = (_NEW_STENCIL), .st = 0, diff --git a/src/mesa/state_tracker/st_atom_stipple.c b/src/mesa/state_tracker/st_atom_stipple.c index dd04d2158c..c91214059a 100644 --- a/src/mesa/state_tracker/st_atom_stipple.c +++ b/src/mesa/state_tracker/st_atom_stipple.c @@ -54,6 +54,7 @@ update_stipple( struct st_context *st ) const struct st_tracked_state st_update_polygon_stipple = { + .name = "st_update_polygon_stipple", .dirty = { .mesa = (_NEW_POLYGONSTIPPLE), .st = 0, diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index bafd38695f..7970bcf2b8 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -72,6 +72,7 @@ update_textures(struct st_context *st) const struct st_tracked_state st_update_texture = { + .name = "st_update_texture", .dirty = { .mesa = _NEW_TEXTURE, .st = 0, diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index a70f4c7434..147aa3c51a 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -86,6 +86,7 @@ update_viewport( struct st_context *st ) const struct st_tracked_state st_update_viewport = { + .name = "st_update_viewport", .dirty = { .mesa = _NEW_BUFFERS | _NEW_VIEWPORT, .st = 0, diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index ab7e2ae4be..a1e6117bde 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -116,6 +116,7 @@ static void update_vs( struct st_context *st ) const struct st_tracked_state st_update_vs = { + .name = "st_update_vs", .dirty = { .mesa = 0, .st = ST_NEW_VERTEX_PROGRAM, diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 9e8015d4c7..cb34994d77 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -48,6 +48,7 @@ struct st_state_flags { }; struct st_tracked_state { + const char *name; struct st_state_flags dirty; void (*update)( struct st_context *st ); }; -- cgit v1.2.3 From 3f8b5e3951cfa4d568f571a3073a78b81040f7fb Mon Sep 17 00:00:00 2001 From: keithw Date: Sat, 25 Aug 2007 22:01:03 +0100 Subject: fix frag vs vertex typo --- src/mesa/state_tracker/st_atom_fs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 767816bf23..d066547616 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -52,9 +52,9 @@ static void compile_fs( struct st_context *st ) tgsi_mesa_compile_fp_program( &fp->Base, fp->tokens, ST_FP_MAX_TOKENS ); fp->fs.inputs_read - = tgsi_mesa_translate_vertex_input_mask(fp->Base.Base.InputsRead); + = tgsi_mesa_translate_fragment_input_mask(fp->Base.Base.InputsRead); fp->fs.outputs_written - = tgsi_mesa_translate_vertex_output_mask(fp->Base.Base.OutputsWritten); + = tgsi_mesa_translate_fragment_output_mask(fp->Base.Base.OutputsWritten); fp->fs.tokens = &fp->tokens[0]; if (TGSI_DEBUG) -- cgit v1.2.3 From b0455958cc5fbb303cd5cd394c7945f357cc9663 Mon Sep 17 00:00:00 2001 From: keithw Date: Sat, 25 Aug 2007 22:05:16 +0100 Subject: add missing files --- src/mesa/state_tracker/st_atom_constbuf.c | 129 +++++++++++++++++++++++++ src/mesa/state_tracker/st_atom_fixedfunction.c | 66 +++++++++++++ 2 files changed, 195 insertions(+) create mode 100644 src/mesa/state_tracker/st_atom_constbuf.c create mode 100644 src/mesa/state_tracker/st_atom_fixedfunction.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c new file mode 100644 index 0000000000..7bea9476f1 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -0,0 +1,129 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "shader/prog_parameter.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" + +#include "st_context.h" +#include "st_atom.h" +#include "st_program.h" + +static void upload_constants( struct st_context *st, + struct gl_program_parameter_list *params, + unsigned id) +{ + struct pipe_winsys *ws = st->pipe->winsys; + struct pipe_constant_buffer *cbuf = &st->state.constants[id]; + + /* update constants */ + if (params && params->NumParameters) { + const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4; + + /* Update our own dependency flags. This works because this + * function will also be called whenever the program changes. + */ + st->constants.tracked_state[id].dirty.mesa = params->StateFlags; + + _mesa_load_state_parameters(st->ctx, params); + + if (!cbuf->buffer) + cbuf->buffer = ws->buffer_create(ws, 1); + + if (0) + { + int i; + + _mesa_printf("%s(%d): %d / %x\n", + __FUNCTION__, id, params->NumParameters, params->StateFlags); + + for (i = 0; i < params->NumParameters; i++) + fprintf(stderr, "%d: %f %f %f %f\n", i, + params->ParameterValues[i][0], + params->ParameterValues[i][1], + params->ParameterValues[i][2], + params->ParameterValues[i][3]); + } + + /* load Mesa constants into the constant buffer */ + ws->buffer_data(ws, cbuf->buffer, paramBytes, params->ParameterValues); + + cbuf->size = paramBytes; + + st->pipe->set_constant_buffer(st->pipe, id, 0, cbuf); + } + else { + st->constants.tracked_state[id].dirty.mesa = 0; + // st->pipe->set_constant_buffer(st->pipe, id, 0, NULL); + } +} + +/* Vertex shader: + */ +static void update_vs_constants(struct st_context *st ) +{ + struct st_vertex_program *vp = st->vp; + struct gl_program_parameter_list *params = vp->Base.Base.Parameters; + + upload_constants( st, params, PIPE_SHADER_VERTEX ); +} + +const struct st_tracked_state st_update_vs_constants = { + .name = "st_update_vs_constants", + .dirty = { + .mesa = 0, + .st = ST_NEW_VERTEX_PROGRAM, + }, + .update = update_vs_constants +}; + +/* Fragment shader: + */ +static void update_fs_constants(struct st_context *st ) +{ + struct st_fragment_program *fp = st->fp; + struct gl_program_parameter_list *params = fp->Base.Base.Parameters; + + upload_constants( st, params, PIPE_SHADER_FRAGMENT ); +} + +const struct st_tracked_state st_update_fs_constants = { + .name = "st_update_fs_constants", + .dirty = { + .mesa = 0, + .st = ST_NEW_FRAGMENT_PROGRAM, + }, + .update = update_fs_constants +}; + diff --git a/src/mesa/state_tracker/st_atom_fixedfunction.c b/src/mesa/state_tracker/st_atom_fixedfunction.c new file mode 100644 index 0000000000..3f137e1633 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_fixedfunction.c @@ -0,0 +1,66 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "tnl/t_vp_build.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + +#include "st_context.h" +#include "st_atom.h" + + +#define TGSI_DEBUG 0 + + +/** + * When TnL state has changed, need to generate new vertex program. + * This should be done before updating the vertes shader (vs) state. + */ +static void update_tnl( struct st_context *st ) +{ + /* Would be good to avoid this when shaders are active: + */ + _tnl_UpdateFixedFunctionProgram( st->ctx ); +} + + +const struct st_tracked_state st_update_tnl = { + .name = "st_update_tnl", + .dirty = { + .mesa = TNL_FIXED_FUNCTION_STATE_FLAGS, + .st = 0 + }, + .update = update_tnl +}; + + -- cgit v1.2.3 From 874b9265601fcc05b1e32e2be029f3ac6a966c97 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 27 Aug 2007 19:30:11 +0100 Subject: Split out vertex shader/cache/fetch functionality from draw_prim.c --- src/mesa/pipe/draw/draw_arrays.c | 6 +- src/mesa/pipe/draw/draw_prim.c | 416 +------------------------------- src/mesa/pipe/draw/draw_vertex.h | 19 ++ src/mesa/pipe/draw/draw_vertex_cache.c | 175 ++++++++++++++ src/mesa/pipe/draw/draw_vertex_fetch.c | 101 ++++++++ src/mesa/pipe/draw/draw_vertex_shader.c | 203 ++++++++++++++++ src/mesa/sources | 3 + 7 files changed, 513 insertions(+), 410 deletions(-) create mode 100644 src/mesa/pipe/draw/draw_vertex_cache.c create mode 100644 src/mesa/pipe/draw/draw_vertex_fetch.c create mode 100644 src/mesa/pipe/draw/draw_vertex_shader.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index f6bf174dc0..b7d06dd5a7 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -55,7 +55,11 @@ draw_arrays(struct draw_context *draw, unsigned prim, /* tell drawing pipeline we're beginning drawing */ draw->pipeline.first->begin( draw->pipeline.first ); - draw_invalidate_vcache( draw ); + /* XXX: Shouldn't really be needed - cache should be invalidated + * after setting new vertex buffers, vertex elements, but not + * between draws. + */ + draw_vertex_cache_invalidate( draw ); draw_set_prim( draw, prim ); diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 09616cf315..b68cca5706 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -58,299 +58,6 @@ static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { }; -static INLINE unsigned -compute_clipmask(float cx, float cy, float cz, float cw) -{ - unsigned mask; -#if defined(macintosh) || defined(__powerpc__) - /* on powerpc cliptest is 17% faster in this way. */ - mask = (((cw < cx) << CLIP_RIGHT_SHIFT)); - mask |= (((cw < -cx) << CLIP_LEFT_SHIFT)); - mask |= (((cw < cy) << CLIP_TOP_SHIFT)); - mask |= (((cw < -cy) << CLIP_BOTTOM_SHIFT)); - mask |= (((cw < cz) << CLIP_FAR_SHIFT)); - mask |= (((cw < -cz) << CLIP_NEAR_SHIFT)); -#else /* !defined(macintosh)) */ - mask = 0x0; - if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT; - if ( cx + cw < 0) mask |= CLIP_LEFT_BIT; - if (-cy + cw < 0) mask |= CLIP_TOP_BIT; - if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT; - if (-cz + cw < 0) mask |= CLIP_FAR_BIT; - if ( cz + cw < 0) mask |= CLIP_NEAR_BIT; -#endif /* defined(macintosh) */ - return mask; -} - - -/** - * Fetch a float[4] vertex attribute from memory, doing format/type - * conversion as needed. - * XXX this might be a temporary thing. - */ -static void -fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) -{ - /* defaults */ - attrib[1] = 0.0; - attrib[2] = 0.0; - attrib[3] = 1.0; - switch (format) { - case PIPE_FORMAT_R32G32B32A32_FLOAT: - attrib[3] = ((float *) ptr)[3]; - /* fall-through */ - case PIPE_FORMAT_R32G32B32_FLOAT: - attrib[2] = ((float *) ptr)[2]; - /* fall-through */ - case PIPE_FORMAT_R32G32_FLOAT: - attrib[1] = ((float *) ptr)[1]; - /* fall-through */ - case PIPE_FORMAT_R32_FLOAT: - attrib[0] = ((float *) ptr)[0]; - break; - default: - assert(0); - } -} - -#if !defined(XSTDCALL) -#if defined(WIN32) -#define XSTDCALL __stdcall -#else -#define XSTDCALL -#endif -#endif - -#if defined(USE_X86_ASM) || defined(SLANG_X86) -typedef void (XSTDCALL *sse2_function)( - const struct tgsi_exec_vector *input, - struct tgsi_exec_vector *output, - float (*constant)[4], - struct tgsi_exec_vector *temporary ); -#endif - -/** - * Transform vertices with the current vertex program/shader - * Up to four vertices can be shaded at a time. - * \param vbuffer the input vertex data - * \param elts indexes of four input vertices - * \param count number of vertices to shade [1..4] - * \param vOut array of pointers to four output vertices - */ -static void -run_vertex_program(struct draw_context *draw, - unsigned elts[4], unsigned count, - struct vertex_header *vOut[]) -{ - struct tgsi_exec_machine machine; - unsigned int j; - -#if 0 - FILE *file = stdout; -#endif - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - - assert(count <= 4); - assert(draw->vertex_shader.outputs_written & (1 << TGSI_ATTRIB_POS)); - -#if 0 - if( file == NULL ) { - file = fopen( "vs-exec.txt", "wt" ); - } -#endif - -#ifdef DEBUG - memset( &machine, 0, sizeof( machine ) ); -#endif - - /* init machine state */ - tgsi_exec_machine_init(&machine, - draw->vertex_shader.tokens, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/ ); - - /* Consts does not require 16 byte alignment. */ - machine.Consts = (float (*)[4]) draw->mapped_constants; - - machine.Inputs = ALIGN16_ASSIGN(inputs); - machine.Outputs = ALIGN16_ASSIGN(outputs); - - -#if 0 - { - unsigned attr; - for (attr = 0; attr < 16; attr++) { - if (draw->vertex_shader.inputs_read & (1 << attr)) { - unsigned buf = draw->vertex_element[attr].vertex_buffer_index; - fprintf(file, "attr %d: buf_off %d src_off %d pitch %d\n", - attr, - draw->vertex_buffer[buf].buffer_offset, - draw->vertex_element[attr].src_offset, - draw->vertex_buffer[buf].pitch); - } - } - } -#endif - - /* load machine inputs */ - for (j = 0; j < count; j++) { - unsigned attr; - for (attr = 0; attr < 16; attr++) { - if (draw->vertex_shader.inputs_read & (1 << attr)) { - unsigned buf = draw->vertex_element[attr].vertex_buffer_index; - const void *src - = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] - + draw->vertex_buffer[buf].buffer_offset - + draw->vertex_element[attr].src_offset - + elts[j] * draw->vertex_buffer[buf].pitch); - float p[4]; - - fetch_attrib4(src, draw->vertex_element[attr].src_format, p); - - machine.Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ - machine.Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ - machine.Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ - machine.Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ -#if 0 - fprintf(file, "Input vertex %d: attr %d: %f %f %f %f\n", - j, attr, p[0], p[1], p[2], p[3]); - fflush( file ); -#endif - } - } - } - -#if 0 - printf("Vertex shader Constants:\n"); - { - int i; - for (i = 0; i < 4; i++) { - printf(" %d: %f %f %f %f\n", i, - machine.Consts[i][0], - machine.Consts[i][1], - machine.Consts[i][2], - machine.Consts[i][3]); - } - } -#endif - - /* run shader */ - if( draw->vertex_shader.executable != NULL ) { -#if defined(USE_X86_ASM) || defined(SLANG_X86) - sse2_function func = (sse2_function) draw->vertex_shader.executable; - func( - machine.Inputs, - machine.Outputs, - machine.Consts, - machine.Temps ); -#else - assert( 0 ); -#endif - } - else { - tgsi_exec_machine_run( &machine ); - } - -#if 0 - for (i = 0; i < 4; i++) { - fprintf(file, "VS result: %f %f %f %f\n", - machine.Outputs[0].xyzw[0].f[i], - machine.Outputs[0].xyzw[1].f[i], - machine.Outputs[0].xyzw[2].f[i], - machine.Outputs[0].xyzw[3].f[i]); - } - fflush( file ); -#endif - - /* store machine results */ - for (j = 0; j < count; j++) { - unsigned slot; - float x, y, z, w; - - /* Handle attr[0] (position) specially: */ - x = vOut[j]->clip[0] = machine.Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine.Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine.Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine.Outputs[0].xyzw[3].f[j]; - - vOut[j]->clipmask = compute_clipmask(x, y, z, w) | draw->user_clipmask; - vOut[j]->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; -#if 0 - fprintf(file, "Vert %d: wincoord: %f %f %f %f\n", j, - vOut[j]->data[0][0], - vOut[j]->data[0][1], - vOut[j]->data[0][2], - vOut[j]->data[0][3]); - fflush( file ); -#endif - - /* remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->vertex_info.num_attribs; slot++) { - vOut[j]->data[slot][0] = machine.Outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine.Outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine.Outputs[slot].xyzw[3].f[j]; -#if 0 - fprintf(file, "output attrib slot %d: %f %f %f %f vert %p\n", - slot, - vOut[j]->data[slot][0], - vOut[j]->data[slot][1], - vOut[j]->data[slot][2], - vOut[j]->data[slot][3], vOut[j]); -#endif - } - } /* loop over vertices */ -} - - -/** - * Called by the draw module when the vertx cache needs to be flushed. - * This involves running the vertex shader. - */ -static void transform_vertices( struct draw_context *draw ) -{ - unsigned i, j; - - /* run vertex shader on vertex cache entries, four per invokation */ - for (i = 0; i < draw->vs.queue_nr; i += 4) { - struct vertex_header *dests[4]; - unsigned elts[4]; - int n; - - for (j = 0; j < 4; j++) { - elts[j] = draw->vs.queue[i + j].elt; - dests[j] = draw->vs.queue[i + j].dest; - } - - n = MIN2(4, draw->vs.queue_nr - i); - assert(n > 0); - assert(n <= 4); - - run_vertex_program(draw, elts, n, dests); - } - - draw->vs.queue_nr = 0; -} - - void draw_flush( struct draw_context *draw ) { struct draw_stage *first = draw->pipeline.first; @@ -358,7 +65,7 @@ void draw_flush( struct draw_context *draw ) /* Make sure all vertices are available: */ - transform_vertices(draw); + draw_vertex_cache_validate(draw); switch (draw->reduced_prim) { case RP_TRI: @@ -385,24 +92,12 @@ void draw_flush( struct draw_context *draw ) } draw->pq.queue_nr = 0; - draw->vcache.referenced = 0; - draw->vcache.overflow = 0; -} - -void draw_invalidate_vcache( struct draw_context *draw ) -{ - unsigned i; - - assert(draw->pq.queue_nr == 0); - assert(draw->vs.queue_nr == 0); - assert(draw->vcache.referenced == 0); - - for (i = 0; i < Elements( draw->vcache.idx ); i++) - draw->vcache.idx[i] = ~0; + draw_vertex_cache_unreference( draw ); } + /* Return a pointer to a freshly queued primitive header. Ensure that * there is room in the vertex cache for a maximum of "nr_verts" new * vertices. Flush primitive and/or vertex queues if necessary to @@ -412,76 +107,15 @@ static struct prim_header *get_queued_prim( struct draw_context *draw, unsigned nr_verts ) { if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH || - draw->vcache.overflow + nr_verts >= VCACHE_OVERFLOW) + draw_vertex_cache_check_space( draw, nr_verts )) + { draw_flush( draw ); - - /* The vs queue is sized so that this can never happen: - */ - assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); - - return &draw->pq.queue[draw->pq.queue_nr++]; -} - - -/* Check if vertex is in cache, otherwise add it. It won't go through - * VS yet, not until there is a flush operation or the VS queue fills up. - */ -static struct vertex_header *get_vertex( struct draw_context *draw, - unsigned i ) -{ - unsigned slot = (i + (i>>5)) & 31; - - /* Cache miss? - */ - if (draw->vcache.idx[slot] != i) { - - /* If slot is in use, use the overflow area: - */ - if (draw->vcache.referenced & (1 << slot)) - slot = VCACHE_SIZE + draw->vcache.overflow++; - else - draw->vcache.referenced |= (1 << slot); /* slot now in use */ - - draw->vcache.idx[slot] = i; - - /* Add to vertex shader queue: - */ - draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; - draw->vs.queue[draw->vs.queue_nr].elt = i; - draw->vs.queue_nr++; - - /* Need to set the vertex's edge flag here. If we're being called - * by do_ef_triangle(), that function needs edge flag info! - */ - draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ } - return draw->vcache.vertex[slot]; -} - - -static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const unsigned *elts = (const unsigned *) draw->mapped_elts; - return get_vertex( draw, elts[i] ); -} - - -static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const ushort *elts = (const ushort *) draw->mapped_elts; - return get_vertex( draw, elts[i] ); + return &draw->pq.queue[draw->pq.queue_nr++]; } -static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const ubyte *elts = (const ubyte *) draw->mapped_elts; - return get_vertex( draw, elts[i] ); -} static void do_point( struct draw_context *draw, @@ -536,7 +170,7 @@ static void do_ef_triangle( struct draw_context *draw, struct vertex_header *v0 = draw->get_vertex( draw, i0 ); struct vertex_header *v1 = draw->get_vertex( draw, i1 ); struct vertex_header *v2 = draw->get_vertex( draw, i2 ); - + prim->reset_line_stipple = reset_stipple; prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | @@ -719,42 +353,6 @@ draw_set_prim( struct draw_context *draw, unsigned prim ) } -/** - * Tell the drawing context about the index/element buffer to use - * (ala glDrawElements) - * If no element buffer is to be used (i.e. glDrawArrays) then this - * should be called with eltSize=0 and elements=NULL. - * - * \param draw the drawing context - * \param eltSize size of each element (1, 2 or 4 bytes) - * \param elements the element buffer ptr - */ -void -draw_set_mapped_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ) -{ - /* choose the get_vertex() function to use */ - switch (eltSize) { - case 0: - draw->get_vertex = get_vertex; - break; - case 1: - draw->get_vertex = get_ubyte_elt_vertex; - break; - case 2: - draw->get_vertex = get_ushort_elt_vertex; - break; - case 4: - draw->get_vertex = get_uint_elt_vertex; - break; - default: - assert(0); - } - draw->mapped_elts = elements; - draw->eltSize = eltSize; -} - - /** * Tell drawing context where to find mapped vertex buffers. */ diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index c104aa1143..5d22e01d5c 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -61,4 +61,23 @@ struct vertex_info +struct draw_context; + +extern int draw_vertex_cache_check_space( struct draw_context *draw, + unsigned nr_verts ); + +extern void draw_vertex_cache_validate( struct draw_context *draw ); +extern void draw_vertex_cache_invalidate( struct draw_context *draw ); +extern void draw_vertex_cache_unreference( struct draw_context *draw ); + +extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); + +struct tgsi_exec_machine; + +extern void draw_vertex_fetch( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ); + + #endif /* DRAW_VERTEX_H */ diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c new file mode 100644 index 0000000000..f1b0cb14bd --- /dev/null +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -0,0 +1,175 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw_private.h" +#include "draw_context.h" +#include "draw_vertex.h" + + +void draw_vertex_cache_invalidate( struct draw_context *draw ) +{ + unsigned i; + + assert(draw->pq.queue_nr == 0); + assert(draw->vs.queue_nr == 0); + assert(draw->vcache.referenced == 0); + + for (i = 0; i < Elements( draw->vcache.idx ); i++) + draw->vcache.idx[i] = ~0; +} + + +/* Check if vertex is in cache, otherwise add it. It won't go through + * VS yet, not until there is a flush operation or the VS queue fills up. + */ +static struct vertex_header *get_vertex( struct draw_context *draw, + unsigned i ) +{ + unsigned slot = (i + (i>>5)) & 31; + + /* Cache miss? + */ + if (draw->vcache.idx[slot] != i) { + + /* If slot is in use, use the overflow area: + */ + if (draw->vcache.referenced & (1 << slot)) + slot = VCACHE_SIZE + draw->vcache.overflow++; + else + draw->vcache.referenced |= (1 << slot); /* slot now in use */ + + draw->vcache.idx[slot] = i; + + /* Add to vertex shader queue: + */ + draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; + draw->vs.queue[draw->vs.queue_nr].elt = i; + draw->vs.queue_nr++; + + /* Need to set the vertex's edge flag here. If we're being called + * by do_ef_triangle(), that function needs edge flag info! + */ + draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ + } + + return draw->vcache.vertex[slot]; +} + + +static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const unsigned *elts = (const unsigned *) draw->mapped_elts; + return get_vertex( draw, elts[i] ); +} + + +static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const ushort *elts = (const ushort *) draw->mapped_elts; + return get_vertex( draw, elts[i] ); +} + + +static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const ubyte *elts = (const ubyte *) draw->mapped_elts; + return get_vertex( draw, elts[i] ); +} + + + +void draw_vertex_cache_validate( struct draw_context *draw ) +{ + draw_vertex_shader_queue_flush( draw ); +} + +void draw_vertex_cache_unreference( struct draw_context *draw ) +{ + draw->vcache.referenced = 0; + draw->vcache.overflow = 0; +} + + +int draw_vertex_cache_check_space( struct draw_context *draw, + unsigned nr_verts ) +{ + if (draw->vcache.overflow + nr_verts < VCACHE_OVERFLOW) { + /* The vs queue is sized so that this can never happen: + */ + assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); + return TRUE; + } + else + return FALSE; +} + + + +/** + * Tell the drawing context about the index/element buffer to use + * (ala glDrawElements) + * If no element buffer is to be used (i.e. glDrawArrays) then this + * should be called with eltSize=0 and elements=NULL. + * + * \param draw the drawing context + * \param eltSize size of each element (1, 2 or 4 bytes) + * \param elements the element buffer ptr + */ +void +draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ) +{ + /* choose the get_vertex() function to use */ + switch (eltSize) { + case 0: + draw->get_vertex = get_vertex; + break; + case 1: + draw->get_vertex = get_ubyte_elt_vertex; + break; + case 2: + draw->get_vertex = get_ushort_elt_vertex; + break; + case 4: + draw->get_vertex = get_uint_elt_vertex; + break; + default: + assert(0); + } + draw->mapped_elts = elements; + draw->eltSize = eltSize; +} + diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c new file mode 100644 index 0000000000..0dbbdf17f2 --- /dev/null +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -0,0 +1,101 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw_private.h" +#include "draw_context.h" +#include "draw_vertex.h" + +#include "pipe/tgsi/exec/tgsi_core.h" + +/** + * Fetch a float[4] vertex attribute from memory, doing format/type + * conversion as needed. + * XXX this might be a temporary thing. + */ +static void +fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) +{ + /* defaults */ + attrib[1] = 0.0; + attrib[2] = 0.0; + attrib[3] = 1.0; + switch (format) { + case PIPE_FORMAT_R32G32B32A32_FLOAT: + attrib[3] = ((float *) ptr)[3]; + /* fall-through */ + case PIPE_FORMAT_R32G32B32_FLOAT: + attrib[2] = ((float *) ptr)[2]; + /* fall-through */ + case PIPE_FORMAT_R32G32_FLOAT: + attrib[1] = ((float *) ptr)[1]; + /* fall-through */ + case PIPE_FORMAT_R32_FLOAT: + attrib[0] = ((float *) ptr)[0]; + break; + default: + assert(0); + } +} + +void draw_vertex_fetch( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + unsigned j; + + + /* load machine inputs */ + for (j = 0; j < count; j++) { + unsigned attr; + for (attr = 0; attr < 16; attr++) { + if (draw->vertex_shader.inputs_read & (1 << attr)) { + unsigned buf = draw->vertex_element[attr].vertex_buffer_index; + const void *src + = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[attr].src_offset + + elts[j] * draw->vertex_buffer[buf].pitch); + float p[4]; + + fetch_attrib4(src, draw->vertex_element[attr].src_format, p); + + machine->Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ + machine->Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ + machine->Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ + machine->Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ + } + } + } +} + diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c new file mode 100644 index 0000000000..8effc74cbe --- /dev/null +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -0,0 +1,203 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "draw_private.h" +#include "draw_context.h" +#include "draw_vertex.h" + +#include "pipe/tgsi/exec/tgsi_core.h" + +static INLINE unsigned +compute_clipmask(float cx, float cy, float cz, float cw) +{ + unsigned mask = 0; + + if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT; + if ( cx + cw < 0) mask |= CLIP_LEFT_BIT; + if (-cy + cw < 0) mask |= CLIP_TOP_BIT; + if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT; + if (-cz + cw < 0) mask |= CLIP_FAR_BIT; + if ( cz + cw < 0) mask |= CLIP_NEAR_BIT; + + return mask; +} + + + + +#if !defined(XSTDCALL) +#if defined(WIN32) +#define XSTDCALL __stdcall +#else +#define XSTDCALL +#endif +#endif + +#if defined(USE_X86_ASM) || defined(SLANG_X86) +typedef void (XSTDCALL *sse2_function)( + const struct tgsi_exec_vector *input, + struct tgsi_exec_vector *output, + float (*constant)[4], + struct tgsi_exec_vector *temporary ); +#endif + +/** + * Transform vertices with the current vertex program/shader + * Up to four vertices can be shaded at a time. + * \param vbuffer the input vertex data + * \param elts indexes of four input vertices + * \param count number of vertices to shade [1..4] + * \param vOut array of pointers to four output vertices + */ +static void +run_vertex_program(struct draw_context *draw, + unsigned elts[4], unsigned count, + struct vertex_header *vOut[]) +{ + struct tgsi_exec_machine machine; + unsigned int j; + + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); + assert(draw->vertex_shader.outputs_written & (1 << TGSI_ATTRIB_POS)); + +#ifdef DEBUG + memset( &machine, 0, sizeof( machine ) ); +#endif + + /* init machine state */ + tgsi_exec_machine_init(&machine, + draw->vertex_shader.tokens, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/ ); + + /* Consts does not require 16 byte alignment. */ + machine.Consts = (float (*)[4]) draw->mapped_constants; + + machine.Inputs = ALIGN16_ASSIGN(inputs); + machine.Outputs = ALIGN16_ASSIGN(outputs); + + draw_vertex_fetch( draw, &machine, elts, count ); + + + /* run shader */ + if( draw->vertex_shader.executable != NULL ) { +#if defined(USE_X86_ASM) || defined(SLANG_X86) + sse2_function func = (sse2_function) draw->vertex_shader.executable; + func( + machine.Inputs, + machine.Outputs, + machine.Consts, + machine.Temps ); +#else + assert( 0 ); +#endif + } + else { + tgsi_exec_machine_run( &machine ); + } + + + /* store machine results */ + for (j = 0; j < count; j++) { + unsigned slot; + float x, y, z, w; + + /* Handle attr[0] (position) specially: */ + x = vOut[j]->clip[0] = machine.Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine.Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine.Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine.Outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(x, y, z, w) | draw->user_clipmask; + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + + /* remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->vertex_info.num_attribs; slot++) { + vOut[j]->data[slot][0] = machine.Outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine.Outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine.Outputs[slot].xyzw[3].f[j]; + } + } /* loop over vertices */ +} + + +/** + * Called by the draw module when the vertx cache needs to be flushed. + * This involves running the vertex shader. + */ +void draw_vertex_shader_queue_flush( struct draw_context *draw ) +{ + unsigned i, j; + + /* run vertex shader on vertex cache entries, four per invokation */ + for (i = 0; i < draw->vs.queue_nr; i += 4) { + struct vertex_header *dests[4]; + unsigned elts[4]; + int n; + + for (j = 0; j < 4; j++) { + elts[j] = draw->vs.queue[i + j].elt; + dests[j] = draw->vs.queue[i + j].dest; + } + + n = MIN2(4, draw->vs.queue_nr - i); + assert(n > 0); + assert(n <= 4); + + run_vertex_program(draw, elts, n, dests); + } + + draw->vs.queue_nr = 0; +} + diff --git a/src/mesa/sources b/src/mesa/sources index 53969774f9..be24f93f74 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -164,6 +164,9 @@ DRAW_SOURCES = \ pipe/draw/draw_offset.c \ pipe/draw/draw_prim.c \ pipe/draw/draw_twoside.c \ + pipe/draw/draw_vertex_cache.c \ + pipe/draw/draw_vertex_fetch.c \ + pipe/draw/draw_vertex_shader.c \ pipe/draw/draw_unfilled.c TGSIEXEC_SOURCES = \ -- cgit v1.2.3 From 5c337508feedfa35ba3f534bf212f08265db2022 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 29 Aug 2007 21:40:18 +0100 Subject: Fix ffs(). --- src/mesa/pipe/p_util.h | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) mode change 100644 => 100755 src/mesa/pipe/p_util.h (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h old mode 100644 new mode 100755 index 3eff3328c7..436bda2139 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -39,10 +39,10 @@ #define Elements(x) sizeof(x)/sizeof(*(x)) /** - * Return pointer aligned to next multiple of 16 bytes. + * Return a pointer aligned to next multiple of 16 bytes. */ static INLINE void * -align16(void *unaligned) +align16( void *unaligned ) { union { void *p; @@ -59,7 +59,12 @@ static INLINE unsigned ffs( unsigned u ) { unsigned i; + if( u == 0 ) { + return 0; + } + __asm bsf eax, [u] + __asm inc eax __asm mov [i], eax return i; -- cgit v1.2.3 From 529c411e4095eeb9be501a6fe657a3a196c3e448 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 29 Aug 2007 22:10:00 +0100 Subject: Break GL dependencies. --- src/mesa/pipe/tgsi/exec/tgsi_build.c | 210 +++++++++++++++++------------------ 1 file changed, 105 insertions(+), 105 deletions(-) mode change 100644 => 100755 src/mesa/pipe/tgsi/exec/tgsi_build.c (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.c b/src/mesa/pipe/tgsi/exec/tgsi_build.c old mode 100644 new mode 100755 index d6f8af656a..20e4cf17f0 --- a/src/mesa/pipe/tgsi/exec/tgsi_build.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_build.c @@ -393,7 +393,7 @@ immediate_grow( struct tgsi_immediate_float32 tgsi_build_immediate_float32( - GLfloat value, + float value, struct tgsi_immediate *immediate, struct tgsi_header *header ) { @@ -406,14 +406,14 @@ tgsi_build_immediate_float32( return immediate_float32; } -GLuint +unsigned tgsi_build_full_immediate( const struct tgsi_full_immediate *full_imm, struct tgsi_token *tokens, struct tgsi_header *header, - GLuint maxsize ) + unsigned maxsize ) { - GLuint size = 0, i; + unsigned size = 0, i; struct tgsi_immediate *immediate; if( maxsize <= size ) @@ -463,10 +463,10 @@ tgsi_default_instruction( void ) struct tgsi_instruction tgsi_build_instruction( - GLuint opcode, - GLuint saturate, - GLuint num_dst_regs, - GLuint num_src_regs, + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, struct tgsi_header *header ) { struct tgsi_instruction instruction; @@ -503,7 +503,7 @@ struct tgsi_full_instruction tgsi_default_full_instruction( void ) { struct tgsi_full_instruction full_instruction; - GLuint i; + unsigned i; full_instruction.Instruction = tgsi_default_instruction(); full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); @@ -519,15 +519,15 @@ tgsi_default_full_instruction( void ) return full_instruction; } -GLuint +unsigned tgsi_build_full_instruction( const struct tgsi_full_instruction *full_inst, struct tgsi_token *tokens, struct tgsi_header *header, - GLuint maxsize ) + unsigned maxsize ) { - GLuint size = 0; - GLuint i; + unsigned size = 0; + unsigned i; struct tgsi_instruction *instruction; struct tgsi_token *prev_token; @@ -835,10 +835,10 @@ tgsi_default_instruction_ext_nv( void ) union token_u32 { - GLuint u32; + unsigned u32; }; -GLuint +unsigned tgsi_compare_instruction_ext_nv( struct tgsi_instruction_ext_nv a, struct tgsi_instruction_ext_nv b ) @@ -850,16 +850,16 @@ tgsi_compare_instruction_ext_nv( struct tgsi_instruction_ext_nv tgsi_build_instruction_ext_nv( - GLuint precision, - GLuint cond_dst_index, - GLuint cond_flow_index, - GLuint cond_mask, - GLuint cond_swizzle_x, - GLuint cond_swizzle_y, - GLuint cond_swizzle_z, - GLuint cond_swizzle_w, - GLuint cond_dst_update, - GLuint cond_flow_update, + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ) @@ -897,19 +897,19 @@ tgsi_default_instruction_ext_label( void ) return instruction_ext_label; } -GLuint +unsigned tgsi_compare_instruction_ext_label( struct tgsi_instruction_ext_label a, struct tgsi_instruction_ext_label b ) { a.Padding = b.Padding = 0; a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; } struct tgsi_instruction_ext_label tgsi_build_instruction_ext_label( - GLuint label, + unsigned label, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ) @@ -938,19 +938,19 @@ tgsi_default_instruction_ext_texture( void ) return instruction_ext_texture; } -GLuint +unsigned tgsi_compare_instruction_ext_texture( struct tgsi_instruction_ext_texture a, struct tgsi_instruction_ext_texture b ) { a.Padding = b.Padding = 0; a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; } struct tgsi_instruction_ext_texture tgsi_build_instruction_ext_texture( - GLuint texture, + unsigned texture, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ) @@ -987,15 +987,15 @@ tgsi_default_src_register( void ) struct tgsi_src_register tgsi_build_src_register( - GLuint file, - GLuint swizzle_x, - GLuint swizzle_y, - GLuint swizzle_z, - GLuint swizzle_w, - GLuint negate, - GLuint indirect, - GLuint dimension, - GLint index, + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, struct tgsi_instruction *instruction, struct tgsi_header *header ) { @@ -1061,42 +1061,42 @@ tgsi_default_src_register_ext_swz( void ) return src_register_ext_swz; } -GLuint +unsigned tgsi_compare_src_register_ext_swz( struct tgsi_src_register_ext_swz a, struct tgsi_src_register_ext_swz b ) { a.Padding = b.Padding = 0; a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; } struct tgsi_src_register_ext_swz tgsi_build_src_register_ext_swz( - GLuint ext_swizzle_x, - GLuint ext_swizzle_y, - GLuint ext_swizzle_z, - GLuint ext_swizzle_w, - GLuint negate_x, - GLuint negate_y, - GLuint negate_z, - GLuint negate_w, - GLuint ext_divide, + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + unsigned ext_divide, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ) { struct tgsi_src_register_ext_swz src_register_ext_swz; - assert (ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE); - assert (ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE); - assert (ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE); - assert (ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE); - assert (negate_x <= 1); - assert (negate_y <= 1); - assert (negate_z <= 1); - assert (negate_w <= 1); - assert (ext_divide <= TGSI_EXTSWIZZLE_ONE); + assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE ); + assert( negate_x <= 1 ); + assert( negate_y <= 1 ); + assert( negate_z <= 1 ); + assert( negate_w <= 1 ); + assert( ext_divide <= TGSI_EXTSWIZZLE_ONE ); src_register_ext_swz = tgsi_default_src_register_ext_swz(); src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; @@ -1132,34 +1132,34 @@ tgsi_default_src_register_ext_mod( void ) return src_register_ext_mod; } -GLuint +unsigned tgsi_compare_src_register_ext_mod( struct tgsi_src_register_ext_mod a, struct tgsi_src_register_ext_mod b ) { a.Padding = b.Padding = 0; a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; } struct tgsi_src_register_ext_mod tgsi_build_src_register_ext_mod( - GLuint complement, - GLuint bias, - GLuint scale_2x, - GLuint absolute, - GLuint negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) { struct tgsi_src_register_ext_mod src_register_ext_mod; - assert (complement <= 1); - assert (bias <= 1); - assert (scale_2x <= 1); - assert (absolute <= 1); - assert (negate <= 1); + assert( complement <= 1 ); + assert( bias <= 1 ); + assert( scale_2x <= 1 ); + assert( absolute <= 1 ); + assert( negate <= 1 ); src_register_ext_mod = tgsi_default_src_register_ext_mod(); src_register_ext_mod.Complement = complement; @@ -1190,8 +1190,8 @@ tgsi_default_dimension( void ) struct tgsi_dimension tgsi_build_dimension( - GLuint indirect, - GLuint index, + unsigned indirect, + unsigned index, struct tgsi_instruction *instruction, struct tgsi_header *header ) { @@ -1224,17 +1224,17 @@ tgsi_default_dst_register( void ) struct tgsi_dst_register tgsi_build_dst_register( - GLuint file, - GLuint mask, - GLint index, + unsigned file, + unsigned mask, + int index, struct tgsi_instruction *instruction, struct tgsi_header *header ) { struct tgsi_dst_register dst_register; - assert (file <= TGSI_FILE_IMMEDIATE); - assert (mask <= TGSI_WRITEMASK_XYZW); - assert (index >= -32768 && index <= 32767); + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( mask <= TGSI_WRITEMASK_XYZW ); + assert( index >= -32768 && index <= 32767 ); dst_register = tgsi_default_dst_register(); dst_register.File = file; @@ -1278,36 +1278,36 @@ tgsi_default_dst_register_ext_concode( void ) return dst_register_ext_concode; } -GLuint +unsigned tgsi_compare_dst_register_ext_concode( struct tgsi_dst_register_ext_concode a, struct tgsi_dst_register_ext_concode b ) { a.Padding = b.Padding = 0; a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; } struct tgsi_dst_register_ext_concode tgsi_build_dst_register_ext_concode( - GLuint cc, - GLuint swizzle_x, - GLuint swizzle_y, - GLuint swizzle_z, - GLuint swizzle_w, - GLint index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) { struct tgsi_dst_register_ext_concode dst_register_ext_concode; - assert (cc <= TGSI_CC_FL); - assert (swizzle_x <= TGSI_SWIZZLE_W); - assert (swizzle_y <= TGSI_SWIZZLE_W); - assert (swizzle_z <= TGSI_SWIZZLE_W); - assert (swizzle_w <= TGSI_SWIZZLE_W); - assert (index >= -32768 && index <= 32767); + assert( cc <= TGSI_CC_FL ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( index >= -32768 && index <= 32767 ); dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); dst_register_ext_concode.CondMask = cc; @@ -1336,26 +1336,26 @@ tgsi_default_dst_register_ext_modulate( void ) return dst_register_ext_modulate; } -GLuint +unsigned tgsi_compare_dst_register_ext_modulate( struct tgsi_dst_register_ext_modulate a, struct tgsi_dst_register_ext_modulate b ) { a.Padding = b.Padding = 0; a.Extended = b.Extended = 0; - return *(GLuint *) &a != *(GLuint *) &b; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; } struct tgsi_dst_register_ext_modulate tgsi_build_dst_register_ext_modulate( - GLuint modulate, - struct tgsi_token *prev_token, + unsigned modulate, + struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ) { struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - assert (modulate <= TGSI_MODULATE_EIGHTH); + assert( modulate <= TGSI_MODULATE_EIGHTH ); dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); dst_register_ext_modulate.Modulate = modulate; -- cgit v1.2.3 From 86fe5f235b174e2aedd02bc14a94df818bc3ea74 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 29 Aug 2007 22:22:02 +0100 Subject: Break GL dependencies. --- src/mesa/pipe/tgsi/exec/tgsi_build.h | 0 src/mesa/pipe/tgsi/exec/tgsi_dump.c | 64 ++++++++++++++++++------------------ 2 files changed, 32 insertions(+), 32 deletions(-) mode change 100644 => 100755 src/mesa/pipe/tgsi/exec/tgsi_build.h mode change 100644 => 100755 src/mesa/pipe/tgsi/exec/tgsi_dump.c (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.h b/src/mesa/pipe/tgsi/exec/tgsi_build.h old mode 100644 new mode 100755 diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c old mode 100644 new mode 100755 index 622d617e29..e6e99d9d75 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -3,15 +3,15 @@ struct text_dump { - FILE *file; - GLuint tabs; + FILE *file; + unsigned tabs; }; static void text_dump_write( struct text_dump *dump, const void *buffer, - GLuint size ) + unsigned size ) { fwrite( buffer, size, 1, dump->file ); } @@ -21,14 +21,14 @@ text_dump_str( struct text_dump *dump, const char *str ) { - GLuint i; - GLuint len = strlen( str ); + unsigned i; + size_t len = strlen( str ); for( i = 0; i < len; i++ ) { text_dump_write( dump, &str[i], 1 ); if( str[i] == '\n' ) { - GLuint i; + unsigned i; for( i = 0; i < dump->tabs; i++ ) { text_dump_write( dump, " ", 4 ); @@ -52,7 +52,7 @@ text_dump_chr( static void text_dump_uix( struct text_dump *dump, - const GLuint ui) + const unsigned ui ) { char str[36]; @@ -63,7 +63,7 @@ text_dump_uix( static void text_dump_uid( struct text_dump *dump, - const GLuint ui ) + const unsigned ui ) { char str[16]; @@ -74,7 +74,7 @@ text_dump_uid( static void text_dump_sid( struct text_dump *dump, - const GLint si ) + const int si ) { char str[16]; @@ -85,7 +85,7 @@ text_dump_sid( static void text_dump_flt( struct text_dump *dump, - const GLfloat f ) + const float f ) { char str[48]; @@ -96,9 +96,9 @@ text_dump_flt( static void text_dump_enum( struct text_dump *dump, - const GLuint e, + const unsigned e, const char **enums, - const GLuint enums_count ) + const unsigned enums_count ) { if( e >= enums_count ) { text_dump_uid( dump, e ); @@ -667,8 +667,8 @@ static void dump_declaration_verbose( struct text_dump *dump, struct tgsi_full_declaration *decl, - GLuint ignored, - GLuint deflt, + unsigned ignored, + unsigned deflt, struct tgsi_full_declaration *fd ) { TXT( "\nFile : " ); @@ -749,7 +749,7 @@ dump_immediate_short( struct text_dump *dump, struct tgsi_full_immediate *imm ) { - GLuint i; + unsigned i; TXT( "\nIMM " ); ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); @@ -776,9 +776,9 @@ static void dump_immediate_verbose( struct text_dump *dump, struct tgsi_full_immediate *imm, - GLuint ignored ) + unsigned ignored ) { - GLuint i; + unsigned i; TXT( "\nDataType : " ); ENM( imm->Immediate.DataType, TGSI_IMMS ); @@ -805,10 +805,10 @@ static void dump_instruction_short( struct text_dump *dump, struct tgsi_full_instruction *inst, - GLuint instno ) + unsigned instno ) { - GLuint i; - GLboolean first_reg = GL_TRUE; + unsigned i; + boolean first_reg = TRUE; CHR( '\n' ); UID( instno ); @@ -858,7 +858,7 @@ dump_instruction_short( } } - first_reg = GL_FALSE; + first_reg = FALSE; } for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { @@ -900,7 +900,7 @@ dump_instruction_short( CHR( '|' ); } - first_reg = GL_FALSE; + first_reg = FALSE; } switch( inst->Instruction.Opcode ) { @@ -916,11 +916,11 @@ static void dump_instruction_verbose( struct text_dump *dump, struct tgsi_full_instruction *inst, - GLuint ignored, - GLuint deflt, + unsigned ignored, + unsigned deflt, struct tgsi_full_instruction *fi ) { - GLuint i; + unsigned i; TXT( "\nOpcode : " ); ENM( inst->Instruction.Opcode, TGSI_OPCODES ); @@ -1259,29 +1259,29 @@ dump_instruction_verbose( void tgsi_dump( const struct tgsi_token *tokens, - GLuint flags ) + unsigned flags ) { struct text_dump _dump; struct text_dump *dump = &_dump; struct tgsi_parse_context parse; struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; - GLuint verbose = flags & TGSI_DUMP_VERBOSE; - GLuint ignored = !(flags & TGSI_DUMP_NO_IGNORED); - GLuint deflt = !(flags & TGSI_DUMP_NO_DEFAULT); - GLuint instno = 0; + unsigned verbose = flags & TGSI_DUMP_VERBOSE; + unsigned ignored = !(flags & TGSI_DUMP_NO_IGNORED); + unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); + unsigned instno = 0; { #if 0 - static GLuint counter = 0; + static unsigned counter = 0; char buffer[64]; sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); dump->file = fopen( buffer, "wt" ); #else dump->file = stderr; - dump->tabs = 0; #endif + dump->tabs = 0; } /* sanity check */ -- cgit v1.2.3 From 6468e68466bfdf93cafafc78d37a156e96af229c Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 29 Aug 2007 22:24:42 +0100 Subject: Break GL dependencies. --- src/mesa/pipe/tgsi/exec/tgsi_parse.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) mode change 100644 => 100755 src/mesa/pipe/tgsi/exec/tgsi_parse.c (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_parse.c b/src/mesa/pipe/tgsi/exec/tgsi_parse.c old mode 100644 new mode 100755 index a6724d762c..1096afffad --- a/src/mesa/pipe/tgsi/exec/tgsi_parse.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_parse.c @@ -12,8 +12,9 @@ void tgsi_full_token_free( union tgsi_full_token *full_token ) { - if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) + if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { free( full_token->FullImmediate.u.Pointer ); + } } unsigned @@ -49,7 +50,7 @@ tgsi_parse_free( tgsi_full_token_free( &ctx->FullToken ); } -GLuint +boolean tgsi_parse_end_of_tokens( struct tgsi_parse_context *ctx ) { -- cgit v1.2.3 From a6e2b76fb570dea8f4a1384e4a9df36f6b81ca84 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 29 Aug 2007 22:26:23 +0100 Subject: Break GL dependencies. --- src/mesa/pipe/tgsi/exec/tgsi_parse.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) mode change 100644 => 100755 src/mesa/pipe/tgsi/exec/tgsi_parse.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_parse.h b/src/mesa/pipe/tgsi/exec/tgsi_parse.h old mode 100644 new mode 100755 index 56a8c254c2..9372da8d5d --- a/src/mesa/pipe/tgsi/exec/tgsi_parse.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_parse.h @@ -105,7 +105,7 @@ void tgsi_parse_free( struct tgsi_parse_context *ctx ); -unsigned +boolean tgsi_parse_end_of_tokens( struct tgsi_parse_context *ctx ); -- cgit v1.2.3 From e62b9241d288da7d9bed28fb85ea17bbb617aa28 Mon Sep 17 00:00:00 2001 From: michal Date: Wed, 29 Aug 2007 22:30:20 +0100 Subject: Add p_compiler.h. --- src/mesa/pipe/tgsi/tgsi_platform.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/tgsi_platform.h b/src/mesa/pipe/tgsi/tgsi_platform.h index 553f0b26a8..b98a1f21be 100644 --- a/src/mesa/pipe/tgsi/tgsi_platform.h +++ b/src/mesa/pipe/tgsi/tgsi_platform.h @@ -9,6 +9,7 @@ extern "C" { #include "mtypes.h" #include "prog_instruction.h" #include "program.h" +#include "pipe/p_compiler.h" #if defined __cplusplus } // extern "C" -- cgit v1.2.3 From 2d187672b7e38fcd8f6caa93f35dceb6fa11aa7f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 30 Aug 2007 11:08:03 -0600 Subject: Sketch out per-vertex point size. The code is all in place, but mostly disabled for now: In t_vp_build.c, write the VERT_RESULT_PSIZE register In sp_state_derived.c, need to emit vertex point size if drawing points. In setup_point() use the point size from the vertex. --- src/mesa/pipe/draw/draw_twoside.c | 2 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 13 ++++++++++--- src/mesa/pipe/softpipe/sp_state_derived.c | 6 ++++++ src/mesa/tnl/t_vp_build.c | 14 ++++++++++++++ 4 files changed, 31 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 5a7697d5f8..a05eea41fc 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -36,7 +36,7 @@ struct twoside_stage { struct draw_stage stage; float sign; /**< +1 or -1 */ - const unsigned *lookup; + const unsigned *lookup; /**< vertex attribute positions */ }; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index c2a5a7907d..a7e1c9e017 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -79,6 +79,8 @@ struct setup_stage { float oneoverarea; + const unsigned *lookup; /**< vertex attribute positions */ + struct tgsi_interp_coef coef[TGSI_ATTRIB_MAX]; struct quad_header quad; @@ -884,10 +886,13 @@ static void setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - /*XXX this should be a vertex attrib! */ - const float halfSize = 0.5f * setup->softpipe->setup.point_size; - const boolean round = setup->softpipe->setup.point_smooth; const struct vertex_header *v0 = prim->v[0]; + + const int sizeAttr = setup->lookup[TGSI_ATTRIB_POINTSIZE]; + const float halfSize + = sizeAttr ? (0.5f * v0->data[sizeAttr][0]) + : (0.5f * setup->softpipe->setup.point_size); + const boolean round = setup->softpipe->setup.point_smooth; const float x = v0->data[TGSI_ATTRIB_POS][0]; const float y = v0->data[TGSI_ATTRIB_POS][1]; unsigned slot, j; @@ -1072,5 +1077,7 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) setup->quad.coef = setup->coef; + setup->lookup = softpipe->draw->vertex_info.attrib_to_slot; + return &setup->stage; } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index b0d79eedda..3bbe9c9fd9 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -102,6 +102,12 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) emit_vertex_attr(vinfo, TGSI_ATTRIB_FOG, FORMAT_1F, INTERP_PERSPECTIVE); } + /* point size */ +#if 0 + /* XXX only emit if drawing points or front/back polygon mode is point mode */ + emit_vertex_attr(vinfo, TGSI_ATTRIB_POINTSIZE, FORMAT_4F, INTERP_CONSTANT); +#endif + /* texcoords and varying vars */ for (i = TGSI_ATTRIB_TEX0; i < TGSI_ATTRIB_VAR7; i++) { if (inputsRead & (1 << i)) { diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index eb5e176895..f923b8e132 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -1364,6 +1364,16 @@ static void build_pointsize( struct tnl_program *p ) release_temp(p, ut); } +/** + * Emit constant point size. + */ +static void constant_pointsize( struct tnl_program *p ) +{ + struct ureg state_size = register_param1(p, STATE_POINT_SIZE); + struct ureg out = register_output(p, VERT_RESULT_PSIZ); + emit_op1(p, OPCODE_MOV, out, WRITEMASK_X, state_size); +} + static void build_tnl_program( struct tnl_program *p ) { /* Emit the program, starting with modelviewproject: */ @@ -1392,6 +1402,10 @@ static void build_tnl_program( struct tnl_program *p ) if (p->state->point_attenuated) build_pointsize(p); +#if 0 + else + constant_pointsize(p); +#endif /* Finish up: */ -- cgit v1.2.3 From aaf03b94861cbf5a602863e4542dd1c2e54ba365 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 30 Aug 2007 14:31:42 -0600 Subject: Fix lost edge flags problem when clipping. In emit_poly() we need to compute header.edgeflags from the vertex edge flags. Also need to set header.det so later stages can determine front/back orientation. --- src/mesa/pipe/draw/draw_clip.c | 12 ++++++++++-- src/mesa/pipe/draw/draw_unfilled.c | 6 ++++++ 2 files changed, 16 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index a23b7a2810..1396b60f45 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -177,11 +177,15 @@ static INLINE void do_tri( struct draw_stage *next, static void emit_poly( struct draw_stage *stage, struct vertex_header **inlist, - unsigned n ) + unsigned n, + const struct prim_header *origPrim) { struct prim_header header; unsigned i; + /* later stages may need the determinant, but only the sign matters */ + header.det = origPrim->det; + for (i = 2; i < n; i++) { header.v[0] = inlist[0]; header.v[1] = inlist[i-1]; @@ -194,6 +198,10 @@ static void emit_poly( struct draw_stage *stage, if (i != 2) header.v[0]->edgeflag = 0; if (i != n-1) header.v[2]->edgeflag = 0; + header.edgeflags = ((header.v[0]->edgeflag << 0) | + (header.v[1]->edgeflag << 1) | + (header.v[2]->edgeflag << 2)); + stage->next->tri( stage->next, &header ); header.v[0]->edgeflag = tmp0; @@ -302,7 +310,7 @@ do_clip_tri( struct draw_stage *stage, /* Emit the polygon as triangles to the setup stage: */ if (n >= 3) - emit_poly( stage, inlist, n ); + emit_poly( stage, inlist, n, header ); } diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index 4215b8b228..b0d6f3d065 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -104,6 +104,12 @@ static void lines( struct draw_stage *stage, struct vertex_header *v1 = header->v[1]; struct vertex_header *v2 = header->v[2]; +#if 0 + assert(((header->edgeflags & 0x1) >> 0) == header->v[0]->edgeflag); + assert(((header->edgeflags & 0x2) >> 1) == header->v[1]->edgeflag); + assert(((header->edgeflags & 0x4) >> 2) == header->v[2]->edgeflag); +#endif + if (header->edgeflags & 0x1) line( stage, v0, v1 ); if (header->edgeflags & 0x2) line( stage, v1, v2 ); if (header->edgeflags & 0x4) line( stage, v2, v0 ); -- cgit v1.2.3 From 942b9bc5bc13d959baa86779a7c669cf96659b9a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 30 Aug 2007 14:41:23 -0600 Subject: In draw_flatshade.c use vertex_info->interp_mode[] to choose attribs/colors to cpy. One less dependency on the TGSI_ATTRIB_x flags. This requires setting the vertex_info->interp_mode[] values in the i915 driver and passing them to draw_set_vertex_attributes(). --- src/mesa/pipe/draw/draw_arrays.c | 16 ++++---- src/mesa/pipe/draw/draw_context.h | 2 +- src/mesa/pipe/draw/draw_flatshade.c | 53 ++++++++++----------------- src/mesa/pipe/draw/draw_vertex.h | 9 +++++ src/mesa/pipe/i915simple/i915_state_derived.c | 18 +++++---- src/mesa/pipe/softpipe/sp_context.h | 7 ---- src/mesa/pipe/softpipe/sp_prim_setup.c | 4 ++ src/mesa/pipe/softpipe/sp_state_derived.c | 4 +- 8 files changed, 56 insertions(+), 57 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index b7d06dd5a7..9e219ed43b 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -75,7 +75,7 @@ draw_arrays(struct draw_context *draw, unsigned prim, static INLINE void -emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) +emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, uint interp) { const uint n = vinfo->num_attribs; vinfo->attr_mask |= (1 << vfAttr); @@ -85,8 +85,7 @@ emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) assert(vfAttr < Elements(vinfo->attrib_to_slot)); vinfo->attrib_to_slot[vfAttr] = n - 2; } - /*printf("Vertex slot %d = vfattrib %d\n", n, vfAttr);*/ - /*vinfo->interp_mode[n] = interpMode;*/ + vinfo->interp_mode[n] = interp; vinfo->format[n] = format; vinfo->num_attribs++; @@ -127,7 +126,8 @@ compute_vertex_size(struct vertex_info *vinfo) void draw_set_vertex_attributes( struct draw_context *draw, - const unsigned *slot_to_vf_attr, + const uint *slot_to_vf_attr, + const uint *interp_mode, unsigned nr_attrs ) { struct vertex_info *vinfo = &draw->vertex_info; @@ -140,15 +140,15 @@ draw_set_vertex_attributes( struct draw_context *draw, /* * First three attribs are always the same: header, clip pos, winpos */ - emit_vertex_attr(vinfo, TGSI_ATTRIB_VERTEX_HEADER, FORMAT_1F); - emit_vertex_attr(vinfo, TGSI_ATTRIB_CLIP_POS, FORMAT_4F); - emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F_VIEWPORT); + emit_vertex_attr(vinfo, TGSI_ATTRIB_VERTEX_HEADER, FORMAT_1F, INTERP_NONE); + emit_vertex_attr(vinfo, TGSI_ATTRIB_CLIP_POS, FORMAT_4F, INTERP_LINEAR); + emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F_VIEWPORT, INTERP_LINEAR); /* * Remaining attribs (color, texcoords, etc) */ for (i = 1; i < nr_attrs; i++) { - emit_vertex_attr(vinfo, slot_to_vf_attr[i], FORMAT_4F); + emit_vertex_attr(vinfo, slot_to_vf_attr[i], FORMAT_4F, interp_mode[i]); } compute_vertex_size(vinfo); diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 1bd0e624fe..4c9e64a12d 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -89,7 +89,7 @@ void draw_set_setup_stage( struct draw_context *draw, struct draw_stage *stage ); void draw_set_vertex_attributes( struct draw_context *draw, - const unsigned *attrs, + const uint *attrs, const uint *interp_mode, unsigned nr_attrs ); unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 5a23e10c03..602a8785ad 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -32,20 +32,6 @@ #include "draw_private.h" -struct flatshade_stage { - struct draw_stage stage; - - const unsigned *lookup; -}; - - - -static INLINE struct flatshade_stage *flatshade_stage( struct draw_stage *stage ) -{ - return (struct flatshade_stage *)stage; -} - - static void flatshade_begin( struct draw_stage *stage ) { stage->next->begin( stage->next ); @@ -69,13 +55,16 @@ static INLINE void copy_colors( struct draw_stage *stage, struct vertex_header *dst, const struct vertex_header *src ) { - const struct flatshade_stage *flatshade = flatshade_stage(stage); - const unsigned *lookup = flatshade->lookup; - - copy_attr( lookup[TGSI_ATTRIB_COLOR0], dst, src ); - copy_attr( lookup[TGSI_ATTRIB_COLOR1], dst, src ); - copy_attr( lookup[TGSI_ATTRIB_BFC0], dst, src ); - copy_attr( lookup[TGSI_ATTRIB_BFC1], dst, src ); + const uint num_attribs = stage->draw->vertex_info.num_attribs; + const uint *interp_mode = stage->draw->vertex_info.interp_mode; + uint i; + + /* Look for constant/flat attribs and duplicate from src to dst vertex */ + for (i = 1; i < num_attribs - 2; i++) { + if (interp_mode[i + 2] == INTERP_CONSTANT) { + copy_attr( i, dst, src ); + } + } } @@ -142,22 +131,20 @@ static void flatshade_reset_stipple_counter( struct draw_stage *stage ) */ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) { - struct flatshade_stage *flatshade = CALLOC_STRUCT(flatshade_stage); + struct draw_stage *flatshade = CALLOC_STRUCT(draw_stage); draw_alloc_tmps( &flatshade->stage, 2 ); - flatshade->stage.draw = draw; - flatshade->stage.next = NULL; - flatshade->stage.begin = flatshade_begin; - flatshade->stage.point = flatshade_point; - flatshade->stage.line = flatshade_line; - flatshade->stage.tri = flatshade_tri; - flatshade->stage.end = flatshade_end; - flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; - - flatshade->lookup = draw->vertex_info.attrib_to_slot; + flatshade->draw = draw; + flatshade->next = NULL; + flatshade->begin = flatshade_begin; + flatshade->point = flatshade_point; + flatshade->line = flatshade_line; + flatshade->tri = flatshade_tri; + flatshade->end = flatshade_end; + flatshade->reset_stipple_counter = flatshade_reset_stipple_counter; - return &flatshade->stage; + return flatshade; } diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 5d22e01d5c..6778df30d1 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -47,6 +47,15 @@ #define FORMAT_4UB 5 +enum interp_mode { + INTERP_NONE, /**< never interpolate vertex header info */ + INTERP_CONSTANT, + INTERP_LINEAR, + INTERP_PERSPECTIVE +}; + + + struct vertex_info { uint num_attribs; diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 9bd0af0f53..e156766a93 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -36,12 +36,13 @@ static INLINE void -emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format) +emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, uint interp) { const uint n = vinfo->num_attribs; vinfo->attr_mask |= (1 << vfAttr); vinfo->slot_to_attrib[n] = vfAttr; vinfo->format[n] = format; + vinfo->interp_mode[n] = interp; vinfo->num_attribs++; } @@ -89,25 +90,27 @@ compute_vertex_size(struct vertex_info *vinfo) static void calculate_vertex_layout( struct i915_context *i915 ) { const unsigned inputsRead = i915->fs.inputs_read; + const uint colorInterp + = i915->setup.flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &i915->current.vertex_info; boolean needW = 0; memset(vinfo, 0, sizeof(*vinfo)); /* pos */ - emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_3F); + emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_3F, INTERP_LINEAR); /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ /* color0 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, FORMAT_4UB); + emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_COLOR; } /* color 1 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { assert(0); /* untested */ - emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, FORMAT_4UB); + emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; } @@ -119,7 +122,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) for (i = TGSI_ATTRIB_TEX0; i <= TGSI_ATTRIB_TEX7; i++) { uint hwtc; if (inputsRead & (1 << i)) { - emit_vertex_attr(vinfo, i, FORMAT_4F); + emit_vertex_attr(vinfo, i, FORMAT_4F, INTERP_PERSPECTIVE); hwtc = TEXCOORDFMT_4D; needW = TRUE; } @@ -146,10 +149,10 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ if (i915->setup.light_twoside) { if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT); + emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT, colorInterp); } if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, FORMAT_OMIT); + emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, FORMAT_OMIT, colorInterp); } } @@ -160,6 +163,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ draw_set_vertex_attributes( i915->draw, vinfo->slot_to_attrib, + vinfo->interp_mode, vinfo->num_attribs); /* Need to set this flag so that the LIS2/4 registers get set. diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index ddd06ab0b7..fd0af6f727 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -46,13 +46,6 @@ struct draw_context; struct draw_stage; -enum interp_mode { - INTERP_CONSTANT, - INTERP_LINEAR, - INTERP_PERSPECTIVE -}; - - #define SP_NEW_VIEWPORT 0x1 #define SP_NEW_SETUP 0x2 #define SP_NEW_FS 0x4 diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index a7e1c9e017..b3c9f51e95 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -487,6 +487,10 @@ static void setup_tri_coefficients( struct setup_stage *setup ) for (j = 0; j < NUM_CHANNELS; j++) tri_persp_coeff(setup, slot, j); break; + + default: + /* invalid interp mode */ + assert(0); } } } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 3bbe9c9fd9..ff0a275fce 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -135,11 +135,13 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* If the attributes have changed, tell the draw module about * the new vertex layout. */ - if (vinfo->attr_mask != softpipe->attr_mask) { + /* XXX we also need to do this when the shading mode (interp modes) change: */ + if (1/*vinfo->attr_mask != softpipe->attr_mask*/) { softpipe->attr_mask = vinfo->attr_mask; draw_set_vertex_attributes( softpipe->draw, vinfo->slot_to_attrib, + vinfo->interp_mode, vinfo->num_attribs); } } -- cgit v1.2.3 From 8f1a0decad6baabed55371415fa4f75309c4c559 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 30 Aug 2007 16:01:23 -0600 Subject: fix unhandled switch/case warning --- src/mesa/pipe/softpipe/sp_prim_setup.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index b3c9f51e95..e7b58e5314 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -717,6 +717,10 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) for (j = 0; j < NUM_CHANNELS; j++) line_persp_coeff(setup, slot, j); break; + + default: + /* invalid interp mode */ + assert(0); } } } -- cgit v1.2.3 From c9e133eab450870c8804e1d3d1e7a44f509454a0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 30 Aug 2007 16:04:10 -0600 Subject: fix breakage from prev commit --- src/mesa/pipe/draw/draw_flatshade.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 602a8785ad..ae7bd24554 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -133,7 +133,7 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) { struct draw_stage *flatshade = CALLOC_STRUCT(draw_stage); - draw_alloc_tmps( &flatshade->stage, 2 ); + draw_alloc_tmps( flatshade, 2 ); flatshade->draw = draw; flatshade->next = NULL; -- cgit v1.2.3 From 963b8a74493a474560447080b23407bfe4c707c5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 30 Aug 2007 16:37:56 -0600 Subject: Remove dependency on TGSI_ATTRIB_x tokens in draw_twoside.c Added a new draw_set_twoside_attributes() function for specifying which vertex attributes are to be copied/replaced when a polygon is back-facing. --- src/mesa/pipe/draw/draw_context.c | 41 +++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_context.h | 4 +++ src/mesa/pipe/draw/draw_private.h | 3 ++ src/mesa/pipe/draw/draw_twoside.c | 26 ++++++++--------- src/mesa/pipe/i915simple/i915_state_derived.c | 22 ++++++++++---- src/mesa/pipe/softpipe/sp_state_derived.c | 25 ++++++++++++---- 6 files changed, 95 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index b14de34141..c15f7652a6 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -67,6 +67,11 @@ struct draw_context *draw_create( void ) draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * MAX_VERTEX_SIZE); } + draw->attrib_front0 = -1; + draw->attrib_back0 = -1; + draw->attrib_front1 = -1; + draw->attrib_back1 = -1; + return draw; } @@ -210,3 +215,39 @@ draw_set_vertex_shader(struct draw_context *draw, { draw->vertex_shader = *shader; } + + +/** + * This function is used to tell the draw module about attributes + * (like colors) that need to be selected based on front/back face + * orientation. + * + * The logic is: + * if (polygon is back-facing) { + * vertex->attrib[front0] = vertex->attrib[back0]; + * vertex->attrib[front1] = vertex->attrib[back1]; + * } + * + * \param front0 first attrib to replace if the polygon is back-facing + * \param back0 first attrib to copy if the polygon is back-facing + * \param front1 second attrib to replace if the polygon is back-facing + * \param back1 second attrib to copy if the polygon is back-facing + * + * Pass -1 to disable two-sided attributes. + */ +void +draw_set_twoside_attributes(struct draw_context *draw, + uint front0, uint back0, + uint front1, uint back1) +{ + /* XXX we could alternately pass an array of front/back attribs if there's + * ever need for more than two. One could imagine a shader extension + * that allows arbitrary attributes to be selected based on polygon + * orientation... + */ + draw->attrib_front0 = front0; + draw->attrib_back0 = back0; + draw->attrib_front1 = front1; + draw->attrib_back1 = back1; +} + diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 4c9e64a12d..21ee18e7cf 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -92,6 +92,10 @@ void draw_set_vertex_attributes( struct draw_context *draw, const uint *attrs, const uint *interp_mode, unsigned nr_attrs ); +void draw_set_twoside_attributes(struct draw_context *draw, + uint front0, uint back0, + uint front1, uint back1); + unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); unsigned draw_trim( unsigned count, unsigned first, unsigned incr ); diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index e61f228d12..80c97ada32 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -158,6 +158,9 @@ struct draw_context /** Describes the layout of post-transformation vertices */ struct vertex_info vertex_info; + /** Two-sided attributes: */ + uint attrib_front0, attrib_back0; + uint attrib_front1, attrib_back1; unsigned nr_vertices; diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index a05eea41fc..98eb088035 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -61,15 +61,11 @@ static void twoside_begin( struct draw_stage *stage ) } -static INLINE void copy_color( unsigned attr_dst, +static INLINE void copy_attrib( unsigned attr_dst, unsigned attr_src, struct vertex_header *v ) { - if (attr_dst && attr_src) { - memcpy( v->data[attr_dst], - v->data[attr_src], - sizeof(v->data[0]) ); - } + COPY_4FV(v->data[attr_dst], v->data[attr_src]); } @@ -78,14 +74,16 @@ static struct vertex_header *copy_bfc( struct twoside_stage *twoside, unsigned idx ) { struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); + const struct draw_context *draw = twoside->stage.draw; - copy_color( twoside->lookup[TGSI_ATTRIB_COLOR0], - twoside->lookup[TGSI_ATTRIB_BFC0], - tmp ); - - copy_color( twoside->lookup[TGSI_ATTRIB_COLOR1], - twoside->lookup[TGSI_ATTRIB_BFC1], - tmp ); + if (draw->attrib_front0) { + assert(draw->attrib_back0); + copy_attrib(draw->attrib_front0, draw->attrib_back0, tmp); + } + if (draw->attrib_front1) { + assert(draw->attrib_back1); + copy_attrib(draw->attrib_front1, draw->attrib_back1, tmp); + } return tmp; } @@ -104,7 +102,7 @@ static void twoside_tri( struct draw_stage *stage, tmp.det = header->det; tmp.edgeflags = header->edgeflags; - /* copy back colors to front color slots */ + /* copy back attribs to front attribs */ tmp.v[0] = copy_bfc(twoside, header->v[0], 0); tmp.v[1] = copy_bfc(twoside, header->v[1], 1); tmp.v[2] = copy_bfc(twoside, header->v[2], 2); diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index e156766a93..4b97223a6e 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -35,8 +35,9 @@ #include "i915_fpc.h" -static INLINE void -emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, uint interp) +static INLINE uint +emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, + uint interp) { const uint n = vinfo->num_attribs; vinfo->attr_mask |= (1 << vfAttr); @@ -44,6 +45,7 @@ emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, uint inter vinfo->format[n] = format; vinfo->interp_mode[n] = interp; vinfo->num_attribs++; + return n; } @@ -93,6 +95,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) const uint colorInterp = i915->setup.flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &i915->current.vertex_info; + uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; boolean needW = 0; memset(vinfo, 0, sizeof(*vinfo)); @@ -103,14 +106,16 @@ static void calculate_vertex_layout( struct i915_context *i915 ) /* color0 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, FORMAT_4UB, colorInterp); + front0 = emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, + FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_COLOR; } /* color 1 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { assert(0); /* untested */ - emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, FORMAT_4UB, colorInterp); + front1 = emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, + FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; } @@ -149,10 +154,12 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ if (i915->setup.light_twoside) { if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT, colorInterp); + back0 = emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, + FORMAT_OMIT, colorInterp); } if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, FORMAT_OMIT, colorInterp); + back1 = emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, + FORMAT_OMIT, colorInterp); } } @@ -166,6 +173,9 @@ static void calculate_vertex_layout( struct i915_context *i915 ) vinfo->interp_mode, vinfo->num_attribs); + draw_set_twoside_attributes(i915->draw, + front0, back0, front1, back1); + /* Need to set this flag so that the LIS2/4 registers get set. * It also means the i915_update_immediate() function must be called * after this one, in i915_update_derived(). diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index ff0a275fce..41b73405ae 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -37,16 +37,21 @@ -static void +/** + * Add another attribute to the given vertex_info object. + * \return slot in which the attribute was added + */ +static uint emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, uint interp) { const uint n = vinfo->num_attribs; vinfo->attr_mask |= (1 << vfAttr); vinfo->slot_to_attrib[n] = vfAttr; - vinfo->interp_mode[n] = interp; vinfo->format[n] = format; + vinfo->interp_mode[n] = interp; vinfo->num_attribs++; + return n; } @@ -62,6 +67,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) const uint colorInterp = softpipe->setup.flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; + uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; uint i; memset(vinfo, 0, sizeof(*vinfo)); @@ -89,12 +95,14 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* color0 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, FORMAT_4F, colorInterp); + front0 = emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, + FORMAT_4F, colorInterp); } /* color1 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, FORMAT_4F, colorInterp); + front1 = emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, + FORMAT_4F, colorInterp); } /* fog */ @@ -124,11 +132,13 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ if (softpipe->setup.light_twoside) { if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT, INTERP_LINEAR); + back0 = emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, + FORMAT_OMIT, colorInterp); } if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, FORMAT_OMIT, INTERP_LINEAR); + back1 = emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, + FORMAT_OMIT, colorInterp); } } @@ -143,6 +153,9 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) vinfo->slot_to_attrib, vinfo->interp_mode, vinfo->num_attribs); + + draw_set_twoside_attributes(softpipe->draw, + front0, back0, front1, back1); } } -- cgit v1.2.3 From 898d68a3762f84f0d435cda2f6aafddd356d1788 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 30 Aug 2007 16:49:24 -0600 Subject: Consolidate vertex-related code in new draw_vertex.c A few functions which were basically duplicated between softpipe and the i915 driver are now re-used: draw_emit_vertex_attr() draw_compute_vertex_size() --- src/mesa/pipe/draw/draw_arrays.c | 79 ------------ src/mesa/pipe/draw/draw_context.c | 36 ------ src/mesa/pipe/draw/draw_context.h | 3 + src/mesa/pipe/draw/draw_vertex.c | 167 ++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_vertex.h | 19 +++ src/mesa/pipe/i915simple/i915_state_derived.c | 71 ++--------- src/mesa/pipe/softpipe/sp_state_derived.c | 46 +++---- src/mesa/sources | 1 + 8 files changed, 215 insertions(+), 207 deletions(-) create mode 100644 src/mesa/pipe/draw/draw_vertex.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c index 9e219ed43b..bbb243c469 100644 --- a/src/mesa/pipe/draw/draw_arrays.c +++ b/src/mesa/pipe/draw/draw_arrays.c @@ -74,82 +74,3 @@ draw_arrays(struct draw_context *draw, unsigned prim, } -static INLINE void -emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, uint interp) -{ - const uint n = vinfo->num_attribs; - vinfo->attr_mask |= (1 << vfAttr); - vinfo->slot_to_attrib[n] = vfAttr; - if (n >= 2) { - /* the first two slots are the vertex header & clippos */ - assert(vfAttr < Elements(vinfo->attrib_to_slot)); - vinfo->attrib_to_slot[vfAttr] = n - 2; - } - vinfo->interp_mode[n] = interp; - vinfo->format[n] = format; - vinfo->num_attribs++; - -} - - -/** XXX this is duplicated in the i915 driver... */ -static void -compute_vertex_size(struct vertex_info *vinfo) -{ - uint i; - - vinfo->size = 0; - for (i = 0; i < vinfo->num_attribs; i++) { - switch (vinfo->format[i]) { - case FORMAT_OMIT: - break; - case FORMAT_4UB: - /* fall-through */ - case FORMAT_1F: - vinfo->size += 1; - break; - case FORMAT_2F: - vinfo->size += 2; - break; - case FORMAT_3F: - vinfo->size += 3; - break; - case FORMAT_4F: - vinfo->size += 4; - break; - default: - assert(0); - } - } -} - - -void -draw_set_vertex_attributes( struct draw_context *draw, - const uint *slot_to_vf_attr, - const uint *interp_mode, - unsigned nr_attrs ) -{ - struct vertex_info *vinfo = &draw->vertex_info; - unsigned i; - - assert(slot_to_vf_attr[0] == TGSI_ATTRIB_POS); - - memset(vinfo, 0, sizeof(*vinfo)); - - /* - * First three attribs are always the same: header, clip pos, winpos - */ - emit_vertex_attr(vinfo, TGSI_ATTRIB_VERTEX_HEADER, FORMAT_1F, INTERP_NONE); - emit_vertex_attr(vinfo, TGSI_ATTRIB_CLIP_POS, FORMAT_4F, INTERP_LINEAR); - emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F_VIEWPORT, INTERP_LINEAR); - - /* - * Remaining attribs (color, texcoords, etc) - */ - for (i = 1; i < nr_attrs; i++) { - emit_vertex_attr(vinfo, slot_to_vf_attr[i], FORMAT_4F, interp_mode[i]); - } - - compute_vertex_size(vinfo); -} diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index c15f7652a6..2936a14d65 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -215,39 +215,3 @@ draw_set_vertex_shader(struct draw_context *draw, { draw->vertex_shader = *shader; } - - -/** - * This function is used to tell the draw module about attributes - * (like colors) that need to be selected based on front/back face - * orientation. - * - * The logic is: - * if (polygon is back-facing) { - * vertex->attrib[front0] = vertex->attrib[back0]; - * vertex->attrib[front1] = vertex->attrib[back1]; - * } - * - * \param front0 first attrib to replace if the polygon is back-facing - * \param back0 first attrib to copy if the polygon is back-facing - * \param front1 second attrib to replace if the polygon is back-facing - * \param back1 second attrib to copy if the polygon is back-facing - * - * Pass -1 to disable two-sided attributes. - */ -void -draw_set_twoside_attributes(struct draw_context *draw, - uint front0, uint back0, - uint front1, uint back1) -{ - /* XXX we could alternately pass an array of front/back attribs if there's - * ever need for more than two. One could imagine a shader extension - * that allows arbitrary attributes to be selected based on polygon - * orientation... - */ - draw->attrib_front0 = front0; - draw->attrib_back0 = back0; - draw->attrib_front1 = front1; - draw->attrib_back1 = back1; -} - diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 21ee18e7cf..03679848b5 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -42,6 +42,7 @@ struct vertex_buffer; +struct vertex_info; struct draw_context; struct draw_stage; @@ -96,6 +97,8 @@ void draw_set_twoside_attributes(struct draw_context *draw, uint front0, uint back0, uint front1, uint back1); +void draw_compute_vertex_size(struct vertex_info *vinfo); + unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); unsigned draw_trim( unsigned count, unsigned first, unsigned incr ); diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c new file mode 100644 index 0000000000..64bc323207 --- /dev/null +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -0,0 +1,167 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * Functions for specifying the post-transformation vertex layout. + * + * Author: + * Brian Paul + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" + +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_vertex.h" + + +static INLINE void +emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, + uint interp) +{ + const uint n = vinfo->num_attribs; + vinfo->attr_mask |= (1 << vfAttr); + vinfo->slot_to_attrib[n] = vfAttr; + if (n >= 2) { + /* the first two slots are the vertex header & clippos */ + assert(vfAttr < Elements(vinfo->attrib_to_slot)); + vinfo->attrib_to_slot[vfAttr] = n - 2; + } + vinfo->interp_mode[n] = interp; + vinfo->format[n] = format; + vinfo->num_attribs++; + +} + + +/** + * Compute the size of a vertex, in dwords/floats, to update the + * vinfo->size field. + */ +void +draw_compute_vertex_size(struct vertex_info *vinfo) +{ + uint i; + + vinfo->size = 0; + for (i = 0; i < vinfo->num_attribs; i++) { + switch (vinfo->format[i]) { + case FORMAT_OMIT: + break; + case FORMAT_4UB: + /* fall-through */ + case FORMAT_1F: + vinfo->size += 1; + break; + case FORMAT_2F: + vinfo->size += 2; + break; + case FORMAT_3F: + vinfo->size += 3; + break; + case FORMAT_4F: + vinfo->size += 4; + break; + default: + assert(0); + } + } +} + + +/** + * Tell the drawing module about the layout of post-transformation vertices + */ +void +draw_set_vertex_attributes( struct draw_context *draw, + const uint *slot_to_vf_attr, + const uint *interp_mode, + unsigned nr_attrs ) +{ + struct vertex_info *vinfo = &draw->vertex_info; + unsigned i; + + assert(slot_to_vf_attr[0] == TGSI_ATTRIB_POS); + + memset(vinfo, 0, sizeof(*vinfo)); + + /* + * First three attribs are always the same: header, clip pos, winpos + */ + emit_vertex_attr(vinfo, TGSI_ATTRIB_VERTEX_HEADER, FORMAT_1F, INTERP_NONE); + emit_vertex_attr(vinfo, TGSI_ATTRIB_CLIP_POS, FORMAT_4F, INTERP_LINEAR); + emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F_VIEWPORT, INTERP_LINEAR); + + /* + * Remaining attribs (color, texcoords, etc) + */ + for (i = 1; i < nr_attrs; i++) { + emit_vertex_attr(vinfo, slot_to_vf_attr[i], FORMAT_4F, interp_mode[i]); + } + + draw_compute_vertex_size(vinfo); +} + + +/** + * This function is used to tell the draw module about attributes + * (like colors) that need to be selected based on front/back face + * orientation. + * + * The logic is: + * if (polygon is back-facing) { + * vertex->attrib[front0] = vertex->attrib[back0]; + * vertex->attrib[front1] = vertex->attrib[back1]; + * } + * + * \param front0 first attrib to replace if the polygon is back-facing + * \param back0 first attrib to copy if the polygon is back-facing + * \param front1 second attrib to replace if the polygon is back-facing + * \param back1 second attrib to copy if the polygon is back-facing + * + * Pass -1 to disable two-sided attributes. + */ +void +draw_set_twoside_attributes(struct draw_context *draw, + uint front0, uint back0, + uint front1, uint back1) +{ + /* XXX we could alternately pass an array of front/back attribs if there's + * ever need for more than two. One could imagine a shader extension + * that allows arbitrary attributes to be selected based on polygon + * orientation... + */ + draw->attrib_front0 = front0; + draw->attrib_back0 = back0; + draw->attrib_front1 = front1; + draw->attrib_back1 = back1; +} diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 6778df30d1..f696475510 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -70,6 +70,25 @@ struct vertex_info +/** + * Add another attribute to the given vertex_info object. + * \return slot in which the attribute was added + */ +static INLINE uint +draw_emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, + uint interp) +{ + const uint n = vinfo->num_attribs; + vinfo->attr_mask |= (1 << vfAttr); + vinfo->slot_to_attrib[n] = vfAttr; + vinfo->format[n] = format; + vinfo->interp_mode[n] = interp; + vinfo->num_attribs++; + return n; +} + + + struct draw_context; extern int draw_vertex_cache_check_space( struct draw_context *draw, diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 4b97223a6e..4347dadbd0 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -35,55 +35,6 @@ #include "i915_fpc.h" -static INLINE uint -emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, - uint interp) -{ - const uint n = vinfo->num_attribs; - vinfo->attr_mask |= (1 << vfAttr); - vinfo->slot_to_attrib[n] = vfAttr; - vinfo->format[n] = format; - vinfo->interp_mode[n] = interp; - vinfo->num_attribs++; - return n; -} - - -/** - * Recompute the vinfo->size field. - */ -static void -compute_vertex_size(struct vertex_info *vinfo) -{ - uint i; - - vinfo->size = 0; - for (i = 0; i < vinfo->num_attribs; i++) { - switch (vinfo->format[i]) { - case FORMAT_OMIT: - break; - case FORMAT_4UB: - /* fall-through */ - case FORMAT_1F: - vinfo->size += 1; - break; - case FORMAT_2F: - vinfo->size += 2; - break; - case FORMAT_3F: - vinfo->size += 3; - break; - case FORMAT_4F: - vinfo->size += 4; - break; - default: - assert(0); - } - } -} - - - /** * Determine which post-transform / pre-rasterization vertex attributes * we need. @@ -101,21 +52,21 @@ static void calculate_vertex_layout( struct i915_context *i915 ) memset(vinfo, 0, sizeof(*vinfo)); /* pos */ - emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_3F, INTERP_LINEAR); + draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_3F, INTERP_LINEAR); /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ /* color0 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - front0 = emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, - FORMAT_4UB, colorInterp); + front0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, + FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_COLOR; } /* color 1 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { assert(0); /* untested */ - front1 = emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, - FORMAT_4UB, colorInterp); + front1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, + FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; } @@ -127,7 +78,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) for (i = TGSI_ATTRIB_TEX0; i <= TGSI_ATTRIB_TEX7; i++) { uint hwtc; if (inputsRead & (1 << i)) { - emit_vertex_attr(vinfo, i, FORMAT_4F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, i, FORMAT_4F, INTERP_PERSPECTIVE); hwtc = TEXCOORDFMT_4D; needW = TRUE; } @@ -154,16 +105,16 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ if (i915->setup.light_twoside) { if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - back0 = emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, - FORMAT_OMIT, colorInterp); + back0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, + FORMAT_OMIT, colorInterp); } if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - back1 = emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, - FORMAT_OMIT, colorInterp); + back1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, + FORMAT_OMIT, colorInterp); } } - compute_vertex_size(vinfo); + draw_compute_vertex_size(vinfo); /* If the attributes have changed, tell the draw module about the new * vertex layout. We'll also update the hardware vertex format info. diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 41b73405ae..4880870e6e 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -36,26 +36,6 @@ #include "pipe/tgsi/exec/tgsi_attribs.h" - -/** - * Add another attribute to the given vertex_info object. - * \return slot in which the attribute was added - */ -static uint -emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, - uint interp) -{ - const uint n = vinfo->num_attribs; - vinfo->attr_mask |= (1 << vfAttr); - vinfo->slot_to_attrib[n] = vfAttr; - vinfo->format[n] = format; - vinfo->interp_mode[n] = interp; - vinfo->num_attribs++; - return n; -} - - - /** * Determine which post-transform / pre-rasterization vertex attributes * we need. @@ -91,35 +71,37 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* position */ /* TODO - Figure out if we need to do perspective divide, etc. */ - emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F, INTERP_LINEAR); + draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F, INTERP_LINEAR); /* color0 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - front0 = emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, - FORMAT_4F, colorInterp); + front0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, + FORMAT_4F, colorInterp); } /* color1 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - front1 = emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, - FORMAT_4F, colorInterp); + front1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, + FORMAT_4F, colorInterp); } /* fog */ if (inputsRead & (1 << TGSI_ATTRIB_FOG)) { - emit_vertex_attr(vinfo, TGSI_ATTRIB_FOG, FORMAT_1F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_FOG, + FORMAT_1F, INTERP_PERSPECTIVE); } /* point size */ #if 0 /* XXX only emit if drawing points or front/back polygon mode is point mode */ - emit_vertex_attr(vinfo, TGSI_ATTRIB_POINTSIZE, FORMAT_4F, INTERP_CONSTANT); + draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POINTSIZE, + FORMAT_4F, INTERP_CONSTANT); #endif /* texcoords and varying vars */ for (i = TGSI_ATTRIB_TEX0; i < TGSI_ATTRIB_VAR7; i++) { if (inputsRead & (1 << i)) { - emit_vertex_attr(vinfo, i, FORMAT_4F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, i, FORMAT_4F, INTERP_PERSPECTIVE); softpipe->need_w = TRUE; } } @@ -132,13 +114,13 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ if (softpipe->setup.light_twoside) { if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - back0 = emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, - FORMAT_OMIT, colorInterp); + back0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, + FORMAT_OMIT, colorInterp); } if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - back1 = emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, - FORMAT_OMIT, colorInterp); + back1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, + FORMAT_OMIT, colorInterp); } } diff --git a/src/mesa/sources b/src/mesa/sources index be24f93f74..ae0797ae24 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -164,6 +164,7 @@ DRAW_SOURCES = \ pipe/draw/draw_offset.c \ pipe/draw/draw_prim.c \ pipe/draw/draw_twoside.c \ + pipe/draw/draw_vertex.c \ pipe/draw/draw_vertex_cache.c \ pipe/draw/draw_vertex_fetch.c \ pipe/draw/draw_vertex_shader.c \ -- cgit v1.2.3 From 2e21058e3d2c484b282cbc0bb5e7169b9a8d4fc7 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 31 Aug 2007 11:27:16 -0600 Subject: Define attrib_format and interp_mode enum typedefs and use where appropriate. --- src/mesa/pipe/draw/draw_context.h | 10 ------ src/mesa/pipe/draw/draw_flatshade.c | 4 +-- src/mesa/pipe/draw/draw_vertex.c | 10 +++--- src/mesa/pipe/draw/draw_vertex.h | 51 +++++++++++++++++++-------- src/mesa/pipe/i915simple/i915_state_derived.c | 4 +-- src/mesa/pipe/softpipe/sp_prim_setup.c | 6 ++-- src/mesa/pipe/softpipe/sp_state_derived.c | 4 +-- 7 files changed, 51 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 03679848b5..7c4f1a1297 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -89,16 +89,6 @@ void draw_set_setup_state( struct draw_context *draw, void draw_set_setup_stage( struct draw_context *draw, struct draw_stage *stage ); -void draw_set_vertex_attributes( struct draw_context *draw, - const uint *attrs, const uint *interp_mode, - unsigned nr_attrs ); - -void draw_set_twoside_attributes(struct draw_context *draw, - uint front0, uint back0, - uint front1, uint back1); - -void draw_compute_vertex_size(struct vertex_info *vinfo); - unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); unsigned draw_trim( unsigned count, unsigned first, unsigned incr ); diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index ae7bd24554..2ba0d5820e 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -56,12 +56,12 @@ static INLINE void copy_colors( struct draw_stage *stage, const struct vertex_header *src ) { const uint num_attribs = stage->draw->vertex_info.num_attribs; - const uint *interp_mode = stage->draw->vertex_info.interp_mode; + const interp_mode *interp = stage->draw->vertex_info.interp_mode; uint i; /* Look for constant/flat attribs and duplicate from src to dst vertex */ for (i = 1; i < num_attribs - 2; i++) { - if (interp_mode[i + 2] == INTERP_CONSTANT) { + if (interp[i + 2] == INTERP_CONSTANT) { copy_attr( i, dst, src ); } } diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index 64bc323207..1c7e1d8662 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -45,8 +45,8 @@ static INLINE void -emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, - uint interp) +emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, + attrib_format format, interp_mode interp) { const uint n = vinfo->num_attribs; vinfo->attr_mask |= (1 << vfAttr); @@ -59,7 +59,6 @@ emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, vinfo->interp_mode[n] = interp; vinfo->format[n] = format; vinfo->num_attribs++; - } @@ -89,6 +88,7 @@ draw_compute_vertex_size(struct vertex_info *vinfo) vinfo->size += 3; break; case FORMAT_4F: + case FORMAT_4F_VIEWPORT: vinfo->size += 4; break; default: @@ -104,7 +104,7 @@ draw_compute_vertex_size(struct vertex_info *vinfo) void draw_set_vertex_attributes( struct draw_context *draw, const uint *slot_to_vf_attr, - const uint *interp_mode, + const interp_mode *interps, unsigned nr_attrs ) { struct vertex_info *vinfo = &draw->vertex_info; @@ -125,7 +125,7 @@ draw_set_vertex_attributes( struct draw_context *draw, * Remaining attribs (color, texcoords, etc) */ for (i = 1; i < nr_attrs; i++) { - emit_vertex_attr(vinfo, slot_to_vf_attr[i], FORMAT_4F, interp_mode[i]); + emit_vertex_attr(vinfo, slot_to_vf_attr[i], FORMAT_4F, interps[i]); } draw_compute_vertex_size(vinfo); diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index f696475510..391058af26 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -38,21 +38,34 @@ #define MAX_VERT_ATTRIBS 12 /* OK? */ -#define FORMAT_OMIT 0 -#define FORMAT_1F 1 -#define FORMAT_2F 2 -#define FORMAT_3F 3 -#define FORMAT_4F 4 -#define FORMAT_4F_VIEWPORT 4 -#define FORMAT_4UB 5 + +struct draw_context; + + + +/** + * Vertex attribute format + */ +typedef enum { + FORMAT_OMIT, + FORMAT_1F, + FORMAT_2F, + FORMAT_3F, + FORMAT_4F, + FORMAT_4F_VIEWPORT, + FORMAT_4UB +} attrib_format; -enum interp_mode { +/** + * Attribute interpolation mode + */ +typedef enum { INTERP_NONE, /**< never interpolate vertex header info */ INTERP_CONSTANT, INTERP_LINEAR, INTERP_PERSPECTIVE -}; +} interp_mode; @@ -63,8 +76,8 @@ struct vertex_info uint attr_mask; /**< mask of VF_ATTR_ bits */ uint slot_to_attrib[MAX_VERT_ATTRIBS]; uint attrib_to_slot[TGSI_ATTRIB_MAX]; - uint interp_mode[MAX_VERT_ATTRIBS]; - uint format[MAX_VERT_ATTRIBS]; /**< FORMAT_x */ + interp_mode interp_mode[MAX_VERT_ATTRIBS]; + attrib_format format[MAX_VERT_ATTRIBS]; /**< FORMAT_x */ uint size; /**< total vertex size in dwords */ }; @@ -75,10 +88,11 @@ struct vertex_info * \return slot in which the attribute was added */ static INLINE uint -draw_emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, - uint interp) +draw_emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, + attrib_format format, interp_mode interp) { const uint n = vinfo->num_attribs; + assert(n < MAX_VERT_ATTRIBS); vinfo->attr_mask |= (1 << vfAttr); vinfo->slot_to_attrib[n] = vfAttr; vinfo->format[n] = format; @@ -88,8 +102,17 @@ draw_emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, uint format, } +extern void draw_set_vertex_attributes( struct draw_context *draw, + const uint *attrs, + const interp_mode *interps, + unsigned nr_attrs ); + +extern void draw_set_twoside_attributes(struct draw_context *draw, + uint front0, uint back0, + uint front1, uint back1); + +extern void draw_compute_vertex_size(struct vertex_info *vinfo); -struct draw_context; extern int draw_vertex_cache_check_space( struct draw_context *draw, unsigned nr_verts ); diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 4347dadbd0..792bb93b17 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -42,8 +42,8 @@ */ static void calculate_vertex_layout( struct i915_context *i915 ) { - const unsigned inputsRead = i915->fs.inputs_read; - const uint colorInterp + const uint inputsRead = i915->fs.inputs_read; + const interp_mode colorInterp = i915->setup.flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &i915->current.vertex_info; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index e7b58e5314..83d317c36f 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -38,9 +38,9 @@ #include "sp_quad.h" #include "sp_prim_setup.h" #include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_vertex.h" #include "pipe/p_util.h" -#include "pipe/draw/draw_vertex.h" /** @@ -461,7 +461,7 @@ static void tri_persp_coeff( struct setup_stage *setup, */ static void setup_tri_coefficients( struct setup_stage *setup ) { - const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + const interp_mode *interp = setup->softpipe->vertex_info.interp_mode; unsigned slot, j; /* z and w are done by linear interpolation: @@ -680,7 +680,7 @@ line_persp_coeff(struct setup_stage *setup, unsigned slot, unsigned i) static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { - const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + const interp_mode *interp = setup->softpipe->vertex_info.interp_mode; unsigned slot, j; /* use setup->vmin, vmax to point to vertices */ diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 4880870e6e..e08ed50a70 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -43,8 +43,8 @@ */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - const unsigned inputsRead = softpipe->fs.inputs_read; - const uint colorInterp + const uint inputsRead = softpipe->fs.inputs_read; + const interp_mode colorInterp = softpipe->setup.flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; -- cgit v1.2.3 From 9a11a4a8b4b66201aeeb9a7057aeb08ef47536e1 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 31 Aug 2007 11:28:31 -0600 Subject: move some protos from draw_vertex.h to draw_private.h --- src/mesa/pipe/draw/draw_private.h | 16 ++++++++++++++++ src/mesa/pipe/draw/draw_vertex.h | 17 ----------------- 2 files changed, 16 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 80c97ada32..9f90dc3e26 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -215,6 +215,22 @@ extern void draw_free_tmps( struct draw_stage *stage ); extern void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ); +extern int draw_vertex_cache_check_space( struct draw_context *draw, + unsigned nr_verts ); + +extern void draw_vertex_cache_validate( struct draw_context *draw ); +extern void draw_vertex_cache_invalidate( struct draw_context *draw ); +extern void draw_vertex_cache_unreference( struct draw_context *draw ); + +extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); + +struct tgsi_exec_machine; + +extern void draw_vertex_fetch( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ); + /** * Get a writeable copy of a vertex. diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 391058af26..4e3e86d86e 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -114,21 +114,4 @@ extern void draw_set_twoside_attributes(struct draw_context *draw, extern void draw_compute_vertex_size(struct vertex_info *vinfo); -extern int draw_vertex_cache_check_space( struct draw_context *draw, - unsigned nr_verts ); - -extern void draw_vertex_cache_validate( struct draw_context *draw ); -extern void draw_vertex_cache_invalidate( struct draw_context *draw ); -extern void draw_vertex_cache_unreference( struct draw_context *draw ); - -extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); - -struct tgsi_exec_machine; - -extern void draw_vertex_fetch( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ); - - #endif /* DRAW_VERTEX_H */ -- cgit v1.2.3 From f2f168d0f8a1427e07133eceae290fa2e845f95f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 6 Sep 2007 14:26:35 -0600 Subject: Move guts of glRasterPos into tnl module, called via new ctx->Driver.RasterPos() function. --- src/mesa/tnl/t_rasterpos.c | 508 +++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 508 insertions(+) create mode 100644 src/mesa/tnl/t_rasterpos.c (limited to 'src') diff --git a/src/mesa/tnl/t_rasterpos.c b/src/mesa/tnl/t_rasterpos.c new file mode 100644 index 0000000000..272f855fc1 --- /dev/null +++ b/src/mesa/tnl/t_rasterpos.c @@ -0,0 +1,508 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#include "glheader.h" +#include "colormac.h" +#include "context.h" +#include "feedback.h" +#include "light.h" +#include "macros.h" +#include "rastpos.h" +#include "simple_list.h" +#include "mtypes.h" + +#include "math/m_matrix.h" +#include "tnl/tnl.h" + + + +/** + * Clip a point against the view volume. + * + * \param v vertex vector describing the point to clip. + * + * \return zero if outside view volume, or one if inside. + */ +static GLuint +viewclip_point( const GLfloat v[] ) +{ + if ( v[0] > v[3] || v[0] < -v[3] + || v[1] > v[3] || v[1] < -v[3] + || v[2] > v[3] || v[2] < -v[3] ) { + return 0; + } + else { + return 1; + } +} + + +/** + * Clip a point against the far/near Z clipping planes. + * + * \param v vertex vector describing the point to clip. + * + * \return zero if outside view volume, or one if inside. + */ +static GLuint +viewclip_point_z( const GLfloat v[] ) +{ + if (v[2] > v[3] || v[2] < -v[3] ) { + return 0; + } + else { + return 1; + } +} + + +/** + * Clip a point against the user clipping planes. + * + * \param ctx GL context. + * \param v vertex vector describing the point to clip. + * + * \return zero if the point was clipped, or one otherwise. + */ +static GLuint +userclip_point( GLcontext *ctx, const GLfloat v[] ) +{ + GLuint p; + + for (p = 0; p < ctx->Const.MaxClipPlanes; p++) { + if (ctx->Transform.ClipPlanesEnabled & (1 << p)) { + GLfloat dot = v[0] * ctx->Transform._ClipUserPlane[p][0] + + v[1] * ctx->Transform._ClipUserPlane[p][1] + + v[2] * ctx->Transform._ClipUserPlane[p][2] + + v[3] * ctx->Transform._ClipUserPlane[p][3]; + if (dot < 0.0F) { + return 0; + } + } + } + + return 1; +} + + +/** + * Compute lighting for the raster position. Both RGB and CI modes computed. + * \param ctx the context + * \param vertex vertex location + * \param normal normal vector + * \param Rcolor returned color + * \param Rspec returned specular color (if separate specular enabled) + * \param Rindex returned color index + */ +static void +shade_rastpos(GLcontext *ctx, + const GLfloat vertex[4], + const GLfloat normal[3], + GLfloat Rcolor[4], + GLfloat Rspec[4], + GLfloat *Rindex) +{ + /*const*/ GLfloat (*base)[3] = ctx->Light._BaseColor; + const struct gl_light *light; + GLfloat diffuseColor[4], specularColor[4]; /* for RGB mode only */ + GLfloat diffuseCI = 0.0, specularCI = 0.0; /* for CI mode only */ + + _mesa_validate_all_lighting_tables( ctx ); + + COPY_3V(diffuseColor, base[0]); + diffuseColor[3] = CLAMP( + ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3], 0.0F, 1.0F ); + ASSIGN_4V(specularColor, 0.0, 0.0, 0.0, 1.0); + + foreach (light, &ctx->Light.EnabledList) { + GLfloat attenuation = 1.0; + GLfloat VP[3]; /* vector from vertex to light pos */ + GLfloat n_dot_VP; + GLfloat diffuseContrib[3], specularContrib[3]; + + if (!(light->_Flags & LIGHT_POSITIONAL)) { + /* light at infinity */ + COPY_3V(VP, light->_VP_inf_norm); + attenuation = light->_VP_inf_spot_attenuation; + } + else { + /* local/positional light */ + GLfloat d; + + /* VP = vector from vertex pos to light[i].pos */ + SUB_3V(VP, light->_Position, vertex); + /* d = length(VP) */ + d = (GLfloat) LEN_3FV( VP ); + if (d > 1.0e-6) { + /* normalize VP */ + GLfloat invd = 1.0F / d; + SELF_SCALE_SCALAR_3V(VP, invd); + } + + /* atti */ + attenuation = 1.0F / (light->ConstantAttenuation + d * + (light->LinearAttenuation + d * + light->QuadraticAttenuation)); + + if (light->_Flags & LIGHT_SPOT) { + GLfloat PV_dot_dir = - DOT3(VP, light->_NormDirection); + + if (PV_dot_dir_CosCutoff) { + continue; + } + else { + double x = PV_dot_dir * (EXP_TABLE_SIZE-1); + int k = (int) x; + GLfloat spot = (GLfloat) (light->_SpotExpTable[k][0] + + (x-k)*light->_SpotExpTable[k][1]); + attenuation *= spot; + } + } + } + + if (attenuation < 1e-3) + continue; + + n_dot_VP = DOT3( normal, VP ); + + if (n_dot_VP < 0.0F) { + ACC_SCALE_SCALAR_3V(diffuseColor, attenuation, light->_MatAmbient[0]); + continue; + } + + /* Ambient + diffuse */ + COPY_3V(diffuseContrib, light->_MatAmbient[0]); + ACC_SCALE_SCALAR_3V(diffuseContrib, n_dot_VP, light->_MatDiffuse[0]); + diffuseCI += n_dot_VP * light->_dli * attenuation; + + /* Specular */ + { + const GLfloat *h; + GLfloat n_dot_h; + + ASSIGN_3V(specularContrib, 0.0, 0.0, 0.0); + + if (ctx->Light.Model.LocalViewer) { + GLfloat v[3]; + COPY_3V(v, vertex); + NORMALIZE_3FV(v); + SUB_3V(VP, VP, v); + NORMALIZE_3FV(VP); + h = VP; + } + else if (light->_Flags & LIGHT_POSITIONAL) { + ACC_3V(VP, ctx->_EyeZDir); + NORMALIZE_3FV(VP); + h = VP; + } + else { + h = light->_h_inf_norm; + } + + n_dot_h = DOT3(normal, h); + + if (n_dot_h > 0.0F) { + GLfloat spec_coef; + GET_SHINE_TAB_ENTRY( ctx->_ShineTable[0], n_dot_h, spec_coef ); + + if (spec_coef > 1.0e-10) { + if (ctx->Light.Model.ColorControl==GL_SEPARATE_SPECULAR_COLOR) { + ACC_SCALE_SCALAR_3V( specularContrib, spec_coef, + light->_MatSpecular[0]); + } + else { + ACC_SCALE_SCALAR_3V( diffuseContrib, spec_coef, + light->_MatSpecular[0]); + } + /*assert(light->_sli > 0.0);*/ + specularCI += spec_coef * light->_sli * attenuation; + } + } + } + + ACC_SCALE_SCALAR_3V( diffuseColor, attenuation, diffuseContrib ); + ACC_SCALE_SCALAR_3V( specularColor, attenuation, specularContrib ); + } + + if (ctx->Visual.rgbMode) { + Rcolor[0] = CLAMP(diffuseColor[0], 0.0F, 1.0F); + Rcolor[1] = CLAMP(diffuseColor[1], 0.0F, 1.0F); + Rcolor[2] = CLAMP(diffuseColor[2], 0.0F, 1.0F); + Rcolor[3] = CLAMP(diffuseColor[3], 0.0F, 1.0F); + Rspec[0] = CLAMP(specularColor[0], 0.0F, 1.0F); + Rspec[1] = CLAMP(specularColor[1], 0.0F, 1.0F); + Rspec[2] = CLAMP(specularColor[2], 0.0F, 1.0F); + Rspec[3] = CLAMP(specularColor[3], 0.0F, 1.0F); + } + else { + GLfloat *ind = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_INDEXES]; + GLfloat d_a = ind[MAT_INDEX_DIFFUSE] - ind[MAT_INDEX_AMBIENT]; + GLfloat s_a = ind[MAT_INDEX_SPECULAR] - ind[MAT_INDEX_AMBIENT]; + GLfloat i = (ind[MAT_INDEX_AMBIENT] + + diffuseCI * (1.0F-specularCI) * d_a + + specularCI * s_a); + if (i > ind[MAT_INDEX_SPECULAR]) { + i = ind[MAT_INDEX_SPECULAR]; + } + *Rindex = i; + } +} + + +/** + * Do texgen needed for glRasterPos. + * \param ctx rendering context + * \param vObj object-space vertex coordinate + * \param vEye eye-space vertex coordinate + * \param normal vertex normal + * \param unit texture unit number + * \param texcoord incoming texcoord and resulting texcoord + */ +static void +compute_texgen(GLcontext *ctx, const GLfloat vObj[4], const GLfloat vEye[4], + const GLfloat normal[3], GLuint unit, GLfloat texcoord[4]) +{ + const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; + + /* always compute sphere map terms, just in case */ + GLfloat u[3], two_nu, rx, ry, rz, m, mInv; + COPY_3V(u, vEye); + NORMALIZE_3FV(u); + two_nu = 2.0F * DOT3(normal, u); + rx = u[0] - normal[0] * two_nu; + ry = u[1] - normal[1] * two_nu; + rz = u[2] - normal[2] * two_nu; + m = rx * rx + ry * ry + (rz + 1.0F) * (rz + 1.0F); + if (m > 0.0F) + mInv = 0.5F * _mesa_inv_sqrtf(m); + else + mInv = 0.0F; + + if (texUnit->TexGenEnabled & S_BIT) { + switch (texUnit->GenModeS) { + case GL_OBJECT_LINEAR: + texcoord[0] = DOT4(vObj, texUnit->ObjectPlaneS); + break; + case GL_EYE_LINEAR: + texcoord[0] = DOT4(vEye, texUnit->EyePlaneS); + break; + case GL_SPHERE_MAP: + texcoord[0] = rx * mInv + 0.5F; + break; + case GL_REFLECTION_MAP: + texcoord[0] = rx; + break; + case GL_NORMAL_MAP: + texcoord[0] = normal[0]; + break; + default: + _mesa_problem(ctx, "Bad S texgen in compute_texgen()"); + return; + } + } + + if (texUnit->TexGenEnabled & T_BIT) { + switch (texUnit->GenModeT) { + case GL_OBJECT_LINEAR: + texcoord[1] = DOT4(vObj, texUnit->ObjectPlaneT); + break; + case GL_EYE_LINEAR: + texcoord[1] = DOT4(vEye, texUnit->EyePlaneT); + break; + case GL_SPHERE_MAP: + texcoord[1] = ry * mInv + 0.5F; + break; + case GL_REFLECTION_MAP: + texcoord[1] = ry; + break; + case GL_NORMAL_MAP: + texcoord[1] = normal[1]; + break; + default: + _mesa_problem(ctx, "Bad T texgen in compute_texgen()"); + return; + } + } + + if (texUnit->TexGenEnabled & R_BIT) { + switch (texUnit->GenModeR) { + case GL_OBJECT_LINEAR: + texcoord[2] = DOT4(vObj, texUnit->ObjectPlaneR); + break; + case GL_EYE_LINEAR: + texcoord[2] = DOT4(vEye, texUnit->EyePlaneR); + break; + case GL_REFLECTION_MAP: + texcoord[2] = rz; + break; + case GL_NORMAL_MAP: + texcoord[2] = normal[2]; + break; + default: + _mesa_problem(ctx, "Bad R texgen in compute_texgen()"); + return; + } + } + + if (texUnit->TexGenEnabled & Q_BIT) { + switch (texUnit->GenModeQ) { + case GL_OBJECT_LINEAR: + texcoord[3] = DOT4(vObj, texUnit->ObjectPlaneQ); + break; + case GL_EYE_LINEAR: + texcoord[3] = DOT4(vEye, texUnit->EyePlaneQ); + break; + default: + _mesa_problem(ctx, "Bad Q texgen in compute_texgen()"); + return; + } + } +} + + +/** + * glRasterPos transformation. Typically called via ctx->Driver.RasterPos(). + * XXX some of this code (such as viewport xform, clip testing and setting + * of ctx->Current.Raster* fields) could get lifted up into the + * main/rasterpos.c code. + * + * \param vObj vertex position in object space + */ +void +_tnl_RasterPos(GLcontext *ctx, const GLfloat vObj[4]) +{ + ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + FLUSH_CURRENT(ctx, 0); + + if (ctx->VertexProgram._Enabled) { + /* XXX implement this */ + _mesa_problem(ctx, "Vertex programs not implemented for glRasterPos"); + return; + } + else { + GLfloat eye[4], clip[4], ndc[3], d; + GLfloat *norm, eyenorm[3]; + GLfloat *objnorm = ctx->Current.Attrib[VERT_ATTRIB_NORMAL]; + + /* apply modelview matrix: eye = MV * obj */ + TRANSFORM_POINT( eye, ctx->ModelviewMatrixStack.Top->m, vObj ); + /* apply projection matrix: clip = Proj * eye */ + TRANSFORM_POINT( clip, ctx->ProjectionMatrixStack.Top->m, eye ); + + /* clip to view volume */ + if (ctx->Transform.RasterPositionUnclipped) { + /* GL_IBM_rasterpos_clip: only clip against Z */ + if (viewclip_point_z(clip) == 0) { + ctx->Current.RasterPosValid = GL_FALSE; + return; + } + } + else if (viewclip_point(clip) == 0) { + /* Normal OpenGL behaviour */ + ctx->Current.RasterPosValid = GL_FALSE; + return; + } + + /* clip to user clipping planes */ + if (ctx->Transform.ClipPlanesEnabled && !userclip_point(ctx, clip)) { + ctx->Current.RasterPosValid = GL_FALSE; + return; + } + + /* ndc = clip / W */ + d = (clip[3] == 0.0F) ? 1.0F : 1.0F / clip[3]; + ndc[0] = clip[0] * d; + ndc[1] = clip[1] * d; + ndc[2] = clip[2] * d; + /* wincoord = viewport_mapping(ndc) */ + ctx->Current.RasterPos[0] = (ndc[0] * ctx->Viewport._WindowMap.m[MAT_SX] + + ctx->Viewport._WindowMap.m[MAT_TX]); + ctx->Current.RasterPos[1] = (ndc[1] * ctx->Viewport._WindowMap.m[MAT_SY] + + ctx->Viewport._WindowMap.m[MAT_TY]); + ctx->Current.RasterPos[2] = (ndc[2] * ctx->Viewport._WindowMap.m[MAT_SZ] + + ctx->Viewport._WindowMap.m[MAT_TZ]) + / ctx->DrawBuffer->_DepthMaxF; + ctx->Current.RasterPos[3] = clip[3]; + + /* compute raster distance */ + if (ctx->Fog.FogCoordinateSource == GL_FOG_COORDINATE_EXT) + ctx->Current.RasterDistance = ctx->Current.Attrib[VERT_ATTRIB_FOG][0]; + else + ctx->Current.RasterDistance = + SQRTF( eye[0]*eye[0] + eye[1]*eye[1] + eye[2]*eye[2] ); + + /* compute transformed normal vector (for lighting or texgen) */ + if (ctx->_NeedEyeCoords) { + const GLfloat *inv = ctx->ModelviewMatrixStack.Top->inv; + TRANSFORM_NORMAL( eyenorm, objnorm, inv ); + norm = eyenorm; + } + else { + norm = objnorm; + } + + /* update raster color */ + if (ctx->Light.Enabled) { + /* lighting */ + shade_rastpos( ctx, vObj, norm, + ctx->Current.RasterColor, + ctx->Current.RasterSecondaryColor, + &ctx->Current.RasterIndex ); + } + else { + /* use current color or index */ + if (ctx->Visual.rgbMode) { + COPY_4FV(ctx->Current.RasterColor, + ctx->Current.Attrib[VERT_ATTRIB_COLOR0]); + COPY_4FV(ctx->Current.RasterSecondaryColor, + ctx->Current.Attrib[VERT_ATTRIB_COLOR1]); + } + else { + ctx->Current.RasterIndex + = ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX][0]; + } + } + + /* texture coords */ + { + GLuint u; + for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) { + GLfloat tc[4]; + COPY_4V(tc, ctx->Current.Attrib[VERT_ATTRIB_TEX0 + u]); + if (ctx->Texture.Unit[u].TexGenEnabled) { + compute_texgen(ctx, vObj, eye, norm, u, tc); + } + TRANSFORM_POINT(ctx->Current.RasterTexCoords[u], + ctx->TextureMatrixStack[u].Top->m, tc); + } + } + + ctx->Current.RasterPosValid = GL_TRUE; + } + + if (ctx->RenderMode == GL_SELECT) { + _mesa_update_hitflag( ctx, ctx->Current.RasterPos[2] ); + } +} -- cgit v1.2.3 From 4c01d498fac14bba751dd87bff235efb5409dca9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 6 Sep 2007 17:02:07 -0600 Subject: Move guts of glRasterPos down into T&L module. --- src/mesa/drivers/common/driverfuncs.c | 1 + src/mesa/main/dd.h | 5 + src/mesa/main/rastpos.c | 564 ++-------------------------------- src/mesa/sources | 1 + src/mesa/tnl/tnl.h | 3 + 5 files changed, 44 insertions(+), 530 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c index 9b1c3f1e06..500dbb2545 100644 --- a/src/mesa/drivers/common/driverfuncs.c +++ b/src/mesa/drivers/common/driverfuncs.c @@ -79,6 +79,7 @@ _mesa_init_driver_functions(struct dd_function_table *driver) /* framebuffer/image functions */ driver->Clear = _swrast_Clear; driver->Accum = _swrast_Accum; + driver->RasterPos = _tnl_RasterPos; driver->DrawPixels = _swrast_DrawPixels; driver->ReadPixels = _swrast_ReadPixels; driver->CopyPixels = _swrast_CopyPixels; diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index caa50dd682..582e0c334d 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -110,6 +110,11 @@ struct dd_function_table { void (*Accum)( GLcontext *ctx, GLenum op, GLfloat value ); + /** + * Execute glRasterPos, updating the ctx->Current.Raster fields + */ + void (*RasterPos)( GLcontext *ctx, const GLfloat v[4] ); + /** * \name Image-related functions */ diff --git a/src/mesa/main/rastpos.c b/src/mesa/main/rastpos.c index 2d71014a18..eef3b09d6d 100644 --- a/src/mesa/main/rastpos.c +++ b/src/mesa/main/rastpos.c @@ -29,674 +29,178 @@ */ #include "glheader.h" -#include "colormac.h" #include "context.h" #include "feedback.h" -#include "light.h" #include "macros.h" #include "rastpos.h" #include "state.h" -#include "simple_list.h" -#include "mtypes.h" - -#include "math/m_matrix.h" - - -/** - * Clip a point against the view volume. - * - * \param v vertex vector describing the point to clip. - * - * \return zero if outside view volume, or one if inside. - */ -static GLuint -viewclip_point( const GLfloat v[] ) -{ - if ( v[0] > v[3] || v[0] < -v[3] - || v[1] > v[3] || v[1] < -v[3] - || v[2] > v[3] || v[2] < -v[3] ) { - return 0; - } - else { - return 1; - } -} - - -/** - * Clip a point against the far/near Z clipping planes. - * - * \param v vertex vector describing the point to clip. - * - * \return zero if outside view volume, or one if inside. - */ -static GLuint -viewclip_point_z( const GLfloat v[] ) -{ - if (v[2] > v[3] || v[2] < -v[3] ) { - return 0; - } - else { - return 1; - } -} - - -/** - * Clip a point against the user clipping planes. - * - * \param ctx GL context. - * \param v vertex vector describing the point to clip. - * - * \return zero if the point was clipped, or one otherwise. - */ -static GLuint -userclip_point( GLcontext *ctx, const GLfloat v[] ) -{ - GLuint p; - - for (p = 0; p < ctx->Const.MaxClipPlanes; p++) { - if (ctx->Transform.ClipPlanesEnabled & (1 << p)) { - GLfloat dot = v[0] * ctx->Transform._ClipUserPlane[p][0] - + v[1] * ctx->Transform._ClipUserPlane[p][1] - + v[2] * ctx->Transform._ClipUserPlane[p][2] - + v[3] * ctx->Transform._ClipUserPlane[p][3]; - if (dot < 0.0F) { - return 0; - } - } - } - - return 1; -} - - -/** - * Compute lighting for the raster position. Both RGB and CI modes computed. - * \param ctx the context - * \param vertex vertex location - * \param normal normal vector - * \param Rcolor returned color - * \param Rspec returned specular color (if separate specular enabled) - * \param Rindex returned color index - */ -static void -shade_rastpos(GLcontext *ctx, - const GLfloat vertex[4], - const GLfloat normal[3], - GLfloat Rcolor[4], - GLfloat Rspec[4], - GLfloat *Rindex) -{ - /*const*/ GLfloat (*base)[3] = ctx->Light._BaseColor; - const struct gl_light *light; - GLfloat diffuseColor[4], specularColor[4]; /* for RGB mode only */ - GLfloat diffuseCI = 0.0, specularCI = 0.0; /* for CI mode only */ - - _mesa_validate_all_lighting_tables( ctx ); - - COPY_3V(diffuseColor, base[0]); - diffuseColor[3] = CLAMP( - ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3], 0.0F, 1.0F ); - ASSIGN_4V(specularColor, 0.0, 0.0, 0.0, 1.0); - - foreach (light, &ctx->Light.EnabledList) { - GLfloat attenuation = 1.0; - GLfloat VP[3]; /* vector from vertex to light pos */ - GLfloat n_dot_VP; - GLfloat diffuseContrib[3], specularContrib[3]; - - if (!(light->_Flags & LIGHT_POSITIONAL)) { - /* light at infinity */ - COPY_3V(VP, light->_VP_inf_norm); - attenuation = light->_VP_inf_spot_attenuation; - } - else { - /* local/positional light */ - GLfloat d; - - /* VP = vector from vertex pos to light[i].pos */ - SUB_3V(VP, light->_Position, vertex); - /* d = length(VP) */ - d = (GLfloat) LEN_3FV( VP ); - if (d > 1.0e-6) { - /* normalize VP */ - GLfloat invd = 1.0F / d; - SELF_SCALE_SCALAR_3V(VP, invd); - } - - /* atti */ - attenuation = 1.0F / (light->ConstantAttenuation + d * - (light->LinearAttenuation + d * - light->QuadraticAttenuation)); - - if (light->_Flags & LIGHT_SPOT) { - GLfloat PV_dot_dir = - DOT3(VP, light->_NormDirection); - - if (PV_dot_dir_CosCutoff) { - continue; - } - else { - double x = PV_dot_dir * (EXP_TABLE_SIZE-1); - int k = (int) x; - GLfloat spot = (GLfloat) (light->_SpotExpTable[k][0] - + (x-k)*light->_SpotExpTable[k][1]); - attenuation *= spot; - } - } - } - - if (attenuation < 1e-3) - continue; - - n_dot_VP = DOT3( normal, VP ); - - if (n_dot_VP < 0.0F) { - ACC_SCALE_SCALAR_3V(diffuseColor, attenuation, light->_MatAmbient[0]); - continue; - } - - /* Ambient + diffuse */ - COPY_3V(diffuseContrib, light->_MatAmbient[0]); - ACC_SCALE_SCALAR_3V(diffuseContrib, n_dot_VP, light->_MatDiffuse[0]); - diffuseCI += n_dot_VP * light->_dli * attenuation; - - /* Specular */ - { - const GLfloat *h; - GLfloat n_dot_h; - - ASSIGN_3V(specularContrib, 0.0, 0.0, 0.0); - - if (ctx->Light.Model.LocalViewer) { - GLfloat v[3]; - COPY_3V(v, vertex); - NORMALIZE_3FV(v); - SUB_3V(VP, VP, v); - NORMALIZE_3FV(VP); - h = VP; - } - else if (light->_Flags & LIGHT_POSITIONAL) { - ACC_3V(VP, ctx->_EyeZDir); - NORMALIZE_3FV(VP); - h = VP; - } - else { - h = light->_h_inf_norm; - } - - n_dot_h = DOT3(normal, h); - - if (n_dot_h > 0.0F) { - GLfloat spec_coef; - GET_SHINE_TAB_ENTRY( ctx->_ShineTable[0], n_dot_h, spec_coef ); - - if (spec_coef > 1.0e-10) { - if (ctx->Light.Model.ColorControl==GL_SEPARATE_SPECULAR_COLOR) { - ACC_SCALE_SCALAR_3V( specularContrib, spec_coef, - light->_MatSpecular[0]); - } - else { - ACC_SCALE_SCALAR_3V( diffuseContrib, spec_coef, - light->_MatSpecular[0]); - } - /*assert(light->_sli > 0.0);*/ - specularCI += spec_coef * light->_sli * attenuation; - } - } - } - - ACC_SCALE_SCALAR_3V( diffuseColor, attenuation, diffuseContrib ); - ACC_SCALE_SCALAR_3V( specularColor, attenuation, specularContrib ); - } - - if (ctx->Visual.rgbMode) { - Rcolor[0] = CLAMP(diffuseColor[0], 0.0F, 1.0F); - Rcolor[1] = CLAMP(diffuseColor[1], 0.0F, 1.0F); - Rcolor[2] = CLAMP(diffuseColor[2], 0.0F, 1.0F); - Rcolor[3] = CLAMP(diffuseColor[3], 0.0F, 1.0F); - Rspec[0] = CLAMP(specularColor[0], 0.0F, 1.0F); - Rspec[1] = CLAMP(specularColor[1], 0.0F, 1.0F); - Rspec[2] = CLAMP(specularColor[2], 0.0F, 1.0F); - Rspec[3] = CLAMP(specularColor[3], 0.0F, 1.0F); - } - else { - GLfloat *ind = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_INDEXES]; - GLfloat d_a = ind[MAT_INDEX_DIFFUSE] - ind[MAT_INDEX_AMBIENT]; - GLfloat s_a = ind[MAT_INDEX_SPECULAR] - ind[MAT_INDEX_AMBIENT]; - GLfloat i = (ind[MAT_INDEX_AMBIENT] - + diffuseCI * (1.0F-specularCI) * d_a - + specularCI * s_a); - if (i > ind[MAT_INDEX_SPECULAR]) { - i = ind[MAT_INDEX_SPECULAR]; - } - *Rindex = i; - } -} /** - * Do texgen needed for glRasterPos. - * \param ctx rendering context - * \param vObj object-space vertex coordinate - * \param vEye eye-space vertex coordinate - * \param normal vertex normal - * \param unit texture unit number - * \param texcoord incoming texcoord and resulting texcoord - */ -static void -compute_texgen(GLcontext *ctx, const GLfloat vObj[4], const GLfloat vEye[4], - const GLfloat normal[3], GLuint unit, GLfloat texcoord[4]) -{ - const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - - /* always compute sphere map terms, just in case */ - GLfloat u[3], two_nu, rx, ry, rz, m, mInv; - COPY_3V(u, vEye); - NORMALIZE_3FV(u); - two_nu = 2.0F * DOT3(normal, u); - rx = u[0] - normal[0] * two_nu; - ry = u[1] - normal[1] * two_nu; - rz = u[2] - normal[2] * two_nu; - m = rx * rx + ry * ry + (rz + 1.0F) * (rz + 1.0F); - if (m > 0.0F) - mInv = 0.5F * _mesa_inv_sqrtf(m); - else - mInv = 0.0F; - - if (texUnit->TexGenEnabled & S_BIT) { - switch (texUnit->GenModeS) { - case GL_OBJECT_LINEAR: - texcoord[0] = DOT4(vObj, texUnit->ObjectPlaneS); - break; - case GL_EYE_LINEAR: - texcoord[0] = DOT4(vEye, texUnit->EyePlaneS); - break; - case GL_SPHERE_MAP: - texcoord[0] = rx * mInv + 0.5F; - break; - case GL_REFLECTION_MAP: - texcoord[0] = rx; - break; - case GL_NORMAL_MAP: - texcoord[0] = normal[0]; - break; - default: - _mesa_problem(ctx, "Bad S texgen in compute_texgen()"); - return; - } - } - - if (texUnit->TexGenEnabled & T_BIT) { - switch (texUnit->GenModeT) { - case GL_OBJECT_LINEAR: - texcoord[1] = DOT4(vObj, texUnit->ObjectPlaneT); - break; - case GL_EYE_LINEAR: - texcoord[1] = DOT4(vEye, texUnit->EyePlaneT); - break; - case GL_SPHERE_MAP: - texcoord[1] = ry * mInv + 0.5F; - break; - case GL_REFLECTION_MAP: - texcoord[1] = ry; - break; - case GL_NORMAL_MAP: - texcoord[1] = normal[1]; - break; - default: - _mesa_problem(ctx, "Bad T texgen in compute_texgen()"); - return; - } - } - - if (texUnit->TexGenEnabled & R_BIT) { - switch (texUnit->GenModeR) { - case GL_OBJECT_LINEAR: - texcoord[2] = DOT4(vObj, texUnit->ObjectPlaneR); - break; - case GL_EYE_LINEAR: - texcoord[2] = DOT4(vEye, texUnit->EyePlaneR); - break; - case GL_REFLECTION_MAP: - texcoord[2] = rz; - break; - case GL_NORMAL_MAP: - texcoord[2] = normal[2]; - break; - default: - _mesa_problem(ctx, "Bad R texgen in compute_texgen()"); - return; - } - } - - if (texUnit->TexGenEnabled & Q_BIT) { - switch (texUnit->GenModeQ) { - case GL_OBJECT_LINEAR: - texcoord[3] = DOT4(vObj, texUnit->ObjectPlaneQ); - break; - case GL_EYE_LINEAR: - texcoord[3] = DOT4(vEye, texUnit->EyePlaneQ); - break; - default: - _mesa_problem(ctx, "Bad Q texgen in compute_texgen()"); - return; - } - } -} - - - -/** - * Set the raster position for pixel operations. - * - * All glRasterPos command call this function to update the current - * raster position. - * - * \param ctx GL context. - * \param x x coordinate for the raster position. - * \param y y coordinate for the raster position. - * \param z z coordinate for the raster position. - * \param w w coordinate for the raster position. - * - * \sa Called by _mesa_RasterPos4f(). - * - * Flushes the vertices, transforms and clips the vertex coordinates, and - * finally sets the current raster position and associated data in - * __GLcontextRec::Current. When in selection mode calls - * _mesa_update_hitflag() with the current raster position. + * Helper function for all the RasterPos functions. */ static void -raster_pos4f(GLcontext *ctx, GLfloat x, GLfloat y, GLfloat z, GLfloat w) +rasterpos(GLfloat x, GLfloat y, GLfloat z, GLfloat w) { - ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); - FLUSH_CURRENT(ctx, 0); + GET_CURRENT_CONTEXT(ctx); + GLfloat p[4]; + + p[0] = x; + p[1] = y; + p[2] = z; + p[3] = w; if (ctx->NewState) _mesa_update_state( ctx ); - if (ctx->VertexProgram._Enabled) { - /* XXX implement this */ - _mesa_problem(ctx, "Vertex programs not implemented for glRasterPos"); - return; - } - else { - GLfloat obj[4], eye[4], clip[4], ndc[3], d; - GLfloat *norm, eyenorm[3]; - GLfloat *objnorm = ctx->Current.Attrib[VERT_ATTRIB_NORMAL]; - - ASSIGN_4V( obj, x, y, z, w ); - /* apply modelview matrix: eye = MV * obj */ - TRANSFORM_POINT( eye, ctx->ModelviewMatrixStack.Top->m, obj ); - /* apply projection matrix: clip = Proj * eye */ - TRANSFORM_POINT( clip, ctx->ProjectionMatrixStack.Top->m, eye ); - - /* clip to view volume */ - if (ctx->Transform.RasterPositionUnclipped) { - /* GL_IBM_rasterpos_clip: only clip against Z */ - if (viewclip_point_z(clip) == 0) { - ctx->Current.RasterPosValid = GL_FALSE; - return; - } - } - else if (viewclip_point(clip) == 0) { - /* Normal OpenGL behaviour */ - ctx->Current.RasterPosValid = GL_FALSE; - return; - } - - /* clip to user clipping planes */ - if (ctx->Transform.ClipPlanesEnabled && !userclip_point(ctx, clip)) { - ctx->Current.RasterPosValid = GL_FALSE; - return; - } - - /* ndc = clip / W */ - d = (clip[3] == 0.0F) ? 1.0F : 1.0F / clip[3]; - ndc[0] = clip[0] * d; - ndc[1] = clip[1] * d; - ndc[2] = clip[2] * d; - /* wincoord = viewport_mapping(ndc) */ - ctx->Current.RasterPos[0] = (ndc[0] * ctx->Viewport._WindowMap.m[MAT_SX] - + ctx->Viewport._WindowMap.m[MAT_TX]); - ctx->Current.RasterPos[1] = (ndc[1] * ctx->Viewport._WindowMap.m[MAT_SY] - + ctx->Viewport._WindowMap.m[MAT_TY]); - ctx->Current.RasterPos[2] = (ndc[2] * ctx->Viewport._WindowMap.m[MAT_SZ] - + ctx->Viewport._WindowMap.m[MAT_TZ]) - / ctx->DrawBuffer->_DepthMaxF; - ctx->Current.RasterPos[3] = clip[3]; - - /* compute raster distance */ - if (ctx->Fog.FogCoordinateSource == GL_FOG_COORDINATE_EXT) - ctx->Current.RasterDistance = ctx->Current.Attrib[VERT_ATTRIB_FOG][0]; - else - ctx->Current.RasterDistance = - SQRTF( eye[0]*eye[0] + eye[1]*eye[1] + eye[2]*eye[2] ); - - /* compute transformed normal vector (for lighting or texgen) */ - if (ctx->_NeedEyeCoords) { - const GLfloat *inv = ctx->ModelviewMatrixStack.Top->inv; - TRANSFORM_NORMAL( eyenorm, objnorm, inv ); - norm = eyenorm; - } - else { - norm = objnorm; - } - - /* update raster color */ - if (ctx->Light.Enabled) { - /* lighting */ - shade_rastpos( ctx, obj, norm, - ctx->Current.RasterColor, - ctx->Current.RasterSecondaryColor, - &ctx->Current.RasterIndex ); - } - else { - /* use current color or index */ - if (ctx->Visual.rgbMode) { - COPY_4FV(ctx->Current.RasterColor, - ctx->Current.Attrib[VERT_ATTRIB_COLOR0]); - COPY_4FV(ctx->Current.RasterSecondaryColor, - ctx->Current.Attrib[VERT_ATTRIB_COLOR1]); - } - else { - ctx->Current.RasterIndex - = ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX][0]; - } - } - - /* texture coords */ - { - GLuint u; - for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) { - GLfloat tc[4]; - COPY_4V(tc, ctx->Current.Attrib[VERT_ATTRIB_TEX0 + u]); - if (ctx->Texture.Unit[u].TexGenEnabled) { - compute_texgen(ctx, obj, eye, norm, u, tc); - } - TRANSFORM_POINT(ctx->Current.RasterTexCoords[u], - ctx->TextureMatrixStack[u].Top->m, tc); - } - } - - ctx->Current.RasterPosValid = GL_TRUE; - } - - if (ctx->RenderMode == GL_SELECT) { - _mesa_update_hitflag( ctx, ctx->Current.RasterPos[2] ); - } + ctx->Driver.RasterPos(ctx, p); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos2d(GLdouble x, GLdouble y) { - _mesa_RasterPos4f((GLfloat) x, (GLfloat) y, 0.0F, 1.0F); + rasterpos(x, y, 0.0F, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos2f(GLfloat x, GLfloat y) { - _mesa_RasterPos4f(x, y, 0.0F, 1.0F); + rasterpos(x, y, 0.0F, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos2i(GLint x, GLint y) { - _mesa_RasterPos4f((GLfloat) x, (GLfloat) y, 0.0F, 1.0F); + rasterpos((GLfloat) x, (GLfloat) y, 0.0F, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos2s(GLshort x, GLshort y) { - _mesa_RasterPos4f(x, y, 0.0F, 1.0F); + rasterpos(x, y, 0.0F, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos3d(GLdouble x, GLdouble y, GLdouble z) { - _mesa_RasterPos4f((GLfloat) x, (GLfloat) y, (GLfloat) z, 1.0F); + rasterpos((GLfloat) x, (GLfloat) y, (GLfloat) z, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos3f(GLfloat x, GLfloat y, GLfloat z) { - _mesa_RasterPos4f(x, y, z, 1.0F); + rasterpos(x, y, z, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos3i(GLint x, GLint y, GLint z) { - _mesa_RasterPos4f((GLfloat) x, (GLfloat) y, (GLfloat) z, 1.0F); + rasterpos((GLfloat) x, (GLfloat) y, (GLfloat) z, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos3s(GLshort x, GLshort y, GLshort z) { - _mesa_RasterPos4f(x, y, z, 1.0F); + rasterpos(x, y, z, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos4d(GLdouble x, GLdouble y, GLdouble z, GLdouble w) { - _mesa_RasterPos4f((GLfloat) x, (GLfloat) y, (GLfloat) z, (GLfloat) w); + rasterpos((GLfloat) x, (GLfloat) y, (GLfloat) z, (GLfloat) w); } -/** Calls raster_pos4f() */ void GLAPIENTRY _mesa_RasterPos4f(GLfloat x, GLfloat y, GLfloat z, GLfloat w) { - GET_CURRENT_CONTEXT(ctx); - raster_pos4f(ctx, x, y, z, w); + rasterpos(x, y, z, w); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos4i(GLint x, GLint y, GLint z, GLint w) { - _mesa_RasterPos4f((GLfloat) x, (GLfloat) y, (GLfloat) z, (GLfloat) w); + rasterpos((GLfloat) x, (GLfloat) y, (GLfloat) z, (GLfloat) w); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos4s(GLshort x, GLshort y, GLshort z, GLshort w) { - _mesa_RasterPos4f(x, y, z, w); + rasterpos(x, y, z, w); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos2dv(const GLdouble *v) { - _mesa_RasterPos4f((GLfloat) v[0], (GLfloat) v[1], 0.0F, 1.0F); + rasterpos((GLfloat) v[0], (GLfloat) v[1], 0.0F, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos2fv(const GLfloat *v) { - _mesa_RasterPos4f(v[0], v[1], 0.0F, 1.0F); + rasterpos(v[0], v[1], 0.0F, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos2iv(const GLint *v) { - _mesa_RasterPos4f((GLfloat) v[0], (GLfloat) v[1], 0.0F, 1.0F); + rasterpos((GLfloat) v[0], (GLfloat) v[1], 0.0F, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos2sv(const GLshort *v) { - _mesa_RasterPos4f(v[0], v[1], 0.0F, 1.0F); + rasterpos(v[0], v[1], 0.0F, 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos3dv(const GLdouble *v) { - _mesa_RasterPos4f((GLfloat) v[0], (GLfloat) v[1], (GLfloat) v[2], 1.0F); + rasterpos((GLfloat) v[0], (GLfloat) v[1], (GLfloat) v[2], 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos3fv(const GLfloat *v) { - _mesa_RasterPos4f(v[0], v[1], v[2], 1.0F); + rasterpos(v[0], v[1], v[2], 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos3iv(const GLint *v) { - _mesa_RasterPos4f((GLfloat) v[0], (GLfloat) v[1], (GLfloat) v[2], 1.0F); + rasterpos((GLfloat) v[0], (GLfloat) v[1], (GLfloat) v[2], 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos3sv(const GLshort *v) { - _mesa_RasterPos4f(v[0], v[1], v[2], 1.0F); + rasterpos(v[0], v[1], v[2], 1.0F); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos4dv(const GLdouble *v) { - _mesa_RasterPos4f((GLfloat) v[0], (GLfloat) v[1], + rasterpos((GLfloat) v[0], (GLfloat) v[1], (GLfloat) v[2], (GLfloat) v[3]); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos4fv(const GLfloat *v) { - _mesa_RasterPos4f(v[0], v[1], v[2], v[3]); + rasterpos(v[0], v[1], v[2], v[3]); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos4iv(const GLint *v) { - _mesa_RasterPos4f((GLfloat) v[0], (GLfloat) v[1], + rasterpos((GLfloat) v[0], (GLfloat) v[1], (GLfloat) v[2], (GLfloat) v[3]); } -/** Calls _mesa_RasterPos4f() */ void GLAPIENTRY _mesa_RasterPos4sv(const GLshort *v) { - _mesa_RasterPos4f(v[0], v[1], v[2], v[3]); + rasterpos(v[0], v[1], v[2], v[3]); } diff --git a/src/mesa/sources b/src/mesa/sources index ae0797ae24..437af0f57b 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -118,6 +118,7 @@ TNL_SOURCES = \ tnl/t_context.c \ tnl/t_pipeline.c \ tnl/t_draw.c \ + tnl/t_rasterpos.c \ tnl/t_vb_program.c \ tnl/t_vb_render.c \ tnl/t_vb_texgen.c \ diff --git a/src/mesa/tnl/tnl.h b/src/mesa/tnl/tnl.h index 047b764dcb..4bdbed92df 100644 --- a/src/mesa/tnl/tnl.h +++ b/src/mesa/tnl/tnl.h @@ -85,4 +85,7 @@ _tnl_draw_prims( GLcontext *ctx, extern void _mesa_load_tracked_matrices(GLcontext *ctx); +extern void +_tnl_RasterPos(GLcontext *ctx, const GLfloat vObj[4]); + #endif -- cgit v1.2.3 From 41ddd76b98e9e236a796f1bb9480ea0b2fb18010 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 6 Sep 2007 17:03:54 -0600 Subject: Initial work for post-transform vertex feedback buffers. This will be used for: GL_NV_transform_feedback, or GL3 equivalent. RasterPos GL select/feedback modes --- src/mesa/pipe/draw/draw_feedback.c | 247 +++++++++++++++++++++++++++++++++++++ 1 file changed, 247 insertions(+) create mode 100644 src/mesa/pipe/draw/draw_feedback.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_feedback.c b/src/mesa/pipe/draw/draw_feedback.c new file mode 100644 index 0000000000..2729f44dc6 --- /dev/null +++ b/src/mesa/pipe/draw/draw_feedback.c @@ -0,0 +1,247 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Primitive/vertex feedback (and/or discard) stage. + * Used to implement transformation feedback/streaming and other things + * which require a post-transformed vertex position (such as rasterpos, + * selection and feedback modes). + * + * Authors: + * Brian Paul + */ + + +#include "pipe/p_util.h" +#include "draw_private.h" + + +struct feedback_stage { + struct draw_stage stage; /**< base class */ + uint num_prim_generated; /**< number of primitives received */ + uint num_prim_emitted; /**< number of primitives fed back */ + uint num_vert_emitted; /**< number of vertices fed back */ + uint max_vert_emit; /**< max number of verts we can emit */ + float *dest[PIPE_MAX_FEEDBACK_ATTRIBS]; /**< dests for vertex attribs */ +}; + + + +/** + * Check if there's space to store 'numVerts' in the feedback buffer(s). + */ +static boolean +check_space(const struct draw_stage *stage, uint numVerts) +{ + const struct feedback_stage *fs = (struct feedback_stage *) stage; + return fs->num_vert_emitted + numVerts <= fs->max_vert_emit; +} + + +/** + * Record the given vertex's attributes into the feedback buffer(s). + */ +static void +feedback_vertex(struct draw_stage *stage, const struct vertex_header *vertex) +{ + struct feedback_stage *fs = (struct feedback_stage *) stage; + const struct pipe_feedback_state *feedback = &stage->draw->feedback; + const uint select = feedback->interleaved ? 0 : 1; + uint i; + + /* + * Note: 'select' is either 0 or 1. By multiplying 'i' by 'select' + * we can either address output buffer 0 (for interleaving) or + * output buffer i (for non-interleaved). + */ + + for (i = 0; i < feedback->num_attribs; i++) { + const uint attr = feedback->attrib[i]; + const float *src = attr ? vertex->data[attr] : vertex->clip; + const uint size = feedback->size[i]; + float *dest = fs->dest[i * select]; + + switch (size) { + case 4: + dest[3] = src[3]; + /* fall-through */ + case 3: + dest[2] = src[2]; + /* fall-through */ + case 2: + dest[1] = src[1]; + /* fall-through */ + case 1: + dest[0] = src[0]; + /* fall-through */ + default: + ; + } + fs->dest[i * select] += size; + } + + fs->num_vert_emitted++; +} + + +static void feedback_begin( struct draw_stage *stage ) +{ + struct feedback_stage *fs = (struct feedback_stage *) stage; + const struct pipe_feedback_state *feedback = &stage->draw->feedback; + + fs->num_prim_generated = 0; + fs->num_prim_emitted = 0; + fs->num_vert_emitted = 0; + + assert(feedback->enabled); + + /* Compute max_vert_emit, the max number of vertices we can emit. + * And, setup dest[] pointers. + */ + if (stage->draw->feedback.interleaved) { + uint i, vertex_size = 0; + /* compute size of each interleaved vertex, in floats */ + for (i = 0; i < feedback->num_attribs; i++) { + vertex_size += feedback->size[i]; + } + /* compute max number of vertices we can feedback */ + fs->max_vert_emit = stage->draw->mapped_feedback_buffer_size[0] + / sizeof(float) / vertex_size; + + fs->dest[0] = (float *) stage->draw->mapped_feedback_buffer[0]; + } + else { + uint i; + uint max = ~0; + for (i = 0; i < feedback->num_attribs; i++) { + uint n = stage->draw->mapped_feedback_buffer_size[i] + / sizeof(float) / feedback->size[i]; + if (n < max) + max = n; + fs->dest[i] = (float *) stage->draw->mapped_feedback_buffer[i]; + } + fs->max_vert_emit = max; + } + + if (!feedback->discard) + stage->next->begin( stage->next ); +} + + +static void feedback_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct feedback_stage *fs = (struct feedback_stage *) stage; + + fs->num_prim_generated++; + + if (stage->draw->feedback.enabled && check_space(stage, 3)) { + feedback_vertex(stage, header->v[0]); + feedback_vertex(stage, header->v[1]); + feedback_vertex(stage, header->v[2]); + fs->num_prim_emitted++; + } + + if (!stage->draw->feedback.discard) + stage->next->tri( stage->next, header ); +} + + +static void feedback_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct feedback_stage *fs = (struct feedback_stage *) stage; + + fs->num_prim_generated++; + + if (stage->draw->feedback.enabled && check_space(stage, 2)) { + feedback_vertex(stage, header->v[0]); + feedback_vertex(stage, header->v[1]); + fs->num_prim_emitted++; + } + + if (!stage->draw->feedback.discard) + stage->next->line( stage->next, header ); +} + + +static void feedback_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct feedback_stage *fs = (struct feedback_stage *) stage; + + fs->num_prim_generated++; + + if (stage->draw->feedback.enabled && check_space(stage, 1)) { + feedback_vertex(stage, header->v[0]); + fs->num_prim_emitted++; + } + + if (!stage->draw->feedback.discard) + stage->next->point( stage->next, header ); +} + + +static void feedback_end( struct draw_stage *stage ) +{ + + /* XXX Unmap the vertex feedback buffers so we can write to them */ + + + if (!stage->draw->feedback.discard) + stage->next->end( stage->next ); +} + + + +static void feedback_reset_stipple_counter( struct draw_stage *stage ) +{ + if (!stage->draw->feedback.discard) + stage->next->reset_stipple_counter( stage->next ); +} + + +/** + * Create feedback drawing stage. + */ +struct draw_stage *draw_feedback_stage( struct draw_context *draw ) +{ + struct feedback_stage *feedback = CALLOC_STRUCT(feedback_stage); + + feedback->stage.draw = draw; + feedback->stage.next = NULL; + feedback->stage.begin = feedback_begin; + feedback->stage.point = feedback_point; + feedback->stage.line = feedback_line; + feedback->stage.tri = feedback_tri; + feedback->stage.end = feedback_end; + feedback->stage.reset_stipple_counter = feedback_reset_stipple_counter; + + return &feedback->stage; +} + + -- cgit v1.2.3 From 63a51aee86f76ae54fb1ce572d584fb0f01aaec5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 6 Sep 2007 17:07:09 -0600 Subject: Initial work for post-transformed vertex feedback buffers. This will be used for: GL_NV_transform_feedback, or similar GL3 functionality glRasterPos GL selection/feedback modes --- src/mesa/pipe/draw/draw_context.c | 42 +++++++++++++++++++++++++++++++++++---- src/mesa/pipe/draw/draw_context.h | 23 ++++++++++++--------- src/mesa/pipe/draw/draw_prim.c | 17 ---------------- src/mesa/pipe/draw/draw_private.h | 9 +++++++++ src/mesa/pipe/p_context.h | 18 +++++++++++++++-- src/mesa/pipe/p_state.h | 15 ++++++++++++++ src/mesa/sources | 1 + 7 files changed, 93 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 2936a14d65..3c095810e5 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -137,7 +137,12 @@ static void validate_pipeline( struct draw_context *draw ) draw->pipeline.flatshade->next = next; next = draw->pipeline.flatshade; } - + + if (draw->feedback.enabled || draw->feedback.discard) { + draw->pipeline.feedback->next = next; + next = draw->pipeline.feedback; + } + draw->pipeline.first = next; } @@ -189,6 +194,14 @@ void draw_set_viewport_state( struct draw_context *draw, } +void +draw_set_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader) +{ + draw->vertex_shader = *shader; +} + + void draw_set_vertex_buffer(struct draw_context *draw, unsigned attr, @@ -209,9 +222,30 @@ draw_set_vertex_element(struct draw_context *draw, } +/** + * Tell drawing context where to find mapped vertex buffers. + */ void -draw_set_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader) +draw_set_mapped_vertex_buffer(struct draw_context *draw, + unsigned attr, const void *buffer) { - draw->vertex_shader = *shader; + draw->mapped_vbuffer[attr] = buffer; +} + + +void +draw_set_mapped_constant_buffer(struct draw_context *draw, + const void *buffer) +{ + draw->mapped_constants = buffer; } + + +void +draw_set_mapped_feedback_buffer(struct draw_context *draw, uint index, + void *buffer, uint size) +{ + draw->mapped_feedback_buffer[index] = buffer; + draw->mapped_feedback_buffer_size[index] = size; /* in bytes */ +} + diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 7c4f1a1297..5031574969 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -93,14 +93,11 @@ unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); unsigned draw_trim( unsigned count, unsigned first, unsigned incr ); -void draw_set_mapped_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ); -void draw_set_mapped_vertex_buffer(struct draw_context *draw, - unsigned attr, const void *buffer); +void +draw_set_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader); -void draw_set_mapped_constant_buffer(struct draw_context *draw, - const void *buffer); void draw_set_vertex_buffer(struct draw_context *draw, @@ -112,10 +109,18 @@ draw_set_vertex_element(struct draw_context *draw, unsigned attr, const struct pipe_vertex_element *element); -void -draw_set_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader); +void draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ); + +void draw_set_mapped_vertex_buffer(struct draw_context *draw, + unsigned attr, const void *buffer); +void draw_set_mapped_constant_buffer(struct draw_context *draw, + const void *buffer); + +void +draw_set_mapped_feedback_buffer(struct draw_context *draw, uint index, + void *buffer, uint size); void draw_arrays(struct draw_context *draw, unsigned prim, diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index b68cca5706..a90d1df7f9 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -353,23 +353,6 @@ draw_set_prim( struct draw_context *draw, unsigned prim ) } -/** - * Tell drawing context where to find mapped vertex buffers. - */ -void draw_set_mapped_vertex_buffer(struct draw_context *draw, - unsigned attr, const void *buffer) -{ - draw->mapped_vbuffer[attr] = buffer; -} - - -void draw_set_mapped_constant_buffer(struct draw_context *draw, - const void *buffer) -{ - draw->mapped_constants = buffer; -} - - unsigned draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) { diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 9f90dc3e26..8bcc3717c4 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -126,6 +126,7 @@ struct draw_context struct draw_stage *first; /**< one of the following */ /* stages (in logical order) */ + struct draw_stage *feedback; struct draw_stage *flatshade; struct draw_stage *clip; struct draw_stage *cull; @@ -137,10 +138,13 @@ struct draw_context /* pipe state that we need: */ struct pipe_setup_state setup; + struct pipe_feedback_state feedback; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; struct pipe_shader_state vertex_shader; + struct pipe_vertex_buffer feedback_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element feedback_element[PIPE_ATTRIB_MAX]; /** The mapped vertex element/index buffer */ const void *mapped_elts; @@ -150,6 +154,10 @@ struct draw_context /** The mapped constant buffers (for vertex shader) */ const void *mapped_constants; + /** The mapped vertex element/index buffer */ + void *mapped_feedback_buffer[PIPE_MAX_FEEDBACK_ATTRIBS]; + uint mapped_feedback_buffer_size[PIPE_MAX_FEEDBACK_ATTRIBS]; /* in bytes */ + /* Clip derived state: */ float plane[12][4]; @@ -203,6 +211,7 @@ struct draw_context +extern struct draw_stage *draw_feedback_stage( struct draw_context *context ); extern struct draw_stage *draw_unfilled_stage( struct draw_context *context ); extern struct draw_stage *draw_twoside_stage( struct draw_context *context ); extern struct draw_stage *draw_offset_stage( struct draw_context *context ); diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index ec9973383c..c9246e883a 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -138,14 +138,28 @@ struct pipe_context { void (*set_viewport_state)( struct pipe_context *, const struct pipe_viewport_state * ); + /* + * Vertex arrays + */ void (*set_vertex_buffer)( struct pipe_context *, unsigned index, const struct pipe_vertex_buffer * ); - + void (*set_vertex_element)( struct pipe_context *, unsigned index, const struct pipe_vertex_element * ); - + + /* + * Vertex feedback + */ + void (*set_feedback_buffer)(struct pipe_context *, + unsigned index, + struct pipe_vertex_buffer *); + + void (*set_feedback_element)(struct pipe_context *, + unsigned index, + const struct pipe_vertex_element *); + /* * Surface functions diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 25e5861e1e..cad96807df 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -49,6 +49,7 @@ #define PIPE_ATTRIB_MAX 32 #define PIPE_MAX_COLOR_BUFS 8 #define PIPE_MAX_TEXTURE_LEVELS 16 +#define PIPE_MAX_FEEDBACK_ATTRIBS 16 /* fwd decl */ @@ -92,6 +93,20 @@ struct pipe_setup_state float offset_scale; }; + +/** + * Post-transform vertex feeback + */ +struct pipe_feedback_state { + uint enabled:1; /**< enable feedback? */ + uint discard:1; /**< discard primitives? */ + uint interleaved:1; /**< interleaved output? */ + uint num_attribs; + uint attrib[PIPE_MAX_FEEDBACK_ATTRIBS]; + uint size[PIPE_MAX_FEEDBACK_ATTRIBS]; +}; + + struct pipe_poly_stipple { unsigned stipple[32]; }; diff --git a/src/mesa/sources b/src/mesa/sources index 437af0f57b..28f865f2ed 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -161,6 +161,7 @@ DRAW_SOURCES = \ pipe/draw/draw_clip.c \ pipe/draw/draw_context.c\ pipe/draw/draw_cull.c \ + pipe/draw/draw_feedback.c \ pipe/draw/draw_flatshade.c \ pipe/draw/draw_offset.c \ pipe/draw/draw_prim.c \ -- cgit v1.2.3 From 21c925f49191df46ed8788ac09bbb3bd4a437c99 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 16:27:07 -0600 Subject: move FLUSH_CURRENT --- src/mesa/main/rastpos.c | 3 +++ src/mesa/tnl/t_rasterpos.c | 3 --- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/rastpos.c b/src/mesa/main/rastpos.c index eef3b09d6d..ee163e0c71 100644 --- a/src/mesa/main/rastpos.c +++ b/src/mesa/main/rastpos.c @@ -53,6 +53,9 @@ rasterpos(GLfloat x, GLfloat y, GLfloat z, GLfloat w) if (ctx->NewState) _mesa_update_state( ctx ); + ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + FLUSH_CURRENT(ctx, 0); + ctx->Driver.RasterPos(ctx, p); } diff --git a/src/mesa/tnl/t_rasterpos.c b/src/mesa/tnl/t_rasterpos.c index 272f855fc1..be963867c0 100644 --- a/src/mesa/tnl/t_rasterpos.c +++ b/src/mesa/tnl/t_rasterpos.c @@ -393,9 +393,6 @@ compute_texgen(GLcontext *ctx, const GLfloat vObj[4], const GLfloat vEye[4], void _tnl_RasterPos(GLcontext *ctx, const GLfloat vObj[4]) { - ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); - FLUSH_CURRENT(ctx, 0); - if (ctx->VertexProgram._Enabled) { /* XXX implement this */ _mesa_problem(ctx, "Vertex programs not implemented for glRasterPos"); -- cgit v1.2.3 From bb193c6d22125f1af62e81c1251acd2a68939608 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 16:28:10 -0600 Subject: glRasterPos function --- src/mesa/state_tracker/st_cb_rasterpos.c | 350 +++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_rasterpos.h | 33 +++ 2 files changed, 383 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_rasterpos.c create mode 100644 src/mesa/state_tracker/st_cb_rasterpos.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c new file mode 100644 index 0000000000..5040c21e51 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -0,0 +1,350 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Brian Paul + */ + +#include "main/imports.h" +#include "main/macros.h" + +#include "st_context.h" +#include "st_atom.h" +#include "st_draw.h" +#include "st_program.h" +#include "st_cb_rasterpos.h" +#include "st_draw.h" +#include "st_format.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/tgsi/exec/tgsi_attribs.h" +#include "shader/prog_instruction.h" +#include "vf/vf.h" + + + +static void +setup_vertex_attribs(GLcontext *ctx) +{ + struct pipe_context *pipe = ctx->st->pipe; + const uint inputAttrs = ctx->st->state.vs.inputs_read; + uint attr; + + /* all attributes come from the default attribute buffer */ + { + struct pipe_vertex_buffer vbuffer; + vbuffer.buffer = ctx->st->default_attrib_buffer; + vbuffer.buffer_offset = 0; + vbuffer.pitch = 0; /* must be zero! */ + vbuffer.max_index = 1; + pipe->set_vertex_buffer(pipe, 0, &vbuffer); + } + + for (attr = 0; attr < 16; attr++) { + struct pipe_vertex_element velement; + + if (inputAttrs & (1 << attr)) { + velement.src_offset = attr * 4 * sizeof(GLfloat); + velement.vertex_buffer_index = 0; + velement.dst_offset = 0; + velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + pipe->set_vertex_element(pipe, attr, &velement); + } + } +} + + +static void +setup_feedback(GLcontext *ctx) +{ + struct pipe_context *pipe = ctx->st->pipe; + const uint outputAttrs = ctx->st->state.vs.outputs_written; + struct pipe_feedback_state feedback; + uint i; + + memset(&feedback, 0, sizeof(feedback)); + feedback.enabled = 1; + feedback.interleaved = 1; + feedback.discard = 1; + feedback.num_attribs = 0; + + for (i = 0; i < TGSI_ATTRIB_VAR0; i++) { + if ((1 << i) & outputAttrs) { + feedback.attrib[feedback.num_attribs] = i; + feedback.size[feedback.num_attribs] = 4; + feedback.num_attribs++; + } + } + + pipe->set_feedback_state(pipe, &feedback); +} + + + + + +/** + * Clip a point against the view volume. + * + * \param v vertex vector describing the point to clip. + * + * \return zero if outside view volume, or one if inside. + */ +static GLuint +viewclip_point( const GLfloat v[] ) +{ + if ( v[0] > v[3] || v[0] < -v[3] + || v[1] > v[3] || v[1] < -v[3] + || v[2] > v[3] || v[2] < -v[3] ) { + return 0; + } + else { + return 1; + } +} + + +/** + * Clip a point against the far/near Z clipping planes. + * + * \param v vertex vector describing the point to clip. + * + * \return zero if outside view volume, or one if inside. + */ +static GLuint +viewclip_point_z( const GLfloat v[] ) +{ + if (v[2] > v[3] || v[2] < -v[3] ) { + return 0; + } + else { + return 1; + } +} + + +/** + * Clip a point against the user clipping planes. + * + * \param ctx GL context. + * \param v vertex vector describing the point to clip. + * + * \return zero if the point was clipped, or one otherwise. + */ +static GLuint +userclip_point( GLcontext *ctx, const GLfloat v[] ) +{ + GLuint p; + + for (p = 0; p < ctx->Const.MaxClipPlanes; p++) { + if (ctx->Transform.ClipPlanesEnabled & (1 << p)) { + GLfloat dot = v[0] * ctx->Transform._ClipUserPlane[p][0] + + v[1] * ctx->Transform._ClipUserPlane[p][1] + + v[2] * ctx->Transform._ClipUserPlane[p][2] + + v[3] * ctx->Transform._ClipUserPlane[p][3]; + if (dot < 0.0F) { + return 0; + } + } + } + + return 1; +} + + +/** + * Update the current raster position. + * Do clip testing, etc. here. + */ +static void +update_rasterpos(GLcontext *ctx, + const float clipPos[4], + const float color0[4], + const float color1[4], + const float *tex) +{ + uint i; + float d, ndc[3]; + + /* clip to view volume */ + if (ctx->Transform.RasterPositionUnclipped) { + /* GL_IBM_rasterpos_clip: only clip against Z */ + if (viewclip_point_z(clipPos) == 0) { + ctx->Current.RasterPosValid = GL_FALSE; + return; + } + } + else if (viewclip_point(clipPos) == 0) { + /* Normal OpenGL behaviour */ + ctx->Current.RasterPosValid = GL_FALSE; + return; + } + + /* clip to user clipping planes */ + if (ctx->Transform.ClipPlanesEnabled && !userclip_point(ctx, clipPos)) { + ctx->Current.RasterPosValid = GL_FALSE; + return; + } + + + /* + * update current raster position + */ + /* ndc = clip / W */ + d = (clipPos[3] == 0.0F) ? 1.0F : 1.0F / clipPos[3]; + ndc[0] = clipPos[0] * d; + ndc[1] = clipPos[1] * d; + ndc[2] = clipPos[2] * d; + /* wincoord = viewport_mapping(ndc) */ + ctx->Current.RasterPos[0] = (ndc[0] * ctx->Viewport._WindowMap.m[MAT_SX] + + ctx->Viewport._WindowMap.m[MAT_TX]); + ctx->Current.RasterPos[1] = (ndc[1] * ctx->Viewport._WindowMap.m[MAT_SY] + + ctx->Viewport._WindowMap.m[MAT_TY]); + ctx->Current.RasterPos[2] = (ndc[2] * ctx->Viewport._WindowMap.m[MAT_SZ] + + ctx->Viewport._WindowMap.m[MAT_TZ]) + / ctx->DrawBuffer->_DepthMaxF; + ctx->Current.RasterPos[3] = clipPos[3]; + + /* compute raster distance */ + if (ctx->Fog.FogCoordinateSource == GL_FOG_COORDINATE_EXT) + ctx->Current.RasterDistance = ctx->Current.Attrib[VERT_ATTRIB_FOG][0]; + else { +#if 0 + /* XXX we don't have an eye coord! */ + ctx->Current.RasterDistance = + SQRTF( eye[0]*eye[0] + eye[1]*eye[1] + eye[2]*eye[2] ); +#endif + } + + /* colors and texcoords */ + COPY_4FV(ctx->Current.RasterColor, color0); + COPY_4FV(ctx->Current.RasterSecondaryColor, color1); + for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { + COPY_4FV(ctx->Current.RasterTexCoords + i, tex + i *4); + } + + ctx->Current.RasterPosValid = GL_TRUE; +} + + + +static void +st_RasterPos(GLcontext *ctx, const GLfloat v[4]) +{ + struct pipe_context *pipe = ctx->st->pipe; + float *buf_map; + struct pipe_feedback_buffer fb_buf; + + st_validate_state(ctx->st); + + /* setup vertex buffers */ + setup_vertex_attribs(ctx); + + /* + * Load the default attribute buffer with current attribs. + */ + { + struct pipe_buffer_handle *buf = ctx->st->default_attrib_buffer; + const unsigned size = sizeof(ctx->Current.Attrib); + const void *data = ctx->Current.Attrib; + /* colors, texcoords, etc */ + pipe->winsys->buffer_data(pipe->winsys, buf, size, data); + /* position */ + pipe->winsys->buffer_subdata(pipe->winsys, buf, + 0, /* offset */ + 4 * sizeof(float), /* size */ + v); /* data */ + } + + + /* setup feedback state */ + setup_feedback(ctx); + + /* setup vertex feedback buffer */ + { + fb_buf.size = 8 * 4 * sizeof(float); + fb_buf.buffer = pipe->winsys->buffer_create(pipe->winsys, 0); + fb_buf.start_offset = 0; + pipe->winsys->buffer_data(pipe->winsys, fb_buf.buffer, + fb_buf.size, + NULL); /* data */ + pipe->set_feedback_buffer(pipe, 0, &fb_buf); + } + + + /* draw a point */ + pipe->draw_arrays(pipe, GL_POINTS, 0, 1); + + /* get feedback */ + buf_map = (float *) pipe->winsys->buffer_map(pipe->winsys, fb_buf.buffer, + PIPE_BUFFER_FLAG_READ); + + /* extract values and update rasterpos state */ + { + const uint outputAttrs = ctx->st->state.vs.outputs_written; + const float *pos, *color0, *color1, *tex0; + float *buf = buf_map; + + assert(outputAttrs & (1 << TGSI_ATTRIB_POS)); + pos = buf; + buf += 4; + + if (outputAttrs & (1 << TGSI_ATTRIB_COLOR0)) { + color0 = buf; + buf += 4; + } + else { + color0 = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; + } + + if (outputAttrs & (1 << TGSI_ATTRIB_COLOR1)) { + color1 = buf; + buf += 4; + } + else { + color1 = ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; + } + + update_rasterpos(ctx, pos, color0, color1, tex0); + } + + + /* free vertex feedback buffer */ + pipe->winsys->buffer_unmap(pipe->winsys, fb_buf.buffer); + pipe->winsys->buffer_unreference(pipe->winsys, &fb_buf.buffer); + + /* restore pipe state */ + pipe->set_feedback_state(pipe, &ctx->st->state.feedback); +} + + +void st_init_rasterpos_functions(struct dd_function_table *functions) +{ + functions->RasterPos = st_RasterPos; +} diff --git a/src/mesa/state_tracker/st_cb_rasterpos.h b/src/mesa/state_tracker/st_cb_rasterpos.h new file mode 100644 index 0000000000..2b992e1405 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_rasterpos.h @@ -0,0 +1,33 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ST_CB_RASTERPOS_H +#define ST_CB_RASTERPOS_H + +extern void st_init_rasterpos_functions(struct dd_function_table *functions); + +#endif -- cgit v1.2.3 From bee148cb7d5c974a1b5534a0307b2c082eea27e2 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 16:28:27 -0600 Subject: plug in rasterpos/feedback code --- src/mesa/state_tracker/st_context.c | 2 ++ src/mesa/state_tracker/st_context.h | 1 + 2 files changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 42263cab64..e66e65e2dd 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -34,6 +34,7 @@ #include "st_cb_drawpixels.h" #include "st_cb_fbo.h" #include "st_cb_queryobj.h" +#include "st_cb_rasterpos.h" #include "st_cb_readpixels.h" #include "st_cb_texture.h" #include "st_cb_flush.h" @@ -121,6 +122,7 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_fbo_functions(functions); st_init_program_functions(functions); st_init_query_functions(functions); + st_init_rasterpos_functions(functions); st_init_readpixels_functions(functions); st_init_texture_functions(functions); st_init_flush_functions(functions); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index cb34994d77..98b78af7c5 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -74,6 +74,7 @@ struct st_context struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; struct pipe_depth_state depth; + struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; -- cgit v1.2.3 From b2dfe2be6cc91c330e7fe21ea663cacfc220582d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 16:28:39 -0600 Subject: added st_cb_rasterpos.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 28f865f2ed..7ff618c6ee 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -212,6 +212,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_program.c \ state_tracker/st_cb_queryobj.c \ + state_tracker/st_cb_rasterpos.c \ state_tracker/st_cb_readpixels.c \ state_tracker/st_cb_strings.c \ state_tracker/st_cb_texture.c \ -- cgit v1.2.3 From d56a3adc30c19c8f00cb7dc585e86babfd3a79a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 16:29:23 -0600 Subject: More work on vertex feedback / glRasterPos. Basic rasterpos works now. --- src/mesa/pipe/draw/draw_context.c | 10 ++++++++++ src/mesa/pipe/draw/draw_context.h | 3 +++ src/mesa/pipe/draw/draw_feedback.c | 3 ++- src/mesa/pipe/p_context.h | 10 ++++------ src/mesa/pipe/p_state.h | 9 +++++++++ src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_context.c | 3 +++ src/mesa/pipe/softpipe/sp_context.h | 4 ++++ src/mesa/pipe/softpipe/sp_draw_arrays.c | 24 ++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_state.h | 9 +++++++++ 10 files changed, 69 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 3c095810e5..4498293e92 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -48,6 +48,7 @@ struct draw_context *draw_create( void ) draw->pipeline.clip = draw_clip_stage( draw ); draw->pipeline.flatshade = draw_flatshade_stage( draw ); draw->pipeline.cull = draw_cull_stage( draw ); + draw->pipeline.feedback = draw_feedback_stage( draw ); ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); @@ -147,6 +148,14 @@ static void validate_pipeline( struct draw_context *draw ) } +void draw_set_feedback_state( struct draw_context *draw, + const struct pipe_feedback_state *feedback ) +{ + draw->feedback = *feedback; + validate_pipeline( draw ); +} + + /** * Register new primitive setup/rendering state. * This causes the drawing pipeline to be rebuilt. @@ -245,6 +254,7 @@ void draw_set_mapped_feedback_buffer(struct draw_context *draw, uint index, void *buffer, uint size) { + assert(index < PIPE_MAX_FEEDBACK_ATTRIBS); draw->mapped_feedback_buffer[index] = buffer; draw->mapped_feedback_buffer_size[index] = size; /* in bytes */ } diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 5031574969..2babc02f45 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -83,6 +83,9 @@ void draw_set_viewport_state( struct draw_context *draw, void draw_set_clip_state( struct draw_context *pipe, const struct pipe_clip_state *clip ); +void draw_set_feedback_state( struct draw_context *draw, + const struct pipe_feedback_state * ); + void draw_set_setup_state( struct draw_context *draw, const struct pipe_setup_state *setup ); diff --git a/src/mesa/pipe/draw/draw_feedback.c b/src/mesa/pipe/draw/draw_feedback.c index 2729f44dc6..ecdd98e83d 100644 --- a/src/mesa/pipe/draw/draw_feedback.c +++ b/src/mesa/pipe/draw/draw_feedback.c @@ -81,7 +81,8 @@ feedback_vertex(struct draw_stage *stage, const struct vertex_header *vertex) for (i = 0; i < feedback->num_attribs; i++) { const uint attr = feedback->attrib[i]; - const float *src = attr ? vertex->data[attr] : vertex->clip; + const uint slot = stage->draw->vertex_info.attrib_to_slot[attr]; + const float *src = attr ? vertex->data[slot] : vertex->clip; const uint size = feedback->size[i]; float *dest = fs->dest[i * select]; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index c9246e883a..27a1128365 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -106,6 +106,9 @@ struct pipe_context { void (*set_depth_state)( struct pipe_context *, const struct pipe_depth_state * ); + void (*set_feedback_state)( struct pipe_context *, + const struct pipe_feedback_state *); + void (*set_framebuffer_state)( struct pipe_context *, const struct pipe_framebuffer_state * ); @@ -154,13 +157,8 @@ struct pipe_context { */ void (*set_feedback_buffer)(struct pipe_context *, unsigned index, - struct pipe_vertex_buffer *); - - void (*set_feedback_element)(struct pipe_context *, - unsigned index, - const struct pipe_vertex_element *); + const struct pipe_feedback_buffer *); - /* * Surface functions * This might go away... diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index cad96807df..ccd40d39e6 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -371,5 +371,14 @@ struct pipe_vertex_element }; +/** + * Vertex feedback buffer + */ +struct pipe_feedback_buffer { + struct pipe_buffer_handle *buffer; + unsigned size; + unsigned start_offset; +}; + #endif diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index ffa31102d5..401df1cfb7 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -26,6 +26,7 @@ DRIVER_SOURCES = \ sp_state_blend.c \ sp_state_clip.c \ sp_state_derived.c \ + sp_state_feedback.c \ sp_state_fs.c \ sp_state_sampler.c \ sp_state_setup.c \ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 5404b7f790..26453d9785 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -239,6 +239,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; softpipe->pipe.set_depth_state = softpipe_set_depth_test_state; + softpipe->pipe.set_feedback_state = softpipe_set_feedback_state; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_fs_state = softpipe_set_fs_state; softpipe->pipe.set_vs_state = softpipe_set_vs_state; @@ -249,8 +250,10 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_stencil_state = softpipe_set_stencil_state; softpipe->pipe.set_texture_state = softpipe_set_texture_state; softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; + softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; + softpipe->pipe.set_feedback_buffer = softpipe_set_feedback_buffer; softpipe->pipe.draw_arrays = softpipe_draw_arrays; softpipe->pipe.draw_elements = softpipe_draw_elements; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index fd0af6f727..2a6b932523 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -78,6 +78,7 @@ struct softpipe_context { struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; struct pipe_depth_state depth_test; + struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; struct pipe_shader_state fs; struct pipe_shader_state vs; @@ -107,6 +108,9 @@ struct softpipe_context { boolean need_z; /**< produce quad/fragment Z values? */ boolean need_w; /**< produce quad/fragment W values? */ + /** Feedback buffers */ + struct pipe_feedback_buffer feedback_buffer[PIPE_MAX_FEEDBACK_ATTRIBS]; + #if 0 /* Stipple derived state: */ diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 5198a493da..7ea29a0a26 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -135,6 +135,19 @@ softpipe_draw_elements(struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } + /* Map feedback buffers if enabled */ + if (sp->feedback.enabled) { + const uint n = sp->feedback.interleaved ? 1 : sp->feedback.num_attribs; + for (i = 0; i < n; i++) { + void *ptr = pipe->winsys->buffer_map(pipe->winsys, + sp->feedback_buffer[i].buffer, + PIPE_BUFFER_FLAG_WRITE); + draw_set_mapped_feedback_buffer(draw, i, ptr, + sp->feedback_buffer[i].size); + } + } + + /* draw! */ draw_arrays(draw, mode, start, count); @@ -152,6 +165,17 @@ softpipe_draw_elements(struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } + /* Unmap feedback buffers if enabled */ + if (sp->feedback.enabled) { + const uint n = sp->feedback.interleaved ? 1 : sp->feedback.num_attribs; + for (i = 0; i < n; i++) { + pipe->winsys->buffer_unmap(pipe->winsys, + sp->feedback_buffer[i].buffer); + draw_set_mapped_feedback_buffer(draw, i, NULL, 0); + } + } + + softpipe_unmap_surfaces(sp); softpipe_unmap_constant_buffers(sp); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 354580b2a5..f40ebe3e2b 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -59,6 +59,9 @@ void softpipe_set_constant_buffer(struct pipe_context *, void softpipe_set_depth_test_state( struct pipe_context *, const struct pipe_depth_state * ); +void softpipe_set_feedback_state( struct pipe_context *, + const struct pipe_feedback_state * ); + void softpipe_set_fs_state( struct pipe_context *, const struct pipe_shader_state * ); @@ -96,6 +99,12 @@ void softpipe_set_vertex_buffer(struct pipe_context *, unsigned index, const struct pipe_vertex_buffer *); +void softpipe_set_feedback_buffer(struct pipe_context *, + uint index, + const struct pipe_feedback_buffer *); + + + void softpipe_update_derived( struct softpipe_context *softpipe ); -- cgit v1.2.3 From b06641eb4a2465dac69467bc436df9e4df56b2f0 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 16:31:24 -0600 Subject: Rename --- src/mesa/drivers/x11/xm_softpipe.c | 272 ------------------------------------- src/mesa/drivers/x11/xm_winsys.c | 272 +++++++++++++++++++++++++++++++++++++ 2 files changed, 272 insertions(+), 272 deletions(-) delete mode 100644 src/mesa/drivers/x11/xm_softpipe.c create mode 100644 src/mesa/drivers/x11/xm_winsys.c (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_softpipe.c b/src/mesa/drivers/x11/xm_softpipe.c deleted file mode 100644 index ff33ad7850..0000000000 --- a/src/mesa/drivers/x11/xm_softpipe.c +++ /dev/null @@ -1,272 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ - -/* - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#include "glxheader.h" -#include "xmesaP.h" -#include "main/macros.h" - -#include "pipe/p_winsys.h" -#include "pipe/softpipe/sp_winsys.h" - - -struct xm_softpipe_winsys -{ - struct softpipe_winsys sws; - XMesaContext xmctx; /* not really needed */ -}; - - -struct xm_buffer -{ - int refcount; - int size; - void *data; - void *mapped; -}; - - - -/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque - * buffer pointer... - */ -static inline struct xm_buffer * -xm_bo( struct pipe_buffer_handle *bo ) -{ - return (struct xm_buffer *) bo; -} - -static inline struct pipe_buffer_handle * -pipe_bo( struct xm_buffer *bo ) -{ - return (struct pipe_buffer_handle *) bo; -} - -/* Turn a softpipe winsys into an xm/softpipe winsys: - */ -static inline struct xm_softpipe_winsys * -xm_softpipe_winsys(struct softpipe_winsys *sws) -{ - return (struct xm_softpipe_winsys *) sws; -} - - -/* Most callbacks map direcly onto dri_bufmgr operations: - */ -static void * -xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, - unsigned flags) -{ - struct xm_buffer *xm_buf = xm_bo(buf); - xm_buf->mapped = xm_buf->data; - return xm_buf->mapped; -} - -static void -xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) -{ - struct xm_buffer *xm_buf = xm_bo(buf); - xm_buf->mapped = NULL; -} - - -static struct pipe_buffer_handle * -xm_buffer_reference(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) -{ - struct xm_buffer *xm_buf = xm_bo(buf); - xm_buf->refcount++; - return buf; -} - -static void -xm_buffer_unreference(struct pipe_winsys *pws, struct pipe_buffer_handle **buf) -{ - if (*buf) { - struct xm_buffer *xm_buf = xm_bo(*buf); - xm_buf->refcount--; - assert(xm_buf->refcount >= 0); - if (xm_buf->refcount == 0) { - if (xm_buf->data) { - free(xm_buf->data); - xm_buf->data = NULL; - } - free(xm_buf); - } - *buf = NULL; - } -} - -static void -xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, - unsigned size, const void *data ) -{ - struct xm_buffer *xm_buf = xm_bo(buf); - if (xm_buf->size != size) { - if (xm_buf->data) - free(xm_buf->data); - xm_buf->data = malloc(size); - xm_buf->size = size; - } - if (data) - memcpy(xm_buf->data, data, size); -} - -static void -xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, - unsigned long offset, unsigned long size, const void *data) -{ - struct xm_buffer *xm_buf = xm_bo(buf); - GLubyte *b = (GLubyte *) xm_buf->data; - assert(b); - memcpy(b + offset, data, size); -} - -static void -xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, - unsigned long offset, unsigned long size, void *data) -{ - const struct xm_buffer *xm_buf = xm_bo(buf); - const GLubyte *b = (GLubyte *) xm_buf->data; - assert(b); - memcpy(data, b + offset, size); -} - -static void -xm_flush_frontbuffer(struct pipe_winsys *pws) -{ - /* - struct intel_context *intel = intel_pipe_winsys(sws)->intel; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelCopyBuffer(dPriv, NULL); - */ -} - -static void -xm_wait_idle(struct pipe_winsys *pws) -{ - /* no-op */ -} - -static void -xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) -{ - va_list args; - va_start( args, fmtString ); - vfprintf(stderr, fmtString, args); - va_end( args ); -} - -static const char * -xm_get_name(struct pipe_winsys *pws) -{ - return "Xlib"; -} - - -/* Softpipe has no concept of pools. We choose the tex/region pool - * for all buffers. - */ -static struct pipe_buffer_handle * -xm_buffer_create(struct pipe_winsys *pws, unsigned alignment) -{ - struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); - buffer->refcount = 1; - return pipe_bo(buffer); -} - - -struct xmesa_pipe_winsys -{ - struct pipe_winsys winsys; - XMesaContext xmesa; -}; - -static struct pipe_winsys * -xmesa_create_pipe_winsys( XMesaContext xmesa ) -{ - struct xmesa_pipe_winsys *xws = CALLOC_STRUCT(xmesa_pipe_winsys); - - /* Fill in this struct with callbacks that pipe will need to - * communicate with the window system, buffer manager, etc. - * - * Pipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate _DriBufferObject. - */ - xws->winsys.buffer_create = xm_buffer_create; - xws->winsys.buffer_map = xm_buffer_map; - xws->winsys.buffer_unmap = xm_buffer_unmap; - xws->winsys.buffer_reference = xm_buffer_reference; - xws->winsys.buffer_unreference = xm_buffer_unreference; - xws->winsys.buffer_data = xm_buffer_data; - xws->winsys.buffer_subdata = xm_buffer_subdata; - xws->winsys.buffer_get_subdata = xm_buffer_get_subdata; - xws->winsys.flush_frontbuffer = xm_flush_frontbuffer; - xws->winsys.wait_idle = xm_wait_idle; - xws->winsys.printf = xm_printf; - xws->winsys.get_name = xm_get_name; - xws->xmesa = xmesa; - - return &xws->winsys; -} - - -struct pipe_context * -xmesa_create_softpipe(XMesaContext xmesa) -{ - struct xm_softpipe_winsys *isws = CALLOC_STRUCT( xm_softpipe_winsys ); - - /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. - * - * Softpipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate xm_buffer. - */ -#if 0 - isws->sws.create_buffer = xm_create_buffer; - isws->sws.buffer_map = xm_buffer_map; - isws->sws.buffer_unmap = xm_buffer_unmap; - isws->sws.buffer_reference = xm_buffer_reference; - isws->sws.buffer_unreference = xm_buffer_unreference; - isws->sws.buffer_data = xm_buffer_data; - isws->sws.buffer_subdata = xm_buffer_subdata; - isws->sws.buffer_get_subdata = xm_buffer_get_subdata; -#endif - - /* Create the softpipe context: - */ - return softpipe_create( xmesa_create_pipe_winsys(xmesa), &isws->sws ); -} diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c new file mode 100644 index 0000000000..ff33ad7850 --- /dev/null +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -0,0 +1,272 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "glxheader.h" +#include "xmesaP.h" +#include "main/macros.h" + +#include "pipe/p_winsys.h" +#include "pipe/softpipe/sp_winsys.h" + + +struct xm_softpipe_winsys +{ + struct softpipe_winsys sws; + XMesaContext xmctx; /* not really needed */ +}; + + +struct xm_buffer +{ + int refcount; + int size; + void *data; + void *mapped; +}; + + + +/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static inline struct xm_buffer * +xm_bo( struct pipe_buffer_handle *bo ) +{ + return (struct xm_buffer *) bo; +} + +static inline struct pipe_buffer_handle * +pipe_bo( struct xm_buffer *bo ) +{ + return (struct pipe_buffer_handle *) bo; +} + +/* Turn a softpipe winsys into an xm/softpipe winsys: + */ +static inline struct xm_softpipe_winsys * +xm_softpipe_winsys(struct softpipe_winsys *sws) +{ + return (struct xm_softpipe_winsys *) sws; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void * +xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned flags) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + xm_buf->mapped = xm_buf->data; + return xm_buf->mapped; +} + +static void +xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + xm_buf->mapped = NULL; +} + + +static struct pipe_buffer_handle * +xm_buffer_reference(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + xm_buf->refcount++; + return buf; +} + +static void +xm_buffer_unreference(struct pipe_winsys *pws, struct pipe_buffer_handle **buf) +{ + if (*buf) { + struct xm_buffer *xm_buf = xm_bo(*buf); + xm_buf->refcount--; + assert(xm_buf->refcount >= 0); + if (xm_buf->refcount == 0) { + if (xm_buf->data) { + free(xm_buf->data); + xm_buf->data = NULL; + } + free(xm_buf); + } + *buf = NULL; + } +} + +static void +xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + if (xm_buf->size != size) { + if (xm_buf->data) + free(xm_buf->data); + xm_buf->data = malloc(size); + xm_buf->size = size; + } + if (data) + memcpy(xm_buf->data, data, size); +} + +static void +xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned long offset, unsigned long size, const void *data) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + GLubyte *b = (GLubyte *) xm_buf->data; + assert(b); + memcpy(b + offset, data, size); +} + +static void +xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned long offset, unsigned long size, void *data) +{ + const struct xm_buffer *xm_buf = xm_bo(buf); + const GLubyte *b = (GLubyte *) xm_buf->data; + assert(b); + memcpy(data, b + offset, size); +} + +static void +xm_flush_frontbuffer(struct pipe_winsys *pws) +{ + /* + struct intel_context *intel = intel_pipe_winsys(sws)->intel; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelCopyBuffer(dPriv, NULL); + */ +} + +static void +xm_wait_idle(struct pipe_winsys *pws) +{ + /* no-op */ +} + +static void +xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) +{ + va_list args; + va_start( args, fmtString ); + vfprintf(stderr, fmtString, args); + va_end( args ); +} + +static const char * +xm_get_name(struct pipe_winsys *pws) +{ + return "Xlib"; +} + + +/* Softpipe has no concept of pools. We choose the tex/region pool + * for all buffers. + */ +static struct pipe_buffer_handle * +xm_buffer_create(struct pipe_winsys *pws, unsigned alignment) +{ + struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); + buffer->refcount = 1; + return pipe_bo(buffer); +} + + +struct xmesa_pipe_winsys +{ + struct pipe_winsys winsys; + XMesaContext xmesa; +}; + +static struct pipe_winsys * +xmesa_create_pipe_winsys( XMesaContext xmesa ) +{ + struct xmesa_pipe_winsys *xws = CALLOC_STRUCT(xmesa_pipe_winsys); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + * + * Pipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + xws->winsys.buffer_create = xm_buffer_create; + xws->winsys.buffer_map = xm_buffer_map; + xws->winsys.buffer_unmap = xm_buffer_unmap; + xws->winsys.buffer_reference = xm_buffer_reference; + xws->winsys.buffer_unreference = xm_buffer_unreference; + xws->winsys.buffer_data = xm_buffer_data; + xws->winsys.buffer_subdata = xm_buffer_subdata; + xws->winsys.buffer_get_subdata = xm_buffer_get_subdata; + xws->winsys.flush_frontbuffer = xm_flush_frontbuffer; + xws->winsys.wait_idle = xm_wait_idle; + xws->winsys.printf = xm_printf; + xws->winsys.get_name = xm_get_name; + xws->xmesa = xmesa; + + return &xws->winsys; +} + + +struct pipe_context * +xmesa_create_softpipe(XMesaContext xmesa) +{ + struct xm_softpipe_winsys *isws = CALLOC_STRUCT( xm_softpipe_winsys ); + + /* Fill in this struct with callbacks that softpipe will need to + * communicate with the window system, buffer manager, etc. + * + * Softpipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate xm_buffer. + */ +#if 0 + isws->sws.create_buffer = xm_create_buffer; + isws->sws.buffer_map = xm_buffer_map; + isws->sws.buffer_unmap = xm_buffer_unmap; + isws->sws.buffer_reference = xm_buffer_reference; + isws->sws.buffer_unreference = xm_buffer_unreference; + isws->sws.buffer_data = xm_buffer_data; + isws->sws.buffer_subdata = xm_buffer_subdata; + isws->sws.buffer_get_subdata = xm_buffer_get_subdata; +#endif + + /* Create the softpipe context: + */ + return softpipe_create( xmesa_create_pipe_winsys(xmesa), &isws->sws ); +} -- cgit v1.2.3 From c3f7f3124a04ba4e02c9c81eb7901e5715940c44 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 16:32:24 -0600 Subject: s/xm_softpipe.c/xm_winsys.c/ --- src/mesa/sources | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 7ff618c6ee..a672ad2fcf 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -320,7 +320,7 @@ X11_DRIVER_SOURCES = \ drivers/x11/xm_dd.c \ drivers/x11/xm_glide.c \ drivers/x11/xm_line.c \ - drivers/x11/xm_softpipe.c \ + drivers/x11/xm_winsys.c \ drivers/x11/xm_span.c \ drivers/x11/xm_surface.c \ drivers/x11/xm_tri.c -- cgit v1.2.3 From 051a2a3028370da66c15c44fa49fa9474b23d668 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 16:35:16 -0600 Subject: clean-ups, comments --- src/mesa/drivers/x11/xm_winsys.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index ff33ad7850..f1ecbf531f 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -41,13 +41,21 @@ #include "pipe/softpipe/sp_winsys.h" -struct xm_softpipe_winsys +/** + * XMesa winsys, derived from softpipe winsys. + * NOTE: there's nothing really X-specific in this winsys layer so + * we could probably lift it up somewhere. + */ +struct xm_winsys { struct softpipe_winsys sws; - XMesaContext xmctx; /* not really needed */ + int foo; /* placeholder */ }; +/** + * Low-level OS/window system memory buffer + */ struct xm_buffer { int refcount; @@ -75,10 +83,10 @@ pipe_bo( struct xm_buffer *bo ) /* Turn a softpipe winsys into an xm/softpipe winsys: */ -static inline struct xm_softpipe_winsys * -xm_softpipe_winsys(struct softpipe_winsys *sws) +static inline struct xm_winsys * +xm_winsys(struct softpipe_winsys *sws) { - return (struct xm_softpipe_winsys *) sws; + return (struct xm_winsys *) sws; } @@ -246,7 +254,7 @@ xmesa_create_pipe_winsys( XMesaContext xmesa ) struct pipe_context * xmesa_create_softpipe(XMesaContext xmesa) { - struct xm_softpipe_winsys *isws = CALLOC_STRUCT( xm_softpipe_winsys ); + struct xm_winsys *isws = CALLOC_STRUCT( xm_winsys ); /* Fill in this struct with callbacks that softpipe will need to * communicate with the window system, buffer manager, etc. -- cgit v1.2.3 From 6275b40063d7d06e0b05767ebca3963ce7e9c34e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 16:52:05 -0600 Subject: merge buffer_unreference(), buffer_reference() --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 18 +++++----- src/mesa/drivers/x11/xm_winsys.c | 42 +++++++++++----------- src/mesa/pipe/i915simple/i915_regions.c | 4 +-- src/mesa/pipe/p_winsys.h | 9 +++-- src/mesa/pipe/softpipe/sp_region.c | 4 +-- src/mesa/pipe/softpipe/sp_state_fs.c | 5 +-- src/mesa/state_tracker/st_cb_bufferobjects.c | 2 +- src/mesa/state_tracker/st_cb_rasterpos.c | 2 +- src/mesa/state_tracker/st_draw.c | 5 +-- 9 files changed, 45 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 33aacf2a43..668bedcbda 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -84,22 +84,21 @@ static void intel_buffer_unmap(struct pipe_winsys *sws, } -static struct pipe_buffer_handle * +static void intel_buffer_reference(struct pipe_winsys *sws, + struct pipe_buffer_handle **ptr, struct pipe_buffer_handle *buf) { - return pipe_bo( driBOReference( dri_bo(buf) ) ); -} - -static void intel_buffer_unreference(struct pipe_winsys *sws, - struct pipe_buffer_handle **buf) -{ - if (*buf) { - driBOUnReference( dri_bo(*buf) ); + if (*ptr) { + driBOUnReference( dri_bo(*ptr) ); *buf = NULL; } + + driBOReference( dri_bo(buf) ); + *ptr = buf; } + /* Grabs the hardware lock! */ static void intel_buffer_data(struct pipe_winsys *sws, @@ -208,7 +207,6 @@ intel_create_pipe_winsys( struct intel_context *intel ) iws->winsys.buffer_map = intel_buffer_map; iws->winsys.buffer_unmap = intel_buffer_unmap; iws->winsys.buffer_reference = intel_buffer_reference; - iws->winsys.buffer_unreference = intel_buffer_unreference; iws->winsys.buffer_data = intel_buffer_data; iws->winsys.buffer_subdata = intel_buffer_subdata; iws->winsys.buffer_get_subdata = intel_buffer_get_subdata; diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index f1ecbf531f..b7b2941490 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -108,30 +108,31 @@ xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) xm_buf->mapped = NULL; } - -static struct pipe_buffer_handle * -xm_buffer_reference(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) -{ - struct xm_buffer *xm_buf = xm_bo(buf); - xm_buf->refcount++; - return buf; -} - static void -xm_buffer_unreference(struct pipe_winsys *pws, struct pipe_buffer_handle **buf) +xm_buffer_reference(struct pipe_winsys *pws, + struct pipe_buffer_handle **ptr, + struct pipe_buffer_handle *buf) { - if (*buf) { - struct xm_buffer *xm_buf = xm_bo(*buf); - xm_buf->refcount--; - assert(xm_buf->refcount >= 0); - if (xm_buf->refcount == 0) { - if (xm_buf->data) { - free(xm_buf->data); - xm_buf->data = NULL; + if (*ptr) { + struct xm_buffer *oldBuf = xm_bo(*ptr); + oldBuf->refcount--; + assert(oldBuf->refcount >= 0); + if (oldBuf->refcount == 0) { + if (oldBuf->data) { + free(oldBuf->data); + oldBuf->data = NULL; } - free(xm_buf); + free(oldBuf); } - *buf = NULL; + *ptr = NULL; + } + + assert(!(*ptr)); + + if (buf) { + struct xm_buffer *newBuf = xm_bo(buf); + newBuf->refcount++; + *ptr = buf; } } @@ -237,7 +238,6 @@ xmesa_create_pipe_winsys( XMesaContext xmesa ) xws->winsys.buffer_map = xm_buffer_map; xws->winsys.buffer_unmap = xm_buffer_unmap; xws->winsys.buffer_reference = xm_buffer_reference; - xws->winsys.buffer_unreference = xm_buffer_unreference; xws->winsys.buffer_data = xm_buffer_data; xws->winsys.buffer_subdata = xm_buffer_subdata; xws->winsys.buffer_get_subdata = xm_buffer_get_subdata; diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index aad4a6a537..bab256c763 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -123,8 +123,8 @@ i915_region_release(struct pipe_context *pipe, struct pipe_region **region) if ((*region)->refcount == 0) { assert((*region)->map_refcount == 0); - i915->pipe.winsys->buffer_unreference( i915->pipe.winsys, - &((*region)->buffer) ); + i915->pipe.winsys->buffer_reference( i915->pipe.winsys, + &((*region)->buffer), NULL ); free(*region); } *region = NULL; diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 09baef020e..4d2cc2196f 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -79,11 +79,10 @@ struct pipe_winsys { void (*buffer_unmap)( struct pipe_winsys *sws, struct pipe_buffer_handle *buf ); - struct pipe_buffer_handle *(*buffer_reference)( struct pipe_winsys *sws, - struct pipe_buffer_handle *buf ); - - void (*buffer_unreference)( struct pipe_winsys *sws, - struct pipe_buffer_handle **buf ); + /** Set ptr = buf, with reference counting */ + void (*buffer_reference)( struct pipe_winsys *sws, + struct pipe_buffer_handle **ptr, + struct pipe_buffer_handle *buf ); void (*buffer_data)(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 25d0a419aa..ae05b1d0de 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -121,8 +121,8 @@ sp_region_release(struct pipe_context *pipe, struct pipe_region **region) if ((*region)->refcount == 0) { assert((*region)->map_refcount == 0); - sp->pipe.winsys->buffer_unreference( sp->pipe.winsys, - &((*region)->buffer) ); + sp->pipe.winsys->buffer_reference( sp->pipe.winsys, + &((*region)->buffer), NULL ); free(*region); } *region = NULL; diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 9e3ff6d35c..5ab246896b 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -68,8 +68,9 @@ void softpipe_set_constant_buffer(struct pipe_context *pipe, assert(index == 0); /* note: reference counting */ - ws->buffer_unreference(ws, &softpipe->constants[shader].buffer); - softpipe->constants[shader].buffer = ws->buffer_reference(ws, buf->buffer); + ws->buffer_reference(ws, + &softpipe->constants[shader].buffer, + buf->buffer); softpipe->constants[shader].size = buf->size; softpipe->dirty |= SP_NEW_CONSTANTS; diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index f1dc4fbaf4..a593bd74d1 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -81,7 +81,7 @@ st_bufferobj_free(GLcontext *ctx, struct gl_buffer_object *obj) struct st_buffer_object *st_obj = st_buffer_object(obj); if (st_obj->buffer) - pipe->winsys->buffer_unreference(pipe->winsys, &st_obj->buffer); + pipe->winsys->buffer_reference(pipe->winsys, &st_obj->buffer, NULL); free(st_obj); } diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 5040c21e51..7bedf3f89f 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -337,7 +337,7 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) /* free vertex feedback buffer */ pipe->winsys->buffer_unmap(pipe->winsys, fb_buf.buffer); - pipe->winsys->buffer_unreference(pipe->winsys, &fb_buf.buffer); + pipe->winsys->buffer_reference(pipe->winsys, &fb_buf.buffer, NULL); /* restore pipe state */ pipe->set_feedback_state(pipe, &ctx->st->state.feedback); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index afb6f96a26..1facc14ccd 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -152,7 +152,8 @@ static void destroy_default_attribs_buffer(struct st_context *st) { struct pipe_context *pipe = st->pipe; - pipe->winsys->buffer_unreference(pipe->winsys, &st->default_attrib_buffer); + pipe->winsys->buffer_reference(pipe->winsys, + &st->default_attrib_buffer, NULL); } @@ -349,7 +350,7 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, pipe->draw_arrays(pipe, prim, 0, numVertex); /* XXX: do one-time */ - pipe->winsys->buffer_unreference(pipe->winsys, &vbuf); + pipe->winsys->buffer_reference(pipe->winsys, &vbuf, NULL); } -- cgit v1.2.3 From 5620d66e362acff99dbc748feec1da1e0f997288 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Sep 2007 17:02:58 -0600 Subject: vertex feedback functions --- src/mesa/pipe/softpipe/sp_state_feedback.c | 72 ++++++++++++++++++++++++++++++ 1 file changed, 72 insertions(+) create mode 100644 src/mesa/pipe/softpipe/sp_state_feedback.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_feedback.c b/src/mesa/pipe/softpipe/sp_state_feedback.c new file mode 100644 index 0000000000..02aaf34e75 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_feedback.c @@ -0,0 +1,72 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Authors: + * Brian Paul + */ + + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_surface.h" + +#include "pipe/p_winsys.h" +#include "pipe/draw/draw_context.h" + + +void +softpipe_set_feedback_state(struct pipe_context *pipe, + const struct pipe_feedback_state *feedback) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->feedback = *feedback; /* struct copy */ + /* + softpipe->dirty |= SP_NEW_FEEDBACK; + */ + + draw_set_feedback_state(softpipe->draw, feedback); +} + + +void +softpipe_set_feedback_buffer(struct pipe_context *pipe, + unsigned index, + const struct pipe_feedback_buffer *feedback) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + assert(index < PIPE_MAX_FEEDBACK_ATTRIBS); + + /* Need to be careful with referencing */ + pipe->winsys->buffer_reference(pipe->winsys, + &softpipe->feedback_buffer[index].buffer, + feedback->buffer); + softpipe->feedback_buffer[index].size = feedback->size; + softpipe->feedback_buffer[index].start_offset = feedback->start_offset; +} -- cgit v1.2.3 From 09fbb3837b6aa5dfc6c94f41ab5443820177c569 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Sep 2007 16:01:17 -0600 Subject: Implement query object interface. This replaces the temporary occlusion counter functions we had before. Added new ctx->Driver.WaitQuery() function which should block until the result is ready. Sketch out some code for vertex transformation feedback counters. --- src/mesa/drivers/common/driverfuncs.c | 5 +- src/mesa/main/dd.h | 6 +- src/mesa/main/mtypes.h | 9 +-- src/mesa/main/queryobj.c | 95 ++++++++++++++++-------------- src/mesa/main/queryobj.h | 14 ++++- src/mesa/pipe/failover/fo_context.c | 4 +- src/mesa/pipe/i915simple/i915_context.c | 21 ++++++- src/mesa/pipe/p_context.h | 9 +-- src/mesa/pipe/p_defines.h | 8 +++ src/mesa/pipe/p_state.h | 10 ++++ src/mesa/pipe/softpipe/sp_context.c | 35 ++++++++--- src/mesa/pipe/softpipe/sp_context.h | 7 ++- src/mesa/pipe/softpipe/sp_quad_occlusion.c | 10 ++-- src/mesa/state_tracker/st_cb_queryobj.c | 83 +++++++++++++++++++++++--- 14 files changed, 232 insertions(+), 84 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c index 500dbb2545..96e5037fa5 100644 --- a/src/mesa/drivers/common/driverfuncs.c +++ b/src/mesa/drivers/common/driverfuncs.c @@ -224,8 +224,9 @@ _mesa_init_driver_functions(struct dd_function_table *driver) /* query objects */ driver->NewQueryObject = _mesa_new_query_object; - driver->BeginQuery = NULL; - driver->EndQuery = NULL; + driver->BeginQuery = _mesa_begin_query; + driver->EndQuery = _mesa_end_query; + driver->WaitQuery = _mesa_wait_query; /* APPLE_vertex_array_object */ driver->NewArrayObject = _mesa_new_array_object; diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index 582e0c334d..f089fcb48f 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -811,9 +811,9 @@ struct dd_function_table { */ /*@{*/ struct gl_query_object * (*NewQueryObject)(GLcontext *ctx, GLuint id); - void (*BeginQuery)(GLcontext *ctx, GLenum target, - struct gl_query_object *q); - void (*EndQuery)(GLcontext *ctx, GLenum target, struct gl_query_object *q); + void (*BeginQuery)(GLcontext *ctx, struct gl_query_object *q); + void (*EndQuery)(GLcontext *ctx, struct gl_query_object *q); + void (*WaitQuery)(GLcontext *ctx, struct gl_query_object *q); /*@}*/ diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 0a64e0c58c..514170dbcf 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2085,10 +2085,11 @@ struct gl_ati_fragment_shader_state */ struct gl_query_object { - GLuint Id; - GLuint64EXT Result; /* the counter */ - GLboolean Active; /* inside Begin/EndQuery */ - GLboolean Ready; /* result is ready */ + GLenum Target; /**< The query target, when active */ + GLuint Id; /**< hash table ID/name */ + GLuint64EXT Result; /**< the counter */ + GLboolean Active; /**< inside Begin/EndQuery */ + GLboolean Ready; /**< result is ready? */ }; diff --git a/src/mesa/main/queryobj.c b/src/mesa/main/queryobj.c index 0e59ba615a..688d0fc7bc 100644 --- a/src/mesa/main/queryobj.c +++ b/src/mesa/main/queryobj.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.1 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -53,6 +53,42 @@ _mesa_new_query_object(GLcontext *ctx, GLuint id) } +/** + * Begin a query. Software driver fallback. + * Called via ctx->Driver.BeginQuery(). + */ +void +_mesa_begin_query(GLcontext *ctx, struct gl_query_object *q) +{ + /* no-op */ +} + + +/** + * End a query. Software driver fallback. + * Called via ctx->Driver.EndQuery(). + */ +void +_mesa_end_query(GLcontext *ctx, struct gl_query_object *q) +{ + q->Ready = GL_TRUE; +} + + +/** + * Wait for query to complete. Software driver fallback. + * Called via ctx->Driver.WaitQuery(). + */ +void +_mesa_wait_query(GLcontext *ctx, struct gl_query_object *q) +{ + /* For software drivers, _mesa_end_query() should have completed the query. + * For real hardware, implement a proper WaitQuery() driver function. + */ + assert(q->Ready); +} + + /** * Delete an occlusion query object. * Not removed from hash table here. @@ -61,7 +97,7 @@ _mesa_new_query_object(GLcontext *ctx, GLuint id) static void delete_query_object(struct gl_query_object *q) { - FREE(q); + _mesa_free(q); } @@ -216,6 +252,7 @@ _mesa_BeginQueryARB(GLenum target, GLuint id) } } + q->Target = target; q->Active = GL_TRUE; q->Result = 0; q->Ready = GL_FALSE; @@ -229,9 +266,7 @@ _mesa_BeginQueryARB(GLenum target, GLuint id) } #endif - if (ctx->Driver.BeginQuery) { - ctx->Driver.BeginQuery(ctx, target, q); - } + ctx->Driver.BeginQuery(ctx, q); } @@ -275,13 +310,7 @@ _mesa_EndQueryARB(GLenum target) } q->Active = GL_FALSE; - if (ctx->Driver.EndQuery) { - ctx->Driver.EndQuery(ctx, target, q); - } - else { - /* if we're using software rendering/querying */ - q->Ready = GL_TRUE; - } + ctx->Driver.EndQuery(ctx, q); } @@ -346,13 +375,8 @@ _mesa_GetQueryObjectivARB(GLuint id, GLenum pname, GLint *params) switch (pname) { case GL_QUERY_RESULT_ARB: - while (!q->Ready) { - /* Wait for the query to finish! */ - /* If using software rendering, the result will always be ready - * by time we get here. Otherwise, we must be using hardware! - */ - ASSERT(ctx->Driver.EndQuery); - } + if (!q->Ready) + ctx->Driver.WaitQuery(ctx, q); /* if result is too large for returned type, clamp to max value */ if (q->Result > 0x7fffffff) { *params = 0x7fffffff; @@ -362,7 +386,6 @@ _mesa_GetQueryObjectivARB(GLuint id, GLenum pname, GLint *params) } break; case GL_QUERY_RESULT_AVAILABLE_ARB: - /* XXX revisit when we have a hardware implementation! */ *params = q->Ready; break; default: @@ -390,13 +413,8 @@ _mesa_GetQueryObjectuivARB(GLuint id, GLenum pname, GLuint *params) switch (pname) { case GL_QUERY_RESULT_ARB: - while (!q->Ready) { - /* Wait for the query to finish! */ - /* If using software rendering, the result will always be ready - * by time we get here. Otherwise, we must be using hardware! - */ - ASSERT(ctx->Driver.EndQuery); - } + if (!q->Ready) + ctx->Driver.WaitQuery(ctx, q); /* if result is too large for returned type, clamp to max value */ if (q->Result > 0xffffffff) { *params = 0xffffffff; @@ -406,7 +424,6 @@ _mesa_GetQueryObjectuivARB(GLuint id, GLenum pname, GLuint *params) } break; case GL_QUERY_RESULT_AVAILABLE_ARB: - /* XXX revisit when we have a hardware implementation! */ *params = q->Ready; break; default: @@ -439,17 +456,11 @@ _mesa_GetQueryObjecti64vEXT(GLuint id, GLenum pname, GLint64EXT *params) switch (pname) { case GL_QUERY_RESULT_ARB: - while (!q->Ready) { - /* Wait for the query to finish! */ - /* If using software rendering, the result will always be ready - * by time we get here. Otherwise, we must be using hardware! - */ - ASSERT(ctx->Driver.EndQuery); - } + if (!q->Ready) + ctx->Driver.WaitQuery(ctx, q); *params = q->Result; break; case GL_QUERY_RESULT_AVAILABLE_ARB: - /* XXX revisit when we have a hardware implementation! */ *params = q->Ready; break; default: @@ -480,17 +491,11 @@ _mesa_GetQueryObjectui64vEXT(GLuint id, GLenum pname, GLuint64EXT *params) switch (pname) { case GL_QUERY_RESULT_ARB: - while (!q->Ready) { - /* Wait for the query to finish! */ - /* If using software rendering, the result will always be ready - * by time we get here. Otherwise, we must be using hardware! - */ - ASSERT(ctx->Driver.EndQuery); - } + if (!q->Ready) + ctx->Driver.WaitQuery(ctx, q); *params = q->Result; break; case GL_QUERY_RESULT_AVAILABLE_ARB: - /* XXX revisit when we have a hardware implementation! */ *params = q->Ready; break; default: diff --git a/src/mesa/main/queryobj.h b/src/mesa/main/queryobj.h index ada8cf8356..d466aae652 100644 --- a/src/mesa/main/queryobj.h +++ b/src/mesa/main/queryobj.h @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5 + * Version: 7.1 * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -36,6 +36,16 @@ _mesa_init_query(GLcontext *ctx); extern void _mesa_free_query_data(GLcontext *ctx); +extern void +_mesa_begin_query(GLcontext *ctx, struct gl_query_object *q); + +extern void +_mesa_end_query(GLcontext *ctx, struct gl_query_object *q); + +extern void +_mesa_wait_query(GLcontext *ctx, struct gl_query_object *q); + + extern void GLAPIENTRY _mesa_GenQueriesARB(GLsizei n, GLuint *ids); diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index b88f1b466c..c58fc9cedc 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -129,8 +129,8 @@ struct pipe_context *failover_create( struct pipe_context *hw, * at this point - if the hardware doesn't support it, don't * advertise it to the application. */ - failover->pipe.reset_occlusion_counter = hw->reset_occlusion_counter; - failover->pipe.get_occlusion_counter = hw->get_occlusion_counter; + failover->pipe.begin_query = hw->begin_query; + failover->pipe.end_query = hw->end_query; failover_init_state_functions( failover ); diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index f4121419f7..6e48b3bd03 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -149,6 +149,22 @@ static void i915_destroy( struct pipe_context *pipe ) +static void +i915_begin_query(struct pipe_context *pipe, struct pipe_query_object *q) +{ + /* should never be called */ + assert(0); +} + + +static void +i915_end_query(struct pipe_context *pipe, struct pipe_query_object *q) +{ + /* should never be called */ + assert(0); +} + + static boolean i915_draw_elements( struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, unsigned indexSize, @@ -257,8 +273,9 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.supported_formats = i915_supported_formats; i915->pipe.max_texture_size = i915_max_texture_size; i915->pipe.clear = i915_clear; - i915->pipe.reset_occlusion_counter = NULL; /* no support */ - i915->pipe.get_occlusion_counter = NULL; + + i915->pipe.begin_query = i915_begin_query; + i915->pipe.end_query = i915_end_query; i915->pipe.draw_arrays = i915_draw_arrays; i915->pipe.draw_elements = i915_draw_elements; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 27a1128365..dafbef410e 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -75,11 +75,12 @@ struct pipe_context { void (*clear)(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue); - /** occlusion counting (XXX this may be temporary - we should probably - * have generic query objects with begin/end methods) + /** + * Query objects */ - void (*reset_occlusion_counter)(struct pipe_context *pipe); - unsigned (*get_occlusion_counter)(struct pipe_context *pipe); + void (*begin_query)(struct pipe_context *pipe, struct pipe_query_object *q); + void (*end_query)(struct pipe_context *pipe, struct pipe_query_object *q); + void (*wait_query)(struct pipe_context *pipe, struct pipe_query_object *q); /* * State functions diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index c1164c5c08..b3ee890576 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -299,4 +299,12 @@ #define PIPE_PRIM_POLYGON 9 +/** + * Query object types + */ +#define PIPE_QUERY_OCCLUSION_COUNTER 0 +#define PIPE_QUERY_PRIMITIVES_GENERATED 1 +#define PIPE_QUERY_PRIMITIVES_EMITTED 2 +#define PIPE_QUERY_TYPES 3 + #endif diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index ccd40d39e6..b994d17ea9 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -381,4 +381,14 @@ struct pipe_feedback_buffer { }; +/** + * Hardware queries (occlusion, transform feedback, timing, etc) + */ +struct pipe_query_object { + uint type:3; /**< PIPE_QUERY_x */ + uint ready:1; /**< is result ready? */ + uint64 count; +}; + + #endif diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 26453d9785..92357808e2 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -195,19 +195,37 @@ static void softpipe_destroy( struct pipe_context *pipe ) } -static void softpipe_reset_occlusion_counter(struct pipe_context *pipe) +static void +softpipe_begin_query(struct pipe_context *pipe, struct pipe_query_object *q) { struct softpipe_context *softpipe = softpipe_context( pipe ); - softpipe->occlusion_counter = 0; + assert(q->type < PIPE_QUERY_TYPES); + assert(!softpipe->queries[q->type]); + softpipe->queries[q->type] = q; } -/* XXX pipe param should be const */ -static unsigned softpipe_get_occlusion_counter(struct pipe_context *pipe) + +static void +softpipe_end_query(struct pipe_context *pipe, struct pipe_query_object *q) { struct softpipe_context *softpipe = softpipe_context( pipe ); - return softpipe->occlusion_counter; + assert(q->type < PIPE_QUERY_TYPES); + assert(softpipe->queries[q->type]); + q->ready = 1; /* software rendering is synchronous */ + softpipe->queries[q->type] = NULL; +} + + +static void +softpipe_wait_query(struct pipe_context *pipe, struct pipe_query_object *q) +{ + /* Should never get here since we indicated that the result was + * ready in softpipe_end_query(). + */ + assert(0); } + static const char *softpipe_get_name( struct pipe_context *pipe ) { return "softpipe"; @@ -260,8 +278,11 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; - softpipe->pipe.reset_occlusion_counter = softpipe_reset_occlusion_counter; - softpipe->pipe.get_occlusion_counter = softpipe_get_occlusion_counter; + + softpipe->pipe.begin_query = softpipe_begin_query; + softpipe->pipe.end_query = softpipe_end_query; + softpipe->pipe.wait_query = softpipe_wait_query; + softpipe->pipe.get_name = softpipe_get_name; softpipe->pipe.get_vendor = softpipe_get_vendor; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 2a6b932523..13d1143c89 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -93,6 +93,11 @@ struct softpipe_context { struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; unsigned dirty; + /* + * Active queries + */ + struct pipe_query_object *queries[PIPE_QUERY_TYPES]; + /* * Mapped vertex buffers */ @@ -120,8 +125,6 @@ struct softpipe_context { /** Derived from scissor and surface bounds: */ struct pipe_scissor_state cliprect; - unsigned occlusion_counter; - unsigned line_stipple_counter; /** Software quad rendering pipeline */ diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c index 6b094a5bef..4f178f0557 100644 --- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c +++ b/src/mesa/pipe/softpipe/sp_quad_occlusion.c @@ -44,11 +44,13 @@ static void occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; + struct pipe_query_object *occ + = softpipe->queries[PIPE_QUERY_OCCLUSION_COUNTER]; - softpipe->occlusion_counter += (quad->mask ) & 1; - softpipe->occlusion_counter += (quad->mask >> 1) & 1; - softpipe->occlusion_counter += (quad->mask >> 2) & 1; - softpipe->occlusion_counter += (quad->mask >> 3) & 1; + occ->count += (quad->mask ) & 1; + occ->count += (quad->mask >> 1) & 1; + occ->count += (quad->mask >> 2) & 1; + occ->count += (quad->mask >> 3) & 1; if (quad->mask) qs->next->run(qs->next, quad); diff --git a/src/mesa/state_tracker/st_cb_queryobj.c b/src/mesa/state_tracker/st_cb_queryobj.c index 3a8fbde8ab..5b95dd7fd3 100644 --- a/src/mesa/state_tracker/st_cb_queryobj.c +++ b/src/mesa/state_tracker/st_cb_queryobj.c @@ -44,33 +44,102 @@ #include "st_public.h" +struct st_query_object +{ + struct gl_query_object base; + struct pipe_query_object pq; +}; + + +/** + * Cast wrapper + */ +static struct st_query_object * +st_query_object(struct gl_query_object *q) +{ + return (struct st_query_object *) q; +} + + +static struct gl_query_object * +st_NewQueryObject(GLcontext *ctx, GLuint id) +{ + struct st_query_object *stq = CALLOC_STRUCT(st_query_object); + if (stq) { + stq->base.Id = id; + stq->base.Ready = GL_TRUE; + return &stq->base; + } + return NULL; +} + + /** * Do glReadPixels by getting rows from the framebuffer surface with * get_tile(). Convert to requested format/type with Mesa image routines. * Image transfer ops are done in software too. */ static void -st_BeginQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q) +st_BeginQuery(GLcontext *ctx, struct gl_query_object *q) { struct pipe_context *pipe = ctx->st->pipe; - if (target == GL_SAMPLES_PASSED_ARB) { - pipe->reset_occlusion_counter(pipe); + struct st_query_object *stq = st_query_object(q); + + stq->pq.count = 0; + + switch (q->Target) { + case GL_SAMPLES_PASSED_ARB: + stq->pq.type = PIPE_QUERY_OCCLUSION_COUNTER; + break; + case GL_PRIMITIVES_GENERATED_NV: + /* someday */ + stq->pq.type = PIPE_QUERY_PRIMITIVES_GENERATED; + break; + case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN_NV: + /* someday */ + stq->pq.type = PIPE_QUERY_PRIMITIVES_EMITTED; + break; + default: + assert(0); } + + pipe->begin_query(pipe, &stq->pq); } static void -st_EndQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q) +st_EndQuery(GLcontext *ctx, struct gl_query_object *q) { struct pipe_context *pipe = ctx->st->pipe; - if (target == GL_SAMPLES_PASSED_ARB) { - q->Result = pipe->get_occlusion_counter(pipe); - } + struct st_query_object *stq = st_query_object(q); + + pipe->end_query(pipe, &stq->pq); + stq->base.Ready = stq->pq.ready; + if (stq->base.Ready) + stq->base.Result = stq->pq.count; +} + + +static void +st_WaitQuery(GLcontext *ctx, struct gl_query_object *q) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_query_object *stq = st_query_object(q); + + /* this function should only be called if we don't have a ready result */ + assert(!stq->base.Ready); + + pipe->wait_query(pipe, &stq->pq); + q->Ready = GL_TRUE; + q->Result = stq->pq.count; } + void st_init_query_functions(struct dd_function_table *functions) { + functions->NewQueryObject = st_NewQueryObject; functions->BeginQuery = st_BeginQuery; functions->EndQuery = st_EndQuery; + functions->WaitQuery = st_WaitQuery; } -- cgit v1.2.3 From c01a5ec3d63a2f414e2b6c5b31c80367943189b4 Mon Sep 17 00:00:00 2001 From: Michel Daenzer Date: Fri, 14 Sep 2007 15:48:54 +0100 Subject: intel_winsys: Adapt to recent drm changes (s/pipes/planes/). --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 38 +++++++++++----------- .../drivers/dri/intel_winsys/intel_swapbuffers.h | 2 +- 2 files changed, 20 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 56d220ab9e..d9adcd3a62 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -46,7 +46,7 @@ #define DRM_VBLANK_FLIP 0x8000000 typedef struct drm_i915_flip { - int pipes; + int planes; } drm_i915_flip_t; #undef DRM_IOCTL_I915_FLIP @@ -273,32 +273,32 @@ intelWindowMoved(struct intel_context *intel) GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); GLuint flags = intel_fb->vblank_flags; GLboolean pf_active; - GLint pf_pipes; + GLint pf_planes; /* Update page flipping info */ - pf_pipes = 0; + pf_planes = 0; if (areaA > 0) - pf_pipes |= 1; + pf_planes |= 1; if (areaB > 0) - pf_pipes |= 2; + pf_planes |= 2; intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3; + (intel_fb->pf_planes & 0x2)) & 0x3; intel_fb->pf_num_pages = 2 /*intel->intelScreen->third.handle ? 3 : 2*/; - pf_active = pf_pipes && (pf_pipes & intel->sarea->pf_active) == pf_pipes; + pf_active = pf_planes && (pf_planes & intel->sarea->pf_active) == pf_planes; if (pf_active != intel_fb->pf_active) DBG(LOCK, "%s - Page flipping %sactive\n", __progname, pf_active ? "" : "in"); if (pf_active) { - /* Sync pages between pipes if we're flipping on both at the same time */ - if (pf_pipes == 0x3 && pf_pipes != intel_fb->pf_pipes && + /* Sync pages between planes if we're flipping on both at the same time */ + if (pf_planes == 0x3 && pf_planes != intel_fb->pf_planes && (intel->sarea->pf_current_page & 0x3) != (((intel->sarea->pf_current_page) >> 2) & 0x3)) { drm_i915_flip_t flip; @@ -314,7 +314,7 @@ intelWindowMoved(struct intel_context *intel) ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % intel_fb->pf_num_pages) << 2; - flip.pipes = 0x2; + flip.planes = 0x2; } else { intel->sarea->pf_current_page = intel->sarea->pf_current_page & (0x3 << 2); @@ -322,13 +322,13 @@ intelWindowMoved(struct intel_context *intel) (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % intel_fb->pf_num_pages; - flip.pipes = 0x1; + flip.planes = 0x1; } drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); } - intel_fb->pf_pipes = pf_pipes; + intel_fb->pf_planes = pf_planes; } intel_fb->pf_active = pf_active; @@ -403,14 +403,14 @@ intel_wait_flips(struct intel_context *intel, GLuint batch_flags) BUFFER_BACK_LEFT); if (intel_fb->Base.Name == 0 && intel_rb->pf_pending == intel_fb->pf_seq) { - GLint pf_pipes = intel_fb->pf_pipes; + GLint pf_planes = intel_fb->pf_planes; BATCH_LOCALS; /* Wait for pending flips to take effect */ BEGIN_BATCH(2, batch_flags); - OUT_BATCH(pf_pipes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) + OUT_BATCH(pf_planes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) : 0); - OUT_BATCH(pf_pipes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) + OUT_BATCH(pf_planes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) : 0); ADVANCE_BATCH(); @@ -449,7 +449,7 @@ intelPageFlip(const __DRIdrawablePrivate * dPriv) if (dPriv->numClipRects && intel_fb->pf_active) { drm_i915_flip_t flip; - flip.pipes = intel_fb->pf_pipes; + flip.planes = intel_fb->pf_planes; ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); } @@ -464,7 +464,7 @@ intelPageFlip(const __DRIdrawablePrivate * dPriv) } intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3; + (intel_fb->pf_planes & 0x2)) & 0x3; if (dPriv->numClipRects != 0) { intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending = @@ -524,7 +524,7 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) swap.seqtype |= DRM_VBLANK_FLIP; intel_fb->pf_current_page = (((intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3) + 1) % + (intel_fb->pf_planes & 0x2)) & 0x3) + 1) % intel_fb->pf_num_pages; } @@ -554,7 +554,7 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) } else { if (swap.seqtype & DRM_VBLANK_FLIP) { intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3) % + (intel_fb->pf_planes & 0x2)) & 0x3) % intel_fb->pf_num_pages; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index d12b5ef96e..0065eac9b2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -42,7 +42,7 @@ struct intel_framebuffer /* Drawable page flipping state */ GLboolean pf_active; GLuint pf_seq; - GLint pf_pipes; + GLint pf_planes; GLint pf_current_page; GLint pf_num_pages; -- cgit v1.2.3 From d59c95bfae0402813a6b9a1242cdd971624c127a Mon Sep 17 00:00:00 2001 From: Michel Daenzer Date: Fri, 14 Sep 2007 15:51:00 +0100 Subject: intel_winsys: Fix intel_buffer_reference. Correctly clear the pointer to the old buffer (not sure how this could build at all before...) and only reference the new one when its pointer is non-NULL. --- src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 668bedcbda..2b2731446d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -91,11 +91,13 @@ intel_buffer_reference(struct pipe_winsys *sws, { if (*ptr) { driBOUnReference( dri_bo(*ptr) ); - *buf = NULL; + *ptr = NULL; } - driBOReference( dri_bo(buf) ); - *ptr = buf; + if (buf) { + driBOReference( dri_bo(buf) ); + *ptr = buf; + } } -- cgit v1.2.3 From bf92350b5b2c8e14515fb8bbfacd8c06502dfb98 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Mon, 17 Sep 2007 13:22:51 +0100 Subject: Fix linux-dri-x86 build. --- src/mesa/state_tracker/st_atom_vs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index a1e6117bde..289a3e4f46 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -72,7 +72,7 @@ static void compile_vs( struct st_context *st ) #if defined(USE_X86_ASM) || defined(SLANG_X86) tgsi_emit_sse2( vp->vs.tokens, - &vp->vs.sse2_program ); + &vp->sse2_program ); #endif vp->dirty = 0; @@ -106,7 +106,7 @@ static void update_vs( struct st_context *st ) compile_vs( st ); #if defined(USE_X86_ASM) || defined(SLANG_X86) - vs.executable = (void *) x86_get_func( &vp->sse2_program ); + st->vp->vs.executable = (void *) x86_get_func( &vp->sse2_program ); #endif st->state.vs = st->vp->vs; -- cgit v1.2.3 From 9ea91974e153858d7c0397bd78448b4be1c10c4c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Sep 2007 10:05:21 -0600 Subject: update begin/end_query() functions --- src/mesa/drivers/x11/xm_dd.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index d7fb3e6e03..f02b40bdc0 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -1015,9 +1015,9 @@ xmesa_new_query_object(GLcontext *ctx, GLuint id) static void -xmesa_begin_query(GLcontext *ctx, GLenum target, struct gl_query_object *q) +xmesa_begin_query(GLcontext *ctx, struct gl_query_object *q) { - if (target == GL_TIME_ELAPSED_EXT) { + if (q->Target == GL_TIME_ELAPSED_EXT) { struct xmesa_query_object *xq = (struct xmesa_query_object *) q; (void) gettimeofday(&xq->StartTime, NULL); } @@ -1042,9 +1042,9 @@ time_diff(const struct timeval *t0, const struct timeval *t1) static void -xmesa_end_query(GLcontext *ctx, GLenum target, struct gl_query_object *q) +xmesa_end_query(GLcontext *ctx, struct gl_query_object *q) { - if (target == GL_TIME_ELAPSED_EXT) { + if (q->Target == GL_TIME_ELAPSED_EXT) { struct xmesa_query_object *xq = (struct xmesa_query_object *) q; struct timeval endTime; (void) gettimeofday(&endTime, NULL); -- cgit v1.2.3 From f48e055c1072ca0f3ce3096606a8d892a559c599 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Sep 2007 10:05:34 -0600 Subject: silence warning --- src/mesa/state_tracker/st_atom_constbuf.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index 7bea9476f1..c2310f9d3f 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -30,6 +30,7 @@ * Brian Paul */ +#include "main/imports.h" #include "shader/prog_parameter.h" #include "pipe/p_context.h" -- cgit v1.2.3 From 5578f8ce3bf349107b04fb0ebb8b8d505bd9424d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 17 Sep 2007 18:02:46 +0100 Subject: Remove non-meaningful region_idle() interface This is already handled by the semanics of map(), etc. --- src/mesa/pipe/failover/fo_context.c | 1 - src/mesa/pipe/i915simple/i915_regions.c | 6 ------ src/mesa/pipe/p_context.h | 2 -- src/mesa/pipe/softpipe/sp_region.c | 7 ------- src/mesa/state_tracker/st_cb_texture.c | 23 ----------------------- 5 files changed, 39 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index c58fc9cedc..076d516583 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -139,7 +139,6 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.region_alloc = hw->region_alloc; failover->pipe.region_release = hw->region_release; - failover->pipe.region_idle = hw->region_idle; failover->pipe.region_map = hw->region_map; failover->pipe.region_unmap = hw->region_unmap; failover->pipe.region_data = hw->region_data; diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index bab256c763..cff90ab5dd 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -37,11 +37,6 @@ #include "i915_blit.h" -static void -i915_region_idle(struct pipe_context *pipe, struct pipe_region *region) -{ - -} static ubyte * @@ -302,7 +297,6 @@ i915_region_fill(struct pipe_context *pipe, void i915_init_region_functions(struct i915_context *i915) { - i915->pipe.region_idle = i915_region_idle; i915->pipe.region_map = i915_region_map; i915->pipe.region_unmap = i915_region_unmap; i915->pipe.region_alloc = i915_region_alloc; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index dafbef410e..ec5555c38e 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -182,8 +182,6 @@ struct pipe_context { void (*region_release)(struct pipe_context *pipe, struct pipe_region **r); - void (*region_idle)(struct pipe_context *pipe, struct pipe_region *region); - ubyte *(*region_map)(struct pipe_context *pipe, struct pipe_region *r); void (*region_unmap)(struct pipe_context *pipe, struct pipe_region *r); diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index ae05b1d0de..697b738e43 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -48,12 +48,6 @@ round_up(unsigned n, unsigned multiple) } -static void -sp_region_idle(struct pipe_context *pipe, struct pipe_region *region) -{ - -} - static ubyte * sp_region_map(struct pipe_context *pipe, struct pipe_region *region) @@ -283,7 +277,6 @@ sp_region_fill(struct pipe_context *pipe, void sp_init_region_functions(struct softpipe_context *sp) { - sp->pipe.region_idle = sp_region_idle; sp->pipe.region_map = sp_region_map; sp->pipe.region_unmap = sp_region_unmap; sp->pipe.region_alloc = sp_region_alloc; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 770c871434..3ae9bd9003 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -705,14 +705,6 @@ st_TexImage(GLcontext * ctx, if (!pixels) return; - - if (stImage->mt) - pipe->region_idle(pipe, stImage->mt->region); - -#if 0 - LOCK_HARDWARE(intel); -#endif - if (stImage->mt) { texImage->Data = st_miptree_image_map(pipe, stImage->mt, @@ -766,10 +758,6 @@ st_TexImage(GLcontext * ctx, texImage->Data = NULL; } -#if 0 - UNLOCK_HARDWARE(intel); -#endif - #if 0 /* GL_SGIS_generate_mipmap -- this can be accelerated now. */ @@ -958,13 +946,6 @@ st_TexSubimage(GLcontext * ctx, if (!pixels) return; - if (stImage->mt) - pipe->region_idle(pipe, stImage->mt->region); - -#if 0 - LOCK_HARDWARE(intel); -#endif - /* Map buffer if necessary. Need to lock to prevent other contexts * from uploading the buffer under us. */ @@ -1004,10 +985,6 @@ st_TexSubimage(GLcontext * ctx, st_miptree_image_unmap(pipe, stImage->mt); texImage->Data = NULL; } - -#if 0 - UNLOCK_HARDWARE(intel); -#endif } -- cgit v1.2.3 From 060a6380270b511fd1e100c1caaa28f43943eb52 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Sep 2007 12:18:32 -0600 Subject: s/bugmgr/bufgr/ --- src/mesa/pipe/p_winsys.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 4d2cc2196f..2861e03c9e 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -60,7 +60,7 @@ struct pipe_winsys { const char *, ... ); - /* The buffer manager is modeled after the dri_bugmgr interface, + /* The buffer manager is modeled after the dri_bufmgr interface, * but this is the subset that softpipe cares about. Remember that * softpipe gets to choose the interface it needs, and the window * systems must then implement that interface (rather than the -- cgit v1.2.3 From 2a40c71fed9461a6ecd17788b17b35655262aec6 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Sep 2007 14:22:49 -0600 Subject: selection/feedback support --- src/mesa/state_tracker/st_cb_feedback.c | 306 ++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_feedback.h | 37 ++++ 2 files changed, 343 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_feedback.c create mode 100644 src/mesa/state_tracker/st_cb_feedback.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c new file mode 100644 index 0000000000..b97b2bb027 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -0,0 +1,306 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * GL_SELECT and GL_FEEDBACK render modes. + * Basically, we use a private instance of the 'draw' module for doing + * selection/feedback. It would be nice to use the transform_feedback + * hardware feature, but it's defined as happening pre-clip and we want + * post-clipped primitives. Also, there's concerns about the efficiency + * of using the hardware for this anyway. + * + * Authors: + * Brian Paul + */ + +#include "main/imports.h" +#include "main/context.h" +#include "main/feedback.h" +#include "main/macros.h" + +#include "vbo/vbo.h" +#include "vbo/vbo_context.h" + +#include "st_context.h" +#include "st_atom.h" +#include "st_draw.h" +#include "st_cb_feedback.h" +#include "st_cb_bufferobjects.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/tgsi/exec/tgsi_attribs.h" +#include "vf/vf.h" + +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_private.h" + + +/** + * This is actually used for both feedback and selection. + */ +struct feedback_stage +{ + struct draw_stage stage; /**< Base class */ + GLcontext *ctx; /**< Rendering context */ + GLboolean reset_stipple_counter; +}; + + +/********************************************************************** + * GL Feedback functions + **********************************************************************/ + +static INLINE struct feedback_stage * +feedback_stage( struct draw_stage *stage ) +{ + return (struct feedback_stage *)stage; +} + + +static void +feedback_vertex(GLcontext *ctx, const struct draw_context *draw, + const struct vertex_header *v) +{ + GLfloat win[4]; + const GLfloat *color, *texcoord; + const GLfloat ci = 0; + GLuint slot; + + win[0] = v->data[0][0]; + win[1] = v->data[0][1]; + win[2] = v->data[0][2]; + win[3] = 1.0F / v->data[0][3]; + + slot = draw->vertex_info.attrib_to_slot[TGSI_ATTRIB_COLOR0]; + if (slot) + color = v->data[slot]; + else + color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; + + slot = draw->vertex_info.attrib_to_slot[TGSI_ATTRIB_TEX0]; + if (slot) + texcoord = v->data[slot]; + else + texcoord = ctx->Current.Attrib[VERT_ATTRIB_TEX0]; + + _mesa_feedback_vertex(ctx, win, color, ci, texcoord); +} + + +static void +feedback_tri( struct draw_stage *stage, struct prim_header *prim ) +{ + struct feedback_stage *fs = feedback_stage(stage); + struct draw_context *draw = stage->draw; + FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_POLYGON_TOKEN); + FEEDBACK_TOKEN(fs->ctx, (GLfloat) 3); /* three vertices */ + feedback_vertex(fs->ctx, draw, prim->v[0]); + feedback_vertex(fs->ctx, draw, prim->v[1]); + feedback_vertex(fs->ctx, draw, prim->v[2]); +} + + +static void +feedback_line( struct draw_stage *stage, struct prim_header *prim ) +{ + struct feedback_stage *fs = feedback_stage(stage); + struct draw_context *draw = stage->draw; + if (fs->reset_stipple_counter) { + FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_LINE_RESET_TOKEN); + fs->reset_stipple_counter = GL_FALSE; + } + else { + FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_LINE_TOKEN); + } + feedback_vertex(fs->ctx, draw, prim->v[0]); + feedback_vertex(fs->ctx, draw, prim->v[1]); +} + + +static void +feedback_point( struct draw_stage *stage, struct prim_header *prim ) +{ + struct feedback_stage *fs = feedback_stage(stage); + struct draw_context *draw = stage->draw; + FEEDBACK_TOKEN(fs->ctx, (GLfloat) GL_POINT_TOKEN); + feedback_vertex(fs->ctx, draw, prim->v[0]); +} + + +static void +feedback_end( struct draw_stage *stage ) +{ + /* no-op */ +} + + +static void +feedback_begin( struct draw_stage *stage ) +{ + /* no-op */ +} + + +static void +feedback_reset_stipple_counter( struct draw_stage *stage ) +{ + struct feedback_stage *fs = feedback_stage(stage); + fs->reset_stipple_counter = GL_TRUE; +} + + +/** + * Create GL feedback drawing stage. + */ +static struct draw_stage * +draw_glfeedback_stage(GLcontext *ctx, struct draw_context *draw) +{ + struct feedback_stage *fs = CALLOC_STRUCT(feedback_stage); + + fs->stage.draw = draw; + fs->stage.next = NULL; + fs->stage.begin = feedback_begin; + fs->stage.point = feedback_point; + fs->stage.line = feedback_line; + fs->stage.tri = feedback_tri; + fs->stage.end = feedback_end; + fs->stage.reset_stipple_counter = feedback_reset_stipple_counter; + fs->ctx = ctx; + + return &fs->stage; +} + + + +/********************************************************************** + * GL Selection functions + **********************************************************************/ + +static void +select_tri( struct draw_stage *stage, struct prim_header *prim ) +{ + struct feedback_stage *fs = feedback_stage(stage); + _mesa_update_hitflag( fs->ctx, prim->v[0]->data[0][2] ); + _mesa_update_hitflag( fs->ctx, prim->v[1]->data[0][2] ); + _mesa_update_hitflag( fs->ctx, prim->v[2]->data[0][2] ); +} + +static void +select_line( struct draw_stage *stage, struct prim_header *prim ) +{ + struct feedback_stage *fs = feedback_stage(stage); + _mesa_update_hitflag( fs->ctx, prim->v[0]->data[0][2] ); + _mesa_update_hitflag( fs->ctx, prim->v[1]->data[0][2] ); +} + + +static void +select_point( struct draw_stage *stage, struct prim_header *prim ) +{ + struct feedback_stage *fs = feedback_stage(stage); + _mesa_update_hitflag( fs->ctx, prim->v[0]->data[0][2] ); +} + + +static void +select_begin( struct draw_stage *stage ) +{ + /* no-op */ +} + + +static void +select_end( struct draw_stage *stage ) +{ + /* no-op */ +} + + +static void +select_reset_stipple_counter( struct draw_stage *stage ) +{ + /* no-op */ +} + + +/** + * Create GL selection mode drawing stage. + */ +static struct draw_stage * +draw_glselect_stage(GLcontext *ctx, struct draw_context *draw) +{ + struct feedback_stage *fs = CALLOC_STRUCT(feedback_stage); + + fs->stage.draw = draw; + fs->stage.next = NULL; + fs->stage.begin = select_begin; + fs->stage.point = select_point; + fs->stage.line = select_line; + fs->stage.tri = select_tri; + fs->stage.end = select_end; + fs->stage.reset_stipple_counter = select_reset_stipple_counter; + fs->ctx = ctx; + + return &fs->stage; +} + + +static void +st_RenderMode(GLcontext *ctx, GLenum newMode ) +{ + struct st_context *st = ctx->st; + struct vbo_context *vbo = (struct vbo_context *) ctx->swtnl_im; + struct draw_context *draw = st->draw; + + if (newMode == GL_RENDER) { + /* restore normal VBO draw function */ + vbo->draw_prims = st_draw_vbo; + } + else if (newMode == GL_SELECT) { + if (!st->selection_stage) + st->selection_stage = draw_glselect_stage(ctx, draw); + draw_set_setup_stage(draw, st->selection_stage); + /* Plug in new vbo draw function */ + vbo->draw_prims = st_feedback_draw_vbo; + } + else { + if (!st->feedback_stage) + st->feedback_stage = draw_glfeedback_stage(ctx, draw); + draw_set_setup_stage(draw, st->feedback_stage); + /* Plug in new vbo draw function */ + vbo->draw_prims = st_feedback_draw_vbo; + } +} + + + +void st_init_feedback_functions(struct dd_function_table *functions) +{ + functions->RenderMode = st_RenderMode; +} diff --git a/src/mesa/state_tracker/st_cb_feedback.h b/src/mesa/state_tracker/st_cb_feedback.h new file mode 100644 index 0000000000..2559ba3817 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_feedback.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_FEEDBACK_H +#define ST_CB_FEEDBACK_H + + +extern void +st_init_feedback_functions(struct dd_function_table *functions); + + +#endif /* ST_CB_FEEDBACK_H */ -- cgit v1.2.3 From b4bacd1fca336f043d1ee3c3a346fbd42a3b02eb Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Sep 2007 14:24:11 -0600 Subject: Plug in selection/feedback code. Not quite finished yet. Selection/feedback are done with a private instance of the 'draw' module in the state tracker. Not quite all the draw context's state is set yet, namely vertex format info. Hold off on that for a bit... --- src/mesa/sources | 1 + src/mesa/state_tracker/st_context.c | 6 ++ src/mesa/state_tracker/st_context.h | 7 ++ src/mesa/state_tracker/st_draw.c | 180 ++++++++++++++++++++++++++++++++++-- src/mesa/state_tracker/st_draw.h | 20 ++++ 5 files changed, 205 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index a672ad2fcf..ec561294c5 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -210,6 +210,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_flush.c \ state_tracker/st_cb_drawpixels.c \ state_tracker/st_cb_fbo.c \ + state_tracker/st_cb_feedback.c \ state_tracker/st_cb_program.c \ state_tracker/st_cb_queryobj.c \ state_tracker/st_cb_rasterpos.c \ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index e66e65e2dd..6ab643f1fe 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -33,6 +33,7 @@ #include "st_cb_clear.h" #include "st_cb_drawpixels.h" #include "st_cb_fbo.h" +#include "st_cb_feedback.h" #include "st_cb_queryobj.h" #include "st_cb_rasterpos.h" #include "st_cb_readpixels.h" @@ -43,6 +44,7 @@ #include "st_draw.h" #include "st_program.h" #include "pipe/p_context.h" +#include "pipe/draw/draw_context.h" void st_invalidate_state(GLcontext * ctx, GLuint new_state) @@ -64,6 +66,8 @@ struct st_context *st_create_context( GLcontext *ctx, st->ctx = ctx; st->pipe = pipe; + st->draw = draw_create(); /* for selection/feedback */ + st->dirty.mesa = ~0; st->dirty.st = ~0; @@ -97,6 +101,7 @@ struct st_context *st_create_context( GLcontext *ctx, void st_destroy_context( struct st_context *st ) { + draw_destroy(st->draw); st_destroy_atoms( st ); st_destroy_draw( st ); @@ -120,6 +125,7 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_clear_functions(functions); st_init_drawpixels_functions(functions); st_init_fbo_functions(functions); + st_init_feedback_functions(functions); st_init_program_functions(functions); st_init_query_functions(functions); st_init_rasterpos_functions(functions); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 98b78af7c5..56cae7d973 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -37,6 +37,9 @@ struct st_region; struct st_texture_object; struct st_texture_image; struct st_fragment_program; +struct draw_context; +struct draw_stage; + #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 @@ -62,6 +65,10 @@ struct st_context struct pipe_context *pipe; + struct draw_context *draw; /**< For selection/feedback */ + struct draw_stage *feedback_stage; /**< For FL_FEEDBACK rendermode */ + struct draw_stage *selection_stage; /**< For GL_SELECT rendermode */ + /* Eventually will use a cache to feed the pipe with * create/bind/delete calls to constant state objects. Not yet * though, we just shove random objects across the interface. diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 1facc14ccd..69f4b7fa5b 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -46,6 +46,8 @@ #include "pipe/p_winsys.h" #include "pipe/tgsi/exec/tgsi_attribs.h" +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_context.h" static GLuint @@ -173,14 +175,14 @@ update_default_attribs_buffer(GLcontext *ctx) * we have something to render. * Basically, translate the information into the format expected by pipe. */ -static void -draw_vbo(GLcontext *ctx, - const struct gl_client_array **arrays, - const struct _mesa_prim *prims, - GLuint nr_prims, - const struct _mesa_index_buffer *ib, - GLuint min_index, - GLuint max_index) +void +st_draw_vbo(GLcontext *ctx, + const struct gl_client_array **arrays, + const struct _mesa_prim *prims, + GLuint nr_prims, + const struct _mesa_index_buffer *ib, + GLuint min_index, + GLuint max_index) { struct pipe_context *pipe = ctx->st->pipe; GLuint attr, i; @@ -356,6 +358,166 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, +/** + * Called by VBO to draw arrays when in selection or feedback mode. + * This is very much like the normal draw_vbo() function above. + * Look at code refactoring some day. + * Might move this into the failover module some day. + */ +void +st_feedback_draw_vbo(GLcontext *ctx, + const struct gl_client_array **arrays, + const struct _mesa_prim *prims, + GLuint nr_prims, + const struct _mesa_index_buffer *ib, + GLuint min_index, + GLuint max_index) +{ + struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; + struct draw_context *draw = st->draw; + GLuint attr, i; + GLbitfield attrsNeeded; + const unsigned attr0_offset = (unsigned) arrays[0]->Ptr; + struct pipe_buffer_handle *index_buffer_handle = 0; + + assert(ctx->RenderMode == GL_SELECT || + ctx->RenderMode == GL_FEEDBACK); + assert(draw); + + /* + * Set up the draw module's state. + * + * We'd like to do this less frequently, but the normal state-update + * code sends state updates to the pipe, not to our private draw module. + */ + assert(draw); + draw_set_viewport_state(draw, &st->state.viewport); + draw_set_clip_state(draw, &st->state.clip); + draw_set_setup_state(draw, &st->state.setup); + draw_set_vertex_shader(draw, &st->state.vs); + /* XXX need to set vertex info too */ + + + update_default_attribs_buffer(ctx); + + /* this must be after state validation */ + attrsNeeded = ctx->st->state.vs.inputs_read; + + /* tell draw module about the vertex array element/attributes */ + for (attr = 0; attr < 16; attr++) { + struct pipe_vertex_buffer vbuffer; + struct pipe_vertex_element velement; + void *map; + + vbuffer.buffer = NULL; + vbuffer.pitch = 0; + velement.src_offset = 0; + velement.vertex_buffer_index = 0; + velement.src_format = 0; + + if (attrsNeeded & (1 << attr)) { + const GLuint mesaAttr = tgsi_attrib_to_mesa_attrib(attr); + struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; + + if (bufobj && bufobj->Name) { + struct st_buffer_object *stobj = st_buffer_object(bufobj); + /* Recall that for VBOs, the gl_client_array->Ptr field is + * really an offset from the start of the VBO, not a pointer. + */ + unsigned offset = (unsigned) arrays[mesaAttr]->Ptr; + + assert(stobj->buffer); + + vbuffer.buffer = stobj->buffer; + vbuffer.buffer_offset = attr0_offset; /* in bytes */ + vbuffer.pitch = arrays[mesaAttr]->StrideB; /* in bytes */ + vbuffer.max_index = 0; /* need this? */ + + velement.src_offset = offset - attr0_offset; /* bytes */ + velement.vertex_buffer_index = attr; + velement.dst_offset = 0; /* need this? */ + velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, + arrays[mesaAttr]->Size); + assert(velement.src_format); + } + else { + /* use the default attribute buffer */ + vbuffer.buffer = ctx->st->default_attrib_buffer; + vbuffer.buffer_offset = 0; + vbuffer.pitch = 0; /* must be zero! */ + vbuffer.max_index = 1; + + velement.src_offset = attr * 4 * sizeof(GLfloat); + velement.vertex_buffer_index = attr; + velement.dst_offset = 0; + velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + } + } + + if (attr == 0) + assert(vbuffer.buffer); + + draw_set_vertex_buffer(draw, attr, &vbuffer); + draw_set_vertex_element(draw, attr, &velement); + + /* map the attrib buffer */ + if (vbuffer.buffer) { + map = pipe->winsys->buffer_map(pipe->winsys, + vbuffer.buffer, + PIPE_BUFFER_FLAG_READ); + draw_set_mapped_vertex_buffer(draw, attr, map); + } + } + + + if (ib) { + unsigned indexSize; + struct gl_buffer_object *bufobj = ib->obj; + struct st_buffer_object *stobj = st_buffer_object(bufobj); + index_buffer_handle = stobj->buffer; + void *map; + + switch (ib->type) { + case GL_UNSIGNED_INT: + indexSize = 4; + break; + case GL_UNSIGNED_SHORT: + indexSize = 2; + break; + default: + assert(0); + } + + map = pipe->winsys->buffer_map(pipe->winsys, + index_buffer_handle, + PIPE_BUFFER_FLAG_READ); + draw_set_mapped_element_buffer(draw, indexSize, map); + } + + + /* draw here */ + for (i = 0; i < nr_prims; i++) { + draw_arrays(draw, prims[i].mode, prims[i].start, prims[i].count); + } + + + /* + * unmap vertex/index buffers + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (draw->vertex_buffer[i].buffer) { + pipe->winsys->buffer_unmap(pipe->winsys, + draw->vertex_buffer[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); + } + } + if (ib) { + pipe->winsys->buffer_unmap(pipe->winsys, index_buffer_handle); + draw_set_mapped_element_buffer(draw, 0, NULL); + } +} + /* This is all a hack to keep using tnl until we have vertex programs @@ -370,7 +532,7 @@ void st_init_draw( struct st_context *st ) assert(vbo); assert(vbo->draw_prims); - vbo->draw_prims = draw_vbo; + vbo->draw_prims = st_draw_vbo; _tnl_ProgramCacheInit( ctx ); } diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h index 0005fbc51f..1fcd1b7e6b 100644 --- a/src/mesa/state_tracker/st_draw.h +++ b/src/mesa/state_tracker/st_draw.h @@ -34,11 +34,31 @@ #ifndef ST_DRAW_H #define ST_DRAW_H +struct _mesa_prim; +struct _mesa_index_buffer; void st_init_draw( struct st_context *st ); void st_destroy_draw( struct st_context *st ); +extern void +st_draw_vbo(GLcontext *ctx, + const struct gl_client_array **arrays, + const struct _mesa_prim *prims, + GLuint nr_prims, + const struct _mesa_index_buffer *ib, + GLuint min_index, + GLuint max_index); + +extern void +st_feedback_draw_vbo(GLcontext *ctx, + const struct gl_client_array **arrays, + const struct _mesa_prim *prims, + GLuint nr_prims, + const struct _mesa_index_buffer *ib, + GLuint min_index, + GLuint max_index); + void st_draw_vertices(GLcontext *ctx, unsigned prim, unsigned numVertex, float *verts, -- cgit v1.2.3 From ffacb1c12a77d71613e8171e31ffc348959881e4 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Sep 2007 15:44:59 -0600 Subject: Replace GL_TEXTURE_* tokens with PIPE_TEXTURE_* Must have missed this a while back. This fixes cube mapping. --- src/mesa/state_tracker/st_mipmap_tree.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index 535004c892..1764bc61f3 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -180,18 +180,22 @@ st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) { static const GLuint zero = 0; - if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) + if (mt->target != PIPE_TEXTURE_3D || mt->level[level].nr_images == 1) return &zero; else return mt->level[level].image_offset; } +/** + * Return the offset to the given mipmap texture image within the + * texture memory buffer, in bytes. + */ GLuint st_miptree_image_offset(const struct pipe_mipmap_tree * mt, GLuint face, GLuint level) { - if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) + if (mt->target == PIPE_TEXTURE_CUBE) return (mt->level[level].level_offset + mt->level[level].image_offset[face] * mt->cpp); else -- cgit v1.2.3 From 9780327c5d95586a88fce94d7b47342355ead118 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 14 Sep 2007 04:08:58 -0400 Subject: First stab at immutable state objects (create/bind/delete) We want our state objects to be immutable, handled via the create/bind/delete calls instead of struct propagation. Only implementing the blend state to see how it would look like and work. --- src/mesa/cso_cache/cso_cache.c | 176 ++++++++++++++++++++++++ src/mesa/cso_cache/cso_cache.h | 75 ++++++++++ src/mesa/pipe/failover/fo_context.h | 3 +- src/mesa/pipe/failover/fo_state.c | 9 +- src/mesa/pipe/failover/fo_state_emit.c | 2 +- src/mesa/pipe/i915simple/i915_context.h | 3 +- src/mesa/pipe/i915simple/i915_state.c | 29 +++- src/mesa/pipe/i915simple/i915_state_dynamic.c | 14 +- src/mesa/pipe/i915simple/i915_state_immediate.c | 20 +-- src/mesa/pipe/p_context.h | 12 +- src/mesa/pipe/softpipe/sp_context.c | 5 +- src/mesa/pipe/softpipe/sp_context.h | 3 +- src/mesa/pipe/softpipe/sp_quad.c | 4 +- src/mesa/pipe/softpipe/sp_quad_blend.c | 12 +- src/mesa/pipe/softpipe/sp_quad_colormask.c | 8 +- src/mesa/pipe/softpipe/sp_state.h | 10 +- src/mesa/pipe/softpipe/sp_state_blend.c | 19 ++- src/mesa/sources | 4 + src/mesa/state_tracker/st_atom_blend.c | 11 +- src/mesa/state_tracker/st_cb_clear.c | 7 +- src/mesa/state_tracker/st_cb_drawpixels.c | 4 +- src/mesa/state_tracker/st_context.c | 4 + src/mesa/state_tracker/st_context.h | 7 +- 23 files changed, 382 insertions(+), 59 deletions(-) create mode 100644 src/mesa/cso_cache/cso_cache.c create mode 100644 src/mesa/cso_cache/cso_cache.h (limited to 'src') diff --git a/src/mesa/cso_cache/cso_cache.c b/src/mesa/cso_cache/cso_cache.c new file mode 100644 index 0000000000..bb853308be --- /dev/null +++ b/src/mesa/cso_cache/cso_cache.c @@ -0,0 +1,176 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Zack Rusin + */ + +#include "cso_cache.h" + +#if 1 +static unsigned hash_key( const void *key, unsigned key_size ) +{ + unsigned *ikey = (unsigned *)key; + unsigned hash = 0, i; + + assert(key_size % 4 == 0); + + /* I'm sure this can be improved on: + */ + for (i = 0; i < key_size/4; i++) + hash ^= ikey[i]; + + return hash; +} +#else +static unsigned hash_key(const unsigned char *p, int n) +{ + unsigned h = 0; + unsigned g; + + while (n--) { + h = (h << 4) + *p++; + if ((g = (h & 0xf0000000)) != 0) + h ^= g >> 23; + h &= ~g; + } + return h; +} +#endif + +unsigned cso_construct_key(void *item, int item_size) +{ + return hash_key((const unsigned char*)(item), item_size); +} + +struct cso_cache_item * +cso_insert_state(struct cso_cache *sc, + unsigned hash_key, + void *state, int state_size) +{ + struct cso_cache_item *found_state = + _mesa_HashLookup(sc->hash, hash_key); + struct cso_cache_item *item = + malloc(sizeof(struct cso_cache_item)); + _mesa_printf("inserting state ========= key = %d\n", hash_key); + item->key = hash_key; + item->state_size = state_size; + item->state = state; + item->next = 0; + + if (found_state) { + while (found_state->next) + found_state = found_state->next; + found_state->next = item; + } else + _mesa_HashInsert(sc->hash, hash_key, item); + return item; +} + +struct cso_cache_item * +cso_find_state(struct cso_cache *sc, + unsigned hash_key, + void *state, int state_size) +{ + struct cso_cache_item *found_state = + _mesa_HashLookup(sc->hash, hash_key); + + while (found_state && + (found_state->state_size != state_size || + memcmp(found_state->state, state, state_size))) { + found_state = found_state->next; + } + + _mesa_printf("finding state ========== %d (%p)\n", hash_key, found_state); + return found_state; +} + +struct cso_cache_item * +cso_remove_state(struct cso_cache *sc, + unsigned hash_key, + void *state, int state_size) +{ + struct cso_cache_item *found_state = + _mesa_HashLookup(sc->hash, hash_key); + struct cso_cache_item *prev = 0; + + while (found_state && + (found_state->state_size != state_size || + memcmp(found_state->state, state, state_size))) { + prev = found_state; + found_state = found_state->next; + } + if (found_state) { + if (prev) + prev->next = found_state->next; + else { + if (found_state->next) + _mesa_HashInsert(sc->hash, hash_key, found_state->next); + else + _mesa_HashRemove(sc->hash, hash_key); + } + } + return found_state; +} + +struct cso_cache *cso_cache_create(void) +{ + struct cso_cache *sc = malloc(sizeof(struct cso_cache)); + + sc->hash = _mesa_NewHashTable(); + + return sc; +} + +void cso_cache_destroy(struct cso_cache *sc) +{ + assert(sc); + assert(sc->hash); + _mesa_DeleteHashTable(sc->hash); + free(sc); +} + +/* This function will either find the state of the given template + * in the cache or it will create a new state state from the given + * template, will insert it in the cache and return it. + */ +struct pipe_blend_state * cso_cached_blend_state( + struct st_context *st, + const struct pipe_blend_state *blend) +{ + unsigned hash_key = cso_construct_key((void*)blend, sizeof(struct pipe_blend_state)); + struct cso_cache_item *cache_item = cso_find_state(st->cache, + hash_key, + (void*)blend, + sizeof(struct pipe_blend_state)); + if (!cache_item) { + const struct pipe_blend_state *created_state = st->pipe->create_blend_state( + st->pipe, blend); + cache_item = cso_insert_state(st->cache, hash_key, + (void*)created_state, sizeof(struct pipe_blend_state)); + } + return (struct pipe_blend_state*)cache_item->state; +} diff --git a/src/mesa/cso_cache/cso_cache.h b/src/mesa/cso_cache/cso_cache.h new file mode 100644 index 0000000000..ca0a2d576a --- /dev/null +++ b/src/mesa/cso_cache/cso_cache.h @@ -0,0 +1,75 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#ifndef CSO_CACHE_H +#define CSO_CACHE_H + +#include "state_tracker/st_context.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" + + +#include "main/hash.h" + +struct cso_cache_item { + unsigned key; + + unsigned state_size; + const void *state; + + struct cso_cache_item *next; +}; + +struct cso_cache { + struct _mesa_HashTable *hash; +}; + +void cso_cache_destroy(struct cso_cache *sc); +struct cso_cache *cso_cache_create(void); + +unsigned cso_construct_key(void *item, int item_size); + +struct cso_cache_item *cso_insert_state(struct cso_cache *sc, + unsigned hash_key, + void *state, int state_size); +struct cso_cache_item *cso_find_state(struct cso_cache *sc, + unsigned hash_key, + void *state, int state_size); +struct cso_cache_item *cso_remove_state(struct cso_cache *sc, + unsigned hash_key, + void *state, int state_size); + +struct pipe_blend_state *cso_cached_blend_state( + struct st_context *pipe, + const struct pipe_blend_state *state); + +#endif diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index 5666d4e830..b065aa832a 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -67,8 +67,9 @@ struct failover_context { /* The most recent drawing state as set by the driver: */ + const struct pipe_blend_state *blend; + struct pipe_alpha_test_state alpha_test; - struct pipe_blend_state blend; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 097acf7d57..2357d7ef5c 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -58,14 +58,14 @@ failover_set_alpha_test_state(struct pipe_context *pipe, static void -failover_set_blend_state( struct pipe_context *pipe, +failover_bind_blend_state( struct pipe_context *pipe, const struct pipe_blend_state *blend ) { struct failover_context *failover = failover_context(pipe); - failover->blend = *blend; + failover->blend = blend; failover->dirty |= FO_NEW_BLEND; - failover->hw->set_blend_state( failover->hw, blend ); + failover->hw->bind_blend_state( failover->hw, blend ); } @@ -266,9 +266,10 @@ failover_set_vertex_element(struct pipe_context *pipe, void failover_init_state_functions( struct failover_context *failover ) { + failover->pipe.bind_blend_state = failover_bind_blend_state; + failover->pipe.set_alpha_test_state = failover_set_alpha_test_state; failover->pipe.set_blend_color = failover_set_blend_color; - failover->pipe.set_blend_state = failover_set_blend_state; failover->pipe.set_clip_state = failover_set_clip_state; failover->pipe.set_clear_color_state = failover_set_clear_color_state; failover->pipe.set_depth_state = failover_set_depth_test_state; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 52fcf5dbc9..77413d100b 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -59,7 +59,7 @@ failover_state_emit( struct failover_context *failover ) failover->sw->set_alpha_test_state( failover->sw, &failover->alpha_test ); if (failover->dirty & FO_NEW_BLEND) - failover->sw->set_blend_state( failover->sw, &failover->blend ); + failover->sw->bind_blend_state( failover->sw, failover->blend ); if (failover->dirty & FO_NEW_BLEND_COLOR) failover->sw->set_blend_color( failover->sw, &failover->blend_color ); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index c30c79d83c..215c5294fa 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -123,8 +123,9 @@ struct i915_context /* The most recent drawing state as set by the driver: */ + const struct pipe_blend_state *blend; + struct pipe_alpha_test_state alpha_test; - struct pipe_blend_state blend; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index f5ea721cc8..478988fd4a 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -37,17 +37,37 @@ /* None of this state is actually used for anything yet. */ -static void i915_set_blend_state( struct pipe_context *pipe, + +static const struct pipe_blend_state * +i915_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + /*struct i915_context *i915 = i915_context(pipe);*/ + + struct pipe_blend_state *new_blend = malloc(sizeof(struct pipe_blend_state)); + memcpy(new_blend, blend, sizeof(struct pipe_blend_state)); + + return new_blend; +} + +static void i915_bind_blend_state( struct pipe_context *pipe, const struct pipe_blend_state *blend ) { struct i915_context *i915 = i915_context(pipe); - i915->blend = *blend; + i915->blend = blend; i915->dirty |= I915_NEW_BLEND; } +static void i915_delete_blend_state( struct pipe_context *pipe, + const struct pipe_blend_state *blend ) +{ + /*struct i915_context *i915 = i915_context(pipe);*/ + free(blend); +} + static void i915_set_blend_color( struct pipe_context *pipe, const struct pipe_blend_color *blend_color ) { @@ -289,9 +309,12 @@ static void i915_set_vertex_element( struct pipe_context *pipe, void i915_init_state_functions( struct i915_context *i915 ) { + i915->pipe.create_blend_state = i915_create_blend_state; + i915->pipe.bind_blend_state = i915_bind_blend_state; + i915->pipe.delete_blend_state = i915_delete_blend_state; + i915->pipe.set_alpha_test_state = i915_set_alpha_test_state; i915->pipe.set_blend_color = i915_set_blend_color; - i915->pipe.set_blend_state = i915_set_blend_state; i915->pipe.set_clip_state = i915_set_clip_state; i915->pipe.set_clear_color_state = i915_set_clear_color_state; i915->pipe.set_constant_buffer = i915_set_constant_buffer; diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index e648357754..49a30fac11 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -82,7 +82,7 @@ static void upload_MODES4( struct i915_context *i915 ) { modes4 |= (_3DSTATE_MODES_4_CMD | ENABLE_LOGIC_OP_FUNC | - LOGIC_OP_FUNC(i915_translate_logic_op(i915->blend.logicop_func))); + LOGIC_OP_FUNC(i915_translate_logic_op(i915->blend->logicop_func))); } /* Always, so that we know when state is in-active: @@ -204,13 +204,13 @@ static void upload_IAB( struct i915_context *i915 ) unsigned iab = 0; { - unsigned eqRGB = i915->blend.rgb_func; - unsigned srcRGB = i915->blend.rgb_src_factor; - unsigned dstRGB = i915->blend.rgb_dst_factor; + unsigned eqRGB = i915->blend->rgb_func; + unsigned srcRGB = i915->blend->rgb_src_factor; + unsigned dstRGB = i915->blend->rgb_dst_factor; - unsigned eqA = i915->blend.alpha_func; - unsigned srcA = i915->blend.alpha_src_factor; - unsigned dstA = i915->blend.alpha_dst_factor; + unsigned eqA = i915->blend->alpha_func; + unsigned srcA = i915->blend->alpha_src_factor; + unsigned dstA = i915->blend->alpha_dst_factor; /* Special handling for MIN/MAX filter modes handled at * state_tracker level. diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 38a24733e1..aaca534f5a 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -145,22 +145,22 @@ static void upload_S5( struct i915_context *i915 ) } /* I915_NEW_BLEND */ - if (i915->blend.logicop_enable) + if (i915->blend->logicop_enable) LIS5 |= S5_LOGICOP_ENABLE; - if (i915->blend.dither) + if (i915->blend->dither) LIS5 |= S5_COLOR_DITHER_ENABLE; - if ((i915->blend.colormask & PIPE_MASK_R) == 0) + if ((i915->blend->colormask & PIPE_MASK_R) == 0) LIS5 |= S5_WRITEDISABLE_RED; - if ((i915->blend.colormask & PIPE_MASK_G) == 0) + if ((i915->blend->colormask & PIPE_MASK_G) == 0) LIS5 |= S5_WRITEDISABLE_GREEN; - if ((i915->blend.colormask & PIPE_MASK_B) == 0) + if ((i915->blend->colormask & PIPE_MASK_B) == 0) LIS5 |= S5_WRITEDISABLE_BLUE; - if ((i915->blend.colormask & PIPE_MASK_A) == 0) + if ((i915->blend->colormask & PIPE_MASK_A) == 0) LIS5 |= S5_WRITEDISABLE_ALPHA; @@ -205,11 +205,11 @@ static void upload_S6( struct i915_context *i915 ) /* I915_NEW_BLEND */ - if (i915->blend.blend_enable) + if (i915->blend->blend_enable) { - unsigned funcRGB = i915->blend.rgb_func; - unsigned srcRGB = i915->blend.rgb_src_factor; - unsigned dstRGB = i915->blend.rgb_dst_factor; + unsigned funcRGB = i915->blend->rgb_func; + unsigned srcRGB = i915->blend->rgb_src_factor; + unsigned dstRGB = i915->blend->rgb_dst_factor; LIS6 |= (S6_CBUF_BLEND_ENABLE | SRC_BLND_FACT(i915_translate_blend_factor(srcRGB)) | diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index ec5555c38e..b9af69fc05 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -31,7 +31,7 @@ #include "p_state.h" #include "p_compiler.h" - +struct pipe_state_cache; /** * Software pipeline rendering context. Basically a collection of * state setting functions, plus VBO drawing entrypoint. @@ -85,12 +85,16 @@ struct pipe_context { /* * State functions */ + const struct pipe_blend_state * (*create_blend_state)(struct pipe_context *, + const struct pipe_blend_state *); + void (*bind_blend_state)(struct pipe_context *, + const struct pipe_blend_state *); + void (*delete_blend_state)(struct pipe_context *, + const struct pipe_blend_state *); + void (*set_alpha_test_state)( struct pipe_context *, const struct pipe_alpha_test_state * ); - void (*set_blend_state)( struct pipe_context *, - const struct pipe_blend_state * ); - void (*set_blend_color)( struct pipe_context *, const struct pipe_blend_color * ); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 92357808e2..b9c7013e9e 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -250,9 +250,12 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.max_texture_size = softpipe_max_texture_size; /* state setters */ + softpipe->pipe.create_blend_state = softpipe_create_blend_state; + softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; + softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; + softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; - softpipe->pipe.set_blend_state = softpipe_set_blend_state; softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 13d1143c89..7fecf2974a 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -71,8 +71,9 @@ struct softpipe_context { /* The most recent drawing state as set by the driver: */ + const struct pipe_blend_state *blend; + struct pipe_alpha_test_state alpha_test; - struct pipe_blend_state blend; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 0f0736479f..2fcbea1f22 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -11,12 +11,12 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.output; - if (sp->blend.colormask != 0xf) { + if (sp->blend->colormask != 0xf) { sp->quad.colormask->next = sp->quad.first; sp->quad.first = sp->quad.colormask; } - if (sp->blend.blend_enable) { + if (sp->blend->blend_enable) { sp->quad.blend->next = sp->quad.first; sp->quad.first = sp->quad.blend; } diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index e5335f3b19..6c7135e53c 100755 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -111,7 +111,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) /* * Compute src/first term RGB */ - switch (softpipe->blend.rgb_src_factor) { + switch (softpipe->blend->rgb_src_factor) { case PIPE_BLENDFACTOR_ONE: VEC4_COPY(source[0], quad->outputs.color[0]); /* R */ VEC4_COPY(source[1], quad->outputs.color[1]); /* G */ @@ -253,7 +253,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) /* * Compute src/first term A */ - switch (softpipe->blend.alpha_src_factor) { + switch (softpipe->blend->alpha_src_factor) { case PIPE_BLENDFACTOR_ONE: VEC4_COPY(source[3], quad->outputs.color[3]); /* A */ break; @@ -275,7 +275,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) /* * Compute dest/second term RGB */ - switch (softpipe->blend.rgb_dst_factor) { + switch (softpipe->blend->rgb_dst_factor) { case PIPE_BLENDFACTOR_ONE: /* dest = dest * 1 NO-OP, leave dest as-is */ break; @@ -301,7 +301,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) /* * Compute dest/second term A */ - switch (softpipe->blend.alpha_dst_factor) { + switch (softpipe->blend->alpha_dst_factor) { case PIPE_BLENDFACTOR_ONE: /* dest = dest * 1 NO-OP, leave dest as-is */ break; @@ -323,7 +323,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) /* * Combine RGB terms */ - switch (softpipe->blend.rgb_func) { + switch (softpipe->blend->rgb_func) { case PIPE_BLEND_ADD: VEC4_ADD(quad->outputs.color[0], source[0], dest[0]); /* R */ VEC4_ADD(quad->outputs.color[1], source[1], dest[1]); /* G */ @@ -356,7 +356,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) /* * Combine A terms */ - switch (softpipe->blend.alpha_func) { + switch (softpipe->blend->alpha_func) { case PIPE_BLEND_ADD: VEC4_ADD(quad->outputs.color[3], source[3], dest[3]); /* A */ break; diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index 7bb65bf8c8..c3e110532a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -49,19 +49,19 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); /* R */ - if (!(softpipe->blend.colormask & PIPE_MASK_R)) + if (!(softpipe->blend->colormask & PIPE_MASK_R)) COPY_4V(quad->outputs.color[0], dest[0]); /* G */ - if (!(softpipe->blend.colormask & PIPE_MASK_G)) + if (!(softpipe->blend->colormask & PIPE_MASK_G)) COPY_4V(quad->outputs.color[1], dest[1]); /* B */ - if (!(softpipe->blend.colormask & PIPE_MASK_B)) + if (!(softpipe->blend->colormask & PIPE_MASK_B)) COPY_4V(quad->outputs.color[2], dest[2]); /* A */ - if (!(softpipe->blend.colormask & PIPE_MASK_A)) + if (!(softpipe->blend->colormask & PIPE_MASK_A)) COPY_4V(quad->outputs.color[3], dest[3]); /* pass quad to next stage */ diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index f40ebe3e2b..e2b1a2a164 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -33,6 +33,13 @@ #include "pipe/p_state.h" +const struct pipe_blend_state * +softpipe_create_blend_state(struct pipe_context *, + const struct pipe_blend_state *); +void softpipe_bind_blend_state(struct pipe_context *, + const struct pipe_blend_state *); +void softpipe_delete_blend_state(struct pipe_context *, + const struct pipe_blend_state *); void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); @@ -40,9 +47,6 @@ void softpipe_set_framebuffer_state( struct pipe_context *, void softpipe_set_alpha_test_state( struct pipe_context *, const struct pipe_alpha_test_state * ); -void softpipe_set_blend_state( struct pipe_context *, - const struct pipe_blend_state * ); - void softpipe_set_blend_color( struct pipe_context *pipe, const struct pipe_blend_color *blend_color ); diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index b2e85d86cc..57f2df7923 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -30,17 +30,32 @@ #include "sp_context.h" #include "sp_state.h" +const struct pipe_blend_state * +softpipe_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + struct pipe_blend_state *new_blend = malloc(sizeof(struct pipe_blend_state)); + memcpy(new_blend, blend, sizeof(struct pipe_blend_state)); + + return new_blend; +} -void softpipe_set_blend_state( struct pipe_context *pipe, +void softpipe_bind_blend_state( struct pipe_context *pipe, const struct pipe_blend_state *blend ) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->blend = *blend; + softpipe->blend = blend; softpipe->dirty |= SP_NEW_BLEND; } +void softpipe_delete_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend ) +{ + free(blend); +} + void softpipe_set_blend_color( struct pipe_context *pipe, const struct pipe_blend_color *blend_color ) diff --git a/src/mesa/sources b/src/mesa/sources index ec561294c5..e57942d664 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -186,6 +186,9 @@ TGSIDECO_SOURCES = \ TGSIMESA_SOURCES = \ pipe/tgsi/mesa/mesa_to_tgsi.c +STATECACHE_SOURCES = \ + cso_cache/cso_cache.c + STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ state_tracker/st_atom_alphatest.c \ @@ -373,6 +376,7 @@ SOLO_SOURCES = \ $(TGSIEXEC_SOURCES) \ $(TGSIDECO_SOURCES) \ $(TGSIMESA_SOURCES) \ + $(STATECACHE_SOURCES) \ $(STATETRACKER_SOURCES) \ $(TNL_SOURCES) \ $(SHADER_SOURCES) \ diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index afd21a6141..d007d50ad3 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -36,6 +36,7 @@ #include "st_atom.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "cso_cache/cso_cache.h" /** @@ -209,10 +210,14 @@ update_blend( struct st_context *st ) if (st->ctx->Color.DitherFlag) blend.dither = 1; - if (memcmp(&blend, &st->state.blend, sizeof(blend)) != 0) { + struct pipe_blend_state *real_blend = + cso_cached_blend_state(st, &blend); + + if (st->state.blend != real_blend) { /* state has changed */ - st->state.blend = blend; /* struct copy */ - st->pipe->set_blend_state(st->pipe, &blend); /* set new state */ + st->state.blend = real_blend; + /* bind new state */ + st->pipe->bind_blend_state(st->pipe, real_blend); } if (memcmp(st->ctx->Color.BlendColor, &st->state.blend_color, 4 * sizeof(GLfloat)) != 0) { diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index fa222df2a4..dc8a84af08 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -49,6 +49,8 @@ #include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "cso_cache/cso_cache.h" + #include "vf/vf.h" @@ -295,7 +297,8 @@ clear_with_quad(GLcontext *ctx, if (st->ctx->Color.DitherFlag) blend.dither = 1; } - pipe->set_blend_state(pipe, &blend); + const struct pipe_blend_state *state = cso_cached_blend_state(st, &blend); + pipe->bind_blend_state(pipe, state); } /* depth state: always pass */ @@ -390,7 +393,7 @@ clear_with_quad(GLcontext *ctx, /* Restore pipe state */ pipe->set_alpha_test_state(pipe, &st->state.alpha_test); - pipe->set_blend_state(pipe, &st->state.blend); + pipe->bind_blend_state(pipe, st->state.blend); pipe->set_depth_state(pipe, &st->state.depth); pipe->set_fs_state(pipe, &st->state.fs); pipe->set_vs_state(pipe, &st->state.vs); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index e2280cdafa..c2d231cdb5 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -474,8 +474,8 @@ static GLboolean any_fragment_ops(const struct st_context *st) { if (st->state.alpha_test.enabled || - st->state.blend.blend_enable || - st->state.blend.logicop_enable || + st->state.blend->blend_enable || + st->state.blend->logicop_enable || st->state.depth.enabled) /* XXX more checks */ return GL_TRUE; diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 6ab643f1fe..26815d5cd4 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -45,6 +45,7 @@ #include "st_program.h" #include "pipe/p_context.h" #include "pipe/draw/draw_context.h" +#include "cso_cache/cso_cache.h" void st_invalidate_state(GLcontext * ctx, GLuint new_state) @@ -71,6 +72,8 @@ struct st_context *st_create_context( GLcontext *ctx, st->dirty.mesa = ~0; st->dirty.st = ~0; + st->cache = cso_cache_create(); + st_init_atoms( st ); st_init_draw( st ); @@ -112,6 +115,7 @@ void st_destroy_context( struct st_context *st ) /*st_destroy_cb_teximage( st );*/ st_destroy_cb_texture( st ); #endif + cso_cache_destroy( st->cache ); st->pipe->destroy( st->pipe ); FREE( st ); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 56cae7d973..bd2efdb960 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -39,7 +39,7 @@ struct st_texture_image; struct st_fragment_program; struct draw_context; struct draw_stage; - +struct cso_cache; #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 @@ -74,8 +74,9 @@ struct st_context * though, we just shove random objects across the interface. */ struct { + const struct pipe_blend_state *blend; + struct pipe_alpha_test_state alpha_test; - struct pipe_blend_state blend; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; @@ -122,6 +123,8 @@ struct st_context struct st_fragment_program *fp; struct pipe_buffer_handle *default_attrib_buffer; + + struct cso_cache *cache; }; -- cgit v1.2.3 From e16c045b83f5c5b4f4064df67623bb76b46b6619 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 17 Sep 2007 07:56:56 -0400 Subject: Implementing a better hash, removing state_tracker dependency from the cache. Replacing mesa's main hash with one that handles collisions, moving state_tracker related caching to the state tracker to keep cso cache independent of it. Cleanups. --- src/mesa/cso_cache/cso_cache.c | 139 +++++------- src/mesa/cso_cache/cso_cache.h | 43 ++-- src/mesa/cso_cache/cso_hash.c | 381 +++++++++++++++++++++++++++++++++ src/mesa/cso_cache/cso_hash.h | 62 ++++++ src/mesa/sources | 2 + src/mesa/state_tracker/st_atom_blend.c | 5 +- src/mesa/state_tracker/st_cache.c | 61 ++++++ src/mesa/state_tracker/st_cache.h | 43 ++++ src/mesa/state_tracker/st_cb_clear.c | 5 +- src/mesa/state_tracker/st_context.c | 2 +- 10 files changed, 627 insertions(+), 116 deletions(-) create mode 100644 src/mesa/cso_cache/cso_hash.c create mode 100644 src/mesa/cso_cache/cso_hash.h create mode 100644 src/mesa/state_tracker/st_cache.c create mode 100644 src/mesa/state_tracker/st_cache.h (limited to 'src') diff --git a/src/mesa/cso_cache/cso_cache.c b/src/mesa/cso_cache/cso_cache.c index bb853308be..784d1f970f 100644 --- a/src/mesa/cso_cache/cso_cache.c +++ b/src/mesa/cso_cache/cso_cache.c @@ -29,9 +29,10 @@ */ #include "cso_cache.h" +#include "cso_hash.h" #if 1 -static unsigned hash_key( const void *key, unsigned key_size ) +static unsigned hash_key(const void *key, unsigned key_size) { unsigned *ikey = (unsigned *)key; unsigned hash = 0, i; @@ -63,114 +64,84 @@ static unsigned hash_key(const unsigned char *p, int n) unsigned cso_construct_key(void *item, int item_size) { - return hash_key((const unsigned char*)(item), item_size); + return hash_key((item), item_size); } -struct cso_cache_item * +static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_type type) +{ + struct cso_hash *hash = 0; + + switch(type) { + case CSO_BLEND: + hash = sc->blend_hash; + } + + return hash; +} + +static int _cso_size_for_type(enum cso_cache_type type) +{ + switch(type) { + case CSO_BLEND: + return sizeof(struct pipe_blend_state); + } + return 0; +} + +struct cso_hash_iter cso_insert_state(struct cso_cache *sc, - unsigned hash_key, - void *state, int state_size) + unsigned hash_key, enum cso_cache_type type, + void *state) { - struct cso_cache_item *found_state = - _mesa_HashLookup(sc->hash, hash_key); - struct cso_cache_item *item = - malloc(sizeof(struct cso_cache_item)); - _mesa_printf("inserting state ========= key = %d\n", hash_key); - item->key = hash_key; - item->state_size = state_size; - item->state = state; - item->next = 0; - - if (found_state) { - while (found_state->next) - found_state = found_state->next; - found_state->next = item; - } else - _mesa_HashInsert(sc->hash, hash_key, item); - return item; + struct cso_hash *hash = _cso_hash_for_type(sc, type); + return cso_hash_insert(hash, hash_key, state); } -struct cso_cache_item * +struct cso_hash_iter cso_find_state(struct cso_cache *sc, - unsigned hash_key, - void *state, int state_size) + unsigned hash_key, enum cso_cache_type type) { - struct cso_cache_item *found_state = - _mesa_HashLookup(sc->hash, hash_key); + struct cso_hash *hash = _cso_hash_for_type(sc, type); - while (found_state && - (found_state->state_size != state_size || - memcmp(found_state->state, state, state_size))) { - found_state = found_state->next; - } - - _mesa_printf("finding state ========== %d (%p)\n", hash_key, found_state); - return found_state; + return cso_hash_find(hash, hash_key); } -struct cso_cache_item * -cso_remove_state(struct cso_cache *sc, - unsigned hash_key, - void *state, int state_size) +struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *templ) { - struct cso_cache_item *found_state = - _mesa_HashLookup(sc->hash, hash_key); - struct cso_cache_item *prev = 0; - - while (found_state && - (found_state->state_size != state_size || - memcmp(found_state->state, state, state_size))) { - prev = found_state; - found_state = found_state->next; - } - if (found_state) { - if (prev) - prev->next = found_state->next; - else { - if (found_state->next) - _mesa_HashInsert(sc->hash, hash_key, found_state->next); - else - _mesa_HashRemove(sc->hash, hash_key); - } + struct cso_hash_iter iter = cso_find_state(sc, hash_key, type); + int size = _cso_size_for_type(type); + while (!cso_hash_iter_is_null(iter)) { + void *iter_data = cso_hash_iter_data(iter); + if (!memcmp(iter_data, templ, size)) + return iter; + iter = cso_hash_iter_next(iter); } - return found_state; + return iter; +} + +void * cso_take_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + return cso_hash_take(hash, hash_key); } struct cso_cache *cso_cache_create(void) { struct cso_cache *sc = malloc(sizeof(struct cso_cache)); - sc->hash = _mesa_NewHashTable(); + sc->blend_hash = cso_hash_create(); return sc; } -void cso_cache_destroy(struct cso_cache *sc) +void cso_cache_delete(struct cso_cache *sc) { assert(sc); - assert(sc->hash); - _mesa_DeleteHashTable(sc->hash); + assert(sc->blend_hash); + cso_hash_delete(sc->blend_hash); free(sc); } -/* This function will either find the state of the given template - * in the cache or it will create a new state state from the given - * template, will insert it in the cache and return it. - */ -struct pipe_blend_state * cso_cached_blend_state( - struct st_context *st, - const struct pipe_blend_state *blend) -{ - unsigned hash_key = cso_construct_key((void*)blend, sizeof(struct pipe_blend_state)); - struct cso_cache_item *cache_item = cso_find_state(st->cache, - hash_key, - (void*)blend, - sizeof(struct pipe_blend_state)); - if (!cache_item) { - const struct pipe_blend_state *created_state = st->pipe->create_blend_state( - st->pipe, blend); - cache_item = cso_insert_state(st->cache, hash_key, - (void*)created_state, sizeof(struct pipe_blend_state)); - } - return (struct pipe_blend_state*)cache_item->state; -} diff --git a/src/mesa/cso_cache/cso_cache.h b/src/mesa/cso_cache/cso_cache.h index ca0a2d576a..c022b98d43 100644 --- a/src/mesa/cso_cache/cso_cache.h +++ b/src/mesa/cso_cache/cso_cache.h @@ -33,43 +33,34 @@ #ifndef CSO_CACHE_H #define CSO_CACHE_H -#include "state_tracker/st_context.h" #include "pipe/p_context.h" #include "pipe/p_state.h" -#include "main/hash.h" - -struct cso_cache_item { - unsigned key; - - unsigned state_size; - const void *state; - - struct cso_cache_item *next; -}; +struct cso_hash; struct cso_cache { - struct _mesa_HashTable *hash; + struct cso_hash *blend_hash; }; -void cso_cache_destroy(struct cso_cache *sc); -struct cso_cache *cso_cache_create(void); +enum cso_cache_type { + CSO_BLEND, +}; unsigned cso_construct_key(void *item, int item_size); -struct cso_cache_item *cso_insert_state(struct cso_cache *sc, - unsigned hash_key, - void *state, int state_size); -struct cso_cache_item *cso_find_state(struct cso_cache *sc, - unsigned hash_key, - void *state, int state_size); -struct cso_cache_item *cso_remove_state(struct cso_cache *sc, - unsigned hash_key, - void *state, int state_size); +struct cso_cache *cso_cache_create(void); +void cso_cache_delete(struct cso_cache *sc); -struct pipe_blend_state *cso_cached_blend_state( - struct st_context *pipe, - const struct pipe_blend_state *state); +struct cso_hash_iter cso_insert_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *state); +struct cso_hash_iter cso_find_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type); +struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *templ); +void * cso_take_state(struct cso_cache *sc, unsigned hash_key, + enum cso_cache_type type); #endif diff --git a/src/mesa/cso_cache/cso_hash.c b/src/mesa/cso_cache/cso_hash.c new file mode 100644 index 0000000000..fe033c9e91 --- /dev/null +++ b/src/mesa/cso_cache/cso_hash.c @@ -0,0 +1,381 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#include "cso_hash.h" + +#include +#include +#include +#include + +#define MAX(a, b) ((a > b) ? (a) : (b)) + +static const int MinNumBits = 4; + +static const unsigned char prime_deltas[] = { + 0, 0, 1, 3, 1, 5, 3, 3, 1, 9, 7, 5, 3, 9, 25, 3, + 1, 21, 3, 21, 7, 15, 9, 5, 3, 29, 15, 0, 0, 0, 0, 0 +}; + +static int primeForNumBits(int numBits) +{ + return (1 << numBits) + prime_deltas[numBits]; +} + +/* + Returns the smallest integer n such that + primeForNumBits(n) >= hint. +*/ +static int countBits(int hint) +{ + int numBits = 0; + int bits = hint; + + while (bits > 1) { + bits >>= 1; + numBits++; + } + + if (numBits >= (int)sizeof(prime_deltas)) { + numBits = sizeof(prime_deltas) - 1; + } else if (primeForNumBits(numBits) < hint) { + ++numBits; + } + return numBits; +} + +struct cso_node { + struct cso_node *next; + unsigned key; + void *value; +}; + +struct cso_hash_data { + struct cso_node *fakeNext; + struct cso_node **buckets; + int size; + int nodeSize; + short userNumBits; + short numBits; + int numBuckets; +}; + +struct cso_hash { + union { + struct cso_hash_data *d; + struct cso_node *e; + } data; +}; + +static void *cso_data_allocate_node(struct cso_hash_data *hash) +{ + return malloc(hash->nodeSize); +} + +static void cso_data_free_node(void *node) +{ + free(node); +} + +static struct cso_node * +cso_hash_create_node(struct cso_hash *hash, + unsigned akey, void *avalue, + struct cso_node **anextNode) +{ + struct cso_node *node = cso_data_allocate_node(hash->data.d); + node->key = akey; + node->value = avalue; + + node->next = (struct cso_node*)(*anextNode); + *anextNode = node; + ++hash->data.d->size; + return node; +} + +static void cso_data_rehash(struct cso_hash_data *hash, int hint) +{ + if (hint < 0) { + hint = countBits(-hint); + if (hint < MinNumBits) + hint = MinNumBits; + hash->userNumBits = hint; + while (primeForNumBits(hint) < (hash->size >> 1)) + ++hint; + } else if (hint < MinNumBits) { + hint = MinNumBits; + } + + if (hash->numBits != hint) { + struct cso_node *e = (struct cso_node *)(hash); + struct cso_node **oldBuckets = hash->buckets; + int oldNumBuckets = hash->numBuckets; + int i = 0; + + hash->numBits = hint; + hash->numBuckets = primeForNumBits(hint); + hash->buckets = malloc(sizeof(struct cso_node*) * hash->numBuckets); + for (i = 0; i < hash->numBuckets; ++i) + hash->buckets[i] = e; + + for (i = 0; i < oldNumBuckets; ++i) { + struct cso_node *firstNode = oldBuckets[i]; + while (firstNode != e) { + unsigned h = firstNode->key; + struct cso_node *lastNode = firstNode; + while (lastNode->next != e && lastNode->next->key == h) + lastNode = lastNode->next; + + struct cso_node *afterLastNode = lastNode->next; + struct cso_node **beforeFirstNode = &hash->buckets[h % hash->numBuckets]; + while (*beforeFirstNode != e) + beforeFirstNode = &(*beforeFirstNode)->next; + lastNode->next = *beforeFirstNode; + *beforeFirstNode = firstNode; + firstNode = afterLastNode; + } + } + free(oldBuckets); + } +} + +static void cso_data_might_grow(struct cso_hash_data *hash) +{ + if (hash->size >= hash->numBuckets) + cso_data_rehash(hash, hash->numBits + 1); +} + +static void cso_data_has_shrunk(struct cso_hash_data *hash) +{ + if (hash->size <= (hash->numBuckets >> 3) && + hash->numBits > hash->userNumBits) { + int max = MAX(hash->numBits-2, hash->userNumBits); + cso_data_rehash(hash, max); + } +} + +static struct cso_node *cso_data_first_node(struct cso_hash_data *hash) +{ + struct cso_node *e = (struct cso_node *)(hash); + struct cso_node **bucket = hash->buckets; + int n = hash->numBuckets; + while (n--) { + if (*bucket != e) + return *bucket; + ++bucket; + } + return e; +} + +static struct cso_node **cso_hash_find_node(struct cso_hash *hash, unsigned akey) +{ + struct cso_node **node; + + if (hash->data.d->numBuckets) { + node = (struct cso_node **)(&hash->data.d->buckets[akey % hash->data.d->numBuckets]); + assert(*node == hash->data.e || (*node)->next); + while (*node != hash->data.e && (*node)->key != akey) + node = &(*node)->next; + } else { + node = (struct cso_node **)((const struct cso_node * const *)(&hash->data.e)); + } + return node; +} + +struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, + unsigned key, void *data) +{ + cso_data_might_grow(hash->data.d); + + struct cso_node **nextNode = cso_hash_find_node(hash, key); + struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); + struct cso_hash_iter iter = {hash, node}; + return iter; +} + +struct cso_hash * cso_hash_create(void) +{ + struct cso_hash *hash = malloc(sizeof(struct cso_hash)); + hash->data.d = malloc(sizeof(struct cso_hash_data)); + hash->data.d->fakeNext = 0; + hash->data.d->buckets = 0; + hash->data.d->size = 0; + hash->data.d->nodeSize = sizeof(struct cso_node); + hash->data.d->userNumBits = MinNumBits; + hash->data.d->numBits = 0; + hash->data.d->numBuckets = 0; + + return hash; +} + +void cso_hash_delete(struct cso_hash *hash) +{ + struct cso_node *e_for_x = (struct cso_node *)(hash->data.d); + struct cso_node **bucket = (struct cso_node **)(hash->data.d->buckets); + int n = hash->data.d->numBuckets; + while (n--) { + struct cso_node *cur = *bucket++; + while (cur != e_for_x) { + struct cso_node *next = cur->next; + cso_data_free_node(cur); + cur = next; + } + } + free(hash->data.d->buckets); + free(hash->data.d); + free(hash); +} + +struct cso_hash_iter cso_hash_find(struct cso_hash *hash, + unsigned key) +{ + struct cso_node **nextNode = cso_hash_find_node(hash, key); + struct cso_hash_iter iter = {hash, *nextNode}; + return iter; +} + +unsigned cso_hash_iter_key(struct cso_hash_iter iter) +{ + if (!iter.node || iter.hash->data.e == iter.node) + return 0; + return iter.node->key; +} + +void * cso_hash_iter_data(struct cso_hash_iter iter) +{ + if (!iter.node || iter.hash->data.e == iter.node) + return 0; + return iter.node->value; +} + +static struct cso_node *cso_hash_data_next(struct cso_node *node) +{ + union { + struct cso_node *next; + struct cso_node *e; + struct cso_hash_data *d; + } a; + a.next = node->next; + if (!a.next) { + fprintf(stderr, "iterating beyond the last element\n"); + return 0; + } + if (a.next->next) + return a.next; + + int start = (node->key % a.d->numBuckets) + 1; + struct cso_node **bucket = a.d->buckets + start; + int n = a.d->numBuckets - start; + while (n--) { + if (*bucket != a.e) + return *bucket; + ++bucket; + } + return a.e; +} + + +static struct cso_node *cso_hash_data_prev(struct cso_node *node) +{ + union { + struct cso_node *e; + struct cso_hash_data *d; + } a; + + a.e = node; + while (a.e->next) + a.e = a.e->next; + + int start; + if (node == a.e) + start = a.d->numBuckets - 1; + else + start = node->key % a.d->numBuckets; + + struct cso_node *sentinel = node; + struct cso_node **bucket = a.d->buckets + start; + while (start >= 0) { + if (*bucket != sentinel) { + struct cso_node *prev = *bucket; + while (prev->next != sentinel) + prev = prev->next; + return prev; + } + + sentinel = a.e; + --bucket; + --start; + } + fprintf(stderr, "iterating backward beyond first element\n"); + return a.e; +} + +struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter) +{ + struct cso_hash_iter next = {iter.hash, cso_hash_data_next(iter.node)}; + return next; +} + +int cso_hash_iter_is_null(struct cso_hash_iter iter) +{ + if (!iter.node || iter.node == iter.hash->data.e) + return 1; + return 0; +} + +void * cso_hash_take(struct cso_hash *hash, + unsigned akey) +{ + struct cso_node **node = cso_hash_find_node(hash, akey); + if (*node != hash->data.e) { + void *t = (*node)->value; + struct cso_node *next = (*node)->next; + cso_data_free_node(*node); + *node = next; + --hash->data.d->size; + cso_data_has_shrunk(hash->data.d); + return t; + } + return 0; +} + +struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter) +{ + struct cso_hash_iter prev = {iter.hash, + cso_hash_data_prev(iter.node)}; + return prev; +} + +struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash) +{ + struct cso_hash_iter iter = {hash, cso_data_first_node(hash->data.d)}; + return iter; +} diff --git a/src/mesa/cso_cache/cso_hash.h b/src/mesa/cso_cache/cso_hash.h new file mode 100644 index 0000000000..b4aa111860 --- /dev/null +++ b/src/mesa/cso_cache/cso_hash.h @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#ifndef CSO_HASH_H +#define CSO_HASH_H + +struct cso_hash; +struct cso_node; + +struct cso_hash_iter { + struct cso_hash *hash; + struct cso_node *node; +}; + +struct cso_hash *cso_hash_create(void); +void cso_hash_delete(struct cso_hash *hash); + +struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, + void *data); +void *cso_hash_take(struct cso_hash *hash, unsigned key); + +struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash); +struct cso_hash_iter cso_hash_find(struct cso_hash *hash, unsigned key); + + +int cso_hash_iter_is_null(struct cso_hash_iter iter); +unsigned cso_hash_iter_key(struct cso_hash_iter iter); +void *cso_hash_iter_data(struct cso_hash_iter iter); + +struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter); +struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); + +#endif diff --git a/src/mesa/sources b/src/mesa/sources index e57942d664..90fa5c65bf 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -187,6 +187,7 @@ TGSIMESA_SOURCES = \ pipe/tgsi/mesa/mesa_to_tgsi.c STATECACHE_SOURCES = \ + cso_cache/cso_hash.c \ cso_cache/cso_cache.c STATETRACKER_SOURCES = \ @@ -220,6 +221,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_readpixels.c \ state_tracker/st_cb_strings.c \ state_tracker/st_cb_texture.c \ + state_tracker/st_cache.c \ state_tracker/st_context.c \ state_tracker/st_draw.c \ state_tracker/st_format.c \ diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index d007d50ad3..d94beb66c7 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -33,10 +33,11 @@ #include "st_context.h" +#include "st_cache.h" #include "st_atom.h" + #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "cso_cache/cso_cache.h" /** @@ -211,7 +212,7 @@ update_blend( struct st_context *st ) blend.dither = 1; struct pipe_blend_state *real_blend = - cso_cached_blend_state(st, &blend); + st_cached_blend_state(st, &blend); if (st->state.blend != real_blend) { /* state has changed */ diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c new file mode 100644 index 0000000000..0205b1cab3 --- /dev/null +++ b/src/mesa/state_tracker/st_cache.c @@ -0,0 +1,61 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#include "st_cache.h" + +#include "st_context.h" + +#include "pipe/p_state.h" + +#include "cso_cache/cso_cache.h" +#include "cso_cache/cso_hash.h" + +/* This function will either find the state of the given template + * in the cache or it will create a new state state from the given + * template, will insert it in the cache and return it. + */ +struct pipe_blend_state * st_cached_blend_state( + struct st_context *st, + const struct pipe_blend_state *blend) +{ + unsigned hash_key = cso_construct_key((void*)blend, sizeof(struct pipe_blend_state)); + struct cso_hash_iter iter = cso_find_state_template(st->cache, + hash_key, CSO_BLEND, + (void*)blend); + if (cso_hash_iter_is_null(iter)) { + const struct pipe_blend_state *created_state = st->pipe->create_blend_state( + st->pipe, blend); + iter = cso_insert_state(st->cache, hash_key, CSO_BLEND, + (void*)created_state); + } + return (struct pipe_blend_state*)(cso_hash_iter_data(iter)); +} diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h new file mode 100644 index 0000000000..29b1d00dca --- /dev/null +++ b/src/mesa/state_tracker/st_cache.h @@ -0,0 +1,43 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#ifndef ST_CACHE_H +#define ST_CACHE_H + +struct pipe_blend_state; +struct st_context; + +struct pipe_blend_state * st_cached_blend_state( + struct st_context *st, + const struct pipe_blend_state *blend); + +#endif diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index dc8a84af08..55e03f644e 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -35,6 +35,7 @@ #include "main/macros.h" #include "shader/prog_instruction.h" #include "st_atom.h" +#include "st_cache.h" #include "st_context.h" #include "st_cb_clear.h" #include "st_cb_fbo.h" @@ -49,8 +50,6 @@ #include "pipe/tgsi/mesa/mesa_to_tgsi.h" -#include "cso_cache/cso_cache.h" - #include "vf/vf.h" @@ -297,7 +296,7 @@ clear_with_quad(GLcontext *ctx, if (st->ctx->Color.DitherFlag) blend.dither = 1; } - const struct pipe_blend_state *state = cso_cached_blend_state(st, &blend); + const struct pipe_blend_state *state = st_cached_blend_state(st, &blend); pipe->bind_blend_state(pipe, state); } diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 26815d5cd4..f9717465f4 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -115,7 +115,7 @@ void st_destroy_context( struct st_context *st ) /*st_destroy_cb_teximage( st );*/ st_destroy_cb_texture( st ); #endif - cso_cache_destroy( st->cache ); + cso_cache_delete( st->cache ); st->pipe->destroy( st->pipe ); FREE( st ); -- cgit v1.2.3 From f117327a3f246713abfd4dc4320d4a1a7f1b811a Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 17 Sep 2007 09:47:41 -0400 Subject: Make sampler an immutable state object. Switch the sample to be an immutable state object. --- src/mesa/cso_cache/cso_cache.c | 7 +++- src/mesa/cso_cache/cso_cache.h | 2 + src/mesa/pipe/failover/fo_context.h | 2 +- src/mesa/pipe/failover/fo_state.c | 8 ++-- src/mesa/pipe/failover/fo_state_emit.c | 4 +- src/mesa/pipe/i915simple/i915_context.h | 4 +- src/mesa/pipe/i915simple/i915_state.c | 54 +++++++++++++++++---------- src/mesa/pipe/i915simple/i915_state_sampler.c | 2 +- src/mesa/pipe/p_context.h | 13 +++++-- src/mesa/pipe/softpipe/sp_context.c | 4 +- src/mesa/pipe/softpipe/sp_context.h | 4 +- src/mesa/pipe/softpipe/sp_state.h | 13 +++++-- src/mesa/pipe/softpipe/sp_state_sampler.c | 26 +++++++++++-- src/mesa/state_tracker/st_atom_sampler.c | 10 +++-- src/mesa/state_tracker/st_cache.c | 17 +++++++++ src/mesa/state_tracker/st_cache.h | 6 +++ src/mesa/state_tracker/st_cb_drawpixels.c | 6 ++- src/mesa/state_tracker/st_context.h | 4 +- 18 files changed, 133 insertions(+), 53 deletions(-) (limited to 'src') diff --git a/src/mesa/cso_cache/cso_cache.c b/src/mesa/cso_cache/cso_cache.c index 784d1f970f..61da590575 100644 --- a/src/mesa/cso_cache/cso_cache.c +++ b/src/mesa/cso_cache/cso_cache.c @@ -74,6 +74,8 @@ static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_ switch(type) { case CSO_BLEND: hash = sc->blend_hash; + case CSO_SAMPLER: + hash = sc->sampler_hash; } return hash; @@ -84,6 +86,8 @@ static int _cso_size_for_type(enum cso_cache_type type) switch(type) { case CSO_BLEND: return sizeof(struct pipe_blend_state); + case CSO_SAMPLER: + return sizeof(struct pipe_sampler_state); } return 0; } @@ -133,6 +137,7 @@ struct cso_cache *cso_cache_create(void) struct cso_cache *sc = malloc(sizeof(struct cso_cache)); sc->blend_hash = cso_hash_create(); + sc->sampler_hash = cso_hash_create(); return sc; } @@ -140,8 +145,8 @@ struct cso_cache *cso_cache_create(void) void cso_cache_delete(struct cso_cache *sc) { assert(sc); - assert(sc->blend_hash); cso_hash_delete(sc->blend_hash); + cso_hash_delete(sc->sampler_hash); free(sc); } diff --git a/src/mesa/cso_cache/cso_cache.h b/src/mesa/cso_cache/cso_cache.h index c022b98d43..05a4cfcbdd 100644 --- a/src/mesa/cso_cache/cso_cache.h +++ b/src/mesa/cso_cache/cso_cache.h @@ -41,10 +41,12 @@ struct cso_hash; struct cso_cache { struct cso_hash *blend_hash; + struct cso_hash *sampler_hash; }; enum cso_cache_type { CSO_BLEND, + CSO_SAMPLER }; unsigned cso_construct_key(void *item, int item_size); diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index b065aa832a..fa336193a8 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -68,6 +68,7 @@ struct failover_context { /* The most recent drawing state as set by the driver: */ const struct pipe_blend_state *blend; + const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -79,7 +80,6 @@ struct failover_context { struct pipe_shader_state vertex_shader; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_setup_state setup; struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 2357d7ef5c..f3a99e4198 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -196,16 +196,16 @@ failover_set_stencil_state(struct pipe_context *pipe, static void -failover_set_sampler_state(struct pipe_context *pipe, +failover_bind_sampler_state(struct pipe_context *pipe, unsigned unit, const struct pipe_sampler_state *sampler) { struct failover_context *failover = failover_context(pipe); - failover->sampler[unit] = *sampler; + failover->sampler[unit] = sampler; failover->dirty |= FO_NEW_SAMPLER; failover->dirty_sampler |= (1<hw->set_sampler_state( failover->hw, unit, sampler ); + failover->hw->bind_sampler_state( failover->hw, unit, sampler ); } @@ -267,6 +267,7 @@ void failover_init_state_functions( struct failover_context *failover ) { failover->pipe.bind_blend_state = failover_bind_blend_state; + failover->pipe.bind_sampler_state = failover_bind_sampler_state; failover->pipe.set_alpha_test_state = failover_set_alpha_test_state; failover->pipe.set_blend_color = failover_set_blend_color; @@ -277,7 +278,6 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.set_fs_state = failover_set_fs_state; failover->pipe.set_vs_state = failover_set_vs_state; failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; - failover->pipe.set_sampler_state = failover_set_sampler_state; failover->pipe.set_scissor_state = failover_set_scissor_state; failover->pipe.set_setup_state = failover_set_setup_state; failover->pipe.set_stencil_state = failover_set_stencil_state; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 77413d100b..9d462678c5 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -100,8 +100,8 @@ failover_state_emit( struct failover_context *failover ) if (failover->dirty & FO_NEW_SAMPLER) { for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { if (failover->dirty_sampler & (1<sw->set_sampler_state( failover->sw, i, - &failover->sampler[i] ); + failover->sw->bind_sampler_state( failover->sw, i, + failover->sampler[i] ); } } } diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 215c5294fa..51baa281ec 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -123,7 +123,8 @@ struct i915_context /* The most recent drawing state as set by the driver: */ - const struct pipe_blend_state *blend; + const struct pipe_blend_state *blend; + const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -135,7 +136,6 @@ struct i915_context struct pipe_shader_state fs; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_setup_state setup; struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 478988fd4a..10060b45a4 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -42,8 +42,6 @@ static const struct pipe_blend_state * i915_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - /*struct i915_context *i915 = i915_context(pipe);*/ - struct pipe_blend_state *new_blend = malloc(sizeof(struct pipe_blend_state)); memcpy(new_blend, blend, sizeof(struct pipe_blend_state)); @@ -62,10 +60,9 @@ static void i915_bind_blend_state( struct pipe_context *pipe, static void i915_delete_blend_state( struct pipe_context *pipe, - const struct pipe_blend_state *blend ) + const struct pipe_blend_state *blend ) { - /*struct i915_context *i915 = i915_context(pipe);*/ - free(blend); + free((void*)blend); } static void i915_set_blend_color( struct pipe_context *pipe, @@ -78,6 +75,34 @@ static void i915_set_blend_color( struct pipe_context *pipe, i915->dirty |= I915_NEW_BLEND; } +static const struct pipe_sampler_state * +i915_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + struct pipe_sampler_state *new_sampler = malloc(sizeof(struct pipe_sampler_state)); + memcpy(new_sampler, sampler, sizeof(struct pipe_sampler_state)); + + return new_sampler; +} + +static void i915_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, + const struct pipe_sampler_state *sampler) +{ + struct i915_context *i915 = i915_context(pipe); + + assert(unit < PIPE_MAX_SAMPLERS); + i915->sampler[unit] = sampler; + + i915->dirty |= I915_NEW_SAMPLER; +} + +static void i915_delete_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + free((struct pipe_sampler_state*)sampler); +} + /** XXX move someday? Or consolidate all these simple state setters * into one file. @@ -189,19 +214,6 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, } -static void i915_set_sampler_state(struct pipe_context *pipe, - unsigned unit, - const struct pipe_sampler_state *sampler) -{ - struct i915_context *i915 = i915_context(pipe); - - assert(unit < PIPE_MAX_SAMPLERS); - i915->sampler[unit] = *sampler; - - i915->dirty |= I915_NEW_SAMPLER; -} - - static void i915_set_texture_state(struct pipe_context *pipe, unsigned unit, struct pipe_mipmap_tree *texture) @@ -293,7 +305,6 @@ static void i915_set_vertex_buffer( struct pipe_context *pipe, /* pass-through to draw module */ draw_set_vertex_buffer(i915->draw, index, buffer); } - static void i915_set_vertex_element( struct pipe_context *pipe, unsigned index, @@ -313,6 +324,10 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.bind_blend_state = i915_bind_blend_state; i915->pipe.delete_blend_state = i915_delete_blend_state; + i915->pipe.create_sampler_state = i915_create_sampler_state; + i915->pipe.bind_sampler_state = i915_bind_sampler_state; + i915->pipe.delete_sampler_state = i915_delete_sampler_state; + i915->pipe.set_alpha_test_state = i915_set_alpha_test_state; i915->pipe.set_blend_color = i915_set_blend_color; i915->pipe.set_clip_state = i915_set_clip_state; @@ -323,7 +338,6 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_fs_state = i915_set_fs_state; i915->pipe.set_vs_state = i915_set_vs_state; i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; - i915->pipe.set_sampler_state = i915_set_sampler_state; i915->pipe.set_scissor_state = i915_set_scissor_state; i915->pipe.set_setup_state = i915_set_setup_state; i915->pipe.set_stencil_state = i915_set_stencil_state; diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 7a595d1022..419a156136 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -269,7 +269,7 @@ void i915_update_samplers( struct i915_context *i915 ) if (i915->texture[unit]) { update_sampler( i915, unit, - i915->sampler + unit, /* sampler state */ + i915->sampler[unit], /* sampler state */ i915->texture[unit], /* mipmap tree */ i915->current.sampler[unit] /* the result */ ); diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index b9af69fc05..0913e49096 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -92,6 +92,15 @@ struct pipe_context { void (*delete_blend_state)(struct pipe_context *, const struct pipe_blend_state *); + const struct pipe_sampler_state * (*create_sampler_state)( + struct pipe_context *, + const struct pipe_sampler_state *); + void (*bind_sampler_state)(struct pipe_context *, + unsigned unit, + const struct pipe_sampler_state *); + void (*delete_sampler_state)(struct pipe_context *, + const struct pipe_sampler_state *); + void (*set_alpha_test_state)( struct pipe_context *, const struct pipe_alpha_test_state * ); @@ -135,10 +144,6 @@ struct pipe_context { void (*set_stencil_state)( struct pipe_context *, const struct pipe_stencil_state * ); - void (*set_sampler_state)( struct pipe_context *, - unsigned unit, - const struct pipe_sampler_state * ); - void (*set_texture_state)( struct pipe_context *, unsigned unit, struct pipe_mipmap_tree * ); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index b9c7013e9e..bab7985e8d 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -253,6 +253,9 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.create_blend_state = softpipe_create_blend_state; softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; + softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; + softpipe->pipe.bind_sampler_state = softpipe_bind_sampler_state; + softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; @@ -265,7 +268,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_fs_state = softpipe_set_fs_state; softpipe->pipe.set_vs_state = softpipe_set_vs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; - softpipe->pipe.set_sampler_state = softpipe_set_sampler_state; softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; softpipe->pipe.set_setup_state = softpipe_set_setup_state; softpipe->pipe.set_stencil_state = softpipe_set_stencil_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 7fecf2974a..5cee1a3cf9 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -71,7 +71,8 @@ struct softpipe_context { /* The most recent drawing state as set by the driver: */ - const struct pipe_blend_state *blend; + const struct pipe_blend_state *blend; + const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -85,7 +86,6 @@ struct softpipe_context { struct pipe_shader_state vs; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_setup_state setup; struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index e2b1a2a164..c8c93709db 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -41,6 +41,15 @@ void softpipe_bind_blend_state(struct pipe_context *, void softpipe_delete_blend_state(struct pipe_context *, const struct pipe_blend_state *); +const struct pipe_sampler_state * +softpipe_create_sampler_state(struct pipe_context *, + const struct pipe_sampler_state *); +void softpipe_bind_sampler_state(struct pipe_context *, + unsigned, + const struct pipe_sampler_state *); +void softpipe_delete_sampler_state(struct pipe_context *, + const struct pipe_sampler_state *); + void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); @@ -81,10 +90,6 @@ void softpipe_set_scissor_state( struct pipe_context *, void softpipe_set_setup_state( struct pipe_context *, const struct pipe_setup_state * ); -void softpipe_set_sampler_state( struct pipe_context *, - unsigned unit, - const struct pipe_sampler_state * ); - void softpipe_set_stencil_state( struct pipe_context *, const struct pipe_stencil_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 2e3d0c3d65..09898eb579 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -34,20 +34,38 @@ +const struct pipe_sampler_state * +softpipe_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + struct pipe_sampler_state *new_sampler = malloc(sizeof(struct pipe_sampler_state)); + memcpy(new_sampler, sampler, sizeof(struct pipe_sampler_state)); + + return new_sampler; +} + void -softpipe_set_sampler_state(struct pipe_context *pipe, - unsigned unit, - const struct pipe_sampler_state *sampler) +softpipe_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, + const struct pipe_sampler_state *sampler) { struct softpipe_context *softpipe = softpipe_context(pipe); assert(unit < PIPE_MAX_SAMPLERS); - softpipe->sampler[unit] = *sampler; + softpipe->sampler[unit] = sampler; softpipe->dirty |= SP_NEW_SAMPLER; } +void +softpipe_delete_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + free((struct pipe_sampler_state*)sampler); +} + + void softpipe_set_texture_state(struct pipe_context *pipe, unsigned unit, diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index d65565f991..9a728e2d79 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -33,6 +33,7 @@ #include "st_context.h" +#include "st_cache.h" #include "st_atom.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -142,10 +143,13 @@ update_samplers(struct st_context *st) /* XXX more sampler state here */ } - if (memcmp(&sampler, &st->state.sampler[u], sizeof(sampler)) != 0) { + const struct pipe_sampler_state *cached_sampler = + st_cached_sampler_state(st, &sampler); + + if (cached_sampler == st->state.sampler[u]) { /* state has changed */ - st->state.sampler[u] = sampler; - st->pipe->set_sampler_state(st->pipe, u, &sampler); + st->state.sampler[u] = cached_sampler; + st->pipe->bind_sampler_state(st->pipe, u, cached_sampler); } } } diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 0205b1cab3..99fb45f00a 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -59,3 +59,20 @@ struct pipe_blend_state * st_cached_blend_state( } return (struct pipe_blend_state*)(cso_hash_iter_data(iter)); } + +struct pipe_sampler_state * st_cached_sampler_state( + struct st_context *st, + const struct pipe_sampler_state *sampler) +{ + unsigned hash_key = cso_construct_key((void*)sampler, sizeof(struct pipe_sampler_state)); + struct cso_hash_iter iter = cso_find_state_template(st->cache, + hash_key, CSO_SAMPLER, + (void*)sampler); + if (cso_hash_iter_is_null(iter)) { + const struct pipe_sampler_state *created_state = st->pipe->create_sampler_state( + st->pipe, sampler); + iter = cso_insert_state(st->cache, hash_key, CSO_SAMPLER, + (void*)created_state); + } + return (struct pipe_sampler_state*)(cso_hash_iter_data(iter)); +} diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index 29b1d00dca..927585141c 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -34,10 +34,16 @@ #define ST_CACHE_H struct pipe_blend_state; +struct pipe_sampler_state; struct st_context; struct pipe_blend_state * st_cached_blend_state( struct st_context *st, const struct pipe_blend_state *blend); +struct pipe_sampler_state * st_cached_sampler_state( + struct st_context *st, + const struct pipe_sampler_state *sampler); + + #endif diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c2d231cdb5..31a37e5cd4 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -34,6 +34,7 @@ #include "st_context.h" #include "st_atom.h" +#include "st_cache.h" #include "st_draw.h" #include "st_program.h" #include "st_cb_drawpixels.h" @@ -359,7 +360,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; - pipe->set_sampler_state(pipe, unit, &sampler); + const struct pipe_sampler_state *state = st_cached_sampler_state(ctx->st, &sampler); + pipe->bind_sampler_state(pipe, unit, state); } /* viewport state: viewport matching window dims */ @@ -402,7 +404,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->set_fs_state(pipe, &ctx->st->state.fs); pipe->set_vs_state(pipe, &ctx->st->state.vs); pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); - pipe->set_sampler_state(pipe, unit, &ctx->st->state.sampler[unit]); + pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); free_mipmap_tree(pipe, mt); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index bd2efdb960..13526ff9e7 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -74,7 +74,8 @@ struct st_context * though, we just shove random objects across the interface. */ struct { - const struct pipe_blend_state *blend; + const struct pipe_blend_state *blend; + const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -86,7 +87,6 @@ struct st_context struct pipe_framebuffer_state framebuffer; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; - struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct pipe_scissor_state scissor; struct pipe_setup_state setup; struct pipe_shader_state fs; -- cgit v1.2.3 From 56edb98d975041cca2e4a3712126b151d80a045a Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 17 Sep 2007 09:50:08 -0400 Subject: Fix the warning. The const is there to enforce the immutable state of the object, which is in reality owned by the pipe so just cast away the constness. --- src/mesa/pipe/softpipe/sp_state_blend.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 57f2df7923..34da613f9d 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -53,7 +53,7 @@ void softpipe_bind_blend_state( struct pipe_context *pipe, void softpipe_delete_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend ) { - free(blend); + free((struct pipe_blend_state *)blend); } -- cgit v1.2.3 From d6ac959833a8e40a27907940969c622692f749b1 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 17 Sep 2007 11:55:18 -0400 Subject: Combing depth and stencil objects and making them immutable. Converting depth and stencil objects into a single state object (d3d10 like) and making it immutable. --- src/mesa/cso_cache/cso_cache.c | 6 + src/mesa/cso_cache/cso_cache.h | 4 +- src/mesa/pipe/failover/fo_context.h | 40 ++++--- src/mesa/pipe/failover/fo_state.c | 25 +---- src/mesa/pipe/failover/fo_state_emit.c | 7 +- src/mesa/pipe/i915simple/i915_context.h | 22 ++-- src/mesa/pipe/i915simple/i915_state.c | 44 +++++--- src/mesa/pipe/i915simple/i915_state_dynamic.c | 24 ++-- src/mesa/pipe/i915simple/i915_state_immediate.c | 22 ++-- src/mesa/pipe/p_context.h | 14 ++- src/mesa/pipe/p_state.h | 48 ++++---- src/mesa/pipe/softpipe/sp_context.c | 5 +- src/mesa/pipe/softpipe/sp_context.h | 26 ++--- src/mesa/pipe/softpipe/sp_quad.c | 8 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 4 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/pipe/softpipe/sp_quad_stencil.c | 32 +++--- src/mesa/pipe/softpipe/sp_state.h | 15 ++- src/mesa/pipe/softpipe/sp_state_blend.c | 34 +++--- src/mesa/pipe/softpipe/sp_state_derived.c | 7 +- src/mesa/sources | 1 - src/mesa/state_tracker/st_atom.c | 3 +- src/mesa/state_tracker/st_atom.h | 3 +- src/mesa/state_tracker/st_atom_depth.c | 108 +++++++++++++++--- src/mesa/state_tracker/st_atom_stencil.c | 141 ------------------------ src/mesa/state_tracker/st_cache.c | 17 +++ src/mesa/state_tracker/st_cache.h | 3 + src/mesa/state_tracker/st_cb_clear.c | 51 ++++----- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_context.h | 3 +- 30 files changed, 337 insertions(+), 384 deletions(-) delete mode 100644 src/mesa/state_tracker/st_atom_stencil.c (limited to 'src') diff --git a/src/mesa/cso_cache/cso_cache.c b/src/mesa/cso_cache/cso_cache.c index 61da590575..a4730394f8 100644 --- a/src/mesa/cso_cache/cso_cache.c +++ b/src/mesa/cso_cache/cso_cache.c @@ -76,6 +76,8 @@ static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_ hash = sc->blend_hash; case CSO_SAMPLER: hash = sc->sampler_hash; + case CSO_DEPTH_STENCIL: + hash = sc->depth_stencil_hash; } return hash; @@ -88,6 +90,8 @@ static int _cso_size_for_type(enum cso_cache_type type) return sizeof(struct pipe_blend_state); case CSO_SAMPLER: return sizeof(struct pipe_sampler_state); + case CSO_DEPTH_STENCIL: + return sizeof(struct pipe_depth_stencil_state); } return 0; } @@ -138,6 +142,7 @@ struct cso_cache *cso_cache_create(void) sc->blend_hash = cso_hash_create(); sc->sampler_hash = cso_hash_create(); + sc->depth_stencil_hash = cso_hash_create(); return sc; } @@ -147,6 +152,7 @@ void cso_cache_delete(struct cso_cache *sc) assert(sc); cso_hash_delete(sc->blend_hash); cso_hash_delete(sc->sampler_hash); + cso_hash_delete(sc->depth_stencil_hash); free(sc); } diff --git a/src/mesa/cso_cache/cso_cache.h b/src/mesa/cso_cache/cso_cache.h index 05a4cfcbdd..c340cf7f67 100644 --- a/src/mesa/cso_cache/cso_cache.h +++ b/src/mesa/cso_cache/cso_cache.h @@ -42,11 +42,13 @@ struct cso_hash; struct cso_cache { struct cso_hash *blend_hash; struct cso_hash *sampler_hash; + struct cso_hash *depth_stencil_hash; }; enum cso_cache_type { CSO_BLEND, - CSO_SAMPLER + CSO_SAMPLER, + CSO_DEPTH_STENCIL }; unsigned cso_construct_key(void *item, int item_size); diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index fa336193a8..63ec7239ab 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -36,25 +36,24 @@ -#define FO_NEW_VIEWPORT 0x1 -#define FO_NEW_SETUP 0x2 -#define FO_NEW_FRAGMENT_SHADER 0x4 -#define FO_NEW_BLEND 0x8 -#define FO_NEW_CLIP 0x10 -#define FO_NEW_SCISSOR 0x20 -#define FO_NEW_STIPPLE 0x40 -#define FO_NEW_FRAMEBUFFER 0x80 -#define FO_NEW_ALPHA_TEST 0x100 -#define FO_NEW_DEPTH_TEST 0x200 -#define FO_NEW_SAMPLER 0x400 -#define FO_NEW_TEXTURE 0x800 -#define FO_NEW_STENCIL 0x1000 -#define FO_NEW_VERTEX 0x2000 -#define FO_NEW_VERTEX_SHADER 0x4000 -#define FO_NEW_BLEND_COLOR 0x8000 -#define FO_NEW_CLEAR_COLOR 0x10000 -#define FO_NEW_VERTEX_BUFFER 0x20000 -#define FO_NEW_VERTEX_ELEMENT 0x40000 +#define FO_NEW_VIEWPORT 0x1 +#define FO_NEW_SETUP 0x2 +#define FO_NEW_FRAGMENT_SHADER 0x4 +#define FO_NEW_BLEND 0x8 +#define FO_NEW_CLIP 0x10 +#define FO_NEW_SCISSOR 0x20 +#define FO_NEW_STIPPLE 0x40 +#define FO_NEW_FRAMEBUFFER 0x80 +#define FO_NEW_ALPHA_TEST 0x100 +#define FO_NEW_DEPTH_STENCIL 0x200 +#define FO_NEW_SAMPLER 0x400 +#define FO_NEW_TEXTURE 0x800 +#define FO_NEW_VERTEX 0x2000 +#define FO_NEW_VERTEX_SHADER 0x4000 +#define FO_NEW_BLEND_COLOR 0x8000 +#define FO_NEW_CLEAR_COLOR 0x10000 +#define FO_NEW_VERTEX_BUFFER 0x20000 +#define FO_NEW_VERTEX_ELEMENT 0x40000 @@ -69,19 +68,18 @@ struct failover_context { */ const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct pipe_depth_stencil_state *depth_stencil; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; - struct pipe_depth_state depth_test; struct pipe_framebuffer_state framebuffer; struct pipe_shader_state fragment_shader; struct pipe_shader_state vertex_shader; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_setup_state setup; - struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index f3a99e4198..43b9757b31 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -103,14 +103,14 @@ failover_set_clear_color_state( struct pipe_context *pipe, } static void -failover_set_depth_test_state(struct pipe_context *pipe, - const struct pipe_depth_state *depth) +failover_bind_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *depth_stencil) { struct failover_context *failover = failover_context(pipe); - failover->depth_test = *depth; - failover->dirty |= FO_NEW_DEPTH_TEST; - failover->hw->set_depth_state( failover->hw, depth ); + failover->depth_stencil = depth_stencil; + failover->dirty |= FO_NEW_DEPTH_STENCIL; + failover->hw->bind_depth_stencil_state( failover->hw, depth_stencil ); } static void @@ -183,18 +183,6 @@ failover_set_scissor_state( struct pipe_context *pipe, failover->hw->set_scissor_state( failover->hw, scissor ); } -static void -failover_set_stencil_state(struct pipe_context *pipe, - const struct pipe_stencil_state *stencil) -{ - struct failover_context *failover = failover_context(pipe); - - failover->stencil = *stencil; - failover->dirty |= FO_NEW_STENCIL; - failover->hw->set_stencil_state( failover->hw, stencil ); -} - - static void failover_bind_sampler_state(struct pipe_context *pipe, unsigned unit, @@ -268,19 +256,18 @@ failover_init_state_functions( struct failover_context *failover ) { failover->pipe.bind_blend_state = failover_bind_blend_state; failover->pipe.bind_sampler_state = failover_bind_sampler_state; + failover->pipe.bind_depth_stencil_state = failover_bind_depth_stencil_state; failover->pipe.set_alpha_test_state = failover_set_alpha_test_state; failover->pipe.set_blend_color = failover_set_blend_color; failover->pipe.set_clip_state = failover_set_clip_state; failover->pipe.set_clear_color_state = failover_set_clear_color_state; - failover->pipe.set_depth_state = failover_set_depth_test_state; failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; failover->pipe.set_fs_state = failover_set_fs_state; failover->pipe.set_vs_state = failover_set_vs_state; failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; failover->pipe.set_scissor_state = failover_set_scissor_state; failover->pipe.set_setup_state = failover_set_setup_state; - failover->pipe.set_stencil_state = failover_set_stencil_state; failover->pipe.set_texture_state = failover_set_texture_state; failover->pipe.set_viewport_state = failover_set_viewport_state; failover->pipe.set_vertex_buffer = failover_set_vertex_buffer; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 9d462678c5..3a1865d766 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -70,8 +70,8 @@ failover_state_emit( struct failover_context *failover ) if (failover->dirty & FO_NEW_CLEAR_COLOR) failover->sw->set_clear_color_state( failover->sw, &failover->clear_color ); - if (failover->dirty & FO_NEW_DEPTH_TEST) - failover->sw->set_depth_state( failover->sw, &failover->depth_test ); + if (failover->dirty & FO_NEW_DEPTH_STENCIL) + failover->sw->bind_depth_stencil_state( failover->sw, failover->depth_stencil ); if (failover->dirty & FO_NEW_FRAMEBUFFER) failover->sw->set_framebuffer_state( failover->sw, &failover->framebuffer ); @@ -91,9 +91,6 @@ failover_state_emit( struct failover_context *failover ) if (failover->dirty & FO_NEW_SCISSOR) failover->sw->set_scissor_state( failover->sw, &failover->scissor ); - if (failover->dirty & FO_NEW_STENCIL) - failover->sw->set_stencil_state( failover->sw, &failover->stencil ); - if (failover->dirty & FO_NEW_VIEWPORT) failover->sw->set_viewport_state( failover->sw, &failover->viewport ); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 51baa281ec..518f780449 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -125,19 +125,18 @@ struct i915_context */ const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct pipe_depth_stencil_state *depth_stencil; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; - struct pipe_depth_state depth_test; struct pipe_framebuffer_state framebuffer; struct pipe_shader_state fs; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_setup_state setup; - struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; @@ -163,16 +162,15 @@ struct i915_context #define I915_NEW_SETUP 0x2 #define I915_NEW_FS 0x4 #define I915_NEW_BLEND 0x8 -#define I915_NEW_CLIP 0x10 -#define I915_NEW_SCISSOR 0x20 -#define I915_NEW_STIPPLE 0x40 -#define I915_NEW_FRAMEBUFFER 0x80 -#define I915_NEW_ALPHA_TEST 0x100 -#define I915_NEW_DEPTH_TEST 0x200 -#define I915_NEW_SAMPLER 0x400 -#define I915_NEW_TEXTURE 0x800 -#define I915_NEW_STENCIL 0x1000 -#define I915_NEW_CONSTANTS 0x2000 +#define I915_NEW_CLIP 0x10 +#define I915_NEW_SCISSOR 0x20 +#define I915_NEW_STIPPLE 0x40 +#define I915_NEW_FRAMEBUFFER 0x80 +#define I915_NEW_ALPHA_TEST 0x100 +#define I915_NEW_DEPTH_STENCIL 0x200 +#define I915_NEW_SAMPLER 0x400 +#define I915_NEW_TEXTURE 0x800 +#define I915_NEW_CONSTANTS 0x1000 /* Driver's internally generated state flags: diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 10060b45a4..5ac2e27d1a 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -107,38 +107,44 @@ static void i915_delete_sampler_state(struct pipe_context *pipe, /** XXX move someday? Or consolidate all these simple state setters * into one file. */ -static void i915_set_depth_test_state(struct pipe_context *pipe, - const struct pipe_depth_state *depth) -{ - struct i915_context *i915 = i915_context(pipe); - i915->depth_test = *depth; +static const struct pipe_depth_stencil_state * +i915_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *depth_stencil) +{ + struct pipe_depth_stencil_state *new_ds = + malloc(sizeof(struct pipe_depth_stencil_state)); + memcpy(new_ds, depth_stencil, sizeof(struct pipe_depth_stencil_state)); - i915->dirty |= I915_NEW_DEPTH_TEST; + return new_ds; } -static void i915_set_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *alpha) +static void i915_bind_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *depth_stencil) { struct i915_context *i915 = i915_context(pipe); - i915->alpha_test = *alpha; + i915->depth_stencil = depth_stencil; - i915->dirty |= I915_NEW_ALPHA_TEST; + i915->dirty |= I915_NEW_DEPTH_STENCIL; } -static void i915_set_stencil_state(struct pipe_context *pipe, - const struct pipe_stencil_state *stencil) +static void i915_delete_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *depth_stencil) +{ + free((struct pipe_depth_stencil_state *)depth_stencil); +} + +static void i915_set_alpha_test_state(struct pipe_context *pipe, + const struct pipe_alpha_test_state *alpha) { struct i915_context *i915 = i915_context(pipe); - i915->stencil = *stencil; + i915->alpha_test = *alpha; - i915->dirty |= I915_NEW_STENCIL; + i915->dirty |= I915_NEW_ALPHA_TEST; } - - static void i915_set_scissor_state( struct pipe_context *pipe, const struct pipe_scissor_state *scissor ) { @@ -328,19 +334,21 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.bind_sampler_state = i915_bind_sampler_state; i915->pipe.delete_sampler_state = i915_delete_sampler_state; + i915->pipe.create_depth_stencil_state = i915_create_depth_stencil_state; + i915->pipe.bind_depth_stencil_state = i915_bind_depth_stencil_state; + i915->pipe.delete_depth_stencil_state = i915_delete_depth_stencil_state; + i915->pipe.set_alpha_test_state = i915_set_alpha_test_state; i915->pipe.set_blend_color = i915_set_blend_color; i915->pipe.set_clip_state = i915_set_clip_state; i915->pipe.set_clear_color_state = i915_set_clear_color_state; i915->pipe.set_constant_buffer = i915_set_constant_buffer; - i915->pipe.set_depth_state = i915_set_depth_test_state; i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; i915->pipe.set_fs_state = i915_set_fs_state; i915->pipe.set_vs_state = i915_set_vs_state; i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; i915->pipe.set_scissor_state = i915_set_scissor_state; i915->pipe.set_setup_state = i915_set_setup_state; - i915->pipe.set_stencil_state = i915_set_stencil_state; i915->pipe.set_texture_state = i915_set_texture_state; i915->pipe.set_viewport_state = i915_set_viewport_state; i915->pipe.set_vertex_buffer = i915_set_vertex_buffer; diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index 49a30fac11..9140eee7c2 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -68,8 +68,8 @@ static void upload_MODES4( struct i915_context *i915 ) /* I915_NEW_STENCIL */ { - int testmask = i915->stencil.value_mask[0] & 0xff; - int writemask = i915->stencil.write_mask[0] & 0xff; + int testmask = i915->depth_stencil->stencil.value_mask[0] & 0xff; + int writemask = i915->depth_stencil->stencil.write_mask[0] & 0xff; modes4 |= (_3DSTATE_MODES_4_CMD | ENABLE_STENCIL_TEST_MASK | @@ -94,7 +94,7 @@ static void upload_MODES4( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_MODES4 = { - .dirty = I915_NEW_BLEND | I915_NEW_STENCIL, + .dirty = I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, .update = upload_MODES4 }; @@ -112,14 +112,14 @@ static void upload_BFO( struct i915_context *i915 ) /* _NEW_STENCIL */ - if (i915->stencil.back_enabled) { - int test = i915_translate_compare_func(i915->stencil.back_func); - int fop = i915_translate_stencil_op(i915->stencil.back_fail_op); - int dfop = i915_translate_stencil_op(i915->stencil.back_zfail_op); - int dpop = i915_translate_stencil_op(i915->stencil.back_zpass_op); - int ref = i915->stencil.ref_value[1] & 0xff; - int tmask = i915->stencil.value_mask[1] & 0xff; - int wmask = i915->stencil.write_mask[1] & 0xff; + if (i915->depth_stencil->stencil.back_enabled) { + int test = i915_translate_compare_func(i915->depth_stencil->stencil.back_func); + int fop = i915_translate_stencil_op(i915->depth_stencil->stencil.back_fail_op); + int dfop = i915_translate_stencil_op(i915->depth_stencil->stencil.back_zfail_op); + int dpop = i915_translate_stencil_op(i915->depth_stencil->stencil.back_zpass_op); + int ref = i915->depth_stencil->stencil.ref_value[1] & 0xff; + int tmask = i915->depth_stencil->stencil.value_mask[1] & 0xff; + int wmask = i915->depth_stencil->stencil.write_mask[1] & 0xff; bf[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_FUNCS | @@ -157,7 +157,7 @@ static void upload_BFO( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_BFO = { - .dirty = I915_NEW_STENCIL, + .dirty = I915_NEW_DEPTH_STENCIL, .update = upload_BFO }; diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index aaca534f5a..484913d308 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -128,12 +128,12 @@ static void upload_S5( struct i915_context *i915 ) unsigned LIS5 = 0; /* I915_NEW_STENCIL */ - if (i915->stencil.front_enabled) { - int test = i915_translate_compare_func(i915->stencil.front_func); - int fop = i915_translate_stencil_op(i915->stencil.front_fail_op); - int dfop = i915_translate_stencil_op(i915->stencil.front_zfail_op); - int dpop = i915_translate_stencil_op(i915->stencil.front_zpass_op); - int ref = i915->stencil.ref_value[0] & 0xff; + if (i915->depth_stencil->stencil.front_enabled) { + int test = i915_translate_compare_func(i915->depth_stencil->stencil.front_func); + int fop = i915_translate_stencil_op(i915->depth_stencil->stencil.front_fail_op); + int dfop = i915_translate_stencil_op(i915->depth_stencil->stencil.front_zfail_op); + int dpop = i915_translate_stencil_op(i915->depth_stencil->stencil.front_zpass_op); + int ref = i915->depth_stencil->stencil.ref_value[0] & 0xff; LIS5 |= (S5_STENCIL_TEST_ENABLE | S5_STENCIL_WRITE_ENABLE | @@ -179,7 +179,7 @@ static void upload_S5( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_S5 = { - .dirty = (I915_NEW_STENCIL | I915_NEW_BLEND | I915_NEW_SETUP), + .dirty = (I915_NEW_DEPTH_STENCIL | I915_NEW_BLEND | I915_NEW_SETUP), .update = upload_S5 }; @@ -219,13 +219,13 @@ static void upload_S6( struct i915_context *i915 ) /* I915_NEW_DEPTH */ - if (i915->depth_test.enabled) { - int func = i915_translate_compare_func(i915->depth_test.func); + if (i915->depth_stencil->depth.enabled) { + int func = i915_translate_compare_func(i915->depth_stencil->depth.func); LIS6 |= (S6_DEPTH_TEST_ENABLE | (func << S6_DEPTH_TEST_FUNC_SHIFT)); - if (i915->depth_test.writemask) + if (i915->depth_stencil->depth.writemask) LIS6 |= S6_DEPTH_WRITE_ENABLE; } @@ -236,7 +236,7 @@ static void upload_S6( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_S6 = { - .dirty = I915_NEW_ALPHA_TEST | I915_NEW_BLEND | I915_NEW_DEPTH_TEST, + .dirty = I915_NEW_ALPHA_TEST | I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, .update = upload_S6 }; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 0913e49096..488f002531 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -101,6 +101,14 @@ struct pipe_context { void (*delete_sampler_state)(struct pipe_context *, const struct pipe_sampler_state *); + const struct pipe_depth_stencil_state * (*create_depth_stencil_state)( + struct pipe_context *, + const struct pipe_depth_stencil_state *); + void (*bind_depth_stencil_state)(struct pipe_context *, + const struct pipe_depth_stencil_state *); + void (*delete_depth_stencil_state)(struct pipe_context *, + const struct pipe_depth_stencil_state *); + void (*set_alpha_test_state)( struct pipe_context *, const struct pipe_alpha_test_state * ); @@ -116,9 +124,6 @@ struct pipe_context { void (*set_constant_buffer)( struct pipe_context *, uint shader, uint index, const struct pipe_constant_buffer *buf ); - - void (*set_depth_state)( struct pipe_context *, - const struct pipe_depth_state * ); void (*set_feedback_state)( struct pipe_context *, const struct pipe_feedback_state *); @@ -141,9 +146,6 @@ struct pipe_context { void (*set_scissor_state)( struct pipe_context *, const struct pipe_scissor_state * ); - void (*set_stencil_state)( struct pipe_context *, - const struct pipe_stencil_state * ); - void (*set_texture_state)( struct pipe_context *, unsigned unit, struct pipe_mipmap_tree * ); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index b994d17ea9..30e559b594 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -146,13 +146,31 @@ struct pipe_shader_state { void *executable; }; -struct pipe_depth_state +struct pipe_depth_stencil_state { - unsigned enabled:1; /**< depth test enabled? */ - unsigned writemask:1; /**< allow depth buffer writes? */ - unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ - unsigned occlusion_count:1; /**< XXX move this elsewhere? */ - float clear; /**< Clear value in [0,1] (XXX correct place?) */ + struct { + unsigned enabled:1; /**< depth test enabled? */ + unsigned writemask:1; /**< allow depth buffer writes? */ + unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ + unsigned occlusion_count:1; /**< XXX move this elsewhere? */ + float clear; /**< Clear value in [0,1] (XXX correct place?) */ + } depth; + struct { + unsigned front_enabled:1; + unsigned front_func:3; /**< PIPE_FUNC_x */ + unsigned front_fail_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned front_zpass_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned front_zfail_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned back_enabled:1; + unsigned back_func:3; /**< PIPE_FUNC_x */ + unsigned back_fail_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned back_zpass_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned back_zfail_op:3; /**< PIPE_STENCIL_OP_x */ + ubyte ref_value[2]; /**< [0] = front, [1] = back */ + ubyte value_mask[2]; + ubyte write_mask[2]; + ubyte clear_value; + } stencil; }; struct pipe_alpha_test_state { @@ -188,24 +206,6 @@ struct pipe_clear_color_state float color[4]; }; -struct pipe_stencil_state { - unsigned front_enabled:1; - unsigned front_func:3; /**< PIPE_FUNC_x */ - unsigned front_fail_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned front_zpass_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned front_zfail_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned back_enabled:1; - unsigned back_func:3; /**< PIPE_FUNC_x */ - unsigned back_fail_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned back_zpass_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned back_zfail_op:3; /**< PIPE_STENCIL_OP_x */ - ubyte ref_value[2]; /**< [0] = front, [1] = back */ - ubyte value_mask[2]; - ubyte write_mask[2]; - ubyte clear_value; -}; - - struct pipe_framebuffer_state { /** multiple colorbuffers for multiple render targets */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index bab7985e8d..9a8b55bb0e 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -256,13 +256,15 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; softpipe->pipe.bind_sampler_state = softpipe_bind_sampler_state; softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; + softpipe->pipe.create_depth_stencil_state = softpipe_create_depth_stencil_state; + softpipe->pipe.bind_depth_stencil_state = softpipe_bind_depth_stencil_state; + softpipe->pipe.delete_depth_stencil_state = softpipe_delete_depth_stencil_state; softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; - softpipe->pipe.set_depth_state = softpipe_set_depth_test_state; softpipe->pipe.set_feedback_state = softpipe_set_feedback_state; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_fs_state = softpipe_set_fs_state; @@ -270,7 +272,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; softpipe->pipe.set_setup_state = softpipe_set_setup_state; - softpipe->pipe.set_stencil_state = softpipe_set_stencil_state; softpipe->pipe.set_texture_state = softpipe_set_texture_state; softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 5cee1a3cf9..4cbb0f891e 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -50,18 +50,17 @@ struct draw_stage; #define SP_NEW_SETUP 0x2 #define SP_NEW_FS 0x4 #define SP_NEW_BLEND 0x8 -#define SP_NEW_CLIP 0x10 -#define SP_NEW_SCISSOR 0x20 -#define SP_NEW_STIPPLE 0x40 -#define SP_NEW_FRAMEBUFFER 0x80 -#define SP_NEW_ALPHA_TEST 0x100 -#define SP_NEW_DEPTH_TEST 0x200 -#define SP_NEW_SAMPLER 0x400 -#define SP_NEW_TEXTURE 0x800 -#define SP_NEW_STENCIL 0x1000 -#define SP_NEW_VERTEX 0x2000 -#define SP_NEW_VS 0x4000 -#define SP_NEW_CONSTANTS 0x8000 +#define SP_NEW_CLIP 0x10 +#define SP_NEW_SCISSOR 0x20 +#define SP_NEW_STIPPLE 0x40 +#define SP_NEW_FRAMEBUFFER 0x80 +#define SP_NEW_ALPHA_TEST 0x100 +#define SP_NEW_DEPTH_STENCIL 0x200 +#define SP_NEW_SAMPLER 0x400 +#define SP_NEW_TEXTURE 0x800 +#define SP_NEW_VERTEX 0x1000 +#define SP_NEW_VS 0x2000 +#define SP_NEW_CONSTANTS 0x4000 struct softpipe_context { @@ -73,13 +72,13 @@ struct softpipe_context { */ const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct pipe_depth_stencil_state *depth_stencil; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; - struct pipe_depth_state depth_test; struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; struct pipe_shader_state fs; @@ -87,7 +86,6 @@ struct softpipe_context { struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_setup_state setup; - struct pipe_stencil_state stencil; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 2fcbea1f22..1f45776d47 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -31,7 +31,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.bufloop; } - if (sp->depth_test.occlusion_count) { + if (sp->depth_stencil->depth.occlusion_count) { sp->quad.occlusion->next = sp->quad.first; sp->quad.first = sp->quad.occlusion; } @@ -43,12 +43,12 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.coverage; } - if ( sp->stencil.front_enabled - || sp->stencil.front_enabled) { + if ( sp->depth_stencil->stencil.front_enabled + || sp->depth_stencil->stencil.back_enabled) { sp->quad.stencil_test->next = sp->quad.first; sp->quad.first = sp->quad.stencil_test; } - else if (sp->depth_test.enabled && + else if (sp->depth_stencil->depth.enabled && sp->framebuffer.zbuf) { sp->quad.depth_test->next = sp->quad.first; sp->quad.first = sp->quad.depth_test; diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 5d46e70393..ff1d84a02d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -77,7 +77,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) /* get zquad from zbuffer */ sps->read_quad_z(sps, quad->x0, quad->y0, bzzzz); - switch (softpipe->depth_test.func) { + switch (softpipe->depth_stencil->depth.func) { case PIPE_FUNC_NEVER: /* zmask = 0 */ break; @@ -129,7 +129,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) quad->mask &= zmask; - if (softpipe->depth_test.writemask) { + if (softpipe->depth_stencil->depth.writemask) { /* This is also efficient with sse / spe instructions: */ diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index cb0b6d8a77..46ad08aaa1 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -179,7 +179,7 @@ static void shade_begin(struct quad_stage *qs) unsigned i, entry; for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - qss->samplers[i].state = &softpipe->sampler[i]; + qss->samplers[i].state = softpipe->sampler[i]; qss->samplers[i].texture = softpipe->texture[i]; qss->samplers[i].get_samples = sp_get_samples; qss->samplers[i].pipe = &softpipe->pipe; diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 47b3b4f089..56cc6907b2 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -207,23 +207,23 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) /* choose front or back face function, operator, etc */ /* XXX we could do these initializations once per primitive */ - if (softpipe->stencil.back_enabled && quad->facing) { - func = softpipe->stencil.back_func; - failOp = softpipe->stencil.back_fail_op; - zFailOp = softpipe->stencil.back_zfail_op; - zPassOp = softpipe->stencil.back_zpass_op; - ref = softpipe->stencil.ref_value[1]; - wrtMask = softpipe->stencil.write_mask[1]; - valMask = softpipe->stencil.value_mask[1]; + if (softpipe->depth_stencil->stencil.back_enabled && quad->facing) { + func = softpipe->depth_stencil->stencil.back_func; + failOp = softpipe->depth_stencil->stencil.back_fail_op; + zFailOp = softpipe->depth_stencil->stencil.back_zfail_op; + zPassOp = softpipe->depth_stencil->stencil.back_zpass_op; + ref = softpipe->depth_stencil->stencil.ref_value[1]; + wrtMask = softpipe->depth_stencil->stencil.write_mask[1]; + valMask = softpipe->depth_stencil->stencil.value_mask[1]; } else { - func = softpipe->stencil.front_func; - failOp = softpipe->stencil.front_fail_op; - zFailOp = softpipe->stencil.front_zfail_op; - zPassOp = softpipe->stencil.front_zpass_op; - ref = softpipe->stencil.ref_value[0]; - wrtMask = softpipe->stencil.write_mask[0]; - valMask = softpipe->stencil.value_mask[0]; + func = softpipe->depth_stencil->stencil.front_func; + failOp = softpipe->depth_stencil->stencil.front_fail_op; + zFailOp = softpipe->depth_stencil->stencil.front_zfail_op; + zPassOp = softpipe->depth_stencil->stencil.front_zpass_op; + ref = softpipe->depth_stencil->stencil.ref_value[0]; + wrtMask = softpipe->depth_stencil->stencil.write_mask[0]; + valMask = softpipe->depth_stencil->stencil.value_mask[0]; } assert(s_surf); /* shouldn't get here if there's no stencil buffer */ @@ -244,7 +244,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) if (quad->mask) { /* now the pixels that passed the stencil test are depth tested */ - if (softpipe->depth_test.enabled) { + if (softpipe->depth_stencil->depth.enabled) { const unsigned origMask = quad->mask; sp_depth_test_quad(qs, quad); /* quad->mask is updated */ diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index c8c93709db..caec3b4519 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -50,6 +50,15 @@ void softpipe_bind_sampler_state(struct pipe_context *, void softpipe_delete_sampler_state(struct pipe_context *, const struct pipe_sampler_state *); + +const struct pipe_depth_stencil_state * +softpipe_create_depth_stencil_state(struct pipe_context *, + const struct pipe_depth_stencil_state *); +void softpipe_bind_depth_stencil_state(struct pipe_context *, + const struct pipe_depth_stencil_state *); +void softpipe_delete_depth_stencil_state(struct pipe_context *, + const struct pipe_depth_stencil_state *); + void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); @@ -69,9 +78,6 @@ void softpipe_set_constant_buffer(struct pipe_context *, uint shader, uint index, const struct pipe_constant_buffer *buf); -void softpipe_set_depth_test_state( struct pipe_context *, - const struct pipe_depth_state * ); - void softpipe_set_feedback_state( struct pipe_context *, const struct pipe_feedback_state * ); @@ -90,9 +96,6 @@ void softpipe_set_scissor_state( struct pipe_context *, void softpipe_set_setup_state( struct pipe_context *, const struct pipe_setup_state * ); -void softpipe_set_stencil_state( struct pipe_context *, - const struct pipe_stencil_state * ); - void softpipe_set_texture_state( struct pipe_context *, unsigned unit, struct pipe_mipmap_tree * ); diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 34da613f9d..83f456ded5 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -71,16 +71,6 @@ void softpipe_set_blend_color( struct pipe_context *pipe, /** XXX move someday? Or consolidate all these simple state setters * into one file. */ -void -softpipe_set_depth_test_state(struct pipe_context *pipe, - const struct pipe_depth_state *depth) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->depth_test = *depth; - - softpipe->dirty |= SP_NEW_DEPTH_TEST; -} void softpipe_set_alpha_test_state(struct pipe_context *pipe, @@ -93,14 +83,30 @@ softpipe_set_alpha_test_state(struct pipe_context *pipe, softpipe->dirty |= SP_NEW_ALPHA_TEST; } +const struct pipe_depth_stencil_state * +softpipe_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *depth_stencil) +{ + struct pipe_depth_stencil_state *new_ds = malloc(sizeof(struct pipe_depth_stencil_state)); + memcpy(new_ds, depth_stencil, sizeof(struct pipe_depth_stencil_state)); + + return new_ds; +} + void -softpipe_set_stencil_state(struct pipe_context *pipe, - const struct pipe_stencil_state *stencil) +softpipe_bind_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *depth_stencil) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->stencil = *stencil; + softpipe->depth_stencil = depth_stencil; - softpipe->dirty |= SP_NEW_STENCIL; + softpipe->dirty |= SP_NEW_DEPTH_STENCIL; } +void +softpipe_delete_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *depth) +{ + free((struct pipe_depth_stencil_state*)depth); +} diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index e08ed50a70..47743e185c 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -55,7 +55,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* Need Z if depth test is enabled or the fragment program uses the * fragment position (XYZW). */ - if (softpipe->depth_test.enabled || + if (softpipe->depth_stencil->depth.enabled || (inputsRead & (1 << TGSI_ATTRIB_POS))) softpipe->need_z = TRUE; else @@ -186,15 +186,14 @@ void softpipe_update_derived( struct softpipe_context *softpipe ) calculate_vertex_layout( softpipe ); if (softpipe->dirty & (SP_NEW_SCISSOR | - SP_NEW_STENCIL | + SP_NEW_DEPTH_STENCIL | SP_NEW_FRAMEBUFFER)) compute_cliprect(softpipe); if (softpipe->dirty & (SP_NEW_BLEND | - SP_NEW_DEPTH_TEST | + SP_NEW_DEPTH_STENCIL | SP_NEW_ALPHA_TEST | SP_NEW_FRAMEBUFFER | - SP_NEW_STENCIL | SP_NEW_SETUP | SP_NEW_FS)) sp_build_quad_pipeline(softpipe); diff --git a/src/mesa/sources b/src/mesa/sources index 90fa5c65bf..22b592df09 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -204,7 +204,6 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_sampler.c \ state_tracker/st_atom_scissor.c \ state_tracker/st_atom_setup.c \ - state_tracker/st_atom_stencil.c \ state_tracker/st_atom_stipple.c \ state_tracker/st_atom_texture.c \ state_tracker/st_atom_viewport.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 66ab5d7c3a..99d0bcb90b 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -46,7 +46,7 @@ static const struct st_tracked_state *atoms[] = { &st_update_framebuffer, &st_update_clear_color, - &st_update_depth, + &st_update_depth_stencil, &st_update_clip, &st_update_tnl, @@ -58,7 +58,6 @@ static const struct st_tracked_state *atoms[] = &st_update_viewport, &st_update_scissor, &st_update_blend, - &st_update_stencil, &st_update_sampler, &st_update_texture, &st_update_vs_constants, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 447430bfef..0e362b1fbf 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -47,7 +47,7 @@ void st_validate_state( struct st_context *st ); const struct st_tracked_state st_update_framebuffer; const struct st_tracked_state st_update_clip; const struct st_tracked_state st_update_clear_color; -const struct st_tracked_state st_update_depth; +const struct st_tracked_state st_update_depth_stencil; const struct st_tracked_state st_update_tnl; const struct st_tracked_state st_update_fs; const struct st_tracked_state st_update_vs; @@ -56,7 +56,6 @@ const struct st_tracked_state st_update_polygon_stipple; const struct st_tracked_state st_update_viewport; const struct st_tracked_state st_update_scissor; const struct st_tracked_state st_update_blend; -const struct st_tracked_state st_update_stencil; const struct st_tracked_state st_update_sampler; const struct st_tracked_state st_update_texture; const struct st_tracked_state st_update_fs_constants; diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index df05c79e36..406773213d 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -29,15 +29,67 @@ * Authors: * Keith Whitwell * Brian Paul + * Zack Rusin */ #include "st_context.h" +#include "st_cache.h" #include "st_atom.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +/** + * Convert GLenum stencil func tokens to pipe tokens. + */ +static GLuint +gl_stencil_func_to_sp(GLenum func) +{ + /* Same values, just biased */ + assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER); + assert(PIPE_FUNC_LESS == GL_LESS - GL_NEVER); + assert(PIPE_FUNC_EQUAL == GL_EQUAL - GL_NEVER); + assert(PIPE_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); + assert(PIPE_FUNC_GREATER == GL_GREATER - GL_NEVER); + assert(PIPE_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); + assert(PIPE_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); + assert(PIPE_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); + assert(func >= GL_NEVER); + assert(func <= GL_ALWAYS); + return func - GL_NEVER; +} + + +/** + * Convert GLenum stencil op tokens to pipe tokens. + */ +static GLuint +gl_stencil_op_to_sp(GLenum func) +{ + switch (func) { + case GL_KEEP: + return PIPE_STENCIL_OP_KEEP; + case GL_ZERO: + return PIPE_STENCIL_OP_ZERO; + case GL_REPLACE: + return PIPE_STENCIL_OP_REPLACE; + case GL_INCR: + return PIPE_STENCIL_OP_INCR; + case GL_DECR: + return PIPE_STENCIL_OP_DECR; + case GL_INCR_WRAP: + return PIPE_STENCIL_OP_INCR_WRAP; + case GL_DECR_WRAP: + return PIPE_STENCIL_OP_DECR_WRAP; + case GL_INVERT: + return PIPE_STENCIL_OP_INVERT; + default: + assert("invalid GL token in gl_stencil_op_to_sp()" == NULL); + return 0; + } +} + /** * Convert GLenum depth func tokens to pipe tokens. */ @@ -59,35 +111,59 @@ gl_depth_func_to_sp(GLenum func) } -static void -update_depth( struct st_context *st ) +static void +update_depth_stencil(struct st_context *st) { - struct pipe_depth_state depth; + struct pipe_depth_stencil_state depth_stencil; - memset(&depth, 0, sizeof(depth)); + memset(&depth_stencil, 0, sizeof(depth_stencil)); - depth.enabled = st->ctx->Depth.Test; - depth.writemask = st->ctx->Depth.Mask; - depth.func = gl_depth_func_to_sp(st->ctx->Depth.Func); - depth.clear = st->ctx->Depth.Clear; + depth_stencil.depth.enabled = st->ctx->Depth.Test; + depth_stencil.depth.writemask = st->ctx->Depth.Mask; + depth_stencil.depth.func = gl_depth_func_to_sp(st->ctx->Depth.Func); + depth_stencil.depth.clear = st->ctx->Depth.Clear; if (st->ctx->Query.CurrentOcclusionObject && st->ctx->Query.CurrentOcclusionObject->Active) - depth.occlusion_count = 1; + depth_stencil.depth.occlusion_count = 1; + + if (st->ctx->Stencil.Enabled) { + depth_stencil.stencil.front_enabled = 1; + depth_stencil.stencil.front_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[0]); + depth_stencil.stencil.front_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[0]); + depth_stencil.stencil.front_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[0]); + depth_stencil.stencil.front_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[0]); + depth_stencil.stencil.ref_value[0] = st->ctx->Stencil.Ref[0] & 0xff; + depth_stencil.stencil.value_mask[0] = st->ctx->Stencil.ValueMask[0] & 0xff; + depth_stencil.stencil.write_mask[0] = st->ctx->Stencil.WriteMask[0] & 0xff; + if (st->ctx->Stencil.TestTwoSide) { + depth_stencil.stencil.back_enabled = 1; + depth_stencil.stencil.back_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[1]); + depth_stencil.stencil.back_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[1]); + depth_stencil.stencil.back_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[1]); + depth_stencil.stencil.back_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[1]); + depth_stencil.stencil.ref_value[1] = st->ctx->Stencil.Ref[1] & 0xff; + depth_stencil.stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1] & 0xff; + depth_stencil.stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1] & 0xff; + } + depth_stencil.stencil.clear_value = st->ctx->Stencil.Clear & 0xff; + } - if (memcmp(&depth, &st->state.depth, sizeof(depth)) != 0) { + struct pipe_depth_stencil_state *cached_state = + st_cached_depth_stencil_state(st, &depth_stencil); + if (st->state.depth_stencil != cached_state) { /* state has changed */ - st->state.depth = depth; /* struct copy */ - st->pipe->set_depth_state(st->pipe, &depth); /* set new state */ + st->state.depth_stencil = cached_state; + st->pipe->bind_depth_stencil_state(st->pipe, cached_state); /* set new state */ } } -const struct st_tracked_state st_update_depth = { - .name = "st_update_depth", +const struct st_tracked_state st_update_depth_stencil = { + .name = "st_update_depth_stencil", .dirty = { - .mesa = (_NEW_DEPTH), + .mesa = (_NEW_DEPTH|_NEW_STENCIL), .st = 0, }, - .update = update_depth + .update = update_depth_stencil }; diff --git a/src/mesa/state_tracker/st_atom_stencil.c b/src/mesa/state_tracker/st_atom_stencil.c deleted file mode 100644 index b8aec0b3b6..0000000000 --- a/src/mesa/state_tracker/st_atom_stencil.c +++ /dev/null @@ -1,141 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#include "st_context.h" -#include "st_atom.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" - - -/** - * Convert GLenum stencil func tokens to pipe tokens. - */ -static GLuint -gl_stencil_func_to_sp(GLenum func) -{ - /* Same values, just biased */ - assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER); - assert(PIPE_FUNC_LESS == GL_LESS - GL_NEVER); - assert(PIPE_FUNC_EQUAL == GL_EQUAL - GL_NEVER); - assert(PIPE_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); - assert(PIPE_FUNC_GREATER == GL_GREATER - GL_NEVER); - assert(PIPE_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); - assert(PIPE_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); - assert(PIPE_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); - assert(func >= GL_NEVER); - assert(func <= GL_ALWAYS); - return func - GL_NEVER; -} - - -/** - * Convert GLenum stencil op tokens to pipe tokens. - */ -static GLuint -gl_stencil_op_to_sp(GLenum func) -{ - switch (func) { - case GL_KEEP: - return PIPE_STENCIL_OP_KEEP; - case GL_ZERO: - return PIPE_STENCIL_OP_ZERO; - case GL_REPLACE: - return PIPE_STENCIL_OP_REPLACE; - case GL_INCR: - return PIPE_STENCIL_OP_INCR; - case GL_DECR: - return PIPE_STENCIL_OP_DECR; - case GL_INCR_WRAP: - return PIPE_STENCIL_OP_INCR_WRAP; - case GL_DECR_WRAP: - return PIPE_STENCIL_OP_DECR_WRAP; - case GL_INVERT: - return PIPE_STENCIL_OP_INVERT; - default: - assert("invalid GL token in gl_stencil_op_to_sp()" == NULL); - return 0; - } -} - - -static void -update_stencil( struct st_context *st ) -{ - struct pipe_stencil_state stencil; - - memset(&stencil, 0, sizeof(stencil)); - - if (st->ctx->Stencil.Enabled) { - stencil.front_enabled = 1; - stencil.front_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[0]); - stencil.front_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[0]); - stencil.front_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[0]); - stencil.front_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[0]); - stencil.ref_value[0] = st->ctx->Stencil.Ref[0] & 0xff; - stencil.value_mask[0] = st->ctx->Stencil.ValueMask[0] & 0xff; - stencil.write_mask[0] = st->ctx->Stencil.WriteMask[0] & 0xff; - if (st->ctx->Stencil.TestTwoSide) { - stencil.back_enabled = 1; - stencil.back_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[1]); - stencil.back_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[1]); - stencil.back_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[1]); - stencil.back_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[1]); - stencil.ref_value[1] = st->ctx->Stencil.Ref[1] & 0xff; - stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1] & 0xff; - stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1] & 0xff; - } - stencil.clear_value = st->ctx->Stencil.Clear & 0xff; - } - - if (memcmp(&stencil, &st->state.stencil, sizeof(stencil)) != 0) { - /* state has changed */ - st->state.stencil = stencil; /* struct copy */ - st->pipe->set_stencil_state(st->pipe, &stencil); /* set new state */ - } -} - - -const struct st_tracked_state st_update_stencil = { - .name = "st_update_stencil", - .dirty = { - .mesa = (_NEW_STENCIL), - .st = 0, - }, - .update = update_stencil -}; - - - - - diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 99fb45f00a..64c03be99d 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -76,3 +76,20 @@ struct pipe_sampler_state * st_cached_sampler_state( } return (struct pipe_sampler_state*)(cso_hash_iter_data(iter)); } + +struct pipe_depth_stencil_state * st_cached_depth_stencil_state( + struct st_context *st, + const struct pipe_depth_stencil_state *depth_stencil) +{ + unsigned hash_key = cso_construct_key((void*)depth_stencil, sizeof(struct pipe_depth_stencil_state)); + struct cso_hash_iter iter = cso_find_state_template(st->cache, + hash_key, CSO_DEPTH_STENCIL, + (void*)depth_stencil); + if (cso_hash_iter_is_null(iter)) { + const struct pipe_depth_stencil_state *created_state = st->pipe->create_depth_stencil_state( + st->pipe, depth_stencil); + iter = cso_insert_state(st->cache, hash_key, CSO_DEPTH_STENCIL, + (void*)created_state); + } + return (struct pipe_depth_stencil_state*)(cso_hash_iter_data(iter)); +} diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index 927585141c..78cb2e774e 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -45,5 +45,8 @@ struct pipe_sampler_state * st_cached_sampler_state( struct st_context *st, const struct pipe_sampler_state *sampler); +struct pipe_depth_stencil_state *st_cached_depth_stencil_state( + struct st_context *st, + const struct pipe_depth_stencil_state *sampler); #endif diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 55e03f644e..e9aabd15b5 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -283,6 +283,7 @@ clear_with_quad(GLcontext *ctx, /* blend state: RGBA masking */ { struct pipe_blend_state blend; + const struct pipe_blend_state *state; memset(&blend, 0, sizeof(blend)); if (color) { if (ctx->Color.ColorMask[0]) @@ -296,20 +297,34 @@ clear_with_quad(GLcontext *ctx, if (st->ctx->Color.DitherFlag) blend.dither = 1; } - const struct pipe_blend_state *state = st_cached_blend_state(st, &blend); + state = st_cached_blend_state(st, &blend); pipe->bind_blend_state(pipe, state); } - /* depth state: always pass */ + /* depth_stencil state: always pass/set to ref value */ { - struct pipe_depth_state depth_test; - memset(&depth_test, 0, sizeof(depth_test)); + struct pipe_depth_stencil_state depth_stencil; + struct pipe_depth_stencil_state *cached; + memset(&depth_stencil, 0, sizeof(depth_stencil)); if (depth) { - depth_test.enabled = 1; - depth_test.writemask = 1; - depth_test.func = PIPE_FUNC_ALWAYS; + depth_stencil.depth.enabled = 1; + depth_stencil.depth.writemask = 1; + depth_stencil.depth.func = PIPE_FUNC_ALWAYS; } - pipe->set_depth_state(pipe, &depth_test); + + if (stencil) { + depth_stencil.stencil.front_enabled = 1; + depth_stencil.stencil.front_func = PIPE_FUNC_ALWAYS; + depth_stencil.stencil.front_fail_op = PIPE_STENCIL_OP_REPLACE; + depth_stencil.stencil.front_zpass_op = PIPE_STENCIL_OP_REPLACE; + depth_stencil.stencil.front_zfail_op = PIPE_STENCIL_OP_REPLACE; + depth_stencil.stencil.ref_value[0] = ctx->Stencil.Clear; + depth_stencil.stencil.value_mask[0] = 0xff; + depth_stencil.stencil.write_mask[0] = ctx->Stencil.WriteMask[0] & 0xff; + } + cached = + st_cached_depth_stencil_state(ctx->st, &depth_stencil); + pipe->bind_depth_stencil_state(pipe, cached); } /* setup state: nothing */ @@ -326,23 +341,6 @@ clear_with_quad(GLcontext *ctx, pipe->set_setup_state(pipe, &setup); } - /* stencil state: always set to ref value */ - { - struct pipe_stencil_state stencil_test; - memset(&stencil_test, 0, sizeof(stencil_test)); - if (stencil) { - stencil_test.front_enabled = 1; - stencil_test.front_func = PIPE_FUNC_ALWAYS; - stencil_test.front_fail_op = PIPE_STENCIL_OP_REPLACE; - stencil_test.front_zpass_op = PIPE_STENCIL_OP_REPLACE; - stencil_test.front_zfail_op = PIPE_STENCIL_OP_REPLACE; - stencil_test.ref_value[0] = ctx->Stencil.Clear; - stencil_test.value_mask[0] = 0xff; - stencil_test.write_mask[0] = ctx->Stencil.WriteMask[0] & 0xff; - } - pipe->set_stencil_state(pipe, &stencil_test); - } - /* fragment shader state: color pass-through program */ { static struct st_fragment_program *stfp = NULL; @@ -393,11 +391,10 @@ clear_with_quad(GLcontext *ctx, /* Restore pipe state */ pipe->set_alpha_test_state(pipe, &st->state.alpha_test); pipe->bind_blend_state(pipe, st->state.blend); - pipe->set_depth_state(pipe, &st->state.depth); + pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil); pipe->set_fs_state(pipe, &st->state.fs); pipe->set_vs_state(pipe, &st->state.vs); pipe->set_setup_state(pipe, &st->state.setup); - pipe->set_stencil_state(pipe, &st->state.stencil); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); /* OR: st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 31a37e5cd4..a0012e3a8c 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -478,7 +478,7 @@ any_fragment_ops(const struct st_context *st) if (st->state.alpha_test.enabled || st->state.blend->blend_enable || st->state.blend->logicop_enable || - st->state.depth.enabled) + st->state.depth_stencil->depth.enabled) /* XXX more checks */ return GL_TRUE; else diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 13526ff9e7..7c887d0b55 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -76,13 +76,13 @@ struct st_context struct { const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct pipe_depth_stencil_state *depth_stencil; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; - struct pipe_depth_state depth; struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; @@ -91,7 +91,6 @@ struct st_context struct pipe_setup_state setup; struct pipe_shader_state fs; struct pipe_shader_state vs; - struct pipe_stencil_state stencil; struct pipe_viewport_state viewport; } state; -- cgit v1.2.3 From 294401814d1d89cc731de1c22c25333aa5d59374 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 17 Sep 2007 12:59:50 -0400 Subject: converting the setup state to immutable object and renaming it to rasterizer state --- src/mesa/cso_cache/cso_cache.c | 6 ++ src/mesa/cso_cache/cso_cache.h | 4 +- src/mesa/pipe/draw/draw_context.c | 16 ++--- src/mesa/pipe/draw/draw_context.h | 2 +- src/mesa/pipe/draw/draw_cull.c | 2 +- src/mesa/pipe/draw/draw_offset.c | 4 +- src/mesa/pipe/draw/draw_private.h | 2 +- src/mesa/pipe/draw/draw_twoside.c | 2 +- src/mesa/pipe/draw/draw_unfilled.c | 4 +- src/mesa/pipe/failover/fo_context.h | 8 +-- src/mesa/pipe/failover/fo_state.c | 14 ++-- src/mesa/pipe/failover/fo_state_emit.c | 4 +- src/mesa/pipe/i915simple/i915_context.h | 4 +- src/mesa/pipe/i915simple/i915_state.c | 30 ++++++-- src/mesa/pipe/i915simple/i915_state_derived.c | 6 +- src/mesa/pipe/i915simple/i915_state_dynamic.c | 16 ++--- src/mesa/pipe/i915simple/i915_state_immediate.c | 30 ++++---- src/mesa/pipe/p_context.h | 11 ++- src/mesa/pipe/p_state.h | 4 +- src/mesa/pipe/softpipe/sp_context.c | 4 +- src/mesa/pipe/softpipe/sp_context.h | 4 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 18 ++--- src/mesa/pipe/softpipe/sp_quad.c | 8 +-- src/mesa/pipe/softpipe/sp_quad_coverage.c | 6 +- src/mesa/pipe/softpipe/sp_state.h | 12 ++-- src/mesa/pipe/softpipe/sp_state_derived.c | 10 +-- src/mesa/pipe/softpipe/sp_state_setup.c | 26 +++++-- src/mesa/state_tracker/st_atom.c | 2 +- src/mesa/state_tracker/st_atom.h | 2 +- src/mesa/state_tracker/st_atom_setup.c | 95 +++++++++++++------------ src/mesa/state_tracker/st_cache.c | 18 +++++ src/mesa/state_tracker/st_cache.h | 6 +- src/mesa/state_tracker/st_cb_clear.c | 12 ++-- src/mesa/state_tracker/st_cb_drawpixels.c | 8 ++- src/mesa/state_tracker/st_context.h | 2 +- src/mesa/state_tracker/st_draw.c | 2 +- 36 files changed, 245 insertions(+), 159 deletions(-) (limited to 'src') diff --git a/src/mesa/cso_cache/cso_cache.c b/src/mesa/cso_cache/cso_cache.c index a4730394f8..4aaadf00e6 100644 --- a/src/mesa/cso_cache/cso_cache.c +++ b/src/mesa/cso_cache/cso_cache.c @@ -78,6 +78,8 @@ static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_ hash = sc->sampler_hash; case CSO_DEPTH_STENCIL: hash = sc->depth_stencil_hash; + case CSO_RASTERIZER: + hash = sc->rasterizer_hash; } return hash; @@ -92,6 +94,8 @@ static int _cso_size_for_type(enum cso_cache_type type) return sizeof(struct pipe_sampler_state); case CSO_DEPTH_STENCIL: return sizeof(struct pipe_depth_stencil_state); + case CSO_RASTERIZER: + return sizeof(struct pipe_rasterizer_state); } return 0; } @@ -143,6 +147,7 @@ struct cso_cache *cso_cache_create(void) sc->blend_hash = cso_hash_create(); sc->sampler_hash = cso_hash_create(); sc->depth_stencil_hash = cso_hash_create(); + sc->rasterizer_hash = cso_hash_create(); return sc; } @@ -153,6 +158,7 @@ void cso_cache_delete(struct cso_cache *sc) cso_hash_delete(sc->blend_hash); cso_hash_delete(sc->sampler_hash); cso_hash_delete(sc->depth_stencil_hash); + cso_hash_delete(sc->rasterizer_hash); free(sc); } diff --git a/src/mesa/cso_cache/cso_cache.h b/src/mesa/cso_cache/cso_cache.h index c340cf7f67..23be9cd713 100644 --- a/src/mesa/cso_cache/cso_cache.h +++ b/src/mesa/cso_cache/cso_cache.h @@ -43,12 +43,14 @@ struct cso_cache { struct cso_hash *blend_hash; struct cso_hash *sampler_hash; struct cso_hash *depth_stencil_hash; + struct cso_hash *rasterizer_hash; }; enum cso_cache_type { CSO_BLEND, CSO_SAMPLER, - CSO_DEPTH_STENCIL + CSO_DEPTH_STENCIL, + CSO_RASTERIZER }; unsigned cso_construct_key(void *item, int item_size); diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 4498293e92..f3236ad59e 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -98,19 +98,19 @@ static void validate_pipeline( struct draw_context *draw ) * shorter pipelines for lines & points. */ - if (draw->setup.fill_cw != PIPE_POLYGON_MODE_FILL || - draw->setup.fill_ccw != PIPE_POLYGON_MODE_FILL) { + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { draw->pipeline.unfilled->next = next; next = draw->pipeline.unfilled; } - if (draw->setup.offset_cw || - draw->setup.offset_ccw) { + if (draw->rasterizer->offset_cw || + draw->rasterizer->offset_ccw) { draw->pipeline.offset->next = next; next = draw->pipeline.offset; } - if (draw->setup.light_twoside) { + if (draw->rasterizer->light_twoside) { draw->pipeline.twoside->next = next; next = draw->pipeline.twoside; } @@ -134,7 +134,7 @@ static void validate_pipeline( struct draw_context *draw ) * this for clipped primitives, ie it is a part of the clip * routine. */ - if (draw->setup.flatshade) { + if (draw->rasterizer->flatshade) { draw->pipeline.flatshade->next = next; next = draw->pipeline.flatshade; } @@ -161,9 +161,9 @@ void draw_set_feedback_state( struct draw_context *draw, * This causes the drawing pipeline to be rebuilt. */ void draw_set_setup_state( struct draw_context *draw, - const struct pipe_setup_state *setup ) + const struct pipe_rasterizer_state *raster ) { - draw->setup = *setup; /* struct copy */ + draw->rasterizer = raster; validate_pipeline( draw ); } diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 2babc02f45..2714252fc5 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -87,7 +87,7 @@ void draw_set_feedback_state( struct draw_context *draw, const struct pipe_feedback_state * ); void draw_set_setup_state( struct draw_context *draw, - const struct pipe_setup_state *setup ); + const struct pipe_rasterizer_state *raster ); void draw_set_setup_stage( struct draw_context *draw, struct draw_stage *stage ); diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index f3d56ad719..f898834ba5 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -54,7 +54,7 @@ static void cull_begin( struct draw_stage *stage ) { struct cull_stage *cull = cull_stage(stage); - cull->winding = stage->draw->setup.cull_mode; + cull->winding = stage->draw->rasterizer->cull_mode; stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 4f653e8c54..6acc7cbcd2 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -57,8 +57,8 @@ static void offset_begin( struct draw_stage *stage ) struct offset_stage *offset = offset_stage(stage); float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ - offset->units = stage->draw->setup.offset_units * mrd; - offset->scale = stage->draw->setup.offset_scale; + offset->units = stage->draw->rasterizer->offset_units * mrd; + offset->scale = stage->draw->rasterizer->offset_scale; stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 8bcc3717c4..fb0aaff40d 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -137,7 +137,7 @@ struct draw_context } pipeline; /* pipe state that we need: */ - struct pipe_setup_state setup; + const struct pipe_rasterizer_state *rasterizer; struct pipe_feedback_state feedback; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 98eb088035..3eb8cce637 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -55,7 +55,7 @@ static void twoside_begin( struct draw_stage *stage ) * if the triangle is back-facing (negative). * sign = -1 for CCW, +1 for CW */ - twoside->sign = (stage->draw->setup.front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; + twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index b0d6f3d065..2d374329d8 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -59,8 +59,8 @@ static void unfilled_begin( struct draw_stage *stage ) { struct unfilled_stage *unfilled = unfilled_stage(stage); - unfilled->mode[0] = stage->draw->setup.fill_ccw; /* front */ - unfilled->mode[1] = stage->draw->setup.fill_cw; /* back */ + unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */ + unfilled->mode[1] = stage->draw->rasterizer->fill_cw; /* back */ stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index 63ec7239ab..b05ceb88ad 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -37,7 +37,7 @@ #define FO_NEW_VIEWPORT 0x1 -#define FO_NEW_SETUP 0x2 +#define FO_NEW_RASTERIZER 0x2 #define FO_NEW_FRAGMENT_SHADER 0x4 #define FO_NEW_BLEND 0x8 #define FO_NEW_CLIP 0x10 @@ -66,9 +66,10 @@ struct failover_context { /* The most recent drawing state as set by the driver: */ - const struct pipe_blend_state *blend; - const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct pipe_blend_state *blend; + const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; + const struct pipe_rasterizer_state *rasterizer; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -79,7 +80,6 @@ struct failover_context { struct pipe_shader_state vertex_shader; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_setup_state setup; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 43b9757b31..8e2b649590 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -160,15 +160,15 @@ failover_set_polygon_stipple( struct pipe_context *pipe, -static void -failover_set_setup_state( struct pipe_context *pipe, - const struct pipe_setup_state *setup ) +static void +failover_bind_rasterizer_state( struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup ) { struct failover_context *failover = failover_context(pipe); - failover->setup = *setup; - failover->dirty |= FO_NEW_SETUP; - failover->hw->set_setup_state( failover->hw, setup ); + failover->rasterizer = setup; + failover->dirty |= FO_NEW_RASTERIZER; + failover->hw->bind_rasterizer_state( failover->hw, setup ); } @@ -257,6 +257,7 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.bind_blend_state = failover_bind_blend_state; failover->pipe.bind_sampler_state = failover_bind_sampler_state; failover->pipe.bind_depth_stencil_state = failover_bind_depth_stencil_state; + failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state; failover->pipe.set_alpha_test_state = failover_set_alpha_test_state; failover->pipe.set_blend_color = failover_set_blend_color; @@ -267,7 +268,6 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.set_vs_state = failover_set_vs_state; failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; failover->pipe.set_scissor_state = failover_set_scissor_state; - failover->pipe.set_setup_state = failover_set_setup_state; failover->pipe.set_texture_state = failover_set_texture_state; failover->pipe.set_viewport_state = failover_set_viewport_state; failover->pipe.set_vertex_buffer = failover_set_vertex_buffer; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 3a1865d766..1c9573a7b0 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -85,8 +85,8 @@ failover_state_emit( struct failover_context *failover ) if (failover->dirty & FO_NEW_STIPPLE) failover->sw->set_polygon_stipple( failover->sw, &failover->poly_stipple ); - if (failover->dirty & FO_NEW_SETUP) - failover->sw->set_setup_state( failover->sw, &failover->setup ); + if (failover->dirty & FO_NEW_RASTERIZER) + failover->sw->bind_rasterizer_state( failover->sw, failover->rasterizer ); if (failover->dirty & FO_NEW_SCISSOR) failover->sw->set_scissor_state( failover->sw, &failover->scissor ); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 518f780449..3fab821fde 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -126,6 +126,7 @@ struct i915_context const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; + const struct pipe_rasterizer_state *rasterizer; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -136,7 +137,6 @@ struct i915_context struct pipe_shader_state fs; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_setup_state setup; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; @@ -159,7 +159,7 @@ struct i915_context /* A flag for each state_tracker state object: */ #define I915_NEW_VIEWPORT 0x1 -#define I915_NEW_SETUP 0x2 +#define I915_NEW_RASTERIZER 0x2 #define I915_NEW_FS 0x4 #define I915_NEW_BLEND 0x8 #define I915_NEW_CLIP 0x10 diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 5ac2e27d1a..00764902bc 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -288,19 +288,36 @@ static void i915_set_viewport_state( struct pipe_context *pipe, } -static void i915_set_setup_state( struct pipe_context *pipe, - const struct pipe_setup_state *setup ) + +static const struct pipe_rasterizer_state * +i915_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup) +{ + struct pipe_rasterizer_state *raster = + malloc(sizeof(struct pipe_rasterizer_state)); + memcpy(raster, setup, sizeof(struct pipe_rasterizer_state)); + + return raster; +} + +static void i915_bind_rasterizer_state( struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup ) { struct i915_context *i915 = i915_context(pipe); - i915->setup = *setup; + i915->rasterizer = setup; /* pass-through to draw module */ draw_set_setup_state(i915->draw, setup); - i915->dirty |= I915_NEW_SETUP; + i915->dirty |= I915_NEW_RASTERIZER; } +static void i915_delete_rasterizer_state( struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup ) +{ + free((struct pipe_rasterizer_state*)setup); +} static void i915_set_vertex_buffer( struct pipe_context *pipe, unsigned index, @@ -338,6 +355,10 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.bind_depth_stencil_state = i915_bind_depth_stencil_state; i915->pipe.delete_depth_stencil_state = i915_delete_depth_stencil_state; + i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; + i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; + i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state; + i915->pipe.set_alpha_test_state = i915_set_alpha_test_state; i915->pipe.set_blend_color = i915_set_blend_color; i915->pipe.set_clip_state = i915_set_clip_state; @@ -348,7 +369,6 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_vs_state = i915_set_vs_state; i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; i915->pipe.set_scissor_state = i915_set_scissor_state; - i915->pipe.set_setup_state = i915_set_setup_state; i915->pipe.set_texture_state = i915_set_texture_state; i915->pipe.set_viewport_state = i915_set_viewport_state; i915->pipe.set_vertex_buffer = i915_set_vertex_buffer; diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 792bb93b17..504bc10a9e 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -44,7 +44,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) { const uint inputsRead = i915->fs.inputs_read; const interp_mode colorInterp - = i915->setup.flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + = i915->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &i915->current.vertex_info; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; boolean needW = 0; @@ -103,7 +103,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) * lighting. Edgeflag is dealt with specially by setting bits in * the vertex header. */ - if (i915->setup.light_twoside) { + if (i915->rasterizer->light_twoside) { if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { back0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT, colorInterp); @@ -142,7 +142,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ void i915_update_derived( struct i915_context *i915 ) { - if (i915->dirty & (I915_NEW_SETUP | I915_NEW_FS)) + if (i915->dirty & (I915_NEW_RASTERIZER | I915_NEW_FS)) calculate_vertex_layout( i915 ); if (i915->dirty & (I915_NEW_SAMPLER | I915_NEW_TEXTURE)) diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index 9140eee7c2..a9791962e2 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -261,10 +261,10 @@ static void upload_DEPTHSCALE( struct i915_context *i915 ) memset( ds, 0, sizeof(ds) ); - /* I915_NEW_SETUP + /* I915_NEW_RASTERIZER */ ds[0].u = _3DSTATE_DEPTH_OFFSET_SCALE; - ds[1].f = i915->setup.offset_scale; + ds[1].f = i915->rasterizer->offset_scale; set_dynamic_indirect( i915, I915_DYNAMIC_DEPTHSCALE_0, @@ -273,7 +273,7 @@ static void upload_DEPTHSCALE( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_DEPTHSCALE = { - .dirty = I915_NEW_SETUP, + .dirty = I915_NEW_RASTERIZER, .update = upload_DEPTHSCALE }; @@ -298,9 +298,9 @@ static void upload_STIPPLE( struct i915_context *i915 ) st[0] = _3DSTATE_STIPPLE; st[1] = 0; - /* I915_NEW_SETUP + /* I915_NEW_RASTERIZER */ - if (i915->setup.poly_stipple_enable) { + if (i915->rasterizer->poly_stipple_enable) { st[1] |= ST1_ENABLE; } @@ -333,7 +333,7 @@ static void upload_STIPPLE( struct i915_context *i915 ) const struct i915_tracked_state i915_upload_STIPPLE = { - .dirty = I915_NEW_SETUP | I915_NEW_STIPPLE, + .dirty = I915_NEW_RASTERIZER | I915_NEW_STIPPLE, .update = upload_STIPPLE }; @@ -346,7 +346,7 @@ static void upload_SCISSOR_ENABLE( struct i915_context *i915 ) { unsigned sc[1]; - if (i915->setup.scissor) + if (i915->rasterizer->scissor) sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT; else sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT; @@ -358,7 +358,7 @@ static void upload_SCISSOR_ENABLE( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_SCISSOR_ENABLE = { - .dirty = I915_NEW_SETUP, + .dirty = I915_NEW_RASTERIZER, .update = upload_SCISSOR_ENABLE }; diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 484913d308..73508f557f 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -62,8 +62,8 @@ static void upload_S2S4(struct i915_context *i915) assert(LIS4); /* should never be zero? */ } - /* I915_NEW_SETUP */ - switch (i915->setup.cull_mode) { + /* I915_NEW_RASTERIZER */ + switch (i915->rasterizer->cull_mode) { case PIPE_WINDING_NONE: LIS4 |= S4_CULLMODE_NONE; break; @@ -78,25 +78,25 @@ static void upload_S2S4(struct i915_context *i915) break; } - /* I915_NEW_SETUP */ + /* I915_NEW_RASTERIZER */ { - int line_width = CLAMP((int)(i915->setup.line_width * 2), 1, 0xf); + int line_width = CLAMP((int)(i915->rasterizer->line_width * 2), 1, 0xf); LIS4 |= line_width << S4_LINE_WIDTH_SHIFT; - if (i915->setup.line_smooth) + if (i915->rasterizer->line_smooth) LIS4 |= S4_LINE_ANTIALIAS_ENABLE; } - /* I915_NEW_SETUP */ + /* I915_NEW_RASTERIZER */ { - int point_size = CLAMP((int) i915->setup.point_size, 1, 0xff); + int point_size = CLAMP((int) i915->rasterizer->point_size, 1, 0xff); LIS4 |= point_size << S4_POINT_WIDTH_SHIFT; } - /* I915_NEW_SETUP */ - if (i915->setup.flatshade) { + /* I915_NEW_RASTERIZER */ + if (i915->rasterizer->flatshade) { LIS4 |= (S4_FLATSHADE_ALPHA | S4_FLATSHADE_COLOR | S4_FLATSHADE_SPECULAR); @@ -114,7 +114,7 @@ static void upload_S2S4(struct i915_context *i915) const struct i915_tracked_state i915_upload_S2S4 = { - .dirty = I915_NEW_SETUP | I915_NEW_VERTEX_FORMAT, + .dirty = I915_NEW_RASTERIZER | I915_NEW_VERTEX_FORMAT, .update = upload_S2S4 }; @@ -165,7 +165,7 @@ static void upload_S5( struct i915_context *i915 ) #if 0 - /* I915_NEW_SETUP */ + /* I915_NEW_RASTERIZER */ if (i915->state.Polygon->OffsetFill) { LIS5 |= S5_GLOBAL_DEPTH_OFFSET_ENABLE; } @@ -179,7 +179,7 @@ static void upload_S5( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_S5 = { - .dirty = (I915_NEW_DEPTH_STENCIL | I915_NEW_BLEND | I915_NEW_SETUP), + .dirty = (I915_NEW_DEPTH_STENCIL | I915_NEW_BLEND | I915_NEW_RASTERIZER), .update = upload_S5 }; @@ -247,9 +247,9 @@ static void upload_S7( struct i915_context *i915 ) { float LIS7; - /* I915_NEW_SETUP + /* I915_NEW_RASTERIZER */ - LIS7 = i915->setup.offset_units; /* probably incorrect */ + LIS7 = i915->rasterizer->offset_units; /* probably incorrect */ if (LIS7 != i915->current.immediate[I915_IMMEDIATE_S7]) { i915->current.immediate[I915_IMMEDIATE_S7] = LIS7; @@ -258,7 +258,7 @@ static void upload_S7( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_S7 = { - .dirty = I915_NEW_SETUP, + .dirty = I915_NEW_RASTERIZER, .update = upload_S7 }; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 488f002531..dda758fe6a 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -101,6 +101,14 @@ struct pipe_context { void (*delete_sampler_state)(struct pipe_context *, const struct pipe_sampler_state *); + const struct pipe_rasterizer_state *(*create_rasterizer_state)( + struct pipe_context *, + const struct pipe_rasterizer_state *); + void (*bind_rasterizer_state)(struct pipe_context *, + const struct pipe_rasterizer_state *); + void (*delete_rasterizer_state)(struct pipe_context *, + const struct pipe_rasterizer_state *); + const struct pipe_depth_stencil_state * (*create_depth_stencil_state)( struct pipe_context *, const struct pipe_depth_stencil_state *); @@ -140,9 +148,6 @@ struct pipe_context { void (*set_polygon_stipple)( struct pipe_context *, const struct pipe_poly_stipple * ); - void (*set_setup_state)( struct pipe_context *, - const struct pipe_setup_state * ); - void (*set_scissor_state)( struct pipe_context *, const struct pipe_scissor_state * ); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 30e559b594..048feede3b 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -65,9 +65,9 @@ struct pipe_buffer_handle; /** - * Primitive (point/line/tri) setup info + * Primitive (point/line/tri) rasterization info */ -struct pipe_setup_state +struct pipe_rasterizer_state { unsigned flatshade:1; unsigned light_twoside:1; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 9a8b55bb0e..7753ce40d7 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -259,6 +259,9 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.create_depth_stencil_state = softpipe_create_depth_stencil_state; softpipe->pipe.bind_depth_stencil_state = softpipe_bind_depth_stencil_state; softpipe->pipe.delete_depth_stencil_state = softpipe_delete_depth_stencil_state; + softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; + softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; + softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; @@ -271,7 +274,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_vs_state = softpipe_set_vs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; - softpipe->pipe.set_setup_state = softpipe_set_setup_state; softpipe->pipe.set_texture_state = softpipe_set_texture_state; softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 4cbb0f891e..f1bb3d39a6 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -47,7 +47,7 @@ struct draw_stage; #define SP_NEW_VIEWPORT 0x1 -#define SP_NEW_SETUP 0x2 +#define SP_NEW_RASTERIZER 0x2 #define SP_NEW_FS 0x4 #define SP_NEW_BLEND 0x8 #define SP_NEW_CLIP 0x10 @@ -73,6 +73,7 @@ struct softpipe_context { const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; + const struct pipe_rasterizer_state *rasterizer; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -85,7 +86,6 @@ struct softpipe_context { struct pipe_shader_state vs; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_setup_state setup; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 83d317c36f..c64a4e9708 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -351,7 +351,7 @@ static boolean setup_sort_vertices( struct setup_stage *setup, * - the GLSL gl_FrontFacing fragment attribute (bool) * - two-sided stencil test */ - setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->setup.front_winding == PIPE_WINDING_CW); + setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); return TRUE; } @@ -830,10 +830,10 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) const int errorDec = error - dx; for (i = 0; i < dx; i++) { - if (!sp->setup.line_stipple_enable || + if (!sp->rasterizer->line_stipple_enable || stipple_test(sp->line_stipple_counter, - sp->setup.line_stipple_pattern, - sp->setup.line_stipple_factor + 1)) { + sp->rasterizer->line_stipple_pattern, + sp->rasterizer->line_stipple_factor + 1)) { plot(setup, x0, y0); } @@ -857,10 +857,10 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) const int errorDec = error - dy; for (i = 0; i < dy; i++) { - if (!sp->setup.line_stipple_enable || + if (!sp->rasterizer->line_stipple_enable || stipple_test(sp->line_stipple_counter, - sp->setup.line_stipple_pattern, - sp->setup.line_stipple_factor + 1)) { + sp->rasterizer->line_stipple_pattern, + sp->rasterizer->line_stipple_factor + 1)) { plot(setup, x0, y0); } @@ -899,8 +899,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const int sizeAttr = setup->lookup[TGSI_ATTRIB_POINTSIZE]; const float halfSize = sizeAttr ? (0.5f * v0->data[sizeAttr][0]) - : (0.5f * setup->softpipe->setup.point_size); - const boolean round = setup->softpipe->setup.point_smooth; + : (0.5f * setup->softpipe->rasterizer->point_size); + const boolean round = setup->softpipe->rasterizer->point_smooth; const float x = v0->data[TGSI_ATTRIB_POS][0]; const float y = v0->data[TGSI_ATTRIB_POS][1]; unsigned slot, j; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 1f45776d47..fc4f8328cf 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -36,9 +36,9 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.occlusion; } - if (sp->setup.poly_smooth || - sp->setup.line_smooth || - sp->setup.point_smooth) { + if (sp->rasterizer->poly_smooth || + sp->rasterizer->line_smooth || + sp->rasterizer->point_smooth) { sp->quad.coverage->next = sp->quad.first; sp->quad.first = sp->quad.coverage; } @@ -65,7 +65,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.shade; } - if (sp->setup.poly_stipple_enable) { + if (sp->rasterizer->poly_stipple_enable) { sp->quad.polygon_stipple->next = sp->quad.first; sp->quad.first = sp->quad.polygon_stipple; } diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c index 8dfec59350..89f50bcca2 100644 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -47,9 +47,9 @@ coverage_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - if ((softpipe->setup.poly_smooth && quad->prim == PRIM_TRI) || - (softpipe->setup.line_smooth && quad->prim == PRIM_LINE) || - (softpipe->setup.point_smooth && quad->prim == PRIM_POINT)) { + if ((softpipe->rasterizer->poly_smooth && quad->prim == PRIM_TRI) || + (softpipe->rasterizer->line_smooth && quad->prim == PRIM_LINE) || + (softpipe->rasterizer->point_smooth && quad->prim == PRIM_POINT)) { unsigned j; for (j = 0; j < QUAD_SIZE; j++) { assert(quad->coverage[j] >= 0.0); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index caec3b4519..62bd26c4df 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -50,7 +50,6 @@ void softpipe_bind_sampler_state(struct pipe_context *, void softpipe_delete_sampler_state(struct pipe_context *, const struct pipe_sampler_state *); - const struct pipe_depth_stencil_state * softpipe_create_depth_stencil_state(struct pipe_context *, const struct pipe_depth_stencil_state *); @@ -59,6 +58,14 @@ void softpipe_bind_depth_stencil_state(struct pipe_context *, void softpipe_delete_depth_stencil_state(struct pipe_context *, const struct pipe_depth_stencil_state *); +const struct pipe_rasterizer_state * +softpipe_create_rasterizer_state(struct pipe_context *, + const struct pipe_rasterizer_state *); +void softpipe_bind_rasterizer_state(struct pipe_context *, + const struct pipe_rasterizer_state *); +void softpipe_delete_rasterizer_state(struct pipe_context *, + const struct pipe_rasterizer_state *); + void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); @@ -93,9 +100,6 @@ void softpipe_set_polygon_stipple( struct pipe_context *, void softpipe_set_scissor_state( struct pipe_context *, const struct pipe_scissor_state * ); -void softpipe_set_setup_state( struct pipe_context *, - const struct pipe_setup_state * ); - void softpipe_set_texture_state( struct pipe_context *, unsigned unit, struct pipe_mipmap_tree * ); diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 47743e185c..8c6bacf65c 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -45,7 +45,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { const uint inputsRead = softpipe->fs.inputs_read; const interp_mode colorInterp - = softpipe->setup.flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; uint i; @@ -112,7 +112,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) * lighting. Edgeflag is dealt with specially by setting bits in * the vertex header. */ - if (softpipe->setup.light_twoside) { + if (softpipe->rasterizer->light_twoside) { if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { back0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT, colorInterp); @@ -160,7 +160,7 @@ compute_cliprect(struct softpipe_context *sp) surfHeight = sp->scissor.maxy; } - if (sp->setup.scissor) { + if (sp->rasterizer->scissor) { /* clip to scissor rect */ sp->cliprect.minx = MAX2(sp->scissor.minx, 0); sp->cliprect.miny = MAX2(sp->scissor.miny, 0); @@ -182,7 +182,7 @@ compute_cliprect(struct softpipe_context *sp) */ void softpipe_update_derived( struct softpipe_context *softpipe ) { - if (softpipe->dirty & (SP_NEW_SETUP | SP_NEW_FS)) + if (softpipe->dirty & (SP_NEW_RASTERIZER | SP_NEW_FS)) calculate_vertex_layout( softpipe ); if (softpipe->dirty & (SP_NEW_SCISSOR | @@ -194,7 +194,7 @@ void softpipe_update_derived( struct softpipe_context *softpipe ) SP_NEW_DEPTH_STENCIL | SP_NEW_ALPHA_TEST | SP_NEW_FRAMEBUFFER | - SP_NEW_SETUP | + SP_NEW_RASTERIZER | SP_NEW_FS)) sp_build_quad_pipeline(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_state_setup.c b/src/mesa/pipe/softpipe/sp_state_setup.c index 4715a26f55..6788396355 100644 --- a/src/mesa/pipe/softpipe/sp_state_setup.c +++ b/src/mesa/pipe/softpipe/sp_state_setup.c @@ -31,17 +31,35 @@ #include "pipe/draw/draw_context.h" -void softpipe_set_setup_state( struct pipe_context *pipe, - const struct pipe_setup_state *setup ) + +const struct pipe_rasterizer_state * +softpipe_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup) +{ + struct pipe_rasterizer_state *raster = + malloc(sizeof(struct pipe_rasterizer_state)); + memcpy(raster, setup, sizeof(struct pipe_rasterizer_state)); + + return raster; +} + +void softpipe_bind_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup) { struct softpipe_context *softpipe = softpipe_context(pipe); /* pass-through to draw module */ draw_set_setup_state(softpipe->draw, setup); - memcpy( &softpipe->setup, setup, sizeof(*setup) ); + softpipe->rasterizer = setup; + + softpipe->dirty |= SP_NEW_RASTERIZER; +} - softpipe->dirty |= SP_NEW_SETUP; +void softpipe_delete_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *rasterizer) +{ + free((struct pipe_rasterizer_state*)rasterizer); } diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 99d0bcb90b..a4af3aeb20 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -53,7 +53,7 @@ static const struct st_tracked_state *atoms[] = &st_update_vs, &st_update_fs, - &st_update_setup, + &st_update_rasterizer, &st_update_polygon_stipple, &st_update_viewport, &st_update_scissor, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 0e362b1fbf..26f6514698 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -51,7 +51,7 @@ const struct st_tracked_state st_update_depth_stencil; const struct st_tracked_state st_update_tnl; const struct st_tracked_state st_update_fs; const struct st_tracked_state st_update_vs; -const struct st_tracked_state st_update_setup; +const struct st_tracked_state st_update_rasterizer; const struct st_tracked_state st_update_polygon_stipple; const struct st_tracked_state st_update_viewport; const struct st_tracked_state st_update_scissor; diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c index 09d921560d..cab8ad5cd6 100644 --- a/src/mesa/state_tracker/st_atom_setup.c +++ b/src/mesa/state_tracker/st_atom_setup.c @@ -32,6 +32,7 @@ #include "st_context.h" +#include "st_cache.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "st_atom.h" @@ -68,20 +69,21 @@ static GLboolean get_offset_flag( GLuint fill_mode, } -static void update_setup_state( struct st_context *st ) +static void update_raster_state( struct st_context *st ) { GLcontext *ctx = st->ctx; - struct pipe_setup_state setup; + struct pipe_rasterizer_state raster; + const struct pipe_rasterizer_state *cached; - memset(&setup, 0, sizeof(setup)); + memset(&raster, 0, sizeof(raster)); /* _NEW_POLYGON, _NEW_BUFFERS */ { if (ctx->Polygon.FrontFace == GL_CCW) - setup.front_winding = PIPE_WINDING_CCW; + raster.front_winding = PIPE_WINDING_CCW; else - setup.front_winding = PIPE_WINDING_CW; + raster.front_winding = PIPE_WINDING_CW; /* XXX * I think the intention here is that user-created framebuffer objects @@ -90,13 +92,13 @@ static void update_setup_state( struct st_context *st ) * But this is an implementation/driver-specific artifact - remove... */ if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) - setup.front_winding ^= PIPE_WINDING_BOTH; + raster.front_winding ^= PIPE_WINDING_BOTH; } /* _NEW_LIGHT */ if (ctx->Light.ShadeModel == GL_FLAT) - setup.flatshade = 1; + raster.flatshade = 1; /* _NEW_LIGHT | _NEW_PROGRAM * @@ -105,23 +107,23 @@ static void update_setup_state( struct st_context *st ) * GL_VERTEX_PROGRAM_TWO_SIDE is set). Note the logic here. */ if (ctx->VertexProgram._Enabled) { - setup.light_twoside = ctx->VertexProgram.TwoSideEnabled; + raster.light_twoside = ctx->VertexProgram.TwoSideEnabled; } else if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { - setup.light_twoside = 1; + raster.light_twoside = 1; } /* _NEW_POLYGON */ if (ctx->Polygon.CullFlag) { if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) { - setup.cull_mode = PIPE_WINDING_BOTH; + raster.cull_mode = PIPE_WINDING_BOTH; } else if (ctx->Polygon.CullFaceMode == GL_FRONT) { - setup.cull_mode = setup.front_winding; + raster.cull_mode = raster.front_winding; } else { - setup.cull_mode = setup.front_winding ^ PIPE_WINDING_BOTH; + raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH; } } @@ -131,23 +133,23 @@ static void update_setup_state( struct st_context *st ) GLuint fill_front = translate_fill( ctx->Polygon.FrontMode ); GLuint fill_back = translate_fill( ctx->Polygon.BackMode ); - if (setup.front_winding == PIPE_WINDING_CW) { - setup.fill_cw = fill_front; - setup.fill_ccw = fill_back; + if (raster.front_winding == PIPE_WINDING_CW) { + raster.fill_cw = fill_front; + raster.fill_ccw = fill_back; } else { - setup.fill_cw = fill_back; - setup.fill_ccw = fill_front; + raster.fill_cw = fill_back; + raster.fill_ccw = fill_front; } /* Simplify when culling is active: */ - if (setup.cull_mode & PIPE_WINDING_CW) { - setup.fill_cw = setup.fill_ccw; + if (raster.cull_mode & PIPE_WINDING_CW) { + raster.fill_cw = raster.fill_ccw; } - if (setup.cull_mode & PIPE_WINDING_CCW) { - setup.fill_ccw = setup.fill_cw; + if (raster.cull_mode & PIPE_WINDING_CCW) { + raster.fill_ccw = raster.fill_cw; } } @@ -155,67 +157,68 @@ static void update_setup_state( struct st_context *st ) */ if (ctx->Polygon.OffsetUnits != 0.0 || ctx->Polygon.OffsetFactor != 0.0) { - setup.offset_cw = get_offset_flag( setup.fill_cw, &ctx->Polygon ); - setup.offset_ccw = get_offset_flag( setup.fill_ccw, &ctx->Polygon ); - setup.offset_units = ctx->Polygon.OffsetUnits; - setup.offset_scale = ctx->Polygon.OffsetFactor; + raster.offset_cw = get_offset_flag( raster.fill_cw, &ctx->Polygon ); + raster.offset_ccw = get_offset_flag( raster.fill_ccw, &ctx->Polygon ); + raster.offset_units = ctx->Polygon.OffsetUnits; + raster.offset_scale = ctx->Polygon.OffsetFactor; } if (ctx->Polygon.SmoothFlag) - setup.poly_smooth = 1; + raster.poly_smooth = 1; if (ctx->Polygon.StippleFlag) - setup.poly_stipple_enable = 1; + raster.poly_stipple_enable = 1; /* _NEW_BUFFERS, _NEW_POLYGON */ - if (setup.fill_cw != PIPE_POLYGON_MODE_FILL || - setup.fill_ccw != PIPE_POLYGON_MODE_FILL) + if (raster.fill_cw != PIPE_POLYGON_MODE_FILL || + raster.fill_ccw != PIPE_POLYGON_MODE_FILL) { GLfloat mrd = (ctx->DrawBuffer ? ctx->DrawBuffer->_MRD : 1.0); - setup.offset_units = ctx->Polygon.OffsetFactor * mrd; - setup.offset_scale = (ctx->Polygon.OffsetUnits * mrd * + raster.offset_units = ctx->Polygon.OffsetFactor * mrd; + raster.offset_scale = (ctx->Polygon.OffsetUnits * mrd * st->polygon_offset_scale); } /* _NEW_POINT */ - setup.point_size = ctx->Point.Size; - setup.point_smooth = ctx->Point.SmoothFlag; + raster.point_size = ctx->Point.Size; + raster.point_smooth = ctx->Point.SmoothFlag; /* _NEW_LINE */ - setup.line_width = ctx->Line.Width; - setup.line_smooth = ctx->Line.SmoothFlag; - setup.line_stipple_enable = ctx->Line.StippleFlag; - setup.line_stipple_pattern = ctx->Line.StipplePattern; + raster.line_width = ctx->Line.Width; + raster.line_smooth = ctx->Line.SmoothFlag; + raster.line_stipple_enable = ctx->Line.StippleFlag; + raster.line_stipple_pattern = ctx->Line.StipplePattern; /* GL stipple factor is in [1,256], remap to [0, 255] here */ - setup.line_stipple_factor = ctx->Line.StippleFactor - 1; + raster.line_stipple_factor = ctx->Line.StippleFactor - 1; /* _NEW_MULTISAMPLE */ if (ctx->Multisample.Enabled) - setup.multisample = 1; + raster.multisample = 1; /* _NEW_SCISSOR */ if (ctx->Scissor.Enabled) - setup.scissor = 1; + raster.scissor = 1; - if (memcmp(&setup, &st->state.setup, sizeof(setup)) != 0) { - st->state.setup = setup; - st->pipe->set_setup_state( st->pipe, &setup ); + cached = st_cached_rasterizer_state(st, &raster); + if (st->state.rasterizer != cached) { + st->state.rasterizer = cached; + st->pipe->bind_rasterizer_state( st->pipe, cached ); } } -const struct st_tracked_state st_update_setup = { - .name = "st_update_setup", +const struct st_tracked_state st_update_rasterizer = { + .name = "st_update_rasterizer", .dirty = { .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | _NEW_SCISSOR | _NEW_POINT | _NEW_BUFFERS | _NEW_MULTISAMPLE), .st = 0, }, - .update = update_setup_state + .update = update_raster_state }; diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 64c03be99d..a687c15587 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -93,3 +93,21 @@ struct pipe_depth_stencil_state * st_cached_depth_stencil_state( } return (struct pipe_depth_stencil_state*)(cso_hash_iter_data(iter)); } + +struct pipe_rasterizer_state * st_cached_rasterizer_state( + struct st_context *st, + const struct pipe_rasterizer_state *raster) +{ + unsigned hash_key = cso_construct_key((void*)raster, + sizeof(struct pipe_rasterizer_state)); + struct cso_hash_iter iter = cso_find_state_template(st->cache, + hash_key, CSO_RASTERIZER, + (void*)raster); + if (cso_hash_iter_is_null(iter)) { + const struct pipe_rasterizer_state *created_state = + st->pipe->create_rasterizer_state(st->pipe, raster); + iter = cso_insert_state(st->cache, hash_key, CSO_RASTERIZER, + (void*)created_state); + } + return (struct pipe_rasterizer_state*)(cso_hash_iter_data(iter)); +} diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index 78cb2e774e..a06af31123 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -47,6 +47,10 @@ struct pipe_sampler_state * st_cached_sampler_state( struct pipe_depth_stencil_state *st_cached_depth_stencil_state( struct st_context *st, - const struct pipe_depth_stencil_state *sampler); + const struct pipe_depth_stencil_state *depth_stencil); + +struct pipe_rasterizer_state *st_cached_rasterizer_state( + struct st_context *st, + const struct pipe_rasterizer_state *raster); #endif diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index e9aabd15b5..584bc1cc2a 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -329,16 +329,18 @@ clear_with_quad(GLcontext *ctx, /* setup state: nothing */ { - struct pipe_setup_state setup; - memset(&setup, 0, sizeof(setup)); + struct pipe_rasterizer_state raster; + const struct pipe_rasterizer_state *cached; + memset(&raster, 0, sizeof(raster)); #if 0 /* don't do per-pixel scissor; we'll just draw a PIPE_PRIM_QUAD * that matches the scissor bounds. */ if (ctx->Scissor.Enabled) - setup.scissor = 1; + raster.scissor = 1; #endif - pipe->set_setup_state(pipe, &setup); + cached = st_cached_rasterizer_state(ctx->st, &raster); + pipe->bind_rasterizer_state(pipe, cached); } /* fragment shader state: color pass-through program */ @@ -394,7 +396,7 @@ clear_with_quad(GLcontext *ctx, pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil); pipe->set_fs_state(pipe, &st->state.fs); pipe->set_vs_state(pipe, &st->state.vs); - pipe->set_setup_state(pipe, &st->state.setup); + pipe->bind_rasterizer_state(pipe, st->state.rasterizer); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); /* OR: st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index a0012e3a8c..78ede8e225 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -316,11 +316,13 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* setup state: just scissor */ { - struct pipe_setup_state setup; + struct pipe_rasterizer_state setup; + struct pipe_rasterizer_state *cached; memset(&setup, 0, sizeof(setup)); if (ctx->Scissor.Enabled) setup.scissor = 1; - pipe->set_setup_state(pipe, &setup); + cached = st_cached_rasterizer_state(ctx->st, &setup); + pipe->bind_rasterizer_state(pipe, cached); } /* fragment shader state: TEX lookup program */ @@ -400,7 +402,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, draw_quad(ctx, x0, y0, z, x1, y1); /* restore GL state */ - pipe->set_setup_state(pipe, &ctx->st->state.setup); + pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer); pipe->set_fs_state(pipe, &ctx->st->state.fs); pipe->set_vs_state(pipe, &ctx->st->state.vs); pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 7c887d0b55..516d319a6e 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -77,6 +77,7 @@ struct st_context const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; + const struct pipe_rasterizer_state *rasterizer; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -88,7 +89,6 @@ struct st_context struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_setup_state setup; struct pipe_shader_state fs; struct pipe_shader_state vs; struct pipe_viewport_state viewport; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 69f4b7fa5b..1ea7799021 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -394,7 +394,7 @@ st_feedback_draw_vbo(GLcontext *ctx, assert(draw); draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); - draw_set_setup_state(draw, &st->state.setup); + draw_set_setup_state(draw, st->state.rasterizer); draw_set_vertex_shader(draw, &st->state.vs); /* XXX need to set vertex info too */ -- cgit v1.2.3 From 29789ca026c03dff64b6286e202561faf111fde9 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 17 Sep 2007 13:49:12 -0400 Subject: Fix a silly bug on setting samplers. Bind when the currently bound state is /not/ the one we just found/created. --- src/mesa/state_tracker/st_atom_sampler.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 9a728e2d79..994d3691d8 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -146,7 +146,7 @@ update_samplers(struct st_context *st) const struct pipe_sampler_state *cached_sampler = st_cached_sampler_state(st, &sampler); - if (cached_sampler == st->state.sampler[u]) { + if (cached_sampler != st->state.sampler[u]) { /* state has changed */ st->state.sampler[u] = cached_sampler; st->pipe->bind_sampler_state(st->pipe, u, cached_sampler); -- cgit v1.2.3 From fbb2f840aef9a540401575b032edb4dfdabd10bb Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 18 Sep 2007 06:24:54 -0400 Subject: move cso cache to the pipe directory --- src/mesa/cso_cache/cso_cache.c | 164 ---------------- src/mesa/cso_cache/cso_cache.h | 72 ------- src/mesa/cso_cache/cso_hash.c | 381 ------------------------------------ src/mesa/cso_cache/cso_hash.h | 62 ------ src/mesa/pipe/cso_cache/cso_cache.c | 164 ++++++++++++++++ src/mesa/pipe/cso_cache/cso_cache.h | 72 +++++++ src/mesa/pipe/cso_cache/cso_hash.c | 381 ++++++++++++++++++++++++++++++++++++ src/mesa/pipe/cso_cache/cso_hash.h | 62 ++++++ src/mesa/sources | 4 +- src/mesa/state_tracker/st_cache.c | 4 +- src/mesa/state_tracker/st_context.c | 2 +- 11 files changed, 684 insertions(+), 684 deletions(-) delete mode 100644 src/mesa/cso_cache/cso_cache.c delete mode 100644 src/mesa/cso_cache/cso_cache.h delete mode 100644 src/mesa/cso_cache/cso_hash.c delete mode 100644 src/mesa/cso_cache/cso_hash.h create mode 100644 src/mesa/pipe/cso_cache/cso_cache.c create mode 100644 src/mesa/pipe/cso_cache/cso_cache.h create mode 100644 src/mesa/pipe/cso_cache/cso_hash.c create mode 100644 src/mesa/pipe/cso_cache/cso_hash.h (limited to 'src') diff --git a/src/mesa/cso_cache/cso_cache.c b/src/mesa/cso_cache/cso_cache.c deleted file mode 100644 index 4aaadf00e6..0000000000 --- a/src/mesa/cso_cache/cso_cache.c +++ /dev/null @@ -1,164 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Zack Rusin - */ - -#include "cso_cache.h" -#include "cso_hash.h" - -#if 1 -static unsigned hash_key(const void *key, unsigned key_size) -{ - unsigned *ikey = (unsigned *)key; - unsigned hash = 0, i; - - assert(key_size % 4 == 0); - - /* I'm sure this can be improved on: - */ - for (i = 0; i < key_size/4; i++) - hash ^= ikey[i]; - - return hash; -} -#else -static unsigned hash_key(const unsigned char *p, int n) -{ - unsigned h = 0; - unsigned g; - - while (n--) { - h = (h << 4) + *p++; - if ((g = (h & 0xf0000000)) != 0) - h ^= g >> 23; - h &= ~g; - } - return h; -} -#endif - -unsigned cso_construct_key(void *item, int item_size) -{ - return hash_key((item), item_size); -} - -static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_type type) -{ - struct cso_hash *hash = 0; - - switch(type) { - case CSO_BLEND: - hash = sc->blend_hash; - case CSO_SAMPLER: - hash = sc->sampler_hash; - case CSO_DEPTH_STENCIL: - hash = sc->depth_stencil_hash; - case CSO_RASTERIZER: - hash = sc->rasterizer_hash; - } - - return hash; -} - -static int _cso_size_for_type(enum cso_cache_type type) -{ - switch(type) { - case CSO_BLEND: - return sizeof(struct pipe_blend_state); - case CSO_SAMPLER: - return sizeof(struct pipe_sampler_state); - case CSO_DEPTH_STENCIL: - return sizeof(struct pipe_depth_stencil_state); - case CSO_RASTERIZER: - return sizeof(struct pipe_rasterizer_state); - } - return 0; -} - -struct cso_hash_iter -cso_insert_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *state) -{ - struct cso_hash *hash = _cso_hash_for_type(sc, type); - return cso_hash_insert(hash, hash_key, state); -} - -struct cso_hash_iter -cso_find_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type) -{ - struct cso_hash *hash = _cso_hash_for_type(sc, type); - - return cso_hash_find(hash, hash_key); -} - -struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *templ) -{ - struct cso_hash_iter iter = cso_find_state(sc, hash_key, type); - int size = _cso_size_for_type(type); - while (!cso_hash_iter_is_null(iter)) { - void *iter_data = cso_hash_iter_data(iter); - if (!memcmp(iter_data, templ, size)) - return iter; - iter = cso_hash_iter_next(iter); - } - return iter; -} - -void * cso_take_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type) -{ - struct cso_hash *hash = _cso_hash_for_type(sc, type); - return cso_hash_take(hash, hash_key); -} - -struct cso_cache *cso_cache_create(void) -{ - struct cso_cache *sc = malloc(sizeof(struct cso_cache)); - - sc->blend_hash = cso_hash_create(); - sc->sampler_hash = cso_hash_create(); - sc->depth_stencil_hash = cso_hash_create(); - sc->rasterizer_hash = cso_hash_create(); - - return sc; -} - -void cso_cache_delete(struct cso_cache *sc) -{ - assert(sc); - cso_hash_delete(sc->blend_hash); - cso_hash_delete(sc->sampler_hash); - cso_hash_delete(sc->depth_stencil_hash); - cso_hash_delete(sc->rasterizer_hash); - free(sc); -} - diff --git a/src/mesa/cso_cache/cso_cache.h b/src/mesa/cso_cache/cso_cache.h deleted file mode 100644 index 23be9cd713..0000000000 --- a/src/mesa/cso_cache/cso_cache.h +++ /dev/null @@ -1,72 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#ifndef CSO_CACHE_H -#define CSO_CACHE_H - -#include "pipe/p_context.h" -#include "pipe/p_state.h" - - -struct cso_hash; - -struct cso_cache { - struct cso_hash *blend_hash; - struct cso_hash *sampler_hash; - struct cso_hash *depth_stencil_hash; - struct cso_hash *rasterizer_hash; -}; - -enum cso_cache_type { - CSO_BLEND, - CSO_SAMPLER, - CSO_DEPTH_STENCIL, - CSO_RASTERIZER -}; - -unsigned cso_construct_key(void *item, int item_size); - -struct cso_cache *cso_cache_create(void); -void cso_cache_delete(struct cso_cache *sc); - -struct cso_hash_iter cso_insert_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *state); -struct cso_hash_iter cso_find_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type); -struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *templ); -void * cso_take_state(struct cso_cache *sc, unsigned hash_key, - enum cso_cache_type type); - -#endif diff --git a/src/mesa/cso_cache/cso_hash.c b/src/mesa/cso_cache/cso_hash.c deleted file mode 100644 index fe033c9e91..0000000000 --- a/src/mesa/cso_cache/cso_hash.c +++ /dev/null @@ -1,381 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#include "cso_hash.h" - -#include -#include -#include -#include - -#define MAX(a, b) ((a > b) ? (a) : (b)) - -static const int MinNumBits = 4; - -static const unsigned char prime_deltas[] = { - 0, 0, 1, 3, 1, 5, 3, 3, 1, 9, 7, 5, 3, 9, 25, 3, - 1, 21, 3, 21, 7, 15, 9, 5, 3, 29, 15, 0, 0, 0, 0, 0 -}; - -static int primeForNumBits(int numBits) -{ - return (1 << numBits) + prime_deltas[numBits]; -} - -/* - Returns the smallest integer n such that - primeForNumBits(n) >= hint. -*/ -static int countBits(int hint) -{ - int numBits = 0; - int bits = hint; - - while (bits > 1) { - bits >>= 1; - numBits++; - } - - if (numBits >= (int)sizeof(prime_deltas)) { - numBits = sizeof(prime_deltas) - 1; - } else if (primeForNumBits(numBits) < hint) { - ++numBits; - } - return numBits; -} - -struct cso_node { - struct cso_node *next; - unsigned key; - void *value; -}; - -struct cso_hash_data { - struct cso_node *fakeNext; - struct cso_node **buckets; - int size; - int nodeSize; - short userNumBits; - short numBits; - int numBuckets; -}; - -struct cso_hash { - union { - struct cso_hash_data *d; - struct cso_node *e; - } data; -}; - -static void *cso_data_allocate_node(struct cso_hash_data *hash) -{ - return malloc(hash->nodeSize); -} - -static void cso_data_free_node(void *node) -{ - free(node); -} - -static struct cso_node * -cso_hash_create_node(struct cso_hash *hash, - unsigned akey, void *avalue, - struct cso_node **anextNode) -{ - struct cso_node *node = cso_data_allocate_node(hash->data.d); - node->key = akey; - node->value = avalue; - - node->next = (struct cso_node*)(*anextNode); - *anextNode = node; - ++hash->data.d->size; - return node; -} - -static void cso_data_rehash(struct cso_hash_data *hash, int hint) -{ - if (hint < 0) { - hint = countBits(-hint); - if (hint < MinNumBits) - hint = MinNumBits; - hash->userNumBits = hint; - while (primeForNumBits(hint) < (hash->size >> 1)) - ++hint; - } else if (hint < MinNumBits) { - hint = MinNumBits; - } - - if (hash->numBits != hint) { - struct cso_node *e = (struct cso_node *)(hash); - struct cso_node **oldBuckets = hash->buckets; - int oldNumBuckets = hash->numBuckets; - int i = 0; - - hash->numBits = hint; - hash->numBuckets = primeForNumBits(hint); - hash->buckets = malloc(sizeof(struct cso_node*) * hash->numBuckets); - for (i = 0; i < hash->numBuckets; ++i) - hash->buckets[i] = e; - - for (i = 0; i < oldNumBuckets; ++i) { - struct cso_node *firstNode = oldBuckets[i]; - while (firstNode != e) { - unsigned h = firstNode->key; - struct cso_node *lastNode = firstNode; - while (lastNode->next != e && lastNode->next->key == h) - lastNode = lastNode->next; - - struct cso_node *afterLastNode = lastNode->next; - struct cso_node **beforeFirstNode = &hash->buckets[h % hash->numBuckets]; - while (*beforeFirstNode != e) - beforeFirstNode = &(*beforeFirstNode)->next; - lastNode->next = *beforeFirstNode; - *beforeFirstNode = firstNode; - firstNode = afterLastNode; - } - } - free(oldBuckets); - } -} - -static void cso_data_might_grow(struct cso_hash_data *hash) -{ - if (hash->size >= hash->numBuckets) - cso_data_rehash(hash, hash->numBits + 1); -} - -static void cso_data_has_shrunk(struct cso_hash_data *hash) -{ - if (hash->size <= (hash->numBuckets >> 3) && - hash->numBits > hash->userNumBits) { - int max = MAX(hash->numBits-2, hash->userNumBits); - cso_data_rehash(hash, max); - } -} - -static struct cso_node *cso_data_first_node(struct cso_hash_data *hash) -{ - struct cso_node *e = (struct cso_node *)(hash); - struct cso_node **bucket = hash->buckets; - int n = hash->numBuckets; - while (n--) { - if (*bucket != e) - return *bucket; - ++bucket; - } - return e; -} - -static struct cso_node **cso_hash_find_node(struct cso_hash *hash, unsigned akey) -{ - struct cso_node **node; - - if (hash->data.d->numBuckets) { - node = (struct cso_node **)(&hash->data.d->buckets[akey % hash->data.d->numBuckets]); - assert(*node == hash->data.e || (*node)->next); - while (*node != hash->data.e && (*node)->key != akey) - node = &(*node)->next; - } else { - node = (struct cso_node **)((const struct cso_node * const *)(&hash->data.e)); - } - return node; -} - -struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, - unsigned key, void *data) -{ - cso_data_might_grow(hash->data.d); - - struct cso_node **nextNode = cso_hash_find_node(hash, key); - struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); - struct cso_hash_iter iter = {hash, node}; - return iter; -} - -struct cso_hash * cso_hash_create(void) -{ - struct cso_hash *hash = malloc(sizeof(struct cso_hash)); - hash->data.d = malloc(sizeof(struct cso_hash_data)); - hash->data.d->fakeNext = 0; - hash->data.d->buckets = 0; - hash->data.d->size = 0; - hash->data.d->nodeSize = sizeof(struct cso_node); - hash->data.d->userNumBits = MinNumBits; - hash->data.d->numBits = 0; - hash->data.d->numBuckets = 0; - - return hash; -} - -void cso_hash_delete(struct cso_hash *hash) -{ - struct cso_node *e_for_x = (struct cso_node *)(hash->data.d); - struct cso_node **bucket = (struct cso_node **)(hash->data.d->buckets); - int n = hash->data.d->numBuckets; - while (n--) { - struct cso_node *cur = *bucket++; - while (cur != e_for_x) { - struct cso_node *next = cur->next; - cso_data_free_node(cur); - cur = next; - } - } - free(hash->data.d->buckets); - free(hash->data.d); - free(hash); -} - -struct cso_hash_iter cso_hash_find(struct cso_hash *hash, - unsigned key) -{ - struct cso_node **nextNode = cso_hash_find_node(hash, key); - struct cso_hash_iter iter = {hash, *nextNode}; - return iter; -} - -unsigned cso_hash_iter_key(struct cso_hash_iter iter) -{ - if (!iter.node || iter.hash->data.e == iter.node) - return 0; - return iter.node->key; -} - -void * cso_hash_iter_data(struct cso_hash_iter iter) -{ - if (!iter.node || iter.hash->data.e == iter.node) - return 0; - return iter.node->value; -} - -static struct cso_node *cso_hash_data_next(struct cso_node *node) -{ - union { - struct cso_node *next; - struct cso_node *e; - struct cso_hash_data *d; - } a; - a.next = node->next; - if (!a.next) { - fprintf(stderr, "iterating beyond the last element\n"); - return 0; - } - if (a.next->next) - return a.next; - - int start = (node->key % a.d->numBuckets) + 1; - struct cso_node **bucket = a.d->buckets + start; - int n = a.d->numBuckets - start; - while (n--) { - if (*bucket != a.e) - return *bucket; - ++bucket; - } - return a.e; -} - - -static struct cso_node *cso_hash_data_prev(struct cso_node *node) -{ - union { - struct cso_node *e; - struct cso_hash_data *d; - } a; - - a.e = node; - while (a.e->next) - a.e = a.e->next; - - int start; - if (node == a.e) - start = a.d->numBuckets - 1; - else - start = node->key % a.d->numBuckets; - - struct cso_node *sentinel = node; - struct cso_node **bucket = a.d->buckets + start; - while (start >= 0) { - if (*bucket != sentinel) { - struct cso_node *prev = *bucket; - while (prev->next != sentinel) - prev = prev->next; - return prev; - } - - sentinel = a.e; - --bucket; - --start; - } - fprintf(stderr, "iterating backward beyond first element\n"); - return a.e; -} - -struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter) -{ - struct cso_hash_iter next = {iter.hash, cso_hash_data_next(iter.node)}; - return next; -} - -int cso_hash_iter_is_null(struct cso_hash_iter iter) -{ - if (!iter.node || iter.node == iter.hash->data.e) - return 1; - return 0; -} - -void * cso_hash_take(struct cso_hash *hash, - unsigned akey) -{ - struct cso_node **node = cso_hash_find_node(hash, akey); - if (*node != hash->data.e) { - void *t = (*node)->value; - struct cso_node *next = (*node)->next; - cso_data_free_node(*node); - *node = next; - --hash->data.d->size; - cso_data_has_shrunk(hash->data.d); - return t; - } - return 0; -} - -struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter) -{ - struct cso_hash_iter prev = {iter.hash, - cso_hash_data_prev(iter.node)}; - return prev; -} - -struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash) -{ - struct cso_hash_iter iter = {hash, cso_data_first_node(hash->data.d)}; - return iter; -} diff --git a/src/mesa/cso_cache/cso_hash.h b/src/mesa/cso_cache/cso_hash.h deleted file mode 100644 index b4aa111860..0000000000 --- a/src/mesa/cso_cache/cso_hash.h +++ /dev/null @@ -1,62 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#ifndef CSO_HASH_H -#define CSO_HASH_H - -struct cso_hash; -struct cso_node; - -struct cso_hash_iter { - struct cso_hash *hash; - struct cso_node *node; -}; - -struct cso_hash *cso_hash_create(void); -void cso_hash_delete(struct cso_hash *hash); - -struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, - void *data); -void *cso_hash_take(struct cso_hash *hash, unsigned key); - -struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash); -struct cso_hash_iter cso_hash_find(struct cso_hash *hash, unsigned key); - - -int cso_hash_iter_is_null(struct cso_hash_iter iter); -unsigned cso_hash_iter_key(struct cso_hash_iter iter); -void *cso_hash_iter_data(struct cso_hash_iter iter); - -struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter); -struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); - -#endif diff --git a/src/mesa/pipe/cso_cache/cso_cache.c b/src/mesa/pipe/cso_cache/cso_cache.c new file mode 100644 index 0000000000..4aaadf00e6 --- /dev/null +++ b/src/mesa/pipe/cso_cache/cso_cache.c @@ -0,0 +1,164 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Zack Rusin + */ + +#include "cso_cache.h" +#include "cso_hash.h" + +#if 1 +static unsigned hash_key(const void *key, unsigned key_size) +{ + unsigned *ikey = (unsigned *)key; + unsigned hash = 0, i; + + assert(key_size % 4 == 0); + + /* I'm sure this can be improved on: + */ + for (i = 0; i < key_size/4; i++) + hash ^= ikey[i]; + + return hash; +} +#else +static unsigned hash_key(const unsigned char *p, int n) +{ + unsigned h = 0; + unsigned g; + + while (n--) { + h = (h << 4) + *p++; + if ((g = (h & 0xf0000000)) != 0) + h ^= g >> 23; + h &= ~g; + } + return h; +} +#endif + +unsigned cso_construct_key(void *item, int item_size) +{ + return hash_key((item), item_size); +} + +static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_type type) +{ + struct cso_hash *hash = 0; + + switch(type) { + case CSO_BLEND: + hash = sc->blend_hash; + case CSO_SAMPLER: + hash = sc->sampler_hash; + case CSO_DEPTH_STENCIL: + hash = sc->depth_stencil_hash; + case CSO_RASTERIZER: + hash = sc->rasterizer_hash; + } + + return hash; +} + +static int _cso_size_for_type(enum cso_cache_type type) +{ + switch(type) { + case CSO_BLEND: + return sizeof(struct pipe_blend_state); + case CSO_SAMPLER: + return sizeof(struct pipe_sampler_state); + case CSO_DEPTH_STENCIL: + return sizeof(struct pipe_depth_stencil_state); + case CSO_RASTERIZER: + return sizeof(struct pipe_rasterizer_state); + } + return 0; +} + +struct cso_hash_iter +cso_insert_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *state) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + return cso_hash_insert(hash, hash_key, state); +} + +struct cso_hash_iter +cso_find_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + + return cso_hash_find(hash, hash_key); +} + +struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *templ) +{ + struct cso_hash_iter iter = cso_find_state(sc, hash_key, type); + int size = _cso_size_for_type(type); + while (!cso_hash_iter_is_null(iter)) { + void *iter_data = cso_hash_iter_data(iter); + if (!memcmp(iter_data, templ, size)) + return iter; + iter = cso_hash_iter_next(iter); + } + return iter; +} + +void * cso_take_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + return cso_hash_take(hash, hash_key); +} + +struct cso_cache *cso_cache_create(void) +{ + struct cso_cache *sc = malloc(sizeof(struct cso_cache)); + + sc->blend_hash = cso_hash_create(); + sc->sampler_hash = cso_hash_create(); + sc->depth_stencil_hash = cso_hash_create(); + sc->rasterizer_hash = cso_hash_create(); + + return sc; +} + +void cso_cache_delete(struct cso_cache *sc) +{ + assert(sc); + cso_hash_delete(sc->blend_hash); + cso_hash_delete(sc->sampler_hash); + cso_hash_delete(sc->depth_stencil_hash); + cso_hash_delete(sc->rasterizer_hash); + free(sc); +} + diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h new file mode 100644 index 0000000000..23be9cd713 --- /dev/null +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -0,0 +1,72 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#ifndef CSO_CACHE_H +#define CSO_CACHE_H + +#include "pipe/p_context.h" +#include "pipe/p_state.h" + + +struct cso_hash; + +struct cso_cache { + struct cso_hash *blend_hash; + struct cso_hash *sampler_hash; + struct cso_hash *depth_stencil_hash; + struct cso_hash *rasterizer_hash; +}; + +enum cso_cache_type { + CSO_BLEND, + CSO_SAMPLER, + CSO_DEPTH_STENCIL, + CSO_RASTERIZER +}; + +unsigned cso_construct_key(void *item, int item_size); + +struct cso_cache *cso_cache_create(void); +void cso_cache_delete(struct cso_cache *sc); + +struct cso_hash_iter cso_insert_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *state); +struct cso_hash_iter cso_find_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type); +struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *templ); +void * cso_take_state(struct cso_cache *sc, unsigned hash_key, + enum cso_cache_type type); + +#endif diff --git a/src/mesa/pipe/cso_cache/cso_hash.c b/src/mesa/pipe/cso_cache/cso_hash.c new file mode 100644 index 0000000000..fe033c9e91 --- /dev/null +++ b/src/mesa/pipe/cso_cache/cso_hash.c @@ -0,0 +1,381 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#include "cso_hash.h" + +#include +#include +#include +#include + +#define MAX(a, b) ((a > b) ? (a) : (b)) + +static const int MinNumBits = 4; + +static const unsigned char prime_deltas[] = { + 0, 0, 1, 3, 1, 5, 3, 3, 1, 9, 7, 5, 3, 9, 25, 3, + 1, 21, 3, 21, 7, 15, 9, 5, 3, 29, 15, 0, 0, 0, 0, 0 +}; + +static int primeForNumBits(int numBits) +{ + return (1 << numBits) + prime_deltas[numBits]; +} + +/* + Returns the smallest integer n such that + primeForNumBits(n) >= hint. +*/ +static int countBits(int hint) +{ + int numBits = 0; + int bits = hint; + + while (bits > 1) { + bits >>= 1; + numBits++; + } + + if (numBits >= (int)sizeof(prime_deltas)) { + numBits = sizeof(prime_deltas) - 1; + } else if (primeForNumBits(numBits) < hint) { + ++numBits; + } + return numBits; +} + +struct cso_node { + struct cso_node *next; + unsigned key; + void *value; +}; + +struct cso_hash_data { + struct cso_node *fakeNext; + struct cso_node **buckets; + int size; + int nodeSize; + short userNumBits; + short numBits; + int numBuckets; +}; + +struct cso_hash { + union { + struct cso_hash_data *d; + struct cso_node *e; + } data; +}; + +static void *cso_data_allocate_node(struct cso_hash_data *hash) +{ + return malloc(hash->nodeSize); +} + +static void cso_data_free_node(void *node) +{ + free(node); +} + +static struct cso_node * +cso_hash_create_node(struct cso_hash *hash, + unsigned akey, void *avalue, + struct cso_node **anextNode) +{ + struct cso_node *node = cso_data_allocate_node(hash->data.d); + node->key = akey; + node->value = avalue; + + node->next = (struct cso_node*)(*anextNode); + *anextNode = node; + ++hash->data.d->size; + return node; +} + +static void cso_data_rehash(struct cso_hash_data *hash, int hint) +{ + if (hint < 0) { + hint = countBits(-hint); + if (hint < MinNumBits) + hint = MinNumBits; + hash->userNumBits = hint; + while (primeForNumBits(hint) < (hash->size >> 1)) + ++hint; + } else if (hint < MinNumBits) { + hint = MinNumBits; + } + + if (hash->numBits != hint) { + struct cso_node *e = (struct cso_node *)(hash); + struct cso_node **oldBuckets = hash->buckets; + int oldNumBuckets = hash->numBuckets; + int i = 0; + + hash->numBits = hint; + hash->numBuckets = primeForNumBits(hint); + hash->buckets = malloc(sizeof(struct cso_node*) * hash->numBuckets); + for (i = 0; i < hash->numBuckets; ++i) + hash->buckets[i] = e; + + for (i = 0; i < oldNumBuckets; ++i) { + struct cso_node *firstNode = oldBuckets[i]; + while (firstNode != e) { + unsigned h = firstNode->key; + struct cso_node *lastNode = firstNode; + while (lastNode->next != e && lastNode->next->key == h) + lastNode = lastNode->next; + + struct cso_node *afterLastNode = lastNode->next; + struct cso_node **beforeFirstNode = &hash->buckets[h % hash->numBuckets]; + while (*beforeFirstNode != e) + beforeFirstNode = &(*beforeFirstNode)->next; + lastNode->next = *beforeFirstNode; + *beforeFirstNode = firstNode; + firstNode = afterLastNode; + } + } + free(oldBuckets); + } +} + +static void cso_data_might_grow(struct cso_hash_data *hash) +{ + if (hash->size >= hash->numBuckets) + cso_data_rehash(hash, hash->numBits + 1); +} + +static void cso_data_has_shrunk(struct cso_hash_data *hash) +{ + if (hash->size <= (hash->numBuckets >> 3) && + hash->numBits > hash->userNumBits) { + int max = MAX(hash->numBits-2, hash->userNumBits); + cso_data_rehash(hash, max); + } +} + +static struct cso_node *cso_data_first_node(struct cso_hash_data *hash) +{ + struct cso_node *e = (struct cso_node *)(hash); + struct cso_node **bucket = hash->buckets; + int n = hash->numBuckets; + while (n--) { + if (*bucket != e) + return *bucket; + ++bucket; + } + return e; +} + +static struct cso_node **cso_hash_find_node(struct cso_hash *hash, unsigned akey) +{ + struct cso_node **node; + + if (hash->data.d->numBuckets) { + node = (struct cso_node **)(&hash->data.d->buckets[akey % hash->data.d->numBuckets]); + assert(*node == hash->data.e || (*node)->next); + while (*node != hash->data.e && (*node)->key != akey) + node = &(*node)->next; + } else { + node = (struct cso_node **)((const struct cso_node * const *)(&hash->data.e)); + } + return node; +} + +struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, + unsigned key, void *data) +{ + cso_data_might_grow(hash->data.d); + + struct cso_node **nextNode = cso_hash_find_node(hash, key); + struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); + struct cso_hash_iter iter = {hash, node}; + return iter; +} + +struct cso_hash * cso_hash_create(void) +{ + struct cso_hash *hash = malloc(sizeof(struct cso_hash)); + hash->data.d = malloc(sizeof(struct cso_hash_data)); + hash->data.d->fakeNext = 0; + hash->data.d->buckets = 0; + hash->data.d->size = 0; + hash->data.d->nodeSize = sizeof(struct cso_node); + hash->data.d->userNumBits = MinNumBits; + hash->data.d->numBits = 0; + hash->data.d->numBuckets = 0; + + return hash; +} + +void cso_hash_delete(struct cso_hash *hash) +{ + struct cso_node *e_for_x = (struct cso_node *)(hash->data.d); + struct cso_node **bucket = (struct cso_node **)(hash->data.d->buckets); + int n = hash->data.d->numBuckets; + while (n--) { + struct cso_node *cur = *bucket++; + while (cur != e_for_x) { + struct cso_node *next = cur->next; + cso_data_free_node(cur); + cur = next; + } + } + free(hash->data.d->buckets); + free(hash->data.d); + free(hash); +} + +struct cso_hash_iter cso_hash_find(struct cso_hash *hash, + unsigned key) +{ + struct cso_node **nextNode = cso_hash_find_node(hash, key); + struct cso_hash_iter iter = {hash, *nextNode}; + return iter; +} + +unsigned cso_hash_iter_key(struct cso_hash_iter iter) +{ + if (!iter.node || iter.hash->data.e == iter.node) + return 0; + return iter.node->key; +} + +void * cso_hash_iter_data(struct cso_hash_iter iter) +{ + if (!iter.node || iter.hash->data.e == iter.node) + return 0; + return iter.node->value; +} + +static struct cso_node *cso_hash_data_next(struct cso_node *node) +{ + union { + struct cso_node *next; + struct cso_node *e; + struct cso_hash_data *d; + } a; + a.next = node->next; + if (!a.next) { + fprintf(stderr, "iterating beyond the last element\n"); + return 0; + } + if (a.next->next) + return a.next; + + int start = (node->key % a.d->numBuckets) + 1; + struct cso_node **bucket = a.d->buckets + start; + int n = a.d->numBuckets - start; + while (n--) { + if (*bucket != a.e) + return *bucket; + ++bucket; + } + return a.e; +} + + +static struct cso_node *cso_hash_data_prev(struct cso_node *node) +{ + union { + struct cso_node *e; + struct cso_hash_data *d; + } a; + + a.e = node; + while (a.e->next) + a.e = a.e->next; + + int start; + if (node == a.e) + start = a.d->numBuckets - 1; + else + start = node->key % a.d->numBuckets; + + struct cso_node *sentinel = node; + struct cso_node **bucket = a.d->buckets + start; + while (start >= 0) { + if (*bucket != sentinel) { + struct cso_node *prev = *bucket; + while (prev->next != sentinel) + prev = prev->next; + return prev; + } + + sentinel = a.e; + --bucket; + --start; + } + fprintf(stderr, "iterating backward beyond first element\n"); + return a.e; +} + +struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter) +{ + struct cso_hash_iter next = {iter.hash, cso_hash_data_next(iter.node)}; + return next; +} + +int cso_hash_iter_is_null(struct cso_hash_iter iter) +{ + if (!iter.node || iter.node == iter.hash->data.e) + return 1; + return 0; +} + +void * cso_hash_take(struct cso_hash *hash, + unsigned akey) +{ + struct cso_node **node = cso_hash_find_node(hash, akey); + if (*node != hash->data.e) { + void *t = (*node)->value; + struct cso_node *next = (*node)->next; + cso_data_free_node(*node); + *node = next; + --hash->data.d->size; + cso_data_has_shrunk(hash->data.d); + return t; + } + return 0; +} + +struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter) +{ + struct cso_hash_iter prev = {iter.hash, + cso_hash_data_prev(iter.node)}; + return prev; +} + +struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash) +{ + struct cso_hash_iter iter = {hash, cso_data_first_node(hash->data.d)}; + return iter; +} diff --git a/src/mesa/pipe/cso_cache/cso_hash.h b/src/mesa/pipe/cso_cache/cso_hash.h new file mode 100644 index 0000000000..b4aa111860 --- /dev/null +++ b/src/mesa/pipe/cso_cache/cso_hash.h @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#ifndef CSO_HASH_H +#define CSO_HASH_H + +struct cso_hash; +struct cso_node; + +struct cso_hash_iter { + struct cso_hash *hash; + struct cso_node *node; +}; + +struct cso_hash *cso_hash_create(void); +void cso_hash_delete(struct cso_hash *hash); + +struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, + void *data); +void *cso_hash_take(struct cso_hash *hash, unsigned key); + +struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash); +struct cso_hash_iter cso_hash_find(struct cso_hash *hash, unsigned key); + + +int cso_hash_iter_is_null(struct cso_hash_iter iter); +unsigned cso_hash_iter_key(struct cso_hash_iter iter); +void *cso_hash_iter_data(struct cso_hash_iter iter); + +struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter); +struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); + +#endif diff --git a/src/mesa/sources b/src/mesa/sources index 22b592df09..017a0167bd 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -187,8 +187,8 @@ TGSIMESA_SOURCES = \ pipe/tgsi/mesa/mesa_to_tgsi.c STATECACHE_SOURCES = \ - cso_cache/cso_hash.c \ - cso_cache/cso_cache.c + pipe/cso_cache/cso_hash.c \ + pipe/cso_cache/cso_cache.c STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index a687c15587..e9c79634bd 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -36,8 +36,8 @@ #include "pipe/p_state.h" -#include "cso_cache/cso_cache.h" -#include "cso_cache/cso_hash.h" +#include "pipe/cso_cache/cso_cache.h" +#include "pipe/cso_cache/cso_hash.h" /* This function will either find the state of the given template * in the cache or it will create a new state state from the given diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index f9717465f4..e0304dd22d 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -45,7 +45,7 @@ #include "st_program.h" #include "pipe/p_context.h" #include "pipe/draw/draw_context.h" -#include "cso_cache/cso_cache.h" +#include "pipe/cso_cache/cso_cache.h" void st_invalidate_state(GLcontext * ctx, GLuint new_state) -- cgit v1.2.3 From de69fc1703f79e5c97e66b654de7a93b7abce8f0 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 18 Sep 2007 10:02:16 -0400 Subject: Finishing up rename of the setup state to the rasterizer state. --- src/mesa/pipe/draw/draw_context.c | 14 +- src/mesa/pipe/draw/draw_context.h | 8 +- src/mesa/pipe/draw/draw_private.h | 2 +- src/mesa/pipe/i915simple/i915_context.c | 2 +- src/mesa/pipe/i915simple/i915_state.c | 2 +- src/mesa/pipe/softpipe/Makefile | 2 +- src/mesa/pipe/softpipe/sp_context.c | 2 +- src/mesa/pipe/softpipe/sp_state_rasterizer.c | 65 ++++++++ src/mesa/pipe/softpipe/sp_state_setup.c | 65 -------- src/mesa/sources | 2 +- src/mesa/state_tracker/st_atom_rasterizer.c | 224 +++++++++++++++++++++++++++ src/mesa/state_tracker/st_atom_setup.c | 224 --------------------------- src/mesa/state_tracker/st_cb_feedback.c | 4 +- src/mesa/state_tracker/st_draw.c | 2 +- 14 files changed, 309 insertions(+), 309 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_state_rasterizer.c delete mode 100644 src/mesa/pipe/softpipe/sp_state_setup.c create mode 100644 src/mesa/state_tracker/st_atom_rasterizer.c delete mode 100644 src/mesa/state_tracker/st_atom_setup.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index f3236ad59e..9acbc53742 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -89,7 +89,7 @@ void draw_destroy( struct draw_context *draw ) */ static void validate_pipeline( struct draw_context *draw ) { - struct draw_stage *next = draw->pipeline.setup; + struct draw_stage *next = draw->pipeline.rasterize; /* * NOTE: we build up the pipeline in end-to-start order. @@ -157,11 +157,11 @@ void draw_set_feedback_state( struct draw_context *draw, /** - * Register new primitive setup/rendering state. + * Register new primitive rasterization/rendering state. * This causes the drawing pipeline to be rebuilt. */ -void draw_set_setup_state( struct draw_context *draw, - const struct pipe_rasterizer_state *raster ) +void draw_set_rasterizer_state( struct draw_context *draw, + const struct pipe_rasterizer_state *raster ) { draw->rasterizer = raster; validate_pipeline( draw ); @@ -172,10 +172,10 @@ void draw_set_setup_state( struct draw_context *draw, * Plug in the primitive rendering/rasterization stage. * This is provided by the device driver. */ -void draw_set_setup_stage( struct draw_context *draw, - struct draw_stage *stage ) +void draw_set_rasterize_stage( struct draw_context *draw, + struct draw_stage *stage ) { - draw->pipeline.setup = stage; + draw->pipeline.rasterize = stage; } diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 2714252fc5..4eb59aab01 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -86,11 +86,11 @@ void draw_set_clip_state( struct draw_context *pipe, void draw_set_feedback_state( struct draw_context *draw, const struct pipe_feedback_state * ); -void draw_set_setup_state( struct draw_context *draw, - const struct pipe_rasterizer_state *raster ); +void draw_set_rasterizer_state( struct draw_context *draw, + const struct pipe_rasterizer_state *raster ); -void draw_set_setup_stage( struct draw_context *draw, - struct draw_stage *stage ); +void draw_set_rasterize_stage( struct draw_context *draw, + struct draw_stage *stage ); unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index fb0aaff40d..bd8c11eb94 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -133,7 +133,7 @@ struct draw_context struct draw_stage *twoside; struct draw_stage *offset; struct draw_stage *unfilled; - struct draw_stage *setup; /* aka render/rasterize */ + struct draw_stage *rasterize; } pipeline; /* pipe state that we need: */ diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 6e48b3bd03..36372898ce 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -285,7 +285,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, */ i915->draw = draw_create(); assert(i915->draw); - draw_set_setup_stage(i915->draw, i915_draw_render_stage(i915)); + draw_set_rasterize_stage(i915->draw, i915_draw_render_stage(i915)); i915_init_region_functions(i915); i915_init_surface_functions(i915); diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 00764902bc..1dfa10ab28 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -308,7 +308,7 @@ static void i915_bind_rasterizer_state( struct pipe_context *pipe, i915->rasterizer = setup; /* pass-through to draw module */ - draw_set_setup_state(i915->draw, setup); + draw_set_rasterizer_state(i915->draw, setup); i915->dirty |= I915_NEW_RASTERIZER; } diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 401df1cfb7..0335f56b99 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -29,7 +29,7 @@ DRIVER_SOURCES = \ sp_state_feedback.c \ sp_state_fs.c \ sp_state_sampler.c \ - sp_state_setup.c \ + sp_state_rasterizer.c \ sp_state_surface.c \ sp_state_vertex.c \ sp_tex_layout.c \ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 7753ce40d7..cf5fc2227e 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -318,7 +318,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, */ softpipe->draw = draw_create(); assert(softpipe->draw); - draw_set_setup_stage(softpipe->draw, sp_draw_render_stage(softpipe)); + draw_set_rasterize_stage(softpipe->draw, sp_draw_render_stage(softpipe)); sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_state_rasterizer.c b/src/mesa/pipe/softpipe/sp_state_rasterizer.c new file mode 100644 index 0000000000..d832adb91b --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_state_rasterizer.c @@ -0,0 +1,65 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "sp_context.h" +#include "sp_state.h" +#include "pipe/draw/draw_context.h" + + + +const struct pipe_rasterizer_state * +softpipe_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup) +{ + struct pipe_rasterizer_state *raster = + malloc(sizeof(struct pipe_rasterizer_state)); + memcpy(raster, setup, sizeof(struct pipe_rasterizer_state)); + + return raster; +} + +void softpipe_bind_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + /* pass-through to draw module */ + draw_set_rasterizer_state(softpipe->draw, setup); + + softpipe->rasterizer = setup; + + softpipe->dirty |= SP_NEW_RASTERIZER; +} + +void softpipe_delete_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *rasterizer) +{ + free((struct pipe_rasterizer_state*)rasterizer); +} + + diff --git a/src/mesa/pipe/softpipe/sp_state_setup.c b/src/mesa/pipe/softpipe/sp_state_setup.c deleted file mode 100644 index 6788396355..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_setup.c +++ /dev/null @@ -1,65 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_defines.h" -#include "sp_context.h" -#include "sp_state.h" -#include "pipe/draw/draw_context.h" - - - -const struct pipe_rasterizer_state * -softpipe_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup) -{ - struct pipe_rasterizer_state *raster = - malloc(sizeof(struct pipe_rasterizer_state)); - memcpy(raster, setup, sizeof(struct pipe_rasterizer_state)); - - return raster; -} - -void softpipe_bind_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - /* pass-through to draw module */ - draw_set_setup_state(softpipe->draw, setup); - - softpipe->rasterizer = setup; - - softpipe->dirty |= SP_NEW_RASTERIZER; -} - -void softpipe_delete_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *rasterizer) -{ - free((struct pipe_rasterizer_state*)rasterizer); -} - - diff --git a/src/mesa/sources b/src/mesa/sources index 017a0167bd..0d4fdc15f4 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -203,7 +203,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_fs.c \ state_tracker/st_atom_sampler.c \ state_tracker/st_atom_scissor.c \ - state_tracker/st_atom_setup.c \ + state_tracker/st_atom_rasterizer.c \ state_tracker/st_atom_stipple.c \ state_tracker/st_atom_texture.c \ state_tracker/st_atom_viewport.c \ diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c new file mode 100644 index 0000000000..cab8ad5cd6 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -0,0 +1,224 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "st_context.h" +#include "st_cache.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "st_atom.h" + +static GLuint translate_fill( GLenum mode ) +{ + switch (mode) { + case GL_POINT: + return PIPE_POLYGON_MODE_POINT; + case GL_LINE: + return PIPE_POLYGON_MODE_LINE; + case GL_FILL: + return PIPE_POLYGON_MODE_FILL; + default: + assert(0); + return 0; + } +} + +static GLboolean get_offset_flag( GLuint fill_mode, + const struct gl_polygon_attrib *p ) +{ + switch (fill_mode) { + case PIPE_POLYGON_MODE_POINT: + return p->OffsetPoint; + case PIPE_POLYGON_MODE_LINE: + return p->OffsetLine; + case PIPE_POLYGON_MODE_FILL: + return p->OffsetFill; + default: + assert(0); + return 0; + } +} + + +static void update_raster_state( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + struct pipe_rasterizer_state raster; + const struct pipe_rasterizer_state *cached; + + memset(&raster, 0, sizeof(raster)); + + /* _NEW_POLYGON, _NEW_BUFFERS + */ + { + if (ctx->Polygon.FrontFace == GL_CCW) + raster.front_winding = PIPE_WINDING_CCW; + else + raster.front_winding = PIPE_WINDING_CW; + + /* XXX + * I think the intention here is that user-created framebuffer objects + * use Y=0=TOP layout instead of OpenGL's normal Y=0=bottom layout. + * Flipping Y changes CW to CCW and vice-versa. + * But this is an implementation/driver-specific artifact - remove... + */ + if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) + raster.front_winding ^= PIPE_WINDING_BOTH; + } + + /* _NEW_LIGHT + */ + if (ctx->Light.ShadeModel == GL_FLAT) + raster.flatshade = 1; + + /* _NEW_LIGHT | _NEW_PROGRAM + * + * Back-face colors can come from traditional lighting (when + * GL_LIGHT_MODEL_TWO_SIDE is set) or from vertex programs (when + * GL_VERTEX_PROGRAM_TWO_SIDE is set). Note the logic here. + */ + if (ctx->VertexProgram._Enabled) { + raster.light_twoside = ctx->VertexProgram.TwoSideEnabled; + } + else if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { + raster.light_twoside = 1; + } + + /* _NEW_POLYGON + */ + if (ctx->Polygon.CullFlag) { + if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) { + raster.cull_mode = PIPE_WINDING_BOTH; + } + else if (ctx->Polygon.CullFaceMode == GL_FRONT) { + raster.cull_mode = raster.front_winding; + } + else { + raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH; + } + } + + /* _NEW_POLYGON + */ + { + GLuint fill_front = translate_fill( ctx->Polygon.FrontMode ); + GLuint fill_back = translate_fill( ctx->Polygon.BackMode ); + + if (raster.front_winding == PIPE_WINDING_CW) { + raster.fill_cw = fill_front; + raster.fill_ccw = fill_back; + } + else { + raster.fill_cw = fill_back; + raster.fill_ccw = fill_front; + } + + /* Simplify when culling is active: + */ + if (raster.cull_mode & PIPE_WINDING_CW) { + raster.fill_cw = raster.fill_ccw; + } + + if (raster.cull_mode & PIPE_WINDING_CCW) { + raster.fill_ccw = raster.fill_cw; + } + } + + /* _NEW_POLYGON + */ + if (ctx->Polygon.OffsetUnits != 0.0 || + ctx->Polygon.OffsetFactor != 0.0) { + raster.offset_cw = get_offset_flag( raster.fill_cw, &ctx->Polygon ); + raster.offset_ccw = get_offset_flag( raster.fill_ccw, &ctx->Polygon ); + raster.offset_units = ctx->Polygon.OffsetUnits; + raster.offset_scale = ctx->Polygon.OffsetFactor; + } + + if (ctx->Polygon.SmoothFlag) + raster.poly_smooth = 1; + + if (ctx->Polygon.StippleFlag) + raster.poly_stipple_enable = 1; + + + /* _NEW_BUFFERS, _NEW_POLYGON + */ + if (raster.fill_cw != PIPE_POLYGON_MODE_FILL || + raster.fill_ccw != PIPE_POLYGON_MODE_FILL) + { + GLfloat mrd = (ctx->DrawBuffer ? + ctx->DrawBuffer->_MRD : + 1.0); + + raster.offset_units = ctx->Polygon.OffsetFactor * mrd; + raster.offset_scale = (ctx->Polygon.OffsetUnits * mrd * + st->polygon_offset_scale); + } + + /* _NEW_POINT + */ + raster.point_size = ctx->Point.Size; + raster.point_smooth = ctx->Point.SmoothFlag; + + /* _NEW_LINE + */ + raster.line_width = ctx->Line.Width; + raster.line_smooth = ctx->Line.SmoothFlag; + raster.line_stipple_enable = ctx->Line.StippleFlag; + raster.line_stipple_pattern = ctx->Line.StipplePattern; + /* GL stipple factor is in [1,256], remap to [0, 255] here */ + raster.line_stipple_factor = ctx->Line.StippleFactor - 1; + + /* _NEW_MULTISAMPLE */ + if (ctx->Multisample.Enabled) + raster.multisample = 1; + + /* _NEW_SCISSOR */ + if (ctx->Scissor.Enabled) + raster.scissor = 1; + + cached = st_cached_rasterizer_state(st, &raster); + if (st->state.rasterizer != cached) { + st->state.rasterizer = cached; + st->pipe->bind_rasterizer_state( st->pipe, cached ); + } +} + +const struct st_tracked_state st_update_rasterizer = { + .name = "st_update_rasterizer", + .dirty = { + .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | _NEW_SCISSOR | + _NEW_POINT | _NEW_BUFFERS | _NEW_MULTISAMPLE), + .st = 0, + }, + .update = update_raster_state +}; diff --git a/src/mesa/state_tracker/st_atom_setup.c b/src/mesa/state_tracker/st_atom_setup.c deleted file mode 100644 index cab8ad5cd6..0000000000 --- a/src/mesa/state_tracker/st_atom_setup.c +++ /dev/null @@ -1,224 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#include "st_context.h" -#include "st_cache.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "st_atom.h" - -static GLuint translate_fill( GLenum mode ) -{ - switch (mode) { - case GL_POINT: - return PIPE_POLYGON_MODE_POINT; - case GL_LINE: - return PIPE_POLYGON_MODE_LINE; - case GL_FILL: - return PIPE_POLYGON_MODE_FILL; - default: - assert(0); - return 0; - } -} - -static GLboolean get_offset_flag( GLuint fill_mode, - const struct gl_polygon_attrib *p ) -{ - switch (fill_mode) { - case PIPE_POLYGON_MODE_POINT: - return p->OffsetPoint; - case PIPE_POLYGON_MODE_LINE: - return p->OffsetLine; - case PIPE_POLYGON_MODE_FILL: - return p->OffsetFill; - default: - assert(0); - return 0; - } -} - - -static void update_raster_state( struct st_context *st ) -{ - GLcontext *ctx = st->ctx; - struct pipe_rasterizer_state raster; - const struct pipe_rasterizer_state *cached; - - memset(&raster, 0, sizeof(raster)); - - /* _NEW_POLYGON, _NEW_BUFFERS - */ - { - if (ctx->Polygon.FrontFace == GL_CCW) - raster.front_winding = PIPE_WINDING_CCW; - else - raster.front_winding = PIPE_WINDING_CW; - - /* XXX - * I think the intention here is that user-created framebuffer objects - * use Y=0=TOP layout instead of OpenGL's normal Y=0=bottom layout. - * Flipping Y changes CW to CCW and vice-versa. - * But this is an implementation/driver-specific artifact - remove... - */ - if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) - raster.front_winding ^= PIPE_WINDING_BOTH; - } - - /* _NEW_LIGHT - */ - if (ctx->Light.ShadeModel == GL_FLAT) - raster.flatshade = 1; - - /* _NEW_LIGHT | _NEW_PROGRAM - * - * Back-face colors can come from traditional lighting (when - * GL_LIGHT_MODEL_TWO_SIDE is set) or from vertex programs (when - * GL_VERTEX_PROGRAM_TWO_SIDE is set). Note the logic here. - */ - if (ctx->VertexProgram._Enabled) { - raster.light_twoside = ctx->VertexProgram.TwoSideEnabled; - } - else if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { - raster.light_twoside = 1; - } - - /* _NEW_POLYGON - */ - if (ctx->Polygon.CullFlag) { - if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) { - raster.cull_mode = PIPE_WINDING_BOTH; - } - else if (ctx->Polygon.CullFaceMode == GL_FRONT) { - raster.cull_mode = raster.front_winding; - } - else { - raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH; - } - } - - /* _NEW_POLYGON - */ - { - GLuint fill_front = translate_fill( ctx->Polygon.FrontMode ); - GLuint fill_back = translate_fill( ctx->Polygon.BackMode ); - - if (raster.front_winding == PIPE_WINDING_CW) { - raster.fill_cw = fill_front; - raster.fill_ccw = fill_back; - } - else { - raster.fill_cw = fill_back; - raster.fill_ccw = fill_front; - } - - /* Simplify when culling is active: - */ - if (raster.cull_mode & PIPE_WINDING_CW) { - raster.fill_cw = raster.fill_ccw; - } - - if (raster.cull_mode & PIPE_WINDING_CCW) { - raster.fill_ccw = raster.fill_cw; - } - } - - /* _NEW_POLYGON - */ - if (ctx->Polygon.OffsetUnits != 0.0 || - ctx->Polygon.OffsetFactor != 0.0) { - raster.offset_cw = get_offset_flag( raster.fill_cw, &ctx->Polygon ); - raster.offset_ccw = get_offset_flag( raster.fill_ccw, &ctx->Polygon ); - raster.offset_units = ctx->Polygon.OffsetUnits; - raster.offset_scale = ctx->Polygon.OffsetFactor; - } - - if (ctx->Polygon.SmoothFlag) - raster.poly_smooth = 1; - - if (ctx->Polygon.StippleFlag) - raster.poly_stipple_enable = 1; - - - /* _NEW_BUFFERS, _NEW_POLYGON - */ - if (raster.fill_cw != PIPE_POLYGON_MODE_FILL || - raster.fill_ccw != PIPE_POLYGON_MODE_FILL) - { - GLfloat mrd = (ctx->DrawBuffer ? - ctx->DrawBuffer->_MRD : - 1.0); - - raster.offset_units = ctx->Polygon.OffsetFactor * mrd; - raster.offset_scale = (ctx->Polygon.OffsetUnits * mrd * - st->polygon_offset_scale); - } - - /* _NEW_POINT - */ - raster.point_size = ctx->Point.Size; - raster.point_smooth = ctx->Point.SmoothFlag; - - /* _NEW_LINE - */ - raster.line_width = ctx->Line.Width; - raster.line_smooth = ctx->Line.SmoothFlag; - raster.line_stipple_enable = ctx->Line.StippleFlag; - raster.line_stipple_pattern = ctx->Line.StipplePattern; - /* GL stipple factor is in [1,256], remap to [0, 255] here */ - raster.line_stipple_factor = ctx->Line.StippleFactor - 1; - - /* _NEW_MULTISAMPLE */ - if (ctx->Multisample.Enabled) - raster.multisample = 1; - - /* _NEW_SCISSOR */ - if (ctx->Scissor.Enabled) - raster.scissor = 1; - - cached = st_cached_rasterizer_state(st, &raster); - if (st->state.rasterizer != cached) { - st->state.rasterizer = cached; - st->pipe->bind_rasterizer_state( st->pipe, cached ); - } -} - -const struct st_tracked_state st_update_rasterizer = { - .name = "st_update_rasterizer", - .dirty = { - .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | _NEW_SCISSOR | - _NEW_POINT | _NEW_BUFFERS | _NEW_MULTISAMPLE), - .st = 0, - }, - .update = update_raster_state -}; diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index b97b2bb027..8e8084fe59 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -285,14 +285,14 @@ st_RenderMode(GLcontext *ctx, GLenum newMode ) else if (newMode == GL_SELECT) { if (!st->selection_stage) st->selection_stage = draw_glselect_stage(ctx, draw); - draw_set_setup_stage(draw, st->selection_stage); + draw_set_rasterize_stage(draw, st->selection_stage); /* Plug in new vbo draw function */ vbo->draw_prims = st_feedback_draw_vbo; } else { if (!st->feedback_stage) st->feedback_stage = draw_glfeedback_stage(ctx, draw); - draw_set_setup_stage(draw, st->feedback_stage); + draw_set_rasterize_stage(draw, st->feedback_stage); /* Plug in new vbo draw function */ vbo->draw_prims = st_feedback_draw_vbo; } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 1ea7799021..f68e449f07 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -394,7 +394,7 @@ st_feedback_draw_vbo(GLcontext *ctx, assert(draw); draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); - draw_set_setup_state(draw, st->state.rasterizer); + draw_set_rasterizer_state(draw, st->state.rasterizer); draw_set_vertex_shader(draw, &st->state.vs); /* XXX need to set vertex info too */ -- cgit v1.2.3 From ccd63b54cfbb6bb241d55f7ac95afcd14819f469 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 18 Sep 2007 13:24:44 -0400 Subject: Convert shader to an immutable state object. --- src/mesa/pipe/cso_cache/cso_cache.c | 13 +++++++++---- src/mesa/pipe/cso_cache/cso_cache.h | 4 +++- src/mesa/pipe/failover/fo_context.h | 4 ++-- src/mesa/pipe/failover/fo_state.c | 19 +++++++++---------- src/mesa/pipe/failover/fo_state_emit.c | 4 ++-- src/mesa/pipe/i915simple/i915_context.h | 2 +- src/mesa/pipe/i915simple/i915_fpc.h | 2 +- src/mesa/pipe/i915simple/i915_fpc_translate.c | 8 ++++---- src/mesa/pipe/i915simple/i915_state.c | 27 ++++++++++++++++++++++----- src/mesa/pipe/i915simple/i915_state_derived.c | 2 +- src/mesa/pipe/p_context.h | 16 ++++++++++------ src/mesa/pipe/softpipe/sp_context.c | 14 ++++++++------ src/mesa/pipe/softpipe/sp_context.h | 4 ++-- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/pipe/softpipe/sp_state.h | 14 +++++++++----- src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- src/mesa/pipe/softpipe/sp_state_fs.c | 24 ++++++++++++++++++++---- src/mesa/state_tracker/st_atom_fs.c | 16 +++++++++++----- src/mesa/state_tracker/st_atom_vs.c | 16 +++++++++++----- src/mesa/state_tracker/st_cache.c | 18 ++++++++++++++++++ src/mesa/state_tracker/st_cache.h | 4 ++++ src/mesa/state_tracker/st_cb_clear.c | 12 ++++++++---- src/mesa/state_tracker/st_cb_drawpixels.c | 12 ++++++++---- src/mesa/state_tracker/st_cb_rasterpos.c | 6 +++--- src/mesa/state_tracker/st_context.h | 4 ++-- src/mesa/state_tracker/st_draw.c | 6 +++--- src/mesa/state_tracker/st_program.h | 6 +++--- 27 files changed, 176 insertions(+), 85 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_cache.c b/src/mesa/pipe/cso_cache/cso_cache.c index 4aaadf00e6..be653d9494 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.c +++ b/src/mesa/pipe/cso_cache/cso_cache.c @@ -80,6 +80,8 @@ static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_ hash = sc->depth_stencil_hash; case CSO_RASTERIZER: hash = sc->rasterizer_hash; + case CSO_SHADER: + hash = sc->shader_hash; } return hash; @@ -96,6 +98,8 @@ static int _cso_size_for_type(enum cso_cache_type type) return sizeof(struct pipe_depth_stencil_state); case CSO_RASTERIZER: return sizeof(struct pipe_rasterizer_state); + case CSO_SHADER: + return sizeof(struct pipe_shader_state); } return 0; } @@ -144,10 +148,11 @@ struct cso_cache *cso_cache_create(void) { struct cso_cache *sc = malloc(sizeof(struct cso_cache)); - sc->blend_hash = cso_hash_create(); - sc->sampler_hash = cso_hash_create(); + sc->blend_hash = cso_hash_create(); + sc->sampler_hash = cso_hash_create(); sc->depth_stencil_hash = cso_hash_create(); - sc->rasterizer_hash = cso_hash_create(); + sc->rasterizer_hash = cso_hash_create(); + sc->shader_hash = cso_hash_create(); return sc; } @@ -159,6 +164,6 @@ void cso_cache_delete(struct cso_cache *sc) cso_hash_delete(sc->sampler_hash); cso_hash_delete(sc->depth_stencil_hash); cso_hash_delete(sc->rasterizer_hash); + cso_hash_delete(sc->shader_hash); free(sc); } - diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h index 23be9cd713..d9793ca855 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -44,13 +44,15 @@ struct cso_cache { struct cso_hash *sampler_hash; struct cso_hash *depth_stencil_hash; struct cso_hash *rasterizer_hash; + struct cso_hash *shader_hash; }; enum cso_cache_type { CSO_BLEND, CSO_SAMPLER, CSO_DEPTH_STENCIL, - CSO_RASTERIZER + CSO_RASTERIZER, + CSO_SHADER }; unsigned cso_construct_key(void *item, int item_size); diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index b05ceb88ad..9556a17e4d 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -70,14 +70,14 @@ struct failover_context { const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; + const struct pipe_shader_state *fragment_shader; + const struct pipe_shader_state *vertex_shader; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_framebuffer_state framebuffer; - struct pipe_shader_state fragment_shader; - struct pipe_shader_state vertex_shader; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 8e2b649590..04ebd33d0d 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -125,28 +125,27 @@ failover_set_framebuffer_state(struct pipe_context *pipe, } static void -failover_set_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *fs) +failover_bind_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *fs) { struct failover_context *failover = failover_context(pipe); - failover->fragment_shader = *fs; + failover->fragment_shader = fs; failover->dirty |= FO_NEW_FRAGMENT_SHADER; - failover->hw->set_fs_state( failover->hw, fs ); + failover->hw->bind_fs_state( failover->hw, fs ); } static void -failover_set_vs_state(struct pipe_context *pipe, +failover_bind_vs_state(struct pipe_context *pipe, const struct pipe_shader_state *vs) { struct failover_context *failover = failover_context(pipe); - failover->vertex_shader = *vs; + failover->vertex_shader = vs; failover->dirty |= FO_NEW_VERTEX_SHADER; - failover->hw->set_vs_state( failover->hw, vs ); + failover->hw->bind_vs_state( failover->hw, vs ); } - static void failover_set_polygon_stipple( struct pipe_context *pipe, const struct pipe_poly_stipple *stipple ) @@ -258,14 +257,14 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.bind_sampler_state = failover_bind_sampler_state; failover->pipe.bind_depth_stencil_state = failover_bind_depth_stencil_state; failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state; + failover->pipe.bind_fs_state = failover_bind_fs_state; + failover->pipe.bind_vs_state = failover_bind_vs_state; failover->pipe.set_alpha_test_state = failover_set_alpha_test_state; failover->pipe.set_blend_color = failover_set_blend_color; failover->pipe.set_clip_state = failover_set_clip_state; failover->pipe.set_clear_color_state = failover_set_clear_color_state; failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; - failover->pipe.set_fs_state = failover_set_fs_state; - failover->pipe.set_vs_state = failover_set_vs_state; failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; failover->pipe.set_scissor_state = failover_set_scissor_state; failover->pipe.set_texture_state = failover_set_texture_state; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 1c9573a7b0..9d304074d0 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -77,10 +77,10 @@ failover_state_emit( struct failover_context *failover ) failover->sw->set_framebuffer_state( failover->sw, &failover->framebuffer ); if (failover->dirty & FO_NEW_FRAGMENT_SHADER) - failover->sw->set_fs_state( failover->sw, &failover->fragment_shader ); + failover->sw->bind_fs_state( failover->sw, failover->fragment_shader ); if (failover->dirty & FO_NEW_VERTEX_SHADER) - failover->sw->set_vs_state( failover->sw, &failover->vertex_shader ); + failover->sw->bind_vs_state( failover->sw, failover->vertex_shader ); if (failover->dirty & FO_NEW_STIPPLE) failover->sw->set_polygon_stipple( failover->sw, &failover->poly_stipple ); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 3fab821fde..9052c92d72 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -127,6 +127,7 @@ struct i915_context const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; + const struct pipe_shader_state *fs; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -134,7 +135,6 @@ struct i915_context struct pipe_clip_state clip; struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; struct pipe_framebuffer_state framebuffer; - struct pipe_shader_state fs; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index 59fcbf40d9..84e4c5a6f3 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -44,7 +44,7 @@ * Program translation state */ struct i915_fp_compile { - struct pipe_shader_state *shader; + const struct pipe_shader_state *shader; struct vertex_info *vertex_info; diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index c2ad80c5d0..32c5600496 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -872,11 +872,11 @@ i915_translate_instructions(struct i915_fp_compile *p, static struct i915_fp_compile * i915_init_compile(struct i915_context *i915, - struct pipe_shader_state *fs) + const struct pipe_shader_state *fs) { struct i915_fp_compile *p = CALLOC_STRUCT(i915_fp_compile); - p->shader = &i915->fs; + p->shader = i915->fs; p->vertex_info = &i915->current.vertex_info; @@ -1032,8 +1032,8 @@ i915_fixup_depth_write(struct i915_fp_compile *p) void i915_translate_fragment_program( struct i915_context *i915 ) { - struct i915_fp_compile *p = i915_init_compile(i915, &i915->fs); - const struct tgsi_token *tokens = i915->fs.tokens; + struct i915_fp_compile *p = i915_init_compile(i915, i915->fs); + const struct tgsi_token *tokens = i915->fs->tokens; i915_find_wpos_space(p); diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 1dfa10ab28..fe835643e0 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -161,19 +161,29 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe, } +static const struct pipe_shader_state * +i915_create_shader_state( struct pipe_context *pipe, + const struct pipe_shader_state *templ ) +{ + + struct pipe_shader_state *shader = malloc(sizeof(struct pipe_shader_state)); + memcpy(shader, templ, sizeof(struct pipe_shader_state)); + + return shader; +} -static void i915_set_fs_state( struct pipe_context *pipe, +static void i915_bind_fs_state( struct pipe_context *pipe, const struct pipe_shader_state *fs ) { struct i915_context *i915 = i915_context(pipe); - memcpy(&i915->fs, fs, sizeof(*fs)); + i915->fs = fs; i915->dirty |= I915_NEW_FS; } -static void i915_set_vs_state( struct pipe_context *pipe, +static void i915_bind_vs_state( struct pipe_context *pipe, const struct pipe_shader_state *vs ) { struct i915_context *i915 = i915_context(pipe); @@ -182,6 +192,11 @@ static void i915_set_vs_state( struct pipe_context *pipe, draw_set_vertex_shader(i915->draw, vs); } +static void i915_delete_shader_state( struct pipe_context *pipe, + const struct pipe_shader_state *shader ) +{ + free((struct pipe_shader_state*)shader); +} static void i915_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, @@ -358,6 +373,10 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state; + i915->pipe.create_shader_state = i915_create_shader_state; + i915->pipe.bind_fs_state = i915_bind_fs_state; + i915->pipe.bind_vs_state = i915_bind_vs_state; + i915->pipe.delete_shader_state = i915_delete_shader_state; i915->pipe.set_alpha_test_state = i915_set_alpha_test_state; i915->pipe.set_blend_color = i915_set_blend_color; @@ -365,8 +384,6 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_clear_color_state = i915_set_clear_color_state; i915->pipe.set_constant_buffer = i915_set_constant_buffer; i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; - i915->pipe.set_fs_state = i915_set_fs_state; - i915->pipe.set_vs_state = i915_set_vs_state; i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; i915->pipe.set_scissor_state = i915_set_scissor_state; i915->pipe.set_texture_state = i915_set_texture_state; diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 504bc10a9e..8d404c55ab 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -42,7 +42,7 @@ */ static void calculate_vertex_layout( struct i915_context *i915 ) { - const uint inputsRead = i915->fs.inputs_read; + const uint inputsRead = i915->fs->inputs_read; const interp_mode colorInterp = i915->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &i915->current.vertex_info; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index dda758fe6a..c405051bce 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -117,6 +117,16 @@ struct pipe_context { void (*delete_depth_stencil_state)(struct pipe_context *, const struct pipe_depth_stencil_state *); + const struct pipe_shader_state * (*create_shader_state)( + struct pipe_context *, + const struct pipe_shader_state *); + void (*bind_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*bind_vs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*delete_shader_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*set_alpha_test_state)( struct pipe_context *, const struct pipe_alpha_test_state * ); @@ -139,12 +149,6 @@ struct pipe_context { void (*set_framebuffer_state)( struct pipe_context *, const struct pipe_framebuffer_state * ); - void (*set_fs_state)( struct pipe_context *, - const struct pipe_shader_state * ); - - void (*set_vs_state)( struct pipe_context *, - const struct pipe_shader_state * ); - void (*set_polygon_stipple)( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index cf5fc2227e..25cb9d8745 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -251,17 +251,21 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, /* state setters */ softpipe->pipe.create_blend_state = softpipe_create_blend_state; - softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; + softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; - softpipe->pipe.bind_sampler_state = softpipe_bind_sampler_state; + softpipe->pipe.bind_sampler_state = softpipe_bind_sampler_state; softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; softpipe->pipe.create_depth_stencil_state = softpipe_create_depth_stencil_state; - softpipe->pipe.bind_depth_stencil_state = softpipe_bind_depth_stencil_state; + softpipe->pipe.bind_depth_stencil_state = softpipe_bind_depth_stencil_state; softpipe->pipe.delete_depth_stencil_state = softpipe_delete_depth_stencil_state; softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; - softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; + softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; + softpipe->pipe.create_shader_state = softpipe_create_shader_state; + softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; + softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; + softpipe->pipe.delete_shader_state = softpipe_delete_shader_state; softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; @@ -270,8 +274,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; softpipe->pipe.set_feedback_state = softpipe_set_feedback_state; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; - softpipe->pipe.set_fs_state = softpipe_set_fs_state; - softpipe->pipe.set_vs_state = softpipe_set_vs_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; softpipe->pipe.set_texture_state = softpipe_set_texture_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index f1bb3d39a6..5c17c47b12 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -74,6 +74,8 @@ struct softpipe_context { const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; + const struct pipe_shader_state *fs; + const struct pipe_shader_state *vs; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -82,8 +84,6 @@ struct softpipe_context { struct pipe_constant_buffer constants[2]; struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; - struct pipe_shader_state fs; - struct pipe_shader_state vs; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 46ad08aaa1..25bc170d8c 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -108,7 +108,7 @@ shade_quad( /* init machine state */ tgsi_exec_machine_init( &machine, - softpipe->fs.tokens, + softpipe->fs->tokens, PIPE_MAX_SAMPLERS, qss->samplers ); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 62bd26c4df..04cc743bd0 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -88,11 +88,15 @@ void softpipe_set_constant_buffer(struct pipe_context *, void softpipe_set_feedback_state( struct pipe_context *, const struct pipe_feedback_state * ); -void softpipe_set_fs_state( struct pipe_context *, - const struct pipe_shader_state * ); - -void softpipe_set_vs_state( struct pipe_context *, - const struct pipe_shader_state * ); +const struct pipe_shader_state * +softpipe_create_shader_state( struct pipe_context *, + const struct pipe_shader_state * ); +void softpipe_bind_fs_state( struct pipe_context *, + const struct pipe_shader_state * ); +void softpipe_bind_vs_state( struct pipe_context *, + const struct pipe_shader_state * ); +void softpipe_delete_shader_state( struct pipe_context *, + const struct pipe_shader_state * ); void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 8c6bacf65c..9611a2ac99 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -43,7 +43,7 @@ */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - const uint inputsRead = softpipe->fs.inputs_read; + const uint inputsRead = softpipe->fs->inputs_read; const interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 5ab246896b..fbbde2f520 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -33,23 +33,33 @@ #include "pipe/draw/draw_context.h" -void softpipe_set_fs_state( struct pipe_context *pipe, +const struct pipe_shader_state * +softpipe_create_shader_state( struct pipe_context *pipe, + const struct pipe_shader_state *templ ) +{ + struct pipe_shader_state *shader = malloc(sizeof(struct pipe_shader_state)); + memcpy(shader, templ, sizeof(struct pipe_shader_state)); + + return shader; +} + +void softpipe_bind_fs_state( struct pipe_context *pipe, const struct pipe_shader_state *fs ) { struct softpipe_context *softpipe = softpipe_context(pipe); - memcpy(&softpipe->fs, fs, sizeof(*fs)); + softpipe->fs = fs; softpipe->dirty |= SP_NEW_FS; } -void softpipe_set_vs_state( struct pipe_context *pipe, +void softpipe_bind_vs_state( struct pipe_context *pipe, const struct pipe_shader_state *vs ) { struct softpipe_context *softpipe = softpipe_context(pipe); - memcpy(&softpipe->vs, vs, sizeof(*vs)); + softpipe->vs = vs; softpipe->dirty |= SP_NEW_VS; @@ -57,6 +67,12 @@ void softpipe_set_vs_state( struct pipe_context *pipe, } +void softpipe_delete_shader_state( struct pipe_context *pipe, + const struct pipe_shader_state *shader ) +{ + free((struct pipe_shader_state*)shader); +} + void softpipe_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, const struct pipe_constant_buffer *buf) diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index d066547616..dc3e5258d8 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -38,6 +38,7 @@ #include "pipe/tgsi/exec/tgsi_dump.h" #include "st_context.h" +#include "st_cache.h" #include "st_atom.h" #include "st_program.h" @@ -46,16 +47,21 @@ static void compile_fs( struct st_context *st ) { struct st_fragment_program *fp = st->fp; + struct pipe_shader_state fs; + struct pipe_shader_state *cached; /* XXX: fix static allocation of tokens: */ tgsi_mesa_compile_fp_program( &fp->Base, fp->tokens, ST_FP_MAX_TOKENS ); - fp->fs.inputs_read + memset(&fs, 0, sizeof(fs)); + fs.inputs_read = tgsi_mesa_translate_fragment_input_mask(fp->Base.Base.InputsRead); - fp->fs.outputs_written + fs.outputs_written = tgsi_mesa_translate_fragment_output_mask(fp->Base.Base.OutputsWritten); - fp->fs.tokens = &fp->tokens[0]; + fs.tokens = &fp->tokens[0]; + cached = st_cached_shader_state(st, &fs); + fp->fsx = cached; if (TGSI_DEBUG) tgsi_dump( fp->tokens, TGSI_DUMP_VERBOSE ); @@ -92,8 +98,8 @@ static void update_fs( struct st_context *st ) if (fp->dirty) compile_fs( st ); - st->state.fs = fp->fs; - st->pipe->set_fs_state(st->pipe, &st->state.fs); + st->state.fs = fp->fsx; + st->pipe->bind_fs_state(st->pipe, st->state.fs); } } diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 289a3e4f46..18be71367a 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -41,6 +41,7 @@ #include "pipe/tgsi/exec/tgsi_core.h" #include "st_context.h" +#include "st_cache.h" #include "st_atom.h" #include "st_program.h" @@ -55,16 +56,21 @@ static void compile_vs( struct st_context *st ) { struct st_vertex_program *vp = st->vp; - + struct pipe_shader_state vs; + struct pipe_shader_state *cached; /* XXX: fix static allocation of tokens: */ tgsi_mesa_compile_vp_program( &vp->Base, vp->tokens, ST_FP_MAX_TOKENS ); - vp->vs.inputs_read + memset(&vs, 0, sizeof(vs)); + vs.inputs_read = tgsi_mesa_translate_vertex_input_mask(vp->Base.Base.InputsRead); - vp->vs.outputs_written + vs.outputs_written = tgsi_mesa_translate_vertex_output_mask(vp->Base.Base.OutputsWritten); - vp->vs.tokens = &vp->tokens[0]; + vs.tokens = &vp->tokens[0]; + + cached = st_cached_shader_state(st, &vs); + vp->vs = cached; if (TGSI_DEBUG) tgsi_dump( vp->tokens, 0 ); @@ -110,7 +116,7 @@ static void update_vs( struct st_context *st ) #endif st->state.vs = st->vp->vs; - st->pipe->set_vs_state(st->pipe, &st->state.vs); + st->pipe->bind_vs_state(st->pipe, st->state.vs); } } diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index e9c79634bd..7b851e3901 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -111,3 +111,21 @@ struct pipe_rasterizer_state * st_cached_rasterizer_state( } return (struct pipe_rasterizer_state*)(cso_hash_iter_data(iter)); } + +struct pipe_shader_state * st_cached_shader_state( + struct st_context *st, + const struct pipe_shader_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_shader_state)); + struct cso_hash_iter iter = cso_find_state_template(st->cache, + hash_key, CSO_SHADER, + (void*)templ); + if (cso_hash_iter_is_null(iter)) { + const struct pipe_shader_state *created_state = + st->pipe->create_shader_state(st->pipe, templ); + iter = cso_insert_state(st->cache, hash_key, CSO_SHADER, + (void*)created_state); + } + return (struct pipe_shader_state*)(cso_hash_iter_data(iter)); +} diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index a06af31123..6a897a9993 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -53,4 +53,8 @@ struct pipe_rasterizer_state *st_cached_rasterizer_state( struct st_context *st, const struct pipe_rasterizer_state *raster); +struct pipe_shader_state *st_cached_shader_state( + struct st_context *st, + const struct pipe_shader_state *templ); + #endif diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 584bc1cc2a..2ea498663b 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -347,6 +347,7 @@ clear_with_quad(GLcontext *ctx, { static struct st_fragment_program *stfp = NULL; struct pipe_shader_state fs; + const struct pipe_shader_state *cached; if (!stfp) { stfp = make_color_shader(st); } @@ -354,13 +355,15 @@ clear_with_quad(GLcontext *ctx, fs.inputs_read = tgsi_mesa_translate_fragment_input_mask(stfp->Base.Base.InputsRead); fs.outputs_written = tgsi_mesa_translate_fragment_output_mask(stfp->Base.Base.OutputsWritten); fs.tokens = &stfp->tokens[0]; - pipe->set_fs_state(pipe, &fs); + cached = st_cached_shader_state(st, &fs); + pipe->bind_fs_state(pipe, cached); } /* vertex shader state: color/position pass-through */ { static struct st_vertex_program *stvp = NULL; struct pipe_shader_state vs; + const struct pipe_shader_state *cached; if (!stvp) { stvp = make_vertex_shader(st); } @@ -368,7 +371,8 @@ clear_with_quad(GLcontext *ctx, vs.inputs_read = stvp->Base.Base.InputsRead; vs.outputs_written = stvp->Base.Base.OutputsWritten; vs.tokens = &stvp->tokens[0]; - pipe->set_vs_state(pipe, &vs); + cached = st_cached_shader_state(st, &vs); + pipe->bind_vs_state(pipe, cached); } /* viewport state: viewport matching window dims */ @@ -394,8 +398,8 @@ clear_with_quad(GLcontext *ctx, pipe->set_alpha_test_state(pipe, &st->state.alpha_test); pipe->bind_blend_state(pipe, st->state.blend); pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil); - pipe->set_fs_state(pipe, &st->state.fs); - pipe->set_vs_state(pipe, &st->state.vs); + pipe->bind_fs_state(pipe, st->state.fs); + pipe->bind_vs_state(pipe, st->state.vs); pipe->bind_rasterizer_state(pipe, st->state.rasterizer); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); /* OR: diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 78ede8e225..37e40636f6 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -329,19 +329,22 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, { static struct st_fragment_program *stfp = NULL; struct pipe_shader_state fs; + struct pipe_shader_state *cached; if (!stfp) { stfp = make_fragment_shader(ctx->st); } memset(&fs, 0, sizeof(fs)); fs.inputs_read = stfp->Base.Base.InputsRead; fs.tokens = &stfp->tokens[0]; - pipe->set_fs_state(pipe, &fs); + cached = st_cached_shader_state(ctx->st, &fs); + pipe->bind_fs_state(pipe, cached); } /* vertex shader state: position + texcoord pass-through */ { static struct st_vertex_program *stvp = NULL; struct pipe_shader_state vs; + struct pipe_shader_state *cached; if (!stvp) { stvp = make_vertex_shader(ctx->st); } @@ -349,7 +352,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, vs.inputs_read = stvp->Base.Base.InputsRead; vs.outputs_written = stvp->Base.Base.OutputsWritten; vs.tokens = &stvp->tokens[0]; - pipe->set_vs_state(pipe, &vs); + cached = st_cached_shader_state(ctx->st, &vs); + pipe->bind_vs_state(pipe, cached); } /* texture sampling state: */ @@ -403,8 +407,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* restore GL state */ pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer); - pipe->set_fs_state(pipe, &ctx->st->state.fs); - pipe->set_vs_state(pipe, &ctx->st->state.vs); + pipe->bind_fs_state(pipe, ctx->st->state.fs); + pipe->bind_vs_state(pipe, ctx->st->state.vs); pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 7bedf3f89f..98efe1a10b 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -53,7 +53,7 @@ static void setup_vertex_attribs(GLcontext *ctx) { struct pipe_context *pipe = ctx->st->pipe; - const uint inputAttrs = ctx->st->state.vs.inputs_read; + const uint inputAttrs = ctx->st->state.vs->inputs_read; uint attr; /* all attributes come from the default attribute buffer */ @@ -84,7 +84,7 @@ static void setup_feedback(GLcontext *ctx) { struct pipe_context *pipe = ctx->st->pipe; - const uint outputAttrs = ctx->st->state.vs.outputs_written; + const uint outputAttrs = ctx->st->state.vs->outputs_written; struct pipe_feedback_state feedback; uint i; @@ -307,7 +307,7 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) /* extract values and update rasterpos state */ { - const uint outputAttrs = ctx->st->state.vs.outputs_written; + const uint outputAttrs = ctx->st->state.vs->outputs_written; const float *pos, *color0, *color1, *tex0; float *buf = buf_map; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 516d319a6e..a8ae5d9c56 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -78,6 +78,8 @@ struct st_context const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; + const struct pipe_shader_state *fs; + const struct pipe_shader_state *vs; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; @@ -89,8 +91,6 @@ struct st_context struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_shader_state fs; - struct pipe_shader_state vs; struct pipe_viewport_state viewport; } state; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index f68e449f07..ac63a38720 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -193,7 +193,7 @@ st_draw_vbo(GLcontext *ctx, update_default_attribs_buffer(ctx); /* this must be after state validation */ - attrsNeeded = ctx->st->state.vs.inputs_read; + attrsNeeded = ctx->st->state.vs->inputs_read; /* tell pipe about the vertex array element/attributes */ for (attr = 0; attr < 16; attr++) { @@ -395,14 +395,14 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); draw_set_rasterizer_state(draw, st->state.rasterizer); - draw_set_vertex_shader(draw, &st->state.vs); + draw_set_vertex_shader(draw, st->state.vs); /* XXX need to set vertex info too */ update_default_attribs_buffer(ctx); /* this must be after state validation */ - attrsNeeded = ctx->st->state.vs.inputs_read; + attrsNeeded = ctx->st->state.vs->inputs_read; /* tell draw module about the vertex array element/attributes */ for (attr = 0; attr < 16; attr++) { diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 7a91983ce9..a2f114b1ba 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -53,8 +53,8 @@ struct st_fragment_program struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; - - struct pipe_shader_state fs; + + const struct pipe_shader_state *fsx; GLuint param_state; }; @@ -75,7 +75,7 @@ struct st_vertex_program struct x86_function sse2_program; #endif - struct pipe_shader_state vs; + const struct pipe_shader_state *vs; GLuint param_state; }; -- cgit v1.2.3 From 72b0a57cfa99e55ad6e1b4ddcd486c202ee1f22d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Sep 2007 12:50:20 -0600 Subject: additional comments for tri_persp_coeff() --- src/mesa/pipe/softpipe/sp_prim_setup.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index c64a4e9708..d0baf0734b 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -428,6 +428,10 @@ static void tri_linear_coeff( struct setup_stage *setup, /** * Compute a0, dadx and dady for a perspective-corrected interpolant, * for a triangle. + * We basically multiply the vertex value by 1/w before computing + * the plane coefficients (a0, dadx, dady). + * Later, when we compute the value at a particular fragment position we'll + * divide the interpolated value by the interpolated W at that fragment. */ static void tri_persp_coeff( struct setup_stage *setup, unsigned slot, -- cgit v1.2.3 From 2a460f6185199bad8b2bca4d0bac319377f801fa Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Sep 2007 12:55:20 -0600 Subject: Fix some issues with perspective-corrected interpolation. In mesa_to_tgsi.c, use TGSI_INTERPOLATE_PERSPECTIVE by default (to match post-transform vertex info convention). More to be done there... In sp_quad_fs.c, interpolate W in addition to Z. This fixes the divide by zero happening in perspective_interpolation() tgsi_exec.c As it was, we were only getting perspective correction of texture coords used by the TGSI_TEX instruction since it does a homogeneous divide. Other coords/varyings were incorrect. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 9 ++++----- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 1 + src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 +- 3 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 25bc170d8c..7d2712a537 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -68,10 +68,9 @@ quad_shade_stage(struct quad_stage *qs) * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy */ static INLINE void -linterp_z(const struct tgsi_interp_coef *coef, - struct tgsi_exec_vector *pos) +linterp(const struct tgsi_interp_coef *coef, + struct tgsi_exec_vector *pos, uint ch) { - uint ch = 2; uint j; for (j = 0; j < QUAD_SIZE; j++) { const float x = pos->xyzw[0].f[j]; @@ -83,7 +82,6 @@ linterp_z(const struct tgsi_interp_coef *coef, } - /* This should be done by the fragment shader execution unit (code * generated from the decl instructions). Do it here for now. */ @@ -131,7 +129,8 @@ shade_quad( machine.Inputs[0].xyzw[1].f[3] = fy + 1.0f; /* interp Z */ - linterp_z(&quad->coef[0], &machine.Inputs[0]); + linterp(&quad->coef[0], &machine.Inputs[0], 2); /* Z */ + linterp(&quad->coef[0], &machine.Inputs[0], 3); /* 1/W */ /* run shader */ tgsi_exec_machine_run( &machine ); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 793f8bc0f6..28dc049658 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1104,6 +1104,7 @@ perspective_interpolation( const float y = mach->Inputs[0].xyzw[1].f[i]; // WPOS.w here is really 1/w const float w = 1.0f / mach->Inputs[0].xyzw[3].f[i]; + assert(mach->Inputs[0].xyzw[3].f[i] != 0.0); mach->Inputs[attrib].xyzw[chan].f[i] = (mach->InterpCoefs[attrib].a0[chan] + diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 182a7eff05..2820706537 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -859,7 +859,7 @@ tgsi_mesa_compile_fp_program( fulldecl = make_frag_input_decl( 1, 1 + count - 1, - TGSI_INTERPOLATE_LINEAR, + TGSI_INTERPOLATE_PERSPECTIVE, TGSI_WRITEMASK_XYZW ); ti += tgsi_build_full_declaration( &fulldecl, -- cgit v1.2.3 From d39dde6ed4ab1927d5b9992d6cc7414e7fa236e8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 18 Sep 2007 21:49:43 +0100 Subject: Fix linux-dri-x86 target. --- src/mesa/state_tracker/st_atom_vs.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 18be71367a..8f1df80a7f 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -77,8 +77,9 @@ static void compile_vs( struct st_context *st ) #if defined(USE_X86_ASM) || defined(SLANG_X86) tgsi_emit_sse2( - vp->vs.tokens, + vp->tokens, &vp->sse2_program ); + cached->executable = (void *) x86_get_func( &vp->sse2_program ); #endif vp->dirty = 0; @@ -111,10 +112,6 @@ static void update_vs( struct st_context *st ) if (vp->dirty) compile_vs( st ); -#if defined(USE_X86_ASM) || defined(SLANG_X86) - st->vp->vs.executable = (void *) x86_get_func( &vp->sse2_program ); -#endif - st->state.vs = st->vp->vs; st->pipe->bind_vs_state(st->pipe, st->state.vs); } -- cgit v1.2.3 From 3f0e04e0e62b853382a018cb6a69eea14e85712d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 18 Sep 2007 21:50:31 +0100 Subject: intel_winsys: Disable scheduled buffer swaps for now. They don't work with private back buffers yet. This gets vsync working. --- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index d9adcd3a62..48c1f68d9c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -496,7 +496,8 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) drm_i915_vblank_swap_t swap; GLboolean ret; - if (!intel_fb->vblank_flags || + /* XXX: Scheduled buffer swaps don't work with private back buffers yet */ + if (1 || !intel_fb->vblank_flags || (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) return GL_FALSE; -- cgit v1.2.3 From 63be96bdc7e9f388a5c49295bd7e150462fd003a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Sep 2007 19:29:26 -0600 Subject: temporarily set the FRAG_BIT_FOGC bit in InputsRead when fog is enabled --- src/mesa/main/texenvprogram.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index 72b54b27d9..37f8fc8090 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -1081,6 +1081,7 @@ create_new_program(GLcontext *ctx, struct state_key *key, * a reduced value and not what is expected in FogOption */ p.program->FogOption = ctx->Fog.Mode; + p.program->Base.InputsRead |= FRAG_BIT_FOGC; /* XXX new */ } else p.program->FogOption = GL_NONE; -- cgit v1.2.3 From bb611c5f1f6aec7ac51d4fa3301422b47f6de795 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Sep 2007 19:37:36 -0600 Subject: Checkpoint: rework shader input/output register mapping. This is a step toward removing TGSI_ATTRIB_ tokens. Basically, when translating Mesa programs to TGSI programs, pass in input and output register re-maps, plus interpolation info. There's some known breakage (cubemap.c) so more to be done... --- src/mesa/pipe/draw/draw_vertex_fetch.c | 39 +++++------ src/mesa/pipe/draw/draw_vertex_shader.c | 14 +++- src/mesa/pipe/p_state.h | 3 + src/mesa/pipe/softpipe/sp_prim_setup.c | 8 +++ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 72 +++++++++++++++------ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 5 ++ src/mesa/state_tracker/st_atom_fs.c | 23 ++++++- src/mesa/state_tracker/st_atom_vs.c | 46 +++++++++++-- src/mesa/state_tracker/st_cb_clear.c | 21 +++++- src/mesa/state_tracker/st_cb_drawpixels.c | 15 ++++- src/mesa/state_tracker/st_draw.c | 103 ++++++++++++++++-------------- src/mesa/state_tracker/st_program.h | 17 +++-- 12 files changed, 260 insertions(+), 106 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 0dbbdf17f2..ce402d681f 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -67,6 +67,10 @@ fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) } } + +/** + * Fetch vertex attributes for 'count' vertices. + */ void draw_vertex_fetch( struct draw_context *draw, struct tgsi_exec_machine *machine, const unsigned *elts, @@ -74,27 +78,26 @@ void draw_vertex_fetch( struct draw_context *draw, { unsigned j; - - /* load machine inputs */ + /* loop over vertices */ for (j = 0; j < count; j++) { - unsigned attr; - for (attr = 0; attr < 16; attr++) { - if (draw->vertex_shader.inputs_read & (1 << attr)) { - unsigned buf = draw->vertex_element[attr].vertex_buffer_index; - const void *src - = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] - + draw->vertex_buffer[buf].buffer_offset - + draw->vertex_element[attr].src_offset - + elts[j] * draw->vertex_buffer[buf].pitch); - float p[4]; + uint attr; + /* loop over vertex attributes (vertex shader inputs) */ + for (attr = 0; attr < draw->vertex_shader.num_inputs; attr++) { + + unsigned buf = draw->vertex_element[attr].vertex_buffer_index; + const void *src + = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[attr].src_offset + + elts[j] * draw->vertex_buffer[buf].pitch); + float p[4]; - fetch_attrib4(src, draw->vertex_element[attr].src_format, p); + fetch_attrib4(src, draw->vertex_element[attr].src_format, p); - machine->Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ - machine->Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ - machine->Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ - machine->Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ - } + machine->Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ + machine->Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ + machine->Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ + machine->Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ } } } diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 8effc74cbe..cb6c605b8d 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -114,7 +114,6 @@ run_vertex_program(struct draw_context *draw, draw_vertex_fetch( draw, &machine, elts, count ); - /* run shader */ if( draw->vertex_shader.executable != NULL ) { #if defined(USE_X86_ASM) || defined(SLANG_X86) @@ -159,14 +158,23 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][2] = z * scale[2] + trans[2]; vOut[j]->data[0][3] = w; - /* remaining attributes are packed into sequential post-transform + /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. + * Skip 0 since we just did it above. + * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. */ - for (slot = 1; slot < draw->vertex_info.num_attribs; slot++) { + for (slot = 1; slot < draw->vertex_info.num_attribs - 2; slot++) { vOut[j]->data[slot][0] = machine.Outputs[slot].xyzw[0].f[j]; vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j]; vOut[j]->data[slot][2] = machine.Outputs[slot].xyzw[2].f[j]; vOut[j]->data[slot][3] = machine.Outputs[slot].xyzw[3].f[j]; + /* + printf("output %d: %f %f %f %f\n", slot, + vOut[j]->data[slot][0], + vOut[j]->data[slot][1], + vOut[j]->data[slot][2], + vOut[j]->data[slot][3]); + */ } } /* loop over vertices */ } diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 048feede3b..d2cc76a59b 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -144,6 +144,9 @@ struct pipe_shader_state { unsigned outputs_written; /**< TGSI_ATTRIB_ bits */ const struct tgsi_token *tokens; void *executable; + + uint num_inputs; + uint num_outputs; }; struct pipe_depth_stencil_state diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index d0baf0734b..8a4be79d11 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -448,6 +448,14 @@ static void tri_persp_coeff( struct setup_stage *setup, float a = setup->ebot.dy * majda - botda * setup->emaj.dy; float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + /* + printf("tri persp %d,%d: %f %f %f\n", slot, i, + setup->vmin->data[slot][i], + setup->vmid->data[slot][i], + setup->vmax->data[slot][i] + ); + */ + assert(slot < TGSI_ATTRIB_MAX); assert(i <= 3); diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 2820706537..1f8d937bc6 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -211,6 +211,7 @@ tgsi_mesa_translate_fragment_output(GLuint attrib) } +#if 01 uint tgsi_mesa_translate_vertex_input_mask(GLbitfield mask) { @@ -224,7 +225,7 @@ tgsi_mesa_translate_vertex_input_mask(GLbitfield mask) } return tgsiMask; } - +#endif uint tgsi_mesa_translate_vertex_output_mask(GLbitfield mask) @@ -318,7 +319,9 @@ map_register_file_index( GLuint processor, GLuint file, GLuint index, - GLbitfield usage_bitmask ) + GLbitfield usage_bitmask, + const GLuint inputMapping[], + const GLuint outputMapping[]) { GLuint mapped_index; GLuint i; @@ -337,6 +340,12 @@ map_register_file_index( * etc. */ assert( index < 32 ); + if (inputMapping) { + printf("New map %d input %d to %d\n", processor, index, + inputMapping[index]); + return inputMapping[index]; + } + assert( usage_bitmask & (1 << index) ); mapped_index = 0; for( i = 0; i < index; i++ ) { @@ -344,7 +353,7 @@ map_register_file_index( mapped_index++; } } - printf("Map input %d to %d\n", index, mapped_index); + printf("Map %d input %d to %d\n", processor, index, mapped_index); break; case TGSI_FILE_OUTPUT: @@ -372,6 +381,8 @@ map_register_file_index( mapped_index++; } } + printf("Map VP output from %d to %d\n", index, mapped_index); + assert(outputMapping[index] == mapped_index); } break; @@ -443,6 +454,8 @@ compile_instruction( struct tgsi_full_instruction *fullinst, GLuint inputs_read, GLuint outputs_written, + const GLuint inputMapping[], + const GLuint outputMapping[], GLuint preamble_size, GLuint processor ) { @@ -462,7 +475,9 @@ compile_instruction( processor, fulldst->DstRegister.File, inst->DstReg.Index, - outputs_written + outputs_written, + NULL, + outputMapping ); fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); @@ -475,7 +490,9 @@ compile_instruction( processor, fullsrc->SrcRegister.File, inst->SrcReg[i].Index, - inputs_read ); + inputs_read, + inputMapping, + outputMapping ); for( j = 0; j < 4; j++ ) { GLuint swz; @@ -789,9 +806,20 @@ make_frag_output_decl( return decl; } + +/** + * Convert Mesa fragment program to TGSI format. + * \param inputMapping array to map original Mesa fragment program inputs + * registers to TGSI generic input indexes + * \param interpMode array[FRAG_ATTRIB_x] of TGSI_INTERPOLATE_LINEAR/PERSP. + * + */ GLboolean tgsi_mesa_compile_fp_program( const struct gl_fragment_program *program, + const GLuint inputMapping[], + const GLuint interpMode[], + const GLuint outputMapping[], struct tgsi_token *tokens, GLuint maxTokens ) { @@ -800,8 +828,10 @@ tgsi_mesa_compile_fp_program( struct tgsi_processor *processor; struct tgsi_full_declaration fulldecl; struct tgsi_full_instruction fullinst; + /* struct tgsi_full_dst_register *fulldst; struct tgsi_full_src_register *fullsrc; + */ GLuint inputs_read; GLboolean reads_wpos; GLuint preamble_size = 0; @@ -822,7 +852,7 @@ tgsi_mesa_compile_fp_program( /* * Declare input attributes. Note that we do not interpolate fragment position. */ - + reads_wpos = 1; /* Fragment position. */ if( reads_wpos ) { fulldecl = make_frag_input_decl( @@ -853,20 +883,16 @@ tgsi_mesa_compile_fp_program( for( i = 1; i < 32; i++ ) { if( inputs_read & (1 << i) ) { count++; + fulldecl = make_frag_input_decl(count, + count, + interpMode[i], + TGSI_WRITEMASK_XYZW ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); } - } - if( count > 0 ) { - fulldecl = make_frag_input_decl( - 1, - 1 + count - 1, - TGSI_INTERPOLATE_PERSPECTIVE, - TGSI_WRITEMASK_XYZW ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } + } /* * Declare output attributes. @@ -932,6 +958,8 @@ tgsi_mesa_compile_fp_program( &fullinst, inputs_read, ~0, /*outputs_written*/ + inputMapping, + outputMapping, preamble_size, TGSI_PROCESSOR_FRAGMENT ) ) { assert( i == program->Base.NumInstructions - 1 ); @@ -955,8 +983,10 @@ tgsi_mesa_compile_fp_program( GLboolean tgsi_mesa_compile_vp_program( const struct gl_vertex_program *program, + const GLuint inputMapping[], + const GLuint outputMapping[], struct tgsi_token *tokens, - GLuint maxTokens ) + GLuint maxTokens) { GLuint i, ti; struct tgsi_header *header; @@ -983,6 +1013,8 @@ tgsi_mesa_compile_vp_program( &fullinst, inputs_read, outputs_written, + inputMapping, + outputMapping, 0, TGSI_PROCESSOR_VERTEX ) ) { assert( i == program->Base.NumInstructions - 1 ); diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index fda3fa397f..017cfce72e 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -10,12 +10,17 @@ struct tgsi_token; GLboolean tgsi_mesa_compile_fp_program( const struct gl_fragment_program *program, + const GLuint inputMapping[], + const GLuint interpMode[], + const GLuint outputMapping[], struct tgsi_token *tokens, GLuint maxTokens ); GLboolean tgsi_mesa_compile_vp_program( const struct gl_vertex_program *program, + const GLuint inputMapping[], + const GLuint outputMapping[], struct tgsi_token *tokens, GLuint maxTokens ); diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index dc3e5258d8..3df2c6750a 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -42,17 +42,34 @@ #include "st_atom.h" #include "st_program.h" -#define TGSI_DEBUG 0 +#define TGSI_DEBUG 1 static void compile_fs( struct st_context *st ) { + /* Map FRAG_RESULT_COLR to output 1, map FRAG_RESULT_DEPR to output 0 */ + static const GLuint outputMapping[2] = {1, 0}; struct st_fragment_program *fp = st->fp; struct pipe_shader_state fs; struct pipe_shader_state *cached; + GLuint interpMode[16]; /* XXX size? */ + GLuint i; + + for (i = 0; i < 16; i++) { + if (fp->Base.Base.InputsRead & (1 << i)) { + if (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1) { + interpMode[i] = TGSI_INTERPOLATE_LINEAR; + } + else { + interpMode[i] = TGSI_INTERPOLATE_PERSPECTIVE; + } + } + } /* XXX: fix static allocation of tokens: */ - tgsi_mesa_compile_fp_program( &fp->Base, fp->tokens, ST_FP_MAX_TOKENS ); + tgsi_mesa_compile_fp_program( &fp->Base, NULL, interpMode, + outputMapping, + fp->tokens, ST_FP_MAX_TOKENS ); memset(&fs, 0, sizeof(fs)); fs.inputs_read @@ -64,7 +81,7 @@ static void compile_fs( struct st_context *st ) fp->fsx = cached; if (TGSI_DEBUG) - tgsi_dump( fp->tokens, TGSI_DUMP_VERBOSE ); + tgsi_dump( fp->tokens, 0/*TGSI_DUMP_VERBOSE*/ ); fp->dirty = 0; } diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 8f1df80a7f..8de19e41ee 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -51,25 +51,61 @@ -/* translate shader to TGSI format -*/ +/** + * Translate Mesa shader to TGSI format + */ static void compile_vs( struct st_context *st ) { struct st_vertex_program *vp = st->vp; struct pipe_shader_state vs; struct pipe_shader_state *cached; + GLuint i; + + memset(&vs, 0, sizeof(vs)); + + /* + * Determine how many inputs there are. + * Also, compute two look-up tables that map between Mesa VERT_ATTRIB_x + * values and TGSI generic input indexes. + */ + for (i = 0; i < MAX_VERTEX_PROGRAM_ATTRIBS; i++) { + if (vp->Base.Base.InputsRead & (1 << i)) { + vp->input_to_index[i] = vs.num_inputs; + vp->index_to_input[vs.num_inputs] = i; + vs.num_inputs++; + } + } + + /* + * Determine output register mapping. + */ + for (i = 0; i < VERT_RESULT_MAX; i++) { + if (vp->Base.Base.OutputsWritten & (1 << i)) { + vp->output_to_index[i] = vs.num_outputs; + vp->index_to_output[vs.num_outputs] = i; + vs.num_outputs++; + } + } + + /* XXX: fix static allocation of tokens: */ - tgsi_mesa_compile_vp_program( &vp->Base, vp->tokens, ST_FP_MAX_TOKENS ); + tgsi_mesa_compile_vp_program( &vp->Base, + vp->input_to_index, + vp->output_to_index, + vp->tokens, ST_FP_MAX_TOKENS ); - memset(&vs, 0, sizeof(vs)); +#if 01 vs.inputs_read = tgsi_mesa_translate_vertex_input_mask(vp->Base.Base.InputsRead); +#endif vs.outputs_written = tgsi_mesa_translate_vertex_output_mask(vp->Base.Base.OutputsWritten); + vs.tokens = &vp->tokens[0]; cached = st_cached_shader_state(st, &vs); + vp->vs = cached; if (TGSI_DEBUG) @@ -121,7 +157,7 @@ static void update_vs( struct st_context *st ) const struct st_tracked_state st_update_vs = { .name = "st_update_vs", .dirty = { - .mesa = 0, + .mesa = _NEW_PROGRAM, .st = ST_NEW_VERTEX_PROGRAM, }, .update = update_vs diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 2ea498663b..65cac9dbde 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -119,12 +119,19 @@ is_depth_stencil_format(GLuint pipeFormat) * Create a simple fragment shader that just passes through the fragment color. */ static struct st_fragment_program * -make_color_shader(struct st_context *st) +make_frag_shader(struct st_context *st) { + static const GLuint outputMapping[] = { 1, 0 }; GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; struct gl_program *p; GLboolean b; + GLuint interpMode[16]; + GLuint i; + + /* XXX temporary */ + for (i = 0; i < 16; i++) + interpMode[i] = TGSI_INTERPOLATE_LINEAR; p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) @@ -151,7 +158,8 @@ make_color_shader(struct st_context *st) stfp = (struct st_fragment_program *) p; /* compile into tgsi format */ - b = tgsi_mesa_compile_fp_program(&stfp->Base, + b = tgsi_mesa_compile_fp_program(&stfp->Base, NULL, interpMode, + outputMapping, stfp->tokens, ST_FP_MAX_TOKENS); assert(b); @@ -166,6 +174,11 @@ make_color_shader(struct st_context *st) static struct st_vertex_program * make_vertex_shader(struct st_context *st) { + /* Map VERT_ATTRIB_POS to 0, VERT_ATTRIB_COLOR0 to 1 */ + static const GLuint inputMapping[4] = { 0, 0, 0, 1 }; + /* Map VERT_RESULT_HPOS to 0, VERT_RESULT_COL0 to 1 */ + static const GLuint outputMapping[2] = { 0, 1 }; + GLcontext *ctx = st->ctx; struct st_vertex_program *stvp; struct gl_program *p; @@ -204,6 +217,8 @@ make_vertex_shader(struct st_context *st) stvp = (struct st_vertex_program *) p; /* compile into tgsi format */ b = tgsi_mesa_compile_vp_program(&stvp->Base, + inputMapping, + outputMapping, stvp->tokens, ST_FP_MAX_TOKENS); assert(b); @@ -349,7 +364,7 @@ clear_with_quad(GLcontext *ctx, struct pipe_shader_state fs; const struct pipe_shader_state *cached; if (!stfp) { - stfp = make_color_shader(st); + stfp = make_frag_shader(st); } memset(&fs, 0, sizeof(fs)); fs.inputs_read = tgsi_mesa_translate_fragment_input_mask(stfp->Base.Base.InputsRead); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 37e40636f6..731c060c11 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -56,10 +56,17 @@ static struct st_fragment_program * make_fragment_shader(struct st_context *st) { + static const GLuint outputMapping[2] = { 1, 0 }; GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; struct gl_program *p; GLboolean b; + GLuint interpMode[16]; + GLuint i; + + /* XXX temporary */ + for (i = 0; i < 16; i++) + interpMode[i] = TGSI_INTERPOLATE_LINEAR; p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) @@ -88,7 +95,8 @@ make_fragment_shader(struct st_context *st) stfp = (struct st_fragment_program *) p; /* compile into tgsi format */ - b = tgsi_mesa_compile_fp_program(&stfp->Base, + b = tgsi_mesa_compile_fp_program(&stfp->Base, NULL, interpMode, + outputMapping, stfp->tokens, ST_FP_MAX_TOKENS); assert(b); @@ -103,6 +111,8 @@ make_fragment_shader(struct st_context *st) static struct st_vertex_program * make_vertex_shader(struct st_context *st) { + /* Map VERT_RESULT_HPOS to 0, VERT_RESULT_TEX0 to 1 */ + static const GLuint outputMapping[] = { 0, 0, 0, 0, 1 }; GLcontext *ctx = st->ctx; struct st_vertex_program *stvp; struct gl_program *p; @@ -140,7 +150,8 @@ make_vertex_shader(struct st_context *st) stvp = (struct st_vertex_program *) p; /* compile into tgsi format */ - b = tgsi_mesa_compile_vp_program(&stvp->Base, + b = tgsi_mesa_compile_vp_program(&stvp->Base, NULL, + outputMapping, stvp->tokens, ST_FP_MAX_TOKENS); assert(b); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index ac63a38720..bff0ad7ef7 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -37,10 +37,12 @@ #include "tnl/t_vp_build.h" -#include "st_context.h" #include "st_atom.h" -#include "st_draw.h" +#include "st_context.h" #include "st_cb_bufferobjects.h" +#include "st_draw.h" +#include "st_program.h" + #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" @@ -185,18 +187,22 @@ st_draw_vbo(GLcontext *ctx, GLuint max_index) { struct pipe_context *pipe = ctx->st->pipe; - GLuint attr, i; - GLbitfield attrsNeeded; + const struct st_vertex_program *vp = ctx->st->vp; + const struct pipe_shader_state *vs; const unsigned attr0_offset = (unsigned) arrays[0]->Ptr; + GLboolean needDefaultAttribs = GL_FALSE; + GLuint attr; st_validate_state(ctx->st); - update_default_attribs_buffer(ctx); - /* this must be after state validation */ - attrsNeeded = ctx->st->state.vs->inputs_read; + /* must do this after state validation! */ + vs = ctx->st->state.vs; + + /* loop over TGSI shader inputs */ + for (attr = 0; attr < vs->num_inputs; attr++) { + const GLuint mesaAttr = vp->index_to_input[attr]; + struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; - /* tell pipe about the vertex array element/attributes */ - for (attr = 0; attr < 16; attr++) { struct pipe_vertex_buffer vbuffer; struct pipe_vertex_element velement; @@ -206,43 +212,40 @@ st_draw_vbo(GLcontext *ctx, velement.vertex_buffer_index = 0; velement.src_format = 0; - if (attrsNeeded & (1 << attr)) { - const GLuint mesaAttr = tgsi_attrib_to_mesa_attrib(attr); - struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; - - if (bufobj && bufobj->Name) { - struct st_buffer_object *stobj = st_buffer_object(bufobj); - /* Recall that for VBOs, the gl_client_array->Ptr field is - * really an offset from the start of the VBO, not a pointer. - */ - unsigned offset = (unsigned) arrays[mesaAttr]->Ptr; - - assert(stobj->buffer); - - vbuffer.buffer = stobj->buffer; - vbuffer.buffer_offset = attr0_offset; /* in bytes */ - vbuffer.pitch = arrays[mesaAttr]->StrideB; /* in bytes */ - vbuffer.max_index = 0; /* need this? */ - - velement.src_offset = offset - attr0_offset; /* bytes */ - velement.vertex_buffer_index = attr; - velement.dst_offset = 0; /* need this? */ - velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, - arrays[mesaAttr]->Size); - assert(velement.src_format); - } - else { - /* use the default attribute buffer */ - vbuffer.buffer = ctx->st->default_attrib_buffer; - vbuffer.buffer_offset = 0; - vbuffer.pitch = 0; /* must be zero! */ - vbuffer.max_index = 1; - - velement.src_offset = attr * 4 * sizeof(GLfloat); - velement.vertex_buffer_index = attr; - velement.dst_offset = 0; - velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - } + if (bufobj && bufobj->Name) { + struct st_buffer_object *stobj = st_buffer_object(bufobj); + /* Recall that for VBOs, the gl_client_array->Ptr field is + * really an offset from the start of the VBO, not a pointer. + */ + unsigned offset = (unsigned) arrays[mesaAttr]->Ptr; + + assert(stobj->buffer); + + vbuffer.buffer = stobj->buffer; + vbuffer.buffer_offset = attr0_offset; /* in bytes */ + vbuffer.pitch = arrays[mesaAttr]->StrideB; /* in bytes */ + vbuffer.max_index = 0; /* need this? */ + + velement.src_offset = offset - attr0_offset; /* bytes */ + velement.vertex_buffer_index = attr; + velement.dst_offset = 0; /* need this? */ + velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, + arrays[mesaAttr]->Size); + assert(velement.src_format); + } + else { + /* use the default attribute buffer */ + needDefaultAttribs = GL_TRUE; + + vbuffer.buffer = ctx->st->default_attrib_buffer; + vbuffer.buffer_offset = 0; + vbuffer.pitch = 0; /* must be zero! */ + vbuffer.max_index = 1; + + velement.src_offset = mesaAttr * 4 * sizeof(GLfloat); + velement.vertex_buffer_index = attr; + velement.dst_offset = 0; + velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; } if (attr == 0) @@ -252,12 +255,17 @@ st_draw_vbo(GLcontext *ctx, pipe->set_vertex_element(pipe, attr, &velement); } + if (needDefaultAttribs) { + update_default_attribs_buffer(ctx); + } + + /* do actual drawing */ if (ib) { /* indexed primitive */ struct gl_buffer_object *bufobj = ib->obj; struct pipe_buffer_handle *bh = NULL; - unsigned indexSize; + unsigned indexSize, i; if (bufobj && bufobj->Name) { /* elements/indexes are in a real VBO */ @@ -285,6 +293,7 @@ st_draw_vbo(GLcontext *ctx, } else { /* non-indexed */ + GLuint i; for (i = 0; i < nr_prims; i++) { pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count); } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index a2f114b1ba..68ceba4d78 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -61,13 +61,20 @@ struct st_fragment_program struct st_vertex_program { - struct gl_vertex_program Base; - GLboolean error; /* If program is malformed for any reason. */ + struct gl_vertex_program Base; /**< The Mesa vertex program */ + GLboolean error; /**< Set if program is malformed for any reason. */ - GLuint id; /* String id, for tracking - * ProgramStringNotify changes. - */ + GLuint id; /**< String id, for tracking ProgramStringNotify changes. */ + + /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ + GLuint input_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; + /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ + GLuint index_to_input[MAX_VERTEX_PROGRAM_ATTRIBS]; + + GLuint output_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; + GLuint index_to_output[MAX_VERTEX_PROGRAM_ATTRIBS]; + /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; -- cgit v1.2.3 From f22e920f478d8732695913ec0d1f7244b451a8f5 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Sep 2007 06:46:32 -0400 Subject: Finish up conversions of shaders to immutable objects. Create/Delete calls should be split since in create we'll be compiling them so we want to know which one it is (vertex/fragment). --- src/mesa/pipe/cso_cache/cso_cache.c | 16 +++++++++++----- src/mesa/pipe/cso_cache/cso_cache.h | 6 ++++-- src/mesa/pipe/i915simple/i915_state.c | 6 ++++-- src/mesa/pipe/p_context.h | 11 ++++++++--- src/mesa/pipe/softpipe/sp_context.c | 10 ++++++---- src/mesa/state_tracker/st_atom_fs.c | 2 +- src/mesa/state_tracker/st_atom_vs.c | 2 +- src/mesa/state_tracker/st_cache.c | 26 ++++++++++++++++++++++---- src/mesa/state_tracker/st_cache.h | 7 ++++++- src/mesa/state_tracker/st_cb_clear.c | 4 ++-- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- 11 files changed, 67 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_cache.c b/src/mesa/pipe/cso_cache/cso_cache.c index be653d9494..e87733c7ab 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.c +++ b/src/mesa/pipe/cso_cache/cso_cache.c @@ -80,8 +80,10 @@ static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_ hash = sc->depth_stencil_hash; case CSO_RASTERIZER: hash = sc->rasterizer_hash; - case CSO_SHADER: - hash = sc->shader_hash; + case CSO_FRAGMENT_SHADER: + hash = sc->fs_hash; + case CSO_VERTEX_SHADER: + hash = sc->vs_hash; } return hash; @@ -98,7 +100,9 @@ static int _cso_size_for_type(enum cso_cache_type type) return sizeof(struct pipe_depth_stencil_state); case CSO_RASTERIZER: return sizeof(struct pipe_rasterizer_state); - case CSO_SHADER: + case CSO_FRAGMENT_SHADER: + return sizeof(struct pipe_shader_state); + case CSO_VERTEX_SHADER: return sizeof(struct pipe_shader_state); } return 0; @@ -152,7 +156,8 @@ struct cso_cache *cso_cache_create(void) sc->sampler_hash = cso_hash_create(); sc->depth_stencil_hash = cso_hash_create(); sc->rasterizer_hash = cso_hash_create(); - sc->shader_hash = cso_hash_create(); + sc->fs_hash = cso_hash_create(); + sc->vs_hash = cso_hash_create(); return sc; } @@ -164,6 +169,7 @@ void cso_cache_delete(struct cso_cache *sc) cso_hash_delete(sc->sampler_hash); cso_hash_delete(sc->depth_stencil_hash); cso_hash_delete(sc->rasterizer_hash); - cso_hash_delete(sc->shader_hash); + cso_hash_delete(sc->fs_hash); + cso_hash_delete(sc->vs_hash); free(sc); } diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h index d9793ca855..352e1a6a59 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -44,7 +44,8 @@ struct cso_cache { struct cso_hash *sampler_hash; struct cso_hash *depth_stencil_hash; struct cso_hash *rasterizer_hash; - struct cso_hash *shader_hash; + struct cso_hash *fs_hash; + struct cso_hash *vs_hash; }; enum cso_cache_type { @@ -52,7 +53,8 @@ enum cso_cache_type { CSO_SAMPLER, CSO_DEPTH_STENCIL, CSO_RASTERIZER, - CSO_SHADER + CSO_FRAGMENT_SHADER, + CSO_VERTEX_SHADER }; unsigned cso_construct_key(void *item, int item_size); diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index fe835643e0..aaf2ccf499 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -373,10 +373,12 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state; - i915->pipe.create_shader_state = i915_create_shader_state; + i915->pipe.create_fs_state = i915_create_shader_state; i915->pipe.bind_fs_state = i915_bind_fs_state; + i915->pipe.delete_fs_state = i915_delete_shader_state; + i915->pipe.create_vs_state = i915_create_shader_state; i915->pipe.bind_vs_state = i915_bind_vs_state; - i915->pipe.delete_shader_state = i915_delete_shader_state; + i915->pipe.delete_vs_state = i915_delete_shader_state; i915->pipe.set_alpha_test_state = i915_set_alpha_test_state; i915->pipe.set_blend_color = i915_set_blend_color; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index c405051bce..5766b2b7df 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -117,15 +117,20 @@ struct pipe_context { void (*delete_depth_stencil_state)(struct pipe_context *, const struct pipe_depth_stencil_state *); - const struct pipe_shader_state * (*create_shader_state)( + const struct pipe_shader_state * (*create_fs_state)( struct pipe_context *, const struct pipe_shader_state *); void (*bind_fs_state)(struct pipe_context *, const struct pipe_shader_state *); + void (*delete_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + const struct pipe_shader_state * (*create_vs_state)( + struct pipe_context *, + const struct pipe_shader_state *); void (*bind_vs_state)(struct pipe_context *, const struct pipe_shader_state *); - void (*delete_shader_state)(struct pipe_context *, - const struct pipe_shader_state *); + void (*delete_vs_state)(struct pipe_context *, + const struct pipe_shader_state *); void (*set_alpha_test_state)( struct pipe_context *, const struct pipe_alpha_test_state * ); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 25cb9d8745..a56793d683 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -262,10 +262,12 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; - softpipe->pipe.create_shader_state = softpipe_create_shader_state; - softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; - softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; - softpipe->pipe.delete_shader_state = softpipe_delete_shader_state; + softpipe->pipe.create_fs_state = softpipe_create_shader_state; + softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; + softpipe->pipe.delete_fs_state = softpipe_delete_shader_state; + softpipe->pipe.create_vs_state = softpipe_create_shader_state; + softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; + softpipe->pipe.delete_vs_state = softpipe_delete_shader_state; softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 3df2c6750a..6dd576a57c 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -77,7 +77,7 @@ static void compile_fs( struct st_context *st ) fs.outputs_written = tgsi_mesa_translate_fragment_output_mask(fp->Base.Base.OutputsWritten); fs.tokens = &fp->tokens[0]; - cached = st_cached_shader_state(st, &fs); + cached = st_cached_fs_state(st, &fs); fp->fsx = cached; if (TGSI_DEBUG) diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 8de19e41ee..166dc70b08 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -104,7 +104,7 @@ static void compile_vs( struct st_context *st ) vs.tokens = &vp->tokens[0]; - cached = st_cached_shader_state(st, &vs); + cached = st_cached_vs_state(st, &vs); vp->vs = cached; diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 7b851e3901..d84a396e18 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -112,19 +112,37 @@ struct pipe_rasterizer_state * st_cached_rasterizer_state( return (struct pipe_rasterizer_state*)(cso_hash_iter_data(iter)); } -struct pipe_shader_state * st_cached_shader_state( +struct pipe_shader_state * st_cached_fs_state( struct st_context *st, const struct pipe_shader_state *templ) { unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_shader_state)); struct cso_hash_iter iter = cso_find_state_template(st->cache, - hash_key, CSO_SHADER, + hash_key, CSO_FRAGMENT_SHADER, (void*)templ); if (cso_hash_iter_is_null(iter)) { const struct pipe_shader_state *created_state = - st->pipe->create_shader_state(st->pipe, templ); - iter = cso_insert_state(st->cache, hash_key, CSO_SHADER, + st->pipe->create_fs_state(st->pipe, templ); + iter = cso_insert_state(st->cache, hash_key, CSO_FRAGMENT_SHADER, + (void*)created_state); + } + return (struct pipe_shader_state*)(cso_hash_iter_data(iter)); +} + +struct pipe_shader_state * st_cached_vs_state( + struct st_context *st, + const struct pipe_shader_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_shader_state)); + struct cso_hash_iter iter = cso_find_state_template(st->cache, + hash_key, CSO_VERTEX_SHADER, + (void*)templ); + if (cso_hash_iter_is_null(iter)) { + const struct pipe_shader_state *created_state = + st->pipe->create_vs_state(st->pipe, templ); + iter = cso_insert_state(st->cache, hash_key, CSO_VERTEX_SHADER, (void*)created_state); } return (struct pipe_shader_state*)(cso_hash_iter_data(iter)); diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index 6a897a9993..bcbe19b823 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -53,7 +53,12 @@ struct pipe_rasterizer_state *st_cached_rasterizer_state( struct st_context *st, const struct pipe_rasterizer_state *raster); -struct pipe_shader_state *st_cached_shader_state( +struct pipe_shader_state *st_cached_fs_state( + struct st_context *st, + const struct pipe_shader_state *templ); + + +struct pipe_shader_state *st_cached_vs_state( struct st_context *st, const struct pipe_shader_state *templ); diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 65cac9dbde..7c669ab457 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -370,7 +370,7 @@ clear_with_quad(GLcontext *ctx, fs.inputs_read = tgsi_mesa_translate_fragment_input_mask(stfp->Base.Base.InputsRead); fs.outputs_written = tgsi_mesa_translate_fragment_output_mask(stfp->Base.Base.OutputsWritten); fs.tokens = &stfp->tokens[0]; - cached = st_cached_shader_state(st, &fs); + cached = st_cached_fs_state(st, &fs); pipe->bind_fs_state(pipe, cached); } @@ -386,7 +386,7 @@ clear_with_quad(GLcontext *ctx, vs.inputs_read = stvp->Base.Base.InputsRead; vs.outputs_written = stvp->Base.Base.OutputsWritten; vs.tokens = &stvp->tokens[0]; - cached = st_cached_shader_state(st, &vs); + cached = st_cached_vs_state(st, &vs); pipe->bind_vs_state(pipe, cached); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 731c060c11..67de781c83 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -347,7 +347,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, memset(&fs, 0, sizeof(fs)); fs.inputs_read = stfp->Base.Base.InputsRead; fs.tokens = &stfp->tokens[0]; - cached = st_cached_shader_state(ctx->st, &fs); + cached = st_cached_fs_state(ctx->st, &fs); pipe->bind_fs_state(pipe, cached); } @@ -363,7 +363,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, vs.inputs_read = stvp->Base.Base.InputsRead; vs.outputs_written = stvp->Base.Base.OutputsWritten; vs.tokens = &stvp->tokens[0]; - cached = st_cached_shader_state(ctx->st, &vs); + cached = st_cached_vs_state(ctx->st, &vs); pipe->bind_vs_state(pipe, cached); } -- cgit v1.2.3 From bcc13fd3f8cc1c235ef5114472b0f919a38c096b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 19 Sep 2007 16:39:04 +0100 Subject: Only generate vertex shader code when necessary. --- src/mesa/state_tracker/st_atom_vs.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 166dc70b08..322fabc456 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -112,10 +112,11 @@ static void compile_vs( struct st_context *st ) tgsi_dump( vp->tokens, 0 ); #if defined(USE_X86_ASM) || defined(SLANG_X86) - tgsi_emit_sse2( - vp->tokens, - &vp->sse2_program ); - cached->executable = (void *) x86_get_func( &vp->sse2_program ); + if (vp->sse2_program.csr == vp->sse2_program.store) + tgsi_emit_sse2( vp->tokens, &vp->sse2_program ); + + if (!cached->executable) + cached->executable = (void *) x86_get_func( &vp->sse2_program ); #endif vp->dirty = 0; -- cgit v1.2.3 From b3cc74aa448f42340cbd01578a51f94eb2949618 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Sep 2007 10:16:46 -0600 Subject: grab vertex program pointer after state validation --- src/mesa/state_tracker/st_draw.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index bff0ad7ef7..6efe3ce8b8 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -187,7 +187,7 @@ st_draw_vbo(GLcontext *ctx, GLuint max_index) { struct pipe_context *pipe = ctx->st->pipe; - const struct st_vertex_program *vp = ctx->st->vp; + const struct st_vertex_program *vp; const struct pipe_shader_state *vs; const unsigned attr0_offset = (unsigned) arrays[0]->Ptr; GLboolean needDefaultAttribs = GL_FALSE; @@ -195,7 +195,8 @@ st_draw_vbo(GLcontext *ctx, st_validate_state(ctx->st); - /* must do this after state validation! */ + /* must get these after state validation! */ + vp = ctx->st->vp; vs = ctx->st->state.vs; /* loop over TGSI shader inputs */ -- cgit v1.2.3 From c0bf7322088715bb411068c3d631b0c4be8cdff5 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Sep 2007 12:35:29 -0400 Subject: Redo the cso cache to map driver data in a lot more pleasing way. Drivers can now create whatever they want from the state template. We use cso_state object to store the template (necessary during lookups), and the driver data. Convert blend state to the new semantics. --- src/mesa/pipe/cso_cache/cso_cache.h | 5 +++++ src/mesa/pipe/failover/fo_context.h | 6 +++++- src/mesa/pipe/failover/fo_state.c | 34 ++++++++++++++++++++++++++++--- src/mesa/pipe/failover/fo_state_emit.c | 3 ++- src/mesa/pipe/i915simple/i915_state.c | 10 ++++----- src/mesa/pipe/p_context.h | 10 ++++----- src/mesa/pipe/softpipe/sp_state.h | 6 +++--- src/mesa/pipe/softpipe/sp_state_blend.c | 10 ++++----- src/mesa/state_tracker/st_atom_blend.c | 8 ++++---- src/mesa/state_tracker/st_cache.c | 21 ++++++++++--------- src/mesa/state_tracker/st_cache.h | 14 +++++++------ src/mesa/state_tracker/st_cb_clear.c | 8 ++++---- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- src/mesa/state_tracker/st_context.h | 3 ++- 14 files changed, 91 insertions(+), 51 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h index 352e1a6a59..7ac5908922 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -48,6 +48,11 @@ struct cso_cache { struct cso_hash *vs_hash; }; +struct cso_blend { + struct pipe_blend_state state; + void *data; +}; + enum cso_cache_type { CSO_BLEND, CSO_SAMPLER, diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index 9556a17e4d..ea7fb109b3 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -60,13 +60,17 @@ #define FO_HW 0 #define FO_SW 1 +struct fo_state { + void *sw_state; + void *hw_state; +}; struct failover_context { struct pipe_context pipe; /**< base class */ /* The most recent drawing state as set by the driver: */ - const struct pipe_blend_state *blend; + const struct fo_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 04ebd33d0d..ba110a6e4f 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -57,17 +57,43 @@ failover_set_alpha_test_state(struct pipe_context *pipe, } -static void +static void * +failover_create_blend_state( struct pipe_context *pipe, + const struct pipe_blend_state *blend ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_blend_state(pipe, blend); + state->hw_state = failover->hw->create_blend_state(pipe, blend); + + return state; +} + +static void failover_bind_blend_state( struct pipe_context *pipe, - const struct pipe_blend_state *blend ) + void *blend ) { struct failover_context *failover = failover_context(pipe); - failover->blend = blend; + failover->blend = (struct fo_state *)blend; failover->dirty |= FO_NEW_BLEND; failover->hw->bind_blend_state( failover->hw, blend ); } +static void +failover_delete_blend_state( struct pipe_context *pipe, + void *blend ) +{ + struct fo_state *state = (struct fo_state*)blend; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_blend_state(pipe, state->sw_state); + failover->hw->delete_blend_state(pipe, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); +} static void failover_set_blend_color( struct pipe_context *pipe, @@ -253,7 +279,9 @@ failover_set_vertex_element(struct pipe_context *pipe, void failover_init_state_functions( struct failover_context *failover ) { + failover->pipe.create_blend_state = failover_create_blend_state; failover->pipe.bind_blend_state = failover_bind_blend_state; + failover->pipe.delete_blend_state = failover_delete_blend_state; failover->pipe.bind_sampler_state = failover_bind_sampler_state; failover->pipe.bind_depth_stencil_state = failover_bind_depth_stencil_state; failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 9d304074d0..72697c01a9 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -59,7 +59,8 @@ failover_state_emit( struct failover_context *failover ) failover->sw->set_alpha_test_state( failover->sw, &failover->alpha_test ); if (failover->dirty & FO_NEW_BLEND) - failover->sw->bind_blend_state( failover->sw, failover->blend ); + failover->sw->bind_blend_state( failover->sw, + failover->blend->sw_state ); if (failover->dirty & FO_NEW_BLEND_COLOR) failover->sw->set_blend_color( failover->sw, &failover->blend_color ); diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index aaf2ccf499..86c108978f 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -38,7 +38,7 @@ /* None of this state is actually used for anything yet. */ -static const struct pipe_blend_state * +static void * i915_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { @@ -49,20 +49,20 @@ i915_create_blend_state(struct pipe_context *pipe, } static void i915_bind_blend_state( struct pipe_context *pipe, - const struct pipe_blend_state *blend ) + void *blend ) { struct i915_context *i915 = i915_context(pipe); - i915->blend = blend; + i915->blend = (struct pipe_blend_state *)blend; i915->dirty |= I915_NEW_BLEND; } static void i915_delete_blend_state( struct pipe_context *pipe, - const struct pipe_blend_state *blend ) + void *blend ) { - free((void*)blend); + free(blend); } static void i915_set_blend_color( struct pipe_context *pipe, diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 5766b2b7df..adca6612d5 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -85,12 +85,10 @@ struct pipe_context { /* * State functions */ - const struct pipe_blend_state * (*create_blend_state)(struct pipe_context *, - const struct pipe_blend_state *); - void (*bind_blend_state)(struct pipe_context *, - const struct pipe_blend_state *); - void (*delete_blend_state)(struct pipe_context *, - const struct pipe_blend_state *); + void * (*create_blend_state)(struct pipe_context *, + const struct pipe_blend_state *); + void (*bind_blend_state)(struct pipe_context *, void *); + void (*delete_blend_state)(struct pipe_context *, void *); const struct pipe_sampler_state * (*create_sampler_state)( struct pipe_context *, diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 04cc743bd0..8e7776a6c7 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -33,13 +33,13 @@ #include "pipe/p_state.h" -const struct pipe_blend_state * +void * softpipe_create_blend_state(struct pipe_context *, const struct pipe_blend_state *); void softpipe_bind_blend_state(struct pipe_context *, - const struct pipe_blend_state *); + void *); void softpipe_delete_blend_state(struct pipe_context *, - const struct pipe_blend_state *); + void *); const struct pipe_sampler_state * softpipe_create_sampler_state(struct pipe_context *, diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 83f456ded5..7a94e82d6f 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -30,7 +30,7 @@ #include "sp_context.h" #include "sp_state.h" -const struct pipe_blend_state * +void * softpipe_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { @@ -41,19 +41,19 @@ softpipe_create_blend_state(struct pipe_context *pipe, } void softpipe_bind_blend_state( struct pipe_context *pipe, - const struct pipe_blend_state *blend ) + void *blend ) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->blend = blend; + softpipe->blend = (const struct pipe_blend_state *)blend; softpipe->dirty |= SP_NEW_BLEND; } void softpipe_delete_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend ) + void *blend ) { - free((struct pipe_blend_state *)blend); + free(blend); } diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index d94beb66c7..d5eadc3541 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -211,14 +211,14 @@ update_blend( struct st_context *st ) if (st->ctx->Color.DitherFlag) blend.dither = 1; - struct pipe_blend_state *real_blend = + const struct cso_blend *cso = st_cached_blend_state(st, &blend); - if (st->state.blend != real_blend) { + if (st->state.blend != cso) { /* state has changed */ - st->state.blend = real_blend; + st->state.blend = cso; /* bind new state */ - st->pipe->bind_blend_state(st->pipe, real_blend); + st->pipe->bind_blend_state(st->pipe, cso->data); } if (memcmp(st->ctx->Color.BlendColor, &st->state.blend_color, 4 * sizeof(GLfloat)) != 0) { diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index d84a396e18..bd6c63b7a1 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -43,21 +43,22 @@ * in the cache or it will create a new state state from the given * template, will insert it in the cache and return it. */ -struct pipe_blend_state * st_cached_blend_state( - struct st_context *st, - const struct pipe_blend_state *blend) +const struct cso_blend * st_cached_blend_state(struct st_context *st, + const struct pipe_blend_state *templ) { - unsigned hash_key = cso_construct_key((void*)blend, sizeof(struct pipe_blend_state)); + unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_blend_state)); struct cso_hash_iter iter = cso_find_state_template(st->cache, hash_key, CSO_BLEND, - (void*)blend); + (void*)templ); if (cso_hash_iter_is_null(iter)) { - const struct pipe_blend_state *created_state = st->pipe->create_blend_state( - st->pipe, blend); - iter = cso_insert_state(st->cache, hash_key, CSO_BLEND, - (void*)created_state); + struct cso_blend *cso = malloc(sizeof(struct cso_blend)); + memcpy(&cso->state, templ, sizeof(struct pipe_blend_state)); + cso->data = st->pipe->create_blend_state(st->pipe, templ); + if (!cso->data) + cso->data = &cso->state; + iter = cso_insert_state(st->cache, hash_key, CSO_BLEND, cso); } - return (struct pipe_blend_state*)(cso_hash_iter_data(iter)); + return ((struct cso_blend *)cso_hash_iter_data(iter)); } struct pipe_sampler_state * st_cached_sampler_state( diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index bcbe19b823..fb0ff0d4db 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -33,17 +33,19 @@ #ifndef ST_CACHE_H #define ST_CACHE_H +#include "pipe/cso_cache/cso_cache.h" + struct pipe_blend_state; struct pipe_sampler_state; struct st_context; -struct pipe_blend_state * st_cached_blend_state( - struct st_context *st, - const struct pipe_blend_state *blend); +const struct cso_blend * +st_cached_blend_state(struct st_context *st, + const struct pipe_blend_state *blend); -struct pipe_sampler_state * st_cached_sampler_state( - struct st_context *st, - const struct pipe_sampler_state *sampler); +struct pipe_sampler_state * +st_cached_sampler_state(struct st_context *st, + const struct pipe_sampler_state *sampler); struct pipe_depth_stencil_state *st_cached_depth_stencil_state( struct st_context *st, diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 7c669ab457..3a6991754a 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -298,7 +298,7 @@ clear_with_quad(GLcontext *ctx, /* blend state: RGBA masking */ { struct pipe_blend_state blend; - const struct pipe_blend_state *state; + const struct cso_blend *cso; memset(&blend, 0, sizeof(blend)); if (color) { if (ctx->Color.ColorMask[0]) @@ -312,8 +312,8 @@ clear_with_quad(GLcontext *ctx, if (st->ctx->Color.DitherFlag) blend.dither = 1; } - state = st_cached_blend_state(st, &blend); - pipe->bind_blend_state(pipe, state); + cso = st_cached_blend_state(st, &blend); + pipe->bind_blend_state(pipe, cso->data); } /* depth_stencil state: always pass/set to ref value */ @@ -411,7 +411,7 @@ clear_with_quad(GLcontext *ctx, /* Restore pipe state */ pipe->set_alpha_test_state(pipe, &st->state.alpha_test); - pipe->bind_blend_state(pipe, st->state.blend); + pipe->bind_blend_state(pipe, st->state.blend->data); pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil); pipe->bind_fs_state(pipe, st->state.fs); pipe->bind_vs_state(pipe, st->state.vs); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 67de781c83..4a554cd5a4 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -493,8 +493,8 @@ static GLboolean any_fragment_ops(const struct st_context *st) { if (st->state.alpha_test.enabled || - st->state.blend->blend_enable || - st->state.blend->logicop_enable || + st->state.blend->state.blend_enable || + st->state.blend->state.logicop_enable || st->state.depth_stencil->depth.enabled) /* XXX more checks */ return GL_TRUE; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index a8ae5d9c56..966574b67c 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -40,6 +40,7 @@ struct st_fragment_program; struct draw_context; struct draw_stage; struct cso_cache; +struct cso_blend; #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 @@ -74,7 +75,7 @@ struct st_context * though, we just shove random objects across the interface. */ struct { - const struct pipe_blend_state *blend; + const struct cso_blend *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; -- cgit v1.2.3 From b6d50abd7d483029469a0faaa28e8e2f2f742c6d Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Sep 2007 13:05:58 -0400 Subject: Convert the i915 driver blend state to the new semantics. --- src/mesa/pipe/i915simple/i915_context.h | 9 ++- src/mesa/pipe/i915simple/i915_state.c | 83 ++++++++++++++++++++++--- src/mesa/pipe/i915simple/i915_state_dynamic.c | 46 ++------------ src/mesa/pipe/i915simple/i915_state_immediate.c | 32 +--------- 4 files changed, 88 insertions(+), 82 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 9052c92d72..ea6d1ce87b 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -113,7 +113,12 @@ struct i915_state unsigned id; /* track lost context events */ }; - +struct i915_blend_state { + unsigned iab; + unsigned modes4; + unsigned LIS5; + unsigned LIS6; +}; struct i915_context { @@ -123,7 +128,7 @@ struct i915_context /* The most recent drawing state as set by the driver: */ - const struct pipe_blend_state *blend; + const struct i915_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 86c108978f..4a4d26be65 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -33,7 +33,9 @@ #include "pipe/p_winsys.h" #include "i915_context.h" +#include "i915_reg.h" #include "i915_state.h" +#include "i915_state_inlines.h" /* None of this state is actually used for anything yet. */ @@ -42,25 +44,90 @@ static void * i915_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - struct pipe_blend_state *new_blend = malloc(sizeof(struct pipe_blend_state)); - memcpy(new_blend, blend, sizeof(struct pipe_blend_state)); + struct i915_blend_state *cso_data = calloc(1, sizeof(struct i915_blend_state)); - return new_blend; + { + unsigned eqRGB = blend->rgb_func; + unsigned srcRGB = blend->rgb_src_factor; + unsigned dstRGB = blend->rgb_dst_factor; + + unsigned eqA = blend->alpha_func; + unsigned srcA = blend->alpha_src_factor; + unsigned dstA = blend->alpha_dst_factor; + + /* Special handling for MIN/MAX filter modes handled at + * state_tracker level. + */ + + if (srcA != srcRGB || + dstA != dstRGB || + eqA != eqRGB) { + + cso_data->iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | + IAB_MODIFY_ENABLE | + IAB_ENABLE | + IAB_MODIFY_FUNC | + IAB_MODIFY_SRC_FACTOR | + IAB_MODIFY_DST_FACTOR | + SRC_ABLND_FACT(i915_translate_blend_factor(srcA)) | + DST_ABLND_FACT(i915_translate_blend_factor(dstA)) | + (i915_translate_blend_func(eqA) << IAB_FUNC_SHIFT)); + } + else { + cso_data->iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | + IAB_MODIFY_ENABLE | + 0); + } + } + + cso_data->modes4 |= (_3DSTATE_MODES_4_CMD | + ENABLE_LOGIC_OP_FUNC | + LOGIC_OP_FUNC(i915_translate_logic_op(blend->logicop_func))); + + if (blend->logicop_enable) + cso_data->LIS5 |= S5_LOGICOP_ENABLE; + + if (blend->dither) + cso_data->LIS5 |= S5_COLOR_DITHER_ENABLE; + + if ((blend->colormask & PIPE_MASK_R) == 0) + cso_data->LIS5 |= S5_WRITEDISABLE_RED; + + if ((blend->colormask & PIPE_MASK_G) == 0) + cso_data->LIS5 |= S5_WRITEDISABLE_GREEN; + + if ((blend->colormask & PIPE_MASK_B) == 0) + cso_data->LIS5 |= S5_WRITEDISABLE_BLUE; + + if ((blend->colormask & PIPE_MASK_A) == 0) + cso_data->LIS5 |= S5_WRITEDISABLE_ALPHA; + + if (blend->blend_enable) { + unsigned funcRGB = blend->rgb_func; + unsigned srcRGB = blend->rgb_src_factor; + unsigned dstRGB = blend->rgb_dst_factor; + + cso_data->LIS6 |= (S6_CBUF_BLEND_ENABLE | + SRC_BLND_FACT(i915_translate_blend_factor(srcRGB)) | + DST_BLND_FACT(i915_translate_blend_factor(dstRGB)) | + (i915_translate_blend_func(funcRGB) << S6_CBUF_BLEND_FUNC_SHIFT)); + } + + return cso_data; } -static void i915_bind_blend_state( struct pipe_context *pipe, - void *blend ) +static void i915_bind_blend_state(struct pipe_context *pipe, + void *blend) { struct i915_context *i915 = i915_context(pipe); - i915->blend = (struct pipe_blend_state *)blend; + i915->blend = (struct i915_blend_state*)blend; i915->dirty |= I915_NEW_BLEND; } -static void i915_delete_blend_state( struct pipe_context *pipe, - void *blend ) +static void i915_delete_blend_state(struct pipe_context *pipe, void *blend) { free(blend); } diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index a9791962e2..262ac947b8 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -79,12 +79,8 @@ static void upload_MODES4( struct i915_context *i915 ) } /* I915_NEW_BLEND */ - { - modes4 |= (_3DSTATE_MODES_4_CMD | - ENABLE_LOGIC_OP_FUNC | - LOGIC_OP_FUNC(i915_translate_logic_op(i915->blend->logicop_func))); - } - + modes4 |= i915->blend->modes4; + /* Always, so that we know when state is in-active: */ set_dynamic_indirect( i915, @@ -201,44 +197,10 @@ const struct i915_tracked_state i915_upload_BLENDCOLOR = { static void upload_IAB( struct i915_context *i915 ) { - unsigned iab = 0; - - { - unsigned eqRGB = i915->blend->rgb_func; - unsigned srcRGB = i915->blend->rgb_src_factor; - unsigned dstRGB = i915->blend->rgb_dst_factor; + unsigned iab = i915->blend->iab; - unsigned eqA = i915->blend->alpha_func; - unsigned srcA = i915->blend->alpha_src_factor; - unsigned dstA = i915->blend->alpha_dst_factor; - /* Special handling for MIN/MAX filter modes handled at - * state_tracker level. - */ - - if (srcA != srcRGB || - dstA != dstRGB || - eqA != eqRGB) { - - iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | - IAB_MODIFY_ENABLE | - IAB_ENABLE | - IAB_MODIFY_FUNC | - IAB_MODIFY_SRC_FACTOR | - IAB_MODIFY_DST_FACTOR | - SRC_ABLND_FACT(i915_translate_blend_factor(srcA)) | - DST_ABLND_FACT(i915_translate_blend_factor(dstA)) | - (i915_translate_blend_func(eqA) << IAB_FUNC_SHIFT)); - } - else { - iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | - IAB_MODIFY_ENABLE | - 0); - } - } - - - set_dynamic_indirect( i915, + set_dynamic_indirect( i915, I915_DYNAMIC_IAB, &iab, 1 ); diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 73508f557f..38cf2adc71 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -144,25 +144,7 @@ static void upload_S5( struct i915_context *i915 ) (dpop << S5_STENCIL_PASS_Z_PASS_SHIFT)); } - /* I915_NEW_BLEND */ - if (i915->blend->logicop_enable) - LIS5 |= S5_LOGICOP_ENABLE; - - if (i915->blend->dither) - LIS5 |= S5_COLOR_DITHER_ENABLE; - - if ((i915->blend->colormask & PIPE_MASK_R) == 0) - LIS5 |= S5_WRITEDISABLE_RED; - - if ((i915->blend->colormask & PIPE_MASK_G) == 0) - LIS5 |= S5_WRITEDISABLE_GREEN; - - if ((i915->blend->colormask & PIPE_MASK_B) == 0) - LIS5 |= S5_WRITEDISABLE_BLUE; - - if ((i915->blend->colormask & PIPE_MASK_A) == 0) - LIS5 |= S5_WRITEDISABLE_ALPHA; - + LIS5 |= i915->blend->LIS5; #if 0 /* I915_NEW_RASTERIZER */ @@ -205,17 +187,7 @@ static void upload_S6( struct i915_context *i915 ) /* I915_NEW_BLEND */ - if (i915->blend->blend_enable) - { - unsigned funcRGB = i915->blend->rgb_func; - unsigned srcRGB = i915->blend->rgb_src_factor; - unsigned dstRGB = i915->blend->rgb_dst_factor; - - LIS6 |= (S6_CBUF_BLEND_ENABLE | - SRC_BLND_FACT(i915_translate_blend_factor(srcRGB)) | - DST_BLND_FACT(i915_translate_blend_factor(dstRGB)) | - (i915_translate_blend_func(funcRGB) << S6_CBUF_BLEND_FUNC_SHIFT)); - } + LIS6 |= i915->blend->LIS6; /* I915_NEW_DEPTH */ -- cgit v1.2.3 From fe555c39bb7fd530298b5be4a8f06bff41726c86 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Sep 2007 14:01:18 -0400 Subject: Convert the rasterizer cso to the new semantics. Basically make cso hold the driver specific struct, while managing the template. --- src/mesa/pipe/cso_cache/cso_cache.h | 5 ++++ src/mesa/pipe/failover/fo_context.h | 2 +- src/mesa/pipe/failover/fo_state.c | 35 ++++++++++++++++++++++++---- src/mesa/pipe/failover/fo_state_emit.c | 3 ++- src/mesa/pipe/i915simple/i915_state.c | 18 ++++++-------- src/mesa/pipe/p_context.h | 11 ++++----- src/mesa/pipe/softpipe/sp_state.h | 8 +++---- src/mesa/pipe/softpipe/sp_state_rasterizer.c | 18 ++++++-------- src/mesa/state_tracker/st_atom_rasterizer.c | 10 ++++---- src/mesa/state_tracker/st_cache.c | 20 +++++++++------- src/mesa/state_tracker/st_cache.h | 6 ++--- src/mesa/state_tracker/st_cb_clear.c | 8 +++---- src/mesa/state_tracker/st_cb_drawpixels.c | 8 +++---- src/mesa/state_tracker/st_context.h | 2 +- src/mesa/state_tracker/st_draw.c | 2 +- 15 files changed, 90 insertions(+), 66 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h index 7ac5908922..cd4b64eec4 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -53,6 +53,11 @@ struct cso_blend { void *data; }; +struct cso_rasterizer { + struct pipe_rasterizer_state state; + void *data; +}; + enum cso_cache_type { CSO_BLEND, CSO_SAMPLER, diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index ea7fb109b3..a649899010 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -73,7 +73,7 @@ struct failover_context { const struct fo_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; - const struct pipe_rasterizer_state *rasterizer; + const struct fo_state *rasterizer; const struct pipe_shader_state *fragment_shader; const struct pipe_shader_state *vertex_shader; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index ba110a6e4f..25725625e0 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -183,17 +183,42 @@ failover_set_polygon_stipple( struct pipe_context *pipe, failover->hw->set_polygon_stipple( failover->hw, stipple ); } +static void * +failover_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + state->sw_state = failover->sw->create_rasterizer_state(pipe, templ); + state->hw_state = failover->hw->create_rasterizer_state(pipe, templ); + + return state; +} static void -failover_bind_rasterizer_state( struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup ) +failover_bind_rasterizer_state(struct pipe_context *pipe, + void *raster) { struct failover_context *failover = failover_context(pipe); - failover->rasterizer = setup; + failover->rasterizer = (struct fo_state *)raster; failover->dirty |= FO_NEW_RASTERIZER; - failover->hw->bind_rasterizer_state( failover->hw, setup ); + failover->hw->bind_rasterizer_state( failover->hw, raster ); +} + +static void +failover_delete_rasterizer_state(struct pipe_context *pipe, + void *raster) +{ + struct fo_state *state = (struct fo_state*)raster; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_rasterizer_state(pipe, state->sw_state); + failover->hw->delete_rasterizer_state(pipe, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); } @@ -284,7 +309,9 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.delete_blend_state = failover_delete_blend_state; failover->pipe.bind_sampler_state = failover_bind_sampler_state; failover->pipe.bind_depth_stencil_state = failover_bind_depth_stencil_state; + failover->pipe.create_rasterizer_state = failover_create_rasterizer_state; failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state; + failover->pipe.delete_rasterizer_state = failover_delete_rasterizer_state; failover->pipe.bind_fs_state = failover_bind_fs_state; failover->pipe.bind_vs_state = failover_bind_vs_state; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 72697c01a9..f2b0b1edc0 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -87,7 +87,8 @@ failover_state_emit( struct failover_context *failover ) failover->sw->set_polygon_stipple( failover->sw, &failover->poly_stipple ); if (failover->dirty & FO_NEW_RASTERIZER) - failover->sw->bind_rasterizer_state( failover->sw, failover->rasterizer ); + failover->sw->bind_rasterizer_state( failover->sw, + failover->rasterizer->sw_state ); if (failover->dirty & FO_NEW_SCISSOR) failover->sw->set_scissor_state( failover->sw, &failover->scissor ); diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 4a4d26be65..66aa9a0274 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -371,23 +371,19 @@ static void i915_set_viewport_state( struct pipe_context *pipe, } -static const struct pipe_rasterizer_state * +static void * i915_create_rasterizer_state(struct pipe_context *pipe, const struct pipe_rasterizer_state *setup) { - struct pipe_rasterizer_state *raster = - malloc(sizeof(struct pipe_rasterizer_state)); - memcpy(raster, setup, sizeof(struct pipe_rasterizer_state)); - - return raster; + return 0; } static void i915_bind_rasterizer_state( struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup ) + void *setup ) { struct i915_context *i915 = i915_context(pipe); - i915->rasterizer = setup; + i915->rasterizer = (struct pipe_rasterizer_state *)setup; /* pass-through to draw module */ draw_set_rasterizer_state(i915->draw, setup); @@ -395,10 +391,10 @@ static void i915_bind_rasterizer_state( struct pipe_context *pipe, i915->dirty |= I915_NEW_RASTERIZER; } -static void i915_delete_rasterizer_state( struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup ) +static void i915_delete_rasterizer_state(struct pipe_context *pipe, + void *setup) { - free((struct pipe_rasterizer_state*)setup); + /* do nothing */ } static void i915_set_vertex_buffer( struct pipe_context *pipe, diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index adca6612d5..1c0ab794f0 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -99,13 +99,10 @@ struct pipe_context { void (*delete_sampler_state)(struct pipe_context *, const struct pipe_sampler_state *); - const struct pipe_rasterizer_state *(*create_rasterizer_state)( - struct pipe_context *, - const struct pipe_rasterizer_state *); - void (*bind_rasterizer_state)(struct pipe_context *, - const struct pipe_rasterizer_state *); - void (*delete_rasterizer_state)(struct pipe_context *, - const struct pipe_rasterizer_state *); + void *(*create_rasterizer_state)(struct pipe_context *, + const struct pipe_rasterizer_state *); + void (*bind_rasterizer_state)(struct pipe_context *, void *); + void (*delete_rasterizer_state)(struct pipe_context *, void *); const struct pipe_depth_stencil_state * (*create_depth_stencil_state)( struct pipe_context *, diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 8e7776a6c7..a20ae1d4a2 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -58,13 +58,13 @@ void softpipe_bind_depth_stencil_state(struct pipe_context *, void softpipe_delete_depth_stencil_state(struct pipe_context *, const struct pipe_depth_stencil_state *); -const struct pipe_rasterizer_state * +void * softpipe_create_rasterizer_state(struct pipe_context *, - const struct pipe_rasterizer_state *); -void softpipe_bind_rasterizer_state(struct pipe_context *, const struct pipe_rasterizer_state *); +void softpipe_bind_rasterizer_state(struct pipe_context *, + void *); void softpipe_delete_rasterizer_state(struct pipe_context *, - const struct pipe_rasterizer_state *); + void *); void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_rasterizer.c b/src/mesa/pipe/softpipe/sp_state_rasterizer.c index d832adb91b..d7845cef82 100644 --- a/src/mesa/pipe/softpipe/sp_state_rasterizer.c +++ b/src/mesa/pipe/softpipe/sp_state_rasterizer.c @@ -32,34 +32,30 @@ -const struct pipe_rasterizer_state * +void * softpipe_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup) + const struct pipe_rasterizer_state *setup) { - struct pipe_rasterizer_state *raster = - malloc(sizeof(struct pipe_rasterizer_state)); - memcpy(raster, setup, sizeof(struct pipe_rasterizer_state)); - - return raster; + return 0; } void softpipe_bind_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup) + void *setup) { struct softpipe_context *softpipe = softpipe_context(pipe); /* pass-through to draw module */ draw_set_rasterizer_state(softpipe->draw, setup); - softpipe->rasterizer = setup; + softpipe->rasterizer = (struct pipe_rasterizer_state *)setup; softpipe->dirty |= SP_NEW_RASTERIZER; } void softpipe_delete_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *rasterizer) + void *rasterizer) { - free((struct pipe_rasterizer_state*)rasterizer); + /* do nothing */ } diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index cab8ad5cd6..e0d83ddaea 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -73,7 +73,7 @@ static void update_raster_state( struct st_context *st ) { GLcontext *ctx = st->ctx; struct pipe_rasterizer_state raster; - const struct pipe_rasterizer_state *cached; + const struct cso_rasterizer *cso; memset(&raster, 0, sizeof(raster)); @@ -206,10 +206,10 @@ static void update_raster_state( struct st_context *st ) if (ctx->Scissor.Enabled) raster.scissor = 1; - cached = st_cached_rasterizer_state(st, &raster); - if (st->state.rasterizer != cached) { - st->state.rasterizer = cached; - st->pipe->bind_rasterizer_state( st->pipe, cached ); + cso = st_cached_rasterizer_state(st, &raster); + if (st->state.rasterizer != cso) { + st->state.rasterizer = cso; + st->pipe->bind_rasterizer_state(st->pipe, cso->data); } } diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index bd6c63b7a1..0f233cea58 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -95,22 +95,24 @@ struct pipe_depth_stencil_state * st_cached_depth_stencil_state( return (struct pipe_depth_stencil_state*)(cso_hash_iter_data(iter)); } -struct pipe_rasterizer_state * st_cached_rasterizer_state( +const struct cso_rasterizer* st_cached_rasterizer_state( struct st_context *st, - const struct pipe_rasterizer_state *raster) + const struct pipe_rasterizer_state *templ) { - unsigned hash_key = cso_construct_key((void*)raster, + unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_rasterizer_state)); struct cso_hash_iter iter = cso_find_state_template(st->cache, hash_key, CSO_RASTERIZER, - (void*)raster); + (void*)templ); if (cso_hash_iter_is_null(iter)) { - const struct pipe_rasterizer_state *created_state = - st->pipe->create_rasterizer_state(st->pipe, raster); - iter = cso_insert_state(st->cache, hash_key, CSO_RASTERIZER, - (void*)created_state); + struct cso_rasterizer *cso = malloc(sizeof(struct cso_rasterizer)); + memcpy(&cso->state, templ, sizeof(struct pipe_rasterizer_state)); + cso->data = st->pipe->create_rasterizer_state(st->pipe, templ); + if (!cso->data) + cso->data = &cso->state; + iter = cso_insert_state(st->cache, hash_key, CSO_RASTERIZER, cso); } - return (struct pipe_rasterizer_state*)(cso_hash_iter_data(iter)); + return (struct cso_rasterizer*)(cso_hash_iter_data(iter)); } struct pipe_shader_state * st_cached_fs_state( diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index fb0ff0d4db..5b8c6168a8 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -51,9 +51,9 @@ struct pipe_depth_stencil_state *st_cached_depth_stencil_state( struct st_context *st, const struct pipe_depth_stencil_state *depth_stencil); -struct pipe_rasterizer_state *st_cached_rasterizer_state( - struct st_context *st, - const struct pipe_rasterizer_state *raster); +const struct cso_rasterizer * +st_cached_rasterizer_state(struct st_context *st, + const struct pipe_rasterizer_state *raster); struct pipe_shader_state *st_cached_fs_state( struct st_context *st, diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 3a6991754a..5d5efd9eae 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -345,7 +345,7 @@ clear_with_quad(GLcontext *ctx, /* setup state: nothing */ { struct pipe_rasterizer_state raster; - const struct pipe_rasterizer_state *cached; + const struct cso_rasterizer *cso; memset(&raster, 0, sizeof(raster)); #if 0 /* don't do per-pixel scissor; we'll just draw a PIPE_PRIM_QUAD @@ -354,8 +354,8 @@ clear_with_quad(GLcontext *ctx, if (ctx->Scissor.Enabled) raster.scissor = 1; #endif - cached = st_cached_rasterizer_state(ctx->st, &raster); - pipe->bind_rasterizer_state(pipe, cached); + cso = st_cached_rasterizer_state(ctx->st, &raster); + pipe->bind_rasterizer_state(pipe, cso->data); } /* fragment shader state: color pass-through program */ @@ -415,7 +415,7 @@ clear_with_quad(GLcontext *ctx, pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil); pipe->bind_fs_state(pipe, st->state.fs); pipe->bind_vs_state(pipe, st->state.vs); - pipe->bind_rasterizer_state(pipe, st->state.rasterizer); + pipe->bind_rasterizer_state(pipe, st->state.rasterizer->data); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); /* OR: st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 4a554cd5a4..0fd728c930 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -328,12 +328,12 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* setup state: just scissor */ { struct pipe_rasterizer_state setup; - struct pipe_rasterizer_state *cached; + const struct cso_rasterizer *cso; memset(&setup, 0, sizeof(setup)); if (ctx->Scissor.Enabled) setup.scissor = 1; - cached = st_cached_rasterizer_state(ctx->st, &setup); - pipe->bind_rasterizer_state(pipe, cached); + cso = st_cached_rasterizer_state(ctx->st, &setup); + pipe->bind_rasterizer_state(pipe, cso->data); } /* fragment shader state: TEX lookup program */ @@ -417,7 +417,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, draw_quad(ctx, x0, y0, z, x1, y1); /* restore GL state */ - pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer); + pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); pipe->bind_fs_state(pipe, ctx->st->state.fs); pipe->bind_vs_state(pipe, ctx->st->state.vs); pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 966574b67c..93b6425480 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -78,7 +78,7 @@ struct st_context const struct cso_blend *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; - const struct pipe_rasterizer_state *rasterizer; + const struct cso_rasterizer *rasterizer; const struct pipe_shader_state *fs; const struct pipe_shader_state *vs; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 6efe3ce8b8..e36c10d595 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -404,7 +404,7 @@ st_feedback_draw_vbo(GLcontext *ctx, assert(draw); draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); - draw_set_rasterizer_state(draw, st->state.rasterizer); + draw_set_rasterizer_state(draw, st->state.rasterizer->data); draw_set_vertex_shader(draw, st->state.vs); /* XXX need to set vertex info too */ -- cgit v1.2.3 From c8650b087440c7ba5ad0d8652dc916852e55bc1d Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Sep 2007 14:06:19 -0400 Subject: Avoid redundant reallocation of the template. cso already allocated the template for us. Returning 0 means we have no driver specific representation and just want the template on the bind. --- src/mesa/pipe/softpipe/sp_state_blend.c | 9 ++++----- src/mesa/state_tracker/st_atom_blend.c | 4 ++-- 2 files changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 7a94e82d6f..7fb47e7aab 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -34,10 +34,9 @@ void * softpipe_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - struct pipe_blend_state *new_blend = malloc(sizeof(struct pipe_blend_state)); - memcpy(new_blend, blend, sizeof(struct pipe_blend_state)); - - return new_blend; + /* means that we just want pipe_blend_state and don't have + * anything specific */ + return 0; } void softpipe_bind_blend_state( struct pipe_context *pipe, @@ -53,7 +52,7 @@ void softpipe_bind_blend_state( struct pipe_context *pipe, void softpipe_delete_blend_state(struct pipe_context *pipe, void *blend ) { - free(blend); + /* do nothing */ } diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index d5eadc3541..f321cd3829 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -156,6 +156,7 @@ static void update_blend( struct st_context *st ) { struct pipe_blend_state blend; + const struct cso_blend *cso; memset(&blend, 0, sizeof(blend)); @@ -211,8 +212,7 @@ update_blend( struct st_context *st ) if (st->ctx->Color.DitherFlag) blend.dither = 1; - const struct cso_blend *cso = - st_cached_blend_state(st, &blend); + cso = st_cached_blend_state(st, &blend); if (st->state.blend != cso) { /* state has changed */ -- cgit v1.2.3 From b8ecec7abefcd1f0720e448c05e901af0224806a Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Sep 2007 14:14:03 -0400 Subject: Adding proper includes often helps with the warnings, or so i hear.. --- src/mesa/state_tracker/st_draw.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index e36c10d595..7075db82e9 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -38,6 +38,7 @@ #include "tnl/t_vp_build.h" #include "st_atom.h" +#include "st_cache.h" #include "st_context.h" #include "st_cb_bufferobjects.h" #include "st_draw.h" -- cgit v1.2.3 From 617b39ce9811b0998ceb746c935cc50cacf8a9bc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 19 Sep 2007 17:59:51 +0100 Subject: Pad surface dimensions to multiples of 2. Avoids an assertion failure with softpipe if requested width or height is odd. --- src/mesa/state_tracker/st_cb_fbo.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 563b5add38..5b4afbd119 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -99,6 +99,10 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe->region_release(pipe, &strb->surface->region); } + /* Softpipe operates on quads, so pad dimensions to multiples of 2 */ + width += width & 1; + height += height & 1; + strb->surface->region = pipe->region_alloc(pipe, cpp, width, height, flags); if (!strb->surface->region) return GL_FALSE; /* out of memory, try s/w buffer? */ -- cgit v1.2.3 From c9a5ef1a1cdee94fe091cb65ea34e9a9bb8b2190 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 19 Sep 2007 20:03:08 +0100 Subject: Fix window resizes. The memcmp is insufficient for eliminating redundant framebuffer state changes. --- src/mesa/state_tracker/st_atom_framebuffer.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 3c4b37e7c5..ccbeacfd6e 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -74,7 +74,10 @@ update_framebuffer_state( struct st_context *st ) framebuffer.sbuf = strb->surface; } - if (memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0) { + /* XXX: The memcmp is insufficient for eliminating redundant state changes, + * but we should probably do more work here to that end. + */ + if (1 /*memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0*/) { st->state.framebuffer = framebuffer; st->pipe->set_framebuffer_state( st->pipe, &framebuffer ); } -- cgit v1.2.3 From a0e948f3aaadf56ab206a6b3e597825f2085f6d1 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Sep 2007 15:29:17 -0400 Subject: Fix a really stupid mistake that Michel hit. Return proper hash. --- src/mesa/pipe/cso_cache/cso_cache.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_cache.c b/src/mesa/pipe/cso_cache/cso_cache.c index e87733c7ab..71f0d08726 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.c +++ b/src/mesa/pipe/cso_cache/cso_cache.c @@ -74,16 +74,22 @@ static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_ switch(type) { case CSO_BLEND: hash = sc->blend_hash; + break; case CSO_SAMPLER: hash = sc->sampler_hash; + break; case CSO_DEPTH_STENCIL: hash = sc->depth_stencil_hash; + break; case CSO_RASTERIZER: hash = sc->rasterizer_hash; + break; case CSO_FRAGMENT_SHADER: hash = sc->fs_hash; + break; case CSO_VERTEX_SHADER: hash = sc->vs_hash; + break; } return hash; -- cgit v1.2.3 From 37cf13ed9a429c755f121daa1776b1b30a985ab3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Sep 2007 18:53:36 -0600 Subject: Checkpoint: replacement of TGSI_ATTRIB_x tokens with input/output semantics. TGSI_ATTRIB_x tokens still present and used in a few places. Expanded set of TGSI_SEMANTIC_x tokens for describing the meaning of inputs/outputs. These tokens are in a crude state ATM. Lots of #if 0 / disabled code to be removed yet, etc... Softpipe and i915 drivers should be in working condition but not heavily tested. --- src/mesa/pipe/draw/draw_vertex_shader.c | 4 + src/mesa/pipe/i915simple/i915_context.c | 14 +++ src/mesa/pipe/i915simple/i915_fpc.h | 2 + src/mesa/pipe/i915simple/i915_fpc_translate.c | 81 +++++++++++++++-- src/mesa/pipe/i915simple/i915_state_derived.c | 60 ++++++++++++- src/mesa/pipe/p_context.h | 3 +- src/mesa/pipe/p_defines.h | 7 ++ src/mesa/pipe/p_state.h | 11 ++- src/mesa/pipe/softpipe/sp_context.c | 10 +++ src/mesa/pipe/softpipe/sp_state_derived.c | 90 +++++++++---------- src/mesa/pipe/tgsi/exec/tgsi_build.c | 2 +- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 18 +++- src/mesa/pipe/tgsi/exec/tgsi_token.h | 31 +++++-- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 118 ++++++++++++++++++++----- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 2 + src/mesa/state_tracker/st_atom_fs.c | 121 ++++++++++++++++++++------ src/mesa/state_tracker/st_atom_vs.c | 114 ++++++++++++++++-------- src/mesa/state_tracker/st_cb_clear.c | 45 ++-------- src/mesa/state_tracker/st_cb_drawpixels.c | 33 +------ src/mesa/state_tracker/st_cb_rasterpos.c | 13 ++- src/mesa/state_tracker/st_draw.c | 8 +- src/mesa/state_tracker/st_program.h | 27 ++++-- 22 files changed, 580 insertions(+), 234 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index cb6c605b8d..2d424632e7 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -94,7 +94,11 @@ run_vertex_program(struct draw_context *draw, const float *trans = draw->viewport.translate; assert(count <= 4); +#if 0 assert(draw->vertex_shader.outputs_written & (1 << TGSI_ATTRIB_POS)); +#else + assert(draw->vertex_shader.output_semantics[0] == TGSI_SEMANTIC_POSITION); +#endif #ifdef DEBUG memset( &machine, 0, sizeof( machine ) ); diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 36372898ce..715e1ef656 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -138,6 +138,18 @@ i915_max_texture_size(struct pipe_context *pipe, unsigned textureType, } +static int +i915_get_param(struct pipe_context *pipe, uint param) +{ + switch (param) { + case PIPE_PARAM_FS_NEEDS_POS: + return 0; + default: + return 0; + } +} + + static void i915_destroy( struct pipe_context *pipe ) { struct i915_context *i915 = i915_context( pipe ); @@ -272,6 +284,8 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.destroy = i915_destroy; i915->pipe.supported_formats = i915_supported_formats; i915->pipe.max_texture_size = i915_max_texture_size; + i915->pipe.get_param = i915_get_param; + i915->pipe.clear = i915_clear; i915->pipe.begin_query = i915_begin_query; diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index 84e4c5a6f3..5c4f2f90e9 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -51,6 +51,8 @@ struct i915_fp_compile { uint declarations[I915_PROGRAM_SIZE]; uint program[I915_PROGRAM_SIZE]; + uint input_semantic[PIPE_MAX_SHADER_INPUTS]; + /** points into the i915->current.constants array: */ float (*constants)[4]; uint num_constants; diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 32c5600496..df8859bec8 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -128,7 +128,7 @@ src_vector(struct i915_fp_compile *p, const struct tgsi_full_src_register *source) { uint index = source->SrcRegister.Index; - uint src; + uint src, sem; switch (source->SrcRegister.File) { case TGSI_FILE_TEMPORARY: @@ -151,6 +151,46 @@ src_vector(struct i915_fp_compile *p, /* use vertex format info to map a slot number to a VF attrib */ assert(index < p->vertex_info->num_attribs); + + sem = p->input_semantic[index]; + +#if 1 + switch (sem) { + case TGSI_SEMANTIC_POSITION: + printf("SKIP SEM POS\n"); + /* + assert(p->wpos_tex != -1); + src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); + */ + break; + case TGSI_SEMANTIC_COLOR0: + src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); + break; + case TGSI_SEMANTIC_COLOR1: + src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); + src = swizzle(src, X, Y, Z, ONE); + break; + case TGSI_SEMANTIC_FOG: + src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); + src = swizzle(src, W, W, W, W); + break; + case TGSI_SEMANTIC_TEX0: + case TGSI_SEMANTIC_TEX1: + case TGSI_SEMANTIC_TEX2: + case TGSI_SEMANTIC_TEX3: + case TGSI_SEMANTIC_TEX4: + case TGSI_SEMANTIC_TEX5: + case TGSI_SEMANTIC_TEX6: + case TGSI_SEMANTIC_TEX7: + src = i915_emit_decl(p, REG_TYPE_T, + T_TEX0 + (sem - TGSI_SEMANTIC_TEX0), + D0_CHANNEL_ALL); + break; + default: + i915_program_error(p, "Bad source->Index"); + return 0; + } +#else index = p->vertex_info->slot_to_attrib[index]; switch (index) { @@ -185,6 +225,7 @@ src_vector(struct i915_fp_compile *p, i915_program_error(p, "Bad source->Index"); return 0; } +#endif break; case TGSI_FILE_CONSTANT: @@ -220,9 +261,12 @@ src_vector(struct i915_fp_compile *p, } /* no abs() or post-abs negation */ +#if 0 + /* XXX assertions disabled to allow arbfplight.c to run */ + /* XXX enable these assertions, or fix things */ assert(!source->SrcRegisterExtMod.Absolute); assert(!source->SrcRegisterExtMod.Negate); - +#endif return src; } @@ -848,7 +892,15 @@ i915_translate_instructions(struct i915_fp_compile *p, switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: - /* XXX no-op? */ + if (parse.FullToken.FullDeclaration.Declaration.File + == TGSI_FILE_INPUT) { + /* save input register info for use in src_vector() */ + uint ind, sem; + ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; + /*printf("FS Input DECL [%u] sem %u\n", ind, sem);*/ + p->input_semantic[ind] = sem; + } break; case TGSI_TOKEN_TYPE_IMMEDIATE: @@ -989,6 +1041,7 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) static void i915_find_wpos_space(struct i915_fp_compile *p) { +#if 0 const uint inputs = p->shader->inputs_read | (1 << TGSI_ATTRIB_POS); /*XXX hack*/ uint i; @@ -1005,6 +1058,14 @@ i915_find_wpos_space(struct i915_fp_compile *p) i915_program_error(p, "No free texcoord for wpos value"); } +#else + if (p->shader->input_semantics[0] == TGSI_SEMANTIC_POSITION) { + /* frag shader using the fragment position input */ +#if 0 + assert(0); +#endif + } +#endif } @@ -1018,13 +1079,17 @@ i915_find_wpos_space(struct i915_fp_compile *p) static void i915_fixup_depth_write(struct i915_fp_compile *p) { - if (p->shader->outputs_written & (1 << TGSI_ATTRIB_POS)) { - uint depth = UREG(REG_TYPE_OD, 0); + /* XXX assuming depth is always in output[0] */ + if (p->shader->output_semantics[0] == TGSI_SEMANTIC_DEPTH) { + const uint depth = UREG(REG_TYPE_OD, 0); i915_emit_arith(p, - A0_MOV, - depth, A0_DEST_CHANNEL_W, 0, - swizzle(depth, X, Y, Z, Z), 0, 0); + A0_MOV, /* opcode */ + depth, /* dest reg */ + A0_DEST_CHANNEL_W, /* write mask */ + 0, /* saturate? */ + swizzle(depth, X, Y, Z, Z), /* src0 */ + 0, 0 /* src1, src2 */); } } diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 8d404c55ab..dece697497 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -33,6 +33,7 @@ #include "i915_state.h" #include "i915_reg.h" #include "i915_fpc.h" +#include "pipe/tgsi/exec/tgsi_token.h" /** @@ -42,19 +43,71 @@ */ static void calculate_vertex_layout( struct i915_context *i915 ) { - const uint inputsRead = i915->fs->inputs_read; + const struct pipe_shader_state *fs = i915->fs; const interp_mode colorInterp = i915->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &i915->current.vertex_info; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; boolean needW = 0; + uint i; + boolean texCoords[8]; + memset(texCoords, 0, sizeof(texCoords)); memset(vinfo, 0, sizeof(*vinfo)); /* pos */ draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_3F, INTERP_LINEAR); /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ + for (i = 0; i < fs->num_inputs; i++) { + switch (fs->input_semantics[i]) { + case TGSI_SEMANTIC_POSITION: + break; + case TGSI_SEMANTIC_COLOR0: + front0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, + FORMAT_4UB, colorInterp); + vinfo->hwfmt[0] |= S4_VFMT_COLOR; + break; + case TGSI_SEMANTIC_COLOR1: + assert(0); /* untested */ + front1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, + FORMAT_4UB, colorInterp); + vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; + break; + case TGSI_SEMANTIC_TEX0: + case TGSI_SEMANTIC_TEX1: + case TGSI_SEMANTIC_TEX2: + case TGSI_SEMANTIC_TEX3: + case TGSI_SEMANTIC_TEX4: + case TGSI_SEMANTIC_TEX5: + case TGSI_SEMANTIC_TEX6: + case TGSI_SEMANTIC_TEX7: + { + const uint unit = fs->input_semantics[i] - TGSI_SEMANTIC_TEX0; + uint hwtc; + texCoords[unit] = TRUE; + draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_TEX0 + i, + FORMAT_4F, INTERP_PERSPECTIVE); + hwtc = TEXCOORDFMT_4D; + needW = TRUE; + vinfo->hwfmt[1] |= hwtc << (unit * 4); + } + break; + default: + assert(0); + } + + } + + /* finish up texcoord fields */ + for (i = 0; i < 8; i++) { + if (!texCoords[i]) { + const uint hwtc = TEXCOORDFMT_NOT_PRESENT; + vinfo->hwfmt[1] |= hwtc << (i* 4); + } + } + +#if 0 /* color0 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { front0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, @@ -88,6 +141,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) vinfo->hwfmt[1] |= hwtc << ((i - TGSI_ATTRIB_TEX0) * 4); } } +#endif /* go back and fill in the vertex position info now that we have needW */ if (needW) { @@ -104,11 +158,11 @@ static void calculate_vertex_layout( struct i915_context *i915 ) * the vertex header. */ if (i915->rasterizer->light_twoside) { - if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { + if (front0) { back0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT, colorInterp); } - if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { + if (back0) { back1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, FORMAT_OMIT, colorInterp); } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 1c0ab794f0..65001dfdf9 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -57,7 +57,8 @@ struct pipe_context { const char *(*get_vendor)( struct pipe_context *pipe ); - + int (*get_param)( struct pipe_context *pipe, int param ); + /* * Drawing. diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index b3ee890576..2a8109b10c 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -307,4 +307,11 @@ #define PIPE_QUERY_PRIMITIVES_EMITTED 2 #define PIPE_QUERY_TYPES 3 + +/** + * Pipe capabilities/queries + */ +#define PIPE_PARAM_FS_NEEDS_POS 1 + + #endif diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index d2cc76a59b..6396d49b84 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -50,6 +50,8 @@ #define PIPE_MAX_COLOR_BUFS 8 #define PIPE_MAX_TEXTURE_LEVELS 16 #define PIPE_MAX_FEEDBACK_ATTRIBS 16 +#define PIPE_MAX_SHADER_INPUTS 16 +#define PIPE_MAX_SHADER_OUTPUTS 16 /* fwd decl */ @@ -140,13 +142,14 @@ struct pipe_constant_buffer { struct pipe_shader_state { - unsigned inputs_read; /**< TGSI_ATTRIB_ bits */ - unsigned outputs_written; /**< TGSI_ATTRIB_ bits */ const struct tgsi_token *tokens; void *executable; - uint num_inputs; - uint num_outputs; + /** These fields somewhat constitute the shader "signature" */ + ubyte num_inputs; + ubyte num_outputs; + ubyte input_semantics[PIPE_MAX_SHADER_INPUTS]; + ubyte output_semantics[PIPE_MAX_SHADER_OUTPUTS]; }; struct pipe_depth_stencil_state diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index a56793d683..ebd5530950 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -236,6 +236,15 @@ static const char *softpipe_get_vendor( struct pipe_context *pipe ) return "Tungsten Graphics, Inc."; } +static int softpipe_get_param(struct pipe_context *pipe, uint param) +{ + switch (param) { + case PIPE_PARAM_FS_NEEDS_POS: + return 1; + default: + return 0; + } +} struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, struct softpipe_winsys *softpipe_winsys ) @@ -248,6 +257,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, /* queries */ softpipe->pipe.supported_formats = softpipe_supported_formats; softpipe->pipe.max_texture_size = softpipe_max_texture_size; + softpipe->pipe.get_param = softpipe_get_param; /* state setters */ softpipe->pipe.create_blend_state = softpipe_create_blend_state; diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 9611a2ac99..0dd0eea0b8 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -34,6 +34,7 @@ #include "sp_state.h" #include "pipe/tgsi/exec/tgsi_attribs.h" +#include "pipe/tgsi/exec/tgsi_token.h" /** @@ -43,7 +44,7 @@ */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - const uint inputsRead = softpipe->fs->inputs_read; + const struct pipe_shader_state *fs = softpipe->fs; const interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; @@ -52,57 +53,59 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) memset(vinfo, 0, sizeof(*vinfo)); - /* Need Z if depth test is enabled or the fragment program uses the - * fragment position (XYZW). - */ - if (softpipe->depth_stencil->depth.enabled || - (inputsRead & (1 << TGSI_ATTRIB_POS))) + if (softpipe->depth_stencil->depth.enabled) softpipe->need_z = TRUE; else softpipe->need_z = FALSE; + softpipe->need_w = FALSE; - /* Need W if we do any perspective-corrected interpolation or the - * fragment program uses the fragment position. - */ - if (inputsRead & (1 << TGSI_ATTRIB_POS)) - softpipe->need_w = TRUE; - else - softpipe->need_w = FALSE; - - /* position */ + /* always emit vertex pos */ /* TODO - Figure out if we need to do perspective divide, etc. */ draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F, INTERP_LINEAR); - - /* color0 */ - if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - front0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, - FORMAT_4F, colorInterp); - } - - /* color1 */ - if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - front1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, - FORMAT_4F, colorInterp); - } - /* fog */ - if (inputsRead & (1 << TGSI_ATTRIB_FOG)) { - draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_FOG, - FORMAT_1F, INTERP_PERSPECTIVE); - } - - /* point size */ + for (i = 0; i < fs->num_inputs; i++) { + switch (fs->input_semantics[i]) { + case TGSI_SEMANTIC_POSITION: + /* Need Z if depth test is enabled or the fragment program uses the + * fragment position (XYZW). + */ + softpipe->need_z = TRUE; + softpipe->need_w = TRUE; + break; + case TGSI_SEMANTIC_COLOR0: + front0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, + FORMAT_4F, colorInterp); + break; + case TGSI_SEMANTIC_COLOR1: + front1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, + FORMAT_4F, colorInterp); + break; + case TGSI_SEMANTIC_FOG: + draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_FOG, + FORMAT_1F, INTERP_PERSPECTIVE); + break; #if 0 - /* XXX only emit if drawing points or front/back polygon mode is point mode */ - draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POINTSIZE, - FORMAT_4F, INTERP_CONSTANT); + case TGSI_SEMANTIC_PSIZE: + /* XXX only emit if drawing points or front/back polygon mode + * is point mode + */ + draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POINTSIZE, + FORMAT_4F, INTERP_CONSTANT); + break; #endif - - /* texcoords and varying vars */ - for (i = TGSI_ATTRIB_TEX0; i < TGSI_ATTRIB_VAR7; i++) { - if (inputsRead & (1 << i)) { + /*case TGSI_SEMANTIC_TEXCOORD:*/ + case TGSI_SEMANTIC_TEX0: + draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_TEX0, + FORMAT_4F, INTERP_PERSPECTIVE); + softpipe->need_w = TRUE; + break; + case TGSI_SEMANTIC_OTHER: draw_emit_vertex_attr(vinfo, i, FORMAT_4F, INTERP_PERSPECTIVE); softpipe->need_w = TRUE; + break; + + default: + assert(0); } } @@ -113,12 +116,11 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) * the vertex header. */ if (softpipe->rasterizer->light_twoside) { - if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { + if (front0) { back0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT, colorInterp); } - - if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { + if (back0) { back1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, FORMAT_OMIT, colorInterp); } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.c b/src/mesa/pipe/tgsi/exec/tgsi_build.c index 20e4cf17f0..1320872c64 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_build.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_build.c @@ -325,7 +325,7 @@ tgsi_build_declaration_semantic( { struct tgsi_declaration_semantic ds; - assert( semantic_name <= TGSI_SEMANTIC_COLOR ); + assert( semantic_name <= TGSI_SEMANTIC_COUNT ); assert( semantic_index <= 0xFFFF ); ds = tgsi_default_declaration_semantic(); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index e6e99d9d75..0a47ad2a8c 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -202,13 +202,27 @@ static const char *TGSI_INTERPOLATES_SHORT[] = static const char *TGSI_SEMANTICS[] = { "SEMANTIC_DEPTH", - "SEMANTIC_COLOR" + "SEMANTIC_COLOR0", + "SEMANTIC_COLOR1", + "SEMANTIC_COLOR0B", + "SEMANTIC_COLOR1B", + "SEMANTIC_POSITION", + "SEMANTIC_FOG", + "SEMANTIC_OTHER," + "SEMANTIC_TEX0", }; static const char *TGSI_SEMANTICS_SHORT[] = { "DEPTH", - "COLOR" + "COLOR0", + "COLOR1", + "COLOR0B", + "COLOR1B", + "POSITION", + "FOG", + "OTHER", + "TEX0" }; static const char *TGSI_IMMS[] = diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h index ca53071a60..a642ba131a 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -73,11 +73,11 @@ struct tgsi_declaration { unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ unsigned Size : 8; /* UINT */ - unsigned File : 4; /* TGSI_FILE_ */ - unsigned Declare : 4; /* TGSI_DECLARE_ */ - unsigned UsageMask : 4; /* TGSI_WRITEMASK_ */ - unsigned Interpolate : 1; /* BOOL */ - unsigned Semantic : 1; /* BOOL */ + unsigned File : 4; /* one of TGSI_FILE_x */ + unsigned Declare : 4; /* one of TGSI_DECLARE_x */ + unsigned UsageMask : 4; /* bitmask of TGSI_WRITEMASK_x flags */ + unsigned Interpolate : 1; /* BOOL, any interpolation info? */ + unsigned Semantic : 1; /* BOOL, any semantic info? */ unsigned Padding : 5; unsigned Extended : 1; /* BOOL */ }; @@ -103,12 +103,27 @@ struct tgsi_declaration_interpolation unsigned Padding : 28; }; -#define TGSI_SEMANTIC_DEPTH 0 -#define TGSI_SEMANTIC_COLOR 1 +#define TGSI_SEMANTIC_DEPTH 0 +#define TGSI_SEMANTIC_COLOR0 1 +#define TGSI_SEMANTIC_COLOR1 2 +#define TGSI_SEMANTIC_COLOR0B 3 /**< back-face primary color */ +#define TGSI_SEMANTIC_COLOR1B 4 /**< back-face secondary color */ +#define TGSI_SEMANTIC_POSITION 5 +#define TGSI_SEMANTIC_FOG 6 +#define TGSI_SEMANTIC_OTHER 7 /* XXX temp */ +#define TGSI_SEMANTIC_TEX0 8 +#define TGSI_SEMANTIC_TEX1 9 +#define TGSI_SEMANTIC_TEX2 10 +#define TGSI_SEMANTIC_TEX3 11 +#define TGSI_SEMANTIC_TEX4 12 +#define TGSI_SEMANTIC_TEX5 13 +#define TGSI_SEMANTIC_TEX6 14 +#define TGSI_SEMANTIC_TEX7 15 +#define TGSI_SEMANTIC_COUNT 16 /**< number of semantic values */ struct tgsi_declaration_semantic { - unsigned SemanticName : 8; /* TGSI_SEMANTIC_ */ + unsigned SemanticName : 8; /* one of TGSI_SEMANTIC_ */ unsigned SemanticIndex : 16; /* UINT */ unsigned Padding : 8; }; diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 1f8d937bc6..fb8365aab5 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -3,9 +3,9 @@ #include "pipe/tgsi/exec/tgsi_attribs.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" -#define TGSI_DEBUG 1 - +#define TGSI_DEBUG 0 +#if 0 /** * Convert a VERT_ATTRIB_x to a TGSI_ATTRIB_y */ @@ -209,9 +209,10 @@ tgsi_mesa_translate_fragment_output(GLuint attrib) return 0; } } +#endif -#if 01 +#if 0 uint tgsi_mesa_translate_vertex_input_mask(GLbitfield mask) { @@ -225,7 +226,6 @@ tgsi_mesa_translate_vertex_input_mask(GLbitfield mask) } return tgsiMask; } -#endif uint tgsi_mesa_translate_vertex_output_mask(GLbitfield mask) @@ -271,6 +271,7 @@ tgsi_mesa_translate_fragment_output_mask(GLbitfield mask) } +#endif @@ -319,12 +320,16 @@ map_register_file_index( GLuint processor, GLuint file, GLuint index, +#if 0 GLbitfield usage_bitmask, +#endif const GLuint inputMapping[], const GLuint outputMapping[]) { GLuint mapped_index; +#if 0 GLuint i; +#endif assert(processor == TGSI_PROCESSOR_FRAGMENT || processor == TGSI_PROCESSOR_VERTEX); @@ -345,7 +350,8 @@ map_register_file_index( inputMapping[index]); return inputMapping[index]; } - + assert(0); +#if 0 assert( usage_bitmask & (1 << index) ); mapped_index = 0; for( i = 0; i < index; i++ ) { @@ -354,6 +360,7 @@ map_register_file_index( } } printf("Map %d input %d to %d\n", processor, index, mapped_index); +#endif break; case TGSI_FILE_OUTPUT: @@ -375,14 +382,17 @@ map_register_file_index( else { /* vertex output slots are tightly packed, find mapped pos */ /* mapped_index = VERT_RESULT_x */ +#if 0 mapped_index = 0; for( i = 0; i < index; i++ ) { if( usage_bitmask & (1 << i) ) { mapped_index++; } } - printf("Map VP output from %d to %d\n", index, mapped_index); assert(outputMapping[index] == mapped_index); +#endif + mapped_index = outputMapping[index]; + printf("Map VP output from %d to %d\n", index, mapped_index); } break; @@ -452,8 +462,10 @@ static GLboolean compile_instruction( const struct prog_instruction *inst, struct tgsi_full_instruction *fullinst, +#if 0 GLuint inputs_read, GLuint outputs_written, +#endif const GLuint inputMapping[], const GLuint outputMapping[], GLuint preamble_size, @@ -475,8 +487,14 @@ compile_instruction( processor, fulldst->DstRegister.File, inst->DstReg.Index, +#if 0 outputs_written, +#endif +#if 0 NULL, +#else + inputMapping, +#endif outputMapping ); fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); @@ -490,7 +508,9 @@ compile_instruction( processor, fullsrc->SrcRegister.File, inst->SrcReg[i].Index, +#if 0 inputs_read, +#endif inputMapping, outputMapping ); @@ -766,10 +786,10 @@ compile_instruction( static struct tgsi_full_declaration make_frag_input_decl( - GLuint first, - GLuint last, + GLuint index, GLuint interpolate, - GLuint usage_mask ) + GLuint usage_mask, + GLuint semantic_name ) { struct tgsi_full_declaration decl; @@ -777,9 +797,11 @@ make_frag_input_decl( decl.Declaration.File = TGSI_FILE_INPUT; decl.Declaration.Declare = TGSI_DECLARE_RANGE; decl.Declaration.UsageMask = usage_mask; + decl.Declaration.Semantic = 1; decl.Declaration.Interpolate = 1; - decl.u.DeclarationRange.First = first; - decl.u.DeclarationRange.Last = last; + decl.u.DeclarationRange.First = index; + decl.u.DeclarationRange.Last = index; + decl.Semantic.SemanticName = semantic_name; decl.Interpolation.Interpolate = interpolate; return decl; @@ -809,15 +831,20 @@ make_frag_output_decl( /** * Convert Mesa fragment program to TGSI format. - * \param inputMapping array to map original Mesa fragment program inputs - * registers to TGSI generic input indexes - * \param interpMode array[FRAG_ATTRIB_x] of TGSI_INTERPOLATE_LINEAR/PERSP. + * \param inputMapping maps Mesa fragment program inputs to TGSI generic + * input indexes + * \param inputSemantic the TGSI_SEMANTIC flag for each input + * \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input + * \param outputMapping maps Mesa fragment program outputs to TGSI + * generic outputs * */ GLboolean tgsi_mesa_compile_fp_program( const struct gl_fragment_program *program, + GLuint numInputs, const GLuint inputMapping[], + const ubyte inputSemantic[], const GLuint interpMode[], const GLuint outputMapping[], struct tgsi_token *tokens, @@ -831,9 +858,9 @@ tgsi_mesa_compile_fp_program( /* struct tgsi_full_dst_register *fulldst; struct tgsi_full_src_register *fullsrc; - */ GLuint inputs_read; GLboolean reads_wpos; + */ GLuint preamble_size = 0; *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -846,6 +873,7 @@ tgsi_mesa_compile_fp_program( ti = 3; +#if 0 reads_wpos = program->Base.InputsRead & (1 << FRAG_ATTRIB_WPOS); inputs_read = program->Base.InputsRead | (1 << FRAG_ATTRIB_WPOS); @@ -856,10 +884,10 @@ tgsi_mesa_compile_fp_program( /* Fragment position. */ if( reads_wpos ) { fulldecl = make_frag_input_decl( - 0, 0, TGSI_INTERPOLATE_CONSTANT, - TGSI_WRITEMASK_XY ); + TGSI_WRITEMASK_XY, + TGSI_SEMANTIC_POSITION ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -869,10 +897,10 @@ tgsi_mesa_compile_fp_program( /* Fragment zw. */ fulldecl = make_frag_input_decl( - 0, 0, TGSI_INTERPOLATE_LINEAR, - reads_wpos ? TGSI_WRITEMASK_ZW : TGSI_WRITEMASK_Z ); + reads_wpos ? TGSI_WRITEMASK_ZW : TGSI_WRITEMASK_Z, + TGSI_SEMANTIC_POSITION ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -884,15 +912,55 @@ tgsi_mesa_compile_fp_program( if( inputs_read & (1 << i) ) { count++; fulldecl = make_frag_input_decl(count, - count, interpMode[i], - TGSI_WRITEMASK_XYZW ); + TGSI_WRITEMASK_XYZW, + inputSemantic[count] ); ti += tgsi_build_full_declaration(&fulldecl, &tokens[ti], header, maxTokens - ti ); } } +#else + + for (i = 0; i < numInputs; i++) { + switch (inputSemantic[i]) { + case TGSI_SEMANTIC_POSITION: + /* Fragment XY pos */ + fulldecl = make_frag_input_decl(i, + TGSI_INTERPOLATE_CONSTANT, + TGSI_WRITEMASK_XY, + TGSI_SEMANTIC_POSITION ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + /* Fragment ZW pos */ + fulldecl = make_frag_input_decl(i, + TGSI_INTERPOLATE_LINEAR, + TGSI_WRITEMASK_ZW, + TGSI_SEMANTIC_POSITION ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + default: + fulldecl = make_frag_input_decl(i, + interpMode[i], + TGSI_WRITEMASK_XYZW, + inputSemantic[i] ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + } + } +#endif + /* * Declare output attributes. @@ -914,7 +982,7 @@ tgsi_mesa_compile_fp_program( if( program->Base.OutputsWritten & (1 << FRAG_RESULT_COLR) ) { fulldecl = make_frag_output_decl( 1, - TGSI_SEMANTIC_COLOR, + TGSI_SEMANTIC_COLOR0, TGSI_WRITEMASK_XYZW ); ti += tgsi_build_full_declaration( &fulldecl, @@ -956,8 +1024,10 @@ tgsi_mesa_compile_fp_program( if( compile_instruction( &program->Base.Instructions[i], &fullinst, +#if 0 inputs_read, ~0, /*outputs_written*/ +#endif inputMapping, outputMapping, preamble_size, @@ -992,10 +1062,12 @@ tgsi_mesa_compile_vp_program( struct tgsi_header *header; struct tgsi_processor *processor; struct tgsi_full_instruction fullinst; +#if 0 GLuint inputs_read = ~0; GLuint outputs_written; outputs_written = program->Base.OutputsWritten; +#endif *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -1011,8 +1083,10 @@ tgsi_mesa_compile_vp_program( if( compile_instruction( &program->Base.Instructions[i], &fullinst, +#if 0 inputs_read, outputs_written, +#endif inputMapping, outputMapping, 0, diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index 017cfce72e..8105e9e738 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -10,7 +10,9 @@ struct tgsi_token; GLboolean tgsi_mesa_compile_fp_program( const struct gl_fragment_program *program, + GLuint numInputs, const GLuint inputMapping[], + const ubyte inputSemantic[], const GLuint interpMode[], const GLuint outputMapping[], struct tgsi_token *tokens, diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 6dd576a57c..94b69c8df7 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -27,6 +27,7 @@ /* * Authors: * Keith Whitwell + * Brian Paul */ #include "shader/prog_parameter.h" @@ -42,55 +43,121 @@ #include "st_atom.h" #include "st_program.h" + #define TGSI_DEBUG 1 -static void compile_fs( struct st_context *st ) + +/** + * Translate a Mesa fragment shader into a TGSI shader. + * \return pointer to cached pipe_shader object. + */ +struct pipe_shader_state * +st_translate_fragment_shader(struct st_context *st, + struct st_fragment_program *stfp) { - /* Map FRAG_RESULT_COLR to output 1, map FRAG_RESULT_DEPR to output 0 */ - static const GLuint outputMapping[2] = {1, 0}; - struct st_fragment_program *fp = st->fp; + GLuint outputMapping[FRAG_RESULT_MAX]; + GLuint inputMapping[PIPE_MAX_SHADER_INPUTS]; struct pipe_shader_state fs; struct pipe_shader_state *cached; GLuint interpMode[16]; /* XXX size? */ GLuint i; + GLbitfield inputsRead = stfp->Base.Base.InputsRead; + + /* Check if all fragment programs need the fragment position (in order + * to do perspective-corrected interpolation). + */ + if (st->pipe->get_param(st->pipe, PIPE_PARAM_FS_NEEDS_POS)) + inputsRead |= FRAG_BIT_WPOS; + + memset(&fs, 0, sizeof(fs)); for (i = 0; i < 16; i++) { - if (fp->Base.Base.InputsRead & (1 << i)) { - if (i == FRAG_ATTRIB_COL0 || i == FRAG_ATTRIB_COL1) { - interpMode[i] = TGSI_INTERPOLATE_LINEAR; + if (inputsRead & (1 << i)) { + inputMapping[i] = fs.num_inputs; + + switch (i) { + case FRAG_ATTRIB_WPOS: + fs.input_semantics[fs.num_inputs] = TGSI_SEMANTIC_POSITION; + interpMode[fs.num_inputs] = TGSI_INTERPOLATE_CONSTANT; + break; + case FRAG_ATTRIB_COL0: + fs.input_semantics[fs.num_inputs] = TGSI_SEMANTIC_COLOR0; + interpMode[fs.num_inputs] = TGSI_INTERPOLATE_LINEAR; + break; + case FRAG_ATTRIB_COL1: + fs.input_semantics[fs.num_inputs] = TGSI_SEMANTIC_COLOR1; + interpMode[fs.num_inputs] = TGSI_INTERPOLATE_LINEAR; + break; + case FRAG_ATTRIB_TEX0: + fs.input_semantics[fs.num_inputs] = TGSI_SEMANTIC_TEX0; + interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; + break; + default: + assert(0); } - else { - interpMode[i] = TGSI_INTERPOLATE_PERSPECTIVE; + + fs.num_inputs++; + } + } + + /* + * Outputs + */ + for (i = 0; i < FRAG_RESULT_MAX; i++) { + if (stfp->Base.Base.OutputsWritten & (1 << i)) { + switch (i) { + case FRAG_RESULT_DEPR: + fs.output_semantics[fs.num_outputs] = TGSI_SEMANTIC_DEPTH; + outputMapping[i] = fs.num_outputs; + break; + case FRAG_RESULT_COLR: + fs.output_semantics[fs.num_outputs] = TGSI_SEMANTIC_COLOR0; + outputMapping[i] = fs.num_outputs; + break; + default: + assert(0); } + fs.num_outputs++; } } /* XXX: fix static allocation of tokens: */ - tgsi_mesa_compile_fp_program( &fp->Base, NULL, interpMode, + tgsi_mesa_compile_fp_program( &stfp->Base, + fs.num_inputs, + inputMapping, + fs.input_semantics, + interpMode, outputMapping, - fp->tokens, ST_FP_MAX_TOKENS ); + stfp->tokens, ST_FP_MAX_TOKENS ); - memset(&fs, 0, sizeof(fs)); +#if 0 fs.inputs_read - = tgsi_mesa_translate_fragment_input_mask(fp->Base.Base.InputsRead); + = tgsi_mesa_translate_fragment_input_mask(stfp->Base.Base.InputsRead); +#endif +#if 0 fs.outputs_written - = tgsi_mesa_translate_fragment_output_mask(fp->Base.Base.OutputsWritten); - fs.tokens = &fp->tokens[0]; + = tgsi_mesa_translate_fragment_output_mask(stfp->Base.Base.OutputsWritten); +#endif + + fs.tokens = &stfp->tokens[0]; + cached = st_cached_fs_state(st, &fs); - fp->fsx = cached; + stfp->fs = cached; if (TGSI_DEBUG) - tgsi_dump( fp->tokens, 0/*TGSI_DUMP_VERBOSE*/ ); + tgsi_dump( stfp->tokens, 0/*TGSI_DUMP_VERBOSE*/ ); + + stfp->dirty = 0; - fp->dirty = 0; + return cached; } static void update_fs( struct st_context *st ) { - struct st_fragment_program *fp = NULL; + struct st_fragment_program *stfp = NULL; /* find active shader and params. Changes to this Mesa state * should be covered by ST_NEW_FRAGMENT_PROGRAM, thanks to the @@ -101,21 +168,21 @@ static void update_fs( struct st_context *st ) st->ctx->Shader.CurrentProgram->FragmentProgram) { struct gl_fragment_program *f = st->ctx->Shader.CurrentProgram->FragmentProgram; - fp = st_fragment_program(f); + stfp = st_fragment_program(f); } else { assert(st->ctx->FragmentProgram._Current); - fp = st_fragment_program(st->ctx->FragmentProgram._Current); + stfp = st_fragment_program(st->ctx->FragmentProgram._Current); } - /* translate shader to TGSI format */ - if (st->fp != fp || fp->dirty) { - st->fp = fp; + /* if new binding, or shader has changed */ + if (st->fp != stfp || stfp->dirty) { + /* Bind the program */ + st->fp = stfp; - if (fp->dirty) - compile_fs( st ); + if (stfp->dirty) + st->state.fs = st_translate_fragment_shader( st, st->fp ); - st->state.fs = fp->fsx; st->pipe->bind_fs_state(st->pipe, st->state.fs); } } diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 322fabc456..cf9dd810e9 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -49,14 +49,15 @@ #define TGSI_DEBUG 1 - - /** - * Translate Mesa shader to TGSI format + * Translate a Mesa vertex shader into a TGSI shader. + * \return pointer to cached pipe_shader object. */ -static void compile_vs( struct st_context *st ) +struct pipe_shader_state * +st_translate_vertex_shader(struct st_context *st, + struct st_vertex_program *stvp) { - struct st_vertex_program *vp = st->vp; + GLuint outputMapping[PIPE_MAX_SHADER_INPUTS]; struct pipe_shader_state vs; struct pipe_shader_state *cached; GLuint i; @@ -69,20 +70,56 @@ static void compile_vs( struct st_context *st ) * values and TGSI generic input indexes. */ for (i = 0; i < MAX_VERTEX_PROGRAM_ATTRIBS; i++) { - if (vp->Base.Base.InputsRead & (1 << i)) { - vp->input_to_index[i] = vs.num_inputs; - vp->index_to_input[vs.num_inputs] = i; + if (stvp->Base.Base.InputsRead & (1 << i)) { + stvp->input_to_index[i] = vs.num_inputs; + stvp->index_to_input[vs.num_inputs] = i; + switch (i) { + case VERT_ATTRIB_POS: + vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_POSITION; + break; + case VERT_ATTRIB_COLOR0: + vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0; + break; + case VERT_ATTRIB_COLOR1: + vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1; + break; + default: + vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_OTHER; + } vs.num_inputs++; } } /* - * Determine output register mapping. + * Determine number of outputs and the register mapping. */ for (i = 0; i < VERT_RESULT_MAX; i++) { - if (vp->Base.Base.OutputsWritten & (1 << i)) { - vp->output_to_index[i] = vs.num_outputs; - vp->index_to_output[vs.num_outputs] = i; + if (stvp->Base.Base.OutputsWritten & (1 << i)) { +#if 0 + stvp->output_to_index[i] = vs.num_outputs; + stvp->index_to_output[vs.num_outputs] = i; +#endif + outputMapping[i] = vs.num_outputs; + + switch (i) { + case VERT_RESULT_HPOS: + vs.output_semantics[vs.num_outputs] = TGSI_SEMANTIC_POSITION; + break; + case VERT_RESULT_COL0: + vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0; + break; + case VERT_RESULT_COL1: + vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1; + break; + case VERT_RESULT_BFC0: + vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0B; + break; + case VERT_RESULT_BFC1: + vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1B; + break; + default: + vs.output_semantics[vs.num_outputs] = TGSI_SEMANTIC_OTHER; + } vs.num_outputs++; } } @@ -90,43 +127,50 @@ static void compile_vs( struct st_context *st ) /* XXX: fix static allocation of tokens: */ - tgsi_mesa_compile_vp_program( &vp->Base, - vp->input_to_index, - vp->output_to_index, - vp->tokens, ST_FP_MAX_TOKENS ); + tgsi_mesa_compile_vp_program( &stvp->Base, + stvp->input_to_index, +#if 0 + stvp->output_to_index, +#else + outputMapping, +#endif + stvp->tokens, ST_FP_MAX_TOKENS ); -#if 01 +#if 0 vs.inputs_read - = tgsi_mesa_translate_vertex_input_mask(vp->Base.Base.InputsRead); + = tgsi_mesa_translate_vertex_input_mask(stvp->Base.Base.InputsRead); #endif +#if 0 vs.outputs_written - = tgsi_mesa_translate_vertex_output_mask(vp->Base.Base.OutputsWritten); + = tgsi_mesa_translate_vertex_output_mask(stvp->Base.Base.OutputsWritten); +#endif - vs.tokens = &vp->tokens[0]; + vs.tokens = &stvp->tokens[0]; cached = st_cached_vs_state(st, &vs); - - vp->vs = cached; + stvp->vs = cached; if (TGSI_DEBUG) - tgsi_dump( vp->tokens, 0 ); + tgsi_dump( stvp->tokens, 0 ); #if defined(USE_X86_ASM) || defined(SLANG_X86) - if (vp->sse2_program.csr == vp->sse2_program.store) - tgsi_emit_sse2( vp->tokens, &vp->sse2_program ); + if (stvp->sse2_program.csr == stvp->sse2_program.store) + tgsi_emit_sse2( stvp->tokens, &stvp->sse2_program ); if (!cached->executable) - cached->executable = (void *) x86_get_func( &vp->sse2_program ); + cached->executable = (void *) x86_get_func( &stvp->sse2_program ); #endif - vp->dirty = 0; + stvp->dirty = 0; + + return cached; } static void update_vs( struct st_context *st ) { - struct st_vertex_program *vp; + struct st_vertex_program *stvp; /* find active shader and params -- Should be covered by * ST_NEW_VERTEX_PROGRAM @@ -136,20 +180,20 @@ static void update_vs( struct st_context *st ) st->ctx->Shader.CurrentProgram->VertexProgram) { struct gl_vertex_program *f = st->ctx->Shader.CurrentProgram->VertexProgram; - vp = st_vertex_program(f); + stvp = st_vertex_program(f); } else { assert(st->ctx->VertexProgram._Current); - vp = st_vertex_program(st->ctx->VertexProgram._Current); + stvp = st_vertex_program(st->ctx->VertexProgram._Current); } - if (st->vp != vp || vp->dirty) { - st->vp = vp; + if (st->vp != stvp || stvp->dirty) { + /* Bind the vertex program */ + st->vp = stvp; - if (vp->dirty) - compile_vs( st ); + if (stvp->dirty) + st->state.vs = st_translate_vertex_shader( st, st->vp ); - st->state.vs = st->vp->vs; st->pipe->bind_vs_state(st->pipe, st->state.vs); } } diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 5d5efd9eae..ee70ce3320 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -121,11 +121,9 @@ is_depth_stencil_format(GLuint pipeFormat) static struct st_fragment_program * make_frag_shader(struct st_context *st) { - static const GLuint outputMapping[] = { 1, 0 }; GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; struct gl_program *p; - GLboolean b; GLuint interpMode[16]; GLuint i; @@ -157,11 +155,7 @@ make_frag_shader(struct st_context *st) p->OutputsWritten = (1 << FRAG_RESULT_COLR); stfp = (struct st_fragment_program *) p; - /* compile into tgsi format */ - b = tgsi_mesa_compile_fp_program(&stfp->Base, NULL, interpMode, - outputMapping, - stfp->tokens, ST_FP_MAX_TOKENS); - assert(b); + st_translate_fragment_shader(st, stfp); return stfp; } @@ -174,15 +168,9 @@ make_frag_shader(struct st_context *st) static struct st_vertex_program * make_vertex_shader(struct st_context *st) { - /* Map VERT_ATTRIB_POS to 0, VERT_ATTRIB_COLOR0 to 1 */ - static const GLuint inputMapping[4] = { 0, 0, 0, 1 }; - /* Map VERT_RESULT_HPOS to 0, VERT_RESULT_COL0 to 1 */ - static const GLuint outputMapping[2] = { 0, 1 }; - GLcontext *ctx = st->ctx; struct st_vertex_program *stvp; struct gl_program *p; - GLboolean b; p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); if (!p) @@ -215,12 +203,8 @@ make_vertex_shader(struct st_context *st) (1 << VERT_RESULT_HPOS)); stvp = (struct st_vertex_program *) p; - /* compile into tgsi format */ - b = tgsi_mesa_compile_vp_program(&stvp->Base, - inputMapping, - outputMapping, - stvp->tokens, ST_FP_MAX_TOKENS); - assert(b); + st_translate_vertex_shader(st, stvp); + assert(stvp->vs); return stvp; } @@ -361,33 +345,19 @@ clear_with_quad(GLcontext *ctx, /* fragment shader state: color pass-through program */ { static struct st_fragment_program *stfp = NULL; - struct pipe_shader_state fs; - const struct pipe_shader_state *cached; if (!stfp) { stfp = make_frag_shader(st); } - memset(&fs, 0, sizeof(fs)); - fs.inputs_read = tgsi_mesa_translate_fragment_input_mask(stfp->Base.Base.InputsRead); - fs.outputs_written = tgsi_mesa_translate_fragment_output_mask(stfp->Base.Base.OutputsWritten); - fs.tokens = &stfp->tokens[0]; - cached = st_cached_fs_state(st, &fs); - pipe->bind_fs_state(pipe, cached); + pipe->bind_fs_state(pipe, stfp->fs); } /* vertex shader state: color/position pass-through */ { static struct st_vertex_program *stvp = NULL; - struct pipe_shader_state vs; - const struct pipe_shader_state *cached; if (!stvp) { stvp = make_vertex_shader(st); } - memset(&vs, 0, sizeof(vs)); - vs.inputs_read = stvp->Base.Base.InputsRead; - vs.outputs_written = stvp->Base.Base.OutputsWritten; - vs.tokens = &stvp->tokens[0]; - cached = st_cached_vs_state(st, &vs); - pipe->bind_vs_state(pipe, cached); + pipe->bind_vs_state(pipe, stvp->vs); } /* viewport state: viewport matching window dims */ @@ -522,12 +492,15 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) assert(strb->surface->format); +#if 01 if (ctx->Scissor.Enabled || (isDS && ctx->DrawBuffer->Visual.stencilBits > 0)) { /* scissoring or we have a combined depth/stencil buffer */ clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_FALSE); } - else { + else +#endif + { /* simple clear of whole buffer */ GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 0fd728c930..d4f260ee54 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -56,11 +56,9 @@ static struct st_fragment_program * make_fragment_shader(struct st_context *st) { - static const GLuint outputMapping[2] = { 1, 0 }; GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; struct gl_program *p; - GLboolean b; GLuint interpMode[16]; GLuint i; @@ -94,11 +92,7 @@ make_fragment_shader(struct st_context *st) p->OutputsWritten = (1 << FRAG_RESULT_COLR); stfp = (struct st_fragment_program *) p; - /* compile into tgsi format */ - b = tgsi_mesa_compile_fp_program(&stfp->Base, NULL, interpMode, - outputMapping, - stfp->tokens, ST_FP_MAX_TOKENS); - assert(b); + st_translate_fragment_shader(st, stfp); return stfp; } @@ -112,11 +106,9 @@ static struct st_vertex_program * make_vertex_shader(struct st_context *st) { /* Map VERT_RESULT_HPOS to 0, VERT_RESULT_TEX0 to 1 */ - static const GLuint outputMapping[] = { 0, 0, 0, 0, 1 }; GLcontext *ctx = st->ctx; struct st_vertex_program *stvp; struct gl_program *p; - GLboolean b; p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); if (!p) @@ -149,11 +141,7 @@ make_vertex_shader(struct st_context *st) (1 << VERT_RESULT_HPOS)); stvp = (struct st_vertex_program *) p; - /* compile into tgsi format */ - b = tgsi_mesa_compile_vp_program(&stvp->Base, NULL, - outputMapping, - stvp->tokens, ST_FP_MAX_TOKENS); - assert(b); + st_translate_vertex_shader(st, stvp); return stvp; } @@ -339,32 +327,19 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* fragment shader state: TEX lookup program */ { static struct st_fragment_program *stfp = NULL; - struct pipe_shader_state fs; - struct pipe_shader_state *cached; if (!stfp) { stfp = make_fragment_shader(ctx->st); } - memset(&fs, 0, sizeof(fs)); - fs.inputs_read = stfp->Base.Base.InputsRead; - fs.tokens = &stfp->tokens[0]; - cached = st_cached_fs_state(ctx->st, &fs); - pipe->bind_fs_state(pipe, cached); + pipe->bind_fs_state(pipe, stfp->fs); } /* vertex shader state: position + texcoord pass-through */ { static struct st_vertex_program *stvp = NULL; - struct pipe_shader_state vs; - struct pipe_shader_state *cached; if (!stvp) { stvp = make_vertex_shader(ctx->st); } - memset(&vs, 0, sizeof(vs)); - vs.inputs_read = stvp->Base.Base.InputsRead; - vs.outputs_written = stvp->Base.Base.OutputsWritten; - vs.tokens = &stvp->tokens[0]; - cached = st_cached_vs_state(ctx->st, &vs); - pipe->bind_vs_state(pipe, cached); + pipe->bind_vs_state(pipe, stvp->vs); } /* texture sampling state: */ diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 98efe1a10b..5245535a65 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -53,6 +53,7 @@ static void setup_vertex_attribs(GLcontext *ctx) { struct pipe_context *pipe = ctx->st->pipe; +#if 0 const uint inputAttrs = ctx->st->state.vs->inputs_read; uint attr; @@ -77,6 +78,9 @@ setup_vertex_attribs(GLcontext *ctx) pipe->set_vertex_element(pipe, attr, &velement); } } +#else + assert(0); +#endif } @@ -84,7 +88,7 @@ static void setup_feedback(GLcontext *ctx) { struct pipe_context *pipe = ctx->st->pipe; - const uint outputAttrs = ctx->st->state.vs->outputs_written; + const struct pipe_shader_state *vs = ctx->st->state.vs; struct pipe_feedback_state feedback; uint i; @@ -94,8 +98,8 @@ setup_feedback(GLcontext *ctx) feedback.discard = 1; feedback.num_attribs = 0; - for (i = 0; i < TGSI_ATTRIB_VAR0; i++) { - if ((1 << i) & outputAttrs) { + for (i = 0; i < vs->num_outputs; i++) { + if (1/***(1 << i) & outputAttrs***/) { feedback.attrib[feedback.num_attribs] = i; feedback.size[feedback.num_attribs] = 4; feedback.num_attribs++; @@ -306,6 +310,7 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) PIPE_BUFFER_FLAG_READ); /* extract values and update rasterpos state */ +#if 0 /* XXX update */ { const uint outputAttrs = ctx->st->state.vs->outputs_written; const float *pos, *color0, *color1, *tex0; @@ -333,7 +338,7 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) update_rasterpos(ctx, pos, color0, color1, tex0); } - +#endif /* free vertex feedback buffer */ pipe->winsys->buffer_unmap(pipe->winsys, fb_buf.buffer); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 7075db82e9..238ade00ac 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -356,7 +356,7 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, velement.vertex_buffer_index = 0; velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; velement.dst_offset = 0; - pipe->set_vertex_element(pipe, attribs[i], &velement); + pipe->set_vertex_element(pipe, i/**attribs[i]**/, &velement); } /* draw */ @@ -411,7 +411,7 @@ st_feedback_draw_vbo(GLcontext *ctx, update_default_attribs_buffer(ctx); - +#if 0 /* this must be after state validation */ attrsNeeded = ctx->st->state.vs->inputs_read; @@ -480,7 +480,9 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_mapped_vertex_buffer(draw, attr, map); } } - +#else + assert(0); +#endif if (ib) { unsigned indexSize; diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 68ceba4d78..4945141d15 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -45,16 +45,15 @@ struct st_fragment_program { struct gl_fragment_program Base; GLboolean error; /* If program is malformed for any reason. */ - - GLuint id; /* String id, for tracking - * ProgramStringNotify changes. - */ + GLuint id; /**< String id, for tracking ProgramStringNotify changes. */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; - const struct pipe_shader_state *fsx; + /** Pointer to the corresponding cached shader */ + const struct pipe_shader_state *fs; + GLuint param_state; }; @@ -63,16 +62,17 @@ struct st_vertex_program { struct gl_vertex_program Base; /**< The Mesa vertex program */ GLboolean error; /**< Set if program is malformed for any reason. */ - - GLuint id; /**< String id, for tracking ProgramStringNotify changes. */ + GLuint id; /**< String id, for tracking ProgramStringNotify changes. */ /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ GLuint input_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ GLuint index_to_input[MAX_VERTEX_PROGRAM_ATTRIBS]; +#if 0 GLuint output_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; GLuint index_to_output[MAX_VERTEX_PROGRAM_ATTRIBS]; +#endif /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; @@ -82,7 +82,9 @@ struct st_vertex_program struct x86_function sse2_program; #endif + /** Pointer to the corresponding cached shader */ const struct pipe_shader_state *vs; + GLuint param_state; }; @@ -102,4 +104,15 @@ st_vertex_program( struct gl_vertex_program *vp ) return (struct st_vertex_program *)vp; } + +extern struct pipe_shader_state * +st_translate_fragment_shader(struct st_context *st, + struct st_fragment_program *fp); + + +extern struct pipe_shader_state * +st_translate_vertex_shader(struct st_context *st, + struct st_vertex_program *vp); + + #endif -- cgit v1.2.3 From daf5b0f41baa50951e7c2f9ea5cd90b119085a7f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 20 Sep 2007 07:50:33 -0400 Subject: Switch fragment/vertex shaders to the new caching semantics. Allow driver custom allocation within cached objects. The shaders are currently twiced (by cso layer and by the program itself). --- src/mesa/pipe/cso_cache/cso_cache.h | 10 +++++ src/mesa/pipe/failover/fo_context.h | 4 +- src/mesa/pipe/failover/fo_state.c | 75 +++++++++++++++++++++++++++---- src/mesa/pipe/failover/fo_state_emit.c | 6 ++- src/mesa/pipe/i915simple/i915_state.c | 28 +++++------- src/mesa/pipe/p_context.h | 23 ++++------ src/mesa/pipe/softpipe/sp_state.h | 11 ++--- src/mesa/pipe/softpipe/sp_state_fs.c | 27 +++++------ src/mesa/state_tracker/st_atom_fs.c | 12 ++--- src/mesa/state_tracker/st_atom_vs.c | 18 ++++---- src/mesa/state_tracker/st_cache.c | 36 ++++++++------- src/mesa/state_tracker/st_cache.h | 12 ++--- src/mesa/state_tracker/st_cb_clear.c | 8 ++-- src/mesa/state_tracker/st_cb_drawpixels.c | 8 ++-- src/mesa/state_tracker/st_cb_rasterpos.c | 3 +- src/mesa/state_tracker/st_context.h | 4 +- src/mesa/state_tracker/st_draw.c | 6 +-- src/mesa/state_tracker/st_program.h | 10 +++-- 18 files changed, 181 insertions(+), 120 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h index cd4b64eec4..57d162b2ac 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -58,6 +58,16 @@ struct cso_rasterizer { void *data; }; +struct cso_fragment_shader { + struct pipe_shader_state state; + void *data; +}; + +struct cso_vertex_shader { + struct pipe_shader_state state; + void *data; +}; + enum cso_cache_type { CSO_BLEND, CSO_SAMPLER, diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index a649899010..a81bfe82dd 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -74,8 +74,8 @@ struct failover_context { const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct fo_state *rasterizer; - const struct pipe_shader_state *fragment_shader; - const struct pipe_shader_state *vertex_shader; + const struct fo_state *fragment_shader; + const struct fo_state *vertex_shader; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 25725625e0..db3aea7756 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -150,26 +150,81 @@ failover_set_framebuffer_state(struct pipe_context *pipe, failover->hw->set_framebuffer_state( failover->hw, framebuffer ); } + +static void * +failover_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_fs_state(pipe, templ); + state->hw_state = failover->hw->create_fs_state(pipe, templ); + + return state; +} + static void failover_bind_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *fs) + void *fs) { struct failover_context *failover = failover_context(pipe); - failover->fragment_shader = fs; + failover->fragment_shader = (struct fo_state *)fs; failover->dirty |= FO_NEW_FRAGMENT_SHADER; - failover->hw->bind_fs_state( failover->hw, fs ); + failover->hw->bind_fs_state(failover->hw, (struct pipe_shader_state *)fs); +} + +static void +failover_delete_fs_state(struct pipe_context *pipe, + void *fs) +{ + struct fo_state *state = (struct fo_state*)fs; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_fs_state(pipe, state->sw_state); + failover->hw->delete_fs_state(pipe, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); +} + +static void * +failover_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_vs_state(pipe, templ); + state->hw_state = failover->hw->create_vs_state(pipe, templ); + + return state; } static void failover_bind_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *vs) + void *vs) { struct failover_context *failover = failover_context(pipe); - failover->vertex_shader = vs; + failover->vertex_shader = (struct fo_state*)vs; failover->dirty |= FO_NEW_VERTEX_SHADER; - failover->hw->bind_vs_state( failover->hw, vs ); + failover->hw->bind_vs_state(failover->hw, vs); +} + +static void +failover_delete_vs_state(struct pipe_context *pipe, + void *vs) +{ + struct fo_state *state = (struct fo_state*)vs; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_vs_state(pipe, state->sw_state); + failover->hw->delete_vs_state(pipe, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); } static void @@ -312,8 +367,12 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.create_rasterizer_state = failover_create_rasterizer_state; failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state; failover->pipe.delete_rasterizer_state = failover_delete_rasterizer_state; - failover->pipe.bind_fs_state = failover_bind_fs_state; - failover->pipe.bind_vs_state = failover_bind_vs_state; + failover->pipe.create_fs_state = failover_create_fs_state; + failover->pipe.bind_fs_state = failover_bind_fs_state; + failover->pipe.delete_fs_state = failover_delete_fs_state; + failover->pipe.create_vs_state = failover_create_vs_state; + failover->pipe.bind_vs_state = failover_bind_vs_state; + failover->pipe.delete_vs_state = failover_delete_vs_state; failover->pipe.set_alpha_test_state = failover_set_alpha_test_state; failover->pipe.set_blend_color = failover_set_blend_color; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index f2b0b1edc0..ec896fd020 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -78,10 +78,12 @@ failover_state_emit( struct failover_context *failover ) failover->sw->set_framebuffer_state( failover->sw, &failover->framebuffer ); if (failover->dirty & FO_NEW_FRAGMENT_SHADER) - failover->sw->bind_fs_state( failover->sw, failover->fragment_shader ); + failover->sw->bind_fs_state( failover->sw, + failover->fragment_shader->sw_state ); if (failover->dirty & FO_NEW_VERTEX_SHADER) - failover->sw->bind_vs_state( failover->sw, failover->vertex_shader ); + failover->sw->bind_vs_state( failover->sw, + failover->vertex_shader->sw_state ); if (failover->dirty & FO_NEW_STIPPLE) failover->sw->set_polygon_stipple( failover->sw, &failover->poly_stipple ); diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 66aa9a0274..1104c9519d 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -228,41 +228,37 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe, } -static const struct pipe_shader_state * -i915_create_shader_state( struct pipe_context *pipe, - const struct pipe_shader_state *templ ) +static void * +i915_create_shader_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { - - struct pipe_shader_state *shader = malloc(sizeof(struct pipe_shader_state)); - memcpy(shader, templ, sizeof(struct pipe_shader_state)); - - return shader; + return 0; } static void i915_bind_fs_state( struct pipe_context *pipe, - const struct pipe_shader_state *fs ) + void *fs ) { struct i915_context *i915 = i915_context(pipe); - i915->fs = fs; + i915->fs = (struct pipe_shader_state *)fs; i915->dirty |= I915_NEW_FS; } -static void i915_bind_vs_state( struct pipe_context *pipe, - const struct pipe_shader_state *vs ) +static void i915_bind_vs_state(struct pipe_context *pipe, + void *vs) { struct i915_context *i915 = i915_context(pipe); /* just pass-through to draw module */ - draw_set_vertex_shader(i915->draw, vs); + draw_set_vertex_shader(i915->draw, (const struct pipe_shader_state *)vs); } -static void i915_delete_shader_state( struct pipe_context *pipe, - const struct pipe_shader_state *shader ) +static void i915_delete_shader_state(struct pipe_context *pipe, + void *shader) { - free((struct pipe_shader_state*)shader); + /*do nothing*/ } static void i915_set_constant_buffer(struct pipe_context *pipe, diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 65001dfdf9..e17faad2c7 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -113,20 +113,15 @@ struct pipe_context { void (*delete_depth_stencil_state)(struct pipe_context *, const struct pipe_depth_stencil_state *); - const struct pipe_shader_state * (*create_fs_state)( - struct pipe_context *, - const struct pipe_shader_state *); - void (*bind_fs_state)(struct pipe_context *, - const struct pipe_shader_state *); - void (*delete_fs_state)(struct pipe_context *, - const struct pipe_shader_state *); - const struct pipe_shader_state * (*create_vs_state)( - struct pipe_context *, - const struct pipe_shader_state *); - void (*bind_vs_state)(struct pipe_context *, - const struct pipe_shader_state *); - void (*delete_vs_state)(struct pipe_context *, - const struct pipe_shader_state *); + void * (*create_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*bind_fs_state)(struct pipe_context *, void *); + void (*delete_fs_state)(struct pipe_context *, void *); + + void * (*create_vs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*bind_vs_state)(struct pipe_context *, void *); + void (*delete_vs_state)(struct pipe_context *, void *); void (*set_alpha_test_state)( struct pipe_context *, const struct pipe_alpha_test_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index a20ae1d4a2..5ed963c21d 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -88,15 +88,12 @@ void softpipe_set_constant_buffer(struct pipe_context *, void softpipe_set_feedback_state( struct pipe_context *, const struct pipe_feedback_state * ); -const struct pipe_shader_state * +void * softpipe_create_shader_state( struct pipe_context *, const struct pipe_shader_state * ); -void softpipe_bind_fs_state( struct pipe_context *, - const struct pipe_shader_state * ); -void softpipe_bind_vs_state( struct pipe_context *, - const struct pipe_shader_state * ); -void softpipe_delete_shader_state( struct pipe_context *, - const struct pipe_shader_state * ); +void softpipe_bind_fs_state( struct pipe_context *, void * ); +void softpipe_bind_vs_state( struct pipe_context *, void * ); +void softpipe_delete_shader_state( struct pipe_context *, void * ); void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index fbbde2f520..8306a95f44 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -33,44 +33,39 @@ #include "pipe/draw/draw_context.h" -const struct pipe_shader_state * -softpipe_create_shader_state( struct pipe_context *pipe, - const struct pipe_shader_state *templ ) +void * softpipe_create_shader_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { - struct pipe_shader_state *shader = malloc(sizeof(struct pipe_shader_state)); - memcpy(shader, templ, sizeof(struct pipe_shader_state)); - - return shader; + /* we just want the pipe_shader_state template in the bind calls */ + return 0; } -void softpipe_bind_fs_state( struct pipe_context *pipe, - const struct pipe_shader_state *fs ) +void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->fs = fs; + softpipe->fs = (struct pipe_shader_state *)fs; softpipe->dirty |= SP_NEW_FS; } -void softpipe_bind_vs_state( struct pipe_context *pipe, - const struct pipe_shader_state *vs ) +void softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->vs = vs; + softpipe->vs = (struct pipe_shader_state *)vs; softpipe->dirty |= SP_NEW_VS; - draw_set_vertex_shader(softpipe->draw, vs); + draw_set_vertex_shader(softpipe->draw, (struct pipe_shader_state *)vs); } void softpipe_delete_shader_state( struct pipe_context *pipe, - const struct pipe_shader_state *shader ) + void *shader ) { - free((struct pipe_shader_state*)shader); + /* do nothing */ } void softpipe_set_constant_buffer(struct pipe_context *pipe, diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 94b69c8df7..91e58f5831 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -51,14 +51,14 @@ * Translate a Mesa fragment shader into a TGSI shader. * \return pointer to cached pipe_shader object. */ -struct pipe_shader_state * +const struct cso_fragment_shader * st_translate_fragment_shader(struct st_context *st, struct st_fragment_program *stfp) { GLuint outputMapping[FRAG_RESULT_MAX]; GLuint inputMapping[PIPE_MAX_SHADER_INPUTS]; struct pipe_shader_state fs; - struct pipe_shader_state *cached; + const struct cso_fragment_shader *cso; GLuint interpMode[16]; /* XXX size? */ GLuint i; GLbitfield inputsRead = stfp->Base.Base.InputsRead; @@ -142,15 +142,15 @@ st_translate_fragment_shader(struct st_context *st, fs.tokens = &stfp->tokens[0]; - cached = st_cached_fs_state(st, &fs); - stfp->fs = cached; + cso = st_cached_fs_state(st, &fs); + stfp->fs = cso; if (TGSI_DEBUG) tgsi_dump( stfp->tokens, 0/*TGSI_DUMP_VERBOSE*/ ); stfp->dirty = 0; - return cached; + return cso; } @@ -183,7 +183,7 @@ static void update_fs( struct st_context *st ) if (stfp->dirty) st->state.fs = st_translate_fragment_shader( st, st->fp ); - st->pipe->bind_fs_state(st->pipe, st->state.fs); + st->pipe->bind_fs_state(st->pipe, st->state.fs->data); } } diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index cf9dd810e9..078c052ae2 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -53,13 +53,13 @@ * Translate a Mesa vertex shader into a TGSI shader. * \return pointer to cached pipe_shader object. */ -struct pipe_shader_state * +const struct cso_vertex_shader * st_translate_vertex_shader(struct st_context *st, struct st_vertex_program *stvp) { GLuint outputMapping[PIPE_MAX_SHADER_INPUTS]; struct pipe_shader_state vs; - struct pipe_shader_state *cached; + const struct cso_vertex_shader *cso; GLuint i; memset(&vs, 0, sizeof(vs)); @@ -147,8 +147,8 @@ st_translate_vertex_shader(struct st_context *st, vs.tokens = &stvp->tokens[0]; - cached = st_cached_vs_state(st, &vs); - stvp->vs = cached; + cso = st_cached_vs_state(st, &vs); + stvp->vs = cso; if (TGSI_DEBUG) tgsi_dump( stvp->tokens, 0 ); @@ -157,13 +157,13 @@ st_translate_vertex_shader(struct st_context *st, if (stvp->sse2_program.csr == stvp->sse2_program.store) tgsi_emit_sse2( stvp->tokens, &stvp->sse2_program ); - if (!cached->executable) - cached->executable = (void *) x86_get_func( &stvp->sse2_program ); + if (!cso->state.executable) + cso->state.executable = (void *) x86_get_func( &stvp->sse2_program ); #endif stvp->dirty = 0; - return cached; + return cso; } @@ -191,10 +191,10 @@ static void update_vs( struct st_context *st ) /* Bind the vertex program */ st->vp = stvp; - if (stvp->dirty) + if (stvp->dirty) st->state.vs = st_translate_vertex_shader( st, st->vp ); - st->pipe->bind_vs_state(st->pipe, st->state.vs); + st->pipe->bind_vs_state(st->pipe, st->state.vs->data); } } diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 0f233cea58..e5ba0592cf 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -115,9 +115,9 @@ const struct cso_rasterizer* st_cached_rasterizer_state( return (struct cso_rasterizer*)(cso_hash_iter_data(iter)); } -struct pipe_shader_state * st_cached_fs_state( - struct st_context *st, - const struct pipe_shader_state *templ) +const struct cso_fragment_shader * +st_cached_fs_state(struct st_context *st, + const struct pipe_shader_state *templ) { unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_shader_state)); @@ -125,17 +125,19 @@ struct pipe_shader_state * st_cached_fs_state( hash_key, CSO_FRAGMENT_SHADER, (void*)templ); if (cso_hash_iter_is_null(iter)) { - const struct pipe_shader_state *created_state = - st->pipe->create_fs_state(st->pipe, templ); - iter = cso_insert_state(st->cache, hash_key, CSO_FRAGMENT_SHADER, - (void*)created_state); + struct cso_fragment_shader *cso = malloc(sizeof(struct cso_fragment_shader)); + memcpy(&cso->state, templ, sizeof(struct pipe_shader_state)); + cso->data = st->pipe->create_fs_state(st->pipe, templ); + if (!cso->data) + cso->data = &cso->state; + iter = cso_insert_state(st->cache, hash_key, CSO_FRAGMENT_SHADER, cso); } - return (struct pipe_shader_state*)(cso_hash_iter_data(iter)); + return (struct cso_fragment_shader*)(cso_hash_iter_data(iter)); } -struct pipe_shader_state * st_cached_vs_state( - struct st_context *st, - const struct pipe_shader_state *templ) +const struct cso_vertex_shader * +st_cached_vs_state(struct st_context *st, + const struct pipe_shader_state *templ) { unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_shader_state)); @@ -143,10 +145,12 @@ struct pipe_shader_state * st_cached_vs_state( hash_key, CSO_VERTEX_SHADER, (void*)templ); if (cso_hash_iter_is_null(iter)) { - const struct pipe_shader_state *created_state = - st->pipe->create_vs_state(st->pipe, templ); - iter = cso_insert_state(st->cache, hash_key, CSO_VERTEX_SHADER, - (void*)created_state); + struct cso_vertex_shader *cso = malloc(sizeof(struct cso_vertex_shader)); + memcpy(&cso->state, templ, sizeof(struct pipe_shader_state)); + cso->data = st->pipe->create_vs_state(st->pipe, templ); + if (!cso->data) + cso->data = &cso->state; + iter = cso_insert_state(st->cache, hash_key, CSO_VERTEX_SHADER, cso); } - return (struct pipe_shader_state*)(cso_hash_iter_data(iter)); + return (struct cso_vertex_shader*)(cso_hash_iter_data(iter)); } diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index 5b8c6168a8..356dd98183 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -55,13 +55,13 @@ const struct cso_rasterizer * st_cached_rasterizer_state(struct st_context *st, const struct pipe_rasterizer_state *raster); -struct pipe_shader_state *st_cached_fs_state( - struct st_context *st, - const struct pipe_shader_state *templ); +const struct cso_fragment_shader * +st_cached_fs_state(struct st_context *st, + const struct pipe_shader_state *templ); -struct pipe_shader_state *st_cached_vs_state( - struct st_context *st, - const struct pipe_shader_state *templ); +const struct cso_vertex_shader * +st_cached_vs_state(struct st_context *st, + const struct pipe_shader_state *templ); #endif diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index ee70ce3320..03a81652cb 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -348,7 +348,7 @@ clear_with_quad(GLcontext *ctx, if (!stfp) { stfp = make_frag_shader(st); } - pipe->bind_fs_state(pipe, stfp->fs); + pipe->bind_fs_state(pipe, stfp->fs->data); } /* vertex shader state: color/position pass-through */ @@ -357,7 +357,7 @@ clear_with_quad(GLcontext *ctx, if (!stvp) { stvp = make_vertex_shader(st); } - pipe->bind_vs_state(pipe, stvp->vs); + pipe->bind_vs_state(pipe, stvp->vs->data); } /* viewport state: viewport matching window dims */ @@ -383,8 +383,8 @@ clear_with_quad(GLcontext *ctx, pipe->set_alpha_test_state(pipe, &st->state.alpha_test); pipe->bind_blend_state(pipe, st->state.blend->data); pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil); - pipe->bind_fs_state(pipe, st->state.fs); - pipe->bind_vs_state(pipe, st->state.vs); + pipe->bind_fs_state(pipe, st->state.fs->data); + pipe->bind_vs_state(pipe, st->state.vs->data); pipe->bind_rasterizer_state(pipe, st->state.rasterizer->data); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); /* OR: diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index d4f260ee54..4e3c24353e 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -330,7 +330,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, if (!stfp) { stfp = make_fragment_shader(ctx->st); } - pipe->bind_fs_state(pipe, stfp->fs); + pipe->bind_fs_state(pipe, stfp->fs->data); } /* vertex shader state: position + texcoord pass-through */ @@ -339,7 +339,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, if (!stvp) { stvp = make_vertex_shader(ctx->st); } - pipe->bind_vs_state(pipe, stvp->vs); + pipe->bind_vs_state(pipe, stvp->vs->data); } /* texture sampling state: */ @@ -393,8 +393,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* restore GL state */ pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); - pipe->bind_fs_state(pipe, ctx->st->state.fs); - pipe->bind_vs_state(pipe, ctx->st->state.vs); + pipe->bind_fs_state(pipe, ctx->st->state.fs->data); + pipe->bind_vs_state(pipe, ctx->st->state.vs->data); pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 5245535a65..04b2016ffc 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -35,6 +35,7 @@ #include "st_context.h" #include "st_atom.h" +#include "st_cache.h" #include "st_draw.h" #include "st_program.h" #include "st_cb_rasterpos.h" @@ -88,7 +89,7 @@ static void setup_feedback(GLcontext *ctx) { struct pipe_context *pipe = ctx->st->pipe; - const struct pipe_shader_state *vs = ctx->st->state.vs; + const struct pipe_shader_state *vs = &ctx->st->state.vs->state; struct pipe_feedback_state feedback; uint i; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 93b6425480..df976260f8 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -79,8 +79,8 @@ struct st_context const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct cso_rasterizer *rasterizer; - const struct pipe_shader_state *fs; - const struct pipe_shader_state *vs; + const struct cso_fragment_shader *fs; + const struct cso_vertex_shader *vs; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 238ade00ac..633e4d9470 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -198,7 +198,7 @@ st_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = ctx->st->state.vs; + vs = &ctx->st->state.vs->state; /* loop over TGSI shader inputs */ for (attr = 0; attr < vs->num_inputs; attr++) { @@ -405,8 +405,8 @@ st_feedback_draw_vbo(GLcontext *ctx, assert(draw); draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); - draw_set_rasterizer_state(draw, st->state.rasterizer->data); - draw_set_vertex_shader(draw, st->state.vs); + draw_set_rasterizer_state(draw, &st->state.rasterizer->state); + draw_set_vertex_shader(draw, &st->state.vs->state); /* XXX need to set vertex info too */ diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 4945141d15..4f9ace3e6a 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -40,6 +40,8 @@ #define ST_FP_MAX_TOKENS 1024 +struct cso_fragment_shader; +struct cso_vertex_shader; struct st_fragment_program { @@ -52,7 +54,7 @@ struct st_fragment_program GLboolean dirty; /** Pointer to the corresponding cached shader */ - const struct pipe_shader_state *fs; + const struct cso_fragment_shader *fs; GLuint param_state; }; @@ -83,7 +85,7 @@ struct st_vertex_program #endif /** Pointer to the corresponding cached shader */ - const struct pipe_shader_state *vs; + const struct cso_vertex_shader *vs; GLuint param_state; }; @@ -105,12 +107,12 @@ st_vertex_program( struct gl_vertex_program *vp ) } -extern struct pipe_shader_state * +extern const struct cso_fragment_shader * st_translate_fragment_shader(struct st_context *st, struct st_fragment_program *fp); -extern struct pipe_shader_state * +extern const struct cso_vertex_shader * st_translate_vertex_shader(struct st_context *st, struct st_vertex_program *vp); -- cgit v1.2.3 From a6c0c5532f7bfa50ae54c36cf4d74ad4b9f926f8 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 20 Sep 2007 08:35:10 -0400 Subject: Convert depth_stencil state to the new semantics. --- src/mesa/pipe/cso_cache/cso_cache.h | 5 +++++ src/mesa/pipe/failover/fo_context.h | 2 +- src/mesa/pipe/failover/fo_state.c | 35 ++++++++++++++++++++++++++++--- src/mesa/pipe/failover/fo_state_emit.c | 3 ++- src/mesa/pipe/i915simple/i915_state.c | 16 ++++++-------- src/mesa/pipe/p_context.h | 11 ++++------ src/mesa/pipe/softpipe/sp_state.h | 14 +++++-------- src/mesa/pipe/softpipe/sp_state_blend.c | 16 ++++++-------- src/mesa/state_tracker/st_atom_depth.c | 10 ++++----- src/mesa/state_tracker/st_cache.c | 23 +++++++++++--------- src/mesa/state_tracker/st_cache.h | 6 +++--- src/mesa/state_tracker/st_cb_clear.c | 9 ++++---- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_context.h | 2 +- 14 files changed, 88 insertions(+), 66 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h index 57d162b2ac..291759d5d1 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -53,6 +53,11 @@ struct cso_blend { void *data; }; +struct cso_depth_stencil { + struct pipe_depth_stencil_state state; + void *data; +}; + struct cso_rasterizer { struct pipe_rasterizer_state state; void *data; diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index a81bfe82dd..7371ad4392 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -72,7 +72,7 @@ struct failover_context { */ const struct fo_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct pipe_depth_stencil_state *depth_stencil; + const struct fo_state *depth_stencil; const struct fo_state *rasterizer; const struct fo_state *fragment_shader; const struct fo_state *vertex_shader; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index db3aea7756..3379f45355 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -128,17 +128,44 @@ failover_set_clear_color_state( struct pipe_context *pipe, failover->hw->set_clear_color_state( failover->hw, clear_color ); } +static void * +failover_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_depth_stencil_state(pipe, templ); + state->hw_state = failover->hw->create_depth_stencil_state(pipe, templ); + + return state; +} + static void failover_bind_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *depth_stencil) + void *depth_stencil) { struct failover_context *failover = failover_context(pipe); - failover->depth_stencil = depth_stencil; + failover->depth_stencil = (struct fo_state *)depth_stencil; failover->dirty |= FO_NEW_DEPTH_STENCIL; failover->hw->bind_depth_stencil_state( failover->hw, depth_stencil ); } +static void +failover_delete_depth_stencil_state(struct pipe_context *pipe, + void *ds) +{ + struct fo_state *state = (struct fo_state*)ds; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_depth_stencil_state(pipe, state->sw_state); + failover->hw->delete_depth_stencil_state(pipe, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); +} + static void failover_set_framebuffer_state(struct pipe_context *pipe, const struct pipe_framebuffer_state *framebuffer) @@ -363,7 +390,9 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.bind_blend_state = failover_bind_blend_state; failover->pipe.delete_blend_state = failover_delete_blend_state; failover->pipe.bind_sampler_state = failover_bind_sampler_state; - failover->pipe.bind_depth_stencil_state = failover_bind_depth_stencil_state; + failover->pipe.create_depth_stencil_state = failover_create_depth_stencil_state; + failover->pipe.bind_depth_stencil_state = failover_bind_depth_stencil_state; + failover->pipe.delete_depth_stencil_state = failover_delete_depth_stencil_state; failover->pipe.create_rasterizer_state = failover_create_rasterizer_state; failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state; failover->pipe.delete_rasterizer_state = failover_delete_rasterizer_state; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index ec896fd020..da12b4e25c 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -72,7 +72,8 @@ failover_state_emit( struct failover_context *failover ) failover->sw->set_clear_color_state( failover->sw, &failover->clear_color ); if (failover->dirty & FO_NEW_DEPTH_STENCIL) - failover->sw->bind_depth_stencil_state( failover->sw, failover->depth_stencil ); + failover->sw->bind_depth_stencil_state( failover->sw, + failover->depth_stencil->sw_state ); if (failover->dirty & FO_NEW_FRAMEBUFFER) failover->sw->set_framebuffer_state( failover->sw, &failover->framebuffer ); diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 1104c9519d..be549ed6fd 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -175,31 +175,27 @@ static void i915_delete_sampler_state(struct pipe_context *pipe, * into one file. */ -static const struct pipe_depth_stencil_state * +static void * i915_create_depth_stencil_state(struct pipe_context *pipe, const struct pipe_depth_stencil_state *depth_stencil) { - struct pipe_depth_stencil_state *new_ds = - malloc(sizeof(struct pipe_depth_stencil_state)); - memcpy(new_ds, depth_stencil, sizeof(struct pipe_depth_stencil_state)); - - return new_ds; + return 0; } static void i915_bind_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *depth_stencil) + void *depth_stencil) { struct i915_context *i915 = i915_context(pipe); - i915->depth_stencil = depth_stencil; + i915->depth_stencil = (const struct pipe_depth_stencil_state *)depth_stencil; i915->dirty |= I915_NEW_DEPTH_STENCIL; } static void i915_delete_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *depth_stencil) + void *depth_stencil) { - free((struct pipe_depth_stencil_state *)depth_stencil); + /* do nothing */ } static void i915_set_alpha_test_state(struct pipe_context *pipe, diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index e17faad2c7..84aca20c58 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -105,13 +105,10 @@ struct pipe_context { void (*bind_rasterizer_state)(struct pipe_context *, void *); void (*delete_rasterizer_state)(struct pipe_context *, void *); - const struct pipe_depth_stencil_state * (*create_depth_stencil_state)( - struct pipe_context *, - const struct pipe_depth_stencil_state *); - void (*bind_depth_stencil_state)(struct pipe_context *, - const struct pipe_depth_stencil_state *); - void (*delete_depth_stencil_state)(struct pipe_context *, - const struct pipe_depth_stencil_state *); + void * (*create_depth_stencil_state)(struct pipe_context *, + const struct pipe_depth_stencil_state *); + void (*bind_depth_stencil_state)(struct pipe_context *, void *); + void (*delete_depth_stencil_state)(struct pipe_context *, void *); void * (*create_fs_state)(struct pipe_context *, const struct pipe_shader_state *); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 5ed963c21d..08dfe208fb 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -50,21 +50,17 @@ void softpipe_bind_sampler_state(struct pipe_context *, void softpipe_delete_sampler_state(struct pipe_context *, const struct pipe_sampler_state *); -const struct pipe_depth_stencil_state * +void * softpipe_create_depth_stencil_state(struct pipe_context *, const struct pipe_depth_stencil_state *); -void softpipe_bind_depth_stencil_state(struct pipe_context *, - const struct pipe_depth_stencil_state *); -void softpipe_delete_depth_stencil_state(struct pipe_context *, - const struct pipe_depth_stencil_state *); +void softpipe_bind_depth_stencil_state(struct pipe_context *, void *); +void softpipe_delete_depth_stencil_state(struct pipe_context *, void *); void * softpipe_create_rasterizer_state(struct pipe_context *, const struct pipe_rasterizer_state *); -void softpipe_bind_rasterizer_state(struct pipe_context *, - void *); -void softpipe_delete_rasterizer_state(struct pipe_context *, - void *); +void softpipe_bind_rasterizer_state(struct pipe_context *, void *); +void softpipe_delete_rasterizer_state(struct pipe_context *, void *); void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 7fb47e7aab..cf47607955 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -82,30 +82,26 @@ softpipe_set_alpha_test_state(struct pipe_context *pipe, softpipe->dirty |= SP_NEW_ALPHA_TEST; } -const struct pipe_depth_stencil_state * +void * softpipe_create_depth_stencil_state(struct pipe_context *pipe, const struct pipe_depth_stencil_state *depth_stencil) { - struct pipe_depth_stencil_state *new_ds = malloc(sizeof(struct pipe_depth_stencil_state)); - memcpy(new_ds, depth_stencil, sizeof(struct pipe_depth_stencil_state)); - - return new_ds; + return 0; } void softpipe_bind_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *depth_stencil) + void *depth_stencil) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->depth_stencil = depth_stencil; + softpipe->depth_stencil = (const struct pipe_depth_stencil_state *)depth_stencil; softpipe->dirty |= SP_NEW_DEPTH_STENCIL; } void -softpipe_delete_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *depth) +softpipe_delete_depth_stencil_state(struct pipe_context *pipe, void *depth) { - free((struct pipe_depth_stencil_state*)depth); + /* do nothing */ } diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index 406773213d..caf51f17ac 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -115,6 +115,7 @@ static void update_depth_stencil(struct st_context *st) { struct pipe_depth_stencil_state depth_stencil; + const struct cso_depth_stencil *cso; memset(&depth_stencil, 0, sizeof(depth_stencil)); @@ -149,12 +150,11 @@ update_depth_stencil(struct st_context *st) depth_stencil.stencil.clear_value = st->ctx->Stencil.Clear & 0xff; } - struct pipe_depth_stencil_state *cached_state = - st_cached_depth_stencil_state(st, &depth_stencil); - if (st->state.depth_stencil != cached_state) { + cso = st_cached_depth_stencil_state(st, &depth_stencil); + if (st->state.depth_stencil != cso) { /* state has changed */ - st->state.depth_stencil = cached_state; - st->pipe->bind_depth_stencil_state(st->pipe, cached_state); /* set new state */ + st->state.depth_stencil = cso; + st->pipe->bind_depth_stencil_state(st->pipe, cso->data); /* bind new state */ } } diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index e5ba0592cf..c1ec130b32 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -78,21 +78,24 @@ struct pipe_sampler_state * st_cached_sampler_state( return (struct pipe_sampler_state*)(cso_hash_iter_data(iter)); } -struct pipe_depth_stencil_state * st_cached_depth_stencil_state( - struct st_context *st, - const struct pipe_depth_stencil_state *depth_stencil) +const struct cso_depth_stencil * +st_cached_depth_stencil_state(struct st_context *st, + const struct pipe_depth_stencil_state *templ) { - unsigned hash_key = cso_construct_key((void*)depth_stencil, sizeof(struct pipe_depth_stencil_state)); + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_depth_stencil_state)); struct cso_hash_iter iter = cso_find_state_template(st->cache, hash_key, CSO_DEPTH_STENCIL, - (void*)depth_stencil); + (void*)templ); if (cso_hash_iter_is_null(iter)) { - const struct pipe_depth_stencil_state *created_state = st->pipe->create_depth_stencil_state( - st->pipe, depth_stencil); - iter = cso_insert_state(st->cache, hash_key, CSO_DEPTH_STENCIL, - (void*)created_state); + struct cso_depth_stencil *cso = malloc(sizeof(struct cso_depth_stencil)); + memcpy(&cso->state, templ, sizeof(struct pipe_depth_stencil_state)); + cso->data = st->pipe->create_depth_stencil_state(st->pipe, templ); + if (!cso->data) + cso->data = &cso->state; + iter = cso_insert_state(st->cache, hash_key, CSO_DEPTH_STENCIL, cso); } - return (struct pipe_depth_stencil_state*)(cso_hash_iter_data(iter)); + return (struct cso_depth_stencil*)(cso_hash_iter_data(iter)); } const struct cso_rasterizer* st_cached_rasterizer_state( diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index 356dd98183..167d9ec11a 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -47,9 +47,9 @@ struct pipe_sampler_state * st_cached_sampler_state(struct st_context *st, const struct pipe_sampler_state *sampler); -struct pipe_depth_stencil_state *st_cached_depth_stencil_state( - struct st_context *st, - const struct pipe_depth_stencil_state *depth_stencil); +const struct cso_depth_stencil * +st_cached_depth_stencil_state(struct st_context *st, + const struct pipe_depth_stencil_state *depth_stencil); const struct cso_rasterizer * st_cached_rasterizer_state(struct st_context *st, diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 03a81652cb..bfc977daa4 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -303,7 +303,7 @@ clear_with_quad(GLcontext *ctx, /* depth_stencil state: always pass/set to ref value */ { struct pipe_depth_stencil_state depth_stencil; - struct pipe_depth_stencil_state *cached; + const struct cso_depth_stencil *cso; memset(&depth_stencil, 0, sizeof(depth_stencil)); if (depth) { depth_stencil.depth.enabled = 1; @@ -321,9 +321,8 @@ clear_with_quad(GLcontext *ctx, depth_stencil.stencil.value_mask[0] = 0xff; depth_stencil.stencil.write_mask[0] = ctx->Stencil.WriteMask[0] & 0xff; } - cached = - st_cached_depth_stencil_state(ctx->st, &depth_stencil); - pipe->bind_depth_stencil_state(pipe, cached); + cso = st_cached_depth_stencil_state(ctx->st, &depth_stencil); + pipe->bind_depth_stencil_state(pipe, cso->data); } /* setup state: nothing */ @@ -382,7 +381,7 @@ clear_with_quad(GLcontext *ctx, /* Restore pipe state */ pipe->set_alpha_test_state(pipe, &st->state.alpha_test); pipe->bind_blend_state(pipe, st->state.blend->data); - pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil); + pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil->data); pipe->bind_fs_state(pipe, st->state.fs->data); pipe->bind_vs_state(pipe, st->state.vs->data); pipe->bind_rasterizer_state(pipe, st->state.rasterizer->data); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 4e3c24353e..95810b7baf 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -470,7 +470,7 @@ any_fragment_ops(const struct st_context *st) if (st->state.alpha_test.enabled || st->state.blend->state.blend_enable || st->state.blend->state.logicop_enable || - st->state.depth_stencil->depth.enabled) + st->state.depth_stencil->state.depth.enabled) /* XXX more checks */ return GL_TRUE; else diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index df976260f8..b82cf24273 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -77,7 +77,7 @@ struct st_context struct { const struct cso_blend *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct pipe_depth_stencil_state *depth_stencil; + const struct cso_depth_stencil *depth_stencil; const struct cso_rasterizer *rasterizer; const struct cso_fragment_shader *fs; const struct cso_vertex_shader *vs; -- cgit v1.2.3 From 7a06c026ad24b74048f6d125383faf25deb1dfbb Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 20 Sep 2007 10:07:10 -0400 Subject: Fix failover state binding and convert the sampler to use the new state constant state object semantics. --- src/mesa/pipe/cso_cache/cso_cache.h | 5 +++ src/mesa/pipe/failover/fo_context.h | 12 +++--- src/mesa/pipe/failover/fo_state.c | 72 ++++++++++++++++++++++--------- src/mesa/pipe/failover/fo_state_emit.c | 2 +- src/mesa/pipe/i915simple/i915_state.c | 15 +++---- src/mesa/pipe/p_context.h | 13 +++--- src/mesa/pipe/softpipe/sp_state.h | 9 ++-- src/mesa/pipe/softpipe/sp_state_sampler.c | 18 +++----- src/mesa/state_tracker/st_atom_sampler.c | 10 ++--- src/mesa/state_tracker/st_cache.c | 22 +++++----- src/mesa/state_tracker/st_cache.h | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 7 +-- src/mesa/state_tracker/st_context.h | 2 +- 13 files changed, 105 insertions(+), 84 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h index 291759d5d1..2acb58c66b 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -73,6 +73,11 @@ struct cso_vertex_shader { void *data; }; +struct cso_sampler { + struct pipe_sampler_state state; + void *data; +}; + enum cso_cache_type { CSO_BLEND, CSO_SAMPLER, diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index 7371ad4392..8a2fbe2be9 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -70,12 +70,12 @@ struct failover_context { /* The most recent drawing state as set by the driver: */ - const struct fo_state *blend; - const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct fo_state *depth_stencil; - const struct fo_state *rasterizer; - const struct fo_state *fragment_shader; - const struct fo_state *vertex_shader; + const struct fo_state *blend; + const struct fo_state *sampler[PIPE_MAX_SAMPLERS]; + const struct fo_state *depth_stencil; + const struct fo_state *rasterizer; + const struct fo_state *fragment_shader; + const struct fo_state *vertex_shader; struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 3379f45355..ce3f0ca635 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -75,10 +75,10 @@ failover_bind_blend_state( struct pipe_context *pipe, void *blend ) { struct failover_context *failover = failover_context(pipe); - - failover->blend = (struct fo_state *)blend; + struct fo_state *state = (struct fo_state *)blend; + failover->blend = state; failover->dirty |= FO_NEW_BLEND; - failover->hw->bind_blend_state( failover->hw, blend ); + failover->hw->bind_blend_state( failover->hw, state->hw_state ); } static void @@ -146,10 +146,10 @@ failover_bind_depth_stencil_state(struct pipe_context *pipe, void *depth_stencil) { struct failover_context *failover = failover_context(pipe); - - failover->depth_stencil = (struct fo_state *)depth_stencil; + struct fo_state *state = (struct fo_state *)depth_stencil; + failover->depth_stencil = state; failover->dirty |= FO_NEW_DEPTH_STENCIL; - failover->hw->bind_depth_stencil_state( failover->hw, depth_stencil ); + failover->hw->bind_depth_stencil_state(failover->hw, state->hw_state); } static void @@ -192,14 +192,13 @@ failover_create_fs_state(struct pipe_context *pipe, } static void -failover_bind_fs_state(struct pipe_context *pipe, - void *fs) +failover_bind_fs_state(struct pipe_context *pipe, void *fs) { struct failover_context *failover = failover_context(pipe); - - failover->fragment_shader = (struct fo_state *)fs; + struct fo_state *state = (struct fo_state*)fs; + failover->fragment_shader = state; failover->dirty |= FO_NEW_FRAGMENT_SHADER; - failover->hw->bind_fs_state(failover->hw, (struct pipe_shader_state *)fs); + failover->hw->bind_fs_state(failover->hw, state->hw_state); } static void @@ -235,9 +234,10 @@ failover_bind_vs_state(struct pipe_context *pipe, { struct failover_context *failover = failover_context(pipe); - failover->vertex_shader = (struct fo_state*)vs; + struct fo_state *state = (struct fo_state*)vs; + failover->vertex_shader = state; failover->dirty |= FO_NEW_VERTEX_SHADER; - failover->hw->bind_vs_state(failover->hw, vs); + failover->hw->bind_vs_state(failover->hw, state->hw_state); } static void @@ -284,9 +284,10 @@ failover_bind_rasterizer_state(struct pipe_context *pipe, { struct failover_context *failover = failover_context(pipe); - failover->rasterizer = (struct fo_state *)raster; + struct fo_state *state = (struct fo_state*)raster; + failover->rasterizer = state; failover->dirty |= FO_NEW_RASTERIZER; - failover->hw->bind_rasterizer_state( failover->hw, raster ); + failover->hw->bind_rasterizer_state(failover->hw, state->hw_state); } static void @@ -315,17 +316,44 @@ failover_set_scissor_state( struct pipe_context *pipe, failover->hw->set_scissor_state( failover->hw, scissor ); } + +static void * +failover_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_sampler_state(pipe, templ); + state->hw_state = failover->hw->create_sampler_state(pipe, templ); + + return state; +} + static void failover_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, - const struct pipe_sampler_state *sampler) + unsigned unit, void *sampler) { struct failover_context *failover = failover_context(pipe); - - failover->sampler[unit] = sampler; + struct fo_state *state = (struct fo_state*)sampler; + failover->sampler[unit] = state; failover->dirty |= FO_NEW_SAMPLER; failover->dirty_sampler |= (1<hw->bind_sampler_state( failover->hw, unit, sampler ); + failover->hw->bind_sampler_state(failover->hw, unit, + state->hw_state); +} + +static void +failover_delete_sampler_state(struct pipe_context *pipe, void *sampler) +{ + struct fo_state *state = (struct fo_state*)sampler; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_sampler_state(pipe, state->sw_state); + failover->hw->delete_sampler_state(pipe, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); } @@ -389,7 +417,9 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.create_blend_state = failover_create_blend_state; failover->pipe.bind_blend_state = failover_bind_blend_state; failover->pipe.delete_blend_state = failover_delete_blend_state; - failover->pipe.bind_sampler_state = failover_bind_sampler_state; + failover->pipe.create_sampler_state = failover_create_sampler_state; + failover->pipe.bind_sampler_state = failover_bind_sampler_state; + failover->pipe.delete_sampler_state = failover_delete_sampler_state; failover->pipe.create_depth_stencil_state = failover_create_depth_stencil_state; failover->pipe.bind_depth_stencil_state = failover_bind_depth_stencil_state; failover->pipe.delete_depth_stencil_state = failover_delete_depth_stencil_state; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index da12b4e25c..c0ea681024 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -103,7 +103,7 @@ failover_state_emit( struct failover_context *failover ) for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { if (failover->dirty_sampler & (1<sw->bind_sampler_state( failover->sw, i, - failover->sampler[i] ); + failover->sampler[i]->sw_state ); } } } diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index be549ed6fd..0fb41e17ab 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -142,32 +142,27 @@ static void i915_set_blend_color( struct pipe_context *pipe, i915->dirty |= I915_NEW_BLEND; } -static const struct pipe_sampler_state * +static void * i915_create_sampler_state(struct pipe_context *pipe, const struct pipe_sampler_state *sampler) { - struct pipe_sampler_state *new_sampler = malloc(sizeof(struct pipe_sampler_state)); - memcpy(new_sampler, sampler, sizeof(struct pipe_sampler_state)); - - return new_sampler; + return 0; } static void i915_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, - const struct pipe_sampler_state *sampler) + unsigned unit, void *sampler) { struct i915_context *i915 = i915_context(pipe); assert(unit < PIPE_MAX_SAMPLERS); - i915->sampler[unit] = sampler; + i915->sampler[unit] = (const struct pipe_sampler_state *)sampler; i915->dirty |= I915_NEW_SAMPLER; } static void i915_delete_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) + void *sampler) { - free((struct pipe_sampler_state*)sampler); } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 84aca20c58..07ee019880 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -91,14 +91,11 @@ struct pipe_context { void (*bind_blend_state)(struct pipe_context *, void *); void (*delete_blend_state)(struct pipe_context *, void *); - const struct pipe_sampler_state * (*create_sampler_state)( - struct pipe_context *, - const struct pipe_sampler_state *); - void (*bind_sampler_state)(struct pipe_context *, - unsigned unit, - const struct pipe_sampler_state *); - void (*delete_sampler_state)(struct pipe_context *, - const struct pipe_sampler_state *); + void * (*create_sampler_state)(struct pipe_context *, + const struct pipe_sampler_state *); + void (*bind_sampler_state)(struct pipe_context *, unsigned unit, + void *); + void (*delete_sampler_state)(struct pipe_context *, void *); void *(*create_rasterizer_state)(struct pipe_context *, const struct pipe_rasterizer_state *); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 08dfe208fb..62323c41c3 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -41,14 +41,11 @@ void softpipe_bind_blend_state(struct pipe_context *, void softpipe_delete_blend_state(struct pipe_context *, void *); -const struct pipe_sampler_state * +void * softpipe_create_sampler_state(struct pipe_context *, const struct pipe_sampler_state *); -void softpipe_bind_sampler_state(struct pipe_context *, - unsigned, - const struct pipe_sampler_state *); -void softpipe_delete_sampler_state(struct pipe_context *, - const struct pipe_sampler_state *); +void softpipe_bind_sampler_state(struct pipe_context *, unsigned, void *); +void softpipe_delete_sampler_state(struct pipe_context *, void *); void * softpipe_create_depth_stencil_state(struct pipe_context *, diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 09898eb579..ad98375735 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -32,27 +32,21 @@ #include "sp_context.h" #include "sp_state.h" - - -const struct pipe_sampler_state * +void * softpipe_create_sampler_state(struct pipe_context *pipe, const struct pipe_sampler_state *sampler) { - struct pipe_sampler_state *new_sampler = malloc(sizeof(struct pipe_sampler_state)); - memcpy(new_sampler, sampler, sizeof(struct pipe_sampler_state)); - - return new_sampler; + return 0; } void softpipe_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, - const struct pipe_sampler_state *sampler) + unsigned unit, void *sampler) { struct softpipe_context *softpipe = softpipe_context(pipe); assert(unit < PIPE_MAX_SAMPLERS); - softpipe->sampler[unit] = sampler; + softpipe->sampler[unit] = (struct pipe_sampler_state *)sampler; softpipe->dirty |= SP_NEW_SAMPLER; } @@ -60,9 +54,9 @@ softpipe_bind_sampler_state(struct pipe_context *pipe, void softpipe_delete_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) + void *sampler) { - free((struct pipe_sampler_state*)sampler); + /* do nothing */ } diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 994d3691d8..23ccd55a05 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -123,6 +123,7 @@ update_samplers(struct st_context *st) const struct gl_texture_object *texobj = st->ctx->Texture.Unit[u]._Current; struct pipe_sampler_state sampler; + const struct cso_sampler *cso; memset(&sampler, 0, sizeof(sampler)); @@ -143,13 +144,12 @@ update_samplers(struct st_context *st) /* XXX more sampler state here */ } - const struct pipe_sampler_state *cached_sampler = - st_cached_sampler_state(st, &sampler); + cso = st_cached_sampler_state(st, &sampler); - if (cached_sampler != st->state.sampler[u]) { + if (cso != st->state.sampler[u]) { /* state has changed */ - st->state.sampler[u] = cached_sampler; - st->pipe->bind_sampler_state(st->pipe, u, cached_sampler); + st->state.sampler[u] = cso; + st->pipe->bind_sampler_state(st->pipe, u, cso->data); } } } diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index c1ec130b32..007a2311e9 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -61,21 +61,23 @@ const struct cso_blend * st_cached_blend_state(struct st_context *st, return ((struct cso_blend *)cso_hash_iter_data(iter)); } -struct pipe_sampler_state * st_cached_sampler_state( - struct st_context *st, - const struct pipe_sampler_state *sampler) +const struct cso_sampler * +st_cached_sampler_state(struct st_context *st, + const struct pipe_sampler_state *templ) { - unsigned hash_key = cso_construct_key((void*)sampler, sizeof(struct pipe_sampler_state)); + unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_sampler_state)); struct cso_hash_iter iter = cso_find_state_template(st->cache, hash_key, CSO_SAMPLER, - (void*)sampler); + (void*)templ); if (cso_hash_iter_is_null(iter)) { - const struct pipe_sampler_state *created_state = st->pipe->create_sampler_state( - st->pipe, sampler); - iter = cso_insert_state(st->cache, hash_key, CSO_SAMPLER, - (void*)created_state); + struct cso_sampler *cso = malloc(sizeof(struct cso_sampler)); + memcpy(&cso->state, templ, sizeof(struct pipe_sampler_state)); + cso->data = st->pipe->create_sampler_state(st->pipe, templ); + if (!cso->data) + cso->data = &cso->state; + iter = cso_insert_state(st->cache, hash_key, CSO_SAMPLER, cso); } - return (struct pipe_sampler_state*)(cso_hash_iter_data(iter)); + return (struct cso_sampler*)(cso_hash_iter_data(iter)); } const struct cso_depth_stencil * diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index 167d9ec11a..483af6fdb4 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -43,7 +43,7 @@ const struct cso_blend * st_cached_blend_state(struct st_context *st, const struct pipe_blend_state *blend); -struct pipe_sampler_state * +const struct cso_sampler * st_cached_sampler_state(struct st_context *st, const struct pipe_sampler_state *sampler); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 95810b7baf..d814e34157 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -345,6 +345,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* texture sampling state: */ { struct pipe_sampler_state sampler; + const struct cso_sampler *cso; memset(&sampler, 0, sizeof(sampler)); sampler.wrap_s = PIPE_TEX_WRAP_REPEAT; sampler.wrap_t = PIPE_TEX_WRAP_REPEAT; @@ -352,8 +353,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; - const struct pipe_sampler_state *state = st_cached_sampler_state(ctx->st, &sampler); - pipe->bind_sampler_state(pipe, unit, state); + cso = st_cached_sampler_state(ctx->st, &sampler); + pipe->bind_sampler_state(pipe, unit, cso->data); } /* viewport state: viewport matching window dims */ @@ -396,7 +397,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->bind_fs_state(pipe, ctx->st->state.fs->data); pipe->bind_vs_state(pipe, ctx->st->state.vs->data); pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); - pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]); + pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]->data); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); free_mipmap_tree(pipe, mt); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index b82cf24273..8a57227c84 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -76,7 +76,7 @@ struct st_context */ struct { const struct cso_blend *blend; - const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct cso_sampler *sampler[PIPE_MAX_SAMPLERS]; const struct cso_depth_stencil *depth_stencil; const struct cso_rasterizer *rasterizer; const struct cso_fragment_shader *fs; -- cgit v1.2.3 From 1d8c31b47bd34a43e1f78a9f21a0c02c57c58479 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 20 Sep 2007 10:28:20 -0400 Subject: Rewrite the depth_stencil state handling in i915. Done to match the new cso semantics. translate in create, use in bind and later delete. --- src/mesa/pipe/i915simple/i915_context.h | 11 +++- src/mesa/pipe/i915simple/i915_state.c | 83 ++++++++++++++++++++++++- src/mesa/pipe/i915simple/i915_state_dynamic.c | 60 +----------------- src/mesa/pipe/i915simple/i915_state_immediate.c | 29 +-------- 4 files changed, 95 insertions(+), 88 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index ea6d1ce87b..1a7df6566b 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -120,6 +120,13 @@ struct i915_blend_state { unsigned LIS6; }; +struct i915_depth_stencil_state { + unsigned stencil_modes4; + unsigned bfo[2]; + unsigned stencil_LIS5; + unsigned depth_LIS6; +}; + struct i915_context { struct pipe_context pipe; @@ -128,9 +135,9 @@ struct i915_context /* The most recent drawing state as set by the driver: */ - const struct i915_blend_state *blend; + const struct i915_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct pipe_depth_stencil_state *depth_stencil; + const struct i915_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct pipe_shader_state *fs; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 0fb41e17ab..9611288961 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -174,7 +174,84 @@ static void * i915_create_depth_stencil_state(struct pipe_context *pipe, const struct pipe_depth_stencil_state *depth_stencil) { - return 0; + struct i915_depth_stencil_state *cso = calloc(1, sizeof(struct i915_depth_stencil_state)); + + { + int testmask = depth_stencil->stencil.value_mask[0] & 0xff; + int writemask = depth_stencil->stencil.write_mask[0] & 0xff; + + cso->stencil_modes4 |= (_3DSTATE_MODES_4_CMD | + ENABLE_STENCIL_TEST_MASK | + STENCIL_TEST_MASK(testmask) | + ENABLE_STENCIL_WRITE_MASK | + STENCIL_WRITE_MASK(writemask)); + } + + if (depth_stencil->stencil.front_enabled) { + int test = i915_translate_compare_func(depth_stencil->stencil.front_func); + int fop = i915_translate_stencil_op(depth_stencil->stencil.front_fail_op); + int dfop = i915_translate_stencil_op(depth_stencil->stencil.front_zfail_op); + int dpop = i915_translate_stencil_op(depth_stencil->stencil.front_zpass_op); + int ref = depth_stencil->stencil.ref_value[0] & 0xff; + + cso->stencil_LIS5 |= (S5_STENCIL_TEST_ENABLE | + S5_STENCIL_WRITE_ENABLE | + (ref << S5_STENCIL_REF_SHIFT) | + (test << S5_STENCIL_TEST_FUNC_SHIFT) | + (fop << S5_STENCIL_FAIL_SHIFT) | + (dfop << S5_STENCIL_PASS_Z_FAIL_SHIFT) | + (dpop << S5_STENCIL_PASS_Z_PASS_SHIFT)); + } + + if (depth_stencil->stencil.back_enabled) { + int test = i915_translate_compare_func(depth_stencil->stencil.back_func); + int fop = i915_translate_stencil_op(depth_stencil->stencil.back_fail_op); + int dfop = i915_translate_stencil_op(depth_stencil->stencil.back_zfail_op); + int dpop = i915_translate_stencil_op(depth_stencil->stencil.back_zpass_op); + int ref = depth_stencil->stencil.ref_value[1] & 0xff; + int tmask = depth_stencil->stencil.value_mask[1] & 0xff; + int wmask = depth_stencil->stencil.write_mask[1] & 0xff; + + cso->bfo[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | + BFO_ENABLE_STENCIL_FUNCS | + BFO_ENABLE_STENCIL_TWO_SIDE | + BFO_ENABLE_STENCIL_REF | + BFO_STENCIL_TWO_SIDE | + (ref << BFO_STENCIL_REF_SHIFT) | + (test << BFO_STENCIL_TEST_SHIFT) | + (fop << BFO_STENCIL_FAIL_SHIFT) | + (dfop << BFO_STENCIL_PASS_Z_FAIL_SHIFT) | + (dpop << BFO_STENCIL_PASS_Z_PASS_SHIFT)); + + cso->bfo[1] = (_3DSTATE_BACKFACE_STENCIL_MASKS | + BFM_ENABLE_STENCIL_TEST_MASK | + BFM_ENABLE_STENCIL_WRITE_MASK | + (tmask << BFM_STENCIL_TEST_MASK_SHIFT) | + (wmask << BFM_STENCIL_WRITE_MASK_SHIFT)); + } + else { + /* This actually disables two-side stencil: The bit set is a + * modify-enable bit to indicate we are changing the two-side + * setting. Then there is a symbolic zero to show that we are + * setting the flag to zero/off. + */ + cso->bfo[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | + BFO_ENABLE_STENCIL_TWO_SIDE | + 0); + cso->bfo[1] = 0; + } + + if (depth_stencil->depth.enabled) { + int func = i915_translate_compare_func(depth_stencil->depth.func); + + cso->depth_LIS6 |= (S6_DEPTH_TEST_ENABLE | + (func << S6_DEPTH_TEST_FUNC_SHIFT)); + + if (depth_stencil->depth.writemask) + cso->depth_LIS6 |= S6_DEPTH_WRITE_ENABLE; + } + + return cso; } static void i915_bind_depth_stencil_state(struct pipe_context *pipe, @@ -182,7 +259,7 @@ static void i915_bind_depth_stencil_state(struct pipe_context *pipe, { struct i915_context *i915 = i915_context(pipe); - i915->depth_stencil = (const struct pipe_depth_stencil_state *)depth_stencil; + i915->depth_stencil = (const struct i915_depth_stencil_state *)depth_stencil; i915->dirty |= I915_NEW_DEPTH_STENCIL; } @@ -190,7 +267,7 @@ static void i915_bind_depth_stencil_state(struct pipe_context *pipe, static void i915_delete_depth_stencil_state(struct pipe_context *pipe, void *depth_stencil) { - /* do nothing */ + free(depth_stencil); } static void i915_set_alpha_test_state(struct pipe_context *pipe, diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index 262ac947b8..d97aeb8d3b 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -67,17 +67,7 @@ static void upload_MODES4( struct i915_context *i915 ) unsigned modes4 = 0; /* I915_NEW_STENCIL */ - { - int testmask = i915->depth_stencil->stencil.value_mask[0] & 0xff; - int writemask = i915->depth_stencil->stencil.write_mask[0] & 0xff; - - modes4 |= (_3DSTATE_MODES_4_CMD | - ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(testmask) | - ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(writemask)); - } - + modes4 |= i915->depth_stencil->stencil_modes4; /* I915_NEW_BLEND */ modes4 |= i915->blend->modes4; @@ -102,53 +92,9 @@ const struct i915_tracked_state i915_upload_MODES4 = { static void upload_BFO( struct i915_context *i915 ) { - unsigned bf[2]; - - memset( bf, 0, sizeof(bf) ); - - /* _NEW_STENCIL - */ - if (i915->depth_stencil->stencil.back_enabled) { - int test = i915_translate_compare_func(i915->depth_stencil->stencil.back_func); - int fop = i915_translate_stencil_op(i915->depth_stencil->stencil.back_fail_op); - int dfop = i915_translate_stencil_op(i915->depth_stencil->stencil.back_zfail_op); - int dpop = i915_translate_stencil_op(i915->depth_stencil->stencil.back_zpass_op); - int ref = i915->depth_stencil->stencil.ref_value[1] & 0xff; - int tmask = i915->depth_stencil->stencil.value_mask[1] & 0xff; - int wmask = i915->depth_stencil->stencil.write_mask[1] & 0xff; - - bf[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | - BFO_ENABLE_STENCIL_FUNCS | - BFO_ENABLE_STENCIL_TWO_SIDE | - BFO_ENABLE_STENCIL_REF | - BFO_STENCIL_TWO_SIDE | - (ref << BFO_STENCIL_REF_SHIFT) | - (test << BFO_STENCIL_TEST_SHIFT) | - (fop << BFO_STENCIL_FAIL_SHIFT) | - (dfop << BFO_STENCIL_PASS_Z_FAIL_SHIFT) | - (dpop << BFO_STENCIL_PASS_Z_PASS_SHIFT)); - - bf[1] = (_3DSTATE_BACKFACE_STENCIL_MASKS | - BFM_ENABLE_STENCIL_TEST_MASK | - BFM_ENABLE_STENCIL_WRITE_MASK | - (tmask << BFM_STENCIL_TEST_MASK_SHIFT) | - (wmask << BFM_STENCIL_WRITE_MASK_SHIFT)); - } - else { - /* This actually disables two-side stencil: The bit set is a - * modify-enable bit to indicate we are changing the two-side - * setting. Then there is a symbolic zero to show that we are - * setting the flag to zero/off. - */ - bf[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | - BFO_ENABLE_STENCIL_TWO_SIDE | - 0); - bf[1] = 0; - } - - set_dynamic_indirect( i915, + set_dynamic_indirect( i915, I915_DYNAMIC_BFO_0, - &bf[0], + &(i915->depth_stencil->bfo[0]), 2 ); } diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 38cf2adc71..7de7f5377e 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -127,22 +127,7 @@ static void upload_S5( struct i915_context *i915 ) { unsigned LIS5 = 0; - /* I915_NEW_STENCIL */ - if (i915->depth_stencil->stencil.front_enabled) { - int test = i915_translate_compare_func(i915->depth_stencil->stencil.front_func); - int fop = i915_translate_stencil_op(i915->depth_stencil->stencil.front_fail_op); - int dfop = i915_translate_stencil_op(i915->depth_stencil->stencil.front_zfail_op); - int dpop = i915_translate_stencil_op(i915->depth_stencil->stencil.front_zpass_op); - int ref = i915->depth_stencil->stencil.ref_value[0] & 0xff; - - LIS5 |= (S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE | - (ref << S5_STENCIL_REF_SHIFT) | - (test << S5_STENCIL_TEST_FUNC_SHIFT) | - (fop << S5_STENCIL_FAIL_SHIFT) | - (dfop << S5_STENCIL_PASS_Z_FAIL_SHIFT) | - (dpop << S5_STENCIL_PASS_Z_PASS_SHIFT)); - } + LIS5 |= i915->depth_stencil->stencil_LIS5; LIS5 |= i915->blend->LIS5; @@ -189,17 +174,9 @@ static void upload_S6( struct i915_context *i915 ) */ LIS6 |= i915->blend->LIS6; - /* I915_NEW_DEPTH + /* I915_NEW_DEPTH */ - if (i915->depth_stencil->depth.enabled) { - int func = i915_translate_compare_func(i915->depth_stencil->depth.func); - - LIS6 |= (S6_DEPTH_TEST_ENABLE | - (func << S6_DEPTH_TEST_FUNC_SHIFT)); - - if (i915->depth_stencil->depth.writemask) - LIS6 |= S6_DEPTH_WRITE_ENABLE; - } + LIS6 |= i915->depth_stencil->depth_LIS6; if (LIS6 != i915->current.immediate[I915_IMMEDIATE_S6]) { i915->current.immediate[I915_IMMEDIATE_S6] = LIS6; -- cgit v1.2.3 From 893f9cda7601fb937138629042dd9f6507eb6b5f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 20 Sep 2007 11:33:05 -0400 Subject: Allow drivers to cache the template in case they want to fallback through softpipe which will require the template. --- src/mesa/state_tracker/st_cache.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 007a2311e9..01d1934232 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -53,7 +53,7 @@ const struct cso_blend * st_cached_blend_state(struct st_context *st, if (cso_hash_iter_is_null(iter)) { struct cso_blend *cso = malloc(sizeof(struct cso_blend)); memcpy(&cso->state, templ, sizeof(struct pipe_blend_state)); - cso->data = st->pipe->create_blend_state(st->pipe, templ); + cso->data = st->pipe->create_blend_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; iter = cso_insert_state(st->cache, hash_key, CSO_BLEND, cso); @@ -72,7 +72,7 @@ st_cached_sampler_state(struct st_context *st, if (cso_hash_iter_is_null(iter)) { struct cso_sampler *cso = malloc(sizeof(struct cso_sampler)); memcpy(&cso->state, templ, sizeof(struct pipe_sampler_state)); - cso->data = st->pipe->create_sampler_state(st->pipe, templ); + cso->data = st->pipe->create_sampler_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; iter = cso_insert_state(st->cache, hash_key, CSO_SAMPLER, cso); @@ -92,7 +92,7 @@ st_cached_depth_stencil_state(struct st_context *st, if (cso_hash_iter_is_null(iter)) { struct cso_depth_stencil *cso = malloc(sizeof(struct cso_depth_stencil)); memcpy(&cso->state, templ, sizeof(struct pipe_depth_stencil_state)); - cso->data = st->pipe->create_depth_stencil_state(st->pipe, templ); + cso->data = st->pipe->create_depth_stencil_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; iter = cso_insert_state(st->cache, hash_key, CSO_DEPTH_STENCIL, cso); @@ -112,7 +112,7 @@ const struct cso_rasterizer* st_cached_rasterizer_state( if (cso_hash_iter_is_null(iter)) { struct cso_rasterizer *cso = malloc(sizeof(struct cso_rasterizer)); memcpy(&cso->state, templ, sizeof(struct pipe_rasterizer_state)); - cso->data = st->pipe->create_rasterizer_state(st->pipe, templ); + cso->data = st->pipe->create_rasterizer_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; iter = cso_insert_state(st->cache, hash_key, CSO_RASTERIZER, cso); @@ -132,7 +132,7 @@ st_cached_fs_state(struct st_context *st, if (cso_hash_iter_is_null(iter)) { struct cso_fragment_shader *cso = malloc(sizeof(struct cso_fragment_shader)); memcpy(&cso->state, templ, sizeof(struct pipe_shader_state)); - cso->data = st->pipe->create_fs_state(st->pipe, templ); + cso->data = st->pipe->create_fs_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; iter = cso_insert_state(st->cache, hash_key, CSO_FRAGMENT_SHADER, cso); @@ -152,7 +152,7 @@ st_cached_vs_state(struct st_context *st, if (cso_hash_iter_is_null(iter)) { struct cso_vertex_shader *cso = malloc(sizeof(struct cso_vertex_shader)); memcpy(&cso->state, templ, sizeof(struct pipe_shader_state)); - cso->data = st->pipe->create_vs_state(st->pipe, templ); + cso->data = st->pipe->create_vs_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; iter = cso_insert_state(st->cache, hash_key, CSO_VERTEX_SHADER, cso); -- cgit v1.2.3 From cc2629f5912d1c608f830ab63f6c4e0875d2fcef Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 20 Sep 2007 11:33:35 -0400 Subject: Make the rasterizer state in i915 use the cso semantics. --- src/mesa/pipe/i915simple/i915_context.h | 16 ++++++- src/mesa/pipe/i915simple/i915_state.c | 64 +++++++++++++++++++++++-- src/mesa/pipe/i915simple/i915_state_derived.c | 5 +- src/mesa/pipe/i915simple/i915_state_dynamic.c | 34 ++++--------- src/mesa/pipe/i915simple/i915_state_immediate.c | 43 +---------------- 5 files changed, 87 insertions(+), 75 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 1a7df6566b..c582b069d7 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -127,6 +127,20 @@ struct i915_depth_stencil_state { unsigned depth_LIS6; }; +struct i915_rasterizer_state { + int light_twoside : 1; + unsigned st; + interp_mode color_interp; + + unsigned LIS4; + unsigned LIS7; + unsigned sc[1]; + + const struct pipe_rasterizer_state *templ; + + union { float f; unsigned u; } ds[2]; +}; + struct i915_context { struct pipe_context pipe; @@ -138,7 +152,7 @@ struct i915_context const struct i915_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct i915_depth_stencil_state *depth_stencil; - const struct pipe_rasterizer_state *rasterizer; + const struct i915_rasterizer_state *rasterizer; const struct pipe_shader_state *fs; struct pipe_alpha_test_state alpha_test; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 9611288961..c7086c17f5 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -31,6 +31,7 @@ #include "pipe/draw/draw_context.h" #include "pipe/p_winsys.h" +#include "pipe/p_util.h" #include "i915_context.h" #include "i915_reg.h" @@ -437,9 +438,64 @@ static void i915_set_viewport_state( struct pipe_context *pipe, static void * i915_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup) + const struct pipe_rasterizer_state *rasterizer) { - return 0; + struct i915_rasterizer_state *cso = calloc(1, sizeof(struct i915_rasterizer_state)); + + cso->templ = rasterizer; + cso->color_interp = rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + cso->light_twoside = rasterizer->light_twoside; + cso->ds[0].u = _3DSTATE_DEPTH_OFFSET_SCALE; + cso->ds[1].f = rasterizer->offset_scale; + if (rasterizer->poly_stipple_enable) { + cso->st |= ST1_ENABLE; + } + + if (rasterizer->scissor) + cso->sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT; + else + cso->sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT; + + switch (rasterizer->cull_mode) { + case PIPE_WINDING_NONE: + cso->LIS4 |= S4_CULLMODE_NONE; + break; + case PIPE_WINDING_CW: + cso->LIS4 |= S4_CULLMODE_CW; + break; + case PIPE_WINDING_CCW: + cso->LIS4 |= S4_CULLMODE_CCW; + break; + case PIPE_WINDING_BOTH: + cso->LIS4 |= S4_CULLMODE_BOTH; + break; + } + + { + int line_width = CLAMP((int)(rasterizer->line_width * 2), 1, 0xf); + + cso->LIS4 |= line_width << S4_LINE_WIDTH_SHIFT; + + if (rasterizer->line_smooth) + cso->LIS4 |= S4_LINE_ANTIALIAS_ENABLE; + } + + { + int point_size = CLAMP((int) rasterizer->point_size, 1, 0xff); + + cso->LIS4 |= point_size << S4_POINT_WIDTH_SHIFT; + } + + if (rasterizer->flatshade) { + cso->LIS4 |= (S4_FLATSHADE_ALPHA | + S4_FLATSHADE_COLOR | + S4_FLATSHADE_SPECULAR); + } + + cso->LIS7 = rasterizer->offset_units; /* probably incorrect */ + + + return cso; } static void i915_bind_rasterizer_state( struct pipe_context *pipe, @@ -447,10 +503,10 @@ static void i915_bind_rasterizer_state( struct pipe_context *pipe, { struct i915_context *i915 = i915_context(pipe); - i915->rasterizer = (struct pipe_rasterizer_state *)setup; + i915->rasterizer = (struct i915_rasterizer_state *)setup; /* pass-through to draw module */ - draw_set_rasterizer_state(i915->draw, setup); + draw_set_rasterizer_state(i915->draw, i915->rasterizer->templ); i915->dirty |= I915_NEW_RASTERIZER; } diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index dece697497..572d270f39 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -44,8 +44,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) { const struct pipe_shader_state *fs = i915->fs; - const interp_mode colorInterp - = i915->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + const interp_mode colorInterp = i915->rasterizer->color_interp; struct vertex_info *vinfo = &i915->current.vertex_info; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; boolean needW = 0; @@ -161,7 +160,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) if (front0) { back0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, FORMAT_OMIT, colorInterp); - } + } if (back0) { back1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, FORMAT_OMIT, colorInterp); diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index d97aeb8d3b..845873baa0 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -165,18 +165,9 @@ const struct i915_tracked_state i915_upload_IAB = { static void upload_DEPTHSCALE( struct i915_context *i915 ) { - union { float f; unsigned u; } ds[2]; - - memset( ds, 0, sizeof(ds) ); - - /* I915_NEW_RASTERIZER - */ - ds[0].u = _3DSTATE_DEPTH_OFFSET_SCALE; - ds[1].f = i915->rasterizer->offset_scale; - - set_dynamic_indirect( i915, + set_dynamic_indirect( i915, I915_DYNAMIC_DEPTHSCALE_0, - &ds[0].u, + &(i915->rasterizer->ds[0].u), 2 ); } @@ -205,12 +196,10 @@ static void upload_STIPPLE( struct i915_context *i915 ) st[0] = _3DSTATE_STIPPLE; st[1] = 0; - - /* I915_NEW_RASTERIZER + + /* I915_NEW_RASTERIZER */ - if (i915->rasterizer->poly_stipple_enable) { - st[1] |= ST1_ENABLE; - } + st[1] |= i915->rasterizer->st; /* I915_NEW_STIPPLE @@ -248,20 +237,13 @@ const struct i915_tracked_state i915_upload_STIPPLE = { /*********************************************************************** - * Scissor. + * Scissor. */ static void upload_SCISSOR_ENABLE( struct i915_context *i915 ) { - unsigned sc[1]; - - if (i915->rasterizer->scissor) - sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT; - else - sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT; - - set_dynamic_indirect( i915, + set_dynamic_indirect( i915, I915_DYNAMIC_SC_ENA_0, - &sc[0], + &(i915->rasterizer->sc[0]), 1 ); } diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 7de7f5377e..874c3819f2 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -62,46 +62,7 @@ static void upload_S2S4(struct i915_context *i915) assert(LIS4); /* should never be zero? */ } - /* I915_NEW_RASTERIZER */ - switch (i915->rasterizer->cull_mode) { - case PIPE_WINDING_NONE: - LIS4 |= S4_CULLMODE_NONE; - break; - case PIPE_WINDING_CW: - LIS4 |= S4_CULLMODE_CW; - break; - case PIPE_WINDING_CCW: - LIS4 |= S4_CULLMODE_CCW; - break; - case PIPE_WINDING_BOTH: - LIS4 |= S4_CULLMODE_BOTH; - break; - } - - /* I915_NEW_RASTERIZER */ - { - int line_width = CLAMP((int)(i915->rasterizer->line_width * 2), 1, 0xf); - - LIS4 |= line_width << S4_LINE_WIDTH_SHIFT; - - if (i915->rasterizer->line_smooth) - LIS4 |= S4_LINE_ANTIALIAS_ENABLE; - } - - /* I915_NEW_RASTERIZER */ - { - int point_size = CLAMP((int) i915->rasterizer->point_size, 1, 0xff); - - LIS4 |= point_size << S4_POINT_WIDTH_SHIFT; - } - - /* I915_NEW_RASTERIZER */ - if (i915->rasterizer->flatshade) { - LIS4 |= (S4_FLATSHADE_ALPHA | - S4_FLATSHADE_COLOR | - S4_FLATSHADE_SPECULAR); - } - + LIS4 |= i915->rasterizer->LIS4; if (LIS2 != i915->current.immediate[I915_IMMEDIATE_S2] || LIS4 != i915->current.immediate[I915_IMMEDIATE_S4]) { @@ -198,7 +159,7 @@ static void upload_S7( struct i915_context *i915 ) /* I915_NEW_RASTERIZER */ - LIS7 = i915->rasterizer->offset_units; /* probably incorrect */ + LIS7 = i915->rasterizer->LIS7; /* probably incorrect */ if (LIS7 != i915->current.immediate[I915_IMMEDIATE_S7]) { i915->current.immediate[I915_IMMEDIATE_S7] = LIS7; -- cgit v1.2.3 From 742e32a40bf5ef1bd90b23aa0f7d451b7b7f0ba3 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 20 Sep 2007 12:34:31 -0400 Subject: Cache the i915 sampler state. --- src/mesa/pipe/i915simple/i915_context.h | 7 +- src/mesa/pipe/i915simple/i915_state.c | 129 ++++++++++++++++++- src/mesa/pipe/i915simple/i915_state_sampler.c | 171 ++++---------------------- 3 files changed, 156 insertions(+), 151 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index c582b069d7..f1e10f3433 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -141,6 +141,11 @@ struct i915_rasterizer_state { union { float f; unsigned u; } ds[2]; }; +struct i915_sampler_state { + unsigned state[3]; + const struct pipe_sampler_state *templ; +}; + struct i915_context { struct pipe_context pipe; @@ -150,7 +155,7 @@ struct i915_context /* The most recent drawing state as set by the driver: */ const struct i915_blend_state *blend; - const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct i915_depth_stencil_state *depth_stencil; const struct i915_rasterizer_state *rasterizer; const struct pipe_shader_state *fs; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index c7086c17f5..525f8ce13a 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -38,9 +38,79 @@ #include "i915_state.h" #include "i915_state_inlines.h" -/* None of this state is actually used for anything yet. + +/* The i915 (and related graphics cores) do not support GL_CLAMP. The + * Intel drivers for "other operating systems" implement GL_CLAMP as + * GL_CLAMP_TO_EDGE, so the same is done here. */ +static unsigned +translate_wrap_mode(unsigned wrap) +{ + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + return TEXCOORDMODE_WRAP; + case PIPE_TEX_WRAP_CLAMP: + return TEXCOORDMODE_CLAMP_EDGE; /* not quite correct */ + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return TEXCOORDMODE_CLAMP_EDGE; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return TEXCOORDMODE_CLAMP_BORDER; +// case PIPE_TEX_WRAP_MIRRORED_REPEAT: +// return TEXCOORDMODE_MIRROR; + default: + return TEXCOORDMODE_WRAP; + } +} +static unsigned translate_img_filter( unsigned filter ) +{ + switch (filter) { + case PIPE_TEX_FILTER_NEAREST: + return FILTER_NEAREST; + case PIPE_TEX_FILTER_LINEAR: + return FILTER_LINEAR; + default: + assert(0); + return FILTER_NEAREST; + } +} + +static unsigned translate_mip_filter( unsigned filter ) +{ + switch (filter) { + case PIPE_TEX_MIPFILTER_NONE: + return MIPFILTER_NONE; + case PIPE_TEX_MIPFILTER_NEAREST: + return MIPFILTER_NEAREST; + case PIPE_TEX_FILTER_LINEAR: + return MIPFILTER_LINEAR; + default: + assert(0); + return MIPFILTER_NONE; + } +} + +static unsigned translate_compare_func(unsigned func) +{ + switch (func) { + case PIPE_FUNC_NEVER: + case PIPE_FUNC_LESS: + case PIPE_FUNC_EQUAL: + case PIPE_FUNC_LEQUAL: + case PIPE_FUNC_GREATER: + case PIPE_FUNC_NOTEQUAL: + case PIPE_FUNC_GEQUAL: + case PIPE_FUNC_ALWAYS: + return 0; + default: + assert(0); + return 0; + } +} + + +/* None of this state is actually used for anything yet. + */ static void * i915_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) @@ -147,7 +217,59 @@ static void * i915_create_sampler_state(struct pipe_context *pipe, const struct pipe_sampler_state *sampler) { - return 0; + struct i915_sampler_state *cso = calloc(1, sizeof(struct i915_sampler_state)); + cso->templ = sampler; + + const unsigned ws = sampler->wrap_s; + const unsigned wt = sampler->wrap_t; + const unsigned wr = sampler->wrap_r; + unsigned minFilt, magFilt; + unsigned mipFilt; + + mipFilt = translate_mip_filter(sampler->min_mip_filter); + if (sampler->max_anisotropy > 1.0) { + minFilt = FILTER_ANISOTROPIC; + magFilt = FILTER_ANISOTROPIC; + } + else { + minFilt = translate_img_filter( sampler->min_img_filter ); + magFilt = translate_img_filter( sampler->mag_img_filter ); + } + + { + int b = sampler->lod_bias * 16.0; + b = CLAMP(b, -256, 255); + cso->state[0] |= ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); + } + + /* Shadow: + */ + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) + { + cso->state[0] |= (SS2_SHADOW_ENABLE | + translate_compare_func(sampler->compare_func)); + + minFilt = FILTER_4X4_FLAT; + magFilt = FILTER_4X4_FLAT; + } + + cso->state[0] |= ((minFilt << SS2_MIN_FILTER_SHIFT) | + (mipFilt << SS2_MIP_FILTER_SHIFT) | + (magFilt << SS2_MAG_FILTER_SHIFT)); + + cso->state[1] |= + ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT)); + + { + ubyte r = float_to_ubyte(sampler->border_color[0]); + ubyte g = float_to_ubyte(sampler->border_color[1]); + ubyte b = float_to_ubyte(sampler->border_color[2]); + ubyte a = float_to_ubyte(sampler->border_color[3]); + cso->state[2] = I915PACKCOLOR8888(r, g, b, a); + } + return cso; } static void i915_bind_sampler_state(struct pipe_context *pipe, @@ -156,7 +278,7 @@ static void i915_bind_sampler_state(struct pipe_context *pipe, struct i915_context *i915 = i915_context(pipe); assert(unit < PIPE_MAX_SAMPLERS); - i915->sampler[unit] = (const struct pipe_sampler_state *)sampler; + i915->sampler[unit] = (const struct i915_sampler_state*)sampler; i915->dirty |= I915_NEW_SAMPLER; } @@ -164,6 +286,7 @@ static void i915_bind_sampler_state(struct pipe_context *pipe, static void i915_delete_sampler_state(struct pipe_context *pipe, void *sampler) { + free(sampler); } diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 419a156136..ebfd29fadb 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -42,81 +42,6 @@ #include "i915_state.h" //#include "i915_cache.h" - - - - - -/* The i915 (and related graphics cores) do not support GL_CLAMP. The - * Intel drivers for "other operating systems" implement GL_CLAMP as - * GL_CLAMP_TO_EDGE, so the same is done here. - */ -static unsigned -translate_wrap_mode(unsigned wrap) -{ - switch (wrap) { - case PIPE_TEX_WRAP_REPEAT: - return TEXCOORDMODE_WRAP; - case PIPE_TEX_WRAP_CLAMP: - return TEXCOORDMODE_CLAMP_EDGE; /* not quite correct */ - case PIPE_TEX_WRAP_CLAMP_TO_EDGE: - return TEXCOORDMODE_CLAMP_EDGE; - case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - return TEXCOORDMODE_CLAMP_BORDER; -// case PIPE_TEX_WRAP_MIRRORED_REPEAT: -// return TEXCOORDMODE_MIRROR; - default: - return TEXCOORDMODE_WRAP; - } -} - -static unsigned translate_img_filter( unsigned filter ) -{ - switch (filter) { - case PIPE_TEX_FILTER_NEAREST: - return FILTER_NEAREST; - case PIPE_TEX_FILTER_LINEAR: - return FILTER_LINEAR; - default: - assert(0); - return FILTER_NEAREST; - } -} - -static unsigned translate_mip_filter( unsigned filter ) -{ - switch (filter) { - case PIPE_TEX_MIPFILTER_NONE: - return MIPFILTER_NONE; - case PIPE_TEX_MIPFILTER_NEAREST: - return MIPFILTER_NEAREST; - case PIPE_TEX_FILTER_LINEAR: - return MIPFILTER_LINEAR; - default: - assert(0); - return MIPFILTER_NONE; - } -} - -static unsigned translate_compare_func(unsigned func) -{ - switch (func) { - case PIPE_FUNC_NEVER: - case PIPE_FUNC_LESS: - case PIPE_FUNC_EQUAL: - case PIPE_FUNC_LEQUAL: - case PIPE_FUNC_GREATER: - case PIPE_FUNC_NOTEQUAL: - case PIPE_FUNC_GEQUAL: - case PIPE_FUNC_ALWAYS: - return 0; - default: - assert(0); - return 0; - } -} - - static uint bitcount(uint k) { @@ -153,60 +78,21 @@ is_power_of_two_texture(const struct pipe_mipmap_tree *mt) */ static void update_sampler(struct i915_context *i915, uint unit, - const struct pipe_sampler_state *sampler, + const struct i915_sampler_state *sampler, const struct pipe_mipmap_tree *mt, unsigned state[3] ) { - const unsigned ws = sampler->wrap_s; - const unsigned wt = sampler->wrap_t; - const unsigned wr = sampler->wrap_r; - /* Need to do this after updating the maps, which call the * intel_finalize_mipmap_tree and hence can update firstLevel: */ - unsigned minFilt, magFilt; - unsigned mipFilt; - - state[0] = state[1] = state[2] = 0; - - mipFilt = translate_mip_filter( sampler->min_mip_filter ); - - if (sampler->max_anisotropy > 1.0) { - minFilt = FILTER_ANISOTROPIC; - magFilt = FILTER_ANISOTROPIC; - } - else { - minFilt = translate_img_filter( sampler->min_img_filter ); - magFilt = translate_img_filter( sampler->mag_img_filter ); - } - - { - int b = sampler->lod_bias * 16.0; - b = CLAMP(b, -256, 255); - state[0] |= ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); - } + state[0] = sampler->state[0]; + state[1] = sampler->state[1]; + state[2] = sampler->state[2]; if (mt->format == PIPE_FORMAT_YCBCR || mt->format == PIPE_FORMAT_YCBCR_REV) state[0] |= SS2_COLORSPACE_CONVERSION; - - /* Shadow: - */ - if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) - { - state[0] |= (SS2_SHADOW_ENABLE | - translate_compare_func(sampler->compare_func)); - - minFilt = FILTER_4X4_FLAT; - magFilt = FILTER_4X4_FLAT; - } - - state[0] |= ((minFilt << SS2_MIN_FILTER_SHIFT) | - (mipFilt << SS2_MIP_FILTER_SHIFT) | - (magFilt << SS2_MAG_FILTER_SHIFT)); - - /* 3D textures don't seem to respect the border color. * Fallback if there's ever a danger that they might refer to * it. @@ -217,45 +103,36 @@ static void update_sampler(struct i915_context *i915, * XXX: Check if this is true on i945. * XXX: Check if this bug got fixed in release silicon. */ - if (mt->target == PIPE_TEXTURE_3D && - (sampler->min_img_filter != PIPE_TEX_FILTER_NEAREST || - sampler->mag_img_filter != PIPE_TEX_FILTER_NEAREST) && - (ws == PIPE_TEX_WRAP_CLAMP || - wt == PIPE_TEX_WRAP_CLAMP || - wr == PIPE_TEX_WRAP_CLAMP || - ws == PIPE_TEX_WRAP_CLAMP_TO_BORDER || - wt == PIPE_TEX_WRAP_CLAMP_TO_BORDER || - wr == PIPE_TEX_WRAP_CLAMP_TO_BORDER)) { #if 0 - if (i915->strict_conformance) { - assert(0); - /* sampler->fallback = true; */ - /* TODO */ + { + const unsigned ws = sampler->templ->wrap_s; + const unsigned wt = sampler->templ->wrap_t; + const unsigned wr = sampler->templ->wrap_r; + if (mt->target == PIPE_TEXTURE_3D && + (sampler->templ->min_img_filter != PIPE_TEX_FILTER_NEAREST || + sampler->templ->mag_img_filter != PIPE_TEX_FILTER_NEAREST) && + (ws == PIPE_TEX_WRAP_CLAMP || + wt == PIPE_TEX_WRAP_CLAMP || + wr == PIPE_TEX_WRAP_CLAMP || + ws == PIPE_TEX_WRAP_CLAMP_TO_BORDER || + wt == PIPE_TEX_WRAP_CLAMP_TO_BORDER || + wr == PIPE_TEX_WRAP_CLAMP_TO_BORDER)) { + if (i915->strict_conformance) { + assert(0); + /* sampler->fallback = true; */ + /* TODO */ + } } -#endif } +#endif - state[1] = - ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) | - (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | - (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT) | - (unit << SS3_TEXTUREMAP_INDEX_SHIFT)); + state[1] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT); if (is_power_of_two_texture(mt)) { state[1] |= SS3_NORMALIZED_COORDS; } - - { - ubyte r = float_to_ubyte(sampler->border_color[0]); - ubyte g = float_to_ubyte(sampler->border_color[1]); - ubyte b = float_to_ubyte(sampler->border_color[2]); - ubyte a = float_to_ubyte(sampler->border_color[3]); - state[2] = I915PACKCOLOR8888(r, g, b, a); - } } - - void i915_update_samplers( struct i915_context *i915 ) { uint unit; -- cgit v1.2.3 From b5908a7671b408835777931e7180b8264f150bbd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 09:43:51 -0600 Subject: fix input/output typos --- src/mesa/state_tracker/st_atom_vs.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 078c052ae2..cc61436f53 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -106,16 +106,16 @@ st_translate_vertex_shader(struct st_context *st, vs.output_semantics[vs.num_outputs] = TGSI_SEMANTIC_POSITION; break; case VERT_RESULT_COL0: - vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0; + vs.output_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0; break; case VERT_RESULT_COL1: - vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1; + vs.output_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1; break; case VERT_RESULT_BFC0: - vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0B; + vs.output_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0B; break; case VERT_RESULT_BFC1: - vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1B; + vs.output_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1B; break; default: vs.output_semantics[vs.num_outputs] = TGSI_SEMANTIC_OTHER; -- cgit v1.2.3 From 768302c53971ad0405ee6e2a30db96fc5f51913b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 09:55:56 -0600 Subject: remove unused lookup[] --- src/mesa/pipe/draw/draw_twoside.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 3eb8cce637..9547f2a67b 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -36,7 +36,9 @@ struct twoside_stage { struct draw_stage stage; float sign; /**< +1 or -1 */ +#if 0 const unsigned *lookup; /**< vertex attribute positions */ +#endif }; @@ -162,7 +164,9 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) twoside->stage.end = twoside_end; twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; +#if 0 twoside->lookup = draw->vertex_info.attrib_to_slot; +#endif return &twoside->stage; } -- cgit v1.2.3 From 745f0cbe0528ac925096f5c1b85de7280fee7fbc Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 13:39:17 -0600 Subject: Always update st->state.vs, not just when the program is dirty. This fixes a regression in the cubemap.c demo which alternates between two different vertex shaders. --- src/mesa/state_tracker/st_atom_fs.c | 8 +++++--- src/mesa/state_tracker/st_atom_vs.c | 9 +++++---- 2 files changed, 10 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 91e58f5831..340c7ec69f 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -177,11 +177,13 @@ static void update_fs( struct st_context *st ) /* if new binding, or shader has changed */ if (st->fp != stfp || stfp->dirty) { - /* Bind the program */ - st->fp = stfp; if (stfp->dirty) - st->state.fs = st_translate_fragment_shader( st, st->fp ); + (void) st_translate_fragment_shader( st, stfp ); + + /* Bind the vertex program and TGSI shader */ + st->fp = stfp; + st->state.fs = stfp->fs; st->pipe->bind_fs_state(st->pipe, st->state.fs->data); } diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index cc61436f53..9c2994fddf 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -188,11 +188,12 @@ static void update_vs( struct st_context *st ) } if (st->vp != stvp || stvp->dirty) { - /* Bind the vertex program */ - st->vp = stvp; - if (stvp->dirty) - st->state.vs = st_translate_vertex_shader( st, st->vp ); + (void) st_translate_vertex_shader( st, stvp ); + + /* Bind the vertex program and TGSI shader */ + st->vp = stvp; + st->state.vs = stvp->vs; st->pipe->bind_vs_state(st->pipe, st->state.vs->data); } -- cgit v1.2.3 From 086734502a614e7778533018846ee66a66df9821 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 13:42:37 -0600 Subject: Checkpoint: vertex attribute clean-up. Remove/disable the attrib/slot mapping arrays in a few places. Work in progress... --- src/mesa/pipe/draw/draw_clip.c | 2 ++ src/mesa/pipe/draw/draw_feedback.c | 3 ++- src/mesa/pipe/draw/draw_private.h | 2 +- src/mesa/pipe/draw/draw_vertex.c | 19 +++++++------------ src/mesa/pipe/draw/draw_vertex.h | 16 ++++++++-------- src/mesa/pipe/draw/draw_vertex_fetch.c | 5 +++++ src/mesa/pipe/i915simple/i915_state_derived.c | 2 +- src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_prim_setup.c | 11 ++++------- src/mesa/pipe/softpipe/sp_state_derived.c | 7 +++++-- src/mesa/state_tracker/st_atom_vs.c | 23 ++++++++++++----------- src/mesa/state_tracker/st_cb_clear.c | 2 +- src/mesa/state_tracker/st_cb_feedback.c | 10 ++++++++-- src/mesa/state_tracker/st_context.h | 12 ++++++++++-- src/mesa/state_tracker/st_program.h | 5 ----- 15 files changed, 67 insertions(+), 53 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 1396b60f45..e2af69e048 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -382,8 +382,10 @@ static void clip_begin( struct draw_stage *stage ) { /* sanity checks. If these fail, review the clip/interp code! */ assert(stage->draw->vertex_info.num_attribs >= 3); +#if 0 assert(stage->draw->vertex_info.slot_to_attrib[0] == TGSI_ATTRIB_VERTEX_HEADER); assert(stage->draw->vertex_info.slot_to_attrib[1] == TGSI_ATTRIB_CLIP_POS); +#endif stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/draw/draw_feedback.c b/src/mesa/pipe/draw/draw_feedback.c index ecdd98e83d..3b8400233e 100644 --- a/src/mesa/pipe/draw/draw_feedback.c +++ b/src/mesa/pipe/draw/draw_feedback.c @@ -78,7 +78,7 @@ feedback_vertex(struct draw_stage *stage, const struct vertex_header *vertex) * we can either address output buffer 0 (for interleaving) or * output buffer i (for non-interleaved). */ - +#if 0 for (i = 0; i < feedback->num_attribs; i++) { const uint attr = feedback->attrib[i]; const uint slot = stage->draw->vertex_info.attrib_to_slot[attr]; @@ -104,6 +104,7 @@ feedback_vertex(struct draw_stage *stage, const struct vertex_header *vertex) } fs->dest[i * select] += size; } +#endif fs->num_vert_emitted++; } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index bd8c11eb94..1285f3200c 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -62,7 +62,7 @@ struct vertex_header { }; /* XXX This is too large */ -#define MAX_VERTEX_SIZE ((2 + TGSI_ATTRIB_MAX) * 4 * sizeof(float)) +#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index 1c7e1d8662..ab4e6c7864 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -45,17 +45,10 @@ static INLINE void -emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, +emit_vertex_attr(struct vertex_info *vinfo, /*uint vfAttr,*/ attrib_format format, interp_mode interp) { const uint n = vinfo->num_attribs; - vinfo->attr_mask |= (1 << vfAttr); - vinfo->slot_to_attrib[n] = vfAttr; - if (n >= 2) { - /* the first two slots are the vertex header & clippos */ - assert(vfAttr < Elements(vinfo->attrib_to_slot)); - vinfo->attrib_to_slot[vfAttr] = n - 2; - } vinfo->interp_mode[n] = interp; vinfo->format[n] = format; vinfo->num_attribs++; @@ -110,22 +103,24 @@ draw_set_vertex_attributes( struct draw_context *draw, struct vertex_info *vinfo = &draw->vertex_info; unsigned i; +#if 0 assert(slot_to_vf_attr[0] == TGSI_ATTRIB_POS); +#endif memset(vinfo, 0, sizeof(*vinfo)); /* * First three attribs are always the same: header, clip pos, winpos */ - emit_vertex_attr(vinfo, TGSI_ATTRIB_VERTEX_HEADER, FORMAT_1F, INTERP_NONE); - emit_vertex_attr(vinfo, TGSI_ATTRIB_CLIP_POS, FORMAT_4F, INTERP_LINEAR); - emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F_VIEWPORT, INTERP_LINEAR); + emit_vertex_attr(vinfo, /*TGSI_ATTRIB_VERTEX_HEADER,*/ FORMAT_1F, INTERP_NONE); + emit_vertex_attr(vinfo, /*TGSI_ATTRIB_CLIP_POS,*/ FORMAT_4F, INTERP_LINEAR); + emit_vertex_attr(vinfo, /*TGSI_ATTRIB_POS,*/ FORMAT_4F_VIEWPORT, INTERP_LINEAR); /* * Remaining attribs (color, texcoords, etc) */ for (i = 1; i < nr_attrs; i++) { - emit_vertex_attr(vinfo, slot_to_vf_attr[i], FORMAT_4F, interps[i]); + emit_vertex_attr(vinfo, /*slot_to_vf_attr[i],*/ FORMAT_4F, interps[i]); } draw_compute_vertex_size(vinfo); diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 4e3e86d86e..1d900b3a4e 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -68,16 +68,15 @@ typedef enum { } interp_mode; - +/** + * Information about post-transformed vertex layout. + */ struct vertex_info { uint num_attribs; uint hwfmt[4]; /**< hardware format info for this format */ - uint attr_mask; /**< mask of VF_ATTR_ bits */ - uint slot_to_attrib[MAX_VERT_ATTRIBS]; - uint attrib_to_slot[TGSI_ATTRIB_MAX]; - interp_mode interp_mode[MAX_VERT_ATTRIBS]; - attrib_format format[MAX_VERT_ATTRIBS]; /**< FORMAT_x */ + interp_mode interp_mode[PIPE_MAX_SHADER_OUTPUTS]; + attrib_format format[PIPE_MAX_SHADER_OUTPUTS]; /**< FORMAT_x */ uint size; /**< total vertex size in dwords */ }; @@ -92,9 +91,10 @@ draw_emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, attrib_format format, interp_mode interp) { const uint n = vinfo->num_attribs; - assert(n < MAX_VERT_ATTRIBS); + assert(n < PIPE_MAX_SHADER_OUTPUTS); + /* vinfo->attr_mask |= (1 << vfAttr); - vinfo->slot_to_attrib[n] = vfAttr; + */ vinfo->format[n] = format; vinfo->interp_mode[n] = interp; vinfo->num_attribs++; diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index ce402d681f..62e8d61be4 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -81,6 +81,9 @@ void draw_vertex_fetch( struct draw_context *draw, /* loop over vertices */ for (j = 0; j < count; j++) { uint attr; + + /*printf("fetch vertex %u: \n", j);*/ + /* loop over vertex attributes (vertex shader inputs) */ for (attr = 0; attr < draw->vertex_shader.num_inputs; attr++) { @@ -94,6 +97,8 @@ void draw_vertex_fetch( struct draw_context *draw, fetch_attrib4(src, draw->vertex_element[attr].src_format, p); + /*printf(" %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]);*/ + machine->Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ machine->Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ machine->Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 572d270f39..91d00fb1c2 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -173,7 +173,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) * vertex layout. We'll also update the hardware vertex format info. */ draw_set_vertex_attributes( i915->draw, - vinfo->slot_to_attrib, + NULL,/*vinfo->slot_to_attrib,*/ vinfo->interp_mode, vinfo->num_attribs); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 5c17c47b12..95215eb640 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -111,6 +111,7 @@ struct softpipe_context { unsigned nr_frag_attrs; /**< number of active fragment attribs */ boolean need_z; /**< produce quad/fragment Z values? */ boolean need_w; /**< produce quad/fragment W values? */ + int psize_slot; /** Feedback buffers */ struct pipe_feedback_buffer feedback_buffer[PIPE_MAX_FEEDBACK_ATTRIBS]; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 8a4be79d11..913ae44601 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -79,8 +79,6 @@ struct setup_stage { float oneoverarea; - const unsigned *lookup; /**< vertex attribute positions */ - struct tgsi_interp_coef coef[TGSI_ATTRIB_MAX]; struct quad_header quad; @@ -249,7 +247,7 @@ static void print_vertex(const struct setup_stage *setup, { int i; printf("Vertex:\n"); - for (i = 0; i < setup->softpipe->nr_attrs; i++) { + for (i = 0; i < setup->quad.nr_attrs; i++) { printf(" %d: %f %f %f\n", i, v->data[i][0], v->data[i][1], v->data[i][2]); } @@ -907,8 +905,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); const struct vertex_header *v0 = prim->v[0]; - - const int sizeAttr = setup->lookup[TGSI_ATTRIB_POINTSIZE]; + const int sizeAttr = setup->softpipe->psize_slot; const float halfSize = sizeAttr ? (0.5f * v0->data[sizeAttr][0]) : (0.5f * setup->softpipe->rasterizer->point_size); @@ -917,6 +914,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const float y = v0->data[TGSI_ATTRIB_POS][1]; unsigned slot, j; + assert(sizeAttr >= 0); + /* For points, all interpolants are constant-valued. * However, for point sprites, we'll need to setup texcoords appropriately. * XXX: which coefficients are the texcoords??? @@ -1097,7 +1096,5 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) setup->quad.coef = setup->coef; - setup->lookup = softpipe->draw->vertex_info.attrib_to_slot; - return &setup->stage; } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 0dd0eea0b8..03b5d7ea3f 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -59,6 +59,8 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) softpipe->need_z = FALSE; softpipe->need_w = FALSE; + softpipe->psize_slot = -1; + /* always emit vertex pos */ /* TODO - Figure out if we need to do perspective divide, etc. */ draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F, INTERP_LINEAR); @@ -93,6 +95,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) FORMAT_4F, INTERP_CONSTANT); break; #endif + softpipe->psize_slot = i; /*case TGSI_SEMANTIC_TEXCOORD:*/ case TGSI_SEMANTIC_TEX0: draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_TEX0, @@ -131,10 +134,10 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ /* XXX we also need to do this when the shading mode (interp modes) change: */ if (1/*vinfo->attr_mask != softpipe->attr_mask*/) { - softpipe->attr_mask = vinfo->attr_mask; + /*softpipe->attr_mask = vinfo->attr_mask;*/ draw_set_vertex_attributes( softpipe->draw, - vinfo->slot_to_attrib, + NULL,/*vinfo->slot_to_attrib,*/ vinfo->interp_mode, vinfo->num_attribs); diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 9c2994fddf..2b8aef5c63 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -57,7 +57,6 @@ const struct cso_vertex_shader * st_translate_vertex_shader(struct st_context *st, struct st_vertex_program *stvp) { - GLuint outputMapping[PIPE_MAX_SHADER_INPUTS]; struct pipe_shader_state vs; const struct cso_vertex_shader *cso; GLuint i; @@ -83,6 +82,9 @@ st_translate_vertex_shader(struct st_context *st, case VERT_ATTRIB_COLOR1: vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1; break; + case VERT_ATTRIB_TEX0: + vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_TEX0; + break; default: vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_OTHER; } @@ -95,11 +97,8 @@ st_translate_vertex_shader(struct st_context *st, */ for (i = 0; i < VERT_RESULT_MAX; i++) { if (stvp->Base.Base.OutputsWritten & (1 << i)) { -#if 0 - stvp->output_to_index[i] = vs.num_outputs; - stvp->index_to_output[vs.num_outputs] = i; -#endif - outputMapping[i] = vs.num_outputs; + /* put this attrib in the next available slot */ + st->vertex_attrib_to_slot[i] = vs.num_outputs; switch (i) { case VERT_RESULT_HPOS: @@ -129,11 +128,7 @@ st_translate_vertex_shader(struct st_context *st, */ tgsi_mesa_compile_vp_program( &stvp->Base, stvp->input_to_index, -#if 0 - stvp->output_to_index, -#else - outputMapping, -#endif + st->vertex_attrib_to_slot, stvp->tokens, ST_FP_MAX_TOKENS ); #if 0 @@ -195,6 +190,12 @@ static void update_vs( struct st_context *st ) st->vp = stvp; st->state.vs = stvp->vs; +#if 0 + printf("###### bind vp tokens: %p %p num_inp=%u\n", + stvp, stvp->tokens, stvp->vs->state.num_inputs); + if (TGSI_DEBUG) + tgsi_dump( stvp->tokens, 0 ); +#endif st->pipe->bind_vs_state(st->pipe, st->state.vs->data); } } diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index bfc977daa4..5e63205088 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -491,7 +491,7 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) assert(strb->surface->format); -#if 01 +#if 0 if (ctx->Scissor.Enabled || (isDS && ctx->DrawBuffer->Visual.stencilBits > 0)) { /* scissoring or we have a combined depth/stencil buffer */ diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 8e8084fe59..e846463c4c 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -87,6 +87,7 @@ static void feedback_vertex(GLcontext *ctx, const struct draw_context *draw, const struct vertex_header *v) { + const struct st_context *st = ctx->st; GLfloat win[4]; const GLfloat *color, *texcoord; const GLfloat ci = 0; @@ -97,13 +98,18 @@ feedback_vertex(GLcontext *ctx, const struct draw_context *draw, win[2] = v->data[0][2]; win[3] = 1.0F / v->data[0][3]; - slot = draw->vertex_info.attrib_to_slot[TGSI_ATTRIB_COLOR0]; + /* XXX + * When we compute vertex layout, save info about position of the + * color and texcoord attribs to use here. + */ + + slot = st->vertex_attrib_to_slot[VERT_RESULT_COL0]; if (slot) color = v->data[slot]; else color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; - slot = draw->vertex_info.attrib_to_slot[TGSI_ATTRIB_TEX0]; + slot = st->vertex_attrib_to_slot[VERT_RESULT_TEX0]; if (slot) texcoord = v->data[slot]; else diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 8a57227c84..55a857f46d 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -119,9 +119,17 @@ struct st_context GLfloat polygon_offset_scale; /* ?? */ - struct st_vertex_program *vp; - struct st_fragment_program *fp; + /** Mapping from VERT_ATTRIB_x to post-transformed vertex slot */ + GLuint vertex_attrib_to_slot[VERT_RESULT_MAX]; + struct st_vertex_program *vp; /**< Currently bound vertex program */ + struct st_fragment_program *fp; /**< Currently bound fragment program */ + + /** + * Buffer object which stores the ctx->Current.Attrib[] values. + * Used for vertex array drawing when we we need an attribute for + * which there's no enabled array. + */ struct pipe_buffer_handle *default_attrib_buffer; struct cso_cache *cache; diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 4f9ace3e6a..c21e27628e 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -71,11 +71,6 @@ struct st_vertex_program /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ GLuint index_to_input[MAX_VERTEX_PROGRAM_ATTRIBS]; -#if 0 - GLuint output_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; - GLuint index_to_output[MAX_VERTEX_PROGRAM_ATTRIBS]; -#endif - /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; -- cgit v1.2.3 From 674d0130693f827ae9ea8e902a38c46d64444384 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 13:50:53 -0600 Subject: checkpoint: TGSI_ATTRIB_x tokens no longer used --- src/mesa/pipe/draw/draw_vertex.h | 2 +- src/mesa/pipe/i915simple/i915_state_derived.c | 25 +++++++++---------------- src/mesa/pipe/softpipe/sp_prim_setup.c | 12 ++++++------ src/mesa/pipe/softpipe/sp_state_derived.c | 20 +++++++++----------- src/mesa/pipe/tgsi/exec/tgsi_attribs.h | 2 ++ src/mesa/state_tracker/st_draw.c | 3 ++- 6 files changed, 29 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 1d900b3a4e..5874c554b1 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -87,7 +87,7 @@ struct vertex_info * \return slot in which the attribute was added */ static INLINE uint -draw_emit_vertex_attr(struct vertex_info *vinfo, uint vfAttr, +draw_emit_vertex_attr(struct vertex_info *vinfo, attrib_format format, interp_mode interp) { const uint n = vinfo->num_attribs; diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 91d00fb1c2..1485d37a62 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -55,7 +55,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) memset(vinfo, 0, sizeof(*vinfo)); /* pos */ - draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_3F, INTERP_LINEAR); + draw_emit_vertex_attr(vinfo, FORMAT_3F, INTERP_LINEAR); /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ for (i = 0; i < fs->num_inputs; i++) { @@ -63,14 +63,12 @@ static void calculate_vertex_layout( struct i915_context *i915 ) case TGSI_SEMANTIC_POSITION: break; case TGSI_SEMANTIC_COLOR0: - front0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, - FORMAT_4UB, colorInterp); + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_COLOR; break; case TGSI_SEMANTIC_COLOR1: assert(0); /* untested */ - front1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, - FORMAT_4UB, colorInterp); + front1 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; break; case TGSI_SEMANTIC_TEX0: @@ -85,8 +83,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) const uint unit = fs->input_semantics[i] - TGSI_SEMANTIC_TEX0; uint hwtc; texCoords[unit] = TRUE; - draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_TEX0 + i, - FORMAT_4F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); hwtc = TEXCOORDFMT_4D; needW = TRUE; vinfo->hwfmt[1] |= hwtc << (unit * 4); @@ -109,16 +106,14 @@ static void calculate_vertex_layout( struct i915_context *i915 ) #if 0 /* color0 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - front0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, - FORMAT_4UB, colorInterp); + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_COLOR; } /* color 1 */ if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { assert(0); /* untested */ - front1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, - FORMAT_4UB, colorInterp); + front1 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; } @@ -130,7 +125,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) for (i = TGSI_ATTRIB_TEX0; i <= TGSI_ATTRIB_TEX7; i++) { uint hwtc; if (inputsRead & (1 << i)) { - draw_emit_vertex_attr(vinfo, i, FORMAT_4F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); hwtc = TEXCOORDFMT_4D; needW = TRUE; } @@ -158,12 +153,10 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ if (i915->rasterizer->light_twoside) { if (front0) { - back0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, - FORMAT_OMIT, colorInterp); + back0 = draw_emit_vertex_attr(vinfo, FORMAT_OMIT, colorInterp); } if (back0) { - back1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, - FORMAT_OMIT, colorInterp); + back1 = draw_emit_vertex_attr(vinfo, FORMAT_OMIT, colorInterp); } } diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 913ae44601..f0f9cf54bd 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -79,7 +79,7 @@ struct setup_stage { float oneoverarea; - struct tgsi_interp_coef coef[TGSI_ATTRIB_MAX]; + struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; struct quad_header quad; struct { @@ -366,7 +366,7 @@ static void const_coeff( struct setup_stage *setup, unsigned slot, unsigned i ) { - assert(slot < TGSI_ATTRIB_MAX); + assert(slot < PIPE_MAX_SHADER_INPUTS); assert(i <= 3); setup->coef[slot].dadx[i] = 0; @@ -391,7 +391,7 @@ static void tri_linear_coeff( struct setup_stage *setup, float a = setup->ebot.dy * majda - botda * setup->emaj.dy; float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - assert(slot < TGSI_ATTRIB_MAX); + assert(slot < PIPE_MAX_SHADER_INPUTS); assert(i <= 3); setup->coef[slot].dadx[i] = a * setup->oneoverarea; @@ -454,7 +454,7 @@ static void tri_persp_coeff( struct setup_stage *setup, ); */ - assert(slot < TGSI_ATTRIB_MAX); + assert(slot < PIPE_MAX_SHADER_INPUTS); assert(i <= 3); setup->coef[slot].dadx[i] = a * setup->oneoverarea; @@ -910,8 +910,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) = sizeAttr ? (0.5f * v0->data[sizeAttr][0]) : (0.5f * setup->softpipe->rasterizer->point_size); const boolean round = setup->softpipe->rasterizer->point_smooth; - const float x = v0->data[TGSI_ATTRIB_POS][0]; - const float y = v0->data[TGSI_ATTRIB_POS][1]; + const float x = v0->data[0][0]; /* Note: data[0] is always position */ + const float y = v0->data[0][1]; unsigned slot, j; assert(sizeAttr >= 0); diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 03b5d7ea3f..3ae207910d 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -63,7 +63,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* always emit vertex pos */ /* TODO - Figure out if we need to do perspective divide, etc. */ - draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POS, FORMAT_4F, INTERP_LINEAR); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR); for (i = 0; i < fs->num_inputs; i++) { switch (fs->input_semantics[i]) { @@ -75,15 +75,15 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) softpipe->need_w = TRUE; break; case TGSI_SEMANTIC_COLOR0: - front0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR0, + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); break; case TGSI_SEMANTIC_COLOR1: - front1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_COLOR1, + front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); break; case TGSI_SEMANTIC_FOG: - draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_FOG, + draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE); break; #if 0 @@ -91,19 +91,19 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* XXX only emit if drawing points or front/back polygon mode * is point mode */ - draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_POINTSIZE, + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_CONSTANT); break; #endif softpipe->psize_slot = i; /*case TGSI_SEMANTIC_TEXCOORD:*/ case TGSI_SEMANTIC_TEX0: - draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_TEX0, + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); softpipe->need_w = TRUE; break; case TGSI_SEMANTIC_OTHER: - draw_emit_vertex_attr(vinfo, i, FORMAT_4F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); softpipe->need_w = TRUE; break; @@ -120,12 +120,10 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ if (softpipe->rasterizer->light_twoside) { if (front0) { - back0 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC0, - FORMAT_OMIT, colorInterp); + back0 = draw_emit_vertex_attr(vinfo, FORMAT_OMIT, colorInterp); } if (back0) { - back1 = draw_emit_vertex_attr(vinfo, TGSI_ATTRIB_BFC1, - FORMAT_OMIT, colorInterp); + back1 = draw_emit_vertex_attr(vinfo, FORMAT_OMIT, colorInterp); } } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_attribs.h b/src/mesa/pipe/tgsi/exec/tgsi_attribs.h index b8f6ba7da8..fa13da12be 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_attribs.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_attribs.h @@ -5,6 +5,7 @@ /** * The specific values here are not important (could remove them). */ +#if 0 enum { TGSI_ATTRIB_POS = 0, TGSI_ATTRIB_WEIGHT = 1, @@ -41,6 +42,7 @@ enum { #define TGSI_MAX_TEXTURE 8 #define TGSI_MAX_VARYING 8 +#endif diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 633e4d9470..e6f4175bf8 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -77,6 +77,7 @@ pipe_vertex_format(GLenum format, GLuint size) } +#if 0 /** * Convert a mesa vertex attribute to a TGSI attribute */ @@ -137,7 +138,7 @@ tgsi_attrib_to_mesa_attrib(GLuint attr) return 0; } } - +#endif /** -- cgit v1.2.3 From c231a9d020bdec8e0749a5547971c79de64f73d8 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 13:58:12 -0600 Subject: remove #includes of tgsi_attribs.h --- src/mesa/pipe/draw/draw_vertex.h | 4 ++-- src/mesa/pipe/softpipe/sp_quad_fs.c | 1 - src/mesa/pipe/softpipe/sp_state_derived.c | 1 - src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 1 - src/mesa/state_tracker/st_cb_feedback.c | 1 - src/mesa/state_tracker/st_cb_rasterpos.c | 1 - src/mesa/state_tracker/st_draw.c | 1 - 7 files changed, 2 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 5874c554b1..5aa0df4e92 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -33,11 +33,11 @@ #ifndef DRAW_VERTEX_H #define DRAW_VERTEX_H -#include "pipe/tgsi/exec/tgsi_attribs.h" +#if 0 #define MAX_VERT_ATTRIBS 12 /* OK? */ - +#endif struct draw_context; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 7d2712a537..13d7eac4f2 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -34,7 +34,6 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" -#include "pipe/tgsi/exec/tgsi_attribs.h" #include "sp_context.h" #include "sp_headers.h" diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 3ae207910d..b51ab66cf3 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -33,7 +33,6 @@ #include "sp_context.h" #include "sp_state.h" -#include "pipe/tgsi/exec/tgsi_attribs.h" #include "pipe/tgsi/exec/tgsi_token.h" diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index fb8365aab5..2d5114a720 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -1,6 +1,5 @@ #include "tgsi_platform.h" #include "tgsi_mesa.h" -#include "pipe/tgsi/exec/tgsi_attribs.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #define TGSI_DEBUG 0 diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index e846463c4c..78cf4c2b4d 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -54,7 +54,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/tgsi/exec/tgsi_attribs.h" #include "vf/vf.h" #include "pipe/draw/draw_context.h" diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 04b2016ffc..56c98916ed 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -44,7 +44,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/tgsi/exec/tgsi_attribs.h" #include "shader/prog_instruction.h" #include "vf/vf.h" diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index e6f4175bf8..0f45bf579a 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -47,7 +47,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/tgsi/exec/tgsi_attribs.h" #include "pipe/draw/draw_private.h" #include "pipe/draw/draw_context.h" -- cgit v1.2.3 From a8834a75a5b469775a983234f9d4bc8925630740 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 13:58:47 -0600 Subject: obsolete --- src/mesa/pipe/tgsi/exec/tgsi_attribs.h | 49 ---------------------------------- 1 file changed, 49 deletions(-) delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_attribs.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_attribs.h b/src/mesa/pipe/tgsi/exec/tgsi_attribs.h deleted file mode 100644 index fa13da12be..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_attribs.h +++ /dev/null @@ -1,49 +0,0 @@ -#ifndef TGSI_ATTRIBS_H -#define TGSI_ATTRIBS_H - - -/** - * The specific values here are not important (could remove them). - */ -#if 0 -enum { - TGSI_ATTRIB_POS = 0, - TGSI_ATTRIB_WEIGHT = 1, - TGSI_ATTRIB_NORMAL = 2, - TGSI_ATTRIB_COLOR0 = 3, - TGSI_ATTRIB_COLOR1 = 4, - TGSI_ATTRIB_FOG = 5, - TGSI_ATTRIB_COLOR_INDEX = 6, /* XXX omit? */ - TGSI_ATTRIB_EDGEFLAG = 7, - TGSI_ATTRIB_TEX0 = 8, - TGSI_ATTRIB_TEX1 = 9, - TGSI_ATTRIB_TEX2 = 10, - TGSI_ATTRIB_TEX3 = 11, - TGSI_ATTRIB_TEX4 = 12, - TGSI_ATTRIB_TEX5 = 13, - TGSI_ATTRIB_TEX6 = 14, - TGSI_ATTRIB_TEX7 = 15, - TGSI_ATTRIB_VAR0 = 16, - TGSI_ATTRIB_VAR1 = 17, - TGSI_ATTRIB_VAR2 = 18, - TGSI_ATTRIB_VAR3 = 19, - TGSI_ATTRIB_VAR4 = 20, - TGSI_ATTRIB_VAR5 = 21, - TGSI_ATTRIB_VAR6 = 22, - TGSI_ATTRIB_VAR7 = 23, - TGSI_ATTRIB_POINTSIZE = 24, - TGSI_ATTRIB_BFC0 = 25, - TGSI_ATTRIB_BFC1 = 26, - TGSI_ATTRIB_CLIP_POS = 27, - TGSI_ATTRIB_VERTEX_HEADER = 28, - TGSI_ATTRIB_MAX = 29 -}; - - -#define TGSI_MAX_TEXTURE 8 -#define TGSI_MAX_VARYING 8 -#endif - - - -#endif /* TGSI_ATTRIBS_H */ -- cgit v1.2.3 From 064daf319cf1868575c794f2380d54ea5b8358af Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 14:03:00 -0600 Subject: remove lots of dead code related to program input/output mapping --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 385 +-------------------------------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 24 -- src/mesa/state_tracker/st_atom_fs.c | 9 - src/mesa/state_tracker/st_atom_vs.c | 9 - 4 files changed, 3 insertions(+), 424 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 2d5114a720..abb0ff755c 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -4,276 +4,6 @@ #define TGSI_DEBUG 0 -#if 0 -/** - * Convert a VERT_ATTRIB_x to a TGSI_ATTRIB_y - */ -uint -tgsi_mesa_translate_vertex_input(GLuint attrib) -{ - /* XXX these could be implemented with array lookups too.... */ - switch (attrib) { - case VERT_ATTRIB_POS: - return TGSI_ATTRIB_POS; - case VERT_ATTRIB_WEIGHT: - return TGSI_ATTRIB_WEIGHT; - case VERT_ATTRIB_NORMAL: - return TGSI_ATTRIB_NORMAL; - case VERT_ATTRIB_COLOR0: - return TGSI_ATTRIB_COLOR0; - case VERT_ATTRIB_COLOR1: - return TGSI_ATTRIB_COLOR1; - case VERT_ATTRIB_FOG: - return TGSI_ATTRIB_FOG; - case VERT_ATTRIB_COLOR_INDEX: - return TGSI_ATTRIB_COLOR_INDEX; - case VERT_ATTRIB_EDGEFLAG: - return TGSI_ATTRIB_EDGEFLAG; - case VERT_ATTRIB_TEX0: - return TGSI_ATTRIB_TEX0; - case VERT_ATTRIB_TEX1: - return TGSI_ATTRIB_TEX1; - case VERT_ATTRIB_TEX2: - return TGSI_ATTRIB_TEX2; - case VERT_ATTRIB_TEX3: - return TGSI_ATTRIB_TEX3; - case VERT_ATTRIB_TEX4: - return TGSI_ATTRIB_TEX4; - case VERT_ATTRIB_TEX5: - return TGSI_ATTRIB_TEX5; - case VERT_ATTRIB_TEX6: - return TGSI_ATTRIB_TEX6; - case VERT_ATTRIB_TEX7: - return TGSI_ATTRIB_TEX7; - case VERT_ATTRIB_GENERIC0: - return TGSI_ATTRIB_VAR0; - case VERT_ATTRIB_GENERIC1: - return TGSI_ATTRIB_VAR1; - case VERT_ATTRIB_GENERIC2: - return TGSI_ATTRIB_VAR2; - case VERT_ATTRIB_GENERIC3: - return TGSI_ATTRIB_VAR3; - case VERT_ATTRIB_GENERIC4: - return TGSI_ATTRIB_VAR4; - case VERT_ATTRIB_GENERIC5: - return TGSI_ATTRIB_VAR5; - case VERT_ATTRIB_GENERIC6: - return TGSI_ATTRIB_VAR6; - case VERT_ATTRIB_GENERIC7: - return TGSI_ATTRIB_VAR7; - default: - assert(0); - return 0; - } -} - - -/** - * Convert VERT_RESULT_x to TGSI_ATTRIB_y - */ -uint -tgsi_mesa_translate_vertex_output(GLuint attrib) -{ - switch (attrib) { - case VERT_RESULT_HPOS: - return TGSI_ATTRIB_POS; - case VERT_RESULT_COL0: - return TGSI_ATTRIB_COLOR0; - case VERT_RESULT_COL1: - return TGSI_ATTRIB_COLOR1; - case VERT_RESULT_FOGC: - return TGSI_ATTRIB_FOG; - case VERT_RESULT_TEX0: - return TGSI_ATTRIB_TEX0; - case VERT_RESULT_TEX1: - return TGSI_ATTRIB_TEX1; - case VERT_RESULT_TEX2: - return TGSI_ATTRIB_TEX2; - case VERT_RESULT_TEX3: - return TGSI_ATTRIB_TEX3; - case VERT_RESULT_TEX4: - return TGSI_ATTRIB_TEX4; - case VERT_RESULT_TEX5: - return TGSI_ATTRIB_TEX5; - case VERT_RESULT_TEX6: - return TGSI_ATTRIB_TEX6; - case VERT_RESULT_TEX7: - return TGSI_ATTRIB_TEX7; - case VERT_RESULT_PSIZ: - return TGSI_ATTRIB_POINTSIZE; - case VERT_RESULT_BFC0: - return TGSI_ATTRIB_BFC0; - case VERT_RESULT_BFC1: - return TGSI_ATTRIB_BFC1; - case VERT_RESULT_VAR0: - return TGSI_ATTRIB_VAR0; - case VERT_RESULT_VAR0 + 1: - return TGSI_ATTRIB_VAR1; - case VERT_RESULT_VAR0 + 2: - return TGSI_ATTRIB_VAR2; - case VERT_RESULT_VAR0 + 3: - return TGSI_ATTRIB_VAR3; - case VERT_RESULT_VAR0 + 4: - return TGSI_ATTRIB_VAR4; - case VERT_RESULT_VAR0 + 5: - return TGSI_ATTRIB_VAR5; - case VERT_RESULT_VAR0 + 6: - return TGSI_ATTRIB_VAR6; - case VERT_RESULT_VAR0 + 7: - return TGSI_ATTRIB_VAR7; - default: - assert(0); - return 0; - } -} - - -/** - * Convert a FRAG_ATTRIB_x to a TGSI_ATTRIB_y - */ -uint -tgsi_mesa_translate_fragment_input(GLuint attrib) -{ - switch (attrib) { - case FRAG_ATTRIB_WPOS: - return TGSI_ATTRIB_POS; - case FRAG_ATTRIB_COL0: - return TGSI_ATTRIB_COLOR0; - case FRAG_ATTRIB_COL1: - return TGSI_ATTRIB_COLOR1; - case FRAG_ATTRIB_FOGC: - return TGSI_ATTRIB_FOG; - case FRAG_ATTRIB_TEX0: - return TGSI_ATTRIB_TEX0; - case FRAG_ATTRIB_TEX1: - return TGSI_ATTRIB_TEX1; - case FRAG_ATTRIB_TEX2: - return TGSI_ATTRIB_TEX2; - case FRAG_ATTRIB_TEX3: - return TGSI_ATTRIB_TEX3; - case FRAG_ATTRIB_TEX4: - return TGSI_ATTRIB_TEX4; - case FRAG_ATTRIB_TEX5: - return TGSI_ATTRIB_TEX5; - case FRAG_ATTRIB_TEX6: - return TGSI_ATTRIB_TEX6; - case FRAG_ATTRIB_TEX7: - return TGSI_ATTRIB_TEX7; - case FRAG_ATTRIB_VAR0: - return TGSI_ATTRIB_VAR0; - case FRAG_ATTRIB_VAR0 + 1: - return TGSI_ATTRIB_VAR1; - case FRAG_ATTRIB_VAR0 + 2: - return TGSI_ATTRIB_VAR2; - case FRAG_ATTRIB_VAR0 + 3: - return TGSI_ATTRIB_VAR3; - case FRAG_ATTRIB_VAR0 + 4: - return TGSI_ATTRIB_VAR4; - case FRAG_ATTRIB_VAR0 + 5: - return TGSI_ATTRIB_VAR5; - case FRAG_ATTRIB_VAR0 + 6: - return TGSI_ATTRIB_VAR6; - case FRAG_ATTRIB_VAR0 + 7: - return TGSI_ATTRIB_VAR7; - default: - assert(0); - return 0; - } -} - - -/** - * Convert FRAG_RESULT_x to TGSI_ATTRIB_y - */ -uint -tgsi_mesa_translate_fragment_output(GLuint attrib) -{ - switch (attrib) { - case FRAG_RESULT_DEPR: - return TGSI_ATTRIB_POS; - case FRAG_RESULT_COLR: - /* fall-through */ - case FRAG_RESULT_COLH: - /* fall-through */ - case FRAG_RESULT_DATA0: - return TGSI_ATTRIB_COLOR0; - case FRAG_RESULT_DATA0 + 1: - return TGSI_ATTRIB_COLOR0 + 1; - case FRAG_RESULT_DATA0 + 2: - return TGSI_ATTRIB_COLOR0 + 2; - case FRAG_RESULT_DATA0 + 3: - return TGSI_ATTRIB_COLOR0 + 3; - default: - assert(0); - return 0; - } -} -#endif - - -#if 0 -uint -tgsi_mesa_translate_vertex_input_mask(GLbitfield mask) -{ - uint tgsiMask = 0x0; - uint i; - for (i = 0; i < VERT_ATTRIB_MAX && mask; i++) { - if (mask & (1 << i)) { - tgsiMask |= 1 << tgsi_mesa_translate_vertex_input(i); - } - mask &= ~(1 << i); - } - return tgsiMask; -} - -uint -tgsi_mesa_translate_vertex_output_mask(GLbitfield mask) -{ - uint tgsiMask = 0x0; - uint i; - for (i = 0; i < VERT_RESULT_MAX && mask; i++) { - if (mask & (1 << i)) { - tgsiMask |= 1 << tgsi_mesa_translate_vertex_output(i); - } - mask &= ~(1 << i); - } - return tgsiMask; -} - -uint -tgsi_mesa_translate_fragment_input_mask(GLbitfield mask) -{ - uint tgsiMask = 0x0; - uint i; - for (i = 0; i < FRAG_ATTRIB_MAX && mask; i++) { - if (mask & (1 << i)) { - tgsiMask |= 1 << tgsi_mesa_translate_fragment_input(i); - } - mask &= ~(1 << i); - } - return tgsiMask; -} - - -uint -tgsi_mesa_translate_fragment_output_mask(GLbitfield mask) -{ - uint tgsiMask = 0x0; - uint i; - for (i = 0; i < FRAG_RESULT_MAX && mask; i++) { - if (mask & (1 << i)) { - tgsiMask |= 1 << tgsi_mesa_translate_fragment_output(i); - } - mask &= ~(1 << i); - } - return tgsiMask; -} - - -#endif - - - /* * Map mesa register file to TGSI register file. @@ -319,16 +49,10 @@ map_register_file_index( GLuint processor, GLuint file, GLuint index, -#if 0 - GLbitfield usage_bitmask, -#endif const GLuint inputMapping[], const GLuint outputMapping[]) { GLuint mapped_index; -#if 0 - GLuint i; -#endif assert(processor == TGSI_PROCESSOR_FRAGMENT || processor == TGSI_PROCESSOR_VERTEX); @@ -343,24 +67,8 @@ map_register_file_index( * If a program uses 3 input attribs, the mapped indexes will be 1,2,3. * etc. */ - assert( index < 32 ); - if (inputMapping) { - printf("New map %d input %d to %d\n", processor, index, - inputMapping[index]); - return inputMapping[index]; - } - assert(0); -#if 0 - assert( usage_bitmask & (1 << index) ); - mapped_index = 0; - for( i = 0; i < index; i++ ) { - if( usage_bitmask & (1 << i) ) { - mapped_index++; - } - } printf("Map %d input %d to %d\n", processor, index, mapped_index); -#endif - break; + return inputMapping[index]; case TGSI_FILE_OUTPUT: /* @@ -381,15 +89,6 @@ map_register_file_index( else { /* vertex output slots are tightly packed, find mapped pos */ /* mapped_index = VERT_RESULT_x */ -#if 0 - mapped_index = 0; - for( i = 0; i < index; i++ ) { - if( usage_bitmask & (1 << i) ) { - mapped_index++; - } - } - assert(outputMapping[index] == mapped_index); -#endif mapped_index = outputMapping[index]; printf("Map VP output from %d to %d\n", index, mapped_index); } @@ -461,10 +160,6 @@ static GLboolean compile_instruction( const struct prog_instruction *inst, struct tgsi_full_instruction *fullinst, -#if 0 - GLuint inputs_read, - GLuint outputs_written, -#endif const GLuint inputMapping[], const GLuint outputMapping[], GLuint preamble_size, @@ -486,14 +181,7 @@ compile_instruction( processor, fulldst->DstRegister.File, inst->DstReg.Index, -#if 0 - outputs_written, -#endif -#if 0 - NULL, -#else inputMapping, -#endif outputMapping ); fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); @@ -507,9 +195,6 @@ compile_instruction( processor, fullsrc->SrcRegister.File, inst->SrcReg[i].Index, -#if 0 - inputs_read, -#endif inputMapping, outputMapping ); @@ -849,7 +534,8 @@ tgsi_mesa_compile_fp_program( struct tgsi_token *tokens, GLuint maxTokens ) { - GLuint i, ti, count; + GLuint i; + GLuint ti; /* token index */ struct tgsi_header *header; struct tgsi_processor *processor; struct tgsi_full_declaration fulldecl; @@ -872,56 +558,6 @@ tgsi_mesa_compile_fp_program( ti = 3; -#if 0 - reads_wpos = program->Base.InputsRead & (1 << FRAG_ATTRIB_WPOS); - inputs_read = program->Base.InputsRead | (1 << FRAG_ATTRIB_WPOS); - - /* - * Declare input attributes. Note that we do not interpolate fragment position. - */ - reads_wpos = 1; - /* Fragment position. */ - if( reads_wpos ) { - fulldecl = make_frag_input_decl( - 0, - TGSI_INTERPOLATE_CONSTANT, - TGSI_WRITEMASK_XY, - TGSI_SEMANTIC_POSITION ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - - /* Fragment zw. */ - fulldecl = make_frag_input_decl( - 0, - TGSI_INTERPOLATE_LINEAR, - reads_wpos ? TGSI_WRITEMASK_ZW : TGSI_WRITEMASK_Z, - TGSI_SEMANTIC_POSITION ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - - count = 0; - for( i = 1; i < 32; i++ ) { - if( inputs_read & (1 << i) ) { - count++; - fulldecl = make_frag_input_decl(count, - interpMode[i], - TGSI_WRITEMASK_XYZW, - inputSemantic[count] ); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } -#else - for (i = 0; i < numInputs; i++) { switch (inputSemantic[i]) { case TGSI_SEMANTIC_POSITION: @@ -958,7 +594,6 @@ tgsi_mesa_compile_fp_program( break; } } -#endif /* @@ -1023,10 +658,6 @@ tgsi_mesa_compile_fp_program( if( compile_instruction( &program->Base.Instructions[i], &fullinst, -#if 0 - inputs_read, - ~0, /*outputs_written*/ -#endif inputMapping, outputMapping, preamble_size, @@ -1061,12 +692,6 @@ tgsi_mesa_compile_vp_program( struct tgsi_header *header; struct tgsi_processor *processor; struct tgsi_full_instruction fullinst; -#if 0 - GLuint inputs_read = ~0; - GLuint outputs_written; - - outputs_written = program->Base.OutputsWritten; -#endif *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -1082,10 +707,6 @@ tgsi_mesa_compile_vp_program( if( compile_instruction( &program->Base.Instructions[i], &fullinst, -#if 0 - inputs_read, - outputs_written, -#endif inputMapping, outputMapping, 0, diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index 8105e9e738..d61d80bab0 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -26,30 +26,6 @@ tgsi_mesa_compile_vp_program( struct tgsi_token *tokens, GLuint maxTokens ); -uint -tgsi_mesa_translate_vertex_input(GLuint attrib); - -uint -tgsi_mesa_translate_vertex_output(GLuint attrib); - -uint -tgsi_mesa_translate_fragment_input(GLuint attrib); - -uint -tgsi_mesa_translate_fragment_output(GLuint attrib); - -uint -tgsi_mesa_translate_vertex_input_mask(GLbitfield mask); - -uint -tgsi_mesa_translate_vertex_output_mask(GLbitfield mask); - -uint -tgsi_mesa_translate_fragment_input_mask(GLbitfield mask); - -uint -tgsi_mesa_translate_fragment_output_mask(GLbitfield mask); - #if defined __cplusplus } // extern "C" diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 340c7ec69f..4d65035143 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -131,15 +131,6 @@ st_translate_fragment_shader(struct st_context *st, outputMapping, stfp->tokens, ST_FP_MAX_TOKENS ); -#if 0 - fs.inputs_read - = tgsi_mesa_translate_fragment_input_mask(stfp->Base.Base.InputsRead); -#endif -#if 0 - fs.outputs_written - = tgsi_mesa_translate_fragment_output_mask(stfp->Base.Base.OutputsWritten); -#endif - fs.tokens = &stfp->tokens[0]; cso = st_cached_fs_state(st, &fs); diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 2b8aef5c63..fcf99bf1c9 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -131,15 +131,6 @@ st_translate_vertex_shader(struct st_context *st, st->vertex_attrib_to_slot, stvp->tokens, ST_FP_MAX_TOKENS ); -#if 0 - vs.inputs_read - = tgsi_mesa_translate_vertex_input_mask(stvp->Base.Base.InputsRead); -#endif -#if 0 - vs.outputs_written - = tgsi_mesa_translate_vertex_output_mask(stvp->Base.Base.OutputsWritten); -#endif - vs.tokens = &stvp->tokens[0]; cso = st_cached_vs_state(st, &vs); -- cgit v1.2.3 From 478d1e2c9c73fc29542375c44d01ab964ce8eccf Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 14:08:09 -0600 Subject: remove old/used code --- src/mesa/pipe/draw/draw_vertex.c | 10 ++++---- src/mesa/pipe/draw/draw_vertex.h | 9 ------- src/mesa/pipe/draw/draw_vertex_shader.c | 4 ---- src/mesa/pipe/i915simple/i915_state_derived.c | 34 --------------------------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 9 +++---- 5 files changed, 8 insertions(+), 58 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index ab4e6c7864..6f81e6e8a3 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -45,7 +45,7 @@ static INLINE void -emit_vertex_attr(struct vertex_info *vinfo, /*uint vfAttr,*/ +emit_vertex_attr(struct vertex_info *vinfo, attrib_format format, interp_mode interp) { const uint n = vinfo->num_attribs; @@ -112,15 +112,15 @@ draw_set_vertex_attributes( struct draw_context *draw, /* * First three attribs are always the same: header, clip pos, winpos */ - emit_vertex_attr(vinfo, /*TGSI_ATTRIB_VERTEX_HEADER,*/ FORMAT_1F, INTERP_NONE); - emit_vertex_attr(vinfo, /*TGSI_ATTRIB_CLIP_POS,*/ FORMAT_4F, INTERP_LINEAR); - emit_vertex_attr(vinfo, /*TGSI_ATTRIB_POS,*/ FORMAT_4F_VIEWPORT, INTERP_LINEAR); + emit_vertex_attr(vinfo, FORMAT_1F, INTERP_NONE); + emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR); + emit_vertex_attr(vinfo, FORMAT_4F_VIEWPORT, INTERP_LINEAR); /* * Remaining attribs (color, texcoords, etc) */ for (i = 1; i < nr_attrs; i++) { - emit_vertex_attr(vinfo, /*slot_to_vf_attr[i],*/ FORMAT_4F, interps[i]); + emit_vertex_attr(vinfo, FORMAT_4F, interps[i]); } draw_compute_vertex_size(vinfo); diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 5aa0df4e92..2968981708 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -34,15 +34,9 @@ #define DRAW_VERTEX_H - -#if 0 -#define MAX_VERT_ATTRIBS 12 /* OK? */ -#endif - struct draw_context; - /** * Vertex attribute format */ @@ -92,9 +86,6 @@ draw_emit_vertex_attr(struct vertex_info *vinfo, { const uint n = vinfo->num_attribs; assert(n < PIPE_MAX_SHADER_OUTPUTS); - /* - vinfo->attr_mask |= (1 << vfAttr); - */ vinfo->format[n] = format; vinfo->interp_mode[n] = interp; vinfo->num_attribs++; diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 2d424632e7..6461e3e3ba 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -94,11 +94,7 @@ run_vertex_program(struct draw_context *draw, const float *trans = draw->viewport.translate; assert(count <= 4); -#if 0 - assert(draw->vertex_shader.outputs_written & (1 << TGSI_ATTRIB_POS)); -#else assert(draw->vertex_shader.output_semantics[0] == TGSI_SEMANTIC_POSITION); -#endif #ifdef DEBUG memset( &machine, 0, sizeof( machine ) ); diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 1485d37a62..2a03650620 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -103,40 +103,6 @@ static void calculate_vertex_layout( struct i915_context *i915 ) } } -#if 0 - /* color0 */ - if (inputsRead & (1 << TGSI_ATTRIB_COLOR0)) { - front0 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); - vinfo->hwfmt[0] |= S4_VFMT_COLOR; - } - - /* color 1 */ - if (inputsRead & (1 << TGSI_ATTRIB_COLOR1)) { - assert(0); /* untested */ - front1 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); - vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; - } - - /* XXX fog? */ - - /* texcoords */ - { - uint i; - for (i = TGSI_ATTRIB_TEX0; i <= TGSI_ATTRIB_TEX7; i++) { - uint hwtc; - if (inputsRead & (1 << i)) { - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); - hwtc = TEXCOORDFMT_4D; - needW = TRUE; - } - else { - hwtc = TEXCOORDFMT_NOT_PRESENT; - } - vinfo->hwfmt[1] |= hwtc << ((i - TGSI_ATTRIB_TEX0) * 4); - } - } -#endif - /* go back and fill in the vertex position info now that we have needW */ if (needW) { vinfo->hwfmt[0] |= S4_VFMT_XYZW; diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index abb0ff755c..115c5d1eb0 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -540,12 +540,6 @@ tgsi_mesa_compile_fp_program( struct tgsi_processor *processor; struct tgsi_full_declaration fulldecl; struct tgsi_full_instruction fullinst; - /* - struct tgsi_full_dst_register *fulldst; - struct tgsi_full_src_register *fullsrc; - GLuint inputs_read; - GLboolean reads_wpos; - */ GLuint preamble_size = 0; *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); @@ -703,6 +697,9 @@ tgsi_mesa_compile_vp_program( ti = 3; + /* XXX todo: input/output declarations + */ + for( i = 0; i < program->Base.NumInstructions; i++ ) { if( compile_instruction( &program->Base.Instructions[i], -- cgit v1.2.3 From f69b5c56feb60791bad27d491ee9592238d4efb0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 14:47:22 -0600 Subject: Clean-up the TGSI_SEMANTIC tokens, introduce semantic indexes. Still need to produce decl instructions for vertex shaders... --- src/mesa/pipe/draw/draw_vertex_shader.c | 3 +- src/mesa/pipe/i915simple/i915_fpc.h | 3 +- src/mesa/pipe/i915simple/i915_fpc_translate.c | 48 +++++++-------- src/mesa/pipe/i915simple/i915_state_derived.c | 32 +++++----- src/mesa/pipe/p_state.h | 8 ++- src/mesa/pipe/softpipe/sp_state_derived.c | 24 ++++---- src/mesa/pipe/tgsi/exec/tgsi_build.c | 2 +- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 22 +++---- src/mesa/pipe/tgsi/exec/tgsi_token.h | 24 +++----- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 23 ++++--- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 3 +- src/mesa/state_tracker/st_atom_fs.c | 35 ++++++++--- src/mesa/state_tracker/st_atom_vs.c | 86 ++++++++++++++++++++++----- 13 files changed, 191 insertions(+), 122 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 6461e3e3ba..ef0399c46e 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -94,7 +94,8 @@ run_vertex_program(struct draw_context *draw, const float *trans = draw->viewport.translate; assert(count <= 4); - assert(draw->vertex_shader.output_semantics[0] == TGSI_SEMANTIC_POSITION); + assert(draw->vertex_shader.output_semantic_name[0] + == TGSI_SEMANTIC_POSITION); #ifdef DEBUG memset( &machine, 0, sizeof( machine ) ); diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index 5c4f2f90e9..5fdc7ef27e 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -51,7 +51,8 @@ struct i915_fp_compile { uint declarations[I915_PROGRAM_SIZE]; uint program[I915_PROGRAM_SIZE]; - uint input_semantic[PIPE_MAX_SHADER_INPUTS]; + uint input_semantic_name[PIPE_MAX_SHADER_INPUTS]; + uint input_semantic_index[PIPE_MAX_SHADER_INPUTS]; /** points into the i915->current.constants array: */ float (*constants)[4]; diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index df8859bec8..6f78c72e04 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -128,7 +128,7 @@ src_vector(struct i915_fp_compile *p, const struct tgsi_full_src_register *source) { uint index = source->SrcRegister.Index; - uint src, sem; + uint src, sem_name, sem_ind; switch (source->SrcRegister.File) { case TGSI_FILE_TEMPORARY: @@ -152,10 +152,11 @@ src_vector(struct i915_fp_compile *p, /* use vertex format info to map a slot number to a VF attrib */ assert(index < p->vertex_info->num_attribs); - sem = p->input_semantic[index]; + sem_name = p->input_semantic_name[index]; + sem_ind = p->input_semantic_index[index]; #if 1 - switch (sem) { + switch (sem_name) { case TGSI_SEMANTIC_POSITION: printf("SKIP SEM POS\n"); /* @@ -163,28 +164,23 @@ src_vector(struct i915_fp_compile *p, src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); */ break; - case TGSI_SEMANTIC_COLOR0: - src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); - break; - case TGSI_SEMANTIC_COLOR1: - src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); - src = swizzle(src, X, Y, Z, ONE); + case TGSI_SEMANTIC_COLOR: + if (sem_ind == 0) { + src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); + } + else { + /* secondary color */ + assert(sem_ind == 1); + src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); + src = swizzle(src, X, Y, Z, ONE); + } break; case TGSI_SEMANTIC_FOG: src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); src = swizzle(src, W, W, W, W); break; - case TGSI_SEMANTIC_TEX0: - case TGSI_SEMANTIC_TEX1: - case TGSI_SEMANTIC_TEX2: - case TGSI_SEMANTIC_TEX3: - case TGSI_SEMANTIC_TEX4: - case TGSI_SEMANTIC_TEX5: - case TGSI_SEMANTIC_TEX6: - case TGSI_SEMANTIC_TEX7: - src = i915_emit_decl(p, REG_TYPE_T, - T_TEX0 + (sem - TGSI_SEMANTIC_TEX0), - D0_CHANNEL_ALL); + case TGSI_SEMANTIC_TEXCOORD: + src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + sem_ind, D0_CHANNEL_ALL); break; default: i915_program_error(p, "Bad source->Index"); @@ -895,11 +891,13 @@ i915_translate_instructions(struct i915_fp_compile *p, if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) { /* save input register info for use in src_vector() */ - uint ind, sem; + uint ind, sem, semi; ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; + semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; /*printf("FS Input DECL [%u] sem %u\n", ind, sem);*/ - p->input_semantic[ind] = sem; + p->input_semantic_name[ind] = sem; + p->input_semantic_index[ind] = semi; } break; @@ -1059,7 +1057,7 @@ i915_find_wpos_space(struct i915_fp_compile *p) i915_program_error(p, "No free texcoord for wpos value"); } #else - if (p->shader->input_semantics[0] == TGSI_SEMANTIC_POSITION) { + if (p->shader->input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { /* frag shader using the fragment position input */ #if 0 assert(0); @@ -1079,8 +1077,8 @@ i915_find_wpos_space(struct i915_fp_compile *p) static void i915_fixup_depth_write(struct i915_fp_compile *p) { - /* XXX assuming depth is always in output[0] */ - if (p->shader->output_semantics[0] == TGSI_SEMANTIC_DEPTH) { + /* XXX assuming pos/depth is always in output[0] */ + if (p->shader->output_semantic_name[0] == TGSI_SEMANTIC_POSITION) { const uint depth = UREG(REG_TYPE_OD, 0); i915_emit_arith(p, diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 2a03650620..e825971199 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -59,28 +59,24 @@ static void calculate_vertex_layout( struct i915_context *i915 ) /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ for (i = 0; i < fs->num_inputs; i++) { - switch (fs->input_semantics[i]) { + switch (fs->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: break; - case TGSI_SEMANTIC_COLOR0: - front0 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); - vinfo->hwfmt[0] |= S4_VFMT_COLOR; - break; - case TGSI_SEMANTIC_COLOR1: - assert(0); /* untested */ - front1 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); - vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; + case TGSI_SEMANTIC_COLOR: + if (fs->input_semantic_index[i] == 0) { + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); + vinfo->hwfmt[0] |= S4_VFMT_COLOR; + } + else { + assert(fs->input_semantic_index[i] == 1); + assert(0); /* untested */ + front1 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); + vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; + } break; - case TGSI_SEMANTIC_TEX0: - case TGSI_SEMANTIC_TEX1: - case TGSI_SEMANTIC_TEX2: - case TGSI_SEMANTIC_TEX3: - case TGSI_SEMANTIC_TEX4: - case TGSI_SEMANTIC_TEX5: - case TGSI_SEMANTIC_TEX6: - case TGSI_SEMANTIC_TEX7: + case TGSI_SEMANTIC_TEXCOORD: { - const uint unit = fs->input_semantics[i] - TGSI_SEMANTIC_TEX0; + const uint unit = fs->input_semantic_index[i]; uint hwtc; texCoords[unit] = TRUE; draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 6396d49b84..d67736c558 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -148,8 +148,12 @@ struct pipe_shader_state { /** These fields somewhat constitute the shader "signature" */ ubyte num_inputs; ubyte num_outputs; - ubyte input_semantics[PIPE_MAX_SHADER_INPUTS]; - ubyte output_semantics[PIPE_MAX_SHADER_OUTPUTS]; + + ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ + ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + + ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ + ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; }; struct pipe_depth_stencil_state diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index b51ab66cf3..66e5cc48f6 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -65,7 +65,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR); for (i = 0; i < fs->num_inputs; i++) { - switch (fs->input_semantics[i]) { + switch (fs->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: /* Need Z if depth test is enabled or the fragment program uses the * fragment position (XYZW). @@ -73,13 +73,16 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) softpipe->need_z = TRUE; softpipe->need_w = TRUE; break; - case TGSI_SEMANTIC_COLOR0: - front0 = draw_emit_vertex_attr(vinfo, - FORMAT_4F, colorInterp); - break; - case TGSI_SEMANTIC_COLOR1: - front1 = draw_emit_vertex_attr(vinfo, - FORMAT_4F, colorInterp); + case TGSI_SEMANTIC_COLOR: + if (fs->input_semantic_index[i] == 0) { + front0 = draw_emit_vertex_attr(vinfo, + FORMAT_4F, colorInterp); + } + else { + assert(fs->input_semantic_index[i] == 1); + front1 = draw_emit_vertex_attr(vinfo, + FORMAT_4F, colorInterp); + } break; case TGSI_SEMANTIC_FOG: draw_emit_vertex_attr(vinfo, @@ -96,12 +99,13 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) #endif softpipe->psize_slot = i; /*case TGSI_SEMANTIC_TEXCOORD:*/ - case TGSI_SEMANTIC_TEX0: + case TGSI_SEMANTIC_TEXCOORD: + /* unit = fs->input_semantic_index[i] */ draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); softpipe->need_w = TRUE; break; - case TGSI_SEMANTIC_OTHER: + case TGSI_SEMANTIC_GENERIC: draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); softpipe->need_w = TRUE; break; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.c b/src/mesa/pipe/tgsi/exec/tgsi_build.c index 1320872c64..78f648aae2 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_build.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_build.c @@ -309,7 +309,7 @@ tgsi_default_declaration_semantic( void ) { struct tgsi_declaration_semantic ds; - ds.SemanticName = TGSI_SEMANTIC_DEPTH; + ds.SemanticName = TGSI_SEMANTIC_POSITION; ds.SemanticIndex = 0; ds.Padding = 0; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index 0a47ad2a8c..315354198a 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -201,28 +201,22 @@ static const char *TGSI_INTERPOLATES_SHORT[] = static const char *TGSI_SEMANTICS[] = { - "SEMANTIC_DEPTH", - "SEMANTIC_COLOR0", - "SEMANTIC_COLOR1", - "SEMANTIC_COLOR0B", - "SEMANTIC_COLOR1B", "SEMANTIC_POSITION", + "SEMANTIC_COLOR", + "SEMANTIC_BCOLOR", "SEMANTIC_FOG", - "SEMANTIC_OTHER," - "SEMANTIC_TEX0", + "SEMANTIC_TEXCOORD", + "SEMANTIC_GENERIC," }; static const char *TGSI_SEMANTICS_SHORT[] = { - "DEPTH", - "COLOR0", - "COLOR1", - "COLOR0B", - "COLOR1B", "POSITION", + "COLOR", + "BCOLOR", "FOG", - "OTHER", - "TEX0" + "TEXCOORD", + "GENERIC", }; static const char *TGSI_IMMS[] = diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h index a642ba131a..2d468b801e 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -103,23 +103,13 @@ struct tgsi_declaration_interpolation unsigned Padding : 28; }; -#define TGSI_SEMANTIC_DEPTH 0 -#define TGSI_SEMANTIC_COLOR0 1 -#define TGSI_SEMANTIC_COLOR1 2 -#define TGSI_SEMANTIC_COLOR0B 3 /**< back-face primary color */ -#define TGSI_SEMANTIC_COLOR1B 4 /**< back-face secondary color */ -#define TGSI_SEMANTIC_POSITION 5 -#define TGSI_SEMANTIC_FOG 6 -#define TGSI_SEMANTIC_OTHER 7 /* XXX temp */ -#define TGSI_SEMANTIC_TEX0 8 -#define TGSI_SEMANTIC_TEX1 9 -#define TGSI_SEMANTIC_TEX2 10 -#define TGSI_SEMANTIC_TEX3 11 -#define TGSI_SEMANTIC_TEX4 12 -#define TGSI_SEMANTIC_TEX5 13 -#define TGSI_SEMANTIC_TEX6 14 -#define TGSI_SEMANTIC_TEX7 15 -#define TGSI_SEMANTIC_COUNT 16 /**< number of semantic values */ +#define TGSI_SEMANTIC_POSITION 0 +#define TGSI_SEMANTIC_COLOR 1 +#define TGSI_SEMANTIC_BCOLOR 2 /**< back-face color */ +#define TGSI_SEMANTIC_FOG 3 +#define TGSI_SEMANTIC_TEXCOORD 4 +#define TGSI_SEMANTIC_GENERIC 5 +#define TGSI_SEMANTIC_COUNT 6 /**< number of semantic values */ struct tgsi_declaration_semantic { diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 115c5d1eb0..c7b62c9b84 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -473,7 +473,8 @@ make_frag_input_decl( GLuint index, GLuint interpolate, GLuint usage_mask, - GLuint semantic_name ) + GLuint semantic_name, + GLuint semantic_index ) { struct tgsi_full_declaration decl; @@ -486,6 +487,7 @@ make_frag_input_decl( decl.u.DeclarationRange.First = index; decl.u.DeclarationRange.Last = index; decl.Semantic.SemanticName = semantic_name; + decl.Semantic.SemanticIndex = semantic_index; decl.Interpolation.Interpolate = interpolate; return decl; @@ -495,6 +497,7 @@ static struct tgsi_full_declaration make_frag_output_decl( GLuint index, GLuint semantic_name, + GLuint semantic_index, GLuint usage_mask ) { struct tgsi_full_declaration decl; @@ -507,7 +510,7 @@ make_frag_output_decl( decl.u.DeclarationRange.First = index; decl.u.DeclarationRange.Last = index; decl.Semantic.SemanticName = semantic_name; - decl.Semantic.SemanticIndex = 0; + decl.Semantic.SemanticIndex = semantic_index; return decl; } @@ -528,7 +531,8 @@ tgsi_mesa_compile_fp_program( const struct gl_fragment_program *program, GLuint numInputs, const GLuint inputMapping[], - const ubyte inputSemantic[], + const ubyte inputSemanticName[], + const ubyte inputSemanticIndex[], const GLuint interpMode[], const GLuint outputMapping[], struct tgsi_token *tokens, @@ -553,13 +557,13 @@ tgsi_mesa_compile_fp_program( ti = 3; for (i = 0; i < numInputs; i++) { - switch (inputSemantic[i]) { + switch (inputSemanticName[i]) { case TGSI_SEMANTIC_POSITION: /* Fragment XY pos */ fulldecl = make_frag_input_decl(i, TGSI_INTERPOLATE_CONSTANT, TGSI_WRITEMASK_XY, - TGSI_SEMANTIC_POSITION ); + TGSI_SEMANTIC_POSITION, 0 ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -569,7 +573,7 @@ tgsi_mesa_compile_fp_program( fulldecl = make_frag_input_decl(i, TGSI_INTERPOLATE_LINEAR, TGSI_WRITEMASK_ZW, - TGSI_SEMANTIC_POSITION ); + TGSI_SEMANTIC_POSITION, 0 ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -580,7 +584,8 @@ tgsi_mesa_compile_fp_program( fulldecl = make_frag_input_decl(i, interpMode[i], TGSI_WRITEMASK_XYZW, - inputSemantic[i] ); + inputSemanticName[i], + inputSemanticIndex[i]); ti += tgsi_build_full_declaration(&fulldecl, &tokens[ti], header, @@ -599,7 +604,7 @@ tgsi_mesa_compile_fp_program( fulldecl = make_frag_output_decl( 0, - TGSI_SEMANTIC_DEPTH, + TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ TGSI_WRITEMASK_Z ); ti += tgsi_build_full_declaration( &fulldecl, @@ -610,7 +615,7 @@ tgsi_mesa_compile_fp_program( if( program->Base.OutputsWritten & (1 << FRAG_RESULT_COLR) ) { fulldecl = make_frag_output_decl( 1, - TGSI_SEMANTIC_COLOR0, + TGSI_SEMANTIC_COLOR, 0, TGSI_WRITEMASK_XYZW ); ti += tgsi_build_full_declaration( &fulldecl, diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index d61d80bab0..b40331981e 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -12,7 +12,8 @@ tgsi_mesa_compile_fp_program( const struct gl_fragment_program *program, GLuint numInputs, const GLuint inputMapping[], - const ubyte inputSemantic[], + const ubyte inputSemanticName[], + const ubyte inputSemanticIndex[], const GLuint interpMode[], const GLuint outputMapping[], struct tgsi_token *tokens, diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 4d65035143..8ee4f407c7 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -77,23 +77,41 @@ st_translate_fragment_shader(struct st_context *st, switch (i) { case FRAG_ATTRIB_WPOS: - fs.input_semantics[fs.num_inputs] = TGSI_SEMANTIC_POSITION; + fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_POSITION; + fs.input_semantic_index[fs.num_inputs] = 0; interpMode[fs.num_inputs] = TGSI_INTERPOLATE_CONSTANT; break; case FRAG_ATTRIB_COL0: - fs.input_semantics[fs.num_inputs] = TGSI_SEMANTIC_COLOR0; + fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_COLOR; + fs.input_semantic_index[fs.num_inputs] = 0; interpMode[fs.num_inputs] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_COL1: - fs.input_semantics[fs.num_inputs] = TGSI_SEMANTIC_COLOR1; + fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_COLOR; + fs.input_semantic_index[fs.num_inputs] = 1; interpMode[fs.num_inputs] = TGSI_INTERPOLATE_LINEAR; break; + case FRAG_ATTRIB_FOGC: + assert(0); + break; case FRAG_ATTRIB_TEX0: - fs.input_semantics[fs.num_inputs] = TGSI_SEMANTIC_TEX0; + case FRAG_ATTRIB_TEX1: + case FRAG_ATTRIB_TEX2: + case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX4: + case FRAG_ATTRIB_TEX5: + case FRAG_ATTRIB_TEX6: + case FRAG_ATTRIB_TEX7: + fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_TEXCOORD; + fs.input_semantic_index[fs.num_inputs] = i - FRAG_ATTRIB_TEX0; interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; break; + case FRAG_ATTRIB_VAR0: + /* fall-through */ default: - assert(0); + fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_GENERIC; + fs.input_semantic_index[fs.num_inputs] = i - FRAG_ATTRIB_VAR0; + interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; } fs.num_inputs++; @@ -107,11 +125,11 @@ st_translate_fragment_shader(struct st_context *st, if (stfp->Base.Base.OutputsWritten & (1 << i)) { switch (i) { case FRAG_RESULT_DEPR: - fs.output_semantics[fs.num_outputs] = TGSI_SEMANTIC_DEPTH; + fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_POSITION; outputMapping[i] = fs.num_outputs; break; case FRAG_RESULT_COLR: - fs.output_semantics[fs.num_outputs] = TGSI_SEMANTIC_COLOR0; + fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_COLOR; outputMapping[i] = fs.num_outputs; break; default: @@ -126,7 +144,8 @@ st_translate_fragment_shader(struct st_context *st, tgsi_mesa_compile_fp_program( &stfp->Base, fs.num_inputs, inputMapping, - fs.input_semantics, + fs.input_semantic_name, + fs.input_semantic_index, interpMode, outputMapping, stfp->tokens, ST_FP_MAX_TOKENS ); diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index fcf99bf1c9..b09f2bece7 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -64,9 +64,8 @@ st_translate_vertex_shader(struct st_context *st, memset(&vs, 0, sizeof(vs)); /* - * Determine how many inputs there are. - * Also, compute two look-up tables that map between Mesa VERT_ATTRIB_x - * values and TGSI generic input indexes. + * Determine number of inputs, the mappings between VERT_ATTRIB_x + * and TGSI generic input indexes, plus input attrib semantic info. */ for (i = 0; i < MAX_VERTEX_PROGRAM_ATTRIBS; i++) { if (stvp->Base.Base.InputsRead & (1 << i)) { @@ -74,26 +73,57 @@ st_translate_vertex_shader(struct st_context *st, stvp->index_to_input[vs.num_inputs] = i; switch (i) { case VERT_ATTRIB_POS: - vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_POSITION; + vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_POSITION; + vs.input_semantic_index[vs.num_inputs] = 0; + break; + case VERT_ATTRIB_WEIGHT: + /* fall-through */ + case VERT_ATTRIB_NORMAL: + /* just label as a generic */ + vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; + vs.input_semantic_index[vs.num_inputs] = 0; break; case VERT_ATTRIB_COLOR0: - vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0; + vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; + vs.input_semantic_index[vs.num_inputs] = 0; break; case VERT_ATTRIB_COLOR1: - vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1; + vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; + vs.input_semantic_index[vs.num_inputs] = 1; break; case VERT_ATTRIB_TEX0: - vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_TEX0; + case VERT_ATTRIB_TEX1: + case VERT_ATTRIB_TEX2: + case VERT_ATTRIB_TEX3: + case VERT_ATTRIB_TEX4: + case VERT_ATTRIB_TEX5: + case VERT_ATTRIB_TEX6: + case VERT_ATTRIB_TEX7: + vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_TEXCOORD; + vs.input_semantic_index[vs.num_inputs] = 1 - VERT_ATTRIB_TEX0; + break; + case VERT_ATTRIB_GENERIC0: + case VERT_ATTRIB_GENERIC1: + case VERT_ATTRIB_GENERIC2: + case VERT_ATTRIB_GENERIC3: + case VERT_ATTRIB_GENERIC4: + case VERT_ATTRIB_GENERIC5: + case VERT_ATTRIB_GENERIC6: + case VERT_ATTRIB_GENERIC7: + assert(i < VERT_ATTRIB_MAX); + vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; + vs.input_semantic_index[vs.num_inputs] = 1 - VERT_ATTRIB_GENERIC0; break; default: - vs.input_semantics[vs.num_inputs] = TGSI_SEMANTIC_OTHER; + assert(0); } vs.num_inputs++; } } /* - * Determine number of outputs and the register mapping. + * Determine number of outputs, the register mapping and + * the semantic information for each vertex output/result. */ for (i = 0; i < VERT_RESULT_MAX; i++) { if (stvp->Base.Base.OutputsWritten & (1 << i)) { @@ -102,23 +132,49 @@ st_translate_vertex_shader(struct st_context *st, switch (i) { case VERT_RESULT_HPOS: - vs.output_semantics[vs.num_outputs] = TGSI_SEMANTIC_POSITION; + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_POSITION; + vs.output_semantic_index[vs.num_inputs] = 0; break; case VERT_RESULT_COL0: - vs.output_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0; + vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; + vs.output_semantic_index[vs.num_inputs] = 0; break; case VERT_RESULT_COL1: - vs.output_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1; + vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; + vs.output_semantic_index[vs.num_inputs] = 1; break; case VERT_RESULT_BFC0: - vs.output_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR0B; + vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_BCOLOR; + vs.output_semantic_index[vs.num_inputs] = 0; break; case VERT_RESULT_BFC1: - vs.output_semantics[vs.num_inputs] = TGSI_SEMANTIC_COLOR1B; + vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_BCOLOR; + vs.output_semantic_index[vs.num_inputs] = 1; + break; + case VERT_RESULT_FOGC: + case VERT_RESULT_PSIZ: + case VERT_RESULT_EDGE: + assert(0); break; + case VERT_RESULT_TEX0: + case VERT_RESULT_TEX1: + case VERT_RESULT_TEX2: + case VERT_RESULT_TEX3: + case VERT_RESULT_TEX4: + case VERT_RESULT_TEX5: + case VERT_RESULT_TEX6: + case VERT_RESULT_TEX7: + vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_TEXCOORD; + vs.output_semantic_index[vs.num_inputs] = i - VERT_RESULT_TEX0; + break; + case VERT_RESULT_VAR0: + /* fall-through */ default: - vs.output_semantics[vs.num_outputs] = TGSI_SEMANTIC_OTHER; + assert(i - VERT_RESULT_VAR0 < MAX_VARYING); + vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; + vs.output_semantic_index[vs.num_inputs] = i - VERT_RESULT_VAR0; } + vs.num_outputs++; } } -- cgit v1.2.3 From 1bf93fccdfcef80619852e36db4f2a516a43ae49 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 15:32:18 -0600 Subject: remove old code in src_vector() --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 37 --------------------------- 1 file changed, 37 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 6f78c72e04..f8266f6adb 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -155,7 +155,6 @@ src_vector(struct i915_fp_compile *p, sem_name = p->input_semantic_name[index]; sem_ind = p->input_semantic_index[index]; -#if 1 switch (sem_name) { case TGSI_SEMANTIC_POSITION: printf("SKIP SEM POS\n"); @@ -186,42 +185,6 @@ src_vector(struct i915_fp_compile *p, i915_program_error(p, "Bad source->Index"); return 0; } -#else - index = p->vertex_info->slot_to_attrib[index]; - - switch (index) { - case TGSI_ATTRIB_POS: - assert(p->wpos_tex != -1); - src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); - break; - case TGSI_ATTRIB_COLOR0: - src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); - break; - case TGSI_ATTRIB_COLOR1: - src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); - src = swizzle(src, X, Y, Z, ONE); - break; - case TGSI_ATTRIB_FOG: - src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); - src = swizzle(src, W, W, W, W); - break; - case TGSI_ATTRIB_TEX0: - case TGSI_ATTRIB_TEX1: - case TGSI_ATTRIB_TEX2: - case TGSI_ATTRIB_TEX3: - case TGSI_ATTRIB_TEX4: - case TGSI_ATTRIB_TEX5: - case TGSI_ATTRIB_TEX6: - case TGSI_ATTRIB_TEX7: - src = i915_emit_decl(p, REG_TYPE_T, - T_TEX0 + (index - TGSI_ATTRIB_TEX0), - D0_CHANNEL_ALL); - break; - default: - i915_program_error(p, "Bad source->Index"); - return 0; - } -#endif break; case TGSI_FILE_CONSTANT: -- cgit v1.2.3 From 154335d53f9eb3a2ae46b46858ffcaa6e144e821 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Sep 2007 19:06:26 -0600 Subject: remove :5 form format field in pipe_surface, makes no difference at this time --- src/mesa/pipe/p_state.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index d67736c558..b4fc01bfc5 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -278,7 +278,7 @@ struct pipe_region struct pipe_surface { struct pipe_region *region; - unsigned format:5; /**< PIPE_FORMAT_x */ + unsigned format; /**< PIPE_FORMAT_x */ unsigned width, height; unsigned offset; /**< offset from start of region, in bytes */ unsigned refcount; -- cgit v1.2.3 From b0fa489eba9170c4b102bf0feb1b1c3f02a34e4e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 21 Sep 2007 04:56:39 -0400 Subject: Silence a few warnings. --- src/mesa/pipe/i915simple/i915_context.c | 2 +- src/mesa/pipe/i915simple/i915_state_sampler.c | 2 +- src/mesa/pipe/softpipe/sp_context.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 715e1ef656..89252e38d5 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -139,7 +139,7 @@ i915_max_texture_size(struct pipe_context *pipe, unsigned textureType, static int -i915_get_param(struct pipe_context *pipe, uint param) +i915_get_param(struct pipe_context *pipe, int param) { switch (param) { case PIPE_PARAM_FS_NEEDS_POS: diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index ebfd29fadb..e6623f4356 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -240,7 +240,7 @@ i915_update_texture(struct i915_context *i915, uint unit, { - struct pipe_buffer_handle *p = driBOReference(mt->region->buffer); + /*struct pipe_buffer_handle *p =*/ driBOReference(mt->region->buffer); } #if 0 diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index ebd5530950..cac44606f7 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -236,7 +236,7 @@ static const char *softpipe_get_vendor( struct pipe_context *pipe ) return "Tungsten Graphics, Inc."; } -static int softpipe_get_param(struct pipe_context *pipe, uint param) +static int softpipe_get_param(struct pipe_context *pipe, int param) { switch (param) { case PIPE_PARAM_FS_NEEDS_POS: -- cgit v1.2.3 From 6cb87cf26f904b891faa42268f373864fa33541d Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 21 Sep 2007 07:00:20 -0400 Subject: Make the alpha test state a cso. --- src/mesa/pipe/cso_cache/cso_cache.c | 7 ++++ src/mesa/pipe/cso_cache/cso_cache.h | 13 +++++-- src/mesa/pipe/failover/fo_context.h | 4 +-- src/mesa/pipe/failover/fo_state.c | 45 +++++++++++++++++++++---- src/mesa/pipe/failover/fo_state_emit.c | 3 +- src/mesa/pipe/i915simple/i915_context.h | 2 +- src/mesa/pipe/i915simple/i915_state.c | 24 ++++++++++--- src/mesa/pipe/i915simple/i915_state_immediate.c | 6 ++-- src/mesa/pipe/p_context.h | 32 ++++++++++-------- src/mesa/pipe/softpipe/sp_context.c | 4 ++- src/mesa/pipe/softpipe/sp_context.h | 2 +- src/mesa/pipe/softpipe/sp_quad.c | 2 +- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 4 +-- src/mesa/pipe/softpipe/sp_state.h | 12 +++++-- src/mesa/pipe/softpipe/sp_state_blend.c | 20 +++++++++-- src/mesa/state_tracker/st_atom.c | 1 + src/mesa/state_tracker/st_atom.h | 1 + src/mesa/state_tracker/st_atom_alphatest.c | 15 ++++----- src/mesa/state_tracker/st_cache.c | 20 +++++++++++ src/mesa/state_tracker/st_cache.h | 4 +++ src/mesa/state_tracker/st_cb_clear.c | 6 ++-- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_context.h | 10 +++--- 23 files changed, 176 insertions(+), 63 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_cache.c b/src/mesa/pipe/cso_cache/cso_cache.c index 71f0d08726..0bba5914dc 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.c +++ b/src/mesa/pipe/cso_cache/cso_cache.c @@ -90,6 +90,9 @@ static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_ case CSO_VERTEX_SHADER: hash = sc->vs_hash; break; + case CSO_ALPHA_TEST: + hash = sc->alpha_hash; + break; } return hash; @@ -110,6 +113,8 @@ static int _cso_size_for_type(enum cso_cache_type type) return sizeof(struct pipe_shader_state); case CSO_VERTEX_SHADER: return sizeof(struct pipe_shader_state); + case CSO_ALPHA_TEST: + return sizeof(struct pipe_alpha_test_state); } return 0; } @@ -164,6 +169,7 @@ struct cso_cache *cso_cache_create(void) sc->rasterizer_hash = cso_hash_create(); sc->fs_hash = cso_hash_create(); sc->vs_hash = cso_hash_create(); + sc->alpha_hash = cso_hash_create(); return sc; } @@ -177,5 +183,6 @@ void cso_cache_delete(struct cso_cache *sc) cso_hash_delete(sc->rasterizer_hash); cso_hash_delete(sc->fs_hash); cso_hash_delete(sc->vs_hash); + cso_hash_delete(sc->alpha_hash); free(sc); } diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h index 2acb58c66b..cd36dd51e9 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -40,12 +40,13 @@ struct cso_hash; struct cso_cache { + struct cso_hash *alpha_hash; struct cso_hash *blend_hash; - struct cso_hash *sampler_hash; struct cso_hash *depth_stencil_hash; - struct cso_hash *rasterizer_hash; struct cso_hash *fs_hash; struct cso_hash *vs_hash; + struct cso_hash *rasterizer_hash; + struct cso_hash *sampler_hash; }; struct cso_blend { @@ -78,13 +79,19 @@ struct cso_sampler { void *data; }; +struct cso_alpha_test { + struct pipe_alpha_test_state state; + void *data; +}; + enum cso_cache_type { CSO_BLEND, CSO_SAMPLER, CSO_DEPTH_STENCIL, CSO_RASTERIZER, CSO_FRAGMENT_SHADER, - CSO_VERTEX_SHADER + CSO_VERTEX_SHADER, + CSO_ALPHA_TEST }; unsigned cso_construct_key(void *item, int item_size); diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index 8a2fbe2be9..7a597013ab 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -64,12 +64,13 @@ struct fo_state { void *sw_state; void *hw_state; }; -struct failover_context { +struct failover_context { struct pipe_context pipe; /**< base class */ /* The most recent drawing state as set by the driver: */ + const struct fo_state *alpha_test; const struct fo_state *blend; const struct fo_state *sampler[PIPE_MAX_SAMPLERS]; const struct fo_state *depth_stencil; @@ -77,7 +78,6 @@ struct failover_context { const struct fo_state *fragment_shader; const struct fo_state *vertex_shader; - struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index ce3f0ca635..f63137f591 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -45,15 +45,44 @@ * lower overheads. */ +static void * +failover_create_alpha_test_state(struct pipe_context *pipe, + const struct pipe_alpha_test_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_alpha_test_state(pipe, templ); + state->hw_state = failover->hw->create_alpha_test_state(pipe, templ); + + return state; +} + static void -failover_set_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *alpha) +failover_bind_alpha_test_state(struct pipe_context *pipe, + void *alpha) { struct failover_context *failover = failover_context(pipe); + struct fo_state *state = (struct fo_state *)alpha; - failover->alpha_test = *alpha; + failover->alpha_test = state; failover->dirty |= FO_NEW_ALPHA_TEST; - failover->hw->set_alpha_test_state( failover->hw, alpha ); + failover->hw->bind_alpha_test_state(failover->hw, + state->hw_state); +} + +static void +failover_delete_alpha_test_state(struct pipe_context *pipe, + void *alpha) +{ + struct fo_state *state = (struct fo_state*)alpha; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_alpha_test_state(pipe, state->sw_state); + failover->hw->delete_alpha_test_state(pipe, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); } @@ -95,7 +124,7 @@ failover_delete_blend_state( struct pipe_context *pipe, free(state); } -static void +static void failover_set_blend_color( struct pipe_context *pipe, const struct pipe_blend_color *blend_color ) { @@ -414,8 +443,11 @@ failover_set_vertex_element(struct pipe_context *pipe, void failover_init_state_functions( struct failover_context *failover ) { + failover->pipe.create_alpha_test_state = failover_create_alpha_test_state; + failover->pipe.bind_alpha_test_state = failover_bind_alpha_test_state; + failover->pipe.delete_alpha_test_state = failover_delete_alpha_test_state; failover->pipe.create_blend_state = failover_create_blend_state; - failover->pipe.bind_blend_state = failover_bind_blend_state; + failover->pipe.bind_blend_state = failover_bind_blend_state; failover->pipe.delete_blend_state = failover_delete_blend_state; failover->pipe.create_sampler_state = failover_create_sampler_state; failover->pipe.bind_sampler_state = failover_bind_sampler_state; @@ -433,7 +465,6 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.bind_vs_state = failover_bind_vs_state; failover->pipe.delete_vs_state = failover_delete_vs_state; - failover->pipe.set_alpha_test_state = failover_set_alpha_test_state; failover->pipe.set_blend_color = failover_set_blend_color; failover->pipe.set_clip_state = failover_set_clip_state; failover->pipe.set_clear_color_state = failover_set_clear_color_state; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index c0ea681024..a3aff8abd2 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -56,7 +56,8 @@ failover_state_emit( struct failover_context *failover ) unsigned i; if (failover->dirty & FO_NEW_ALPHA_TEST) - failover->sw->set_alpha_test_state( failover->sw, &failover->alpha_test ); + failover->sw->bind_alpha_test_state( failover->sw, + failover->alpha_test->sw_state ); if (failover->dirty & FO_NEW_BLEND) failover->sw->bind_blend_state( failover->sw, diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index f1e10f3433..9958a8592d 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -154,13 +154,13 @@ struct i915_context /* The most recent drawing state as set by the driver: */ + const struct pipe_alpha_test_state *alpha_test; const struct i915_blend_state *blend; const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct i915_depth_stencil_state *depth_stencil; const struct i915_rasterizer_state *rasterizer; const struct pipe_shader_state *fs; - struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 525f8ce13a..8bfd2da3b5 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -394,16 +394,29 @@ static void i915_delete_depth_stencil_state(struct pipe_context *pipe, free(depth_stencil); } -static void i915_set_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *alpha) + +static void * +i915_create_alpha_test_state(struct pipe_context *pipe, + const struct pipe_alpha_test_state *alpha) +{ + return 0; +} + +static void i915_bind_alpha_test_state(struct pipe_context *pipe, + void *alpha) { struct i915_context *i915 = i915_context(pipe); - i915->alpha_test = *alpha; + i915->alpha_test = (const struct pipe_alpha_test_state*)alpha; i915->dirty |= I915_NEW_ALPHA_TEST; } +static void i915_delete_alpha_test_state(struct pipe_context *pipe, + void *alpha) +{ +} + static void i915_set_scissor_state( struct pipe_context *pipe, const struct pipe_scissor_state *scissor ) { @@ -664,6 +677,10 @@ static void i915_set_vertex_element( struct pipe_context *pipe, void i915_init_state_functions( struct i915_context *i915 ) { + i915->pipe.create_alpha_test_state = i915_create_alpha_test_state; + i915->pipe.bind_alpha_test_state = i915_bind_alpha_test_state; + i915->pipe.delete_alpha_test_state = i915_delete_alpha_test_state; + i915->pipe.create_blend_state = i915_create_blend_state; i915->pipe.bind_blend_state = i915_bind_blend_state; i915->pipe.delete_blend_state = i915_delete_blend_state; @@ -686,7 +703,6 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.bind_vs_state = i915_bind_vs_state; i915->pipe.delete_vs_state = i915_delete_shader_state; - i915->pipe.set_alpha_test_state = i915_set_alpha_test_state; i915->pipe.set_blend_color = i915_set_blend_color; i915->pipe.set_clip_state = i915_set_clip_state; i915->pipe.set_clear_color_state = i915_set_clear_color_state; diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 874c3819f2..014fddfdda 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -121,9 +121,9 @@ static void upload_S6( struct i915_context *i915 ) /* I915_NEW_ALPHA_TEST */ - if (i915->alpha_test.enabled) { - int test = i915_translate_compare_func(i915->alpha_test.func); - ubyte refByte = float_to_ubyte(i915->alpha_test.ref); + if (i915->alpha_test->enabled) { + int test = i915_translate_compare_func(i915->alpha_test->func); + ubyte refByte = float_to_ubyte(i915->alpha_test->ref); LIS6 |= (S6_ALPHA_TEST_ENABLE | diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 07ee019880..8ba1031efe 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -86,26 +86,30 @@ struct pipe_context { /* * State functions */ + void * (*create_alpha_test_state)(struct pipe_context *, + const struct pipe_alpha_test_state *); + void (*bind_alpha_test_state)(struct pipe_context *, void *); + void (*delete_alpha_test_state)(struct pipe_context *, void *); + void * (*create_blend_state)(struct pipe_context *, const struct pipe_blend_state *); - void (*bind_blend_state)(struct pipe_context *, void *); - void (*delete_blend_state)(struct pipe_context *, void *); + void (*bind_blend_state)(struct pipe_context *, void *); + void (*delete_blend_state)(struct pipe_context *, void *); void * (*create_sampler_state)(struct pipe_context *, const struct pipe_sampler_state *); - void (*bind_sampler_state)(struct pipe_context *, unsigned unit, - void *); - void (*delete_sampler_state)(struct pipe_context *, void *); + void (*bind_sampler_state)(struct pipe_context *, unsigned unit, void *); + void (*delete_sampler_state)(struct pipe_context *, void *); - void *(*create_rasterizer_state)(struct pipe_context *, - const struct pipe_rasterizer_state *); - void (*bind_rasterizer_state)(struct pipe_context *, void *); - void (*delete_rasterizer_state)(struct pipe_context *, void *); + void * (*create_rasterizer_state)(struct pipe_context *, + const struct pipe_rasterizer_state *); + void (*bind_rasterizer_state)(struct pipe_context *, void *); + void (*delete_rasterizer_state)(struct pipe_context *, void *); void * (*create_depth_stencil_state)(struct pipe_context *, const struct pipe_depth_stencil_state *); - void (*bind_depth_stencil_state)(struct pipe_context *, void *); - void (*delete_depth_stencil_state)(struct pipe_context *, void *); + void (*bind_depth_stencil_state)(struct pipe_context *, void *); + void (*delete_depth_stencil_state)(struct pipe_context *, void *); void * (*create_fs_state)(struct pipe_context *, const struct pipe_shader_state *); @@ -117,9 +121,9 @@ struct pipe_context { void (*bind_vs_state)(struct pipe_context *, void *); void (*delete_vs_state)(struct pipe_context *, void *); - void (*set_alpha_test_state)( struct pipe_context *, - const struct pipe_alpha_test_state * ); - + /* The following look more properties than states. + * maybe combine a few of them into states or pass them + * in the bind calls to the state */ void (*set_blend_color)( struct pipe_context *, const struct pipe_blend_color * ); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index cac44606f7..a6ab45314c 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -260,6 +260,9 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.get_param = softpipe_get_param; /* state setters */ + softpipe->pipe.create_alpha_test_state = softpipe_create_alpha_test_state; + softpipe->pipe.bind_alpha_test_state = softpipe_bind_alpha_test_state; + softpipe->pipe.delete_alpha_test_state = softpipe_delete_alpha_test_state; softpipe->pipe.create_blend_state = softpipe_create_blend_state; softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; @@ -279,7 +282,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; softpipe->pipe.delete_vs_state = softpipe_delete_shader_state; - softpipe->pipe.set_alpha_test_state = softpipe_set_alpha_test_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 95215eb640..7b1518cfac 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -70,6 +70,7 @@ struct softpipe_context { /* The most recent drawing state as set by the driver: */ + const struct pipe_alpha_test_state *alpha_test; const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; @@ -77,7 +78,6 @@ struct softpipe_context { const struct pipe_shader_state *fs; const struct pipe_shader_state *vs; - struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index fc4f8328cf..74a21dc54b 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -54,7 +54,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.depth_test; } - if (sp->alpha_test.enabled) { + if (sp->alpha_test->enabled) { sp->quad.alpha_test->next = sp->quad.first; sp->quad.first = sp->quad.alpha_test; } diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index 64c1624bd0..4f28414b0e 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -14,10 +14,10 @@ static void alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - const float ref = softpipe->alpha_test.ref; + const float ref = softpipe->alpha_test->ref; unsigned passMask = 0x0, j; - switch (softpipe->alpha_test.func) { + switch (softpipe->alpha_test->func) { case PIPE_FUNC_NEVER: quad->mask = 0x0; break; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 62323c41c3..f0e1461d25 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -33,6 +33,15 @@ #include "pipe/p_state.h" + +void * +softpipe_create_alpha_test_state(struct pipe_context *, + const struct pipe_alpha_test_state *); +void +softpipe_bind_alpha_test_state(struct pipe_context *, void *); +void +softpipe_delete_alpha_test_state(struct pipe_context *, void *); + void * softpipe_create_blend_state(struct pipe_context *, const struct pipe_blend_state *); @@ -62,9 +71,6 @@ void softpipe_delete_rasterizer_state(struct pipe_context *, void *); void softpipe_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); -void softpipe_set_alpha_test_state( struct pipe_context *, - const struct pipe_alpha_test_state * ); - void softpipe_set_blend_color( struct pipe_context *pipe, const struct pipe_blend_color *blend_color ); diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index cf47607955..cb0921f687 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -71,17 +71,31 @@ void softpipe_set_blend_color( struct pipe_context *pipe, * into one file. */ +void * +softpipe_create_alpha_test_state(struct pipe_context *pipe, + const struct pipe_alpha_test_state *alpha) +{ + return 0; +} + void -softpipe_set_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *alpha) +softpipe_bind_alpha_test_state(struct pipe_context *pipe, + void *alpha) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->alpha_test = *alpha; + softpipe->alpha_test = (const struct pipe_alpha_test_state *)alpha; softpipe->dirty |= SP_NEW_ALPHA_TEST; } +void +softpipe_delete_alpha_test_state(struct pipe_context *pipe, + void *alpha) +{ + /* do nothing */ +} + void * softpipe_create_depth_stencil_state(struct pipe_context *pipe, const struct pipe_depth_stencil_state *depth_stencil) diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index a4af3aeb20..fc339b91ed 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -62,6 +62,7 @@ static const struct st_tracked_state *atoms[] = &st_update_texture, &st_update_vs_constants, &st_update_fs_constants, + &st_update_alpha_test }; diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 26f6514698..6710c1a269 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -60,6 +60,7 @@ const struct st_tracked_state st_update_sampler; const struct st_tracked_state st_update_texture; const struct st_tracked_state st_update_fs_constants; const struct st_tracked_state st_update_vs_constants; +const struct st_tracked_state st_update_alpha_test; #endif diff --git a/src/mesa/state_tracker/st_atom_alphatest.c b/src/mesa/state_tracker/st_atom_alphatest.c index 4378154053..873520ab02 100644 --- a/src/mesa/state_tracker/st_atom_alphatest.c +++ b/src/mesa/state_tracker/st_atom_alphatest.c @@ -33,6 +33,7 @@ #include "st_context.h" +#include "st_cache.h" #include "st_atom.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -63,6 +64,7 @@ static void update_alpha_test( struct st_context *st ) { struct pipe_alpha_test_state alpha; + const struct cso_alpha_test *cso; memset(&alpha, 0, sizeof(alpha)); @@ -71,11 +73,11 @@ update_alpha_test( struct st_context *st ) alpha.func = gl_alpha_func_to_sp(st->ctx->Color.AlphaFunc); alpha.ref = st->ctx->Color.AlphaRef; } - - if (memcmp(&alpha, &st->state.alpha_test, sizeof(alpha)) != 0) { + cso = st_cached_alpha_test_state(st, &alpha); + if (st->state.alpha_test != cso) { /* state has changed */ - st->state.alpha_test = alpha; /* struct copy */ - st->pipe->set_alpha_test_state(st->pipe, &alpha); /* set new state */ + st->state.alpha_test = cso; + st->pipe->bind_alpha_test_state(st->pipe, cso->data); /* bind new state */ } } @@ -88,8 +90,3 @@ const struct st_tracked_state st_update_alpha_test = { }, .update = update_alpha_test }; - - - - - diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 01d1934232..1686721990 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -159,3 +159,23 @@ st_cached_vs_state(struct st_context *st, } return (struct cso_vertex_shader*)(cso_hash_iter_data(iter)); } + +const struct cso_alpha_test * +st_cached_alpha_test_state(struct st_context *st, + const struct pipe_alpha_test_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_alpha_test_state)); + struct cso_hash_iter iter = cso_find_state_template(st->cache, + hash_key, CSO_ALPHA_TEST, + (void*)templ); + if (cso_hash_iter_is_null(iter)) { + struct cso_alpha_test *cso = malloc(sizeof(struct cso_alpha_test)); + memcpy(&cso->state, templ, sizeof(struct pipe_alpha_test_state)); + cso->data = st->pipe->create_alpha_test_state(st->pipe, &cso->state); + if (!cso->data) + cso->data = &cso->state; + iter = cso_insert_state(st->cache, hash_key, CSO_ALPHA_TEST, cso); + } + return ((struct cso_alpha_test *)cso_hash_iter_data(iter)); +} diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index 483af6fdb4..422f668c56 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -39,6 +39,10 @@ struct pipe_blend_state; struct pipe_sampler_state; struct st_context; +const struct cso_alpha_test * +st_cached_alpha_test_state(struct st_context *st, + const struct pipe_alpha_test_state *alpha); + const struct cso_blend * st_cached_blend_state(struct st_context *st, const struct pipe_blend_state *blend); diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 5e63205088..c53446d588 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -275,8 +275,10 @@ clear_with_quad(GLcontext *ctx, /* alpha state: disabled */ { struct pipe_alpha_test_state alpha_test; + const struct cso_alpha_test *cso; memset(&alpha_test, 0, sizeof(alpha_test)); - pipe->set_alpha_test_state(pipe, &alpha_test); + cso = st_cached_alpha_test_state(st, &alpha_test); + pipe->bind_alpha_test_state(pipe, cso->data); } /* blend state: RGBA masking */ @@ -379,7 +381,7 @@ clear_with_quad(GLcontext *ctx, draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); /* Restore pipe state */ - pipe->set_alpha_test_state(pipe, &st->state.alpha_test); + pipe->bind_alpha_test_state(pipe, st->state.alpha_test->data); pipe->bind_blend_state(pipe, st->state.blend->data); pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil->data); pipe->bind_fs_state(pipe, st->state.fs->data); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index d814e34157..65c5465546 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -468,7 +468,7 @@ compatible_formats(GLenum format, GLenum type, GLuint pipeFormat) static GLboolean any_fragment_ops(const struct st_context *st) { - if (st->state.alpha_test.enabled || + if (st->state.alpha_test->state.enabled || st->state.blend->state.blend_enable || st->state.blend->state.logicop_enable || st->state.depth_stencil->state.depth.enabled) diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 55a857f46d..3713328eb1 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -75,14 +75,14 @@ struct st_context * though, we just shove random objects across the interface. */ struct { - const struct cso_blend *blend; - const struct cso_sampler *sampler[PIPE_MAX_SAMPLERS]; - const struct cso_depth_stencil *depth_stencil; - const struct cso_rasterizer *rasterizer; + const struct cso_alpha_test *alpha_test; + const struct cso_blend *blend; + const struct cso_sampler *sampler[PIPE_MAX_SAMPLERS]; + const struct cso_depth_stencil *depth_stencil; + const struct cso_rasterizer *rasterizer; const struct cso_fragment_shader *fs; const struct cso_vertex_shader *vs; - struct pipe_alpha_test_state alpha_test; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; -- cgit v1.2.3 From 42df29d3b9139f7b8bb757893bb8bf930585cc33 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 21 Sep 2007 07:06:19 -0400 Subject: Convert i915 to use cso semantics of alpha_test. --- src/mesa/pipe/i915simple/i915_context.h | 8 ++++++-- src/mesa/pipe/i915simple/i915_state.c | 19 +++++++++++++++---- src/mesa/pipe/i915simple/i915_state_immediate.c | 10 +--------- 3 files changed, 22 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 9958a8592d..dc19b6efeb 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -146,15 +146,19 @@ struct i915_sampler_state { const struct pipe_sampler_state *templ; }; +struct i915_alpha_test_state { + unsigned LIS6; +}; + struct i915_context { - struct pipe_context pipe; + struct pipe_context pipe; struct i915_winsys *winsys; struct draw_context *draw; /* The most recent drawing state as set by the driver: */ - const struct pipe_alpha_test_state *alpha_test; + const struct i915_alpha_test_state *alpha_test; const struct i915_blend_state *blend; const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct i915_depth_stencil_state *depth_stencil; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 8bfd2da3b5..5c6d37a466 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -397,9 +397,19 @@ static void i915_delete_depth_stencil_state(struct pipe_context *pipe, static void * i915_create_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *alpha) + const struct pipe_alpha_test_state *alpha_test) { - return 0; + struct i915_alpha_test_state *cso = calloc(1, sizeof(struct i915_alpha_test_state)); + + if (alpha_test->enabled) { + int test = i915_translate_compare_func(alpha_test->func); + ubyte refByte = float_to_ubyte(alpha_test->ref); + + cso->LIS6 |= (S6_ALPHA_TEST_ENABLE | + (test << S6_ALPHA_TEST_FUNC_SHIFT) | + (((unsigned) refByte) << S6_ALPHA_REF_SHIFT)); + } + return cso; } static void i915_bind_alpha_test_state(struct pipe_context *pipe, @@ -407,7 +417,7 @@ static void i915_bind_alpha_test_state(struct pipe_context *pipe, { struct i915_context *i915 = i915_context(pipe); - i915->alpha_test = (const struct pipe_alpha_test_state*)alpha; + i915->alpha_test = (const struct i915_alpha_test_state*)alpha; i915->dirty |= I915_NEW_ALPHA_TEST; } @@ -415,6 +425,7 @@ static void i915_bind_alpha_test_state(struct pipe_context *pipe, static void i915_delete_alpha_test_state(struct pipe_context *pipe, void *alpha) { + free(alpha); } static void i915_set_scissor_state( struct pipe_context *pipe, @@ -650,7 +661,7 @@ static void i915_bind_rasterizer_state( struct pipe_context *pipe, static void i915_delete_rasterizer_state(struct pipe_context *pipe, void *setup) { - /* do nothing */ + free(setup); } static void i915_set_vertex_buffer( struct pipe_context *pipe, diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 014fddfdda..479d72a87f 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -121,15 +121,7 @@ static void upload_S6( struct i915_context *i915 ) /* I915_NEW_ALPHA_TEST */ - if (i915->alpha_test->enabled) { - int test = i915_translate_compare_func(i915->alpha_test->func); - ubyte refByte = float_to_ubyte(i915->alpha_test->ref); - - - LIS6 |= (S6_ALPHA_TEST_ENABLE | - (test << S6_ALPHA_TEST_FUNC_SHIFT) | - (((unsigned) refByte) << S6_ALPHA_REF_SHIFT)); - } + LIS6 |= i915->alpha_test->LIS6; /* I915_NEW_BLEND */ -- cgit v1.2.3 From 562c9410630c7faeb3b04bd8b4fa8d29c9193eaa Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 10:38:18 -0600 Subject: fix 1/i typos --- src/mesa/state_tracker/st_atom_vs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index b09f2bece7..dcd3564252 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -100,7 +100,7 @@ st_translate_vertex_shader(struct st_context *st, case VERT_ATTRIB_TEX6: case VERT_ATTRIB_TEX7: vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_TEXCOORD; - vs.input_semantic_index[vs.num_inputs] = 1 - VERT_ATTRIB_TEX0; + vs.input_semantic_index[vs.num_inputs] = i - VERT_ATTRIB_TEX0; break; case VERT_ATTRIB_GENERIC0: case VERT_ATTRIB_GENERIC1: @@ -112,7 +112,7 @@ st_translate_vertex_shader(struct st_context *st, case VERT_ATTRIB_GENERIC7: assert(i < VERT_ATTRIB_MAX); vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[vs.num_inputs] = 1 - VERT_ATTRIB_GENERIC0; + vs.input_semantic_index[vs.num_inputs] = i - VERT_ATTRIB_GENERIC0; break; default: assert(0); -- cgit v1.2.3 From ed10ece932162ac7c3c5ee125a0ca35c805240ad Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 10:40:18 -0600 Subject: Pass vertex program input semantics into tgsi_mesa_compile_vp_program() to produce DCLs. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 44 +++++++++++++++++++++++----------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 3 +++ src/mesa/state_tracker/st_atom_vs.c | 3 +++ 3 files changed, 36 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index c7b62c9b84..ea6716f1a2 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -469,7 +469,7 @@ compile_instruction( } static struct tgsi_full_declaration -make_frag_input_decl( +make_input_decl( GLuint index, GLuint interpolate, GLuint usage_mask, @@ -560,20 +560,20 @@ tgsi_mesa_compile_fp_program( switch (inputSemanticName[i]) { case TGSI_SEMANTIC_POSITION: /* Fragment XY pos */ - fulldecl = make_frag_input_decl(i, - TGSI_INTERPOLATE_CONSTANT, - TGSI_WRITEMASK_XY, - TGSI_SEMANTIC_POSITION, 0 ); + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_CONSTANT, + TGSI_WRITEMASK_XY, + TGSI_SEMANTIC_POSITION, 0 ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], header, maxTokens - ti ); /* Fragment ZW pos */ - fulldecl = make_frag_input_decl(i, - TGSI_INTERPOLATE_LINEAR, - TGSI_WRITEMASK_ZW, - TGSI_SEMANTIC_POSITION, 0 ); + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_LINEAR, + TGSI_WRITEMASK_ZW, + TGSI_SEMANTIC_POSITION, 0 ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -581,11 +581,11 @@ tgsi_mesa_compile_fp_program( maxTokens - ti ); break; default: - fulldecl = make_frag_input_decl(i, - interpMode[i], - TGSI_WRITEMASK_XYZW, - inputSemanticName[i], - inputSemanticIndex[i]); + fulldecl = make_input_decl(i, + interpMode[i], + TGSI_WRITEMASK_XYZW, + inputSemanticName[i], + inputSemanticIndex[i]); ti += tgsi_build_full_declaration(&fulldecl, &tokens[ti], header, @@ -682,7 +682,10 @@ tgsi_mesa_compile_fp_program( GLboolean tgsi_mesa_compile_vp_program( const struct gl_vertex_program *program, + GLuint numInputs, const GLuint inputMapping[], + const ubyte inputSemanticName[], + const ubyte inputSemanticIndex[], const GLuint outputMapping[], struct tgsi_token *tokens, GLuint maxTokens) @@ -704,6 +707,19 @@ tgsi_mesa_compile_vp_program( /* XXX todo: input/output declarations */ + for (i = 0; i < numInputs; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_CONSTANT, /* no interp */ + TGSI_WRITEMASK_XYZW, + inputSemanticName[i], + inputSemanticIndex[i]); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + for( i = 0; i < program->Base.NumInstructions; i++ ) { if( compile_instruction( diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index b40331981e..dc020cefaa 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -22,7 +22,10 @@ tgsi_mesa_compile_fp_program( GLboolean tgsi_mesa_compile_vp_program( const struct gl_vertex_program *program, + GLuint numInputs, const GLuint inputMapping[], + const ubyte inputSemanticName[], + const ubyte inputSemanticIndex[], const GLuint outputMapping[], struct tgsi_token *tokens, GLuint maxTokens ); diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index dcd3564252..cf0414dd1e 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -183,7 +183,10 @@ st_translate_vertex_shader(struct st_context *st, /* XXX: fix static allocation of tokens: */ tgsi_mesa_compile_vp_program( &stvp->Base, + vs.num_inputs, stvp->input_to_index, + vs.input_semantic_name, + vs.input_semantic_index, st->vertex_attrib_to_slot, stvp->tokens, ST_FP_MAX_TOKENS ); -- cgit v1.2.3 From f38881c542a6e96ba45f185e372940cbb183647d Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 10:53:45 -0600 Subject: fix in/out typos --- src/mesa/state_tracker/st_atom_vs.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index cf0414dd1e..d9ae0cdfac 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -133,23 +133,23 @@ st_translate_vertex_shader(struct st_context *st, switch (i) { case VERT_RESULT_HPOS: vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_POSITION; - vs.output_semantic_index[vs.num_inputs] = 0; + vs.output_semantic_index[vs.num_outputs] = 0; break; case VERT_RESULT_COL0: vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; - vs.output_semantic_index[vs.num_inputs] = 0; + vs.output_semantic_index[vs.num_outputs] = 0; break; case VERT_RESULT_COL1: vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; - vs.output_semantic_index[vs.num_inputs] = 1; + vs.output_semantic_index[vs.num_outputs] = 1; break; case VERT_RESULT_BFC0: - vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_BCOLOR; - vs.output_semantic_index[vs.num_inputs] = 0; + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_BCOLOR; + vs.output_semantic_index[vs.num_outputs] = 0; break; case VERT_RESULT_BFC1: - vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_BCOLOR; - vs.output_semantic_index[vs.num_inputs] = 1; + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_BCOLOR; + vs.output_semantic_index[vs.num_outputs] = 1; break; case VERT_RESULT_FOGC: case VERT_RESULT_PSIZ: @@ -164,15 +164,15 @@ st_translate_vertex_shader(struct st_context *st, case VERT_RESULT_TEX5: case VERT_RESULT_TEX6: case VERT_RESULT_TEX7: - vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_TEXCOORD; - vs.output_semantic_index[vs.num_inputs] = i - VERT_RESULT_TEX0; + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_TEXCOORD; + vs.output_semantic_index[vs.num_outputs] = i - VERT_RESULT_TEX0; break; case VERT_RESULT_VAR0: /* fall-through */ default: assert(i - VERT_RESULT_VAR0 < MAX_VARYING); - vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[vs.num_inputs] = i - VERT_RESULT_VAR0; + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_GENERIC; + vs.output_semantic_index[vs.num_outputs] = i - VERT_RESULT_VAR0; } vs.num_outputs++; -- cgit v1.2.3 From 9f96c5d619a478a61c809e88ef53e59005471054 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 10:56:32 -0600 Subject: pass vs output semantics to tgsi_mesa_compile_vp_program() to generate output DECs --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 19 +++++++++++++++++-- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 3 +++ src/mesa/state_tracker/st_atom_vs.c | 7 +++++-- 3 files changed, 25 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index ea6716f1a2..8a3fda2c63 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -686,7 +686,10 @@ tgsi_mesa_compile_vp_program( const GLuint inputMapping[], const ubyte inputSemanticName[], const ubyte inputSemanticIndex[], + GLuint numOutputs, const GLuint outputMapping[], + const ubyte outputSemanticName[], + const ubyte outputSemanticIndex[], struct tgsi_token *tokens, GLuint maxTokens) { @@ -705,8 +708,7 @@ tgsi_mesa_compile_vp_program( ti = 3; - /* XXX todo: input/output declarations - */ + /* input decls */ for (i = 0; i < numInputs; i++) { struct tgsi_full_declaration fulldecl; fulldecl = make_input_decl(i, @@ -720,6 +722,19 @@ tgsi_mesa_compile_vp_program( maxTokens - ti ); } + /* output decls */ + for (i = 0; i < numOutputs; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_frag_output_decl(i, + outputSemanticName[i], + outputSemanticIndex[i], + TGSI_WRITEMASK_XYZW ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + for( i = 0; i < program->Base.NumInstructions; i++ ) { if( compile_instruction( diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index dc020cefaa..cf85eb9c0b 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -26,7 +26,10 @@ tgsi_mesa_compile_vp_program( const GLuint inputMapping[], const ubyte inputSemanticName[], const ubyte inputSemanticIndex[], + GLuint numOutputs, const GLuint outputMapping[], + const ubyte outputSemanticName[], + const ubyte outputSemanticIndex[], struct tgsi_token *tokens, GLuint maxTokens ); diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index d9ae0cdfac..c1bcdbe5a1 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -136,11 +136,11 @@ st_translate_vertex_shader(struct st_context *st, vs.output_semantic_index[vs.num_outputs] = 0; break; case VERT_RESULT_COL0: - vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_COLOR; vs.output_semantic_index[vs.num_outputs] = 0; break; case VERT_RESULT_COL1: - vs.output_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_COLOR; vs.output_semantic_index[vs.num_outputs] = 1; break; case VERT_RESULT_BFC0: @@ -187,7 +187,10 @@ st_translate_vertex_shader(struct st_context *st, stvp->input_to_index, vs.input_semantic_name, vs.input_semantic_index, + vs.num_outputs, st->vertex_attrib_to_slot, + vs.output_semantic_name, + vs.output_semantic_index, stvp->tokens, ST_FP_MAX_TOKENS ); vs.tokens = &stvp->tokens[0]; -- cgit v1.2.3 From 8d987d3c4edf489880e5e852e31b6883a15b6a14 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 10:57:30 -0600 Subject: minor renaming --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 8a3fda2c63..b1cf122695 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -494,7 +494,7 @@ make_input_decl( } static struct tgsi_full_declaration -make_frag_output_decl( +make_output_decl( GLuint index, GLuint semantic_name, GLuint semantic_index, @@ -602,7 +602,7 @@ tgsi_mesa_compile_fp_program( program->Base.OutputsWritten == (program->Base.OutputsWritten & ((1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR))) ); - fulldecl = make_frag_output_decl( + fulldecl = make_output_decl( 0, TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ TGSI_WRITEMASK_Z ); @@ -613,7 +613,7 @@ tgsi_mesa_compile_fp_program( maxTokens - ti ); if( program->Base.OutputsWritten & (1 << FRAG_RESULT_COLR) ) { - fulldecl = make_frag_output_decl( + fulldecl = make_output_decl( 1, TGSI_SEMANTIC_COLOR, 0, TGSI_WRITEMASK_XYZW ); @@ -725,10 +725,10 @@ tgsi_mesa_compile_vp_program( /* output decls */ for (i = 0; i < numOutputs; i++) { struct tgsi_full_declaration fulldecl; - fulldecl = make_frag_output_decl(i, - outputSemanticName[i], - outputSemanticIndex[i], - TGSI_WRITEMASK_XYZW ); + fulldecl = make_output_decl(i, + outputSemanticName[i], + outputSemanticIndex[i], + TGSI_WRITEMASK_XYZW ); ti += tgsi_build_full_declaration(&fulldecl, &tokens[ti], header, -- cgit v1.2.3 From 97b81eb98f57e9405ce3665df24eedd1a1af39fc Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 11:01:26 -0600 Subject: rename index var from i to attr, use FRAG_ATTRIB_MAX instead of 16 --- src/mesa/state_tracker/st_atom_fs.c | 24 ++++++++++++------------ src/mesa/state_tracker/st_atom_vs.c | 32 ++++++++++++++++---------------- 2 files changed, 28 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 8ee4f407c7..dd9d2476ba 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -60,7 +60,7 @@ st_translate_fragment_shader(struct st_context *st, struct pipe_shader_state fs; const struct cso_fragment_shader *cso; GLuint interpMode[16]; /* XXX size? */ - GLuint i; + GLuint attr; GLbitfield inputsRead = stfp->Base.Base.InputsRead; /* Check if all fragment programs need the fragment position (in order @@ -71,11 +71,11 @@ st_translate_fragment_shader(struct st_context *st, memset(&fs, 0, sizeof(fs)); - for (i = 0; i < 16; i++) { - if (inputsRead & (1 << i)) { - inputMapping[i] = fs.num_inputs; + for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) { + if (inputsRead & (1 << attr)) { + inputMapping[attr] = fs.num_inputs; - switch (i) { + switch (attr) { case FRAG_ATTRIB_WPOS: fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_POSITION; fs.input_semantic_index[fs.num_inputs] = 0; @@ -103,14 +103,14 @@ st_translate_fragment_shader(struct st_context *st, case FRAG_ATTRIB_TEX6: case FRAG_ATTRIB_TEX7: fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_TEXCOORD; - fs.input_semantic_index[fs.num_inputs] = i - FRAG_ATTRIB_TEX0; + fs.input_semantic_index[fs.num_inputs] = attr - FRAG_ATTRIB_TEX0; interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; break; case FRAG_ATTRIB_VAR0: /* fall-through */ default: fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_GENERIC; - fs.input_semantic_index[fs.num_inputs] = i - FRAG_ATTRIB_VAR0; + fs.input_semantic_index[fs.num_inputs] = attr - FRAG_ATTRIB_VAR0; interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; } @@ -121,16 +121,16 @@ st_translate_fragment_shader(struct st_context *st, /* * Outputs */ - for (i = 0; i < FRAG_RESULT_MAX; i++) { - if (stfp->Base.Base.OutputsWritten & (1 << i)) { - switch (i) { + for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { + if (stfp->Base.Base.OutputsWritten & (1 << attr)) { + switch (attr) { case FRAG_RESULT_DEPR: fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_POSITION; - outputMapping[i] = fs.num_outputs; + outputMapping[attr] = fs.num_outputs; break; case FRAG_RESULT_COLR: fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_COLOR; - outputMapping[i] = fs.num_outputs; + outputMapping[attr] = fs.num_outputs; break; default: assert(0); diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index c1bcdbe5a1..2add3dcfef 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -59,7 +59,7 @@ st_translate_vertex_shader(struct st_context *st, { struct pipe_shader_state vs; const struct cso_vertex_shader *cso; - GLuint i; + GLuint attr; memset(&vs, 0, sizeof(vs)); @@ -67,11 +67,11 @@ st_translate_vertex_shader(struct st_context *st, * Determine number of inputs, the mappings between VERT_ATTRIB_x * and TGSI generic input indexes, plus input attrib semantic info. */ - for (i = 0; i < MAX_VERTEX_PROGRAM_ATTRIBS; i++) { - if (stvp->Base.Base.InputsRead & (1 << i)) { - stvp->input_to_index[i] = vs.num_inputs; - stvp->index_to_input[vs.num_inputs] = i; - switch (i) { + for (attr = 0; attr < MAX_VERTEX_PROGRAM_ATTRIBS; attr++) { + if (stvp->Base.Base.InputsRead & (1 << attr)) { + stvp->input_to_index[attr] = vs.num_inputs; + stvp->index_to_input[vs.num_inputs] = attr; + switch (attr) { case VERT_ATTRIB_POS: vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_POSITION; vs.input_semantic_index[vs.num_inputs] = 0; @@ -100,7 +100,7 @@ st_translate_vertex_shader(struct st_context *st, case VERT_ATTRIB_TEX6: case VERT_ATTRIB_TEX7: vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_TEXCOORD; - vs.input_semantic_index[vs.num_inputs] = i - VERT_ATTRIB_TEX0; + vs.input_semantic_index[vs.num_inputs] = attr - VERT_ATTRIB_TEX0; break; case VERT_ATTRIB_GENERIC0: case VERT_ATTRIB_GENERIC1: @@ -110,9 +110,9 @@ st_translate_vertex_shader(struct st_context *st, case VERT_ATTRIB_GENERIC5: case VERT_ATTRIB_GENERIC6: case VERT_ATTRIB_GENERIC7: - assert(i < VERT_ATTRIB_MAX); + assert(attr < VERT_ATTRIB_MAX); vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[vs.num_inputs] = i - VERT_ATTRIB_GENERIC0; + vs.input_semantic_index[vs.num_inputs] = attr - VERT_ATTRIB_GENERIC0; break; default: assert(0); @@ -125,12 +125,12 @@ st_translate_vertex_shader(struct st_context *st, * Determine number of outputs, the register mapping and * the semantic information for each vertex output/result. */ - for (i = 0; i < VERT_RESULT_MAX; i++) { - if (stvp->Base.Base.OutputsWritten & (1 << i)) { + for (attr = 0; attr < VERT_RESULT_MAX; attr++) { + if (stvp->Base.Base.OutputsWritten & (1 << attr)) { /* put this attrib in the next available slot */ - st->vertex_attrib_to_slot[i] = vs.num_outputs; + st->vertex_attrib_to_slot[attr] = vs.num_outputs; - switch (i) { + switch (attr) { case VERT_RESULT_HPOS: vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_POSITION; vs.output_semantic_index[vs.num_outputs] = 0; @@ -165,14 +165,14 @@ st_translate_vertex_shader(struct st_context *st, case VERT_RESULT_TEX6: case VERT_RESULT_TEX7: vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_TEXCOORD; - vs.output_semantic_index[vs.num_outputs] = i - VERT_RESULT_TEX0; + vs.output_semantic_index[vs.num_outputs] = attr - VERT_RESULT_TEX0; break; case VERT_RESULT_VAR0: /* fall-through */ default: - assert(i - VERT_RESULT_VAR0 < MAX_VARYING); + assert(attr - VERT_RESULT_VAR0 < MAX_VARYING); vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[vs.num_outputs] = i - VERT_RESULT_VAR0; + vs.output_semantic_index[vs.num_outputs] = attr - VERT_RESULT_VAR0; } vs.num_outputs++; -- cgit v1.2.3 From e9259ad0d7f674a1f2f9156ba0baeedc0e7d0d54 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 11:20:29 -0600 Subject: Get rid of TGSI_SEMANTIC_TEXCOORD and just use TGSI_SEMANTIC_GENERIC. --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 3 ++- src/mesa/pipe/i915simple/i915_state_derived.c | 3 ++- src/mesa/pipe/softpipe/sp_state_derived.c | 7 +------ src/mesa/pipe/tgsi/exec/tgsi_token.h | 5 ++--- src/mesa/state_tracker/st_atom_fs.c | 2 +- src/mesa/state_tracker/st_atom_vs.c | 4 ++-- 6 files changed, 10 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index f8266f6adb..24f21f6633 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -178,7 +178,8 @@ src_vector(struct i915_fp_compile *p, src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); src = swizzle(src, W, W, W, W); break; - case TGSI_SEMANTIC_TEXCOORD: + case TGSI_SEMANTIC_GENERIC: + /* usually a texcoord */ src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + sem_ind, D0_CHANNEL_ALL); break; default: diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index e825971199..30a11f1ef2 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -74,7 +74,8 @@ static void calculate_vertex_layout( struct i915_context *i915 ) vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; } break; - case TGSI_SEMANTIC_TEXCOORD: + case TGSI_SEMANTIC_GENERIC: + /* usually a texcoord */ { const uint unit = fs->input_semantic_index[i]; uint hwtc; diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 66e5cc48f6..0e78209e30 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -99,13 +99,8 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) #endif softpipe->psize_slot = i; /*case TGSI_SEMANTIC_TEXCOORD:*/ - case TGSI_SEMANTIC_TEXCOORD: - /* unit = fs->input_semantic_index[i] */ - draw_emit_vertex_attr(vinfo, - FORMAT_4F, INTERP_PERSPECTIVE); - softpipe->need_w = TRUE; - break; case TGSI_SEMANTIC_GENERIC: + /* this includes texcoords and varying vars */ draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); softpipe->need_w = TRUE; break; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h index 2d468b801e..e6f884c5cf 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -107,9 +107,8 @@ struct tgsi_declaration_interpolation #define TGSI_SEMANTIC_COLOR 1 #define TGSI_SEMANTIC_BCOLOR 2 /**< back-face color */ #define TGSI_SEMANTIC_FOG 3 -#define TGSI_SEMANTIC_TEXCOORD 4 -#define TGSI_SEMANTIC_GENERIC 5 -#define TGSI_SEMANTIC_COUNT 6 /**< number of semantic values */ +#define TGSI_SEMANTIC_GENERIC 4 +#define TGSI_SEMANTIC_COUNT 5 /**< number of semantic values */ struct tgsi_declaration_semantic { diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index dd9d2476ba..20042fc6dc 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -102,7 +102,7 @@ st_translate_fragment_shader(struct st_context *st, case FRAG_ATTRIB_TEX5: case FRAG_ATTRIB_TEX6: case FRAG_ATTRIB_TEX7: - fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_TEXCOORD; + fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_GENERIC; fs.input_semantic_index[fs.num_inputs] = attr - FRAG_ATTRIB_TEX0; interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; break; diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 2add3dcfef..ef3ed802d7 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -99,7 +99,7 @@ st_translate_vertex_shader(struct st_context *st, case VERT_ATTRIB_TEX5: case VERT_ATTRIB_TEX6: case VERT_ATTRIB_TEX7: - vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_TEXCOORD; + vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; vs.input_semantic_index[vs.num_inputs] = attr - VERT_ATTRIB_TEX0; break; case VERT_ATTRIB_GENERIC0: @@ -164,7 +164,7 @@ st_translate_vertex_shader(struct st_context *st, case VERT_RESULT_TEX5: case VERT_RESULT_TEX6: case VERT_RESULT_TEX7: - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_TEXCOORD; + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_GENERIC; vs.output_semantic_index[vs.num_outputs] = attr - VERT_RESULT_TEX0; break; case VERT_RESULT_VAR0: -- cgit v1.2.3 From 11996050883ade4439a4a1e3192a02ba6f6aaa14 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 11:22:34 -0600 Subject: remove SEMANTIC_TEXCOORD --- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index 315354198a..e246f21c6f 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -205,7 +205,6 @@ static const char *TGSI_SEMANTICS[] = "SEMANTIC_COLOR", "SEMANTIC_BCOLOR", "SEMANTIC_FOG", - "SEMANTIC_TEXCOORD", "SEMANTIC_GENERIC," }; @@ -215,7 +214,6 @@ static const char *TGSI_SEMANTICS_SHORT[] = "COLOR", "BCOLOR", "FOG", - "TEXCOORD", "GENERIC", }; -- cgit v1.2.3 From 1ce50d8bd9a3bc64552c55a235a947de9e6f776c Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 11:34:20 -0600 Subject: use VERT_ATTRIB_MAX for loop limit --- src/mesa/state_tracker/st_atom_vs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index ef3ed802d7..e4e9902e65 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -67,7 +67,7 @@ st_translate_vertex_shader(struct st_context *st, * Determine number of inputs, the mappings between VERT_ATTRIB_x * and TGSI generic input indexes, plus input attrib semantic info. */ - for (attr = 0; attr < MAX_VERTEX_PROGRAM_ATTRIBS; attr++) { + for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { if (stvp->Base.Base.InputsRead & (1 << attr)) { stvp->input_to_index[attr] = vs.num_inputs; stvp->index_to_input[vs.num_inputs] = attr; -- cgit v1.2.3 From a3a8887e5742bc18476deafcf45f220de89c1c1d Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 11:34:28 -0600 Subject: comments --- src/mesa/state_tracker/st_atom_fs.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 20042fc6dc..1e7886a469 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -71,6 +71,9 @@ st_translate_fragment_shader(struct st_context *st, memset(&fs, 0, sizeof(fs)); + /* + * Convert Mesa program inputs to TGSI input register semantics. + */ for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) { if (inputsRead & (1 << attr)) { inputMapping[attr] = fs.num_inputs; @@ -119,7 +122,7 @@ st_translate_fragment_shader(struct st_context *st, } /* - * Outputs + * Semantics for outputs */ for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { if (stfp->Base.Base.OutputsWritten & (1 << attr)) { -- cgit v1.2.3 From 440a5c7c24e65dd19689ba2d2ce231a7ea7e4704 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 11:54:57 -0600 Subject: remove debug printfs --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index b1cf122695..d0d97ab0f8 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -67,7 +67,6 @@ map_register_file_index( * If a program uses 3 input attribs, the mapped indexes will be 1,2,3. * etc. */ - printf("Map %d input %d to %d\n", processor, index, mapped_index); return inputMapping[index]; case TGSI_FILE_OUTPUT: @@ -90,7 +89,6 @@ map_register_file_index( /* vertex output slots are tightly packed, find mapped pos */ /* mapped_index = VERT_RESULT_x */ mapped_index = outputMapping[index]; - printf("Map VP output from %d to %d\n", index, mapped_index); } break; -- cgit v1.2.3 From 46f606e9715145218331a04f0d1f66fb9f8531d6 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 11:55:28 -0600 Subject: reenable some clear code that was temporarily disabled --- src/mesa/state_tracker/st_cb_clear.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index c53446d588..639e0ceb40 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -493,15 +493,12 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) assert(strb->surface->format); -#if 0 if (ctx->Scissor.Enabled || (isDS && ctx->DrawBuffer->Visual.stencilBits > 0)) { /* scissoring or we have a combined depth/stencil buffer */ clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_FALSE); } - else -#endif - { + else { /* simple clear of whole buffer */ GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); -- cgit v1.2.3 From da45890818ab5ae94592208e3581b5c2febaa6b4 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 12:06:08 -0600 Subject: Fix up some point size breakage. Start on fogcoord too. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 4 +- src/mesa/pipe/softpipe/sp_state_derived.c | 62 +++++++++++++++++++++---------- src/mesa/pipe/tgsi/exec/tgsi_token.h | 5 ++- src/mesa/state_tracker/st_atom_fs.c | 4 +- src/mesa/state_tracker/st_atom_vs.c | 10 +++++ src/mesa/state_tracker/st_cb_rasterpos.c | 3 ++ 6 files changed, 63 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index f0f9cf54bd..6d63cc9412 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -907,15 +907,13 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const struct vertex_header *v0 = prim->v[0]; const int sizeAttr = setup->softpipe->psize_slot; const float halfSize - = sizeAttr ? (0.5f * v0->data[sizeAttr][0]) + = sizeAttr > 0 ? (0.5f * v0->data[sizeAttr][0]) : (0.5f * setup->softpipe->rasterizer->point_size); const boolean round = setup->softpipe->rasterizer->point_smooth; const float x = v0->data[0][0]; /* Note: data[0] is always position */ const float y = v0->data[0][1]; unsigned slot, j; - assert(sizeAttr >= 0); - /* For points, all interpolants are constant-valued. * However, for point sprites, we'll need to setup texcoords appropriately. * XXX: which coefficients are the texcoords??? diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 0e78209e30..35ba32cd81 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -43,6 +43,7 @@ */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { + const struct pipe_shader_state *vs = softpipe->vs; const struct pipe_shader_state *fs = softpipe->fs; const interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; @@ -58,47 +59,68 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) softpipe->need_z = FALSE; softpipe->need_w = FALSE; + if (fs->input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { + /* Need Z if depth test is enabled or the fragment program uses the + * fragment position (XYZW). + */ + softpipe->need_z = TRUE; + softpipe->need_w = TRUE; + } + softpipe->psize_slot = -1; /* always emit vertex pos */ - /* TODO - Figure out if we need to do perspective divide, etc. */ draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR); - for (i = 0; i < fs->num_inputs; i++) { - switch (fs->input_semantic_name[i]) { + /* + * XXX I think we need to reconcile the vertex shader outputs with + * the fragment shader inputs here to make sure the slots line up. + * Might just be getting lucky so far. + * Or maybe do that in the state tracker? + */ + + for (i = 0; i < vs->num_outputs; i++) { + switch (vs->output_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: - /* Need Z if depth test is enabled or the fragment program uses the - * fragment position (XYZW). + /* vertex programs always emit position, but might not be + * needed for fragment progs. */ - softpipe->need_z = TRUE; - softpipe->need_w = TRUE; + /* no-op */ break; + case TGSI_SEMANTIC_COLOR: if (fs->input_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(vinfo, - FORMAT_4F, colorInterp); + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + } + else { + assert(fs->input_semantic_index[i] == 1); + front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + } + break; + + case TGSI_SEMANTIC_BCOLOR: + if (fs->input_semantic_index[i] == 0) { + back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); } else { assert(fs->input_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(vinfo, - FORMAT_4F, colorInterp); + back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); } break; + case TGSI_SEMANTIC_FOG: - draw_emit_vertex_attr(vinfo, - FORMAT_1F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE); break; -#if 0 + case TGSI_SEMANTIC_PSIZE: /* XXX only emit if drawing points or front/back polygon mode * is point mode */ - draw_emit_vertex_attr(vinfo, - FORMAT_4F, INTERP_CONSTANT); - break; -#endif + draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT); softpipe->psize_slot = i; - /*case TGSI_SEMANTIC_TEXCOORD:*/ + break; + case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); @@ -112,6 +134,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) softpipe->nr_frag_attrs = vinfo->num_attribs; +#if 0 /* Additional attributes required for setup: Just twosided * lighting. Edgeflag is dealt with specially by setting bits in * the vertex header. @@ -124,6 +147,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) back1 = draw_emit_vertex_attr(vinfo, FORMAT_OMIT, colorInterp); } } +#endif /* If the attributes have changed, tell the draw module about * the new vertex layout. diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h index e6f884c5cf..1d99a50dde 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -107,8 +107,9 @@ struct tgsi_declaration_interpolation #define TGSI_SEMANTIC_COLOR 1 #define TGSI_SEMANTIC_BCOLOR 2 /**< back-face color */ #define TGSI_SEMANTIC_FOG 3 -#define TGSI_SEMANTIC_GENERIC 4 -#define TGSI_SEMANTIC_COUNT 5 /**< number of semantic values */ +#define TGSI_SEMANTIC_PSIZE 4 +#define TGSI_SEMANTIC_GENERIC 5 +#define TGSI_SEMANTIC_COUNT 6 /**< number of semantic values */ struct tgsi_declaration_semantic { diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index 1e7886a469..f8a1dc83cf 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -95,7 +95,9 @@ st_translate_fragment_shader(struct st_context *st, interpMode[fs.num_inputs] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_FOGC: - assert(0); + fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_FOG; + fs.input_semantic_index[fs.num_inputs] = 0; + interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; break; case FRAG_ATTRIB_TEX0: case FRAG_ATTRIB_TEX1: diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index e4e9902e65..ae51e9764b 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -91,6 +91,10 @@ st_translate_vertex_shader(struct st_context *st, vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; vs.input_semantic_index[vs.num_inputs] = 1; break; + case VERT_ATTRIB_FOG: + vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_FOG; + vs.input_semantic_index[vs.num_inputs] = 0; + break; case VERT_ATTRIB_TEX0: case VERT_ATTRIB_TEX1: case VERT_ATTRIB_TEX2: @@ -152,7 +156,13 @@ st_translate_vertex_shader(struct st_context *st, vs.output_semantic_index[vs.num_outputs] = 1; break; case VERT_RESULT_FOGC: + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_FOG; + vs.output_semantic_index[vs.num_outputs] = 0; + break; case VERT_RESULT_PSIZ: + vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_PSIZE; + vs.output_semantic_index[vs.num_outputs] = 0; + break; case VERT_RESULT_EDGE: assert(0); break; diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 56c98916ed..2311bddc65 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -265,6 +265,9 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) float *buf_map; struct pipe_feedback_buffer fb_buf; + /** XXX TEMPORARILY DISABLE */ + return; + st_validate_state(ctx->st); /* setup vertex buffers */ -- cgit v1.2.3 From 890125d3adc4c5011e9241210c44d40dfdd6c4eb Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 12:21:05 -0600 Subject: fix some post-transform vertex layout issues (bcolors, psize emitted last --- src/mesa/pipe/softpipe/sp_state_derived.c | 32 +++++++++++++++---------------- 1 file changed, 16 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 35ba32cd81..6c6e798069 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -48,6 +48,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) const interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; + boolean emitBack0 = FALSE, emitBack1 = FALSE, emitPsize = FALSE; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; uint i; @@ -101,11 +102,11 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) case TGSI_SEMANTIC_BCOLOR: if (fs->input_semantic_index[i] == 0) { - back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + emitBack0 = TRUE; } else { assert(fs->input_semantic_index[i] == 1); - back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + emitBack1 = TRUE; } break; @@ -117,8 +118,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) /* XXX only emit if drawing points or front/back polygon mode * is point mode */ - draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT); - softpipe->psize_slot = i; + emitPsize = TRUE; break; case TGSI_SEMANTIC_GENERIC: @@ -134,20 +134,20 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) softpipe->nr_frag_attrs = vinfo->num_attribs; -#if 0 - /* Additional attributes required for setup: Just twosided - * lighting. Edgeflag is dealt with specially by setting bits in - * the vertex header. + /* We want these after all other attribs since they won't get passed + * to the fragment shader. All prior vertex output attribs should match + * up 1:1 with the fragment shader inputs. */ - if (softpipe->rasterizer->light_twoside) { - if (front0) { - back0 = draw_emit_vertex_attr(vinfo, FORMAT_OMIT, colorInterp); - } - if (back0) { - back1 = draw_emit_vertex_attr(vinfo, FORMAT_OMIT, colorInterp); - } + if (emitBack0) { + back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + } + if (emitBack1) { + back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + } + if (emitPsize) { + softpipe->psize_slot + = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT); } -#endif /* If the attributes have changed, tell the draw module about * the new vertex layout. -- cgit v1.2.3 From ac394414fcf244186557fbfa18e7962fa05dc39a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 12:36:59 -0600 Subject: added PSIZE --- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index e246f21c6f..2532d5c831 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -205,6 +205,7 @@ static const char *TGSI_SEMANTICS[] = "SEMANTIC_COLOR", "SEMANTIC_BCOLOR", "SEMANTIC_FOG", + "SEMANTIC_PSIZE", "SEMANTIC_GENERIC," }; @@ -214,6 +215,7 @@ static const char *TGSI_SEMANTICS_SHORT[] = "COLOR", "BCOLOR", "FOG", + "PSIZE", "GENERIC", }; -- cgit v1.2.3 From 1ae7373ac520c277d9aa7b915526bb0e1567b9a5 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 12:37:13 -0600 Subject: invert draw_vertex_cache_check_space() --- src/mesa/pipe/draw/draw_prim.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index a90d1df7f9..e82e48b90b 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -107,7 +107,7 @@ static struct prim_header *get_queued_prim( struct draw_context *draw, unsigned nr_verts ) { if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH || - draw_vertex_cache_check_space( draw, nr_verts )) + !draw_vertex_cache_check_space( draw, nr_verts )) { draw_flush( draw ); } -- cgit v1.2.3 From c0afc92f00e5153a168fb6df518b7a2e6b3b9406 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 24 Sep 2007 05:43:07 -0400 Subject: In here we actually do want to assign it. Fix compilation. --- src/mesa/state_tracker/st_atom_vs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index ae51e9764b..9050f30cab 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -216,7 +216,7 @@ st_translate_vertex_shader(struct st_context *st, tgsi_emit_sse2( stvp->tokens, &stvp->sse2_program ); if (!cso->state.executable) - cso->state.executable = (void *) x86_get_func( &stvp->sse2_program ); + ((struct cso_vertex_shader*)cso)->state.executable = (void *) x86_get_func( &stvp->sse2_program ); #endif stvp->dirty = 0; -- cgit v1.2.3 From c0dd02219d47f45ce469abbef8044431f6d85d0a Mon Sep 17 00:00:00 2001 From: michal Date: Mon, 24 Sep 2007 12:32:26 +0100 Subject: Enable SSE2 for FS. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 59 ++++---- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 248 ++++++++++++++++++++++++++++++++- src/mesa/pipe/tgsi/exec/tgsi_sse2.h | 5 + src/mesa/state_tracker/st_atom_fs.c | 9 ++ src/mesa/state_tracker/st_cb_program.c | 21 ++- src/mesa/state_tracker/st_program.h | 6 +- 6 files changed, 310 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 13d7eac4f2..bff5525b0f 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -56,30 +56,22 @@ quad_shade_stage(struct quad_stage *qs) } +#if !defined(XSTDCALL) +#if defined(WIN32) +#define XSTDCALL __stdcall +#else +#define XSTDCALL +#endif +#endif - -/** - * Compute quad's attribute values by linear interpolation. - * - * Push into the fp: - * - * INPUT[attr] = MAD COEF_A0[attr], COEF_DADX[attr], INPUT_WPOS.xxxx - * INPUT[attr] = MAD INPUT[attr], COEF_DADY[attr], INPUT_WPOS.yyyy - */ -static INLINE void -linterp(const struct tgsi_interp_coef *coef, - struct tgsi_exec_vector *pos, uint ch) -{ - uint j; - for (j = 0; j < QUAD_SIZE; j++) { - const float x = pos->xyzw[0].f[j]; - const float y = pos->xyzw[1].f[j]; - pos->xyzw[ch].f[j] = (coef->a0[ch] + - coef->dadx[ch] * x + - coef->dady[ch] * y); - } -} - +#if defined(USE_X86_ASM) || defined(SLANG_X86) +typedef void (XSTDCALL *sse2_function)( + const struct tgsi_exec_vector *input, + struct tgsi_exec_vector *output, + float (*constant)[4], + struct tgsi_exec_vector *temporary, + const struct tgsi_interp_coef *coef ); +#endif /* This should be done by the fragment shader execution unit (code * generated from the decl instructions). Do it here for now. @@ -127,12 +119,23 @@ shade_quad( machine.Inputs[0].xyzw[1].f[2] = fy + 1.0f; machine.Inputs[0].xyzw[1].f[3] = fy + 1.0f; - /* interp Z */ - linterp(&quad->coef[0], &machine.Inputs[0], 2); /* Z */ - linterp(&quad->coef[0], &machine.Inputs[0], 3); /* 1/W */ - /* run shader */ - tgsi_exec_machine_run( &machine ); + if( softpipe->fs->executable != NULL ) { +#if defined(USE_X86_ASM) || defined(SLANG_X86) + sse2_function func = (sse2_function) softpipe->fs->executable; + func( + machine.Inputs, + machine.Outputs, + machine.Consts, + machine.Temps, + machine.InterpCoefs ); +#else + assert( 0 ); +#endif + } + else { + tgsi_exec_machine_run( &machine ); + } /* store result color (always in output[1]) */ memcpy( diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index d89bb19970..cf5e386ddf 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -114,6 +114,23 @@ get_temp( (vec * 4 + chan) * 16 ); } +static struct x86_reg +get_coef_base( void ) +{ + return get_output_base(); +} + +static struct x86_reg +get_coef( + unsigned vec, + unsigned chan, + unsigned member ) +{ + return x86_make_disp( + get_coef_base(), + ((vec * 3 + member) * 4 + chan) * 4 ); +} + static struct x86_reg get_addr( unsigned vec, @@ -143,7 +160,7 @@ emit_const( } static void -emit_input( +emit_inputf( struct x86_function *func, unsigned xmm, unsigned vec, @@ -155,6 +172,19 @@ emit_input( get_input( vec, chan ) ); } +static void +emit_inputs( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movups( + func, + get_input( vec, chan ), + make_xmm( xmm ) ); +} + static void emit_output( struct x86_function *func, @@ -182,7 +212,7 @@ emit_tempf( } static void -emit_temps ( +emit_temps( struct x86_function *func, unsigned xmm, unsigned vec, @@ -194,6 +224,70 @@ emit_temps ( make_xmm( xmm ) ); } +static void +emit_coef( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan, + unsigned member ) +{ + sse_movss( + func, + make_xmm( xmm ), + get_coef( vec, chan, member ) ); + sse_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + +static void +emit_coef_a0( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 0 ); +} + +static void +emit_coef_dadx( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 1 ); +} + +static void +emit_coef_dady( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 2 ); +} + static void emit_addrf( struct x86_function *func, @@ -676,7 +770,7 @@ emit_fetch( break; case TGSI_FILE_INPUT: - emit_input( + emit_inputf( func, xmm, reg->SrcRegister.Index, @@ -1658,6 +1752,76 @@ emit_instruction( } } +static void +emit_declaration( + struct x86_function *func, + struct tgsi_full_declaration *decl ) +{ + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + unsigned i, j; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if( first == 0 ) { + mask &= ~TGSI_WRITEMASK_XY; + if( mask == TGSI_WRITEMASK_NONE ) { + first++; + } + } + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + switch( decl->Interpolation.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + emit_coef_a0( func, 0, i, j ); + emit_inputs( func, 0, i, j ); + break; + + case TGSI_INTERPOLATE_LINEAR: + emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_coef_dadx( func, 1, i, j ); + emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_coef_dady( func, 3, i, j ); + emit_mul( func, 0, 1 ); /* x * dadx */ + emit_coef_a0( func, 4, i, j ); + emit_mul( func, 2, 3 ); /* y * dady */ + emit_add( func, 0, 4 ); /* x * dadx + a0 */ + emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ + emit_inputs( func, 0, i, j ); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_coef_dadx( func, 1, i, j ); + emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_coef_dady( func, 3, i, j ); + emit_mul( func, 0, 1 ); /* x * dadx */ + emit_inputf( func, 4, 0, TGSI_SWIZZLE_W ); + emit_coef_a0( func, 5, i, j ); + emit_rcp( func, 4, 4 ); /* 1.0 / w */ + emit_mul( func, 2, 3 ); /* y * dady */ + emit_add( func, 0, 5 ); /* x * dadx + a0 */ + emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ + emit_mul( func, 0, 4 ); /* (x * dadx + y * dady + a0) / w */ + emit_inputs( func, 0, i, j ); + break; + + default: + assert( 0 ); + } + } + } + } + } +} + unsigned tgsi_emit_sse2( struct tgsi_token *tokens, @@ -1715,4 +1879,82 @@ tgsi_emit_sse2( return 1; } +/** + * Fragment shaders are responsible for interpolating shader inputs. Because on + * x86 we have only 4 GP registers, and here we have 5 shader arguments (input, + * output, const, temp and coef), the code is split into two phases -- + * DECLARATION and INSTRUCTION phase. + * GP register holding the output argument is aliased with the coeff argument, + * as outputs are not needed in the DECLARATION phase. + */ +unsigned +tgsi_emit_sse2_fs( + struct tgsi_token *tokens, + struct x86_function *func ) +{ + struct tgsi_parse_context parse; + boolean instruction_phase = FALSE; + + func->csr = func->store; + + /* DECLARATION phase, do not load output argument. */ + x86_mov( + func, + get_input_base(), + get_argument( 0 ) ); + x86_mov( + func, + get_const_base(), + get_argument( 2 ) ); + x86_mov( + func, + get_temp_base(), + get_argument( 3 ) ); + x86_mov( + func, + get_coef_base(), + get_argument( 4 ) ); + + tgsi_parse_init( &parse, tokens ); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + emit_declaration( + func, + &parse.FullToken.FullDeclaration ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + if( !instruction_phase ) { + /* INSTRUCTION phase, overwrite coeff with output. */ + instruction_phase = TRUE; + x86_mov( + func, + get_output_base(), + get_argument( 1 ) ); + } + emit_instruction( + func, + &parse.FullToken.FullInstruction ); + break; + + default: + assert( 0 ); + } + } + + tgsi_parse_free( &parse ); + +#ifdef WIN32 + x86_retw( func, 16 ); +#else + x86_ret( func ); +#endif + + return 1; +} + #endif diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h index 4a39658484..9bee371766 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h @@ -13,6 +13,11 @@ tgsi_emit_sse2( struct tgsi_token *tokens, struct x86_function *function ); +unsigned +tgsi_emit_sse2_fs( + struct tgsi_token *tokens, + struct x86_function *function ); + #if defined __cplusplus } // extern "C" #endif // defined __cplusplus diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index f8a1dc83cf..ef5b941c17 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -36,6 +36,7 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "pipe/tgsi/exec/tgsi_core.h" #include "pipe/tgsi/exec/tgsi_dump.h" #include "st_context.h" @@ -163,6 +164,14 @@ st_translate_fragment_shader(struct st_context *st, if (TGSI_DEBUG) tgsi_dump( stfp->tokens, 0/*TGSI_DUMP_VERBOSE*/ ); +#if defined(USE_X86_ASM) || defined(SLANG_X86) + if (stfp->sse2_program.csr == stfp->sse2_program.store) + tgsi_emit_sse2_fs( stfp->tokens, &stfp->sse2_program ); + + if (!cso->state.executable) + ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); +#endif + stfp->dirty = 0; return cso; diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index b9c19bdd3e..9f46f9e93f 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -99,13 +99,16 @@ static struct gl_program *st_new_program( GLcontext *ctx, } case GL_FRAGMENT_PROGRAM_ARB: - case GL_FRAGMENT_PROGRAM_NV: - { + case GL_FRAGMENT_PROGRAM_NV: { struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program); prog->id = program_id++; prog->dirty = 1; +#if defined(USE_X86_ASM) || defined(SLANG_X86) + x86_init_func( &prog->sse2_program ); +#endif + return _mesa_init_fragment_program( ctx, &prog->Base, target, @@ -121,8 +124,7 @@ static void st_delete_program( GLcontext *ctx, struct gl_program *prog ) { switch( prog->Target ) { - case GL_VERTEX_PROGRAM_ARB: - { + case GL_VERTEX_PROGRAM_ARB: { #if defined(USE_X86_ASM) || defined(SLANG_X86) struct st_vertex_program *p = (struct st_vertex_program *) prog; @@ -130,7 +132,14 @@ static void st_delete_program( GLcontext *ctx, #endif break; } + case GL_FRAGMENT_PROGRAM_ARB: { +#if defined(USE_X86_ASM) || defined(SLANG_X86) + struct st_fragment_program *p = (struct st_fragment_program *) prog; + x86_release_func( &p->sse2_program ); +#endif + break; + } } _mesa_delete_program( ctx, prog ); } @@ -156,7 +165,7 @@ static void st_program_string_notify( GLcontext *ctx, if (prog == &ctx->FragmentProgram._Current->Base) st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - p->id = program_id++; + p->id = program_id++; p->param_state = p->Base.Base.Parameters->StateFlags; } else if (target == GL_VERTEX_PROGRAM_ARB) { @@ -165,7 +174,7 @@ static void st_program_string_notify( GLcontext *ctx, if (prog == &ctx->VertexProgram._Current->Base) st->dirty.st |= ST_NEW_VERTEX_PROGRAM; - p->id = program_id++; + p->id = program_id++; p->param_state = p->Base.Base.Parameters->StateFlags; /* Also tell tnl about it: diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index c21e27628e..419afa4e78 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -49,10 +49,14 @@ struct st_fragment_program GLboolean error; /* If program is malformed for any reason. */ GLuint id; /**< String id, for tracking ProgramStringNotify changes. */ - + /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; GLboolean dirty; +#if defined(USE_X86_ASM) || defined(SLANG_X86) + struct x86_function sse2_program; +#endif + /** Pointer to the corresponding cached shader */ const struct cso_fragment_shader *fs; -- cgit v1.2.3 From 0940059f8b2dceb817e9b1778e78baf1ccc4c438 Mon Sep 17 00:00:00 2001 From: keithw Date: Tue, 25 Sep 2007 12:08:16 +0100 Subject: test against -1 for front/back attrib --- src/mesa/pipe/draw/draw_twoside.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 9547f2a67b..26d26e659e 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -78,12 +78,12 @@ static struct vertex_header *copy_bfc( struct twoside_stage *twoside, struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); const struct draw_context *draw = twoside->stage.draw; - if (draw->attrib_front0) { - assert(draw->attrib_back0); + if (draw->attrib_front0 != -1) { + assert(draw->attrib_back0 != -1); copy_attrib(draw->attrib_front0, draw->attrib_back0, tmp); } - if (draw->attrib_front1) { - assert(draw->attrib_back1); + if (draw->attrib_front1 != -1) { + assert(draw->attrib_back1 != -1); copy_attrib(draw->attrib_front1, draw->attrib_back1, tmp); } -- cgit v1.2.3 From 45b37dc078130b64e110d1a965960c60c56f0e60 Mon Sep 17 00:00:00 2001 From: keithw Date: Tue, 25 Sep 2007 13:16:25 +0100 Subject: turn off shader debug --- src/mesa/state_tracker/st_atom_fs.c | 2 +- src/mesa/state_tracker/st_atom_vs.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index ef5b941c17..dd4cdf0855 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -45,7 +45,7 @@ #include "st_program.h" -#define TGSI_DEBUG 1 +#define TGSI_DEBUG 0 /** diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index 9050f30cab..a6c0d159d4 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -46,7 +46,7 @@ #include "st_program.h" -#define TGSI_DEBUG 1 +#define TGSI_DEBUG 0 /** -- cgit v1.2.3 From a37e0daeb97bb36ba10038b12a909e22e08b52c4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 25 Sep 2007 13:20:53 +0100 Subject: First attempt at building vertex buffers post-clip. Build a buffer of contigous vertices and indices at the backend of our software transformation/clipping path. This will become the mechanism for emitting buffers of vertices to rasterization hardware. This is similar to but not the same as the post-transform vertex cache. In particular, these vertices are subject to clipping, culling, poly offset, etc. The vertices emitted will all be used by hardware. TODOs include the actual transformation to hardware vertex formats, moving this out of softpipe to somewhere more useful and allowing >1 primitive to share the generated VB. --- src/mesa/pipe/draw/draw_clip.c | 1 + src/mesa/pipe/draw/draw_prim.c | 9 +- src/mesa/pipe/draw/draw_private.h | 6 +- src/mesa/pipe/draw/draw_vertex_cache.c | 20 ++- src/mesa/pipe/draw/draw_vertex_shader.c | 2 + src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_context.c | 14 +- src/mesa/pipe/softpipe/sp_context.h | 2 + src/mesa/pipe/softpipe/sp_prim_setup.c | 96 +++++++++- src/mesa/pipe/softpipe/sp_prim_setup.h | 34 ++++ src/mesa/pipe/softpipe/sp_prim_vbuf.c | 305 ++++++++++++++++++++++++++++++++ 11 files changed, 476 insertions(+), 14 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_prim_vbuf.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index e2af69e048..3ccc408bc0 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -109,6 +109,7 @@ static void interp( const struct clipper *clip, dst->clipmask = 0; dst->edgeflag = 0; dst->pad = 0; + dst->vertex_id = 0; } /* Clip coordinates: interpolate normally diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index e82e48b90b..be2f987b9a 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -106,9 +106,12 @@ void draw_flush( struct draw_context *draw ) static struct prim_header *get_queued_prim( struct draw_context *draw, unsigned nr_verts ) { - if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH || - !draw_vertex_cache_check_space( draw, nr_verts )) - { + if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH) { +// fprintf(stderr, "p"); + draw_flush( draw ); + } + else if (!draw_vertex_cache_check_space( draw, nr_verts )) { +// fprintf(stderr, "v"); draw_flush( draw ); } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 1285f3200c..04d38c4e0c 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -54,7 +54,8 @@ struct vertex_header { unsigned clipmask:12; unsigned edgeflag:1; - unsigned pad:19; + unsigned pad:3; + unsigned vertex_id:16; float clip[4]; @@ -230,6 +231,8 @@ extern int draw_vertex_cache_check_space( struct draw_context *draw, extern void draw_vertex_cache_validate( struct draw_context *draw ); extern void draw_vertex_cache_invalidate( struct draw_context *draw ); extern void draw_vertex_cache_unreference( struct draw_context *draw ); +extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); + extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); @@ -255,6 +258,7 @@ dup_vert( struct draw_stage *stage, { struct vertex_header *tmp = stage->tmp[idx]; memcpy(tmp, vert, stage->draw->vertex_info.size * sizeof(float) ); + tmp->vertex_id = ~0; return tmp; } diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index f1b0cb14bd..a226798123 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -46,6 +46,8 @@ void draw_vertex_cache_invalidate( struct draw_context *draw ) for (i = 0; i < Elements( draw->vcache.idx ); i++) draw->vcache.idx[i] = ~0; + +// fprintf(stderr, "x\n"); } @@ -63,10 +65,14 @@ static struct vertex_header *get_vertex( struct draw_context *draw, /* If slot is in use, use the overflow area: */ - if (draw->vcache.referenced & (1 << slot)) + if (draw->vcache.referenced & (1 << slot)) { +// fprintf(stderr, "o"); slot = VCACHE_SIZE + draw->vcache.overflow++; - else + } + else { +// fprintf(stderr, "."); draw->vcache.referenced |= (1 << slot); /* slot now in use */ + } draw->vcache.idx[slot] = i; @@ -79,7 +85,10 @@ static struct vertex_header *get_vertex( struct draw_context *draw, /* Need to set the vertex's edge flag here. If we're being called * by do_ef_triangle(), that function needs edge flag info! */ + draw->vcache.vertex[slot]->clipmask = 0; draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ + draw->vcache.vertex[slot]->pad = 0; + draw->vcache.vertex[slot]->vertex_id = ~0; } return draw->vcache.vertex[slot]; @@ -110,6 +119,13 @@ static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, } +void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) +{ + unsigned i; + + for (i = 0; i < Elements(draw->vcache.vertex); i++) + draw->vcache.vertex[i]->vertex_id = ~0; +} void draw_vertex_cache_validate( struct draw_context *draw ) { diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index ef0399c46e..fe4f124dd2 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -189,6 +189,8 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) { unsigned i, j; +// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); + /* run vertex shader on vertex cache entries, four per invokation */ for (i = 0; i < draw->vs.queue_nr; i += 4) { struct vertex_header *dests[4]; diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 0335f56b99..63eb6b5761 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -10,6 +10,7 @@ DRIVER_SOURCES = \ sp_context.c \ sp_draw_arrays.c \ sp_prim_setup.c \ + sp_prim_vbuf.c \ sp_quad.c \ sp_quad_alpha_test.c \ sp_quad_blend.c \ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index a6ab45314c..e415966dae 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -334,7 +334,19 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, */ softpipe->draw = draw_create(); assert(softpipe->draw); - draw_set_rasterize_stage(softpipe->draw, sp_draw_render_stage(softpipe)); + softpipe->setup = sp_draw_render_stage(softpipe); + + if (getenv("SP_VBUF")) { + softpipe->vbuf = sp_draw_vbuf_stage(softpipe->draw, + &softpipe->pipe, + sp_vbuf_setup_draw); + + draw_set_rasterize_stage(softpipe->draw, softpipe->vbuf); + } + else { + draw_set_rasterize_stage(softpipe->draw, softpipe->setup); + } + sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 7b1518cfac..c0a681f3d6 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -146,6 +146,8 @@ struct softpipe_context { /** The primitive drawing context */ struct draw_context *draw; + struct draw_stage *setup; + struct draw_stage *vbuf; struct pipe_surface *cbuf; /**< current color buffer (one of cbufs) */ }; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 6d63cc9412..2e27d00acf 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -41,7 +41,7 @@ #include "pipe/draw/draw_vertex.h" #include "pipe/p_util.h" - +#define DEBUG_VERTS 0 /** * Triangle edge info @@ -241,15 +241,15 @@ static void flush_spans( struct setup_stage *setup ) setup->span.right[1] = 0; } -#if 0 +#if DEBUG_VERTS static void print_vertex(const struct setup_stage *setup, const struct vertex_header *v) { int i; - printf("Vertex:\n"); + fprintf(stderr, "Vertex: (%p)\n", v); for (i = 0; i < setup->quad.nr_attrs; i++) { - printf(" %d: %f %f %f\n", i, - v->data[i][0], v->data[i][1], v->data[i][2]); + fprintf(stderr, " %d: %f %f %f %f\n", i, + v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); } } #endif @@ -261,8 +261,8 @@ static boolean setup_sort_vertices( struct setup_stage *setup, const struct vertex_header *v1 = prim->v[1]; const struct vertex_header *v2 = prim->v[2]; -#if 0 - printf("Triangle:\n"); +#if DEBUG_VERTS + fprintf(stderr, "Triangle:\n"); print_vertex(setup, v0); print_vertex(setup, v1); print_vertex(setup, v2); @@ -1096,3 +1096,85 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) return &setup->stage; } + + +/* Recalculate det. This is only used in the test harness below: + */ +static void calc_det( struct prim_header *header ) +{ + /* Window coords: */ + const float *v0 = header->v[0]->data[0]; + const float *v1 = header->v[1]->data[0]; + const float *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0] - v2[0]; + const float ey = v0[1] - v2[1]; + const float fx = v1[0] - v2[0]; + const float fy = v1[1] - v2[1]; + + /* det = cross(e,f).z */ + header->det = ex * fy - ey * fx; +} + + + +/* Test harness - feed vertex buffer back into prim pipeline. + * + * The big issue at this point is that reset_stipple doesn't make it + * through the interface. Probably need to split primitives at reset + * stipple, perhaps using the ~0 index marker. + */ +void sp_vbuf_setup_draw( struct pipe_context *pipe, + unsigned primitive, + const ushort *elements, + unsigned nr_elements, + const void *vertex_buffer, + unsigned nr_vertices ) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + struct setup_stage *setup = setup_stage( softpipe->setup ); + struct prim_header prim; + unsigned vertex_size = setup->stage.draw->vertex_info.size * sizeof(float); + int i, j; + + prim.det = 0; + prim.reset_line_stipple = 0; + prim.edgeflags = 0; + prim.pad = 0; + + setup->stage.begin( &setup->stage ); + + switch (primitive) { + case PIPE_PRIM_TRIANGLES: + for (i = 0; i < nr_elements; i += 3) { + for (j = 0; j < 3; j++) + prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + + elements[i+j] * vertex_size); + + calc_det(&prim); + setup->stage.tri( &setup->stage, &prim ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i < nr_elements; i += 2) { + for (j = 0; j < 2; j++) + prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + + elements[i+j] * vertex_size); + + setup->stage.line( &setup->stage, &prim ); + } + break; + + case PIPE_PRIM_POINTS: + for (i = 0; i < nr_elements; i += 2) { + prim.v[i] = (struct vertex_header *)((char *)vertex_buffer + + elements[i] * vertex_size); + setup->stage.point( &setup->stage, &prim ); + } + break; + } + + setup->stage.end( &setup->stage ); +} diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h index dece42fe45..598394f73e 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.h +++ b/src/mesa/pipe/softpipe/sp_prim_setup.h @@ -44,5 +44,39 @@ struct softpipe_context; extern struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ); +/* A special stage to gather the stream of triangles, lines, points + * together and reconstruct vertex buffers for hardware upload. + * + * First attempt, work in progress. + * + * TODO: + * - separate out vertex buffer building and primitive emit, ie >1 draw per vb. + * - tell vbuf stage how to build hw vertices directly + * - pass vbuf stage a buffer pointer for direct emit to agp/vram. + */ +typedef void (*vbuf_draw_func)( struct pipe_context *pipe, + unsigned prim, + const ushort *elements, + unsigned nr_elements, + const void *vertex_buffer, + unsigned nr_vertices ); + + +extern struct draw_stage *sp_draw_vbuf_stage( struct draw_context *draw_context, + struct pipe_context *pipe, + vbuf_draw_func draw ); + + + +/* Test harness + */ +void sp_vbuf_setup_draw( struct pipe_context *pipe, + unsigned prim, + const ushort *elements, + unsigned nr_elements, + const void *vertex_buffer, + unsigned nr_vertices ); + + #endif /* SP_PRIM_SETUP_H */ diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c new file mode 100644 index 0000000000..235903ca6d --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -0,0 +1,305 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Build post-transformation, post-clipping vertex buffers and element + * lists by hooking into the end of the primitive pipeline and + * manipulating the vertex_id field in the vertex headers. + * + * Keith Whitwell + */ + + +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "sp_prim_setup.h" +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_vertex.h" +#include "pipe/p_util.h" + +static void vbuf_flush_elements( struct draw_stage *stage ); + + +#define VBUF_SIZE (64*1024) +#define IBUF_SIZE (16*1024) + + +/** + * Vertex buffer emit stage. + */ +struct vbuf_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct draw_context *draw_context; + struct pipe_context *pipe; + vbuf_draw_func draw; + + /* Vertices are passed in as an array of floats making up each + * attribute in turn. Will eventually convert to hardware format + * in this stage. + */ + char *vertex_map; + char *vertex_ptr; + unsigned vertex_size; + unsigned nr_vertices; + + unsigned max_vertices; + + ushort *element_map; + unsigned nr_elements; + + unsigned prim; + +}; + +/** + * Basically a cast wrapper. + */ +static INLINE struct vbuf_stage *vbuf_stage( struct draw_stage *stage ) +{ + return (struct vbuf_stage *)stage; +} + + + + +static boolean overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) +{ + unsigned long used = (char *)ptr - (char *)map; + return (used + bytes) > bufsz; +} + + +static boolean check_space( struct vbuf_stage *vbuf ) +{ + if (overflow( vbuf->vertex_map, + vbuf->vertex_ptr, + 4 * vbuf->vertex_size, + VBUF_SIZE )) + return FALSE; + + + if (vbuf->nr_elements + 4 > IBUF_SIZE / sizeof(ushort) ) + return FALSE; + + return TRUE; +} + + +static void emit_vertex( struct vbuf_stage *vbuf, + struct vertex_header *vertex ) +{ + fprintf(stderr, "emit vertex %d to %p\n", + vbuf->nr_vertices, vbuf->vertex_ptr); + + vertex->vertex_id = vbuf->nr_vertices++; + + //vbuf->emit_vertex( vbuf->vertex_ptr, vertex ); + memcpy(vbuf->vertex_ptr, vertex, vbuf->vertex_size); + + vbuf->vertex_ptr += vbuf->vertex_size; +} + + + +/** + * + */ +static void vbuf_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + if (!check_space( vbuf )) + vbuf_flush_elements( stage ); + + for (i = 0; i < 3; i++) { + if (prim->v[i]->vertex_id == 0xffff) + emit_vertex( vbuf, prim->v[i] ); + + vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; + } +} + + +static void vbuf_line(struct draw_stage *stage, + struct prim_header *prim) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + if (!check_space( vbuf )) + vbuf_flush_elements( stage ); + + for (i = 0; i < 2; i++) { + if (prim->v[i]->vertex_id == 0xffff) + emit_vertex( vbuf, prim->v[i] ); + + vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; + } +} + + +static void vbuf_point(struct draw_stage *stage, + struct prim_header *prim) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + if (!check_space( vbuf )) + vbuf_flush_elements( stage ); + + if (prim->v[0]->vertex_id == 0xffff) + emit_vertex( vbuf, prim->v[0] ); + + vbuf->element_map[vbuf->nr_elements++] = prim->v[0]->vertex_id; +} + + +static void vbuf_first_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_elements( stage ); + stage->tri = vbuf_tri; + stage->tri( stage, prim ); + vbuf->prim = PIPE_PRIM_TRIANGLES; +} + +static void vbuf_first_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_elements( stage ); + stage->line = vbuf_line; + stage->line( stage, prim ); + vbuf->prim = PIPE_PRIM_LINES; +} + +static void vbuf_first_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_elements( stage ); + stage->point = vbuf_point; + stage->point( stage, prim ); + vbuf->prim = PIPE_PRIM_POINTS; +} + + + +static void vbuf_flush_elements( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + if (vbuf->nr_elements) { + fprintf(stderr, "%s (%d elts)\n", __FUNCTION__, vbuf->nr_elements); + + /* Draw now or add to list of primitives??? + */ + vbuf->draw( vbuf->pipe, + vbuf->prim, + vbuf->element_map, + vbuf->nr_elements, + vbuf->vertex_map, + (vbuf->vertex_ptr - vbuf->vertex_map) / vbuf->vertex_size ); + + vbuf->nr_elements = 0; + + vbuf->vertex_ptr = vbuf->vertex_map; + vbuf->nr_vertices = 0; + + /* Reset vertex ids? Actually, want to not do that unless our + * vertex buffer is full. Would like separate + * flush-on-index-full and flush-on-vb-full, but may raise + * issues uploading vertices if the hardware wants to flush when + * we flush. + */ + draw_vertex_cache_reset_vertex_ids( vbuf->draw_context ); + } + + stage->tri = vbuf_first_tri; + stage->line = vbuf_first_line; + stage->point = vbuf_first_point; +} + + +static void vbuf_begin( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage(stage); + + vbuf->vertex_size = vbuf->draw_context->vertex_info.size * sizeof(float); +} + + +static void vbuf_end( struct draw_stage *stage ) +{ + /* Overkill. + */ + vbuf_flush_elements( stage ); +} + + +static void reset_stipple_counter( struct draw_stage *stage ) +{ + /* XXX: This doesn't work. + */ +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *sp_draw_vbuf_stage( struct draw_context *draw_context, + struct pipe_context *pipe, + vbuf_draw_func draw ) +{ + struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); + + vbuf->stage.begin = vbuf_begin; + vbuf->stage.point = vbuf_first_point; + vbuf->stage.line = vbuf_first_line; + vbuf->stage.tri = vbuf_first_tri; + vbuf->stage.end = vbuf_end; + vbuf->stage.reset_stipple_counter = reset_stipple_counter; + + vbuf->pipe = pipe; + vbuf->draw = draw; + vbuf->draw_context = draw_context; + + vbuf->element_map = malloc( IBUF_SIZE ); + vbuf->vertex_map = malloc( VBUF_SIZE ); + + vbuf->vertex_ptr = vbuf->vertex_map; + + + return &vbuf->stage; +} -- cgit v1.2.3 From 608914aeb068a0d59da5c239fb5ff28b522e064d Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 12:39:26 -0600 Subject: remove unused lookup array --- src/mesa/pipe/draw/draw_twoside.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 26d26e659e..7f0cb98b30 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -36,9 +36,6 @@ struct twoside_stage { struct draw_stage stage; float sign; /**< +1 or -1 */ -#if 0 - const unsigned *lookup; /**< vertex attribute positions */ -#endif }; @@ -164,9 +161,5 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) twoside->stage.end = twoside_end; twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; -#if 0 - twoside->lookup = draw->vertex_info.attrib_to_slot; -#endif - return &twoside->stage; } -- cgit v1.2.3 From ff08c7b032fb17e4ffcda7497488dc5e263a7d6e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Sep 2007 12:41:24 -0600 Subject: test for presence of both front and back color vertex slots --- src/mesa/pipe/draw/draw_twoside.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 7f0cb98b30..d987b00598 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -68,6 +68,9 @@ static INLINE void copy_attrib( unsigned attr_dst, } +/** + * Copy back color(s) to front color(s). + */ static struct vertex_header *copy_bfc( struct twoside_stage *twoside, const struct vertex_header *v, unsigned idx ) @@ -75,12 +78,12 @@ static struct vertex_header *copy_bfc( struct twoside_stage *twoside, struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); const struct draw_context *draw = twoside->stage.draw; - if (draw->attrib_front0 != -1) { - assert(draw->attrib_back0 != -1); + if (draw->attrib_front0 && draw->attrib_back0) { + assert(draw->attrib_back0); copy_attrib(draw->attrib_front0, draw->attrib_back0, tmp); } - if (draw->attrib_front1 != -1) { - assert(draw->attrib_back1 != -1); + if (draw->attrib_front1 && draw->attrib_back1) { + assert(draw->attrib_back1); copy_attrib(draw->attrib_front1, draw->attrib_back1, tmp); } -- cgit v1.2.3 From 07ef3e42451daa26fa5426026bcc4c89729644e0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 14:21:18 -0600 Subject: vertex/fragment shader state and linking --- src/mesa/state_tracker/st_atom_shader.c | 422 ++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_atom_shader.h | 9 + 2 files changed, 431 insertions(+) create mode 100644 src/mesa/state_tracker/st_atom_shader.c create mode 100644 src/mesa/state_tracker/st_atom_shader.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c new file mode 100644 index 0000000000..f3bb9effde --- /dev/null +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -0,0 +1,422 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * State validation for vertex/fragment shaders. + * Note that we have to delay most vertex/fragment shader translation + * until rendering time since the linkage between the vertex outputs and + * fragment inputs can vary depending on the pairing of shaders. + * + * Authors: + * Brian Paul + */ + + + +#include "main/imports.h" +#include "main/mtypes.h" + +#include "pipe/p_context.h" +#include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "pipe/tgsi/exec/tgsi_core.h" + +#include "st_context.h" +#include "st_cache.h" +#include "st_atom.h" +#include "st_program.h" +#include "st_atom_shader.h" + + + +/** + * Structure to describe a (vertex program, fragment program) pair + * which is linked together (used together to render something). This + * linkage basically servers the same purpose as the OpenGL Shading + * Language linker, but also applies to ARB programs and Mesa's + * fixed-function-generated programs. + * + * More background: + * + * The translation from Mesa programs to TGSI programs depends on the + * linkage between the vertex program and the fragment program. This is + * because we tightly pack the inputs and outputs of shaders into + * consecutive "slots". + * + * Suppose an app uses one vertex program "VP" (outputting pos, color and tex0) + * and two fragment programs: + * FP1: uses tex0 input only (input slot 0) + * FP2: uses color input only (input slot 0) + * + * When VP is used with FP1 we want VP.output[2] to match FP1.input[0], but + * when VP is used with FP2 we want VP.output[1] to match FP1.input[0]. + * + * We don't want to re-translate the vertex and/or fragment programs + * each time the VP/FP bindings/linkings change. The solution is this + * structure which stores the translated TGSI shaders on a per-linkage + * basis. + * + */ +struct linked_program_pair +{ + struct st_vertex_program *vprog; /**< never changes */ + struct st_fragment_program *fprog; /**< never changes */ + + struct tgsi_token vs_tokens[ST_FP_MAX_TOKENS]; + struct tgsi_token fs_tokens[ST_FP_MAX_TOKENS]; + + const struct cso_vertex_shader *vs; + const struct cso_fragment_shader *fs; + + GLuint vertSerialNo, fragSerialNo; + + /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ + GLuint vp_input_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; + /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ + GLuint vp_index_to_input[MAX_VERTEX_PROGRAM_ATTRIBS]; + + GLuint vp_result_to_slot[VERT_RESULT_MAX]; + + struct linked_program_pair *next; +}; + + +/** XXX temporary - use some kind of hash table instead */ +static struct linked_program_pair *Pairs = NULL; + + +static void +find_and_remove(struct gl_program *prog) +{ + struct linked_program_pair *pair, *prev = NULL, *next; + for (pair = Pairs; pair; pair = next) { + next = pair->next; + if (pair->vprog == (struct st_vertex_program *) prog || + pair->fprog == (struct st_fragment_program *) prog) { + /* unlink */ + if (prev) + prev->next = next; + else + Pairs = next; + /* delete pair->vs */ + /* delete pair->fs */ + free(pair); + } + else { + prev = pair; + } + } +} + + +/** + * Delete any known program pairs that use the given vertex program. + */ +void +st_remove_vertex_program(struct st_context *st, struct st_vertex_program *stvp) +{ + find_and_remove(&stvp->Base.Base); +} + + +/** + * Delete any known program pairs that use the given fragment program. + */ +void +st_remove_fragment_program(struct st_context *st, + struct st_fragment_program *stfp) +{ + find_and_remove(&stfp->Base.Base); +} + + + +/** + * Given a vertex program output attribute, return the corresponding + * fragment program input attribute. + * \return -1 for vertex outputs that have no corresponding fragment input + */ +static GLint +vp_out_to_fp_in(GLuint vertResult) +{ + if (vertResult >= VERT_RESULT_TEX0 && + vertResult < VERT_RESULT_TEX0 + MAX_TEXTURE_COORD_UNITS) + return FRAG_ATTRIB_TEX0 + (vertResult - VERT_RESULT_TEX0); + + if (vertResult >= VERT_RESULT_VAR0 && + vertResult < VERT_RESULT_VAR0 + MAX_VARYING) + return FRAG_ATTRIB_VAR0 + (vertResult - VERT_RESULT_VAR0); + + switch (vertResult) { + case VERT_RESULT_HPOS: + return FRAG_ATTRIB_WPOS; + case VERT_RESULT_COL0: + return FRAG_ATTRIB_COL0; + case VERT_RESULT_COL1: + return FRAG_ATTRIB_COL1; + case VERT_RESULT_FOGC: + return FRAG_ATTRIB_FOGC; + default: + /* Back-face colors, edge flags, etc */ + return -1; + } +} + + +/** + * Examine the outputs written by a vertex program and the inputs read + * by a fragment program to determine which match up and where they + * should be mapped into the generic shader output/input slots. + * \param vert_output_map returns the vertex output register mapping + * \param frag_input_map returns the fragment input register mapping + */ +static GLuint +link_outputs_to_inputs(GLbitfield outputsWritten, + GLbitfield inputsRead, + GLuint vert_output_map[], + GLuint frag_input_map[]) +{ + static const GLuint UNUSED = ~0; + GLint vert_slot_to_attr[50], frag_slot_to_attr[50]; + GLuint outAttr, inAttr; + GLuint numIn = 0, dummySlot; + + for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) { + if (inputsRead & (1 << inAttr)) { + frag_input_map[inAttr] = numIn; + frag_slot_to_attr[numIn] = inAttr; + numIn++; + } + else { + frag_input_map[inAttr] = UNUSED; + } + } + + for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { + if (outputsWritten & (1 << outAttr)) { + /* see if the frag prog wants this vert output */ + GLint fpIn = vp_out_to_fp_in(outAttr); + + if (fpIn >= 0) { + GLuint frag_slot = frag_input_map[fpIn]; + vert_output_map[outAttr] = frag_slot; + vert_slot_to_attr[frag_slot] = outAttr; + } + else { + vert_output_map[outAttr] = UNUSED; + } + } + else { + vert_output_map[outAttr] = UNUSED; + } + } + + /* + * We'll map all unused vertex program outputs to this slot. + * We'll also map all undefined fragment program inputs to this slot. + */ + dummySlot = numIn; + + /* Map vert program outputs that aren't used to the dummy slot */ + for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { + if (outputsWritten & (1 << outAttr)) { + if (vert_output_map[outAttr] == UNUSED) + vert_output_map[outAttr] = dummySlot; + } + } + + /* Map frag program inputs that aren't defined to the dummy slot */ + for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) { + if (inputsRead & (1 << inAttr)) { + if (frag_input_map[inAttr] == UNUSED) + frag_input_map[inAttr] = dummySlot; + } + } + +#if 0 + printf("vOut W slot\n"); + for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { + printf("%4d %c %4d\n", outAttr, + " *"[(outputsWritten >> outAttr) & 1], + vert_output_map[outAttr]); + } + printf("vIn R slot\n"); + for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) { + printf("%3d %c %4d\n", inAttr, + " *"[(inputsRead >> inAttr) & 1], + frag_input_map[inAttr]); + } +#endif + + return numIn; +} + + +static struct linked_program_pair * +lookup_program_pair(struct st_context *st, + struct st_vertex_program *vprog, + struct st_fragment_program *fprog) +{ + struct linked_program_pair *pair; + + /* search */ + for (pair = Pairs; pair; pair = pair->next) { + if (pair->vprog == vprog && pair->fprog == fprog) { + /* found it */ + break; + } + } + + /* + * Examine the outputs of the vertex shader and the inputs of the + * fragment shader to determine how to match both to a common set + * of slots. + */ + if (!pair) { + pair = CALLOC_STRUCT(linked_program_pair); + if (pair) { + pair->vprog = vprog; + pair->fprog = fprog; + } + } + + return pair; +} + + +static void +link_shaders(struct st_context *st, struct linked_program_pair *pair) +{ + struct st_vertex_program *vprog = pair->vprog; + struct st_fragment_program *fprog = pair->fprog; + + assert(vprog); + assert(fprog); + + if (pair->vertSerialNo != vprog->serialNo || + pair->fragSerialNo != fprog->serialNo) { + /* re-link and re-translate */ + GLuint vert_output_mapping[VERT_RESULT_MAX]; + GLuint frag_input_mapping[FRAG_ATTRIB_MAX]; + + link_outputs_to_inputs(vprog->Base.Base.OutputsWritten, + fprog->Base.Base.InputsRead | FRAG_BIT_WPOS, + vert_output_mapping, + frag_input_mapping); + + /* xlate vp to vs + vs tokens */ + st_translate_vertex_program(st, vprog, + vert_output_mapping, + pair->vs_tokens, ST_FP_MAX_TOKENS); + + pair->vprog = vprog; + /* temp hacks */ + pair->vs = vprog->vs; + vprog->vs = NULL; + + + /* xlate fp to fs + fs tokens */ + st_translate_fragment_program(st, fprog, + frag_input_mapping, + pair->fs_tokens, ST_FP_MAX_TOKENS); + pair->fprog = fprog; + /* temp hacks */ + pair->fs = fprog->fs; + fprog->fs = NULL; + + /* save pair */ + pair->next = Pairs; + Pairs = pair; + + pair->vertSerialNo = vprog->serialNo; + pair->fragSerialNo = fprog->serialNo; + } +} + + +static void +update_linkage( struct st_context *st ) +{ + struct linked_program_pair *pair; + struct st_vertex_program *stvp; + struct st_fragment_program *stfp; + + /* find active shader and params -- Should be covered by + * ST_NEW_VERTEX_PROGRAM + */ + if (st->ctx->Shader.CurrentProgram && + st->ctx->Shader.CurrentProgram->LinkStatus && + st->ctx->Shader.CurrentProgram->VertexProgram) { + struct gl_vertex_program *f + = st->ctx->Shader.CurrentProgram->VertexProgram; + stvp = st_vertex_program(f); + } + else { + assert(st->ctx->VertexProgram._Current); + stvp = st_vertex_program(st->ctx->VertexProgram._Current); + } + + + if (st->ctx->Shader.CurrentProgram && + st->ctx->Shader.CurrentProgram->LinkStatus && + st->ctx->Shader.CurrentProgram->FragmentProgram) { + struct gl_fragment_program *f + = st->ctx->Shader.CurrentProgram->FragmentProgram; + stfp = st_fragment_program(f); + } + else { + assert(st->ctx->FragmentProgram._Current); + stfp = st_fragment_program(st->ctx->FragmentProgram._Current); + } + + + pair = lookup_program_pair(st, stvp, stfp); + assert(pair); + link_shaders(st, pair); + + + /* Bind the vertex program and TGSI shader */ + st->vp = stvp; + st->state.vs = pair->vs; + st->pipe->bind_vs_state(st->pipe, st->state.vs->data); + + /* Bind the fragment program and TGSI shader */ + st->fp = stfp; + st->state.fs = pair->fs; + st->pipe->bind_fs_state(st->pipe, st->state.fs->data); + + st->vertex_result_to_slot = pair->vp_result_to_slot; +} + + +const struct st_tracked_state st_update_shader = { + .name = "st_update_shader", + .dirty = { + .mesa = 0, + .st = ST_NEW_LINKAGE + }, + .update = update_linkage +}; diff --git a/src/mesa/state_tracker/st_atom_shader.h b/src/mesa/state_tracker/st_atom_shader.h new file mode 100644 index 0000000000..48e5348d73 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_shader.h @@ -0,0 +1,9 @@ + + + +extern void +st_remove_vertex_program(struct st_context *, struct st_vertex_program *); + +extern void +st_remove_fragment_program(struct st_context *, struct st_fragment_program *); + -- cgit v1.2.3 From f9ed2fdaace0d4d7f091a4423a8638945e920b0d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 14:22:13 -0600 Subject: translate Mesa programs to TGSI programs (formerly in st_atom_[fv]s.c) --- src/mesa/state_tracker/st_program.c | 413 ++++++++++++++++++++++++++++++++++++ 1 file changed, 413 insertions(+) create mode 100644 src/mesa/state_tracker/st_program.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c new file mode 100644 index 0000000000..88c02e5b75 --- /dev/null +++ b/src/mesa/state_tracker/st_program.c @@ -0,0 +1,413 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "shader/prog_parameter.h" +#include "shader/prog_print.h" +#include "tnl/t_vp_build.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "pipe/tgsi/exec/tgsi_core.h" + +#include "st_context.h" +#include "st_cache.h" +#include "st_atom.h" +#include "st_program.h" + + +#define TGSI_DEBUG 1 + + +/** + * Translate a Mesa vertex shader into a TGSI shader. + * \param outputMapping to map vertex program output registers to TGSI + * output slots + * \param tokensOut destination for TGSI tokens + * \return pointer to cached pipe_shader object. + */ +const struct cso_vertex_shader * +st_translate_vertex_program(struct st_context *st, + struct st_vertex_program *stvp, + const GLuint outputMapping[], + struct tgsi_token *tokensOut, + GLuint maxTokens) +{ + GLuint defaultOutputMapping[VERT_RESULT_MAX]; + struct pipe_shader_state vs; + const struct cso_vertex_shader *cso; + GLuint attr, i; + + memset(&vs, 0, sizeof(vs)); + + /* + * Determine number of inputs, the mappings between VERT_ATTRIB_x + * and TGSI generic input indexes, plus input attrib semantic info. + */ + for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { + if (stvp->Base.Base.InputsRead & (1 << attr)) { + const GLuint slot = vs.num_inputs; + + vs.num_inputs++; + + stvp->input_to_index[attr] = slot; + stvp->index_to_input[slot] = attr; + + switch (attr) { + case VERT_ATTRIB_POS: + vs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + vs.input_semantic_index[slot] = 0; + break; + case VERT_ATTRIB_WEIGHT: + /* fall-through */ + case VERT_ATTRIB_NORMAL: + /* just label as a generic */ + vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.input_semantic_index[slot] = 0; + break; + case VERT_ATTRIB_COLOR0: + vs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs.input_semantic_index[slot] = 0; + break; + case VERT_ATTRIB_COLOR1: + vs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs.input_semantic_index[slot] = 1; + break; + case VERT_ATTRIB_FOG: + vs.input_semantic_name[slot] = TGSI_SEMANTIC_FOG; + vs.input_semantic_index[slot] = 0; + break; + case VERT_ATTRIB_TEX0: + case VERT_ATTRIB_TEX1: + case VERT_ATTRIB_TEX2: + case VERT_ATTRIB_TEX3: + case VERT_ATTRIB_TEX4: + case VERT_ATTRIB_TEX5: + case VERT_ATTRIB_TEX6: + case VERT_ATTRIB_TEX7: + vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.input_semantic_index[slot] = attr - VERT_ATTRIB_TEX0; + break; + case VERT_ATTRIB_GENERIC0: + case VERT_ATTRIB_GENERIC1: + case VERT_ATTRIB_GENERIC2: + case VERT_ATTRIB_GENERIC3: + case VERT_ATTRIB_GENERIC4: + case VERT_ATTRIB_GENERIC5: + case VERT_ATTRIB_GENERIC6: + case VERT_ATTRIB_GENERIC7: + assert(attr < VERT_ATTRIB_MAX); + vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.input_semantic_index[slot] = attr - VERT_ATTRIB_GENERIC0; + break; + default: + assert(0); + } + } + } + + /* initialize output semantics to defaults */ + for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) { + vs.output_semantic_name[i] = TGSI_SEMANTIC_GENERIC; + vs.output_semantic_index[i] = 0; + } + + /* + * Determine number of outputs, the (default) output register + * mapping and the semantic information for each output. + */ + for (attr = 0; attr < VERT_RESULT_MAX; attr++) { + if (stvp->Base.Base.OutputsWritten & (1 << attr)) { + GLuint slot; + + if (outputMapping) { + slot = outputMapping[attr]; + assert(slot != ~0); + } + else { + slot = vs.num_outputs; + vs.num_outputs++; + defaultOutputMapping[attr] = slot; + } + + /* + printf("Output %u -> slot %u\n", attr, slot); + */ + + switch (attr) { + case VERT_RESULT_HPOS: + vs.output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + vs.output_semantic_index[slot] = 0; + break; + case VERT_RESULT_COL0: + vs.output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs.output_semantic_index[slot] = 0; + break; + case VERT_RESULT_COL1: + vs.output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs.output_semantic_index[slot] = 1; + break; + case VERT_RESULT_BFC0: + vs.output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + vs.output_semantic_index[slot] = 0; + break; + case VERT_RESULT_BFC1: + vs.output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + vs.output_semantic_index[slot] = 1; + break; + case VERT_RESULT_FOGC: + vs.output_semantic_name[slot] = TGSI_SEMANTIC_FOG; + vs.output_semantic_index[slot] = 0; + break; + case VERT_RESULT_PSIZ: + vs.output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; + vs.output_semantic_index[slot] = 0; + break; + case VERT_RESULT_EDGE: + assert(0); + break; + case VERT_RESULT_TEX0: + case VERT_RESULT_TEX1: + case VERT_RESULT_TEX2: + case VERT_RESULT_TEX3: + case VERT_RESULT_TEX4: + case VERT_RESULT_TEX5: + case VERT_RESULT_TEX6: + case VERT_RESULT_TEX7: + vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.output_semantic_index[slot] = attr - VERT_RESULT_TEX0; + break; + case VERT_RESULT_VAR0: + /* fall-through */ + default: + assert(attr - VERT_RESULT_VAR0 < MAX_VARYING); + vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.output_semantic_index[slot] = attr - VERT_RESULT_VAR0; + } + } + } + + + if (outputMapping) { + /* find max output slot referenced to compute vs.num_outputs */ + GLuint maxSlot = 0; + for (attr = 0; attr < VERT_RESULT_MAX; attr++) { + if (outputMapping[attr] != ~0 && outputMapping[attr] > maxSlot) + maxSlot = outputMapping[attr]; + } + vs.num_outputs = maxSlot + 1; + } + else { + outputMapping = defaultOutputMapping; + } + + /* XXX: fix static allocation of tokens: + */ + tgsi_mesa_compile_vp_program( &stvp->Base, + /* inputs */ + vs.num_inputs, + stvp->input_to_index, + vs.input_semantic_name, + vs.input_semantic_index, + /* outputs */ + vs.num_outputs, + outputMapping, + vs.output_semantic_name, + vs.output_semantic_index, + /* tokenized result */ + tokensOut, maxTokens); + + vs.tokens = tokensOut; + cso = st_cached_vs_state(st, &vs); + stvp->vs = cso; + + if (TGSI_DEBUG) + tgsi_dump( tokensOut, 0 ); + +#if defined(USE_X86_ASM) || defined(SLANG_X86) + if (stvp->sse2_program.csr == stvp->sse2_program.store) + tgsi_emit_sse2( tokensOut, &stvp->sse2_program ); + + if (!cso->state.executable) + ((struct cso_vertex_shader*)cso)->state.executable = (void *) x86_get_func( &stvp->sse2_program ); +#endif + + return cso; +} + + + +/** + * Translate a Mesa fragment shader into a TGSI shader. + * \param inputMapping to map fragment program input registers to TGSI + * input slots + * \param tokensOut destination for TGSI tokens + * \return pointer to cached pipe_shader object. + */ +const struct cso_fragment_shader * +st_translate_fragment_program(struct st_context *st, + struct st_fragment_program *stfp, + const GLuint inputMapping[], + struct tgsi_token *tokensOut, + GLuint maxTokens) +{ + GLuint outputMapping[FRAG_RESULT_MAX]; + GLuint defaultInputMapping[FRAG_ATTRIB_MAX]; + struct pipe_shader_state fs; + const struct cso_fragment_shader *cso; + GLuint interpMode[16]; /* XXX size? */ + GLuint attr; + GLbitfield inputsRead = stfp->Base.Base.InputsRead; + + /* Check if all fragment programs need the fragment position (in order + * to do perspective-corrected interpolation). + */ + /* XXX temporary! */ + if (st->pipe->get_param(st->pipe, PIPE_PARAM_FS_NEEDS_POS)) + inputsRead |= FRAG_BIT_WPOS; + + memset(&fs, 0, sizeof(fs)); + + /* + * Convert Mesa program inputs to TGSI input register semantics. + */ + for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) { + if (inputsRead & (1 << attr)) { + const GLuint slot = fs.num_inputs; + + fs.num_inputs++; + + defaultInputMapping[attr] = slot; + + switch (attr) { + case FRAG_ATTRIB_WPOS: + fs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + fs.input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_CONSTANT; + break; + case FRAG_ATTRIB_COL0: + fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + fs.input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_LINEAR; + break; + case FRAG_ATTRIB_COL1: + fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + fs.input_semantic_index[slot] = 1; + interpMode[slot] = TGSI_INTERPOLATE_LINEAR; + break; + case FRAG_ATTRIB_FOGC: + fs.input_semantic_name[slot] = TGSI_SEMANTIC_FOG; + fs.input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; + break; + case FRAG_ATTRIB_TEX0: + case FRAG_ATTRIB_TEX1: + case FRAG_ATTRIB_TEX2: + case FRAG_ATTRIB_TEX3: + case FRAG_ATTRIB_TEX4: + case FRAG_ATTRIB_TEX5: + case FRAG_ATTRIB_TEX6: + case FRAG_ATTRIB_TEX7: + fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + fs.input_semantic_index[slot] = attr - FRAG_ATTRIB_TEX0; + interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; + break; + case FRAG_ATTRIB_VAR0: + /* fall-through */ + default: + fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + fs.input_semantic_index[slot] = attr - FRAG_ATTRIB_VAR0; + interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; + } + } + } + + /* + * Semantics for outputs + */ + for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { + if (stfp->Base.Base.OutputsWritten & (1 << attr)) { + switch (attr) { + case FRAG_RESULT_DEPR: + fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_POSITION; + outputMapping[attr] = fs.num_outputs; + break; + case FRAG_RESULT_COLR: + fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_COLOR; + outputMapping[attr] = fs.num_outputs; + break; + default: + assert(0); + } + fs.num_outputs++; + } + } + + if (!inputMapping) + inputMapping = defaultInputMapping; + + /* XXX: fix static allocation of tokens: + */ + tgsi_mesa_compile_fp_program( &stfp->Base, + /* inputs */ + fs.num_inputs, + inputMapping, + fs.input_semantic_name, + fs.input_semantic_index, + interpMode, + /* outputs */ + outputMapping, + /* tokenized result */ + tokensOut, maxTokens); + + + fs.tokens = tokensOut; + + cso = st_cached_fs_state(st, &fs); + stfp->fs = cso; + + if (TGSI_DEBUG) + tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); + +#if defined(USE_X86_ASM) || defined(SLANG_X86) + if (stfp->sse2_program.csr == stfp->sse2_program.store) + tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); + + if (!cso->state.executable) + ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); +#endif + + return cso; +} + -- cgit v1.2.3 From 40c543eb71368c646259afb87d5c76551f6b45b7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 14:29:11 -0600 Subject: Translate mesa vertex/fragment programs to TGSI programs at same time to do proper linking. Previously, programs were translated independently during validation. The problem is the translation to TGSI format, which packs shader input/outputs into continuous slots, depends on which vertex program is being paired with which fragment shader. Now, we look at the outputs of the vertex program in conjunction with the inputs of the fragment shader to be sure the attributes match up correctly. The new 'linked_program_pair' class keeps track of the associations between vertex and fragment shaders. It's also the place where the TGSI tokens are kept since they're no longer per-program state but per-linkage. Still a few loose ends, like implementing some kind of hash/lookup table for linked_program_pairs. --- src/mesa/pipe/draw/draw_feedback.c | 7 +- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 4 + src/mesa/sources | 4 +- src/mesa/state_tracker/st_atom.c | 3 +- src/mesa/state_tracker/st_atom.h | 3 +- src/mesa/state_tracker/st_atom_fs.c | 96 +++++++++++------- src/mesa/state_tracker/st_atom_vs.c | 156 +++++++++++++++++++----------- src/mesa/state_tracker/st_cb_clear.c | 6 +- src/mesa/state_tracker/st_cb_drawpixels.c | 6 +- src/mesa/state_tracker/st_cb_feedback.c | 4 +- src/mesa/state_tracker/st_cb_program.c | 88 ++++++++++------- src/mesa/state_tracker/st_cb_rasterpos.c | 61 ++++++------ src/mesa/state_tracker/st_context.h | 6 +- src/mesa/state_tracker/st_program.h | 27 ++++-- 14 files changed, 287 insertions(+), 184 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_feedback.c b/src/mesa/pipe/draw/draw_feedback.c index 3b8400233e..ee54db0ad5 100644 --- a/src/mesa/pipe/draw/draw_feedback.c +++ b/src/mesa/pipe/draw/draw_feedback.c @@ -78,11 +78,9 @@ feedback_vertex(struct draw_stage *stage, const struct vertex_header *vertex) * we can either address output buffer 0 (for interleaving) or * output buffer i (for non-interleaved). */ -#if 0 for (i = 0; i < feedback->num_attribs; i++) { - const uint attr = feedback->attrib[i]; - const uint slot = stage->draw->vertex_info.attrib_to_slot[attr]; - const float *src = attr ? vertex->data[slot] : vertex->clip; + const uint slot = feedback->attrib[i]; + const float *src = slot ? vertex->data[slot] : vertex->clip; const uint size = feedback->size[i]; float *dest = fs->dest[i * select]; @@ -104,7 +102,6 @@ feedback_vertex(struct draw_stage *stage, const struct vertex_header *vertex) } fs->dest[i * select] += size; } -#endif fs->num_vert_emitted++; } diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index d0d97ab0f8..fa27fd3cd0 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -476,6 +476,8 @@ make_input_decl( { struct tgsi_full_declaration decl; + assert(semantic_name < TGSI_SEMANTIC_COUNT); + decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_INPUT; decl.Declaration.Declare = TGSI_DECLARE_RANGE; @@ -500,6 +502,8 @@ make_output_decl( { struct tgsi_full_declaration decl; + assert(semantic_name < TGSI_SEMANTIC_COUNT); + decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_OUTPUT; decl.Declaration.Declare = TGSI_DECLARE_RANGE; diff --git a/src/mesa/sources b/src/mesa/sources index 0d4fdc15f4..985bd2dce6 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -200,14 +200,13 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_depth.c \ state_tracker/st_atom_fixedfunction.c \ state_tracker/st_atom_framebuffer.c \ - state_tracker/st_atom_fs.c \ state_tracker/st_atom_sampler.c \ state_tracker/st_atom_scissor.c \ + state_tracker/st_atom_shader.c \ state_tracker/st_atom_rasterizer.c \ state_tracker/st_atom_stipple.c \ state_tracker/st_atom_texture.c \ state_tracker/st_atom_viewport.c \ - state_tracker/st_atom_vs.c \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_flush.c \ @@ -224,6 +223,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_context.c \ state_tracker/st_draw.c \ state_tracker/st_format.c \ + state_tracker/st_program.c \ state_tracker/st_mipmap_tree.c SHADER_SOURCES = \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index fc339b91ed..326042cb34 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -50,8 +50,7 @@ static const struct st_tracked_state *atoms[] = &st_update_clip, &st_update_tnl, - &st_update_vs, - &st_update_fs, + &st_update_shader, &st_update_rasterizer, &st_update_polygon_stipple, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 6710c1a269..94cb7bee7a 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -49,8 +49,7 @@ const struct st_tracked_state st_update_clip; const struct st_tracked_state st_update_clear_color; const struct st_tracked_state st_update_depth_stencil; const struct st_tracked_state st_update_tnl; -const struct st_tracked_state st_update_fs; -const struct st_tracked_state st_update_vs; +const struct st_tracked_state st_update_shader; const struct st_tracked_state st_update_rasterizer; const struct st_tracked_state st_update_polygon_stipple; const struct st_tracked_state st_update_viewport; diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c index dd4cdf0855..28019858f7 100644 --- a/src/mesa/state_tracker/st_atom_fs.c +++ b/src/mesa/state_tracker/st_atom_fs.c @@ -24,11 +24,12 @@ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ + +/* + * Authors: + * Keith Whitwell + * Brian Paul + */ #include "shader/prog_parameter.h" @@ -50,14 +51,20 @@ /** * Translate a Mesa fragment shader into a TGSI shader. + * \param inputMapping to map fragment program input registers to TGSI + * input slots + * \param tokensOut destination for TGSI tokens * \return pointer to cached pipe_shader object. */ const struct cso_fragment_shader * -st_translate_fragment_shader(struct st_context *st, - struct st_fragment_program *stfp) +st_translate_fragment_shader(const struct st_context *st, + struct st_fragment_program *stfp, + const GLuint inputMapping[], + struct tgsi_token *tokensOut, + GLuint maxTokens) { GLuint outputMapping[FRAG_RESULT_MAX]; - GLuint inputMapping[PIPE_MAX_SHADER_INPUTS]; + GLuint defaultInputMapping[FRAG_ATTRIB_MAX]; struct pipe_shader_state fs; const struct cso_fragment_shader *cso; GLuint interpMode[16]; /* XXX size? */ @@ -67,6 +74,7 @@ st_translate_fragment_shader(struct st_context *st, /* Check if all fragment programs need the fragment position (in order * to do perspective-corrected interpolation). */ + /* XXX temporary! */ if (st->pipe->get_param(st->pipe, PIPE_PARAM_FS_NEEDS_POS)) inputsRead |= FRAG_BIT_WPOS; @@ -77,28 +85,32 @@ st_translate_fragment_shader(struct st_context *st, */ for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) { if (inputsRead & (1 << attr)) { - inputMapping[attr] = fs.num_inputs; + const GLuint slot = fs.num_inputs; + + fs.num_inputs++; + + defaultInputMapping[attr] = slot; switch (attr) { case FRAG_ATTRIB_WPOS: - fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_POSITION; - fs.input_semantic_index[fs.num_inputs] = 0; - interpMode[fs.num_inputs] = TGSI_INTERPOLATE_CONSTANT; + fs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + fs.input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_CONSTANT; break; case FRAG_ATTRIB_COL0: - fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_COLOR; - fs.input_semantic_index[fs.num_inputs] = 0; - interpMode[fs.num_inputs] = TGSI_INTERPOLATE_LINEAR; + fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + fs.input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_COL1: - fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_COLOR; - fs.input_semantic_index[fs.num_inputs] = 1; - interpMode[fs.num_inputs] = TGSI_INTERPOLATE_LINEAR; + fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + fs.input_semantic_index[slot] = 1; + interpMode[slot] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_FOGC: - fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_FOG; - fs.input_semantic_index[fs.num_inputs] = 0; - interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; + fs.input_semantic_name[slot] = TGSI_SEMANTIC_FOG; + fs.input_semantic_index[slot] = 0; + interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; case FRAG_ATTRIB_TEX0: case FRAG_ATTRIB_TEX1: @@ -108,19 +120,17 @@ st_translate_fragment_shader(struct st_context *st, case FRAG_ATTRIB_TEX5: case FRAG_ATTRIB_TEX6: case FRAG_ATTRIB_TEX7: - fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_GENERIC; - fs.input_semantic_index[fs.num_inputs] = attr - FRAG_ATTRIB_TEX0; - interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; + fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + fs.input_semantic_index[slot] = attr - FRAG_ATTRIB_TEX0; + interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; case FRAG_ATTRIB_VAR0: /* fall-through */ default: - fs.input_semantic_name[fs.num_inputs] = TGSI_SEMANTIC_GENERIC; - fs.input_semantic_index[fs.num_inputs] = attr - FRAG_ATTRIB_VAR0; - interpMode[fs.num_inputs] = TGSI_INTERPOLATE_PERSPECTIVE; + fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + fs.input_semantic_index[slot] = attr - FRAG_ATTRIB_VAR0; + interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; } - - fs.num_inputs++; } } @@ -145,35 +155,40 @@ st_translate_fragment_shader(struct st_context *st, } } + if (!inputMapping) + inputMapping = defaultInputMapping; + /* XXX: fix static allocation of tokens: */ tgsi_mesa_compile_fp_program( &stfp->Base, + /* inputs */ fs.num_inputs, inputMapping, fs.input_semantic_name, fs.input_semantic_index, interpMode, + /* outputs */ outputMapping, - stfp->tokens, ST_FP_MAX_TOKENS ); + /* tokenized result */ + tokensOut, maxTokens); + - fs.tokens = &stfp->tokens[0]; + fs.tokens = tokensOut; cso = st_cached_fs_state(st, &fs); stfp->fs = cso; if (TGSI_DEBUG) - tgsi_dump( stfp->tokens, 0/*TGSI_DUMP_VERBOSE*/ ); + tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); #if defined(USE_X86_ASM) || defined(SLANG_X86) if (stfp->sse2_program.csr == stfp->sse2_program.store) - tgsi_emit_sse2_fs( stfp->tokens, &stfp->sse2_program ); + tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); if (!cso->state.executable) ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); #endif - stfp->dirty = 0; - return cso; } @@ -200,8 +215,9 @@ static void update_fs( struct st_context *st ) } /* if new binding, or shader has changed */ - if (st->fp != stfp || stfp->dirty) { + if (st->fp != stfp /**|| stfp->dirty**/) { +#if 0 if (stfp->dirty) (void) st_translate_fragment_shader( st, stfp ); @@ -210,10 +226,17 @@ static void update_fs( struct st_context *st ) st->state.fs = stfp->fs; st->pipe->bind_fs_state(st->pipe, st->state.fs->data); +#else + + /* NEW */ + st->dirty.st |= ST_NEW_LINKAGE; + +#endif } } +#if 0 const struct st_tracked_state st_update_fs = { .name = "st_update_fs", .dirty = { @@ -222,3 +245,4 @@ const struct st_tracked_state st_update_fs = { }, .update = update_fs }; +#endif diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c index a6c0d159d4..0f07906a96 100644 --- a/src/mesa/state_tracker/st_atom_vs.c +++ b/src/mesa/state_tracker/st_atom_vs.c @@ -51,15 +51,22 @@ /** * Translate a Mesa vertex shader into a TGSI shader. + * \param outputMapping to map vertex program output registers to TGSI + * output slots + * \param tokensOut destination for TGSI tokens * \return pointer to cached pipe_shader object. */ const struct cso_vertex_shader * -st_translate_vertex_shader(struct st_context *st, - struct st_vertex_program *stvp) +st_translate_vertex_shader(const struct st_context *st, + struct st_vertex_program *stvp, + const GLuint outputMapping[], + struct tgsi_token *tokensOut, + GLuint maxTokens) { + GLuint defaultOutputMapping[VERT_RESULT_MAX]; struct pipe_shader_state vs; const struct cso_vertex_shader *cso; - GLuint attr; + GLuint attr, i; memset(&vs, 0, sizeof(vs)); @@ -69,31 +76,36 @@ st_translate_vertex_shader(struct st_context *st, */ for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { if (stvp->Base.Base.InputsRead & (1 << attr)) { - stvp->input_to_index[attr] = vs.num_inputs; - stvp->index_to_input[vs.num_inputs] = attr; + const GLuint slot = vs.num_inputs; + + vs.num_inputs++; + + stvp->input_to_index[attr] = slot; + stvp->index_to_input[slot] = attr; + switch (attr) { case VERT_ATTRIB_POS: - vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_POSITION; - vs.input_semantic_index[vs.num_inputs] = 0; + vs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + vs.input_semantic_index[slot] = 0; break; case VERT_ATTRIB_WEIGHT: /* fall-through */ case VERT_ATTRIB_NORMAL: /* just label as a generic */ - vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[vs.num_inputs] = 0; + vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.input_semantic_index[slot] = 0; break; case VERT_ATTRIB_COLOR0: - vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; - vs.input_semantic_index[vs.num_inputs] = 0; + vs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs.input_semantic_index[slot] = 0; break; case VERT_ATTRIB_COLOR1: - vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_COLOR; - vs.input_semantic_index[vs.num_inputs] = 1; + vs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs.input_semantic_index[slot] = 1; break; case VERT_ATTRIB_FOG: - vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_FOG; - vs.input_semantic_index[vs.num_inputs] = 0; + vs.input_semantic_name[slot] = TGSI_SEMANTIC_FOG; + vs.input_semantic_index[slot] = 0; break; case VERT_ATTRIB_TEX0: case VERT_ATTRIB_TEX1: @@ -103,8 +115,8 @@ st_translate_vertex_shader(struct st_context *st, case VERT_ATTRIB_TEX5: case VERT_ATTRIB_TEX6: case VERT_ATTRIB_TEX7: - vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[vs.num_inputs] = attr - VERT_ATTRIB_TEX0; + vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.input_semantic_index[slot] = attr - VERT_ATTRIB_TEX0; break; case VERT_ATTRIB_GENERIC0: case VERT_ATTRIB_GENERIC1: @@ -115,53 +127,71 @@ st_translate_vertex_shader(struct st_context *st, case VERT_ATTRIB_GENERIC6: case VERT_ATTRIB_GENERIC7: assert(attr < VERT_ATTRIB_MAX); - vs.input_semantic_name[vs.num_inputs] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[vs.num_inputs] = attr - VERT_ATTRIB_GENERIC0; + vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.input_semantic_index[slot] = attr - VERT_ATTRIB_GENERIC0; break; default: assert(0); } - vs.num_inputs++; } } + /* initialize output semantics to defaults */ + for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) { + vs.output_semantic_name[i] = TGSI_SEMANTIC_GENERIC; + vs.output_semantic_index[i] = 0; + } + /* - * Determine number of outputs, the register mapping and - * the semantic information for each vertex output/result. + * Determine number of outputs, the (default) output register + * mapping and the semantic information for each output. */ for (attr = 0; attr < VERT_RESULT_MAX; attr++) { if (stvp->Base.Base.OutputsWritten & (1 << attr)) { - /* put this attrib in the next available slot */ - st->vertex_attrib_to_slot[attr] = vs.num_outputs; + GLuint slot; + + if (outputMapping) { + slot = outputMapping[attr]; + assert(slot != ~0); + } + else { + slot = vs.num_outputs; + vs.num_outputs++; + defaultOutputMapping[attr] = slot; + } + + /* + printf("Output %u -> slot %u\n", attr, slot); + */ switch (attr) { case VERT_RESULT_HPOS: - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_POSITION; - vs.output_semantic_index[vs.num_outputs] = 0; + vs.output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + vs.output_semantic_index[slot] = 0; break; case VERT_RESULT_COL0: - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_COLOR; - vs.output_semantic_index[vs.num_outputs] = 0; + vs.output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs.output_semantic_index[slot] = 0; break; case VERT_RESULT_COL1: - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_COLOR; - vs.output_semantic_index[vs.num_outputs] = 1; + vs.output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs.output_semantic_index[slot] = 1; break; case VERT_RESULT_BFC0: - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_BCOLOR; - vs.output_semantic_index[vs.num_outputs] = 0; + vs.output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + vs.output_semantic_index[slot] = 0; break; case VERT_RESULT_BFC1: - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_BCOLOR; - vs.output_semantic_index[vs.num_outputs] = 1; + vs.output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + vs.output_semantic_index[slot] = 1; break; case VERT_RESULT_FOGC: - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_FOG; - vs.output_semantic_index[vs.num_outputs] = 0; + vs.output_semantic_name[slot] = TGSI_SEMANTIC_FOG; + vs.output_semantic_index[slot] = 0; break; case VERT_RESULT_PSIZ: - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_PSIZE; - vs.output_semantic_index[vs.num_outputs] = 0; + vs.output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; + vs.output_semantic_index[slot] = 0; break; case VERT_RESULT_EDGE: assert(0); @@ -174,46 +204,59 @@ st_translate_vertex_shader(struct st_context *st, case VERT_RESULT_TEX5: case VERT_RESULT_TEX6: case VERT_RESULT_TEX7: - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[vs.num_outputs] = attr - VERT_RESULT_TEX0; + vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.output_semantic_index[slot] = attr - VERT_RESULT_TEX0; break; case VERT_RESULT_VAR0: /* fall-through */ default: assert(attr - VERT_RESULT_VAR0 < MAX_VARYING); - vs.output_semantic_name[vs.num_outputs] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[vs.num_outputs] = attr - VERT_RESULT_VAR0; + vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs.output_semantic_index[slot] = attr - VERT_RESULT_VAR0; } - - vs.num_outputs++; } } + if (outputMapping) { + /* find max output slot referenced to compute vs.num_outputs */ + GLuint maxSlot = 0; + for (attr = 0; attr < VERT_RESULT_MAX; attr++) { + if (outputMapping[attr] != ~0 && outputMapping[attr] > maxSlot) + maxSlot = outputMapping[attr]; + } + vs.num_outputs = maxSlot + 1; + } + else { + outputMapping = defaultOutputMapping; + } + /* XXX: fix static allocation of tokens: */ tgsi_mesa_compile_vp_program( &stvp->Base, + /* inputs */ vs.num_inputs, stvp->input_to_index, vs.input_semantic_name, vs.input_semantic_index, + /* outputs */ vs.num_outputs, - st->vertex_attrib_to_slot, + outputMapping, vs.output_semantic_name, vs.output_semantic_index, - stvp->tokens, ST_FP_MAX_TOKENS ); - - vs.tokens = &stvp->tokens[0]; + /* tokenized result */ + tokensOut, maxTokens); + vs.tokens = tokensOut; cso = st_cached_vs_state(st, &vs); stvp->vs = cso; if (TGSI_DEBUG) - tgsi_dump( stvp->tokens, 0 ); + tgsi_dump( tokensOut, 0 ); #if defined(USE_X86_ASM) || defined(SLANG_X86) if (stvp->sse2_program.csr == stvp->sse2_program.store) - tgsi_emit_sse2( stvp->tokens, &stvp->sse2_program ); + tgsi_emit_sse2( tokensOut, &stvp->sse2_program ); if (!cso->state.executable) ((struct cso_vertex_shader*)cso)->state.executable = (void *) x86_get_func( &stvp->sse2_program ); @@ -246,6 +289,7 @@ static void update_vs( struct st_context *st ) } if (st->vp != stvp || stvp->dirty) { +#if 0 if (stvp->dirty) (void) st_translate_vertex_shader( st, stvp ); @@ -260,10 +304,15 @@ static void update_vs( struct st_context *st ) tgsi_dump( stvp->tokens, 0 ); #endif st->pipe->bind_vs_state(st->pipe, st->state.vs->data); +#else + /* NEW */ + st->dirty.st |= ST_NEW_LINKAGE; + +#endif } } - +#if 0 const struct st_tracked_state st_update_vs = { .name = "st_update_vs", .dirty = { @@ -272,7 +321,4 @@ const struct st_tracked_state st_update_vs = { }, .update = update_vs }; - - - - +#endif diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 639e0ceb40..367ae06cf3 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -155,7 +155,8 @@ make_frag_shader(struct st_context *st) p->OutputsWritten = (1 << FRAG_RESULT_COLR); stfp = (struct st_fragment_program *) p; - st_translate_fragment_shader(st, stfp); + st_translate_fragment_program(st, stfp, NULL, + stfp->tokens, ST_FP_MAX_TOKENS); return stfp; } @@ -203,7 +204,8 @@ make_vertex_shader(struct st_context *st) (1 << VERT_RESULT_HPOS)); stvp = (struct st_vertex_program *) p; - st_translate_vertex_shader(st, stvp); + st_translate_vertex_program(st, stvp, NULL, + stvp->tokens, ST_FP_MAX_TOKENS); assert(stvp->vs); return stvp; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 65c5465546..619c5d8ab7 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -92,7 +92,8 @@ make_fragment_shader(struct st_context *st) p->OutputsWritten = (1 << FRAG_RESULT_COLR); stfp = (struct st_fragment_program *) p; - st_translate_fragment_shader(st, stfp); + st_translate_fragment_program(st, stfp, NULL, + stfp->tokens, ST_FP_MAX_TOKENS); return stfp; } @@ -141,7 +142,8 @@ make_vertex_shader(struct st_context *st) (1 << VERT_RESULT_HPOS)); stvp = (struct st_vertex_program *) p; - st_translate_vertex_shader(st, stvp); + st_translate_vertex_program(st, stvp, NULL, + stvp->tokens, ST_FP_MAX_TOKENS); return stvp; } diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 78cf4c2b4d..537a58f39d 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -102,13 +102,13 @@ feedback_vertex(GLcontext *ctx, const struct draw_context *draw, * color and texcoord attribs to use here. */ - slot = st->vertex_attrib_to_slot[VERT_RESULT_COL0]; + slot = st->vertex_result_to_slot[VERT_RESULT_COL0]; if (slot) color = v->data[slot]; else color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; - slot = st->vertex_attrib_to_slot[VERT_RESULT_TEX0]; + slot = st->vertex_result_to_slot[VERT_RESULT_TEX0]; if (slot) texcoord = v->data[slot]; else diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 9f46f9e93f..aee316df6f 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -30,24 +30,26 @@ * Keith Whitwell */ +#include "main/glheader.h" +#include "main/macros.h" +#include "main/enums.h" +#include "shader/prog_instruction.h" +#include "shader/prog_parameter.h" +#include "shader/program.h" +#include "shader/programopt.h" + #include "st_context.h" -#include "st_program.h" -#include "glheader.h" -#include "macros.h" -#include "enums.h" -#include "prog_instruction.h" -#include "prog_parameter.h" -#include "program.h" -#include "programopt.h" +#include "st_program.h" +#include "st_atom_shader.h" + #include "tnl/tnl.h" #include "pipe/tgsi/mesa/tgsi_mesa.h" -/* Counter to track program string changes: +/** + * Called via ctx->Driver.BindProgram() to bind an ARB vertex or + * fragment program. */ -static GLuint program_id = 0; - - static void st_bind_program( GLcontext *ctx, GLenum target, struct gl_program *prog ) @@ -62,8 +64,14 @@ static void st_bind_program( GLcontext *ctx, st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; break; } + st->dirty.st |= ST_NEW_LINKAGE; } + +/** + * Called via ctx->Driver.UseProgram() to bind a linked GLSL program + * (vertex shader + fragment shader). + */ static void st_use_program( GLcontext *ctx, GLuint program ) { @@ -71,6 +79,7 @@ static void st_use_program( GLcontext *ctx, st->dirty.st |= ST_NEW_VERTEX_PROGRAM; st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; + st->dirty.st |= ST_NEW_LINKAGE; } @@ -79,14 +88,13 @@ static struct gl_program *st_new_program( GLcontext *ctx, GLenum target, GLuint id ) { -// struct st_context *st = st_context(ctx); + struct st_context *st = st_context(ctx); switch (target) { case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *prog = CALLOC_STRUCT(st_vertex_program); - prog->id = program_id++; - prog->dirty = 1; + prog->serialNo = 1; #if defined(USE_X86_ASM) || defined(SLANG_X86) x86_init_func( &prog->sse2_program ); @@ -102,8 +110,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, case GL_FRAGMENT_PROGRAM_NV: { struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program); - prog->id = program_id++; - prog->dirty = 1; + prog->serialNo = 1; #if defined(USE_X86_ASM) || defined(SLANG_X86) x86_init_func( &prog->sse2_program ); @@ -118,29 +125,40 @@ static struct gl_program *st_new_program( GLcontext *ctx, default: return _mesa_new_program(ctx, target, id); } + + st->dirty.st |= ST_NEW_LINKAGE; } + static void st_delete_program( GLcontext *ctx, struct gl_program *prog ) { + struct st_context *st = st_context(ctx); + switch( prog->Target ) { - case GL_VERTEX_PROGRAM_ARB: { + case GL_VERTEX_PROGRAM_ARB: + { + struct st_vertex_program *stvp = (struct st_vertex_program *) prog; #if defined(USE_X86_ASM) || defined(SLANG_X86) - struct st_vertex_program *p = (struct st_vertex_program *) prog; - - x86_release_func( &p->sse2_program ); + x86_release_func( &stvp->sse2_program ); #endif + st_remove_vertex_program(st, stvp); + } break; - } - case GL_FRAGMENT_PROGRAM_ARB: { + case GL_FRAGMENT_PROGRAM_ARB: + { + struct st_fragment_program *stfp + = (struct st_fragment_program *) prog; #if defined(USE_X86_ASM) || defined(SLANG_X86) - struct st_fragment_program *p = (struct st_fragment_program *) prog; - - x86_release_func( &p->sse2_program ); + x86_release_func( &stfp->sse2_program ); #endif + st_remove_fragment_program(st, stfp); + } break; + default: + assert(0); /* problem */ } - } + _mesa_delete_program( ctx, prog ); } @@ -160,27 +178,31 @@ static void st_program_string_notify( GLcontext *ctx, struct st_context *st = st_context(ctx); if (target == GL_FRAGMENT_PROGRAM_ARB) { - struct st_fragment_program *p = (struct st_fragment_program *)prog; + struct st_fragment_program *stfp = (struct st_fragment_program *) prog; if (prog == &ctx->FragmentProgram._Current->Base) st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - p->id = program_id++; - p->param_state = p->Base.Base.Parameters->StateFlags; + stfp->serialNo++; + + stfp->param_state = stfp->Base.Base.Parameters->StateFlags; } else if (target == GL_VERTEX_PROGRAM_ARB) { - struct st_vertex_program *p = (struct st_vertex_program *)prog; + struct st_vertex_program *stvp = (struct st_vertex_program *) prog; if (prog == &ctx->VertexProgram._Current->Base) st->dirty.st |= ST_NEW_VERTEX_PROGRAM; - p->id = program_id++; - p->param_state = p->Base.Base.Parameters->StateFlags; + stvp->serialNo++; + + stvp->param_state = stvp->Base.Base.Parameters->StateFlags; /* Also tell tnl about it: */ _tnl_program_string(ctx, target, prog); } + + st->dirty.st |= ST_NEW_LINKAGE; } diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 2311bddc65..661d155e6d 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -53,9 +53,9 @@ static void setup_vertex_attribs(GLcontext *ctx) { struct pipe_context *pipe = ctx->st->pipe; -#if 0 - const uint inputAttrs = ctx->st->state.vs->inputs_read; - uint attr; + const struct cso_vertex_shader *vs = ctx->st->state.vs; + const struct st_vertex_program *stvp = ctx->st->vp; + uint slot; /* all attributes come from the default attribute buffer */ { @@ -67,20 +67,16 @@ setup_vertex_attribs(GLcontext *ctx) pipe->set_vertex_buffer(pipe, 0, &vbuffer); } - for (attr = 0; attr < 16; attr++) { + for (slot = 0; slot < vs->state.num_inputs; slot++) { struct pipe_vertex_element velement; + const GLuint attr = stvp->index_to_input[slot]; - if (inputAttrs & (1 << attr)) { - velement.src_offset = attr * 4 * sizeof(GLfloat); - velement.vertex_buffer_index = 0; - velement.dst_offset = 0; - velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - pipe->set_vertex_element(pipe, attr, &velement); - } + velement.src_offset = attr * 4 * sizeof(GLfloat); + velement.vertex_buffer_index = 0; + velement.dst_offset = 0; + velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + pipe->set_vertex_element(pipe, slot, &velement); } -#else - assert(0); -#endif } @@ -98,12 +94,11 @@ setup_feedback(GLcontext *ctx) feedback.discard = 1; feedback.num_attribs = 0; + /* feedback all results from vertex shader */ for (i = 0; i < vs->num_outputs; i++) { - if (1/***(1 << i) & outputAttrs***/) { - feedback.attrib[feedback.num_attribs] = i; - feedback.size[feedback.num_attribs] = 4; - feedback.num_attribs++; - } + feedback.attrib[feedback.num_attribs] = i; + feedback.size[feedback.num_attribs] = 4; + feedback.num_attribs++; } pipe->set_feedback_state(pipe, &feedback); @@ -261,13 +256,11 @@ update_rasterpos(GLcontext *ctx, static void st_RasterPos(GLcontext *ctx, const GLfloat v[4]) { - struct pipe_context *pipe = ctx->st->pipe; + const struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; float *buf_map; struct pipe_feedback_buffer fb_buf; - /** XXX TEMPORARILY DISABLE */ - return; - st_validate_state(ctx->st); /* setup vertex buffers */ @@ -277,7 +270,7 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) * Load the default attribute buffer with current attribs. */ { - struct pipe_buffer_handle *buf = ctx->st->default_attrib_buffer; + struct pipe_buffer_handle *buf = st->default_attrib_buffer; const unsigned size = sizeof(ctx->Current.Attrib); const void *data = ctx->Current.Attrib; /* colors, texcoords, etc */ @@ -313,17 +306,16 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) PIPE_BUFFER_FLAG_READ); /* extract values and update rasterpos state */ -#if 0 /* XXX update */ { - const uint outputAttrs = ctx->st->state.vs->outputs_written; + const GLuint *outputMapping = st->vertex_result_to_slot; const float *pos, *color0, *color1, *tex0; float *buf = buf_map; - assert(outputAttrs & (1 << TGSI_ATTRIB_POS)); + assert(outputMapping[VERT_RESULT_HPOS] != ~0); pos = buf; buf += 4; - if (outputAttrs & (1 << TGSI_ATTRIB_COLOR0)) { + if (outputMapping[VERT_RESULT_COL0] != ~0) { color0 = buf; buf += 4; } @@ -331,7 +323,7 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) color0 = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; } - if (outputAttrs & (1 << TGSI_ATTRIB_COLOR1)) { + if (outputMapping[VERT_RESULT_COL1] != ~0) { color1 = buf; buf += 4; } @@ -339,16 +331,23 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) color1 = ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; } + if (outputMapping[VERT_RESULT_TEX0] != ~0) { + tex0 = buf; + buf += 4; + } + else { + tex0 = ctx->Current.Attrib[VERT_ATTRIB_TEX0]; + } + update_rasterpos(ctx, pos, color0, color1, tex0); } -#endif /* free vertex feedback buffer */ pipe->winsys->buffer_unmap(pipe->winsys, fb_buf.buffer); pipe->winsys->buffer_reference(pipe->winsys, &fb_buf.buffer, NULL); /* restore pipe state */ - pipe->set_feedback_state(pipe, &ctx->st->state.feedback); + pipe->set_feedback_state(pipe, &st->state.feedback); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 3713328eb1..24f0ff9aaf 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -45,6 +45,8 @@ struct cso_blend; #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 #define ST_NEW_VERTEX_PROGRAM 0x4 +#define ST_NEW_LINKAGE 0x8 + struct st_state_flags { GLuint mesa; @@ -119,8 +121,8 @@ struct st_context GLfloat polygon_offset_scale; /* ?? */ - /** Mapping from VERT_ATTRIB_x to post-transformed vertex slot */ - GLuint vertex_attrib_to_slot[VERT_RESULT_MAX]; + /** Mapping from VERT_RESULT_x to post-transformed vertex slot */ + const GLuint *vertex_result_to_slot; struct st_vertex_program *vp; /**< Currently bound vertex program */ struct st_fragment_program *fp; /**< Currently bound fragment program */ diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 419afa4e78..355dee574b 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -47,11 +47,11 @@ struct st_fragment_program { struct gl_fragment_program Base; GLboolean error; /* If program is malformed for any reason. */ - GLuint id; /**< String id, for tracking ProgramStringNotify changes. */ + + GLuint serialNo; /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; - GLboolean dirty; #if defined(USE_X86_ASM) || defined(SLANG_X86) struct x86_function sse2_program; @@ -68,7 +68,8 @@ struct st_vertex_program { struct gl_vertex_program Base; /**< The Mesa vertex program */ GLboolean error; /**< Set if program is malformed for any reason. */ - GLuint id; /**< String id, for tracking ProgramStringNotify changes. */ + + GLuint serialNo; /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ GLuint input_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; @@ -77,7 +78,6 @@ struct st_vertex_program /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; - GLboolean dirty; #if defined(USE_X86_ASM) || defined(SLANG_X86) struct x86_function sse2_program; @@ -90,7 +90,8 @@ struct st_vertex_program }; -extern void st_init_program_functions(struct dd_function_table *functions); +extern void +st_init_program_functions(struct dd_function_table *functions); static inline struct st_fragment_program * @@ -99,6 +100,7 @@ st_fragment_program( struct gl_fragment_program *fp ) return (struct st_fragment_program *)fp; } + static inline struct st_vertex_program * st_vertex_program( struct gl_vertex_program *vp ) { @@ -107,13 +109,18 @@ st_vertex_program( struct gl_vertex_program *vp ) extern const struct cso_fragment_shader * -st_translate_fragment_shader(struct st_context *st, - struct st_fragment_program *fp); +st_translate_fragment_program(struct st_context *st, + struct st_fragment_program *fp, + const GLuint inputMapping[], + struct tgsi_token *tokens, + GLuint maxTokens); extern const struct cso_vertex_shader * -st_translate_vertex_shader(struct st_context *st, - struct st_vertex_program *vp); - +st_translate_vertex_program(struct st_context *st, + struct st_vertex_program *vp, + const GLuint vert_output_to_slot[], + struct tgsi_token *tokens, + GLuint maxTokens); #endif -- cgit v1.2.3 From 6b1d2fa8154e04eea7fa55e07b951aef3b23993f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 14:30:26 -0600 Subject: replaced by st_atom_shader.c --- src/mesa/state_tracker/st_atom_fs.c | 248 --------------------------- src/mesa/state_tracker/st_atom_vs.c | 324 ------------------------------------ 2 files changed, 572 deletions(-) delete mode 100644 src/mesa/state_tracker/st_atom_fs.c delete mode 100644 src/mesa/state_tracker/st_atom_vs.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fs.c b/src/mesa/state_tracker/st_atom_fs.c deleted file mode 100644 index 28019858f7..0000000000 --- a/src/mesa/state_tracker/st_atom_fs.c +++ /dev/null @@ -1,248 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* - * Authors: - * Keith Whitwell - * Brian Paul - */ - -#include "shader/prog_parameter.h" - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_winsys.h" -#include "pipe/tgsi/mesa/mesa_to_tgsi.h" -#include "pipe/tgsi/exec/tgsi_core.h" -#include "pipe/tgsi/exec/tgsi_dump.h" - -#include "st_context.h" -#include "st_cache.h" -#include "st_atom.h" -#include "st_program.h" - - -#define TGSI_DEBUG 0 - - -/** - * Translate a Mesa fragment shader into a TGSI shader. - * \param inputMapping to map fragment program input registers to TGSI - * input slots - * \param tokensOut destination for TGSI tokens - * \return pointer to cached pipe_shader object. - */ -const struct cso_fragment_shader * -st_translate_fragment_shader(const struct st_context *st, - struct st_fragment_program *stfp, - const GLuint inputMapping[], - struct tgsi_token *tokensOut, - GLuint maxTokens) -{ - GLuint outputMapping[FRAG_RESULT_MAX]; - GLuint defaultInputMapping[FRAG_ATTRIB_MAX]; - struct pipe_shader_state fs; - const struct cso_fragment_shader *cso; - GLuint interpMode[16]; /* XXX size? */ - GLuint attr; - GLbitfield inputsRead = stfp->Base.Base.InputsRead; - - /* Check if all fragment programs need the fragment position (in order - * to do perspective-corrected interpolation). - */ - /* XXX temporary! */ - if (st->pipe->get_param(st->pipe, PIPE_PARAM_FS_NEEDS_POS)) - inputsRead |= FRAG_BIT_WPOS; - - memset(&fs, 0, sizeof(fs)); - - /* - * Convert Mesa program inputs to TGSI input register semantics. - */ - for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) { - if (inputsRead & (1 << attr)) { - const GLuint slot = fs.num_inputs; - - fs.num_inputs++; - - defaultInputMapping[attr] = slot; - - switch (attr) { - case FRAG_ATTRIB_WPOS: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - fs.input_semantic_index[slot] = 0; - interpMode[slot] = TGSI_INTERPOLATE_CONSTANT; - break; - case FRAG_ATTRIB_COL0: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - fs.input_semantic_index[slot] = 0; - interpMode[slot] = TGSI_INTERPOLATE_LINEAR; - break; - case FRAG_ATTRIB_COL1: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - fs.input_semantic_index[slot] = 1; - interpMode[slot] = TGSI_INTERPOLATE_LINEAR; - break; - case FRAG_ATTRIB_FOGC: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_FOG; - fs.input_semantic_index[slot] = 0; - interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; - break; - case FRAG_ATTRIB_TEX0: - case FRAG_ATTRIB_TEX1: - case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: - case FRAG_ATTRIB_TEX4: - case FRAG_ATTRIB_TEX5: - case FRAG_ATTRIB_TEX6: - case FRAG_ATTRIB_TEX7: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - fs.input_semantic_index[slot] = attr - FRAG_ATTRIB_TEX0; - interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; - break; - case FRAG_ATTRIB_VAR0: - /* fall-through */ - default: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - fs.input_semantic_index[slot] = attr - FRAG_ATTRIB_VAR0; - interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; - } - } - } - - /* - * Semantics for outputs - */ - for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { - if (stfp->Base.Base.OutputsWritten & (1 << attr)) { - switch (attr) { - case FRAG_RESULT_DEPR: - fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_POSITION; - outputMapping[attr] = fs.num_outputs; - break; - case FRAG_RESULT_COLR: - fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_COLOR; - outputMapping[attr] = fs.num_outputs; - break; - default: - assert(0); - } - fs.num_outputs++; - } - } - - if (!inputMapping) - inputMapping = defaultInputMapping; - - /* XXX: fix static allocation of tokens: - */ - tgsi_mesa_compile_fp_program( &stfp->Base, - /* inputs */ - fs.num_inputs, - inputMapping, - fs.input_semantic_name, - fs.input_semantic_index, - interpMode, - /* outputs */ - outputMapping, - /* tokenized result */ - tokensOut, maxTokens); - - - fs.tokens = tokensOut; - - cso = st_cached_fs_state(st, &fs); - stfp->fs = cso; - - if (TGSI_DEBUG) - tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); - -#if defined(USE_X86_ASM) || defined(SLANG_X86) - if (stfp->sse2_program.csr == stfp->sse2_program.store) - tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); - - if (!cso->state.executable) - ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); -#endif - - return cso; -} - - - -static void update_fs( struct st_context *st ) -{ - struct st_fragment_program *stfp = NULL; - - /* find active shader and params. Changes to this Mesa state - * should be covered by ST_NEW_FRAGMENT_PROGRAM, thanks to the - * logic in st_cb_program.c - */ - if (st->ctx->Shader.CurrentProgram && - st->ctx->Shader.CurrentProgram->LinkStatus && - st->ctx->Shader.CurrentProgram->FragmentProgram) { - struct gl_fragment_program *f - = st->ctx->Shader.CurrentProgram->FragmentProgram; - stfp = st_fragment_program(f); - } - else { - assert(st->ctx->FragmentProgram._Current); - stfp = st_fragment_program(st->ctx->FragmentProgram._Current); - } - - /* if new binding, or shader has changed */ - if (st->fp != stfp /**|| stfp->dirty**/) { - -#if 0 - if (stfp->dirty) - (void) st_translate_fragment_shader( st, stfp ); - - /* Bind the vertex program and TGSI shader */ - st->fp = stfp; - st->state.fs = stfp->fs; - - st->pipe->bind_fs_state(st->pipe, st->state.fs->data); -#else - - /* NEW */ - st->dirty.st |= ST_NEW_LINKAGE; - -#endif - } -} - - -#if 0 -const struct st_tracked_state st_update_fs = { - .name = "st_update_fs", - .dirty = { - .mesa = 0, - .st = ST_NEW_FRAGMENT_PROGRAM, - }, - .update = update_fs -}; -#endif diff --git a/src/mesa/state_tracker/st_atom_vs.c b/src/mesa/state_tracker/st_atom_vs.c deleted file mode 100644 index 0f07906a96..0000000000 --- a/src/mesa/state_tracker/st_atom_vs.c +++ /dev/null @@ -1,324 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - -#include "shader/prog_parameter.h" -#include "shader/prog_print.h" -#include "tnl/t_vp_build.h" - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_winsys.h" -#include "pipe/tgsi/mesa/mesa_to_tgsi.h" -#include "pipe/tgsi/exec/tgsi_core.h" - -#include "st_context.h" -#include "st_cache.h" -#include "st_atom.h" -#include "st_program.h" - - -#define TGSI_DEBUG 0 - - -/** - * Translate a Mesa vertex shader into a TGSI shader. - * \param outputMapping to map vertex program output registers to TGSI - * output slots - * \param tokensOut destination for TGSI tokens - * \return pointer to cached pipe_shader object. - */ -const struct cso_vertex_shader * -st_translate_vertex_shader(const struct st_context *st, - struct st_vertex_program *stvp, - const GLuint outputMapping[], - struct tgsi_token *tokensOut, - GLuint maxTokens) -{ - GLuint defaultOutputMapping[VERT_RESULT_MAX]; - struct pipe_shader_state vs; - const struct cso_vertex_shader *cso; - GLuint attr, i; - - memset(&vs, 0, sizeof(vs)); - - /* - * Determine number of inputs, the mappings between VERT_ATTRIB_x - * and TGSI generic input indexes, plus input attrib semantic info. - */ - for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { - if (stvp->Base.Base.InputsRead & (1 << attr)) { - const GLuint slot = vs.num_inputs; - - vs.num_inputs++; - - stvp->input_to_index[attr] = slot; - stvp->index_to_input[slot] = attr; - - switch (attr) { - case VERT_ATTRIB_POS: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - vs.input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_WEIGHT: - /* fall-through */ - case VERT_ATTRIB_NORMAL: - /* just label as a generic */ - vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_COLOR0: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs.input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_COLOR1: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs.input_semantic_index[slot] = 1; - break; - case VERT_ATTRIB_FOG: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_FOG; - vs.input_semantic_index[slot] = 0; - break; - case VERT_ATTRIB_TEX0: - case VERT_ATTRIB_TEX1: - case VERT_ATTRIB_TEX2: - case VERT_ATTRIB_TEX3: - case VERT_ATTRIB_TEX4: - case VERT_ATTRIB_TEX5: - case VERT_ATTRIB_TEX6: - case VERT_ATTRIB_TEX7: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[slot] = attr - VERT_ATTRIB_TEX0; - break; - case VERT_ATTRIB_GENERIC0: - case VERT_ATTRIB_GENERIC1: - case VERT_ATTRIB_GENERIC2: - case VERT_ATTRIB_GENERIC3: - case VERT_ATTRIB_GENERIC4: - case VERT_ATTRIB_GENERIC5: - case VERT_ATTRIB_GENERIC6: - case VERT_ATTRIB_GENERIC7: - assert(attr < VERT_ATTRIB_MAX); - vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[slot] = attr - VERT_ATTRIB_GENERIC0; - break; - default: - assert(0); - } - } - } - - /* initialize output semantics to defaults */ - for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) { - vs.output_semantic_name[i] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[i] = 0; - } - - /* - * Determine number of outputs, the (default) output register - * mapping and the semantic information for each output. - */ - for (attr = 0; attr < VERT_RESULT_MAX; attr++) { - if (stvp->Base.Base.OutputsWritten & (1 << attr)) { - GLuint slot; - - if (outputMapping) { - slot = outputMapping[attr]; - assert(slot != ~0); - } - else { - slot = vs.num_outputs; - vs.num_outputs++; - defaultOutputMapping[attr] = slot; - } - - /* - printf("Output %u -> slot %u\n", attr, slot); - */ - - switch (attr) { - case VERT_RESULT_HPOS: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - vs.output_semantic_index[slot] = 0; - break; - case VERT_RESULT_COL0: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs.output_semantic_index[slot] = 0; - break; - case VERT_RESULT_COL1: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs.output_semantic_index[slot] = 1; - break; - case VERT_RESULT_BFC0: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; - vs.output_semantic_index[slot] = 0; - break; - case VERT_RESULT_BFC1: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; - vs.output_semantic_index[slot] = 1; - break; - case VERT_RESULT_FOGC: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_FOG; - vs.output_semantic_index[slot] = 0; - break; - case VERT_RESULT_PSIZ: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; - vs.output_semantic_index[slot] = 0; - break; - case VERT_RESULT_EDGE: - assert(0); - break; - case VERT_RESULT_TEX0: - case VERT_RESULT_TEX1: - case VERT_RESULT_TEX2: - case VERT_RESULT_TEX3: - case VERT_RESULT_TEX4: - case VERT_RESULT_TEX5: - case VERT_RESULT_TEX6: - case VERT_RESULT_TEX7: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[slot] = attr - VERT_RESULT_TEX0; - break; - case VERT_RESULT_VAR0: - /* fall-through */ - default: - assert(attr - VERT_RESULT_VAR0 < MAX_VARYING); - vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[slot] = attr - VERT_RESULT_VAR0; - } - } - } - - - if (outputMapping) { - /* find max output slot referenced to compute vs.num_outputs */ - GLuint maxSlot = 0; - for (attr = 0; attr < VERT_RESULT_MAX; attr++) { - if (outputMapping[attr] != ~0 && outputMapping[attr] > maxSlot) - maxSlot = outputMapping[attr]; - } - vs.num_outputs = maxSlot + 1; - } - else { - outputMapping = defaultOutputMapping; - } - - /* XXX: fix static allocation of tokens: - */ - tgsi_mesa_compile_vp_program( &stvp->Base, - /* inputs */ - vs.num_inputs, - stvp->input_to_index, - vs.input_semantic_name, - vs.input_semantic_index, - /* outputs */ - vs.num_outputs, - outputMapping, - vs.output_semantic_name, - vs.output_semantic_index, - /* tokenized result */ - tokensOut, maxTokens); - - vs.tokens = tokensOut; - cso = st_cached_vs_state(st, &vs); - stvp->vs = cso; - - if (TGSI_DEBUG) - tgsi_dump( tokensOut, 0 ); - -#if defined(USE_X86_ASM) || defined(SLANG_X86) - if (stvp->sse2_program.csr == stvp->sse2_program.store) - tgsi_emit_sse2( tokensOut, &stvp->sse2_program ); - - if (!cso->state.executable) - ((struct cso_vertex_shader*)cso)->state.executable = (void *) x86_get_func( &stvp->sse2_program ); -#endif - - stvp->dirty = 0; - - return cso; -} - - - -static void update_vs( struct st_context *st ) -{ - struct st_vertex_program *stvp; - - /* find active shader and params -- Should be covered by - * ST_NEW_VERTEX_PROGRAM - */ - if (st->ctx->Shader.CurrentProgram && - st->ctx->Shader.CurrentProgram->LinkStatus && - st->ctx->Shader.CurrentProgram->VertexProgram) { - struct gl_vertex_program *f - = st->ctx->Shader.CurrentProgram->VertexProgram; - stvp = st_vertex_program(f); - } - else { - assert(st->ctx->VertexProgram._Current); - stvp = st_vertex_program(st->ctx->VertexProgram._Current); - } - - if (st->vp != stvp || stvp->dirty) { -#if 0 - if (stvp->dirty) - (void) st_translate_vertex_shader( st, stvp ); - - /* Bind the vertex program and TGSI shader */ - st->vp = stvp; - st->state.vs = stvp->vs; - -#if 0 - printf("###### bind vp tokens: %p %p num_inp=%u\n", - stvp, stvp->tokens, stvp->vs->state.num_inputs); - if (TGSI_DEBUG) - tgsi_dump( stvp->tokens, 0 ); -#endif - st->pipe->bind_vs_state(st->pipe, st->state.vs->data); -#else - /* NEW */ - st->dirty.st |= ST_NEW_LINKAGE; - -#endif - } -} - -#if 0 -const struct st_tracked_state st_update_vs = { - .name = "st_update_vs", - .dirty = { - .mesa = _NEW_PROGRAM, - .st = ST_NEW_VERTEX_PROGRAM, - }, - .update = update_vs -}; -#endif -- cgit v1.2.3 From 1201eb852966a64742645a4cd3a2879bed81e29d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 14:46:18 -0600 Subject: disable TGSI_DEBUG --- src/mesa/state_tracker/st_program.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 88c02e5b75..fc5d00020a 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -47,7 +47,7 @@ #include "st_program.h" -#define TGSI_DEBUG 1 +#define TGSI_DEBUG 0 /** -- cgit v1.2.3 From 07f31a29b41be572c48fc1b440eaeb6a6b010cc8 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 14:46:49 -0600 Subject: some clean-up of ST_NEW_ shader flags --- src/mesa/state_tracker/st_atom_shader.c | 2 +- src/mesa/state_tracker/st_cb_program.c | 24 ++++-------------------- src/mesa/state_tracker/st_context.h | 2 +- 3 files changed, 6 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index f3bb9effde..383184e575 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -416,7 +416,7 @@ const struct st_tracked_state st_update_shader = { .name = "st_update_shader", .dirty = { .mesa = 0, - .st = ST_NEW_LINKAGE + .st = ST_NEW_SHADER }, .update = update_linkage }; diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index aee316df6f..f57d22652f 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -56,15 +56,7 @@ static void st_bind_program( GLcontext *ctx, { struct st_context *st = st_context(ctx); - switch (target) { - case GL_VERTEX_PROGRAM_ARB: - st->dirty.st |= ST_NEW_VERTEX_PROGRAM; - break; - case GL_FRAGMENT_PROGRAM_ARB: - st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - break; - } - st->dirty.st |= ST_NEW_LINKAGE; + st->dirty.st |= ST_NEW_SHADER; } @@ -77,9 +69,7 @@ static void st_use_program( GLcontext *ctx, { struct st_context *st = st_context(ctx); - st->dirty.st |= ST_NEW_VERTEX_PROGRAM; - st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - st->dirty.st |= ST_NEW_LINKAGE; + st->dirty.st |= ST_NEW_SHADER; } @@ -126,7 +116,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, return _mesa_new_program(ctx, target, id); } - st->dirty.st |= ST_NEW_LINKAGE; + st->dirty.st |= ST_NEW_SHADER; } @@ -180,9 +170,6 @@ static void st_program_string_notify( GLcontext *ctx, if (target == GL_FRAGMENT_PROGRAM_ARB) { struct st_fragment_program *stfp = (struct st_fragment_program *) prog; - if (prog == &ctx->FragmentProgram._Current->Base) - st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; - stfp->serialNo++; stfp->param_state = stfp->Base.Base.Parameters->StateFlags; @@ -190,9 +177,6 @@ static void st_program_string_notify( GLcontext *ctx, else if (target == GL_VERTEX_PROGRAM_ARB) { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; - if (prog == &ctx->VertexProgram._Current->Base) - st->dirty.st |= ST_NEW_VERTEX_PROGRAM; - stvp->serialNo++; stvp->param_state = stvp->Base.Base.Parameters->StateFlags; @@ -202,7 +186,7 @@ static void st_program_string_notify( GLcontext *ctx, _tnl_program_string(ctx, target, prog); } - st->dirty.st |= ST_NEW_LINKAGE; + st->dirty.st |= ST_NEW_SHADER; } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 24f0ff9aaf..da4fd0d4e4 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -45,7 +45,7 @@ struct cso_blend; #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 #define ST_NEW_VERTEX_PROGRAM 0x4 -#define ST_NEW_LINKAGE 0x8 +#define ST_NEW_SHADER 0x8 struct st_state_flags { -- cgit v1.2.3 From 3bf8d2ac7108a7f0f1722e411161e013bb8573f0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 15:18:51 -0600 Subject: Disable vertex shader fog, compute fog in fragment shader. --- src/mesa/main/texenvprogram.c | 16 +++++++++++----- src/mesa/state_tracker/st_context.c | 5 +++++ 2 files changed, 16 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index 37f8fc8090..48c28d45ca 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -1097,15 +1097,21 @@ create_new_program(GLcontext *ctx, struct state_key *key, ASSERT(p.program->Base.NumInstructions <= MAX_INSTRUCTIONS); /* Allocate final instruction array */ - program->Base.Instructions - = _mesa_alloc_instructions(program->Base.NumInstructions); - if (!program->Base.Instructions) { + p.program->Base.Instructions + = _mesa_alloc_instructions(p.program->Base.NumInstructions); + if (!p.program->Base.Instructions) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "generating tex env program"); return; } - _mesa_copy_instructions(program->Base.Instructions, instBuffer, - program->Base.NumInstructions); + _mesa_copy_instructions(p.program->Base.Instructions, instBuffer, + p.program->Base.NumInstructions); + + if (p.program->FogOption) { + _mesa_append_fog_code(ctx, p.program); + p.program->FogOption = GL_NONE; + } + /* Notify driver the fragment program has (actually) changed. */ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index e0304dd22d..7c20b036a4 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -26,6 +26,8 @@ **************************************************************************/ #include "main/imports.h" +#include "main/extensions.h" +#include "tnl/tnl.h" #include "vbo/vbo.h" #include "st_public.h" #include "st_context.h" @@ -98,6 +100,9 @@ struct st_context *st_create_context( GLcontext *ctx, /* XXXX This is temporary! */ _mesa_enable_sw_extensions(ctx); + /* we'll always do per-pixel fog in the fragment shader */ + _tnl_allow_vertex_fog(ctx, GL_FALSE); + return st; } -- cgit v1.2.3 From 83fad68ec1989c719646a76f4cc5e0b3d23537ed Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 15:20:04 -0600 Subject: include programopt.h --- src/mesa/main/texenvprogram.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index 48c28d45ca..cdf35b4636 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -32,6 +32,7 @@ #include "shader/prog_instruction.h" #include "shader/prog_print.h" #include "shader/prog_statevars.h" +#include "shader/programopt.h" #include "texenvprogram.h" /** -- cgit v1.2.3 From 38504011c8dc55cbcfdf16e39eb7a1473e1bdfbd Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 15:20:58 -0600 Subject: better debug output --- src/mesa/state_tracker/st_atom_constbuf.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index c2310f9d3f..cf52a923d3 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -66,15 +66,15 @@ static void upload_constants( struct st_context *st, { int i; - _mesa_printf("%s(%d): %d / %x\n", - __FUNCTION__, id, params->NumParameters, params->StateFlags); + printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", + __FUNCTION__, id, params->NumParameters, params->StateFlags); for (i = 0; i < params->NumParameters; i++) - fprintf(stderr, "%d: %f %f %f %f\n", i, - params->ParameterValues[i][0], - params->ParameterValues[i][1], - params->ParameterValues[i][2], - params->ParameterValues[i][3]); + printf("%d: %f %f %f %f\n", i, + params->ParameterValues[i][0], + params->ParameterValues[i][1], + params->ParameterValues[i][2], + params->ParameterValues[i][3]); } /* load Mesa constants into the constant buffer */ -- cgit v1.2.3 From b5051bf6bbe54269c78f4dd545791e51bc7ce9e0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 15:21:19 -0600 Subject: clean-up #includes --- src/mesa/state_tracker/st_program.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index fc5d00020a..9f363adc5c 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -31,13 +31,11 @@ */ -#include "shader/prog_parameter.h" -#include "shader/prog_print.h" -#include "tnl/t_vp_build.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/p_winsys.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/exec/tgsi_core.h" -- cgit v1.2.3 From 02ea8b81414b50936b8b6b7b8031511e12ef55cd Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 15:48:14 -0600 Subject: re-org/prep for glBitmap --- src/mesa/state_tracker/st_cb_drawpixels.c | 27 ++++++++++----------------- 1 file changed, 10 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 619c5d8ab7..c30d4ede4a 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -30,7 +30,8 @@ * Brian Paul */ -#include "imports.h" +#include "main/imports.h" +#include "main/image.h" #include "st_context.h" #include "st_atom.h" @@ -59,12 +60,6 @@ make_fragment_shader(struct st_context *st) GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; struct gl_program *p; - GLuint interpMode[16]; - GLuint i; - - /* XXX temporary */ - for (i = 0; i < 16; i++) - interpMode[i] = TGSI_INTERPOLATE_LINEAR; p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) @@ -297,13 +292,11 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, static void draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, - GLsizei width, GLsizei height, GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels) + GLsizei width, GLsizei height, + struct pipe_mipmap_tree *mt) { const GLuint unit = 0; struct pipe_context *pipe = ctx->st->pipe; - struct pipe_mipmap_tree *mt; GLfloat x0, y0, x1, y1; GLuint maxWidth, maxHeight; @@ -377,8 +370,6 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* mipmap tree state: */ { - mt = make_mipmap_tree(ctx->st, width, height, format, type, - unpack, pixels); pipe->set_texture_state(pipe, unit, mt); } @@ -401,8 +392,6 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]->data); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); - - free_mipmap_tree(pipe, mt); } @@ -547,8 +536,12 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, any_pixel_transfer_ops(st) || !compatible_formats(format, type, ps->format)) { /* textured quad */ - draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, - format, type, unpack, pixels); + struct pipe_mipmap_tree *mt + = make_mipmap_tree(ctx->st, width, height, format, type, + unpack, pixels); + draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], + width, height, mt); + free_mipmap_tree(st->pipe, mt); } else { /* blit */ -- cgit v1.2.3 From ccff14de0d6291aa0866ce5d207af416caec69e7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 16:52:38 -0600 Subject: Simple implementation of glBitmap rendering. Create a texture matching the bitmap image and use a fragment program to modulate current raster color by the boolean-valued texture. Need to eventually use fragment culling (see comments in code). --- src/mesa/state_tracker/st_cb_drawpixels.c | 364 ++++++++++++++++++++++++++---- 1 file changed, 320 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c30d4ede4a..49120a4dea 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -55,36 +55,79 @@ * the fragment color. */ static struct st_fragment_program * -make_fragment_shader(struct st_context *st) +make_fragment_shader(struct st_context *st, GLboolean bitmapMode) { GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; struct gl_program *p; + GLuint ic = 0; p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) return NULL; - p->NumInstructions = 2; - p->Instructions = _mesa_alloc_instructions(2); + if (bitmapMode) + p->NumInstructions = 3; + else + p->NumInstructions = 2; + + p->Instructions = _mesa_alloc_instructions(p->NumInstructions); if (!p->Instructions) { ctx->Driver.DeleteProgram(ctx, p); return NULL; } - _mesa_init_instructions(p->Instructions, 2); - /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ - p->Instructions[0].Opcode = OPCODE_TEX; - p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[0].DstReg.Index = FRAG_RESULT_COLR; - p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[0].SrcReg[0].Index = FRAG_ATTRIB_TEX0; - p->Instructions[0].TexSrcUnit = 0; - p->Instructions[0].TexSrcTarget = TEXTURE_2D_INDEX; + _mesa_init_instructions(p->Instructions, p->NumInstructions); + if (bitmapMode) { + /* + * XXX This is temporary + * We actually need to cull the fragment if the texture value is zero. + * But TGSI doesn't support conditionals yet. + * Also, we need to compose this fragment shader with the current + * user-provided fragment shader so the fragment program is applied + * to the fragments which aren't culled. + */ + /* TEX temp0, fragment.texcoord[0], texture[0], 2D; */ + p->Instructions[ic].Opcode = OPCODE_TEX; + p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; + p->Instructions[ic].DstReg.Index = 0; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + p->Instructions[ic].TexSrcUnit = 0; + p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; + /* MUL result.color, temp0.xxxx, fragment.color */ + p->Instructions[ic].Opcode = OPCODE_MUL; + p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; + p->Instructions[ic].SrcReg[1].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[1].Index = FRAG_ATTRIB_COL0; + ic++; + } + else { + /* DrawPixels mode */ + /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ + p->Instructions[0].Opcode = OPCODE_TEX; + p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[0].DstReg.Index = FRAG_RESULT_COLR; + p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[0].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + p->Instructions[0].TexSrcUnit = 0; + p->Instructions[0].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; + } /* END; */ - p->Instructions[1].Opcode = OPCODE_END; + p->Instructions[ic++].Opcode = OPCODE_END; + + assert(ic == p->NumInstructions); p->InputsRead = FRAG_BIT_TEX0; p->OutputsWritten = (1 << FRAG_RESULT_COLR); + if (bitmapMode) { + p->InputsRead |= FRAG_BIT_COL0; + } stfp = (struct st_fragment_program *) p; st_translate_fragment_program(st, stfp, NULL, @@ -96,45 +139,67 @@ make_fragment_shader(struct st_context *st) /** * Create a simple vertex shader that just passes through the - * vertex position and color. + * vertex position and texcoord (and color). */ static struct st_vertex_program * -make_vertex_shader(struct st_context *st) +make_vertex_shader(struct st_context *st, GLboolean passColor) { - /* Map VERT_RESULT_HPOS to 0, VERT_RESULT_TEX0 to 1 */ GLcontext *ctx = st->ctx; struct st_vertex_program *stvp; struct gl_program *p; + GLuint ic = 0; p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); if (!p) return NULL; - p->NumInstructions = 3; - p->Instructions = _mesa_alloc_instructions(3); + if (passColor) + p->NumInstructions = 4; + else + p->NumInstructions = 3; + + p->Instructions = _mesa_alloc_instructions(p->NumInstructions); if (!p->Instructions) { ctx->Driver.DeleteProgram(ctx, p); return NULL; } - _mesa_init_instructions(p->Instructions, 3); + _mesa_init_instructions(p->Instructions, p->NumInstructions); /* MOV result.pos, vertex.pos; */ p->Instructions[0].Opcode = OPCODE_MOV; p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; p->Instructions[0].DstReg.Index = VERT_RESULT_HPOS; p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; p->Instructions[0].SrcReg[0].Index = VERT_ATTRIB_POS; - /* MOV result.color, vertex.color; */ + /* MOV result.texcoord0, vertex.texcoord0; */ p->Instructions[1].Opcode = OPCODE_MOV; p->Instructions[1].DstReg.File = PROGRAM_OUTPUT; p->Instructions[1].DstReg.Index = VERT_RESULT_TEX0; p->Instructions[1].SrcReg[0].File = PROGRAM_INPUT; p->Instructions[1].SrcReg[0].Index = VERT_ATTRIB_TEX0; + ic = 2; + if (passColor) { + /* MOV result.color0, vertex.color0; */ + p->Instructions[ic].Opcode = OPCODE_MOV; + p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[ic].DstReg.Index = VERT_RESULT_COL0; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = VERT_ATTRIB_COLOR0; + ic++; + } + /* END; */ - p->Instructions[2].Opcode = OPCODE_END; + p->Instructions[ic].Opcode = OPCODE_END; + ic++; + + assert(ic == p->NumInstructions); p->InputsRead = VERT_BIT_POS | VERT_BIT_TEX0; p->OutputsWritten = ((1 << VERT_RESULT_TEX0) | (1 << VERT_RESULT_HPOS)); + if (passColor) { + p->InputsRead |= VERT_BIT_COLOR0; + p->OutputsWritten |= (1 << VERT_RESULT_COL0); + } stvp = (struct st_vertex_program *) p; st_translate_vertex_program(st, stvp, NULL, @@ -255,10 +320,10 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, GLuint i; /* upper-left */ - verts[0][0][0] = x0; - verts[0][0][1] = y0; - verts[0][1][0] = 0.0; - verts[0][1][1] = 0.0; + verts[0][0][0] = x0; /* attr[0].x */ + verts[0][0][1] = y0; /* attr[0].x */ + verts[0][1][0] = 0.0; /* attr[1].s */ + verts[0][1][1] = 0.0; /* attr[1].t */ /* upper-right */ verts[1][0][0] = x1; @@ -290,10 +355,66 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, } +static void +draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, + GLfloat x1, GLfloat y1, const GLfloat *color) +{ + static const GLuint attribs[3] = { + 0, /* pos */ + 1, /* color */ + 8 /* tex0 */ + }; + GLfloat verts[4][3][4]; /* four verts, three attribs, XYZW */ + GLuint i; + + /* upper-left */ + verts[0][0][0] = x0; /* attr[0].x */ + verts[0][0][1] = y0; /* attr[0].x */ + verts[0][2][0] = 0.0; /* attr[2].s */ + verts[0][2][1] = 0.0; /* attr[2].t */ + + /* upper-right */ + verts[1][0][0] = x1; + verts[1][0][1] = y0; + verts[1][2][0] = 1.0; + verts[1][2][1] = 0.0; + + /* lower-right */ + verts[2][0][0] = x1; + verts[2][0][1] = y1; + verts[2][2][0] = 1.0; + verts[2][2][1] = 1.0; + + /* lower-left */ + verts[3][0][0] = x0; + verts[3][0][1] = y1; + verts[3][2][0] = 0.0; + verts[3][2][1] = 1.0; + + /* same for all verts: */ + for (i = 0; i < 4; i++) { + verts[i][0][2] = z; /*Z*/ + verts[i][0][3] = 1.0; /*W*/ + verts[i][1][0] = color[0]; + verts[i][1][1] = color[1]; + verts[i][1][2] = color[2]; + verts[i][1][3] = color[3]; + verts[i][2][2] = 0.0; /*R*/ + verts[i][2][3] = 1.0; /*Q*/ + } + + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 3, attribs); +} + + + static void draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, GLsizei width, GLsizei height, - struct pipe_mipmap_tree *mt) + struct pipe_mipmap_tree *mt, + struct st_vertex_program *stvp, + struct st_fragment_program *stfp, + const GLfloat *color) { const GLuint unit = 0; struct pipe_context *pipe = ctx->st->pipe; @@ -320,22 +441,10 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, } /* fragment shader state: TEX lookup program */ - { - static struct st_fragment_program *stfp = NULL; - if (!stfp) { - stfp = make_fragment_shader(ctx->st); - } - pipe->bind_fs_state(pipe, stfp->fs->data); - } + pipe->bind_fs_state(pipe, stfp->fs->data); /* vertex shader state: position + texcoord pass-through */ - { - static struct st_vertex_program *stvp = NULL; - if (!stvp) { - stvp = make_vertex_shader(ctx->st); - } - pipe->bind_vs_state(pipe, stvp->vs->data); - } + pipe->bind_vs_state(pipe, stvp->vs->data); /* texture sampling state: */ { @@ -383,7 +492,10 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, y1 = y + height * ctx->Pixel.ZoomY; /* draw textured quad */ - draw_quad(ctx, x0, y0, z, x1, y1); + if (color) + draw_quad_colored(ctx, x0, y0, z, x1, y1, color); + else + draw_quad(ctx, x0, y0, z, x1, y1); /* restore GL state */ pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); @@ -514,10 +626,21 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { + static struct st_fragment_program *stfp = NULL; + static struct st_vertex_program *stvp = NULL; struct st_context *st = ctx->st; struct pipe_surface *ps; GLuint bufferFormat; + /* create the fragment program if needed */ + if (!stfp) { + stfp = make_fragment_shader(ctx->st, GL_FALSE); + } + /* and vertex program */ + if (!stvp) { + stvp = make_vertex_shader(ctx->st, GL_FALSE); + } + st_validate_state(st); if (format == GL_DEPTH_COMPONENT) { @@ -540,7 +663,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, = make_mipmap_tree(ctx->st, width, height, format, type, unpack, pixels); draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], - width, height, mt); + width, height, mt, stvp, stfp, NULL); free_mipmap_tree(st->pipe, mt); } else { @@ -550,16 +673,169 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } + +/** + * Create a texture which represents a bitmap image. + */ +static struct pipe_mipmap_tree * +create_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap) +{ + struct pipe_context *pipe = ctx->st->pipe; + const uint flags = PIPE_SURFACE_FLAG_TEXTURE; + uint numFormats, i, format = 0, cpp, comp, pitch; + const uint *formats = ctx->st->pipe->supported_formats(pipe, &numFormats); + ubyte *dest; + struct pipe_mipmap_tree *mt; + int row, col; + + /* find a texture format we know */ + for (i = 0; i < numFormats; i++) { + switch (formats[i]) { + case PIPE_FORMAT_U_I8: + format = formats[i]; + cpp = 1; + comp = 0; + break; + case PIPE_FORMAT_U_A8_R8_G8_B8: + format = formats[i]; + cpp = 4; + comp = 3; /* alpha channel */ /*XXX little-endian dependency */ + break; + default: + /* XXX support more formats */ + ; + } + } + assert(format); + + /** + * Create a mipmap tree. + */ + mt = CALLOC_STRUCT(pipe_mipmap_tree); + + /* allocate texture region/storage */ + mt->region = pipe->region_alloc(pipe, cpp, width, height, flags); + pitch = mt->region->pitch; + + /* map texture region */ + dest = pipe->region_map(pipe, mt->region); + + /* Put image into texture region. + * Note that the image is actually going to be upside down in + * the texture. We deal with that with texcoords. + */ + + for (row = 0; row < height; row++) { + const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, + bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); + ubyte *destRow = dest + row * pitch * cpp; + + if (unpack->LsbFirst) { + /* Lsb first */ + GLubyte mask = 1U << (unpack->SkipPixels & 0x7); + assert(0); + for (col = 0; col < width; col++) { + + /* set texel to 255 if bit is set */ + destRow[comp] = 255; //(*src & mask) ? 255 : 0; + destRow += cpp; + + if (mask == 128U) { + src++; + mask = 1U; + } + else { + mask = mask << 1; + } + } + + /* get ready for next row */ + if (mask != 1) + src++; + } + else { + /* Msb first */ + GLubyte mask = 128U >> (unpack->SkipPixels & 0x7); + for (col = 0; col < width; col++) { + + /* set texel to 255 if bit is set */ + destRow[comp] =(*src & mask) ? 255 : 0; + destRow += cpp; + + if (mask == 1U) { + src++; + mask = 128U; + } + else { + mask = mask >> 1; + } + } + + /* get ready for next row */ + if (mask != 128) + src++; + } + + } /* row */ + + /* unmap */ + pipe->region_unmap(pipe, mt->region); + + mt->target = PIPE_TEXTURE_2D; + mt->internal_format = GL_RGBA; + mt->format = format; + mt->first_level = 0; + mt->last_level = 0; + mt->width0 = width; + mt->height0 = height; + mt->depth0 = 1; + mt->cpp = cpp; + mt->compressed = 0; + mt->pitch = mt->region->pitch; + mt->depth_pitch = 0; + mt->total_height = height; + mt->level[0].level_offset = 0; + mt->level[0].width = width; + mt->level[0].height = height; + mt->level[0].depth = 1; + mt->level[0].nr_images = 1; + mt->level[0].image_offset = NULL; + mt->refcount = 1; + + return mt; +} + + + static void st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap ) { + static struct st_vertex_program *stvp = NULL; + static struct st_fragment_program *stfp = NULL; struct st_context *st = ctx->st; + struct pipe_mipmap_tree *mt; + + /* create the fragment program if needed */ + if (!stfp) { + stfp = make_fragment_shader(ctx->st, GL_TRUE); + } + /* and vertex program */ + if (!stvp) { + stvp = make_vertex_shader(ctx->st, GL_TRUE); + } st_validate_state(st); - fprintf(stderr, "st_Bitmap not implemented yet\n"); - /* XXX to do */ + mt = create_bitmap_texture(ctx, width, height, unpack, bitmap); + + draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], + width, height, mt, stvp, stfp, + ctx->Current.RasterColor); + + free_mipmap_tree(st->pipe, mt); } -- cgit v1.2.3 From 0dfa5506a318b202ac955a59cc7c9b22b5ff3867 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 16:56:35 -0600 Subject: st_draw_vertices() no longer needs attribs[] array parameter --- src/mesa/state_tracker/st_cb_clear.c | 6 +----- src/mesa/state_tracker/st_cb_drawpixels.c | 13 ++----------- src/mesa/state_tracker/st_draw.c | 5 ++--- src/mesa/state_tracker/st_draw.h | 2 +- 4 files changed, 6 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 367ae06cf3..3a70d2f121 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -223,10 +223,6 @@ draw_quad(GLcontext *ctx, float x0, float y0, float x1, float y1, GLfloat z, const GLfloat color[4]) { - static const GLuint attribs[2] = { - 0, /* pos */ - 3 /* color */ - }; GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ GLuint i; @@ -253,7 +249,7 @@ draw_quad(GLcontext *ctx, verts[i][1][3] = color[3]; } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, attribs); + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 49120a4dea..9efd718834 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -312,10 +312,6 @@ static void draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, GLfloat x1, GLfloat y1) { - static const GLuint attribs[2] = { - 0, /* pos */ - 8 /* tex0 */ - }; GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ GLuint i; @@ -351,7 +347,7 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, verts[i][1][3] = 1.0; /*Q*/ } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, attribs); + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2); } @@ -359,11 +355,6 @@ static void draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, GLfloat x1, GLfloat y1, const GLfloat *color) { - static const GLuint attribs[3] = { - 0, /* pos */ - 1, /* color */ - 8 /* tex0 */ - }; GLfloat verts[4][3][4]; /* four verts, three attribs, XYZW */ GLuint i; @@ -403,7 +394,7 @@ draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, verts[i][2][3] = 1.0; /*Q*/ } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 3, attribs); + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 3); } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 0f45bf579a..ce5bf0c8a9 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -316,7 +316,7 @@ st_draw_vbo(GLcontext *ctx, void st_draw_vertices(GLcontext *ctx, unsigned prim, unsigned numVertex, float *verts, - unsigned numAttribs, const unsigned attribs[]) + unsigned numAttribs) { const float width = ctx->DrawBuffer->Width; const float height = ctx->DrawBuffer->Height; @@ -328,7 +328,6 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, unsigned i; assert(numAttribs > 0); - assert(attribs[0] == 0); /* position */ /* convert to clip coords */ for (i = 0; i < numVertex; i++) { @@ -356,7 +355,7 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, velement.vertex_buffer_index = 0; velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; velement.dst_offset = 0; - pipe->set_vertex_element(pipe, i/**attribs[i]**/, &velement); + pipe->set_vertex_element(pipe, i, &velement); } /* draw */ diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h index 1fcd1b7e6b..89ee790c57 100644 --- a/src/mesa/state_tracker/st_draw.h +++ b/src/mesa/state_tracker/st_draw.h @@ -62,7 +62,7 @@ st_feedback_draw_vbo(GLcontext *ctx, void st_draw_vertices(GLcontext *ctx, unsigned prim, unsigned numVertex, float *verts, - unsigned numAttribs, const unsigned attribs[]); + unsigned numAttribs); #endif -- cgit v1.2.3 From b28f4a27d586b3ba3a7e87db761b6d624dd3e672 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 17:03:19 -0600 Subject: don't apply pixelzoom to bitmaps --- src/mesa/state_tracker/st_cb_drawpixels.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 9efd718834..30449658e9 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -402,6 +402,7 @@ draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, static void draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, GLsizei width, GLsizei height, + GLfloat zoomX, GLfloat zoomY, struct pipe_mipmap_tree *mt, struct st_vertex_program *stvp, struct st_fragment_program *stfp, @@ -654,7 +655,8 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, = make_mipmap_tree(ctx->st, width, height, format, type, unpack, pixels); draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], - width, height, mt, stvp, stfp, NULL); + width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, + mt, stvp, stfp, NULL); free_mipmap_tree(st->pipe, mt); } else { @@ -823,7 +825,8 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, mt = create_bitmap_texture(ctx, width, height, unpack, bitmap); draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], - width, height, mt, stvp, stfp, + width, height, 1.0, 1.0, + mt, stvp, stfp, ctx->Current.RasterColor); free_mipmap_tree(st->pipe, mt); -- cgit v1.2.3 From 960fe21008eb00bf778f82476e3d00df0a5c34db Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Sep 2007 17:07:59 -0600 Subject: comments, null ptr checks, etc. --- src/mesa/state_tracker/st_cb_drawpixels.c | 54 ++++++++++++++++++++++--------- 1 file changed, 38 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 30449658e9..fb89ceef8c 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -209,7 +209,9 @@ make_vertex_shader(struct st_context *st, GLboolean passColor) } - +/** + * Make mipmap tree containing the glDrawPixels image. + */ static struct pipe_mipmap_tree * make_mipmap_tree(struct st_context *st, GLsizei width, GLsizei height, GLenum format, GLenum type, @@ -230,11 +232,14 @@ make_mipmap_tree(struct st_context *st, cpp = st_sizeof_format(pipeFormat); mt = CALLOC_STRUCT(pipe_mipmap_tree); + if (!mt) + return NULL; if (unpack->BufferObj && unpack->BufferObj->Name) { /* mt->region = buffer_object_region(unpack->BufferObj); */ + printf("st_DrawPixels (sourcing from PBO not implemented yet)\n"); } else { static const GLuint dstImageOffsets = 0; @@ -654,10 +659,12 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct pipe_mipmap_tree *mt = make_mipmap_tree(ctx->st, width, height, format, type, unpack, pixels); - draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], - width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, - mt, stvp, stfp, NULL); - free_mipmap_tree(st->pipe, mt); + if (mt) { + draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], + width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, + mt, stvp, stfp, NULL); + free_mipmap_tree(st->pipe, mt); + } } else { /* blit */ @@ -671,9 +678,9 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, * Create a texture which represents a bitmap image. */ static struct pipe_mipmap_tree * -create_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, - const GLubyte *bitmap) +make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap) { struct pipe_context *pipe = ctx->st->pipe; const uint flags = PIPE_SURFACE_FLAG_TEXTURE; @@ -707,6 +714,16 @@ create_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, * Create a mipmap tree. */ mt = CALLOC_STRUCT(pipe_mipmap_tree); + if (!mt) + return NULL; + + if (unpack->BufferObj && unpack->BufferObj->Name) { + /* + mt->region = buffer_object_region(unpack->BufferObj); + */ + printf("st_Bitmap (sourcing from PBO not implemented yet)\n"); + } + /* allocate texture region/storage */ mt->region = pipe->region_alloc(pipe, cpp, width, height, flags); @@ -714,6 +731,10 @@ create_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, /* map texture region */ dest = pipe->region_map(pipe, mt->region); + if (!dest) { + printf("st_Bitmap region_map() failed!?!"); + return NULL; + } /* Put image into texture region. * Note that the image is actually going to be upside down in @@ -732,7 +753,7 @@ create_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, for (col = 0; col < width; col++) { /* set texel to 255 if bit is set */ - destRow[comp] = 255; //(*src & mask) ? 255 : 0; + destRow[comp] = (*src & mask) ? 255 : 0; destRow += cpp; if (mask == 128U) { @@ -822,14 +843,15 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, st_validate_state(st); - mt = create_bitmap_texture(ctx, width, height, unpack, bitmap); - - draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], - width, height, 1.0, 1.0, - mt, stvp, stfp, - ctx->Current.RasterColor); + mt = make_bitmap_texture(ctx, width, height, unpack, bitmap); + if (mt) { + draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], + width, height, 1.0, 1.0, + mt, stvp, stfp, + ctx->Current.RasterColor); - free_mipmap_tree(st->pipe, mt); + free_mipmap_tree(st->pipe, mt); + } } -- cgit v1.2.3 From 715e5edff6ade0a1ef02c8f8d9e4d61316113721 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 15:34:31 -0600 Subject: better debugging --- src/mesa/state_tracker/st_atom_constbuf.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index cf52a923d3..5285485c94 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -32,6 +32,7 @@ #include "main/imports.h" #include "shader/prog_parameter.h" +#include "shader/prog_print.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -69,12 +70,15 @@ static void upload_constants( struct st_context *st, printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", __FUNCTION__, id, params->NumParameters, params->StateFlags); + _mesa_print_parameter_list(params); + /* for (i = 0; i < params->NumParameters; i++) printf("%d: %f %f %f %f\n", i, params->ParameterValues[i][0], params->ParameterValues[i][1], params->ParameterValues[i][2], params->ParameterValues[i][3]); + */ } /* load Mesa constants into the constant buffer */ -- cgit v1.2.3 From bad4e10af746ce16b730a3e7a4e2ff53ecb6d0f6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 15:34:51 -0600 Subject: check for _NEW_PROGRAM mesa state (fixes arbfplight) --- src/mesa/state_tracker/st_atom_shader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 383184e575..9bba78f7a1 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -415,7 +415,7 @@ update_linkage( struct st_context *st ) const struct st_tracked_state st_update_shader = { .name = "st_update_shader", .dirty = { - .mesa = 0, + .mesa = _NEW_PROGRAM, .st = ST_NEW_SHADER }, .update = update_linkage -- cgit v1.2.3 From 324ecadbfdf9b944e059832f146451e4151dcb21 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 17:03:11 -0600 Subject: Added new _mesa_clip_copytexsubimage() function to do avoid clipping down in the drivers. This should probably be pulled into main-line Mesa... --- src/mesa/main/image.c | 31 +++++++++++++++++++++++++++++++ src/mesa/main/image.h | 6 ++++++ src/mesa/main/teximage.c | 30 +++++++++++++++++++++++------- 3 files changed, 60 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c index ba46cdc1b1..ae3c82b810 100644 --- a/src/mesa/main/image.c +++ b/src/mesa/main/image.c @@ -4624,6 +4624,37 @@ _mesa_clip_readpixels(const GLcontext *ctx, } +/** + * Do clipping for a glCopyTexSubImage call. + * The framebuffer source region might extend outside the framebuffer + * bounds. Clip the source region against the framebuffer bounds and + * adjust the texture/dest position and size accordingly. + * + * \return GL_FALSE if region is totally clipped, GL_TRUE otherwise. + */ +GLboolean +_mesa_clip_copytexsubimage(const GLcontext *ctx, + GLint *destX, GLint *destY, + GLint *srcX, GLint *srcY, + GLsizei *width, GLsizei *height) +{ + const struct gl_framebuffer *fb = ctx->ReadBuffer; + const GLint srcX0 = *srcX, srcY0 = *srcY; + + if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, + srcX, srcY, width, height)) { + *destX = *destX + *srcX - srcX0; + *destY = *destY + *srcY - srcY0; + + return GL_TRUE; + } + else { + return GL_FALSE; + } +} + + + /** * Clip the rectangle defined by (x, y, width, height) against the bounds * specified by [xmin, xmax) and [ymin, ymax). diff --git a/src/mesa/main/image.h b/src/mesa/main/image.h index 2a16989fa7..c379d5fa7d 100644 --- a/src/mesa/main/image.h +++ b/src/mesa/main/image.h @@ -224,6 +224,12 @@ _mesa_clip_readpixels(const GLcontext *ctx, GLsizei *width, GLsizei *height, struct gl_pixelstore_attrib *pack); +extern GLboolean +_mesa_clip_copytexsubimage(const GLcontext *ctx, + GLint *destX, GLint *destY, + GLint *srcX, GLint *srcY, + GLsizei *width, GLsizei *height); + extern GLboolean _mesa_clip_to_region(GLint xmin, GLint ymin, GLint xmax, GLint ymax, diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c index 3420d8e2ba..09ec0d4553 100644 --- a/src/mesa/main/teximage.c +++ b/src/mesa/main/teximage.c @@ -3024,6 +3024,9 @@ _mesa_CopyTexSubImage1D( GLenum target, GLint level, struct gl_texture_object *texObj; struct gl_texture_image *texImage; GLsizei postConvWidth = width; + GLint yoffset = 0; + GLsizei height = 1; + GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); @@ -3053,8 +3056,13 @@ _mesa_CopyTexSubImage1D( GLenum target, GLint level, /* If we have a border, xoffset=-1 is legal. Bias by border width */ xoffset += texImage->Border; - ASSERT(ctx->Driver.CopyTexSubImage1D); - (*ctx->Driver.CopyTexSubImage1D)(ctx, target, level, xoffset, x, y, width); + if (_mesa_clip_copytexsubimage(ctx, &xoffset, &yoffset, &x, &y, + &width, &height)) { + ASSERT(ctx->Driver.CopyTexSubImage1D); + ctx->Driver.CopyTexSubImage1D(ctx, target, level, + xoffset, x, y, width); + } + ctx->NewState |= _NEW_TEXTURE; } out: @@ -3099,10 +3107,14 @@ _mesa_CopyTexSubImage2D( GLenum target, GLint level, /* If we have a border, xoffset=-1 is legal. Bias by border width */ xoffset += texImage->Border; yoffset += texImage->Border; - - ASSERT(ctx->Driver.CopyTexSubImage2D); - (*ctx->Driver.CopyTexSubImage2D)(ctx, target, level, + + if (_mesa_clip_copytexsubimage(ctx, &xoffset, &yoffset, &x, &y, + &width, &height)) { + ASSERT(ctx->Driver.CopyTexSubImage2D); + ctx->Driver.CopyTexSubImage2D(ctx, target, level, xoffset, yoffset, x, y, width, height); + } + ctx->NewState |= _NEW_TEXTURE; } out: @@ -3150,10 +3162,14 @@ _mesa_CopyTexSubImage3D( GLenum target, GLint level, yoffset += texImage->Border; zoffset += texImage->Border; - ASSERT(ctx->Driver.CopyTexSubImage3D); - (*ctx->Driver.CopyTexSubImage3D)(ctx, target, level, + if (_mesa_clip_copytexsubimage(ctx, &xoffset, &yoffset, &x, &y, + &width, &height)) { + ASSERT(ctx->Driver.CopyTexSubImage3D); + ctx->Driver.CopyTexSubImage3D(ctx, target, level, xoffset, yoffset, zoffset, x, y, width, height); + } + ctx->NewState |= _NEW_TEXTURE; } out: -- cgit v1.2.3 From b3bb1e3ebf260f18a2097a9a8fe8ab52f104dab1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 17:03:40 -0600 Subject: s/_mesa_copy_rect/copy_rect/ --- src/mesa/pipe/softpipe/sp_region.c | 51 +++++++++++++++++++------------------- 1 file changed, 26 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 697b738e43..faf2737d6b 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -123,21 +123,22 @@ sp_region_release(struct pipe_context *pipe, struct pipe_region **region) } -/* - * XXX Move this into core Mesa? +/** + * Copy 2D rect from one place to another. + * Position and sizes are in pixels. */ static void -_mesa_copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte * src, - unsigned src_pitch, - unsigned src_x, - unsigned src_y) +copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + unsigned src_pitch, + unsigned src_x, + unsigned src_y) { unsigned i; @@ -176,10 +177,10 @@ sp_region_data(struct pipe_context *pipe, const void *src, unsigned src_pitch, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); + copy_rect(pipe->region_map(pipe, dst) + dst_offset, + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); pipe->region_unmap(pipe, dst); } @@ -198,14 +199,14 @@ sp_region_copy(struct pipe_context *pipe, { assert( dst->cpp == src->cpp ); - _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - pipe->region_map(pipe, src) + src_offset, - src->pitch, - srcx, srcy); + copy_rect(pipe->region_map(pipe, dst) + dst_offset, + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe->region_map(pipe, src) + src_offset, + src->pitch, + srcx, srcy); pipe->region_unmap(pipe, src); pipe->region_unmap(pipe, dst); -- cgit v1.2.3 From b27498c7cad7dae4e3b3ef9517b9c0aef58f73f6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 17:18:42 -0600 Subject: checkpoint: glCopyTexImage work --- src/mesa/state_tracker/st_cb_texture.c | 163 +++++++++++++++------------------ 1 file changed, 73 insertions(+), 90 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 3ae9bd9003..ad9b52e21a 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -37,6 +37,7 @@ #include "main/texstore.h" #include "state_tracker/st_context.h" +#include "state_tracker/st_cb_fbo.h" #include "state_tracker/st_cb_texture.h" #include "state_tracker/st_format.h" #include "state_tracker/st_mipmap_tree.h" @@ -1049,63 +1050,39 @@ st_TexSubImage1D(GLcontext * ctx, /** - * Get the pipe_region which is the source for any glCopyTex[Sub]Image call. - * - * Do the best we can using the blitter. A future project is to use - * the texture engine and fragment programs for these copies. + * Do a CopyTex[Sub]Image using an optimized hardware (blit) path. + * Note that the region to copy has already been clip tested. + * \return GL_TRUE if success, GL_FALSE if failure (use a fallback) */ -static const struct pipe_region * -get_teximage_source(GLcontext *ctx, GLenum internalFormat) -{ -#if 00 - struct intel_renderbuffer *irb; - - DBG("%s %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(internalFormat)); - - switch (internalFormat) { - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16_ARB: - irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); - if (irb && irb->region && irb->region->cpp == 2) - return irb->region; - return NULL; - case GL_DEPTH24_STENCIL8_EXT: - case GL_DEPTH_STENCIL_EXT: - irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); - if (irb && irb->region && irb->region->cpp == 4) - return irb->region; - return NULL; - case GL_RGBA: - case GL_RGBA8: - return intel_readbuf_region(intel); - case GL_RGB: - if (intel->intelScreen->front.cpp == 2) - return intel_readbuf_region(intel); - return NULL; - default: - return NULL; - } -#else - return NULL; -#endif -} - - static GLboolean do_copy_texsubimage(GLcontext *ctx, struct st_texture_image *stImage, - GLenum internalFormat, - GLint dstx, GLint dsty, - GLint x, GLint y, GLsizei width, GLsizei height) + GLenum baseFormat, + GLint destX, GLint destY, + GLint srcX, GLint srcY, GLsizei width, GLsizei height) { - const struct pipe_region *src = - get_teximage_source(ctx, internalFormat); + struct gl_framebuffer *fb = ctx->ReadBuffer; + struct st_renderbuffer *strb; + struct pipe_context *pipe = ctx->st->pipe; + struct pipe_region *src_region, *dest_region; + uint dest_offset, src_offset; + uint dest_format, src_format; - if (!stImage->mt || !src) { - DBG("%s fail %p %p\n", __FUNCTION__, (void *) stImage->mt, (void *) src); - return GL_FALSE; + /* determine if copying depth or color data */ + if (baseFormat == GL_DEPTH_COMPONENT) { + strb = st_renderbuffer(fb->_DepthBuffer); + } + else if (baseFormat == GL_DEPTH_STENCIL_EXT) { + strb = st_renderbuffer(fb->_StencilBuffer); } + else { + /* baseFormat == GL_RGB, GL_RGBA, GL_ALPHA, etc */ + strb = st_renderbuffer(fb->_ColorReadBuffer); + } + + assert(strb); + assert(strb->surface); + assert(stImage->mt); #if 00 /* XXX FIX flush/locking */ intelFlush(ctx); @@ -1113,49 +1090,55 @@ do_copy_texsubimage(GLcontext *ctx, LOCK_HARDWARE(intel); #endif - { - GLuint image_offset = st_miptree_image_offset(stImage->mt, - stImage->face, - stImage->level); - const GLint orig_x = x; - const GLint orig_y = y; - const struct gl_framebuffer *fb = ctx->DrawBuffer; - - if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, - &x, &y, &width, &height)) { - /* Update dst for clipped src. Need to also clip the source rect. - */ - dstx += x - orig_x; - dsty += y - orig_y; + src_format = strb->surface->format; + dest_format = stImage->mt->format; + if (src_format != dest_format) + return GL_FALSE; - if (!(ctx->ReadBuffer->Name == 0)) { - /* XXX this looks bogus ? */ - /* FBO: invert Y */ - y = ctx->ReadBuffer->Height - y - 1; - } + src_region = strb->surface->region; + dest_region = stImage->mt->region; + if (!src_region || !dest_region) + return GL_FALSE; + if (src_region->cpp != dest_region->cpp) + return GL_FALSE; + + src_offset = 0; + dest_offset = st_miptree_image_offset(stImage->mt, + stImage->face, + stImage->level); + + /* XXX may need to invert image depending on window vs. user-created FBO */ - /* A bit of fiddling to get the blitter to work with -ve - * pitches. But we get a nice inverted blit this way, so it's - * worth it: - */ #if 0 - intelEmitCopyBlit(intel, - stImage->mt->cpp, - -src->pitch, - src->buffer, - src->height * src->pitch * src->cpp, - stImage->mt->pitch, - stImage->mt->region->buffer, - image_offset, - x, y + height, dstx, dsty, width, height, - GL_COPY); /* ? */ - intel_batchbuffer_flush(intel->batch); + /* A bit of fiddling to get the blitter to work with -ve + * pitches. But we get a nice inverted blit this way, so it's + * worth it: + */ + intelEmitCopyBlit(intel, + stImage->mt->cpp, + -src->pitch, + src->buffer, + src->height * src->pitch * src->cpp, + stImage->mt->pitch, + stImage->mt->region->buffer, + dest_offset, + x, y + height, dstx, dsty, width, height, + GL_COPY); /* ? */ + intel_batchbuffer_flush(intel->batch); #else - /* XXX use pipe->region_copy() ??? */ - (void) image_offset; + + pipe->region_copy(pipe, + /* dest */ + dest_region, + dest_offset, + destX, destY, + /* src */ + src_region, + src_offset, + srcX, srcY, + /* size */ + width, height); #endif - } - } #if 0 UNLOCK_HARDWARE(intel); @@ -1266,7 +1249,7 @@ st_CopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, _mesa_select_tex_object(ctx, texUnit, target); struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); - GLenum internalFormat = texImage->InternalFormat; + const GLenum baseFormat = texImage->TexFormat->BaseFormat; /* XXX need to check as in above function? */ @@ -1275,7 +1258,7 @@ st_CopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, if (!do_copy_texsubimage(ctx, st_texture_image(texImage), - internalFormat, xoffset, 0, x, y, width, 1)) { + baseFormat, xoffset, 0, x, y, width, 1)) { #if 0 _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); #endif -- cgit v1.2.3 From 038cb561eb094af2f2ba06e18e61246fc0c87c3c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 17:57:15 -0600 Subject: More work on glCopyTexSubImage. Start sketching out a fallback path based on surface->get_tile(), put_tile() which will do format convertion and GL's pixel transfer ops. --- src/mesa/state_tracker/st_cb_texture.c | 274 +++++++++++++++++++-------------- 1 file changed, 157 insertions(+), 117 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index ad9b52e21a..052035f988 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1049,18 +1049,83 @@ st_TexSubImage1D(GLcontext * ctx, +/** + * Return 0 for GL_TEXTURE_CUBE_MAP_POSITIVE_X, + * 1 for GL_TEXTURE_CUBE_MAP_NEGATIVE_X, + * etc. + * XXX duplicated from main/teximage.c + */ +static uint +texture_face(GLenum target) +{ + if (target >= GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB && + target <= GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB) + return (GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X; + else + return 0; +} + + + +/** + * Do a CopyTexSubImage operation by mapping the source region and + * dest region and copying/converting pixels. + */ +static void +fallback_copy_texsubimage(GLcontext *ctx, + GLenum target, + GLint level, + struct st_renderbuffer *strb, + struct st_texture_image *stImage, + GLenum baseFormat, + GLint destX, GLint destY, GLint destZ, + GLint srcX, GLint srcY, + GLsizei width, GLsizei height) +{ + struct pipe_context *pipe = ctx->st->pipe; + const uint face = texture_face(target); + struct pipe_mipmap_tree *mt = stImage->mt; + struct pipe_surface *src_surf, *dest_surf; + GLfloat *data; + + src_surf = strb->surface; + + dest_surf = pipe->get_tex_surface(pipe, mt, + face, level, destZ); + + data = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + src_surf->get_tile(src_surf, srcX, srcY, width, height, data); + + /* process pixels */ + + dest_surf->put_tile(dest_surf, destX, destY, width, height, data); + + free(data); +} + + + + /** * Do a CopyTex[Sub]Image using an optimized hardware (blit) path. * Note that the region to copy has already been clip tested. * \return GL_TRUE if success, GL_FALSE if failure (use a fallback) */ -static GLboolean +static void do_copy_texsubimage(GLcontext *ctx, - struct st_texture_image *stImage, - GLenum baseFormat, - GLint destX, GLint destY, - GLint srcX, GLint srcY, GLsizei width, GLsizei height) + GLenum target, GLint level, + GLint destX, GLint destY, GLint destZ, + GLint srcX, GLint srcY, + GLsizei width, GLsizei height) { + struct gl_texture_unit *texUnit = + &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + struct gl_texture_object *texObj = + _mesa_select_tex_object(ctx, texUnit, target); + struct gl_texture_image *texImage = + _mesa_select_tex_image(ctx, texObj, target, level); + struct st_texture_image *stImage = st_texture_image(texImage); + GLenum baseFormat = texImage->InternalFormat; struct gl_framebuffer *fb = ctx->ReadBuffer; struct st_renderbuffer *strb; struct pipe_context *pipe = ctx->st->pipe; @@ -1068,6 +1133,8 @@ do_copy_texsubimage(GLcontext *ctx, uint dest_offset, src_offset; uint dest_format, src_format; + (void) texImage; + /* determine if copying depth or color data */ if (baseFormat == GL_DEPTH_COMPONENT) { strb = st_renderbuffer(fb->_DepthBuffer); @@ -1084,65 +1151,75 @@ do_copy_texsubimage(GLcontext *ctx, assert(strb->surface); assert(stImage->mt); -#if 00 /* XXX FIX flush/locking */ - intelFlush(ctx); - /* XXX still need the lock ? */ - LOCK_HARDWARE(intel); -#endif - src_format = strb->surface->format; dest_format = stImage->mt->format; - if (src_format != dest_format) - return GL_FALSE; src_region = strb->surface->region; dest_region = stImage->mt->region; - if (!src_region || !dest_region) - return GL_FALSE; - if (src_region->cpp != dest_region->cpp) - return GL_FALSE; - src_offset = 0; - dest_offset = st_miptree_image_offset(stImage->mt, - stImage->face, - stImage->level); + if (src_format == dest_format && + /* XXX also check GL pixel transfer ops here */ + src_region && + dest_region && + src_region->cpp == dest_region->cpp) { + /* do blit-style copy */ + src_offset = 0; + dest_offset = st_miptree_image_offset(stImage->mt, + stImage->face, + stImage->level); + + /* XXX may need to invert image depending on window + * vs. user-created FBO + */ - /* XXX may need to invert image depending on window vs. user-created FBO */ +#if 00 /* XXX FIX flush/locking */ + intelFlush(ctx); + /* XXX still need the lock ? */ + LOCK_HARDWARE(intel); +#endif #if 0 - /* A bit of fiddling to get the blitter to work with -ve - * pitches. But we get a nice inverted blit this way, so it's - * worth it: - */ - intelEmitCopyBlit(intel, - stImage->mt->cpp, - -src->pitch, - src->buffer, - src->height * src->pitch * src->cpp, - stImage->mt->pitch, - stImage->mt->region->buffer, - dest_offset, - x, y + height, dstx, dsty, width, height, - GL_COPY); /* ? */ - intel_batchbuffer_flush(intel->batch); + /* A bit of fiddling to get the blitter to work with -ve + * pitches. But we get a nice inverted blit this way, so it's + * worth it: + */ + intelEmitCopyBlit(intel, + stImage->mt->cpp, + -src->pitch, + src->buffer, + src->height * src->pitch * src->cpp, + stImage->mt->pitch, + stImage->mt->region->buffer, + dest_offset, + x, y + height, dstx, dsty, width, height, + GL_COPY); /* ? */ + intel_batchbuffer_flush(intel->batch); #else - pipe->region_copy(pipe, - /* dest */ - dest_region, - dest_offset, - destX, destY, - /* src */ - src_region, - src_offset, - srcX, srcY, - /* size */ - width, height); + pipe->region_copy(pipe, + /* dest */ + dest_region, + dest_offset, + destX, destY, + /* src */ + src_region, + src_offset, + srcX, srcY, + /* size */ + width, height); #endif #if 0 - UNLOCK_HARDWARE(intel); + UNLOCK_HARDWARE(intel); #endif + } + else { + fallback_copy_texsubimage(ctx, target, level, + strb, stImage, baseFormat, + destX, destY, destZ, + srcX, srcY, width, height); + } + #if 0 /* GL_SGIS_generate_mipmap -- this can be accelerated now. @@ -1155,10 +1232,10 @@ do_copy_texsubimage(GLcontext *ctx, } #endif - return GL_TRUE; } + static void st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, GLenum internalFormat, @@ -1171,8 +1248,10 @@ st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); +#if 0 if (border) goto fail; +#endif /* Setup or redefine the texture object, mipmap tree and texture * image. Don't populate yet. @@ -1182,19 +1261,9 @@ st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, GL_RGBA, CHAN_TYPE, NULL, &ctx->DefaultPacking, texObj, texImage); - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, 0, 0, x, y, width, 1)) - goto fail; - - return; - - fail: -#if 0 - _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, - width, border); -#endif - ; + do_copy_texsubimage(ctx, target, level, + 0, 0, 0, + x, y, width, 1); } @@ -1211,8 +1280,10 @@ st_CopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); +#if 0 if (border) goto fail; +#endif /* Setup or redefine the texture object, mipmap tree and texture * image. Don't populate yet. @@ -1223,19 +1294,9 @@ st_CopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, &ctx->DefaultPacking, texObj, texImage); - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, 0, 0, x, y, width, height)) - goto fail; - - return; - - fail: -#if 0 - _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, - width, height, border); -#endif - assert(0); + do_copy_texsubimage(ctx, target, level, + 0, 0, 0, + x, y, width, height); } @@ -1243,27 +1304,11 @@ static void st_CopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint x, GLint y, GLsizei width) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - const GLenum baseFormat = texImage->TexFormat->BaseFormat; - - /* XXX need to check as in above function? */ - - /* Need to check texture is compatible with source format. - */ - - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - baseFormat, xoffset, 0, x, y, width, 1)) { -#if 0 - _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); -#endif - assert(0); - } + const GLint yoffset = 0, zoffset = 0; + const GLsizei height = 1; + do_copy_texsubimage(ctx, target, level, + xoffset, yoffset, zoffset, + x, y, width, height); } @@ -1272,28 +1317,21 @@ st_CopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint x, GLint y, GLsizei width, GLsizei height) { - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - GLenum internalFormat = texImage->InternalFormat; - + const GLint zoffset = 0; + do_copy_texsubimage(ctx, target, level, + xoffset, yoffset, zoffset, + x, y, width, height); +} - /* Need to check texture is compatible with source format. - */ - if (!do_copy_texsubimage(ctx, - st_texture_image(texImage), - internalFormat, - xoffset, yoffset, x, y, width, height)) { -#if 0 - _swrast_copy_texsubimage2d(ctx, target, level, - xoffset, yoffset, x, y, width, height); -#endif - assert(0); - } +static void +st_CopyTexSubImage3D(GLcontext * ctx, GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + do_copy_texsubimage(ctx, target, level, + xoffset, yoffset, zoffset, + x, y, width, height); } @@ -1588,6 +1626,8 @@ st_init_texture_functions(struct dd_function_table *functions) functions->CopyTexImage2D = st_CopyTexImage2D; functions->CopyTexSubImage1D = st_CopyTexSubImage1D; functions->CopyTexSubImage2D = st_CopyTexSubImage2D; + functions->CopyTexSubImage3D = st_CopyTexSubImage3D; + functions->GetTexImage = st_GetTexImage; /* compressed texture functions */ -- cgit v1.2.3 From fcd4eeb743c717d48166e38a57fcd4a1752e32ab Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 18:34:13 -0600 Subject: don't use scissored bounds in _mesa_clip_copytexsubimage() --- src/mesa/main/image.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c index ae3c82b810..76e105e65e 100644 --- a/src/mesa/main/image.c +++ b/src/mesa/main/image.c @@ -4641,7 +4641,7 @@ _mesa_clip_copytexsubimage(const GLcontext *ctx, const struct gl_framebuffer *fb = ctx->ReadBuffer; const GLint srcX0 = *srcX, srcY0 = *srcY; - if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, + if (_mesa_clip_to_region(0, 0, fb->Width, fb->Height, srcX, srcY, width, height)) { *destX = *destX + *srcX - srcX0; *destY = *destY + *srcY - srcY0; -- cgit v1.2.3 From 78008dbcaaf74cac3b66dae103f631de94df6137 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 18:34:43 -0600 Subject: added a8r8g8b8_put_tile() --- src/mesa/pipe/softpipe/sp_surface.c | 41 +++++++++++++++++++++++++++++++++++++ 1 file changed, 41 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 75e36e3e3f..b06f6813c0 100755 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -133,6 +133,46 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, } +static void +a8r8g8b8_put_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + unsigned *dst + = ((unsigned *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp clipping hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + dst[j] = (a << 24) | (r << 16) | (g << 8) | b; + pRow += 4; + } + dst += ps->region->pitch; + p += w0 * 4; + } +} + + /*** PIPE_FORMAT_U_A1_R5_G5_B5 ***/ static void @@ -705,6 +745,7 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) sps->read_quad_f_swz = a8r8g8b8_read_quad_f_swz; sps->write_quad_f_swz = a8r8g8b8_write_quad_f_swz; sps->surface.get_tile = a8r8g8b8_get_tile; + sps->surface.put_tile = a8r8g8b8_put_tile; break; case PIPE_FORMAT_U_A1_R5_G5_B5: sps->surface.get_tile = a1r5g5b5_get_tile; -- cgit v1.2.3 From c6717a86420d7141013165f7acd50b3c3f751756 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 18:39:14 -0600 Subject: fallback_copy_texsubimage() basically works now (at least w/ Xlib driver). Have to map regions before calling get_tile()/put_tile(). Need to invert srcY of glCopyTexSubImage() depending on renderbuffers up/down orientation. Still need to invert image in fallback_copy_texsubimage() when needed. --- src/mesa/state_tracker/st_cb_texture.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 052035f988..e3a747fa74 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1070,6 +1070,8 @@ texture_face(GLenum target) /** * Do a CopyTexSubImage operation by mapping the source region and * dest region and copying/converting pixels. + * + * Note: srcY=0=TOP of renderbuffer */ static void fallback_copy_texsubimage(GLcontext *ctx, @@ -1093,13 +1095,20 @@ fallback_copy_texsubimage(GLcontext *ctx, dest_surf = pipe->get_tex_surface(pipe, mt, face, level, destZ); + (void) pipe->region_map(pipe, dest_surf->region); + (void) pipe->region_map(pipe, src_surf->region); + data = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); src_surf->get_tile(src_surf, srcX, srcY, width, height, data); - /* process pixels */ + /* Do GL pixel transfer ops here */ + /* Also, invert image if strb orientation is Y_0_TOP */ dest_surf->put_tile(dest_surf, destX, destY, width, height, data); + (void) pipe->region_unmap(pipe, dest_surf->region); + (void) pipe->region_unmap(pipe, src_surf->region); + free(data); } @@ -1109,6 +1118,9 @@ fallback_copy_texsubimage(GLcontext *ctx, /** * Do a CopyTex[Sub]Image using an optimized hardware (blit) path. * Note that the region to copy has already been clip tested. + * + * Note: srcY=0=Bottom of renderbuffer + * * \return GL_TRUE if success, GL_FALSE if failure (use a fallback) */ static void @@ -1151,6 +1163,10 @@ do_copy_texsubimage(GLcontext *ctx, assert(strb->surface); assert(stImage->mt); + if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { + srcY = strb->Base.Height - srcY - height; + } + src_format = strb->surface->format; dest_format = stImage->mt->format; -- cgit v1.2.3 From 5c83f1371978472fbe4bba8f686733c6b519874a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Sep 2007 18:54:20 -0600 Subject: Do image flipping in do_copy_texsubimage() and GL pixel transfer ops (except convolution). --- src/mesa/state_tracker/st_cb_texture.c | 35 +++++++++++++++++++++++++++------- 1 file changed, 28 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index e3a747fa74..13c803c6a7 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1069,7 +1069,7 @@ texture_face(GLenum target) /** * Do a CopyTexSubImage operation by mapping the source region and - * dest region and copying/converting pixels. + * dest region and using get_tile()/put_tile() to access the pixels/texels. * * Note: srcY=0=TOP of renderbuffer */ @@ -1089,6 +1089,16 @@ fallback_copy_texsubimage(GLcontext *ctx, struct pipe_mipmap_tree *mt = stImage->mt; struct pipe_surface *src_surf, *dest_surf; GLfloat *data; + GLint row, yStep; + + /* determine bottom-to-top vs. top-to-bottom order */ + if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { + destY = height - 1 - destY; + yStep = -1; + } + else { + yStep = 1; + } src_surf = strb->surface; @@ -1098,13 +1108,24 @@ fallback_copy_texsubimage(GLcontext *ctx, (void) pipe->region_map(pipe, dest_surf->region); (void) pipe->region_map(pipe, src_surf->region); - data = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - src_surf->get_tile(src_surf, srcX, srcY, width, height, data); + /* buffer for one row */ + data = (GLfloat *) malloc(width * 4 * sizeof(GLfloat)); - /* Do GL pixel transfer ops here */ - /* Also, invert image if strb orientation is Y_0_TOP */ + /* do copy row by row */ + for (row = 0; row < height; row++) { + src_surf->get_tile(src_surf, srcX, srcY + row, width, 1, data); + + /* XXX we're ignoring convolution for now */ + if (ctx->_ImageTransferState) { + _mesa_apply_rgba_transfer_ops(ctx, + ctx->_ImageTransferState & ~IMAGE_CONVOLUTION_BIT, + width, (GLfloat (*)[4])data); + } + + dest_surf->put_tile(dest_surf, destX, destY, width, 1, data); + destY += yStep; + } - dest_surf->put_tile(dest_surf, destX, destY, width, height, data); (void) pipe->region_unmap(pipe, dest_surf->region); (void) pipe->region_unmap(pipe, src_surf->region); @@ -1174,7 +1195,7 @@ do_copy_texsubimage(GLcontext *ctx, dest_region = stImage->mt->region; if (src_format == dest_format && - /* XXX also check GL pixel transfer ops here */ + ctx->_ImageTransferState == 0x0 && src_region && dest_region && src_region->cpp == dest_region->cpp) { -- cgit v1.2.3 From 7770acf8d4360ecfcaeece6e366f5adc6c0c9dee Mon Sep 17 00:00:00 2001 From: keithw Date: Wed, 26 Sep 2007 11:55:08 +0100 Subject: Use Gallium in the renderer string --- src/mesa/state_tracker/st_cb_strings.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_strings.c b/src/mesa/state_tracker/st_cb_strings.c index 776f518e2b..c344df0ff1 100644 --- a/src/mesa/state_tracker/st_cb_strings.c +++ b/src/mesa/state_tracker/st_cb_strings.c @@ -66,7 +66,8 @@ st_get_string(GLcontext * ctx, GLenum name) } case GL_RENDERER: - snprintf(st->renderer, sizeof(st->renderer), "TG3D, %s on %s", + snprintf(st->renderer, sizeof(st->renderer), "Gallium %s, %s on %s", + ST_VERSION_STRING, pipe->get_name( pipe ), pipe->winsys->get_name( pipe->winsys )); -- cgit v1.2.3 From 08589f71051e588b0bb7d0c8b529976c85398dd1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 26 Sep 2007 11:56:17 +0100 Subject: Make flushing more lazy in the draw module. --- src/mesa/pipe/draw/draw_arrays.c | 76 -------------- src/mesa/pipe/draw/draw_clip.c | 4 - src/mesa/pipe/draw/draw_context.c | 118 +++++++++++----------- src/mesa/pipe/draw/draw_context.h | 24 +++-- src/mesa/pipe/draw/draw_debug.c | 115 +++++++++++++++++++++ src/mesa/pipe/draw/draw_feedback.c | 2 + src/mesa/pipe/draw/draw_prim.c | 172 +++++++++++++------------------- src/mesa/pipe/draw/draw_prim.h | 5 - src/mesa/pipe/draw/draw_private.h | 23 ++++- src/mesa/pipe/draw/draw_validate.c | 124 +++++++++++++++++++++++ src/mesa/pipe/draw/draw_vertex_cache.c | 17 ++-- src/mesa/pipe/draw/draw_vertex_fetch.c | 2 + src/mesa/pipe/draw/draw_vertex_shader.c | 5 +- src/mesa/pipe/softpipe/sp_draw_arrays.c | 17 ++-- src/mesa/pipe/softpipe/sp_prim_setup.c | 2 +- src/mesa/sources | 3 +- 16 files changed, 430 insertions(+), 279 deletions(-) delete mode 100644 src/mesa/pipe/draw/draw_arrays.c create mode 100644 src/mesa/pipe/draw/draw_debug.c create mode 100644 src/mesa/pipe/draw/draw_validate.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_arrays.c b/src/mesa/pipe/draw/draw_arrays.c deleted file mode 100644 index bbb243c469..0000000000 --- a/src/mesa/pipe/draw/draw_arrays.c +++ /dev/null @@ -1,76 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Brian Paul - * Keith Whitwell - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_context.h" -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" - -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_prim.h" - - -/** - * Draw vertex arrays - * This is the main entrypoint into the drawing module. - * \param prim one of PIPE_PRIM_x - * \param start index of first vertex to draw - * \param count number of vertices to draw - */ -void -draw_arrays(struct draw_context *draw, unsigned prim, - unsigned start, unsigned count) -{ - /* tell drawing pipeline we're beginning drawing */ - draw->pipeline.first->begin( draw->pipeline.first ); - - /* XXX: Shouldn't really be needed - cache should be invalidated - * after setting new vertex buffers, vertex elements, but not - * between draws. - */ - draw_vertex_cache_invalidate( draw ); - - draw_set_prim( draw, prim ); - - /* drawing done here: */ - draw_prim(draw, start, count); - - /* draw any left-over buffered prims */ - draw_flush(draw); - - /* tell drawing pipeline we're done drawing */ - draw->pipeline.first->end( draw->pipeline.first ); -} - - diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 3ccc408bc0..a505146047 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -248,8 +248,6 @@ do_clip_tri( struct draw_stage *stage, inlist[1] = header->v[1]; inlist[2] = header->v[2]; - clipmask &= ~CLIP_CULL_BIT; - while (clipmask && n >= 3) { const unsigned plane_idx = ffs(clipmask)-1; const float *plane = clipper->plane[plane_idx]; @@ -331,8 +329,6 @@ do_clip_line( struct draw_stage *stage, float t1 = 0.0F; struct prim_header newprim; - clipmask &= ~CLIP_CULL_BIT; - while (clipmask) { const unsigned plane_idx = ffs(clipmask)-1; const float *plane = clipper->plane[plane_idx]; diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 9acbc53742..66c66ff698 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -49,6 +49,8 @@ struct draw_context *draw_create( void ) draw->pipeline.flatshade = draw_flatshade_stage( draw ); draw->pipeline.cull = draw_cull_stage( draw ); draw->pipeline.feedback = draw_feedback_stage( draw ); + draw->pipeline.validate = draw_validate_stage( draw ); + draw->pipeline.first = draw->pipeline.validate; ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); @@ -73,6 +75,8 @@ struct draw_context *draw_create( void ) draw->attrib_front1 = -1; draw->attrib_back1 = -1; + draw_vertex_cache_invalidate( draw ); + return draw; } @@ -84,75 +88,19 @@ void draw_destroy( struct draw_context *draw ) } -/** - * Rebuild the rendering pipeline. - */ -static void validate_pipeline( struct draw_context *draw ) -{ - struct draw_stage *next = draw->pipeline.rasterize; - - /* - * NOTE: we build up the pipeline in end-to-start order. - * - * TODO: make the current primitive part of the state and build - * shorter pipelines for lines & points. - */ - - if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { - draw->pipeline.unfilled->next = next; - next = draw->pipeline.unfilled; - } - - if (draw->rasterizer->offset_cw || - draw->rasterizer->offset_ccw) { - draw->pipeline.offset->next = next; - next = draw->pipeline.offset; - } - - if (draw->rasterizer->light_twoside) { - draw->pipeline.twoside->next = next; - next = draw->pipeline.twoside; - } - - /* Always run the cull stage as we calculate determinant there - * also. Fix this.. - */ - { - draw->pipeline.cull->next = next; - next = draw->pipeline.cull; - } - - /* Clip stage - */ - { - draw->pipeline.clip->next = next; - next = draw->pipeline.clip; - } - - /* Do software flatshading prior to clipping. XXX: should only do - * this for clipped primitives, ie it is a part of the clip - * routine. - */ - if (draw->rasterizer->flatshade) { - draw->pipeline.flatshade->next = next; - next = draw->pipeline.flatshade; - } - - if (draw->feedback.enabled || draw->feedback.discard) { - draw->pipeline.feedback->next = next; - next = draw->pipeline.feedback; - } - draw->pipeline.first = next; +void draw_flush( struct draw_context *draw ) +{ + if (draw->drawing) + draw_do_flush( draw, DRAW_FLUSH_DRAW ); } void draw_set_feedback_state( struct draw_context *draw, const struct pipe_feedback_state *feedback ) { + draw_flush( draw ); draw->feedback = *feedback; - validate_pipeline( draw ); } @@ -163,8 +111,8 @@ void draw_set_feedback_state( struct draw_context *draw, void draw_set_rasterizer_state( struct draw_context *draw, const struct pipe_rasterizer_state *raster ) { + draw_flush( draw ); draw->rasterizer = raster; - validate_pipeline( draw ); } @@ -175,6 +123,7 @@ void draw_set_rasterizer_state( struct draw_context *draw, void draw_set_rasterize_stage( struct draw_context *draw, struct draw_stage *stage ) { + draw_flush( draw ); draw->pipeline.rasterize = stage; } @@ -185,6 +134,8 @@ void draw_set_rasterize_stage( struct draw_context *draw, void draw_set_clip_state( struct draw_context *draw, const struct pipe_clip_state *clip ) { + draw_flush( draw ); + assert(clip->nr <= PIPE_MAX_CLIP_PLANES); memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); draw->nr_planes = 6 + clip->nr; @@ -199,6 +150,7 @@ void draw_set_clip_state( struct draw_context *draw, void draw_set_viewport_state( struct draw_context *draw, const struct pipe_viewport_state *viewport ) { + draw_flush( draw ); draw->viewport = *viewport; /* struct copy */ } @@ -207,6 +159,7 @@ void draw_set_vertex_shader(struct draw_context *draw, const struct pipe_shader_state *shader) { + draw_flush( draw ); draw->vertex_shader = *shader; } @@ -216,6 +169,8 @@ draw_set_vertex_buffer(struct draw_context *draw, unsigned attr, const struct pipe_vertex_buffer *buffer) { + draw_flush( draw ); + assert(attr < PIPE_ATTRIB_MAX); draw->vertex_buffer[attr] = *buffer; } @@ -226,6 +181,8 @@ draw_set_vertex_element(struct draw_context *draw, unsigned attr, const struct pipe_vertex_element *element) { + draw_flush( draw ); + assert(attr < PIPE_ATTRIB_MAX); draw->vertex_element[attr] = *element; } @@ -238,6 +195,8 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw, unsigned attr, const void *buffer) { + draw_flush( draw ); + draw->mapped_vbuffer[attr] = buffer; } @@ -246,6 +205,8 @@ void draw_set_mapped_constant_buffer(struct draw_context *draw, const void *buffer) { + draw_flush( draw ); + draw->mapped_constants = buffer; } @@ -254,8 +215,41 @@ void draw_set_mapped_feedback_buffer(struct draw_context *draw, uint index, void *buffer, uint size) { + draw_flush( draw ); + assert(index < PIPE_MAX_FEEDBACK_ATTRIBS); draw->mapped_feedback_buffer[index] = buffer; draw->mapped_feedback_buffer_size[index] = size; /* in bytes */ } + + + + +/** + * Allocate space for temporary post-transform vertices, such as for clipping. + */ +void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) +{ + stage->nr_tmps = nr; + + if (nr) { + ubyte *store = (ubyte *) malloc(MAX_VERTEX_SIZE * nr); + unsigned i; + + stage->tmp = (struct vertex_header **) malloc(sizeof(struct vertex_header *) * nr); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } +} + +void draw_free_tmps( struct draw_stage *stage ) +{ + if (stage->tmp) { + free(stage->tmp[0]); + free(stage->tmp); + } +} + + diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 4eb59aab01..0ccf5f6046 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -57,7 +57,6 @@ struct draw_stage; #define CLIP_BOTTOM_BIT 0x08 #define CLIP_NEAR_BIT 0x10 #define CLIP_FAR_BIT 0x20 -#define CLIP_CULL_BIT (1 << (6 + PIPE_MAX_CLIP_PLANES)) /*unused? */ /*@}*/ /** @@ -92,10 +91,6 @@ void draw_set_rasterizer_state( struct draw_context *draw, void draw_set_rasterize_stage( struct draw_context *draw, struct draw_stage *stage ); -unsigned draw_prim_info( unsigned prim, unsigned *first, unsigned *incr ); - -unsigned draw_trim( unsigned count, unsigned first, unsigned incr ); - void draw_set_vertex_shader(struct draw_context *draw, @@ -125,9 +120,22 @@ void draw_set_mapped_feedback_buffer(struct draw_context *draw, uint index, void *buffer, uint size); -void -draw_arrays(struct draw_context *draw, unsigned prim, - unsigned start, unsigned count); + +/*********************************************************************** + * draw_prim.c + */ + +void draw_arrays(struct draw_context *draw, unsigned prim, + unsigned start, unsigned count); + +void draw_flush(struct draw_context *draw); + +/*********************************************************************** + * draw_debug.c + */ +boolean draw_validate_prim( unsigned prim, unsigned length ); +unsigned draw_trim_prim( unsigned mode, unsigned count ); + #endif /* DRAW_CONTEXT_H */ diff --git a/src/mesa/pipe/draw/draw_debug.c b/src/mesa/pipe/draw/draw_debug.c new file mode 100644 index 0000000000..246e20fb30 --- /dev/null +++ b/src/mesa/pipe/draw/draw_debug.c @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw_private.h" +#include "draw_context.h" +#include "draw_prim.h" + + + +static void +draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) +{ + assert(prim >= PIPE_PRIM_POINTS); + assert(prim <= PIPE_PRIM_POLYGON); + + switch (prim) { + case PIPE_PRIM_POINTS: + *first = 1; + *incr = 1; + break; + case PIPE_PRIM_LINES: + *first = 2; + *incr = 2; + break; + case PIPE_PRIM_LINE_STRIP: + *first = 2; + *incr = 1; + break; + case PIPE_PRIM_LINE_LOOP: + *first = 2; + *incr = 1; + break; + case PIPE_PRIM_TRIANGLES: + *first = 3; + *incr = 3; + break; + case PIPE_PRIM_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + break; + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + *first = 3; + *incr = 1; + break; + case PIPE_PRIM_QUADS: + *first = 4; + *incr = 4; + break; + case PIPE_PRIM_QUAD_STRIP: + *first = 4; + *incr = 2; + break; + default: + assert(0); + *first = 1; + *incr = 1; + break; + } +} + + +unsigned +draw_trim_prim( unsigned mode, unsigned count ) +{ + unsigned length, first, incr; + + draw_prim_info( mode, &first, &incr ); + + if (count < first) + length = 0; + else + length = count - (count - first) % incr; + + return length; +} + + +boolean +draw_validate_prim( unsigned mode, unsigned count ) +{ + return (count > 0 && + count == draw_trim_prim( mode, count )); +} + diff --git a/src/mesa/pipe/draw/draw_feedback.c b/src/mesa/pipe/draw/draw_feedback.c index ee54db0ad5..15fad4da80 100644 --- a/src/mesa/pipe/draw/draw_feedback.c +++ b/src/mesa/pipe/draw/draw_feedback.c @@ -69,6 +69,8 @@ static void feedback_vertex(struct draw_stage *stage, const struct vertex_header *vertex) { struct feedback_stage *fs = (struct feedback_stage *) stage; + +#if 0 const struct pipe_feedback_state *feedback = &stage->draw->feedback; const uint select = feedback->interleaved ? 0 : 1; uint i; diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index be2f987b9a..a497a350f7 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -58,14 +58,15 @@ static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { }; -void draw_flush( struct draw_context *draw ) +static void draw_prim_queue_flush( struct draw_context *draw ) { struct draw_stage *first = draw->pipeline.first; unsigned i; /* Make sure all vertices are available: */ - draw_vertex_cache_validate(draw); + if (draw->vs.queue_nr) + draw_vertex_shader_queue_flush(draw); switch (draw->reduced_prim) { case RP_TRI: @@ -91,12 +92,41 @@ void draw_flush( struct draw_context *draw ) break; } - draw->pq.queue_nr = 0; - + draw->pq.queue_nr = 0; draw_vertex_cache_unreference( draw ); } +void draw_do_flush( struct draw_context *draw, + unsigned flush ) +{ + if ((flush & (DRAW_FLUSH_PRIM_QUEUE | + DRAW_FLUSH_VERTEX_CACHE_INVALIDATE | + DRAW_FLUSH_DRAW)) && + draw->pq.queue_nr) + { + draw_prim_queue_flush(draw); + } + + if ((flush & (DRAW_FLUSH_VERTEX_CACHE_INVALIDATE | + DRAW_FLUSH_DRAW)) && + draw->drawing) + { + draw_vertex_cache_invalidate(draw); + } + + if ((flush & DRAW_FLUSH_DRAW) && + draw->drawing) + { + draw->pipeline.first->end( draw->pipeline.first ); + draw->drawing = 0; + draw->prim = ~0; + draw->pipeline.first = draw->pipeline.validate; + } + +} + + /* Return a pointer to a freshly queued primitive header. Ensure that * there is room in the vertex cache for a maximum of "nr_verts" new @@ -106,13 +136,13 @@ void draw_flush( struct draw_context *draw ) static struct prim_header *get_queued_prim( struct draw_context *draw, unsigned nr_verts ) { - if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH) { -// fprintf(stderr, "p"); - draw_flush( draw ); - } - else if (!draw_vertex_cache_check_space( draw, nr_verts )) { + if (!draw_vertex_cache_check_space( draw, nr_verts )) { // fprintf(stderr, "v"); - draw_flush( draw ); + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE_INVALIDATE ); + } + else if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH) { +// fprintf(stderr, "p"); + draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); } return &draw->pq.queue[draw->pq.queue_nr++]; @@ -129,7 +159,7 @@ static void do_point( struct draw_context *draw, prim->reset_line_stipple = 0; prim->edgeflags = 1; prim->pad = 0; - prim->v[0] = draw->get_vertex( draw, i0 ); + prim->v[0] = draw->vcache.get_vertex( draw, i0 ); } @@ -143,8 +173,8 @@ static void do_line( struct draw_context *draw, prim->reset_line_stipple = reset_stipple; prim->edgeflags = 1; prim->pad = 0; - prim->v[0] = draw->get_vertex( draw, i0 ); - prim->v[1] = draw->get_vertex( draw, i1 ); + prim->v[0] = draw->vcache.get_vertex( draw, i0 ); + prim->v[1] = draw->vcache.get_vertex( draw, i1 ); } static void do_triangle( struct draw_context *draw, @@ -157,9 +187,9 @@ static void do_triangle( struct draw_context *draw, prim->reset_line_stipple = 1; prim->edgeflags = ~0; prim->pad = 0; - prim->v[0] = draw->get_vertex( draw, i0 ); - prim->v[1] = draw->get_vertex( draw, i1 ); - prim->v[2] = draw->get_vertex( draw, i2 ); + prim->v[0] = draw->vcache.get_vertex( draw, i0 ); + prim->v[1] = draw->vcache.get_vertex( draw, i1 ); + prim->v[2] = draw->vcache.get_vertex( draw, i2 ); } static void do_ef_triangle( struct draw_context *draw, @@ -170,9 +200,9 @@ static void do_ef_triangle( struct draw_context *draw, unsigned i2 ) { struct prim_header *prim = get_queued_prim( draw, 3 ); - struct vertex_header *v0 = draw->get_vertex( draw, i0 ); - struct vertex_header *v1 = draw->get_vertex( draw, i1 ); - struct vertex_header *v2 = draw->get_vertex( draw, i2 ); + struct vertex_header *v0 = draw->vcache.get_vertex( draw, i0 ); + struct vertex_header *v1 = draw->vcache.get_vertex( draw, i1 ); + struct vertex_header *v2 = draw->vcache.get_vertex( draw, i2 ); prim->reset_line_stipple = reset_stipple; @@ -202,7 +232,7 @@ static void do_quad( struct draw_context *draw, /** * Main entrypoint to draw some number of points/lines/triangles */ -void +static void draw_prim( struct draw_context *draw, unsigned start, unsigned count ) { unsigned i; @@ -341,14 +371,15 @@ draw_prim( struct draw_context *draw, unsigned start, unsigned count ) } -void +static void draw_set_prim( struct draw_context *draw, unsigned prim ) { + _mesa_printf("%s %d\n", __FUNCTION__, prim); assert(prim >= PIPE_PRIM_POINTS); assert(prim <= PIPE_PRIM_POLYGON); if (reduced_prim[prim] != draw->reduced_prim) { - draw_flush( draw ); + draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); draw->reduced_prim = reduced_prim[prim]; } @@ -356,91 +387,32 @@ draw_set_prim( struct draw_context *draw, unsigned prim ) } -unsigned -draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) -{ - assert(prim >= PIPE_PRIM_POINTS); - assert(prim <= PIPE_PRIM_POLYGON); - - switch (prim) { - case PIPE_PRIM_POINTS: - *first = 1; - *incr = 1; - return 0; - case PIPE_PRIM_LINES: - *first = 2; - *incr = 2; - return 0; - case PIPE_PRIM_LINE_STRIP: - *first = 2; - *incr = 1; - return 0; - case PIPE_PRIM_LINE_LOOP: - *first = 2; - *incr = 1; - return 1; - case PIPE_PRIM_TRIANGLES: - *first = 3; - *incr = 3; - return 0; - case PIPE_PRIM_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - return 0; - case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: - *first = 3; - *incr = 1; - return 1; - case PIPE_PRIM_QUADS: - *first = 4; - *incr = 4; - return 0; - case PIPE_PRIM_QUAD_STRIP: - *first = 4; - *incr = 2; - return 0; - default: - assert(0); - *first = 1; - *incr = 1; - return 0; - } -} - - -unsigned -draw_trim( unsigned count, unsigned first, unsigned incr ) -{ - if (count < first) - return 0; - else - return count - (count - first) % incr; -} /** - * Allocate space for temporary post-transform vertices, such as for clipping. + * Draw vertex arrays + * This is the main entrypoint into the drawing module. + * \param prim one of PIPE_PRIM_x + * \param start index of first vertex to draw + * \param count number of vertices to draw */ -void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) +void +draw_arrays(struct draw_context *draw, unsigned prim, + unsigned start, unsigned count) { - stage->nr_tmps = nr; - - if (nr) { - ubyte *store = (ubyte *) malloc(MAX_VERTEX_SIZE * nr); - unsigned i; + if (!draw->drawing) { + draw->drawing = 1; - stage->tmp = (struct vertex_header **) malloc(sizeof(struct vertex_header *) * nr); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + /* tell drawing pipeline we're beginning drawing */ + draw->pipeline.first->begin( draw->pipeline.first ); } -} -void draw_free_tmps( struct draw_stage *stage ) -{ - if (stage->tmp) { - free(stage->tmp[0]); - free(stage->tmp); + if (draw->prim != prim) { + draw_set_prim( draw, prim ); } + + /* drawing done here: */ + draw_prim(draw, start, count); } + + diff --git a/src/mesa/pipe/draw/draw_prim.h b/src/mesa/pipe/draw/draw_prim.h index 4c55b02978..07cd3e2c9f 100644 --- a/src/mesa/pipe/draw/draw_prim.h +++ b/src/mesa/pipe/draw/draw_prim.h @@ -6,11 +6,6 @@ void draw_invalidate_vcache( struct draw_context *draw ); -void draw_set_prim( struct draw_context *draw, unsigned prim ); - -void draw_prim( struct draw_context *draw, unsigned start, unsigned count ); - -void draw_flush( struct draw_context *draw ); #endif /* DRAW_PRIM_H */ diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 04d38c4e0c..ebef5347ab 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -126,6 +126,8 @@ struct draw_context struct { struct draw_stage *first; /**< one of the following */ + struct draw_stage *validate; + /* stages (in logical order) */ struct draw_stage *feedback; struct draw_stage *flatshade; @@ -171,13 +173,10 @@ struct draw_context uint attrib_front0, attrib_back0; uint attrib_front1, attrib_back1; - unsigned nr_vertices; - + unsigned drawing; unsigned prim; /**< current prim type: PIPE_PRIM_x */ unsigned reduced_prim; - struct vertex_header *(*get_vertex)( struct draw_context *draw, - unsigned i ); /* Post-tnl vertex cache: */ @@ -186,6 +185,9 @@ struct draw_context unsigned idx[VCACHE_SIZE + VCACHE_OVERFLOW]; struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; unsigned overflow; + + struct vertex_header *(*get_vertex)( struct draw_context *draw, + unsigned i ); } vcache; /* Vertex shader queue: @@ -219,6 +221,7 @@ extern struct draw_stage *draw_offset_stage( struct draw_context *context ); extern struct draw_stage *draw_clip_stage( struct draw_context *context ); extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); extern struct draw_stage *draw_cull_stage( struct draw_context *context ); +extern struct draw_stage *draw_validate_stage( struct draw_context *context ); extern void draw_free_tmps( struct draw_stage *stage ); @@ -228,7 +231,6 @@ extern void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ); extern int draw_vertex_cache_check_space( struct draw_context *draw, unsigned nr_verts ); -extern void draw_vertex_cache_validate( struct draw_context *draw ); extern void draw_vertex_cache_invalidate( struct draw_context *draw ); extern void draw_vertex_cache_unreference( struct draw_context *draw ); extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); @@ -244,6 +246,17 @@ extern void draw_vertex_fetch( struct draw_context *draw, unsigned count ); +#define DRAW_FLUSH_PRIM_QUEUE 0x1 +#define DRAW_FLUSH_VERTEX_CACHE_INVALIDATE 0x2 +#define DRAW_FLUSH_DRAW 0x4 + + +void draw_do_flush( struct draw_context *draw, + unsigned flags ); + + + + /** * Get a writeable copy of a vertex. * \param stage drawing stage info diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c new file mode 100644 index 0000000000..7c5a9dceca --- /dev/null +++ b/src/mesa/pipe/draw/draw_validate.c @@ -0,0 +1,124 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + + + + +/** + * Rebuild the rendering pipeline. + */ +static void validate_begin( struct draw_stage *stage ) +{ + struct draw_context *draw = stage->draw; + struct draw_stage *next = draw->pipeline.rasterize; + + /* + * NOTE: we build up the pipeline in end-to-start order. + * + * TODO: make the current primitive part of the state and build + * shorter pipelines for lines & points. + */ + + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { + draw->pipeline.unfilled->next = next; + next = draw->pipeline.unfilled; + } + + if (draw->rasterizer->offset_cw || + draw->rasterizer->offset_ccw) { + draw->pipeline.offset->next = next; + next = draw->pipeline.offset; + } + + if (draw->rasterizer->light_twoside) { + draw->pipeline.twoside->next = next; + next = draw->pipeline.twoside; + } + + /* Always run the cull stage as we calculate determinant there + * also. Fix this.. + */ + { + draw->pipeline.cull->next = next; + next = draw->pipeline.cull; + } + + /* Clip stage + */ + { + draw->pipeline.clip->next = next; + next = draw->pipeline.clip; + } + + /* Do software flatshading prior to clipping. XXX: should only do + * this for clipped primitives, ie it is a part of the clip + * routine. + */ + if (draw->rasterizer->flatshade) { + draw->pipeline.flatshade->next = next; + next = draw->pipeline.flatshade; + } + + if (draw->feedback.enabled || draw->feedback.discard) { + draw->pipeline.feedback->next = next; + next = draw->pipeline.feedback; + } + + draw->pipeline.first = next; + draw->pipeline.first->begin( draw->pipeline.first ); +} + + + + +/** + * Create validate pipeline stage. + */ +struct draw_stage *draw_validate_stage( struct draw_context *draw ) +{ + struct draw_stage *stage = CALLOC_STRUCT(draw_stage); + + stage->draw = draw; + stage->next = NULL; + stage->begin = validate_begin; + stage->point = NULL; + stage->line = NULL; + stage->tri = NULL; + stage->end = NULL; + stage->reset_stipple_counter = NULL; + + return stage; +} diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index a226798123..1f4adedede 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -90,6 +90,9 @@ static struct vertex_header *get_vertex( struct draw_context *draw, draw->vcache.vertex[slot]->pad = 0; draw->vcache.vertex[slot]->vertex_id = ~0; } + else { +// fprintf(stderr, "*"); + } return draw->vcache.vertex[slot]; } @@ -127,10 +130,6 @@ void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) draw->vcache.vertex[i]->vertex_id = ~0; } -void draw_vertex_cache_validate( struct draw_context *draw ) -{ - draw_vertex_shader_queue_flush( draw ); -} void draw_vertex_cache_unreference( struct draw_context *draw ) { @@ -168,19 +167,21 @@ void draw_set_mapped_element_buffer( struct draw_context *draw, unsigned eltSize, void *elements ) { +// draw_statechange( draw ); + /* choose the get_vertex() function to use */ switch (eltSize) { case 0: - draw->get_vertex = get_vertex; + draw->vcache.get_vertex = get_vertex; break; case 1: - draw->get_vertex = get_ubyte_elt_vertex; + draw->vcache.get_vertex = get_ubyte_elt_vertex; break; case 2: - draw->get_vertex = get_ushort_elt_vertex; + draw->vcache.get_vertex = get_ushort_elt_vertex; break; case 4: - draw->get_vertex = get_uint_elt_vertex; + draw->vcache.get_vertex = get_uint_elt_vertex; break; default: assert(0); diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 62e8d61be4..2b839d641e 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -99,6 +99,8 @@ void draw_vertex_fetch( struct draw_context *draw, /*printf(" %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]);*/ + /* Transform to AoS xxxx/yyyy/zzzz/wwww representation: + */ machine->Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ machine->Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ machine->Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index fe4f124dd2..b9e4cc13cc 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -169,13 +169,12 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j]; vOut[j]->data[slot][2] = machine.Outputs[slot].xyzw[2].f[j]; vOut[j]->data[slot][3] = machine.Outputs[slot].xyzw[3].f[j]; - /* + printf("output %d: %f %f %f %f\n", slot, vOut[j]->data[slot][0], vOut[j]->data[slot][1], vOut[j]->data[slot][2], vOut[j]->data[slot][3]); - */ } } /* loop over vertices */ } @@ -189,7 +188,7 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) { unsigned i, j; -// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); + fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); /* run vertex shader on vertex cache entries, four per invokation */ for (i = 0; i < draw->vs.queue_nr; i += 4) { diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 7ea29a0a26..21c30b53f3 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -95,14 +95,16 @@ softpipe_draw_elements(struct pipe_context *pipe, { struct softpipe_context *sp = softpipe_context(pipe); struct draw_context *draw = sp->draw; - unsigned length, first, incr, i; + unsigned i; - /* first, check that the primitive is not malformed */ - draw_prim_info( mode, &first, &incr ); - length = draw_trim( count, first, incr ); - if (!length) - return TRUE; + /* first, check that the primitive is not malformed. It is the + * state tracker's responsibility to do send only correctly formed + * primitives down. + */ +// count = draw_trim_prim( mode, count ); + if (!draw_validate_prim( mode, count )) + assert(0); if (sp->dirty) softpipe_update_derived( sp ); @@ -151,6 +153,9 @@ softpipe_draw_elements(struct pipe_context *pipe, /* draw! */ draw_arrays(draw, mode, start, count); + /* always flush for now */ + draw_flush(draw); + /* * unmap vertex/index buffers */ diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 2e27d00acf..621a44512c 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -41,7 +41,7 @@ #include "pipe/draw/draw_vertex.h" #include "pipe/p_util.h" -#define DEBUG_VERTS 0 +#define DEBUG_VERTS 1 /** * Triangle edge info diff --git a/src/mesa/sources b/src/mesa/sources index 985bd2dce6..069be2e01d 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -157,15 +157,16 @@ VF_SOURCES = \ DRAW_SOURCES = \ - pipe/draw/draw_arrays.c \ pipe/draw/draw_clip.c \ pipe/draw/draw_context.c\ pipe/draw/draw_cull.c \ + pipe/draw/draw_debug.c \ pipe/draw/draw_feedback.c \ pipe/draw/draw_flatshade.c \ pipe/draw/draw_offset.c \ pipe/draw/draw_prim.c \ pipe/draw/draw_twoside.c \ + pipe/draw/draw_validate.c \ pipe/draw/draw_vertex.c \ pipe/draw/draw_vertex_cache.c \ pipe/draw/draw_vertex_fetch.c \ -- cgit v1.2.3 From e078f910dcdba2081a52b74308ca3d330115e912 Mon Sep 17 00:00:00 2001 From: keithw Date: Wed, 26 Sep 2007 11:57:28 +0100 Subject: restore primitive trimming in sp_draw_arrays.c --- src/mesa/pipe/softpipe/sp_draw_arrays.c | 9 ++++++--- src/mesa/pipe/softpipe/sp_prim_setup.c | 2 +- 2 files changed, 7 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 21c30b53f3..405659fb46 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -99,12 +99,15 @@ softpipe_draw_elements(struct pipe_context *pipe, /* first, check that the primitive is not malformed. It is the * state tracker's responsibility to do send only correctly formed - * primitives down. + * primitives down. It currently isn't doing that though... */ -// count = draw_trim_prim( mode, count ); - +#if 1 + count = draw_trim_prim( mode, count ); +#else if (!draw_validate_prim( mode, count )) assert(0); +#endif + if (sp->dirty) softpipe_update_derived( sp ); diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 621a44512c..2e27d00acf 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -41,7 +41,7 @@ #include "pipe/draw/draw_vertex.h" #include "pipe/p_util.h" -#define DEBUG_VERTS 1 +#define DEBUG_VERTS 0 /** * Triangle edge info -- cgit v1.2.3 From c5cfa639880d5309a2276dc0c47a73ca0752d696 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 26 Sep 2007 12:08:08 +0100 Subject: disable debug --- src/mesa/pipe/draw/draw_prim.c | 1 - src/mesa/pipe/draw/draw_vertex_shader.c | 5 +++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index a497a350f7..698557d6bd 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -374,7 +374,6 @@ draw_prim( struct draw_context *draw, unsigned start, unsigned count ) static void draw_set_prim( struct draw_context *draw, unsigned prim ) { - _mesa_printf("%s %d\n", __FUNCTION__, prim); assert(prim >= PIPE_PRIM_POINTS); assert(prim <= PIPE_PRIM_POLYGON); diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index b9e4cc13cc..fe4f124dd2 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -169,12 +169,13 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j]; vOut[j]->data[slot][2] = machine.Outputs[slot].xyzw[2].f[j]; vOut[j]->data[slot][3] = machine.Outputs[slot].xyzw[3].f[j]; - + /* printf("output %d: %f %f %f %f\n", slot, vOut[j]->data[slot][0], vOut[j]->data[slot][1], vOut[j]->data[slot][2], vOut[j]->data[slot][3]); + */ } } /* loop over vertices */ } @@ -188,7 +189,7 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) { unsigned i, j; - fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); +// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); /* run vertex shader on vertex cache entries, four per invokation */ for (i = 0; i < draw->vs.queue_nr; i += 4) { -- cgit v1.2.3 From 035a824730ae3804b2bfaadec94015aac5c6ebcc Mon Sep 17 00:00:00 2001 From: keithw Date: Wed, 26 Sep 2007 12:10:46 +0100 Subject: fix merge --- src/mesa/pipe/draw/draw_feedback.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_feedback.c b/src/mesa/pipe/draw/draw_feedback.c index 15fad4da80..ee54db0ad5 100644 --- a/src/mesa/pipe/draw/draw_feedback.c +++ b/src/mesa/pipe/draw/draw_feedback.c @@ -69,8 +69,6 @@ static void feedback_vertex(struct draw_stage *stage, const struct vertex_header *vertex) { struct feedback_stage *fs = (struct feedback_stage *) stage; - -#if 0 const struct pipe_feedback_state *feedback = &stage->draw->feedback; const uint select = feedback->interleaved ? 0 : 1; uint i; -- cgit v1.2.3 From 65e3af51efc9d688ef8face0a44429a90c5dd4c9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 26 Sep 2007 12:46:45 +0100 Subject: Enable codegen based whenever __i386__ is defined. --- src/mesa/pipe/draw/draw_vertex_shader.c | 10 ++-------- src/mesa/pipe/softpipe/sp_quad_fs.c | 10 ++-------- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 2 +- src/mesa/state_tracker/st_cb_program.c | 8 ++++---- src/mesa/state_tracker/st_program.c | 4 ++-- src/mesa/state_tracker/st_program.h | 4 ++-- src/mesa/x86/rtasm/x86sse.c | 2 +- src/mesa/x86/rtasm/x86sse.h | 2 +- 8 files changed, 15 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index fe4f124dd2..a2e1cdc472 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -64,13 +64,11 @@ compute_clipmask(float cx, float cy, float cz, float cw) #endif #endif -#if defined(USE_X86_ASM) || defined(SLANG_X86) -typedef void (XSTDCALL *sse2_function)( +typedef void (XSTDCALL *codegen_function) ( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, float (*constant)[4], struct tgsi_exec_vector *temporary ); -#endif /** * Transform vertices with the current vertex program/shader @@ -117,16 +115,12 @@ run_vertex_program(struct draw_context *draw, /* run shader */ if( draw->vertex_shader.executable != NULL ) { -#if defined(USE_X86_ASM) || defined(SLANG_X86) - sse2_function func = (sse2_function) draw->vertex_shader.executable; + codegen_function func = (codegen_function) draw->vertex_shader.executable; func( machine.Inputs, machine.Outputs, machine.Consts, machine.Temps ); -#else - assert( 0 ); -#endif } else { tgsi_exec_machine_run( &machine ); diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index bff5525b0f..673d339f41 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -64,14 +64,12 @@ quad_shade_stage(struct quad_stage *qs) #endif #endif -#if defined(USE_X86_ASM) || defined(SLANG_X86) -typedef void (XSTDCALL *sse2_function)( +typedef void (XSTDCALL *codegen_function)( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, float (*constant)[4], struct tgsi_exec_vector *temporary, const struct tgsi_interp_coef *coef ); -#endif /* This should be done by the fragment shader execution unit (code * generated from the decl instructions). Do it here for now. @@ -121,17 +119,13 @@ shade_quad( /* run shader */ if( softpipe->fs->executable != NULL ) { -#if defined(USE_X86_ASM) || defined(SLANG_X86) - sse2_function func = (sse2_function) softpipe->fs->executable; + codegen_function func = (codegen_function) softpipe->fs->executable; func( machine.Inputs, machine.Outputs, machine.Consts, machine.Temps, machine.InterpCoefs ); -#else - assert( 0 ); -#endif } else { tgsi_exec_machine_run( &machine ); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index cf5e386ddf..e7a6112e6e 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -2,7 +2,7 @@ #include "tgsi_core.h" #include "x86/rtasm/x86sse.h" -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) #define FOR_EACH_CHANNEL( CHAN )\ for( CHAN = 0; CHAN < 4; CHAN++ ) diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index f57d22652f..25dbc2c244 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -86,7 +86,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, prog->serialNo = 1; -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) x86_init_func( &prog->sse2_program ); #endif @@ -102,7 +102,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, prog->serialNo = 1; -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) x86_init_func( &prog->sse2_program ); #endif @@ -129,7 +129,7 @@ static void st_delete_program( GLcontext *ctx, case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) x86_release_func( &stvp->sse2_program ); #endif st_remove_vertex_program(st, stvp); @@ -139,7 +139,7 @@ static void st_delete_program( GLcontext *ctx, { struct st_fragment_program *stfp = (struct st_fragment_program *) prog; -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) x86_release_func( &stfp->sse2_program ); #endif st_remove_fragment_program(st, stfp); diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 9f363adc5c..44cc33b09f 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -253,7 +253,7 @@ st_translate_vertex_program(struct st_context *st, if (TGSI_DEBUG) tgsi_dump( tokensOut, 0 ); -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) if (stvp->sse2_program.csr == stvp->sse2_program.store) tgsi_emit_sse2( tokensOut, &stvp->sse2_program ); @@ -398,7 +398,7 @@ st_translate_fragment_program(struct st_context *st, if (TGSI_DEBUG) tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) if (stfp->sse2_program.csr == stfp->sse2_program.store) tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 355dee574b..6e106c16b0 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -53,7 +53,7 @@ struct st_fragment_program /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) struct x86_function sse2_program; #endif @@ -79,7 +79,7 @@ struct st_vertex_program /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) struct x86_function sse2_program; #endif diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index f5b0ccdb9b..96c1301d7f 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -1,4 +1,4 @@ -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) #include "imports.h" #include "x86sse.h" diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h index c6236395b2..4816bd2ad2 100644 --- a/src/mesa/x86/rtasm/x86sse.h +++ b/src/mesa/x86/rtasm/x86sse.h @@ -2,7 +2,7 @@ #ifndef _X86SSE_H_ #define _X86SSE_H_ -#if defined(USE_X86_ASM) || defined(SLANG_X86) +#if defined(__i386__) || defined(__386__) #include "glheader.h" -- cgit v1.2.3 From 82e2d3226893487d33152f15763516473187c07d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 26 Sep 2007 13:30:01 +0100 Subject: Fix some compiler warnings with -pedantic --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 18 +++++++++--------- src/mesa/x86/rtasm/x86sse.c | 12 +++++++++--- src/mesa/x86/rtasm/x86sse.h | 2 +- 3 files changed, 19 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index e7a6112e6e..d840d64cae 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -408,7 +408,7 @@ emit_func_call1( struct x86_function *func, unsigned xmm_dst, unsigned storage, - unsigned char *code ) + void (*code)() ) { x86_push( func, @@ -443,7 +443,7 @@ emit_func_call2( unsigned xmm_dst, unsigned xmm_src, unsigned storage, - unsigned char *code ) + void (*code)() ) { x86_push( func, @@ -506,7 +506,7 @@ emit_cos( func, xmm_dst, ALIGN16( (unsigned) g_cos_storage ), - (unsigned char *) cos4f ); + cos4f ); } /* XXX: move into machine context */ @@ -538,7 +538,7 @@ emit_sin (struct x86_function *func, func, xmm_dst, ALIGN16( (unsigned) g_sin_storage ), - (unsigned char *) sin4f ); + sin4f ); } static void @@ -596,7 +596,7 @@ emit_pow( xmm_dst, xmm_src, ALIGN16( (unsigned) g_pow_storage ), - (unsigned char *) pow4f ); + pow4f ); } /* XXX: move into machine context */ @@ -629,7 +629,7 @@ emit_ex2( func, xmm_dst, ALIGN16( (unsigned) g_ex2_storage ), - (unsigned char *) ex24f ); + ex24f ); } /* XXX: move into machine context */ @@ -655,7 +655,7 @@ emit_lg2( func, xmm_dst, ALIGN16( (unsigned) g_lg2_storage ), - (unsigned char *) lg24f ); + lg24f ); } /* XXX: move into machine context */ @@ -681,7 +681,7 @@ emit_flr( func, xmm_dst, ALIGN16( (unsigned) g_flr_storage ), - (unsigned char *) flr4f ); + flr4f ); } /* XXX: move into machine context */ @@ -707,7 +707,7 @@ emit_frc( func, xmm_dst, ALIGN16( (unsigned) g_frc_storage ), - (unsigned char *) frc4f ); + frc4f ); } static void diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index 96c1301d7f..a6cfa40f91 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -6,6 +6,12 @@ #define DISASSEM 0 #define X86_TWOB 0x0f +static GLubyte *cptr( void (*label)() ) +{ + return (char *)(unsigned long)label; +} + + /* Emit bytes to the instruction stream: */ static void emit_1b( struct x86_function *p, GLbyte b0 ) @@ -252,10 +258,10 @@ void x86_jmp( struct x86_function *p, GLubyte *label) emit_1i(p, label - x86_get_label(p) - 4); } -void x86_call( struct x86_function *p, GLubyte *label) +void x86_call( struct x86_function *p, void (*label)()) { emit_1ub(p, 0xe8); - emit_1i(p, label - x86_get_label(p) - 4); + emit_1i(p, cptr(label) - x86_get_label(p) - 4); } /* michal: @@ -1138,7 +1144,7 @@ void (*x86_get_func( struct x86_function *p ))(void) { if (DISASSEM) _mesa_printf("disassemble %p %p\n", p->store, p->csr); - return (void (*)(void))p->store; + return (void (*)(void)) (unsigned long) p->store; } #else diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h index 4816bd2ad2..c1ca06088b 100644 --- a/src/mesa/x86/rtasm/x86sse.h +++ b/src/mesa/x86/rtasm/x86sse.h @@ -120,7 +120,7 @@ void x86_fixup_fwd_jump( struct x86_function *p, void x86_jmp( struct x86_function *p, GLubyte *label ); -void x86_call( struct x86_function *p, GLubyte *label ); +void x86_call( struct x86_function *p, void (*label)() ); /* michal: * Temporary. As I need immediate operands, and dont want to mess with the codegen, -- cgit v1.2.3 From 48385cd13ac9410e07856924a3cd113b78f198db Mon Sep 17 00:00:00 2001 From: keithw Date: Thu, 27 Sep 2007 12:50:02 +0100 Subject: reduced debug --- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index 235903ca6d..a950e72e56 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -114,8 +114,8 @@ static boolean check_space( struct vbuf_stage *vbuf ) static void emit_vertex( struct vbuf_stage *vbuf, struct vertex_header *vertex ) { - fprintf(stderr, "emit vertex %d to %p\n", - vbuf->nr_vertices, vbuf->vertex_ptr); +// fprintf(stderr, "emit vertex %d to %p\n", +// vbuf->nr_vertices, vbuf->vertex_ptr); vertex->vertex_id = vbuf->nr_vertices++; -- cgit v1.2.3 From e0271e16da6427436ed755f170e9b402ada5dd7c Mon Sep 17 00:00:00 2001 From: keithw Date: Thu, 27 Sep 2007 16:20:31 +0100 Subject: remove dead code --- src/mesa/state_tracker/st_cb_texture.c | 83 ---------------------------------- 1 file changed, 83 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 13c803c6a7..e7be3df722 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -487,8 +487,6 @@ try_pbo_upload(GLcontext *ctx, dst_stride = stImage->mt->pitch; - intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); { struct _DriBufferObject *src_buffer = intel_bufferobj_buffer(intel, pbo, INTEL_READ); @@ -505,10 +503,7 @@ try_pbo_upload(GLcontext *ctx, dst_stride, dst_buffer, dst_offset, 0, 0, 0, 0, width, height, GL_COPY); - - intel_batchbuffer_flush(intel->batch); } - UNLOCK_HARDWARE(intel); return GL_TRUE; #endif @@ -556,10 +551,6 @@ st_TexImage(GLcontext * ctx, DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); -#if 0 - intelFlush(ctx); -#endif - stImage->face = target_to_face(target); stImage->level = level; @@ -937,10 +928,6 @@ st_TexSubimage(GLcontext * ctx, _mesa_lookup_enum_by_nr(target), level, xoffset, yoffset, width, height); -#if 0 - intelFlush(ctx); -#endif - pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format, type, pixels, packing, "glTexSubImage2D"); @@ -1209,12 +1196,6 @@ do_copy_texsubimage(GLcontext *ctx, * vs. user-created FBO */ -#if 00 /* XXX FIX flush/locking */ - intelFlush(ctx); - /* XXX still need the lock ? */ - LOCK_HARDWARE(intel); -#endif - #if 0 /* A bit of fiddling to get the blitter to work with -ve * pitches. But we get a nice inverted blit this way, so it's @@ -1230,7 +1211,6 @@ do_copy_texsubimage(GLcontext *ctx, dest_offset, x, y + height, dstx, dsty, width, height, GL_COPY); /* ? */ - intel_batchbuffer_flush(intel->batch); #else pipe->region_copy(pipe, @@ -1245,10 +1225,6 @@ do_copy_texsubimage(GLcontext *ctx, /* size */ width, height); #endif - -#if 0 - UNLOCK_HARDWARE(intel); -#endif } else { fallback_copy_texsubimage(ctx, target, level, @@ -1583,70 +1559,11 @@ st_finalize_mipmap_tree(GLcontext *ctx, } } - /** - if (need_flush) - intel_batchbuffer_flush(intel->batch); - **/ return GL_TRUE; } -#if 0 /* unused? */ -void -st_tex_map_images(struct pipe_context *pipe, - struct st_texture_object *stObj) -{ - GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - GLuint face, i; - - DBG("%s\n", __FUNCTION__); - - for (face = 0; face < nr_faces; face++) { - for (i = stObj->firstLevel; i <= stObj->lastLevel; i++) { - struct st_texture_image *stImage = - st_texture_image(stObj->base.Image[face][i]); - - if (stImage->mt) { - stImage->base.Data - = st_miptree_image_map(pipe, - stImage->mt, - stImage->face, - stImage->level, - &stImage->base.RowStride, - stImage->base.ImageOffsets); - /* convert stride to texels, not bytes */ - stImage->base.RowStride /= stImage->mt->cpp; -/* stImage->base.ImageStride /= stImage->mt->cpp; */ - } - } - } -} - - - -void -st_tex_unmap_images(struct pipe_context *pipe, - struct st_texture_object *stObj) -{ - GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - GLuint face, i; - - for (face = 0; face < nr_faces; face++) { - for (i = stObj->firstLevel; i <= stObj->lastLevel; i++) { - struct st_texture_image *stImage = - st_texture_image(stObj->base.Image[face][i]); - - if (stImage->mt) { - st_miptree_image_unmap(pipe, stImage->mt); - stImage->base.Data = NULL; - } - } - } -} -#endif - - void -- cgit v1.2.3 From 28b315dc1aed36bebadfacbd55e481e7baacfcb5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Sep 2007 16:31:13 -0600 Subject: comments --- src/mesa/state_tracker/st_cb_texture.c | 7 ++++--- src/mesa/state_tracker/st_mipmap_tree.c | 6 +++--- 2 files changed, 7 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index e7be3df722..eea47b9b41 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1414,9 +1414,10 @@ copy_image_data_to_tree(struct pipe_context *pipe, /* Copy potentially with the blitter: */ st_miptree_image_copy(pipe, - stObj->mt, - stImage->face, - stImage->level, stImage->mt); + stObj->mt, /* dest miptree */ + stImage->face, stImage->level, + stImage->mt /* src miptree */ + ); st_miptree_release(pipe, &stImage->mt); } diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index 1764bc61f3..27bbf17608 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -309,10 +309,10 @@ st_miptree_image_copy(struct pipe_context *pipe, for (i = 0; i < depth; i++) { pipe->region_copy(pipe, dst->region, dst_offset + dst_depth_offset[i], - 0, - 0, + 0, 0, /* destX, Y */ src->region, src_offset + src_depth_offset[i], - 0, 0, width, height); + 0, 0, /* srcX, Y */ + width, height); } } -- cgit v1.2.3 From 7966e479dca22bf2d2b844d50ac5bef70614e15a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Sep 2007 16:33:01 -0600 Subject: set miptree pitch to region pitch after allocating the region in st_miptree_create() This fixes rendering with small (4x4) textures with softpipe. Haven't yet tested with i915. --- src/mesa/state_tracker/st_mipmap_tree.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index 27bbf17608..faca148d80 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -89,6 +89,7 @@ st_miptree_create(struct pipe_context *pipe, /* note: it's OK to pass 'pitch' as 'width' here: */ mt->region = pipe->region_alloc(pipe, mt->cpp, mt->pitch, mt->total_height, flags); + mt->pitch = mt->region->pitch; /*XXX NEW */ } if (!mt->region) { -- cgit v1.2.3 From 6dcfddb8e2ec2bfb6187b912807fa65f28da2c5e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 28 Sep 2007 04:33:55 -0400 Subject: Redoing the way we handle vertex shaders for the draw module. --- src/mesa/pipe/draw/draw_context.c | 8 ---- src/mesa/pipe/draw/draw_context.h | 9 ++-- src/mesa/pipe/draw/draw_private.h | 14 ++++++- src/mesa/pipe/draw/draw_vertex_fetch.c | 2 +- src/mesa/pipe/draw/draw_vertex_shader.c | 50 +++++++++++++++++++++-- src/mesa/pipe/i915simple/i915_state.c | 44 ++++++++++++-------- src/mesa/pipe/softpipe/sp_context.c | 8 ++-- src/mesa/pipe/softpipe/sp_context.h | 8 +++- src/mesa/pipe/softpipe/sp_state.h | 14 ++++--- src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- src/mesa/pipe/softpipe/sp_state_fs.c | 49 +++++++++++++++++----- src/mesa/state_tracker/st_cb_program.c | 7 ---- src/mesa/state_tracker/st_draw.c | 2 +- src/mesa/state_tracker/st_program.c | 8 ---- src/mesa/state_tracker/st_program.h | 4 -- src/mesa/x86/rtasm/x86sse.c | 68 +++++++++++++++---------------- src/mesa/x86/rtasm/x86sse.h | 49 +++++++++++----------- 17 files changed, 211 insertions(+), 135 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 66c66ff698..5efb173228 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -155,14 +155,6 @@ void draw_set_viewport_state( struct draw_context *draw, } -void -draw_set_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader) -{ - draw_flush( draw ); - draw->vertex_shader = *shader; -} - void draw_set_vertex_buffer(struct draw_context *draw, diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 0ccf5f6046..398e96d94d 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -92,9 +92,12 @@ void draw_set_rasterize_stage( struct draw_context *draw, struct draw_stage *stage ); -void -draw_set_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader); +void * draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader); +void draw_bind_vertex_shader(struct draw_context *draw, + void *vcso); +void draw_delete_vertex_shader(struct draw_context *draw, + void *vcso); void diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index ebef5347ab..2f52299f74 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -46,6 +46,7 @@ #include "draw_vertex.h" +#include "x86/rtasm/x86sse.h" /** * Basic vertex info. @@ -104,7 +105,7 @@ struct draw_stage void (*tri)( struct draw_stage *, struct prim_header * ); - + void (*end)( struct draw_stage * ); void (*reset_stipple_counter)( struct draw_stage * ); @@ -116,6 +117,15 @@ struct draw_stage #define VCACHE_OVERFLOW 4 #define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ +/** + * Private version of the compiled vertex_shader + */ +struct draw_vertex_shader { + const struct pipe_shader_state *state; +#if defined(__i386__) || defined(__386__) + struct x86_function sse2_program; +#endif +}; /** * Private context for the drawing module. @@ -145,7 +155,7 @@ struct draw_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - struct pipe_shader_state vertex_shader; + const struct draw_vertex_shader *vertex_shader; struct pipe_vertex_buffer feedback_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element feedback_element[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 2b839d641e..bc37e18c34 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -85,7 +85,7 @@ void draw_vertex_fetch( struct draw_context *draw, /*printf("fetch vertex %u: \n", j);*/ /* loop over vertex attributes (vertex shader inputs) */ - for (attr = 0; attr < draw->vertex_shader.num_inputs; attr++) { + for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { unsigned buf = draw->vertex_element[attr].vertex_buffer_index; const void *src diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index a2e1cdc472..f7ddbd1909 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -36,6 +36,8 @@ #include "draw_context.h" #include "draw_vertex.h" +#include "x86/rtasm/x86sse.h" + #include "pipe/tgsi/exec/tgsi_core.h" static INLINE unsigned @@ -70,6 +72,7 @@ typedef void (XSTDCALL *codegen_function) ( float (*constant)[4], struct tgsi_exec_vector *temporary ); + /** * Transform vertices with the current vertex program/shader * Up to four vertices can be shaded at a time. @@ -92,7 +95,7 @@ run_vertex_program(struct draw_context *draw, const float *trans = draw->viewport.translate; assert(count <= 4); - assert(draw->vertex_shader.output_semantic_name[0] + assert(draw->vertex_shader->state->output_semantic_name[0] == TGSI_SEMANTIC_POSITION); #ifdef DEBUG @@ -101,7 +104,7 @@ run_vertex_program(struct draw_context *draw, /* init machine state */ tgsi_exec_machine_init(&machine, - draw->vertex_shader.tokens, + draw->vertex_shader->state->tokens, PIPE_MAX_SAMPLERS, NULL /*samplers*/ ); @@ -114,8 +117,8 @@ run_vertex_program(struct draw_context *draw, draw_vertex_fetch( draw, &machine, elts, count ); /* run shader */ - if( draw->vertex_shader.executable != NULL ) { - codegen_function func = (codegen_function) draw->vertex_shader.executable; + if( draw->vertex_shader->state->executable != NULL ) { + codegen_function func = (codegen_function) draw->vertex_shader->state->executable; func( machine.Inputs, machine.Outputs, @@ -206,3 +209,42 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) draw->vs.queue_nr = 0; } + +void * +draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader) +{ + struct draw_vertex_shader *vs = calloc(1, sizeof(struct draw_vertex_shader)); + + vs->state = shader; +#if defined(__i386__) || defined(__386__) + x86_init_func(&vs->sse2_program); + + tgsi_emit_sse2(shader->tokens, &vs->sse2_program); + + ((struct pipe_shader_state*)(vs->state))->executable = + x86_get_func(&vs->sse2_program); +#endif + + return vs; +} + +void draw_bind_vertex_shader(struct draw_context *draw, + void *vcso) +{ + draw_flush(draw); + draw->vertex_shader = (struct draw_vertex_shader*)(vcso); +} + +void draw_delete_vertex_shader(struct draw_context *draw, + void *vcso) +{ + struct draw_vertex_shader *vs = (struct draw_vertex_shader*)(vcso); +#if defined(__i386__) || defined(__386__) + x86_release_func(&vs->sse2_program); +#endif + free(vcso); +} + + + diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 5c6d37a466..19ca5e575f 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -443,16 +443,13 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe, { } - -static void * -i915_create_shader_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) +static void * i915_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { return 0; } -static void i915_bind_fs_state( struct pipe_context *pipe, - void *fs ) +static void i915_bind_fs_state(struct pipe_context *pipe, void *fs) { struct i915_context *i915 = i915_context(pipe); @@ -461,20 +458,35 @@ static void i915_bind_fs_state( struct pipe_context *pipe, i915->dirty |= I915_NEW_FS; } +static void i915_delete_fs_state(struct pipe_context *pipe, void *shader) +{ + /*do nothing*/ +} -static void i915_bind_vs_state(struct pipe_context *pipe, - void *vs) +static void * +i915_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { struct i915_context *i915 = i915_context(pipe); /* just pass-through to draw module */ - draw_set_vertex_shader(i915->draw, (const struct pipe_shader_state *)vs); + return draw_create_vertex_shader(i915->draw, templ); } -static void i915_delete_shader_state(struct pipe_context *pipe, - void *shader) +static void i915_bind_vs_state(struct pipe_context *pipe, void *vs) { - /*do nothing*/ + struct i915_context *i915 = i915_context(pipe); + + /* just pass-through to draw module */ + draw_bind_vertex_shader(i915->draw, vs); +} + +static void i915_delete_vs_state(struct pipe_context *pipe, void *shader) +{ + struct i915_context *i915 = i915_context(pipe); + + /* just pass-through to draw module */ + draw_delete_vertex_shader(i915->draw, shader); } static void i915_set_constant_buffer(struct pipe_context *pipe, @@ -707,12 +719,12 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state; - i915->pipe.create_fs_state = i915_create_shader_state; + i915->pipe.create_fs_state = i915_create_fs_state; i915->pipe.bind_fs_state = i915_bind_fs_state; - i915->pipe.delete_fs_state = i915_delete_shader_state; - i915->pipe.create_vs_state = i915_create_shader_state; + i915->pipe.delete_fs_state = i915_delete_fs_state; + i915->pipe.create_vs_state = i915_create_vs_state; i915->pipe.bind_vs_state = i915_bind_vs_state; - i915->pipe.delete_vs_state = i915_delete_shader_state; + i915->pipe.delete_vs_state = i915_delete_vs_state; i915->pipe.set_blend_color = i915_set_blend_color; i915->pipe.set_clip_state = i915_set_clip_state; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index e415966dae..695bf1a9e0 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -275,12 +275,12 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; - softpipe->pipe.create_fs_state = softpipe_create_shader_state; + softpipe->pipe.create_fs_state = softpipe_create_fs_state; softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; - softpipe->pipe.delete_fs_state = softpipe_delete_shader_state; - softpipe->pipe.create_vs_state = softpipe_create_shader_state; + softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; + softpipe->pipe.create_vs_state = softpipe_create_vs_state; softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; - softpipe->pipe.delete_vs_state = softpipe_delete_shader_state; + softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; softpipe->pipe.set_clip_state = softpipe_set_clip_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index c0a681f3d6..ccf29b5683 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -62,8 +62,12 @@ struct draw_stage; #define SP_NEW_VS 0x2000 #define SP_NEW_CONSTANTS 0x4000 +struct sp_vertex_shader_state { + const struct pipe_shader_state *state; + void *draw_data; +}; -struct softpipe_context { +struct softpipe_context { struct pipe_context pipe; /**< base class */ struct softpipe_winsys *winsys; /**< window system interface */ @@ -76,7 +80,7 @@ struct softpipe_context { const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct pipe_shader_state *fs; - const struct pipe_shader_state *vs; + const struct sp_vertex_shader_state *vs; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index f0e1461d25..f9061e86e5 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -87,12 +87,14 @@ void softpipe_set_constant_buffer(struct pipe_context *, void softpipe_set_feedback_state( struct pipe_context *, const struct pipe_feedback_state * ); -void * -softpipe_create_shader_state( struct pipe_context *, - const struct pipe_shader_state * ); -void softpipe_bind_fs_state( struct pipe_context *, void * ); -void softpipe_bind_vs_state( struct pipe_context *, void * ); -void softpipe_delete_shader_state( struct pipe_context *, void * ); +void *softpipe_create_fs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_fs_state(struct pipe_context *, void *); +void softpipe_delete_fs_state(struct pipe_context *, void *); +void *softpipe_create_vs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_vs_state(struct pipe_context *, void *); +void softpipe_delete_vs_state(struct pipe_context *, void *); void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 6c6e798069..b6145df8e2 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -43,7 +43,7 @@ */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - const struct pipe_shader_state *vs = softpipe->vs; + const struct pipe_shader_state *vs = softpipe->vs->state; const struct pipe_shader_state *fs = softpipe->fs; const interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 8306a95f44..f1bec2c73a 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -33,10 +33,13 @@ #include "pipe/draw/draw_context.h" -void * softpipe_create_shader_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) +void * softpipe_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { - /* we just want the pipe_shader_state template in the bind calls */ + /* Decide whether we'll be codegenerating this shader and if so do + * that now. + */ + return 0; } @@ -49,25 +52,51 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) softpipe->dirty |= SP_NEW_FS; } +void softpipe_delete_fs_state(struct pipe_context *pipe, + void *shader) +{ +} + + +void * softpipe_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + struct sp_vertex_shader_state *state = + malloc(sizeof(struct sp_vertex_shader_state)); + + state->state = templ; + state->draw_data = draw_create_vertex_shader(softpipe->draw, + state->state); + + return state; +} void softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->vs = (struct pipe_shader_state *)vs; + softpipe->vs = (const struct sp_vertex_shader_state *)vs; - softpipe->dirty |= SP_NEW_VS; + draw_bind_vertex_shader(softpipe->draw, softpipe->vs->draw_data); - draw_set_vertex_shader(softpipe->draw, (struct pipe_shader_state *)vs); + softpipe->dirty |= SP_NEW_VS; } - -void softpipe_delete_shader_state( struct pipe_context *pipe, - void *shader ) +void softpipe_delete_vs_state(struct pipe_context *pipe, + void *vs) { - /* do nothing */ + struct softpipe_context *softpipe = softpipe_context(pipe); + + struct sp_vertex_shader_state *state = + (struct sp_vertex_shader_state *)vs; + + draw_delete_vertex_shader(softpipe->draw, state->draw_data); + free(state); } + + void softpipe_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, const struct pipe_constant_buffer *buf) diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 25dbc2c244..3a7ce9405e 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -86,10 +86,6 @@ static struct gl_program *st_new_program( GLcontext *ctx, prog->serialNo = 1; -#if defined(__i386__) || defined(__386__) - x86_init_func( &prog->sse2_program ); -#endif - return _mesa_init_vertex_program( ctx, &prog->Base, target, @@ -129,9 +125,6 @@ static void st_delete_program( GLcontext *ctx, case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; -#if defined(__i386__) || defined(__386__) - x86_release_func( &stvp->sse2_program ); -#endif st_remove_vertex_program(st, stvp); } break; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index ce5bf0c8a9..7c836ba2c1 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -405,7 +405,7 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); draw_set_rasterizer_state(draw, &st->state.rasterizer->state); - draw_set_vertex_shader(draw, &st->state.vs->state); + draw_bind_vertex_shader(draw, st->state.vs->data); /* XXX need to set vertex info too */ diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 44cc33b09f..116b59a067 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -253,14 +253,6 @@ st_translate_vertex_program(struct st_context *st, if (TGSI_DEBUG) tgsi_dump( tokensOut, 0 ); -#if defined(__i386__) || defined(__386__) - if (stvp->sse2_program.csr == stvp->sse2_program.store) - tgsi_emit_sse2( tokensOut, &stvp->sse2_program ); - - if (!cso->state.executable) - ((struct cso_vertex_shader*)cso)->state.executable = (void *) x86_get_func( &stvp->sse2_program ); -#endif - return cso; } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 6e106c16b0..2b79201313 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -79,10 +79,6 @@ struct st_vertex_program /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif - /** Pointer to the corresponding cached shader */ const struct cso_vertex_shader *vs; diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index a6cfa40f91..f9a9352e46 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -6,22 +6,22 @@ #define DISASSEM 0 #define X86_TWOB 0x0f -static GLubyte *cptr( void (*label)() ) +static unsigned char *cptr( void (*label)() ) { - return (char *)(unsigned long)label; + return (unsigned char *)(unsigned long)label; } /* Emit bytes to the instruction stream: */ -static void emit_1b( struct x86_function *p, GLbyte b0 ) +static void emit_1b( struct x86_function *p, char b0 ) { - *(GLbyte *)(p->csr++) = b0; + *(char *)(p->csr++) = b0; } -static void emit_1i( struct x86_function *p, GLint i0 ) +static void emit_1i( struct x86_function *p, int i0 ) { - *(GLint *)(p->csr) = i0; + *(int *)(p->csr) = i0; p->csr += 4; } @@ -35,20 +35,20 @@ static void disassem( struct x86_function *p, const char *fn ) #endif } -static void emit_1ub_fn( struct x86_function *p, GLubyte b0, const char *fn ) +static void emit_1ub_fn( struct x86_function *p, unsigned char b0, const char *fn ) { disassem(p, fn); *(p->csr++) = b0; } -static void emit_2ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, const char *fn ) +static void emit_2ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, const char *fn ) { disassem(p, fn); *(p->csr++) = b0; *(p->csr++) = b1; } -static void emit_3ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, GLubyte b2, const char *fn ) +static void emit_3ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2, const char *fn ) { disassem(p, fn); *(p->csr++) = b0; @@ -69,7 +69,7 @@ static void emit_modrm( struct x86_function *p, struct x86_reg reg, struct x86_reg regmem ) { - GLubyte val = 0; + unsigned char val = 0; assert(reg.mod == mod_REG); @@ -104,7 +104,7 @@ static void emit_modrm( struct x86_function *p, static void emit_modrm_noreg( struct x86_function *p, - GLuint op, + unsigned op, struct x86_reg regmem ) { struct x86_reg dummy = x86_make_reg(file_REG32, op); @@ -117,8 +117,8 @@ static void emit_modrm_noreg( struct x86_function *p, * the arguments presented. */ static void emit_op_modrm( struct x86_function *p, - GLubyte op_dst_is_reg, - GLubyte op_dst_is_mem, + unsigned char op_dst_is_reg, + unsigned char op_dst_is_mem, struct x86_reg dst, struct x86_reg src ) { @@ -162,7 +162,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file, } struct x86_reg x86_make_disp( struct x86_reg reg, - GLint disp ) + int disp ) { assert(reg.file == file_REG32); @@ -191,7 +191,7 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg ) return x86_make_reg( reg.file, reg.idx ); } -GLubyte *x86_get_label( struct x86_function *p ) +unsigned char *x86_get_label( struct x86_function *p ) { return p->csr; } @@ -205,13 +205,13 @@ GLubyte *x86_get_label( struct x86_function *p ) void x86_jcc( struct x86_function *p, enum x86_cc cc, - GLubyte *label ) + unsigned char *label ) { - GLint offset = label - (x86_get_label(p) + 2); + int offset = label - (x86_get_label(p) + 2); if (offset <= 127 && offset >= -128) { emit_1ub(p, 0x70 + cc); - emit_1b(p, (GLbyte) offset); + emit_1b(p, (char) offset); } else { offset = label - (x86_get_label(p) + 6); @@ -222,7 +222,7 @@ void x86_jcc( struct x86_function *p, /* Always use a 32bit offset for forward jumps: */ -GLubyte *x86_jcc_forward( struct x86_function *p, +unsigned char *x86_jcc_forward( struct x86_function *p, enum x86_cc cc ) { emit_2ub(p, 0x0f, 0x80 + cc); @@ -230,14 +230,14 @@ GLubyte *x86_jcc_forward( struct x86_function *p, return x86_get_label(p); } -GLubyte *x86_jmp_forward( struct x86_function *p) +unsigned char *x86_jmp_forward( struct x86_function *p) { emit_1ub(p, 0xe9); emit_1i(p, 0); return x86_get_label(p); } -GLubyte *x86_call_forward( struct x86_function *p) +unsigned char *x86_call_forward( struct x86_function *p) { emit_1ub(p, 0xe8); emit_1i(p, 0); @@ -247,12 +247,12 @@ GLubyte *x86_call_forward( struct x86_function *p) /* Fixup offset from forward jump: */ void x86_fixup_fwd_jump( struct x86_function *p, - GLubyte *fixup ) + unsigned char *fixup ) { *(int *)(fixup - 4) = x86_get_label(p) - fixup; } -void x86_jmp( struct x86_function *p, GLubyte *label) +void x86_jmp( struct x86_function *p, unsigned char *label) { emit_1ub(p, 0xe9); emit_1i(p, label - x86_get_label(p) - 4); @@ -268,7 +268,7 @@ void x86_call( struct x86_function *p, void (*label)()) * Temporary. As I need immediate operands, and dont want to mess with the codegen, * I load the immediate into general purpose register and use it. */ -void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm ) +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) { assert(dst.mod == mod_REG); emit_1ub(p, 0xb8 + dst.idx); @@ -595,7 +595,7 @@ void sse_cvtps2pi( struct x86_function *p, void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - GLubyte shuf) + unsigned char shuf) { emit_2ub(p, X86_TWOB, 0xC6); emit_modrm(p, dest, arg0); @@ -605,7 +605,7 @@ void sse_shufps( struct x86_function *p, void sse_cmpps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - GLubyte cc) + unsigned char cc) { emit_2ub(p, X86_TWOB, 0xC2); emit_modrm(p, dest, arg0); @@ -630,7 +630,7 @@ void sse_pmovmskb( struct x86_function *p, void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - GLubyte shuf) + unsigned char shuf) { emit_3ub(p, 0x66, X86_TWOB, 0x70); emit_modrm(p, dest, arg0); @@ -772,11 +772,11 @@ void x87_fclex( struct x86_function *p ) static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg, - GLubyte dst0ub0, - GLubyte dst0ub1, - GLubyte arg0ub0, - GLubyte arg0ub1, - GLubyte argmem_noreg) + unsigned char dst0ub0, + unsigned char dst0ub1, + unsigned char arg0ub0, + unsigned char arg0ub1, + unsigned char argmem_noreg) { assert(dst.file == file_x87); @@ -1116,7 +1116,7 @@ void mmx_movq( struct x86_function *p, * account any push/pop activity: */ struct x86_reg x86_fn_arg( struct x86_function *p, - GLuint arg ) + unsigned arg ) { return x86_make_disp(x86_make_reg(file_REG32, reg_SP), p->stack_offset + arg * 4); /* ??? */ @@ -1128,7 +1128,7 @@ void x86_init_func( struct x86_function *p ) x86_init_func_size(p, 1024); } -void x86_init_func_size( struct x86_function *p, GLuint code_size ) +void x86_init_func_size( struct x86_function *p, unsigned code_size ) { p->store = _mesa_exec_malloc(code_size); p->csr = p->store; diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h index c1ca06088b..63b9a36392 100644 --- a/src/mesa/x86/rtasm/x86sse.h +++ b/src/mesa/x86/rtasm/x86sse.h @@ -4,24 +4,22 @@ #if defined(__i386__) || defined(__386__) -#include "glheader.h" - /* It is up to the caller to ensure that instructions issued are * suitable for the host cpu. There are no checks made in this module * for mmx/sse/sse2 support on the cpu. */ struct x86_reg { - GLuint file:3; - GLuint idx:3; - GLuint mod:2; /* mod_REG if this is just a register */ - GLint disp:24; /* only +/- 23bits of offset - should be enough... */ + unsigned file:3; + unsigned idx:3; + unsigned mod:2; /* mod_REG if this is just a register */ + int disp:24; /* only +/- 23bits of offset - should be enough... */ }; struct x86_function { - GLubyte *store; - GLubyte *csr; - GLuint stack_offset; - GLint need_emms; + unsigned char *store; + unsigned char *csr; + unsigned stack_offset; + int need_emms; const char *fn; }; @@ -81,7 +79,7 @@ enum sse_cc { void x86_init_func( struct x86_function *p ); -void x86_init_func_size( struct x86_function *p, GLuint code_size ); +void x86_init_func_size( struct x86_function *p, unsigned code_size ); void x86_release_func( struct x86_function *p ); void (*x86_get_func( struct x86_function *p ))( void ); @@ -93,7 +91,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file, enum x86_reg_name idx ); struct x86_reg x86_make_disp( struct x86_reg reg, - GLint disp ); + int disp ); struct x86_reg x86_deref( struct x86_reg reg ); @@ -102,23 +100,23 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg ); /* Labels, jumps and fixup: */ -GLubyte *x86_get_label( struct x86_function *p ); +unsigned char *x86_get_label( struct x86_function *p ); void x86_jcc( struct x86_function *p, enum x86_cc cc, - GLubyte *label ); + unsigned char *label ); -GLubyte *x86_jcc_forward( struct x86_function *p, +unsigned char *x86_jcc_forward( struct x86_function *p, enum x86_cc cc ); -GLubyte *x86_jmp_forward( struct x86_function *p); +unsigned char *x86_jmp_forward( struct x86_function *p); -GLubyte *x86_call_forward( struct x86_function *p); +unsigned char *x86_call_forward( struct x86_function *p); void x86_fixup_fwd_jump( struct x86_function *p, - GLubyte *fixup ); + unsigned char *fixup ); -void x86_jmp( struct x86_function *p, GLubyte *label ); +void x86_jmp( struct x86_function *p, unsigned char *label ); void x86_call( struct x86_function *p, void (*label)() ); @@ -126,7 +124,7 @@ void x86_call( struct x86_function *p, void (*label)() ); * Temporary. As I need immediate operands, and dont want to mess with the codegen, * I load the immediate into general purpose register and use it. */ -void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm ); +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ); /* Macro for sse_shufps() and sse2_pshufd(): @@ -147,7 +145,8 @@ void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf ); +void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, + unsigned char shuf ); void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -157,7 +156,8 @@ void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg sr void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, GLubyte cc ); +void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, + unsigned char cc ); void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -175,7 +175,8 @@ void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf ); +void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, + unsigned char shuf ); void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -247,7 +248,7 @@ void x87_fucom( struct x86_function *p, struct x86_reg arg ); * account any push/pop activity. Note - doesn't track explict * manipulation of ESP by other instructions. */ -struct x86_reg x86_fn_arg( struct x86_function *p, GLuint arg ); +struct x86_reg x86_fn_arg( struct x86_function *p, unsigned arg ); #endif #endif -- cgit v1.2.3 From 57b5d3605745c96ddc2b6de7d50c93db65ba1257 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 28 Sep 2007 07:33:54 -0400 Subject: Use sse only if GALLIUM_SSE is defined --- src/mesa/pipe/draw/draw_context.c | 11 +++++++++++ src/mesa/pipe/draw/draw_context.h | 1 + src/mesa/pipe/draw/draw_private.h | 2 ++ src/mesa/pipe/draw/draw_vertex_shader.c | 9 +++++---- src/mesa/state_tracker/st_program.c | 11 +++++++---- 5 files changed, 26 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 5efb173228..3fb667ab1a 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -41,6 +41,12 @@ struct draw_context *draw_create( void ) { struct draw_context *draw = CALLOC_STRUCT( draw_context ); +#if defined(__i386__) || defined(__386__) + draw->use_sse = getenv("GALLIUM_SSE") != NULL; +#else + draw->use_sse = false; +#endif + /* create pipeline stages */ draw->pipeline.unfilled = draw_unfilled_stage( draw ); draw->pipeline.twoside = draw_twoside_stage( draw ); @@ -244,4 +250,9 @@ void draw_free_tmps( struct draw_stage *stage ) } } +boolean draw_use_sse(struct draw_context *draw) +{ + return draw->use_sse; +} + diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 398e96d94d..8e2232244c 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -99,6 +99,7 @@ void draw_bind_vertex_shader(struct draw_context *draw, void draw_delete_vertex_shader(struct draw_context *draw, void *vcso); +boolean draw_use_sse(struct draw_context *draw); void draw_set_vertex_buffer(struct draw_context *draw, diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 2f52299f74..12a970a671 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -220,6 +220,8 @@ struct draw_context struct prim_header queue[PRIM_QUEUE_LENGTH]; unsigned queue_nr; } pq; + + int use_sse : 1; }; diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index f7ddbd1909..3518bd52a3 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -220,10 +220,11 @@ draw_create_vertex_shader(struct draw_context *draw, #if defined(__i386__) || defined(__386__) x86_init_func(&vs->sse2_program); - tgsi_emit_sse2(shader->tokens, &vs->sse2_program); - - ((struct pipe_shader_state*)(vs->state))->executable = - x86_get_func(&vs->sse2_program); + if (draw->use_sse) { + tgsi_emit_sse2(shader->tokens, &vs->sse2_program); + ((struct pipe_shader_state*)(vs->state))->executable = + x86_get_func(&vs->sse2_program); + } #endif return vs; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 116b59a067..a41b953a21 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -36,6 +36,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/draw/draw_context.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/exec/tgsi_core.h" @@ -391,11 +392,13 @@ st_translate_fragment_program(struct st_context *st, tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); #if defined(__i386__) || defined(__386__) - if (stfp->sse2_program.csr == stfp->sse2_program.store) - tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); + if (draw_use_sse(st->draw)) { + if (stfp->sse2_program.csr == stfp->sse2_program.store) + tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); - if (!cso->state.executable) - ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); + if (!cso->state.executable) + ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); + } #endif return cso; -- cgit v1.2.3 From f2a33a63f1f41681375baa2a9ad261cb60db2a85 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 28 Sep 2007 12:28:06 -0400 Subject: Revert "Use sse only if GALLIUM_SSE is defined" This reverts commit 57b5d3605745c96ddc2b6de7d50c93db65ba1257. --- src/mesa/pipe/draw/draw_context.c | 11 ----------- src/mesa/pipe/draw/draw_context.h | 1 - src/mesa/pipe/draw/draw_private.h | 2 -- src/mesa/pipe/draw/draw_vertex_shader.c | 9 ++++----- src/mesa/state_tracker/st_program.c | 11 ++++------- 5 files changed, 8 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 3fb667ab1a..5efb173228 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -41,12 +41,6 @@ struct draw_context *draw_create( void ) { struct draw_context *draw = CALLOC_STRUCT( draw_context ); -#if defined(__i386__) || defined(__386__) - draw->use_sse = getenv("GALLIUM_SSE") != NULL; -#else - draw->use_sse = false; -#endif - /* create pipeline stages */ draw->pipeline.unfilled = draw_unfilled_stage( draw ); draw->pipeline.twoside = draw_twoside_stage( draw ); @@ -250,9 +244,4 @@ void draw_free_tmps( struct draw_stage *stage ) } } -boolean draw_use_sse(struct draw_context *draw) -{ - return draw->use_sse; -} - diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 8e2232244c..398e96d94d 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -99,7 +99,6 @@ void draw_bind_vertex_shader(struct draw_context *draw, void draw_delete_vertex_shader(struct draw_context *draw, void *vcso); -boolean draw_use_sse(struct draw_context *draw); void draw_set_vertex_buffer(struct draw_context *draw, diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 12a970a671..2f52299f74 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -220,8 +220,6 @@ struct draw_context struct prim_header queue[PRIM_QUEUE_LENGTH]; unsigned queue_nr; } pq; - - int use_sse : 1; }; diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 3518bd52a3..f7ddbd1909 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -220,11 +220,10 @@ draw_create_vertex_shader(struct draw_context *draw, #if defined(__i386__) || defined(__386__) x86_init_func(&vs->sse2_program); - if (draw->use_sse) { - tgsi_emit_sse2(shader->tokens, &vs->sse2_program); - ((struct pipe_shader_state*)(vs->state))->executable = - x86_get_func(&vs->sse2_program); - } + tgsi_emit_sse2(shader->tokens, &vs->sse2_program); + + ((struct pipe_shader_state*)(vs->state))->executable = + x86_get_func(&vs->sse2_program); #endif return vs; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index a41b953a21..116b59a067 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -36,7 +36,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/draw/draw_context.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/exec/tgsi_core.h" @@ -392,13 +391,11 @@ st_translate_fragment_program(struct st_context *st, tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); #if defined(__i386__) || defined(__386__) - if (draw_use_sse(st->draw)) { - if (stfp->sse2_program.csr == stfp->sse2_program.store) - tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); + if (stfp->sse2_program.csr == stfp->sse2_program.store) + tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); - if (!cso->state.executable) - ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); - } + if (!cso->state.executable) + ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); #endif return cso; -- cgit v1.2.3 From 901577e07fcab0cf90a272fee900cb0831ae84c3 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 28 Sep 2007 12:28:16 -0400 Subject: Revert "Redoing the way we handle vertex shaders for the draw module." This reverts commit 6dcfddb8e2ec2bfb6187b912807fa65f28da2c5e. --- src/mesa/pipe/draw/draw_context.c | 8 ++++ src/mesa/pipe/draw/draw_context.h | 9 ++-- src/mesa/pipe/draw/draw_private.h | 14 +------ src/mesa/pipe/draw/draw_vertex_fetch.c | 2 +- src/mesa/pipe/draw/draw_vertex_shader.c | 50 ++--------------------- src/mesa/pipe/i915simple/i915_state.c | 44 ++++++++------------ src/mesa/pipe/softpipe/sp_context.c | 8 ++-- src/mesa/pipe/softpipe/sp_context.h | 8 +--- src/mesa/pipe/softpipe/sp_state.h | 14 +++---- src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- src/mesa/pipe/softpipe/sp_state_fs.c | 49 +++++----------------- src/mesa/state_tracker/st_cb_program.c | 7 ++++ src/mesa/state_tracker/st_draw.c | 2 +- src/mesa/state_tracker/st_program.c | 8 ++++ src/mesa/state_tracker/st_program.h | 4 ++ src/mesa/x86/rtasm/x86sse.c | 68 +++++++++++++++---------------- src/mesa/x86/rtasm/x86sse.h | 49 +++++++++++----------- 17 files changed, 135 insertions(+), 211 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 5efb173228..66c66ff698 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -155,6 +155,14 @@ void draw_set_viewport_state( struct draw_context *draw, } +void +draw_set_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader) +{ + draw_flush( draw ); + draw->vertex_shader = *shader; +} + void draw_set_vertex_buffer(struct draw_context *draw, diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 398e96d94d..0ccf5f6046 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -92,12 +92,9 @@ void draw_set_rasterize_stage( struct draw_context *draw, struct draw_stage *stage ); -void * draw_create_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader); -void draw_bind_vertex_shader(struct draw_context *draw, - void *vcso); -void draw_delete_vertex_shader(struct draw_context *draw, - void *vcso); +void +draw_set_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader); void diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 2f52299f74..ebef5347ab 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -46,7 +46,6 @@ #include "draw_vertex.h" -#include "x86/rtasm/x86sse.h" /** * Basic vertex info. @@ -105,7 +104,7 @@ struct draw_stage void (*tri)( struct draw_stage *, struct prim_header * ); - + void (*end)( struct draw_stage * ); void (*reset_stipple_counter)( struct draw_stage * ); @@ -117,15 +116,6 @@ struct draw_stage #define VCACHE_OVERFLOW 4 #define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ -/** - * Private version of the compiled vertex_shader - */ -struct draw_vertex_shader { - const struct pipe_shader_state *state; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif -}; /** * Private context for the drawing module. @@ -155,7 +145,7 @@ struct draw_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - const struct draw_vertex_shader *vertex_shader; + struct pipe_shader_state vertex_shader; struct pipe_vertex_buffer feedback_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element feedback_element[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index bc37e18c34..2b839d641e 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -85,7 +85,7 @@ void draw_vertex_fetch( struct draw_context *draw, /*printf("fetch vertex %u: \n", j);*/ /* loop over vertex attributes (vertex shader inputs) */ - for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { + for (attr = 0; attr < draw->vertex_shader.num_inputs; attr++) { unsigned buf = draw->vertex_element[attr].vertex_buffer_index; const void *src diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index f7ddbd1909..a2e1cdc472 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -36,8 +36,6 @@ #include "draw_context.h" #include "draw_vertex.h" -#include "x86/rtasm/x86sse.h" - #include "pipe/tgsi/exec/tgsi_core.h" static INLINE unsigned @@ -72,7 +70,6 @@ typedef void (XSTDCALL *codegen_function) ( float (*constant)[4], struct tgsi_exec_vector *temporary ); - /** * Transform vertices with the current vertex program/shader * Up to four vertices can be shaded at a time. @@ -95,7 +92,7 @@ run_vertex_program(struct draw_context *draw, const float *trans = draw->viewport.translate; assert(count <= 4); - assert(draw->vertex_shader->state->output_semantic_name[0] + assert(draw->vertex_shader.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); #ifdef DEBUG @@ -104,7 +101,7 @@ run_vertex_program(struct draw_context *draw, /* init machine state */ tgsi_exec_machine_init(&machine, - draw->vertex_shader->state->tokens, + draw->vertex_shader.tokens, PIPE_MAX_SAMPLERS, NULL /*samplers*/ ); @@ -117,8 +114,8 @@ run_vertex_program(struct draw_context *draw, draw_vertex_fetch( draw, &machine, elts, count ); /* run shader */ - if( draw->vertex_shader->state->executable != NULL ) { - codegen_function func = (codegen_function) draw->vertex_shader->state->executable; + if( draw->vertex_shader.executable != NULL ) { + codegen_function func = (codegen_function) draw->vertex_shader.executable; func( machine.Inputs, machine.Outputs, @@ -209,42 +206,3 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) draw->vs.queue_nr = 0; } - -void * -draw_create_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader) -{ - struct draw_vertex_shader *vs = calloc(1, sizeof(struct draw_vertex_shader)); - - vs->state = shader; -#if defined(__i386__) || defined(__386__) - x86_init_func(&vs->sse2_program); - - tgsi_emit_sse2(shader->tokens, &vs->sse2_program); - - ((struct pipe_shader_state*)(vs->state))->executable = - x86_get_func(&vs->sse2_program); -#endif - - return vs; -} - -void draw_bind_vertex_shader(struct draw_context *draw, - void *vcso) -{ - draw_flush(draw); - draw->vertex_shader = (struct draw_vertex_shader*)(vcso); -} - -void draw_delete_vertex_shader(struct draw_context *draw, - void *vcso) -{ - struct draw_vertex_shader *vs = (struct draw_vertex_shader*)(vcso); -#if defined(__i386__) || defined(__386__) - x86_release_func(&vs->sse2_program); -#endif - free(vcso); -} - - - diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 19ca5e575f..5c6d37a466 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -443,13 +443,16 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe, { } -static void * i915_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) + +static void * +i915_create_shader_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { return 0; } -static void i915_bind_fs_state(struct pipe_context *pipe, void *fs) +static void i915_bind_fs_state( struct pipe_context *pipe, + void *fs ) { struct i915_context *i915 = i915_context(pipe); @@ -458,35 +461,20 @@ static void i915_bind_fs_state(struct pipe_context *pipe, void *fs) i915->dirty |= I915_NEW_FS; } -static void i915_delete_fs_state(struct pipe_context *pipe, void *shader) -{ - /*do nothing*/ -} - -static void * -i915_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - struct i915_context *i915 = i915_context(pipe); - - /* just pass-through to draw module */ - return draw_create_vertex_shader(i915->draw, templ); -} -static void i915_bind_vs_state(struct pipe_context *pipe, void *vs) +static void i915_bind_vs_state(struct pipe_context *pipe, + void *vs) { struct i915_context *i915 = i915_context(pipe); /* just pass-through to draw module */ - draw_bind_vertex_shader(i915->draw, vs); + draw_set_vertex_shader(i915->draw, (const struct pipe_shader_state *)vs); } -static void i915_delete_vs_state(struct pipe_context *pipe, void *shader) +static void i915_delete_shader_state(struct pipe_context *pipe, + void *shader) { - struct i915_context *i915 = i915_context(pipe); - - /* just pass-through to draw module */ - draw_delete_vertex_shader(i915->draw, shader); + /*do nothing*/ } static void i915_set_constant_buffer(struct pipe_context *pipe, @@ -719,12 +707,12 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state; - i915->pipe.create_fs_state = i915_create_fs_state; + i915->pipe.create_fs_state = i915_create_shader_state; i915->pipe.bind_fs_state = i915_bind_fs_state; - i915->pipe.delete_fs_state = i915_delete_fs_state; - i915->pipe.create_vs_state = i915_create_vs_state; + i915->pipe.delete_fs_state = i915_delete_shader_state; + i915->pipe.create_vs_state = i915_create_shader_state; i915->pipe.bind_vs_state = i915_bind_vs_state; - i915->pipe.delete_vs_state = i915_delete_vs_state; + i915->pipe.delete_vs_state = i915_delete_shader_state; i915->pipe.set_blend_color = i915_set_blend_color; i915->pipe.set_clip_state = i915_set_clip_state; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 695bf1a9e0..e415966dae 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -275,12 +275,12 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; - softpipe->pipe.create_fs_state = softpipe_create_fs_state; + softpipe->pipe.create_fs_state = softpipe_create_shader_state; softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; - softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; - softpipe->pipe.create_vs_state = softpipe_create_vs_state; + softpipe->pipe.delete_fs_state = softpipe_delete_shader_state; + softpipe->pipe.create_vs_state = softpipe_create_shader_state; softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; - softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; + softpipe->pipe.delete_vs_state = softpipe_delete_shader_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; softpipe->pipe.set_clip_state = softpipe_set_clip_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index ccf29b5683..c0a681f3d6 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -62,12 +62,8 @@ struct draw_stage; #define SP_NEW_VS 0x2000 #define SP_NEW_CONSTANTS 0x4000 -struct sp_vertex_shader_state { - const struct pipe_shader_state *state; - void *draw_data; -}; -struct softpipe_context { +struct softpipe_context { struct pipe_context pipe; /**< base class */ struct softpipe_winsys *winsys; /**< window system interface */ @@ -80,7 +76,7 @@ struct softpipe_context { const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct pipe_shader_state *fs; - const struct sp_vertex_shader_state *vs; + const struct pipe_shader_state *vs; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index f9061e86e5..f0e1461d25 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -87,14 +87,12 @@ void softpipe_set_constant_buffer(struct pipe_context *, void softpipe_set_feedback_state( struct pipe_context *, const struct pipe_feedback_state * ); -void *softpipe_create_fs_state(struct pipe_context *, - const struct pipe_shader_state *); -void softpipe_bind_fs_state(struct pipe_context *, void *); -void softpipe_delete_fs_state(struct pipe_context *, void *); -void *softpipe_create_vs_state(struct pipe_context *, - const struct pipe_shader_state *); -void softpipe_bind_vs_state(struct pipe_context *, void *); -void softpipe_delete_vs_state(struct pipe_context *, void *); +void * +softpipe_create_shader_state( struct pipe_context *, + const struct pipe_shader_state * ); +void softpipe_bind_fs_state( struct pipe_context *, void * ); +void softpipe_bind_vs_state( struct pipe_context *, void * ); +void softpipe_delete_shader_state( struct pipe_context *, void * ); void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index b6145df8e2..6c6e798069 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -43,7 +43,7 @@ */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - const struct pipe_shader_state *vs = softpipe->vs->state; + const struct pipe_shader_state *vs = softpipe->vs; const struct pipe_shader_state *fs = softpipe->fs; const interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index f1bec2c73a..8306a95f44 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -33,13 +33,10 @@ #include "pipe/draw/draw_context.h" -void * softpipe_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) +void * softpipe_create_shader_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { - /* Decide whether we'll be codegenerating this shader and if so do - * that now. - */ - + /* we just want the pipe_shader_state template in the bind calls */ return 0; } @@ -52,50 +49,24 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) softpipe->dirty |= SP_NEW_FS; } -void softpipe_delete_fs_state(struct pipe_context *pipe, - void *shader) -{ -} - - -void * softpipe_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - struct sp_vertex_shader_state *state = - malloc(sizeof(struct sp_vertex_shader_state)); - - state->state = templ; - state->draw_data = draw_create_vertex_shader(softpipe->draw, - state->state); - - return state; -} void softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->vs = (const struct sp_vertex_shader_state *)vs; - - draw_bind_vertex_shader(softpipe->draw, softpipe->vs->draw_data); + softpipe->vs = (struct pipe_shader_state *)vs; softpipe->dirty |= SP_NEW_VS; -} -void softpipe_delete_vs_state(struct pipe_context *pipe, - void *vs) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - struct sp_vertex_shader_state *state = - (struct sp_vertex_shader_state *)vs; - - draw_delete_vertex_shader(softpipe->draw, state->draw_data); - free(state); + draw_set_vertex_shader(softpipe->draw, (struct pipe_shader_state *)vs); } +void softpipe_delete_shader_state( struct pipe_context *pipe, + void *shader ) +{ + /* do nothing */ +} void softpipe_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 3a7ce9405e..25dbc2c244 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -86,6 +86,10 @@ static struct gl_program *st_new_program( GLcontext *ctx, prog->serialNo = 1; +#if defined(__i386__) || defined(__386__) + x86_init_func( &prog->sse2_program ); +#endif + return _mesa_init_vertex_program( ctx, &prog->Base, target, @@ -125,6 +129,9 @@ static void st_delete_program( GLcontext *ctx, case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; +#if defined(__i386__) || defined(__386__) + x86_release_func( &stvp->sse2_program ); +#endif st_remove_vertex_program(st, stvp); } break; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 7c836ba2c1..ce5bf0c8a9 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -405,7 +405,7 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); draw_set_rasterizer_state(draw, &st->state.rasterizer->state); - draw_bind_vertex_shader(draw, st->state.vs->data); + draw_set_vertex_shader(draw, &st->state.vs->state); /* XXX need to set vertex info too */ diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 116b59a067..44cc33b09f 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -253,6 +253,14 @@ st_translate_vertex_program(struct st_context *st, if (TGSI_DEBUG) tgsi_dump( tokensOut, 0 ); +#if defined(__i386__) || defined(__386__) + if (stvp->sse2_program.csr == stvp->sse2_program.store) + tgsi_emit_sse2( tokensOut, &stvp->sse2_program ); + + if (!cso->state.executable) + ((struct cso_vertex_shader*)cso)->state.executable = (void *) x86_get_func( &stvp->sse2_program ); +#endif + return cso; } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 2b79201313..6e106c16b0 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -79,6 +79,10 @@ struct st_vertex_program /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; +#if defined(__i386__) || defined(__386__) + struct x86_function sse2_program; +#endif + /** Pointer to the corresponding cached shader */ const struct cso_vertex_shader *vs; diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index f9a9352e46..a6cfa40f91 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -6,22 +6,22 @@ #define DISASSEM 0 #define X86_TWOB 0x0f -static unsigned char *cptr( void (*label)() ) +static GLubyte *cptr( void (*label)() ) { - return (unsigned char *)(unsigned long)label; + return (char *)(unsigned long)label; } /* Emit bytes to the instruction stream: */ -static void emit_1b( struct x86_function *p, char b0 ) +static void emit_1b( struct x86_function *p, GLbyte b0 ) { - *(char *)(p->csr++) = b0; + *(GLbyte *)(p->csr++) = b0; } -static void emit_1i( struct x86_function *p, int i0 ) +static void emit_1i( struct x86_function *p, GLint i0 ) { - *(int *)(p->csr) = i0; + *(GLint *)(p->csr) = i0; p->csr += 4; } @@ -35,20 +35,20 @@ static void disassem( struct x86_function *p, const char *fn ) #endif } -static void emit_1ub_fn( struct x86_function *p, unsigned char b0, const char *fn ) +static void emit_1ub_fn( struct x86_function *p, GLubyte b0, const char *fn ) { disassem(p, fn); *(p->csr++) = b0; } -static void emit_2ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, const char *fn ) +static void emit_2ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, const char *fn ) { disassem(p, fn); *(p->csr++) = b0; *(p->csr++) = b1; } -static void emit_3ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2, const char *fn ) +static void emit_3ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, GLubyte b2, const char *fn ) { disassem(p, fn); *(p->csr++) = b0; @@ -69,7 +69,7 @@ static void emit_modrm( struct x86_function *p, struct x86_reg reg, struct x86_reg regmem ) { - unsigned char val = 0; + GLubyte val = 0; assert(reg.mod == mod_REG); @@ -104,7 +104,7 @@ static void emit_modrm( struct x86_function *p, static void emit_modrm_noreg( struct x86_function *p, - unsigned op, + GLuint op, struct x86_reg regmem ) { struct x86_reg dummy = x86_make_reg(file_REG32, op); @@ -117,8 +117,8 @@ static void emit_modrm_noreg( struct x86_function *p, * the arguments presented. */ static void emit_op_modrm( struct x86_function *p, - unsigned char op_dst_is_reg, - unsigned char op_dst_is_mem, + GLubyte op_dst_is_reg, + GLubyte op_dst_is_mem, struct x86_reg dst, struct x86_reg src ) { @@ -162,7 +162,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file, } struct x86_reg x86_make_disp( struct x86_reg reg, - int disp ) + GLint disp ) { assert(reg.file == file_REG32); @@ -191,7 +191,7 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg ) return x86_make_reg( reg.file, reg.idx ); } -unsigned char *x86_get_label( struct x86_function *p ) +GLubyte *x86_get_label( struct x86_function *p ) { return p->csr; } @@ -205,13 +205,13 @@ unsigned char *x86_get_label( struct x86_function *p ) void x86_jcc( struct x86_function *p, enum x86_cc cc, - unsigned char *label ) + GLubyte *label ) { - int offset = label - (x86_get_label(p) + 2); + GLint offset = label - (x86_get_label(p) + 2); if (offset <= 127 && offset >= -128) { emit_1ub(p, 0x70 + cc); - emit_1b(p, (char) offset); + emit_1b(p, (GLbyte) offset); } else { offset = label - (x86_get_label(p) + 6); @@ -222,7 +222,7 @@ void x86_jcc( struct x86_function *p, /* Always use a 32bit offset for forward jumps: */ -unsigned char *x86_jcc_forward( struct x86_function *p, +GLubyte *x86_jcc_forward( struct x86_function *p, enum x86_cc cc ) { emit_2ub(p, 0x0f, 0x80 + cc); @@ -230,14 +230,14 @@ unsigned char *x86_jcc_forward( struct x86_function *p, return x86_get_label(p); } -unsigned char *x86_jmp_forward( struct x86_function *p) +GLubyte *x86_jmp_forward( struct x86_function *p) { emit_1ub(p, 0xe9); emit_1i(p, 0); return x86_get_label(p); } -unsigned char *x86_call_forward( struct x86_function *p) +GLubyte *x86_call_forward( struct x86_function *p) { emit_1ub(p, 0xe8); emit_1i(p, 0); @@ -247,12 +247,12 @@ unsigned char *x86_call_forward( struct x86_function *p) /* Fixup offset from forward jump: */ void x86_fixup_fwd_jump( struct x86_function *p, - unsigned char *fixup ) + GLubyte *fixup ) { *(int *)(fixup - 4) = x86_get_label(p) - fixup; } -void x86_jmp( struct x86_function *p, unsigned char *label) +void x86_jmp( struct x86_function *p, GLubyte *label) { emit_1ub(p, 0xe9); emit_1i(p, label - x86_get_label(p) - 4); @@ -268,7 +268,7 @@ void x86_call( struct x86_function *p, void (*label)()) * Temporary. As I need immediate operands, and dont want to mess with the codegen, * I load the immediate into general purpose register and use it. */ -void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm ) { assert(dst.mod == mod_REG); emit_1ub(p, 0xb8 + dst.idx); @@ -595,7 +595,7 @@ void sse_cvtps2pi( struct x86_function *p, void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - unsigned char shuf) + GLubyte shuf) { emit_2ub(p, X86_TWOB, 0xC6); emit_modrm(p, dest, arg0); @@ -605,7 +605,7 @@ void sse_shufps( struct x86_function *p, void sse_cmpps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - unsigned char cc) + GLubyte cc) { emit_2ub(p, X86_TWOB, 0xC2); emit_modrm(p, dest, arg0); @@ -630,7 +630,7 @@ void sse_pmovmskb( struct x86_function *p, void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - unsigned char shuf) + GLubyte shuf) { emit_3ub(p, 0x66, X86_TWOB, 0x70); emit_modrm(p, dest, arg0); @@ -772,11 +772,11 @@ void x87_fclex( struct x86_function *p ) static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg, - unsigned char dst0ub0, - unsigned char dst0ub1, - unsigned char arg0ub0, - unsigned char arg0ub1, - unsigned char argmem_noreg) + GLubyte dst0ub0, + GLubyte dst0ub1, + GLubyte arg0ub0, + GLubyte arg0ub1, + GLubyte argmem_noreg) { assert(dst.file == file_x87); @@ -1116,7 +1116,7 @@ void mmx_movq( struct x86_function *p, * account any push/pop activity: */ struct x86_reg x86_fn_arg( struct x86_function *p, - unsigned arg ) + GLuint arg ) { return x86_make_disp(x86_make_reg(file_REG32, reg_SP), p->stack_offset + arg * 4); /* ??? */ @@ -1128,7 +1128,7 @@ void x86_init_func( struct x86_function *p ) x86_init_func_size(p, 1024); } -void x86_init_func_size( struct x86_function *p, unsigned code_size ) +void x86_init_func_size( struct x86_function *p, GLuint code_size ) { p->store = _mesa_exec_malloc(code_size); p->csr = p->store; diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h index 63b9a36392..c1ca06088b 100644 --- a/src/mesa/x86/rtasm/x86sse.h +++ b/src/mesa/x86/rtasm/x86sse.h @@ -4,22 +4,24 @@ #if defined(__i386__) || defined(__386__) +#include "glheader.h" + /* It is up to the caller to ensure that instructions issued are * suitable for the host cpu. There are no checks made in this module * for mmx/sse/sse2 support on the cpu. */ struct x86_reg { - unsigned file:3; - unsigned idx:3; - unsigned mod:2; /* mod_REG if this is just a register */ - int disp:24; /* only +/- 23bits of offset - should be enough... */ + GLuint file:3; + GLuint idx:3; + GLuint mod:2; /* mod_REG if this is just a register */ + GLint disp:24; /* only +/- 23bits of offset - should be enough... */ }; struct x86_function { - unsigned char *store; - unsigned char *csr; - unsigned stack_offset; - int need_emms; + GLubyte *store; + GLubyte *csr; + GLuint stack_offset; + GLint need_emms; const char *fn; }; @@ -79,7 +81,7 @@ enum sse_cc { void x86_init_func( struct x86_function *p ); -void x86_init_func_size( struct x86_function *p, unsigned code_size ); +void x86_init_func_size( struct x86_function *p, GLuint code_size ); void x86_release_func( struct x86_function *p ); void (*x86_get_func( struct x86_function *p ))( void ); @@ -91,7 +93,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file, enum x86_reg_name idx ); struct x86_reg x86_make_disp( struct x86_reg reg, - int disp ); + GLint disp ); struct x86_reg x86_deref( struct x86_reg reg ); @@ -100,23 +102,23 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg ); /* Labels, jumps and fixup: */ -unsigned char *x86_get_label( struct x86_function *p ); +GLubyte *x86_get_label( struct x86_function *p ); void x86_jcc( struct x86_function *p, enum x86_cc cc, - unsigned char *label ); + GLubyte *label ); -unsigned char *x86_jcc_forward( struct x86_function *p, +GLubyte *x86_jcc_forward( struct x86_function *p, enum x86_cc cc ); -unsigned char *x86_jmp_forward( struct x86_function *p); +GLubyte *x86_jmp_forward( struct x86_function *p); -unsigned char *x86_call_forward( struct x86_function *p); +GLubyte *x86_call_forward( struct x86_function *p); void x86_fixup_fwd_jump( struct x86_function *p, - unsigned char *fixup ); + GLubyte *fixup ); -void x86_jmp( struct x86_function *p, unsigned char *label ); +void x86_jmp( struct x86_function *p, GLubyte *label ); void x86_call( struct x86_function *p, void (*label)() ); @@ -124,7 +126,7 @@ void x86_call( struct x86_function *p, void (*label)() ); * Temporary. As I need immediate operands, and dont want to mess with the codegen, * I load the immediate into general purpose register and use it. */ -void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ); +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm ); /* Macro for sse_shufps() and sse2_pshufd(): @@ -145,8 +147,7 @@ void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - unsigned char shuf ); +void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf ); void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -156,8 +157,7 @@ void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg sr void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, - unsigned char cc ); +void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, GLubyte cc ); void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -175,8 +175,7 @@ void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - unsigned char shuf ); +void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf ); void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -248,7 +247,7 @@ void x87_fucom( struct x86_function *p, struct x86_reg arg ); * account any push/pop activity. Note - doesn't track explict * manipulation of ESP by other instructions. */ -struct x86_reg x86_fn_arg( struct x86_function *p, unsigned arg ); +struct x86_reg x86_fn_arg( struct x86_function *p, GLuint arg ); #endif #endif -- cgit v1.2.3 From f78193f444dad2a1aee219f538bc0da3f040c2e1 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 28 Sep 2007 04:33:55 -0400 Subject: Remove gl dependency from the x86 assembler code --- src/mesa/x86/rtasm/x86sse.c | 68 ++++++++++++++++++++++----------------------- src/mesa/x86/rtasm/x86sse.h | 49 ++++++++++++++++---------------- 2 files changed, 59 insertions(+), 58 deletions(-) (limited to 'src') diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index a6cfa40f91..f9a9352e46 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -6,22 +6,22 @@ #define DISASSEM 0 #define X86_TWOB 0x0f -static GLubyte *cptr( void (*label)() ) +static unsigned char *cptr( void (*label)() ) { - return (char *)(unsigned long)label; + return (unsigned char *)(unsigned long)label; } /* Emit bytes to the instruction stream: */ -static void emit_1b( struct x86_function *p, GLbyte b0 ) +static void emit_1b( struct x86_function *p, char b0 ) { - *(GLbyte *)(p->csr++) = b0; + *(char *)(p->csr++) = b0; } -static void emit_1i( struct x86_function *p, GLint i0 ) +static void emit_1i( struct x86_function *p, int i0 ) { - *(GLint *)(p->csr) = i0; + *(int *)(p->csr) = i0; p->csr += 4; } @@ -35,20 +35,20 @@ static void disassem( struct x86_function *p, const char *fn ) #endif } -static void emit_1ub_fn( struct x86_function *p, GLubyte b0, const char *fn ) +static void emit_1ub_fn( struct x86_function *p, unsigned char b0, const char *fn ) { disassem(p, fn); *(p->csr++) = b0; } -static void emit_2ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, const char *fn ) +static void emit_2ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, const char *fn ) { disassem(p, fn); *(p->csr++) = b0; *(p->csr++) = b1; } -static void emit_3ub_fn( struct x86_function *p, GLubyte b0, GLubyte b1, GLubyte b2, const char *fn ) +static void emit_3ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2, const char *fn ) { disassem(p, fn); *(p->csr++) = b0; @@ -69,7 +69,7 @@ static void emit_modrm( struct x86_function *p, struct x86_reg reg, struct x86_reg regmem ) { - GLubyte val = 0; + unsigned char val = 0; assert(reg.mod == mod_REG); @@ -104,7 +104,7 @@ static void emit_modrm( struct x86_function *p, static void emit_modrm_noreg( struct x86_function *p, - GLuint op, + unsigned op, struct x86_reg regmem ) { struct x86_reg dummy = x86_make_reg(file_REG32, op); @@ -117,8 +117,8 @@ static void emit_modrm_noreg( struct x86_function *p, * the arguments presented. */ static void emit_op_modrm( struct x86_function *p, - GLubyte op_dst_is_reg, - GLubyte op_dst_is_mem, + unsigned char op_dst_is_reg, + unsigned char op_dst_is_mem, struct x86_reg dst, struct x86_reg src ) { @@ -162,7 +162,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file, } struct x86_reg x86_make_disp( struct x86_reg reg, - GLint disp ) + int disp ) { assert(reg.file == file_REG32); @@ -191,7 +191,7 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg ) return x86_make_reg( reg.file, reg.idx ); } -GLubyte *x86_get_label( struct x86_function *p ) +unsigned char *x86_get_label( struct x86_function *p ) { return p->csr; } @@ -205,13 +205,13 @@ GLubyte *x86_get_label( struct x86_function *p ) void x86_jcc( struct x86_function *p, enum x86_cc cc, - GLubyte *label ) + unsigned char *label ) { - GLint offset = label - (x86_get_label(p) + 2); + int offset = label - (x86_get_label(p) + 2); if (offset <= 127 && offset >= -128) { emit_1ub(p, 0x70 + cc); - emit_1b(p, (GLbyte) offset); + emit_1b(p, (char) offset); } else { offset = label - (x86_get_label(p) + 6); @@ -222,7 +222,7 @@ void x86_jcc( struct x86_function *p, /* Always use a 32bit offset for forward jumps: */ -GLubyte *x86_jcc_forward( struct x86_function *p, +unsigned char *x86_jcc_forward( struct x86_function *p, enum x86_cc cc ) { emit_2ub(p, 0x0f, 0x80 + cc); @@ -230,14 +230,14 @@ GLubyte *x86_jcc_forward( struct x86_function *p, return x86_get_label(p); } -GLubyte *x86_jmp_forward( struct x86_function *p) +unsigned char *x86_jmp_forward( struct x86_function *p) { emit_1ub(p, 0xe9); emit_1i(p, 0); return x86_get_label(p); } -GLubyte *x86_call_forward( struct x86_function *p) +unsigned char *x86_call_forward( struct x86_function *p) { emit_1ub(p, 0xe8); emit_1i(p, 0); @@ -247,12 +247,12 @@ GLubyte *x86_call_forward( struct x86_function *p) /* Fixup offset from forward jump: */ void x86_fixup_fwd_jump( struct x86_function *p, - GLubyte *fixup ) + unsigned char *fixup ) { *(int *)(fixup - 4) = x86_get_label(p) - fixup; } -void x86_jmp( struct x86_function *p, GLubyte *label) +void x86_jmp( struct x86_function *p, unsigned char *label) { emit_1ub(p, 0xe9); emit_1i(p, label - x86_get_label(p) - 4); @@ -268,7 +268,7 @@ void x86_call( struct x86_function *p, void (*label)()) * Temporary. As I need immediate operands, and dont want to mess with the codegen, * I load the immediate into general purpose register and use it. */ -void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm ) +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) { assert(dst.mod == mod_REG); emit_1ub(p, 0xb8 + dst.idx); @@ -595,7 +595,7 @@ void sse_cvtps2pi( struct x86_function *p, void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - GLubyte shuf) + unsigned char shuf) { emit_2ub(p, X86_TWOB, 0xC6); emit_modrm(p, dest, arg0); @@ -605,7 +605,7 @@ void sse_shufps( struct x86_function *p, void sse_cmpps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - GLubyte cc) + unsigned char cc) { emit_2ub(p, X86_TWOB, 0xC2); emit_modrm(p, dest, arg0); @@ -630,7 +630,7 @@ void sse_pmovmskb( struct x86_function *p, void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - GLubyte shuf) + unsigned char shuf) { emit_3ub(p, 0x66, X86_TWOB, 0x70); emit_modrm(p, dest, arg0); @@ -772,11 +772,11 @@ void x87_fclex( struct x86_function *p ) static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg, - GLubyte dst0ub0, - GLubyte dst0ub1, - GLubyte arg0ub0, - GLubyte arg0ub1, - GLubyte argmem_noreg) + unsigned char dst0ub0, + unsigned char dst0ub1, + unsigned char arg0ub0, + unsigned char arg0ub1, + unsigned char argmem_noreg) { assert(dst.file == file_x87); @@ -1116,7 +1116,7 @@ void mmx_movq( struct x86_function *p, * account any push/pop activity: */ struct x86_reg x86_fn_arg( struct x86_function *p, - GLuint arg ) + unsigned arg ) { return x86_make_disp(x86_make_reg(file_REG32, reg_SP), p->stack_offset + arg * 4); /* ??? */ @@ -1128,7 +1128,7 @@ void x86_init_func( struct x86_function *p ) x86_init_func_size(p, 1024); } -void x86_init_func_size( struct x86_function *p, GLuint code_size ) +void x86_init_func_size( struct x86_function *p, unsigned code_size ) { p->store = _mesa_exec_malloc(code_size); p->csr = p->store; diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h index c1ca06088b..63b9a36392 100644 --- a/src/mesa/x86/rtasm/x86sse.h +++ b/src/mesa/x86/rtasm/x86sse.h @@ -4,24 +4,22 @@ #if defined(__i386__) || defined(__386__) -#include "glheader.h" - /* It is up to the caller to ensure that instructions issued are * suitable for the host cpu. There are no checks made in this module * for mmx/sse/sse2 support on the cpu. */ struct x86_reg { - GLuint file:3; - GLuint idx:3; - GLuint mod:2; /* mod_REG if this is just a register */ - GLint disp:24; /* only +/- 23bits of offset - should be enough... */ + unsigned file:3; + unsigned idx:3; + unsigned mod:2; /* mod_REG if this is just a register */ + int disp:24; /* only +/- 23bits of offset - should be enough... */ }; struct x86_function { - GLubyte *store; - GLubyte *csr; - GLuint stack_offset; - GLint need_emms; + unsigned char *store; + unsigned char *csr; + unsigned stack_offset; + int need_emms; const char *fn; }; @@ -81,7 +79,7 @@ enum sse_cc { void x86_init_func( struct x86_function *p ); -void x86_init_func_size( struct x86_function *p, GLuint code_size ); +void x86_init_func_size( struct x86_function *p, unsigned code_size ); void x86_release_func( struct x86_function *p ); void (*x86_get_func( struct x86_function *p ))( void ); @@ -93,7 +91,7 @@ struct x86_reg x86_make_reg( enum x86_reg_file file, enum x86_reg_name idx ); struct x86_reg x86_make_disp( struct x86_reg reg, - GLint disp ); + int disp ); struct x86_reg x86_deref( struct x86_reg reg ); @@ -102,23 +100,23 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg ); /* Labels, jumps and fixup: */ -GLubyte *x86_get_label( struct x86_function *p ); +unsigned char *x86_get_label( struct x86_function *p ); void x86_jcc( struct x86_function *p, enum x86_cc cc, - GLubyte *label ); + unsigned char *label ); -GLubyte *x86_jcc_forward( struct x86_function *p, +unsigned char *x86_jcc_forward( struct x86_function *p, enum x86_cc cc ); -GLubyte *x86_jmp_forward( struct x86_function *p); +unsigned char *x86_jmp_forward( struct x86_function *p); -GLubyte *x86_call_forward( struct x86_function *p); +unsigned char *x86_call_forward( struct x86_function *p); void x86_fixup_fwd_jump( struct x86_function *p, - GLubyte *fixup ); + unsigned char *fixup ); -void x86_jmp( struct x86_function *p, GLubyte *label ); +void x86_jmp( struct x86_function *p, unsigned char *label ); void x86_call( struct x86_function *p, void (*label)() ); @@ -126,7 +124,7 @@ void x86_call( struct x86_function *p, void (*label)() ); * Temporary. As I need immediate operands, and dont want to mess with the codegen, * I load the immediate into general purpose register and use it. */ -void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, GLint imm ); +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ); /* Macro for sse_shufps() and sse2_pshufd(): @@ -147,7 +145,8 @@ void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf ); +void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, + unsigned char shuf ); void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -157,7 +156,8 @@ void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg sr void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, GLubyte cc ); +void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, + unsigned char cc ); void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -175,7 +175,8 @@ void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, GLubyte shuf ); +void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, + unsigned char shuf ); void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -247,7 +248,7 @@ void x87_fucom( struct x86_function *p, struct x86_reg arg ); * account any push/pop activity. Note - doesn't track explict * manipulation of ESP by other instructions. */ -struct x86_reg x86_fn_arg( struct x86_function *p, GLuint arg ); +struct x86_reg x86_fn_arg( struct x86_function *p, unsigned arg ); #endif #endif -- cgit v1.2.3 From a1a989f0be8dc34082b52bb3b3a6eacb36d9e75e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 28 Sep 2007 04:33:55 -0400 Subject: Redoing the way we handle vertex shaders for the draw module. --- src/mesa/pipe/draw/draw_context.c | 8 ----- src/mesa/pipe/draw/draw_context.h | 9 ++++-- src/mesa/pipe/draw/draw_private.h | 14 +++++++-- src/mesa/pipe/draw/draw_vertex_fetch.c | 2 +- src/mesa/pipe/draw/draw_vertex_shader.c | 50 ++++++++++++++++++++++++++++--- src/mesa/pipe/i915simple/i915_state.c | 44 +++++++++++++++++---------- src/mesa/pipe/softpipe/sp_context.c | 8 ++--- src/mesa/pipe/softpipe/sp_context.h | 8 +++-- src/mesa/pipe/softpipe/sp_state.h | 14 +++++---- src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- src/mesa/pipe/softpipe/sp_state_fs.c | 49 +++++++++++++++++++++++------- src/mesa/state_tracker/st_cb_program.c | 7 ----- src/mesa/state_tracker/st_draw.c | 2 +- src/mesa/state_tracker/st_program.c | 8 ----- src/mesa/state_tracker/st_program.h | 4 --- 15 files changed, 152 insertions(+), 77 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 66c66ff698..5efb173228 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -155,14 +155,6 @@ void draw_set_viewport_state( struct draw_context *draw, } -void -draw_set_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader) -{ - draw_flush( draw ); - draw->vertex_shader = *shader; -} - void draw_set_vertex_buffer(struct draw_context *draw, diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 0ccf5f6046..398e96d94d 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -92,9 +92,12 @@ void draw_set_rasterize_stage( struct draw_context *draw, struct draw_stage *stage ); -void -draw_set_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader); +void * draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader); +void draw_bind_vertex_shader(struct draw_context *draw, + void *vcso); +void draw_delete_vertex_shader(struct draw_context *draw, + void *vcso); void diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index ebef5347ab..2f52299f74 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -46,6 +46,7 @@ #include "draw_vertex.h" +#include "x86/rtasm/x86sse.h" /** * Basic vertex info. @@ -104,7 +105,7 @@ struct draw_stage void (*tri)( struct draw_stage *, struct prim_header * ); - + void (*end)( struct draw_stage * ); void (*reset_stipple_counter)( struct draw_stage * ); @@ -116,6 +117,15 @@ struct draw_stage #define VCACHE_OVERFLOW 4 #define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ +/** + * Private version of the compiled vertex_shader + */ +struct draw_vertex_shader { + const struct pipe_shader_state *state; +#if defined(__i386__) || defined(__386__) + struct x86_function sse2_program; +#endif +}; /** * Private context for the drawing module. @@ -145,7 +155,7 @@ struct draw_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - struct pipe_shader_state vertex_shader; + const struct draw_vertex_shader *vertex_shader; struct pipe_vertex_buffer feedback_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element feedback_element[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 2b839d641e..bc37e18c34 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -85,7 +85,7 @@ void draw_vertex_fetch( struct draw_context *draw, /*printf("fetch vertex %u: \n", j);*/ /* loop over vertex attributes (vertex shader inputs) */ - for (attr = 0; attr < draw->vertex_shader.num_inputs; attr++) { + for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { unsigned buf = draw->vertex_element[attr].vertex_buffer_index; const void *src diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index a2e1cdc472..f7ddbd1909 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -36,6 +36,8 @@ #include "draw_context.h" #include "draw_vertex.h" +#include "x86/rtasm/x86sse.h" + #include "pipe/tgsi/exec/tgsi_core.h" static INLINE unsigned @@ -70,6 +72,7 @@ typedef void (XSTDCALL *codegen_function) ( float (*constant)[4], struct tgsi_exec_vector *temporary ); + /** * Transform vertices with the current vertex program/shader * Up to four vertices can be shaded at a time. @@ -92,7 +95,7 @@ run_vertex_program(struct draw_context *draw, const float *trans = draw->viewport.translate; assert(count <= 4); - assert(draw->vertex_shader.output_semantic_name[0] + assert(draw->vertex_shader->state->output_semantic_name[0] == TGSI_SEMANTIC_POSITION); #ifdef DEBUG @@ -101,7 +104,7 @@ run_vertex_program(struct draw_context *draw, /* init machine state */ tgsi_exec_machine_init(&machine, - draw->vertex_shader.tokens, + draw->vertex_shader->state->tokens, PIPE_MAX_SAMPLERS, NULL /*samplers*/ ); @@ -114,8 +117,8 @@ run_vertex_program(struct draw_context *draw, draw_vertex_fetch( draw, &machine, elts, count ); /* run shader */ - if( draw->vertex_shader.executable != NULL ) { - codegen_function func = (codegen_function) draw->vertex_shader.executable; + if( draw->vertex_shader->state->executable != NULL ) { + codegen_function func = (codegen_function) draw->vertex_shader->state->executable; func( machine.Inputs, machine.Outputs, @@ -206,3 +209,42 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) draw->vs.queue_nr = 0; } + +void * +draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader) +{ + struct draw_vertex_shader *vs = calloc(1, sizeof(struct draw_vertex_shader)); + + vs->state = shader; +#if defined(__i386__) || defined(__386__) + x86_init_func(&vs->sse2_program); + + tgsi_emit_sse2(shader->tokens, &vs->sse2_program); + + ((struct pipe_shader_state*)(vs->state))->executable = + x86_get_func(&vs->sse2_program); +#endif + + return vs; +} + +void draw_bind_vertex_shader(struct draw_context *draw, + void *vcso) +{ + draw_flush(draw); + draw->vertex_shader = (struct draw_vertex_shader*)(vcso); +} + +void draw_delete_vertex_shader(struct draw_context *draw, + void *vcso) +{ + struct draw_vertex_shader *vs = (struct draw_vertex_shader*)(vcso); +#if defined(__i386__) || defined(__386__) + x86_release_func(&vs->sse2_program); +#endif + free(vcso); +} + + + diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 5c6d37a466..19ca5e575f 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -443,16 +443,13 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe, { } - -static void * -i915_create_shader_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) +static void * i915_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { return 0; } -static void i915_bind_fs_state( struct pipe_context *pipe, - void *fs ) +static void i915_bind_fs_state(struct pipe_context *pipe, void *fs) { struct i915_context *i915 = i915_context(pipe); @@ -461,20 +458,35 @@ static void i915_bind_fs_state( struct pipe_context *pipe, i915->dirty |= I915_NEW_FS; } +static void i915_delete_fs_state(struct pipe_context *pipe, void *shader) +{ + /*do nothing*/ +} -static void i915_bind_vs_state(struct pipe_context *pipe, - void *vs) +static void * +i915_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { struct i915_context *i915 = i915_context(pipe); /* just pass-through to draw module */ - draw_set_vertex_shader(i915->draw, (const struct pipe_shader_state *)vs); + return draw_create_vertex_shader(i915->draw, templ); } -static void i915_delete_shader_state(struct pipe_context *pipe, - void *shader) +static void i915_bind_vs_state(struct pipe_context *pipe, void *vs) { - /*do nothing*/ + struct i915_context *i915 = i915_context(pipe); + + /* just pass-through to draw module */ + draw_bind_vertex_shader(i915->draw, vs); +} + +static void i915_delete_vs_state(struct pipe_context *pipe, void *shader) +{ + struct i915_context *i915 = i915_context(pipe); + + /* just pass-through to draw module */ + draw_delete_vertex_shader(i915->draw, shader); } static void i915_set_constant_buffer(struct pipe_context *pipe, @@ -707,12 +719,12 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state; - i915->pipe.create_fs_state = i915_create_shader_state; + i915->pipe.create_fs_state = i915_create_fs_state; i915->pipe.bind_fs_state = i915_bind_fs_state; - i915->pipe.delete_fs_state = i915_delete_shader_state; - i915->pipe.create_vs_state = i915_create_shader_state; + i915->pipe.delete_fs_state = i915_delete_fs_state; + i915->pipe.create_vs_state = i915_create_vs_state; i915->pipe.bind_vs_state = i915_bind_vs_state; - i915->pipe.delete_vs_state = i915_delete_shader_state; + i915->pipe.delete_vs_state = i915_delete_vs_state; i915->pipe.set_blend_color = i915_set_blend_color; i915->pipe.set_clip_state = i915_set_clip_state; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index e415966dae..695bf1a9e0 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -275,12 +275,12 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; - softpipe->pipe.create_fs_state = softpipe_create_shader_state; + softpipe->pipe.create_fs_state = softpipe_create_fs_state; softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; - softpipe->pipe.delete_fs_state = softpipe_delete_shader_state; - softpipe->pipe.create_vs_state = softpipe_create_shader_state; + softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; + softpipe->pipe.create_vs_state = softpipe_create_vs_state; softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; - softpipe->pipe.delete_vs_state = softpipe_delete_shader_state; + softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; softpipe->pipe.set_blend_color = softpipe_set_blend_color; softpipe->pipe.set_clip_state = softpipe_set_clip_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index c0a681f3d6..ccf29b5683 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -62,8 +62,12 @@ struct draw_stage; #define SP_NEW_VS 0x2000 #define SP_NEW_CONSTANTS 0x4000 +struct sp_vertex_shader_state { + const struct pipe_shader_state *state; + void *draw_data; +}; -struct softpipe_context { +struct softpipe_context { struct pipe_context pipe; /**< base class */ struct softpipe_winsys *winsys; /**< window system interface */ @@ -76,7 +80,7 @@ struct softpipe_context { const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct pipe_shader_state *fs; - const struct pipe_shader_state *vs; + const struct sp_vertex_shader_state *vs; struct pipe_blend_color blend_color; struct pipe_clear_color_state clear_color; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index f0e1461d25..f9061e86e5 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -87,12 +87,14 @@ void softpipe_set_constant_buffer(struct pipe_context *, void softpipe_set_feedback_state( struct pipe_context *, const struct pipe_feedback_state * ); -void * -softpipe_create_shader_state( struct pipe_context *, - const struct pipe_shader_state * ); -void softpipe_bind_fs_state( struct pipe_context *, void * ); -void softpipe_bind_vs_state( struct pipe_context *, void * ); -void softpipe_delete_shader_state( struct pipe_context *, void * ); +void *softpipe_create_fs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_fs_state(struct pipe_context *, void *); +void softpipe_delete_fs_state(struct pipe_context *, void *); +void *softpipe_create_vs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_vs_state(struct pipe_context *, void *); +void softpipe_delete_vs_state(struct pipe_context *, void *); void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 6c6e798069..b6145df8e2 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -43,7 +43,7 @@ */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - const struct pipe_shader_state *vs = softpipe->vs; + const struct pipe_shader_state *vs = softpipe->vs->state; const struct pipe_shader_state *fs = softpipe->fs; const interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 8306a95f44..f1bec2c73a 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -33,10 +33,13 @@ #include "pipe/draw/draw_context.h" -void * softpipe_create_shader_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) +void * softpipe_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { - /* we just want the pipe_shader_state template in the bind calls */ + /* Decide whether we'll be codegenerating this shader and if so do + * that now. + */ + return 0; } @@ -49,25 +52,51 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) softpipe->dirty |= SP_NEW_FS; } +void softpipe_delete_fs_state(struct pipe_context *pipe, + void *shader) +{ +} + + +void * softpipe_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + struct sp_vertex_shader_state *state = + malloc(sizeof(struct sp_vertex_shader_state)); + + state->state = templ; + state->draw_data = draw_create_vertex_shader(softpipe->draw, + state->state); + + return state; +} void softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->vs = (struct pipe_shader_state *)vs; + softpipe->vs = (const struct sp_vertex_shader_state *)vs; - softpipe->dirty |= SP_NEW_VS; + draw_bind_vertex_shader(softpipe->draw, softpipe->vs->draw_data); - draw_set_vertex_shader(softpipe->draw, (struct pipe_shader_state *)vs); + softpipe->dirty |= SP_NEW_VS; } - -void softpipe_delete_shader_state( struct pipe_context *pipe, - void *shader ) +void softpipe_delete_vs_state(struct pipe_context *pipe, + void *vs) { - /* do nothing */ + struct softpipe_context *softpipe = softpipe_context(pipe); + + struct sp_vertex_shader_state *state = + (struct sp_vertex_shader_state *)vs; + + draw_delete_vertex_shader(softpipe->draw, state->draw_data); + free(state); } + + void softpipe_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, const struct pipe_constant_buffer *buf) diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 25dbc2c244..3a7ce9405e 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -86,10 +86,6 @@ static struct gl_program *st_new_program( GLcontext *ctx, prog->serialNo = 1; -#if defined(__i386__) || defined(__386__) - x86_init_func( &prog->sse2_program ); -#endif - return _mesa_init_vertex_program( ctx, &prog->Base, target, @@ -129,9 +125,6 @@ static void st_delete_program( GLcontext *ctx, case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; -#if defined(__i386__) || defined(__386__) - x86_release_func( &stvp->sse2_program ); -#endif st_remove_vertex_program(st, stvp); } break; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index ce5bf0c8a9..7c836ba2c1 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -405,7 +405,7 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); draw_set_rasterizer_state(draw, &st->state.rasterizer->state); - draw_set_vertex_shader(draw, &st->state.vs->state); + draw_bind_vertex_shader(draw, st->state.vs->data); /* XXX need to set vertex info too */ diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 44cc33b09f..116b59a067 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -253,14 +253,6 @@ st_translate_vertex_program(struct st_context *st, if (TGSI_DEBUG) tgsi_dump( tokensOut, 0 ); -#if defined(__i386__) || defined(__386__) - if (stvp->sse2_program.csr == stvp->sse2_program.store) - tgsi_emit_sse2( tokensOut, &stvp->sse2_program ); - - if (!cso->state.executable) - ((struct cso_vertex_shader*)cso)->state.executable = (void *) x86_get_func( &stvp->sse2_program ); -#endif - return cso; } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 6e106c16b0..2b79201313 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -79,10 +79,6 @@ struct st_vertex_program /** The program in TGSI format */ struct tgsi_token tokens[ST_FP_MAX_TOKENS]; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif - /** Pointer to the corresponding cached shader */ const struct cso_vertex_shader *vs; -- cgit v1.2.3 From 8731e391f5a8bb45adb53242b0a473707c0c024c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 28 Sep 2007 07:33:54 -0400 Subject: Use sse only if GALLIUM_SSE is defined --- src/mesa/pipe/draw/draw_context.c | 11 +++++++++++ src/mesa/pipe/draw/draw_context.h | 1 + src/mesa/pipe/draw/draw_private.h | 2 ++ src/mesa/pipe/draw/draw_vertex_shader.c | 9 +++++---- src/mesa/state_tracker/st_program.c | 11 +++++++---- 5 files changed, 26 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 5efb173228..3fb667ab1a 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -41,6 +41,12 @@ struct draw_context *draw_create( void ) { struct draw_context *draw = CALLOC_STRUCT( draw_context ); +#if defined(__i386__) || defined(__386__) + draw->use_sse = getenv("GALLIUM_SSE") != NULL; +#else + draw->use_sse = false; +#endif + /* create pipeline stages */ draw->pipeline.unfilled = draw_unfilled_stage( draw ); draw->pipeline.twoside = draw_twoside_stage( draw ); @@ -244,4 +250,9 @@ void draw_free_tmps( struct draw_stage *stage ) } } +boolean draw_use_sse(struct draw_context *draw) +{ + return draw->use_sse; +} + diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 398e96d94d..8e2232244c 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -99,6 +99,7 @@ void draw_bind_vertex_shader(struct draw_context *draw, void draw_delete_vertex_shader(struct draw_context *draw, void *vcso); +boolean draw_use_sse(struct draw_context *draw); void draw_set_vertex_buffer(struct draw_context *draw, diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 2f52299f74..12a970a671 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -220,6 +220,8 @@ struct draw_context struct prim_header queue[PRIM_QUEUE_LENGTH]; unsigned queue_nr; } pq; + + int use_sse : 1; }; diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index f7ddbd1909..3518bd52a3 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -220,10 +220,11 @@ draw_create_vertex_shader(struct draw_context *draw, #if defined(__i386__) || defined(__386__) x86_init_func(&vs->sse2_program); - tgsi_emit_sse2(shader->tokens, &vs->sse2_program); - - ((struct pipe_shader_state*)(vs->state))->executable = - x86_get_func(&vs->sse2_program); + if (draw->use_sse) { + tgsi_emit_sse2(shader->tokens, &vs->sse2_program); + ((struct pipe_shader_state*)(vs->state))->executable = + x86_get_func(&vs->sse2_program); + } #endif return vs; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 116b59a067..a41b953a21 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -36,6 +36,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/draw/draw_context.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/exec/tgsi_core.h" @@ -391,11 +392,13 @@ st_translate_fragment_program(struct st_context *st, tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); #if defined(__i386__) || defined(__386__) - if (stfp->sse2_program.csr == stfp->sse2_program.store) - tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); + if (draw_use_sse(st->draw)) { + if (stfp->sse2_program.csr == stfp->sse2_program.store) + tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); - if (!cso->state.executable) - ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); + if (!cso->state.executable) + ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); + } #endif return cso; -- cgit v1.2.3 From f14ece2d2c9add5ebf21171746f34ce60ff0df3b Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Sep 2007 13:49:50 -0600 Subject: Use texture->first_level, not 0, when not mipmapping. Fixes crash when GL_BASE_LEVEL!=0. Also, remove old assertion. --- src/mesa/pipe/softpipe/sp_tex_sample.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 40d0cf40ae..0c990294f2 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -479,10 +479,8 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, { if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { /* no mipmap selection needed */ - assert(sampler->state->min_img_filter == - sampler->state->mag_img_filter); *imgFilter = sampler->state->mag_img_filter; - *level0 = *level1 = 0; + *level0 = *level1 = sampler->texture->first_level; } else { float lambda; @@ -497,7 +495,7 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, if (lambda < 0.0) { /* XXX threshold depends on the filter */ /* magnifying */ *imgFilter = sampler->state->mag_img_filter; - *level0 = *level1 = 0; + *level0 = *level1 = sampler->texture->first_level; } else { /* minifying */ -- cgit v1.2.3 From 636480cc9c7836daf879cb45644900922cf31f47 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Sep 2007 15:39:09 -0600 Subject: Instead of linked program pairs, keep a list of vertex programs translated for each fragment program. --- src/mesa/state_tracker/st_atom_shader.c | 337 ++++++++++-------------------- src/mesa/state_tracker/st_cb_clear.c | 4 +- src/mesa/state_tracker/st_cb_drawpixels.c | 4 +- src/mesa/state_tracker/st_cb_program.c | 12 ++ src/mesa/state_tracker/st_program.h | 30 ++- 5 files changed, 149 insertions(+), 238 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 9bba78f7a1..d7a89d1e95 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -51,105 +51,56 @@ #include "st_atom_shader.h" - /** - * Structure to describe a (vertex program, fragment program) pair - * which is linked together (used together to render something). This - * linkage basically servers the same purpose as the OpenGL Shading - * Language linker, but also applies to ARB programs and Mesa's - * fixed-function-generated programs. - * - * More background: - * - * The translation from Mesa programs to TGSI programs depends on the - * linkage between the vertex program and the fragment program. This is - * because we tightly pack the inputs and outputs of shaders into - * consecutive "slots". - * - * Suppose an app uses one vertex program "VP" (outputting pos, color and tex0) - * and two fragment programs: - * FP1: uses tex0 input only (input slot 0) - * FP2: uses color input only (input slot 0) - * - * When VP is used with FP1 we want VP.output[2] to match FP1.input[0], but - * when VP is used with FP2 we want VP.output[1] to match FP1.input[0]. - * - * We don't want to re-translate the vertex and/or fragment programs - * each time the VP/FP bindings/linkings change. The solution is this - * structure which stores the translated TGSI shaders on a per-linkage - * basis. - * + * This represents a vertex program, especially translated to match + * the inputs of a particular fragment shader. */ -struct linked_program_pair +struct translated_vertex_program { - struct st_vertex_program *vprog; /**< never changes */ - struct st_fragment_program *fprog; /**< never changes */ + /** The fragment shader "signature" this vertex shader is meant for: */ + GLbitfield frag_inputs; - struct tgsi_token vs_tokens[ST_FP_MAX_TOKENS]; - struct tgsi_token fs_tokens[ST_FP_MAX_TOKENS]; + /** Compared against master vertex program's serialNo: */ + GLuint serialNo; - const struct cso_vertex_shader *vs; - const struct cso_fragment_shader *fs; - - GLuint vertSerialNo, fragSerialNo; + /** Maps VERT_RESULT_x to slot */ + GLuint output_to_slot[VERT_RESULT_MAX]; - /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ - GLuint vp_input_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; - /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ - GLuint vp_index_to_input[MAX_VERTEX_PROGRAM_ATTRIBS]; + /** The program in TGSI format */ + struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; - GLuint vp_result_to_slot[VERT_RESULT_MAX]; + /** Pointer to the translated, cached vertex shader */ + const struct cso_vertex_shader *vs; - struct linked_program_pair *next; + struct translated_vertex_program *next; /**< next in linked list */ }; -/** XXX temporary - use some kind of hash table instead */ -static struct linked_program_pair *Pairs = NULL; - - -static void -find_and_remove(struct gl_program *prog) -{ - struct linked_program_pair *pair, *prev = NULL, *next; - for (pair = Pairs; pair; pair = next) { - next = pair->next; - if (pair->vprog == (struct st_vertex_program *) prog || - pair->fprog == (struct st_fragment_program *) prog) { - /* unlink */ - if (prev) - prev->next = next; - else - Pairs = next; - /* delete pair->vs */ - /* delete pair->fs */ - free(pair); - } - else { - prev = pair; - } - } -} - /** - * Delete any known program pairs that use the given vertex program. + * Free data hanging off the st vert prog. */ void st_remove_vertex_program(struct st_context *st, struct st_vertex_program *stvp) { - find_and_remove(&stvp->Base.Base); + /* no-op, for now? */ } /** - * Delete any known program pairs that use the given fragment program. + * Free data hanging off the st frag prog. */ void st_remove_fragment_program(struct st_context *st, struct st_fragment_program *stfp) { - find_and_remove(&stfp->Base.Base); + struct translated_vertex_program *xvp, *next; + + for (xvp = stfp->vertex_programs; xvp; xvp = next) { + next = xvp->next; + /* XXX free xvp->vs */ + free(xvp); + } } @@ -187,182 +138,122 @@ vp_out_to_fp_in(GLuint vertResult) /** - * Examine the outputs written by a vertex program and the inputs read - * by a fragment program to determine which match up and where they - * should be mapped into the generic shader output/input slots. - * \param vert_output_map returns the vertex output register mapping - * \param frag_input_map returns the fragment input register mapping + * Find a translated vertex program that corresponds to stvp and + * has outputs matched to stfp's inputs. + * This performs vertex and fragment translation (to TGSI) when needed. */ -static GLuint -link_outputs_to_inputs(GLbitfield outputsWritten, - GLbitfield inputsRead, - GLuint vert_output_map[], - GLuint frag_input_map[]) +static struct translated_vertex_program * +find_translated_vp(struct st_context *st, + struct st_vertex_program *stvp, + struct st_fragment_program *stfp) { static const GLuint UNUSED = ~0; - GLint vert_slot_to_attr[50], frag_slot_to_attr[50]; - GLuint outAttr, inAttr; - GLuint numIn = 0, dummySlot; - - for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) { - if (inputsRead & (1 << inAttr)) { - frag_input_map[inAttr] = numIn; - frag_slot_to_attr[numIn] = inAttr; - numIn++; - } - else { - frag_input_map[inAttr] = UNUSED; - } - } + struct translated_vertex_program *xvp; + const GLbitfield fragInputsRead + = stfp->Base.Base.InputsRead | FRAG_BIT_WPOS; - for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - if (outputsWritten & (1 << outAttr)) { - /* see if the frag prog wants this vert output */ - GLint fpIn = vp_out_to_fp_in(outAttr); + /* + * Translate fragment program if needed. + */ + if (!stfp->fs) { + GLuint inAttr, numIn = 0; - if (fpIn >= 0) { - GLuint frag_slot = frag_input_map[fpIn]; - vert_output_map[outAttr] = frag_slot; - vert_slot_to_attr[frag_slot] = outAttr; + for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) { + if (fragInputsRead & (1 << inAttr)) { + stfp->input_to_slot[inAttr] = numIn; + numIn++; } else { - vert_output_map[outAttr] = UNUSED; + stfp->input_to_slot[inAttr] = UNUSED; } } - else { - vert_output_map[outAttr] = UNUSED; - } - } - /* - * We'll map all unused vertex program outputs to this slot. - * We'll also map all undefined fragment program inputs to this slot. - */ - dummySlot = numIn; + stfp->num_input_slots = numIn; - /* Map vert program outputs that aren't used to the dummy slot */ - for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - if (outputsWritten & (1 << outAttr)) { - if (vert_output_map[outAttr] == UNUSED) - vert_output_map[outAttr] = dummySlot; - } + (void) st_translate_fragment_program(st, stfp, + stfp->input_to_slot, + stfp->tokens, + ST_MAX_SHADER_TOKENS); + assert(stfp->fs); } - /* Map frag program inputs that aren't defined to the dummy slot */ - for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) { - if (inputsRead & (1 << inAttr)) { - if (frag_input_map[inAttr] == UNUSED) - frag_input_map[inAttr] = dummySlot; - } - } -#if 0 - printf("vOut W slot\n"); - for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - printf("%4d %c %4d\n", outAttr, - " *"[(outputsWritten >> outAttr) & 1], - vert_output_map[outAttr]); - } - printf("vIn R slot\n"); - for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) { - printf("%3d %c %4d\n", inAttr, - " *"[(inputsRead >> inAttr) & 1], - frag_input_map[inAttr]); + /* See if we've got a translated vertex program whose outputs match + * the fragment program's inputs. + * XXX This could be a hash lookup, using InputsRead as the key. + */ + for (xvp = stfp->vertex_programs; xvp; xvp = xvp->next) { + if (xvp->frag_inputs == stfp->Base.Base.InputsRead) { + break; + } } -#endif - return numIn; -} + /* No? Allocate translated vp object now */ + if (!xvp) { + xvp = CALLOC_STRUCT(translated_vertex_program); + xvp->frag_inputs = fragInputsRead; + xvp->next = stfp->vertex_programs; + stfp->vertex_programs = xvp; + } -static struct linked_program_pair * -lookup_program_pair(struct st_context *st, - struct st_vertex_program *vprog, - struct st_fragment_program *fprog) -{ - struct linked_program_pair *pair; - - /* search */ - for (pair = Pairs; pair; pair = pair->next) { - if (pair->vprog == vprog && pair->fprog == fprog) { - /* found it */ - break; + /* See if we need to translate vertex program to TGSI form */ + if (xvp->serialNo != stvp->serialNo) { + GLuint outAttr, dummySlot; + const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten; + + /* Compute mapping of vertex program outputs to slots, which depends + * on the fragment program's input->slot mapping. + */ + for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { + /* set default: */ + xvp->output_to_slot[outAttr] = UNUSED; + + if (outputsWritten & (1 << outAttr)) { + /* see if the frag prog wants this vert output */ + GLint fpIn = vp_out_to_fp_in(outAttr); + if (fpIn >= 0) { + xvp->output_to_slot[outAttr] = stfp->input_to_slot[fpIn]; + } + } } - } - /* - * Examine the outputs of the vertex shader and the inputs of the - * fragment shader to determine how to match both to a common set - * of slots. - */ - if (!pair) { - pair = CALLOC_STRUCT(linked_program_pair); - if (pair) { - pair->vprog = vprog; - pair->fprog = fprog; + /* Unneeded vertex program outputs will go to this slot. + * We could use this info to do dead code elimination in the + * vertex program. + */ + dummySlot = stfp->num_input_slots; + + /* Map vert program outputs that aren't used to the dummy slot */ + for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { + if (outputsWritten & (1 << outAttr)) { + if (xvp->output_to_slot[outAttr] == UNUSED) + xvp->output_to_slot[outAttr] = dummySlot; + } } - } - return pair; -} + xvp->vs = st_translate_vertex_program(st, stvp, + xvp->output_to_slot, + xvp->tokens, + ST_MAX_SHADER_TOKENS); + assert(xvp->vs); + stvp->vs = NULL; /* don't want to use this */ -static void -link_shaders(struct st_context *st, struct linked_program_pair *pair) -{ - struct st_vertex_program *vprog = pair->vprog; - struct st_fragment_program *fprog = pair->fprog; - - assert(vprog); - assert(fprog); - - if (pair->vertSerialNo != vprog->serialNo || - pair->fragSerialNo != fprog->serialNo) { - /* re-link and re-translate */ - GLuint vert_output_mapping[VERT_RESULT_MAX]; - GLuint frag_input_mapping[FRAG_ATTRIB_MAX]; - - link_outputs_to_inputs(vprog->Base.Base.OutputsWritten, - fprog->Base.Base.InputsRead | FRAG_BIT_WPOS, - vert_output_mapping, - frag_input_mapping); - - /* xlate vp to vs + vs tokens */ - st_translate_vertex_program(st, vprog, - vert_output_mapping, - pair->vs_tokens, ST_FP_MAX_TOKENS); - - pair->vprog = vprog; - /* temp hacks */ - pair->vs = vprog->vs; - vprog->vs = NULL; - - - /* xlate fp to fs + fs tokens */ - st_translate_fragment_program(st, fprog, - frag_input_mapping, - pair->fs_tokens, ST_FP_MAX_TOKENS); - pair->fprog = fprog; - /* temp hacks */ - pair->fs = fprog->fs; - fprog->fs = NULL; - - /* save pair */ - pair->next = Pairs; - Pairs = pair; - - pair->vertSerialNo = vprog->serialNo; - pair->fragSerialNo = fprog->serialNo; + /* translated VP is up to date now */ + xvp->serialNo = stvp->serialNo; } + + return xvp; } static void update_linkage( struct st_context *st ) { - struct linked_program_pair *pair; struct st_vertex_program *stvp; struct st_fragment_program *stfp; + struct translated_vertex_program *xvp; /* find active shader and params -- Should be covered by * ST_NEW_VERTEX_PROGRAM @@ -392,23 +283,17 @@ update_linkage( struct st_context *st ) stfp = st_fragment_program(st->ctx->FragmentProgram._Current); } + xvp = find_translated_vp(st, stvp, stfp); - pair = lookup_program_pair(st, stvp, stfp); - assert(pair); - link_shaders(st, pair); - - - /* Bind the vertex program and TGSI shader */ st->vp = stvp; - st->state.vs = pair->vs; + st->state.vs = xvp->vs; st->pipe->bind_vs_state(st->pipe, st->state.vs->data); - /* Bind the fragment program and TGSI shader */ st->fp = stfp; - st->state.fs = pair->fs; + st->state.fs = stfp->fs; st->pipe->bind_fs_state(st->pipe, st->state.fs->data); - st->vertex_result_to_slot = pair->vp_result_to_slot; + st->vertex_result_to_slot = xvp->output_to_slot; } diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 3a70d2f121..cca4fa19a9 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -156,7 +156,7 @@ make_frag_shader(struct st_context *st) stfp = (struct st_fragment_program *) p; st_translate_fragment_program(st, stfp, NULL, - stfp->tokens, ST_FP_MAX_TOKENS); + stfp->tokens, ST_MAX_SHADER_TOKENS); return stfp; } @@ -205,7 +205,7 @@ make_vertex_shader(struct st_context *st) stvp = (struct st_vertex_program *) p; st_translate_vertex_program(st, stvp, NULL, - stvp->tokens, ST_FP_MAX_TOKENS); + stvp->tokens, ST_MAX_SHADER_TOKENS); assert(stvp->vs); return stvp; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index fb89ceef8c..f58b5d947d 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -131,7 +131,7 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) stfp = (struct st_fragment_program *) p; st_translate_fragment_program(st, stfp, NULL, - stfp->tokens, ST_FP_MAX_TOKENS); + stfp->tokens, ST_MAX_SHADER_TOKENS); return stfp; } @@ -203,7 +203,7 @@ make_vertex_shader(struct st_context *st, GLboolean passColor) stvp = (struct st_vertex_program *) p; st_translate_vertex_program(st, stvp, NULL, - stvp->tokens, ST_FP_MAX_TOKENS); + stvp->tokens, ST_MAX_SHADER_TOKENS); return stvp; } diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 3a7ce9405e..5c00dd1ae1 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -165,6 +165,12 @@ static void st_program_string_notify( GLcontext *ctx, stfp->serialNo++; + if (stfp->fs) { + /* free the TGSI code */ + // cso_delete(stfp->vs); + stfp->fs = NULL; + } + stfp->param_state = stfp->Base.Base.Parameters->StateFlags; } else if (target == GL_VERTEX_PROGRAM_ARB) { @@ -172,6 +178,12 @@ static void st_program_string_notify( GLcontext *ctx, stvp->serialNo++; + if (stvp->vs) { + /* free the TGSI code */ + // cso_delete(stfp->vs); + stvp->vs = NULL; + } + stvp->param_state = stvp->Base.Base.Parameters->StateFlags; /* Also tell tnl about it: diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 2b79201313..a714f3f5b0 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -38,20 +38,28 @@ #include "pipe/tgsi/exec/tgsi_token.h" #include "x86/rtasm/x86sse.h" -#define ST_FP_MAX_TOKENS 1024 + +#define ST_MAX_SHADER_TOKENS 1024 + struct cso_fragment_shader; struct cso_vertex_shader; +struct translated_vertex_program; + +/** + * Derived from Mesa gl_fragment_program: + */ struct st_fragment_program { struct gl_fragment_program Base; - GLboolean error; /* If program is malformed for any reason. */ - GLuint serialNo; + GLuint input_to_slot[FRAG_ATTRIB_MAX]; /**< Maps FRAG_ATTRIB_x to slot */ + GLuint num_input_slots; + /** The program in TGSI format */ - struct tgsi_token tokens[ST_FP_MAX_TOKENS]; + struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; #if defined(__i386__) || defined(__386__) struct x86_function sse2_program; @@ -61,23 +69,29 @@ struct st_fragment_program const struct cso_fragment_shader *fs; GLuint param_state; + + /** List of vertex programs which have been translated such that their + * outputs match this fragment program's inputs. + */ + struct translated_vertex_program *vertex_programs; }; +/** + * Derived from Mesa gl_fragment_program: + */ struct st_vertex_program { struct gl_vertex_program Base; /**< The Mesa vertex program */ - GLboolean error; /**< Set if program is malformed for any reason. */ - GLuint serialNo; /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ GLuint input_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ - GLuint index_to_input[MAX_VERTEX_PROGRAM_ATTRIBS]; + GLuint index_to_input[PIPE_MAX_SHADER_INPUTS]; /** The program in TGSI format */ - struct tgsi_token tokens[ST_FP_MAX_TOKENS]; + struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; /** Pointer to the corresponding cached shader */ const struct cso_vertex_shader *vs; -- cgit v1.2.3 From 78876bf61c6f6e3c3fec471431f91307f51c107a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Sep 2007 16:19:51 -0600 Subject: fix st_use_program() --- src/mesa/state_tracker/st_cb_program.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 5c00dd1ae1..8df8b263d3 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -37,6 +37,7 @@ #include "shader/prog_parameter.h" #include "shader/program.h" #include "shader/programopt.h" +#include "shader/shader_api.h" #include "st_context.h" #include "st_program.h" @@ -70,6 +71,8 @@ static void st_use_program( GLcontext *ctx, struct st_context *st = st_context(ctx); st->dirty.st |= ST_NEW_SHADER; + + _mesa_use_program(ctx, program); } -- cgit v1.2.3 From bbcbd362a94386f2a19570dc09ec74c9ef4db067 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Sep 2007 16:20:10 -0600 Subject: remove old debug code, fix warning --- src/mesa/state_tracker/st_atom_constbuf.c | 11 ----------- 1 file changed, 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index 5285485c94..d140e0ef50 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -65,20 +65,9 @@ static void upload_constants( struct st_context *st, if (0) { - int i; - printf("%s(shader=%d, numParams=%d, stateFlags=0x%x)\n", __FUNCTION__, id, params->NumParameters, params->StateFlags); - _mesa_print_parameter_list(params); - /* - for (i = 0; i < params->NumParameters; i++) - printf("%d: %f %f %f %f\n", i, - params->ParameterValues[i][0], - params->ParameterValues[i][1], - params->ParameterValues[i][2], - params->ParameterValues[i][3]); - */ } /* load Mesa constants into the constant buffer */ -- cgit v1.2.3 From ae7f200788947cdde15b279ea3a4558f6abd9922 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Sep 2007 20:31:24 -0600 Subject: fix bug causing unnecessary translations --- src/mesa/state_tracker/st_atom_shader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index d7a89d1e95..0a905fb563 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -183,7 +183,7 @@ find_translated_vp(struct st_context *st, * XXX This could be a hash lookup, using InputsRead as the key. */ for (xvp = stfp->vertex_programs; xvp; xvp = xvp->next) { - if (xvp->frag_inputs == stfp->Base.Base.InputsRead) { + if (xvp->frag_inputs == fragInputsRead) { break; } } -- cgit v1.2.3 From fc38c827716d333c92017348add1e80ea78af948 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Sep 2007 21:04:34 -0600 Subject: Initial implementation of IF/ELSE/ENDIF using conditional masking. mach->CondMask controls writing to each of the 4 components in a quad. mach->CondMaskStack handles nested conditionals. Only a few of the micro ops (add/sub/mul/store) have been updated to obey CondMask at this time. --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 303 ++++++++++++++---------------------- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 12 +- 2 files changed, 124 insertions(+), 191 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 28dc049658..0ddc3f2965 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -96,6 +96,8 @@ tgsi_exec_machine_init( mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; } + + mach->CondMask = 0xf; } void @@ -175,12 +177,17 @@ static void micro_add( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) + const union tgsi_exec_channel *src1, + uint mask) { - dst->f[0] = src0->f[0] + src1->f[0]; - dst->f[1] = src0->f[1] + src1->f[1]; - dst->f[2] = src0->f[2] + src1->f[2]; - dst->f[3] = src0->f[3] + src1->f[3]; + if (mask & 0x1) + dst->f[0] = src0->f[0] + src1->f[0]; + if (mask & 0x2) + dst->f[1] = src0->f[1] + src1->f[1]; + if (mask & 0x4) + dst->f[2] = src0->f[2] + src1->f[2]; + if (mask & 0x8) + dst->f[3] = src0->f[3] + src1->f[3]; } static void @@ -524,12 +531,17 @@ static void micro_mul( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) + const union tgsi_exec_channel *src1, + uint condMask) { - dst->f[0] = src0->f[0] * src1->f[0]; - dst->f[1] = src0->f[1] * src1->f[1]; - dst->f[2] = src0->f[2] * src1->f[2]; - dst->f[3] = src0->f[3] * src1->f[3]; + if (condMask & 0x1) + dst->f[0] = src0->f[0] * src1->f[0]; + if (condMask & 0x2) + dst->f[1] = src0->f[1] * src1->f[1]; + if (condMask & 0x4) + dst->f[2] = src0->f[2] * src1->f[2]; + if (condMask & 0x8) + dst->f[3] = src0->f[3] * src1->f[3]; } static void @@ -720,12 +732,17 @@ static void micro_sub( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) + const union tgsi_exec_channel *src1, + uint mask) { - dst->f[0] = src0->f[0] - src1->f[0]; - dst->f[1] = src0->f[1] - src1->f[1]; - dst->f[2] = src0->f[2] - src1->f[2]; - dst->f[3] = src0->f[3] - src1->f[3]; + if (mask & 0x1) + dst->f[0] = src0->f[0] - src1->f[0]; + if (mask & 0x2) + dst->f[1] = src0->f[1] - src1->f[1]; + if (mask & 0x4) + dst->f[2] = src0->f[2] - src1->f[2]; + if (mask & 0x8) + dst->f[3] = src0->f[3] - src1->f[3]; } static void @@ -940,7 +957,8 @@ store_dest( const union tgsi_exec_channel *chan, const struct tgsi_full_dst_register *reg, const struct tgsi_full_instruction *inst, - GLuint chan_index ) + GLuint chan_index, + uint mask) { union tgsi_exec_channel *dst; @@ -968,7 +986,18 @@ store_dest( switch (inst->Instruction.Saturate) { case TGSI_SAT_NONE: +#if 0 *dst = *chan; +#else + if (mask & 0x1) + dst->i[0] = chan->i[0]; + if (mask & 0x2) + dst->i[1] = chan->i[1]; + if (mask & 0x4) + dst->i[2] = chan->i[2]; + if (mask & 0x8) + dst->i[3] = chan->i[3]; +#endif break; case TGSI_SAT_ZERO_ONE: @@ -989,7 +1018,10 @@ store_dest( fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) #define STORE(VAL,INDEX,CHAN)\ - store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN) + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN, ~0) + +#define STORE_MASKED(VAL,INDEX,CHAN,MASK) \ + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN, MASK) static void exec_kil (struct tgsi_exec_machine *mach, @@ -1113,6 +1145,7 @@ perspective_interpolation( } } + typedef void (* interpolation_func)( struct tgsi_exec_machine *mach, unsigned attrib, @@ -1209,7 +1242,7 @@ exec_instruction( /* TGSI_OPCODE_SWZ */ FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - STORE( &r[0], 0, chan_index ); + STORE_MASKED( &r[0], 0, chan_index, mach->CondMask ); } break; @@ -1276,7 +1309,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); STORE(&r[0], 0, chan_index); } @@ -1286,7 +1319,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); STORE( &r[0], 0, chan_index ); } break; @@ -1295,17 +1328,17 @@ exec_instruction( /* TGSI_OPCODE_DOT3 */ FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); FETCH( &r[1], 0, CHAN_Y ); FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); FETCH( &r[1], 0, CHAN_Z ); FETCH( &r[2], 1, CHAN_Z ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1317,25 +1350,25 @@ exec_instruction( FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 1, CHAN_Y); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); FETCH(&r[1], 0, CHAN_Z); FETCH(&r[2], 1, CHAN_Z); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); FETCH(&r[1], 0, CHAN_W); FETCH(&r[2], 1, CHAN_W); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1350,7 +1383,7 @@ exec_instruction( if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { FETCH( &r[0], 0, CHAN_Y ); FETCH( &r[1], 1, CHAN_Y); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); STORE( &r[0], 0, CHAN_Y ); } @@ -1412,9 +1445,9 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); FETCH( &r[1], 2, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); STORE( &r[0], 0, chan_index ); } break; @@ -1424,7 +1457,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - micro_sub( &r[0], &r[0], &r[1] ); + micro_sub( &r[0], &r[0], &r[1], mach->CondMask ); STORE(&r[0], 0, chan_index); } @@ -1437,9 +1470,9 @@ exec_instruction( FETCH(&r[1], 1, chan_index); FETCH(&r[2], 2, chan_index); - micro_sub( &r[1], &r[1], &r[2] ); - micro_mul( &r[0], &r[0], &r[1] ); - micro_add( &r[0], &r[0], &r[2] ); + micro_sub( &r[1], &r[1], &r[2], mach->CondMask ); + micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); + micro_add( &r[0], &r[0], &r[2], mach->CondMask ); STORE(&r[0], 0, chan_index); } @@ -1533,13 +1566,13 @@ exec_instruction( FETCH(&r[0], 0, CHAN_Y); FETCH(&r[1], 1, CHAN_Z); - micro_mul( &r[2], &r[0], &r[1] ); + micro_mul( &r[2], &r[0], &r[1], mach->CondMask ); FETCH(&r[3], 0, CHAN_Z); FETCH(&r[4], 1, CHAN_Y); - micro_mul( &r[5], &r[3], &r[4] ); - micro_sub( &r[2], &r[2], &r[5] ); + micro_mul( &r[5], &r[3], &r[4], mach->CondMask ); + micro_sub( &r[2], &r[2], &r[5], mach->CondMask ); if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { STORE( &r[2], 0, CHAN_X ); @@ -1547,20 +1580,20 @@ exec_instruction( FETCH(&r[2], 1, CHAN_X); - micro_mul( &r[3], &r[3], &r[2] ); + micro_mul( &r[3], &r[3], &r[2], mach->CondMask ); FETCH(&r[5], 0, CHAN_X); - micro_mul( &r[1], &r[1], &r[5] ); - micro_sub( &r[3], &r[3], &r[1] ); + micro_mul( &r[1], &r[1], &r[5], mach->CondMask ); + micro_sub( &r[3], &r[3], &r[1], mach->CondMask ); if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { STORE( &r[3], 0, CHAN_Y ); } - micro_mul( &r[5], &r[5], &r[4] ); - micro_mul( &r[0], &r[0], &r[2] ); - micro_sub( &r[5], &r[5], &r[0] ); + micro_mul( &r[5], &r[5], &r[4], mach->CondMask ); + micro_mul( &r[0], &r[0], &r[2], mach->CondMask ); + micro_sub( &r[5], &r[5], &r[0], mach->CondMask ); if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { STORE( &r[5], 0, CHAN_Z ); @@ -1593,23 +1626,23 @@ exec_instruction( FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 1, CHAN_Y); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); FETCH(&r[1], 0, CHAN_Z); FETCH(&r[2], 1, CHAN_Z); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); FETCH(&r[1], 1, CHAN_W); - micro_add( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1892,12 +1925,12 @@ exec_instruction( case TGSI_OPCODE_DP2: FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); FETCH( &r[1], 0, CHAN_Y ); FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1], mach->CondMask ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1913,42 +1946,21 @@ exec_instruction( break; case TGSI_OPCODE_IF: - { - GLuint cond = 0; - struct tgsi_exec_cond_state *state; - - /* Allocate condition state. */ - assert( mach->CondStack.Index > 0 ); - mach->CondStack.Index--; - - /* Evaluate the condition mask. */ - FETCH( &r[0], 0, CHAN_X ); - if( r[0].u[0] ) { - cond |= 1; - } - if( r[0].u[1] ) { - cond |= 2; - } - if( r[0].u[2] ) { - cond |= 4; - } - if( r[0].u[3] ) { - cond |= 8; - } - - state = &mach->CondStack.States[mach->CondStack.Index]; - - /* Initialize the If portion of condition state. */ - memcpy( - state->IfPortion.TempsAddrs, - mach->Temps, - sizeof( state->IfPortion.TempsAddrs ) ); - memcpy( - state->IfPortion.Outputs, - mach->Outputs, - sizeof( state->IfPortion.Outputs ) ); - state->Condition = cond; - state->WasElse = GL_FALSE; + /* push CondMask */ + mach->condStack[mach->CondStackTop++] = mach->CondMask; + FETCH( &r[0], 0, CHAN_X ); + /* update CondMask */ + if( ! r[0].u[0] ) { + mach->CondMask &= ~0x1; + } + if( ! r[0].u[1] ) { + mach->CondMask &= ~0x2; + } + if( ! r[0].u[2] ) { + mach->CondMask &= ~0x4; + } + if( ! r[0].u[3] ) { + mach->CondMask &= ~0x8; } break; @@ -1961,106 +1973,19 @@ exec_instruction( break; case TGSI_OPCODE_ELSE: + /* invert CondMask wrt previous mask */ { - struct tgsi_exec_cond_state *state; - struct tgsi_exec_cond_regs temp; - - state = &mach->CondStack.States[mach->CondStack.Index]; - - /* Copy the results of the If portion to temporary storage. */ - memcpy( - temp.TempsAddrs, - mach->Temps, - sizeof( temp.TempsAddrs ) ); - memcpy( - temp.Outputs, - mach->Outputs, - sizeof( temp.Outputs ) ); - - /* Restore the state of registers from before the If statement. */ - memcpy( - mach->Temps, - state->IfPortion.TempsAddrs, - sizeof( state->IfPortion.TempsAddrs ) ); - memcpy( - mach->Outputs, - state->IfPortion.Outputs, - sizeof( state->IfPortion.Outputs ) ); - - /* Save the results of If portion. */ - memcpy( - &state->IfPortion, - &temp, - sizeof( state->IfPortion ) ); - state->WasElse = GL_TRUE; + uint prevMask; + assert(mach->CondStackTop > 0); + prevMask = mach->condStack[mach->CondStackTop - 1]; + mach->CondMask = ~mach->CondMask & prevMask; } break; case TGSI_OPCODE_ENDIF: - { - struct tgsi_exec_cond_state *state; - GLuint i; - - state = &mach->CondStack.States[mach->CondStack.Index]; - - if( state->WasElse ) { - /* Save the results of Else portion. */ - memcpy( - state->ElsePortion.TempsAddrs, - mach->Temps, - sizeof( state->ElsePortion.TempsAddrs ) ); - memcpy( - state->ElsePortion.Outputs, - mach->Outputs, - sizeof( state->ElsePortion.Outputs ) ); - } - else { - /* Copy the state of registers from before the If statement to Else portion. */ - memcpy( - &state->ElsePortion, - &state->IfPortion, - sizeof( state->ElsePortion ) ); - - /* Save the results of the If portion. */ - memcpy( - state->IfPortion.TempsAddrs, - mach->Temps, - sizeof( state->IfPortion.TempsAddrs ) ); - memcpy( - state->IfPortion.Outputs, - mach->Outputs, - sizeof( state->IfPortion.Outputs ) ); - } - - /* Mix the If and Else portions based on condition mask. */ - for( i = 0; i < 4; i++ ) { - struct tgsi_exec_cond_regs *regs; - GLuint j; - - if( state->Condition & (1 << i) ) { - regs = &state->IfPortion; - } - else { - regs = &state->ElsePortion; - } - - for( j = 0; j < TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS; j++ ) { - mach->Temps[j].xyzw[0].u[i] = regs->TempsAddrs[j].xyzw[0].u[i]; - mach->Temps[j].xyzw[1].u[i] = regs->TempsAddrs[j].xyzw[1].u[i]; - mach->Temps[j].xyzw[2].u[i] = regs->TempsAddrs[j].xyzw[2].u[i]; - mach->Temps[j].xyzw[3].u[i] = regs->TempsAddrs[j].xyzw[3].u[i]; - } - for( j = 0; j < 2; j++ ) { - mach->Outputs[j].xyzw[0].u[i] = regs->Outputs[j].xyzw[0].u[i]; - mach->Outputs[j].xyzw[1].u[i] = regs->Outputs[j].xyzw[1].u[i]; - mach->Outputs[j].xyzw[2].u[i] = regs->Outputs[j].xyzw[2].u[i]; - mach->Outputs[j].xyzw[3].u[i] = regs->Outputs[j].xyzw[3].u[i]; - } - } - - /* Release condition state. */ - mach->CondStack.Index++; - } + assert(mach->CondStackTop > 0); + /* pop CondMask */ + mach->CondMask = mach->condStack[--mach->CondStackTop]; break; case TGSI_OPCODE_ENDLOOP: @@ -2247,8 +2172,6 @@ tgsi_exec_machine_run2( mach->Primitives[0] = 0; } - mach->CondStack.Index = 8; - k = tgsi_parse_init( &parse, mach->Tokens ); if (k != TGSI_PARSE_OK) { printf("Problem parsing!\n"); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index d50f786029..f87d6e8862 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -120,6 +120,10 @@ struct tgsi_exec_cond_stack unsigned Index; /* into States[] */ }; + +#define TGSI_EXEC_MAX_COND_NESTING 10 + + struct tgsi_exec_machine { /* @@ -152,9 +156,15 @@ struct tgsi_exec_machine /* FRAGMENT processor only. */ const struct tgsi_interp_coef *InterpCoefs; - struct tgsi_exec_cond_stack CondStack; + /* Conditional execution mask */ + uint CondMask; + + /* Condition mask stack (for nested conditionals) */ + uint condStack[TGSI_EXEC_MAX_COND_NESTING]; + int CondStackTop; }; + void tgsi_exec_machine_init( struct tgsi_exec_machine *mach, -- cgit v1.2.3 From 91550f0a17488f41da4829a627db90f0081250dc Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 29 Sep 2007 09:51:11 -0600 Subject: remove unused tgsi_exec_cond_* types --- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 24 ------------------------ 1 file changed, 24 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index f87d6e8862..54b38cb4b2 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -97,30 +97,6 @@ struct tgsi_exec_labels #define TGSI_EXEC_NUM_TEMPS (32 + 4) #define TGSI_EXEC_NUM_ADDRS 1 -/* XXX: This is temporary */ -struct tgsi_exec_cond_regs -{ - struct tgsi_exec_vector TempsAddrs[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS]; - struct tgsi_exec_vector Outputs[2]; /* XXX: That's just enough for fragment shader only! */ -}; - -/* XXX: This is temporary */ -struct tgsi_exec_cond_state -{ - struct tgsi_exec_cond_regs IfPortion; - struct tgsi_exec_cond_regs ElsePortion; - unsigned Condition; - boolean WasElse; -}; - -/* XXX: This is temporary */ -struct tgsi_exec_cond_stack -{ - struct tgsi_exec_cond_state States[8]; - unsigned Index; /* into States[] */ -}; - - #define TGSI_EXEC_MAX_COND_NESTING 10 -- cgit v1.2.3 From 78f3cd1e08d68111a2db308c000b94cc580b3cad Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 29 Sep 2007 10:43:29 -0600 Subject: Initial conditional execution support for loops and BRK instruction. Also, instead of passing cond mask to each micro op, just apply it in the store_dest() function. --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 188 +++++++++++++++++++----------------- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 16 ++- 2 files changed, 110 insertions(+), 94 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 0ddc3f2965..90da2ee2db 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -54,6 +54,12 @@ FOR_EACH_CHANNEL( CHAN )\ if (IS_CHANNEL_ENABLED2( INST, CHAN )) + +/** The execution mask depends on the conditional mask and the loop mask */ +#define UPDATE_EXEC_MASK(MACH) \ + MACH->ExecMask = MACH->CondMask & MACH->LoopMask + + #define CHAN_X 0 #define CHAN_Y 1 #define CHAN_Z 2 @@ -98,6 +104,8 @@ tgsi_exec_machine_init( } mach->CondMask = 0xf; + mach->LoopMask = 0xf; + mach->ExecMask = 0xf; } void @@ -177,17 +185,12 @@ static void micro_add( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - uint mask) + const union tgsi_exec_channel *src1 ) { - if (mask & 0x1) - dst->f[0] = src0->f[0] + src1->f[0]; - if (mask & 0x2) - dst->f[1] = src0->f[1] + src1->f[1]; - if (mask & 0x4) - dst->f[2] = src0->f[2] + src1->f[2]; - if (mask & 0x8) - dst->f[3] = src0->f[3] + src1->f[3]; + dst->f[0] = src0->f[0] + src1->f[0]; + dst->f[1] = src0->f[1] + src1->f[1]; + dst->f[2] = src0->f[2] + src1->f[2]; + dst->f[3] = src0->f[3] + src1->f[3]; } static void @@ -313,7 +316,7 @@ micro_ieq( static void micro_exp2( union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) + const union tgsi_exec_channel *src) { dst->f[0] = (GLfloat) pow( 2.0, (GLdouble) src->f[0] ); dst->f[1] = (GLfloat) pow( 2.0, (GLdouble) src->f[1] ); @@ -531,17 +534,12 @@ static void micro_mul( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - uint condMask) + const union tgsi_exec_channel *src1 ) { - if (condMask & 0x1) - dst->f[0] = src0->f[0] * src1->f[0]; - if (condMask & 0x2) - dst->f[1] = src0->f[1] * src1->f[1]; - if (condMask & 0x4) - dst->f[2] = src0->f[2] * src1->f[2]; - if (condMask & 0x8) - dst->f[3] = src0->f[3] * src1->f[3]; + dst->f[0] = src0->f[0] * src1->f[0]; + dst->f[1] = src0->f[1] * src1->f[1]; + dst->f[2] = src0->f[2] * src1->f[2]; + dst->f[3] = src0->f[3] * src1->f[3]; } static void @@ -732,17 +730,12 @@ static void micro_sub( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - uint mask) + const union tgsi_exec_channel *src1 ) { - if (mask & 0x1) - dst->f[0] = src0->f[0] - src1->f[0]; - if (mask & 0x2) - dst->f[1] = src0->f[1] - src1->f[1]; - if (mask & 0x4) - dst->f[2] = src0->f[2] - src1->f[2]; - if (mask & 0x8) - dst->f[3] = src0->f[3] - src1->f[3]; + dst->f[0] = src0->f[0] - src1->f[0]; + dst->f[1] = src0->f[1] - src1->f[1]; + dst->f[2] = src0->f[2] - src1->f[2]; + dst->f[3] = src0->f[3] - src1->f[3]; } static void @@ -957,8 +950,7 @@ store_dest( const union tgsi_exec_channel *chan, const struct tgsi_full_dst_register *reg, const struct tgsi_full_instruction *inst, - GLuint chan_index, - uint mask) + GLuint chan_index ) { union tgsi_exec_channel *dst; @@ -989,13 +981,13 @@ store_dest( #if 0 *dst = *chan; #else - if (mask & 0x1) + if (mach->ExecMask & 0x1) dst->i[0] = chan->i[0]; - if (mask & 0x2) + if (mach->ExecMask & 0x2) dst->i[1] = chan->i[1]; - if (mask & 0x4) + if (mach->ExecMask & 0x4) dst->i[2] = chan->i[2]; - if (mask & 0x8) + if (mach->ExecMask & 0x8) dst->i[3] = chan->i[3]; #endif break; @@ -1018,10 +1010,8 @@ store_dest( fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) #define STORE(VAL,INDEX,CHAN)\ - store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN, ~0) + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) -#define STORE_MASKED(VAL,INDEX,CHAN,MASK) \ - store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN, MASK) static void exec_kil (struct tgsi_exec_machine *mach, @@ -1242,7 +1232,7 @@ exec_instruction( /* TGSI_OPCODE_SWZ */ FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - STORE_MASKED( &r[0], 0, chan_index, mach->CondMask ); + STORE( &r[0], 0, chan_index ); } break; @@ -1309,7 +1299,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[0], &r[0], &r[1] ); STORE(&r[0], 0, chan_index); } @@ -1319,7 +1309,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1] ); STORE( &r[0], 0, chan_index ); } break; @@ -1328,17 +1318,17 @@ exec_instruction( /* TGSI_OPCODE_DOT3 */ FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[0], &r[0], &r[1] ); FETCH( &r[1], 0, CHAN_Y ); FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); FETCH( &r[1], 0, CHAN_Z ); FETCH( &r[2], 1, CHAN_Z ); - micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1350,25 +1340,25 @@ exec_instruction( FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); - micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[0], &r[0], &r[1] ); FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 1, CHAN_Y); - micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); FETCH(&r[1], 0, CHAN_Z); FETCH(&r[2], 1, CHAN_Z); - micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); FETCH(&r[1], 0, CHAN_W); FETCH(&r[2], 1, CHAN_W); - micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1383,7 +1373,7 @@ exec_instruction( if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { FETCH( &r[0], 0, CHAN_Y ); FETCH( &r[1], 1, CHAN_Y); - micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[0], &r[0], &r[1] ); STORE( &r[0], 0, CHAN_Y ); } @@ -1416,7 +1406,7 @@ exec_instruction( micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); - STORE(&r[0], 0, chan_index); + STORE(&r[0], 0, chan_index ); } break; @@ -1445,9 +1435,9 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[0], &r[0], &r[1] ); FETCH( &r[1], 2, chan_index ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1] ); STORE( &r[0], 0, chan_index ); } break; @@ -1457,7 +1447,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - micro_sub( &r[0], &r[0], &r[1], mach->CondMask ); + micro_sub( &r[0], &r[0], &r[1] ); STORE(&r[0], 0, chan_index); } @@ -1470,9 +1460,9 @@ exec_instruction( FETCH(&r[1], 1, chan_index); FETCH(&r[2], 2, chan_index); - micro_sub( &r[1], &r[1], &r[2], mach->CondMask ); - micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); - micro_add( &r[0], &r[0], &r[2], mach->CondMask ); + micro_sub( &r[1], &r[1], &r[2] ); + micro_mul( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[2] ); STORE(&r[0], 0, chan_index); } @@ -1566,13 +1556,13 @@ exec_instruction( FETCH(&r[0], 0, CHAN_Y); FETCH(&r[1], 1, CHAN_Z); - micro_mul( &r[2], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[2], &r[0], &r[1] ); FETCH(&r[3], 0, CHAN_Z); FETCH(&r[4], 1, CHAN_Y); - micro_mul( &r[5], &r[3], &r[4], mach->CondMask ); - micro_sub( &r[2], &r[2], &r[5], mach->CondMask ); + micro_mul( &r[5], &r[3], &r[4] ); + micro_sub( &r[2], &r[2], &r[5] ); if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { STORE( &r[2], 0, CHAN_X ); @@ -1580,20 +1570,20 @@ exec_instruction( FETCH(&r[2], 1, CHAN_X); - micro_mul( &r[3], &r[3], &r[2], mach->CondMask ); + micro_mul( &r[3], &r[3], &r[2] ); FETCH(&r[5], 0, CHAN_X); - micro_mul( &r[1], &r[1], &r[5], mach->CondMask ); - micro_sub( &r[3], &r[3], &r[1], mach->CondMask ); + micro_mul( &r[1], &r[1], &r[5] ); + micro_sub( &r[3], &r[3], &r[1] ); if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { STORE( &r[3], 0, CHAN_Y ); } - micro_mul( &r[5], &r[5], &r[4], mach->CondMask ); - micro_mul( &r[0], &r[0], &r[2], mach->CondMask ); - micro_sub( &r[5], &r[5], &r[0], mach->CondMask ); + micro_mul( &r[5], &r[5], &r[4] ); + micro_mul( &r[0], &r[0], &r[2] ); + micro_sub( &r[5], &r[5], &r[0] ); if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { STORE( &r[5], 0, CHAN_Z ); @@ -1626,23 +1616,23 @@ exec_instruction( FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); - micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[0], &r[0], &r[1] ); FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 1, CHAN_Y); - micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); FETCH(&r[1], 0, CHAN_Z); FETCH(&r[2], 1, CHAN_Z); - micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); FETCH(&r[1], 1, CHAN_W); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_add( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1925,12 +1915,12 @@ exec_instruction( case TGSI_OPCODE_DP2: FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[0], &r[0], &r[1] ); FETCH( &r[1], 0, CHAN_Y ); FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2], mach->CondMask ); - micro_add( &r[0], &r[0], &r[1], mach->CondMask ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1942,12 +1932,15 @@ exec_instruction( break; case TGSI_OPCODE_BRK: - assert (0); + /* turn off loop channels for each enabled exec channel */ + mach->LoopMask &= ~mach->ExecMask; + UPDATE_EXEC_MASK(mach); break; case TGSI_OPCODE_IF: /* push CondMask */ - mach->condStack[mach->CondStackTop++] = mach->CondMask; + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; FETCH( &r[0], 0, CHAN_X ); /* update CondMask */ if( ! r[0].u[0] ) { @@ -1962,10 +1955,13 @@ exec_instruction( if( ! r[0].u[3] ) { mach->CondMask &= ~0x8; } + UPDATE_EXEC_MASK(mach); break; case TGSI_OPCODE_LOOP: - assert (0); + /* push LoopMask */ + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; break; case TGSI_OPCODE_REP: @@ -1977,20 +1973,25 @@ exec_instruction( { uint prevMask; assert(mach->CondStackTop > 0); - prevMask = mach->condStack[mach->CondStackTop - 1]; + prevMask = mach->CondStack[mach->CondStackTop - 1]; mach->CondMask = ~mach->CondMask & prevMask; + UPDATE_EXEC_MASK(mach); } break; case TGSI_OPCODE_ENDIF: - assert(mach->CondStackTop > 0); /* pop CondMask */ - mach->CondMask = mach->condStack[--mach->CondStackTop]; + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + UPDATE_EXEC_MASK(mach); break; case TGSI_OPCODE_ENDLOOP: - assert (0); - break; + /* pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + UPDATE_EXEC_MASK(mach); + break; case TGSI_OPCODE_ENDREP: assert (0); @@ -2109,7 +2110,9 @@ exec_instruction( break; case TGSI_OPCODE_BGNLOOP2: - assert( 0 ); + /* push LoopMask */ + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; break; case TGSI_OPCODE_BGNSUB: @@ -2117,7 +2120,10 @@ exec_instruction( break; case TGSI_OPCODE_ENDLOOP2: - assert( 0 ); + /* pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + UPDATE_EXEC_MASK(mach); break; case TGSI_OPCODE_ENDSUB: diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 54b38cb4b2..e67a8138e8 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -98,8 +98,12 @@ struct tgsi_exec_labels #define TGSI_EXEC_NUM_ADDRS 1 #define TGSI_EXEC_MAX_COND_NESTING 10 +#define TGSI_EXEC_MAX_LOOP_NESTING 10 +/** + * Run-time virtual machine state for executing TGSI shader. + */ struct tgsi_exec_machine { /* @@ -132,12 +136,18 @@ struct tgsi_exec_machine /* FRAGMENT processor only. */ const struct tgsi_interp_coef *InterpCoefs; - /* Conditional execution mask */ + /* Conditional execution masks */ uint CondMask; + uint LoopMask; + uint ExecMask; /**< = CondMask & LoopMask */ - /* Condition mask stack (for nested conditionals) */ - uint condStack[TGSI_EXEC_MAX_COND_NESTING]; + /** Condition mask stack (for nested conditionals) */ + uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; int CondStackTop; + + /** Loop mask stack (for nested loops) */ + uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopStackTop; }; -- cgit v1.2.3 From b13618c3167002a823e9bb1e744a0a316fd0e2bf Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 1 Oct 2007 13:45:53 -0600 Subject: fix a few vert/frag program items to get i915 driver going again --- src/mesa/pipe/i915simple/i915_context.c | 2 -- src/mesa/pipe/p_defines.h | 6 ------ src/mesa/pipe/softpipe/sp_context.c | 2 -- src/mesa/state_tracker/st_atom_shader.c | 5 ++++- src/mesa/state_tracker/st_program.c | 9 ++++----- 5 files changed, 8 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 89252e38d5..b27ab48a50 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -142,8 +142,6 @@ static int i915_get_param(struct pipe_context *pipe, int param) { switch (param) { - case PIPE_PARAM_FS_NEEDS_POS: - return 0; default: return 0; } diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 2a8109b10c..d336f83998 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -308,10 +308,4 @@ #define PIPE_QUERY_TYPES 3 -/** - * Pipe capabilities/queries - */ -#define PIPE_PARAM_FS_NEEDS_POS 1 - - #endif diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 695bf1a9e0..610e7ad689 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -239,8 +239,6 @@ static const char *softpipe_get_vendor( struct pipe_context *pipe ) static int softpipe_get_param(struct pipe_context *pipe, int param) { switch (param) { - case PIPE_PARAM_FS_NEEDS_POS: - return 1; default: return 0; } diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 0a905fb563..8aa977179a 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -57,6 +57,8 @@ */ struct translated_vertex_program { + struct st_vertex_program *master; + /** The fragment shader "signature" this vertex shader is meant for: */ GLbitfield frag_inputs; @@ -183,7 +185,7 @@ find_translated_vp(struct st_context *st, * XXX This could be a hash lookup, using InputsRead as the key. */ for (xvp = stfp->vertex_programs; xvp; xvp = xvp->next) { - if (xvp->frag_inputs == fragInputsRead) { + if (xvp->master == stvp && xvp->frag_inputs == fragInputsRead) { break; } } @@ -192,6 +194,7 @@ find_translated_vp(struct st_context *st, if (!xvp) { xvp = CALLOC_STRUCT(translated_vertex_program); xvp->frag_inputs = fragInputsRead; + xvp->master = stvp; xvp->next = stfp->vertex_programs; stfp->vertex_programs = xvp; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index a41b953a21..a00f296d15 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -281,12 +281,11 @@ st_translate_fragment_program(struct st_context *st, GLuint attr; GLbitfield inputsRead = stfp->Base.Base.InputsRead; - /* Check if all fragment programs need the fragment position (in order - * to do perspective-corrected interpolation). + /* For software rendering, we always need the fragment input position + * in order to calculate interpolated values. + * For i915, we always want to emit the semantic info for position. */ - /* XXX temporary! */ - if (st->pipe->get_param(st->pipe, PIPE_PARAM_FS_NEEDS_POS)) - inputsRead |= FRAG_BIT_WPOS; + inputsRead |= FRAG_BIT_WPOS; memset(&fs, 0, sizeof(fs)); -- cgit v1.2.3 From dab288b9821572b7fd3d6c9ceb4b8b3a83cd15bf Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 1 Oct 2007 13:50:24 -0600 Subject: remove specular color assertion (the path works) --- src/mesa/pipe/i915simple/i915_state_derived.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 30a11f1ef2..4b5b6a4fc8 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -69,7 +69,6 @@ static void calculate_vertex_layout( struct i915_context *i915 ) } else { assert(fs->input_semantic_index[i] == 1); - assert(0); /* untested */ front1 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; } -- cgit v1.2.3 From 726060680ba69aaec659f78e24f2db58acd780cb Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 1 Oct 2007 13:54:24 -0600 Subject: don't crash when fog enabled, still not rendered correctly though --- src/mesa/pipe/i915simple/i915_state_derived.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 4b5b6a4fc8..7eb3f3a5fa 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -85,6 +85,10 @@ static void calculate_vertex_layout( struct i915_context *i915 ) vinfo->hwfmt[1] |= hwtc << (unit * 4); } break; + case TGSI_SEMANTIC_FOG: + fprintf(stderr, "i915 fogcoord not implemented yet\n"); + draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE); + break; default: assert(0); } -- cgit v1.2.3 From 3c41589420c90fa3f81da75c1c90be39f68915e9 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 1 Oct 2007 13:54:37 -0600 Subject: disable debug printf --- src/mesa/pipe/i915simple/i915_state_sampler.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index e6623f4356..3b736f9c55 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -257,7 +257,9 @@ i915_update_texture(struct i915_context *i915, uint unit, assert(format); assert(pitch); + /* printf("texture format = 0x%x\n", format); + */ /* MS3 state */ state[0] = -- cgit v1.2.3 From a24031d50c6b4c584aae08316dc3c00e18e24b58 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 1 Oct 2007 13:55:16 -0600 Subject: don't crash in RasterPos if feedback not implemented yet --- src/mesa/state_tracker/st_cb_rasterpos.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 661d155e6d..13580e400b 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -101,7 +101,8 @@ setup_feedback(GLcontext *ctx) feedback.num_attribs++; } - pipe->set_feedback_state(pipe, &feedback); + if (pipe->set_feedback_state) + pipe->set_feedback_state(pipe, &feedback); } @@ -294,7 +295,8 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) pipe->winsys->buffer_data(pipe->winsys, fb_buf.buffer, fb_buf.size, NULL); /* data */ - pipe->set_feedback_buffer(pipe, 0, &fb_buf); + if (pipe->set_feedback_buffer) + pipe->set_feedback_buffer(pipe, 0, &fb_buf); } @@ -347,7 +349,8 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) pipe->winsys->buffer_reference(pipe->winsys, &fb_buf.buffer, NULL); /* restore pipe state */ - pipe->set_feedback_state(pipe, &st->state.feedback); + if (pipe->set_feedback_state) + pipe->set_feedback_state(pipe, &st->state.feedback); } -- cgit v1.2.3 From f8b2148a2b0c28ea5970be275f1bd678aa32094d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 1 Oct 2007 14:46:19 -0600 Subject: avoid updating vertex format when it doesn't really change --- src/mesa/pipe/i915simple/i915_state_derived.c | 74 ++++++++++++++------------- 1 file changed, 39 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 7eb3f3a5fa..95f705929e 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -45,17 +45,17 @@ static void calculate_vertex_layout( struct i915_context *i915 ) { const struct pipe_shader_state *fs = i915->fs; const interp_mode colorInterp = i915->rasterizer->color_interp; - struct vertex_info *vinfo = &i915->current.vertex_info; + struct vertex_info vinfo; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; boolean needW = 0; uint i; boolean texCoords[8]; memset(texCoords, 0, sizeof(texCoords)); - memset(vinfo, 0, sizeof(*vinfo)); + memset(&vinfo, 0, sizeof(vinfo)); /* pos */ - draw_emit_vertex_attr(vinfo, FORMAT_3F, INTERP_LINEAR); + draw_emit_vertex_attr(&vinfo, FORMAT_3F, INTERP_LINEAR); /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ for (i = 0; i < fs->num_inputs; i++) { @@ -64,13 +64,13 @@ static void calculate_vertex_layout( struct i915_context *i915 ) break; case TGSI_SEMANTIC_COLOR: if (fs->input_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); - vinfo->hwfmt[0] |= S4_VFMT_COLOR; + front0 = draw_emit_vertex_attr(&vinfo, FORMAT_4UB, colorInterp); + vinfo.hwfmt[0] |= S4_VFMT_COLOR; } else { assert(fs->input_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(vinfo, FORMAT_4UB, colorInterp); - vinfo->hwfmt[0] |= S4_VFMT_SPEC_FOG; + front1 = draw_emit_vertex_attr(&vinfo, FORMAT_4UB, colorInterp); + vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG; } break; case TGSI_SEMANTIC_GENERIC: @@ -79,15 +79,15 @@ static void calculate_vertex_layout( struct i915_context *i915 ) const uint unit = fs->input_semantic_index[i]; uint hwtc; texCoords[unit] = TRUE; - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(&vinfo, FORMAT_4F, INTERP_PERSPECTIVE); hwtc = TEXCOORDFMT_4D; needW = TRUE; - vinfo->hwfmt[1] |= hwtc << (unit * 4); + vinfo.hwfmt[1] |= hwtc << (unit * 4); } break; case TGSI_SEMANTIC_FOG: fprintf(stderr, "i915 fogcoord not implemented yet\n"); - draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(&vinfo, FORMAT_1F, INTERP_PERSPECTIVE); break; default: assert(0); @@ -99,18 +99,18 @@ static void calculate_vertex_layout( struct i915_context *i915 ) for (i = 0; i < 8; i++) { if (!texCoords[i]) { const uint hwtc = TEXCOORDFMT_NOT_PRESENT; - vinfo->hwfmt[1] |= hwtc << (i* 4); + vinfo.hwfmt[1] |= hwtc << (i* 4); } } /* go back and fill in the vertex position info now that we have needW */ if (needW) { - vinfo->hwfmt[0] |= S4_VFMT_XYZW; - vinfo->format[0] = FORMAT_4F; + vinfo.hwfmt[0] |= S4_VFMT_XYZW; + vinfo.format[0] = FORMAT_4F; } else { - vinfo->hwfmt[0] |= S4_VFMT_XYZ; - vinfo->format[0] = FORMAT_3F; + vinfo.hwfmt[0] |= S4_VFMT_XYZ; + vinfo.format[0] = FORMAT_3F; } /* Additional attributes required for setup: Just twosided @@ -119,31 +119,35 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ if (i915->rasterizer->light_twoside) { if (front0) { - back0 = draw_emit_vertex_attr(vinfo, FORMAT_OMIT, colorInterp); + back0 = draw_emit_vertex_attr(&vinfo, FORMAT_OMIT, colorInterp); } if (back0) { - back1 = draw_emit_vertex_attr(vinfo, FORMAT_OMIT, colorInterp); + back1 = draw_emit_vertex_attr(&vinfo, FORMAT_OMIT, colorInterp); } } - draw_compute_vertex_size(vinfo); - - /* If the attributes have changed, tell the draw module about the new - * vertex layout. We'll also update the hardware vertex format info. - */ - draw_set_vertex_attributes( i915->draw, - NULL,/*vinfo->slot_to_attrib,*/ - vinfo->interp_mode, - vinfo->num_attribs); - - draw_set_twoside_attributes(i915->draw, - front0, back0, front1, back1); - - /* Need to set this flag so that the LIS2/4 registers get set. - * It also means the i915_update_immediate() function must be called - * after this one, in i915_update_derived(). - */ - i915->dirty |= I915_NEW_VERTEX_FORMAT; + draw_compute_vertex_size(&vinfo); + + if (memcmp(&i915->current.vertex_info, &vinfo, sizeof(vinfo))) { + /* If the attributes have changed, tell the draw module about the new + * vertex layout. We'll also update the hardware vertex format info. + */ + draw_set_vertex_attributes( i915->draw, + NULL,/*vinfo.slot_to_attrib,*/ + vinfo.interp_mode, + vinfo.num_attribs); + + draw_set_twoside_attributes(i915->draw, + front0, back0, front1, back1); + + /* Need to set this flag so that the LIS2/4 registers get set. + * It also means the i915_update_immediate() function must be called + * after this one, in i915_update_derived(). + */ + i915->dirty |= I915_NEW_VERTEX_FORMAT; + + memcpy(&i915->current.vertex_info, &vinfo, sizeof(vinfo)); + } } -- cgit v1.2.3 From 4c19f95a3997be7ca27c314952478790a63aeeb7 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 1 Oct 2007 15:39:15 -0600 Subject: use i915_miptree_layout when using an i915 --- src/mesa/pipe/i915simple/i915_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index b27ab48a50..5d97eed247 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -311,7 +311,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, if (i915->flags.is_i945) i915->pipe.mipmap_tree_layout = i945_miptree_layout; else - i915->pipe.mipmap_tree_layout = i945_miptree_layout; + i915->pipe.mipmap_tree_layout = i915_miptree_layout; i915->dirty = ~0; -- cgit v1.2.3 From cdf99a9fad0fcb080244faa7d32b33b1cfaf6ff7 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 1 Oct 2007 18:10:53 -0600 Subject: comments/code for min/max_lod - revisit someday --- src/mesa/state_tracker/st_atom_sampler.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 23ccd55a05..151d724863 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -137,8 +137,18 @@ update_samplers(struct st_context *st) sampler.mag_img_filter = gl_filter_to_img_filter(texobj->MagFilter); sampler.lod_bias = st->ctx->Texture.Unit[u].LodBias; +#if 1 sampler.min_lod = texobj->MinLod; sampler.max_lod = texobj->MaxLod; +#else + /* min/max lod should really be as follows (untested). + * Also, calculate_first_last_level() needs to be overhauled + * since today's hardware had real support for LOD clamping. + */ + sampler.min_lod = MAX2(texobj->BaseLevel, texobj->MinLod); + sampler.max_lod = MIN2(texobj->MaxLevel, texobj->MaxLod); +#endif + sampler.max_anisotropy = texobj->MaxAnisotropy; /* XXX more sampler state here */ -- cgit v1.2.3 From 57d3770f35730bef17e5d93bd424a59eb6daec4c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 10:38:56 -0600 Subject: checkpoint: unpack all instructions before interpretation. Actual looping works. --- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 2 + src/mesa/pipe/tgsi/exec/tgsi_exec.c | 129 ++++++++++++++++++++++++++++++--- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 6 ++ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 + 4 files changed, 130 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index 2532d5c831..4cf3397162 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -914,6 +914,8 @@ dump_instruction_short( switch( inst->Instruction.Opcode ) { case TGSI_OPCODE_IF: case TGSI_OPCODE_ELSE: + case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_ENDLOOP2: TXT( " :" ); UID( inst->InstructionExtLabel.Label ); break; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 90da2ee2db..77a24ec1d8 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1213,12 +1213,13 @@ static void exec_instruction( struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst, - struct tgsi_exec_labels *labels, - GLuint *programCounter ) + int *pc ) { GLuint chan_index; union tgsi_exec_channel r[8]; + (*pc)++; + switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { @@ -1690,7 +1691,12 @@ exec_instruction( break; case TGSI_OPCODE_SEQ: - assert (0); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } break; case TGSI_OPCODE_SFL: @@ -1934,6 +1940,7 @@ exec_instruction( case TGSI_OPCODE_BRK: /* turn off loop channels for each enabled exec channel */ mach->LoopMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ UPDATE_EXEC_MASK(mach); break; @@ -1956,6 +1963,7 @@ exec_instruction( mach->CondMask &= ~0x8; } UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ELSE */ break; case TGSI_OPCODE_LOOP: @@ -1976,6 +1984,7 @@ exec_instruction( prevMask = mach->CondStack[mach->CondStackTop - 1]; mach->CondMask = ~mach->CondMask & prevMask; UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ENDIF */ } break; @@ -2120,10 +2129,16 @@ exec_instruction( break; case TGSI_OPCODE_ENDLOOP2: - /* pop LoopMask */ - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - UPDATE_EXEC_MASK(mach); + if (mach->LoopMask) { + /* repeat loop: jump to instruction just past BGNLOOP */ + *pc = inst->InstructionExtLabel.Label + 1; + } + else { + /* exit loop: pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + UPDATE_EXEC_MASK(mach); + } break; case TGSI_OPCODE_ENDSUB: @@ -2154,12 +2169,80 @@ exec_instruction( } } + +static void +expand_program(struct tgsi_exec_machine *mach ) +{ + struct tgsi_full_instruction *instructions; + struct tgsi_full_declaration *declarations; + struct tgsi_parse_context parse; + uint k; + uint maxInstructions = 10, numInstructions = 0; + uint maxDeclarations = 10, numDeclarations = 0; + + k = tgsi_parse_init( &parse, mach->Tokens ); + if (k != TGSI_PARSE_OK) { + printf("Problem parsing!\n"); + return; + } + + declarations = (struct tgsi_full_declaration *) + malloc(maxDeclarations * sizeof(struct tgsi_full_declaration)); + + instructions = (struct tgsi_full_instruction *) + malloc(maxInstructions * sizeof(struct tgsi_full_instruction)); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + /* + exec_declaration( mach, &parse.FullToken.FullDeclaration ); + */ + if (numDeclarations == maxDeclarations) { + maxDeclarations += 10; + declarations = realloc(declarations, + maxDeclarations + * sizeof(struct tgsi_full_instruction)); + } + memcpy(declarations + numDeclarations, + &parse.FullToken.FullInstruction, + sizeof(declarations[0])); + numDeclarations++; + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + if (numInstructions == maxInstructions) { + maxInstructions += 10; + instructions = realloc(instructions, + maxInstructions + * sizeof(struct tgsi_full_instruction)); + } + memcpy(instructions + numInstructions, + &parse.FullToken.FullInstruction, + sizeof(instructions[0])); + numInstructions++; + break; + default: + assert( 0 ); + } + } + tgsi_parse_free (&parse); + + mach->Instructions = instructions; + mach->NumInstructions = numInstructions; + mach->Declarations = declarations; + mach->NumDeclarations = numDeclarations; +} + + void tgsi_exec_machine_run2( struct tgsi_exec_machine *mach, struct tgsi_exec_labels *labels ) { -#if MESA +#if 0 && MESA GET_CURRENT_CONTEXT(ctx); GLuint i; #endif @@ -2167,8 +2250,14 @@ tgsi_exec_machine_run2( #if XXX_SSE mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; #else +#if 0 struct tgsi_parse_context parse; GLuint k; +#endif + + if (!mach->Instructions) { + expand_program(mach); + } mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; @@ -2178,12 +2267,13 @@ tgsi_exec_machine_run2( mach->Primitives[0] = 0; } + +#if 0 k = tgsi_parse_init( &parse, mach->Tokens ); if (k != TGSI_PARSE_OK) { printf("Problem parsing!\n"); return; } - while( !tgsi_parse_end_of_tokens( &parse ) ) { tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { @@ -2200,6 +2290,26 @@ tgsi_exec_machine_run2( } } tgsi_parse_free (&parse); +#else + { + uint i; + int pc; + + for (i = 0; i < mach->NumDeclarations; i++) { + exec_declaration( mach, mach->Declarations+i ); + } + + pc = 0; + + while (pc != 99 && pc < mach->NumInstructions) { + exec_instruction( mach, mach->Instructions + pc, &pc ); + } + + free(mach->Declarations); + free(mach->Instructions); + } +#endif + #endif #if 0 @@ -2214,3 +2324,4 @@ tgsi_exec_machine_run2( #endif } + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index e67a8138e8..8997ea9c09 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -148,6 +148,12 @@ struct tgsi_exec_machine /** Loop mask stack (for nested loops) */ uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; int LoopStackTop; + + struct tgsi_full_instruction *Instructions; + uint NumInstructions; + + struct tgsi_full_declaration *Declarations; + uint NumDeclarations; }; diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index fa27fd3cd0..5daf50ddef 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -260,6 +260,7 @@ compile_instruction( break; case OPCODE_BGNLOOP: fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; break; case OPCODE_BGNSUB: fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB; @@ -306,6 +307,7 @@ compile_instruction( break; case OPCODE_ENDLOOP: fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; break; case OPCODE_ENDSUB: fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB; -- cgit v1.2.3 From 0d13ade0cdd38759936a74824efbd6ac8b563aed Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 11:46:11 -0600 Subject: Move tgsi machine state init/allocations so they're done less frequently. This, plus expanding all instructions ahead of time, seems to have improved the performance of program execution by 8x or so. --- src/mesa/pipe/draw/draw_private.h | 4 + src/mesa/pipe/draw/draw_vertex_shader.c | 52 ++++++---- src/mesa/pipe/softpipe/sp_quad_fs.c | 81 +++++++-------- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 176 ++++++++++++++++++-------------- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 8 +- 5 files changed, 179 insertions(+), 142 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 12a970a671..a54fef41e7 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -47,6 +47,8 @@ #include "draw_vertex.h" #include "x86/rtasm/x86sse.h" +#include "pipe/tgsi/exec/tgsi_core.h" + /** * Basic vertex info. @@ -187,6 +189,8 @@ struct draw_context unsigned prim; /**< current prim type: PIPE_PRIM_x */ unsigned reduced_prim; + /** TGSI program interpreter runtime state */ + struct tgsi_exec_machine machine; /* Post-tnl vertex cache: */ diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 3518bd52a3..e3bcd35334 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -86,7 +86,7 @@ run_vertex_program(struct draw_context *draw, unsigned elts[4], unsigned count, struct vertex_header *vOut[]) { - struct tgsi_exec_machine machine; + struct tgsi_exec_machine *machine = &draw->machine; unsigned int j; ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); @@ -98,35 +98,39 @@ run_vertex_program(struct draw_context *draw, assert(draw->vertex_shader->state->output_semantic_name[0] == TGSI_SEMANTIC_POSITION); -#ifdef DEBUG - memset( &machine, 0, sizeof( machine ) ); +#ifdef DEBUG_foo + memset( machine, 0, sizeof( *machine ) ); #endif +#if 0 /* init machine state */ - tgsi_exec_machine_init(&machine, + tgsi_exec_machine_init(machine, draw->vertex_shader->state->tokens, PIPE_MAX_SAMPLERS, NULL /*samplers*/ ); +#endif /* Consts does not require 16 byte alignment. */ - machine.Consts = (float (*)[4]) draw->mapped_constants; + machine->Consts = (float (*)[4]) draw->mapped_constants; - machine.Inputs = ALIGN16_ASSIGN(inputs); - machine.Outputs = ALIGN16_ASSIGN(outputs); + machine->Inputs = ALIGN16_ASSIGN(inputs); + machine->Outputs = ALIGN16_ASSIGN(outputs); - draw_vertex_fetch( draw, &machine, elts, count ); + draw_vertex_fetch( draw, machine, elts, count ); /* run shader */ if( draw->vertex_shader->state->executable != NULL ) { + /* SSE */ codegen_function func = (codegen_function) draw->vertex_shader->state->executable; func( - machine.Inputs, - machine.Outputs, - machine.Consts, - machine.Temps ); + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps ); } else { - tgsi_exec_machine_run( &machine ); + /* interpreter */ + tgsi_exec_machine_run( machine ); } @@ -136,10 +140,10 @@ run_vertex_program(struct draw_context *draw, float x, y, z, w; /* Handle attr[0] (position) specially: */ - x = vOut[j]->clip[0] = machine.Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine.Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine.Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine.Outputs[0].xyzw[3].f[j]; + x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; vOut[j]->clipmask = compute_clipmask(x, y, z, w) | draw->user_clipmask; vOut[j]->edgeflag = 1; @@ -162,10 +166,10 @@ run_vertex_program(struct draw_context *draw, * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. */ for (slot = 1; slot < draw->vertex_info.num_attribs - 2; slot++) { - vOut[j]->data[slot][0] = machine.Outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine.Outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine.Outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine.Outputs[slot].xyzw[3].f[j]; + vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; /* printf("output %d: %f %f %f %f\n", slot, vOut[j]->data[slot][0], @@ -235,6 +239,12 @@ void draw_bind_vertex_shader(struct draw_context *draw, { draw_flush(draw); draw->vertex_shader = (struct draw_vertex_shader*)(vcso); + + /* init machine state */ + tgsi_exec_machine_init(&draw->machine, + draw->vertex_shader->state->tokens, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/ ); } void draw_delete_vertex_shader(struct draw_context *draw, diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 673d339f41..57c01dcfcc 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -45,6 +45,8 @@ struct quad_shade_stage { struct quad_stage stage; struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; + struct tgsi_exec_machine machine; + struct tgsi_exec_vector *inputs, *outputs; }; @@ -83,58 +85,41 @@ shade_quad( struct softpipe_context *softpipe = qs->softpipe; const float fx = (float) quad->x0; const float fy = (float) quad->y0; - struct tgsi_exec_machine machine; - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); - -#ifdef DEBUG - memset( &machine, 0, sizeof( machine ) ); -#endif - - /* init machine state */ - tgsi_exec_machine_init( - &machine, - softpipe->fs->tokens, - PIPE_MAX_SAMPLERS, - qss->samplers ); + struct tgsi_exec_machine *machine = &qss->machine; /* Consts does not require 16 byte alignment. */ - machine.Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; - - machine.Inputs = ALIGN16_ASSIGN(inputs); - machine.Outputs = ALIGN16_ASSIGN(outputs); + machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; - machine.InterpCoefs = quad->coef; + machine->InterpCoefs = quad->coef; - machine.Inputs[0].xyzw[0].f[0] = fx; - machine.Inputs[0].xyzw[0].f[1] = fx + 1.0f; - machine.Inputs[0].xyzw[0].f[2] = fx; - machine.Inputs[0].xyzw[0].f[3] = fx + 1.0f; + machine->Inputs[0].xyzw[0].f[0] = fx; + machine->Inputs[0].xyzw[0].f[1] = fx + 1.0f; + machine->Inputs[0].xyzw[0].f[2] = fx; + machine->Inputs[0].xyzw[0].f[3] = fx + 1.0f; - machine.Inputs[0].xyzw[1].f[0] = fy; - machine.Inputs[0].xyzw[1].f[1] = fy; - machine.Inputs[0].xyzw[1].f[2] = fy + 1.0f; - machine.Inputs[0].xyzw[1].f[3] = fy + 1.0f; + machine->Inputs[0].xyzw[1].f[0] = fy; + machine->Inputs[0].xyzw[1].f[1] = fy; + machine->Inputs[0].xyzw[1].f[2] = fy + 1.0f; + machine->Inputs[0].xyzw[1].f[3] = fy + 1.0f; /* run shader */ if( softpipe->fs->executable != NULL ) { codegen_function func = (codegen_function) softpipe->fs->executable; func( - machine.Inputs, - machine.Outputs, - machine.Consts, - machine.Temps, - machine.InterpCoefs ); + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps, + machine->InterpCoefs ); } else { - tgsi_exec_machine_run( &machine ); + tgsi_exec_machine_run( machine ); } /* store result color (always in output[1]) */ memcpy( quad->outputs.color, - &machine.Outputs[1].xyzw[0].f[0], + &machine->Outputs[1].xyzw[0].f[0], sizeof( quad->outputs.color ) ); #if 0 @@ -142,14 +127,14 @@ shade_quad( /* XXX temporary */ memcpy( quad->outputs.depth, - &machine.Outputs[0].xyzw[2], + machine->Outputs[0].xyzw[2], sizeof( quad->outputs.depth ) ); } #else { uint i; for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine.Inputs[0].xyzw[2].f[i]; + quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; #if 0 printf("output z %f\n", quad->outputs.depth[i]); #endif @@ -188,6 +173,12 @@ static void shade_begin(struct quad_stage *qs) } } + /* XXX only do this if the fragment shader changes... */ + tgsi_exec_machine_init(&qss->machine, + softpipe->fs->tokens, + PIPE_MAX_SAMPLERS, + qss->samplers ); + if (qs->next) qs->next->begin(qs->next); } @@ -195,11 +186,17 @@ static void shade_begin(struct quad_stage *qs) struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) { - struct quad_shade_stage *stage = CALLOC_STRUCT(quad_shade_stage); + struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); + + /* allocate storage for program inputs/outputs, aligned to 16 bytes */ + qss->inputs = malloc(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); + qss->outputs = malloc(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); + qss->machine.Inputs = align16(qss->inputs); + qss->machine.Outputs = align16(qss->outputs); - stage->stage.softpipe = softpipe; - stage->stage.begin = shade_begin; - stage->stage.run = shade_quad; + qss->stage.softpipe = softpipe; + qss->stage.begin = shade_begin; + qss->stage.run = shade_quad; - return &stage->stage; + return &qss->stage; } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 77a24ec1d8..1c515a26e3 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -65,6 +65,80 @@ #define CHAN_Z 2 #define CHAN_W 3 + +static void +expand_program(struct tgsi_exec_machine *mach ) +{ + struct tgsi_full_instruction *instructions; + struct tgsi_full_declaration *declarations; + struct tgsi_parse_context parse; + uint k; + uint maxInstructions = 10, numInstructions = 0; + uint maxDeclarations = 10, numDeclarations = 0; + + k = tgsi_parse_init( &parse, mach->Tokens ); + if (k != TGSI_PARSE_OK) { + printf("Problem parsing!\n"); + return; + } + + declarations = (struct tgsi_full_declaration *) + malloc(maxDeclarations * sizeof(struct tgsi_full_declaration)); + + instructions = (struct tgsi_full_instruction *) + malloc(maxInstructions * sizeof(struct tgsi_full_instruction)); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + /* + exec_declaration( mach, &parse.FullToken.FullDeclaration ); + */ + if (numDeclarations == maxDeclarations) { + maxDeclarations += 10; + declarations = realloc(declarations, + maxDeclarations + * sizeof(struct tgsi_full_instruction)); + } + memcpy(declarations + numDeclarations, + &parse.FullToken.FullInstruction, + sizeof(declarations[0])); + numDeclarations++; + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + if (numInstructions == maxInstructions) { + maxInstructions += 10; + instructions = realloc(instructions, + maxInstructions + * sizeof(struct tgsi_full_instruction)); + } + memcpy(instructions + numInstructions, + &parse.FullToken.FullInstruction, + sizeof(instructions[0])); + numInstructions++; + break; + default: + assert( 0 ); + } + } + tgsi_parse_free (&parse); + + assert(!mach->Instructions); + mach->Instructions = instructions; + mach->NumInstructions = numInstructions; + mach->Declarations = declarations; + mach->NumDeclarations = numDeclarations; +} + + +/** + * Initialize machine state by expanding tokens to full instructions, + * allocating temporary storage, setting up constants, etc. + * After this, we can call tgsi_exec_machine_run() many times. + */ void tgsi_exec_machine_init( struct tgsi_exec_machine *mach, @@ -103,16 +177,32 @@ tgsi_exec_machine_init( mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; } + if (mach->Declarations) { + free(mach->Declarations); + mach->Declarations = NULL; + mach->NumDeclarations = 0; + } + if (mach->Instructions) { + free(mach->Instructions); + mach->Instructions = NULL; + mach->NumInstructions = 0; + } + mach->CondMask = 0xf; mach->LoopMask = 0xf; mach->ExecMask = 0xf; + +#if 01 + tgsi_exec_prepare( mach ); + expand_program(mach); +#endif } void tgsi_exec_prepare( - struct tgsi_exec_machine *mach, - struct tgsi_exec_labels *labels ) + struct tgsi_exec_machine *mach ) { + struct tgsi_exec_labels *labels = &mach->Labels; struct tgsi_parse_context parse; GLuint k; GLuint instno = 0; @@ -164,10 +254,10 @@ void tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) { - struct tgsi_exec_labels labels; - - tgsi_exec_prepare( mach, &labels ); - tgsi_exec_machine_run2( mach, &labels ); +#if 0 + tgsi_exec_prepare( mach ); +#endif + tgsi_exec_machine_run2( mach ); } static void @@ -2170,77 +2260,9 @@ exec_instruction( } -static void -expand_program(struct tgsi_exec_machine *mach ) -{ - struct tgsi_full_instruction *instructions; - struct tgsi_full_declaration *declarations; - struct tgsi_parse_context parse; - uint k; - uint maxInstructions = 10, numInstructions = 0; - uint maxDeclarations = 10, numDeclarations = 0; - - k = tgsi_parse_init( &parse, mach->Tokens ); - if (k != TGSI_PARSE_OK) { - printf("Problem parsing!\n"); - return; - } - - declarations = (struct tgsi_full_declaration *) - malloc(maxDeclarations * sizeof(struct tgsi_full_declaration)); - - instructions = (struct tgsi_full_instruction *) - malloc(maxInstructions * sizeof(struct tgsi_full_instruction)); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - /* - exec_declaration( mach, &parse.FullToken.FullDeclaration ); - */ - if (numDeclarations == maxDeclarations) { - maxDeclarations += 10; - declarations = realloc(declarations, - maxDeclarations - * sizeof(struct tgsi_full_instruction)); - } - memcpy(declarations + numDeclarations, - &parse.FullToken.FullInstruction, - sizeof(declarations[0])); - numDeclarations++; - break; - case TGSI_TOKEN_TYPE_IMMEDIATE: - break; - case TGSI_TOKEN_TYPE_INSTRUCTION: - if (numInstructions == maxInstructions) { - maxInstructions += 10; - instructions = realloc(instructions, - maxInstructions - * sizeof(struct tgsi_full_instruction)); - } - memcpy(instructions + numInstructions, - &parse.FullToken.FullInstruction, - sizeof(instructions[0])); - numInstructions++; - break; - default: - assert( 0 ); - } - } - tgsi_parse_free (&parse); - - mach->Instructions = instructions; - mach->NumInstructions = numInstructions; - mach->Declarations = declarations; - mach->NumDeclarations = numDeclarations; -} - - void tgsi_exec_machine_run2( - struct tgsi_exec_machine *mach, - struct tgsi_exec_labels *labels ) + struct tgsi_exec_machine *mach ) { #if 0 && MESA GET_CURRENT_CONTEXT(ctx); @@ -2255,9 +2277,11 @@ tgsi_exec_machine_run2( GLuint k; #endif +#if 0 if (!mach->Instructions) { expand_program(mach); } +#endif mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; @@ -2305,8 +2329,10 @@ tgsi_exec_machine_run2( exec_instruction( mach, mach->Instructions + pc, &pc ); } +#if 0 free(mach->Declarations); free(mach->Instructions); +#endif } #endif diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 8997ea9c09..2b493ff682 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -154,6 +154,8 @@ struct tgsi_exec_machine struct tgsi_full_declaration *Declarations; uint NumDeclarations; + + struct tgsi_exec_labels Labels; }; @@ -166,8 +168,7 @@ tgsi_exec_machine_init( void tgsi_exec_prepare( - struct tgsi_exec_machine *mach, - struct tgsi_exec_labels *labels ); + struct tgsi_exec_machine *mach ); void tgsi_exec_machine_run( @@ -175,8 +176,7 @@ tgsi_exec_machine_run( void tgsi_exec_machine_run2( - struct tgsi_exec_machine *mach, - struct tgsi_exec_labels *labels ); + struct tgsi_exec_machine *mach ); #if defined __cplusplus } // extern "C" -- cgit v1.2.3 From 9ac03856f563e0e830f6cdeef321875682f79943 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 11:47:21 -0600 Subject: remove dead code --- src/mesa/pipe/draw/draw_vertex_shader.c | 14 +------------- 1 file changed, 1 insertion(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index e3bcd35334..6e8036825d 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -98,18 +98,6 @@ run_vertex_program(struct draw_context *draw, assert(draw->vertex_shader->state->output_semantic_name[0] == TGSI_SEMANTIC_POSITION); -#ifdef DEBUG_foo - memset( machine, 0, sizeof( *machine ) ); -#endif - -#if 0 - /* init machine state */ - tgsi_exec_machine_init(machine, - draw->vertex_shader->state->tokens, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/ ); -#endif - /* Consts does not require 16 byte alignment. */ machine->Consts = (float (*)[4]) draw->mapped_constants; @@ -240,7 +228,7 @@ void draw_bind_vertex_shader(struct draw_context *draw, draw_flush(draw); draw->vertex_shader = (struct draw_vertex_shader*)(vcso); - /* init machine state */ + /* specify the fragment program to interpret/execute */ tgsi_exec_machine_init(&draw->machine, draw->vertex_shader->state->tokens, PIPE_MAX_SAMPLERS, -- cgit v1.2.3 From e2009d0871ac230e4b0c84a23eabef4cd37d81e2 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 11:49:43 -0600 Subject: no more need for tgsi_exec_machine_run2(), remove dead code --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 56 +------------------------------------ src/mesa/pipe/tgsi/exec/tgsi_exec.h | 4 --- 2 files changed, 1 insertion(+), 59 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 1c515a26e3..6b33914c15 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -250,15 +250,6 @@ tgsi_exec_prepare( tgsi_parse_free (&parse); } -void -tgsi_exec_machine_run( - struct tgsi_exec_machine *mach ) -{ -#if 0 - tgsi_exec_prepare( mach ); -#endif - tgsi_exec_machine_run2( mach ); -} static void micro_abs( @@ -2261,27 +2252,11 @@ exec_instruction( void -tgsi_exec_machine_run2( - struct tgsi_exec_machine *mach ) +tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) { -#if 0 && MESA - GET_CURRENT_CONTEXT(ctx); - GLuint i; -#endif - #if XXX_SSE mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; #else -#if 0 - struct tgsi_parse_context parse; - GLuint k; -#endif - -#if 0 - if (!mach->Instructions) { - expand_program(mach); - } -#endif mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; @@ -2292,29 +2267,6 @@ tgsi_exec_machine_run2( } -#if 0 - k = tgsi_parse_init( &parse, mach->Tokens ); - if (k != TGSI_PARSE_OK) { - printf("Problem parsing!\n"); - return; - } - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - exec_declaration( mach, &parse.FullToken.FullDeclaration ); - break; - case TGSI_TOKEN_TYPE_IMMEDIATE: - break; - case TGSI_TOKEN_TYPE_INSTRUCTION: - exec_instruction( mach, &parse.FullToken.FullInstruction, labels, &parse.Position ); - break; - default: - assert( 0 ); - } - } - tgsi_parse_free (&parse); -#else { uint i; int pc; @@ -2328,13 +2280,7 @@ tgsi_exec_machine_run2( while (pc != 99 && pc < mach->NumInstructions) { exec_instruction( mach, mach->Instructions + pc, &pc ); } - -#if 0 - free(mach->Declarations); - free(mach->Instructions); -#endif } -#endif #endif diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 2b493ff682..cf298b7c65 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -174,10 +174,6 @@ void tgsi_exec_machine_run( struct tgsi_exec_machine *mach ); -void -tgsi_exec_machine_run2( - struct tgsi_exec_machine *mach ); - #if defined __cplusplus } // extern "C" #endif // defined __cplusplus -- cgit v1.2.3 From 4726489248283380c0693ebf4eddbe3015caf8a3 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 13:41:56 -0600 Subject: Fold expand_program() into tgsi_exec_prepare() --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 132 +++++++++++++----------------------- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 4 -- 2 files changed, 48 insertions(+), 88 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 6b33914c15..ca397bde6a 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -66,21 +66,21 @@ #define CHAN_W 3 + static void -expand_program(struct tgsi_exec_machine *mach ) +tgsi_exec_prepare( struct tgsi_exec_machine *mach ) { + struct tgsi_exec_labels *labels = &mach->Labels; + struct tgsi_parse_context parse; struct tgsi_full_instruction *instructions; struct tgsi_full_declaration *declarations; - struct tgsi_parse_context parse; - uint k; uint maxInstructions = 10, numInstructions = 0; uint maxDeclarations = 10, numDeclarations = 0; + GLuint k; + GLuint instno = 0; - k = tgsi_parse_init( &parse, mach->Tokens ); - if (k != TGSI_PARSE_OK) { - printf("Problem parsing!\n"); - return; - } + mach->ImmLimit = 0; + labels->count = 0; declarations = (struct tgsi_full_declaration *) malloc(maxDeclarations * sizeof(struct tgsi_full_declaration)); @@ -88,13 +88,20 @@ expand_program(struct tgsi_exec_machine *mach ) instructions = (struct tgsi_full_instruction *) malloc(maxInstructions * sizeof(struct tgsi_full_instruction)); + k = tgsi_parse_init( &parse, mach->Tokens ); + if (k != TGSI_PARSE_OK) { + fprintf(stderr, "Problem parsing!\n"); + return; + } + while( !tgsi_parse_end_of_tokens( &parse ) ) { + GLuint pointer = parse.Position; + GLuint i; + tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: - /* - exec_declaration( mach, &parse.FullToken.FullDeclaration ); - */ + /* save expanded declaration */ if (numDeclarations == maxDeclarations) { maxDeclarations += 10; declarations = realloc(declarations, @@ -106,9 +113,25 @@ expand_program(struct tgsi_exec_machine *mach ) sizeof(declarations[0])); numDeclarations++; break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + assert( (parse.FullToken.FullImmediate.Immediate.Size - 1) % 4 == 0 ); + assert( mach->ImmLimit + (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4 <= 256 ); + + for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size - 1; i++ ) { + mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } + mach->ImmLimit += (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4; break; + case TGSI_TOKEN_TYPE_INSTRUCTION: + assert( labels->count < 128 ); + + labels->labels[labels->count][0] = instno; + labels->labels[labels->count][1] = pointer; + labels->count++; + + /* save expanded instruction */ if (numInstructions == maxInstructions) { maxInstructions += 10; instructions = realloc(instructions, @@ -120,17 +143,24 @@ expand_program(struct tgsi_exec_machine *mach ) sizeof(instructions[0])); numInstructions++; break; + default: assert( 0 ); } } tgsi_parse_free (&parse); - assert(!mach->Instructions); - mach->Instructions = instructions; - mach->NumInstructions = numInstructions; + if (mach->Declarations) { + free(mach->Declarations); + } mach->Declarations = declarations; mach->NumDeclarations = numDeclarations; + + if (mach->Instructions) { + free(mach->Instructions); + } + mach->Instructions = instructions; + mach->NumInstructions = numInstructions; } @@ -177,77 +207,7 @@ tgsi_exec_machine_init( mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; } - if (mach->Declarations) { - free(mach->Declarations); - mach->Declarations = NULL; - mach->NumDeclarations = 0; - } - if (mach->Instructions) { - free(mach->Instructions); - mach->Instructions = NULL; - mach->NumInstructions = 0; - } - - mach->CondMask = 0xf; - mach->LoopMask = 0xf; - mach->ExecMask = 0xf; - -#if 01 tgsi_exec_prepare( mach ); - expand_program(mach); -#endif -} - -void -tgsi_exec_prepare( - struct tgsi_exec_machine *mach ) -{ - struct tgsi_exec_labels *labels = &mach->Labels; - struct tgsi_parse_context parse; - GLuint k; - GLuint instno = 0; - - mach->ImmLimit = 0; - labels->count = 0; - - k = tgsi_parse_init( &parse, mach->Tokens ); - if (k != TGSI_PARSE_OK) { - printf("Problem parsing!\n"); - return; - } - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - GLuint pointer = parse.Position; - GLuint i; - - tgsi_parse_token( &parse ); - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - assert( (parse.FullToken.FullImmediate.Immediate.Size - 1) % 4 == 0 ); - assert( mach->ImmLimit + (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4 <= 256 ); - - for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size - 1; i++ ) { - mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; - } - mach->ImmLimit += (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4; - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - assert( labels->count < 128 ); - - labels->labels[labels->count][0] = instno; - labels->labels[labels->count][1] = pointer; - labels->count++; - break; - - default: - assert( 0 ); - } - } - tgsi_parse_free (&parse); } @@ -2258,6 +2218,10 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; #else + mach->CondMask = 0xf; + mach->LoopMask = 0xf; + mach->ExecMask = 0xf; + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index cf298b7c65..0f2dd7d987 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -166,10 +166,6 @@ tgsi_exec_machine_init( unsigned numSamplers, struct tgsi_sampler *samplers); -void -tgsi_exec_prepare( - struct tgsi_exec_machine *mach ); - void tgsi_exec_machine_run( struct tgsi_exec_machine *mach ); -- cgit v1.2.3 From 5e49ec339df1d23b1f1790c49c9f102098f42c0d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 13:49:38 -0600 Subject: Added TGSI_OPCODE_END Halt program execution when we get to END instruction. The GLSL compiler puts subroutines after the end instruction so we have to stop before then. --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 4 ++++ src/mesa/pipe/tgsi/exec/tgsi_dump.c | 5 ++++- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 12 +++++++---- src/mesa/pipe/tgsi/exec/tgsi_token.h | 4 +++- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 29 +++++++-------------------- 5 files changed, 26 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 24f21f6633..24bb1d3e2a 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -486,6 +486,10 @@ i915_translate_instruction(struct i915_fp_compile *p, swizzle(src1, ONE, Y, ONE, W), 0); break; + case TGSI_OPCODE_END: + /* no-op */ + break; + case TGSI_OPCODE_EX2: src0 = src_vector(p, &inst->FullSrcRegisters[0]); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index 4cf3397162..0b273cd6e5 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -363,7 +363,8 @@ static const char *TGSI_OPCODES[] = "OPCODE_CALLNZ", "OPCODE_IFC", "OPCODE_BREAKC", - "OPCODE_TXP" + "OPCODE_TXP", + "OPCODE_END" }; static const char *TGSI_OPCODES_SHORT[] = @@ -500,6 +501,8 @@ static const char *TGSI_OPCODES_SHORT[] = "CALLNZ", "IFC", "BREAKC", + "TXP", + "END" }; static const char *TGSI_SATS[] = diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index ca397bde6a..27154d6883 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -2029,6 +2029,11 @@ exec_instruction( } break; + case TGSI_OPCODE_END: + /* halt execution */ + *pc = -1; + break; + case TGSI_OPCODE_ENDIF: /* pop CondMask */ assert(mach->CondStackTop > 0); @@ -2233,15 +2238,14 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) { uint i; - int pc; + int pc = 0; for (i = 0; i < mach->NumDeclarations; i++) { exec_declaration( mach, mach->Declarations+i ); } - pc = 0; - - while (pc != 99 && pc < mach->NumInstructions) { + while (pc != -1) { + assert(pc < mach->NumInstructions); exec_instruction( mach, mach->Instructions + pc, &pc ); } } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h index 1d99a50dde..d2fa813815 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -1100,7 +1100,9 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MOVA */ /* TGSI_OPCODE_LOGP */ -#define TGSI_OPCODE_LAST 133 +#define TGSI_OPCODE_END 133 /* aka HALT */ + +#define TGSI_OPCODE_LAST 134 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 5daf50ddef..5c987436be 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -154,7 +154,7 @@ convert_writemask( return writemask; } -static GLboolean +static void compile_instruction( const struct prog_instruction *inst, struct tgsi_full_instruction *fullinst, @@ -460,12 +460,11 @@ compile_instruction( fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; break; case OPCODE_END: - return GL_TRUE; + fullinst->Instruction.Opcode = TGSI_OPCODE_END; + break; default: assert( 0 ); } - - return GL_FALSE; } static struct tgsi_full_declaration @@ -658,20 +657,13 @@ tgsi_mesa_compile_fp_program( #endif for( i = 0; i < program->Base.NumInstructions; i++ ) { - if( compile_instruction( + compile_instruction( &program->Base.Instructions[i], &fullinst, inputMapping, outputMapping, preamble_size, - TGSI_PROCESSOR_FRAGMENT ) ) { - assert( i == program->Base.NumInstructions - 1 ); - - if( TGSI_DEBUG ) { - tgsi_dump( tokens, 0 ); - } - break; - } + TGSI_PROCESSOR_FRAGMENT ); ti += tgsi_build_full_instruction( &fullinst, @@ -741,20 +733,13 @@ tgsi_mesa_compile_vp_program( for( i = 0; i < program->Base.NumInstructions; i++ ) { - if( compile_instruction( + compile_instruction( &program->Base.Instructions[i], &fullinst, inputMapping, outputMapping, 0, - TGSI_PROCESSOR_VERTEX ) ) { - assert( i == program->Base.NumInstructions - 1 ); - - if( TGSI_DEBUG ) { - tgsi_dump( tokens, 0 ); - } - break; - } + TGSI_PROCESSOR_VERTEX ); ti += tgsi_build_full_instruction( &fullinst, -- cgit v1.2.3 From 4ad80ad5cfdb9bc23eee15938645a334d227cd72 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 14:05:21 -0600 Subject: Implement CAL/RET and a call stack for subroutines. The glsl/brick.c shader demo runs now. --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 48 ++++++++++++++++++---------------- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 4 +++ src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 7 +++++ 3 files changed, 37 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 27154d6883..d757df48d8 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1903,12 +1903,18 @@ exec_instruction( break; case TGSI_OPCODE_CAL: - assert (0); + /* note that PC was already incremented above */ + mach->CallStack[mach->CallStackTop++] = *pc; + *pc = inst->InstructionExtLabel.Label; break; case TGSI_OPCODE_RET: - /* XXX: end of shader! */ - /*assert (0);*/ + assert(mach->CallStackTop >= 0); + if (mach->CallStackTop == 0) { + /* XXX error? */ + return; + } + *pc = mach->CallStack[--mach->CallStackTop]; break; case TGSI_OPCODE_SSG: @@ -2041,13 +2047,6 @@ exec_instruction( UPDATE_EXEC_MASK(mach); break; - case TGSI_OPCODE_ENDLOOP: - /* pop LoopMask */ - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - UPDATE_EXEC_MASK(mach); - break; - case TGSI_OPCODE_ENDREP: assert (0); break; @@ -2171,9 +2170,11 @@ exec_instruction( break; case TGSI_OPCODE_BGNSUB: - assert( 0 ); + /* no-op */ break; + case TGSI_OPCODE_ENDLOOP: + /* fall-through (for now at least) */ case TGSI_OPCODE_ENDLOOP2: if (mach->LoopMask) { /* repeat loop: jump to instruction just past BGNLOOP */ @@ -2219,6 +2220,9 @@ exec_instruction( void tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) { + uint i; + int pc = 0; + #if XXX_SSE mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; #else @@ -2226,6 +2230,9 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) mach->CondMask = 0xf; mach->LoopMask = 0xf; mach->ExecMask = 0xf; + assert(mach->CondStackTop == 0); + assert(mach->LoopStackTop == 0); + assert(mach->CallStackTop == 0); mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; @@ -2236,18 +2243,15 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) } - { - uint i; - int pc = 0; - - for (i = 0; i < mach->NumDeclarations; i++) { - exec_declaration( mach, mach->Declarations+i ); - } + /* execute declarations (interpolants) */ + for (i = 0; i < mach->NumDeclarations; i++) { + exec_declaration( mach, mach->Declarations+i ); + } - while (pc != -1) { - assert(pc < mach->NumInstructions); - exec_instruction( mach, mach->Instructions + pc, &pc ); - } + /* execute instructions, until pc is set to -1 */ + while (pc != -1) { + assert(pc < mach->NumInstructions); + exec_instruction( mach, mach->Instructions + pc, &pc ); } #endif diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 0f2dd7d987..ae5ce8a7c9 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -99,6 +99,7 @@ struct tgsi_exec_labels #define TGSI_EXEC_MAX_COND_NESTING 10 #define TGSI_EXEC_MAX_LOOP_NESTING 10 +#define TGSI_EXEC_MAX_CALL_NESTING 10 /** @@ -149,6 +150,9 @@ struct tgsi_exec_machine uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; int LoopStackTop; + uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; + int CallStackTop; + struct tgsi_full_instruction *Instructions; uint NumInstructions; diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 5c987436be..431b82a98f 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -271,6 +271,10 @@ compile_instruction( case OPCODE_BRK: fullinst->Instruction.Opcode = TGSI_OPCODE_BRK; break; + case OPCODE_CAL: + fullinst->Instruction.Opcode = TGSI_OPCODE_CAL; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; case OPCODE_CMP: fullinst->Instruction.Opcode = TGSI_OPCODE_CMP; break; @@ -382,6 +386,9 @@ compile_instruction( case OPCODE_RCP: fullinst->Instruction.Opcode = TGSI_OPCODE_RCP; break; + case OPCODE_RET: + fullinst->Instruction.Opcode = TGSI_OPCODE_RET; + break; case OPCODE_RSQ: fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ; tgsi_util_set_full_src_register_sign_mode( -- cgit v1.2.3 From 6aab3e3d269b71e8306928088e3603daf082223f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 14:17:47 -0600 Subject: Added TGSI_FILE_OUTPUT case in fetch_src_file_channel(). GLSL output variables (such as vertex varyings) can be read (they're not write-only). --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index d757df48d8..e428b250f0 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -855,6 +855,14 @@ fetch_src_file_channel( chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; break; + case TGSI_FILE_OUTPUT: + /* vertex varying/output vars can be read too */ + chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; + break; + default: assert( 0 ); } -- cgit v1.2.3 From 8955bc34581ac716cf5cabb02fde5260ab3ae207 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 14:25:43 -0600 Subject: Minor optimization: no-op CAL when ExecMask==0x0. --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index e428b250f0..5a50bd5aed 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1911,9 +1911,12 @@ exec_instruction( break; case TGSI_OPCODE_CAL: - /* note that PC was already incremented above */ - mach->CallStack[mach->CallStackTop++] = *pc; - *pc = inst->InstructionExtLabel.Label; + /* skip the call if no execution channels are enabled */ + if (mach->ExecMask) { + /* note that PC was already incremented above */ + mach->CallStack[mach->CallStackTop++] = *pc; + *pc = inst->InstructionExtLabel.Label; + } break; case TGSI_OPCODE_RET: -- cgit v1.2.3 From fe1d15acc7b0ed5d6eb22829f2d8547a36a852a9 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 15:17:37 -0600 Subject: added micro_trunc(), re-order some code --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 79 ++++++++++++++++++++++--------------- 1 file changed, 47 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 5a50bd5aed..4a42238636 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -179,6 +179,10 @@ tgsi_exec_machine_init( GLuint i, k; struct tgsi_parse_context parse; +#if 0 + tgsi_dump(tokens, 0); +#endif + mach->Tokens = tokens; mach->Samplers = samplers; @@ -734,6 +738,17 @@ micro_ishr( dst->i[3] = src0->i[3] >> src1->i[3]; } +static void +micro_trunc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0 ) +{ + dst->f[0] = (float) (int) src0->u[0]; + dst->f[1] = (float) (int) src0->u[1]; + dst->f[2] = (float) (int) src0->u[2]; + dst->f[3] = (float) (int) src0->u[3]; +} + static void micro_ushr( union tgsi_exec_channel *dst, @@ -1995,13 +2010,6 @@ exec_instruction( assert (0); break; - case TGSI_OPCODE_BRK: - /* turn off loop channels for each enabled exec channel */ - mach->LoopMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); - break; - case TGSI_OPCODE_IF: /* push CondMask */ assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); @@ -2024,16 +2032,6 @@ exec_instruction( /* Todo: If CondMask==0, jump to ELSE */ break; - case TGSI_OPCODE_LOOP: - /* push LoopMask */ - assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; - break; - - case TGSI_OPCODE_REP: - assert (0); - break; - case TGSI_OPCODE_ELSE: /* invert CondMask wrt previous mask */ { @@ -2046,11 +2044,6 @@ exec_instruction( } break; - case TGSI_OPCODE_END: - /* halt execution */ - *pc = -1; - break; - case TGSI_OPCODE_ENDIF: /* pop CondMask */ assert(mach->CondStackTop > 0); @@ -2058,6 +2051,15 @@ exec_instruction( UPDATE_EXEC_MASK(mach); break; + case TGSI_OPCODE_END: + /* halt execution */ + *pc = -1; + break; + + case TGSI_OPCODE_REP: + assert (0); + break; + case TGSI_OPCODE_ENDREP: assert (0); break; @@ -2095,8 +2097,11 @@ exec_instruction( break; case TGSI_OPCODE_TRUNC: - /* TGSI_OPCODE_INT */ - assert (0); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_trunc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } break; case TGSI_OPCODE_SHL: @@ -2160,10 +2165,6 @@ exec_instruction( assert (0); break; - case TGSI_OPCODE_CONT: - assert (0); - break; - case TGSI_OPCODE_EMIT: mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; @@ -2174,16 +2175,14 @@ exec_instruction( mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; break; + case TGSI_OPCODE_LOOP: + /* fall-through (for now) */ case TGSI_OPCODE_BGNLOOP2: /* push LoopMask */ assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; break; - case TGSI_OPCODE_BGNSUB: - /* no-op */ - break; - case TGSI_OPCODE_ENDLOOP: /* fall-through (for now at least) */ case TGSI_OPCODE_ENDLOOP2: @@ -2199,6 +2198,22 @@ exec_instruction( } break; + case TGSI_OPCODE_BRK: + /* turn off loop channels for each enabled exec channel */ + mach->LoopMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_CONT: + assert (0); + break; + + + case TGSI_OPCODE_BGNSUB: + /* no-op */ + break; + case TGSI_OPCODE_ENDSUB: assert( 0 ); break; -- cgit v1.2.3 From 53a6a55c7c56c6811a9e627c8624c05e83d4e04b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 16:05:07 -0600 Subject: Implement CONT statement. --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 25 +++++++++++++++++++------ src/mesa/pipe/tgsi/exec/tgsi_exec.h | 9 +++++++-- 2 files changed, 26 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 4a42238636..03720d2fa0 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -57,7 +57,7 @@ /** The execution mask depends on the conditional mask and the loop mask */ #define UPDATE_EXEC_MASK(MACH) \ - MACH->ExecMask = MACH->CondMask & MACH->LoopMask + MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask #define CHAN_X 0 @@ -2178,14 +2178,19 @@ exec_instruction( case TGSI_OPCODE_LOOP: /* fall-through (for now) */ case TGSI_OPCODE_BGNLOOP2: - /* push LoopMask */ + /* push LoopMask and ContMasks */ assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; break; case TGSI_OPCODE_ENDLOOP: /* fall-through (for now at least) */ case TGSI_OPCODE_ENDLOOP2: + /* Restore ContMask, but don't pop */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; if (mach->LoopMask) { /* repeat loop: jump to instruction just past BGNLOOP */ *pc = inst->InstructionExtLabel.Label + 1; @@ -2194,8 +2199,11 @@ exec_instruction( /* exit loop: pop LoopMask */ assert(mach->LoopStackTop > 0); mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - UPDATE_EXEC_MASK(mach); + /* pop ContMask */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; } + UPDATE_EXEC_MASK(mach); break; case TGSI_OPCODE_BRK: @@ -2206,16 +2214,18 @@ exec_instruction( break; case TGSI_OPCODE_CONT: - assert (0); + /* turn off cont channels for each enabled exec channel */ + mach->ContMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); break; - case TGSI_OPCODE_BGNSUB: /* no-op */ break; case TGSI_OPCODE_ENDSUB: - assert( 0 ); + /* no-op */ break; case TGSI_OPCODE_NOISE1: @@ -2255,9 +2265,12 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) mach->CondMask = 0xf; mach->LoopMask = 0xf; + mach->ContMask = 0xf; mach->ExecMask = 0xf; + assert(mach->CondStackTop == 0); assert(mach->LoopStackTop == 0); + assert(mach->ContStackTop == 0); assert(mach->CallStackTop == 0); mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index ae5ce8a7c9..e7827ecce1 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -138,8 +138,9 @@ struct tgsi_exec_machine const struct tgsi_interp_coef *InterpCoefs; /* Conditional execution masks */ - uint CondMask; - uint LoopMask; + uint CondMask; /**< For IF/ELSE/ENDIF */ + uint LoopMask; /**< For BGNLOOP/ENDLOOP */ + uint ContMask; /**< For loop CONT statements */ uint ExecMask; /**< = CondMask & LoopMask */ /** Condition mask stack (for nested conditionals) */ @@ -150,6 +151,10 @@ struct tgsi_exec_machine uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; int LoopStackTop; + /** Loop continue mask stack (see comments in tgsi_exec.c) */ + uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int ContStackTop; + uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; int CallStackTop; -- cgit v1.2.3 From 4d155a32d2c44af4d85715fcc982ad4bc18d38fa Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 16:17:34 -0600 Subject: added license and interpreter comments --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 52 +++++++++++++++++++++++++++++++++++++ 1 file changed, 52 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 03720d2fa0..60abecaef1 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1,3 +1,55 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI interpretor/executor. + * + * Flow control information: + * + * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) + * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special + * care since a condition may be true for some quad components but false + * for other components. + * + * We basically execute all statements (even if they're in the part of + * an IF/ELSE clause that's "not taken") and use a special mask to + * control writing to destination registers. This is the ExecMask. + * See store_dest(). + * + * The ExecMask is computed from three other masks (CondMask, LoopMask and + * ContMask) which are controlled by the flow control instructions (namely: + * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). + * + * + * Authors: + * Michael Krol + * Brian Paul + */ + #include "tgsi_platform.h" #include "tgsi_core.h" #include "pipe/p_state.h" -- cgit v1.2.3 From c9dceb17c0bf255252f58a4c947522b958fe015c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 16:24:40 -0600 Subject: Push mask stacks upon CAL, pop upon RET. Still need to handle conditional RET statements... --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 60abecaef1..edd46100c5 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1980,6 +1980,16 @@ exec_instruction( case TGSI_OPCODE_CAL: /* skip the call if no execution channels are enabled */ if (mach->ExecMask) { + /* do the call */ + + /* push Cond, Loop, Cont stacks */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + /* note that PC was already incremented above */ mach->CallStack[mach->CallStackTop++] = *pc; *pc = inst->InstructionExtLabel.Label; @@ -1987,6 +1997,15 @@ exec_instruction( break; case TGSI_OPCODE_RET: + /* XXX examine ExecMask to determine if we should _really_ return */ + /* pop Cond, Loop, Cont stacks */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->CallStackTop >= 0); if (mach->CallStackTop == 0) { /* XXX error? */ -- cgit v1.2.3 From d781cdc8fadc802a1f2edbeb13ccb1ee768ce803 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 16:55:21 -0600 Subject: Generate a texenv fragment program if there's a GLSL vertex shader but no GLSL fragment shader. This allows Glean glsl1 test to get pretty far. --- src/mesa/main/texenvprogram.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index cdf35b4636..a554c033c4 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -1247,7 +1247,8 @@ _mesa_UpdateTexEnvProgram( GLcontext *ctx ) /* If a conventional fragment program/shader isn't in effect... */ if (!ctx->FragmentProgram._Enabled && - !ctx->Shader.CurrentProgram) { + (!ctx->Shader.CurrentProgram || + !ctx->Shader.CurrentProgram->FragmentProgram) ) { make_state_key(ctx, &key); hash = hash_key(&key); -- cgit v1.2.3 From ae28d19bbf527e50be9ff9e52cb69f6afdf1d506 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 16:56:02 -0600 Subject: added MESA_FORMAT_Z16 in st_mesa_format_to_pipe_format(), for Glean --- src/mesa/state_tracker/st_format.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 84e6a2d5fd..8dd7df4495 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -123,6 +123,8 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) return PIPE_FORMAT_U_L8; case MESA_FORMAT_I8: return PIPE_FORMAT_U_I8; + case MESA_FORMAT_Z16: + return PIPE_FORMAT_U_Z16; default: assert(0); return 0; -- cgit v1.2.3 From ca34912bf7e57a01b180a3bb3b6e36331f442558 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 16:57:19 -0600 Subject: assert that program length > 1 (some code, plus END) --- src/mesa/state_tracker/st_atom_shader.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 8aa977179a..dc6210cd6c 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -172,6 +172,8 @@ find_translated_vp(struct st_context *st, stfp->num_input_slots = numIn; + assert(stfp->Base.Base.NumInstructions > 1); + (void) st_translate_fragment_program(st, stfp, stfp->input_to_slot, stfp->tokens, @@ -235,6 +237,7 @@ find_translated_vp(struct st_context *st, } } + assert(stvp->Base.Base.NumInstructions > 1); xvp->vs = st_translate_vertex_program(st, stvp, xvp->output_to_slot, -- cgit v1.2.3 From 066ccec49485f0b6d314ef555e4b2a68f68c3804 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 17:07:30 -0600 Subject: Separate TGSI_OPCODE_KIL and TGSI_OPCODE_KILP (predicated). These correspond to the NV and ARB-style fragment program KIL instructions. The former is really supposed to examine the NV condition codes but Mesa's GLSL compiler always emits unconditional KIL instructions. --- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 6 ++++-- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 18 +++++++++++++++++- src/mesa/pipe/tgsi/exec/tgsi_token.h | 7 ++++--- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 6 ++++++ 4 files changed, 31 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index 0b273cd6e5..c179659aee 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -270,7 +270,7 @@ static const char *TGSI_OPCODES[] = "OPCODE_COS", "OPCODE_DDX", "OPCODE_DDY", - "OPCODE_KIL", + "OPCODE_KILP", "OPCODE_PK2H", "OPCODE_PK2US", "OPCODE_PK4B", @@ -364,6 +364,7 @@ static const char *TGSI_OPCODES[] = "OPCODE_IFC", "OPCODE_BREAKC", "OPCODE_TXP", + "OPCODE_KIL", "OPCODE_END" }; @@ -408,7 +409,7 @@ static const char *TGSI_OPCODES_SHORT[] = "COS", "DDX", "DDY", - "KIL", + "KILP", "PK2H", "PK2US", "PK4B", @@ -502,6 +503,7 @@ static const char *TGSI_OPCODES_SHORT[] = "IFC", "BREAKC", "TXP", + "KIL", "END" }; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index edd46100c5..b88620d71d 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1129,8 +1129,12 @@ store_dest( store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) +/** + * Execute ARB-style KIL which is predicated by a src register. + * Kill fragment if any of the four values is less than zero. + */ static void -exec_kil (struct tgsi_exec_machine *mach, +exec_kilp(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst) { GLuint uniquemask; @@ -1168,6 +1172,14 @@ exec_kil (struct tgsi_exec_machine *mach, } +static void +exec_kil(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + /* for enabled ExecMask bits, set the killed bit */ + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; +} + /* @@ -1782,6 +1794,10 @@ exec_instruction( } break; + case TGSI_OPCODE_KILP: + exec_kilp (mach, inst); + break; + case TGSI_OPCODE_KIL: exec_kil (mach, inst); break; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h index d2fa813815..8d5992facb 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -236,7 +236,7 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_EX2 TGSI_OPCODE_EXPBASE2 #define TGSI_OPCODE_FLR TGSI_OPCODE_FLOOR #define TGSI_OPCODE_FRC TGSI_OPCODE_FRAC -#define TGSI_OPCODE_KIL 39 +#define TGSI_OPCODE_KILP 39 /* predicated kill */ #define TGSI_OPCODE_LG2 TGSI_OPCODE_LOGBASE2 /* TGSI_OPCODE_LIT */ #define TGSI_OPCODE_LRP TGSI_OPCODE_LERP @@ -1100,9 +1100,10 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MOVA */ /* TGSI_OPCODE_LOGP */ -#define TGSI_OPCODE_END 133 /* aka HALT */ +#define TGSI_OPCODE_KIL 133 /* unpredicated kill */ +#define TGSI_OPCODE_END 134 /* aka HALT */ -#define TGSI_OPCODE_LAST 134 +#define TGSI_OPCODE_LAST 135 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 431b82a98f..2a2a1a0430 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -336,6 +336,12 @@ compile_instruction( fullinst->Instruction.Opcode = TGSI_OPCODE_INT; break; case OPCODE_KIL: + /* predicated w/ a register */ + fullinst->Instruction.Opcode = TGSI_OPCODE_KILP; + break; + case OPCODE_KIL_NV: + /* unpredicated */ + assert(inst->DstReg.CondMask == COND_TR); fullinst->Instruction.Opcode = TGSI_OPCODE_KIL; break; case OPCODE_LG2: -- cgit v1.2.3 From ff9949d05caa5543f4e57d831e603cba52f62e8b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 17:13:06 -0600 Subject: Implement TGSI_OPCODE_SNE with micro_ne() --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 21 ++++++++++++++++++++- 1 file changed, 20 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index b88620d71d..158abba270 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -698,6 +698,20 @@ micro_movc( dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; } +static void +micro_ne( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] != src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] != src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] != src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] != src1->f[3] ? src2->f[3] : src3->f[3]; +} + static void micro_neg( union tgsi_exec_channel *dst, @@ -1862,7 +1876,12 @@ exec_instruction( break; case TGSI_OPCODE_SNE: - assert (0); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ne( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } break; case TGSI_OPCODE_STR: -- cgit v1.2.3 From 355f8f7eedf9ddfac7edd2244f09c5a47fd8af86 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 19:16:57 -0600 Subject: Implement/use fragment kill results --- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 14 +++++++------- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 2 +- 3 files changed, 9 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 57c01dcfcc..f394c587dc 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -113,7 +113,7 @@ shade_quad( machine->InterpCoefs ); } else { - tgsi_exec_machine_run( machine ); + quad->mask &= tgsi_exec_machine_run( machine ); } /* store result color (always in output[1]) */ diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 158abba270..a0b3bc1c8e 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -2359,16 +2359,16 @@ exec_instruction( } -void +/** + * Run TGSI interpreter. + * \return bitmask of "alive" quad components + */ +uint tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) { uint i; int pc = 0; -#if XXX_SSE - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; -#else - mach->CondMask = 0xf; mach->LoopMask = 0xf; mach->ContMask = 0xf; @@ -2399,8 +2399,6 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) exec_instruction( mach, mach->Instructions + pc, &pc ); } -#endif - #if 0 /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { @@ -2411,6 +2409,8 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; } #endif + + return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index e7827ecce1..354d5df6c2 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -175,7 +175,7 @@ tgsi_exec_machine_init( unsigned numSamplers, struct tgsi_sampler *samplers); -void +uint tgsi_exec_machine_run( struct tgsi_exec_machine *mach ); -- cgit v1.2.3 From 4b6cc36b2b4892c3ce0862789c3b294c52356805 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 2 Oct 2007 19:44:55 -0600 Subject: clean-ups --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 33 ++++++++++++--------------------- 1 file changed, 12 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index a0b3bc1c8e..3f170ffcfc 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1089,7 +1089,8 @@ store_dest( return; case TGSI_FILE_OUTPUT: - dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + reg->DstRegister.Index].xyzw[chan_index]; + dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + + reg->DstRegister.Index].xyzw[chan_index]; break; case TGSI_FILE_TEMPORARY: @@ -1108,9 +1109,6 @@ store_dest( switch (inst->Instruction.Saturate) { case TGSI_SAT_NONE: -#if 0 - *dst = *chan; -#else if (mach->ExecMask & 0x1) dst->i[0] = chan->i[0]; if (mach->ExecMask & 0x2) @@ -1119,12 +1117,14 @@ store_dest( dst->i[2] = chan->i[2]; if (mach->ExecMask & 0x8) dst->i[3] = chan->i[3]; -#endif break; case TGSI_SAT_ZERO_ONE: - micro_lt( dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], chan ); - micro_lt( dst, chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + /* XXX need to obey ExecMask here */ + micro_lt( dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], + &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], chan ); + micro_lt( dst, chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], + chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); break; case TGSI_SAT_MINUS_PLUS_ONE: @@ -1186,16 +1186,6 @@ exec_kilp(struct tgsi_exec_machine *mach, } -static void -exec_kil(struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst) -{ - /* for enabled ExecMask bits, set the killed bit */ - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; -} - - - /* * Fetch a texel using STR texture coordinates. */ @@ -1809,12 +1799,13 @@ exec_instruction( break; case TGSI_OPCODE_KILP: - exec_kilp (mach, inst); - break; + exec_kilp (mach, inst); + break; case TGSI_OPCODE_KIL: - exec_kil (mach, inst); - break; + /* for enabled ExecMask bits, set the killed bit */ + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; + break; case TGSI_OPCODE_PK2H: assert (0); -- cgit v1.2.3 From 51345cb3c4d85a9e88ac35b59e938b0692df6205 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 3 Oct 2007 08:47:36 -0400 Subject: Make softpipe behave more like a real driver by always allocating something in the state functions. --- src/mesa/pipe/softpipe/sp_state_blend.c | 23 ++++++++++++++--------- src/mesa/pipe/softpipe/sp_state_fs.c | 5 ++++- src/mesa/pipe/softpipe/sp_state_rasterizer.c | 7 +++++-- src/mesa/pipe/softpipe/sp_state_sampler.c | 6 ++++-- 4 files changed, 27 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index cb0921f687..6376c3c61c 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -34,9 +34,9 @@ void * softpipe_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - /* means that we just want pipe_blend_state and don't have - * anything specific */ - return 0; + struct pipe_blend_state *state = malloc(sizeof(struct pipe_blend_state)); + memcpy(state, blend, sizeof(struct pipe_blend_state)); + return state; } void softpipe_bind_blend_state( struct pipe_context *pipe, @@ -50,9 +50,9 @@ void softpipe_bind_blend_state( struct pipe_context *pipe, } void softpipe_delete_blend_state(struct pipe_context *pipe, - void *blend ) + void *blend) { - /* do nothing */ + free(blend); } @@ -75,7 +75,9 @@ void * softpipe_create_alpha_test_state(struct pipe_context *pipe, const struct pipe_alpha_test_state *alpha) { - return 0; + struct pipe_alpha_test_state *state = malloc(sizeof(struct pipe_alpha_test_state)); + memcpy(state, alpha, sizeof(struct pipe_alpha_test_state)); + return state; } void @@ -93,14 +95,17 @@ void softpipe_delete_alpha_test_state(struct pipe_context *pipe, void *alpha) { - /* do nothing */ + free(alpha); } void * softpipe_create_depth_stencil_state(struct pipe_context *pipe, const struct pipe_depth_stencil_state *depth_stencil) { - return 0; + struct pipe_depth_stencil_state *state = + malloc(sizeof(struct pipe_depth_stencil_state)); + memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_state)); + return state; } void @@ -117,5 +122,5 @@ softpipe_bind_depth_stencil_state(struct pipe_context *pipe, void softpipe_delete_depth_stencil_state(struct pipe_context *pipe, void *depth) { - /* do nothing */ + free(depth); } diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index f1bec2c73a..390340fa8d 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -40,7 +40,9 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, * that now. */ - return 0; + struct pipe_shader_state *state = malloc(sizeof(struct pipe_shader_state)); + memcpy(state, templ, sizeof(struct pipe_shader_state)); + return state; } void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) @@ -55,6 +57,7 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) void softpipe_delete_fs_state(struct pipe_context *pipe, void *shader) { + free(shader); } diff --git a/src/mesa/pipe/softpipe/sp_state_rasterizer.c b/src/mesa/pipe/softpipe/sp_state_rasterizer.c index d7845cef82..076ea66d9e 100644 --- a/src/mesa/pipe/softpipe/sp_state_rasterizer.c +++ b/src/mesa/pipe/softpipe/sp_state_rasterizer.c @@ -36,7 +36,10 @@ void * softpipe_create_rasterizer_state(struct pipe_context *pipe, const struct pipe_rasterizer_state *setup) { - return 0; + struct pipe_rasterizer_state *state = + malloc(sizeof(struct pipe_rasterizer_state)); + memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); + return state; } void softpipe_bind_rasterizer_state(struct pipe_context *pipe, @@ -55,7 +58,7 @@ void softpipe_bind_rasterizer_state(struct pipe_context *pipe, void softpipe_delete_rasterizer_state(struct pipe_context *pipe, void *rasterizer) { - /* do nothing */ + free(rasterizer); } diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index ad98375735..4084e9163f 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -36,7 +36,9 @@ void * softpipe_create_sampler_state(struct pipe_context *pipe, const struct pipe_sampler_state *sampler) { - return 0; + struct pipe_sampler_state *state = malloc(sizeof(struct pipe_sampler_state)); + memcpy(state, sampler, sizeof(struct pipe_sampler_state)); + return state; } void @@ -56,7 +58,7 @@ void softpipe_delete_sampler_state(struct pipe_context *pipe, void *sampler) { - /* do nothing */ + free(sampler); } -- cgit v1.2.3 From 5e4309ee92560caed70ff6975c4a59aaa5d86542 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 3 Oct 2007 08:50:27 -0400 Subject: Explain a little better what we're doing here. --- src/mesa/state_tracker/st_cache.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 1686721990..c0f712ba1d 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -39,10 +39,17 @@ #include "pipe/cso_cache/cso_cache.h" #include "pipe/cso_cache/cso_hash.h" -/* This function will either find the state of the given template - * in the cache or it will create a new state state from the given - * template, will insert it in the cache and return it. + +/* Those function will either find the state of the given template + * in the cache or they will create a new state from the given + * template, insert it in the cache and return it. + */ + +/* + * If the driver returns 0 from the create method then they will assign + * the data member of the cso to be the template itself. */ + const struct cso_blend * st_cached_blend_state(struct st_context *st, const struct pipe_blend_state *templ) { -- cgit v1.2.3 From a7e997cfc5d909eebbc16ba5b0362e9778f01561 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 1 Oct 2007 06:25:23 -0400 Subject: Get rid of ST_NEW_SHADER flag and the dependency on _NEW_PROGRAM which is being hit all the time. Done by Keith really. --- src/mesa/state_tracker/st_atom_shader.c | 4 ++-- src/mesa/state_tracker/st_cb_program.c | 26 ++++++++++++++++---------- src/mesa/state_tracker/st_context.h | 1 - 3 files changed, 18 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index dc6210cd6c..919461bcb8 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -306,8 +306,8 @@ update_linkage( struct st_context *st ) const struct st_tracked_state st_update_shader = { .name = "st_update_shader", .dirty = { - .mesa = _NEW_PROGRAM, - .st = ST_NEW_SHADER + .mesa = 0, + .st = ST_NEW_VERTEX_PROGRAM | ST_NEW_FRAGMENT_PROGRAM }, .update = update_linkage }; diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 8df8b263d3..1902c8d7d6 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -43,7 +43,6 @@ #include "st_program.h" #include "st_atom_shader.h" -#include "tnl/tnl.h" #include "pipe/tgsi/mesa/tgsi_mesa.h" @@ -57,7 +56,14 @@ static void st_bind_program( GLcontext *ctx, { struct st_context *st = st_context(ctx); - st->dirty.st |= ST_NEW_SHADER; + switch (target) { + case GL_VERTEX_PROGRAM_ARB: + st->dirty.st |= ST_NEW_VERTEX_PROGRAM; + break; + case GL_FRAGMENT_PROGRAM_ARB: + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; + break; + } } @@ -70,7 +76,8 @@ static void st_use_program( GLcontext *ctx, { struct st_context *st = st_context(ctx); - st->dirty.st |= ST_NEW_SHADER; + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; + st->dirty.st |= ST_NEW_VERTEX_PROGRAM; _mesa_use_program(ctx, program); } @@ -114,8 +121,6 @@ static struct gl_program *st_new_program( GLcontext *ctx, default: return _mesa_new_program(ctx, target, id); } - - st->dirty.st |= ST_NEW_SHADER; } @@ -175,6 +180,9 @@ static void st_program_string_notify( GLcontext *ctx, } stfp->param_state = stfp->Base.Base.Parameters->StateFlags; + + if (st->fp == stfp) + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; } else if (target == GL_VERTEX_PROGRAM_ARB) { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; @@ -189,12 +197,10 @@ static void st_program_string_notify( GLcontext *ctx, stvp->param_state = stvp->Base.Base.Parameters->StateFlags; - /* Also tell tnl about it: - */ - _tnl_program_string(ctx, target, prog); - } + if (st->vp == stvp) + st->dirty.st |= ST_NEW_VERTEX_PROGRAM; - st->dirty.st |= ST_NEW_SHADER; + } } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index da4fd0d4e4..b4ae041d36 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -45,7 +45,6 @@ struct cso_blend; #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 #define ST_NEW_VERTEX_PROGRAM 0x4 -#define ST_NEW_SHADER 0x8 struct st_state_flags { -- cgit v1.2.3 From cdd38d487a311e6c71b76382d428f5dc26caf067 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 3 Oct 2007 10:08:45 -0400 Subject: Unify handling of userplanes and regular planes to simplify the clipping code. (really done by Keith) --- src/mesa/pipe/draw/draw_clip.c | 52 ++++----------------------------- src/mesa/pipe/draw/draw_context.c | 2 -- src/mesa/pipe/draw/draw_private.h | 1 - src/mesa/pipe/draw/draw_vertex_shader.c | 33 +++++++++++++++------ 4 files changed, 30 insertions(+), 58 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index a505146047..4e1483f025 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -50,13 +50,6 @@ #endif -/** bitmask for the first view-frustum clip planes */ -#define VIEWPLANE_MASK ((1 << 6) - 1) - -/** bitmask for the user-defined clip planes */ -#define USERPLANE_MASK (((1 << PIPE_MAX_CLIP_PLANES) - 1) << 6) - - struct clipper { struct draw_stage stage; /**< base class */ @@ -335,10 +328,6 @@ do_clip_line( struct draw_stage *stage, const float dp0 = dot4( pos0, plane ); const float dp1 = dot4( pos1, plane ); - /* need this to handle user-clip planes properly */ - if (dp0 < 0.0F && dp1 < 0.0F) - return; - if (dp1 < 0.0F) { float t = dp1 / (dp1 - dp0); t1 = MAX2(t1, t); @@ -392,21 +381,8 @@ static void clip_point( struct draw_stage *stage, struct prim_header *header ) { - if (header->v[0]->clipmask == 0) { + if (header->v[0]->clipmask == 0) stage->next->point( stage->next, header ); - } - else if (header->v[0]->clipmask & USERPLANE_MASK) { - /* test against user clip planes now */ - const struct clipper *clipper = clipper_stage( stage ); - uint i; - for (i = 6; i < stage->draw->nr_planes; i++) { - float dot = dot4(clipper->plane[i], header->v[0]->clip); - if (dot < 0.0F) - return; /* clipped! */ - } - /* not clipped */ - stage->next->point( stage->next, header ); - } } @@ -421,22 +397,8 @@ clip_line( struct draw_stage *stage, /* no clipping needed */ stage->next->line( stage->next, header ); } - else if (((header->v[0]->clipmask & - header->v[1]->clipmask & - VIEWPLANE_MASK) == 0) || - (clipmask & USERPLANE_MASK)) { - /* About the above predicate: the clipmask bits for the view volume - * exactly indicate whether the coordinate is inside or outside each - * frustum plane. However, the bits for user-defined planes are set - * if the plane is enabled, and does not really indicate if the - * coordinate is inside or outside the user-defined plane. - * - * To change this (so that the user-plane bits really indicate - * inside/outside) we'd have to compute the dot products earlier - * in the draw_prim.c code (see compute_clipmask()). - * We will probably do that when we have support for user clip coord - * in vertex shaders... - */ + else if ((header->v[0]->clipmask & + header->v[1]->clipmask) == 0) { do_clip_line(stage, header, clipmask); } /* else, totally clipped */ @@ -455,11 +417,9 @@ clip_tri( struct draw_stage *stage, /* no clipping needed */ stage->next->tri( stage->next, header ); } - else if (((header->v[0]->clipmask & - header->v[1]->clipmask & - header->v[2]->clipmask & - VIEWPLANE_MASK) == 0) || - (clipmask & USERPLANE_MASK)) { + else if ((header->v[0]->clipmask & + header->v[1]->clipmask & + header->v[2]->clipmask) == 0) { do_clip_tri(stage, header, clipmask); } } diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 3fb667ab1a..44e770f364 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -145,8 +145,6 @@ void draw_set_clip_state( struct draw_context *draw, assert(clip->nr <= PIPE_MAX_CLIP_PLANES); memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); draw->nr_planes = 6 + clip->nr; - /* bitmask of the enabled user-defined clip planes */ - draw->user_clipmask = ((1 << clip->nr) - 1) << 6; } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index a54fef41e7..ff38925fc0 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -177,7 +177,6 @@ struct draw_context */ float plane[12][4]; unsigned nr_planes; - unsigned user_clipmask; /** Describes the layout of post-transformation vertices */ struct vertex_info vertex_info; diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 6e8036825d..d17496a24f 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -40,17 +40,27 @@ #include "pipe/tgsi/exec/tgsi_core.h" + +static INLINE float dot4(const float *a, const float *b) +{ + float result = (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); + + return result; +} + static INLINE unsigned -compute_clipmask(float cx, float cy, float cz, float cw) +compute_clipmask(const float *clip, const float (*plane)[4], unsigned nr) { unsigned mask = 0; + unsigned i; - if (-cx + cw < 0) mask |= CLIP_RIGHT_BIT; - if ( cx + cw < 0) mask |= CLIP_LEFT_BIT; - if (-cy + cw < 0) mask |= CLIP_TOP_BIT; - if ( cy + cw < 0) mask |= CLIP_BOTTOM_BIT; - if (-cz + cw < 0) mask |= CLIP_FAR_BIT; - if ( cz + cw < 0) mask |= CLIP_NEAR_BIT; + for (i = 0; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<clip[0] = machine->Outputs[0].xyzw[0].f[j]; y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - vOut[j]->clipmask = compute_clipmask(x, y, z, w) | draw->user_clipmask; + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); vOut[j]->edgeflag = 1; /* divide by w */ -- cgit v1.2.3 From b46926c4beff4ac14ac0cf2e3bc52f8118ed4d2a Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 3 Oct 2007 10:31:42 -0400 Subject: Unify the definitions of the 4 component dot product into one location. --- src/mesa/pipe/draw/draw_clip.c | 14 -------------- src/mesa/pipe/draw/draw_private.h | 10 ++++++++++ src/mesa/pipe/draw/draw_vertex_shader.c | 11 ----------- 3 files changed, 10 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 4e1483f025..222022d6c1 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -138,20 +138,6 @@ static void interp( const struct clipper *clip, } } - - -static INLINE float dot4( const float *a, - const float *b ) -{ - float result = (a[0]*b[0] + - a[1]*b[1] + - a[2]*b[2] + - a[3]*b[3]); - - return result; -} - - #if 0 static INLINE void do_tri( struct draw_stage *next, struct prim_header *header ) diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index ff38925fc0..b3f1c4d23e 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -290,5 +290,15 @@ dup_vert( struct draw_stage *stage, return tmp; } +static INLINE float +dot4(const float *a, const float *b) +{ + float result = (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); + + return result; +} #endif /* DRAW_PRIVATE_H */ diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index d17496a24f..49ff3b32ff 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -40,17 +40,6 @@ #include "pipe/tgsi/exec/tgsi_core.h" - -static INLINE float dot4(const float *a, const float *b) -{ - float result = (a[0]*b[0] + - a[1]*b[1] + - a[2]*b[2] + - a[3]*b[3]); - - return result; -} - static INLINE unsigned compute_clipmask(const float *clip, const float (*plane)[4], unsigned nr) { -- cgit v1.2.3 From 07cd46d111b063ed37391ad7c316db9d7927ad03 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 3 Oct 2007 12:00:52 -0400 Subject: In vertex shaders also expect the template to be stack allocated structure so allocate a private copy of it. --- src/mesa/pipe/softpipe/sp_state_fs.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 390340fa8d..88d9bd9716 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -67,8 +67,11 @@ void * softpipe_create_vs_state(struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_vertex_shader_state *state = malloc(sizeof(struct sp_vertex_shader_state)); + struct pipe_shader_state *templ_copy = + malloc(sizeof(struct pipe_shader_state)); + memcpy(templ_copy, templ, sizeof(struct pipe_shader_state)); - state->state = templ; + state->state = templ_copy; state->draw_data = draw_create_vertex_shader(softpipe->draw, state->state); @@ -95,6 +98,7 @@ void softpipe_delete_vs_state(struct pipe_context *pipe, (struct sp_vertex_shader_state *)vs; draw_delete_vertex_shader(softpipe->draw, state->draw_data); + free(state->state); free(state); } -- cgit v1.2.3 From ce765a7fb77e12ff083a9068ec232a15bcf41f66 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 2 Oct 2007 13:01:07 +0200 Subject: intel_winsys: Adapt to DRM changes (again). It was decided after all to stick to 'pipes' here, even though the actual meaning is now 'planes'. --- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 48c1f68d9c..a6d0d43e4d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -46,7 +46,7 @@ #define DRM_VBLANK_FLIP 0x8000000 typedef struct drm_i915_flip { - int planes; + int pipes; } drm_i915_flip_t; #undef DRM_IOCTL_I915_FLIP @@ -314,7 +314,7 @@ intelWindowMoved(struct intel_context *intel) ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % intel_fb->pf_num_pages) << 2; - flip.planes = 0x2; + flip.pipes = 0x2; } else { intel->sarea->pf_current_page = intel->sarea->pf_current_page & (0x3 << 2); @@ -322,7 +322,7 @@ intelWindowMoved(struct intel_context *intel) (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % intel_fb->pf_num_pages; - flip.planes = 0x1; + flip.pipes = 0x1; } drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); @@ -449,7 +449,7 @@ intelPageFlip(const __DRIdrawablePrivate * dPriv) if (dPriv->numClipRects && intel_fb->pf_active) { drm_i915_flip_t flip; - flip.planes = intel_fb->pf_planes; + flip.pipes = intel_fb->pf_planes; ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); } -- cgit v1.2.3 From 344464bf2e4e151968cfb101c2477e440508b1f0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 3 Oct 2007 20:33:23 +0200 Subject: Track fragment and vertex shader code generation via pipe shader state objects. Unfortunately, the generated fragment shader code is effectively unusable until it handles quad->mask. --- src/mesa/pipe/draw/draw_private.h | 3 --- src/mesa/pipe/draw/draw_vertex_shader.c | 16 ++++++++++------ src/mesa/pipe/p_state.h | 5 +++++ src/mesa/pipe/softpipe/sp_context.c | 6 ++++++ src/mesa/pipe/softpipe/sp_context.h | 2 ++ src/mesa/pipe/softpipe/sp_quad_fs.c | 3 ++- src/mesa/pipe/softpipe/sp_state_fs.c | 20 ++++++++++++++++++++ src/mesa/state_tracker/st_cb_program.c | 7 ------- src/mesa/state_tracker/st_program.c | 10 ---------- src/mesa/state_tracker/st_program.h | 4 ---- 10 files changed, 45 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index b3f1c4d23e..79b2176c59 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -124,9 +124,6 @@ struct draw_stage */ struct draw_vertex_shader { const struct pipe_shader_state *state; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif }; /** diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 49ff3b32ff..e99537f16c 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -214,12 +214,13 @@ draw_create_vertex_shader(struct draw_context *draw, vs->state = shader; #if defined(__i386__) || defined(__386__) - x86_init_func(&vs->sse2_program); - if (draw->use_sse) { - tgsi_emit_sse2(shader->tokens, &vs->sse2_program); - ((struct pipe_shader_state*)(vs->state))->executable = - x86_get_func(&vs->sse2_program); + x86_init_func( &shader->sse2_program ); + + tgsi_emit_sse2( shader->tokens, &shader->sse2_program ); + + ((struct pipe_shader_state *)shader)->executable = (void *) + x86_get_func( &shader->sse2_program ); } #endif @@ -243,9 +244,12 @@ void draw_delete_vertex_shader(struct draw_context *draw, void *vcso) { struct draw_vertex_shader *vs = (struct draw_vertex_shader*)(vcso); + #if defined(__i386__) || defined(__386__) - x86_release_func(&vs->sse2_program); + x86_release_func(&vs->state->sse2_program); #endif + + free(vs->state); free(vcso); } diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index b4fc01bfc5..ac9d7009cc 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -40,6 +40,8 @@ #include "p_compiler.h" +#include "x86/rtasm/x86sse.h" + /** * Implementation limits */ @@ -143,6 +145,9 @@ struct pipe_constant_buffer { struct pipe_shader_state { const struct tgsi_token *tokens; +#if defined(__i386__) || defined(__386__) + struct x86_function sse2_program; +#endif void *executable; /** These fields somewhat constitute the shader "signature" */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 610e7ad689..daccc8fe23 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -249,6 +249,12 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); +#if defined(__i386__) || defined(__386__) + softpipe->use_sse = getenv("GALLIUM_SSE") != NULL; +#else + softpipe->use_sse = false; +#endif + softpipe->pipe.winsys = pipe_winsys; softpipe->pipe.destroy = softpipe_destroy; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index ccf29b5683..4f429e8139 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -154,6 +154,8 @@ struct softpipe_context { struct draw_stage *vbuf; struct pipe_surface *cbuf; /**< current color buffer (one of cbufs) */ + + int use_sse : 1; }; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index f394c587dc..a8c25c4868 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -103,7 +103,8 @@ shade_quad( machine->Inputs[0].xyzw[1].f[3] = fy + 1.0f; /* run shader */ - if( softpipe->fs->executable != NULL ) { + /* XXX: Generated code effectively unusable until it handles quad->mask */ + if( !quad->mask && softpipe->fs->executable != NULL ) { codegen_function func = (codegen_function) softpipe->fs->executable; func( machine->Inputs, diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 88d9bd9716..5547c849ac 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -31,17 +31,31 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "pipe/draw/draw_context.h" +#include "pipe/tgsi/exec/tgsi_core.h" void * softpipe_create_fs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { + struct softpipe_context *softpipe = softpipe_context(pipe); + /* Decide whether we'll be codegenerating this shader and if so do * that now. */ struct pipe_shader_state *state = malloc(sizeof(struct pipe_shader_state)); memcpy(state, templ, sizeof(struct pipe_shader_state)); + +#if defined(__i386__) || defined(__386__) + if (softpipe->use_sse) { + x86_init_func( &state->sse2_program ); + + tgsi_emit_sse2_fs( state->tokens, &state->sse2_program ); + + state->executable = (void *)x86_get_func( &state->sse2_program ); + } +#endif + return state; } @@ -57,6 +71,12 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) void softpipe_delete_fs_state(struct pipe_context *pipe, void *shader) { +#if defined(__i386__) || defined(__386__) + struct pipe_shader_state *state = shader; + + x86_release_func( &state->sse2_program ); +#endif + free(shader); } diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 1902c8d7d6..23a7bf473e 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -108,10 +108,6 @@ static struct gl_program *st_new_program( GLcontext *ctx, prog->serialNo = 1; -#if defined(__i386__) || defined(__386__) - x86_init_func( &prog->sse2_program ); -#endif - return _mesa_init_fragment_program( ctx, &prog->Base, target, @@ -140,9 +136,6 @@ static void st_delete_program( GLcontext *ctx, { struct st_fragment_program *stfp = (struct st_fragment_program *) prog; -#if defined(__i386__) || defined(__386__) - x86_release_func( &stfp->sse2_program ); -#endif st_remove_fragment_program(st, stfp); } break; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index a00f296d15..807cd8cb05 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -390,16 +390,6 @@ st_translate_fragment_program(struct st_context *st, if (TGSI_DEBUG) tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); -#if defined(__i386__) || defined(__386__) - if (draw_use_sse(st->draw)) { - if (stfp->sse2_program.csr == stfp->sse2_program.store) - tgsi_emit_sse2_fs( tokensOut, &stfp->sse2_program ); - - if (!cso->state.executable) - ((struct cso_fragment_shader*)cso)->state.executable = (void *) x86_get_func( &stfp->sse2_program ); - } -#endif - return cso; } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index a714f3f5b0..ae89055e82 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -61,10 +61,6 @@ struct st_fragment_program /** The program in TGSI format */ struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif - /** Pointer to the corresponding cached shader */ const struct cso_fragment_shader *fs; -- cgit v1.2.3 From 86a03e43ced13e7c00f16d037b27faddfbcb2333 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 3 Oct 2007 20:35:19 +0200 Subject: Double amount of memory allocated for generated shader code. The code generated for the glxgears vertex shader didn't fit, causing a crash. --- src/mesa/x86/rtasm/x86sse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index f9a9352e46..56c211eee0 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -1125,7 +1125,7 @@ struct x86_reg x86_fn_arg( struct x86_function *p, void x86_init_func( struct x86_function *p ) { - x86_init_func_size(p, 1024); + x86_init_func_size(p, 2048); } void x86_init_func_size( struct x86_function *p, unsigned code_size ) -- cgit v1.2.3 From e9df20c2fa7df8458a9f0781f31de51c4944d41f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 3 Oct 2007 20:37:59 +0200 Subject: Some minor cleanups. --- src/mesa/state_tracker/st_cb_clear.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index cca4fa19a9..a854378b18 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -264,7 +264,7 @@ clear_with_quad(GLcontext *ctx, GLboolean color, GLboolean depth, GLboolean stencil) { struct st_context *st = ctx->st; - struct pipe_context *pipe = ctx->st->pipe; + struct pipe_context *pipe = st->pipe; const GLfloat x0 = ctx->DrawBuffer->_Xmin; const GLfloat y0 = ctx->DrawBuffer->_Ymin; const GLfloat x1 = ctx->DrawBuffer->_Xmax; @@ -321,7 +321,7 @@ clear_with_quad(GLcontext *ctx, depth_stencil.stencil.value_mask[0] = 0xff; depth_stencil.stencil.write_mask[0] = ctx->Stencil.WriteMask[0] & 0xff; } - cso = st_cached_depth_stencil_state(ctx->st, &depth_stencil); + cso = st_cached_depth_stencil_state(st, &depth_stencil); pipe->bind_depth_stencil_state(pipe, cso->data); } @@ -337,7 +337,7 @@ clear_with_quad(GLcontext *ctx, if (ctx->Scissor.Enabled) raster.scissor = 1; #endif - cso = st_cached_rasterizer_state(ctx->st, &raster); + cso = st_cached_rasterizer_state(st, &raster); pipe->bind_rasterizer_state(pipe, cso->data); } @@ -385,7 +385,7 @@ clear_with_quad(GLcontext *ctx, pipe->bind_fs_state(pipe, st->state.fs->data); pipe->bind_vs_state(pipe, st->state.vs->data); pipe->bind_rasterizer_state(pipe, st->state.rasterizer->data); - pipe->set_viewport_state(pipe, &ctx->st->state.viewport); + pipe->set_viewport_state(pipe, &st->state.viewport); /* OR: st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); */ @@ -491,8 +491,7 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) assert(strb->surface->format); - if (ctx->Scissor.Enabled || - (isDS && ctx->DrawBuffer->Visual.stencilBits > 0)) { + if (check_clear_depth_with_quad(ctx, rb)) { /* scissoring or we have a combined depth/stencil buffer */ clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_FALSE); } -- cgit v1.2.3 From 1a18ebc6b31ebb8e950b84790744dd5e3113f014 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 09:55:06 -0600 Subject: move TEX code into exec_tex() --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 214 +++++++++++++++++++++--------------- 1 file changed, 125 insertions(+), 89 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 3f170ffcfc..28211c9507 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1194,7 +1194,7 @@ fetch_texel( struct tgsi_sampler *sampler, const union tgsi_exec_channel *s, const union tgsi_exec_channel *t, const union tgsi_exec_channel *p, - float lodbias, + float lodbias, /* XXX should be float[4] */ union tgsi_exec_channel *r, union tgsi_exec_channel *g, union tgsi_exec_channel *b, @@ -1213,6 +1213,124 @@ fetch_texel( struct tgsi_sampler *sampler, } } + +static void +exec_tex(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + boolean biasLod) +{ + const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + union tgsi_exec_channel r[8]; + uint chan_index; + float lodBias; + + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: + + FETCH(&r[0], 0, CHAN_X); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[1], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[1] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[1], 0, CHAN_W); + lodBias = r[1].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ + &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ + break; + + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[2], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[2] ); + micro_div( &r[1], &r[1], &r[2] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[1], 0, CHAN_W); + lodBias = r[1].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], NULL, lodBias, + &r[0], &r[1], &r[2], &r[3]); + break; + + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[1], 0, CHAN_W); + lodBias = r[1].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, + &r[0], &r[1], &r[2], &r[3]); + break; + + default: + assert (0); + } + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } +} + + + static void constant_interpolation( struct tgsi_exec_machine *mach, @@ -1880,91 +1998,13 @@ exec_instruction( break; case TGSI_OPCODE_TEX: - { - const GLuint unit = inst->FullSrcRegisters[1].SrcRegister.Index; - switch (inst->InstructionExtTexture.Texture) { - case TGSI_TEXTURE_1D: - - FETCH(&r[0], 0, CHAN_X); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[1], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[1] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - fetch_texel(&mach->Samplers[unit], - &r[0], NULL, NULL, 0.0, - &r[0], &r[1], &r[2], &r[3]); - break; - - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[2], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[2] ); - micro_div( &r[1], &r[1], &r[2] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], NULL, 0.0, - &r[0], &r[1], &r[2], &r[3]); - break; - - case TGSI_TEXTURE_3D: - case TGSI_TEXTURE_CUBE: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], 0.0, - &r[0], &r[1], &r[2], &r[3]); - break; - - default: - assert (0); - } + /* src arg0 is the texcoord */ + exec_tex(mach, inst, FALSE); + break; - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[chan_index], 0, chan_index ); - } - } + case TGSI_OPCODE_TXB: + /* Texture lookup with lod bias (src0.w) */ + exec_tex(mach, inst, TRUE); break; case TGSI_OPCODE_TXD: @@ -2076,10 +2116,6 @@ exec_instruction( } break; - case TGSI_OPCODE_TXB: - assert (0); - break; - case TGSI_OPCODE_NRM: assert (0); break; -- cgit v1.2.3 From df235f33a6e851e0487755e9ad5f9d5cfd041c18 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 09:55:36 -0600 Subject: fix f/u mix-up in micro_trunc() --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 28211c9507..c075223728 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -809,10 +809,10 @@ micro_trunc( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src0 ) { - dst->f[0] = (float) (int) src0->u[0]; - dst->f[1] = (float) (int) src0->u[1]; - dst->f[2] = (float) (int) src0->u[2]; - dst->f[3] = (float) (int) src0->u[3]; + dst->f[0] = (float) (int) src0->f[0]; + dst->f[1] = (float) (int) src0->f[1]; + dst->f[2] = (float) (int) src0->f[2]; + dst->f[3] = (float) (int) src0->f[3]; } static void -- cgit v1.2.3 From 76e61556e7b69b6ce11be4bd600590a049bfbbef Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 10:33:05 -0600 Subject: fix eq/neq bugs --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 20 ++++---------------- 1 file changed, 4 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index c075223728..6116db8dcb 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -698,20 +698,6 @@ micro_movc( dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; } -static void -micro_ne( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] != src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] != src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] != src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] != src1->f[3] ? src2->f[3] : src3->f[3]; -} - static void micro_neg( union tgsi_exec_channel *dst, @@ -1949,7 +1935,9 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + micro_eq( &r[0], &r[0], &r[1], + &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], + &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); STORE( &r[0], 0, chan_index ); } break; @@ -1988,7 +1976,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_ne( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); STORE( &r[0], 0, chan_index ); } break; -- cgit v1.2.3 From e6f4af6b236fcf077cef2ea74346fa9b97106b95 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 13:44:53 -0600 Subject: added s8z24_get_tile() --- src/mesa/pipe/softpipe/sp_surface.c | 37 +++++++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index b06f6813c0..574b474ba3 100755 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -695,6 +695,42 @@ s8z24_write_quad_stencil(struct softpipe_surface *sps, } +/** + * Note, the actual returned pixels are uint, not float + */ +static void +s8z24_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const uint *src + = ((const uint *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_S8_Z24); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp clipping hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + uint *pRow = (uint *) p; + for (j = 0; j < w; j++) { + pRow[j] = src[j]; + } + src += ps->region->pitch; + p += w0; + } +} + + /*** PIPE_FORMAT_U_S8 ***/ static void @@ -784,6 +820,7 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) sps->write_quad_z = s8z24_write_quad_z; sps->read_quad_stencil = s8z24_read_quad_stencil; sps->write_quad_stencil = s8z24_write_quad_stencil; + sps->surface.get_tile = s8z24_get_tile; break; case PIPE_FORMAT_U_S8: -- cgit v1.2.3 From a9f0330061471ba47beb2369884d7661b715722e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 13:45:24 -0600 Subject: basic support for reading GL_DEPTH_COMPONENT --- src/mesa/state_tracker/st_cb_readpixels.c | 37 ++++++++++++++++++++++++++++--- 1 file changed, 34 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 98604e5b6b..a82c4e526a 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -86,7 +86,12 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } - strb = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + if (format == GL_DEPTH_COMPONENT) { + strb = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer); + } + else { + strb = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + } if (!strb) return; @@ -98,6 +103,14 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, height, format, type, 0, 0); dfStride = width * 4; } +#if 0 + else if (format == GL_DEPTH_COMPONENT && type == GL_FLOAT) { + /* write tile(row) directly into user's buffer */ + df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width, + height, format, type, 0, 0); + dfStride = width; + } +#endif else { /* write tile(row) into temp row buffer */ df = (GLfloat *) temp; @@ -122,8 +135,26 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, /* convert GLfloat to user's format/type */ GLvoid *dst = _mesa_image_address2d(&clippedPacking, dest, width, height, format, type, i, 0); - _mesa_pack_rgba_span_float(ctx, width, temp, format, type, dst, - &clippedPacking, transferOps); + if (format == GL_DEPTH_COMPONENT) { + float z[MAX_WIDTH]; + if (strb->surface->format == PIPE_FORMAT_S8_Z24) { + const double scale = 1.0 / ((1 << 24) - 1); + const uint *zs = (const uint *) temp; + uint k; + for (k = 0; k < width; k++) { + z[k] = (zs[k] & 0xffffff) * scale; + } + } + else { + assert(0); + } + _mesa_pack_depth_span(ctx, width, dst, type, + z, &clippedPacking); + } + else { + _mesa_pack_rgba_span_float(ctx, width, temp, format, type, dst, + &clippedPacking, transferOps); + } } } -- cgit v1.2.3 From 5356ba250c078c861f3a31a587b79570308cfc23 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 13:46:48 -0600 Subject: handle frag progs that write Z --- src/mesa/pipe/softpipe/sp_quad_fs.c | 23 ++++++++++------------- 1 file changed, 10 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index a8c25c4868..d9a2e6b9cf 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -123,25 +123,22 @@ shade_quad( &machine->Outputs[1].xyzw[0].f[0], sizeof( quad->outputs.color ) ); -#if 0 - if( softpipe->need_z ) { - /* XXX temporary */ - memcpy( - quad->outputs.depth, - machine->Outputs[0].xyzw[2], - sizeof( quad->outputs.depth ) ); + /* Z */ + if (qss->stage.softpipe->fs->output_semantic_name[0] + == TGSI_SEMANTIC_POSITION) { + /* output[0] is new Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; + } } -#else - { + else { + /* pass input Z to output Z */ uint i; for (i = 0; i < 4; i++) { quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; -#if 0 - printf("output z %f\n", quad->outputs.depth[i]); -#endif } } -#endif /* shader may cull fragments */ if( quad->mask ) { -- cgit v1.2.3 From 65996f27f48e2212bf36e515aba82b6876398bfd Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 13:49:31 -0600 Subject: Move XSTDCALL definition to p_compiler.h --- src/mesa/pipe/draw/draw_vertex_shader.c | 10 ---------- src/mesa/pipe/p_compiler.h | 10 ++++++++++ src/mesa/pipe/softpipe/sp_quad_fs.c | 8 -------- 3 files changed, 10 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index e99537f16c..ea3fc2128d 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -55,16 +55,6 @@ compute_clipmask(const float *clip, const float (*plane)[4], unsigned nr) } - - -#if !defined(XSTDCALL) -#if defined(WIN32) -#define XSTDCALL __stdcall -#else -#define XSTDCALL -#endif -#endif - typedef void (XSTDCALL *codegen_function) ( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h index 3dd6bc5c5a..020977b05b 100644 --- a/src/mesa/pipe/p_compiler.h +++ b/src/mesa/pipe/p_compiler.h @@ -80,4 +80,14 @@ typedef unsigned long long uint64; +/** For calling code-gen'd functions */ +#if !defined(XSTDCALL) +#if defined(WIN32) +#define XSTDCALL __stdcall +#else +#define XSTDCALL +#endif +#endif + + #endif /* P_COMPILER_H */ diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index d9a2e6b9cf..d9349a2280 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -58,14 +58,6 @@ quad_shade_stage(struct quad_stage *qs) } -#if !defined(XSTDCALL) -#if defined(WIN32) -#define XSTDCALL __stdcall -#else -#define XSTDCALL -#endif -#endif - typedef void (XSTDCALL *codegen_function)( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, -- cgit v1.2.3 From 635c4c41bdf111462939da292d65328595d314e8 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 14:43:57 -0600 Subject: Add outputs_written bitfield to pipe_shader_state, use it to determine if fragment shader writes Z. --- src/mesa/pipe/p_state.h | 4 +-- src/mesa/pipe/softpipe/sp_quad_fs.c | 3 +-- src/mesa/state_tracker/st_program.c | 49 ++++++++++++++++++++++++++----------- 3 files changed, 37 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index ac9d7009cc..64c5f13f23 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -150,13 +150,11 @@ struct pipe_shader_state { #endif void *executable; - /** These fields somewhat constitute the shader "signature" */ ubyte num_inputs; ubyte num_outputs; - + uint outputs_written; /**< bitmask */ ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; - ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; }; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index d9349a2280..2b0c4366bd 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -116,8 +116,7 @@ shade_quad( sizeof( quad->outputs.color ) ); /* Z */ - if (qss->stage.softpipe->fs->output_semantic_name[0] - == TGSI_SEMANTIC_POSITION) { + if (qss->stage.softpipe->fs->outputs_written & 0x1) { /* output[0] is new Z */ uint i; for (i = 0; i < 4; i++) { diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 807cd8cb05..86aaaad6bb 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -160,6 +160,8 @@ st_translate_vertex_program(struct st_context *st, defaultOutputMapping[attr] = slot; } + vs.outputs_written |= (1 << slot); + /* printf("Output %u -> slot %u\n", attr, slot); */ @@ -346,21 +348,40 @@ st_translate_fragment_program(struct st_context *st, /* * Semantics for outputs */ - for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { - if (stfp->Base.Base.OutputsWritten & (1 << attr)) { - switch (attr) { - case FRAG_RESULT_DEPR: - fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_POSITION; - outputMapping[attr] = fs.num_outputs; - break; - case FRAG_RESULT_COLR: - fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_COLOR; - outputMapping[attr] = fs.num_outputs; - break; - default: - assert(0); + { + uint numColors = 0; + GLbitfield outputsWritten = stfp->Base.Base.OutputsWritten; + + /* output[0] is always Z, but may not really be written */ + fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_POSITION; + fs.output_semantic_index[fs.num_outputs] = 0; + outputMapping[FRAG_RESULT_DEPR] = fs.num_outputs; + fs.num_outputs++; + + if (outputsWritten & (1 << FRAG_RESULT_DEPR)) { + fs.outputs_written |= 0x1; + outputsWritten &= ~(1 << FRAG_RESULT_DEPR); + } + + /* color outputs begin at output [1] */ + for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { + if (outputsWritten & (1 << attr)) { + switch (attr) { + case FRAG_RESULT_DEPR: + assert(0); + break; + case FRAG_RESULT_COLR: + fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_COLOR; + fs.output_semantic_index[fs.num_outputs] = numColors; + outputMapping[attr] = fs.num_outputs; + fs.outputs_written |= (0x2 << numColors); + numColors++; + break; + default: + assert(0); + } + fs.num_outputs++; } - fs.num_outputs++; } } -- cgit v1.2.3 From 68c4254d757012cbb5994d3b55769eb250747fce Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 16:18:44 -0600 Subject: updated comments --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 18 ++++-------------- 1 file changed, 4 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 2a2a1a0430..189c7f3eda 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -59,22 +59,13 @@ map_register_file_index( switch( file ) { case TGSI_FILE_INPUT: - /* - * The fragment/vertex program input indexes (FRAG/VERT_ATTRIB_x) get - * mapped to a packed sequence of integers. - * If a program uses one input attribute, the mapped index will be 1. - * If a program uses two input attribs, the mapped indexes will be 1,2. - * If a program uses 3 input attribs, the mapped indexes will be 1,2,3. - * etc. - */ + /* inputs are mapped according to the user-defined map */ return inputMapping[index]; case TGSI_FILE_OUTPUT: - /* - assert( usage_bitmask == 0x0 ); - */ if( processor == TGSI_PROCESSOR_FRAGMENT ) { - /* depth result -> index 0 + /* fragment program outputs are hard-coded: + * depth result -> index 0 * color results -> index 1, 2, ... */ if( index == FRAG_RESULT_DEPR ) { @@ -86,8 +77,7 @@ map_register_file_index( } } else { - /* vertex output slots are tightly packed, find mapped pos */ - /* mapped_index = VERT_RESULT_x */ + /* vertex outputs are mapped according to the user-defined map */ mapped_index = outputMapping[index]; } break; -- cgit v1.2.3 From 1a69db9d2dfae16df5aa606ada7d631d98944449 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 3 Oct 2007 16:34:04 -0600 Subject: use different temps in exec_tex() for bias --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 6116db8dcb..52848a2a73 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1230,7 +1230,7 @@ exec_tex(struct tgsi_exec_machine *mach, if (biasLod) { FETCH(&r[1], 0, CHAN_W); - lodBias = r[1].f[0]; + lodBias = r[2].f[0]; } else lodBias = 0.0; @@ -1261,8 +1261,8 @@ exec_tex(struct tgsi_exec_machine *mach, } if (biasLod) { - FETCH(&r[1], 0, CHAN_W); - lodBias = r[1].f[0]; + FETCH(&r[2], 0, CHAN_W); + lodBias = r[2].f[0]; } else lodBias = 0.0; @@ -1295,8 +1295,8 @@ exec_tex(struct tgsi_exec_machine *mach, } if (biasLod) { - FETCH(&r[1], 0, CHAN_W); - lodBias = r[1].f[0]; + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; } else lodBias = 0.0; -- cgit v1.2.3 From 94490f63c8294efbce8822cf2be6da9be67c8789 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 9 Oct 2007 14:12:26 -0600 Subject: Use linked fragment shader inputs (if present) to determine which outputs are needed. --- src/mesa/tnl/t_vp_build.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index f923b8e132..b28102b1b9 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -118,10 +118,17 @@ static struct state_key *make_state_key( GLcontext *ctx ) { TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; - const struct gl_fragment_program *fp = ctx->FragmentProgram._Current; + const struct gl_fragment_program *fp; struct state_key *key = CALLOC_STRUCT(state_key); GLuint i; + if (ctx->Shader.CurrentProgram && + ctx->Shader.CurrentProgram->LinkStatus && + ctx->Shader.CurrentProgram->FragmentProgram) + fp = ctx->Shader.CurrentProgram->FragmentProgram; + else + fp = ctx->FragmentProgram._Current; + /* This now relies on texenvprogram.c being active: */ assert(fp); -- cgit v1.2.3 From 47ff9428ee09a948e6b412f7e3e46ed1305881a1 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 9 Oct 2007 14:31:45 -0600 Subject: Don't emit semantic info for vertex program inputs. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 189c7f3eda..e4a93cb60d 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -475,6 +475,7 @@ make_input_decl( GLuint index, GLuint interpolate, GLuint usage_mask, + GLboolean semantic_info, GLuint semantic_name, GLuint semantic_index ) { @@ -486,12 +487,14 @@ make_input_decl( decl.Declaration.File = TGSI_FILE_INPUT; decl.Declaration.Declare = TGSI_DECLARE_RANGE; decl.Declaration.UsageMask = usage_mask; - decl.Declaration.Semantic = 1; + decl.Declaration.Semantic = semantic_info; decl.Declaration.Interpolate = 1; decl.u.DeclarationRange.First = index; decl.u.DeclarationRange.Last = index; - decl.Semantic.SemanticName = semantic_name; - decl.Semantic.SemanticIndex = semantic_index; + if (semantic_info) { + decl.Semantic.SemanticName = semantic_name; + decl.Semantic.SemanticIndex = semantic_index; + } decl.Interpolation.Interpolate = interpolate; return decl; @@ -569,7 +572,7 @@ tgsi_mesa_compile_fp_program( fulldecl = make_input_decl(i, TGSI_INTERPOLATE_CONSTANT, TGSI_WRITEMASK_XY, - TGSI_SEMANTIC_POSITION, 0 ); + GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -579,7 +582,7 @@ tgsi_mesa_compile_fp_program( fulldecl = make_input_decl(i, TGSI_INTERPOLATE_LINEAR, TGSI_WRITEMASK_ZW, - TGSI_SEMANTIC_POSITION, 0 ); + GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); ti += tgsi_build_full_declaration( &fulldecl, &tokens[ti], @@ -590,7 +593,7 @@ tgsi_mesa_compile_fp_program( fulldecl = make_input_decl(i, interpMode[i], TGSI_WRITEMASK_XYZW, - inputSemanticName[i], + GL_TRUE, inputSemanticName[i], inputSemanticIndex[i]); ti += tgsi_build_full_declaration(&fulldecl, &tokens[ti], @@ -713,7 +716,7 @@ tgsi_mesa_compile_vp_program( fulldecl = make_input_decl(i, TGSI_INTERPOLATE_CONSTANT, /* no interp */ TGSI_WRITEMASK_XYZW, - inputSemanticName[i], + GL_FALSE, inputSemanticName[i], inputSemanticIndex[i]); ti += tgsi_build_full_declaration(&fulldecl, &tokens[ti], -- cgit v1.2.3 From 31bd01f93b0f5795e290897e79c3f16d90e1f202 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 9 Oct 2007 14:36:14 -0600 Subject: New ATTRIB interpolation type for vertex program inputs/declarations. --- src/mesa/pipe/tgsi/exec/tgsi_build.c | 2 +- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 6 ++++-- src/mesa/pipe/tgsi/exec/tgsi_token.h | 1 + src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 +- 4 files changed, 7 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.c b/src/mesa/pipe/tgsi/exec/tgsi_build.c index 78f648aae2..56827726f5 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_build.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_build.c @@ -294,7 +294,7 @@ tgsi_build_declaration_interpolation( { struct tgsi_declaration_interpolation di; - assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); + assert( interpolate <= TGSI_INTERPOLATE_ATTRIB ); di = tgsi_default_declaration_interpolation(); di.Interpolate = interpolate; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index c179659aee..463f841dde 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -189,14 +189,16 @@ static const char *TGSI_INTERPOLATES[] = { "INTERPOLATE_CONSTANT", "INTERPOLATE_LINEAR", - "INTERPOLATE_PERSPECTIVE" + "INTERPOLATE_PERSPECTIVE", + "INTERPOLATE_ATTRIB" }; static const char *TGSI_INTERPOLATES_SHORT[] = { "CONSTANT", "LINEAR", - "PERSPECTIVE" + "PERSPECTIVE", + "ATTRIB" }; static const char *TGSI_SEMANTICS[] = diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h index 8d5992facb..80c3fcd434 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -96,6 +96,7 @@ struct tgsi_declaration_mask #define TGSI_INTERPOLATE_CONSTANT 0 #define TGSI_INTERPOLATE_LINEAR 1 #define TGSI_INTERPOLATE_PERSPECTIVE 2 +#define TGSI_INTERPOLATE_ATTRIB 3 /**< Vertex shader input attrib */ struct tgsi_declaration_interpolation { diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index e4a93cb60d..88de85994a 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -714,7 +714,7 @@ tgsi_mesa_compile_vp_program( for (i = 0; i < numInputs; i++) { struct tgsi_full_declaration fulldecl; fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_CONSTANT, /* no interp */ + TGSI_INTERPOLATE_ATTRIB, TGSI_WRITEMASK_XYZW, GL_FALSE, inputSemanticName[i], inputSemanticIndex[i]); -- cgit v1.2.3 From aec1f8e4f8315df23dc51a0b5a5fad90a03be851 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 9 Oct 2007 14:40:11 -0600 Subject: More work on flow control for CAL/RET. --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 37 +++++++++++++++++++++++++------------ src/mesa/pipe/tgsi/exec/tgsi_exec.h | 6 ++++++ 2 files changed, 31 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 52848a2a73..6bbff76a87 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -109,7 +109,7 @@ /** The execution mask depends on the conditional mask and the loop mask */ #define UPDATE_EXEC_MASK(MACH) \ - MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask + MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask #define CHAN_X 0 @@ -2044,6 +2044,9 @@ exec_instruction( assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); mach->ContStack[mach->ContStackTop++] = mach->ContMask; + assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); + mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; + /* note that PC was already incremented above */ mach->CallStack[mach->CallStackTop++] = *pc; *pc = inst->InstructionExtLabel.Label; @@ -2053,19 +2056,28 @@ exec_instruction( case TGSI_OPCODE_RET: /* XXX examine ExecMask to determine if we should _really_ return */ /* pop Cond, Loop, Cont stacks */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; + mach->FuncMask &= ~mach->ExecMask; + UPDATE_EXEC_MASK(mach); - assert(mach->CallStackTop >= 0); - if (mach->CallStackTop == 0) { - /* XXX error? */ - return; + if (mach->ExecMask == 0x0) { + /* really return now (otherwise, keep executing */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + + assert(mach->FuncStackTop > 0); + mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; + + assert(mach->CallStackTop >= 0); + if (mach->CallStackTop == 0) { + /* XXX error? */ + return; + } + *pc = mach->CallStack[--mach->CallStackTop]; } - *pc = mach->CallStack[--mach->CallStackTop]; break; case TGSI_OPCODE_SSG: @@ -2387,6 +2399,7 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) mach->CondMask = 0xf; mach->LoopMask = 0xf; mach->ContMask = 0xf; + mach->FuncMask = 0xf; mach->ExecMask = 0xf; assert(mach->CondStackTop == 0); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 354d5df6c2..20647e72e2 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -141,6 +141,7 @@ struct tgsi_exec_machine uint CondMask; /**< For IF/ELSE/ENDIF */ uint LoopMask; /**< For BGNLOOP/ENDLOOP */ uint ContMask; /**< For loop CONT statements */ + uint FuncMask; /**< For function calls */ uint ExecMask; /**< = CondMask & LoopMask */ /** Condition mask stack (for nested conditionals) */ @@ -155,6 +156,11 @@ struct tgsi_exec_machine uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; int ContStackTop; + /** Function execution mask stack (for executing subroutine code) */ + uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; + int FuncStackTop; + + /** Function call stack for saving/restoring the program counter */ uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; int CallStackTop; -- cgit v1.2.3 From 342bc50c3d8765ea4ab50aa7d77df5c86c478c61 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 9 Oct 2007 14:52:26 -0600 Subject: update some comments --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 6bbff76a87..fda5584cf3 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -2036,7 +2036,7 @@ exec_instruction( if (mach->ExecMask) { /* do the call */ - /* push Cond, Loop, Cont stacks */ + /* push the Cond, Loop, Cont stacks */ assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); mach->CondStack[mach->CondStackTop++] = mach->CondMask; assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); @@ -2054,13 +2054,13 @@ exec_instruction( break; case TGSI_OPCODE_RET: - /* XXX examine ExecMask to determine if we should _really_ return */ - /* pop Cond, Loop, Cont stacks */ mach->FuncMask &= ~mach->ExecMask; UPDATE_EXEC_MASK(mach); if (mach->ExecMask == 0x0) { /* really return now (otherwise, keep executing */ + + /* pop the Cond, Loop, Cont stacks */ assert(mach->CondStackTop > 0); mach->CondMask = mach->CondStack[--mach->CondStackTop]; assert(mach->LoopStackTop > 0); -- cgit v1.2.3 From bc139a19b00f8686caa8db7c56af2087f26e369a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 9 Oct 2007 14:55:22 -0600 Subject: Pack fragment program outputs to be consistant with vertex programs. Previously, output[0] was always Z and output[1] was color. Now output[0] will be color if Z is not written. In shade_quad() use the semantic info to determine which quantity is in which output slot. --- src/mesa/pipe/p_state.h | 1 - src/mesa/pipe/softpipe/sp_quad_fs.c | 23 ++++++++++++++--------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 31 ++----------------------------- src/mesa/state_tracker/st_program.c | 20 ++++++++------------ 4 files changed, 24 insertions(+), 51 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 64c5f13f23..99ec574124 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -152,7 +152,6 @@ struct pipe_shader_state { ubyte num_inputs; ubyte num_outputs; - uint outputs_written; /**< bitmask */ ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 2b0c4366bd..b6005c6a69 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -78,6 +78,7 @@ shade_quad( const float fx = (float) quad->x0; const float fy = (float) quad->y0; struct tgsi_exec_machine *machine = &qss->machine; + uint colorOut; /* Consts does not require 16 byte alignment. */ machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; @@ -109,28 +110,32 @@ shade_quad( quad->mask &= tgsi_exec_machine_run( machine ); } - /* store result color (always in output[1]) */ - memcpy( - quad->outputs.color, - &machine->Outputs[1].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); - - /* Z */ - if (qss->stage.softpipe->fs->outputs_written & 0x1) { + if (qss->stage.softpipe->fs->output_semantic_name[0] == TGSI_SEMANTIC_POSITION) { /* output[0] is new Z */ uint i; for (i = 0; i < 4; i++) { quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; } + colorOut = 1; } else { - /* pass input Z to output Z */ + /* pass input Z (which was interpolated by the executor) to output Z */ uint i; for (i = 0; i < 4; i++) { quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; } + colorOut = 0; } + /* store result color */ + /* XXX need to handle multiple color outputs someday */ + assert(qss->stage.softpipe->fs->output_semantic_name[colorOut] + == TGSI_SEMANTIC_COLOR); + memcpy( + quad->outputs.color, + &machine->Outputs[colorOut].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); + /* shader may cull fragments */ if( quad->mask ) { qs->next->run( qs->next, quad ); diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 88de85994a..5a1ec3553e 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -46,47 +46,22 @@ map_register_file( */ static GLuint map_register_file_index( - GLuint processor, GLuint file, GLuint index, const GLuint inputMapping[], const GLuint outputMapping[]) { - GLuint mapped_index; - - assert(processor == TGSI_PROCESSOR_FRAGMENT - || processor == TGSI_PROCESSOR_VERTEX); - switch( file ) { case TGSI_FILE_INPUT: /* inputs are mapped according to the user-defined map */ return inputMapping[index]; case TGSI_FILE_OUTPUT: - if( processor == TGSI_PROCESSOR_FRAGMENT ) { - /* fragment program outputs are hard-coded: - * depth result -> index 0 - * color results -> index 1, 2, ... - */ - if( index == FRAG_RESULT_DEPR ) { - mapped_index = 0; /**TGSI_ATTRIB_POS;**/ - } - else { - assert( index == FRAG_RESULT_COLR ); - mapped_index = 1; /**TGSI_ATTRIB_COLOR0;**/ - } - } - else { - /* vertex outputs are mapped according to the user-defined map */ - mapped_index = outputMapping[index]; - } - break; + return outputMapping[index]; default: - mapped_index = index; + return index; } - - return mapped_index; } /* @@ -166,7 +141,6 @@ compile_instruction( fulldst = &fullinst->FullDstRegisters[0]; fulldst->DstRegister.File = map_register_file( inst->DstReg.File ); fulldst->DstRegister.Index = map_register_file_index( - processor, fulldst->DstRegister.File, inst->DstReg.Index, inputMapping, @@ -180,7 +154,6 @@ compile_instruction( fullsrc = &fullinst->FullSrcRegisters[i]; fullsrc->SrcRegister.File = map_register_file( inst->SrcReg[i].File ); fullsrc->SrcRegister.Index = map_register_file_index( - processor, fullsrc->SrcRegister.File, inst->SrcReg[i].Index, inputMapping, diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 86aaaad6bb..0b14830353 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -160,8 +160,6 @@ st_translate_vertex_program(struct st_context *st, defaultOutputMapping[attr] = slot; } - vs.outputs_written |= (1 << slot); - /* printf("Output %u -> slot %u\n", attr, slot); */ @@ -346,35 +344,33 @@ st_translate_fragment_program(struct st_context *st, } /* - * Semantics for outputs + * Semantics and mapping for outputs */ { uint numColors = 0; GLbitfield outputsWritten = stfp->Base.Base.OutputsWritten; - /* output[0] is always Z, but may not really be written */ - fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_POSITION; - fs.output_semantic_index[fs.num_outputs] = 0; - outputMapping[FRAG_RESULT_DEPR] = fs.num_outputs; - fs.num_outputs++; - + /* if z is written, emit that first */ if (outputsWritten & (1 << FRAG_RESULT_DEPR)) { - fs.outputs_written |= 0x1; + fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_POSITION; + fs.output_semantic_index[fs.num_outputs] = 0; + outputMapping[FRAG_RESULT_DEPR] = fs.num_outputs; + fs.num_outputs++; outputsWritten &= ~(1 << FRAG_RESULT_DEPR); } - /* color outputs begin at output [1] */ + /* handle remaning outputs (color) */ for (attr = 0; attr < FRAG_RESULT_MAX; attr++) { if (outputsWritten & (1 << attr)) { switch (attr) { case FRAG_RESULT_DEPR: + /* handled above */ assert(0); break; case FRAG_RESULT_COLR: fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_COLOR; fs.output_semantic_index[fs.num_outputs] = numColors; outputMapping[attr] = fs.num_outputs; - fs.outputs_written |= (0x2 << numColors); numColors++; break; default: -- cgit v1.2.3 From 578962d6240252567cb9725a71e9854ad0dc72a0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 9 Oct 2007 16:20:00 -0600 Subject: Finish up decls for packed fragment program outputs. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 52 ++++++++++++++++++---------------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 3 ++ src/mesa/state_tracker/st_program.c | 3 ++ 3 files changed, 34 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 5a1ec3553e..66b8db07f9 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -516,7 +516,10 @@ tgsi_mesa_compile_fp_program( const ubyte inputSemanticName[], const ubyte inputSemanticIndex[], const GLuint interpMode[], + GLuint numOutputs, const GLuint outputMapping[], + const ubyte outputSemanticName[], + const ubyte outputSemanticIndex[], struct tgsi_token *tokens, GLuint maxTokens ) { @@ -580,30 +583,31 @@ tgsi_mesa_compile_fp_program( /* * Declare output attributes. */ - assert( - program->Base.OutputsWritten == - (program->Base.OutputsWritten & ((1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR))) ); - - fulldecl = make_output_decl( - 0, - TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ - TGSI_WRITEMASK_Z ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - - if( program->Base.OutputsWritten & (1 << FRAG_RESULT_COLR) ) { - fulldecl = make_output_decl( - 1, - TGSI_SEMANTIC_COLOR, 0, - TGSI_WRITEMASK_XYZW ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); + for (i = 0; i < numOutputs; i++) { + switch (outputSemanticName[i]) { + case TGSI_SEMANTIC_POSITION: + fulldecl = make_output_decl(i, + TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ + TGSI_WRITEMASK_Z ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + case TGSI_SEMANTIC_COLOR: + fulldecl = make_output_decl(i, + TGSI_SEMANTIC_COLOR, 0, + TGSI_WRITEMASK_XYZW ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + default: + abort(); + } } /* diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index cf85eb9c0b..29e8053369 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -15,7 +15,10 @@ tgsi_mesa_compile_fp_program( const ubyte inputSemanticName[], const ubyte inputSemanticIndex[], const GLuint interpMode[], + GLuint numOutputs, const GLuint outputMapping[], + const ubyte outputSemanticName[], + const ubyte outputSemanticIndex[], struct tgsi_token *tokens, GLuint maxTokens ); diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 0b14830353..1c5af7b75f 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -394,7 +394,10 @@ st_translate_fragment_program(struct st_context *st, fs.input_semantic_index, interpMode, /* outputs */ + fs.num_outputs, outputMapping, + fs.output_semantic_name, + fs.output_semantic_index, /* tokenized result */ tokensOut, maxTokens); -- cgit v1.2.3 From 11a46c61e0552c055513050ef3ec28050a143f91 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 9 Oct 2007 16:28:01 -0600 Subject: Generate DCLs for temporaries. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 78 +++++++++++++++++++++++++++++++++- 1 file changed, 77 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 66b8db07f9..750314fb43 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -498,6 +498,54 @@ make_output_decl( } +static struct tgsi_full_declaration +make_temp_decl(GLuint index) +{ + struct tgsi_full_declaration decl; + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.u.DeclarationRange.First = index; + decl.u.DeclarationRange.Last = index; + return decl; +} + + +/** + * Find the temporaries which are used in the given program. + * Put the indices of the temporaries in 'tempsUsed'. + * \return number of temporaries used + */ +static GLuint +find_temporaries(const struct gl_program *program, + GLuint tempsUsed[MAX_PROGRAM_TEMPS]) +{ + GLuint i, j, count; + + for (i = 0; i < MAX_PROGRAM_TEMPS; i++) + tempsUsed[i] = GL_FALSE; + + for (i = 0; i < program->NumInstructions; i++) { + const struct prog_instruction *inst = program->Instructions + i; + const GLuint n = _mesa_num_inst_src_regs( inst->Opcode ); + for (j = 0; j < n; j++) { + if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) + tempsUsed[inst->SrcReg[j].Index] = GL_TRUE; + if (inst->DstReg.File == PROGRAM_TEMPORARY) + tempsUsed[inst->DstReg.Index] = GL_TRUE; + } + } + + /* convert flags to list of indices */ + count = 0; + for (i = 0; i < MAX_PROGRAM_TEMPS; i++) { + if (tempsUsed[i]) + tempsUsed[count++] = i; + } + return count; +} + + /** * Convert Mesa fragment program to TGSI format. * \param inputMapping maps Mesa fragment program inputs to TGSI generic @@ -527,7 +575,6 @@ tgsi_mesa_compile_fp_program( GLuint ti; /* token index */ struct tgsi_header *header; struct tgsi_processor *processor; - struct tgsi_full_declaration fulldecl; struct tgsi_full_instruction fullinst; GLuint preamble_size = 0; @@ -542,6 +589,7 @@ tgsi_mesa_compile_fp_program( ti = 3; for (i = 0; i < numInputs; i++) { + struct tgsi_full_declaration fulldecl; switch (inputSemanticName[i]) { case TGSI_SEMANTIC_POSITION: /* Fragment XY pos */ @@ -584,6 +632,7 @@ tgsi_mesa_compile_fp_program( * Declare output attributes. */ for (i = 0; i < numOutputs; i++) { + struct tgsi_full_declaration fulldecl; switch (outputSemanticName[i]) { case TGSI_SEMANTIC_POSITION: fulldecl = make_output_decl(i, @@ -610,6 +659,20 @@ tgsi_mesa_compile_fp_program( } } + { + GLuint tempsUsed[MAX_PROGRAM_TEMPS]; + uint numTemps = find_temporaries(&program->Base, tempsUsed); + for (i = 0; i < numTemps; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_temp_decl(tempsUsed[i]); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + /* * Copy fragment z if the shader does not write it. */ @@ -714,6 +777,19 @@ tgsi_mesa_compile_vp_program( maxTokens - ti ); } + { + GLuint tempsUsed[MAX_PROGRAM_TEMPS]; + uint numTemps = find_temporaries(&program->Base, tempsUsed); + for (i = 0; i < numTemps; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_temp_decl(tempsUsed[i]); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } for( i = 0; i < program->Base.NumInstructions; i++ ) { compile_instruction( -- cgit v1.2.3 From d4c3f26662c48235462ae25b84c7b63e48af1c22 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 9 Oct 2007 16:56:25 -0600 Subject: start on a unified vertex/fragment program translation function --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 194 +++++++++++++++++++++++++++++++++ 1 file changed, 194 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 750314fb43..1bf3be40dd 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -546,6 +546,200 @@ find_temporaries(const struct gl_program *program, } + + +/** + * Translate Mesa program to TGSI format. + * \param program the program to translate + * \param numInputs number of input registers used + * \param inputMapping maps Mesa fragment program inputs to TGSI generic + * input indexes + * \param inputSemanticName the TGSI_SEMANTIC flag for each input + * \param inputSemanticIndex the semantic index (ex: which texcoord) for each input + * \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input + + * \param numOutputs number of output registers used + * \param outputMapping maps Mesa fragment program outputs to TGSI + * generic outputs + * \param outputSemanticName the TGSI_SEMANTIC flag for each output + * \param outputSemanticIndex the semantic index (ex: which texcoord) for each output + * \param tokens array to store translated tokens in + * \param maxTokens size of the tokens array + * + */ +#if 0 +static GLboolean +tgsi_translate_program( + uint procType, + const struct gl_program *program, + GLuint numInputs, + const GLuint inputMapping[], + const ubyte inputSemanticName[], + const ubyte inputSemanticIndex[], + const GLuint interpMode[], + GLuint numOutputs, + const GLuint outputMapping[], + const ubyte outputSemanticName[], + const ubyte outputSemanticIndex[], + struct tgsi_token *tokens, + GLuint maxTokens ) +{ + GLuint i; + GLuint ti; /* token index */ + struct tgsi_header *header; + struct tgsi_processor *processor; + struct tgsi_full_instruction fullinst; + GLuint preamble_size = 0; + + assert(procType == TGSI_PROCESSOR_FRAGMENT || + procType == TGSI_PROCESSOR_VERTEX); + + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header(); + + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor( procType, header ); + + ti = 3; + + /* + * Declare input attributes. + */ + if (procType == TGSI_PROCESSOR_FRAGMENT) { + for (i = 0; i < numInputs; i++) { + struct tgsi_full_declaration fulldecl; + switch (inputSemanticName[i]) { + case TGSI_SEMANTIC_POSITION: + /* Fragment XY pos */ + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_CONSTANT, + TGSI_WRITEMASK_XY, + GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + /* Fragment ZW pos */ + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_LINEAR, + TGSI_WRITEMASK_ZW, + GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + default: + fulldecl = make_input_decl(i, + interpMode[i], + TGSI_WRITEMASK_XYZW, + GL_TRUE, inputSemanticName[i], + inputSemanticIndex[i]); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + } + } + } + else { + /* vertex prog */ + for (i = 0; i < numInputs; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_ATTRIB, + TGSI_WRITEMASK_XYZW, + GL_FALSE, inputSemanticName[i], + inputSemanticIndex[i]); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + + /* + * Declare output attributes. + */ + if (procType == TGSI_PROCESSOR_FRAGMENT) { + for (i = 0; i < numOutputs; i++) { + struct tgsi_full_declaration fulldecl; + switch (outputSemanticName[i]) { + case TGSI_SEMANTIC_POSITION: + fulldecl = make_output_decl(i, + TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ + TGSI_WRITEMASK_Z ); + break; + case TGSI_SEMANTIC_COLOR: + fulldecl = make_output_decl(i, + TGSI_SEMANTIC_COLOR, 0, + TGSI_WRITEMASK_XYZW ); + break; + default: + abort(); + } + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + else { + /* vertex prog */ + for (i = 0; i < numOutputs; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_output_decl(i, + outputSemanticName[i], + outputSemanticIndex[i], + TGSI_WRITEMASK_XYZW ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + + /* temporary decls */ + { + GLuint tempsUsed[MAX_PROGRAM_TEMPS]; + uint numTemps = find_temporaries(program, tempsUsed); + for (i = 0; i < numTemps; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_temp_decl(tempsUsed[i]); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + + for( i = 0; i < program->NumInstructions; i++ ) { + compile_instruction( + &program->Instructions[i], + &fullinst, + inputMapping, + outputMapping, + preamble_size, + procType ); + + ti += tgsi_build_full_instruction( + &fullinst, + &tokens[ti], + header, + maxTokens - ti ); + } + + return GL_TRUE; +} +#endif + + + /** * Convert Mesa fragment program to TGSI format. * \param inputMapping maps Mesa fragment program inputs to TGSI generic -- cgit v1.2.3 From 013ce15595df19005996046ed4f9f42771299d14 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 09:43:37 -0600 Subject: remove unused var --- src/mesa/state_tracker/st_cb_program.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 23a7bf473e..26609e9645 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -88,8 +88,6 @@ static struct gl_program *st_new_program( GLcontext *ctx, GLenum target, GLuint id ) { - struct st_context *st = st_context(ctx); - switch (target) { case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *prog = CALLOC_STRUCT(st_vertex_program); -- cgit v1.2.3 From 44200421e8e15e603464e4a3e9d10f46787fc737 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 09:44:54 -0600 Subject: Merge tgsi_mesa_compile_fp_program() and tgsi_mesa_compile_vp_program() into tgsi_translate_mesa_program(). --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 288 ++------------------------------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 19 +-- src/mesa/state_tracker/st_program.c | 56 +++---- 3 files changed, 45 insertions(+), 318 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 1bf3be40dd..8975b4fd57 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -39,10 +39,10 @@ map_register_file( /** * Map mesa register file index to TGSI index. * Take special care when processing input and output indices. - * \param processor either TGSI_PROCESSOR_FRAGMENT or TGSI_PROCESSOR_VERTEX * \param file one of TGSI_FILE_x * \param index the mesa register file index - * \param usage_bitmask ??? + * \param inputMapping maps Mesa input indexes to TGSI input indexes + * \param outputMapping maps Mesa output indexes to TGSI output indexes */ static GLuint map_register_file_index( @@ -443,6 +443,9 @@ compile_instruction( } } +/** + * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens + */ static struct tgsi_full_declaration make_input_decl( GLuint index, @@ -450,7 +453,7 @@ make_input_decl( GLuint usage_mask, GLboolean semantic_info, GLuint semantic_name, - GLuint semantic_index ) + GLbitfield semantic_index ) { struct tgsi_full_declaration decl; @@ -473,12 +476,15 @@ make_input_decl( return decl; } +/** + * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens + */ static struct tgsi_full_declaration make_output_decl( GLuint index, GLuint semantic_name, GLuint semantic_index, - GLuint usage_mask ) + GLbitfield usage_mask ) { struct tgsi_full_declaration decl; @@ -567,10 +573,9 @@ find_temporaries(const struct gl_program *program, * \param maxTokens size of the tokens array * */ -#if 0 -static GLboolean -tgsi_translate_program( - uint procType, +GLboolean +tgsi_translate_mesa_program( + uint procType, const struct gl_program *program, GLuint numInputs, const GLuint inputMapping[], @@ -736,271 +741,4 @@ tgsi_translate_program( return GL_TRUE; } -#endif - - - -/** - * Convert Mesa fragment program to TGSI format. - * \param inputMapping maps Mesa fragment program inputs to TGSI generic - * input indexes - * \param inputSemantic the TGSI_SEMANTIC flag for each input - * \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input - * \param outputMapping maps Mesa fragment program outputs to TGSI - * generic outputs - * - */ -GLboolean -tgsi_mesa_compile_fp_program( - const struct gl_fragment_program *program, - GLuint numInputs, - const GLuint inputMapping[], - const ubyte inputSemanticName[], - const ubyte inputSemanticIndex[], - const GLuint interpMode[], - GLuint numOutputs, - const GLuint outputMapping[], - const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - struct tgsi_token *tokens, - GLuint maxTokens ) -{ - GLuint i; - GLuint ti; /* token index */ - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_instruction fullinst; - GLuint preamble_size = 0; - - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( TGSI_PROCESSOR_FRAGMENT, header ); - - ti = 3; - - for (i = 0; i < numInputs; i++) { - struct tgsi_full_declaration fulldecl; - switch (inputSemanticName[i]) { - case TGSI_SEMANTIC_POSITION: - /* Fragment XY pos */ - fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_CONSTANT, - TGSI_WRITEMASK_XY, - GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - /* Fragment ZW pos */ - fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_LINEAR, - TGSI_WRITEMASK_ZW, - GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - default: - fulldecl = make_input_decl(i, - interpMode[i], - TGSI_WRITEMASK_XYZW, - GL_TRUE, inputSemanticName[i], - inputSemanticIndex[i]); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - } - } - - - /* - * Declare output attributes. - */ - for (i = 0; i < numOutputs; i++) { - struct tgsi_full_declaration fulldecl; - switch (outputSemanticName[i]) { - case TGSI_SEMANTIC_POSITION: - fulldecl = make_output_decl(i, - TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ - TGSI_WRITEMASK_Z ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - case TGSI_SEMANTIC_COLOR: - fulldecl = make_output_decl(i, - TGSI_SEMANTIC_COLOR, 0, - TGSI_WRITEMASK_XYZW ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - default: - abort(); - } - } - - { - GLuint tempsUsed[MAX_PROGRAM_TEMPS]; - uint numTemps = find_temporaries(&program->Base, tempsUsed); - for (i = 0; i < numTemps; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_temp_decl(tempsUsed[i]); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - - /* - * Copy fragment z if the shader does not write it. - */ -#if 0 - if( !(program->Base.OutputsWritten & (1 << FRAG_RESULT_DEPR)) ) { - fullinst = tgsi_default_full_instruction(); - - fullinst.Instruction.Opcode = TGSI_OPCODE_MOV; - fullinst.Instruction.NumDstRegs = 1; - fullinst.Instruction.NumSrcRegs = 1; - - fulldst = &fullinst.FullDstRegisters[0]; - fulldst->DstRegister.File = TGSI_FILE_OUTPUT; - fulldst->DstRegister.Index = 0; - fulldst->DstRegister.WriteMask = TGSI_WRITEMASK_Z; - - fullsrc = &fullinst.FullSrcRegisters[0]; - fullsrc->SrcRegister.File = TGSI_FILE_INPUT; - fullsrc->SrcRegister.Index = 0; - - ti += tgsi_build_full_instruction( - &fullinst, - &tokens[ti], - header, - maxTokens - ti ); - preamble_size++; - } -#endif - - for( i = 0; i < program->Base.NumInstructions; i++ ) { - compile_instruction( - &program->Base.Instructions[i], - &fullinst, - inputMapping, - outputMapping, - preamble_size, - TGSI_PROCESSOR_FRAGMENT ); - - ti += tgsi_build_full_instruction( - &fullinst, - &tokens[ti], - header, - maxTokens - ti ); - } - - return GL_TRUE; -} - -GLboolean -tgsi_mesa_compile_vp_program( - const struct gl_vertex_program *program, - GLuint numInputs, - const GLuint inputMapping[], - const ubyte inputSemanticName[], - const ubyte inputSemanticIndex[], - GLuint numOutputs, - const GLuint outputMapping[], - const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - struct tgsi_token *tokens, - GLuint maxTokens) -{ - GLuint i, ti; - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_instruction fullinst; - - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( TGSI_PROCESSOR_VERTEX, header ); - - ti = 3; - - /* input decls */ - for (i = 0; i < numInputs; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_ATTRIB, - TGSI_WRITEMASK_XYZW, - GL_FALSE, inputSemanticName[i], - inputSemanticIndex[i]); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - - /* output decls */ - for (i = 0; i < numOutputs; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_output_decl(i, - outputSemanticName[i], - outputSemanticIndex[i], - TGSI_WRITEMASK_XYZW ); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - - { - GLuint tempsUsed[MAX_PROGRAM_TEMPS]; - uint numTemps = find_temporaries(&program->Base, tempsUsed); - for (i = 0; i < numTemps; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_temp_decl(tempsUsed[i]); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - - for( i = 0; i < program->Base.NumInstructions; i++ ) { - compile_instruction( - &program->Base.Instructions[i], - &fullinst, - inputMapping, - outputMapping, - 0, - TGSI_PROCESSOR_VERTEX ); - - ti += tgsi_build_full_instruction( - &fullinst, - &tokens[ti], - header, - maxTokens - ti ); - } - - return GL_TRUE; -} diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h index 29e8053369..13372d75fd 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h @@ -8,8 +8,9 @@ extern "C" { struct tgsi_token; GLboolean -tgsi_mesa_compile_fp_program( - const struct gl_fragment_program *program, +tgsi_translate_mesa_program( + uint procType, + const struct gl_program *program, GLuint numInputs, const GLuint inputMapping[], const ubyte inputSemanticName[], @@ -22,20 +23,6 @@ tgsi_mesa_compile_fp_program( struct tgsi_token *tokens, GLuint maxTokens ); -GLboolean -tgsi_mesa_compile_vp_program( - const struct gl_vertex_program *program, - GLuint numInputs, - const GLuint inputMapping[], - const ubyte inputSemanticName[], - const ubyte inputSemanticIndex[], - GLuint numOutputs, - const GLuint outputMapping[], - const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - struct tgsi_token *tokens, - GLuint maxTokens ); - #if defined __cplusplus } // extern "C" diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 1c5af7b75f..5b6fa70dff 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -233,19 +233,21 @@ st_translate_vertex_program(struct st_context *st, /* XXX: fix static allocation of tokens: */ - tgsi_mesa_compile_vp_program( &stvp->Base, - /* inputs */ - vs.num_inputs, - stvp->input_to_index, - vs.input_semantic_name, - vs.input_semantic_index, - /* outputs */ - vs.num_outputs, - outputMapping, - vs.output_semantic_name, + tgsi_translate_mesa_program( TGSI_PROCESSOR_VERTEX, + &stvp->Base.Base, + /* inputs */ + vs.num_inputs, + stvp->input_to_index, + vs.input_semantic_name, + vs.input_semantic_index, + NULL, + /* outputs */ + vs.num_outputs, + outputMapping, + vs.output_semantic_name, vs.output_semantic_index, - /* tokenized result */ - tokensOut, maxTokens); + /* tokenized result */ + tokensOut, maxTokens); vs.tokens = tokensOut; cso = st_cached_vs_state(st, &vs); @@ -386,21 +388,21 @@ st_translate_fragment_program(struct st_context *st, /* XXX: fix static allocation of tokens: */ - tgsi_mesa_compile_fp_program( &stfp->Base, - /* inputs */ - fs.num_inputs, - inputMapping, - fs.input_semantic_name, - fs.input_semantic_index, - interpMode, - /* outputs */ - fs.num_outputs, - outputMapping, - fs.output_semantic_name, - fs.output_semantic_index, - /* tokenized result */ - tokensOut, maxTokens); - + tgsi_translate_mesa_program( TGSI_PROCESSOR_FRAGMENT, + &stfp->Base.Base, + /* inputs */ + fs.num_inputs, + inputMapping, + fs.input_semantic_name, + fs.input_semantic_index, + interpMode, + /* outputs */ + fs.num_outputs, + outputMapping, + fs.output_semantic_name, + fs.output_semantic_index, + /* tokenized result */ + tokensOut, maxTokens); fs.tokens = tokensOut; -- cgit v1.2.3 From fa7a5898761fc106c9dbef6f640ce4654eae5136 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 10:33:38 -0600 Subject: Size of input_to_index array should be VERT_ATTRIB_MAX. This fixes an out of bounds array write that was causing the glsl/bump demo to render incorrectly. --- src/mesa/state_tracker/st_program.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index ae89055e82..51877af8da 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -82,7 +82,7 @@ struct st_vertex_program GLuint serialNo; /** maps a Mesa VERT_ATTRIB_x to a packed TGSI input index */ - GLuint input_to_index[MAX_VERTEX_PROGRAM_ATTRIBS]; + GLuint input_to_index[VERT_ATTRIB_MAX]; /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ GLuint index_to_input[PIPE_MAX_SHADER_INPUTS]; -- cgit v1.2.3 From 500e3af175cf8ef66bad23ae3b9e440670421ecd Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 11:04:48 -0600 Subject: fix width/height padding that caused failed assertion upon window resize --- src/mesa/state_tracker/st_cb_fbo.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 5b4afbd119..4a21ff5371 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -64,6 +64,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, const struct pipe_format_info *info = st_get_format_info(pipeFormat); GLuint cpp; GLbitfield flags = PIPE_SURFACE_FLAG_RENDER; /* want to render to surface */ + GLuint width2, height2; assert(info); if (!info) @@ -100,10 +101,10 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } /* Softpipe operates on quads, so pad dimensions to multiples of 2 */ - width += width & 1; - height += height & 1; + width2 = (width + 1) & ~1; + height2 = (height + 1) & ~1; - strb->surface->region = pipe->region_alloc(pipe, cpp, width, height, flags); + strb->surface->region = pipe->region_alloc(pipe, cpp, width2, height2, flags); if (!strb->surface->region) return GL_FALSE; /* out of memory, try s/w buffer? */ -- cgit v1.2.3 From e23a7b2ea4a311af6b661f10747716e28fa451db Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 11:40:04 -0600 Subject: use pipe_region_reference() in softpipe_get_tex_surface() to fix refcount error --- src/mesa/pipe/softpipe/sp_surface.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 574b474ba3..27d4f064be 100755 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -882,7 +882,7 @@ softpipe_get_tex_surface(struct pipe_context *pipe, if (ps) { assert(ps->format); assert(ps->refcount); - ps->region = mt->region; + pipe_region_reference(&ps->region, mt->region); ps->width = mt->level[level].width; ps->height = mt->level[level].height; ps->offset = offset; -- cgit v1.2.3 From 2b31b413f97c73816a2845782aee9a49e22d2e0b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 14:13:21 -0600 Subject: use correct mipmap level in st_render_texture() --- src/mesa/state_tracker/st_cb_fbo.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 4a21ff5371..3e4aeab523 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -316,10 +316,10 @@ st_render_texture(GLcontext *ctx, /* get the mipmap tree for the texture */ mt = st_get_texobj_mipmap_tree(att->Texture); assert(mt); - assert(mt->level[0].width); + assert(mt->level[att->TextureLevel].width); - rb->Width = mt->level[0].width; - rb->Height = mt->level[0].height; + rb->Width = mt->level[att->TextureLevel].width; + rb->Height = mt->level[att->TextureLevel].height; /* the renderbuffer's surface is inside the mipmap_tree: */ strb->surface = pipe->get_tex_surface(pipe, mt, -- cgit v1.2.3 From 8b68104b90b1d136b5f67a196c6cbdf9fe4dbfe1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 15:05:24 -0600 Subject: free/unreference pipe_surface in get_texel() --- src/mesa/pipe/softpipe/sp_tex_sample.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 0c990294f2..b89adf8480 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -594,6 +594,8 @@ get_texel(struct tgsi_sampler *sampler, ty * TEX_CACHE_TILE_SIZE, TEX_CACHE_TILE_SIZE, TEX_CACHE_TILE_SIZE, (float *) sampler->cache[entry].data); + + pipe_surface_reference(&ps, NULL); } else { /* -- cgit v1.2.3 From 9f8cfa75ca0882a8015a714887b41f8a7a10fd19 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 15:09:57 -0600 Subject: fix softpipe_clear() to handle ps->offset!=0 (such as when rendering to texture and level!=0) --- src/mesa/pipe/softpipe/sp_clear.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 2d5344f424..539da1d58b 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -50,11 +50,17 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, softpipe_update_derived(softpipe); /* not needed?? */ - x = 0; - y = 0; w = softpipe->framebuffer.cbufs[0]->width; h = softpipe->framebuffer.cbufs[0]->height; + /* Use the X coord to trick region_fill() into filling at an offset + * from the start of the region. Perhaps pipe_region should have the + * 'offset' field, not pipe_surface??? + */ + assert(ps->offset % ps->region->cpp == 0); + x = ps->offset / ps->region->cpp; + y = 0; + assert(w <= ps->region->pitch); assert(h <= ps->region->height); -- cgit v1.2.3 From d0bca086ab6d032909e9a429720fea297c536f97 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 15:45:57 -0600 Subject: fix get_result_vector() since fp output slots are not fixed anymore --- src/mesa/pipe/i915simple/i915_fpc.h | 3 +++ src/mesa/pipe/i915simple/i915_fpc_translate.c | 30 ++++++++++++++++++++------- 2 files changed, 25 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index 5fdc7ef27e..c25f5abf0e 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -54,6 +54,9 @@ struct i915_fp_compile { uint input_semantic_name[PIPE_MAX_SHADER_INPUTS]; uint input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + uint output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; + uint output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; + /** points into the i915->current.constants array: */ float (*constants)[4]; uint num_constants; diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 24bb1d3e2a..f34225bfed 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -240,14 +240,17 @@ get_result_vector(struct i915_fp_compile *p, { switch (dest->DstRegister.File) { case TGSI_FILE_OUTPUT: - switch (dest->DstRegister.Index) { - case 0: /**TGSI_ATTRIB_POS:**/ - return UREG(REG_TYPE_OD, 0); - case 1: /**TGSI_ATTRIB_COLOR0:**/ - return UREG(REG_TYPE_OC, 0); - default: - i915_program_error(p, "Bad inst->DstReg.Index"); - return 0; + { + uint sem_name = p->output_semantic_name[dest->DstRegister.Index]; + switch (sem_name) { + case TGSI_SEMANTIC_POSITION: + return UREG(REG_TYPE_OD, 0); + case TGSI_SEMANTIC_COLOR: + return UREG(REG_TYPE_OC, 0); + default: + i915_program_error(p, "Bad inst->DstReg.Index/semantics"); + return 0; + } } case TGSI_FILE_TEMPORARY: return UREG(REG_TYPE_R, dest->DstRegister.Index); @@ -867,6 +870,17 @@ i915_translate_instructions(struct i915_fp_compile *p, p->input_semantic_name[ind] = sem; p->input_semantic_index[ind] = semi; } + else if (parse.FullToken.FullDeclaration.Declaration.File + == TGSI_FILE_OUTPUT) { + /* save output register info for use in get_result_vector() */ + uint ind, sem, semi; + ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; + semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; + /*printf("FS Output DECL [%u] sem %u\n", ind, sem);*/ + p->output_semantic_name[ind] = sem; + p->output_semantic_index[ind] = semi; + } break; case TGSI_TOKEN_TYPE_IMMEDIATE: -- cgit v1.2.3 From d37e048cd1e24952214267a2c33cbaf346c10b26 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 16:19:49 -0600 Subject: disable some assertions which pop up during window resize, but can be ignored for now --- src/mesa/pipe/softpipe/sp_surface.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 27d4f064be..29b5c9631c 100755 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -56,9 +56,10 @@ a8r8g8b8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, unsigned i, j; assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); +#if 0 assert(x < (int) sps->surface.width - 1); assert(y < (int) sps->surface.height - 1); - +#endif for (i = 0; i < 2; i++) { /* loop over pixel row */ for (j = 0; j < 2; j++) { /* loop over pixel column */ const unsigned p = src[j]; -- cgit v1.2.3 From babdd1f6516a21d774a15432290358640aa099d8 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 16:21:28 -0600 Subject: minor optimization for color/z results --- src/mesa/pipe/softpipe/sp_quad_fs.c | 45 ++++++++++++++++++++++++------------- 1 file changed, 30 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index b6005c6a69..c850f08c57 100755 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -47,6 +47,7 @@ struct quad_shade_stage struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; struct tgsi_exec_machine machine; struct tgsi_exec_vector *inputs, *outputs; + int colorOutSlot, depthOutSlot; }; @@ -78,7 +79,6 @@ shade_quad( const float fx = (float) quad->x0; const float fy = (float) quad->y0; struct tgsi_exec_machine *machine = &qss->machine; - uint colorOut; /* Consts does not require 16 byte alignment. */ machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; @@ -110,32 +110,33 @@ shade_quad( quad->mask &= tgsi_exec_machine_run( machine ); } - if (qss->stage.softpipe->fs->output_semantic_name[0] == TGSI_SEMANTIC_POSITION) { - /* output[0] is new Z */ + /* store result color */ + if (qss->colorOutSlot >= 0) { + /* XXX need to handle multiple color outputs someday */ + assert(qss->stage.softpipe->fs->output_semantic_name[qss->colorOutSlot] + == TGSI_SEMANTIC_COLOR); + memcpy( + quad->outputs.color, + &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); + } + + /* store result Z */ + if (qss->depthOutSlot >= 0) { + /* output[slot] is new Z */ uint i; for (i = 0; i < 4; i++) { quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; } - colorOut = 1; } else { - /* pass input Z (which was interpolated by the executor) to output Z */ + /* copy input Z (which was interpolated by the executor) to output Z */ uint i; for (i = 0; i < 4; i++) { quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; } - colorOut = 0; } - /* store result color */ - /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->output_semantic_name[colorOut] - == TGSI_SEMANTIC_COLOR); - memcpy( - quad->outputs.color, - &machine->Outputs[colorOut].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); - /* shader may cull fragments */ if( quad->mask ) { qs->next->run( qs->next, quad ); @@ -173,6 +174,20 @@ static void shade_begin(struct quad_stage *qs) PIPE_MAX_SAMPLERS, qss->samplers ); + /* find output slots for depth, color */ + qss->colorOutSlot = -1; + qss->depthOutSlot = -1; + for (i = 0; i < qss->stage.softpipe->fs->num_outputs; i++) { + switch (qss->stage.softpipe->fs->output_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + qss->depthOutSlot = i; + break; + case TGSI_SEMANTIC_COLOR: + qss->colorOutSlot = i; + break; + } + } + if (qs->next) qs->next->begin(qs->next); } -- cgit v1.2.3 From 178beffd35c886894c5d213bbe3bdde9fa47e863 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 16:51:09 -0600 Subject: update comments (s/softpipe/gallium/ etc) --- src/mesa/pipe/p_winsys.h | 38 ++++++++++++++++++-------------------- 1 file changed, 18 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 2861e03c9e..73ac61c33a 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -29,13 +29,13 @@ #define P_WINSYS_H -/* This is the interface that softpipe requires any window system - * hosting it to implement. This is the only include file in softpipe +/** + * This is the interface that Gallium3D requires any window system + * hosting it to implement. This is the only include file in Gallium3D * which is public. - */ - - -/* Pipe drivers are (meant to be!) independent of both GL and the + * + * + * Gallium3D drivers are (meant to be!) independent of both GL and the * window system. The window system provides a buffer manager and a * set of additional hooks for things like command buffer submission, * etc. @@ -45,33 +45,33 @@ * etc. */ -struct pipe_buffer_handle; -struct pipe_winsys { +/** Opaque type */ +struct pipe_buffer_handle; - /* Do any special operations to ensure frontbuffer contents are +struct pipe_winsys +{ + /** + * Do any special operations to ensure frontbuffer contents are * displayed, eg copy fake frontbuffer. */ void (*flush_frontbuffer)( struct pipe_winsys *sws ); - /* debug output - */ + /** for debug output */ void (*printf)( struct pipe_winsys *sws, const char *, ... ); /* The buffer manager is modeled after the dri_bufmgr interface, - * but this is the subset that softpipe cares about. Remember that - * softpipe gets to choose the interface it needs, and the window + * but this is the subset that gallium cares about. Remember that + * gallium gets to choose the interface it needs, and the window * systems must then implement that interface (rather than the * other way around...). - * - * Softpipe only really wants to make system memory allocations, - * right?? */ struct pipe_buffer_handle *(*buffer_create)(struct pipe_winsys *sws, unsigned alignment ); + /** flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE */ void *(*buffer_map)( struct pipe_winsys *sws, struct pipe_buffer_handle *buf, unsigned flags ); @@ -101,12 +101,10 @@ struct pipe_winsys { void *data); - /* Wait for any hw swapbuffers, etc. to finish: - */ + /** Wait for any hw swapbuffers, etc. to finish */ void (*wait_idle)( struct pipe_winsys *sws ); - /* Queries: - */ + /** Queries */ const char *(*get_name)( struct pipe_winsys *sws ); }; -- cgit v1.2.3 From 3f226d4bafcc7b174ee42e4b5a2a4fb3756758ad Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 17:50:31 -0600 Subject: added SSCALED cases in fetch_attrib4() --- src/mesa/pipe/draw/draw_vertex_fetch.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index bc37e18c34..de1cd06da6 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -62,6 +62,20 @@ fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) case PIPE_FORMAT_R32_FLOAT: attrib[0] = ((float *) ptr)[0]; break; + + case PIPE_FORMAT_R32G32B32A32_SSCALED: + attrib[3] = ((int *) ptr)[3]; + /* fall-through */ + case PIPE_FORMAT_R32G32B32_SSCALED: + attrib[2] = ((int *) ptr)[2]; + /* fall-through */ + case PIPE_FORMAT_R32G32_SSCALED: + attrib[1] = ((int *) ptr)[1]; + /* fall-through */ + case PIPE_FORMAT_R32_SSCALED: + attrib[0] = ((int *) ptr)[0]; + break; + default: assert(0); } -- cgit v1.2.3 From fcdd34de4b169dc12991f7894266ee834e9bf768 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 17:57:47 -0600 Subject: Added user_buffer_create() method This is used to wrap user-space data in a pipe buffer object to allow more uniform treatment of various things. For example, wrapping client-side vertex arrays so they look like VBOs. --- src/mesa/pipe/p_winsys.h | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 73ac61c33a..c80559c730 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -71,6 +71,12 @@ struct pipe_winsys struct pipe_buffer_handle *(*buffer_create)(struct pipe_winsys *sws, unsigned alignment ); + /** Create a buffer that wraps user-space data */ + struct pipe_buffer_handle *(*user_buffer_create)(struct pipe_winsys *sws, + void *ptr, + unsigned bytes); + + /** flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE */ void *(*buffer_map)( struct pipe_winsys *sws, struct pipe_buffer_handle *buf, -- cgit v1.2.3 From 66ab569590f3bb9232eb7a04988ed477fe7fe2c5 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 17:58:12 -0600 Subject: implement the user_buffer_create() function --- src/mesa/drivers/x11/xm_winsys.c | 28 +++++++++++++++++++++++----- 1 file changed, 23 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index b7b2941490..7e29580b50 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -58,8 +58,9 @@ struct xm_winsys */ struct xm_buffer { + boolean userBuffer; /** Is this a user-space buffer? */ int refcount; - int size; + unsigned size; void *data; void *mapped; }; @@ -119,7 +120,8 @@ xm_buffer_reference(struct pipe_winsys *pws, assert(oldBuf->refcount >= 0); if (oldBuf->refcount == 0) { if (oldBuf->data) { - free(oldBuf->data); + if (!oldBuf->userBuffer) + free(oldBuf->data); oldBuf->data = NULL; } free(oldBuf); @@ -141,6 +143,7 @@ xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, unsigned size, const void *data ) { struct xm_buffer *xm_buf = xm_bo(buf); + assert(!xm_buf->userBuffer); if (xm_buf->size != size) { if (xm_buf->data) free(xm_buf->data); @@ -157,6 +160,7 @@ xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, { struct xm_buffer *xm_buf = xm_bo(buf); GLubyte *b = (GLubyte *) xm_buf->data; + assert(!xm_buf->userBuffer); assert(b); memcpy(b + offset, data, size); } @@ -167,6 +171,7 @@ xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, { const struct xm_buffer *xm_buf = xm_bo(buf); const GLubyte *b = (GLubyte *) xm_buf->data; + assert(!xm_buf->userBuffer); assert(b); memcpy(data, b + offset, size); } @@ -204,9 +209,6 @@ xm_get_name(struct pipe_winsys *pws) } -/* Softpipe has no concept of pools. We choose the tex/region pool - * for all buffers. - */ static struct pipe_buffer_handle * xm_buffer_create(struct pipe_winsys *pws, unsigned alignment) { @@ -216,6 +218,21 @@ xm_buffer_create(struct pipe_winsys *pws, unsigned alignment) } +/** + * Create buffer which wraps user-space data. + */ +static struct pipe_buffer_handle * +xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) +{ + struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); + buffer->userBuffer = TRUE; + buffer->refcount = 1; + buffer->data = ptr; + buffer->size = bytes; + return pipe_bo(buffer); +} + + struct xmesa_pipe_winsys { struct pipe_winsys winsys; @@ -235,6 +252,7 @@ xmesa_create_pipe_winsys( XMesaContext xmesa ) * that rendering be done to an appropriate _DriBufferObject. */ xws->winsys.buffer_create = xm_buffer_create; + xws->winsys.user_buffer_create = xm_user_buffer_create; xws->winsys.buffer_map = xm_buffer_map; xws->winsys.buffer_unmap = xm_buffer_unmap; xws->winsys.buffer_reference = xm_buffer_reference; -- cgit v1.2.3 From 453ac4b1ebc70aeabb182f2b336c6abb3324323e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 10 Oct 2007 18:00:12 -0600 Subject: Use winsys->user_buffer_create() to wrap user-space vertex arrays and element buffers. Now client-side vertex arrays and glDrawElements work. More testing/debug/clean-up to come... --- src/mesa/state_tracker/st_draw.c | 72 ++++++++++++++++++++++++++++------------ 1 file changed, 50 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 7c836ba2c1..22f09c7a98 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -31,6 +31,7 @@ */ #include "main/imports.h" +#include "main/image.h" #include "vbo/vbo.h" #include "vbo/vbo_context.h" @@ -61,6 +62,12 @@ pipe_vertex_format(GLenum format, GLuint size) PIPE_FORMAT_R32G32B32_FLOAT, PIPE_FORMAT_R32G32B32A32_FLOAT, }; + static const GLuint int_fmts[4] = { + PIPE_FORMAT_R32_SSCALED, + PIPE_FORMAT_R32G32_SSCALED, + PIPE_FORMAT_R32G32B32_SSCALED, + PIPE_FORMAT_R32G32B32A32_SSCALED, + }; assert(format >= GL_BYTE); assert(format <= GL_DOUBLE); @@ -70,6 +77,8 @@ pipe_vertex_format(GLenum format, GLuint size) switch (format) { case GL_FLOAT: return float_fmts[size - 1]; + case GL_INT: + return int_fmts[size - 1]; default: assert(0); } @@ -215,6 +224,7 @@ st_draw_vbo(GLcontext *ctx, velement.src_format = 0; if (bufobj && bufobj->Name) { + /* attribute data is in a VBO */ struct st_buffer_object *stobj = st_buffer_object(bufobj); /* Recall that for VBOs, the gl_client_array->Ptr field is * really an offset from the start of the VBO, not a pointer. @@ -236,23 +246,30 @@ st_draw_vbo(GLcontext *ctx, assert(velement.src_format); } else { - /* use the default attribute buffer */ - needDefaultAttribs = GL_TRUE; + /* attribute data is in user-space memory, not a VBO */ + uint bytes = (arrays[mesaAttr]->Size + * _mesa_sizeof_type(arrays[mesaAttr]->Type) + * (max_index + 1)); + + /* wrap user data */ + vbuffer.buffer + = pipe->winsys->user_buffer_create(pipe->winsys, + (void *) arrays[mesaAttr]->Ptr, + bytes); + + /* XXX need to deref/free this buffer.vbuffer after drawing! */ - vbuffer.buffer = ctx->st->default_attrib_buffer; vbuffer.buffer_offset = 0; - vbuffer.pitch = 0; /* must be zero! */ - vbuffer.max_index = 1; + vbuffer.pitch = arrays[mesaAttr]->StrideB; /* in bytes */ + vbuffer.max_index = 0; /* need this? */ - velement.src_offset = mesaAttr * 4 * sizeof(GLfloat); + velement.src_offset = 0; velement.vertex_buffer_index = attr; - velement.dst_offset = 0; - velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + velement.dst_offset = 0; /* need this? */ + velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, + arrays[mesaAttr]->Size); } - if (attr == 0) - assert(vbuffer.buffer); - pipe->set_vertex_buffer(pipe, attr, &vbuffer); pipe->set_vertex_element(pipe, attr, &velement); } @@ -268,30 +285,41 @@ st_draw_vbo(GLcontext *ctx, struct gl_buffer_object *bufobj = ib->obj; struct pipe_buffer_handle *bh = NULL; unsigned indexSize, i; + GLboolean userBuffer = GL_FALSE; + + switch (ib->type) { + case GL_UNSIGNED_INT: + indexSize = 4; + break; + case GL_UNSIGNED_SHORT: + indexSize = 2; + break; + default: + assert(0); + } if (bufobj && bufobj->Name) { /* elements/indexes are in a real VBO */ struct st_buffer_object *stobj = st_buffer_object(bufobj); bh = stobj->buffer; - switch (ib->type) { - case GL_UNSIGNED_INT: - indexSize = 4; - break; - case GL_UNSIGNED_SHORT: - indexSize = 2; - break; - default: - assert(0); - } + /* XXX reference buffer here, don't special case userBuffer below */ } else { - assert(0); + /* element/indicies are in user space memory */ + bh = pipe->winsys->user_buffer_create(pipe->winsys, + (void *) ib->ptr, + ib->count * indexSize); + userBuffer = GL_TRUE; } for (i = 0; i < nr_prims; i++) { pipe->draw_elements(pipe, bh, indexSize, prims[i].mode, prims[i].start, prims[i].count); } + + if (userBuffer) { + pipe->winsys->buffer_reference(pipe->winsys, &bh, NULL); + } } else { /* non-indexed */ -- cgit v1.2.3 From cdc156d52f4ac33898b4a0e4930a1a9ebe3c81cf Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 11 Oct 2007 09:10:19 -0600 Subject: Initial support for user-space buffer objects. Basically, add userBuffer/Data/Size fields to _DriBufferObject, check those fields in driBOMap/Unmap(). New driGenUserBuffer() function. --- src/mesa/drivers/dri/common/dri_bufmgr.c | 54 ++++++++++++++++++++++++++------ src/mesa/drivers/dri/common/dri_bufmgr.h | 4 +++ 2 files changed, 49 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/common/dri_bufmgr.c b/src/mesa/drivers/dri/common/dri_bufmgr.c index eaa4fb09c7..5747307f3b 100644 --- a/src/mesa/drivers/dri/common/dri_bufmgr.c +++ b/src/mesa/drivers/dri/common/dri_bufmgr.c @@ -67,6 +67,10 @@ typedef struct _DriBufferObject unsigned hint; unsigned alignment; void *private; + /* user-space buffer: */ + unsigned userBuffer; + void *userData; + unsigned userSize; } DriBufferObject; @@ -205,22 +209,29 @@ driBOWaitIdle(struct _DriBufferObject *buf, int lazy) void * driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint) { - void *virtual; + if (buf->userBuffer) { + return buf->userData; + } + else { + void *virtual; - assert(buf->private != NULL); + assert(buf->private != NULL); - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(buf->pool->map(buf->pool, buf->private, flags, hint, &virtual)); - _glthread_UNLOCK_MUTEX(buf->mutex); - return virtual; + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(buf->pool->map(buf->pool, buf->private, flags, hint, &virtual)); + _glthread_UNLOCK_MUTEX(buf->mutex); + return virtual; + } } void driBOUnmap(struct _DriBufferObject *buf) { - assert(buf->private != NULL); + if (!buf->userBuffer) { + assert(buf->private != NULL); - buf->pool->unmap(buf->pool, buf->private); + buf->pool->unmap(buf->pool, buf->private); + } } unsigned long @@ -272,7 +283,8 @@ driBOUnReference(struct _DriBufferObject *buf) tmp = --buf->refCount; _glthread_UNLOCK_MUTEX(bmMutex); if (!tmp) { - buf->pool->destroy(buf->pool, buf->private); + if (buf->private) + buf->pool->destroy(buf->pool, buf->private); free(buf); } } @@ -285,6 +297,8 @@ driBOData(struct _DriBufferObject *buf, int newBuffer; struct _DriBufferPool *pool; + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + _glthread_LOCK_MUTEX(buf->mutex); pool = buf->pool; if (!pool->create) { @@ -323,6 +337,8 @@ driBOSubData(struct _DriBufferObject *buf, { void *virtual; + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + _glthread_LOCK_MUTEX(buf->mutex); if (size && data) { BM_CKFATAL(buf->pool->map(buf->pool, buf->private, @@ -339,6 +355,8 @@ driBOGetSubData(struct _DriBufferObject *buf, { void *virtual; + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + _glthread_LOCK_MUTEX(buf->mutex); if (size && data) { BM_CKFATAL(buf->pool->map(buf->pool, buf->private, @@ -354,6 +372,8 @@ driBOSetStatic(struct _DriBufferObject *buf, unsigned long offset, unsigned long size, void *virtual, unsigned flags) { + assert(!buf->userBuffer); /* XXX what to do? */ + _glthread_LOCK_MUTEX(buf->mutex); if (buf->private != NULL) { _mesa_error(NULL, GL_INVALID_OPERATION, @@ -415,6 +435,22 @@ driGenBuffers(struct _DriBufferPool *pool, } } +void +driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject **buffers, + void *ptr, unsigned bytes) +{ + const unsigned alignment = 1, flags = 0, hint = 0; + + driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); + + (*buffers)->userBuffer = 1; + (*buffers)->userData = ptr; + (*buffers)->userSize = bytes; +} + + void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]) { diff --git a/src/mesa/drivers/dri/common/dri_bufmgr.h b/src/mesa/drivers/dri/common/dri_bufmgr.h index 01f149ae4e..ee4ce2cbde 100644 --- a/src/mesa/drivers/dri/common/dri_bufmgr.h +++ b/src/mesa/drivers/dri/common/dri_bufmgr.h @@ -78,6 +78,10 @@ extern void driGenBuffers(struct _DriBufferPool *pool, unsigned n, struct _DriBufferObject *buffers[], unsigned alignment, unsigned flags, unsigned hint); +extern void driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject *buffers[], + void *ptr, unsigned bytes); extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]); extern void driInitBufMgr(int fd); extern void driBOCreateList(int target, drmBOList * list); -- cgit v1.2.3 From 98e638aa8df9f5b789b68a7a742666a4b06b6cb3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 11 Oct 2007 09:10:43 -0600 Subject: implement intel_user_buffer_create() --- src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 2b2731446d..72fb24acbf 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -151,6 +151,21 @@ intel_buffer_create(struct pipe_winsys *sws, } +static struct pipe_buffer_handle * +intel_user_buffer_create(struct pipe_winsys *sws, void *ptr, unsigned bytes) +{ + struct intel_context *intel = intel_pipe_winsys(sws)->intel; + struct _DriBufferObject *buffer; + + LOCK_HARDWARE( intel ); + driGenUserBuffer( intel->intelScreen->regionPool, + "pipe user buffer", &buffer, ptr, bytes); + UNLOCK_HARDWARE( intel ); + + return pipe_bo(buffer); +} + + static void intel_wait_idle( struct pipe_winsys *sws ) { struct intel_context *intel = intel_pipe_winsys(sws)->intel; @@ -206,6 +221,7 @@ intel_create_pipe_winsys( struct intel_context *intel ) * that rendering be done to an appropriate _DriBufferObject. */ iws->winsys.buffer_create = intel_buffer_create; + iws->winsys.user_buffer_create = intel_user_buffer_create; iws->winsys.buffer_map = intel_buffer_map; iws->winsys.buffer_unmap = intel_buffer_unmap; iws->winsys.buffer_reference = intel_buffer_reference; -- cgit v1.2.3 From 71efb6cf69724da2f1e783a13c3074d4e8645c64 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 11 Oct 2007 09:11:13 -0600 Subject: clean-up, streamline st_draw_vbo() --- src/mesa/state_tracker/st_draw.c | 26 ++++++++++++-------------- 1 file changed, 12 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 22f09c7a98..bfaa352fdc 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -197,6 +197,7 @@ st_draw_vbo(GLcontext *ctx, GLuint max_index) { struct pipe_context *pipe = ctx->st->pipe; + struct pipe_winsys *winsys = pipe->winsys; const struct st_vertex_program *vp; const struct pipe_shader_state *vs; const unsigned attr0_offset = (unsigned) arrays[0]->Ptr; @@ -253,9 +254,9 @@ st_draw_vbo(GLcontext *ctx, /* wrap user data */ vbuffer.buffer - = pipe->winsys->user_buffer_create(pipe->winsys, - (void *) arrays[mesaAttr]->Ptr, - bytes); + = winsys->user_buffer_create(winsys, + (void *) arrays[mesaAttr]->Ptr, + bytes); /* XXX need to deref/free this buffer.vbuffer after drawing! */ @@ -283,9 +284,8 @@ st_draw_vbo(GLcontext *ctx, if (ib) { /* indexed primitive */ struct gl_buffer_object *bufobj = ib->obj; - struct pipe_buffer_handle *bh = NULL; + struct pipe_buffer_handle *indexBuf = NULL; unsigned indexSize, i; - GLboolean userBuffer = GL_FALSE; switch (ib->type) { case GL_UNSIGNED_INT: @@ -298,28 +298,26 @@ st_draw_vbo(GLcontext *ctx, assert(0); } + /* get/create the index buffer object */ if (bufobj && bufobj->Name) { /* elements/indexes are in a real VBO */ struct st_buffer_object *stobj = st_buffer_object(bufobj); - bh = stobj->buffer; - /* XXX reference buffer here, don't special case userBuffer below */ + winsys->buffer_reference(winsys, &indexBuf, stobj->buffer); } else { /* element/indicies are in user space memory */ - bh = pipe->winsys->user_buffer_create(pipe->winsys, + indexBuf = winsys->user_buffer_create(winsys, (void *) ib->ptr, ib->count * indexSize); - userBuffer = GL_TRUE; } + /* draw */ for (i = 0; i < nr_prims; i++) { - pipe->draw_elements(pipe, bh, indexSize, - prims[i].mode, prims[i].start, prims[i].count); + pipe->draw_elements(pipe, indexBuf, indexSize, + prims[i].mode, prims[i].start, prims[i].count); } - if (userBuffer) { - pipe->winsys->buffer_reference(pipe->winsys, &bh, NULL); - } + winsys->buffer_reference(winsys, &indexBuf, NULL); } else { /* non-indexed */ -- cgit v1.2.3 From 4d79f2c4a2ef87595558051b3bdaeb54b4bd0339 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 11 Oct 2007 09:48:06 -0600 Subject: streamline st_draw_vbo() a bit more, free user-space buffers at end --- src/mesa/state_tracker/st_draw.c | 76 +++++++++++++++++----------------------- 1 file changed, 32 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index bfaa352fdc..f8a3440bc9 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -200,9 +200,11 @@ st_draw_vbo(GLcontext *ctx, struct pipe_winsys *winsys = pipe->winsys; const struct st_vertex_program *vp; const struct pipe_shader_state *vs; - const unsigned attr0_offset = (unsigned) arrays[0]->Ptr; - GLboolean needDefaultAttribs = GL_FALSE; GLuint attr; + struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; + + /* sanity check for pointer arithmetic below */ + assert(sizeof(arrays[0]->Ptr[0]) == 1); st_validate_state(ctx->st); @@ -210,41 +212,26 @@ st_draw_vbo(GLcontext *ctx, vp = ctx->st->vp; vs = &ctx->st->state.vs->state; - /* loop over TGSI shader inputs */ + /* loop over TGSI shader inputs to determine vertex buffer + * and attribute info + */ for (attr = 0; attr < vs->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; - - struct pipe_vertex_buffer vbuffer; struct pipe_vertex_element velement; - vbuffer.buffer = NULL; - vbuffer.pitch = 0; - velement.src_offset = 0; - velement.vertex_buffer_index = 0; - velement.src_format = 0; - if (bufobj && bufobj->Name) { - /* attribute data is in a VBO */ - struct st_buffer_object *stobj = st_buffer_object(bufobj); - /* Recall that for VBOs, the gl_client_array->Ptr field is + /* Attribute data is in a VBO. + * Recall that for VBOs, the gl_client_array->Ptr field is * really an offset from the start of the VBO, not a pointer. */ - unsigned offset = (unsigned) arrays[mesaAttr]->Ptr; - + struct st_buffer_object *stobj = st_buffer_object(bufobj); assert(stobj->buffer); - vbuffer.buffer = stobj->buffer; - vbuffer.buffer_offset = attr0_offset; /* in bytes */ - vbuffer.pitch = arrays[mesaAttr]->StrideB; /* in bytes */ - vbuffer.max_index = 0; /* need this? */ - - velement.src_offset = offset - attr0_offset; /* bytes */ - velement.vertex_buffer_index = attr; - velement.dst_offset = 0; /* need this? */ - velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, - arrays[mesaAttr]->Size); - assert(velement.src_format); + vbuffer[attr].buffer = NULL; + winsys->buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); + vbuffer[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ + velement.src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; } else { /* attribute data is in user-space memory, not a VBO */ @@ -253,32 +240,28 @@ st_draw_vbo(GLcontext *ctx, * (max_index + 1)); /* wrap user data */ - vbuffer.buffer + vbuffer[attr].buffer = winsys->user_buffer_create(winsys, (void *) arrays[mesaAttr]->Ptr, bytes); - - /* XXX need to deref/free this buffer.vbuffer after drawing! */ - - vbuffer.buffer_offset = 0; - vbuffer.pitch = arrays[mesaAttr]->StrideB; /* in bytes */ - vbuffer.max_index = 0; /* need this? */ - + vbuffer[attr].buffer_offset = 0; velement.src_offset = 0; - velement.vertex_buffer_index = attr; - velement.dst_offset = 0; /* need this? */ - velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, - arrays[mesaAttr]->Size); } - pipe->set_vertex_buffer(pipe, attr, &vbuffer); + /* common-case setup */ + vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ + vbuffer[attr].max_index = 0; /* need this? */ + velement.vertex_buffer_index = attr; + velement.dst_offset = 0; /* need this? */ + velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, + arrays[mesaAttr]->Size); + assert(velement.src_format); + + /* tell pipe about this attribute */ + pipe->set_vertex_buffer(pipe, attr, &vbuffer[attr]); pipe->set_vertex_element(pipe, attr, &velement); } - if (needDefaultAttribs) { - update_default_attribs_buffer(ctx); - } - /* do actual drawing */ if (ib) { @@ -326,6 +309,11 @@ st_draw_vbo(GLcontext *ctx, pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count); } } + + /* unreference buffers (frees wrapped user-space buffer objects) */ + for (attr = 0; attr < vs->num_inputs; attr++) { + winsys->buffer_reference(winsys, &vbuffer[attr].buffer, NULL); + } } -- cgit v1.2.3 From 55dd7d794719ebdbf62b58cc15e022ce1e430873 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 11 Oct 2007 10:00:54 -0600 Subject: update st_feedback_draw_vbo() --- src/mesa/state_tracker/st_draw.c | 138 ++++++++++++++++----------------------- 1 file changed, 58 insertions(+), 80 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index f8a3440bc9..aa7fedf9b8 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -171,17 +171,6 @@ destroy_default_attribs_buffer(struct st_context *st) } -static void -update_default_attribs_buffer(GLcontext *ctx) -{ - struct pipe_context *pipe = ctx->st->pipe; - struct pipe_buffer_handle *buf = ctx->st->default_attrib_buffer; - const unsigned size = sizeof(ctx->Current.Attrib); - const void *data = ctx->Current.Attrib; - pipe->winsys->buffer_data(pipe->winsys, buf, size, data); -} - - /** * This function gets plugged into the VBO module and is called when * we have something to render. @@ -200,8 +189,8 @@ st_draw_vbo(GLcontext *ctx, struct pipe_winsys *winsys = pipe->winsys; const struct st_vertex_program *vp; const struct pipe_shader_state *vs; - GLuint attr; struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; + GLuint attr; /* sanity check for pointer arithmetic below */ assert(sizeof(arrays[0]->Ptr[0]) == 1); @@ -400,15 +389,23 @@ st_feedback_draw_vbo(GLcontext *ctx, struct st_context *st = ctx->st; struct pipe_context *pipe = st->pipe; struct draw_context *draw = st->draw; - GLuint attr, i; - GLbitfield attrsNeeded; - const unsigned attr0_offset = (unsigned) arrays[0]->Ptr; + struct pipe_winsys *winsys = pipe->winsys; + const struct st_vertex_program *vp; + const struct pipe_shader_state *vs; struct pipe_buffer_handle *index_buffer_handle = 0; + struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; + GLuint attr, i; assert(ctx->RenderMode == GL_SELECT || ctx->RenderMode == GL_FEEDBACK); assert(draw); + st_validate_state(ctx->st); + + /* must get these after state validation! */ + vp = ctx->st->vp; + vs = &ctx->st->state.vs->state; + /* * Set up the draw module's state. * @@ -420,82 +417,63 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_clip_state(draw, &st->state.clip); draw_set_rasterizer_state(draw, &st->state.rasterizer->state); draw_bind_vertex_shader(draw, st->state.vs->data); - /* XXX need to set vertex info too */ - - update_default_attribs_buffer(ctx); -#if 0 - /* this must be after state validation */ - attrsNeeded = ctx->st->state.vs->inputs_read; - - /* tell draw module about the vertex array element/attributes */ - for (attr = 0; attr < 16; attr++) { - struct pipe_vertex_buffer vbuffer; + /* loop over TGSI shader inputs to determine vertex buffer + * and attribute info + */ + for (attr = 0; attr < vs->num_inputs; attr++) { + const GLuint mesaAttr = vp->index_to_input[attr]; + struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; struct pipe_vertex_element velement; void *map; - vbuffer.buffer = NULL; - vbuffer.pitch = 0; - velement.src_offset = 0; - velement.vertex_buffer_index = 0; - velement.src_format = 0; - - if (attrsNeeded & (1 << attr)) { - const GLuint mesaAttr = tgsi_attrib_to_mesa_attrib(attr); - struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; - - if (bufobj && bufobj->Name) { - struct st_buffer_object *stobj = st_buffer_object(bufobj); - /* Recall that for VBOs, the gl_client_array->Ptr field is - * really an offset from the start of the VBO, not a pointer. - */ - unsigned offset = (unsigned) arrays[mesaAttr]->Ptr; - - assert(stobj->buffer); - - vbuffer.buffer = stobj->buffer; - vbuffer.buffer_offset = attr0_offset; /* in bytes */ - vbuffer.pitch = arrays[mesaAttr]->StrideB; /* in bytes */ - vbuffer.max_index = 0; /* need this? */ - - velement.src_offset = offset - attr0_offset; /* bytes */ - velement.vertex_buffer_index = attr; - velement.dst_offset = 0; /* need this? */ - velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, - arrays[mesaAttr]->Size); - assert(velement.src_format); - } - else { - /* use the default attribute buffer */ - vbuffer.buffer = ctx->st->default_attrib_buffer; - vbuffer.buffer_offset = 0; - vbuffer.pitch = 0; /* must be zero! */ - vbuffer.max_index = 1; - - velement.src_offset = attr * 4 * sizeof(GLfloat); - velement.vertex_buffer_index = attr; - velement.dst_offset = 0; - velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - } + if (bufobj && bufobj->Name) { + /* Attribute data is in a VBO. + * Recall that for VBOs, the gl_client_array->Ptr field is + * really an offset from the start of the VBO, not a pointer. + */ + struct st_buffer_object *stobj = st_buffer_object(bufobj); + assert(stobj->buffer); + + vbuffer[attr].buffer = NULL; + winsys->buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); + vbuffer[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ + velement.src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; + } + else { + /* attribute data is in user-space memory, not a VBO */ + uint bytes = (arrays[mesaAttr]->Size + * _mesa_sizeof_type(arrays[mesaAttr]->Type) + * (max_index + 1)); + + /* wrap user data */ + vbuffer[attr].buffer + = winsys->user_buffer_create(winsys, + (void *) arrays[mesaAttr]->Ptr, + bytes); + vbuffer[attr].buffer_offset = 0; + velement.src_offset = 0; } - if (attr == 0) - assert(vbuffer.buffer); + /* common-case setup */ + vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ + vbuffer[attr].max_index = 0; /* need this? */ + velement.vertex_buffer_index = attr; + velement.dst_offset = 0; /* need this? */ + velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, + arrays[mesaAttr]->Size); + assert(velement.src_format); - draw_set_vertex_buffer(draw, attr, &vbuffer); + /* tell draw about this attribute */ + draw_set_vertex_buffer(draw, attr, &vbuffer[attr]); draw_set_vertex_element(draw, attr, &velement); /* map the attrib buffer */ - if (vbuffer.buffer) { - map = pipe->winsys->buffer_map(pipe->winsys, - vbuffer.buffer, - PIPE_BUFFER_FLAG_READ); - draw_set_mapped_vertex_buffer(draw, attr, map); - } + map = pipe->winsys->buffer_map(pipe->winsys, + vbuffer[attr].buffer, + PIPE_BUFFER_FLAG_READ); + draw_set_mapped_vertex_buffer(draw, attr, map); } -#else - assert(0); -#endif if (ib) { unsigned indexSize; -- cgit v1.2.3 From 3ab326e1d7ef0ae9bd03dbce33e791449b865ba1 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 11 Oct 2007 10:01:17 -0600 Subject: remove unused tgsi_attrib_to_mesa_attrib() --- src/mesa/state_tracker/st_draw.c | 64 ---------------------------------------- 1 file changed, 64 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index aa7fedf9b8..26cd76f939 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -85,70 +85,6 @@ pipe_vertex_format(GLenum format, GLuint size) } -#if 0 -/** - * Convert a mesa vertex attribute to a TGSI attribute - */ -static GLuint -tgsi_attrib_to_mesa_attrib(GLuint attr) -{ - switch (attr) { - case TGSI_ATTRIB_POS: - return VERT_ATTRIB_POS; - case TGSI_ATTRIB_WEIGHT: - return VERT_ATTRIB_WEIGHT; - case TGSI_ATTRIB_NORMAL: - return VERT_ATTRIB_NORMAL; - case TGSI_ATTRIB_COLOR0: - return VERT_ATTRIB_COLOR0; - case TGSI_ATTRIB_COLOR1: - return VERT_ATTRIB_COLOR1; - case TGSI_ATTRIB_FOG: - return VERT_ATTRIB_FOG; - case TGSI_ATTRIB_COLOR_INDEX: - return VERT_ATTRIB_COLOR_INDEX; - case TGSI_ATTRIB_EDGEFLAG: - return VERT_ATTRIB_EDGEFLAG; - case TGSI_ATTRIB_TEX0: - return VERT_ATTRIB_TEX0; - case TGSI_ATTRIB_TEX1: - return VERT_ATTRIB_TEX1; - case TGSI_ATTRIB_TEX2: - return VERT_ATTRIB_TEX2; - case TGSI_ATTRIB_TEX3: - return VERT_ATTRIB_TEX3; - case TGSI_ATTRIB_TEX4: - return VERT_ATTRIB_TEX4; - case TGSI_ATTRIB_TEX5: - return VERT_ATTRIB_TEX5; - case TGSI_ATTRIB_TEX6: - return VERT_ATTRIB_TEX6; - case TGSI_ATTRIB_TEX7: - return VERT_ATTRIB_TEX7; - case TGSI_ATTRIB_VAR0: - return VERT_ATTRIB_GENERIC0; - case TGSI_ATTRIB_VAR1: - return VERT_ATTRIB_GENERIC1; - case TGSI_ATTRIB_VAR2: - return VERT_ATTRIB_GENERIC2; - case TGSI_ATTRIB_VAR3: - return VERT_ATTRIB_GENERIC3; - case TGSI_ATTRIB_VAR4: - return VERT_ATTRIB_GENERIC4; - case TGSI_ATTRIB_VAR5: - return VERT_ATTRIB_GENERIC5; - case TGSI_ATTRIB_VAR6: - return VERT_ATTRIB_GENERIC6; - case TGSI_ATTRIB_VAR7: - return VERT_ATTRIB_GENERIC7; - default: - assert(0); - return 0; - } -} -#endif - - /** * The default attribute buffer is basically a copy of the * ctx->Current.Attrib[] array. It's used when the vertex program -- cgit v1.2.3 From afd6bd3cb0036089f7fe01eb140b5d9ee2654ce9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 11 Oct 2007 10:03:59 -0600 Subject: rename some vars, updated comments --- src/mesa/state_tracker/st_draw.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 26cd76f939..f62f2b3a3a 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -53,8 +53,11 @@ #include "pipe/draw/draw_context.h" +/** + * Return a PIPE_FORMAT_x for the given GL datatype and size. + */ static GLuint -pipe_vertex_format(GLenum format, GLuint size) +pipe_vertex_format(GLenum type, GLuint size) { static const GLuint float_fmts[4] = { PIPE_FORMAT_R32_FLOAT, @@ -69,12 +72,12 @@ pipe_vertex_format(GLenum format, GLuint size) PIPE_FORMAT_R32G32B32A32_SSCALED, }; - assert(format >= GL_BYTE); - assert(format <= GL_DOUBLE); + assert(type >= GL_BYTE); + assert(type <= GL_DOUBLE); assert(size >= 1); assert(size <= 4); - switch (format) { + switch (type) { case GL_FLOAT: return float_fmts[size - 1]; case GL_INT: @@ -250,7 +253,6 @@ st_draw_vbo(GLcontext *ctx, * \param numVertex number of vertices * \param verts vertex data (all attributes are float[4]) * \param numAttribs number of attributes per vertex - * \param attribs index of each attribute (0=pos, 3=color, etc) */ void st_draw_vertices(GLcontext *ctx, unsigned prim, @@ -460,14 +462,12 @@ st_feedback_draw_vbo(GLcontext *ctx, -/* This is all a hack to keep using tnl until we have vertex programs - * up and running. - */ void st_init_draw( struct st_context *st ) { GLcontext *ctx = st->ctx; struct vbo_context *vbo = (struct vbo_context *) ctx->swtnl_im; + /* actually, not used here, but elsewhere */ create_default_attribs_buffer(st); assert(vbo); -- cgit v1.2.3 From 3fb88639af09af2f77203fd633c19ea736a7c0e5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 11 Oct 2007 11:09:36 -0600 Subject: Init draw->prim = ~0 We weren't rendering correctly if the first thing drawn was a point (PRIM_MODE_POINT=0). --- src/mesa/pipe/draw/draw_context.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 44e770f364..e252148903 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -81,6 +81,8 @@ struct draw_context *draw_create( void ) draw->attrib_front1 = -1; draw->attrib_back1 = -1; + draw->prim = ~0; /* != any of PIPE_PRIM_x */ + draw_vertex_cache_invalidate( draw ); return draw; -- cgit v1.2.3 From e552b9bd099b9fce1cd58ce7922a0c9f74cad034 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 12 Oct 2007 09:42:06 -0600 Subject: initial use of KIL for glBitmap rendering --- src/mesa/state_tracker/st_cb_drawpixels.c | 51 ++++++++++++++++++++++++++----- 1 file changed, 44 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index f58b5d947d..f3d624f14b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -66,8 +66,14 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) if (!p) return NULL; +#define CULL 1 /* Use KIL to cull 0 bits/pixels in bitmap? */ + if (bitmapMode) +#if CULL + p->NumInstructions = 7; +#else p->NumInstructions = 3; +#endif else p->NumInstructions = 2; @@ -95,6 +101,36 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) p->Instructions[ic].TexSrcUnit = 0; p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; ic++; +#if CULL + /* IF temp0 */ + p->Instructions[ic].Opcode = OPCODE_IF; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; + p->Instructions[ic].BranchTarget = ic + 2; + ic++; + + /* MOV result.color, fragment.color */ + p->Instructions[ic].Opcode = OPCODE_MOV; + p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0; + ic++; + + /* ELSE */ + p->Instructions[ic].Opcode = OPCODE_ELSE; + p->Instructions[ic].BranchTarget = ic + 2; + ic++; + + /* KILL */ + p->Instructions[ic].Opcode = OPCODE_KIL_NV; + ic++; + + /* ENDIF */ + p->Instructions[ic].Opcode = OPCODE_ENDIF; + ic++; +#else /* MUL result.color, temp0.xxxx, fragment.color */ p->Instructions[ic].Opcode = OPCODE_MUL; p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; @@ -105,17 +141,18 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) p->Instructions[ic].SrcReg[1].File = PROGRAM_INPUT; p->Instructions[ic].SrcReg[1].Index = FRAG_ATTRIB_COL0; ic++; +#endif } else { /* DrawPixels mode */ /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ - p->Instructions[0].Opcode = OPCODE_TEX; - p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[0].DstReg.Index = FRAG_RESULT_COLR; - p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[0].SrcReg[0].Index = FRAG_ATTRIB_TEX0; - p->Instructions[0].TexSrcUnit = 0; - p->Instructions[0].TexSrcTarget = TEXTURE_2D_INDEX; + p->Instructions[ic].Opcode = OPCODE_TEX; + p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + p->Instructions[ic].TexSrcUnit = 0; + p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; ic++; } /* END; */ -- cgit v1.2.3 From 4becfdfdd871bba5304d87159892a0ff9df222b8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 12 Oct 2007 16:11:58 -0600 Subject: change exec_kilp() to only touch lowest four bits in kilmask --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index fda5584cf3..4a608c445e 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1137,9 +1137,9 @@ static void exec_kilp(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst) { - GLuint uniquemask; + GLbitfield uniquemask; GLuint chan_index; - GLuint kilmask = 0; + GLbitfield kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ union tgsi_exec_channel r[1]; /* This mask stores component bits that were already tested. Note that @@ -1165,7 +1165,7 @@ exec_kilp(struct tgsi_exec_machine *mach, FETCH(&r[0], 0, chan_index); for (i = 0; i < 4; i++) if (r[0].f[i] < 0.0f) - kilmask |= 1 << (i * 4); + kilmask |= 1 << i; } mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; -- cgit v1.2.3 From 2a699038dc2412311f89f9fd843522e784ad8fa3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 12 Oct 2007 16:12:22 -0600 Subject: added code to print extended swizzles --- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index 463f841dde..76241748c7 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -589,6 +589,16 @@ static const char *TGSI_EXTSWIZZLES[] = "EXTSWIZZLE_ONE" }; +static const char *TGSI_EXTSWIZZLES_SHORT[] = +{ + "x", + "y", + "z", + "w", + "0", + "1" +}; + static const char *TGSI_WRITEMASKS[] = { "0", @@ -900,6 +910,16 @@ dump_instruction_short( SID( src->SrcRegister.Index ); CHR( ']' ); + if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { + CHR( '.' ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); + } if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || -- cgit v1.2.3 From 47fc06753ec4319a3ac6470060a4582975daceda Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 12 Oct 2007 16:13:15 -0600 Subject: added assertion to be sure we don't exceed bitfield size --- src/mesa/state_tracker/st_draw.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index f62f2b3a3a..cc204640eb 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -160,6 +160,7 @@ st_draw_vbo(GLcontext *ctx, winsys->buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); vbuffer[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ velement.src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; + assert(velement.src_offset <= 2048); /* 11-bit field */ } else { /* attribute data is in user-space memory, not a VBO */ -- cgit v1.2.3 From 89414d951d329cd70f947a3b85dee714c518dcc8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 12 Oct 2007 16:16:20 -0600 Subject: Use KIL in glBitmap shader to cull the fragments for 0 bits. --- src/mesa/state_tracker/st_cb_drawpixels.c | 71 +++++++++++++------------------ 1 file changed, 30 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index f3d624f14b..f4968b5634 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -66,14 +66,8 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) if (!p) return NULL; -#define CULL 1 /* Use KIL to cull 0 bits/pixels in bitmap? */ - if (bitmapMode) -#if CULL - p->NumInstructions = 7; -#else - p->NumInstructions = 3; -#endif + p->NumInstructions = 6; else p->NumInstructions = 2; @@ -85,14 +79,11 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) _mesa_init_instructions(p->Instructions, p->NumInstructions); if (bitmapMode) { /* - * XXX This is temporary - * We actually need to cull the fragment if the texture value is zero. - * But TGSI doesn't support conditionals yet. - * Also, we need to compose this fragment shader with the current + * XXX, we need to compose this fragment shader with the current * user-provided fragment shader so the fragment program is applied * to the fragments which aren't culled. */ - /* TEX temp0, fragment.texcoord[0], texture[0], 2D; */ + /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */ p->Instructions[ic].Opcode = OPCODE_TEX; p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; p->Instructions[ic].DstReg.Index = 0; @@ -101,13 +92,36 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) p->Instructions[ic].TexSrcUnit = 0; p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; ic++; -#if CULL - /* IF temp0 */ - p->Instructions[ic].Opcode = OPCODE_IF; + + /* SWZ tmp0.x, tmp0.x, 1111; # tmp0.x = 1.0 */ + p->Instructions[ic].Opcode = OPCODE_SWZ; + p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; + p->Instructions[ic].DstReg.Index = 0; + p->Instructions[ic].DstReg.WriteMask = WRITEMASK_X; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + p->Instructions[ic].SrcReg[0].Swizzle + = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE ); + ic++; + + /* SUB tmp0.w, tmp0.w, tmp0.x; # tmp0.w -= 1 */ + p->Instructions[ic].Opcode = OPCODE_SUB; + p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; + p->Instructions[ic].DstReg.Index = 0; + p->Instructions[ic].DstReg.WriteMask = WRITEMASK_W; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; + p->Instructions[ic].SrcReg[1].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[1].Index = 0; + p->Instructions[ic].SrcReg[1].Swizzle = SWIZZLE_XXXX; /* 1.0 */ + ic++; + + /* KIL if tmp0.w < 0 */ + p->Instructions[ic].Opcode = OPCODE_KIL; p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; p->Instructions[ic].SrcReg[0].Index = 0; p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; - p->Instructions[ic].BranchTarget = ic + 2; ic++; /* MOV result.color, fragment.color */ @@ -117,31 +131,6 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0; ic++; - - /* ELSE */ - p->Instructions[ic].Opcode = OPCODE_ELSE; - p->Instructions[ic].BranchTarget = ic + 2; - ic++; - - /* KILL */ - p->Instructions[ic].Opcode = OPCODE_KIL_NV; - ic++; - - /* ENDIF */ - p->Instructions[ic].Opcode = OPCODE_ENDIF; - ic++; -#else - /* MUL result.color, temp0.xxxx, fragment.color */ - p->Instructions[ic].Opcode = OPCODE_MUL; - p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; - p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[0].Index = 0; - p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; - p->Instructions[ic].SrcReg[1].File = PROGRAM_INPUT; - p->Instructions[ic].SrcReg[1].Index = FRAG_ATTRIB_COL0; - ic++; -#endif } else { /* DrawPixels mode */ -- cgit v1.2.3 From a1e332950bf5f6460af81ad29f5917c84649f249 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 13 Oct 2007 09:04:35 -0600 Subject: added z16/z32_git_tile(), change s8z24_get_tile() to return Z as float, ignore stencil --- src/mesa/pipe/softpipe/sp_surface.c | 83 +++++++++++++++++++++++++++++++++++-- 1 file changed, 80 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 29b5c9631c..07def21d42 100755 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -240,6 +240,44 @@ z16_write_quad_z(struct softpipe_surface *sps, dst[1] = zzzz[3]; } +/** + * Return as floats in [0,1]. + */ +static void +z16_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const ushort *src + = ((const ushort *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + const float scale = 1.0 / 65535.0; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_Z16); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp clipping hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j] = src[j] * scale; + } + src += ps->region->pitch; + p += w0; + } +} + + + /*** PIPE_FORMAT_U_L8 ***/ @@ -619,6 +657,42 @@ z32_write_quad_z(struct softpipe_surface *sps, dst[1] = zzzz[3]; } +/** + * Return as floats in [0,1]. + */ +static void +z32_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const uint *src + = ((const uint *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + const double scale = 1.0 / (double) 0xffffffff; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_Z16); + +#if 0 + assert(x + w <= ps->width); + assert(y + h <= ps->height); +#else + /* temp clipping hack */ + if (x + w > ps->width) + w = ps->width - x; + if (y + h > ps->height) + h = ps->height -y; +#endif + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j] = src[j] * scale; + } + src += ps->region->pitch; + p += w0; + } +} + /*** PIPE_FORMAT_S8_Z24 ***/ @@ -697,7 +771,7 @@ s8z24_write_quad_stencil(struct softpipe_surface *sps, /** - * Note, the actual returned pixels are uint, not float + * Return Z component as float in [0,1]. Stencil part ignored. */ static void s8z24_get_tile(struct pipe_surface *ps, @@ -706,6 +780,7 @@ s8z24_get_tile(struct pipe_surface *ps, const uint *src = ((const uint *) (ps->region->map + ps->offset)) + y * ps->region->pitch + x; + const double scale = 1.0 / ((1 << 24) - 1); unsigned i, j; unsigned w0 = w; @@ -722,9 +797,9 @@ s8z24_get_tile(struct pipe_surface *ps, h = ps->height -y; #endif for (i = 0; i < h; i++) { - uint *pRow = (uint *) p; + float *pRow = p; for (j = 0; j < w; j++) { - pRow[j] = src[j]; + pRow[j] = (src[j] & 0xffffff) * scale; } src += ps->region->pitch; p += w0; @@ -811,10 +886,12 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) case PIPE_FORMAT_U_Z16: sps->read_quad_z = z16_read_quad_z; sps->write_quad_z = z16_write_quad_z; + sps->surface.get_tile = z16_get_tile; break; case PIPE_FORMAT_U_Z32: sps->read_quad_z = z32_read_quad_z; sps->write_quad_z = z32_write_quad_z; + sps->surface.get_tile = z32_get_tile; break; case PIPE_FORMAT_S8_Z24: sps->read_quad_z = s8z24_read_quad_z; -- cgit v1.2.3 From eb389aaf720a08045bc2492dd6cf50f1a2e44e87 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 13 Oct 2007 09:05:20 -0600 Subject: get_tile() of Z surfaces returns floats now --- src/mesa/state_tracker/st_cb_readpixels.c | 14 +------------- 1 file changed, 1 insertion(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index a82c4e526a..35cf64bae2 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -136,20 +136,8 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLvoid *dst = _mesa_image_address2d(&clippedPacking, dest, width, height, format, type, i, 0); if (format == GL_DEPTH_COMPONENT) { - float z[MAX_WIDTH]; - if (strb->surface->format == PIPE_FORMAT_S8_Z24) { - const double scale = 1.0 / ((1 << 24) - 1); - const uint *zs = (const uint *) temp; - uint k; - for (k = 0; k < width; k++) { - z[k] = (zs[k] & 0xffffff) * scale; - } - } - else { - assert(0); - } _mesa_pack_depth_span(ctx, width, dst, type, - z, &clippedPacking); + (GLfloat *) temp, &clippedPacking); } else { _mesa_pack_rgba_span_float(ctx, width, temp, format, type, dst, -- cgit v1.2.3 From d75acc8ffa47dbd85d864b2f6ab028652e45044d Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 13 Oct 2007 11:11:11 -0600 Subject: Check texture format in get_texel() to handle depth textures. --- src/mesa/pipe/softpipe/sp_tex_sample.c | 21 +++++++++++++++++---- 1 file changed, 17 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index b89adf8480..aab26f0b0e 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -606,10 +606,23 @@ get_texel(struct tgsi_sampler *sampler, /* get the texel from cache entry */ tx = x % TEX_CACHE_TILE_SIZE; ty = y % TEX_CACHE_TILE_SIZE; - rgba[0][j] = sampler->cache[entry].data[ty][tx][0]; - rgba[1][j] = sampler->cache[entry].data[ty][tx][1]; - rgba[2][j] = sampler->cache[entry].data[ty][tx][2]; - rgba[3][j] = sampler->cache[entry].data[ty][tx][3]; + if (sampler->texture->format == PIPE_FORMAT_U_Z16 || + sampler->texture->format == PIPE_FORMAT_U_Z32 || + sampler->texture->format == PIPE_FORMAT_S8_Z24) { + /* get_tile() returned one float per texel */ + float *src = (float *) sampler->cache[entry].data; + rgba[0][j] = + rgba[1][j] = + rgba[2][j] = + rgba[3][j] = src[ty * TEX_CACHE_TILE_SIZE + tx]; + } + else { + /* get_tile() returned four floats per texel */ + rgba[0][j] = sampler->cache[entry].data[ty][tx][0]; + rgba[1][j] = sampler->cache[entry].data[ty][tx][1]; + rgba[2][j] = sampler->cache[entry].data[ty][tx][2]; + rgba[3][j] = sampler->cache[entry].data[ty][tx][3]; + } } -- cgit v1.2.3 From a2c8b0e861e7c84d927b636663ac39f8cdc504a3 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 13 Oct 2007 11:11:59 -0600 Subject: glDrawPixels support for GL_DEPTH_COMPONENT --- src/mesa/state_tracker/st_cb_drawpixels.c | 125 +++++++++++++++++++++++++++--- 1 file changed, 115 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index f4968b5634..15abc8550b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -163,6 +163,67 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) } +/** + * Create fragment shader that does a TEX() instruction to get a Z + * value, then writes to FRAG_RESULT_DEPR. + */ +static struct st_fragment_program * +make_fragment_shader_z(struct st_context *st) +{ + GLcontext *ctx = st->ctx; + struct st_fragment_program *stfp; + struct gl_program *p; + GLuint ic = 0; + + p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + if (!p) + return NULL; + + p->NumInstructions = 3; + + p->Instructions = _mesa_alloc_instructions(p->NumInstructions); + if (!p->Instructions) { + ctx->Driver.DeleteProgram(ctx, p); + return NULL; + } + _mesa_init_instructions(p->Instructions, p->NumInstructions); + + /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ + p->Instructions[ic].Opcode = OPCODE_TEX; + p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[ic].DstReg.Index = FRAG_RESULT_DEPR; + p->Instructions[ic].DstReg.WriteMask = WRITEMASK_Z; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + p->Instructions[ic].TexSrcUnit = 0; + p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; + + /* MOV result.color, fragment.color */ + p->Instructions[ic].Opcode = OPCODE_MOV; + p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0; + ic++; + + /* END; */ + p->Instructions[ic++].Opcode = OPCODE_END; + + assert(ic == p->NumInstructions); + + p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0; + p->OutputsWritten = (1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR); + + stfp = (struct st_fragment_program *) p; + st_translate_fragment_program(st, stfp, NULL, + stfp->tokens, ST_MAX_SHADER_TOKENS); + + return stfp; +} + + + /** * Create a simple vertex shader that just passes through the * vertex position and texcoord (and color). @@ -235,6 +296,21 @@ make_vertex_shader(struct st_context *st, GLboolean passColor) } +static GLenum +_mesa_base_format(GLenum format) +{ + switch (format) { + case GL_DEPTH_COMPONENT: + return GL_DEPTH_COMPONENT; + case GL_STENCIL_INDEX: + return GL_STENCIL_INDEX; + default: + return GL_RGBA; + } +} + + + /** * Make mipmap tree containing the glDrawPixels image. */ @@ -249,8 +325,11 @@ make_mipmap_tree(struct st_context *st, const GLbitfield flags = PIPE_SURFACE_FLAG_TEXTURE; struct pipe_mipmap_tree *mt; GLuint pipeFormat, cpp; + GLenum baseFormat; - mformat = st_ChooseTextureFormat(st->ctx, GL_RGBA, format, type); + baseFormat = _mesa_base_format(format); + + mformat = st_ChooseTextureFormat(st->ctx, baseFormat, format, type); assert(mformat); pipeFormat = st_mesa_format_to_pipe_format(mformat->MesaFormat); @@ -285,7 +364,7 @@ make_mipmap_tree(struct st_context *st, * the texture. We deal with that with texcoords. */ success = mformat->StoreImage(st->ctx, 2, /* dims */ - GL_RGBA, /* baseInternalFormat */ + baseFormat, /* baseInternalFormat */ mformat, /* gl_texture_format */ dest, /* dest */ 0, 0, 0, /* dstX/Y/Zoffset */ @@ -649,31 +728,51 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { - static struct st_fragment_program *stfp = NULL; - static struct st_vertex_program *stvp = NULL; + static struct st_fragment_program *stfp_c = NULL; /* color */ + static struct st_fragment_program *stfp_z = NULL; /* z */ + static struct st_vertex_program *stvp_t = NULL; /* just emit texcoord */ + static struct st_vertex_program *stvp_c = NULL; /* emit color too */ + struct st_fragment_program *stfp; + struct st_vertex_program *stvp; struct st_context *st = ctx->st; struct pipe_surface *ps; GLuint bufferFormat; + const GLfloat *color; /* create the fragment program if needed */ - if (!stfp) { - stfp = make_fragment_shader(ctx->st, GL_FALSE); + if (!stfp_c) { + stfp_c = make_fragment_shader(ctx->st, GL_FALSE); } + if (!stfp_z) { + stfp_z = make_fragment_shader_z(ctx->st); + } + /* and vertex program */ - if (!stvp) { - stvp = make_vertex_shader(ctx->st, GL_FALSE); + if (!stvp_t) { + stvp_t = make_vertex_shader(ctx->st, GL_FALSE); + } + if (!stvp_c) { + stvp_c = make_vertex_shader(ctx->st, GL_TRUE); } st_validate_state(st); if (format == GL_DEPTH_COMPONENT) { ps = st->state.framebuffer.zbuf; + stfp = stfp_z; + stvp = stvp_c; + color = ctx->Current.RasterColor; } else if (format == GL_STENCIL_INDEX) { ps = st->state.framebuffer.sbuf; + /* XXX special case - can't use texture map */ + color = NULL; } else { ps = st->state.framebuffer.cbufs[0]; + stfp = stfp_c; + stvp = stvp_t; + color = NULL; } bufferFormat = ps->format; @@ -688,7 +787,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (mt) { draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, - mt, stvp, stfp, NULL); + mt, stvp, stfp, color); free_mipmap_tree(st->pipe, mt); } } @@ -890,8 +989,14 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, st_validate_state(st); + /* allocate a texture of size width x height */ + + /* blit/copy framebuffer region into texture */ + + /* draw textured quad */ + + fprintf(stderr, "st_CopyPixels not implemented yet\n"); - /* XXX to do */ } -- cgit v1.2.3 From 1e2770e2e6244a0decbf077dfd34a3763428a2c0 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 13 Oct 2007 12:27:10 -0600 Subject: inequality tests were backward --- src/mesa/pipe/softpipe/sp_quad_stencil.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 56cc6907b2..bf72bb23cd 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -42,42 +42,42 @@ do_stencil_test(const ubyte stencilVals[QUAD_SIZE], unsigned func, break; case PIPE_FUNC_LESS: for (j = 0; j < QUAD_SIZE; j++) { - if ((stencilVals[j] & valMask) < ref) { + if (ref < (stencilVals[j] & valMask)) { passMask |= (1 << j); } } break; case PIPE_FUNC_EQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if ((stencilVals[j] & valMask) == ref) { + if (ref == (stencilVals[j] & valMask)) { passMask |= (1 << j); } } break; case PIPE_FUNC_LEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if ((stencilVals[j] & valMask) <= ref) { + if (ref <= (stencilVals[j] & valMask)) { passMask |= (1 << j); } } break; case PIPE_FUNC_GREATER: for (j = 0; j < QUAD_SIZE; j++) { - if ((stencilVals[j] & valMask) > ref) { + if (ref > (stencilVals[j] & valMask)) { passMask |= (1 << j); } } break; case PIPE_FUNC_NOTEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if ((stencilVals[j] & valMask) != ref) { + if (ref != (stencilVals[j] & valMask)) { passMask |= (1 << j); } } break; case PIPE_FUNC_GEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if ((stencilVals[j] & valMask) >= ref) { + if (ref >= (stencilVals[j] & valMask)) { passMask |= (1 << j); } } -- cgit v1.2.3 From 12f41c8884616b8b0c0e445a8bd3f0587a970d1a Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 13 Oct 2007 12:28:34 -0600 Subject: format info for Z16/Z32 --- src/mesa/state_tracker/st_format.c | 16 +++++++++++++++- 1 file changed, 15 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 8dd7df4495..2d20892cb0 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -79,7 +79,20 @@ st_get_format_info(GLuint format) 0, 0, /* depth, stencil */ 2 /* size in bytes */ }, - /* XXX lots more */ + { + PIPE_FORMAT_U_Z16, + GL_DEPTH_COMPONENT, /* base_format */ + 0, 0, 0, 0, 0, 0, /* color bits */ + 16, 0, /* depth, stencil */ + 2 /* size in bytes */ + }, + { + PIPE_FORMAT_U_Z32, + GL_DEPTH_COMPONENT, /* base_format */ + 0, 0, 0, 0, 0, 0, /* color bits */ + 32, 0, /* depth, stencil */ + 4 /* size in bytes */ + }, { PIPE_FORMAT_S8_Z24, GL_DEPTH_STENCIL_EXT, /* base_format */ @@ -87,6 +100,7 @@ st_get_format_info(GLuint format) 24, 8, /* depth, stencil */ 4 /* size in bytes */ } + /* XXX lots more cases to add */ }; GLuint i; -- cgit v1.2.3 From fa1d442879f1279cf510a52a1002367b904d35ad Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 13 Oct 2007 12:28:53 -0600 Subject: GL_STENCIL_INDEX support --- src/mesa/state_tracker/st_cb_drawpixels.c | 85 ++++++++++++++++++++++++++++++- 1 file changed, 83 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 15abc8550b..92705cdc44 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -32,6 +32,7 @@ #include "main/imports.h" #include "main/image.h" +#include "main/macros.h" #include "st_context.h" #include "st_atom.h" @@ -720,6 +721,81 @@ draw_blit(struct st_context *st, } +static void +draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, + GLsizei width, GLsizei height, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels) +{ + struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; + struct pipe_surface *ps = st->state.framebuffer.sbuf; + const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0; + GLint skipPixels; + ubyte *stmap; + + /* map the stencil buffer */ + stmap = pipe->region_map(pipe, ps->region); + + /* if width > MAX_WIDTH, have to process image in chunks */ + skipPixels = 0; + while (skipPixels < width) { + const GLint spanX = x + skipPixels; + const GLint spanWidth = MIN2(width - skipPixels, MAX_WIDTH); + GLint row; + for (row = 0; row < height; row++) { + GLint spanY = y + row; + GLubyte values[MAX_WIDTH]; + GLenum destType = GL_UNSIGNED_BYTE; + const GLvoid *source = _mesa_image_address2d(unpack, pixels, + width, height, + GL_COLOR_INDEX, type, + row, skipPixels); + _mesa_unpack_stencil_span(ctx, spanWidth, destType, values, + type, source, unpack, + ctx->_ImageTransferState); + if (zoom) { + /* + _swrast_write_zoomed_stencil_span(ctx, x, y, spanWidth, + spanX, spanY, values); + */ + } + else { + if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { + spanY = ctx->DrawBuffer->Height - spanY - 1; + } + + switch (ps->format) { + case PIPE_FORMAT_U_S8: + { + ubyte *dest = stmap + spanY * ps->region->pitch + spanX; + memcpy(dest, values, spanWidth); + } + break; + case PIPE_FORMAT_S8_Z24: + { + uint *dest = (uint *) stmap + spanY * ps->region->pitch + spanX; + GLint k; + for (k = 0; k < spanWidth; k++) { + uint p = dest[k]; + p = (p & 0xffffff) | (values[k] << 24); + dest[k] = p; + } + } + break; + default: + assert(0); + } + } + } + skipPixels += spanWidth; + } + + /* unmap the stencil buffer */ + pipe->region_unmap(pipe, ps->region); +} + + /** * Called via ctx->Driver.DrawPixels() */ @@ -739,7 +815,12 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLuint bufferFormat; const GLfloat *color; - /* create the fragment program if needed */ + if (format == GL_STENCIL_INDEX) { + draw_stencil_pixels(ctx, x, y, width, height, type, unpack, pixels); + return; + } + + /* create the fragment programs if needed */ if (!stfp_c) { stfp_c = make_fragment_shader(ctx->st, GL_FALSE); } @@ -747,7 +828,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, stfp_z = make_fragment_shader_z(ctx->st); } - /* and vertex program */ + /* and vertex programs */ if (!stvp_t) { stvp_t = make_vertex_shader(ctx->st, GL_FALSE); } -- cgit v1.2.3 From 69466c60a5286328d140a5100580c90e9bf62327 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 13 Oct 2007 12:29:11 -0600 Subject: GL_STENCIL_INDEX support --- src/mesa/state_tracker/st_cb_readpixels.c | 78 ++++++++++++++++++++++++++++++- 1 file changed, 77 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 35cf64bae2..3fcce2463c 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -46,6 +46,78 @@ #include "st_public.h" +/** + * Special case for reading stencil buffer. + * For color/depth we use get_tile(). For stencil, map the stencil buffer. + */ +static void +read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, + GLsizei width, GLsizei height, GLenum type, + const struct gl_pixelstore_attrib *packing, + GLvoid *pixels) +{ + struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; + struct gl_framebuffer *fb = ctx->ReadBuffer; + struct st_renderbuffer *strb = st_renderbuffer(fb->_StencilBuffer); + struct pipe_surface *ps = strb->surface; + ubyte *stmap; + GLint j; + + /* map the stencil buffer */ + stmap = pipe->region_map(pipe, ps->region); + + /* width should never be > MAX_WIDTH since we did clipping earlier */ + ASSERT(width <= MAX_WIDTH); + + /* process image row by row */ + for (j = 0; j < height; j++, y++) { + GLvoid *dest; + GLstencil values[MAX_WIDTH]; + GLint srcY; + + if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { + srcY = ctx->DrawBuffer->Height - y - 1; + } + else { + srcY = y; + } + + /* get stencil values */ + switch (ps->format) { + case PIPE_FORMAT_U_S8: + { + const ubyte *src = stmap + y * ps->region->pitch + x; + memcpy(values, src, width); + } + break; + case PIPE_FORMAT_S8_Z24: + { + const uint *src = (const uint *) stmap + y * ps->region->pitch + x; + GLint k; + for (k = 0; k < width; k++) { + values[k] = src[k] >> 24; + } + } + break; + default: + assert(0); + } + + /* store */ + dest = _mesa_image_address2d(packing, pixels, width, height, + GL_STENCIL_INDEX, type, j, 0); + + _mesa_pack_stencil_span(ctx, width, type, dest, values, packing); + } + + + /* unmap the stencil buffer */ + pipe->region_unmap(pipe, ps->region); +} + + + /** * Do glReadPixels by getting rows from the framebuffer surface with * get_tile(). Convert to requested format/type with Mesa image routines. @@ -86,7 +158,11 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } - if (format == GL_DEPTH_COMPONENT) { + if (format == GL_STENCIL_INDEX) { + read_stencil_pixels(ctx, x, y, width, height, type, pack, dest); + return; + } + else if (format == GL_DEPTH_COMPONENT) { strb = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer); } else { -- cgit v1.2.3 From 9f797d8cb323f5f61ac8cffcba3329b6f42721aa Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 14 Oct 2007 11:52:00 -0600 Subject: Added accum function/files. --- src/mesa/sources | 1 + src/mesa/state_tracker/st_cb_accum.c | 90 ++++++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_accum.h | 36 +++++++++++++++ src/mesa/state_tracker/st_context.c | 2 + 4 files changed, 129 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_accum.c create mode 100644 src/mesa/state_tracker/st_cb_accum.h (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 069be2e01d..cf307401b2 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -208,6 +208,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_stipple.c \ state_tracker/st_atom_texture.c \ state_tracker/st_atom_viewport.c \ + state_tracker/st_cb_accum.c \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_flush.c \ diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c new file mode 100644 index 0000000000..c8fef6c135 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -0,0 +1,90 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Brian Paul + */ + +#include "main/imports.h" +#include "main/image.h" +#include "main/macros.h" + +#include "st_context.h" +#include "st_atom.h" +#include "st_cache.h" +#include "st_draw.h" +#include "st_program.h" +#include "st_cb_accum.h" +#include "st_draw.h" +#include "st_format.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + + + +static void +st_Accum(GLcontext *ctx, GLenum op, GLfloat value) +{ + const GLint xpos = ctx->DrawBuffer->_Xmin; + const GLint ypos = ctx->DrawBuffer->_Ymin; + const GLint width = ctx->DrawBuffer->_Xmax - xpos; + const GLint height = ctx->DrawBuffer->_Ymax - ypos; + + switch (op) { + case GL_ADD: + if (value != 0.0F) { + accum_add(ctx, value, xpos, ypos, width, height); + } + break; + case GL_MULT: + if (value != 1.0F) { + accum_mult(ctx, value, xpos, ypos, width, height); + } + break; + case GL_ACCUM: + if (value != 0.0F) { + accum_accum(ctx, value, xpos, ypos, width, height); + } + break; + case GL_LOAD: + accum_load(ctx, value, xpos, ypos, width, height); + break; + case GL_RETURN: + accum_return(ctx, value, xpos, ypos, width, height); + break; + default: + assert(0); + } +} + + + +void st_init_accum_functions(struct dd_function_table *functions) +{ + functions->Accum = st_Accum; +} diff --git a/src/mesa/state_tracker/st_cb_accum.h b/src/mesa/state_tracker/st_cb_accum.h new file mode 100644 index 0000000000..8a4101c627 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_accum.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_ACCUM_H +#define ST_CB_ACCUM_H + + +extern void st_init_accum_functions(struct dd_function_table *functions); + + +#endif /* ST_CB_ACCUM_H */ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 7c20b036a4..09d9b1ea3b 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -31,6 +31,7 @@ #include "vbo/vbo.h" #include "st_public.h" #include "st_context.h" +#include "st_cb_accum.h" #include "st_cb_bufferobjects.h" #include "st_cb_clear.h" #include "st_cb_drawpixels.h" @@ -130,6 +131,7 @@ void st_destroy_context( struct st_context *st ) void st_init_driver_functions(struct dd_function_table *functions) { + st_init_accum_functions(functions); st_init_bufferobject_functions(functions); st_init_clear_functions(functions); st_init_drawpixels_functions(functions); -- cgit v1.2.3 From d45fdc3f1f8200c1f1703bdcd5a74a153c74b1c4 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 14 Oct 2007 11:53:15 -0600 Subject: 16-bit RGBA surface format for accum buffers --- src/mesa/pipe/softpipe/sp_region.c | 77 +++++++++----- src/mesa/pipe/softpipe/sp_surface.c | 206 ++++++++++++++++++++---------------- 2 files changed, 168 insertions(+), 115 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index faf2737d6b..982e081f60 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -212,9 +212,7 @@ sp_region_copy(struct pipe_context *pipe, pipe->region_unmap(pipe, dst); } -/* Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ + static ubyte * get_pointer(struct pipe_region *dst, unsigned x, unsigned y) { @@ -222,6 +220,13 @@ get_pointer(struct pipe_region *dst, unsigned x, unsigned y) } +#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8)) + + +/** + * Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ static void sp_region_fill(struct pipe_context *pipe, struct pipe_region *dst, @@ -237,32 +242,54 @@ sp_region_fill(struct pipe_context *pipe, (void)pipe->region_map(pipe, dst); switch (dst->cpp) { - case 1: { - ubyte *row = get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); + case 1: + { + ubyte *row = get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); row += dst->pitch; + } } - } - break; - case 2: { - ushort *row = (ushort *) get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; + break; + case 2: + { + ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } } - } - break; - case 4: { - unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; + break; + case 4: + { + unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } } - } - break; + break; + case 8: + { + /* expand the 4-byte clear value to an 8-byte value */ + ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); + ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); + ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); + ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) { + row[j*4+0] = val0; + row[j*4+1] = val1; + row[j*4+2] = val2; + row[j*4+3] = val3; + } + row += dst->pitch * 4; + } + } + break; default: assert(0); break; diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 07def21d42..02a072fe9b 100755 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -43,6 +43,31 @@ */ +/** Convert GLshort in [-32768,32767] to GLfloat in [-1.0,1.0] */ +#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) + +#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ + us = ( (short) ( CLAMP((f), 0.0, 1.0) * 32767.0F) ) + + + +#if 0 +#define CLIP_TILE \ + do { \ + assert(x + w <= ps->width); \ + assert(y + h <= ps->height); \ + } while(0) + +#else +#define CLIP_TILE \ + do { \ + if (x + w > ps->width) \ + w = ps->width - x; \ + if (y + h > ps->height) \ + h = ps->height -y; \ + } while(0) +#endif + /*** PIPE_FORMAT_U_A8_R8_G8_B8 ***/ @@ -108,16 +133,8 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif + CLIP_TILE; + for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { @@ -147,16 +164,8 @@ a8r8g8b8_put_tile(struct pipe_surface *ps, assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif + CLIP_TILE; + for (i = 0; i < h; i++) { const float *pRow = p; for (j = 0; j < w; j++) { @@ -256,16 +265,8 @@ z16_get_tile(struct pipe_surface *ps, assert(ps->format == PIPE_FORMAT_U_Z16); -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif + CLIP_TILE; + for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { @@ -338,16 +339,8 @@ l8_get_tile(struct pipe_surface *ps, assert(ps->format == PIPE_FORMAT_U_L8); -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif + CLIP_TILE; + for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { @@ -422,16 +415,8 @@ a8_get_tile(struct pipe_surface *ps, assert(ps->format == PIPE_FORMAT_U_A8); -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif + CLIP_TILE; + for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { @@ -447,6 +432,74 @@ a8_get_tile(struct pipe_surface *ps, } +/*** PIPE_FORMAT_S_R16_G16_B16_A16 ***/ + +static void +r16g16b16a16_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const short *src + = ((const short *) (ps->region->map + ps->offset)) + + (y * ps->region->pitch + x) * 4; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_S_R16_G16_B16_A16); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + const short *pixel = src; + for (j = 0; j < w; j++) { + pRow[0] = SHORT_TO_FLOAT(pixel[0]); + pRow[1] = SHORT_TO_FLOAT(pixel[1]); + pRow[2] = SHORT_TO_FLOAT(pixel[2]); + pRow[3] = SHORT_TO_FLOAT(pixel[3]); + pRow += 4; + pixel += 4; + } + src += ps->region->pitch * 4; + p += w0 * 4; + } +} + + +static void +r16g16b16a16_put_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + short *dst + = ((short *) (ps->region->map + ps->offset)) + + (y * ps->region->pitch + x) * 4; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_S_R16_G16_B16_A16); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + short r, g, b, a; + UNCLAMPED_FLOAT_TO_SHORT(r, pRow[0]); + UNCLAMPED_FLOAT_TO_SHORT(g, pRow[1]); + UNCLAMPED_FLOAT_TO_SHORT(b, pRow[2]); + UNCLAMPED_FLOAT_TO_SHORT(a, pRow[3]); + dst[j*4+0] = r; + dst[j*4+1] = g; + dst[j*4+2] = b; + dst[j*4+3] = a; + pRow += 4; + } + dst += ps->region->pitch * 4; + p += w0 * 4; + } +} + + /*** PIPE_FORMAT_U_I8 ***/ @@ -507,16 +560,8 @@ i8_get_tile(struct pipe_surface *ps, assert(ps->format == PIPE_FORMAT_U_I8); -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif + CLIP_TILE; + for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { @@ -593,16 +638,8 @@ a8_l8_get_tile(struct pipe_surface *ps, assert(ps->format == PIPE_FORMAT_U_A8_L8); -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif + CLIP_TILE; + for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { @@ -673,16 +710,8 @@ z32_get_tile(struct pipe_surface *ps, assert(ps->format == PIPE_FORMAT_U_Z16); -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif + CLIP_TILE; + for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { @@ -786,16 +815,8 @@ s8z24_get_tile(struct pipe_surface *ps, assert(ps->format == PIPE_FORMAT_S8_Z24); -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif + CLIP_TILE; + for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { @@ -883,6 +904,11 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) sps->surface.get_tile = a8_l8_get_tile; break; + case PIPE_FORMAT_S_R16_G16_B16_A16: + sps->surface.get_tile = r16g16b16a16_get_tile; + sps->surface.put_tile = r16g16b16a16_put_tile; + break; + case PIPE_FORMAT_U_Z16: sps->read_quad_z = z16_read_quad_z; sps->write_quad_z = z16_write_quad_z; -- cgit v1.2.3 From 7b0b694406e4043cb163f9832c9c02934fa54568 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 14 Oct 2007 11:55:31 -0600 Subject: 16-bit rgba surface/format for accum --- src/mesa/state_tracker/st_cb_fbo.c | 12 ++++------ src/mesa/state_tracker/st_format.c | 49 +++++++++++++++++++++++++++++++++++++- src/mesa/state_tracker/st_format.h | 5 ++++ 3 files changed, 57 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 3e4aeab523..e1cc7d98bd 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -70,14 +70,6 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, if (!info) return GL_FALSE; - switch (pipeFormat) { - case PIPE_FORMAT_S8_Z24: - strb->Base.DataType = GL_UNSIGNED_INT; - break; - default: - strb->Base.DataType = GL_UNSIGNED_BYTE; /* XXX fix */ - } - strb->Base._ActualFormat = info->base_format; strb->Base.RedBits = info->red_bits; strb->Base.GreenBits = info->green_bits; @@ -85,6 +77,9 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, strb->Base.AlphaBits = info->alpha_bits; strb->Base.DepthBits = info->depth_bits; strb->Base.StencilBits = info->stencil_bits; + strb->Base.DataType = st_format_datatype(pipeFormat); + + assert(strb->Base.DataType); cpp = info->size; @@ -229,6 +224,7 @@ st_new_renderbuffer_fb(GLenum intFormat) switch (intFormat) { case GL_RGB5: case GL_RGBA8: + case GL_RGBA16: strb->Base._BaseFormat = GL_RGBA; break; case GL_DEPTH_COMPONENT16: diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 2d20892cb0..7564c6014e 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -54,6 +54,7 @@ st_get_format_info(GLuint format) { PIPE_FORMAT_U_R8_G8_B8_A8, /* format */ GL_RGBA, /* base_format */ + GL_UNSIGNED_BYTE, /* datatype for renderbuffers */ 8, 8, 8, 8, 0, 0, /* color bits */ 0, 0, /* depth, stencil */ 4 /* size in bytes */ @@ -61,6 +62,7 @@ st_get_format_info(GLuint format) { PIPE_FORMAT_U_A8_R8_G8_B8, GL_RGBA, /* base_format */ + GL_UNSIGNED_BYTE, /* datatype for renderbuffers */ 8, 8, 8, 8, 0, 0, /* color bits */ 0, 0, /* depth, stencil */ 4 /* size in bytes */ @@ -68,6 +70,7 @@ st_get_format_info(GLuint format) { PIPE_FORMAT_U_A1_R5_G5_B5, GL_RGBA, /* base_format */ + GL_UNSIGNED_SHORT, /* datatype for renderbuffers */ 5, 5, 5, 1, 0, 0, /* color bits */ 0, 0, /* depth, stencil */ 2 /* size in bytes */ @@ -75,13 +78,23 @@ st_get_format_info(GLuint format) { PIPE_FORMAT_U_R5_G6_B5, GL_RGBA, /* base_format */ + GL_UNSIGNED_SHORT, /* datatype for renderbuffers */ 5, 6, 5, 0, 0, 0, /* color bits */ 0, 0, /* depth, stencil */ 2 /* size in bytes */ }, + { + PIPE_FORMAT_S_R16_G16_B16_A16, + GL_RGBA, /* base_format */ + GL_UNSIGNED_SHORT, /* datatype for renderbuffers */ + 16, 16, 16, 16, 0, 0, /* color bits */ + 0, 0, /* depth, stencil */ + 8 /* size in bytes */ + }, { PIPE_FORMAT_U_Z16, GL_DEPTH_COMPONENT, /* base_format */ + GL_UNSIGNED_SHORT, /* datatype for renderbuffers */ 0, 0, 0, 0, 0, 0, /* color bits */ 16, 0, /* depth, stencil */ 2 /* size in bytes */ @@ -89,6 +102,7 @@ st_get_format_info(GLuint format) { PIPE_FORMAT_U_Z32, GL_DEPTH_COMPONENT, /* base_format */ + GL_UNSIGNED_INT, /* datatype for renderbuffers */ 0, 0, 0, 0, 0, 0, /* color bits */ 32, 0, /* depth, stencil */ 4 /* size in bytes */ @@ -96,6 +110,7 @@ st_get_format_info(GLuint format) { PIPE_FORMAT_S8_Z24, GL_DEPTH_STENCIL_EXT, /* base_format */ + GL_UNSIGNED_INT, /* datatype for renderbuffers */ 0, 0, 0, 0, 0, 0, /* color bits */ 24, 8, /* depth, stencil */ 4 /* size in bytes */ @@ -112,6 +127,9 @@ st_get_format_info(GLuint format) } +/** + * Return bytes per pixel for the given format. + */ GLuint st_sizeof_format(GLuint pipeFormat) { @@ -121,6 +139,18 @@ st_sizeof_format(GLuint pipeFormat) } +/** + * Return bytes per pixel for the given format. + */ +GLenum +st_format_datatype(GLuint pipeFormat) +{ + const struct pipe_format_info *info = st_get_format_info(pipeFormat); + assert(info); + return info->datatype; +} + + GLuint st_mesa_format_to_pipe_format(GLuint mesaFormat) { @@ -164,6 +194,22 @@ default_rgba_format(const GLuint formats[], GLuint num) } +/** + * Search list of formats for first RGBA format with >8 bits/channel. + */ +static GLuint +default_deep_rgba_format(const GLuint formats[], GLuint num) +{ + GLuint i; + for (i = 0; i < num; i++) { + if (formats[i] == PIPE_FORMAT_S_R16_G16_B16_A16) { + return formats[i]; + } + } + return PIPE_FORMAT_NONE; +} + + /** * Search list of formats for first depth/Z format. */ @@ -244,8 +290,9 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: - case GL_RGBA16: return default_rgba_format(supported, n); + case GL_RGBA16: + return default_deep_rgba_format(supported, n); case GL_RGBA4: case GL_RGBA2: diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index 6b3cba0f63..17d0985411 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -34,6 +34,7 @@ struct pipe_format_info { GLuint format; GLenum base_format; + GLenum datatype; GLubyte red_bits; GLubyte green_bits; GLubyte blue_bits; @@ -54,6 +55,10 @@ extern GLuint st_sizeof_format(GLuint pipeFormat); +extern GLenum +st_format_datatype(GLuint pipeFormat); + + extern GLuint st_mesa_format_to_pipe_format(GLuint mesaFormat); -- cgit v1.2.3 From 90f7ae2370630c45acb9287465806e5fcac44033 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 14 Oct 2007 11:55:45 -0600 Subject: accum buffer support --- src/mesa/drivers/x11/xm_api.c | 14 ++++++++++---- src/mesa/drivers/x11/xm_surface.c | 3 ++- 2 files changed, 12 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 254d6863ba..1d3f799f36 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -442,15 +442,21 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_STENCIL, rb); } + if (vis->mesa_visual.accumRedBits > 0) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_RGBA16); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_ACCUM, rb); + } + /* * Other renderbuffer (depth, stencil, etc) */ _mesa_add_soft_renderbuffers(&b->mesa_buffer, - GL_FALSE, /* color */ - GL_FALSE,/*vis->mesa_visual.haveDepthBuffer,*/ - GL_FALSE, - vis->mesa_visual.haveAccumBuffer, + GL_FALSE, /* color */ + GL_FALSE, /*vis->mesa_visual.haveDepthBuffer,*/ + GL_FALSE, /* stencil */ + GL_FALSE, /* accum */ b->swAlpha, vis->mesa_visual.numAuxBuffers > 0 ); diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index d40d9eda84..70fa97b339 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -277,10 +277,11 @@ xmesa_supported_formats(struct pipe_context *pipe, GLuint *numFormats) { static const GLuint formats[] = { PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_S_R16_G16_B16_A16, PIPE_FORMAT_S8_Z24 }; - *numFormats = 2; + *numFormats = sizeof(formats) / sizeof(formats[0]); return formats; } -- cgit v1.2.3 From a7611dce40e532f6c768ed7011725fcfb6424883 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 14 Oct 2007 12:32:39 -0600 Subject: implement put_tile() --- src/mesa/drivers/x11/xm_surface.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 70fa97b339..75ad450360 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -192,7 +192,18 @@ static void put_tile(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, const GLfloat *p) { - assert(0); + struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); + GLubyte tmp[MAX_WIDTH * 4]; + GLuint i, j; + GET_CURRENT_CONTEXT(ctx); + FLIP(y); + for (i = 0; i < h; i++) { + for (j = 0; j < w * 4; j++) { + CLAMPED_FLOAT_TO_UBYTE(tmp[j], p[j]); + } + xrb->St.Base.PutRow(ctx, &xrb->St.Base, w, x, y - i, tmp, NULL); + p += w * 4; + } } -- cgit v1.2.3 From e7af94416240f79ac609ff096690a1929664944e Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 14 Oct 2007 12:34:55 -0600 Subject: implement accum ops --- src/mesa/state_tracker/st_cb_accum.c | 170 +++++++++++++++++++++++++++++++++-- 1 file changed, 162 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index c8fef6c135..03c65db70b 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -35,21 +35,175 @@ #include "main/macros.h" #include "st_context.h" -#include "st_atom.h" #include "st_cache.h" -#include "st_draw.h" -#include "st_program.h" #include "st_cb_accum.h" +#include "st_cb_fbo.h" #include "st_draw.h" #include "st_format.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +/** + * For hardware that supports deep color buffers, we could accelerate + * most/all the accum operations with blending/texturing. + * For now, just use the get/put_tile() functions and do things in software. + */ + + +/** For ADD/MULT */ +static void +accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, + GLint xpos, GLint ypos, GLint width, GLint height, + struct pipe_surface *acc_ps) +{ + GLfloat *accBuf; + GLint i; + + accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + (void) pipe->region_map(pipe, acc_ps->region); + + acc_ps->get_tile(acc_ps, xpos, ypos, width, height, accBuf); + + for (i = 0; i < 4 * width * height; i++) { + GLfloat val = accBuf[i] * scale + bias; + accBuf[i] = CLAMP(val, 0.0, 1.0); + } + + acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); + + _mesa_free(accBuf); + + pipe->region_unmap(pipe, acc_ps->region); +} + + +static void +accum_accum(struct pipe_context *pipe, GLfloat value, + GLint xpos, GLint ypos, GLint width, GLint height, + struct pipe_surface *acc_ps, + struct pipe_surface *color_ps) +{ + ubyte *colorMap, *accMap; + GLfloat *colorBuf, *accBuf; + GLint i; + + colorBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + colorMap = pipe->region_map(pipe, color_ps->region); + accMap = pipe->region_map(pipe, acc_ps->region); + + color_ps->get_tile(color_ps, xpos, ypos, width, height, colorBuf); + acc_ps->get_tile(acc_ps, xpos, ypos, width, height, accBuf); + + for (i = 0; i < 4 * width * height; i++) { + GLfloat val = accBuf[i] + colorBuf[i] * value; + accBuf[i] = CLAMP(val, 0.0, 1.0); + } + + acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); + + _mesa_free(colorBuf); + _mesa_free(accBuf); + + pipe->region_unmap(pipe, color_ps->region); + pipe->region_unmap(pipe, acc_ps->region); +} + + +static void +accum_load(struct pipe_context *pipe, GLfloat value, + GLint xpos, GLint ypos, GLint width, GLint height, + struct pipe_surface *acc_ps, + struct pipe_surface *color_ps) +{ + GLfloat *buf; + GLint i; + + buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + (void) pipe->region_map(pipe, color_ps->region); + (void) pipe->region_map(pipe, acc_ps->region); + + color_ps->get_tile(color_ps, xpos, ypos, width, height, buf); + + for (i = 0; i < 4 * width * height; i++) { + GLfloat val = buf[i] * value; + buf[i] = CLAMP(val, 0.0, 1.0); + } + + acc_ps->put_tile(acc_ps, xpos, ypos, width, height, buf); + + _mesa_free(buf); + + pipe->region_unmap(pipe, color_ps->region); + pipe->region_unmap(pipe, acc_ps->region); +} + + +static void +accum_return(GLcontext *ctx, GLfloat value, + GLint xpos, GLint ypos, GLint width, GLint height, + struct pipe_surface *acc_ps, + struct pipe_surface *color_ps) +{ + struct pipe_context *pipe = ctx->st->pipe; + const GLboolean writeR = ctx->Color.ColorMask[0]; + const GLboolean writeG = ctx->Color.ColorMask[1]; + const GLboolean writeB = ctx->Color.ColorMask[2]; + const GLboolean writeA = ctx->Color.ColorMask[3]; + GLfloat *buf; + GLint i; + + buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + (void) pipe->region_map(pipe, color_ps->region); + (void) pipe->region_map(pipe, acc_ps->region); + + acc_ps->get_tile(acc_ps, xpos, ypos, width, height, buf); + + for (i = 0; i < width * height; i++) { + if (writeR) { + GLfloat r = buf[i * 4 + 0] * value; + buf[i * 4 + 0] = CLAMP(r, 0.0, 1.0); + } + if (writeG) { + GLfloat g = buf[i * 4 + 1] * value; + buf[i * 4 + 1] = CLAMP(g, 0.0, 1.0); + } + if (writeB) { + GLfloat b = buf[i * 4 + 2] * value; + buf[i * 4 + 2] = CLAMP(b, 0.0, 1.0); + } + if (writeA) { + GLfloat a = buf[i * 4 + 3] * value; + buf[i * 4 + 3] = CLAMP(a, 0.0, 1.0); + } + } + + color_ps->put_tile(color_ps, xpos, ypos, width, height, buf); + + _mesa_free(buf); + + pipe->region_unmap(pipe, color_ps->region); + pipe->region_unmap(pipe, acc_ps->region); +} + static void st_Accum(GLcontext *ctx, GLenum op, GLfloat value) { + struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; + struct st_renderbuffer *acc_strb + = st_renderbuffer(ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer); + struct st_renderbuffer *color_strb + = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + struct pipe_surface *acc_ps = acc_strb->surface; + struct pipe_surface *color_ps = color_strb->surface; + const GLint xpos = ctx->DrawBuffer->_Xmin; const GLint ypos = ctx->DrawBuffer->_Ymin; const GLint width = ctx->DrawBuffer->_Xmax - xpos; @@ -58,24 +212,24 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value) switch (op) { case GL_ADD: if (value != 0.0F) { - accum_add(ctx, value, xpos, ypos, width, height); + accum_mad(pipe, 1.0, value, xpos, ypos, width, height, acc_ps); } break; case GL_MULT: if (value != 1.0F) { - accum_mult(ctx, value, xpos, ypos, width, height); + accum_mad(pipe, value, 0.0, xpos, ypos, width, height, acc_ps); } break; case GL_ACCUM: if (value != 0.0F) { - accum_accum(ctx, value, xpos, ypos, width, height); + accum_accum(pipe, value, xpos, ypos, width, height, acc_ps, color_ps); } break; case GL_LOAD: - accum_load(ctx, value, xpos, ypos, width, height); + accum_load(pipe, value, xpos, ypos, width, height, acc_ps, color_ps); break; case GL_RETURN: - accum_return(ctx, value, xpos, ypos, width, height); + accum_return(ctx, value, xpos, ypos, width, height, acc_ps, color_ps); break; default: assert(0); -- cgit v1.2.3 From 29db69e222b04b5a616942b06bd118c9ac75ec41 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 10:25:42 -0600 Subject: added print_texkil_op() --- src/mesa/pipe/i915simple/i915_debug_fp.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_debug_fp.c b/src/mesa/pipe/i915simple/i915_debug_fp.c index a108f1ad4f..87fc3b2f9a 100644 --- a/src/mesa/pipe/i915simple/i915_debug_fp.c +++ b/src/mesa/pipe/i915simple/i915_debug_fp.c @@ -298,6 +298,19 @@ print_tex_op(struct debug_stream *stream, PRINTF("\n"); } +static void +print_texkil_op(struct debug_stream *stream, + unsigned opcode, const unsigned * program) +{ + PRINTF("TEXKIL "); + + print_reg_type_nr(stream, + (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & + REG_TYPE_MASK, + (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); + PRINTF("\n"); +} + static void print_dcl_op(struct debug_stream *stream, unsigned opcode, const unsigned * program) @@ -328,8 +341,10 @@ i915_disassemble_program(struct debug_stream *stream, if ((int) opcode >= A0_NOP && opcode <= A0_SLT) print_arith_op(stream, opcode >> 24, program); - else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) + else if (opcode >= T0_TEXLD && opcode < T0_TEXKILL) print_tex_op(stream, opcode >> 24, program); + else if (opcode == T0_TEXKILL) + print_texkil_op(stream, opcode >> 24, program); else if (opcode == D0_DCL) print_dcl_op(stream, opcode >> 24, program); else -- cgit v1.2.3 From 41b3fcbf0404bc4a4f6588c0feef553bb65d4440 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 10:27:12 -0600 Subject: check for extended swizzles, added TGSI_OPCODE_KILP --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 33 +++++++++++++++++++++++---- 1 file changed, 28 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index f34225bfed..e041f89140 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -197,11 +197,21 @@ src_vector(struct i915_fp_compile *p, return 0; } - src = swizzle(src, - source->SrcRegister.SwizzleX, - source->SrcRegister.SwizzleY, - source->SrcRegister.SwizzleZ, - source->SrcRegister.SwizzleW); + if (source->SrcRegister.Extended) { + src = swizzle(src, + source->SrcRegisterExtSwz.ExtSwizzleX, + source->SrcRegisterExtSwz.ExtSwizzleY, + source->SrcRegisterExtSwz.ExtSwizzleZ, + source->SrcRegisterExtSwz.ExtSwizzleW); + } + else { + src = swizzle(src, + source->SrcRegister.SwizzleX, + source->SrcRegister.SwizzleY, + source->SrcRegister.SwizzleZ, + source->SrcRegister.SwizzleW); + } + /* There's both negate-all-components and per-component negation. * Try to handle both here. @@ -512,6 +522,19 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_KIL: + /* unconditional kill */ + assert(0); /* not tested yet */ +#if 0 + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ + 0, src0, T0_TEXKILL); +#endif + break; + + case TGSI_OPCODE_KILP: + /* kill if src[0].x < 0 || src[0].y < 0 ... */ src0 = src_vector(p, &inst->FullSrcRegisters[0]); tmp = i915_get_utemp(p); -- cgit v1.2.3 From 583c11d38375c535d3835d89012fa5318bbd34ce Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 10:28:18 -0600 Subject: simplify KIL usage for bitmap fragprog --- src/mesa/state_tracker/st_cb_drawpixels.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 92705cdc44..e3479deb79 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -105,11 +105,10 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE ); ic++; - /* SUB tmp0.w, tmp0.w, tmp0.x; # tmp0.w -= 1 */ + /* SUB tmp0, tmp0.wwww, tmp0.xxxx; # tmp0.w -= 1 */ p->Instructions[ic].Opcode = OPCODE_SUB; p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; p->Instructions[ic].DstReg.Index = 0; - p->Instructions[ic].DstReg.WriteMask = WRITEMASK_W; p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; p->Instructions[ic].SrcReg[0].Index = 0; p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; @@ -118,11 +117,10 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) p->Instructions[ic].SrcReg[1].Swizzle = SWIZZLE_XXXX; /* 1.0 */ ic++; - /* KIL if tmp0.w < 0 */ + /* KIL if tmp0 < 0 */ p->Instructions[ic].Opcode = OPCODE_KIL; p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; p->Instructions[ic].SrcReg[0].Index = 0; - p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; ic++; /* MOV result.color, fragment.color */ -- cgit v1.2.3 From 66ea4b7c7c6c61cfacb3640cfc264cae575062cd Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 10:30:09 -0600 Subject: remove unused includes --- src/mesa/pipe/draw/draw_vertex.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index 6f81e6e8a3..1204a944de 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -35,8 +35,6 @@ #include "pipe/p_defines.h" -#include "pipe/p_context.h" -#include "pipe/p_winsys.h" #include "pipe/p_util.h" #include "pipe/draw/draw_private.h" -- cgit v1.2.3 From 3eeef8aabe399a404d5c86dcc23066d9b5453a7f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 10:30:38 -0600 Subject: Change 'drawing' field to boolean. --- src/mesa/pipe/draw/draw_prim.c | 4 ++-- src/mesa/pipe/draw/draw_private.h | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 698557d6bd..cf8943b40d 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -119,7 +119,7 @@ void draw_do_flush( struct draw_context *draw, draw->drawing) { draw->pipeline.first->end( draw->pipeline.first ); - draw->drawing = 0; + draw->drawing = FALSE; draw->prim = ~0; draw->pipeline.first = draw->pipeline.validate; } @@ -400,7 +400,7 @@ draw_arrays(struct draw_context *draw, unsigned prim, unsigned start, unsigned count) { if (!draw->drawing) { - draw->drawing = 1; + draw->drawing = TRUE; /* tell drawing pipeline we're beginning drawing */ draw->pipeline.first->begin( draw->pipeline.first ); diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 79b2176c59..6aa4cd44a0 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -181,7 +181,7 @@ struct draw_context uint attrib_front0, attrib_back0; uint attrib_front1, attrib_back1; - unsigned drawing; + boolean drawing; /**< do we presently have something queued for drawing? */ unsigned prim; /**< current prim type: PIPE_PRIM_x */ unsigned reduced_prim; -- cgit v1.2.3 From c8bf63e992f902f1bef0c20e5b50f397c4d219a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 11:28:25 -0600 Subject: feedback/rasterpos fix-ups --- src/mesa/pipe/i915simple/i915_context.c | 23 +++++++++++++++++++++++ src/mesa/pipe/i915simple/i915_context.h | 4 ++++ src/mesa/pipe/i915simple/i915_state.c | 30 ++++++++++++++++++++++++++++++ 3 files changed, 57 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 5d97eed247..8150fcfbf5 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -212,6 +212,19 @@ static boolean i915_draw_elements( struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } + /* Map feedback buffers if enabled */ + if (i915->feedback.enabled) { + const uint n = i915->feedback.interleaved ? 1 : i915->feedback.num_attribs; + for (i = 0; i < n; i++) { + void *ptr = pipe->winsys->buffer_map(pipe->winsys, + i915->feedback_buffer[i].buffer, + PIPE_BUFFER_FLAG_WRITE); + draw_set_mapped_feedback_buffer(draw, i, ptr, + i915->feedback_buffer[i].size); + } + } + + draw_set_mapped_constant_buffer(draw, i915->current.constants[PIPE_SHADER_VERTEX]); @@ -232,6 +245,16 @@ static boolean i915_draw_elements( struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } + /* Unmap feedback buffers if enabled */ + if (i915->feedback.enabled) { + const uint n = i915->feedback.interleaved ? 1 : i915->feedback.num_attribs; + for (i = 0; i < n; i++) { + pipe->winsys->buffer_unmap(pipe->winsys, + i915->feedback_buffer[i].buffer); + draw_set_mapped_feedback_buffer(draw, i, NULL, 0); + } + } + return TRUE; } diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index dc19b6efeb..a3dd392e75 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -169,6 +169,7 @@ struct i915_context struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; + struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; @@ -176,6 +177,9 @@ struct i915_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + /** Feedback buffers */ + struct pipe_feedback_buffer feedback_buffer[PIPE_MAX_FEEDBACK_ATTRIBS]; + unsigned dirty; unsigned *batch_start; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 19ca5e575f..4a31747fe8 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -550,6 +550,33 @@ static void i915_set_framebuffer_state(struct pipe_context *pipe, } +static void +i915_set_feedback_state(struct pipe_context *pipe, + const struct pipe_feedback_state *feedback) +{ + struct i915_context *i915 = i915_context(pipe); + i915->feedback = *feedback; + draw_set_feedback_state(i915->draw, feedback); +} + + +static void +i915_set_feedback_buffer(struct pipe_context *pipe, + unsigned index, + const struct pipe_feedback_buffer *feedback) +{ + struct i915_context *i915 = i915_context(pipe); + + assert(index < PIPE_MAX_FEEDBACK_ATTRIBS); + + /* Need to be careful with referencing */ + pipe->winsys->buffer_reference(pipe->winsys, + &i915->feedback_buffer[index].buffer, + feedback->buffer); + i915->feedback_buffer[index].size = feedback->size; + i915->feedback_buffer[index].start_offset = feedback->start_offset; +} + static void i915_set_clear_color_state(struct pipe_context *pipe, @@ -731,6 +758,9 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_clear_color_state = i915_set_clear_color_state; i915->pipe.set_constant_buffer = i915_set_constant_buffer; i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; + i915->pipe.set_feedback_state = i915_set_feedback_state; + i915->pipe.set_feedback_buffer = i915_set_feedback_buffer; + i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; i915->pipe.set_scissor_state = i915_set_scissor_state; i915->pipe.set_texture_state = i915_set_texture_state; -- cgit v1.2.3 From 369eefc34c8d7acdb881ea5b0516406d71344fc4 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 11:47:53 -0600 Subject: add 'normalized_coords' field to pipe_sampler_state This controls whether texcoords are interpreted as-is or scaled up from [0,1]. Fixes glDrawPixels/glBitmap problems on i915 when image is non power-of-two. Also, cleans up the CSO sampler state for i915 a bit. --- src/mesa/pipe/i915simple/i915_state.c | 3 +++ src/mesa/pipe/i915simple/i915_state_sampler.c | 4 ---- src/mesa/pipe/p_state.h | 1 + src/mesa/pipe/softpipe/sp_tex_sample.c | 24 ++++++++++++++++-------- src/mesa/state_tracker/st_atom_sampler.c | 3 +++ src/mesa/state_tracker/st_cb_drawpixels.c | 1 + 6 files changed, 24 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 4a31747fe8..8da5662e3f 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -262,6 +262,9 @@ i915_create_sampler_state(struct pipe_context *pipe, (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT)); + if (sampler->normalized_coords) + cso->state[1] |= SS3_NORMALIZED_COORDS; + { ubyte r = float_to_ubyte(sampler->border_color[0]); ubyte g = float_to_ubyte(sampler->border_color[1]); diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 3b736f9c55..8dec6781cd 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -127,10 +127,6 @@ static void update_sampler(struct i915_context *i915, #endif state[1] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT); - - if (is_power_of_two_texture(mt)) { - state[1] |= SS3_NORMALIZED_COORDS; - } } void i915_update_samplers( struct i915_context *i915 ) diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 99ec574124..da62aa1b27 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -243,6 +243,7 @@ struct pipe_sampler_state unsigned compare:1; /**< shadow/depth compare enabled? */ unsigned compare_mode:1; /**< PIPE_TEX_COMPARE_x */ unsigned compare_func:3; /**< PIPE_FUNC_x */ + unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */ float shadow_ambient; /**< shadow test fail color/intensity */ float min_lod; float max_lod; diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index aab26f0b0e..6ff84119e5 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -427,8 +427,9 @@ compute_lambda(struct tgsi_sampler *sampler, float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; dsdx = FABSF(dsdx); dsdy = FABSF(dsdy); - /* XXX only multiply by width for NORMALIZEd texcoords */ - rho = MAX2(dsdx, dsdy) * sampler->texture->width0; + rho = MAX2(dsdx, dsdy); + if (sampler->state->normalized_coords) + rho *= sampler->texture->width0; } if (t) { float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; @@ -436,8 +437,9 @@ compute_lambda(struct tgsi_sampler *sampler, float max; dtdx = FABSF(dtdx); dtdy = FABSF(dtdy); - /* XXX only multiply by height for NORMALIZEd texcoords */ - max = MAX2(dtdx, dtdy) * sampler->texture->height0; + max = MAX2(dtdx, dtdy); + if (sampler->state->normalized_coords) + max *= sampler->texture->height0; rho = MAX2(rho, max); } if (p) { @@ -446,8 +448,9 @@ compute_lambda(struct tgsi_sampler *sampler, float max; dpdx = FABSF(dpdx); dpdy = FABSF(dpdy); - /* XXX only multiply by depth for NORMALIZEd texcoords */ - max = MAX2(dpdx, dpdy) * sampler->texture->depth0; + max = MAX2(dpdx, dpdy); + if (sampler->state->normalized_coords) + max *= sampler->texture->depth0; rho = MAX2(rho, max); } @@ -647,8 +650,13 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, choose_mipmap_levels(sampler, s, t, p, lodbias, &level0, &level1, &levelBlend, &imgFilter); - width = sampler->texture->level[level0].width; - height = sampler->texture->level[level0].height; + if (sampler->state->normalized_coords) { + width = sampler->texture->level[level0].width; + height = sampler->texture->level[level0].height; + } + else { + width = height = 1.0; + } assert(width > 0); diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 151d724863..38de35933e 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -136,6 +136,9 @@ update_samplers(struct st_context *st) sampler.min_mip_filter = gl_filter_to_mip_filter(texobj->MinFilter); sampler.mag_img_filter = gl_filter_to_img_filter(texobj->MagFilter); + if (texobj->Target != GL_TEXTURE_RECTANGLE_ARB) + sampler.normalized_coords = 1; + sampler.lod_bias = st->ctx->Texture.Unit[u].LodBias; #if 1 sampler.min_lod = texobj->MinLod; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index e3479deb79..7d115209f4 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -558,6 +558,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.normalized_coords = 1; cso = st_cached_sampler_state(ctx->st, &sampler); pipe->bind_sampler_state(pipe, unit, cso->data); } -- cgit v1.2.3 From e3a729a681cfdc9733c833c9574459f577058bbf Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 11:48:20 -0600 Subject: fix logic for printing extended swizzles --- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 28 +++++++++++++++------------- 1 file changed, 15 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index 76241748c7..7fd503d81b 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -910,20 +910,22 @@ dump_instruction_short( SID( src->SrcRegister.Index ); CHR( ']' ); - if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || - src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || - src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || - src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { - CHR( '.' ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); + if (src->SrcRegister.Extended) { + if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { + CHR( '.' ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); + } } - if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || - src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || - src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || - src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { + else if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { CHR( '.' ); ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); -- cgit v1.2.3 From 8630e5edb36f00de30117d56c1691363d471c0cc Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 13:05:15 -0600 Subject: call pipe->set_vertex_buffer() after drawing/unreferencing. This fixes potential a stale reference to vertex buffers. Fixes cubemap demo on i915. --- src/mesa/state_tracker/st_draw.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index cc204640eb..640fa8263e 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -242,6 +242,8 @@ st_draw_vbo(GLcontext *ctx, /* unreference buffers (frees wrapped user-space buffer objects) */ for (attr = 0; attr < vs->num_inputs; attr++) { winsys->buffer_reference(winsys, &vbuffer[attr].buffer, NULL); + assert(!vbuffer[attr].buffer); + pipe->set_vertex_buffer(pipe, attr, &vbuffer[attr]); } } -- cgit v1.2.3 From 230225c5d746e4f62d71dda8c2832cc25eb28afe Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 13:06:53 -0600 Subject: formatting --- src/mesa/pipe/i915simple/i915_context.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 8150fcfbf5..161f8ce697 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -175,10 +175,11 @@ i915_end_query(struct pipe_context *pipe, struct pipe_query_object *q) } -static boolean i915_draw_elements( struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, - unsigned prim, unsigned start, unsigned count) +static boolean +i915_draw_elements( struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) { struct i915_context *i915 = i915_context( pipe ); struct draw_context *draw = i915->draw; @@ -187,8 +188,7 @@ static boolean i915_draw_elements( struct pipe_context *pipe, if (i915->dirty) i915_update_derived( i915 ); - - /* + /* * Map vertex buffers */ for (i = 0; i < PIPE_ATTRIB_MAX; i++) { -- cgit v1.2.3 From 520ad5f854723955c89584e4fa0bd172d0969e09 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 13:25:13 -0600 Subject: GL_SELECT mode works now --- src/mesa/state_tracker/st_cb_feedback.c | 7 +++++++ src/mesa/state_tracker/st_draw.c | 15 +++++++++++++++ 2 files changed, 22 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 537a58f39d..f0c6bf6185 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -293,6 +293,13 @@ st_RenderMode(GLcontext *ctx, GLenum newMode ) draw_set_rasterize_stage(draw, st->selection_stage); /* Plug in new vbo draw function */ vbo->draw_prims = st_feedback_draw_vbo; + /* setup post-transform vertex attribs */ + { + /* just emit pos as GLfloat[4] */ + static const uint attrs[1] = { FORMAT_4F }; + const interp_mode *interp = NULL; + draw_set_vertex_attributes(draw, attrs, interp, 1); + } } else { if (!st->feedback_stage) diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 640fa8263e..7fcc4d8e65 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -336,6 +336,7 @@ st_feedback_draw_vbo(GLcontext *ctx, struct pipe_buffer_handle *index_buffer_handle = 0; struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; GLuint attr, i; + ubyte *mapped_constants; assert(ctx->RenderMode == GL_SELECT || ctx->RenderMode == GL_FEEDBACK); @@ -439,6 +440,17 @@ st_feedback_draw_vbo(GLcontext *ctx, PIPE_BUFFER_FLAG_READ); draw_set_mapped_element_buffer(draw, indexSize, map); } + else { + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + + /* map constant buffers */ + mapped_constants = winsys->buffer_map(winsys, + st->state.constants[PIPE_SHADER_VERTEX].buffer, + PIPE_BUFFER_FLAG_READ); + draw_set_mapped_constant_buffer(st->draw, mapped_constants); /* draw here */ @@ -447,6 +459,9 @@ st_feedback_draw_vbo(GLcontext *ctx, } + /* unmap constant buffers */ + winsys->buffer_unmap(winsys, st->state.constants[PIPE_SHADER_VERTEX].buffer); + /* * unmap vertex/index buffers */ -- cgit v1.2.3 From 65b1f6947f0d4d83a942478383d27ba84a02b20f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 13:42:56 -0600 Subject: setup vertex format for GL_FEEDBACK mode --- src/mesa/state_tracker/st_cb_feedback.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index f0c6bf6185..5e334e3c85 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -54,6 +54,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" +#include "pipe/cso_cache/cso_cache.h" #include "vf/vf.h" #include "pipe/draw/draw_context.h" @@ -307,6 +308,19 @@ st_RenderMode(GLcontext *ctx, GLenum newMode ) draw_set_rasterize_stage(draw, st->feedback_stage); /* Plug in new vbo draw function */ vbo->draw_prims = st_feedback_draw_vbo; + /* setup post-transform vertex attribs */ + { + /* emit all attribs as GLfloat[4] */ + uint attrs[PIPE_MAX_SHADER_OUTPUTS]; + interp_mode interp[PIPE_MAX_SHADER_OUTPUTS]; + GLuint n = st->state.vs->state.num_outputs; + GLuint i; + for (i = 0; i < n; i++) { + attrs[i] = FORMAT_4F; + interp[i] = INTERP_NONE; + } + draw_set_vertex_attributes(draw, attrs, interp, n); + } } } -- cgit v1.2.3 From 0585b4e746f35efbc8577311eab6679951217ac3 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 15:05:08 -0600 Subject: Undo prev changes. --- src/mesa/state_tracker/st_cb_feedback.c | 29 ++++++----------------------- src/mesa/state_tracker/st_draw.c | 30 ++++++++++++++++++++++++++++++ src/mesa/tnl/t_vp_build.c | 7 +++++++ 3 files changed, 43 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 5e334e3c85..923e1cdb89 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -93,8 +93,9 @@ feedback_vertex(GLcontext *ctx, const struct draw_context *draw, const GLfloat ci = 0; GLuint slot; + /* Recall that Y=0=Top of window for Gallium wincoords */ win[0] = v->data[0][0]; - win[1] = v->data[0][1]; + win[1] = ctx->DrawBuffer->Height - v->data[0][1]; win[2] = v->data[0][2]; win[3] = 1.0F / v->data[0][3]; @@ -104,13 +105,13 @@ feedback_vertex(GLcontext *ctx, const struct draw_context *draw, */ slot = st->vertex_result_to_slot[VERT_RESULT_COL0]; - if (slot) + if (slot != ~0) color = v->data[slot]; else color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; slot = st->vertex_result_to_slot[VERT_RESULT_TEX0]; - if (slot) + if (slot != ~0) texcoord = v->data[slot]; else texcoord = ctx->Current.Attrib[VERT_ATTRIB_TEX0]; @@ -294,13 +295,6 @@ st_RenderMode(GLcontext *ctx, GLenum newMode ) draw_set_rasterize_stage(draw, st->selection_stage); /* Plug in new vbo draw function */ vbo->draw_prims = st_feedback_draw_vbo; - /* setup post-transform vertex attribs */ - { - /* just emit pos as GLfloat[4] */ - static const uint attrs[1] = { FORMAT_4F }; - const interp_mode *interp = NULL; - draw_set_vertex_attributes(draw, attrs, interp, 1); - } } else { if (!st->feedback_stage) @@ -308,19 +302,8 @@ st_RenderMode(GLcontext *ctx, GLenum newMode ) draw_set_rasterize_stage(draw, st->feedback_stage); /* Plug in new vbo draw function */ vbo->draw_prims = st_feedback_draw_vbo; - /* setup post-transform vertex attribs */ - { - /* emit all attribs as GLfloat[4] */ - uint attrs[PIPE_MAX_SHADER_OUTPUTS]; - interp_mode interp[PIPE_MAX_SHADER_OUTPUTS]; - GLuint n = st->state.vs->state.num_outputs; - GLuint i; - for (i = 0; i < n; i++) { - attrs[i] = FORMAT_4F; - interp[i] = INTERP_NONE; - } - draw_set_vertex_attributes(draw, attrs, interp, n); - } + /* need to generate/use a vertex program that emits pos/color/tex */ + st->dirty.st |= ST_NEW_VERTEX_PROGRAM; } } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 7fcc4d8e65..1806a1888b 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -310,6 +310,35 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, } +/** + * Set the (private) draw module's post-transformed vertex format when in + * GL_SELECT or GL_FEEDBACK mode. + */ +static void +set_feedback_vertex_format(GLcontext *ctx) +{ + struct st_context *st = ctx->st; + uint attrs[PIPE_MAX_SHADER_OUTPUTS]; + interp_mode interp[PIPE_MAX_SHADER_OUTPUTS]; + GLuint n, i; + + if (ctx->RenderMode == GL_FEEDBACK) { + /* emit all attribs (pos, color, texcoord) as GLfloat[4] */ + n = st->state.vs->state.num_outputs; + for (i = 0; i < n; i++) { + attrs[i] = FORMAT_4F; + interp[i] = INTERP_NONE; + } + } + else { + assert(ctx->RenderMode == GL_SELECT); + n = 1; + attrs[0] = FORMAT_4F; + interp[0] = INTERP_NONE; + } + + draw_set_vertex_attributes(st->draw, attrs, interp, n); +} /** @@ -359,6 +388,7 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_clip_state(draw, &st->state.clip); draw_set_rasterizer_state(draw, &st->state.rasterizer->state); draw_bind_vertex_shader(draw, st->state.vs->data); + set_feedback_vertex_format(ctx); /* loop over TGSI shader inputs to determine vertex buffer * and attribute info diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index b28102b1b9..a588b0d16b 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -134,6 +134,13 @@ static struct state_key *make_state_key( GLcontext *ctx ) assert(fp); key->fragprog_inputs_read = fp->Base.InputsRead; + if (ctx->RenderMode == GL_FEEDBACK) { + /* This is a bit of a hack, but you can imagine feedback mode as using + * a special no-op fragment shader that just requires particular + * inputs (inputs which satisfy feedback mode). + */ + key->fragprog_inputs_read |= FRAG_BIT_COL0 | FRAG_BIT_TEX0; + } key->separate_specular = (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR); -- cgit v1.2.3 From b4deb73fabf8aa545ba6ac25ca08f5d05ede2178 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 15:05:39 -0600 Subject: remove feedback hack --- src/mesa/tnl/t_vp_build.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index a588b0d16b..b28102b1b9 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -134,13 +134,6 @@ static struct state_key *make_state_key( GLcontext *ctx ) assert(fp); key->fragprog_inputs_read = fp->Base.InputsRead; - if (ctx->RenderMode == GL_FEEDBACK) { - /* This is a bit of a hack, but you can imagine feedback mode as using - * a special no-op fragment shader that just requires particular - * inputs (inputs which satisfy feedback mode). - */ - key->fragprog_inputs_read |= FRAG_BIT_COL0 | FRAG_BIT_TEX0; - } key->separate_specular = (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR); -- cgit v1.2.3 From b1f136dba21c688a2fa347a56ab8bf8c784dd51f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 15:30:13 -0600 Subject: Fix useabs logic in build_fog(). We always need to compute the absolute value of the fogcoord if we're passing it through for per-fragment fog. --- src/mesa/tnl/t_vp_build.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index b28102b1b9..e7caa2fc1d 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -1110,8 +1110,6 @@ static void build_fog( struct tnl_program *p ) { struct ureg fog = register_output(p, VERT_RESULT_FOGC); struct ureg input; - GLuint useabs = p->state->fog_source_is_depth && p->state->fog_mode && - (p->state->fog_mode != FOG_EXP2); if (p->state->fog_source_is_depth) { input = swizzle1(get_eye_position(p), Z); @@ -1125,6 +1123,8 @@ static void build_fog( struct tnl_program *p ) STATE_FOG_PARAMS_OPTIMIZED); struct ureg tmp = get_temp(p); + GLboolean useabs = (p->state->fog_mode != FOG_EXP2); + if (useabs) { emit_op1(p, OPCODE_ABS, tmp, 0, input); } @@ -1156,7 +1156,10 @@ static void build_fog( struct tnl_program *p ) /* results = incoming fog coords (compute fog per-fragment later) * * KW: Is it really necessary to do anything in this case? + * BP: Yes, we always need to compute the absolute value, unless + * we want to push that down into the fragment program... */ + GLboolean useabs = GL_TRUE; emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, WRITEMASK_X, input); } } -- cgit v1.2.3 From ce8988018ca2d838c93df904271b2afc62d6b021 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 15:30:45 -0600 Subject: check for width or height = 0 before calling driver Bitmap func --- src/mesa/main/drawpix.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c index c82abccc41..296dcf8f0b 100644 --- a/src/mesa/main/drawpix.c +++ b/src/mesa/main/drawpix.c @@ -341,7 +341,7 @@ _mesa_Bitmap( GLsizei width, GLsizei height, } if (ctx->RenderMode == GL_RENDER) { - if (bitmap) { + if (bitmap && width && height) { /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */ GLint x = IFLOOR(ctx->Current.RasterPos[0] - xorig); GLint y = IFLOOR(ctx->Current.RasterPos[1] - yorig); -- cgit v1.2.3 From 566c11254f09259311f64816786375eeeee9e61b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 15:31:24 -0600 Subject: remove commented out #include --- src/mesa/pipe/i915simple/i915_clear.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_clear.c b/src/mesa/pipe/i915simple/i915_clear.c index a4dba1a764..e8087df5a4 100644 --- a/src/mesa/pipe/i915simple/i915_clear.c +++ b/src/mesa/pipe/i915simple/i915_clear.c @@ -33,7 +33,6 @@ #include "pipe/p_defines.h" #include "i915_context.h" #include "i915_state.h" -//#include "colormac.h" /** -- cgit v1.2.3 From 7abc78a9678cf09e7a39869549ed369802c43d58 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 17:26:01 -0600 Subject: implement logicop --- src/mesa/pipe/softpipe/sp_quad.c | 29 +++++++- src/mesa/pipe/softpipe/sp_quad_blend.c | 118 ++++++++++++++++++++++++++++++++- 2 files changed, 145 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 74a21dc54b..429497e9b2 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -1,3 +1,29 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ #include "sp_context.h" @@ -16,7 +42,8 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->quad.first = sp->quad.colormask; } - if (sp->blend->blend_enable) { + if (sp->blend->blend_enable || + sp->blend->logicop_enable) { sp->quad.blend->next = sp->quad.first; sp->quad.first = sp->quad.blend; } diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 6c7135e53c..b56e51c104 100755 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -96,6 +96,116 @@ do { \ +static void +logicop_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); + float dest[4][QUAD_SIZE]; + ubyte src[4][4], dst[4][4], res[4][4]; + uint *src4 = (uint *) src; + uint *dst4 = (uint *) dst; + uint *res4 = (uint *) res; + uint j; + + /* get colors from framebuffer */ + sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + /* convert to ubyte */ + for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][0], dest[j][0]); /* P0 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][1], dest[j][1]); /* P1 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][2], dest[j][2]); /* P2 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][3], dest[j][3]); /* P3 */ + + UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quad->outputs.color[j][0]); /* P0 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quad->outputs.color[j][1]); /* P1 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quad->outputs.color[j][2]); /* P2 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quad->outputs.color[j][3]); /* P3 */ + } + + switch (softpipe->blend->logicop_func) { + case PIPE_LOGICOP_CLEAR: + for (j = 0; j < 4; j++) + res4[j] = 0; + break; + case PIPE_LOGICOP_NOR: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] | dst4[j]); + break; + case PIPE_LOGICOP_AND_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j] & dst4[j]; + break; + case PIPE_LOGICOP_COPY_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j]; + break; + case PIPE_LOGICOP_AND_REVERSE: + for (j = 0; j < 4; j++) + res4[j] = src4[j] & ~dst4[j]; + break; + case PIPE_LOGICOP_INVERT: + for (j = 0; j < 4; j++) + res4[j] = ~dst4[j]; + break; + case PIPE_LOGICOP_XOR: + for (j = 0; j < 4; j++) + res4[j] = dst4[j] ^ src4[j]; + break; + case PIPE_LOGICOP_NAND: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] & dst4[j]); + break; + case PIPE_LOGICOP_AND: + for (j = 0; j < 4; j++) + res4[j] = src4[j] & dst4[j]; + break; + case PIPE_LOGICOP_EQUIV: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] ^ dst4[j]); + break; + case PIPE_LOGICOP_NOOP: + for (j = 0; j < 4; j++) + res4[j] = dst4[j]; + break; + case PIPE_LOGICOP_OR_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j] | dst4[j]; + break; + case PIPE_LOGICOP_COPY: + for (j = 0; j < 4; j++) + res4[j] = src4[j]; + break; + case PIPE_LOGICOP_OR_REVERSE: + for (j = 0; j < 4; j++) + res4[j] = src4[j] | ~dst4[j]; + break; + case PIPE_LOGICOP_OR: + for (j = 0; j < 4; j++) + res4[j] = src4[j] | dst4[j]; + break; + case PIPE_LOGICOP_SET: + for (j = 0; j < 4; j++) + res4[j] = ~0; + break; + default: + assert(0); + } + + for (j = 0; j < 4; j++) { + quad->outputs.color[j][0] = UBYTE_TO_FLOAT(res[j][0]); + quad->outputs.color[j][1] = UBYTE_TO_FLOAT(res[j][1]); + quad->outputs.color[j][2] = UBYTE_TO_FLOAT(res[j][2]); + quad->outputs.color[j][3] = UBYTE_TO_FLOAT(res[j][3]); + } + + /* pass quad to next stage */ + qs->next->run(qs->next, quad); +} + + + + static void blend_quad(struct quad_stage *qs, struct quad_header *quad) { @@ -104,7 +214,12 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; - + + if (softpipe->blend->logicop_enable) { + logicop_quad(qs, quad); + return; + } + /* get colors from framebuffer */ sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); @@ -371,6 +486,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) break; case PIPE_BLEND_MAX: VEC4_MAX(quad->outputs.color[3], source[3], dest[3]); /* A */ + break; default: abort(); } -- cgit v1.2.3 From b2605d5f803f51f216fd815374b59dfc616f7e13 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 17:27:54 -0600 Subject: GL_UNSIGNED_BYTE indexes --- src/mesa/state_tracker/st_draw.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 1806a1888b..fbf7713f58 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -206,6 +206,9 @@ st_draw_vbo(GLcontext *ctx, case GL_UNSIGNED_SHORT: indexSize = 2; break; + case GL_UNSIGNED_BYTE: + indexSize = 1; + break; default: assert(0); } -- cgit v1.2.3 From cd54414b75476b793867b863da0cbb9f079f613c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 17:59:16 -0600 Subject: fix incorrect register in get_eye_normal() for rescale_normals case --- src/mesa/tnl/t_vp_build.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index e7caa2fc1d..2c0883d0b0 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -693,7 +693,7 @@ static struct ureg get_eye_normal( struct tnl_program *p ) struct ureg rescale = register_param2(p, STATE_INTERNAL, STATE_NORMAL_SCALE); - emit_op2( p, OPCODE_MUL, p->eye_normal, 0, normal, + emit_op2( p, OPCODE_MUL, p->eye_normal, 0, p->eye_normal, swizzle1(rescale, X)); } } @@ -1122,7 +1122,6 @@ static void build_fog( struct tnl_program *p ) struct ureg params = register_param2(p, STATE_INTERNAL, STATE_FOG_PARAMS_OPTIMIZED); struct ureg tmp = get_temp(p); - GLboolean useabs = (p->state->fog_mode != FOG_EXP2); if (useabs) { -- cgit v1.2.3 From c838d4c8763135a912616eceb6b6ced09f5762e0 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 15 Oct 2007 19:14:40 -0600 Subject: finish remaining blend modes --- src/mesa/pipe/softpipe/sp_quad_blend.c | 231 +++++++++++++++++++++++++++++++-- 1 file changed, 218 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index b56e51c104..3af5ab5b29 100755 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -259,7 +259,14 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) } break; case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - assert(0); /* to do */ + { + const float *alpha = quad->outputs.color[3]; + float diff[4]; + VEC4_SUB(diff, one, dest[3]); + VEC4_MIN(source[0], alpha, diff); /* R */ + VEC4_MIN(source[1], alpha, diff); /* G */ + VEC4_MIN(source[2], alpha, diff); /* B */ + } break; case PIPE_BLENDFACTOR_CONST_COLOR: { @@ -372,16 +379,67 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_ONE: VEC4_COPY(source[3], quad->outputs.color[3]); /* A */ break; + case PIPE_BLENDFACTOR_SRC_COLOR: + /* fall-through */ case PIPE_BLENDFACTOR_SRC_ALPHA: { const float *alpha = quad->outputs.color[3]; VEC4_MUL(source[3], quad->outputs.color[3], alpha); /* A */ } break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(source[3], quad->outputs.color[3], dest[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + { + const float *alpha = quad->outputs.color[3]; + float diff[4]; + VEC4_SUB(diff, one, dest[3]); + VEC4_MIN(source[3], alpha, diff); /* A */ + } + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(source[3], quad->outputs.color[3], comp); /* A */ + } + break; case PIPE_BLENDFACTOR_ZERO: VEC4_COPY(source[3], zero); /* A */ break; - /* XXX fill in remaining terms */ + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, quad->outputs.color[3]); + VEC4_MUL(source[3], quad->outputs.color[3], inv_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, dest[3]); + VEC4_MUL(source[3], quad->outputs.color[3], inv_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + /* A */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(source[3], quad->outputs.color[3], inv_comp); + } + break; default: abort(); } @@ -394,6 +452,70 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_ONE: /* dest = dest * 1 NO-OP, leave dest as-is */ break; + case PIPE_BLENDFACTOR_SRC_COLOR: + VEC4_MUL(dest[0], dest[0], quad->outputs.color[0]); /* R */ + VEC4_MUL(dest[1], dest[1], quad->outputs.color[1]); /* G */ + VEC4_MUL(dest[2], dest[2], quad->outputs.color[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + VEC4_MUL(dest[0], dest[0], quad->outputs.color[3]); /* R * A */ + VEC4_MUL(dest[1], dest[1], quad->outputs.color[3]); /* G * A */ + VEC4_MUL(dest[2], dest[2], quad->outputs.color[3]); /* B * A */ + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */ + VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */ + VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */ + break; + case PIPE_BLENDFACTOR_DST_COLOR: + VEC4_MUL(dest[0], dest[0], dest[0]); /* R */ + VEC4_MUL(dest[1], dest[1], dest[1]); /* G */ + VEC4_MUL(dest[2], dest[2], dest[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* illegal */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ + VEC4_MUL(dest[0], dest[0], comp); /* R */ + VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ + VEC4_MUL(dest[1], dest[1], comp); /* G */ + VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ + VEC4_MUL(dest[2], dest[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(dest[0], dest[0], comp); /* R */ + VEC4_MUL(dest[1], dest[1], comp); /* G */ + VEC4_MUL(dest[2], dest[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[0], zero); /* R */ + VEC4_COPY(dest[1], zero); /* G */ + VEC4_COPY(dest[2], zero); /* B */ + break; + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: + /* XXX what are these? */ + assert(0); + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quad->outputs.color[0]); /* R */ + VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ + VEC4_SUB(inv_comp, one, quad->outputs.color[1]); /* G */ + VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ + VEC4_SUB(inv_comp, one, quad->outputs.color[2]); /* B */ + VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ + } + break; case PIPE_BLENDFACTOR_INV_SRC_ALPHA: { float one_minus_alpha[QUAD_SIZE]; @@ -403,14 +525,56 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ } break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[0], zero); /* R */ - VEC4_COPY(dest[1], zero); /* G */ - VEC4_COPY(dest[2], zero); /* B */ + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quad->outputs.color[3]); /* A */ + VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ + VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ + VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[0]); /* R */ + VEC4_MUL(dest[0], dest[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, dest[1]); /* G */ + VEC4_MUL(dest[1], dest[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, dest[2]); /* B */ + VEC4_MUL(dest[2], dest[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + { + float inv_comp[4]; + /* R */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); + VEC4_MUL(dest[0], dest[0], inv_comp); + /* G */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); + VEC4_MUL(dest[1], dest[1], inv_comp); + /* B */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); + VEC4_MUL(dest[2], dest[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(dest[0], dest[0], inv_comp); + VEC4_MUL(dest[1], dest[1], inv_comp); + VEC4_MUL(dest[2], dest[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + /* XXX what are these? */ + assert(0); break; - /* XXX fill in remaining terms */ default: - abort(); + assert(0); } /* @@ -420,6 +584,33 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_ONE: /* dest = dest * 1 NO-OP, leave dest as-is */ break; + case PIPE_BLENDFACTOR_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_SRC_ALPHA: + VEC4_MUL(dest[3], dest[3], quad->outputs.color[3]); /* A * A */ + break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(dest[3], dest[3], dest[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* illegal */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(dest[3], dest[3], comp); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[3], zero); /* A */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* fall-through */ case PIPE_BLENDFACTOR_INV_SRC_ALPHA: { float one_minus_alpha[QUAD_SIZE]; @@ -427,12 +618,26 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ } break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[3], zero); /* A */ + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[3]); /* A */ + VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(dest[3], dest[3], inv_comp); + } break; - /* XXX fill in remaining terms */ default: - abort(); + assert(0); } /* @@ -465,7 +670,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) VEC4_MAX(quad->outputs.color[2], source[2], dest[2]); /* B */ break; default: - abort(); + assert(0); } /* -- cgit v1.2.3 From ce0f2e88e3f71427dabf0dc37900ce0b47ae8003 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 08:53:08 -0600 Subject: fix blend term translation bug, rename some funcs --- src/mesa/state_tracker/st_atom_blend.c | 26 +++++++++++++------------- 1 file changed, 13 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index f321cd3829..2a9d209153 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -45,7 +45,7 @@ * Both blend factors and blend funcs are accepted. */ static GLuint -gl_blend_to_sp(GLenum blend) +translate_blend(GLenum blend) { switch (blend) { /* blend functions */ @@ -88,9 +88,9 @@ gl_blend_to_sp(GLenum blend) case GL_ONE_MINUS_SRC_ALPHA: return PIPE_BLENDFACTOR_INV_SRC_ALPHA; case GL_ONE_MINUS_DST_COLOR: - return PIPE_BLENDFACTOR_INV_DST_ALPHA; - case GL_ONE_MINUS_DST_ALPHA: return PIPE_BLENDFACTOR_INV_DST_COLOR; + case GL_ONE_MINUS_DST_ALPHA: + return PIPE_BLENDFACTOR_INV_DST_ALPHA; case GL_ONE_MINUS_CONSTANT_COLOR: return PIPE_BLENDFACTOR_INV_CONST_COLOR; case GL_ONE_MINUS_CONSTANT_ALPHA: @@ -100,7 +100,7 @@ gl_blend_to_sp(GLenum blend) return PIPE_BLENDFACTOR_INV_SRC1_ALPHA; */ default: - assert("invalid GL token in gl_blend_to_sp()" == NULL); + assert("invalid GL token in translate_blend()" == NULL); return 0; } } @@ -110,7 +110,7 @@ gl_blend_to_sp(GLenum blend) * Convert GLenum logicop tokens to pipe tokens. */ static GLuint -gl_logicop_to_sp(GLenum logicop) +translate_logicop(GLenum logicop) { switch (logicop) { case GL_CLEAR: @@ -146,7 +146,7 @@ gl_logicop_to_sp(GLenum logicop) case GL_SET: return PIPE_LOGICOP_SET; default: - assert("invalid GL token in gl_logicop_to_sp()" == NULL); + assert("invalid GL token in translate_logicop()" == NULL); return 0; } } @@ -165,13 +165,13 @@ update_blend( struct st_context *st ) st->ctx->Color.BlendEquationRGB == GL_LOGIC_OP)) { /* logicop enabled */ blend.logicop_enable = 1; - blend.logicop_func = gl_logicop_to_sp(st->ctx->Color.LogicOp); + blend.logicop_func = translate_logicop(st->ctx->Color.LogicOp); } else if (st->ctx->Color.BlendEnabled) { /* blending enabled */ blend.blend_enable = 1; - blend.rgb_func = gl_blend_to_sp(st->ctx->Color.BlendEquationRGB); + blend.rgb_func = translate_blend(st->ctx->Color.BlendEquationRGB); if (st->ctx->Color.BlendEquationRGB == GL_MIN || st->ctx->Color.BlendEquationRGB == GL_MAX) { /* Min/max are special */ @@ -179,11 +179,11 @@ update_blend( struct st_context *st ) blend.rgb_dst_factor = PIPE_BLENDFACTOR_ONE; } else { - blend.rgb_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcRGB); - blend.rgb_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstRGB); + blend.rgb_src_factor = translate_blend(st->ctx->Color.BlendSrcRGB); + blend.rgb_dst_factor = translate_blend(st->ctx->Color.BlendDstRGB); } - blend.alpha_func = gl_blend_to_sp(st->ctx->Color.BlendEquationA); + blend.alpha_func = translate_blend(st->ctx->Color.BlendEquationA); if (st->ctx->Color.BlendEquationA == GL_MIN || st->ctx->Color.BlendEquationA == GL_MAX) { /* Min/max are special */ @@ -191,8 +191,8 @@ update_blend( struct st_context *st ) blend.alpha_dst_factor = PIPE_BLENDFACTOR_ONE; } else { - blend.alpha_src_factor = gl_blend_to_sp(st->ctx->Color.BlendSrcA); - blend.alpha_dst_factor = gl_blend_to_sp(st->ctx->Color.BlendDstA); + blend.alpha_src_factor = translate_blend(st->ctx->Color.BlendSrcA); + blend.alpha_dst_factor = translate_blend(st->ctx->Color.BlendDstA); } } else { -- cgit v1.2.3 From 63b0b5b6c74fcca7d84ac13b893a1f1f5becf39d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 12:04:16 -0600 Subject: fix perspective-interpolated attribs for points, liens --- src/mesa/pipe/softpipe/sp_prim_setup.c | 50 +++++++++++++++++++++++++++++----- 1 file changed, 43 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 2e27d00acf..7c81faf9df 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -659,9 +659,9 @@ static void setup_tri( struct draw_stage *stage, static void line_linear_coeff(struct setup_stage *setup, unsigned slot, unsigned i) { - const float dz = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - const float dadx = dz * setup->emaj.dx * setup->oneoverarea; - const float dady = dz * setup->emaj.dy * setup->oneoverarea; + const float da = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + const float dadx = da * setup->emaj.dx * setup->oneoverarea; + const float dady = da * setup->emaj.dy * setup->oneoverarea; setup->coef[slot].dadx[i] = dadx; setup->coef[slot].dady[i] = dady; setup->coef[slot].a0[i] @@ -678,8 +678,19 @@ line_linear_coeff(struct setup_stage *setup, unsigned slot, unsigned i) static void line_persp_coeff(struct setup_stage *setup, unsigned slot, unsigned i) { - /* XXX to do */ - line_linear_coeff(setup, slot, i); /* XXX temporary */ + /* XXX double-check/verify this arithmetic */ + const float a0 = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; + const float a1 = setup->vmax->data[slot][i] * setup->vmin->data[0][3]; + const float da = a1 - a0; + const float dadx = da * setup->emaj.dx * setup->oneoverarea; + const float dady = da * setup->emaj.dy * setup->oneoverarea; + setup->coef[slot].dadx[i] = dadx; + setup->coef[slot].dady[i] = dady; + setup->coef[slot].a0[i] + = (setup->vmin->data[slot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); + } @@ -895,6 +906,18 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) } +static void +point_persp_coeff(struct setup_stage *setup, const struct vertex_header *vert, + uint slot, uint i) +{ + assert(slot < PIPE_MAX_SHADER_INPUTS); + assert(i <= 3); + setup->coef[slot].dadx[i] = 0.0F; + setup->coef[slot].dady[i] = 0.0F; + setup->coef[slot].a0[i] = vert->data[slot][i] * vert->data[0][3]; +} + + /** * Do setup for point rasterization, then render the point. * Round or square points... @@ -904,6 +927,7 @@ static void setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); + const interp_mode *interp = setup->softpipe->vertex_info.interp_mode; const struct vertex_header *v0 = prim->v[0]; const int sizeAttr = setup->softpipe->psize_slot; const float halfSize @@ -934,8 +958,20 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const_coeff(setup, 0, 2); const_coeff(setup, 0, 3); for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); + switch (interp[slot]) { + case INTERP_CONSTANT: + /* fall-through */ + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + point_persp_coeff(setup, v0, slot, j); + break; + default: + assert(0); + } } setup->quad.prim = PRIM_POINT; -- cgit v1.2.3 From 0edd490a96a53f83d2fb18a570cf20a4a0c5ee40 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 17:36:37 -0600 Subject: Redefine QUAD_TOP_LEFT, TOP_RIGHT, etc. to reflect Y=0=TOP raster layout. --- src/mesa/pipe/softpipe/sp_headers.h | 19 ++++++++++--------- src/mesa/pipe/softpipe/sp_prim_setup.c | 12 ++++++------ 2 files changed, 16 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index b377c34712..e23742f803 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -39,17 +39,18 @@ /* The rasterizer generates 2x2 quads of fragment and feeds them to - * the current fp_machine (see below). + * the current fp_machine (see below). + * Remember that Y=0=top with Y increasing down the window. */ -#define QUAD_BOTTOM_LEFT 0 -#define QUAD_BOTTOM_RIGHT 1 -#define QUAD_TOP_LEFT 2 -#define QUAD_TOP_RIGHT 3 +#define QUAD_TOP_LEFT 0 +#define QUAD_TOP_RIGHT 1 +#define QUAD_BOTTOM_LEFT 2 +#define QUAD_BOTTOM_RIGHT 3 -#define MASK_BOTTOM_LEFT 0x1 -#define MASK_BOTTOM_RIGHT 0x2 -#define MASK_TOP_LEFT 0x4 -#define MASK_TOP_RIGHT 0x8 +#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) +#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) +#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) +#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) #define MASK_ALL 0xf diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 7c81faf9df..598c2f6c22 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -125,11 +125,11 @@ quad_clip(struct setup_stage *setup) if (setup->quad.x0 < minx) setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); if (setup->quad.y0 < miny) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); if (setup->quad.x0 == maxx - 1) setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); if (setup->quad.y0 == maxy - 1) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); } @@ -184,16 +184,16 @@ static unsigned calculate_mask( struct setup_stage *setup, unsigned mask = 0; if (x >= setup->span.left[0] && x < setup->span.right[0]) - mask |= MASK_BOTTOM_LEFT; + mask |= MASK_TOP_LEFT; if (x >= setup->span.left[1] && x < setup->span.right[1]) - mask |= MASK_TOP_LEFT; + mask |= MASK_BOTTOM_LEFT; if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) - mask |= MASK_BOTTOM_RIGHT; + mask |= MASK_TOP_RIGHT; if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) - mask |= MASK_TOP_RIGHT; + mask |= MASK_BOTTOM_RIGHT; return mask; } -- cgit v1.2.3 From bd35c53143560177a045b314c9b4196c229f4a4c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 17:37:46 -0600 Subject: Fix broken polygon stippling (see comments for details). --- src/mesa/pipe/softpipe/sp_quad_stipple.c | 32 +++++++++++++++++++++++++++----- 1 file changed, 27 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index cb127095d7..b2658ff158 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -16,18 +16,40 @@ static void stipple_quad(struct quad_stage *qs, struct quad_header *quad) { + static const uint bit31 = 1 << 31; + static const uint bit30 = 1 << 30; + if (quad->prim == PRIM_TRI) { struct softpipe_context *softpipe = qs->softpipe; + /* need to invert Y to index into OpenGL's stipple pattern */ + const int y0 = softpipe->framebuffer.cbufs[0]->height - 1 - quad->y0; + const int y1 = y0 - 1; + const unsigned stipple0 = softpipe->poly_stipple.stipple[y0 % 32]; + const unsigned stipple1 = softpipe->poly_stipple.stipple[y1 % 32]; + +#if 1 const int col0 = quad->x0 % 32; - const int row0 = quad->y0 % 32; - const unsigned stipple0 = softpipe->poly_stipple.stipple[row0]; - const unsigned stipple1 = softpipe->poly_stipple.stipple[row0 + 1]; + if ((stipple0 & (bit31 >> col0)) == 0) + quad->mask &= ~MASK_TOP_LEFT; + + if ((stipple0 & (bit30 >> col0)) == 0) + quad->mask &= ~MASK_TOP_RIGHT; + + if ((stipple1 & (bit31 >> col0)) == 0) + quad->mask &= ~MASK_BOTTOM_LEFT; - /* XXX there may be a better way to lay out the stored stipple - * values to further simplify this computation. + if ((stipple1 & (bit30 >> col0)) == 0) + quad->mask &= ~MASK_BOTTOM_RIGHT; +#else + /* We'd like to use this code, but we'd need to redefine + * MASK_TOP_LEFT to be (1 << 1) and MASK_TOP_RIGHT to be (1 << 0), + * and similarly for the BOTTOM bits. But that may have undesirable + * side effects elsewhere. */ + const int col0 = 30 - (quad->x0 % 32); quad->mask &= (((stipple0 >> col0) & 0x3) | (((stipple1 >> col0) & 0x3) << 2)); +#endif if (quad->mask) qs->next->run(qs->next, quad); -- cgit v1.2.3 From efdacc90b4f8262706715f4bf918cf11378922be Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 17:43:41 -0600 Subject: Update mask calculations for point drawing. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 598c2f6c22..4749f20697 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -1011,8 +1011,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) dist2 = dx * dx + dy * dy; if (dist2 <= rmax2) { cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_BOTTOM_LEFT; + setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_TOP_LEFT; } dx = (ix + 1.5f) - x; @@ -1020,8 +1020,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) dist2 = dx * dx + dy * dy; if (dist2 <= rmax2) { cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_BOTTOM_RIGHT; + setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_TOP_RIGHT; } dx = (ix + 0.5f) - x; @@ -1029,8 +1029,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) dist2 = dx * dx + dy * dy; if (dist2 <= rmax2) { cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_TOP_LEFT; + setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_BOTTOM_LEFT; } dx = (ix + 1.5f) - x; @@ -1038,8 +1038,8 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) dist2 = dx * dx + dy * dy; if (dist2 <= rmax2) { cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_TOP_RIGHT; + setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_BOTTOM_RIGHT; } if (setup->quad.mask) { @@ -1068,12 +1068,12 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) if (iy + 0.5 < y - halfSize) { /* below the bottom edge */ - setup->quad.mask &= ~(MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + setup->quad.mask &= ~(MASK_TOP_LEFT | MASK_TOP_RIGHT); } if (iy + 1.5 > y + halfSize) { /* above the top edge */ - setup->quad.mask &= ~(MASK_TOP_LEFT | MASK_TOP_RIGHT); + setup->quad.mask &= ~(MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); } if (setup->quad.mask) { -- cgit v1.2.3 From 29cfec1581769c1bf25a221dbc89623722d422c2 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 17:56:13 -0600 Subject: remove debug assertion --- src/mesa/state_tracker/st_cb_drawpixels.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 7d115209f4..97d4f41ed9 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -954,7 +954,6 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, if (unpack->LsbFirst) { /* Lsb first */ GLubyte mask = 1U << (unpack->SkipPixels & 0x7); - assert(0); for (col = 0; col < width; col++) { /* set texel to 255 if bit is set */ -- cgit v1.2.3 From 495961716430b1ff99078abe1a7ac9bc07dd6116 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 17:59:24 -0600 Subject: front/back determination was wrong --- src/mesa/pipe/draw/draw_unfilled.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index 2d374329d8..6cab086a45 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -125,7 +125,7 @@ static void unfilled_tri( struct draw_stage *stage, struct prim_header *header ) { struct unfilled_stage *unfilled = unfilled_stage(stage); - unsigned mode = unfilled->mode[header->det < 0.0]; + unsigned mode = unfilled->mode[header->det >= 0.0]; switch (mode) { case PIPE_POLYGON_MODE_FILL: -- cgit v1.2.3 From 2de9477feea4b2d3996a2118b36924fe8474a7d5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 18:45:28 -0600 Subject: New st_clear_accum_buffer() function (can't use pipe->clear() since it doesn't handle negative color values) Also, remove unneeded clamping in the accum ops. --- src/mesa/state_tracker/st_cb_accum.c | 44 +++++++++++++++++++++++++++++++----- src/mesa/state_tracker/st_cb_accum.h | 3 +++ 2 files changed, 41 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 03c65db70b..192bdbab63 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -51,6 +51,41 @@ */ +void +st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_renderbuffer *acc_strb = st_renderbuffer(rb); + struct pipe_surface *acc_ps = acc_strb->surface; + const GLint xpos = ctx->DrawBuffer->_Xmin; + const GLint ypos = ctx->DrawBuffer->_Ymin; + const GLint width = ctx->DrawBuffer->_Xmax - xpos; + const GLint height = ctx->DrawBuffer->_Ymax - ypos; + const GLfloat r = ctx->Accum.ClearColor[0]; + const GLfloat g = ctx->Accum.ClearColor[1]; + const GLfloat b = ctx->Accum.ClearColor[2]; + const GLfloat a = ctx->Accum.ClearColor[3]; + GLfloat *accBuf; + GLint i; + + (void) pipe->region_map(pipe, acc_ps->region); + + accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + for (i = 0; i < width * height; i++) { + accBuf[i * 4 + 0] = r; + accBuf[i * 4 + 1] = g; + accBuf[i * 4 + 2] = b; + accBuf[i * 4 + 3] = a; + } + + acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); + + pipe->region_unmap(pipe, acc_ps->region); +} + + + /** For ADD/MULT */ static void accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, @@ -67,8 +102,7 @@ accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, acc_ps->get_tile(acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { - GLfloat val = accBuf[i] * scale + bias; - accBuf[i] = CLAMP(val, 0.0, 1.0); + accBuf[i] = accBuf[i] * scale + bias; } acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); @@ -99,8 +133,7 @@ accum_accum(struct pipe_context *pipe, GLfloat value, acc_ps->get_tile(acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { - GLfloat val = accBuf[i] + colorBuf[i] * value; - accBuf[i] = CLAMP(val, 0.0, 1.0); + accBuf[i] = accBuf[i] + colorBuf[i] * value; } acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); @@ -130,8 +163,7 @@ accum_load(struct pipe_context *pipe, GLfloat value, color_ps->get_tile(color_ps, xpos, ypos, width, height, buf); for (i = 0; i < 4 * width * height; i++) { - GLfloat val = buf[i] * value; - buf[i] = CLAMP(val, 0.0, 1.0); + buf[i] = buf[i] * value; } acc_ps->put_tile(acc_ps, xpos, ypos, width, height, buf); diff --git a/src/mesa/state_tracker/st_cb_accum.h b/src/mesa/state_tracker/st_cb_accum.h index 8a4101c627..ed9b7dab94 100644 --- a/src/mesa/state_tracker/st_cb_accum.h +++ b/src/mesa/state_tracker/st_cb_accum.h @@ -30,6 +30,9 @@ #define ST_CB_ACCUM_H +extern void +st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb); + extern void st_init_accum_functions(struct dd_function_table *functions); -- cgit v1.2.3 From 45700ac280ddd5e23c57763129257d7fba171d9d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 18:45:42 -0600 Subject: use new st_clear_accum_buffer() function --- src/mesa/state_tracker/st_cb_clear.c | 27 +++------------------------ 1 file changed, 3 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index a854378b18..4f269ed3f7 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -37,6 +37,7 @@ #include "st_atom.h" #include "st_cache.h" #include "st_context.h" +#include "st_cb_accum.h" #include "st_cb_clear.h" #include "st_cb_fbo.h" #include "st_draw.h" @@ -461,28 +462,6 @@ clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) } -static void -clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) -{ - struct st_renderbuffer *strb = st_renderbuffer(rb); - - if (!ctx->Scissor.Enabled) { - /* clear whole buffer w/out masking */ - GLuint clearValue - = color_value(strb->surface->format, ctx->Accum.ClearColor); - /* Note that clearValue is 32 bits but the accum buffer will - * typically be 64bpp... - */ - ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); - } - else { - /* scissoring */ - /* XXX point framebuffer.cbufs[0] at the accum buffer */ - clear_with_quad(ctx, GL_TRUE, GL_FALSE, GL_FALSE); - } -} - - static void clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { @@ -605,8 +584,8 @@ static void st_clear(GLcontext *ctx, GLbitfield mask) } if (mask & BUFFER_BIT_ACCUM) { - clear_accum_buffer(ctx, - ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer); + st_clear_accum_buffer(ctx, + ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer); } if ((mask & BUFFER_BITS_DS) == BUFFER_BITS_DS && depthRb == stencilRb) { -- cgit v1.2.3 From 7db3cf8c48a3fb73cd742dbc02394f10027db9ed Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 16 Oct 2007 18:46:14 -0600 Subject: lower clamp bound for UNCLAMPED_FLOAT_TO_SHORT() is -1 --- src/mesa/pipe/softpipe/sp_surface.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 02a072fe9b..292676a302 100755 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -47,7 +47,7 @@ #define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) #define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ - us = ( (short) ( CLAMP((f), 0.0, 1.0) * 32767.0F) ) + us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) -- cgit v1.2.3 From 863cc0af74f8c02711e09cb78bf2c24d8783daae Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 10:51:55 -0600 Subject: fix square point rasterization --- src/mesa/pipe/softpipe/sp_prim_setup.c | 74 ++++++++++++++++++++-------------- 1 file changed, 43 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 4749f20697..73b18b776b 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -930,9 +930,10 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const interp_mode *interp = setup->softpipe->vertex_info.interp_mode; const struct vertex_header *v0 = prim->v[0]; const int sizeAttr = setup->softpipe->psize_slot; - const float halfSize - = sizeAttr > 0 ? (0.5f * v0->data[sizeAttr][0]) - : (0.5f * setup->softpipe->rasterizer->point_size); + const float size + = sizeAttr > 0 ? v0->data[sizeAttr][0] + : setup->softpipe->rasterizer->point_size; + const float halfSize = 0.5F * size; const boolean round = setup->softpipe->rasterizer->point_smooth; const float x = v0->data[0][0]; /* Note: data[0] is always position */ const float y = v0->data[0][1]; @@ -986,19 +987,18 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) clip_emit_quad(setup); } else { - const int ixmin = block((int) (x - halfSize)); - const int ixmax = block((int) (x + halfSize)); - const int iymin = block((int) (y - halfSize)); - const int iymax = block((int) (y + halfSize)); - int ix, iy; - if (round) { /* rounded points */ + const int ixmin = block((int) (x - halfSize)); + const int ixmax = block((int) (x + halfSize)); + const int iymin = block((int) (y - halfSize)); + const int iymax = block((int) (y + halfSize)); const float rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ const float rmax = halfSize + 0.7071F; const float rmin2 = MAX2(0.0F, rmin * rmin); const float rmax2 = rmax * rmax; const float cscale = 1.0F / (rmax2 - rmin2); + int ix, iy; for (iy = iymin; iy <= iymax; iy += 2) { for (ix = ixmin; ix <= ixmax; ix += 2) { @@ -1052,35 +1052,47 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) } else { /* square points */ + const int xmin = (int) (x + 0.75 - halfSize); + const int ymin = (int) (y + 0.25 - halfSize); + const int xmax = xmin + (int) size; + const int ymax = ymin + (int) size; + /* XXX could apply scissor to xmin,ymin,xmax,ymax now */ + const int ixmin = block(xmin); + const int ixmax = block(xmax - 1); + const int iymin = block(ymin); + const int iymax = block(ymax - 1); + int ix, iy; + + /* + printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax); + */ for (iy = iymin; iy <= iymax; iy += 2) { + uint rowMask = 0xf; + if (iy < ymin) { + /* above the top edge */ + rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + } + if (iy + 1 >= ymax) { + /* below the bottom edge */ + rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + } + for (ix = ixmin; ix <= ixmax; ix += 2) { - setup->quad.mask = 0xf; + uint mask = rowMask; - if (ix + 0.5 < x - halfSize) { + if (ix < xmin) { /* fragment is past left edge of point, turn off left bits */ - setup->quad.mask &= ~(MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); } - - if (ix + 1.5 > x + halfSize) { + if (ix + 1 >= xmax) { /* past the right edge */ - setup->quad.mask &= ~(MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - } - - if (iy + 0.5 < y - halfSize) { - /* below the bottom edge */ - setup->quad.mask &= ~(MASK_TOP_LEFT | MASK_TOP_RIGHT); - } - - if (iy + 1.5 > y + halfSize) { - /* above the top edge */ - setup->quad.mask &= ~(MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - } - - if (setup->quad.mask) { - setup->quad.x0 = ix; - setup->quad.y0 = iy; - clip_emit_quad(setup); + mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); } + + setup->quad.mask = mask; + setup->quad.x0 = ix; + setup->quad.y0 = iy; + clip_emit_quad(setup); } } } -- cgit v1.2.3 From 4f23468bd0d14b8ed687a641003d587b91ad39a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 10:52:30 -0600 Subject: don't crash if texture border is used (unsupported for now) --- src/mesa/pipe/softpipe/sp_tex_sample.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 6ff84119e5..78f00b2bf0 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -799,6 +799,9 @@ sp_get_samples(struct tgsi_sampler *sampler, float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]) { + if (!sampler->texture) + return; + switch (sampler->texture->target) { case PIPE_TEXTURE_1D: sp_get_samples_1d(sampler, s, t, p, lodbias, rgba); -- cgit v1.2.3 From 2dd27cfdd981b3b2c973066082b1168c4cb6f42c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 11:24:06 -0600 Subject: Fix viewport Z scale/bias to get the right Z values from drawing the quad. --- src/mesa/state_tracker/st_cb_clear.c | 6 +++--- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- 2 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 4f269ed3f7..c0ea1a4bf6 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -326,7 +326,7 @@ clear_with_quad(GLcontext *ctx, pipe->bind_depth_stencil_state(pipe, cso->data); } - /* setup state: nothing */ + /* rasterizer state: nothing */ { struct pipe_rasterizer_state raster; const struct cso_rasterizer *cso; @@ -367,11 +367,11 @@ clear_with_quad(GLcontext *ctx, struct pipe_viewport_state vp; vp.scale[0] = 0.5 * width; vp.scale[1] = -0.5 * height; - vp.scale[2] = 0.5; + vp.scale[2] = 1.0; vp.scale[3] = 1.0; vp.translate[0] = 0.5 * width; vp.translate[1] = 0.5 * height; - vp.translate[2] = 0.5; + vp.translate[2] = 0.0; vp.translate[3] = 0.0; pipe->set_viewport_state(pipe, &vp); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 97d4f41ed9..9031ef46df 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -570,11 +570,11 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, struct pipe_viewport_state vp; vp.scale[0] = 0.5 * width; vp.scale[1] = -0.5 * height; - vp.scale[2] = 0.5; + vp.scale[2] = 1.0; vp.scale[3] = 1.0; vp.translate[0] = 0.5 * width; vp.translate[1] = 0.5 * height; - vp.translate[2] = 0.5; + vp.translate[2] = 0.0; vp.translate[3] = 0.0; pipe->set_viewport_state(pipe, &vp); } -- cgit v1.2.3 From 4338f0f709255cc4d69d01b9aca579bca31ac269 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 12:27:40 -0600 Subject: Update framebuffer state in response to _NEW_COLOR (set by glDrawBuffer) --- src/mesa/state_tracker/st_atom_framebuffer.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index ccbeacfd6e..a6caf3f5ff 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -84,10 +84,14 @@ update_framebuffer_state( struct st_context *st ) } +/** + * Note that glDrawBuffer() sets _NEW_COLOR, not _NEW_BUFFER. + */ + const struct st_tracked_state st_update_framebuffer = { .name = "st_update_framebuffer", .dirty = { - .mesa = _NEW_BUFFERS, + .mesa = (_NEW_BUFFERS | _NEW_COLOR), .st = 0, }, .update = update_framebuffer_state -- cgit v1.2.3 From 38743e2ef1091304a7059c04c157fde80bd977ec Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 12:32:21 -0600 Subject: generate selection hit if in selection mode and pos is not clipped --- src/mesa/state_tracker/st_cb_rasterpos.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 13580e400b..653801001a 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -31,6 +31,7 @@ */ #include "main/imports.h" +#include "main/feedback.h" #include "main/macros.h" #include "st_context.h" @@ -250,6 +251,10 @@ update_rasterpos(GLcontext *ctx, } ctx->Current.RasterPosValid = GL_TRUE; + + if (ctx->RenderMode == GL_SELECT) { + _mesa_update_hitflag( ctx, ctx->Current.RasterPos[2] ); + } } -- cgit v1.2.3 From 7392dac9621ae2e87485d24d17845558a165519d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 12:42:29 -0600 Subject: When in GL_FEEDBACK mode, make sure we emit color and texcoord registers. --- src/mesa/tnl/t_vp_build.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index 2c0883d0b0..cbffed5155 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -135,6 +135,11 @@ static struct state_key *make_state_key( GLcontext *ctx ) key->fragprog_inputs_read = fp->Base.InputsRead; + if (ctx->RenderMode == GL_FEEDBACK) { + /* make sure the vertprog emits color and tex0 */ + key->fragprog_inputs_read |= (FRAG_BIT_COL0 | FRAG_BIT_TEX0); + } + key->separate_specular = (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR); -- cgit v1.2.3 From 3d6f9d904f07b7676cc971eb3faf9dd8e7c58e50 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 14:13:21 -0600 Subject: implement 3D texture sampling --- src/mesa/pipe/softpipe/sp_tex_sample.c | 112 ++++++++++++++++++++++++++++++++- 1 file changed, 111 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 78f00b2bf0..b8d7e49a7e 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -697,7 +697,6 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, if (level0 != level1) { /* get texels from second mipmap level and blend */ float rgba2[4][4]; - unsigned c; x0 = x0 / 2; y0 = y0 / 2; x1 = x1 / 2; @@ -759,6 +758,117 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, float rgba[NUM_CHANNELS][QUAD_SIZE]) { /* get/map pipe_surfaces corresponding to 3D tex slices */ + unsigned level0, level1, j, imgFilter; + int width, height, depth; + float levelBlend; + const uint face = 0; + + choose_mipmap_levels(sampler, s, t, p, lodbias, + &level0, &level1, &levelBlend, &imgFilter); + + if (sampler->state->normalized_coords) { + width = sampler->texture->level[level0].width; + height = sampler->texture->level[level0].height; + depth = sampler->texture->level[level0].depth; + } + else { + width = height = depth = 1.0; + } + + assert(width > 0); + assert(height > 0); + assert(depth > 0); + + switch (imgFilter) { + case PIPE_TEX_FILTER_NEAREST: + for (j = 0; j < QUAD_SIZE; j++) { + int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); + int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); + int z = nearest_texcoord(sampler->state->wrap_r, p[j], depth); + get_texel(sampler, face, level0, z, x, y, rgba, j); + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + unsigned c; + x /= 2; + y /= 2; + z /= 2; + get_texel(sampler, face, level1, z, x, y, rgba2, j); + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = LERP(levelBlend, rgba2[c][j], rgba[c][j]); + } + } + } + break; + case PIPE_TEX_FILTER_LINEAR: + for (j = 0; j < QUAD_SIZE; j++) { + float texel0[4][4], texel1[4][4], xw, yw, zw; + int x0, x1, y0, y1, z0, z1, c; + linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &xw); + linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &yw); + linear_texcoord(sampler->state->wrap_r, p[j], depth, &z0, &z1, &zw); + get_texel(sampler, face, level0, z0, x0, y0, texel0, 0); + get_texel(sampler, face, level0, z0, x1, y0, texel0, 1); + get_texel(sampler, face, level0, z0, x0, y1, texel0, 2); + get_texel(sampler, face, level0, z0, x1, y1, texel0, 3); + get_texel(sampler, face, level0, z1, x0, y0, texel1, 0); + get_texel(sampler, face, level0, z1, x1, y0, texel1, 1); + get_texel(sampler, face, level0, z1, x0, y1, texel1, 2); + get_texel(sampler, face, level0, z1, x1, y1, texel1, 3); + + /* 3D lerp */ + for (c = 0; c < 4; c++) { + float ctemp0[4][4], ctemp1[4][4]; + ctemp0[c][j] = lerp_2d(xw, yw, + texel0[c][0], texel0[c][1], + texel0[c][2], texel0[c][3]); + ctemp1[c][j] = lerp_2d(xw, yw, + texel1[c][0], texel1[c][1], + texel1[c][2], texel1[c][3]); + rgba[c][j] = LERP(zw, ctemp0[c][j], ctemp1[c][j]); + } + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + x0 /= 2; + y0 /= 2; + z0 /= 2; + x1 /= 2; + y1 /= 2; + z1 /= 2; + get_texel(sampler, face, level1, z0, x0, y0, texel0, 0); + get_texel(sampler, face, level1, z0, x1, y0, texel0, 1); + get_texel(sampler, face, level1, z0, x0, y1, texel0, 2); + get_texel(sampler, face, level1, z0, x1, y1, texel0, 3); + get_texel(sampler, face, level1, z1, x0, y0, texel1, 0); + get_texel(sampler, face, level1, z1, x1, y0, texel1, 1); + get_texel(sampler, face, level1, z1, x0, y1, texel1, 2); + get_texel(sampler, face, level1, z1, x1, y1, texel1, 3); + + /* 3D lerp */ + for (c = 0; c < 4; c++) { + float ctemp0[4][4], ctemp1[4][4]; + ctemp0[c][j] = lerp_2d(xw, yw, + texel0[c][0], texel0[c][1], + texel0[c][2], texel0[c][3]); + ctemp1[c][j] = lerp_2d(xw, yw, + texel1[c][0], texel1[c][1], + texel1[c][2], texel1[c][3]); + rgba2[c][j] = LERP(zw, ctemp0[c][j], ctemp1[c][j]); + } + + /* blend mipmap levels */ + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); + } + } + } + break; + default: + assert(0); + } } -- cgit v1.2.3 From df4410a59784482fcbd48f82788dd0a9f5a62c15 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 14:16:42 -0600 Subject: reorder params to get_texel() --- src/mesa/pipe/softpipe/sp_tex_sample.c | 63 +++++++++++++++++----------------- 1 file changed, 32 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index b8d7e49a7e..c5f4602d2b 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -556,15 +556,15 @@ compute_cache_pos(unsigned face, unsigned level, unsigned zslice, * * \param face the cube face in 0..5 * \param level the mipmap level - * \param zslice which slice of a 3D texture * \param x the x coord of texel within 2D image * \param y the y coord of texel within 2D image + * \param zslice which slice of a 3D texture * \param rgba the quad to put the texel/color into * \param j which element of the rgba quad to write to */ static void get_texel(struct tgsi_sampler *sampler, - unsigned face, unsigned level, unsigned zslice, int x, int y, + unsigned face, unsigned level, int x, int y, unsigned zslice, float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j) { int tx = x / TEX_CACHE_TILE_SIZE; @@ -665,7 +665,7 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, for (j = 0; j < QUAD_SIZE; j++) { int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); - get_texel(sampler, faces[j], level0, 0, x, y, rgba, j); + get_texel(sampler, faces[j], level0, x, y, 0, rgba, j); if (level0 != level1) { /* get texels from second mipmap level and blend */ @@ -673,7 +673,7 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, unsigned c; x = x / 2; y = y / 2; - get_texel(sampler, faces[j], level1, 0, x, y, rgba2, j); + get_texel(sampler, faces[j], level1, x, y, 0, rgba2, j); for (c = 0; c < NUM_CHANNELS; c++) { rgba[c][j] = LERP(levelBlend, rgba2[c][j], rgba[c][j]); } @@ -686,10 +686,10 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, int x0, y0, x1, y1, c; linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &a); linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &b); - get_texel(sampler, faces[j], level0, 0, x0, y0, tx, 0); - get_texel(sampler, faces[j], level0, 0, x1, y0, tx, 1); - get_texel(sampler, faces[j], level0, 0, x0, y1, tx, 2); - get_texel(sampler, faces[j], level0, 0, x1, y1, tx, 3); + get_texel(sampler, faces[j], level0, x0, y0, 0, tx, 0); + get_texel(sampler, faces[j], level0, x1, y0, 0, tx, 1); + get_texel(sampler, faces[j], level0, x0, y1, 0, tx, 2); + get_texel(sampler, faces[j], level0, x1, y1, 0, tx, 3); for (c = 0; c < 4; c++) { rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); } @@ -701,10 +701,10 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, y0 = y0 / 2; x1 = x1 / 2; y1 = y1 / 2; - get_texel(sampler, faces[j], level1, 0, x0, y0, tx, 0); - get_texel(sampler, faces[j], level1, 0, x1, y0, tx, 1); - get_texel(sampler, faces[j], level1, 0, x0, y1, tx, 2); - get_texel(sampler, faces[j], level1, 0, x1, y1, tx, 3); + get_texel(sampler, faces[j], level1, x0, y0, 0, tx, 0); + get_texel(sampler, faces[j], level1, x1, y0, 0, tx, 1); + get_texel(sampler, faces[j], level1, x0, y1, 0, tx, 2); + get_texel(sampler, faces[j], level1, x1, y1, 0, tx, 3); for (c = 0; c < 4; c++) { rgba2[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); @@ -785,7 +785,7 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); int z = nearest_texcoord(sampler->state->wrap_r, p[j], depth); - get_texel(sampler, face, level0, z, x, y, rgba, j); + get_texel(sampler, face, level0, x, y, z, rgba, j); if (level0 != level1) { /* get texels from second mipmap level and blend */ @@ -794,7 +794,7 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, x /= 2; y /= 2; z /= 2; - get_texel(sampler, face, level1, z, x, y, rgba2, j); + get_texel(sampler, face, level1, x, y, z, rgba2, j); for (c = 0; c < NUM_CHANNELS; c++) { rgba[c][j] = LERP(levelBlend, rgba2[c][j], rgba[c][j]); } @@ -803,19 +803,20 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, break; case PIPE_TEX_FILTER_LINEAR: for (j = 0; j < QUAD_SIZE; j++) { - float texel0[4][4], texel1[4][4], xw, yw, zw; + float texel0[4][4], texel1[4][4]; + float xw, yw, zw; /* interpolation weights */ int x0, x1, y0, y1, z0, z1, c; linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &xw); linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &yw); linear_texcoord(sampler->state->wrap_r, p[j], depth, &z0, &z1, &zw); - get_texel(sampler, face, level0, z0, x0, y0, texel0, 0); - get_texel(sampler, face, level0, z0, x1, y0, texel0, 1); - get_texel(sampler, face, level0, z0, x0, y1, texel0, 2); - get_texel(sampler, face, level0, z0, x1, y1, texel0, 3); - get_texel(sampler, face, level0, z1, x0, y0, texel1, 0); - get_texel(sampler, face, level0, z1, x1, y0, texel1, 1); - get_texel(sampler, face, level0, z1, x0, y1, texel1, 2); - get_texel(sampler, face, level0, z1, x1, y1, texel1, 3); + get_texel(sampler, face, level0, x0, y0, z0, texel0, 0); + get_texel(sampler, face, level0, x1, y0, z0, texel0, 1); + get_texel(sampler, face, level0, x0, y1, z0, texel0, 2); + get_texel(sampler, face, level0, x1, y1, z0, texel0, 3); + get_texel(sampler, face, level0, x0, y0, z1, texel1, 0); + get_texel(sampler, face, level0, x1, y0, z1, texel1, 1); + get_texel(sampler, face, level0, x0, y1, z1, texel1, 2); + get_texel(sampler, face, level0, x1, y1, z1, texel1, 3); /* 3D lerp */ for (c = 0; c < 4; c++) { @@ -838,14 +839,14 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, x1 /= 2; y1 /= 2; z1 /= 2; - get_texel(sampler, face, level1, z0, x0, y0, texel0, 0); - get_texel(sampler, face, level1, z0, x1, y0, texel0, 1); - get_texel(sampler, face, level1, z0, x0, y1, texel0, 2); - get_texel(sampler, face, level1, z0, x1, y1, texel0, 3); - get_texel(sampler, face, level1, z1, x0, y0, texel1, 0); - get_texel(sampler, face, level1, z1, x1, y0, texel1, 1); - get_texel(sampler, face, level1, z1, x0, y1, texel1, 2); - get_texel(sampler, face, level1, z1, x1, y1, texel1, 3); + get_texel(sampler, face, level1, x0, y0, z0, texel0, 0); + get_texel(sampler, face, level1, x1, y0, z0, texel0, 1); + get_texel(sampler, face, level1, x0, y1, z0, texel0, 2); + get_texel(sampler, face, level1, x1, y1, z0, texel0, 3); + get_texel(sampler, face, level1, x0, y0, z1, texel1, 0); + get_texel(sampler, face, level1, x1, y0, z1, texel1, 1); + get_texel(sampler, face, level1, x0, y1, z1, texel1, 2); + get_texel(sampler, face, level1, x1, y1, z1, texel1, 3); /* 3D lerp */ for (c = 0; c < 4; c++) { -- cgit v1.2.3 From 906768316d9521a32d9a7eebc9edaf76c06a98a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 14:29:12 -0600 Subject: Replace repeat_remainder() with simpler macro that just casts args to unsigned. --- src/mesa/pipe/softpipe/sp_tex_sample.c | 19 ++++++------------- 1 file changed, 6 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index c5f4602d2b..7add74e98a 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -77,17 +77,10 @@ lerp_2d(float a, float b, /** - * Compute the remainder of a divided by b, but be careful with - * negative values so that REPEAT mode works right. + * If A is a signed integer, A % B doesn't give the right value for A < 0 + * (in terms of texture repeat). Just casting to unsigned fixes that. */ -static INLINE int -repeat_remainder(int a, int b) -{ - if (a >= 0) - return a % b; - else - return (a + 1) % b + b - 1; -} +#define REMAINDER(A, B) ((unsigned) (A) % (unsigned) (B)) /** @@ -106,7 +99,7 @@ nearest_texcoord(unsigned wrapMode, float s, unsigned size) /* s limited to [0,1) */ /* i limited to [0,size-1] */ i = ifloor(s * size); - i = repeat_remainder(i, size); + i = REMAINDER(i, size); return i; case PIPE_TEX_WRAP_CLAMP: /* s limited to [0,1] */ @@ -231,8 +224,8 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, switch (wrapMode) { case PIPE_TEX_WRAP_REPEAT: u = s * size - 0.5F; - *i0 = repeat_remainder(ifloor(u), size); - *i1 = repeat_remainder(*i0 + 1, size); + *i0 = REMAINDER(ifloor(u), size); + *i1 = REMAINDER(*i0 + 1, size); break; case PIPE_TEX_WRAP_CLAMP: if (s <= 0.0F) -- cgit v1.2.3 From e48ea925105fb74312f6e2f608b909f897a834e8 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 14:56:55 -0600 Subject: fix addressing error in read_stencil_pixels() --- src/mesa/state_tracker/st_cb_readpixels.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 3fcce2463c..8c9c47a535 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -87,13 +87,13 @@ read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, switch (ps->format) { case PIPE_FORMAT_U_S8: { - const ubyte *src = stmap + y * ps->region->pitch + x; + const ubyte *src = stmap + srcY * ps->region->pitch + x; memcpy(values, src, width); } break; case PIPE_FORMAT_S8_Z24: { - const uint *src = (const uint *) stmap + y * ps->region->pitch + x; + const uint *src = (uint *) stmap + srcY * ps->region->pitch + x; GLint k; for (k = 0; k < width; k++) { values[k] = src[k] >> 24; -- cgit v1.2.3 From 5cb0d749b07be4ecbf4d4d7cde18aecdfe2dba38 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 14:57:15 -0600 Subject: formatting fix --- src/mesa/state_tracker/st_cb_texture.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index eea47b9b41..1621b56336 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1500,7 +1500,9 @@ st_finalize_mipmap_tree(GLcontext *ctx, comp_byte = compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); cpp = comp_byte; } - else cpp = firstImage->base.TexFormat->TexelBytes; + else { + cpp = firstImage->base.TexFormat->TexelBytes; + } /* Check tree can hold all active levels. Check tree matches * target, imageFormat, etc. -- cgit v1.2.3 From 420c062cb68e4e313ef34bd879852a003101169e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 15:59:59 -0600 Subject: Assert that there's no outstanding current state that needs to be flushed from the vbo in _mesa_load_state_parameters(). Several of the state parameters use current state (such as materials). Need to make sure those values are not stale. --- src/mesa/shader/prog_statevars.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index d37d7fb9bf..9a745a77e5 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -815,6 +815,8 @@ _mesa_load_state_parameters(GLcontext *ctx, if (!paramList) return; + assert(ctx->Driver.NeedFlush == 0); + for (i = 0; i < paramList->NumParameters; i++) { if (paramList->Parameters[i].Type == PROGRAM_STATE_VAR) { _mesa_fetch_state(ctx, -- cgit v1.2.3 From 49adf51eeec31c9f3c995a70acc5008522689708 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 16:18:45 -0600 Subject: Add some FLUSH_CURRENT() calls. Without these we can find ourselves in _mesa_load_state_parameters() computing derived lighting/material values whhen the current material properties haven't been updated from the VBO. This may be a somewhat wide-spread problem that needs more attention... --- src/mesa/main/buffers.c | 2 ++ src/mesa/main/drawpix.c | 2 ++ 2 files changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index 0e6ca8ea1c..7a918c53a7 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -118,6 +118,8 @@ _mesa_Clear( GLbitfield mask ) GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + FLUSH_CURRENT(ctx, 0); + if (MESA_VERBOSE & VERBOSE_API) _mesa_debug(ctx, "glClear 0x%x\n", mask); diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c index 296dcf8f0b..e4be0d496e 100644 --- a/src/mesa/main/drawpix.c +++ b/src/mesa/main/drawpix.c @@ -276,6 +276,8 @@ _mesa_ReadPixels( GLint x, GLint y, GLsizei width, GLsizei height, GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + FLUSH_CURRENT(ctx, 0); + if (width < 0 || height < 0) { _mesa_error( ctx, GL_INVALID_VALUE, "glReadPixels(width=%d height=%d)", width, height ); -- cgit v1.2.3 From 52111366cc34e367993b5ec785fe331891a3da5e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 16:19:10 -0600 Subject: disable the NeedFlush assertion for now --- src/mesa/shader/prog_statevars.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 9a745a77e5..1bb7142c60 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -815,7 +815,7 @@ _mesa_load_state_parameters(GLcontext *ctx, if (!paramList) return; - assert(ctx->Driver.NeedFlush == 0); + /*assert(ctx->Driver.NeedFlush == 0);*/ for (i = 0; i < paramList->NumParameters; i++) { if (paramList->Parameters[i].Type == PROGRAM_STATE_VAR) { -- cgit v1.2.3 From f953c223df26293f955f7d0621a6f917e9cc9768 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 16:23:25 -0600 Subject: remove #include vf.h --- src/mesa/state_tracker/st_cb_clear.c | 2 -- src/mesa/state_tracker/st_cb_drawpixels.c | 1 - src/mesa/state_tracker/st_cb_feedback.c | 1 - src/mesa/state_tracker/st_cb_rasterpos.c | 1 - 4 files changed, 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index c0ea1a4bf6..2e7d9f1a30 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -51,8 +51,6 @@ #include "pipe/tgsi/mesa/mesa_to_tgsi.h" -#include "vf/vf.h" - diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 9031ef46df..bb8a083883 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -47,7 +47,6 @@ #include "pipe/p_defines.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "shader/prog_instruction.h" -#include "vf/vf.h" diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 923e1cdb89..a9fd2579a2 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -55,7 +55,6 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "pipe/cso_cache/cso_cache.h" -#include "vf/vf.h" #include "pipe/draw/draw_context.h" #include "pipe/draw/draw_private.h" diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 653801001a..eb6ee51939 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -46,7 +46,6 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "shader/prog_instruction.h" -#include "vf/vf.h" -- cgit v1.2.3 From 17c2f56dc3f2f58ba89d8e305e7d9b423e3cae16 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 17 Oct 2007 17:22:06 -0600 Subject: fix interpolation bug in nearest-image/linear-mipmap filtering --- src/mesa/pipe/softpipe/sp_tex_sample.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 7add74e98a..64cb94d944 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -668,7 +668,7 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, y = y / 2; get_texel(sampler, faces[j], level1, x, y, 0, rgba2, j); for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(levelBlend, rgba2[c][j], rgba[c][j]); + rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); } } } -- cgit v1.2.3 From 7cd58433fa59fe9db38b515a48feeef23405630e Mon Sep 17 00:00:00 2001 From: Oliver McFadden Date: Thu, 18 Oct 2007 14:45:32 +0000 Subject: sp_context.c:255: error: 'false' undeclared (first use in this function) --- src/mesa/pipe/softpipe/sp_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index daccc8fe23..f2f2e36869 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -252,7 +252,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, #if defined(__i386__) || defined(__386__) softpipe->use_sse = getenv("GALLIUM_SSE") != NULL; #else - softpipe->use_sse = false; + softpipe->use_sse = FALSE; #endif softpipe->pipe.winsys = pipe_winsys; -- cgit v1.2.3 From d16b844ff64082b62ad481f543fcef587a1c1612 Mon Sep 17 00:00:00 2001 From: Oliver McFadden Date: Thu, 18 Oct 2007 16:01:42 +0000 Subject: pipe/draw/draw_context.c:47: error: 'false' undeclared (first use in this function) --- src/mesa/pipe/draw/draw_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index e252148903..d2015bd514 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -44,7 +44,7 @@ struct draw_context *draw_create( void ) #if defined(__i386__) || defined(__386__) draw->use_sse = getenv("GALLIUM_SSE") != NULL; #else - draw->use_sse = false; + draw->use_sse = FALSE; #endif /* create pipeline stages */ -- cgit v1.2.3 From a22fafbb4b6b772c531651ada53661749cd1f29e Mon Sep 17 00:00:00 2001 From: Oliver McFadden Date: Thu, 18 Oct 2007 16:36:04 +0000 Subject: Corrected the file permissions in src/mesa/pipe. --- src/mesa/pipe/p_util.h | 0 src/mesa/pipe/softpipe/sp_quad_blend.c | 0 src/mesa/pipe/softpipe/sp_quad_fs.c | 0 src/mesa/pipe/softpipe/sp_surface.c | 0 src/mesa/pipe/softpipe/sp_tex_layout.c | 0 src/mesa/pipe/tgsi/exec/tgsi_build.c | 0 src/mesa/pipe/tgsi/exec/tgsi_build.h | 0 src/mesa/pipe/tgsi/exec/tgsi_dump.c | 0 src/mesa/pipe/tgsi/exec/tgsi_parse.c | 0 src/mesa/pipe/tgsi/exec/tgsi_parse.h | 0 src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 0 src/mesa/pipe/tgsi/exec/tgsi_sse2.h | 0 12 files changed, 0 insertions(+), 0 deletions(-) mode change 100755 => 100644 src/mesa/pipe/p_util.h mode change 100755 => 100644 src/mesa/pipe/softpipe/sp_quad_blend.c mode change 100755 => 100644 src/mesa/pipe/softpipe/sp_quad_fs.c mode change 100755 => 100644 src/mesa/pipe/softpipe/sp_surface.c mode change 100755 => 100644 src/mesa/pipe/softpipe/sp_tex_layout.c mode change 100755 => 100644 src/mesa/pipe/tgsi/exec/tgsi_build.c mode change 100755 => 100644 src/mesa/pipe/tgsi/exec/tgsi_build.h mode change 100755 => 100644 src/mesa/pipe/tgsi/exec/tgsi_dump.c mode change 100755 => 100644 src/mesa/pipe/tgsi/exec/tgsi_parse.c mode change 100755 => 100644 src/mesa/pipe/tgsi/exec/tgsi_parse.h mode change 100755 => 100644 src/mesa/pipe/tgsi/exec/tgsi_sse2.c mode change 100755 => 100644 src/mesa/pipe/tgsi/exec/tgsi_sse2.h (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.c b/src/mesa/pipe/tgsi/exec/tgsi_build.c old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.h b/src/mesa/pipe/tgsi/exec/tgsi_build.h old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/tgsi/exec/tgsi_parse.c b/src/mesa/pipe/tgsi/exec/tgsi_parse.c old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/tgsi/exec/tgsi_parse.h b/src/mesa/pipe/tgsi/exec/tgsi_parse.h old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c old mode 100755 new mode 100644 diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h old mode 100755 new mode 100644 -- cgit v1.2.3 From 376fb1c23efd437109da88cd2e53fca9d1c77bf2 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 10:09:01 -0600 Subject: fix logic error in stipple_quad() (point/line fragments were getting dropped) --- src/mesa/pipe/softpipe/sp_quad_stipple.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index b2658ff158..4af5059d67 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -50,10 +50,11 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) quad->mask &= (((stipple0 >> col0) & 0x3) | (((stipple1 >> col0) & 0x3) << 2)); #endif - - if (quad->mask) - qs->next->run(qs->next, quad); + if (!quad->mask) + return; } + + qs->next->run(qs->next, quad); } -- cgit v1.2.3 From 958c92e0c74f8cd4b303e7a7485a1c8b8cfde2cb Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 10:16:28 -0600 Subject: debug functions --- src/mesa/state_tracker/st_debug.c | 60 +++++++++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_debug.h | 36 +++++++++++++++++++++++ 2 files changed, 96 insertions(+) create mode 100644 src/mesa/state_tracker/st_debug.c create mode 100644 src/mesa/state_tracker/st_debug.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c new file mode 100644 index 0000000000..a88888b17c --- /dev/null +++ b/src/mesa/state_tracker/st_debug.c @@ -0,0 +1,60 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "main/context.h" +#include "shader/prog_print.h" + +#include "pipe/p_state.h" +#include "pipe/tgsi/exec/tgsi_core.h" + +#include "pipe/cso_cache/cso_cache.h" + +#include "st_context.h" +#include "st_debug.h" +#include "st_program.h" + + + +/** + * Print current state. May be called from inside gdb to see currently + * bound vertex/fragment shaders and associated constants. + */ +void +st_print_current(void) +{ + GET_CURRENT_CONTEXT(ctx); + struct st_context *st = ctx->st; + + tgsi_dump( st->state.vs->state.tokens, 0 ); + if (st->vp->Base.Base.Parameters) + _mesa_print_parameter_list(st->vp->Base.Base.Parameters); + + tgsi_dump( st->state.fs->state.tokens, 0 ); + if (st->fp->Base.Base.Parameters) + _mesa_print_parameter_list(st->fp->Base.Base.Parameters); +} diff --git a/src/mesa/state_tracker/st_debug.h b/src/mesa/state_tracker/st_debug.h new file mode 100644 index 0000000000..49d752e1b2 --- /dev/null +++ b/src/mesa/state_tracker/st_debug.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_DEBUG_H +#define ST_DEBUG_H + +extern void +st_print_current(void); + + +#endif /* ST_DEBUG_H */ -- cgit v1.2.3 From 66586762f1d2a3a310cf83e177d3c659a93486c8 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 10:16:40 -0600 Subject: added st_debug.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index cf307401b2..d53e75c820 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -223,6 +223,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_texture.c \ state_tracker/st_cache.c \ state_tracker/st_context.c \ + state_tracker/st_debug.c \ state_tracker/st_draw.c \ state_tracker/st_format.c \ state_tracker/st_program.c \ -- cgit v1.2.3 From 5bd119f9438b680f5e42458ef0b250662af36235 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 10:42:47 -0600 Subject: handle fogcoord/raster distance --- src/mesa/state_tracker/st_cb_rasterpos.c | 19 +++++++++++++++---- 1 file changed, 15 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index eb6ee51939..40807fc05a 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -187,6 +187,7 @@ update_rasterpos(GLcontext *ctx, const float clipPos[4], const float color0[4], const float color1[4], + const float *fog, const float *tex) { uint i; @@ -232,15 +233,17 @@ update_rasterpos(GLcontext *ctx, ctx->Current.RasterPos[3] = clipPos[3]; /* compute raster distance */ +#if 0 if (ctx->Fog.FogCoordinateSource == GL_FOG_COORDINATE_EXT) ctx->Current.RasterDistance = ctx->Current.Attrib[VERT_ATTRIB_FOG][0]; else { -#if 0 /* XXX we don't have an eye coord! */ ctx->Current.RasterDistance = SQRTF( eye[0]*eye[0] + eye[1]*eye[1] + eye[2]*eye[2] ); -#endif } +#else + ctx->Current.RasterDistance = fog[0]; +#endif /* colors and texcoords */ COPY_4FV(ctx->Current.RasterColor, color0); @@ -314,7 +317,7 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) /* extract values and update rasterpos state */ { const GLuint *outputMapping = st->vertex_result_to_slot; - const float *pos, *color0, *color1, *tex0; + const float *pos, *color0, *color1, *fog, *tex0; float *buf = buf_map; assert(outputMapping[VERT_RESULT_HPOS] != ~0); @@ -337,6 +340,14 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) color1 = ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; } + if (outputMapping[VERT_RESULT_FOGC] != ~0) { + fog = buf; + buf += 4; + } + else { + fog = ctx->Current.Attrib[VERT_ATTRIB_FOG]; + } + if (outputMapping[VERT_RESULT_TEX0] != ~0) { tex0 = buf; buf += 4; @@ -345,7 +356,7 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) tex0 = ctx->Current.Attrib[VERT_ATTRIB_TEX0]; } - update_rasterpos(ctx, pos, color0, color1, tex0); + update_rasterpos(ctx, pos, color0, color1, fog, tex0); } /* free vertex feedback buffer */ -- cgit v1.2.3 From 54c5262eaff45bea8fa01305fa423341998fdc4e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 10:55:26 -0600 Subject: print vertex input mapping --- src/mesa/state_tracker/st_debug.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index a88888b17c..e00684ac03 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -49,6 +49,12 @@ st_print_current(void) { GET_CURRENT_CONTEXT(ctx); struct st_context *st = ctx->st; + int i; + + printf("Vertex Transform Inputs:\n"); + for (i = 0; i < st->state.vs->state.num_inputs; i++) { + printf(" Slot %d: VERT_ATTRIB_%d\n", i, st->vp->index_to_input[i]); + } tgsi_dump( st->state.vs->state.tokens, 0 ); if (st->vp->Base.Base.Parameters) -- cgit v1.2.3 From 2ae9f53c9d56ca1898d30cbd359c1fd31027ae6c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 11:03:39 -0600 Subject: better debug code, silence some warnings --- src/mesa/pipe/draw/draw_vertex_shader.c | 19 +++++++++++++------ 1 file changed, 13 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index ea3fc2128d..8519daf17c 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -40,8 +40,12 @@ #include "pipe/tgsi/exec/tgsi_core.h" + +#define DBG 0 + + static INLINE unsigned -compute_clipmask(const float *clip, const float (*plane)[4], unsigned nr) +compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) { unsigned mask = 0; unsigned i; @@ -142,6 +146,9 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][2] = z * scale[2] + trans[2]; vOut[j]->data[0][3] = w; +#if DBG + printf("output[%d]win: %f %f %f %f\n", x, y, z, w); +#endif /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. * Skip 0 since we just did it above. @@ -152,13 +159,13 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; - /* - printf("output %d: %f %f %f %f\n", slot, +#if DBG + printf("output[%d][%d]: %f %f %f %f\n", j, slot, vOut[j]->data[slot][0], vOut[j]->data[slot][1], vOut[j]->data[slot][2], vOut[j]->data[slot][3]); - */ +#endif } } /* loop over vertices */ } @@ -236,10 +243,10 @@ void draw_delete_vertex_shader(struct draw_context *draw, struct draw_vertex_shader *vs = (struct draw_vertex_shader*)(vcso); #if defined(__i386__) || defined(__386__) - x86_release_func(&vs->state->sse2_program); + x86_release_func((struct x86_function *) &vs->state->sse2_program); #endif - free(vs->state); + free((void *) vs->state); free(vcso); } -- cgit v1.2.3 From b21f4ac15fc190b474e1686d50a6992caed9c388 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 11:05:03 -0600 Subject: added const qualifiers --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 4 ++-- src/mesa/pipe/tgsi/exec/tgsi_sse2.h | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) mode change 100644 => 100755 src/mesa/pipe/tgsi/exec/tgsi_sse2.c mode change 100644 => 100755 src/mesa/pipe/tgsi/exec/tgsi_sse2.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c old mode 100644 new mode 100755 index d840d64cae..d0d5719c93 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -1824,7 +1824,7 @@ emit_declaration( unsigned tgsi_emit_sse2( - struct tgsi_token *tokens, + const struct tgsi_token *tokens, struct x86_function *func ) { struct tgsi_parse_context parse; @@ -1889,7 +1889,7 @@ tgsi_emit_sse2( */ unsigned tgsi_emit_sse2_fs( - struct tgsi_token *tokens, + const struct tgsi_token *tokens, struct x86_function *func ) { struct tgsi_parse_context parse; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h old mode 100644 new mode 100755 index 9bee371766..60fdd5f61f --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h @@ -10,12 +10,12 @@ struct x86_function; unsigned tgsi_emit_sse2( - struct tgsi_token *tokens, + const struct tgsi_token *tokens, struct x86_function *function ); unsigned tgsi_emit_sse2_fs( - struct tgsi_token *tokens, + const struct tgsi_token *tokens, struct x86_function *function ); #if defined __cplusplus -- cgit v1.2.3 From 447c93dd1b3bb31fe19fa81b32d1c3c62ab65c54 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 11:09:12 -0600 Subject: Change type of shader->executable field from void * to generic function pointer. Fix warnings in draw_create_vertex_shader() --- src/mesa/pipe/draw/draw_vertex_shader.c | 10 ++++++---- src/mesa/pipe/p_state.h | 2 +- 2 files changed, 7 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 8519daf17c..fd0183f8cd 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -212,12 +212,14 @@ draw_create_vertex_shader(struct draw_context *draw, vs->state = shader; #if defined(__i386__) || defined(__386__) if (draw->use_sse) { - x86_init_func( &shader->sse2_program ); + /* cast-away const */ + struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; - tgsi_emit_sse2( shader->tokens, &shader->sse2_program ); + x86_init_func( &sh->sse2_program ); - ((struct pipe_shader_state *)shader)->executable = (void *) - x86_get_func( &shader->sse2_program ); + tgsi_emit_sse2( sh->tokens, &sh->sse2_program ); + + sh->executable = x86_get_func( &sh->sse2_program ); } #endif diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index da62aa1b27..1c3b2811a4 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -148,7 +148,7 @@ struct pipe_shader_state { #if defined(__i386__) || defined(__386__) struct x86_function sse2_program; #endif - void *executable; + void (*executable)(); ubyte num_inputs; ubyte num_outputs; -- cgit v1.2.3 From 8f3fb395ff9bec371942e0769763daa4c59f876b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 12:31:05 -0600 Subject: make read_stencil_pixels() non-static --- src/mesa/state_tracker/st_cb_readpixels.c | 10 +++++----- src/mesa/state_tracker/st_cb_readpixels.h | 6 ++++++ 2 files changed, 11 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 8c9c47a535..7bce7e9a27 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -50,11 +50,11 @@ * Special case for reading stencil buffer. * For color/depth we use get_tile(). For stencil, map the stencil buffer. */ -static void -read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, - GLsizei width, GLsizei height, GLenum type, - const struct gl_pixelstore_attrib *packing, - GLvoid *pixels) +void +st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, + GLsizei width, GLsizei height, GLenum type, + const struct gl_pixelstore_attrib *packing, + GLvoid *pixels) { struct st_context *st = ctx->st; struct pipe_context *pipe = st->pipe; diff --git a/src/mesa/state_tracker/st_cb_readpixels.h b/src/mesa/state_tracker/st_cb_readpixels.h index 1dbe9727a5..79acdad88e 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.h +++ b/src/mesa/state_tracker/st_cb_readpixels.h @@ -29,6 +29,12 @@ #ifndef ST_CB_READPIXELS_H #define ST_CB_READPIXELS_H +extern void +st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, + GLsizei width, GLsizei height, GLenum type, + const struct gl_pixelstore_attrib *packing, + GLvoid *pixels); + extern void st_init_readpixels_functions(struct dd_function_table *functions); -- cgit v1.2.3 From bdc574c5bd3cf1a493d70863436b773d0a8a73a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 12:31:35 -0600 Subject: start implementing glCopyPixels (stencil works) --- src/mesa/state_tracker/st_cb_drawpixels.c | 79 +++++++++++++++++++++++++++++++ 1 file changed, 79 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index bb8a083883..03fc1340e7 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -40,6 +40,8 @@ #include "st_draw.h" #include "st_program.h" #include "st_cb_drawpixels.h" +#include "st_cb_readpixels.h" +#include "st_cb_fbo.h" #include "st_cb_texture.h" #include "st_draw.h" #include "st_format.h" @@ -1058,6 +1060,78 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } +static void +copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, + GLsizei width, GLsizei height, + GLint dstx, GLint dsty) +{ + struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; + struct st_renderbuffer *rbRead = st_renderbuffer(ctx->ReadBuffer->_StencilBuffer); + struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer); + struct pipe_surface *psRead = rbRead->surface; + struct pipe_surface *psDraw = rbDraw->surface; + ubyte *readMap, *drawMap; + ubyte *buffer; + int i; + + buffer = malloc(width * height * sizeof(ubyte)); + if (!buffer) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glCopyPixels(stencil)"); + return; + } + + /* map the stencil buffers */ + readMap = pipe->region_map(pipe, psRead->region); + drawMap = pipe->region_map(pipe, psDraw->region); + + /* this will do stencil pixel transfer ops */ + st_read_stencil_pixels(ctx, srcx, srcy, width, height, GL_UNSIGNED_BYTE, + &ctx->DefaultPacking, buffer); + + /* draw */ + /* XXX PixelZoom not handled yet */ + for (i = 0; i < height; i++) { + ubyte *dst; + const ubyte *src; + int y; + + y = dsty + i; + + if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { + y = ctx->DrawBuffer->Height - y - 1; + } + + dst = drawMap + (y * psDraw->region->pitch + dstx) * psDraw->region->cpp; + src = buffer + i * width; + + switch (psDraw->format) { + case PIPE_FORMAT_S8_Z24: + { + uint *dst4 = (uint *) dst; + int j; + for (j = 0; j < width; j++) { + *dst4 = (*dst4 & 0xffffff) | (src[j] << 24); + dst4++; + } + } + break; + case PIPE_FORMAT_U_S8: + memcpy(dst, src, width); + break; + default: + assert(0); + } + } + + free(buffer); + + /* unmap the stencil buffers */ + pipe->region_unmap(pipe, psRead->region); + pipe->region_unmap(pipe, psDraw->region); +} + + static void st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, GLsizei width, GLsizei height, @@ -1067,6 +1141,11 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, st_validate_state(st); + if (type == GL_STENCIL) { + copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty); + return; + } + /* allocate a texture of size width x height */ /* blit/copy framebuffer region into texture */ -- cgit v1.2.3 From 563584a4ee853c45b7a4b60c68ac8dea4d92942d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 13:27:10 -0600 Subject: checkpoint: code refactoring for glCopyPixels --- src/mesa/state_tracker/st_cb_drawpixels.c | 173 +++++++++++++++++++++--------- 1 file changed, 123 insertions(+), 50 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 03fc1340e7..858ce176fe 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -55,15 +55,21 @@ /** * Create a simple fragment shader that does a TEX() instruction to get * the fragment color. + * If bitmapMode, use KIL instruction to kill the "0-pixels". */ static struct st_fragment_program * make_fragment_shader(struct st_context *st, GLboolean bitmapMode) { + /* only make programs once and re-use */ + static struct st_fragment_program *progs[2] = { NULL, NULL }; GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; struct gl_program *p; GLuint ic = 0; + if (progs[bitmapMode]) + return progs[bitmapMode]; + p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) return NULL; @@ -159,6 +165,8 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) st_translate_fragment_program(st, stfp, NULL, stfp->tokens, ST_MAX_SHADER_TOKENS); + progs[bitmapMode] = stfp; + return stfp; } @@ -166,15 +174,20 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) /** * Create fragment shader that does a TEX() instruction to get a Z * value, then writes to FRAG_RESULT_DEPR. + * Pass fragment color through as-is. */ static struct st_fragment_program * make_fragment_shader_z(struct st_context *st) { GLcontext *ctx = st->ctx; - struct st_fragment_program *stfp; + /* only make programs once and re-use */ + static struct st_fragment_program *stfp = NULL; struct gl_program *p; GLuint ic = 0; + if (stfp) + return stfp; + p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) return NULL; @@ -226,16 +239,21 @@ make_fragment_shader_z(struct st_context *st) /** * Create a simple vertex shader that just passes through the - * vertex position and texcoord (and color). + * vertex position and texcoord (and optionally, color). */ static struct st_vertex_program * make_vertex_shader(struct st_context *st, GLboolean passColor) { + /* only make programs once and re-use */ + static struct st_vertex_program *progs[2] = { NULL, NULL }; GLcontext *ctx = st->ctx; struct st_vertex_program *stvp; struct gl_program *p; GLuint ic = 0; + if (progs[passColor]) + return progs[passColor]; + p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); if (!p) return NULL; @@ -292,6 +310,8 @@ make_vertex_shader(struct st_context *st, GLboolean passColor) st_translate_vertex_program(st, stvp, NULL, stvp->tokens, ST_MAX_SHADER_TOKENS); + progs[passColor] = stvp; + return stvp; } @@ -311,8 +331,51 @@ _mesa_base_format(GLenum format) +static struct pipe_mipmap_tree * +alloc_mipmap_tree(struct st_context *st, + GLsizei width, GLsizei height, uint pipeFormat) +{ + const GLbitfield flags = PIPE_SURFACE_FLAG_TEXTURE; + struct pipe_mipmap_tree *mt; + GLuint cpp; + + mt = CALLOC_STRUCT(pipe_mipmap_tree); + if (!mt) + return NULL; + + cpp = st_sizeof_format(pipeFormat); + + /* allocate texture region/storage */ + mt->region = st->pipe->region_alloc(st->pipe, cpp, width, height, flags); + + mt->target = PIPE_TEXTURE_2D; + mt->internal_format = GL_RGBA; + mt->format = pipeFormat; + mt->first_level = 0; + mt->last_level = 0; + mt->width0 = width; + mt->height0 = height; + mt->depth0 = 1; + mt->cpp = cpp; + mt->compressed = 0; + mt->pitch = mt->region->pitch; + mt->depth_pitch = 0; + mt->total_height = height; + mt->level[0].level_offset = 0; + mt->level[0].width = width; + mt->level[0].height = height; + mt->level[0].depth = 1; + mt->level[0].nr_images = 1; + mt->level[0].image_offset = NULL; + mt->refcount = 1; + + return mt; +} + + /** - * Make mipmap tree containing the glDrawPixels image. + * Make mipmap tree containing an image for glDrawPixels image. + * If 'pixels' is NULL, leave the texture image data undefined. */ static struct pipe_mipmap_tree * make_mipmap_tree(struct st_context *st, @@ -322,7 +385,6 @@ make_mipmap_tree(struct st_context *st, { struct pipe_context *pipe = st->pipe; const struct gl_texture_format *mformat; - const GLbitfield flags = PIPE_SURFACE_FLAG_TEXTURE; struct pipe_mipmap_tree *mt; GLuint pipeFormat, cpp; GLenum baseFormat; @@ -336,7 +398,7 @@ make_mipmap_tree(struct st_context *st, assert(pipeFormat); cpp = st_sizeof_format(pipeFormat); - mt = CALLOC_STRUCT(pipe_mipmap_tree); + mt = alloc_mipmap_tree(st, width, height, pipeFormat); if (!mt) return NULL; @@ -346,15 +408,12 @@ make_mipmap_tree(struct st_context *st, */ printf("st_DrawPixels (sourcing from PBO not implemented yet)\n"); } - else { + + { static const GLuint dstImageOffsets = 0; GLboolean success; + GLuint pitch = mt->region->pitch; GLubyte *dest; - GLuint pitch; - - /* allocate texture region/storage */ - mt->region = st->pipe->region_alloc(st->pipe, cpp, width, height, flags); - pitch = mt->region->pitch; /* map texture region */ dest = pipe->region_map(pipe, mt->region); @@ -377,10 +436,10 @@ make_mipmap_tree(struct st_context *st, /* unmap */ pipe->region_unmap(pipe, mt->region); - assert(success); } +#if 0 mt->target = PIPE_TEXTURE_2D; mt->internal_format = GL_RGBA; mt->format = pipeFormat; @@ -401,7 +460,7 @@ make_mipmap_tree(struct st_context *st, mt->level[0].nr_images = 1; mt->level[0].image_offset = NULL; mt->refcount = 1; - +#endif return mt; } @@ -804,10 +863,6 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLenum format, GLenum type, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { - static struct st_fragment_program *stfp_c = NULL; /* color */ - static struct st_fragment_program *stfp_z = NULL; /* z */ - static struct st_vertex_program *stvp_t = NULL; /* just emit texcoord */ - static struct st_vertex_program *stvp_c = NULL; /* emit color too */ struct st_fragment_program *stfp; struct st_vertex_program *stvp; struct st_context *st = ctx->st; @@ -820,28 +875,12 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, return; } - /* create the fragment programs if needed */ - if (!stfp_c) { - stfp_c = make_fragment_shader(ctx->st, GL_FALSE); - } - if (!stfp_z) { - stfp_z = make_fragment_shader_z(ctx->st); - } - - /* and vertex programs */ - if (!stvp_t) { - stvp_t = make_vertex_shader(ctx->st, GL_FALSE); - } - if (!stvp_c) { - stvp_c = make_vertex_shader(ctx->st, GL_TRUE); - } - st_validate_state(st); if (format == GL_DEPTH_COMPONENT) { ps = st->state.framebuffer.zbuf; - stfp = stfp_z; - stvp = stvp_c; + stfp = make_fragment_shader_z(ctx->st); + stvp = make_vertex_shader(ctx->st, GL_TRUE); color = ctx->Current.RasterColor; } else if (format == GL_STENCIL_INDEX) { @@ -851,8 +890,8 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } else { ps = st->state.framebuffer.cbufs[0]; - stfp = stfp_c; - stvp = stvp_t; + stfp = make_fragment_shader(ctx->st, GL_FALSE); + stvp = make_vertex_shader(ctx->st, GL_FALSE); color = NULL; } @@ -1032,19 +1071,16 @@ static void st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap ) { - static struct st_vertex_program *stvp = NULL; - static struct st_fragment_program *stfp = NULL; + struct st_vertex_program *stvp; + struct st_fragment_program *stfp; struct st_context *st = ctx->st; struct pipe_mipmap_tree *mt; - /* create the fragment program if needed */ - if (!stfp) { - stfp = make_fragment_shader(ctx->st, GL_TRUE); - } + /* create the fragment program */ + stfp = make_fragment_shader(ctx->st, GL_TRUE); + /* and vertex program */ - if (!stvp) { - stvp = make_vertex_shader(ctx->st, GL_TRUE); - } + stvp = make_vertex_shader(ctx->st, GL_TRUE); st_validate_state(st); @@ -1138,22 +1174,59 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, GLint dstx, GLint dsty, GLenum type) { struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; + struct st_renderbuffer *rbRead; + struct st_vertex_program *stvp; + struct st_fragment_program *stfp; + struct pipe_surface *psRead; + struct pipe_mipmap_tree *mt; + GLfloat *color; + uint format; st_validate_state(st); if (type == GL_STENCIL) { + /* can't use texturing to do stencil */ copy_stencil_pixels(ctx, srcx, srcy, width, height, dstx, dsty); return; } - /* allocate a texture of size width x height */ + if (type == GL_COLOR) { + rbRead = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + color = NULL; + stfp = make_fragment_shader(ctx->st, GL_FALSE); + stvp = make_vertex_shader(ctx->st, GL_FALSE); + } + else { + rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer); + color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; + stfp = make_fragment_shader_z(ctx->st); + stvp = make_vertex_shader(ctx->st, GL_TRUE); + } + + psRead = rbRead->surface; + format = psRead->format; - /* blit/copy framebuffer region into texture */ + mt = alloc_mipmap_tree(ctx->st, width, height, format); + if (!mt) + return; - /* draw textured quad */ + /* copy source framebuffer region into mipmap/texture */ + pipe->region_copy(pipe, + mt->region, /* dest */ + 0, /* dest_offset */ + 0, 0, /* destx/y */ + psRead->region, + 0, /* src_offset */ + srcx, srcy, width, height); - fprintf(stderr, "st_CopyPixels not implemented yet\n"); + /* draw textured quad */ + draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2], + width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, + mt, stvp, stfp, color); + + free_mipmap_tree(st->pipe, mt); } -- cgit v1.2.3 From 68b88fab882d882abbdc555ddd2c6df37a7e125e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 14:20:38 -0600 Subject: fix function name --- src/mesa/state_tracker/st_cb_readpixels.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 7bce7e9a27..2c6847e372 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -159,7 +159,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } if (format == GL_STENCIL_INDEX) { - read_stencil_pixels(ctx, x, y, width, height, type, pack, dest); + st_read_stencil_pixels(ctx, x, y, width, height, type, pack, dest); return; } else if (format == GL_DEPTH_COMPONENT) { -- cgit v1.2.3 From d44e515fd77d088862c5f19ef9a7aa92b04b5f13 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 14:22:38 -0600 Subject: fix upside down CopyPixels image --- src/mesa/state_tracker/st_cb_drawpixels.c | 42 ++++++++++++++++++------------- 1 file changed, 25 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 858ce176fe..e4eaed1643 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -479,34 +479,35 @@ free_mipmap_tree(struct pipe_context *pipe, struct pipe_mipmap_tree *mt) */ static void draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, - GLfloat x1, GLfloat y1) + GLfloat x1, GLfloat y1, GLboolean invertTex) { GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ GLuint i; + GLfloat tTop = invertTex, tBot = 1.0 - tTop; /* upper-left */ verts[0][0][0] = x0; /* attr[0].x */ verts[0][0][1] = y0; /* attr[0].x */ verts[0][1][0] = 0.0; /* attr[1].s */ - verts[0][1][1] = 0.0; /* attr[1].t */ + verts[0][1][1] = tTop; /* attr[1].t */ /* upper-right */ verts[1][0][0] = x1; verts[1][0][1] = y0; verts[1][1][0] = 1.0; - verts[1][1][1] = 0.0; + verts[1][1][1] = tTop; /* lower-right */ verts[2][0][0] = x1; verts[2][0][1] = y1; verts[2][1][0] = 1.0; - verts[2][1][1] = 1.0; + verts[2][1][1] = tBot; /* lower-left */ verts[3][0][0] = x0; verts[3][0][1] = y1; verts[3][1][0] = 0.0; - verts[3][1][1] = 1.0; + verts[3][1][1] = tBot; /* same for all verts: */ for (i = 0; i < 4; i++) { @@ -522,34 +523,36 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, static void draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, - GLfloat x1, GLfloat y1, const GLfloat *color) + GLfloat x1, GLfloat y1, const GLfloat *color, + GLboolean invertTex) { GLfloat verts[4][3][4]; /* four verts, three attribs, XYZW */ GLuint i; + GLfloat tTop = invertTex, tBot = 1.0 - tTop; /* upper-left */ verts[0][0][0] = x0; /* attr[0].x */ - verts[0][0][1] = y0; /* attr[0].x */ + verts[0][0][1] = y0; /* attr[0].y */ verts[0][2][0] = 0.0; /* attr[2].s */ - verts[0][2][1] = 0.0; /* attr[2].t */ + verts[0][2][1] = tTop; /* attr[2].t */ /* upper-right */ verts[1][0][0] = x1; verts[1][0][1] = y0; verts[1][2][0] = 1.0; - verts[1][2][1] = 0.0; + verts[1][2][1] = tTop; /* lower-right */ verts[2][0][0] = x1; verts[2][0][1] = y1; verts[2][2][0] = 1.0; - verts[2][2][1] = 1.0; + verts[2][2][1] = tBot; /* lower-left */ verts[3][0][0] = x0; verts[3][0][1] = y1; verts[3][2][0] = 0.0; - verts[3][2][1] = 1.0; + verts[3][2][1] = tBot; /* same for all verts: */ for (i = 0; i < 4; i++) { @@ -575,7 +578,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, struct pipe_mipmap_tree *mt, struct st_vertex_program *stvp, struct st_fragment_program *stfp, - const GLfloat *color) + const GLfloat *color, + GLboolean invertTex) { const GLuint unit = 0; struct pipe_context *pipe = ctx->st->pipe; @@ -655,9 +659,9 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* draw textured quad */ if (color) - draw_quad_colored(ctx, x0, y0, z, x1, y1, color); + draw_quad_colored(ctx, x0, y0, z, x1, y1, color, invertTex); else - draw_quad(ctx, x0, y0, z, x1, y1); + draw_quad(ctx, x0, y0, z, x1, y1, invertTex); /* restore GL state */ pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); @@ -907,7 +911,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (mt) { draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, - mt, stvp, stfp, color); + mt, stvp, stfp, color, GL_FALSE); free_mipmap_tree(st->pipe, mt); } } @@ -1089,7 +1093,7 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, 1.0, 1.0, mt, stvp, stfp, - ctx->Current.RasterColor); + ctx->Current.RasterColor, GL_FALSE); free_mipmap_tree(st->pipe, mt); } @@ -1211,6 +1215,10 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, if (!mt) return; + if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { + srcy = ctx->DrawBuffer->Height - srcy - height - 1; + } + /* copy source framebuffer region into mipmap/texture */ pipe->region_copy(pipe, mt->region, /* dest */ @@ -1224,7 +1232,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, /* draw textured quad */ draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, - mt, stvp, stfp, color); + mt, stvp, stfp, color, GL_TRUE); free_mipmap_tree(st->pipe, mt); } -- cgit v1.2.3 From 874b6b9960ee3778182cf331f6faf59d477eb214 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 14:30:03 -0600 Subject: fix off-by-one error in CopyPixels src position --- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index e4eaed1643..b88b96e3b2 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1216,7 +1216,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, return; if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { - srcy = ctx->DrawBuffer->Height - srcy - height - 1; + srcy = ctx->DrawBuffer->Height - srcy - height; } /* copy source framebuffer region into mipmap/texture */ -- cgit v1.2.3 From 5d39f4f9fdba4cd3333a39af583a73b85eb6bb74 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 15:14:50 -0600 Subject: fix bug in copy_rect(), use temp vars in sp_region_copy() to aid debugging --- src/mesa/pipe/softpipe/sp_region.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 982e081f60..0f4d2f0575 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -147,7 +147,7 @@ copy_rect(ubyte * dst, dst += dst_x * cpp; src += src_x * cpp; dst += dst_y * dst_pitch; - src += src_y * dst_pitch; + src += src_y * src_pitch; width *= cpp; if (width == dst_pitch && width == src_pitch) @@ -197,15 +197,18 @@ sp_region_copy(struct pipe_context *pipe, unsigned src_offset, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { + ubyte *src_map, *dst_map; assert( dst->cpp == src->cpp ); - copy_rect(pipe->region_map(pipe, dst) + dst_offset, + dst_map = pipe->region_map(pipe, dst); + src_map = pipe->region_map(pipe, src); + copy_rect(dst_map + dst_offset, dst->cpp, dst->pitch, - dstx, dsty, - width, height, - pipe->region_map(pipe, src) + src_offset, - src->pitch, + dstx, dsty, + width, height, + src_map + src_offset, + src->pitch, srcx, srcy); pipe->region_unmap(pipe, src); -- cgit v1.2.3 From 0007cd7ba0a61fcbcf9c9d19e014408be25ae496 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 15:18:55 -0600 Subject: Alternate CopyPixels path based on get/put_tile(). For some drivers (like Xlib) it's not possible to treat the front/back color buffers as pipe_regions. So pipe->region_copy() won't work. Added a new state tracker field indicating if we can use regions for colorbuffer accesses. This should probably be re-considered someday... --- src/mesa/drivers/x11/xm_api.c | 2 ++ src/mesa/state_tracker/st_cb_drawpixels.c | 42 +++++++++++++++++++++++++------ src/mesa/state_tracker/st_context.c | 1 + src/mesa/state_tracker/st_context.h | 6 +++++ 4 files changed, 43 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 1d3f799f36..ff83dab075 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -1603,6 +1603,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) mesaCtx->st->pipe->surface_alloc = xmesa_surface_alloc; mesaCtx->st->pipe->supported_formats = xmesa_supported_formats; + mesaCtx->st->haveFramebufferRegions = GL_FALSE; + /* special pipe->clear function */ mesaCtx->st->pipe->clear = xmesa_clear; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index b88b96e3b2..90f91c39b5 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1219,14 +1219,40 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, srcy = ctx->DrawBuffer->Height - srcy - height; } - /* copy source framebuffer region into mipmap/texture */ - pipe->region_copy(pipe, - mt->region, /* dest */ - 0, /* dest_offset */ - 0, 0, /* destx/y */ - psRead->region, - 0, /* src_offset */ - srcx, srcy, width, height); + /* For some drivers (like Xlib) it's not possible to treat the + * front/back color buffers as regions (they're XImages and Pixmaps). + * So, this var tells us if we can use region_copy here... + */ + if (st->haveFramebufferRegions) { + /* copy source framebuffer region into mipmap/texture */ + pipe->region_copy(pipe, + mt->region, /* dest */ + 0, /* dest_offset */ + 0, 0, /* destx/y */ + psRead->region, + 0, /* src_offset */ + srcx, srcy, width, height); + } + else { + /* alternate path using get/put_tile() */ + struct pipe_surface *psTex; + GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + psTex = pipe->get_tex_surface(pipe, mt, 0, 0, 0); + + (void) pipe->region_map(pipe, psRead->region); + (void) pipe->region_map(pipe, psTex->region); + + psRead->get_tile(psRead, srcx, srcy, width, height, buf); + psTex->put_tile(psTex, 0, 0, width, height, buf); + + pipe->region_unmap(pipe, psRead->region); + pipe->region_unmap(pipe, psTex->region); + + pipe_surface_reference(&psTex, NULL); + + free(buf); + } /* draw textured quad */ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 09d9b1ea3b..8ced3f504c 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -90,6 +90,7 @@ struct st_context *st_create_context( GLcontext *ctx, st->ctx->VertexProgram._MaintainTnlProgram = GL_TRUE; + st->haveFramebufferRegions = GL_TRUE; #if 0 st_init_cb_clear( st ); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index b4ae041d36..bacc0b9b3f 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -111,6 +111,12 @@ struct st_context char vendor[100]; char renderer[100]; + /** Can we access the front/back color buffers as pipe_regions? + * We can't with the Xlib driver... + * This is a hack that should be fixed someday. + */ + GLboolean haveFramebufferRegions; + /* State to be validated: */ struct st_tracked_state **atoms; -- cgit v1.2.3 From b57f573b9907df351ccc9a7853e94b0efdf35218 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 18 Oct 2007 16:27:01 -0600 Subject: remove old comment --- src/mesa/pipe/softpipe/sp_context.c | 5 ----- 1 file changed, 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index f2f2e36869..44cc685105 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -355,10 +355,5 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); - /* - * XXX we could plug GL selection/feedback into the drawing pipeline - * by specifying a different setup/render stage. - */ - return &softpipe->pipe; } -- cgit v1.2.3 From 5cf4fc5832cef855158337c2bb71ebcc4e3caa1d Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 19 Oct 2007 10:07:55 -0600 Subject: clamp/clip in put_tile(), move softpipe_init_surface_funcs() call --- src/mesa/drivers/x11/xm_surface.c | 25 +++++++++++++++++++++---- 1 file changed, 21 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 75ad450360..86fecdb2f7 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -49,6 +49,15 @@ #include "state_tracker/st_context.h" +#define CLIP_TILE \ + do { \ + if (x + w > ps->width) \ + w = ps->width - x; \ + if (y + h > ps->height) \ + h = ps->height -y; \ + } while(0) + + static INLINE struct xmesa_surface * xmesa_surf(struct softpipe_surface *sps) { @@ -176,14 +185,18 @@ get_tile(struct pipe_surface *ps, struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); GLubyte tmp[MAX_WIDTH * 4]; GLuint i, j; + unsigned w0 = w; GET_CURRENT_CONTEXT(ctx); + + CLIP_TILE; + FLIP(y); for (i = 0; i < h; i++) { xrb->St.Base.GetRow(ctx, &xrb->St.Base, w, x, y - i, tmp); for (j = 0; j < w * 4; j++) { p[j] = UBYTE_TO_FLOAT(tmp[j]); } - p += w * 4; + p += w0 * 4; } } @@ -195,14 +208,16 @@ put_tile(struct pipe_surface *ps, struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); GLubyte tmp[MAX_WIDTH * 4]; GLuint i, j; + unsigned w0 = w; GET_CURRENT_CONTEXT(ctx); + CLIP_TILE; FLIP(y); for (i = 0; i < h; i++) { for (j = 0; j < w * 4; j++) { - CLAMPED_FLOAT_TO_UBYTE(tmp[j], p[j]); + UNCLAMPED_FLOAT_TO_UBYTE(tmp[j], p[j]); } xrb->St.Base.PutRow(ctx, &xrb->St.Base, w, x, y - i, tmp, NULL); - p += w * 4; + p += w0 * 4; } } @@ -223,6 +238,9 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) xms->surface.surface.format = pipeFormat; xms->surface.surface.refcount = 1; + /* some of the functions plugged in by this call will get overridden */ + softpipe_init_surface_funcs(&xms->surface); + switch (pipeFormat) { case PIPE_FORMAT_U_A8_R8_G8_B8: xms->surface.read_quad_f_swz = read_quad_f; @@ -236,7 +254,6 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) xms->surface.surface.put_tile = put_tile; break; case PIPE_FORMAT_S8_Z24: - softpipe_init_surface_funcs(&xms->surface); /* xms->surface.read_quad_z = 1; xms->surface.write_quad_z = 1; -- cgit v1.2.3 From 46c3cf18315345effd15a69987294c1195843e2a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 19 Oct 2007 10:08:24 -0600 Subject: Tile cache functions. --- src/mesa/pipe/softpipe/sp_tile_cache.c | 219 +++++++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_tile_cache.h | 70 +++++++++++ 2 files changed, 289 insertions(+) create mode 100644 src/mesa/pipe/softpipe/sp_tile_cache.c create mode 100644 src/mesa/pipe/softpipe/sp_tile_cache.h (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c new file mode 100644 index 0000000000..c7e6e6e6d4 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -0,0 +1,219 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Framebuffer/surface tile caching. + * + * Author: + * Brian Paul + */ + + +#include "sp_context.h" +#include "sp_surface.h" +#include "sp_tile_cache.h" + + +#define NUM_ENTRIES 20 + + +/** XXX move these */ +#define MAX_WIDTH 2048 +#define MAX_HEIGHT 2048 + + +struct softpipe_tile_cache +{ + struct softpipe_cached_tile entries[NUM_ENTRIES]; + uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; +}; + + +/** + * Return the position in the cache for the tile that contains win pos (x,y). + * We currently use a direct mapped cache so this is like a hack key. + * At some point we should investige something more sophisticated, like + * a LRU replacement policy. + */ +#define CACHE_POS(x, y) \ + (((x) / TILE_SIZE + ((y) / TILE_SIZE) * 5) % NUM_ENTRIES) + + + +static uint +is_clear_flag_set(const uint *bitvec, int x, int y) +{ + int pos, bit; + x /= TILE_SIZE; + y /= TILE_SIZE; + pos = y * (MAX_WIDTH / TILE_SIZE) + x; + bit = bitvec[pos / 32] & (1 << (pos & 31)); + return bit; +} + + +static void +clear_clear_flag(uint *bitvec, int x, int y) +{ + int pos; + x /= TILE_SIZE; + y /= TILE_SIZE; + pos = y * (MAX_WIDTH / TILE_SIZE) + x; + bitvec[pos / 32] &= ~(1 << (pos & 31)); +} + + +struct softpipe_tile_cache * +sp_create_tile_cache(void) +{ + struct softpipe_tile_cache *tc; + uint pos; + + tc = calloc(1, sizeof(*tc)); + if (tc) { + for (pos = 0; pos < NUM_ENTRIES; pos++) { + tc->entries[pos].x = + tc->entries[pos].y = -1; + } + } + return tc; +} + + +void +sp_destroy_tile_cache(struct softpipe_tile_cache *tc) +{ + uint pos; + for (pos = 0; pos < NUM_ENTRIES; pos++) { + assert(tc->entries[pos].x < 0); + } + free(tc); +} + + + +void +sp_flush_tile_cache(struct softpipe_surface *sps) +{ + struct softpipe_tile_cache *tc = sps->tc; + /* + struct softpipe_surface *zsurf = softpipe_surface(softpipe->zbuf); + */ + int inuse = 0, pos; + + for (pos = 0; pos < NUM_ENTRIES; pos++) { + struct softpipe_cached_tile *tile = tc->entries + pos; + if (tile->x >= 0) { + sps->surface.put_tile(&sps->surface, + tile->x, + tile->y, + TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + /* + sps->surface.put_tile(&zsurf->surface, + tile->x, + tile->y, + TILE_SIZE, TILE_SIZE, + (float *) tile->depth); + */ + + tile->x = tile->y = -1; /* mark as empty */ + inuse++; + } + } + + printf("flushed tiles in use: %d\n", inuse); + +} + + +struct softpipe_cached_tile * +sp_get_cached_tile(struct softpipe_surface *sps, int x, int y) +{ + struct softpipe_tile_cache *tc = sps->tc; + + /* tile pos in framebuffer: */ + const int tile_x = x & ~(TILE_SIZE - 1); + const int tile_y = y & ~(TILE_SIZE - 1); + + /* cache pos/entry: */ + const int pos = CACHE_POS(x, y); + struct softpipe_cached_tile *tile = tc->entries + pos; + + /* + printf("output %d, %d at pos %d\n", x, y, pos); + */ + + if (tile_x != tile->x || + tile_y != tile->y) { + + /* + printf("new tile at %d, %d, pos %d\n", tile_x, tile_y, pos); + */ + + if (tile->x != -1) { + /* put dirty tile back in framebuffer */ + sps->surface.put_tile(&sps->surface, tile->x, tile->y, + TILE_SIZE, TILE_SIZE, (float *) tile->data.color); + } + + if (is_clear_flag_set(tc->clear_flags, x, y)) { + printf("clear tile\n"); +#if 0 + uint i, j; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.color[i][j][0] = 0.5; + tile->data.color[i][j][1] = 0.5; + tile->data.color[i][j][2] = 0.5; + tile->data.color[i][j][3] = 0.5; + } + } +#endif + memset(tile->data.color, 0, sizeof(tile->data.color)); + clear_clear_flag(tc->clear_flags, x, y); + } + else { + /* get new tile from framebuffer */ + sps->surface.get_tile(&sps->surface, tile_x, tile_y, + TILE_SIZE, TILE_SIZE, (float *) tile->data.color); + } + + tile->x = tile_x; + tile->y = tile_y; + } + + return tile; +} + + +void +sp_clear_tile_cache(struct softpipe_surface *sps, unsigned clearval) +{ + (void) clearval; /* XXX use this */ + memset(sps->tc->clear_flags, 255, sizeof(sps->tc->clear_flags)); +} diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h new file mode 100644 index 0000000000..4248361f99 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -0,0 +1,70 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_TILE_CACHE_H +#define SP_TILE_CACHE_H + + +#include "p_compiler.h" + + +struct softpipe_context; +struct softpipe_tile_cache; + + +#define TILE_SIZE 64 + + + +struct softpipe_cached_tile +{ + int x, y; /** pos of tile in window coords */ + union { + float color[TILE_SIZE][TILE_SIZE][4]; + uint depth[TILE_SIZE][TILE_SIZE]; + } data; +}; + + +extern struct softpipe_tile_cache * +sp_create_tile_cache(void); + +extern void +sp_destroy_tile_cache(struct softpipe_tile_cache *tc); + +extern void +sp_flush_tile_cache(struct softpipe_surface *sps); + +extern void +sp_clear_tile_cache(struct softpipe_surface *sps, unsigned clearval); + +extern struct softpipe_cached_tile * +sp_get_cached_tile(struct softpipe_surface *sps, int x, int y); + + +#endif /* SP_TILE_CACHE_H */ + -- cgit v1.2.3 From 2b2f761e2b0dc160793be2f48e811d2d455e1e22 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 19 Oct 2007 10:10:08 -0600 Subject: Initial implementation of surface tile caching. Instead of using read/write_quad() functions, do framebuffer accesses via get/put_tile(). A cache of tiles is used to avoid frequent get/put() calls. Only implemented for color buffers right now. --- src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_clear.c | 5 ++ src/mesa/pipe/softpipe/sp_flush.c | 12 ++++ src/mesa/pipe/softpipe/sp_quad_blend.c | 33 +++++++-- src/mesa/pipe/softpipe/sp_quad_colormask.c | 18 ++++- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 20 ++++++ src/mesa/pipe/softpipe/sp_quad_output.c | 101 +++++++++++----------------- src/mesa/pipe/softpipe/sp_surface.c | 8 ++- src/mesa/pipe/softpipe/sp_surface.h | 3 + 9 files changed, 129 insertions(+), 72 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 63eb6b5761..2cbd9e5b51 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -35,6 +35,7 @@ DRIVER_SOURCES = \ sp_state_vertex.c \ sp_tex_layout.c \ sp_tex_sample.c \ + sp_tile_cache.c \ sp_surface.c C_SOURCES = \ diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 539da1d58b..46370b4ed9 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -35,6 +35,7 @@ #include "sp_context.h" #include "sp_surface.h" #include "sp_state.h" +#include "sp_tile_cache.h" /** @@ -46,6 +47,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); + struct softpipe_surface *sps = softpipe_surface(ps); unsigned x, y, w, h; softpipe_update_derived(softpipe); /* not needed?? */ @@ -64,5 +66,8 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, assert(w <= ps->region->pitch); assert(h <= ps->region->height); + /* XXX skip this fill if we're using tile cache */ pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue); + + sp_clear_tile_cache(sps, clearValue); } diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index cdf4a53c83..d4eada8228 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -33,8 +33,11 @@ #include "pipe/p_defines.h" #include "sp_flush.h" #include "sp_context.h" +#include "sp_surface.h" +#include "sp_tile_cache.h" #include "sp_winsys.h" + /* There will be actual work to do here. In future we may want a * fence-like interface instead of finish, and perhaps flush will take * flags to indicate what type of flush is required. @@ -43,9 +46,18 @@ void softpipe_flush( struct pipe_context *pipe, unsigned flags ) { + struct softpipe_context *softpipe = softpipe_context(pipe); + uint i; + /* - flush the quad pipeline * - flush the texture cache * - flush the render cache */ + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.cbufs[i]); + if (sps) + sp_flush_tile_cache(sps); + } } diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 3af5ab5b29..43294e4cd4 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -35,6 +35,7 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" +#include "sp_tile_cache.h" #include "sp_quad.h" @@ -106,10 +107,19 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) uint *src4 = (uint *) src; uint *dst4 = (uint *) dst; uint *res4 = (uint *) res; - uint j; + struct softpipe_cached_tile * + tile = sp_get_cached_tile(sps, quad->x0, quad->y0); + uint i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } + } - /* get colors from framebuffer */ - sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); /* convert to ubyte */ for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */ UNCLAMPED_FLOAT_TO_UBYTE(dst[j][0], dest[j][0]); /* P0 */ @@ -209,19 +219,28 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) static void blend_quad(struct quad_stage *qs, struct quad_header *quad) { - static const float zero[4] = { 0, 0, 0, 0 }; - static const float one[4] = { 1, 1, 1, 1 }; struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); + static const float zero[4] = { 0, 0, 0, 0 }; + static const float one[4] = { 1, 1, 1, 1 }; float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; + struct softpipe_cached_tile * + tile = sp_get_cached_tile(sps, quad->x0, quad->y0); + uint i, j; if (softpipe->blend->logicop_enable) { logicop_quad(qs, quad); return; } - /* get colors from framebuffer */ - sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } + } /* * Compute src/first term RGB diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index c3e110532a..aa69f53580 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -36,17 +36,31 @@ #include "sp_headers.h" #include "sp_surface.h" #include "sp_quad.h" +#include "sp_tile_cache.h" +/** + * XXX colormask could be rolled into blending... + */ static void colormask_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); float dest[4][QUAD_SIZE]; - - sps->read_quad_f_swz(sps, quad->x0, quad->y0, dest); + struct softpipe_cached_tile * + tile = sp_get_cached_tile(sps, quad->x0, quad->y0); + uint i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } + } /* R */ if (!(softpipe->blend->colormask & PIPE_MASK_R)) diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index ff1d84a02d..efd08b981c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -32,6 +32,7 @@ #include "sp_headers.h" #include "sp_surface.h" #include "sp_quad.h" +#include "sp_tile_cache.h" /** @@ -48,6 +49,9 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) unsigned zmask = 0; unsigned j; float scale; +#if 0 + struct cached_tile *tile = sp_get_cached_tile(softpipe, quad->x0, quad->y0); +#endif assert(sps); /* shouldn't get here if there's no zbuffer */ @@ -74,8 +78,16 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); } +#if 0 + for (j = 0; j < 4; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + bzzzz[j] = tile->depth[y][x]; + } +#else /* get zquad from zbuffer */ sps->read_quad_z(sps, quad->x0, quad->y0, bzzzz); +#endif switch (softpipe->depth_stencil->depth.func) { case PIPE_FUNC_NEVER: @@ -139,8 +151,16 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } } +#if 1 /* write updated zquad to zbuffer */ sps->write_quad_z(sps, quad->x0, quad->y0, bzzzz); +#else + for (j = 0; j < 4; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + tile->depth[y][x] = bzzzz[j]; + } +#endif } } diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index 49420c4fe7..01ee06a69c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -1,57 +1,36 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ #include "pipe/p_util.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" #include "sp_quad.h" - - -static void mask_copy( float (*dest)[4], - float (*src)[4], - unsigned mask ) -{ - unsigned i, j; - - for (i = 0; i < 4; i++) { - if (mask & (1<softpipe; struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); - - if (quad->mask != MASK_ALL) { - float tmp[4][QUAD_SIZE]; - - /* XXX probably add a masked-write function someday */ - - sps->read_quad_f_swz(sps, quad->x0, quad->y0, tmp); - - mask_copy( tmp, quad->outputs.color, quad->mask ); - - sps->write_quad_f_swz(sps, quad->x0, quad->y0, tmp); - } - else if (quad->mask) { - sps->write_quad_f_swz(sps, quad->x0, quad->y0, quad->outputs.color); + struct softpipe_cached_tile *tile + = sp_get_cached_tile(sps, quad->x0, quad->y0); + /* in-tile pos: */ + const int itx = quad->x0 % TILE_SIZE; + const int ity = quad->y0 % TILE_SIZE; + int i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + int x = itx + (j & 1); + int y = ity + (j >> 1); + for (i = 0; i < 4; i++) { /* loop over color chans */ + tile->data.color[y][x][i] = quad->outputs.color[i][j]; + } + } } } diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 292676a302..e5b4f249fa 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -25,12 +25,12 @@ * **************************************************************************/ +#include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_state.h" #include "sp_surface.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" - +#include "sp_tile_cache.h" /** * Softpipe surface functions. @@ -871,6 +871,8 @@ s8_write_quad_stencil(struct softpipe_surface *sps, void softpipe_init_surface_funcs(struct softpipe_surface *sps) { + sps->tc = sp_create_tile_cache(); + assert(sps->surface.format); switch (sps->surface.format) { diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 545983f431..431303553f 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -37,6 +37,7 @@ struct pipe_context; struct softpipe_surface; struct softpipe_context; +struct softpipe_tile_cache; /** @@ -45,6 +46,8 @@ struct softpipe_context; struct softpipe_surface { struct pipe_surface surface; + struct softpipe_tile_cache *tc; + /** * Functions for read/writing surface data */ -- cgit v1.2.3 From 96b06ac557b721f0b1b9cc05b1d2b3a289701ae5 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 19 Oct 2007 10:12:00 -0600 Subject: call Driver.Flush() in _mesa_notifySwapBuffers() --- src/mesa/main/context.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 00e4c8328e..e8335e19ca 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -163,9 +163,11 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ); * We have to finish any pending rendering. */ void -_mesa_notifySwapBuffers(__GLcontext *gc) +_mesa_notifySwapBuffers(__GLcontext *ctx) { - FLUSH_VERTICES( gc, 0 ); + if (ctx->Driver.Flush) { + ctx->Driver.Flush(ctx); + } } -- cgit v1.2.3 From 257f0da6a2e9cc687c0cdc76c0fb326eab9e9ac9 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 19 Oct 2007 10:13:55 -0600 Subject: disable debug printf --- src/mesa/pipe/softpipe/sp_tile_cache.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index c7e6e6e6d4..2ecdfc89e4 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -146,8 +146,7 @@ sp_flush_tile_cache(struct softpipe_surface *sps) } } - printf("flushed tiles in use: %d\n", inuse); - + /*printf("flushed tiles in use: %d\n", inuse);*/ } -- cgit v1.2.3 From ffd37b1fda1c880028f2c353edbf4807d55b6844 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 19 Oct 2007 12:42:05 -0600 Subject: don't alloc region in xmesa_surface_alloc(), fixes a mem leak --- src/mesa/drivers/x11/xm_surface.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 86fecdb2f7..8d2c2ca5f3 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -284,6 +284,7 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) { struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + assert(pipe); assert(pipeFormat); xms->surface.surface.format = pipeFormat; @@ -293,9 +294,6 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) */ softpipe_init_surface_funcs(&xms->surface); - assert(pipe); - xms->surface.surface.region = pipe->region_alloc(pipe, 1, 1, 1, 0x0); - return &xms->surface.surface; } -- cgit v1.2.3 From d4a9d4bdefb0670554c2e8ee35129d40c06b64ad Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 19 Oct 2007 12:45:54 -0600 Subject: added get/put_tile_raw() methods --- src/mesa/pipe/p_state.h | 21 +++++++++++++++++---- 1 file changed, 17 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 1c3b2811a4..daa74ec053 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -286,13 +286,26 @@ struct pipe_surface unsigned offset; /**< offset from start of region, in bytes */ unsigned refcount; - /** get block/tile of pixels from surface */ + /** + * Get block/tile of pixels from surface as floats + * If color surface, return float[4]. If depth surface, return float[1]. + */ void (*get_tile)(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p); + uint x, uint y, uint w, uint h, float *p); - /** put block/tile of pixels into surface */ + /** + * Put block/tile of pixels into surface as floats + * If color surface, data is float[4]. If depth surface, data is float[1]. + */ void (*put_tile)(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, const float *p); + uint x, uint y, uint w, uint h, const float *p); + + /** As above, but data is raw pixel data */ + void (*get_tile_raw)(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, void *p); + /** As above, but data is raw pixel data */ + void (*put_tile_raw)(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const void *p); }; -- cgit v1.2.3 From c492725abfab8df545a5adea3cd124cba3ba5c4c Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 19 Oct 2007 12:47:05 -0600 Subject: get/put_tile_raw() funcs for 16/32bpp surfaces --- src/mesa/pipe/softpipe/sp_surface.c | 115 ++++++++++++++++++++++++++++++++++++ 1 file changed, 115 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index e5b4f249fa..8aad5f6c13 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -865,6 +865,115 @@ s8_write_quad_stencil(struct softpipe_surface *sps, } +/** + * Get raw tile, any 32-bit pixel format. + */ +static void +get_tile_raw32(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, void *p) +{ + const uint *src + = ((const uint *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + uint *pDest = (uint *) p; + unsigned i; + unsigned w0 = w; + + assert(ps->region->map); + assert(ps->format == PIPE_FORMAT_S8_Z24 || + ps->format == PIPE_FORMAT_U_Z32); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + memcpy(pDest, src, w * sizeof(uint)); + src += ps->region->pitch; + pDest += w0; + } +} + + +/** + * put raw tile, any 32-bit pixel format. + */ +static void +put_tile_raw32(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const void *p) +{ + uint *dst + = ((uint *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + const uint *pSrc = (const uint *) p; + unsigned i; + unsigned w0 = w; + + assert(ps->region->map); + assert(ps->format == PIPE_FORMAT_S8_Z24 || + ps->format == PIPE_FORMAT_U_Z32); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + memcpy(dst, pSrc, w * sizeof(uint)); + dst += ps->region->pitch; + pSrc += w0; + } +} + + +/** + * Get raw tile, any 16-bit pixel format. + */ +static void +get_tile_raw16(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, void *p) +{ + const ushort *src + = ((const ushort *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + ushort *pDest = (ushort *) p; + uint i; + uint w0 = w; + + assert(ps->format == PIPE_FORMAT_U_Z16); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + memcpy(pDest, src, w * sizeof(ushort)); + src += ps->region->pitch; + pDest += w0; + } +} + +/** + * Put raw tile, any 16-bit pixel format. + */ +static void +put_tile_raw16(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const void *p) +{ + ushort *dst + = ((ushort *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + const ushort *pSrc = (const ushort *) p; + unsigned i; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_Z16); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + memcpy(dst, pSrc, w * sizeof(ushort)); + dst += ps->region->pitch; + pSrc += w0; + } +} + + + + /** * Initialize the quad_read/write and get/put_tile() methods. */ @@ -915,11 +1024,15 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) sps->read_quad_z = z16_read_quad_z; sps->write_quad_z = z16_write_quad_z; sps->surface.get_tile = z16_get_tile; + sps->surface.get_tile_raw = get_tile_raw16; + sps->surface.put_tile_raw = put_tile_raw16; break; case PIPE_FORMAT_U_Z32: sps->read_quad_z = z32_read_quad_z; sps->write_quad_z = z32_write_quad_z; sps->surface.get_tile = z32_get_tile; + sps->surface.get_tile_raw = get_tile_raw32; + sps->surface.put_tile_raw = put_tile_raw32; break; case PIPE_FORMAT_S8_Z24: sps->read_quad_z = s8z24_read_quad_z; @@ -927,6 +1040,8 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) sps->read_quad_stencil = s8z24_read_quad_stencil; sps->write_quad_stencil = s8z24_write_quad_stencil; sps->surface.get_tile = s8z24_get_tile; + sps->surface.get_tile_raw = get_tile_raw32; + sps->surface.put_tile_raw = put_tile_raw32; break; case PIPE_FORMAT_U_S8: -- cgit v1.2.3 From 832e73bc098fa8fd680d70cb495f495b33769630 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 10:09:12 -0600 Subject: added case for TGSI_OPCODE_END --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index d0d5719c93..77ba7bb944 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -1747,6 +1747,10 @@ emit_instruction( assert( 0 ); break; + case TGSI_OPCODE_END: + /* nothing */ + break; + default: assert( 0 ); } -- cgit v1.2.3 From 7e8396399824108d62dc3e02b2af0422e98aab8e Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 15:18:02 -0600 Subject: Convert Z/stencil ops to use cached tiles like colors. Also, quite a bit of re-org of the tile caches and surface mapping/unmapping. Leave surfaces mapped between primitives now. --- src/mesa/pipe/softpipe/sp_clear.c | 2 + src/mesa/pipe/softpipe/sp_context.c | 82 ++++++++++++------ src/mesa/pipe/softpipe/sp_context.h | 8 ++ src/mesa/pipe/softpipe/sp_draw_arrays.c | 5 +- src/mesa/pipe/softpipe/sp_flush.c | 14 +-- src/mesa/pipe/softpipe/sp_quad_blend.c | 6 +- src/mesa/pipe/softpipe/sp_quad_colormask.c | 3 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 127 +++++++++++++++++++--------- src/mesa/pipe/softpipe/sp_quad_output.c | 3 +- src/mesa/pipe/softpipe/sp_quad_stencil.c | 52 ++++++++++-- src/mesa/pipe/softpipe/sp_state.h | 6 ++ src/mesa/pipe/softpipe/sp_state_surface.c | 88 ++++++++++++++++++- src/mesa/pipe/softpipe/sp_surface.c | 9 +- src/mesa/pipe/softpipe/sp_surface.h | 2 - src/mesa/pipe/softpipe/sp_tile_cache.c | 97 ++++++++++++++------- src/mesa/pipe/softpipe/sp_tile_cache.h | 17 +++- 16 files changed, 395 insertions(+), 126 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 46370b4ed9..4f04f8243f 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -69,5 +69,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, /* XXX skip this fill if we're using tile cache */ pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue); +#if 0 sp_clear_tile_cache(sps, clearValue); +#endif } diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 44cc685105..87eaf6fb54 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -39,6 +39,7 @@ #include "sp_region.h" #include "sp_state.h" #include "sp_surface.h" +#include "sp_tile_cache.h" #include "sp_tex_layout.h" #include "sp_winsys.h" @@ -113,6 +114,9 @@ softpipe_max_texture_size(struct pipe_context *pipe, unsigned textureType, } +/** + * Map any drawing surfaces which aren't already mapped + */ void softpipe_map_surfaces(struct softpipe_context *sp) { @@ -120,68 +124,88 @@ softpipe_map_surfaces(struct softpipe_context *sp) unsigned i; for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); - if (sps->surface.region) - pipe->region_map(pipe, sps->surface.region); + struct pipe_surface *ps = sp->framebuffer.cbufs[i]; + if (ps->region && !ps->region->map) { + pipe->region_map(pipe, ps->region); + } } if (sp->framebuffer.zbuf) { - struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.zbuf); - if (sps->surface.region) - pipe->region_map(pipe, sps->surface.region); + struct pipe_surface *ps = sp->framebuffer.zbuf; + if (ps->region && !ps->region->map) { + pipe->region_map(pipe, ps->region); + } } if (sp->framebuffer.sbuf) { - struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.sbuf); - if (sps->surface.region) - pipe->region_map(pipe, sps->surface.region); + struct pipe_surface *ps = sp->framebuffer.sbuf; + if (ps->region && !ps->region->map) { + pipe->region_map(pipe, ps->region); + } } +} + + +void +softpipe_map_texture_surfaces(struct softpipe_context *sp) +{ + struct pipe_context *pipe = &sp->pipe; + uint i; - /* textures */ for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { struct pipe_mipmap_tree *mt = sp->texture[i]; if (mt) { pipe->region_map(pipe, mt->region); } } - - /* XXX depth & stencil bufs */ } +/** + * Unmap any mapped drawing surfaces + */ void softpipe_unmap_surfaces(struct softpipe_context *sp) { struct pipe_context *pipe = &sp->pipe; - unsigned i; + uint i; + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + sp_flush_tile_cache(sp->cbuf_cache[i]); + sp_flush_tile_cache(sp->zbuf_cache); + sp_flush_tile_cache(sp->sbuf_cache); for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.cbufs[i]); - if (sps->surface.region) - pipe->region_unmap(pipe, sps->surface.region); + struct pipe_surface *ps = sp->framebuffer.cbufs[i]; + if (ps->region) + pipe->region_unmap(pipe, ps->region); } if (sp->framebuffer.zbuf) { - struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.zbuf); - if (sps->surface.region) - pipe->region_unmap(pipe, sps->surface.region); + struct pipe_surface *ps = sp->framebuffer.zbuf; + if (ps->region) + pipe->region_unmap(pipe, ps->region); } - if (sp->framebuffer.sbuf) { - struct softpipe_surface *sps = softpipe_surface(sp->framebuffer.sbuf); - if (sps->surface.region) - pipe->region_unmap(pipe, sps->surface.region); + if (sp->framebuffer.sbuf && sp->framebuffer.sbuf != sp->framebuffer.zbuf) { + struct pipe_surface *ps = sp->framebuffer.sbuf; + if (ps->region) + pipe->region_unmap(pipe, ps->region); } +} - /* textures */ + +void +softpipe_unmap_texture_surfaces(struct softpipe_context *sp) +{ + struct pipe_context *pipe = &sp->pipe; + uint i; for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { struct pipe_mipmap_tree *mt = sp->texture[i]; if (mt) { pipe->region_unmap(pipe, mt->region); } } - - /* XXX depth & stencil bufs */ } @@ -248,6 +272,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, struct softpipe_winsys *softpipe_winsys ) { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); + uint i; #if defined(__i386__) || defined(__386__) softpipe->use_sse = getenv("GALLIUM_SSE") != NULL; @@ -355,5 +380,10 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + softpipe->cbuf_cache[i] = sp_create_tile_cache(); + softpipe->zbuf_cache = sp_create_tile_cache(); + softpipe->sbuf_cache_sep = sp_create_tile_cache(); + return &softpipe->pipe; } diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 4f429e8139..ea05f80d59 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -44,6 +44,7 @@ struct softpipe_surface; struct softpipe_winsys; struct draw_context; struct draw_stage; +struct softpipe_tile_cache; #define SP_NEW_VIEWPORT 0x1 @@ -155,6 +156,13 @@ struct softpipe_context { struct pipe_surface *cbuf; /**< current color buffer (one of cbufs) */ + struct softpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS]; + struct softpipe_tile_cache *zbuf_cache; + /** Stencil buffer cache, for stencil separate from Z */ + struct softpipe_tile_cache *sbuf_cache_sep; + /** This either points to zbuf_cache or sbuf_cache_sep */ + struct softpipe_tile_cache *sbuf_cache; + int use_sse : 1; }; diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 405659fb46..64a4fbe333 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -113,7 +113,7 @@ softpipe_draw_elements(struct pipe_context *pipe, softpipe_update_derived( sp ); softpipe_map_surfaces(sp); - + softpipe_map_texture_surfaces(sp); softpipe_map_constant_buffers(sp); /* @@ -184,7 +184,8 @@ softpipe_draw_elements(struct pipe_context *pipe, } - softpipe_unmap_surfaces(sp); + /* Note: leave drawing surfaces mapped */ + softpipe_unmap_texture_surfaces(sp); softpipe_unmap_constant_buffers(sp); return TRUE; diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index d4eada8228..f2186dbb65 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -54,10 +54,14 @@ softpipe_flush( struct pipe_context *pipe, * - flush the render cache */ - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.cbufs[i]); - if (sps) - sp_flush_tile_cache(sps); - } + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + if (softpipe->cbuf_cache[i]) + sp_flush_tile_cache(softpipe->cbuf_cache[i]); + + if (softpipe->zbuf_cache) + sp_flush_tile_cache(softpipe->zbuf_cache); + + if (softpipe->sbuf_cache) + sp_flush_tile_cache(softpipe->sbuf_cache); } diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 43294e4cd4..5787b89588 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -101,14 +101,13 @@ static void logicop_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); float dest[4][QUAD_SIZE]; ubyte src[4][4], dst[4][4], res[4][4]; uint *src4 = (uint *) src; uint *dst4 = (uint *) dst; uint *res4 = (uint *) res; struct softpipe_cached_tile * - tile = sp_get_cached_tile(sps, quad->x0, quad->y0); + tile = sp_get_cached_tile(softpipe->cbuf_cache[0], quad->x0, quad->y0); uint i, j; /* get/swizzle dest colors */ @@ -220,12 +219,11 @@ static void blend_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); static const float zero[4] = { 0, 0, 0, 0 }; static const float one[4] = { 1, 1, 1, 1 }; float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; struct softpipe_cached_tile * - tile = sp_get_cached_tile(sps, quad->x0, quad->y0); + tile = sp_get_cached_tile(softpipe->cbuf_cache[0], quad->x0, quad->y0); uint i, j; if (softpipe->blend->logicop_enable) { diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index aa69f53580..3c0196dd5d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -47,10 +47,9 @@ static void colormask_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); float dest[4][QUAD_SIZE]; struct softpipe_cached_tile * - tile = sp_get_cached_tile(sps, quad->x0, quad->y0); + tile = sp_get_cached_tile(softpipe->cbuf_cache[0], quad->x0, quad->y0); uint i, j; /* get/swizzle dest colors */ diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index efd08b981c..29231322b8 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -39,55 +39,86 @@ * Do depth testing for a quad. * Not static since it's used by the stencil code. */ + +/* + * To increase efficiency, we should probably have multiple versions + * of this function that are specifically for Z16, Z32 and FP Z buffers. + * Try to effectively do that with codegen... + */ + void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); + const uint format = sps->surface.format; unsigned bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ unsigned qzzzz[QUAD_SIZE]; /**< Z values from the quad */ unsigned zmask = 0; unsigned j; - float scale; -#if 0 - struct cached_tile *tile = sp_get_cached_tile(softpipe, quad->x0, quad->y0); -#endif + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe->zbuf_cache, quad->x0, quad->y0); assert(sps); /* shouldn't get here if there's no zbuffer */ /* - * To increase efficiency, we should probably have multiple versions - * of this function that are specifically for Z16, Z32 and FP Z buffers. - * Try to effectively do that with codegen... - */ - if (sps->surface.format == PIPE_FORMAT_U_Z16) - scale = 65535.0; - else if (sps->surface.format == PIPE_FORMAT_S8_Z24) - scale = (float) ((1 << 24) - 1); - else - assert(0); /* XXX fix this someday */ - - /* - * Convert quad's float depth values to int depth values. + * Convert quad's float depth values to int depth values (qzzzz). * If the Z buffer stores integer values, we _have_ to do the depth * compares with integers (not floats). Otherwise, the float->int->float * conversion of Z values (which isn't an identity function) will cause * Z-fighting errors. + * + * Also, get the zbuffer values (bzzzz) from the cached tile. */ - for (j = 0; j < QUAD_SIZE; j++) { - qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); - } - -#if 0 - for (j = 0; j < 4; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - bzzzz[j] = tile->depth[y][x]; + switch (format) { + case PIPE_FORMAT_U_Z16: + { + float scale = 65535.0; + + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); + } + + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + bzzzz[j] = tile->data.depth16[y][x]; + } + } + break; + case PIPE_FORMAT_U_Z32: + { + double scale = (double) (uint) ~0UL; + + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); + } + + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + bzzzz[j] = tile->data.depth32[y][x]; + } + } + break; + case PIPE_FORMAT_S8_Z24: + { + float scale = (float) ((1 << 24) - 1); + + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); + } + + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + bzzzz[j] = tile->data.depth32[y][x] & 0xffffff; + } + } + break; + default: + assert(0); } -#else - /* get zquad from zbuffer */ - sps->read_quad_z(sps, quad->x0, quad->y0, bzzzz); -#endif switch (softpipe->depth_stencil->depth.func) { case PIPE_FUNC_NEVER: @@ -151,16 +182,34 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } } -#if 1 - /* write updated zquad to zbuffer */ - sps->write_quad_z(sps, quad->x0, quad->y0, bzzzz); -#else - for (j = 0; j < 4; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - tile->depth[y][x] = bzzzz[j]; + /* put updated Z values back into cached tile */ + switch (format) { + case PIPE_FORMAT_U_Z16: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + tile->data.depth16[y][x] = bzzzz[j]; + } + break; + case PIPE_FORMAT_U_Z32: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + tile->data.depth32[y][x] = bzzzz[j]; + } + break; + case PIPE_FORMAT_S8_Z24: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + uint s8z24 = tile->data.depth32[y][x]; + s8z24 = (s8z24 & 0xff000000) | bzzzz[j]; + tile->data.depth32[y][x] = s8z24; + } + break; + default: + assert(0); } -#endif } } diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index 01ee06a69c..e86f42be46 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -42,9 +42,8 @@ static void output_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_surface *sps = softpipe_surface(softpipe->cbuf); struct softpipe_cached_tile *tile - = sp_get_cached_tile(sps, quad->x0, quad->y0); + = sp_get_cached_tile(softpipe->cbuf_cache[0], quad->x0, quad->y0); /* in-tile pos: */ const int itx = quad->x0 % TILE_SIZE; const int ity = quad->y0 % TILE_SIZE; diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index bf72bb23cd..4a3823d646 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -7,6 +7,7 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" +#include "sp_tile_cache.h" #include "sp_quad.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" @@ -200,10 +201,13 @@ static void stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_surface *s_surf = softpipe_surface(softpipe->framebuffer.sbuf); + struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.sbuf); unsigned func, zFailOp, zPassOp, failOp; ubyte ref, wrtMask, valMask; ubyte stencilVals[QUAD_SIZE]; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe->sbuf_cache, quad->x0, quad->y0); + uint j; /* choose front or back face function, operator, etc */ /* XXX we could do these initializations once per primitive */ @@ -226,8 +230,27 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) valMask = softpipe->depth_stencil->stencil.value_mask[0]; } - assert(s_surf); /* shouldn't get here if there's no stencil buffer */ - s_surf->read_quad_stencil(s_surf, quad->x0, quad->y0, stencilVals); + assert(sps); /* shouldn't get here if there's no stencil buffer */ + + /* get stencil values from cached tile */ + switch (sps->surface.format) { + case PIPE_FORMAT_S8_Z24: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + stencilVals[j] = tile->data.depth32[y][x] >> 24; + } + break; + case PIPE_FORMAT_U_S8: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + stencilVals[j] = tile->data.stencil8[y][x]; + } + break; + default: + assert(0); + } /* do the stencil test first */ { @@ -242,7 +265,6 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) } if (quad->mask) { - /* now the pixels that passed the stencil test are depth tested */ if (softpipe->depth_stencil->depth.enabled) { const unsigned origMask = quad->mask; @@ -267,7 +289,27 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) } - s_surf->write_quad_stencil(s_surf, quad->x0, quad->y0, stencilVals); + /* put new stencil values into cached tile */ + switch (sps->surface.format) { + case PIPE_FORMAT_S8_Z24: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + uint s8z24 = tile->data.depth32[y][x]; + s8z24 = (stencilVals[j] << 24) | (s8z24 & 0xffffff); + tile->data.depth32[y][x] = s8z24; + } + break; + case PIPE_FORMAT_U_S8: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + tile->data.stencil8[y][x] = stencilVals[j]; + } + break; + default: + assert(0); + } if (quad->mask) qs->next->run(qs->next, quad); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index f9061e86e5..c194f0ea0d 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -141,4 +141,10 @@ softpipe_map_surfaces(struct softpipe_context *sp); void softpipe_unmap_surfaces(struct softpipe_context *sp); +void +softpipe_map_texture_surfaces(struct softpipe_context *sp); + +void +softpipe_unmap_texture_surfaces(struct softpipe_context *sp); + #endif diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index a534ffb2c2..cd1e75c563 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -30,20 +30,100 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_surface.h" +#include "sp_tile_cache.h" -/* +/** * XXX this might get moved someday + * Set the framebuffer surface info: color buffers, zbuffer, stencil buffer. + * Here, we map the surfaces and update the tile cache to point to the new + * surfaces. */ void softpipe_set_framebuffer_state(struct pipe_context *pipe, const struct pipe_framebuffer_state *fb) { - struct softpipe_context *softpipe = softpipe_context(pipe); + struct softpipe_context *sp = softpipe_context(pipe); + struct softpipe_surface *sps; + uint i; + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + /* check if changing cbuf */ + if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) { + /* flush old */ + sp_flush_tile_cache(sp->cbuf_cache[i]); + /* unmap old */ + sps = softpipe_surface(sp->framebuffer.cbufs[i]); + if (sps && sps->surface.region) + pipe->region_unmap(pipe, sps->surface.region); + /* map new */ + sps = softpipe_surface(fb->cbufs[i]); + if (sps) + pipe->region_map(pipe, sps->surface.region); + /* assign new */ + sp->framebuffer.cbufs[i] = fb->cbufs[i]; + + /* update cache */ + sp_tile_cache_set_surface(sp->cbuf_cache[i], sps); + } + } + + sp->framebuffer.num_cbufs = fb->num_cbufs; + + /* zbuf changing? */ + if (sp->framebuffer.zbuf != fb->zbuf) { + /* flush old */ + sp_flush_tile_cache(sp->zbuf_cache); + /* unmap old */ + sps = softpipe_surface(sp->framebuffer.zbuf); + if (sps && sps->surface.region) + pipe->region_unmap(pipe, sps->surface.region); + if (sp->framebuffer.sbuf == sp->framebuffer.zbuf) { + /* combined z/stencil */ + sp->framebuffer.sbuf = NULL; + } + /* map new */ + sps = softpipe_surface(fb->zbuf); + if (sps) + pipe->region_map(pipe, sps->surface.region); + /* assign new */ + sp->framebuffer.zbuf = fb->zbuf; + + /* update cache */ + sp_tile_cache_set_surface(sp->zbuf_cache, sps); + } + + /* XXX combined depth/stencil here */ + + /* sbuf changing? */ + if (sp->framebuffer.sbuf != fb->sbuf) { + /* flush old */ + sp_flush_tile_cache(sp->sbuf_cache_sep); + /* unmap old */ + sps = softpipe_surface(sp->framebuffer.sbuf); + if (sps && sps->surface.region) + pipe->region_unmap(pipe, sps->surface.region); + /* map new */ + sps = softpipe_surface(fb->sbuf); + if (sps && fb->sbuf != fb->zbuf) + pipe->region_map(pipe, sps->surface.region); + /* assign new */ + sp->framebuffer.sbuf = fb->sbuf; - softpipe->framebuffer = *fb; /* struct copy */ + /* update cache */ + if (fb->sbuf != fb->zbuf) { + /* separate stencil buf */ + sp->sbuf_cache = sp->sbuf_cache_sep; + sp_tile_cache_set_surface(sp->sbuf_cache, sps); + } + else { + /* combined depth/stencil */ + sp->sbuf_cache = sp->zbuf_cache; + sp_tile_cache_set_surface(sp->sbuf_cache, sps); + } + } - softpipe->dirty |= SP_NEW_FRAMEBUFFER; + sp->dirty |= SP_NEW_FRAMEBUFFER; } diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 8aad5f6c13..fb8c8d0826 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -61,6 +61,10 @@ #else #define CLIP_TILE \ do { \ + if (x >= ps->width) \ + return; \ + if (y >= ps->height) \ + return; \ if (x + w > ps->width) \ w = ps->width - x; \ if (y + h > ps->height) \ @@ -907,13 +911,16 @@ put_tile_raw32(struct pipe_surface *ps, unsigned i; unsigned w0 = w; + assert(w < 1000); assert(ps->region->map); assert(ps->format == PIPE_FORMAT_S8_Z24 || ps->format == PIPE_FORMAT_U_Z32); + assert(w < 1000); CLIP_TILE; for (i = 0; i < h; i++) { + assert(w < 1000); memcpy(dst, pSrc, w * sizeof(uint)); dst += ps->region->pitch; pSrc += w0; @@ -980,8 +987,6 @@ put_tile_raw16(struct pipe_surface *ps, void softpipe_init_surface_funcs(struct softpipe_surface *sps) { - sps->tc = sp_create_tile_cache(); - assert(sps->surface.format); switch (sps->surface.format) { diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 431303553f..522f7612ab 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -46,8 +46,6 @@ struct softpipe_tile_cache; struct softpipe_surface { struct pipe_surface surface; - struct softpipe_tile_cache *tc; - /** * Functions for read/writing surface data */ diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 2ecdfc89e4..74bd4a3d11 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -48,6 +48,7 @@ struct softpipe_tile_cache { + struct softpipe_surface *surface; /**< the surface we're caching */ struct softpipe_cached_tile entries[NUM_ENTRIES]; uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; }; @@ -115,45 +116,64 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc) } +void +sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, + struct softpipe_surface *sps) +{ + tc->surface = sps; +} + + + void -sp_flush_tile_cache(struct softpipe_surface *sps) +sp_flush_tile_cache(struct softpipe_tile_cache *tc) { - struct softpipe_tile_cache *tc = sps->tc; - /* - struct softpipe_surface *zsurf = softpipe_surface(softpipe->zbuf); - */ + struct pipe_surface *ps = &tc->surface->surface; + boolean is_depth_stencil; int inuse = 0, pos; + if (!ps || !ps->region || !ps->region->map) + return; + + is_depth_stencil = (ps->format == PIPE_FORMAT_S8_Z24 || + ps->format == PIPE_FORMAT_U_Z16 || + ps->format == PIPE_FORMAT_U_Z32 || + ps->format == PIPE_FORMAT_U_S8); + for (pos = 0; pos < NUM_ENTRIES; pos++) { struct softpipe_cached_tile *tile = tc->entries + pos; if (tile->x >= 0) { - sps->surface.put_tile(&sps->surface, - tile->x, - tile->y, - TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - /* - sps->surface.put_tile(&zsurf->surface, - tile->x, - tile->y, - TILE_SIZE, TILE_SIZE, - (float *) tile->depth); - */ + if (is_depth_stencil) { + ps->put_tile_raw(ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32); + } + else { + ps->put_tile(ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } tile->x = tile->y = -1; /* mark as empty */ inuse++; } } - - /*printf("flushed tiles in use: %d\n", inuse);*/ + /* + printf("flushed tiles in use: %d\n", inuse); + */ } struct softpipe_cached_tile * -sp_get_cached_tile(struct softpipe_surface *sps, int x, int y) +sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y) { - struct softpipe_tile_cache *tc = sps->tc; + struct pipe_surface *ps = &tc->surface->surface; + boolean is_depth_stencil + = (ps->format == PIPE_FORMAT_S8_Z24 || + ps->format == PIPE_FORMAT_U_Z16 || + ps->format == PIPE_FORMAT_U_Z32 || + ps->format == PIPE_FORMAT_U_S8); /* tile pos in framebuffer: */ const int tile_x = x & ~(TILE_SIZE - 1); @@ -176,13 +196,22 @@ sp_get_cached_tile(struct softpipe_surface *sps, int x, int y) if (tile->x != -1) { /* put dirty tile back in framebuffer */ - sps->surface.put_tile(&sps->surface, tile->x, tile->y, - TILE_SIZE, TILE_SIZE, (float *) tile->data.color); + if (is_depth_stencil) { + ps->put_tile_raw(ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32); + } + else { + ps->put_tile(ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } } - if (is_clear_flag_set(tc->clear_flags, x, y)) { - printf("clear tile\n"); + if (0/*is_clear_flag_set(tc->clear_flags, x, y)*/) { + /* don't get tile from framebuffer, just clear it */ #if 0 + printf("clear tile\n"); uint i, j; for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { @@ -192,14 +221,24 @@ sp_get_cached_tile(struct softpipe_surface *sps, int x, int y) tile->data.color[i][j][3] = 0.5; } } +#else + (void) is_clear_flag_set; #endif memset(tile->data.color, 0, sizeof(tile->data.color)); clear_clear_flag(tc->clear_flags, x, y); } else { /* get new tile from framebuffer */ - sps->surface.get_tile(&sps->surface, tile_x, tile_y, - TILE_SIZE, TILE_SIZE, (float *) tile->data.color); + if (is_depth_stencil) { + ps->get_tile_raw(ps, + tile_x, tile_y, TILE_SIZE, TILE_SIZE, + tile->data.depth32); + } + else { + ps->get_tile(ps, + tile_x, tile_y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } } tile->x = tile_x; @@ -211,8 +250,8 @@ sp_get_cached_tile(struct softpipe_surface *sps, int x, int y) void -sp_clear_tile_cache(struct softpipe_surface *sps, unsigned clearval) +sp_clear_tile_cache(struct softpipe_tile_cache *tc, unsigned clearval) { (void) clearval; /* XXX use this */ - memset(sps->tc->clear_flags, 255, sizeof(sps->tc->clear_flags)); + memset(tc->clear_flags, 255, sizeof(tc->clear_flags)); } diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index 4248361f99..80bcac6904 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -36,6 +36,9 @@ struct softpipe_context; struct softpipe_tile_cache; +/** + * Cache tile size (width and height). This needs to be a power of two. + */ #define TILE_SIZE 64 @@ -45,7 +48,9 @@ struct softpipe_cached_tile int x, y; /** pos of tile in window coords */ union { float color[TILE_SIZE][TILE_SIZE][4]; - uint depth[TILE_SIZE][TILE_SIZE]; + uint depth32[TILE_SIZE][TILE_SIZE]; + ushort depth16[TILE_SIZE][TILE_SIZE]; + ubyte stencil8[TILE_SIZE][TILE_SIZE]; } data; }; @@ -57,13 +62,17 @@ extern void sp_destroy_tile_cache(struct softpipe_tile_cache *tc); extern void -sp_flush_tile_cache(struct softpipe_surface *sps); +sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, + struct softpipe_surface *sps); extern void -sp_clear_tile_cache(struct softpipe_surface *sps, unsigned clearval); +sp_flush_tile_cache(struct softpipe_tile_cache *tc); + +extern void +sp_clear_tile_cache(struct softpipe_tile_cache *tc, unsigned clearval); extern struct softpipe_cached_tile * -sp_get_cached_tile(struct softpipe_surface *sps, int x, int y); +sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y); #endif /* SP_TILE_CACHE_H */ -- cgit v1.2.3 From 46e2d2bb0a7423122412e3f119fdd89fedd9aef6 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 15:19:14 -0600 Subject: use combined depth/stencil buffer when possible --- src/mesa/drivers/x11/xm_api.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index ff83dab075..187663e66c 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -430,16 +430,26 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, b->swAlpha = GL_FALSE; } - if (vis->mesa_visual.depthBits > 0) { + if (vis->mesa_visual.depthBits > 0 && + vis->mesa_visual.stencilBits > 0) { + /* combined depth/stencil */ struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32); + = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_DEPTH, rb); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_STENCIL, rb); } + else { + if (vis->mesa_visual.depthBits > 0) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_DEPTH, rb); + } - if (vis->mesa_visual.stencilBits > 0) { - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT); - _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_STENCIL, rb); + if (vis->mesa_visual.stencilBits > 0) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_STENCIL, rb); + } } if (vis->mesa_visual.accumRedBits > 0) { -- cgit v1.2.3 From fd3876e9e3fb17df61dbf45d400796d682a8fbe0 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 15:20:17 -0600 Subject: renderbuffer tweaks in update_framebuffer_state() --- src/mesa/state_tracker/st_atom_framebuffer.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index a6caf3f5ff..e776c9112d 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -45,6 +45,7 @@ static void update_framebuffer_state( struct st_context *st ) { struct pipe_framebuffer_state framebuffer; + struct gl_framebuffer *fb = st->ctx->DrawBuffer; struct st_renderbuffer *strb; GLuint i; @@ -53,21 +54,21 @@ update_framebuffer_state( struct st_context *st ) /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state * to determine which surfaces to draw to */ - framebuffer.num_cbufs = st->ctx->DrawBuffer->_NumColorDrawBuffers[0]; + framebuffer.num_cbufs = fb->_NumColorDrawBuffers[0]; for (i = 0; i < framebuffer.num_cbufs; i++) { - strb = st_renderbuffer(st->ctx->DrawBuffer->_ColorDrawBuffers[0][i]); + strb = st_renderbuffer(fb->_ColorDrawBuffers[0][i]); assert(strb->surface); framebuffer.cbufs[i] = strb->surface; } - strb = st_renderbuffer(st->ctx->DrawBuffer->_DepthBuffer); + strb = st_renderbuffer(fb->Attachment[BUFFER_DEPTH].Renderbuffer); if (strb) { strb = st_renderbuffer(strb->Base.Wrapped); assert(strb->surface); framebuffer.zbuf = strb->surface; } - strb = st_renderbuffer(st->ctx->DrawBuffer->_StencilBuffer); + strb = st_renderbuffer(fb->Attachment[BUFFER_STENCIL].Renderbuffer); if (strb) { strb = st_renderbuffer(strb->Base.Wrapped); assert(strb->surface); -- cgit v1.2.3 From a1633c07164e8bc9630f1fa77769e7ceee585a59 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 15:20:46 -0600 Subject: unmap regions when reallocating renderbuffer storage --- src/mesa/state_tracker/st_cb_fbo.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index e1cc7d98bd..64a64fd6b9 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -92,6 +92,10 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, /* free old region */ if (strb->surface->region) { + /* loop here since mapping is refcounted */ + struct pipe_region *r = strb->surface->region; + while (r->map) + pipe->region_unmap(pipe, r); pipe->region_release(pipe, &strb->surface->region); } -- cgit v1.2.3 From bb3d61551c8bb3fc60ca48bd2394d4ba7ba004d2 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 15:21:02 -0600 Subject: flush the pipe before accum ops --- src/mesa/state_tracker/st_cb_accum.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 192bdbab63..ddf9dc2adf 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -241,6 +241,9 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value) const GLint width = ctx->DrawBuffer->_Xmax - xpos; const GLint height = ctx->DrawBuffer->_Ymax - ypos; + /* make sure color bufs aren't cached */ + pipe->flush(pipe, 0); + switch (op) { case GL_ADD: if (value != 0.0F) { -- cgit v1.2.3 From 7c8b2f7ce33a672a90de8000bcbf15ec764e9d85 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 15:45:24 -0600 Subject: In region_unmap(), check if region is mapped before decrementing refcount. --- src/mesa/pipe/i915simple/i915_regions.c | 11 +++++++---- src/mesa/pipe/softpipe/sp_region.c | 11 +++++++---- 2 files changed, 14 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index cff90ab5dd..577a6adfd8 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -59,10 +59,13 @@ i915_region_unmap(struct pipe_context *pipe, struct pipe_region *region) { struct i915_context *i915 = i915_context( pipe ); - if (!--region->map_refcount) { - i915->pipe.winsys->buffer_unmap( i915->pipe.winsys, - region->buffer ); - region->map = NULL; + if (region->map_refcount > 0) { + assert(region->map); + if (!--region->map_refcount) { + i915->pipe.winsys->buffer_unmap( i915->pipe.winsys, + region->buffer ); + region->map = NULL; + } } } diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 0f4d2f0575..4317a9ea1b 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -69,10 +69,13 @@ sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) { struct softpipe_context *sp = softpipe_context( pipe ); - if (!--region->map_refcount) { - sp->pipe.winsys->buffer_unmap( sp->pipe.winsys, - region->buffer ); - region->map = NULL; + if (region->map_refcount > 0) { + assert(region->map); + if (!--region->map_refcount) { + sp->pipe.winsys->buffer_unmap( sp->pipe.winsys, + region->buffer ); + region->map = NULL; + } } } -- cgit v1.2.3 From f9aa75718708076e50033287fde993799878ecf6 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 15:52:36 -0600 Subject: Call softpipe_unmap_surfaces() in softpipe_flush(). This fixes a DRM BO failure upon swapbuffers caused by the color buffer still being mapped. This is a bit heavy handed since we don't always need to unmap buffers when flushing. Need to pass a flag to flush() or design a new function. --- src/mesa/pipe/softpipe/sp_flush.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index f2186dbb65..5eb4d3367d 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -63,5 +63,13 @@ softpipe_flush( struct pipe_context *pipe, if (softpipe->sbuf_cache) sp_flush_tile_cache(softpipe->sbuf_cache); + + /* Need this call for hardware buffers before swapbuffers. + * + * there should probably be another/different flush-type function + * that's called before swapbuffers because we don't always want + * to unmap surfaces when flushing. + */ + softpipe_unmap_surfaces(softpipe); } -- cgit v1.2.3 From 03145d864ce21094592ae847fbe8da57c419374e Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 15:52:59 -0600 Subject: init sp->sbuf_cache to avoid possible segfault --- src/mesa/pipe/softpipe/sp_context.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 87eaf6fb54..10c53a257b 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -384,6 +384,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->cbuf_cache[i] = sp_create_tile_cache(); softpipe->zbuf_cache = sp_create_tile_cache(); softpipe->sbuf_cache_sep = sp_create_tile_cache(); + softpipe->sbuf_cache = softpipe->sbuf_cache_sep; /* initial value */ return &softpipe->pipe; } -- cgit v1.2.3 From 49848208cf1faba49a83fb8872a29f6fa910127d Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 20 Oct 2007 16:09:17 -0600 Subject: Remove obsolete read/write_quad() functions --- src/mesa/drivers/x11/xm_surface.c | 113 --------- src/mesa/pipe/softpipe/sp_surface.c | 493 ++++-------------------------------- src/mesa/pipe/softpipe/sp_surface.h | 39 +-- 3 files changed, 54 insertions(+), 591 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 8d2c2ca5f3..562ec76636 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -76,108 +76,6 @@ xmesa_rb(struct softpipe_surface *sps) #define FLIP(Y) Y = xrb->St.Base.Height - (Y) - 1; -/** - * quad reading/writing - * These functions are just wrappers around the existing renderbuffer - * functions. - * Note that Y=0 is the top of the surface. - */ - -static void -read_quad_f(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS]) -{ - struct xmesa_renderbuffer *xrb = xmesa_rb(sps); - GLubyte temp[16]; - GLfloat *dst = (GLfloat *) rgba; - GLuint i; - GET_CURRENT_CONTEXT(ctx); - FLIP(y); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, temp); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y - 1, temp + 8); - for (i = 0; i < 16; i++) { - dst[i] = UBYTE_TO_FLOAT(temp[i]); - } -} - -static void -read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) -{ - struct xmesa_renderbuffer *xrb = xmesa_rb(sps); - GLubyte temp[16]; - GLfloat *dst = (GLfloat *) rrrr; - GLuint i, j; - GET_CURRENT_CONTEXT(ctx); - FLIP(y); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, temp); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y - 1, temp + 8); - for (i = 0; i < 4; i++) { - for (j = 0; j < 4; j++) { - dst[j * 4 + i] = UBYTE_TO_FLOAT(temp[i * 4 + j]); - } - } -} - -static void -write_quad_f(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rgba)[NUM_CHANNELS]) -{ - struct xmesa_renderbuffer *xrb = xmesa_rb(sps); - GLubyte temp[16]; - const GLfloat *src = (const GLfloat *) rgba; - GLuint i; - GET_CURRENT_CONTEXT(ctx); - FLIP(y); - for (i = 0; i < 16; i++) { - UNCLAMPED_FLOAT_TO_UBYTE(temp[i], src[i]); - } - xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y, temp, NULL); - xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y - 1, temp + 8, NULL); -} - -static void -write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) -{ - struct xmesa_renderbuffer *xrb = xmesa_rb(sps); - GLubyte temp[16]; - const GLfloat *src = (const GLfloat *) rrrr; - GLuint i, j; - GET_CURRENT_CONTEXT(ctx); - FLIP(y); - for (i = 0; i < 4; i++) { - for (j = 0; j < 4; j++) { - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + i], src[i * 4 + j]); - } - } - xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y, temp, NULL); - xrb->St.Base.PutRow(ctx, &xrb->St.Base, 2, x, y - 1, temp + 8, NULL); -} - -static void -read_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS]) -{ - struct xmesa_renderbuffer *xrb = xmesa_rb(sps); - GET_CURRENT_CONTEXT(ctx); - FLIP(y); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, rgba); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y - 1, rgba + 2); -} - -static void -write_quad_ub(struct softpipe_surface *sps, GLint x, GLint y, - GLubyte (*rgba)[NUM_CHANNELS]) -{ - struct xmesa_renderbuffer *xrb = xmesa_rb(sps); - GET_CURRENT_CONTEXT(ctx); - FLIP(y); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y, rgba); - xrb->St.Base.GetRow(ctx, &xrb->St.Base, 2, x, y - 1, rgba + 2); -} - - static void get_tile(struct pipe_surface *ps, GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) @@ -243,21 +141,10 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) switch (pipeFormat) { case PIPE_FORMAT_U_A8_R8_G8_B8: - xms->surface.read_quad_f_swz = read_quad_f; - xms->surface.read_quad_f = read_quad_f; - xms->surface.read_quad_f_swz = read_quad_f_swz; - xms->surface.read_quad_ub = read_quad_ub; - xms->surface.write_quad_f = write_quad_f; - xms->surface.write_quad_f_swz = write_quad_f_swz; - xms->surface.write_quad_ub = write_quad_ub; xms->surface.surface.get_tile = get_tile; xms->surface.surface.put_tile = put_tile; break; case PIPE_FORMAT_S8_Z24: - /* - xms->surface.read_quad_z = 1; - xms->surface.write_quad_z = 1; - */ break; default: abort(); diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index fb8c8d0826..40f045800f 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -34,12 +34,6 @@ /** * Softpipe surface functions. - * Basically, create surface of a particular type, then plug in default - * read/write_quad and get/put_tile() functions. - * Note that these quad funcs assume the buffer/region is in a linear - * layout with Y=0=top. - * If we had swizzled/AOS buffers the read/write quad functions could be - * simplified a lot.... */ @@ -75,56 +69,6 @@ /*** PIPE_FORMAT_U_A8_R8_G8_B8 ***/ -static void -a8r8g8b8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - const unsigned *src - = ((const unsigned *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); -#if 0 - assert(x < (int) sps->surface.width - 1); - assert(y < (int) sps->surface.height - 1); -#endif - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - const unsigned p = src[j]; - rrrr[0][i * 2 + j] = UBYTE_TO_FLOAT((p >> 16) & 0xff); /*R*/ - rrrr[1][i * 2 + j] = UBYTE_TO_FLOAT((p >> 8) & 0xff); /*G*/ - rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT((p ) & 0xff); /*B*/ - rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT((p >> 24) & 0xff); /*A*/ - } - src += sps->surface.region->pitch; - } -} - -static void -a8r8g8b8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - unsigned *dst - = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - ubyte r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/ - UNCLAMPED_FLOAT_TO_UBYTE(g, rrrr[1][i * 2 + j]); /*G*/ - UNCLAMPED_FLOAT_TO_UBYTE(b, rrrr[2][i * 2 + j]); /*B*/ - UNCLAMPED_FLOAT_TO_UBYTE(a, rrrr[3][i * 2 + j]); /*A*/ - dst[j] = (a << 24) | (r << 16) | (g << 8) | b; - } - dst += sps->surface.region->pitch; - } -} - static void a8r8g8b8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) @@ -217,42 +161,6 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_U_Z16 ***/ -static void -z16_read_quad_z(struct softpipe_surface *sps, - int x, int y, unsigned zzzz[QUAD_SIZE]) -{ - const ushort *src - = ((const ushort *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_Z16); - - /* converting ushort to unsigned: */ - zzzz[0] = src[0]; - zzzz[1] = src[1]; - src += sps->surface.region->pitch; - zzzz[2] = src[0]; - zzzz[3] = src[1]; -} - -static void -z16_write_quad_z(struct softpipe_surface *sps, - int x, int y, const unsigned zzzz[QUAD_SIZE]) -{ - ushort *dst - = ((ushort *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_Z16); - - /* converting unsigned to ushort: */ - dst[0] = zzzz[0]; - dst[1] = zzzz[1]; - dst += sps->surface.region->pitch; - dst[0] = zzzz[2]; - dst[1] = zzzz[3]; -} - /** * Return as floats in [0,1]. */ @@ -286,51 +194,6 @@ z16_get_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_U_L8 ***/ -static void -l8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - const ubyte *src - = ((const ubyte *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_L8); - assert(x < (int) sps->surface.width - 1); - assert(y < (int) sps->surface.height - 1); - - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - rrrr[0][i * 2 + j] = - rrrr[1][i * 2 + j] = - rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT(src[j]); - rrrr[3][i * 2 + j] = 1.0F; - } - src += sps->surface.region->pitch; - } -} - -static void -l8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - ubyte *dst - = ((ubyte *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_L8); - - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - ubyte r; - UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/ - dst[j] = r; - } - dst += sps->surface.region->pitch; - } -} - static void l8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) @@ -362,51 +225,6 @@ l8_get_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_U_A8 ***/ -static void -a8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - const ubyte *src - = ((const ubyte *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8); - assert(x < (int) sps->surface.width - 1); - assert(y < (int) sps->surface.height - 1); - - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - rrrr[0][i * 2 + j] = - rrrr[1][i * 2 + j] = - rrrr[2][i * 2 + j] = 0.0F; - rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT(src[j]); - } - src += sps->surface.region->pitch; - } -} - -static void -a8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - ubyte *dst - = ((ubyte *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8); - - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - ubyte r; - UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[3][i * 2 + j]); /*A*/ - dst[j] = r; - } - dst += sps->surface.region->pitch; - } -} - static void a8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) @@ -507,51 +325,6 @@ r16g16b16a16_put_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_U_I8 ***/ -static void -i8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - const ubyte *src - = ((const ubyte *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_I8); - assert(x < (int) sps->surface.width - 1); - assert(y < (int) sps->surface.height - 1); - - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - rrrr[0][i * 2 + j] = - rrrr[1][i * 2 + j] = - rrrr[2][i * 2 + j] = - rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT(src[j]); - } - src += sps->surface.region->pitch; - } -} - -static void -i8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - ubyte *dst - = ((ubyte *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_I8); - - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - ubyte r; - UNCLAMPED_FLOAT_TO_UBYTE(r, rrrr[0][i * 2 + j]); /*R*/ - dst[j] = r; - } - dst += sps->surface.region->pitch; - } -} - static void i8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) @@ -583,53 +356,6 @@ i8_get_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_U_A8_L8 ***/ -static void -a8_l8_read_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - const ushort *src - = ((const ushort *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_L8); - assert(x < (int) sps->surface.width - 1); - assert(y < (int) sps->surface.height - 1); - - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - const ushort p = src[j]; - rrrr[0][i * 2 + j] = - rrrr[1][i * 2 + j] = - rrrr[2][i * 2 + j] = UBYTE_TO_FLOAT(p >> 8); - rrrr[3][i * 2 + j] = UBYTE_TO_FLOAT(p & 0xff); - } - src += sps->surface.region->pitch; - } -} - -static void -a8_l8_write_quad_f_swz(struct softpipe_surface *sps, int x, int y, - float (*rrrr)[QUAD_SIZE]) -{ - ushort *dst - = ((ushort *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - unsigned i, j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_L8); - - for (i = 0; i < 2; i++) { /* loop over pixel row */ - for (j = 0; j < 2; j++) { /* loop over pixel column */ - ubyte l, a; - UNCLAMPED_FLOAT_TO_UBYTE(l, rrrr[0][i * 2 + j]); /*R*/ - UNCLAMPED_FLOAT_TO_UBYTE(a, rrrr[3][i * 2 + j]); /*A*/ - dst[j] = (l << 8) | a; - } - dst += sps->surface.region->pitch; - } -} - static void a8_l8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) @@ -664,40 +390,6 @@ a8_l8_get_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_U_Z32 ***/ -static void -z32_read_quad_z(struct softpipe_surface *sps, - int x, int y, unsigned zzzz[QUAD_SIZE]) -{ - const unsigned *src - = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_Z32); - - zzzz[0] = src[0]; - zzzz[1] = src[1]; - src += sps->surface.region->pitch; - zzzz[2] = src[0]; - zzzz[3] = src[1]; -} - -static void -z32_write_quad_z(struct softpipe_surface *sps, - int x, int y, const unsigned zzzz[QUAD_SIZE]) -{ - unsigned *dst - = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_Z32); - - dst[0] = zzzz[0]; - dst[1] = zzzz[1]; - dst += sps->surface.region->pitch; - dst[0] = zzzz[2]; - dst[1] = zzzz[3]; -} - /** * Return as floats in [0,1]. */ @@ -729,80 +421,6 @@ z32_get_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_S8_Z24 ***/ -static void -s8z24_read_quad_z(struct softpipe_surface *sps, - int x, int y, unsigned zzzz[QUAD_SIZE]) -{ - static const unsigned mask = 0x00ffffff; - const unsigned *src - = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - /* extract lower three bytes */ - zzzz[0] = src[0] & mask; - zzzz[1] = src[1] & mask; - src += sps->surface.region->pitch; - zzzz[2] = src[0] & mask; - zzzz[3] = src[1] & mask; -} - -static void -s8z24_write_quad_z(struct softpipe_surface *sps, - int x, int y, const unsigned zzzz[QUAD_SIZE]) -{ - static const unsigned mask = 0xff000000; - unsigned *dst - = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - assert(zzzz[0] <= 0xffffff); - - dst[0] = (dst[0] & mask) | zzzz[0]; - dst[1] = (dst[1] & mask) | zzzz[1]; - dst += sps->surface.region->pitch; - dst[0] = (dst[0] & mask) | zzzz[2]; - dst[1] = (dst[1] & mask) | zzzz[3]; -} - -static void -s8z24_read_quad_stencil(struct softpipe_surface *sps, - int x, int y, ubyte ssss[QUAD_SIZE]) -{ - const unsigned *src - = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - ssss[0] = src[0] >> 24; - ssss[1] = src[1] >> 24; - src += sps->surface.region->pitch; - ssss[2] = src[0] >> 24; - ssss[3] = src[1] >> 24; -} - -static void -s8z24_write_quad_stencil(struct softpipe_surface *sps, - int x, int y, const ubyte ssss[QUAD_SIZE]) -{ - static const unsigned mask = 0x00ffffff; - unsigned *dst - = ((unsigned *) (sps->surface.region->map + sps->surface.offset)) - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_S8_Z24); - - dst[0] = (dst[0] & mask) | (ssss[0] << 24); - dst[1] = (dst[1] & mask) | (ssss[1] << 24); - dst += sps->surface.region->pitch; - dst[0] = (dst[0] & mask) | (ssss[2] << 24); - dst[1] = (dst[1] & mask) | (ssss[3] << 24); -} - - /** * Return Z component as float in [0,1]. Stencil part ignored. */ @@ -832,43 +450,6 @@ s8z24_get_tile(struct pipe_surface *ps, } -/*** PIPE_FORMAT_U_S8 ***/ - -static void -s8_read_quad_stencil(struct softpipe_surface *sps, - int x, int y, ubyte ssss[QUAD_SIZE]) -{ - const ubyte *src - = sps->surface.region->map + sps->surface.offset - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_S8); - - ssss[0] = src[0]; - ssss[1] = src[1]; - src += sps->surface.region->pitch; - ssss[2] = src[0]; - ssss[3] = src[1]; -} - -static void -s8_write_quad_stencil(struct softpipe_surface *sps, - int x, int y, const ubyte ssss[QUAD_SIZE]) -{ - ubyte *dst - = sps->surface.region->map + sps->surface.offset - + y * sps->surface.region->pitch + x; - - assert(sps->surface.format == PIPE_FORMAT_U_S8); - - dst[0] = ssss[0]; - dst[1] = ssss[1]; - dst += sps->surface.region->pitch; - dst[0] = ssss[2]; - dst[1] = ssss[3]; -} - - /** * Get raw tile, any 32-bit pixel format. */ @@ -979,6 +560,57 @@ put_tile_raw16(struct pipe_surface *ps, } +/** + * Get raw tile, any 16-bit pixel format. + */ +static void +get_tile_raw8(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, void *p) +{ + const ubyte *src + = ((const ubyte *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + ubyte *pDest = (ubyte *) p; + uint i; + uint w0 = w; + + assert(ps->format == PIPE_FORMAT_U_Z16); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + memcpy(pDest, src, w * sizeof(ubyte)); + src += ps->region->pitch; + pDest += w0; + } +} + +/** + * Put raw tile, any 16-bit pixel format. + */ +static void +put_tile_raw8(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const void *p) +{ + ubyte *dst + = ((ubyte *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + const ubyte *pSrc = (const ubyte *) p; + unsigned i; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_Z16); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + memcpy(dst, pSrc, w * sizeof(ubyte)); + dst += ps->region->pitch; + pSrc += w0; + } +} + + /** @@ -991,8 +623,6 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) switch (sps->surface.format) { case PIPE_FORMAT_U_A8_R8_G8_B8: - sps->read_quad_f_swz = a8r8g8b8_read_quad_f_swz; - sps->write_quad_f_swz = a8r8g8b8_write_quad_f_swz; sps->surface.get_tile = a8r8g8b8_get_tile; sps->surface.put_tile = a8r8g8b8_put_tile; break; @@ -1000,23 +630,15 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) sps->surface.get_tile = a1r5g5b5_get_tile; break; case PIPE_FORMAT_U_L8: - sps->read_quad_f_swz = l8_read_quad_f_swz; - sps->write_quad_f_swz = l8_write_quad_f_swz; sps->surface.get_tile = l8_get_tile; break; case PIPE_FORMAT_U_A8: - sps->read_quad_f_swz = a8_read_quad_f_swz; - sps->write_quad_f_swz = a8_write_quad_f_swz; sps->surface.get_tile = a8_get_tile; break; case PIPE_FORMAT_U_I8: - sps->read_quad_f_swz = i8_read_quad_f_swz; - sps->write_quad_f_swz = i8_write_quad_f_swz; sps->surface.get_tile = i8_get_tile; break; case PIPE_FORMAT_U_A8_L8: - sps->read_quad_f_swz = a8_l8_read_quad_f_swz; - sps->write_quad_f_swz = a8_l8_write_quad_f_swz; sps->surface.get_tile = a8_l8_get_tile; break; @@ -1026,32 +648,23 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) break; case PIPE_FORMAT_U_Z16: - sps->read_quad_z = z16_read_quad_z; - sps->write_quad_z = z16_write_quad_z; sps->surface.get_tile = z16_get_tile; sps->surface.get_tile_raw = get_tile_raw16; sps->surface.put_tile_raw = put_tile_raw16; break; case PIPE_FORMAT_U_Z32: - sps->read_quad_z = z32_read_quad_z; - sps->write_quad_z = z32_write_quad_z; sps->surface.get_tile = z32_get_tile; sps->surface.get_tile_raw = get_tile_raw32; sps->surface.put_tile_raw = put_tile_raw32; break; case PIPE_FORMAT_S8_Z24: - sps->read_quad_z = s8z24_read_quad_z; - sps->write_quad_z = s8z24_write_quad_z; - sps->read_quad_stencil = s8z24_read_quad_stencil; - sps->write_quad_stencil = s8z24_write_quad_stencil; sps->surface.get_tile = s8z24_get_tile; sps->surface.get_tile_raw = get_tile_raw32; sps->surface.put_tile_raw = put_tile_raw32; break; - case PIPE_FORMAT_U_S8: - sps->read_quad_stencil = s8_read_quad_stencil; - sps->write_quad_stencil = s8_write_quad_stencil; + sps->surface.get_tile_raw = get_tile_raw8; + sps->surface.put_tile_raw = put_tile_raw8; break; default: assert(0); diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 522f7612ab..06c0a01aee 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -46,44 +46,7 @@ struct softpipe_tile_cache; struct softpipe_surface { struct pipe_surface surface; - /** - * Functions for read/writing surface data - */ - void (*read_quad_f)( struct softpipe_surface *, - int x, int y, - float (*rgba)[NUM_CHANNELS] ); - - void (*read_quad_f_swz)( struct softpipe_surface *, - int x, int y, - float (*rrrr)[QUAD_SIZE] ); - - void (*read_quad_ub)( struct softpipe_surface *, - int x, int y, - ubyte (*rgba)[NUM_CHANNELS] ); - - - void (*write_quad_f)( struct softpipe_surface *, - int x, int y, - float (*rgba)[NUM_CHANNELS] ); - - void (*write_quad_f_swz)( struct softpipe_surface *, - int x, int y, - float (*rrrr)[QUAD_SIZE] ); - - - void (*write_quad_ub)( struct softpipe_surface *, - int x, int y, - ubyte (*rgba)[NUM_CHANNELS] ); - - void (*read_quad_z)(struct softpipe_surface *, - int x, int y, unsigned zzzz[QUAD_SIZE]); - void (*write_quad_z)(struct softpipe_surface *, - int x, int y, const unsigned zzzz[QUAD_SIZE]); - - void (*read_quad_stencil)(struct softpipe_surface *, - int x, int y, ubyte ssss[QUAD_SIZE]); - void (*write_quad_stencil)(struct softpipe_surface *, - int x, int y, const ubyte ssss[QUAD_SIZE]); + /* no softpipe-specific extras now */ }; -- cgit v1.2.3 From cae640eae63544211710a2848e23f6d1d74f827f Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 21 Oct 2007 17:14:46 -0600 Subject: silence warning --- src/mesa/pipe/softpipe/sp_clear.c | 2 +- src/mesa/pipe/softpipe/sp_flush.c | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 4f04f8243f..96bca4b171 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -47,7 +47,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); - struct softpipe_surface *sps = softpipe_surface(ps); + /*struct softpipe_surface *sps = softpipe_surface(ps);*/ unsigned x, y, w, h; softpipe_update_derived(softpipe); /* not needed?? */ diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index 5eb4d3367d..2b077c2021 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -34,6 +34,7 @@ #include "sp_flush.h" #include "sp_context.h" #include "sp_surface.h" +#include "sp_state.h" #include "sp_tile_cache.h" #include "sp_winsys.h" -- cgit v1.2.3 From c2322333b8d1732f4c6d4b71ff5ee2ea772d3cb5 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 21 Oct 2007 17:15:07 -0600 Subject: rename some vars --- src/mesa/pipe/softpipe/sp_quad_blend.c | 50 +++++++++++++++++----------------- 1 file changed, 25 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 5787b89588..b056e477b1 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -55,44 +55,44 @@ do { \ DST[3] = SRC; \ } while(0) -#define VEC4_ADD(SUM, A, B) \ +#define VEC4_ADD(R, A, B) \ do { \ - SUM[0] = A[0] + B[0]; \ - SUM[1] = A[1] + B[1]; \ - SUM[2] = A[2] + B[2]; \ - SUM[3] = A[3] + B[3]; \ + R[0] = A[0] + B[0]; \ + R[1] = A[1] + B[1]; \ + R[2] = A[2] + B[2]; \ + R[3] = A[3] + B[3]; \ } while (0) -#define VEC4_SUB(SUM, A, B) \ +#define VEC4_SUB(R, A, B) \ do { \ - SUM[0] = A[0] - B[0]; \ - SUM[1] = A[1] - B[1]; \ - SUM[2] = A[2] - B[2]; \ - SUM[3] = A[3] - B[3]; \ + R[0] = A[0] - B[0]; \ + R[1] = A[1] - B[1]; \ + R[2] = A[2] - B[2]; \ + R[3] = A[3] - B[3]; \ } while (0) -#define VEC4_MUL(SUM, A, B) \ +#define VEC4_MUL(R, A, B) \ do { \ - SUM[0] = A[0] * B[0]; \ - SUM[1] = A[1] * B[1]; \ - SUM[2] = A[2] * B[2]; \ - SUM[3] = A[3] * B[3]; \ + R[0] = A[0] * B[0]; \ + R[1] = A[1] * B[1]; \ + R[2] = A[2] * B[2]; \ + R[3] = A[3] * B[3]; \ } while (0) -#define VEC4_MIN(SUM, A, B) \ +#define VEC4_MIN(R, A, B) \ do { \ - SUM[0] = (A[0] < B[0]) ? A[0] : B[0]; \ - SUM[1] = (A[1] < B[1]) ? A[1] : B[1]; \ - SUM[2] = (A[2] < B[2]) ? A[2] : B[2]; \ - SUM[3] = (A[3] < B[3]) ? A[3] : B[3]; \ + R[0] = (A[0] < B[0]) ? A[0] : B[0]; \ + R[1] = (A[1] < B[1]) ? A[1] : B[1]; \ + R[2] = (A[2] < B[2]) ? A[2] : B[2]; \ + R[3] = (A[3] < B[3]) ? A[3] : B[3]; \ } while (0) -#define VEC4_MAX(SUM, A, B) \ +#define VEC4_MAX(R, A, B) \ do { \ - SUM[0] = (A[0] > B[0]) ? A[0] : B[0]; \ - SUM[1] = (A[1] > B[1]) ? A[1] : B[1]; \ - SUM[2] = (A[2] > B[2]) ? A[2] : B[2]; \ - SUM[3] = (A[3] > B[3]) ? A[3] : B[3]; \ + R[0] = (A[0] > B[0]) ? A[0] : B[0]; \ + R[1] = (A[1] > B[1]) ? A[1] : B[1]; \ + R[2] = (A[2] > B[2]) ? A[2] : B[2]; \ + R[3] = (A[3] > B[3]) ? A[3] : B[3]; \ } while (0) -- cgit v1.2.3 From b3204c2aff3f3d442ada04f241f352155a3af205 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 21 Oct 2007 18:06:35 -0600 Subject: Start implementing cache routines for textures. First step to consolidating surface/texture caching... --- src/mesa/pipe/softpipe/sp_tile_cache.c | 61 ++++++++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_tile_cache.h | 11 +++++- 2 files changed, 71 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 74bd4a3d11..d88bce4619 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -49,6 +49,7 @@ struct softpipe_tile_cache { struct softpipe_surface *surface; /**< the surface we're caching */ + struct pipe_mipmap_tree *texture; /**< if caching a texture */ struct softpipe_cached_tile entries[NUM_ENTRIES]; uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; }; @@ -124,6 +125,12 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, } +void +sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, + struct pipe_mipmap_tree *texture) +{ + tc->texture = texture; +} void @@ -249,6 +256,60 @@ sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y) } +/** + * Given the texture face, level, zslice, x and y values, compute + * the cache entry position/index where we'd hope to find the + * cached texture tile. + * This is basically a direct-map cache. + * XXX There's probably lots of ways in which we can improve this. + */ +static uint +tex_cache_pos(int x, int y, int z, int face, int level) +{ + uint entry = x + y * 2 + z * 4 + face + level; + return entry % NUM_ENTRIES; +} + + +/** + * Similar to sp_get_cached_tile() but for textures. + * Tiles are read-only and indexed with more params. + */ +struct softpipe_cached_tile * +sp_get_cached_tile_tex(struct softpipe_tile_cache *tc, int x, int y, int z, + int face, int level) +{ + struct pipe_context *pipe; /* XXX need this */ + + /* tile pos in framebuffer: */ + const int tile_x = x & ~(TILE_SIZE - 1); + const int tile_y = y & ~(TILE_SIZE - 1); + + /* cache pos/entry: */ + const int pos = tex_cache_pos(x / TILE_SIZE, y / TILE_SIZE, + z, face, level); + struct softpipe_cached_tile *tile = tc->entries + pos; + + if (tile_x != tile->x || + tile_y != tile->y || + z != tile->z || + face != tile->face || + level != tile->level) { + struct pipe_surface *ps + = pipe->get_tex_surface(pipe, tc->texture, face, level, z); + + ps->get_tile(ps, + tile_x, tile_y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + + pipe_surface_reference(&ps, NULL); + } + + return tile; +} + + + void sp_clear_tile_cache(struct softpipe_tile_cache *tc, unsigned clearval) { diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index 80bcac6904..9879b1821c 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -45,7 +45,8 @@ struct softpipe_tile_cache; struct softpipe_cached_tile { - int x, y; /** pos of tile in window coords */ + int x, y; /**< pos of tile in window coords */ + int z, face, level; /**< Extra texture indexes */ union { float color[TILE_SIZE][TILE_SIZE][4]; uint depth32[TILE_SIZE][TILE_SIZE]; @@ -65,6 +66,10 @@ extern void sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, struct softpipe_surface *sps); +extern void +sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, + struct pipe_mipmap_tree *texture); + extern void sp_flush_tile_cache(struct softpipe_tile_cache *tc); @@ -74,6 +79,10 @@ sp_clear_tile_cache(struct softpipe_tile_cache *tc, unsigned clearval); extern struct softpipe_cached_tile * sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y); +extern struct softpipe_cached_tile * +sp_get_cached_tile_tex(struct softpipe_tile_cache *tc, int x, int y, int z, + int face, int level); + #endif /* SP_TILE_CACHE_H */ -- cgit v1.2.3 From ec3bd21c465f27d0a8e313e00338109d21019fc0 Mon Sep 17 00:00:00 2001 From: keithw Date: Fri, 19 Oct 2007 16:41:13 +0100 Subject: pull clip/ module wide and stippled lines/points code --- src/mesa/pipe/draw/draw_linestipple.c | 266 ++++++++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_wide_prims.c | 275 ++++++++++++++++++++++++++++++++++ 2 files changed, 541 insertions(+) create mode 100644 src/mesa/pipe/draw/draw_linestipple.c create mode 100644 src/mesa/pipe/draw/draw_wide_prims.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_linestipple.c b/src/mesa/pipe/draw/draw_linestipple.c new file mode 100644 index 0000000000..1fac1ebe66 --- /dev/null +++ b/src/mesa/pipe/draw/draw_linestipple.c @@ -0,0 +1,266 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +/* Implement line stipple by cutting lines up into smaller lines. + * There are hundreds of ways to implement line stipple, this is one + * choice that should work in all situations, requires no state + * manipulations, but with a penalty in terms of large amounts of + * generated geometry. + */ + +#include "imports.h" +#include "macros.h" + +#define CLIP_PRIVATE +#include "clip/clip_context.h" + +#define CLIP_PIPE_PRIVATE +#include "clip/clip_pipe.h" + + + +struct stipple_stage { + struct clip_pipe_stage stage; + + GLuint hw_data_offset; + + GLfloat counter; + GLuint pattern; + GLuint factor; +}; + + + + +static INLINE struct stipple_stage *stipple_stage( struct clip_pipe_stage *stage ) +{ + return (struct stipple_stage *)stage; +} + + + + +static void interp_attr( const struct vf_attr *a, + GLubyte *vdst, + GLfloat t, + const GLubyte *vin, + const GLubyte *vout ) +{ + GLuint offset = a->vertoffset; + GLfloat fin[4], fout[4], fdst[4]; + + a->extract( a, fin, vin + offset ); + a->extract( a, fout, vout + offset ); + + fdst[0] = LINTERP( t, fout[0], fin[0] ); + fdst[1] = LINTERP( t, fout[1], fin[1] ); + fdst[2] = LINTERP( t, fout[2], fin[2] ); + fdst[3] = LINTERP( t, fout[3], fin[3] ); + + a->insert[4-1]( a, vdst + offset, fdst ); +} + + + + +/* Weird screen-space interpolation?? Otherwise do something special + * with pos.w or fix vertices to always have clip coords available. + */ +static void screen_interp( struct vertex_fetch *vf, + struct vertex_header *dst, + GLfloat t, + const struct vertex_header *out, + const struct vertex_header *in ) +{ + GLubyte *vdst = (GLubyte *)dst; + const GLubyte *vin = (const GLubyte *)in; + const GLubyte *vout = (const GLubyte *)out; + + const struct vf_attr *a = vf->attr; + const GLuint attr_count = vf->attr_count; + GLuint j; + + /* Vertex header. + */ + { + assert(a[0].attrib == VF_ATTRIB_VERTEX_HEADER); + dst->clipmask = 0; + dst->edgeflag = 0; + dst->pad = 0; + dst->index = 0xffff; + } + + + /* Other attributes + */ + for (j = 1; j < attr_count; j++) { + interp_attr(&a[j], vdst, t, vin, vout); + } +} + + + +/* Clip a line against the viewport and user clip planes. + */ +static void draw_line_segment( struct clip_pipe_stage *stage, + struct prim_header *header, + GLfloat t0, + GLfloat t1 ) +{ + struct vertex_fetch *vf = stage->pipe->draw->vb.vf; + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + + struct prim_header newprim = *header; + header = &newprim; + + _mesa_printf("%s %f %f\n", __FUNCTION__, t0, t1); + + if (t0 > 0.0) { + screen_interp( vf, stage->tmp[0], t0, v0, v1 ); + header->v[0] = stage->tmp[0]; + } + + if (t1 < 1.0) { + screen_interp( vf, stage->tmp[1], t1, v0, v1 ); + header->v[1] = stage->tmp[1]; + } + + stage->next->line( stage->next, header ); +} + + + +/* XXX: don't really want to iterate like this. + */ +static INLINE unsigned +stipple_test(int counter, ushort pattern, int factor) +{ + int b = (counter / factor) & 0xf; + return (1 << b) & pattern; +} + + + +/* XXX: Need to have precalculated flatshading already. + */ +static void stipple_line( struct clip_pipe_stage *stage, + struct prim_header *header ) +{ + struct stipple_stage *stipple = stipple_stage(stage); + GLuint hw_data_offset = stipple->hw_data_offset; + const GLfloat *pos0 = (GLfloat *)&(header->v[0]->data[hw_data_offset]); + const GLfloat *pos1 = (GLfloat *)&(header->v[1]->data[hw_data_offset]); + GLfloat start = 0; + int state = 0; + + GLfloat x0 = (GLfloat)pos0[0]; + GLfloat x1 = (GLfloat)pos1[0]; + GLfloat y0 = (GLfloat)pos0[1]; + GLfloat y1 = (GLfloat)pos1[1]; + + GLfloat dx = x0 > x1 ? x0 - x1 : x1 - x0; + GLfloat dy = y0 > y1 ? y0 - y1 : y1 - y0; + + GLfloat length = MAX2(dx, dy); + GLint i; + + if (header->reset_line_stipple) + stipple->counter = 0; + + /* XXX: iterating like this is lame + */ + for (i = 0; i < length; i++) { + int result = !!stipple_test( stipple->counter+i, stipple->pattern, stipple->factor ); +// _mesa_printf("%d %f %d\n", i, length, result); + if (result != state) { + if (state) { + if (start != i) + draw_line_segment( stage, header, start / length, i / length ); + } + else { + start = i; + } + state = result; + } + } + + if (state && start < length) + draw_line_segment( stage, header, start / length, 1.0 ); + + stipple->counter += length; +} + + + +static void stipple_begin( struct clip_pipe_stage *stage ) +{ + struct stipple_stage *stipple = stipple_stage(stage); + struct clip_context *draw = stage->pipe->draw; + + if (stage->pipe->draw->vb_state.clipped_prims) + stipple->hw_data_offset = 16; + else + stipple->hw_data_offset = 0; + + stipple->stage.tri = clip_passthrough_tri; + stipple->stage.point = clip_passthrough_point; + + stipple->stage.line = stipple_line; + stipple->factor = draw->state.line_stipple_factor + 1; + stipple->pattern = draw->state.line_stipple_pattern; + + stage->next->begin( stage->next ); +} + + + +static void stipple_end( struct clip_pipe_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct clip_pipe_stage *clip_pipe_stipple( struct clip_pipeline *pipe ) +{ + struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); + + clip_pipe_alloc_tmps( &stipple->stage, 4 ); + + stipple->stage.pipe = pipe; + stipple->stage.next = NULL; + stipple->stage.begin = stipple_begin; + stipple->stage.point = clip_passthrough_point; + stipple->stage.line = stipple_line; + stipple->stage.tri = clip_passthrough_tri; + stipple->stage.reset_tmps = clip_pipe_reset_tmps; + stipple->stage.end = stipple_end; + + return &stipple->stage; +} diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c new file mode 100644 index 0000000000..20151f1538 --- /dev/null +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -0,0 +1,275 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + + +struct wide_stage { + struct draw_stage stage; + + unsigned hw_data_offset; + + float half_line_width; + float half_point_size; +}; + + + +static INLINE struct wide_stage *wide_stage( struct draw_stage *stage ) +{ + return (struct wide_stage *)stage; +} + + + + + +static void tri( struct draw_stage *next, + struct vertex_header *v0, + struct vertex_header *v1, + struct vertex_header *v2 ) +{ + struct prim_header tmp; + + tmp.det = 1.0; + tmp.v[0] = v0; + tmp.v[1] = v1; + tmp.v[2] = v2; + next->tri( next, &tmp ); +} + +static void quad( struct draw_stage *next, + struct vertex_header *v0, + struct vertex_header *v1, + struct vertex_header *v2, + struct vertex_header *v3 ) +{ + /* XXX: Need to disable tri-stipple + */ + tri( next, v0, v1, v3 ); + tri( next, v2, v0, v3 ); +} + +static void wide_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct wide_stage *wide = wide_stage(stage); + unsigned hw_data_offset = wide->hw_data_offset; + float half_width = wide->half_line_width; + + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); + + float *pos0 = (float *)&(v0->data[hw_data_offset]); + float *pos1 = (float *)&(v1->data[hw_data_offset]); + float *pos2 = (float *)&(v2->data[hw_data_offset]); + float *pos3 = (float *)&(v3->data[hw_data_offset]); + + float dx = FABSF(pos0[0] - pos2[0]); + float dy = FABSF(pos0[1] - pos2[1]); + + if (dx > dy) { + pos0[1] -= half_width; + pos1[1] += half_width; + pos2[1] -= half_width; + pos3[1] += half_width; + } + else { + pos0[0] -= half_width; + pos1[0] += half_width; + pos2[0] -= half_width; + pos3[0] += half_width; + } + + quad( stage->next, v0, v1, v2, v3 ); +} + + +static void make_wide_point( struct draw_stage *stage, + const struct vertex_header *vin, + struct vertex_header *v[] ) +{ + struct wide_stage *wide = wide_stage(stage); + unsigned hw_data_offset = wide->hw_data_offset; + float half_size = wide->half_point_size; + float *pos[4]; + + v[0] = dup_vert(stage, vin, 0); + pos[0] = (float *)&(v[0]->data[hw_data_offset]); + + /* Probably not correct: + */ + pos[0][0] = pos[0][0]; + pos[0][1] = pos[0][1] - .5; + + v[1] = dup_vert(stage, v[0], 1); + v[2] = dup_vert(stage, v[0], 2); + v[3] = dup_vert(stage, v[0], 3); + + pos[1] = (float *)&(v[1]->data[hw_data_offset]); + pos[2] = (float *)&(v[2]->data[hw_data_offset]); + pos[3] = (float *)&(v[3]->data[hw_data_offset]); + + _mesa_printf("point %f %f, %f\n", pos[0][0], pos[0][1], half_size); + + pos[0][0] -= half_size; + pos[0][1] -= half_size; + + pos[1][0] -= half_size; + pos[1][1] += half_size; + + pos[2][0] += half_size; + pos[2][1] -= half_size; + + pos[3][0] += half_size; + pos[3][1] += half_size; + +// quad( stage->next, v[0], v[1], v[2], v[3] ); +} + +static void wide_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v[4]; + make_wide_point(stage, header->v[0], v ); + quad( stage->next, v[0], v[1], v[2], v[3] ); +} + + +static void set_texcoord( struct vertex_fetch *vf, + struct vertex_header *v, + const float *val ) +{ + GLubyte *dst = (GLubyte *)v; + const struct vf_attr *a = vf->attr; + const unsigned attr_count = vf->attr_count; + unsigned j; + + /* XXX: precompute which attributes need to be set. + */ + for (j = 1; j < attr_count; j++) { + if (a[j].attrib >= VF_ATTRIB_TEX0 && + a[j].attrib <= VF_ATTRIB_TEX7) + a[j].insert[4-1]( &a[j], dst + a[j].vertoffset, val ); + } +} + + + +/* If there are lots of sprite points (and why wouldn't there be?) it + * would probably be more sensible to change hardware setup to + * optimize this rather than doing the whole thing in software like + * this. + */ +static void sprite_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v[4]; + struct vertex_fetch *vf = stage->pipe->draw->vb.vf; + + static const float tex00[4] = { 0, 0, 0, 1 }; + static const float tex01[4] = { 0, 1, 0, 1 }; + static const float tex11[4] = { 1, 1, 0, 1 }; + static const float tex10[4] = { 1, 0, 0, 1 }; + + make_wide_point(stage, header->v[0], &v[0] ); + + set_texcoord( vf, v[0], tex00 ); + set_texcoord( vf, v[1], tex01 ); + set_texcoord( vf, v[2], tex10 ); + set_texcoord( vf, v[3], tex11 ); + + quad( stage->next, v[0], v[1], v[2], v[3] ); +} + + + +static void wide_begin( struct draw_stage *stage ) +{ + struct wide_stage *wide = wide_stage(stage); + struct clip_context *draw = stage->pipe->draw; + + if (draw->vb_state.clipped_prims) + wide->hw_data_offset = 16; + else + wide->hw_data_offset = 0; + + wide->half_point_size = draw->state.point_size / 2; + wide->half_line_width = draw->state.line_width / 2; + + if (draw->state.line_width != 1.0) { + wide->stage.line = wide_line; + } + else { + wide->stage.line = clip_passthrough_line; + } + + if (draw->state.point_sprite) { + wide->stage.point = sprite_point; + } + else if (draw->state.point_size != 1.0) { + wide->stage.point = wide_point; + } + else { + wide->stage.point = clip_passthrough_point; + } + + + stage->next->begin( stage->next ); +} + + + +static void wide_end( struct draw_stage *stage ) +{ + stage->next->end( stage->next ); +} + +struct draw_stage *clip_pipe_wide( struct clip_pipeline *pipe ) +{ + struct wide_stage *wide = CALLOC_STRUCT(wide_stage); + + clip_pipe_alloc_tmps( &wide->stage, 4 ); + + wide->stage.pipe = pipe; + wide->stage.next = NULL; + wide->stage.begin = wide_begin; + wide->stage.point = wide_point; + wide->stage.line = wide_line; + wide->stage.tri = clip_passthrough_tri; + wide->stage.reset_tmps = clip_pipe_reset_tmps; + wide->stage.end = wide_end; + + return &wide->stage; +} -- cgit v1.2.3 From 70eb7996f265f3634dabda078f13d1be3533cc65 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 09:37:26 -0600 Subject: Finish unifying the surface and texture tile caches. --- src/mesa/pipe/softpipe/sp_context.c | 20 +++++-- src/mesa/pipe/softpipe/sp_context.h | 2 + src/mesa/pipe/softpipe/sp_quad_fs.c | 20 +++---- src/mesa/pipe/softpipe/sp_state_sampler.c | 4 ++ src/mesa/pipe/softpipe/sp_tex_sample.c | 98 +++++-------------------------- src/mesa/pipe/softpipe/sp_tile_cache.c | 31 +++++++--- src/mesa/pipe/softpipe/sp_tile_cache.h | 5 +- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 10 +--- 8 files changed, 72 insertions(+), 118 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 10c53a257b..bf61019f62 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -343,6 +343,20 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.mipmap_tree_layout = softpipe_mipmap_tree_layout; softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; + /* + * Alloc caches for accessing drawing surfaces and textures. + * Must be before quad stage setup! + */ + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + softpipe->cbuf_cache[i] = sp_create_tile_cache(); + softpipe->zbuf_cache = sp_create_tile_cache(); + softpipe->sbuf_cache_sep = sp_create_tile_cache(); + softpipe->sbuf_cache = softpipe->sbuf_cache_sep; /* initial value */ + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + softpipe->tex_cache[i] = sp_create_tile_cache(); + + /* setup quad rendering stages */ softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); softpipe->quad.shade = sp_quad_shade_stage(softpipe); @@ -380,11 +394,5 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) - softpipe->cbuf_cache[i] = sp_create_tile_cache(); - softpipe->zbuf_cache = sp_create_tile_cache(); - softpipe->sbuf_cache_sep = sp_create_tile_cache(); - softpipe->sbuf_cache = softpipe->sbuf_cache_sep; /* initial value */ - return &softpipe->pipe; } diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index ea05f80d59..3e77bd6b85 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -163,6 +163,8 @@ struct softpipe_context { /** This either points to zbuf_cache or sbuf_cache_sep */ struct softpipe_tile_cache *sbuf_cache; + struct softpipe_tile_cache *tex_cache[PIPE_MAX_SAMPLERS]; + int use_sse : 1; }; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index c850f08c57..674e02d0f4 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -153,19 +153,10 @@ static void shade_begin(struct quad_stage *qs) struct softpipe_context *softpipe = qs->softpipe; unsigned i, entry; + /* set TGSI sampler state that varies */ for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { qss->samplers[i].state = softpipe->sampler[i]; qss->samplers[i].texture = softpipe->texture[i]; - qss->samplers[i].get_samples = sp_get_samples; - qss->samplers[i].pipe = &softpipe->pipe; - /* init cache info here */ - for (entry = 0; entry < TEX_CACHE_NUM_ENTRIES; entry++) { - qss->samplers[i].cache[entry].x = -1; - qss->samplers[i].cache[entry].y = -1; - qss->samplers[i].cache[entry].level = -1; - qss->samplers[i].cache[entry].face = -1; - qss->samplers[i].cache[entry].zslice = -1; - } } /* XXX only do this if the fragment shader changes... */ @@ -196,6 +187,7 @@ static void shade_begin(struct quad_stage *qs) struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) { struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); + uint i; /* allocate storage for program inputs/outputs, aligned to 16 bytes */ qss->inputs = malloc(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); @@ -207,5 +199,13 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) qss->stage.begin = shade_begin; qss->stage.run = shade_quad; + /* set TGSI sampler state that's constant */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + assert(softpipe->tex_cache[i]); + qss->samplers[i].get_samples = sp_get_samples; + qss->samplers[i].pipe = &softpipe->pipe; + qss->samplers[i].cache = softpipe->tex_cache[i]; + } + return &qss->stage; } diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 4084e9163f..c00e815f2d 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -31,6 +31,8 @@ #include "sp_context.h" #include "sp_state.h" +#include "sp_tile_cache.h" + void * softpipe_create_sampler_state(struct pipe_context *pipe, @@ -72,5 +74,7 @@ softpipe_set_texture_state(struct pipe_context *pipe, assert(unit < PIPE_MAX_SAMPLERS); softpipe->texture[unit] = texture; /* ptr, not struct */ + sp_tile_cache_set_texture(softpipe->tex_cache[unit], texture); + softpipe->dirty |= SP_NEW_TEXTURE; } diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 64cb94d944..385b8042c3 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -36,6 +36,7 @@ #include "sp_context.h" #include "sp_surface.h" #include "sp_tex_sample.h" +#include "sp_tile_cache.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" @@ -522,28 +523,6 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, } - -/** - * Given the texture face, level, zslice, x and y values, compute - * the cache entry position/index where we'd hope to find the - * cached texture tile. - * This is basically a direct-map cache. - * XXX There's probably lots of ways in which we can improve - * texture caching.... - */ -static unsigned -compute_cache_pos(unsigned face, unsigned level, unsigned zslice, - int tx, int ty) -{ -#if 01 - unsigned entry = tx + ty * 2 + zslice * 4 + face + level; - return entry % TEX_CACHE_NUM_ENTRIES; -#else - return 0; -#endif -} - - /** * Get a texel from a texture, using the texture tile cache. * @@ -551,74 +530,27 @@ compute_cache_pos(unsigned face, unsigned level, unsigned zslice, * \param level the mipmap level * \param x the x coord of texel within 2D image * \param y the y coord of texel within 2D image - * \param zslice which slice of a 3D texture + * \param z which slice of a 3D texture * \param rgba the quad to put the texel/color into * \param j which element of the rgba quad to write to + * + * XXX maybe move this into sp_tile_cache.c and merge with the + * sp_get_cached_tile_tex() function. Also, get 4 texels instead of 1... */ static void get_texel(struct tgsi_sampler *sampler, - unsigned face, unsigned level, int x, int y, unsigned zslice, + unsigned face, unsigned level, int x, int y, int z, float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j) { - int tx = x / TEX_CACHE_TILE_SIZE; - int ty = y / TEX_CACHE_TILE_SIZE; - unsigned entry = compute_cache_pos(face, level, zslice, tx, ty); - - if (tx != sampler->cache[entry].x || - ty != sampler->cache[entry].y || - face != sampler->cache[entry].face || - level != sampler->cache[entry].level || - zslice != sampler->cache[entry].zslice) { - /* entry is not what's expected */ - struct pipe_context *pipe = (struct pipe_context *) sampler->pipe; - struct pipe_surface *ps - = pipe->get_tex_surface(pipe, sampler->texture, face, level, zslice); - - /* - printf("cache miss (%d, %d) face %u\n", tx, ty, face); - */ - - assert(ps->width == sampler->texture->level[level].width); - assert(ps->height == sampler->texture->level[level].height); - sampler->cache[entry].x = tx; - sampler->cache[entry].y = ty; - sampler->cache[entry].level = level; - sampler->cache[entry].face = face; - sampler->cache[entry].zslice = zslice; - ps->get_tile(ps, - tx * TEX_CACHE_TILE_SIZE, - ty * TEX_CACHE_TILE_SIZE, - TEX_CACHE_TILE_SIZE, TEX_CACHE_TILE_SIZE, - (float *) sampler->cache[entry].data); - - pipe_surface_reference(&ps, NULL); - } - else { - /* - printf("cache hit (%d, %d)\n", x, y); - */ - } - - /* get the texel from cache entry */ - tx = x % TEX_CACHE_TILE_SIZE; - ty = y % TEX_CACHE_TILE_SIZE; - if (sampler->texture->format == PIPE_FORMAT_U_Z16 || - sampler->texture->format == PIPE_FORMAT_U_Z32 || - sampler->texture->format == PIPE_FORMAT_S8_Z24) { - /* get_tile() returned one float per texel */ - float *src = (float *) sampler->cache[entry].data; - rgba[0][j] = - rgba[1][j] = - rgba[2][j] = - rgba[3][j] = src[ty * TEX_CACHE_TILE_SIZE + tx]; - } - else { - /* get_tile() returned four floats per texel */ - rgba[0][j] = sampler->cache[entry].data[ty][tx][0]; - rgba[1][j] = sampler->cache[entry].data[ty][tx][1]; - rgba[2][j] = sampler->cache[entry].data[ty][tx][2]; - rgba[3][j] = sampler->cache[entry].data[ty][tx][3]; - } + const int tx = x % TILE_SIZE; + const int ty = y % TILE_SIZE; + const struct softpipe_cached_tile *tile + = sp_get_cached_tile_tex(sampler->pipe, sampler->cache, + x, y, z, face, level); + rgba[0][j] = tile->data.color[ty][tx][0]; + rgba[1][j] = tile->data.color[ty][tx][1]; + rgba[2][j] = tile->data.color[ty][tx][2]; + rgba[3][j] = tile->data.color[ty][tx][3]; } diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index d88bce4619..1e287c91a4 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -129,7 +129,15 @@ void sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, struct pipe_mipmap_tree *texture) { + uint i; + tc->texture = texture; + + /* mark as entries as invalid/empty */ + /* XXX we should try to avoid this when the teximage hasn't changed */ + for (i = 0; i < NUM_ENTRIES; i++) { + tc->entries[i].x = -1; + } } @@ -263,10 +271,10 @@ sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y) * This is basically a direct-map cache. * XXX There's probably lots of ways in which we can improve this. */ -static uint +static INLINE uint tex_cache_pos(int x, int y, int z, int face, int level) { - uint entry = x + y * 2 + z * 4 + face + level; + uint entry = x + y * 5 + z * 4 + face + level; return entry % NUM_ENTRIES; } @@ -275,19 +283,17 @@ tex_cache_pos(int x, int y, int z, int face, int level) * Similar to sp_get_cached_tile() but for textures. * Tiles are read-only and indexed with more params. */ -struct softpipe_cached_tile * -sp_get_cached_tile_tex(struct softpipe_tile_cache *tc, int x, int y, int z, +const struct softpipe_cached_tile * +sp_get_cached_tile_tex(struct pipe_context *pipe, + struct softpipe_tile_cache *tc, int x, int y, int z, int face, int level) { - struct pipe_context *pipe; /* XXX need this */ - /* tile pos in framebuffer: */ const int tile_x = x & ~(TILE_SIZE - 1); const int tile_y = y & ~(TILE_SIZE - 1); - /* cache pos/entry: */ - const int pos = tex_cache_pos(x / TILE_SIZE, y / TILE_SIZE, - z, face, level); + const uint pos = tex_cache_pos(x / TILE_SIZE, y / TILE_SIZE, z, + face, level); struct softpipe_cached_tile *tile = tc->entries + pos; if (tile_x != tile->x || @@ -295,6 +301,7 @@ sp_get_cached_tile_tex(struct softpipe_tile_cache *tc, int x, int y, int z, z != tile->z || face != tile->face || level != tile->level) { + /* XXX this call is a bit heavier than we'd like: */ struct pipe_surface *ps = pipe->get_tex_surface(pipe, tc->texture, face, level, z); @@ -303,6 +310,12 @@ sp_get_cached_tile_tex(struct softpipe_tile_cache *tc, int x, int y, int z, (float *) tile->data.color); pipe_surface_reference(&ps, NULL); + + tile->x = tile_x; + tile->y = tile_y; + tile->z = z; + tile->face = face; + tile->level = level; } return tile; diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index 9879b1821c..a1cc387a12 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -79,8 +79,9 @@ sp_clear_tile_cache(struct softpipe_tile_cache *tc, unsigned clearval); extern struct softpipe_cached_tile * sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y); -extern struct softpipe_cached_tile * -sp_get_cached_tile_tex(struct softpipe_tile_cache *tc, int x, int y, int z, +extern const struct softpipe_cached_tile * +sp_get_cached_tile_tex(struct pipe_context *pipe, + struct softpipe_tile_cache *tc, int x, int y, int z, int face, int level); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 20647e72e2..978c2d574e 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -29,14 +29,8 @@ struct tgsi_interp_coef float dady[NUM_CHANNELS]; }; -#define TEX_CACHE_TILE_SIZE 8 -#define TEX_CACHE_NUM_ENTRIES 8 -struct tgsi_texture_cache_entry -{ - int x, y, face, level, zslice; - float data[TEX_CACHE_TILE_SIZE][TEX_CACHE_TILE_SIZE][4]; -}; +struct softpipe_tile_cache; /**< Opaque to TGSI */ struct tgsi_sampler { @@ -50,7 +44,7 @@ struct tgsi_sampler float lodbias, float rgba[NUM_CHANNELS][QUAD_SIZE]); void *pipe; /*XXX temporary*/ - struct tgsi_texture_cache_entry cache[TEX_CACHE_NUM_ENTRIES]; + struct softpipe_tile_cache *cache; }; struct tgsi_exec_labels -- cgit v1.2.3 From 80d2bb7c643092e489dba77e4bd02a8684fcc42b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 11:01:02 -0600 Subject: update to working condition --- src/mesa/pipe/draw/draw_wide_prims.c | 202 ++++++++++++++++------------------- 1 file changed, 91 insertions(+), 111 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 20151f1538..8e1c4a8ae7 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -27,6 +27,7 @@ /* Authors: Keith Whitwell */ + #include "pipe/p_util.h" #include "pipe/p_defines.h" #include "draw_private.h" @@ -36,8 +37,6 @@ struct wide_stage { struct draw_stage stage; - unsigned hw_data_offset; - float half_line_width; float half_point_size; }; @@ -50,54 +49,50 @@ static INLINE struct wide_stage *wide_stage( struct draw_stage *stage ) } - - - -static void tri( struct draw_stage *next, - struct vertex_header *v0, - struct vertex_header *v1, - struct vertex_header *v2 ) +static void passthrough_point( struct draw_stage *stage, + struct prim_header *header ) { - struct prim_header tmp; + stage->next->point( stage->next, header ); +} - tmp.det = 1.0; - tmp.v[0] = v0; - tmp.v[1] = v1; - tmp.v[2] = v2; - next->tri( next, &tmp ); +static void passthrough_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line(stage->next, header); } -static void quad( struct draw_stage *next, - struct vertex_header *v0, - struct vertex_header *v1, - struct vertex_header *v2, - struct vertex_header *v3 ) +static void passthrough_tri( struct draw_stage *stage, + struct prim_header *header ) { - /* XXX: Need to disable tri-stipple - */ - tri( next, v0, v1, v3 ); - tri( next, v2, v0, v3 ); + stage->next->tri(stage->next, header); } + +/** + * Draw a wide line by drawing a quad (two triangles). + * XXX still need line stipple. + * XXX need to disable polygon stipple. + */ static void wide_line( struct draw_stage *stage, struct prim_header *header ) { - struct wide_stage *wide = wide_stage(stage); - unsigned hw_data_offset = wide->hw_data_offset; - float half_width = wide->half_line_width; + const struct wide_stage *wide = wide_stage(stage); + const float half_width = wide->half_line_width; + + struct prim_header tri; struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); - float *pos0 = (float *)&(v0->data[hw_data_offset]); - float *pos1 = (float *)&(v1->data[hw_data_offset]); - float *pos2 = (float *)&(v2->data[hw_data_offset]); - float *pos3 = (float *)&(v3->data[hw_data_offset]); - - float dx = FABSF(pos0[0] - pos2[0]); - float dy = FABSF(pos0[1] - pos2[1]); + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + const float dx = FABSF(pos0[0] - pos2[0]); + const float dy = FABSF(pos0[1] - pos2[1]); if (dx > dy) { pos0[1] -= half_width; @@ -112,81 +107,63 @@ static void wide_line( struct draw_stage *stage, pos3[0] += half_width; } - quad( stage->next, v0, v1, v2, v3 ); + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); } -static void make_wide_point( struct draw_stage *stage, - const struct vertex_header *vin, - struct vertex_header *v[] ) +static void wide_point( struct draw_stage *stage, + struct prim_header *header ) { - struct wide_stage *wide = wide_stage(stage); - unsigned hw_data_offset = wide->hw_data_offset; + const struct wide_stage *wide = wide_stage(stage); float half_size = wide->half_point_size; - float *pos[4]; - - v[0] = dup_vert(stage, vin, 0); - pos[0] = (float *)&(v[0]->data[hw_data_offset]); - /* Probably not correct: - */ - pos[0][0] = pos[0][0]; - pos[0][1] = pos[0][1] - .5; + struct prim_header tri; - v[1] = dup_vert(stage, v[0], 1); - v[2] = dup_vert(stage, v[0], 2); - v[3] = dup_vert(stage, v[0], 3); - - pos[1] = (float *)&(v[1]->data[hw_data_offset]); - pos[2] = (float *)&(v[2]->data[hw_data_offset]); - pos[3] = (float *)&(v[3]->data[hw_data_offset]); - - _mesa_printf("point %f %f, %f\n", pos[0][0], pos[0][1], half_size); + /* four dups of original vertex */ + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); - pos[0][0] -= half_size; - pos[0][1] -= half_size; + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; - pos[1][0] -= half_size; - pos[1][1] += half_size; + pos0[0] -= half_size; + pos0[1] -= half_size; - pos[2][0] += half_size; - pos[2][1] -= half_size; + pos1[0] -= half_size; + pos1[1] += half_size; - pos[3][0] += half_size; - pos[3][1] += half_size; - -// quad( stage->next, v[0], v[1], v[2], v[3] ); -} + pos2[0] += half_size; + pos2[1] -= half_size; -static void wide_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v[4]; - make_wide_point(stage, header->v[0], v ); - quad( stage->next, v[0], v[1], v[2], v[3] ); -} + pos3[0] += half_size; + pos3[1] += half_size; + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); -static void set_texcoord( struct vertex_fetch *vf, - struct vertex_header *v, - const float *val ) -{ - GLubyte *dst = (GLubyte *)v; - const struct vf_attr *a = vf->attr; - const unsigned attr_count = vf->attr_count; - unsigned j; - - /* XXX: precompute which attributes need to be set. - */ - for (j = 1; j < attr_count; j++) { - if (a[j].attrib >= VF_ATTRIB_TEX0 && - a[j].attrib <= VF_ATTRIB_TEX7) - a[j].insert[4-1]( &a[j], dst + a[j].vertoffset, val ); - } + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); } - /* If there are lots of sprite points (and why wouldn't there be?) it * would probably be more sensible to change hardware setup to * optimize this rather than doing the whole thing in software like @@ -195,6 +172,7 @@ static void set_texcoord( struct vertex_fetch *vf, static void sprite_point( struct draw_stage *stage, struct prim_header *header ) { +#if 0 struct vertex_header *v[4]; struct vertex_fetch *vf = stage->pipe->draw->vb.vf; @@ -204,13 +182,14 @@ static void sprite_point( struct draw_stage *stage, static const float tex10[4] = { 1, 0, 0, 1 }; make_wide_point(stage, header->v[0], &v[0] ); - + set_texcoord( vf, v[0], tex00 ); set_texcoord( vf, v[1], tex01 ); set_texcoord( vf, v[2], tex10 ); set_texcoord( vf, v[3], tex11 ); quad( stage->next, v[0], v[1], v[2], v[3] ); +#endif } @@ -218,31 +197,26 @@ static void sprite_point( struct draw_stage *stage, static void wide_begin( struct draw_stage *stage ) { struct wide_stage *wide = wide_stage(stage); - struct clip_context *draw = stage->pipe->draw; - - if (draw->vb_state.clipped_prims) - wide->hw_data_offset = 16; - else - wide->hw_data_offset = 0; + struct draw_context *draw = stage->draw; - wide->half_point_size = draw->state.point_size / 2; - wide->half_line_width = draw->state.line_width / 2; + wide->half_point_size = 0.5 * draw->rasterizer->point_size; + wide->half_line_width = 0.5 * draw->rasterizer->line_width; - if (draw->state.line_width != 1.0) { + if (draw->rasterizer->line_width != 1.0) { wide->stage.line = wide_line; } else { - wide->stage.line = clip_passthrough_line; + wide->stage.line = passthrough_line; } - if (draw->state.point_sprite) { + if (0/*draw->state.point_sprite*/) { wide->stage.point = sprite_point; } - else if (draw->state.point_size != 1.0) { + else if (draw->rasterizer->point_size != 1.0) { wide->stage.point = wide_point; } else { - wide->stage.point = clip_passthrough_point; + wide->stage.point = passthrough_point; } @@ -256,20 +230,26 @@ static void wide_end( struct draw_stage *stage ) stage->next->end( stage->next ); } -struct draw_stage *clip_pipe_wide( struct clip_pipeline *pipe ) + +static void draw_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + +struct draw_stage *draw_wide_stage( struct draw_context *draw ) { struct wide_stage *wide = CALLOC_STRUCT(wide_stage); - clip_pipe_alloc_tmps( &wide->stage, 4 ); + draw_alloc_tmps( &wide->stage, 4 ); - wide->stage.pipe = pipe; + wide->stage.draw = draw; wide->stage.next = NULL; wide->stage.begin = wide_begin; wide->stage.point = wide_point; wide->stage.line = wide_line; - wide->stage.tri = clip_passthrough_tri; - wide->stage.reset_tmps = clip_pipe_reset_tmps; + wide->stage.tri = passthrough_tri; wide->stage.end = wide_end; + wide->stage.reset_stipple_counter = draw_reset_stipple_counter; return &wide->stage; } -- cgit v1.2.3 From e3444deec5a369e4ffabfeb9f6c257dd6b8e5a30 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 11:01:34 -0600 Subject: plug the wide prims code into the pipeline --- src/mesa/pipe/draw/draw_context.c | 1 + src/mesa/pipe/draw/draw_private.h | 2 ++ src/mesa/pipe/draw/draw_validate.c | 6 ++++++ src/mesa/sources | 3 ++- 4 files changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index d2015bd514..544c423c2d 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -48,6 +48,7 @@ struct draw_context *draw_create( void ) #endif /* create pipeline stages */ + draw->pipeline.wide = draw_wide_stage( draw ); draw->pipeline.unfilled = draw_unfilled_stage( draw ); draw->pipeline.twoside = draw_twoside_stage( draw ); draw->pipeline.offset = draw_offset_stage( draw ); diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 6aa4cd44a0..7e5d5db5ab 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -145,6 +145,7 @@ struct draw_context struct draw_stage *twoside; struct draw_stage *offset; struct draw_stage *unfilled; + struct draw_stage *wide; struct draw_stage *rasterize; } pipeline; @@ -233,6 +234,7 @@ extern struct draw_stage *draw_offset_stage( struct draw_context *context ); extern struct draw_stage *draw_clip_stage( struct draw_context *context ); extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); extern struct draw_stage *draw_cull_stage( struct draw_context *context ); +extern struct draw_stage *draw_wide_stage( struct draw_context *context ); extern struct draw_stage *draw_validate_stage( struct draw_context *context ); diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index 7c5a9dceca..5ec581d8b3 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -51,6 +51,12 @@ static void validate_begin( struct draw_stage *stage ) * shorter pipelines for lines & points. */ + if (draw->rasterizer->line_width != 1.0 || + draw->rasterizer->point_size != 1.0) { + draw->pipeline.wide->next = next; + next = draw->pipeline.wide; + } + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { draw->pipeline.unfilled->next = next; diff --git a/src/mesa/sources b/src/mesa/sources index d53e75c820..fea3234db3 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -166,12 +166,13 @@ DRAW_SOURCES = \ pipe/draw/draw_offset.c \ pipe/draw/draw_prim.c \ pipe/draw/draw_twoside.c \ + pipe/draw/draw_unfilled.c \ pipe/draw/draw_validate.c \ pipe/draw/draw_vertex.c \ pipe/draw/draw_vertex_cache.c \ pipe/draw/draw_vertex_fetch.c \ pipe/draw/draw_vertex_shader.c \ - pipe/draw/draw_unfilled.c + pipe/draw/draw_wide_prims.c TGSIEXEC_SOURCES = \ pipe/tgsi/exec/tgsi_build.c \ -- cgit v1.2.3 From 8d244159870d6f8d67b844e3c84b0d32a387e59c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 11:38:40 -0600 Subject: tweak point corners to pass conform test --- src/mesa/pipe/draw/draw_wide_prims.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 8e1c4a8ae7..e7e90fabae 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -125,6 +125,7 @@ static void wide_point( struct draw_stage *stage, { const struct wide_stage *wide = wide_stage(stage); float half_size = wide->half_point_size; + float left_adj, right_adj; struct prim_header tri; @@ -139,16 +140,19 @@ static void wide_point( struct draw_stage *stage, float *pos2 = v2->data[0]; float *pos3 = v3->data[0]; - pos0[0] -= half_size; + left_adj = -half_size + 0.25; + right_adj = half_size + 0.25; + + pos0[0] += left_adj; pos0[1] -= half_size; - pos1[0] -= half_size; + pos1[0] += left_adj; pos1[1] += half_size; - pos2[0] += half_size; + pos2[0] += right_adj; pos2[1] -= half_size; - pos3[0] += half_size; + pos3[0] += right_adj; pos3[1] += half_size; tri.det = header->det; /* only the sign matters */ -- cgit v1.2.3 From 34abb93ea10855840a86695a97ccbc89e6d2b9f8 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 11:41:17 -0600 Subject: remove unused var --- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 674e02d0f4..9f85f7c30c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -151,7 +151,7 @@ static void shade_begin(struct quad_stage *qs) { struct quad_shade_stage *qss = quad_shade_stage(qs); struct softpipe_context *softpipe = qs->softpipe; - unsigned i, entry; + unsigned i; /* set TGSI sampler state that varies */ for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { -- cgit v1.2.3 From cd4d732773e06e462e78b8f5bc9f3f1552a198ac Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 11:41:31 -0600 Subject: add point_sprite flag to rasterizer state --- src/mesa/pipe/p_state.h | 1 + src/mesa/state_tracker/st_atom_rasterizer.c | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index daa74ec053..fd3f4f22b9 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -85,6 +85,7 @@ struct pipe_rasterizer_state unsigned poly_smooth:1; unsigned poly_stipple_enable:1; unsigned point_smooth:1; + unsigned point_sprite:1; unsigned multisample:1; /* XXX maybe more ms state in future */ unsigned line_smooth:1; unsigned line_stipple_enable:1; diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index e0d83ddaea..9fbd7dc09e 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -188,6 +188,7 @@ static void update_raster_state( struct st_context *st ) */ raster.point_size = ctx->Point.Size; raster.point_smooth = ctx->Point.SmoothFlag; + raster.point_sprite = ctx->Point.PointSprite; /* _NEW_LINE */ -- cgit v1.2.3 From 22e5c4f0f09cff64039b171c5cee6def07395e1f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 11:59:26 -0600 Subject: implement point sprite mode --- src/mesa/pipe/draw/draw_validate.c | 3 +- src/mesa/pipe/draw/draw_wide_prims.c | 80 +++++++++++++++++++++--------------- 2 files changed, 48 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index 5ec581d8b3..fdeb1be8ca 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -52,7 +52,8 @@ static void validate_begin( struct draw_stage *stage ) */ if (draw->rasterizer->line_width != 1.0 || - draw->rasterizer->point_size != 1.0) { + draw->rasterizer->point_size != 1.0 || + draw->rasterizer->point_sprite) { draw->pipeline.wide->next = next; next = draw->pipeline.wide; } diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index e7e90fabae..0d02e70814 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -39,6 +39,9 @@ struct wide_stage { float half_line_width; float half_point_size; + + uint texcoord[PIPE_MAX_SHADER_OUTPUTS]; + uint num_texcoords; }; @@ -120,10 +123,30 @@ static void wide_line( struct draw_stage *stage, } +static void set_texcoords(const struct wide_stage *wide, + struct vertex_header *v, const float tc[4]) +{ + uint i; + for (i = 0; i < wide->num_texcoords; i++) { + uint j = wide->texcoord[i]; + v->data[j][0] = tc[0]; + v->data[j][1] = tc[1]; + v->data[j][2] = tc[2]; + v->data[j][3] = tc[3]; + } +} + + +/* If there are lots of sprite points (and why wouldn't there be?) it + * would probably be more sensible to change hardware setup to + * optimize this rather than doing the whole thing in software like + * this. + */ static void wide_point( struct draw_stage *stage, struct prim_header *header ) { const struct wide_stage *wide = wide_stage(stage); + const boolean sprite = stage->draw->rasterizer->point_sprite; float half_size = wide->half_point_size; float left_adj, right_adj; @@ -155,6 +178,17 @@ static void wide_point( struct draw_stage *stage, pos3[0] += right_adj; pos3[1] += half_size; + if (sprite) { + static const float tex00[4] = { 0, 0, 0, 1 }; + static const float tex01[4] = { 0, 1, 0, 1 }; + static const float tex11[4] = { 1, 1, 0, 1 }; + static const float tex10[4] = { 1, 0, 0, 1 }; + set_texcoords( wide, v0, tex00 ); + set_texcoords( wide, v1, tex01 ); + set_texcoords( wide, v2, tex10 ); + set_texcoords( wide, v3, tex11 ); + } + tri.det = header->det; /* only the sign matters */ tri.v[0] = v0; tri.v[1] = v2; @@ -168,36 +202,6 @@ static void wide_point( struct draw_stage *stage, } -/* If there are lots of sprite points (and why wouldn't there be?) it - * would probably be more sensible to change hardware setup to - * optimize this rather than doing the whole thing in software like - * this. - */ -static void sprite_point( struct draw_stage *stage, - struct prim_header *header ) -{ -#if 0 - struct vertex_header *v[4]; - struct vertex_fetch *vf = stage->pipe->draw->vb.vf; - - static const float tex00[4] = { 0, 0, 0, 1 }; - static const float tex01[4] = { 0, 1, 0, 1 }; - static const float tex11[4] = { 1, 1, 0, 1 }; - static const float tex10[4] = { 1, 0, 0, 1 }; - - make_wide_point(stage, header->v[0], &v[0] ); - - set_texcoord( vf, v[0], tex00 ); - set_texcoord( vf, v[1], tex01 ); - set_texcoord( vf, v[2], tex10 ); - set_texcoord( vf, v[3], tex11 ); - - quad( stage->next, v[0], v[1], v[2], v[3] ); -#endif -} - - - static void wide_begin( struct draw_stage *stage ) { struct wide_stage *wide = wide_stage(stage); @@ -213,16 +217,24 @@ static void wide_begin( struct draw_stage *stage ) wide->stage.line = passthrough_line; } - if (0/*draw->state.point_sprite*/) { - wide->stage.point = sprite_point; - } - else if (draw->rasterizer->point_size != 1.0) { + if (draw->rasterizer->point_size != 1.0) { wide->stage.point = wide_point; } else { wide->stage.point = passthrough_point; } + if (draw->rasterizer->point_sprite) { + /* find vertex shader texcoord outputs */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i, j = 0; + for (i = 0; i < vs->state->num_outputs; i++) { + if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { + wide->texcoord[j++] = i; + } + } + wide->num_texcoords = j; + } stage->next->begin( stage->next ); } -- cgit v1.2.3 From 1b4852345954af9b582b03a91a3d8399b8fb0e92 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 12:10:30 -0600 Subject: add support for sprite texcoord modes --- src/mesa/pipe/draw/draw_wide_prims.c | 27 ++++++++++++++++++++------- src/mesa/pipe/p_defines.h | 7 +++++++ src/mesa/pipe/p_state.h | 1 + src/mesa/state_tracker/st_atom_rasterizer.c | 12 ++++++++++++ 4 files changed, 40 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 0d02e70814..917944a611 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -40,7 +40,8 @@ struct wide_stage { float half_line_width; float half_point_size; - uint texcoord[PIPE_MAX_SHADER_OUTPUTS]; + uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; + uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; uint num_texcoords; }; @@ -123,16 +124,26 @@ static void wide_line( struct draw_stage *stage, } +/** + * Set the vertex texcoords for sprite mode. + * Coords may be left untouched or set to a right-side-up or upside-down + * orientation. + */ static void set_texcoords(const struct wide_stage *wide, struct vertex_header *v, const float tc[4]) { uint i; for (i = 0; i < wide->num_texcoords; i++) { - uint j = wide->texcoord[i]; - v->data[j][0] = tc[0]; - v->data[j][1] = tc[1]; - v->data[j][2] = tc[2]; - v->data[j][3] = tc[3]; + if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { + uint j = wide->texcoord_slot[i]; + v->data[j][0] = tc[0]; + if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) + v->data[j][1] = 1.0 - tc[1]; + else + v->data[j][1] = tc[1]; + v->data[j][2] = tc[2]; + v->data[j][3] = tc[3]; + } } } @@ -230,7 +241,9 @@ static void wide_begin( struct draw_stage *stage ) uint i, j = 0; for (i = 0; i < vs->state->num_outputs; i++) { if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { - wide->texcoord[j++] = i; + wide->texcoord_slot[j] = i; + wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; + j++; } } wide->num_texcoords = j; diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index d336f83998..9281d40bd3 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -308,4 +308,11 @@ #define PIPE_QUERY_TYPES 3 +/** + * Point sprite coord modes + */ +#define PIPE_SPRITE_COORD_NONE 0 +#define PIPE_SPRITE_COORD_UPPER_LEFT 1 +#define PIPE_SPRITE_COORD_LOWER_LEFT 2 + #endif diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index fd3f4f22b9..ecbcbd098a 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -96,6 +96,7 @@ struct pipe_rasterizer_state float point_size; /**< used when no per-vertex size */ float offset_units; float offset_scale; + ubyte sprite_coord_mode[PIPE_MAX_SHADER_OUTPUTS]; /**< PIPE_SPRITE_COORD_ */ }; diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 9fbd7dc09e..9f857e2837 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -74,6 +74,7 @@ static void update_raster_state( struct st_context *st ) GLcontext *ctx = st->ctx; struct pipe_rasterizer_state raster; const struct cso_rasterizer *cso; + uint i; memset(&raster, 0, sizeof(raster)); @@ -189,6 +190,17 @@ static void update_raster_state( struct st_context *st ) raster.point_size = ctx->Point.Size; raster.point_smooth = ctx->Point.SmoothFlag; raster.point_sprite = ctx->Point.PointSprite; + for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) { + if (ctx->Point.CoordReplace[i]) { + if (ctx->Point.SpriteOrigin == GL_UPPER_LEFT) + raster.sprite_coord_mode[i] = PIPE_SPRITE_COORD_UPPER_LEFT; + else + raster.sprite_coord_mode[i] = PIPE_SPRITE_COORD_LOWER_LEFT; + } + else { + raster.sprite_coord_mode[i] = PIPE_SPRITE_COORD_NONE; + } + } /* _NEW_LINE */ -- cgit v1.2.3 From beefc6011bce9e99cb46430186de1c13f027cb05 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 12:19:54 -0600 Subject: new flag to control psize (from vertex shader or fixed size) --- src/mesa/pipe/draw/draw_wide_prims.c | 25 ++++++++++++++++++++++++- src/mesa/pipe/p_state.h | 1 + src/mesa/state_tracker/st_atom_rasterizer.c | 1 + 3 files changed, 26 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 917944a611..e61cc2e025 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -43,6 +43,8 @@ struct wide_stage { uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; uint num_texcoords; + + int psize_slot; }; @@ -158,7 +160,7 @@ static void wide_point( struct draw_stage *stage, { const struct wide_stage *wide = wide_stage(stage); const boolean sprite = stage->draw->rasterizer->point_sprite; - float half_size = wide->half_point_size; + float half_size; float left_adj, right_adj; struct prim_header tri; @@ -174,6 +176,14 @@ static void wide_point( struct draw_stage *stage, float *pos2 = v2->data[0]; float *pos3 = v3->data[0]; + /* point size is either per-vertex or fixed size */ + if (wide->psize_slot >= 0) { + half_size = header->v[0]->data[wide->psize_slot][0]; + } + else { + half_size = wide->half_point_size; + } + left_adj = -half_size + 0.25; right_adj = half_size + 0.25; @@ -248,6 +258,19 @@ static void wide_begin( struct draw_stage *stage ) } wide->num_texcoords = j; } + + wide->psize_slot = -1; + if (draw->rasterizer->point_size_per_vertex) { + /* find PSIZ vertex output */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i; + for (i = 0; i < vs->state->num_outputs; i++) { + if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + wide->psize_slot = i; + break; + } + } + } stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index ecbcbd098a..4a18d25ab8 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -86,6 +86,7 @@ struct pipe_rasterizer_state unsigned poly_stipple_enable:1; unsigned point_smooth:1; unsigned point_sprite:1; + unsigned point_size_per_vertex:1; /**< size computed in vertex shader */ unsigned multisample:1; /* XXX maybe more ms state in future */ unsigned line_smooth:1; unsigned line_stipple_enable:1; diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 9f857e2837..2a7128dd27 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -201,6 +201,7 @@ static void update_raster_state( struct st_context *st ) raster.sprite_coord_mode[i] = PIPE_SPRITE_COORD_NONE; } } + raster.point_size_per_vertex = ctx->VertexProgram.PointSizeEnabled; /* _NEW_LINE */ -- cgit v1.2.3 From 0a3eaeadb96b313df0ac1e45eba07d39f432db72 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 17:20:08 -0600 Subject: don't use GL types for get/put_tile() params --- src/mesa/drivers/x11/xm_surface.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 562ec76636..42d5d6bbc4 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -78,12 +78,12 @@ xmesa_rb(struct softpipe_surface *sps) static void get_tile(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, GLfloat *p) + uint x, uint y, uint w, uint h, float *p) { struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); GLubyte tmp[MAX_WIDTH * 4]; GLuint i, j; - unsigned w0 = w; + uint w0 = w; GET_CURRENT_CONTEXT(ctx); CLIP_TILE; @@ -101,12 +101,12 @@ get_tile(struct pipe_surface *ps, static void put_tile(struct pipe_surface *ps, - GLuint x, GLuint y, GLuint w, GLuint h, const GLfloat *p) + uint x, uint y, uint w, uint h, const float *p) { struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); GLubyte tmp[MAX_WIDTH * 4]; GLuint i, j; - unsigned w0 = w; + uint w0 = w; GET_CURRENT_CONTEXT(ctx); CLIP_TILE; FLIP(y); -- cgit v1.2.3 From 455a08d87d6f838a92bb9260c9f27757e3740593 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 17:20:56 -0600 Subject: fix masking bug, memory leak --- src/mesa/state_tracker/st_cb_accum.c | 57 ++++++++++++++++++------------------ 1 file changed, 28 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index ddf9dc2adf..7c154baa15 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -81,6 +81,8 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); + free(accBuf); + pipe->region_unmap(pipe, acc_ps->region); } @@ -107,7 +109,7 @@ accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); - _mesa_free(accBuf); + free(accBuf); pipe->region_unmap(pipe, acc_ps->region); } @@ -138,8 +140,8 @@ accum_accum(struct pipe_context *pipe, GLfloat value, acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); - _mesa_free(colorBuf); - _mesa_free(accBuf); + free(colorBuf); + free(accBuf); pipe->region_unmap(pipe, color_ps->region); pipe->region_unmap(pipe, acc_ps->region); @@ -168,7 +170,7 @@ accum_load(struct pipe_context *pipe, GLfloat value, acc_ps->put_tile(acc_ps, xpos, ypos, width, height, buf); - _mesa_free(buf); + free(buf); pipe->region_unmap(pipe, color_ps->region); pipe->region_unmap(pipe, acc_ps->region); @@ -182,42 +184,39 @@ accum_return(GLcontext *ctx, GLfloat value, struct pipe_surface *color_ps) { struct pipe_context *pipe = ctx->st->pipe; - const GLboolean writeR = ctx->Color.ColorMask[0]; - const GLboolean writeG = ctx->Color.ColorMask[1]; - const GLboolean writeB = ctx->Color.ColorMask[2]; - const GLboolean writeA = ctx->Color.ColorMask[3]; - GLfloat *buf; - GLint i; + const GLubyte *colormask = ctx->Color.ColorMask; + GLfloat *abuf, *cbuf = NULL; + GLint i, ch; - buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + abuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); (void) pipe->region_map(pipe, color_ps->region); (void) pipe->region_map(pipe, acc_ps->region); - acc_ps->get_tile(acc_ps, xpos, ypos, width, height, buf); + acc_ps->get_tile(acc_ps, xpos, ypos, width, height, abuf); + + if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) { + cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + color_ps->get_tile(color_ps, xpos, ypos, width, height, cbuf); + } for (i = 0; i < width * height; i++) { - if (writeR) { - GLfloat r = buf[i * 4 + 0] * value; - buf[i * 4 + 0] = CLAMP(r, 0.0, 1.0); - } - if (writeG) { - GLfloat g = buf[i * 4 + 1] * value; - buf[i * 4 + 1] = CLAMP(g, 0.0, 1.0); - } - if (writeB) { - GLfloat b = buf[i * 4 + 2] * value; - buf[i * 4 + 2] = CLAMP(b, 0.0, 1.0); - } - if (writeA) { - GLfloat a = buf[i * 4 + 3] * value; - buf[i * 4 + 3] = CLAMP(a, 0.0, 1.0); + for (ch = 0; ch < 4; ch++) { + if (colormask[ch]) { + GLfloat val = abuf[i * 4 + ch] * value; + abuf[i * 4 + ch] = CLAMP(val, 0.0, 1.0); + } + else { + abuf[i * 4 + ch] = cbuf[i * 4 + ch]; + } } } - color_ps->put_tile(color_ps, xpos, ypos, width, height, buf); + color_ps->put_tile(color_ps, xpos, ypos, width, height, abuf); - _mesa_free(buf); + free(abuf); + if (cbuf) + free(cbuf); pipe->region_unmap(pipe, color_ps->region); pipe->region_unmap(pipe, acc_ps->region); -- cgit v1.2.3 From 588c91eb0a93bdb1ae2819db63b081c9d78833d1 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 22 Oct 2007 17:50:00 -0600 Subject: don't apply ColorMask to main gc --- src/mesa/drivers/x11/xm_dd.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index f02b40bdc0..3ee44bfbaf 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -163,7 +163,6 @@ index_mask( GLcontext *ctx, GLuint mask ) m = (unsigned long) mask; } XMesaSetPlaneMask( xmesa->display, xmbuf->cleargc, m ); - XMesaSetPlaneMask( xmesa->display, xmbuf->gc, m ); } } @@ -195,7 +194,6 @@ color_mask(GLcontext *ctx, if (bmask) m |= GET_BLUEMASK(xmesa->xm_visual); } XMesaSetPlaneMask( xmesa->display, xmbuf->cleargc, m ); - XMesaSetPlaneMask( xmesa->display, xmbuf->gc, m ); } } @@ -454,13 +452,13 @@ xmesa_DrawPixels_8R8G8B( GLcontext *ctx, ctx->_ImageTransferState == 0 && /* no color tables, scale/bias, etc */ ctx->Pixel.ZoomX == 1.0 && /* no zooming */ ctx->Pixel.ZoomY == 1.0 && - xrb->pixmap && + xrb->pixmap && /* drawing to pixmap or window */ xrb->St.Base.AlphaBits == 0) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); XMesaDisplay *dpy = xmesa->xm_visual->display; - const XMesaGC gc = xmbuf->gc; + const XMesaGC gc = xmbuf->cleargc; /* effected by glColorMask */ int dstX = x; int dstY = y; int w = width; @@ -555,7 +553,7 @@ xmesa_DrawPixels_5R6G5B( GLcontext *ctx, const SWcontext *swrast = SWRAST_CONTEXT( ctx ); XMesaDisplay *dpy = xmesa->xm_visual->display; XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - const XMesaGC gc = xmbuf->gc; + const XMesaGC gc = xmbuf->cleargc; /* effected by glColorMask */ ASSERT(dpy); ASSERT(gc); @@ -564,7 +562,7 @@ xmesa_DrawPixels_5R6G5B( GLcontext *ctx, if (swrast->NewState) _swrast_validate_derived( ctx ); - if (xrb->pixmap && + if (xrb->pixmap && /* drawing to pixmap or window */ format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 && !ctx->Color.DitherFlag && /* no dithering */ @@ -655,7 +653,8 @@ xmesa_CopyPixels( GLcontext *ctx, const XMesaContext xmesa = XMESA_CONTEXT(ctx); const SWcontext *swrast = SWRAST_CONTEXT( ctx ); XMesaDisplay *dpy = xmesa->xm_visual->display; - const XMesaGC gc = ((XMesaBuffer) ctx->DrawBuffer)->gc; + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + const XMesaGC gc = xmbuf->cleargc; /* effected by glColorMask */ struct xmesa_renderbuffer *srcXrb = xmesa_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer->Wrapped); struct xmesa_renderbuffer *dstXrb -- cgit v1.2.3 From 5c79c088cd0a2c512891b87b67a3c4f810595658 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 08:30:36 -0600 Subject: Don't pad renderbuffers to multiple of two pixels anymore. This was only needed to avoid out-of-bounds memory accesses with the 2x2 quad_read/write() functions which no longer exist. --- src/mesa/state_tracker/st_cb_fbo.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 64a64fd6b9..94e286feab 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -64,7 +64,6 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, const struct pipe_format_info *info = st_get_format_info(pipeFormat); GLuint cpp; GLbitfield flags = PIPE_SURFACE_FLAG_RENDER; /* want to render to surface */ - GLuint width2, height2; assert(info); if (!info) @@ -99,11 +98,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe->region_release(pipe, &strb->surface->region); } - /* Softpipe operates on quads, so pad dimensions to multiples of 2 */ - width2 = (width + 1) & ~1; - height2 = (height + 1) & ~1; - - strb->surface->region = pipe->region_alloc(pipe, cpp, width2, height2, flags); + strb->surface->region = pipe->region_alloc(pipe, cpp, width, height, flags); if (!strb->surface->region) return GL_FALSE; /* out of memory, try s/w buffer? */ -- cgit v1.2.3 From e69943e6dda102df8418a8261b95155350181a2f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 10:23:01 -0600 Subject: bump up MAX_INSTRUCTIONS and add an assertion to catch emitting too many instructions --- src/mesa/main/texenvprogram.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index a554c033c4..2614440a74 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -36,10 +36,11 @@ #include "texenvprogram.h" /** - * According to Glean's texCombine test, no more than 21 instructions - * are needed. Allow a few extra just in case. + * This MAX is probably a bit generous, but that's OK. There can be + * up to four instructions per texture unit (TEX + 3 for combine), + * then there's fog and specular add. */ -#define MAX_INSTRUCTIONS 24 +#define MAX_INSTRUCTIONS ((MAX_TEXTURE_UNITS * 4) + 12) #define DISASSEM (MESA_VERBOSE & VERBOSE_DISASSEM) @@ -478,6 +479,8 @@ emit_op(struct texenv_fragment_program *p, GLuint nr = p->program->Base.NumInstructions++; struct prog_instruction *inst = &p->program->Base.Instructions[nr]; + assert(nr < MAX_INSTRUCTIONS); + _mesa_init_instructions(inst, 1); inst->Opcode = op; -- cgit v1.2.3 From 112a1580f658948e553fe04399a20958dca67c16 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 10:54:50 -0600 Subject: properly init dst reg's CondMask/Swizzle fields --- src/mesa/main/texenvprogram.c | 4 ++-- src/mesa/tnl/t_vp_build.c | 6 +++--- 2 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index 2614440a74..abc2567134 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -462,8 +462,8 @@ static void emit_dst( struct prog_dst_register *dst, dst->File = ureg.file; dst->Index = ureg.idx; dst->WriteMask = mask; - dst->CondMask = 0; - dst->CondSwizzle = 0; + dst->CondMask = COND_TR; /* always pass cond test */ + dst->CondSwizzle = SWIZZLE_NOOP; } static struct prog_instruction * diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index cbffed5155..dc2d5e0065 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -488,8 +488,8 @@ static void emit_dst( struct prog_dst_register *dst, dst->Index = reg.idx; /* allow zero as a shorthand for xyzw */ dst->WriteMask = mask ? mask : WRITEMASK_XYZW; - dst->CondMask = COND_TR; - dst->CondSwizzle = 0; + dst->CondMask = COND_TR; /* always pass cond test */ + dst->CondSwizzle = SWIZZLE_NOOP; dst->CondSrc = 0; dst->pad = 0; } @@ -512,7 +512,7 @@ static void debug_insn( struct prog_instruction *inst, const char *fn, static void emit_op3fn(struct tnl_program *p, - GLuint op, + enum prog_opcode op, struct ureg dest, GLuint mask, struct ureg src0, -- cgit v1.2.3 From be049999829d21a1c9ec9a2c616e99186208266b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 11:38:17 -0600 Subject: adjust coords in wide_line() to be conformant --- src/mesa/pipe/draw/draw_wide_prims.c | 55 ++++++++++++++++++++++++++++++------ 1 file changed, 47 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index e61cc2e025..76b8a5319b 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -100,17 +100,56 @@ static void wide_line( struct draw_stage *stage, const float dx = FABSF(pos0[0] - pos2[0]); const float dy = FABSF(pos0[1] - pos2[1]); + /* + * Draw wide line as a quad (two tris) by "stretching" the line along + * X or Y. + * XXX For AA lines, the quad corners have to be computed in a + * more sophisticated way. + */ + + /* need to tweak coords in several ways to be conformant here */ + if (dx > dy) { - pos0[1] -= half_width; - pos1[1] += half_width; - pos2[1] -= half_width; - pos3[1] += half_width; + /* x-major line */ + pos0[1] = pos0[1] - half_width - 0.25; + pos1[1] = pos1[1] + half_width - 0.25; + pos2[1] = pos2[1] - half_width - 0.25; + pos3[1] = pos3[1] + half_width - 0.25; + if (pos0[0] < pos2[0]) { + /* left to right line */ + pos0[0] -= 0.5; + pos1[0] -= 0.5; + pos2[0] -= 0.5; + pos3[0] -= 0.5; + } + else { + /* right to left line */ + pos0[0] += 0.5; + pos1[0] += 0.5; + pos2[0] += 0.5; + pos3[0] += 0.5; + } } else { - pos0[0] -= half_width; - pos1[0] += half_width; - pos2[0] -= half_width; - pos3[0] += half_width; + /* y-major line */ + pos0[0] = pos0[0] - half_width + 0.25; + pos1[0] = pos1[0] + half_width + 0.25; + pos2[0] = pos2[0] - half_width + 0.25; + pos3[0] = pos3[0] + half_width + 0.25; + if (pos0[1] < pos2[1]) { + /* top to bottom line */ + pos0[1] -= 0.5; + pos1[1] -= 0.5; + pos2[1] -= 0.5; + pos3[1] -= 0.5; + } + else { + /* bottom to top line */ + pos0[1] += 0.5; + pos1[1] += 0.5; + pos2[1] += 0.5; + pos3[1] += 0.5; + } } tri.det = header->det; /* only the sign matters */ -- cgit v1.2.3 From 40e46d0727920dffdcc23aba150aa3c37c6ecf65 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 12:32:02 -0600 Subject: In get_vertex(), slot was computed using & 31. Replace with % VCACHE_SIZE. Also, assert that index is not too large before indexing array. --- src/mesa/pipe/draw/draw_vertex_cache.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index 1f4adedede..dc939d6d14 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -57,7 +57,7 @@ void draw_vertex_cache_invalidate( struct draw_context *draw ) static struct vertex_header *get_vertex( struct draw_context *draw, unsigned i ) { - unsigned slot = (i + (i>>5)) & 31; + unsigned slot = (i + (i>>5)) % VCACHE_SIZE; /* Cache miss? */ @@ -74,6 +74,8 @@ static struct vertex_header *get_vertex( struct draw_context *draw, draw->vcache.referenced |= (1 << slot); /* slot now in use */ } + assert(slot < Elements(draw->vcache.idx)); + draw->vcache.idx[slot] = i; /* Add to vertex shader queue: -- cgit v1.2.3 From 3df65af849280863a15fefcb11f8304b83c9ffa0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 15:02:02 -0600 Subject: added comment --- src/mesa/pipe/draw/draw_private.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 7e5d5db5ab..b7811e891e 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -192,7 +192,7 @@ struct draw_context /* Post-tnl vertex cache: */ struct { - unsigned referenced; + unsigned referenced; /**< bitfield */ unsigned idx[VCACHE_SIZE + VCACHE_OVERFLOW]; struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; unsigned overflow; -- cgit v1.2.3 From 4664261f8d8fefa347bf38a224f6584d0fdeebfc Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 15:08:54 -0600 Subject: Fix vertex cache bug that allows multiple vertices to fall into the same slot. Need to set the slot's bit in draw->vcache.referenced even when there was a cache hit since flushing the primitive buffer will have cleared the bitfield but not the cache's vertex indexes. Fixes a bug found when drawing long triangle fans but could be hit by other prim types as well. An alternate fix would be to call draw_vertex_cache_invalidate() from draw_vertex_cache_unreference(). --- src/mesa/pipe/draw/draw_vertex_cache.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index dc939d6d14..56a4c34df5 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -59,6 +59,8 @@ static struct vertex_header *get_vertex( struct draw_context *draw, { unsigned slot = (i + (i>>5)) % VCACHE_SIZE; + assert(slot < 32); /* so we don't exceed the bitfield size below */ + /* Cache miss? */ if (draw->vcache.idx[slot] != i) { @@ -80,6 +82,7 @@ static struct vertex_header *get_vertex( struct draw_context *draw, /* Add to vertex shader queue: */ + assert(draw->vs.queue_nr < VS_QUEUE_LENGTH); draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; draw->vs.queue[draw->vs.queue_nr].elt = i; draw->vs.queue_nr++; @@ -94,8 +97,14 @@ static struct vertex_header *get_vertex( struct draw_context *draw, } else { // fprintf(stderr, "*"); + /* primitive flushing may have cleared the bitfield but did not + * clear the idx[] array values. Set the bit now. This fixes a + * bug found when drawing long triangle fans. + */ + draw->vcache.referenced |= (1 << slot); } + return draw->vcache.vertex[slot]; } -- cgit v1.2.3 From b2e529982eb702ea039f6436c9dece39401a4c9c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 8 Oct 2007 10:24:35 -0400 Subject: Initial stab at LLVM integration. --- src/mesa/Makefile | 4 +- src/mesa/pipe/llvm/llvmtgsi.cpp | 158 ++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/llvmtgsi.h | 27 ++++++ src/mesa/pipe/p_state.h | 3 + src/mesa/pipe/softpipe/sp_quad_fs.c | 3 + src/mesa/sources | 10 ++- src/mesa/state_tracker/st_program.c | 4 +- 7 files changed, 205 insertions(+), 4 deletions(-) create mode 100644 src/mesa/pipe/llvm/llvmtgsi.cpp create mode 100644 src/mesa/pipe/llvm/llvmtgsi.h (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 9dde96ee21..cb7c466f47 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -16,10 +16,10 @@ SOFTPIPE_LIB = $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a .SUFFIXES : .cpp .c.o: - $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + $(CC) -c $(LLVM_CFLAGS) $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ .cpp.o: - $(CXX) -c $(INCLUDE_DIRS) $(CXXFLAGS) $< -o $@ + $(CXX) -c $(LLVM_CXXFLAGS) $(INCLUDE_DIRS) $(CXXFLAGS) $< -o $@ .S.o: $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp new file mode 100644 index 0000000000..857eb1cce1 --- /dev/null +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -0,0 +1,158 @@ +#include "llvmtgsi.h" + +#include "pipe/tgsi/exec/tgsi_exec.h" +#include "pipe/tgsi/exec/tgsi_token.h" +#include "pipe/tgsi/exec/tgsi_build.h" +#include "pipe/tgsi/exec/tgsi_util.h" +#include "pipe/tgsi/exec/tgsi_parse.h" +#include "pipe/tgsi/exec/tgsi_dump.h" +//#include "pipe/tgsi/tgsi_platform.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + + +static void +translate_declaration(llvm::Module *module, + struct tgsi_full_declaration *decl, + struct tgsi_full_declaration *fd) +{ +} + + +static void +translate_immediate(llvm::Module *module, + struct tgsi_full_immediate *imm) +{ +} + + +static void +translate_instruction(llvm::Module *module, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi) +{ +} + + +static llvm::Module * +tgsi_to_llvm(const struct tgsi_token *tokens) +{ + llvm::Module *mod = new llvm::Module("tgsi"); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + + tgsi_parse_init(&parse, tokens); + + //parse.FullHeader.Processor.Processor + + //parse.FullVersion.Version.MajorVersion + //parse.FullVersion.Version.MinorVersion + + //parse.FullHeader.Header.HeaderSize + //parse.FullHeader.Header.BodySize + //parse.FullHeader.Processor.Processor + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + fprintf(stderr, "Translating %d\n", parse.FullToken.Token.Type); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declaration(mod, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediate(mod, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + translate_instruction(mod, + &parse.FullToken.FullInstruction, + &fi); + break; + + default: + assert(0); + } + } + + //TXT("\ntgsi-dump end -------------------\n"); + + tgsi_parse_free(&parse); + + return mod; +} + +struct ga_llvm_prog * +ga_llvm_from_tgsi(const struct tgsi_token *tokens) +{ + std::cout << "Creating llvm " <module = mod; + ga_llvm->engine = 0;//ee; + fprintf(stderr, "DUMPX \n"); + //tgsi_dump(tokens, TGSI_DUMP_VERBOSE); + tgsi_dump(tokens, 0); + fprintf(stderr, "DUMPEND \n"); + + return ga_llvm; +} + +void ga_llvm_prog_delete(struct ga_llvm_prog *prog) +{ + llvm::Module *mod = static_cast(prog->module); + delete mod; + prog->module = 0; + prog->engine = 0; + free(prog); +} + +int ga_llvm_prog_exec(struct ga_llvm_prog *prog) +{ + //std::cout << "START "<(prog->module); + llvm::Function *func = mod->getFunction("main"); + llvm::ExecutionEngine *ee = static_cast(prog->engine); + + std::vector args(0); + //args[0] = GenericValue(&st); + //std::cout << "Mod is "<<*mod; + //std::cout << "\n\nRunning llvm: " << std::endl; + if (func) { + std::cout << "Func is "<runFunction(func, args); + } + +//delete ee; +//delete mp; + + return 0; +} diff --git a/src/mesa/pipe/llvm/llvmtgsi.h b/src/mesa/pipe/llvm/llvmtgsi.h new file mode 100644 index 0000000000..c0cee915b9 --- /dev/null +++ b/src/mesa/pipe/llvm/llvmtgsi.h @@ -0,0 +1,27 @@ +#ifndef LLVMTGSI_H +#define LLVMTGSI_H + +#if defined __cplusplus +extern "C" { +#endif + +struct tgsi_exec_machine; +struct tgsi_token; +struct tgsi_sampler; + +struct ga_llvm_prog { + void *module; + void *engine; +}; +struct ga_llvm_prog * +ga_llvm_from_tgsi(const struct tgsi_token *tokens); + +void ga_llvm_prog_delete(struct ga_llvm_prog *prog); + +int ga_llvm_prog_exec(struct ga_llvm_prog *prog); + +#if defined __cplusplus +} // extern "C" +#endif + +#endif diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 4a18d25ab8..69b11588e0 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -58,6 +58,7 @@ /* fwd decl */ struct pipe_surface; +struct ga_llvm_prog; /* opaque type */ struct pipe_buffer_handle; @@ -153,6 +154,8 @@ struct pipe_shader_state { #endif void (*executable)(); + const struct ga_llvm_prog *llvm_prog; + ubyte num_inputs; ubyte num_outputs; ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 9f85f7c30c..035f05170b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -35,6 +35,8 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" +#include "pipe/llvm/llvmtgsi.h" + #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" @@ -107,6 +109,7 @@ shade_quad( machine->InterpCoefs ); } else { + ga_llvm_prog_exec(softpipe->fs->llvm_prog); quad->mask &= tgsi_exec_machine_run( machine ); } diff --git a/src/mesa/sources b/src/mesa/sources index fea3234db3..41c3ae387e 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -188,6 +188,11 @@ TGSIDECO_SOURCES = \ TGSIMESA_SOURCES = \ pipe/tgsi/mesa/mesa_to_tgsi.c +ifeq ($(MESA_NO_LLVM),0) + LLVMTGSI_SOURCES = \ + pipe/llvm/llvmtgsi.cpp +endif + STATECACHE_SOURCES = \ pipe/cso_cache/cso_hash.c \ pipe/cso_cache/cso_cache.c @@ -366,6 +371,7 @@ FBDEV_DRIVER_SOURCES = \ ALL_SOURCES = \ $(GLAPI_SOURCES) \ $(SOLO_SOURCES) \ + $(LLVMTGSI_SOURCES) \ $(ASM_SOURCES) \ $(COMMON_DRIVER_SOURCES)\ $(X11_DRIVER_SOURCES) \ @@ -392,13 +398,15 @@ SOLO_SOURCES = \ CORE_SOURCES = \ $(GLAPI_SOURCES) \ - $(SOLO_SOURCES) + $(SOLO_SOURCES) \ + $(LLVMTGSI_SOURCES) ### Object files SOLO_OBJECTS = \ $(SOLO_SOURCES:.c=.o) \ + $(LLVMTGSI_SOURCES:.cpp=.o) \ $(ASM_SOURCES:.S=.o) GLAPI_OBJECTS = \ diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 5b6fa70dff..edb2703101 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -39,6 +39,7 @@ #include "pipe/draw/draw_context.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/exec/tgsi_core.h" +#include "pipe/llvm/llvmtgsi.h" #include "st_context.h" #include "st_cache.h" @@ -250,6 +251,7 @@ st_translate_vertex_program(struct st_context *st, tokensOut, maxTokens); vs.tokens = tokensOut; + vs.llvm_prog = ga_llvm_from_tgsi(vs.tokens); cso = st_cached_vs_state(st, &vs); stvp->vs = cso; @@ -405,7 +407,7 @@ st_translate_fragment_program(struct st_context *st, tokensOut, maxTokens); fs.tokens = tokensOut; - + fs.llvm_prog = ga_llvm_from_tgsi(fs.tokens); cso = st_cached_fs_state(st, &fs); stfp->fs = cso; -- cgit v1.2.3 From 11bc1f015a781760c419bdd53b326132b5146971 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 10 Oct 2007 07:21:02 -0400 Subject: Stub out some conversion. --- src/mesa/pipe/llvm/llvmtgsi.cpp | 291 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 291 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 857eb1cce1..2221a2c353 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -29,6 +29,7 @@ translate_declaration(llvm::Module *module, struct tgsi_full_declaration *decl, struct tgsi_full_declaration *fd) { + /* i think this is going to be a noop */ } @@ -44,6 +45,296 @@ translate_instruction(llvm::Module *module, struct tgsi_full_instruction *inst, struct tgsi_full_instruction *fi) { + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: + break; + case TGSI_OPCODE_MOV: + break; + case TGSI_OPCODE_LIT: + break; + case TGSI_OPCODE_RCP: + break; + case TGSI_OPCODE_RSQ: + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: + break; + case TGSI_OPCODE_ADD: + break; + case TGSI_OPCODE_DP3: + break; + case TGSI_OPCODE_DP4: + break; + case TGSI_OPCODE_DST: + break; + case TGSI_OPCODE_MIN: + break; + case TGSI_OPCODE_MAX: + break; + case TGSI_OPCODE_SLT: + break; + case TGSI_OPCODE_SGE: + break; + case TGSI_OPCODE_MAD: + break; + case TGSI_OPCODE_SUB: + break; + case TGSI_OPCODE_LERP: + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: + break; + case TGSI_OPCODE_LOGBASE2: + break; + case TGSI_OPCODE_POWER: + break; + case TGSI_OPCODE_CROSSPRODUCT: + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: + break; + case TGSI_OPCODE_COS: + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: + break; + case TGSI_OPCODE_SIN: + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_TXP: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: + break; + case TGSI_OPCODE_RET: + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: + break; + case TGSI_OPCODE_SCS: + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: + break; + case TGSI_OPCODE_IF: + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: + break; + case TGSI_OPCODE_ENDIF: + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: + break; + case TGSI_OPCODE_BGNSUB: + break; + case TGSI_OPCODE_ENDLOOP2: + break; + case TGSI_OPCODE_ENDSUB: + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + /*TXT( "_SAT" );*/ + break; + case TGSI_SAT_MINUS_PLUS_ONE: + /*TXT( "_SAT[-1,1]" );*/ + break; + default: + assert( 0 ); + } } -- cgit v1.2.3 From 2dbba8b024720c11cb2d812b5ccb61ecb9887faa Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 12 Oct 2007 13:57:53 -0400 Subject: Implement the conversion and do the initial execution pass. --- src/mesa/pipe/draw/draw_private.h | 1 + src/mesa/pipe/draw/draw_vertex_fetch.c | 4 +- src/mesa/pipe/draw/draw_vertex_shader.c | 13 +-- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 129 +++++++++++++++++++++++++++ src/mesa/pipe/llvm/llvm_builtins.c | 90 +++++++++++++++++++ src/mesa/pipe/llvm/llvmtgsi.cpp | 6 +- src/mesa/pipe/llvm/llvmtgsi.h | 6 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/sources | 1 + 9 files changed, 242 insertions(+), 10 deletions(-) create mode 100644 src/mesa/pipe/draw/draw_vertex_shader_llvm.c create mode 100644 src/mesa/pipe/llvm/llvm_builtins.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index b7811e891e..b672cc2e8b 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -251,6 +251,7 @@ extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); +extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw ); struct tgsi_exec_machine; diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index de1cd06da6..1b093b7342 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -96,7 +96,7 @@ void draw_vertex_fetch( struct draw_context *draw, for (j = 0; j < count; j++) { uint attr; - /*printf("fetch vertex %u: \n", j);*/ + printf("fetch vertex %u: \n", j); /* loop over vertex attributes (vertex shader inputs) */ for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { @@ -111,7 +111,7 @@ void draw_vertex_fetch( struct draw_context *draw, fetch_attrib4(src, draw->vertex_element[attr].src_format, p); - /*printf(" %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]);*/ + printf("> %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]); /* Transform to AoS xxxx/yyyy/zzzz/wwww representation: */ diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index fd0183f8cd..874259e727 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -91,6 +91,8 @@ run_vertex_program(struct draw_context *draw, assert(draw->vertex_shader->state->output_semantic_name[0] == TGSI_SEMANTIC_POSITION); + fprintf(stderr, "------ run_vertex\n"); + /* Consts does not require 16 byte alignment. */ machine->Consts = (float (*)[4]) draw->mapped_constants; @@ -161,10 +163,6 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; #if DBG printf("output[%d][%d]: %f %f %f %f\n", j, slot, - vOut[j]->data[slot][0], - vOut[j]->data[slot][1], - vOut[j]->data[slot][2], - vOut[j]->data[slot][3]); #endif } } /* loop over vertices */ @@ -179,7 +177,12 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) { unsigned i, j; -// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); + fprintf(stderr, "XX q(%d) ", draw->vs.queue_nr ); + + if (draw->vertex_shader->state->llvm_prog) { + draw_vertex_shader_queue_flush_llvm(draw); + return; + } /* run vertex shader on vertex cache entries, four per invokation */ for (i = 0; i < draw->vs.queue_nr; i += 4) { diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c new file mode 100644 index 0000000000..b38498efc5 --- /dev/null +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -0,0 +1,129 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#include "pipe/p_util.h" +#include "draw_private.h" +#include "draw_context.h" +#include "draw_vertex.h" + +#include "pipe/llvm/llvmtgsi.h" +#include "pipe/tgsi/exec/tgsi_core.h" + +static INLINE void +fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) +{ + /* defaults */ + attrib[1] = 0.0; + attrib[2] = 0.0; + attrib[3] = 1.0; + switch (format) { + case PIPE_FORMAT_R32G32B32A32_FLOAT: + attrib[3] = ((float *) ptr)[3]; + /* fall-through */ + case PIPE_FORMAT_R32G32B32_FLOAT: + attrib[2] = ((float *) ptr)[2]; + /* fall-through */ + case PIPE_FORMAT_R32G32_FLOAT: + attrib[1] = ((float *) ptr)[1]; + /* fall-through */ + case PIPE_FORMAT_R32_FLOAT: + attrib[0] = ((float *) ptr)[0]; + break; + default: + assert(0); + } +} + + +/** + * Fetch vertex attributes for 'count' vertices. + */ +static INLINE +void vertex_fetch(struct draw_context *draw, + const unsigned elt, + float (*inputs)[4]) +{ + uint attr; + + printf("fetch vertex %u: \n", elt); + + /* loop over vertex attributes (vertex shader inputs) */ + for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { + + unsigned buf = draw->vertex_element[attr].vertex_buffer_index; + const void *src + = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[attr].src_offset + + elt * draw->vertex_buffer[buf].pitch); + float p[4]; + + fetch_attrib4(src, draw->vertex_element[attr].src_format, p); + + printf("> %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]); + + inputs[attr][0] = p[0]; /*X*/ + inputs[attr][1] = p[1]; /*Y*/ + inputs[attr][2] = p[2]; /*Z*/ + inputs[attr][3] = p[3]; /*W*/ + } +} + + +/** + * Called by the draw module when the vertx cache needs to be flushed. + * This involves running the vertex shader. + */ +void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) +{ + unsigned i; + + struct vertex_header *dests[VS_QUEUE_LENGTH]; + float inputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; + float (*consts)[4] = (float (*)[4]) draw->mapped_constants; + struct ga_llvm_prog *prog = draw->vertex_shader->state->llvm_prog; + + fprintf(stderr, "XX q(%d) ", draw->vs.queue_nr); + + /* fetch the inputs */ + for (i = 0; i < draw->vs.queue_nr; ++i) { + unsigned elt = draw->vs.queue[i].elt; + dests[i] = draw->vs.queue[i].dest; + vertex_fetch(draw, elt, inputs[i]); + } + + /* batch execute the shaders on all the vertices */ + ga_llvm_prog_exec(prog, inputs, dests, consts, + draw->vs.queue_nr); + + draw->vs.queue_nr = 0; +} diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c new file mode 100644 index 0000000000..0f0efeb303 --- /dev/null +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -0,0 +1,90 @@ + + +inline float4 compute_clip(float4 vec, float4 scale, float4 trans) +{ + return vec*scale + trans; +} + + +inline float +dot4(const float4 a, const float4 b) +{ + float4 c = a*b; + return c.x + c.y + c.z + c.w; +} + +inline unsigned +compute_clipmask(float4 clip, const float4 (*plane), unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + for (i = 0; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<clip[0] = clip.x; + y = vOut->clip[1] = clip.y; + z = vOut->clip[2] = clip.z; + w = vOut->clip[3] = clip.w; + + vOut[i]->clipmask = compute_clipmask(res0, planes, nr_planes); + vOut[i]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + res0.x = x; res0.y = y; res0.z = z; res0.w = 1; + + /* Viewport mapping */ + res = res * scale + trans; + vOut->data[0][0] = res.x; + vOut->data[0][1] = res.y; + vOut->data[0][2] = res.z; + vOut->data[0][3] = w; + + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + * Skip 0 since we just did it above. + * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. + */ + for (slot = 1; slot < draw->vertex_info.num_attribs - 2; slot++) { + float4 vec = results[slot]; + vOut->data[slot][0] = vec.x; + vOut->data[slot][1] = vec.y; + vOut->data[slot][2] = vec.z; + vOut->data[slot][3] = vec.w; + + printf("output %d: %f %f %f %f\n", slot, + vOut->data[slot][0], + vOut->data[slot][1], + vOut->data[slot][2], + vOut->data[slot][3]); + } +} + +void run_vertex_shader(float ainputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4], + struct vertex_header *dests[VS_QUEUE_LENGTH], + float *aconsts[4] + int count) +{ + float4 inputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS]; + float4 *consts; +} diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 2221a2c353..46b7561b5e 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -426,7 +426,11 @@ void ga_llvm_prog_delete(struct ga_llvm_prog *prog) free(prog); } -int ga_llvm_prog_exec(struct ga_llvm_prog *prog) +int ga_llvm_prog_exec(struct ga_llvm_prog *prog, + float (*inputs)[32][4], + void *dests[16*32*4], + float (*consts)[4], + int count) { //std::cout << "START "<(prog->module); diff --git a/src/mesa/pipe/llvm/llvmtgsi.h b/src/mesa/pipe/llvm/llvmtgsi.h index c0cee915b9..9fbb0ea8f9 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.h +++ b/src/mesa/pipe/llvm/llvmtgsi.h @@ -18,7 +18,11 @@ ga_llvm_from_tgsi(const struct tgsi_token *tokens); void ga_llvm_prog_delete(struct ga_llvm_prog *prog); -int ga_llvm_prog_exec(struct ga_llvm_prog *prog); +int ga_llvm_prog_exec(struct ga_llvm_prog *prog, + float (*inputs)[32][4], + void *dests[16*32*4], + float (*consts)[4], + int count); #if defined __cplusplus } // extern "C" diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 035f05170b..00afcd930a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -109,7 +109,7 @@ shade_quad( machine->InterpCoefs ); } else { - ga_llvm_prog_exec(softpipe->fs->llvm_prog); + //ga_llvm_prog_exec(softpipe->fs->llvm_prog); quad->mask &= tgsi_exec_machine_run( machine ); } diff --git a/src/mesa/sources b/src/mesa/sources index 41c3ae387e..c2544ed322 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -172,6 +172,7 @@ DRAW_SOURCES = \ pipe/draw/draw_vertex_cache.c \ pipe/draw/draw_vertex_fetch.c \ pipe/draw/draw_vertex_shader.c \ + pipe/draw/draw_vertex_shader_llvm.c \ pipe/draw/draw_wide_prims.c TGSIEXEC_SOURCES = \ -- cgit v1.2.3 From 9e6d58fac26a12246e9e560f3802ebcbce2423bc Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 15 Oct 2007 08:12:22 -0400 Subject: Generate the base shader. --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 5 +- src/mesa/pipe/llvm/llvm_builtins.c | 115 ++++++++++++++++++++------ src/mesa/pipe/llvm/llvmtgsi.cpp | 117 ++++++++++++++++++++++----- src/mesa/pipe/llvm/llvmtgsi.h | 4 +- 4 files changed, 194 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index b38498efc5..fd49901051 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -112,7 +112,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) float (*consts)[4] = (float (*)[4]) draw->mapped_constants; struct ga_llvm_prog *prog = draw->vertex_shader->state->llvm_prog; - fprintf(stderr, "XX q(%d) ", draw->vs.queue_nr); + fprintf(stderr, "--- XX q(%d) ", draw->vs.queue_nr); /* fetch the inputs */ for (i = 0; i < draw->vs.queue_nr; ++i) { @@ -123,7 +123,8 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) /* batch execute the shaders on all the vertices */ ga_llvm_prog_exec(prog, inputs, dests, consts, - draw->vs.queue_nr); + draw->vs.queue_nr, + draw->vertex_info.num_attribs); draw->vs.queue_nr = 0; } diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index 0f0efeb303..c7a9ea0d5a 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -1,20 +1,29 @@ +/* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm-dis */ +/* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -funcname=createBaseShader */ +typedef __attribute__(( ocu_vector_type(4) )) float float4; +#if 0 +//clang doesn't suppoer "struct->member" notation yet +struct vertex_header { + unsigned clipmask:12; + unsigned edgeflag:1; + unsigned pad:3; + unsigned vertex_id:16; -inline float4 compute_clip(float4 vec, float4 scale, float4 trans) -{ - return vec*scale + trans; -} + float clip[4]; + float data[][4]; +}; inline float -dot4(const float4 a, const float4 b) +dot4(float4 a, float4 b) { float4 c = a*b; return c.x + c.y + c.z + c.w; } inline unsigned -compute_clipmask(float4 clip, const float4 (*plane), unsigned nr) +compute_clipmask(float4 clip, float4 (*plane), unsigned nr) { unsigned mask = 0; unsigned i; @@ -29,7 +38,8 @@ compute_clipmask(float4 clip, const float4 (*plane), unsigned nr) inline void collect_results(float4 *results, struct vertex_header *vOut, float4 *planes, int nr_planes, - float4 scale, float4 trans) + float4 scale, float4 trans, + int num_attribs) { /* store results */ unsigned slot; @@ -38,13 +48,14 @@ inline void collect_results(float4 *results, struct vertex_header *vOut, /* Handle attr[0] (position) specially: */ float4 res0 = results[0]; - x = vOut->clip[0] = clip.x; - y = vOut->clip[1] = clip.y; - z = vOut->clip[2] = clip.z; - w = vOut->clip[3] = clip.w; + float *clip = vOut->clip; + x = clip[0] = res0.x; + y = clip[1] = res0.y; + z = clip[2] = res0.z; + w = clip[3] = res0.w; - vOut[i]->clipmask = compute_clipmask(res0, planes, nr_planes); - vOut[i]->edgeflag = 1; + vOut->clipmask = compute_clipmask(res0, planes, nr_planes); + vOut->edgeflag = 1; /* divide by w */ w = 1.0f / w; @@ -54,10 +65,10 @@ inline void collect_results(float4 *results, struct vertex_header *vOut, res0.x = x; res0.y = y; res0.z = z; res0.w = 1; /* Viewport mapping */ - res = res * scale + trans; - vOut->data[0][0] = res.x; - vOut->data[0][1] = res.y; - vOut->data[0][2] = res.z; + res0 = res0 * scale + trans; + vOut->data[0][0] = res0.x; + vOut->data[0][1] = res0.y; + vOut->data[0][2] = res0.z; vOut->data[0][3] = w; /* Remaining attributes are packed into sequential post-transform @@ -65,7 +76,7 @@ inline void collect_results(float4 *results, struct vertex_header *vOut, * Skip 0 since we just did it above. * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. */ - for (slot = 1; slot < draw->vertex_info.num_attribs - 2; slot++) { + for (slot = 1; slot < num_attribs - 2; slot++) { float4 vec = results[slot]; vOut->data[slot][0] = vec.x; vOut->data[slot][1] = vec.y; @@ -79,12 +90,68 @@ inline void collect_results(float4 *results, struct vertex_header *vOut, vOut->data[slot][3]); } } +#endif -void run_vertex_shader(float ainputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4], - struct vertex_header *dests[VS_QUEUE_LENGTH], - float *aconsts[4] - int count) +void from_array(float4 (*res)[32], float (*ainputs)[32][4], + int count, int num_attribs) { - float4 inputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS]; - float4 *consts; + for (int i = 0; i < count; ++i) { + for (int j = 0; j < num_attribs; ++j) { + float4 vec; + vec.x = ainputs[i][j][0]; + vec.y = ainputs[i][j][1]; + vec.z = ainputs[i][j][2]; + vec.w = ainputs[i][j][3]; + res[i][j] = vec; + } + } +} + +void from_consts(float4 *res, float (*ainputs)[4], + int count) +{ + for (int i = 0; i < count; ++i) { + float4 vec; + vec.x = ainputs[i][0]; + vec.y = ainputs[i][1]; + vec.z = ainputs[i][2]; + vec.w = ainputs[i][3]; + res[i] = vec; + } +} + +void to_array(float (*dests)[4], float4 *in, int num_attribs) +{ + for (int i = 0; i < num_attribs; ++i) { + float *rd = dests[i]; + float4 ri = in[i]; + rd[0] = ri.x; + rd[1] = ri.y; + rd[2] = ri.z; + rd[3] = ri.w; + } +} + +extern void execute_shader(float4 *dests, float4 *inputs, + float4 *consts); + +void run_vertex_shader(float (*ainputs)[32][4], + float (*dests)[32][4], + float (*aconsts)[4], + int count, + int num_attribs) +{ + float4 inputs[16*32*4][32]; + float4 consts[32]; + float4 results[16*32*4][32]; + + printf("XXXXXXXXXXX run_vertex_shader\n"); + from_array(inputs, ainputs, count, num_attribs); + from_consts(consts, aconsts, 32); + for (int i = 0; i < count; ++i) { + float4 *in = inputs[i]; + float4 *res = results[i]; + to_array(dests[i], results[i], num_attribs); + execute_shader(res, in, consts); + } } diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 46b7561b5e..1abc148521 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -14,15 +14,88 @@ #include #include #include +#include +#include #include #include #include #include #include #include +#include +#include +#include +#include #include #include +using namespace llvm; +#include "llvm_base_shader.cpp" + + +static inline void addPass(PassManager &PM, Pass *P) { + // Add the pass to the pass manager... + PM.add(P); +} + +static inline void AddStandardCompilePasses(PassManager &PM) { + PM.add(createVerifierPass()); // Verify that input is correct + + addPass(PM, createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp + + // If the -strip-debug command line option was specified, do it. + //if (StripDebug) + // addPass(PM, createStripSymbolsPass(true)); + + addPass(PM, createRaiseAllocationsPass()); // call %malloc -> malloc inst + addPass(PM, createCFGSimplificationPass()); // Clean up disgusting code + addPass(PM, createPromoteMemoryToRegisterPass());// Kill useless allocas + addPass(PM, createGlobalOptimizerPass()); // Optimize out global vars + addPass(PM, createGlobalDCEPass()); // Remove unused fns and globs + addPass(PM, createIPConstantPropagationPass());// IP Constant Propagation + addPass(PM, createDeadArgEliminationPass()); // Dead argument elimination + addPass(PM, createInstructionCombiningPass()); // Clean up after IPCP & DAE + addPass(PM, createCFGSimplificationPass()); // Clean up after IPCP & DAE + + addPass(PM, createPruneEHPass()); // Remove dead EH info + + //if (!DisableInline) + addPass(PM, createFunctionInliningPass()); // Inline small functions + addPass(PM, createArgumentPromotionPass()); // Scalarize uninlined fn args + + addPass(PM, createTailDuplicationPass()); // Simplify cfg by copying code + addPass(PM, createInstructionCombiningPass()); // Cleanup for scalarrepl. + addPass(PM, createCFGSimplificationPass()); // Merge & remove BBs + addPass(PM, createScalarReplAggregatesPass()); // Break up aggregate allocas + addPass(PM, createInstructionCombiningPass()); // Combine silly seq's + addPass(PM, createCondPropagationPass()); // Propagate conditionals + + addPass(PM, createTailCallEliminationPass()); // Eliminate tail calls + addPass(PM, createCFGSimplificationPass()); // Merge & remove BBs + addPass(PM, createReassociatePass()); // Reassociate expressions + addPass(PM, createLoopRotatePass()); + addPass(PM, createLICMPass()); // Hoist loop invariants + addPass(PM, createLoopUnswitchPass()); // Unswitch loops. + addPass(PM, createLoopIndexSplitPass()); // Index split loops. + addPass(PM, createInstructionCombiningPass()); // Clean up after LICM/reassoc + addPass(PM, createIndVarSimplifyPass()); // Canonicalize indvars + addPass(PM, createLoopUnrollPass()); // Unroll small loops + addPass(PM, createInstructionCombiningPass()); // Clean up after the unroller + addPass(PM, createGVNPass()); // Remove redundancies + addPass(PM, createSCCPPass()); // Constant prop with SCCP + + // Run instcombine after redundancy elimination to exploit opportunities + // opened up by them. + addPass(PM, createInstructionCombiningPass()); + addPass(PM, createCondPropagationPass()); // Propagate conditionals + + addPass(PM, createDeadStoreEliminationPass()); // Delete dead stores + addPass(PM, createAggressiveDCEPass()); // SSA based 'Aggressive DCE' + addPass(PM, createCFGSimplificationPass()); // Merge & remove BBs + addPass(PM, createSimplifyLibCallsPass()); // Library Call Optimizations + addPass(PM, createDeadTypeEliminationPass()); // Eliminate dead types + addPass(PM, createConstantMergePass()); // Merge dup global constants +} static void translate_declaration(llvm::Module *module, @@ -341,7 +414,7 @@ translate_instruction(llvm::Module *module, static llvm::Module * tgsi_to_llvm(const struct tgsi_token *tokens) { - llvm::Module *mod = new llvm::Module("tgsi"); + llvm::Module *mod = createBaseShader(); struct tgsi_parse_context parse; struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; @@ -402,18 +475,33 @@ ga_llvm_from_tgsi(const struct tgsi_token *tokens) struct ga_llvm_prog *ga_llvm = (struct ga_llvm_prog *)malloc(sizeof(struct ga_llvm_prog)); llvm::Module *mod = tgsi_to_llvm(tokens); + + /* Run optimization passes over it */ + PassManager passes; + // Add an appropriate TargetData instance for this module... + passes.add(new TargetData(mod)); + AddStandardCompilePasses(passes); + std::cout<<"Running optimization passes..."<getFunction("run_vertex_shader"); + std::cout << "run_vertex_shader = "<function = ee->getPointerToFunctionOrStub(func); + std::cout << " -- FUNC is " <function; + return ga_llvm; } @@ -423,6 +511,7 @@ void ga_llvm_prog_delete(struct ga_llvm_prog *prog) delete mod; prog->module = 0; prog->engine = 0; + prog->function = 0; free(prog); } @@ -430,24 +519,12 @@ int ga_llvm_prog_exec(struct ga_llvm_prog *prog, float (*inputs)[32][4], void *dests[16*32*4], float (*consts)[4], - int count) + int count, + int num_attribs) { - //std::cout << "START "<(prog->module); - llvm::Function *func = mod->getFunction("main"); - llvm::ExecutionEngine *ee = static_cast(prog->engine); - - std::vector args(0); - //args[0] = GenericValue(&st); - //std::cout << "Mod is "<<*mod; - //std::cout << "\n\nRunning llvm: " << std::endl; - if (func) { - std::cout << "Func is "<runFunction(func, args); - } + std::cout << "---- START LLVM Execution "< Date: Mon, 15 Oct 2007 09:59:19 -0400 Subject: Execution engine is a singleton, for now keep it in the pipe. --- src/mesa/Makefile | 6 ++--- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 34 ++++++++++++++++++++++++++-- src/mesa/pipe/llvm/llvmtgsi.cpp | 30 ++++++++++++++++-------- src/mesa/pipe/llvm/llvmtgsi.h | 5 ++-- src/mesa/pipe/p_context.h | 4 +++- src/mesa/state_tracker/st_program.c | 4 ++-- 6 files changed, 64 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index cb7c466f47..5b56f76e0b 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -113,11 +113,11 @@ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) - @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ + $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ - $(MKLIB_OPTIONS) $(GL_LIB_DEPS) $(STAND_ALONE_OBJECTS) \ - $(SOFTPIPE_LIB) + $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ + $(SOFTPIPE_LIB) $(GL_LIB_DEPS) # Make the OSMesa library $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECTS) diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index fd49901051..37315f7566 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -98,6 +98,10 @@ void vertex_fetch(struct draw_context *draw, } } +void execute_shader(void *dests, void *inputs, void *consts) +{ + fprintf(stderr, "EXECUTING--\n"); +} /** * Called by the draw module when the vertx cache needs to be flushed. @@ -109,10 +113,33 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) struct vertex_header *dests[VS_QUEUE_LENGTH]; float inputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; + float outputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; float (*consts)[4] = (float (*)[4]) draw->mapped_constants; struct ga_llvm_prog *prog = draw->vertex_shader->state->llvm_prog; - fprintf(stderr, "--- XX q(%d) ", draw->vs.queue_nr); + fprintf(stderr, "--- XX q(%d) \n", draw->vs.queue_nr); + /* Consts does not require 16 byte alignment. */ + fprintf(stderr, "0AAAAA = %f %f %f %f\n", + consts[0][0], + consts[0][1], + consts[0][2], + consts[0][3]); + + fprintf(stderr, "1AAAAA = %f %f %f %f\n", + consts[1][0], + consts[1][1], + consts[1][2], + consts[1][3]); + fprintf(stderr, "2AAAAA = %f %f %f %f\n", + consts[2][0], + consts[2][1], + consts[2][2], + consts[2][3]); + fprintf(stderr, "3AAAAA = %f %f %f %f\n", + consts[3][0], + consts[3][1], + consts[3][2], + consts[3][3]); /* fetch the inputs */ for (i = 0; i < draw->vs.queue_nr; ++i) { @@ -122,9 +149,12 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) } /* batch execute the shaders on all the vertices */ - ga_llvm_prog_exec(prog, inputs, dests, consts, + ga_llvm_prog_exec(prog, inputs, outputs, consts, draw->vs.queue_nr, draw->vertex_info.num_attribs); + /* FIXME: finish conversion */ + /* dests = outputs */ + draw->vs.queue_nr = 0; } diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 1abc148521..a616355f8f 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -1,5 +1,6 @@ #include "llvmtgsi.h" +#include "pipe/p_context.h" #include "pipe/tgsi/exec/tgsi_exec.h" #include "pipe/tgsi/exec/tgsi_token.h" #include "pipe/tgsi/exec/tgsi_build.h" @@ -469,7 +470,7 @@ tgsi_to_llvm(const struct tgsi_token *tokens) } struct ga_llvm_prog * -ga_llvm_from_tgsi(const struct tgsi_token *tokens) +ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) { std::cout << "Creating llvm " <llvm_execution_engine) { + ee = llvm::ExecutionEngine::create(mp, false); + pipe->llvm_execution_engine = ee; + } else { + ee = (llvm::ExecutionEngine*)pipe->llvm_execution_engine; + ee->addModuleProvider(mp); + } ga_llvm->module = mod; - ga_llvm->engine = ee; fprintf(stderr, "DUMPX \n"); //tgsi_dump(tokens, TGSI_DUMP_VERBOSE); tgsi_dump(tokens, 0); fprintf(stderr, "DUMPEND \n"); Function *func = mod->getFunction("run_vertex_shader"); - std::cout << "run_vertex_shader = "<function; + std::cout << " -- FUNC is " <function<(prog->function); + runner(inputs, dests, consts, count, num_attribs); std::cout << "---- END LLVM Execution "<pipe, vs.tokens); cso = st_cached_vs_state(st, &vs); stvp->vs = cso; @@ -407,7 +407,7 @@ st_translate_fragment_program(struct st_context *st, tokensOut, maxTokens); fs.tokens = tokensOut; - fs.llvm_prog = ga_llvm_from_tgsi(fs.tokens); + fs.llvm_prog = ga_llvm_from_tgsi(st->pipe, fs.tokens); cso = st_cached_fs_state(st, &fs); stfp->fs = cso; -- cgit v1.2.3 From 5e0205023e8e6a08b0eb61286e15eb095f32ab3d Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 16 Oct 2007 05:41:06 -0400 Subject: Cleanup some of the testing code. Implement first pass at actually running shaders in llvm. --- src/mesa/pipe/draw/draw_vertex_shader.c | 1 + src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 29 +--- src/mesa/pipe/llvm/llvm_builtins.c | 19 +-- src/mesa/pipe/llvm/llvmtgsi.cpp | 195 ++++++++++++++++++++++++++- src/mesa/pipe/llvm/llvmtgsi.h | 6 +- 5 files changed, 208 insertions(+), 42 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 874259e727..e36ecdc849 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -132,6 +132,7 @@ run_vertex_program(struct draw_context *draw, y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + printf("output %d: %f %f %f %f\n", 0, x, y, z, w); vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); vOut[j]->edgeflag = 1; diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 37315f7566..97a0480b5c 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -98,11 +98,6 @@ void vertex_fetch(struct draw_context *draw, } } -void execute_shader(void *dests, void *inputs, void *consts) -{ - fprintf(stderr, "EXECUTING--\n"); -} - /** * Called by the draw module when the vertx cache needs to be flushed. * This involves running the vertex shader. @@ -118,28 +113,6 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) struct ga_llvm_prog *prog = draw->vertex_shader->state->llvm_prog; fprintf(stderr, "--- XX q(%d) \n", draw->vs.queue_nr); - /* Consts does not require 16 byte alignment. */ - fprintf(stderr, "0AAAAA = %f %f %f %f\n", - consts[0][0], - consts[0][1], - consts[0][2], - consts[0][3]); - - fprintf(stderr, "1AAAAA = %f %f %f %f\n", - consts[1][0], - consts[1][1], - consts[1][2], - consts[1][3]); - fprintf(stderr, "2AAAAA = %f %f %f %f\n", - consts[2][0], - consts[2][1], - consts[2][2], - consts[2][3]); - fprintf(stderr, "3AAAAA = %f %f %f %f\n", - consts[3][0], - consts[3][1], - consts[3][2], - consts[3][3]); /* fetch the inputs */ for (i = 0; i < draw->vs.queue_nr; ++i) { @@ -151,7 +124,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) /* batch execute the shaders on all the vertices */ ga_llvm_prog_exec(prog, inputs, outputs, consts, draw->vs.queue_nr, - draw->vertex_info.num_attribs); + draw->vertex_shader->state->num_inputs); /* FIXME: finish conversion */ /* dests = outputs */ diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index c7a9ea0d5a..e135b4781d 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -92,7 +92,7 @@ inline void collect_results(float4 *results, struct vertex_header *vOut, } #endif -void from_array(float4 (*res)[32], float (*ainputs)[32][4], +void from_array(float4 (*res)[16], float (*ainputs)[16][4], int count, int num_attribs) { for (int i = 0; i < count; ++i) { @@ -102,6 +102,7 @@ void from_array(float4 (*res)[32], float (*ainputs)[32][4], vec.y = ainputs[i][j][1]; vec.z = ainputs[i][j][2]; vec.w = ainputs[i][j][3]; + //printf("FAR(%d %d) %f %f %f %f\n", i, j, vec.x, vec.y, vec.z, vec.w); res[i][j] = vec; } } @@ -116,6 +117,7 @@ void from_consts(float4 *res, float (*ainputs)[4], vec.y = ainputs[i][1]; vec.z = ainputs[i][2]; vec.w = ainputs[i][3]; + //printf("VCONST %f %f %f %f\n", vec.x, vec.y, vec.z, vec.w); res[i] = vec; } } @@ -125,6 +127,7 @@ void to_array(float (*dests)[4], float4 *in, int num_attribs) for (int i = 0; i < num_attribs; ++i) { float *rd = dests[i]; float4 ri = in[i]; + printf("DEST = %f %f %f %f\n", ri.x, ri.y, ri.z, ri.w); rd[0] = ri.x; rd[1] = ri.y; rd[2] = ri.z; @@ -135,23 +138,23 @@ void to_array(float (*dests)[4], float4 *in, int num_attribs) extern void execute_shader(float4 *dests, float4 *inputs, float4 *consts); -void run_vertex_shader(float (*ainputs)[32][4], - float (*dests)[32][4], +void run_vertex_shader(float (*ainputs)[16][4], + float (*dests)[16][4], float (*aconsts)[4], int count, int num_attribs) { - float4 inputs[16*32*4][32]; - float4 consts[32]; - float4 results[16*32*4][32]; + float4 inputs[16*32*4][16]; + float4 consts[16]; + float4 results[16*32*4][16]; printf("XXXXXXXXXXX run_vertex_shader\n"); from_array(inputs, ainputs, count, num_attribs); - from_consts(consts, aconsts, 32); + from_consts(consts, aconsts, 5); for (int i = 0; i < count; ++i) { float4 *in = inputs[i]; float4 *res = results[i]; - to_array(dests[i], results[i], num_attribs); execute_shader(res, in, consts); + to_array(dests[i], res, num_attribs); } } diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index a616355f8f..4ff6281c74 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -420,6 +420,19 @@ tgsi_to_llvm(const struct tgsi_token *tokens) struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; + Function* shader = mod->getFunction("execute_shader"); + shader->setName("execute_shader_2"); + + Function::arg_iterator args = shader->arg_begin(); + Value *ptr_OUT = args++; + ptr_OUT->setName("OUT"); + Value *ptr_IN = args++; + ptr_IN->setName("IN"); + Value *ptr_CONST = args++; + ptr_CONST->setName("CONST"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + tgsi_parse_init(&parse, tokens); //parse.FullHeader.Processor.Processor @@ -462,10 +475,184 @@ tgsi_to_llvm(const struct tgsi_token *tokens) } } + + Function* func_printf = mod->getFunction("printf"); +#if 1 + + // Type Definitions + ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 19); + + PointerType* PointerTy_1 = PointerType::get(ArrayTy_0); + + VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); + + PointerType* PointerTy_3 = PointerType::get(VectorTy_4); + + + VectorType* VectorTy_5 = VectorType::get(IntegerType::get(32), 4); + + + PointerType* PointerTy_8 = PointerType::get(IntegerType::get(8)); + + + // Global Variable Declarations + + + GlobalVariable* gvar_array__str = new GlobalVariable( + /*Type=*/ArrayTy_0, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str", + mod); + + GlobalVariable* gvar_array__str1 = new GlobalVariable( + /*Type=*/ArrayTy_0, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str1", + mod); + + GlobalVariable* gvar_array__str2 = new GlobalVariable( + /*Type=*/ArrayTy_0, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str2", + mod); + + // Constant Definitions + Constant* const_array_9 = ConstantArray::get("const %f %f %f %f\x0A", true); + Constant* const_array_10 = ConstantArray::get("resul %f %f %f %f\x0A", true); + Constant* const_array_11 = ConstantArray::get("outpu %f %f %f %f\x0A", true); + UndefValue* const_packed_12 = UndefValue::get(VectorTy_4); + Constant* const_packed_13 = Constant::getNullValue(VectorTy_5); + std::vector const_packed_14_elems; + ConstantInt* const_int32_15 = ConstantInt::get(APInt(32, "1", 10)); + const_packed_14_elems.push_back(const_int32_15); + const_packed_14_elems.push_back(const_int32_15); + const_packed_14_elems.push_back(const_int32_15); + const_packed_14_elems.push_back(const_int32_15); + Constant* const_packed_14 = ConstantVector::get(VectorTy_5, const_packed_14_elems); + std::vector const_packed_16_elems; + ConstantInt* const_int32_17 = ConstantInt::get(APInt(32, "2", 10)); + const_packed_16_elems.push_back(const_int32_17); + const_packed_16_elems.push_back(const_int32_17); + const_packed_16_elems.push_back(const_int32_17); + const_packed_16_elems.push_back(const_int32_17); + Constant* const_packed_16 = ConstantVector::get(VectorTy_5, const_packed_16_elems); + std::vector const_packed_18_elems; + ConstantInt* const_int32_19 = ConstantInt::get(APInt(32, "3", 10)); + const_packed_18_elems.push_back(const_int32_19); + const_packed_18_elems.push_back(const_int32_19); + const_packed_18_elems.push_back(const_int32_19); + const_packed_18_elems.push_back(const_int32_19); + Constant* const_packed_18 = ConstantVector::get(VectorTy_5, const_packed_18_elems); + std::vector const_ptr_20_indices; + Constant* const_int32_21 = Constant::getNullValue(IntegerType::get(32)); + const_ptr_20_indices.push_back(const_int32_21); + const_ptr_20_indices.push_back(const_int32_21); + Constant* const_ptr_20 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_20_indices[0], const_ptr_20_indices.size() ); + UndefValue* const_double_22 = UndefValue::get(Type::DoubleTy); + std::vector const_ptr_23_indices; + const_ptr_23_indices.push_back(const_int32_21); + const_ptr_23_indices.push_back(const_int32_21); + Constant* const_ptr_23 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_23_indices[0], const_ptr_23_indices.size() ); + std::vector const_ptr_24_indices; + const_ptr_24_indices.push_back(const_int32_21); + const_ptr_24_indices.push_back(const_int32_21); + Constant* const_ptr_24 = ConstantExpr::getGetElementPtr(gvar_array__str2, &const_ptr_24_indices[0], const_ptr_24_indices.size() ); + + // Global Variable Definitions + gvar_array__str->setInitializer(const_array_9); + gvar_array__str1->setInitializer(const_array_10); + gvar_array__str2->setInitializer(const_array_11); + + // Function Definitions + + // Function: execute_shader (func_execute_shader) + { + // Block entry (label_entry) + LoadInst* packed_tmp1 = new LoadInst(ptr_IN, "tmp1", false, label_entry); + ShuffleVectorInst* packed_tmp3 = new ShuffleVectorInst(packed_tmp1, const_packed_12, const_packed_13, "tmp3", label_entry); + LoadInst* packed_tmp6 = new LoadInst(ptr_CONST, "tmp6", false, label_entry); + BinaryOperator* packed_mul = BinaryOperator::create(Instruction::Mul, packed_tmp3, packed_tmp6, "mul", label_entry); + ShuffleVectorInst* packed_tmp8 = new ShuffleVectorInst(packed_tmp1, const_packed_12, const_packed_14, "tmp8", label_entry); + GetElementPtrInst* ptr_arrayidx10 = new GetElementPtrInst(ptr_CONST, const_int32_15, "arrayidx10", label_entry); + LoadInst* packed_tmp11 = new LoadInst(ptr_arrayidx10, "tmp11", false, label_entry); + BinaryOperator* packed_mul12 = BinaryOperator::create(Instruction::Mul, packed_tmp8, packed_tmp11, "mul12", label_entry); + BinaryOperator* packed_add = BinaryOperator::create(Instruction::Add, packed_mul12, packed_mul, "add", label_entry); + ShuffleVectorInst* packed_tmp15 = new ShuffleVectorInst(packed_tmp1, const_packed_12, const_packed_16, "tmp15", label_entry); + GetElementPtrInst* ptr_arrayidx17 = new GetElementPtrInst(ptr_CONST, const_int32_17, "arrayidx17", label_entry); + LoadInst* packed_tmp18 = new LoadInst(ptr_arrayidx17, "tmp18", false, label_entry); + BinaryOperator* packed_mul19 = BinaryOperator::create(Instruction::Mul, packed_tmp15, packed_tmp18, "mul19", label_entry); + BinaryOperator* packed_add21 = BinaryOperator::create(Instruction::Add, packed_mul19, packed_add, "add21", label_entry); + ShuffleVectorInst* packed_tmp25 = new ShuffleVectorInst(packed_tmp1, const_packed_12, const_packed_18, "tmp25", label_entry); + GetElementPtrInst* ptr_arrayidx27 = new GetElementPtrInst(ptr_CONST, const_int32_19, "arrayidx27", label_entry); + LoadInst* packed_tmp28 = new LoadInst(ptr_arrayidx27, "tmp28", false, label_entry); + BinaryOperator* packed_mul29 = BinaryOperator::create(Instruction::Mul, packed_tmp25, packed_tmp28, "mul29", label_entry); + BinaryOperator* packed_add31 = BinaryOperator::create(Instruction::Add, packed_mul29, packed_add21, "add31", label_entry); + StoreInst* void_25 = new StoreInst(packed_add31, ptr_OUT, false, label_entry); + GetElementPtrInst* ptr_arrayidx33 = new GetElementPtrInst(ptr_OUT, const_int32_15, "arrayidx33", label_entry); + GetElementPtrInst* ptr_arrayidx35 = new GetElementPtrInst(ptr_IN, const_int32_15, "arrayidx35", label_entry); + LoadInst* packed_tmp36 = new LoadInst(ptr_arrayidx35, "tmp36", false, label_entry); + StoreInst* void_26 = new StoreInst(packed_tmp36, ptr_arrayidx33, false, label_entry); + std::vector int32_call_params; + int32_call_params.push_back(const_ptr_20); + int32_call_params.push_back(const_double_22); + int32_call_params.push_back(const_double_22); + int32_call_params.push_back(const_double_22); + int32_call_params.push_back(const_double_22); + //CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry); + //int32_call->setCallingConv(CallingConv::C); + //int32_call->setTailCall(true); + ExtractElementInst* float_tmp52 = new ExtractElementInst(packed_tmp1, const_int32_21, "tmp52", label_entry); + CastInst* double_conv53 = new FPExtInst(float_tmp52, Type::DoubleTy, "conv53", label_entry); + ExtractElementInst* float_tmp55 = new ExtractElementInst(packed_tmp1, const_int32_15, "tmp55", label_entry); + CastInst* double_conv56 = new FPExtInst(float_tmp55, Type::DoubleTy, "conv56", label_entry); + ExtractElementInst* float_tmp58 = new ExtractElementInst(packed_tmp1, const_int32_17, "tmp58", label_entry); + CastInst* double_conv59 = new FPExtInst(float_tmp58, Type::DoubleTy, "conv59", label_entry); + ExtractElementInst* float_tmp61 = new ExtractElementInst(packed_tmp1, const_int32_19, "tmp61", label_entry); + CastInst* double_conv62 = new FPExtInst(float_tmp61, Type::DoubleTy, "conv62", label_entry); + std::vector int32_call63_params; + int32_call63_params.push_back(const_ptr_23); + int32_call63_params.push_back(double_conv53); + int32_call63_params.push_back(double_conv56); + int32_call63_params.push_back(double_conv59); + int32_call63_params.push_back(double_conv62); + //CallInst* int32_call63 = new CallInst(func_printf, int32_call63_params.begin(), int32_call63_params.end(), "call63", label_entry); + //int32_call63->setCallingConv(CallingConv::C); + //int32_call63->setTailCall(true); + ExtractElementInst* float_tmp65 = new ExtractElementInst(packed_add31, const_int32_21, "tmp65", label_entry); + CastInst* double_conv66 = new FPExtInst(float_tmp65, Type::DoubleTy, "conv66", label_entry); + ExtractElementInst* float_tmp68 = new ExtractElementInst(packed_add31, const_int32_15, "tmp68", label_entry); + CastInst* double_conv69 = new FPExtInst(float_tmp68, Type::DoubleTy, "conv69", label_entry); + ExtractElementInst* float_tmp71 = new ExtractElementInst(packed_add31, const_int32_17, "tmp71", label_entry); + CastInst* double_conv72 = new FPExtInst(float_tmp71, Type::DoubleTy, "conv72", label_entry); + ExtractElementInst* float_tmp74 = new ExtractElementInst(packed_add31, const_int32_19, "tmp74", label_entry); + CastInst* double_conv75 = new FPExtInst(float_tmp74, Type::DoubleTy, "conv75", label_entry); + std::vector int32_call76_params; + int32_call76_params.push_back(const_ptr_24); + int32_call76_params.push_back(double_conv66); + int32_call76_params.push_back(double_conv69); + int32_call76_params.push_back(double_conv72); + int32_call76_params.push_back(double_conv75); + //CallInst* int32_call76 = new CallInst(func_printf, int32_call76_params.begin(), int32_call76_params.end(), "call76", label_entry); + //int32_call76->setCallingConv(CallingConv::C); + //int32_call76->setTailCall(true); + } +#endif + + new ReturnInst(label_entry); + //TXT("\ntgsi-dump end -------------------\n"); tgsi_parse_free(&parse); + std::cout<<"XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"< Date: Tue, 16 Oct 2007 13:23:43 -0400 Subject: Draw first triangle. Start on the llvm builder. --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 74 ++++++++++++++- src/mesa/pipe/llvm/llvmtgsi.cpp | 53 ++++++----- src/mesa/pipe/llvm/tgsillvmbuilder.cpp | 134 +++++++++++++++++++++++++++ 3 files changed, 237 insertions(+), 24 deletions(-) create mode 100644 src/mesa/pipe/llvm/tgsillvmbuilder.cpp (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 97a0480b5c..2ff35ace24 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -98,6 +98,21 @@ void vertex_fetch(struct draw_context *draw, } } +static INLINE unsigned +compute_clipmask(const float *clip, const float (*plane)[4], unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + for (i = 0; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<mapped_constants; - struct ga_llvm_prog *prog = draw->vertex_shader->state->llvm_prog; + float (*consts)[4] = (float (*)[4]) draw->mapped_constants; + struct ga_llvm_prog *prog = (struct ga_llvm_prog *)draw->vertex_shader->state->llvm_prog; + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; fprintf(stderr, "--- XX q(%d) \n", draw->vs.queue_nr); @@ -129,5 +146,58 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) /* FIXME: finish conversion */ /* dests = outputs */ + /* store machine results */ + for (int i = 0; i < draw->vs.queue_nr; ++i) { + unsigned slot; + float x, y, z, w; + struct vertex_header *vOut = draw->vs.queue[i].dest; + float (*dests)[4] = outputs[i]; + + /* Handle attr[0] (position) specially: + * + * XXX: Computing the clipmask should be done in the vertex + * program as a set of DP4 instructions appended to the + * user-provided code. + */ + x = vOut->clip[0] = dests[0][0]; + y = vOut->clip[1] = dests[0][1]; + z = vOut->clip[2] = dests[0][2]; + w = vOut->clip[3] = dests[0][3]; + printf("output %d: %f %f %f %f\n", 0, x, y, z, w); + + vOut->clipmask = compute_clipmask(vOut->clip, draw->plane, draw->nr_planes); + vOut->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut->data[0][0] = x * scale[0] + trans[0]; + vOut->data[0][1] = y * scale[1] + trans[1]; + vOut->data[0][2] = z * scale[2] + trans[2]; + vOut->data[0][3] = w; + + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + * Skip 0 since we just did it above. + * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. + */ + for (slot = 1; slot < draw->vertex_info.num_attribs - 2; slot++) { + vOut->data[slot][0] = dests[slot][0]; + vOut->data[slot][1] = dests[slot][1]; + vOut->data[slot][2] = dests[slot][2]; + vOut->data[slot][3] = dests[slot][3]; + + printf("output %d: %f %f %f %f\n", slot, + vOut->data[slot][0], + vOut->data[slot][1], + vOut->data[slot][2], + vOut->data[slot][3]); + } + } /* loop over vertices */ + draw->vs.queue_nr = 0; } diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 4ff6281c74..375e61c0db 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -32,6 +32,7 @@ using namespace llvm; #include "llvm_base_shader.cpp" +#include "tgsillvmbuilder.cpp" static inline void addPass(PassManager &PM, Pass *P) { @@ -116,6 +117,7 @@ translate_immediate(llvm::Module *module, static void translate_instruction(llvm::Module *module, + VertexShaderBuilder *builder, struct tgsi_full_instruction *inst, struct tgsi_full_instruction *fi) { @@ -135,6 +137,7 @@ translate_instruction(llvm::Module *module, case TGSI_OPCODE_LOG: break; case TGSI_OPCODE_MUL: + break; case TGSI_OPCODE_ADD: break; @@ -435,6 +438,7 @@ tgsi_to_llvm(const struct tgsi_token *tokens) tgsi_parse_init(&parse, tokens); + Function* func_printf = mod->getFunction("printf"); //parse.FullHeader.Processor.Processor //parse.FullVersion.Version.MajorVersion @@ -446,7 +450,7 @@ tgsi_to_llvm(const struct tgsi_token *tokens) fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); - + VertexShaderBuilder builder(label_entry, ptr_IN, ptr_CONST); while(!tgsi_parse_end_of_tokens(&parse)) { tgsi_parse_token(&parse); @@ -465,7 +469,7 @@ tgsi_to_llvm(const struct tgsi_token *tokens) break; case TGSI_TOKEN_TYPE_INSTRUCTION: - translate_instruction(mod, + translate_instruction(mod, &builder, &parse.FullToken.FullInstruction, &fi); break; @@ -475,29 +479,26 @@ tgsi_to_llvm(const struct tgsi_token *tokens) } } - - Function* func_printf = mod->getFunction("printf"); -#if 1 - +#if 0 // Type Definitions ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 19); - + PointerType* PointerTy_1 = PointerType::get(ArrayTy_0); - + VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); - + PointerType* PointerTy_3 = PointerType::get(VectorTy_4); - - + + VectorType* VectorTy_5 = VectorType::get(IntegerType::get(32), 4); - - + + PointerType* PointerTy_8 = PointerType::get(IntegerType::get(8)); - - + + // Global Variable Declarations - + GlobalVariable* gvar_array__str = new GlobalVariable( /*Type=*/ArrayTy_0, /*isConstant=*/true, @@ -505,7 +506,7 @@ tgsi_to_llvm(const struct tgsi_token *tokens) /*Initializer=*/0, // has initializer, specified below /*Name=*/".str", mod); - + GlobalVariable* gvar_array__str1 = new GlobalVariable( /*Type=*/ArrayTy_0, /*isConstant=*/true, @@ -513,7 +514,7 @@ tgsi_to_llvm(const struct tgsi_token *tokens) /*Initializer=*/0, // has initializer, specified below /*Name=*/".str1", mod); - + GlobalVariable* gvar_array__str2 = new GlobalVariable( /*Type=*/ArrayTy_0, /*isConstant=*/true, @@ -521,7 +522,7 @@ tgsi_to_llvm(const struct tgsi_token *tokens) /*Initializer=*/0, // has initializer, specified below /*Name=*/".str2", mod); - + // Constant Definitions Constant* const_array_9 = ConstantArray::get("const %f %f %f %f\x0A", true); Constant* const_array_10 = ConstantArray::get("resul %f %f %f %f\x0A", true); @@ -563,14 +564,14 @@ tgsi_to_llvm(const struct tgsi_token *tokens) const_ptr_24_indices.push_back(const_int32_21); const_ptr_24_indices.push_back(const_int32_21); Constant* const_ptr_24 = ConstantExpr::getGetElementPtr(gvar_array__str2, &const_ptr_24_indices[0], const_ptr_24_indices.size() ); - + // Global Variable Definitions gvar_array__str->setInitializer(const_array_9); gvar_array__str1->setInitializer(const_array_10); gvar_array__str2->setInitializer(const_array_11); - + // Function Definitions - + // Function: execute_shader (func_execute_shader) { // Block entry (label_entry) @@ -725,5 +726,13 @@ int ga_llvm_prog_exec(struct ga_llvm_prog *prog, runner(inputs, dests, consts, count, num_attribs); std::cout << "---- END LLVM Execution "< + +class VertexShaderBuilder +{ + typedef std::map LoadMap; +public: + VertexShaderBuilder(llvm::BasicBlock *block, llvm::Value *in, llvm::Value *consts); + + llvm::ConstantInt *constantInt(int); + llvm::Constant *shuffleMask(int vec); + llvm::Value *inputElement(int idx); + llvm::Value *constElement(int idx); + + llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); + + +private: + llvm::BasicBlock *m_block; + llvm::Value *m_IN; + llvm::Value *m_CONST; + + std::map m_constInts; + std::map m_intVecs; + LoadMap m_inputs; + LoadMap m_consts; + + VectorType *m_floatVecType; + VectorType *m_intVecType; + + Value *m_undefFloatVec; + Value *m_undefIntVec; + + int m_shuffleId; +}; + +VertexShaderBuilder::VertexShaderBuilder(llvm::BasicBlock *block, llvm::Value *in, llvm::Value *consts) + : m_block(block), m_IN(in), m_CONST(consts) +{ + m_floatVecType = VectorType::get(Type::FloatTy, 4); + m_intVecType = VectorType::get(IntegerType::get(32), 4); + + m_undefFloatVec = UndefValue::get(m_floatVecType); + m_undefIntVec = UndefValue::get(m_intVecType); + + m_shuffleId = 0; +} + +//can only build vectors with all members in the [0, 9] range +llvm::Constant *VertexShaderBuilder::shuffleMask(int vec) +{ + if (m_intVecs.find(vec) != m_intVecs.end()) { + return m_intVecs[vec]; + } + int origVec = vec; + Constant* const_vec = 0; + if (origVec == 0) { + const_vec = Constant::getNullValue(m_intVecType); + } else { + int x = vec / 1000; vec -= x * 1000; + int y = vec / 100; vec -= y * 100; + int z = vec / 10; vec -= z * 10; + int w = vec; + std::vector elems; + elems.push_back(constantInt(x)); + elems.push_back(constantInt(y)); + elems.push_back(constantInt(z)); + elems.push_back(constantInt(w)); + const_vec = ConstantVector::get(m_intVecType, elems); + } + + m_intVecs[origVec] = const_vec; + return const_vec; +} + +llvm::ConstantInt *VertexShaderBuilder::constantInt(int idx) +{ + if (m_constInts.find(idx) != m_constInts.end()) { + return m_constInts[idx]; + } + ConstantInt *const_int = ConstantInt::get(APInt(32, idx)); + m_constInts[idx] = const_int; + return const_int; +} + +llvm::Value *VertexShaderBuilder::inputElement(int idx) +{ + if (m_inputs.find(idx) != m_inputs.end()) { + return m_inputs[idx]; + } + char ptrName[13]; + char name[9]; + snprintf(ptrName, 13, "input_ptr%d", idx); + snprintf(name, 9, "input%d", idx); + GetElementPtrInst *getElem = new GetElementPtrInst(m_IN, + constantInt(idx), + ptrName, + m_block); + LoadInst *load = new LoadInst(getElem, name, + false, m_block); + m_inputs[idx] = load; + return load; +} + +llvm::Value *VertexShaderBuilder::constElement(int idx) +{ + if (m_consts.find(idx) != m_consts.end()) { + return m_consts[idx]; + } + char ptrName[13]; + char name[9]; + snprintf(ptrName, 13, "const_ptr%d", idx); + snprintf(name, 9, "const%d", idx); + GetElementPtrInst *getElem = new GetElementPtrInst(m_CONST, + constantInt(idx), + ptrName, + m_block); + LoadInst *load = new LoadInst(getElem, name, + false, m_block); + m_consts[idx] = load; + return load; +} + +llvm::Value *VertexShaderBuilder::shuffleVector(llvm::Value *vec, int shuffle) +{ + Constant *mask = shuffleMask(shuffle); + ++m_shuffleId; + char name[11]; + snprintf(name, 11, "shuffle%d", m_shuffleId); + ShuffleVectorInst *res = + new ShuffleVectorInst(vec, m_undefFloatVec, mask, + name, m_block); + return res; +} -- cgit v1.2.3 From 3975f34fd36f8b04d499bb6b3d48eaeef5cab24e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 17 Oct 2007 11:27:46 -0400 Subject: Implement basic opcode translation and storage translation. --- src/mesa/pipe/llvm/instructions.cpp | 162 +++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 43 +++++++++ src/mesa/pipe/llvm/llvmtgsi.cpp | 101 ++++++++++++++++++-- src/mesa/pipe/llvm/tgsillvmbuilder.cpp | 68 ++++++++++---- src/mesa/sources | 3 +- 5 files changed, 350 insertions(+), 27 deletions(-) create mode 100644 src/mesa/pipe/llvm/instructions.cpp create mode 100644 src/mesa/pipe/llvm/instructions.h (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp new file mode 100644 index 0000000000..147a1b64f2 --- /dev/null +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -0,0 +1,162 @@ +#include "instructions.h" + +#include +#include +#include +#include +#include + +using namespace llvm; + +Instructions::Instructions(llvm::Module *mod, llvm::BasicBlock *block) + : m_mod(mod), m_block(block), m_idx(0) +{ + m_floatVecType = VectorType::get(Type::FloatTy, 4); + m_llvmFSqrt = 0; + m_llvmFAbs = 0; +} + +llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) +{ + BinaryOperator *res = BinaryOperator::create(Instruction::Add, in1, in2, + name("add"), + m_block); + return res; +} + +llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3) +{ + Value *mulRes = mul(in1, in2); + return add(mulRes, in3); +} + +llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2) +{ + BinaryOperator *res = BinaryOperator::create(Instruction::Mul, in1, in2, + name("mul"), + m_block); + return res; +} + +const char * Instructions::name(const char *prefix) +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + ExtractElementInst *x = new ExtractElementInst(mulRes, unsigned(0), + name("extractx"), + m_block); + ExtractElementInst *y = new ExtractElementInst(mulRes, unsigned(1), + name("extracty"), + m_block); + ExtractElementInst *z = new ExtractElementInst(mulRes, unsigned(2), + name("extractz"), + m_block); + BinaryOperator *xy = BinaryOperator::create(Instruction::Add, x, y, + name("xy"), + m_block); + BinaryOperator *dot3 = BinaryOperator::create(Instruction::Add, xy, z, + name("dot3"), + m_block); + return vectorFromVals(dot3, dot3, dot3, dot3); +} + +llvm::Value *Instructions::callFSqrt(llvm::Value *val) +{ + if (!m_llvmFSqrt) { + // predeclare the intrinsic + std::vector fsqrtArgs; + fsqrtArgs.push_back(Type::FloatTy); + ParamAttrsList *fsqrtPal = 0; + FunctionType* fsqrtType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/fsqrtArgs, + /*isVarArg=*/false, + /*ParamAttrs=*/fsqrtPal); + m_llvmFSqrt = new Function( + /*Type=*/fsqrtType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"llvm.sqrt.f32", m_mod); + m_llvmFSqrt->setCallingConv(CallingConv::C); + } + CallInst *call = new CallInst(m_llvmFSqrt, val, + name("sqrt"), + m_block); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::rsq(llvm::Value *in1) +{ + ExtractElementInst *x = new ExtractElementInst(in1, unsigned(0), + name("extractx"), + m_block); + Value *abs = callFAbs(x); + Value *sqrt = callFSqrt(abs); + + BinaryOperator *rsqrt = BinaryOperator::create(Instruction::FDiv, + ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + sqrt, + name("rsqrt"), + m_block); + return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt); +} + +llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w) +{ + Constant *const_vec = Constant::getNullValue(m_floatVecType); + InsertElementInst *res = new InsertElementInst(const_vec, x, unsigned(0), + name("vecx"), m_block); + res = new InsertElementInst(res, y, unsigned(1), + name("vecxy"), + m_block); + res = new InsertElementInst(res, z, unsigned(2), + name("vecxyz"), + m_block); + if (w) + res = new InsertElementInst(res, w, unsigned(3), + name("vecxyw"), + m_block); + return res; +} + +llvm::Value *Instructions::callFAbs(llvm::Value *val) +{ + if (!m_llvmFAbs) { + // predeclare the intrinsic + std::vector fabsArgs; + fabsArgs.push_back(Type::FloatTy); + ParamAttrsList *fabsPal = 0; + FunctionType* fabsType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/fabsArgs, + /*isVarArg=*/false, + /*ParamAttrs=*/fabsPal); + m_llvmFAbs = new Function( + /*Type=*/fabsType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"fabs", m_mod); + m_llvmFAbs->setCallingConv(CallingConv::C); + } + CallInst *call = new CallInst(m_llvmFAbs, val, + name("fabs"), + m_block); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::lit(llvm::Value *in1) +{ + return in1; +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h new file mode 100644 index 0000000000..c6e77710ea --- /dev/null +++ b/src/mesa/pipe/llvm/instructions.h @@ -0,0 +1,43 @@ +#ifndef INSTRUCTIONS_H +#define INSTRUCTIONS_H + +#include +#include +#include + +namespace llvm { + class VectorType; +} + +class Instructions +{ +public: + Instructions(llvm::Module *mod, llvm::BasicBlock *block); + + llvm::Value *add(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); + llvm::Value *lit(llvm::Value *in1); + llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in2); + llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); + llvm::Value *rsq(llvm::Value *in1); +private: + const char *name(const char *prefix); + + llvm::Value *callFSqrt(llvm::Value *val); + llvm::Value *callFAbs(llvm::Value *val); + + llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w=0); +private: + llvm::Module *m_mod; + char m_name[32]; + llvm::BasicBlock *m_block; + int m_idx; + llvm::Function *m_llvmFSqrt; + llvm::Function *m_llvmFAbs; + + llvm::VectorType *m_floatVecType; +}; + +#endif diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 375e61c0db..5cbd0ae89f 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -30,6 +30,7 @@ #include #include +#include "instructions.h" using namespace llvm; #include "llvm_base_shader.cpp" #include "tgsillvmbuilder.cpp" @@ -117,31 +118,82 @@ translate_immediate(llvm::Module *module, static void translate_instruction(llvm::Module *module, - VertexShaderBuilder *builder, + Storage *storage, + Instructions *instr, struct tgsi_full_instruction *inst, struct tgsi_full_instruction *fi) { + llvm::Value *inputs[4]; + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + llvm::Value *val = 0; + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->constElement(src->SrcRegister.Index); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->inputElement(src->SrcRegister.Index); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->tempElement(src->SrcRegister.Index); + } else { + fprintf(stderr, "ERROR: not support llvm source\n"); + return; + } + + if (src->SrcRegister.Extended) { + if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { + int swizzle = src->SrcRegisterExtSwz.ExtSwizzleX * 1000; + swizzle += src->SrcRegisterExtSwz.ExtSwizzleY * 100; + swizzle += src->SrcRegisterExtSwz.ExtSwizzleZ * 10; + swizzle += src->SrcRegisterExtSwz.ExtSwizzleW * 1; + val = storage->shuffleVector(val, swizzle); + } + } else if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) { + int swizzle = src->SrcRegister.SwizzleX * 1000; + swizzle += src->SrcRegister.SwizzleY * 100; + swizzle += src->SrcRegister.SwizzleZ * 10; + swizzle += src->SrcRegister.SwizzleW * 1; + val = storage->shuffleVector(val, swizzle); + } + inputs[i] = val; + } + + llvm::Value *out = 0; switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: break; - case TGSI_OPCODE_MOV: + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } break; - case TGSI_OPCODE_LIT: + case TGSI_OPCODE_LIT: { + //out = instr->lit(inputs[0]); + return; + } break; case TGSI_OPCODE_RCP: break; - case TGSI_OPCODE_RSQ: + case TGSI_OPCODE_RSQ: { + out = instr->rsq(inputs[0]); + } break; case TGSI_OPCODE_EXP: break; case TGSI_OPCODE_LOG: break; - case TGSI_OPCODE_MUL: - + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_ADD: break; - case TGSI_OPCODE_DP3: + case TGSI_OPCODE_DP3: { + out = instr->dp3(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_DP4: break; @@ -155,7 +207,9 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_SGE: break; - case TGSI_OPCODE_MAD: + case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); + } break; case TGSI_OPCODE_SUB: break; @@ -400,6 +454,7 @@ translate_instruction(llvm::Module *module, break; } + switch( inst->Instruction.Saturate ) { case TGSI_SAT_NONE: break; @@ -412,6 +467,31 @@ translate_instruction(llvm::Module *module, default: assert( 0 ); } + + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->store(dst->DstRegister.Index, out); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->setTempElement(dst->DstRegister.Index, out); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + } + +#if 0 + if (dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW) { + if (dst->DstRegister.WriteMask & TGSI_WRITEMASK_X) { + } + if (dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y) { + } + if (dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z) { + } + if (dst->DstRegister.WriteMask & TGSI_WRITEMASK_W) { + } + } +#endif + } } @@ -450,7 +530,8 @@ tgsi_to_llvm(const struct tgsi_token *tokens) fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); - VertexShaderBuilder builder(label_entry, ptr_IN, ptr_CONST); + Storage storage(label_entry, ptr_OUT, ptr_IN, ptr_CONST); + Instructions instr(mod, label_entry); while(!tgsi_parse_end_of_tokens(&parse)) { tgsi_parse_token(&parse); @@ -469,7 +550,7 @@ tgsi_to_llvm(const struct tgsi_token *tokens) break; case TGSI_TOKEN_TYPE_INSTRUCTION: - translate_instruction(mod, &builder, + translate_instruction(mod, &storage, &instr, &parse.FullToken.FullInstruction, &fi); break; diff --git a/src/mesa/pipe/llvm/tgsillvmbuilder.cpp b/src/mesa/pipe/llvm/tgsillvmbuilder.cpp index 4088da11f8..ca70a46648 100644 --- a/src/mesa/pipe/llvm/tgsillvmbuilder.cpp +++ b/src/mesa/pipe/llvm/tgsillvmbuilder.cpp @@ -1,41 +1,52 @@ #include -class VertexShaderBuilder +class Storage { typedef std::map LoadMap; public: - VertexShaderBuilder(llvm::BasicBlock *block, llvm::Value *in, llvm::Value *consts); + Storage(llvm::BasicBlock *block, + llvm::Value *out, + llvm::Value *in, llvm::Value *consts); llvm::ConstantInt *constantInt(int); llvm::Constant *shuffleMask(int vec); llvm::Value *inputElement(int idx); llvm::Value *constElement(int idx); + llvm::Value *tempElement(int idx) const; + void setTempElement(int idx, llvm::Value *val); + llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); + void store(int dstIdx, llvm::Value *val); private: llvm::BasicBlock *m_block; + llvm::Value *m_OUT; llvm::Value *m_IN; llvm::Value *m_CONST; std::map m_constInts; - std::map m_intVecs; - LoadMap m_inputs; - LoadMap m_consts; + std::map m_intVecs; + std::vector m_temps; + LoadMap m_inputs; + LoadMap m_consts; - VectorType *m_floatVecType; - VectorType *m_intVecType; + llvm::VectorType *m_floatVecType; + llvm::VectorType *m_intVecType; - Value *m_undefFloatVec; - Value *m_undefIntVec; + llvm::Value *m_undefFloatVec; + llvm::Value *m_undefIntVec; int m_shuffleId; }; -VertexShaderBuilder::VertexShaderBuilder(llvm::BasicBlock *block, llvm::Value *in, llvm::Value *consts) - : m_block(block), m_IN(in), m_CONST(consts) +Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, + llvm::Value *in, llvm::Value *consts) + : m_block(block), m_OUT(out), + m_IN(in), m_CONST(consts), + m_temps(32) { m_floatVecType = VectorType::get(Type::FloatTy, 4); m_intVecType = VectorType::get(IntegerType::get(32), 4); @@ -47,7 +58,7 @@ VertexShaderBuilder::VertexShaderBuilder(llvm::BasicBlock *block, llvm::Value *i } //can only build vectors with all members in the [0, 9] range -llvm::Constant *VertexShaderBuilder::shuffleMask(int vec) +llvm::Constant *Storage::shuffleMask(int vec) { if (m_intVecs.find(vec) != m_intVecs.end()) { return m_intVecs[vec]; @@ -73,7 +84,7 @@ llvm::Constant *VertexShaderBuilder::shuffleMask(int vec) return const_vec; } -llvm::ConstantInt *VertexShaderBuilder::constantInt(int idx) +llvm::ConstantInt *Storage::constantInt(int idx) { if (m_constInts.find(idx) != m_constInts.end()) { return m_constInts[idx]; @@ -83,7 +94,7 @@ llvm::ConstantInt *VertexShaderBuilder::constantInt(int idx) return const_int; } -llvm::Value *VertexShaderBuilder::inputElement(int idx) +llvm::Value *Storage::inputElement(int idx) { if (m_inputs.find(idx) != m_inputs.end()) { return m_inputs[idx]; @@ -102,7 +113,7 @@ llvm::Value *VertexShaderBuilder::inputElement(int idx) return load; } -llvm::Value *VertexShaderBuilder::constElement(int idx) +llvm::Value *Storage::constElement(int idx) { if (m_consts.find(idx) != m_consts.end()) { return m_consts[idx]; @@ -121,7 +132,7 @@ llvm::Value *VertexShaderBuilder::constElement(int idx) return load; } -llvm::Value *VertexShaderBuilder::shuffleVector(llvm::Value *vec, int shuffle) +llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) { Constant *mask = shuffleMask(shuffle); ++m_shuffleId; @@ -132,3 +143,28 @@ llvm::Value *VertexShaderBuilder::shuffleVector(llvm::Value *vec, int shuffle) name, m_block); return res; } + + +llvm::Value *Storage::tempElement(int idx) const +{ + Value *ret = m_temps[idx]; + if (!ret) + return m_undefFloatVec; + return ret; +} + +void Storage::setTempElement(int idx, llvm::Value *val) +{ + m_temps[idx] = val; +} + +void Storage::store(int dstIdx, llvm::Value *val) +{ + char ptrName[13]; + snprintf(ptrName, 13, "out_ptr%d", dstIdx); + GetElementPtrInst *getElem = new GetElementPtrInst(m_OUT, + constantInt(dstIdx), + ptrName, + m_block); + new StoreInst(val, getElem, false, m_block); +} diff --git a/src/mesa/sources b/src/mesa/sources index c2544ed322..7e1d909e28 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -191,7 +191,8 @@ TGSIMESA_SOURCES = \ ifeq ($(MESA_NO_LLVM),0) LLVMTGSI_SOURCES = \ - pipe/llvm/llvmtgsi.cpp + pipe/llvm/llvmtgsi.cpp \ + pipe/llvm/instructions.cpp endif STATECACHE_SOURCES = \ -- cgit v1.2.3 From b0f80693434cb203f63d8fbab56c1522000ed88f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 17 Oct 2007 11:28:26 -0400 Subject: Refactor the tgsi->llvm storage translator --- src/mesa/pipe/llvm/llvmtgsi.cpp | 5 +- src/mesa/pipe/llvm/storage.cpp | 140 +++++++++++++++++++++++++++ src/mesa/pipe/llvm/storage.h | 57 +++++++++++ src/mesa/pipe/llvm/tgsillvmbuilder.cpp | 170 --------------------------------- src/mesa/sources | 1 + 5 files changed, 201 insertions(+), 172 deletions(-) create mode 100644 src/mesa/pipe/llvm/storage.cpp create mode 100644 src/mesa/pipe/llvm/storage.h delete mode 100644 src/mesa/pipe/llvm/tgsillvmbuilder.cpp (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 5cbd0ae89f..c6cf2de4fc 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -1,5 +1,8 @@ #include "llvmtgsi.h" +#include "instructions.h" +#include "storage.h" + #include "pipe/p_context.h" #include "pipe/tgsi/exec/tgsi_exec.h" #include "pipe/tgsi/exec/tgsi_token.h" @@ -30,10 +33,8 @@ #include #include -#include "instructions.h" using namespace llvm; #include "llvm_base_shader.cpp" -#include "tgsillvmbuilder.cpp" static inline void addPass(PassManager &PM, Pass *P) { diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp new file mode 100644 index 0000000000..c6e86ea4b4 --- /dev/null +++ b/src/mesa/pipe/llvm/storage.cpp @@ -0,0 +1,140 @@ +#include "storage.h" + +#include +#include +#include + +#include +#include +#include +#include +#include + +using namespace llvm; + +Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, + llvm::Value *in, llvm::Value *consts) + : m_block(block), m_OUT(out), + m_IN(in), m_CONST(consts), + m_temps(32) +{ + m_floatVecType = VectorType::get(Type::FloatTy, 4); + m_intVecType = VectorType::get(IntegerType::get(32), 4); + + m_undefFloatVec = UndefValue::get(m_floatVecType); + m_undefIntVec = UndefValue::get(m_intVecType); + + m_shuffleId = 0; +} + +//can only build vectors with all members in the [0, 9] range +llvm::Constant *Storage::shuffleMask(int vec) +{ + if (m_intVecs.find(vec) != m_intVecs.end()) { + return m_intVecs[vec]; + } + int origVec = vec; + Constant* const_vec = 0; + if (origVec == 0) { + const_vec = Constant::getNullValue(m_intVecType); + } else { + int x = vec / 1000; vec -= x * 1000; + int y = vec / 100; vec -= y * 100; + int z = vec / 10; vec -= z * 10; + int w = vec; + std::vector elems; + elems.push_back(constantInt(x)); + elems.push_back(constantInt(y)); + elems.push_back(constantInt(z)); + elems.push_back(constantInt(w)); + const_vec = ConstantVector::get(m_intVecType, elems); + } + + m_intVecs[origVec] = const_vec; + return const_vec; +} + +llvm::ConstantInt *Storage::constantInt(int idx) +{ + if (m_constInts.find(idx) != m_constInts.end()) { + return m_constInts[idx]; + } + ConstantInt *const_int = ConstantInt::get(APInt(32, idx)); + m_constInts[idx] = const_int; + return const_int; +} + +llvm::Value *Storage::inputElement(int idx) +{ + if (m_inputs.find(idx) != m_inputs.end()) { + return m_inputs[idx]; + } + char ptrName[13]; + char name[9]; + snprintf(ptrName, 13, "input_ptr%d", idx); + snprintf(name, 9, "input%d", idx); + GetElementPtrInst *getElem = new GetElementPtrInst(m_IN, + constantInt(idx), + ptrName, + m_block); + LoadInst *load = new LoadInst(getElem, name, + false, m_block); + m_inputs[idx] = load; + return load; +} + +llvm::Value *Storage::constElement(int idx) +{ + if (m_consts.find(idx) != m_consts.end()) { + return m_consts[idx]; + } + char ptrName[13]; + char name[9]; + snprintf(ptrName, 13, "const_ptr%d", idx); + snprintf(name, 9, "const%d", idx); + GetElementPtrInst *getElem = new GetElementPtrInst(m_CONST, + constantInt(idx), + ptrName, + m_block); + LoadInst *load = new LoadInst(getElem, name, + false, m_block); + m_consts[idx] = load; + return load; +} + +llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) +{ + Constant *mask = shuffleMask(shuffle); + ++m_shuffleId; + char name[11]; + snprintf(name, 11, "shuffle%d", m_shuffleId); + ShuffleVectorInst *res = + new ShuffleVectorInst(vec, m_undefFloatVec, mask, + name, m_block); + return res; +} + + +llvm::Value *Storage::tempElement(int idx) const +{ + Value *ret = m_temps[idx]; + if (!ret) + return m_undefFloatVec; + return ret; +} + +void Storage::setTempElement(int idx, llvm::Value *val) +{ + m_temps[idx] = val; +} + +void Storage::store(int dstIdx, llvm::Value *val) +{ + char ptrName[13]; + snprintf(ptrName, 13, "out_ptr%d", dstIdx); + GetElementPtrInst *getElem = new GetElementPtrInst(m_OUT, + constantInt(dstIdx), + ptrName, + m_block); + new StoreInst(val, getElem, false, m_block); +} diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h new file mode 100644 index 0000000000..b69c8d614e --- /dev/null +++ b/src/mesa/pipe/llvm/storage.h @@ -0,0 +1,57 @@ +#ifndef STORAGE_H +#define STORAGE_H + +#include +#include + +namespace llvm { + class BasicBlock; + class Constant; + class ConstantInt; + class LoadInst; + class Value; + class VectorType; +} + +class Storage +{ + typedef std::map LoadMap; +public: + Storage(llvm::BasicBlock *block, + llvm::Value *out, + llvm::Value *in, llvm::Value *consts); + + llvm::ConstantInt *constantInt(int); + llvm::Constant *shuffleMask(int vec); + llvm::Value *inputElement(int idx); + llvm::Value *constElement(int idx); + + llvm::Value *tempElement(int idx) const; + void setTempElement(int idx, llvm::Value *val); + + llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); + + + void store(int dstIdx, llvm::Value *val); +private: + llvm::BasicBlock *m_block; + llvm::Value *m_OUT; + llvm::Value *m_IN; + llvm::Value *m_CONST; + + std::map m_constInts; + std::map m_intVecs; + std::vector m_temps; + LoadMap m_inputs; + LoadMap m_consts; + + llvm::VectorType *m_floatVecType; + llvm::VectorType *m_intVecType; + + llvm::Value *m_undefFloatVec; + llvm::Value *m_undefIntVec; + + int m_shuffleId; +}; + +#endif diff --git a/src/mesa/pipe/llvm/tgsillvmbuilder.cpp b/src/mesa/pipe/llvm/tgsillvmbuilder.cpp deleted file mode 100644 index ca70a46648..0000000000 --- a/src/mesa/pipe/llvm/tgsillvmbuilder.cpp +++ /dev/null @@ -1,170 +0,0 @@ - -#include - -class Storage -{ - typedef std::map LoadMap; -public: - Storage(llvm::BasicBlock *block, - llvm::Value *out, - llvm::Value *in, llvm::Value *consts); - - llvm::ConstantInt *constantInt(int); - llvm::Constant *shuffleMask(int vec); - llvm::Value *inputElement(int idx); - llvm::Value *constElement(int idx); - - llvm::Value *tempElement(int idx) const; - void setTempElement(int idx, llvm::Value *val); - - llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); - - - void store(int dstIdx, llvm::Value *val); -private: - llvm::BasicBlock *m_block; - llvm::Value *m_OUT; - llvm::Value *m_IN; - llvm::Value *m_CONST; - - std::map m_constInts; - std::map m_intVecs; - std::vector m_temps; - LoadMap m_inputs; - LoadMap m_consts; - - llvm::VectorType *m_floatVecType; - llvm::VectorType *m_intVecType; - - llvm::Value *m_undefFloatVec; - llvm::Value *m_undefIntVec; - - int m_shuffleId; -}; - -Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, - llvm::Value *in, llvm::Value *consts) - : m_block(block), m_OUT(out), - m_IN(in), m_CONST(consts), - m_temps(32) -{ - m_floatVecType = VectorType::get(Type::FloatTy, 4); - m_intVecType = VectorType::get(IntegerType::get(32), 4); - - m_undefFloatVec = UndefValue::get(m_floatVecType); - m_undefIntVec = UndefValue::get(m_intVecType); - - m_shuffleId = 0; -} - -//can only build vectors with all members in the [0, 9] range -llvm::Constant *Storage::shuffleMask(int vec) -{ - if (m_intVecs.find(vec) != m_intVecs.end()) { - return m_intVecs[vec]; - } - int origVec = vec; - Constant* const_vec = 0; - if (origVec == 0) { - const_vec = Constant::getNullValue(m_intVecType); - } else { - int x = vec / 1000; vec -= x * 1000; - int y = vec / 100; vec -= y * 100; - int z = vec / 10; vec -= z * 10; - int w = vec; - std::vector elems; - elems.push_back(constantInt(x)); - elems.push_back(constantInt(y)); - elems.push_back(constantInt(z)); - elems.push_back(constantInt(w)); - const_vec = ConstantVector::get(m_intVecType, elems); - } - - m_intVecs[origVec] = const_vec; - return const_vec; -} - -llvm::ConstantInt *Storage::constantInt(int idx) -{ - if (m_constInts.find(idx) != m_constInts.end()) { - return m_constInts[idx]; - } - ConstantInt *const_int = ConstantInt::get(APInt(32, idx)); - m_constInts[idx] = const_int; - return const_int; -} - -llvm::Value *Storage::inputElement(int idx) -{ - if (m_inputs.find(idx) != m_inputs.end()) { - return m_inputs[idx]; - } - char ptrName[13]; - char name[9]; - snprintf(ptrName, 13, "input_ptr%d", idx); - snprintf(name, 9, "input%d", idx); - GetElementPtrInst *getElem = new GetElementPtrInst(m_IN, - constantInt(idx), - ptrName, - m_block); - LoadInst *load = new LoadInst(getElem, name, - false, m_block); - m_inputs[idx] = load; - return load; -} - -llvm::Value *Storage::constElement(int idx) -{ - if (m_consts.find(idx) != m_consts.end()) { - return m_consts[idx]; - } - char ptrName[13]; - char name[9]; - snprintf(ptrName, 13, "const_ptr%d", idx); - snprintf(name, 9, "const%d", idx); - GetElementPtrInst *getElem = new GetElementPtrInst(m_CONST, - constantInt(idx), - ptrName, - m_block); - LoadInst *load = new LoadInst(getElem, name, - false, m_block); - m_consts[idx] = load; - return load; -} - -llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) -{ - Constant *mask = shuffleMask(shuffle); - ++m_shuffleId; - char name[11]; - snprintf(name, 11, "shuffle%d", m_shuffleId); - ShuffleVectorInst *res = - new ShuffleVectorInst(vec, m_undefFloatVec, mask, - name, m_block); - return res; -} - - -llvm::Value *Storage::tempElement(int idx) const -{ - Value *ret = m_temps[idx]; - if (!ret) - return m_undefFloatVec; - return ret; -} - -void Storage::setTempElement(int idx, llvm::Value *val) -{ - m_temps[idx] = val; -} - -void Storage::store(int dstIdx, llvm::Value *val) -{ - char ptrName[13]; - snprintf(ptrName, 13, "out_ptr%d", dstIdx); - GetElementPtrInst *getElem = new GetElementPtrInst(m_OUT, - constantInt(dstIdx), - ptrName, - m_block); - new StoreInst(val, getElem, false, m_block); -} diff --git a/src/mesa/sources b/src/mesa/sources index 7e1d909e28..1bc45c9813 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -192,6 +192,7 @@ TGSIMESA_SOURCES = \ ifeq ($(MESA_NO_LLVM),0) LLVMTGSI_SOURCES = \ pipe/llvm/llvmtgsi.cpp \ + pipe/llvm/storage.cpp \ pipe/llvm/instructions.cpp endif -- cgit v1.2.3 From fcbde5e9f44ee7254b6618b6fe2be98a1c803ed2 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 17 Oct 2007 12:13:33 -0400 Subject: Add the "add" opcode and assert one vp if an opcode isn't supported --- src/mesa/pipe/llvm/llvmtgsi.cpp | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index c6cf2de4fc..c934c002f0 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -190,7 +190,9 @@ translate_instruction(llvm::Module *module, out = instr->mul(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_ADD: + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_DP3: { out = instr->dp3(inputs[0], inputs[1]); @@ -447,6 +449,7 @@ translate_instruction(llvm::Module *module, case TGSI_OPCODE_KIL: break; case TGSI_OPCODE_END: + return; break; default: fprintf(stderr, "ERROR: Unknown opcode %d\n", @@ -455,6 +458,11 @@ translate_instruction(llvm::Module *module, break; } + if (!out) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } switch( inst->Instruction.Saturate ) { case TGSI_SAT_NONE: -- cgit v1.2.3 From e20294be114c2593035afaf6fe0726e0ce628ed0 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 17 Oct 2007 13:27:25 -0400 Subject: Implement pow/rcp and sub opcodes --- src/mesa/pipe/llvm/instructions.cpp | 66 +++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 10 ++++-- src/mesa/pipe/llvm/llvmtgsi.cpp | 20 +++++++++-- 3 files changed, 91 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 147a1b64f2..31729d0f58 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -12,8 +12,10 @@ Instructions::Instructions(llvm::Module *mod, llvm::BasicBlock *block) : m_mod(mod), m_block(block), m_idx(0) { m_floatVecType = VectorType::get(Type::FloatTy, 4); + m_llvmFSqrt = 0; m_llvmFAbs = 0; + m_llvmPow = 0; } llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) @@ -160,3 +162,67 @@ llvm::Value * Instructions::lit(llvm::Value *in1) return in1; } +llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) +{ + BinaryOperator *res = BinaryOperator::create(Instruction::Sub, in1, in2, + name("sub"), + m_block); + return res; +} + +llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) +{ + if (!m_llvmPow) { + // predeclare the intrinsic + std::vector powArgs; + powArgs.push_back(Type::FloatTy); + powArgs.push_back(Type::FloatTy); + ParamAttrsList *powPal = 0; + FunctionType* powType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/powArgs, + /*isVarArg=*/false, + /*ParamAttrs=*/powPal); + m_llvmPow = new Function( + /*Type=*/powType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"llvm.pow.f32", m_mod); + m_llvmPow->setCallingConv(CallingConv::C); + } + std::vector params; + params.push_back(val1); + params.push_back(val2); + CallInst *call = new CallInst(m_llvmPow, params.begin(), params.end(), + name("pow"), + m_block); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2) +{ + ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), + name("x1"), + m_block); + ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), + name("x2"), + m_block); + llvm::Value *val = callPow(x1, x2); + return vectorFromVals(val, val, val, val); +} + +llvm::Value * Instructions::rcp(llvm::Value *in1) +{ + ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), + name("x1"), + m_block); + BinaryOperator *res = BinaryOperator::create(Instruction::FDiv, + ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + x1, + name("rcp"), + m_block); + return vectorFromVals(res, res, res, res); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index c6e77710ea..18b5f91131 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -20,12 +20,16 @@ public: llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in2); llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); + llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); + llvm::Value *rcp(llvm::Value *in1); llvm::Value *rsq(llvm::Value *in1); + llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); private: const char *name(const char *prefix); llvm::Value *callFSqrt(llvm::Value *val); llvm::Value *callFAbs(llvm::Value *val); + llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2); llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, llvm::Value *z, llvm::Value *w=0); @@ -34,10 +38,12 @@ private: char m_name[32]; llvm::BasicBlock *m_block; int m_idx; - llvm::Function *m_llvmFSqrt; - llvm::Function *m_llvmFAbs; llvm::VectorType *m_floatVecType; + + llvm::Function *m_llvmFSqrt; + llvm::Function *m_llvmFAbs; + llvm::Function *m_llvmPow; }; #endif diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index c934c002f0..45114abe4e 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -125,6 +125,7 @@ translate_instruction(llvm::Module *module, struct tgsi_full_instruction *fi) { llvm::Value *inputs[4]; + printf("translate instr START\n"); for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; llvm::Value *val = 0; @@ -136,6 +137,7 @@ translate_instruction(llvm::Module *module, val = storage->tempElement(src->SrcRegister.Index); } else { fprintf(stderr, "ERROR: not support llvm source\n"); + printf("translate instr END\n"); return; } @@ -154,6 +156,9 @@ translate_instruction(llvm::Module *module, src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) { + fprintf(stderr, "SWIZZLE is %d %d %d %d\n", + src->SrcRegister.SwizzleX, src->SrcRegister.SwizzleY, + src->SrcRegister.SwizzleZ, src->SrcRegister.SwizzleW); int swizzle = src->SrcRegister.SwizzleX * 1000; swizzle += src->SrcRegister.SwizzleY * 100; swizzle += src->SrcRegister.SwizzleZ * 10; @@ -176,7 +181,9 @@ translate_instruction(llvm::Module *module, return; } break; - case TGSI_OPCODE_RCP: + case TGSI_OPCODE_RCP: { + out = instr->rcp(inputs[0]); + } break; case TGSI_OPCODE_RSQ: { out = instr->rsq(inputs[0]); @@ -214,7 +221,9 @@ translate_instruction(llvm::Module *module, out = instr->madd(inputs[0], inputs[1], inputs[2]); } break; - case TGSI_OPCODE_SUB: + case TGSI_OPCODE_SUB: { + out = instr->sub(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_LERP: break; @@ -240,7 +249,9 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_LOGBASE2: break; - case TGSI_OPCODE_POWER: + case TGSI_OPCODE_POWER: { + out = instr->pow(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_CROSSPRODUCT: break; @@ -449,6 +460,7 @@ translate_instruction(llvm::Module *module, case TGSI_OPCODE_KIL: break; case TGSI_OPCODE_END: + printf("translate instr END\n"); return; break; default: @@ -481,6 +493,7 @@ translate_instruction(llvm::Module *module, struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + printf("--- storing to %d %p\n", dst->DstRegister.Index, out); storage->store(dst->DstRegister.Index, out); } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { storage->setTempElement(dst->DstRegister.Index, out); @@ -501,6 +514,7 @@ translate_instruction(llvm::Module *module, } #endif } + printf("translate instr END\n"); } -- cgit v1.2.3 From 7abe3364b2c463fd3e96c2bc9d07aaa91bcfbc2c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 17 Oct 2007 13:34:25 -0400 Subject: Implement dot4 opcode --- src/mesa/pipe/llvm/instructions.cpp | 27 +++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 1 + src/mesa/pipe/llvm/llvmtgsi.cpp | 12 +++++++----- 3 files changed, 35 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 31729d0f58..9e3b989600 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -226,3 +226,30 @@ llvm::Value * Instructions::rcp(llvm::Value *in1) return vectorFromVals(res, res, res, res); } +llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + ExtractElementInst *x = new ExtractElementInst(mulRes, unsigned(0), + name("extractx"), + m_block); + ExtractElementInst *y = new ExtractElementInst(mulRes, unsigned(1), + name("extracty"), + m_block); + ExtractElementInst *z = new ExtractElementInst(mulRes, unsigned(2), + name("extractz"), + m_block); + ExtractElementInst *w = new ExtractElementInst(mulRes, unsigned(3), + name("extractw"), + m_block); + BinaryOperator *xy = BinaryOperator::create(Instruction::Add, x, y, + name("xy"), + m_block); + BinaryOperator *xyz = BinaryOperator::create(Instruction::Add, xy, z, + name("xyz"), + m_block); + BinaryOperator *dot4 = BinaryOperator::create(Instruction::Add, xyz, w, + name("dot4"), + m_block); + return vectorFromVals(dot4, dot4, dot4, dot4); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 18b5f91131..c35cdb75f0 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -16,6 +16,7 @@ public: llvm::Value *add(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); llvm::Value *lit(llvm::Value *in1); llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in2); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 45114abe4e..45bc96cb1a 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -205,7 +205,9 @@ translate_instruction(llvm::Module *module, out = instr->dp3(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_DP4: + case TGSI_OPCODE_DP4: { + out = instr->dp4(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_DST: break; @@ -767,6 +769,10 @@ ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) std::cout << "Creating llvm " <addModuleProvider(mp); } ga_llvm->module = mod; - fprintf(stderr, "DUMPX \n"); - //tgsi_dump(tokens, TGSI_DUMP_VERBOSE); - tgsi_dump(tokens, 0); - fprintf(stderr, "DUMPEND \n"); Function *func = mod->getFunction("run_vertex_shader"); std::cout << "run_vertex_shader = "< Date: Wed, 17 Oct 2007 13:38:53 -0400 Subject: Implement dph opcode --- src/mesa/pipe/llvm/instructions.cpp | 27 +++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 1 + src/mesa/pipe/llvm/llvmtgsi.cpp | 4 +++- 3 files changed, 31 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 9e3b989600..62289ba62a 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -253,3 +253,30 @@ llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) return vectorFromVals(dot4, dot4, dot4, dot4); } +llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + ExtractElementInst *x = new ExtractElementInst(mulRes, unsigned(0), + name("extractx"), + m_block); + ExtractElementInst *y = new ExtractElementInst(mulRes, unsigned(1), + name("extracty"), + m_block); + ExtractElementInst *z = new ExtractElementInst(mulRes, unsigned(2), + name("extractz"), + m_block); + ExtractElementInst *w = new ExtractElementInst(in2, unsigned(3), + name("val2w"), + m_block); + BinaryOperator *xy = BinaryOperator::create(Instruction::Add, x, y, + name("xy"), + m_block); + BinaryOperator *xyz = BinaryOperator::create(Instruction::Add, xy, z, + name("xyz"), + m_block); + BinaryOperator *dph = BinaryOperator::create(Instruction::Add, xyz, w, + name("dph"), + m_block); + return vectorFromVals(dph, dph, dph, dph); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index c35cdb75f0..fb8ac9b8cd 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -17,6 +17,7 @@ public: llvm::Value *add(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); llvm::Value *lit(llvm::Value *in1); llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in2); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 45bc96cb1a..9a5d3b8d9c 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -263,7 +263,9 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_RCC: break; - case TGSI_OPCODE_DPH: + case TGSI_OPCODE_DPH: { + out = instr->dph(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_COS: break; -- cgit v1.2.3 From bd38f4d92125e648b276344ebb2ab9d517b3e7a5 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 17 Oct 2007 13:46:00 -0400 Subject: Implement dst opcode --- src/mesa/pipe/llvm/instructions.cpp | 22 ++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 1 + src/mesa/pipe/llvm/llvmtgsi.cpp | 4 +++- 3 files changed, 26 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 62289ba62a..ee67ae0eed 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -280,3 +280,25 @@ llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) return vectorFromVals(dph, dph, dph, dph); } +llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) +{ + ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), + name("y1"), + m_block); + ExtractElementInst *z = new ExtractElementInst(in1, unsigned(2), + name("z"), + m_block); + ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), + name("y2"), + m_block); + ExtractElementInst *w = new ExtractElementInst(in2, unsigned(3), + name("w"), + m_block); + BinaryOperator *ry = BinaryOperator::create(Instruction::Mul, + y1, y2, + name("tyuy"), + m_block); + return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), + ry, z, w); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index fb8ac9b8cd..0ca9bd25a7 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -18,6 +18,7 @@ public: llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); llvm::Value *lit(llvm::Value *in1); llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in2); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 9a5d3b8d9c..91c2d9beb8 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -209,7 +209,9 @@ translate_instruction(llvm::Module *module, out = instr->dp4(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_DST: + case TGSI_OPCODE_DST: { + out = instr->dst(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_MIN: break; -- cgit v1.2.3 From 1c955171c84f5d062f5f0be983de636a1817c6f7 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 18 Oct 2007 05:51:01 -0400 Subject: Implement ex2 opcode --- src/mesa/pipe/llvm/instructions.cpp | 9 +++++++++ src/mesa/pipe/llvm/instructions.h | 1 + src/mesa/pipe/llvm/llvmtgsi.cpp | 4 +++- 3 files changed, 13 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index ee67ae0eed..1e6b74eac6 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -302,3 +302,12 @@ llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) ry, z, w); } +llvm::Value * Instructions::ex2(llvm::Value *in) +{ + llvm::Value *val = callPow(ConstantFP::get(Type::FloatTy, APFloat(2.f)), + new ExtractElementInst(in, unsigned(0), + name("x1"), + m_block)); + return vectorFromVals(val, val, val, val); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 0ca9bd25a7..de33b79e37 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -19,6 +19,7 @@ public: llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); + llvm::Value *ex2(llvm::Value *in1); llvm::Value *lit(llvm::Value *in1); llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in2); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 91c2d9beb8..5ad45ff6cc 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -249,7 +249,9 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_ROUND: break; - case TGSI_OPCODE_EXPBASE2: + case TGSI_OPCODE_EXPBASE2: { + out = instr->ex2(inputs[0]); + } break; case TGSI_OPCODE_LOGBASE2: break; -- cgit v1.2.3 From c6a1beb18fa5c556c3889f3a5ebdffe51981ad85 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 18 Oct 2007 07:03:37 -0400 Subject: Implement frc and floor --- src/mesa/pipe/llvm/instructions.cpp | 51 +++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 6 ++++- src/mesa/pipe/llvm/llvmtgsi.cpp | 8 ++++-- 3 files changed, 62 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 1e6b74eac6..ee9104434f 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -16,6 +16,7 @@ Instructions::Instructions(llvm::Module *mod, llvm::BasicBlock *block) m_llvmFSqrt = 0; m_llvmFAbs = 0; m_llvmPow = 0; + m_llvmFloor = 0; } llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) @@ -311,3 +312,53 @@ llvm::Value * Instructions::ex2(llvm::Value *in) return vectorFromVals(val, val, val, val); } +llvm::Value * Instructions::callFloor(llvm::Value *val) +{ + if (!m_llvmFloor) { + // predeclare the intrinsic + std::vector floorArgs; + floorArgs.push_back(Type::FloatTy); + ParamAttrsList *floorPal = 0; + FunctionType* floorType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/floorArgs, + /*isVarArg=*/false, + /*ParamAttrs=*/floorPal); + m_llvmFloor = new Function( + /*Type=*/floorType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"floorf", m_mod); + m_llvmFloor->setCallingConv(CallingConv::C); + } + CallInst *call = new CallInst(m_llvmFloor, val, + name("floorf"), + m_block); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::floor(llvm::Value *in) +{ + ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), + name("extractx"), + m_block); + ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), + name("extracty"), + m_block); + ExtractElementInst *z = new ExtractElementInst(in, unsigned(2), + name("extractz"), + m_block); + ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), + name("extractw"), + m_block); + return vectorFromVals(callFloor(x), callFloor(y), + callFloor(z), callFloor(w)); +} + +llvm::Value * Instructions::frc(llvm::Value *in) +{ + llvm::Value *flr = floor(in); + return sub(in, flr); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index de33b79e37..17165ac9e7 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -20,6 +20,8 @@ public: llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); llvm::Value *ex2(llvm::Value *in1); + llvm::Value *floor(llvm::Value *in1); + llvm::Value *frc(llvm::Value *in1); llvm::Value *lit(llvm::Value *in1); llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in2); @@ -31,8 +33,9 @@ public: private: const char *name(const char *prefix); - llvm::Value *callFSqrt(llvm::Value *val); llvm::Value *callFAbs(llvm::Value *val); + llvm::Value *callFloor(llvm::Value *val); + llvm::Value *callFSqrt(llvm::Value *val); llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2); llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, @@ -48,6 +51,7 @@ private: llvm::Function *m_llvmFSqrt; llvm::Function *m_llvmFAbs; llvm::Function *m_llvmPow; + llvm::Function *m_llvmFloor; }; #endif diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 5ad45ff6cc..44caa59d52 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -241,11 +241,15 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_NEGATE: break; - case TGSI_OPCODE_FRAC: + case TGSI_OPCODE_FRAC: { + out = instr->frc(inputs[0]); + } break; case TGSI_OPCODE_CLAMP: break; - case TGSI_OPCODE_FLOOR: + case TGSI_OPCODE_FLOOR: { + out = instr->floor(inputs[0]); + } break; case TGSI_OPCODE_ROUND: break; -- cgit v1.2.3 From 3ae767dd073e8c24fc159cb86f89e61ea7a0e85f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 18 Oct 2007 07:20:27 -0400 Subject: Fix some warnings and implement lg2 --- src/mesa/pipe/llvm/instructions.cpp | 55 ++- src/mesa/pipe/llvm/instructions.h | 15 +- src/mesa/pipe/llvm/llvm_base_shader.cpp | 810 ++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/llvmtgsi.cpp | 4 +- 4 files changed, 875 insertions(+), 9 deletions(-) create mode 100644 src/mesa/pipe/llvm/llvm_base_shader.cpp (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index ee9104434f..d43a617d35 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -14,9 +14,10 @@ Instructions::Instructions(llvm::Module *mod, llvm::BasicBlock *block) m_floatVecType = VectorType::get(Type::FloatTy, 4); m_llvmFSqrt = 0; - m_llvmFAbs = 0; - m_llvmPow = 0; + m_llvmFAbs = 0; + m_llvmPow = 0; m_llvmFloor = 0; + m_llvmFlog = 0; } llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) @@ -362,3 +363,53 @@ llvm::Value * Instructions::frc(llvm::Value *in) return sub(in, flr); } +llvm::Value * Instructions::callFLog(llvm::Value *val) +{ + if (!m_llvmFlog) { + // predeclare the intrinsic + std::vector flogArgs; + flogArgs.push_back(Type::FloatTy); + ParamAttrsList *flogPal = 0; + FunctionType* flogType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/flogArgs, + /*isVarArg=*/false, + /*ParamAttrs=*/flogPal); + m_llvmFlog = new Function( + /*Type=*/flogType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"logf", m_mod); + m_llvmFlog->setCallingConv(CallingConv::C); + } + CallInst *call = new CallInst(m_llvmFlog, val, + name("logf"), + m_block); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::lg2(llvm::Value *in) +{ + ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), + name("extractx"), + m_block); + ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), + name("extracty"), + m_block); + ExtractElementInst *z = new ExtractElementInst(in, unsigned(2), + name("extractz"), + m_block); + ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), + name("extractw"), + m_block); + llvm::Value *const_vec = vectorFromVals( + ConstantFP::get(Type::FloatTy, APFloat(1.442695f)), + ConstantFP::get(Type::FloatTy, APFloat(1.442695f)), + ConstantFP::get(Type::FloatTy, APFloat(1.442695f)), + ConstantFP::get(Type::FloatTy, APFloat(1.442695f)) + ); + return mul(vectorFromVals(callFLog(x), callFLog(y), + callFLog(z), callFLog(w)), const_vec); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 17165ac9e7..e6ad47ebd2 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -19,16 +19,17 @@ public: llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); - llvm::Value *ex2(llvm::Value *in1); - llvm::Value *floor(llvm::Value *in1); - llvm::Value *frc(llvm::Value *in1); - llvm::Value *lit(llvm::Value *in1); + llvm::Value *ex2(llvm::Value *in); + llvm::Value *floor(llvm::Value *in); + llvm::Value *frc(llvm::Value *in); + llvm::Value *lit(llvm::Value *in); + llvm::Value *lg2(llvm::Value *in); llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in2); llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); - llvm::Value *rcp(llvm::Value *in1); - llvm::Value *rsq(llvm::Value *in1); + llvm::Value *rcp(llvm::Value *in); + llvm::Value *rsq(llvm::Value *in); llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); private: const char *name(const char *prefix); @@ -36,6 +37,7 @@ private: llvm::Value *callFAbs(llvm::Value *val); llvm::Value *callFloor(llvm::Value *val); llvm::Value *callFSqrt(llvm::Value *val); + llvm::Value *callFLog(llvm::Value *val); llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2); llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, @@ -52,6 +54,7 @@ private: llvm::Function *m_llvmFAbs; llvm::Function *m_llvmPow; llvm::Function *m_llvmFloor; + llvm::Function *m_llvmFlog; }; #endif diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp new file mode 100644 index 0000000000..58dd57f6b9 --- /dev/null +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -0,0 +1,810 @@ +// Generated by llvm2cpp - DO NOT MODIFY! + + +Module* createBaseShader() { + // Module Construction + Module* mod = new Module("Shader"); + + // Type Definitions + ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 20); + + ArrayType* ArrayTy_2 = ArrayType::get(IntegerType::get(8), 31); + + std::vectorFuncTy_4_args; + VectorType* VectorTy_7 = VectorType::get(Type::FloatTy, 4); + + ArrayType* ArrayTy_6 = ArrayType::get(VectorTy_7, 16); + + PointerType* PointerTy_5 = PointerType::get(ArrayTy_6); + + FuncTy_4_args.push_back(PointerTy_5); + ArrayType* ArrayTy_10 = ArrayType::get(Type::FloatTy, 4); + + ArrayType* ArrayTy_9 = ArrayType::get(ArrayTy_10, 16); + + PointerType* PointerTy_8 = PointerType::get(ArrayTy_9); + + FuncTy_4_args.push_back(PointerTy_8); + FuncTy_4_args.push_back(IntegerType::get(32)); + FuncTy_4_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_4_PAL = 0; + FunctionType* FuncTy_4 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_4_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_4_PAL); + + PointerType* PointerTy_12 = PointerType::get(VectorTy_7); + + std::vectorFuncTy_13_args; + FuncTy_13_args.push_back(PointerTy_12); + PointerType* PointerTy_14 = PointerType::get(ArrayTy_10); + + FuncTy_13_args.push_back(PointerTy_14); + FuncTy_13_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_13_PAL = 0; + FunctionType* FuncTy_13 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_13_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_13_PAL); + + std::vectorFuncTy_15_args; + FuncTy_15_args.push_back(PointerTy_14); + FuncTy_15_args.push_back(PointerTy_12); + FuncTy_15_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_15_PAL = 0; + FunctionType* FuncTy_15 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_15_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_15_PAL); + + std::vectorFuncTy_17_args; + ParamAttrsList *FuncTy_17_PAL = 0; + FunctionType* FuncTy_17 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_17_args, + /*isVarArg=*/true, + /*ParamAttrs=*/FuncTy_17_PAL); + + std::vectorFuncTy_19_args; + FuncTy_19_args.push_back(PointerTy_8); + FuncTy_19_args.push_back(PointerTy_8); + FuncTy_19_args.push_back(PointerTy_14); + FuncTy_19_args.push_back(IntegerType::get(32)); + FuncTy_19_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_19_PAL = 0; + FunctionType* FuncTy_19 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_19_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_19_PAL); + + ArrayType* ArrayTy_21 = ArrayType::get(ArrayTy_6, 2048); + + std::vectorFuncTy_23_args; + FuncTy_23_args.push_back(PointerTy_12); + FuncTy_23_args.push_back(PointerTy_12); + FuncTy_23_args.push_back(PointerTy_12); + ParamAttrsList *FuncTy_23_PAL = 0; + FunctionType* FuncTy_23 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_23_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_23_PAL); + + + // Function Declarations + + Function* func_from_array = new Function( + /*Type=*/FuncTy_4, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"from_array", mod); + func_from_array->setCallingConv(CallingConv::C); + + Function* func_from_consts = new Function( + /*Type=*/FuncTy_13, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"from_consts", mod); + func_from_consts->setCallingConv(CallingConv::C); + + Function* func_to_array = new Function( + /*Type=*/FuncTy_15, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"to_array", mod); + func_to_array->setCallingConv(CallingConv::C); + + Function* func_printf = new Function( + /*Type=*/FuncTy_17, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", mod); // (external, no body) + func_printf->setCallingConv(CallingConv::C); + + Function* func_run_vertex_shader = new Function( + /*Type=*/FuncTy_19, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"run_vertex_shader", mod); + func_run_vertex_shader->setCallingConv(CallingConv::C); + + Function* func_execute_shader = new Function( + /*Type=*/FuncTy_23, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"execute_shader", mod); // (external, no body) + func_execute_shader->setCallingConv(CallingConv::C); + + // Global Variable Declarations + + + GlobalVariable* gvar_array__str = new GlobalVariable( + /*Type=*/ArrayTy_0, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str", + mod); + + GlobalVariable* gvar_array__str1 = new GlobalVariable( + /*Type=*/ArrayTy_2, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str1", + mod); + + // Constant Definitions + Constant* const_array_24 = ConstantArray::get("DEST = %f %f %f %f\x0A", true); + Constant* const_array_25 = ConstantArray::get("XXXXXXXXXXX run_vertex_shader\x0A", true); + Constant* const_int32_26 = Constant::getNullValue(IntegerType::get(32)); + UndefValue* const_packed_27 = UndefValue::get(VectorTy_7); + ConstantInt* const_int32_28 = ConstantInt::get(APInt(32, "1", 10)); + ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "2", 10)); + ConstantInt* const_int32_30 = ConstantInt::get(APInt(32, "3", 10)); + std::vector const_ptr_31_indices; + const_ptr_31_indices.push_back(const_int32_26); + const_ptr_31_indices.push_back(const_int32_26); + Constant* const_ptr_31 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_31_indices[0], const_ptr_31_indices.size() ); + std::vector const_ptr_32_indices; + const_ptr_32_indices.push_back(const_int32_26); + const_ptr_32_indices.push_back(const_int32_26); + Constant* const_ptr_32 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_32_indices[0], const_ptr_32_indices.size() ); + ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "4", 10)); + + // Global Variable Definitions + gvar_array__str->setInitializer(const_array_24); + gvar_array__str1->setInitializer(const_array_25); + + // Function Definitions + + // Function: from_array (func_from_array) + { + Function::arg_iterator args = func_from_array->arg_begin(); + Value* ptr_res = args++; + ptr_res->setName("res"); + Value* ptr_ainputs = args++; + ptr_ainputs->setName("ainputs"); + Value* int32_count = args++; + int32_count->setName("count"); + Value* int32_num_attribs = args++; + int32_num_attribs->setName("num_attribs"); + + BasicBlock* label_entry = new BasicBlock("entry",func_from_array,0); + BasicBlock* label_forbody6 = new BasicBlock("forbody6",func_from_array,0); + BasicBlock* label_forinc57 = new BasicBlock("forinc57",func_from_array,0); + BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); + + // Block entry (label_entry) + ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_26, "cmp", label_entry); + ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_26, "cmp5", label_entry); + BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); + new BranchInst(label_forbody6, label_afterfor60, int1_bothcond, label_entry); + + // Block forbody6 (label_forbody6) + Argument* fwdref_35 = new Argument(IntegerType::get(32)); + Argument* fwdref_36 = new Argument(IntegerType::get(32)); + PHINode* int32_i_013_0 = new PHINode(IntegerType::get(32), "i.013.0", label_forbody6); + int32_i_013_0->reserveOperandSpace(3); + int32_i_013_0->addIncoming(const_int32_26, label_entry); + int32_i_013_0->addIncoming(fwdref_35, label_forinc57); + int32_i_013_0->addIncoming(fwdref_36, label_forbody6); + + Argument* fwdref_37 = new Argument(IntegerType::get(32)); + PHINode* int32_j_03_0 = new PHINode(IntegerType::get(32), "j.03.0", label_forbody6); + int32_j_03_0->reserveOperandSpace(3); + int32_j_03_0->addIncoming(fwdref_37, label_forbody6); + int32_j_03_0->addIncoming(const_int32_26, label_forinc57); + int32_j_03_0->addIncoming(const_int32_26, label_entry); + + Argument* fwdref_38 = new Argument(VectorTy_7); + PHINode* packed_vec_01_0 = new PHINode(VectorTy_7, "vec.01.0", label_forbody6); + packed_vec_01_0->reserveOperandSpace(3); + packed_vec_01_0->addIncoming(fwdref_38, label_forbody6); + packed_vec_01_0->addIncoming(const_packed_27, label_entry); + packed_vec_01_0->addIncoming(fwdref_38, label_forinc57); + + std::vector ptr_arraydecay11_indices; + ptr_arraydecay11_indices.push_back(int32_i_013_0); + ptr_arraydecay11_indices.push_back(int32_j_03_0); + ptr_arraydecay11_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); + LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); + InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_01_0, float_tmp13, const_int32_26, "tmp15", label_forbody6); + std::vector ptr_arrayidx23_indices; + ptr_arrayidx23_indices.push_back(int32_i_013_0); + ptr_arrayidx23_indices.push_back(int32_j_03_0); + ptr_arrayidx23_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); + LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_28, "tmp26", label_forbody6); + std::vector ptr_arrayidx34_indices; + ptr_arrayidx34_indices.push_back(int32_i_013_0); + ptr_arrayidx34_indices.push_back(int32_j_03_0); + ptr_arrayidx34_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6); + LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6); + InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_29, "tmp37", label_forbody6); + std::vector ptr_arrayidx45_indices; + ptr_arrayidx45_indices.push_back(int32_i_013_0); + ptr_arrayidx45_indices.push_back(int32_j_03_0); + ptr_arrayidx45_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6); + LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6); + InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_30, "tmp48", label_forbody6); + std::vector ptr_arrayidx54_indices; + ptr_arrayidx54_indices.push_back(int32_i_013_0); + ptr_arrayidx54_indices.push_back(int32_j_03_0); + Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); + new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); + BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_03_0, const_int32_28, "inc", label_forbody6); + ICmpInst* int1_cmp511 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp511", label_forbody6); + new BranchInst(label_forbody6, label_forinc57, int1_cmp511, label_forbody6); + + // Block forinc57 (label_forinc57) + BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_013_0, const_int32_28, "inc59", label_forinc57); + ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59, int32_count, "cmp21", label_forinc57); + new BranchInst(label_forbody6, label_afterfor60, int1_cmp21, label_forinc57); + + // Block afterfor60 (label_afterfor60) + new ReturnInst(label_afterfor60); + + // Resolve Forward References + fwdref_36->replaceAllUsesWith(int32_i_013_0); delete fwdref_36; + fwdref_38->replaceAllUsesWith(packed_tmp48); delete fwdref_38; + fwdref_37->replaceAllUsesWith(int32_inc); delete fwdref_37; + fwdref_35->replaceAllUsesWith(int32_inc59); delete fwdref_35; + + } + + // Function: from_consts (func_from_consts) + { + Function::arg_iterator args = func_from_consts->arg_begin(); + Value* ptr_res_43 = args++; + ptr_res_43->setName("res"); + Value* ptr_ainputs_44 = args++; + ptr_ainputs_44->setName("ainputs"); + Value* int32_count_45 = args++; + int32_count_45->setName("count"); + + BasicBlock* label_entry_46 = new BasicBlock("entry",func_from_consts,0); + BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); + BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); + + // Block entry (label_entry_46) + ICmpInst* int1_cmp_47 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_45, const_int32_26, "cmp", label_entry_46); + new BranchInst(label_forbody, label_afterfor, int1_cmp_47, label_entry_46); + + // Block forbody (label_forbody) + Argument* fwdref_49 = new Argument(IntegerType::get(32)); + PHINode* int32_i_02_0 = new PHINode(IntegerType::get(32), "i.02.0", label_forbody); + int32_i_02_0->reserveOperandSpace(2); + int32_i_02_0->addIncoming(const_int32_26, label_entry_46); + int32_i_02_0->addIncoming(fwdref_49, label_forbody); + + Argument* fwdref_51 = new Argument(VectorTy_7); + PHINode* packed_vec_01_0_50 = new PHINode(VectorTy_7, "vec.01.0", label_forbody); + packed_vec_01_0_50->reserveOperandSpace(2); + packed_vec_01_0_50->addIncoming(const_packed_27, label_entry_46); + packed_vec_01_0_50->addIncoming(fwdref_51, label_forbody); + + std::vector ptr_arraydecay_indices; + ptr_arraydecay_indices.push_back(int32_i_02_0); + ptr_arraydecay_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_44, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); + LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); + InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_01_0_50, float_tmp5, const_int32_26, "tmp7", label_forbody); + std::vector ptr_arrayidx12_indices; + ptr_arrayidx12_indices.push_back(int32_i_02_0); + ptr_arrayidx12_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_44, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); + LoadInst* float_tmp13_52 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); + InsertElementInst* packed_tmp15_53 = new InsertElementInst(packed_tmp7, float_tmp13_52, const_int32_28, "tmp15", label_forbody); + std::vector ptr_arrayidx20_indices; + ptr_arrayidx20_indices.push_back(int32_i_02_0); + ptr_arrayidx20_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_44, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); + LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); + InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_53, float_tmp21, const_int32_29, "tmp23", label_forbody); + std::vector ptr_arrayidx28_indices; + ptr_arrayidx28_indices.push_back(int32_i_02_0); + ptr_arrayidx28_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_44, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); + LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); + InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_30, "tmp31", label_forbody); + GetElementPtrInst* ptr_arrayidx34_54 = new GetElementPtrInst(ptr_res_43, int32_i_02_0, "arrayidx34", label_forbody); + new StoreInst(packed_tmp31, ptr_arrayidx34_54, false, label_forbody); + BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_02_0, const_int32_28, "indvar.next", label_forbody); + ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_45, "exitcond", label_forbody); + new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); + + // Block afterfor (label_afterfor) + new ReturnInst(label_afterfor); + + // Resolve Forward References + fwdref_51->replaceAllUsesWith(packed_tmp31); delete fwdref_51; + fwdref_49->replaceAllUsesWith(int32_indvar_next); delete fwdref_49; + + } + + // Function: to_array (func_to_array) + { + Function::arg_iterator args = func_to_array->arg_begin(); + Value* ptr_dests = args++; + ptr_dests->setName("dests"); + Value* ptr_in = args++; + ptr_in->setName("in"); + Value* int32_num_attribs_58 = args++; + int32_num_attribs_58->setName("num_attribs"); + + BasicBlock* label_entry_59 = new BasicBlock("entry",func_to_array,0); + BasicBlock* label_forbody_60 = new BasicBlock("forbody",func_to_array,0); + BasicBlock* label_afterfor_61 = new BasicBlock("afterfor",func_to_array,0); + + // Block entry (label_entry_59) + ICmpInst* int1_cmp_62 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_58, const_int32_26, "cmp", label_entry_59); + new BranchInst(label_forbody_60, label_afterfor_61, int1_cmp_62, label_entry_59); + + // Block forbody (label_forbody_60) + Argument* fwdref_64 = new Argument(IntegerType::get(32)); + PHINode* int32_i_01_0 = new PHINode(IntegerType::get(32), "i.01.0", label_forbody_60); + int32_i_01_0->reserveOperandSpace(2); + int32_i_01_0->addIncoming(const_int32_26, label_entry_59); + int32_i_01_0->addIncoming(fwdref_64, label_forbody_60); + + std::vector ptr_arraydecay_65_indices; + ptr_arraydecay_65_indices.push_back(int32_i_01_0); + ptr_arraydecay_65_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay_65 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_65_indices.begin(), ptr_arraydecay_65_indices.end(), "arraydecay", label_forbody_60); + GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_01_0, "arrayidx6", label_forbody_60); + LoadInst* packed_tmp7_66 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_60); + ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_tmp7_66, const_int32_26, "tmp9", label_forbody_60); + CastInst* double_conv = new FPExtInst(float_tmp9, Type::DoubleTy, "conv", label_forbody_60); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_66, const_int32_28, "tmp11", label_forbody_60); + CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_forbody_60); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp7_66, const_int32_29, "tmp14", label_forbody_60); + CastInst* double_conv15 = new FPExtInst(float_tmp14, Type::DoubleTy, "conv15", label_forbody_60); + ExtractElementInst* float_tmp17 = new ExtractElementInst(packed_tmp7_66, const_int32_30, "tmp17", label_forbody_60); + CastInst* double_conv18 = new FPExtInst(float_tmp17, Type::DoubleTy, "conv18", label_forbody_60); + std::vector int32_call_params; + int32_call_params.push_back(const_ptr_31); + int32_call_params.push_back(double_conv); + int32_call_params.push_back(double_conv12); + int32_call_params.push_back(double_conv15); + int32_call_params.push_back(double_conv18); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_forbody_60); + int32_call->setCallingConv(CallingConv::C); + int32_call->setTailCall(true); + new StoreInst(float_tmp9, ptr_arraydecay_65, false, label_forbody_60); + std::vector ptr_arrayidx24_indices; + ptr_arrayidx24_indices.push_back(int32_i_01_0); + ptr_arrayidx24_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx24 = new GetElementPtrInst(ptr_dests, ptr_arrayidx24_indices.begin(), ptr_arrayidx24_indices.end(), "arrayidx24", label_forbody_60); + new StoreInst(float_tmp11, ptr_arrayidx24, false, label_forbody_60); + std::vector ptr_arrayidx28_69_indices; + ptr_arrayidx28_69_indices.push_back(int32_i_01_0); + ptr_arrayidx28_69_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx28_69 = new GetElementPtrInst(ptr_dests, ptr_arrayidx28_69_indices.begin(), ptr_arrayidx28_69_indices.end(), "arrayidx28", label_forbody_60); + new StoreInst(float_tmp14, ptr_arrayidx28_69, false, label_forbody_60); + std::vector ptr_arrayidx32_indices; + ptr_arrayidx32_indices.push_back(int32_i_01_0); + ptr_arrayidx32_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx32 = new GetElementPtrInst(ptr_dests, ptr_arrayidx32_indices.begin(), ptr_arrayidx32_indices.end(), "arrayidx32", label_forbody_60); + new StoreInst(float_tmp17, ptr_arrayidx32, false, label_forbody_60); + BinaryOperator* int32_indvar_next_72 = BinaryOperator::create(Instruction::Add, int32_i_01_0, const_int32_28, "indvar.next", label_forbody_60); + ICmpInst* int1_exitcond_73 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_72, int32_num_attribs_58, "exitcond", label_forbody_60); + new BranchInst(label_afterfor_61, label_forbody_60, int1_exitcond_73, label_forbody_60); + + // Block afterfor (label_afterfor_61) + new ReturnInst(label_afterfor_61); + + // Resolve Forward References + fwdref_64->replaceAllUsesWith(int32_indvar_next_72); delete fwdref_64; + + } + + // Function: run_vertex_shader (func_run_vertex_shader) + { + Function::arg_iterator args = func_run_vertex_shader->arg_begin(); + Value* ptr_ainputs_76 = args++; + ptr_ainputs_76->setName("ainputs"); + Value* ptr_dests_77 = args++; + ptr_dests_77->setName("dests"); + Value* ptr_aconsts = args++; + ptr_aconsts->setName("aconsts"); + Value* int32_count_78 = args++; + int32_count_78->setName("count"); + Value* int32_num_attribs_79 = args++; + int32_num_attribs_79->setName("num_attribs"); + + BasicBlock* label_entry_80 = new BasicBlock("entry",func_run_vertex_shader,0); + BasicBlock* label_forbody6_i = new BasicBlock("forbody6.i",func_run_vertex_shader,0); + BasicBlock* label_forinc57_i = new BasicBlock("forinc57.i",func_run_vertex_shader,0); + BasicBlock* label_forbody_i11_preheader = new BasicBlock("forbody.i11.preheader",func_run_vertex_shader,0); + BasicBlock* label_forbody_preheader = new BasicBlock("forbody.preheader",func_run_vertex_shader,0); + BasicBlock* label_forbody_us = new BasicBlock("forbody.us",func_run_vertex_shader,0); + BasicBlock* label_to_array_exit_us = new BasicBlock("to_array.exit.us",func_run_vertex_shader,0); + BasicBlock* label_forbody_i_us = new BasicBlock("forbody.i.us",func_run_vertex_shader,0); + BasicBlock* label_forbody_81 = new BasicBlock("forbody",func_run_vertex_shader,0); + BasicBlock* label_afterfor_82 = new BasicBlock("afterfor",func_run_vertex_shader,0); + + // Block entry (label_entry_80) + AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_21, "inputs", label_entry_80); + AllocaInst* ptr_consts = new AllocaInst(ArrayTy_6, "consts", label_entry_80); + AllocaInst* ptr_results = new AllocaInst(ArrayTy_21, "results", label_entry_80); + CallInst* int32_call_83 = new CallInst(func_printf, const_ptr_32, "call", label_entry_80); + int32_call_83->setCallingConv(CallingConv::C); + int32_call_83->setTailCall(false); + ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_78, const_int32_26, "cmp.i", label_entry_80); + ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_79, const_int32_26, "cmp5.i", label_entry_80); + BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_80); + new BranchInst(label_forbody6_i, label_forbody_i11_preheader, int1_bothcond_i, label_entry_80); + + // Block forbody6.i (label_forbody6_i) + Argument* fwdref_85 = new Argument(IntegerType::get(32)); + Argument* fwdref_86 = new Argument(IntegerType::get(32)); + PHINode* int32_i_013_0_i_ph = new PHINode(IntegerType::get(32), "i.013.0.i.ph", label_forbody6_i); + int32_i_013_0_i_ph->reserveOperandSpace(3); + int32_i_013_0_i_ph->addIncoming(const_int32_26, label_entry_80); + int32_i_013_0_i_ph->addIncoming(fwdref_85, label_forinc57_i); + int32_i_013_0_i_ph->addIncoming(fwdref_86, label_forbody6_i); + + Argument* fwdref_87 = new Argument(IntegerType::get(32)); + PHINode* int32_j_03_0_i = new PHINode(IntegerType::get(32), "j.03.0.i", label_forbody6_i); + int32_j_03_0_i->reserveOperandSpace(3); + int32_j_03_0_i->addIncoming(fwdref_87, label_forbody6_i); + int32_j_03_0_i->addIncoming(const_int32_26, label_forinc57_i); + int32_j_03_0_i->addIncoming(const_int32_26, label_entry_80); + + Argument* fwdref_88 = new Argument(VectorTy_7); + PHINode* packed_vec_01_0_i = new PHINode(VectorTy_7, "vec.01.0.i", label_forbody6_i); + packed_vec_01_0_i->reserveOperandSpace(3); + packed_vec_01_0_i->addIncoming(fwdref_88, label_forbody6_i); + packed_vec_01_0_i->addIncoming(const_packed_27, label_entry_80); + packed_vec_01_0_i->addIncoming(fwdref_88, label_forinc57_i); + + std::vector ptr_arraydecay11_i_indices; + ptr_arraydecay11_i_indices.push_back(int32_i_013_0_i_ph); + ptr_arraydecay11_i_indices.push_back(int32_j_03_0_i); + ptr_arraydecay11_i_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_76, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); + LoadInst* float_tmp13_i = new LoadInst(ptr_arraydecay11_i, "tmp13.i", false, label_forbody6_i); + InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_01_0_i, float_tmp13_i, const_int32_26, "tmp15.i", label_forbody6_i); + std::vector ptr_arrayidx23_i_indices; + ptr_arrayidx23_i_indices.push_back(int32_i_013_0_i_ph); + ptr_arrayidx23_i_indices.push_back(int32_j_03_0_i); + ptr_arrayidx23_i_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_76, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); + LoadInst* float_tmp24_i = new LoadInst(ptr_arrayidx23_i, "tmp24.i", false, label_forbody6_i); + InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_28, "tmp26.i", label_forbody6_i); + std::vector ptr_arrayidx34_i_indices; + ptr_arrayidx34_i_indices.push_back(int32_i_013_0_i_ph); + ptr_arrayidx34_i_indices.push_back(int32_j_03_0_i); + ptr_arrayidx34_i_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_76, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); + LoadInst* float_tmp35_i = new LoadInst(ptr_arrayidx34_i, "tmp35.i", false, label_forbody6_i); + InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_29, "tmp37.i", label_forbody6_i); + std::vector ptr_arrayidx45_i_indices; + ptr_arrayidx45_i_indices.push_back(int32_i_013_0_i_ph); + ptr_arrayidx45_i_indices.push_back(int32_j_03_0_i); + ptr_arrayidx45_i_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_76, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); + LoadInst* float_tmp46_i = new LoadInst(ptr_arrayidx45_i, "tmp46.i", false, label_forbody6_i); + InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_30, "tmp48.i", label_forbody6_i); + std::vector ptr_arrayidx54_i_indices; + ptr_arrayidx54_i_indices.push_back(const_int32_26); + ptr_arrayidx54_i_indices.push_back(int32_i_013_0_i_ph); + ptr_arrayidx54_i_indices.push_back(int32_j_03_0_i); + Instruction* ptr_arrayidx54_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx54_i_indices.begin(), ptr_arrayidx54_i_indices.end(), "arrayidx54.i", label_forbody6_i); + new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); + BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_03_0_i, const_int32_28, "inc.i", label_forbody6_i); + ICmpInst* int1_cmp511_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_attribs_79, "cmp511.i", label_forbody6_i); + new BranchInst(label_forbody6_i, label_forinc57_i, int1_cmp511_i, label_forbody6_i); + + // Block forinc57.i (label_forinc57_i) + BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_013_0_i_ph, const_int32_28, "inc59.i", label_forinc57_i); + ICmpInst* int1_cmp21_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59_i, int32_count_78, "cmp21.i", label_forinc57_i); + new BranchInst(label_forbody6_i, label_forbody_i11_preheader, int1_cmp21_i, label_forinc57_i); + + // Block forbody.i11.preheader (label_forbody_i11_preheader) + std::vector ptr_arraydecay_i3_indices; + ptr_arraydecay_i3_indices.push_back(const_int32_26); + ptr_arraydecay_i3_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay_i3 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_indices.begin(), ptr_arraydecay_i3_indices.end(), "arraydecay.i3", label_forbody_i11_preheader); + LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i3, "tmp5.i", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp7_i4 = new InsertElementInst(const_packed_27, float_tmp5_i, const_int32_26, "tmp7.i4", label_forbody_i11_preheader); + std::vector ptr_arrayidx12_i_indices; + ptr_arrayidx12_i_indices.push_back(const_int32_26); + ptr_arrayidx12_i_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i11_preheader); + LoadInst* float_tmp13_i5 = new LoadInst(ptr_arrayidx12_i, "tmp13.i5", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp15_i6 = new InsertElementInst(packed_tmp7_i4, float_tmp13_i5, const_int32_28, "tmp15.i6", label_forbody_i11_preheader); + std::vector ptr_arrayidx20_i_indices; + ptr_arrayidx20_i_indices.push_back(const_int32_26); + ptr_arrayidx20_i_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i11_preheader); + LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp23_i = new InsertElementInst(packed_tmp15_i6, float_tmp21_i, const_int32_29, "tmp23.i", label_forbody_i11_preheader); + std::vector ptr_arrayidx28_i7_indices; + ptr_arrayidx28_i7_indices.push_back(const_int32_26); + ptr_arrayidx28_i7_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx28_i7 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_indices.begin(), ptr_arrayidx28_i7_indices.end(), "arrayidx28.i7", label_forbody_i11_preheader); + LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i7, "tmp29.i", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i, float_tmp29_i, const_int32_30, "tmp31.i", label_forbody_i11_preheader); + std::vector ptr_arrayidx34_i8_indices; + ptr_arrayidx34_i8_indices.push_back(const_int32_26); + ptr_arrayidx34_i8_indices.push_back(const_int32_26); + Instruction* ptr_arrayidx34_i8 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_indices.begin(), ptr_arrayidx34_i8_indices.end(), "arrayidx34.i8", label_forbody_i11_preheader); + new StoreInst(packed_tmp31_i, ptr_arrayidx34_i8, false, label_forbody_i11_preheader); + std::vector ptr_arraydecay_i3_1_indices; + ptr_arraydecay_i3_1_indices.push_back(const_int32_28); + ptr_arraydecay_i3_1_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay_i3_1 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_1_indices.begin(), ptr_arraydecay_i3_1_indices.end(), "arraydecay.i3.1", label_forbody_i11_preheader); + LoadInst* float_tmp5_i_1 = new LoadInst(ptr_arraydecay_i3_1, "tmp5.i.1", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp7_i4_1 = new InsertElementInst(packed_tmp31_i, float_tmp5_i_1, const_int32_26, "tmp7.i4.1", label_forbody_i11_preheader); + std::vector ptr_arrayidx12_i_1_indices; + ptr_arrayidx12_i_1_indices.push_back(const_int32_28); + ptr_arrayidx12_i_1_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx12_i_1 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_1_indices.begin(), ptr_arrayidx12_i_1_indices.end(), "arrayidx12.i.1", label_forbody_i11_preheader); + LoadInst* float_tmp13_i5_1 = new LoadInst(ptr_arrayidx12_i_1, "tmp13.i5.1", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp15_i6_1 = new InsertElementInst(packed_tmp7_i4_1, float_tmp13_i5_1, const_int32_28, "tmp15.i6.1", label_forbody_i11_preheader); + std::vector ptr_arrayidx20_i_1_indices; + ptr_arrayidx20_i_1_indices.push_back(const_int32_28); + ptr_arrayidx20_i_1_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx20_i_1 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_1_indices.begin(), ptr_arrayidx20_i_1_indices.end(), "arrayidx20.i.1", label_forbody_i11_preheader); + LoadInst* float_tmp21_i_1 = new LoadInst(ptr_arrayidx20_i_1, "tmp21.i.1", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp23_i_1 = new InsertElementInst(packed_tmp15_i6_1, float_tmp21_i_1, const_int32_29, "tmp23.i.1", label_forbody_i11_preheader); + std::vector ptr_arrayidx28_i7_1_indices; + ptr_arrayidx28_i7_1_indices.push_back(const_int32_28); + ptr_arrayidx28_i7_1_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx28_i7_1 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_1_indices.begin(), ptr_arrayidx28_i7_1_indices.end(), "arrayidx28.i7.1", label_forbody_i11_preheader); + LoadInst* float_tmp29_i_1 = new LoadInst(ptr_arrayidx28_i7_1, "tmp29.i.1", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp31_i_1 = new InsertElementInst(packed_tmp23_i_1, float_tmp29_i_1, const_int32_30, "tmp31.i.1", label_forbody_i11_preheader); + std::vector ptr_arrayidx34_i8_1_indices; + ptr_arrayidx34_i8_1_indices.push_back(const_int32_26); + ptr_arrayidx34_i8_1_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx34_i8_1 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_1_indices.begin(), ptr_arrayidx34_i8_1_indices.end(), "arrayidx34.i8.1", label_forbody_i11_preheader); + new StoreInst(packed_tmp31_i_1, ptr_arrayidx34_i8_1, false, label_forbody_i11_preheader); + std::vector ptr_arraydecay_i3_2_indices; + ptr_arraydecay_i3_2_indices.push_back(const_int32_29); + ptr_arraydecay_i3_2_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay_i3_2 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_2_indices.begin(), ptr_arraydecay_i3_2_indices.end(), "arraydecay.i3.2", label_forbody_i11_preheader); + LoadInst* float_tmp5_i_2 = new LoadInst(ptr_arraydecay_i3_2, "tmp5.i.2", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp7_i4_2 = new InsertElementInst(packed_tmp31_i_1, float_tmp5_i_2, const_int32_26, "tmp7.i4.2", label_forbody_i11_preheader); + std::vector ptr_arrayidx12_i_2_indices; + ptr_arrayidx12_i_2_indices.push_back(const_int32_29); + ptr_arrayidx12_i_2_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx12_i_2 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_2_indices.begin(), ptr_arrayidx12_i_2_indices.end(), "arrayidx12.i.2", label_forbody_i11_preheader); + LoadInst* float_tmp13_i5_2 = new LoadInst(ptr_arrayidx12_i_2, "tmp13.i5.2", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp15_i6_2 = new InsertElementInst(packed_tmp7_i4_2, float_tmp13_i5_2, const_int32_28, "tmp15.i6.2", label_forbody_i11_preheader); + std::vector ptr_arrayidx20_i_2_indices; + ptr_arrayidx20_i_2_indices.push_back(const_int32_29); + ptr_arrayidx20_i_2_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx20_i_2 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_2_indices.begin(), ptr_arrayidx20_i_2_indices.end(), "arrayidx20.i.2", label_forbody_i11_preheader); + LoadInst* float_tmp21_i_2 = new LoadInst(ptr_arrayidx20_i_2, "tmp21.i.2", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp23_i_2 = new InsertElementInst(packed_tmp15_i6_2, float_tmp21_i_2, const_int32_29, "tmp23.i.2", label_forbody_i11_preheader); + std::vector ptr_arrayidx28_i7_2_indices; + ptr_arrayidx28_i7_2_indices.push_back(const_int32_29); + ptr_arrayidx28_i7_2_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx28_i7_2 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_2_indices.begin(), ptr_arrayidx28_i7_2_indices.end(), "arrayidx28.i7.2", label_forbody_i11_preheader); + LoadInst* float_tmp29_i_2 = new LoadInst(ptr_arrayidx28_i7_2, "tmp29.i.2", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp31_i_2 = new InsertElementInst(packed_tmp23_i_2, float_tmp29_i_2, const_int32_30, "tmp31.i.2", label_forbody_i11_preheader); + std::vector ptr_arrayidx34_i8_2_indices; + ptr_arrayidx34_i8_2_indices.push_back(const_int32_26); + ptr_arrayidx34_i8_2_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx34_i8_2 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_2_indices.begin(), ptr_arrayidx34_i8_2_indices.end(), "arrayidx34.i8.2", label_forbody_i11_preheader); + new StoreInst(packed_tmp31_i_2, ptr_arrayidx34_i8_2, false, label_forbody_i11_preheader); + std::vector ptr_arraydecay_i3_3_indices; + ptr_arraydecay_i3_3_indices.push_back(const_int32_30); + ptr_arraydecay_i3_3_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay_i3_3 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_3_indices.begin(), ptr_arraydecay_i3_3_indices.end(), "arraydecay.i3.3", label_forbody_i11_preheader); + LoadInst* float_tmp5_i_3 = new LoadInst(ptr_arraydecay_i3_3, "tmp5.i.3", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp7_i4_3 = new InsertElementInst(packed_tmp31_i_2, float_tmp5_i_3, const_int32_26, "tmp7.i4.3", label_forbody_i11_preheader); + std::vector ptr_arrayidx12_i_3_indices; + ptr_arrayidx12_i_3_indices.push_back(const_int32_30); + ptr_arrayidx12_i_3_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx12_i_3 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_3_indices.begin(), ptr_arrayidx12_i_3_indices.end(), "arrayidx12.i.3", label_forbody_i11_preheader); + LoadInst* float_tmp13_i5_3 = new LoadInst(ptr_arrayidx12_i_3, "tmp13.i5.3", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp15_i6_3 = new InsertElementInst(packed_tmp7_i4_3, float_tmp13_i5_3, const_int32_28, "tmp15.i6.3", label_forbody_i11_preheader); + std::vector ptr_arrayidx20_i_3_indices; + ptr_arrayidx20_i_3_indices.push_back(const_int32_30); + ptr_arrayidx20_i_3_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx20_i_3 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_3_indices.begin(), ptr_arrayidx20_i_3_indices.end(), "arrayidx20.i.3", label_forbody_i11_preheader); + LoadInst* float_tmp21_i_3 = new LoadInst(ptr_arrayidx20_i_3, "tmp21.i.3", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp23_i_3 = new InsertElementInst(packed_tmp15_i6_3, float_tmp21_i_3, const_int32_29, "tmp23.i.3", label_forbody_i11_preheader); + std::vector ptr_arrayidx28_i7_3_indices; + ptr_arrayidx28_i7_3_indices.push_back(const_int32_30); + ptr_arrayidx28_i7_3_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx28_i7_3 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_3_indices.begin(), ptr_arrayidx28_i7_3_indices.end(), "arrayidx28.i7.3", label_forbody_i11_preheader); + LoadInst* float_tmp29_i_3 = new LoadInst(ptr_arrayidx28_i7_3, "tmp29.i.3", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp31_i_3 = new InsertElementInst(packed_tmp23_i_3, float_tmp29_i_3, const_int32_30, "tmp31.i.3", label_forbody_i11_preheader); + std::vector ptr_arrayidx34_i8_3_indices; + ptr_arrayidx34_i8_3_indices.push_back(const_int32_26); + ptr_arrayidx34_i8_3_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx34_i8_3 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_3_indices.begin(), ptr_arrayidx34_i8_3_indices.end(), "arrayidx34.i8.3", label_forbody_i11_preheader); + new StoreInst(packed_tmp31_i_3, ptr_arrayidx34_i8_3, false, label_forbody_i11_preheader); + std::vector ptr_arraydecay_i3_4_indices; + ptr_arraydecay_i3_4_indices.push_back(const_int32_33); + ptr_arraydecay_i3_4_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay_i3_4 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_4_indices.begin(), ptr_arraydecay_i3_4_indices.end(), "arraydecay.i3.4", label_forbody_i11_preheader); + LoadInst* float_tmp5_i_4 = new LoadInst(ptr_arraydecay_i3_4, "tmp5.i.4", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp7_i4_4 = new InsertElementInst(packed_tmp31_i_3, float_tmp5_i_4, const_int32_26, "tmp7.i4.4", label_forbody_i11_preheader); + std::vector ptr_arrayidx12_i_4_indices; + ptr_arrayidx12_i_4_indices.push_back(const_int32_33); + ptr_arrayidx12_i_4_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx12_i_4 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_4_indices.begin(), ptr_arrayidx12_i_4_indices.end(), "arrayidx12.i.4", label_forbody_i11_preheader); + LoadInst* float_tmp13_i5_4 = new LoadInst(ptr_arrayidx12_i_4, "tmp13.i5.4", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp15_i6_4 = new InsertElementInst(packed_tmp7_i4_4, float_tmp13_i5_4, const_int32_28, "tmp15.i6.4", label_forbody_i11_preheader); + std::vector ptr_arrayidx20_i_4_indices; + ptr_arrayidx20_i_4_indices.push_back(const_int32_33); + ptr_arrayidx20_i_4_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx20_i_4 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_4_indices.begin(), ptr_arrayidx20_i_4_indices.end(), "arrayidx20.i.4", label_forbody_i11_preheader); + LoadInst* float_tmp21_i_4 = new LoadInst(ptr_arrayidx20_i_4, "tmp21.i.4", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp23_i_4 = new InsertElementInst(packed_tmp15_i6_4, float_tmp21_i_4, const_int32_29, "tmp23.i.4", label_forbody_i11_preheader); + std::vector ptr_arrayidx28_i7_4_indices; + ptr_arrayidx28_i7_4_indices.push_back(const_int32_33); + ptr_arrayidx28_i7_4_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx28_i7_4 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_4_indices.begin(), ptr_arrayidx28_i7_4_indices.end(), "arrayidx28.i7.4", label_forbody_i11_preheader); + LoadInst* float_tmp29_i_4 = new LoadInst(ptr_arrayidx28_i7_4, "tmp29.i.4", false, label_forbody_i11_preheader); + InsertElementInst* packed_tmp31_i_4 = new InsertElementInst(packed_tmp23_i_4, float_tmp29_i_4, const_int32_30, "tmp31.i.4", label_forbody_i11_preheader); + std::vector ptr_arrayidx34_i8_4_indices; + ptr_arrayidx34_i8_4_indices.push_back(const_int32_26); + ptr_arrayidx34_i8_4_indices.push_back(const_int32_33); + Instruction* ptr_arrayidx34_i8_4 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_4_indices.begin(), ptr_arrayidx34_i8_4_indices.end(), "arrayidx34.i8.4", label_forbody_i11_preheader); + new StoreInst(packed_tmp31_i_4, ptr_arrayidx34_i8_4, false, label_forbody_i11_preheader); + new BranchInst(label_forbody_preheader, label_afterfor_82, int1_cmp_i, label_forbody_i11_preheader); + + // Block forbody.preheader (label_forbody_preheader) + new BranchInst(label_forbody_us, label_forbody_81, int1_cmp5_i, label_forbody_preheader); + + // Block forbody.us (label_forbody_us) + Argument* fwdref_99 = new Argument(IntegerType::get(32)); + PHINode* int32_i_012_0_us = new PHINode(IntegerType::get(32), "i.012.0.us", label_forbody_us); + int32_i_012_0_us->reserveOperandSpace(2); + int32_i_012_0_us->addIncoming(const_int32_26, label_forbody_preheader); + int32_i_012_0_us->addIncoming(fwdref_99, label_to_array_exit_us); + + std::vector ptr_arraydecay9_us_indices; + ptr_arraydecay9_us_indices.push_back(const_int32_26); + ptr_arraydecay9_us_indices.push_back(int32_i_012_0_us); + ptr_arraydecay9_us_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay9_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay9_us_indices.begin(), ptr_arraydecay9_us_indices.end(), "arraydecay9.us", label_forbody_us); + std::vector ptr_arraydecay13_us_indices; + ptr_arraydecay13_us_indices.push_back(const_int32_26); + ptr_arraydecay13_us_indices.push_back(int32_i_012_0_us); + ptr_arraydecay13_us_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay13_us = new GetElementPtrInst(ptr_results, ptr_arraydecay13_us_indices.begin(), ptr_arraydecay13_us_indices.end(), "arraydecay13.us", label_forbody_us); + std::vector void_100_params; + void_100_params.push_back(ptr_arraydecay13_us); + void_100_params.push_back(ptr_arraydecay9_us); + void_100_params.push_back(ptr_arrayidx34_i8); + CallInst* void_100 = new CallInst(func_execute_shader, void_100_params.begin(), void_100_params.end(), "", label_forbody_us); + void_100->setCallingConv(CallingConv::C); + void_100->setTailCall(false); + new BranchInst(label_forbody_i_us, label_forbody_us); + + // Block to_array.exit.us (label_to_array_exit_us) + BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_012_0_us, const_int32_28, "inc.us", label_to_array_exit_us); + ICmpInst* int1_cmp18_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_count_78, "cmp18.us", label_to_array_exit_us); + new BranchInst(label_forbody_us, label_afterfor_82, int1_cmp18_us, label_to_array_exit_us); + + // Block forbody.i.us (label_forbody_i_us) + Argument* fwdref_103 = new Argument(IntegerType::get(32)); + PHINode* int32_i_01_0_i_us = new PHINode(IntegerType::get(32), "i.01.0.i.us", label_forbody_i_us); + int32_i_01_0_i_us->reserveOperandSpace(2); + int32_i_01_0_i_us->addIncoming(const_int32_26, label_forbody_us); + int32_i_01_0_i_us->addIncoming(fwdref_103, label_forbody_i_us); + + std::vector ptr_arraydecay_i_us_indices; + ptr_arraydecay_i_us_indices.push_back(int32_i_012_0_us); + ptr_arraydecay_i_us_indices.push_back(int32_i_01_0_i_us); + ptr_arraydecay_i_us_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_77, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); + std::vector ptr_arrayidx6_i_us_indices; + ptr_arrayidx6_i_us_indices.push_back(const_int32_26); + ptr_arrayidx6_i_us_indices.push_back(int32_i_012_0_us); + ptr_arrayidx6_i_us_indices.push_back(int32_i_01_0_i_us); + Instruction* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_results, ptr_arrayidx6_i_us_indices.begin(), ptr_arrayidx6_i_us_indices.end(), "arrayidx6.i.us", label_forbody_i_us); + LoadInst* packed_tmp7_i_us = new LoadInst(ptr_arrayidx6_i_us, "tmp7.i.us", false, label_forbody_i_us); + ExtractElementInst* float_tmp9_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_26, "tmp9.i.us", label_forbody_i_us); + CastInst* double_conv_i_us = new FPExtInst(float_tmp9_i_us, Type::DoubleTy, "conv.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_28, "tmp11.i.us", label_forbody_i_us); + CastInst* double_conv12_i_us = new FPExtInst(float_tmp11_i_us, Type::DoubleTy, "conv12.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp14_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_29, "tmp14.i.us", label_forbody_i_us); + CastInst* double_conv15_i_us = new FPExtInst(float_tmp14_i_us, Type::DoubleTy, "conv15.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp17_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_30, "tmp17.i.us", label_forbody_i_us); + CastInst* double_conv18_i_us = new FPExtInst(float_tmp17_i_us, Type::DoubleTy, "conv18.i.us", label_forbody_i_us); + std::vector int32_call_i_us_params; + int32_call_i_us_params.push_back(const_ptr_31); + int32_call_i_us_params.push_back(double_conv_i_us); + int32_call_i_us_params.push_back(double_conv12_i_us); + int32_call_i_us_params.push_back(double_conv15_i_us); + int32_call_i_us_params.push_back(double_conv18_i_us); + CallInst* int32_call_i_us = new CallInst(func_printf, int32_call_i_us_params.begin(), int32_call_i_us_params.end(), "call.i.us", label_forbody_i_us); + int32_call_i_us->setCallingConv(CallingConv::C); + int32_call_i_us->setTailCall(false); + new StoreInst(float_tmp9_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); + std::vector ptr_arrayidx24_i_us_indices; + ptr_arrayidx24_i_us_indices.push_back(int32_i_012_0_us); + ptr_arrayidx24_i_us_indices.push_back(int32_i_01_0_i_us); + ptr_arrayidx24_i_us_indices.push_back(const_int32_28); + Instruction* ptr_arrayidx24_i_us = new GetElementPtrInst(ptr_dests_77, ptr_arrayidx24_i_us_indices.begin(), ptr_arrayidx24_i_us_indices.end(), "arrayidx24.i.us", label_forbody_i_us); + new StoreInst(float_tmp11_i_us, ptr_arrayidx24_i_us, false, label_forbody_i_us); + std::vector ptr_arrayidx28_i_us_indices; + ptr_arrayidx28_i_us_indices.push_back(int32_i_012_0_us); + ptr_arrayidx28_i_us_indices.push_back(int32_i_01_0_i_us); + ptr_arrayidx28_i_us_indices.push_back(const_int32_29); + Instruction* ptr_arrayidx28_i_us = new GetElementPtrInst(ptr_dests_77, ptr_arrayidx28_i_us_indices.begin(), ptr_arrayidx28_i_us_indices.end(), "arrayidx28.i.us", label_forbody_i_us); + new StoreInst(float_tmp14_i_us, ptr_arrayidx28_i_us, false, label_forbody_i_us); + std::vector ptr_arrayidx32_i_us_indices; + ptr_arrayidx32_i_us_indices.push_back(int32_i_012_0_us); + ptr_arrayidx32_i_us_indices.push_back(int32_i_01_0_i_us); + ptr_arrayidx32_i_us_indices.push_back(const_int32_30); + Instruction* ptr_arrayidx32_i_us = new GetElementPtrInst(ptr_dests_77, ptr_arrayidx32_i_us_indices.begin(), ptr_arrayidx32_i_us_indices.end(), "arrayidx32.i.us", label_forbody_i_us); + new StoreInst(float_tmp17_i_us, ptr_arrayidx32_i_us, false, label_forbody_i_us); + BinaryOperator* int32_indvar_next_108 = BinaryOperator::create(Instruction::Add, int32_i_01_0_i_us, const_int32_28, "indvar.next", label_forbody_i_us); + ICmpInst* int1_exitcond_109 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_108, int32_num_attribs_79, "exitcond", label_forbody_i_us); + new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_109, label_forbody_i_us); + + // Block forbody (label_forbody_81) + Argument* fwdref_111 = new Argument(IntegerType::get(32)); + PHINode* int32_i_012_0 = new PHINode(IntegerType::get(32), "i.012.0", label_forbody_81); + int32_i_012_0->reserveOperandSpace(2); + int32_i_012_0->addIncoming(const_int32_26, label_forbody_preheader); + int32_i_012_0->addIncoming(fwdref_111, label_forbody_81); + + std::vector ptr_arraydecay9_indices; + ptr_arraydecay9_indices.push_back(const_int32_26); + ptr_arraydecay9_indices.push_back(int32_i_012_0); + ptr_arraydecay9_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay9 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay9_indices.begin(), ptr_arraydecay9_indices.end(), "arraydecay9", label_forbody_81); + std::vector ptr_arraydecay13_indices; + ptr_arraydecay13_indices.push_back(const_int32_26); + ptr_arraydecay13_indices.push_back(int32_i_012_0); + ptr_arraydecay13_indices.push_back(const_int32_26); + Instruction* ptr_arraydecay13 = new GetElementPtrInst(ptr_results, ptr_arraydecay13_indices.begin(), ptr_arraydecay13_indices.end(), "arraydecay13", label_forbody_81); + std::vector void_112_params; + void_112_params.push_back(ptr_arraydecay13); + void_112_params.push_back(ptr_arraydecay9); + void_112_params.push_back(ptr_arrayidx34_i8); + CallInst* void_112 = new CallInst(func_execute_shader, void_112_params.begin(), void_112_params.end(), "", label_forbody_81); + void_112->setCallingConv(CallingConv::C); + void_112->setTailCall(false); + BinaryOperator* int32_inc_113 = BinaryOperator::create(Instruction::Add, int32_i_012_0, const_int32_28, "inc", label_forbody_81); + ICmpInst* int1_cmp18 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_113, int32_count_78, "cmp18", label_forbody_81); + new BranchInst(label_forbody_81, label_afterfor_82, int1_cmp18, label_forbody_81); + + // Block afterfor (label_afterfor_82) + new ReturnInst(label_afterfor_82); + + // Resolve Forward References + fwdref_86->replaceAllUsesWith(int32_i_013_0_i_ph); delete fwdref_86; + fwdref_88->replaceAllUsesWith(packed_tmp48_i); delete fwdref_88; + fwdref_87->replaceAllUsesWith(int32_inc_i); delete fwdref_87; + fwdref_85->replaceAllUsesWith(int32_inc59_i); delete fwdref_85; + fwdref_99->replaceAllUsesWith(int32_inc_us); delete fwdref_99; + fwdref_103->replaceAllUsesWith(int32_indvar_next_108); delete fwdref_103; + fwdref_111->replaceAllUsesWith(int32_inc_113); delete fwdref_111; + + } + + return mod; +} diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 44caa59d52..b2f6d17e96 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -257,7 +257,9 @@ translate_instruction(llvm::Module *module, out = instr->ex2(inputs[0]); } break; - case TGSI_OPCODE_LOGBASE2: + case TGSI_OPCODE_LOGBASE2: { + out = instr->lg2(inputs[0]); + } break; case TGSI_OPCODE_POWER: { out = instr->pow(inputs[0], inputs[1]); -- cgit v1.2.3 From e9a623d6a69718e3a9cc46dbb54cb4e7bd79f09c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 18 Oct 2007 08:12:48 -0400 Subject: implement min and max --- src/mesa/pipe/llvm/instructions.cpp | 102 ++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 2 + src/mesa/pipe/llvm/llvmtgsi.cpp | 8 ++- 3 files changed, 110 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index d43a617d35..2dd1d1861c 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -413,3 +413,105 @@ llvm::Value * Instructions::lg2(llvm::Value *in) callFLog(z), callFLog(w)), const_vec); } +llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) +{ + ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), + name("x1"), + m_block); + ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), + name("y1"), + m_block); + ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), + name("z1"), + m_block); + ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), + name("w1"), + m_block); + + ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), + name("x2"), + m_block); + ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), + name("y2"), + m_block); + ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), + name("z2"), + m_block); + ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), + name("w2"), + m_block); + + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OLT, x1, x2, + name("xcmp"), m_block); + SelectInst *selx = new SelectInst(xcmp, x1, x2, + name("selx"), m_block); + + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OLT, y1, y2, + name("ycmp"), m_block); + SelectInst *sely = new SelectInst(ycmp, y1, y2, + name("sely"), m_block); + + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OLT, z1, z2, + name("zcmp"), m_block); + SelectInst *selz = new SelectInst(zcmp, z1, z2, + name("selz"), m_block); + + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OLT, w1, w2, + name("wcmp"), m_block); + SelectInst *selw = new SelectInst(wcmp, w1, w2, + name("selw"), m_block); + + return vectorFromVals(selx, sely, selz, selw); +} + +llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) +{ + ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), + name("x1"), + m_block); + ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), + name("y1"), + m_block); + ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), + name("z1"), + m_block); + ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), + name("w1"), + m_block); + + ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), + name("x2"), + m_block); + ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), + name("y2"), + m_block); + ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), + name("z2"), + m_block); + ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), + name("w2"), + m_block); + + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGT, x1, x2, + name("xcmp"), m_block); + SelectInst *selx = new SelectInst(xcmp, x1, x2, + name("selx"), m_block); + + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGT, y1, y2, + name("ycmp"), m_block); + SelectInst *sely = new SelectInst(ycmp, y1, y2, + name("sely"), m_block); + + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGT, z1, z2, + name("zcmp"), m_block); + SelectInst *selz = new SelectInst(zcmp, z1, z2, + name("selz"), m_block); + + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGT, w1, w2, + name("wcmp"), m_block); + SelectInst *selw = new SelectInst(wcmp, w1, w2, + name("selw"), m_block); + + return vectorFromVals(selx, sely, selz, selw); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index e6ad47ebd2..dd1e76728a 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -26,6 +26,8 @@ public: llvm::Value *lg2(llvm::Value *in); llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, llvm::Value *in2); + llvm::Value *min(llvm::Value *in1, llvm::Value *in2); + llvm::Value *max(llvm::Value *in1, llvm::Value *in2); llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); llvm::Value *rcp(llvm::Value *in); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index b2f6d17e96..9b4a5f024e 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -213,9 +213,13 @@ translate_instruction(llvm::Module *module, out = instr->dst(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_MIN: + case TGSI_OPCODE_MIN: { + out = instr->min(inputs[0], inputs[1]); + } break; - case TGSI_OPCODE_MAX: + case TGSI_OPCODE_MAX: { + out = instr->max(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_SLT: break; -- cgit v1.2.3 From d4d8d7c468c8ba45e302e163dd87b4e45426e1da Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sat, 20 Oct 2007 16:55:23 -0400 Subject: Muchos fixos. gears kinda works. and cases don't crash. glorious --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 3 +- src/mesa/pipe/llvm/instructions.cpp | 397 ++++++++- src/mesa/pipe/llvm/instructions.h | 10 +- src/mesa/pipe/llvm/llvm_base_shader.cpp | 1179 ++++++++++++++------------ src/mesa/pipe/llvm/llvm_builtins.c | 25 +- src/mesa/pipe/llvm/llvmtgsi.cpp | 217 +---- src/mesa/pipe/llvm/llvmtgsi.h | 4 +- src/mesa/pipe/llvm/storage.cpp | 78 +- src/mesa/pipe/llvm/storage.h | 14 +- 9 files changed, 1149 insertions(+), 778 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 2ff35ace24..d658544f30 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -141,7 +141,8 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) /* batch execute the shaders on all the vertices */ ga_llvm_prog_exec(prog, inputs, outputs, consts, draw->vs.queue_nr, - draw->vertex_shader->state->num_inputs); + draw->vertex_shader->state->num_inputs, + draw->vertex_info.num_attribs); /* FIXME: finish conversion */ /* dests = outputs */ diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 2dd1d1861c..3105c49bcc 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -3,13 +3,17 @@ #include #include #include +#include #include #include using namespace llvm; -Instructions::Instructions(llvm::Module *mod, llvm::BasicBlock *block) - : m_mod(mod), m_block(block), m_idx(0) + +Function* makeLitFunction(Module *mod); + +Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block) + : m_mod(mod), m_func(func), m_block(block), m_idx(0) { m_floatVecType = VectorType::get(Type::FloatTy, 4); @@ -18,6 +22,8 @@ Instructions::Instructions(llvm::Module *mod, llvm::BasicBlock *block) m_llvmPow = 0; m_llvmFloor = 0; m_llvmFlog = 0; + m_llvmLit = 0; + m_fmtPtr = 0; } llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) @@ -128,7 +134,7 @@ llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, m_block); if (w) res = new InsertElementInst(res, w, unsigned(3), - name("vecxyw"), + name("vecxyzw"), m_block); return res; } @@ -159,9 +165,36 @@ llvm::Value *Instructions::callFAbs(llvm::Value *val) return call; } -llvm::Value * Instructions::lit(llvm::Value *in1) +llvm::Value * Instructions::lit(llvm::Value *in) { - return in1; +#if 1 + printVector(in); + return in; + + ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), + name("x"), + m_block); + + ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), + name("y"), + m_block); + + ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), + name("w"), + m_block); + return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), + ConstantFP::get(Type::FloatTy, APFloat(0.f)), + ConstantFP::get(Type::FloatTy, APFloat(0.f)), + ConstantFP::get(Type::FloatTy, APFloat(1.f))); +#else + if (!m_llvmLit) { + m_llvmLit = makeLitFunction(m_mod); + } + CallInst *call = new CallInst(m_llvmLit, in, name("litres"), m_block); + //call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +#endif } llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) @@ -515,3 +548,357 @@ llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) return vectorFromVals(selx, sely, selz, selw); } + +/* + Generated from: +extern float exp2f(float x); +extern float log2f(float x); + +float4 lit(float4 tmp) +{ + tmp.w = (tmp.w < -128.0) ? -128.0f : ((tmp.w > 128.0f) ? 128.f : tmp.w); + float4 result; + result.x = 1.0; + result.y = tmp.x; + result.z = (tmp.x > 0) ? exp2f(tmp.w * log2f(tmp.y)) : 0.0; + result.w = 1.0; + return result; +} +with: +clang --emit-llvm lit.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-function -for=lit + +*/ +Function* makeLitFunction(Module *mod) { + +// Type Definitions +ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 27); + +PointerType* PointerTy_1 = PointerType::get(ArrayTy_0); + +ArrayType* ArrayTy_2 = ArrayType::get(IntegerType::get(8), 28); + +PointerType* PointerTy_3 = PointerType::get(ArrayTy_2); + +ArrayType* ArrayTy_4 = ArrayType::get(IntegerType::get(8), 8); + +PointerType* PointerTy_5 = PointerType::get(ArrayTy_4); + +ArrayType* ArrayTy_6 = ArrayType::get(IntegerType::get(8), 33); + +PointerType* PointerTy_7 = PointerType::get(ArrayTy_6); + +std::vectorFuncTy_8_args; +FuncTy_8_args.push_back(Type::FloatTy); +FuncTy_8_args.push_back(Type::FloatTy); +ParamAttrsList *FuncTy_8_PAL = 0; +FunctionType* FuncTy_8 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_8_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_8_PAL); + +std::vectorFuncTy_10_args; +ParamAttrsList *FuncTy_10_PAL = 0; +FunctionType* FuncTy_10 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_10_args, + /*isVarArg=*/true, + /*ParamAttrs=*/FuncTy_10_PAL); + +PointerType* PointerTy_9 = PointerType::get(FuncTy_10); + +PointerType* PointerTy_11 = PointerType::get(IntegerType::get(8)); + +PointerType* PointerTy_12 = PointerType::get(FuncTy_8); + +VectorType* VectorTy_13 = VectorType::get(Type::FloatTy, 4); + +std::vectorFuncTy_14_args; +FuncTy_14_args.push_back(VectorTy_13); +ParamAttrsList *FuncTy_14_PAL = 0; +FunctionType* FuncTy_14 = FunctionType::get( + /*Result=*/VectorTy_13, + /*Params=*/FuncTy_14_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_14_PAL); + + +// Function Declarations + +Function* func_approx = new Function( + /*Type=*/FuncTy_8, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"approx", mod); +func_approx->setCallingConv(CallingConv::C); + +Function* func_printf = mod->getFunction("printf"); + +Function* func_powf = new Function( + /*Type=*/FuncTy_8, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"powf", mod); // (external, no body) +func_powf->setCallingConv(CallingConv::C); + +Function* func_lit = new Function( + /*Type=*/FuncTy_14, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"lit", mod); +func_lit->setCallingConv(CallingConv::C); + +// Global Variable Declarations + + +GlobalVariable* gvar_array__str = new GlobalVariable( +/*Type=*/ArrayTy_0, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str", +mod); + +GlobalVariable* gvar_array__str1 = new GlobalVariable( +/*Type=*/ArrayTy_2, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str1", +mod); + +GlobalVariable* gvar_array__str2 = new GlobalVariable( +/*Type=*/ArrayTy_4, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str2", +mod); + +GlobalVariable* gvar_array__str3 = new GlobalVariable( +/*Type=*/ArrayTy_6, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str3", +mod); + +// Constant Definitions +Constant* const_array_15 = ConstantArray::get("After test with '%f' '%f'\x0A", true); +Constant* const_array_16 = ConstantArray::get("Calling pow with '%f' '%f'\x0A", true); +Constant* const_array_17 = ConstantArray::get("IN LIT\x0A", true); +Constant* const_array_18 = ConstantArray::get("About to approx with '%f' '%f'\x0A", true); +ConstantFP* const_float_19 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); +ConstantFP* const_float_20 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); +std::vector const_ptr_21_indices; +Constant* const_int32_22 = Constant::getNullValue(IntegerType::get(32)); +const_ptr_21_indices.push_back(const_int32_22); +const_ptr_21_indices.push_back(const_int32_22); +Constant* const_ptr_21 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_21_indices[0], const_ptr_21_indices.size() ); +Constant* const_float_23 = Constant::getNullValue(Type::FloatTy); +std::vector const_ptr_24_indices; +const_ptr_24_indices.push_back(const_int32_22); +const_ptr_24_indices.push_back(const_int32_22); +Constant* const_ptr_24 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_24_indices[0], const_ptr_24_indices.size() ); +std::vector const_ptr_25_indices; +const_ptr_25_indices.push_back(const_int32_22); +const_ptr_25_indices.push_back(const_int32_22); +Constant* const_ptr_25 = ConstantExpr::getGetElementPtr(gvar_array__str2, &const_ptr_25_indices[0], const_ptr_25_indices.size() ); +std::vector const_packed_26_elems; +ConstantFP* const_float_27 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); +const_packed_26_elems.push_back(const_float_27); +UndefValue* const_float_28 = UndefValue::get(Type::FloatTy); +const_packed_26_elems.push_back(const_float_28); +const_packed_26_elems.push_back(const_float_28); +const_packed_26_elems.push_back(const_float_27); +Constant* const_packed_26 = ConstantVector::get(VectorTy_13, const_packed_26_elems); +ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "1", 10)); +ConstantInt* const_int32_30 = ConstantInt::get(APInt(32, "3", 10)); +std::vector const_ptr_31_indices; +const_ptr_31_indices.push_back(const_int32_22); +const_ptr_31_indices.push_back(const_int32_22); +Constant* const_ptr_31 = ConstantExpr::getGetElementPtr(gvar_array__str3, &const_ptr_31_indices[0], const_ptr_31_indices.size() ); +ConstantInt* const_int32_32 = ConstantInt::get(APInt(32, "2", 10)); +std::vector const_packed_33_elems; +const_packed_33_elems.push_back(const_float_27); +const_packed_33_elems.push_back(const_float_23); +const_packed_33_elems.push_back(const_float_23); +const_packed_33_elems.push_back(const_float_27); +Constant* const_packed_33 = ConstantVector::get(VectorTy_13, const_packed_33_elems); + +// Global Variable Definitions +gvar_array__str->setInitializer(const_array_15); +gvar_array__str1->setInitializer(const_array_16); +gvar_array__str2->setInitializer(const_array_17); +gvar_array__str3->setInitializer(const_array_18); + +// Function Definitions + +// Function: approx (func_approx) +{ + Function::arg_iterator args = func_approx->arg_begin(); + Value* float_a = args++; + float_a->setName("a"); + Value* float_b = args++; + float_b->setName("b"); + + BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); + + // Block entry (label_entry) + FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_19, "cmp", label_entry); + SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_19, float_b, "b.addr.0", label_entry); + FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_20, "cmp3", label_entry); + SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_20, float_b_addr_0, "b.addr.1", label_entry); + CastInst* double_conv = new FPExtInst(float_a, Type::DoubleTy, "conv", label_entry); + CastInst* double_conv8 = new FPExtInst(float_b_addr_1, Type::DoubleTy, "conv8", label_entry); + std::vector int32_call_params; + int32_call_params.push_back(const_ptr_21); + int32_call_params.push_back(double_conv); + int32_call_params.push_back(double_conv8); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry); + int32_call->setCallingConv(CallingConv::C); + int32_call->setTailCall(true); + FCmpInst* int1_cmp11 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_23, "cmp11", label_entry); + SelectInst* float_a_addr_0 = new SelectInst(int1_cmp11, const_float_23, float_a, "a.addr.0", label_entry); + CastInst* double_conv16 = new FPExtInst(float_a_addr_0, Type::DoubleTy, "conv16", label_entry); + std::vector int32_call19_params; + int32_call19_params.push_back(const_ptr_24); + int32_call19_params.push_back(double_conv16); + int32_call19_params.push_back(double_conv8); + CallInst* int32_call19 = new CallInst(func_printf, int32_call19_params.begin(), int32_call19_params.end(), "call19", label_entry); + int32_call19->setCallingConv(CallingConv::C); + int32_call19->setTailCall(true); + std::vector float_call22_params; + float_call22_params.push_back(float_a_addr_0); + float_call22_params.push_back(float_b_addr_1); + CallInst* float_call22 = new CallInst(func_powf, float_call22_params.begin(), float_call22_params.end(), "call22", label_entry); + float_call22->setCallingConv(CallingConv::C); + float_call22->setTailCall(true); + new ReturnInst(float_call22, label_entry); + +} + +// Function: lit (func_lit) +{ + Function::arg_iterator args = func_lit->arg_begin(); + Value* packed_tmp = args++; + packed_tmp->setName("tmp"); + + BasicBlock* label_entry_35 = new BasicBlock("entry",func_lit,0); + BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); + BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); + + // Block entry (label_entry_35) + CallInst* int32_call_36 = new CallInst(func_printf, const_ptr_25, "call", label_entry_35); + int32_call_36->setCallingConv(CallingConv::C); + int32_call_36->setTailCall(true); + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_22, "tmp7", label_entry_35); + FCmpInst* int1_cmp_37 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_23, "cmp", label_entry_35); + new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_37, label_entry_35); + + // Block ifthen (label_ifthen) + InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_26, float_tmp7, const_int32_29, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_29, "tmp14", label_ifthen); + CastInst* double_conv15 = new FPExtInst(float_tmp14, Type::DoubleTy, "conv15", label_ifthen); + ExtractElementInst* float_tmp17 = new ExtractElementInst(packed_tmp, const_int32_30, "tmp17", label_ifthen); + CastInst* double_conv18 = new FPExtInst(float_tmp17, Type::DoubleTy, "conv18", label_ifthen); + std::vector int32_call19_39_params; + int32_call19_39_params.push_back(const_ptr_31); + int32_call19_39_params.push_back(double_conv15); + int32_call19_39_params.push_back(double_conv18); + CallInst* int32_call19_39 = new CallInst(func_printf, int32_call19_39_params.begin(), int32_call19_39_params.end(), "call19", label_ifthen); + int32_call19_39->setCallingConv(CallingConv::C); + int32_call19_39->setTailCall(true); + FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp17, const_float_19, "cmp.i", label_ifthen); + SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_19, float_tmp17, "b.addr.0.i", label_ifthen); + FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_20, "cmp3.i", label_ifthen); + SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_20, float_b_addr_0_i, "b.addr.1.i", label_ifthen); + CastInst* double_conv8_i = new FPExtInst(float_b_addr_1_i, Type::DoubleTy, "conv8.i", label_ifthen); + std::vector int32_call_i_params; + int32_call_i_params.push_back(const_ptr_21); + int32_call_i_params.push_back(double_conv15); + int32_call_i_params.push_back(double_conv8_i); + CallInst* int32_call_i = new CallInst(func_printf, int32_call_i_params.begin(), int32_call_i_params.end(), "call.i", label_ifthen); + int32_call_i->setCallingConv(CallingConv::C); + int32_call_i->setTailCall(true); + FCmpInst* int1_cmp11_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_23, "cmp11.i", label_ifthen); + SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp11_i, const_float_23, float_tmp14, "a.addr.0.i", label_ifthen); + CastInst* double_conv16_i = new FPExtInst(float_a_addr_0_i, Type::DoubleTy, "conv16.i", label_ifthen); + std::vector int32_call19_i_params; + int32_call19_i_params.push_back(const_ptr_24); + int32_call19_i_params.push_back(double_conv16_i); + int32_call19_i_params.push_back(double_conv8_i); + CallInst* int32_call19_i = new CallInst(func_printf, int32_call19_i_params.begin(), int32_call19_i_params.end(), "call19.i", label_ifthen); + int32_call19_i->setCallingConv(CallingConv::C); + int32_call19_i->setTailCall(true); + std::vector float_call22_i_params; + float_call22_i_params.push_back(float_a_addr_0_i); + float_call22_i_params.push_back(float_b_addr_1_i); + CallInst* float_call22_i = new CallInst(func_powf, float_call22_i_params.begin(), float_call22_i_params.end(), "call22.i", label_ifthen); + float_call22_i->setCallingConv(CallingConv::C); + float_call22_i->setTailCall(true); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp12, float_call22_i, const_int32_32, "tmp26", label_ifthen); + new ReturnInst(packed_tmp26, label_ifthen); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock) + new ReturnInst(const_packed_33, label_UnifiedReturnBlock); + +} + +return func_lit; + +} + +void Instructions::printVector(llvm::Value *val) +{ + static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A"; + + if (!m_fmtPtr) { + Constant *format = ConstantArray::get(frmt, true); + ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1); + GlobalVariable* globalFormat = new GlobalVariable( + /*Type=*/arrayTy, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name(".str"), + m_mod); + globalFormat->setInitializer(format); + + Constant* const_int0 = Constant::getNullValue(IntegerType::get(32)); + std::vector const_ptr_21_indices; + const_ptr_21_indices.push_back(const_int0); + const_ptr_21_indices.push_back(const_int0); + m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat, + &const_ptr_21_indices[0], const_ptr_21_indices.size()); + } + + Function *func_printf = m_mod->getFunction("printf"); + ExtractElementInst *x = new ExtractElementInst(val, unsigned(0), + name("extractx"), + m_block); + ExtractElementInst *y = new ExtractElementInst(val, unsigned(1), + name("extracty"), + m_block); + ExtractElementInst *z = new ExtractElementInst(val, unsigned(2), + name("extractz"), + m_block); + ExtractElementInst *w = new ExtractElementInst(val, unsigned(3), + name("extractw"), + m_block); + CastInst *dx = new FPExtInst(x, Type::DoubleTy, name("dx"), m_block); + CastInst *dy = new FPExtInst(y, Type::DoubleTy, name("dy"), m_block); + CastInst *dz = new FPExtInst(z, Type::DoubleTy, name("dz"), m_block); + CastInst *dw = new FPExtInst(w, Type::DoubleTy, name("dw"), m_block); + std::vector params; + params.push_back(m_fmtPtr); + params.push_back(dx); + params.push_back(dy); + params.push_back(dz); + params.push_back(dw); + CallInst* call = new CallInst(func_printf, params.begin(), params.end(), "printf", m_block); + call->setCallingConv(CallingConv::C); + call->setTailCall(true); +} + +llvm::Value * Instructions::swizzleWrite(llvm::Value *in, int mask) +{ + +} diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index dd1e76728a..bb71cc8520 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -7,12 +7,13 @@ namespace llvm { class VectorType; + class Function; } class Instructions { public: - Instructions(llvm::Module *mod, llvm::BasicBlock *block); + Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block); llvm::Value *add(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); @@ -33,6 +34,9 @@ public: llvm::Value *rcp(llvm::Value *in); llvm::Value *rsq(llvm::Value *in); llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); + llvm::Value *swizzleWrite(llvm::Value *in, int mask); + + void printVector(llvm::Value *val); private: const char *name(const char *prefix); @@ -46,6 +50,7 @@ private: llvm::Value *z, llvm::Value *w=0); private: llvm::Module *m_mod; + llvm::Function *m_func; char m_name[32]; llvm::BasicBlock *m_block; int m_idx; @@ -57,6 +62,9 @@ private: llvm::Function *m_llvmPow; llvm::Function *m_llvmFloor; llvm::Function *m_llvmFlog; + llvm::Function *m_llvmLit; + + llvm::Constant *m_fmtPtr; }; #endif diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index 58dd57f6b9..f1854b5c37 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -6,59 +6,47 @@ Module* createBaseShader() { Module* mod = new Module("Shader"); // Type Definitions - ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 20); - - ArrayType* ArrayTy_2 = ArrayType::get(IntegerType::get(8), 31); + ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 24); - std::vectorFuncTy_4_args; - VectorType* VectorTy_7 = VectorType::get(Type::FloatTy, 4); + PointerType* PointerTy_1 = PointerType::get(ArrayTy_0); - ArrayType* ArrayTy_6 = ArrayType::get(VectorTy_7, 16); + ArrayType* ArrayTy_2 = ArrayType::get(IntegerType::get(8), 20); - PointerType* PointerTy_5 = PointerType::get(ArrayTy_6); + PointerType* PointerTy_3 = PointerType::get(ArrayTy_2); - FuncTy_4_args.push_back(PointerTy_5); - ArrayType* ArrayTy_10 = ArrayType::get(Type::FloatTy, 4); + ArrayType* ArrayTy_4 = ArrayType::get(IntegerType::get(8), 68); - ArrayType* ArrayTy_9 = ArrayType::get(ArrayTy_10, 16); + PointerType* PointerTy_5 = PointerType::get(ArrayTy_4); - PointerType* PointerTy_8 = PointerType::get(ArrayTy_9); + ArrayType* ArrayTy_6 = ArrayType::get(IntegerType::get(8), 33); - FuncTy_4_args.push_back(PointerTy_8); - FuncTy_4_args.push_back(IntegerType::get(32)); - FuncTy_4_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_4_PAL = 0; - FunctionType* FuncTy_4 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_4_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_4_PAL); + PointerType* PointerTy_7 = PointerType::get(ArrayTy_6); - PointerType* PointerTy_12 = PointerType::get(VectorTy_7); + std::vectorFuncTy_8_args; + VectorType* VectorTy_11 = VectorType::get(Type::FloatTy, 4); - std::vectorFuncTy_13_args; - FuncTy_13_args.push_back(PointerTy_12); - PointerType* PointerTy_14 = PointerType::get(ArrayTy_10); + ArrayType* ArrayTy_10 = ArrayType::get(VectorTy_11, 16); - FuncTy_13_args.push_back(PointerTy_14); - FuncTy_13_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_13_PAL = 0; - FunctionType* FuncTy_13 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_13_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_13_PAL); - - std::vectorFuncTy_15_args; - FuncTy_15_args.push_back(PointerTy_14); - FuncTy_15_args.push_back(PointerTy_12); - FuncTy_15_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_15_PAL = 0; - FunctionType* FuncTy_15 = FunctionType::get( + PointerType* PointerTy_9 = PointerType::get(ArrayTy_10); + + FuncTy_8_args.push_back(PointerTy_9); + ArrayType* ArrayTy_14 = ArrayType::get(Type::FloatTy, 4); + + ArrayType* ArrayTy_13 = ArrayType::get(ArrayTy_14, 16); + + PointerType* PointerTy_12 = PointerType::get(ArrayTy_13); + + FuncTy_8_args.push_back(PointerTy_12); + FuncTy_8_args.push_back(IntegerType::get(32)); + FuncTy_8_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_8_PAL = 0; + FunctionType* FuncTy_8 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_15_args, + /*Params=*/FuncTy_8_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_15_PAL); + /*ParamAttrs=*/FuncTy_8_PAL); + + PointerType* PointerTy_15 = PointerType::get(Type::FloatTy); std::vectorFuncTy_17_args; ParamAttrsList *FuncTy_17_PAL = 0; @@ -68,25 +56,44 @@ Module* createBaseShader() { /*isVarArg=*/true, /*ParamAttrs=*/FuncTy_17_PAL); - std::vectorFuncTy_19_args; - FuncTy_19_args.push_back(PointerTy_8); - FuncTy_19_args.push_back(PointerTy_8); - FuncTy_19_args.push_back(PointerTy_14); - FuncTy_19_args.push_back(IntegerType::get(32)); - FuncTy_19_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_19_PAL = 0; - FunctionType* FuncTy_19 = FunctionType::get( + PointerType* PointerTy_16 = PointerType::get(FuncTy_17); + + PointerType* PointerTy_18 = PointerType::get(IntegerType::get(8)); + + PointerType* PointerTy_19 = PointerType::get(VectorTy_11); + + std::vectorFuncTy_20_args; + FuncTy_20_args.push_back(PointerTy_19); + PointerType* PointerTy_21 = PointerType::get(ArrayTy_14); + + FuncTy_20_args.push_back(PointerTy_21); + FuncTy_20_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_20_PAL = 0; + FunctionType* FuncTy_20 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_19_args, + /*Params=*/FuncTy_20_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_19_PAL); - - ArrayType* ArrayTy_21 = ArrayType::get(ArrayTy_6, 2048); + /*ParamAttrs=*/FuncTy_20_PAL); + + std::vectorFuncTy_22_args; + FuncTy_22_args.push_back(PointerTy_21); + FuncTy_22_args.push_back(PointerTy_19); + FuncTy_22_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_22_PAL = 0; + FunctionType* FuncTy_22 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_22_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_22_PAL); std::vectorFuncTy_23_args; FuncTy_23_args.push_back(PointerTy_12); FuncTy_23_args.push_back(PointerTy_12); - FuncTy_23_args.push_back(PointerTy_12); + FuncTy_23_args.push_back(PointerTy_21); + FuncTy_23_args.push_back(IntegerType::get(32)); + FuncTy_23_args.push_back(IntegerType::get(32)); + FuncTy_23_args.push_back(IntegerType::get(32)); + FuncTy_23_args.push_back(IntegerType::get(32)); ParamAttrsList *FuncTy_23_PAL = 0; FunctionType* FuncTy_23 = FunctionType::get( /*Result=*/Type::VoidTy, @@ -94,41 +101,62 @@ Module* createBaseShader() { /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_23_PAL); + ArrayType* ArrayTy_25 = ArrayType::get(ArrayTy_10, 2048); + + PointerType* PointerTy_24 = PointerType::get(ArrayTy_25); + + ArrayType* ArrayTy_27 = ArrayType::get(VectorTy_11, 32); + + PointerType* PointerTy_26 = PointerType::get(ArrayTy_27); + + std::vectorFuncTy_29_args; + FuncTy_29_args.push_back(PointerTy_19); + FuncTy_29_args.push_back(PointerTy_19); + FuncTy_29_args.push_back(PointerTy_19); + ParamAttrsList *FuncTy_29_PAL = 0; + FunctionType* FuncTy_29 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_29_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_29_PAL); + + PointerType* PointerTy_28 = PointerType::get(FuncTy_29); + // Function Declarations Function* func_from_array = new Function( - /*Type=*/FuncTy_4, + /*Type=*/FuncTy_8, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_array", mod); func_from_array->setCallingConv(CallingConv::C); + Function* func_printf = new Function( + /*Type=*/FuncTy_17, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", mod); // (external, no body) + func_printf->setCallingConv(CallingConv::C); + Function* func_from_consts = new Function( - /*Type=*/FuncTy_13, + /*Type=*/FuncTy_20, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_consts", mod); func_from_consts->setCallingConv(CallingConv::C); Function* func_to_array = new Function( - /*Type=*/FuncTy_15, + /*Type=*/FuncTy_22, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"to_array", mod); func_to_array->setCallingConv(CallingConv::C); - Function* func_printf = new Function( - /*Type=*/FuncTy_17, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", mod); // (external, no body) - func_printf->setCallingConv(CallingConv::C); - Function* func_run_vertex_shader = new Function( - /*Type=*/FuncTy_19, + /*Type=*/FuncTy_23, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_vertex_shader", mod); func_run_vertex_shader->setCallingConv(CallingConv::C); Function* func_execute_shader = new Function( - /*Type=*/FuncTy_23, + /*Type=*/FuncTy_29, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"execute_shader", mod); // (external, no body) func_execute_shader->setCallingConv(CallingConv::C); @@ -145,34 +173,89 @@ Module* createBaseShader() { mod); GlobalVariable* gvar_array__str1 = new GlobalVariable( - /*Type=*/ArrayTy_2, + /*Type=*/ArrayTy_0, /*isConstant=*/true, /*Linkage=*/GlobalValue::InternalLinkage, /*Initializer=*/0, // has initializer, specified below /*Name=*/".str1", mod); + GlobalVariable* gvar_array__str2 = new GlobalVariable( + /*Type=*/ArrayTy_2, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str2", + mod); + + GlobalVariable* gvar_array__str3 = new GlobalVariable( + /*Type=*/ArrayTy_4, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str3", + mod); + + GlobalVariable* gvar_array__str4 = new GlobalVariable( + /*Type=*/ArrayTy_6, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str4", + mod); + + GlobalVariable* gvar_array__str5 = new GlobalVariable( + /*Type=*/ArrayTy_0, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str5", + mod); + // Constant Definitions - Constant* const_array_24 = ConstantArray::get("DEST = %f %f %f %f\x0A", true); - Constant* const_array_25 = ConstantArray::get("XXXXXXXXXXX run_vertex_shader\x0A", true); - Constant* const_int32_26 = Constant::getNullValue(IntegerType::get(32)); - UndefValue* const_packed_27 = UndefValue::get(VectorTy_7); - ConstantInt* const_int32_28 = ConstantInt::get(APInt(32, "1", 10)); - ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "2", 10)); - ConstantInt* const_int32_30 = ConstantInt::get(APInt(32, "3", 10)); - std::vector const_ptr_31_indices; - const_ptr_31_indices.push_back(const_int32_26); - const_ptr_31_indices.push_back(const_int32_26); - Constant* const_ptr_31 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_31_indices[0], const_ptr_31_indices.size() ); - std::vector const_ptr_32_indices; - const_ptr_32_indices.push_back(const_int32_26); - const_ptr_32_indices.push_back(const_int32_26); - Constant* const_ptr_32 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_32_indices[0], const_ptr_32_indices.size() ); - ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "4", 10)); + Constant* const_array_30 = ConstantArray::get("FRA(%d %d) %f %f %f %f\x0A", true); + Constant* const_array_31 = ConstantArray::get("VCONST(%d) %f %f %f %f\x0A", true); + Constant* const_array_32 = ConstantArray::get("DEST = %f %f %f %f\x0A", true); + Constant* const_array_33 = ConstantArray::get("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, consts = %d\x0A", true); + Constant* const_array_34 = ConstantArray::get(" after conversion\x0A", true); + Constant* const_array_35 = ConstantArray::get("after executing shader\x0A", true); + Constant* const_int32_36 = Constant::getNullValue(IntegerType::get(32)); + UndefValue* const_packed_37 = UndefValue::get(VectorTy_11); + ConstantInt* const_int32_38 = ConstantInt::get(APInt(32, "1", 10)); + ConstantInt* const_int32_39 = ConstantInt::get(APInt(32, "2", 10)); + ConstantInt* const_int32_40 = ConstantInt::get(APInt(32, "3", 10)); + std::vector const_ptr_41_indices; + const_ptr_41_indices.push_back(const_int32_36); + const_ptr_41_indices.push_back(const_int32_36); + Constant* const_ptr_41 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_41_indices[0], const_ptr_41_indices.size() ); + std::vector const_ptr_42_indices; + const_ptr_42_indices.push_back(const_int32_36); + const_ptr_42_indices.push_back(const_int32_36); + Constant* const_ptr_42 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_42_indices[0], const_ptr_42_indices.size() ); + std::vector const_ptr_43_indices; + const_ptr_43_indices.push_back(const_int32_36); + const_ptr_43_indices.push_back(const_int32_36); + Constant* const_ptr_43 = ConstantExpr::getGetElementPtr(gvar_array__str2, &const_ptr_43_indices[0], const_ptr_43_indices.size() ); + std::vector const_ptr_44_indices; + const_ptr_44_indices.push_back(const_int32_36); + const_ptr_44_indices.push_back(const_int32_36); + Constant* const_ptr_44 = ConstantExpr::getGetElementPtr(gvar_array__str3, &const_ptr_44_indices[0], const_ptr_44_indices.size() ); + std::vector const_ptr_45_indices; + const_ptr_45_indices.push_back(const_int32_36); + const_ptr_45_indices.push_back(const_int32_36); + Constant* const_ptr_45 = ConstantExpr::getGetElementPtr(gvar_array__str4, &const_ptr_45_indices[0], const_ptr_45_indices.size() ); + std::vector const_ptr_46_indices; + const_ptr_46_indices.push_back(const_int32_36); + const_ptr_46_indices.push_back(const_int32_36); + Constant* const_ptr_46 = ConstantExpr::getGetElementPtr(gvar_array__str5, &const_ptr_46_indices[0], const_ptr_46_indices.size() ); // Global Variable Definitions - gvar_array__str->setInitializer(const_array_24); - gvar_array__str1->setInitializer(const_array_25); + gvar_array__str->setInitializer(const_array_30); + gvar_array__str1->setInitializer(const_array_31); + gvar_array__str2->setInitializer(const_array_32); + gvar_array__str3->setInitializer(const_array_33); + gvar_array__str4->setInitializer(const_array_34); + gvar_array__str5->setInitializer(const_array_35); // Function Definitions @@ -190,158 +273,187 @@ Module* createBaseShader() { BasicBlock* label_entry = new BasicBlock("entry",func_from_array,0); BasicBlock* label_forbody6 = new BasicBlock("forbody6",func_from_array,0); - BasicBlock* label_forinc57 = new BasicBlock("forinc57",func_from_array,0); - BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); + BasicBlock* label_forinc70 = new BasicBlock("forinc70",func_from_array,0); + BasicBlock* label_afterfor73 = new BasicBlock("afterfor73",func_from_array,0); // Block entry (label_entry) - ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_26, "cmp", label_entry); - ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_26, "cmp5", label_entry); + ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_36, "cmp", label_entry); + ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_36, "cmp5", label_entry); BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); - new BranchInst(label_forbody6, label_afterfor60, int1_bothcond, label_entry); + new BranchInst(label_forbody6, label_afterfor73, int1_bothcond, label_entry); // Block forbody6 (label_forbody6) - Argument* fwdref_35 = new Argument(IntegerType::get(32)); - Argument* fwdref_36 = new Argument(IntegerType::get(32)); - PHINode* int32_i_013_0 = new PHINode(IntegerType::get(32), "i.013.0", label_forbody6); - int32_i_013_0->reserveOperandSpace(3); - int32_i_013_0->addIncoming(const_int32_26, label_entry); - int32_i_013_0->addIncoming(fwdref_35, label_forinc57); - int32_i_013_0->addIncoming(fwdref_36, label_forbody6); - - Argument* fwdref_37 = new Argument(IntegerType::get(32)); + Argument* fwdref_48 = new Argument(IntegerType::get(32)); + Argument* fwdref_49 = new Argument(IntegerType::get(32)); + PHINode* int32_i_014_0 = new PHINode(IntegerType::get(32), "i.014.0", label_forbody6); + int32_i_014_0->reserveOperandSpace(3); + int32_i_014_0->addIncoming(const_int32_36, label_entry); + int32_i_014_0->addIncoming(fwdref_48, label_forinc70); + int32_i_014_0->addIncoming(fwdref_49, label_forbody6); + + Argument* fwdref_50 = new Argument(IntegerType::get(32)); PHINode* int32_j_03_0 = new PHINode(IntegerType::get(32), "j.03.0", label_forbody6); int32_j_03_0->reserveOperandSpace(3); - int32_j_03_0->addIncoming(fwdref_37, label_forbody6); - int32_j_03_0->addIncoming(const_int32_26, label_forinc57); - int32_j_03_0->addIncoming(const_int32_26, label_entry); + int32_j_03_0->addIncoming(fwdref_50, label_forbody6); + int32_j_03_0->addIncoming(const_int32_36, label_forinc70); + int32_j_03_0->addIncoming(const_int32_36, label_entry); - Argument* fwdref_38 = new Argument(VectorTy_7); - PHINode* packed_vec_01_0 = new PHINode(VectorTy_7, "vec.01.0", label_forbody6); + Argument* fwdref_51 = new Argument(VectorTy_11); + PHINode* packed_vec_01_0 = new PHINode(VectorTy_11, "vec.01.0", label_forbody6); packed_vec_01_0->reserveOperandSpace(3); - packed_vec_01_0->addIncoming(fwdref_38, label_forbody6); - packed_vec_01_0->addIncoming(const_packed_27, label_entry); - packed_vec_01_0->addIncoming(fwdref_38, label_forinc57); + packed_vec_01_0->addIncoming(fwdref_51, label_forbody6); + packed_vec_01_0->addIncoming(const_packed_37, label_entry); + packed_vec_01_0->addIncoming(fwdref_51, label_forinc70); std::vector ptr_arraydecay11_indices; - ptr_arraydecay11_indices.push_back(int32_i_013_0); + ptr_arraydecay11_indices.push_back(int32_i_014_0); ptr_arraydecay11_indices.push_back(int32_j_03_0); - ptr_arraydecay11_indices.push_back(const_int32_26); + ptr_arraydecay11_indices.push_back(const_int32_36); Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); - InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_01_0, float_tmp13, const_int32_26, "tmp15", label_forbody6); + InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_01_0, float_tmp13, const_int32_36, "tmp15", label_forbody6); std::vector ptr_arrayidx23_indices; - ptr_arrayidx23_indices.push_back(int32_i_013_0); + ptr_arrayidx23_indices.push_back(int32_i_014_0); ptr_arrayidx23_indices.push_back(int32_j_03_0); - ptr_arrayidx23_indices.push_back(const_int32_28); + ptr_arrayidx23_indices.push_back(const_int32_38); Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_28, "tmp26", label_forbody6); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_38, "tmp26", label_forbody6); std::vector ptr_arrayidx34_indices; - ptr_arrayidx34_indices.push_back(int32_i_013_0); + ptr_arrayidx34_indices.push_back(int32_i_014_0); ptr_arrayidx34_indices.push_back(int32_j_03_0); - ptr_arrayidx34_indices.push_back(const_int32_29); + ptr_arrayidx34_indices.push_back(const_int32_39); Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6); LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6); - InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_29, "tmp37", label_forbody6); + InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_39, "tmp37", label_forbody6); std::vector ptr_arrayidx45_indices; - ptr_arrayidx45_indices.push_back(int32_i_013_0); + ptr_arrayidx45_indices.push_back(int32_i_014_0); ptr_arrayidx45_indices.push_back(int32_j_03_0); - ptr_arrayidx45_indices.push_back(const_int32_30); + ptr_arrayidx45_indices.push_back(const_int32_40); Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6); LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6); - InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_30, "tmp48", label_forbody6); - std::vector ptr_arrayidx54_indices; - ptr_arrayidx54_indices.push_back(int32_i_013_0); - ptr_arrayidx54_indices.push_back(int32_j_03_0); - Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); - new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); - BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_03_0, const_int32_28, "inc", label_forbody6); - ICmpInst* int1_cmp511 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp511", label_forbody6); - new BranchInst(label_forbody6, label_forinc57, int1_cmp511, label_forbody6); - - // Block forinc57 (label_forinc57) - BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_013_0, const_int32_28, "inc59", label_forinc57); - ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59, int32_count, "cmp21", label_forinc57); - new BranchInst(label_forbody6, label_afterfor60, int1_cmp21, label_forinc57); - - // Block afterfor60 (label_afterfor60) - new ReturnInst(label_afterfor60); + InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_40, "tmp48", label_forbody6); + CastInst* double_conv = new FPExtInst(float_tmp13, Type::DoubleTy, "conv", label_forbody6); + CastInst* double_conv55 = new FPExtInst(float_tmp24, Type::DoubleTy, "conv55", label_forbody6); + CastInst* double_conv58 = new FPExtInst(float_tmp35, Type::DoubleTy, "conv58", label_forbody6); + CastInst* double_conv61 = new FPExtInst(float_tmp46, Type::DoubleTy, "conv61", label_forbody6); + std::vector int32_call_params; + int32_call_params.push_back(const_ptr_41); + int32_call_params.push_back(int32_i_014_0); + int32_call_params.push_back(int32_j_03_0); + int32_call_params.push_back(double_conv); + int32_call_params.push_back(double_conv55); + int32_call_params.push_back(double_conv58); + int32_call_params.push_back(double_conv61); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_forbody6); + int32_call->setCallingConv(CallingConv::C); + int32_call->setTailCall(true); + std::vector ptr_arrayidx67_indices; + ptr_arrayidx67_indices.push_back(int32_i_014_0); + ptr_arrayidx67_indices.push_back(int32_j_03_0); + Instruction* ptr_arrayidx67 = new GetElementPtrInst(ptr_res, ptr_arrayidx67_indices.begin(), ptr_arrayidx67_indices.end(), "arrayidx67", label_forbody6); + StoreInst* void_52 = new StoreInst(packed_tmp48, ptr_arrayidx67, false, label_forbody6); + BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_03_0, const_int32_38, "inc", label_forbody6); + ICmpInst* int1_cmp512 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp512", label_forbody6); + new BranchInst(label_forbody6, label_forinc70, int1_cmp512, label_forbody6); + + // Block forinc70 (label_forinc70) + BinaryOperator* int32_inc72 = BinaryOperator::create(Instruction::Add, int32_i_014_0, const_int32_38, "inc72", label_forinc70); + ICmpInst* int1_cmp23 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc72, int32_count, "cmp23", label_forinc70); + new BranchInst(label_forbody6, label_afterfor73, int1_cmp23, label_forinc70); + + // Block afterfor73 (label_afterfor73) + new ReturnInst(label_afterfor73); // Resolve Forward References - fwdref_36->replaceAllUsesWith(int32_i_013_0); delete fwdref_36; - fwdref_38->replaceAllUsesWith(packed_tmp48); delete fwdref_38; - fwdref_37->replaceAllUsesWith(int32_inc); delete fwdref_37; - fwdref_35->replaceAllUsesWith(int32_inc59); delete fwdref_35; + fwdref_49->replaceAllUsesWith(int32_i_014_0); delete fwdref_49; + fwdref_51->replaceAllUsesWith(packed_tmp48); delete fwdref_51; + fwdref_50->replaceAllUsesWith(int32_inc); delete fwdref_50; + fwdref_48->replaceAllUsesWith(int32_inc72); delete fwdref_48; } // Function: from_consts (func_from_consts) { Function::arg_iterator args = func_from_consts->arg_begin(); - Value* ptr_res_43 = args++; - ptr_res_43->setName("res"); - Value* ptr_ainputs_44 = args++; - ptr_ainputs_44->setName("ainputs"); - Value* int32_count_45 = args++; - int32_count_45->setName("count"); - - BasicBlock* label_entry_46 = new BasicBlock("entry",func_from_consts,0); + Value* ptr_res_56 = args++; + ptr_res_56->setName("res"); + Value* ptr_ainputs_57 = args++; + ptr_ainputs_57->setName("ainputs"); + Value* int32_count_58 = args++; + int32_count_58->setName("count"); + + BasicBlock* label_entry_59 = new BasicBlock("entry",func_from_consts,0); BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); - // Block entry (label_entry_46) - ICmpInst* int1_cmp_47 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_45, const_int32_26, "cmp", label_entry_46); - new BranchInst(label_forbody, label_afterfor, int1_cmp_47, label_entry_46); + // Block entry (label_entry_59) + ICmpInst* int1_cmp_60 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_58, const_int32_36, "cmp", label_entry_59); + new BranchInst(label_forbody, label_afterfor, int1_cmp_60, label_entry_59); // Block forbody (label_forbody) - Argument* fwdref_49 = new Argument(IntegerType::get(32)); + Argument* fwdref_62 = new Argument(IntegerType::get(32)); PHINode* int32_i_02_0 = new PHINode(IntegerType::get(32), "i.02.0", label_forbody); int32_i_02_0->reserveOperandSpace(2); - int32_i_02_0->addIncoming(const_int32_26, label_entry_46); - int32_i_02_0->addIncoming(fwdref_49, label_forbody); + int32_i_02_0->addIncoming(const_int32_36, label_entry_59); + int32_i_02_0->addIncoming(fwdref_62, label_forbody); - Argument* fwdref_51 = new Argument(VectorTy_7); - PHINode* packed_vec_01_0_50 = new PHINode(VectorTy_7, "vec.01.0", label_forbody); - packed_vec_01_0_50->reserveOperandSpace(2); - packed_vec_01_0_50->addIncoming(const_packed_27, label_entry_46); - packed_vec_01_0_50->addIncoming(fwdref_51, label_forbody); + Argument* fwdref_64 = new Argument(VectorTy_11); + PHINode* packed_vec_01_0_63 = new PHINode(VectorTy_11, "vec.01.0", label_forbody); + packed_vec_01_0_63->reserveOperandSpace(2); + packed_vec_01_0_63->addIncoming(const_packed_37, label_entry_59); + packed_vec_01_0_63->addIncoming(fwdref_64, label_forbody); std::vector ptr_arraydecay_indices; ptr_arraydecay_indices.push_back(int32_i_02_0); - ptr_arraydecay_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_44, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); + ptr_arraydecay_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_57, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); - InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_01_0_50, float_tmp5, const_int32_26, "tmp7", label_forbody); + InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_01_0_63, float_tmp5, const_int32_36, "tmp7", label_forbody); std::vector ptr_arrayidx12_indices; ptr_arrayidx12_indices.push_back(int32_i_02_0); - ptr_arrayidx12_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_44, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); - LoadInst* float_tmp13_52 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); - InsertElementInst* packed_tmp15_53 = new InsertElementInst(packed_tmp7, float_tmp13_52, const_int32_28, "tmp15", label_forbody); + ptr_arrayidx12_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_57, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); + LoadInst* float_tmp13_65 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); + InsertElementInst* packed_tmp15_66 = new InsertElementInst(packed_tmp7, float_tmp13_65, const_int32_38, "tmp15", label_forbody); std::vector ptr_arrayidx20_indices; ptr_arrayidx20_indices.push_back(int32_i_02_0); - ptr_arrayidx20_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_44, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); + ptr_arrayidx20_indices.push_back(const_int32_39); + Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_57, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); - InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_53, float_tmp21, const_int32_29, "tmp23", label_forbody); + InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_66, float_tmp21, const_int32_39, "tmp23", label_forbody); std::vector ptr_arrayidx28_indices; ptr_arrayidx28_indices.push_back(int32_i_02_0); - ptr_arrayidx28_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_44, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); + ptr_arrayidx28_indices.push_back(const_int32_40); + Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_57, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); - InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_30, "tmp31", label_forbody); - GetElementPtrInst* ptr_arrayidx34_54 = new GetElementPtrInst(ptr_res_43, int32_i_02_0, "arrayidx34", label_forbody); - new StoreInst(packed_tmp31, ptr_arrayidx34_54, false, label_forbody); - BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_02_0, const_int32_28, "indvar.next", label_forbody); - ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_45, "exitcond", label_forbody); + InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_40, "tmp31", label_forbody); + CastInst* double_conv_67 = new FPExtInst(float_tmp5, Type::DoubleTy, "conv", label_forbody); + CastInst* double_conv37 = new FPExtInst(float_tmp13_65, Type::DoubleTy, "conv37", label_forbody); + CastInst* double_conv40 = new FPExtInst(float_tmp21, Type::DoubleTy, "conv40", label_forbody); + CastInst* double_conv43 = new FPExtInst(float_tmp29, Type::DoubleTy, "conv43", label_forbody); + std::vector int32_call_68_params; + int32_call_68_params.push_back(const_ptr_42); + int32_call_68_params.push_back(int32_i_02_0); + int32_call_68_params.push_back(double_conv_67); + int32_call_68_params.push_back(double_conv37); + int32_call_68_params.push_back(double_conv40); + int32_call_68_params.push_back(double_conv43); + CallInst* int32_call_68 = new CallInst(func_printf, int32_call_68_params.begin(), int32_call_68_params.end(), "call", label_forbody); + int32_call_68->setCallingConv(CallingConv::C); + int32_call_68->setTailCall(true); + GetElementPtrInst* ptr_arrayidx46 = new GetElementPtrInst(ptr_res_56, int32_i_02_0, "arrayidx46", label_forbody); + StoreInst* void_69 = new StoreInst(packed_tmp31, ptr_arrayidx46, false, label_forbody); + BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_02_0, const_int32_38, "indvar.next", label_forbody); + ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_58, "exitcond", label_forbody); new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); // Block afterfor (label_afterfor) new ReturnInst(label_afterfor); // Resolve Forward References - fwdref_51->replaceAllUsesWith(packed_tmp31); delete fwdref_51; - fwdref_49->replaceAllUsesWith(int32_indvar_next); delete fwdref_49; + fwdref_64->replaceAllUsesWith(packed_tmp31); delete fwdref_64; + fwdref_62->replaceAllUsesWith(int32_indvar_next); delete fwdref_62; } @@ -352,457 +464,418 @@ Module* createBaseShader() { ptr_dests->setName("dests"); Value* ptr_in = args++; ptr_in->setName("in"); - Value* int32_num_attribs_58 = args++; - int32_num_attribs_58->setName("num_attribs"); + Value* int32_num_attribs_72 = args++; + int32_num_attribs_72->setName("num_attribs"); - BasicBlock* label_entry_59 = new BasicBlock("entry",func_to_array,0); - BasicBlock* label_forbody_60 = new BasicBlock("forbody",func_to_array,0); - BasicBlock* label_afterfor_61 = new BasicBlock("afterfor",func_to_array,0); + BasicBlock* label_entry_73 = new BasicBlock("entry",func_to_array,0); + BasicBlock* label_forbody_74 = new BasicBlock("forbody",func_to_array,0); + BasicBlock* label_afterfor_75 = new BasicBlock("afterfor",func_to_array,0); - // Block entry (label_entry_59) - ICmpInst* int1_cmp_62 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_58, const_int32_26, "cmp", label_entry_59); - new BranchInst(label_forbody_60, label_afterfor_61, int1_cmp_62, label_entry_59); + // Block entry (label_entry_73) + ICmpInst* int1_cmp_76 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_72, const_int32_36, "cmp", label_entry_73); + new BranchInst(label_forbody_74, label_afterfor_75, int1_cmp_76, label_entry_73); - // Block forbody (label_forbody_60) - Argument* fwdref_64 = new Argument(IntegerType::get(32)); - PHINode* int32_i_01_0 = new PHINode(IntegerType::get(32), "i.01.0", label_forbody_60); + // Block forbody (label_forbody_74) + Argument* fwdref_78 = new Argument(IntegerType::get(32)); + PHINode* int32_i_01_0 = new PHINode(IntegerType::get(32), "i.01.0", label_forbody_74); int32_i_01_0->reserveOperandSpace(2); - int32_i_01_0->addIncoming(const_int32_26, label_entry_59); - int32_i_01_0->addIncoming(fwdref_64, label_forbody_60); - - std::vector ptr_arraydecay_65_indices; - ptr_arraydecay_65_indices.push_back(int32_i_01_0); - ptr_arraydecay_65_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay_65 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_65_indices.begin(), ptr_arraydecay_65_indices.end(), "arraydecay", label_forbody_60); - GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_01_0, "arrayidx6", label_forbody_60); - LoadInst* packed_tmp7_66 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_60); - ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_tmp7_66, const_int32_26, "tmp9", label_forbody_60); - CastInst* double_conv = new FPExtInst(float_tmp9, Type::DoubleTy, "conv", label_forbody_60); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_66, const_int32_28, "tmp11", label_forbody_60); - CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_forbody_60); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp7_66, const_int32_29, "tmp14", label_forbody_60); - CastInst* double_conv15 = new FPExtInst(float_tmp14, Type::DoubleTy, "conv15", label_forbody_60); - ExtractElementInst* float_tmp17 = new ExtractElementInst(packed_tmp7_66, const_int32_30, "tmp17", label_forbody_60); - CastInst* double_conv18 = new FPExtInst(float_tmp17, Type::DoubleTy, "conv18", label_forbody_60); - std::vector int32_call_params; - int32_call_params.push_back(const_ptr_31); - int32_call_params.push_back(double_conv); - int32_call_params.push_back(double_conv12); - int32_call_params.push_back(double_conv15); - int32_call_params.push_back(double_conv18); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_forbody_60); - int32_call->setCallingConv(CallingConv::C); - int32_call->setTailCall(true); - new StoreInst(float_tmp9, ptr_arraydecay_65, false, label_forbody_60); + int32_i_01_0->addIncoming(const_int32_36, label_entry_73); + int32_i_01_0->addIncoming(fwdref_78, label_forbody_74); + + std::vector ptr_arraydecay_79_indices; + ptr_arraydecay_79_indices.push_back(int32_i_01_0); + ptr_arraydecay_79_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay_79 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_79_indices.begin(), ptr_arraydecay_79_indices.end(), "arraydecay", label_forbody_74); + GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_01_0, "arrayidx6", label_forbody_74); + LoadInst* packed_tmp7_80 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_74); + ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_tmp7_80, const_int32_36, "tmp9", label_forbody_74); + CastInst* double_conv_81 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv", label_forbody_74); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_80, const_int32_38, "tmp11", label_forbody_74); + CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_forbody_74); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp7_80, const_int32_39, "tmp14", label_forbody_74); + CastInst* double_conv15 = new FPExtInst(float_tmp14, Type::DoubleTy, "conv15", label_forbody_74); + ExtractElementInst* float_tmp17 = new ExtractElementInst(packed_tmp7_80, const_int32_40, "tmp17", label_forbody_74); + CastInst* double_conv18 = new FPExtInst(float_tmp17, Type::DoubleTy, "conv18", label_forbody_74); + std::vector int32_call_82_params; + int32_call_82_params.push_back(const_ptr_43); + int32_call_82_params.push_back(double_conv_81); + int32_call_82_params.push_back(double_conv12); + int32_call_82_params.push_back(double_conv15); + int32_call_82_params.push_back(double_conv18); + CallInst* int32_call_82 = new CallInst(func_printf, int32_call_82_params.begin(), int32_call_82_params.end(), "call", label_forbody_74); + int32_call_82->setCallingConv(CallingConv::C); + int32_call_82->setTailCall(true); + StoreInst* void_83 = new StoreInst(float_tmp9, ptr_arraydecay_79, false, label_forbody_74); std::vector ptr_arrayidx24_indices; ptr_arrayidx24_indices.push_back(int32_i_01_0); - ptr_arrayidx24_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx24 = new GetElementPtrInst(ptr_dests, ptr_arrayidx24_indices.begin(), ptr_arrayidx24_indices.end(), "arrayidx24", label_forbody_60); - new StoreInst(float_tmp11, ptr_arrayidx24, false, label_forbody_60); - std::vector ptr_arrayidx28_69_indices; - ptr_arrayidx28_69_indices.push_back(int32_i_01_0); - ptr_arrayidx28_69_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx28_69 = new GetElementPtrInst(ptr_dests, ptr_arrayidx28_69_indices.begin(), ptr_arrayidx28_69_indices.end(), "arrayidx28", label_forbody_60); - new StoreInst(float_tmp14, ptr_arrayidx28_69, false, label_forbody_60); + ptr_arrayidx24_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx24 = new GetElementPtrInst(ptr_dests, ptr_arrayidx24_indices.begin(), ptr_arrayidx24_indices.end(), "arrayidx24", label_forbody_74); + StoreInst* void_84 = new StoreInst(float_tmp11, ptr_arrayidx24, false, label_forbody_74); + std::vector ptr_arrayidx28_85_indices; + ptr_arrayidx28_85_indices.push_back(int32_i_01_0); + ptr_arrayidx28_85_indices.push_back(const_int32_39); + Instruction* ptr_arrayidx28_85 = new GetElementPtrInst(ptr_dests, ptr_arrayidx28_85_indices.begin(), ptr_arrayidx28_85_indices.end(), "arrayidx28", label_forbody_74); + StoreInst* void_86 = new StoreInst(float_tmp14, ptr_arrayidx28_85, false, label_forbody_74); std::vector ptr_arrayidx32_indices; ptr_arrayidx32_indices.push_back(int32_i_01_0); - ptr_arrayidx32_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx32 = new GetElementPtrInst(ptr_dests, ptr_arrayidx32_indices.begin(), ptr_arrayidx32_indices.end(), "arrayidx32", label_forbody_60); - new StoreInst(float_tmp17, ptr_arrayidx32, false, label_forbody_60); - BinaryOperator* int32_indvar_next_72 = BinaryOperator::create(Instruction::Add, int32_i_01_0, const_int32_28, "indvar.next", label_forbody_60); - ICmpInst* int1_exitcond_73 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_72, int32_num_attribs_58, "exitcond", label_forbody_60); - new BranchInst(label_afterfor_61, label_forbody_60, int1_exitcond_73, label_forbody_60); + ptr_arrayidx32_indices.push_back(const_int32_40); + Instruction* ptr_arrayidx32 = new GetElementPtrInst(ptr_dests, ptr_arrayidx32_indices.begin(), ptr_arrayidx32_indices.end(), "arrayidx32", label_forbody_74); + StoreInst* void_87 = new StoreInst(float_tmp17, ptr_arrayidx32, false, label_forbody_74); + BinaryOperator* int32_indvar_next_88 = BinaryOperator::create(Instruction::Add, int32_i_01_0, const_int32_38, "indvar.next", label_forbody_74); + ICmpInst* int1_exitcond_89 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_88, int32_num_attribs_72, "exitcond", label_forbody_74); + new BranchInst(label_afterfor_75, label_forbody_74, int1_exitcond_89, label_forbody_74); - // Block afterfor (label_afterfor_61) - new ReturnInst(label_afterfor_61); + // Block afterfor (label_afterfor_75) + new ReturnInst(label_afterfor_75); // Resolve Forward References - fwdref_64->replaceAllUsesWith(int32_indvar_next_72); delete fwdref_64; + fwdref_78->replaceAllUsesWith(int32_indvar_next_88); delete fwdref_78; } // Function: run_vertex_shader (func_run_vertex_shader) { Function::arg_iterator args = func_run_vertex_shader->arg_begin(); - Value* ptr_ainputs_76 = args++; - ptr_ainputs_76->setName("ainputs"); - Value* ptr_dests_77 = args++; - ptr_dests_77->setName("dests"); + Value* ptr_ainputs_92 = args++; + ptr_ainputs_92->setName("ainputs"); + Value* ptr_dests_93 = args++; + ptr_dests_93->setName("dests"); Value* ptr_aconsts = args++; ptr_aconsts->setName("aconsts"); - Value* int32_count_78 = args++; - int32_count_78->setName("count"); - Value* int32_num_attribs_79 = args++; - int32_num_attribs_79->setName("num_attribs"); - - BasicBlock* label_entry_80 = new BasicBlock("entry",func_run_vertex_shader,0); + Value* int32_num_vertices = args++; + int32_num_vertices->setName("num_vertices"); + Value* int32_num_inputs = args++; + int32_num_inputs->setName("num_inputs"); + Value* int32_num_attribs_94 = args++; + int32_num_attribs_94->setName("num_attribs"); + Value* int32_num_consts = args++; + int32_num_consts->setName("num_consts"); + + BasicBlock* label_entry_95 = new BasicBlock("entry",func_run_vertex_shader,0); BasicBlock* label_forbody6_i = new BasicBlock("forbody6.i",func_run_vertex_shader,0); - BasicBlock* label_forinc57_i = new BasicBlock("forinc57.i",func_run_vertex_shader,0); - BasicBlock* label_forbody_i11_preheader = new BasicBlock("forbody.i11.preheader",func_run_vertex_shader,0); + BasicBlock* label_forinc70_i = new BasicBlock("forinc70.i",func_run_vertex_shader,0); + BasicBlock* label_from_array_exit = new BasicBlock("from_array.exit",func_run_vertex_shader,0); + BasicBlock* label_forbody_i15 = new BasicBlock("forbody.i15",func_run_vertex_shader,0); + BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_vertex_shader,0); BasicBlock* label_forbody_preheader = new BasicBlock("forbody.preheader",func_run_vertex_shader,0); BasicBlock* label_forbody_us = new BasicBlock("forbody.us",func_run_vertex_shader,0); BasicBlock* label_to_array_exit_us = new BasicBlock("to_array.exit.us",func_run_vertex_shader,0); BasicBlock* label_forbody_i_us = new BasicBlock("forbody.i.us",func_run_vertex_shader,0); - BasicBlock* label_forbody_81 = new BasicBlock("forbody",func_run_vertex_shader,0); - BasicBlock* label_afterfor_82 = new BasicBlock("afterfor",func_run_vertex_shader,0); - - // Block entry (label_entry_80) - AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_21, "inputs", label_entry_80); - AllocaInst* ptr_consts = new AllocaInst(ArrayTy_6, "consts", label_entry_80); - AllocaInst* ptr_results = new AllocaInst(ArrayTy_21, "results", label_entry_80); - CallInst* int32_call_83 = new CallInst(func_printf, const_ptr_32, "call", label_entry_80); - int32_call_83->setCallingConv(CallingConv::C); - int32_call_83->setTailCall(false); - ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_78, const_int32_26, "cmp.i", label_entry_80); - ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_79, const_int32_26, "cmp5.i", label_entry_80); - BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_80); - new BranchInst(label_forbody6_i, label_forbody_i11_preheader, int1_bothcond_i, label_entry_80); + BasicBlock* label_forbody_96 = new BasicBlock("forbody",func_run_vertex_shader,0); + BasicBlock* label_afterfor_97 = new BasicBlock("afterfor",func_run_vertex_shader,0); + + // Block entry (label_entry_95) + AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_25, "inputs", label_entry_95); + AllocaInst* ptr_consts = new AllocaInst(ArrayTy_27, "consts", label_entry_95); + AllocaInst* ptr_results = new AllocaInst(ArrayTy_25, "results", label_entry_95); + std::vector int32_call_98_params; + int32_call_98_params.push_back(const_ptr_44); + int32_call_98_params.push_back(int32_num_vertices); + int32_call_98_params.push_back(int32_num_inputs); + int32_call_98_params.push_back(int32_num_consts); + CallInst* int32_call_98 = new CallInst(func_printf, int32_call_98_params.begin(), int32_call_98_params.end(), "call", label_entry_95); + int32_call_98->setCallingConv(CallingConv::C); + int32_call_98->setTailCall(false); + ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_36, "cmp.i", label_entry_95); + ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_36, "cmp5.i", label_entry_95); + BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_95); + new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_95); // Block forbody6.i (label_forbody6_i) - Argument* fwdref_85 = new Argument(IntegerType::get(32)); - Argument* fwdref_86 = new Argument(IntegerType::get(32)); - PHINode* int32_i_013_0_i_ph = new PHINode(IntegerType::get(32), "i.013.0.i.ph", label_forbody6_i); - int32_i_013_0_i_ph->reserveOperandSpace(3); - int32_i_013_0_i_ph->addIncoming(const_int32_26, label_entry_80); - int32_i_013_0_i_ph->addIncoming(fwdref_85, label_forinc57_i); - int32_i_013_0_i_ph->addIncoming(fwdref_86, label_forbody6_i); - - Argument* fwdref_87 = new Argument(IntegerType::get(32)); + Argument* fwdref_100 = new Argument(IntegerType::get(32)); + Argument* fwdref_101 = new Argument(IntegerType::get(32)); + PHINode* int32_i_014_0_i_ph = new PHINode(IntegerType::get(32), "i.014.0.i.ph", label_forbody6_i); + int32_i_014_0_i_ph->reserveOperandSpace(3); + int32_i_014_0_i_ph->addIncoming(const_int32_36, label_entry_95); + int32_i_014_0_i_ph->addIncoming(fwdref_100, label_forinc70_i); + int32_i_014_0_i_ph->addIncoming(fwdref_101, label_forbody6_i); + + Argument* fwdref_102 = new Argument(IntegerType::get(32)); PHINode* int32_j_03_0_i = new PHINode(IntegerType::get(32), "j.03.0.i", label_forbody6_i); int32_j_03_0_i->reserveOperandSpace(3); - int32_j_03_0_i->addIncoming(fwdref_87, label_forbody6_i); - int32_j_03_0_i->addIncoming(const_int32_26, label_forinc57_i); - int32_j_03_0_i->addIncoming(const_int32_26, label_entry_80); + int32_j_03_0_i->addIncoming(fwdref_102, label_forbody6_i); + int32_j_03_0_i->addIncoming(const_int32_36, label_forinc70_i); + int32_j_03_0_i->addIncoming(const_int32_36, label_entry_95); - Argument* fwdref_88 = new Argument(VectorTy_7); - PHINode* packed_vec_01_0_i = new PHINode(VectorTy_7, "vec.01.0.i", label_forbody6_i); + Argument* fwdref_103 = new Argument(VectorTy_11); + PHINode* packed_vec_01_0_i = new PHINode(VectorTy_11, "vec.01.0.i", label_forbody6_i); packed_vec_01_0_i->reserveOperandSpace(3); - packed_vec_01_0_i->addIncoming(fwdref_88, label_forbody6_i); - packed_vec_01_0_i->addIncoming(const_packed_27, label_entry_80); - packed_vec_01_0_i->addIncoming(fwdref_88, label_forinc57_i); + packed_vec_01_0_i->addIncoming(fwdref_103, label_forbody6_i); + packed_vec_01_0_i->addIncoming(const_packed_37, label_entry_95); + packed_vec_01_0_i->addIncoming(fwdref_103, label_forinc70_i); std::vector ptr_arraydecay11_i_indices; - ptr_arraydecay11_i_indices.push_back(int32_i_013_0_i_ph); + ptr_arraydecay11_i_indices.push_back(int32_i_014_0_i_ph); ptr_arraydecay11_i_indices.push_back(int32_j_03_0_i); - ptr_arraydecay11_i_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_76, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); + ptr_arraydecay11_i_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_92, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); LoadInst* float_tmp13_i = new LoadInst(ptr_arraydecay11_i, "tmp13.i", false, label_forbody6_i); - InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_01_0_i, float_tmp13_i, const_int32_26, "tmp15.i", label_forbody6_i); + InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_01_0_i, float_tmp13_i, const_int32_36, "tmp15.i", label_forbody6_i); std::vector ptr_arrayidx23_i_indices; - ptr_arrayidx23_i_indices.push_back(int32_i_013_0_i_ph); + ptr_arrayidx23_i_indices.push_back(int32_i_014_0_i_ph); ptr_arrayidx23_i_indices.push_back(int32_j_03_0_i); - ptr_arrayidx23_i_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_76, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); + ptr_arrayidx23_i_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_92, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); LoadInst* float_tmp24_i = new LoadInst(ptr_arrayidx23_i, "tmp24.i", false, label_forbody6_i); - InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_28, "tmp26.i", label_forbody6_i); + InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_38, "tmp26.i", label_forbody6_i); std::vector ptr_arrayidx34_i_indices; - ptr_arrayidx34_i_indices.push_back(int32_i_013_0_i_ph); + ptr_arrayidx34_i_indices.push_back(int32_i_014_0_i_ph); ptr_arrayidx34_i_indices.push_back(int32_j_03_0_i); - ptr_arrayidx34_i_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_76, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); + ptr_arrayidx34_i_indices.push_back(const_int32_39); + Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_92, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); LoadInst* float_tmp35_i = new LoadInst(ptr_arrayidx34_i, "tmp35.i", false, label_forbody6_i); - InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_29, "tmp37.i", label_forbody6_i); + InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_39, "tmp37.i", label_forbody6_i); std::vector ptr_arrayidx45_i_indices; - ptr_arrayidx45_i_indices.push_back(int32_i_013_0_i_ph); + ptr_arrayidx45_i_indices.push_back(int32_i_014_0_i_ph); ptr_arrayidx45_i_indices.push_back(int32_j_03_0_i); - ptr_arrayidx45_i_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_76, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); + ptr_arrayidx45_i_indices.push_back(const_int32_40); + Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_92, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); LoadInst* float_tmp46_i = new LoadInst(ptr_arrayidx45_i, "tmp46.i", false, label_forbody6_i); - InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_30, "tmp48.i", label_forbody6_i); - std::vector ptr_arrayidx54_i_indices; - ptr_arrayidx54_i_indices.push_back(const_int32_26); - ptr_arrayidx54_i_indices.push_back(int32_i_013_0_i_ph); - ptr_arrayidx54_i_indices.push_back(int32_j_03_0_i); - Instruction* ptr_arrayidx54_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx54_i_indices.begin(), ptr_arrayidx54_i_indices.end(), "arrayidx54.i", label_forbody6_i); - new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); - BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_03_0_i, const_int32_28, "inc.i", label_forbody6_i); - ICmpInst* int1_cmp511_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_attribs_79, "cmp511.i", label_forbody6_i); - new BranchInst(label_forbody6_i, label_forinc57_i, int1_cmp511_i, label_forbody6_i); - - // Block forinc57.i (label_forinc57_i) - BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_013_0_i_ph, const_int32_28, "inc59.i", label_forinc57_i); - ICmpInst* int1_cmp21_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59_i, int32_count_78, "cmp21.i", label_forinc57_i); - new BranchInst(label_forbody6_i, label_forbody_i11_preheader, int1_cmp21_i, label_forinc57_i); - - // Block forbody.i11.preheader (label_forbody_i11_preheader) - std::vector ptr_arraydecay_i3_indices; - ptr_arraydecay_i3_indices.push_back(const_int32_26); - ptr_arraydecay_i3_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay_i3 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_indices.begin(), ptr_arraydecay_i3_indices.end(), "arraydecay.i3", label_forbody_i11_preheader); - LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i3, "tmp5.i", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp7_i4 = new InsertElementInst(const_packed_27, float_tmp5_i, const_int32_26, "tmp7.i4", label_forbody_i11_preheader); + InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_40, "tmp48.i", label_forbody6_i); + CastInst* double_conv_i = new FPExtInst(float_tmp13_i, Type::DoubleTy, "conv.i", label_forbody6_i); + CastInst* double_conv55_i = new FPExtInst(float_tmp24_i, Type::DoubleTy, "conv55.i", label_forbody6_i); + CastInst* double_conv58_i = new FPExtInst(float_tmp35_i, Type::DoubleTy, "conv58.i", label_forbody6_i); + CastInst* double_conv61_i = new FPExtInst(float_tmp46_i, Type::DoubleTy, "conv61.i", label_forbody6_i); + std::vector int32_call_i_params; + int32_call_i_params.push_back(const_ptr_41); + int32_call_i_params.push_back(int32_i_014_0_i_ph); + int32_call_i_params.push_back(int32_j_03_0_i); + int32_call_i_params.push_back(double_conv_i); + int32_call_i_params.push_back(double_conv55_i); + int32_call_i_params.push_back(double_conv58_i); + int32_call_i_params.push_back(double_conv61_i); + CallInst* int32_call_i = new CallInst(func_printf, int32_call_i_params.begin(), int32_call_i_params.end(), "call.i", label_forbody6_i); + int32_call_i->setCallingConv(CallingConv::C); + int32_call_i->setTailCall(false); + std::vector ptr_arrayidx67_i_indices; + ptr_arrayidx67_i_indices.push_back(const_int32_36); + ptr_arrayidx67_i_indices.push_back(int32_i_014_0_i_ph); + ptr_arrayidx67_i_indices.push_back(int32_j_03_0_i); + Instruction* ptr_arrayidx67_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx67_i_indices.begin(), ptr_arrayidx67_i_indices.end(), "arrayidx67.i", label_forbody6_i); + StoreInst* void_104 = new StoreInst(packed_tmp48_i, ptr_arrayidx67_i, false, label_forbody6_i); + BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_03_0_i, const_int32_38, "inc.i", label_forbody6_i); + ICmpInst* int1_cmp512_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_inputs, "cmp512.i", label_forbody6_i); + new BranchInst(label_forbody6_i, label_forinc70_i, int1_cmp512_i, label_forbody6_i); + + // Block forinc70.i (label_forinc70_i) + BinaryOperator* int32_inc72_i = BinaryOperator::create(Instruction::Add, int32_i_014_0_i_ph, const_int32_38, "inc72.i", label_forinc70_i); + ICmpInst* int1_cmp23_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc72_i, int32_num_vertices, "cmp23.i", label_forinc70_i); + new BranchInst(label_forbody6_i, label_from_array_exit, int1_cmp23_i, label_forinc70_i); + + // Block from_array.exit (label_from_array_exit) + ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_36, "cmp.i4", label_from_array_exit); + new BranchInst(label_forbody_i15, label_from_consts_exit, int1_cmp_i4, label_from_array_exit); + + // Block forbody.i15 (label_forbody_i15) + Argument* fwdref_108 = new Argument(IntegerType::get(32)); + PHINode* int32_i_02_0_i = new PHINode(IntegerType::get(32), "i.02.0.i", label_forbody_i15); + int32_i_02_0_i->reserveOperandSpace(2); + int32_i_02_0_i->addIncoming(const_int32_36, label_from_array_exit); + int32_i_02_0_i->addIncoming(fwdref_108, label_forbody_i15); + + Argument* fwdref_109 = new Argument(VectorTy_11); + PHINode* packed_vec_01_0_i5 = new PHINode(VectorTy_11, "vec.01.0.i5", label_forbody_i15); + packed_vec_01_0_i5->reserveOperandSpace(2); + packed_vec_01_0_i5->addIncoming(const_packed_37, label_from_array_exit); + packed_vec_01_0_i5->addIncoming(fwdref_109, label_forbody_i15); + + std::vector ptr_arraydecay_i6_indices; + ptr_arraydecay_i6_indices.push_back(int32_i_02_0_i); + ptr_arraydecay_i6_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay_i6 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i6_indices.begin(), ptr_arraydecay_i6_indices.end(), "arraydecay.i6", label_forbody_i15); + LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i6, "tmp5.i", false, label_forbody_i15); + InsertElementInst* packed_tmp7_i7 = new InsertElementInst(packed_vec_01_0_i5, float_tmp5_i, const_int32_36, "tmp7.i7", label_forbody_i15); std::vector ptr_arrayidx12_i_indices; - ptr_arrayidx12_i_indices.push_back(const_int32_26); - ptr_arrayidx12_i_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i11_preheader); - LoadInst* float_tmp13_i5 = new LoadInst(ptr_arrayidx12_i, "tmp13.i5", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp15_i6 = new InsertElementInst(packed_tmp7_i4, float_tmp13_i5, const_int32_28, "tmp15.i6", label_forbody_i11_preheader); + ptr_arrayidx12_i_indices.push_back(int32_i_02_0_i); + ptr_arrayidx12_i_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i15); + LoadInst* float_tmp13_i8 = new LoadInst(ptr_arrayidx12_i, "tmp13.i8", false, label_forbody_i15); + InsertElementInst* packed_tmp15_i9 = new InsertElementInst(packed_tmp7_i7, float_tmp13_i8, const_int32_38, "tmp15.i9", label_forbody_i15); std::vector ptr_arrayidx20_i_indices; - ptr_arrayidx20_i_indices.push_back(const_int32_26); - ptr_arrayidx20_i_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i11_preheader); - LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp23_i = new InsertElementInst(packed_tmp15_i6, float_tmp21_i, const_int32_29, "tmp23.i", label_forbody_i11_preheader); - std::vector ptr_arrayidx28_i7_indices; - ptr_arrayidx28_i7_indices.push_back(const_int32_26); - ptr_arrayidx28_i7_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx28_i7 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_indices.begin(), ptr_arrayidx28_i7_indices.end(), "arrayidx28.i7", label_forbody_i11_preheader); - LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i7, "tmp29.i", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i, float_tmp29_i, const_int32_30, "tmp31.i", label_forbody_i11_preheader); - std::vector ptr_arrayidx34_i8_indices; - ptr_arrayidx34_i8_indices.push_back(const_int32_26); - ptr_arrayidx34_i8_indices.push_back(const_int32_26); - Instruction* ptr_arrayidx34_i8 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_indices.begin(), ptr_arrayidx34_i8_indices.end(), "arrayidx34.i8", label_forbody_i11_preheader); - new StoreInst(packed_tmp31_i, ptr_arrayidx34_i8, false, label_forbody_i11_preheader); - std::vector ptr_arraydecay_i3_1_indices; - ptr_arraydecay_i3_1_indices.push_back(const_int32_28); - ptr_arraydecay_i3_1_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay_i3_1 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_1_indices.begin(), ptr_arraydecay_i3_1_indices.end(), "arraydecay.i3.1", label_forbody_i11_preheader); - LoadInst* float_tmp5_i_1 = new LoadInst(ptr_arraydecay_i3_1, "tmp5.i.1", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp7_i4_1 = new InsertElementInst(packed_tmp31_i, float_tmp5_i_1, const_int32_26, "tmp7.i4.1", label_forbody_i11_preheader); - std::vector ptr_arrayidx12_i_1_indices; - ptr_arrayidx12_i_1_indices.push_back(const_int32_28); - ptr_arrayidx12_i_1_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx12_i_1 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_1_indices.begin(), ptr_arrayidx12_i_1_indices.end(), "arrayidx12.i.1", label_forbody_i11_preheader); - LoadInst* float_tmp13_i5_1 = new LoadInst(ptr_arrayidx12_i_1, "tmp13.i5.1", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp15_i6_1 = new InsertElementInst(packed_tmp7_i4_1, float_tmp13_i5_1, const_int32_28, "tmp15.i6.1", label_forbody_i11_preheader); - std::vector ptr_arrayidx20_i_1_indices; - ptr_arrayidx20_i_1_indices.push_back(const_int32_28); - ptr_arrayidx20_i_1_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx20_i_1 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_1_indices.begin(), ptr_arrayidx20_i_1_indices.end(), "arrayidx20.i.1", label_forbody_i11_preheader); - LoadInst* float_tmp21_i_1 = new LoadInst(ptr_arrayidx20_i_1, "tmp21.i.1", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp23_i_1 = new InsertElementInst(packed_tmp15_i6_1, float_tmp21_i_1, const_int32_29, "tmp23.i.1", label_forbody_i11_preheader); - std::vector ptr_arrayidx28_i7_1_indices; - ptr_arrayidx28_i7_1_indices.push_back(const_int32_28); - ptr_arrayidx28_i7_1_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx28_i7_1 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_1_indices.begin(), ptr_arrayidx28_i7_1_indices.end(), "arrayidx28.i7.1", label_forbody_i11_preheader); - LoadInst* float_tmp29_i_1 = new LoadInst(ptr_arrayidx28_i7_1, "tmp29.i.1", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp31_i_1 = new InsertElementInst(packed_tmp23_i_1, float_tmp29_i_1, const_int32_30, "tmp31.i.1", label_forbody_i11_preheader); - std::vector ptr_arrayidx34_i8_1_indices; - ptr_arrayidx34_i8_1_indices.push_back(const_int32_26); - ptr_arrayidx34_i8_1_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx34_i8_1 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_1_indices.begin(), ptr_arrayidx34_i8_1_indices.end(), "arrayidx34.i8.1", label_forbody_i11_preheader); - new StoreInst(packed_tmp31_i_1, ptr_arrayidx34_i8_1, false, label_forbody_i11_preheader); - std::vector ptr_arraydecay_i3_2_indices; - ptr_arraydecay_i3_2_indices.push_back(const_int32_29); - ptr_arraydecay_i3_2_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay_i3_2 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_2_indices.begin(), ptr_arraydecay_i3_2_indices.end(), "arraydecay.i3.2", label_forbody_i11_preheader); - LoadInst* float_tmp5_i_2 = new LoadInst(ptr_arraydecay_i3_2, "tmp5.i.2", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp7_i4_2 = new InsertElementInst(packed_tmp31_i_1, float_tmp5_i_2, const_int32_26, "tmp7.i4.2", label_forbody_i11_preheader); - std::vector ptr_arrayidx12_i_2_indices; - ptr_arrayidx12_i_2_indices.push_back(const_int32_29); - ptr_arrayidx12_i_2_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx12_i_2 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_2_indices.begin(), ptr_arrayidx12_i_2_indices.end(), "arrayidx12.i.2", label_forbody_i11_preheader); - LoadInst* float_tmp13_i5_2 = new LoadInst(ptr_arrayidx12_i_2, "tmp13.i5.2", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp15_i6_2 = new InsertElementInst(packed_tmp7_i4_2, float_tmp13_i5_2, const_int32_28, "tmp15.i6.2", label_forbody_i11_preheader); - std::vector ptr_arrayidx20_i_2_indices; - ptr_arrayidx20_i_2_indices.push_back(const_int32_29); - ptr_arrayidx20_i_2_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx20_i_2 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_2_indices.begin(), ptr_arrayidx20_i_2_indices.end(), "arrayidx20.i.2", label_forbody_i11_preheader); - LoadInst* float_tmp21_i_2 = new LoadInst(ptr_arrayidx20_i_2, "tmp21.i.2", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp23_i_2 = new InsertElementInst(packed_tmp15_i6_2, float_tmp21_i_2, const_int32_29, "tmp23.i.2", label_forbody_i11_preheader); - std::vector ptr_arrayidx28_i7_2_indices; - ptr_arrayidx28_i7_2_indices.push_back(const_int32_29); - ptr_arrayidx28_i7_2_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx28_i7_2 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_2_indices.begin(), ptr_arrayidx28_i7_2_indices.end(), "arrayidx28.i7.2", label_forbody_i11_preheader); - LoadInst* float_tmp29_i_2 = new LoadInst(ptr_arrayidx28_i7_2, "tmp29.i.2", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp31_i_2 = new InsertElementInst(packed_tmp23_i_2, float_tmp29_i_2, const_int32_30, "tmp31.i.2", label_forbody_i11_preheader); - std::vector ptr_arrayidx34_i8_2_indices; - ptr_arrayidx34_i8_2_indices.push_back(const_int32_26); - ptr_arrayidx34_i8_2_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx34_i8_2 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_2_indices.begin(), ptr_arrayidx34_i8_2_indices.end(), "arrayidx34.i8.2", label_forbody_i11_preheader); - new StoreInst(packed_tmp31_i_2, ptr_arrayidx34_i8_2, false, label_forbody_i11_preheader); - std::vector ptr_arraydecay_i3_3_indices; - ptr_arraydecay_i3_3_indices.push_back(const_int32_30); - ptr_arraydecay_i3_3_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay_i3_3 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_3_indices.begin(), ptr_arraydecay_i3_3_indices.end(), "arraydecay.i3.3", label_forbody_i11_preheader); - LoadInst* float_tmp5_i_3 = new LoadInst(ptr_arraydecay_i3_3, "tmp5.i.3", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp7_i4_3 = new InsertElementInst(packed_tmp31_i_2, float_tmp5_i_3, const_int32_26, "tmp7.i4.3", label_forbody_i11_preheader); - std::vector ptr_arrayidx12_i_3_indices; - ptr_arrayidx12_i_3_indices.push_back(const_int32_30); - ptr_arrayidx12_i_3_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx12_i_3 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_3_indices.begin(), ptr_arrayidx12_i_3_indices.end(), "arrayidx12.i.3", label_forbody_i11_preheader); - LoadInst* float_tmp13_i5_3 = new LoadInst(ptr_arrayidx12_i_3, "tmp13.i5.3", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp15_i6_3 = new InsertElementInst(packed_tmp7_i4_3, float_tmp13_i5_3, const_int32_28, "tmp15.i6.3", label_forbody_i11_preheader); - std::vector ptr_arrayidx20_i_3_indices; - ptr_arrayidx20_i_3_indices.push_back(const_int32_30); - ptr_arrayidx20_i_3_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx20_i_3 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_3_indices.begin(), ptr_arrayidx20_i_3_indices.end(), "arrayidx20.i.3", label_forbody_i11_preheader); - LoadInst* float_tmp21_i_3 = new LoadInst(ptr_arrayidx20_i_3, "tmp21.i.3", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp23_i_3 = new InsertElementInst(packed_tmp15_i6_3, float_tmp21_i_3, const_int32_29, "tmp23.i.3", label_forbody_i11_preheader); - std::vector ptr_arrayidx28_i7_3_indices; - ptr_arrayidx28_i7_3_indices.push_back(const_int32_30); - ptr_arrayidx28_i7_3_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx28_i7_3 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_3_indices.begin(), ptr_arrayidx28_i7_3_indices.end(), "arrayidx28.i7.3", label_forbody_i11_preheader); - LoadInst* float_tmp29_i_3 = new LoadInst(ptr_arrayidx28_i7_3, "tmp29.i.3", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp31_i_3 = new InsertElementInst(packed_tmp23_i_3, float_tmp29_i_3, const_int32_30, "tmp31.i.3", label_forbody_i11_preheader); - std::vector ptr_arrayidx34_i8_3_indices; - ptr_arrayidx34_i8_3_indices.push_back(const_int32_26); - ptr_arrayidx34_i8_3_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx34_i8_3 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_3_indices.begin(), ptr_arrayidx34_i8_3_indices.end(), "arrayidx34.i8.3", label_forbody_i11_preheader); - new StoreInst(packed_tmp31_i_3, ptr_arrayidx34_i8_3, false, label_forbody_i11_preheader); - std::vector ptr_arraydecay_i3_4_indices; - ptr_arraydecay_i3_4_indices.push_back(const_int32_33); - ptr_arraydecay_i3_4_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay_i3_4 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i3_4_indices.begin(), ptr_arraydecay_i3_4_indices.end(), "arraydecay.i3.4", label_forbody_i11_preheader); - LoadInst* float_tmp5_i_4 = new LoadInst(ptr_arraydecay_i3_4, "tmp5.i.4", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp7_i4_4 = new InsertElementInst(packed_tmp31_i_3, float_tmp5_i_4, const_int32_26, "tmp7.i4.4", label_forbody_i11_preheader); - std::vector ptr_arrayidx12_i_4_indices; - ptr_arrayidx12_i_4_indices.push_back(const_int32_33); - ptr_arrayidx12_i_4_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx12_i_4 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_4_indices.begin(), ptr_arrayidx12_i_4_indices.end(), "arrayidx12.i.4", label_forbody_i11_preheader); - LoadInst* float_tmp13_i5_4 = new LoadInst(ptr_arrayidx12_i_4, "tmp13.i5.4", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp15_i6_4 = new InsertElementInst(packed_tmp7_i4_4, float_tmp13_i5_4, const_int32_28, "tmp15.i6.4", label_forbody_i11_preheader); - std::vector ptr_arrayidx20_i_4_indices; - ptr_arrayidx20_i_4_indices.push_back(const_int32_33); - ptr_arrayidx20_i_4_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx20_i_4 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_4_indices.begin(), ptr_arrayidx20_i_4_indices.end(), "arrayidx20.i.4", label_forbody_i11_preheader); - LoadInst* float_tmp21_i_4 = new LoadInst(ptr_arrayidx20_i_4, "tmp21.i.4", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp23_i_4 = new InsertElementInst(packed_tmp15_i6_4, float_tmp21_i_4, const_int32_29, "tmp23.i.4", label_forbody_i11_preheader); - std::vector ptr_arrayidx28_i7_4_indices; - ptr_arrayidx28_i7_4_indices.push_back(const_int32_33); - ptr_arrayidx28_i7_4_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx28_i7_4 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i7_4_indices.begin(), ptr_arrayidx28_i7_4_indices.end(), "arrayidx28.i7.4", label_forbody_i11_preheader); - LoadInst* float_tmp29_i_4 = new LoadInst(ptr_arrayidx28_i7_4, "tmp29.i.4", false, label_forbody_i11_preheader); - InsertElementInst* packed_tmp31_i_4 = new InsertElementInst(packed_tmp23_i_4, float_tmp29_i_4, const_int32_30, "tmp31.i.4", label_forbody_i11_preheader); - std::vector ptr_arrayidx34_i8_4_indices; - ptr_arrayidx34_i8_4_indices.push_back(const_int32_26); - ptr_arrayidx34_i8_4_indices.push_back(const_int32_33); - Instruction* ptr_arrayidx34_i8_4 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i8_4_indices.begin(), ptr_arrayidx34_i8_4_indices.end(), "arrayidx34.i8.4", label_forbody_i11_preheader); - new StoreInst(packed_tmp31_i_4, ptr_arrayidx34_i8_4, false, label_forbody_i11_preheader); - new BranchInst(label_forbody_preheader, label_afterfor_82, int1_cmp_i, label_forbody_i11_preheader); + ptr_arrayidx20_i_indices.push_back(int32_i_02_0_i); + ptr_arrayidx20_i_indices.push_back(const_int32_39); + Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i15); + LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i15); + InsertElementInst* packed_tmp23_i = new InsertElementInst(packed_tmp15_i9, float_tmp21_i, const_int32_39, "tmp23.i", label_forbody_i15); + std::vector ptr_arrayidx28_i10_indices; + ptr_arrayidx28_i10_indices.push_back(int32_i_02_0_i); + ptr_arrayidx28_i10_indices.push_back(const_int32_40); + Instruction* ptr_arrayidx28_i10 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i10_indices.begin(), ptr_arrayidx28_i10_indices.end(), "arrayidx28.i10", label_forbody_i15); + LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i10, "tmp29.i", false, label_forbody_i15); + InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i, float_tmp29_i, const_int32_40, "tmp31.i", label_forbody_i15); + CastInst* double_conv_i11 = new FPExtInst(float_tmp5_i, Type::DoubleTy, "conv.i11", label_forbody_i15); + CastInst* double_conv37_i = new FPExtInst(float_tmp13_i8, Type::DoubleTy, "conv37.i", label_forbody_i15); + CastInst* double_conv40_i = new FPExtInst(float_tmp21_i, Type::DoubleTy, "conv40.i", label_forbody_i15); + CastInst* double_conv43_i = new FPExtInst(float_tmp29_i, Type::DoubleTy, "conv43.i", label_forbody_i15); + std::vector int32_call_i12_params; + int32_call_i12_params.push_back(const_ptr_42); + int32_call_i12_params.push_back(int32_i_02_0_i); + int32_call_i12_params.push_back(double_conv_i11); + int32_call_i12_params.push_back(double_conv37_i); + int32_call_i12_params.push_back(double_conv40_i); + int32_call_i12_params.push_back(double_conv43_i); + CallInst* int32_call_i12 = new CallInst(func_printf, int32_call_i12_params.begin(), int32_call_i12_params.end(), "call.i12", label_forbody_i15); + int32_call_i12->setCallingConv(CallingConv::C); + int32_call_i12->setTailCall(false); + std::vector ptr_arrayidx46_i_indices; + ptr_arrayidx46_i_indices.push_back(const_int32_36); + ptr_arrayidx46_i_indices.push_back(int32_i_02_0_i); + Instruction* ptr_arrayidx46_i = new GetElementPtrInst(ptr_consts, ptr_arrayidx46_i_indices.begin(), ptr_arrayidx46_i_indices.end(), "arrayidx46.i", label_forbody_i15); + StoreInst* void_110 = new StoreInst(packed_tmp31_i, ptr_arrayidx46_i, false, label_forbody_i15); + BinaryOperator* int32_indvar_next25 = BinaryOperator::create(Instruction::Add, int32_i_02_0_i, const_int32_38, "indvar.next25", label_forbody_i15); + ICmpInst* int1_exitcond26 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next25, int32_num_consts, "exitcond26", label_forbody_i15); + new BranchInst(label_from_consts_exit, label_forbody_i15, int1_exitcond26, label_forbody_i15); + + // Block from_consts.exit (label_from_consts_exit) + CallInst* int32_call9 = new CallInst(func_printf, const_ptr_45, "call9", label_from_consts_exit); + int32_call9->setCallingConv(CallingConv::C); + int32_call9->setTailCall(false); + new BranchInst(label_forbody_preheader, label_afterfor_97, int1_cmp_i, label_from_consts_exit); // Block forbody.preheader (label_forbody_preheader) - new BranchInst(label_forbody_us, label_forbody_81, int1_cmp5_i, label_forbody_preheader); + std::vector ptr_arraydecay21_indices; + ptr_arraydecay21_indices.push_back(const_int32_36); + ptr_arraydecay21_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay21 = new GetElementPtrInst(ptr_consts, ptr_arraydecay21_indices.begin(), ptr_arraydecay21_indices.end(), "arraydecay21", label_forbody_preheader); + ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_94, const_int32_36, "cmp.i1", label_forbody_preheader); + new BranchInst(label_forbody_us, label_forbody_96, int1_cmp_i1, label_forbody_preheader); // Block forbody.us (label_forbody_us) - Argument* fwdref_99 = new Argument(IntegerType::get(32)); - PHINode* int32_i_012_0_us = new PHINode(IntegerType::get(32), "i.012.0.us", label_forbody_us); - int32_i_012_0_us->reserveOperandSpace(2); - int32_i_012_0_us->addIncoming(const_int32_26, label_forbody_preheader); - int32_i_012_0_us->addIncoming(fwdref_99, label_to_array_exit_us); - - std::vector ptr_arraydecay9_us_indices; - ptr_arraydecay9_us_indices.push_back(const_int32_26); - ptr_arraydecay9_us_indices.push_back(int32_i_012_0_us); - ptr_arraydecay9_us_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay9_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay9_us_indices.begin(), ptr_arraydecay9_us_indices.end(), "arraydecay9.us", label_forbody_us); - std::vector ptr_arraydecay13_us_indices; - ptr_arraydecay13_us_indices.push_back(const_int32_26); - ptr_arraydecay13_us_indices.push_back(int32_i_012_0_us); - ptr_arraydecay13_us_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay13_us = new GetElementPtrInst(ptr_results, ptr_arraydecay13_us_indices.begin(), ptr_arraydecay13_us_indices.end(), "arraydecay13.us", label_forbody_us); - std::vector void_100_params; - void_100_params.push_back(ptr_arraydecay13_us); - void_100_params.push_back(ptr_arraydecay9_us); - void_100_params.push_back(ptr_arrayidx34_i8); - CallInst* void_100 = new CallInst(func_execute_shader, void_100_params.begin(), void_100_params.end(), "", label_forbody_us); - void_100->setCallingConv(CallingConv::C); - void_100->setTailCall(false); + Argument* fwdref_114 = new Argument(IntegerType::get(32)); + PHINode* int32_i_016_0_us = new PHINode(IntegerType::get(32), "i.016.0.us", label_forbody_us); + int32_i_016_0_us->reserveOperandSpace(2); + int32_i_016_0_us->addIncoming(const_int32_36, label_forbody_preheader); + int32_i_016_0_us->addIncoming(fwdref_114, label_to_array_exit_us); + + std::vector ptr_arraydecay14_us_indices; + ptr_arraydecay14_us_indices.push_back(const_int32_36); + ptr_arraydecay14_us_indices.push_back(int32_i_016_0_us); + ptr_arraydecay14_us_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay14_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay14_us_indices.begin(), ptr_arraydecay14_us_indices.end(), "arraydecay14.us", label_forbody_us); + std::vector ptr_arraydecay18_us_indices; + ptr_arraydecay18_us_indices.push_back(const_int32_36); + ptr_arraydecay18_us_indices.push_back(int32_i_016_0_us); + ptr_arraydecay18_us_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay18_us = new GetElementPtrInst(ptr_results, ptr_arraydecay18_us_indices.begin(), ptr_arraydecay18_us_indices.end(), "arraydecay18.us", label_forbody_us); + std::vector void_115_params; + void_115_params.push_back(ptr_arraydecay18_us); + void_115_params.push_back(ptr_arraydecay14_us); + void_115_params.push_back(ptr_arraydecay21); + CallInst* void_115 = new CallInst(func_execute_shader, void_115_params.begin(), void_115_params.end(), "", label_forbody_us); + void_115->setCallingConv(CallingConv::C); + void_115->setTailCall(false); + CallInst* int32_call22_us = new CallInst(func_printf, const_ptr_46, "call22.us", label_forbody_us); + int32_call22_us->setCallingConv(CallingConv::C); + int32_call22_us->setTailCall(false); new BranchInst(label_forbody_i_us, label_forbody_us); // Block to_array.exit.us (label_to_array_exit_us) - BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_012_0_us, const_int32_28, "inc.us", label_to_array_exit_us); - ICmpInst* int1_cmp18_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_count_78, "cmp18.us", label_to_array_exit_us); - new BranchInst(label_forbody_us, label_afterfor_82, int1_cmp18_us, label_to_array_exit_us); + BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_016_0_us, const_int32_38, "inc.us", label_to_array_exit_us); + ICmpInst* int1_cmp22_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_num_vertices, "cmp22.us", label_to_array_exit_us); + new BranchInst(label_forbody_us, label_afterfor_97, int1_cmp22_us, label_to_array_exit_us); // Block forbody.i.us (label_forbody_i_us) - Argument* fwdref_103 = new Argument(IntegerType::get(32)); + Argument* fwdref_118 = new Argument(IntegerType::get(32)); PHINode* int32_i_01_0_i_us = new PHINode(IntegerType::get(32), "i.01.0.i.us", label_forbody_i_us); int32_i_01_0_i_us->reserveOperandSpace(2); - int32_i_01_0_i_us->addIncoming(const_int32_26, label_forbody_us); - int32_i_01_0_i_us->addIncoming(fwdref_103, label_forbody_i_us); + int32_i_01_0_i_us->addIncoming(const_int32_36, label_forbody_us); + int32_i_01_0_i_us->addIncoming(fwdref_118, label_forbody_i_us); std::vector ptr_arraydecay_i_us_indices; - ptr_arraydecay_i_us_indices.push_back(int32_i_012_0_us); + ptr_arraydecay_i_us_indices.push_back(int32_i_016_0_us); ptr_arraydecay_i_us_indices.push_back(int32_i_01_0_i_us); - ptr_arraydecay_i_us_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_77, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); + ptr_arraydecay_i_us_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_93, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); std::vector ptr_arrayidx6_i_us_indices; - ptr_arrayidx6_i_us_indices.push_back(const_int32_26); - ptr_arrayidx6_i_us_indices.push_back(int32_i_012_0_us); + ptr_arrayidx6_i_us_indices.push_back(const_int32_36); + ptr_arrayidx6_i_us_indices.push_back(int32_i_016_0_us); ptr_arrayidx6_i_us_indices.push_back(int32_i_01_0_i_us); Instruction* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_results, ptr_arrayidx6_i_us_indices.begin(), ptr_arrayidx6_i_us_indices.end(), "arrayidx6.i.us", label_forbody_i_us); LoadInst* packed_tmp7_i_us = new LoadInst(ptr_arrayidx6_i_us, "tmp7.i.us", false, label_forbody_i_us); - ExtractElementInst* float_tmp9_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_26, "tmp9.i.us", label_forbody_i_us); - CastInst* double_conv_i_us = new FPExtInst(float_tmp9_i_us, Type::DoubleTy, "conv.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_28, "tmp11.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp9_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_36, "tmp9.i.us", label_forbody_i_us); + CastInst* double_conv_i2_us = new FPExtInst(float_tmp9_i_us, Type::DoubleTy, "conv.i2.us", label_forbody_i_us); + ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_38, "tmp11.i.us", label_forbody_i_us); CastInst* double_conv12_i_us = new FPExtInst(float_tmp11_i_us, Type::DoubleTy, "conv12.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp14_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_29, "tmp14.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp14_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_39, "tmp14.i.us", label_forbody_i_us); CastInst* double_conv15_i_us = new FPExtInst(float_tmp14_i_us, Type::DoubleTy, "conv15.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp17_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_30, "tmp17.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp17_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_40, "tmp17.i.us", label_forbody_i_us); CastInst* double_conv18_i_us = new FPExtInst(float_tmp17_i_us, Type::DoubleTy, "conv18.i.us", label_forbody_i_us); - std::vector int32_call_i_us_params; - int32_call_i_us_params.push_back(const_ptr_31); - int32_call_i_us_params.push_back(double_conv_i_us); - int32_call_i_us_params.push_back(double_conv12_i_us); - int32_call_i_us_params.push_back(double_conv15_i_us); - int32_call_i_us_params.push_back(double_conv18_i_us); - CallInst* int32_call_i_us = new CallInst(func_printf, int32_call_i_us_params.begin(), int32_call_i_us_params.end(), "call.i.us", label_forbody_i_us); - int32_call_i_us->setCallingConv(CallingConv::C); - int32_call_i_us->setTailCall(false); - new StoreInst(float_tmp9_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); + std::vector int32_call_i3_us_params; + int32_call_i3_us_params.push_back(const_ptr_43); + int32_call_i3_us_params.push_back(double_conv_i2_us); + int32_call_i3_us_params.push_back(double_conv12_i_us); + int32_call_i3_us_params.push_back(double_conv15_i_us); + int32_call_i3_us_params.push_back(double_conv18_i_us); + CallInst* int32_call_i3_us = new CallInst(func_printf, int32_call_i3_us_params.begin(), int32_call_i3_us_params.end(), "call.i3.us", label_forbody_i_us); + int32_call_i3_us->setCallingConv(CallingConv::C); + int32_call_i3_us->setTailCall(false); + StoreInst* void_119 = new StoreInst(float_tmp9_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx24_i_us_indices; - ptr_arrayidx24_i_us_indices.push_back(int32_i_012_0_us); + ptr_arrayidx24_i_us_indices.push_back(int32_i_016_0_us); ptr_arrayidx24_i_us_indices.push_back(int32_i_01_0_i_us); - ptr_arrayidx24_i_us_indices.push_back(const_int32_28); - Instruction* ptr_arrayidx24_i_us = new GetElementPtrInst(ptr_dests_77, ptr_arrayidx24_i_us_indices.begin(), ptr_arrayidx24_i_us_indices.end(), "arrayidx24.i.us", label_forbody_i_us); - new StoreInst(float_tmp11_i_us, ptr_arrayidx24_i_us, false, label_forbody_i_us); + ptr_arrayidx24_i_us_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx24_i_us = new GetElementPtrInst(ptr_dests_93, ptr_arrayidx24_i_us_indices.begin(), ptr_arrayidx24_i_us_indices.end(), "arrayidx24.i.us", label_forbody_i_us); + StoreInst* void_120 = new StoreInst(float_tmp11_i_us, ptr_arrayidx24_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx28_i_us_indices; - ptr_arrayidx28_i_us_indices.push_back(int32_i_012_0_us); + ptr_arrayidx28_i_us_indices.push_back(int32_i_016_0_us); ptr_arrayidx28_i_us_indices.push_back(int32_i_01_0_i_us); - ptr_arrayidx28_i_us_indices.push_back(const_int32_29); - Instruction* ptr_arrayidx28_i_us = new GetElementPtrInst(ptr_dests_77, ptr_arrayidx28_i_us_indices.begin(), ptr_arrayidx28_i_us_indices.end(), "arrayidx28.i.us", label_forbody_i_us); - new StoreInst(float_tmp14_i_us, ptr_arrayidx28_i_us, false, label_forbody_i_us); + ptr_arrayidx28_i_us_indices.push_back(const_int32_39); + Instruction* ptr_arrayidx28_i_us = new GetElementPtrInst(ptr_dests_93, ptr_arrayidx28_i_us_indices.begin(), ptr_arrayidx28_i_us_indices.end(), "arrayidx28.i.us", label_forbody_i_us); + StoreInst* void_121 = new StoreInst(float_tmp14_i_us, ptr_arrayidx28_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx32_i_us_indices; - ptr_arrayidx32_i_us_indices.push_back(int32_i_012_0_us); + ptr_arrayidx32_i_us_indices.push_back(int32_i_016_0_us); ptr_arrayidx32_i_us_indices.push_back(int32_i_01_0_i_us); - ptr_arrayidx32_i_us_indices.push_back(const_int32_30); - Instruction* ptr_arrayidx32_i_us = new GetElementPtrInst(ptr_dests_77, ptr_arrayidx32_i_us_indices.begin(), ptr_arrayidx32_i_us_indices.end(), "arrayidx32.i.us", label_forbody_i_us); - new StoreInst(float_tmp17_i_us, ptr_arrayidx32_i_us, false, label_forbody_i_us); - BinaryOperator* int32_indvar_next_108 = BinaryOperator::create(Instruction::Add, int32_i_01_0_i_us, const_int32_28, "indvar.next", label_forbody_i_us); - ICmpInst* int1_exitcond_109 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_108, int32_num_attribs_79, "exitcond", label_forbody_i_us); - new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_109, label_forbody_i_us); - - // Block forbody (label_forbody_81) - Argument* fwdref_111 = new Argument(IntegerType::get(32)); - PHINode* int32_i_012_0 = new PHINode(IntegerType::get(32), "i.012.0", label_forbody_81); - int32_i_012_0->reserveOperandSpace(2); - int32_i_012_0->addIncoming(const_int32_26, label_forbody_preheader); - int32_i_012_0->addIncoming(fwdref_111, label_forbody_81); - - std::vector ptr_arraydecay9_indices; - ptr_arraydecay9_indices.push_back(const_int32_26); - ptr_arraydecay9_indices.push_back(int32_i_012_0); - ptr_arraydecay9_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay9 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay9_indices.begin(), ptr_arraydecay9_indices.end(), "arraydecay9", label_forbody_81); - std::vector ptr_arraydecay13_indices; - ptr_arraydecay13_indices.push_back(const_int32_26); - ptr_arraydecay13_indices.push_back(int32_i_012_0); - ptr_arraydecay13_indices.push_back(const_int32_26); - Instruction* ptr_arraydecay13 = new GetElementPtrInst(ptr_results, ptr_arraydecay13_indices.begin(), ptr_arraydecay13_indices.end(), "arraydecay13", label_forbody_81); - std::vector void_112_params; - void_112_params.push_back(ptr_arraydecay13); - void_112_params.push_back(ptr_arraydecay9); - void_112_params.push_back(ptr_arrayidx34_i8); - CallInst* void_112 = new CallInst(func_execute_shader, void_112_params.begin(), void_112_params.end(), "", label_forbody_81); - void_112->setCallingConv(CallingConv::C); - void_112->setTailCall(false); - BinaryOperator* int32_inc_113 = BinaryOperator::create(Instruction::Add, int32_i_012_0, const_int32_28, "inc", label_forbody_81); - ICmpInst* int1_cmp18 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_113, int32_count_78, "cmp18", label_forbody_81); - new BranchInst(label_forbody_81, label_afterfor_82, int1_cmp18, label_forbody_81); - - // Block afterfor (label_afterfor_82) - new ReturnInst(label_afterfor_82); + ptr_arrayidx32_i_us_indices.push_back(const_int32_40); + Instruction* ptr_arrayidx32_i_us = new GetElementPtrInst(ptr_dests_93, ptr_arrayidx32_i_us_indices.begin(), ptr_arrayidx32_i_us_indices.end(), "arrayidx32.i.us", label_forbody_i_us); + StoreInst* void_122 = new StoreInst(float_tmp17_i_us, ptr_arrayidx32_i_us, false, label_forbody_i_us); + BinaryOperator* int32_indvar_next_123 = BinaryOperator::create(Instruction::Add, int32_i_01_0_i_us, const_int32_38, "indvar.next", label_forbody_i_us); + ICmpInst* int1_exitcond_124 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_123, int32_num_attribs_94, "exitcond", label_forbody_i_us); + new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_124, label_forbody_i_us); + + // Block forbody (label_forbody_96) + Argument* fwdref_126 = new Argument(IntegerType::get(32)); + PHINode* int32_i_016_0 = new PHINode(IntegerType::get(32), "i.016.0", label_forbody_96); + int32_i_016_0->reserveOperandSpace(2); + int32_i_016_0->addIncoming(const_int32_36, label_forbody_preheader); + int32_i_016_0->addIncoming(fwdref_126, label_forbody_96); + + std::vector ptr_arraydecay14_indices; + ptr_arraydecay14_indices.push_back(const_int32_36); + ptr_arraydecay14_indices.push_back(int32_i_016_0); + ptr_arraydecay14_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_96); + std::vector ptr_arraydecay18_indices; + ptr_arraydecay18_indices.push_back(const_int32_36); + ptr_arraydecay18_indices.push_back(int32_i_016_0); + ptr_arraydecay18_indices.push_back(const_int32_36); + Instruction* ptr_arraydecay18 = new GetElementPtrInst(ptr_results, ptr_arraydecay18_indices.begin(), ptr_arraydecay18_indices.end(), "arraydecay18", label_forbody_96); + std::vector void_127_params; + void_127_params.push_back(ptr_arraydecay18); + void_127_params.push_back(ptr_arraydecay14); + void_127_params.push_back(ptr_arraydecay21); + CallInst* void_127 = new CallInst(func_execute_shader, void_127_params.begin(), void_127_params.end(), "", label_forbody_96); + void_127->setCallingConv(CallingConv::C); + void_127->setTailCall(false); + CallInst* int32_call22 = new CallInst(func_printf, const_ptr_46, "call22", label_forbody_96); + int32_call22->setCallingConv(CallingConv::C); + int32_call22->setTailCall(false); + BinaryOperator* int32_inc_128 = BinaryOperator::create(Instruction::Add, int32_i_016_0, const_int32_38, "inc", label_forbody_96); + ICmpInst* int1_cmp22 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_128, int32_num_vertices, "cmp22", label_forbody_96); + new BranchInst(label_forbody_96, label_afterfor_97, int1_cmp22, label_forbody_96); + + // Block afterfor (label_afterfor_97) + new ReturnInst(label_afterfor_97); // Resolve Forward References - fwdref_86->replaceAllUsesWith(int32_i_013_0_i_ph); delete fwdref_86; - fwdref_88->replaceAllUsesWith(packed_tmp48_i); delete fwdref_88; - fwdref_87->replaceAllUsesWith(int32_inc_i); delete fwdref_87; - fwdref_85->replaceAllUsesWith(int32_inc59_i); delete fwdref_85; - fwdref_99->replaceAllUsesWith(int32_inc_us); delete fwdref_99; - fwdref_103->replaceAllUsesWith(int32_indvar_next_108); delete fwdref_103; - fwdref_111->replaceAllUsesWith(int32_inc_113); delete fwdref_111; + fwdref_101->replaceAllUsesWith(int32_i_014_0_i_ph); delete fwdref_101; + fwdref_103->replaceAllUsesWith(packed_tmp48_i); delete fwdref_103; + fwdref_102->replaceAllUsesWith(int32_inc_i); delete fwdref_102; + fwdref_100->replaceAllUsesWith(int32_inc72_i); delete fwdref_100; + fwdref_109->replaceAllUsesWith(packed_tmp31_i); delete fwdref_109; + fwdref_108->replaceAllUsesWith(int32_indvar_next25); delete fwdref_108; + fwdref_114->replaceAllUsesWith(int32_inc_us); delete fwdref_114; + fwdref_118->replaceAllUsesWith(int32_indvar_next_123); delete fwdref_118; + fwdref_126->replaceAllUsesWith(int32_inc_128); delete fwdref_126; } diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index e135b4781d..82ae59446e 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -102,14 +102,14 @@ void from_array(float4 (*res)[16], float (*ainputs)[16][4], vec.y = ainputs[i][j][1]; vec.z = ainputs[i][j][2]; vec.w = ainputs[i][j][3]; - //printf("FAR(%d %d) %f %f %f %f\n", i, j, vec.x, vec.y, vec.z, vec.w); + printf("FRA(%d %d) %f %f %f %f\n", i, j, vec.x, vec.y, vec.z, vec.w); res[i][j] = vec; } } } void from_consts(float4 *res, float (*ainputs)[4], - int count) + int count) { for (int i = 0; i < count; ++i) { float4 vec; @@ -117,7 +117,7 @@ void from_consts(float4 *res, float (*ainputs)[4], vec.y = ainputs[i][1]; vec.z = ainputs[i][2]; vec.w = ainputs[i][3]; - //printf("VCONST %f %f %f %f\n", vec.x, vec.y, vec.z, vec.w); + printf("VCONST(%d) %f %f %f %f\n", i, vec.x, vec.y, vec.z, vec.w); res[i] = vec; } } @@ -141,20 +141,25 @@ extern void execute_shader(float4 *dests, float4 *inputs, void run_vertex_shader(float (*ainputs)[16][4], float (*dests)[16][4], float (*aconsts)[4], - int count, - int num_attribs) + int num_vertices, + int num_inputs, + int num_attribs, + int num_consts) { float4 inputs[16*32*4][16]; - float4 consts[16]; + float4 consts[32]; float4 results[16*32*4][16]; - printf("XXXXXXXXXXX run_vertex_shader\n"); - from_array(inputs, ainputs, count, num_attribs); - from_consts(consts, aconsts, 5); - for (int i = 0; i < count; ++i) { + printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, consts = %d\n", + num_vertices, num_inputs, num_consts); + from_array(inputs, ainputs, num_vertices, num_inputs); + from_consts(consts, aconsts, num_consts); + printf(" after conversion\n"); + for (int i = 0; i < num_vertices; ++i) { float4 *in = inputs[i]; float4 *res = results[i]; execute_shader(res, in, consts); + printf("after executing shader\n"); to_array(dests[i], res, num_attribs); } } diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 9b4a5f024e..2a86a4e442 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -125,6 +125,10 @@ translate_instruction(llvm::Module *module, struct tgsi_full_instruction *fi) { llvm::Value *inputs[4]; + inputs[0] = 0; + inputs[1] = 0; + inputs[2] = 0; + inputs[3] = 0; printf("translate instr START\n"); for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; @@ -168,7 +172,10 @@ translate_instruction(llvm::Module *module, inputs[i] = val; } + if (inputs[0]) + instr->printVector(inputs[0]); llvm::Value *out = 0; + printf("Opcode is %d\n", inst->Instruction.Opcode); switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: break; @@ -512,32 +519,19 @@ translate_instruction(llvm::Module *module, if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { printf("--- storing to %d %p\n", dst->DstRegister.Index, out); - storage->store(dst->DstRegister.Index, out); + storage->store(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->setTempElement(dst->DstRegister.Index, out); + storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else { fprintf(stderr, "ERROR: unsupported LLVM destination!"); } - -#if 0 - if (dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW) { - if (dst->DstRegister.WriteMask & TGSI_WRITEMASK_X) { - } - if (dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y) { - } - if (dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z) { - } - if (dst->DstRegister.WriteMask & TGSI_WRITEMASK_W) { - } - } -#endif } printf("translate instr END\n"); } static llvm::Module * -tgsi_to_llvm(const struct tgsi_token *tokens) +tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) { llvm::Module *mod = createBaseShader(); struct tgsi_parse_context parse; @@ -572,7 +566,7 @@ tgsi_to_llvm(const struct tgsi_token *tokens) fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); Storage storage(label_entry, ptr_OUT, ptr_IN, ptr_CONST); - Instructions instr(mod, label_entry); + Instructions instr(mod, shader, label_entry); while(!tgsi_parse_end_of_tokens(&parse)) { tgsi_parse_token(&parse); @@ -601,178 +595,14 @@ tgsi_to_llvm(const struct tgsi_token *tokens) } } -#if 0 - // Type Definitions - ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 19); - - PointerType* PointerTy_1 = PointerType::get(ArrayTy_0); - - VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); - - PointerType* PointerTy_3 = PointerType::get(VectorTy_4); - - - VectorType* VectorTy_5 = VectorType::get(IntegerType::get(32), 4); - - - PointerType* PointerTy_8 = PointerType::get(IntegerType::get(8)); - - - // Global Variable Declarations - - - GlobalVariable* gvar_array__str = new GlobalVariable( - /*Type=*/ArrayTy_0, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str", - mod); - - GlobalVariable* gvar_array__str1 = new GlobalVariable( - /*Type=*/ArrayTy_0, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str1", - mod); - - GlobalVariable* gvar_array__str2 = new GlobalVariable( - /*Type=*/ArrayTy_0, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str2", - mod); - - // Constant Definitions - Constant* const_array_9 = ConstantArray::get("const %f %f %f %f\x0A", true); - Constant* const_array_10 = ConstantArray::get("resul %f %f %f %f\x0A", true); - Constant* const_array_11 = ConstantArray::get("outpu %f %f %f %f\x0A", true); - UndefValue* const_packed_12 = UndefValue::get(VectorTy_4); - Constant* const_packed_13 = Constant::getNullValue(VectorTy_5); - std::vector const_packed_14_elems; - ConstantInt* const_int32_15 = ConstantInt::get(APInt(32, "1", 10)); - const_packed_14_elems.push_back(const_int32_15); - const_packed_14_elems.push_back(const_int32_15); - const_packed_14_elems.push_back(const_int32_15); - const_packed_14_elems.push_back(const_int32_15); - Constant* const_packed_14 = ConstantVector::get(VectorTy_5, const_packed_14_elems); - std::vector const_packed_16_elems; - ConstantInt* const_int32_17 = ConstantInt::get(APInt(32, "2", 10)); - const_packed_16_elems.push_back(const_int32_17); - const_packed_16_elems.push_back(const_int32_17); - const_packed_16_elems.push_back(const_int32_17); - const_packed_16_elems.push_back(const_int32_17); - Constant* const_packed_16 = ConstantVector::get(VectorTy_5, const_packed_16_elems); - std::vector const_packed_18_elems; - ConstantInt* const_int32_19 = ConstantInt::get(APInt(32, "3", 10)); - const_packed_18_elems.push_back(const_int32_19); - const_packed_18_elems.push_back(const_int32_19); - const_packed_18_elems.push_back(const_int32_19); - const_packed_18_elems.push_back(const_int32_19); - Constant* const_packed_18 = ConstantVector::get(VectorTy_5, const_packed_18_elems); - std::vector const_ptr_20_indices; - Constant* const_int32_21 = Constant::getNullValue(IntegerType::get(32)); - const_ptr_20_indices.push_back(const_int32_21); - const_ptr_20_indices.push_back(const_int32_21); - Constant* const_ptr_20 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_20_indices[0], const_ptr_20_indices.size() ); - UndefValue* const_double_22 = UndefValue::get(Type::DoubleTy); - std::vector const_ptr_23_indices; - const_ptr_23_indices.push_back(const_int32_21); - const_ptr_23_indices.push_back(const_int32_21); - Constant* const_ptr_23 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_23_indices[0], const_ptr_23_indices.size() ); - std::vector const_ptr_24_indices; - const_ptr_24_indices.push_back(const_int32_21); - const_ptr_24_indices.push_back(const_int32_21); - Constant* const_ptr_24 = ConstantExpr::getGetElementPtr(gvar_array__str2, &const_ptr_24_indices[0], const_ptr_24_indices.size() ); - - // Global Variable Definitions - gvar_array__str->setInitializer(const_array_9); - gvar_array__str1->setInitializer(const_array_10); - gvar_array__str2->setInitializer(const_array_11); - - // Function Definitions - - // Function: execute_shader (func_execute_shader) - { - // Block entry (label_entry) - LoadInst* packed_tmp1 = new LoadInst(ptr_IN, "tmp1", false, label_entry); - ShuffleVectorInst* packed_tmp3 = new ShuffleVectorInst(packed_tmp1, const_packed_12, const_packed_13, "tmp3", label_entry); - LoadInst* packed_tmp6 = new LoadInst(ptr_CONST, "tmp6", false, label_entry); - BinaryOperator* packed_mul = BinaryOperator::create(Instruction::Mul, packed_tmp3, packed_tmp6, "mul", label_entry); - ShuffleVectorInst* packed_tmp8 = new ShuffleVectorInst(packed_tmp1, const_packed_12, const_packed_14, "tmp8", label_entry); - GetElementPtrInst* ptr_arrayidx10 = new GetElementPtrInst(ptr_CONST, const_int32_15, "arrayidx10", label_entry); - LoadInst* packed_tmp11 = new LoadInst(ptr_arrayidx10, "tmp11", false, label_entry); - BinaryOperator* packed_mul12 = BinaryOperator::create(Instruction::Mul, packed_tmp8, packed_tmp11, "mul12", label_entry); - BinaryOperator* packed_add = BinaryOperator::create(Instruction::Add, packed_mul12, packed_mul, "add", label_entry); - ShuffleVectorInst* packed_tmp15 = new ShuffleVectorInst(packed_tmp1, const_packed_12, const_packed_16, "tmp15", label_entry); - GetElementPtrInst* ptr_arrayidx17 = new GetElementPtrInst(ptr_CONST, const_int32_17, "arrayidx17", label_entry); - LoadInst* packed_tmp18 = new LoadInst(ptr_arrayidx17, "tmp18", false, label_entry); - BinaryOperator* packed_mul19 = BinaryOperator::create(Instruction::Mul, packed_tmp15, packed_tmp18, "mul19", label_entry); - BinaryOperator* packed_add21 = BinaryOperator::create(Instruction::Add, packed_mul19, packed_add, "add21", label_entry); - ShuffleVectorInst* packed_tmp25 = new ShuffleVectorInst(packed_tmp1, const_packed_12, const_packed_18, "tmp25", label_entry); - GetElementPtrInst* ptr_arrayidx27 = new GetElementPtrInst(ptr_CONST, const_int32_19, "arrayidx27", label_entry); - LoadInst* packed_tmp28 = new LoadInst(ptr_arrayidx27, "tmp28", false, label_entry); - BinaryOperator* packed_mul29 = BinaryOperator::create(Instruction::Mul, packed_tmp25, packed_tmp28, "mul29", label_entry); - BinaryOperator* packed_add31 = BinaryOperator::create(Instruction::Add, packed_mul29, packed_add21, "add31", label_entry); - StoreInst* void_25 = new StoreInst(packed_add31, ptr_OUT, false, label_entry); - GetElementPtrInst* ptr_arrayidx33 = new GetElementPtrInst(ptr_OUT, const_int32_15, "arrayidx33", label_entry); - GetElementPtrInst* ptr_arrayidx35 = new GetElementPtrInst(ptr_IN, const_int32_15, "arrayidx35", label_entry); - LoadInst* packed_tmp36 = new LoadInst(ptr_arrayidx35, "tmp36", false, label_entry); - StoreInst* void_26 = new StoreInst(packed_tmp36, ptr_arrayidx33, false, label_entry); - std::vector int32_call_params; - int32_call_params.push_back(const_ptr_20); - int32_call_params.push_back(const_double_22); - int32_call_params.push_back(const_double_22); - int32_call_params.push_back(const_double_22); - int32_call_params.push_back(const_double_22); - //CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry); - //int32_call->setCallingConv(CallingConv::C); - //int32_call->setTailCall(true); - ExtractElementInst* float_tmp52 = new ExtractElementInst(packed_tmp1, const_int32_21, "tmp52", label_entry); - CastInst* double_conv53 = new FPExtInst(float_tmp52, Type::DoubleTy, "conv53", label_entry); - ExtractElementInst* float_tmp55 = new ExtractElementInst(packed_tmp1, const_int32_15, "tmp55", label_entry); - CastInst* double_conv56 = new FPExtInst(float_tmp55, Type::DoubleTy, "conv56", label_entry); - ExtractElementInst* float_tmp58 = new ExtractElementInst(packed_tmp1, const_int32_17, "tmp58", label_entry); - CastInst* double_conv59 = new FPExtInst(float_tmp58, Type::DoubleTy, "conv59", label_entry); - ExtractElementInst* float_tmp61 = new ExtractElementInst(packed_tmp1, const_int32_19, "tmp61", label_entry); - CastInst* double_conv62 = new FPExtInst(float_tmp61, Type::DoubleTy, "conv62", label_entry); - std::vector int32_call63_params; - int32_call63_params.push_back(const_ptr_23); - int32_call63_params.push_back(double_conv53); - int32_call63_params.push_back(double_conv56); - int32_call63_params.push_back(double_conv59); - int32_call63_params.push_back(double_conv62); - //CallInst* int32_call63 = new CallInst(func_printf, int32_call63_params.begin(), int32_call63_params.end(), "call63", label_entry); - //int32_call63->setCallingConv(CallingConv::C); - //int32_call63->setTailCall(true); - ExtractElementInst* float_tmp65 = new ExtractElementInst(packed_add31, const_int32_21, "tmp65", label_entry); - CastInst* double_conv66 = new FPExtInst(float_tmp65, Type::DoubleTy, "conv66", label_entry); - ExtractElementInst* float_tmp68 = new ExtractElementInst(packed_add31, const_int32_15, "tmp68", label_entry); - CastInst* double_conv69 = new FPExtInst(float_tmp68, Type::DoubleTy, "conv69", label_entry); - ExtractElementInst* float_tmp71 = new ExtractElementInst(packed_add31, const_int32_17, "tmp71", label_entry); - CastInst* double_conv72 = new FPExtInst(float_tmp71, Type::DoubleTy, "conv72", label_entry); - ExtractElementInst* float_tmp74 = new ExtractElementInst(packed_add31, const_int32_19, "tmp74", label_entry); - CastInst* double_conv75 = new FPExtInst(float_tmp74, Type::DoubleTy, "conv75", label_entry); - std::vector int32_call76_params; - int32_call76_params.push_back(const_ptr_24); - int32_call76_params.push_back(double_conv66); - int32_call76_params.push_back(double_conv69); - int32_call76_params.push_back(double_conv72); - int32_call76_params.push_back(double_conv75); - //CallInst* int32_call76 = new CallInst(func_printf, int32_call76_params.begin(), int32_call76_params.end(), "call76", label_entry); - //int32_call76->setCallingConv(CallingConv::C); - //int32_call76->setTailCall(true); - } -#endif - new ReturnInst(label_entry); //TXT("\ntgsi-dump end -------------------\n"); tgsi_parse_free(&parse); + prog->num_consts = storage.numConsts(); + std::cout<<"XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"<(prog->function); - runner(inputs, dests, consts, count, num_attribs); + runner(inputs, dests, consts, num_vertices, num_inputs, + num_attribs, prog->num_consts); std::cout << "---- END LLVM Execution "< #include #include @@ -16,7 +17,8 @@ Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, llvm::Value *in, llvm::Value *consts) : m_block(block), m_OUT(out), m_IN(in), m_CONST(consts), - m_temps(32) + m_temps(32), m_dstCache(32), + m_idx(0) { m_floatVecType = VectorType::get(Type::FloatTy, 4); m_intVecType = VectorType::get(IntegerType::get(32), 4); @@ -25,6 +27,7 @@ Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, m_undefIntVec = UndefValue::get(m_intVecType); m_shuffleId = 0; + m_numConsts = 0; } //can only build vectors with all members in the [0, 9] range @@ -69,15 +72,11 @@ llvm::Value *Storage::inputElement(int idx) if (m_inputs.find(idx) != m_inputs.end()) { return m_inputs[idx]; } - char ptrName[13]; - char name[9]; - snprintf(ptrName, 13, "input_ptr%d", idx); - snprintf(name, 9, "input%d", idx); GetElementPtrInst *getElem = new GetElementPtrInst(m_IN, constantInt(idx), - ptrName, + name("input_ptr"), m_block); - LoadInst *load = new LoadInst(getElem, name, + LoadInst *load = new LoadInst(getElem, name("input"), false, m_block); m_inputs[idx] = load; return load; @@ -85,6 +84,7 @@ llvm::Value *Storage::inputElement(int idx) llvm::Value *Storage::constElement(int idx) { + m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; if (m_consts.find(idx) != m_consts.end()) { return m_consts[idx]; } @@ -123,18 +123,70 @@ llvm::Value *Storage::tempElement(int idx) const return ret; } -void Storage::setTempElement(int idx, llvm::Value *val) +void Storage::setTempElement(int idx, llvm::Value *val, int mask) { + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = m_temps[idx]; + val = maskWrite(val, mask, templ); + } m_temps[idx] = val; } -void Storage::store(int dstIdx, llvm::Value *val) +void Storage::store(int dstIdx, llvm::Value *val, int mask) { - char ptrName[13]; - snprintf(ptrName, 13, "out_ptr%d", dstIdx); + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = m_dstCache[dstIdx]; + val = maskWrite(val, mask, templ); + } + GetElementPtrInst *getElem = new GetElementPtrInst(m_OUT, constantInt(dstIdx), - ptrName, + name("out_ptr"), m_block); - new StoreInst(val, getElem, false, m_block); + StoreInst *st = new StoreInst(val, getElem, false, m_block); + //m_dstCache[dstIdx] = st; +} + +llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ) +{ + llvm::Value *dst = templ; + if (!dst) + dst = Constant::getNullValue(m_floatVecType); + if ((mask & TGSI_WRITEMASK_X)) { + llvm::Value *x = new ExtractElementInst(src, unsigned(0), + name("x"), m_block); + dst = new InsertElementInst(dst, x, unsigned(0), + name("dstx"), m_block); + } + if ((mask & TGSI_WRITEMASK_Y)) { + llvm::Value *y = new ExtractElementInst(src, unsigned(1), + name("y"), m_block); + dst = new InsertElementInst(dst, y, unsigned(1), + name("dsty"), m_block); + } + if ((mask & TGSI_WRITEMASK_Z)) { + llvm::Value *z = new ExtractElementInst(src, unsigned(2), + name("z"), m_block); + dst = new InsertElementInst(dst, z, unsigned(2), + name("dstz"), m_block); + } + if ((mask & TGSI_WRITEMASK_W)) { + llvm::Value *w = new ExtractElementInst(src, unsigned(3), + name("w"), m_block); + dst = new InsertElementInst(dst, w, unsigned(3), + name("dstw"), m_block); + } + return dst; +} + +const char * Storage::name(const char *prefix) +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +int Storage::numConsts() const +{ + return m_numConsts; } diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index b69c8d614e..f39d308fdd 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -27,12 +27,17 @@ public: llvm::Value *constElement(int idx); llvm::Value *tempElement(int idx) const; - void setTempElement(int idx, llvm::Value *val); + void setTempElement(int idx, llvm::Value *val, int mask); llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); - void store(int dstIdx, llvm::Value *val); + void store(int dstIdx, llvm::Value *val, int mask); + + int numConsts() const; +private: + llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); + const char *name(const char *prefix); private: llvm::BasicBlock *m_block; llvm::Value *m_OUT; @@ -42,6 +47,7 @@ private: std::map m_constInts; std::map m_intVecs; std::vector m_temps; + std::vector m_dstCache; LoadMap m_inputs; LoadMap m_consts; @@ -52,6 +58,10 @@ private: llvm::Value *m_undefIntVec; int m_shuffleId; + char m_name[32]; + int m_idx; + + int m_numConsts; }; #endif -- cgit v1.2.3 From 3b772a277fe032c94654648d0671ff42fd3fdce6 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sat, 20 Oct 2007 16:59:37 -0400 Subject: Make lit work but in a very lame way - will have to be rewritten. --- src/mesa/pipe/llvm/instructions.cpp | 9 ++------- src/mesa/pipe/llvm/instructions.h | 1 - src/mesa/pipe/llvm/llvmtgsi.cpp | 3 +-- 3 files changed, 3 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 3105c49bcc..5a40888696 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -167,7 +167,7 @@ llvm::Value *Instructions::callFAbs(llvm::Value *val) llvm::Value * Instructions::lit(llvm::Value *in) { -#if 1 +#if 0 printVector(in); return in; @@ -191,7 +191,7 @@ llvm::Value * Instructions::lit(llvm::Value *in) m_llvmLit = makeLitFunction(m_mod); } CallInst *call = new CallInst(m_llvmLit, in, name("litres"), m_block); - //call->setCallingConv(CallingConv::C); + call->setCallingConv(CallingConv::C); call->setTailCall(false); return call; #endif @@ -897,8 +897,3 @@ void Instructions::printVector(llvm::Value *val) call->setCallingConv(CallingConv::C); call->setTailCall(true); } - -llvm::Value * Instructions::swizzleWrite(llvm::Value *in, int mask) -{ - -} diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index bb71cc8520..87203fdc10 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -34,7 +34,6 @@ public: llvm::Value *rcp(llvm::Value *in); llvm::Value *rsq(llvm::Value *in); llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); - llvm::Value *swizzleWrite(llvm::Value *in, int mask); void printVector(llvm::Value *val); private: diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 2a86a4e442..049bf0fa6e 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -184,8 +184,7 @@ translate_instruction(llvm::Module *module, } break; case TGSI_OPCODE_LIT: { - //out = instr->lit(inputs[0]); - return; + out = instr->lit(inputs[0]); } break; case TGSI_OPCODE_RCP: { -- cgit v1.2.3 From 6815a9af6c98934e3537507ccbf077d69a5d5320 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sun, 21 Oct 2007 11:16:01 -0400 Subject: Cleanup some of the debugging output code --- src/mesa/Makefile | 2 +- src/mesa/pipe/draw/draw_vertex_shader.c | 5 ---- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 11 +------- src/mesa/pipe/llvm/instructions.cpp | 21 --------------- src/mesa/pipe/llvm/llvmtgsi.cpp | 40 ++++------------------------ src/mesa/pipe/llvm/storage.cpp | 14 +++------- src/mesa/pipe/llvm/storage.h | 1 - 7 files changed, 10 insertions(+), 84 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 5b56f76e0b..e9dfb95e5c 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -113,7 +113,7 @@ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) - $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ + @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index e36ecdc849..fb20dfa4e1 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -91,8 +91,6 @@ run_vertex_program(struct draw_context *draw, assert(draw->vertex_shader->state->output_semantic_name[0] == TGSI_SEMANTIC_POSITION); - fprintf(stderr, "------ run_vertex\n"); - /* Consts does not require 16 byte alignment. */ machine->Consts = (float (*)[4]) draw->mapped_constants; @@ -132,7 +130,6 @@ run_vertex_program(struct draw_context *draw, y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - printf("output %d: %f %f %f %f\n", 0, x, y, z, w); vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); vOut[j]->edgeflag = 1; @@ -178,8 +175,6 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) { unsigned i, j; - fprintf(stderr, "XX q(%d) ", draw->vs.queue_nr ); - if (draw->vertex_shader->state->llvm_prog) { draw_vertex_shader_queue_flush_llvm(draw); return; diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index d658544f30..05c3fe511b 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -74,8 +74,6 @@ void vertex_fetch(struct draw_context *draw, { uint attr; - printf("fetch vertex %u: \n", elt); - /* loop over vertex attributes (vertex shader inputs) */ for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { @@ -89,8 +87,6 @@ void vertex_fetch(struct draw_context *draw, fetch_attrib4(src, draw->vertex_element[attr].src_format, p); - printf("> %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]); - inputs[attr][0] = p[0]; /*X*/ inputs[attr][1] = p[1]; /*Y*/ inputs[attr][2] = p[2]; /*Z*/ @@ -105,7 +101,7 @@ compute_clipmask(const float *clip, const float (*plane)[4], unsigned nr) unsigned i; for (i = 0; i < nr; i++) { - if (dot4(clip, plane[i]) < 0) + if (dot4(clip, plane[i]) < 0) mask |= (1<viewport.scale; const float *trans = draw->viewport.translate; - fprintf(stderr, "--- XX q(%d) \n", draw->vs.queue_nr); - /* fetch the inputs */ for (i = 0; i < draw->vs.queue_nr; ++i) { unsigned elt = draw->vs.queue[i].elt; @@ -144,9 +138,6 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) draw->vertex_shader->state->num_inputs, draw->vertex_info.num_attribs); - /* FIXME: finish conversion */ - /* dests = outputs */ - /* store machine results */ for (int i = 0; i < draw->vs.queue_nr; ++i) { unsigned slot; diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 5a40888696..037bec653f 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -167,26 +167,6 @@ llvm::Value *Instructions::callFAbs(llvm::Value *val) llvm::Value * Instructions::lit(llvm::Value *in) { -#if 0 - printVector(in); - return in; - - ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), - name("x"), - m_block); - - ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), - name("y"), - m_block); - - ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), - name("w"), - m_block); - return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), - ConstantFP::get(Type::FloatTy, APFloat(0.f)), - ConstantFP::get(Type::FloatTy, APFloat(0.f)), - ConstantFP::get(Type::FloatTy, APFloat(1.f))); -#else if (!m_llvmLit) { m_llvmLit = makeLitFunction(m_mod); } @@ -194,7 +174,6 @@ llvm::Value * Instructions::lit(llvm::Value *in) call->setCallingConv(CallingConv::C); call->setTailCall(false); return call; -#endif } llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 049bf0fa6e..56b488fb1a 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -141,7 +141,6 @@ translate_instruction(llvm::Module *module, val = storage->tempElement(src->SrcRegister.Index); } else { fprintf(stderr, "ERROR: not support llvm source\n"); - printf("translate instr END\n"); return; } @@ -160,9 +159,6 @@ translate_instruction(llvm::Module *module, src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) { - fprintf(stderr, "SWIZZLE is %d %d %d %d\n", - src->SrcRegister.SwizzleX, src->SrcRegister.SwizzleY, - src->SrcRegister.SwizzleZ, src->SrcRegister.SwizzleW); int swizzle = src->SrcRegister.SwizzleX * 1000; swizzle += src->SrcRegister.SwizzleY * 100; swizzle += src->SrcRegister.SwizzleZ * 10; @@ -517,7 +513,6 @@ translate_instruction(llvm::Module *module, struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - printf("--- storing to %d %p\n", dst->DstRegister.Index, out); storage->store(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); @@ -552,16 +547,6 @@ tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) tgsi_parse_init(&parse, tokens); - Function* func_printf = mod->getFunction("printf"); - //parse.FullHeader.Processor.Processor - - //parse.FullVersion.Version.MajorVersion - //parse.FullVersion.Version.MinorVersion - - //parse.FullHeader.Header.HeaderSize - //parse.FullHeader.Header.BodySize - //parse.FullHeader.Processor.Processor - fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); Storage storage(label_entry, ptr_OUT, ptr_IN, ptr_CONST); @@ -569,8 +554,6 @@ tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) while(!tgsi_parse_end_of_tokens(&parse)) { tgsi_parse_token(&parse); - fprintf(stderr, "Translating %d\n", parse.FullToken.Token.Type); - switch (parse.FullToken.Token.Type) { case TGSI_TOKEN_TYPE_DECLARATION: translate_declaration(mod, @@ -596,8 +579,6 @@ tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) new ReturnInst(label_entry); - //TXT("\ntgsi-dump end -------------------\n"); - tgsi_parse_free(&parse); prog->num_consts = storage.numConsts(); @@ -611,25 +592,21 @@ tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) struct ga_llvm_prog * ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) { - std::cout << "Creating llvm " < Date: Mon, 22 Oct 2007 08:46:28 -0400 Subject: Align the load/stores. Reduce the amount of debugging output. --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 6 + src/mesa/pipe/llvm/instructions.cpp | 459 ++++------- src/mesa/pipe/llvm/instructions.h | 2 + src/mesa/pipe/llvm/llvm_base_shader.cpp | 1131 +++++++++++--------------- src/mesa/pipe/llvm/llvm_builtins.c | 13 +- src/mesa/pipe/llvm/llvmtgsi.cpp | 23 +- src/mesa/pipe/llvm/llvmtgsi.h | 1 + src/mesa/pipe/llvm/storage.cpp | 3 + 8 files changed, 665 insertions(+), 973 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 05c3fe511b..bcc44f7980 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -38,6 +38,8 @@ #include "pipe/llvm/llvmtgsi.h" #include "pipe/tgsi/exec/tgsi_core.h" +#define DBG 0 + static INLINE void fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) { @@ -155,7 +157,9 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) y = vOut->clip[1] = dests[0][1]; z = vOut->clip[2] = dests[0][2]; w = vOut->clip[3] = dests[0][3]; +#if DBG printf("output %d: %f %f %f %f\n", 0, x, y, z, w); +#endif vOut->clipmask = compute_clipmask(vOut->clip, draw->plane, draw->nr_planes); vOut->edgeflag = 1; @@ -183,11 +187,13 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) vOut->data[slot][2] = dests[slot][2]; vOut->data[slot][3] = dests[slot][3]; +#if DBG printf("output %d: %f %f %f %f\n", slot, vOut->data[slot][0], vOut->data[slot][1], vOut->data[slot][2], vOut->data[slot][3]); +#endif } } /* loop over vertices */ diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 037bec653f..044b447846 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -527,99 +527,153 @@ llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) return vectorFromVals(selx, sely, selz, selw); } +void Instructions::printVector(llvm::Value *val) +{ + static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A"; + + if (!m_fmtPtr) { + Constant *format = ConstantArray::get(frmt, true); + ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1); + GlobalVariable* globalFormat = new GlobalVariable( + /*Type=*/arrayTy, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name(".str"), + m_mod); + globalFormat->setInitializer(format); + + Constant* const_int0 = Constant::getNullValue(IntegerType::get(32)); + std::vector const_ptr_21_indices; + const_ptr_21_indices.push_back(const_int0); + const_ptr_21_indices.push_back(const_int0); + m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat, + &const_ptr_21_indices[0], const_ptr_21_indices.size()); + } + + Function *func_printf = m_mod->getFunction("printf"); + if (!func_printf) + func_printf = declarePrintf(); + assert(func_printf); + ExtractElementInst *x = new ExtractElementInst(val, unsigned(0), + name("extractx"), + m_block); + ExtractElementInst *y = new ExtractElementInst(val, unsigned(1), + name("extracty"), + m_block); + ExtractElementInst *z = new ExtractElementInst(val, unsigned(2), + name("extractz"), + m_block); + ExtractElementInst *w = new ExtractElementInst(val, unsigned(3), + name("extractw"), + m_block); + CastInst *dx = new FPExtInst(x, Type::DoubleTy, name("dx"), m_block); + CastInst *dy = new FPExtInst(y, Type::DoubleTy, name("dy"), m_block); + CastInst *dz = new FPExtInst(z, Type::DoubleTy, name("dz"), m_block); + CastInst *dw = new FPExtInst(w, Type::DoubleTy, name("dw"), m_block); + std::vector params; + params.push_back(m_fmtPtr); + params.push_back(dx); + params.push_back(dy); + params.push_back(dz); + params.push_back(dw); + CallInst* call = new CallInst(func_printf, params.begin(), params.end(), + name("printf"), m_block); + call->setCallingConv(CallingConv::C); + call->setTailCall(true); +} + +llvm::Function * Instructions::declarePrintf() +{ + std::vector args; + ParamAttrsList *params = 0; + FunctionType* funcTy = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/args, + /*isVarArg=*/true, + /*ParamAttrs=*/params); + Function* func_printf = new Function( + /*Type=*/funcTy, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", m_mod); + func_printf->setCallingConv(CallingConv::C); + return func_printf; +} + /* - Generated from: -extern float exp2f(float x); -extern float log2f(float x); +typedef __attribute__(( ocu_vector_type(4) )) float float4; + +extern float powf(float a, float b); + +inline float approx(float a, float b) +{ + if (b < -128.0f) b = -128.0f; + if (b > 128.0f) b = 128.0f; + if (a < 0) a = 0; + return powf(a, b); +} float4 lit(float4 tmp) { - tmp.w = (tmp.w < -128.0) ? -128.0f : ((tmp.w > 128.0f) ? 128.f : tmp.w); float4 result; result.x = 1.0; - result.y = tmp.x; - result.z = (tmp.x > 0) ? exp2f(tmp.w * log2f(tmp.y)) : 0.0; result.w = 1.0; + if (tmp.x > 0) { + result.y = tmp.x; + result.z = approx(tmp.y, tmp.w); + } else { + result.y = 0; + result.z = 0; + } return result; } with: -clang --emit-llvm lit.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-function -for=lit - +clang --emit-llvm lit.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -for=lit */ Function* makeLitFunction(Module *mod) { // Type Definitions -ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 27); - -PointerType* PointerTy_1 = PointerType::get(ArrayTy_0); - -ArrayType* ArrayTy_2 = ArrayType::get(IntegerType::get(8), 28); - -PointerType* PointerTy_3 = PointerType::get(ArrayTy_2); - -ArrayType* ArrayTy_4 = ArrayType::get(IntegerType::get(8), 8); - -PointerType* PointerTy_5 = PointerType::get(ArrayTy_4); - -ArrayType* ArrayTy_6 = ArrayType::get(IntegerType::get(8), 33); - -PointerType* PointerTy_7 = PointerType::get(ArrayTy_6); - -std::vectorFuncTy_8_args; -FuncTy_8_args.push_back(Type::FloatTy); -FuncTy_8_args.push_back(Type::FloatTy); -ParamAttrsList *FuncTy_8_PAL = 0; -FunctionType* FuncTy_8 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_8_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_8_PAL); - -std::vectorFuncTy_10_args; -ParamAttrsList *FuncTy_10_PAL = 0; -FunctionType* FuncTy_10 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_10_args, - /*isVarArg=*/true, - /*ParamAttrs=*/FuncTy_10_PAL); - -PointerType* PointerTy_9 = PointerType::get(FuncTy_10); - -PointerType* PointerTy_11 = PointerType::get(IntegerType::get(8)); - -PointerType* PointerTy_12 = PointerType::get(FuncTy_8); - -VectorType* VectorTy_13 = VectorType::get(Type::FloatTy, 4); - -std::vectorFuncTy_14_args; -FuncTy_14_args.push_back(VectorTy_13); -ParamAttrsList *FuncTy_14_PAL = 0; -FunctionType* FuncTy_14 = FunctionType::get( - /*Result=*/VectorTy_13, - /*Params=*/FuncTy_14_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_14_PAL); + std::vectorFuncTy_0_args; + FuncTy_0_args.push_back(Type::FloatTy); + FuncTy_0_args.push_back(Type::FloatTy); + ParamAttrsList *FuncTy_0_PAL = 0; + FunctionType* FuncTy_0 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_0_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_0_PAL); + + PointerType* PointerTy_1 = PointerType::get(FuncTy_0); + + VectorType* VectorTy_2 = VectorType::get(Type::FloatTy, 4); + + std::vectorFuncTy_3_args; + FuncTy_3_args.push_back(VectorTy_2); + ParamAttrsList *FuncTy_3_PAL = 0; + FunctionType* FuncTy_3 = FunctionType::get( + /*Result=*/VectorTy_2, + /*Params=*/FuncTy_3_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_3_PAL); // Function Declarations -Function* func_approx = new Function( - /*Type=*/FuncTy_8, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"approx", mod); -func_approx->setCallingConv(CallingConv::C); - -Function* func_printf = mod->getFunction("printf"); + Function* func_approx = new Function( + /*Type=*/FuncTy_0, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"approx", mod); + func_approx->setCallingConv(CallingConv::C); Function* func_powf = new Function( - /*Type=*/FuncTy_8, + /*Type=*/FuncTy_0, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"powf", mod); // (external, no body) func_powf->setCallingConv(CallingConv::C); Function* func_lit = new Function( - /*Type=*/FuncTy_14, + /*Type=*/FuncTy_3, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"lit", mod); func_lit->setCallingConv(CallingConv::C); @@ -627,86 +681,30 @@ func_lit->setCallingConv(CallingConv::C); // Global Variable Declarations -GlobalVariable* gvar_array__str = new GlobalVariable( -/*Type=*/ArrayTy_0, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str", -mod); - -GlobalVariable* gvar_array__str1 = new GlobalVariable( -/*Type=*/ArrayTy_2, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str1", -mod); - -GlobalVariable* gvar_array__str2 = new GlobalVariable( -/*Type=*/ArrayTy_4, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str2", -mod); - -GlobalVariable* gvar_array__str3 = new GlobalVariable( -/*Type=*/ArrayTy_6, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str3", -mod); - // Constant Definitions -Constant* const_array_15 = ConstantArray::get("After test with '%f' '%f'\x0A", true); -Constant* const_array_16 = ConstantArray::get("Calling pow with '%f' '%f'\x0A", true); -Constant* const_array_17 = ConstantArray::get("IN LIT\x0A", true); -Constant* const_array_18 = ConstantArray::get("About to approx with '%f' '%f'\x0A", true); -ConstantFP* const_float_19 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); -ConstantFP* const_float_20 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); -std::vector const_ptr_21_indices; -Constant* const_int32_22 = Constant::getNullValue(IntegerType::get(32)); -const_ptr_21_indices.push_back(const_int32_22); -const_ptr_21_indices.push_back(const_int32_22); -Constant* const_ptr_21 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_21_indices[0], const_ptr_21_indices.size() ); -Constant* const_float_23 = Constant::getNullValue(Type::FloatTy); -std::vector const_ptr_24_indices; -const_ptr_24_indices.push_back(const_int32_22); -const_ptr_24_indices.push_back(const_int32_22); -Constant* const_ptr_24 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_24_indices[0], const_ptr_24_indices.size() ); -std::vector const_ptr_25_indices; -const_ptr_25_indices.push_back(const_int32_22); -const_ptr_25_indices.push_back(const_int32_22); -Constant* const_ptr_25 = ConstantExpr::getGetElementPtr(gvar_array__str2, &const_ptr_25_indices[0], const_ptr_25_indices.size() ); -std::vector const_packed_26_elems; -ConstantFP* const_float_27 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); -const_packed_26_elems.push_back(const_float_27); -UndefValue* const_float_28 = UndefValue::get(Type::FloatTy); -const_packed_26_elems.push_back(const_float_28); -const_packed_26_elems.push_back(const_float_28); -const_packed_26_elems.push_back(const_float_27); -Constant* const_packed_26 = ConstantVector::get(VectorTy_13, const_packed_26_elems); -ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "1", 10)); -ConstantInt* const_int32_30 = ConstantInt::get(APInt(32, "3", 10)); -std::vector const_ptr_31_indices; -const_ptr_31_indices.push_back(const_int32_22); -const_ptr_31_indices.push_back(const_int32_22); -Constant* const_ptr_31 = ConstantExpr::getGetElementPtr(gvar_array__str3, &const_ptr_31_indices[0], const_ptr_31_indices.size() ); -ConstantInt* const_int32_32 = ConstantInt::get(APInt(32, "2", 10)); -std::vector const_packed_33_elems; -const_packed_33_elems.push_back(const_float_27); -const_packed_33_elems.push_back(const_float_23); -const_packed_33_elems.push_back(const_float_23); -const_packed_33_elems.push_back(const_float_27); -Constant* const_packed_33 = ConstantVector::get(VectorTy_13, const_packed_33_elems); +ConstantFP* const_float_4 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); +ConstantFP* const_float_5 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); +Constant* const_float_6 = Constant::getNullValue(Type::FloatTy); +Constant* const_int32_7 = Constant::getNullValue(IntegerType::get(32)); +std::vector const_packed_8_elems; +ConstantFP* const_float_9 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); +const_packed_8_elems.push_back(const_float_9); +UndefValue* const_float_10 = UndefValue::get(Type::FloatTy); +const_packed_8_elems.push_back(const_float_10); +const_packed_8_elems.push_back(const_float_10); +const_packed_8_elems.push_back(const_float_9); +Constant* const_packed_8 = ConstantVector::get(VectorTy_2, const_packed_8_elems); +ConstantInt* const_int32_11 = ConstantInt::get(APInt(32, "1", 10)); +ConstantInt* const_int32_12 = ConstantInt::get(APInt(32, "3", 10)); +ConstantInt* const_int32_13 = ConstantInt::get(APInt(32, "2", 10)); +std::vector const_packed_14_elems; +const_packed_14_elems.push_back(const_float_9); +const_packed_14_elems.push_back(const_float_6); +const_packed_14_elems.push_back(const_float_6); +const_packed_14_elems.push_back(const_float_9); +Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elems); // Global Variable Definitions -gvar_array__str->setInitializer(const_array_15); -gvar_array__str1->setInitializer(const_array_16); -gvar_array__str2->setInitializer(const_array_17); -gvar_array__str3->setInitializer(const_array_18); // Function Definitions @@ -721,36 +719,19 @@ gvar_array__str3->setInitializer(const_array_18); BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); // Block entry (label_entry) - FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_19, "cmp", label_entry); - SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_19, float_b, "b.addr.0", label_entry); - FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_20, "cmp3", label_entry); - SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_20, float_b_addr_0, "b.addr.1", label_entry); - CastInst* double_conv = new FPExtInst(float_a, Type::DoubleTy, "conv", label_entry); - CastInst* double_conv8 = new FPExtInst(float_b_addr_1, Type::DoubleTy, "conv8", label_entry); - std::vector int32_call_params; - int32_call_params.push_back(const_ptr_21); - int32_call_params.push_back(double_conv); - int32_call_params.push_back(double_conv8); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry); - int32_call->setCallingConv(CallingConv::C); - int32_call->setTailCall(true); - FCmpInst* int1_cmp11 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_23, "cmp11", label_entry); - SelectInst* float_a_addr_0 = new SelectInst(int1_cmp11, const_float_23, float_a, "a.addr.0", label_entry); - CastInst* double_conv16 = new FPExtInst(float_a_addr_0, Type::DoubleTy, "conv16", label_entry); - std::vector int32_call19_params; - int32_call19_params.push_back(const_ptr_24); - int32_call19_params.push_back(double_conv16); - int32_call19_params.push_back(double_conv8); - CallInst* int32_call19 = new CallInst(func_printf, int32_call19_params.begin(), int32_call19_params.end(), "call19", label_entry); - int32_call19->setCallingConv(CallingConv::C); - int32_call19->setTailCall(true); - std::vector float_call22_params; - float_call22_params.push_back(float_a_addr_0); - float_call22_params.push_back(float_b_addr_1); - CallInst* float_call22 = new CallInst(func_powf, float_call22_params.begin(), float_call22_params.end(), "call22", label_entry); - float_call22->setCallingConv(CallingConv::C); - float_call22->setTailCall(true); - new ReturnInst(float_call22, label_entry); + FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_4, "cmp", label_entry); + SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_4, float_b, "b.addr.0", label_entry); + FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_5, "cmp3", label_entry); + SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_5, float_b_addr_0, "b.addr.1", label_entry); + FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_6, "cmp7", label_entry); + SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_6, float_a, "a.addr.0", label_entry); + std::vector float_call_params; + float_call_params.push_back(float_a_addr_0); + float_call_params.push_back(float_b_addr_1); + CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); + float_call->setCallingConv(CallingConv::C); + float_call->setTailCall(true); + new ReturnInst(float_call, label_entry); } @@ -760,119 +741,37 @@ gvar_array__str3->setInitializer(const_array_18); Value* packed_tmp = args++; packed_tmp->setName("tmp"); - BasicBlock* label_entry_35 = new BasicBlock("entry",func_lit,0); + BasicBlock* label_entry_16 = new BasicBlock("entry",func_lit,0); BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); - // Block entry (label_entry_35) - CallInst* int32_call_36 = new CallInst(func_printf, const_ptr_25, "call", label_entry_35); - int32_call_36->setCallingConv(CallingConv::C); - int32_call_36->setTailCall(true); - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_22, "tmp7", label_entry_35); - FCmpInst* int1_cmp_37 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_23, "cmp", label_entry_35); - new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_37, label_entry_35); + // Block entry (label_entry_16) + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_7, "tmp7", label_entry_16); + FCmpInst* int1_cmp_17 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_6, "cmp", label_entry_16); + new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_17, label_entry_16); // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_26, float_tmp7, const_int32_29, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_29, "tmp14", label_ifthen); - CastInst* double_conv15 = new FPExtInst(float_tmp14, Type::DoubleTy, "conv15", label_ifthen); - ExtractElementInst* float_tmp17 = new ExtractElementInst(packed_tmp, const_int32_30, "tmp17", label_ifthen); - CastInst* double_conv18 = new FPExtInst(float_tmp17, Type::DoubleTy, "conv18", label_ifthen); - std::vector int32_call19_39_params; - int32_call19_39_params.push_back(const_ptr_31); - int32_call19_39_params.push_back(double_conv15); - int32_call19_39_params.push_back(double_conv18); - CallInst* int32_call19_39 = new CallInst(func_printf, int32_call19_39_params.begin(), int32_call19_39_params.end(), "call19", label_ifthen); - int32_call19_39->setCallingConv(CallingConv::C); - int32_call19_39->setTailCall(true); - FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp17, const_float_19, "cmp.i", label_ifthen); - SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_19, float_tmp17, "b.addr.0.i", label_ifthen); - FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_20, "cmp3.i", label_ifthen); - SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_20, float_b_addr_0_i, "b.addr.1.i", label_ifthen); - CastInst* double_conv8_i = new FPExtInst(float_b_addr_1_i, Type::DoubleTy, "conv8.i", label_ifthen); - std::vector int32_call_i_params; - int32_call_i_params.push_back(const_ptr_21); - int32_call_i_params.push_back(double_conv15); - int32_call_i_params.push_back(double_conv8_i); - CallInst* int32_call_i = new CallInst(func_printf, int32_call_i_params.begin(), int32_call_i_params.end(), "call.i", label_ifthen); - int32_call_i->setCallingConv(CallingConv::C); - int32_call_i->setTailCall(true); - FCmpInst* int1_cmp11_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_23, "cmp11.i", label_ifthen); - SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp11_i, const_float_23, float_tmp14, "a.addr.0.i", label_ifthen); - CastInst* double_conv16_i = new FPExtInst(float_a_addr_0_i, Type::DoubleTy, "conv16.i", label_ifthen); - std::vector int32_call19_i_params; - int32_call19_i_params.push_back(const_ptr_24); - int32_call19_i_params.push_back(double_conv16_i); - int32_call19_i_params.push_back(double_conv8_i); - CallInst* int32_call19_i = new CallInst(func_printf, int32_call19_i_params.begin(), int32_call19_i_params.end(), "call19.i", label_ifthen); - int32_call19_i->setCallingConv(CallingConv::C); - int32_call19_i->setTailCall(true); - std::vector float_call22_i_params; - float_call22_i_params.push_back(float_a_addr_0_i); - float_call22_i_params.push_back(float_b_addr_1_i); - CallInst* float_call22_i = new CallInst(func_powf, float_call22_i_params.begin(), float_call22_i_params.end(), "call22.i", label_ifthen); - float_call22_i->setCallingConv(CallingConv::C); - float_call22_i->setTailCall(true); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp12, float_call22_i, const_int32_32, "tmp26", label_ifthen); - new ReturnInst(packed_tmp26, label_ifthen); + InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_8, float_tmp7, const_int32_11, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_11, "tmp14", label_ifthen); + ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_12, "tmp16", label_ifthen); + FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_4, "cmp.i", label_ifthen); + SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_4, float_tmp16, "b.addr.0.i", label_ifthen); + FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_5, "cmp3.i", label_ifthen); + SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_5, float_b_addr_0_i, "b.addr.1.i", label_ifthen); + FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_6, "cmp7.i", label_ifthen); + SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_6, float_tmp14, "a.addr.0.i", label_ifthen); + std::vector float_call_i_params; + float_call_i_params.push_back(float_a_addr_0_i); + float_call_i_params.push_back(float_b_addr_1_i); + CallInst* float_call_i = new CallInst(func_powf, float_call_i_params.begin(), float_call_i_params.end(), "call.i", label_ifthen); + float_call_i->setCallingConv(CallingConv::C); + float_call_i->setTailCall(true); + InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_13, "tmp18", label_ifthen); + new ReturnInst(packed_tmp18, label_ifthen); // Block UnifiedReturnBlock (label_UnifiedReturnBlock) - new ReturnInst(const_packed_33, label_UnifiedReturnBlock); + new ReturnInst(const_packed_14, label_UnifiedReturnBlock); } - -return func_lit; - -} - -void Instructions::printVector(llvm::Value *val) -{ - static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A"; - - if (!m_fmtPtr) { - Constant *format = ConstantArray::get(frmt, true); - ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1); - GlobalVariable* globalFormat = new GlobalVariable( - /*Type=*/arrayTy, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name(".str"), - m_mod); - globalFormat->setInitializer(format); - - Constant* const_int0 = Constant::getNullValue(IntegerType::get(32)); - std::vector const_ptr_21_indices; - const_ptr_21_indices.push_back(const_int0); - const_ptr_21_indices.push_back(const_int0); - m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat, - &const_ptr_21_indices[0], const_ptr_21_indices.size()); - } - - Function *func_printf = m_mod->getFunction("printf"); - ExtractElementInst *x = new ExtractElementInst(val, unsigned(0), - name("extractx"), - m_block); - ExtractElementInst *y = new ExtractElementInst(val, unsigned(1), - name("extracty"), - m_block); - ExtractElementInst *z = new ExtractElementInst(val, unsigned(2), - name("extractz"), - m_block); - ExtractElementInst *w = new ExtractElementInst(val, unsigned(3), - name("extractw"), - m_block); - CastInst *dx = new FPExtInst(x, Type::DoubleTy, name("dx"), m_block); - CastInst *dy = new FPExtInst(y, Type::DoubleTy, name("dy"), m_block); - CastInst *dz = new FPExtInst(z, Type::DoubleTy, name("dz"), m_block); - CastInst *dw = new FPExtInst(w, Type::DoubleTy, name("dw"), m_block); - std::vector params; - params.push_back(m_fmtPtr); - params.push_back(dx); - params.push_back(dy); - params.push_back(dz); - params.push_back(dw); - CallInst* call = new CallInst(func_printf, params.begin(), params.end(), "printf", m_block); - call->setCallingConv(CallingConv::C); - call->setTailCall(true); + return func_lit; } diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 87203fdc10..b4948ed240 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -47,6 +47,8 @@ private: llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, llvm::Value *z, llvm::Value *w=0); + + llvm::Function *declarePrintf(); private: llvm::Module *m_mod; llvm::Function *m_func; diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index f1854b5c37..3a458cb11f 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -6,157 +6,123 @@ Module* createBaseShader() { Module* mod = new Module("Shader"); // Type Definitions - ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 24); + std::vectorFuncTy_0_args; + VectorType* VectorTy_3 = VectorType::get(Type::FloatTy, 4); - PointerType* PointerTy_1 = PointerType::get(ArrayTy_0); + ArrayType* ArrayTy_2 = ArrayType::get(VectorTy_3, 16); - ArrayType* ArrayTy_2 = ArrayType::get(IntegerType::get(8), 20); + PointerType* PointerTy_1 = PointerType::get(ArrayTy_2); - PointerType* PointerTy_3 = PointerType::get(ArrayTy_2); + FuncTy_0_args.push_back(PointerTy_1); + ArrayType* ArrayTy_6 = ArrayType::get(Type::FloatTy, 4); - ArrayType* ArrayTy_4 = ArrayType::get(IntegerType::get(8), 68); + ArrayType* ArrayTy_5 = ArrayType::get(ArrayTy_6, 16); - PointerType* PointerTy_5 = PointerType::get(ArrayTy_4); + PointerType* PointerTy_4 = PointerType::get(ArrayTy_5); - ArrayType* ArrayTy_6 = ArrayType::get(IntegerType::get(8), 33); - - PointerType* PointerTy_7 = PointerType::get(ArrayTy_6); - - std::vectorFuncTy_8_args; - VectorType* VectorTy_11 = VectorType::get(Type::FloatTy, 4); - - ArrayType* ArrayTy_10 = ArrayType::get(VectorTy_11, 16); - - PointerType* PointerTy_9 = PointerType::get(ArrayTy_10); - - FuncTy_8_args.push_back(PointerTy_9); - ArrayType* ArrayTy_14 = ArrayType::get(Type::FloatTy, 4); - - ArrayType* ArrayTy_13 = ArrayType::get(ArrayTy_14, 16); - - PointerType* PointerTy_12 = PointerType::get(ArrayTy_13); - - FuncTy_8_args.push_back(PointerTy_12); - FuncTy_8_args.push_back(IntegerType::get(32)); - FuncTy_8_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_8_PAL = 0; - FunctionType* FuncTy_8 = FunctionType::get( + FuncTy_0_args.push_back(PointerTy_4); + FuncTy_0_args.push_back(IntegerType::get(32)); + FuncTy_0_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_0_PAL = 0; + FunctionType* FuncTy_0 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_8_args, + /*Params=*/FuncTy_0_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_8_PAL); - - PointerType* PointerTy_15 = PointerType::get(Type::FloatTy); + /*ParamAttrs=*/FuncTy_0_PAL); - std::vectorFuncTy_17_args; - ParamAttrsList *FuncTy_17_PAL = 0; - FunctionType* FuncTy_17 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_17_args, - /*isVarArg=*/true, - /*ParamAttrs=*/FuncTy_17_PAL); + PointerType* PointerTy_7 = PointerType::get(Type::FloatTy); - PointerType* PointerTy_16 = PointerType::get(FuncTy_17); + PointerType* PointerTy_8 = PointerType::get(VectorTy_3); - PointerType* PointerTy_18 = PointerType::get(IntegerType::get(8)); + std::vectorFuncTy_9_args; + FuncTy_9_args.push_back(PointerTy_8); + PointerType* PointerTy_10 = PointerType::get(ArrayTy_6); - PointerType* PointerTy_19 = PointerType::get(VectorTy_11); - - std::vectorFuncTy_20_args; - FuncTy_20_args.push_back(PointerTy_19); - PointerType* PointerTy_21 = PointerType::get(ArrayTy_14); - - FuncTy_20_args.push_back(PointerTy_21); - FuncTy_20_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_20_PAL = 0; - FunctionType* FuncTy_20 = FunctionType::get( + FuncTy_9_args.push_back(PointerTy_10); + FuncTy_9_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_9_PAL = 0; + FunctionType* FuncTy_9 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_20_args, + /*Params=*/FuncTy_9_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_20_PAL); - - std::vectorFuncTy_22_args; - FuncTy_22_args.push_back(PointerTy_21); - FuncTy_22_args.push_back(PointerTy_19); - FuncTy_22_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_22_PAL = 0; - FunctionType* FuncTy_22 = FunctionType::get( + /*ParamAttrs=*/FuncTy_9_PAL); + + std::vectorFuncTy_11_args; + FuncTy_11_args.push_back(PointerTy_10); + FuncTy_11_args.push_back(PointerTy_8); + FuncTy_11_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_11_PAL = 0; + FunctionType* FuncTy_11 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_22_args, + /*Params=*/FuncTy_11_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_22_PAL); - - std::vectorFuncTy_23_args; - FuncTy_23_args.push_back(PointerTy_12); - FuncTy_23_args.push_back(PointerTy_12); - FuncTy_23_args.push_back(PointerTy_21); - FuncTy_23_args.push_back(IntegerType::get(32)); - FuncTy_23_args.push_back(IntegerType::get(32)); - FuncTy_23_args.push_back(IntegerType::get(32)); - FuncTy_23_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_23_PAL = 0; - FunctionType* FuncTy_23 = FunctionType::get( + /*ParamAttrs=*/FuncTy_11_PAL); + + std::vectorFuncTy_12_args; + FuncTy_12_args.push_back(PointerTy_4); + FuncTy_12_args.push_back(PointerTy_4); + FuncTy_12_args.push_back(PointerTy_10); + FuncTy_12_args.push_back(IntegerType::get(32)); + FuncTy_12_args.push_back(IntegerType::get(32)); + FuncTy_12_args.push_back(IntegerType::get(32)); + FuncTy_12_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_12_PAL = 0; + FunctionType* FuncTy_12 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_23_args, + /*Params=*/FuncTy_12_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_23_PAL); + /*ParamAttrs=*/FuncTy_12_PAL); - ArrayType* ArrayTy_25 = ArrayType::get(ArrayTy_10, 2048); + ArrayType* ArrayTy_14 = ArrayType::get(ArrayTy_2, 2048); - PointerType* PointerTy_24 = PointerType::get(ArrayTy_25); + PointerType* PointerTy_13 = PointerType::get(ArrayTy_14); - ArrayType* ArrayTy_27 = ArrayType::get(VectorTy_11, 32); + ArrayType* ArrayTy_16 = ArrayType::get(VectorTy_3, 32); - PointerType* PointerTy_26 = PointerType::get(ArrayTy_27); + PointerType* PointerTy_15 = PointerType::get(ArrayTy_16); - std::vectorFuncTy_29_args; - FuncTy_29_args.push_back(PointerTy_19); - FuncTy_29_args.push_back(PointerTy_19); - FuncTy_29_args.push_back(PointerTy_19); - ParamAttrsList *FuncTy_29_PAL = 0; - FunctionType* FuncTy_29 = FunctionType::get( + std::vectorFuncTy_18_args; + FuncTy_18_args.push_back(PointerTy_8); + FuncTy_18_args.push_back(PointerTy_8); + FuncTy_18_args.push_back(PointerTy_8); + ParamAttrsList *FuncTy_18_PAL = 0; + FunctionType* FuncTy_18 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_29_args, + /*Params=*/FuncTy_18_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_29_PAL); + /*ParamAttrs=*/FuncTy_18_PAL); - PointerType* PointerTy_28 = PointerType::get(FuncTy_29); + PointerType* PointerTy_17 = PointerType::get(FuncTy_18); // Function Declarations Function* func_from_array = new Function( - /*Type=*/FuncTy_8, + /*Type=*/FuncTy_0, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_array", mod); func_from_array->setCallingConv(CallingConv::C); - Function* func_printf = new Function( - /*Type=*/FuncTy_17, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", mod); // (external, no body) - func_printf->setCallingConv(CallingConv::C); - Function* func_from_consts = new Function( - /*Type=*/FuncTy_20, + /*Type=*/FuncTy_9, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_consts", mod); func_from_consts->setCallingConv(CallingConv::C); Function* func_to_array = new Function( - /*Type=*/FuncTy_22, + /*Type=*/FuncTy_11, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"to_array", mod); func_to_array->setCallingConv(CallingConv::C); Function* func_run_vertex_shader = new Function( - /*Type=*/FuncTy_23, + /*Type=*/FuncTy_12, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_vertex_shader", mod); func_run_vertex_shader->setCallingConv(CallingConv::C); Function* func_execute_shader = new Function( - /*Type=*/FuncTy_29, + /*Type=*/FuncTy_18, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"execute_shader", mod); // (external, no body) func_execute_shader->setCallingConv(CallingConv::C); @@ -164,98 +130,14 @@ Module* createBaseShader() { // Global Variable Declarations - GlobalVariable* gvar_array__str = new GlobalVariable( - /*Type=*/ArrayTy_0, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str", - mod); - - GlobalVariable* gvar_array__str1 = new GlobalVariable( - /*Type=*/ArrayTy_0, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str1", - mod); - - GlobalVariable* gvar_array__str2 = new GlobalVariable( - /*Type=*/ArrayTy_2, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str2", - mod); - - GlobalVariable* gvar_array__str3 = new GlobalVariable( - /*Type=*/ArrayTy_4, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str3", - mod); - - GlobalVariable* gvar_array__str4 = new GlobalVariable( - /*Type=*/ArrayTy_6, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str4", - mod); - - GlobalVariable* gvar_array__str5 = new GlobalVariable( - /*Type=*/ArrayTy_0, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str5", - mod); - // Constant Definitions - Constant* const_array_30 = ConstantArray::get("FRA(%d %d) %f %f %f %f\x0A", true); - Constant* const_array_31 = ConstantArray::get("VCONST(%d) %f %f %f %f\x0A", true); - Constant* const_array_32 = ConstantArray::get("DEST = %f %f %f %f\x0A", true); - Constant* const_array_33 = ConstantArray::get("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, consts = %d\x0A", true); - Constant* const_array_34 = ConstantArray::get(" after conversion\x0A", true); - Constant* const_array_35 = ConstantArray::get("after executing shader\x0A", true); - Constant* const_int32_36 = Constant::getNullValue(IntegerType::get(32)); - UndefValue* const_packed_37 = UndefValue::get(VectorTy_11); - ConstantInt* const_int32_38 = ConstantInt::get(APInt(32, "1", 10)); - ConstantInt* const_int32_39 = ConstantInt::get(APInt(32, "2", 10)); - ConstantInt* const_int32_40 = ConstantInt::get(APInt(32, "3", 10)); - std::vector const_ptr_41_indices; - const_ptr_41_indices.push_back(const_int32_36); - const_ptr_41_indices.push_back(const_int32_36); - Constant* const_ptr_41 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_41_indices[0], const_ptr_41_indices.size() ); - std::vector const_ptr_42_indices; - const_ptr_42_indices.push_back(const_int32_36); - const_ptr_42_indices.push_back(const_int32_36); - Constant* const_ptr_42 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_42_indices[0], const_ptr_42_indices.size() ); - std::vector const_ptr_43_indices; - const_ptr_43_indices.push_back(const_int32_36); - const_ptr_43_indices.push_back(const_int32_36); - Constant* const_ptr_43 = ConstantExpr::getGetElementPtr(gvar_array__str2, &const_ptr_43_indices[0], const_ptr_43_indices.size() ); - std::vector const_ptr_44_indices; - const_ptr_44_indices.push_back(const_int32_36); - const_ptr_44_indices.push_back(const_int32_36); - Constant* const_ptr_44 = ConstantExpr::getGetElementPtr(gvar_array__str3, &const_ptr_44_indices[0], const_ptr_44_indices.size() ); - std::vector const_ptr_45_indices; - const_ptr_45_indices.push_back(const_int32_36); - const_ptr_45_indices.push_back(const_int32_36); - Constant* const_ptr_45 = ConstantExpr::getGetElementPtr(gvar_array__str4, &const_ptr_45_indices[0], const_ptr_45_indices.size() ); - std::vector const_ptr_46_indices; - const_ptr_46_indices.push_back(const_int32_36); - const_ptr_46_indices.push_back(const_int32_36); - Constant* const_ptr_46 = ConstantExpr::getGetElementPtr(gvar_array__str5, &const_ptr_46_indices[0], const_ptr_46_indices.size() ); + Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); + UndefValue* const_packed_20 = UndefValue::get(VectorTy_3); + ConstantInt* const_int32_21 = ConstantInt::get(APInt(32, "1", 10)); + ConstantInt* const_int32_22 = ConstantInt::get(APInt(32, "2", 10)); + ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "3", 10)); // Global Variable Definitions - gvar_array__str->setInitializer(const_array_30); - gvar_array__str1->setInitializer(const_array_31); - gvar_array__str2->setInitializer(const_array_32); - gvar_array__str3->setInitializer(const_array_33); - gvar_array__str4->setInitializer(const_array_34); - gvar_array__str5->setInitializer(const_array_35); // Function Definitions @@ -273,187 +155,158 @@ Module* createBaseShader() { BasicBlock* label_entry = new BasicBlock("entry",func_from_array,0); BasicBlock* label_forbody6 = new BasicBlock("forbody6",func_from_array,0); - BasicBlock* label_forinc70 = new BasicBlock("forinc70",func_from_array,0); - BasicBlock* label_afterfor73 = new BasicBlock("afterfor73",func_from_array,0); + BasicBlock* label_forinc57 = new BasicBlock("forinc57",func_from_array,0); + BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); // Block entry (label_entry) - ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_36, "cmp", label_entry); - ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_36, "cmp5", label_entry); + ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_19, "cmp", label_entry); + ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_19, "cmp5", label_entry); BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); - new BranchInst(label_forbody6, label_afterfor73, int1_bothcond, label_entry); + new BranchInst(label_forbody6, label_afterfor60, int1_bothcond, label_entry); // Block forbody6 (label_forbody6) - Argument* fwdref_48 = new Argument(IntegerType::get(32)); - Argument* fwdref_49 = new Argument(IntegerType::get(32)); - PHINode* int32_i_014_0 = new PHINode(IntegerType::get(32), "i.014.0", label_forbody6); - int32_i_014_0->reserveOperandSpace(3); - int32_i_014_0->addIncoming(const_int32_36, label_entry); - int32_i_014_0->addIncoming(fwdref_48, label_forinc70); - int32_i_014_0->addIncoming(fwdref_49, label_forbody6); - - Argument* fwdref_50 = new Argument(IntegerType::get(32)); - PHINode* int32_j_03_0 = new PHINode(IntegerType::get(32), "j.03.0", label_forbody6); - int32_j_03_0->reserveOperandSpace(3); - int32_j_03_0->addIncoming(fwdref_50, label_forbody6); - int32_j_03_0->addIncoming(const_int32_36, label_forinc70); - int32_j_03_0->addIncoming(const_int32_36, label_entry); - - Argument* fwdref_51 = new Argument(VectorTy_11); - PHINode* packed_vec_01_0 = new PHINode(VectorTy_11, "vec.01.0", label_forbody6); - packed_vec_01_0->reserveOperandSpace(3); - packed_vec_01_0->addIncoming(fwdref_51, label_forbody6); - packed_vec_01_0->addIncoming(const_packed_37, label_entry); - packed_vec_01_0->addIncoming(fwdref_51, label_forinc70); + Argument* fwdref_25 = new Argument(IntegerType::get(32)); + Argument* fwdref_26 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody6); + int32_i_0_reg2mem_0->reserveOperandSpace(3); + int32_i_0_reg2mem_0->addIncoming(const_int32_19, label_entry); + int32_i_0_reg2mem_0->addIncoming(fwdref_25, label_forinc57); + int32_i_0_reg2mem_0->addIncoming(fwdref_26, label_forbody6); + + Argument* fwdref_27 = new Argument(IntegerType::get(32)); + PHINode* int32_j_0_reg2mem_0 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0", label_forbody6); + int32_j_0_reg2mem_0->reserveOperandSpace(3); + int32_j_0_reg2mem_0->addIncoming(fwdref_27, label_forbody6); + int32_j_0_reg2mem_0->addIncoming(const_int32_19, label_forinc57); + int32_j_0_reg2mem_0->addIncoming(const_int32_19, label_entry); + + Argument* fwdref_28 = new Argument(VectorTy_3); + PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_3, "vec.0.reg2mem.0", label_forbody6); + packed_vec_0_reg2mem_0->reserveOperandSpace(3); + packed_vec_0_reg2mem_0->addIncoming(fwdref_28, label_forbody6); + packed_vec_0_reg2mem_0->addIncoming(const_packed_20, label_entry); + packed_vec_0_reg2mem_0->addIncoming(fwdref_28, label_forinc57); std::vector ptr_arraydecay11_indices; - ptr_arraydecay11_indices.push_back(int32_i_014_0); - ptr_arraydecay11_indices.push_back(int32_j_03_0); - ptr_arraydecay11_indices.push_back(const_int32_36); + ptr_arraydecay11_indices.push_back(int32_i_0_reg2mem_0); + ptr_arraydecay11_indices.push_back(int32_j_0_reg2mem_0); + ptr_arraydecay11_indices.push_back(const_int32_19); Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); - InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_01_0, float_tmp13, const_int32_36, "tmp15", label_forbody6); + InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_19, "tmp15", label_forbody6); std::vector ptr_arrayidx23_indices; - ptr_arrayidx23_indices.push_back(int32_i_014_0); - ptr_arrayidx23_indices.push_back(int32_j_03_0); - ptr_arrayidx23_indices.push_back(const_int32_38); + ptr_arrayidx23_indices.push_back(int32_i_0_reg2mem_0); + ptr_arrayidx23_indices.push_back(int32_j_0_reg2mem_0); + ptr_arrayidx23_indices.push_back(const_int32_21); Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_38, "tmp26", label_forbody6); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_21, "tmp26", label_forbody6); std::vector ptr_arrayidx34_indices; - ptr_arrayidx34_indices.push_back(int32_i_014_0); - ptr_arrayidx34_indices.push_back(int32_j_03_0); - ptr_arrayidx34_indices.push_back(const_int32_39); + ptr_arrayidx34_indices.push_back(int32_i_0_reg2mem_0); + ptr_arrayidx34_indices.push_back(int32_j_0_reg2mem_0); + ptr_arrayidx34_indices.push_back(const_int32_22); Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6); LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6); - InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_39, "tmp37", label_forbody6); + InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_22, "tmp37", label_forbody6); std::vector ptr_arrayidx45_indices; - ptr_arrayidx45_indices.push_back(int32_i_014_0); - ptr_arrayidx45_indices.push_back(int32_j_03_0); - ptr_arrayidx45_indices.push_back(const_int32_40); + ptr_arrayidx45_indices.push_back(int32_i_0_reg2mem_0); + ptr_arrayidx45_indices.push_back(int32_j_0_reg2mem_0); + ptr_arrayidx45_indices.push_back(const_int32_23); Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6); LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6); - InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_40, "tmp48", label_forbody6); - CastInst* double_conv = new FPExtInst(float_tmp13, Type::DoubleTy, "conv", label_forbody6); - CastInst* double_conv55 = new FPExtInst(float_tmp24, Type::DoubleTy, "conv55", label_forbody6); - CastInst* double_conv58 = new FPExtInst(float_tmp35, Type::DoubleTy, "conv58", label_forbody6); - CastInst* double_conv61 = new FPExtInst(float_tmp46, Type::DoubleTy, "conv61", label_forbody6); - std::vector int32_call_params; - int32_call_params.push_back(const_ptr_41); - int32_call_params.push_back(int32_i_014_0); - int32_call_params.push_back(int32_j_03_0); - int32_call_params.push_back(double_conv); - int32_call_params.push_back(double_conv55); - int32_call_params.push_back(double_conv58); - int32_call_params.push_back(double_conv61); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_forbody6); - int32_call->setCallingConv(CallingConv::C); - int32_call->setTailCall(true); - std::vector ptr_arrayidx67_indices; - ptr_arrayidx67_indices.push_back(int32_i_014_0); - ptr_arrayidx67_indices.push_back(int32_j_03_0); - Instruction* ptr_arrayidx67 = new GetElementPtrInst(ptr_res, ptr_arrayidx67_indices.begin(), ptr_arrayidx67_indices.end(), "arrayidx67", label_forbody6); - StoreInst* void_52 = new StoreInst(packed_tmp48, ptr_arrayidx67, false, label_forbody6); - BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_03_0, const_int32_38, "inc", label_forbody6); - ICmpInst* int1_cmp512 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp512", label_forbody6); - new BranchInst(label_forbody6, label_forinc70, int1_cmp512, label_forbody6); - - // Block forinc70 (label_forinc70) - BinaryOperator* int32_inc72 = BinaryOperator::create(Instruction::Add, int32_i_014_0, const_int32_38, "inc72", label_forinc70); - ICmpInst* int1_cmp23 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc72, int32_count, "cmp23", label_forinc70); - new BranchInst(label_forbody6, label_afterfor73, int1_cmp23, label_forinc70); - - // Block afterfor73 (label_afterfor73) - new ReturnInst(label_afterfor73); + InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_23, "tmp48", label_forbody6); + std::vector ptr_arrayidx54_indices; + ptr_arrayidx54_indices.push_back(int32_i_0_reg2mem_0); + ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0); + Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); + StoreInst* void_29 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); + BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_21, "inc", label_forbody6); + ICmpInst* int1_cmp59 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp59", label_forbody6); + new BranchInst(label_forbody6, label_forinc57, int1_cmp59, label_forbody6); + + // Block forinc57 (label_forinc57) + BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_21, "inc59", label_forinc57); + ICmpInst* int1_cmp17 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59, int32_count, "cmp17", label_forinc57); + new BranchInst(label_forbody6, label_afterfor60, int1_cmp17, label_forinc57); + + // Block afterfor60 (label_afterfor60) + new ReturnInst(label_afterfor60); // Resolve Forward References - fwdref_49->replaceAllUsesWith(int32_i_014_0); delete fwdref_49; - fwdref_51->replaceAllUsesWith(packed_tmp48); delete fwdref_51; - fwdref_50->replaceAllUsesWith(int32_inc); delete fwdref_50; - fwdref_48->replaceAllUsesWith(int32_inc72); delete fwdref_48; + fwdref_26->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_26; + fwdref_28->replaceAllUsesWith(packed_tmp48); delete fwdref_28; + fwdref_27->replaceAllUsesWith(int32_inc); delete fwdref_27; + fwdref_25->replaceAllUsesWith(int32_inc59); delete fwdref_25; } // Function: from_consts (func_from_consts) { Function::arg_iterator args = func_from_consts->arg_begin(); - Value* ptr_res_56 = args++; - ptr_res_56->setName("res"); - Value* ptr_ainputs_57 = args++; - ptr_ainputs_57->setName("ainputs"); - Value* int32_count_58 = args++; - int32_count_58->setName("count"); - - BasicBlock* label_entry_59 = new BasicBlock("entry",func_from_consts,0); + Value* ptr_res_33 = args++; + ptr_res_33->setName("res"); + Value* ptr_ainputs_34 = args++; + ptr_ainputs_34->setName("ainputs"); + Value* int32_count_35 = args++; + int32_count_35->setName("count"); + + BasicBlock* label_entry_36 = new BasicBlock("entry",func_from_consts,0); BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); - // Block entry (label_entry_59) - ICmpInst* int1_cmp_60 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_58, const_int32_36, "cmp", label_entry_59); - new BranchInst(label_forbody, label_afterfor, int1_cmp_60, label_entry_59); + // Block entry (label_entry_36) + ICmpInst* int1_cmp_37 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_35, const_int32_19, "cmp", label_entry_36); + new BranchInst(label_forbody, label_afterfor, int1_cmp_37, label_entry_36); // Block forbody (label_forbody) - Argument* fwdref_62 = new Argument(IntegerType::get(32)); - PHINode* int32_i_02_0 = new PHINode(IntegerType::get(32), "i.02.0", label_forbody); - int32_i_02_0->reserveOperandSpace(2); - int32_i_02_0->addIncoming(const_int32_36, label_entry_59); - int32_i_02_0->addIncoming(fwdref_62, label_forbody); - - Argument* fwdref_64 = new Argument(VectorTy_11); - PHINode* packed_vec_01_0_63 = new PHINode(VectorTy_11, "vec.01.0", label_forbody); - packed_vec_01_0_63->reserveOperandSpace(2); - packed_vec_01_0_63->addIncoming(const_packed_37, label_entry_59); - packed_vec_01_0_63->addIncoming(fwdref_64, label_forbody); + Argument* fwdref_40 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_39 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); + int32_i_0_reg2mem_0_39->reserveOperandSpace(2); + int32_i_0_reg2mem_0_39->addIncoming(const_int32_19, label_entry_36); + int32_i_0_reg2mem_0_39->addIncoming(fwdref_40, label_forbody); + + Argument* fwdref_42 = new Argument(VectorTy_3); + PHINode* packed_vec_0_reg2mem_0_41 = new PHINode(VectorTy_3, "vec.0.reg2mem.0", label_forbody); + packed_vec_0_reg2mem_0_41->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_41->addIncoming(const_packed_20, label_entry_36); + packed_vec_0_reg2mem_0_41->addIncoming(fwdref_42, label_forbody); std::vector ptr_arraydecay_indices; - ptr_arraydecay_indices.push_back(int32_i_02_0); - ptr_arraydecay_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_57, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); + ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_39); + ptr_arraydecay_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_34, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); - InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_01_0_63, float_tmp5, const_int32_36, "tmp7", label_forbody); + InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_41, float_tmp5, const_int32_19, "tmp7", label_forbody); std::vector ptr_arrayidx12_indices; - ptr_arrayidx12_indices.push_back(int32_i_02_0); - ptr_arrayidx12_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_57, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); - LoadInst* float_tmp13_65 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); - InsertElementInst* packed_tmp15_66 = new InsertElementInst(packed_tmp7, float_tmp13_65, const_int32_38, "tmp15", label_forbody); + ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_39); + ptr_arrayidx12_indices.push_back(const_int32_21); + Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_34, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); + LoadInst* float_tmp13_43 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); + InsertElementInst* packed_tmp15_44 = new InsertElementInst(packed_tmp7, float_tmp13_43, const_int32_21, "tmp15", label_forbody); std::vector ptr_arrayidx20_indices; - ptr_arrayidx20_indices.push_back(int32_i_02_0); - ptr_arrayidx20_indices.push_back(const_int32_39); - Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_57, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); + ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_39); + ptr_arrayidx20_indices.push_back(const_int32_22); + Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_34, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); - InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_66, float_tmp21, const_int32_39, "tmp23", label_forbody); + InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_44, float_tmp21, const_int32_22, "tmp23", label_forbody); std::vector ptr_arrayidx28_indices; - ptr_arrayidx28_indices.push_back(int32_i_02_0); - ptr_arrayidx28_indices.push_back(const_int32_40); - Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_57, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); + ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_39); + ptr_arrayidx28_indices.push_back(const_int32_23); + Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_34, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); - InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_40, "tmp31", label_forbody); - CastInst* double_conv_67 = new FPExtInst(float_tmp5, Type::DoubleTy, "conv", label_forbody); - CastInst* double_conv37 = new FPExtInst(float_tmp13_65, Type::DoubleTy, "conv37", label_forbody); - CastInst* double_conv40 = new FPExtInst(float_tmp21, Type::DoubleTy, "conv40", label_forbody); - CastInst* double_conv43 = new FPExtInst(float_tmp29, Type::DoubleTy, "conv43", label_forbody); - std::vector int32_call_68_params; - int32_call_68_params.push_back(const_ptr_42); - int32_call_68_params.push_back(int32_i_02_0); - int32_call_68_params.push_back(double_conv_67); - int32_call_68_params.push_back(double_conv37); - int32_call_68_params.push_back(double_conv40); - int32_call_68_params.push_back(double_conv43); - CallInst* int32_call_68 = new CallInst(func_printf, int32_call_68_params.begin(), int32_call_68_params.end(), "call", label_forbody); - int32_call_68->setCallingConv(CallingConv::C); - int32_call_68->setTailCall(true); - GetElementPtrInst* ptr_arrayidx46 = new GetElementPtrInst(ptr_res_56, int32_i_02_0, "arrayidx46", label_forbody); - StoreInst* void_69 = new StoreInst(packed_tmp31, ptr_arrayidx46, false, label_forbody); - BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_02_0, const_int32_38, "indvar.next", label_forbody); - ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_58, "exitcond", label_forbody); + InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_23, "tmp31", label_forbody); + GetElementPtrInst* ptr_arrayidx34_45 = new GetElementPtrInst(ptr_res_33, int32_i_0_reg2mem_0_39, "arrayidx34", label_forbody); + StoreInst* void_46 = new StoreInst(packed_tmp31, ptr_arrayidx34_45, false, label_forbody); + BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_39, const_int32_21, "indvar.next", label_forbody); + ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_35, "exitcond", label_forbody); new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); // Block afterfor (label_afterfor) new ReturnInst(label_afterfor); // Resolve Forward References - fwdref_64->replaceAllUsesWith(packed_tmp31); delete fwdref_64; - fwdref_62->replaceAllUsesWith(int32_indvar_next); delete fwdref_62; + fwdref_42->replaceAllUsesWith(packed_tmp31); delete fwdref_42; + fwdref_40->replaceAllUsesWith(int32_indvar_next); delete fwdref_40; } @@ -464,418 +317,346 @@ Module* createBaseShader() { ptr_dests->setName("dests"); Value* ptr_in = args++; ptr_in->setName("in"); - Value* int32_num_attribs_72 = args++; - int32_num_attribs_72->setName("num_attribs"); - - BasicBlock* label_entry_73 = new BasicBlock("entry",func_to_array,0); - BasicBlock* label_forbody_74 = new BasicBlock("forbody",func_to_array,0); - BasicBlock* label_afterfor_75 = new BasicBlock("afterfor",func_to_array,0); - - // Block entry (label_entry_73) - ICmpInst* int1_cmp_76 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_72, const_int32_36, "cmp", label_entry_73); - new BranchInst(label_forbody_74, label_afterfor_75, int1_cmp_76, label_entry_73); - - // Block forbody (label_forbody_74) - Argument* fwdref_78 = new Argument(IntegerType::get(32)); - PHINode* int32_i_01_0 = new PHINode(IntegerType::get(32), "i.01.0", label_forbody_74); - int32_i_01_0->reserveOperandSpace(2); - int32_i_01_0->addIncoming(const_int32_36, label_entry_73); - int32_i_01_0->addIncoming(fwdref_78, label_forbody_74); - - std::vector ptr_arraydecay_79_indices; - ptr_arraydecay_79_indices.push_back(int32_i_01_0); - ptr_arraydecay_79_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay_79 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_79_indices.begin(), ptr_arraydecay_79_indices.end(), "arraydecay", label_forbody_74); - GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_01_0, "arrayidx6", label_forbody_74); - LoadInst* packed_tmp7_80 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_74); - ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_tmp7_80, const_int32_36, "tmp9", label_forbody_74); - CastInst* double_conv_81 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv", label_forbody_74); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_80, const_int32_38, "tmp11", label_forbody_74); - CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_forbody_74); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp7_80, const_int32_39, "tmp14", label_forbody_74); - CastInst* double_conv15 = new FPExtInst(float_tmp14, Type::DoubleTy, "conv15", label_forbody_74); - ExtractElementInst* float_tmp17 = new ExtractElementInst(packed_tmp7_80, const_int32_40, "tmp17", label_forbody_74); - CastInst* double_conv18 = new FPExtInst(float_tmp17, Type::DoubleTy, "conv18", label_forbody_74); - std::vector int32_call_82_params; - int32_call_82_params.push_back(const_ptr_43); - int32_call_82_params.push_back(double_conv_81); - int32_call_82_params.push_back(double_conv12); - int32_call_82_params.push_back(double_conv15); - int32_call_82_params.push_back(double_conv18); - CallInst* int32_call_82 = new CallInst(func_printf, int32_call_82_params.begin(), int32_call_82_params.end(), "call", label_forbody_74); - int32_call_82->setCallingConv(CallingConv::C); - int32_call_82->setTailCall(true); - StoreInst* void_83 = new StoreInst(float_tmp9, ptr_arraydecay_79, false, label_forbody_74); - std::vector ptr_arrayidx24_indices; - ptr_arrayidx24_indices.push_back(int32_i_01_0); - ptr_arrayidx24_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx24 = new GetElementPtrInst(ptr_dests, ptr_arrayidx24_indices.begin(), ptr_arrayidx24_indices.end(), "arrayidx24", label_forbody_74); - StoreInst* void_84 = new StoreInst(float_tmp11, ptr_arrayidx24, false, label_forbody_74); - std::vector ptr_arrayidx28_85_indices; - ptr_arrayidx28_85_indices.push_back(int32_i_01_0); - ptr_arrayidx28_85_indices.push_back(const_int32_39); - Instruction* ptr_arrayidx28_85 = new GetElementPtrInst(ptr_dests, ptr_arrayidx28_85_indices.begin(), ptr_arrayidx28_85_indices.end(), "arrayidx28", label_forbody_74); - StoreInst* void_86 = new StoreInst(float_tmp14, ptr_arrayidx28_85, false, label_forbody_74); - std::vector ptr_arrayidx32_indices; - ptr_arrayidx32_indices.push_back(int32_i_01_0); - ptr_arrayidx32_indices.push_back(const_int32_40); - Instruction* ptr_arrayidx32 = new GetElementPtrInst(ptr_dests, ptr_arrayidx32_indices.begin(), ptr_arrayidx32_indices.end(), "arrayidx32", label_forbody_74); - StoreInst* void_87 = new StoreInst(float_tmp17, ptr_arrayidx32, false, label_forbody_74); - BinaryOperator* int32_indvar_next_88 = BinaryOperator::create(Instruction::Add, int32_i_01_0, const_int32_38, "indvar.next", label_forbody_74); - ICmpInst* int1_exitcond_89 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_88, int32_num_attribs_72, "exitcond", label_forbody_74); - new BranchInst(label_afterfor_75, label_forbody_74, int1_exitcond_89, label_forbody_74); - - // Block afterfor (label_afterfor_75) - new ReturnInst(label_afterfor_75); + Value* int32_num_attribs_49 = args++; + int32_num_attribs_49->setName("num_attribs"); + + BasicBlock* label_entry_50 = new BasicBlock("entry",func_to_array,0); + BasicBlock* label_forbody_51 = new BasicBlock("forbody",func_to_array,0); + BasicBlock* label_afterfor_52 = new BasicBlock("afterfor",func_to_array,0); + + // Block entry (label_entry_50) + ICmpInst* int1_cmp_53 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_49, const_int32_19, "cmp", label_entry_50); + new BranchInst(label_forbody_51, label_afterfor_52, int1_cmp_53, label_entry_50); + + // Block forbody (label_forbody_51) + Argument* fwdref_56 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_55 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_51); + int32_i_0_reg2mem_0_55->reserveOperandSpace(2); + int32_i_0_reg2mem_0_55->addIncoming(const_int32_19, label_entry_50); + int32_i_0_reg2mem_0_55->addIncoming(fwdref_56, label_forbody_51); + + std::vector ptr_arraydecay_57_indices; + ptr_arraydecay_57_indices.push_back(int32_i_0_reg2mem_0_55); + ptr_arraydecay_57_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay_57 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_57_indices.begin(), ptr_arraydecay_57_indices.end(), "arraydecay", label_forbody_51); + GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_55, "arrayidx6", label_forbody_51); + LoadInst* packed_tmp7_58 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_51); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_58, const_int32_19, "tmp11", label_forbody_51); + StoreInst* void_59 = new StoreInst(float_tmp11, ptr_arraydecay_57, false, label_forbody_51); + std::vector ptr_arrayidx13_indices; + ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_55); + ptr_arrayidx13_indices.push_back(const_int32_21); + Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_51); + ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_58, const_int32_21, "tmp15", label_forbody_51); + StoreInst* void_60 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_51); + std::vector ptr_arrayidx17_indices; + ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_55); + ptr_arrayidx17_indices.push_back(const_int32_22); + Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_51); + ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_58, const_int32_22, "tmp19", label_forbody_51); + StoreInst* void_61 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_51); + std::vector ptr_arrayidx21_indices; + ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_55); + ptr_arrayidx21_indices.push_back(const_int32_23); + Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_51); + ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_58, const_int32_23, "tmp23", label_forbody_51); + StoreInst* void_62 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_51); + BinaryOperator* int32_indvar_next_63 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_55, const_int32_21, "indvar.next", label_forbody_51); + ICmpInst* int1_exitcond_64 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_63, int32_num_attribs_49, "exitcond", label_forbody_51); + new BranchInst(label_afterfor_52, label_forbody_51, int1_exitcond_64, label_forbody_51); + + // Block afterfor (label_afterfor_52) + new ReturnInst(label_afterfor_52); // Resolve Forward References - fwdref_78->replaceAllUsesWith(int32_indvar_next_88); delete fwdref_78; + fwdref_56->replaceAllUsesWith(int32_indvar_next_63); delete fwdref_56; } // Function: run_vertex_shader (func_run_vertex_shader) { Function::arg_iterator args = func_run_vertex_shader->arg_begin(); - Value* ptr_ainputs_92 = args++; - ptr_ainputs_92->setName("ainputs"); - Value* ptr_dests_93 = args++; - ptr_dests_93->setName("dests"); + Value* ptr_ainputs_67 = args++; + ptr_ainputs_67->setName("ainputs"); + Value* ptr_dests_68 = args++; + ptr_dests_68->setName("dests"); Value* ptr_aconsts = args++; ptr_aconsts->setName("aconsts"); Value* int32_num_vertices = args++; int32_num_vertices->setName("num_vertices"); Value* int32_num_inputs = args++; int32_num_inputs->setName("num_inputs"); - Value* int32_num_attribs_94 = args++; - int32_num_attribs_94->setName("num_attribs"); + Value* int32_num_attribs_69 = args++; + int32_num_attribs_69->setName("num_attribs"); Value* int32_num_consts = args++; int32_num_consts->setName("num_consts"); - BasicBlock* label_entry_95 = new BasicBlock("entry",func_run_vertex_shader,0); + BasicBlock* label_entry_70 = new BasicBlock("entry",func_run_vertex_shader,0); BasicBlock* label_forbody6_i = new BasicBlock("forbody6.i",func_run_vertex_shader,0); - BasicBlock* label_forinc70_i = new BasicBlock("forinc70.i",func_run_vertex_shader,0); + BasicBlock* label_forinc57_i = new BasicBlock("forinc57.i",func_run_vertex_shader,0); BasicBlock* label_from_array_exit = new BasicBlock("from_array.exit",func_run_vertex_shader,0); BasicBlock* label_forbody_i15 = new BasicBlock("forbody.i15",func_run_vertex_shader,0); - BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_vertex_shader,0); + BasicBlock* label_forcond = new BasicBlock("forcond",func_run_vertex_shader,0); BasicBlock* label_forbody_preheader = new BasicBlock("forbody.preheader",func_run_vertex_shader,0); BasicBlock* label_forbody_us = new BasicBlock("forbody.us",func_run_vertex_shader,0); BasicBlock* label_to_array_exit_us = new BasicBlock("to_array.exit.us",func_run_vertex_shader,0); BasicBlock* label_forbody_i_us = new BasicBlock("forbody.i.us",func_run_vertex_shader,0); - BasicBlock* label_forbody_96 = new BasicBlock("forbody",func_run_vertex_shader,0); - BasicBlock* label_afterfor_97 = new BasicBlock("afterfor",func_run_vertex_shader,0); - - // Block entry (label_entry_95) - AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_25, "inputs", label_entry_95); - AllocaInst* ptr_consts = new AllocaInst(ArrayTy_27, "consts", label_entry_95); - AllocaInst* ptr_results = new AllocaInst(ArrayTy_25, "results", label_entry_95); - std::vector int32_call_98_params; - int32_call_98_params.push_back(const_ptr_44); - int32_call_98_params.push_back(int32_num_vertices); - int32_call_98_params.push_back(int32_num_inputs); - int32_call_98_params.push_back(int32_num_consts); - CallInst* int32_call_98 = new CallInst(func_printf, int32_call_98_params.begin(), int32_call_98_params.end(), "call", label_entry_95); - int32_call_98->setCallingConv(CallingConv::C); - int32_call_98->setTailCall(false); - ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_36, "cmp.i", label_entry_95); - ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_36, "cmp5.i", label_entry_95); - BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_95); - new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_95); + BasicBlock* label_forbody_71 = new BasicBlock("forbody",func_run_vertex_shader,0); + BasicBlock* label_afterfor_72 = new BasicBlock("afterfor",func_run_vertex_shader,0); + + // Block entry (label_entry_70) + AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_14, "inputs", label_entry_70); + AllocaInst* ptr_consts = new AllocaInst(ArrayTy_16, "consts", label_entry_70); + AllocaInst* ptr_results = new AllocaInst(ArrayTy_14, "results", label_entry_70); + ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_19, "cmp.i", label_entry_70); + ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_19, "cmp5.i", label_entry_70); + BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_70); + new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_70); // Block forbody6.i (label_forbody6_i) - Argument* fwdref_100 = new Argument(IntegerType::get(32)); - Argument* fwdref_101 = new Argument(IntegerType::get(32)); - PHINode* int32_i_014_0_i_ph = new PHINode(IntegerType::get(32), "i.014.0.i.ph", label_forbody6_i); - int32_i_014_0_i_ph->reserveOperandSpace(3); - int32_i_014_0_i_ph->addIncoming(const_int32_36, label_entry_95); - int32_i_014_0_i_ph->addIncoming(fwdref_100, label_forinc70_i); - int32_i_014_0_i_ph->addIncoming(fwdref_101, label_forbody6_i); - - Argument* fwdref_102 = new Argument(IntegerType::get(32)); - PHINode* int32_j_03_0_i = new PHINode(IntegerType::get(32), "j.03.0.i", label_forbody6_i); - int32_j_03_0_i->reserveOperandSpace(3); - int32_j_03_0_i->addIncoming(fwdref_102, label_forbody6_i); - int32_j_03_0_i->addIncoming(const_int32_36, label_forinc70_i); - int32_j_03_0_i->addIncoming(const_int32_36, label_entry_95); - - Argument* fwdref_103 = new Argument(VectorTy_11); - PHINode* packed_vec_01_0_i = new PHINode(VectorTy_11, "vec.01.0.i", label_forbody6_i); - packed_vec_01_0_i->reserveOperandSpace(3); - packed_vec_01_0_i->addIncoming(fwdref_103, label_forbody6_i); - packed_vec_01_0_i->addIncoming(const_packed_37, label_entry_95); - packed_vec_01_0_i->addIncoming(fwdref_103, label_forinc70_i); + Argument* fwdref_74 = new Argument(IntegerType::get(32)); + Argument* fwdref_75 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_i_ph = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i.ph", label_forbody6_i); + int32_i_0_reg2mem_0_i_ph->reserveOperandSpace(3); + int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_19, label_entry_70); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_74, label_forinc57_i); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_75, label_forbody6_i); + + Argument* fwdref_76 = new Argument(IntegerType::get(32)); + PHINode* int32_j_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i); + int32_j_0_reg2mem_0_i->reserveOperandSpace(3); + int32_j_0_reg2mem_0_i->addIncoming(fwdref_76, label_forbody6_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_19, label_forinc57_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_19, label_entry_70); + + Argument* fwdref_77 = new Argument(VectorTy_3); + PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_3, "vec.0.reg2mem.0.i", label_forbody6_i); + packed_vec_0_reg2mem_0_i->reserveOperandSpace(3); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_77, label_forbody6_i); + packed_vec_0_reg2mem_0_i->addIncoming(const_packed_20, label_entry_70); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_77, label_forinc57_i); std::vector ptr_arraydecay11_i_indices; - ptr_arraydecay11_i_indices.push_back(int32_i_014_0_i_ph); - ptr_arraydecay11_i_indices.push_back(int32_j_03_0_i); - ptr_arraydecay11_i_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_92, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); + ptr_arraydecay11_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); + ptr_arraydecay11_i_indices.push_back(int32_j_0_reg2mem_0_i); + ptr_arraydecay11_i_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_67, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); LoadInst* float_tmp13_i = new LoadInst(ptr_arraydecay11_i, "tmp13.i", false, label_forbody6_i); - InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_01_0_i, float_tmp13_i, const_int32_36, "tmp15.i", label_forbody6_i); + InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_19, "tmp15.i", label_forbody6_i); std::vector ptr_arrayidx23_i_indices; - ptr_arrayidx23_i_indices.push_back(int32_i_014_0_i_ph); - ptr_arrayidx23_i_indices.push_back(int32_j_03_0_i); - ptr_arrayidx23_i_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_92, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); + ptr_arrayidx23_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); + ptr_arrayidx23_i_indices.push_back(int32_j_0_reg2mem_0_i); + ptr_arrayidx23_i_indices.push_back(const_int32_21); + Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_67, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); LoadInst* float_tmp24_i = new LoadInst(ptr_arrayidx23_i, "tmp24.i", false, label_forbody6_i); - InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_38, "tmp26.i", label_forbody6_i); + InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_21, "tmp26.i", label_forbody6_i); std::vector ptr_arrayidx34_i_indices; - ptr_arrayidx34_i_indices.push_back(int32_i_014_0_i_ph); - ptr_arrayidx34_i_indices.push_back(int32_j_03_0_i); - ptr_arrayidx34_i_indices.push_back(const_int32_39); - Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_92, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); + ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); + ptr_arrayidx34_i_indices.push_back(int32_j_0_reg2mem_0_i); + ptr_arrayidx34_i_indices.push_back(const_int32_22); + Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_67, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); LoadInst* float_tmp35_i = new LoadInst(ptr_arrayidx34_i, "tmp35.i", false, label_forbody6_i); - InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_39, "tmp37.i", label_forbody6_i); + InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_22, "tmp37.i", label_forbody6_i); std::vector ptr_arrayidx45_i_indices; - ptr_arrayidx45_i_indices.push_back(int32_i_014_0_i_ph); - ptr_arrayidx45_i_indices.push_back(int32_j_03_0_i); - ptr_arrayidx45_i_indices.push_back(const_int32_40); - Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_92, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); + ptr_arrayidx45_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); + ptr_arrayidx45_i_indices.push_back(int32_j_0_reg2mem_0_i); + ptr_arrayidx45_i_indices.push_back(const_int32_23); + Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_67, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); LoadInst* float_tmp46_i = new LoadInst(ptr_arrayidx45_i, "tmp46.i", false, label_forbody6_i); - InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_40, "tmp48.i", label_forbody6_i); - CastInst* double_conv_i = new FPExtInst(float_tmp13_i, Type::DoubleTy, "conv.i", label_forbody6_i); - CastInst* double_conv55_i = new FPExtInst(float_tmp24_i, Type::DoubleTy, "conv55.i", label_forbody6_i); - CastInst* double_conv58_i = new FPExtInst(float_tmp35_i, Type::DoubleTy, "conv58.i", label_forbody6_i); - CastInst* double_conv61_i = new FPExtInst(float_tmp46_i, Type::DoubleTy, "conv61.i", label_forbody6_i); - std::vector int32_call_i_params; - int32_call_i_params.push_back(const_ptr_41); - int32_call_i_params.push_back(int32_i_014_0_i_ph); - int32_call_i_params.push_back(int32_j_03_0_i); - int32_call_i_params.push_back(double_conv_i); - int32_call_i_params.push_back(double_conv55_i); - int32_call_i_params.push_back(double_conv58_i); - int32_call_i_params.push_back(double_conv61_i); - CallInst* int32_call_i = new CallInst(func_printf, int32_call_i_params.begin(), int32_call_i_params.end(), "call.i", label_forbody6_i); - int32_call_i->setCallingConv(CallingConv::C); - int32_call_i->setTailCall(false); - std::vector ptr_arrayidx67_i_indices; - ptr_arrayidx67_i_indices.push_back(const_int32_36); - ptr_arrayidx67_i_indices.push_back(int32_i_014_0_i_ph); - ptr_arrayidx67_i_indices.push_back(int32_j_03_0_i); - Instruction* ptr_arrayidx67_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx67_i_indices.begin(), ptr_arrayidx67_i_indices.end(), "arrayidx67.i", label_forbody6_i); - StoreInst* void_104 = new StoreInst(packed_tmp48_i, ptr_arrayidx67_i, false, label_forbody6_i); - BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_03_0_i, const_int32_38, "inc.i", label_forbody6_i); - ICmpInst* int1_cmp512_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_inputs, "cmp512.i", label_forbody6_i); - new BranchInst(label_forbody6_i, label_forinc70_i, int1_cmp512_i, label_forbody6_i); - - // Block forinc70.i (label_forinc70_i) - BinaryOperator* int32_inc72_i = BinaryOperator::create(Instruction::Add, int32_i_014_0_i_ph, const_int32_38, "inc72.i", label_forinc70_i); - ICmpInst* int1_cmp23_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc72_i, int32_num_vertices, "cmp23.i", label_forinc70_i); - new BranchInst(label_forbody6_i, label_from_array_exit, int1_cmp23_i, label_forinc70_i); + InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_23, "tmp48.i", label_forbody6_i); + std::vector ptr_arrayidx54_i_indices; + ptr_arrayidx54_i_indices.push_back(const_int32_19); + ptr_arrayidx54_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); + ptr_arrayidx54_i_indices.push_back(int32_j_0_reg2mem_0_i); + Instruction* ptr_arrayidx54_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx54_i_indices.begin(), ptr_arrayidx54_i_indices.end(), "arrayidx54.i", label_forbody6_i); + StoreInst* void_78 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); + BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_21, "inc.i", label_forbody6_i); + ICmpInst* int1_cmp59_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_inputs, "cmp59.i", label_forbody6_i); + new BranchInst(label_forbody6_i, label_forinc57_i, int1_cmp59_i, label_forbody6_i); + + // Block forinc57.i (label_forinc57_i) + BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_21, "inc59.i", label_forinc57_i); + ICmpInst* int1_cmp17_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59_i, int32_num_vertices, "cmp17.i", label_forinc57_i); + new BranchInst(label_forbody6_i, label_from_array_exit, int1_cmp17_i, label_forinc57_i); // Block from_array.exit (label_from_array_exit) - ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_36, "cmp.i4", label_from_array_exit); - new BranchInst(label_forbody_i15, label_from_consts_exit, int1_cmp_i4, label_from_array_exit); + ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_19, "cmp.i4", label_from_array_exit); + new BranchInst(label_forbody_i15, label_forcond, int1_cmp_i4, label_from_array_exit); // Block forbody.i15 (label_forbody_i15) - Argument* fwdref_108 = new Argument(IntegerType::get(32)); - PHINode* int32_i_02_0_i = new PHINode(IntegerType::get(32), "i.02.0.i", label_forbody_i15); - int32_i_02_0_i->reserveOperandSpace(2); - int32_i_02_0_i->addIncoming(const_int32_36, label_from_array_exit); - int32_i_02_0_i->addIncoming(fwdref_108, label_forbody_i15); - - Argument* fwdref_109 = new Argument(VectorTy_11); - PHINode* packed_vec_01_0_i5 = new PHINode(VectorTy_11, "vec.01.0.i5", label_forbody_i15); - packed_vec_01_0_i5->reserveOperandSpace(2); - packed_vec_01_0_i5->addIncoming(const_packed_37, label_from_array_exit); - packed_vec_01_0_i5->addIncoming(fwdref_109, label_forbody_i15); - - std::vector ptr_arraydecay_i6_indices; - ptr_arraydecay_i6_indices.push_back(int32_i_02_0_i); - ptr_arraydecay_i6_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay_i6 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i6_indices.begin(), ptr_arraydecay_i6_indices.end(), "arraydecay.i6", label_forbody_i15); - LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i6, "tmp5.i", false, label_forbody_i15); - InsertElementInst* packed_tmp7_i7 = new InsertElementInst(packed_vec_01_0_i5, float_tmp5_i, const_int32_36, "tmp7.i7", label_forbody_i15); + Argument* fwdref_82 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_i5 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i5", label_forbody_i15); + int32_i_0_reg2mem_0_i5->reserveOperandSpace(2); + int32_i_0_reg2mem_0_i5->addIncoming(const_int32_19, label_from_array_exit); + int32_i_0_reg2mem_0_i5->addIncoming(fwdref_82, label_forbody_i15); + + Argument* fwdref_83 = new Argument(VectorTy_3); + PHINode* packed_vec_0_reg2mem_0_i6 = new PHINode(VectorTy_3, "vec.0.reg2mem.0.i6", label_forbody_i15); + packed_vec_0_reg2mem_0_i6->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_20, label_from_array_exit); + packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_83, label_forbody_i15); + + std::vector ptr_arraydecay_i7_indices; + ptr_arraydecay_i7_indices.push_back(int32_i_0_reg2mem_0_i5); + ptr_arraydecay_i7_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay_i7 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i7_indices.begin(), ptr_arraydecay_i7_indices.end(), "arraydecay.i7", label_forbody_i15); + LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i7, "tmp5.i", false, label_forbody_i15); + InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_19, "tmp7.i8", label_forbody_i15); std::vector ptr_arrayidx12_i_indices; - ptr_arrayidx12_i_indices.push_back(int32_i_02_0_i); - ptr_arrayidx12_i_indices.push_back(const_int32_38); + ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i5); + ptr_arrayidx12_i_indices.push_back(const_int32_21); Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i15); - LoadInst* float_tmp13_i8 = new LoadInst(ptr_arrayidx12_i, "tmp13.i8", false, label_forbody_i15); - InsertElementInst* packed_tmp15_i9 = new InsertElementInst(packed_tmp7_i7, float_tmp13_i8, const_int32_38, "tmp15.i9", label_forbody_i15); + LoadInst* float_tmp13_i9 = new LoadInst(ptr_arrayidx12_i, "tmp13.i9", false, label_forbody_i15); + InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_21, "tmp15.i10", label_forbody_i15); std::vector ptr_arrayidx20_i_indices; - ptr_arrayidx20_i_indices.push_back(int32_i_02_0_i); - ptr_arrayidx20_i_indices.push_back(const_int32_39); + ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i5); + ptr_arrayidx20_i_indices.push_back(const_int32_22); Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i15); LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i15); - InsertElementInst* packed_tmp23_i = new InsertElementInst(packed_tmp15_i9, float_tmp21_i, const_int32_39, "tmp23.i", label_forbody_i15); - std::vector ptr_arrayidx28_i10_indices; - ptr_arrayidx28_i10_indices.push_back(int32_i_02_0_i); - ptr_arrayidx28_i10_indices.push_back(const_int32_40); - Instruction* ptr_arrayidx28_i10 = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i10_indices.begin(), ptr_arrayidx28_i10_indices.end(), "arrayidx28.i10", label_forbody_i15); - LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i10, "tmp29.i", false, label_forbody_i15); - InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i, float_tmp29_i, const_int32_40, "tmp31.i", label_forbody_i15); - CastInst* double_conv_i11 = new FPExtInst(float_tmp5_i, Type::DoubleTy, "conv.i11", label_forbody_i15); - CastInst* double_conv37_i = new FPExtInst(float_tmp13_i8, Type::DoubleTy, "conv37.i", label_forbody_i15); - CastInst* double_conv40_i = new FPExtInst(float_tmp21_i, Type::DoubleTy, "conv40.i", label_forbody_i15); - CastInst* double_conv43_i = new FPExtInst(float_tmp29_i, Type::DoubleTy, "conv43.i", label_forbody_i15); - std::vector int32_call_i12_params; - int32_call_i12_params.push_back(const_ptr_42); - int32_call_i12_params.push_back(int32_i_02_0_i); - int32_call_i12_params.push_back(double_conv_i11); - int32_call_i12_params.push_back(double_conv37_i); - int32_call_i12_params.push_back(double_conv40_i); - int32_call_i12_params.push_back(double_conv43_i); - CallInst* int32_call_i12 = new CallInst(func_printf, int32_call_i12_params.begin(), int32_call_i12_params.end(), "call.i12", label_forbody_i15); - int32_call_i12->setCallingConv(CallingConv::C); - int32_call_i12->setTailCall(false); - std::vector ptr_arrayidx46_i_indices; - ptr_arrayidx46_i_indices.push_back(const_int32_36); - ptr_arrayidx46_i_indices.push_back(int32_i_02_0_i); - Instruction* ptr_arrayidx46_i = new GetElementPtrInst(ptr_consts, ptr_arrayidx46_i_indices.begin(), ptr_arrayidx46_i_indices.end(), "arrayidx46.i", label_forbody_i15); - StoreInst* void_110 = new StoreInst(packed_tmp31_i, ptr_arrayidx46_i, false, label_forbody_i15); - BinaryOperator* int32_indvar_next25 = BinaryOperator::create(Instruction::Add, int32_i_02_0_i, const_int32_38, "indvar.next25", label_forbody_i15); - ICmpInst* int1_exitcond26 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next25, int32_num_consts, "exitcond26", label_forbody_i15); - new BranchInst(label_from_consts_exit, label_forbody_i15, int1_exitcond26, label_forbody_i15); - - // Block from_consts.exit (label_from_consts_exit) - CallInst* int32_call9 = new CallInst(func_printf, const_ptr_45, "call9", label_from_consts_exit); - int32_call9->setCallingConv(CallingConv::C); - int32_call9->setTailCall(false); - new BranchInst(label_forbody_preheader, label_afterfor_97, int1_cmp_i, label_from_consts_exit); + InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_22, "tmp23.i11", label_forbody_i15); + std::vector ptr_arrayidx28_i_indices; + ptr_arrayidx28_i_indices.push_back(int32_i_0_reg2mem_0_i5); + ptr_arrayidx28_i_indices.push_back(const_int32_23); + Instruction* ptr_arrayidx28_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i_indices.begin(), ptr_arrayidx28_i_indices.end(), "arrayidx28.i", label_forbody_i15); + LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i15); + InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_23, "tmp31.i", label_forbody_i15); + std::vector ptr_arrayidx34_i12_indices; + ptr_arrayidx34_i12_indices.push_back(const_int32_19); + ptr_arrayidx34_i12_indices.push_back(int32_i_0_reg2mem_0_i5); + Instruction* ptr_arrayidx34_i12 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i12_indices.begin(), ptr_arrayidx34_i12_indices.end(), "arrayidx34.i12", label_forbody_i15); + StoreInst* void_84 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); + BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_21, "indvar.next24", label_forbody_i15); + ICmpInst* int1_exitcond25 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next24, int32_num_consts, "exitcond25", label_forbody_i15); + new BranchInst(label_forcond, label_forbody_i15, int1_exitcond25, label_forbody_i15); + + // Block forcond (label_forcond) + new BranchInst(label_forbody_preheader, label_afterfor_72, int1_cmp_i, label_forcond); // Block forbody.preheader (label_forbody_preheader) - std::vector ptr_arraydecay21_indices; - ptr_arraydecay21_indices.push_back(const_int32_36); - ptr_arraydecay21_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay21 = new GetElementPtrInst(ptr_consts, ptr_arraydecay21_indices.begin(), ptr_arraydecay21_indices.end(), "arraydecay21", label_forbody_preheader); - ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_94, const_int32_36, "cmp.i1", label_forbody_preheader); - new BranchInst(label_forbody_us, label_forbody_96, int1_cmp_i1, label_forbody_preheader); + std::vector ptr_arraydecay17_indices; + ptr_arraydecay17_indices.push_back(const_int32_19); + ptr_arraydecay17_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay17 = new GetElementPtrInst(ptr_consts, ptr_arraydecay17_indices.begin(), ptr_arraydecay17_indices.end(), "arraydecay17", label_forbody_preheader); + ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_69, const_int32_19, "cmp.i1", label_forbody_preheader); + new BranchInst(label_forbody_us, label_forbody_71, int1_cmp_i1, label_forbody_preheader); // Block forbody.us (label_forbody_us) - Argument* fwdref_114 = new Argument(IntegerType::get(32)); - PHINode* int32_i_016_0_us = new PHINode(IntegerType::get(32), "i.016.0.us", label_forbody_us); - int32_i_016_0_us->reserveOperandSpace(2); - int32_i_016_0_us->addIncoming(const_int32_36, label_forbody_preheader); - int32_i_016_0_us->addIncoming(fwdref_114, label_to_array_exit_us); - + Argument* fwdref_88 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.us", label_forbody_us); + int32_i_0_reg2mem_0_us->reserveOperandSpace(2); + int32_i_0_reg2mem_0_us->addIncoming(const_int32_19, label_forbody_preheader); + int32_i_0_reg2mem_0_us->addIncoming(fwdref_88, label_to_array_exit_us); + + std::vector ptr_arraydecay10_us_indices; + ptr_arraydecay10_us_indices.push_back(const_int32_19); + ptr_arraydecay10_us_indices.push_back(int32_i_0_reg2mem_0_us); + ptr_arraydecay10_us_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay10_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_us_indices.begin(), ptr_arraydecay10_us_indices.end(), "arraydecay10.us", label_forbody_us); std::vector ptr_arraydecay14_us_indices; - ptr_arraydecay14_us_indices.push_back(const_int32_36); - ptr_arraydecay14_us_indices.push_back(int32_i_016_0_us); - ptr_arraydecay14_us_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay14_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay14_us_indices.begin(), ptr_arraydecay14_us_indices.end(), "arraydecay14.us", label_forbody_us); - std::vector ptr_arraydecay18_us_indices; - ptr_arraydecay18_us_indices.push_back(const_int32_36); - ptr_arraydecay18_us_indices.push_back(int32_i_016_0_us); - ptr_arraydecay18_us_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay18_us = new GetElementPtrInst(ptr_results, ptr_arraydecay18_us_indices.begin(), ptr_arraydecay18_us_indices.end(), "arraydecay18.us", label_forbody_us); - std::vector void_115_params; - void_115_params.push_back(ptr_arraydecay18_us); - void_115_params.push_back(ptr_arraydecay14_us); - void_115_params.push_back(ptr_arraydecay21); - CallInst* void_115 = new CallInst(func_execute_shader, void_115_params.begin(), void_115_params.end(), "", label_forbody_us); - void_115->setCallingConv(CallingConv::C); - void_115->setTailCall(false); - CallInst* int32_call22_us = new CallInst(func_printf, const_ptr_46, "call22.us", label_forbody_us); - int32_call22_us->setCallingConv(CallingConv::C); - int32_call22_us->setTailCall(false); + ptr_arraydecay14_us_indices.push_back(const_int32_19); + ptr_arraydecay14_us_indices.push_back(int32_i_0_reg2mem_0_us); + ptr_arraydecay14_us_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay14_us = new GetElementPtrInst(ptr_results, ptr_arraydecay14_us_indices.begin(), ptr_arraydecay14_us_indices.end(), "arraydecay14.us", label_forbody_us); + std::vector void_89_params; + void_89_params.push_back(ptr_arraydecay14_us); + void_89_params.push_back(ptr_arraydecay10_us); + void_89_params.push_back(ptr_arraydecay17); + CallInst* void_89 = new CallInst(func_execute_shader, void_89_params.begin(), void_89_params.end(), "", label_forbody_us); + void_89->setCallingConv(CallingConv::C); + void_89->setTailCall(false); new BranchInst(label_forbody_i_us, label_forbody_us); // Block to_array.exit.us (label_to_array_exit_us) - BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_016_0_us, const_int32_38, "inc.us", label_to_array_exit_us); - ICmpInst* int1_cmp22_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_num_vertices, "cmp22.us", label_to_array_exit_us); - new BranchInst(label_forbody_us, label_afterfor_97, int1_cmp22_us, label_to_array_exit_us); + BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_21, "inc.us", label_to_array_exit_us); + ICmpInst* int1_cmp21_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_num_vertices, "cmp21.us", label_to_array_exit_us); + new BranchInst(label_forbody_us, label_afterfor_72, int1_cmp21_us, label_to_array_exit_us); // Block forbody.i.us (label_forbody_i_us) - Argument* fwdref_118 = new Argument(IntegerType::get(32)); - PHINode* int32_i_01_0_i_us = new PHINode(IntegerType::get(32), "i.01.0.i.us", label_forbody_i_us); - int32_i_01_0_i_us->reserveOperandSpace(2); - int32_i_01_0_i_us->addIncoming(const_int32_36, label_forbody_us); - int32_i_01_0_i_us->addIncoming(fwdref_118, label_forbody_i_us); + Argument* fwdref_92 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_i2_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i2.us", label_forbody_i_us); + int32_i_0_reg2mem_0_i2_us->reserveOperandSpace(2); + int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_19, label_forbody_us); + int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_92, label_forbody_i_us); std::vector ptr_arraydecay_i_us_indices; - ptr_arraydecay_i_us_indices.push_back(int32_i_016_0_us); - ptr_arraydecay_i_us_indices.push_back(int32_i_01_0_i_us); - ptr_arraydecay_i_us_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_93, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); + ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_us); + ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); + ptr_arraydecay_i_us_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); std::vector ptr_arrayidx6_i_us_indices; - ptr_arrayidx6_i_us_indices.push_back(const_int32_36); - ptr_arrayidx6_i_us_indices.push_back(int32_i_016_0_us); - ptr_arrayidx6_i_us_indices.push_back(int32_i_01_0_i_us); + ptr_arrayidx6_i_us_indices.push_back(const_int32_19); + ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_us); + ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); Instruction* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_results, ptr_arrayidx6_i_us_indices.begin(), ptr_arrayidx6_i_us_indices.end(), "arrayidx6.i.us", label_forbody_i_us); LoadInst* packed_tmp7_i_us = new LoadInst(ptr_arrayidx6_i_us, "tmp7.i.us", false, label_forbody_i_us); - ExtractElementInst* float_tmp9_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_36, "tmp9.i.us", label_forbody_i_us); - CastInst* double_conv_i2_us = new FPExtInst(float_tmp9_i_us, Type::DoubleTy, "conv.i2.us", label_forbody_i_us); - ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_38, "tmp11.i.us", label_forbody_i_us); - CastInst* double_conv12_i_us = new FPExtInst(float_tmp11_i_us, Type::DoubleTy, "conv12.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp14_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_39, "tmp14.i.us", label_forbody_i_us); - CastInst* double_conv15_i_us = new FPExtInst(float_tmp14_i_us, Type::DoubleTy, "conv15.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp17_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_40, "tmp17.i.us", label_forbody_i_us); - CastInst* double_conv18_i_us = new FPExtInst(float_tmp17_i_us, Type::DoubleTy, "conv18.i.us", label_forbody_i_us); - std::vector int32_call_i3_us_params; - int32_call_i3_us_params.push_back(const_ptr_43); - int32_call_i3_us_params.push_back(double_conv_i2_us); - int32_call_i3_us_params.push_back(double_conv12_i_us); - int32_call_i3_us_params.push_back(double_conv15_i_us); - int32_call_i3_us_params.push_back(double_conv18_i_us); - CallInst* int32_call_i3_us = new CallInst(func_printf, int32_call_i3_us_params.begin(), int32_call_i3_us_params.end(), "call.i3.us", label_forbody_i_us); - int32_call_i3_us->setCallingConv(CallingConv::C); - int32_call_i3_us->setTailCall(false); - StoreInst* void_119 = new StoreInst(float_tmp9_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); - std::vector ptr_arrayidx24_i_us_indices; - ptr_arrayidx24_i_us_indices.push_back(int32_i_016_0_us); - ptr_arrayidx24_i_us_indices.push_back(int32_i_01_0_i_us); - ptr_arrayidx24_i_us_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx24_i_us = new GetElementPtrInst(ptr_dests_93, ptr_arrayidx24_i_us_indices.begin(), ptr_arrayidx24_i_us_indices.end(), "arrayidx24.i.us", label_forbody_i_us); - StoreInst* void_120 = new StoreInst(float_tmp11_i_us, ptr_arrayidx24_i_us, false, label_forbody_i_us); - std::vector ptr_arrayidx28_i_us_indices; - ptr_arrayidx28_i_us_indices.push_back(int32_i_016_0_us); - ptr_arrayidx28_i_us_indices.push_back(int32_i_01_0_i_us); - ptr_arrayidx28_i_us_indices.push_back(const_int32_39); - Instruction* ptr_arrayidx28_i_us = new GetElementPtrInst(ptr_dests_93, ptr_arrayidx28_i_us_indices.begin(), ptr_arrayidx28_i_us_indices.end(), "arrayidx28.i.us", label_forbody_i_us); - StoreInst* void_121 = new StoreInst(float_tmp14_i_us, ptr_arrayidx28_i_us, false, label_forbody_i_us); - std::vector ptr_arrayidx32_i_us_indices; - ptr_arrayidx32_i_us_indices.push_back(int32_i_016_0_us); - ptr_arrayidx32_i_us_indices.push_back(int32_i_01_0_i_us); - ptr_arrayidx32_i_us_indices.push_back(const_int32_40); - Instruction* ptr_arrayidx32_i_us = new GetElementPtrInst(ptr_dests_93, ptr_arrayidx32_i_us_indices.begin(), ptr_arrayidx32_i_us_indices.end(), "arrayidx32.i.us", label_forbody_i_us); - StoreInst* void_122 = new StoreInst(float_tmp17_i_us, ptr_arrayidx32_i_us, false, label_forbody_i_us); - BinaryOperator* int32_indvar_next_123 = BinaryOperator::create(Instruction::Add, int32_i_01_0_i_us, const_int32_38, "indvar.next", label_forbody_i_us); - ICmpInst* int1_exitcond_124 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_123, int32_num_attribs_94, "exitcond", label_forbody_i_us); - new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_124, label_forbody_i_us); - - // Block forbody (label_forbody_96) - Argument* fwdref_126 = new Argument(IntegerType::get(32)); - PHINode* int32_i_016_0 = new PHINode(IntegerType::get(32), "i.016.0", label_forbody_96); - int32_i_016_0->reserveOperandSpace(2); - int32_i_016_0->addIncoming(const_int32_36, label_forbody_preheader); - int32_i_016_0->addIncoming(fwdref_126, label_forbody_96); - + ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_19, "tmp11.i.us", label_forbody_i_us); + StoreInst* void_93 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); + std::vector ptr_arrayidx13_i_us_indices; + ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_us); + ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); + ptr_arrayidx13_i_us_indices.push_back(const_int32_21); + Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_21, "tmp15.i3.us", label_forbody_i_us); + StoreInst* void_94 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); + std::vector ptr_arrayidx17_i_us_indices; + ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_us); + ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); + ptr_arrayidx17_i_us_indices.push_back(const_int32_22); + Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_22, "tmp19.i.us", label_forbody_i_us); + StoreInst* void_95 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); + std::vector ptr_arrayidx21_i_us_indices; + ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_us); + ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); + ptr_arrayidx21_i_us_indices.push_back(const_int32_23); + Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_23, "tmp23.i.us", label_forbody_i_us); + StoreInst* void_96 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); + BinaryOperator* int32_indvar_next_97 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_21, "indvar.next", label_forbody_i_us); + ICmpInst* int1_exitcond_98 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_97, int32_num_attribs_69, "exitcond", label_forbody_i_us); + new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_98, label_forbody_i_us); + + // Block forbody (label_forbody_71) + Argument* fwdref_101 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_100 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_71); + int32_i_0_reg2mem_0_100->reserveOperandSpace(2); + int32_i_0_reg2mem_0_100->addIncoming(const_int32_19, label_forbody_preheader); + int32_i_0_reg2mem_0_100->addIncoming(fwdref_101, label_forbody_71); + + std::vector ptr_arraydecay10_indices; + ptr_arraydecay10_indices.push_back(const_int32_19); + ptr_arraydecay10_indices.push_back(int32_i_0_reg2mem_0_100); + ptr_arraydecay10_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_forbody_71); std::vector ptr_arraydecay14_indices; - ptr_arraydecay14_indices.push_back(const_int32_36); - ptr_arraydecay14_indices.push_back(int32_i_016_0); - ptr_arraydecay14_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_96); - std::vector ptr_arraydecay18_indices; - ptr_arraydecay18_indices.push_back(const_int32_36); - ptr_arraydecay18_indices.push_back(int32_i_016_0); - ptr_arraydecay18_indices.push_back(const_int32_36); - Instruction* ptr_arraydecay18 = new GetElementPtrInst(ptr_results, ptr_arraydecay18_indices.begin(), ptr_arraydecay18_indices.end(), "arraydecay18", label_forbody_96); - std::vector void_127_params; - void_127_params.push_back(ptr_arraydecay18); - void_127_params.push_back(ptr_arraydecay14); - void_127_params.push_back(ptr_arraydecay21); - CallInst* void_127 = new CallInst(func_execute_shader, void_127_params.begin(), void_127_params.end(), "", label_forbody_96); - void_127->setCallingConv(CallingConv::C); - void_127->setTailCall(false); - CallInst* int32_call22 = new CallInst(func_printf, const_ptr_46, "call22", label_forbody_96); - int32_call22->setCallingConv(CallingConv::C); - int32_call22->setTailCall(false); - BinaryOperator* int32_inc_128 = BinaryOperator::create(Instruction::Add, int32_i_016_0, const_int32_38, "inc", label_forbody_96); - ICmpInst* int1_cmp22 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_128, int32_num_vertices, "cmp22", label_forbody_96); - new BranchInst(label_forbody_96, label_afterfor_97, int1_cmp22, label_forbody_96); - - // Block afterfor (label_afterfor_97) - new ReturnInst(label_afterfor_97); + ptr_arraydecay14_indices.push_back(const_int32_19); + ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_100); + ptr_arraydecay14_indices.push_back(const_int32_19); + Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_results, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_71); + std::vector void_102_params; + void_102_params.push_back(ptr_arraydecay14); + void_102_params.push_back(ptr_arraydecay10); + void_102_params.push_back(ptr_arraydecay17); + CallInst* void_102 = new CallInst(func_execute_shader, void_102_params.begin(), void_102_params.end(), "", label_forbody_71); + void_102->setCallingConv(CallingConv::C); + void_102->setTailCall(false); + BinaryOperator* int32_inc_103 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_100, const_int32_21, "inc", label_forbody_71); + ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_103, int32_num_vertices, "cmp21", label_forbody_71); + new BranchInst(label_forbody_71, label_afterfor_72, int1_cmp21, label_forbody_71); + + // Block afterfor (label_afterfor_72) + new ReturnInst(label_afterfor_72); // Resolve Forward References - fwdref_101->replaceAllUsesWith(int32_i_014_0_i_ph); delete fwdref_101; - fwdref_103->replaceAllUsesWith(packed_tmp48_i); delete fwdref_103; - fwdref_102->replaceAllUsesWith(int32_inc_i); delete fwdref_102; - fwdref_100->replaceAllUsesWith(int32_inc72_i); delete fwdref_100; - fwdref_109->replaceAllUsesWith(packed_tmp31_i); delete fwdref_109; - fwdref_108->replaceAllUsesWith(int32_indvar_next25); delete fwdref_108; - fwdref_114->replaceAllUsesWith(int32_inc_us); delete fwdref_114; - fwdref_118->replaceAllUsesWith(int32_indvar_next_123); delete fwdref_118; - fwdref_126->replaceAllUsesWith(int32_inc_128); delete fwdref_126; + fwdref_75->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_75; + fwdref_77->replaceAllUsesWith(packed_tmp48_i); delete fwdref_77; + fwdref_76->replaceAllUsesWith(int32_inc_i); delete fwdref_76; + fwdref_74->replaceAllUsesWith(int32_inc59_i); delete fwdref_74; + fwdref_83->replaceAllUsesWith(packed_tmp31_i); delete fwdref_83; + fwdref_82->replaceAllUsesWith(int32_indvar_next24); delete fwdref_82; + fwdref_88->replaceAllUsesWith(int32_inc_us); delete fwdref_88; + fwdref_92->replaceAllUsesWith(int32_indvar_next_97); delete fwdref_92; + fwdref_101->replaceAllUsesWith(int32_inc_103); delete fwdref_101; } diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index 82ae59446e..2269b76ea1 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -102,7 +102,6 @@ void from_array(float4 (*res)[16], float (*ainputs)[16][4], vec.y = ainputs[i][j][1]; vec.z = ainputs[i][j][2]; vec.w = ainputs[i][j][3]; - printf("FRA(%d %d) %f %f %f %f\n", i, j, vec.x, vec.y, vec.z, vec.w); res[i][j] = vec; } } @@ -117,7 +116,6 @@ void from_consts(float4 *res, float (*ainputs)[4], vec.y = ainputs[i][1]; vec.z = ainputs[i][2]; vec.w = ainputs[i][3]; - printf("VCONST(%d) %f %f %f %f\n", i, vec.x, vec.y, vec.z, vec.w); res[i] = vec; } } @@ -127,7 +125,6 @@ void to_array(float (*dests)[4], float4 *in, int num_attribs) for (int i = 0; i < num_attribs; ++i) { float *rd = dests[i]; float4 ri = in[i]; - printf("DEST = %f %f %f %f\n", ri.x, ri.y, ri.z, ri.w); rd[0] = ri.x; rd[1] = ri.y; rd[2] = ri.z; @@ -135,8 +132,8 @@ void to_array(float (*dests)[4], float4 *in, int num_attribs) } } -extern void execute_shader(float4 *dests, float4 *inputs, - float4 *consts); +extern void execute_shader(float4 dests[16], float4 inputs[16], + float4 consts[32]); void run_vertex_shader(float (*ainputs)[16][4], float (*dests)[16][4], @@ -150,16 +147,14 @@ void run_vertex_shader(float (*ainputs)[16][4], float4 consts[32]; float4 results[16*32*4][16]; - printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, consts = %d\n", - num_vertices, num_inputs, num_consts); + /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", + num_vertices, num_inputs, num_attribs, num_consts);*/ from_array(inputs, ainputs, num_vertices, num_inputs); from_consts(consts, aconsts, num_consts); - printf(" after conversion\n"); for (int i = 0; i < num_vertices; ++i) { float4 *in = inputs[i]; float4 *res = results[i]; execute_shader(res, in, consts); - printf("after executing shader\n"); to_array(dests[i], res, num_attribs); } } diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 56b488fb1a..00f8b86efc 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -31,12 +31,17 @@ #include #include #include + +#include #include + using namespace llvm; #include "llvm_base_shader.cpp" +static int GLOBAL_ID = 0; + static inline void addPass(PassManager &PM, Pass *P) { // Add the pass to the pass manager... PM.add(P); @@ -168,8 +173,8 @@ translate_instruction(llvm::Module *module, inputs[i] = val; } - if (inputs[0]) - instr->printVector(inputs[0]); + /*if (inputs[0]) + instr->printVector(inputs[0]);*/ llvm::Value *out = 0; printf("Opcode is %d\n", inst->Instruction.Opcode); switch (inst->Instruction.Opcode) { @@ -523,7 +528,6 @@ translate_instruction(llvm::Module *module, printf("translate instr END\n"); } - static llvm::Module * tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) { @@ -533,7 +537,11 @@ tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) struct tgsi_full_declaration fd; Function* shader = mod->getFunction("execute_shader"); - shader->setName("execute_shader_2"); + std::ostringstream stream; + stream << "execute_shader"; + stream << prog->id; + std::string func_name = stream.str(); + shader->setName(func_name.c_str()); Function::arg_iterator args = shader->arg_begin(); Value *ptr_OUT = args++; @@ -593,8 +601,10 @@ struct ga_llvm_prog * ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) { std::cout << "Creating llvm from: " <id = GLOBAL_ID; fprintf(stderr, "----- TGSI Start ---- \n"); tgsi_dump(tokens, 0); fprintf(stderr, "----- TGSI End ---- \n"); @@ -618,9 +628,7 @@ ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) ga_llvm->module = mod; Function *func = mod->getFunction("run_vertex_shader"); - std::cout << "run_vertex_shader = "<function = ee->getPointerToFunctionOrStub(func); - std::cout << " -- FUNC is " <function<(prog->function); runner(inputs, dests, consts, num_vertices, num_inputs, num_attribs, prog->num_consts); - std::cout << "---- END LLVM Execution "<setAlignment(8); m_inputs[idx] = load; return load; } @@ -93,6 +94,7 @@ llvm::Value *Storage::constElement(int idx) m_block); LoadInst *load = new LoadInst(getElem, name("const"), false, m_block); + load->setAlignment(8); m_consts[idx] = load; return load; } @@ -136,6 +138,7 @@ void Storage::store(int dstIdx, llvm::Value *val, int mask) name("out_ptr"), m_block); StoreInst *st = new StoreInst(val, getElem, false, m_block); + st->setAlignment(8); //m_dstCache[dstIdx] = st; } -- cgit v1.2.3 From b04430efd963ca541c435c6c1007feccf5474040 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 22 Oct 2007 11:00:28 -0400 Subject: Introduce linux-llvm target. Remove the llvm code from the default build for now. Cleanup some of the debugging output. --- src/mesa/pipe/llvm/llvmtgsi.cpp | 50 ++++++++++++++++++++++++++++++++--------- src/mesa/pipe/llvm/llvmtgsi.h | 11 ++++----- 2 files changed, 43 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 00f8b86efc..e1bb281a94 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -33,9 +33,17 @@ #include #include +#include #include +struct ga_llvm_prog { + void *module; + void *function; + int num_consts; + int id; +}; + using namespace llvm; #include "llvm_base_shader.cpp" @@ -134,7 +142,7 @@ translate_instruction(llvm::Module *module, inputs[1] = 0; inputs[2] = 0; inputs[3] = 0; - printf("translate instr START\n"); + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; llvm::Value *val = 0; @@ -176,7 +184,6 @@ translate_instruction(llvm::Module *module, /*if (inputs[0]) instr->printVector(inputs[0]);*/ llvm::Value *out = 0; - printf("Opcode is %d\n", inst->Instruction.Opcode); switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: break; @@ -485,7 +492,6 @@ translate_instruction(llvm::Module *module, case TGSI_OPCODE_KIL: break; case TGSI_OPCODE_END: - printf("translate instr END\n"); return; break; default: @@ -525,7 +531,6 @@ translate_instruction(llvm::Module *module, fprintf(stderr, "ERROR: unsupported LLVM destination!"); } } - printf("translate instr END\n"); } static llvm::Module * @@ -590,10 +595,6 @@ tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) tgsi_parse_free(&parse); prog->num_consts = storage.numConsts(); - - std::cout<<"XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"<id = GLOBAL_ID; - fprintf(stderr, "----- TGSI Start ---- \n"); tgsi_dump(tokens, 0); - fprintf(stderr, "----- TGSI End ---- \n"); llvm::Module *mod = tgsi_to_llvm(ga_llvm, tokens); /* Run optimization passes over it */ @@ -630,6 +629,8 @@ ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) Function *func = mod->getFunction("run_vertex_shader"); ga_llvm->function = ee->getPointerToFunctionOrStub(func); + ga_llvm_prog_dump(ga_llvm, 0); + return ga_llvm; } @@ -638,7 +639,6 @@ void ga_llvm_prog_delete(struct ga_llvm_prog *prog) llvm::Module *mod = static_cast(prog->module); delete mod; prog->module = 0; - prog->engine = 0; prog->function = 0; free(prog); } @@ -665,3 +665,31 @@ int ga_llvm_prog_exec(struct ga_llvm_prog *prog, return 0; } + +void ga_llvm_prog_dump(struct ga_llvm_prog *prog, const char *file_prefix) +{ + llvm::Module *mod; + if (!prog || !prog->module) + return; + + mod = static_cast(prog->module); + + if (file_prefix) { + std::ostringstream stream; + stream << file_prefix; + stream << prog->id; + stream << ".ll"; + std::string name = stream.str(); + std::ofstream out(name.c_str()); + if (!out) { + std::cerr<<"Can't open file : "<id<id< Date: Mon, 22 Oct 2007 11:27:34 -0400 Subject: Implement SGE and SLT --- src/mesa/pipe/llvm/instructions.cpp | 57 +++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 2 ++ src/mesa/pipe/llvm/llvmtgsi.cpp | 8 ++++-- 3 files changed, 65 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 044b447846..1c42a37c74 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -600,6 +600,62 @@ llvm::Function * Instructions::declarePrintf() return func_printf; } +llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), name("x1"), m_block); + ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), name("x2"), m_block); + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGE, x1, x2, name("xcmp"), m_block); + SelectInst *x = new SelectInst(xcmp, const1f, const0f, name("xsel"), m_block); + + ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), name("y1"), m_block); + ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), name("y2"), m_block); + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGE, y1, y2, name("ycmp"), m_block); + SelectInst *y = new SelectInst(ycmp, const1f, const0f, name("ysel"), m_block); + + ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), name("z1"), m_block); + ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), name("z2"), m_block); + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGE, z1, z2, name("zcmp"), m_block); + SelectInst *z = new SelectInst(zcmp, const1f, const0f, name("zsel"), m_block); + + ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), name("w1"), m_block); + ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), name("w2"), m_block); + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGE, w1, w2, name("wcmp"), m_block); + SelectInst *w = new SelectInst(wcmp, const1f, const0f, name("wsel"), m_block); + + return vectorFromVals(x, y, z, w); +} + + +llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), name("x1"), m_block); + ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), name("x2"), m_block); + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OLT, x1, x2, name("xcmp"), m_block); + SelectInst *x = new SelectInst(xcmp, const1f, const0f, name("xsel"), m_block); + + ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), name("y1"), m_block); + ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), name("y2"), m_block); + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OLT, y1, y2, name("ycmp"), m_block); + SelectInst *y = new SelectInst(ycmp, const1f, const0f, name("ysel"), m_block); + + ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), name("z1"), m_block); + ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), name("z2"), m_block); + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OLT, z1, z2, name("zcmp"), m_block); + SelectInst *z = new SelectInst(zcmp, const1f, const0f, name("zsel"), m_block); + + ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), name("w1"), m_block); + ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), name("w2"), m_block); + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OLT, w1, w2, name("wcmp"), m_block); + SelectInst *w = new SelectInst(wcmp, const1f, const0f, name("wsel"), m_block); + + return vectorFromVals(x, y, z, w); +} /* typedef __attribute__(( ocu_vector_type(4) )) float float4; @@ -775,3 +831,4 @@ Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elem } return func_lit; } + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index b4948ed240..557ae3730b 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -33,6 +33,8 @@ public: llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); llvm::Value *rcp(llvm::Value *in); llvm::Value *rsq(llvm::Value *in); + llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); + llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); void printVector(llvm::Value *val); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index e1bb281a94..d40a7cc380 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -235,9 +235,13 @@ translate_instruction(llvm::Module *module, out = instr->max(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_SLT: + case TGSI_OPCODE_SLT: { + out = instr->slt(inputs[0], inputs[1]); + } break; - case TGSI_OPCODE_SGE: + case TGSI_OPCODE_SGE: { + out = instr->sge(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_MAD: { out = instr->madd(inputs[0], inputs[1], inputs[2]); -- cgit v1.2.3 From 1248b9776bfeec1f61962604b21212d2cf336283 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 22 Oct 2007 14:01:52 -0400 Subject: Implement extended swizzling. --- src/mesa/pipe/llvm/llvmtgsi.cpp | 29 ++++++++++++++++++++++++----- src/mesa/pipe/llvm/storage.cpp | 25 ++++++++++++++++++++++++- src/mesa/pipe/llvm/storage.h | 2 ++ 3 files changed, 50 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index d40a7cc380..69d54bf9af 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -162,10 +162,26 @@ translate_instruction(llvm::Module *module, src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { - int swizzle = src->SrcRegisterExtSwz.ExtSwizzleX * 1000; - swizzle += src->SrcRegisterExtSwz.ExtSwizzleY * 100; - swizzle += src->SrcRegisterExtSwz.ExtSwizzleZ * 10; - swizzle += src->SrcRegisterExtSwz.ExtSwizzleW * 1; + int swizzle = 0; + + if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X) + swizzle = src->SrcRegisterExtSwz.ExtSwizzleX * 1000; + else + swizzle = src->SrcRegister.SwizzleX * 1000; + if (src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y) + swizzle += src->SrcRegisterExtSwz.ExtSwizzleY * 100; + else + swizzle += src->SrcRegister.SwizzleY * 100; + if (src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z) + swizzle += src->SrcRegisterExtSwz.ExtSwizzleZ * 10; + else + swizzle += src->SrcRegister.SwizzleZ * 10; + if (src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) + swizzle += src->SrcRegisterExtSwz.ExtSwizzleW * 1; + else + swizzle += src->SrcRegister.SwizzleW * 1; + /*fprintf(stderr, "EXT XXXXXXXX swizzle x = %d\n", swizzle);*/ + val = storage->shuffleVector(val, swizzle); } } else if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || @@ -176,13 +192,16 @@ translate_instruction(llvm::Module *module, swizzle += src->SrcRegister.SwizzleY * 100; swizzle += src->SrcRegister.SwizzleZ * 10; swizzle += src->SrcRegister.SwizzleW * 1; + /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ val = storage->shuffleVector(val, swizzle); } inputs[i] = val; } /*if (inputs[0]) - instr->printVector(inputs[0]);*/ + instr->printVector(inputs[0]); + if (inputs[1]) + instr->printVector(inputs[1]);*/ llvm::Value *out = 0; switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 0fe973b78e..3bf37c7567 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -25,6 +25,7 @@ Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, m_undefFloatVec = UndefValue::get(m_floatVecType); m_undefIntVec = UndefValue::get(m_intVecType); + m_extSwizzleVec = 0; m_numConsts = 0; } @@ -32,6 +33,27 @@ Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, //can only build vectors with all members in the [0, 9] range llvm::Constant *Storage::shuffleMask(int vec) { + if (!m_extSwizzleVec) { + Constant *const_vec = Constant::getNullValue(m_floatVecType); + InsertElementInst *res = new InsertElementInst(const_vec, + ConstantFP::get(Type::FloatTy, APFloat(0.f)), + unsigned(0), + name("extswx"), m_block); + res = new InsertElementInst(res, ConstantFP::get(Type::FloatTy, APFloat(1.f)), + unsigned(1), + name("extswy"), + m_block); + res = new InsertElementInst(res, ConstantFP::get(Type::FloatTy, APFloat(0.f)), + unsigned(2), + name("extswz"), + m_block); + res = new InsertElementInst(res, ConstantFP::get(Type::FloatTy, APFloat(1.f)), + unsigned(3), + name("extsww"), + m_block); + m_extSwizzleVec = res; + } + if (m_intVecs.find(vec) != m_intVecs.end()) { return m_intVecs[vec]; } @@ -79,6 +101,7 @@ llvm::Value *Storage::inputElement(int idx) false, m_block); load->setAlignment(8); m_inputs[idx] = load; + return load; } @@ -103,7 +126,7 @@ llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) { Constant *mask = shuffleMask(shuffle); ShuffleVectorInst *res = - new ShuffleVectorInst(vec, m_undefFloatVec, mask, + new ShuffleVectorInst(vec, m_extSwizzleVec, mask, name("shuffle"), m_block); return res; } diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index 66a3363332..da7b0dbe01 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -57,6 +57,8 @@ private: llvm::Value *m_undefFloatVec; llvm::Value *m_undefIntVec; + llvm::Value *m_extSwizzleVec; + char m_name[32]; int m_idx; -- cgit v1.2.3 From ba823b3ded1b6ec47b8a0e26ed08a229fe1a9140 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 23 Oct 2007 07:11:39 -0400 Subject: Implement cross product and abs opcode --- src/mesa/pipe/llvm/instructions.cpp | 57 +++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 2 ++ src/mesa/pipe/llvm/llvmtgsi.cpp | 8 ++++-- 3 files changed, 65 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 1c42a37c74..78d2658724 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -657,6 +657,61 @@ llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) return vectorFromVals(x, y, z, w); } +llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) +{ + ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), + name("x1"), + m_block); + ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), + name("y1"), + m_block); + ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), + name("z1"), + m_block); + + ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), + name("x2"), + m_block); + ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), + name("y2"), + m_block); + ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), + name("z2"), + m_block); + Value *y1z2 = mul(y1, z2); + Value *z1y2 = mul(z1, y2); + + Value *z1x2 = mul(z1, x2); + Value *x1z2 = mul(x1, z2); + + Value *x1y2 = mul(x1, y2); + Value *y1x2 = mul(y1, x2); + + return vectorFromVals(sub(y1z2, z1y2), sub(z1x2, x1z2), sub(x1y2, y1x2)); +} + + +llvm::Value * Instructions::abs(llvm::Value *in) +{ + ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), + name("extractx"), + m_block); + ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), + name("extracty"), + m_block); + ExtractElementInst *z = new ExtractElementInst(in, unsigned(2), + name("extractz"), + m_block); + ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), + name("extractw"), + m_block); + Value *xabs = callFAbs(x); + Value *yabs = callFAbs(y); + Value *zabs = callFAbs(z); + Value *wabs = callFAbs(w); + return vectorFromVals(xabs, yabs, zabs, wabs); +} + /* typedef __attribute__(( ocu_vector_type(4) )) float float4; @@ -832,3 +887,5 @@ Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elem return func_lit; } + + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 557ae3730b..f7bc4a5839 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -15,7 +15,9 @@ class Instructions public: Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block); + llvm::Value *abs(llvm::Value *in1); llvm::Value *add(llvm::Value *in1, llvm::Value *in2); + llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 69d54bf9af..8983f16297 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -306,11 +306,15 @@ translate_instruction(llvm::Module *module, out = instr->pow(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_CROSSPRODUCT: + case TGSI_OPCODE_CROSSPRODUCT: { + out = instr->cross(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_MULTIPLYMATRIX: break; - case TGSI_OPCODE_ABS: + case TGSI_OPCODE_ABS: { + out = instr->abs(inputs[0]); + } break; case TGSI_OPCODE_RCC: break; -- cgit v1.2.3 From 5040eefbb758fb0e02125ad3a6bfb3dba13cb7fa Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 24 Oct 2007 07:49:24 -0400 Subject: Implement arl, lerp opcodes and do a first stab at if/endif handling and branching support. --- src/mesa/pipe/llvm/instructions.cpp | 304 +++++++++++++++++++++++------------- src/mesa/pipe/llvm/instructions.h | 12 ++ src/mesa/pipe/llvm/llvmtgsi.cpp | 51 ++++-- src/mesa/pipe/llvm/storage.cpp | 130 +++++++++++++-- src/mesa/pipe/llvm/storage.h | 26 ++- 5 files changed, 388 insertions(+), 135 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 78d2658724..66c34f8101 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -369,6 +369,11 @@ llvm::Value * Instructions::floor(llvm::Value *in) callFloor(z), callFloor(w)); } +llvm::Value * Instructions::arl(llvm::Value *in) +{ + return floor(in); +} + llvm::Value * Instructions::frc(llvm::Value *in) { llvm::Value *flr = floor(in); @@ -600,6 +605,34 @@ llvm::Function * Instructions::declarePrintf() return func_printf; } + +llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), name("x1"), m_block); + ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), name("x2"), m_block); + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGT, x1, x2, name("xcmp"), m_block); + SelectInst *x = new SelectInst(xcmp, const1f, const0f, name("xsel"), m_block); + + ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), name("y1"), m_block); + ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), name("y2"), m_block); + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGT, y1, y2, name("ycmp"), m_block); + SelectInst *y = new SelectInst(ycmp, const1f, const0f, name("ysel"), m_block); + + ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), name("z1"), m_block); + ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), name("z2"), m_block); + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGT, z1, z2, name("zcmp"), m_block); + SelectInst *z = new SelectInst(zcmp, const1f, const0f, name("zsel"), m_block); + + ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), name("w1"), m_block); + ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), name("w2"), m_block); + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGT, w1, w2, name("wcmp"), m_block); + SelectInst *w = new SelectInst(wcmp, const1f, const0f, name("wsel"), m_block); + + return vectorFromVals(x, y, z, w); +} llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) { Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); @@ -713,6 +746,8 @@ llvm::Value * Instructions::abs(llvm::Value *in) } /* +FIXME: hand write the lit function. Currently it's +generated from: typedef __attribute__(( ocu_vector_type(4) )) float float4; extern float powf(float a, float b); @@ -744,7 +779,7 @@ clang --emit-llvm lit.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -f */ Function* makeLitFunction(Module *mod) { -// Type Definitions + // Type Definitions std::vectorFuncTy_0_args; FuncTy_0_args.push_back(Type::FloatTy); FuncTy_0_args.push_back(Type::FloatTy); @@ -769,123 +804,172 @@ Function* makeLitFunction(Module *mod) { /*ParamAttrs=*/FuncTy_3_PAL); -// Function Declarations + // Function Declarations Function* func_approx = new Function( /*Type=*/FuncTy_0, /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"approx", mod); + /*Name=*/"approx", mod); func_approx->setCallingConv(CallingConv::C); -Function* func_powf = new Function( - /*Type=*/FuncTy_0, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"powf", mod); // (external, no body) -func_powf->setCallingConv(CallingConv::C); - -Function* func_lit = new Function( - /*Type=*/FuncTy_3, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"lit", mod); -func_lit->setCallingConv(CallingConv::C); - -// Global Variable Declarations - - -// Constant Definitions -ConstantFP* const_float_4 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); -ConstantFP* const_float_5 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); -Constant* const_float_6 = Constant::getNullValue(Type::FloatTy); -Constant* const_int32_7 = Constant::getNullValue(IntegerType::get(32)); -std::vector const_packed_8_elems; -ConstantFP* const_float_9 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); -const_packed_8_elems.push_back(const_float_9); -UndefValue* const_float_10 = UndefValue::get(Type::FloatTy); -const_packed_8_elems.push_back(const_float_10); -const_packed_8_elems.push_back(const_float_10); -const_packed_8_elems.push_back(const_float_9); -Constant* const_packed_8 = ConstantVector::get(VectorTy_2, const_packed_8_elems); -ConstantInt* const_int32_11 = ConstantInt::get(APInt(32, "1", 10)); -ConstantInt* const_int32_12 = ConstantInt::get(APInt(32, "3", 10)); -ConstantInt* const_int32_13 = ConstantInt::get(APInt(32, "2", 10)); -std::vector const_packed_14_elems; -const_packed_14_elems.push_back(const_float_9); -const_packed_14_elems.push_back(const_float_6); -const_packed_14_elems.push_back(const_float_6); -const_packed_14_elems.push_back(const_float_9); -Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elems); - -// Global Variable Definitions - -// Function Definitions - -// Function: approx (func_approx) -{ - Function::arg_iterator args = func_approx->arg_begin(); - Value* float_a = args++; - float_a->setName("a"); - Value* float_b = args++; - float_b->setName("b"); - - BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); - - // Block entry (label_entry) - FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_4, "cmp", label_entry); - SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_4, float_b, "b.addr.0", label_entry); - FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_5, "cmp3", label_entry); - SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_5, float_b_addr_0, "b.addr.1", label_entry); - FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_6, "cmp7", label_entry); - SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_6, float_a, "a.addr.0", label_entry); - std::vector float_call_params; - float_call_params.push_back(float_a_addr_0); - float_call_params.push_back(float_b_addr_1); - CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); - float_call->setCallingConv(CallingConv::C); - float_call->setTailCall(true); - new ReturnInst(float_call, label_entry); - -} - -// Function: lit (func_lit) -{ - Function::arg_iterator args = func_lit->arg_begin(); - Value* packed_tmp = args++; - packed_tmp->setName("tmp"); - - BasicBlock* label_entry_16 = new BasicBlock("entry",func_lit,0); - BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); - BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); - - // Block entry (label_entry_16) - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_7, "tmp7", label_entry_16); - FCmpInst* int1_cmp_17 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_6, "cmp", label_entry_16); - new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_17, label_entry_16); - - // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_8, float_tmp7, const_int32_11, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_11, "tmp14", label_ifthen); - ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_12, "tmp16", label_ifthen); - FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_4, "cmp.i", label_ifthen); - SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_4, float_tmp16, "b.addr.0.i", label_ifthen); - FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_5, "cmp3.i", label_ifthen); - SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_5, float_b_addr_0_i, "b.addr.1.i", label_ifthen); - FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_6, "cmp7.i", label_ifthen); - SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_6, float_tmp14, "a.addr.0.i", label_ifthen); - std::vector float_call_i_params; - float_call_i_params.push_back(float_a_addr_0_i); - float_call_i_params.push_back(float_b_addr_1_i); - CallInst* float_call_i = new CallInst(func_powf, float_call_i_params.begin(), float_call_i_params.end(), "call.i", label_ifthen); - float_call_i->setCallingConv(CallingConv::C); - float_call_i->setTailCall(true); - InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_13, "tmp18", label_ifthen); - new ReturnInst(packed_tmp18, label_ifthen); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock) - new ReturnInst(const_packed_14, label_UnifiedReturnBlock); - -} + Function* func_powf = new Function( + /*Type=*/FuncTy_0, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"powf", mod); // (external, no body) + func_powf->setCallingConv(CallingConv::C); + + Function* func_lit = new Function( + /*Type=*/FuncTy_3, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"lit", mod); + func_lit->setCallingConv(CallingConv::C); + + // Global Variable Declarations + + + // Constant Definitions + ConstantFP* const_float_4 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); + ConstantFP* const_float_5 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); + Constant* const_float_6 = Constant::getNullValue(Type::FloatTy); + Constant* const_int32_7 = Constant::getNullValue(IntegerType::get(32)); + std::vector const_packed_8_elems; + ConstantFP* const_float_9 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + const_packed_8_elems.push_back(const_float_9); + UndefValue* const_float_10 = UndefValue::get(Type::FloatTy); + const_packed_8_elems.push_back(const_float_10); + const_packed_8_elems.push_back(const_float_10); + const_packed_8_elems.push_back(const_float_9); + Constant* const_packed_8 = ConstantVector::get(VectorTy_2, const_packed_8_elems); + ConstantInt* const_int32_11 = ConstantInt::get(APInt(32, "1", 10)); + ConstantInt* const_int32_12 = ConstantInt::get(APInt(32, "3", 10)); + ConstantInt* const_int32_13 = ConstantInt::get(APInt(32, "2", 10)); + std::vector const_packed_14_elems; + const_packed_14_elems.push_back(const_float_9); + const_packed_14_elems.push_back(const_float_6); + const_packed_14_elems.push_back(const_float_6); + const_packed_14_elems.push_back(const_float_9); + Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elems); + + // Global Variable Definitions + + // Function Definitions + + // Function: approx (func_approx) + { + Function::arg_iterator args = func_approx->arg_begin(); + Value* float_a = args++; + float_a->setName("a"); + Value* float_b = args++; + float_b->setName("b"); + + BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); + + // Block entry (label_entry) + FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_4, "cmp", label_entry); + SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_4, float_b, "b.addr.0", label_entry); + FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_5, "cmp3", label_entry); + SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_5, float_b_addr_0, "b.addr.1", label_entry); + FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_6, "cmp7", label_entry); + SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_6, float_a, "a.addr.0", label_entry); + std::vector float_call_params; + float_call_params.push_back(float_a_addr_0); + float_call_params.push_back(float_b_addr_1); + CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); + float_call->setCallingConv(CallingConv::C); + float_call->setTailCall(true); + new ReturnInst(float_call, label_entry); + } + + // Function: lit (func_lit) + { + Function::arg_iterator args = func_lit->arg_begin(); + Value* packed_tmp = args++; + packed_tmp->setName("tmp"); + + BasicBlock* label_entry_16 = new BasicBlock("entry",func_lit,0); + BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); + BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); + + // Block entry (label_entry_16) + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_7, "tmp7", label_entry_16); + FCmpInst* int1_cmp_17 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_6, "cmp", label_entry_16); + new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_17, label_entry_16); + + // Block ifthen (label_ifthen) + InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_8, float_tmp7, const_int32_11, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_11, "tmp14", label_ifthen); + ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_12, "tmp16", label_ifthen); + FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_4, "cmp.i", label_ifthen); + SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_4, float_tmp16, "b.addr.0.i", label_ifthen); + FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_5, "cmp3.i", label_ifthen); + SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_5, float_b_addr_0_i, "b.addr.1.i", label_ifthen); + FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_6, "cmp7.i", label_ifthen); + SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_6, float_tmp14, "a.addr.0.i", label_ifthen); + std::vector float_call_i_params; + float_call_i_params.push_back(float_a_addr_0_i); + float_call_i_params.push_back(float_b_addr_1_i); + CallInst* float_call_i = new CallInst(func_powf, float_call_i_params.begin(), float_call_i_params.end(), "call.i", label_ifthen); + float_call_i->setCallingConv(CallingConv::C); + float_call_i->setTailCall(true); + InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_13, "tmp18", label_ifthen); + new ReturnInst(packed_tmp18, label_ifthen); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock) + new ReturnInst(const_packed_14, label_UnifiedReturnBlock); + + } return func_lit; } +void Instructions::ifop(llvm::Value *in) +{ + BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); + BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); + + //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0); + //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0); + //BasicBlock *wblock = new BasicBlock(name("wblock"), m_func,0); + + Constant *float0 = Constant::getNullValue(Type::FloatTy); + + ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), name("extractx"), + m_block); + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_UNE, x, float0, + name("xcmp"), m_block); + new BranchInst(ifend, ifthen, xcmp, m_block); + //m_block = yblock; + + + m_block = ifthen; + m_ifStack.push(ifend); +} + +llvm::BasicBlock * Instructions::currentBlock() const +{ + return m_block; +} +void Instructions::endif() +{ + assert(!m_ifStack.empty()); + new BranchInst(m_ifStack.top(), m_block); + m_block = m_ifStack.top(); + m_ifStack.pop(); +} + +llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3) +{ + llvm::Value *m = mul(in1, in2); + llvm::Value *vec1 = vectorFromVals(ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + ConstantFP::get(Type::FloatTy, + APFloat(1.f))); + llvm::Value *s = sub(vec1, in1); + return add(m, mul(s, in3)); +} diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index f7bc4a5839..ce2ce796d9 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -5,6 +5,8 @@ #include #include +#include + namespace llvm { class VectorType; class Function; @@ -15,16 +17,23 @@ class Instructions public: Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block); + llvm::BasicBlock *currentBlock() const; + llvm::Value *abs(llvm::Value *in1); + llvm::Value *arl(llvm::Value *in1); llvm::Value *add(llvm::Value *in1, llvm::Value *in2); llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); + void endif(); llvm::Value *ex2(llvm::Value *in); llvm::Value *floor(llvm::Value *in); llvm::Value *frc(llvm::Value *in); + void ifop(llvm::Value *in); + llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3); llvm::Value *lit(llvm::Value *in); llvm::Value *lg2(llvm::Value *in); llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, @@ -36,6 +45,7 @@ public: llvm::Value *rcp(llvm::Value *in); llvm::Value *rsq(llvm::Value *in); llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); @@ -70,6 +80,8 @@ private: llvm::Function *m_llvmLit; llvm::Constant *m_fmtPtr; + + std::stack m_ifStack; }; #endif diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 8983f16297..ade4573fb8 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -146,10 +146,16 @@ translate_instruction(llvm::Module *module, for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; llvm::Value *val = 0; + llvm::Value *indIdx = 0; + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + indIdx = storage->extractIndex(indIdx); + } if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->constElement(src->SrcRegister.Index); + val = storage->constElement(src->SrcRegister.Index, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->inputElement(src->SrcRegister.Index); + val = storage->inputElement(src->SrcRegister.Index, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { val = storage->tempElement(src->SrcRegister.Index); } else { @@ -204,7 +210,9 @@ translate_instruction(llvm::Module *module, instr->printVector(inputs[1]);*/ llvm::Value *out = 0; switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: + case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); + } break; case TGSI_OPCODE_MOV: { out = inputs[0]; @@ -270,7 +278,9 @@ translate_instruction(llvm::Module *module, out = instr->sub(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_LERP: + case TGSI_OPCODE_LERP: { + out = instr->lerp(inputs[0], inputs[1], inputs[2]); + } break; case TGSI_OPCODE_CND: break; @@ -344,7 +354,9 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_SFL: break; - case TGSI_OPCODE_SGT: + case TGSI_OPCODE_SGT: { + out = instr->sgt(inputs[0], inputs[1]); + } break; case TGSI_OPCODE_SIN: break; @@ -398,7 +410,11 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_BRK: break; - case TGSI_OPCODE_IF: + case TGSI_OPCODE_IF: { + instr->ifop(inputs[0]); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } break; case TGSI_OPCODE_LOOP: break; @@ -406,7 +422,12 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_ELSE: break; - case TGSI_OPCODE_ENDIF: + case TGSI_OPCODE_ENDIF: { + instr->endif(); + storage->setCurrentBlock(instr->currentBlock()); + storage->popPhiNode(); + return; //just update the state + } break; case TGSI_OPCODE_ENDLOOP: break; @@ -554,8 +575,11 @@ translate_instruction(llvm::Module *module, storage->store(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else { fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); } } } @@ -617,7 +641,7 @@ tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) } } - new ReturnInst(label_entry); + new ReturnInst(instr.currentBlock()); tgsi_parse_free(&parse); @@ -635,7 +659,8 @@ ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) ga_llvm->id = GLOBAL_ID; tgsi_dump(tokens, 0); llvm::Module *mod = tgsi_to_llvm(ga_llvm, tokens); - + ga_llvm->module = mod; + ga_llvm_prog_dump(ga_llvm, 0); /* Run optimization passes over it */ PassManager passes; passes.add(new TargetData(mod)); @@ -715,8 +740,14 @@ void ga_llvm_prog_dump(struct ga_llvm_prog *prog, const char *file_prefix) out << (*mod); out.close(); } else { + std::ostringstream stream; + stream << "execute_shader"; + stream << prog->id; + std::string func_name = stream.str(); + llvm::Function *func = mod->getFunction(func_name.c_str()); + assert(func); std::cout<<"; ---------- Start shader "<id<id<setAlignment(8); @@ -105,16 +120,29 @@ llvm::Value *Storage::inputElement(int idx) return load; } -llvm::Value *Storage::constElement(int idx) +llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) { m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; - if (m_consts.find(idx) != m_consts.end()) { + if (!indIdx && m_consts.find(idx) != m_consts.end()) { return m_consts[idx]; } - GetElementPtrInst *getElem = new GetElementPtrInst(m_CONST, - constantInt(idx), - name("const_ptr"), - m_block); + + GetElementPtrInst *getElem = 0; + + if (indIdx) + getElem = new GetElementPtrInst(m_CONST, + BinaryOperator::create(Instruction::Add, + indIdx, + constantInt(idx), + name("add"), + m_block), + name("const_ptr"), + m_block); + else + getElem = new GetElementPtrInst(m_CONST, + constantInt(idx), + name("const_ptr"), + m_block); LoadInst *load = new LoadInst(getElem, name("const"), false, m_block); load->setAlignment(8); @@ -146,7 +174,17 @@ void Storage::setTempElement(int idx, llvm::Value *val, int mask) llvm::Value *templ = m_temps[idx]; val = maskWrite(val, mask, templ); } + llvm::Value *templ = m_temps[idx]; + if (templ) { + BasicBlock *block = m_varBlocks[templ]; + if (block != m_block) { + addPhiNode(idx, val, m_block, templ, block); + } else + updatePhiNode(idx, val); + } + m_temps[idx] = val; + m_varBlocks[val] = m_block; } void Storage::store(int dstIdx, llvm::Value *val, int mask) @@ -208,3 +246,69 @@ int Storage::numConsts() const { return m_numConsts; } + +llvm::Value * Storage::addrElement(int idx) const +{ + Value *ret = m_addrs[idx]; + if (!ret) + return m_undefFloatVec; + return ret; +} + +void Storage::setAddrElement(int idx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = m_addrs[idx]; + val = maskWrite(val, mask, templ); + } + m_addrs[idx] = val; +} + +llvm::Value * Storage::extractIndex(llvm::Value *vec) +{ + llvm::Value *x = new ExtractElementInst(vec, unsigned(0), + name("x"), m_block); + return new FPToSIInst(x, IntegerType::get(32), name("intidx"), m_block); +} + +void Storage::setCurrentBlock(llvm::BasicBlock *block) +{ + m_block = block; +} + +void Storage::addPhiNode(int idx, llvm::Value *val1, llvm::BasicBlock *blk1, + llvm::Value *val2, llvm::BasicBlock *blk2) +{ + PhiNode node; + node.val1 = val1; + node.block1 = blk1; + node.val2 = val2; + node.block2 = blk2; + m_phiNodes[idx] = node; +} + +void Storage::updatePhiNode(int idx, llvm::Value *val1) +{ + if (m_phiNodes.find(idx) == m_phiNodes.end()) + return; + PhiNode node = m_phiNodes[idx]; + node.val1 = val1; + m_phiNodes[idx] = node; +} + +void Storage::popPhiNode() +{ + if (!m_phiNodes.empty()) { + std::map::const_iterator itr; + for (itr = m_phiNodes.begin(); itr != m_phiNodes.end(); ++itr) { + PhiNode node = (*itr).second; + PHINode *dest = new PHINode(m_floatVecType, + name("phiDest"), m_block); + dest->reserveOperandSpace(2); + dest->addIncoming(node.val1, node.block1); + dest->addIncoming(node.val2, node.block2); + m_temps[(*itr).first] = dest; + } + } + m_phiNodes.clear(); +} diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index da7b0dbe01..919fbc9ded 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -21,19 +21,27 @@ public: llvm::Value *out, llvm::Value *in, llvm::Value *consts); + void setCurrentBlock(llvm::BasicBlock *block); + llvm::ConstantInt *constantInt(int); llvm::Constant *shuffleMask(int vec); - llvm::Value *inputElement(int idx); - llvm::Value *constElement(int idx); + llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); + llvm::Value *constElement(int idx, llvm::Value *indIdx =0); llvm::Value *tempElement(int idx) const; void setTempElement(int idx, llvm::Value *val, int mask); + llvm::Value *addrElement(int idx) const; + void setAddrElement(int idx, llvm::Value *val, int mask); + llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); + llvm::Value *extractIndex(llvm::Value *vec); void store(int dstIdx, llvm::Value *val, int mask); + void popPhiNode(); + int numConsts() const; private: llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); @@ -47,6 +55,7 @@ private: std::map m_constInts; std::map m_intVecs; std::vector m_temps; + std::vector m_addrs; std::vector m_dstCache; LoadMap m_inputs; LoadMap m_consts; @@ -63,6 +72,19 @@ private: int m_idx; int m_numConsts; + + void addPhiNode(int, llvm::Value*, llvm::BasicBlock*, + llvm::Value*, llvm::BasicBlock*); + void updatePhiNode(int, llvm::Value*); + struct PhiNode { + llvm::Value *val1; + llvm::BasicBlock *block1; + llvm::Value *val2; + llvm::BasicBlock *block2; + }; + + std::map m_varBlocks; + std::map m_phiNodes; }; #endif -- cgit v1.2.3 From d76a7b61bb2de2425289f462e07a678cf3c4ba59 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 24 Oct 2007 08:34:01 -0400 Subject: Cleanup some code. --- src/mesa/pipe/draw/draw_private.h | 2 ++ src/mesa/pipe/draw/draw_vertex_fetch.c | 4 ++-- src/mesa/pipe/draw/draw_vertex_shader.c | 7 +++++++ src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 4 ++++ src/mesa/pipe/llvm/llvmtgsi.cpp | 6 ++++-- src/mesa/pipe/llvm/llvmtgsi.h | 4 ++++ src/mesa/pipe/softpipe/sp_quad_fs.c | 4 +++- src/mesa/sources | 2 +- src/mesa/state_tracker/st_program.c | 6 +++++- 9 files changed, 32 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index b672cc2e8b..e4c0551124 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -251,7 +251,9 @@ extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); +#ifdef MESA_LLVM extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw ); +#endif struct tgsi_exec_machine; diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 1b093b7342..de1cd06da6 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -96,7 +96,7 @@ void draw_vertex_fetch( struct draw_context *draw, for (j = 0; j < count; j++) { uint attr; - printf("fetch vertex %u: \n", j); + /*printf("fetch vertex %u: \n", j);*/ /* loop over vertex attributes (vertex shader inputs) */ for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { @@ -111,7 +111,7 @@ void draw_vertex_fetch( struct draw_context *draw, fetch_attrib4(src, draw->vertex_element[attr].src_format, p); - printf("> %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]); + /*printf(" %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]);*/ /* Transform to AoS xxxx/yyyy/zzzz/wwww representation: */ diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index fb20dfa4e1..24cbf20fc7 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -161,6 +161,10 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; #if DBG printf("output[%d][%d]: %f %f %f %f\n", j, slot, + vOut[j]->data[slot][0], + vOut[j]->data[slot][1], + vOut[j]->data[slot][2], + vOut[j]->data[slot][3]); #endif } } /* loop over vertices */ @@ -175,10 +179,13 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) { unsigned i, j; +// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); +#ifdef MESA_LLVM if (draw->vertex_shader->state->llvm_prog) { draw_vertex_shader_queue_flush_llvm(draw); return; } +#endif /* run vertex shader on vertex cache entries, four per invokation */ for (i = 0; i < draw->vs.queue_nr; i += 4) { diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index bcc44f7980..99f0aca9e5 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -35,6 +35,8 @@ #include "draw_context.h" #include "draw_vertex.h" +#ifdef MESA_LLVM + #include "pipe/llvm/llvmtgsi.h" #include "pipe/tgsi/exec/tgsi_core.h" @@ -199,3 +201,5 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) draw->vs.queue_nr = 0; } + +#endif /* MESA_LLVM */ diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index ade4573fb8..066175cdf9 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -10,7 +10,6 @@ #include "pipe/tgsi/exec/tgsi_util.h" #include "pipe/tgsi/exec/tgsi_parse.h" #include "pipe/tgsi/exec/tgsi_dump.h" -//#include "pipe/tgsi/tgsi_platform.h" #include #include @@ -36,6 +35,7 @@ #include #include +#ifdef MESA_LLVM struct ga_llvm_prog { void *module; @@ -159,7 +159,7 @@ translate_instruction(llvm::Module *module, } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { val = storage->tempElement(src->SrcRegister.Index); } else { - fprintf(stderr, "ERROR: not support llvm source\n"); + fprintf(stderr, "ERROR: not supported llvm source\n"); return; } @@ -751,3 +751,5 @@ void ga_llvm_prog_dump(struct ga_llvm_prog *prog, const char *file_prefix) std::cout<<"; ---------- End shader "<id<InterpCoefs ); } else { - //ga_llvm_prog_exec(softpipe->fs->llvm_prog); +#ifdef MESA_LLVM + /*ga_llvm_prog_exec(softpipe->fs->llvm_prog);*/ +#endif quad->mask &= tgsi_exec_machine_run( machine ); } diff --git a/src/mesa/sources b/src/mesa/sources index 1bc45c9813..6d1ba9b9bd 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -189,7 +189,7 @@ TGSIDECO_SOURCES = \ TGSIMESA_SOURCES = \ pipe/tgsi/mesa/mesa_to_tgsi.c -ifeq ($(MESA_NO_LLVM),0) +ifeq ($(MESA_LLVM),1) LLVMTGSI_SOURCES = \ pipe/llvm/llvmtgsi.cpp \ pipe/llvm/storage.cpp \ diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 0f9d769264..6794227e13 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -251,7 +251,9 @@ st_translate_vertex_program(struct st_context *st, tokensOut, maxTokens); vs.tokens = tokensOut; +#ifdef MESA_LLVM vs.llvm_prog = ga_llvm_from_tgsi(st->pipe, vs.tokens); +#endif cso = st_cached_vs_state(st, &vs); stvp->vs = cso; @@ -407,7 +409,9 @@ st_translate_fragment_program(struct st_context *st, tokensOut, maxTokens); fs.tokens = tokensOut; - fs.llvm_prog = ga_llvm_from_tgsi(st->pipe, fs.tokens); +#ifdef MESA_LLVM + /*fs.llvm_prog = ga_llvm_from_tgsi(st->pipe, fs.tokens);*/ +#endif cso = st_cached_fs_state(st, &fs); stfp->fs = cso; -- cgit v1.2.3 From 9d4b51aab8f5a0c0a4a0c14886d1c87828d9eeb0 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 24 Oct 2007 11:41:28 -0400 Subject: Remove the silly function and remove the bogus comment. --- src/mesa/pipe/llvm/llvmtgsi.cpp | 108 +++++++++++++++++++--------------------- 1 file changed, 50 insertions(+), 58 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 066175cdf9..02a82a2a08 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -47,71 +47,62 @@ struct ga_llvm_prog { using namespace llvm; #include "llvm_base_shader.cpp" - static int GLOBAL_ID = 0; -static inline void addPass(PassManager &PM, Pass *P) { - // Add the pass to the pass manager... - PM.add(P); -} - static inline void AddStandardCompilePasses(PassManager &PM) { PM.add(createVerifierPass()); // Verify that input is correct - addPass(PM, createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp - - // If the -strip-debug command line option was specified, do it. - //if (StripDebug) - // addPass(PM, createStripSymbolsPass(true)); - - addPass(PM, createRaiseAllocationsPass()); // call %malloc -> malloc inst - addPass(PM, createCFGSimplificationPass()); // Clean up disgusting code - addPass(PM, createPromoteMemoryToRegisterPass());// Kill useless allocas - addPass(PM, createGlobalOptimizerPass()); // Optimize out global vars - addPass(PM, createGlobalDCEPass()); // Remove unused fns and globs - addPass(PM, createIPConstantPropagationPass());// IP Constant Propagation - addPass(PM, createDeadArgEliminationPass()); // Dead argument elimination - addPass(PM, createInstructionCombiningPass()); // Clean up after IPCP & DAE - addPass(PM, createCFGSimplificationPass()); // Clean up after IPCP & DAE - - addPass(PM, createPruneEHPass()); // Remove dead EH info - - //if (!DisableInline) - addPass(PM, createFunctionInliningPass()); // Inline small functions - addPass(PM, createArgumentPromotionPass()); // Scalarize uninlined fn args - - addPass(PM, createTailDuplicationPass()); // Simplify cfg by copying code - addPass(PM, createInstructionCombiningPass()); // Cleanup for scalarrepl. - addPass(PM, createCFGSimplificationPass()); // Merge & remove BBs - addPass(PM, createScalarReplAggregatesPass()); // Break up aggregate allocas - addPass(PM, createInstructionCombiningPass()); // Combine silly seq's - addPass(PM, createCondPropagationPass()); // Propagate conditionals - - addPass(PM, createTailCallEliminationPass()); // Eliminate tail calls - addPass(PM, createCFGSimplificationPass()); // Merge & remove BBs - addPass(PM, createReassociatePass()); // Reassociate expressions - addPass(PM, createLoopRotatePass()); - addPass(PM, createLICMPass()); // Hoist loop invariants - addPass(PM, createLoopUnswitchPass()); // Unswitch loops. - addPass(PM, createLoopIndexSplitPass()); // Index split loops. - addPass(PM, createInstructionCombiningPass()); // Clean up after LICM/reassoc - addPass(PM, createIndVarSimplifyPass()); // Canonicalize indvars - addPass(PM, createLoopUnrollPass()); // Unroll small loops - addPass(PM, createInstructionCombiningPass()); // Clean up after the unroller - addPass(PM, createGVNPass()); // Remove redundancies - addPass(PM, createSCCPPass()); // Constant prop with SCCP + PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp + + //PM.add(createStripSymbolsPass(true)); + + PM.add(createRaiseAllocationsPass()); // call %malloc -> malloc inst + PM.add(createCFGSimplificationPass()); // Clean up disgusting code + PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas + PM.add(createGlobalOptimizerPass()); // Optimize out global vars + PM.add(createGlobalDCEPass()); // Remove unused fns and globs + PM.add(createIPConstantPropagationPass());// IP Constant Propagation + PM.add(createDeadArgEliminationPass()); // Dead argument elimination + PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE + PM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE + + PM.add(createPruneEHPass()); // Remove dead EH info + + PM.add(createFunctionInliningPass()); // Inline small functions + PM.add(createArgumentPromotionPass()); // Scalarize uninlined fn args + + PM.add(createTailDuplicationPass()); // Simplify cfg by copying code + PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl. + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas + PM.add(createInstructionCombiningPass()); // Combine silly seq's + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createTailCallEliminationPass()); // Eliminate tail calls + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createReassociatePass()); // Reassociate expressions + PM.add(createLoopRotatePass()); + PM.add(createLICMPass()); // Hoist loop invariants + PM.add(createLoopUnswitchPass()); // Unswitch loops. + PM.add(createLoopIndexSplitPass()); // Index split loops. + PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc + PM.add(createIndVarSimplifyPass()); // Canonicalize indvars + PM.add(createLoopUnrollPass()); // Unroll small loops + PM.add(createInstructionCombiningPass()); // Clean up after the unroller + PM.add(createGVNPass()); // Remove redundancies + PM.add(createSCCPPass()); // Constant prop with SCCP // Run instcombine after redundancy elimination to exploit opportunities // opened up by them. - addPass(PM, createInstructionCombiningPass()); - addPass(PM, createCondPropagationPass()); // Propagate conditionals - - addPass(PM, createDeadStoreEliminationPass()); // Delete dead stores - addPass(PM, createAggressiveDCEPass()); // SSA based 'Aggressive DCE' - addPass(PM, createCFGSimplificationPass()); // Merge & remove BBs - addPass(PM, createSimplifyLibCallsPass()); // Library Call Optimizations - addPass(PM, createDeadTypeEliminationPass()); // Eliminate dead types - addPass(PM, createConstantMergePass()); // Merge dup global constants + PM.add(createInstructionCombiningPass()); + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createDeadStoreEliminationPass()); // Delete dead stores + PM.add(createAggressiveDCEPass()); // SSA based 'Aggressive DCE' + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createSimplifyLibCallsPass()); // Library Call Optimizations + PM.add(createDeadTypeEliminationPass()); // Eliminate dead types + PM.add(createConstantMergePass()); // Merge dup global constants } static void @@ -119,7 +110,6 @@ translate_declaration(llvm::Module *module, struct tgsi_full_declaration *decl, struct tgsi_full_declaration *fd) { - /* i think this is going to be a noop */ } @@ -658,9 +648,11 @@ ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) (struct ga_llvm_prog *)malloc(sizeof(struct ga_llvm_prog)); ga_llvm->id = GLOBAL_ID; tgsi_dump(tokens, 0); + llvm::Module *mod = tgsi_to_llvm(ga_llvm, tokens); ga_llvm->module = mod; ga_llvm_prog_dump(ga_llvm, 0); + /* Run optimization passes over it */ PassManager passes; passes.add(new TargetData(mod)); -- cgit v1.2.3 From 02cf317ed69c67c672c852451c45986d7798413e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 24 Oct 2007 12:48:06 -0400 Subject: Add copyright headers and do some cleanups. --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 4 +- src/mesa/pipe/llvm/instructions.cpp | 32 ++++++++++++++ src/mesa/pipe/llvm/instructions.h | 32 ++++++++++++++ src/mesa/pipe/llvm/llvm_builtins.c | 32 ++++++++++++++ src/mesa/pipe/llvm/llvmtgsi.cpp | 66 +++++++++++++++++++++------- src/mesa/pipe/llvm/llvmtgsi.h | 44 ++++++++++++++++--- src/mesa/pipe/llvm/storage.cpp | 32 ++++++++++++++ src/mesa/pipe/llvm/storage.h | 32 ++++++++++++++ src/mesa/state_tracker/st_program.c | 4 +- 9 files changed, 251 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 99f0aca9e5..c0720d2872 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -125,7 +125,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) float inputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; float outputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; float (*consts)[4] = (float (*)[4]) draw->mapped_constants; - struct ga_llvm_prog *prog = (struct ga_llvm_prog *)draw->vertex_shader->state->llvm_prog; + struct gallivm_prog *prog = (struct gallivm_prog *)draw->vertex_shader->state->llvm_prog; const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; @@ -137,7 +137,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) } /* batch execute the shaders on all the vertices */ - ga_llvm_prog_exec(prog, inputs, outputs, consts, + gallivm_prog_exec(prog, inputs, outputs, consts, draw->vs.queue_nr, draw->vertex_shader->state->num_inputs, draw->vertex_info.num_attribs); diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 66c34f8101..6ded3f1249 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -1,3 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + #include "instructions.h" #include diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index ce2ce796d9..8b47accd15 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -1,3 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + #ifndef INSTRUCTIONS_H #define INSTRUCTIONS_H diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index 2269b76ea1..b0f7cae43b 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -1,3 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + /* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm-dis */ /* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -funcname=createBaseShader */ typedef __attribute__(( ocu_vector_type(4) )) float float4; diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 02a82a2a08..b8e4e39316 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -1,3 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + #include "llvmtgsi.h" #include "instructions.h" @@ -37,7 +69,7 @@ #ifdef MESA_LLVM -struct ga_llvm_prog { +struct gallivm_prog { void *module; void *function; int num_consts; @@ -575,7 +607,7 @@ translate_instruction(llvm::Module *module, } static llvm::Module * -tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) +tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) { llvm::Module *mod = createBaseShader(); struct tgsi_parse_context parse; @@ -639,19 +671,19 @@ tgsi_to_llvm(struct ga_llvm_prog *prog, const struct tgsi_token *tokens) return mod; } -struct ga_llvm_prog * -ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) +struct gallivm_prog * +gallivm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) { std::cout << "Creating llvm from: " <id = GLOBAL_ID; + struct gallivm_prog *gallivm = + (struct gallivm_prog *)malloc(sizeof(struct gallivm_prog)); + gallivm->id = GLOBAL_ID; tgsi_dump(tokens, 0); - llvm::Module *mod = tgsi_to_llvm(ga_llvm, tokens); - ga_llvm->module = mod; - ga_llvm_prog_dump(ga_llvm, 0); + llvm::Module *mod = tgsi_to_llvm(gallivm, tokens); + gallivm->module = mod; + gallivm_prog_dump(gallivm, 0); /* Run optimization passes over it */ PassManager passes; @@ -668,17 +700,17 @@ ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) ee = (llvm::ExecutionEngine*)pipe->llvm_execution_engine; ee->addModuleProvider(mp); } - ga_llvm->module = mod; + gallivm->module = mod; Function *func = mod->getFunction("run_vertex_shader"); - ga_llvm->function = ee->getPointerToFunctionOrStub(func); + gallivm->function = ee->getPointerToFunctionOrStub(func); - ga_llvm_prog_dump(ga_llvm, 0); + gallivm_prog_dump(gallivm, 0); - return ga_llvm; + return gallivm; } -void ga_llvm_prog_delete(struct ga_llvm_prog *prog) +void gallivm_prog_delete(struct gallivm_prog *prog) { llvm::Module *mod = static_cast(prog->module); delete mod; @@ -695,7 +727,7 @@ typedef void (*vertex_shader_runner)(float (*ainputs)[PIPE_MAX_SHADER_INPUTS][4] int num_attribs, int num_consts); -int ga_llvm_prog_exec(struct ga_llvm_prog *prog, +int gallivm_prog_exec(struct gallivm_prog *prog, float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], float (*dests)[PIPE_MAX_SHADER_INPUTS][4], float (*consts)[4], @@ -710,7 +742,7 @@ int ga_llvm_prog_exec(struct ga_llvm_prog *prog, return 0; } -void ga_llvm_prog_dump(struct ga_llvm_prog *prog, const char *file_prefix) +void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) { llvm::Module *mod; if (!prog || !prog->module) diff --git a/src/mesa/pipe/llvm/llvmtgsi.h b/src/mesa/pipe/llvm/llvmtgsi.h index 40798f9217..d4cf61ca97 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.h +++ b/src/mesa/pipe/llvm/llvmtgsi.h @@ -1,3 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + #ifndef LLVMTGSI_H #define LLVMTGSI_H @@ -14,14 +46,14 @@ struct tgsi_token; struct tgsi_sampler; struct pipe_context; -struct ga_llvm_prog; +struct gallivm_prog; -struct ga_llvm_prog * -ga_llvm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens); +struct gallivm_prog * +gallivm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens); -void ga_llvm_prog_delete(struct ga_llvm_prog *prog); +void gallivm_prog_delete(struct gallivm_prog *prog); -int ga_llvm_prog_exec(struct ga_llvm_prog *prog, +int gallivm_prog_exec(struct gallivm_prog *prog, float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], float (*dests)[PIPE_MAX_SHADER_INPUTS][4], float (*consts)[4], @@ -29,7 +61,7 @@ int ga_llvm_prog_exec(struct ga_llvm_prog *prog, int num_inputs, int num_attribs); -void ga_llvm_prog_dump(struct ga_llvm_prog *prog, const char *file_prefix); +void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); #endif /* MESA_LLVM */ diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 5b1f05190e..75ee18567d 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -1,3 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + #include "storage.h" #include "pipe/tgsi/exec/tgsi_token.h" diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index 919fbc9ded..bc8cffdf66 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -1,3 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + #ifndef STORAGE_H #define STORAGE_H diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 6794227e13..9fc0798ec5 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -252,7 +252,7 @@ st_translate_vertex_program(struct st_context *st, vs.tokens = tokensOut; #ifdef MESA_LLVM - vs.llvm_prog = ga_llvm_from_tgsi(st->pipe, vs.tokens); + vs.llvm_prog = (void*)gallivm_from_tgsi(st->pipe, vs.tokens); #endif cso = st_cached_vs_state(st, &vs); stvp->vs = cso; @@ -410,7 +410,7 @@ st_translate_fragment_program(struct st_context *st, fs.tokens = tokensOut; #ifdef MESA_LLVM - /*fs.llvm_prog = ga_llvm_from_tgsi(st->pipe, fs.tokens);*/ + /*fs.llvm_prog = gallivm_from_tgsi(st->pipe, fs.tokens);*/ #endif cso = st_cached_fs_state(st, &fs); stfp->fs = cso; -- cgit v1.2.3 From af960431675ddb51c24da29ac183399401b62362 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 17:16:38 -0600 Subject: added a pipe->flush() call in xmesa_clear() --- src/mesa/drivers/x11/xm_surface.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 42d5d6bbc4..b2ef70ea90 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -117,6 +117,12 @@ put_tile(struct pipe_surface *ps, xrb->St.Base.PutRow(ctx, &xrb->St.Base, w, x, y - i, tmp, NULL); p += w0 * 4; } +#if 0 /* debug: flush */ + { + XMesaContext xm = XMESA_CONTEXT(ctx); + XSync(xm->display, 0); + } +#endif } @@ -208,6 +214,11 @@ xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) { struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); + /* XXX actually, we should just discard any cached tiles from this + * surface since we don't want to accidentally re-use them after clearing. + */ + pipe->flush(pipe, 0); + if (xrb && xrb->ximage) { /* clearing back color buffer */ GET_CURRENT_CONTEXT(ctx); -- cgit v1.2.3 From 47f2e97019d367e544439d0604b824b7bd2643c7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 17:45:38 -0600 Subject: get_tile() for z16, z32, s8z24 surfaces needs to return 4 floats per pixel (for depth texture sampling) --- src/mesa/pipe/softpipe/sp_surface.c | 27 ++++++++++++++++++--------- 1 file changed, 18 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 40f045800f..057a311cd9 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -162,7 +162,7 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_U_Z16 ***/ /** - * Return as floats in [0,1]. + * Return each Z value as four floats in [0,1]. */ static void z16_get_tile(struct pipe_surface *ps, @@ -182,10 +182,13 @@ z16_get_tile(struct pipe_surface *ps, for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { - pRow[j] = src[j] * scale; + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = src[j] * scale; } src += ps->region->pitch; - p += w0; + p += 4 * w0; } } @@ -391,7 +394,7 @@ a8_l8_get_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_U_Z32 ***/ /** - * Return as floats in [0,1]. + * Return each Z value as four floats in [0,1]. */ static void z32_get_tile(struct pipe_surface *ps, @@ -411,10 +414,13 @@ z32_get_tile(struct pipe_surface *ps, for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { - pRow[j] = src[j] * scale; + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = src[j] * scale; } src += ps->region->pitch; - p += w0; + p += 4 * w0; } } @@ -422,7 +428,7 @@ z32_get_tile(struct pipe_surface *ps, /*** PIPE_FORMAT_S8_Z24 ***/ /** - * Return Z component as float in [0,1]. Stencil part ignored. + * Return Z component as four float in [0,1]. Stencil part ignored. */ static void s8z24_get_tile(struct pipe_surface *ps, @@ -442,10 +448,13 @@ s8z24_get_tile(struct pipe_surface *ps, for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { - pRow[j] = (src[j] & 0xffffff) * scale; + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = (src[j] & 0xffffff) * scale; } src += ps->region->pitch; - p += w0; + p += 4 * w0; } } -- cgit v1.2.3 From b78e90807abc31f58492992cdfe5e01bfd53e68b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 17:45:55 -0600 Subject: flush pipe in draw_stencil_pixels() --- src/mesa/state_tracker/st_cb_drawpixels.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 90f91c39b5..4fdf1cef2b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -797,6 +797,8 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, GLint skipPixels; ubyte *stmap; + pipe->flush(pipe, 0); + /* map the stencil buffer */ stmap = pipe->region_map(pipe, ps->region); -- cgit v1.2.3 From 76a4fd098f44ae4f226d4747b9fdaf9db5d40270 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 23 Oct 2007 18:49:19 -0600 Subject: a bit more work for optimizing clears in tile cache (not enabled yet) --- src/mesa/drivers/x11/xm_surface.c | 14 +++++ src/mesa/pipe/softpipe/sp_clear.c | 18 +++++- src/mesa/pipe/softpipe/sp_tile_cache.c | 104 +++++++++++++++++++++++++-------- src/mesa/pipe/softpipe/sp_tile_cache.h | 7 ++- 4 files changed, 115 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index b2ef70ea90..f83c7917c4 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -46,6 +46,7 @@ #include "pipe/p_defines.h" #include "pipe/softpipe/sp_context.h" #include "pipe/softpipe/sp_clear.h" +#include "pipe/softpipe/sp_tile_cache.h" #include "state_tracker/st_context.h" @@ -219,6 +220,19 @@ xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) */ pipe->flush(pipe, 0); + { + struct softpipe_context *sp = softpipe_context(pipe); + struct softpipe_surface *sps = softpipe_surface(ps); + if (sps == sp_tile_cache_get_surface(sp->cbuf_cache[0])) { + float clear[4]; + clear[0] = 0.2; /* XXX hack */ + clear[1] = 0.2; + clear[2] = 0.2; + clear[3] = 0.2; + sp_tile_cache_clear(sp->cbuf_cache[0], clear); + } + } + if (xrb && xrb->ximage) { /* clearing back color buffer */ GET_CURRENT_CONTEXT(ctx); diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 96bca4b171..08b87417aa 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -47,7 +47,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); - /*struct softpipe_surface *sps = softpipe_surface(ps);*/ + struct softpipe_surface *sps = softpipe_surface(ps); unsigned x, y, w, h; softpipe_update_derived(softpipe); /* not needed?? */ @@ -66,9 +66,23 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, assert(w <= ps->region->pitch); assert(h <= ps->region->height); - /* XXX skip this fill if we're using tile cache */ + if (sps == sp_tile_cache_get_surface(softpipe->zbuf_cache)) { + float clear[4]; + clear[0] = 1.0; /* XXX hack */ + sp_tile_cache_clear(softpipe->zbuf_cache, clear); + } + else if (sps == sp_tile_cache_get_surface(softpipe->cbuf_cache[0])) { + float clear[4]; + clear[0] = 0.2; /* XXX hack */ + clear[1] = 0.2; /* XXX hack */ + clear[2] = 0.2; /* XXX hack */ + clear[3] = 0.2; /* XXX hack */ + sp_tile_cache_clear(softpipe->cbuf_cache[0], clear); + } + pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue); + #if 0 sp_clear_tile_cache(sps, clearValue); #endif diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 1e287c91a4..129785d26d 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -37,6 +37,7 @@ #include "sp_surface.h" #include "sp_tile_cache.h" +#define CLEAR_OPTIMIZATION 0 #define NUM_ENTRIES 20 @@ -52,6 +53,7 @@ struct softpipe_tile_cache struct pipe_mipmap_tree *texture; /**< if caching a texture */ struct softpipe_cached_tile entries[NUM_ENTRIES]; uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; + float clear_value[4]; }; @@ -73,6 +75,7 @@ is_clear_flag_set(const uint *bitvec, int x, int y) x /= TILE_SIZE; y /= TILE_SIZE; pos = y * (MAX_WIDTH / TILE_SIZE) + x; + assert(pos / 32 < (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32); bit = bitvec[pos / 32] & (1 << (pos & 31)); return bit; } @@ -85,6 +88,7 @@ clear_clear_flag(uint *bitvec, int x, int y) x /= TILE_SIZE; y /= TILE_SIZE; pos = y * (MAX_WIDTH / TILE_SIZE) + x; + assert(pos / 32 < (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32); bitvec[pos / 32] &= ~(1 << (pos & 31)); } @@ -125,6 +129,13 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, } +struct softpipe_surface * +sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) +{ + return tc->surface; +} + + void sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, struct pipe_mipmap_tree *texture) @@ -174,6 +185,7 @@ sp_flush_tile_cache(struct softpipe_tile_cache *tc) inuse++; } } + /* printf("flushed tiles in use: %d\n", inuse); */ @@ -198,17 +210,9 @@ sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y) const int pos = CACHE_POS(x, y); struct softpipe_cached_tile *tile = tc->entries + pos; - /* - printf("output %d, %d at pos %d\n", x, y, pos); - */ - if (tile_x != tile->x || tile_y != tile->y) { - /* - printf("new tile at %d, %d, pos %d\n", tile_x, tile_y, pos); - */ - if (tile->x != -1) { /* put dirty tile back in framebuffer */ if (is_depth_stencil) { @@ -223,23 +227,63 @@ sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y) } } - if (0/*is_clear_flag_set(tc->clear_flags, x, y)*/) { + if (is_clear_flag_set(tc->clear_flags, x, y)) { /* don't get tile from framebuffer, just clear it */ -#if 0 - printf("clear tile\n"); uint i, j; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.color[i][j][0] = 0.5; - tile->data.color[i][j][1] = 0.5; - tile->data.color[i][j][2] = 0.5; - tile->data.color[i][j][3] = 0.5; + /* XXX these loops could be optimized */ + switch (ps->format) { + case PIPE_FORMAT_U_Z16: + { + ushort clear_val = (ushort) (tc->clear_value[0] * 0xffff); + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.depth16[i][j] = clear_val; + } + } + } + break; + case PIPE_FORMAT_U_Z32: + { + uint clear_val = (uint) (tc->clear_value[0] * 0xffffffff); + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.depth32[i][j] = clear_val; + } + } + } + break; + case PIPE_FORMAT_S8_Z24: + { + uint clear_val = (uint) (tc->clear_value[0] * 0xffffff); + clear_val |= ((uint) tc->clear_value[1]) << 24; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.depth32[i][j] = clear_val; + } + } + } + break; + case PIPE_FORMAT_U_S8: + { + ubyte clear_val = (uint) tc->clear_value[0]; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.stencil8[i][j] = clear_val; + } + } + } + break; + default: + /* color */ + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.color[i][j][0] = tc->clear_value[0]; + tile->data.color[i][j][1] = tc->clear_value[1]; + tile->data.color[i][j][2] = tc->clear_value[2]; + tile->data.color[i][j][3] = tc->clear_value[3]; + } } } -#else - (void) is_clear_flag_set; -#endif - memset(tile->data.color, 0, sizeof(tile->data.color)); clear_clear_flag(tc->clear_flags, x, y); } else { @@ -322,10 +366,22 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, } - +/** + * When a whole surface is being cleared to a value we can avoid + * fetching tiles above. + * Save the color and set a 'clearflag' for each tile of the screen. + */ void -sp_clear_tile_cache(struct softpipe_tile_cache *tc, unsigned clearval) +sp_tile_cache_clear(struct softpipe_tile_cache *tc, const float value[4]) { - (void) clearval; /* XXX use this */ + tc->clear_value[0] = value[0]; + tc->clear_value[1] = value[1]; + tc->clear_value[2] = value[2]; + tc->clear_value[3] = value[3]; + +#if CLEAR_OPTIMIZATION memset(tc->clear_flags, 255, sizeof(tc->clear_flags)); +#else + memset(tc->clear_flags, 0, sizeof(tc->clear_flags)); +#endif } diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index a1cc387a12..15245a2efb 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -29,7 +29,7 @@ #define SP_TILE_CACHE_H -#include "p_compiler.h" +#include "pipe/p_compiler.h" struct softpipe_context; @@ -66,6 +66,9 @@ extern void sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, struct softpipe_surface *sps); +extern struct softpipe_surface * +sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); + extern void sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, struct pipe_mipmap_tree *texture); @@ -74,7 +77,7 @@ extern void sp_flush_tile_cache(struct softpipe_tile_cache *tc); extern void -sp_clear_tile_cache(struct softpipe_tile_cache *tc, unsigned clearval); +sp_tile_cache_clear(struct softpipe_tile_cache *tc, const float value[4]); extern struct softpipe_cached_tile * sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y); -- cgit v1.2.3 From 4eafe69a2ac27e8e92ce0fc503a5d6cc24b8f257 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 12:30:49 -0600 Subject: better debug output --- src/mesa/pipe/draw/draw_vertex_shader.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 24cbf20fc7..d8bd5a96d6 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -147,7 +147,11 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][3] = w; #if DBG - printf("output[%d]win: %f %f %f %f\n", x, y, z, w); + printf("output[%d]win: %f %f %f %f\n", j, + vOut[j]->data[0][0], + vOut[j]->data[0][1], + vOut[j]->data[0][2], + vOut[j]->data[0][3]); #endif /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. -- cgit v1.2.3 From e506514dbbaf6850f477125f89068afa967fa9b4 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 12:31:14 -0600 Subject: fix bug in RET code --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 4a608c445e..113a0f2c18 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -2077,6 +2077,7 @@ exec_instruction( return; } *pc = mach->CallStack[--mach->CallStackTop]; + UPDATE_EXEC_MASK(mach); } break; -- cgit v1.2.3 From 3a345e000350c1590367f01be5db6f0f10701283 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 12:31:29 -0600 Subject: print CAL label --- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index 7fd503d81b..9d21da0965 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -945,6 +945,7 @@ dump_instruction_short( case TGSI_OPCODE_ELSE: case TGSI_OPCODE_BGNLOOP2: case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_CAL: TXT( " :" ); UID( inst->InstructionExtLabel.Label ); break; -- cgit v1.2.3 From f68fce1f1f61997f284b77cca5b5765cd7443240 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 11:37:05 -0600 Subject: Implement gl_PointCoord attribute for GLSL fragment shaders. Contains the normalized fragment position within a point sprite. --- src/mesa/shader/slang/slang_codegen.c | 6 +- src/mesa/shader/slang/slang_compile.c | 11 ++++ src/mesa/swrast/s_fog.c | 1 - src/mesa/swrast/s_fragprog.c | 1 + src/mesa/swrast/s_points.c | 117 +++++++++++++++++++--------------- src/mesa/swrast/s_span.c | 2 +- 6 files changed, 85 insertions(+), 53 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 8b2bdd74b5..fb1f9d5a20 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -282,6 +282,8 @@ sampler_to_texture_index(const slang_type_specifier_type type) } +#define SWIZZLE_ZWWW MAKE_SWIZZLE4(SWIZZLE_Z, SWIZZLE_W, SWIZZLE_W, SWIZZLE_W) + /** * Return the VERT_ATTRIB_* or FRAG_ATTRIB_* value that corresponds to * a vertex or fragment program input variable. Return -1 if the input @@ -316,9 +318,11 @@ _slang_input_index(const char *name, GLenum target, GLuint *swizzleOut) { "gl_FragCoord", FRAG_ATTRIB_WPOS, SWIZZLE_NOOP }, { "gl_Color", FRAG_ATTRIB_COL0, SWIZZLE_NOOP }, { "gl_SecondaryColor", FRAG_ATTRIB_COL1, SWIZZLE_NOOP }, - { "gl_FogFragCoord", FRAG_ATTRIB_FOGC, SWIZZLE_XXXX }, { "gl_TexCoord", FRAG_ATTRIB_TEX0, SWIZZLE_NOOP }, + /* note: we're packing several quantities into the fogcoord vector */ + { "gl_FogFragCoord", FRAG_ATTRIB_FOGC, SWIZZLE_XXXX }, { "gl_FrontFacing", FRAG_ATTRIB_FOGC, SWIZZLE_YYYY }, /*XXX*/ + { "gl_PointCoord", FRAG_ATTRIB_FOGC, SWIZZLE_ZWWW }, { NULL, 0, SWIZZLE_NOOP } }; GLuint i; diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 4e29e8dcc9..2be89a5ce0 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -1983,6 +1983,10 @@ static const byte slang_120_core_gc[] = { #include "library/slang_120_core_gc.h" }; +static const byte slang_120_fragment_gc[] = { +#include "library/slang_builtin_120_fragment_gc.h" +}; + static const byte slang_common_builtin_gc[] = { #include "library/slang_common_builtin_gc.h" }; @@ -2059,6 +2063,13 @@ compile_object(grammar * id, const char *source, slang_code_object * object, SLANG_UNIT_FRAGMENT_BUILTIN, infolog, NULL, &object->builtin[SLANG_BUILTIN_COMMON], NULL)) return GL_FALSE; +#if FEATURE_ARB_shading_language_120 + if (!compile_binary(slang_120_fragment_gc, + &object->builtin[SLANG_BUILTIN_TARGET], + SLANG_UNIT_FRAGMENT_BUILTIN, infolog, NULL, + &object->builtin[SLANG_BUILTIN_COMMON], NULL)) + return GL_FALSE; +#endif } else if (type == SLANG_UNIT_VERTEX_SHADER) { if (!compile_binary(slang_vertex_builtin_gc, diff --git a/src/mesa/swrast/s_fog.c b/src/mesa/swrast/s_fog.c index ed47964a66..7b143f6e5b 100644 --- a/src/mesa/swrast/s_fog.c +++ b/src/mesa/swrast/s_fog.c @@ -168,7 +168,6 @@ _swrast_fog_rgba_span( const GLcontext *ctx, SWspan *span ) GLfloat rFog, gFog, bFog; ASSERT(swrast->_FogEnabled); - ASSERT(swrast->_ActiveAttribMask & FRAG_BIT_FOGC); ASSERT(span->arrayMask & SPAN_RGBA); /* compute (scaled) fog color */ diff --git a/src/mesa/swrast/s_fragprog.c b/src/mesa/swrast/s_fragprog.c index 14c9868c18..6656ebc0d0 100644 --- a/src/mesa/swrast/s_fragprog.c +++ b/src/mesa/swrast/s_fragprog.c @@ -113,6 +113,7 @@ init_machine(GLcontext *ctx, struct gl_program_machine *machine, machine->DerivY = (GLfloat (*)[4]) span->attrStepY; machine->NumDeriv = FRAG_ATTRIB_MAX; + /* if running a GLSL program (not ARB_fragment_program) */ if (ctx->Shader.CurrentProgram) { /* Store front/back facing value in register FOGC.Y */ machine->Attribs[FRAG_ATTRIB_FOGC][col][1] = (GLfloat) ctx->_Facing; diff --git a/src/mesa/swrast/s_points.c b/src/mesa/swrast/s_points.c index 4768fbea97..c0f32e9c91 100644 --- a/src/mesa/swrast/s_points.c +++ b/src/mesa/swrast/s_points.c @@ -55,7 +55,7 @@ sprite_point(GLcontext *ctx, const SWvertex *vert) SWcontext *swrast = SWRAST_CONTEXT(ctx); SWspan span; GLfloat size; - GLuint tCoords[MAX_TEXTURE_COORD_UNITS]; + GLuint tCoords[MAX_TEXTURE_COORD_UNITS + 1]; GLuint numTcoords = 0; GLfloat t0, dtdy; @@ -99,57 +99,71 @@ sprite_point(GLcontext *ctx, const SWvertex *vert) span.attrStepX[FRAG_ATTRIB_WPOS][3] = 0.0F; span.attrStepY[FRAG_ATTRIB_WPOS][3] = 0.0F; - ATTRIB_LOOP_BEGIN - if (attr >= FRAG_ATTRIB_TEX0 && attr < FRAG_ATTRIB_VAR0) { - const GLuint u = attr - FRAG_ATTRIB_TEX0; - /* a texcoord */ - if (ctx->Point.CoordReplace[u]) { - GLfloat s, r, dsdx; - - s = 0.0; - dsdx = 1.0 / size; - - if (ctx->Point.SpriteOrigin == GL_LOWER_LEFT) { - t0 = 0.0; - dtdy = 1.0 / size; - } - else { - /* GL_UPPER_LEFT */ - t0 = 1.0; - dtdy = -1.0 / size; - } - tCoords[numTcoords++] = attr; - - if (ctx->Point.SpriteRMode == GL_ZERO) - r = 0.0F; - else if (ctx->Point.SpriteRMode == GL_S) - r = vert->attrib[attr][0]; - else /* GL_R */ - r = vert->attrib[attr][2]; - - span.attrStart[attr][0] = s; - span.attrStart[attr][1] = 0.0; /* overwritten below */ - span.attrStart[attr][2] = r; - span.attrStart[attr][3] = 1.0; - - span.attrStepX[attr][0] = dsdx; - span.attrStepX[attr][1] = 0.0; - span.attrStepX[attr][2] = 0.0; - span.attrStepX[attr][3] = 0.0; - - span.attrStepY[attr][0] = 0.0; - span.attrStepY[attr][1] = dtdy; - span.attrStepY[attr][2] = 0.0; - span.attrStepY[attr][3] = 0.0; + { + GLfloat s, r, dsdx; + + /* texcoord / pointcoord interpolants */ + s = 0.0; + dsdx = 1.0 / size; + if (ctx->Point.SpriteOrigin == GL_LOWER_LEFT) { + t0 = 0.0; + dtdy = 1.0 / size; + } + else { + /* GL_UPPER_LEFT */ + t0 = 1.0; + dtdy = -1.0 / size; + } + ATTRIB_LOOP_BEGIN + if (attr >= FRAG_ATTRIB_TEX0 && attr < FRAG_ATTRIB_VAR0) { + const GLuint u = attr - FRAG_ATTRIB_TEX0; + /* a texcoord */ + if (ctx->Point.CoordReplace[u]) { + tCoords[numTcoords++] = attr; + + if (ctx->Point.SpriteRMode == GL_ZERO) + r = 0.0F; + else if (ctx->Point.SpriteRMode == GL_S) + r = vert->attrib[attr][0]; + else /* GL_R */ + r = vert->attrib[attr][2]; + + span.attrStart[attr][0] = s; + span.attrStart[attr][1] = 0.0; /* overwritten below */ + span.attrStart[attr][2] = r; + span.attrStart[attr][3] = 1.0; + + span.attrStepX[attr][0] = dsdx; + span.attrStepX[attr][1] = 0.0; + span.attrStepX[attr][2] = 0.0; + span.attrStepX[attr][3] = 0.0; + + span.attrStepY[attr][0] = 0.0; + span.attrStepY[attr][1] = dtdy; + span.attrStepY[attr][2] = 0.0; + span.attrStepY[attr][3] = 0.0; + + continue; + } + } + else if (attr == FRAG_ATTRIB_FOGC) { + /* GLSL gl_PointCoord is stored in fog.zw */ + span.attrStart[FRAG_ATTRIB_FOGC][2] = 0.0; + span.attrStart[FRAG_ATTRIB_FOGC][3] = 0.0; /* t0 set below */ + span.attrStepX[FRAG_ATTRIB_FOGC][2] = dsdx; + span.attrStepX[FRAG_ATTRIB_FOGC][3] = 0.0; + span.attrStepY[FRAG_ATTRIB_FOGC][2] = 0.0; + span.attrStepY[FRAG_ATTRIB_FOGC][3] = dtdy; + tCoords[numTcoords++] = FRAG_ATTRIB_FOGC; continue; } - } - /* use vertex's texcoord/attrib */ - COPY_4V(span.attrStart[attr], vert->attrib[attr]); - ASSIGN_4V(span.attrStepX[attr], 0, 0, 0, 0); - ASSIGN_4V(span.attrStepY[attr], 0, 0, 0, 0); - ATTRIB_LOOP_END + /* use vertex's texcoord/attrib */ + COPY_4V(span.attrStart[attr], vert->attrib[attr]); + ASSIGN_4V(span.attrStepX[attr], 0, 0, 0, 0); + ASSIGN_4V(span.attrStepY[attr], 0, 0, 0, 0); + ATTRIB_LOOP_END; + } /* compute pos, bounds and render */ { @@ -183,7 +197,10 @@ sprite_point(GLcontext *ctx, const SWvertex *vert) GLuint i; /* setup texcoord T for this row */ for (i = 0; i < numTcoords; i++) { - span.attrStart[tCoords[i]][1] = tcoord; + if (tCoords[i] == FRAG_ATTRIB_FOGC) + span.attrStart[FRAG_ATTRIB_FOGC][3] = tcoord; + else + span.attrStart[tCoords[i]][1] = tcoord; } /* these might get changed by span clipping */ diff --git a/src/mesa/swrast/s_span.c b/src/mesa/swrast/s_span.c index cfc65bee87..d9c1d1bec7 100644 --- a/src/mesa/swrast/s_span.c +++ b/src/mesa/swrast/s_span.c @@ -1180,7 +1180,7 @@ shade_texture_span(GLcontext *ctx, SWspan *span) if (span->primitive == GL_BITMAP && span->array->ChanType != GL_FLOAT) { convert_color_type(span, GL_FLOAT, 0); } - if (span->primitive != GL_POINT) { + if (span->primitive != GL_POINT || ctx->Point.PointSprite) { /* for points, we populated the arrays already */ interpolate_active_attribs(ctx, span, ~0); } -- cgit v1.2.3 From 3733b1f2e95642b47218604aee6a34eacb0c7293 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 12:43:02 -0600 Subject: add to git --- .../slang/library/slang_builtin_120_common_gc.h | 104 +++++++++++++++++++++ .../slang/library/slang_builtin_120_fragment_gc.h | 5 + 2 files changed, 109 insertions(+) create mode 100644 src/mesa/shader/slang/library/slang_builtin_120_common_gc.h create mode 100644 src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h b/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h new file mode 100644 index 0000000000..fc1a944217 --- /dev/null +++ b/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h @@ -0,0 +1,104 @@ + +/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ +/* slang_builtin_120_common.gc */ + +3,1,0,26,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,26,109,0,0,1,0,0,26,110,0, +0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,28,0,109,97,116,114,105,120,67,111,109,112,77,117, +108,116,0,1,0,0,28,109,0,0,1,0,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57, +18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,27,0,109, +97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,27,109,0,0,1,0,0,27,110,0,0,0,1,8,58,109, +97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110, +0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,30,0,109,97,116, +114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,30,109,0,0,1,0,0,30,110,0,0,0,1,8,58,109,97,116, +51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10, +49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,29,0,109,97,116,114,105, +120,67,111,109,112,77,117,108,116,0,1,0,0,29,109,0,0,1,0,0,29,110,0,0,0,1,8,58,109,97,116,52,120, +50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0, +57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10, +51,0,57,48,0,0,0,0,1,0,31,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,31,109,0, +0,1,0,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48, +0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57, +48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,0,13,0,111,117,116,101,114,80,114, +111,100,117,99,116,0,1,0,0,10,99,0,0,1,0,0,10,114,0,0,0,1,8,58,109,97,116,50,0,18,99,0,59,120,0,18, +114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48, +0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,0,0,0,1,0,14,0,111,117,116,101,114,80,114,111,100,117,99, +116,0,1,0,0,11,99,0,0,1,0,0,11,114,0,0,0,1,8,58,109,97,116,51,0,18,99,0,59,120,0,18,114,0,59,120,0, +48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59, +120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59, +121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0, +59,122,0,18,114,0,59,122,0,48,0,0,0,0,1,0,15,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0, +0,12,99,0,0,1,0,0,12,114,0,0,0,1,8,58,109,97,116,52,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18, +99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18, +114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48, +0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0, +18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0, +48,0,18,99,0,59,119,0,18,114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59, +121,0,18,114,0,59,119,0,48,0,18,99,0,59,122,0,18,114,0,59,119,0,48,0,18,99,0,59,119,0,18,114,0,59, +119,0,48,0,0,0,0,1,0,26,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,11,99,0,0,1,0,0,10, +114,0,0,0,1,8,58,109,97,116,50,120,51,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0, +18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0, +48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,0,0,0,1,0,27, +0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,10,99,0,0,1,0,0,11,114,0,0,0,1,8,58,109,97, +116,51,120,50,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18, +99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18, +114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,0,0,0,1,0,28,0,111,117,116,101,114,80, +114,111,100,117,99,116,0,1,0,0,12,99,0,0,1,0,0,10,114,0,0,0,1,8,58,109,97,116,50,120,52,0,18,99,0, +59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0, +59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18, +99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,119,0,18, +114,0,59,121,0,48,0,0,0,0,1,0,29,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,10,99,0,0, +1,0,0,12,114,0,0,0,1,8,58,109,97,116,52,120,50,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0, +59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0, +59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18, +99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48,0,0,0,0,1,0,30,0,111, +117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,12,99,0,0,1,0,0,11,114,0,0,0,1,8,58,109,97,116, +51,120,52,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99, +0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114, +0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18, +99,0,59,119,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18, +114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,18,99,0,59,119,0,18,114,0,59,122,0,48, +0,0,0,0,1,0,31,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,11,99,0,0,1,0,0,12,114,0,0, +0,1,8,58,109,97,116,52,120,51,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0, +59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18, +99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18, +114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48, +0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48,0,18,99,0,59,122,0, +18,114,0,59,119,0,48,0,0,0,0,1,0,13,0,116,114,97,110,115,112,111,115,101,0,1,0,0,13,109,0,0,0,1,8, +58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16, +8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,0,0,0,1,0,14,0,116,114,97,110,115,112,111, +115,101,0,1,0,0,14,109,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16, +10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0, +16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109, +0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,0,0,0,1,0,15,0,116,114,97,110,115,112, +111,115,101,0,1,0,0,15,109,0,0,0,1,8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0, +16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,0,0,18, +109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0, +18,109,0,16,10,51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0, +0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59,122,0,0,18,109,0,16,8,48,0,57,59,119, +0,0,18,109,0,16,10,49,0,57,59,119,0,0,18,109,0,16,10,50,0,57,59,119,0,0,18,109,0,16,10,51,0,57,59, +119,0,0,0,0,0,1,0,26,0,116,114,97,110,115,112,111,115,101,0,1,0,0,27,109,0,0,0,1,8,58,109,97,116, +50,120,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0, +57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50, +0,57,59,121,0,0,0,0,0,1,0,27,0,116,114,97,110,115,112,111,115,101,0,1,0,0,26,109,0,0,0,1,8,58,109, +97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16, +8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0, +16,10,49,0,57,59,122,0,0,0,0,0,1,0,28,0,116,114,97,110,115,112,111,115,101,0,1,0,0,29,109,0,0,0,1, +8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18, +109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0, +18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121, +0,0,0,0,0,1,0,29,0,116,114,97,110,115,112,111,115,101,0,1,0,0,28,109,0,0,0,1,8,58,109,97,116,52, +120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,8,48,0,57, +59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0, +57,59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,0,0,0,1,0,30,0, +116,114,97,110,115,112,111,115,101,0,1,0,0,31,109,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16, +8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0, +16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109, +0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18, +109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59,122,0,0, +0,0,0,1,0,31,0,116,114,97,110,115,112,111,115,101,0,1,0,0,30,109,0,0,0,1,8,58,109,97,116,52,120,51, +0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120, +0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59, +121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57, +59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,18,109,0,16,10,50,0, +57,59,119,0,0,0,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h b/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h new file mode 100644 index 0000000000..2400b273d8 --- /dev/null +++ b/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h @@ -0,0 +1,5 @@ + +/* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ +/* slang_builtin_120_fragment.gc */ + +3,2,2,3,10,1,103,108,95,80,111,105,110,116,67,111,111,114,100,0,0,0,0 -- cgit v1.2.3 From 11ce6244a79106c592364b30434b6ddade3fd6bb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 14:34:40 -0600 Subject: Undo indexOffset change (I think, git???) --- src/mesa/pipe/failover/fo_context.c | 6 ++++-- src/mesa/pipe/i915simple/i915_context.c | 7 ++++--- src/mesa/pipe/p_context.h | 2 +- src/mesa/pipe/softpipe/sp_draw_arrays.c | 7 ++++--- src/mesa/pipe/softpipe/sp_state.h | 2 +- src/mesa/state_tracker/st_draw.c | 6 ++++-- 6 files changed, 18 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index 076d516583..b836ef29f3 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -47,7 +47,7 @@ static void failover_destroy( struct pipe_context *pipe ) static boolean failover_draw_elements( struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, + unsigned indexSize, unsigned indexOffset, unsigned prim, unsigned start, unsigned count) { struct failover_context *failover = failover_context( pipe ); @@ -65,6 +65,7 @@ static boolean failover_draw_elements( struct pipe_context *pipe, if (!failover->hw->draw_elements( failover->hw, indexBuffer, indexSize, + indexOffset, prim, start, count )) { @@ -84,6 +85,7 @@ static boolean failover_draw_elements( struct pipe_context *pipe, failover->sw->draw_elements( failover->sw, indexBuffer, indexSize, + indexOffset, prim, start, count ); @@ -102,7 +104,7 @@ static boolean failover_draw_elements( struct pipe_context *pipe, static boolean failover_draw_arrays( struct pipe_context *pipe, unsigned prim, unsigned start, unsigned count) { - return failover_draw_elements(pipe, NULL, 0, prim, start, count); + return failover_draw_elements(pipe, NULL, 0, 0, prim, start, count); } diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 161f8ce697..adc91fb473 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -178,7 +178,7 @@ i915_end_query(struct pipe_context *pipe, struct pipe_query_object *q) static boolean i915_draw_elements( struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, + unsigned indexSize, unsigned indexOffset, unsigned prim, unsigned start, unsigned count) { struct i915_context *i915 = i915_context( pipe ); @@ -202,9 +202,10 @@ i915_draw_elements( struct pipe_context *pipe, } /* Map index buffer, if present */ if (indexBuffer) { - void *mapped_indexes + ubyte *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, PIPE_BUFFER_FLAG_READ); + mapped_indexes += indexOffset; draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { @@ -262,7 +263,7 @@ i915_draw_elements( struct pipe_context *pipe, static boolean i915_draw_arrays( struct pipe_context *pipe, unsigned prim, unsigned start, unsigned count) { - return i915_draw_elements(pipe, NULL, 0, prim, start, count); + return i915_draw_elements(pipe, NULL, 0, 0, prim, start, count); } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 2558a6341c..08082f9367 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -71,7 +71,7 @@ struct pipe_context { boolean (*draw_elements)( struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, + unsigned indexSize, unsigned indexOffset, unsigned mode, unsigned start, unsigned count); /** Clear a surface to given value (no scissor; clear whole surface) */ diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 64a4fbe333..d5cd6beae8 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -75,7 +75,7 @@ boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { - return softpipe_draw_elements(pipe, NULL, 0, mode, start, count); + return softpipe_draw_elements(pipe, NULL, 0, 0, mode, start, count); } @@ -90,7 +90,7 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, boolean softpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, + unsigned indexSize, unsigned indexOffset, unsigned mode, unsigned start, unsigned count) { struct softpipe_context *sp = softpipe_context(pipe); @@ -130,9 +130,10 @@ softpipe_draw_elements(struct pipe_context *pipe, } /* Map index buffer, if present */ if (indexBuffer) { - void *mapped_indexes + ubyte *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, PIPE_BUFFER_FLAG_READ); + mapped_indexes += indexOffset; draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index c194f0ea0d..8747f34cdd 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -131,7 +131,7 @@ boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, boolean softpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, + unsigned indexSize, unsigned indexOffset, unsigned mode, unsigned start, unsigned count); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index fbf7713f58..50251a816b 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -197,7 +197,7 @@ st_draw_vbo(GLcontext *ctx, /* indexed primitive */ struct gl_buffer_object *bufobj = ib->obj; struct pipe_buffer_handle *indexBuf = NULL; - unsigned indexSize, i; + unsigned indexSize, indexOffset, i; switch (ib->type) { case GL_UNSIGNED_INT: @@ -218,17 +218,19 @@ st_draw_vbo(GLcontext *ctx, /* elements/indexes are in a real VBO */ struct st_buffer_object *stobj = st_buffer_object(bufobj); winsys->buffer_reference(winsys, &indexBuf, stobj->buffer); + indexOffset = (unsigned) ib->ptr; } else { /* element/indicies are in user space memory */ indexBuf = winsys->user_buffer_create(winsys, (void *) ib->ptr, ib->count * indexSize); + indexOffset = 0; } /* draw */ for (i = 0; i < nr_prims; i++) { - pipe->draw_elements(pipe, indexBuf, indexSize, + pipe->draw_elements(pipe, indexBuf, indexSize, indexOffset, prims[i].mode, prims[i].start, prims[i].count); } -- cgit v1.2.3 From 40a0b053f78acd0a08f6876518b8e9301480431f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 14:35:40 -0600 Subject: Revert "Undo indexOffset change (I think, git???)" This reverts commit 11ce6244a79106c592364b30434b6ddade3fd6bb. --- src/mesa/pipe/failover/fo_context.c | 6 ++---- src/mesa/pipe/i915simple/i915_context.c | 7 +++---- src/mesa/pipe/p_context.h | 2 +- src/mesa/pipe/softpipe/sp_draw_arrays.c | 7 +++---- src/mesa/pipe/softpipe/sp_state.h | 2 +- src/mesa/state_tracker/st_draw.c | 6 ++---- 6 files changed, 12 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index b836ef29f3..076d516583 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -47,7 +47,7 @@ static void failover_destroy( struct pipe_context *pipe ) static boolean failover_draw_elements( struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, unsigned indexOffset, + unsigned indexSize, unsigned prim, unsigned start, unsigned count) { struct failover_context *failover = failover_context( pipe ); @@ -65,7 +65,6 @@ static boolean failover_draw_elements( struct pipe_context *pipe, if (!failover->hw->draw_elements( failover->hw, indexBuffer, indexSize, - indexOffset, prim, start, count )) { @@ -85,7 +84,6 @@ static boolean failover_draw_elements( struct pipe_context *pipe, failover->sw->draw_elements( failover->sw, indexBuffer, indexSize, - indexOffset, prim, start, count ); @@ -104,7 +102,7 @@ static boolean failover_draw_elements( struct pipe_context *pipe, static boolean failover_draw_arrays( struct pipe_context *pipe, unsigned prim, unsigned start, unsigned count) { - return failover_draw_elements(pipe, NULL, 0, 0, prim, start, count); + return failover_draw_elements(pipe, NULL, 0, prim, start, count); } diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index adc91fb473..161f8ce697 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -178,7 +178,7 @@ i915_end_query(struct pipe_context *pipe, struct pipe_query_object *q) static boolean i915_draw_elements( struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, unsigned indexOffset, + unsigned indexSize, unsigned prim, unsigned start, unsigned count) { struct i915_context *i915 = i915_context( pipe ); @@ -202,10 +202,9 @@ i915_draw_elements( struct pipe_context *pipe, } /* Map index buffer, if present */ if (indexBuffer) { - ubyte *mapped_indexes + void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, PIPE_BUFFER_FLAG_READ); - mapped_indexes += indexOffset; draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { @@ -263,7 +262,7 @@ i915_draw_elements( struct pipe_context *pipe, static boolean i915_draw_arrays( struct pipe_context *pipe, unsigned prim, unsigned start, unsigned count) { - return i915_draw_elements(pipe, NULL, 0, 0, prim, start, count); + return i915_draw_elements(pipe, NULL, 0, prim, start, count); } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 08082f9367..2558a6341c 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -71,7 +71,7 @@ struct pipe_context { boolean (*draw_elements)( struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, unsigned indexOffset, + unsigned indexSize, unsigned mode, unsigned start, unsigned count); /** Clear a surface to given value (no scissor; clear whole surface) */ diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index d5cd6beae8..64a4fbe333 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -75,7 +75,7 @@ boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) { - return softpipe_draw_elements(pipe, NULL, 0, 0, mode, start, count); + return softpipe_draw_elements(pipe, NULL, 0, mode, start, count); } @@ -90,7 +90,7 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, boolean softpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, unsigned indexOffset, + unsigned indexSize, unsigned mode, unsigned start, unsigned count) { struct softpipe_context *sp = softpipe_context(pipe); @@ -130,10 +130,9 @@ softpipe_draw_elements(struct pipe_context *pipe, } /* Map index buffer, if present */ if (indexBuffer) { - ubyte *mapped_indexes + void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, PIPE_BUFFER_FLAG_READ); - mapped_indexes += indexOffset; draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 8747f34cdd..c194f0ea0d 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -131,7 +131,7 @@ boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, boolean softpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, unsigned indexOffset, + unsigned indexSize, unsigned mode, unsigned start, unsigned count); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 50251a816b..fbf7713f58 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -197,7 +197,7 @@ st_draw_vbo(GLcontext *ctx, /* indexed primitive */ struct gl_buffer_object *bufobj = ib->obj; struct pipe_buffer_handle *indexBuf = NULL; - unsigned indexSize, indexOffset, i; + unsigned indexSize, i; switch (ib->type) { case GL_UNSIGNED_INT: @@ -218,19 +218,17 @@ st_draw_vbo(GLcontext *ctx, /* elements/indexes are in a real VBO */ struct st_buffer_object *stobj = st_buffer_object(bufobj); winsys->buffer_reference(winsys, &indexBuf, stobj->buffer); - indexOffset = (unsigned) ib->ptr; } else { /* element/indicies are in user space memory */ indexBuf = winsys->user_buffer_create(winsys, (void *) ib->ptr, ib->count * indexSize); - indexOffset = 0; } /* draw */ for (i = 0; i < nr_prims; i++) { - pipe->draw_elements(pipe, indexBuf, indexSize, indexOffset, + pipe->draw_elements(pipe, indexBuf, indexSize, prims[i].mode, prims[i].start, prims[i].count); } -- cgit v1.2.3 From 7636aac54bb357ce907ae6cc2c41f32957faecf3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 14:40:09 -0600 Subject: Fix glDrawElements + VBO rendering when ptr/offset is not zero. When an index VBO is bound, glDrawElement's indices pointer is really an offset into the element buffer. Add that offset to the prims[i].start value before calling pipe->draw_elements(). Fixes vbo.c conform test. --- src/mesa/state_tracker/st_draw.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index fbf7713f58..d7d9a1d203 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -197,7 +197,7 @@ st_draw_vbo(GLcontext *ctx, /* indexed primitive */ struct gl_buffer_object *bufobj = ib->obj; struct pipe_buffer_handle *indexBuf = NULL; - unsigned indexSize, i; + unsigned indexSize, indexOffset, i; switch (ib->type) { case GL_UNSIGNED_INT: @@ -218,18 +218,21 @@ st_draw_vbo(GLcontext *ctx, /* elements/indexes are in a real VBO */ struct st_buffer_object *stobj = st_buffer_object(bufobj); winsys->buffer_reference(winsys, &indexBuf, stobj->buffer); + indexOffset = (unsigned) ib->ptr / indexSize; } else { /* element/indicies are in user space memory */ indexBuf = winsys->user_buffer_create(winsys, (void *) ib->ptr, ib->count * indexSize); + indexOffset = 0; } /* draw */ for (i = 0; i < nr_prims; i++) { pipe->draw_elements(pipe, indexBuf, indexSize, - prims[i].mode, prims[i].start, prims[i].count); + prims[i].mode, + prims[i].start + indexOffset, prims[i].count); } winsys->buffer_reference(winsys, &indexBuf, NULL); -- cgit v1.2.3 From 1c5fec714d30c02d6d00d95215c2e302c4c65cc3 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 24 Oct 2007 19:20:19 -0400 Subject: Simplify branch handling. Prepare for proper decleration extraction. --- src/mesa/pipe/llvm/instructions.cpp | 2 +- src/mesa/pipe/llvm/llvmtgsi.cpp | 22 +++++++++++-- src/mesa/pipe/llvm/storage.cpp | 66 +++++++++---------------------------- src/mesa/pipe/llvm/storage.h | 18 ++-------- 4 files changed, 38 insertions(+), 70 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 6ded3f1249..622d420f64 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -968,7 +968,7 @@ void Instructions::ifop(llvm::Value *in) m_block); FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_UNE, x, float0, name("xcmp"), m_block); - new BranchInst(ifend, ifthen, xcmp, m_block); + new BranchInst(ifthen, ifend, xcmp, m_block); //m_block = yblock; diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index b8e4e39316..ec284a624a 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -139,9 +139,24 @@ static inline void AddStandardCompilePasses(PassManager &PM) { static void translate_declaration(llvm::Module *module, + Storage *storage, struct tgsi_full_declaration *decl, struct tgsi_full_declaration *fd) { + if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: { + int start = decl->u.DeclarationRange.First; + int end = decl->u.DeclarationRange.Last; + for (int i = start; i <= end; ++i) { + storage->declareTemp(i); + } + } + break; + default: + assert( 0 ); + } + } } @@ -180,8 +195,10 @@ translate_instruction(llvm::Module *module, val = storage->inputElement(src->SrcRegister.Index, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { val = storage->tempElement(src->SrcRegister.Index); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + fprintf(stderr, "FIXME: do somethign with immediates?\n"); } else { - fprintf(stderr, "ERROR: not supported llvm source\n"); + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); return; } @@ -447,7 +464,6 @@ translate_instruction(llvm::Module *module, case TGSI_OPCODE_ENDIF: { instr->endif(); storage->setCurrentBlock(instr->currentBlock()); - storage->popPhiNode(); return; //just update the state } break; @@ -642,7 +658,7 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) switch (parse.FullToken.Token.Type) { case TGSI_TOKEN_TYPE_DECLARATION: - translate_declaration(mod, + translate_declaration(mod, &storage, &parse.FullToken.FullDeclaration, &fd); break; diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 75ee18567d..6721f387c8 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -192,31 +192,29 @@ llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) } -llvm::Value *Storage::tempElement(int idx) const +llvm::Value *Storage::tempElement(int idx) { - Value *ret = m_temps[idx]; - if (!ret) + Value *ptr = m_temps[idx]; + if (!ptr) return m_undefFloatVec; - return ret; + llvm::LoadInst *li = new LoadInst(ptr, name("temp"), + false, m_block); + li->setAlignment(8); + return li; } void Storage::setTempElement(int idx, llvm::Value *val, int mask) { if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = m_temps[idx]; + llvm::Value *templ = 0; + if (m_temps[idx]) + templ = tempElement(idx); val = maskWrite(val, mask, templ); } - llvm::Value *templ = m_temps[idx]; - if (templ) { - BasicBlock *block = m_varBlocks[templ]; - if (block != m_block) { - addPhiNode(idx, val, m_block, templ, block); - } else - updatePhiNode(idx, val); - } - - m_temps[idx] = val; - m_varBlocks[val] = m_block; + llvm::Value *ptr = m_temps[idx]; + assert(ptr); + StoreInst *st = new StoreInst(val, ptr, false, m_block); + st->setAlignment(8); } void Storage::store(int dstIdx, llvm::Value *val, int mask) @@ -308,39 +306,7 @@ void Storage::setCurrentBlock(llvm::BasicBlock *block) m_block = block; } -void Storage::addPhiNode(int idx, llvm::Value *val1, llvm::BasicBlock *blk1, - llvm::Value *val2, llvm::BasicBlock *blk2) -{ - PhiNode node; - node.val1 = val1; - node.block1 = blk1; - node.val2 = val2; - node.block2 = blk2; - m_phiNodes[idx] = node; -} - -void Storage::updatePhiNode(int idx, llvm::Value *val1) -{ - if (m_phiNodes.find(idx) == m_phiNodes.end()) - return; - PhiNode node = m_phiNodes[idx]; - node.val1 = val1; - m_phiNodes[idx] = node; -} - -void Storage::popPhiNode() +void Storage::declareTemp(int idx) { - if (!m_phiNodes.empty()) { - std::map::const_iterator itr; - for (itr = m_phiNodes.begin(); itr != m_phiNodes.end(); ++itr) { - PhiNode node = (*itr).second; - PHINode *dest = new PHINode(m_floatVecType, - name("phiDest"), m_block); - dest->reserveOperandSpace(2); - dest->addIncoming(node.val1, node.block1); - dest->addIncoming(node.val2, node.block2); - m_temps[(*itr).first] = dest; - } - } - m_phiNodes.clear(); + m_temps[idx] = new AllocaInst(m_floatVecType, name("temp"), m_block); } diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index bc8cffdf66..26b2579b07 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -60,8 +60,9 @@ public: llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); llvm::Value *constElement(int idx, llvm::Value *indIdx =0); - llvm::Value *tempElement(int idx) const; + llvm::Value *tempElement(int idx); void setTempElement(int idx, llvm::Value *val, int mask); + void declareTemp(int idx); llvm::Value *addrElement(int idx) const; void setAddrElement(int idx, llvm::Value *val, int mask); @@ -72,8 +73,6 @@ public: void store(int dstIdx, llvm::Value *val, int mask); - void popPhiNode(); - int numConsts() const; private: llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); @@ -104,19 +103,6 @@ private: int m_idx; int m_numConsts; - - void addPhiNode(int, llvm::Value*, llvm::BasicBlock*, - llvm::Value*, llvm::BasicBlock*); - void updatePhiNode(int, llvm::Value*); - struct PhiNode { - llvm::Value *val1; - llvm::BasicBlock *block1; - llvm::Value *val2; - llvm::BasicBlock *block2; - }; - - std::map m_varBlocks; - std::map m_phiNodes; }; #endif -- cgit v1.2.3 From 5022ee43fc7916f16bdce8b076bad78fca262a62 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 24 Oct 2007 19:26:09 -0400 Subject: Make branching work :) Simply allow output variables as valid operand inputs when they've been assigned already. --- src/mesa/pipe/llvm/llvmtgsi.cpp | 4 ++-- src/mesa/pipe/llvm/storage.cpp | 28 ++++++++++++++++++++++++++++ src/mesa/pipe/llvm/storage.h | 1 + 3 files changed, 31 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index ec284a624a..93dd2ea46a 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -195,8 +195,8 @@ translate_instruction(llvm::Module *module, val = storage->inputElement(src->SrcRegister.Index, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { val = storage->tempElement(src->SrcRegister.Index); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - fprintf(stderr, "FIXME: do somethign with immediates?\n"); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->outputElement(src->SrcRegister.Index, indIdx); } else { fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); return; diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 6721f387c8..603053c0e8 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -310,3 +310,31 @@ void Storage::declareTemp(int idx) { m_temps[idx] = new AllocaInst(m_floatVecType, name("temp"), m_block); } + +llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx ) +{ + GetElementPtrInst *getElem = 0; + + if (indIdx) { + getElem = new GetElementPtrInst(m_IN, + BinaryOperator::create(Instruction::Add, + indIdx, + constantInt(idx), + name("add"), + m_block), + name("output_ptr"), + m_block); + } else { + getElem = new GetElementPtrInst(m_IN, + constantInt(idx), + name("output_ptr"), + m_block); + } + + LoadInst *load = new LoadInst(getElem, name("output"), + false, m_block); + load->setAlignment(8); + m_inputs[idx] = load; + + return load; +} diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index 26b2579b07..dd5de35073 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -59,6 +59,7 @@ public: llvm::Constant *shuffleMask(int vec); llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); llvm::Value *constElement(int idx, llvm::Value *indIdx =0); + llvm::Value *outputElement(int idx, llvm::Value *indIdx =0); llvm::Value *tempElement(int idx); void setTempElement(int idx, llvm::Value *val, int mask); -- cgit v1.2.3 From 731352f03ee998050331104d90abb47ee1377b3f Mon Sep 17 00:00:00 2001 From: keithw Date: Thu, 25 Oct 2007 10:18:52 +0100 Subject: set vcache.referenced bit always --- src/mesa/pipe/draw/draw_vertex_cache.c | 18 +++++------------- 1 file changed, 5 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index 56a4c34df5..511f371ac8 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -68,13 +68,8 @@ static struct vertex_header *get_vertex( struct draw_context *draw, /* If slot is in use, use the overflow area: */ if (draw->vcache.referenced & (1 << slot)) { -// fprintf(stderr, "o"); slot = VCACHE_SIZE + draw->vcache.overflow++; } - else { -// fprintf(stderr, "."); - draw->vcache.referenced |= (1 << slot); /* slot now in use */ - } assert(slot < Elements(draw->vcache.idx)); @@ -95,16 +90,13 @@ static struct vertex_header *get_vertex( struct draw_context *draw, draw->vcache.vertex[slot]->pad = 0; draw->vcache.vertex[slot]->vertex_id = ~0; } - else { -// fprintf(stderr, "*"); - /* primitive flushing may have cleared the bitfield but did not - * clear the idx[] array values. Set the bit now. This fixes a - * bug found when drawing long triangle fans. - */ - draw->vcache.referenced |= (1 << slot); - } + /* primitive flushing may have cleared the bitfield but did not + * clear the idx[] array values. Set the bit now. This fixes a + * bug found when drawing long triangle fans. + */ + draw->vcache.referenced |= (1 << slot); return draw->vcache.vertex[slot]; } -- cgit v1.2.3 From 7073ef96824242669735a8681519e1a0cee14309 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 25 Oct 2007 07:18:01 -0400 Subject: Implement else ops. --- src/mesa/pipe/llvm/instructions.cpp | 13 ++++++++++++- src/mesa/pipe/llvm/instructions.h | 1 + src/mesa/pipe/llvm/llvmtgsi.cpp | 6 +++++- 3 files changed, 18 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 622d420f64..a0645c5804 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -956,7 +956,7 @@ Function* makeLitFunction(Module *mod) { void Instructions::ifop(llvm::Value *in) { BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); - BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); + BasicBlock *ifend = new BasicBlock(name("ifthenend"), m_func,0); //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0); //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0); @@ -981,6 +981,17 @@ llvm::BasicBlock * Instructions::currentBlock() const return m_block; } +void Instructions::elseop() +{ + assert(!m_ifStack.empty()); + BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); + new BranchInst(ifend, m_block); + m_block = m_ifStack.top(); + m_block->setName(name("ifelse")); + m_ifStack.pop(); + m_ifStack.push(ifend); +} + void Instructions::endif() { assert(!m_ifStack.empty()); diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 8b47accd15..b0608e1da6 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -60,6 +60,7 @@ public: llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); void endif(); + void elseop(); llvm::Value *ex2(llvm::Value *in); llvm::Value *floor(llvm::Value *in); llvm::Value *frc(llvm::Value *in); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 93dd2ea46a..6ff4bc2270 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -459,7 +459,11 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_REP: break; - case TGSI_OPCODE_ELSE: + case TGSI_OPCODE_ELSE: { + instr->elseop(); + storage->setCurrentBlock(instr->currentBlock()); + return; //only state update + } break; case TGSI_OPCODE_ENDIF: { instr->endif(); -- cgit v1.2.3 From 1d26e9c447fd9746b2219edbf65b1991521bcfe7 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 25 Oct 2007 07:52:59 -0400 Subject: Fix nested conditionals --- src/mesa/pipe/llvm/storage.cpp | 9 --------- src/mesa/pipe/llvm/storage.h | 5 ++--- 2 files changed, 2 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 603053c0e8..1072917198 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -123,9 +123,6 @@ llvm::ConstantInt *Storage::constantInt(int idx) llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) { - if (!indIdx && m_inputs.find(idx) != m_inputs.end()) { - return m_inputs[idx]; - } GetElementPtrInst *getElem = 0; if (indIdx) { @@ -147,7 +144,6 @@ llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) LoadInst *load = new LoadInst(getElem, name("input"), false, m_block); load->setAlignment(8); - m_inputs[idx] = load; return load; } @@ -155,9 +151,6 @@ llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) { m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; - if (!indIdx && m_consts.find(idx) != m_consts.end()) { - return m_consts[idx]; - } GetElementPtrInst *getElem = 0; @@ -178,7 +171,6 @@ llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) LoadInst *load = new LoadInst(getElem, name("const"), false, m_block); load->setAlignment(8); - m_consts[idx] = load; return load; } @@ -334,7 +326,6 @@ llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx ) LoadInst *load = new LoadInst(getElem, name("output"), false, m_block); load->setAlignment(8); - m_inputs[idx] = load; return load; } diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index dd5de35073..a844d1c30f 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -47,7 +47,6 @@ namespace llvm { class Storage { - typedef std::map LoadMap; public: Storage(llvm::BasicBlock *block, llvm::Value *out, @@ -75,9 +74,11 @@ public: void store(int dstIdx, llvm::Value *val, int mask); int numConsts() const; + private: llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); const char *name(const char *prefix); + private: llvm::BasicBlock *m_block; llvm::Value *m_OUT; @@ -89,8 +90,6 @@ private: std::vector m_temps; std::vector m_addrs; std::vector m_dstCache; - LoadMap m_inputs; - LoadMap m_consts; llvm::VectorType *m_floatVecType; llvm::VectorType *m_intVecType; -- cgit v1.2.3 From 1d17cb721afaa53317614af90488a45c26e083e3 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 25 Oct 2007 09:03:53 -0400 Subject: Fix nested swizzles. Actually fetch the destination contents instead of input. --- src/mesa/pipe/llvm/storage.cpp | 10 ++++++---- src/mesa/pipe/llvm/storage.h | 3 +++ 2 files changed, 9 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 1072917198..cba719a8be 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -212,7 +212,9 @@ void Storage::setTempElement(int idx, llvm::Value *val, int mask) void Storage::store(int dstIdx, llvm::Value *val, int mask) { if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = m_dstCache[dstIdx]; + llvm::Value *templ = 0; + if (m_destWriteMap[dstIdx]) + templ = outputElement(dstIdx); val = maskWrite(val, mask, templ); } @@ -222,7 +224,7 @@ void Storage::store(int dstIdx, llvm::Value *val, int mask) m_block); StoreInst *st = new StoreInst(val, getElem, false, m_block); st->setAlignment(8); - //m_dstCache[dstIdx] = st; + m_destWriteMap[dstIdx] = true; } llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ) @@ -308,7 +310,7 @@ llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx ) GetElementPtrInst *getElem = 0; if (indIdx) { - getElem = new GetElementPtrInst(m_IN, + getElem = new GetElementPtrInst(m_OUT, BinaryOperator::create(Instruction::Add, indIdx, constantInt(idx), @@ -317,7 +319,7 @@ llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx ) name("output_ptr"), m_block); } else { - getElem = new GetElementPtrInst(m_IN, + getElem = new GetElementPtrInst(m_OUT, constantInt(idx), name("output_ptr"), m_block); diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index a844d1c30f..ebdfcdefd6 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -34,6 +34,7 @@ #define STORAGE_H #include +#include #include namespace llvm { @@ -103,6 +104,8 @@ private: int m_idx; int m_numConsts; + + std::map m_destWriteMap; }; #endif -- cgit v1.2.3 From 4a4e6f2cabdf39e9032e3e513f9966f471454077 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 25 Oct 2007 11:47:25 -0400 Subject: Implement loops --- src/mesa/pipe/llvm/instructions.cpp | 64 +++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.h | 11 ++++++- src/mesa/pipe/llvm/llvmtgsi.cpp | 21 +++++++++--- 3 files changed, 91 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index a0645c5804..5b54af8717 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -1016,3 +1016,67 @@ llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, return add(m, mul(s, in3)); } +void Instructions::beginLoop() +{ + BasicBlock *begin = new BasicBlock(name("loop"), m_func,0); + BasicBlock *end = new BasicBlock(name("endloop"), m_func,0); + + new BranchInst(begin, m_block); + Loop loop; + loop.begin = begin; + loop.end = end; + m_block = begin; + m_loopStack.push(loop); +} + +void Instructions::endLoop() +{ + assert(!m_loopStack.empty()); + Loop loop = m_loopStack.top(); + new BranchInst(loop.begin, m_block); + loop.end->moveAfter(m_block); + m_block = loop.end; + m_loopStack.pop(); +} + +void Instructions::brk() +{ + assert(!m_loopStack.empty()); + BasicBlock *unr = new BasicBlock(name("unreachable"), m_func,0); + new BranchInst(m_loopStack.top().end, m_block); + m_block = unr; +} + +llvm::Value * Instructions::trunc(llvm::Value *in) +{ + ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), + name("extractx"), + m_block); + ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), + name("extracty"), + m_block); + ExtractElementInst *z = new ExtractElementInst(in, unsigned(2), + name("extractz"), + m_block); + ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), + name("extractw"), + m_block); + CastInst *icastx = new FPToSIInst(x, IntegerType::get(32), + name("ftoix"), m_block); + CastInst *icasty = new FPToSIInst(y, IntegerType::get(32), + name("ftoiy"), m_block); + CastInst *icastz = new FPToSIInst(z, IntegerType::get(32), + name("ftoiz"), m_block); + CastInst *icastw = new FPToSIInst(w, IntegerType::get(32), + name("ftoiw"), m_block); + CastInst *fx = new SIToFPInst(icastx, Type::FloatTy, + name("fx"), m_block); + CastInst *fy = new SIToFPInst(icasty, Type::FloatTy, + name("fy"), m_block); + CastInst *fz = new SIToFPInst(icastz, Type::FloatTy, + name("fz"), m_block); + CastInst *fw = new SIToFPInst(icastw, Type::FloatTy, + name("fw"), m_block); + return vectorFromVals(fx, fy, fz, fw); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index b0608e1da6..29ae168e76 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -54,13 +54,16 @@ public: llvm::Value *abs(llvm::Value *in1); llvm::Value *arl(llvm::Value *in1); llvm::Value *add(llvm::Value *in1, llvm::Value *in2); + void beginLoop(); + void brk(); llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); - void endif(); void elseop(); + void endif(); + void endLoop(); llvm::Value *ex2(llvm::Value *in); llvm::Value *floor(llvm::Value *in); llvm::Value *frc(llvm::Value *in); @@ -81,6 +84,7 @@ public: llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); + llvm::Value *trunc(llvm::Value *in); void printVector(llvm::Value *val); private: @@ -115,6 +119,11 @@ private: llvm::Constant *m_fmtPtr; std::stack m_ifStack; + struct Loop { + llvm::BasicBlock *begin; + llvm::BasicBlock *end; + }; + std::stack m_loopStack; }; #endif diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 6ff4bc2270..6eae46d3ae 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -447,7 +447,10 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_TXL: break; - case TGSI_OPCODE_BRK: + case TGSI_OPCODE_BRK: { + instr->brk(); + return; + } break; case TGSI_OPCODE_IF: { instr->ifop(inputs[0]); @@ -485,7 +488,9 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_NOT: break; - case TGSI_OPCODE_TRUNC: + case TGSI_OPCODE_TRUNC: { + out = instr->trunc(inputs[0]); + } break; case TGSI_OPCODE_SHL: break; @@ -511,11 +516,19 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_ENDPRIM: break; - case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_BGNLOOP2: { + instr->beginLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } break; case TGSI_OPCODE_BGNSUB: break; - case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_ENDLOOP2: { + instr->endLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } break; case TGSI_OPCODE_ENDSUB: break; -- cgit v1.2.3 From 8568fadaf22787bc1652bb6f818b9c5c1ec81318 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 25 Oct 2007 18:24:51 +0200 Subject: LLVM requires linking with the C++ compiler. It's basically luck if it works with the C compiler on some platforms... --- src/mesa/Makefile | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index e9dfb95e5c..521cbeb606 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -67,7 +67,7 @@ linux-solo: depend subdirs libmesa.a # Stand-alone Mesa libGL, no built-in drivers (DirectFB) libgl-core: $(CORE_OBJECTS) - @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ + @ $(TOP)/bin/mklib -o $(GL_LIB) \ -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) $(CORE_OBJECTS) \ $(GL_LIB_DEPS) @@ -80,7 +80,7 @@ directfb: depend subdirs libgl-core # fbdev Mesa driver (libGL.so) fbdev: $(CORE_OBJECTS) $(FBDEV_DRIVER_OBJECTS) $(COMMON_DRIVER_OBJECTS) - @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ + @ $(TOP)/bin/mklib -o $(GL_LIB) \ -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ $(CORE_OBJECTS) $(FBDEV_DRIVER_OBJECTS) \ @@ -113,7 +113,7 @@ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) - @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ + @ $(TOP)/bin/mklib -o $(GL_LIB) \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ @@ -122,13 +122,13 @@ $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) # Make the OSMesa library $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECTS) @ if [ "${DRIVER_DIRS}" = "osmesa" ] ; then \ - $(TOP)/bin/mklib -o $(OSMESA_LIB) -linker '$(CC)' \ + $(TOP)/bin/mklib -o $(OSMESA_LIB) \ -major $(MESA_MAJOR) \ -minor $(MESA_MINOR) -patch $(MESA_TINY) \ -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ $(OSMESA_LIB_DEPS) $(OSMESA16_OBJECTS) ; \ else \ - $(TOP)/bin/mklib -o $(OSMESA_LIB) -linker '$(CC)' \ + $(TOP)/bin/mklib -o $(OSMESA_LIB) \ -major $(MESA_MAJOR) \ -minor $(MESA_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ -- cgit v1.2.3 From 18a22a43024474cb63a8bb26d5136b4f38db213b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 12 Oct 2007 00:23:17 +0100 Subject: Briefly describe the buffer management interface. --- src/mesa/pipe/p_winsys.h | 31 ++++++++++++++++++++----------- 1 file changed, 20 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index c80559c730..7e5d394a90 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -30,11 +30,17 @@ /** + * \file * This is the interface that Gallium3D requires any window system * hosting it to implement. This is the only include file in Gallium3D * which is public. - * - * + */ + + +/** Opaque type for a buffer */ +struct pipe_buffer_handle; + +/** * Gallium3D drivers are (meant to be!) independent of both GL and the * window system. The window system provides a buffer manager and a * set of additional hooks for things like command buffer submission, @@ -44,11 +50,6 @@ * driver and the hardware driver about the format of command buffers, * etc. */ - - -/** Opaque type */ -struct pipe_buffer_handle; - struct pipe_winsys { /** @@ -57,12 +58,14 @@ struct pipe_winsys */ void (*flush_frontbuffer)( struct pipe_winsys *sws ); - /** for debug output */ + /** Debug output */ void (*printf)( struct pipe_winsys *sws, const char *, ... ); - /* The buffer manager is modeled after the dri_bufmgr interface, + /** + * The buffer manager is modeled after the dri_bufmgr interface, which + * in turn is modeled after the ARB_vertex_buffer_object extension, * but this is the subset that gallium cares about. Remember that * gallium gets to choose the interface it needs, and the window * systems must then implement that interface (rather than the @@ -77,7 +80,10 @@ struct pipe_winsys unsigned bytes); - /** flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE */ + /** + * Map the entire data store of a buffer object into the client's address. + * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. + */ void *(*buffer_map)( struct pipe_winsys *sws, struct pipe_buffer_handle *buf, unsigned flags ); @@ -90,16 +96,19 @@ struct pipe_winsys struct pipe_buffer_handle **ptr, struct pipe_buffer_handle *buf ); + /** Create the data store of a buffer and optionally initialize it */ void (*buffer_data)(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, unsigned size, const void *data ); + /** Modify some or all of the data contained in a buffer's data store */ void (*buffer_subdata)(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, const void *data); + /** Query some or all of the data contained in a buffer's data store */ void (*buffer_get_subdata)(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, unsigned long offset, @@ -117,4 +126,4 @@ struct pipe_winsys -#endif /* SP_WINSYS_H */ +#endif /* P_WINSYS_H */ -- cgit v1.2.3 From 86f09fd1a78c5ea060b62e6b657fd693a2d688a8 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 16:58:00 -0600 Subject: s/GLfloat/float/, s/GLint/int/, etc --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 162 ++++++++++++++++++------------------ 1 file changed, 81 insertions(+), 81 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 113a0f2c18..9c5c1f26be 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -128,8 +128,8 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) struct tgsi_full_declaration *declarations; uint maxInstructions = 10, numInstructions = 0; uint maxDeclarations = 10, numDeclarations = 0; - GLuint k; - GLuint instno = 0; + uint k; + uint instno = 0; mach->ImmLimit = 0; labels->count = 0; @@ -147,8 +147,8 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) } while( !tgsi_parse_end_of_tokens( &parse ) ) { - GLuint pointer = parse.Position; - GLuint i; + uint pointer = parse.Position; + uint i; tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { @@ -225,10 +225,10 @@ void tgsi_exec_machine_init( struct tgsi_exec_machine *mach, const struct tgsi_token *tokens, - GLuint numSamplers, + uint numSamplers, struct tgsi_sampler *samplers) { - GLuint i, k; + uint i, k; struct tgsi_parse_context parse; #if 0 @@ -272,10 +272,10 @@ micro_abs( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) fabs( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) fabs( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) fabs( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) fabs( (GLdouble) src->f[3] ); + dst->f[0] = (float) fabs( (double) src->f[0] ); + dst->f[1] = (float) fabs( (double) src->f[1] ); + dst->f[2] = (float) fabs( (double) src->f[2] ); + dst->f[3] = (float) fabs( (double) src->f[3] ); } static void @@ -319,10 +319,10 @@ micro_ceil( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) ceil( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) ceil( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) ceil( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) ceil( (GLdouble) src->f[3] ); + dst->f[0] = (float) ceil( (double) src->f[0] ); + dst->f[1] = (float) ceil( (double) src->f[1] ); + dst->f[2] = (float) ceil( (double) src->f[2] ); + dst->f[3] = (float) ceil( (double) src->f[3] ); } static void @@ -330,10 +330,10 @@ micro_cos( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) cos( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) cos( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) cos( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) cos( (GLdouble) src->f[3] ); + dst->f[0] = (float) cos( (double) src->f[0] ); + dst->f[1] = (float) cos( (double) src->f[1] ); + dst->f[2] = (float) cos( (double) src->f[2] ); + dst->f[3] = (float) cos( (double) src->f[3] ); } static void @@ -415,10 +415,10 @@ micro_exp2( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src) { - dst->f[0] = (GLfloat) pow( 2.0, (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) pow( 2.0, (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) pow( 2.0, (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) pow( 2.0, (GLdouble) src->f[3] ); + dst->f[0] = (float) pow( 2.0, (double) src->f[0] ); + dst->f[1] = (float) pow( 2.0, (double) src->f[1] ); + dst->f[2] = (float) pow( 2.0, (double) src->f[2] ); + dst->f[3] = (float) pow( 2.0, (double) src->f[3] ); } static void @@ -426,10 +426,10 @@ micro_f2it( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->i[0] = (GLint) src->f[0]; - dst->i[1] = (GLint) src->f[1]; - dst->i[2] = (GLint) src->f[2]; - dst->i[3] = (GLint) src->f[3]; + dst->i[0] = (int) src->f[0]; + dst->i[1] = (int) src->f[1]; + dst->i[2] = (int) src->f[2]; + dst->i[3] = (int) src->f[3]; } static void @@ -437,10 +437,10 @@ micro_f2ut( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->u[0] = (GLuint) src->f[0]; - dst->u[1] = (GLuint) src->f[1]; - dst->u[2] = (GLuint) src->f[2]; - dst->u[3] = (GLuint) src->f[3]; + dst->u[0] = (uint) src->f[0]; + dst->u[1] = (uint) src->f[1]; + dst->u[2] = (uint) src->f[2]; + dst->u[3] = (uint) src->f[3]; } static void @@ -448,10 +448,10 @@ micro_flr( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) floor( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) floor( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) floor( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) floor( (GLdouble) src->f[3] ); + dst->f[0] = (float) floor( (double) src->f[0] ); + dst->f[1] = (float) floor( (double) src->f[1] ); + dst->f[2] = (float) floor( (double) src->f[2] ); + dst->f[3] = (float) floor( (double) src->f[3] ); } static void @@ -459,10 +459,10 @@ micro_frc( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = src->f[0] - (GLfloat) floor( (GLdouble) src->f[0] ); - dst->f[1] = src->f[1] - (GLfloat) floor( (GLdouble) src->f[1] ); - dst->f[2] = src->f[2] - (GLfloat) floor( (GLdouble) src->f[2] ); - dst->f[3] = src->f[3] - (GLfloat) floor( (GLdouble) src->f[3] ); + dst->f[0] = src->f[0] - (float) floor( (double) src->f[0] ); + dst->f[1] = src->f[1] - (float) floor( (double) src->f[1] ); + dst->f[2] = src->f[2] - (float) floor( (double) src->f[2] ); + dst->f[3] = src->f[3] - (float) floor( (double) src->f[3] ); } static void @@ -484,10 +484,10 @@ micro_i2f( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) src->i[0]; - dst->f[1] = (GLfloat) src->i[1]; - dst->f[2] = (GLfloat) src->i[2]; - dst->f[3] = (GLfloat) src->i[3]; + dst->f[0] = (float) src->i[0]; + dst->f[1] = (float) src->i[1]; + dst->f[2] = (float) src->i[2]; + dst->f[3] = (float) src->i[3]; } static void @@ -495,10 +495,10 @@ micro_lg2( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) log( (GLdouble) src->f[0] ) * 1.442695f; - dst->f[1] = (GLfloat) log( (GLdouble) src->f[1] ) * 1.442695f; - dst->f[2] = (GLfloat) log( (GLdouble) src->f[2] ) * 1.442695f; - dst->f[3] = (GLfloat) log( (GLdouble) src->f[3] ) * 1.442695f; + dst->f[0] = (float) log( (double) src->f[0] ) * 1.442695f; + dst->f[1] = (float) log( (double) src->f[1] ) * 1.442695f; + dst->f[2] = (float) log( (double) src->f[2] ) * 1.442695f; + dst->f[3] = (float) log( (double) src->f[3] ) * 1.442695f; } static void @@ -749,10 +749,10 @@ micro_pow( const union tgsi_exec_channel *src0, const union tgsi_exec_channel *src1 ) { - dst->f[0] = (GLfloat) pow( (GLdouble) src0->f[0], (GLdouble) src1->f[0] ); - dst->f[1] = (GLfloat) pow( (GLdouble) src0->f[1], (GLdouble) src1->f[1] ); - dst->f[2] = (GLfloat) pow( (GLdouble) src0->f[2], (GLdouble) src1->f[2] ); - dst->f[3] = (GLfloat) pow( (GLdouble) src0->f[3], (GLdouble) src1->f[3] ); + dst->f[0] = (float) pow( (double) src0->f[0], (double) src1->f[0] ); + dst->f[1] = (float) pow( (double) src0->f[1], (double) src1->f[1] ); + dst->f[2] = (float) pow( (double) src0->f[2], (double) src1->f[2] ); + dst->f[3] = (float) pow( (double) src0->f[3], (double) src1->f[3] ); } static void @@ -760,10 +760,10 @@ micro_rnd( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) floor( (GLdouble) (src->f[0] + 0.5f) ); - dst->f[1] = (GLfloat) floor( (GLdouble) (src->f[1] + 0.5f) ); - dst->f[2] = (GLfloat) floor( (GLdouble) (src->f[2] + 0.5f) ); - dst->f[3] = (GLfloat) floor( (GLdouble) (src->f[3] + 0.5f) ); + dst->f[0] = (float) floor( (double) (src->f[0] + 0.5f) ); + dst->f[1] = (float) floor( (double) (src->f[1] + 0.5f) ); + dst->f[2] = (float) floor( (double) (src->f[2] + 0.5f) ); + dst->f[3] = (float) floor( (double) (src->f[3] + 0.5f) ); } static void @@ -818,20 +818,20 @@ micro_sin( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) sin( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) sin( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) sin( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) sin( (GLdouble) src->f[3] ); + dst->f[0] = (float) sin( (double) src->f[0] ); + dst->f[1] = (float) sin( (double) src->f[1] ); + dst->f[2] = (float) sin( (double) src->f[2] ); + dst->f[3] = (float) sin( (double) src->f[3] ); } static void micro_sqrt( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) sqrt( (GLdouble) src->f[0] ); - dst->f[1] = (GLfloat) sqrt( (GLdouble) src->f[1] ); - dst->f[2] = (GLfloat) sqrt( (GLdouble) src->f[2] ); - dst->f[3] = (GLfloat) sqrt( (GLdouble) src->f[3] ); + dst->f[0] = (float) sqrt( (double) src->f[0] ); + dst->f[1] = (float) sqrt( (double) src->f[1] ); + dst->f[2] = (float) sqrt( (double) src->f[2] ); + dst->f[3] = (float) sqrt( (double) src->f[3] ); } static void @@ -851,10 +851,10 @@ micro_u2f( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (GLfloat) src->u[0]; - dst->f[1] = (GLfloat) src->u[1]; - dst->f[2] = (GLfloat) src->u[2]; - dst->f[3] = (GLfloat) src->u[3]; + dst->f[0] = (float) src->u[0]; + dst->f[1] = (float) src->u[1]; + dst->f[2] = (float) src->u[2]; + dst->f[3] = (float) src->u[3]; } static void @@ -872,8 +872,8 @@ micro_xor( static void fetch_src_file_channel( const struct tgsi_exec_machine *mach, - const GLuint file, - const GLuint swizzle, + const uint file, + const uint swizzle, const union tgsi_exec_channel *index, union tgsi_exec_channel *chan ) { @@ -905,13 +905,13 @@ fetch_src_file_channel( break; case TGSI_FILE_IMMEDIATE: - assert( index->i[0] < (GLint) mach->ImmLimit ); + assert( index->i[0] < (int) mach->ImmLimit ); chan->f[0] = mach->Imms[index->i[0]][swizzle]; - assert( index->i[1] < (GLint) mach->ImmLimit ); + assert( index->i[1] < (int) mach->ImmLimit ); chan->f[1] = mach->Imms[index->i[1]][swizzle]; - assert( index->i[2] < (GLint) mach->ImmLimit ); + assert( index->i[2] < (int) mach->ImmLimit ); chan->f[2] = mach->Imms[index->i[2]][swizzle]; - assert( index->i[3] < (GLint) mach->ImmLimit ); + assert( index->i[3] < (int) mach->ImmLimit ); chan->f[3] = mach->Imms[index->i[3]][swizzle]; break; @@ -953,10 +953,10 @@ fetch_source( const struct tgsi_exec_machine *mach, union tgsi_exec_channel *chan, const struct tgsi_full_src_register *reg, - const GLuint chan_index ) + const uint chan_index ) { union tgsi_exec_channel index; - GLuint swizzle; + uint swizzle; index.i[0] = index.i[1] = @@ -1066,7 +1066,7 @@ store_dest( const union tgsi_exec_channel *chan, const struct tgsi_full_dst_register *reg, const struct tgsi_full_instruction *inst, - GLuint chan_index ) + uint chan_index ) { union tgsi_exec_channel *dst; @@ -1138,7 +1138,7 @@ exec_kilp(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst) { GLbitfield uniquemask; - GLuint chan_index; + uint chan_index; GLbitfield kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ union tgsi_exec_channel r[1]; @@ -1149,8 +1149,8 @@ exec_kilp(struct tgsi_exec_machine *mach, for (chan_index = 0; chan_index < 4; chan_index++) { - GLuint swizzle; - GLuint i; + uint swizzle; + uint i; /* unswizzle channel */ swizzle = tgsi_util_get_full_src_register_extswizzle ( @@ -1186,8 +1186,8 @@ fetch_texel( struct tgsi_sampler *sampler, union tgsi_exec_channel *b, union tgsi_exec_channel *a ) { - GLuint j; - GLfloat rgba[NUM_CHANNELS][QUAD_SIZE]; + uint j; + float rgba[NUM_CHANNELS][QUAD_SIZE]; sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); @@ -1451,7 +1451,7 @@ exec_instruction( const struct tgsi_full_instruction *inst, int *pc ) { - GLuint chan_index; + uint chan_index; union tgsi_exec_channel r[8]; (*pc)++; -- cgit v1.2.3 From 3dde3715db671e32261bfb8b11fb599e5a0e6c5a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 16:58:47 -0600 Subject: update TILE_TOP/BOTTOM_* values to match softpipe --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 9c5c1f26be..24ee98061d 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -60,10 +60,10 @@ #include "main/macros.h" #endif -#define TILE_BOTTOM_LEFT 0 -#define TILE_BOTTOM_RIGHT 1 -#define TILE_TOP_LEFT 2 -#define TILE_TOP_RIGHT 3 +#define TILE_TOP_LEFT 0 +#define TILE_TOP_RIGHT 1 +#define TILE_BOTTOM_LEFT 2 +#define TILE_BOTTOM_RIGHT 3 #define TEMP_0_I TGSI_EXEC_TEMP_00000000_I #define TEMP_0_C TGSI_EXEC_TEMP_00000000_C -- cgit v1.2.3 From aaac436c6c06232e383d5cee3c6dfc69eb820c83 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 24 Oct 2007 17:01:23 -0600 Subject: added some comments --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 3 +++ src/mesa/pipe/tgsi/exec/tgsi_exec.h | 3 +++ 2 files changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 24ee98061d..d5a2639af1 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -65,6 +65,9 @@ #define TILE_BOTTOM_LEFT 2 #define TILE_BOTTOM_RIGHT 3 +/* + * Shorthand locations of various utility registers (_I = Index, _C = Channel) + */ #define TEMP_0_I TGSI_EXEC_TEMP_00000000_I #define TEMP_0_C TGSI_EXEC_TEMP_00000000_C #define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 978c2d574e..6c1d368342 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -53,6 +53,9 @@ struct tgsi_exec_labels unsigned count; }; +/* + * Locations of various utility registers (_I = Index, _C = Channel) + */ #define TGSI_EXEC_TEMP_00000000_I 32 #define TGSI_EXEC_TEMP_00000000_C 0 -- cgit v1.2.3 From afd19177e4e6571858fc94ab6be1b12bb54a04ed Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 25 Oct 2007 18:30:06 -0600 Subject: tex comments --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index d5a2639af1..3f464372ca 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -926,7 +926,7 @@ fetch_src_file_channel( break; case TGSI_FILE_OUTPUT: - /* vertex varying/output vars can be read too */ + /* vertex/fragment output vars can be read too */ chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; @@ -1989,16 +1989,24 @@ exec_instruction( break; case TGSI_OPCODE_TEX: - /* src arg0 is the texcoord */ + /* simple texture lookup */ + /* src[0] is the texcoord */ + /* src[1] is the sampler unit */ exec_tex(mach, inst, FALSE); break; case TGSI_OPCODE_TXB: - /* Texture lookup with lod bias (src0.w) */ + /* Texture lookup with lod bias */ + /* src[0] is the texcoord (src[0].w = load bias) */ + /* src[1] is the sampler unit */ exec_tex(mach, inst, TRUE); break; case TGSI_OPCODE_TXD: + /* Texture lookup with explict partial derivatives */ + /* src[0] = texcoord */ + /* src[1] = d[strq]/dx */ + /* src[2] = d[strq]/dy */ assert (0); break; -- cgit v1.2.3 From ee80e0b620c5b9af62dac8ad64a84042b46f5264 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 25 Oct 2007 18:50:15 -0600 Subject: Move the get/put_tile() functions to pipe_context. The _rgba versions are temporary until the state tracker is updated. --- src/mesa/drivers/x11/xm_surface.c | 4 +- src/mesa/pipe/i915simple/i915_surface.c | 17 +++--- src/mesa/pipe/p_context.h | 18 ++++++ src/mesa/pipe/p_state.h | 21 ------- src/mesa/pipe/softpipe/sp_context.c | 6 +- src/mesa/pipe/softpipe/sp_flush.c | 6 +- src/mesa/pipe/softpipe/sp_quad_blend.c | 8 ++- src/mesa/pipe/softpipe/sp_quad_colormask.c | 5 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 2 +- src/mesa/pipe/softpipe/sp_quad_output.c | 3 +- src/mesa/pipe/softpipe/sp_quad_stencil.c | 2 +- src/mesa/pipe/softpipe/sp_state_surface.c | 6 +- src/mesa/pipe/softpipe/sp_surface.c | 91 ++++++++++++++++++++++------- src/mesa/pipe/softpipe/sp_surface.h | 9 ++- src/mesa/pipe/softpipe/sp_tile_cache.c | 50 ++++++++-------- src/mesa/pipe/softpipe/sp_tile_cache.h | 6 +- src/mesa/state_tracker/st_cb_accum.c | 22 +++---- src/mesa/state_tracker/st_cb_drawpixels.c | 4 +- src/mesa/state_tracker/st_cb_readpixels.c | 2 +- src/mesa/state_tracker/st_cb_texture.c | 4 +- 20 files changed, 177 insertions(+), 109 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index f83c7917c4..340f796bc8 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -148,8 +148,8 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) switch (pipeFormat) { case PIPE_FORMAT_U_A8_R8_G8_B8: - xms->surface.surface.get_tile = get_tile; - xms->surface.surface.put_tile = put_tile; + xms->surface.get_tile = get_tile; + xms->surface.put_tile = put_tile; break; case PIPE_FORMAT_S8_Z24: break; diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index c094cf9ec4..a07a21c13b 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -29,7 +29,6 @@ #include "i915_state.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" -//#include "main/imports.h" struct i915_surface @@ -44,8 +43,9 @@ struct i915_surface * Share it someday. */ static void -i915_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) +i915_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, float *p) { const unsigned *src = ((const unsigned *) (ps->region->map + ps->offset)) @@ -82,8 +82,9 @@ i915_get_tile(struct pipe_surface *ps, static void -i915_put_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, const float *p) +i915_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const float *p) { /* any need to put tiles into i915 surfaces? */ assert(0); @@ -102,8 +103,8 @@ i915_surface_alloc(struct pipe_context *pipe, unsigned format) surf->surface.format = format; surf->surface.refcount = 1; - surf->surface.get_tile = i915_get_tile; - surf->surface.put_tile = i915_put_tile; + // surf->surface.get_tile = i915_get_tile; + // surf->surface.put_tile = i915_put_tile; return &surf->surface; } @@ -113,4 +114,6 @@ void i915_init_surface_functions(struct i915_context *i915) { i915->pipe.surface_alloc = i915_surface_alloc; + i915->pipe.get_tile_rgba = i915_get_tile_rgba; + i915->pipe.put_tile_rgba = i915_put_tile_rgba; } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 2558a6341c..a22ea3a8a0 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -188,6 +188,24 @@ struct pipe_context { unsigned face, unsigned level, unsigned zslice); + /** Get a block of raw pixel data from a surface */ + void (*get_tile)(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride); + /** Put a block of raw pixel data into a surface */ + void (*put_tile)(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride); + + /* XXX temporary here, move these to softpipe */ + void (*get_tile_rgba)(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, float *p); + void (*put_tile_rgba)(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const float *p); + + /* * Memory region functions * Some of these may go away... diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 69b11588e0..03045e4213 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -291,27 +291,6 @@ struct pipe_surface unsigned width, height; unsigned offset; /**< offset from start of region, in bytes */ unsigned refcount; - - /** - * Get block/tile of pixels from surface as floats - * If color surface, return float[4]. If depth surface, return float[1]. - */ - void (*get_tile)(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *p); - - /** - * Put block/tile of pixels into surface as floats - * If color surface, data is float[4]. If depth surface, data is float[1]. - */ - void (*put_tile)(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *p); - - /** As above, but data is raw pixel data */ - void (*get_tile_raw)(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, void *p); - /** As above, but data is raw pixel data */ - void (*put_tile_raw)(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const void *p); }; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index bf61019f62..476d4ac01c 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -171,9 +171,9 @@ softpipe_unmap_surfaces(struct softpipe_context *sp) uint i; for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) - sp_flush_tile_cache(sp->cbuf_cache[i]); - sp_flush_tile_cache(sp->zbuf_cache); - sp_flush_tile_cache(sp->sbuf_cache); + sp_flush_tile_cache(sp, sp->cbuf_cache[i]); + sp_flush_tile_cache(sp, sp->zbuf_cache); + sp_flush_tile_cache(sp, sp->sbuf_cache); for (i = 0; i < sp->framebuffer.num_cbufs; i++) { struct pipe_surface *ps = sp->framebuffer.cbufs[i]; diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index 2b077c2021..1010924bf6 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -57,13 +57,13 @@ softpipe_flush( struct pipe_context *pipe, for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) if (softpipe->cbuf_cache[i]) - sp_flush_tile_cache(softpipe->cbuf_cache[i]); + sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]); if (softpipe->zbuf_cache) - sp_flush_tile_cache(softpipe->zbuf_cache); + sp_flush_tile_cache(softpipe, softpipe->zbuf_cache); if (softpipe->sbuf_cache) - sp_flush_tile_cache(softpipe->sbuf_cache); + sp_flush_tile_cache(softpipe, softpipe->sbuf_cache); /* Need this call for hardware buffers before swapbuffers. * diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index b056e477b1..9b7a48669d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -107,7 +107,8 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) uint *dst4 = (uint *) dst; uint *res4 = (uint *) res; struct softpipe_cached_tile * - tile = sp_get_cached_tile(softpipe->cbuf_cache[0], quad->x0, quad->y0); + tile = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[0], + quad->x0, quad->y0); uint i, j; /* get/swizzle dest colors */ @@ -222,8 +223,9 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) static const float zero[4] = { 0, 0, 0, 0 }; static const float one[4] = { 1, 1, 1, 1 }; float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; - struct softpipe_cached_tile * - tile = sp_get_cached_tile(softpipe->cbuf_cache[0], quad->x0, quad->y0); + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[0], + quad->x0, quad->y0); uint i, j; if (softpipe->blend->logicop_enable) { diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index 3c0196dd5d..8872825555 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -48,8 +48,9 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; float dest[4][QUAD_SIZE]; - struct softpipe_cached_tile * - tile = sp_get_cached_tile(softpipe->cbuf_cache[0], quad->x0, quad->y0); + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[0], quad->x0, quad->y0); uint i, j; /* get/swizzle dest colors */ diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 29231322b8..05dafeca7c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -57,7 +57,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) unsigned zmask = 0; unsigned j; struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe->zbuf_cache, quad->x0, quad->y0); + = sp_get_cached_tile(softpipe, softpipe->zbuf_cache, quad->x0, quad->y0); assert(sps); /* shouldn't get here if there's no zbuffer */ diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index e86f42be46..f757a43927 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -43,7 +43,8 @@ output_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe->cbuf_cache[0], quad->x0, quad->y0); + = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[0], + quad->x0, quad->y0); /* in-tile pos: */ const int itx = quad->x0 % TILE_SIZE; const int ity = quad->y0 % TILE_SIZE; diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 4a3823d646..8475bf3b96 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -206,7 +206,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) ubyte ref, wrtMask, valMask; ubyte stencilVals[QUAD_SIZE]; struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe->sbuf_cache, quad->x0, quad->y0); + = sp_get_cached_tile(softpipe, softpipe->sbuf_cache, quad->x0, quad->y0); uint j; /* choose front or back face function, operator, etc */ diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index cd1e75c563..0960fc45ab 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -51,7 +51,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* check if changing cbuf */ if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) { /* flush old */ - sp_flush_tile_cache(sp->cbuf_cache[i]); + sp_flush_tile_cache(sp, sp->cbuf_cache[i]); /* unmap old */ sps = softpipe_surface(sp->framebuffer.cbufs[i]); if (sps && sps->surface.region) @@ -73,7 +73,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* zbuf changing? */ if (sp->framebuffer.zbuf != fb->zbuf) { /* flush old */ - sp_flush_tile_cache(sp->zbuf_cache); + sp_flush_tile_cache(sp, sp->zbuf_cache); /* unmap old */ sps = softpipe_surface(sp->framebuffer.zbuf); if (sps && sps->surface.region) @@ -98,7 +98,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* sbuf changing? */ if (sp->framebuffer.sbuf != fb->sbuf) { /* flush old */ - sp_flush_tile_cache(sp->sbuf_cache_sep); + sp_flush_tile_cache(sp, sp->sbuf_cache_sep); /* unmap old */ sps = softpipe_surface(sp->framebuffer.sbuf); if (sps && sps->surface.region) diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 057a311cd9..b8af1c561f 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -632,48 +632,48 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) switch (sps->surface.format) { case PIPE_FORMAT_U_A8_R8_G8_B8: - sps->surface.get_tile = a8r8g8b8_get_tile; - sps->surface.put_tile = a8r8g8b8_put_tile; + sps->get_tile = a8r8g8b8_get_tile; + sps->put_tile = a8r8g8b8_put_tile; break; case PIPE_FORMAT_U_A1_R5_G5_B5: - sps->surface.get_tile = a1r5g5b5_get_tile; + sps->get_tile = a1r5g5b5_get_tile; break; case PIPE_FORMAT_U_L8: - sps->surface.get_tile = l8_get_tile; + sps->get_tile = l8_get_tile; break; case PIPE_FORMAT_U_A8: - sps->surface.get_tile = a8_get_tile; + sps->get_tile = a8_get_tile; break; case PIPE_FORMAT_U_I8: - sps->surface.get_tile = i8_get_tile; + sps->get_tile = i8_get_tile; break; case PIPE_FORMAT_U_A8_L8: - sps->surface.get_tile = a8_l8_get_tile; + sps->get_tile = a8_l8_get_tile; break; case PIPE_FORMAT_S_R16_G16_B16_A16: - sps->surface.get_tile = r16g16b16a16_get_tile; - sps->surface.put_tile = r16g16b16a16_put_tile; + sps->get_tile = r16g16b16a16_get_tile; + sps->put_tile = r16g16b16a16_put_tile; break; case PIPE_FORMAT_U_Z16: - sps->surface.get_tile = z16_get_tile; - sps->surface.get_tile_raw = get_tile_raw16; - sps->surface.put_tile_raw = put_tile_raw16; + sps->get_tile = z16_get_tile; + sps->get_tile_raw = get_tile_raw16; + sps->put_tile_raw = put_tile_raw16; break; case PIPE_FORMAT_U_Z32: - sps->surface.get_tile = z32_get_tile; - sps->surface.get_tile_raw = get_tile_raw32; - sps->surface.put_tile_raw = put_tile_raw32; + sps->get_tile = z32_get_tile; + sps->get_tile_raw = get_tile_raw32; + sps->put_tile_raw = put_tile_raw32; break; case PIPE_FORMAT_S8_Z24: - sps->surface.get_tile = s8z24_get_tile; - sps->surface.get_tile_raw = get_tile_raw32; - sps->surface.put_tile_raw = put_tile_raw32; + sps->get_tile = s8z24_get_tile; + sps->get_tile_raw = get_tile_raw32; + sps->put_tile_raw = put_tile_raw32; break; case PIPE_FORMAT_U_S8: - sps->surface.get_tile_raw = get_tile_raw8; - sps->surface.put_tile_raw = put_tile_raw8; + sps->get_tile_raw = get_tile_raw8; + sps->put_tile_raw = put_tile_raw8; break; default: assert(0); @@ -739,8 +739,59 @@ softpipe_get_tex_surface(struct pipe_context *pipe, } +static void +get_tile_generic(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride) +{ + struct softpipe_surface *sps = softpipe_surface(ps); + sps->get_tile_raw(ps, x, y, w, h, p); +} + + +static void +put_tile_generic(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride) +{ + struct softpipe_surface *sps = softpipe_surface(ps); + sps->put_tile_raw(ps, x, y, w, h, p); +} + + +static void +get_tile_rgba_generic(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p) +{ + struct softpipe_surface *sps = softpipe_surface(ps); + sps->get_tile(ps, x, y, w, h, p); +} + + +static void +put_tile_rgba_generic(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p) +{ + struct softpipe_surface *sps = softpipe_surface(ps); + sps->put_tile(ps, x, y, w, h, p); +} + + + void sp_init_surface_functions(struct softpipe_context *sp) { sp->pipe.surface_alloc = softpipe_surface_alloc; + + sp->pipe.get_tile = get_tile_generic; + sp->pipe.put_tile = put_tile_generic; + + sp->pipe.get_tile_rgba = get_tile_rgba_generic; + sp->pipe.put_tile_rgba = put_tile_rgba_generic; } diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 06c0a01aee..0c2486a171 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -46,7 +46,14 @@ struct softpipe_tile_cache; struct softpipe_surface { struct pipe_surface surface; - /* no softpipe-specific extras now */ + void (*get_tile)(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, float *p); + void (*put_tile)(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const float *p); + void (*get_tile_raw)(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, void *p); + void (*put_tile_raw)(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const void *p); }; diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 129785d26d..421d7bdb1a 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -153,8 +153,10 @@ sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, void -sp_flush_tile_cache(struct softpipe_tile_cache *tc) +sp_flush_tile_cache(struct softpipe_context *softpipe, + struct softpipe_tile_cache *tc) { + struct pipe_context *pipe = &softpipe->pipe; struct pipe_surface *ps = &tc->surface->surface; boolean is_depth_stencil; int inuse = 0, pos; @@ -171,14 +173,14 @@ sp_flush_tile_cache(struct softpipe_tile_cache *tc) struct softpipe_cached_tile *tile = tc->entries + pos; if (tile->x >= 0) { if (is_depth_stencil) { - ps->put_tile_raw(ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32); + pipe->put_tile(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); } else { - ps->put_tile(ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + pipe->put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); } tile->x = tile->y = -1; /* mark as empty */ @@ -193,8 +195,10 @@ sp_flush_tile_cache(struct softpipe_tile_cache *tc) struct softpipe_cached_tile * -sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y) +sp_get_cached_tile(struct softpipe_context *softpipe, + struct softpipe_tile_cache *tc, int x, int y) { + struct pipe_context *pipe = &softpipe->pipe; struct pipe_surface *ps = &tc->surface->surface; boolean is_depth_stencil = (ps->format == PIPE_FORMAT_S8_Z24 || @@ -216,14 +220,14 @@ sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y) if (tile->x != -1) { /* put dirty tile back in framebuffer */ if (is_depth_stencil) { - ps->put_tile_raw(ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32); + pipe->put_tile(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0 /*STRIDE*/); } else { - ps->put_tile(ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + pipe->put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); } } @@ -289,14 +293,14 @@ sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y) else { /* get new tile from framebuffer */ if (is_depth_stencil) { - ps->get_tile_raw(ps, - tile_x, tile_y, TILE_SIZE, TILE_SIZE, - tile->data.depth32); + pipe->get_tile(pipe, ps, + tile_x, tile_y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); } else { - ps->get_tile(ps, - tile_x, tile_y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + pipe->get_tile_rgba(pipe, ps, + tile_x, tile_y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); } } @@ -349,9 +353,9 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, struct pipe_surface *ps = pipe->get_tex_surface(pipe, tc->texture, face, level, z); - ps->get_tile(ps, - tile_x, tile_y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + pipe->get_tile_rgba(pipe, ps, + tile_x, tile_y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); pipe_surface_reference(&ps, NULL); diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index 15245a2efb..e122d70d13 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -74,13 +74,15 @@ sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, struct pipe_mipmap_tree *texture); extern void -sp_flush_tile_cache(struct softpipe_tile_cache *tc); +sp_flush_tile_cache(struct softpipe_context *softpipe, + struct softpipe_tile_cache *tc); extern void sp_tile_cache_clear(struct softpipe_tile_cache *tc, const float value[4]); extern struct softpipe_cached_tile * -sp_get_cached_tile(struct softpipe_tile_cache *tc, int x, int y); +sp_get_cached_tile(struct softpipe_context *softpipe, + struct softpipe_tile_cache *tc, int x, int y); extern const struct softpipe_cached_tile * sp_get_cached_tile_tex(struct pipe_context *pipe, diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 7c154baa15..7a245b0ed6 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -79,7 +79,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) accBuf[i * 4 + 3] = a; } - acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); + pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(accBuf); @@ -101,13 +101,13 @@ accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, (void) pipe->region_map(pipe, acc_ps->region); - acc_ps->get_tile(acc_ps, xpos, ypos, width, height, accBuf); + pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] * scale + bias; } - acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); + pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(accBuf); @@ -131,14 +131,14 @@ accum_accum(struct pipe_context *pipe, GLfloat value, colorMap = pipe->region_map(pipe, color_ps->region); accMap = pipe->region_map(pipe, acc_ps->region); - color_ps->get_tile(color_ps, xpos, ypos, width, height, colorBuf); - acc_ps->get_tile(acc_ps, xpos, ypos, width, height, accBuf); + pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, colorBuf); + pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] + colorBuf[i] * value; } - acc_ps->put_tile(acc_ps, xpos, ypos, width, height, accBuf); + pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(colorBuf); free(accBuf); @@ -162,13 +162,13 @@ accum_load(struct pipe_context *pipe, GLfloat value, (void) pipe->region_map(pipe, color_ps->region); (void) pipe->region_map(pipe, acc_ps->region); - color_ps->get_tile(color_ps, xpos, ypos, width, height, buf); + pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, buf); for (i = 0; i < 4 * width * height; i++) { buf[i] = buf[i] * value; } - acc_ps->put_tile(acc_ps, xpos, ypos, width, height, buf); + pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, buf); free(buf); @@ -193,11 +193,11 @@ accum_return(GLcontext *ctx, GLfloat value, (void) pipe->region_map(pipe, color_ps->region); (void) pipe->region_map(pipe, acc_ps->region); - acc_ps->get_tile(acc_ps, xpos, ypos, width, height, abuf); + pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) { cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - color_ps->get_tile(color_ps, xpos, ypos, width, height, cbuf); + pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, cbuf); } for (i = 0; i < width * height; i++) { @@ -212,7 +212,7 @@ accum_return(GLcontext *ctx, GLfloat value, } } - color_ps->put_tile(color_ps, xpos, ypos, width, height, abuf); + pipe->put_tile_rgba(pipe, color_ps, xpos, ypos, width, height, abuf); free(abuf); if (cbuf) diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 4fdf1cef2b..c4a954c43d 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1245,8 +1245,8 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, (void) pipe->region_map(pipe, psRead->region); (void) pipe->region_map(pipe, psTex->region); - psRead->get_tile(psRead, srcx, srcy, width, height, buf); - psTex->put_tile(psTex, 0, 0, width, height, buf); + pipe->get_tile_rgba(pipe, psRead, srcx, srcy, width, height, buf); + pipe->put_tile_rgba(pipe, psTex, 0, 0, width, height, buf); pipe->region_unmap(pipe, psRead->region); pipe->region_unmap(pipe, psTex->region); diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 2c6847e372..c347a0b688 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -204,7 +204,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, /* Do a row at a time to flip image data vertically */ for (i = 0; i < height; i++) { - strb->surface->get_tile(strb->surface, x, y, width, 1, df); + pipe->get_tile_rgba(pipe, strb->surface, x, y, width, 1, df); y += yStep; df += dfStride; if (!dfStride) { diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 1621b56336..e1cf4861d0 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1100,7 +1100,7 @@ fallback_copy_texsubimage(GLcontext *ctx, /* do copy row by row */ for (row = 0; row < height; row++) { - src_surf->get_tile(src_surf, srcX, srcY + row, width, 1, data); + pipe->get_tile_rgba(pipe, src_surf, srcX, srcY + row, width, 1, data); /* XXX we're ignoring convolution for now */ if (ctx->_ImageTransferState) { @@ -1109,7 +1109,7 @@ fallback_copy_texsubimage(GLcontext *ctx, width, (GLfloat (*)[4])data); } - dest_surf->put_tile(dest_surf, destX, destY, width, 1, data); + pipe->put_tile_rgba(pipe, dest_surf, destX, destY, width, 1, data); destY += yStep; } -- cgit v1.2.3 From 616112ea2e0eefea356be228bff8754ee955d8b3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 25 Oct 2007 19:19:51 -0600 Subject: silence warning --- src/mesa/state_tracker/st_cb_clear.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 2e7d9f1a30..30672e0dd4 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -464,7 +464,9 @@ static void clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { struct st_renderbuffer *strb = st_renderbuffer(rb); + /* const GLboolean isDS = is_depth_stencil_format(strb->surface->format); + */ assert(strb->surface->format); -- cgit v1.2.3 From f684120417c6b3ca9e7486ffeb24fe88e428834d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 25 Oct 2007 19:27:29 -0600 Subject: Move region_alloc() and region_release() to pipe_winsys. This allows regions to be allocated w/out a rendering context. --- .../drivers/dri/intel_winsys/intel_winsys_i915.c | 61 ++++++++++++++++ src/mesa/drivers/x11/xm_buffer.c | 4 +- src/mesa/drivers/x11/xm_surface.c | 4 +- src/mesa/drivers/x11/xm_winsys.c | 83 ++++++++++++++++------ src/mesa/pipe/failover/fo_context.c | 2 - src/mesa/pipe/i915simple/i915_regions.c | 61 ---------------- src/mesa/pipe/p_context.h | 13 ++-- src/mesa/pipe/p_winsys.h | 14 ++++ src/mesa/pipe/softpipe/sp_region.c | 57 --------------- src/mesa/state_tracker/st_cb_drawpixels.c | 9 ++- src/mesa/state_tracker/st_cb_fbo.c | 8 ++- src/mesa/state_tracker/st_mipmap_tree.c | 7 +- 12 files changed, 163 insertions(+), 160 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index f6eb050152..c481495309 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -120,6 +120,64 @@ static void intel_i915_batch_flush( struct i915_winsys *sws ) +static struct pipe_region * +intel_i915_region_alloc(struct pipe_winsys *winsys, + unsigned cpp, unsigned width, + unsigned height, unsigned flags) +{ + struct pipe_region *region = calloc(sizeof(*region), 1); + const unsigned alignment = 64; + + /* Choose a pitch to match hardware requirements - requires 64 byte + * alignment of render targets. + * + * XXX: is this ok for textures?? + * clearly want to be able to render to textures under some + * circumstances, but maybe not always a requirement. + */ + unsigned pitch; + + /* XXX is the pitch different for textures vs. drawables? */ + if (flags & PIPE_SURFACE_FLAG_TEXTURE) /* or PIPE_SURFACE_FLAG_RENDER? */ + pitch = ((cpp * width + 63) & ~63) / cpp; + else + pitch = ((cpp * width + 63) & ~63) / cpp; + + region->cpp = cpp; + region->pitch = pitch; + region->height = height; /* needed? */ + region->refcount = 1; + + region->buffer = winsys->buffer_create( winsys, alignment ); + + winsys->buffer_data( winsys, + region->buffer, + pitch * cpp * height, + NULL ); + + return region; +} + +static void +intel_i915_region_release(struct pipe_winsys *winsys, + struct pipe_region **region) +{ + if (!*region) + return; + + assert((*region)->refcount > 0); + (*region)->refcount--; + + if ((*region)->refcount == 0) { + assert((*region)->map_refcount == 0); + + winsys->buffer_reference( winsys, + &((*region)->buffer), NULL ); + free(*region); + } + *region = NULL; +} + struct pipe_context * intel_create_i915simple( struct intel_context *intel ) @@ -135,6 +193,9 @@ intel_create_i915simple( struct intel_context *intel ) iws->winsys.batch_flush = intel_i915_batch_flush; iws->intel = intel; + iws->winsys.region_alloc = intel_i915_region_alloc; + iws->winsys.region_release = intel_i915_region_release; + /* Create the i915simple context: */ return i915_create( intel_create_pipe_winsys(intel), diff --git a/src/mesa/drivers/x11/xm_buffer.c b/src/mesa/drivers/x11/xm_buffer.c index b8d3df1379..09356c7329 100644 --- a/src/mesa/drivers/x11/xm_buffer.c +++ b/src/mesa/drivers/x11/xm_buffer.c @@ -37,6 +37,7 @@ #include "renderbuffer.h" #include "pipe/p_state.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #include "state_tracker/st_context.h" @@ -254,7 +255,8 @@ finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) struct pipe_context *pipe = ctx->st->pipe; if (!xrb->St.surface->region) { int w = 1, h = 1; - xrb->St.surface->region = pipe->region_alloc(pipe, 1, w, h, 0x0); + xrb->St.surface->region = pipe->winsys->region_alloc(pipe->winsys, + 1, w, h, 0x0); } } diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 340f796bc8..337033f8ad 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -44,6 +44,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #include "pipe/softpipe/sp_context.h" #include "pipe/softpipe/sp_clear.h" #include "pipe/softpipe/sp_tile_cache.h" @@ -163,7 +164,8 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) * functions. */ if (pipe) - xms->surface.surface.region = pipe->region_alloc(pipe, 1, 0, 0, 0x0); + xms->surface.surface.region = pipe->winsys->region_alloc(pipe->winsys, + 1, 0, 0, 0x0); return &xms->surface.surface; } diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index 7e29580b50..624d28dd01 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -233,6 +233,63 @@ xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) } + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + + +static struct pipe_region * +xm_region_alloc(struct pipe_winsys *winsys, + unsigned cpp, unsigned width, unsigned height, unsigned flags) +{ + struct pipe_region *region = CALLOC_STRUCT(pipe_region); + const unsigned alignment = 64; + + region->cpp = cpp; + region->pitch = round_up(width, alignment / cpp); + region->height = height; + region->refcount = 1; + + assert(region->pitch > 0); + + region->buffer = winsys->buffer_create( winsys, alignment ) +; + + /* NULL data --> just allocate the space */ + winsys->buffer_data( winsys, + region->buffer, + region->pitch * cpp * height, + NULL ); + return region; +} + + +static void +xm_region_release(struct pipe_winsys *winsys, struct pipe_region **region) +{ + if (!*region) + return; + + assert((*region)->refcount > 0); + (*region)->refcount--; + + if ((*region)->refcount == 0) { + assert((*region)->map_refcount == 0); + + winsys->buffer_reference( winsys, &((*region)->buffer), NULL ); + free(*region); + } + *region = NULL; +} + + + struct xmesa_pipe_winsys { struct pipe_winsys winsys; @@ -259,6 +316,10 @@ xmesa_create_pipe_winsys( XMesaContext xmesa ) xws->winsys.buffer_data = xm_buffer_data; xws->winsys.buffer_subdata = xm_buffer_subdata; xws->winsys.buffer_get_subdata = xm_buffer_get_subdata; + + xws->winsys.region_alloc = xm_region_alloc; + xws->winsys.region_release = xm_region_release; + xws->winsys.flush_frontbuffer = xm_flush_frontbuffer; xws->winsys.wait_idle = xm_wait_idle; xws->winsys.printf = xm_printf; @@ -272,27 +333,9 @@ xmesa_create_pipe_winsys( XMesaContext xmesa ) struct pipe_context * xmesa_create_softpipe(XMesaContext xmesa) { - struct xm_winsys *isws = CALLOC_STRUCT( xm_winsys ); + struct xm_winsys *xm_ws = CALLOC_STRUCT( xm_winsys ); - /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. - * - * Softpipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate xm_buffer. - */ -#if 0 - isws->sws.create_buffer = xm_create_buffer; - isws->sws.buffer_map = xm_buffer_map; - isws->sws.buffer_unmap = xm_buffer_unmap; - isws->sws.buffer_reference = xm_buffer_reference; - isws->sws.buffer_unreference = xm_buffer_unreference; - isws->sws.buffer_data = xm_buffer_data; - isws->sws.buffer_subdata = xm_buffer_subdata; - isws->sws.buffer_get_subdata = xm_buffer_get_subdata; -#endif - /* Create the softpipe context: */ - return softpipe_create( xmesa_create_pipe_winsys(xmesa), &isws->sws ); + return softpipe_create( xmesa_create_pipe_winsys(xmesa), &xm_ws->sws ); } diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index 076d516583..a3a0296598 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -137,8 +137,6 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.surface_alloc = hw->surface_alloc; failover->pipe.get_tex_surface = hw->get_tex_surface; - failover->pipe.region_alloc = hw->region_alloc; - failover->pipe.region_release = hw->region_release; failover->pipe.region_map = hw->region_map; failover->pipe.region_unmap = hw->region_unmap; failover->pipe.region_data = hw->region_data; diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index 577a6adfd8..0410446326 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -69,65 +69,6 @@ i915_region_unmap(struct pipe_context *pipe, struct pipe_region *region) } } -static struct pipe_region * -i915_region_alloc(struct pipe_context *pipe, - unsigned cpp, unsigned width, unsigned height, unsigned flags) -{ - struct i915_context *i915 = i915_context( pipe ); - struct pipe_region *region = calloc(sizeof(*region), 1); - const unsigned alignment = 64; - - /* Choose a pitch to match hardware requirements - requires 64 byte - * alignment of render targets. - * - * XXX: is this ok for textures?? - * clearly want to be able to render to textures under some - * circumstances, but maybe not always a requirement. - */ - unsigned pitch; - - /* XXX is the pitch different for textures vs. drawables? */ - if (flags & PIPE_SURFACE_FLAG_TEXTURE) /* or PIPE_SURFACE_FLAG_RENDER? */ - pitch = ((cpp * width + 63) & ~63) / cpp; - else - pitch = ((cpp * width + 63) & ~63) / cpp; - - region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ - region->refcount = 1; - - region->buffer = i915->pipe.winsys->buffer_create( i915->pipe.winsys, alignment ); - - i915->pipe.winsys->buffer_data( i915->pipe.winsys, - region->buffer, - pitch * cpp * height, - NULL ); - - return region; -} - -static void -i915_region_release(struct pipe_context *pipe, struct pipe_region **region) -{ - struct i915_context *i915 = i915_context( pipe ); - - if (!*region) - return; - - assert((*region)->refcount > 0); - (*region)->refcount--; - - if ((*region)->refcount == 0) { - assert((*region)->map_refcount == 0); - - i915->pipe.winsys->buffer_reference( i915->pipe.winsys, - &((*region)->buffer), NULL ); - free(*region); - } - *region = NULL; -} - /* * XXX Move this into core Mesa? @@ -302,8 +243,6 @@ i915_init_region_functions(struct i915_context *i915) { i915->pipe.region_map = i915_region_map; i915->pipe.region_unmap = i915_region_unmap; - i915->pipe.region_alloc = i915_region_alloc; - i915->pipe.region_release = i915_region_release; i915->pipe.region_data = i915_region_data; i915->pipe.region_copy = i915_region_copy; i915->pipe.region_fill = i915_region_fill; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index a22ea3a8a0..b9de3667e5 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -208,14 +208,7 @@ struct pipe_context { /* * Memory region functions - * Some of these may go away... */ - struct pipe_region *(*region_alloc)(struct pipe_context *pipe, - unsigned cpp, unsigned width, unsigned height, - unsigned flags); - - void (*region_release)(struct pipe_context *pipe, struct pipe_region **r); - ubyte *(*region_map)(struct pipe_context *pipe, struct pipe_region *r); void (*region_unmap)(struct pipe_context *pipe, struct pipe_region *r); @@ -225,7 +218,8 @@ struct pipe_context { unsigned dest_offset, unsigned destx, unsigned desty, const void *src, unsigned src_stride, - unsigned srcx, unsigned srcy, unsigned width, unsigned height); + unsigned srcx, unsigned srcy, + unsigned width, unsigned height); void (*region_copy)(struct pipe_context *pipe, struct pipe_region *dest, @@ -234,7 +228,8 @@ struct pipe_context { struct pipe_region *src, /* don't make this const - need to map/unmap */ unsigned src_offset, - unsigned srcx, unsigned srcy, unsigned width, unsigned height); + unsigned srcx, unsigned srcy, + unsigned width, unsigned height); void (*region_fill)(struct pipe_context *pipe, struct pipe_region *dst, diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 7e5d394a90..3b04c44733 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -50,6 +50,13 @@ struct pipe_buffer_handle; * driver and the hardware driver about the format of command buffers, * etc. */ + + +struct pipe_region; + +/** Opaque type */ +struct pipe_buffer_handle; + struct pipe_winsys { /** @@ -63,6 +70,13 @@ struct pipe_winsys const char *, ... ); + struct pipe_region *(*region_alloc)(struct pipe_winsys *ws, + unsigned cpp, unsigned width, + unsigned height, unsigned flags); + + void (*region_release)(struct pipe_winsys *ws, struct pipe_region **r); + + /** * The buffer manager is modeled after the dri_bufmgr interface, which * in turn is modeled after the ARB_vertex_buffer_object extension, diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 4317a9ea1b..80a67dcabc 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -38,16 +38,6 @@ #include "pipe/p_defines.h" -/** - * Round n up to next multiple. - */ -static INLINE unsigned -round_up(unsigned n, unsigned multiple) -{ - return (n + multiple - 1) & ~(multiple - 1); -} - - static ubyte * sp_region_map(struct pipe_context *pipe, struct pipe_region *region) @@ -79,51 +69,6 @@ sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) } } -static struct pipe_region * -sp_region_alloc(struct pipe_context *pipe, - unsigned cpp, unsigned width, unsigned height, unsigned flags) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - struct pipe_region *region = CALLOC_STRUCT(pipe_region); - const unsigned alignment = 64; - - region->cpp = cpp; - region->pitch = round_up(width, alignment / cpp); - region->height = height; - region->refcount = 1; - - assert(region->pitch > 0); - - region->buffer = sp->pipe.winsys->buffer_create( sp->pipe.winsys, alignment ); - - /* NULL data --> just allocate the space */ - sp->pipe.winsys->buffer_data( sp->pipe.winsys, - region->buffer, - region->pitch * cpp * height, - NULL ); - return region; -} - -static void -sp_region_release(struct pipe_context *pipe, struct pipe_region **region) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - - if (!*region) - return; - - assert((*region)->refcount > 0); - (*region)->refcount--; - - if ((*region)->refcount == 0) { - assert((*region)->map_refcount == 0); - - sp->pipe.winsys->buffer_reference( sp->pipe.winsys, - &((*region)->buffer), NULL ); - free(*region); - } - *region = NULL; -} /** @@ -313,8 +258,6 @@ sp_init_region_functions(struct softpipe_context *sp) { sp->pipe.region_map = sp_region_map; sp->pipe.region_unmap = sp_region_unmap; - sp->pipe.region_alloc = sp_region_alloc; - sp->pipe.region_release = sp_region_release; sp->pipe.region_data = sp_region_data; sp->pipe.region_copy = sp_region_copy; sp->pipe.region_fill = sp_region_fill; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c4a954c43d..5d8890e022 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -47,6 +47,7 @@ #include "st_format.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "shader/prog_instruction.h" @@ -346,7 +347,8 @@ alloc_mipmap_tree(struct st_context *st, cpp = st_sizeof_format(pipeFormat); /* allocate texture region/storage */ - mt->region = st->pipe->region_alloc(st->pipe, cpp, width, height, flags); + mt->region = st->pipe->winsys->region_alloc(st->pipe->winsys, + cpp, width, height, flags); mt->target = PIPE_TEXTURE_2D; mt->internal_format = GL_RGBA; @@ -468,7 +470,7 @@ make_mipmap_tree(struct st_context *st, static void free_mipmap_tree(struct pipe_context *pipe, struct pipe_mipmap_tree *mt) { - pipe->region_release(pipe, &mt->region); + pipe->winsys->region_release(pipe->winsys, &mt->region); free(mt); } @@ -977,7 +979,8 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, /* allocate texture region/storage */ - mt->region = pipe->region_alloc(pipe, cpp, width, height, flags); + mt->region = pipe->winsys->region_alloc(pipe->winsys, + cpp, width, height, flags); pitch = mt->region->pitch; /* map texture region */ diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 94e286feab..f22132b3cf 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -41,6 +41,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #include "st_context.h" #include "st_cb_fbo.h" #include "st_cb_texture.h" @@ -95,10 +96,11 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, struct pipe_region *r = strb->surface->region; while (r->map) pipe->region_unmap(pipe, r); - pipe->region_release(pipe, &strb->surface->region); + pipe->winsys->region_release(pipe->winsys, &strb->surface->region); } - strb->surface->region = pipe->region_alloc(pipe, cpp, width, height, flags); + strb->surface->region = pipe->winsys->region_alloc(pipe->winsys, cpp, + width, height, flags); if (!strb->surface->region) return GL_FALSE; /* out of memory, try s/w buffer? */ @@ -125,7 +127,7 @@ st_renderbuffer_delete(struct gl_renderbuffer *rb) ASSERT(strb); if (strb && strb->surface) { if (strb->surface->region) { - pipe->region_release(pipe, &strb->surface->region); + pipe->winsys->region_release(pipe->winsys, &strb->surface->region); } free(strb->surface); } diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index faca148d80..d1db590bee 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -31,6 +31,7 @@ #include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_winsys.h" #define DBG if(0) printf @@ -87,8 +88,8 @@ st_miptree_create(struct pipe_context *pipe, ok = pipe->mipmap_tree_layout(pipe, mt); if (ok) { /* note: it's OK to pass 'pitch' as 'width' here: */ - mt->region = pipe->region_alloc(pipe, mt->cpp, mt->pitch, - mt->total_height, flags); + mt->region = pipe->winsys->region_alloc(pipe->winsys, mt->cpp, mt->pitch, + mt->total_height, flags); mt->pitch = mt->region->pitch; /*XXX NEW */ } @@ -124,7 +125,7 @@ st_miptree_release(struct pipe_context *pipe, DBG("%s deleting %p\n", __FUNCTION__, (void *) *mt); - pipe->region_release(pipe, &((*mt)->region)); + pipe->winsys->region_release(pipe->winsys, &((*mt)->region)); for (i = 0; i < MAX_TEXTURE_LEVELS; i++) if ((*mt)->level[i].image_offset) -- cgit v1.2.3 From ec854674577dc8162fd336e2a5369ec274271929 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 25 Oct 2007 20:31:13 -0600 Subject: clean-up, simplify some tile code, more to come --- src/mesa/pipe/softpipe/sp_surface.c | 247 +++++++++--------------------------- src/mesa/pipe/softpipe/sp_surface.h | 5 +- 2 files changed, 63 insertions(+), 189 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index b8af1c561f..4accafa384 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -459,169 +459,6 @@ s8z24_get_tile(struct pipe_surface *ps, } -/** - * Get raw tile, any 32-bit pixel format. - */ -static void -get_tile_raw32(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, void *p) -{ - const uint *src - = ((const uint *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; - uint *pDest = (uint *) p; - unsigned i; - unsigned w0 = w; - - assert(ps->region->map); - assert(ps->format == PIPE_FORMAT_S8_Z24 || - ps->format == PIPE_FORMAT_U_Z32); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - memcpy(pDest, src, w * sizeof(uint)); - src += ps->region->pitch; - pDest += w0; - } -} - - -/** - * put raw tile, any 32-bit pixel format. - */ -static void -put_tile_raw32(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const void *p) -{ - uint *dst - = ((uint *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; - const uint *pSrc = (const uint *) p; - unsigned i; - unsigned w0 = w; - - assert(w < 1000); - assert(ps->region->map); - assert(ps->format == PIPE_FORMAT_S8_Z24 || - ps->format == PIPE_FORMAT_U_Z32); - - assert(w < 1000); - CLIP_TILE; - - for (i = 0; i < h; i++) { - assert(w < 1000); - memcpy(dst, pSrc, w * sizeof(uint)); - dst += ps->region->pitch; - pSrc += w0; - } -} - - -/** - * Get raw tile, any 16-bit pixel format. - */ -static void -get_tile_raw16(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, void *p) -{ - const ushort *src - = ((const ushort *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; - ushort *pDest = (ushort *) p; - uint i; - uint w0 = w; - - assert(ps->format == PIPE_FORMAT_U_Z16); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - memcpy(pDest, src, w * sizeof(ushort)); - src += ps->region->pitch; - pDest += w0; - } -} - -/** - * Put raw tile, any 16-bit pixel format. - */ -static void -put_tile_raw16(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const void *p) -{ - ushort *dst - = ((ushort *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; - const ushort *pSrc = (const ushort *) p; - unsigned i; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_Z16); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - memcpy(dst, pSrc, w * sizeof(ushort)); - dst += ps->region->pitch; - pSrc += w0; - } -} - - -/** - * Get raw tile, any 16-bit pixel format. - */ -static void -get_tile_raw8(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, void *p) -{ - const ubyte *src - = ((const ubyte *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; - ubyte *pDest = (ubyte *) p; - uint i; - uint w0 = w; - - assert(ps->format == PIPE_FORMAT_U_Z16); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - memcpy(pDest, src, w * sizeof(ubyte)); - src += ps->region->pitch; - pDest += w0; - } -} - -/** - * Put raw tile, any 16-bit pixel format. - */ -static void -put_tile_raw8(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const void *p) -{ - ubyte *dst - = ((ubyte *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; - const ubyte *pSrc = (const ubyte *) p; - unsigned i; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_Z16); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - memcpy(dst, pSrc, w * sizeof(ubyte)); - dst += ps->region->pitch; - pSrc += w0; - } -} - - - - /** * Initialize the quad_read/write and get/put_tile() methods. */ @@ -658,22 +495,14 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) case PIPE_FORMAT_U_Z16: sps->get_tile = z16_get_tile; - sps->get_tile_raw = get_tile_raw16; - sps->put_tile_raw = put_tile_raw16; break; case PIPE_FORMAT_U_Z32: sps->get_tile = z32_get_tile; - sps->get_tile_raw = get_tile_raw32; - sps->put_tile_raw = put_tile_raw32; break; case PIPE_FORMAT_S8_Z24: sps->get_tile = s8z24_get_tile; - sps->get_tile_raw = get_tile_raw32; - sps->put_tile_raw = put_tile_raw32; break; case PIPE_FORMAT_U_S8: - sps->get_tile_raw = get_tile_raw8; - sps->put_tile_raw = put_tile_raw8; break; default: assert(0); @@ -739,28 +568,75 @@ softpipe_get_tex_surface(struct pipe_context *pipe, } +/** + * Move raw block of pixels from surface to user memory. + */ static void -get_tile_generic(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride) +softpipe_get_tile(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride) { - struct softpipe_surface *sps = softpipe_surface(ps); - sps->get_tile_raw(ps, x, y, w, h, p); + const uint cpp = ps->region->cpp; + const uint w0 = w; + const ubyte *pSrc; + ubyte *pDest; + uint i; + + assert(ps->region->map); + + CLIP_TILE; + + if (dst_stride == 0) { + dst_stride = w0 * cpp; + } + + pSrc = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; + pDest = (ubyte *) p; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w0 * cpp); + pDest += dst_stride; + pSrc += ps->region->pitch * cpp; + } } +/** + * Move raw block of pixels from user memory to surface. + */ static void -put_tile_generic(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride) +softpipe_put_tile(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride) { - struct softpipe_surface *sps = softpipe_surface(ps); - sps->put_tile_raw(ps, x, y, w, h, p); + const uint cpp = ps->region->cpp; + const uint w0 = w; + const ubyte *pSrc; + ubyte *pDest; + uint i; + + assert(ps->region->map); + + CLIP_TILE; + + if (src_stride == 0) { + src_stride = w0 * cpp; + } + + pSrc = (const ubyte *) p; + pDest = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w0 * cpp); + pDest += ps->region->pitch * cpp; + pSrc += src_stride; + } } +/* XXX TEMPORARY */ static void get_tile_rgba_generic(struct pipe_context *pipe, struct pipe_surface *ps, @@ -772,6 +648,7 @@ get_tile_rgba_generic(struct pipe_context *pipe, } +/* XXX TEMPORARY */ static void put_tile_rgba_generic(struct pipe_context *pipe, struct pipe_surface *ps, @@ -789,8 +666,8 @@ sp_init_surface_functions(struct softpipe_context *sp) { sp->pipe.surface_alloc = softpipe_surface_alloc; - sp->pipe.get_tile = get_tile_generic; - sp->pipe.put_tile = put_tile_generic; + sp->pipe.get_tile = softpipe_get_tile; + sp->pipe.put_tile = softpipe_put_tile; sp->pipe.get_tile_rgba = get_tile_rgba_generic; sp->pipe.put_tile_rgba = put_tile_rgba_generic; diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 0c2486a171..af6533d4f0 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -46,14 +46,11 @@ struct softpipe_tile_cache; struct softpipe_surface { struct pipe_surface surface; + /* XXX these are temporary here */ void (*get_tile)(struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p); void (*put_tile)(struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p); - void (*get_tile_raw)(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, void *p); - void (*put_tile_raw)(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const void *p); }; -- cgit v1.2.3 From f7be1b419aab80c4e011183611964eb4d7c023c2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 26 Oct 2007 13:31:04 +0100 Subject: Actually move region_alloc() and region_release() to intel_winsys. --- .../drivers/dri/intel_winsys/intel_winsys_i915.c | 62 --------------------- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 64 ++++++++++++++++++++++ 2 files changed, 64 insertions(+), 62 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index c481495309..e07fadb278 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -120,65 +120,6 @@ static void intel_i915_batch_flush( struct i915_winsys *sws ) -static struct pipe_region * -intel_i915_region_alloc(struct pipe_winsys *winsys, - unsigned cpp, unsigned width, - unsigned height, unsigned flags) -{ - struct pipe_region *region = calloc(sizeof(*region), 1); - const unsigned alignment = 64; - - /* Choose a pitch to match hardware requirements - requires 64 byte - * alignment of render targets. - * - * XXX: is this ok for textures?? - * clearly want to be able to render to textures under some - * circumstances, but maybe not always a requirement. - */ - unsigned pitch; - - /* XXX is the pitch different for textures vs. drawables? */ - if (flags & PIPE_SURFACE_FLAG_TEXTURE) /* or PIPE_SURFACE_FLAG_RENDER? */ - pitch = ((cpp * width + 63) & ~63) / cpp; - else - pitch = ((cpp * width + 63) & ~63) / cpp; - - region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ - region->refcount = 1; - - region->buffer = winsys->buffer_create( winsys, alignment ); - - winsys->buffer_data( winsys, - region->buffer, - pitch * cpp * height, - NULL ); - - return region; -} - -static void -intel_i915_region_release(struct pipe_winsys *winsys, - struct pipe_region **region) -{ - if (!*region) - return; - - assert((*region)->refcount > 0); - (*region)->refcount--; - - if ((*region)->refcount == 0) { - assert((*region)->map_refcount == 0); - - winsys->buffer_reference( winsys, - &((*region)->buffer), NULL ); - free(*region); - } - *region = NULL; -} - - struct pipe_context * intel_create_i915simple( struct intel_context *intel ) { @@ -193,9 +134,6 @@ intel_create_i915simple( struct intel_context *intel ) iws->winsys.batch_flush = intel_i915_batch_flush; iws->intel = intel; - iws->winsys.region_alloc = intel_i915_region_alloc; - iws->winsys.region_release = intel_i915_region_release; - /* Create the i915simple context: */ return i915_create( intel_create_pipe_winsys(intel), diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 72fb24acbf..6f244e1100 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -41,6 +41,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_defines.h" +#include "pipe/p_state.h" @@ -192,6 +193,66 @@ intel_flush_frontbuffer( struct pipe_winsys *sws ) intelCopyBuffer(dPriv, NULL); } + +static struct pipe_region * +intel_i915_region_alloc(struct pipe_winsys *winsys, + unsigned cpp, unsigned width, + unsigned height, unsigned flags) +{ + struct pipe_region *region = calloc(sizeof(*region), 1); + const unsigned alignment = 64; + + /* Choose a pitch to match hardware requirements - requires 64 byte + * alignment of render targets. + * + * XXX: is this ok for textures?? + * clearly want to be able to render to textures under some + * circumstances, but maybe not always a requirement. + */ + unsigned pitch; + + /* XXX is the pitch different for textures vs. drawables? */ + if (flags & PIPE_SURFACE_FLAG_TEXTURE) /* or PIPE_SURFACE_FLAG_RENDER? */ + pitch = ((cpp * width + 63) & ~63) / cpp; + else + pitch = ((cpp * width + 63) & ~63) / cpp; + + region->cpp = cpp; + region->pitch = pitch; + region->height = height; /* needed? */ + region->refcount = 1; + + region->buffer = winsys->buffer_create( winsys, alignment ); + + winsys->buffer_data( winsys, + region->buffer, + pitch * cpp * height, + NULL ); + + return region; +} + +static void +intel_i915_region_release(struct pipe_winsys *winsys, + struct pipe_region **region) +{ + if (!*region) + return; + + assert((*region)->refcount > 0); + (*region)->refcount--; + + if ((*region)->refcount == 0) { + assert((*region)->map_refcount == 0); + + winsys->buffer_reference( winsys, + &((*region)->buffer), NULL ); + free(*region); + } + *region = NULL; +} + + static void intel_printf( struct pipe_winsys *sws, const char *fmtString, ... ) { @@ -234,5 +295,8 @@ intel_create_pipe_winsys( struct intel_context *intel ) iws->winsys.get_name = intel_get_name; iws->intel = intel; + iws->winsys.region_alloc = intel_i915_region_alloc; + iws->winsys.region_release = intel_i915_region_release; + return &iws->winsys; } -- cgit v1.2.3 From e4f6f0ec02133e9297c3f2db787dee14bf0ae6e1 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 26 Oct 2007 10:45:42 -0600 Subject: surface_alloc() is now a winsys function. This allows surfaces to be allocated without a rendering context. A few loose ends to resolve, but in working condition. --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 14 ++ src/mesa/drivers/x11/xm_api.c | 7 + src/mesa/drivers/x11/xm_surface.c | 123 +++++++++----- src/mesa/drivers/x11/xm_winsys.c | 27 +++ src/mesa/drivers/x11/xmesaP.h | 13 +- src/mesa/pipe/failover/fo_context.c | 2 + src/mesa/pipe/i915simple/i915_surface.c | 189 ++++++++++++++++----- src/mesa/pipe/p_context.h | 9 +- src/mesa/pipe/p_winsys.h | 4 + src/mesa/pipe/softpipe/sp_surface.c | 107 +++++++++--- src/mesa/pipe/softpipe/sp_surface.h | 15 +- src/mesa/state_tracker/st_cb_fbo.c | 2 +- 12 files changed, 393 insertions(+), 119 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 6f244e1100..4569b1e3bf 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -253,6 +253,18 @@ intel_i915_region_release(struct pipe_winsys *winsys, } +static struct pipe_surface * +intel_i915_surface_alloc(struct pipe_winsys *winsys, unsigned format) +{ + struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); + if (surf) { + surf->format = format; + surf->refcount = 1; + } + return surf; +} + + static void intel_printf( struct pipe_winsys *sws, const char *fmtString, ... ) { @@ -298,5 +310,7 @@ intel_create_pipe_winsys( struct intel_context *intel ) iws->winsys.region_alloc = intel_i915_region_alloc; iws->winsys.region_release = intel_i915_region_release; + iws->winsys.surface_alloc = intel_i915_surface_alloc; + return &iws->winsys; } diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 187663e66c..1d2b93d100 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -1610,8 +1610,15 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) st_create_context( mesaCtx, xmesa_create_softpipe( c ) ); + /* override these functions, as if the xlib driver were derived from + * the softpipe driver. + */ +#if 0 mesaCtx->st->pipe->surface_alloc = xmesa_surface_alloc; +#endif mesaCtx->st->pipe->supported_formats = xmesa_supported_formats; + mesaCtx->st->pipe->get_tile_rgba = xmesa_get_tile_rgba; + mesaCtx->st->pipe->put_tile_rgba = xmesa_put_tile_rgba; mesaCtx->st->haveFramebufferRegions = GL_FALSE; diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 337033f8ad..3655a55e4e 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -48,6 +48,7 @@ #include "pipe/softpipe/sp_context.h" #include "pipe/softpipe/sp_clear.h" #include "pipe/softpipe/sp_tile_cache.h" +#include "pipe/softpipe/sp_surface.h" #include "state_tracker/st_context.h" @@ -67,6 +68,13 @@ xmesa_surf(struct softpipe_surface *sps) } +static INLINE struct xmesa_surface * +xmesa_surface(struct pipe_surface *ps) +{ + return (struct xmesa_surface *) ps; +} + + static INLINE struct xmesa_renderbuffer * xmesa_rb(struct softpipe_surface *sps) { @@ -78,53 +86,71 @@ xmesa_rb(struct softpipe_surface *sps) #define FLIP(Y) Y = xrb->St.Base.Height - (Y) - 1; -static void -get_tile(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *p) +void +xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, float *p) { - struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); - GLubyte tmp[MAX_WIDTH * 4]; - GLuint i, j; - uint w0 = w; - GET_CURRENT_CONTEXT(ctx); - - CLIP_TILE; - - FLIP(y); - for (i = 0; i < h; i++) { - xrb->St.Base.GetRow(ctx, &xrb->St.Base, w, x, y - i, tmp); - for (j = 0; j < w * 4; j++) { - p[j] = UBYTE_TO_FLOAT(tmp[j]); + struct xmesa_surface *xms = xmesa_surface(ps); + struct xmesa_renderbuffer *xrb = xms->xrb; + + if (xrb) { + /* this is a front/back color buffer */ + GLubyte tmp[MAX_WIDTH * 4]; + GLuint i, j; + uint w0 = w; + GET_CURRENT_CONTEXT(ctx); + + CLIP_TILE; + + FLIP(y); + for (i = 0; i < h; i++) { + xrb->St.Base.GetRow(ctx, &xrb->St.Base, w, x, y - i, tmp); + for (j = 0; j < w * 4; j++) { + p[j] = UBYTE_TO_FLOAT(tmp[j]); + } + p += w0 * 4; } - p += w0 * 4; + } + else { + /* other softpipe surface */ + softpipe_get_tile_rgba(pipe, ps, x, y, w, h, p); } } -static void -put_tile(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *p) +void +xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const float *p) { - struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); - GLubyte tmp[MAX_WIDTH * 4]; - GLuint i, j; - uint w0 = w; - GET_CURRENT_CONTEXT(ctx); - CLIP_TILE; - FLIP(y); - for (i = 0; i < h; i++) { - for (j = 0; j < w * 4; j++) { - UNCLAMPED_FLOAT_TO_UBYTE(tmp[j], p[j]); + struct xmesa_surface *xms = xmesa_surface(ps); + struct xmesa_renderbuffer *xrb = xms->xrb; + + if (xrb) { + /* this is a front/back color buffer */ + GLubyte tmp[MAX_WIDTH * 4]; + GLuint i, j; + uint w0 = w; + GET_CURRENT_CONTEXT(ctx); + CLIP_TILE; + FLIP(y); + for (i = 0; i < h; i++) { + for (j = 0; j < w * 4; j++) { + UNCLAMPED_FLOAT_TO_UBYTE(tmp[j], p[j]); + } + xrb->St.Base.PutRow(ctx, &xrb->St.Base, w, x, y - i, tmp, NULL); + p += w0 * 4; } - xrb->St.Base.PutRow(ctx, &xrb->St.Base, w, x, y - i, tmp, NULL); - p += w0 * 4; - } #if 0 /* debug: flush */ - { - XMesaContext xm = XMESA_CONTEXT(ctx); - XSync(xm->display, 0); - } + { + XMesaContext xm = XMESA_CONTEXT(ctx); + XSync(xm->display, 0); + } #endif + } + else { + /* other softpipe surface */ + softpipe_put_tile_rgba(pipe, ps, x, y, w, h, p); + } } @@ -141,12 +167,15 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) assert(pipeFormat); - xms->surface.surface.format = pipeFormat; - xms->surface.surface.refcount = 1; + xms->surface.format = pipeFormat; + xms->surface.refcount = 1; +#if 0 /* some of the functions plugged in by this call will get overridden */ softpipe_init_surface_funcs(&xms->surface); +#endif +#if 0 switch (pipeFormat) { case PIPE_FORMAT_U_A8_R8_G8_B8: xms->surface.get_tile = get_tile; @@ -157,6 +186,7 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) default: abort(); } +#endif /* Note, the region we allocate doesn't actually have any storage * since we're drawing into an XImage or Pixmap. @@ -164,10 +194,10 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) * functions. */ if (pipe) - xms->surface.surface.region = pipe->winsys->region_alloc(pipe->winsys, - 1, 0, 0, 0x0); + xms->surface.region = pipe->winsys->region_alloc(pipe->winsys, + 1, 0, 0, 0x0); - return &xms->surface.surface; + return &xms->surface; } @@ -183,14 +213,15 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) assert(pipe); assert(pipeFormat); - xms->surface.surface.format = pipeFormat; - xms->surface.surface.refcount = 1; + xms->surface.format = pipeFormat; + xms->surface.refcount = 1; +#if 0 /* * This is really just a softpipe surface, not an XImage/Pixmap surface. */ softpipe_init_surface_funcs(&xms->surface); - - return &xms->surface.surface; +#endif + return &xms->surface; } diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index 624d28dd01..36805437f0 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -289,6 +289,31 @@ xm_region_release(struct pipe_winsys *winsys, struct pipe_region **region) } +/** + * Called via pipe->surface_alloc() to create new surfaces (textures, + * renderbuffers, etc. + */ +static struct pipe_surface * +xm_surface_alloc(struct pipe_winsys *ws, GLuint pipeFormat) +{ + struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + + assert(ws); + assert(pipeFormat); + + xms->surface.format = pipeFormat; + xms->surface.refcount = 1; +#if 0 + /* + * This is really just a softpipe surface, not an XImage/Pixmap surface. + */ + softpipe_init_surface_funcs(&xms->surface); +#endif + return &xms->surface; +} + + + struct xmesa_pipe_winsys { @@ -320,6 +345,8 @@ xmesa_create_pipe_winsys( XMesaContext xmesa ) xws->winsys.region_alloc = xm_region_alloc; xws->winsys.region_release = xm_region_release; + xws->winsys.surface_alloc = xm_surface_alloc; + xws->winsys.flush_frontbuffer = xm_flush_frontbuffer; xws->winsys.wait_idle = xm_wait_idle; xws->winsys.printf = xm_printf; diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 9cbe8670f9..8af95504fb 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -598,7 +598,7 @@ struct pipe_context; struct xmesa_surface { - struct softpipe_surface surface; + struct pipe_surface surface; struct xmesa_renderbuffer *xrb; }; @@ -618,8 +618,15 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); extern struct pipe_surface * xmesa_new_color_surface(struct pipe_context *pipe, GLuint format); -extern const GLuint * -xmesa_supported_formats(struct pipe_context *pipe, GLuint *numFormats); +extern const uint * +xmesa_supported_formats(struct pipe_context *pipe, uint *numFormats); +extern void +xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, float *p); + +extern void +xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const float *p); #endif diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index a3a0296598..7e02b751bb 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -134,7 +134,9 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover_init_state_functions( failover ); +#if 0 failover->pipe.surface_alloc = hw->surface_alloc; +#endif failover->pipe.get_tex_surface = hw->get_tex_surface; failover->pipe.region_map = hw->region_map; diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index a07a21c13b..b4b5bd1ce5 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -29,13 +29,20 @@ #include "i915_state.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "pipe/p_winsys.h" -struct i915_surface -{ - struct pipe_surface surface; - /* anything else? */ -}; +#define CLIP_TILE \ + do { \ + if (x >= ps->width) \ + return; \ + if (y >= ps->height) \ + return; \ + if (x + w > ps->width) \ + w = ps->width - x; \ + if (y + h > ps->height) \ + h = ps->height -y; \ + } while(0) /** @@ -53,30 +60,44 @@ i915_get_tile_rgba(struct pipe_context *pipe, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); - -#if 0 - assert(x + w <= ps->width); - assert(y + h <= ps->height); -#else - /* temp clipping hack */ - if (x + w > ps->width) - w = ps->width - x; - if (y + h > ps->height) - h = ps->height -y; -#endif - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow += 4; + CLIP_TILE; + + switch (ps->format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const unsigned pixel = src[j]; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow += 4; + } + src += ps->region->pitch; + p += w0 * 4; + } + break; + case PIPE_FORMAT_S8_Z24: + { + const float scale = 1.0 / (float) 0xffffff; + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const unsigned pixel = src[j]; + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (pixel & 0xffffff) * scale; + pRow += 4; + } + src += ps->region->pitch; + p += w0 * 4; + } } - src += ps->region->pitch; - p += w0 * 4; + break; + default: + assert(0); } } @@ -86,34 +107,122 @@ i915_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p) { - /* any need to put tiles into i915 surfaces? */ + /* TODO */ assert(0); } +/* + * XXX note: same as code in sp_surface.c + */ +static void +i915_get_tile(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride) +{ + const uint cpp = ps->region->cpp; + const uint w0 = w; + const ubyte *pSrc; + ubyte *pDest; + uint i; -static struct pipe_surface * -i915_surface_alloc(struct pipe_context *pipe, unsigned format) + assert(ps->region->map); + + CLIP_TILE; + + if (dst_stride == 0) { + dst_stride = w0 * cpp; + } + + pSrc = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; + pDest = (ubyte *) p; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w0 * cpp); + pDest += dst_stride; + pSrc += ps->region->pitch * cpp; + } +} + + +/* + * XXX note: same as code in sp_surface.c + */ +static void +i915_put_tile(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride) { - struct i915_surface *surf = CALLOC_STRUCT(i915_surface); + const uint cpp = ps->region->cpp; + const uint w0 = w; + const ubyte *pSrc; + ubyte *pDest; + uint i; + + assert(ps->region->map); + + CLIP_TILE; + + if (src_stride == 0) { + src_stride = w0 * cpp; + } + + pSrc = (const ubyte *) p; + pDest = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w0 * cpp); + pDest += ps->region->pitch * cpp; + pSrc += src_stride; + } +} - if (!surf) - return NULL; - surf->surface.format = format; - surf->surface.refcount = 1; +/* + * XXX note: same as code in sp_surface.c + */ +static struct pipe_surface * +i915_get_tex_surface(struct pipe_context *pipe, + struct pipe_mipmap_tree *mt, + unsigned face, unsigned level, unsigned zslice) +{ + struct pipe_surface *ps; + unsigned offset; /* in bytes */ - // surf->surface.get_tile = i915_get_tile; - // surf->surface.put_tile = i915_put_tile; + offset = mt->level[level].level_offset; - return &surf->surface; + if (mt->target == PIPE_TEXTURE_CUBE) { + offset += mt->level[level].image_offset[face] * mt->cpp; + } + else if (mt->target == PIPE_TEXTURE_3D) { + offset += mt->level[level].image_offset[zslice] * mt->cpp; + } + else { + assert(face == 0); + assert(zslice == 0); + } + + ps = pipe->winsys->surface_alloc(pipe->winsys, mt->format); + if (ps) { + assert(ps->format); + assert(ps->refcount); + pipe_region_reference(&ps->region, mt->region); + ps->width = mt->level[level].width; + ps->height = mt->level[level].height; + ps->offset = offset; + } + return ps; } void i915_init_surface_functions(struct i915_context *i915) { - i915->pipe.surface_alloc = i915_surface_alloc; + i915->pipe.get_tex_surface = i915_get_tex_surface; + i915->pipe.get_tile = i915_get_tile; + i915->pipe.put_tile = i915_put_tile; i915->pipe.get_tile_rgba = i915_get_tile_rgba; i915->pipe.put_tile_rgba = i915_put_tile_rgba; } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index b9de3667e5..3a041f158b 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -176,13 +176,7 @@ struct pipe_context { unsigned index, const struct pipe_feedback_buffer *); - /* - * Surface functions - * This might go away... - */ - struct pipe_surface *(*surface_alloc)(struct pipe_context *pipe, - unsigned format); - + /** Get a surface which is a "view" into a texture */ struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, struct pipe_mipmap_tree *texture, unsigned face, unsigned level, @@ -198,7 +192,6 @@ struct pipe_context { struct pipe_surface *ps, uint x, uint y, uint w, uint h, const void *p, int src_stride); - /* XXX temporary here, move these to softpipe */ void (*get_tile_rgba)(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p); diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 3b04c44733..10a2caf1a2 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -77,6 +77,10 @@ struct pipe_winsys void (*region_release)(struct pipe_winsys *ws, struct pipe_region **r); + /** allocate a new surface (no context dependency) */ + struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws, + unsigned format); + /** * The buffer manager is modeled after the dri_bufmgr interface, which * in turn is modeled after the ARB_vertex_buffer_object extension, diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 4accafa384..d1aa2aba97 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -27,6 +27,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "pipe/p_winsys.h" #include "sp_context.h" #include "sp_state.h" #include "sp_surface.h" @@ -361,7 +362,7 @@ i8_get_tile(struct pipe_surface *ps, static void a8_l8_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) + unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ushort *src = ((const ushort *) (ps->region->map + ps->offset)) @@ -466,7 +467,7 @@ void softpipe_init_surface_funcs(struct softpipe_surface *sps) { assert(sps->surface.format); - +#if 0 switch (sps->surface.format) { case PIPE_FORMAT_U_A8_R8_G8_B8: sps->get_tile = a8r8g8b8_get_tile; @@ -507,6 +508,7 @@ softpipe_init_surface_funcs(struct softpipe_surface *sps) default: assert(0); } +#endif } @@ -555,7 +557,7 @@ softpipe_get_tex_surface(struct pipe_context *pipe, assert(zslice == 0); } - ps = pipe->surface_alloc(pipe, mt->format); + ps = pipe->winsys->surface_alloc(pipe->winsys, mt->format); if (ps) { assert(ps->format); assert(ps->refcount); @@ -637,26 +639,90 @@ softpipe_put_tile(struct pipe_context *pipe, /* XXX TEMPORARY */ -static void -get_tile_rgba_generic(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p) +void +softpipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p) { - struct softpipe_surface *sps = softpipe_surface(ps); - sps->get_tile(ps, x, y, w, h, p); + switch (ps->format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + a8r8g8b8_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_A1_R5_G5_B5: + a1r5g5b5_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_L8: + l8_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_A8: + a8_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_I8: + i8_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_A8_L8: + a8_l8_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_S_R16_G16_B16_A16: + r16g16b16a16_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_Z16: + z16_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_Z32: + z32_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_S8_Z24: + s8z24_get_tile(ps, x, y, w, h, p); + break; + default: + assert(0); + } } /* XXX TEMPORARY */ -static void -put_tile_rgba_generic(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p) +void +softpipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p) { - struct softpipe_surface *sps = softpipe_surface(ps); - sps->put_tile(ps, x, y, w, h, p); + switch (ps->format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + a8r8g8b8_put_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_A1_R5_G5_B5: + /*a1r5g5b5_put_tile(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_L8: + /*l8_put_tile(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_A8: + /*a8_put_tile(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_I8: + /*i8_put_tile(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_A8_L8: + /*a8_l8_put_tile(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_S_R16_G16_B16_A16: + r16g16b16a16_put_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_Z16: + /*z16_put_tile(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_Z32: + /*z32_put_tile(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_S8_Z24: + /*s8z24_put_tile(ps, x, y, w, h, p);*/ + break; + default: + assert(0); + } } @@ -664,11 +730,12 @@ put_tile_rgba_generic(struct pipe_context *pipe, void sp_init_surface_functions(struct softpipe_context *sp) { +#if 0 sp->pipe.surface_alloc = softpipe_surface_alloc; - +#endif sp->pipe.get_tile = softpipe_get_tile; sp->pipe.put_tile = softpipe_put_tile; - sp->pipe.get_tile_rgba = get_tile_rgba_generic; - sp->pipe.put_tile_rgba = put_tile_rgba_generic; + sp->pipe.get_tile_rgba = softpipe_get_tile_rgba; + sp->pipe.put_tile_rgba = softpipe_put_tile_rgba; } diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index af6533d4f0..359a438c86 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -46,20 +46,33 @@ struct softpipe_tile_cache; struct softpipe_surface { struct pipe_surface surface; +#if 0 /* XXX these are temporary here */ void (*get_tile)(struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p); void (*put_tile)(struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p); +#endif }; - extern struct pipe_surface * softpipe_get_tex_surface(struct pipe_context *pipe, struct pipe_mipmap_tree *mt, unsigned face, unsigned level, unsigned zslice); +extern void +softpipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p); + +extern void +softpipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p); + extern void softpipe_init_surface_funcs(struct softpipe_surface *sps); diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index f22132b3cf..160edc4274 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -84,7 +84,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, cpp = info->size; if (!strb->surface) { - strb->surface = pipe->surface_alloc(pipe, pipeFormat); + strb->surface = pipe->winsys->surface_alloc(pipe->winsys, pipeFormat); assert(strb->surface); if (!strb->surface) return GL_FALSE; -- cgit v1.2.3 From 61d998c966d10e7a44b06d378c54b5f21ad69b53 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 25 Oct 2007 12:14:07 -0400 Subject: Make ret part of the instruction block instead of implicitely closing at the end of parsing. --- src/mesa/pipe/llvm/instructions.cpp | 5 +++++ src/mesa/pipe/llvm/instructions.h | 1 + src/mesa/pipe/llvm/llvmtgsi.cpp | 3 +-- 3 files changed, 7 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 5b54af8717..3fca522324 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -1080,3 +1080,8 @@ llvm::Value * Instructions::trunc(llvm::Value *in) return vectorFromVals(fx, fy, fz, fw); } +void Instructions::end() +{ + new ReturnInst(m_block); +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 29ae168e76..82d871d410 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -64,6 +64,7 @@ public: void elseop(); void endif(); void endLoop(); + void end(); llvm::Value *ex2(llvm::Value *in); llvm::Value *floor(llvm::Value *in); llvm::Value *frc(llvm::Value *in); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 6eae46d3ae..b1175d7f9b 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -595,6 +595,7 @@ translate_instruction(llvm::Module *module, case TGSI_OPCODE_KIL: break; case TGSI_OPCODE_END: + instr->end(); return; break; default: @@ -696,8 +697,6 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) } } - new ReturnInst(instr.currentBlock()); - tgsi_parse_free(&parse); prog->num_consts = storage.numConsts(); -- cgit v1.2.3 From 56da35ef76b062f2d37f37a4d3d986e333c4954d Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 26 Oct 2007 08:20:55 -0400 Subject: Fix swizzle fetching in tgsi_util_get_full_src_register_extswizzle. Shorten the swizzle coding in LLVM compilation using it and fix some warnings. --- src/mesa/pipe/llvm/llvm_base_shader.cpp | 35 +++++++++-------------- src/mesa/pipe/llvm/llvmtgsi.cpp | 49 ++++++++------------------------- src/mesa/pipe/tgsi/exec/tgsi_util.c | 2 +- 3 files changed, 25 insertions(+), 61 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index 3a458cb11f..f6fc83be9a 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -30,8 +30,6 @@ Module* createBaseShader() { /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_0_PAL); - PointerType* PointerTy_7 = PointerType::get(Type::FloatTy); - PointerType* PointerTy_8 = PointerType::get(VectorTy_3); std::vectorFuncTy_9_args; @@ -75,12 +73,8 @@ Module* createBaseShader() { ArrayType* ArrayTy_14 = ArrayType::get(ArrayTy_2, 2048); - PointerType* PointerTy_13 = PointerType::get(ArrayTy_14); - ArrayType* ArrayTy_16 = ArrayType::get(VectorTy_3, 32); - PointerType* PointerTy_15 = PointerType::get(ArrayTy_16); - std::vectorFuncTy_18_args; FuncTy_18_args.push_back(PointerTy_8); FuncTy_18_args.push_back(PointerTy_8); @@ -91,10 +85,7 @@ Module* createBaseShader() { /*Params=*/FuncTy_18_args, /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_18_PAL); - - PointerType* PointerTy_17 = PointerType::get(FuncTy_18); - - + // Function Declarations Function* func_from_array = new Function( @@ -219,7 +210,7 @@ Module* createBaseShader() { ptr_arrayidx54_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0); Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); - StoreInst* void_29 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); + new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_21, "inc", label_forbody6); ICmpInst* int1_cmp59 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp59", label_forbody6); new BranchInst(label_forbody6, label_forinc57, int1_cmp59, label_forbody6); @@ -296,7 +287,7 @@ Module* createBaseShader() { LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_23, "tmp31", label_forbody); GetElementPtrInst* ptr_arrayidx34_45 = new GetElementPtrInst(ptr_res_33, int32_i_0_reg2mem_0_39, "arrayidx34", label_forbody); - StoreInst* void_46 = new StoreInst(packed_tmp31, ptr_arrayidx34_45, false, label_forbody); + new StoreInst(packed_tmp31, ptr_arrayidx34_45, false, label_forbody); BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_39, const_int32_21, "indvar.next", label_forbody); ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_35, "exitcond", label_forbody); new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); @@ -342,25 +333,25 @@ Module* createBaseShader() { GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_55, "arrayidx6", label_forbody_51); LoadInst* packed_tmp7_58 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_51); ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_58, const_int32_19, "tmp11", label_forbody_51); - StoreInst* void_59 = new StoreInst(float_tmp11, ptr_arraydecay_57, false, label_forbody_51); + new StoreInst(float_tmp11, ptr_arraydecay_57, false, label_forbody_51); std::vector ptr_arrayidx13_indices; ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_55); ptr_arrayidx13_indices.push_back(const_int32_21); Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_51); ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_58, const_int32_21, "tmp15", label_forbody_51); - StoreInst* void_60 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_51); + new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_51); std::vector ptr_arrayidx17_indices; ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_55); ptr_arrayidx17_indices.push_back(const_int32_22); Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_51); ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_58, const_int32_22, "tmp19", label_forbody_51); - StoreInst* void_61 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_51); + new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_51); std::vector ptr_arrayidx21_indices; ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_55); ptr_arrayidx21_indices.push_back(const_int32_23); Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_51); ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_58, const_int32_23, "tmp23", label_forbody_51); - StoreInst* void_62 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_51); + new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_51); BinaryOperator* int32_indvar_next_63 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_55, const_int32_21, "indvar.next", label_forbody_51); ICmpInst* int1_exitcond_64 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_63, int32_num_attribs_49, "exitcond", label_forbody_51); new BranchInst(label_afterfor_52, label_forbody_51, int1_exitcond_64, label_forbody_51); @@ -469,7 +460,7 @@ Module* createBaseShader() { ptr_arrayidx54_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx54_i_indices.push_back(int32_j_0_reg2mem_0_i); Instruction* ptr_arrayidx54_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx54_i_indices.begin(), ptr_arrayidx54_i_indices.end(), "arrayidx54.i", label_forbody6_i); - StoreInst* void_78 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); + new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_21, "inc.i", label_forbody6_i); ICmpInst* int1_cmp59_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_inputs, "cmp59.i", label_forbody6_i); new BranchInst(label_forbody6_i, label_forinc57_i, int1_cmp59_i, label_forbody6_i); @@ -524,7 +515,7 @@ Module* createBaseShader() { ptr_arrayidx34_i12_indices.push_back(const_int32_19); ptr_arrayidx34_i12_indices.push_back(int32_i_0_reg2mem_0_i5); Instruction* ptr_arrayidx34_i12 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i12_indices.begin(), ptr_arrayidx34_i12_indices.end(), "arrayidx34.i12", label_forbody_i15); - StoreInst* void_84 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); + new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_21, "indvar.next24", label_forbody_i15); ICmpInst* int1_exitcond25 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next24, int32_num_consts, "exitcond25", label_forbody_i15); new BranchInst(label_forcond, label_forbody_i15, int1_exitcond25, label_forbody_i15); @@ -590,28 +581,28 @@ Module* createBaseShader() { Instruction* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_results, ptr_arrayidx6_i_us_indices.begin(), ptr_arrayidx6_i_us_indices.end(), "arrayidx6.i.us", label_forbody_i_us); LoadInst* packed_tmp7_i_us = new LoadInst(ptr_arrayidx6_i_us, "tmp7.i.us", false, label_forbody_i_us); ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_19, "tmp11.i.us", label_forbody_i_us); - StoreInst* void_93 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); + new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx13_i_us_indices; ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); ptr_arrayidx13_i_us_indices.push_back(const_int32_21); Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_21, "tmp15.i3.us", label_forbody_i_us); - StoreInst* void_94 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); + new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx17_i_us_indices; ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); ptr_arrayidx17_i_us_indices.push_back(const_int32_22); Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_22, "tmp19.i.us", label_forbody_i_us); - StoreInst* void_95 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); + new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx21_i_us_indices; ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); ptr_arrayidx21_i_us_indices.push_back(const_int32_23); Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_23, "tmp23.i.us", label_forbody_i_us); - StoreInst* void_96 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); + new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); BinaryOperator* int32_indvar_next_97 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_21, "indvar.next", label_forbody_i_us); ICmpInst* int1_exitcond_98 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_97, int32_num_attribs_69, "exitcond", label_forbody_i_us); new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_98, label_forbody_i_us); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index b1175d7f9b..6dfd7926fb 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -166,7 +166,6 @@ translate_immediate(llvm::Module *module, { } - static void translate_instruction(llvm::Module *module, Storage *storage, @@ -202,41 +201,15 @@ translate_instruction(llvm::Module *module, return; } - if (src->SrcRegister.Extended) { - if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || - src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || - src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || - src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { - int swizzle = 0; - - if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X) - swizzle = src->SrcRegisterExtSwz.ExtSwizzleX * 1000; - else - swizzle = src->SrcRegister.SwizzleX * 1000; - if (src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y) - swizzle += src->SrcRegisterExtSwz.ExtSwizzleY * 100; - else - swizzle += src->SrcRegister.SwizzleY * 100; - if (src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z) - swizzle += src->SrcRegisterExtSwz.ExtSwizzleZ * 10; - else - swizzle += src->SrcRegister.SwizzleZ * 10; - if (src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) - swizzle += src->SrcRegisterExtSwz.ExtSwizzleW * 1; - else - swizzle += src->SrcRegister.SwizzleW * 1; - /*fprintf(stderr, "EXT XXXXXXXX swizzle x = %d\n", swizzle);*/ - - val = storage->shuffleVector(val, swizzle); - } - } else if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || - src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || - src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || - src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) { - int swizzle = src->SrcRegister.SwizzleX * 1000; - swizzle += src->SrcRegister.SwizzleY * 100; - swizzle += src->SrcRegister.SwizzleZ * 10; - swizzle += src->SrcRegister.SwizzleW * 1; + int swizzle = 0; + int xstart = 1000; + const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + + TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; + for (int k = 0; k < 4; ++k) { + swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * xstart; + xstart /= 10; + } + if (swizzle != NO_SWIZZLE) { /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ val = storage->shuffleVector(val, swizzle); } @@ -245,8 +218,8 @@ translate_instruction(llvm::Module *module, /*if (inputs[0]) instr->printVector(inputs[0]); - if (inputs[1]) - instr->printVector(inputs[1]);*/ + if (inputs[1]) + instr->printVector(inputs[1]);*/ llvm::Value *out = 0; switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: { diff --git a/src/mesa/pipe/tgsi/exec/tgsi_util.c b/src/mesa/pipe/tgsi/exec/tgsi_util.c index 38d6d6e6bc..67e32ccbd0 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_util.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_util.c @@ -88,7 +88,7 @@ tgsi_util_get_full_src_register_extswizzle( if( swizzle <= TGSI_SWIZZLE_W ) { swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegister, - swizzle ); + component ); } return swizzle; -- cgit v1.2.3 From 67e4b8299620db2e2f33795621b23e9827604bb1 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 26 Oct 2007 14:52:10 -0400 Subject: Get basic function calls working in the shaders. --- src/mesa/pipe/llvm/instructions.cpp | 110 +++++++++++++++++++++++++++----- src/mesa/pipe/llvm/instructions.h | 11 ++++ src/mesa/pipe/llvm/llvm_base_shader.cpp | 2 +- src/mesa/pipe/llvm/llvmtgsi.cpp | 36 ++++++++--- src/mesa/pipe/llvm/storage.cpp | 40 +++++++++++- src/mesa/pipe/llvm/storage.h | 16 +++++ 6 files changed, 190 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 3fca522324..645ab9106f 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -32,6 +32,8 @@ #include "instructions.h" +#include "storage.h" + #include #include #include @@ -39,11 +41,22 @@ #include #include -using namespace llvm; +#include +#include +#include +using namespace llvm; Function* makeLitFunction(Module *mod); +static inline std::string createFuncName(int label) +{ + std::ostringstream stream; + stream << "function"; + stream << label; + return stream.str(); +} + Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block) : m_mod(mod), m_func(func), m_block(block), m_idx(0) { @@ -623,18 +636,18 @@ void Instructions::printVector(llvm::Value *val) llvm::Function * Instructions::declarePrintf() { std::vector args; - ParamAttrsList *params = 0; - FunctionType* funcTy = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/args, - /*isVarArg=*/true, - /*ParamAttrs=*/params); - Function* func_printf = new Function( - /*Type=*/funcTy, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", m_mod); - func_printf->setCallingConv(CallingConv::C); - return func_printf; + ParamAttrsList *params = 0; + FunctionType* funcTy = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/args, + /*isVarArg=*/true, + /*ParamAttrs=*/params); + Function* func_printf = new Function( + /*Type=*/funcTy, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", m_mod); + func_printf->setCallingConv(CallingConv::C); + return func_printf; } @@ -822,8 +835,6 @@ Function* makeLitFunction(Module *mod) { /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_0_PAL); - PointerType* PointerTy_1 = PointerType::get(FuncTy_0); - VectorType* VectorTy_2 = VectorType::get(Type::FloatTy, 4); std::vectorFuncTy_3_args; @@ -1085,3 +1096,72 @@ void Instructions::end() new ReturnInst(m_block); } +void Instructions::cal(int label, llvm::Value *out, llvm::Value *in, + llvm::Value *cst) +{ + std::vector params; + params.push_back(out); + params.push_back(in); + params.push_back(cst); + llvm::Function *func = findFunction(label); + + new CallInst(func, params.begin(), params.end(), std::string(), m_block); +} + +llvm::Function * Instructions::declareFunc(int label) +{ + PointerType *vecPtr = PointerType::get(m_floatVecType); + std::vector args; + args.push_back(vecPtr); + args.push_back(vecPtr); + args.push_back(vecPtr); + ParamAttrsList *params = 0; + FunctionType *funcType = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/args, + /*isVarArg=*/false, + /*ParamAttrs=*/params); + std::string name = createFuncName(label); + Function *func = new Function( + /*Type=*/funcType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/name.c_str(), m_mod); + func->setCallingConv(CallingConv::C); + return func; +} + +void Instructions::bgnSub(unsigned label, Storage *storage) +{ + llvm::Function *func = findFunction(label); + + Function::arg_iterator args = func->arg_begin(); + Value *ptr_OUT = args++; + ptr_OUT->setName("OUT"); + Value *ptr_IN = args++; + ptr_IN->setName("IN"); + Value *ptr_CONST = args++; + ptr_CONST->setName("CONST"); + storage->pushArguments(ptr_OUT, ptr_IN, ptr_CONST); + + llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); + + m_func = func; + m_block = entry; +} + +void Instructions::endSub() +{ + m_func = 0; + m_block = 0; +} + +llvm::Function * Instructions::findFunction(int label) +{ + llvm::Function *func = m_functions[label]; + if (!func) { + func = declareFunc(label); + m_functions[label] = func; + } + return func; +} + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 82d871d410..85feb1665d 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -37,6 +37,7 @@ #include #include +#include #include namespace llvm { @@ -44,6 +45,8 @@ namespace llvm { class Function; } +class Storage; + class Instructions { public: @@ -55,7 +58,10 @@ public: llvm::Value *arl(llvm::Value *in1); llvm::Value *add(llvm::Value *in1, llvm::Value *in2); void beginLoop(); + void bgnSub(unsigned, Storage *); void brk(); + void cal(int label, llvm::Value *out, llvm::Value *in, + llvm::Value *cst); llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); @@ -65,6 +71,7 @@ public: void endif(); void endLoop(); void end(); + void endSub(); llvm::Value *ex2(llvm::Value *in); llvm::Value *floor(llvm::Value *in); llvm::Value *frc(llvm::Value *in); @@ -101,6 +108,9 @@ private: llvm::Value *z, llvm::Value *w=0); llvm::Function *declarePrintf(); + llvm::Function *declareFunc(int label); + + llvm::Function *findFunction(int label); private: llvm::Module *m_mod; llvm::Function *m_func; @@ -125,6 +135,7 @@ private: llvm::BasicBlock *end; }; std::stack m_loopStack; + std::map m_functions; }; #endif diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index f6fc83be9a..3f058258ee 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -634,7 +634,7 @@ Module* createBaseShader() { BinaryOperator* int32_inc_103 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_100, const_int32_21, "inc", label_forbody_71); ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_103, int32_num_vertices, "cmp21", label_forbody_71); new BranchInst(label_forbody_71, label_afterfor_72, int1_cmp21, label_forbody_71); - + // Block afterfor (label_afterfor_72) new ReturnInst(label_afterfor_72); diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 6dfd7926fb..cfeb19e4ba 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -171,7 +171,8 @@ translate_instruction(llvm::Module *module, Storage *storage, Instructions *instr, struct tgsi_full_instruction *inst, - struct tgsi_full_instruction *fi) + struct tgsi_full_instruction *fi, + unsigned instno) { llvm::Value *inputs[4]; inputs[0] = 0; @@ -400,9 +401,18 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_BRA: break; - case TGSI_OPCODE_CAL: + case TGSI_OPCODE_CAL: { + instr->cal(inst->InstructionExtLabel.Label, + storage->outputPtr(), + storage->inputPtr(), + storage->constPtr()); + return; + } break; - case TGSI_OPCODE_RET: + case TGSI_OPCODE_RET: { + instr->end(); + return; + } break; case TGSI_OPCODE_SSG: break; @@ -495,15 +505,24 @@ translate_instruction(llvm::Module *module, return; } break; - case TGSI_OPCODE_BGNSUB: + case TGSI_OPCODE_BGNSUB: { + instr->bgnSub(instno, storage); + storage->setCurrentBlock(instr->currentBlock()); + return; + } break; case TGSI_OPCODE_ENDLOOP2: { instr->endLoop(); storage->setCurrentBlock(instr->currentBlock()); + storage->popArguments(); return; } break; - case TGSI_OPCODE_ENDSUB: + case TGSI_OPCODE_ENDSUB: { + instr->endSub(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } break; case TGSI_OPCODE_NOISE1: break; @@ -620,7 +639,7 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) struct tgsi_parse_context parse; struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; - + unsigned instno = 0; Function* shader = mod->getFunction("execute_shader"); std::ostringstream stream; stream << "execute_shader"; @@ -662,7 +681,8 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) case TGSI_TOKEN_TYPE_INSTRUCTION: translate_instruction(mod, &storage, &instr, &parse.FullToken.FullInstruction, - &fi); + &fi, instno); + ++instno; break; default: @@ -776,7 +796,7 @@ void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) llvm::Function *func = mod->getFunction(func_name.c_str()); assert(func); std::cout<<"; ---------- Start shader "<id<id< #include +#include #include namespace llvm { @@ -53,6 +54,10 @@ public: llvm::Value *out, llvm::Value *in, llvm::Value *consts); + llvm::Value *inputPtr() const; + llvm::Value *outputPtr() const; + llvm::Value *constPtr() const; + void setCurrentBlock(llvm::BasicBlock *block); llvm::ConstantInt *constantInt(int); @@ -76,6 +81,10 @@ public: int numConsts() const; + void pushArguments(llvm::Value *out, llvm::Value *in, + llvm::Value *constPtr); + void popArguments(); + private: llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); const char *name(const char *prefix); @@ -106,6 +115,13 @@ private: int m_numConsts; std::map m_destWriteMap; + + struct Args { + llvm::Value *out; + llvm::Value *in; + llvm::Value *cst; + }; + std::stack m_argStack; }; #endif -- cgit v1.2.3 From 6b30f3888e46c3981f1e4fc34c155c7539275420 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 26 Oct 2007 12:20:17 -0600 Subject: Initial support for immediate values in TGSI programs. These can be evaluated at compile time. Code disabled pending clarifications of TGSI immediate data structures. --- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 6 +++- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 11 +++++++ src/mesa/pipe/tgsi/exec/tgsi_exec.h | 4 +-- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 60 +++++++++++++++++++++++++++++++--- 4 files changed, 74 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index 9d21da0965..e7eb811d18 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -89,7 +89,7 @@ text_dump_flt( { char str[48]; - sprintf( str, "%40.6f", f ); + sprintf( str, "%10.4f", f ); text_dump_str( dump, str ); } @@ -780,7 +780,11 @@ dump_immediate_short( ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); TXT( " { " ); +#if 0 for( i = 0; i < imm->Immediate.Size - 1; i++ ) { +#else + for( i = 0; i < imm->Immediate.Size; i++ ) { +#endif switch( imm->Immediate.DataType ) { case TGSI_IMM_FLOAT32: FLT( imm->u.ImmediateFloat32[i].Float ); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 3f464372ca..42aed9bd6b 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -170,6 +170,7 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) break; case TGSI_TOKEN_TYPE_IMMEDIATE: +#if 0 assert( (parse.FullToken.FullImmediate.Immediate.Size - 1) % 4 == 0 ); assert( mach->ImmLimit + (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4 <= 256 ); @@ -177,6 +178,16 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; } mach->ImmLimit += (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4; +#else + /* Add this immediate value (vector of 1,2,3,4 floats) to immediates array */ + assert( parse.FullToken.FullImmediate.Immediate.Size <= 4 ); + assert( mach->ImmLimit < TGSI_EXEC_NUM_IMMEDIATES ); + + for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size; i++ ) { + mach->Imms[mach->ImmLimit][i] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } + mach->ImmLimit++; +#endif break; case TGSI_TOKEN_TYPE_INSTRUCTION: diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 6c1d368342..1805e72487 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -93,12 +93,12 @@ struct tgsi_exec_labels #define TGSI_EXEC_NUM_TEMPS (32 + 4) #define TGSI_EXEC_NUM_ADDRS 1 +#define TGSI_EXEC_NUM_IMMEDIATES 256 #define TGSI_EXEC_MAX_COND_NESTING 10 #define TGSI_EXEC_MAX_LOOP_NESTING 10 #define TGSI_EXEC_MAX_CALL_NESTING 10 - /** * Run-time virtual machine state for executing TGSI shader. */ @@ -120,7 +120,7 @@ struct tgsi_exec_machine struct tgsi_sampler *Samplers; - float Imms[256][4]; + float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; unsigned ImmLimit; float (*Consts)[4]; struct tgsi_exec_vector *Inputs; diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 8975b4fd57..4fac61777c 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -1,9 +1,12 @@ #include "tgsi_platform.h" #include "tgsi_mesa.h" #include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "shader/prog_parameter.h" #define TGSI_DEBUG 0 +#define EMIT_IMMEDIATES 0 + /* * Map mesa register file to TGSI register file. @@ -21,9 +24,14 @@ map_register_file( //case PROGRAM_ENV_PARAM: case PROGRAM_STATE_VAR: case PROGRAM_NAMED_PARAM: - case PROGRAM_CONSTANT: case PROGRAM_UNIFORM: return TGSI_FILE_CONSTANT; + case PROGRAM_CONSTANT: +#if EMIT_IMMEDIATES + return TGSI_FILE_IMMEDIATE; +#else + return TGSI_FILE_CONSTANT; +#endif case PROGRAM_INPUT: return TGSI_FILE_INPUT; case PROGRAM_OUTPUT: @@ -49,7 +57,8 @@ map_register_file_index( GLuint file, GLuint index, const GLuint inputMapping[], - const GLuint outputMapping[]) + const GLuint outputMapping[], + const GLuint immediateMapping[]) { switch( file ) { case TGSI_FILE_INPUT: @@ -59,6 +68,11 @@ map_register_file_index( case TGSI_FILE_OUTPUT: return outputMapping[index]; +#if EMIT_IMMEDIATES + case TGSI_FILE_IMMEDIATE: + return immediateMapping[index]; +#endif + default: return index; } @@ -119,12 +133,26 @@ convert_writemask( return writemask; } +#if EMIT_IMMEDIATES +static struct tgsi_full_immediate +make_immediate(const float *value, uint size) +{ + struct tgsi_full_immediate imm; + imm.Immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + imm.Immediate.Size = size; + imm.Immediate.DataType = TGSI_IMM_FLOAT32; + imm.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; + return imm; +} +#endif + static void compile_instruction( const struct prog_instruction *inst, struct tgsi_full_instruction *fullinst, const GLuint inputMapping[], const GLuint outputMapping[], + const GLuint immediateMapping[], GLuint preamble_size, GLuint processor ) { @@ -144,7 +172,8 @@ compile_instruction( fulldst->DstRegister.File, inst->DstReg.Index, inputMapping, - outputMapping + outputMapping, + NULL ); fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); @@ -157,7 +186,8 @@ compile_instruction( fullsrc->SrcRegister.File, inst->SrcReg[i].Index, inputMapping, - outputMapping ); + outputMapping, + immediateMapping); for( j = 0; j < 4; j++ ) { GLuint swz; @@ -595,6 +625,10 @@ tgsi_translate_mesa_program( struct tgsi_processor *processor; struct tgsi_full_instruction fullinst; GLuint preamble_size = 0; + GLuint immediates[1000]; +#if EMIT_IMMEDIATES + GLuint numImmediates = 0; +#endif assert(procType == TGSI_PROCESSOR_FRAGMENT || procType == TGSI_PROCESSOR_VERTEX); @@ -723,12 +757,30 @@ tgsi_translate_mesa_program( } } + /* immediates/literals */ +#if EMIT_IMMEDIATES + for (i = 0; i < program->Parameters->NumParameters; i++) { + if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { + struct tgsi_full_immediate fullimm + = make_immediate(program->Parameters->ParameterValues[i], + program->Parameters->Parameters[i].Size); + ti += tgsi_build_full_immediate(&fullimm, + &tokens[ti], + header, + maxTokens - ti); + immediates[i] = numImmediates; + numImmediates++; + } + } +#endif + for( i = 0; i < program->NumInstructions; i++ ) { compile_instruction( &program->Instructions[i], &fullinst, inputMapping, outputMapping, + immediates, preamble_size, procType ); -- cgit v1.2.3 From f92083c338069c5c22d200e4696d8cd908258492 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 26 Oct 2007 12:24:28 -0600 Subject: if we hit RET w/ empty call stack, halt --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 42aed9bd6b..0125f40dd2 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -2082,6 +2082,13 @@ exec_instruction( if (mach->ExecMask == 0x0) { /* really return now (otherwise, keep executing */ + if (mach->CallStackTop == 0) { + /* returning from main() */ + *pc = -1; + return; + } + *pc = mach->CallStack[--mach->CallStackTop]; + /* pop the Cond, Loop, Cont stacks */ assert(mach->CondStackTop > 0); mach->CondMask = mach->CondStack[--mach->CondStackTop]; @@ -2089,16 +2096,9 @@ exec_instruction( mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; assert(mach->ContStackTop > 0); mach->ContMask = mach->ContStack[--mach->ContStackTop]; - assert(mach->FuncStackTop > 0); mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; - assert(mach->CallStackTop >= 0); - if (mach->CallStackTop == 0) { - /* XXX error? */ - return; - } - *pc = mach->CallStack[--mach->CallStackTop]; UPDATE_EXEC_MASK(mach); } break; -- cgit v1.2.3 From 78c1f8b2decf168d183c52e7b414adb29dd18988 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 26 Oct 2007 13:02:40 -0600 Subject: convert OPCODE_END -> TGSI_OPCODE_RET --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 4fac61777c..6ce7ab9b71 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -466,7 +466,7 @@ compile_instruction( fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; break; case OPCODE_END: - fullinst->Instruction.Opcode = TGSI_OPCODE_END; + fullinst->Instruction.Opcode = TGSI_OPCODE_RET; break; default: assert( 0 ); -- cgit v1.2.3 From 789d248558061fe4d65f664d6770a12b90fa2e34 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 26 Oct 2007 19:12:02 -0400 Subject: Hold a stack of temporaries so that we can redeclare them for all defined functions. Fixes crashes in function calls. --- src/mesa/pipe/llvm/llvmtgsi.cpp | 4 +++- src/mesa/pipe/llvm/storage.cpp | 21 +++++++++++++++++++++ src/mesa/pipe/llvm/storage.h | 3 +++ 3 files changed, 27 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index cfeb19e4ba..af602326ae 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -508,19 +508,21 @@ translate_instruction(llvm::Module *module, case TGSI_OPCODE_BGNSUB: { instr->bgnSub(instno, storage); storage->setCurrentBlock(instr->currentBlock()); + storage->pushTemps(); return; } break; case TGSI_OPCODE_ENDLOOP2: { instr->endLoop(); storage->setCurrentBlock(instr->currentBlock()); - storage->popArguments(); return; } break; case TGSI_OPCODE_ENDSUB: { instr->endSub(); storage->setCurrentBlock(instr->currentBlock()); + storage->popArguments(); + storage->popTemps(); return; } break; diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 88ef6711cf..7300cdfef0 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -369,3 +369,24 @@ void Storage::popArguments() m_CONST = arg.cst; m_argStack.pop(); } + +void Storage::pushTemps() +{ + m_tempStack.push(m_temps); + std::vector oldTemps = m_temps; + m_temps = std::vector(32); + int i = 0; + for (std::vector::iterator itr = oldTemps.begin(); + itr != oldTemps.end(); ++itr) { + if (*itr) { + declareTemp(i); + } + ++i; + } +} + +void Storage::popTemps() +{ + m_temps = m_tempStack.top(); + m_tempStack.pop(); +} diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index b8d6eb0604..eeadaa0506 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -84,6 +84,8 @@ public: void pushArguments(llvm::Value *out, llvm::Value *in, llvm::Value *constPtr); void popArguments(); + void pushTemps(); + void popTemps(); private: llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); @@ -122,6 +124,7 @@ private: llvm::Value *cst; }; std::stack m_argStack; + std::stack > m_tempStack; }; #endif -- cgit v1.2.3 From 8fed2466e4056668a76a87cf935b5fbff8ae15ca Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 26 Oct 2007 19:19:09 -0600 Subject: Re-implement GLSL texture sampler variables. GLSL sampler variables indicate which texture unit to use for TEX instructions. Previously, this was baked into the fragment/vertex program and couldn't be readily changed once set. Now, SamplerUnits[] array indicates which texture unit is to be used for each sampler variable. These values are set with glUniform1i(). This is extra state that must be passed to the fragment/vertex program executor at runtime. --- src/mesa/main/config.h | 1 + src/mesa/main/mtypes.h | 6 ++ src/mesa/pipe/failover/fo_context.h | 1 + src/mesa/pipe/failover/fo_state.c | 14 ++++ src/mesa/pipe/i915simple/i915_context.h | 1 + src/mesa/pipe/i915simple/i915_state.c | 10 +++ src/mesa/pipe/p_context.h | 3 + src/mesa/pipe/softpipe/sp_context.c | 1 + src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_quad_fs.c | 1 + src/mesa/pipe/softpipe/sp_state.h | 3 + src/mesa/pipe/softpipe/sp_state_sampler.c | 14 ++++ src/mesa/pipe/tgsi/exec/tgsi_exec.c | 5 +- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 1 + src/mesa/shader/prog_execute.c | 16 ++--- src/mesa/shader/prog_execute.h | 2 + src/mesa/shader/prog_instruction.c | 11 +++ src/mesa/shader/prog_instruction.h | 5 ++ src/mesa/shader/prog_parameter.c | 7 +- src/mesa/shader/prog_parameter.h | 2 +- src/mesa/shader/prog_print.c | 7 ++ src/mesa/shader/program.c | 5 ++ src/mesa/shader/shader_api.c | 113 +++++++++++++++++++----------- src/mesa/shader/slang/slang_codegen.c | 15 +++- src/mesa/shader/slang/slang_compile.c | 6 ++ src/mesa/shader/slang/slang_emit.c | 6 +- src/mesa/shader/slang/slang_link.c | 71 +++++++------------ src/mesa/shader/slang/slang_link.h | 4 -- src/mesa/shader/slang/slang_typeinfo.h | 1 + src/mesa/state_tracker/st_atom_sampler.c | 33 ++++++++- src/mesa/swrast/s_fragprog.c | 2 + 31 files changed, 262 insertions(+), 106 deletions(-) (limited to 'src') diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h index cebef1c383..8c64248845 100644 --- a/src/mesa/main/config.h +++ b/src/mesa/main/config.h @@ -207,6 +207,7 @@ #define MAX_PROGRAM_ADDRESS_REGS 2 #define MAX_UNIFORMS 128 #define MAX_VARYING 8 +#define MAX_SAMPLERS 8 /*@}*/ /** For GL_ARB_vertex_shader */ diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 514170dbcf..8adc4e3373 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -1901,6 +1901,7 @@ struct gl_program GLbitfield InputsRead; /**< Bitmask of which input regs are read */ GLbitfield OutputsWritten; /**< Bitmask of which output regs are written to */ GLbitfield TexturesUsed[MAX_TEXTURE_IMAGE_UNITS]; /**< TEXTURE_x_BIT bitmask */ + GLbitfield SamplersUsed; /**< Bitfield of which samplers are used */ GLbitfield ShadowSamplers; /**< Texture units used for shadow sampling. */ /** Named parameters, constants, etc. from program text */ @@ -1913,6 +1914,11 @@ struct gl_program /** Vertex program user-defined attributes */ struct gl_program_parameter_list *Attributes; + /** Map from sampler unit to texture unit (set by glUniform1i()) */ + GLubyte SamplerUnits[MAX_SAMPLERS]; + /** Which texture target is being sampled (TEXTURE_1D/2D/3D/etc_INDEX) */ + GLubyte SamplerTargets[MAX_SAMPLERS]; + /** Logical counts */ /*@{*/ GLuint NumInstructions; diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index 7a597013ab..759b53ccbe 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -84,6 +84,7 @@ struct failover_context { struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; + uint sampler_units[PIPE_MAX_SAMPLERS]; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index f63137f591..2cd1a50b20 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -294,6 +294,19 @@ failover_set_polygon_stipple( struct pipe_context *pipe, failover->hw->set_polygon_stipple( failover->hw, stipple ); } +static void +failover_set_sampler_units( struct pipe_context *pipe, + uint num_samplers, const uint *units ) +{ + struct failover_context *failover = failover_context(pipe); + uint i; + + for (i = 0; i < num_samplers; i++) + failover->sampler_units[i] = units[i]; + failover->dirty |= FO_NEW_SAMPLER; + failover->hw->set_sampler_units(failover->hw, num_samplers, units); +} + static void * failover_create_rasterizer_state(struct pipe_context *pipe, const struct pipe_rasterizer_state *templ) @@ -470,6 +483,7 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.set_clear_color_state = failover_set_clear_color_state; failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; + failover->pipe.set_sampler_units = failover_set_sampler_units; failover->pipe.set_scissor_state = failover_set_scissor_state; failover->pipe.set_texture_state = failover_set_texture_state; failover->pipe.set_viewport_state = failover_set_viewport_state; diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index a3dd392e75..5a3ecedad2 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -173,6 +173,7 @@ struct i915_context struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; + uint sampler_units[PIPE_MAX_SAMPLERS]; struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 8da5662e3f..05f8a6e1fd 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -446,6 +446,15 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe, { } +static void i915_set_sampler_units(struct pipe_context *pipe, + uint numSamplers, const uint *units) +{ + struct i915_context *i915 = i915_context(pipe); + uint i; + for (i = 0; i < numSamplers; i++) + i915->sampler_units[i] = units[i]; +} + static void * i915_create_fs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { @@ -765,6 +774,7 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_feedback_buffer = i915_set_feedback_buffer; i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; + i915->pipe.set_sampler_units = i915_set_sampler_units; i915->pipe.set_scissor_state = i915_set_scissor_state; i915->pipe.set_texture_state = i915_set_texture_state; i915->pipe.set_viewport_state = i915_set_viewport_state; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 3a041f158b..5497f50f73 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -148,6 +148,9 @@ struct pipe_context { void (*set_polygon_stipple)( struct pipe_context *, const struct pipe_poly_stipple * ); + void (*set_sampler_units)( struct pipe_context *, + uint num_samplers, const uint *units ); + void (*set_scissor_state)( struct pipe_context *, const struct pipe_scissor_state * ); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 476d4ac01c..58ef744f30 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -318,6 +318,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_feedback_state = softpipe_set_feedback_state; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; + softpipe->pipe.set_sampler_units = softpipe_set_sampler_units; softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; softpipe->pipe.set_texture_state = softpipe_set_texture_state; softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 3e77bd6b85..88a418d3c7 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -95,6 +95,7 @@ struct softpipe_context { struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + uint sampler_units[PIPE_MAX_SAMPLERS]; unsigned dirty; /* diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 3371b109fc..9b9504cd15 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -85,6 +85,7 @@ shade_quad( /* Consts does not require 16 byte alignment. */ machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; + machine->SamplerUnits = softpipe->sampler_units; machine->InterpCoefs = quad->coef; machine->Inputs[0].xyzw[0].f[0] = fx; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index c194f0ea0d..61532bcdeb 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -99,6 +99,9 @@ void softpipe_delete_vs_state(struct pipe_context *, void *); void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); +void softpipe_set_sampler_units( struct pipe_context *, + uint numSamplers, const uint *units ); + void softpipe_set_scissor_state( struct pipe_context *, const struct pipe_scissor_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index c00e815f2d..e70eabd578 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -78,3 +78,17 @@ softpipe_set_texture_state(struct pipe_context *pipe, softpipe->dirty |= SP_NEW_TEXTURE; } + + +void +softpipe_set_sampler_units(struct pipe_context *pipe, + uint num_samplers, const uint *units ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + uint i; + for (i = 0; i < num_samplers; i++) + softpipe->sampler_units[i] = units[i]; + softpipe->dirty |= SP_NEW_SAMPLER; +} + + diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 0125f40dd2..66a81b4bd8 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1219,11 +1219,14 @@ exec_tex(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst, boolean biasLod) { - const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + const uint sampler = inst->FullSrcRegisters[1].SrcRegister.Index; + const uint unit = mach->SamplerUnits[sampler]; union tgsi_exec_channel r[8]; uint chan_index; float lodBias; + // printf("Sampler %u unit %u\n", sampler, unit); + switch (inst->InstructionExtTexture.Texture) { case TGSI_TEXTURE_1D: diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 1805e72487..38f9218520 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -118,6 +118,7 @@ struct tgsi_exec_machine struct tgsi_exec_vector *Temps; struct tgsi_exec_vector *Addrs; + uint *SamplerUnits; struct tgsi_sampler *Samplers; float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c index 28d195d0ee..bd64b57eb9 100644 --- a/src/mesa/shader/prog_execute.c +++ b/src/mesa/shader/prog_execute.c @@ -1316,22 +1316,22 @@ _mesa_execute_program(GLcontext * ctx, * The rest of the time, just use zero (until we get a more * sophisticated way of computing lambda). */ + const GLuint unit = machine->Samplers[inst->TexSrcUnit]; GLfloat coord[4], color[4], lambda; #if 0 if (inst->SrcReg[0].File == PROGRAM_INPUT && - inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) - lambda = span->array->lambda[inst->TexSrcUnit][column]; + inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + unit) + lambda = span->array->lambda[unit][column]; else #endif lambda = 0.0; fetch_vector4(&inst->SrcReg[0], machine, coord); - machine->FetchTexelLod(ctx, coord, lambda, inst->TexSrcUnit, - color); + machine->FetchTexelLod(ctx, coord, lambda, unit, color); if (DEBUG_PROG) { printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g], " "lod %f\n", color[0], color[1], color[2], color[3], - inst->TexSrcUnit, + unit, coord[0], coord[1], coord[2], coord[3], lambda); } store_vector4(inst, machine, color); @@ -1375,11 +1375,12 @@ _mesa_execute_program(GLcontext * ctx, case OPCODE_TXP: /* GL_ARB_fragment_program only */ /* Texture lookup w/ projective divide */ { + const GLuint unit = machine->Samplers[inst->TexSrcUnit]; GLfloat texcoord[4], color[4], lambda; #if 0 if (inst->SrcReg[0].File == PROGRAM_INPUT && inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) - lambda = span->array->lambda[inst->TexSrcUnit][column]; + lambda = span->array->lambda[unit][column]; else #endif lambda = 0.0; @@ -1393,8 +1394,7 @@ _mesa_execute_program(GLcontext * ctx, texcoord[1] /= texcoord[3]; texcoord[2] /= texcoord[3]; } - machine->FetchTexelLod(ctx, texcoord, lambda, - inst->TexSrcUnit, color); + machine->FetchTexelLod(ctx, texcoord, lambda, unit, color); store_vector4(inst, machine, color); } break; diff --git a/src/mesa/shader/prog_execute.h b/src/mesa/shader/prog_execute.h index be29eceeda..db7bcee516 100644 --- a/src/mesa/shader/prog_execute.h +++ b/src/mesa/shader/prog_execute.h @@ -62,6 +62,8 @@ struct gl_program_machine GLuint CondCodes[4]; /**< COND_* value for x/y/z/w */ GLint AddressReg[MAX_PROGRAM_ADDRESS_REGS][4]; + GLuint *Samplers; /** Array mapping sampler var to tex unit */ + GLuint CallStack[MAX_PROGRAM_CALL_DEPTH]; /**< For CAL/RET instructions */ GLuint StackDepth; /**< Index/ptr to top of CallStack[] */ diff --git a/src/mesa/shader/prog_instruction.c b/src/mesa/shader/prog_instruction.c index c84c76fd5b..066129037a 100644 --- a/src/mesa/shader/prog_instruction.c +++ b/src/mesa/shader/prog_instruction.c @@ -246,6 +246,17 @@ _mesa_num_inst_dst_regs(gl_inst_opcode opcode) } +GLboolean +_mesa_is_tex_instruction(gl_inst_opcode opcode) +{ + return (opcode == OPCODE_TEX || + opcode == OPCODE_TXB || + opcode == OPCODE_TXD || + opcode == OPCODE_TXL || + opcode == OPCODE_TXP); +} + + /** * Return string name for given program opcode. */ diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h index 643969b367..e8a2407ea8 100644 --- a/src/mesa/shader/prog_instruction.h +++ b/src/mesa/shader/prog_instruction.h @@ -413,11 +413,13 @@ struct prog_instruction */ GLint BranchTarget; +#if 0 /** * For TEX instructions in shaders, the sampler to use for the * texture lookup. */ GLint Sampler; +#endif const char *Comment; }; @@ -443,6 +445,9 @@ _mesa_num_inst_src_regs(gl_inst_opcode opcode); extern GLuint _mesa_num_inst_dst_regs(gl_inst_opcode opcode); +extern GLboolean +_mesa_is_tex_instruction(gl_inst_opcode opcode); + extern const char * _mesa_opcode_string(gl_inst_opcode opcode); diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index 9e3d3fecf2..b4008abbd9 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -283,22 +283,25 @@ _mesa_add_uniform(struct gl_program_parameter_list *paramList, * Add a sampler to the parameter list. * \param name uniform's name * \param datatype GL_SAMPLER_2D, GL_SAMPLER_2D_RECT_ARB, etc. + * \param index the sampler number (as seen in TEX instructions) */ GLint _mesa_add_sampler(struct gl_program_parameter_list *paramList, - const char *name, GLenum datatype) + const char *name, GLenum datatype, GLuint index) { GLint i = _mesa_lookup_parameter_index(paramList, -1, name); if (i >= 0 && paramList->Parameters[i].Type == PROGRAM_SAMPLER) { ASSERT(paramList->Parameters[i].Size == 1); ASSERT(paramList->Parameters[i].DataType == datatype); + ASSERT(paramList->ParameterValues[i][0] == index); /* already in list */ return i; } else { + GLfloat indexf = index; const GLint size = 1; /* a sampler is basically a texture unit number */ i = _mesa_add_parameter(paramList, PROGRAM_SAMPLER, name, - size, datatype, NULL, NULL); + size, datatype, &indexf, NULL); return i; } } diff --git a/src/mesa/shader/prog_parameter.h b/src/mesa/shader/prog_parameter.h index 09ff851ea7..40c8c09e09 100644 --- a/src/mesa/shader/prog_parameter.h +++ b/src/mesa/shader/prog_parameter.h @@ -104,7 +104,7 @@ _mesa_add_uniform(struct gl_program_parameter_list *paramList, extern GLint _mesa_add_sampler(struct gl_program_parameter_list *paramList, - const char *name, GLenum datatype); + const char *name, GLenum datatype, GLuint index); extern GLint _mesa_add_varying(struct gl_program_parameter_list *paramList, diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index e92837f739..c421b1228b 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -719,6 +719,8 @@ _mesa_print_program_opt(const struct gl_program *prog, void _mesa_print_program_parameters(GLcontext *ctx, const struct gl_program *prog) { + GLuint i; + _mesa_printf("InputsRead: 0x%x\n", prog->InputsRead); _mesa_printf("OutputsWritten: 0x%x\n", prog->OutputsWritten); _mesa_printf("NumInstructions=%d\n", prog->NumInstructions); @@ -726,6 +728,11 @@ _mesa_print_program_parameters(GLcontext *ctx, const struct gl_program *prog) _mesa_printf("NumParameters=%d\n", prog->NumParameters); _mesa_printf("NumAttributes=%d\n", prog->NumAttributes); _mesa_printf("NumAddressRegs=%d\n", prog->NumAddressRegs); + _mesa_printf("Samplers=[ "); + for (i = 0; i < MAX_SAMPLERS; i++) { + _mesa_printf("%d ", prog->SamplerUnits[i]); + } + _mesa_printf("]\n"); _mesa_load_state_parameters(ctx, prog->Parameters); diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 1f227390af..cafc0dcfaa 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -187,12 +187,17 @@ _mesa_init_program_struct( GLcontext *ctx, struct gl_program *prog, { (void) ctx; if (prog) { + GLuint i; _mesa_bzero(prog, sizeof(*prog)); prog->Id = id; prog->Target = target; prog->Resident = GL_TRUE; prog->RefCount = 1; prog->Format = GL_PROGRAM_FORMAT_ASCII_ARB; + + /* default mapping from samplers to texture units */ + for (i = 0; i < MAX_SAMPLERS; i++) + prog->SamplerUnits[i] = i; } return prog; diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 9f6c54dd4c..bc5ecdaa15 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -1045,6 +1045,28 @@ _mesa_use_program(GLcontext *ctx, GLuint program) } + +/** + * Update the vertex and fragment program's TexturesUsed arrays. + */ +static void +update_textures_used(struct gl_program *prog) +{ + GLuint s; + + memset(prog->TexturesUsed, 0, sizeof(prog->TexturesUsed)); + + for (s = 0; s < MAX_SAMPLERS; s++) { + if (prog->SamplersUsed & (1 << s)) { + GLuint u = prog->SamplerUnits[s]; + GLuint t = prog->SamplerTargets[s]; + assert(u < MAX_TEXTURE_IMAGE_UNITS); + prog->TexturesUsed[u] |= (1 << t); + } + } +} + + /** * Called via ctx->Driver.Uniform(). */ @@ -1067,26 +1089,6 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, FLUSH_VERTICES(ctx, _NEW_PROGRAM); - /* - * If we're setting a sampler, we must use glUniformi1()! - */ - if (shProg->Uniforms->Parameters[location].Type == PROGRAM_SAMPLER) { - GLint unit; - if (type != GL_INT || count != 1) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glUniform(only glUniform1i can be used " - "to set sampler uniforms)"); - return; - } - /* check that the sampler (tex unit index) is legal */ - unit = ((GLint *) values)[0]; - if (unit >= ctx->Const.MaxTextureImageUnits) { - _mesa_error(ctx, GL_INVALID_VALUE, - "glUniform1(invalid sampler/tex unit index)"); - return; - } - } - if (count < 0) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniform(count < 0)"); return; @@ -1119,32 +1121,61 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, return; } - for (k = 0; k < count; k++) { - GLfloat *uniformVal = shProg->Uniforms->ParameterValues[location + k]; - if (type == GL_INT || - type == GL_INT_VEC2 || - type == GL_INT_VEC3 || - type == GL_INT_VEC4) { - const GLint *iValues = ((const GLint *) values) + k * elems; - for (i = 0; i < elems; i++) { - uniformVal[i] = (GLfloat) iValues[i]; - } + if (shProg->Uniforms->Parameters[location].Type == PROGRAM_SAMPLER) { + /* This controls which texture unit which is used by a sampler */ + GLuint texUnit, sampler; + + /* data type for setting samplers must be int */ + if (type != GL_INT || count != 1) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glUniform(only glUniform1i can be used " + "to set sampler uniforms)"); + return; } - else { - const GLfloat *fValues = ((const GLfloat *) values) + k * elems; - for (i = 0; i < elems; i++) { - uniformVal[i] = fValues[i]; - } + + sampler = (GLuint) shProg->Uniforms->ParameterValues[location][0]; + texUnit = ((GLuint *) values)[0]; + + /* check that the sampler (tex unit index) is legal */ + if (texUnit >= ctx->Const.MaxTextureImageUnits) { + _mesa_error(ctx, GL_INVALID_VALUE, + "glUniform1(invalid sampler/tex unit index)"); + return; } - } - if (shProg->Uniforms->Parameters[location].Type == PROGRAM_SAMPLER) { - if (shProg->VertexProgram) - _slang_resolve_samplers(shProg, &shProg->VertexProgram->Base); - if (shProg->FragmentProgram) - _slang_resolve_samplers(shProg, &shProg->FragmentProgram->Base); + if (shProg->VertexProgram) { + shProg->VertexProgram->Base.SamplerUnits[sampler] = texUnit; + update_textures_used(&shProg->VertexProgram->Base); + } + if (shProg->FragmentProgram) { + shProg->FragmentProgram->Base.SamplerUnits[sampler] = texUnit; + update_textures_used(&shProg->FragmentProgram->Base); + } + + FLUSH_VERTICES(ctx, _NEW_TEXTURE); } + else { + /* ordinary uniform variable */ + for (k = 0; k < count; k++) { + GLfloat *uniformVal = shProg->Uniforms->ParameterValues[location + k]; + if (type == GL_INT || + type == GL_INT_VEC2 || + type == GL_INT_VEC3 || + type == GL_INT_VEC4) { + const GLint *iValues = ((const GLint *) values) + k * elems; + for (i = 0; i < elems; i++) { + uniformVal[i] = (GLfloat) iValues[i]; + } + } + else { + const GLfloat *fValues = ((const GLfloat *) values) + k * elems; + for (i = 0; i < elems; i++) { + uniformVal[i] = fValues[i]; + } + } + } + } } diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index fb1f9d5a20..767ba3ffb4 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2874,14 +2874,23 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, const GLint texIndex = sampler_to_texture_index(var->type.specifier.type); if (texIndex != -1) { - /* Texture sampler: + /* This is a texture sampler variable... * store->File = PROGRAM_SAMPLER - * store->Index = sampler uniform location + * store->Index = sampler number (0..7, typically) * store->Size = texture type index (1D, 2D, 3D, cube, etc) */ +#if 0 GLint samplerUniform = _mesa_add_sampler(prog->Parameters, varName, datatype); - store = _slang_new_ir_storage(PROGRAM_SAMPLER, samplerUniform, texIndex); +#elif 0 + GLint samplerUniform + = _mesa_add_sampler(prog->Samplers, varName, datatype); + (void) _mesa_add_sampler(prog->Parameters, varName, datatype); /* dummy entry */ +#else + const GLint sampNum = A->numSamplers++; + _mesa_add_sampler(prog->Parameters, varName, datatype, sampNum); +#endif + store = _slang_new_ir_storage(PROGRAM_SAMPLER, sampNum, texIndex); if (dbg) printf("SAMPLER "); } else if (var->type.qualifier == SLANG_QUAL_UNIFORM) { diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 2be89a5ce0..bfb9ca4db6 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -1618,6 +1618,7 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O, A.program = O->program; A.vartable = O->vartable; A.curFuncEndLabel = NULL; + A.numSamplers = 0; if (!_slang_codegen_global_variable(&A, var, C->type)) return 0; } @@ -1640,6 +1641,7 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O, A.space.funcs = O->funs; A.space.structs = O->structs; A.space.vars = O->vars; + A.numSamplers = 0; if (!initialize_global(&A, var)) return 0; } @@ -1773,6 +1775,7 @@ parse_function(slang_parse_ctx * C, slang_output_ctx * O, int definition, A.program = O->program; A.vartable = O->vartable; A.log = C->L; + A.numSamplers = 0; _slang_codegen_function(&A, *parsed_func_ret); } @@ -2152,6 +2155,9 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) shader->Programs[0]->Parameters = _mesa_new_parameter_list(); shader->Programs[0]->Varying = _mesa_new_parameter_list(); shader->Programs[0]->Attributes = _mesa_new_parameter_list(); +#if 0 + shader->Programs[0]->Samplers = _mesa_new_parameter_list(); +#endif } slang_info_log_construct(&info_log); diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index fe13f2865c..2b08e7020f 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -906,11 +906,15 @@ emit_tex(slang_emit_info *emitInfo, slang_ir_node *n) assert(n->Children[0]->Store->Size >= TEXTURE_1D_INDEX); assert(n->Children[0]->Store->Size <= TEXTURE_RECT_INDEX); - inst->Sampler = n->Children[0]->Store->Index; /* i.e. uniform's index */ inst->TexSrcTarget = n->Children[0]->Store->Size; +#if 0 inst->TexSrcUnit = 27; /* Dummy value; the TexSrcUnit will be computed at * link time, using the sampler uniform's value. */ + inst->Sampler = n->Children[0]->Store->Index; /* i.e. uniform's index */ +#else + inst->TexSrcUnit = n->Children[0]->Store->Index; /* i.e. uniform's index */ +#endif return inst; } diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index eaa29ba094..32f7168553 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -144,10 +144,14 @@ is_uniform(GLuint file) } +static GLuint shProg_NumSamplers = 0; /** XXX temporary */ + + static GLboolean link_uniform_vars(struct gl_shader_program *shProg, struct gl_program *prog) { GLuint *map, i; + GLuint samplerMap[MAX_SAMPLERS]; #if 0 printf("================ pre link uniforms ===============\n"); @@ -168,10 +172,13 @@ link_uniform_vars(struct gl_shader_program *shProg, struct gl_program *prog) /* sanity check */ assert(is_uniform(p->Type)); + /* See if this uniform is already in the linked program's list */ if (p->Name) { + /* this is a named uniform */ j = _mesa_lookup_parameter_index(shProg->Uniforms, -1, p->Name); } else { + /* this is an unnamed constant */ /*GLuint swizzle;*/ ASSERT(p->Type == PROGRAM_CONSTANT); if (_mesa_lookup_parameter_constant(shProg->Uniforms, pVals, @@ -184,7 +191,8 @@ link_uniform_vars(struct gl_shader_program *shProg, struct gl_program *prog) } if (j >= 0) { - /* already in list, check size XXX check this */ + /* already in linked program's list */ + /* check size XXX check this */ #if 0 assert(p->Size == shProg->Uniforms->Parameters[j].Size); #endif @@ -205,7 +213,15 @@ link_uniform_vars(struct gl_shader_program *shProg, struct gl_program *prog) j = _mesa_add_uniform(shProg->Uniforms, p->Name, p->Size, p->DataType); break; case PROGRAM_SAMPLER: - j = _mesa_add_sampler(shProg->Uniforms, p->Name, p->DataType); + { + GLuint sampNum = shProg_NumSamplers++; + GLuint oldSampNum; + j = _mesa_add_sampler(shProg->Uniforms, p->Name, + p->DataType, sampNum); + oldSampNum = (GLuint) prog->Parameters->ParameterValues[i][0]; + assert(oldSampNum < MAX_SAMPLERS); + samplerMap[oldSampNum] = sampNum; + } break; default: _mesa_problem(NULL, "bad parameter type in link_uniform_vars()"); @@ -243,6 +259,7 @@ link_uniform_vars(struct gl_shader_program *shProg, struct gl_program *prog) /* OK, now scan the program/shader instructions looking for uniform vars, * replacing the old index with the new index. */ + prog->SamplersUsed = 0x0; for (i = 0; i < prog->NumInstructions; i++) { struct prog_instruction *inst = prog->Instructions + i; GLuint j; @@ -257,14 +274,15 @@ link_uniform_vars(struct gl_shader_program *shProg, struct gl_program *prog) } } - if (inst->Opcode == OPCODE_TEX || - inst->Opcode == OPCODE_TXB || - inst->Opcode == OPCODE_TXP) { + if (_mesa_is_tex_instruction(inst->Opcode)) { /* printf("====== remap sampler from %d to %d\n", inst->Sampler, map[ inst->Sampler ]); */ - inst->Sampler = map[ inst->Sampler ]; + /* here, texUnit is really samplerUnit */ + inst->TexSrcUnit = samplerMap[inst->TexSrcUnit]; + prog->SamplerTargets[inst->TexSrcUnit] = inst->TexSrcTarget; + prog->SamplersUsed |= (1 << inst->TexSrcUnit); } } @@ -404,36 +422,6 @@ _slang_remap_attribute(struct gl_program *prog, GLuint oldAttrib, GLuint newAttr -/** - * Scan program for texture instructions, lookup sampler/uniform's value - * to determine which texture unit to use. - * Also, update the program's TexturesUsed[] array. - */ -void -_slang_resolve_samplers(struct gl_shader_program *shProg, - struct gl_program *prog) -{ - GLuint i; - - for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++) - prog->TexturesUsed[i] = 0; - - for (i = 0; i < prog->NumInstructions; i++) { - struct prog_instruction *inst = prog->Instructions + i; - if (inst->Opcode == OPCODE_TEX || - inst->Opcode == OPCODE_TXB || - inst->Opcode == OPCODE_TXP) { - GLint sampleUnit = (GLint) shProg->Uniforms->ParameterValues[inst->Sampler][0]; - assert(sampleUnit < MAX_TEXTURE_IMAGE_UNITS); - inst->TexSrcUnit = sampleUnit; - - prog->TexturesUsed[inst->TexSrcUnit] |= (1 << inst->TexSrcTarget); - } - } -} - - - /** cast wrapper */ static struct gl_vertex_program * vertex_program(struct gl_program *prog) @@ -490,6 +478,8 @@ _slang_link(GLcontext *ctx, const struct gl_fragment_program *fragProg; GLuint i; + shProg_NumSamplers = 0; /** XXX temporary */ + _mesa_clear_shader_program_data(ctx, shProg); shProg->Uniforms = _mesa_new_parameter_list(); @@ -549,13 +539,6 @@ _slang_link(GLcontext *ctx, shProg->FragmentProgram->Base.Parameters = shProg->Uniforms; } - if (shProg->VertexProgram) { - _slang_resolve_samplers(shProg, &shProg->VertexProgram->Base); - } - if (shProg->FragmentProgram) { - _slang_resolve_samplers(shProg, &shProg->FragmentProgram->Base); - } - if (shProg->VertexProgram) { if (!_slang_resolve_attributes(shProg, &shProg->VertexProgram->Base)) { /*goto cleanup;*/ @@ -601,7 +584,7 @@ _slang_link(GLcontext *ctx, _mesa_print_program(&fragProg->Base); _mesa_print_program_parameters(ctx, &fragProg->Base); #endif -#if 0 +#if 01 printf("************** linked fragment prog\n"); _mesa_print_program(&shProg->FragmentProgram->Base); _mesa_print_program_parameters(ctx, &shProg->FragmentProgram->Base); diff --git a/src/mesa/shader/slang/slang_link.h b/src/mesa/shader/slang/slang_link.h index 606b9e46b1..8ef8a6b4b3 100644 --- a/src/mesa/shader/slang/slang_link.h +++ b/src/mesa/shader/slang/slang_link.h @@ -32,10 +32,6 @@ extern void _slang_link(GLcontext *ctx, GLhandleARB h, struct gl_shader_program *shProg); -extern void -_slang_resolve_samplers(struct gl_shader_program *shProg, - struct gl_program *prog); - extern void _slang_remap_attribute(struct gl_program *prog, GLuint oldAttrib, GLuint newAttrib); diff --git a/src/mesa/shader/slang/slang_typeinfo.h b/src/mesa/shader/slang/slang_typeinfo.h index 587331e8b1..ad5aa3e195 100644 --- a/src/mesa/shader/slang/slang_typeinfo.h +++ b/src/mesa/shader/slang/slang_typeinfo.h @@ -65,6 +65,7 @@ typedef struct slang_assemble_ctx_ struct slang_label_ *curFuncEndLabel; struct slang_ir_node_ *CurLoop; struct slang_function_ *CurFunction; + GLuint numSamplers; } slang_assemble_ctx; diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 38de35933e..80b8cae013 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -113,6 +113,23 @@ gl_filter_to_img_filter(GLenum filter) } +static struct gl_fragment_program * +current_fragment_program(GLcontext *ctx) +{ + struct gl_fragment_program *f; + + if (ctx->Shader.CurrentProgram && + ctx->Shader.CurrentProgram->LinkStatus && + ctx->Shader.CurrentProgram->FragmentProgram) { + f = ctx->Shader.CurrentProgram->FragmentProgram; + } + else { + f = ctx->FragmentProgram._Current; + assert(f); + } + return f; +} + static void update_samplers(struct st_context *st) @@ -165,13 +182,27 @@ update_samplers(struct st_context *st) st->pipe->bind_sampler_state(st->pipe, u, cso->data); } } + + + /* mapping from sampler vars to texture units */ + { + struct gl_fragment_program *fprog = current_fragment_program(st->ctx); + const GLubyte *samplerUnits = fprog->Base.SamplerUnits; + uint sample_units[PIPE_MAX_SAMPLERS]; + uint s; + for (s = 0; s < PIPE_MAX_SAMPLERS; s++) { + sample_units[s] = fprog->Base.SamplerUnits[s]; + } + + st->pipe->set_sampler_units(st->pipe, PIPE_MAX_SAMPLERS, sample_units); + } } const struct st_tracked_state st_update_sampler = { .name = "st_update_sampler", .dirty = { - .mesa = _NEW_TEXTURE, + .mesa = _NEW_TEXTURE | _NEW_PROGRAM, .st = 0, }, .update = update_samplers diff --git a/src/mesa/swrast/s_fragprog.c b/src/mesa/swrast/s_fragprog.c index 6656ebc0d0..6ee8bfd0a5 100644 --- a/src/mesa/swrast/s_fragprog.c +++ b/src/mesa/swrast/s_fragprog.c @@ -113,6 +113,8 @@ init_machine(GLcontext *ctx, struct gl_program_machine *machine, machine->DerivY = (GLfloat (*)[4]) span->attrStepY; machine->NumDeriv = FRAG_ATTRIB_MAX; + machine->Samplers = program->Base.SamplerUnits; + /* if running a GLSL program (not ARB_fragment_program) */ if (ctx->Shader.CurrentProgram) { /* Store front/back facing value in register FOGC.Y */ -- cgit v1.2.3 From 19710c95da683097372aa54b7538eaa2de778aa5 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 26 Oct 2007 19:29:38 -0600 Subject: turn off debug output --- src/mesa/shader/slang/slang_link.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 32f7168553..390ae56aa5 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -584,7 +584,7 @@ _slang_link(GLcontext *ctx, _mesa_print_program(&fragProg->Base); _mesa_print_program_parameters(ctx, &fragProg->Base); #endif -#if 01 +#if 0 printf("************** linked fragment prog\n"); _mesa_print_program(&shProg->FragmentProgram->Base); _mesa_print_program_parameters(ctx, &shProg->FragmentProgram->Base); -- cgit v1.2.3 From bafbfb4dce7a10fe91c8e0564750e69358ac9461 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 26 Oct 2007 19:31:35 -0600 Subject: Fix up handling of immediate values for TGSI shaders. Still disabled pending LLVM updates. --- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 4 ---- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 24 ++++++++---------------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 2 +- 3 files changed, 9 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index e7eb811d18..f740500dec 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -780,11 +780,7 @@ dump_immediate_short( ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); TXT( " { " ); -#if 0 for( i = 0; i < imm->Immediate.Size - 1; i++ ) { -#else - for( i = 0; i < imm->Immediate.Size; i++ ) { -#endif switch( imm->Immediate.DataType ) { case TGSI_IMM_FLOAT32: FLT( imm->u.ImmediateFloat32[i].Float ); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 66a81b4bd8..f577f004d7 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -170,24 +170,16 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) break; case TGSI_TOKEN_TYPE_IMMEDIATE: -#if 0 - assert( (parse.FullToken.FullImmediate.Immediate.Size - 1) % 4 == 0 ); - assert( mach->ImmLimit + (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4 <= 256 ); + { + uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; + assert( size % 4 == 0 ); + assert( mach->ImmLimit + size / 4 <= 256 ); - for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size - 1; i++ ) { - mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; - } - mach->ImmLimit += (parse.FullToken.FullImmediate.Immediate.Size - 1) / 4; -#else - /* Add this immediate value (vector of 1,2,3,4 floats) to immediates array */ - assert( parse.FullToken.FullImmediate.Immediate.Size <= 4 ); - assert( mach->ImmLimit < TGSI_EXEC_NUM_IMMEDIATES ); - - for( i = 0; i < parse.FullToken.FullImmediate.Immediate.Size; i++ ) { - mach->Imms[mach->ImmLimit][i] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + for( i = 0; i < size; i++ ) { + mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } + mach->ImmLimit += size / 4; } - mach->ImmLimit++; -#endif break; case TGSI_TOKEN_TYPE_INSTRUCTION: diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c index 6ce7ab9b71..2c33f26d48 100644 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c @@ -139,7 +139,7 @@ make_immediate(const float *value, uint size) { struct tgsi_full_immediate imm; imm.Immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; - imm.Immediate.Size = size; + imm.Immediate.Size = 1 + size; /* one for the token itself */ imm.Immediate.DataType = TGSI_IMM_FLOAT32; imm.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; return imm; -- cgit v1.2.3 From 3cf6644c00ad0b265c64645d0b14de9dc90ba851 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:02:40 -0600 Subject: Move mesa_to_tgsi.[ch] into state tracker. --- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c | 796 ------------------------------- src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h | 32 -- src/mesa/state_tracker/st_mesa_to_tgsi.c | 796 +++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_mesa_to_tgsi.h | 32 ++ 4 files changed, 828 insertions(+), 828 deletions(-) delete mode 100644 src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c delete mode 100644 src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h create mode 100644 src/mesa/state_tracker/st_mesa_to_tgsi.c create mode 100644 src/mesa/state_tracker/st_mesa_to_tgsi.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c deleted file mode 100644 index 2c33f26d48..0000000000 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.c +++ /dev/null @@ -1,796 +0,0 @@ -#include "tgsi_platform.h" -#include "tgsi_mesa.h" -#include "pipe/tgsi/mesa/mesa_to_tgsi.h" -#include "shader/prog_parameter.h" - -#define TGSI_DEBUG 0 - -#define EMIT_IMMEDIATES 0 - - -/* - * Map mesa register file to TGSI register file. - */ -static GLuint -map_register_file( - enum register_file file ) -{ - switch( file ) { - case PROGRAM_UNDEFINED: - return TGSI_FILE_NULL; - case PROGRAM_TEMPORARY: - return TGSI_FILE_TEMPORARY; - //case PROGRAM_LOCAL_PARAM: - //case PROGRAM_ENV_PARAM: - case PROGRAM_STATE_VAR: - case PROGRAM_NAMED_PARAM: - case PROGRAM_UNIFORM: - return TGSI_FILE_CONSTANT; - case PROGRAM_CONSTANT: -#if EMIT_IMMEDIATES - return TGSI_FILE_IMMEDIATE; -#else - return TGSI_FILE_CONSTANT; -#endif - case PROGRAM_INPUT: - return TGSI_FILE_INPUT; - case PROGRAM_OUTPUT: - return TGSI_FILE_OUTPUT; - case PROGRAM_ADDRESS: - return TGSI_FILE_ADDRESS; - default: - assert( 0 ); - return TGSI_FILE_NULL; - } -} - -/** - * Map mesa register file index to TGSI index. - * Take special care when processing input and output indices. - * \param file one of TGSI_FILE_x - * \param index the mesa register file index - * \param inputMapping maps Mesa input indexes to TGSI input indexes - * \param outputMapping maps Mesa output indexes to TGSI output indexes - */ -static GLuint -map_register_file_index( - GLuint file, - GLuint index, - const GLuint inputMapping[], - const GLuint outputMapping[], - const GLuint immediateMapping[]) -{ - switch( file ) { - case TGSI_FILE_INPUT: - /* inputs are mapped according to the user-defined map */ - return inputMapping[index]; - - case TGSI_FILE_OUTPUT: - return outputMapping[index]; - -#if EMIT_IMMEDIATES - case TGSI_FILE_IMMEDIATE: - return immediateMapping[index]; -#endif - - default: - return index; - } -} - -/* - * Map mesa texture target to TGSI texture target. - */ -static GLuint -map_texture_target( - GLuint textarget ) -{ - switch( textarget ) { - case TEXTURE_1D_INDEX: - return TGSI_TEXTURE_1D; - case TEXTURE_2D_INDEX: - return TGSI_TEXTURE_2D; - case TEXTURE_3D_INDEX: - return TGSI_TEXTURE_3D; - case TEXTURE_CUBE_INDEX: - return TGSI_TEXTURE_CUBE; - case TEXTURE_RECT_INDEX: - return TGSI_TEXTURE_RECT; - default: - assert( 0 ); - } - - return TGSI_TEXTURE_1D; -} - -static GLuint -convert_sat( - GLuint sat ) -{ - switch( sat ) { - case SATURATE_OFF: - return TGSI_SAT_NONE; - case SATURATE_ZERO_ONE: - return TGSI_SAT_ZERO_ONE; - case SATURATE_PLUS_MINUS_ONE: - return TGSI_SAT_MINUS_PLUS_ONE; - default: - assert( 0 ); - return TGSI_SAT_NONE; - } -} - -static GLuint -convert_writemask( - GLuint writemask ) -{ - assert( WRITEMASK_X == TGSI_WRITEMASK_X ); - assert( WRITEMASK_Y == TGSI_WRITEMASK_Y ); - assert( WRITEMASK_Z == TGSI_WRITEMASK_Z ); - assert( WRITEMASK_W == TGSI_WRITEMASK_W ); - assert( (writemask & ~TGSI_WRITEMASK_XYZW) == 0 ); - - return writemask; -} - -#if EMIT_IMMEDIATES -static struct tgsi_full_immediate -make_immediate(const float *value, uint size) -{ - struct tgsi_full_immediate imm; - imm.Immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; - imm.Immediate.Size = 1 + size; /* one for the token itself */ - imm.Immediate.DataType = TGSI_IMM_FLOAT32; - imm.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; - return imm; -} -#endif - -static void -compile_instruction( - const struct prog_instruction *inst, - struct tgsi_full_instruction *fullinst, - const GLuint inputMapping[], - const GLuint outputMapping[], - const GLuint immediateMapping[], - GLuint preamble_size, - GLuint processor ) -{ - GLuint i; - struct tgsi_full_dst_register *fulldst; - struct tgsi_full_src_register *fullsrc; - - *fullinst = tgsi_default_full_instruction(); - - fullinst->Instruction.Saturate = convert_sat( inst->SaturateMode ); - fullinst->Instruction.NumDstRegs = _mesa_num_inst_dst_regs( inst->Opcode ); - fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode ); - - fulldst = &fullinst->FullDstRegisters[0]; - fulldst->DstRegister.File = map_register_file( inst->DstReg.File ); - fulldst->DstRegister.Index = map_register_file_index( - fulldst->DstRegister.File, - inst->DstReg.Index, - inputMapping, - outputMapping, - NULL - ); - fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); - - for( i = 0; i < fullinst->Instruction.NumSrcRegs; i++ ) { - GLuint j; - - fullsrc = &fullinst->FullSrcRegisters[i]; - fullsrc->SrcRegister.File = map_register_file( inst->SrcReg[i].File ); - fullsrc->SrcRegister.Index = map_register_file_index( - fullsrc->SrcRegister.File, - inst->SrcReg[i].Index, - inputMapping, - outputMapping, - immediateMapping); - - for( j = 0; j < 4; j++ ) { - GLuint swz; - - swz = GET_SWZ( inst->SrcReg[i].Swizzle, j ); - if( swz > SWIZZLE_W ) { - tgsi_util_set_src_register_extswizzle( - &fullsrc->SrcRegisterExtSwz, - swz, - j ); - } - else { - tgsi_util_set_src_register_swizzle( - &fullsrc->SrcRegister, - swz, - j ); - } - } - - if( inst->SrcReg[i].NegateBase == NEGATE_XYZW ) { - fullsrc->SrcRegister.Negate = 1; - } - else if( inst->SrcReg[i].NegateBase != NEGATE_NONE ) { - if( inst->SrcReg[i].NegateBase & NEGATE_X ) { - fullsrc->SrcRegisterExtSwz.NegateX = 1; - } - if( inst->SrcReg[i].NegateBase & NEGATE_Y ) { - fullsrc->SrcRegisterExtSwz.NegateY = 1; - } - if( inst->SrcReg[i].NegateBase & NEGATE_Z ) { - fullsrc->SrcRegisterExtSwz.NegateZ = 1; - } - if( inst->SrcReg[i].NegateBase & NEGATE_W ) { - fullsrc->SrcRegisterExtSwz.NegateW = 1; - } - } - - if( inst->SrcReg[i].Abs ) { - fullsrc->SrcRegisterExtMod.Absolute = 1; - } - - if( inst->SrcReg[i].NegateAbs ) { - fullsrc->SrcRegisterExtMod.Negate = 1; - } - - if( inst->SrcReg[i].RelAddr ) { - fullsrc->SrcRegister.Indirect = 1; - - fullsrc->SrcRegisterInd.File = TGSI_FILE_ADDRESS; - fullsrc->SrcRegisterInd.Index = 0; - } - } - - switch( inst->Opcode ) { - case OPCODE_ARL: - fullinst->Instruction.Opcode = TGSI_OPCODE_ARL; - break; - case OPCODE_ABS: - fullinst->Instruction.Opcode = TGSI_OPCODE_ABS; - break; - case OPCODE_ADD: - fullinst->Instruction.Opcode = TGSI_OPCODE_ADD; - break; - case OPCODE_BGNLOOP: - fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_BGNSUB: - fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB; - break; - case OPCODE_BRA: - fullinst->Instruction.Opcode = TGSI_OPCODE_BRA; - break; - case OPCODE_BRK: - fullinst->Instruction.Opcode = TGSI_OPCODE_BRK; - break; - case OPCODE_CAL: - fullinst->Instruction.Opcode = TGSI_OPCODE_CAL; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_CMP: - fullinst->Instruction.Opcode = TGSI_OPCODE_CMP; - break; - case OPCODE_CONT: - fullinst->Instruction.Opcode = TGSI_OPCODE_CONT; - break; - case OPCODE_COS: - fullinst->Instruction.Opcode = TGSI_OPCODE_COS; - break; - case OPCODE_DDX: - fullinst->Instruction.Opcode = TGSI_OPCODE_DDX; - break; - case OPCODE_DDY: - fullinst->Instruction.Opcode = TGSI_OPCODE_DDY; - break; - case OPCODE_DP3: - fullinst->Instruction.Opcode = TGSI_OPCODE_DP3; - break; - case OPCODE_DP4: - fullinst->Instruction.Opcode = TGSI_OPCODE_DP4; - break; - case OPCODE_DPH: - fullinst->Instruction.Opcode = TGSI_OPCODE_DPH; - break; - case OPCODE_DST: - fullinst->Instruction.Opcode = TGSI_OPCODE_DST; - break; - case OPCODE_ELSE: - fullinst->Instruction.Opcode = TGSI_OPCODE_ELSE; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_ENDIF: - fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF; - break; - case OPCODE_ENDLOOP: - fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_ENDSUB: - fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB; - break; - case OPCODE_EX2: - fullinst->Instruction.Opcode = TGSI_OPCODE_EX2; - break; - case OPCODE_EXP: - fullinst->Instruction.Opcode = TGSI_OPCODE_EXP; - break; - case OPCODE_FLR: - fullinst->Instruction.Opcode = TGSI_OPCODE_FLR; - break; - case OPCODE_FRC: - fullinst->Instruction.Opcode = TGSI_OPCODE_FRC; - break; - case OPCODE_IF: - fullinst->Instruction.Opcode = TGSI_OPCODE_IF; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_INT: - fullinst->Instruction.Opcode = TGSI_OPCODE_INT; - break; - case OPCODE_KIL: - /* predicated w/ a register */ - fullinst->Instruction.Opcode = TGSI_OPCODE_KILP; - break; - case OPCODE_KIL_NV: - /* unpredicated */ - assert(inst->DstReg.CondMask == COND_TR); - fullinst->Instruction.Opcode = TGSI_OPCODE_KIL; - break; - case OPCODE_LG2: - fullinst->Instruction.Opcode = TGSI_OPCODE_LG2; - break; - case OPCODE_LOG: - fullinst->Instruction.Opcode = TGSI_OPCODE_LOG; - break; - case OPCODE_LIT: - fullinst->Instruction.Opcode = TGSI_OPCODE_LIT; - break; - case OPCODE_LRP: - fullinst->Instruction.Opcode = TGSI_OPCODE_LRP; - break; - case OPCODE_MAD: - fullinst->Instruction.Opcode = TGSI_OPCODE_MAD; - break; - case OPCODE_MAX: - fullinst->Instruction.Opcode = TGSI_OPCODE_MAX; - break; - case OPCODE_MIN: - fullinst->Instruction.Opcode = TGSI_OPCODE_MIN; - break; - case OPCODE_MOV: - fullinst->Instruction.Opcode = TGSI_OPCODE_MOV; - break; - case OPCODE_MUL: - fullinst->Instruction.Opcode = TGSI_OPCODE_MUL; - break; - case OPCODE_NOISE1: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE1; - break; - case OPCODE_NOISE2: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE2; - break; - case OPCODE_NOISE3: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE3; - break; - case OPCODE_NOISE4: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE4; - break; - case OPCODE_NOP: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOP; - break; - case OPCODE_POW: - fullinst->Instruction.Opcode = TGSI_OPCODE_POW; - break; - case OPCODE_RCP: - fullinst->Instruction.Opcode = TGSI_OPCODE_RCP; - break; - case OPCODE_RET: - fullinst->Instruction.Opcode = TGSI_OPCODE_RET; - break; - case OPCODE_RSQ: - fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ; - tgsi_util_set_full_src_register_sign_mode( - &fullinst->FullSrcRegisters[0], - TGSI_UTIL_SIGN_CLEAR ); - break; - case OPCODE_SCS: - fullinst->Instruction.Opcode = TGSI_OPCODE_SCS; - fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XY; - break; - case OPCODE_SEQ: - fullinst->Instruction.Opcode = TGSI_OPCODE_SEQ; - break; - case OPCODE_SGE: - fullinst->Instruction.Opcode = TGSI_OPCODE_SGE; - break; - case OPCODE_SGT: - fullinst->Instruction.Opcode = TGSI_OPCODE_SGT; - break; - case OPCODE_SIN: - fullinst->Instruction.Opcode = TGSI_OPCODE_SIN; - break; - case OPCODE_SLE: - fullinst->Instruction.Opcode = TGSI_OPCODE_SLE; - break; - case OPCODE_SLT: - fullinst->Instruction.Opcode = TGSI_OPCODE_SLT; - break; - case OPCODE_SNE: - fullinst->Instruction.Opcode = TGSI_OPCODE_SNE; - break; - case OPCODE_SUB: - fullinst->Instruction.Opcode = TGSI_OPCODE_SUB; - break; - case OPCODE_SWZ: - fullinst->Instruction.Opcode = TGSI_OPCODE_SWZ; - break; - case OPCODE_TEX: - fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; - fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_TXB: - fullinst->Instruction.Opcode = TGSI_OPCODE_TXB; - fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_TXD: - fullinst->Instruction.Opcode = TGSI_OPCODE_TXD; - fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_TXL: - fullinst->Instruction.Opcode = TGSI_OPCODE_TXL; - fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_TXP: - fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; - fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide = TGSI_EXTSWIZZLE_W; - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_XPD: - fullinst->Instruction.Opcode = TGSI_OPCODE_XPD; - fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; - break; - case OPCODE_END: - fullinst->Instruction.Opcode = TGSI_OPCODE_RET; - break; - default: - assert( 0 ); - } -} - -/** - * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens - */ -static struct tgsi_full_declaration -make_input_decl( - GLuint index, - GLuint interpolate, - GLuint usage_mask, - GLboolean semantic_info, - GLuint semantic_name, - GLbitfield semantic_index ) -{ - struct tgsi_full_declaration decl; - - assert(semantic_name < TGSI_SEMANTIC_COUNT); - - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.Declaration.UsageMask = usage_mask; - decl.Declaration.Semantic = semantic_info; - decl.Declaration.Interpolate = 1; - decl.u.DeclarationRange.First = index; - decl.u.DeclarationRange.Last = index; - if (semantic_info) { - decl.Semantic.SemanticName = semantic_name; - decl.Semantic.SemanticIndex = semantic_index; - } - decl.Interpolation.Interpolate = interpolate; - - return decl; -} - -/** - * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens - */ -static struct tgsi_full_declaration -make_output_decl( - GLuint index, - GLuint semantic_name, - GLuint semantic_index, - GLbitfield usage_mask ) -{ - struct tgsi_full_declaration decl; - - assert(semantic_name < TGSI_SEMANTIC_COUNT); - - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.Declaration.UsageMask = usage_mask; - decl.Declaration.Semantic = 1; - decl.u.DeclarationRange.First = index; - decl.u.DeclarationRange.Last = index; - decl.Semantic.SemanticName = semantic_name; - decl.Semantic.SemanticIndex = semantic_index; - - return decl; -} - - -static struct tgsi_full_declaration -make_temp_decl(GLuint index) -{ - struct tgsi_full_declaration decl; - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.u.DeclarationRange.First = index; - decl.u.DeclarationRange.Last = index; - return decl; -} - - -/** - * Find the temporaries which are used in the given program. - * Put the indices of the temporaries in 'tempsUsed'. - * \return number of temporaries used - */ -static GLuint -find_temporaries(const struct gl_program *program, - GLuint tempsUsed[MAX_PROGRAM_TEMPS]) -{ - GLuint i, j, count; - - for (i = 0; i < MAX_PROGRAM_TEMPS; i++) - tempsUsed[i] = GL_FALSE; - - for (i = 0; i < program->NumInstructions; i++) { - const struct prog_instruction *inst = program->Instructions + i; - const GLuint n = _mesa_num_inst_src_regs( inst->Opcode ); - for (j = 0; j < n; j++) { - if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) - tempsUsed[inst->SrcReg[j].Index] = GL_TRUE; - if (inst->DstReg.File == PROGRAM_TEMPORARY) - tempsUsed[inst->DstReg.Index] = GL_TRUE; - } - } - - /* convert flags to list of indices */ - count = 0; - for (i = 0; i < MAX_PROGRAM_TEMPS; i++) { - if (tempsUsed[i]) - tempsUsed[count++] = i; - } - return count; -} - - - - -/** - * Translate Mesa program to TGSI format. - * \param program the program to translate - * \param numInputs number of input registers used - * \param inputMapping maps Mesa fragment program inputs to TGSI generic - * input indexes - * \param inputSemanticName the TGSI_SEMANTIC flag for each input - * \param inputSemanticIndex the semantic index (ex: which texcoord) for each input - * \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input - - * \param numOutputs number of output registers used - * \param outputMapping maps Mesa fragment program outputs to TGSI - * generic outputs - * \param outputSemanticName the TGSI_SEMANTIC flag for each output - * \param outputSemanticIndex the semantic index (ex: which texcoord) for each output - * \param tokens array to store translated tokens in - * \param maxTokens size of the tokens array - * - */ -GLboolean -tgsi_translate_mesa_program( - uint procType, - const struct gl_program *program, - GLuint numInputs, - const GLuint inputMapping[], - const ubyte inputSemanticName[], - const ubyte inputSemanticIndex[], - const GLuint interpMode[], - GLuint numOutputs, - const GLuint outputMapping[], - const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - struct tgsi_token *tokens, - GLuint maxTokens ) -{ - GLuint i; - GLuint ti; /* token index */ - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_instruction fullinst; - GLuint preamble_size = 0; - GLuint immediates[1000]; -#if EMIT_IMMEDIATES - GLuint numImmediates = 0; -#endif - - assert(procType == TGSI_PROCESSOR_FRAGMENT || - procType == TGSI_PROCESSOR_VERTEX); - - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); - - ti = 3; - - /* - * Declare input attributes. - */ - if (procType == TGSI_PROCESSOR_FRAGMENT) { - for (i = 0; i < numInputs; i++) { - struct tgsi_full_declaration fulldecl; - switch (inputSemanticName[i]) { - case TGSI_SEMANTIC_POSITION: - /* Fragment XY pos */ - fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_CONSTANT, - TGSI_WRITEMASK_XY, - GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - /* Fragment ZW pos */ - fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_LINEAR, - TGSI_WRITEMASK_ZW, - GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - default: - fulldecl = make_input_decl(i, - interpMode[i], - TGSI_WRITEMASK_XYZW, - GL_TRUE, inputSemanticName[i], - inputSemanticIndex[i]); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - } - } - } - else { - /* vertex prog */ - for (i = 0; i < numInputs; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_ATTRIB, - TGSI_WRITEMASK_XYZW, - GL_FALSE, inputSemanticName[i], - inputSemanticIndex[i]); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - - /* - * Declare output attributes. - */ - if (procType == TGSI_PROCESSOR_FRAGMENT) { - for (i = 0; i < numOutputs; i++) { - struct tgsi_full_declaration fulldecl; - switch (outputSemanticName[i]) { - case TGSI_SEMANTIC_POSITION: - fulldecl = make_output_decl(i, - TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ - TGSI_WRITEMASK_Z ); - break; - case TGSI_SEMANTIC_COLOR: - fulldecl = make_output_decl(i, - TGSI_SEMANTIC_COLOR, 0, - TGSI_WRITEMASK_XYZW ); - break; - default: - abort(); - } - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - else { - /* vertex prog */ - for (i = 0; i < numOutputs; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_output_decl(i, - outputSemanticName[i], - outputSemanticIndex[i], - TGSI_WRITEMASK_XYZW ); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - - /* temporary decls */ - { - GLuint tempsUsed[MAX_PROGRAM_TEMPS]; - uint numTemps = find_temporaries(program, tempsUsed); - for (i = 0; i < numTemps; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_temp_decl(tempsUsed[i]); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - - /* immediates/literals */ -#if EMIT_IMMEDIATES - for (i = 0; i < program->Parameters->NumParameters; i++) { - if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { - struct tgsi_full_immediate fullimm - = make_immediate(program->Parameters->ParameterValues[i], - program->Parameters->Parameters[i].Size); - ti += tgsi_build_full_immediate(&fullimm, - &tokens[ti], - header, - maxTokens - ti); - immediates[i] = numImmediates; - numImmediates++; - } - } -#endif - - for( i = 0; i < program->NumInstructions; i++ ) { - compile_instruction( - &program->Instructions[i], - &fullinst, - inputMapping, - outputMapping, - immediates, - preamble_size, - procType ); - - ti += tgsi_build_full_instruction( - &fullinst, - &tokens[ti], - header, - maxTokens - ti ); - } - - return GL_TRUE; -} - diff --git a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h b/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h deleted file mode 100644 index 13372d75fd..0000000000 --- a/src/mesa/pipe/tgsi/mesa/mesa_to_tgsi.h +++ /dev/null @@ -1,32 +0,0 @@ -#if !defined MESA_TO_TGSI_H -#define MESA_TO_TGSI_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_token; - -GLboolean -tgsi_translate_mesa_program( - uint procType, - const struct gl_program *program, - GLuint numInputs, - const GLuint inputMapping[], - const ubyte inputSemanticName[], - const ubyte inputSemanticIndex[], - const GLuint interpMode[], - GLuint numOutputs, - const GLuint outputMapping[], - const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - struct tgsi_token *tokens, - GLuint maxTokens ); - - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined MESA_TO_TGSI_H - diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c new file mode 100644 index 0000000000..2c33f26d48 --- /dev/null +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -0,0 +1,796 @@ +#include "tgsi_platform.h" +#include "tgsi_mesa.h" +#include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "shader/prog_parameter.h" + +#define TGSI_DEBUG 0 + +#define EMIT_IMMEDIATES 0 + + +/* + * Map mesa register file to TGSI register file. + */ +static GLuint +map_register_file( + enum register_file file ) +{ + switch( file ) { + case PROGRAM_UNDEFINED: + return TGSI_FILE_NULL; + case PROGRAM_TEMPORARY: + return TGSI_FILE_TEMPORARY; + //case PROGRAM_LOCAL_PARAM: + //case PROGRAM_ENV_PARAM: + case PROGRAM_STATE_VAR: + case PROGRAM_NAMED_PARAM: + case PROGRAM_UNIFORM: + return TGSI_FILE_CONSTANT; + case PROGRAM_CONSTANT: +#if EMIT_IMMEDIATES + return TGSI_FILE_IMMEDIATE; +#else + return TGSI_FILE_CONSTANT; +#endif + case PROGRAM_INPUT: + return TGSI_FILE_INPUT; + case PROGRAM_OUTPUT: + return TGSI_FILE_OUTPUT; + case PROGRAM_ADDRESS: + return TGSI_FILE_ADDRESS; + default: + assert( 0 ); + return TGSI_FILE_NULL; + } +} + +/** + * Map mesa register file index to TGSI index. + * Take special care when processing input and output indices. + * \param file one of TGSI_FILE_x + * \param index the mesa register file index + * \param inputMapping maps Mesa input indexes to TGSI input indexes + * \param outputMapping maps Mesa output indexes to TGSI output indexes + */ +static GLuint +map_register_file_index( + GLuint file, + GLuint index, + const GLuint inputMapping[], + const GLuint outputMapping[], + const GLuint immediateMapping[]) +{ + switch( file ) { + case TGSI_FILE_INPUT: + /* inputs are mapped according to the user-defined map */ + return inputMapping[index]; + + case TGSI_FILE_OUTPUT: + return outputMapping[index]; + +#if EMIT_IMMEDIATES + case TGSI_FILE_IMMEDIATE: + return immediateMapping[index]; +#endif + + default: + return index; + } +} + +/* + * Map mesa texture target to TGSI texture target. + */ +static GLuint +map_texture_target( + GLuint textarget ) +{ + switch( textarget ) { + case TEXTURE_1D_INDEX: + return TGSI_TEXTURE_1D; + case TEXTURE_2D_INDEX: + return TGSI_TEXTURE_2D; + case TEXTURE_3D_INDEX: + return TGSI_TEXTURE_3D; + case TEXTURE_CUBE_INDEX: + return TGSI_TEXTURE_CUBE; + case TEXTURE_RECT_INDEX: + return TGSI_TEXTURE_RECT; + default: + assert( 0 ); + } + + return TGSI_TEXTURE_1D; +} + +static GLuint +convert_sat( + GLuint sat ) +{ + switch( sat ) { + case SATURATE_OFF: + return TGSI_SAT_NONE; + case SATURATE_ZERO_ONE: + return TGSI_SAT_ZERO_ONE; + case SATURATE_PLUS_MINUS_ONE: + return TGSI_SAT_MINUS_PLUS_ONE; + default: + assert( 0 ); + return TGSI_SAT_NONE; + } +} + +static GLuint +convert_writemask( + GLuint writemask ) +{ + assert( WRITEMASK_X == TGSI_WRITEMASK_X ); + assert( WRITEMASK_Y == TGSI_WRITEMASK_Y ); + assert( WRITEMASK_Z == TGSI_WRITEMASK_Z ); + assert( WRITEMASK_W == TGSI_WRITEMASK_W ); + assert( (writemask & ~TGSI_WRITEMASK_XYZW) == 0 ); + + return writemask; +} + +#if EMIT_IMMEDIATES +static struct tgsi_full_immediate +make_immediate(const float *value, uint size) +{ + struct tgsi_full_immediate imm; + imm.Immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + imm.Immediate.Size = 1 + size; /* one for the token itself */ + imm.Immediate.DataType = TGSI_IMM_FLOAT32; + imm.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; + return imm; +} +#endif + +static void +compile_instruction( + const struct prog_instruction *inst, + struct tgsi_full_instruction *fullinst, + const GLuint inputMapping[], + const GLuint outputMapping[], + const GLuint immediateMapping[], + GLuint preamble_size, + GLuint processor ) +{ + GLuint i; + struct tgsi_full_dst_register *fulldst; + struct tgsi_full_src_register *fullsrc; + + *fullinst = tgsi_default_full_instruction(); + + fullinst->Instruction.Saturate = convert_sat( inst->SaturateMode ); + fullinst->Instruction.NumDstRegs = _mesa_num_inst_dst_regs( inst->Opcode ); + fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode ); + + fulldst = &fullinst->FullDstRegisters[0]; + fulldst->DstRegister.File = map_register_file( inst->DstReg.File ); + fulldst->DstRegister.Index = map_register_file_index( + fulldst->DstRegister.File, + inst->DstReg.Index, + inputMapping, + outputMapping, + NULL + ); + fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); + + for( i = 0; i < fullinst->Instruction.NumSrcRegs; i++ ) { + GLuint j; + + fullsrc = &fullinst->FullSrcRegisters[i]; + fullsrc->SrcRegister.File = map_register_file( inst->SrcReg[i].File ); + fullsrc->SrcRegister.Index = map_register_file_index( + fullsrc->SrcRegister.File, + inst->SrcReg[i].Index, + inputMapping, + outputMapping, + immediateMapping); + + for( j = 0; j < 4; j++ ) { + GLuint swz; + + swz = GET_SWZ( inst->SrcReg[i].Swizzle, j ); + if( swz > SWIZZLE_W ) { + tgsi_util_set_src_register_extswizzle( + &fullsrc->SrcRegisterExtSwz, + swz, + j ); + } + else { + tgsi_util_set_src_register_swizzle( + &fullsrc->SrcRegister, + swz, + j ); + } + } + + if( inst->SrcReg[i].NegateBase == NEGATE_XYZW ) { + fullsrc->SrcRegister.Negate = 1; + } + else if( inst->SrcReg[i].NegateBase != NEGATE_NONE ) { + if( inst->SrcReg[i].NegateBase & NEGATE_X ) { + fullsrc->SrcRegisterExtSwz.NegateX = 1; + } + if( inst->SrcReg[i].NegateBase & NEGATE_Y ) { + fullsrc->SrcRegisterExtSwz.NegateY = 1; + } + if( inst->SrcReg[i].NegateBase & NEGATE_Z ) { + fullsrc->SrcRegisterExtSwz.NegateZ = 1; + } + if( inst->SrcReg[i].NegateBase & NEGATE_W ) { + fullsrc->SrcRegisterExtSwz.NegateW = 1; + } + } + + if( inst->SrcReg[i].Abs ) { + fullsrc->SrcRegisterExtMod.Absolute = 1; + } + + if( inst->SrcReg[i].NegateAbs ) { + fullsrc->SrcRegisterExtMod.Negate = 1; + } + + if( inst->SrcReg[i].RelAddr ) { + fullsrc->SrcRegister.Indirect = 1; + + fullsrc->SrcRegisterInd.File = TGSI_FILE_ADDRESS; + fullsrc->SrcRegisterInd.Index = 0; + } + } + + switch( inst->Opcode ) { + case OPCODE_ARL: + fullinst->Instruction.Opcode = TGSI_OPCODE_ARL; + break; + case OPCODE_ABS: + fullinst->Instruction.Opcode = TGSI_OPCODE_ABS; + break; + case OPCODE_ADD: + fullinst->Instruction.Opcode = TGSI_OPCODE_ADD; + break; + case OPCODE_BGNLOOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_BGNSUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB; + break; + case OPCODE_BRA: + fullinst->Instruction.Opcode = TGSI_OPCODE_BRA; + break; + case OPCODE_BRK: + fullinst->Instruction.Opcode = TGSI_OPCODE_BRK; + break; + case OPCODE_CAL: + fullinst->Instruction.Opcode = TGSI_OPCODE_CAL; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_CMP: + fullinst->Instruction.Opcode = TGSI_OPCODE_CMP; + break; + case OPCODE_CONT: + fullinst->Instruction.Opcode = TGSI_OPCODE_CONT; + break; + case OPCODE_COS: + fullinst->Instruction.Opcode = TGSI_OPCODE_COS; + break; + case OPCODE_DDX: + fullinst->Instruction.Opcode = TGSI_OPCODE_DDX; + break; + case OPCODE_DDY: + fullinst->Instruction.Opcode = TGSI_OPCODE_DDY; + break; + case OPCODE_DP3: + fullinst->Instruction.Opcode = TGSI_OPCODE_DP3; + break; + case OPCODE_DP4: + fullinst->Instruction.Opcode = TGSI_OPCODE_DP4; + break; + case OPCODE_DPH: + fullinst->Instruction.Opcode = TGSI_OPCODE_DPH; + break; + case OPCODE_DST: + fullinst->Instruction.Opcode = TGSI_OPCODE_DST; + break; + case OPCODE_ELSE: + fullinst->Instruction.Opcode = TGSI_OPCODE_ELSE; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_ENDIF: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF; + break; + case OPCODE_ENDLOOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_ENDSUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB; + break; + case OPCODE_EX2: + fullinst->Instruction.Opcode = TGSI_OPCODE_EX2; + break; + case OPCODE_EXP: + fullinst->Instruction.Opcode = TGSI_OPCODE_EXP; + break; + case OPCODE_FLR: + fullinst->Instruction.Opcode = TGSI_OPCODE_FLR; + break; + case OPCODE_FRC: + fullinst->Instruction.Opcode = TGSI_OPCODE_FRC; + break; + case OPCODE_IF: + fullinst->Instruction.Opcode = TGSI_OPCODE_IF; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_INT: + fullinst->Instruction.Opcode = TGSI_OPCODE_INT; + break; + case OPCODE_KIL: + /* predicated w/ a register */ + fullinst->Instruction.Opcode = TGSI_OPCODE_KILP; + break; + case OPCODE_KIL_NV: + /* unpredicated */ + assert(inst->DstReg.CondMask == COND_TR); + fullinst->Instruction.Opcode = TGSI_OPCODE_KIL; + break; + case OPCODE_LG2: + fullinst->Instruction.Opcode = TGSI_OPCODE_LG2; + break; + case OPCODE_LOG: + fullinst->Instruction.Opcode = TGSI_OPCODE_LOG; + break; + case OPCODE_LIT: + fullinst->Instruction.Opcode = TGSI_OPCODE_LIT; + break; + case OPCODE_LRP: + fullinst->Instruction.Opcode = TGSI_OPCODE_LRP; + break; + case OPCODE_MAD: + fullinst->Instruction.Opcode = TGSI_OPCODE_MAD; + break; + case OPCODE_MAX: + fullinst->Instruction.Opcode = TGSI_OPCODE_MAX; + break; + case OPCODE_MIN: + fullinst->Instruction.Opcode = TGSI_OPCODE_MIN; + break; + case OPCODE_MOV: + fullinst->Instruction.Opcode = TGSI_OPCODE_MOV; + break; + case OPCODE_MUL: + fullinst->Instruction.Opcode = TGSI_OPCODE_MUL; + break; + case OPCODE_NOISE1: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE1; + break; + case OPCODE_NOISE2: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE2; + break; + case OPCODE_NOISE3: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE3; + break; + case OPCODE_NOISE4: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE4; + break; + case OPCODE_NOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOP; + break; + case OPCODE_POW: + fullinst->Instruction.Opcode = TGSI_OPCODE_POW; + break; + case OPCODE_RCP: + fullinst->Instruction.Opcode = TGSI_OPCODE_RCP; + break; + case OPCODE_RET: + fullinst->Instruction.Opcode = TGSI_OPCODE_RET; + break; + case OPCODE_RSQ: + fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ; + tgsi_util_set_full_src_register_sign_mode( + &fullinst->FullSrcRegisters[0], + TGSI_UTIL_SIGN_CLEAR ); + break; + case OPCODE_SCS: + fullinst->Instruction.Opcode = TGSI_OPCODE_SCS; + fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XY; + break; + case OPCODE_SEQ: + fullinst->Instruction.Opcode = TGSI_OPCODE_SEQ; + break; + case OPCODE_SGE: + fullinst->Instruction.Opcode = TGSI_OPCODE_SGE; + break; + case OPCODE_SGT: + fullinst->Instruction.Opcode = TGSI_OPCODE_SGT; + break; + case OPCODE_SIN: + fullinst->Instruction.Opcode = TGSI_OPCODE_SIN; + break; + case OPCODE_SLE: + fullinst->Instruction.Opcode = TGSI_OPCODE_SLE; + break; + case OPCODE_SLT: + fullinst->Instruction.Opcode = TGSI_OPCODE_SLT; + break; + case OPCODE_SNE: + fullinst->Instruction.Opcode = TGSI_OPCODE_SNE; + break; + case OPCODE_SUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_SUB; + break; + case OPCODE_SWZ: + fullinst->Instruction.Opcode = TGSI_OPCODE_SWZ; + break; + case OPCODE_TEX: + fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXB: + fullinst->Instruction.Opcode = TGSI_OPCODE_TXB; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXD: + fullinst->Instruction.Opcode = TGSI_OPCODE_TXD; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXL: + fullinst->Instruction.Opcode = TGSI_OPCODE_TXL; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXP: + fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide = TGSI_EXTSWIZZLE_W; + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_XPD: + fullinst->Instruction.Opcode = TGSI_OPCODE_XPD; + fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; + break; + case OPCODE_END: + fullinst->Instruction.Opcode = TGSI_OPCODE_RET; + break; + default: + assert( 0 ); + } +} + +/** + * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens + */ +static struct tgsi_full_declaration +make_input_decl( + GLuint index, + GLuint interpolate, + GLuint usage_mask, + GLboolean semantic_info, + GLuint semantic_name, + GLbitfield semantic_index ) +{ + struct tgsi_full_declaration decl; + + assert(semantic_name < TGSI_SEMANTIC_COUNT); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.Declaration.UsageMask = usage_mask; + decl.Declaration.Semantic = semantic_info; + decl.Declaration.Interpolate = 1; + decl.u.DeclarationRange.First = index; + decl.u.DeclarationRange.Last = index; + if (semantic_info) { + decl.Semantic.SemanticName = semantic_name; + decl.Semantic.SemanticIndex = semantic_index; + } + decl.Interpolation.Interpolate = interpolate; + + return decl; +} + +/** + * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens + */ +static struct tgsi_full_declaration +make_output_decl( + GLuint index, + GLuint semantic_name, + GLuint semantic_index, + GLbitfield usage_mask ) +{ + struct tgsi_full_declaration decl; + + assert(semantic_name < TGSI_SEMANTIC_COUNT); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.Declaration.UsageMask = usage_mask; + decl.Declaration.Semantic = 1; + decl.u.DeclarationRange.First = index; + decl.u.DeclarationRange.Last = index; + decl.Semantic.SemanticName = semantic_name; + decl.Semantic.SemanticIndex = semantic_index; + + return decl; +} + + +static struct tgsi_full_declaration +make_temp_decl(GLuint index) +{ + struct tgsi_full_declaration decl; + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.u.DeclarationRange.First = index; + decl.u.DeclarationRange.Last = index; + return decl; +} + + +/** + * Find the temporaries which are used in the given program. + * Put the indices of the temporaries in 'tempsUsed'. + * \return number of temporaries used + */ +static GLuint +find_temporaries(const struct gl_program *program, + GLuint tempsUsed[MAX_PROGRAM_TEMPS]) +{ + GLuint i, j, count; + + for (i = 0; i < MAX_PROGRAM_TEMPS; i++) + tempsUsed[i] = GL_FALSE; + + for (i = 0; i < program->NumInstructions; i++) { + const struct prog_instruction *inst = program->Instructions + i; + const GLuint n = _mesa_num_inst_src_regs( inst->Opcode ); + for (j = 0; j < n; j++) { + if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) + tempsUsed[inst->SrcReg[j].Index] = GL_TRUE; + if (inst->DstReg.File == PROGRAM_TEMPORARY) + tempsUsed[inst->DstReg.Index] = GL_TRUE; + } + } + + /* convert flags to list of indices */ + count = 0; + for (i = 0; i < MAX_PROGRAM_TEMPS; i++) { + if (tempsUsed[i]) + tempsUsed[count++] = i; + } + return count; +} + + + + +/** + * Translate Mesa program to TGSI format. + * \param program the program to translate + * \param numInputs number of input registers used + * \param inputMapping maps Mesa fragment program inputs to TGSI generic + * input indexes + * \param inputSemanticName the TGSI_SEMANTIC flag for each input + * \param inputSemanticIndex the semantic index (ex: which texcoord) for each input + * \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input + + * \param numOutputs number of output registers used + * \param outputMapping maps Mesa fragment program outputs to TGSI + * generic outputs + * \param outputSemanticName the TGSI_SEMANTIC flag for each output + * \param outputSemanticIndex the semantic index (ex: which texcoord) for each output + * \param tokens array to store translated tokens in + * \param maxTokens size of the tokens array + * + */ +GLboolean +tgsi_translate_mesa_program( + uint procType, + const struct gl_program *program, + GLuint numInputs, + const GLuint inputMapping[], + const ubyte inputSemanticName[], + const ubyte inputSemanticIndex[], + const GLuint interpMode[], + GLuint numOutputs, + const GLuint outputMapping[], + const ubyte outputSemanticName[], + const ubyte outputSemanticIndex[], + struct tgsi_token *tokens, + GLuint maxTokens ) +{ + GLuint i; + GLuint ti; /* token index */ + struct tgsi_header *header; + struct tgsi_processor *processor; + struct tgsi_full_instruction fullinst; + GLuint preamble_size = 0; + GLuint immediates[1000]; +#if EMIT_IMMEDIATES + GLuint numImmediates = 0; +#endif + + assert(procType == TGSI_PROCESSOR_FRAGMENT || + procType == TGSI_PROCESSOR_VERTEX); + + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header(); + + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor( procType, header ); + + ti = 3; + + /* + * Declare input attributes. + */ + if (procType == TGSI_PROCESSOR_FRAGMENT) { + for (i = 0; i < numInputs; i++) { + struct tgsi_full_declaration fulldecl; + switch (inputSemanticName[i]) { + case TGSI_SEMANTIC_POSITION: + /* Fragment XY pos */ + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_CONSTANT, + TGSI_WRITEMASK_XY, + GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + /* Fragment ZW pos */ + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_LINEAR, + TGSI_WRITEMASK_ZW, + GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + default: + fulldecl = make_input_decl(i, + interpMode[i], + TGSI_WRITEMASK_XYZW, + GL_TRUE, inputSemanticName[i], + inputSemanticIndex[i]); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + } + } + } + else { + /* vertex prog */ + for (i = 0; i < numInputs; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_ATTRIB, + TGSI_WRITEMASK_XYZW, + GL_FALSE, inputSemanticName[i], + inputSemanticIndex[i]); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + + /* + * Declare output attributes. + */ + if (procType == TGSI_PROCESSOR_FRAGMENT) { + for (i = 0; i < numOutputs; i++) { + struct tgsi_full_declaration fulldecl; + switch (outputSemanticName[i]) { + case TGSI_SEMANTIC_POSITION: + fulldecl = make_output_decl(i, + TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ + TGSI_WRITEMASK_Z ); + break; + case TGSI_SEMANTIC_COLOR: + fulldecl = make_output_decl(i, + TGSI_SEMANTIC_COLOR, 0, + TGSI_WRITEMASK_XYZW ); + break; + default: + abort(); + } + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + else { + /* vertex prog */ + for (i = 0; i < numOutputs; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_output_decl(i, + outputSemanticName[i], + outputSemanticIndex[i], + TGSI_WRITEMASK_XYZW ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + + /* temporary decls */ + { + GLuint tempsUsed[MAX_PROGRAM_TEMPS]; + uint numTemps = find_temporaries(program, tempsUsed); + for (i = 0; i < numTemps; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_temp_decl(tempsUsed[i]); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + + /* immediates/literals */ +#if EMIT_IMMEDIATES + for (i = 0; i < program->Parameters->NumParameters; i++) { + if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { + struct tgsi_full_immediate fullimm + = make_immediate(program->Parameters->ParameterValues[i], + program->Parameters->Parameters[i].Size); + ti += tgsi_build_full_immediate(&fullimm, + &tokens[ti], + header, + maxTokens - ti); + immediates[i] = numImmediates; + numImmediates++; + } + } +#endif + + for( i = 0; i < program->NumInstructions; i++ ) { + compile_instruction( + &program->Instructions[i], + &fullinst, + inputMapping, + outputMapping, + immediates, + preamble_size, + procType ); + + ti += tgsi_build_full_instruction( + &fullinst, + &tokens[ti], + header, + maxTokens - ti ); + } + + return GL_TRUE; +} + diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h new file mode 100644 index 0000000000..13372d75fd --- /dev/null +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -0,0 +1,32 @@ +#if !defined MESA_TO_TGSI_H +#define MESA_TO_TGSI_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_token; + +GLboolean +tgsi_translate_mesa_program( + uint procType, + const struct gl_program *program, + GLuint numInputs, + const GLuint inputMapping[], + const ubyte inputSemanticName[], + const ubyte inputSemanticIndex[], + const GLuint interpMode[], + GLuint numOutputs, + const GLuint outputMapping[], + const ubyte outputSemanticName[], + const ubyte outputSemanticIndex[], + struct tgsi_token *tokens, + GLuint maxTokens ); + + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined MESA_TO_TGSI_H + -- cgit v1.2.3 From ef6940f17220f1149dce6daf548bd0103d91a281 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:03:15 -0600 Subject: Move mesa_to_tgsi.[ch] to state_tracker --- src/mesa/sources | 4 +--- src/mesa/state_tracker/st_atom_shader.c | 2 +- src/mesa/state_tracker/st_cb_clear.c | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_cb_program.c | 1 - src/mesa/state_tracker/st_mesa_to_tgsi.c | 37 ++++++++++++++++++++++++++-- src/mesa/state_tracker/st_mesa_to_tgsi.h | 40 ++++++++++++++++++++++++++----- src/mesa/state_tracker/st_program.c | 2 +- 8 files changed, 74 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 6d1ba9b9bd..30d3c32b83 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -186,8 +186,6 @@ TGSIEXEC_SOURCES = \ TGSIDECO_SOURCES = \ pipe/tgsi/deco/deco_caps.c -TGSIMESA_SOURCES = \ - pipe/tgsi/mesa/mesa_to_tgsi.c ifeq ($(MESA_LLVM),1) LLVMTGSI_SOURCES = \ @@ -235,6 +233,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_debug.c \ state_tracker/st_draw.c \ state_tracker/st_format.c \ + state_tracker/st_mesa_to_tgsi.c \ state_tracker/st_program.c \ state_tracker/st_mipmap_tree.c @@ -389,7 +388,6 @@ SOLO_SOURCES = \ $(DRAW_SOURCES) \ $(TGSIEXEC_SOURCES) \ $(TGSIDECO_SOURCES) \ - $(TGSIMESA_SOURCES) \ $(STATECACHE_SOURCES) \ $(STATETRACKER_SOURCES) \ $(TNL_SOURCES) \ diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 919461bcb8..92ca22851e 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -41,7 +41,6 @@ #include "main/mtypes.h" #include "pipe/p_context.h" -#include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/exec/tgsi_core.h" #include "st_context.h" @@ -49,6 +48,7 @@ #include "st_atom.h" #include "st_program.h" #include "st_atom_shader.h" +#include "st_mesa_to_tgsi.h" /** diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 30672e0dd4..2c79e2890d 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -43,13 +43,13 @@ #include "st_draw.h" #include "st_program.h" #include "st_public.h" +#include "st_mesa_to_tgsi.h" #include "pipe/p_context.h" #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/tgsi/mesa/mesa_to_tgsi.h" diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 5d8890e022..896ba2b905 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -45,10 +45,10 @@ #include "st_cb_texture.h" #include "st_draw.h" #include "st_format.h" +#include "st_mesa_to_tgsi.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "shader/prog_instruction.h" diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 26609e9645..a330c1c922 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -43,7 +43,6 @@ #include "st_program.h" #include "st_atom_shader.h" -#include "pipe/tgsi/mesa/tgsi_mesa.h" /** diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 2c33f26d48..131e50bece 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -1,6 +1,39 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * \author + * Michal Krol + */ + + #include "tgsi_platform.h" -#include "tgsi_mesa.h" -#include "pipe/tgsi/mesa/mesa_to_tgsi.h" +#include "pipe/tgsi/exec/tgsi_core.h" +#include "st_mesa_to_tgsi.h" #include "shader/prog_parameter.h" #define TGSI_DEBUG 0 diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index 13372d75fd..941a75ab05 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -1,9 +1,37 @@ -#if !defined MESA_TO_TGSI_H -#define MESA_TO_TGSI_H +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_MESA_TO_TGSI_H +#define ST_MESA_TO_TGSI_H #if defined __cplusplus extern "C" { -#endif // defined __cplusplus +#endif struct tgsi_token; @@ -25,8 +53,8 @@ tgsi_translate_mesa_program( #if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus +} /* extern "C" */ +#endif -#endif // !defined MESA_TO_TGSI_H +#endif /* ST_MESA_TO_TGSI_H */ diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 9fc0798ec5..8c61815b9b 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -37,7 +37,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/draw/draw_context.h" -#include "pipe/tgsi/mesa/mesa_to_tgsi.h" #include "pipe/tgsi/exec/tgsi_core.h" #include "pipe/llvm/llvmtgsi.h" @@ -45,6 +44,7 @@ #include "st_cache.h" #include "st_atom.h" #include "st_program.h" +#include "st_mesa_to_tgsi.h" #define TGSI_DEBUG 0 -- cgit v1.2.3 From 4f24568dc7d7cc0de56928b99684b602091e4218 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:04:43 -0600 Subject: Obsolete --- src/mesa/pipe/tgsi/mesa/Makefile | 3 --- src/mesa/pipe/tgsi/mesa/tgsi_mesa.h | 8 -------- 2 files changed, 11 deletions(-) delete mode 100644 src/mesa/pipe/tgsi/mesa/Makefile delete mode 100644 src/mesa/pipe/tgsi/mesa/tgsi_mesa.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/mesa/Makefile b/src/mesa/pipe/tgsi/mesa/Makefile deleted file mode 100644 index eb8b14e0e8..0000000000 --- a/src/mesa/pipe/tgsi/mesa/Makefile +++ /dev/null @@ -1,3 +0,0 @@ -default: - cd ../../.. ; make - diff --git a/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h b/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h deleted file mode 100644 index a4e16e708e..0000000000 --- a/src/mesa/pipe/tgsi/mesa/tgsi_mesa.h +++ /dev/null @@ -1,8 +0,0 @@ -#if !defined TGSI_MESA_H -#define TGSI_MESA_H - -#include "../exec/tgsi_core.h" -#include "mesa_to_tgsi.h" - -#endif // !defined TGSI_MESA_H - -- cgit v1.2.3 From f6a73c3f2815c4c84563c186bba6c8e67bb42ae9 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:26:59 -0600 Subject: Remove remnants of softpipe_surface. This is the last of the clean-up following the change which moved surface allocation to the winsys layer. --- src/mesa/drivers/x11/xm_surface.c | 43 +++-------------- src/mesa/pipe/softpipe/sp_clear.c | 7 ++- src/mesa/pipe/softpipe/sp_context.h | 1 - src/mesa/pipe/softpipe/sp_quad_depth_test.c | 6 +-- src/mesa/pipe/softpipe/sp_quad_stencil.c | 8 +-- src/mesa/pipe/softpipe/sp_state_surface.c | 46 +++++++++--------- src/mesa/pipe/softpipe/sp_surface.c | 75 ----------------------------- src/mesa/pipe/softpipe/sp_surface.h | 27 ----------- src/mesa/pipe/softpipe/sp_tile_cache.c | 12 ++--- src/mesa/pipe/softpipe/sp_tile_cache.h | 4 +- 10 files changed, 47 insertions(+), 182 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 3655a55e4e..dbbf26e33f 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -27,7 +27,7 @@ * \file xm_surface.c * Code to allow the softpipe code to write to X windows/buffers. * This is a bit of a hack for now. We've basically got two different - * abstractions for color buffers: gl_renderbuffer and softpipe_surface. + * abstractions for color buffers: gl_renderbuffer and pipe_surface. * They'll need to get merged someday... * For now, they're separate things that point to each other. */ @@ -61,13 +61,6 @@ } while(0) -static INLINE struct xmesa_surface * -xmesa_surf(struct softpipe_surface *sps) -{ - return (struct xmesa_surface *) sps; -} - - static INLINE struct xmesa_surface * xmesa_surface(struct pipe_surface *ps) { @@ -76,9 +69,9 @@ xmesa_surface(struct pipe_surface *ps) static INLINE struct xmesa_renderbuffer * -xmesa_rb(struct softpipe_surface *sps) +xmesa_rb(struct pipe_surface *ps) { - struct xmesa_surface *xms = xmesa_surf(sps); + struct xmesa_surface *xms = xmesa_surface(ps); return xms->xrb; } @@ -170,24 +163,6 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) xms->surface.format = pipeFormat; xms->surface.refcount = 1; -#if 0 - /* some of the functions plugged in by this call will get overridden */ - softpipe_init_surface_funcs(&xms->surface); -#endif - -#if 0 - switch (pipeFormat) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - xms->surface.get_tile = get_tile; - xms->surface.put_tile = put_tile; - break; - case PIPE_FORMAT_S8_Z24: - break; - default: - abort(); - } -#endif - /* Note, the region we allocate doesn't actually have any storage * since we're drawing into an XImage or Pixmap. * The region's size will get set in the xmesa_alloc_front/back_storage() @@ -215,12 +190,7 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) xms->surface.format = pipeFormat; xms->surface.refcount = 1; -#if 0 - /* - * This is really just a softpipe surface, not an XImage/Pixmap surface. - */ - softpipe_init_surface_funcs(&xms->surface); -#endif + return &xms->surface; } @@ -246,7 +216,7 @@ xmesa_supported_formats(struct pipe_context *pipe, GLuint *numFormats) void xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) { - struct xmesa_renderbuffer *xrb = xmesa_rb((struct softpipe_surface *) ps); + struct xmesa_renderbuffer *xrb = xmesa_rb(ps); /* XXX actually, we should just discard any cached tiles from this * surface since we don't want to accidentally re-use them after clearing. @@ -255,8 +225,7 @@ xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) { struct softpipe_context *sp = softpipe_context(pipe); - struct softpipe_surface *sps = softpipe_surface(ps); - if (sps == sp_tile_cache_get_surface(sp->cbuf_cache[0])) { + if (ps == sp_tile_cache_get_surface(sp->cbuf_cache[0])) { float clear[4]; clear[0] = 0.2; /* XXX hack */ clear[1] = 0.2; diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 08b87417aa..df0537be1d 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -47,7 +47,6 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); - struct softpipe_surface *sps = softpipe_surface(ps); unsigned x, y, w, h; softpipe_update_derived(softpipe); /* not needed?? */ @@ -66,12 +65,12 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, assert(w <= ps->region->pitch); assert(h <= ps->region->height); - if (sps == sp_tile_cache_get_surface(softpipe->zbuf_cache)) { + if (ps == sp_tile_cache_get_surface(softpipe->zbuf_cache)) { float clear[4]; clear[0] = 1.0; /* XXX hack */ sp_tile_cache_clear(softpipe->zbuf_cache, clear); } - else if (sps == sp_tile_cache_get_surface(softpipe->cbuf_cache[0])) { + else if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[0])) { float clear[4]; clear[0] = 0.2; /* XXX hack */ clear[1] = 0.2; /* XXX hack */ @@ -84,6 +83,6 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, #if 0 - sp_clear_tile_cache(sps, clearValue); + sp_clear_tile_cache(ps, clearValue); #endif } diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 88a418d3c7..e51e81c7b8 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -40,7 +40,6 @@ #include "sp_quad.h" -struct softpipe_surface; struct softpipe_winsys; struct draw_context; struct draw_stage; diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 05dafeca7c..188509065b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -50,8 +50,8 @@ void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.zbuf); - const uint format = sps->surface.format; + struct pipe_surface *ps = softpipe->framebuffer.zbuf; + const uint format = ps->format; unsigned bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ unsigned qzzzz[QUAD_SIZE]; /**< Z values from the quad */ unsigned zmask = 0; @@ -59,7 +59,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) struct softpipe_cached_tile *tile = sp_get_cached_tile(softpipe, softpipe->zbuf_cache, quad->x0, quad->y0); - assert(sps); /* shouldn't get here if there's no zbuffer */ + assert(ps); /* shouldn't get here if there's no zbuffer */ /* * Convert quad's float depth values to int depth values (qzzzz). diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 8475bf3b96..c4240f1f8b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -201,7 +201,7 @@ static void stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_surface *sps = softpipe_surface(softpipe->framebuffer.sbuf); + struct pipe_surface *ps = softpipe->framebuffer.sbuf; unsigned func, zFailOp, zPassOp, failOp; ubyte ref, wrtMask, valMask; ubyte stencilVals[QUAD_SIZE]; @@ -230,10 +230,10 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) valMask = softpipe->depth_stencil->stencil.value_mask[0]; } - assert(sps); /* shouldn't get here if there's no stencil buffer */ + assert(ps); /* shouldn't get here if there's no stencil buffer */ /* get stencil values from cached tile */ - switch (sps->surface.format) { + switch (ps->format) { case PIPE_FORMAT_S8_Z24: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); @@ -290,7 +290,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) } /* put new stencil values into cached tile */ - switch (sps->surface.format) { + switch (ps->format) { case PIPE_FORMAT_S8_Z24: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index 0960fc45ab..d4e0bd1e15 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -44,7 +44,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, const struct pipe_framebuffer_state *fb) { struct softpipe_context *sp = softpipe_context(pipe); - struct softpipe_surface *sps; + struct pipe_surface *ps; uint i; for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { @@ -53,18 +53,18 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* flush old */ sp_flush_tile_cache(sp, sp->cbuf_cache[i]); /* unmap old */ - sps = softpipe_surface(sp->framebuffer.cbufs[i]); - if (sps && sps->surface.region) - pipe->region_unmap(pipe, sps->surface.region); + ps = sp->framebuffer.cbufs[i]; + if (ps && ps->region) + pipe->region_unmap(pipe, ps->region); /* map new */ - sps = softpipe_surface(fb->cbufs[i]); - if (sps) - pipe->region_map(pipe, sps->surface.region); + ps = fb->cbufs[i]; + if (ps) + pipe->region_map(pipe, ps->region); /* assign new */ sp->framebuffer.cbufs[i] = fb->cbufs[i]; /* update cache */ - sp_tile_cache_set_surface(sp->cbuf_cache[i], sps); + sp_tile_cache_set_surface(sp->cbuf_cache[i], ps); } } @@ -75,22 +75,22 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* flush old */ sp_flush_tile_cache(sp, sp->zbuf_cache); /* unmap old */ - sps = softpipe_surface(sp->framebuffer.zbuf); - if (sps && sps->surface.region) - pipe->region_unmap(pipe, sps->surface.region); + ps = sp->framebuffer.zbuf; + if (ps && ps->region) + pipe->region_unmap(pipe, ps->region); if (sp->framebuffer.sbuf == sp->framebuffer.zbuf) { /* combined z/stencil */ sp->framebuffer.sbuf = NULL; } /* map new */ - sps = softpipe_surface(fb->zbuf); - if (sps) - pipe->region_map(pipe, sps->surface.region); + ps = fb->zbuf; + if (ps) + pipe->region_map(pipe, ps->region); /* assign new */ sp->framebuffer.zbuf = fb->zbuf; /* update cache */ - sp_tile_cache_set_surface(sp->zbuf_cache, sps); + sp_tile_cache_set_surface(sp->zbuf_cache, ps); } /* XXX combined depth/stencil here */ @@ -100,13 +100,13 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* flush old */ sp_flush_tile_cache(sp, sp->sbuf_cache_sep); /* unmap old */ - sps = softpipe_surface(sp->framebuffer.sbuf); - if (sps && sps->surface.region) - pipe->region_unmap(pipe, sps->surface.region); + ps = sp->framebuffer.sbuf; + if (ps && ps->region) + pipe->region_unmap(pipe, ps->region); /* map new */ - sps = softpipe_surface(fb->sbuf); - if (sps && fb->sbuf != fb->zbuf) - pipe->region_map(pipe, sps->surface.region); + ps = fb->sbuf; + if (ps && fb->sbuf != fb->zbuf) + pipe->region_map(pipe, ps->region); /* assign new */ sp->framebuffer.sbuf = fb->sbuf; @@ -114,12 +114,12 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, if (fb->sbuf != fb->zbuf) { /* separate stencil buf */ sp->sbuf_cache = sp->sbuf_cache_sep; - sp_tile_cache_set_surface(sp->sbuf_cache, sps); + sp_tile_cache_set_surface(sp->sbuf_cache, ps); } else { /* combined depth/stencil */ sp->sbuf_cache = sp->zbuf_cache; - sp_tile_cache_set_surface(sp->sbuf_cache, sps); + sp_tile_cache_set_surface(sp->sbuf_cache, ps); } } diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index d1aa2aba97..bc343140ba 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -460,78 +460,6 @@ s8z24_get_tile(struct pipe_surface *ps, } -/** - * Initialize the quad_read/write and get/put_tile() methods. - */ -void -softpipe_init_surface_funcs(struct softpipe_surface *sps) -{ - assert(sps->surface.format); -#if 0 - switch (sps->surface.format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - sps->get_tile = a8r8g8b8_get_tile; - sps->put_tile = a8r8g8b8_put_tile; - break; - case PIPE_FORMAT_U_A1_R5_G5_B5: - sps->get_tile = a1r5g5b5_get_tile; - break; - case PIPE_FORMAT_U_L8: - sps->get_tile = l8_get_tile; - break; - case PIPE_FORMAT_U_A8: - sps->get_tile = a8_get_tile; - break; - case PIPE_FORMAT_U_I8: - sps->get_tile = i8_get_tile; - break; - case PIPE_FORMAT_U_A8_L8: - sps->get_tile = a8_l8_get_tile; - break; - - case PIPE_FORMAT_S_R16_G16_B16_A16: - sps->get_tile = r16g16b16a16_get_tile; - sps->put_tile = r16g16b16a16_put_tile; - break; - - case PIPE_FORMAT_U_Z16: - sps->get_tile = z16_get_tile; - break; - case PIPE_FORMAT_U_Z32: - sps->get_tile = z32_get_tile; - break; - case PIPE_FORMAT_S8_Z24: - sps->get_tile = s8z24_get_tile; - break; - case PIPE_FORMAT_U_S8: - break; - default: - assert(0); - } -#endif -} - - -static struct pipe_surface * -softpipe_surface_alloc(struct pipe_context *pipe, unsigned pipeFormat) -{ - struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); - if (!sps) - return NULL; - - assert(pipeFormat < PIPE_FORMAT_COUNT); - - sps->surface.format = pipeFormat; - sps->surface.refcount = 1; - softpipe_init_surface_funcs(sps); - - return &sps->surface; -} - - - - - /** * Called via pipe->get_tex_surface() * XXX is this in the right place? @@ -730,9 +658,6 @@ softpipe_put_tile_rgba(struct pipe_context *pipe, void sp_init_surface_functions(struct softpipe_context *sp) { -#if 0 - sp->pipe.surface_alloc = softpipe_surface_alloc; -#endif sp->pipe.get_tile = softpipe_get_tile; sp->pipe.put_tile = softpipe_put_tile; diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 359a438c86..7e08ce7a2b 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -35,26 +35,10 @@ #include "pipe/p_state.h" struct pipe_context; -struct softpipe_surface; struct softpipe_context; struct softpipe_tile_cache; -/** - * Softpipe surface is derived from pipe_surface. - */ -struct softpipe_surface { - struct pipe_surface surface; - -#if 0 - /* XXX these are temporary here */ - void (*get_tile)(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *p); - void (*put_tile)(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *p); -#endif -}; - extern struct pipe_surface * softpipe_get_tex_surface(struct pipe_context *pipe, struct pipe_mipmap_tree *mt, @@ -73,17 +57,6 @@ softpipe_put_tile_rgba(struct pipe_context *pipe, uint x, uint y, uint w, uint h, const float *p); -extern void -softpipe_init_surface_funcs(struct softpipe_surface *sps); - - -/** Cast wrapper */ -static INLINE struct softpipe_surface * -softpipe_surface(struct pipe_surface *ps) -{ - return (struct softpipe_surface *) ps; -} - extern void sp_init_surface_functions(struct softpipe_context *sp); diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 421d7bdb1a..1b600aadf1 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -49,7 +49,7 @@ struct softpipe_tile_cache { - struct softpipe_surface *surface; /**< the surface we're caching */ + struct pipe_surface *surface; /**< the surface we're caching */ struct pipe_mipmap_tree *texture; /**< if caching a texture */ struct softpipe_cached_tile entries[NUM_ENTRIES]; uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; @@ -123,13 +123,13 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc) void sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, - struct softpipe_surface *sps) + struct pipe_surface *ps) { - tc->surface = sps; + tc->surface = ps; } -struct softpipe_surface * +struct pipe_surface * sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) { return tc->surface; @@ -157,7 +157,7 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc) { struct pipe_context *pipe = &softpipe->pipe; - struct pipe_surface *ps = &tc->surface->surface; + struct pipe_surface *ps = tc->surface; boolean is_depth_stencil; int inuse = 0, pos; @@ -199,7 +199,7 @@ sp_get_cached_tile(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc, int x, int y) { struct pipe_context *pipe = &softpipe->pipe; - struct pipe_surface *ps = &tc->surface->surface; + struct pipe_surface *ps = tc->surface; boolean is_depth_stencil = (ps->format == PIPE_FORMAT_S8_Z24 || ps->format == PIPE_FORMAT_U_Z16 || diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index e122d70d13..e66fec2e20 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -64,9 +64,9 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc); extern void sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, - struct softpipe_surface *sps); + struct pipe_surface *sps); -extern struct softpipe_surface * +extern struct pipe_surface * sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); extern void -- cgit v1.2.3 From 6ada4e3a69476ed086955fe4060b894559e1eb95 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:32:50 -0600 Subject: Comments about texture instructions and the src regs --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 20 ++++++++++++-------- src/mesa/state_tracker/st_mesa_to_tgsi.c | 13 ++++++++++--- 2 files changed, 22 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index f577f004d7..d98f46cc0d 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1996,15 +1996,15 @@ exec_instruction( case TGSI_OPCODE_TEX: /* simple texture lookup */ - /* src[0] is the texcoord */ - /* src[1] is the sampler unit */ + /* src[0] = texcoord */ + /* src[1] = sampler unit */ exec_tex(mach, inst, FALSE); break; case TGSI_OPCODE_TXB: /* Texture lookup with lod bias */ - /* src[0] is the texcoord (src[0].w = load bias) */ - /* src[1] is the sampler unit */ + /* src[0] = texcoord (src[0].w = load bias) */ + /* src[1] = sampler unit */ exec_tex(mach, inst, TRUE); break; @@ -2013,9 +2013,17 @@ exec_instruction( /* src[0] = texcoord */ /* src[1] = d[strq]/dx */ /* src[2] = d[strq]/dy */ + /* src[3] = sampler unit */ assert (0); break; + case TGSI_OPCODE_TXL: + /* Texture lookup with explit LOD */ + /* src[0] = texcoord (src[0].w = load bias) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE); + break; + case TGSI_OPCODE_UP2H: assert (0); break; @@ -2157,10 +2165,6 @@ exec_instruction( } break; - case TGSI_OPCODE_TXL: - assert (0); - break; - case TGSI_OPCODE_IF: /* push CondMask */ assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 131e50bece..621159eb85 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -459,6 +459,7 @@ compile_instruction( fullinst->Instruction.Opcode = TGSI_OPCODE_SWZ; break; case OPCODE_TEX: + /* ordinary texture lookup */ fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; fullinst->Instruction.NumSrcRegs = 2; fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); @@ -466,6 +467,7 @@ compile_instruction( fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; break; case OPCODE_TXB: + /* texture lookup with LOD bias */ fullinst->Instruction.Opcode = TGSI_OPCODE_TXB; fullinst->Instruction.NumSrcRegs = 2; fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); @@ -473,13 +475,16 @@ compile_instruction( fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; break; case OPCODE_TXD: + /* texture lookup with explicit partial derivatives */ fullinst->Instruction.Opcode = TGSI_OPCODE_TXD; - fullinst->Instruction.NumSrcRegs = 2; + fullinst->Instruction.NumSrcRegs = 4; fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + /* src[0] = coord, src[1] = d[strq]/dx, src[2] = d[strq]/dy */ + fullinst->FullSrcRegisters[3].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[3].SrcRegister.Index = inst->TexSrcUnit; break; case OPCODE_TXL: + /* texture lookup with explicit LOD */ fullinst->Instruction.Opcode = TGSI_OPCODE_TXL; fullinst->Instruction.NumSrcRegs = 2; fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); @@ -487,6 +492,8 @@ compile_instruction( fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; break; case OPCODE_TXP: + /* texture lookup with divide by Q component */ + /* convert to TEX w/ special flag for division */ fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; fullinst->Instruction.NumSrcRegs = 2; fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); -- cgit v1.2.3 From 0c2bcedf9a1c924d672f4ec1b0f7d8af7a9df876 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:33:50 -0600 Subject: indentation fixes --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 40 ++++++++++++++++++------------------- 1 file changed, 20 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index d98f46cc0d..c503bce196 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1535,21 +1535,21 @@ exec_instruction( assert (0); break; - case TGSI_OPCODE_LOG: - assert (0); - break; + case TGSI_OPCODE_LOG: + assert (0); + break; - case TGSI_OPCODE_MUL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); + case TGSI_OPCODE_MUL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); - micro_mul( &r[0], &r[0], &r[1] ); + micro_mul( &r[0], &r[0], &r[1] ); - STORE(&r[0], 0, chan_index); - } - break; + STORE(&r[0], 0, chan_index); + } + break; case TGSI_OPCODE_ADD: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { @@ -1688,16 +1688,16 @@ exec_instruction( } break; - case TGSI_OPCODE_SUB: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); + case TGSI_OPCODE_SUB: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); - micro_sub( &r[0], &r[0], &r[1] ); + micro_sub( &r[0], &r[0], &r[1] ); - STORE(&r[0], 0, chan_index); - } - break; + STORE(&r[0], 0, chan_index); + } + break; case TGSI_OPCODE_LERP: /* TGSI_OPCODE_LRP */ -- cgit v1.2.3 From 88c021a0d1ce81f5e3e6d972ae86c1efdb882e3c Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:35:51 -0600 Subject: s/256/TGSI_EXEC_NUM_IMMEDIATES/ --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index c503bce196..d16b8d50b6 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -173,7 +173,7 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) { uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; assert( size % 4 == 0 ); - assert( mach->ImmLimit + size / 4 <= 256 ); + assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES ); for( i = 0; i < size; i++ ) { mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; -- cgit v1.2.3 From e807b1900ff0a5661246d8eeff8b8e230231ad60 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:36:43 -0600 Subject: remove #define MESA stuff --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 5 ----- 1 file changed, 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index d16b8d50b6..bd57076d75 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -54,11 +54,6 @@ #include "tgsi_core.h" #include "pipe/p_state.h" -#define MESA 1 -#if MESA -#include "main/context.h" -#include "main/macros.h" -#endif #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 -- cgit v1.2.3 From d7c189b85484cb15570edc09862ed848bdd4c7d2 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:37:40 -0600 Subject: s/GLbitfield/uint/ --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index bd57076d75..1545f0d479 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1138,9 +1138,9 @@ static void exec_kilp(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst) { - GLbitfield uniquemask; + uint uniquemask; uint chan_index; - GLbitfield kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ union tgsi_exec_channel r[1]; /* This mask stores component bits that were already tested. Note that -- cgit v1.2.3 From d75ff8447cd0b993be36f399ccd360aaf412b1f3 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:38:41 -0600 Subject: indentation fixes --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 52 ++++++++++++++++++------------------- 1 file changed, 26 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 1545f0d479..df4abc6066 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1139,37 +1139,37 @@ exec_kilp(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst) { uint uniquemask; - uint chan_index; - uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ - union tgsi_exec_channel r[1]; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - for (chan_index = 0; chan_index < 4; chan_index++) - { - uint swizzle; - uint i; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle ( + uint chan_index; + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + union tgsi_exec_channel r[1]; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + uint swizzle; + uint i; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle ( &inst->FullSrcRegisters[0], chan_index); - /* check if the component has not been already tested */ - if (uniquemask & (1 << swizzle)) - continue; - uniquemask |= 1 << swizzle; + /* check if the component has not been already tested */ + if (uniquemask & (1 << swizzle)) + continue; + uniquemask |= 1 << swizzle; - FETCH(&r[0], 0, chan_index); - for (i = 0; i < 4; i++) - if (r[0].f[i] < 0.0f) - kilmask |= 1 << i; - } + FETCH(&r[0], 0, chan_index); + for (i = 0; i < 4; i++) + if (r[0].f[i] < 0.0f) + kilmask |= 1 << i; + } - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; } -- cgit v1.2.3 From 79d8e78442c08082083261d517cdf260e0bd309f Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 27 Oct 2007 09:43:28 -0600 Subject: New comments, replace //-style with /* */ --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 6 ++--- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 53 +++++++++++++++++++++++++++++++++---- 2 files changed, 51 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index df4abc6066..e4f23d3eef 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -1212,7 +1212,7 @@ exec_tex(struct tgsi_exec_machine *mach, uint chan_index; float lodBias; - // printf("Sampler %u unit %u\n", sampler, unit); + /* printf("Sampler %u unit %u\n", sampler, unit); */ switch (inst->InstructionExtTexture.Texture) { case TGSI_TEXTURE_1D: @@ -1364,7 +1364,7 @@ perspective_interpolation( for( i = 0; i < QUAD_SIZE; i++ ) { const float x = mach->Inputs[0].xyzw[0].f[i]; const float y = mach->Inputs[0].xyzw[1].f[i]; - // WPOS.w here is really 1/w + /* WPOS.w here is really 1/w */ const float w = 1.0f / mach->Inputs[0].xyzw[3].f[i]; assert(mach->Inputs[0].xyzw[3].f[i] != 0.0); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 38f9218520..1d497e97fb 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #if !defined TGSI_EXEC_H #define TGSI_EXEC_H @@ -5,11 +32,14 @@ #if defined __cplusplus extern "C" { -#endif // defined __cplusplus +#endif #define NUM_CHANNELS 4 /* R,G,B,A */ #define QUAD_SIZE 4 /* 4 pixel/quad */ +/** + * Registers may be treated as float, signed int or unsigned int. + */ union tgsi_exec_channel { float f[QUAD_SIZE]; @@ -17,11 +47,18 @@ union tgsi_exec_channel unsigned u[QUAD_SIZE]; }; +/** + * A vector[RGBA] of channels[4 pixels] + */ struct tgsi_exec_vector { union tgsi_exec_channel xyzw[NUM_CHANNELS]; }; +/** + * For fragment programs, information for computing fragment input + * values from plane equation of the triangle/line. + */ struct tgsi_interp_coef { float a0[NUM_CHANNELS]; /* in an xyzw layout */ @@ -32,6 +69,10 @@ struct tgsi_interp_coef struct softpipe_tile_cache; /**< Opaque to TGSI */ +/** + * Information for sampling textures, which must be implemented + * by code outside the TGSI executor. + */ struct tgsi_sampler { const struct pipe_sampler_state *state; @@ -47,6 +88,9 @@ struct tgsi_sampler struct softpipe_tile_cache *cache; }; +/** + * For branching/calling subroutines. + */ struct tgsi_exec_labels { unsigned labels[128][2]; @@ -184,8 +228,7 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ); #if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_EXEC_H +} /* extern "C" */ +#endif +#endif /* TGSI_EXEC_H */ -- cgit v1.2.3 From 0e96a53aec7a4eb0236fa31f4682734b9310cb80 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:12:25 +0100 Subject: Remove llvm dependencies. --- src/mesa/pipe/p_context.h | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 5497f50f73..dbbf48c576 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -39,8 +39,6 @@ struct pipe_state_cache; struct pipe_context { struct pipe_winsys *winsys; - void *llvm_execution_engine; - void (*destroy)( struct pipe_context * ); /* @@ -261,8 +259,8 @@ pipe_region_reference(struct pipe_region **ptr, struct pipe_region *region) if (*ptr) { /* unreference the old thing */ struct pipe_region *oldReg = *ptr; + assert(oldReg->refcount > 0); oldReg->refcount--; - assert(oldReg->refcount >= 0); if (oldReg->refcount == 0) { /* free the old region */ assert(oldReg->map_refcount == 0); @@ -289,8 +287,8 @@ pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) if (*ptr) { /* unreference the old thing */ struct pipe_surface *oldSurf = *ptr; + assert(oldSurf->refcount > 0); oldSurf->refcount--; - assert(oldSurf->refcount >= 0); if (oldSurf->refcount == 0) { /* free the old region */ pipe_region_reference(&oldSurf->region, NULL); @@ -307,3 +305,4 @@ pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) #endif /* PIPE_CONTEXT_H */ + -- cgit v1.2.3 From 11a19866f6e425d23f0a1dd404ae836b3c9f190b Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:14:29 +0100 Subject: Silence compiler warnings. --- src/mesa/pipe/draw/draw_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 544c423c2d..61f9e4909b 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -253,7 +253,7 @@ void draw_free_tmps( struct draw_stage *stage ) boolean draw_use_sse(struct draw_context *draw) { - return draw->use_sse; + return (boolean) draw->use_sse; } -- cgit v1.2.3 From bc99ea96fbde347aa4cad13477d5e3c30cb8069e Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:15:30 +0100 Subject: Add sse2_program to draw_vertex_shader. --- src/mesa/pipe/draw/draw_private.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index e4c0551124..44cf0b1c98 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -123,7 +123,8 @@ struct draw_stage * Private version of the compiled vertex_shader */ struct draw_vertex_shader { - const struct pipe_shader_state *state; + const struct pipe_shader_state *state; + struct x86_function sse2_program; }; /** -- cgit v1.2.3 From 26df9d1a304f03e9692e7bac0a3bc83330f30c6a Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:17:52 +0100 Subject: Move format definitions from p_defines.h to p_format.h. PIPE_FORMAT is not an ordinary enum -- it encodes its format description in the value that defines it. --- src/mesa/pipe/p_defines.h | 94 +------------------ src/mesa/pipe/p_format.h | 226 ++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 228 insertions(+), 92 deletions(-) create mode 100644 src/mesa/pipe/p_format.h (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 9281d40bd3..8982428636 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -28,6 +28,8 @@ #ifndef PIPE_DEFINES_H #define PIPE_DEFINES_H +#include "p_format.h" + #define PIPE_BLENDFACTOR_ONE 0x1 #define PIPE_BLENDFACTOR_SRC_COLOR 0x2 #define PIPE_BLENDFACTOR_SRC_ALPHA 0x3 @@ -158,98 +160,6 @@ #define PIPE_TEX_FACE_NEG_Z 5 #define PIPE_TEX_FACE_MAX 6 - -/** - * Texture/surface image formats (preliminary) - */ - -/* KW: Added lots of surface formats to support vertex element layout - * definitions, and eventually render-to-vertex-buffer. Could - * consider making float/int/uint/scaled/normalized a separate - * parameter, but on the other hand there are special cases like - * z24s8, compressed textures, ycbcr, etc that won't fit that model. - */ - -#define PIPE_FORMAT_NONE 0 /**< unstructured */ -#define PIPE_FORMAT_U_A8_R8_G8_B8 2 /**< ubyte[4] ARGB */ -#define PIPE_FORMAT_U_A1_R5_G5_B5 3 /**< 16-bit packed RGBA */ -#define PIPE_FORMAT_U_A4_R4_G4_B4 4 /**< 16-bit packed RGBA */ -#define PIPE_FORMAT_U_R5_G6_B5 5 /**< 16-bit packed RGB */ -#define PIPE_FORMAT_U_L8 6 /**< ubyte luminance */ -#define PIPE_FORMAT_U_A8 7 /**< ubyte alpha */ -#define PIPE_FORMAT_U_I8 8 /**< ubyte intensity */ -#define PIPE_FORMAT_U_A8_L8 9 /**< ubyte alpha, luminance */ -#define PIPE_FORMAT_S_R16_G16_B16_A16 10 /**< signed 16-bit RGBA (accum) */ -#define PIPE_FORMAT_YCBCR 11 -#define PIPE_FORMAT_YCBCR_REV 12 -#define PIPE_FORMAT_U_Z16 13 /**< ushort Z/depth */ -#define PIPE_FORMAT_U_Z32 14 /**< uint Z/depth */ -#define PIPE_FORMAT_F_Z32 15 /**< float Z/depth */ -#define PIPE_FORMAT_S8_Z24 16 /**< 8-bit stencil + 24-bit Z */ -#define PIPE_FORMAT_U_S8 17 /**< 8-bit stencil */ -#define PIPE_FORMAT_R64_FLOAT 0x20 -#define PIPE_FORMAT_R64G64_FLOAT 0x21 -#define PIPE_FORMAT_R64G64B64_FLOAT 0x22 -#define PIPE_FORMAT_R64G64B64A64_FLOAT 0x23 -#define PIPE_FORMAT_R32_FLOAT 0x24 -#define PIPE_FORMAT_R32G32_FLOAT 0x25 -#define PIPE_FORMAT_R32G32B32_FLOAT 0x26 -#define PIPE_FORMAT_R32G32B32A32_FLOAT 0x27 -#define PIPE_FORMAT_R32_UNORM 0x28 -#define PIPE_FORMAT_R32G32_UNORM 0x29 -#define PIPE_FORMAT_R32G32B32_UNORM 0x2a -#define PIPE_FORMAT_R32G32B32A32_UNORM 0x2b -#define PIPE_FORMAT_R32_USCALED 0x2c -#define PIPE_FORMAT_R32G32_USCALED 0x2d -#define PIPE_FORMAT_R32G32B32_USCALED 0x2e -#define PIPE_FORMAT_R32G32B32A32_USCALED 0x2f -#define PIPE_FORMAT_R32_SNORM 0x30 -#define PIPE_FORMAT_R32G32_SNORM 0x31 -#define PIPE_FORMAT_R32G32B32_SNORM 0x32 -#define PIPE_FORMAT_R32G32B32A32_SNORM 0x33 -#define PIPE_FORMAT_R32_SSCALED 0x34 -#define PIPE_FORMAT_R32G32_SSCALED 0x35 -#define PIPE_FORMAT_R32G32B32_SSCALED 0x36 -#define PIPE_FORMAT_R32G32B32A32_SSCALED 0x37 -#define PIPE_FORMAT_R16_UNORM 0x38 -#define PIPE_FORMAT_R16G16_UNORM 0x39 -#define PIPE_FORMAT_R16G16B16_UNORM 0x3a -#define PIPE_FORMAT_R16G16B16A16_UNORM 0x3b -#define PIPE_FORMAT_R16_USCALED 0x3c -#define PIPE_FORMAT_R16G16_USCALED 0x3d -#define PIPE_FORMAT_R16G16B16_USCALED 0x3e -#define PIPE_FORMAT_R16G16B16A16_USCALED 0x3f -#define PIPE_FORMAT_R16_SNORM 0x40 -#define PIPE_FORMAT_R16G16_SNORM 0x41 -#define PIPE_FORMAT_R16G16B16_SNORM 0x42 -#define PIPE_FORMAT_R16G16B16A16_SNORM 0x43 -#define PIPE_FORMAT_R16_SSCALED 0x44 -#define PIPE_FORMAT_R16G16_SSCALED 0x45 -#define PIPE_FORMAT_R16G16B16_SSCALED 0x46 -#define PIPE_FORMAT_R16G16B16A16_SSCALED 0x47 -#define PIPE_FORMAT_R8_UNORM 0x48 -#define PIPE_FORMAT_R8G8_UNORM 0x49 -#define PIPE_FORMAT_R8G8B8_UNORM 0x4a -#define PIPE_FORMAT_R8G8B8A8_UNORM 0x4b -#define PIPE_FORMAT_R8_USCALED 0x4c -#define PIPE_FORMAT_R8G8_USCALED 0x4d -#define PIPE_FORMAT_R8G8B8_USCALED 0x4e -#define PIPE_FORMAT_R8G8B8A8_USCALED 0x4f -#define PIPE_FORMAT_R8_SNORM 0x50 -#define PIPE_FORMAT_R8G8_SNORM 0x51 -#define PIPE_FORMAT_R8G8B8_SNORM 0x52 -#define PIPE_FORMAT_R8G8B8A8_SNORM 0x53 -#define PIPE_FORMAT_R8_SSCALED 0x54 -#define PIPE_FORMAT_R8G8_SSCALED 0x55 -#define PIPE_FORMAT_R8G8B8_SSCALED 0x56 -#define PIPE_FORMAT_R8G8B8A8_SSCALED 0x57 - -#define PIPE_FORMAT_COUNT 0x58 /**< number of formats */ - -/* Duplicated formats: - */ -#define PIPE_FORMAT_U_R8_G8_B8_A8 PIPE_FORMAT_R8G8B8A8_UNORM - /** * Surface flags */ diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h new file mode 100644 index 0000000000..edb6fa4f1e --- /dev/null +++ b/src/mesa/pipe/p_format.h @@ -0,0 +1,226 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef PIPE_FORMAT_H +#define PIPE_FORMAT_H + +#include "p_compiler.h" + +#define PIPE_FORMAT_LAYOUT_RGBAZS 0 +#define PIPE_FORMAT_LAYOUT_YCBCR 1 + +#define PIPE_FORMAT_COMP_R 0 +#define PIPE_FORMAT_COMP_G 1 +#define PIPE_FORMAT_COMP_B 2 +#define PIPE_FORMAT_COMP_A 3 +#define PIPE_FORMAT_COMP_0 4 +#define PIPE_FORMAT_COMP_1 5 +#define PIPE_FORMAT_COMP_Z 6 +#define PIPE_FORMAT_COMP_S 7 + +#define PIPE_FORMAT_TYPE_UNKNOWN 0 +#define PIPE_FORMAT_TYPE_FLOAT 1 +#define PIPE_FORMAT_TYPE_UNORM 2 +#define PIPE_FORMAT_TYPE_SNORM 3 +#define PIPE_FORMAT_TYPE_USCALED 4 +#define PIPE_FORMAT_TYPE_SSCALED 5 + +struct pipe_format_header +{ + uint layout : 2; /**< PIPE_FORMAT_LAYOUT_ */ + uint padding : 30; +}; + +/** + * RGBAZS Format Layout. + */ +struct pipe_format_rgbazs +{ + uint layout : 2; /**< PIPE_FORMAT_LAYOUT_RGBAZS */ + uint swizzleX : 3; /**< PIPE_FORMAT_COMP_ */ + uint swizzleY : 3; /**< PIPE_FORMAT_COMP_ */ + uint swizzleZ : 3; /**< PIPE_FORMAT_COMP_ */ + uint swizzleW : 3; /**< PIPE_FORMAT_COMP_ */ + uint sizeX : 3; /**< Size of X - 1 */ + uint sizeY : 3; /**< Size of Y - 1 */ + uint sizeZ : 3; /**< Size of Z - 1 */ + uint sizeW : 3; /**< Size of W - 1 */ + uint exp8 : 2; /**< Scale size by 8 ^ exp8 */ + uint type : 3; /**< PIPE_FORMAT_TYPE_ */ + uint padding : 1; +}; + +#define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP8, TYPE ) (\ + (PIPE_FORMAT_LAYOUT_RGBAZS << 0) |\ + ((SWZ) << 2) |\ + ((SIZEX) << 14) |\ + ((SIZEY) << 17) |\ + ((SIZEZ) << 20) |\ + ((SIZEW) << 23) |\ + ((EXP8) << 26) |\ + ((TYPE) << 28) ) + +#define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 2) | ((SWZY) << 5) | ((SWZZ) << 8) | ((SWZW) << 11)) + +#define _PIPE_FORMAT_RGBAZS_1U( SWZ, SIZE, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 0, TYPE ) + +#define _PIPE_FORMAT_RGBAZS_1N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE ) + +#define _PIPE_FORMAT_RGBAZS_8U( SWZ, SIZE, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 1, TYPE ) + +#define _PIPE_FORMAT_RGBAZS_8N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) + +#define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 2, TYPE ) + +#define _PIPE_FORMAT_R000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_RG00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) +#define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R ) +#define _PIPE_FORMAT_RRR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) +#define _PIPE_FORMAT_RRRR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R ) +#define _PIPE_FORMAT_RRRG _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G ) +#define _PIPE_FORMAT_Z000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) + +struct pipe_format_ycbcr +{ + uint layout : 2; /**< PIPE_FORMAT_LAYOUT_YCBCR */ + uint reversed : 1; + uint padding : 29; +}; + +#define _PIPE_FORMAT_YCBCR( REV ) (\ + (PIPE_FORMAT_LAYOUT_YCBCR << 0) |\ + ((REVERSED) << 2) ) + +union pipe_format +{ + uint value32; + struct pipe_format_header header; + struct pipe_format_rgbazs rgbazs; + struct pipe_format_ycbcr ycbcr; +}; + +/** + * Texture/surface image formats (preliminary) + */ + +/* KW: Added lots of surface formats to support vertex element layout + * definitions, and eventually render-to-vertex-buffer. Could + * consider making float/int/uint/scaled/normalized a separate + * parameter, but on the other hand there are special cases like + * z24s8, compressed textures, ycbcr, etc that won't fit that model. + */ + +#define PIPE_FORMAT_NONE _PIPE_FORMAT_RGBAZS_1U( _PIPE_FORMAT_0000, 0, PIPE_FORMAT_TYPE_UNKNOWN ) /**< unstructured */ +#define PIPE_FORMAT_U_A8_R8_G8_B8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_ARGB, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte[4] ARGB */ +#define PIPE_FORMAT_U_A1_R5_G5_B5 _PIPE_FORMAT_RGBAZS_1N( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGBA */ +#define PIPE_FORMAT_U_A4_R4_G4_B4 _PIPE_FORMAT_RGBAZS_1U( _PIPE_FORMAT_ARGB, 4, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGBA */ +#define PIPE_FORMAT_U_R5_G6_B5 _PIPE_FORMAT_RGBAZS_1N( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGB */ +#define PIPE_FORMAT_U_L8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRR1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte luminance */ +#define PIPE_FORMAT_U_A8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_000R, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha */ +#define PIPE_FORMAT_U_I8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRRR, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte intensity */ +#define PIPE_FORMAT_U_A8_L8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRRG, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha, luminance */ +#define PIPE_FORMAT_S_R16_G16_B16_A16 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RGBA, 2, PIPE_FORMAT_TYPE_SNORM ) /**< signed 16-bit RGBA (accum) */ +#define PIPE_FORMAT_YCBCR _PIPE_FORMAT_YCBCR( 0 ) +#define PIPE_FORMAT_YCBCR_REV _PIPE_FORMAT_YCBCR( 1 ) +#define PIPE_FORMAT_U_Z16 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< ushort Z/depth */ +#define PIPE_FORMAT_U_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< uint Z/depth */ +#define PIPE_FORMAT_F_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) /**< float Z/depth */ +#define PIPE_FORMAT_S8_Z24 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil + 24-bit Z */ +#define PIPE_FORMAT_U_S8 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil */ +#define PIPE_FORMAT_R64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R64G64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R64G64B64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R64G64B64A64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32G32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32G32B32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32G32B32A32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32G32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32G32B32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32G32B32A32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32G32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32G32B32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32G32B32A32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32G32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32G32B32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32G32B32A32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32G32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32G32B32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32G32B32A32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16G16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16G16B16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16G16B16A16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16G16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16G16B16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16G16B16A16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16G16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16G16B16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16G16B16A16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16G16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16G16B16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16G16B16A16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8G8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8G8B8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8G8B8A8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8G8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8G8B8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8G8B8A8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8G8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8G8B8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8G8B8A8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8G8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8G8B8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8G8B8A8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) + +/* Duplicated formats: + */ +#define PIPE_FORMAT_U_R8_G8_B8_A8 PIPE_FORMAT_R8G8B8A8_UNORM + +#endif -- cgit v1.2.3 From 21b5ff5dccd448cd581203ff177244c835e407a8 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:19:20 +0100 Subject: Remove llvm and sse2 dependencies. --- src/mesa/pipe/p_state.h | 10 ---------- 1 file changed, 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 03045e4213..a61cbaf8e1 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -40,8 +40,6 @@ #include "p_compiler.h" -#include "x86/rtasm/x86sse.h" - /** * Implementation limits */ @@ -58,7 +56,6 @@ /* fwd decl */ struct pipe_surface; -struct ga_llvm_prog; /* opaque type */ struct pipe_buffer_handle; @@ -149,13 +146,6 @@ struct pipe_constant_buffer { struct pipe_shader_state { const struct tgsi_token *tokens; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif - void (*executable)(); - - const struct ga_llvm_prog *llvm_prog; - ubyte num_inputs; ubyte num_outputs; ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ -- cgit v1.2.3 From 3755840e005dd11db0a51effa420672ac4f119d4 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:20:11 +0100 Subject: Call quad stage destructors. --- src/mesa/pipe/softpipe/sp_context.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 58ef744f30..999047889a 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -215,6 +215,18 @@ static void softpipe_destroy( struct pipe_context *pipe ) draw_destroy( softpipe->draw ); + softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); + softpipe->quad.shade->destroy( softpipe->quad.shade ); + softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); + softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); + softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); + softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); + softpipe->quad.coverage->destroy( softpipe->quad.coverage ); + softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); + softpipe->quad.blend->destroy( softpipe->quad.blend ); + softpipe->quad.colormask->destroy( softpipe->quad.colormask ); + softpipe->quad.output->destroy( softpipe->quad.output ); + free( softpipe ); } -- cgit v1.2.3 From 187c164bb6fa794f59181df89e72ff8c543238c6 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:20:53 +0100 Subject: Silence compiler warnings. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 73b18b776b..0b1b592bbe 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -853,7 +853,7 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) for (i = 0; i < dx; i++) { if (!sp->rasterizer->line_stipple_enable || stipple_test(sp->line_stipple_counter, - sp->rasterizer->line_stipple_pattern, + (ushort) sp->rasterizer->line_stipple_pattern, sp->rasterizer->line_stipple_factor + 1)) { plot(setup, x0, y0); } @@ -880,7 +880,7 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) for (i = 0; i < dy; i++) { if (!sp->rasterizer->line_stipple_enable || stipple_test(sp->line_stipple_counter, - sp->rasterizer->line_stipple_pattern, + (ushort) sp->rasterizer->line_stipple_pattern, sp->rasterizer->line_stipple_factor + 1)) { plot(setup, x0, y0); } @@ -1184,7 +1184,7 @@ void sp_vbuf_setup_draw( struct pipe_context *pipe, struct setup_stage *setup = setup_stage( softpipe->setup ); struct prim_header prim; unsigned vertex_size = setup->stage.draw->vertex_info.size * sizeof(float); - int i, j; + unsigned i, j; prim.det = 0; prim.reset_line_stipple = 0; -- cgit v1.2.3 From dee9406e4847f98b346f0fff72d16df46e9584a4 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:21:30 +0100 Subject: Silence compiler warnings. --- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index a950e72e56..a26179e45c 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -90,7 +90,7 @@ static INLINE struct vbuf_stage *vbuf_stage( struct draw_stage *stage ) static boolean overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) { - unsigned long used = (char *)ptr - (char *)map; + unsigned long used = (unsigned long) ((char *) ptr - (char *) map); return (used + bytes) > bufsz; } @@ -143,7 +143,7 @@ static void vbuf_tri( struct draw_stage *stage, if (prim->v[i]->vertex_id == 0xffff) emit_vertex( vbuf, prim->v[i] ); - vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; + vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; } } @@ -161,7 +161,7 @@ static void vbuf_line(struct draw_stage *stage, if (prim->v[i]->vertex_id == 0xffff) emit_vertex( vbuf, prim->v[i] ); - vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; + vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; } } @@ -177,7 +177,7 @@ static void vbuf_point(struct draw_stage *stage, if (prim->v[0]->vertex_id == 0xffff) emit_vertex( vbuf, prim->v[0] ); - vbuf->element_map[vbuf->nr_elements++] = prim->v[0]->vertex_id; + vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[0]->vertex_id; } @@ -230,7 +230,7 @@ static void vbuf_flush_elements( struct draw_stage *stage ) vbuf->element_map, vbuf->nr_elements, vbuf->vertex_map, - (vbuf->vertex_ptr - vbuf->vertex_map) / vbuf->vertex_size ); + (unsigned) (vbuf->vertex_ptr - vbuf->vertex_map) / vbuf->vertex_size ); vbuf->nr_elements = 0; -- cgit v1.2.3 From 6961769cb23c8b9ed2fb56d8ce6e649848412357 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:25:09 +0100 Subject: Define destroy method called by softpipe's destructor. --- src/mesa/pipe/softpipe/sp_quad.h | 2 ++ src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 7 +++++++ src/mesa/pipe/softpipe/sp_quad_blend.c | 7 +++++++ src/mesa/pipe/softpipe/sp_quad_bufloop.c | 7 +++++++ src/mesa/pipe/softpipe/sp_quad_colormask.c | 7 +++++++ src/mesa/pipe/softpipe/sp_quad_coverage.c | 7 +++++++ src/mesa/pipe/softpipe/sp_quad_depth_test.c | 7 +++++++ src/mesa/pipe/softpipe/sp_quad_fs.c | 7 +++++++ src/mesa/pipe/softpipe/sp_quad_occlusion.c | 6 ++++++ src/mesa/pipe/softpipe/sp_quad_output.c | 7 +++++++ src/mesa/pipe/softpipe/sp_quad_stencil.c | 7 +++++++ src/mesa/pipe/softpipe/sp_quad_stipple.c | 7 +++++++ 12 files changed, 78 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index cca7dbdcbe..534541122b 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -45,6 +45,8 @@ struct quad_stage { /** the stage action */ void (*run)(struct quad_stage *qs, struct quad_header *quad); + + void (*destroy)(struct quad_stage *qs); }; diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index 4f28414b0e..d9b914d896 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -88,6 +88,12 @@ static void alpha_test_begin(struct quad_stage *qs) } +static void alpha_test_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + struct quad_stage * sp_quad_alpha_test_stage( struct softpipe_context *softpipe ) { @@ -96,6 +102,7 @@ sp_quad_alpha_test_stage( struct softpipe_context *softpipe ) stage->softpipe = softpipe; stage->begin = alpha_test_begin; stage->run = alpha_test_quad; + stage->destroy = alpha_test_destroy; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 9b7a48669d..cecf8af29b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -727,6 +727,12 @@ static void blend_begin(struct quad_stage *qs) } +static void blend_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); @@ -734,6 +740,7 @@ struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ) stage->softpipe = softpipe; stage->begin = blend_begin; stage->run = blend_quad; + stage->destroy = blend_destroy; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c index ae2fb5ef97..bdf491d4b1 100644 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -50,6 +50,12 @@ static void cbuf_loop_begin(struct quad_stage *qs) } +static void cbuf_loop_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + /** * Create the colorbuffer loop stage. * This is used to implement multiple render targets and GL_FRONT_AND_BACK @@ -62,6 +68,7 @@ struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ) stage->softpipe = softpipe; stage->begin = cbuf_loop_begin; stage->run = cbuf_loop_quad; + stage->destroy = cbuf_loop_destroy; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index 8872825555..f0aa657e95 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -90,6 +90,12 @@ static void colormask_begin(struct quad_stage *qs) } +static void colormask_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); @@ -97,6 +103,7 @@ struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ) stage->softpipe = softpipe; stage->begin = colormask_begin; stage->run = colormask_quad; + stage->destroy = colormask_destroy; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c index 89f50bcca2..25e7b033b9 100644 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -69,6 +69,12 @@ static void coverage_begin(struct quad_stage *qs) } +static void coverage_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); @@ -76,6 +82,7 @@ struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ) stage->softpipe = softpipe; stage->begin = coverage_begin; stage->run = coverage_quad; + stage->destroy = coverage_destroy; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 188509065b..c24232bf1e 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -231,6 +231,12 @@ static void depth_test_begin(struct quad_stage *qs) } +static void depth_test_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); @@ -238,6 +244,7 @@ struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ) stage->softpipe = softpipe; stage->begin = depth_test_begin; stage->run = depth_test_quad; + stage->destroy = depth_test_destroy; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 9b9504cd15..dc8d6e0e23 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -190,6 +190,12 @@ static void shade_begin(struct quad_stage *qs) } +static void shade_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) { struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); @@ -204,6 +210,7 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) qss->stage.softpipe = softpipe; qss->stage.begin = shade_begin; qss->stage.run = shade_quad; + qss->stage.destroy = shade_destroy; /* set TGSI sampler state that's constant */ for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c index 4f178f0557..18a36e8286 100644 --- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c +++ b/src/mesa/pipe/softpipe/sp_quad_occlusion.c @@ -64,6 +64,11 @@ static void occlusion_begin(struct quad_stage *qs) } +static void occlusion_destroy(struct quad_stage *qs) +{ + free( qs ); +} + struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ) { @@ -72,6 +77,7 @@ struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ) stage->softpipe = softpipe; stage->begin = occlusion_begin; stage->run = occlusion_count_quad; + stage->destroy = occlusion_destroy; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index f757a43927..09194c14a6 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -70,6 +70,12 @@ static void output_begin(struct quad_stage *qs) } +static void output_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); @@ -77,6 +83,7 @@ struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ) stage->softpipe = softpipe; stage->begin = output_begin; stage->run = output_quad; + stage->destroy = output_destroy; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index c4240f1f8b..14aa7df279 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -323,6 +323,12 @@ static void stencil_begin(struct quad_stage *qs) } +static void stencil_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ) { struct quad_stage *stage = CALLOC_STRUCT(quad_stage); @@ -330,6 +336,7 @@ struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe stage->softpipe = softpipe; stage->begin = stencil_begin; stage->run = stencil_test_quad; + stage->destroy = stencil_destroy; return stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index 4af5059d67..7916dcd0ca 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -65,6 +65,12 @@ static void stipple_begin(struct quad_stage *qs) } +static void stipple_destroy(struct quad_stage *qs) +{ + free( qs ); +} + + struct quad_stage * sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ) { @@ -73,6 +79,7 @@ sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ) stage->softpipe = softpipe; stage->begin = stipple_begin; stage->run = stipple_quad; + stage->destroy = stipple_destroy; return stage; } -- cgit v1.2.3 From 9053fcabcbf1b1c969a9a52585d232bb778eba63 Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 17:53:56 +0100 Subject: Remove GL dependencies. --- src/mesa/pipe/tgsi/exec/tgsi_util.c | 42 ++++++++++++++++++------------------- 1 file changed, 21 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_util.c b/src/mesa/pipe/tgsi/exec/tgsi_util.c index 67e32ccbd0..d1bc935526 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_util.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_util.c @@ -19,10 +19,10 @@ tgsi_align_128bit( return ph.pointer; } -GLuint +unsigned tgsi_util_get_src_register_swizzle( const struct tgsi_src_register *reg, - GLuint component ) + unsigned component ) { switch( component ) { case 0: @@ -39,10 +39,10 @@ tgsi_util_get_src_register_swizzle( return 0; } -GLuint +unsigned tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - GLuint component ) + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) { switch( component ) { case 0: @@ -59,12 +59,12 @@ tgsi_util_get_src_register_extswizzle( return 0; } -GLuint +unsigned tgsi_util_get_full_src_register_extswizzle( const struct tgsi_full_src_register *reg, - GLuint component ) + unsigned component ) { - GLuint swizzle; + unsigned swizzle; /* * First, calculate the extended swizzle for a given channel. This will give @@ -97,8 +97,8 @@ tgsi_util_get_full_src_register_extswizzle( void tgsi_util_set_src_register_swizzle( struct tgsi_src_register *reg, - GLuint swizzle, - GLuint component ) + unsigned swizzle, + unsigned component ) { switch( component ) { case 0: @@ -121,8 +121,8 @@ tgsi_util_set_src_register_swizzle( void tgsi_util_set_src_register_extswizzle( struct tgsi_src_register_ext_swz *reg, - GLuint swizzle, - GLuint component ) + unsigned swizzle, + unsigned component ) { switch( component ) { case 0: @@ -142,10 +142,10 @@ tgsi_util_set_src_register_extswizzle( } } -GLuint +unsigned tgsi_util_get_src_register_extnegate( const struct tgsi_src_register_ext_swz *reg, - GLuint component ) + unsigned component ) { switch( component ) { case 0: @@ -165,8 +165,8 @@ tgsi_util_get_src_register_extnegate( void tgsi_util_set_src_register_extnegate( struct tgsi_src_register_ext_swz *reg, - GLuint negate, - GLuint component ) + unsigned negate, + unsigned component ) { switch( component ) { case 0: @@ -186,12 +186,12 @@ tgsi_util_set_src_register_extnegate( } } -GLuint +unsigned tgsi_util_get_full_src_register_sign_mode( const struct tgsi_full_src_register *reg, - GLuint component ) + unsigned component ) { - GLuint sign_mode; + unsigned sign_mode; if( reg->SrcRegisterExtMod.Absolute ) { /* Consider only the post-abs negation. */ @@ -206,7 +206,7 @@ tgsi_util_get_full_src_register_sign_mode( else { /* Accumulate the three negations. */ - GLuint negate; + unsigned negate; negate = reg->SrcRegister.Negate; if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { @@ -230,7 +230,7 @@ tgsi_util_get_full_src_register_sign_mode( void tgsi_util_set_full_src_register_sign_mode( struct tgsi_full_src_register *reg, - GLuint sign_mode ) + unsigned sign_mode ) { reg->SrcRegisterExtSwz.NegateX = 0; reg->SrcRegisterExtSwz.NegateY = 0; -- cgit v1.2.3 From bf229cb6f1aa01ae005afd2d90b3563ff495eaaa Mon Sep 17 00:00:00 2001 From: michal Date: Fri, 26 Oct 2007 18:00:35 +0100 Subject: Fix alignment problems. Remove NOP opcode. TEX opcode returns solid white (a hack for D3D state tracker). --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 61 +++++++++++++++++++++---------------- src/mesa/pipe/tgsi/exec/tgsi_sse2.h | 4 +-- 2 files changed, 37 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 77ba7bb944..1e1002f54e 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -2,8 +2,6 @@ #include "tgsi_core.h" #include "x86/rtasm/x86sse.h" -#if defined(__i386__) || defined(__386__) - #define FOR_EACH_CHANNEL( CHAN )\ for( CHAN = 0; CHAN < 4; CHAN++ ) @@ -22,7 +20,17 @@ #define CHAN_Z 2 #define CHAN_W 3 -#define ALIGN16( ADDR ) (((ADDR) + 15) & ~15) +static unsigned +p2u( void *p ) +{ + union { + void *p; + unsigned u; + } pu; + + pu.p = p; + return pu.u; +} static struct x86_reg get_argument( @@ -482,7 +490,7 @@ static float g_cos_storage[4 + 3]; static void cos4f( void ) { - float *store = (float *) ALIGN16( (unsigned) g_cos_storage ); + float *store = (float *) tgsi_align_128bit( g_cos_storage ); #ifdef WIN32 store[0] = (float) cos( (double) store[0] ); @@ -505,7 +513,7 @@ emit_cos( emit_func_call1( func, xmm_dst, - ALIGN16( (unsigned) g_cos_storage ), + p2u( tgsi_align_128bit( g_cos_storage ) ), cos4f ); } @@ -515,7 +523,7 @@ static float g_sin_storage[4 + 3]; static void sin4f( void ) { - float *store = (float *) ALIGN16( (unsigned) g_sin_storage ); + float *store = (float *) tgsi_align_128bit( g_sin_storage ); #ifdef WIN32 store[0] = (float) sin( (double) store[0] ); @@ -537,7 +545,7 @@ emit_sin (struct x86_function *func, emit_func_call1( func, xmm_dst, - ALIGN16( (unsigned) g_sin_storage ), + p2u( tgsi_align_128bit( g_sin_storage ) ), sin4f ); } @@ -570,7 +578,7 @@ static float g_pow_storage[4 + 4 + 3]; static void pow4f( void ) { - float *store = (float *) ALIGN16( (unsigned) g_pow_storage ); + float *store = (float *) tgsi_align_128bit( g_pow_storage ); #ifdef WIN32 store[0] = (float) pow( (double) store[0], (double) store[4] ); @@ -595,7 +603,7 @@ emit_pow( func, xmm_dst, xmm_src, - ALIGN16( (unsigned) g_pow_storage ), + p2u( tgsi_align_128bit( g_pow_storage ) ), pow4f ); } @@ -605,7 +613,7 @@ static float g_ex2_storage[4 + 3]; static void ex24f( void ) { - float *store = (float *) ALIGN16( (unsigned) g_ex2_storage ); + float *store = (float *) tgsi_align_128bit( g_ex2_storage ); #ifdef WIN32 store[0] = (float) pow( 2.0, (double) store[0] ); @@ -628,7 +636,7 @@ emit_ex2( emit_func_call1( func, xmm_dst, - ALIGN16( (unsigned) g_ex2_storage ), + p2u( tgsi_align_128bit( g_ex2_storage ) ), ex24f ); } @@ -638,7 +646,7 @@ static float g_lg2_storage[4 + 3]; static void lg24f( void ) { - float *store = (float *) ALIGN16( (unsigned) g_lg2_storage ); + float *store = (float *) tgsi_align_128bit( g_lg2_storage ); store[0] = LOG2( store[0] ); store[1] = LOG2( store[1] ); @@ -654,7 +662,7 @@ emit_lg2( emit_func_call1( func, xmm_dst, - ALIGN16( (unsigned) g_lg2_storage ), + p2u( tgsi_align_128bit( g_lg2_storage ) ), lg24f ); } @@ -664,7 +672,7 @@ static float g_flr_storage[4 + 3]; static void flr4f( void ) { - float *store = (float *) ALIGN16( (unsigned) g_flr_storage ); + float *store = (float *) tgsi_align_128bit( g_flr_storage ); store[0] = (float) floor( (double) store[0] ); store[1] = (float) floor( (double) store[1] ); @@ -680,7 +688,7 @@ emit_flr( emit_func_call1( func, xmm_dst, - ALIGN16( (unsigned) g_flr_storage ), + p2u( tgsi_align_128bit( g_flr_storage ) ), flr4f ); } @@ -690,7 +698,7 @@ static float g_frc_storage[4 + 3]; static void frc4f( void ) { - float *store = (float *) ALIGN16( (unsigned) g_frc_storage ); + float *store = (float *) tgsi_align_128bit( g_frc_storage ); store[0] -= (float) floor( (double) store[0] ); store[1] -= (float) floor( (double) store[1] ); @@ -706,7 +714,7 @@ emit_frc( emit_func_call1( func, xmm_dst, - ALIGN16( (unsigned) g_frc_storage ), + p2u( tgsi_align_128bit( g_frc_storage ) ), frc4f ); } @@ -1547,7 +1555,14 @@ emit_instruction( break; case TGSI_OPCODE_TEX: - assert( 0 ); + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } break; case TGSI_OPCODE_TXD: @@ -1747,10 +1762,6 @@ emit_instruction( assert( 0 ); break; - case TGSI_OPCODE_END: - /* nothing */ - break; - default: assert( 0 ); } @@ -1828,7 +1839,7 @@ emit_declaration( unsigned tgsi_emit_sse2( - const struct tgsi_token *tokens, + struct tgsi_token *tokens, struct x86_function *func ) { struct tgsi_parse_context parse; @@ -1893,7 +1904,7 @@ tgsi_emit_sse2( */ unsigned tgsi_emit_sse2_fs( - const struct tgsi_token *tokens, + struct tgsi_token *tokens, struct x86_function *func ) { struct tgsi_parse_context parse; @@ -1960,5 +1971,3 @@ tgsi_emit_sse2_fs( return 1; } - -#endif diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h index 60fdd5f61f..9bee371766 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h @@ -10,12 +10,12 @@ struct x86_function; unsigned tgsi_emit_sse2( - const struct tgsi_token *tokens, + struct tgsi_token *tokens, struct x86_function *function ); unsigned tgsi_emit_sse2_fs( - const struct tgsi_token *tokens, + struct tgsi_token *tokens, struct x86_function *function ); #if defined __cplusplus -- cgit v1.2.3 From 7362fe5d163a32cb8ba76de907ccf53dc69ff066 Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 13:35:46 +0100 Subject: Fix YCBCR macro. Fix line endings. --- src/mesa/pipe/p_format.h | 448 +++++++++++++++++++++++------------------------ 1 file changed, 224 insertions(+), 224 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index edb6fa4f1e..3046538f12 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -1,226 +1,226 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef PIPE_FORMAT_H -#define PIPE_FORMAT_H - -#include "p_compiler.h" - -#define PIPE_FORMAT_LAYOUT_RGBAZS 0 -#define PIPE_FORMAT_LAYOUT_YCBCR 1 - -#define PIPE_FORMAT_COMP_R 0 -#define PIPE_FORMAT_COMP_G 1 -#define PIPE_FORMAT_COMP_B 2 -#define PIPE_FORMAT_COMP_A 3 -#define PIPE_FORMAT_COMP_0 4 -#define PIPE_FORMAT_COMP_1 5 -#define PIPE_FORMAT_COMP_Z 6 -#define PIPE_FORMAT_COMP_S 7 - -#define PIPE_FORMAT_TYPE_UNKNOWN 0 -#define PIPE_FORMAT_TYPE_FLOAT 1 -#define PIPE_FORMAT_TYPE_UNORM 2 -#define PIPE_FORMAT_TYPE_SNORM 3 -#define PIPE_FORMAT_TYPE_USCALED 4 -#define PIPE_FORMAT_TYPE_SSCALED 5 - -struct pipe_format_header -{ - uint layout : 2; /**< PIPE_FORMAT_LAYOUT_ */ - uint padding : 30; -}; - -/** - * RGBAZS Format Layout. - */ -struct pipe_format_rgbazs -{ - uint layout : 2; /**< PIPE_FORMAT_LAYOUT_RGBAZS */ - uint swizzleX : 3; /**< PIPE_FORMAT_COMP_ */ - uint swizzleY : 3; /**< PIPE_FORMAT_COMP_ */ - uint swizzleZ : 3; /**< PIPE_FORMAT_COMP_ */ - uint swizzleW : 3; /**< PIPE_FORMAT_COMP_ */ - uint sizeX : 3; /**< Size of X - 1 */ - uint sizeY : 3; /**< Size of Y - 1 */ - uint sizeZ : 3; /**< Size of Z - 1 */ - uint sizeW : 3; /**< Size of W - 1 */ - uint exp8 : 2; /**< Scale size by 8 ^ exp8 */ - uint type : 3; /**< PIPE_FORMAT_TYPE_ */ - uint padding : 1; -}; - -#define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP8, TYPE ) (\ - (PIPE_FORMAT_LAYOUT_RGBAZS << 0) |\ - ((SWZ) << 2) |\ - ((SIZEX) << 14) |\ - ((SIZEY) << 17) |\ - ((SIZEZ) << 20) |\ - ((SIZEW) << 23) |\ - ((EXP8) << 26) |\ - ((TYPE) << 28) ) - -#define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 2) | ((SWZY) << 5) | ((SWZZ) << 8) | ((SWZW) << 11)) - -#define _PIPE_FORMAT_RGBAZS_1U( SWZ, SIZE, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 0, TYPE ) - -#define _PIPE_FORMAT_RGBAZS_1N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE ) - -#define _PIPE_FORMAT_RGBAZS_8U( SWZ, SIZE, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 1, TYPE ) - -#define _PIPE_FORMAT_RGBAZS_8N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) - -#define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 2, TYPE ) - -#define _PIPE_FORMAT_R000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_RG00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_A ) -#define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) -#define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) -#define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) -#define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R ) -#define _PIPE_FORMAT_RRR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) -#define _PIPE_FORMAT_RRRR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R ) -#define _PIPE_FORMAT_RRRG _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G ) -#define _PIPE_FORMAT_Z000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) - -struct pipe_format_ycbcr -{ - uint layout : 2; /**< PIPE_FORMAT_LAYOUT_YCBCR */ - uint reversed : 1; - uint padding : 29; -}; - -#define _PIPE_FORMAT_YCBCR( REV ) (\ - (PIPE_FORMAT_LAYOUT_YCBCR << 0) |\ - ((REVERSED) << 2) ) - -union pipe_format -{ - uint value32; - struct pipe_format_header header; - struct pipe_format_rgbazs rgbazs; - struct pipe_format_ycbcr ycbcr; -}; - -/** - * Texture/surface image formats (preliminary) - */ - -/* KW: Added lots of surface formats to support vertex element layout - * definitions, and eventually render-to-vertex-buffer. Could - * consider making float/int/uint/scaled/normalized a separate - * parameter, but on the other hand there are special cases like - * z24s8, compressed textures, ycbcr, etc that won't fit that model. - */ - -#define PIPE_FORMAT_NONE _PIPE_FORMAT_RGBAZS_1U( _PIPE_FORMAT_0000, 0, PIPE_FORMAT_TYPE_UNKNOWN ) /**< unstructured */ -#define PIPE_FORMAT_U_A8_R8_G8_B8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_ARGB, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte[4] ARGB */ -#define PIPE_FORMAT_U_A1_R5_G5_B5 _PIPE_FORMAT_RGBAZS_1N( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGBA */ -#define PIPE_FORMAT_U_A4_R4_G4_B4 _PIPE_FORMAT_RGBAZS_1U( _PIPE_FORMAT_ARGB, 4, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGBA */ -#define PIPE_FORMAT_U_R5_G6_B5 _PIPE_FORMAT_RGBAZS_1N( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGB */ -#define PIPE_FORMAT_U_L8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRR1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte luminance */ -#define PIPE_FORMAT_U_A8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_000R, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha */ -#define PIPE_FORMAT_U_I8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRRR, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte intensity */ -#define PIPE_FORMAT_U_A8_L8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRRG, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha, luminance */ -#define PIPE_FORMAT_S_R16_G16_B16_A16 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RGBA, 2, PIPE_FORMAT_TYPE_SNORM ) /**< signed 16-bit RGBA (accum) */ -#define PIPE_FORMAT_YCBCR _PIPE_FORMAT_YCBCR( 0 ) -#define PIPE_FORMAT_YCBCR_REV _PIPE_FORMAT_YCBCR( 1 ) -#define PIPE_FORMAT_U_Z16 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< ushort Z/depth */ -#define PIPE_FORMAT_U_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< uint Z/depth */ -#define PIPE_FORMAT_F_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) /**< float Z/depth */ -#define PIPE_FORMAT_S8_Z24 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil + 24-bit Z */ -#define PIPE_FORMAT_U_S8 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil */ -#define PIPE_FORMAT_R64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R64G64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R64G64B64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R64G64B64A64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32G32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32G32B32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32G32B32A32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32G32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32G32B32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32G32B32A32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32G32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32G32B32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32G32B32A32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32G32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32G32B32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32G32B32A32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R32G32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R32G32B32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R32G32B32A32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16G16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16G16B16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16G16B16A16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16G16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16G16B16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16G16B16A16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16G16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16G16B16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16G16B16A16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16G16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16G16B16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16G16B16A16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8G8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8G8B8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8G8B8A8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8G8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8G8B8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8G8B8A8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8G8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8G8B8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8G8B8A8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8G8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8G8B8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8G8B8A8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) - +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef PIPE_FORMAT_H +#define PIPE_FORMAT_H + +#include "p_compiler.h" + +#define PIPE_FORMAT_LAYOUT_RGBAZS 0 +#define PIPE_FORMAT_LAYOUT_YCBCR 1 + +#define PIPE_FORMAT_COMP_R 0 +#define PIPE_FORMAT_COMP_G 1 +#define PIPE_FORMAT_COMP_B 2 +#define PIPE_FORMAT_COMP_A 3 +#define PIPE_FORMAT_COMP_0 4 +#define PIPE_FORMAT_COMP_1 5 +#define PIPE_FORMAT_COMP_Z 6 +#define PIPE_FORMAT_COMP_S 7 + +#define PIPE_FORMAT_TYPE_UNKNOWN 0 +#define PIPE_FORMAT_TYPE_FLOAT 1 +#define PIPE_FORMAT_TYPE_UNORM 2 +#define PIPE_FORMAT_TYPE_SNORM 3 +#define PIPE_FORMAT_TYPE_USCALED 4 +#define PIPE_FORMAT_TYPE_SSCALED 5 + +struct pipe_format_header +{ + uint layout : 2; /**< PIPE_FORMAT_LAYOUT_ */ + uint padding : 30; +}; + +/** + * RGBAZS Format Layout. + */ +struct pipe_format_rgbazs +{ + uint layout : 2; /**< PIPE_FORMAT_LAYOUT_RGBAZS */ + uint swizzleX : 3; /**< PIPE_FORMAT_COMP_ */ + uint swizzleY : 3; /**< PIPE_FORMAT_COMP_ */ + uint swizzleZ : 3; /**< PIPE_FORMAT_COMP_ */ + uint swizzleW : 3; /**< PIPE_FORMAT_COMP_ */ + uint sizeX : 3; /**< Size of X - 1 */ + uint sizeY : 3; /**< Size of Y - 1 */ + uint sizeZ : 3; /**< Size of Z - 1 */ + uint sizeW : 3; /**< Size of W - 1 */ + uint exp8 : 2; /**< Scale size by 8 ^ exp8 */ + uint type : 3; /**< PIPE_FORMAT_TYPE_ */ + uint padding : 1; +}; + +#define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP8, TYPE ) (\ + (PIPE_FORMAT_LAYOUT_RGBAZS << 0) |\ + ((SWZ) << 2) |\ + ((SIZEX) << 14) |\ + ((SIZEY) << 17) |\ + ((SIZEZ) << 20) |\ + ((SIZEW) << 23) |\ + ((EXP8) << 26) |\ + ((TYPE) << 28) ) + +#define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 2) | ((SWZY) << 5) | ((SWZZ) << 8) | ((SWZW) << 11)) + +#define _PIPE_FORMAT_RGBAZS_1U( SWZ, SIZE, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 0, TYPE ) + +#define _PIPE_FORMAT_RGBAZS_1N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE ) + +#define _PIPE_FORMAT_RGBAZS_8U( SWZ, SIZE, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 1, TYPE ) + +#define _PIPE_FORMAT_RGBAZS_8N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) + +#define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 2, TYPE ) + +#define _PIPE_FORMAT_R000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_RG00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) +#define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R ) +#define _PIPE_FORMAT_RRR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) +#define _PIPE_FORMAT_RRRR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R ) +#define _PIPE_FORMAT_RRRG _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G ) +#define _PIPE_FORMAT_Z000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) + +struct pipe_format_ycbcr +{ + uint layout : 2; /**< PIPE_FORMAT_LAYOUT_YCBCR */ + uint reversed : 1; + uint padding : 29; +}; + +#define _PIPE_FORMAT_YCBCR( REV ) (\ + (PIPE_FORMAT_LAYOUT_YCBCR << 0) |\ + ((REV) << 2) ) + +union pipe_format +{ + uint value32; + struct pipe_format_header header; + struct pipe_format_rgbazs rgbazs; + struct pipe_format_ycbcr ycbcr; +}; + +/** + * Texture/surface image formats (preliminary) + */ + +/* KW: Added lots of surface formats to support vertex element layout + * definitions, and eventually render-to-vertex-buffer. Could + * consider making float/int/uint/scaled/normalized a separate + * parameter, but on the other hand there are special cases like + * z24s8, compressed textures, ycbcr, etc that won't fit that model. + */ + +#define PIPE_FORMAT_NONE _PIPE_FORMAT_RGBAZS_1U( _PIPE_FORMAT_0000, 0, PIPE_FORMAT_TYPE_UNKNOWN ) /**< unstructured */ +#define PIPE_FORMAT_U_A8_R8_G8_B8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_ARGB, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte[4] ARGB */ +#define PIPE_FORMAT_U_A1_R5_G5_B5 _PIPE_FORMAT_RGBAZS_1N( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGBA */ +#define PIPE_FORMAT_U_A4_R4_G4_B4 _PIPE_FORMAT_RGBAZS_1U( _PIPE_FORMAT_ARGB, 4, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGBA */ +#define PIPE_FORMAT_U_R5_G6_B5 _PIPE_FORMAT_RGBAZS_1N( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGB */ +#define PIPE_FORMAT_U_L8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRR1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte luminance */ +#define PIPE_FORMAT_U_A8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_000R, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha */ +#define PIPE_FORMAT_U_I8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRRR, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte intensity */ +#define PIPE_FORMAT_U_A8_L8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRRG, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha, luminance */ +#define PIPE_FORMAT_S_R16_G16_B16_A16 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RGBA, 2, PIPE_FORMAT_TYPE_SNORM ) /**< signed 16-bit RGBA (accum) */ +#define PIPE_FORMAT_YCBCR _PIPE_FORMAT_YCBCR( 0 ) +#define PIPE_FORMAT_YCBCR_REV _PIPE_FORMAT_YCBCR( 1 ) +#define PIPE_FORMAT_U_Z16 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< ushort Z/depth */ +#define PIPE_FORMAT_U_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< uint Z/depth */ +#define PIPE_FORMAT_F_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) /**< float Z/depth */ +#define PIPE_FORMAT_S8_Z24 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil + 24-bit Z */ +#define PIPE_FORMAT_U_S8 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil */ +#define PIPE_FORMAT_R64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R64G64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R64G64B64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R64G64B64A64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32G32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32G32B32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32G32B32A32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32G32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32G32B32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32G32B32A32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32G32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32G32B32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32G32B32A32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32G32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32G32B32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32G32B32A32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32G32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32G32B32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32G32B32A32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16G16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16G16B16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16G16B16A16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16G16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16G16B16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16G16B16A16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16G16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16G16B16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16G16B16A16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16G16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16G16B16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16G16B16A16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8G8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8G8B8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8G8B8A8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8G8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8G8B8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8G8B8A8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8G8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8G8B8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8G8B8A8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8G8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8G8B8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8G8B8A8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) + /* Duplicated formats: */ -#define PIPE_FORMAT_U_R8_G8_B8_A8 PIPE_FORMAT_R8G8B8A8_UNORM - -#endif +#define PIPE_FORMAT_U_R8_G8_B8_A8 PIPE_FORMAT_R8G8B8A8_UNORM + +#endif -- cgit v1.2.3 From 31b4b261083e546998eba37178ac196049e4e501 Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 14:01:39 +0100 Subject: Enable SSE2 for vertex shaders. --- src/mesa/pipe/draw/draw_private.h | 2 ++ src/mesa/pipe/draw/draw_vertex_shader.c | 22 ++++++++++------------ 2 files changed, 12 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 44cf0b1c98..ac47d2a76f 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -124,7 +124,9 @@ struct draw_stage */ struct draw_vertex_shader { const struct pipe_shader_state *state; +#if defined(__i386__) || defined(__386__) struct x86_function sse2_program; +#endif }; /** diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index d8bd5a96d6..a2e64b1e4c 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -100,20 +100,20 @@ run_vertex_program(struct draw_context *draw, draw_vertex_fetch( draw, machine, elts, count ); /* run shader */ - if( draw->vertex_shader->state->executable != NULL ) { +#if defined(__i386__) || defined(__386__) + { /* SSE */ - codegen_function func = (codegen_function) draw->vertex_shader->state->executable; + codegen_function func = (codegen_function) x86_get_func( &draw->vertex_shader->sse2_program ); func( machine->Inputs, machine->Outputs, machine->Consts, machine->Temps ); } - else { - /* interpreter */ - tgsi_exec_machine_run( machine ); - } - +#else + /* interpreter */ + tgsi_exec_machine_run( machine ); +#endif /* store machine results */ for (j = 0; j < count; j++) { @@ -225,11 +225,9 @@ draw_create_vertex_shader(struct draw_context *draw, /* cast-away const */ struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; - x86_init_func( &sh->sse2_program ); - - tgsi_emit_sse2( sh->tokens, &sh->sse2_program ); + x86_init_func( &vs->sse2_program ); - sh->executable = x86_get_func( &sh->sse2_program ); + tgsi_emit_sse2( sh->tokens, &vs->sse2_program ); } #endif @@ -255,7 +253,7 @@ void draw_delete_vertex_shader(struct draw_context *draw, struct draw_vertex_shader *vs = (struct draw_vertex_shader*)(vcso); #if defined(__i386__) || defined(__386__) - x86_release_func((struct x86_function *) &vs->state->sse2_program); + x86_release_func((struct x86_function *) &vs->sse2_program); #endif free((void *) vs->state); -- cgit v1.2.3 From 1ab8f6e69615c4f39110cf9202ae4b52238c0bf2 Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 14:05:13 +0100 Subject: Enable SSE2 for fragment shaders. --- src/mesa/pipe/softpipe/sp_context.h | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 21 ++++++++++++--------- src/mesa/pipe/softpipe/sp_state.h | 11 +++++++++++ src/mesa/pipe/softpipe/sp_state_fs.c | 12 +++++------- 4 files changed, 29 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index e51e81c7b8..548151b378 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -79,7 +79,7 @@ struct softpipe_context { const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; - const struct pipe_shader_state *fs; + const struct sp_fragment_shader_state *fs; const struct sp_vertex_shader_state *vs; struct pipe_blend_color blend_color; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index dc8d6e0e23..f079ab9583 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -37,7 +37,10 @@ #include "pipe/llvm/llvmtgsi.h" +#include "x86/rtasm/x86sse.h" + #include "sp_context.h" +#include "sp_state.h" #include "sp_headers.h" #include "sp_quad.h" #include "sp_tex_sample.h" @@ -100,8 +103,8 @@ shade_quad( /* run shader */ /* XXX: Generated code effectively unusable until it handles quad->mask */ - if( !quad->mask && softpipe->fs->executable != NULL ) { - codegen_function func = (codegen_function) softpipe->fs->executable; + if( !quad->mask ) { + codegen_function func = (codegen_function) x86_get_func( &softpipe->fs->sse2_program ); func( machine->Inputs, machine->Outputs, @@ -119,7 +122,7 @@ shade_quad( /* store result color */ if (qss->colorOutSlot >= 0) { /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->output_semantic_name[qss->colorOutSlot] + assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] == TGSI_SEMANTIC_COLOR); memcpy( quad->outputs.color, @@ -167,15 +170,15 @@ static void shade_begin(struct quad_stage *qs) /* XXX only do this if the fragment shader changes... */ tgsi_exec_machine_init(&qss->machine, - softpipe->fs->tokens, + softpipe->fs->shader.tokens, PIPE_MAX_SAMPLERS, qss->samplers ); /* find output slots for depth, color */ qss->colorOutSlot = -1; qss->depthOutSlot = -1; - for (i = 0; i < qss->stage.softpipe->fs->num_outputs; i++) { - switch (qss->stage.softpipe->fs->output_semantic_name[i]) { + for (i = 0; i < qss->stage.softpipe->fs->shader.num_outputs; i++) { + switch (qss->stage.softpipe->fs->shader.output_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: qss->depthOutSlot = i; break; @@ -190,9 +193,9 @@ static void shade_begin(struct quad_stage *qs) } -static void shade_destroy(struct quad_stage *qs) -{ - free( qs ); +static void shade_destroy(struct quad_stage *qs) +{ + free( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 61532bcdeb..676ad06bfc 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -33,6 +33,17 @@ #include "pipe/p_state.h" +#include "x86/rtasm/x86sse.h" + +/** + * Softpipe fs state is derived from pipe_shader_state. + */ +struct sp_fragment_shader_state { + struct pipe_shader_state shader; +#if defined(__i386__) || defined(__386__) + struct x86_function sse2_program; +#endif +}; void * softpipe_create_alpha_test_state(struct pipe_context *, diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 5547c849ac..86aa80c0fc 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -43,16 +43,14 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, * that now. */ - struct pipe_shader_state *state = malloc(sizeof(struct pipe_shader_state)); - memcpy(state, templ, sizeof(struct pipe_shader_state)); + struct sp_fragment_shader_state *state = malloc(sizeof(struct sp_fragment_shader_state)); + state->shader = *templ; #if defined(__i386__) || defined(__386__) if (softpipe->use_sse) { x86_init_func( &state->sse2_program ); - tgsi_emit_sse2_fs( state->tokens, &state->sse2_program ); - - state->executable = (void *)x86_get_func( &state->sse2_program ); + tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); } #endif @@ -63,7 +61,7 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->fs = (struct pipe_shader_state *)fs; + softpipe->fs = (struct sp_fragment_shader_state *) fs; softpipe->dirty |= SP_NEW_FS; } @@ -72,7 +70,7 @@ void softpipe_delete_fs_state(struct pipe_context *pipe, void *shader) { #if defined(__i386__) || defined(__386__) - struct pipe_shader_state *state = shader; + struct sp_fragment_shader_state *state = shader; x86_release_func( &state->sse2_program ); #endif -- cgit v1.2.3 From e9ce69b1c9cb6a84e4c0403cf3090653e125c91d Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 14:06:14 +0100 Subject: Remove PIPE_FORMAT_COUNT references. --- src/mesa/state_tracker/st_format.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 7564c6014e..cd44f19a76 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -247,12 +247,12 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, GLenum format, GLenum type) { const GLuint *supported; - GLboolean allow[PIPE_FORMAT_COUNT]; + GLboolean allow[256]; /* XXX: this will go away */ GLuint i, n; /* query supported formats and fill in bool allow[] table */ supported = pipe->supported_formats(pipe, &n); - assert(n < PIPE_FORMAT_COUNT); /* sanity check */ + assert(n < 256); /* sanity check */ /* XXX: this will go away */ memset(allow, 0, sizeof(allow)); for (i = 0; i < n; i++) { allow[supported[i]] = 1; -- cgit v1.2.3 From f27dcb51f70d6a2efab39770ddeff3d5603deaaa Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 14:40:22 +0100 Subject: Respect use_sse flag. --- src/mesa/pipe/draw/draw_vertex_shader.c | 27 +++++++++++++++------------ 1 file changed, 15 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index a2e64b1e4c..263b5a7d9e 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -101,7 +101,7 @@ run_vertex_program(struct draw_context *draw, /* run shader */ #if defined(__i386__) || defined(__386__) - { + if (draw->use_sse) { /* SSE */ codegen_function func = (codegen_function) x86_get_func( &draw->vertex_shader->sse2_program ); func( @@ -110,9 +110,12 @@ run_vertex_program(struct draw_context *draw, machine->Consts, machine->Temps ); } + else #else - /* interpreter */ - tgsi_exec_machine_run( machine ); + { + /* interpreter */ + tgsi_exec_machine_run( machine ); + } #endif /* store machine results */ @@ -217,16 +220,17 @@ void * draw_create_vertex_shader(struct draw_context *draw, const struct pipe_shader_state *shader) { - struct draw_vertex_shader *vs = calloc(1, sizeof(struct draw_vertex_shader)); + struct draw_vertex_shader *vs; + vs = calloc( 1, sizeof( struct draw_vertex_shader ) ); vs->state = shader; + #if defined(__i386__) || defined(__386__) if (draw->use_sse) { /* cast-away const */ struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; x86_init_func( &vs->sse2_program ); - tgsi_emit_sse2( sh->tokens, &vs->sse2_program ); } #endif @@ -250,15 +254,14 @@ void draw_bind_vertex_shader(struct draw_context *draw, void draw_delete_vertex_shader(struct draw_context *draw, void *vcso) { - struct draw_vertex_shader *vs = (struct draw_vertex_shader*)(vcso); + struct draw_vertex_shader *vs; + + vs = (struct draw_vertex_shader *) vcso; #if defined(__i386__) || defined(__386__) - x86_release_func((struct x86_function *) &vs->sse2_program); + x86_release_func( (struct x86_function *) &vs->sse2_program ); #endif - free((void *) vs->state); - free(vcso); + free( vs->state ); + free( vs ); } - - - -- cgit v1.2.3 From b23350700e476e397fefefa45f77d2a5a33c111a Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 14:41:16 +0100 Subject: vertex_element::src_format needs 32 bits. --- src/mesa/pipe/p_state.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index a61cbaf8e1..0e1a038c2e 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -368,7 +368,7 @@ struct pipe_vertex_element unsigned vertex_buffer_index:5; unsigned dst_offset:8; - unsigned src_format:8; /**< PIPE_FORMAT_* */ + unsigned src_format; /**< PIPE_FORMAT_* */ }; -- cgit v1.2.3 From 478b5692de2d9be0ff96b5a61a79f18c8ff64bb6 Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 14:42:17 +0100 Subject: Add #ifs. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index f079ab9583..1ff4406a4b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -102,6 +102,7 @@ shade_quad( machine->Inputs[0].xyzw[1].f[3] = fy + 1.0f; /* run shader */ +#if defined(__i386__) || defined(__386__) /* XXX: Generated code effectively unusable until it handles quad->mask */ if( !quad->mask ) { codegen_function func = (codegen_function) x86_get_func( &softpipe->fs->sse2_program ); @@ -112,7 +113,9 @@ shade_quad( machine->Temps, machine->InterpCoefs ); } - else { + else +#endif + { #ifdef MESA_LLVM /*ga_llvm_prog_exec(softpipe->fs->llvm_prog);*/ #endif -- cgit v1.2.3 From 9dfc27edf11b3fd00c364894578f5a3a05c4ca54 Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 15:07:09 +0100 Subject: Fix SSE bug. --- src/mesa/pipe/draw/draw_vertex_shader.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 263b5a7d9e..9dbb317f2a 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -111,12 +111,11 @@ run_vertex_program(struct draw_context *draw, machine->Temps ); } else -#else +#endif { /* interpreter */ tgsi_exec_machine_run( machine ); } -#endif /* store machine results */ for (j = 0; j < count; j++) { -- cgit v1.2.3 From 205d4e49001123ea0bb10139aa3236cc1f2141d0 Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 15:08:00 +0100 Subject: Respect use_sse flag. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/pipe/softpipe/sp_state_fs.c | 5 ++--- 2 files changed, 3 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 1ff4406a4b..701a68ec6a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -104,7 +104,7 @@ shade_quad( /* run shader */ #if defined(__i386__) || defined(__386__) /* XXX: Generated code effectively unusable until it handles quad->mask */ - if( !quad->mask ) { + if( !quad->mask && softpipe->use_sse ) { codegen_function func = (codegen_function) x86_get_func( &softpipe->fs->sse2_program ); func( machine->Inputs, diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 86aa80c0fc..a94ec1e92c 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -49,7 +49,6 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, #if defined(__i386__) || defined(__386__) if (softpipe->use_sse) { x86_init_func( &state->sse2_program ); - tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); } #endif @@ -69,13 +68,13 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) void softpipe_delete_fs_state(struct pipe_context *pipe, void *shader) { -#if defined(__i386__) || defined(__386__) struct sp_fragment_shader_state *state = shader; +#if defined(__i386__) || defined(__386__) x86_release_func( &state->sse2_program ); #endif - free(shader); + free( state ); } -- cgit v1.2.3 From d8b59ba1775ea730130fa0e2a1f054161d948820 Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 15:55:11 +0100 Subject: Refactor supported format queries. --- src/mesa/state_tracker/st_format.c | 140 +++++++++++++++++++------------------ 1 file changed, 73 insertions(+), 67 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index cd44f19a76..5ccbe721eb 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -175,20 +175,39 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) } } +/* XXX: This function should be implemented by pipe object. */ +static GLboolean +allow_format( + struct pipe_context *pipe, + unsigned format ) +{ + const GLuint *supported; + GLuint i, n; + + supported = pipe->supported_formats( pipe, &n ); + for (i = 0; i < n; i++) { + if (supported[i] == format) { + return GL_TRUE; + } + } + return GL_FALSE; +} /** * Search list of formats for first RGBA format. */ static GLuint -default_rgba_format(const GLuint formats[], GLuint num) +default_rgba_format( + struct pipe_context *pipe ) { - GLuint i; - for (i = 0; i < num; i++) { - if (formats[i] == PIPE_FORMAT_U_R8_G8_B8_A8 || - formats[i] == PIPE_FORMAT_U_A8_R8_G8_B8 || - formats[i] == PIPE_FORMAT_U_R5_G6_B5) { - return formats[i]; - } + if (allow_format( pipe, PIPE_FORMAT_U_R8_G8_B8_A8 )) { + return PIPE_FORMAT_U_R8_G8_B8_A8; + } + if (allow_format( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) { + return PIPE_FORMAT_U_A8_R8_G8_B8; + } + if (allow_format( pipe, PIPE_FORMAT_U_R5_G6_B5 )) { + return PIPE_FORMAT_U_R5_G6_B5; } return PIPE_FORMAT_NONE; } @@ -198,13 +217,11 @@ default_rgba_format(const GLuint formats[], GLuint num) * Search list of formats for first RGBA format with >8 bits/channel. */ static GLuint -default_deep_rgba_format(const GLuint formats[], GLuint num) +default_deep_rgba_format( + struct pipe_context *pipe ) { - GLuint i; - for (i = 0; i < num; i++) { - if (formats[i] == PIPE_FORMAT_S_R16_G16_B16_A16) { - return formats[i]; - } + if (allow_format( pipe, PIPE_FORMAT_S_R16_G16_B16_A16 )) { + return PIPE_FORMAT_S_R16_G16_B16_A16; } return PIPE_FORMAT_NONE; } @@ -214,20 +231,21 @@ default_deep_rgba_format(const GLuint formats[], GLuint num) * Search list of formats for first depth/Z format. */ static GLuint -default_depth_format(const GLuint formats[], GLuint num) +default_depth_format( + struct pipe_context *pipe ) { - GLuint i; - for (i = 0; i < num; i++) { - if (formats[i] == PIPE_FORMAT_U_Z16 || - formats[i] == PIPE_FORMAT_U_Z32 || - formats[i] == PIPE_FORMAT_S8_Z24) { - return formats[i]; - } + if (allow_format( pipe, PIPE_FORMAT_U_Z16 )) { + return PIPE_FORMAT_U_Z16; + } + if (allow_format( pipe, PIPE_FORMAT_U_Z32 )) { + return PIPE_FORMAT_U_Z32; + } + if (allow_format( pipe, PIPE_FORMAT_S8_Z24 )) { + return PIPE_FORMAT_S8_Z24; } return PIPE_FORMAT_NONE; } - /** * Choose the PIPE_FORMAT_ to use for storing a texture image based * on the user's internalFormat, format and type parameters. @@ -246,77 +264,65 @@ GLuint st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, GLenum format, GLenum type) { - const GLuint *supported; - GLboolean allow[256]; /* XXX: this will go away */ - GLuint i, n; - - /* query supported formats and fill in bool allow[] table */ - supported = pipe->supported_formats(pipe, &n); - assert(n < 256); /* sanity check */ /* XXX: this will go away */ - memset(allow, 0, sizeof(allow)); - for (i = 0; i < n; i++) { - allow[supported[i]] = 1; - } - switch (internalFormat) { case 4: case GL_RGBA: case GL_COMPRESSED_RGBA: if (format == GL_BGRA) { if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - if (allow[PIPE_FORMAT_U_A8_R8_G8_B8]) + if (allow_format( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) return PIPE_FORMAT_U_A8_R8_G8_B8; } else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - if (allow[PIPE_FORMAT_U_A4_R4_G4_B4]) + if (allow_format( pipe, PIPE_FORMAT_U_A4_R4_G4_B4 )) return PIPE_FORMAT_U_A4_R4_G4_B4; } else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) + if (allow_format( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) return PIPE_FORMAT_U_A1_R5_G5_B5; } } - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case 3: case GL_RGB: case GL_COMPRESSED_RGB: if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - if (allow[PIPE_FORMAT_U_R5_G6_B5]) + if (allow_format( pipe, PIPE_FORMAT_U_R5_G6_B5 )) return PIPE_FORMAT_U_R5_G6_B5; } - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case GL_RGBA16: - return default_deep_rgba_format(supported, n); + return default_deep_rgba_format( pipe ); case GL_RGBA4: case GL_RGBA2: - if (allow[PIPE_FORMAT_U_A4_R4_G4_B4]) + if (allow_format( pipe, PIPE_FORMAT_U_A4_R4_G4_B4 )) return PIPE_FORMAT_U_A4_R4_G4_B4; - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case GL_RGB5_A1: - if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) + if (allow_format( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) return PIPE_FORMAT_U_A1_R5_G5_B5; - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - if (allow[PIPE_FORMAT_U_A1_R5_G5_B5]) + if (allow_format( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) return PIPE_FORMAT_U_A1_R5_G5_B5; - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case GL_ALPHA: case GL_ALPHA4: @@ -324,9 +330,9 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - if (allow[PIPE_FORMAT_U_A8]) + if (allow_format( pipe, PIPE_FORMAT_U_A8 )) return PIPE_FORMAT_U_A8; - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case 1: case GL_LUMINANCE: @@ -335,9 +341,9 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - if (allow[PIPE_FORMAT_U_A8]) + if (allow_format( pipe, PIPE_FORMAT_U_A8 )) return PIPE_FORMAT_U_A8; - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case 2: case GL_LUMINANCE_ALPHA: @@ -348,9 +354,9 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - if (allow[PIPE_FORMAT_U_A8_L8]) + if (allow_format( pipe, PIPE_FORMAT_U_A8_L8 )) return PIPE_FORMAT_U_A8_L8; - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case GL_INTENSITY: case GL_INTENSITY4: @@ -358,17 +364,17 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - if (allow[PIPE_FORMAT_U_I8]) + if (allow_format( pipe, PIPE_FORMAT_U_I8 )) return PIPE_FORMAT_U_I8; - return default_rgba_format(supported, n); + return default_rgba_format( pipe ); case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) { - if (allow[PIPE_FORMAT_YCBCR]) + if (allow_format( pipe, PIPE_FORMAT_YCBCR )) return PIPE_FORMAT_YCBCR; } else { - if (allow[PIPE_FORMAT_YCBCR_REV]) + if (allow_format( pipe, PIPE_FORMAT_YCBCR_REV )) return PIPE_FORMAT_YCBCR_REV; } return PIPE_FORMAT_NONE; @@ -397,34 +403,34 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, #endif case GL_DEPTH_COMPONENT16: - if (allow[PIPE_FORMAT_U_Z16]) + if (allow_format( pipe, PIPE_FORMAT_U_Z16 )) return PIPE_FORMAT_U_Z16; /* fall-through */ case GL_DEPTH_COMPONENT24: - if (allow[PIPE_FORMAT_S8_Z24]) + if (allow_format( pipe, PIPE_FORMAT_S8_Z24 )) return PIPE_FORMAT_S8_Z24; /* fall-through */ case GL_DEPTH_COMPONENT32: - if (allow[PIPE_FORMAT_U_Z32]) + if (allow_format( pipe, PIPE_FORMAT_U_Z32 )) return PIPE_FORMAT_U_Z32; /* fall-through */ case GL_DEPTH_COMPONENT: - return default_depth_format(supported, n); + return default_depth_format( pipe ); case GL_STENCIL_INDEX: case GL_STENCIL_INDEX1_EXT: case GL_STENCIL_INDEX4_EXT: case GL_STENCIL_INDEX8_EXT: case GL_STENCIL_INDEX16_EXT: - if (allow[PIPE_FORMAT_U_S8]) + if (allow_format( pipe, PIPE_FORMAT_U_S8 )) return PIPE_FORMAT_U_S8; - if (allow[PIPE_FORMAT_S8_Z24]) + if (allow_format( pipe, PIPE_FORMAT_S8_Z24 )) return PIPE_FORMAT_S8_Z24; return PIPE_FORMAT_NONE; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - if (allow[PIPE_FORMAT_S8_Z24]) + if (allow_format( pipe, PIPE_FORMAT_S8_Z24 )) return PIPE_FORMAT_S8_Z24; return PIPE_FORMAT_NONE; -- cgit v1.2.3 From f16f23b3ab4caf6588ce713fc682aac6e9f808fb Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 17:30:23 +0100 Subject: Use PIPE_FORMAT in state tracker. Fix PIPE_FORMAT field encoding. Re-implement st_get_format_info. --- src/mesa/pipe/p_format.h | 2 +- src/mesa/state_tracker/st_cb_fbo.c | 23 ++--- src/mesa/state_tracker/st_format.c | 183 +++++++++++++++++++++++++++++++++++-- src/mesa/state_tracker/st_format.h | 6 +- 4 files changed, 192 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index 3046538f12..1cab3c27b2 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -84,7 +84,7 @@ struct pipe_format_rgbazs ((EXP8) << 26) |\ ((TYPE) << 28) ) -#define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 2) | ((SWZY) << 5) | ((SWZZ) << 8) | ((SWZW) << 11)) +#define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 0) | ((SWZY) << 3) | ((SWZZ) << 6) | ((SWZW) << 9)) #define _PIPE_FORMAT_RGBAZS_1U( SWZ, SIZE, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 0, TYPE ) diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 160edc4274..5512bdeed8 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -62,26 +62,27 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, struct st_renderbuffer *strb = st_renderbuffer(rb); const GLuint pipeFormat = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE); - const struct pipe_format_info *info = st_get_format_info(pipeFormat); + struct pipe_format_info info; GLuint cpp; GLbitfield flags = PIPE_SURFACE_FLAG_RENDER; /* want to render to surface */ - assert(info); - if (!info) + if (!st_get_format_info( pipeFormat, &info )) { + assert( 0 ); return GL_FALSE; + } - strb->Base._ActualFormat = info->base_format; - strb->Base.RedBits = info->red_bits; - strb->Base.GreenBits = info->green_bits; - strb->Base.BlueBits = info->blue_bits; - strb->Base.AlphaBits = info->alpha_bits; - strb->Base.DepthBits = info->depth_bits; - strb->Base.StencilBits = info->stencil_bits; + strb->Base._ActualFormat = info.base_format; + strb->Base.RedBits = info.red_bits; + strb->Base.GreenBits = info.green_bits; + strb->Base.BlueBits = info.blue_bits; + strb->Base.AlphaBits = info.alpha_bits; + strb->Base.DepthBits = info.depth_bits; + strb->Base.StencilBits = info.stencil_bits; strb->Base.DataType = st_format_datatype(pipeFormat); assert(strb->Base.DataType); - cpp = info->size; + cpp = info.size; if (!strb->surface) { strb->surface = pipe->winsys->surface_alloc(pipe->winsys, pipeFormat); diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 5ccbe721eb..5a50d2d63a 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -36,20 +36,74 @@ #include "main/texstore.h" #include "main/texformat.h" #include "main/enums.h" +#include "main/macros.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "st_context.h" #include "st_format.h" +static GLuint +format_bits( + struct pipe_format_rgbazs info, + GLuint comp ) +{ + GLuint size; + if (info.swizzleX == comp) { + size = info.sizeX; + } + else if (info.swizzleY == comp) { + size = info.sizeY; + } + else if (info.swizzleZ == comp) { + size = info.sizeZ; + } + else if (info.swizzleW == comp) { + size = info.sizeW; + } + else { + size = 0; + } + return size << (info.exp8 * 3); +} + +static GLuint +format_max_bits( + struct pipe_format_rgbazs info ) +{ + GLuint size = format_bits( info, PIPE_FORMAT_COMP_R ); + + size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_G ) ); + size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_B ) ); + size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_A ) ); + size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_Z ) ); + size = MAX2( size, format_bits( info, PIPE_FORMAT_COMP_S ) ); + return size; +} + +static GLuint +format_size( + struct pipe_format_rgbazs info ) +{ + return + format_bits( info, PIPE_FORMAT_COMP_R ) + + format_bits( info, PIPE_FORMAT_COMP_G ) + + format_bits( info, PIPE_FORMAT_COMP_B ) + + format_bits( info, PIPE_FORMAT_COMP_A ) + + format_bits( info, PIPE_FORMAT_COMP_Z ) + + format_bits( info, PIPE_FORMAT_COMP_S ); +} /* * XXX temporary here */ -const struct pipe_format_info * -st_get_format_info(GLuint format) +GLboolean +st_get_format_info( + GLuint format, + struct pipe_format_info *pinfo ) { +#if 0 static const struct pipe_format_info info[] = { { PIPE_FORMAT_U_R8_G8_B8_A8, /* format */ @@ -124,6 +178,113 @@ st_get_format_info(GLuint format) return info + i; } return NULL; +#endif + + union pipe_format fmt; + + fmt.value32 = format; + if (fmt.header.layout == PIPE_FORMAT_LAYOUT_RGBAZS) { + struct pipe_format_rgbazs info; + + info = fmt.rgbazs; + +#if 0 + printf( + "PIPE_FORMAT: X(%u), Y(%u), Z(%u), W(%u)\n", + info.sizeX, + info.sizeY, + info.sizeZ, + info.sizeW ); +#endif + + /* Data type */ + if (format == PIPE_FORMAT_U_A1_R5_G5_B5 || format == PIPE_FORMAT_U_R5_G6_B5) { + pinfo->datatype = GL_UNSIGNED_SHORT; + } + else { + GLuint size; + + assert( info.type == PIPE_FORMAT_TYPE_UNORM ); + + size = format_max_bits( info ); + if (size == 8) { + pinfo->datatype = GL_UNSIGNED_BYTE; + } + else if (size == 16) { + pinfo->datatype = GL_UNSIGNED_SHORT; + } + else { + assert( size <= 32 ); + + pinfo->datatype = GL_UNSIGNED_INT; + } + } + + /* Component bits */ + pinfo->red_bits = format_bits( info, PIPE_FORMAT_COMP_R ); + pinfo->green_bits = format_bits( info, PIPE_FORMAT_COMP_G ); + pinfo->blue_bits = format_bits( info, PIPE_FORMAT_COMP_B ); + pinfo->alpha_bits = format_bits( info, PIPE_FORMAT_COMP_A ); + pinfo->depth_bits = format_bits( info, PIPE_FORMAT_COMP_Z ); + pinfo->stencil_bits = format_bits( info, PIPE_FORMAT_COMP_S ); + + /* Format size */ + pinfo->size = format_size( info ) / 8; + + /* Luminance & Intensity bits */ + if( info.swizzleX == PIPE_FORMAT_COMP_R && info.swizzleY == PIPE_FORMAT_COMP_R && info.swizzleZ == PIPE_FORMAT_COMP_R ) { + if( info.swizzleW == PIPE_FORMAT_COMP_R ) { + pinfo->luminance_bits = 0; + pinfo->intensity_bits = pinfo->red_bits; + } + else { + pinfo->luminance_bits = pinfo->red_bits; + pinfo->intensity_bits = 0; + } + pinfo->red_bits = 0; + } + + /* Base format */ + if (pinfo->depth_bits) { + if (pinfo->stencil_bits) { + pinfo->base_format = GL_DEPTH_STENCIL_EXT; + } + else { + pinfo->base_format = GL_DEPTH_COMPONENT; + } + } + else if (pinfo->stencil_bits) { + pinfo->base_format = GL_STENCIL_INDEX; + } + else { + pinfo->base_format = GL_RGBA; + } + } + else { + struct pipe_format_ycbcr info; + + assert( fmt.header.layout == PIPE_FORMAT_LAYOUT_YCBCR ); + + info = fmt.ycbcr; + + /* TODO */ + assert( 0 ); + } + +#if 0 + printf( + "ST_FORMAT: R(%u), G(%u), B(%u), A(%u), Z(%u), S(%u)\n", + pinfo->red_bits, + pinfo->green_bits, + pinfo->blue_bits, + pinfo->alpha_bits, + pinfo->depth_bits, + pinfo->stencil_bits ); +#endif + + pinfo->format = format; + + return GL_TRUE; } @@ -133,9 +294,12 @@ st_get_format_info(GLuint format) GLuint st_sizeof_format(GLuint pipeFormat) { - const struct pipe_format_info *info = st_get_format_info(pipeFormat); - assert(info); - return info->size; + struct pipe_format_info info; + if (!st_get_format_info( pipeFormat, &info )) { + assert( 0 ); + return 0; + } + return info.size; } @@ -145,9 +309,12 @@ st_sizeof_format(GLuint pipeFormat) GLenum st_format_datatype(GLuint pipeFormat) { - const struct pipe_format_info *info = st_get_format_info(pipeFormat); - assert(info); - return info->datatype; + struct pipe_format_info info; + if (!st_get_format_info( pipeFormat, &info )) { + assert( 0 ); + return 0; + } + return info.datatype; } diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index 17d0985411..23abc36edf 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -47,8 +47,10 @@ struct pipe_format_info }; -extern const struct pipe_format_info * -st_get_format_info(GLuint format); +extern GLboolean +st_get_format_info( + GLuint format, + struct pipe_format_info *pinfo ); extern GLuint -- cgit v1.2.3 From a846d7d0273e879371b69f1bbcb8a6698a24a903 Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 18:43:49 +0100 Subject: Enable SSE path. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 701a68ec6a..43c4d58ac6 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -103,8 +103,7 @@ shade_quad( /* run shader */ #if defined(__i386__) || defined(__386__) - /* XXX: Generated code effectively unusable until it handles quad->mask */ - if( !quad->mask && softpipe->use_sse ) { + if( softpipe->use_sse ) { codegen_function func = (codegen_function) x86_get_func( &softpipe->fs->sse2_program ); func( machine->Inputs, @@ -112,6 +111,7 @@ shade_quad( machine->Consts, machine->Temps, machine->InterpCoefs ); + quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]); } else #endif -- cgit v1.2.3 From f93b9dc09a8f9289d7bd5c0f99c935f28016691e Mon Sep 17 00:00:00 2001 From: michal Date: Sat, 27 Oct 2007 18:53:38 +0100 Subject: Use FREE instead of free. Fix newlines. --- src/mesa/pipe/p_util.h | 2 ++ src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_blend.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_bufloop.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_colormask.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_coverage.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_fs.c | 6 +++++- src/mesa/pipe/softpipe/sp_quad_occlusion.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_output.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_stencil.c | 6 +++--- src/mesa/pipe/softpipe/sp_quad_stipple.c | 6 +++--- 12 files changed, 37 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 436bda2139..319bb790ae 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -32,6 +32,8 @@ #define CALLOC_STRUCT(T) (struct T *) calloc(1, sizeof(struct T)) +#define FREE( PTR ) free( PTR ) + #define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index d9b914d896..6a9cf29e47 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -88,9 +88,9 @@ static void alpha_test_begin(struct quad_stage *qs) } -static void alpha_test_destroy(struct quad_stage *qs) -{ - free( qs ); +static void alpha_test_destroy(struct quad_stage *qs) +{ + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index cecf8af29b..696e252af1 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -727,9 +727,9 @@ static void blend_begin(struct quad_stage *qs) } -static void blend_destroy(struct quad_stage *qs) -{ - free( qs ); +static void blend_destroy(struct quad_stage *qs) +{ + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c index bdf491d4b1..aac70e2b04 100644 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -50,9 +50,9 @@ static void cbuf_loop_begin(struct quad_stage *qs) } -static void cbuf_loop_destroy(struct quad_stage *qs) -{ - free( qs ); +static void cbuf_loop_destroy(struct quad_stage *qs) +{ + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index f0aa657e95..c585aa3edd 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -90,9 +90,9 @@ static void colormask_begin(struct quad_stage *qs) } -static void colormask_destroy(struct quad_stage *qs) -{ - free( qs ); +static void colormask_destroy(struct quad_stage *qs) +{ + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c index 25e7b033b9..9dfad7c580 100644 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -69,9 +69,9 @@ static void coverage_begin(struct quad_stage *qs) } -static void coverage_destroy(struct quad_stage *qs) -{ - free( qs ); +static void coverage_destroy(struct quad_stage *qs) +{ + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index c24232bf1e..9cb01d7b2f 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -231,9 +231,9 @@ static void depth_test_begin(struct quad_stage *qs) } -static void depth_test_destroy(struct quad_stage *qs) -{ - free( qs ); +static void depth_test_destroy(struct quad_stage *qs) +{ + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 43c4d58ac6..81e405d3db 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -198,7 +198,11 @@ static void shade_begin(struct quad_stage *qs) static void shade_destroy(struct quad_stage *qs) { - free( qs ); + struct quad_shade_stage *qss = (struct quad_shade_stage *) qs; + + FREE( qss->inputs ); + FREE( qss->outputs ); + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c index 18a36e8286..028d30c92f 100644 --- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c +++ b/src/mesa/pipe/softpipe/sp_quad_occlusion.c @@ -64,9 +64,9 @@ static void occlusion_begin(struct quad_stage *qs) } -static void occlusion_destroy(struct quad_stage *qs) -{ - free( qs ); +static void occlusion_destroy(struct quad_stage *qs) +{ + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index 09194c14a6..cebfec18f7 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -70,9 +70,9 @@ static void output_begin(struct quad_stage *qs) } -static void output_destroy(struct quad_stage *qs) -{ - free( qs ); +static void output_destroy(struct quad_stage *qs) +{ + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 14aa7df279..831ad8bad0 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -323,9 +323,9 @@ static void stencil_begin(struct quad_stage *qs) } -static void stencil_destroy(struct quad_stage *qs) -{ - free( qs ); +static void stencil_destroy(struct quad_stage *qs) +{ + FREE( qs ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index 7916dcd0ca..fcbbf00c7d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -65,9 +65,9 @@ static void stipple_begin(struct quad_stage *qs) } -static void stipple_destroy(struct quad_stage *qs) -{ - free( qs ); +static void stipple_destroy(struct quad_stage *qs) +{ + FREE( qs ); } -- cgit v1.2.3 From 90e47c8c7bd72c751b9d478fa583d758aa26b06d Mon Sep 17 00:00:00 2001 From: michal Date: Sun, 28 Oct 2007 14:26:04 +0000 Subject: Fix newlines. --- src/mesa/pipe/softpipe/sp_context.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 999047889a..53c4b03756 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -215,16 +215,16 @@ static void softpipe_destroy( struct pipe_context *pipe ) draw_destroy( softpipe->draw ); - softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); - softpipe->quad.shade->destroy( softpipe->quad.shade ); - softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); - softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); - softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); - softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); - softpipe->quad.coverage->destroy( softpipe->quad.coverage ); - softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); - softpipe->quad.blend->destroy( softpipe->quad.blend ); - softpipe->quad.colormask->destroy( softpipe->quad.colormask ); + softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); + softpipe->quad.shade->destroy( softpipe->quad.shade ); + softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); + softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); + softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); + softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); + softpipe->quad.coverage->destroy( softpipe->quad.coverage ); + softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); + softpipe->quad.blend->destroy( softpipe->quad.blend ); + softpipe->quad.colormask->destroy( softpipe->quad.colormask ); softpipe->quad.output->destroy( softpipe->quad.output ); free( softpipe ); -- cgit v1.2.3 From 80ab2ab335717db4f3b9593ff1e111cc2191d83f Mon Sep 17 00:00:00 2001 From: michal Date: Sun, 28 Oct 2007 14:27:02 +0000 Subject: Control FS TGSI dumping with GALLIUM_DUMP_FS env variable. --- src/mesa/pipe/softpipe/sp_context.c | 2 ++ src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_state_fs.c | 6 ++++++ 3 files changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 53c4b03756..22493f4703 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -292,6 +292,8 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->use_sse = FALSE; #endif + softpipe->dump_fs = getenv( "GALLIUM_DUMP_FS" ) != NULL; + softpipe->pipe.winsys = pipe_winsys; softpipe->pipe.destroy = softpipe_destroy; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 548151b378..720297653f 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -166,6 +166,7 @@ struct softpipe_context { struct softpipe_tile_cache *tex_cache[PIPE_MAX_SAMPLERS]; int use_sse : 1; + int dump_fs : 1; }; diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index a94ec1e92c..ad8ab561ef 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -46,6 +46,12 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, struct sp_fragment_shader_state *state = malloc(sizeof(struct sp_fragment_shader_state)); state->shader = *templ; + if( softpipe->dump_fs ) { + tgsi_dump( + state->shader.tokens, + 0 ); + } + #if defined(__i386__) || defined(__386__) if (softpipe->use_sse) { x86_init_func( &state->sse2_program ); -- cgit v1.2.3 From e420e9d48577de57f912ab39d59c2d1d4d14b8f6 Mon Sep 17 00:00:00 2001 From: michal Date: Sun, 28 Oct 2007 14:42:26 +0000 Subject: Declare temporaries in a more compact fashion. The following declarations: DCL TEMP[0] DCL TEMP[1] DCL TEMP[2] DCL TEMP[4] become: DCL TEMP[0..2] DCL TEMP[4] --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 116 ++++++++++++++++--------------- 1 file changed, 61 insertions(+), 55 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 621159eb85..cb0c96169e 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -572,34 +572,36 @@ make_output_decl( return decl; } - - -static struct tgsi_full_declaration -make_temp_decl(GLuint index) -{ - struct tgsi_full_declaration decl; - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.u.DeclarationRange.First = index; - decl.u.DeclarationRange.Last = index; - return decl; -} - + + +static struct tgsi_full_declaration +make_temp_decl( + GLuint start_index, + GLuint end_index ) +{ + struct tgsi_full_declaration decl; + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.u.DeclarationRange.First = start_index; + decl.u.DeclarationRange.Last = end_index; + return decl; +} + /** * Find the temporaries which are used in the given program. - * Put the indices of the temporaries in 'tempsUsed'. - * \return number of temporaries used - */ -static GLuint -find_temporaries(const struct gl_program *program, - GLuint tempsUsed[MAX_PROGRAM_TEMPS]) -{ - GLuint i, j, count; - - for (i = 0; i < MAX_PROGRAM_TEMPS; i++) - tempsUsed[i] = GL_FALSE; + * Put the indices of the temporaries in 'tempsUsed'. + * \return number of temporaries used + */ +static void +find_temporaries(const struct gl_program *program, + GLboolean tempsUsed[MAX_PROGRAM_TEMPS]) +{ + GLuint i, j; + + for (i = 0; i < MAX_PROGRAM_TEMPS; i++) + tempsUsed[i] = GL_FALSE; for (i = 0; i < program->NumInstructions; i++) { const struct prog_instruction *inst = program->Instructions + i; @@ -608,20 +610,12 @@ find_temporaries(const struct gl_program *program, if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) tempsUsed[inst->SrcReg[j].Index] = GL_TRUE; if (inst->DstReg.File == PROGRAM_TEMPORARY) - tempsUsed[inst->DstReg.Index] = GL_TRUE; - } - } - - /* convert flags to list of indices */ - count = 0; - for (i = 0; i < MAX_PROGRAM_TEMPS; i++) { - if (tempsUsed[i]) - tempsUsed[count++] = i; - } - return count; -} - - + tempsUsed[inst->DstReg.Index] = GL_TRUE; + } + } +} + + /** @@ -781,22 +775,34 @@ tgsi_translate_mesa_program( maxTokens - ti ); } } - - /* temporary decls */ - { - GLuint tempsUsed[MAX_PROGRAM_TEMPS]; - uint numTemps = find_temporaries(program, tempsUsed); - for (i = 0; i < numTemps; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_temp_decl(tempsUsed[i]); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - + + /* temporary decls */ + { + GLboolean tempsUsed[MAX_PROGRAM_TEMPS + 1]; + GLboolean inside_range = GL_FALSE; + GLuint start_range; + + find_temporaries(program, tempsUsed); + tempsUsed[MAX_PROGRAM_TEMPS] = GL_FALSE; + for (i = 0; i < MAX_PROGRAM_TEMPS + 1; i++) { + if (tempsUsed[i] && !inside_range) { + inside_range = GL_TRUE; + start_range = i; + } + else if (!tempsUsed[i] && inside_range) { + struct tgsi_full_declaration fulldecl; + + inside_range = GL_FALSE; + fulldecl = make_temp_decl( start_range, i - 1 ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + } + /* immediates/literals */ #if EMIT_IMMEDIATES for (i = 0; i < program->Parameters->NumParameters; i++) { -- cgit v1.2.3 From c5ad88e9f1d6a915d9464df0b8fa4de65a3513d2 Mon Sep 17 00:00:00 2001 From: michal Date: Sun, 28 Oct 2007 14:42:26 +0000 Subject: Declare temporaries in a more compact fashion. The following declarations: DCL TEMP[0] DCL TEMP[1] DCL TEMP[2] DCL TEMP[4] become: DCL TEMP[0..2] DCL TEMP[4] --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 122 +++++++++++++++---------------- 1 file changed, 61 insertions(+), 61 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index cb0c96169e..0e3533af4d 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -572,36 +572,36 @@ make_output_decl( return decl; } - - -static struct tgsi_full_declaration -make_temp_decl( - GLuint start_index, - GLuint end_index ) -{ - struct tgsi_full_declaration decl; - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.u.DeclarationRange.First = start_index; - decl.u.DeclarationRange.Last = end_index; - return decl; -} - + + +static struct tgsi_full_declaration +make_temp_decl( + GLuint start_index, + GLuint end_index ) +{ + struct tgsi_full_declaration decl; + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.u.DeclarationRange.First = start_index; + decl.u.DeclarationRange.Last = end_index; + return decl; +} + /** * Find the temporaries which are used in the given program. - * Put the indices of the temporaries in 'tempsUsed'. - * \return number of temporaries used - */ -static void -find_temporaries(const struct gl_program *program, - GLboolean tempsUsed[MAX_PROGRAM_TEMPS]) -{ - GLuint i, j; - - for (i = 0; i < MAX_PROGRAM_TEMPS; i++) - tempsUsed[i] = GL_FALSE; + * Put the indices of the temporaries in 'tempsUsed'. + * \return number of temporaries used + */ +static void +find_temporaries(const struct gl_program *program, + GLboolean tempsUsed[MAX_PROGRAM_TEMPS]) +{ + GLuint i, j; + + for (i = 0; i < MAX_PROGRAM_TEMPS; i++) + tempsUsed[i] = GL_FALSE; for (i = 0; i < program->NumInstructions; i++) { const struct prog_instruction *inst = program->Instructions + i; @@ -610,12 +610,12 @@ find_temporaries(const struct gl_program *program, if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) tempsUsed[inst->SrcReg[j].Index] = GL_TRUE; if (inst->DstReg.File == PROGRAM_TEMPORARY) - tempsUsed[inst->DstReg.Index] = GL_TRUE; - } - } -} - - + tempsUsed[inst->DstReg.Index] = GL_TRUE; + } + } +} + + /** @@ -775,34 +775,34 @@ tgsi_translate_mesa_program( maxTokens - ti ); } } - - /* temporary decls */ - { - GLboolean tempsUsed[MAX_PROGRAM_TEMPS + 1]; - GLboolean inside_range = GL_FALSE; - GLuint start_range; - - find_temporaries(program, tempsUsed); - tempsUsed[MAX_PROGRAM_TEMPS] = GL_FALSE; - for (i = 0; i < MAX_PROGRAM_TEMPS + 1; i++) { - if (tempsUsed[i] && !inside_range) { - inside_range = GL_TRUE; - start_range = i; - } - else if (!tempsUsed[i] && inside_range) { - struct tgsi_full_declaration fulldecl; - - inside_range = GL_FALSE; - fulldecl = make_temp_decl( start_range, i - 1 ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - } - + + /* temporary decls */ + { + GLboolean tempsUsed[MAX_PROGRAM_TEMPS + 1]; + GLboolean inside_range = GL_FALSE; + GLuint start_range; + + find_temporaries(program, tempsUsed); + tempsUsed[MAX_PROGRAM_TEMPS] = GL_FALSE; + for (i = 0; i < MAX_PROGRAM_TEMPS + 1; i++) { + if (tempsUsed[i] && !inside_range) { + inside_range = GL_TRUE; + start_range = i; + } + else if (!tempsUsed[i] && inside_range) { + struct tgsi_full_declaration fulldecl; + + inside_range = GL_FALSE; + fulldecl = make_temp_decl( start_range, i - 1 ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + } + /* immediates/literals */ #if EMIT_IMMEDIATES for (i = 0; i < program->Parameters->NumParameters; i++) { -- cgit v1.2.3 From 2953415223fc61e92305453d60714c7e919b604d Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 28 Oct 2007 16:16:28 +0000 Subject: Rework sse-utility function calls. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 265 +++++++++++++++--------------------- 1 file changed, 111 insertions(+), 154 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 1e1002f54e..b9a87fc6b8 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -20,17 +20,7 @@ #define CHAN_Z 2 #define CHAN_W 3 -static unsigned -p2u( void *p ) -{ - union { - void *p; - unsigned u; - } pu; - - pu.p = p; - return pu.u; -} +#define TEMP_R0 TGSI_EXEC_TEMP_R0 static struct x86_reg get_argument( @@ -376,132 +366,112 @@ emit_add( } static void -emit_push_abcd( +emit_push_gp( struct x86_function *func ) { x86_push( func, - x86_make_reg( file_REG32, reg_AX ) ); + get_const_base() ); x86_push( func, - x86_make_reg( file_REG32, reg_BX ) ); + get_input_base() ); x86_push( func, - x86_make_reg( file_REG32, reg_CX ) ); + get_output_base() ); + + /* It is important on non-win32 platforms that temp base is pushed last. + */ x86_push( func, - x86_make_reg( file_REG32, reg_DX ) ); + get_temp_base() ); } static void -emit_pop_dcba( +emit_pop_gp( struct x86_function *func ) { + /* Restore GP registers in a reverse order. + */ x86_pop( func, - x86_make_reg( file_REG32, reg_DX ) ); + get_temp_base() ); x86_pop( func, - x86_make_reg( file_REG32, reg_CX ) ); + get_output_base() ); x86_pop( func, - x86_make_reg( file_REG32, reg_BX ) ); + get_input_base() ); x86_pop( func, - x86_make_reg( file_REG32, reg_AX ) ); + get_const_base() ); } static void -emit_func_call1( +emit_func_call_dst( struct x86_function *func, unsigned xmm_dst, - unsigned storage, void (*code)() ) { - x86_push( - func, - x86_make_reg( file_REG32, reg_AX ) ); - x86_mov_reg_imm( - func, - x86_make_reg( file_REG32, reg_AX ), - storage ); sse_movaps( func, - x86_deref( x86_make_reg( file_REG32, reg_AX ) ), + get_temp( TEMP_R0, 0 ), make_xmm( xmm_dst ) ); - emit_push_abcd( + + emit_push_gp( func ); + +#ifdef WIN32 + x86_push( + func, + get_temp( TEMP_R0, 0 ) ); +#endif + x86_call( func, code ); - emit_pop_dcba( + + emit_pop_gp( func ); + sse_movaps( func, make_xmm( xmm_dst ), - x86_deref( x86_make_reg( file_REG32, reg_AX ) ) ); - x86_pop( - func, - x86_make_reg( file_REG32, reg_AX ) ); + get_temp( TEMP_R0, 0 ) ); } static void -emit_func_call2( +emit_func_call_dst_src( struct x86_function *func, unsigned xmm_dst, unsigned xmm_src, - unsigned storage, void (*code)() ) { - x86_push( - func, - x86_make_reg( file_REG32, reg_AX ) ); - x86_mov_reg_imm( - func, - x86_make_reg( file_REG32, reg_AX ), - storage ); sse_movaps( func, - x86_deref( x86_make_reg( file_REG32, reg_AX ) ), - make_xmm( xmm_dst ) ); - sse_movaps( - func, - x86_make_disp( x86_make_reg( file_REG32, reg_AX ), 16 ), + get_temp( TEMP_R0, 1 ), make_xmm( xmm_src ) ); - emit_push_abcd( - func ); - x86_call( + + emit_func_call_dst( func, + xmm_dst, code ); - emit_pop_dcba( - func ); - sse_movaps( - func, - make_xmm( xmm_dst ), - x86_deref( x86_make_reg( file_REG32, reg_AX ) ) ); - x86_pop( - func, - x86_make_reg( file_REG32, reg_AX ) ); } -/* XXX: move into machine context */ -static float g_cos_storage[4 + 3]; - -static void -cos4f( void ) +static void XSTDCALL +cos4f( + float *store ) { - float *store = (float *) tgsi_align_128bit( g_cos_storage ); - #ifdef WIN32 store[0] = (float) cos( (double) store[0] ); store[1] = (float) cos( (double) store[1] ); store[2] = (float) cos( (double) store[2] ); store[3] = (float) cos( (double) store[3] ); #else - store[0] = cosf( store[0] ); - store[1] = cosf( store[1] ); - store[2] = cosf( store[2] ); - store[3] = cosf( store[3] ); + const unsigned X = TEMP_R0 * 16; + store[X + 0] = cosf( store[X + 0] ); + store[X + 1] = cosf( store[X + 1] ); + store[X + 2] = cosf( store[X + 2] ); + store[X + 3] = cosf( store[X + 3] ); #endif } @@ -510,31 +480,27 @@ emit_cos( struct x86_function *func, unsigned xmm_dst ) { - emit_func_call1( + emit_func_call_dst( func, xmm_dst, - p2u( tgsi_align_128bit( g_cos_storage ) ), cos4f ); } -/* XXX: move into machine context */ -static float g_sin_storage[4 + 3]; - -static void -sin4f( void ) +static void XSTDCALL +sin4f( + float *store ) { - float *store = (float *) tgsi_align_128bit( g_sin_storage ); - #ifdef WIN32 store[0] = (float) sin( (double) store[0] ); store[1] = (float) sin( (double) store[1] ); store[2] = (float) sin( (double) store[2] ); store[3] = (float) sin( (double) store[3] ); #else - store[0] = sinf( store[0] ); - store[1] = sinf( store[1] ); - store[2] = sinf( store[2] ); - store[3] = sinf( store[3] ); + const unsigned X = TEMP_R0 * 16; + store[X + 0] = sinf( store[X + 0] ); + store[X + 1] = sinf( store[X + 1] ); + store[X + 2] = sinf( store[X + 2] ); + store[X + 3] = sinf( store[X + 3] ); #endif } @@ -542,10 +508,9 @@ static void emit_sin (struct x86_function *func, unsigned xmm_dst) { - emit_func_call1( + emit_func_call_dst( func, xmm_dst, - p2u( tgsi_align_128bit( g_sin_storage ) ), sin4f ); } @@ -572,24 +537,21 @@ emit_mul (struct x86_function *func, make_xmm( xmm_src ) ); } -/* XXX: move into machine context */ -static float g_pow_storage[4 + 4 + 3]; - -static void -pow4f( void ) +static void XSTDCALL +pow4f( + float *store ) { - float *store = (float *) tgsi_align_128bit( g_pow_storage ); - #ifdef WIN32 store[0] = (float) pow( (double) store[0], (double) store[4] ); store[1] = (float) pow( (double) store[1], (double) store[5] ); store[2] = (float) pow( (double) store[2], (double) store[6] ); store[3] = (float) pow( (double) store[3], (double) store[7] ); #else - store[0] = powf( store[0], store[4] ); - store[1] = powf( store[1], store[5] ); - store[2] = powf( store[2], store[6] ); - store[3] = powf( store[3], store[7] ); + const unsigned X = TEMP_R0 * 16; + store[X + 0] = powf( store[X + 0], store[X + 4] ); + store[X + 1] = powf( store[X + 1], store[X + 5] ); + store[X + 2] = powf( store[X + 2], store[X + 6] ); + store[X + 3] = powf( store[X + 3], store[X + 7] ); #endif } @@ -599,32 +561,28 @@ emit_pow( unsigned xmm_dst, unsigned xmm_src ) { - emit_func_call2( + emit_func_call_dst_src( func, xmm_dst, xmm_src, - p2u( tgsi_align_128bit( g_pow_storage ) ), pow4f ); } -/* XXX: move into machine context */ -static float g_ex2_storage[4 + 3]; - -static void -ex24f( void ) +static void XSTDCALL +ex24f( + float *store ) { - float *store = (float *) tgsi_align_128bit( g_ex2_storage ); - #ifdef WIN32 store[0] = (float) pow( 2.0, (double) store[0] ); store[1] = (float) pow( 2.0, (double) store[1] ); store[2] = (float) pow( 2.0, (double) store[2] ); store[3] = (float) pow( 2.0, (double) store[3] ); #else - store[0] = powf( 2.0f, store[0] ); - store[1] = powf( 2.0f, store[1] ); - store[2] = powf( 2.0f, store[2] ); - store[3] = powf( 2.0f, store[3] ); + const unsigned X = TEMP_R0 * 16; + store[X + 0] = powf( 2.0f, store[X + 0] ); + store[X + 1] = powf( 2.0f, store[X + 1] ); + store[X + 2] = powf( 2.0f, store[X + 2] ); + store[X + 3] = powf( 2.0f, store[X + 3] ); #endif } @@ -633,25 +591,25 @@ emit_ex2( struct x86_function *func, unsigned xmm_dst ) { - emit_func_call1( + emit_func_call_dst( func, xmm_dst, - p2u( tgsi_align_128bit( g_ex2_storage ) ), ex24f ); } -/* XXX: move into machine context */ -static float g_lg2_storage[4 + 3]; - -static void -lg24f( void ) +static void XSTDCALL +lg24f( + float *store ) { - float *store = (float *) tgsi_align_128bit( g_lg2_storage ); - - store[0] = LOG2( store[0] ); - store[1] = LOG2( store[1] ); - store[2] = LOG2( store[2] ); - store[3] = LOG2( store[3] ); +#ifdef WIN32 + const unsigned X = 0; +#else + const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] = LOG2( store[X + 0] ); + store[X + 1] = LOG2( store[X + 1] ); + store[X + 2] = LOG2( store[X + 2] ); + store[X + 3] = LOG2( store[X + 3] ); } static void @@ -659,25 +617,25 @@ emit_lg2( struct x86_function *func, unsigned xmm_dst ) { - emit_func_call1( + emit_func_call_dst( func, xmm_dst, - p2u( tgsi_align_128bit( g_lg2_storage ) ), lg24f ); } -/* XXX: move into machine context */ -static float g_flr_storage[4 + 3]; - -static void -flr4f( void ) +static void XSTDCALL +flr4f( + float *store ) { - float *store = (float *) tgsi_align_128bit( g_flr_storage ); - - store[0] = (float) floor( (double) store[0] ); - store[1] = (float) floor( (double) store[1] ); - store[2] = (float) floor( (double) store[2] ); - store[3] = (float) floor( (double) store[3] ); +#ifdef WIN32 + const unsigned X = 0; +#else + const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] = (float) floor( (double) store[X + 0] ); + store[X + 1] = (float) floor( (double) store[X + 1] ); + store[X + 2] = (float) floor( (double) store[X + 2] ); + store[X + 3] = (float) floor( (double) store[X + 3] ); } static void @@ -685,25 +643,25 @@ emit_flr( struct x86_function *func, unsigned xmm_dst ) { - emit_func_call1( + emit_func_call_dst( func, xmm_dst, - p2u( tgsi_align_128bit( g_flr_storage ) ), flr4f ); } -/* XXX: move into machine context */ -static float g_frc_storage[4 + 3]; - -static void -frc4f( void ) +static void XSTDCALL +frc4f( + float *store ) { - float *store = (float *) tgsi_align_128bit( g_frc_storage ); - - store[0] -= (float) floor( (double) store[0] ); - store[1] -= (float) floor( (double) store[1] ); - store[2] -= (float) floor( (double) store[2] ); - store[3] -= (float) floor( (double) store[3] ); +#ifdef WIN32 + const unsigned X = 0; +#else + const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] -= (float) floor( (double) store[X + 0] ); + store[X + 1] -= (float) floor( (double) store[X + 1] ); + store[X + 2] -= (float) floor( (double) store[X + 2] ); + store[X + 3] -= (float) floor( (double) store[X + 3] ); } static void @@ -711,10 +669,9 @@ emit_frc( struct x86_function *func, unsigned xmm_dst ) { - emit_func_call1( + emit_func_call_dst( func, xmm_dst, - p2u( tgsi_align_128bit( g_frc_storage ) ), frc4f ); } -- cgit v1.2.3 From b85cd7b70096cf7c922aed56ae8255fb4b8f0709 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 28 Oct 2007 16:36:07 +0000 Subject: Use FREE and MALLOC instead of free and malloc. --- src/mesa/pipe/p_util.h | 2 ++ src/mesa/pipe/softpipe/sp_context.c | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 4 ++-- 3 files changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 319bb790ae..4f68f5e1bd 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -32,6 +32,8 @@ #define CALLOC_STRUCT(T) (struct T *) calloc(1, sizeof(struct T)) +#define MALLOC( SIZE ) malloc( SIZE ) + #define FREE( PTR ) free( PTR ) #define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 22493f4703..3a1861cb62 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -227,7 +227,7 @@ static void softpipe_destroy( struct pipe_context *pipe ) softpipe->quad.colormask->destroy( softpipe->quad.colormask ); softpipe->quad.output->destroy( softpipe->quad.output ); - free( softpipe ); + FREE( softpipe ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 81e405d3db..c4f90e5064 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -212,8 +212,8 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) uint i; /* allocate storage for program inputs/outputs, aligned to 16 bytes */ - qss->inputs = malloc(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); - qss->outputs = malloc(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); + qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); + qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); qss->machine.Inputs = align16(qss->inputs); qss->machine.Outputs = align16(qss->outputs); -- cgit v1.2.3 From 3c8121967224f91bfcd5431b4069d66eecbc5952 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 28 Oct 2007 17:19:39 +0000 Subject: Replace supported_formats with is_format_supported interface. The old supported_formats interface returned a list of formats supported by a pipe/winsys implementation. This was reasonable when gallium had a fixed list of predefined format. Now things has changed and the definition of PIPE_FORMAT is more flexible. The new shiny is_format_supported interface gets PIPE_FORMAT as an argument and returns a boolean whether this particular format is supported. --- src/mesa/drivers/x11/xm_api.c | 2 +- src/mesa/drivers/x11/xm_surface.c | 18 ++++---- src/mesa/drivers/x11/xmesaP.h | 4 +- src/mesa/pipe/failover/fo_context.c | 2 +- src/mesa/pipe/i915simple/i915_context.c | 24 +++++------ src/mesa/pipe/p_context.h | 4 +- src/mesa/pipe/softpipe/sp_context.c | 12 +++--- src/mesa/pipe/softpipe/sp_winsys.h | 4 +- src/mesa/state_tracker/st_cb_drawpixels.c | 33 +++++++-------- src/mesa/state_tracker/st_format.c | 70 ++++++++++++------------------- 10 files changed, 75 insertions(+), 98 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 1d2b93d100..4b31447bbd 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -1616,7 +1616,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) #if 0 mesaCtx->st->pipe->surface_alloc = xmesa_surface_alloc; #endif - mesaCtx->st->pipe->supported_formats = xmesa_supported_formats; + mesaCtx->st->pipe->is_format_supported = xmesa_is_format_supported; mesaCtx->st->pipe->get_tile_rgba = xmesa_get_tile_rgba; mesaCtx->st->pipe->put_tile_rgba = xmesa_put_tile_rgba; diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index dbbf26e33f..5533158ece 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -195,18 +195,16 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) } -const GLuint * -xmesa_supported_formats(struct pipe_context *pipe, GLuint *numFormats) +boolean +xmesa_is_format_supported(struct pipe_context *pipe, uint format) { - static const GLuint formats[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_S_R16_G16_B16_A16, - PIPE_FORMAT_S8_Z24 + switch( format ) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_S_R16_G16_B16_A16: + case PIPE_FORMAT_S8_Z24: + return TRUE; }; - - *numFormats = sizeof(formats) / sizeof(formats[0]); - - return formats; + return FALSE; } diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 8af95504fb..4709d63394 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -618,8 +618,8 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); extern struct pipe_surface * xmesa_new_color_surface(struct pipe_context *pipe, GLuint format); -extern const uint * -xmesa_supported_formats(struct pipe_context *pipe, uint *numFormats); +extern boolean +xmesa_is_format_supported(struct pipe_context *pipe, uint format); extern void xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index 7e02b751bb..aa5d0885e6 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -116,7 +116,7 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.winsys = hw->winsys; failover->pipe.destroy = failover_destroy; - failover->pipe.supported_formats = hw->supported_formats; + failover->pipe.is_format_supported = hw->is_format_supported; failover->pipe.max_texture_size = hw->max_texture_size; failover->pipe.get_name = hw->get_name; failover->pipe.get_vendor = hw->get_vendor; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 161f8ce697..6541f0e848 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -40,16 +40,16 @@ /** - * Return list of supported surface/texture formats. + * Query format support. * If we find texture and drawable support differs, add a selector * parameter or another function. */ -static const unsigned * -i915_supported_formats(struct pipe_context *pipe, -// unsigned type, - unsigned *numFormats) +static boolean +i915_is_format_supported( struct pipe_context *pipe, + uint format ) { #if 0 + /* XXX: This is broken -- rewrite if still needed. */ static const unsigned tex_supported[] = { PIPE_FORMAT_U_R8_G8_B8_A8, PIPE_FORMAT_U_A8_R8_G8_B8, @@ -97,13 +97,13 @@ i915_supported_formats(struct pipe_context *pipe, return NULL; } #else - static const unsigned render_supported[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_S8_Z24, + switch( format ) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_U_R5_G6_B5: + case PIPE_FORMAT_S8_Z24: + return TRUE; }; - *numFormats = 3; - return render_supported; + return FALSE; #endif } @@ -303,7 +303,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.winsys = pipe_winsys; i915->pipe.destroy = i915_destroy; - i915->pipe.supported_formats = i915_supported_formats; + i915->pipe.is_format_supported = i915_is_format_supported; i915->pipe.max_texture_size = i915_max_texture_size; i915->pipe.get_param = i915_get_param; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index dbbf48c576..a1441a7e43 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -44,8 +44,8 @@ struct pipe_context { /* * Queries */ - const unsigned *(*supported_formats)(struct pipe_context *pipe, - unsigned *numFormats); + boolean (*is_format_supported)( struct pipe_context *pipe, + uint format ); void (*max_texture_size)(struct pipe_context *pipe, unsigned textureType, /* PIPE_TEXTURE_x */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 3a1861cb62..f67588783b 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -46,14 +46,16 @@ /** - * Return list of supported surface/texture formats. + * Query format support. * If we find texture and drawable support differs, add a selector * parameter or another function. */ -static const unsigned * -softpipe_supported_formats(struct pipe_context *pipe, unsigned *numFormats) +static boolean +softpipe_is_format_supported( struct pipe_context *pipe, + uint format ) { #if 0 + /* XXX: This is broken -- rewrite if still needed. */ static const unsigned supported[] = { PIPE_FORMAT_U_R8_G8_B8_A8, PIPE_FORMAT_U_A8_R8_G8_B8, @@ -76,7 +78,7 @@ softpipe_supported_formats(struct pipe_context *pipe, unsigned *numFormats) return supported; #else struct softpipe_context *softpipe = softpipe_context( pipe ); - return softpipe->winsys->supported_formats( softpipe->winsys, numFormats ); + return softpipe->winsys->is_format_supported( softpipe->winsys, format ); #endif } @@ -298,7 +300,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.destroy = softpipe_destroy; /* queries */ - softpipe->pipe.supported_formats = softpipe_supported_formats; + softpipe->pipe.is_format_supported = softpipe_is_format_supported; softpipe->pipe.max_texture_size = softpipe_max_texture_size; softpipe->pipe.get_param = softpipe_get_param; diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h index 726e4c8bb6..d8ae971188 100644 --- a/src/mesa/pipe/softpipe/sp_winsys.h +++ b/src/mesa/pipe/softpipe/sp_winsys.h @@ -35,8 +35,8 @@ */ struct softpipe_winsys { - const unsigned *(*supported_formats)(struct softpipe_winsys *sws, - unsigned *numFormats); + boolean (*is_format_supported)( struct softpipe_winsys *sws, + uint format ); }; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 896ba2b905..99e1e3ed25 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -937,31 +937,26 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, { struct pipe_context *pipe = ctx->st->pipe; const uint flags = PIPE_SURFACE_FLAG_TEXTURE; - uint numFormats, i, format = 0, cpp, comp, pitch; - const uint *formats = ctx->st->pipe->supported_formats(pipe, &numFormats); + uint format = 0, cpp, comp, pitch; ubyte *dest; struct pipe_mipmap_tree *mt; int row, col; /* find a texture format we know */ - for (i = 0; i < numFormats; i++) { - switch (formats[i]) { - case PIPE_FORMAT_U_I8: - format = formats[i]; - cpp = 1; - comp = 0; - break; - case PIPE_FORMAT_U_A8_R8_G8_B8: - format = formats[i]; - cpp = 4; - comp = 3; /* alpha channel */ /*XXX little-endian dependency */ - break; - default: - /* XXX support more formats */ - ; - } + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8 )) { + format = PIPE_FORMAT_U_I8; + cpp = 1; + comp = 0; + } + else if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) { + format = PIPE_FORMAT_U_A8_R8_G8_B8; + cpp = 4; + comp = 3; /* alpha channel */ /*XXX little-endian dependency */ + } + else { + /* XXX support more formats */ + assert( 0 ); } - assert(format); /** * Create a mipmap tree. diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 5a50d2d63a..c0e1a79bad 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -342,24 +342,6 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) } } -/* XXX: This function should be implemented by pipe object. */ -static GLboolean -allow_format( - struct pipe_context *pipe, - unsigned format ) -{ - const GLuint *supported; - GLuint i, n; - - supported = pipe->supported_formats( pipe, &n ); - for (i = 0; i < n; i++) { - if (supported[i] == format) { - return GL_TRUE; - } - } - return GL_FALSE; -} - /** * Search list of formats for first RGBA format. */ @@ -367,13 +349,13 @@ static GLuint default_rgba_format( struct pipe_context *pipe ) { - if (allow_format( pipe, PIPE_FORMAT_U_R8_G8_B8_A8 )) { + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_R8_G8_B8_A8 )) { return PIPE_FORMAT_U_R8_G8_B8_A8; } - if (allow_format( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) { + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) { return PIPE_FORMAT_U_A8_R8_G8_B8; } - if (allow_format( pipe, PIPE_FORMAT_U_R5_G6_B5 )) { + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_R5_G6_B5 )) { return PIPE_FORMAT_U_R5_G6_B5; } return PIPE_FORMAT_NONE; @@ -387,7 +369,7 @@ static GLuint default_deep_rgba_format( struct pipe_context *pipe ) { - if (allow_format( pipe, PIPE_FORMAT_S_R16_G16_B16_A16 )) { + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S_R16_G16_B16_A16 )) { return PIPE_FORMAT_S_R16_G16_B16_A16; } return PIPE_FORMAT_NONE; @@ -401,13 +383,13 @@ static GLuint default_depth_format( struct pipe_context *pipe ) { - if (allow_format( pipe, PIPE_FORMAT_U_Z16 )) { + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_Z16 )) { return PIPE_FORMAT_U_Z16; } - if (allow_format( pipe, PIPE_FORMAT_U_Z32 )) { + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_Z32 )) { return PIPE_FORMAT_U_Z32; } - if (allow_format( pipe, PIPE_FORMAT_S8_Z24 )) { + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) { return PIPE_FORMAT_S8_Z24; } return PIPE_FORMAT_NONE; @@ -437,15 +419,15 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_COMPRESSED_RGBA: if (format == GL_BGRA) { if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - if (allow_format( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) return PIPE_FORMAT_U_A8_R8_G8_B8; } else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - if (allow_format( pipe, PIPE_FORMAT_U_A4_R4_G4_B4 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A4_R4_G4_B4 )) return PIPE_FORMAT_U_A4_R4_G4_B4; } else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - if (allow_format( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) return PIPE_FORMAT_U_A1_R5_G5_B5; } } @@ -455,7 +437,7 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_RGB: case GL_COMPRESSED_RGB: if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - if (allow_format( pipe, PIPE_FORMAT_U_R5_G6_B5 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_R5_G6_B5 )) return PIPE_FORMAT_U_R5_G6_B5; } return default_rgba_format( pipe ); @@ -469,12 +451,12 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_RGBA4: case GL_RGBA2: - if (allow_format( pipe, PIPE_FORMAT_U_A4_R4_G4_B4 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A4_R4_G4_B4 )) return PIPE_FORMAT_U_A4_R4_G4_B4; return default_rgba_format( pipe ); case GL_RGB5_A1: - if (allow_format( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) return PIPE_FORMAT_U_A1_R5_G5_B5; return default_rgba_format( pipe ); @@ -487,7 +469,7 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - if (allow_format( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) return PIPE_FORMAT_U_A1_R5_G5_B5; return default_rgba_format( pipe ); @@ -497,7 +479,7 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - if (allow_format( pipe, PIPE_FORMAT_U_A8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8 )) return PIPE_FORMAT_U_A8; return default_rgba_format( pipe ); @@ -508,7 +490,7 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - if (allow_format( pipe, PIPE_FORMAT_U_A8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8 )) return PIPE_FORMAT_U_A8; return default_rgba_format( pipe ); @@ -521,7 +503,7 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - if (allow_format( pipe, PIPE_FORMAT_U_A8_L8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_L8 )) return PIPE_FORMAT_U_A8_L8; return default_rgba_format( pipe ); @@ -531,17 +513,17 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - if (allow_format( pipe, PIPE_FORMAT_U_I8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8 )) return PIPE_FORMAT_U_I8; return default_rgba_format( pipe ); case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) { - if (allow_format( pipe, PIPE_FORMAT_YCBCR )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_YCBCR )) return PIPE_FORMAT_YCBCR; } else { - if (allow_format( pipe, PIPE_FORMAT_YCBCR_REV )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_YCBCR_REV )) return PIPE_FORMAT_YCBCR_REV; } return PIPE_FORMAT_NONE; @@ -570,15 +552,15 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, #endif case GL_DEPTH_COMPONENT16: - if (allow_format( pipe, PIPE_FORMAT_U_Z16 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_Z16 )) return PIPE_FORMAT_U_Z16; /* fall-through */ case GL_DEPTH_COMPONENT24: - if (allow_format( pipe, PIPE_FORMAT_S8_Z24 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) return PIPE_FORMAT_S8_Z24; /* fall-through */ case GL_DEPTH_COMPONENT32: - if (allow_format( pipe, PIPE_FORMAT_U_Z32 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_Z32 )) return PIPE_FORMAT_U_Z32; /* fall-through */ case GL_DEPTH_COMPONENT: @@ -589,15 +571,15 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_STENCIL_INDEX4_EXT: case GL_STENCIL_INDEX8_EXT: case GL_STENCIL_INDEX16_EXT: - if (allow_format( pipe, PIPE_FORMAT_U_S8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_S8 )) return PIPE_FORMAT_U_S8; - if (allow_format( pipe, PIPE_FORMAT_S8_Z24 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) return PIPE_FORMAT_S8_Z24; return PIPE_FORMAT_NONE; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - if (allow_format( pipe, PIPE_FORMAT_S8_Z24 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) return PIPE_FORMAT_S8_Z24; return PIPE_FORMAT_NONE; -- cgit v1.2.3 From 2bbd714fda7b0fb21184bea8ed0c08f155cba528 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 28 Oct 2007 17:48:46 +0000 Subject: Update comments. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 0e3533af4d..af680b9e99 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -591,8 +591,6 @@ make_temp_decl( /** * Find the temporaries which are used in the given program. - * Put the indices of the temporaries in 'tempsUsed'. - * \return number of temporaries used */ static void find_temporaries(const struct gl_program *program, -- cgit v1.2.3 From c1b9f0eb7a4a0feb549218cdec3b7964d4c91bc9 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 28 Oct 2007 17:53:02 +0000 Subject: Remove unused static functions. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 24 ------------------------ 1 file changed, 24 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index b9a87fc6b8..46905236e2 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -129,16 +129,6 @@ get_coef( ((vec * 3 + member) * 4 + chan) * 4 ); } -static struct x86_reg -get_addr( - unsigned vec, - unsigned chan ) -{ - return get_temp( - vec + TGSI_EXEC_NUM_TEMPS, - chan ); -} - static void emit_const( struct x86_function *func, @@ -286,20 +276,6 @@ emit_coef_dady( 2 ); } -static void -emit_addrf( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_tempf( - func, - xmm, - vec + TGSI_EXEC_NUM_TEMPS, - chan ); -} - static void emit_addrs( struct x86_function *func, -- cgit v1.2.3 From af5061cf5aced1d59d0a051c44ca5f8d633ddb7f Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 29 Oct 2007 11:03:37 +0000 Subject: Fix newlines. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 1680 +++++++++++++++--------------- 1 file changed, 840 insertions(+), 840 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index af680b9e99..37ba6a90b5 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -1,840 +1,840 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* - * \author - * Michal Krol - */ - - -#include "tgsi_platform.h" -#include "pipe/tgsi/exec/tgsi_core.h" -#include "st_mesa_to_tgsi.h" -#include "shader/prog_parameter.h" - -#define TGSI_DEBUG 0 - -#define EMIT_IMMEDIATES 0 - - -/* - * Map mesa register file to TGSI register file. - */ -static GLuint -map_register_file( - enum register_file file ) -{ - switch( file ) { - case PROGRAM_UNDEFINED: - return TGSI_FILE_NULL; - case PROGRAM_TEMPORARY: - return TGSI_FILE_TEMPORARY; - //case PROGRAM_LOCAL_PARAM: - //case PROGRAM_ENV_PARAM: - case PROGRAM_STATE_VAR: - case PROGRAM_NAMED_PARAM: - case PROGRAM_UNIFORM: - return TGSI_FILE_CONSTANT; - case PROGRAM_CONSTANT: -#if EMIT_IMMEDIATES - return TGSI_FILE_IMMEDIATE; -#else - return TGSI_FILE_CONSTANT; -#endif - case PROGRAM_INPUT: - return TGSI_FILE_INPUT; - case PROGRAM_OUTPUT: - return TGSI_FILE_OUTPUT; - case PROGRAM_ADDRESS: - return TGSI_FILE_ADDRESS; - default: - assert( 0 ); - return TGSI_FILE_NULL; - } -} - -/** - * Map mesa register file index to TGSI index. - * Take special care when processing input and output indices. - * \param file one of TGSI_FILE_x - * \param index the mesa register file index - * \param inputMapping maps Mesa input indexes to TGSI input indexes - * \param outputMapping maps Mesa output indexes to TGSI output indexes - */ -static GLuint -map_register_file_index( - GLuint file, - GLuint index, - const GLuint inputMapping[], - const GLuint outputMapping[], - const GLuint immediateMapping[]) -{ - switch( file ) { - case TGSI_FILE_INPUT: - /* inputs are mapped according to the user-defined map */ - return inputMapping[index]; - - case TGSI_FILE_OUTPUT: - return outputMapping[index]; - -#if EMIT_IMMEDIATES - case TGSI_FILE_IMMEDIATE: - return immediateMapping[index]; -#endif - - default: - return index; - } -} - -/* - * Map mesa texture target to TGSI texture target. - */ -static GLuint -map_texture_target( - GLuint textarget ) -{ - switch( textarget ) { - case TEXTURE_1D_INDEX: - return TGSI_TEXTURE_1D; - case TEXTURE_2D_INDEX: - return TGSI_TEXTURE_2D; - case TEXTURE_3D_INDEX: - return TGSI_TEXTURE_3D; - case TEXTURE_CUBE_INDEX: - return TGSI_TEXTURE_CUBE; - case TEXTURE_RECT_INDEX: - return TGSI_TEXTURE_RECT; - default: - assert( 0 ); - } - - return TGSI_TEXTURE_1D; -} - -static GLuint -convert_sat( - GLuint sat ) -{ - switch( sat ) { - case SATURATE_OFF: - return TGSI_SAT_NONE; - case SATURATE_ZERO_ONE: - return TGSI_SAT_ZERO_ONE; - case SATURATE_PLUS_MINUS_ONE: - return TGSI_SAT_MINUS_PLUS_ONE; - default: - assert( 0 ); - return TGSI_SAT_NONE; - } -} - -static GLuint -convert_writemask( - GLuint writemask ) -{ - assert( WRITEMASK_X == TGSI_WRITEMASK_X ); - assert( WRITEMASK_Y == TGSI_WRITEMASK_Y ); - assert( WRITEMASK_Z == TGSI_WRITEMASK_Z ); - assert( WRITEMASK_W == TGSI_WRITEMASK_W ); - assert( (writemask & ~TGSI_WRITEMASK_XYZW) == 0 ); - - return writemask; -} - -#if EMIT_IMMEDIATES -static struct tgsi_full_immediate -make_immediate(const float *value, uint size) -{ - struct tgsi_full_immediate imm; - imm.Immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; - imm.Immediate.Size = 1 + size; /* one for the token itself */ - imm.Immediate.DataType = TGSI_IMM_FLOAT32; - imm.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; - return imm; -} -#endif - -static void -compile_instruction( - const struct prog_instruction *inst, - struct tgsi_full_instruction *fullinst, - const GLuint inputMapping[], - const GLuint outputMapping[], - const GLuint immediateMapping[], - GLuint preamble_size, - GLuint processor ) -{ - GLuint i; - struct tgsi_full_dst_register *fulldst; - struct tgsi_full_src_register *fullsrc; - - *fullinst = tgsi_default_full_instruction(); - - fullinst->Instruction.Saturate = convert_sat( inst->SaturateMode ); - fullinst->Instruction.NumDstRegs = _mesa_num_inst_dst_regs( inst->Opcode ); - fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode ); - - fulldst = &fullinst->FullDstRegisters[0]; - fulldst->DstRegister.File = map_register_file( inst->DstReg.File ); - fulldst->DstRegister.Index = map_register_file_index( - fulldst->DstRegister.File, - inst->DstReg.Index, - inputMapping, - outputMapping, - NULL - ); - fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); - - for( i = 0; i < fullinst->Instruction.NumSrcRegs; i++ ) { - GLuint j; - - fullsrc = &fullinst->FullSrcRegisters[i]; - fullsrc->SrcRegister.File = map_register_file( inst->SrcReg[i].File ); - fullsrc->SrcRegister.Index = map_register_file_index( - fullsrc->SrcRegister.File, - inst->SrcReg[i].Index, - inputMapping, - outputMapping, - immediateMapping); - - for( j = 0; j < 4; j++ ) { - GLuint swz; - - swz = GET_SWZ( inst->SrcReg[i].Swizzle, j ); - if( swz > SWIZZLE_W ) { - tgsi_util_set_src_register_extswizzle( - &fullsrc->SrcRegisterExtSwz, - swz, - j ); - } - else { - tgsi_util_set_src_register_swizzle( - &fullsrc->SrcRegister, - swz, - j ); - } - } - - if( inst->SrcReg[i].NegateBase == NEGATE_XYZW ) { - fullsrc->SrcRegister.Negate = 1; - } - else if( inst->SrcReg[i].NegateBase != NEGATE_NONE ) { - if( inst->SrcReg[i].NegateBase & NEGATE_X ) { - fullsrc->SrcRegisterExtSwz.NegateX = 1; - } - if( inst->SrcReg[i].NegateBase & NEGATE_Y ) { - fullsrc->SrcRegisterExtSwz.NegateY = 1; - } - if( inst->SrcReg[i].NegateBase & NEGATE_Z ) { - fullsrc->SrcRegisterExtSwz.NegateZ = 1; - } - if( inst->SrcReg[i].NegateBase & NEGATE_W ) { - fullsrc->SrcRegisterExtSwz.NegateW = 1; - } - } - - if( inst->SrcReg[i].Abs ) { - fullsrc->SrcRegisterExtMod.Absolute = 1; - } - - if( inst->SrcReg[i].NegateAbs ) { - fullsrc->SrcRegisterExtMod.Negate = 1; - } - - if( inst->SrcReg[i].RelAddr ) { - fullsrc->SrcRegister.Indirect = 1; - - fullsrc->SrcRegisterInd.File = TGSI_FILE_ADDRESS; - fullsrc->SrcRegisterInd.Index = 0; - } - } - - switch( inst->Opcode ) { - case OPCODE_ARL: - fullinst->Instruction.Opcode = TGSI_OPCODE_ARL; - break; - case OPCODE_ABS: - fullinst->Instruction.Opcode = TGSI_OPCODE_ABS; - break; - case OPCODE_ADD: - fullinst->Instruction.Opcode = TGSI_OPCODE_ADD; - break; - case OPCODE_BGNLOOP: - fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_BGNSUB: - fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB; - break; - case OPCODE_BRA: - fullinst->Instruction.Opcode = TGSI_OPCODE_BRA; - break; - case OPCODE_BRK: - fullinst->Instruction.Opcode = TGSI_OPCODE_BRK; - break; - case OPCODE_CAL: - fullinst->Instruction.Opcode = TGSI_OPCODE_CAL; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_CMP: - fullinst->Instruction.Opcode = TGSI_OPCODE_CMP; - break; - case OPCODE_CONT: - fullinst->Instruction.Opcode = TGSI_OPCODE_CONT; - break; - case OPCODE_COS: - fullinst->Instruction.Opcode = TGSI_OPCODE_COS; - break; - case OPCODE_DDX: - fullinst->Instruction.Opcode = TGSI_OPCODE_DDX; - break; - case OPCODE_DDY: - fullinst->Instruction.Opcode = TGSI_OPCODE_DDY; - break; - case OPCODE_DP3: - fullinst->Instruction.Opcode = TGSI_OPCODE_DP3; - break; - case OPCODE_DP4: - fullinst->Instruction.Opcode = TGSI_OPCODE_DP4; - break; - case OPCODE_DPH: - fullinst->Instruction.Opcode = TGSI_OPCODE_DPH; - break; - case OPCODE_DST: - fullinst->Instruction.Opcode = TGSI_OPCODE_DST; - break; - case OPCODE_ELSE: - fullinst->Instruction.Opcode = TGSI_OPCODE_ELSE; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_ENDIF: - fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF; - break; - case OPCODE_ENDLOOP: - fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_ENDSUB: - fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB; - break; - case OPCODE_EX2: - fullinst->Instruction.Opcode = TGSI_OPCODE_EX2; - break; - case OPCODE_EXP: - fullinst->Instruction.Opcode = TGSI_OPCODE_EXP; - break; - case OPCODE_FLR: - fullinst->Instruction.Opcode = TGSI_OPCODE_FLR; - break; - case OPCODE_FRC: - fullinst->Instruction.Opcode = TGSI_OPCODE_FRC; - break; - case OPCODE_IF: - fullinst->Instruction.Opcode = TGSI_OPCODE_IF; - fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; - break; - case OPCODE_INT: - fullinst->Instruction.Opcode = TGSI_OPCODE_INT; - break; - case OPCODE_KIL: - /* predicated w/ a register */ - fullinst->Instruction.Opcode = TGSI_OPCODE_KILP; - break; - case OPCODE_KIL_NV: - /* unpredicated */ - assert(inst->DstReg.CondMask == COND_TR); - fullinst->Instruction.Opcode = TGSI_OPCODE_KIL; - break; - case OPCODE_LG2: - fullinst->Instruction.Opcode = TGSI_OPCODE_LG2; - break; - case OPCODE_LOG: - fullinst->Instruction.Opcode = TGSI_OPCODE_LOG; - break; - case OPCODE_LIT: - fullinst->Instruction.Opcode = TGSI_OPCODE_LIT; - break; - case OPCODE_LRP: - fullinst->Instruction.Opcode = TGSI_OPCODE_LRP; - break; - case OPCODE_MAD: - fullinst->Instruction.Opcode = TGSI_OPCODE_MAD; - break; - case OPCODE_MAX: - fullinst->Instruction.Opcode = TGSI_OPCODE_MAX; - break; - case OPCODE_MIN: - fullinst->Instruction.Opcode = TGSI_OPCODE_MIN; - break; - case OPCODE_MOV: - fullinst->Instruction.Opcode = TGSI_OPCODE_MOV; - break; - case OPCODE_MUL: - fullinst->Instruction.Opcode = TGSI_OPCODE_MUL; - break; - case OPCODE_NOISE1: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE1; - break; - case OPCODE_NOISE2: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE2; - break; - case OPCODE_NOISE3: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE3; - break; - case OPCODE_NOISE4: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE4; - break; - case OPCODE_NOP: - fullinst->Instruction.Opcode = TGSI_OPCODE_NOP; - break; - case OPCODE_POW: - fullinst->Instruction.Opcode = TGSI_OPCODE_POW; - break; - case OPCODE_RCP: - fullinst->Instruction.Opcode = TGSI_OPCODE_RCP; - break; - case OPCODE_RET: - fullinst->Instruction.Opcode = TGSI_OPCODE_RET; - break; - case OPCODE_RSQ: - fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ; - tgsi_util_set_full_src_register_sign_mode( - &fullinst->FullSrcRegisters[0], - TGSI_UTIL_SIGN_CLEAR ); - break; - case OPCODE_SCS: - fullinst->Instruction.Opcode = TGSI_OPCODE_SCS; - fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XY; - break; - case OPCODE_SEQ: - fullinst->Instruction.Opcode = TGSI_OPCODE_SEQ; - break; - case OPCODE_SGE: - fullinst->Instruction.Opcode = TGSI_OPCODE_SGE; - break; - case OPCODE_SGT: - fullinst->Instruction.Opcode = TGSI_OPCODE_SGT; - break; - case OPCODE_SIN: - fullinst->Instruction.Opcode = TGSI_OPCODE_SIN; - break; - case OPCODE_SLE: - fullinst->Instruction.Opcode = TGSI_OPCODE_SLE; - break; - case OPCODE_SLT: - fullinst->Instruction.Opcode = TGSI_OPCODE_SLT; - break; - case OPCODE_SNE: - fullinst->Instruction.Opcode = TGSI_OPCODE_SNE; - break; - case OPCODE_SUB: - fullinst->Instruction.Opcode = TGSI_OPCODE_SUB; - break; - case OPCODE_SWZ: - fullinst->Instruction.Opcode = TGSI_OPCODE_SWZ; - break; - case OPCODE_TEX: - /* ordinary texture lookup */ - fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; - fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_TXB: - /* texture lookup with LOD bias */ - fullinst->Instruction.Opcode = TGSI_OPCODE_TXB; - fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_TXD: - /* texture lookup with explicit partial derivatives */ - fullinst->Instruction.Opcode = TGSI_OPCODE_TXD; - fullinst->Instruction.NumSrcRegs = 4; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - /* src[0] = coord, src[1] = d[strq]/dx, src[2] = d[strq]/dy */ - fullinst->FullSrcRegisters[3].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[3].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_TXL: - /* texture lookup with explicit LOD */ - fullinst->Instruction.Opcode = TGSI_OPCODE_TXL; - fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_TXP: - /* texture lookup with divide by Q component */ - /* convert to TEX w/ special flag for division */ - fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; - fullinst->Instruction.NumSrcRegs = 2; - fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide = TGSI_EXTSWIZZLE_W; - fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; - break; - case OPCODE_XPD: - fullinst->Instruction.Opcode = TGSI_OPCODE_XPD; - fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; - break; - case OPCODE_END: - fullinst->Instruction.Opcode = TGSI_OPCODE_RET; - break; - default: - assert( 0 ); - } -} - -/** - * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens - */ -static struct tgsi_full_declaration -make_input_decl( - GLuint index, - GLuint interpolate, - GLuint usage_mask, - GLboolean semantic_info, - GLuint semantic_name, - GLbitfield semantic_index ) -{ - struct tgsi_full_declaration decl; - - assert(semantic_name < TGSI_SEMANTIC_COUNT); - - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.Declaration.UsageMask = usage_mask; - decl.Declaration.Semantic = semantic_info; - decl.Declaration.Interpolate = 1; - decl.u.DeclarationRange.First = index; - decl.u.DeclarationRange.Last = index; - if (semantic_info) { - decl.Semantic.SemanticName = semantic_name; - decl.Semantic.SemanticIndex = semantic_index; - } - decl.Interpolation.Interpolate = interpolate; - - return decl; -} - -/** - * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens - */ -static struct tgsi_full_declaration -make_output_decl( - GLuint index, - GLuint semantic_name, - GLuint semantic_index, - GLbitfield usage_mask ) -{ - struct tgsi_full_declaration decl; - - assert(semantic_name < TGSI_SEMANTIC_COUNT); - - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.Declaration.UsageMask = usage_mask; - decl.Declaration.Semantic = 1; - decl.u.DeclarationRange.First = index; - decl.u.DeclarationRange.Last = index; - decl.Semantic.SemanticName = semantic_name; - decl.Semantic.SemanticIndex = semantic_index; - - return decl; -} - - -static struct tgsi_full_declaration -make_temp_decl( - GLuint start_index, - GLuint end_index ) -{ - struct tgsi_full_declaration decl; - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.u.DeclarationRange.First = start_index; - decl.u.DeclarationRange.Last = end_index; - return decl; -} - - -/** - * Find the temporaries which are used in the given program. - */ -static void -find_temporaries(const struct gl_program *program, - GLboolean tempsUsed[MAX_PROGRAM_TEMPS]) -{ - GLuint i, j; - - for (i = 0; i < MAX_PROGRAM_TEMPS; i++) - tempsUsed[i] = GL_FALSE; - - for (i = 0; i < program->NumInstructions; i++) { - const struct prog_instruction *inst = program->Instructions + i; - const GLuint n = _mesa_num_inst_src_regs( inst->Opcode ); - for (j = 0; j < n; j++) { - if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) - tempsUsed[inst->SrcReg[j].Index] = GL_TRUE; - if (inst->DstReg.File == PROGRAM_TEMPORARY) - tempsUsed[inst->DstReg.Index] = GL_TRUE; - } - } -} - - - - -/** - * Translate Mesa program to TGSI format. - * \param program the program to translate - * \param numInputs number of input registers used - * \param inputMapping maps Mesa fragment program inputs to TGSI generic - * input indexes - * \param inputSemanticName the TGSI_SEMANTIC flag for each input - * \param inputSemanticIndex the semantic index (ex: which texcoord) for each input - * \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input - - * \param numOutputs number of output registers used - * \param outputMapping maps Mesa fragment program outputs to TGSI - * generic outputs - * \param outputSemanticName the TGSI_SEMANTIC flag for each output - * \param outputSemanticIndex the semantic index (ex: which texcoord) for each output - * \param tokens array to store translated tokens in - * \param maxTokens size of the tokens array - * - */ -GLboolean -tgsi_translate_mesa_program( - uint procType, - const struct gl_program *program, - GLuint numInputs, - const GLuint inputMapping[], - const ubyte inputSemanticName[], - const ubyte inputSemanticIndex[], - const GLuint interpMode[], - GLuint numOutputs, - const GLuint outputMapping[], - const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - struct tgsi_token *tokens, - GLuint maxTokens ) -{ - GLuint i; - GLuint ti; /* token index */ - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_instruction fullinst; - GLuint preamble_size = 0; - GLuint immediates[1000]; -#if EMIT_IMMEDIATES - GLuint numImmediates = 0; -#endif - - assert(procType == TGSI_PROCESSOR_FRAGMENT || - procType == TGSI_PROCESSOR_VERTEX); - - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); - - ti = 3; - - /* - * Declare input attributes. - */ - if (procType == TGSI_PROCESSOR_FRAGMENT) { - for (i = 0; i < numInputs; i++) { - struct tgsi_full_declaration fulldecl; - switch (inputSemanticName[i]) { - case TGSI_SEMANTIC_POSITION: - /* Fragment XY pos */ - fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_CONSTANT, - TGSI_WRITEMASK_XY, - GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - /* Fragment ZW pos */ - fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_LINEAR, - TGSI_WRITEMASK_ZW, - GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - default: - fulldecl = make_input_decl(i, - interpMode[i], - TGSI_WRITEMASK_XYZW, - GL_TRUE, inputSemanticName[i], - inputSemanticIndex[i]); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - } - } - } - else { - /* vertex prog */ - for (i = 0; i < numInputs; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_ATTRIB, - TGSI_WRITEMASK_XYZW, - GL_FALSE, inputSemanticName[i], - inputSemanticIndex[i]); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - - /* - * Declare output attributes. - */ - if (procType == TGSI_PROCESSOR_FRAGMENT) { - for (i = 0; i < numOutputs; i++) { - struct tgsi_full_declaration fulldecl; - switch (outputSemanticName[i]) { - case TGSI_SEMANTIC_POSITION: - fulldecl = make_output_decl(i, - TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ - TGSI_WRITEMASK_Z ); - break; - case TGSI_SEMANTIC_COLOR: - fulldecl = make_output_decl(i, - TGSI_SEMANTIC_COLOR, 0, - TGSI_WRITEMASK_XYZW ); - break; - default: - abort(); - } - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - else { - /* vertex prog */ - for (i = 0; i < numOutputs; i++) { - struct tgsi_full_declaration fulldecl; - fulldecl = make_output_decl(i, - outputSemanticName[i], - outputSemanticIndex[i], - TGSI_WRITEMASK_XYZW ); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - - /* temporary decls */ - { - GLboolean tempsUsed[MAX_PROGRAM_TEMPS + 1]; - GLboolean inside_range = GL_FALSE; - GLuint start_range; - - find_temporaries(program, tempsUsed); - tempsUsed[MAX_PROGRAM_TEMPS] = GL_FALSE; - for (i = 0; i < MAX_PROGRAM_TEMPS + 1; i++) { - if (tempsUsed[i] && !inside_range) { - inside_range = GL_TRUE; - start_range = i; - } - else if (!tempsUsed[i] && inside_range) { - struct tgsi_full_declaration fulldecl; - - inside_range = GL_FALSE; - fulldecl = make_temp_decl( start_range, i - 1 ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - } - } - } - - /* immediates/literals */ -#if EMIT_IMMEDIATES - for (i = 0; i < program->Parameters->NumParameters; i++) { - if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { - struct tgsi_full_immediate fullimm - = make_immediate(program->Parameters->ParameterValues[i], - program->Parameters->Parameters[i].Size); - ti += tgsi_build_full_immediate(&fullimm, - &tokens[ti], - header, - maxTokens - ti); - immediates[i] = numImmediates; - numImmediates++; - } - } -#endif - - for( i = 0; i < program->NumInstructions; i++ ) { - compile_instruction( - &program->Instructions[i], - &fullinst, - inputMapping, - outputMapping, - immediates, - preamble_size, - procType ); - - ti += tgsi_build_full_instruction( - &fullinst, - &tokens[ti], - header, - maxTokens - ti ); - } - - return GL_TRUE; -} - +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * \author + * Michal Krol + */ + + +#include "tgsi_platform.h" +#include "pipe/tgsi/exec/tgsi_core.h" +#include "st_mesa_to_tgsi.h" +#include "shader/prog_parameter.h" + +#define TGSI_DEBUG 0 + +#define EMIT_IMMEDIATES 0 + + +/* + * Map mesa register file to TGSI register file. + */ +static GLuint +map_register_file( + enum register_file file ) +{ + switch( file ) { + case PROGRAM_UNDEFINED: + return TGSI_FILE_NULL; + case PROGRAM_TEMPORARY: + return TGSI_FILE_TEMPORARY; + //case PROGRAM_LOCAL_PARAM: + //case PROGRAM_ENV_PARAM: + case PROGRAM_STATE_VAR: + case PROGRAM_NAMED_PARAM: + case PROGRAM_UNIFORM: + return TGSI_FILE_CONSTANT; + case PROGRAM_CONSTANT: +#if EMIT_IMMEDIATES + return TGSI_FILE_IMMEDIATE; +#else + return TGSI_FILE_CONSTANT; +#endif + case PROGRAM_INPUT: + return TGSI_FILE_INPUT; + case PROGRAM_OUTPUT: + return TGSI_FILE_OUTPUT; + case PROGRAM_ADDRESS: + return TGSI_FILE_ADDRESS; + default: + assert( 0 ); + return TGSI_FILE_NULL; + } +} + +/** + * Map mesa register file index to TGSI index. + * Take special care when processing input and output indices. + * \param file one of TGSI_FILE_x + * \param index the mesa register file index + * \param inputMapping maps Mesa input indexes to TGSI input indexes + * \param outputMapping maps Mesa output indexes to TGSI output indexes + */ +static GLuint +map_register_file_index( + GLuint file, + GLuint index, + const GLuint inputMapping[], + const GLuint outputMapping[], + const GLuint immediateMapping[]) +{ + switch( file ) { + case TGSI_FILE_INPUT: + /* inputs are mapped according to the user-defined map */ + return inputMapping[index]; + + case TGSI_FILE_OUTPUT: + return outputMapping[index]; + +#if EMIT_IMMEDIATES + case TGSI_FILE_IMMEDIATE: + return immediateMapping[index]; +#endif + + default: + return index; + } +} + +/* + * Map mesa texture target to TGSI texture target. + */ +static GLuint +map_texture_target( + GLuint textarget ) +{ + switch( textarget ) { + case TEXTURE_1D_INDEX: + return TGSI_TEXTURE_1D; + case TEXTURE_2D_INDEX: + return TGSI_TEXTURE_2D; + case TEXTURE_3D_INDEX: + return TGSI_TEXTURE_3D; + case TEXTURE_CUBE_INDEX: + return TGSI_TEXTURE_CUBE; + case TEXTURE_RECT_INDEX: + return TGSI_TEXTURE_RECT; + default: + assert( 0 ); + } + + return TGSI_TEXTURE_1D; +} + +static GLuint +convert_sat( + GLuint sat ) +{ + switch( sat ) { + case SATURATE_OFF: + return TGSI_SAT_NONE; + case SATURATE_ZERO_ONE: + return TGSI_SAT_ZERO_ONE; + case SATURATE_PLUS_MINUS_ONE: + return TGSI_SAT_MINUS_PLUS_ONE; + default: + assert( 0 ); + return TGSI_SAT_NONE; + } +} + +static GLuint +convert_writemask( + GLuint writemask ) +{ + assert( WRITEMASK_X == TGSI_WRITEMASK_X ); + assert( WRITEMASK_Y == TGSI_WRITEMASK_Y ); + assert( WRITEMASK_Z == TGSI_WRITEMASK_Z ); + assert( WRITEMASK_W == TGSI_WRITEMASK_W ); + assert( (writemask & ~TGSI_WRITEMASK_XYZW) == 0 ); + + return writemask; +} + +#if EMIT_IMMEDIATES +static struct tgsi_full_immediate +make_immediate(const float *value, uint size) +{ + struct tgsi_full_immediate imm; + imm.Immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + imm.Immediate.Size = 1 + size; /* one for the token itself */ + imm.Immediate.DataType = TGSI_IMM_FLOAT32; + imm.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; + return imm; +} +#endif + +static void +compile_instruction( + const struct prog_instruction *inst, + struct tgsi_full_instruction *fullinst, + const GLuint inputMapping[], + const GLuint outputMapping[], + const GLuint immediateMapping[], + GLuint preamble_size, + GLuint processor ) +{ + GLuint i; + struct tgsi_full_dst_register *fulldst; + struct tgsi_full_src_register *fullsrc; + + *fullinst = tgsi_default_full_instruction(); + + fullinst->Instruction.Saturate = convert_sat( inst->SaturateMode ); + fullinst->Instruction.NumDstRegs = _mesa_num_inst_dst_regs( inst->Opcode ); + fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode ); + + fulldst = &fullinst->FullDstRegisters[0]; + fulldst->DstRegister.File = map_register_file( inst->DstReg.File ); + fulldst->DstRegister.Index = map_register_file_index( + fulldst->DstRegister.File, + inst->DstReg.Index, + inputMapping, + outputMapping, + NULL + ); + fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); + + for( i = 0; i < fullinst->Instruction.NumSrcRegs; i++ ) { + GLuint j; + + fullsrc = &fullinst->FullSrcRegisters[i]; + fullsrc->SrcRegister.File = map_register_file( inst->SrcReg[i].File ); + fullsrc->SrcRegister.Index = map_register_file_index( + fullsrc->SrcRegister.File, + inst->SrcReg[i].Index, + inputMapping, + outputMapping, + immediateMapping); + + for( j = 0; j < 4; j++ ) { + GLuint swz; + + swz = GET_SWZ( inst->SrcReg[i].Swizzle, j ); + if( swz > SWIZZLE_W ) { + tgsi_util_set_src_register_extswizzle( + &fullsrc->SrcRegisterExtSwz, + swz, + j ); + } + else { + tgsi_util_set_src_register_swizzle( + &fullsrc->SrcRegister, + swz, + j ); + } + } + + if( inst->SrcReg[i].NegateBase == NEGATE_XYZW ) { + fullsrc->SrcRegister.Negate = 1; + } + else if( inst->SrcReg[i].NegateBase != NEGATE_NONE ) { + if( inst->SrcReg[i].NegateBase & NEGATE_X ) { + fullsrc->SrcRegisterExtSwz.NegateX = 1; + } + if( inst->SrcReg[i].NegateBase & NEGATE_Y ) { + fullsrc->SrcRegisterExtSwz.NegateY = 1; + } + if( inst->SrcReg[i].NegateBase & NEGATE_Z ) { + fullsrc->SrcRegisterExtSwz.NegateZ = 1; + } + if( inst->SrcReg[i].NegateBase & NEGATE_W ) { + fullsrc->SrcRegisterExtSwz.NegateW = 1; + } + } + + if( inst->SrcReg[i].Abs ) { + fullsrc->SrcRegisterExtMod.Absolute = 1; + } + + if( inst->SrcReg[i].NegateAbs ) { + fullsrc->SrcRegisterExtMod.Negate = 1; + } + + if( inst->SrcReg[i].RelAddr ) { + fullsrc->SrcRegister.Indirect = 1; + + fullsrc->SrcRegisterInd.File = TGSI_FILE_ADDRESS; + fullsrc->SrcRegisterInd.Index = 0; + } + } + + switch( inst->Opcode ) { + case OPCODE_ARL: + fullinst->Instruction.Opcode = TGSI_OPCODE_ARL; + break; + case OPCODE_ABS: + fullinst->Instruction.Opcode = TGSI_OPCODE_ABS; + break; + case OPCODE_ADD: + fullinst->Instruction.Opcode = TGSI_OPCODE_ADD; + break; + case OPCODE_BGNLOOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_BGNLOOP2; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_BGNSUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB; + break; + case OPCODE_BRA: + fullinst->Instruction.Opcode = TGSI_OPCODE_BRA; + break; + case OPCODE_BRK: + fullinst->Instruction.Opcode = TGSI_OPCODE_BRK; + break; + case OPCODE_CAL: + fullinst->Instruction.Opcode = TGSI_OPCODE_CAL; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_CMP: + fullinst->Instruction.Opcode = TGSI_OPCODE_CMP; + break; + case OPCODE_CONT: + fullinst->Instruction.Opcode = TGSI_OPCODE_CONT; + break; + case OPCODE_COS: + fullinst->Instruction.Opcode = TGSI_OPCODE_COS; + break; + case OPCODE_DDX: + fullinst->Instruction.Opcode = TGSI_OPCODE_DDX; + break; + case OPCODE_DDY: + fullinst->Instruction.Opcode = TGSI_OPCODE_DDY; + break; + case OPCODE_DP3: + fullinst->Instruction.Opcode = TGSI_OPCODE_DP3; + break; + case OPCODE_DP4: + fullinst->Instruction.Opcode = TGSI_OPCODE_DP4; + break; + case OPCODE_DPH: + fullinst->Instruction.Opcode = TGSI_OPCODE_DPH; + break; + case OPCODE_DST: + fullinst->Instruction.Opcode = TGSI_OPCODE_DST; + break; + case OPCODE_ELSE: + fullinst->Instruction.Opcode = TGSI_OPCODE_ELSE; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_ENDIF: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDIF; + break; + case OPCODE_ENDLOOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDLOOP2; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_ENDSUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB; + break; + case OPCODE_EX2: + fullinst->Instruction.Opcode = TGSI_OPCODE_EX2; + break; + case OPCODE_EXP: + fullinst->Instruction.Opcode = TGSI_OPCODE_EXP; + break; + case OPCODE_FLR: + fullinst->Instruction.Opcode = TGSI_OPCODE_FLR; + break; + case OPCODE_FRC: + fullinst->Instruction.Opcode = TGSI_OPCODE_FRC; + break; + case OPCODE_IF: + fullinst->Instruction.Opcode = TGSI_OPCODE_IF; + fullinst->InstructionExtLabel.Label = inst->BranchTarget + preamble_size; + break; + case OPCODE_INT: + fullinst->Instruction.Opcode = TGSI_OPCODE_INT; + break; + case OPCODE_KIL: + /* predicated w/ a register */ + fullinst->Instruction.Opcode = TGSI_OPCODE_KILP; + break; + case OPCODE_KIL_NV: + /* unpredicated */ + assert(inst->DstReg.CondMask == COND_TR); + fullinst->Instruction.Opcode = TGSI_OPCODE_KIL; + break; + case OPCODE_LG2: + fullinst->Instruction.Opcode = TGSI_OPCODE_LG2; + break; + case OPCODE_LOG: + fullinst->Instruction.Opcode = TGSI_OPCODE_LOG; + break; + case OPCODE_LIT: + fullinst->Instruction.Opcode = TGSI_OPCODE_LIT; + break; + case OPCODE_LRP: + fullinst->Instruction.Opcode = TGSI_OPCODE_LRP; + break; + case OPCODE_MAD: + fullinst->Instruction.Opcode = TGSI_OPCODE_MAD; + break; + case OPCODE_MAX: + fullinst->Instruction.Opcode = TGSI_OPCODE_MAX; + break; + case OPCODE_MIN: + fullinst->Instruction.Opcode = TGSI_OPCODE_MIN; + break; + case OPCODE_MOV: + fullinst->Instruction.Opcode = TGSI_OPCODE_MOV; + break; + case OPCODE_MUL: + fullinst->Instruction.Opcode = TGSI_OPCODE_MUL; + break; + case OPCODE_NOISE1: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE1; + break; + case OPCODE_NOISE2: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE2; + break; + case OPCODE_NOISE3: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE3; + break; + case OPCODE_NOISE4: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOISE4; + break; + case OPCODE_NOP: + fullinst->Instruction.Opcode = TGSI_OPCODE_NOP; + break; + case OPCODE_POW: + fullinst->Instruction.Opcode = TGSI_OPCODE_POW; + break; + case OPCODE_RCP: + fullinst->Instruction.Opcode = TGSI_OPCODE_RCP; + break; + case OPCODE_RET: + fullinst->Instruction.Opcode = TGSI_OPCODE_RET; + break; + case OPCODE_RSQ: + fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ; + tgsi_util_set_full_src_register_sign_mode( + &fullinst->FullSrcRegisters[0], + TGSI_UTIL_SIGN_CLEAR ); + break; + case OPCODE_SCS: + fullinst->Instruction.Opcode = TGSI_OPCODE_SCS; + fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XY; + break; + case OPCODE_SEQ: + fullinst->Instruction.Opcode = TGSI_OPCODE_SEQ; + break; + case OPCODE_SGE: + fullinst->Instruction.Opcode = TGSI_OPCODE_SGE; + break; + case OPCODE_SGT: + fullinst->Instruction.Opcode = TGSI_OPCODE_SGT; + break; + case OPCODE_SIN: + fullinst->Instruction.Opcode = TGSI_OPCODE_SIN; + break; + case OPCODE_SLE: + fullinst->Instruction.Opcode = TGSI_OPCODE_SLE; + break; + case OPCODE_SLT: + fullinst->Instruction.Opcode = TGSI_OPCODE_SLT; + break; + case OPCODE_SNE: + fullinst->Instruction.Opcode = TGSI_OPCODE_SNE; + break; + case OPCODE_SUB: + fullinst->Instruction.Opcode = TGSI_OPCODE_SUB; + break; + case OPCODE_SWZ: + fullinst->Instruction.Opcode = TGSI_OPCODE_SWZ; + break; + case OPCODE_TEX: + /* ordinary texture lookup */ + fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXB: + /* texture lookup with LOD bias */ + fullinst->Instruction.Opcode = TGSI_OPCODE_TXB; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXD: + /* texture lookup with explicit partial derivatives */ + fullinst->Instruction.Opcode = TGSI_OPCODE_TXD; + fullinst->Instruction.NumSrcRegs = 4; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + /* src[0] = coord, src[1] = d[strq]/dx, src[2] = d[strq]/dy */ + fullinst->FullSrcRegisters[3].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[3].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXL: + /* texture lookup with explicit LOD */ + fullinst->Instruction.Opcode = TGSI_OPCODE_TXL; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_TXP: + /* texture lookup with divide by Q component */ + /* convert to TEX w/ special flag for division */ + fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; + fullinst->Instruction.NumSrcRegs = 2; + fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); + fullinst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide = TGSI_EXTSWIZZLE_W; + fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; + break; + case OPCODE_XPD: + fullinst->Instruction.Opcode = TGSI_OPCODE_XPD; + fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; + break; + case OPCODE_END: + fullinst->Instruction.Opcode = TGSI_OPCODE_RET; + break; + default: + assert( 0 ); + } +} + +/** + * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens + */ +static struct tgsi_full_declaration +make_input_decl( + GLuint index, + GLuint interpolate, + GLuint usage_mask, + GLboolean semantic_info, + GLuint semantic_name, + GLbitfield semantic_index ) +{ + struct tgsi_full_declaration decl; + + assert(semantic_name < TGSI_SEMANTIC_COUNT); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.Declaration.UsageMask = usage_mask; + decl.Declaration.Semantic = semantic_info; + decl.Declaration.Interpolate = 1; + decl.u.DeclarationRange.First = index; + decl.u.DeclarationRange.Last = index; + if (semantic_info) { + decl.Semantic.SemanticName = semantic_name; + decl.Semantic.SemanticIndex = semantic_index; + } + decl.Interpolation.Interpolate = interpolate; + + return decl; +} + +/** + * \param usage_mask bitfield of TGSI_WRITEMASK_{XYZW} tokens + */ +static struct tgsi_full_declaration +make_output_decl( + GLuint index, + GLuint semantic_name, + GLuint semantic_index, + GLbitfield usage_mask ) +{ + struct tgsi_full_declaration decl; + + assert(semantic_name < TGSI_SEMANTIC_COUNT); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.Declaration.UsageMask = usage_mask; + decl.Declaration.Semantic = 1; + decl.u.DeclarationRange.First = index; + decl.u.DeclarationRange.Last = index; + decl.Semantic.SemanticName = semantic_name; + decl.Semantic.SemanticIndex = semantic_index; + + return decl; +} + + +static struct tgsi_full_declaration +make_temp_decl( + GLuint start_index, + GLuint end_index ) +{ + struct tgsi_full_declaration decl; + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.u.DeclarationRange.First = start_index; + decl.u.DeclarationRange.Last = end_index; + return decl; +} + + +/** + * Find the temporaries which are used in the given program. + */ +static void +find_temporaries(const struct gl_program *program, + GLboolean tempsUsed[MAX_PROGRAM_TEMPS]) +{ + GLuint i, j; + + for (i = 0; i < MAX_PROGRAM_TEMPS; i++) + tempsUsed[i] = GL_FALSE; + + for (i = 0; i < program->NumInstructions; i++) { + const struct prog_instruction *inst = program->Instructions + i; + const GLuint n = _mesa_num_inst_src_regs( inst->Opcode ); + for (j = 0; j < n; j++) { + if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) + tempsUsed[inst->SrcReg[j].Index] = GL_TRUE; + if (inst->DstReg.File == PROGRAM_TEMPORARY) + tempsUsed[inst->DstReg.Index] = GL_TRUE; + } + } +} + + + + +/** + * Translate Mesa program to TGSI format. + * \param program the program to translate + * \param numInputs number of input registers used + * \param inputMapping maps Mesa fragment program inputs to TGSI generic + * input indexes + * \param inputSemanticName the TGSI_SEMANTIC flag for each input + * \param inputSemanticIndex the semantic index (ex: which texcoord) for each input + * \param interpMode the TGSI_INTERPOLATE_LINEAR/PERSP mode for each input + + * \param numOutputs number of output registers used + * \param outputMapping maps Mesa fragment program outputs to TGSI + * generic outputs + * \param outputSemanticName the TGSI_SEMANTIC flag for each output + * \param outputSemanticIndex the semantic index (ex: which texcoord) for each output + * \param tokens array to store translated tokens in + * \param maxTokens size of the tokens array + * + */ +GLboolean +tgsi_translate_mesa_program( + uint procType, + const struct gl_program *program, + GLuint numInputs, + const GLuint inputMapping[], + const ubyte inputSemanticName[], + const ubyte inputSemanticIndex[], + const GLuint interpMode[], + GLuint numOutputs, + const GLuint outputMapping[], + const ubyte outputSemanticName[], + const ubyte outputSemanticIndex[], + struct tgsi_token *tokens, + GLuint maxTokens ) +{ + GLuint i; + GLuint ti; /* token index */ + struct tgsi_header *header; + struct tgsi_processor *processor; + struct tgsi_full_instruction fullinst; + GLuint preamble_size = 0; + GLuint immediates[1000]; +#if EMIT_IMMEDIATES + GLuint numImmediates = 0; +#endif + + assert(procType == TGSI_PROCESSOR_FRAGMENT || + procType == TGSI_PROCESSOR_VERTEX); + + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header(); + + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor( procType, header ); + + ti = 3; + + /* + * Declare input attributes. + */ + if (procType == TGSI_PROCESSOR_FRAGMENT) { + for (i = 0; i < numInputs; i++) { + struct tgsi_full_declaration fulldecl; + switch (inputSemanticName[i]) { + case TGSI_SEMANTIC_POSITION: + /* Fragment XY pos */ + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_CONSTANT, + TGSI_WRITEMASK_XY, + GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + /* Fragment ZW pos */ + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_LINEAR, + TGSI_WRITEMASK_ZW, + GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + default: + fulldecl = make_input_decl(i, + interpMode[i], + TGSI_WRITEMASK_XYZW, + GL_TRUE, inputSemanticName[i], + inputSemanticIndex[i]); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + break; + } + } + } + else { + /* vertex prog */ + for (i = 0; i < numInputs; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_input_decl(i, + TGSI_INTERPOLATE_ATTRIB, + TGSI_WRITEMASK_XYZW, + GL_FALSE, inputSemanticName[i], + inputSemanticIndex[i]); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + + /* + * Declare output attributes. + */ + if (procType == TGSI_PROCESSOR_FRAGMENT) { + for (i = 0; i < numOutputs; i++) { + struct tgsi_full_declaration fulldecl; + switch (outputSemanticName[i]) { + case TGSI_SEMANTIC_POSITION: + fulldecl = make_output_decl(i, + TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ + TGSI_WRITEMASK_Z ); + break; + case TGSI_SEMANTIC_COLOR: + fulldecl = make_output_decl(i, + TGSI_SEMANTIC_COLOR, 0, + TGSI_WRITEMASK_XYZW ); + break; + default: + abort(); + } + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + else { + /* vertex prog */ + for (i = 0; i < numOutputs; i++) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_output_decl(i, + outputSemanticName[i], + outputSemanticIndex[i], + TGSI_WRITEMASK_XYZW ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + + /* temporary decls */ + { + GLboolean tempsUsed[MAX_PROGRAM_TEMPS + 1]; + GLboolean inside_range = GL_FALSE; + GLuint start_range; + + find_temporaries(program, tempsUsed); + tempsUsed[MAX_PROGRAM_TEMPS] = GL_FALSE; + for (i = 0; i < MAX_PROGRAM_TEMPS + 1; i++) { + if (tempsUsed[i] && !inside_range) { + inside_range = GL_TRUE; + start_range = i; + } + else if (!tempsUsed[i] && inside_range) { + struct tgsi_full_declaration fulldecl; + + inside_range = GL_FALSE; + fulldecl = make_temp_decl( start_range, i - 1 ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + } + + /* immediates/literals */ +#if EMIT_IMMEDIATES + for (i = 0; i < program->Parameters->NumParameters; i++) { + if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { + struct tgsi_full_immediate fullimm + = make_immediate(program->Parameters->ParameterValues[i], + program->Parameters->Parameters[i].Size); + ti += tgsi_build_full_immediate(&fullimm, + &tokens[ti], + header, + maxTokens - ti); + immediates[i] = numImmediates; + numImmediates++; + } + } +#endif + + for( i = 0; i < program->NumInstructions; i++ ) { + compile_instruction( + &program->Instructions[i], + &fullinst, + inputMapping, + outputMapping, + immediates, + preamble_size, + procType ); + + ti += tgsi_build_full_instruction( + &fullinst, + &tokens[ti], + header, + maxTokens - ti ); + } + + return GL_TRUE; +} + -- cgit v1.2.3 From dbb33a9710e64243ba8c69f44a149376300da2aa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 26 Oct 2007 19:30:04 +0100 Subject: Start a vertex buffer constuction stage for i915 based on Keith's draft done on softpipe. --- src/mesa/pipe/i915simple/Makefile | 1 + src/mesa/pipe/i915simple/i915_context.h | 6 + src/mesa/pipe/i915simple/i915_prim_vbuf.c | 314 ++++++++++++++++++++++++++++++ 3 files changed, 321 insertions(+) create mode 100644 src/mesa/pipe/i915simple/i915_prim_vbuf.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 670350c0a5..391a084915 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -21,6 +21,7 @@ DRIVER_SOURCES = \ i915_state_sampler.c \ i915_strings.c \ i915_prim_emit.c \ + i915_prim_vbuf.c \ i915_tex_layout.c \ i915_fpc_emit.c \ i915_fpc_translate.c \ diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 5a3ecedad2..5d312286f6 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -236,6 +236,12 @@ struct i915_context struct draw_stage *i915_draw_render_stage( struct i915_context *i915 ); +/*********************************************************************** + * i915_prim_vbuf.c: + */ +struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ); + + /*********************************************************************** * i915_state_emit.c: */ diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c new file mode 100644 index 0000000000..a80af00137 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -0,0 +1,314 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Build post-transformation, post-clipping vertex buffers and element + * lists by hooking into the end of the primitive pipeline and + * manipulating the vertex_id field in the vertex headers. + * + * XXX: work in progress + * + * \author José Fonseca + * \author Keith Whitwell + */ + + +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_vertex.h" +#include "pipe/p_util.h" + +#include "softpipe/sp_context.h" +#include "softpipe/sp_headers.h" +#include "softpipe/sp_quad.h" +#include "softpipe/sp_prim_setup.h" + +#include "i915_context.h" + + +static void vbuf_flush_elements( struct draw_stage *stage ); + + +#define VBUF_SIZE (64*1024) +#define IBUF_SIZE (16*1024) + + +/** + * Vertex buffer emit stage. + */ +struct vbuf_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + vbuf_draw_func draw; + + /* Vertices are passed in as an array of floats making up each + * attribute in turn. Will eventually convert to hardware format + * in this stage. + */ + char *vertex_map; + char *vertex_ptr; + unsigned vertex_size; + unsigned nr_vertices; + + unsigned max_vertices; + + ushort *element_map; + unsigned nr_elements; + + unsigned prim; + + struct i915_context *i915; +}; + + +/** + * Basically a cast wrapper. + */ +static INLINE struct vbuf_stage *vbuf_stage( struct draw_stage *stage ) +{ + return (struct vbuf_stage *)stage; +} + + +static boolean overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) +{ + unsigned long used = (char *)ptr - (char *)map; + return (used + bytes) > bufsz; +} + + +static boolean check_space( struct vbuf_stage *vbuf ) +{ + if (overflow( vbuf->vertex_map, + vbuf->vertex_ptr, + 4 * vbuf->vertex_size, + VBUF_SIZE )) + return FALSE; + + + if (vbuf->nr_elements + 4 > IBUF_SIZE / sizeof(ushort) ) + return FALSE; + + return TRUE; +} + + +static void emit_vertex( struct vbuf_stage *vbuf, + struct vertex_header *vertex ) +{ +// fprintf(stderr, "emit vertex %d to %p\n", +// vbuf->nr_vertices, vbuf->vertex_ptr); + + vertex->vertex_id = vbuf->nr_vertices++; + + //vbuf->emit_vertex( vbuf->vertex_ptr, vertex ); + memcpy(vbuf->vertex_ptr, vertex, vbuf->vertex_size); + + vbuf->vertex_ptr += vbuf->vertex_size; +} + + +static void vbuf_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + if (!check_space( vbuf )) + vbuf_flush_elements( stage ); + + for (i = 0; i < 3; i++) { + if (prim->v[i]->vertex_id == 0xffff) + emit_vertex( vbuf, prim->v[i] ); + + vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; + } +} + + +static void vbuf_line(struct draw_stage *stage, + struct prim_header *prim) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + if (!check_space( vbuf )) + vbuf_flush_elements( stage ); + + for (i = 0; i < 2; i++) { + if (prim->v[i]->vertex_id == 0xffff) + emit_vertex( vbuf, prim->v[i] ); + + vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; + } +} + + +static void vbuf_point(struct draw_stage *stage, + struct prim_header *prim) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + if (!check_space( vbuf )) + vbuf_flush_elements( stage ); + + if (prim->v[0]->vertex_id == 0xffff) + emit_vertex( vbuf, prim->v[0] ); + + vbuf->element_map[vbuf->nr_elements++] = prim->v[0]->vertex_id; +} + + +static void vbuf_first_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_elements( stage ); + stage->tri = vbuf_tri; + stage->tri( stage, prim ); + vbuf->prim = PIPE_PRIM_TRIANGLES; +} + + +static void vbuf_first_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_elements( stage ); + stage->line = vbuf_line; + stage->line( stage, prim ); + vbuf->prim = PIPE_PRIM_LINES; +} + + +static void vbuf_first_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_elements( stage ); + stage->point = vbuf_point; + stage->point( stage, prim ); + vbuf->prim = PIPE_PRIM_POINTS; +} + + +static void vbuf_draw( struct pipe_context *pipe, + unsigned prim, + const ushort *elements, + unsigned nr_elements, + const void *vertex_buffer, + unsigned nr_vertices ) +{ + /* FIXME: */ +} + + +static void vbuf_flush_elements( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + if (vbuf->nr_elements) { + fprintf(stderr, "%s (%d elts)\n", __FUNCTION__, vbuf->nr_elements); + + /* Draw now or add to list of primitives??? + */ + vbuf_draw( &vbuf->i915->pipe, + vbuf->prim, + vbuf->element_map, + vbuf->nr_elements, + vbuf->vertex_map, + (vbuf->vertex_ptr - vbuf->vertex_map) / vbuf->vertex_size ); + + vbuf->nr_elements = 0; + + vbuf->vertex_ptr = vbuf->vertex_map; + vbuf->nr_vertices = 0; + + /* Reset vertex ids? Actually, want to not do that unless our + * vertex buffer is full. Would like separate + * flush-on-index-full and flush-on-vb-full, but may raise + * issues uploading vertices if the hardware wants to flush when + * we flush. + */ + draw_vertex_cache_reset_vertex_ids( vbuf->i915->draw ); + } + + stage->tri = vbuf_first_tri; + stage->line = vbuf_first_line; + stage->point = vbuf_first_point; +} + + +static void vbuf_begin( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage(stage); + + vbuf->vertex_size = vbuf->i915->draw->vertex_info.size * sizeof(float); +} + + +static void vbuf_end( struct draw_stage *stage ) +{ + /* Overkill. + */ + vbuf_flush_elements( stage ); +} + + +static void reset_stipple_counter( struct draw_stage *stage ) +{ + /* XXX: This doesn't work. + */ +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) +{ + struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); + + vbuf->i915 = i915; + vbuf->stage.draw = i915->draw; + vbuf->stage.begin = vbuf_begin; + vbuf->stage.point = vbuf_first_point; + vbuf->stage.line = vbuf_first_line; + vbuf->stage.tri = vbuf_first_tri; + vbuf->stage.end = vbuf_end; + vbuf->stage.reset_stipple_counter = reset_stipple_counter; + + vbuf->element_map = malloc( IBUF_SIZE ); + vbuf->vertex_map = malloc( VBUF_SIZE ); + + vbuf->vertex_ptr = vbuf->vertex_map; + + return &vbuf->stage; +} -- cgit v1.2.3 From 02091e0f9c041250ea5abea6a5aa739d8e19c852 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 29 Oct 2007 10:25:02 +0000 Subject: Reserve the accurate number of dwords in the batch buffer. --- src/mesa/pipe/i915simple/i915_prim_emit.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index fcf3c99cac..40e33e6e09 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -134,7 +134,7 @@ emit_prim( struct draw_stage *stage, if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); - ptr = BEGIN_BATCH( nr * vertex_size, 0 ); + ptr = BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 ); if (ptr == 0) { FLUSH_BATCH(); @@ -143,7 +143,7 @@ emit_prim( struct draw_stage *stage, i915_update_derived( i915 ); i915_emit_hardware_state( i915 ); - ptr = BEGIN_BATCH( nr * vertex_size, 0 ); + ptr = BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 ); if (ptr == 0) { assert(0); return; @@ -157,10 +157,8 @@ emit_prim( struct draw_stage *stage, hwprim | ((4 + vertex_size * nr)/4 - 2)); - for (i = 0; i < nr; i++) { + for (i = 0; i < nr; i++) emit_hw_vertex(i915, prim->v[i]); - ptr += vertex_size / sizeof(int); - } } -- cgit v1.2.3 From 46aeff5814ae16544874ceafa5bd1e9d6577ca9d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 29 Oct 2007 11:49:59 +0000 Subject: Get vertex buffer stage in a minimally working state. --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 160 +++++++++++++++++++++++++----- 1 file changed, 133 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index a80af00137..33e532ad02 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -48,6 +48,10 @@ #include "softpipe/sp_prim_setup.h" #include "i915_context.h" +#include "i915_reg.h" +#include "i915_winsys.h" +#include "i915_batch.h" +#include "i915_state.h" static void vbuf_flush_elements( struct draw_stage *stage ); @@ -63,14 +67,14 @@ static void vbuf_flush_elements( struct draw_stage *stage ); struct vbuf_stage { struct draw_stage stage; /**< This must be first (base class) */ - vbuf_draw_func draw; - + /* FIXME: we have no guarantee that 'unsigned' is 32bit */ + /* Vertices are passed in as an array of floats making up each * attribute in turn. Will eventually convert to hardware format * in this stage. */ - char *vertex_map; - char *vertex_ptr; + unsigned *vertex_map; + unsigned *vertex_ptr; unsigned vertex_size; unsigned nr_vertices; @@ -94,7 +98,8 @@ static INLINE struct vbuf_stage *vbuf_stage( struct draw_stage *stage ) } -static boolean overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) +static inline +boolean overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) { unsigned long used = (char *)ptr - (char *)map; return (used + bytes) > bufsz; @@ -109,7 +114,6 @@ static boolean check_space( struct vbuf_stage *vbuf ) VBUF_SIZE )) return FALSE; - if (vbuf->nr_elements + 4 > IBUF_SIZE / sizeof(ushort) ) return FALSE; @@ -117,18 +121,66 @@ static boolean check_space( struct vbuf_stage *vbuf ) } -static void emit_vertex( struct vbuf_stage *vbuf, - struct vertex_header *vertex ) +/** + * Extract the needed fields from vertex_header and emit i915 dwords. + * Recall that the vertices are constructed by the 'draw' module and + * have a couple of slots at the beginning (1-dword header, 4-dword + * clip pos) that we ignore here. + */ +static inline void +emit_vertex( struct vbuf_stage *vbuf, + struct vertex_header *vertex ) { + struct i915_context *i915 = vbuf->i915; + const struct vertex_info *vinfo = &i915->current.vertex_info; + uint i; + uint count = 0; /* for debug/sanity */ + // fprintf(stderr, "emit vertex %d to %p\n", // vbuf->nr_vertices, vbuf->vertex_ptr); - vertex->vertex_id = vbuf->nr_vertices++; - - //vbuf->emit_vertex( vbuf->vertex_ptr, vertex ); - memcpy(vbuf->vertex_ptr, vertex, vbuf->vertex_size); - - vbuf->vertex_ptr += vbuf->vertex_size; + /* TODO: reuse vertices */ + /* vertex->vertex_id = */ vbuf->nr_vertices++; + + for (i = 0; i < vinfo->num_attribs; i++) { + switch (vinfo->format[i]) { + case FORMAT_OMIT: + /* no-op */ + break; + case FORMAT_1F: + *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); + count++; + break; + case FORMAT_2F: + *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); + count += 2; + break; + case FORMAT_3F: + *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][2]); + count += 3; + break; + case FORMAT_4F: + *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][2]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][3]); + count += 4; + break; + case FORMAT_4UB: + *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[i][2] ), + float_to_ubyte( vertex->data[i][1] ), + float_to_ubyte( vertex->data[i][0] ), + float_to_ubyte( vertex->data[i][3] )); + count += 1; + break; + default: + assert(0); + } + } + assert(count == vinfo->size); } @@ -219,14 +271,66 @@ static void vbuf_first_point( struct draw_stage *stage, } -static void vbuf_draw( struct pipe_context *pipe, +static void vbuf_draw( struct draw_stage *stage, unsigned prim, const ushort *elements, unsigned nr_elements, const void *vertex_buffer, - unsigned nr_vertices ) + unsigned nr ) { - /* FIXME: */ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + struct i915_context *i915 = vbuf->i915; + unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ + unsigned hwprim; + unsigned *ptr; + unsigned i; + + switch(prim) { + case PIPE_PRIM_POINTS: + hwprim = PRIM3D_POINTLIST; + break; + case PIPE_PRIM_LINES: + hwprim = PRIM3D_LINELIST; + break; + case PIPE_PRIM_TRIANGLES: + hwprim = PRIM3D_TRILIST; + break; + default: + assert(0); + return; + } + + if (i915->dirty) + i915_update_derived( i915 ); + + if (i915->hardware_dirty) + i915_emit_hardware_state( i915 ); + + assert(vbuf->vertex_ptr - vbuf->vertex_map == nr * vertex_size / 4); + + ptr = BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 ); + if (ptr == 0) { + FLUSH_BATCH(); + + /* Make sure state is re-emitted after a flush: + */ + i915_update_derived( i915 ); + i915_emit_hardware_state( i915 ); + + ptr = BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 ); + if (ptr == 0) { + assert(0); + return; + } + } + + /* TODO: Fire a DMA buffer */ + OUT_BATCH(_3DPRIMITIVE | + hwprim | + ((4 + vertex_size * nr)/4 - 2)); + + for (i = 0; i < nr * vertex_size / 4; i++) + OUT_BATCH(vbuf->vertex_map[i]); } @@ -235,16 +339,19 @@ static void vbuf_flush_elements( struct draw_stage *stage ) struct vbuf_stage *vbuf = vbuf_stage( stage ); if (vbuf->nr_elements) { - fprintf(stderr, "%s (%d elts)\n", __FUNCTION__, vbuf->nr_elements); + fprintf(stderr, "%s (%d elts, %d verts)\n", + __FUNCTION__, + vbuf->nr_elements, + vbuf->nr_vertices); /* Draw now or add to list of primitives??? */ - vbuf_draw( &vbuf->i915->pipe, - vbuf->prim, - vbuf->element_map, - vbuf->nr_elements, - vbuf->vertex_map, - (vbuf->vertex_ptr - vbuf->vertex_map) / vbuf->vertex_size ); + vbuf_draw( stage, + vbuf->prim, + vbuf->element_map, + vbuf->nr_elements, + vbuf->vertex_map, + vbuf->nr_vertices ); vbuf->nr_elements = 0; @@ -270,7 +377,7 @@ static void vbuf_begin( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage(stage); - vbuf->vertex_size = vbuf->i915->draw->vertex_info.size * sizeof(float); + vbuf->vertex_size = vbuf->i915->current.vertex_info.size * 4; } @@ -284,8 +391,6 @@ static void vbuf_end( struct draw_stage *stage ) static void reset_stipple_counter( struct draw_stage *stage ) { - /* XXX: This doesn't work. - */ } @@ -305,6 +410,7 @@ struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) vbuf->stage.end = vbuf_end; vbuf->stage.reset_stipple_counter = reset_stipple_counter; + /* FIXME: free this memory on takedown */ vbuf->element_map = malloc( IBUF_SIZE ); vbuf->vertex_map = malloc( VBUF_SIZE ); -- cgit v1.2.3 From a9e1fcf98a871b182f82dc37c15d0f69bf1a3187 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 29 Oct 2007 11:51:04 +0000 Subject: Enable the vertex buffer stage according to the I915_VBUF environment var. --- src/mesa/pipe/i915simple/i915_context.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 6541f0e848..fc878c175c 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -320,7 +320,12 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, */ i915->draw = draw_create(); assert(i915->draw); - draw_set_rasterize_stage(i915->draw, i915_draw_render_stage(i915)); + if (getenv("I915_VBUF")) { + draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915)); + } + else { + draw_set_rasterize_stage(i915->draw, i915_draw_render_stage(i915)); + } i915_init_region_functions(i915); i915_init_surface_functions(i915); -- cgit v1.2.3 From 1039a755142e2fd45dc291d891c514fdfa7e033c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 29 Oct 2007 12:14:52 +0000 Subject: Reuse hardware vertice representation. --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 31 ++++++++++++------------------- 1 file changed, 12 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 33e532ad02..0adf987721 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -139,8 +139,7 @@ emit_vertex( struct vbuf_stage *vbuf, // fprintf(stderr, "emit vertex %d to %p\n", // vbuf->nr_vertices, vbuf->vertex_ptr); - /* TODO: reuse vertices */ - /* vertex->vertex_id = */ vbuf->nr_vertices++; + vertex->vertex_id = vbuf->nr_vertices++; for (i = 0; i < vinfo->num_attribs; i++) { switch (vinfo->format[i]) { @@ -271,21 +270,17 @@ static void vbuf_first_point( struct draw_stage *stage, } -static void vbuf_draw( struct draw_stage *stage, - unsigned prim, - const ushort *elements, - unsigned nr_elements, - const void *vertex_buffer, - unsigned nr ) +static void vbuf_draw( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); struct i915_context *i915 = vbuf->i915; + unsigned nr = vbuf->nr_elements; unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ unsigned hwprim; unsigned *ptr; - unsigned i; + unsigned i, j; - switch(prim) { + switch(vbuf->prim) { case PIPE_PRIM_POINTS: hwprim = PRIM3D_POINTLIST; break; @@ -306,7 +301,7 @@ static void vbuf_draw( struct draw_stage *stage, if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); - assert(vbuf->vertex_ptr - vbuf->vertex_map == nr * vertex_size / 4); + assert(vbuf->vertex_ptr - vbuf->vertex_map == vbuf->nr_vertices * vertex_size / 4); ptr = BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 ); if (ptr == 0) { @@ -329,8 +324,9 @@ static void vbuf_draw( struct draw_stage *stage, hwprim | ((4 + vertex_size * nr)/4 - 2)); - for (i = 0; i < nr * vertex_size / 4; i++) - OUT_BATCH(vbuf->vertex_map[i]); + for (i = 0; i < nr; i++) + for (j = 0; j < vertex_size / 4; j++) + OUT_BATCH(vbuf->vertex_map[vbuf->element_map[i]*vertex_size/4 + j]); } @@ -339,19 +335,16 @@ static void vbuf_flush_elements( struct draw_stage *stage ) struct vbuf_stage *vbuf = vbuf_stage( stage ); if (vbuf->nr_elements) { +#if 0 fprintf(stderr, "%s (%d elts, %d verts)\n", __FUNCTION__, vbuf->nr_elements, vbuf->nr_vertices); +#endif /* Draw now or add to list of primitives??? */ - vbuf_draw( stage, - vbuf->prim, - vbuf->element_map, - vbuf->nr_elements, - vbuf->vertex_map, - vbuf->nr_vertices ); + vbuf_draw( stage ); vbuf->nr_elements = 0; -- cgit v1.2.3 From 242b8659e40416f893157c7a0919964dabc957cf Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 29 Oct 2007 12:29:06 +0000 Subject: Fix i915simple build. --- src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c | 6 +++--- src/mesa/pipe/softpipe/sp_winsys.h | 3 +++ 2 files changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c index 917e933641..1660046f13 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c @@ -46,8 +46,8 @@ struct intel_softpipe_winsys { * Return list of surface formats supported by this driver. */ static const unsigned * -intel_supported_formats(struct softpipe_winsys *sws, - unsigned *numFormats) +intel_is_format_supported(struct softpipe_winsys *sws, + unsigned *numFormats) { static const GLuint formats[] = { PIPE_FORMAT_U_A8_R8_G8_B8, @@ -69,7 +69,7 @@ intel_create_softpipe( struct intel_context *intel ) /* Fill in this struct with callbacks that softpipe will need to * communicate with the window system, buffer manager, etc. */ - isws->sws.supported_formats = intel_supported_formats; + isws->sws.is_format_supported = intel_is_format_supported; isws->intel = intel; /* Create the softpipe context: diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h index d8ae971188..1912e59b9f 100644 --- a/src/mesa/pipe/softpipe/sp_winsys.h +++ b/src/mesa/pipe/softpipe/sp_winsys.h @@ -29,6 +29,9 @@ #define SP_WINSYS_H +#include "pipe/p_compiler.h" // for boolean + + /* This is the interface that softpipe requires any window system * hosting it to implement. This is the only include file in softpipe * which is public. -- cgit v1.2.3 From e15ca7963e8b2e80eb79f6352f0761d0c1581fb5 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 29 Oct 2007 11:40:09 +0000 Subject: Remove TGSI_INTERPOLATE_ATTRIB. --- src/mesa/pipe/tgsi/exec/tgsi_build.c | 2 +- src/mesa/pipe/tgsi/exec/tgsi_token.h | 1 - src/mesa/state_tracker/st_mesa_to_tgsi.c | 18 ++++++++++-------- 3 files changed, 11 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.c b/src/mesa/pipe/tgsi/exec/tgsi_build.c index 56827726f5..78f648aae2 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_build.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_build.c @@ -294,7 +294,7 @@ tgsi_build_declaration_interpolation( { struct tgsi_declaration_interpolation di; - assert( interpolate <= TGSI_INTERPOLATE_ATTRIB ); + assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); di = tgsi_default_declaration_interpolation(); di.Interpolate = interpolate; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h index 80c3fcd434..8d5992facb 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -96,7 +96,6 @@ struct tgsi_declaration_mask #define TGSI_INTERPOLATE_CONSTANT 0 #define TGSI_INTERPOLATE_LINEAR 1 #define TGSI_INTERPOLATE_PERSPECTIVE 2 -#define TGSI_INTERPOLATE_ATTRIB 3 /**< Vertex shader input attrib */ struct tgsi_declaration_interpolation { diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 37ba6a90b5..75195968e5 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -519,6 +519,7 @@ compile_instruction( static struct tgsi_full_declaration make_input_decl( GLuint index, + GLboolean interpolate_info, GLuint interpolate, GLuint usage_mask, GLboolean semantic_info, @@ -534,14 +535,16 @@ make_input_decl( decl.Declaration.Declare = TGSI_DECLARE_RANGE; decl.Declaration.UsageMask = usage_mask; decl.Declaration.Semantic = semantic_info; - decl.Declaration.Interpolate = 1; decl.u.DeclarationRange.First = index; decl.u.DeclarationRange.Last = index; if (semantic_info) { decl.Semantic.SemanticName = semantic_name; decl.Semantic.SemanticIndex = semantic_index; } - decl.Interpolation.Interpolate = interpolate; + if (interpolate_info) { + decl.Declaration.Interpolate = 1; + decl.Interpolation.Interpolate = interpolate; + } return decl; } @@ -685,7 +688,7 @@ tgsi_translate_mesa_program( case TGSI_SEMANTIC_POSITION: /* Fragment XY pos */ fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_CONSTANT, + GL_TRUE, TGSI_INTERPOLATE_CONSTANT, TGSI_WRITEMASK_XY, GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); ti += tgsi_build_full_declaration( @@ -695,7 +698,7 @@ tgsi_translate_mesa_program( maxTokens - ti ); /* Fragment ZW pos */ fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_LINEAR, + GL_TRUE, TGSI_INTERPOLATE_LINEAR, TGSI_WRITEMASK_ZW, GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); ti += tgsi_build_full_declaration(&fulldecl, @@ -705,7 +708,7 @@ tgsi_translate_mesa_program( break; default: fulldecl = make_input_decl(i, - interpMode[i], + GL_TRUE, interpMode[i], TGSI_WRITEMASK_XYZW, GL_TRUE, inputSemanticName[i], inputSemanticIndex[i]); @@ -722,10 +725,9 @@ tgsi_translate_mesa_program( for (i = 0; i < numInputs; i++) { struct tgsi_full_declaration fulldecl; fulldecl = make_input_decl(i, - TGSI_INTERPOLATE_ATTRIB, + GL_FALSE, 0, TGSI_WRITEMASK_XYZW, - GL_FALSE, inputSemanticName[i], - inputSemanticIndex[i]); + GL_FALSE, 0, 0); ti += tgsi_build_full_declaration(&fulldecl, &tokens[ti], header, -- cgit v1.2.3 From bd922c6437fb2081082bb164bcb0304559f6d7c4 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 29 Oct 2007 11:45:42 +0000 Subject: Implement RET opcode. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 18 +++++------------- 1 file changed, 5 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 46905236e2..b8edcf0a2e 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -1539,7 +1539,11 @@ emit_instruction( break; case TGSI_OPCODE_RET: - assert( 0 ); +#ifdef WIN32 + x86_retw( func, 16 ); +#else + x86_ret( func ); +#endif break; case TGSI_OPCODE_SSG: @@ -1818,12 +1822,6 @@ tgsi_emit_sse2( tgsi_parse_free( &parse ); -#ifdef WIN32 - x86_retw( func, 16 ); -#else - x86_ret( func ); -#endif - return 1; } @@ -1896,11 +1894,5 @@ tgsi_emit_sse2_fs( tgsi_parse_free( &parse ); -#ifdef WIN32 - x86_retw( func, 16 ); -#else - x86_ret( func ); -#endif - return 1; } -- cgit v1.2.3 From 1eabc29ed126649f0de4a4f26166489bdcb9c01d Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 29 Oct 2007 12:14:54 +0000 Subject: Code re-org. Add comments. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 484 ++++++++++++++++++++---------------- 1 file changed, 264 insertions(+), 220 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index b8edcf0a2e..abdebd6f97 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -22,14 +22,9 @@ #define TEMP_R0 TGSI_EXEC_TEMP_R0 -static struct x86_reg -get_argument( - unsigned index ) -{ - return x86_make_disp( - x86_make_reg( file_REG32, reg_SP ), - (index + 1) * 4 ); -} +/** + * X86 utility functions. + */ static struct x86_reg make_xmm( @@ -40,6 +35,10 @@ make_xmm( (enum x86_reg_name) xmm ); } +/** + * X86 register mapping helpers. + */ + static struct x86_reg get_const_base( void ) { @@ -48,16 +47,6 @@ get_const_base( void ) reg_CX ); } -static struct x86_reg -get_const( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_const_base(), - (vec * 4 + chan) * 4 ); -} - static struct x86_reg get_input_base( void ) { @@ -67,55 +56,78 @@ get_input_base( void ) } static struct x86_reg -get_input( - unsigned vec, - unsigned chan ) +get_output_base( void ) { - return x86_make_disp( - get_input_base(), - (vec * 4 + chan) * 16 ); + return x86_make_reg( + file_REG32, + reg_DX ); } static struct x86_reg -get_output_base( void ) +get_temp_base( void ) { return x86_make_reg( file_REG32, - reg_DX ); + reg_BX ); } static struct x86_reg -get_output( +get_coef_base( void ) +{ + return get_output_base(); +} + +/** + * Data access helpers. + */ + +static struct x86_reg +get_argument( + unsigned index ) +{ + return x86_make_disp( + x86_make_reg( file_REG32, reg_SP ), + (index + 1) * 4 ); +} + +static struct x86_reg +get_const( unsigned vec, unsigned chan ) { return x86_make_disp( - get_output_base(), - (vec * 4 + chan) * 16 ); + get_const_base(), + (vec * 4 + chan) * 4 ); } static struct x86_reg -get_temp_base( void ) +get_input( + unsigned vec, + unsigned chan ) { - return x86_make_reg( - file_REG32, - reg_BX ); + return x86_make_disp( + get_input_base(), + (vec * 4 + chan) * 16 ); } static struct x86_reg -get_temp( +get_output( unsigned vec, unsigned chan ) { return x86_make_disp( - get_temp_base(), + get_output_base(), (vec * 4 + chan) * 16 ); } static struct x86_reg -get_coef_base( void ) +get_temp( + unsigned vec, + unsigned chan ) { - return get_output_base(); + return x86_make_disp( + get_temp_base(), + (vec * 4 + chan) * 16 ); } static struct x86_reg @@ -129,6 +141,10 @@ get_coef( ((vec * 3 + member) * 4 + chan) * 4 ); } +/** + * Data fetch helpers. + */ + static void emit_const( struct x86_function *func, @@ -160,19 +176,6 @@ emit_inputf( get_input( vec, chan ) ); } -static void -emit_inputs( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movups( - func, - get_input( vec, chan ), - make_xmm( xmm ) ); -} - static void emit_output( struct x86_function *func, @@ -199,19 +202,6 @@ emit_tempf( get_temp( vec, chan ) ); } -static void -emit_temps( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movaps( - func, - get_temp( vec, chan ), - make_xmm( xmm ) ); -} - static void emit_coef( struct x86_function *func, @@ -231,49 +221,34 @@ emit_coef( SHUF( 0, 0, 0, 0 ) ); } -static void -emit_coef_a0( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 0 ); -} +/** + * Data store helpers. + */ static void -emit_coef_dadx( +emit_inputs( struct x86_function *func, unsigned xmm, unsigned vec, unsigned chan ) { - emit_coef( + sse_movups( func, - xmm, - vec, - chan, - 1 ); + get_input( vec, chan ), + make_xmm( xmm ) ); } static void -emit_coef_dady( +emit_temps( struct x86_function *func, unsigned xmm, unsigned vec, unsigned chan ) { - emit_coef( + sse_movaps( func, - xmm, - vec, - chan, - 2 ); + get_temp( vec, chan ), + make_xmm( xmm ) ); } static void @@ -290,57 +265,59 @@ emit_addrs( chan ); } -static void -emit_abs( - struct x86_function *func, - unsigned xmm ) -{ - sse_andps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_7FFFFFFF_I, - TGSI_EXEC_TEMP_7FFFFFFF_C ) ); -} +/** + * Coefficent fetch helpers. + */ static void -emit_neg( +emit_coef_a0( struct x86_function *func, - unsigned xmm ) + unsigned xmm, + unsigned vec, + unsigned chan ) { - sse_xorps( + emit_coef( func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C ) ); + xmm, + vec, + chan, + 0 ); } static void -emit_setsign( +emit_coef_dadx( struct x86_function *func, - unsigned xmm ) + unsigned xmm, + unsigned vec, + unsigned chan ) { - sse_orps( + emit_coef( func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C ) ); + xmm, + vec, + chan, + 1 ); } static void -emit_add( +emit_coef_dady( struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) + unsigned xmm, + unsigned vec, + unsigned chan ) { - sse_addps( + emit_coef( func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); + xmm, + vec, + chan, + 2 ); } +/** + * Function call helpers. + */ + static void emit_push_gp( struct x86_function *func ) @@ -433,6 +410,35 @@ emit_func_call_dst_src( code ); } +/** + * Low-level instruction translators. + */ + +static void +emit_abs( + struct x86_function *func, + unsigned xmm ) +{ + sse_andps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_7FFFFFFF_I, + TGSI_EXEC_TEMP_7FFFFFFF_C ) ); +} + +static void +emit_add( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + sse_addps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + static void XSTDCALL cos4f( float *store ) @@ -463,114 +469,95 @@ emit_cos( } static void XSTDCALL -sin4f( +ex24f( float *store ) { #ifdef WIN32 - store[0] = (float) sin( (double) store[0] ); - store[1] = (float) sin( (double) store[1] ); - store[2] = (float) sin( (double) store[2] ); - store[3] = (float) sin( (double) store[3] ); + store[0] = (float) pow( 2.0, (double) store[0] ); + store[1] = (float) pow( 2.0, (double) store[1] ); + store[2] = (float) pow( 2.0, (double) store[2] ); + store[3] = (float) pow( 2.0, (double) store[3] ); #else const unsigned X = TEMP_R0 * 16; - store[X + 0] = sinf( store[X + 0] ); - store[X + 1] = sinf( store[X + 1] ); - store[X + 2] = sinf( store[X + 2] ); - store[X + 3] = sinf( store[X + 3] ); + store[X + 0] = powf( 2.0f, store[X + 0] ); + store[X + 1] = powf( 2.0f, store[X + 1] ); + store[X + 2] = powf( 2.0f, store[X + 2] ); + store[X + 3] = powf( 2.0f, store[X + 3] ); #endif } static void -emit_sin (struct x86_function *func, - unsigned xmm_dst) +emit_ex2( + struct x86_function *func, + unsigned xmm_dst ) { emit_func_call_dst( func, xmm_dst, - sin4f ); + ex24f ); } static void -emit_mov( +emit_f2it( struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - sse_movups( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_mul (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) + unsigned xmm ) { - sse_mulps( + sse2_cvttps2dq( func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); + make_xmm( xmm ), + make_xmm( xmm ) ); } static void XSTDCALL -pow4f( +flr4f( float *store ) { #ifdef WIN32 - store[0] = (float) pow( (double) store[0], (double) store[4] ); - store[1] = (float) pow( (double) store[1], (double) store[5] ); - store[2] = (float) pow( (double) store[2], (double) store[6] ); - store[3] = (float) pow( (double) store[3], (double) store[7] ); + const unsigned X = 0; #else const unsigned X = TEMP_R0 * 16; - store[X + 0] = powf( store[X + 0], store[X + 4] ); - store[X + 1] = powf( store[X + 1], store[X + 5] ); - store[X + 2] = powf( store[X + 2], store[X + 6] ); - store[X + 3] = powf( store[X + 3], store[X + 7] ); #endif + store[X + 0] = (float) floor( (double) store[X + 0] ); + store[X + 1] = (float) floor( (double) store[X + 1] ); + store[X + 2] = (float) floor( (double) store[X + 2] ); + store[X + 3] = (float) floor( (double) store[X + 3] ); } static void -emit_pow( +emit_flr( struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) + unsigned xmm_dst ) { - emit_func_call_dst_src( + emit_func_call_dst( func, xmm_dst, - xmm_src, - pow4f ); + flr4f ); } static void XSTDCALL -ex24f( +frc4f( float *store ) { #ifdef WIN32 - store[0] = (float) pow( 2.0, (double) store[0] ); - store[1] = (float) pow( 2.0, (double) store[1] ); - store[2] = (float) pow( 2.0, (double) store[2] ); - store[3] = (float) pow( 2.0, (double) store[3] ); + const unsigned X = 0; #else const unsigned X = TEMP_R0 * 16; - store[X + 0] = powf( 2.0f, store[X + 0] ); - store[X + 1] = powf( 2.0f, store[X + 1] ); - store[X + 2] = powf( 2.0f, store[X + 2] ); - store[X + 3] = powf( 2.0f, store[X + 3] ); #endif + store[X + 0] -= (float) floor( (double) store[X + 0] ); + store[X + 1] -= (float) floor( (double) store[X + 1] ); + store[X + 2] -= (float) floor( (double) store[X + 2] ); + store[X + 3] -= (float) floor( (double) store[X + 3] ); } static void -emit_ex2( +emit_frc( struct x86_function *func, unsigned xmm_dst ) { emit_func_call_dst( func, xmm_dst, - ex24f ); + frc4f ); } static void XSTDCALL @@ -599,56 +586,71 @@ emit_lg2( lg24f ); } -static void XSTDCALL -flr4f( - float *store ) +static void +emit_mov( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) { -#ifdef WIN32 - const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif - store[X + 0] = (float) floor( (double) store[X + 0] ); - store[X + 1] = (float) floor( (double) store[X + 1] ); - store[X + 2] = (float) floor( (double) store[X + 2] ); - store[X + 3] = (float) floor( (double) store[X + 3] ); + sse_movups( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); } static void -emit_flr( +emit_mul (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + sse_mulps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_neg( struct x86_function *func, - unsigned xmm_dst ) + unsigned xmm ) { - emit_func_call_dst( + sse_xorps( func, - xmm_dst, - flr4f ); + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); } static void XSTDCALL -frc4f( +pow4f( float *store ) { #ifdef WIN32 - const unsigned X = 0; + store[0] = (float) pow( (double) store[0], (double) store[4] ); + store[1] = (float) pow( (double) store[1], (double) store[5] ); + store[2] = (float) pow( (double) store[2], (double) store[6] ); + store[3] = (float) pow( (double) store[3], (double) store[7] ); #else const unsigned X = TEMP_R0 * 16; + store[X + 0] = powf( store[X + 0], store[X + 4] ); + store[X + 1] = powf( store[X + 1], store[X + 5] ); + store[X + 2] = powf( store[X + 2], store[X + 6] ); + store[X + 3] = powf( store[X + 3], store[X + 7] ); #endif - store[X + 0] -= (float) floor( (double) store[X + 0] ); - store[X + 1] -= (float) floor( (double) store[X + 1] ); - store[X + 2] -= (float) floor( (double) store[X + 2] ); - store[X + 3] -= (float) floor( (double) store[X + 3] ); } static void -emit_frc( +emit_pow( struct x86_function *func, - unsigned xmm_dst ) + unsigned xmm_dst, + unsigned xmm_src ) { - emit_func_call_dst( + emit_func_call_dst_src( func, xmm_dst, - frc4f ); + xmm_src, + pow4f ); } static void @@ -675,6 +677,47 @@ emit_rsqrt( make_xmm( xmm_src ) ); } +static void +emit_setsign( + struct x86_function *func, + unsigned xmm ) +{ + sse_orps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); +} + +static void XSTDCALL +sin4f( + float *store ) +{ +#ifdef WIN32 + store[0] = (float) sin( (double) store[0] ); + store[1] = (float) sin( (double) store[1] ); + store[2] = (float) sin( (double) store[2] ); + store[3] = (float) sin( (double) store[3] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = sinf( store[X + 0] ); + store[X + 1] = sinf( store[X + 1] ); + store[X + 2] = sinf( store[X + 2] ); + store[X + 3] = sinf( store[X + 3] ); +#endif +} + +static void +emit_sin (struct x86_function *func, + unsigned xmm_dst) +{ + emit_func_call_dst( + func, + xmm_dst, + sin4f ); +} + static void emit_sub( struct x86_function *func, @@ -687,6 +730,10 @@ emit_sub( make_xmm( xmm_src ) ); } +/** + * Register fetch. + */ + static void emit_fetch( struct x86_function *func, @@ -769,6 +816,13 @@ emit_fetch( } } +#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\ + emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN ) + +/** + * Register store. + */ + static void emit_store( struct x86_function *func, @@ -820,6 +874,13 @@ emit_store( } } +#define STORE( FUNC, INST, XMM, INDEX, CHAN )\ + emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) + +/** + * High-level instruction translators. + */ + static void emit_kil( struct x86_function *func, @@ -915,12 +976,6 @@ emit_kil( x86_make_reg( file_REG32, reg_AX ) ); } -#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\ - emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN ) - -#define STORE( FUNC, INST, XMM, INDEX, CHAN )\ - emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) - static void emit_setcc( struct x86_function *func, @@ -981,17 +1036,6 @@ emit_cmp( } } -static void -emit_f2it( - struct x86_function *func, - unsigned xmm ) -{ - sse2_cvttps2dq( - func, - make_xmm( xmm ), - make_xmm( xmm ) ); -} - static void emit_instruction( struct x86_function *func, -- cgit v1.2.3 From 3b25ce9eca9643a7a008735ec69c4e7ffddbe742 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 29 Oct 2007 13:21:17 +0000 Subject: Add SSE dump facilities. Wrap x86_, sse_ and sse2 rtasm calls in emit_ calls. Those emit_ calls, if required, dump instructions to stdout. SSE dumping disabled by default. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 498 +++++++++++++++++++++++++++++++----- 1 file changed, 428 insertions(+), 70 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index abdebd6f97..a71c5155c0 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -2,6 +2,104 @@ #include "tgsi_core.h" #include "x86/rtasm/x86sse.h" +#define DUMP_SSE 0 + +#if DUMP_SSE + +static void +_print_reg( + struct x86_reg reg ) +{ + switch( reg.file ) { + case file_REG32: + switch( reg.idx ) { + case reg_AX: + printf( "EAX" ); + break; + case reg_CX: + printf( "ECX" ); + break; + case reg_DX: + printf( "EDX" ); + break; + case reg_BX: + printf( "EBX" ); + break; + case reg_SP: + printf( "ESP" ); + break; + case reg_BP: + printf( "EBP" ); + break; + case reg_SI: + printf( "ESI" ); + break; + case reg_DI: + printf( "EDI" ); + break; + } + break; + case file_MMX: + assert( 0 ); + break; + case file_XMM: + printf( "XMM%u", reg.idx ); + break; + case file_x87: + assert( 0 ); + break; + } +} + +static void +_fill( + const char *op ) +{ + unsigned count = 10 - strlen( op ); + + while( count-- ) { + printf( " " ); + } +} + +#define DUMP_START() printf( "\nsse-dump start ----------------" ) +#define DUMP_END() printf( "\nsse-dump end ----------------\n" ) +#define DUMP( OP ) printf( "\n%s", OP ) +#define DUMP_I( OP, I ) do {\ + printf( "\n%s", OP );\ + _fill( OP );\ + printf( "%u", I ); } while( 0 ) +#define DUMP_R( OP, R0 ) do {\ + printf( "\n%s", OP );\ + _fill( OP );\ + _print_reg( R0 ); } while( 0 ) +#define DUMP_RR( OP, R0, R1 ) do {\ + printf( "\n%s", OP );\ + _fill( OP );\ + _print_reg( R0 );\ + printf( ", " );\ + _print_reg( R1 ); } while( 0 ) +#define DUMP_RRI( OP, R0, R1, I ) do {\ + printf( "\n%s", OP );\ + _fill( OP );\ + _print_reg( R0 );\ + printf( ", " );\ + _print_reg( R1 );\ + printf( ", " );\ + printf( "%u", I ); } while( 0 ) + +#else + +#define DUMP_START() +#define DUMP_END() +#define DUMP( OP ) +#define DUMP_I( OP, I ) +#define DUMP_R( OP, R0 ) +#define DUMP_RR( OP, R0, R1 ) +#define DUMP_RRI( OP, R0, R1, I ) + +#endif + #define FOR_EACH_CHANNEL( CHAN )\ for( CHAN = 0; CHAN < 4; CHAN++ ) @@ -141,6 +239,258 @@ get_coef( ((vec * 3 + member) * 4 + chan) * 4 ); } +/** + * X86 rtasm wrappers. + */ + +static void +emit_addps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ADDPS", dst, src ); + sse_addps( func, dst, src ); +} + +static void +emit_andnps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ANDNPS", dst, src ); + sse_andnps( func, dst, src ); +} + +static void +emit_andps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ANDPS", dst, src ); + sse_andps( func, dst, src ); +} + +static void +emit_call( + struct x86_function *func, + void (* addr)() ) +{ + DUMP_I( "CALL", addr ); + x86_call( func, addr ); +} + +static void +emit_cmpps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src, + enum sse_cc cc ) +{ + DUMP_RRI( "CMPPS", dst, src, cc ); + sse_cmpps( func, dst, src, cc ); +} + +static void +emit_cvttps2dq( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "CVTTPS2DQ", dst, src ); + sse2_cvttps2dq( func, dst, src ); +} + +static void +emit_maxps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MAXPS", dst, src ); + sse_maxps( func, dst, src ); +} + +static void +emit_minps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MINPS", dst, src ); + sse_minps( func, dst, src ); +} + +static void +emit_mov( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOV", dst, src ); + x86_mov( func, dst, src ); +} + +static void +emit_movaps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOVAPS", dst, src ); + sse_movaps( func, dst, src ); +} + +static void +emit_movss( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOVSS", dst, src ); + sse_movss( func, dst, src ); +} + +static void +emit_movups( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOVUPS", dst, src ); + sse_movups( func, dst, src ); +} + +static void +emit_mulps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MULPS", dst, src ); + sse_mulps( func, dst, src ); +} + +static void +emit_or( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "OR", dst, src ); + x86_or( func, dst, src ); +} + +static void +emit_orps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ORPS", dst, src ); + sse_orps( func, dst, src ); +} + +static void +emit_pmovmskb( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "PMOVMSKB", dst, src ); + sse_pmovmskb( func, dst, src ); +} + +static void +emit_pop( + struct x86_function *func, + struct x86_reg dst ) +{ + DUMP_R( "POP", dst ); + x86_pop( func, dst ); +} + +static void +emit_push( + struct x86_function *func, + struct x86_reg dst ) +{ + DUMP_R( "PUSH", dst ); + x86_push( func, dst ); +} + +static void +emit_rcpps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "RCPPS", dst, src ); + sse2_rcpps( func, dst, src ); +} + +#ifdef WIN32 +static void +emit_retw( + struct x86_function *func, + unsigned size ) +{ + DUMP_I( "RET", size ); + x86_retw( func, size ); +} +#else +static void +emit_ret( + struct x86_function *func ) +{ + DUMP( "RET" ); + x86_ret( func ); +} +#endif + +static void +emit_rsqrtps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "RSQRTPS", dst, src ); + sse_rsqrtps( func, dst, src ); +} + +static void +emit_shufps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src, + unsigned shuf ) +{ + DUMP_RRI( "SHUFPS", dst, src, shuf ); + sse_shufps( func, dst, src, shuf ); +} + +static void +emit_subps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "SUBPS", dst, src ); + sse_subps( func, dst, src ); +} + +static void +emit_xorps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "XORPS", dst, src ); + sse_xorps( func, dst, src ); +} + /** * Data fetch helpers. */ @@ -152,11 +502,11 @@ emit_const( unsigned vec, unsigned chan ) { - sse_movss( + emit_movss( func, make_xmm( xmm ), get_const( vec, chan ) ); - sse_shufps( + emit_shufps( func, make_xmm( xmm ), make_xmm( xmm ), @@ -170,7 +520,7 @@ emit_inputf( unsigned vec, unsigned chan ) { - sse_movups( + emit_movups( func, make_xmm( xmm ), get_input( vec, chan ) ); @@ -183,7 +533,7 @@ emit_output( unsigned vec, unsigned chan ) { - sse_movups( + emit_movups( func, get_output( vec, chan ), make_xmm( xmm ) ); @@ -196,7 +546,7 @@ emit_tempf( unsigned vec, unsigned chan ) { - sse_movaps( + emit_movaps( func, make_xmm( xmm ), get_temp( vec, chan ) ); @@ -210,11 +560,11 @@ emit_coef( unsigned chan, unsigned member ) { - sse_movss( + emit_movss( func, make_xmm( xmm ), get_coef( vec, chan, member ) ); - sse_shufps( + emit_shufps( func, make_xmm( xmm ), make_xmm( xmm ), @@ -232,7 +582,7 @@ emit_inputs( unsigned vec, unsigned chan ) { - sse_movups( + emit_movups( func, get_input( vec, chan ), make_xmm( xmm ) ); @@ -245,7 +595,7 @@ emit_temps( unsigned vec, unsigned chan ) { - sse_movaps( + emit_movaps( func, get_temp( vec, chan ), make_xmm( xmm ) ); @@ -322,19 +672,19 @@ static void emit_push_gp( struct x86_function *func ) { - x86_push( + emit_push( func, get_const_base() ); - x86_push( + emit_push( func, get_input_base() ); - x86_push( + emit_push( func, get_output_base() ); /* It is important on non-win32 platforms that temp base is pushed last. */ - x86_push( + emit_push( func, get_temp_base() ); } @@ -345,16 +695,16 @@ emit_pop_gp( { /* Restore GP registers in a reverse order. */ - x86_pop( + emit_pop( func, get_temp_base() ); - x86_pop( + emit_pop( func, get_output_base() ); - x86_pop( + emit_pop( func, get_input_base() ); - x86_pop( + emit_pop( func, get_const_base() ); } @@ -365,7 +715,7 @@ emit_func_call_dst( unsigned xmm_dst, void (*code)() ) { - sse_movaps( + emit_movaps( func, get_temp( TEMP_R0, 0 ), make_xmm( xmm_dst ) ); @@ -374,19 +724,19 @@ emit_func_call_dst( func ); #ifdef WIN32 - x86_push( + emit_push( func, get_temp( TEMP_R0, 0 ) ); #endif - x86_call( + emit_call( func, code ); emit_pop_gp( func ); - sse_movaps( + emit_movaps( func, make_xmm( xmm_dst ), get_temp( TEMP_R0, 0 ) ); @@ -399,7 +749,7 @@ emit_func_call_dst_src( unsigned xmm_src, void (*code)() ) { - sse_movaps( + emit_movaps( func, get_temp( TEMP_R0, 1 ), make_xmm( xmm_src ) ); @@ -419,7 +769,7 @@ emit_abs( struct x86_function *func, unsigned xmm ) { - sse_andps( + emit_andps( func, make_xmm( xmm ), get_temp( @@ -433,7 +783,7 @@ emit_add( unsigned xmm_dst, unsigned xmm_src ) { - sse_addps( + emit_addps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -502,7 +852,7 @@ emit_f2it( struct x86_function *func, unsigned xmm ) { - sse2_cvttps2dq( + emit_cvttps2dq( func, make_xmm( xmm ), make_xmm( xmm ) ); @@ -587,12 +937,12 @@ emit_lg2( } static void -emit_mov( +emit_MOV( struct x86_function *func, unsigned xmm_dst, unsigned xmm_src ) { - sse_movups( + emit_movups( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -603,7 +953,7 @@ emit_mul (struct x86_function *func, unsigned xmm_dst, unsigned xmm_src) { - sse_mulps( + emit_mulps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -614,7 +964,7 @@ emit_neg( struct x86_function *func, unsigned xmm ) { - sse_xorps( + emit_xorps( func, make_xmm( xmm ), get_temp( @@ -659,7 +1009,7 @@ emit_rcp ( unsigned xmm_dst, unsigned xmm_src ) { - sse2_rcpps( + emit_rcpps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -671,7 +1021,7 @@ emit_rsqrt( unsigned xmm_dst, unsigned xmm_src ) { - sse_rsqrtps( + emit_rsqrtps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -682,7 +1032,7 @@ emit_setsign( struct x86_function *func, unsigned xmm ) { - sse_orps( + emit_orps( func, make_xmm( xmm ), get_temp( @@ -724,7 +1074,7 @@ emit_sub( unsigned xmm_dst, unsigned xmm_src ) { - sse_subps( + emit_subps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -925,16 +1275,16 @@ emit_kil( } } - x86_push( + emit_push( func, x86_make_reg( file_REG32, reg_AX ) ); - x86_push( + emit_push( func, x86_make_reg( file_REG32, reg_DX ) ); FOR_EACH_CHANNEL( chan_index ) { if( uniquemask & (1 << chan_index) ) { - sse_cmpps( + emit_cmpps( func, make_xmm( registers[chan_index] ), get_temp( @@ -943,17 +1293,17 @@ emit_kil( cc_LessThan ); if( chan_index == firstchan ) { - sse_pmovmskb( + emit_pmovmskb( func, x86_make_reg( file_REG32, reg_AX ), make_xmm( registers[chan_index] ) ); } else { - sse_pmovmskb( + emit_pmovmskb( func, x86_make_reg( file_REG32, reg_DX ), make_xmm( registers[chan_index] ) ); - x86_or( + emit_or( func, x86_make_reg( file_REG32, reg_AX ), x86_make_reg( file_REG32, reg_DX ) ); @@ -961,17 +1311,17 @@ emit_kil( } } - x86_or( + emit_or( func, get_temp( TGSI_EXEC_TEMP_KILMASK_I, TGSI_EXEC_TEMP_KILMASK_C ), x86_make_reg( file_REG32, reg_AX ) ); - x86_pop( + emit_pop( func, x86_make_reg( file_REG32, reg_DX ) ); - x86_pop( + emit_pop( func, x86_make_reg( file_REG32, reg_AX ) ); } @@ -987,12 +1337,12 @@ emit_setcc( FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); FETCH( func, *inst, 1, 1, chan_index ); - sse_cmpps( + emit_cmpps( func, make_xmm( 0 ), make_xmm( 1 ), cc ); - sse_andps( + emit_andps( func, make_xmm( 0 ), get_temp( @@ -1013,22 +1363,22 @@ emit_cmp( FETCH( func, *inst, 0, 0, chan_index ); FETCH( func, *inst, 1, 1, chan_index ); FETCH( func, *inst, 2, 2, chan_index ); - sse_cmpps( + emit_cmpps( func, make_xmm( 0 ), get_temp( TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ), cc_LessThan ); - sse_andps( + emit_andps( func, make_xmm( 1 ), make_xmm( 0 ) ); - sse_andnps( + emit_andnps( func, make_xmm( 0 ), make_xmm( 2 ) ); - sse_orps( + emit_orps( func, make_xmm( 0 ), make_xmm( 1 ) ); @@ -1079,7 +1429,7 @@ emit_instruction( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { FETCH( func, *inst, 0, 0, CHAN_X ); - sse_maxps( + emit_maxps( func, make_xmm( 0 ), get_temp( @@ -1089,20 +1439,20 @@ emit_instruction( } if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { FETCH( func, *inst, 1, 0, CHAN_Y ); - sse_maxps( + emit_maxps( func, make_xmm( 1 ), get_temp( TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ) ); FETCH( func, *inst, 2, 0, CHAN_W ); - sse_minps( + emit_minps( func, make_xmm( 2 ), get_temp( TGSI_EXEC_TEMP_128_I, TGSI_EXEC_TEMP_128_C ) ); - sse_maxps( + emit_maxps( func, make_xmm( 2 ), get_temp( @@ -1110,16 +1460,16 @@ emit_instruction( TGSI_EXEC_TEMP_MINUS_128_C ) ); emit_pow( func, 1, 2 ); FETCH( func, *inst, 0, 0, CHAN_X ); - sse_xorps( + emit_xorps( func, make_xmm( 2 ), make_xmm( 2 ) ); - sse_cmpps( + emit_cmpps( func, make_xmm( 2 ), make_xmm( 0 ), cc_LessThanEqual ); - sse_andps( + emit_andps( func, make_xmm( 2 ), make_xmm( 1 ) ); @@ -1241,7 +1591,7 @@ emit_instruction( FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); FETCH( func, *inst, 1, 1, chan_index ); - sse_minps( + emit_minps( func, make_xmm( 0 ), make_xmm( 1 ) ); @@ -1253,7 +1603,7 @@ emit_instruction( FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); FETCH( func, *inst, 1, 1, chan_index ); - sse_maxps( + emit_maxps( func, make_xmm( 0 ), make_xmm( 1 ) ); @@ -1393,9 +1743,9 @@ emit_instruction( FETCH( func, *inst, 4, 1, CHAN_Y ); } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - emit_mov( func, 2, 0 ); + emit_MOV( func, 2, 0 ); emit_mul( func, 2, 1 ); - emit_mov( func, 5, 3 ); + emit_MOV( func, 5, 3 ); emit_mul( func, 5, 4 ); emit_sub( func, 2, 5 ); STORE( func, *inst, 2, 0, CHAN_X ); @@ -1584,9 +1934,9 @@ emit_instruction( case TGSI_OPCODE_RET: #ifdef WIN32 - x86_retw( func, 16 ); + emit_retw( func, 16 ); #else - x86_ret( func ); + emit_ret( func ); #endif break; @@ -1825,21 +2175,23 @@ tgsi_emit_sse2( { struct tgsi_parse_context parse; + DUMP_START(); + func->csr = func->store; - x86_mov( + emit_mov( func, get_input_base(), get_argument( 0 ) ); - x86_mov( + emit_mov( func, get_output_base(), get_argument( 1 ) ); - x86_mov( + emit_mov( func, get_const_base(), get_argument( 2 ) ); - x86_mov( + emit_mov( func, get_temp_base(), get_argument( 3 ) ); @@ -1866,6 +2218,8 @@ tgsi_emit_sse2( tgsi_parse_free( &parse ); + DUMP_END(); + return 1; } @@ -1885,22 +2239,24 @@ tgsi_emit_sse2_fs( struct tgsi_parse_context parse; boolean instruction_phase = FALSE; + DUMP_START(); + func->csr = func->store; /* DECLARATION phase, do not load output argument. */ - x86_mov( + emit_mov( func, get_input_base(), get_argument( 0 ) ); - x86_mov( + emit_mov( func, get_const_base(), get_argument( 2 ) ); - x86_mov( + emit_mov( func, get_temp_base(), get_argument( 3 ) ); - x86_mov( + emit_mov( func, get_coef_base(), get_argument( 4 ) ); @@ -1921,7 +2277,7 @@ tgsi_emit_sse2_fs( if( !instruction_phase ) { /* INSTRUCTION phase, overwrite coeff with output. */ instruction_phase = TRUE; - x86_mov( + emit_mov( func, get_output_base(), get_argument( 1 ) ); @@ -1938,5 +2294,7 @@ tgsi_emit_sse2_fs( tgsi_parse_free( &parse ); + DUMP_END(); + return 1; } -- cgit v1.2.3 From abe8cd19171def0de000e58b9f71c43adf4c6336 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 26 Oct 2007 19:53:53 -0400 Subject: Make sure the swizzling vector is being recreated for each function. This makes GLSL bricks work. --- src/mesa/pipe/llvm/storage.cpp | 1 + src/mesa/pipe/llvm/storage.h | 9 ++++----- 2 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 7300cdfef0..dca8b39958 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -383,6 +383,7 @@ void Storage::pushTemps() } ++i; } + m_extSwizzleVec = 0; } void Storage::popTemps() diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index eeadaa0506..ac3e4d5af9 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -106,11 +106,6 @@ private: llvm::VectorType *m_floatVecType; llvm::VectorType *m_intVecType; - llvm::Value *m_undefFloatVec; - llvm::Value *m_undefIntVec; - - llvm::Value *m_extSwizzleVec; - char m_name[32]; int m_idx; @@ -118,6 +113,10 @@ private: std::map m_destWriteMap; + llvm::Value *m_undefFloatVec; + llvm::Value *m_undefIntVec; + llvm::Value *m_extSwizzleVec; + struct Args { llvm::Value *out; llvm::Value *in; -- cgit v1.2.3 From 25b17b213b7ba0d1b93ec37211504ee489944ce8 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 29 Oct 2007 08:27:32 -0400 Subject: Refactor the LLVM code a bit. Move the CPU vertex shader execution code to the draw module, remove traces of LLVM from the state tracker, abstract execution engine for the purposes of the draw module. --- src/mesa/pipe/draw/draw_private.h | 8 +++ src/mesa/pipe/draw/draw_vertex_shader.c | 10 +++- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 2 +- src/mesa/pipe/llvm/llvmtgsi.cpp | 86 +++++++++++++++++++++++----- src/mesa/pipe/llvm/llvmtgsi.h | 13 +++-- src/mesa/state_tracker/st_program.c | 4 +- 6 files changed, 97 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index ac47d2a76f..f52ff0bd44 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -50,6 +50,8 @@ #include "pipe/tgsi/exec/tgsi_core.h" +struct gallivm_prog; +struct gallivm_cpu_engine; /** * Basic vertex info. * Carry some useful information around with the vertices in the prim pipe. @@ -127,6 +129,9 @@ struct draw_vertex_shader { #if defined(__i386__) || defined(__386__) struct x86_function sse2_program; #endif +#ifdef MESA_LLVM + struct gallivm_prog *llvm_prog; +#endif }; /** @@ -226,6 +231,9 @@ struct draw_context } pq; int use_sse : 1; +#ifdef MESA_LLVM + struct gallivm_cpu_engine *engine; +#endif }; diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 9dbb317f2a..7fd17292b8 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -39,6 +39,7 @@ #include "x86/rtasm/x86sse.h" #include "pipe/tgsi/exec/tgsi_core.h" +#include "pipe/llvm/llvmtgsi.h" #define DBG 0 @@ -187,7 +188,7 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) // fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); #ifdef MESA_LLVM - if (draw->vertex_shader->state->llvm_prog) { + if (draw->vertex_shader->llvm_prog) { draw_vertex_shader_queue_flush_llvm(draw); return; } @@ -233,6 +234,13 @@ draw_create_vertex_shader(struct draw_context *draw, tgsi_emit_sse2( sh->tokens, &vs->sse2_program ); } #endif +#ifdef MESA_LLVM + vs->llvm_prog = gallivm_from_tgsi(shader->tokens); + if (!draw->engine) + draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); + else + gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); +#endif return vs; } diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index c0720d2872..b340ab38fd 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -125,7 +125,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) float inputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; float outputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; float (*consts)[4] = (float (*)[4]) draw->mapped_constants; - struct gallivm_prog *prog = (struct gallivm_prog *)draw->vertex_shader->state->llvm_prog; + struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index af602326ae..b57a0f8366 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -70,12 +70,16 @@ #ifdef MESA_LLVM struct gallivm_prog { - void *module; + llvm::Module *module; void *function; int num_consts; int id; }; +struct gallivm_cpu_engine { + llvm::ExecutionEngine *engine; +}; + using namespace llvm; #include "llvm_base_shader.cpp" @@ -698,13 +702,21 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) return mod; } +/*! + Translates the TGSI tokens into LLVM format. Translated representation + is stored in the gallivm_prog and returned. + After calling this function the gallivm_prog can either be used with a custom + code generator to generate machine code for the GPU which the code generator + addresses or it can be jit compiled with gallivm_cpu_jit_compile and executed + with gallivm_prog_exec to run the module on the CPU. + */ struct gallivm_prog * -gallivm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) +gallivm_from_tgsi(const struct tgsi_token *tokens) { std::cout << "Creating llvm from: " <id = GLOBAL_ID; tgsi_dump(tokens, 0); @@ -718,20 +730,8 @@ gallivm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens) AddStandardCompilePasses(passes); passes.run(*mod); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = 0; - if (!pipe->llvm_execution_engine) { - ee = llvm::ExecutionEngine::create(mp, false); - pipe->llvm_execution_engine = ee; - } else { - ee = (llvm::ExecutionEngine*)pipe->llvm_execution_engine; - ee->addModuleProvider(mp); - } gallivm->module = mod; - Function *func = mod->getFunction("run_vertex_shader"); - gallivm->function = ee->getPointerToFunctionOrStub(func); - gallivm_prog_dump(gallivm, 0); return gallivm; @@ -754,6 +754,12 @@ typedef void (*vertex_shader_runner)(float (*ainputs)[PIPE_MAX_SHADER_INPUTS][4] int num_attribs, int num_consts); + +/*! + This function is used to execute the gallivm_prog in software. Before calling + this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile + function. + */ int gallivm_prog_exec(struct gallivm_prog *prog, float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], float (*dests)[PIPE_MAX_SHADER_INPUTS][4], @@ -763,6 +769,7 @@ int gallivm_prog_exec(struct gallivm_prog *prog, int num_attribs) { vertex_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); runner(inputs, dests, consts, num_vertices, num_inputs, num_attribs, prog->num_consts); @@ -803,4 +810,53 @@ void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) } } + +/*! + This function creates a CPU based execution engine for the given gallivm_prog. + gallivm_cpu_engine should be used as a singleton throughout the library. Before + executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile. + The gallivm_prog instance which is being passed to the constructor is being + automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile + with it again. + */ +struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) +{ + struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *) + calloc(1, sizeof(struct gallivm_cpu_engine)); + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); + cpu->engine = ee; + + llvm::Function *func = mod->getFunction("run_vertex_shader"); + prog->function = ee->getPointerToFunctionOrStub(func); + return cpu; +} + + +/*! + This function JIT compiles the given gallivm_prog with the given cpu based execution engine. + The reference to the generated machine code entry point will be stored + in the gallivm_prog program. After executing this function one can call gallivm_prog_exec + in order to execute the gallivm_prog on the CPU. + */ +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) +{ + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = cpu->engine; + assert(ee); + ee->addModuleProvider(mp); + + llvm::Function *func = mod->getFunction("run_vertex_shader"); + prog->function = ee->getPointerToFunctionOrStub(func); +} + +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) +{ + free(cpu); +} + #endif /* MESA_LLVM */ + + diff --git a/src/mesa/pipe/llvm/llvmtgsi.h b/src/mesa/pipe/llvm/llvmtgsi.h index d4cf61ca97..a76dfd60b5 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.h +++ b/src/mesa/pipe/llvm/llvmtgsi.h @@ -44,15 +44,12 @@ extern "C" { struct tgsi_exec_machine; struct tgsi_token; struct tgsi_sampler; -struct pipe_context; struct gallivm_prog; +struct gallivm_cpu_engine; -struct gallivm_prog * -gallivm_from_tgsi(struct pipe_context *pipe, const struct tgsi_token *tokens); - +struct gallivm_prog *gallivm_from_tgsi(const struct tgsi_token *tokens); void gallivm_prog_delete(struct gallivm_prog *prog); - int gallivm_prog_exec(struct gallivm_prog *prog, float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], float (*dests)[PIPE_MAX_SHADER_INPUTS][4], @@ -60,9 +57,13 @@ int gallivm_prog_exec(struct gallivm_prog *prog, int num_vertices, int num_inputs, int num_attribs); - void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); + +struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); + #endif /* MESA_LLVM */ #if defined __cplusplus diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 8c61815b9b..706238cbf5 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -251,9 +251,7 @@ st_translate_vertex_program(struct st_context *st, tokensOut, maxTokens); vs.tokens = tokensOut; -#ifdef MESA_LLVM - vs.llvm_prog = (void*)gallivm_from_tgsi(st->pipe, vs.tokens); -#endif + cso = st_cached_vs_state(st, &vs); stvp->vs = cso; -- cgit v1.2.3 From a70c5e37f1c2e43738469e4799ad2b9e7c604782 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 29 Oct 2007 08:35:08 -0400 Subject: Remove typedefs from enums. typedefs are rather evil, remove them and use the enum keyword explicitely. --- src/mesa/pipe/draw/draw_flatshade.c | 2 +- src/mesa/pipe/draw/draw_vertex.c | 4 ++-- src/mesa/pipe/draw/draw_vertex.h | 16 ++++++++-------- src/mesa/pipe/i915simple/i915_context.h | 2 +- src/mesa/pipe/i915simple/i915_state_derived.c | 2 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 6 +++--- src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- src/mesa/state_tracker/st_draw.c | 2 +- 8 files changed, 18 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 2ba0d5820e..3b22c01b34 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -56,7 +56,7 @@ static INLINE void copy_colors( struct draw_stage *stage, const struct vertex_header *src ) { const uint num_attribs = stage->draw->vertex_info.num_attribs; - const interp_mode *interp = stage->draw->vertex_info.interp_mode; + const enum interp_mode *interp = stage->draw->vertex_info.interp_mode; uint i; /* Look for constant/flat attribs and duplicate from src to dst vertex */ diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index 1204a944de..ce76cb043c 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -44,7 +44,7 @@ static INLINE void emit_vertex_attr(struct vertex_info *vinfo, - attrib_format format, interp_mode interp) + enum attrib_format format, enum interp_mode interp) { const uint n = vinfo->num_attribs; vinfo->interp_mode[n] = interp; @@ -95,7 +95,7 @@ draw_compute_vertex_size(struct vertex_info *vinfo) void draw_set_vertex_attributes( struct draw_context *draw, const uint *slot_to_vf_attr, - const interp_mode *interps, + const enum interp_mode *interps, unsigned nr_attrs ) { struct vertex_info *vinfo = &draw->vertex_info; diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 2968981708..a1fa7aae5a 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -40,7 +40,7 @@ struct draw_context; /** * Vertex attribute format */ -typedef enum { +enum attrib_format { FORMAT_OMIT, FORMAT_1F, FORMAT_2F, @@ -48,18 +48,18 @@ typedef enum { FORMAT_4F, FORMAT_4F_VIEWPORT, FORMAT_4UB -} attrib_format; +}; /** * Attribute interpolation mode */ -typedef enum { +enum interp_mode { INTERP_NONE, /**< never interpolate vertex header info */ INTERP_CONSTANT, INTERP_LINEAR, INTERP_PERSPECTIVE -} interp_mode; +}; /** @@ -69,8 +69,8 @@ struct vertex_info { uint num_attribs; uint hwfmt[4]; /**< hardware format info for this format */ - interp_mode interp_mode[PIPE_MAX_SHADER_OUTPUTS]; - attrib_format format[PIPE_MAX_SHADER_OUTPUTS]; /**< FORMAT_x */ + enum interp_mode interp_mode[PIPE_MAX_SHADER_OUTPUTS]; + enum attrib_format format[PIPE_MAX_SHADER_OUTPUTS]; /**< FORMAT_x */ uint size; /**< total vertex size in dwords */ }; @@ -82,7 +82,7 @@ struct vertex_info */ static INLINE uint draw_emit_vertex_attr(struct vertex_info *vinfo, - attrib_format format, interp_mode interp) + enum attrib_format format, enum interp_mode interp) { const uint n = vinfo->num_attribs; assert(n < PIPE_MAX_SHADER_OUTPUTS); @@ -95,7 +95,7 @@ draw_emit_vertex_attr(struct vertex_info *vinfo, extern void draw_set_vertex_attributes( struct draw_context *draw, const uint *attrs, - const interp_mode *interps, + const enum interp_mode *interps, unsigned nr_attrs ); extern void draw_set_twoside_attributes(struct draw_context *draw, diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 5d312286f6..5c748ea896 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -130,7 +130,7 @@ struct i915_depth_stencil_state { struct i915_rasterizer_state { int light_twoside : 1; unsigned st; - interp_mode color_interp; + enum interp_mode color_interp; unsigned LIS4; unsigned LIS7; diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 95f705929e..ed1521fcce 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -44,7 +44,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) { const struct pipe_shader_state *fs = i915->fs; - const interp_mode colorInterp = i915->rasterizer->color_interp; + const enum interp_mode colorInterp = i915->rasterizer->color_interp; struct vertex_info vinfo; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; boolean needW = 0; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 0b1b592bbe..21d643bfe0 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -471,7 +471,7 @@ static void tri_persp_coeff( struct setup_stage *setup, */ static void setup_tri_coefficients( struct setup_stage *setup ) { - const interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; unsigned slot, j; /* z and w are done by linear interpolation: @@ -701,7 +701,7 @@ line_persp_coeff(struct setup_stage *setup, unsigned slot, unsigned i) static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { - const interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; unsigned slot, j; /* use setup->vmin, vmax to point to vertices */ @@ -927,7 +927,7 @@ static void setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - const interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; const struct vertex_header *v0 = prim->v[0]; const int sizeAttr = setup->softpipe->psize_slot; const float size diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index b6145df8e2..e5ca953e60 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -45,7 +45,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { const struct pipe_shader_state *vs = softpipe->vs->state; const struct pipe_shader_state *fs = softpipe->fs; - const interp_mode colorInterp + const enum interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; boolean emitBack0 = FALSE, emitBack1 = FALSE, emitPsize = FALSE; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index d7d9a1d203..065e157bc6 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -325,7 +325,7 @@ set_feedback_vertex_format(GLcontext *ctx) { struct st_context *st = ctx->st; uint attrs[PIPE_MAX_SHADER_OUTPUTS]; - interp_mode interp[PIPE_MAX_SHADER_OUTPUTS]; + enum interp_mode interp[PIPE_MAX_SHADER_OUTPUTS]; GLuint n, i; if (ctx->RenderMode == GL_FEEDBACK) { -- cgit v1.2.3 From a6a3d8cb755c36d4ae1dc4dbfb310f06862a37b4 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 29 Oct 2007 10:01:29 -0400 Subject: Remove conditionals from the makefiles. --- src/mesa/pipe/llvm/instructions.cpp | 2 ++ src/mesa/pipe/llvm/llvmtgsi.cpp | 3 +-- src/mesa/pipe/llvm/storage.cpp | 2 ++ src/mesa/sources | 4 +--- 4 files changed, 6 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 645ab9106f..b0aa0cc746 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -29,6 +29,7 @@ * Authors: * Zack Rusin zack@tungstengraphics.com */ +#ifdef MESA_LLVM #include "instructions.h" @@ -1165,3 +1166,4 @@ llvm::Function * Instructions::findFunction(int label) return func; } +#endif //MESA_LLVM diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index b57a0f8366..4611ac1766 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -29,6 +29,7 @@ * Authors: * Zack Rusin zack@tungstengraphics.com */ +#ifdef MESA_LLVM #include "llvmtgsi.h" @@ -67,8 +68,6 @@ #include #include -#ifdef MESA_LLVM - struct gallivm_prog { llvm::Module *module; void *function; diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index dca8b39958..ff62fcf3e4 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -29,6 +29,7 @@ * Authors: * Zack Rusin zack@tungstengraphics.com */ +#ifdef MESA_LLVM #include "storage.h" @@ -391,3 +392,4 @@ void Storage::popTemps() m_temps = m_tempStack.top(); m_tempStack.pop(); } +#endif //MESA_LLVM diff --git a/src/mesa/sources b/src/mesa/sources index 30d3c32b83..e27f02a9bc 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -187,12 +187,10 @@ TGSIDECO_SOURCES = \ pipe/tgsi/deco/deco_caps.c -ifeq ($(MESA_LLVM),1) - LLVMTGSI_SOURCES = \ +LLVMTGSI_SOURCES = \ pipe/llvm/llvmtgsi.cpp \ pipe/llvm/storage.cpp \ pipe/llvm/instructions.cpp -endif STATECACHE_SOURCES = \ pipe/cso_cache/cso_hash.c \ -- cgit v1.2.3 From 2a3f3679eba802dcb4b46f90c66326c9195cdbcb Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 09:23:46 -0600 Subject: Disable the else clause which assigns the default fragment program to ctx->FragmentProgram._Current The _Current field should either point to the fragment program which is to be run (GLSL, ARB_f_p, fixed-func-generated, etc) or be NULL if conventional fixed-function code is to be used. Matches TNL program code. --- src/mesa/main/texenvprogram.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index abc2567134..21a290402f 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -1279,9 +1279,11 @@ _mesa_UpdateTexEnvProgram( GLcontext *ctx ) _mesa_printf("Found existing texenv program for key %x\n", hash); } } +#if 0 else { ctx->FragmentProgram._Current = ctx->FragmentProgram.Current; } +#endif /* Tell the driver about the change. Could define a new target for * this? -- cgit v1.2.3 From cc0b55c242e9f53a8e45c5d4392f964af44f396e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 09:24:29 -0600 Subject: simplify code which access the current vertex/fragment shaders --- src/mesa/state_tracker/st_atom_sampler.c | 21 +-------------------- src/mesa/state_tracker/st_atom_shader.c | 27 ++++----------------------- 2 files changed, 5 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 80b8cae013..bdc6024477 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -113,24 +113,6 @@ gl_filter_to_img_filter(GLenum filter) } -static struct gl_fragment_program * -current_fragment_program(GLcontext *ctx) -{ - struct gl_fragment_program *f; - - if (ctx->Shader.CurrentProgram && - ctx->Shader.CurrentProgram->LinkStatus && - ctx->Shader.CurrentProgram->FragmentProgram) { - f = ctx->Shader.CurrentProgram->FragmentProgram; - } - else { - f = ctx->FragmentProgram._Current; - assert(f); - } - return f; -} - - static void update_samplers(struct st_context *st) { @@ -186,8 +168,7 @@ update_samplers(struct st_context *st) /* mapping from sampler vars to texture units */ { - struct gl_fragment_program *fprog = current_fragment_program(st->ctx); - const GLubyte *samplerUnits = fprog->Base.SamplerUnits; + struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current; uint sample_units[PIPE_MAX_SAMPLERS]; uint s; for (s = 0; s < PIPE_MAX_SAMPLERS; s++) { diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 92ca22851e..d4976941f9 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -264,30 +264,11 @@ update_linkage( struct st_context *st ) /* find active shader and params -- Should be covered by * ST_NEW_VERTEX_PROGRAM */ - if (st->ctx->Shader.CurrentProgram && - st->ctx->Shader.CurrentProgram->LinkStatus && - st->ctx->Shader.CurrentProgram->VertexProgram) { - struct gl_vertex_program *f - = st->ctx->Shader.CurrentProgram->VertexProgram; - stvp = st_vertex_program(f); - } - else { - assert(st->ctx->VertexProgram._Current); - stvp = st_vertex_program(st->ctx->VertexProgram._Current); - } - + assert(st->ctx->VertexProgram._Current); + stvp = st_vertex_program(st->ctx->VertexProgram._Current); - if (st->ctx->Shader.CurrentProgram && - st->ctx->Shader.CurrentProgram->LinkStatus && - st->ctx->Shader.CurrentProgram->FragmentProgram) { - struct gl_fragment_program *f - = st->ctx->Shader.CurrentProgram->FragmentProgram; - stfp = st_fragment_program(f); - } - else { - assert(st->ctx->FragmentProgram._Current); - stfp = st_fragment_program(st->ctx->FragmentProgram._Current); - } + assert(st->ctx->FragmentProgram._Current); + stfp = st_fragment_program(st->ctx->FragmentProgram._Current); xvp = find_translated_vp(st, stvp, stfp); -- cgit v1.2.3 From b58b64f361bef38a76b199427a4e248b1fab9f65 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 09:41:53 -0600 Subject: fix bad fragment shader pointer assignment --- src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index e5ca953e60..81f70f0f24 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -44,7 +44,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { const struct pipe_shader_state *vs = softpipe->vs->state; - const struct pipe_shader_state *fs = softpipe->fs; + const struct pipe_shader_state *fs = &softpipe->fs->shader; const enum interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; -- cgit v1.2.3 From ebe1642d7a65acd5e41632b3b23655f67de85cfc Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 10:36:10 -0600 Subject: check for signed vs. unsigned in st_get_format_info() - fixes accum buffer failure --- src/mesa/state_tracker/st_format.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index c0e1a79bad..bffc39be87 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -204,19 +204,25 @@ st_get_format_info( else { GLuint size; - assert( info.type == PIPE_FORMAT_TYPE_UNORM ); - size = format_max_bits( info ); if (size == 8) { - pinfo->datatype = GL_UNSIGNED_BYTE; + if (info.type == PIPE_FORMAT_TYPE_UNORM) + pinfo->datatype = GL_UNSIGNED_BYTE; + else + pinfo->datatype = GL_BYTE; } else if (size == 16) { - pinfo->datatype = GL_UNSIGNED_SHORT; + if (info.type == PIPE_FORMAT_TYPE_UNORM) + pinfo->datatype = GL_UNSIGNED_SHORT; + else + pinfo->datatype = GL_SHORT; } else { assert( size <= 32 ); - - pinfo->datatype = GL_UNSIGNED_INT; + if (info.type == PIPE_FORMAT_TYPE_UNORM) + pinfo->datatype = GL_UNSIGNED_INT; + else + pinfo->datatype = GL_INT; } } -- cgit v1.2.3 From d37eb130c090abc413a54e57e0d3800ac4cd59f4 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 10:37:12 -0600 Subject: remove dead code --- src/mesa/state_tracker/st_format.c | 77 -------------------------------------- 1 file changed, 77 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index bffc39be87..d5e8581b21 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -103,83 +103,6 @@ st_get_format_info( GLuint format, struct pipe_format_info *pinfo ) { -#if 0 - static const struct pipe_format_info info[] = { - { - PIPE_FORMAT_U_R8_G8_B8_A8, /* format */ - GL_RGBA, /* base_format */ - GL_UNSIGNED_BYTE, /* datatype for renderbuffers */ - 8, 8, 8, 8, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 4 /* size in bytes */ - }, - { - PIPE_FORMAT_U_A8_R8_G8_B8, - GL_RGBA, /* base_format */ - GL_UNSIGNED_BYTE, /* datatype for renderbuffers */ - 8, 8, 8, 8, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 4 /* size in bytes */ - }, - { - PIPE_FORMAT_U_A1_R5_G5_B5, - GL_RGBA, /* base_format */ - GL_UNSIGNED_SHORT, /* datatype for renderbuffers */ - 5, 5, 5, 1, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 2 /* size in bytes */ - }, - { - PIPE_FORMAT_U_R5_G6_B5, - GL_RGBA, /* base_format */ - GL_UNSIGNED_SHORT, /* datatype for renderbuffers */ - 5, 6, 5, 0, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 2 /* size in bytes */ - }, - { - PIPE_FORMAT_S_R16_G16_B16_A16, - GL_RGBA, /* base_format */ - GL_UNSIGNED_SHORT, /* datatype for renderbuffers */ - 16, 16, 16, 16, 0, 0, /* color bits */ - 0, 0, /* depth, stencil */ - 8 /* size in bytes */ - }, - { - PIPE_FORMAT_U_Z16, - GL_DEPTH_COMPONENT, /* base_format */ - GL_UNSIGNED_SHORT, /* datatype for renderbuffers */ - 0, 0, 0, 0, 0, 0, /* color bits */ - 16, 0, /* depth, stencil */ - 2 /* size in bytes */ - }, - { - PIPE_FORMAT_U_Z32, - GL_DEPTH_COMPONENT, /* base_format */ - GL_UNSIGNED_INT, /* datatype for renderbuffers */ - 0, 0, 0, 0, 0, 0, /* color bits */ - 32, 0, /* depth, stencil */ - 4 /* size in bytes */ - }, - { - PIPE_FORMAT_S8_Z24, - GL_DEPTH_STENCIL_EXT, /* base_format */ - GL_UNSIGNED_INT, /* datatype for renderbuffers */ - 0, 0, 0, 0, 0, 0, /* color bits */ - 24, 8, /* depth, stencil */ - 4 /* size in bytes */ - } - /* XXX lots more cases to add */ - }; - GLuint i; - - for (i = 0; i < sizeof(info) / sizeof(info[0]); i++) { - if (info[i].format == format) - return info + i; - } - return NULL; -#endif - union pipe_format fmt; fmt.value32 = format; -- cgit v1.2.3 From ee295fccdd0c94cb6b8af4dfb30283e39f548223 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 29 Oct 2007 16:20:45 +0000 Subject: Make gallium compile in win32. Use FREE, MALLOC, CALLOC, GETENV wrappers. Silence compiler warnings. Add proper copyrights. --- src/mesa/pipe/draw/draw_context.c | 16 ++-- src/mesa/pipe/draw/draw_vertex_fetch.c | 8 +- src/mesa/pipe/draw/draw_vertex_shader.c | 15 ++-- src/mesa/pipe/draw/draw_wide_prims.c | 60 ++++++------- src/mesa/pipe/p_context.h | 6 +- src/mesa/pipe/p_util.h | 126 ++++++++++++++++++++++----- src/mesa/pipe/softpipe/sp_clear.c | 8 +- src/mesa/pipe/softpipe/sp_context.c | 6 +- src/mesa/pipe/softpipe/sp_context.h | 3 +- src/mesa/pipe/softpipe/sp_draw_arrays.c | 1 - src/mesa/pipe/softpipe/sp_prim_setup.c | 14 +-- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 4 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 47 +++++----- src/mesa/pipe/softpipe/sp_quad_fs.c | 47 +++++----- src/mesa/pipe/softpipe/sp_region.c | 2 +- src/mesa/pipe/softpipe/sp_state_blend.c | 12 +-- src/mesa/pipe/softpipe/sp_state_fs.c | 33 ++++--- src/mesa/pipe/softpipe/sp_state_rasterizer.c | 6 +- src/mesa/pipe/softpipe/sp_state_sampler.c | 4 +- src/mesa/pipe/softpipe/sp_surface.c | 8 +- src/mesa/pipe/softpipe/sp_tex_layout.c | 6 +- src/mesa/pipe/softpipe/sp_tex_sample.c | 5 +- src/mesa/pipe/softpipe/sp_tile_cache.c | 4 +- src/mesa/pipe/tgsi/exec/tgsi_dump.c | 44 ++++++---- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 22 +++-- src/mesa/pipe/tgsi/exec/tgsi_parse.c | 31 ++++++- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 29 +++++- 27 files changed, 366 insertions(+), 201 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 61f9e4909b..80317dd6cf 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -42,7 +42,7 @@ struct draw_context *draw_create( void ) struct draw_context *draw = CALLOC_STRUCT( draw_context ); #if defined(__i386__) || defined(__386__) - draw->use_sse = getenv("GALLIUM_SSE") != NULL; + draw->use_sse = GETENV( "GALLIUM_SSE" ) != NULL; #else draw->use_sse = FALSE; #endif @@ -71,7 +71,7 @@ struct draw_context *draw_create( void ) */ { int i; - char *tmp = malloc(Elements(draw->vcache.vertex) * MAX_VERTEX_SIZE); + char *tmp = MALLOC( Elements(draw->vcache.vertex) * MAX_VERTEX_SIZE ); for (i = 0; i < Elements(draw->vcache.vertex); i++) draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * MAX_VERTEX_SIZE); @@ -92,8 +92,8 @@ struct draw_context *draw_create( void ) void draw_destroy( struct draw_context *draw ) { - free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ - free( draw ); + FREE( draw->vcache.vertex[0] ); /* Frees all the vertices. */ + FREE( draw ); } @@ -233,10 +233,10 @@ void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) stage->nr_tmps = nr; if (nr) { - ubyte *store = (ubyte *) malloc(MAX_VERTEX_SIZE * nr); + ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); unsigned i; - stage->tmp = (struct vertex_header **) malloc(sizeof(struct vertex_header *) * nr); + stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); for (i = 0; i < nr; i++) stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); @@ -246,8 +246,8 @@ void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) void draw_free_tmps( struct draw_stage *stage ) { if (stage->tmp) { - free(stage->tmp[0]); - free(stage->tmp); + FREE( stage->tmp[0] ); + FREE( stage->tmp ); } } diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index de1cd06da6..5a7e6febe9 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -64,16 +64,16 @@ fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) break; case PIPE_FORMAT_R32G32B32A32_SSCALED: - attrib[3] = ((int *) ptr)[3]; + attrib[3] = (float) ((int *) ptr)[3]; /* fall-through */ case PIPE_FORMAT_R32G32B32_SSCALED: - attrib[2] = ((int *) ptr)[2]; + attrib[2] = (float) ((int *) ptr)[2]; /* fall-through */ case PIPE_FORMAT_R32G32_SSCALED: - attrib[1] = ((int *) ptr)[1]; + attrib[1] = (float) ((int *) ptr)[1]; /* fall-through */ case PIPE_FORMAT_R32_SSCALED: - attrib[0] = ((int *) ptr)[0]; + attrib[0] = (float) ((int *) ptr)[0]; break; default: diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 7fd17292b8..056ad007cf 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -42,7 +42,7 @@ #include "pipe/llvm/llvmtgsi.h" -#define DBG 0 +#define DBG_VS 0 static INLINE unsigned @@ -149,7 +149,7 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][2] = z * scale[2] + trans[2]; vOut[j]->data[0][3] = w; -#if DBG +#if DBG_VS printf("output[%d]win: %f %f %f %f\n", j, vOut[j]->data[0][0], vOut[j]->data[0][1], @@ -166,7 +166,7 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; -#if DBG +#if DBG_VS printf("output[%d][%d]: %f %f %f %f\n", j, slot, vOut[j]->data[slot][0], vOut[j]->data[slot][1], @@ -222,7 +222,11 @@ draw_create_vertex_shader(struct draw_context *draw, { struct draw_vertex_shader *vs; - vs = calloc( 1, sizeof( struct draw_vertex_shader ) ); + vs = CALLOC_STRUCT( draw_vertex_shader ); + if (vs == NULL) { + return NULL; + } + vs->state = shader; #if defined(__i386__) || defined(__386__) @@ -269,6 +273,5 @@ void draw_delete_vertex_shader(struct draw_context *draw, x86_release_func( (struct x86_function *) &vs->sse2_program ); #endif - free( vs->state ); - free( vs ); + FREE( vs ); } diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 76b8a5319b..f3b4478f9a 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -111,44 +111,44 @@ static void wide_line( struct draw_stage *stage, if (dx > dy) { /* x-major line */ - pos0[1] = pos0[1] - half_width - 0.25; - pos1[1] = pos1[1] + half_width - 0.25; - pos2[1] = pos2[1] - half_width - 0.25; - pos3[1] = pos3[1] + half_width - 0.25; + pos0[1] = pos0[1] - half_width - 0.25f; + pos1[1] = pos1[1] + half_width - 0.25f; + pos2[1] = pos2[1] - half_width - 0.25f; + pos3[1] = pos3[1] + half_width - 0.25f; if (pos0[0] < pos2[0]) { /* left to right line */ - pos0[0] -= 0.5; - pos1[0] -= 0.5; - pos2[0] -= 0.5; - pos3[0] -= 0.5; + pos0[0] -= 0.5f; + pos1[0] -= 0.5f; + pos2[0] -= 0.5f; + pos3[0] -= 0.5f; } else { /* right to left line */ - pos0[0] += 0.5; - pos1[0] += 0.5; - pos2[0] += 0.5; - pos3[0] += 0.5; + pos0[0] += 0.5f; + pos1[0] += 0.5f; + pos2[0] += 0.5f; + pos3[0] += 0.5f; } } else { /* y-major line */ - pos0[0] = pos0[0] - half_width + 0.25; - pos1[0] = pos1[0] + half_width + 0.25; - pos2[0] = pos2[0] - half_width + 0.25; - pos3[0] = pos3[0] + half_width + 0.25; + pos0[0] = pos0[0] - half_width + 0.25f; + pos1[0] = pos1[0] + half_width + 0.25f; + pos2[0] = pos2[0] - half_width + 0.25f; + pos3[0] = pos3[0] + half_width + 0.25f; if (pos0[1] < pos2[1]) { /* top to bottom line */ - pos0[1] -= 0.5; - pos1[1] -= 0.5; - pos2[1] -= 0.5; - pos3[1] -= 0.5; + pos0[1] -= 0.5f; + pos1[1] -= 0.5f; + pos2[1] -= 0.5f; + pos3[1] -= 0.5f; } else { /* bottom to top line */ - pos0[1] += 0.5; - pos1[1] += 0.5; - pos2[1] += 0.5; - pos3[1] += 0.5; + pos0[1] += 0.5f; + pos1[1] += 0.5f; + pos2[1] += 0.5f; + pos3[1] += 0.5f; } } @@ -179,7 +179,7 @@ static void set_texcoords(const struct wide_stage *wide, uint j = wide->texcoord_slot[i]; v->data[j][0] = tc[0]; if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) - v->data[j][1] = 1.0 - tc[1]; + v->data[j][1] = 1.0f - tc[1]; else v->data[j][1] = tc[1]; v->data[j][2] = tc[2]; @@ -198,7 +198,7 @@ static void wide_point( struct draw_stage *stage, struct prim_header *header ) { const struct wide_stage *wide = wide_stage(stage); - const boolean sprite = stage->draw->rasterizer->point_sprite; + const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; float half_size; float left_adj, right_adj; @@ -223,8 +223,8 @@ static void wide_point( struct draw_stage *stage, half_size = wide->half_point_size; } - left_adj = -half_size + 0.25; - right_adj = half_size + 0.25; + left_adj = -half_size + 0.25f; + right_adj = half_size + 0.25f; pos0[0] += left_adj; pos0[1] -= half_size; @@ -267,8 +267,8 @@ static void wide_begin( struct draw_stage *stage ) struct wide_stage *wide = wide_stage(stage); struct draw_context *draw = stage->draw; - wide->half_point_size = 0.5 * draw->rasterizer->point_size; - wide->half_line_width = 0.5 * draw->rasterizer->line_width; + wide->half_point_size = 0.5f * draw->rasterizer->point_size; + wide->half_line_width = 0.5f * draw->rasterizer->line_width; if (draw->rasterizer->line_width != 1.0) { wide->stage.line = wide_line; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index a1441a7e43..3bd266674f 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -29,7 +29,7 @@ #define PIPE_CONTEXT_H #include "p_state.h" -#include "p_compiler.h" +#include "p_util.h" struct pipe_state_cache; /** @@ -265,7 +265,7 @@ pipe_region_reference(struct pipe_region **ptr, struct pipe_region *region) /* free the old region */ assert(oldReg->map_refcount == 0); /* XXX dereference the region->buffer */ - free(oldReg); + FREE( oldReg ); } *ptr = NULL; } @@ -292,7 +292,7 @@ pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) if (oldSurf->refcount == 0) { /* free the old region */ pipe_region_reference(&oldSurf->region, NULL); - free(oldSurf); + FREE( oldSurf ); } *ptr = NULL; } diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 4f68f5e1bd..7897bc90e4 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -1,28 +1,29 @@ - -/* - * Mesa 3-D graphics library - * Version: 6.5.2 - * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. - * +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ #ifndef P_UTIL_H #define P_UTIL_H @@ -30,12 +31,84 @@ #include "p_compiler.h" #include -#define CALLOC_STRUCT(T) (struct T *) calloc(1, sizeof(struct T)) +#ifdef WIN32 + +#ifdef __cplusplus +extern "C" +{ +#endif + +void * __stdcall +EngAllocMem( + unsigned long Flags, + unsigned long MemSize, + unsigned long Tag ); + +void __stdcall +EngFreeMem( + void *Mem ); + +#ifdef __cplusplus +} +#endif + +static INLINE void * +MALLOC( unsigned size ) +{ + return EngAllocMem( 0, size, 'D3AG' ); +} + +static INLINE void * +CALLOC( unsigned count, unsigned size ) +{ + void *ptr = MALLOC( count * size ); + if( ptr ) { + memset( ptr, 0, count * size ); + } + return ptr; +} + +static INLINE void +FREE( void *ptr ) +{ + if( ptr ) { + EngFreeMem( ptr ); + } +} + +static INLINE void * +REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) +{ + void *new_ptr; + if( new_size <= old_size ) { + return old_ptr; + } + new_ptr = MALLOC( new_size ); + if( new_ptr ) { + memcpy( new_ptr, old_ptr, old_size ); + } + FREE( old_ptr ); + return new_ptr; +} + +#define GETENV( X ) NULL + +#else // WIN32 #define MALLOC( SIZE ) malloc( SIZE ) +#define CALLOC( COUNT, SIZE ) calloc( COUNT, SIZE ) + #define FREE( PTR ) free( PTR ) +#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) + +#define GETENV( X ) getenv( X ) + +#endif // WIN32 + +#define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) + #define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) @@ -221,5 +294,10 @@ static INLINE float LOG2(float val) return num.f + log_2; } +#if defined(__GNUC__) +#define CEILF(x) ceilf(x) +#else +#define CEILF(x) ((float) ceil(x)) +#endif #endif diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index df0537be1d..078b6a7964 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -72,10 +72,10 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, } else if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[0])) { float clear[4]; - clear[0] = 0.2; /* XXX hack */ - clear[1] = 0.2; /* XXX hack */ - clear[2] = 0.2; /* XXX hack */ - clear[3] = 0.2; /* XXX hack */ + clear[0] = 0.2f; /* XXX hack */ + clear[1] = 0.2f; /* XXX hack */ + clear[2] = 0.2f; /* XXX hack */ + clear[3] = 0.2f; /* XXX hack */ sp_tile_cache_clear(softpipe->cbuf_cache[0], clear); } diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index f67588783b..46f591e425 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -289,12 +289,12 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, uint i; #if defined(__i386__) || defined(__386__) - softpipe->use_sse = getenv("GALLIUM_SSE") != NULL; + softpipe->use_sse = GETENV( "GALLIUM_SSE" ) != NULL; #else softpipe->use_sse = FALSE; #endif - softpipe->dump_fs = getenv( "GALLIUM_DUMP_FS" ) != NULL; + softpipe->dump_fs = GETENV( "GALLIUM_DUMP_FS" ) != NULL; softpipe->pipe.winsys = pipe_winsys; softpipe->pipe.destroy = softpipe_destroy; @@ -396,7 +396,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, assert(softpipe->draw); softpipe->setup = sp_draw_render_stage(softpipe); - if (getenv("SP_VBUF")) { + if (GETENV( "SP_VBUF" ) != NULL) { softpipe->vbuf = sp_draw_vbuf_stage(softpipe->draw, &softpipe->pipe, sp_vbuf_setup_draw); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 720297653f..a411969bc0 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -31,7 +31,6 @@ #ifndef SP_CONTEXT_H #define SP_CONTEXT_H -#include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -63,7 +62,7 @@ struct softpipe_tile_cache; #define SP_NEW_CONSTANTS 0x4000 struct sp_vertex_shader_state { - const struct pipe_shader_state *state; + struct pipe_shader_state *state; void *draw_data; }; diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 64a4fbe333..8a82cdfe1a 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -34,7 +34,6 @@ #include "pipe/p_defines.h" #include "pipe/p_context.h" #include "pipe/p_winsys.h" -#include "pipe/p_util.h" #include "sp_context.h" #include "sp_state.h" diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 21d643bfe0..9f5a43a827 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -516,18 +516,18 @@ static void setup_tri_edges( struct setup_stage *setup ) float vmid_y = setup->vmid->data[0][1] - 0.5f; float vmax_y = setup->vmax->data[0][1] - 0.5f; - setup->emaj.sy = ceilf(vmin_y); - setup->emaj.lines = (int) ceilf(vmax_y - setup->emaj.sy); + setup->emaj.sy = CEILF(vmin_y); + setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; - setup->etop.sy = ceilf(vmid_y); - setup->etop.lines = (int) ceilf(vmax_y - setup->etop.sy); + setup->etop.sy = CEILF(vmid_y); + setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); setup->etop.dxdy = setup->etop.dx / setup->etop.dy; setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; - setup->ebot.sy = ceilf(vmin_y); - setup->ebot.lines = (int) ceilf(vmid_y - setup->ebot.sy); + setup->ebot.sy = CEILF(vmin_y); + setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; } @@ -934,7 +934,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) = sizeAttr > 0 ? v0->data[sizeAttr][0] : setup->softpipe->rasterizer->point_size; const float halfSize = 0.5F * size; - const boolean round = setup->softpipe->rasterizer->point_smooth; + const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth; const float x = v0->data[0][0]; /* Note: data[0] is always position */ const float y = v0->data[0][1]; unsigned slot, j; diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index a26179e45c..ddf662eead 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -295,8 +295,8 @@ struct draw_stage *sp_draw_vbuf_stage( struct draw_context *draw_context, vbuf->draw = draw; vbuf->draw_context = draw_context; - vbuf->element_map = malloc( IBUF_SIZE ); - vbuf->vertex_map = malloc( VBUF_SIZE ); + vbuf->element_map = MALLOC( IBUF_SIZE ); + vbuf->vertex_map = MALLOC( VBUF_SIZE ); vbuf->vertex_ptr = vbuf->vertex_map; diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 9cb01d7b2f..f7e93af784 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -1,26 +1,29 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ /** * \brief Quad depth testing @@ -188,7 +191,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); - tile->data.depth16[y][x] = bzzzz[j]; + tile->data.depth16[y][x] = (ushort) bzzzz[j]; } break; case PIPE_FORMAT_U_Z32: diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index c4f90e5064..d6d7eb6489 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -1,26 +1,29 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ /* Vertices are just an array of floats, with all the attributes * packed. We currently assume a layout like: @@ -35,8 +38,6 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" -#include "pipe/llvm/llvmtgsi.h" - #include "x86/rtasm/x86sse.h" #include "sp_context.h" diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index 80a67dcabc..fef63ef2f6 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -207,7 +207,7 @@ sp_region_fill(struct pipe_context *pipe, ushort *row = (ushort *) get_pointer(dst, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) - row[j] = value; + row[j] = (ushort) value; row += dst->pitch; } } diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 6376c3c61c..e9e6a55e0d 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -34,7 +34,7 @@ void * softpipe_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - struct pipe_blend_state *state = malloc(sizeof(struct pipe_blend_state)); + struct pipe_blend_state *state = MALLOC( sizeof(struct pipe_blend_state) ); memcpy(state, blend, sizeof(struct pipe_blend_state)); return state; } @@ -52,7 +52,7 @@ void softpipe_bind_blend_state( struct pipe_context *pipe, void softpipe_delete_blend_state(struct pipe_context *pipe, void *blend) { - free(blend); + FREE( blend ); } @@ -75,7 +75,7 @@ void * softpipe_create_alpha_test_state(struct pipe_context *pipe, const struct pipe_alpha_test_state *alpha) { - struct pipe_alpha_test_state *state = malloc(sizeof(struct pipe_alpha_test_state)); + struct pipe_alpha_test_state *state = MALLOC( sizeof(struct pipe_alpha_test_state) ); memcpy(state, alpha, sizeof(struct pipe_alpha_test_state)); return state; } @@ -95,7 +95,7 @@ void softpipe_delete_alpha_test_state(struct pipe_context *pipe, void *alpha) { - free(alpha); + FREE( alpha ); } void * @@ -103,7 +103,7 @@ softpipe_create_depth_stencil_state(struct pipe_context *pipe, const struct pipe_depth_stencil_state *depth_stencil) { struct pipe_depth_stencil_state *state = - malloc(sizeof(struct pipe_depth_stencil_state)); + MALLOC( sizeof(struct pipe_depth_stencil_state) ); memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_state)); return state; } @@ -122,5 +122,5 @@ softpipe_bind_depth_stencil_state(struct pipe_context *pipe, void softpipe_delete_depth_stencil_state(struct pipe_context *pipe, void *depth) { - free(depth); + FREE( depth ); } diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index ad8ab561ef..351c2954f5 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -29,6 +29,7 @@ #include "sp_state.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "pipe/p_winsys.h" #include "pipe/draw/draw_context.h" #include "pipe/tgsi/exec/tgsi_core.h" @@ -43,7 +44,7 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, * that now. */ - struct sp_fragment_shader_state *state = malloc(sizeof(struct sp_fragment_shader_state)); + struct sp_fragment_shader_state *state = MALLOC( sizeof(struct sp_fragment_shader_state) ); state->shader = *templ; if( softpipe->dump_fs ) { @@ -80,7 +81,7 @@ void softpipe_delete_fs_state(struct pipe_context *pipe, x86_release_func( &state->sse2_program ); #endif - free( state ); + FREE( state ); } @@ -88,15 +89,27 @@ void * softpipe_create_vs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { struct softpipe_context *softpipe = softpipe_context(pipe); - struct sp_vertex_shader_state *state = - malloc(sizeof(struct sp_vertex_shader_state)); - struct pipe_shader_state *templ_copy = - malloc(sizeof(struct pipe_shader_state)); - memcpy(templ_copy, templ, sizeof(struct pipe_shader_state)); + struct sp_vertex_shader_state *state; + + state = MALLOC( sizeof(struct sp_vertex_shader_state) ); + if (state == NULL ) { + return NULL; + } + + state->state = MALLOC( sizeof(struct pipe_shader_state) ); + if (state->state == NULL) { + FREE( state ); + return NULL; + } + memcpy( state->state, templ, sizeof(struct pipe_shader_state) ); - state->state = templ_copy; state->draw_data = draw_create_vertex_shader(softpipe->draw, state->state); + if (state->draw_data == NULL) { + FREE( state->state ); + FREE( state ); + return NULL; + } return state; } @@ -121,8 +134,8 @@ void softpipe_delete_vs_state(struct pipe_context *pipe, (struct sp_vertex_shader_state *)vs; draw_delete_vertex_shader(softpipe->draw, state->draw_data); - free(state->state); - free(state); + FREE( state->state ); + FREE( state ); } diff --git a/src/mesa/pipe/softpipe/sp_state_rasterizer.c b/src/mesa/pipe/softpipe/sp_state_rasterizer.c index 076ea66d9e..a69e0d9012 100644 --- a/src/mesa/pipe/softpipe/sp_state_rasterizer.c +++ b/src/mesa/pipe/softpipe/sp_state_rasterizer.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -37,7 +37,7 @@ softpipe_create_rasterizer_state(struct pipe_context *pipe, const struct pipe_rasterizer_state *setup) { struct pipe_rasterizer_state *state = - malloc(sizeof(struct pipe_rasterizer_state)); + MALLOC( sizeof(struct pipe_rasterizer_state) ); memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); return state; } @@ -58,7 +58,7 @@ void softpipe_bind_rasterizer_state(struct pipe_context *pipe, void softpipe_delete_rasterizer_state(struct pipe_context *pipe, void *rasterizer) { - free(rasterizer); + FREE( rasterizer ); } diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index e70eabd578..7b528fbaf3 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -38,7 +38,7 @@ void * softpipe_create_sampler_state(struct pipe_context *pipe, const struct pipe_sampler_state *sampler) { - struct pipe_sampler_state *state = malloc(sizeof(struct pipe_sampler_state)); + struct pipe_sampler_state *state = MALLOC( sizeof(struct pipe_sampler_state) ); memcpy(state, sampler, sizeof(struct pipe_sampler_state)); return state; } @@ -60,7 +60,7 @@ void softpipe_delete_sampler_state(struct pipe_context *pipe, void *sampler) { - free(sampler); + FREE( sampler ); } diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index bc343140ba..9d5529d8d6 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -172,7 +172,7 @@ z16_get_tile(struct pipe_surface *ps, const ushort *src = ((const ushort *) (ps->region->map + ps->offset)) + y * ps->region->pitch + x; - const float scale = 1.0 / 65535.0; + const float scale = 1.0f / 65535.0f; unsigned i, j; unsigned w0 = w; @@ -418,7 +418,7 @@ z32_get_tile(struct pipe_surface *ps, pRow[j * 4 + 0] = pRow[j * 4 + 1] = pRow[j * 4 + 2] = - pRow[j * 4 + 3] = src[j] * scale; + pRow[j * 4 + 3] = (float) (scale * src[j]); } src += ps->region->pitch; p += 4 * w0; @@ -452,7 +452,7 @@ s8z24_get_tile(struct pipe_surface *ps, pRow[j * 4 + 0] = pRow[j * 4 + 1] = pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (src[j] & 0xffffff) * scale; + pRow[j * 4 + 3] = (float) (scale * (src[j] & 0xffffff)); } src += ps->region->pitch; p += 4 * w0; diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c index 15447d6600..3ac7bfcb73 100644 --- a/src/mesa/pipe/softpipe/sp_tex_layout.c +++ b/src/mesa/pipe/softpipe/sp_tex_layout.c @@ -30,10 +30,8 @@ * Michel Dänzer */ -#include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" #include "sp_tex_layout.h" @@ -76,14 +74,14 @@ sp_miptree_set_level_info(struct pipe_mipmap_tree *mt, /* Not sure when this would happen, but anyway: */ if (mt->level[level].image_offset) { - free(mt->level[level].image_offset); + FREE( mt->level[level].image_offset ); mt->level[level].image_offset = NULL; } assert(nr_images); assert(!mt->level[level].image_offset); - mt->level[level].image_offset = (unsigned *) malloc(nr_images * sizeof(unsigned)); + mt->level[level].image_offset = (unsigned *) MALLOC( nr_images * sizeof(unsigned) ); mt->level[level].image_offset[0] = 0; } diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 385b8042c3..0b91ef0e37 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -39,7 +39,6 @@ #include "sp_tile_cache.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" #include "pipe/tgsi/exec/tgsi_exec.h" @@ -580,7 +579,7 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, height = sampler->texture->level[level0].height; } else { - width = height = 1.0; + width = height = 1; } assert(width > 0); @@ -697,7 +696,7 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, depth = sampler->texture->level[level0].depth; } else { - width = height = depth = 1.0; + width = height = depth = 1; } assert(width > 0); diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 1b600aadf1..473316b674 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -99,7 +99,7 @@ sp_create_tile_cache(void) struct softpipe_tile_cache *tc; uint pos; - tc = calloc(1, sizeof(*tc)); + tc = CALLOC_STRUCT( softpipe_tile_cache ); if (tc) { for (pos = 0; pos < NUM_ENTRIES; pos++) { tc->entries[pos].x = @@ -117,7 +117,7 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc) for (pos = 0; pos < NUM_ENTRIES; pos++) { assert(tc->entries[pos].x < 0); } - free(tc); + FREE( tc ); } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c index f740500dec..1df71efff7 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_dump.c @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #include "tgsi_platform.h" #include "tgsi_core.h" @@ -7,15 +34,6 @@ struct text_dump unsigned tabs; }; -static void -text_dump_write( - struct text_dump *dump, - const void *buffer, - unsigned size ) -{ - fwrite( buffer, size, 1, dump->file ); -} - static void text_dump_str( struct text_dump *dump, @@ -25,13 +43,13 @@ text_dump_str( size_t len = strlen( str ); for( i = 0; i < len; i++ ) { - text_dump_write( dump, &str[i], 1 ); + fprintf( dump->file, "%c", str[i] ); if( str[i] == '\n' ) { unsigned i; for( i = 0; i < dump->tabs; i++ ) { - text_dump_write( dump, " ", 4 ); + fprintf( dump->file, " " ); } } } @@ -1434,9 +1452,5 @@ tgsi_dump( TXT( "\ntgsi-dump end -------------------\n" ); tgsi_parse_free( &parse ); - - if (dump->file != stderr && - dump->file != stdout) - fclose( dump->file ); } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index e4f23d3eef..7607ded63a 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -133,10 +133,10 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) labels->count = 0; declarations = (struct tgsi_full_declaration *) - malloc(maxDeclarations * sizeof(struct tgsi_full_declaration)); + MALLOC( maxDeclarations * sizeof(struct tgsi_full_declaration) ); instructions = (struct tgsi_full_instruction *) - malloc(maxInstructions * sizeof(struct tgsi_full_instruction)); + MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) ); k = tgsi_parse_init( &parse, mach->Tokens ); if (k != TGSI_PARSE_OK) { @@ -153,10 +153,12 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) case TGSI_TOKEN_TYPE_DECLARATION: /* save expanded declaration */ if (numDeclarations == maxDeclarations) { - maxDeclarations += 10; - declarations = realloc(declarations, + declarations = REALLOC(declarations, maxDeclarations + * sizeof(struct tgsi_full_instruction), + (maxDeclarations + 10) * sizeof(struct tgsi_full_instruction)); + maxDeclarations += 10; } memcpy(declarations + numDeclarations, &parse.FullToken.FullInstruction, @@ -186,10 +188,12 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) /* save expanded instruction */ if (numInstructions == maxInstructions) { - maxInstructions += 10; - instructions = realloc(instructions, + instructions = REALLOC(instructions, maxInstructions + * sizeof(struct tgsi_full_instruction), + (maxInstructions + 10) * sizeof(struct tgsi_full_instruction)); + maxInstructions += 10; } memcpy(instructions + numInstructions, &parse.FullToken.FullInstruction, @@ -204,13 +208,13 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) tgsi_parse_free (&parse); if (mach->Declarations) { - free(mach->Declarations); + FREE( mach->Declarations ); } mach->Declarations = declarations; mach->NumDeclarations = numDeclarations; if (mach->Instructions) { - free(mach->Instructions); + FREE( mach->Instructions ); } mach->Instructions = instructions; mach->NumInstructions = numInstructions; @@ -242,7 +246,7 @@ tgsi_exec_machine_init( k = tgsi_parse_init (&parse, mach->Tokens); if (k != TGSI_PARSE_OK) { - printf("Problem parsing!\n"); + fprintf( stderr, "Problem parsing!\n" ); return; } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_parse.c b/src/mesa/pipe/tgsi/exec/tgsi_parse.c index 1096afffad..aee71decb3 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_parse.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_parse.c @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #include "tgsi_platform.h" #include "tgsi_core.h" @@ -13,7 +40,7 @@ tgsi_full_token_free( union tgsi_full_token *full_token ) { if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { - free( full_token->FullImmediate.u.Pointer ); + FREE( full_token->FullImmediate.u.Pointer ); } } @@ -123,7 +150,7 @@ tgsi_parse_token( switch (imm->Immediate.DataType) { case TGSI_IMM_FLOAT32: - imm->u.Pointer = malloc( + imm->u.Pointer = MALLOC( sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); for( i = 0; i < imm->Immediate.Size - 1; i++ ) { next_token( ctx, &imm->u.ImmediateFloat32[i] ); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index a71c5155c0..bc20d9ca05 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #include "tgsi_platform.h" #include "tgsi_core.h" #include "x86/rtasm/x86sse.h" @@ -465,7 +492,7 @@ emit_shufps( struct x86_function *func, struct x86_reg dst, struct x86_reg src, - unsigned shuf ) + unsigned char shuf ) { DUMP_RRI( "SHUFPS", dst, src, shuf ); sse_shufps( func, dst, src, shuf ); -- cgit v1.2.3 From 7de874ec2c7b9e3aff7f81b7e30045b45381fbad Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 29 Oct 2007 16:59:57 +0000 Subject: Make it compile under linux. Move pipe_region/surface_reference functions to pipe/p_inlines.h. Remove #include "p_util.h" from pipe/p_context.h. --- src/mesa/pipe/i915simple/i915_surface.c | 1 + src/mesa/pipe/p_context.h | 60 ------------------- src/mesa/pipe/p_inlines.h | 89 ++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_state_blend.c | 2 + src/mesa/pipe/softpipe/sp_state_rasterizer.c | 1 + src/mesa/pipe/softpipe/sp_state_sampler.c | 1 + src/mesa/pipe/softpipe/sp_surface.c | 1 + src/mesa/pipe/softpipe/sp_tex_layout.c | 1 + src/mesa/pipe/softpipe/sp_tex_sample.c | 2 +- src/mesa/pipe/softpipe/sp_tile_cache.c | 3 +- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 6 +- src/mesa/state_tracker/st_cb_drawpixels.c | 1 + src/mesa/state_tracker/st_cb_fbo.c | 1 + 13 files changed, 104 insertions(+), 65 deletions(-) create mode 100644 src/mesa/pipe/p_inlines.h (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index b4b5bd1ce5..afe15fb525 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -29,6 +29,7 @@ #include "i915_state.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/p_winsys.h" diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 3bd266674f..33d1878158 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -29,7 +29,6 @@ #define PIPE_CONTEXT_H #include "p_state.h" -#include "p_util.h" struct pipe_state_cache; /** @@ -246,63 +245,4 @@ struct pipe_context { unsigned flags ); }; - -/** - * Set 'ptr' to point to 'region' and update reference counting. - * The old thing pointed to, if any, will be unreferenced first. - * 'region' may be NULL. - */ -static INLINE void -pipe_region_reference(struct pipe_region **ptr, struct pipe_region *region) -{ - assert(ptr); - if (*ptr) { - /* unreference the old thing */ - struct pipe_region *oldReg = *ptr; - assert(oldReg->refcount > 0); - oldReg->refcount--; - if (oldReg->refcount == 0) { - /* free the old region */ - assert(oldReg->map_refcount == 0); - /* XXX dereference the region->buffer */ - FREE( oldReg ); - } - *ptr = NULL; - } - if (region) { - /* reference the new thing */ - region->refcount++; - *ptr = region; - } -} - - -/** - * \sa pipe_region_reference - */ -static INLINE void -pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) -{ - assert(ptr); - if (*ptr) { - /* unreference the old thing */ - struct pipe_surface *oldSurf = *ptr; - assert(oldSurf->refcount > 0); - oldSurf->refcount--; - if (oldSurf->refcount == 0) { - /* free the old region */ - pipe_region_reference(&oldSurf->region, NULL); - FREE( oldSurf ); - } - *ptr = NULL; - } - if (surf) { - /* reference the new thing */ - surf->refcount++; - *ptr = surf; - } -} - - #endif /* PIPE_CONTEXT_H */ - diff --git a/src/mesa/pipe/p_inlines.h b/src/mesa/pipe/p_inlines.h new file mode 100644 index 0000000000..ea666fa20a --- /dev/null +++ b/src/mesa/pipe/p_inlines.h @@ -0,0 +1,89 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_INLINES_H +#define P_INLINES_H + +#include "p_context.h" +//#include "p_util.h" + +/** + * Set 'ptr' to point to 'region' and update reference counting. + * The old thing pointed to, if any, will be unreferenced first. + * 'region' may be NULL. + */ +static INLINE void +pipe_region_reference(struct pipe_region **ptr, struct pipe_region *region) +{ + assert(ptr); + if (*ptr) { + /* unreference the old thing */ + struct pipe_region *oldReg = *ptr; + assert(oldReg->refcount > 0); + oldReg->refcount--; + if (oldReg->refcount == 0) { + /* free the old region */ + assert(oldReg->map_refcount == 0); + /* XXX dereference the region->buffer */ + FREE( oldReg ); + } + *ptr = NULL; + } + if (region) { + /* reference the new thing */ + region->refcount++; + *ptr = region; + } +} + +/** + * \sa pipe_region_reference + */ +static INLINE void +pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) +{ + assert(ptr); + if (*ptr) { + /* unreference the old thing */ + struct pipe_surface *oldSurf = *ptr; + assert(oldSurf->refcount > 0); + oldSurf->refcount--; + if (oldSurf->refcount == 0) { + /* free the old region */ + pipe_region_reference(&oldSurf->region, NULL); + FREE( oldSurf ); + } + *ptr = NULL; + } + if (surf) { + /* reference the new thing */ + surf->refcount++; + *ptr = surf; + } +} + +#endif /* P_INLINES_H */ diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index e9e6a55e0d..5ceec2513f 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -27,6 +27,8 @@ /* Authors: Keith Whitwell */ + +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_state.h" diff --git a/src/mesa/pipe/softpipe/sp_state_rasterizer.c b/src/mesa/pipe/softpipe/sp_state_rasterizer.c index a69e0d9012..ce8fa4f2b8 100644 --- a/src/mesa/pipe/softpipe/sp_state_rasterizer.c +++ b/src/mesa/pipe/softpipe/sp_state_rasterizer.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_state.h" #include "pipe/draw/draw_context.h" diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 7b528fbaf3..246a7d6eda 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -29,6 +29,7 @@ * Brian Paul */ +#include "pipe/p_util.h" #include "sp_context.h" #include "sp_state.h" #include "sp_tile_cache.h" diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 9d5529d8d6..eeaf98ce36 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -27,6 +27,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "sp_context.h" #include "sp_state.h" diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c index 3ac7bfcb73..8156b00301 100644 --- a/src/mesa/pipe/softpipe/sp_tex_layout.c +++ b/src/mesa/pipe/softpipe/sp_tex_layout.c @@ -32,6 +32,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "sp_tex_layout.h" diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 0b91ef0e37..e3e607d27c 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -32,13 +32,13 @@ * Brian Paul */ - #include "sp_context.h" #include "sp_surface.h" #include "sp_tex_sample.h" #include "sp_tile_cache.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "pipe/tgsi/exec/tgsi_exec.h" diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 473316b674..9542ec3765 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -32,7 +32,8 @@ * Brian Paul */ - +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "sp_context.h" #include "sp_surface.h" #include "sp_tile_cache.h" diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 7607ded63a..3494b21599 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -50,10 +50,10 @@ * Brian Paul */ -#include "tgsi_platform.h" -#include "tgsi_core.h" +#include "pipe/p_compiler.h" #include "pipe/p_state.h" - +#include "pipe/p_util.h" +#include "tgsi_core.h" #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 99e1e3ed25..3c57317415 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -48,6 +48,7 @@ #include "st_mesa_to_tgsi.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "shader/prog_instruction.h" diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 5512bdeed8..1ffd1a30b8 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -41,6 +41,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "st_context.h" #include "st_cb_fbo.h" -- cgit v1.2.3 From da9815e17f42b5f07beb47a08146e44c367c1ac2 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 29 Oct 2007 17:36:20 +0000 Subject: Add detailed comments. --- src/mesa/pipe/p_format.h | 84 ++++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 77 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index 1cab3c27b2..89d83a44f5 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -30,9 +30,31 @@ #include "p_compiler.h" +/** + * The PIPE_FORMAT is a 32-bit wide bitfield that encodes all the information + * needed to uniquely describe a pixel format. + */ + +/** + * Possible format layouts -- occupy first 2 bits. The interpretation of + * the remaining 30 bits depends on a particual format layout. + */ #define PIPE_FORMAT_LAYOUT_RGBAZS 0 #define PIPE_FORMAT_LAYOUT_YCBCR 1 +struct pipe_format_header +{ + uint layout : 2; /**< PIPE_FORMAT_LAYOUT_ */ + uint padding : 30; +}; + +/** + * RGBAZS Format Layout. + */ + +/** + * Format component selectors for RGBAZS layout. + */ #define PIPE_FORMAT_COMP_R 0 #define PIPE_FORMAT_COMP_G 1 #define PIPE_FORMAT_COMP_B 2 @@ -42,6 +64,9 @@ #define PIPE_FORMAT_COMP_Z 6 #define PIPE_FORMAT_COMP_S 7 +/** + * Format types for RGBAZS layout. + */ #define PIPE_FORMAT_TYPE_UNKNOWN 0 #define PIPE_FORMAT_TYPE_FLOAT 1 #define PIPE_FORMAT_TYPE_UNORM 2 @@ -49,14 +74,24 @@ #define PIPE_FORMAT_TYPE_USCALED 4 #define PIPE_FORMAT_TYPE_SSCALED 5 -struct pipe_format_header -{ - uint layout : 2; /**< PIPE_FORMAT_LAYOUT_ */ - uint padding : 30; -}; - /** - * RGBAZS Format Layout. + * In a nutshell, this bitfield contains instructions how to unpack + * a given format to a full-blown 4-component vector suitable for FS. + * Because the destination vector is assumed to be RGBA FLOAT, we + * need to know how to swizzle and expand components from the source + * vector. + * Let's take U_A1_R5_G5_B5 as an example. SwizzleX is A, sizeX + * is 1 bit and type is UNORM. So we take the most significant bit + * from source vector, convert 0 to 0.0 and 1 to 1.0 and save it + * in the last component of the destination RGBA component. + * Next, swizzleY is R, sizeY is 5 and type is UNORM. We normalize + * those 5 bits into [0.0; 1.0] range and put it into second + * component of the destination vector. Rinse and repeat for + * components Z and W. + * If any of size fields is zero, it means the source format contains + * less than four components. + * If any swizzle is 0 or 1, the corresponding destination component + * should be filled with 0.0 and 1.0, respectively. */ struct pipe_format_rgbazs { @@ -74,6 +109,9 @@ struct pipe_format_rgbazs uint padding : 1; }; +/** + * Helper macro to encode the above structure into a 32-bit value. + */ #define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP8, TYPE ) (\ (PIPE_FORMAT_LAYOUT_RGBAZS << 0) |\ ((SWZ) << 2) |\ @@ -84,23 +122,44 @@ struct pipe_format_rgbazs ((EXP8) << 26) |\ ((TYPE) << 28) ) +/** + * Helper macro to encode the swizzle part of the structure above. + */ #define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 0) | ((SWZY) << 3) | ((SWZZ) << 6) | ((SWZW) << 9)) +/** + * Shorthand macro for RGBAZS layout with uniform component sizes in 1-bit units. + */ #define _PIPE_FORMAT_RGBAZS_1U( SWZ, SIZE, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 0, TYPE ) +/** + * Shorthand macro for RGBAZS layout with non-uniform component sizes in 1-bit units. + */ #define _PIPE_FORMAT_RGBAZS_1N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE ) +/** + * Shorthand macro for RGBAZS layout with uniform component sizes in 8-bit units. + */ #define _PIPE_FORMAT_RGBAZS_8U( SWZ, SIZE, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 1, TYPE ) +/** + * Shorthand macro for RGBAZS layout with non-uniform component sizes in 8-bit units. + */ #define _PIPE_FORMAT_RGBAZS_8N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) +/** + * Shorthand macro for RGBAZS layout with non-uniform component sizes in 64-bit units. + */ #define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 2, TYPE ) +/** + * Shorthand macro for common format swizzles. + */ #define _PIPE_FORMAT_R000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_RG00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_A ) #define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) @@ -115,6 +174,14 @@ struct pipe_format_rgbazs #define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +/** + * YCBCR Format Layout. + */ + +/** + * This bitfields is simple. It only contains a flag that indicates whether the + * format is reversed or not. + */ struct pipe_format_ycbcr { uint layout : 2; /**< PIPE_FORMAT_LAYOUT_YCBCR */ @@ -122,6 +189,9 @@ struct pipe_format_ycbcr uint padding : 29; }; +/** + * Helper macro to encode the above structure into a 32-bit value. + */ #define _PIPE_FORMAT_YCBCR( REV ) (\ (PIPE_FORMAT_LAYOUT_YCBCR << 0) |\ ((REV) << 2) ) -- cgit v1.2.3 From f33ced441c934b625b9c126d054439855e822215 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 29 Oct 2007 17:36:37 +0000 Subject: Update intel_is_format_supported. --- .../dri/intel_winsys/intel_winsys_softpipe.c | 22 ++++++++++------------ 1 file changed, 10 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c index 1660046f13..f1358fef1a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c @@ -45,22 +45,20 @@ struct intel_softpipe_winsys { /** * Return list of surface formats supported by this driver. */ -static const unsigned * -intel_is_format_supported(struct softpipe_winsys *sws, - unsigned *numFormats) +static boolean +intel_is_format_supported(struct softpipe_winsys *sws, uint format) { - static const GLuint formats[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_S8_Z24, - }; - - *numFormats = sizeof(formats) / sizeof(formats[0]); - return formats; + switch(format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_U_R5_G6_B5: + case PIPE_FORMAT_S8_Z24: + return TRUE; + default: + return FALSE; + } } - struct pipe_context * intel_create_softpipe( struct intel_context *intel ) { -- cgit v1.2.3 From 918ea5168baaecdf70f5ea37e5815cacf9558163 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 29 Oct 2007 19:50:10 +0000 Subject: Rename 'mms-config.' to 'mms.config'. It looks like Windows does not like filenames ending with a dot, in effect renaming it to 'mms-config'. --- src/descrip.mms | 2 +- src/glu/mesa/descrip.mms | 2 +- src/glu/sgi/descrip.mms | 2 +- src/glut/glx/descrip.mms | 2 +- src/mesa/drivers/common/descrip.mms | 2 +- src/mesa/drivers/osmesa/descrip.mms | 2 +- src/mesa/drivers/x11/descrip.mms | 2 +- src/mesa/glapi/descrip.mms | 2 +- src/mesa/main/descrip.mms | 2 +- src/mesa/math/descrip.mms | 2 +- src/mesa/shader/descrip.mms | 2 +- src/mesa/shader/grammar/descrip.mms | 2 +- src/mesa/shader/slang/descrip.mms | 2 +- src/mesa/swrast/descrip.mms | 2 +- src/mesa/swrast_setup/descrip.mms | 2 +- src/mesa/tnl/descrip.mms | 2 +- src/mesa/vbo/descrip.mms | 2 +- 17 files changed, 17 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/descrip.mms b/src/descrip.mms index 71b8ea16ac..79c7d98d25 100644 --- a/src/descrip.mms +++ b/src/descrip.mms @@ -1,7 +1,7 @@ # Makefile for Mesa for VMS # contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl -.include [-]mms-config. +.include [-]mms.config all : set default [.mesa] diff --git a/src/glu/mesa/descrip.mms b/src/glu/mesa/descrip.mms index 2b3f64d8bc..12eb6a437a 100644 --- a/src/glu/mesa/descrip.mms +++ b/src/glu/mesa/descrip.mms @@ -4,7 +4,7 @@ .first define gl [-.include.gl] -.include [-]mms-config. +.include [-]mms.config ##### MACROS ##### diff --git a/src/glu/sgi/descrip.mms b/src/glu/sgi/descrip.mms index 5abc8b2e04..49d618e6c9 100644 --- a/src/glu/sgi/descrip.mms +++ b/src/glu/sgi/descrip.mms @@ -4,7 +4,7 @@ .first define gl [---.include.gl] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### diff --git a/src/glut/glx/descrip.mms b/src/glut/glx/descrip.mms index 2e858309b6..358b417511 100644 --- a/src/glut/glx/descrip.mms +++ b/src/glut/glx/descrip.mms @@ -4,7 +4,7 @@ .first define gl [---.include.gl] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### GLUT_MAJOR = 3 diff --git a/src/mesa/drivers/common/descrip.mms b/src/mesa/drivers/common/descrip.mms index 3cb856d12c..c2c119db7f 100644 --- a/src/mesa/drivers/common/descrip.mms +++ b/src/mesa/drivers/common/descrip.mms @@ -8,7 +8,7 @@ define tnl [--.tnl] define swrast [--.swrast] -.include [----]mms-config. +.include [----]mms.config ##### MACROS ##### diff --git a/src/mesa/drivers/osmesa/descrip.mms b/src/mesa/drivers/osmesa/descrip.mms index 4035b24e4e..5be194bcee 100644 --- a/src/mesa/drivers/osmesa/descrip.mms +++ b/src/mesa/drivers/osmesa/descrip.mms @@ -11,7 +11,7 @@ define array_cache [--.array_cache] define drivers [-] -.include [----]mms-config. +.include [----]mms.config ##### MACROS ##### diff --git a/src/mesa/drivers/x11/descrip.mms b/src/mesa/drivers/x11/descrip.mms index 6c6184b2c3..f181707a14 100644 --- a/src/mesa/drivers/x11/descrip.mms +++ b/src/mesa/drivers/x11/descrip.mms @@ -11,7 +11,7 @@ define array_cache [--.array_cache] define drivers [-] -.include [----]mms-config. +.include [----]mms.config ##### MACROS ##### diff --git a/src/mesa/glapi/descrip.mms b/src/mesa/glapi/descrip.mms index f17e5329b6..16bf6387e8 100644 --- a/src/mesa/glapi/descrip.mms +++ b/src/mesa/glapi/descrip.mms @@ -5,7 +5,7 @@ .first define gl [---.include.gl] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### diff --git a/src/mesa/main/descrip.mms b/src/mesa/main/descrip.mms index d09c57b321..2bd388be89 100644 --- a/src/mesa/main/descrip.mms +++ b/src/mesa/main/descrip.mms @@ -7,7 +7,7 @@ define math [-.math] define shader [-.shader] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### diff --git a/src/mesa/math/descrip.mms b/src/mesa/math/descrip.mms index a3f20c2f25..5b9e9915f8 100644 --- a/src/mesa/math/descrip.mms +++ b/src/mesa/math/descrip.mms @@ -6,7 +6,7 @@ define gl [---.include.gl] define math [-.math] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### diff --git a/src/mesa/shader/descrip.mms b/src/mesa/shader/descrip.mms index d70cec3830..f5d04cfa78 100644 --- a/src/mesa/shader/descrip.mms +++ b/src/mesa/shader/descrip.mms @@ -7,7 +7,7 @@ define swrast [-.swrast] define array_cache [-.array_cache] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### diff --git a/src/mesa/shader/grammar/descrip.mms b/src/mesa/shader/grammar/descrip.mms index f7fbee96bc..cff53ee872 100644 --- a/src/mesa/shader/grammar/descrip.mms +++ b/src/mesa/shader/grammar/descrip.mms @@ -8,7 +8,7 @@ define swrast [--.swrast] define array_cache [--.array_cache] -.include [----]mms-config. +.include [----]mms.config ##### MACROS ##### diff --git a/src/mesa/shader/slang/descrip.mms b/src/mesa/shader/slang/descrip.mms index c86763718a..8b9cd822b8 100644 --- a/src/mesa/shader/slang/descrip.mms +++ b/src/mesa/shader/slang/descrip.mms @@ -8,7 +8,7 @@ define swrast [--.swrast] define array_cache [--.array_cache] -.include [----]mms-config. +.include [----]mms.config ##### MACROS ##### diff --git a/src/mesa/swrast/descrip.mms b/src/mesa/swrast/descrip.mms index 4d49673b5d..4d446600ba 100644 --- a/src/mesa/swrast/descrip.mms +++ b/src/mesa/swrast/descrip.mms @@ -8,7 +8,7 @@ define swrast [-.swrast] define array_cache [-.array_cache] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### diff --git a/src/mesa/swrast_setup/descrip.mms b/src/mesa/swrast_setup/descrip.mms index e5e48afb3d..0ab81c07d3 100644 --- a/src/mesa/swrast_setup/descrip.mms +++ b/src/mesa/swrast_setup/descrip.mms @@ -9,7 +9,7 @@ define swrast [-.swrast] define array_cache [-.array_cache] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### diff --git a/src/mesa/tnl/descrip.mms b/src/mesa/tnl/descrip.mms index 91e32bf978..a9aed89f1d 100644 --- a/src/mesa/tnl/descrip.mms +++ b/src/mesa/tnl/descrip.mms @@ -8,7 +8,7 @@ define shader [-.shader] define array_cache [-.array_cache] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### diff --git a/src/mesa/vbo/descrip.mms b/src/mesa/vbo/descrip.mms index 4ab22e4005..e00b6703aa 100644 --- a/src/mesa/vbo/descrip.mms +++ b/src/mesa/vbo/descrip.mms @@ -11,7 +11,7 @@ define swrast [-.swrast] define swrast_setup [-.swrast_setup] -.include [---]mms-config. +.include [---]mms.config ##### MACROS ##### -- cgit v1.2.3 From 83ce6c51d7189094cf2a13fdcc0882a051a3bd41 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 11:23:02 -0600 Subject: Refactor _mesa_UpdateTexEnvProgram() Will be replaced by _mesa_get_fixed_func_fragment_program(). --- src/mesa/main/texenvprogram.c | 68 +++++++++++++++++++++++-------------------- src/mesa/main/texenvprogram.h | 3 ++ 2 files changed, 39 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index 21a290402f..d8f3314c70 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -1234,6 +1234,39 @@ static GLuint hash_key( const struct state_key *key ) } +/** + * Return a fragment program which implements the current + * fixed-function texture, fog and color-sum operations. + */ +struct gl_fragment_program * +_mesa_get_fixed_func_fragment_program(GLcontext *ctx) +{ + struct gl_fragment_program *prog; + struct state_key key; + GLuint hash; + + make_state_key(ctx, &key); + hash = hash_key(&key); + + prog = search_cache(&ctx->Texture.env_fp_cache, hash, &key, sizeof(key)); + + if (!prog) { + if (0) + _mesa_printf("Building new texenv proggy for key %x\n", hash); + + prog = (struct gl_fragment_program *) + ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + + create_new_program(ctx, &key, prog); + + cache_item(&ctx->Texture.env_fp_cache, hash, &key, prog); + } + + return prog; +} + + + /** * If _MaintainTexEnvProgram is set we'll generate a fragment program that * implements the current texture env/combine mode. @@ -1242,8 +1275,6 @@ static GLuint hash_key( const struct state_key *key ) void _mesa_UpdateTexEnvProgram( GLcontext *ctx ) { - struct state_key key; - GLuint hash; const struct gl_fragment_program *prev = ctx->FragmentProgram._Current; ASSERT(ctx->FragmentProgram._MaintainTexEnvProgram); @@ -1252,38 +1283,11 @@ _mesa_UpdateTexEnvProgram( GLcontext *ctx ) if (!ctx->FragmentProgram._Enabled && (!ctx->Shader.CurrentProgram || !ctx->Shader.CurrentProgram->FragmentProgram) ) { - make_state_key(ctx, &key); - hash = hash_key(&key); - - ctx->FragmentProgram._Current = - ctx->FragmentProgram._TexEnvProgram = - search_cache(&ctx->Texture.env_fp_cache, hash, &key, sizeof(key)); - - if (!ctx->FragmentProgram._TexEnvProgram) { - if (0) - _mesa_printf("Building new texenv proggy for key %x\n", hash); - /* create new tex env program */ - ctx->FragmentProgram._Current = - ctx->FragmentProgram._TexEnvProgram = - (struct gl_fragment_program *) - ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); - - create_new_program(ctx, &key, ctx->FragmentProgram._TexEnvProgram); - - cache_item(&ctx->Texture.env_fp_cache, hash, &key, - ctx->FragmentProgram._TexEnvProgram); - } - else { - if (0) - _mesa_printf("Found existing texenv program for key %x\n", hash); - } + ctx->FragmentProgram._Current + = ctx->FragmentProgram._TexEnvProgram + = _mesa_get_fixed_func_fragment_program(ctx); } -#if 0 - else { - ctx->FragmentProgram._Current = ctx->FragmentProgram.Current; - } -#endif /* Tell the driver about the change. Could define a new target for * this? diff --git a/src/mesa/main/texenvprogram.h b/src/mesa/main/texenvprogram.h index 6f017767c8..ac73910cb5 100644 --- a/src/mesa/main/texenvprogram.h +++ b/src/mesa/main/texenvprogram.h @@ -34,6 +34,9 @@ #include "mtypes.h" +extern struct gl_fragment_program * +_mesa_get_fixed_func_fragment_program(GLcontext *ctx); + extern void _mesa_UpdateTexEnvProgram( GLcontext *ctx ); extern void _mesa_TexEnvProgramCacheInit( GLcontext *ctx ); extern void _mesa_TexEnvProgramCacheDestroy( GLcontext *ctx ); -- cgit v1.2.3 From 27153bf02dab57d11565fa7730de4767618ce62d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 11:30:09 -0600 Subject: Refactor _tnl_UpdateFixedFunctionProgram(). New _mesa_get_fixed_func_vertex_program() function... --- src/mesa/tnl/t_vp_build.c | 81 ++++++++++++++++++++++++++++------------------- src/mesa/tnl/t_vp_build.h | 11 +++++-- 2 files changed, 57 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index dc2d5e0065..86c6ccc1c0 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -1534,49 +1534,66 @@ static GLuint hash_key( struct state_key *key ) return hash; } -void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ) + +/** + * Return a vertex program which implements the current fixed-function + * transform/lighting/texgen operations. + * XXX move this into core mesa (main/) + */ +struct gl_vertex_program * +_mesa_get_fixed_func_vertex_program(GLcontext *ctx) { TNLcontext *tnl = TNL_CONTEXT(ctx); + struct gl_vertex_program *prog; struct state_key *key; GLuint hash; - const struct gl_vertex_program *prev = ctx->VertexProgram._Current; - if (!ctx->VertexProgram._Current || - ctx->VertexProgram._Current == ctx->VertexProgram._TnlProgram) { - /* Grab all the relevent state and put it in a single structure: - */ - key = make_state_key(ctx); - hash = hash_key(key); + /* Grab all the relevent state and put it in a single structure: + */ + key = make_state_key(ctx); + hash = hash_key(key); - /* Look for an already-prepared program for this state: - */ - ctx->VertexProgram._TnlProgram = (struct gl_vertex_program *) - search_cache( tnl->vp_cache, hash, key, sizeof(*key) ); + /* Look for an already-prepared program for this state: + */ + prog = (struct gl_vertex_program *) + search_cache( tnl->vp_cache, hash, key, sizeof(*key) ); - /* OK, we'll have to build a new one: - */ - if (!ctx->VertexProgram._TnlProgram) { - if (0) - _mesa_printf("Build new TNL program\n"); + if (!prog) { + /* OK, we'll have to build a new one */ + if (0) + _mesa_printf("Build new TNL program\n"); - ctx->VertexProgram._TnlProgram = (struct gl_vertex_program *) - ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); + prog = (struct gl_vertex_program *) + ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); - create_new_program( key, ctx->VertexProgram._TnlProgram, - ctx->Const.VertexProgram.MaxTemps ); + create_new_program( key, prog, + ctx->Const.VertexProgram.MaxTemps ); - if (ctx->Driver.ProgramStringNotify) - ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB, - &ctx->VertexProgram._TnlProgram->Base ); +#if 0 + if (ctx->Driver.ProgramStringNotify) + ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB, + &prog->Base ); +#endif + cache_item(tnl->vp_cache, hash, key, prog); + } + else { + /* use cached program */ + _mesa_free(key); + } - cache_item(tnl->vp_cache, hash, key, ctx->VertexProgram._TnlProgram ); - } - else { - FREE(key); - if (0) - _mesa_printf("Found existing TNL program for key %x\n", hash); - } - ctx->VertexProgram._Current = ctx->VertexProgram._TnlProgram; + return prog; +} + + +void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ) +{ + const struct gl_vertex_program *prev = ctx->VertexProgram._Current; + + if (!ctx->VertexProgram._Current || + ctx->VertexProgram._Current == ctx->VertexProgram._TnlProgram) { + ctx->VertexProgram._Current + = ctx->VertexProgram._TnlProgram + = _mesa_get_fixed_func_vertex_program(ctx); } /* Tell the driver about the change. Could define a new target for diff --git a/src/mesa/tnl/t_vp_build.h b/src/mesa/tnl/t_vp_build.h index 5e22fcf8c4..034701d8c4 100644 --- a/src/mesa/tnl/t_vp_build.h +++ b/src/mesa/tnl/t_vp_build.h @@ -17,24 +17,29 @@ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * TUNGSTEN GRAPHICS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#ifndef _T_ARB_BUILD_H -#define _T_ARB_BUILD_H +#ifndef T_VP_BUILD_H +#define T_VP_BUILD_H #include "mtypes.h" #define TNL_FIXED_FUNCTION_STATE_FLAGS (_NEW_PROGRAM | \ _NEW_LIGHT | \ _NEW_TEXTURE | \ + _NEW_TEXTURE_MATRIX | \ _NEW_TRANSFORM | \ _NEW_FOG | \ _NEW_POINT) +extern struct gl_vertex_program * +_mesa_get_fixed_func_vertex_program(GLcontext *ctx); + extern void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ); extern void _tnl_ProgramCacheInit( GLcontext *ctx ); -- cgit v1.2.3 From 783cedcdc14feda9c8ad0ca754752254bfa589eb Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 11:54:13 -0600 Subject: simplify getting of current frag prog --- src/mesa/tnl/t_vp_build.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index 86c6ccc1c0..056eb0db8e 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -122,12 +122,7 @@ static struct state_key *make_state_key( GLcontext *ctx ) struct state_key *key = CALLOC_STRUCT(state_key); GLuint i; - if (ctx->Shader.CurrentProgram && - ctx->Shader.CurrentProgram->LinkStatus && - ctx->Shader.CurrentProgram->FragmentProgram) - fp = ctx->Shader.CurrentProgram->FragmentProgram; - else - fp = ctx->FragmentProgram._Current; + fp = ctx->FragmentProgram._Current; /* This now relies on texenvprogram.c being active: */ -- cgit v1.2.3 From 9946012949ad294b2255f914fee4c80cff5e0040 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 12:18:58 -0600 Subject: refactoring to begin removing dependency on tnl context --- src/mesa/tnl/t_vp_build.c | 39 +++++++++++++++++++++++++++++++++------ 1 file changed, 33 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index 056eb0db8e..2b7d8eebe0 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -114,10 +114,40 @@ static GLuint translate_texgen( GLboolean enabled, GLenum mode ) } } -static struct state_key *make_state_key( GLcontext *ctx ) + +/** + * Returns bitmask of flags indicating which materials are set per-vertex + * in the current VB. + * XXX get these from the VBO... + */ +static GLbitfield +tnl_get_per_vertex_materials(GLcontext *ctx) { TNLcontext *tnl = TNL_CONTEXT(ctx); struct vertex_buffer *VB = &tnl->vb; + GLuint i; + GLbitfield mask = 0x0; + + for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) + if (VB->AttribPtr[i] && VB->AttribPtr[i]->stride) + mask |= 1 << (i - _TNL_FIRST_MAT); + + return mask; +} + +/** + * Should fog be computed per-vertex? + */ +static GLbitfield +tnl_get_per_vertex_fog(GLcontext *ctx) +{ + TNLcontext *tnl = TNL_CONTEXT(ctx); + return tnl->_DoVertexFog; +} + + +static struct state_key *make_state_key( GLcontext *ctx ) +{ const struct gl_fragment_program *fp; struct state_key *key = CALLOC_STRUCT(state_key); GLuint i; @@ -152,9 +182,7 @@ static struct state_key *make_state_key( GLcontext *ctx ) key->light_color_material_mask = ctx->Light.ColorMaterialBitmask; } - for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) - if (VB->AttribPtr[i] && VB->AttribPtr[i]->stride) - key->light_material_mask |= 1<<(i-_TNL_ATTRIB_MAT_FRONT_AMBIENT); + key->light_material_mask = tnl_get_per_vertex_materials(ctx); for (i = 0; i < MAX_LIGHTS; i++) { struct gl_light *light = &ctx->Light.Light[i]; @@ -187,8 +215,7 @@ static struct state_key *make_state_key( GLcontext *ctx ) if (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT) key->fog_source_is_depth = 1; - if (tnl->_DoVertexFog) - key->tnl_do_vertex_fog = 1; + key->tnl_do_vertex_fog = tnl_get_per_vertex_fog(ctx); if (ctx->Point._Attenuated) key->point_attenuated = 1; -- cgit v1.2.3 From f18d4e058ed979c6e42e868c7febde4fa62c5810 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 12:25:46 -0600 Subject: Remove ctx field from texenvprog_cache --- src/mesa/main/mtypes.h | 1 - src/mesa/main/texenvprogram.c | 23 +++++++++++------------ 2 files changed, 11 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 8adc4e3373..44587ae962 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -1566,7 +1566,6 @@ struct texenvprog_cache_item { struct texenvprog_cache { struct texenvprog_cache_item **items; GLuint size, n_items; - GLcontext *ctx; }; /** diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index d8f3314c70..cf92503c34 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -399,9 +399,9 @@ static struct ureg get_tex_temp( struct texenv_fragment_program *p ) } -static void release_temps( struct texenv_fragment_program *p ) +static void release_temps(GLcontext *ctx, struct texenv_fragment_program *p ) { - GLuint max_temp = p->ctx->Const.FragmentProgram.MaxTemps; + GLuint max_temp = ctx->Const.FragmentProgram.MaxTemps; /* KW: To support tex_env_crossbar, don't release the registers in * temps_output. @@ -1037,7 +1037,7 @@ create_new_program(GLcontext *ctx, struct state_key *key, p.one = undef; p.last_tex_stage = 0; - release_temps(&p); + release_temps(ctx, &p); if (key->enabled_units) { /* First pass - to support texture_env_crossbar, first identify @@ -1055,7 +1055,7 @@ create_new_program(GLcontext *ctx, struct state_key *key, for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) if (key->enabled_units & (1<size = size; } -static void clear_cache( struct texenvprog_cache *cache ) +static void clear_cache( GLcontext *ctx, struct texenvprog_cache *cache ) { struct texenvprog_cache_item *c, *next; GLuint i; @@ -1178,8 +1178,7 @@ static void clear_cache( struct texenvprog_cache *cache ) for (c = cache->items[i]; c; c = next) { next = c->next; _mesa_free(c->key); - cache->ctx->Driver.DeleteProgram(cache->ctx, - (struct gl_program *) c->data); + ctx->Driver.DeleteProgram(ctx, (struct gl_program *) c->data); _mesa_free(c); } cache->items[i] = NULL; @@ -1190,7 +1189,8 @@ static void clear_cache( struct texenvprog_cache *cache ) } -static void cache_item( struct texenvprog_cache *cache, +static void cache_item( GLcontext *ctx, + struct texenvprog_cache *cache, GLuint hash, const struct state_key *key, void *data ) @@ -1208,7 +1208,7 @@ static void cache_item( struct texenvprog_cache *cache, if (cache->size < 1000) rehash(cache); else - clear_cache(cache); + clear_cache(ctx, cache); } cache->n_items++; @@ -1259,7 +1259,7 @@ _mesa_get_fixed_func_fragment_program(GLcontext *ctx) create_new_program(ctx, &key, prog); - cache_item(&ctx->Texture.env_fp_cache, hash, &key, prog); + cache_item(ctx, &ctx->Texture.env_fp_cache, hash, &key, prog); } return prog; @@ -1301,7 +1301,6 @@ _mesa_UpdateTexEnvProgram( GLcontext *ctx ) void _mesa_TexEnvProgramCacheInit( GLcontext *ctx ) { - ctx->Texture.env_fp_cache.ctx = ctx; ctx->Texture.env_fp_cache.size = 17; ctx->Texture.env_fp_cache.n_items = 0; ctx->Texture.env_fp_cache.items = (struct texenvprog_cache_item **) @@ -1312,6 +1311,6 @@ void _mesa_TexEnvProgramCacheInit( GLcontext *ctx ) void _mesa_TexEnvProgramCacheDestroy( GLcontext *ctx ) { - clear_cache(&ctx->Texture.env_fp_cache); + clear_cache(ctx, &ctx->Texture.env_fp_cache); _mesa_free(ctx->Texture.env_fp_cache.items); } -- cgit v1.2.3 From 1553eba50c7a2577bce069a3400a29e4d49c0f31 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 14:13:57 -0600 Subject: Rewrite update_program() to use _mesa_get_fixed_func_fragment/vertex_program(). --- src/mesa/main/state.c | 126 +++++++++++++++++++++++++++++--------------------- 1 file changed, 74 insertions(+), 52 deletions(-) (limited to 'src') diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 444f227760..41c657e9fd 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -92,6 +92,7 @@ #include "shader/program.h" #include "texenvprogram.h" #endif +#include "tnl/t_vp_build.h" #if FEATURE_ARB_shader_objects #include "shaders.h" #endif @@ -938,13 +939,12 @@ update_arrays( GLcontext *ctx ) } -/** - * Update derived vertex/fragment program state. - */ static void update_program(GLcontext *ctx) { const struct gl_shader_program *shProg = ctx->Shader.CurrentProgram; + const struct gl_vertex_program *prevVP = ctx->VertexProgram._Current; + const struct gl_fragment_program *prevFP = ctx->FragmentProgram._Current; /* These _Enabled flags indicate if the program is enabled AND valid. */ ctx->VertexProgram._Enabled = ctx->VertexProgram.Enabled @@ -956,64 +956,83 @@ update_program(GLcontext *ctx) /* * Set the ctx->VertexProgram._Current and ctx->FragmentProgram._Current - * pointers to the programs that should be enabled/used. + * pointers to the programs that should be enabled/used. These will only + * be NULL if we need to use the fixed-function code. * * These programs may come from several sources. The priority is as * follows: * 1. OpenGL 2.0/ARB vertex/fragment shaders * 2. ARB/NV vertex/fragment programs * 3. Programs derived from fixed-function state. + * + * Note: it's possible for a vertex shader to get used with a fragment + * program (and vice versa) here, but in practice that shouldn't ever + * come up, or matter. */ - ctx->FragmentProgram._Current = NULL; - - if (shProg && shProg->LinkStatus) { - /* Use shader programs */ - /* XXX this isn't quite right, since we may have either a vertex - * _or_ fragment shader (not always both). - */ - ctx->VertexProgram._Current = shProg->VertexProgram; + /** + ** Fragment program + **/ +#if 1 + /* XXX get rid of this someday? */ + ctx->FragmentProgram._Active = GL_FALSE; +#endif + if (shProg && shProg->LinkStatus && shProg->FragmentProgram) { + /* user-defined fragment shader */ ctx->FragmentProgram._Current = shProg->FragmentProgram; } + else if (ctx->FragmentProgram._Enabled) { + /* use user-defined fragment program */ + ctx->FragmentProgram._Current = ctx->FragmentProgram.Current; + } + else if (ctx->FragmentProgram._MaintainTexEnvProgram) { + /* fragment program generated from fixed-function state */ + ctx->FragmentProgram._Current = _mesa_get_fixed_func_fragment_program(ctx); + ctx->FragmentProgram._TexEnvProgram = ctx->FragmentProgram._Current; + + /* XXX get rid of this confusing stuff someday? */ + ctx->FragmentProgram._Active = ctx->FragmentProgram._Enabled; + if (ctx->FragmentProgram._UseTexEnvProgram) + ctx->FragmentProgram._Active = GL_TRUE; + } else { - if (ctx->VertexProgram._Enabled) { - /* use user-defined vertex program */ - ctx->VertexProgram._Current = ctx->VertexProgram.Current; - } - else if (ctx->VertexProgram._MaintainTnlProgram) { - /* Use vertex program generated from fixed-function state. - * The _Current pointer will get set in - * _tnl_UpdateFixedFunctionProgram() later if appropriate. - */ - ctx->VertexProgram._Current = NULL; - } - else { - /* no vertex program */ - ctx->VertexProgram._Current = NULL; - } + /* no fragment program */ + ctx->FragmentProgram._Current = NULL; + } - if (ctx->FragmentProgram._Enabled) { - /* use user-defined vertex program */ - ctx->FragmentProgram._Current = ctx->FragmentProgram.Current; - } - else if (ctx->FragmentProgram._MaintainTexEnvProgram) { - /* Use fragment program generated from fixed-function state. - * The _Current pointer will get set in _mesa_UpdateTexEnvProgram() - * later if appropriate. - */ - ctx->FragmentProgram._Current = NULL; - } - else { - /* no fragment program */ - ctx->FragmentProgram._Current = NULL; - } + if (ctx->FragmentProgram._Current != prevFP && ctx->Driver.BindProgram) { + ctx->Driver.BindProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, + (struct gl_program *) ctx->FragmentProgram._Current); } - ctx->FragmentProgram._Active = ctx->FragmentProgram._Enabled; - if (ctx->FragmentProgram._MaintainTexEnvProgram && - !ctx->FragmentProgram._Enabled) { - if (ctx->FragmentProgram._UseTexEnvProgram) - ctx->FragmentProgram._Active = GL_TRUE; + /** + ** Vertex program + **/ +#if 1 + /* XXX get rid of this someday? */ + ctx->VertexProgram._TnlProgram = NULL; +#endif + if (shProg && shProg->LinkStatus && shProg->VertexProgram) { + /* user-defined vertex shader */ + ctx->VertexProgram._Current = shProg->VertexProgram; + } + else if (ctx->VertexProgram._Enabled) { + /* use user-defined vertex program */ + ctx->VertexProgram._Current = ctx->VertexProgram.Current; + } + else if (ctx->VertexProgram._MaintainTnlProgram) { + /* vertex program generated from fixed-function state */ + ctx->VertexProgram._Current = _mesa_get_fixed_func_vertex_program(ctx); + ctx->VertexProgram._TnlProgram = ctx->VertexProgram._Current; + } + else { + /* no vertex program / used fixed-function code */ + ctx->VertexProgram._Current = NULL; + } + + if (ctx->VertexProgram._Current != prevVP && ctx->Driver.BindProgram) { + ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB, + (struct gl_program *) ctx->VertexProgram._Current); } } @@ -1140,13 +1159,11 @@ void _mesa_update_state_locked( GLcontext *ctx ) { GLbitfield new_state = ctx->NewState; + GLbitfield prog_flags = _NEW_PROGRAM; if (MESA_VERBOSE & VERBOSE_STATE) _mesa_print_state("_mesa_update_state", new_state); - if (new_state & _NEW_PROGRAM) - update_program( ctx ); - if (new_state & (_NEW_MODELVIEW|_NEW_PROJECTION)) _mesa_update_modelview_project( ctx, new_state ); @@ -1182,9 +1199,14 @@ _mesa_update_state_locked( GLcontext *ctx ) update_tricaps( ctx, new_state ); if (ctx->FragmentProgram._MaintainTexEnvProgram) { - if (new_state & (_NEW_TEXTURE | _DD_NEW_SEPARATE_SPECULAR | _NEW_FOG)) - _mesa_UpdateTexEnvProgram(ctx); + prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR); } + if (ctx->VertexProgram._MaintainTnlProgram) { + prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _NEW_LIGHT); + } + if (new_state & prog_flags) + update_program( ctx ); + /* ctx->_NeedEyeCoords is now up to date. * -- cgit v1.2.3 From 64ffc9540ba9c0208f8eca98996d03d632c33064 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 15:07:23 -0600 Subject: fix memcpy bugs --- src/mesa/pipe/softpipe/sp_surface.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index eeaf98ce36..c18fe56f89 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -526,7 +526,7 @@ softpipe_get_tile(struct pipe_context *pipe, pDest = (ubyte *) p; for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w0 * cpp); + memcpy(pDest, pSrc, w * cpp); pDest += dst_stride; pSrc += ps->region->pitch * cpp; } @@ -560,7 +560,7 @@ softpipe_put_tile(struct pipe_context *pipe, pDest = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w0 * cpp); + memcpy(pDest, pSrc, w * cpp); pDest += ps->region->pitch * cpp; pSrc += src_stride; } -- cgit v1.2.3 From 8b2f997af572dd3aca00c4888adf2b0a60656331 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 15:08:43 -0600 Subject: minor code simplification --- src/mesa/pipe/softpipe/sp_surface.c | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index c18fe56f89..3aa72ef569 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -509,19 +509,18 @@ softpipe_get_tile(struct pipe_context *pipe, void *p, int dst_stride) { const uint cpp = ps->region->cpp; - const uint w0 = w; const ubyte *pSrc; ubyte *pDest; uint i; assert(ps->region->map); - CLIP_TILE; - if (dst_stride == 0) { - dst_stride = w0 * cpp; + dst_stride = w * cpp; } + CLIP_TILE; + pSrc = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; pDest = (ubyte *) p; @@ -543,19 +542,18 @@ softpipe_put_tile(struct pipe_context *pipe, const void *p, int src_stride) { const uint cpp = ps->region->cpp; - const uint w0 = w; const ubyte *pSrc; ubyte *pDest; uint i; assert(ps->region->map); - CLIP_TILE; - if (src_stride == 0) { - src_stride = w0 * cpp; + src_stride = w * cpp; } + CLIP_TILE; + pSrc = (const ubyte *) p; pDest = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; -- cgit v1.2.3 From b0f3b5910ebd0737600ab7b1fdc135d74f2617f4 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 29 Oct 2007 10:59:24 -0400 Subject: Enable immediates in TGSI and work with them in LLVM code. Enables immediates by default in the TGSI translation code and adds code handling it in llvm tgsi translation. --- src/mesa/pipe/llvm/llvmtgsi.cpp | 20 ++++++++++++++++++-- src/mesa/pipe/llvm/storage.cpp | 15 +++++++++++++++ src/mesa/pipe/llvm/storage.h | 4 ++++ src/mesa/state_tracker/st_mesa_to_tgsi.c | 14 -------------- 4 files changed, 37 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 4611ac1766..ec38c695b9 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -164,9 +164,23 @@ translate_declaration(llvm::Module *module, static void -translate_immediate(llvm::Module *module, +translate_immediate(Storage *storage, struct tgsi_full_immediate *imm) { + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert( 0 ); + } + } + printf("-------------- VEC = %f %f %f %f\n", + vec[0], vec[1], vec[2], vec[3]); + storage->addImmediate(vec); } static void @@ -200,6 +214,8 @@ translate_instruction(llvm::Module *module, val = storage->tempElement(src->SrcRegister.Index); } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { val = storage->outputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->immediateElement(src->SrcRegister.Index); } else { fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); return; @@ -679,7 +695,7 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) break; case TGSI_TOKEN_TYPE_IMMEDIATE: - translate_immediate(mod, + translate_immediate(&storage, &parse.FullToken.FullImmediate); break; diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index ff62fcf3e4..1aaabbe882 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -393,3 +393,18 @@ void Storage::popTemps() m_tempStack.pop(); } #endif //MESA_LLVM + +llvm::Value * Storage::immediateElement(int idx) +{ + return m_immediates[idx]; +} + +void Storage::addImmediate(float *val) +{ + std::vector vec(4); + vec[0] = ConstantFP::get(Type::FloatTy, APFloat(val[0])); + vec[1] = ConstantFP::get(Type::FloatTy, APFloat(val[1])); + vec[2] = ConstantFP::get(Type::FloatTy, APFloat(val[2])); + vec[3] = ConstantFP::get(Type::FloatTy, APFloat(val[3])); + m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); +} diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index ac3e4d5af9..f9a82ef857 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -65,6 +65,7 @@ public: llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); llvm::Value *constElement(int idx, llvm::Value *indIdx =0); llvm::Value *outputElement(int idx, llvm::Value *indIdx =0); + llvm::Value *immediateElement(int idx); llvm::Value *tempElement(int idx); void setTempElement(int idx, llvm::Value *val, int mask); @@ -87,6 +88,8 @@ public: void pushTemps(); void popTemps(); + void addImmediate(float *val); + private: llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); const char *name(const char *prefix); @@ -102,6 +105,7 @@ private: std::vector m_temps; std::vector m_addrs; std::vector m_dstCache; + std::vector m_immediates; llvm::VectorType *m_floatVecType; llvm::VectorType *m_intVecType; diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 75195968e5..36b41ea7e8 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -38,8 +38,6 @@ #define TGSI_DEBUG 0 -#define EMIT_IMMEDIATES 0 - /* * Map mesa register file to TGSI register file. @@ -60,11 +58,7 @@ map_register_file( case PROGRAM_UNIFORM: return TGSI_FILE_CONSTANT; case PROGRAM_CONSTANT: -#if EMIT_IMMEDIATES return TGSI_FILE_IMMEDIATE; -#else - return TGSI_FILE_CONSTANT; -#endif case PROGRAM_INPUT: return TGSI_FILE_INPUT; case PROGRAM_OUTPUT: @@ -101,10 +95,8 @@ map_register_file_index( case TGSI_FILE_OUTPUT: return outputMapping[index]; -#if EMIT_IMMEDIATES case TGSI_FILE_IMMEDIATE: return immediateMapping[index]; -#endif default: return index; @@ -166,7 +158,6 @@ convert_writemask( return writemask; } -#if EMIT_IMMEDIATES static struct tgsi_full_immediate make_immediate(const float *value, uint size) { @@ -177,7 +168,6 @@ make_immediate(const float *value, uint size) imm.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; return imm; } -#endif static void compile_instruction( @@ -661,9 +651,7 @@ tgsi_translate_mesa_program( struct tgsi_full_instruction fullinst; GLuint preamble_size = 0; GLuint immediates[1000]; -#if EMIT_IMMEDIATES GLuint numImmediates = 0; -#endif assert(procType == TGSI_PROCESSOR_FRAGMENT || procType == TGSI_PROCESSOR_VERTEX); @@ -804,7 +792,6 @@ tgsi_translate_mesa_program( } /* immediates/literals */ -#if EMIT_IMMEDIATES for (i = 0; i < program->Parameters->NumParameters; i++) { if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { struct tgsi_full_immediate fullimm @@ -818,7 +805,6 @@ tgsi_translate_mesa_program( numImmediates++; } } -#endif for( i = 0; i < program->NumInstructions; i++ ) { compile_instruction( -- cgit v1.2.3 From fd908ce234d1f553b59d65afb7e4243ffee24018 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 29 Oct 2007 11:42:22 -0400 Subject: Change the way we handle temporaries in LLVM translation. TGSI uses TEMP, among others, as a way of passing arguments from one function to another. Instead of trying to figure out which temp's a function needs and trying to dynamically adjust its signature just pass the whole array of temporaries to them. --- src/mesa/pipe/llvm/instructions.cpp | 8 +- src/mesa/pipe/llvm/instructions.h | 2 +- src/mesa/pipe/llvm/llvm_base_shader.cpp | 589 +++++++++++++++++--------------- src/mesa/pipe/llvm/llvm_builtins.c | 5 +- src/mesa/pipe/llvm/llvmtgsi.cpp | 23 +- src/mesa/pipe/llvm/storage.cpp | 84 +++-- src/mesa/pipe/llvm/storage.h | 15 +- 7 files changed, 376 insertions(+), 350 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index b0aa0cc746..2eac0edbc4 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -1098,12 +1098,13 @@ void Instructions::end() } void Instructions::cal(int label, llvm::Value *out, llvm::Value *in, - llvm::Value *cst) + llvm::Value *cst, llvm::Value *temp) { std::vector params; params.push_back(out); params.push_back(in); params.push_back(cst); + params.push_back(temp); llvm::Function *func = findFunction(label); new CallInst(func, params.begin(), params.end(), std::string(), m_block); @@ -1116,6 +1117,7 @@ llvm::Function * Instructions::declareFunc(int label) args.push_back(vecPtr); args.push_back(vecPtr); args.push_back(vecPtr); + args.push_back(vecPtr); ParamAttrsList *params = 0; FunctionType *funcType = FunctionType::get( /*Result=*/Type::VoidTy, @@ -1142,7 +1144,9 @@ void Instructions::bgnSub(unsigned label, Storage *storage) ptr_IN->setName("IN"); Value *ptr_CONST = args++; ptr_CONST->setName("CONST"); - storage->pushArguments(ptr_OUT, ptr_IN, ptr_CONST); + Value *ptr_TEMP = args++; + ptr_TEMP->setName("TEMP"); + storage->pushArguments(ptr_OUT, ptr_IN, ptr_CONST, ptr_TEMP); llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 85feb1665d..8a1aed3181 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -61,7 +61,7 @@ public: void bgnSub(unsigned, Storage *); void brk(); void cal(int label, llvm::Value *out, llvm::Value *in, - llvm::Value *cst); + llvm::Value *cst, llvm::Value *tmp); llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index 3f058258ee..bf95ba3846 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -30,6 +30,8 @@ Module* createBaseShader() { /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_0_PAL); + PointerType* PointerTy_7 = PointerType::get(Type::FloatTy); + PointerType* PointerTy_8 = PointerType::get(VectorTy_3); std::vectorFuncTy_9_args; @@ -73,19 +75,31 @@ Module* createBaseShader() { ArrayType* ArrayTy_14 = ArrayType::get(ArrayTy_2, 2048); + PointerType* PointerTy_13 = PointerType::get(ArrayTy_14); + ArrayType* ArrayTy_16 = ArrayType::get(VectorTy_3, 32); - std::vectorFuncTy_18_args; - FuncTy_18_args.push_back(PointerTy_8); - FuncTy_18_args.push_back(PointerTy_8); - FuncTy_18_args.push_back(PointerTy_8); - ParamAttrsList *FuncTy_18_PAL = 0; - FunctionType* FuncTy_18 = FunctionType::get( + PointerType* PointerTy_15 = PointerType::get(ArrayTy_16); + + ArrayType* ArrayTy_18 = ArrayType::get(VectorTy_3, 128); + + PointerType* PointerTy_17 = PointerType::get(ArrayTy_18); + + std::vectorFuncTy_20_args; + FuncTy_20_args.push_back(PointerTy_8); + FuncTy_20_args.push_back(PointerTy_8); + FuncTy_20_args.push_back(PointerTy_8); + FuncTy_20_args.push_back(PointerTy_8); + ParamAttrsList *FuncTy_20_PAL = 0; + FunctionType* FuncTy_20 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_18_args, + /*Params=*/FuncTy_20_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_18_PAL); - + /*ParamAttrs=*/FuncTy_20_PAL); + + PointerType* PointerTy_19 = PointerType::get(FuncTy_20); + + // Function Declarations Function* func_from_array = new Function( @@ -113,7 +127,7 @@ Module* createBaseShader() { func_run_vertex_shader->setCallingConv(CallingConv::C); Function* func_execute_shader = new Function( - /*Type=*/FuncTy_18, + /*Type=*/FuncTy_20, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"execute_shader", mod); // (external, no body) func_execute_shader->setCallingConv(CallingConv::C); @@ -122,11 +136,11 @@ Module* createBaseShader() { // Constant Definitions - Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); - UndefValue* const_packed_20 = UndefValue::get(VectorTy_3); - ConstantInt* const_int32_21 = ConstantInt::get(APInt(32, "1", 10)); - ConstantInt* const_int32_22 = ConstantInt::get(APInt(32, "2", 10)); - ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "3", 10)); + Constant* const_int32_21 = Constant::getNullValue(IntegerType::get(32)); + UndefValue* const_packed_22 = UndefValue::get(VectorTy_3); + ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); + ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "2", 10)); + ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "3", 10)); // Global Variable Definitions @@ -150,73 +164,73 @@ Module* createBaseShader() { BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); // Block entry (label_entry) - ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_19, "cmp", label_entry); - ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_19, "cmp5", label_entry); + ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_21, "cmp", label_entry); + ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_21, "cmp5", label_entry); BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); new BranchInst(label_forbody6, label_afterfor60, int1_bothcond, label_entry); // Block forbody6 (label_forbody6) - Argument* fwdref_25 = new Argument(IntegerType::get(32)); - Argument* fwdref_26 = new Argument(IntegerType::get(32)); + Argument* fwdref_27 = new Argument(IntegerType::get(32)); + Argument* fwdref_28 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody6); int32_i_0_reg2mem_0->reserveOperandSpace(3); - int32_i_0_reg2mem_0->addIncoming(const_int32_19, label_entry); - int32_i_0_reg2mem_0->addIncoming(fwdref_25, label_forinc57); - int32_i_0_reg2mem_0->addIncoming(fwdref_26, label_forbody6); + int32_i_0_reg2mem_0->addIncoming(const_int32_21, label_entry); + int32_i_0_reg2mem_0->addIncoming(fwdref_27, label_forinc57); + int32_i_0_reg2mem_0->addIncoming(fwdref_28, label_forbody6); - Argument* fwdref_27 = new Argument(IntegerType::get(32)); + Argument* fwdref_29 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0", label_forbody6); int32_j_0_reg2mem_0->reserveOperandSpace(3); - int32_j_0_reg2mem_0->addIncoming(fwdref_27, label_forbody6); - int32_j_0_reg2mem_0->addIncoming(const_int32_19, label_forinc57); - int32_j_0_reg2mem_0->addIncoming(const_int32_19, label_entry); + int32_j_0_reg2mem_0->addIncoming(fwdref_29, label_forbody6); + int32_j_0_reg2mem_0->addIncoming(const_int32_21, label_forinc57); + int32_j_0_reg2mem_0->addIncoming(const_int32_21, label_entry); - Argument* fwdref_28 = new Argument(VectorTy_3); + Argument* fwdref_30 = new Argument(VectorTy_3); PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_3, "vec.0.reg2mem.0", label_forbody6); packed_vec_0_reg2mem_0->reserveOperandSpace(3); - packed_vec_0_reg2mem_0->addIncoming(fwdref_28, label_forbody6); - packed_vec_0_reg2mem_0->addIncoming(const_packed_20, label_entry); - packed_vec_0_reg2mem_0->addIncoming(fwdref_28, label_forinc57); + packed_vec_0_reg2mem_0->addIncoming(fwdref_30, label_forbody6); + packed_vec_0_reg2mem_0->addIncoming(const_packed_22, label_entry); + packed_vec_0_reg2mem_0->addIncoming(fwdref_30, label_forinc57); std::vector ptr_arraydecay11_indices; ptr_arraydecay11_indices.push_back(int32_i_0_reg2mem_0); ptr_arraydecay11_indices.push_back(int32_j_0_reg2mem_0); - ptr_arraydecay11_indices.push_back(const_int32_19); + ptr_arraydecay11_indices.push_back(const_int32_21); Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); - InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_19, "tmp15", label_forbody6); + InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_21, "tmp15", label_forbody6); std::vector ptr_arrayidx23_indices; ptr_arrayidx23_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx23_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx23_indices.push_back(const_int32_21); + ptr_arrayidx23_indices.push_back(const_int32_23); Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_21, "tmp26", label_forbody6); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_23, "tmp26", label_forbody6); std::vector ptr_arrayidx34_indices; ptr_arrayidx34_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx34_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx34_indices.push_back(const_int32_22); + ptr_arrayidx34_indices.push_back(const_int32_24); Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6); LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6); - InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_22, "tmp37", label_forbody6); + InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_24, "tmp37", label_forbody6); std::vector ptr_arrayidx45_indices; ptr_arrayidx45_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx45_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx45_indices.push_back(const_int32_23); + ptr_arrayidx45_indices.push_back(const_int32_25); Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6); LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6); - InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_23, "tmp48", label_forbody6); + InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_25, "tmp48", label_forbody6); std::vector ptr_arrayidx54_indices; ptr_arrayidx54_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0); Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); - new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); - BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_21, "inc", label_forbody6); + StoreInst* void_31 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); + BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_23, "inc", label_forbody6); ICmpInst* int1_cmp59 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp59", label_forbody6); new BranchInst(label_forbody6, label_forinc57, int1_cmp59, label_forbody6); // Block forinc57 (label_forinc57) - BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_21, "inc59", label_forinc57); + BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_23, "inc59", label_forinc57); ICmpInst* int1_cmp17 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59, int32_count, "cmp17", label_forinc57); new BranchInst(label_forbody6, label_afterfor60, int1_cmp17, label_forinc57); @@ -224,80 +238,80 @@ Module* createBaseShader() { new ReturnInst(label_afterfor60); // Resolve Forward References - fwdref_26->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_26; - fwdref_28->replaceAllUsesWith(packed_tmp48); delete fwdref_28; - fwdref_27->replaceAllUsesWith(int32_inc); delete fwdref_27; - fwdref_25->replaceAllUsesWith(int32_inc59); delete fwdref_25; + fwdref_28->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_28; + fwdref_30->replaceAllUsesWith(packed_tmp48); delete fwdref_30; + fwdref_29->replaceAllUsesWith(int32_inc); delete fwdref_29; + fwdref_27->replaceAllUsesWith(int32_inc59); delete fwdref_27; } // Function: from_consts (func_from_consts) { Function::arg_iterator args = func_from_consts->arg_begin(); - Value* ptr_res_33 = args++; - ptr_res_33->setName("res"); - Value* ptr_ainputs_34 = args++; - ptr_ainputs_34->setName("ainputs"); - Value* int32_count_35 = args++; - int32_count_35->setName("count"); - - BasicBlock* label_entry_36 = new BasicBlock("entry",func_from_consts,0); + Value* ptr_res_35 = args++; + ptr_res_35->setName("res"); + Value* ptr_ainputs_36 = args++; + ptr_ainputs_36->setName("ainputs"); + Value* int32_count_37 = args++; + int32_count_37->setName("count"); + + BasicBlock* label_entry_38 = new BasicBlock("entry",func_from_consts,0); BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); - // Block entry (label_entry_36) - ICmpInst* int1_cmp_37 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_35, const_int32_19, "cmp", label_entry_36); - new BranchInst(label_forbody, label_afterfor, int1_cmp_37, label_entry_36); + // Block entry (label_entry_38) + ICmpInst* int1_cmp_39 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_37, const_int32_21, "cmp", label_entry_38); + new BranchInst(label_forbody, label_afterfor, int1_cmp_39, label_entry_38); // Block forbody (label_forbody) - Argument* fwdref_40 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_39 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); - int32_i_0_reg2mem_0_39->reserveOperandSpace(2); - int32_i_0_reg2mem_0_39->addIncoming(const_int32_19, label_entry_36); - int32_i_0_reg2mem_0_39->addIncoming(fwdref_40, label_forbody); - - Argument* fwdref_42 = new Argument(VectorTy_3); - PHINode* packed_vec_0_reg2mem_0_41 = new PHINode(VectorTy_3, "vec.0.reg2mem.0", label_forbody); - packed_vec_0_reg2mem_0_41->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_41->addIncoming(const_packed_20, label_entry_36); - packed_vec_0_reg2mem_0_41->addIncoming(fwdref_42, label_forbody); + Argument* fwdref_42 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_41 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); + int32_i_0_reg2mem_0_41->reserveOperandSpace(2); + int32_i_0_reg2mem_0_41->addIncoming(const_int32_21, label_entry_38); + int32_i_0_reg2mem_0_41->addIncoming(fwdref_42, label_forbody); + + Argument* fwdref_44 = new Argument(VectorTy_3); + PHINode* packed_vec_0_reg2mem_0_43 = new PHINode(VectorTy_3, "vec.0.reg2mem.0", label_forbody); + packed_vec_0_reg2mem_0_43->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_43->addIncoming(const_packed_22, label_entry_38); + packed_vec_0_reg2mem_0_43->addIncoming(fwdref_44, label_forbody); std::vector ptr_arraydecay_indices; - ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_39); - ptr_arraydecay_indices.push_back(const_int32_19); - Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_34, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); + ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_41); + ptr_arraydecay_indices.push_back(const_int32_21); + Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_36, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); - InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_41, float_tmp5, const_int32_19, "tmp7", label_forbody); + InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_43, float_tmp5, const_int32_21, "tmp7", label_forbody); std::vector ptr_arrayidx12_indices; - ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_39); - ptr_arrayidx12_indices.push_back(const_int32_21); - Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_34, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); - LoadInst* float_tmp13_43 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); - InsertElementInst* packed_tmp15_44 = new InsertElementInst(packed_tmp7, float_tmp13_43, const_int32_21, "tmp15", label_forbody); + ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_41); + ptr_arrayidx12_indices.push_back(const_int32_23); + Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_36, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); + LoadInst* float_tmp13_45 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); + InsertElementInst* packed_tmp15_46 = new InsertElementInst(packed_tmp7, float_tmp13_45, const_int32_23, "tmp15", label_forbody); std::vector ptr_arrayidx20_indices; - ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_39); - ptr_arrayidx20_indices.push_back(const_int32_22); - Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_34, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); + ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_41); + ptr_arrayidx20_indices.push_back(const_int32_24); + Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_36, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); - InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_44, float_tmp21, const_int32_22, "tmp23", label_forbody); + InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_46, float_tmp21, const_int32_24, "tmp23", label_forbody); std::vector ptr_arrayidx28_indices; - ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_39); - ptr_arrayidx28_indices.push_back(const_int32_23); - Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_34, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); + ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_41); + ptr_arrayidx28_indices.push_back(const_int32_25); + Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_36, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); - InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_23, "tmp31", label_forbody); - GetElementPtrInst* ptr_arrayidx34_45 = new GetElementPtrInst(ptr_res_33, int32_i_0_reg2mem_0_39, "arrayidx34", label_forbody); - new StoreInst(packed_tmp31, ptr_arrayidx34_45, false, label_forbody); - BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_39, const_int32_21, "indvar.next", label_forbody); - ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_35, "exitcond", label_forbody); + InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_25, "tmp31", label_forbody); + GetElementPtrInst* ptr_arrayidx34_47 = new GetElementPtrInst(ptr_res_35, int32_i_0_reg2mem_0_41, "arrayidx34", label_forbody); + StoreInst* void_48 = new StoreInst(packed_tmp31, ptr_arrayidx34_47, false, label_forbody); + BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_41, const_int32_23, "indvar.next", label_forbody); + ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_37, "exitcond", label_forbody); new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); // Block afterfor (label_afterfor) new ReturnInst(label_afterfor); // Resolve Forward References - fwdref_42->replaceAllUsesWith(packed_tmp31); delete fwdref_42; - fwdref_40->replaceAllUsesWith(int32_indvar_next); delete fwdref_40; + fwdref_44->replaceAllUsesWith(packed_tmp31); delete fwdref_44; + fwdref_42->replaceAllUsesWith(int32_indvar_next); delete fwdref_42; } @@ -308,81 +322,81 @@ Module* createBaseShader() { ptr_dests->setName("dests"); Value* ptr_in = args++; ptr_in->setName("in"); - Value* int32_num_attribs_49 = args++; - int32_num_attribs_49->setName("num_attribs"); - - BasicBlock* label_entry_50 = new BasicBlock("entry",func_to_array,0); - BasicBlock* label_forbody_51 = new BasicBlock("forbody",func_to_array,0); - BasicBlock* label_afterfor_52 = new BasicBlock("afterfor",func_to_array,0); - - // Block entry (label_entry_50) - ICmpInst* int1_cmp_53 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_49, const_int32_19, "cmp", label_entry_50); - new BranchInst(label_forbody_51, label_afterfor_52, int1_cmp_53, label_entry_50); - - // Block forbody (label_forbody_51) - Argument* fwdref_56 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_55 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_51); - int32_i_0_reg2mem_0_55->reserveOperandSpace(2); - int32_i_0_reg2mem_0_55->addIncoming(const_int32_19, label_entry_50); - int32_i_0_reg2mem_0_55->addIncoming(fwdref_56, label_forbody_51); - - std::vector ptr_arraydecay_57_indices; - ptr_arraydecay_57_indices.push_back(int32_i_0_reg2mem_0_55); - ptr_arraydecay_57_indices.push_back(const_int32_19); - Instruction* ptr_arraydecay_57 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_57_indices.begin(), ptr_arraydecay_57_indices.end(), "arraydecay", label_forbody_51); - GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_55, "arrayidx6", label_forbody_51); - LoadInst* packed_tmp7_58 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_51); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_58, const_int32_19, "tmp11", label_forbody_51); - new StoreInst(float_tmp11, ptr_arraydecay_57, false, label_forbody_51); + Value* int32_num_attribs_51 = args++; + int32_num_attribs_51->setName("num_attribs"); + + BasicBlock* label_entry_52 = new BasicBlock("entry",func_to_array,0); + BasicBlock* label_forbody_53 = new BasicBlock("forbody",func_to_array,0); + BasicBlock* label_afterfor_54 = new BasicBlock("afterfor",func_to_array,0); + + // Block entry (label_entry_52) + ICmpInst* int1_cmp_55 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_51, const_int32_21, "cmp", label_entry_52); + new BranchInst(label_forbody_53, label_afterfor_54, int1_cmp_55, label_entry_52); + + // Block forbody (label_forbody_53) + Argument* fwdref_58 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_57 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_53); + int32_i_0_reg2mem_0_57->reserveOperandSpace(2); + int32_i_0_reg2mem_0_57->addIncoming(const_int32_21, label_entry_52); + int32_i_0_reg2mem_0_57->addIncoming(fwdref_58, label_forbody_53); + + std::vector ptr_arraydecay_59_indices; + ptr_arraydecay_59_indices.push_back(int32_i_0_reg2mem_0_57); + ptr_arraydecay_59_indices.push_back(const_int32_21); + Instruction* ptr_arraydecay_59 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_59_indices.begin(), ptr_arraydecay_59_indices.end(), "arraydecay", label_forbody_53); + GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_57, "arrayidx6", label_forbody_53); + LoadInst* packed_tmp7_60 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_53); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_60, const_int32_21, "tmp11", label_forbody_53); + StoreInst* void_61 = new StoreInst(float_tmp11, ptr_arraydecay_59, false, label_forbody_53); std::vector ptr_arrayidx13_indices; - ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_55); - ptr_arrayidx13_indices.push_back(const_int32_21); - Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_51); - ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_58, const_int32_21, "tmp15", label_forbody_51); - new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_51); + ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_57); + ptr_arrayidx13_indices.push_back(const_int32_23); + Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_53); + ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_60, const_int32_23, "tmp15", label_forbody_53); + StoreInst* void_62 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_53); std::vector ptr_arrayidx17_indices; - ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_55); - ptr_arrayidx17_indices.push_back(const_int32_22); - Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_51); - ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_58, const_int32_22, "tmp19", label_forbody_51); - new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_51); + ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_57); + ptr_arrayidx17_indices.push_back(const_int32_24); + Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_53); + ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_60, const_int32_24, "tmp19", label_forbody_53); + StoreInst* void_63 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_53); std::vector ptr_arrayidx21_indices; - ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_55); - ptr_arrayidx21_indices.push_back(const_int32_23); - Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_51); - ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_58, const_int32_23, "tmp23", label_forbody_51); - new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_51); - BinaryOperator* int32_indvar_next_63 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_55, const_int32_21, "indvar.next", label_forbody_51); - ICmpInst* int1_exitcond_64 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_63, int32_num_attribs_49, "exitcond", label_forbody_51); - new BranchInst(label_afterfor_52, label_forbody_51, int1_exitcond_64, label_forbody_51); - - // Block afterfor (label_afterfor_52) - new ReturnInst(label_afterfor_52); + ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_57); + ptr_arrayidx21_indices.push_back(const_int32_25); + Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_53); + ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_60, const_int32_25, "tmp23", label_forbody_53); + StoreInst* void_64 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_53); + BinaryOperator* int32_indvar_next_65 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_57, const_int32_23, "indvar.next", label_forbody_53); + ICmpInst* int1_exitcond_66 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_65, int32_num_attribs_51, "exitcond", label_forbody_53); + new BranchInst(label_afterfor_54, label_forbody_53, int1_exitcond_66, label_forbody_53); + + // Block afterfor (label_afterfor_54) + new ReturnInst(label_afterfor_54); // Resolve Forward References - fwdref_56->replaceAllUsesWith(int32_indvar_next_63); delete fwdref_56; + fwdref_58->replaceAllUsesWith(int32_indvar_next_65); delete fwdref_58; } // Function: run_vertex_shader (func_run_vertex_shader) { Function::arg_iterator args = func_run_vertex_shader->arg_begin(); - Value* ptr_ainputs_67 = args++; - ptr_ainputs_67->setName("ainputs"); - Value* ptr_dests_68 = args++; - ptr_dests_68->setName("dests"); + Value* ptr_ainputs_69 = args++; + ptr_ainputs_69->setName("ainputs"); + Value* ptr_dests_70 = args++; + ptr_dests_70->setName("dests"); Value* ptr_aconsts = args++; ptr_aconsts->setName("aconsts"); Value* int32_num_vertices = args++; int32_num_vertices->setName("num_vertices"); Value* int32_num_inputs = args++; int32_num_inputs->setName("num_inputs"); - Value* int32_num_attribs_69 = args++; - int32_num_attribs_69->setName("num_attribs"); + Value* int32_num_attribs_71 = args++; + int32_num_attribs_71->setName("num_attribs"); Value* int32_num_consts = args++; int32_num_consts->setName("num_consts"); - BasicBlock* label_entry_70 = new BasicBlock("entry",func_run_vertex_shader,0); + BasicBlock* label_entry_72 = new BasicBlock("entry",func_run_vertex_shader,0); BasicBlock* label_forbody6_i = new BasicBlock("forbody6.i",func_run_vertex_shader,0); BasicBlock* label_forinc57_i = new BasicBlock("forinc57.i",func_run_vertex_shader,0); BasicBlock* label_from_array_exit = new BasicBlock("from_array.exit",func_run_vertex_shader,0); @@ -392,262 +406,269 @@ Module* createBaseShader() { BasicBlock* label_forbody_us = new BasicBlock("forbody.us",func_run_vertex_shader,0); BasicBlock* label_to_array_exit_us = new BasicBlock("to_array.exit.us",func_run_vertex_shader,0); BasicBlock* label_forbody_i_us = new BasicBlock("forbody.i.us",func_run_vertex_shader,0); - BasicBlock* label_forbody_71 = new BasicBlock("forbody",func_run_vertex_shader,0); - BasicBlock* label_afterfor_72 = new BasicBlock("afterfor",func_run_vertex_shader,0); - - // Block entry (label_entry_70) - AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_14, "inputs", label_entry_70); - AllocaInst* ptr_consts = new AllocaInst(ArrayTy_16, "consts", label_entry_70); - AllocaInst* ptr_results = new AllocaInst(ArrayTy_14, "results", label_entry_70); - ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_19, "cmp.i", label_entry_70); - ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_19, "cmp5.i", label_entry_70); - BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_70); - new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_70); + BasicBlock* label_forbody_73 = new BasicBlock("forbody",func_run_vertex_shader,0); + BasicBlock* label_afterfor_74 = new BasicBlock("afterfor",func_run_vertex_shader,0); + + // Block entry (label_entry_72) + AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_14, "inputs", label_entry_72); + AllocaInst* ptr_consts = new AllocaInst(ArrayTy_16, "consts", label_entry_72); + AllocaInst* ptr_results = new AllocaInst(ArrayTy_14, "results", label_entry_72); + AllocaInst* ptr_temps = new AllocaInst(ArrayTy_18, "temps", label_entry_72); + ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_21, "cmp.i", label_entry_72); + ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_21, "cmp5.i", label_entry_72); + BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_72); + new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_72); // Block forbody6.i (label_forbody6_i) - Argument* fwdref_74 = new Argument(IntegerType::get(32)); - Argument* fwdref_75 = new Argument(IntegerType::get(32)); + Argument* fwdref_76 = new Argument(IntegerType::get(32)); + Argument* fwdref_77 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i_ph = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i.ph", label_forbody6_i); int32_i_0_reg2mem_0_i_ph->reserveOperandSpace(3); - int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_19, label_entry_70); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_74, label_forinc57_i); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_75, label_forbody6_i); + int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_21, label_entry_72); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_76, label_forinc57_i); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_77, label_forbody6_i); - Argument* fwdref_76 = new Argument(IntegerType::get(32)); + Argument* fwdref_78 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i); int32_j_0_reg2mem_0_i->reserveOperandSpace(3); - int32_j_0_reg2mem_0_i->addIncoming(fwdref_76, label_forbody6_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_19, label_forinc57_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_19, label_entry_70); + int32_j_0_reg2mem_0_i->addIncoming(fwdref_78, label_forbody6_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_21, label_forinc57_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_21, label_entry_72); - Argument* fwdref_77 = new Argument(VectorTy_3); + Argument* fwdref_79 = new Argument(VectorTy_3); PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_3, "vec.0.reg2mem.0.i", label_forbody6_i); packed_vec_0_reg2mem_0_i->reserveOperandSpace(3); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_77, label_forbody6_i); - packed_vec_0_reg2mem_0_i->addIncoming(const_packed_20, label_entry_70); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_77, label_forinc57_i); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_79, label_forbody6_i); + packed_vec_0_reg2mem_0_i->addIncoming(const_packed_22, label_entry_72); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_79, label_forinc57_i); std::vector ptr_arraydecay11_i_indices; ptr_arraydecay11_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arraydecay11_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arraydecay11_i_indices.push_back(const_int32_19); - Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_67, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); + ptr_arraydecay11_i_indices.push_back(const_int32_21); + Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_69, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); LoadInst* float_tmp13_i = new LoadInst(ptr_arraydecay11_i, "tmp13.i", false, label_forbody6_i); - InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_19, "tmp15.i", label_forbody6_i); + InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_21, "tmp15.i", label_forbody6_i); std::vector ptr_arrayidx23_i_indices; ptr_arrayidx23_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx23_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx23_i_indices.push_back(const_int32_21); - Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_67, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); + ptr_arrayidx23_i_indices.push_back(const_int32_23); + Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_69, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); LoadInst* float_tmp24_i = new LoadInst(ptr_arrayidx23_i, "tmp24.i", false, label_forbody6_i); - InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_21, "tmp26.i", label_forbody6_i); + InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_23, "tmp26.i", label_forbody6_i); std::vector ptr_arrayidx34_i_indices; ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx34_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx34_i_indices.push_back(const_int32_22); - Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_67, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); + ptr_arrayidx34_i_indices.push_back(const_int32_24); + Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_69, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); LoadInst* float_tmp35_i = new LoadInst(ptr_arrayidx34_i, "tmp35.i", false, label_forbody6_i); - InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_22, "tmp37.i", label_forbody6_i); + InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_24, "tmp37.i", label_forbody6_i); std::vector ptr_arrayidx45_i_indices; ptr_arrayidx45_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx45_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx45_i_indices.push_back(const_int32_23); - Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_67, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); + ptr_arrayidx45_i_indices.push_back(const_int32_25); + Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_69, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); LoadInst* float_tmp46_i = new LoadInst(ptr_arrayidx45_i, "tmp46.i", false, label_forbody6_i); - InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_23, "tmp48.i", label_forbody6_i); + InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_25, "tmp48.i", label_forbody6_i); std::vector ptr_arrayidx54_i_indices; - ptr_arrayidx54_i_indices.push_back(const_int32_19); + ptr_arrayidx54_i_indices.push_back(const_int32_21); ptr_arrayidx54_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx54_i_indices.push_back(int32_j_0_reg2mem_0_i); Instruction* ptr_arrayidx54_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx54_i_indices.begin(), ptr_arrayidx54_i_indices.end(), "arrayidx54.i", label_forbody6_i); - new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); - BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_21, "inc.i", label_forbody6_i); + StoreInst* void_80 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); + BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_23, "inc.i", label_forbody6_i); ICmpInst* int1_cmp59_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_inputs, "cmp59.i", label_forbody6_i); new BranchInst(label_forbody6_i, label_forinc57_i, int1_cmp59_i, label_forbody6_i); // Block forinc57.i (label_forinc57_i) - BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_21, "inc59.i", label_forinc57_i); + BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_23, "inc59.i", label_forinc57_i); ICmpInst* int1_cmp17_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59_i, int32_num_vertices, "cmp17.i", label_forinc57_i); new BranchInst(label_forbody6_i, label_from_array_exit, int1_cmp17_i, label_forinc57_i); // Block from_array.exit (label_from_array_exit) - ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_19, "cmp.i4", label_from_array_exit); + ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_21, "cmp.i4", label_from_array_exit); new BranchInst(label_forbody_i15, label_forcond, int1_cmp_i4, label_from_array_exit); // Block forbody.i15 (label_forbody_i15) - Argument* fwdref_82 = new Argument(IntegerType::get(32)); + Argument* fwdref_84 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i5 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i5", label_forbody_i15); int32_i_0_reg2mem_0_i5->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i5->addIncoming(const_int32_19, label_from_array_exit); - int32_i_0_reg2mem_0_i5->addIncoming(fwdref_82, label_forbody_i15); + int32_i_0_reg2mem_0_i5->addIncoming(const_int32_21, label_from_array_exit); + int32_i_0_reg2mem_0_i5->addIncoming(fwdref_84, label_forbody_i15); - Argument* fwdref_83 = new Argument(VectorTy_3); + Argument* fwdref_85 = new Argument(VectorTy_3); PHINode* packed_vec_0_reg2mem_0_i6 = new PHINode(VectorTy_3, "vec.0.reg2mem.0.i6", label_forbody_i15); packed_vec_0_reg2mem_0_i6->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_20, label_from_array_exit); - packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_83, label_forbody_i15); + packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_22, label_from_array_exit); + packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_85, label_forbody_i15); std::vector ptr_arraydecay_i7_indices; ptr_arraydecay_i7_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arraydecay_i7_indices.push_back(const_int32_19); + ptr_arraydecay_i7_indices.push_back(const_int32_21); Instruction* ptr_arraydecay_i7 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i7_indices.begin(), ptr_arraydecay_i7_indices.end(), "arraydecay.i7", label_forbody_i15); LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i7, "tmp5.i", false, label_forbody_i15); - InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_19, "tmp7.i8", label_forbody_i15); + InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_21, "tmp7.i8", label_forbody_i15); std::vector ptr_arrayidx12_i_indices; ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx12_i_indices.push_back(const_int32_21); + ptr_arrayidx12_i_indices.push_back(const_int32_23); Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i15); LoadInst* float_tmp13_i9 = new LoadInst(ptr_arrayidx12_i, "tmp13.i9", false, label_forbody_i15); - InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_21, "tmp15.i10", label_forbody_i15); + InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_23, "tmp15.i10", label_forbody_i15); std::vector ptr_arrayidx20_i_indices; ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx20_i_indices.push_back(const_int32_22); + ptr_arrayidx20_i_indices.push_back(const_int32_24); Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i15); LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i15); - InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_22, "tmp23.i11", label_forbody_i15); + InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_24, "tmp23.i11", label_forbody_i15); std::vector ptr_arrayidx28_i_indices; ptr_arrayidx28_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx28_i_indices.push_back(const_int32_23); + ptr_arrayidx28_i_indices.push_back(const_int32_25); Instruction* ptr_arrayidx28_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i_indices.begin(), ptr_arrayidx28_i_indices.end(), "arrayidx28.i", label_forbody_i15); LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i15); - InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_23, "tmp31.i", label_forbody_i15); + InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_25, "tmp31.i", label_forbody_i15); std::vector ptr_arrayidx34_i12_indices; - ptr_arrayidx34_i12_indices.push_back(const_int32_19); + ptr_arrayidx34_i12_indices.push_back(const_int32_21); ptr_arrayidx34_i12_indices.push_back(int32_i_0_reg2mem_0_i5); Instruction* ptr_arrayidx34_i12 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i12_indices.begin(), ptr_arrayidx34_i12_indices.end(), "arrayidx34.i12", label_forbody_i15); - new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); - BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_21, "indvar.next24", label_forbody_i15); + StoreInst* void_86 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); + BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_23, "indvar.next24", label_forbody_i15); ICmpInst* int1_exitcond25 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next24, int32_num_consts, "exitcond25", label_forbody_i15); new BranchInst(label_forcond, label_forbody_i15, int1_exitcond25, label_forbody_i15); // Block forcond (label_forcond) - new BranchInst(label_forbody_preheader, label_afterfor_72, int1_cmp_i, label_forcond); + new BranchInst(label_forbody_preheader, label_afterfor_74, int1_cmp_i, label_forcond); // Block forbody.preheader (label_forbody_preheader) std::vector ptr_arraydecay17_indices; - ptr_arraydecay17_indices.push_back(const_int32_19); - ptr_arraydecay17_indices.push_back(const_int32_19); + ptr_arraydecay17_indices.push_back(const_int32_21); + ptr_arraydecay17_indices.push_back(const_int32_21); Instruction* ptr_arraydecay17 = new GetElementPtrInst(ptr_consts, ptr_arraydecay17_indices.begin(), ptr_arraydecay17_indices.end(), "arraydecay17", label_forbody_preheader); - ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_69, const_int32_19, "cmp.i1", label_forbody_preheader); - new BranchInst(label_forbody_us, label_forbody_71, int1_cmp_i1, label_forbody_preheader); + std::vector ptr_arraydecay18_indices; + ptr_arraydecay18_indices.push_back(const_int32_21); + ptr_arraydecay18_indices.push_back(const_int32_21); + Instruction* ptr_arraydecay18 = new GetElementPtrInst(ptr_temps, ptr_arraydecay18_indices.begin(), ptr_arraydecay18_indices.end(), "arraydecay18", label_forbody_preheader); + ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_71, const_int32_21, "cmp.i1", label_forbody_preheader); + new BranchInst(label_forbody_us, label_forbody_73, int1_cmp_i1, label_forbody_preheader); // Block forbody.us (label_forbody_us) - Argument* fwdref_88 = new Argument(IntegerType::get(32)); + Argument* fwdref_90 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.us", label_forbody_us); int32_i_0_reg2mem_0_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_us->addIncoming(const_int32_19, label_forbody_preheader); - int32_i_0_reg2mem_0_us->addIncoming(fwdref_88, label_to_array_exit_us); + int32_i_0_reg2mem_0_us->addIncoming(const_int32_21, label_forbody_preheader); + int32_i_0_reg2mem_0_us->addIncoming(fwdref_90, label_to_array_exit_us); std::vector ptr_arraydecay10_us_indices; - ptr_arraydecay10_us_indices.push_back(const_int32_19); + ptr_arraydecay10_us_indices.push_back(const_int32_21); ptr_arraydecay10_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay10_us_indices.push_back(const_int32_19); + ptr_arraydecay10_us_indices.push_back(const_int32_21); Instruction* ptr_arraydecay10_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_us_indices.begin(), ptr_arraydecay10_us_indices.end(), "arraydecay10.us", label_forbody_us); std::vector ptr_arraydecay14_us_indices; - ptr_arraydecay14_us_indices.push_back(const_int32_19); + ptr_arraydecay14_us_indices.push_back(const_int32_21); ptr_arraydecay14_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay14_us_indices.push_back(const_int32_19); + ptr_arraydecay14_us_indices.push_back(const_int32_21); Instruction* ptr_arraydecay14_us = new GetElementPtrInst(ptr_results, ptr_arraydecay14_us_indices.begin(), ptr_arraydecay14_us_indices.end(), "arraydecay14.us", label_forbody_us); - std::vector void_89_params; - void_89_params.push_back(ptr_arraydecay14_us); - void_89_params.push_back(ptr_arraydecay10_us); - void_89_params.push_back(ptr_arraydecay17); - CallInst* void_89 = new CallInst(func_execute_shader, void_89_params.begin(), void_89_params.end(), "", label_forbody_us); - void_89->setCallingConv(CallingConv::C); - void_89->setTailCall(false); + std::vector void_91_params; + void_91_params.push_back(ptr_arraydecay14_us); + void_91_params.push_back(ptr_arraydecay10_us); + void_91_params.push_back(ptr_arraydecay17); + void_91_params.push_back(ptr_arraydecay18); + CallInst* void_91 = new CallInst(func_execute_shader, void_91_params.begin(), void_91_params.end(), "", label_forbody_us); + void_91->setCallingConv(CallingConv::C); + void_91->setTailCall(false); new BranchInst(label_forbody_i_us, label_forbody_us); // Block to_array.exit.us (label_to_array_exit_us) - BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_21, "inc.us", label_to_array_exit_us); + BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_23, "inc.us", label_to_array_exit_us); ICmpInst* int1_cmp21_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_num_vertices, "cmp21.us", label_to_array_exit_us); - new BranchInst(label_forbody_us, label_afterfor_72, int1_cmp21_us, label_to_array_exit_us); + new BranchInst(label_forbody_us, label_afterfor_74, int1_cmp21_us, label_to_array_exit_us); // Block forbody.i.us (label_forbody_i_us) - Argument* fwdref_92 = new Argument(IntegerType::get(32)); + Argument* fwdref_94 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i2_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i2.us", label_forbody_i_us); int32_i_0_reg2mem_0_i2_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_19, label_forbody_us); - int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_92, label_forbody_i_us); + int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_21, label_forbody_us); + int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_94, label_forbody_i_us); std::vector ptr_arraydecay_i_us_indices; ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arraydecay_i_us_indices.push_back(const_int32_19); - Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); + ptr_arraydecay_i_us_indices.push_back(const_int32_21); + Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_70, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); std::vector ptr_arrayidx6_i_us_indices; - ptr_arrayidx6_i_us_indices.push_back(const_int32_19); + ptr_arrayidx6_i_us_indices.push_back(const_int32_21); ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); Instruction* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_results, ptr_arrayidx6_i_us_indices.begin(), ptr_arrayidx6_i_us_indices.end(), "arrayidx6.i.us", label_forbody_i_us); LoadInst* packed_tmp7_i_us = new LoadInst(ptr_arrayidx6_i_us, "tmp7.i.us", false, label_forbody_i_us); - ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_19, "tmp11.i.us", label_forbody_i_us); - new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); + ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_21, "tmp11.i.us", label_forbody_i_us); + StoreInst* void_95 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx13_i_us_indices; ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx13_i_us_indices.push_back(const_int32_21); - Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_21, "tmp15.i3.us", label_forbody_i_us); - new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); + ptr_arrayidx13_i_us_indices.push_back(const_int32_23); + Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_70, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_23, "tmp15.i3.us", label_forbody_i_us); + StoreInst* void_96 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx17_i_us_indices; ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx17_i_us_indices.push_back(const_int32_22); - Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_22, "tmp19.i.us", label_forbody_i_us); - new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); + ptr_arrayidx17_i_us_indices.push_back(const_int32_24); + Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_70, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_24, "tmp19.i.us", label_forbody_i_us); + StoreInst* void_97 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx21_i_us_indices; ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx21_i_us_indices.push_back(const_int32_23); - Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_68, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_23, "tmp23.i.us", label_forbody_i_us); - new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); - BinaryOperator* int32_indvar_next_97 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_21, "indvar.next", label_forbody_i_us); - ICmpInst* int1_exitcond_98 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_97, int32_num_attribs_69, "exitcond", label_forbody_i_us); - new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_98, label_forbody_i_us); - - // Block forbody (label_forbody_71) - Argument* fwdref_101 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_100 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_71); - int32_i_0_reg2mem_0_100->reserveOperandSpace(2); - int32_i_0_reg2mem_0_100->addIncoming(const_int32_19, label_forbody_preheader); - int32_i_0_reg2mem_0_100->addIncoming(fwdref_101, label_forbody_71); + ptr_arrayidx21_i_us_indices.push_back(const_int32_25); + Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_70, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_25, "tmp23.i.us", label_forbody_i_us); + StoreInst* void_98 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); + BinaryOperator* int32_indvar_next_99 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_23, "indvar.next", label_forbody_i_us); + ICmpInst* int1_exitcond_100 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_99, int32_num_attribs_71, "exitcond", label_forbody_i_us); + new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_100, label_forbody_i_us); + + // Block forbody (label_forbody_73) + Argument* fwdref_103 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_102 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_73); + int32_i_0_reg2mem_0_102->reserveOperandSpace(2); + int32_i_0_reg2mem_0_102->addIncoming(const_int32_21, label_forbody_preheader); + int32_i_0_reg2mem_0_102->addIncoming(fwdref_103, label_forbody_73); std::vector ptr_arraydecay10_indices; - ptr_arraydecay10_indices.push_back(const_int32_19); - ptr_arraydecay10_indices.push_back(int32_i_0_reg2mem_0_100); - ptr_arraydecay10_indices.push_back(const_int32_19); - Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_forbody_71); + ptr_arraydecay10_indices.push_back(const_int32_21); + ptr_arraydecay10_indices.push_back(int32_i_0_reg2mem_0_102); + ptr_arraydecay10_indices.push_back(const_int32_21); + Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_forbody_73); std::vector ptr_arraydecay14_indices; - ptr_arraydecay14_indices.push_back(const_int32_19); - ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_100); - ptr_arraydecay14_indices.push_back(const_int32_19); - Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_results, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_71); - std::vector void_102_params; - void_102_params.push_back(ptr_arraydecay14); - void_102_params.push_back(ptr_arraydecay10); - void_102_params.push_back(ptr_arraydecay17); - CallInst* void_102 = new CallInst(func_execute_shader, void_102_params.begin(), void_102_params.end(), "", label_forbody_71); - void_102->setCallingConv(CallingConv::C); - void_102->setTailCall(false); - BinaryOperator* int32_inc_103 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_100, const_int32_21, "inc", label_forbody_71); - ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_103, int32_num_vertices, "cmp21", label_forbody_71); - new BranchInst(label_forbody_71, label_afterfor_72, int1_cmp21, label_forbody_71); - - // Block afterfor (label_afterfor_72) - new ReturnInst(label_afterfor_72); + ptr_arraydecay14_indices.push_back(const_int32_21); + ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_102); + ptr_arraydecay14_indices.push_back(const_int32_21); + Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_results, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_73); + std::vector void_104_params; + void_104_params.push_back(ptr_arraydecay14); + void_104_params.push_back(ptr_arraydecay10); + void_104_params.push_back(ptr_arraydecay17); + void_104_params.push_back(ptr_arraydecay18); + CallInst* void_104 = new CallInst(func_execute_shader, void_104_params.begin(), void_104_params.end(), "", label_forbody_73); + void_104->setCallingConv(CallingConv::C); + void_104->setTailCall(false); + BinaryOperator* int32_inc_105 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_102, const_int32_23, "inc", label_forbody_73); + ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_105, int32_num_vertices, "cmp21", label_forbody_73); + new BranchInst(label_forbody_73, label_afterfor_74, int1_cmp21, label_forbody_73); + + // Block afterfor (label_afterfor_74) + new ReturnInst(label_afterfor_74); // Resolve Forward References - fwdref_75->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_75; - fwdref_77->replaceAllUsesWith(packed_tmp48_i); delete fwdref_77; - fwdref_76->replaceAllUsesWith(int32_inc_i); delete fwdref_76; - fwdref_74->replaceAllUsesWith(int32_inc59_i); delete fwdref_74; - fwdref_83->replaceAllUsesWith(packed_tmp31_i); delete fwdref_83; - fwdref_82->replaceAllUsesWith(int32_indvar_next24); delete fwdref_82; - fwdref_88->replaceAllUsesWith(int32_inc_us); delete fwdref_88; - fwdref_92->replaceAllUsesWith(int32_indvar_next_97); delete fwdref_92; - fwdref_101->replaceAllUsesWith(int32_inc_103); delete fwdref_101; + fwdref_77->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_77; + fwdref_79->replaceAllUsesWith(packed_tmp48_i); delete fwdref_79; + fwdref_78->replaceAllUsesWith(int32_inc_i); delete fwdref_78; + fwdref_76->replaceAllUsesWith(int32_inc59_i); delete fwdref_76; + fwdref_85->replaceAllUsesWith(packed_tmp31_i); delete fwdref_85; + fwdref_84->replaceAllUsesWith(int32_indvar_next24); delete fwdref_84; + fwdref_90->replaceAllUsesWith(int32_inc_us); delete fwdref_90; + fwdref_94->replaceAllUsesWith(int32_indvar_next_99); delete fwdref_94; + fwdref_103->replaceAllUsesWith(int32_inc_105); delete fwdref_103; } diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index b0f7cae43b..c8ed384ab9 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -165,7 +165,7 @@ void to_array(float (*dests)[4], float4 *in, int num_attribs) } extern void execute_shader(float4 dests[16], float4 inputs[16], - float4 consts[32]); + float4 consts[32], float4 temps[128]); void run_vertex_shader(float (*ainputs)[16][4], float (*dests)[16][4], @@ -178,6 +178,7 @@ void run_vertex_shader(float (*ainputs)[16][4], float4 inputs[16*32*4][16]; float4 consts[32]; float4 results[16*32*4][16]; + float4 temps[128];//MAX_PROGRAM_TEMPS /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", num_vertices, num_inputs, num_attribs, num_consts);*/ @@ -186,7 +187,7 @@ void run_vertex_shader(float (*ainputs)[16][4], for (int i = 0; i < num_vertices; ++i) { float4 *in = inputs[i]; float4 *res = results[i]; - execute_shader(res, in, consts); + execute_shader(res, in, consts, temps); to_array(dests[i], res, num_attribs); } } diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index ec38c695b9..797843c7c3 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -146,20 +146,6 @@ translate_declaration(llvm::Module *module, struct tgsi_full_declaration *decl, struct tgsi_full_declaration *fd) { - if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: { - int start = decl->u.DeclarationRange.First; - int end = decl->u.DeclarationRange.Last; - for (int i = start; i <= end; ++i) { - storage->declareTemp(i); - } - } - break; - default: - assert( 0 ); - } - } } @@ -424,7 +410,8 @@ translate_instruction(llvm::Module *module, instr->cal(inst->InstructionExtLabel.Label, storage->outputPtr(), storage->inputPtr(), - storage->constPtr()); + storage->constPtr(), + storage->tempPtr()); return; } break; @@ -641,7 +628,7 @@ translate_instruction(llvm::Module *module, struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->store(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { @@ -675,6 +662,8 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) ptr_IN->setName("IN"); Value *ptr_CONST = args++; ptr_CONST->setName("CONST"); + Value *ptr_TEMPS = args++; + ptr_TEMPS->setName("TEMPS"); BasicBlock *label_entry = new BasicBlock("entry", shader, 0); @@ -682,7 +671,7 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); - Storage storage(label_entry, ptr_OUT, ptr_IN, ptr_CONST); + Storage storage(label_entry, ptr_OUT, ptr_IN, ptr_CONST, ptr_TEMPS); Instructions instr(mod, shader, label_entry); while(!tgsi_parse_end_of_tokens(&parse)) { tgsi_parse_token(&parse); diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 1aaabbe882..6ed243d9f3 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -47,10 +47,10 @@ using namespace llvm; Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, - llvm::Value *in, llvm::Value *consts) + llvm::Value *in, llvm::Value *consts, llvm::Value *temps) : m_block(block), m_OUT(out), - m_IN(in), m_CONST(consts), - m_temps(32), m_addrs(32), + m_IN(in), m_CONST(consts), m_TEMPS(temps), + m_addrs(32), m_dstCache(32), m_idx(0) { @@ -185,32 +185,51 @@ llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) } -llvm::Value *Storage::tempElement(int idx) +llvm::Value *Storage::tempElement(int idx, llvm::Value *indIdx) { - Value *ptr = m_temps[idx]; - if (!ptr) - return m_undefFloatVec; - llvm::LoadInst *li = new LoadInst(ptr, name("temp"), - false, m_block); - li->setAlignment(8); - return li; + GetElementPtrInst *getElem = 0; + + if (indIdx) { + getElem = new GetElementPtrInst(m_TEMPS, + BinaryOperator::create(Instruction::Add, + indIdx, + constantInt(idx), + name("add"), + m_block), + name("temp_ptr"), + m_block); + } else { + getElem = new GetElementPtrInst(m_TEMPS, + constantInt(idx), + name("temp_ptr"), + m_block); + } + + LoadInst *load = new LoadInst(getElem, name("temp"), + false, m_block); + load->setAlignment(8); + + return load; } void Storage::setTempElement(int idx, llvm::Value *val, int mask) { if (mask != TGSI_WRITEMASK_XYZW) { llvm::Value *templ = 0; - if (m_temps[idx]) + if (m_tempWriteMap[idx]) templ = tempElement(idx); val = maskWrite(val, mask, templ); } - llvm::Value *ptr = m_temps[idx]; - assert(ptr); - StoreInst *st = new StoreInst(val, ptr, false, m_block); + GetElementPtrInst *getElem = new GetElementPtrInst(m_TEMPS, + constantInt(idx), + name("temp_ptr"), + m_block); + StoreInst *st = new StoreInst(val, getElem, false, m_block); st->setAlignment(8); + m_tempWriteMap[idx] = true; } -void Storage::store(int dstIdx, llvm::Value *val, int mask) +void Storage::setOutputElement(int dstIdx, llvm::Value *val, int mask) { if (mask != TGSI_WRITEMASK_XYZW) { llvm::Value *templ = 0; @@ -301,12 +320,7 @@ void Storage::setCurrentBlock(llvm::BasicBlock *block) m_block = block; } -void Storage::declareTemp(int idx) -{ - m_temps[idx] = new AllocaInst(m_floatVecType, name("temp"), m_block); -} - -llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx ) +llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx) { GetElementPtrInst *getElem = 0; @@ -348,18 +362,25 @@ llvm::Value * Storage::constPtr() const return m_CONST; } +llvm::Value * Storage::tempPtr() const +{ + return m_TEMPS; +} + void Storage::pushArguments(llvm::Value *out, llvm::Value *in, - llvm::Value *constPtr) + llvm::Value *constPtr, llvm::Value *temp) { Args arg; arg.out = m_OUT; arg.in = m_IN; arg.cst = m_CONST; + arg.temp = m_TEMPS; m_argStack.push(arg); m_OUT = out; m_IN = in; m_CONST = constPtr; + m_TEMPS = temp; } void Storage::popArguments() @@ -368,31 +389,18 @@ void Storage::popArguments() m_OUT = arg.out; m_IN = arg.in; m_CONST = arg.cst; + m_TEMPS = arg.temp; m_argStack.pop(); } void Storage::pushTemps() { - m_tempStack.push(m_temps); - std::vector oldTemps = m_temps; - m_temps = std::vector(32); - int i = 0; - for (std::vector::iterator itr = oldTemps.begin(); - itr != oldTemps.end(); ++itr) { - if (*itr) { - declareTemp(i); - } - ++i; - } m_extSwizzleVec = 0; } void Storage::popTemps() { - m_temps = m_tempStack.top(); - m_tempStack.pop(); } -#endif //MESA_LLVM llvm::Value * Storage::immediateElement(int idx) { @@ -408,3 +416,5 @@ void Storage::addImmediate(float *val) vec[3] = ConstantFP::get(Type::FloatTy, APFloat(val[3])); m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); } + +#endif //MESA_LLVM diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index f9a82ef857..aa02f02f85 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -52,11 +52,12 @@ class Storage public: Storage(llvm::BasicBlock *block, llvm::Value *out, - llvm::Value *in, llvm::Value *consts); + llvm::Value *in, llvm::Value *consts, llvm::Value *temps); llvm::Value *inputPtr() const; llvm::Value *outputPtr() const; llvm::Value *constPtr() const; + llvm::Value *tempPtr() const; void setCurrentBlock(llvm::BasicBlock *block); @@ -65,11 +66,11 @@ public: llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); llvm::Value *constElement(int idx, llvm::Value *indIdx =0); llvm::Value *outputElement(int idx, llvm::Value *indIdx =0); + llvm::Value *tempElement(int idx, llvm::Value *indIdx =0); llvm::Value *immediateElement(int idx); - llvm::Value *tempElement(int idx); + void setOutputElement(int dstIdx, llvm::Value *val, int mask); void setTempElement(int idx, llvm::Value *val, int mask); - void declareTemp(int idx); llvm::Value *addrElement(int idx) const; void setAddrElement(int idx, llvm::Value *val, int mask); @@ -78,12 +79,10 @@ public: llvm::Value *extractIndex(llvm::Value *vec); - void store(int dstIdx, llvm::Value *val, int mask); - int numConsts() const; void pushArguments(llvm::Value *out, llvm::Value *in, - llvm::Value *constPtr); + llvm::Value *constPtr, llvm::Value *temp); void popArguments(); void pushTemps(); void popTemps(); @@ -99,10 +98,10 @@ private: llvm::Value *m_OUT; llvm::Value *m_IN; llvm::Value *m_CONST; + llvm::Value *m_TEMPS; std::map m_constInts; std::map m_intVecs; - std::vector m_temps; std::vector m_addrs; std::vector m_dstCache; std::vector m_immediates; @@ -116,6 +115,7 @@ private: int m_numConsts; std::map m_destWriteMap; + std::map m_tempWriteMap; llvm::Value *m_undefFloatVec; llvm::Value *m_undefIntVec; @@ -125,6 +125,7 @@ private: llvm::Value *out; llvm::Value *in; llvm::Value *cst; + llvm::Value *temp; }; std::stack m_argStack; std::stack > m_tempStack; -- cgit v1.2.3 From 75a9018fb90c93033ee5cbade9dd2febdc195d11 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 29 Oct 2007 13:20:55 -0400 Subject: Dump only relevant functions when in the debugging output. --- src/mesa/pipe/llvm/llvmtgsi.cpp | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 797843c7c3..94fc708606 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -802,14 +802,20 @@ void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) out << (*mod); out.close(); } else { - std::ostringstream stream; - stream << "execute_shader"; - stream << prog->id; - std::string func_name = stream.str(); - llvm::Function *func = mod->getFunction(func_name.c_str()); - assert(func); + const llvm::Module::FunctionListType &funcs = mod->getFunctionList(); + llvm::Module::FunctionListType::const_iterator itr; std::cout<<"; ---------- Start shader "<id<id< Date: Mon, 29 Oct 2007 13:42:58 -0400 Subject: Cleanup constant vector handling a bit. --- src/mesa/pipe/llvm/instructions.cpp | 28 ++++++++++++++-------------- src/mesa/pipe/llvm/instructions.h | 2 ++ src/mesa/pipe/llvm/storage.cpp | 24 ++++++------------------ 3 files changed, 22 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 2eac0edbc4..3c68b764aa 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -466,12 +466,8 @@ llvm::Value * Instructions::lg2(llvm::Value *in) ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), name("extractw"), m_block); - llvm::Value *const_vec = vectorFromVals( - ConstantFP::get(Type::FloatTy, APFloat(1.442695f)), - ConstantFP::get(Type::FloatTy, APFloat(1.442695f)), - ConstantFP::get(Type::FloatTy, APFloat(1.442695f)), - ConstantFP::get(Type::FloatTy, APFloat(1.442695f)) - ); + llvm::Value *const_vec = constVector(1.442695f, 1.442695f, + 1.442695f, 1.442695f); return mul(vectorFromVals(callFLog(x), callFLog(y), callFLog(z), callFLog(w)), const_vec); } @@ -1016,14 +1012,7 @@ llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3) { llvm::Value *m = mul(in1, in2); - llvm::Value *vec1 = vectorFromVals(ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - ConstantFP::get(Type::FloatTy, - APFloat(1.f))); + llvm::Value *vec1 = constVector(1.f, 1.f, 1.f, 1.f); llvm::Value *s = sub(vec1, in1); return add(m, mul(s, in3)); } @@ -1170,4 +1159,15 @@ llvm::Function * Instructions::findFunction(int label) return func; } +llvm::Value * Instructions::constVector(float x, float y, float z, float w) +{ + std::vector vec(4); + vec[0] = ConstantFP::get(Type::FloatTy, APFloat(x)); + vec[1] = ConstantFP::get(Type::FloatTy, APFloat(y)); + vec[2] = ConstantFP::get(Type::FloatTy, APFloat(z)); + vec[3] = ConstantFP::get(Type::FloatTy, APFloat(w)); + return ConstantVector::get(m_floatVecType, vec); +} + #endif //MESA_LLVM + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 8a1aed3181..b7b5129694 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -107,6 +107,8 @@ private: llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, llvm::Value *z, llvm::Value *w=0); + llvm::Value *constVector(float x, float y, float z, float w); + llvm::Function *declarePrintf(); llvm::Function *declareFunc(int label); diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 6ed243d9f3..1715bd4de4 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -68,24 +68,12 @@ Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, llvm::Constant *Storage::shuffleMask(int vec) { if (!m_extSwizzleVec) { - Constant *const_vec = Constant::getNullValue(m_floatVecType); - InsertElementInst *res = new InsertElementInst(const_vec, - ConstantFP::get(Type::FloatTy, APFloat(0.f)), - unsigned(0), - name("extswx"), m_block); - res = new InsertElementInst(res, ConstantFP::get(Type::FloatTy, APFloat(1.f)), - unsigned(1), - name("extswy"), - m_block); - res = new InsertElementInst(res, ConstantFP::get(Type::FloatTy, APFloat(0.f)), - unsigned(2), - name("extswz"), - m_block); - res = new InsertElementInst(res, ConstantFP::get(Type::FloatTy, APFloat(1.f)), - unsigned(3), - name("extsww"), - m_block); - m_extSwizzleVec = res; + std::vector elems; + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); + m_extSwizzleVec = ConstantVector::get(m_floatVecType, elems); } if (m_intVecs.find(vec) != m_intVecs.end()) { -- cgit v1.2.3 From ea2c74a25ef4d8bd0cb6fab21913c6a32bcee26f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 30 Oct 2007 06:33:30 -0400 Subject: Code cleanups. --- src/mesa/pipe/llvm/llvmtgsi.cpp | 37 ++++++++++++++++++++++--------------- src/mesa/pipe/llvm/llvmtgsi.h | 2 -- 2 files changed, 22 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index 94fc708606..e7cfeb2c91 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -164,11 +164,28 @@ translate_immediate(Storage *storage, assert( 0 ); } } - printf("-------------- VEC = %f %f %f %f\n", - vec[0], vec[1], vec[2], vec[3]); storage->addImmediate(vec); } +static inline llvm::Value * +swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, + Storage *storage) +{ + int swizzle = 0; + int start = 1000; + const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + + TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; + for (int k = 0; k < 4; ++k) { + swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; + start /= 10; + } + if (swizzle != NO_SWIZZLE) { + /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ + val = storage->shuffleVector(val, swizzle); + } + return val; +} + static void translate_instruction(llvm::Module *module, Storage *storage, @@ -207,19 +224,7 @@ translate_instruction(llvm::Module *module, return; } - int swizzle = 0; - int xstart = 1000; - const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + - TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; - for (int k = 0; k < 4; ++k) { - swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * xstart; - xstart /= 10; - } - if (swizzle != NO_SWIZZLE) { - /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ - val = storage->shuffleVector(val, swizzle); - } - inputs[i] = val; + inputs[i] = swizzleVector(val, src, storage); } /*if (inputs[0]) @@ -611,6 +616,7 @@ translate_instruction(llvm::Module *module, assert(!"Unsupported opcode"); } + /* # not sure if we need this */ switch( inst->Instruction.Saturate ) { case TGSI_SAT_NONE: break; @@ -624,6 +630,7 @@ translate_instruction(llvm::Module *module, assert( 0 ); } + /* store results */ for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; diff --git a/src/mesa/pipe/llvm/llvmtgsi.h b/src/mesa/pipe/llvm/llvmtgsi.h index a76dfd60b5..dadc0ea455 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.h +++ b/src/mesa/pipe/llvm/llvmtgsi.h @@ -41,9 +41,7 @@ extern "C" { #ifdef MESA_LLVM -struct tgsi_exec_machine; struct tgsi_token; -struct tgsi_sampler; struct gallivm_prog; struct gallivm_cpu_engine; -- cgit v1.2.3 From a01341a762baf7d5e18bd1a55950da7958958927 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 30 Oct 2007 06:55:08 -0400 Subject: Rename to better reflect what it's really doing. --- src/mesa/pipe/llvm/llvm_builtins.c | 193 ------------------------------------- src/mesa/pipe/llvm/llvm_entry.c | 193 +++++++++++++++++++++++++++++++++++++ 2 files changed, 193 insertions(+), 193 deletions(-) delete mode 100644 src/mesa/pipe/llvm/llvm_builtins.c create mode 100644 src/mesa/pipe/llvm/llvm_entry.c (limited to 'src') diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c deleted file mode 100644 index c8ed384ab9..0000000000 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ /dev/null @@ -1,193 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -/* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm-dis */ -/* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -funcname=createBaseShader */ -typedef __attribute__(( ocu_vector_type(4) )) float float4; - -#if 0 -//clang doesn't suppoer "struct->member" notation yet -struct vertex_header { - unsigned clipmask:12; - unsigned edgeflag:1; - unsigned pad:3; - unsigned vertex_id:16; - - float clip[4]; - - float data[][4]; -}; - -inline float -dot4(float4 a, float4 b) -{ - float4 c = a*b; - return c.x + c.y + c.z + c.w; -} - -inline unsigned -compute_clipmask(float4 clip, float4 (*plane), unsigned nr) -{ - unsigned mask = 0; - unsigned i; - - for (i = 0; i < nr; i++) { - if (dot4(clip, plane[i]) < 0) - mask |= (1<clip; - x = clip[0] = res0.x; - y = clip[1] = res0.y; - z = clip[2] = res0.z; - w = clip[3] = res0.w; - - vOut->clipmask = compute_clipmask(res0, planes, nr_planes); - vOut->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - res0.x = x; res0.y = y; res0.z = z; res0.w = 1; - - /* Viewport mapping */ - res0 = res0 * scale + trans; - vOut->data[0][0] = res0.x; - vOut->data[0][1] = res0.y; - vOut->data[0][2] = res0.z; - vOut->data[0][3] = w; - - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - * Skip 0 since we just did it above. - * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. - */ - for (slot = 1; slot < num_attribs - 2; slot++) { - float4 vec = results[slot]; - vOut->data[slot][0] = vec.x; - vOut->data[slot][1] = vec.y; - vOut->data[slot][2] = vec.z; - vOut->data[slot][3] = vec.w; - - printf("output %d: %f %f %f %f\n", slot, - vOut->data[slot][0], - vOut->data[slot][1], - vOut->data[slot][2], - vOut->data[slot][3]); - } -} -#endif - -void from_array(float4 (*res)[16], float (*ainputs)[16][4], - int count, int num_attribs) -{ - for (int i = 0; i < count; ++i) { - for (int j = 0; j < num_attribs; ++j) { - float4 vec; - vec.x = ainputs[i][j][0]; - vec.y = ainputs[i][j][1]; - vec.z = ainputs[i][j][2]; - vec.w = ainputs[i][j][3]; - res[i][j] = vec; - } - } -} - -void from_consts(float4 *res, float (*ainputs)[4], - int count) -{ - for (int i = 0; i < count; ++i) { - float4 vec; - vec.x = ainputs[i][0]; - vec.y = ainputs[i][1]; - vec.z = ainputs[i][2]; - vec.w = ainputs[i][3]; - res[i] = vec; - } -} - -void to_array(float (*dests)[4], float4 *in, int num_attribs) -{ - for (int i = 0; i < num_attribs; ++i) { - float *rd = dests[i]; - float4 ri = in[i]; - rd[0] = ri.x; - rd[1] = ri.y; - rd[2] = ri.z; - rd[3] = ri.w; - } -} - -extern void execute_shader(float4 dests[16], float4 inputs[16], - float4 consts[32], float4 temps[128]); - -void run_vertex_shader(float (*ainputs)[16][4], - float (*dests)[16][4], - float (*aconsts)[4], - int num_vertices, - int num_inputs, - int num_attribs, - int num_consts) -{ - float4 inputs[16*32*4][16]; - float4 consts[32]; - float4 results[16*32*4][16]; - float4 temps[128];//MAX_PROGRAM_TEMPS - - /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", - num_vertices, num_inputs, num_attribs, num_consts);*/ - from_array(inputs, ainputs, num_vertices, num_inputs); - from_consts(consts, aconsts, num_consts); - for (int i = 0; i < num_vertices; ++i) { - float4 *in = inputs[i]; - float4 *res = results[i]; - execute_shader(res, in, consts, temps); - to_array(dests[i], res, num_attribs); - } -} diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c new file mode 100644 index 0000000000..67066455ae --- /dev/null +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -0,0 +1,193 @@ +/* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -funcname=createBaseShader */ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +/* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm-dis */ +typedef __attribute__(( ocu_vector_type(4) )) float float4; + +#if 0 +//clang doesn't suppoer "struct->member" notation yet +struct vertex_header { + unsigned clipmask:12; + unsigned edgeflag:1; + unsigned pad:3; + unsigned vertex_id:16; + + float clip[4]; + + float data[][4]; +}; + +inline float +dot4(float4 a, float4 b) +{ + float4 c = a*b; + return c.x + c.y + c.z + c.w; +} + +inline unsigned +compute_clipmask(float4 clip, float4 (*plane), unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + for (i = 0; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<clip; + x = clip[0] = res0.x; + y = clip[1] = res0.y; + z = clip[2] = res0.z; + w = clip[3] = res0.w; + + vOut->clipmask = compute_clipmask(res0, planes, nr_planes); + vOut->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + res0.x = x; res0.y = y; res0.z = z; res0.w = 1; + + /* Viewport mapping */ + res0 = res0 * scale + trans; + vOut->data[0][0] = res0.x; + vOut->data[0][1] = res0.y; + vOut->data[0][2] = res0.z; + vOut->data[0][3] = w; + + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + * Skip 0 since we just did it above. + * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. + */ + for (slot = 1; slot < num_attribs - 2; slot++) { + float4 vec = results[slot]; + vOut->data[slot][0] = vec.x; + vOut->data[slot][1] = vec.y; + vOut->data[slot][2] = vec.z; + vOut->data[slot][3] = vec.w; + + printf("output %d: %f %f %f %f\n", slot, + vOut->data[slot][0], + vOut->data[slot][1], + vOut->data[slot][2], + vOut->data[slot][3]); + } +} +#endif + +void from_array(float4 (*res)[16], float (*ainputs)[16][4], + int count, int num_attribs) +{ + for (int i = 0; i < count; ++i) { + for (int j = 0; j < num_attribs; ++j) { + float4 vec; + vec.x = ainputs[i][j][0]; + vec.y = ainputs[i][j][1]; + vec.z = ainputs[i][j][2]; + vec.w = ainputs[i][j][3]; + res[i][j] = vec; + } + } +} + +void from_consts(float4 *res, float (*ainputs)[4], + int count) +{ + for (int i = 0; i < count; ++i) { + float4 vec; + vec.x = ainputs[i][0]; + vec.y = ainputs[i][1]; + vec.z = ainputs[i][2]; + vec.w = ainputs[i][3]; + res[i] = vec; + } +} + +void to_array(float (*dests)[4], float4 *in, int num_attribs) +{ + for (int i = 0; i < num_attribs; ++i) { + float *rd = dests[i]; + float4 ri = in[i]; + rd[0] = ri.x; + rd[1] = ri.y; + rd[2] = ri.z; + rd[3] = ri.w; + } +} + +extern void execute_shader(float4 dests[16], float4 inputs[16], + float4 consts[32], float4 temps[128]); + +void run_vertex_shader(float (*ainputs)[16][4], + float (*dests)[16][4], + float (*aconsts)[4], + int num_vertices, + int num_inputs, + int num_attribs, + int num_consts) +{ + float4 inputs[16*32*4][16]; + float4 consts[32]; + float4 results[16*32*4][16]; + float4 temps[128];//MAX_PROGRAM_TEMPS + + /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", + num_vertices, num_inputs, num_attribs, num_consts);*/ + from_array(inputs, ainputs, num_vertices, num_inputs); + from_consts(consts, aconsts, num_consts); + for (int i = 0; i < num_vertices; ++i) { + float4 *in = inputs[i]; + float4 *res = results[i]; + execute_shader(res, in, consts, temps); + to_array(dests[i], res, num_attribs); + } +} -- cgit v1.2.3 From dfd774318d1549c1e52364870a2c16a220dcd64c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 30 Oct 2007 07:16:22 -0400 Subject: Export true C based builtins to a file of its own. --- src/mesa/pipe/llvm/gallivm_builtins.cpp | 151 ++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.cpp | 181 +------------------------------- src/mesa/pipe/llvm/llvm_builtins.c | 59 +++++++++++ 3 files changed, 215 insertions(+), 176 deletions(-) create mode 100644 src/mesa/pipe/llvm/gallivm_builtins.cpp create mode 100644 src/mesa/pipe/llvm/llvm_builtins.c (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm_builtins.cpp b/src/mesa/pipe/llvm/gallivm_builtins.cpp new file mode 100644 index 0000000000..4acbedc566 --- /dev/null +++ b/src/mesa/pipe/llvm/gallivm_builtins.cpp @@ -0,0 +1,151 @@ +// Generated by llvm2cpp - DO NOT MODIFY! + + +Module* createGallivmBuiltins(Module *mod) { + +mod->setModuleIdentifier("shader"); + +// Type Definitions +std::vectorFuncTy_0_args; +FuncTy_0_args.push_back(Type::FloatTy); +FuncTy_0_args.push_back(Type::FloatTy); +ParamAttrsList *FuncTy_0_PAL = 0; +FunctionType* FuncTy_0 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_0_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_0_PAL); + +PointerType* PointerTy_1 = PointerType::get(FuncTy_0); + +VectorType* VectorTy_2 = VectorType::get(Type::FloatTy, 4); + +std::vectorFuncTy_3_args; +FuncTy_3_args.push_back(VectorTy_2); +ParamAttrsList *FuncTy_3_PAL = 0; +FunctionType* FuncTy_3 = FunctionType::get( + /*Result=*/VectorTy_2, + /*Params=*/FuncTy_3_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_3_PAL); + + +// Function Declarations + +Function* func_approx = new Function( + /*Type=*/FuncTy_0, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"approx", mod); +func_approx->setCallingConv(CallingConv::C); + +Function* func_powf = new Function( + /*Type=*/FuncTy_0, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"powf", mod); // (external, no body) +func_powf->setCallingConv(CallingConv::C); + +Function* func_lit = new Function( + /*Type=*/FuncTy_3, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"lit", mod); +func_lit->setCallingConv(CallingConv::C); + +// Global Variable Declarations + + +// Constant Definitions +ConstantFP* const_float_4 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); +ConstantFP* const_float_5 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); +Constant* const_float_6 = Constant::getNullValue(Type::FloatTy); +Constant* const_int32_7 = Constant::getNullValue(IntegerType::get(32)); +std::vector const_packed_8_elems; +ConstantFP* const_float_9 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); +const_packed_8_elems.push_back(const_float_9); +UndefValue* const_float_10 = UndefValue::get(Type::FloatTy); +const_packed_8_elems.push_back(const_float_10); +const_packed_8_elems.push_back(const_float_10); +const_packed_8_elems.push_back(const_float_9); +Constant* const_packed_8 = ConstantVector::get(VectorTy_2, const_packed_8_elems); +ConstantInt* const_int32_11 = ConstantInt::get(APInt(32, "1", 10)); +ConstantInt* const_int32_12 = ConstantInt::get(APInt(32, "3", 10)); +ConstantInt* const_int32_13 = ConstantInt::get(APInt(32, "2", 10)); +std::vector const_packed_14_elems; +const_packed_14_elems.push_back(const_float_9); +const_packed_14_elems.push_back(const_float_6); +const_packed_14_elems.push_back(const_float_6); +const_packed_14_elems.push_back(const_float_9); +Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elems); + +// Global Variable Definitions + +// Function Definitions + +// Function: approx (func_approx) +{ + Function::arg_iterator args = func_approx->arg_begin(); + Value* float_a = args++; + float_a->setName("a"); + Value* float_b = args++; + float_b->setName("b"); + + BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); + + // Block entry (label_entry) + FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_4, "cmp", label_entry); + SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_4, float_b, "b.addr.0", label_entry); + FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_5, "cmp3", label_entry); + SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_5, float_b_addr_0, "b.addr.1", label_entry); + FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_6, "cmp7", label_entry); + SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_6, float_a, "a.addr.0", label_entry); + std::vector float_call_params; + float_call_params.push_back(float_a_addr_0); + float_call_params.push_back(float_b_addr_1); + CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); + float_call->setCallingConv(CallingConv::C); + float_call->setTailCall(true); + new ReturnInst(float_call, label_entry); + +} + +// Function: lit (func_lit) +{ + Function::arg_iterator args = func_lit->arg_begin(); + Value* packed_tmp = args++; + packed_tmp->setName("tmp"); + + BasicBlock* label_entry_16 = new BasicBlock("entry",func_lit,0); + BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); + BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); + + // Block entry (label_entry_16) + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_7, "tmp7", label_entry_16); + FCmpInst* int1_cmp_17 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_6, "cmp", label_entry_16); + new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_17, label_entry_16); + + // Block ifthen (label_ifthen) + InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_8, float_tmp7, const_int32_11, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_11, "tmp14", label_ifthen); + ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_12, "tmp16", label_ifthen); + FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_4, "cmp.i", label_ifthen); + SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_4, float_tmp16, "b.addr.0.i", label_ifthen); + FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_5, "cmp3.i", label_ifthen); + SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_5, float_b_addr_0_i, "b.addr.1.i", label_ifthen); + FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_6, "cmp7.i", label_ifthen); + SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_6, float_tmp14, "a.addr.0.i", label_ifthen); + std::vector float_call_i_params; + float_call_i_params.push_back(float_a_addr_0_i); + float_call_i_params.push_back(float_b_addr_1_i); + CallInst* float_call_i = new CallInst(func_powf, float_call_i_params.begin(), float_call_i_params.end(), "call.i", label_ifthen); + float_call_i->setCallingConv(CallingConv::C); + float_call_i->setTailCall(true); + InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_13, "tmp18", label_ifthen); + new ReturnInst(packed_tmp18, label_ifthen); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock) + new ReturnInst(const_packed_14, label_UnifiedReturnBlock); + +} + +return mod; + +} diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 3c68b764aa..4e116f2a98 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -48,7 +48,7 @@ using namespace llvm; -Function* makeLitFunction(Module *mod); +#include "gallivm_builtins.cpp" static inline std::string createFuncName(int label) { @@ -70,6 +70,8 @@ Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicB m_llvmFlog = 0; m_llvmLit = 0; m_fmtPtr = 0; + + createGallivmBuiltins(m_mod); } llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) @@ -214,7 +216,7 @@ llvm::Value *Instructions::callFAbs(llvm::Value *val) llvm::Value * Instructions::lit(llvm::Value *in) { if (!m_llvmLit) { - m_llvmLit = makeLitFunction(m_mod); + m_llvmLit = m_mod->getFunction("lit"); } CallInst *call = new CallInst(m_llvmLit, in, name("litres"), m_block); call->setCallingConv(CallingConv::C); @@ -787,180 +789,6 @@ llvm::Value * Instructions::abs(llvm::Value *in) return vectorFromVals(xabs, yabs, zabs, wabs); } -/* -FIXME: hand write the lit function. Currently it's -generated from: -typedef __attribute__(( ocu_vector_type(4) )) float float4; - -extern float powf(float a, float b); - -inline float approx(float a, float b) -{ - if (b < -128.0f) b = -128.0f; - if (b > 128.0f) b = 128.0f; - if (a < 0) a = 0; - return powf(a, b); -} - -float4 lit(float4 tmp) -{ - float4 result; - result.x = 1.0; - result.w = 1.0; - if (tmp.x > 0) { - result.y = tmp.x; - result.z = approx(tmp.y, tmp.w); - } else { - result.y = 0; - result.z = 0; - } - return result; -} -with: -clang --emit-llvm lit.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -for=lit -*/ -Function* makeLitFunction(Module *mod) { - - // Type Definitions - std::vectorFuncTy_0_args; - FuncTy_0_args.push_back(Type::FloatTy); - FuncTy_0_args.push_back(Type::FloatTy); - ParamAttrsList *FuncTy_0_PAL = 0; - FunctionType* FuncTy_0 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_0_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_0_PAL); - - VectorType* VectorTy_2 = VectorType::get(Type::FloatTy, 4); - - std::vectorFuncTy_3_args; - FuncTy_3_args.push_back(VectorTy_2); - ParamAttrsList *FuncTy_3_PAL = 0; - FunctionType* FuncTy_3 = FunctionType::get( - /*Result=*/VectorTy_2, - /*Params=*/FuncTy_3_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_3_PAL); - - - // Function Declarations - - Function* func_approx = new Function( - /*Type=*/FuncTy_0, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"approx", mod); - func_approx->setCallingConv(CallingConv::C); - - Function* func_powf = new Function( - /*Type=*/FuncTy_0, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"powf", mod); // (external, no body) - func_powf->setCallingConv(CallingConv::C); - - Function* func_lit = new Function( - /*Type=*/FuncTy_3, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"lit", mod); - func_lit->setCallingConv(CallingConv::C); - - // Global Variable Declarations - - - // Constant Definitions - ConstantFP* const_float_4 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); - ConstantFP* const_float_5 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); - Constant* const_float_6 = Constant::getNullValue(Type::FloatTy); - Constant* const_int32_7 = Constant::getNullValue(IntegerType::get(32)); - std::vector const_packed_8_elems; - ConstantFP* const_float_9 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - const_packed_8_elems.push_back(const_float_9); - UndefValue* const_float_10 = UndefValue::get(Type::FloatTy); - const_packed_8_elems.push_back(const_float_10); - const_packed_8_elems.push_back(const_float_10); - const_packed_8_elems.push_back(const_float_9); - Constant* const_packed_8 = ConstantVector::get(VectorTy_2, const_packed_8_elems); - ConstantInt* const_int32_11 = ConstantInt::get(APInt(32, "1", 10)); - ConstantInt* const_int32_12 = ConstantInt::get(APInt(32, "3", 10)); - ConstantInt* const_int32_13 = ConstantInt::get(APInt(32, "2", 10)); - std::vector const_packed_14_elems; - const_packed_14_elems.push_back(const_float_9); - const_packed_14_elems.push_back(const_float_6); - const_packed_14_elems.push_back(const_float_6); - const_packed_14_elems.push_back(const_float_9); - Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elems); - - // Global Variable Definitions - - // Function Definitions - - // Function: approx (func_approx) - { - Function::arg_iterator args = func_approx->arg_begin(); - Value* float_a = args++; - float_a->setName("a"); - Value* float_b = args++; - float_b->setName("b"); - - BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); - - // Block entry (label_entry) - FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_4, "cmp", label_entry); - SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_4, float_b, "b.addr.0", label_entry); - FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_5, "cmp3", label_entry); - SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_5, float_b_addr_0, "b.addr.1", label_entry); - FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_6, "cmp7", label_entry); - SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_6, float_a, "a.addr.0", label_entry); - std::vector float_call_params; - float_call_params.push_back(float_a_addr_0); - float_call_params.push_back(float_b_addr_1); - CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); - float_call->setCallingConv(CallingConv::C); - float_call->setTailCall(true); - new ReturnInst(float_call, label_entry); - } - - // Function: lit (func_lit) - { - Function::arg_iterator args = func_lit->arg_begin(); - Value* packed_tmp = args++; - packed_tmp->setName("tmp"); - - BasicBlock* label_entry_16 = new BasicBlock("entry",func_lit,0); - BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); - BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); - - // Block entry (label_entry_16) - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_7, "tmp7", label_entry_16); - FCmpInst* int1_cmp_17 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_6, "cmp", label_entry_16); - new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_17, label_entry_16); - - // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_8, float_tmp7, const_int32_11, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_11, "tmp14", label_ifthen); - ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_12, "tmp16", label_ifthen); - FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_4, "cmp.i", label_ifthen); - SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_4, float_tmp16, "b.addr.0.i", label_ifthen); - FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_5, "cmp3.i", label_ifthen); - SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_5, float_b_addr_0_i, "b.addr.1.i", label_ifthen); - FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_6, "cmp7.i", label_ifthen); - SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_6, float_tmp14, "a.addr.0.i", label_ifthen); - std::vector float_call_i_params; - float_call_i_params.push_back(float_a_addr_0_i); - float_call_i_params.push_back(float_b_addr_1_i); - CallInst* float_call_i = new CallInst(func_powf, float_call_i_params.begin(), float_call_i_params.end(), "call.i", label_ifthen); - float_call_i->setCallingConv(CallingConv::C); - float_call_i->setTailCall(true); - InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_13, "tmp18", label_ifthen); - new ReturnInst(packed_tmp18, label_ifthen); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock) - new ReturnInst(const_packed_14, label_UnifiedReturnBlock); - - } - return func_lit; -} - void Instructions::ifop(llvm::Value *in) { BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); @@ -1171,3 +999,4 @@ llvm::Value * Instructions::constVector(float x, float y, float z, float w) #endif //MESA_LLVM + diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c new file mode 100644 index 0000000000..9d3d6141a6 --- /dev/null +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -0,0 +1,59 @@ +/*clang --emit-llvm llvm_builtins.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=gallivm_builtins.cpp -f -for=shader -funcname=createGallivmBuiltins*/ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +typedef __attribute__(( ocu_vector_type(4) )) float float4; + + +extern float powf(float a, float b); + +inline float approx(float a, float b) +{ + if (b < -128.0f) b = -128.0f; + if (b > 128.0f) b = 128.0f; + if (a < 0) a = 0; + return powf(a, b); +} + +float4 lit(float4 tmp) +{ + float4 result; + result.x = 1.0; + result.w = 1.0; + if (tmp.x > 0) { + result.y = tmp.x; + result.z = approx(tmp.y, tmp.w); + } else { + result.y = 0; + result.z = 0; + } + return result; +} -- cgit v1.2.3 From c97c03da46292abe72f94747fe527eb4e4623248 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 30 Oct 2007 07:43:02 -0400 Subject: Refactor vector extraction. --- src/mesa/pipe/llvm/instructions.cpp | 301 +++++++++++------------------------- src/mesa/pipe/llvm/instructions.h | 2 + 2 files changed, 89 insertions(+), 214 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 4e116f2a98..dca5f4c995 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -291,25 +291,14 @@ llvm::Value * Instructions::rcp(llvm::Value *in1) llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) { Value *mulRes = mul(in1, in2); - ExtractElementInst *x = new ExtractElementInst(mulRes, unsigned(0), - name("extractx"), - m_block); - ExtractElementInst *y = new ExtractElementInst(mulRes, unsigned(1), - name("extracty"), - m_block); - ExtractElementInst *z = new ExtractElementInst(mulRes, unsigned(2), - name("extractz"), - m_block); - ExtractElementInst *w = new ExtractElementInst(mulRes, unsigned(3), - name("extractw"), - m_block); - BinaryOperator *xy = BinaryOperator::create(Instruction::Add, x, y, - name("xy"), - m_block); - BinaryOperator *xyz = BinaryOperator::create(Instruction::Add, xy, z, + std::vector vec = extractVector(mulRes); + BinaryOperator *xy = BinaryOperator::create(Instruction::Add, vec[0], vec[1], + name("xy"), + m_block); + BinaryOperator *xyz = BinaryOperator::create(Instruction::Add, xy, vec[2], name("xyz"), m_block); - BinaryOperator *dot4 = BinaryOperator::create(Instruction::Add, xyz, w, + BinaryOperator *dot4 = BinaryOperator::create(Instruction::Add, xyz, vec[3], name("dot4"), m_block); return vectorFromVals(dot4, dot4, dot4, dot4); @@ -318,25 +307,14 @@ llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) { Value *mulRes = mul(in1, in2); - ExtractElementInst *x = new ExtractElementInst(mulRes, unsigned(0), - name("extractx"), - m_block); - ExtractElementInst *y = new ExtractElementInst(mulRes, unsigned(1), - name("extracty"), - m_block); - ExtractElementInst *z = new ExtractElementInst(mulRes, unsigned(2), - name("extractz"), - m_block); - ExtractElementInst *w = new ExtractElementInst(in2, unsigned(3), - name("val2w"), - m_block); - BinaryOperator *xy = BinaryOperator::create(Instruction::Add, x, y, + std::vector vec1 = extractVector(mulRes); + BinaryOperator *xy = BinaryOperator::create(Instruction::Add, vec1[0], vec1[1], name("xy"), m_block); - BinaryOperator *xyz = BinaryOperator::create(Instruction::Add, xy, z, + BinaryOperator *xyz = BinaryOperator::create(Instruction::Add, xy, vec1[2], name("xyz"), m_block); - BinaryOperator *dph = BinaryOperator::create(Instruction::Add, xyz, w, + BinaryOperator *dph = BinaryOperator::create(Instruction::Add, xyz, vec1[3], name("dph"), m_block); return vectorFromVals(dph, dph, dph, dph); @@ -401,20 +379,9 @@ llvm::Value * Instructions::callFloor(llvm::Value *val) llvm::Value * Instructions::floor(llvm::Value *in) { - ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), - name("extractx"), - m_block); - ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), - name("extracty"), - m_block); - ExtractElementInst *z = new ExtractElementInst(in, unsigned(2), - name("extractz"), - m_block); - ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), - name("extractw"), - m_block); - return vectorFromVals(callFloor(x), callFloor(y), - callFloor(z), callFloor(w)); + std::vector vec = extractVector(in); + return vectorFromVals(callFloor(vec[0]), callFloor(vec[1]), + callFloor(vec[2]), callFloor(vec[3])); } llvm::Value * Instructions::arl(llvm::Value *in) @@ -456,70 +423,36 @@ llvm::Value * Instructions::callFLog(llvm::Value *val) llvm::Value * Instructions::lg2(llvm::Value *in) { - ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), - name("extractx"), - m_block); - ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), - name("extracty"), - m_block); - ExtractElementInst *z = new ExtractElementInst(in, unsigned(2), - name("extractz"), - m_block); - ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), - name("extractw"), - m_block); + std::vector vec = extractVector(in); llvm::Value *const_vec = constVector(1.442695f, 1.442695f, 1.442695f, 1.442695f); - return mul(vectorFromVals(callFLog(x), callFLog(y), - callFLog(z), callFLog(w)), const_vec); + return mul(vectorFromVals(callFLog(vec[0]), callFLog(vec[1]), + callFLog(vec[2]), callFLog(vec[3])), const_vec); } llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) { - ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), - name("x1"), - m_block); - ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), - name("y1"), - m_block); - ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), - name("z1"), - m_block); - ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), - name("w1"), - m_block); - - ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), - name("x2"), - m_block); - ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), - name("y2"), - m_block); - ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), - name("z2"), - m_block); - ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), - name("w2"), - m_block); + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OLT, x1, x2, + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[0], vec2[0], name("xcmp"), m_block); - SelectInst *selx = new SelectInst(xcmp, x1, x2, + SelectInst *selx = new SelectInst(xcmp, vec1[0], vec2[0], name("selx"), m_block); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OLT, y1, y2, + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[1], vec2[1], name("ycmp"), m_block); - SelectInst *sely = new SelectInst(ycmp, y1, y2, + SelectInst *sely = new SelectInst(ycmp, vec1[1], vec2[1], name("sely"), m_block); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OLT, z1, z2, + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[2], vec2[2], name("zcmp"), m_block); - SelectInst *selz = new SelectInst(zcmp, z1, z2, + SelectInst *selz = new SelectInst(zcmp, vec1[2], vec2[2], name("selz"), m_block); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OLT, w1, w2, + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[3], vec2[3], name("wcmp"), m_block); - SelectInst *selw = new SelectInst(wcmp, w1, w2, + SelectInst *selw = new SelectInst(wcmp, vec1[3], vec2[3], name("selw"), m_block); return vectorFromVals(selx, sely, selz, selw); @@ -527,50 +460,27 @@ llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) { - ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), - name("x1"), - m_block); - ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), - name("y1"), - m_block); - ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), - name("z1"), - m_block); - ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), - name("w1"), - m_block); + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); - ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), - name("x2"), - m_block); - ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), - name("y2"), - m_block); - ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), - name("z2"), - m_block); - ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), - name("w2"), - m_block); - - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGT, x1, x2, + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[0], vec2[0], name("xcmp"), m_block); - SelectInst *selx = new SelectInst(xcmp, x1, x2, + SelectInst *selx = new SelectInst(xcmp, vec1[0], vec2[0], name("selx"), m_block); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGT, y1, y2, + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[1], vec2[1], name("ycmp"), m_block); - SelectInst *sely = new SelectInst(ycmp, y1, y2, + SelectInst *sely = new SelectInst(ycmp, vec1[1], vec2[1], name("sely"), m_block); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGT, z1, z2, + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[2], vec2[2], name("zcmp"), m_block); - SelectInst *selz = new SelectInst(zcmp, z1, z2, + SelectInst *selz = new SelectInst(zcmp, vec1[2], vec2[2], name("selz"), m_block); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGT, w1, w2, + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[3], vec2[3], name("wcmp"), m_block); - SelectInst *selw = new SelectInst(wcmp, w1, w2, + SelectInst *selw = new SelectInst(wcmp, vec1[3], vec2[3], name("selw"), m_block); return vectorFromVals(selx, sely, selz, selw); @@ -604,22 +514,11 @@ void Instructions::printVector(llvm::Value *val) if (!func_printf) func_printf = declarePrintf(); assert(func_printf); - ExtractElementInst *x = new ExtractElementInst(val, unsigned(0), - name("extractx"), - m_block); - ExtractElementInst *y = new ExtractElementInst(val, unsigned(1), - name("extracty"), - m_block); - ExtractElementInst *z = new ExtractElementInst(val, unsigned(2), - name("extractz"), - m_block); - ExtractElementInst *w = new ExtractElementInst(val, unsigned(3), - name("extractw"), - m_block); - CastInst *dx = new FPExtInst(x, Type::DoubleTy, name("dx"), m_block); - CastInst *dy = new FPExtInst(y, Type::DoubleTy, name("dy"), m_block); - CastInst *dz = new FPExtInst(z, Type::DoubleTy, name("dz"), m_block); - CastInst *dw = new FPExtInst(w, Type::DoubleTy, name("dw"), m_block); + std::vector vec = extractVector(val); + CastInst *dx = new FPExtInst(vec[0], Type::DoubleTy, name("dx"), m_block); + CastInst *dy = new FPExtInst(vec[1], Type::DoubleTy, name("dy"), m_block); + CastInst *dz = new FPExtInst(vec[2], Type::DoubleTy, name("dz"), m_block); + CastInst *dw = new FPExtInst(vec[3], Type::DoubleTy, name("dw"), m_block); std::vector params; params.push_back(m_fmtPtr); params.push_back(dx); @@ -655,24 +554,18 @@ llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); Constant *const0f = Constant::getNullValue(Type::FloatTy); - ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), name("x1"), m_block); - ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), name("x2"), m_block); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGT, x1, x2, name("xcmp"), m_block); + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[0], vec2[0], name("xcmp"), m_block); SelectInst *x = new SelectInst(xcmp, const1f, const0f, name("xsel"), m_block); - ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), name("y1"), m_block); - ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), name("y2"), m_block); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGT, y1, y2, name("ycmp"), m_block); + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[1], vec2[1], name("ycmp"), m_block); SelectInst *y = new SelectInst(ycmp, const1f, const0f, name("ysel"), m_block); - ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), name("z1"), m_block); - ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), name("z2"), m_block); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGT, z1, z2, name("zcmp"), m_block); + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[2], vec2[2], name("zcmp"), m_block); SelectInst *z = new SelectInst(zcmp, const1f, const0f, name("zsel"), m_block); - ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), name("w1"), m_block); - ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), name("w2"), m_block); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGT, w1, w2, name("wcmp"), m_block); + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[3], vec2[3], name("wcmp"), m_block); SelectInst *w = new SelectInst(wcmp, const1f, const0f, name("wsel"), m_block); return vectorFromVals(x, y, z, w); @@ -682,24 +575,19 @@ llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); Constant *const0f = Constant::getNullValue(Type::FloatTy); - ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), name("x1"), m_block); - ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), name("x2"), m_block); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGE, x1, x2, name("xcmp"), m_block); + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGE, vec1[0], vec2[0], name("xcmp"), m_block); SelectInst *x = new SelectInst(xcmp, const1f, const0f, name("xsel"), m_block); - ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), name("y1"), m_block); - ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), name("y2"), m_block); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGE, y1, y2, name("ycmp"), m_block); + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGE, vec1[1], vec2[1], name("ycmp"), m_block); SelectInst *y = new SelectInst(ycmp, const1f, const0f, name("ysel"), m_block); - ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), name("z1"), m_block); - ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), name("z2"), m_block); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGE, z1, z2, name("zcmp"), m_block); + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGE, vec1[2], vec2[2], name("zcmp"), m_block); SelectInst *z = new SelectInst(zcmp, const1f, const0f, name("zsel"), m_block); - ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), name("w1"), m_block); - ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), name("w2"), m_block); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGE, w1, w2, name("wcmp"), m_block); + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGE, vec1[3], vec2[3], name("wcmp"), m_block); SelectInst *w = new SelectInst(wcmp, const1f, const0f, name("wsel"), m_block); return vectorFromVals(x, y, z, w); @@ -711,24 +599,19 @@ llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); Constant *const0f = Constant::getNullValue(Type::FloatTy); - ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), name("x1"), m_block); - ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), name("x2"), m_block); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OLT, x1, x2, name("xcmp"), m_block); + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[0], vec2[0], name("xcmp"), m_block); SelectInst *x = new SelectInst(xcmp, const1f, const0f, name("xsel"), m_block); - ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), name("y1"), m_block); - ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), name("y2"), m_block); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OLT, y1, y2, name("ycmp"), m_block); + FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[1], vec2[1], name("ycmp"), m_block); SelectInst *y = new SelectInst(ycmp, const1f, const0f, name("ysel"), m_block); - ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), name("z1"), m_block); - ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), name("z2"), m_block); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OLT, z1, z2, name("zcmp"), m_block); + FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[2], vec2[2], name("zcmp"), m_block); SelectInst *z = new SelectInst(zcmp, const1f, const0f, name("zsel"), m_block); - ExtractElementInst *w1 = new ExtractElementInst(in1, unsigned(3), name("w1"), m_block); - ExtractElementInst *w2 = new ExtractElementInst(in2, unsigned(3), name("w2"), m_block); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OLT, w1, w2, name("wcmp"), m_block); + FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[3], vec2[3], name("wcmp"), m_block); SelectInst *w = new SelectInst(wcmp, const1f, const0f, name("wsel"), m_block); return vectorFromVals(x, y, z, w); @@ -770,22 +653,11 @@ llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) llvm::Value * Instructions::abs(llvm::Value *in) { - ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), - name("extractx"), - m_block); - ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), - name("extracty"), - m_block); - ExtractElementInst *z = new ExtractElementInst(in, unsigned(2), - name("extractz"), - m_block); - ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), - name("extractw"), - m_block); - Value *xabs = callFAbs(x); - Value *yabs = callFAbs(y); - Value *zabs = callFAbs(z); - Value *wabs = callFAbs(w); + std::vector vec = extractVector(in); + Value *xabs = callFAbs(vec[0]); + Value *yabs = callFAbs(vec[1]); + Value *zabs = callFAbs(vec[2]); + Value *wabs = callFAbs(vec[3]); return vectorFromVals(xabs, yabs, zabs, wabs); } @@ -878,26 +750,15 @@ void Instructions::brk() llvm::Value * Instructions::trunc(llvm::Value *in) { - ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), - name("extractx"), - m_block); - ExtractElementInst *y = new ExtractElementInst(in, unsigned(1), - name("extracty"), - m_block); - ExtractElementInst *z = new ExtractElementInst(in, unsigned(2), - name("extractz"), - m_block); - ExtractElementInst *w = new ExtractElementInst(in, unsigned(3), - name("extractw"), - m_block); - CastInst *icastx = new FPToSIInst(x, IntegerType::get(32), - name("ftoix"), m_block); - CastInst *icasty = new FPToSIInst(y, IntegerType::get(32), - name("ftoiy"), m_block); - CastInst *icastz = new FPToSIInst(z, IntegerType::get(32), - name("ftoiz"), m_block); - CastInst *icastw = new FPToSIInst(w, IntegerType::get(32), - name("ftoiw"), m_block); + std::vector vec = extractVector(in); + CastInst *icastx = new FPToSIInst(vec[0], IntegerType::get(32), + name("ftoix"), m_block); + CastInst *icasty = new FPToSIInst(vec[1], IntegerType::get(32), + name("ftoiy"), m_block); + CastInst *icastz = new FPToSIInst(vec[2], IntegerType::get(32), + name("ftoiz"), m_block); + CastInst *icastw = new FPToSIInst(vec[3], IntegerType::get(32), + name("ftoiw"), m_block); CastInst *fx = new SIToFPInst(icastx, Type::FloatTy, name("fx"), m_block); CastInst *fy = new SIToFPInst(icasty, Type::FloatTy, @@ -997,6 +858,18 @@ llvm::Value * Instructions::constVector(float x, float y, float z, float w) return ConstantVector::get(m_floatVecType, vec); } + +std::vector Instructions::extractVector(llvm::Value *vec) +{ + std::vector elems(4); + elems[0] = new ExtractElementInst(vec, unsigned(0), name("x"), m_block); + elems[1] = new ExtractElementInst(vec, unsigned(1), name("y"), m_block); + elems[2] = new ExtractElementInst(vec, unsigned(2), name("z"), m_block); + elems[3] = new ExtractElementInst(vec, unsigned(3), name("w"), m_block); + return elems; +} + #endif //MESA_LLVM + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index b7b5129694..85ffc7b920 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -113,6 +113,8 @@ private: llvm::Function *declareFunc(int label); llvm::Function *findFunction(int label); + + std::vector extractVector(llvm::Value *vec); private: llvm::Module *m_mod; llvm::Function *m_func; -- cgit v1.2.3 From 449562cde0008c755d20b3de83cbd3266bc001ac Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 30 Oct 2007 09:00:36 -0400 Subject: Switch to using LLVM builder. Instead of creating all the instructions by hand, switching to using LLVMBuilder. --- src/mesa/pipe/llvm/instructions.cpp | 449 ++++++++++++++++-------------------- src/mesa/pipe/llvm/instructions.h | 17 +- src/mesa/pipe/llvm/llvmtgsi.cpp | 4 +- 3 files changed, 217 insertions(+), 253 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index dca5f4c995..aa6e108598 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -58,8 +58,10 @@ static inline std::string createFuncName(int label) return stream.str(); } -Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block) - : m_mod(mod), m_func(func), m_block(block), m_idx(0) +Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, + Storage *storage) + : m_mod(mod), m_func(func), m_builder(block), m_idx(0), + m_storage(storage) { m_floatVecType = VectorType::get(Type::FloatTy, 4); @@ -76,10 +78,7 @@ Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicB llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) { - BinaryOperator *res = BinaryOperator::create(Instruction::Add, in1, in2, - name("add"), - m_block); - return res; + return m_builder.CreateAdd(in1, in2, name("add")); } llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, @@ -91,10 +90,7 @@ llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2) { - BinaryOperator *res = BinaryOperator::create(Instruction::Mul, in1, in2, - name("mul"), - m_block); - return res; + return m_builder.CreateMul(in1, in2, name("mul")); } const char * Instructions::name(const char *prefix) @@ -107,21 +103,17 @@ const char * Instructions::name(const char *prefix) llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2) { Value *mulRes = mul(in1, in2); - ExtractElementInst *x = new ExtractElementInst(mulRes, unsigned(0), - name("extractx"), - m_block); - ExtractElementInst *y = new ExtractElementInst(mulRes, unsigned(1), - name("extracty"), - m_block); - ExtractElementInst *z = new ExtractElementInst(mulRes, unsigned(2), - name("extractz"), - m_block); - BinaryOperator *xy = BinaryOperator::create(Instruction::Add, x, y, - name("xy"), - m_block); - BinaryOperator *dot3 = BinaryOperator::create(Instruction::Add, xy, z, - name("dot3"), - m_block); + ExtractElementInst *x = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(0), + name("extractx")); + ExtractElementInst *y = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(1), + name("extracty")); + ExtractElementInst *z = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(2), + name("extractz")); + Value *xy = m_builder.CreateAdd(x, y,name("xy")); + Value *dot3 = m_builder.CreateAdd(xy, z, name("dot3")); return vectorFromVals(dot3, dot3, dot3, dot3); } @@ -143,9 +135,8 @@ llvm::Value *Instructions::callFSqrt(llvm::Value *val) /*Name=*/"llvm.sqrt.f32", m_mod); m_llvmFSqrt->setCallingConv(CallingConv::C); } - CallInst *call = new CallInst(m_llvmFSqrt, val, - name("sqrt"), - m_block); + CallInst *call = m_builder.CreateCall(m_llvmFSqrt, val, + name("sqrt")); call->setCallingConv(CallingConv::C); call->setTailCall(false); return call; @@ -153,18 +144,16 @@ llvm::Value *Instructions::callFSqrt(llvm::Value *val) llvm::Value * Instructions::rsq(llvm::Value *in1) { - ExtractElementInst *x = new ExtractElementInst(in1, unsigned(0), - name("extractx"), - m_block); + ExtractElementInst *x = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("extractx")); Value *abs = callFAbs(x); Value *sqrt = callFSqrt(abs); - BinaryOperator *rsqrt = BinaryOperator::create(Instruction::FDiv, - ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - sqrt, - name("rsqrt"), - m_block); + Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + sqrt, + name("rsqrt")); return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt); } @@ -172,18 +161,16 @@ llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, llvm::Value *z, llvm::Value *w) { Constant *const_vec = Constant::getNullValue(m_floatVecType); - InsertElementInst *res = new InsertElementInst(const_vec, x, unsigned(0), - name("vecx"), m_block); - res = new InsertElementInst(res, y, unsigned(1), - name("vecxy"), - m_block); - res = new InsertElementInst(res, z, unsigned(2), - name("vecxyz"), - m_block); + InsertElementInst *res = m_builder.CreateInsertElement(const_vec, x, + m_storage->constantInt(0), + name("vecx")); + res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), + name("vecxy")); + res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), + name("vecxyz")); if (w) - res = new InsertElementInst(res, w, unsigned(3), - name("vecxyzw"), - m_block); + res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), + name("vecxyzw")); return res; } @@ -205,9 +192,8 @@ llvm::Value *Instructions::callFAbs(llvm::Value *val) /*Name=*/"fabs", m_mod); m_llvmFAbs->setCallingConv(CallingConv::C); } - CallInst *call = new CallInst(m_llvmFAbs, val, - name("fabs"), - m_block); + CallInst *call = m_builder.CreateCall(m_llvmFAbs, val, + name("fabs")); call->setCallingConv(CallingConv::C); call->setTailCall(false); return call; @@ -218,7 +204,7 @@ llvm::Value * Instructions::lit(llvm::Value *in) if (!m_llvmLit) { m_llvmLit = m_mod->getFunction("lit"); } - CallInst *call = new CallInst(m_llvmLit, in, name("litres"), m_block); + CallInst *call = m_builder.CreateCall(m_llvmLit, in, name("litres")); call->setCallingConv(CallingConv::C); call->setTailCall(false); return call; @@ -226,9 +212,7 @@ llvm::Value * Instructions::lit(llvm::Value *in) llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) { - BinaryOperator *res = BinaryOperator::create(Instruction::Sub, in1, in2, - name("sub"), - m_block); + Value *res = m_builder.CreateSub(in1, in2, name("sub")); return res; } @@ -254,9 +238,8 @@ llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) std::vector params; params.push_back(val1); params.push_back(val2); - CallInst *call = new CallInst(m_llvmPow, params.begin(), params.end(), - name("pow"), - m_block); + CallInst *call = m_builder.CreateCall(m_llvmPow, params.begin(), params.end(), + name("pow")); call->setCallingConv(CallingConv::C); call->setTailCall(false); return call; @@ -264,27 +247,24 @@ llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2) { - ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), - name("x1"), - m_block); - ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), - name("x2"), - m_block); + ExtractElementInst *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + ExtractElementInst *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); llvm::Value *val = callPow(x1, x2); return vectorFromVals(val, val, val, val); } llvm::Value * Instructions::rcp(llvm::Value *in1) { - ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), - name("x1"), - m_block); - BinaryOperator *res = BinaryOperator::create(Instruction::FDiv, - ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - x1, - name("rcp"), - m_block); + ExtractElementInst *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *res = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + x1, name("rcp")); return vectorFromVals(res, res, res, res); } @@ -292,15 +272,9 @@ llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) { Value *mulRes = mul(in1, in2); std::vector vec = extractVector(mulRes); - BinaryOperator *xy = BinaryOperator::create(Instruction::Add, vec[0], vec[1], - name("xy"), - m_block); - BinaryOperator *xyz = BinaryOperator::create(Instruction::Add, xy, vec[2], - name("xyz"), - m_block); - BinaryOperator *dot4 = BinaryOperator::create(Instruction::Add, xyz, vec[3], - name("dot4"), - m_block); + Value *xy = m_builder.CreateAdd(vec[0], vec[1], name("xy")); + Value *xyz = m_builder.CreateAdd(xy, vec[2], name("xyz")); + Value *dot4 = m_builder.CreateAdd(xyz, vec[3], name("dot4")); return vectorFromVals(dot4, dot4, dot4, dot4); } @@ -308,36 +282,27 @@ llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) { Value *mulRes = mul(in1, in2); std::vector vec1 = extractVector(mulRes); - BinaryOperator *xy = BinaryOperator::create(Instruction::Add, vec1[0], vec1[1], - name("xy"), - m_block); - BinaryOperator *xyz = BinaryOperator::create(Instruction::Add, xy, vec1[2], - name("xyz"), - m_block); - BinaryOperator *dph = BinaryOperator::create(Instruction::Add, xyz, vec1[3], - name("dph"), - m_block); + Value *xy = m_builder.CreateAdd(vec1[0], vec1[1], name("xy")); + Value *xyz = m_builder.CreateAdd(xy, vec1[2], name("xyz")); + Value *dph = m_builder.CreateAdd(xyz, vec1[3], name("dph")); return vectorFromVals(dph, dph, dph, dph); } llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) { - ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), - name("y1"), - m_block); - ExtractElementInst *z = new ExtractElementInst(in1, unsigned(2), - name("z"), - m_block); - ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), - name("y2"), - m_block); - ExtractElementInst *w = new ExtractElementInst(in2, unsigned(3), - name("w"), - m_block); - BinaryOperator *ry = BinaryOperator::create(Instruction::Mul, - y1, y2, - name("tyuy"), - m_block); + ExtractElementInst *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + ExtractElementInst *z = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z")); + ExtractElementInst *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + ExtractElementInst *w = m_builder.CreateExtractElement(in2, + m_storage->constantInt(3), + name("w")); + Value *ry = m_builder.CreateMul(y1, y2, name("tyuy")); return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), ry, z, w); } @@ -345,9 +310,9 @@ llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) llvm::Value * Instructions::ex2(llvm::Value *in) { llvm::Value *val = callPow(ConstantFP::get(Type::FloatTy, APFloat(2.f)), - new ExtractElementInst(in, unsigned(0), - name("x1"), - m_block)); + m_builder.CreateExtractElement( + in, m_storage->constantInt(0), + name("x1"))); return vectorFromVals(val, val, val, val); } @@ -369,9 +334,8 @@ llvm::Value * Instructions::callFloor(llvm::Value *val) /*Name=*/"floorf", m_mod); m_llvmFloor->setCallingConv(CallingConv::C); } - CallInst *call = new CallInst(m_llvmFloor, val, - name("floorf"), - m_block); + CallInst *call = m_builder.CreateCall(m_llvmFloor, val, + name("floorf")); call->setCallingConv(CallingConv::C); call->setTailCall(false); return call; @@ -413,9 +377,8 @@ llvm::Value * Instructions::callFLog(llvm::Value *val) /*Name=*/"logf", m_mod); m_llvmFlog->setCallingConv(CallingConv::C); } - CallInst *call = new CallInst(m_llvmFlog, val, - name("logf"), - m_block); + CallInst *call = m_builder.CreateCall(m_llvmFlog, val, + name("logf")); call->setCallingConv(CallingConv::C); call->setTailCall(false); return call; @@ -435,25 +398,21 @@ llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) std::vector vec1 = extractVector(in1); std::vector vec2 = extractVector(in2); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[0], vec2[0], - name("xcmp"), m_block); - SelectInst *selx = new SelectInst(xcmp, vec1[0], vec2[0], - name("selx"), m_block); + Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); + SelectInst *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[1], vec2[1], - name("ycmp"), m_block); - SelectInst *sely = new SelectInst(ycmp, vec1[1], vec2[1], - name("sely"), m_block); + Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); + SelectInst *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[2], vec2[2], - name("zcmp"), m_block); - SelectInst *selz = new SelectInst(zcmp, vec1[2], vec2[2], - name("selz"), m_block); + Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); + SelectInst *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[3], vec2[3], - name("wcmp"), m_block); - SelectInst *selw = new SelectInst(wcmp, vec1[3], vec2[3], - name("selw"), m_block); + Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); + SelectInst *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); return vectorFromVals(selx, sely, selz, selw); } @@ -463,25 +422,25 @@ llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) std::vector vec1 = extractVector(in1); std::vector vec2 = extractVector(in2); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[0], vec2[0], - name("xcmp"), m_block); - SelectInst *selx = new SelectInst(xcmp, vec1[0], vec2[0], - name("selx"), m_block); + Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], + name("xcmp")); + SelectInst *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[1], vec2[1], - name("ycmp"), m_block); - SelectInst *sely = new SelectInst(ycmp, vec1[1], vec2[1], - name("sely"), m_block); + Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], + name("ycmp")); + SelectInst *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[2], vec2[2], - name("zcmp"), m_block); - SelectInst *selz = new SelectInst(zcmp, vec1[2], vec2[2], - name("selz"), m_block); + Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], + name("zcmp")); + SelectInst *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[3], vec2[3], - name("wcmp"), m_block); - SelectInst *selw = new SelectInst(wcmp, vec1[3], vec2[3], - name("selw"), m_block); + Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], + name("wcmp")); + SelectInst *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); return vectorFromVals(selx, sely, selz, selw); } @@ -515,18 +474,18 @@ void Instructions::printVector(llvm::Value *val) func_printf = declarePrintf(); assert(func_printf); std::vector vec = extractVector(val); - CastInst *dx = new FPExtInst(vec[0], Type::DoubleTy, name("dx"), m_block); - CastInst *dy = new FPExtInst(vec[1], Type::DoubleTy, name("dy"), m_block); - CastInst *dz = new FPExtInst(vec[2], Type::DoubleTy, name("dz"), m_block); - CastInst *dw = new FPExtInst(vec[3], Type::DoubleTy, name("dw"), m_block); + CastInst *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx")); + CastInst *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy")); + CastInst *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz")); + CastInst *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw")); std::vector params; params.push_back(m_fmtPtr); params.push_back(dx); params.push_back(dy); params.push_back(dz); params.push_back(dw); - CallInst* call = new CallInst(func_printf, params.begin(), params.end(), - name("printf"), m_block); + CallInst* call = m_builder.CreateCall(func_printf, params.begin(), params.end(), + name("printf")); call->setCallingConv(CallingConv::C); call->setTailCall(true); } @@ -556,17 +515,17 @@ llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) std::vector vec1 = extractVector(in1); std::vector vec2 = extractVector(in2); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[0], vec2[0], name("xcmp"), m_block); - SelectInst *x = new SelectInst(xcmp, const1f, const0f, name("xsel"), m_block); + Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], name("xcmp")); + SelectInst *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[1], vec2[1], name("ycmp"), m_block); - SelectInst *y = new SelectInst(ycmp, const1f, const0f, name("ysel"), m_block); + Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], name("ycmp")); + SelectInst *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[2], vec2[2], name("zcmp"), m_block); - SelectInst *z = new SelectInst(zcmp, const1f, const0f, name("zsel"), m_block); + Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], name("zcmp")); + SelectInst *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGT, vec1[3], vec2[3], name("wcmp"), m_block); - SelectInst *w = new SelectInst(wcmp, const1f, const0f, name("wsel"), m_block); + Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], name("wcmp")); + SelectInst *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); return vectorFromVals(x, y, z, w); } @@ -578,17 +537,17 @@ llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) std::vector vec1 = extractVector(in1); std::vector vec2 = extractVector(in2); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OGE, vec1[0], vec2[0], name("xcmp"), m_block); - SelectInst *x = new SelectInst(xcmp, const1f, const0f, name("xsel"), m_block); + Value *xcmp = m_builder.CreateFCmpOGE(vec1[0], vec2[0], name("xcmp")); + SelectInst *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OGE, vec1[1], vec2[1], name("ycmp"), m_block); - SelectInst *y = new SelectInst(ycmp, const1f, const0f, name("ysel"), m_block); + Value *ycmp = m_builder.CreateFCmpOGE(vec1[1], vec2[1], name("ycmp")); + SelectInst *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OGE, vec1[2], vec2[2], name("zcmp"), m_block); - SelectInst *z = new SelectInst(zcmp, const1f, const0f, name("zsel"), m_block); + Value *zcmp = m_builder.CreateFCmpOGE(vec1[2], vec2[2], name("zcmp")); + SelectInst *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OGE, vec1[3], vec2[3], name("wcmp"), m_block); - SelectInst *w = new SelectInst(wcmp, const1f, const0f, name("wsel"), m_block); + Value *wcmp = m_builder.CreateFCmpOGE(vec1[3], vec2[3], name("wcmp")); + SelectInst *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); return vectorFromVals(x, y, z, w); } @@ -602,42 +561,42 @@ llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) std::vector vec1 = extractVector(in1); std::vector vec2 = extractVector(in2); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[0], vec2[0], name("xcmp"), m_block); - SelectInst *x = new SelectInst(xcmp, const1f, const0f, name("xsel"), m_block); + Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); + SelectInst *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - FCmpInst *ycmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[1], vec2[1], name("ycmp"), m_block); - SelectInst *y = new SelectInst(ycmp, const1f, const0f, name("ysel"), m_block); + Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); + SelectInst *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - FCmpInst *zcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[2], vec2[2], name("zcmp"), m_block); - SelectInst *z = new SelectInst(zcmp, const1f, const0f, name("zsel"), m_block); + Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); + SelectInst *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - FCmpInst *wcmp = new FCmpInst(FCmpInst::FCMP_OLT, vec1[3], vec2[3], name("wcmp"), m_block); - SelectInst *w = new SelectInst(wcmp, const1f, const0f, name("wsel"), m_block); + Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); + SelectInst *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); return vectorFromVals(x, y, z, w); } llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) { - ExtractElementInst *x1 = new ExtractElementInst(in1, unsigned(0), - name("x1"), - m_block); - ExtractElementInst *y1 = new ExtractElementInst(in1, unsigned(1), - name("y1"), - m_block); - ExtractElementInst *z1 = new ExtractElementInst(in1, unsigned(2), - name("z1"), - m_block); - - ExtractElementInst *x2 = new ExtractElementInst(in2, unsigned(0), - name("x2"), - m_block); - ExtractElementInst *y2 = new ExtractElementInst(in2, unsigned(1), - name("y2"), - m_block); - ExtractElementInst *z2 = new ExtractElementInst(in2, unsigned(2), - name("z2"), - m_block); + ExtractElementInst *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + ExtractElementInst *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + ExtractElementInst *z1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z1")); + + ExtractElementInst *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); + ExtractElementInst *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + ExtractElementInst *z2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(2), + name("z2")); Value *y1z2 = mul(y1, z2); Value *z1y2 = mul(z1, y2); @@ -672,30 +631,28 @@ void Instructions::ifop(llvm::Value *in) Constant *float0 = Constant::getNullValue(Type::FloatTy); - ExtractElementInst *x = new ExtractElementInst(in, unsigned(0), name("extractx"), - m_block); - FCmpInst *xcmp = new FCmpInst(FCmpInst::FCMP_UNE, x, float0, - name("xcmp"), m_block); - new BranchInst(ifthen, ifend, xcmp, m_block); - //m_block = yblock; - + ExtractElementInst *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0), + name("extractx")); + Value *xcmp = m_builder.CreateFCmpUNE(x, float0, name("xcmp")); + m_builder.CreateCondBr(xcmp, ifthen, ifend); + //m_builder.SetInsertPoint(yblock); - m_block = ifthen; + m_builder.SetInsertPoint(ifthen); m_ifStack.push(ifend); } llvm::BasicBlock * Instructions::currentBlock() const { - return m_block; + return m_builder.GetInsertBlock(); } void Instructions::elseop() { assert(!m_ifStack.empty()); BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); - new BranchInst(ifend, m_block); - m_block = m_ifStack.top(); - m_block->setName(name("ifelse")); + m_builder.CreateBr(ifend); + m_builder.SetInsertPoint(m_ifStack.top()); + currentBlock()->setName(name("ifelse")); m_ifStack.pop(); m_ifStack.push(ifend); } @@ -703,8 +660,8 @@ void Instructions::elseop() void Instructions::endif() { assert(!m_ifStack.empty()); - new BranchInst(m_ifStack.top(), m_block); - m_block = m_ifStack.top(); + m_builder.CreateBr(m_ifStack.top()); + m_builder.SetInsertPoint(m_ifStack.top()); m_ifStack.pop(); } @@ -722,11 +679,11 @@ void Instructions::beginLoop() BasicBlock *begin = new BasicBlock(name("loop"), m_func,0); BasicBlock *end = new BasicBlock(name("endloop"), m_func,0); - new BranchInst(begin, m_block); + m_builder.CreateBr(begin); Loop loop; loop.begin = begin; loop.end = end; - m_block = begin; + m_builder.SetInsertPoint(begin); m_loopStack.push(loop); } @@ -734,9 +691,9 @@ void Instructions::endLoop() { assert(!m_loopStack.empty()); Loop loop = m_loopStack.top(); - new BranchInst(loop.begin, m_block); - loop.end->moveAfter(m_block); - m_block = loop.end; + m_builder.CreateBr(loop.begin); + loop.end->moveAfter(currentBlock()); + m_builder.SetInsertPoint(loop.end); m_loopStack.pop(); } @@ -744,35 +701,35 @@ void Instructions::brk() { assert(!m_loopStack.empty()); BasicBlock *unr = new BasicBlock(name("unreachable"), m_func,0); - new BranchInst(m_loopStack.top().end, m_block); - m_block = unr; + m_builder.CreateBr(m_loopStack.top().end); + m_builder.SetInsertPoint(unr); } llvm::Value * Instructions::trunc(llvm::Value *in) { std::vector vec = extractVector(in); - CastInst *icastx = new FPToSIInst(vec[0], IntegerType::get(32), - name("ftoix"), m_block); - CastInst *icasty = new FPToSIInst(vec[1], IntegerType::get(32), - name("ftoiy"), m_block); - CastInst *icastz = new FPToSIInst(vec[2], IntegerType::get(32), - name("ftoiz"), m_block); - CastInst *icastw = new FPToSIInst(vec[3], IntegerType::get(32), - name("ftoiw"), m_block); - CastInst *fx = new SIToFPInst(icastx, Type::FloatTy, - name("fx"), m_block); - CastInst *fy = new SIToFPInst(icasty, Type::FloatTy, - name("fy"), m_block); - CastInst *fz = new SIToFPInst(icastz, Type::FloatTy, - name("fz"), m_block); - CastInst *fw = new SIToFPInst(icastw, Type::FloatTy, - name("fw"), m_block); + CastInst *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32), + name("ftoix")); + CastInst *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32), + name("ftoiy")); + CastInst *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32), + name("ftoiz")); + CastInst *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32), + name("ftoiw")); + CastInst *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy, + name("fx")); + CastInst *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy, + name("fy")); + CastInst *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy, + name("fz")); + CastInst *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy, + name("fw")); return vectorFromVals(fx, fy, fz, fw); } void Instructions::end() { - new ReturnInst(m_block); + m_builder.CreateRetVoid(); } void Instructions::cal(int label, llvm::Value *out, llvm::Value *in, @@ -785,7 +742,7 @@ void Instructions::cal(int label, llvm::Value *out, llvm::Value *in, params.push_back(temp); llvm::Function *func = findFunction(label); - new CallInst(func, params.begin(), params.end(), std::string(), m_block); + m_builder.CreateCall(func, params.begin(), params.end()); } llvm::Function * Instructions::declareFunc(int label) @@ -811,7 +768,7 @@ llvm::Function * Instructions::declareFunc(int label) return func; } -void Instructions::bgnSub(unsigned label, Storage *storage) +void Instructions::bgnSub(unsigned label) { llvm::Function *func = findFunction(label); @@ -824,18 +781,18 @@ void Instructions::bgnSub(unsigned label, Storage *storage) ptr_CONST->setName("CONST"); Value *ptr_TEMP = args++; ptr_TEMP->setName("TEMP"); - storage->pushArguments(ptr_OUT, ptr_IN, ptr_CONST, ptr_TEMP); + m_storage->pushArguments(ptr_OUT, ptr_IN, ptr_CONST, ptr_TEMP); llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); m_func = func; - m_block = entry; + m_builder.SetInsertPoint(entry); } void Instructions::endSub() { m_func = 0; - m_block = 0; + m_builder.SetInsertPoint(0); } llvm::Function * Instructions::findFunction(int label) @@ -862,10 +819,14 @@ llvm::Value * Instructions::constVector(float x, float y, float z, float w) std::vector Instructions::extractVector(llvm::Value *vec) { std::vector elems(4); - elems[0] = new ExtractElementInst(vec, unsigned(0), name("x"), m_block); - elems[1] = new ExtractElementInst(vec, unsigned(1), name("y"), m_block); - elems[2] = new ExtractElementInst(vec, unsigned(2), name("z"), m_block); - elems[3] = new ExtractElementInst(vec, unsigned(3), name("w"), m_block); + elems[0] = m_builder.CreateExtractElement(vec, m_storage->constantInt(0), + name("x")); + elems[1] = m_builder.CreateExtractElement(vec, m_storage->constantInt(1), + name("y")); + elems[2] = m_builder.CreateExtractElement(vec, m_storage->constantInt(2), + name("z")); + elems[3] = m_builder.CreateExtractElement(vec, m_storage->constantInt(3), + name("w")); return elems; } diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 85ffc7b920..83d07e2f2e 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -36,6 +36,7 @@ #include #include #include +#include #include #include @@ -50,7 +51,8 @@ class Storage; class Instructions { public: - Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block); + Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, + Storage *storage); llvm::BasicBlock *currentBlock() const; @@ -58,7 +60,7 @@ public: llvm::Value *arl(llvm::Value *in1); llvm::Value *add(llvm::Value *in1, llvm::Value *in2); void beginLoop(); - void bgnSub(unsigned, Storage *); + void bgnSub(unsigned); void brk(); void cal(int label, llvm::Value *out, llvm::Value *in, llvm::Value *cst, llvm::Value *tmp); @@ -116,11 +118,11 @@ private: std::vector extractVector(llvm::Value *vec); private: - llvm::Module *m_mod; - llvm::Function *m_func; - char m_name[32]; - llvm::BasicBlock *m_block; - int m_idx; + llvm::Module *m_mod; + llvm::Function *m_func; + char m_name[32]; + llvm::LLVMFoldingBuilder m_builder; + int m_idx; llvm::VectorType *m_floatVecType; @@ -140,6 +142,7 @@ private: }; std::stack m_loopStack; std::map m_functions; + Storage *m_storage; }; #endif diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp index e7cfeb2c91..c787f3ab9f 100644 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ b/src/mesa/pipe/llvm/llvmtgsi.cpp @@ -517,7 +517,7 @@ translate_instruction(llvm::Module *module, } break; case TGSI_OPCODE_BGNSUB: { - instr->bgnSub(instno, storage); + instr->bgnSub(instno); storage->setCurrentBlock(instr->currentBlock()); storage->pushTemps(); return; @@ -679,7 +679,7 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); Storage storage(label_entry, ptr_OUT, ptr_IN, ptr_CONST, ptr_TEMPS); - Instructions instr(mod, shader, label_entry); + Instructions instr(mod, shader, label_entry, &storage); while(!tgsi_parse_end_of_tokens(&parse)) { tgsi_parse_token(&parse); -- cgit v1.2.3 From 4b30d177b47953023b77494ef31fb1399beb7405 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 16:00:08 -0600 Subject: added _mesa_combine_parameter_lists() --- src/mesa/shader/prog_parameter.c | 33 +++++++++++++++++++++++++++++++++ src/mesa/shader/prog_parameter.h | 10 ++++++++++ 2 files changed, 43 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index b4008abbd9..bb964274b9 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -607,6 +607,39 @@ _mesa_clone_parameter_list(const struct gl_program_parameter_list *list) } +/** + * Return a new parameter list which is listA + listB. + */ +struct gl_program_parameter_list * +_mesa_combine_parameter_lists(const struct gl_program_parameter_list *listA, + const struct gl_program_parameter_list *listB) +{ + struct gl_program_parameter_list *list; + + if (listA) { + list = _mesa_clone_parameter_list(listA); + if (list && listB) { + GLuint i; + for (i = 0; i < listB->NumParameters; i++) { + struct gl_program_parameter *param = listB->Parameters + i; + _mesa_add_parameter(list, param->Type, param->Name, param->Size, + param->DataType, + listB->ParameterValues[i], + param->StateIndexes); + } + } + } + else if (listB) { + list = _mesa_clone_parameter_list(listB); + } + else { + list = NULL; + } + return list; +} + + + /** * Find longest name of all uniform parameters in list. */ diff --git a/src/mesa/shader/prog_parameter.h b/src/mesa/shader/prog_parameter.h index 40c8c09e09..d6cc03448c 100644 --- a/src/mesa/shader/prog_parameter.h +++ b/src/mesa/shader/prog_parameter.h @@ -78,6 +78,16 @@ _mesa_free_parameter_list(struct gl_program_parameter_list *paramList); extern struct gl_program_parameter_list * _mesa_clone_parameter_list(const struct gl_program_parameter_list *list); +extern struct gl_program_parameter_list * +_mesa_combine_parameter_lists(const struct gl_program_parameter_list *a, + const struct gl_program_parameter_list *b); + +static INLINE GLuint +_mesa_num_parameters(const struct gl_program_parameter_list *list) +{ + return list ? list->NumParameters : 0; +} + extern GLint _mesa_add_parameter(struct gl_program_parameter_list *paramList, enum register_file type, const char *name, -- cgit v1.2.3 From 9ffd88911f69babaa2214fb27f1ab66abd282639 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 16:35:59 -0600 Subject: Added _mesa_combine_programs() for concatenating two programs. --- src/mesa/shader/program.c | 149 ++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/shader/program.h | 8 +++ 2 files changed, 157 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index cafc0dcfaa..0e420179bc 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -409,6 +409,155 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog) +/** + * Search instructions for registers that match (oldFile, oldIndex), + * replacing them with (newFile, newIndex). + */ +static void +replace_registers(struct prog_instruction *inst, GLuint numInst, + GLuint oldFile, GLuint oldIndex, + GLuint newFile, GLuint newIndex) +{ + GLuint i, j; + for (i = 0; i < numInst; i++) { + for (j = 0; j < _mesa_num_inst_src_regs(inst->Opcode); j++) { + if (inst[i].SrcReg[j].File == oldFile && + inst[i].SrcReg[j].Index == oldIndex) { + inst[i].SrcReg[j].File = newFile; + inst[i].SrcReg[j].Index = newIndex; + } + } + } +} + + +/** + * Search instructions for references to program parameters. When found, + * increment the parameter index by 'offset'. + * Used when combining programs. + */ +static void +adjust_param_indexes(struct prog_instruction *inst, GLuint numInst, + GLuint offset) +{ + GLuint i, j; + for (i = 0; i < numInst; i++) { + for (j = 0; j < _mesa_num_inst_src_regs(inst->Opcode); j++) { + GLuint f = inst[i].SrcReg[j].File; + if (f == PROGRAM_CONSTANT || + f == PROGRAM_UNIFORM || + f == PROGRAM_STATE_VAR) { + inst[i].SrcReg[j].Index += offset; + } + } + } +} + + +/** + * Combine two programs into one. Fix instructions so the outputs of + * the first program go to the inputs of the second program. + */ +struct gl_program * +_mesa_combine_programs(GLcontext *ctx, + struct gl_program *progA, struct gl_program *progB) +{ + struct prog_instruction *newInst; + struct gl_program *newProg; + const GLuint lenA = progA->NumInstructions - 1; /* omit END instr */ + const GLuint lenB = progB->NumInstructions; + const GLuint numParamsA = _mesa_num_parameters(progA->Parameters); + const GLuint newLength = lenA + lenB; + GLuint i; + + ASSERT(progA->Target == progB->Target); + + newInst = _mesa_alloc_instructions(newLength); + if (!newInst) + return GL_FALSE; + + _mesa_copy_instructions(newInst, progA->Instructions, lenA); + _mesa_copy_instructions(newInst + lenA, progB->Instructions, lenB); + + /* adjust branch / instruction addresses for B's instructions */ + for (i = 0; i < lenB; i++) { + newInst[lenA + i].BranchTarget += lenA; + } + + newProg = ctx->Driver.NewProgram(ctx, progA->Target, 0); + newProg->Instructions = newInst; + newProg->NumInstructions = newLength; + + if (newProg->Target == GL_FRAGMENT_PROGRAM_ARB) { + /* connect color outputs/inputs */ + if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) && + (progB->InputsRead & (1 << FRAG_ATTRIB_COL0))) { + replace_registers(newInst + lenA, lenB, + PROGRAM_INPUT, FRAG_ATTRIB_COL0, + PROGRAM_OUTPUT, FRAG_RESULT_COLR); + } + + newProg->InputsRead = progA->InputsRead; + newProg->InputsRead |= (progB->InputsRead & ~(1 << FRAG_ATTRIB_COL0)); + newProg->OutputsWritten = progB->OutputsWritten; + } + else { + /* vertex program */ + assert(0); /* XXX todo */ + } + + /* + * Merge parameters (uniforms, constants, etc) + */ + newProg->Parameters = _mesa_combine_parameter_lists(progA->Parameters, + progB->Parameters); + + adjust_param_indexes(newInst + lenA, lenB, numParamsA); + + + return newProg; +} + + + + +/** + * Scan the given program to find a free register of the given type. + * \param regFile - PROGRAM_INPUT, PROGRAM_OUTPUT or PROGRAM_TEMPORARY + */ +GLint +_mesa_find_free_register(const struct gl_program *prog, GLuint regFile) +{ + GLboolean used[MAX_PROGRAM_TEMPS]; + GLuint i, k; + + assert(regFile == PROGRAM_INPUT || + regFile == PROGRAM_OUTPUT || + regFile == PROGRAM_TEMPORARY); + + _mesa_memset(used, 0, sizeof(used)); + + for (i = 0; i < prog->NumInstructions; i++) { + const struct prog_instruction *inst = prog->Instructions + i; + const GLuint n = _mesa_num_inst_src_regs(inst->Opcode); + + for (k = 0; k < n; k++) { + if (inst->SrcReg[k].File == regFile) { + used[inst->SrcReg[k].Index] = GL_TRUE; + } + } + } + + for (i = 0; i < MAX_PROGRAM_TEMPS; i++) { + if (!used[i]) + return i; + } + + return -1; +} + + + /** * Mixing ARB and NV vertex/fragment programs can be tricky. * Note: GL_VERTEX_PROGRAM_ARB == GL_VERTEX_PROGRAM_NV diff --git a/src/mesa/shader/program.h b/src/mesa/shader/program.h index 52704856ab..02faa706d9 100644 --- a/src/mesa/shader/program.h +++ b/src/mesa/shader/program.h @@ -88,6 +88,14 @@ extern struct gl_program * _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog); +extern struct gl_program * +_mesa_combine_programs(GLcontext *ctx, + struct gl_program *progA, struct gl_program *progB); + +extern GLint +_mesa_find_free_register(const struct gl_program *prog, GLuint regFile); + + /* * API functions common to ARB/NV_vertex/fragment_program */ -- cgit v1.2.3 From cf56a99fb03e2c6e2739a6a9dbd6230c607813dc Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 16:37:49 -0600 Subject: Initial version of st_get_pixel_transfer_program(). For generating fragment programs which implement pixel transfer ops. --- src/mesa/state_tracker/st_pixeltransfer.c | 131 ++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_pixeltransfer.h | 36 ++++++++ 2 files changed, 167 insertions(+) create mode 100644 src/mesa/state_tracker/st_pixeltransfer.c create mode 100644 src/mesa/state_tracker/st_pixeltransfer.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_pixeltransfer.c b/src/mesa/state_tracker/st_pixeltransfer.c new file mode 100644 index 0000000000..52e2a4c89d --- /dev/null +++ b/src/mesa/state_tracker/st_pixeltransfer.c @@ -0,0 +1,131 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * Generate fragment programs to implement pixel transfer ops, such as + * scale/bias, colormatrix, colortable, convolution... + * + * Authors: + * Brian Paul + */ + +#include "main/imports.h" +#include "main/image.h" +#include "main/macros.h" +#include "shader/program.h" +#include "shader/prog_instruction.h" +#include "shader/prog_parameter.h" +#include "shader/prog_print.h" + +#include "st_pixeltransfer.h" + + +#define MAX_INST 100 + +/** + * Returns a fragment program which implements the current pixel transfer ops. + */ +struct gl_fragment_program * +st_get_pixel_transfer_program(GLcontext *ctx) +{ + struct prog_instruction inst[MAX_INST]; + struct gl_program_parameter_list *params; + struct gl_fragment_program *fp; + GLuint ic = 0; + + params = _mesa_new_parameter_list(); + + /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_TEX; + inst[ic].DstReg.File = PROGRAM_OUTPUT; + inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].SrcReg[0].File = PROGRAM_INPUT; + inst[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + inst[ic].TexSrcUnit = 0; + inst[ic].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; + + /* MAD result.color, result.color, scale, bias; */ + if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 || + ctx->Pixel.GreenBias != 0.0 || ctx->Pixel.GreenScale != 1.0 || + ctx->Pixel.BlueBias != 0.0 || ctx->Pixel.BlueScale != 1.0 || + ctx->Pixel.AlphaBias != 0.0 || ctx->Pixel.AlphaScale != 1.0) { + GLfloat scale[4], bias[4]; + GLint scale_p, bias_p; + scale[0] = ctx->Pixel.RedScale; + scale[1] = ctx->Pixel.GreenScale; + scale[2] = ctx->Pixel.BlueScale; + scale[3] = ctx->Pixel.AlphaScale; + bias[0] = ctx->Pixel.RedBias; + bias[1] = ctx->Pixel.GreenBias; + bias[2] = ctx->Pixel.BlueBias; + bias[3] = ctx->Pixel.AlphaBias; + + scale_p = _mesa_add_named_constant(params, "RGBA_scale", scale, 4); + bias_p = _mesa_add_named_constant(params, "RGBA_bias", bias, 4); + + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_MAD; + inst[ic].DstReg.File = PROGRAM_OUTPUT; + inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].SrcReg[0].File = PROGRAM_OUTPUT; + inst[ic].SrcReg[0].Index = FRAG_RESULT_COLR; + inst[ic].SrcReg[1].File = PROGRAM_CONSTANT; + inst[ic].SrcReg[1].Index = scale_p; + inst[ic].SrcReg[2].File = PROGRAM_CONSTANT; + inst[ic].SrcReg[2].Index = bias_p; + ic++; + } + + /* END; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_END; + ic++; + + assert(ic <= MAX_INST); + + + fp = (struct gl_fragment_program *) + ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + if (!fp) + return NULL; + + fp->Base.Instructions = _mesa_alloc_instructions(ic); + if (!fp->Base.Instructions) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, + "generating pixel transfer program"); + return NULL; + } + + _mesa_copy_instructions(fp->Base.Instructions, inst, + fp->Base.NumInstructions); + + fp->Base.NumInstructions = ic; + + return fp; +} diff --git a/src/mesa/state_tracker/st_pixeltransfer.h b/src/mesa/state_tracker/st_pixeltransfer.h new file mode 100644 index 0000000000..35f0044a30 --- /dev/null +++ b/src/mesa/state_tracker/st_pixeltransfer.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_PIXELTRANSFER_H +#define ST_PIXELTRANSFER_H + +extern struct gl_fragment_program * +st_get_pixel_transfer_program(GLcontext *ctx); + + +#endif /* ST_PIXELTRANSFER_H */ -- cgit v1.2.3 From bcc026893b4719122711d6dd3a391cca10820665 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 16:38:23 -0600 Subject: added st_pixeltransfer.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index e27f02a9bc..77d169c5d6 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -232,6 +232,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_draw.c \ state_tracker/st_format.c \ state_tracker/st_mesa_to_tgsi.c \ + state_tracker/st_pixeltransfer.c \ state_tracker/st_program.c \ state_tracker/st_mipmap_tree.c -- cgit v1.2.3 From 1203f54686896ed21bb4e2e57880e0e034091b14 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 16:38:53 -0600 Subject: make _mesa_combine_programs() params const --- src/mesa/shader/program.c | 3 ++- src/mesa/shader/program.h | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 0e420179bc..2d8b887791 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -460,7 +460,8 @@ adjust_param_indexes(struct prog_instruction *inst, GLuint numInst, */ struct gl_program * _mesa_combine_programs(GLcontext *ctx, - struct gl_program *progA, struct gl_program *progB) + const struct gl_program *progA, + const struct gl_program *progB) { struct prog_instruction *newInst; struct gl_program *newProg; diff --git a/src/mesa/shader/program.h b/src/mesa/shader/program.h index 02faa706d9..4b7297e4c6 100644 --- a/src/mesa/shader/program.h +++ b/src/mesa/shader/program.h @@ -90,7 +90,8 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog); extern struct gl_program * _mesa_combine_programs(GLcontext *ctx, - struct gl_program *progA, struct gl_program *progB); + const struct gl_program *progA, + const struct gl_program *progB); extern GLint _mesa_find_free_register(const struct gl_program *prog, GLuint regFile); -- cgit v1.2.3 From cc9b4df5cbcb4e77cd5ee06e97e450fc0d78006e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 17:14:23 -0600 Subject: update comments, fix typo --- src/mesa/state_tracker/st_context.h | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index bacc0b9b3f..329599bc61 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -68,12 +68,11 @@ struct st_context struct pipe_context *pipe; struct draw_context *draw; /**< For selection/feedback */ - struct draw_stage *feedback_stage; /**< For FL_FEEDBACK rendermode */ + struct draw_stage *feedback_stage; /**< For GL_FEEDBACK rendermode */ struct draw_stage *selection_stage; /**< For GL_SELECT rendermode */ - /* Eventually will use a cache to feed the pipe with - * create/bind/delete calls to constant state objects. Not yet - * though, we just shove random objects across the interface. + /* Some state is contained in constant objects. + * Other state is just parameter values. */ struct { const struct cso_alpha_test *alpha_test; -- cgit v1.2.3 From b19a93393043371776af6d50662a3eb0a9a965ce Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 17:32:42 -0600 Subject: comment unused 'cb' field --- src/mesa/state_tracker/st_context.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 329599bc61..ecab0e2932 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -99,6 +99,7 @@ struct st_context struct st_tracked_state tracked_state[2]; } constants; + /* XXX unused: */ struct { struct gl_fragment_program *fragment_program; } cb; -- cgit v1.2.3 From ba0fcc47d61be6caa2f4a5f4eb0c36eba9e2cb59 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 29 Oct 2007 17:36:39 -0600 Subject: Set _NEW_BUFFERS in glRead/DrawBuffer(). Previously, we set _NEW_PIXEL and _NEW_COLOR in these functions, respectively. That correponds to the GL attribute groups, but doesn't make much sense otherwise. This could improve validation efficiency in a few places too. It looks like all the drivers are already checking for _NEW_BUFFERS in the right places (since that's the bit for FBO state) so we can trim out _NEW_PIXEL and _NEW_COLOR at any time. --- src/mesa/drivers/x11/xm_dd.c | 2 +- src/mesa/main/buffers.c | 4 ++-- src/mesa/main/state.c | 2 +- src/mesa/state_tracker/st_atom_framebuffer.c | 6 +----- 4 files changed, 5 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index 3ee44bfbaf..8ae243ae66 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -841,7 +841,7 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) * GL_DITHER, GL_READ/DRAW_BUFFER, buffer binding state, etc. effect * renderbuffer span/clear funcs. */ - if (new_state & (_NEW_COLOR | _NEW_PIXEL | _NEW_BUFFERS)) { + if (new_state & (_NEW_COLOR | _NEW_BUFFERS)) { XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); struct xmesa_renderbuffer *front_xrb, *back_xrb; diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index 7a918c53a7..3cbd671bab 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -527,7 +527,7 @@ _mesa_drawbuffers(GLcontext *ctx, GLuint n, const GLenum *buffers, set_color_output(ctx, output, GL_NONE, 0x0); } - ctx->NewState |= _NEW_COLOR; + ctx->NewState |= _NEW_BUFFERS; } @@ -588,7 +588,7 @@ _mesa_ReadBuffer(GLenum buffer) if (!_mesa_readbuffer_update_fields(ctx, buffer)) return; - ctx->NewState |= _NEW_PIXEL; + ctx->NewState |= _NEW_BUFFERS; /* * Call device driver function. diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 41c657e9fd..0e4cf047c8 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -1170,7 +1170,7 @@ _mesa_update_state_locked( GLcontext *ctx ) if (new_state & (_NEW_PROGRAM|_NEW_TEXTURE|_NEW_TEXTURE_MATRIX)) _mesa_update_texture( ctx, new_state ); - if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) + if (new_state & _NEW_BUFFERS) _mesa_update_framebuffer(ctx); if (new_state & (_NEW_SCISSOR | _NEW_BUFFERS | _NEW_VIEWPORT)) diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index e776c9112d..aec51f5eed 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -85,14 +85,10 @@ update_framebuffer_state( struct st_context *st ) } -/** - * Note that glDrawBuffer() sets _NEW_COLOR, not _NEW_BUFFER. - */ - const struct st_tracked_state st_update_framebuffer = { .name = "st_update_framebuffer", .dirty = { - .mesa = (_NEW_BUFFERS | _NEW_COLOR), + .mesa = _NEW_BUFFERS, .st = 0, }, .update = update_framebuffer_state -- cgit v1.2.3 From 8234935b658c99b829564f75a2c7840a1301ca36 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 11:14:33 -0600 Subject: check for NULL ptr in _mesa_print_parameter_list() --- src/mesa/shader/prog_print.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index c421b1228b..f410a9099e 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -753,6 +753,9 @@ _mesa_print_parameter_list(const struct gl_program_parameter_list *list) const gl_prog_print_mode mode = PROG_PRINT_DEBUG; GLuint i; + if (!list) + return; + _mesa_printf("param list %p\n", (void *) list); for (i = 0; i < list->NumParameters; i++){ struct gl_program_parameter *param = list->Parameters + i; -- cgit v1.2.3 From 64a97680a714eed17968dbe16589bfbc6c8a62a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 11:15:09 -0600 Subject: bug fixes, implement state atom for pixel transfer --- src/mesa/state_tracker/st_pixeltransfer.c | 52 ++++++++++++++++++++++++------- 1 file changed, 41 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_pixeltransfer.c b/src/mesa/state_tracker/st_pixeltransfer.c index 52e2a4c89d..b1fce81755 100644 --- a/src/mesa/state_tracker/st_pixeltransfer.c +++ b/src/mesa/state_tracker/st_pixeltransfer.c @@ -41,7 +41,7 @@ #include "shader/prog_parameter.h" #include "shader/prog_print.h" -#include "st_pixeltransfer.h" +#include "st_context.h" #define MAX_INST 100 @@ -49,14 +49,19 @@ /** * Returns a fragment program which implements the current pixel transfer ops. */ -struct gl_fragment_program * -st_get_pixel_transfer_program(GLcontext *ctx) +static struct gl_fragment_program * +get_pixel_transfer_program(GLcontext *ctx) { struct prog_instruction inst[MAX_INST]; struct gl_program_parameter_list *params; struct gl_fragment_program *fp; GLuint ic = 0; + fp = (struct gl_fragment_program *) + ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + if (!fp) + return NULL; + params = _mesa_new_parameter_list(); /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ @@ -69,6 +74,8 @@ st_get_pixel_transfer_program(GLcontext *ctx) inst[ic].TexSrcUnit = 0; inst[ic].TexSrcTarget = TEXTURE_2D_INDEX; ic++; + fp->Base.InputsRead = (1 << FRAG_ATTRIB_TEX0); + fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR); /* MAD result.color, result.color, scale, bias; */ if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 || @@ -110,11 +117,6 @@ st_get_pixel_transfer_program(GLcontext *ctx) assert(ic <= MAX_INST); - fp = (struct gl_fragment_program *) - ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); - if (!fp) - return NULL; - fp->Base.Instructions = _mesa_alloc_instructions(ic); if (!fp->Base.Instructions) { _mesa_error(ctx, GL_OUT_OF_MEMORY, @@ -122,10 +124,38 @@ st_get_pixel_transfer_program(GLcontext *ctx) return NULL; } - _mesa_copy_instructions(fp->Base.Instructions, inst, - fp->Base.NumInstructions); - + _mesa_copy_instructions(fp->Base.Instructions, inst, ic); fp->Base.NumInstructions = ic; + fp->Base.Parameters = params; + + printf("========= pixel transfer prog\n"); + _mesa_print_program(&fp->Base); + _mesa_print_parameter_list(fp->Base.Parameters); return fp; } + + + +static void +update_pixel_transfer(struct st_context *st) +{ + /* XXX temporary - implement a program cache */ + GLcontext *ctx = st->ctx; + if (st->pixel_transfer_program) { + ctx->Driver.DeleteProgram(ctx, &st->pixel_transfer_program->Base); + } + + st->pixel_transfer_program = get_pixel_transfer_program(ctx); +} + + + +const struct st_tracked_state st_update_pixel_transfer = { + .name = "st_update_pixel_transfer", + .dirty = { + .mesa = _NEW_PIXEL, + .st = 0, + }, + .update = update_pixel_transfer +}; -- cgit v1.2.3 From 088e80f6dbdaa8bdcac2d9a46e8d5f387c116371 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 11:15:29 -0600 Subject: added st_update_pixel_transfer atom --- src/mesa/state_tracker/st_atom.c | 3 ++- src/mesa/state_tracker/st_atom.h | 1 + 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 326042cb34..d967b9b978 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -61,7 +61,8 @@ static const struct st_tracked_state *atoms[] = &st_update_texture, &st_update_vs_constants, &st_update_fs_constants, - &st_update_alpha_test + &st_update_alpha_test, + &st_update_pixel_transfer }; diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 94cb7bee7a..a7caca5459 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -60,6 +60,7 @@ const struct st_tracked_state st_update_texture; const struct st_tracked_state st_update_fs_constants; const struct st_tracked_state st_update_vs_constants; const struct st_tracked_state st_update_alpha_test; +const struct st_tracked_state st_update_pixel_transfer; #endif -- cgit v1.2.3 From 0be546fa53753119a770d6368ced380bdf69c69b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 11:15:59 -0600 Subject: added pixel_transfer_program field --- src/mesa/state_tracker/st_context.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index ecab0e2932..3774a999ab 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -132,6 +132,8 @@ struct st_context struct st_vertex_program *vp; /**< Currently bound vertex program */ struct st_fragment_program *fp; /**< Currently bound fragment program */ + struct gl_fragment_program *pixel_transfer_program; + /** * Buffer object which stores the ctx->Current.Attrib[] values. * Used for vertex array drawing when we we need an attribute for -- cgit v1.2.3 From b8042fb7fb15a4f99a10f9d9e48d43301622c2bf Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 11:16:26 -0600 Subject: make st_upload_constants() public --- src/mesa/state_tracker/st_atom_constbuf.c | 19 ++++++++++++---- src/mesa/state_tracker/st_atom_constbuf.h | 38 +++++++++++++++++++++++++++++++ 2 files changed, 52 insertions(+), 5 deletions(-) create mode 100644 src/mesa/state_tracker/st_atom_constbuf.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index d140e0ef50..7da7136676 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -40,15 +40,24 @@ #include "st_context.h" #include "st_atom.h" +#include "st_atom_constbuf.h" #include "st_program.h" -static void upload_constants( struct st_context *st, - struct gl_program_parameter_list *params, - unsigned id) + +/** + * Pass the given program parameters to the graphics pipe as a + * constant buffer. + * \param id either PIPE_SHADER_VERTEX or PIPE_SHADER_FRAGMENT + */ +void st_upload_constants( struct st_context *st, + struct gl_program_parameter_list *params, + unsigned id) { struct pipe_winsys *ws = st->pipe->winsys; struct pipe_constant_buffer *cbuf = &st->state.constants[id]; + assert(id == PIPE_SHADER_VERTEX || id == PIPE_SHADER_FRAGMENT); + /* update constants */ if (params && params->NumParameters) { const uint paramBytes = params->NumParameters * sizeof(GLfloat) * 4; @@ -90,7 +99,7 @@ static void update_vs_constants(struct st_context *st ) struct st_vertex_program *vp = st->vp; struct gl_program_parameter_list *params = vp->Base.Base.Parameters; - upload_constants( st, params, PIPE_SHADER_VERTEX ); + st_upload_constants( st, params, PIPE_SHADER_VERTEX ); } const struct st_tracked_state st_update_vs_constants = { @@ -109,7 +118,7 @@ static void update_fs_constants(struct st_context *st ) struct st_fragment_program *fp = st->fp; struct gl_program_parameter_list *params = fp->Base.Base.Parameters; - upload_constants( st, params, PIPE_SHADER_FRAGMENT ); + st_upload_constants( st, params, PIPE_SHADER_FRAGMENT ); } const struct st_tracked_state st_update_fs_constants = { diff --git a/src/mesa/state_tracker/st_atom_constbuf.h b/src/mesa/state_tracker/st_atom_constbuf.h new file mode 100644 index 0000000000..f707534e2c --- /dev/null +++ b/src/mesa/state_tracker/st_atom_constbuf.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_ATOM_CONSTBUF_H +#define ST_ATOM_CONSTBUF_H + + +void st_upload_constants( struct st_context *st, + struct gl_program_parameter_list *params, + unsigned id); + + +#endif /* ST_ATOM_CONSTBUF_H */ -- cgit v1.2.3 From de38d9c498b999d80b51679ece6b4caccd80a2b8 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 11:16:55 -0600 Subject: checkpoint: combine pixel transfer and user shader for glDrawPixels --- src/mesa/state_tracker/st_cb_drawpixels.c | 34 ++++++++++++++++++++++++++++++- 1 file changed, 33 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 3c57317415..ed0ef71192 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -33,9 +33,13 @@ #include "main/imports.h" #include "main/image.h" #include "main/macros.h" +#include "shader/program.h" +#include "shader/prog_parameter.h" +#include "shader/prog_print.h" #include "st_context.h" #include "st_atom.h" +#include "st_atom_constbuf.h" #include "st_cache.h" #include "st_draw.h" #include "st_program.h" @@ -173,6 +177,34 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) } +static struct st_fragment_program * +make_drawpix_fragment_shader(struct st_context *st) +{ + GLcontext *ctx = st->ctx; + struct st_fragment_program *stfp; + struct gl_program *p; + + printf("====== drawpix combine progs\n"); + p = _mesa_combine_programs(ctx, + &st->pixel_transfer_program->Base, + &ctx->FragmentProgram._Current->Base); + + _mesa_print_program(p); + printf("InputsRead: 0x%x\n", p->InputsRead); + printf("OutputsWritten: 0x%x\n", p->OutputsWritten); + _mesa_print_parameter_list(p->Parameters); + + stfp = (struct st_fragment_program *) p; + st_translate_fragment_program(st, stfp, NULL, + stfp->tokens, ST_MAX_SHADER_TOKENS); + + + st_upload_constants( st, p->Parameters, PIPE_SHADER_FRAGMENT ); + + return stfp; +} + + /** * Create fragment shader that does a TEX() instruction to get a Z * value, then writes to FRAG_RESULT_DEPR. @@ -899,7 +931,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } else { ps = st->state.framebuffer.cbufs[0]; - stfp = make_fragment_shader(ctx->st, GL_FALSE); + stfp = make_drawpix_fragment_shader(ctx->st); stvp = make_vertex_shader(ctx->st, GL_FALSE); color = NULL; } -- cgit v1.2.3 From 7ce99a11037e577e36480f3e29b2685b853f0330 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 11:18:31 -0600 Subject: Rename file since it's now a state atom --- src/mesa/sources | 2 +- src/mesa/state_tracker/st_atom_pixeltransfer.c | 161 +++++++++++++++++++++++++ src/mesa/state_tracker/st_atom_pixeltransfer.h | 36 ++++++ 3 files changed, 198 insertions(+), 1 deletion(-) create mode 100644 src/mesa/state_tracker/st_atom_pixeltransfer.c create mode 100644 src/mesa/state_tracker/st_atom_pixeltransfer.h (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 77d169c5d6..a4af293da0 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -206,6 +206,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_depth.c \ state_tracker/st_atom_fixedfunction.c \ state_tracker/st_atom_framebuffer.c \ + state_tracker/st_atom_pixeltransfer.c \ state_tracker/st_atom_sampler.c \ state_tracker/st_atom_scissor.c \ state_tracker/st_atom_shader.c \ @@ -232,7 +233,6 @@ STATETRACKER_SOURCES = \ state_tracker/st_draw.c \ state_tracker/st_format.c \ state_tracker/st_mesa_to_tgsi.c \ - state_tracker/st_pixeltransfer.c \ state_tracker/st_program.c \ state_tracker/st_mipmap_tree.c diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c new file mode 100644 index 0000000000..b1fce81755 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -0,0 +1,161 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * Generate fragment programs to implement pixel transfer ops, such as + * scale/bias, colormatrix, colortable, convolution... + * + * Authors: + * Brian Paul + */ + +#include "main/imports.h" +#include "main/image.h" +#include "main/macros.h" +#include "shader/program.h" +#include "shader/prog_instruction.h" +#include "shader/prog_parameter.h" +#include "shader/prog_print.h" + +#include "st_context.h" + + +#define MAX_INST 100 + +/** + * Returns a fragment program which implements the current pixel transfer ops. + */ +static struct gl_fragment_program * +get_pixel_transfer_program(GLcontext *ctx) +{ + struct prog_instruction inst[MAX_INST]; + struct gl_program_parameter_list *params; + struct gl_fragment_program *fp; + GLuint ic = 0; + + fp = (struct gl_fragment_program *) + ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + if (!fp) + return NULL; + + params = _mesa_new_parameter_list(); + + /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_TEX; + inst[ic].DstReg.File = PROGRAM_OUTPUT; + inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].SrcReg[0].File = PROGRAM_INPUT; + inst[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + inst[ic].TexSrcUnit = 0; + inst[ic].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; + fp->Base.InputsRead = (1 << FRAG_ATTRIB_TEX0); + fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR); + + /* MAD result.color, result.color, scale, bias; */ + if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 || + ctx->Pixel.GreenBias != 0.0 || ctx->Pixel.GreenScale != 1.0 || + ctx->Pixel.BlueBias != 0.0 || ctx->Pixel.BlueScale != 1.0 || + ctx->Pixel.AlphaBias != 0.0 || ctx->Pixel.AlphaScale != 1.0) { + GLfloat scale[4], bias[4]; + GLint scale_p, bias_p; + scale[0] = ctx->Pixel.RedScale; + scale[1] = ctx->Pixel.GreenScale; + scale[2] = ctx->Pixel.BlueScale; + scale[3] = ctx->Pixel.AlphaScale; + bias[0] = ctx->Pixel.RedBias; + bias[1] = ctx->Pixel.GreenBias; + bias[2] = ctx->Pixel.BlueBias; + bias[3] = ctx->Pixel.AlphaBias; + + scale_p = _mesa_add_named_constant(params, "RGBA_scale", scale, 4); + bias_p = _mesa_add_named_constant(params, "RGBA_bias", bias, 4); + + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_MAD; + inst[ic].DstReg.File = PROGRAM_OUTPUT; + inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].SrcReg[0].File = PROGRAM_OUTPUT; + inst[ic].SrcReg[0].Index = FRAG_RESULT_COLR; + inst[ic].SrcReg[1].File = PROGRAM_CONSTANT; + inst[ic].SrcReg[1].Index = scale_p; + inst[ic].SrcReg[2].File = PROGRAM_CONSTANT; + inst[ic].SrcReg[2].Index = bias_p; + ic++; + } + + /* END; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_END; + ic++; + + assert(ic <= MAX_INST); + + + fp->Base.Instructions = _mesa_alloc_instructions(ic); + if (!fp->Base.Instructions) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, + "generating pixel transfer program"); + return NULL; + } + + _mesa_copy_instructions(fp->Base.Instructions, inst, ic); + fp->Base.NumInstructions = ic; + fp->Base.Parameters = params; + + printf("========= pixel transfer prog\n"); + _mesa_print_program(&fp->Base); + _mesa_print_parameter_list(fp->Base.Parameters); + + return fp; +} + + + +static void +update_pixel_transfer(struct st_context *st) +{ + /* XXX temporary - implement a program cache */ + GLcontext *ctx = st->ctx; + if (st->pixel_transfer_program) { + ctx->Driver.DeleteProgram(ctx, &st->pixel_transfer_program->Base); + } + + st->pixel_transfer_program = get_pixel_transfer_program(ctx); +} + + + +const struct st_tracked_state st_update_pixel_transfer = { + .name = "st_update_pixel_transfer", + .dirty = { + .mesa = _NEW_PIXEL, + .st = 0, + }, + .update = update_pixel_transfer +}; diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.h b/src/mesa/state_tracker/st_atom_pixeltransfer.h new file mode 100644 index 0000000000..35f0044a30 --- /dev/null +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_PIXELTRANSFER_H +#define ST_PIXELTRANSFER_H + +extern struct gl_fragment_program * +st_get_pixel_transfer_program(GLcontext *ctx); + + +#endif /* ST_PIXELTRANSFER_H */ -- cgit v1.2.3 From cbdff56f8cbb0560ca5eeb6e4b96eb3996e67fd1 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 11:19:31 -0600 Subject: Remove obsolete files --- src/mesa/state_tracker/st_atom_pixeltransfer.h | 36 ------ src/mesa/state_tracker/st_pixeltransfer.c | 161 ------------------------- src/mesa/state_tracker/st_pixeltransfer.h | 36 ------ 3 files changed, 233 deletions(-) delete mode 100644 src/mesa/state_tracker/st_atom_pixeltransfer.h delete mode 100644 src/mesa/state_tracker/st_pixeltransfer.c delete mode 100644 src/mesa/state_tracker/st_pixeltransfer.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.h b/src/mesa/state_tracker/st_atom_pixeltransfer.h deleted file mode 100644 index 35f0044a30..0000000000 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.h +++ /dev/null @@ -1,36 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef ST_PIXELTRANSFER_H -#define ST_PIXELTRANSFER_H - -extern struct gl_fragment_program * -st_get_pixel_transfer_program(GLcontext *ctx); - - -#endif /* ST_PIXELTRANSFER_H */ diff --git a/src/mesa/state_tracker/st_pixeltransfer.c b/src/mesa/state_tracker/st_pixeltransfer.c deleted file mode 100644 index b1fce81755..0000000000 --- a/src/mesa/state_tracker/st_pixeltransfer.c +++ /dev/null @@ -1,161 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* - * Generate fragment programs to implement pixel transfer ops, such as - * scale/bias, colormatrix, colortable, convolution... - * - * Authors: - * Brian Paul - */ - -#include "main/imports.h" -#include "main/image.h" -#include "main/macros.h" -#include "shader/program.h" -#include "shader/prog_instruction.h" -#include "shader/prog_parameter.h" -#include "shader/prog_print.h" - -#include "st_context.h" - - -#define MAX_INST 100 - -/** - * Returns a fragment program which implements the current pixel transfer ops. - */ -static struct gl_fragment_program * -get_pixel_transfer_program(GLcontext *ctx) -{ - struct prog_instruction inst[MAX_INST]; - struct gl_program_parameter_list *params; - struct gl_fragment_program *fp; - GLuint ic = 0; - - fp = (struct gl_fragment_program *) - ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); - if (!fp) - return NULL; - - params = _mesa_new_parameter_list(); - - /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ - _mesa_init_instructions(inst + ic, 1); - inst[ic].Opcode = OPCODE_TEX; - inst[ic].DstReg.File = PROGRAM_OUTPUT; - inst[ic].DstReg.Index = FRAG_RESULT_COLR; - inst[ic].SrcReg[0].File = PROGRAM_INPUT; - inst[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; - inst[ic].TexSrcUnit = 0; - inst[ic].TexSrcTarget = TEXTURE_2D_INDEX; - ic++; - fp->Base.InputsRead = (1 << FRAG_ATTRIB_TEX0); - fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR); - - /* MAD result.color, result.color, scale, bias; */ - if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 || - ctx->Pixel.GreenBias != 0.0 || ctx->Pixel.GreenScale != 1.0 || - ctx->Pixel.BlueBias != 0.0 || ctx->Pixel.BlueScale != 1.0 || - ctx->Pixel.AlphaBias != 0.0 || ctx->Pixel.AlphaScale != 1.0) { - GLfloat scale[4], bias[4]; - GLint scale_p, bias_p; - scale[0] = ctx->Pixel.RedScale; - scale[1] = ctx->Pixel.GreenScale; - scale[2] = ctx->Pixel.BlueScale; - scale[3] = ctx->Pixel.AlphaScale; - bias[0] = ctx->Pixel.RedBias; - bias[1] = ctx->Pixel.GreenBias; - bias[2] = ctx->Pixel.BlueBias; - bias[3] = ctx->Pixel.AlphaBias; - - scale_p = _mesa_add_named_constant(params, "RGBA_scale", scale, 4); - bias_p = _mesa_add_named_constant(params, "RGBA_bias", bias, 4); - - _mesa_init_instructions(inst + ic, 1); - inst[ic].Opcode = OPCODE_MAD; - inst[ic].DstReg.File = PROGRAM_OUTPUT; - inst[ic].DstReg.Index = FRAG_RESULT_COLR; - inst[ic].SrcReg[0].File = PROGRAM_OUTPUT; - inst[ic].SrcReg[0].Index = FRAG_RESULT_COLR; - inst[ic].SrcReg[1].File = PROGRAM_CONSTANT; - inst[ic].SrcReg[1].Index = scale_p; - inst[ic].SrcReg[2].File = PROGRAM_CONSTANT; - inst[ic].SrcReg[2].Index = bias_p; - ic++; - } - - /* END; */ - _mesa_init_instructions(inst + ic, 1); - inst[ic].Opcode = OPCODE_END; - ic++; - - assert(ic <= MAX_INST); - - - fp->Base.Instructions = _mesa_alloc_instructions(ic); - if (!fp->Base.Instructions) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, - "generating pixel transfer program"); - return NULL; - } - - _mesa_copy_instructions(fp->Base.Instructions, inst, ic); - fp->Base.NumInstructions = ic; - fp->Base.Parameters = params; - - printf("========= pixel transfer prog\n"); - _mesa_print_program(&fp->Base); - _mesa_print_parameter_list(fp->Base.Parameters); - - return fp; -} - - - -static void -update_pixel_transfer(struct st_context *st) -{ - /* XXX temporary - implement a program cache */ - GLcontext *ctx = st->ctx; - if (st->pixel_transfer_program) { - ctx->Driver.DeleteProgram(ctx, &st->pixel_transfer_program->Base); - } - - st->pixel_transfer_program = get_pixel_transfer_program(ctx); -} - - - -const struct st_tracked_state st_update_pixel_transfer = { - .name = "st_update_pixel_transfer", - .dirty = { - .mesa = _NEW_PIXEL, - .st = 0, - }, - .update = update_pixel_transfer -}; diff --git a/src/mesa/state_tracker/st_pixeltransfer.h b/src/mesa/state_tracker/st_pixeltransfer.h deleted file mode 100644 index 35f0044a30..0000000000 --- a/src/mesa/state_tracker/st_pixeltransfer.h +++ /dev/null @@ -1,36 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef ST_PIXELTRANSFER_H -#define ST_PIXELTRANSFER_H - -extern struct gl_fragment_program * -st_get_pixel_transfer_program(GLcontext *ctx); - - -#endif /* ST_PIXELTRANSFER_H */ -- cgit v1.2.3 From ab3f6015aa7227da3137b60456deb3905680f95f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 11:32:16 -0600 Subject: Disable debug code. Basic code for PixelTranslfer ops and glDrawPixels works now. A pixel transfer program is generated and combined with the current fragment shader. --- src/mesa/state_tracker/st_atom_pixeltransfer.c | 2 ++ src/mesa/state_tracker/st_cb_drawpixels.c | 10 ++++++++-- 2 files changed, 10 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index b1fce81755..d3abf87d7d 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -128,9 +128,11 @@ get_pixel_transfer_program(GLcontext *ctx) fp->Base.NumInstructions = ic; fp->Base.Parameters = params; +#if 0 printf("========= pixel transfer prog\n"); _mesa_print_program(&fp->Base); _mesa_print_parameter_list(fp->Base.Parameters); +#endif return fp; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index ed0ef71192..ff7aa7d74b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -184,15 +184,21 @@ make_drawpix_fragment_shader(struct st_context *st) struct st_fragment_program *stfp; struct gl_program *p; - printf("====== drawpix combine progs\n"); + /* + * XXX Use st_program's serial numbers to determine when the + * user-provided program and pixel-transfer program to avoid + * needless combining/translation here. + */ + p = _mesa_combine_programs(ctx, &st->pixel_transfer_program->Base, &ctx->FragmentProgram._Current->Base); - +#if 0 _mesa_print_program(p); printf("InputsRead: 0x%x\n", p->InputsRead); printf("OutputsWritten: 0x%x\n", p->OutputsWritten); _mesa_print_parameter_list(p->Parameters); +#endif stfp = (struct st_fragment_program *) p; st_translate_fragment_program(st, stfp, NULL, -- cgit v1.2.3 From 2f496ff9574ad99f8d85b5d63e97b15faf5aee39 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 30 Oct 2007 18:36:05 +0100 Subject: tgsi_translate_mesa_program: Don't crash when program->Parameters == NULL. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 36b41ea7e8..ab7aa504a1 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -792,7 +792,8 @@ tgsi_translate_mesa_program( } /* immediates/literals */ - for (i = 0; i < program->Parameters->NumParameters; i++) { + for (i = 0; program->Parameters && i < program->Parameters->NumParameters; + i++) { if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { struct tgsi_full_immediate fullimm = make_immediate(program->Parameters->ParameterValues[i], -- cgit v1.2.3 From 0ee15e050bddad45ef6424e67c2ce75095d90745 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 12:23:22 -0600 Subject: Program cache functions. Generic version of cache code from texenvprogram.c and t_vp_build.c Not used by those files just yet, will also be used for pixel_transfer programs. --- src/mesa/shader/prog_cache.c | 191 +++++++++++++++++++++++++++++++++++++++++++ src/mesa/shader/prog_cache.h | 55 +++++++++++++ 2 files changed, 246 insertions(+) create mode 100644 src/mesa/shader/prog_cache.c create mode 100644 src/mesa/shader/prog_cache.h (limited to 'src') diff --git a/src/mesa/shader/prog_cache.c b/src/mesa/shader/prog_cache.c new file mode 100644 index 0000000000..d97e27dd6b --- /dev/null +++ b/src/mesa/shader/prog_cache.c @@ -0,0 +1,191 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "main/glheader.h" +#include "main/mtypes.h" +#include "main/imports.h" +#include "shader/prog_cache.h" + + +struct cache_item +{ + GLuint hash; + void *key; + struct gl_program *program; + struct cache_item *next; +}; + +struct gl_program_cache +{ + struct cache_item **items; + GLuint size, n_items; +}; + + + +/** + * Compute hash index from state key. + */ +static GLuint +hash_key(const void *key, GLuint key_size) +{ + const GLuint *ikey = (const GLuint *) key; + GLuint hash = 0, i; + + assert(key_size >= 4); + + /* Make a slightly better attempt at a hash function: + */ + for (i = 0; i < key_size / sizeof(*ikey); i++) + { + hash += ikey[i]; + hash += (hash << 10); + hash ^= (hash >> 6); + } + + return hash; +} + + +/** + * Rebuild/expand the hash table to accomodate more entries + */ +static void +rehash(struct gl_program_cache *cache) +{ + struct cache_item **items; + struct cache_item *c, *next; + GLuint size, i; + + size = cache->size * 3; + items = (struct cache_item**) _mesa_malloc(size * sizeof(*items)); + _mesa_memset(items, 0, size * sizeof(*items)); + + for (i = 0; i < cache->size; i++) + for (c = cache->items[i]; c; c = next) { + next = c->next; + c->next = items[c->hash % size]; + items[c->hash % size] = c; + } + + _mesa_free(cache->items); + cache->items = items; + cache->size = size; +} + + +static void +clear_cache(GLcontext *ctx, struct gl_program_cache *cache) +{ + struct cache_item *c, *next; + GLuint i; + + for (i = 0; i < cache->size; i++) { + for (c = cache->items[i]; c; c = next) { + next = c->next; + _mesa_free(c->key); + ctx->Driver.DeleteProgram(ctx, c->program); + _mesa_free(c); + } + cache->items[i] = NULL; + } + + + cache->n_items = 0; +} + + + +struct gl_program_cache * +_mesa_new_program_cache(void) +{ + struct gl_program_cache *cache = CALLOC_STRUCT(gl_program_cache); + if (cache) { + cache->size = 17; + cache->items = (struct cache_item **) + _mesa_calloc(cache->size * sizeof(struct cache_item)); + if (!cache->items) { + _mesa_free(cache); + return NULL; + } + } + return cache; +} + + +void +_mesa_delete_program_cache(GLcontext *ctx, struct gl_program_cache *cache) +{ + clear_cache(ctx, cache); + _mesa_free(cache->items); +} + + +struct gl_program * +_mesa_search_program_cache(const struct gl_program_cache *cache, + const void *key, GLuint keysize) +{ + const GLuint hash = hash_key(key, keysize); + struct cache_item *c; + + for (c = cache->items[hash % cache->size]; c; c = c->next) { + if (c->hash == hash && memcmp(c->key, key, keysize) == 0) + return c->program; + } + + return NULL; +} + + +void +_mesa_program_cache_insert(GLcontext *ctx, + struct gl_program_cache *cache, + const void *key, GLuint keysize, + struct gl_program *program) +{ + const GLuint hash = hash_key(key, keysize); + struct cache_item *c = CALLOC_STRUCT(cache_item); + + c->hash = hash; + + c->key = _mesa_malloc(keysize); + memcpy(c->key, key, keysize); + + c->program = program; + + if (cache->n_items > cache->size * 1.5) { + if (cache->size < 1000) + rehash(cache); + else + clear_cache(ctx, cache); + } + + cache->n_items++; + c->next = cache->items[hash % cache->size]; + cache->items[hash % cache->size] = c; +} diff --git a/src/mesa/shader/prog_cache.h b/src/mesa/shader/prog_cache.h new file mode 100644 index 0000000000..a8c91fba01 --- /dev/null +++ b/src/mesa/shader/prog_cache.h @@ -0,0 +1,55 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef PROG_CACHE_H +#define PROG_CACHE_H + + +/** Opaque type */ +struct gl_program_cache; + + +extern struct gl_program_cache * +_mesa_new_program_cache(void); + +extern void +_mesa_delete_program_cache(GLcontext *ctx, struct gl_program_cache *pc); + + +extern struct gl_program * +_mesa_search_program_cache(const struct gl_program_cache *cache, + const void *key, GLuint keysize); + +extern void +_mesa_program_cache_insert(GLcontext *ctx, + struct gl_program_cache *cache, + const void *key, GLuint keysize, + struct gl_program *program); + + +#endif /* PROG_CACHE_H */ -- cgit v1.2.3 From 3e317996400778c3197bdaa7d95f8712205e589b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 12:23:37 -0600 Subject: added prog_cache.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index a4af293da0..727872378e 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -245,6 +245,7 @@ SHADER_SOURCES = \ shader/nvprogram.c \ shader/nvvertparse.c \ shader/program.c \ + shader/prog_cache.c \ shader/prog_debug.c \ shader/prog_execute.c \ shader/prog_instruction.c \ -- cgit v1.2.3 From df174bdb818d4e21652d49c73ff17722e34a38b9 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 12:24:05 -0600 Subject: added pixel_transfer_cache --- src/mesa/state_tracker/st_context.c | 4 ++++ src/mesa/state_tracker/st_context.h | 4 +++- 2 files changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 8ced3f504c..e872e8b12c 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -92,6 +92,8 @@ struct st_context *st_create_context( GLcontext *ctx, st->haveFramebufferRegions = GL_TRUE; + st->pixel_transfer_cache = _mesa_new_program_cache(); + #if 0 st_init_cb_clear( st ); st_init_cb_program( st ); @@ -124,6 +126,8 @@ void st_destroy_context( struct st_context *st ) #endif cso_cache_delete( st->cache ); + _mesa_delete_program_cache(st->ctx, st->pixel_transfer_cache); + st->pipe->destroy( st->pipe ); FREE( st ); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 3774a999ab..4de70a3b2e 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -28,7 +28,8 @@ #ifndef ST_CONTEXT_H #define ST_CONTEXT_H -#include "mtypes.h" +#include "main/mtypes.h" +#include "shader/prog_cache.h" #include "pipe/p_state.h" @@ -133,6 +134,7 @@ struct st_context struct st_fragment_program *fp; /**< Currently bound fragment program */ struct gl_fragment_program *pixel_transfer_program; + struct gl_program_cache *pixel_transfer_cache; /** * Buffer object which stores the ctx->Current.Attrib[] values. -- cgit v1.2.3 From 8aa42546add1fef4949e2d4ceded62e2d1dd0215 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 12:24:27 -0600 Subject: start using program cache --- src/mesa/state_tracker/st_atom_pixeltransfer.c | 75 ++++++++++++++++++++++++-- 1 file changed, 70 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index d3abf87d7d..f01196bee4 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -44,6 +44,60 @@ #include "st_context.h" + +struct state_key +{ + GLuint scaleAndBias:1; + GLuint colorMatrix:1; + +#if 0 + GLfloat Maps[3][256][4]; + int NumMaps; + GLint NumStages; + pipeline_stage Stages[STAGE_MAX]; + GLboolean StagesUsed[STAGE_MAX]; + GLfloat Scale1[4], Bias1[4]; + GLfloat Scale2[4], Bias2[4]; +#endif +}; + + +static GLboolean +is_identity(const GLfloat m[16]) +{ + GLuint i; + for (i = 0; i < 16; i++) { + const int row = i % 4, col = i / 4; + const float val = (row == col); + if (m[i] != val) + return GL_FALSE; + } + return GL_TRUE; +} + + +static void +make_state_key(GLcontext *ctx, struct state_key *key) +{ + /*GLuint i, j;*/ + + memset(key, 0, sizeof(*key)); + + if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 || + ctx->Pixel.GreenBias != 0.0 || ctx->Pixel.GreenScale != 1.0 || + ctx->Pixel.BlueBias != 0.0 || ctx->Pixel.BlueScale != 1.0 || + ctx->Pixel.AlphaBias != 0.0 || ctx->Pixel.AlphaScale != 1.0) { + key->scaleAndBias = 1; + } + + if (!is_identity(ctx->ColorMatrixStack.Top->m)) { + key->colorMatrix = 1; + } +} + + + + #define MAX_INST 100 /** @@ -142,13 +196,24 @@ get_pixel_transfer_program(GLcontext *ctx) static void update_pixel_transfer(struct st_context *st) { - /* XXX temporary - implement a program cache */ - GLcontext *ctx = st->ctx; - if (st->pixel_transfer_program) { - ctx->Driver.DeleteProgram(ctx, &st->pixel_transfer_program->Base); + struct state_key key; + struct gl_fragment_program *fp; + + make_state_key(st->ctx, &key); + + fp = (struct gl_fragment_program *) + _mesa_search_program_cache(st->pixel_transfer_cache, &key, sizeof(key)); + if (!fp) { + printf("Cached program not found\n"); + fp = get_pixel_transfer_program(st->ctx); + _mesa_program_cache_insert(st->ctx, st->pixel_transfer_cache, + &key, sizeof(key), &fp->Base); + } + else { + printf("Use cached program\n"); } - st->pixel_transfer_program = get_pixel_transfer_program(ctx); + st->pixel_transfer_program = fp; } -- cgit v1.2.3 From 068c7bd912283e051a55b2fd5c4568685d10f3ed Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 12:39:58 -0600 Subject: add state vars for color matrix, pixel transfer scale&bias --- src/mesa/shader/prog_statevars.c | 16 ++++++++++++++++ src/mesa/shader/prog_statevars.h | 3 +++ 2 files changed, 19 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 1bb7142c60..68eaf0be94 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -278,6 +278,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], case STATE_MVP_MATRIX: case STATE_TEXTURE_MATRIX: case STATE_PROGRAM_MATRIX: + case STATE_COLOR_MATRIX: { /* state[0] = modelview, projection, texture, etc. */ /* state[1] = which texture matrix or program matrix */ @@ -311,6 +312,9 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], else if (mat == STATE_PROGRAM_MATRIX) { matrix = ctx->ProgramMatrixStack[index].Top; } + else if (mat == STATE_COLOR_MATRIX) { + matrix = ctx->ColorMatrixStack.Top; + } else { _mesa_problem(ctx, "Bad matrix name in _mesa_fetch_state()"); return; @@ -434,6 +438,18 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], value[3] = ctx->Light.Light[ln]._CosCutoff; return; } + case STATE_PT_SCALE: + value[0] = ctx->Pixel.RedScale; + value[1] = ctx->Pixel.GreenScale; + value[2] = ctx->Pixel.BlueScale; + value[3] = ctx->Pixel.AlphaScale; + break; + case STATE_PT_BIAS: + value[0] = ctx->Pixel.RedBias; + value[1] = ctx->Pixel.GreenBias; + value[2] = ctx->Pixel.BlueBias; + value[3] = ctx->Pixel.AlphaBias; + break; default: /* unknown state indexes are silently ignored * should be handled by the driver. diff --git a/src/mesa/shader/prog_statevars.h b/src/mesa/shader/prog_statevars.h index 3281a4a2a0..9fedcfab8a 100644 --- a/src/mesa/shader/prog_statevars.h +++ b/src/mesa/shader/prog_statevars.h @@ -67,6 +67,7 @@ typedef enum gl_state_index_ { STATE_MVP_MATRIX, STATE_TEXTURE_MATRIX, STATE_PROGRAM_MATRIX, + STATE_COLOR_MATRIX, STATE_MATRIX_INVERSE, STATE_MATRIX_TRANSPOSE, STATE_MATRIX_INVTRANS, @@ -108,6 +109,8 @@ typedef enum gl_state_index_ { STATE_POSITION_NORMALIZED, /* normalized light position */ STATE_FOG_PARAMS_OPTIMIZED, /* for faster fog calc */ STATE_SPOT_DIR_NORMALIZED, /* pre-normalized spot dir */ + STATE_PT_SCALE, /**< Pixel transfer RGBA scale */ + STATE_PT_BIAS, /**< Pixel transfer RGBA bias */ STATE_INTERNAL_DRIVER /* first available state index for drivers (must be last) */ } gl_state_index; -- cgit v1.2.3 From 0abd1fca91b5807a4906ec4b520e68594db5b36c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 13:53:38 -0600 Subject: more additions for COLOR_MATRIX, etc --- src/mesa/shader/prog_statevars.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 68eaf0be94..6cc490bc77 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -508,6 +508,8 @@ _mesa_program_state_flags(const gl_state_index state[STATE_LENGTH]) return _NEW_TEXTURE_MATRIX; case STATE_PROGRAM_MATRIX: return _NEW_TRACK_MATRIX; + case STATE_COLOR_MATRIX: + return _NEW_COLOR_MATRIX; case STATE_DEPTH_RANGE: return _NEW_VIEWPORT; @@ -601,6 +603,9 @@ append_token(char *dst, gl_state_index k) case STATE_PROGRAM_MATRIX: append(dst, "matrix.program"); break; + case STATE_COLOR_MATRIX: + append(dst, "matrix.color"); + break; case STATE_MATRIX_INVERSE: append(dst, ".inverse"); break; @@ -686,6 +691,12 @@ append_token(char *dst, gl_state_index k) case STATE_POSITION_NORMALIZED: append(dst, "(internal)"); break; + case STATE_PT_SCALE: + append(dst, "PTscale"); + break; + case STATE_PT_BIAS: + append(dst, "PTbias"); + break; default: ; } @@ -764,6 +775,7 @@ _mesa_program_state_string(const gl_state_index state[STATE_LENGTH]) case STATE_MVP_MATRIX: case STATE_TEXTURE_MATRIX: case STATE_PROGRAM_MATRIX: + case STATE_COLOR_MATRIX: { /* state[0] = modelview, projection, texture, etc. */ /* state[1] = which texture matrix or program matrix */ -- cgit v1.2.3 From 4836217850114e0972900a68fd7d93e2e241819b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 13:54:24 -0600 Subject: color matrix --- src/mesa/state_tracker/st_atom_pixeltransfer.c | 101 +++++++++++++++++++++---- 1 file changed, 86 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index f01196bee4..b190c39343 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -104,7 +104,7 @@ make_state_key(GLcontext *ctx, struct state_key *key) * Returns a fragment program which implements the current pixel transfer ops. */ static struct gl_fragment_program * -get_pixel_transfer_program(GLcontext *ctx) +get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) { struct prog_instruction inst[MAX_INST]; struct gl_program_parameter_list *params; @@ -132,12 +132,14 @@ get_pixel_transfer_program(GLcontext *ctx) fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR); /* MAD result.color, result.color, scale, bias; */ - if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 || - ctx->Pixel.GreenBias != 0.0 || ctx->Pixel.GreenScale != 1.0 || - ctx->Pixel.BlueBias != 0.0 || ctx->Pixel.BlueScale != 1.0 || - ctx->Pixel.AlphaBias != 0.0 || ctx->Pixel.AlphaScale != 1.0) { + if (key->scaleAndBias) { + static const gl_state_index scale_state[STATE_LENGTH] = + { STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 }; + static const gl_state_index bias_state[STATE_LENGTH] = + { STATE_INTERNAL, STATE_PT_BIAS, 0, 0, 0 }; GLfloat scale[4], bias[4]; GLint scale_p, bias_p; + scale[0] = ctx->Pixel.RedScale; scale[1] = ctx->Pixel.GreenScale; scale[2] = ctx->Pixel.BlueScale; @@ -147,8 +149,8 @@ get_pixel_transfer_program(GLcontext *ctx) bias[2] = ctx->Pixel.BlueBias; bias[3] = ctx->Pixel.AlphaBias; - scale_p = _mesa_add_named_constant(params, "RGBA_scale", scale, 4); - bias_p = _mesa_add_named_constant(params, "RGBA_bias", bias, 4); + scale_p = _mesa_add_state_reference(params, scale_state); + bias_p = _mesa_add_state_reference(params, bias_state); _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_MAD; @@ -156,13 +158,86 @@ get_pixel_transfer_program(GLcontext *ctx) inst[ic].DstReg.Index = FRAG_RESULT_COLR; inst[ic].SrcReg[0].File = PROGRAM_OUTPUT; inst[ic].SrcReg[0].Index = FRAG_RESULT_COLR; - inst[ic].SrcReg[1].File = PROGRAM_CONSTANT; + inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = scale_p; - inst[ic].SrcReg[2].File = PROGRAM_CONSTANT; + inst[ic].SrcReg[2].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[2].Index = bias_p; ic++; } + if (key->colorMatrix) { + static const gl_state_index row0_state[STATE_LENGTH] = + { STATE_COLOR_MATRIX, 0, 0, 0, 0 }; + static const gl_state_index row1_state[STATE_LENGTH] = + { STATE_COLOR_MATRIX, 0, 1, 1, 0 }; + static const gl_state_index row2_state[STATE_LENGTH] = + { STATE_COLOR_MATRIX, 0, 2, 2, 0 }; + static const gl_state_index row3_state[STATE_LENGTH] = + { STATE_COLOR_MATRIX, 0, 3, 3, 0 }; + + GLint row0_p = _mesa_add_state_reference(params, row0_state); + GLint row1_p = _mesa_add_state_reference(params, row1_state); + GLint row2_p = _mesa_add_state_reference(params, row2_state); + GLint row3_p = _mesa_add_state_reference(params, row3_state); + + /* MOV temp0, result.color; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_MOV; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = 0; + inst[ic].SrcReg[0].File = PROGRAM_OUTPUT; + inst[ic].SrcReg[0].Index = FRAG_RESULT_COLR; + ic++; + + /* DP4 result.color.x, tmp0, matrow0; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_DP4; + inst[ic].DstReg.File = PROGRAM_OUTPUT; + inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].DstReg.WriteMask = WRITEMASK_X; + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = 0; + inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; + inst[ic].SrcReg[1].Index = row0_p; + ic++; + + /* DP4 result.color.y, tmp0, matrow1; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_DP4; + inst[ic].DstReg.File = PROGRAM_OUTPUT; + inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].DstReg.WriteMask = WRITEMASK_Y; + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = 0; + inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; + inst[ic].SrcReg[1].Index = row1_p; + ic++; + + /* DP4 result.color.z, tmp0, matrow2; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_DP4; + inst[ic].DstReg.File = PROGRAM_OUTPUT; + inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].DstReg.WriteMask = WRITEMASK_Z; + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = 0; + inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; + inst[ic].SrcReg[1].Index = row2_p; + ic++; + + /* DP4 result.color.w, tmp0, matrow3; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_DP4; + inst[ic].DstReg.File = PROGRAM_OUTPUT; + inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].DstReg.WriteMask = WRITEMASK_W; + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = 0; + inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; + inst[ic].SrcReg[1].Index = row3_p; + ic++; + } + /* END; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_END; @@ -204,14 +279,10 @@ update_pixel_transfer(struct st_context *st) fp = (struct gl_fragment_program *) _mesa_search_program_cache(st->pixel_transfer_cache, &key, sizeof(key)); if (!fp) { - printf("Cached program not found\n"); - fp = get_pixel_transfer_program(st->ctx); + fp = get_pixel_transfer_program(st->ctx, &key); _mesa_program_cache_insert(st->ctx, st->pixel_transfer_cache, &key, sizeof(key), &fp->Base); } - else { - printf("Use cached program\n"); - } st->pixel_transfer_program = fp; } @@ -221,7 +292,7 @@ update_pixel_transfer(struct st_context *st) const struct st_tracked_state st_update_pixel_transfer = { .name = "st_update_pixel_transfer", .dirty = { - .mesa = _NEW_PIXEL, + .mesa = _NEW_PIXEL | _NEW_COLOR_MATRIX, .st = 0, }, .update = update_pixel_transfer -- cgit v1.2.3 From ee70c02b62f7692de6293045928c47168652bd1c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 13:55:56 -0600 Subject: Before calling texformat->StoreImage() set _ImageTransferState=0 since we'll do pixel transfer in the fragment program. --- src/mesa/state_tracker/st_cb_drawpixels.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index ff7aa7d74b..3359338385 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -424,6 +424,7 @@ make_mipmap_tree(struct st_context *st, const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels) { + GLcontext *ctx = st->ctx; struct pipe_context *pipe = st->pipe; const struct gl_texture_format *mformat; struct pipe_mipmap_tree *mt; @@ -432,7 +433,7 @@ make_mipmap_tree(struct st_context *st, baseFormat = _mesa_base_format(format); - mformat = st_ChooseTextureFormat(st->ctx, baseFormat, format, type); + mformat = st_ChooseTextureFormat(ctx, baseFormat, format, type); assert(mformat); pipeFormat = st_mesa_format_to_pipe_format(mformat->MesaFormat); @@ -455,6 +456,10 @@ make_mipmap_tree(struct st_context *st, GLboolean success; GLuint pitch = mt->region->pitch; GLubyte *dest; + const GLbitfield imageTransferStateSave = ctx->_ImageTransferState; + + /* we'll do pixel transfer in a fragment shader */ + ctx->_ImageTransferState = 0x0; /* map texture region */ dest = pipe->region_map(pipe, mt->region); @@ -463,7 +468,7 @@ make_mipmap_tree(struct st_context *st, * Note that the image is actually going to be upside down in * the texture. We deal with that with texcoords. */ - success = mformat->StoreImage(st->ctx, 2, /* dims */ + success = mformat->StoreImage(ctx, 2, /* dims */ baseFormat, /* baseInternalFormat */ mformat, /* gl_texture_format */ dest, /* dest */ @@ -478,6 +483,9 @@ make_mipmap_tree(struct st_context *st, /* unmap */ pipe->region_unmap(pipe, mt->region); assert(success); + + /* restore */ + ctx->_ImageTransferState = imageTransferStateSave; } #if 0 -- cgit v1.2.3 From d6a739f6b0658414a81715bf690159f7cfdb4961 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 16:13:37 -0600 Subject: Use program serial numbers to avoid re-generating fragment programs for glDrawPixels. --- src/mesa/state_tracker/st_atom_pixeltransfer.c | 14 ++++--- src/mesa/state_tracker/st_cb_drawpixels.c | 52 +++++++++++++++++--------- src/mesa/state_tracker/st_cb_program.c | 6 ++- src/mesa/state_tracker/st_context.c | 4 +- src/mesa/state_tracker/st_context.h | 10 ++++- 5 files changed, 58 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index b190c39343..8cf340e685 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -42,6 +42,7 @@ #include "shader/prog_print.h" #include "st_context.h" +#include "st_program.h" @@ -79,8 +80,6 @@ is_identity(const GLfloat m[16]) static void make_state_key(GLcontext *ctx, struct state_key *key) { - /*GLuint i, j;*/ - memset(key, 0, sizeof(*key)); if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 || @@ -189,6 +188,8 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) inst[ic].SrcReg[0].Index = FRAG_RESULT_COLR; ic++; + /* XXX reimplement in terms of MUL/MAD (see t_vp_build.c) */ + /* DP4 result.color.x, tmp0, matrow0; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; @@ -268,6 +269,9 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) +/** + * Update st->pixel_xfer.program in response to new pixel-transfer state. + */ static void update_pixel_transfer(struct st_context *st) { @@ -277,14 +281,14 @@ update_pixel_transfer(struct st_context *st) make_state_key(st->ctx, &key); fp = (struct gl_fragment_program *) - _mesa_search_program_cache(st->pixel_transfer_cache, &key, sizeof(key)); + _mesa_search_program_cache(st->pixel_xfer.cache, &key, sizeof(key)); if (!fp) { fp = get_pixel_transfer_program(st->ctx, &key); - _mesa_program_cache_insert(st->ctx, st->pixel_transfer_cache, + _mesa_program_cache_insert(st->ctx, st->pixel_xfer.cache, &key, sizeof(key), &fp->Base); } - st->pixel_transfer_program = fp; + st->pixel_xfer.program = (struct st_fragment_program *) fp; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 3359338385..c74c8fa641 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -182,30 +182,48 @@ make_drawpix_fragment_shader(struct st_context *st) { GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; - struct gl_program *p; - /* - * XXX Use st_program's serial numbers to determine when the - * user-provided program and pixel-transfer program to avoid - * needless combining/translation here. - */ + if (st->pixel_xfer.program->serialNo == st->pixel_xfer.xfer_prog_sn + && st->fp->serialNo == st->pixel_xfer.user_prog_sn) { + /* the pixel tranfer program has not changed and the user-defined + * shader has not changed, so re-use the combined program. + */ + stfp = st->pixel_xfer.combined_prog; + } + else { + /* Concatenate the pixel transfer program with the current user- + * defined shader. + */ + stfp = (struct st_fragment_program *) + _mesa_combine_programs(ctx, + &st->pixel_xfer.program->Base.Base, + &st->fp->Base.Base); - p = _mesa_combine_programs(ctx, - &st->pixel_transfer_program->Base, - &ctx->FragmentProgram._Current->Base); #if 0 - _mesa_print_program(p); - printf("InputsRead: 0x%x\n", p->InputsRead); - printf("OutputsWritten: 0x%x\n", p->OutputsWritten); - _mesa_print_parameter_list(p->Parameters); + { + struct gl_program *p = &stfp->Base.Base; + _mesa_print_program(p); + printf("InputsRead: 0x%x\n", p->InputsRead); + printf("OutputsWritten: 0x%x\n", p->OutputsWritten); + _mesa_print_parameter_list(p->Parameters); + } #endif - stfp = (struct st_fragment_program *) p; - st_translate_fragment_program(st, stfp, NULL, - stfp->tokens, ST_MAX_SHADER_TOKENS); + /* translate to TGSI tokens */ + st_translate_fragment_program(st, stfp, NULL, + stfp->tokens, ST_MAX_SHADER_TOKENS); + /* save new program, update serial numbers */ + st->pixel_xfer.xfer_prog_sn = st->pixel_xfer.program->serialNo; + st->pixel_xfer.user_prog_sn = st->fp->serialNo; + st->pixel_xfer.combined_prog_sn = stfp->serialNo; + st->pixel_xfer.combined_prog = stfp; + } - st_upload_constants( st, p->Parameters, PIPE_SHADER_FRAGMENT ); + /* Ideally we'd have updated the pipe constants during the normal + * st/atom mechanism. But we can't since this is specific to glDrawPixels. + */ + st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT); return stfp; } diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index a330c1c922..aed6b1ee97 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -44,6 +44,8 @@ #include "st_atom_shader.h" +static GLuint SerialNo = 1; + /** * Called via ctx->Driver.BindProgram() to bind an ARB vertex or @@ -91,7 +93,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *prog = CALLOC_STRUCT(st_vertex_program); - prog->serialNo = 1; + prog->serialNo = SerialNo++; return _mesa_init_vertex_program( ctx, &prog->Base, @@ -103,7 +105,7 @@ static struct gl_program *st_new_program( GLcontext *ctx, case GL_FRAGMENT_PROGRAM_NV: { struct st_fragment_program *prog = CALLOC_STRUCT(st_fragment_program); - prog->serialNo = 1; + prog->serialNo = SerialNo++; return _mesa_init_fragment_program( ctx, &prog->Base, diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index e872e8b12c..c7398233d8 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -92,7 +92,7 @@ struct st_context *st_create_context( GLcontext *ctx, st->haveFramebufferRegions = GL_TRUE; - st->pixel_transfer_cache = _mesa_new_program_cache(); + st->pixel_xfer.cache = _mesa_new_program_cache(); #if 0 st_init_cb_clear( st ); @@ -126,7 +126,7 @@ void st_destroy_context( struct st_context *st ) #endif cso_cache_delete( st->cache ); - _mesa_delete_program_cache(st->ctx, st->pixel_transfer_cache); + _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache); st->pipe->destroy( st->pipe ); FREE( st ); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 4de70a3b2e..8bbfb0d6fe 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -133,8 +133,14 @@ struct st_context struct st_vertex_program *vp; /**< Currently bound vertex program */ struct st_fragment_program *fp; /**< Currently bound fragment program */ - struct gl_fragment_program *pixel_transfer_program; - struct gl_program_cache *pixel_transfer_cache; + struct { + struct gl_program_cache *cache; + struct st_fragment_program *program; /**< cur pixel transfer prog */ + GLuint xfer_prog_sn; /**< pixel xfer program serial no. */ + GLuint user_prog_sn; /**< user fragment program serial no. */ + struct st_fragment_program *combined_prog; + GLuint combined_prog_sn; + } pixel_xfer; /** * Buffer object which stores the ctx->Current.Attrib[] values. -- cgit v1.2.3 From 2dfe0c4a24feef8b7ffc5951d9d1867661493bf2 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 17:46:40 -0600 Subject: combine shaders for glCopyPixels --- src/mesa/state_tracker/st_cb_drawpixels.c | 143 +++++++++++++----------------- 1 file changed, 64 insertions(+), 79 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c74c8fa641..1864d65ee8 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -64,26 +64,23 @@ * If bitmapMode, use KIL instruction to kill the "0-pixels". */ static struct st_fragment_program * -make_fragment_shader(struct st_context *st, GLboolean bitmapMode) +make_bitmap_fragment_shader(struct st_context *st) { /* only make programs once and re-use */ - static struct st_fragment_program *progs[2] = { NULL, NULL }; + static struct st_fragment_program *prog = NULL; GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; struct gl_program *p; GLuint ic = 0; - if (progs[bitmapMode]) - return progs[bitmapMode]; + if (prog) + return prog; p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) return NULL; - if (bitmapMode) - p->NumInstructions = 6; - else - p->NumInstructions = 2; + p->NumInstructions = 6; p->Instructions = _mesa_alloc_instructions(p->NumInstructions); if (!p->Instructions) { @@ -91,96 +88,81 @@ make_fragment_shader(struct st_context *st, GLboolean bitmapMode) return NULL; } _mesa_init_instructions(p->Instructions, p->NumInstructions); - if (bitmapMode) { - /* - * XXX, we need to compose this fragment shader with the current - * user-provided fragment shader so the fragment program is applied - * to the fragments which aren't culled. - */ - /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */ - p->Instructions[ic].Opcode = OPCODE_TEX; - p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; - p->Instructions[ic].DstReg.Index = 0; - p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; - p->Instructions[ic].TexSrcUnit = 0; - p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; - ic++; - /* SWZ tmp0.x, tmp0.x, 1111; # tmp0.x = 1.0 */ - p->Instructions[ic].Opcode = OPCODE_SWZ; - p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; - p->Instructions[ic].DstReg.Index = 0; - p->Instructions[ic].DstReg.WriteMask = WRITEMASK_X; - p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[0].Index = 0; - p->Instructions[ic].SrcReg[0].Swizzle - = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE ); - ic++; + /* + * XXX, we need to compose this fragment shader with the current + * user-provided fragment shader so the fragment program is applied + * to the fragments which aren't culled. + */ + /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */ + p->Instructions[ic].Opcode = OPCODE_TEX; + p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; + p->Instructions[ic].DstReg.Index = 0; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + p->Instructions[ic].TexSrcUnit = 0; + p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; - /* SUB tmp0, tmp0.wwww, tmp0.xxxx; # tmp0.w -= 1 */ - p->Instructions[ic].Opcode = OPCODE_SUB; - p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; - p->Instructions[ic].DstReg.Index = 0; - p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[0].Index = 0; - p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; - p->Instructions[ic].SrcReg[1].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[1].Index = 0; - p->Instructions[ic].SrcReg[1].Swizzle = SWIZZLE_XXXX; /* 1.0 */ - ic++; + /* SWZ tmp0.x, tmp0.x, 1111; # tmp0.x = 1.0 */ + p->Instructions[ic].Opcode = OPCODE_SWZ; + p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; + p->Instructions[ic].DstReg.Index = 0; + p->Instructions[ic].DstReg.WriteMask = WRITEMASK_X; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + p->Instructions[ic].SrcReg[0].Swizzle + = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE ); + ic++; - /* KIL if tmp0 < 0 */ - p->Instructions[ic].Opcode = OPCODE_KIL; - p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[0].Index = 0; - ic++; + /* SUB tmp0, tmp0.wwww, tmp0.xxxx; # tmp0.w -= 1 */ + p->Instructions[ic].Opcode = OPCODE_SUB; + p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; + p->Instructions[ic].DstReg.Index = 0; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; + p->Instructions[ic].SrcReg[1].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[1].Index = 0; + p->Instructions[ic].SrcReg[1].Swizzle = SWIZZLE_XXXX; /* 1.0 */ + ic++; + + /* KIL if tmp0 < 0 */ + p->Instructions[ic].Opcode = OPCODE_KIL; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + ic++; + + /* MOV result.color, fragment.color */ + p->Instructions[ic].Opcode = OPCODE_MOV; + p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0; + ic++; - /* MOV result.color, fragment.color */ - p->Instructions[ic].Opcode = OPCODE_MOV; - p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; - p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0; - ic++; - } - else { - /* DrawPixels mode */ - /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ - p->Instructions[ic].Opcode = OPCODE_TEX; - p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; - p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; - p->Instructions[ic].TexSrcUnit = 0; - p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; - ic++; - } /* END; */ p->Instructions[ic++].Opcode = OPCODE_END; assert(ic == p->NumInstructions); - p->InputsRead = FRAG_BIT_TEX0; + p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0; p->OutputsWritten = (1 << FRAG_RESULT_COLR); - if (bitmapMode) { - p->InputsRead |= FRAG_BIT_COL0; - } stfp = (struct st_fragment_program *) p; st_translate_fragment_program(st, stfp, NULL, stfp->tokens, ST_MAX_SHADER_TOKENS); - progs[bitmapMode] = stfp; + prog = stfp; return stfp; } static struct st_fragment_program * -make_drawpix_fragment_shader(struct st_context *st) +make_drawpix_fragment_shader(GLcontext *ctx) { - GLcontext *ctx = st->ctx; + struct st_context *st = ctx->st; struct st_fragment_program *stfp; if (st->pixel_xfer.program->serialNo == st->pixel_xfer.xfer_prog_sn @@ -963,7 +945,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } else { ps = st->state.framebuffer.cbufs[0]; - stfp = make_drawpix_fragment_shader(ctx->st); + stfp = make_drawpix_fragment_shader(ctx); stvp = make_vertex_shader(ctx->st, GL_FALSE); color = NULL; } @@ -1146,7 +1128,7 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct pipe_mipmap_tree *mt; /* create the fragment program */ - stfp = make_fragment_shader(ctx->st, GL_TRUE); + stfp = make_bitmap_fragment_shader(ctx->st); /* and vertex program */ stvp = make_vertex_shader(ctx->st, GL_TRUE); @@ -1252,6 +1234,9 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, GLfloat *color; uint format; + /* make sure rendering has completed */ + pipe->flush(pipe, 0x0); + st_validate_state(st); if (type == GL_STENCIL) { @@ -1263,7 +1248,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, if (type == GL_COLOR) { rbRead = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); color = NULL; - stfp = make_fragment_shader(ctx->st, GL_FALSE); + stfp = make_drawpix_fragment_shader(ctx); stvp = make_vertex_shader(ctx->st, GL_FALSE); } else { -- cgit v1.2.3 From c1b2b97b8a7f2c9ff0f19c5bba352dfae11de4b2 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 18:24:54 -0600 Subject: special-case KIL/KIL_NV --- src/mesa/shader/prog_print.c | 20 +++++++++++++++++++- 1 file changed, 19 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index f410a9099e..969b9757cc 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -525,7 +525,7 @@ _mesa_print_instruction_opt(const struct prog_instruction *inst, GLint indent, _mesa_printf("_SAT"); _mesa_printf(" "); print_dst_reg(&inst->DstReg, mode, prog); - _mesa_printf("%s[%d], %s", + _mesa_printf(", %s[%d], %s", file_string((enum register_file) inst->SrcReg[0].File, mode), inst->SrcReg[0].Index, @@ -535,6 +535,7 @@ _mesa_print_instruction_opt(const struct prog_instruction *inst, GLint indent, break; case OPCODE_TEX: case OPCODE_TXP: + case OPCODE_TXL: case OPCODE_TXB: _mesa_printf("%s", _mesa_opcode_string(inst->Opcode)); if (inst->SaturateMode == SATURATE_ZERO_ONE) @@ -555,6 +556,23 @@ _mesa_print_instruction_opt(const struct prog_instruction *inst, GLint indent, } print_comment(inst); break; + + case OPCODE_KIL: + _mesa_printf("%s", _mesa_opcode_string(inst->Opcode)); + _mesa_printf(" "); + print_src_reg(&inst->SrcReg[0], mode, prog); + print_comment(inst); + break; + case OPCODE_KIL_NV: + _mesa_printf("%s", _mesa_opcode_string(inst->Opcode)); + _mesa_printf(" "); + _mesa_printf("%s.%s", + _mesa_condcode_string(inst->DstReg.CondMask), + _mesa_swizzle_string(inst->DstReg.CondSwizzle, + GL_FALSE, GL_FALSE)); + print_comment(inst); + break; + case OPCODE_ARL: _mesa_printf("ARL addr.x, "); print_src_reg(&inst->SrcReg[0], mode, prog); -- cgit v1.2.3 From fb9cf48259a4534219c7b8dff5ceaed7afde56ad Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 18:26:34 -0600 Subject: fix InputsRead bug in _mesa_combine_programs() --- src/mesa/shader/program.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 2d8b887791..ed1aacd068 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -469,6 +469,7 @@ _mesa_combine_programs(GLcontext *ctx, const GLuint lenB = progB->NumInstructions; const GLuint numParamsA = _mesa_num_parameters(progA->Parameters); const GLuint newLength = lenA + lenB; + GLbitfield inputsB; GLuint i; ASSERT(progA->Target == progB->Target); @@ -498,8 +499,11 @@ _mesa_combine_programs(GLcontext *ctx, PROGRAM_OUTPUT, FRAG_RESULT_COLR); } - newProg->InputsRead = progA->InputsRead; - newProg->InputsRead |= (progB->InputsRead & ~(1 << FRAG_ATTRIB_COL0)); + inputsB = progB->InputsRead; + if (progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) { + inputsB &= ~(1 << FRAG_ATTRIB_COL0); + } + newProg->InputsRead = progA->InputsRead | inputsB; newProg->OutputsWritten = progB->OutputsWritten; } else { -- cgit v1.2.3 From 520c71161ad0929a4b43e8e1e75067e10920f596 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 18:38:32 -0600 Subject: Implement shader concatenation for glBitmap. --- src/mesa/state_tracker/st_cb_drawpixels.c | 114 ++++++++++++++++++++---------- src/mesa/state_tracker/st_context.h | 6 ++ 2 files changed, 81 insertions(+), 39 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 1864d65ee8..ffaa34d7d3 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -59,28 +59,22 @@ /** - * Create a simple fragment shader that does a TEX() instruction to get - * the fragment color. - * If bitmapMode, use KIL instruction to kill the "0-pixels". + * Make fragment program for glBitmap: + * Sample the texture and kill the fragment if the bit is 0. + * This program will be combined with the user's fragment program. */ static struct st_fragment_program * -make_bitmap_fragment_shader(struct st_context *st) +make_bitmap_fragment_program(GLcontext *ctx) { - /* only make programs once and re-use */ - static struct st_fragment_program *prog = NULL; - GLcontext *ctx = st->ctx; struct st_fragment_program *stfp; struct gl_program *p; GLuint ic = 0; - if (prog) - return prog; - p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) return NULL; - p->NumInstructions = 6; + p->NumInstructions = 5; p->Instructions = _mesa_alloc_instructions(p->NumInstructions); if (!p->Instructions) { @@ -89,11 +83,6 @@ make_bitmap_fragment_shader(struct st_context *st) } _mesa_init_instructions(p->Instructions, p->NumInstructions); - /* - * XXX, we need to compose this fragment shader with the current - * user-provided fragment shader so the fragment program is applied - * to the fragments which aren't culled. - */ /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */ p->Instructions[ic].Opcode = OPCODE_TEX; p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; @@ -133,34 +122,84 @@ make_bitmap_fragment_shader(struct st_context *st) p->Instructions[ic].SrcReg[0].Index = 0; ic++; - /* MOV result.color, fragment.color */ - p->Instructions[ic].Opcode = OPCODE_MOV; - p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; - p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0; - ic++; - /* END; */ p->Instructions[ic++].Opcode = OPCODE_END; assert(ic == p->NumInstructions); - p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0; - p->OutputsWritten = (1 << FRAG_RESULT_COLR); + p->InputsRead = FRAG_BIT_TEX0; + p->OutputsWritten = 0x0; stfp = (struct st_fragment_program *) p; - st_translate_fragment_program(st, stfp, NULL, + st_translate_fragment_program(ctx->st, stfp, NULL, stfp->tokens, ST_MAX_SHADER_TOKENS); - prog = stfp; + return stfp; +} + + +/** + * Combine basic bitmap fragment program with the user-defined program. + */ +static struct st_fragment_program * +combined_bitmap_fragment_program(GLcontext *ctx) +{ + struct st_context *st = ctx->st; + struct st_fragment_program *stfp; + + if (!st->bitmap.program) { + /* create the basic bitmap fragment program */ + st->bitmap.program = make_bitmap_fragment_program(ctx); + } + + if (st->bitmap.user_prog_sn == st->fp->serialNo) { + /* re-use */ + stfp = st->bitmap.combined_prog; + } + else { + /* Concatenate the bitmap program with the current user-defined program. + */ + stfp = (struct st_fragment_program *) + _mesa_combine_programs(ctx, + &st->bitmap.program->Base.Base, + &st->fp->Base.Base); + +#if 0 + { + struct gl_program *p = &stfp->Base.Base; + printf("Combined bitmap program:\n"); + _mesa_print_program(p); + printf("InputsRead: 0x%x\n", p->InputsRead); + printf("OutputsWritten: 0x%x\n", p->OutputsWritten); + _mesa_print_parameter_list(p->Parameters); + } +#endif + + /* translate to TGSI tokens */ + st_translate_fragment_program(st, stfp, NULL, + stfp->tokens, ST_MAX_SHADER_TOKENS); + + /* save new program, update serial numbers */ + st->bitmap.user_prog_sn = st->fp->serialNo; + st->bitmap.combined_prog = stfp; + } + + /* Ideally we'd have updated the pipe constants during the normal + * st/atom mechanism. But we can't since this is specific to glBitmap. + */ + st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT); return stfp; } + +/** + * Make fragment shader for glDraw/CopyPixels. This shader is made + * by combining the pixel transfer shader with the user-defined shader. + */ static struct st_fragment_program * -make_drawpix_fragment_shader(GLcontext *ctx) +combined_drawpix_fragment_program(GLcontext *ctx) { struct st_context *st = ctx->st; struct st_fragment_program *stfp; @@ -168,13 +207,13 @@ make_drawpix_fragment_shader(GLcontext *ctx) if (st->pixel_xfer.program->serialNo == st->pixel_xfer.xfer_prog_sn && st->fp->serialNo == st->pixel_xfer.user_prog_sn) { /* the pixel tranfer program has not changed and the user-defined - * shader has not changed, so re-use the combined program. + * program has not changed, so re-use the combined program. */ stfp = st->pixel_xfer.combined_prog; } else { /* Concatenate the pixel transfer program with the current user- - * defined shader. + * defined program. */ stfp = (struct st_fragment_program *) _mesa_combine_programs(ctx, @@ -241,7 +280,7 @@ make_fragment_shader_z(struct st_context *st) } _mesa_init_instructions(p->Instructions, p->NumInstructions); - /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ + /* TEX result.depth, fragment.texcoord[0], texture[0], 2D; */ p->Instructions[ic].Opcode = OPCODE_TEX; p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; p->Instructions[ic].DstReg.Index = FRAG_RESULT_DEPR; @@ -945,7 +984,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } else { ps = st->state.framebuffer.cbufs[0]; - stfp = make_drawpix_fragment_shader(ctx); + stfp = combined_drawpix_fragment_program(ctx); stvp = make_vertex_shader(ctx->st, GL_FALSE); color = NULL; } @@ -1122,16 +1161,13 @@ static void st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap ) { - struct st_vertex_program *stvp; struct st_fragment_program *stfp; + struct st_vertex_program *stvp; struct st_context *st = ctx->st; struct pipe_mipmap_tree *mt; - /* create the fragment program */ - stfp = make_bitmap_fragment_shader(ctx->st); - - /* and vertex program */ stvp = make_vertex_shader(ctx->st, GL_TRUE); + stfp = combined_bitmap_fragment_program(ctx); st_validate_state(st); @@ -1248,7 +1284,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, if (type == GL_COLOR) { rbRead = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); color = NULL; - stfp = make_drawpix_fragment_shader(ctx); + stfp = combined_drawpix_fragment_program(ctx); stvp = make_vertex_shader(ctx->st, GL_FALSE); } else { diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 8bbfb0d6fe..7ab69df7c7 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -142,6 +142,12 @@ struct st_context GLuint combined_prog_sn; } pixel_xfer; + struct { + struct st_fragment_program *program; /**< bitmap tex/kil program */ + GLuint user_prog_sn; /**< user fragment program serial no. */ + struct st_fragment_program *combined_prog; + } bitmap; + /** * Buffer object which stores the ctx->Current.Attrib[] values. * Used for vertex array drawing when we we need an attribute for -- cgit v1.2.3 From 13acc263859a4b4689f519be8b5519c784baf3ba Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 31 Oct 2007 13:20:01 +0000 Subject: Chain vertex buffers into the batch buffer. --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 72 ++++++++++++++++++++----------- src/mesa/pipe/i915simple/i915_reg.h | 4 ++ 2 files changed, 52 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 0adf987721..8b5a08a3bb 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -67,19 +67,17 @@ static void vbuf_flush_elements( struct draw_stage *stage ); struct vbuf_stage { struct draw_stage stage; /**< This must be first (base class) */ + /** Vertex size in bytes */ + unsigned vertex_size; + /* FIXME: we have no guarantee that 'unsigned' is 32bit */ - - /* Vertices are passed in as an array of floats making up each - * attribute in turn. Will eventually convert to hardware format - * in this stage. - */ + + /** Vertices in hardware format */ unsigned *vertex_map; unsigned *vertex_ptr; - unsigned vertex_size; - unsigned nr_vertices; - unsigned max_vertices; - + unsigned nr_vertices; + ushort *element_map; unsigned nr_elements; @@ -274,11 +272,13 @@ static void vbuf_draw( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); struct i915_context *i915 = vbuf->i915; + struct pipe_winsys *winsys = i915->pipe.winsys; unsigned nr = vbuf->nr_elements; unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ unsigned hwprim; - unsigned *ptr; - unsigned i, j; + unsigned i; + char *ptr; + struct pipe_buffer_handle *buf; switch(vbuf->prim) { case PIPE_PRIM_POINTS: @@ -295,15 +295,37 @@ static void vbuf_draw( struct draw_stage *stage ) return; } + assert(vbuf->vertex_ptr - vbuf->vertex_map == vbuf->nr_vertices * vertex_size / 4); + + /* FIXME: handle failure */ + buf = winsys->buffer_create(winsys, 64); + winsys->buffer_data(winsys, buf, 8 + nr * vertex_size, NULL); + ptr = winsys->buffer_map(winsys, buf, PIPE_BUFFER_FLAG_WRITE); + *(unsigned *)ptr = _3DPRIMITIVE | + hwprim | + ((4 + vertex_size * nr)/4 - 2); + ptr += 4; + for (i = 0; i < nr; i++) { + memcpy(ptr, + (char*)vbuf->vertex_map + vbuf->element_map[i]*vertex_size, + vertex_size ); + ptr += vertex_size; + } + *(unsigned *)ptr = MI_BATCH_BUFFER_END; + ptr += 4; + winsys->buffer_unmap(winsys, buf); + if (i915->dirty) i915_update_derived( i915 ); if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); - assert(vbuf->vertex_ptr - vbuf->vertex_map == vbuf->nr_vertices * vertex_size / 4); - - ptr = BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 ); + ptr = BEGIN_BATCH( 2, 1 ); +#if 1 + assert(ptr); +#else + /* XXX: below is bogues as ptr always nonzero except in fatal errors */ if (ptr == 0) { FLUSH_BATCH(); @@ -312,21 +334,23 @@ static void vbuf_draw( struct draw_stage *stage ) i915_update_derived( i915 ); i915_emit_hardware_state( i915 ); - ptr = BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 ); + ptr = BEGIN_BATCH( 2, 1 ); if (ptr == 0) { assert(0); return; } } - - /* TODO: Fire a DMA buffer */ - OUT_BATCH(_3DPRIMITIVE | - hwprim | - ((4 + vertex_size * nr)/4 - 2)); - - for (i = 0; i < nr; i++) - for (j = 0; j < vertex_size / 4; j++) - OUT_BATCH(vbuf->vertex_map[vbuf->element_map[i]*vertex_size/4 + j]); +#endif + + /* chain the vertex buffer in the batch buffer */ + OUT_BATCH(MI_BATCH_BUFFER_START + | (2 << 6) /* GTT-mapped memory */); + OUT_RELOC( buf, I915_BUFFER_ACCESS_READ, 0 ); + /* FIXME: we need to flush here since control after chained buffers returns + * directly to the ring buffer */ + FLUSH_BATCH(); + + winsys->buffer_reference(winsys, &buf, NULL); } diff --git a/src/mesa/pipe/i915simple/i915_reg.h b/src/mesa/pipe/i915simple/i915_reg.h index f4070ef93e..04620fec68 100644 --- a/src/mesa/pipe/i915simple/i915_reg.h +++ b/src/mesa/pipe/i915simple/i915_reg.h @@ -907,6 +907,10 @@ #define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) #define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) +#define MI_BATCH_BUFFER (0x30<<23) +#define MI_BATCH_BUFFER_START (0x31<<23) +#define MI_BATCH_BUFFER_END (0xa<<23) + #define COMPAREFUNC_ALWAYS 0 -- cgit v1.2.3 From af60ebc4eedcfec2090e1ddd7630137948c26866 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 31 Oct 2007 14:16:00 +0000 Subject: Allow more verbose error output. --- src/mesa/pipe/i915simple/i915_fpc.h | 2 +- src/mesa/pipe/i915simple/i915_fpc_translate.c | 16 +++++++++++++--- 2 files changed, 14 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h index c25f5abf0e..8c7b68aefb 100644 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ b/src/mesa/pipe/i915simple/i915_fpc.h @@ -204,7 +204,7 @@ extern void i915_disassemble_program(const uint * program, uint sz); */ extern void -i915_program_error(struct i915_fp_compile *p, const char *msg); +i915_program_error(struct i915_fp_compile *p, const char *msg, ...); extern void i915_translate_fragment_program(struct i915_context *i915); diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index e041f89140..7db7e4f6ba 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -25,6 +25,9 @@ * **************************************************************************/ + +#include + #include "i915_reg.h" #include "i915_context.h" #include "i915_fpc.h" @@ -111,9 +114,16 @@ i915_use_passthrough_shader(struct i915_context *i915) void -i915_program_error(struct i915_fp_compile *p, const char *msg) +i915_program_error(struct i915_fp_compile *p, const char *msg, ...) { - fprintf(stderr, "i915_program_error: %s\n", msg); + va_list args; + + fprintf(stderr, "i915_program_error: "); + va_start( args, msg ); + vfprintf( stderr, msg, args ); + va_end( args ); + fprintf(stderr, "\n"); + p->error = 1; } @@ -855,7 +865,7 @@ i915_translate_instruction(struct i915_fp_compile *p, break; default: - i915_program_error(p, "bad opcode"); + i915_program_error(p, "bad opcode %d", inst->Instruction.Opcode); return; } -- cgit v1.2.3 From 44e6944e9dda3b78f2f4a9bdd583a4feeae1dcba Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 31 Oct 2007 14:58:43 +0000 Subject: Handle TGSI_OPCODE_RET. --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 7db7e4f6ba..7838031c17 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -673,6 +673,10 @@ i915_translate_instruction(struct i915_fp_compile *p, flags, 0, swizzle(tmp, X, X, X, X), 0, 0); break; + case TGSI_OPCODE_RET: + /* XXX: no-op? */ + break; + case TGSI_OPCODE_RCP: src0 = src_vector(p, &inst->FullSrcRegisters[0]); -- cgit v1.2.3 From fe55eab8fa976fdf3d30b1d8160e4bb10e754e2d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 31 Oct 2007 15:11:18 +0000 Subject: Hold the vertex buffer handle. --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 17 +++++++++-------- src/mesa/pipe/i915simple/i915_winsys.h | 4 ++++ 2 files changed, 13 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 8b5a08a3bb..dc6e091681 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -81,6 +81,8 @@ struct vbuf_stage { ushort *element_map; unsigned nr_elements; + struct pipe_buffer_handle *buf; + unsigned prim; struct i915_context *i915; @@ -278,7 +280,6 @@ static void vbuf_draw( struct draw_stage *stage ) unsigned hwprim; unsigned i; char *ptr; - struct pipe_buffer_handle *buf; switch(vbuf->prim) { case PIPE_PRIM_POINTS: @@ -298,9 +299,11 @@ static void vbuf_draw( struct draw_stage *stage ) assert(vbuf->vertex_ptr - vbuf->vertex_map == vbuf->nr_vertices * vertex_size / 4); /* FIXME: handle failure */ - buf = winsys->buffer_create(winsys, 64); - winsys->buffer_data(winsys, buf, 8 + nr * vertex_size, NULL); - ptr = winsys->buffer_map(winsys, buf, PIPE_BUFFER_FLAG_WRITE); + if(!vbuf->buf) + vbuf->buf = winsys->buffer_create(winsys, 64); + + winsys->buffer_data(winsys, vbuf->buf, 8 + nr * vertex_size, NULL); + ptr = winsys->buffer_map(winsys, vbuf->buf, PIPE_BUFFER_FLAG_WRITE); *(unsigned *)ptr = _3DPRIMITIVE | hwprim | ((4 + vertex_size * nr)/4 - 2); @@ -313,7 +316,7 @@ static void vbuf_draw( struct draw_stage *stage ) } *(unsigned *)ptr = MI_BATCH_BUFFER_END; ptr += 4; - winsys->buffer_unmap(winsys, buf); + winsys->buffer_unmap(winsys, vbuf->buf); if (i915->dirty) i915_update_derived( i915 ); @@ -345,12 +348,10 @@ static void vbuf_draw( struct draw_stage *stage ) /* chain the vertex buffer in the batch buffer */ OUT_BATCH(MI_BATCH_BUFFER_START | (2 << 6) /* GTT-mapped memory */); - OUT_RELOC( buf, I915_BUFFER_ACCESS_READ, 0 ); + OUT_RELOC( vbuf->buf, I915_BUFFER_ACCESS_READ, 0 ); /* FIXME: we need to flush here since control after chained buffers returns * directly to the ring buffer */ FLUSH_BATCH(); - - winsys->buffer_reference(winsys, &buf, NULL); } diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index a95927d803..544763644c 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -69,6 +69,10 @@ struct i915_winsys { unsigned delta ); void (*batch_flush)( struct i915_winsys *sws ); +#if 0 + void (*batch_chain)( struct i915_winsys *sws, + struct pipe_buffer_handle *buf ): +#endif }; #define I915_BUFFER_ACCESS_WRITE 0x1 -- cgit v1.2.3 From 2860f609949ced58caeb4828e7d2a34439d767da Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 10:52:44 -0600 Subject: re-enable fb size assertions which were previously disabled --- src/mesa/drivers/dri/common/drirenderbuffer.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/common/drirenderbuffer.c b/src/mesa/drivers/dri/common/drirenderbuffer.c index d4f7f2527b..d34da53479 100644 --- a/src/mesa/drivers/dri/common/drirenderbuffer.c +++ b/src/mesa/drivers/dri/common/drirenderbuffer.c @@ -209,11 +209,9 @@ driUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv) struct gl_framebuffer *fb = (struct gl_framebuffer *) dPriv->driverPrivate; if (fb && (dPriv->w != fb->Width || dPriv->h != fb->Height)) { ctx->Driver.ResizeBuffers(ctx, fb, dPriv->w, dPriv->h); -#if 0 /* if the driver needs the hw lock for ResizeBuffers, the drawable might have changed again by now */ assert(fb->Width == dPriv->w); assert(fb->Height == dPriv->h); -#endif } } -- cgit v1.2.3 From d775509fb7736aac2d4c7903a93c0d7cb8015a91 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 11:08:07 -0600 Subject: plug st_invalidate_state() into ctx->Driver.UpdateState Start lifting Mesa stuff up out of winsys/driver code. --- src/mesa/state_tracker/st_context.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index c7398233d8..95c5eec26e 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -51,12 +51,20 @@ #include "pipe/cso_cache/cso_cache.h" +/** + * Called via ctx->Driver.UpdateState() + */ void st_invalidate_state(GLcontext * ctx, GLuint new_state) { struct st_context *st = st_context(ctx); st->dirty.mesa |= new_state; st->dirty.st |= ST_NEW_MESA; + + /* This is the only core Mesa module we depend upon. + * No longer use swrast, swsetup, tnl. + */ + _vbo_InvalidateState(ctx, new_state); } @@ -149,4 +157,6 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_texture_functions(functions); st_init_flush_functions(functions); st_init_string_functions(functions); + + functions->UpdateState = st_invalidate_state; } -- cgit v1.2.3 From b31e37f14d75231724a1cbb0c7fe7031a2315671 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 11:08:31 -0600 Subject: remove intelInvalidateState --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 13 ------------- 1 file changed, 13 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 71bc5003d2..975085b168 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -151,16 +151,6 @@ static const struct dri_debug_control debug_control[] = { }; #endif -static void -intelInvalidateState(GLcontext * ctx, GLuint new_state) -{ - _vbo_InvalidateState(ctx, new_state); - _tnl_InvalidateState(ctx, new_state); - _tnl_invalidate_vertex_state(ctx, new_state); - - st_invalidate_state( ctx, new_state ); -} - void intelFlush(GLcontext * ctx) @@ -180,9 +170,6 @@ static void intelInitDriverFunctions(struct dd_function_table *functions) { _mesa_init_driver_functions(functions); - - functions->UpdateState = intelInvalidateState; - st_init_driver_functions(functions); } -- cgit v1.2.3 From 8984a283389a13f9b4315aa7b7df1eaaa612ba28 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 11:35:50 -0600 Subject: Lift VBO/tnl stuff up out of drivers --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 18 +++--------------- src/mesa/drivers/x11/xm_api.c | 11 +++++++++++ src/mesa/state_tracker/st_context.c | 8 ++++++++ 3 files changed, 22 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 975085b168..790f67b552 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -30,10 +30,6 @@ #include "context.h" #include "extensions.h" -#include "tnl/tnl.h" -#include "tnl/t_pipeline.h" -#include "tnl/t_vertex.h" - #include "drivers/common/driverfuncs.h" #include "i830_dri.h" @@ -210,11 +206,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->driScreen->myNum, "i915"); - /* Initialize the software rasterizer and helper modules. */ - _vbo_CreateContext(ctx); - _tnl_CreateContext(ctx); - - /* * memory pools */ @@ -230,9 +221,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->driFd = sPriv->fd; intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; - TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline; - - fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); intel->iw.irq_seq = -1; intel->irqsEmitted = 0; @@ -309,8 +297,6 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) //intel->vtbl.destroy(intel); release_texture_heaps = (intel->ctx.Shared->RefCount == 1); - _tnl_DestroyContext(&intel->ctx); - _vbo_DestroyContext(&intel->ctx); intel_batchbuffer_free(intel->batch); @@ -332,8 +318,10 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) */ } - /* free the Mesa context */ + /* free the Mesa context data */ _mesa_free_context_data(&intel->ctx); + + st_destroy_context(intel->ctx.st); } } diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 4b31447bbd..ffe8361d26 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -76,9 +76,11 @@ #include "swrast/swrast.h" #include "swrast_setup/swrast_setup.h" #include "vbo/vbo.h" +#if 0 #include "tnl/tnl.h" #include "tnl/t_context.h" #include "tnl/t_pipeline.h" +#endif #include "drivers/common/driverfuncs.h" #include "state_tracker/st_public.h" @@ -1528,7 +1530,9 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) XMesaContext c; GLcontext *mesaCtx; struct dd_function_table functions; +#if 0 TNLcontext *tnl; +#endif if (firstTime) { _glthread_INIT_MUTEX(_xmesa_lock); @@ -1591,17 +1595,22 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) /* Initialize the software rasterizer and helper modules. */ if (!_swrast_CreateContext( mesaCtx ) || +#if 0 !_vbo_CreateContext( mesaCtx ) || !_tnl_CreateContext( mesaCtx ) || +#endif !_swsetup_CreateContext( mesaCtx )) { _mesa_free_context_data(&c->mesa); _mesa_free(c); return NULL; } +#if 0 /* tnl setup */ tnl = TNL_CONTEXT(mesaCtx); tnl->Driver.RunPipeline = _tnl_run_pipeline; +#endif + /* swrast setup */ xmesa_register_swrast_functions( mesaCtx ); _swsetup_Wakeup(mesaCtx); @@ -1641,8 +1650,10 @@ void XMesaDestroyContext( XMesaContext c ) _swsetup_DestroyContext( mesaCtx ); _swrast_DestroyContext( mesaCtx ); +#if 0 _tnl_DestroyContext( mesaCtx ); _vbo_DestroyContext( mesaCtx ); +#endif _mesa_free_context_data( mesaCtx ); _mesa_free( c ); } diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 95c5eec26e..97e178bbe1 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -78,6 +78,11 @@ struct st_context *st_create_context( GLcontext *ctx, st->ctx = ctx; st->pipe = pipe; + /* state tracker needs the VBO module */ + _vbo_CreateContext(ctx); + /* XXX temporary */ + _tnl_CreateContext(ctx); + st->draw = draw_create(); /* for selection/feedback */ st->dirty.mesa = ~0; @@ -125,6 +130,9 @@ void st_destroy_context( struct st_context *st ) st_destroy_atoms( st ); st_destroy_draw( st ); + _vbo_DestroyContext(st->ctx); + _tnl_DestroyContext(st->ctx); /* XXX temporary */ + #if 0 st_destroy_cb_clear( st ); st_destroy_cb_program( st ); -- cgit v1.2.3 From 3798395af5106c2100ea56b23762ad76890c6351 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:07:03 -0600 Subject: move a few lines of code --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 790f67b552..6c880498e5 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -225,6 +225,10 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->iw.irq_seq = -1; intel->irqsEmitted = 0; + intel->batch = intel_batchbuffer_alloc(intel); + intel->last_swap_fence = NULL; + intel->first_swap_fence = NULL; + /* Disable imaging extension until convolution is working in * teximage paths: */ @@ -232,11 +236,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, /* GL_TRUE, */ GL_FALSE); - - intel->batch = intel_batchbuffer_alloc(intel); - intel->last_swap_fence = NULL; - intel->first_swap_fence = NULL; - if (intel->ctx.Mesa_DXTn) { _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); _mesa_enable_extension(ctx, "GL_S3_s3tc"); -- cgit v1.2.3 From b26aae67f5fe4194b48a5d3ddf704797b804b58c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:00:38 -0600 Subject: alloc caches for fixed-func vertex/fragment progs --- src/mesa/main/mtypes.h | 7 +++++++ src/mesa/shader/program.c | 4 ++++ 2 files changed, 11 insertions(+) (limited to 'src') diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 44587ae962..b435c29793 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -123,6 +123,7 @@ typedef int GLfixed; /*@{*/ struct _mesa_HashTable; struct gl_pixelstore_attrib; +struct gl_program_cache; struct gl_texture_format; struct gl_texture_image; struct gl_texture_object; @@ -2000,6 +2001,9 @@ struct gl_vertex_program_state /** Program to emulate fixed-function T&L (see above) */ struct gl_vertex_program *_TnlProgram; + /** Cache of fixed-function programs */ + struct gl_program_cache *Cache; + #if FEATURE_MESA_program_debug GLprogramcallbackMESA Callback; GLvoid *CallbackData; @@ -2033,6 +2037,9 @@ struct gl_fragment_program_state /** Program to emulate fixed-function texture env/combine (see above) */ struct gl_fragment_program *_TexEnvProgram; + /** Cache of fixed-function programs */ + struct gl_program_cache *Cache; + #if FEATURE_MESA_program_debug GLprogramcallbackMESA Callback; GLvoid *CallbackData; diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index ed1aacd068..f13aa18e95 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -33,6 +33,7 @@ #include "context.h" #include "hash.h" #include "program.h" +#include "prog_cache.h" #include "prog_parameter.h" #include "prog_instruction.h" @@ -66,6 +67,7 @@ _mesa_init_program(GLcontext *ctx) ctx->VertexProgram.TrackMatrix[i] = GL_NONE; ctx->VertexProgram.TrackMatrixTransform[i] = GL_IDENTITY_NV; } + ctx->VertexProgram.Cache = _mesa_new_program_cache(); #endif #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program @@ -73,8 +75,10 @@ _mesa_init_program(GLcontext *ctx) ctx->FragmentProgram.Current = (struct gl_fragment_program *) ctx->Shared->DefaultFragmentProgram; assert(ctx->FragmentProgram.Current); ctx->FragmentProgram.Current->Base.RefCount++; + ctx->FragmentProgram.Cache = _mesa_new_program_cache(); #endif + /* XXX probably move this stuff */ #if FEATURE_ATI_fragment_shader ctx->ATIFragmentShader.Enabled = GL_FALSE; -- cgit v1.2.3 From 287d573f5d992416e8355b597224e817c137387b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:01:08 -0600 Subject: fix type for Samplers field --- src/mesa/shader/prog_execute.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/prog_execute.h b/src/mesa/shader/prog_execute.h index db7bcee516..af77d4278b 100644 --- a/src/mesa/shader/prog_execute.h +++ b/src/mesa/shader/prog_execute.h @@ -62,7 +62,7 @@ struct gl_program_machine GLuint CondCodes[4]; /**< COND_* value for x/y/z/w */ GLint AddressReg[MAX_PROGRAM_ADDRESS_REGS][4]; - GLuint *Samplers; /** Array mapping sampler var to tex unit */ + const GLubyte *Samplers; /** Array mapping sampler var to tex unit */ GLuint CallStack[MAX_PROGRAM_CALL_DEPTH]; /**< For CAL/RET instructions */ GLuint StackDepth; /**< Index/ptr to top of CallStack[] */ -- cgit v1.2.3 From cf3f601682297d94482a1448eff3f36b26514ab1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:03:55 -0600 Subject: Lift fixed function vertex program generation up from tnl module. --- src/mesa/main/ffvertex_prog.c | 1550 +++++++++++++++++++++++++++++++++++++++++ src/mesa/main/ffvertex_prog.h | 38 + 2 files changed, 1588 insertions(+) create mode 100644 src/mesa/main/ffvertex_prog.c create mode 100644 src/mesa/main/ffvertex_prog.h (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c new file mode 100644 index 0000000000..8fcb9e5b14 --- /dev/null +++ b/src/mesa/main/ffvertex_prog.c @@ -0,0 +1,1550 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file ffvertex_prog. + * + * Create a vertex program to execute the current fixed function T&L pipeline. + * \author Keith Whitwell + */ + + +#include "main/glheader.h" +#include "main/mtypes.h" +#include "main/macros.h" +#include "main/enums.h" +#include "main/ffvertex_prog.h" +#include "shader/program.h" +#include "shader/prog_cache.h" +#include "shader/prog_instruction.h" +#include "shader/prog_parameter.h" +#include "shader/prog_print.h" +#include "shader/prog_statevars.h" + + +struct state_key { + unsigned light_global_enabled:1; + unsigned light_local_viewer:1; + unsigned light_twoside:1; + unsigned light_color_material:1; + unsigned light_color_material_mask:12; + unsigned light_material_mask:12; + + unsigned normalize:1; + unsigned rescale_normals:1; + unsigned fog_source_is_depth:1; + unsigned tnl_do_vertex_fog:1; + unsigned separate_specular:1; + unsigned fog_mode:2; + unsigned point_attenuated:1; + unsigned texture_enabled_global:1; + unsigned fragprog_inputs_read:12; + + struct { + unsigned light_enabled:1; + unsigned light_eyepos3_is_zero:1; + unsigned light_spotcutoff_is_180:1; + unsigned light_attenuated:1; + unsigned texunit_really_enabled:1; + unsigned texmat_enabled:1; + unsigned texgen_enabled:4; + unsigned texgen_mode0:4; + unsigned texgen_mode1:4; + unsigned texgen_mode2:4; + unsigned texgen_mode3:4; + } unit[8]; +}; + + + +#define FOG_NONE 0 +#define FOG_LINEAR 1 +#define FOG_EXP 2 +#define FOG_EXP2 3 + +static GLuint translate_fog_mode( GLenum mode ) +{ + switch (mode) { + case GL_LINEAR: return FOG_LINEAR; + case GL_EXP: return FOG_EXP; + case GL_EXP2: return FOG_EXP2; + default: return FOG_NONE; + } +} + +#define TXG_NONE 0 +#define TXG_OBJ_LINEAR 1 +#define TXG_EYE_LINEAR 2 +#define TXG_SPHERE_MAP 3 +#define TXG_REFLECTION_MAP 4 +#define TXG_NORMAL_MAP 5 + +static GLuint translate_texgen( GLboolean enabled, GLenum mode ) +{ + if (!enabled) + return TXG_NONE; + + switch (mode) { + case GL_OBJECT_LINEAR: return TXG_OBJ_LINEAR; + case GL_EYE_LINEAR: return TXG_EYE_LINEAR; + case GL_SPHERE_MAP: return TXG_SPHERE_MAP; + case GL_REFLECTION_MAP_NV: return TXG_REFLECTION_MAP; + case GL_NORMAL_MAP_NV: return TXG_NORMAL_MAP; + default: return TXG_NONE; + } +} + + +/** + * Returns bitmask of flags indicating which materials are set per-vertex + * in the current VB. + * XXX get these from the VBO... + */ +static GLbitfield +tnl_get_per_vertex_materials(GLcontext *ctx) +{ + GLbitfield mask = 0x0; +#if 0 + TNLcontext *tnl = TNL_CONTEXT(ctx); + struct vertex_buffer *VB = &tnl->vb; + GLuint i; + + for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) + if (VB->AttribPtr[i] && VB->AttribPtr[i]->stride) + mask |= 1 << (i - _TNL_FIRST_MAT); +#endif + return mask; +} + +/** + * Should fog be computed per-vertex? + */ +static GLboolean +tnl_get_per_vertex_fog(GLcontext *ctx) +{ +#if 0 + TNLcontext *tnl = TNL_CONTEXT(ctx); + return tnl->_DoVertexFog; +#else + return GL_FALSE; +#endif +} + + +static struct state_key *make_state_key( GLcontext *ctx ) +{ + const struct gl_fragment_program *fp; + struct state_key *key = CALLOC_STRUCT(state_key); + GLuint i; + + fp = ctx->FragmentProgram._Current; + + /* This now relies on texenvprogram.c being active: + */ + assert(fp); + + key->fragprog_inputs_read = fp->Base.InputsRead; + + if (ctx->RenderMode == GL_FEEDBACK) { + /* make sure the vertprog emits color and tex0 */ + key->fragprog_inputs_read |= (FRAG_BIT_COL0 | FRAG_BIT_TEX0); + } + + key->separate_specular = (ctx->Light.Model.ColorControl == + GL_SEPARATE_SPECULAR_COLOR); + + if (ctx->Light.Enabled) { + key->light_global_enabled = 1; + + if (ctx->Light.Model.LocalViewer) + key->light_local_viewer = 1; + + if (ctx->Light.Model.TwoSide) + key->light_twoside = 1; + + if (ctx->Light.ColorMaterialEnabled) { + key->light_color_material = 1; + key->light_color_material_mask = ctx->Light.ColorMaterialBitmask; + } + + key->light_material_mask = tnl_get_per_vertex_materials(ctx); + + for (i = 0; i < MAX_LIGHTS; i++) { + struct gl_light *light = &ctx->Light.Light[i]; + + if (light->Enabled) { + key->unit[i].light_enabled = 1; + + if (light->EyePosition[3] == 0.0) + key->unit[i].light_eyepos3_is_zero = 1; + + if (light->SpotCutoff == 180.0) + key->unit[i].light_spotcutoff_is_180 = 1; + + if (light->ConstantAttenuation != 1.0 || + light->LinearAttenuation != 0.0 || + light->QuadraticAttenuation != 0.0) + key->unit[i].light_attenuated = 1; + } + } + } + + if (ctx->Transform.Normalize) + key->normalize = 1; + + if (ctx->Transform.RescaleNormals) + key->rescale_normals = 1; + + key->fog_mode = translate_fog_mode(fp->FogOption); + + if (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT) + key->fog_source_is_depth = 1; + + key->tnl_do_vertex_fog = tnl_get_per_vertex_fog(ctx); + + if (ctx->Point._Attenuated) + key->point_attenuated = 1; + + if (ctx->Texture._TexGenEnabled || + ctx->Texture._TexMatEnabled || + ctx->Texture._EnabledUnits) + key->texture_enabled_global = 1; + + for (i = 0; i < MAX_TEXTURE_UNITS; i++) { + struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i]; + + if (texUnit->_ReallyEnabled) + key->unit[i].texunit_really_enabled = 1; + + if (ctx->Texture._TexMatEnabled & ENABLE_TEXMAT(i)) + key->unit[i].texmat_enabled = 1; + + if (texUnit->TexGenEnabled) { + key->unit[i].texgen_enabled = 1; + + key->unit[i].texgen_mode0 = + translate_texgen( texUnit->TexGenEnabled & (1<<0), + texUnit->GenModeS ); + key->unit[i].texgen_mode1 = + translate_texgen( texUnit->TexGenEnabled & (1<<1), + texUnit->GenModeT ); + key->unit[i].texgen_mode2 = + translate_texgen( texUnit->TexGenEnabled & (1<<2), + texUnit->GenModeR ); + key->unit[i].texgen_mode3 = + translate_texgen( texUnit->TexGenEnabled & (1<<3), + texUnit->GenModeQ ); + } + } + + return key; +} + + + +/* Very useful debugging tool - produces annotated listing of + * generated program with line/function references for each + * instruction back into this file: + */ +#define DISASSEM (MESA_VERBOSE&VERBOSE_DISASSEM) + +/* Should be tunable by the driver - do we want to do matrix + * multiplications with DP4's or with MUL/MAD's? SSE works better + * with the latter, drivers may differ. + */ +#define PREFER_DP4 0 + +#define MAX_INSN 256 + +/* Use uregs to represent registers internally, translate to Mesa's + * expected formats on emit. + * + * NOTE: These are passed by value extensively in this file rather + * than as usual by pointer reference. If this disturbs you, try + * remembering they are just 32bits in size. + * + * GCC is smart enough to deal with these dword-sized structures in + * much the same way as if I had defined them as dwords and was using + * macros to access and set the fields. This is much nicer and easier + * to evolve. + */ +struct ureg { + GLuint file:4; + GLint idx:8; /* relative addressing may be negative */ + GLuint negate:1; + GLuint swz:12; + GLuint pad:7; +}; + + +struct tnl_program { + const struct state_key *state; + struct gl_vertex_program *program; + + GLuint temp_in_use; + GLuint temp_reserved; + + struct ureg eye_position; + struct ureg eye_position_normalized; + struct ureg eye_normal; + struct ureg identity; + + GLuint materials; + GLuint color_materials; +}; + + +static const struct ureg undef = { + PROGRAM_UNDEFINED, + ~0, + 0, + 0, + 0 +}; + +/* Local shorthand: + */ +#define X SWIZZLE_X +#define Y SWIZZLE_Y +#define Z SWIZZLE_Z +#define W SWIZZLE_W + + +/* Construct a ureg: + */ +static struct ureg make_ureg(GLuint file, GLint idx) +{ + struct ureg reg; + reg.file = file; + reg.idx = idx; + reg.negate = 0; + reg.swz = SWIZZLE_NOOP; + reg.pad = 0; + return reg; +} + + + +static struct ureg negate( struct ureg reg ) +{ + reg.negate ^= 1; + return reg; +} + + +static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w ) +{ + reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x), + GET_SWZ(reg.swz, y), + GET_SWZ(reg.swz, z), + GET_SWZ(reg.swz, w)); + + return reg; +} + +static struct ureg swizzle1( struct ureg reg, int x ) +{ + return swizzle(reg, x, x, x, x); +} + +static struct ureg get_temp( struct tnl_program *p ) +{ + int bit = _mesa_ffs( ~p->temp_in_use ); + if (!bit) { + _mesa_problem(NULL, "%s: out of temporaries\n", __FILE__); + _mesa_exit(1); + } + + if ((GLuint) bit > p->program->Base.NumTemporaries) + p->program->Base.NumTemporaries = bit; + + p->temp_in_use |= 1<<(bit-1); + return make_ureg(PROGRAM_TEMPORARY, bit-1); +} + +static struct ureg reserve_temp( struct tnl_program *p ) +{ + struct ureg temp = get_temp( p ); + p->temp_reserved |= 1<temp_in_use &= ~(1<temp_in_use |= p->temp_reserved; /* can't release reserved temps */ + } +} + +static void release_temps( struct tnl_program *p ) +{ + p->temp_in_use = p->temp_reserved; +} + + + +static struct ureg register_input( struct tnl_program *p, GLuint input ) +{ + p->program->Base.InputsRead |= (1<program->Base.OutputsWritten |= (1<program->Base.Parameters, values, 4, + &swizzle ); + ASSERT(swizzle == SWIZZLE_NOOP); + return make_ureg(PROGRAM_STATE_VAR, idx); +} + +#define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1) +#define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0) +#define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1) +#define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1) + +static GLboolean is_undef( struct ureg reg ) +{ + return reg.file == PROGRAM_UNDEFINED; +} + +static struct ureg get_identity_param( struct tnl_program *p ) +{ + if (is_undef(p->identity)) + p->identity = register_const4f(p, 0,0,0,1); + + return p->identity; +} + +static struct ureg register_param5(struct tnl_program *p, + GLint s0, + GLint s1, + GLint s2, + GLint s3, + GLint s4) +{ + gl_state_index tokens[STATE_LENGTH]; + GLint idx; + tokens[0] = s0; + tokens[1] = s1; + tokens[2] = s2; + tokens[3] = s3; + tokens[4] = s4; + idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens ); + return make_ureg(PROGRAM_STATE_VAR, idx); +} + + +#define register_param1(p,s0) register_param5(p,s0,0,0,0,0) +#define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0) +#define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0) +#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0) + + +static void register_matrix_param5( struct tnl_program *p, + GLint s0, /* modelview, projection, etc */ + GLint s1, /* texture matrix number */ + GLint s2, /* first row */ + GLint s3, /* last row */ + GLint s4, /* inverse, transpose, etc */ + struct ureg *matrix ) +{ + GLint i; + + /* This is a bit sad as the support is there to pull the whole + * matrix out in one go: + */ + for (i = 0; i <= s3 - s2; i++) + matrix[i] = register_param5( p, s0, s1, i, i, s4 ); +} + + +static void emit_arg( struct prog_src_register *src, + struct ureg reg ) +{ + src->File = reg.file; + src->Index = reg.idx; + src->Swizzle = reg.swz; + src->NegateBase = reg.negate ? NEGATE_XYZW : 0; + src->Abs = 0; + src->NegateAbs = 0; + src->RelAddr = 0; +} + +static void emit_dst( struct prog_dst_register *dst, + struct ureg reg, GLuint mask ) +{ + dst->File = reg.file; + dst->Index = reg.idx; + /* allow zero as a shorthand for xyzw */ + dst->WriteMask = mask ? mask : WRITEMASK_XYZW; + dst->CondMask = COND_TR; /* always pass cond test */ + dst->CondSwizzle = SWIZZLE_NOOP; + dst->CondSrc = 0; + dst->pad = 0; +} + +static void debug_insn( struct prog_instruction *inst, const char *fn, + GLuint line ) +{ + if (DISASSEM) { + static const char *last_fn; + + if (fn != last_fn) { + last_fn = fn; + _mesa_printf("%s:\n", fn); + } + + _mesa_printf("%d:\t", line); + _mesa_print_instruction(inst); + } +} + + +static void emit_op3fn(struct tnl_program *p, + enum prog_opcode op, + struct ureg dest, + GLuint mask, + struct ureg src0, + struct ureg src1, + struct ureg src2, + const char *fn, + GLuint line) +{ + GLuint nr = p->program->Base.NumInstructions++; + struct prog_instruction *inst = &p->program->Base.Instructions[nr]; + + if (p->program->Base.NumInstructions > MAX_INSN) { + _mesa_problem(0, "Out of instructions in emit_op3fn\n"); + return; + } + + inst->Opcode = (enum prog_opcode) op; + inst->StringPos = 0; + inst->Data = 0; + + emit_arg( &inst->SrcReg[0], src0 ); + emit_arg( &inst->SrcReg[1], src1 ); + emit_arg( &inst->SrcReg[2], src2 ); + + emit_dst( &inst->DstReg, dest, mask ); + + debug_insn(inst, fn, line); +} + + +#define emit_op3(p, op, dst, mask, src0, src1, src2) \ + emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__) + +#define emit_op2(p, op, dst, mask, src0, src1) \ + emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__) + +#define emit_op1(p, op, dst, mask, src0) \ + emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__) + + +static struct ureg make_temp( struct tnl_program *p, struct ureg reg ) +{ + if (reg.file == PROGRAM_TEMPORARY && + !(p->temp_reserved & (1<eye_position)) { + struct ureg pos = register_input( p, VERT_ATTRIB_POS ); + struct ureg modelview[4]; + + p->eye_position = reserve_temp(p); + + if (PREFER_DP4) { + register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3, + 0, modelview ); + + emit_matrix_transform_vec4(p, p->eye_position, modelview, pos); + } + else { + register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3, + STATE_MATRIX_TRANSPOSE, modelview ); + + emit_transpose_matrix_transform_vec4(p, p->eye_position, modelview, pos); + } + } + + return p->eye_position; +} + + +static struct ureg get_eye_position_normalized( struct tnl_program *p ) +{ + if (is_undef(p->eye_position_normalized)) { + struct ureg eye = get_eye_position(p); + p->eye_position_normalized = reserve_temp(p); + emit_normalize_vec3(p, p->eye_position_normalized, eye); + } + + return p->eye_position_normalized; +} + + +static struct ureg get_eye_normal( struct tnl_program *p ) +{ + if (is_undef(p->eye_normal)) { + struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL ); + struct ureg mvinv[3]; + + register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 2, + STATE_MATRIX_INVTRANS, mvinv ); + + p->eye_normal = reserve_temp(p); + + /* Transform to eye space: + */ + emit_matrix_transform_vec3( p, p->eye_normal, mvinv, normal ); + + /* Normalize/Rescale: + */ + if (p->state->normalize) { + emit_normalize_vec3( p, p->eye_normal, p->eye_normal ); + } + else if (p->state->rescale_normals) { + struct ureg rescale = register_param2(p, STATE_INTERNAL, + STATE_NORMAL_SCALE); + + emit_op2( p, OPCODE_MUL, p->eye_normal, 0, p->eye_normal, + swizzle1(rescale, X)); + } + } + + return p->eye_normal; +} + + + +static void build_hpos( struct tnl_program *p ) +{ + struct ureg pos = register_input( p, VERT_ATTRIB_POS ); + struct ureg hpos = register_output( p, VERT_RESULT_HPOS ); + struct ureg mvp[4]; + + if (PREFER_DP4) { + register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, + 0, mvp ); + emit_matrix_transform_vec4( p, hpos, mvp, pos ); + } + else { + register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, + STATE_MATRIX_TRANSPOSE, mvp ); + emit_transpose_matrix_transform_vec4( p, hpos, mvp, pos ); + } +} + + +static GLuint material_attrib( GLuint side, GLuint property ) +{ + return ((property - STATE_AMBIENT) * 2 + + side); +} + +/* Get a bitmask of which material values vary on a per-vertex basis. + */ +static void set_material_flags( struct tnl_program *p ) +{ + p->color_materials = 0; + p->materials = 0; + + if (p->state->light_color_material) { + p->materials = + p->color_materials = p->state->light_color_material_mask; + } + + p->materials |= p->state->light_material_mask; +} + + +/* XXX temporary!!! */ +#define _TNL_ATTRIB_MAT_FRONT_AMBIENT 32 + +static struct ureg get_material( struct tnl_program *p, GLuint side, + GLuint property ) +{ + GLuint attrib = material_attrib(side, property); + + if (p->color_materials & (1<materials & (1<materials & SCENE_COLOR_BITS(side)) { + struct ureg lm_ambient = register_param1(p, STATE_LIGHTMODEL_AMBIENT); + struct ureg material_emission = get_material(p, side, STATE_EMISSION); + struct ureg material_ambient = get_material(p, side, STATE_AMBIENT); + struct ureg material_diffuse = get_material(p, side, STATE_DIFFUSE); + struct ureg tmp = make_temp(p, material_diffuse); + emit_op3(p, OPCODE_MAD, tmp, WRITEMASK_XYZ, lm_ambient, + material_ambient, material_emission); + return tmp; + } + else + return register_param2( p, STATE_LIGHTMODEL_SCENECOLOR, side ); +} + + +static struct ureg get_lightprod( struct tnl_program *p, GLuint light, + GLuint side, GLuint property ) +{ + GLuint attrib = material_attrib(side, property); + if (p->materials & (1<state->unit[i].light_spotcutoff_is_180) { + struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL, + STATE_SPOT_DIR_NORMALIZED, i); + struct ureg spot = get_temp(p); + struct ureg slt = get_temp(p); + + emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot_dir_norm); + emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot); + emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W)); + emit_op2(p, OPCODE_MUL, att, 0, slt, spot); + + release_temp(p, spot); + release_temp(p, slt); + } + + /* Calculate distance attenuation: + */ + if (p->state->unit[i].light_attenuated) { + + /* 1/d,d,d,1/d */ + emit_op1(p, OPCODE_RCP, dist, WRITEMASK_YZ, dist); + /* 1,d,d*d,1/d */ + emit_op2(p, OPCODE_MUL, dist, WRITEMASK_XZ, dist, swizzle1(dist,Y)); + /* 1/dist-atten */ + emit_op2(p, OPCODE_DP3, dist, 0, attenuation, dist); + + if (!p->state->unit[i].light_spotcutoff_is_180) { + /* dist-atten */ + emit_op1(p, OPCODE_RCP, dist, 0, dist); + /* spot-atten * dist-atten */ + emit_op2(p, OPCODE_MUL, att, 0, dist, att); + } else { + /* dist-atten */ + emit_op1(p, OPCODE_RCP, att, 0, dist); + } + } + + return att; +} + + + + + +/* Need to add some addtional parameters to allow lighting in object + * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye + * space lighting. + */ +static void build_lighting( struct tnl_program *p ) +{ + const GLboolean twoside = p->state->light_twoside; + const GLboolean separate = p->state->separate_specular; + GLuint nr_lights = 0, count = 0; + struct ureg normal = get_eye_normal(p); + struct ureg lit = get_temp(p); + struct ureg dots = get_temp(p); + struct ureg _col0 = undef, _col1 = undef; + struct ureg _bfc0 = undef, _bfc1 = undef; + GLuint i; + + for (i = 0; i < MAX_LIGHTS; i++) + if (p->state->unit[i].light_enabled) + nr_lights++; + + set_material_flags(p); + + { + struct ureg shininess = get_material(p, 0, STATE_SHININESS); + emit_op1(p, OPCODE_MOV, dots, WRITEMASK_W, swizzle1(shininess,X)); + release_temp(p, shininess); + + _col0 = make_temp(p, get_scenecolor(p, 0)); + if (separate) + _col1 = make_temp(p, get_identity_param(p)); + else + _col1 = _col0; + + } + + if (twoside) { + struct ureg shininess = get_material(p, 1, STATE_SHININESS); + emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z, + negate(swizzle1(shininess,X))); + release_temp(p, shininess); + + _bfc0 = make_temp(p, get_scenecolor(p, 1)); + if (separate) + _bfc1 = make_temp(p, get_identity_param(p)); + else + _bfc1 = _bfc0; + } + + + /* If no lights, still need to emit the scenecolor. + */ + { + struct ureg res0 = register_output( p, VERT_RESULT_COL0 ); + emit_op1(p, OPCODE_MOV, res0, 0, _col0); + } + + if (separate) { + struct ureg res1 = register_output( p, VERT_RESULT_COL1 ); + emit_op1(p, OPCODE_MOV, res1, 0, _col1); + } + + if (twoside) { + struct ureg res0 = register_output( p, VERT_RESULT_BFC0 ); + emit_op1(p, OPCODE_MOV, res0, 0, _bfc0); + } + + if (twoside && separate) { + struct ureg res1 = register_output( p, VERT_RESULT_BFC1 ); + emit_op1(p, OPCODE_MOV, res1, 0, _bfc1); + } + + if (nr_lights == 0) { + release_temps(p); + return; + } + + + for (i = 0; i < MAX_LIGHTS; i++) { + if (p->state->unit[i].light_enabled) { + struct ureg half = undef; + struct ureg att = undef, VPpli = undef; + + count++; + + if (p->state->unit[i].light_eyepos3_is_zero) { + /* Can used precomputed constants in this case. + * Attenuation never applies to infinite lights. + */ + VPpli = register_param3(p, STATE_LIGHT, i, + STATE_POSITION_NORMALIZED); + if (p->state->light_local_viewer) { + struct ureg eye_hat = get_eye_position_normalized(p); + half = get_temp(p); + emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); + emit_normalize_vec3(p, half, half); + } else { + half = register_param3(p, STATE_LIGHT, i, STATE_HALF_VECTOR); + } + } + else { + struct ureg Ppli = register_param3(p, STATE_LIGHT, i, + STATE_POSITION); + struct ureg V = get_eye_position(p); + struct ureg dist = get_temp(p); + + VPpli = get_temp(p); + half = get_temp(p); + + /* Calulate VPpli vector + */ + emit_op2(p, OPCODE_SUB, VPpli, 0, Ppli, V); + + /* Normalize VPpli. The dist value also used in + * attenuation below. + */ + emit_op2(p, OPCODE_DP3, dist, 0, VPpli, VPpli); + emit_op1(p, OPCODE_RSQ, dist, 0, dist); + emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist); + + + /* Calculate attenuation: + */ + if (!p->state->unit[i].light_spotcutoff_is_180 || + p->state->unit[i].light_attenuated) { + att = calculate_light_attenuation(p, i, VPpli, dist); + } + + + /* Calculate viewer direction, or use infinite viewer: + */ + if (p->state->light_local_viewer) { + struct ureg eye_hat = get_eye_position_normalized(p); + emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); + } + else { + struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z); + emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir); + } + + emit_normalize_vec3(p, half, half); + + release_temp(p, dist); + } + + /* Calculate dot products: + */ + emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); + emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half); + + + /* Front face lighting: + */ + { + struct ureg ambient = get_lightprod(p, i, 0, STATE_AMBIENT); + struct ureg diffuse = get_lightprod(p, i, 0, STATE_DIFFUSE); + struct ureg specular = get_lightprod(p, i, 0, STATE_SPECULAR); + struct ureg res0, res1; + GLuint mask0, mask1; + + emit_op1(p, OPCODE_LIT, lit, 0, dots); + + if (!is_undef(att)) + emit_op2(p, OPCODE_MUL, lit, 0, lit, att); + + + if (count == nr_lights) { + if (separate) { + mask0 = WRITEMASK_XYZ; + mask1 = WRITEMASK_XYZ; + res0 = register_output( p, VERT_RESULT_COL0 ); + res1 = register_output( p, VERT_RESULT_COL1 ); + } + else { + mask0 = 0; + mask1 = WRITEMASK_XYZ; + res0 = _col0; + res1 = register_output( p, VERT_RESULT_COL0 ); + } + } else { + mask0 = 0; + mask1 = 0; + res0 = _col0; + res1 = _col1; + } + + emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0); + emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0); + emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1); + + release_temp(p, ambient); + release_temp(p, diffuse); + release_temp(p, specular); + } + + /* Back face lighting: + */ + if (twoside) { + struct ureg ambient = get_lightprod(p, i, 1, STATE_AMBIENT); + struct ureg diffuse = get_lightprod(p, i, 1, STATE_DIFFUSE); + struct ureg specular = get_lightprod(p, i, 1, STATE_SPECULAR); + struct ureg res0, res1; + GLuint mask0, mask1; + + emit_op1(p, OPCODE_LIT, lit, 0, negate(swizzle(dots,X,Y,W,Z))); + + if (!is_undef(att)) + emit_op2(p, OPCODE_MUL, lit, 0, lit, att); + + if (count == nr_lights) { + if (separate) { + mask0 = WRITEMASK_XYZ; + mask1 = WRITEMASK_XYZ; + res0 = register_output( p, VERT_RESULT_BFC0 ); + res1 = register_output( p, VERT_RESULT_BFC1 ); + } + else { + mask0 = 0; + mask1 = WRITEMASK_XYZ; + res0 = _bfc0; + res1 = register_output( p, VERT_RESULT_BFC0 ); + } + } else { + res0 = _bfc0; + res1 = _bfc1; + mask0 = 0; + mask1 = 0; + } + + emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0); + emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0); + emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1); + + release_temp(p, ambient); + release_temp(p, diffuse); + release_temp(p, specular); + } + + release_temp(p, half); + release_temp(p, VPpli); + release_temp(p, att); + } + } + + release_temps( p ); +} + + +static void build_fog( struct tnl_program *p ) +{ + struct ureg fog = register_output(p, VERT_RESULT_FOGC); + struct ureg input; + + if (p->state->fog_source_is_depth) { + input = swizzle1(get_eye_position(p), Z); + } + else { + input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X); + } + + if (p->state->fog_mode && p->state->tnl_do_vertex_fog) { + struct ureg params = register_param2(p, STATE_INTERNAL, + STATE_FOG_PARAMS_OPTIMIZED); + struct ureg tmp = get_temp(p); + GLboolean useabs = (p->state->fog_mode != FOG_EXP2); + + if (useabs) { + emit_op1(p, OPCODE_ABS, tmp, 0, input); + } + + switch (p->state->fog_mode) { + case FOG_LINEAR: { + struct ureg id = get_identity_param(p); + emit_op3(p, OPCODE_MAD, tmp, 0, useabs ? tmp : input, + swizzle1(params,X), swizzle1(params,Y)); + emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */ + emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W)); + break; + } + case FOG_EXP: + emit_op2(p, OPCODE_MUL, tmp, 0, useabs ? tmp : input, + swizzle1(params,Z)); + emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp)); + break; + case FOG_EXP2: + emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W)); + emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp); + emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp)); + break; + } + + release_temp(p, tmp); + } + else { + /* results = incoming fog coords (compute fog per-fragment later) + * + * KW: Is it really necessary to do anything in this case? + * BP: Yes, we always need to compute the absolute value, unless + * we want to push that down into the fragment program... + */ + GLboolean useabs = GL_TRUE; + emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, WRITEMASK_X, input); + } +} + +static void build_reflect_texgen( struct tnl_program *p, + struct ureg dest, + GLuint writemask ) +{ + struct ureg normal = get_eye_normal(p); + struct ureg eye_hat = get_eye_position_normalized(p); + struct ureg tmp = get_temp(p); + + /* n.u */ + emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); + /* 2n.u */ + emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); + /* (-2n.u)n + u */ + emit_op3(p, OPCODE_MAD, dest, writemask, negate(tmp), normal, eye_hat); + + release_temp(p, tmp); +} + +static void build_sphere_texgen( struct tnl_program *p, + struct ureg dest, + GLuint writemask ) +{ + struct ureg normal = get_eye_normal(p); + struct ureg eye_hat = get_eye_position_normalized(p); + struct ureg tmp = get_temp(p); + struct ureg half = register_scalar_const(p, .5); + struct ureg r = get_temp(p); + struct ureg inv_m = get_temp(p); + struct ureg id = get_identity_param(p); + + /* Could share the above calculations, but it would be + * a fairly odd state for someone to set (both sphere and + * reflection active for different texture coordinate + * components. Of course - if two texture units enable + * reflect and/or sphere, things start to tilt in favour + * of seperating this out: + */ + + /* n.u */ + emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); + /* 2n.u */ + emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); + /* (-2n.u)n + u */ + emit_op3(p, OPCODE_MAD, r, 0, negate(tmp), normal, eye_hat); + /* r + 0,0,1 */ + emit_op2(p, OPCODE_ADD, tmp, 0, r, swizzle(id,X,Y,W,Z)); + /* rx^2 + ry^2 + (rz+1)^2 */ + emit_op2(p, OPCODE_DP3, tmp, 0, tmp, tmp); + /* 2/m */ + emit_op1(p, OPCODE_RSQ, tmp, 0, tmp); + /* 1/m */ + emit_op2(p, OPCODE_MUL, inv_m, 0, tmp, half); + /* r/m + 1/2 */ + emit_op3(p, OPCODE_MAD, dest, writemask, r, inv_m, half); + + release_temp(p, tmp); + release_temp(p, r); + release_temp(p, inv_m); +} + + +static void build_texture_transform( struct tnl_program *p ) +{ + GLuint i, j; + + for (i = 0; i < MAX_TEXTURE_UNITS; i++) { + + if (!(p->state->fragprog_inputs_read & FRAG_BIT_TEX(i))) + continue; + + if (p->state->unit[i].texgen_enabled || + p->state->unit[i].texmat_enabled) { + + GLuint texmat_enabled = p->state->unit[i].texmat_enabled; + struct ureg out = register_output(p, VERT_RESULT_TEX0 + i); + struct ureg out_texgen = undef; + + if (p->state->unit[i].texgen_enabled) { + GLuint copy_mask = 0; + GLuint sphere_mask = 0; + GLuint reflect_mask = 0; + GLuint normal_mask = 0; + GLuint modes[4]; + + if (texmat_enabled) + out_texgen = get_temp(p); + else + out_texgen = out; + + modes[0] = p->state->unit[i].texgen_mode0; + modes[1] = p->state->unit[i].texgen_mode1; + modes[2] = p->state->unit[i].texgen_mode2; + modes[3] = p->state->unit[i].texgen_mode3; + + for (j = 0; j < 4; j++) { + switch (modes[j]) { + case TXG_OBJ_LINEAR: { + struct ureg obj = register_input(p, VERT_ATTRIB_POS); + struct ureg plane = + register_param3(p, STATE_TEXGEN, i, + STATE_TEXGEN_OBJECT_S + j); + + emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, + obj, plane ); + break; + } + case TXG_EYE_LINEAR: { + struct ureg eye = get_eye_position(p); + struct ureg plane = + register_param3(p, STATE_TEXGEN, i, + STATE_TEXGEN_EYE_S + j); + + emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, + eye, plane ); + break; + } + case TXG_SPHERE_MAP: + sphere_mask |= WRITEMASK_X << j; + break; + case TXG_REFLECTION_MAP: + reflect_mask |= WRITEMASK_X << j; + break; + case TXG_NORMAL_MAP: + normal_mask |= WRITEMASK_X << j; + break; + case TXG_NONE: + copy_mask |= WRITEMASK_X << j; + } + + } + + + if (sphere_mask) { + build_sphere_texgen(p, out_texgen, sphere_mask); + } + + if (reflect_mask) { + build_reflect_texgen(p, out_texgen, reflect_mask); + } + + if (normal_mask) { + struct ureg normal = get_eye_normal(p); + emit_op1(p, OPCODE_MOV, out_texgen, normal_mask, normal ); + } + + if (copy_mask) { + struct ureg in = register_input(p, VERT_ATTRIB_TEX0+i); + emit_op1(p, OPCODE_MOV, out_texgen, copy_mask, in ); + } + } + + if (texmat_enabled) { + struct ureg texmat[4]; + struct ureg in = (!is_undef(out_texgen) ? + out_texgen : + register_input(p, VERT_ATTRIB_TEX0+i)); + if (PREFER_DP4) { + register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3, + 0, texmat ); + emit_matrix_transform_vec4( p, out, texmat, in ); + } + else { + register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3, + STATE_MATRIX_TRANSPOSE, texmat ); + emit_transpose_matrix_transform_vec4( p, out, texmat, in ); + } + } + + release_temps(p); + } + else { + emit_passthrough(p, VERT_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i); + } + } +} + + +static void build_pointsize( struct tnl_program *p ) +{ + struct ureg eye = get_eye_position(p); + struct ureg state_size = register_param1(p, STATE_POINT_SIZE); + struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION); + struct ureg out = register_output(p, VERT_RESULT_PSIZ); + struct ureg ut = get_temp(p); + + /* dist = |eyez| */ + emit_op1(p, OPCODE_ABS, ut, WRITEMASK_Y, swizzle1(eye, Z)); + /* p1 + dist * (p2 + dist * p3); */ + emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y), + swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y)); + emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y), + ut, swizzle1(state_attenuation, X)); + + /* 1 / sqrt(factor) */ + emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut ); + +#if 1 + /* out = pointSize / sqrt(factor) */ + emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size); +#else + /* not sure, might make sense to do clamping here, + but it's not done in t_vb_points neither */ + emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size); + emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y)); + emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z)); +#endif + + release_temp(p, ut); +} + +/** + * Emit constant point size. + */ +static void constant_pointsize( struct tnl_program *p ) +{ + struct ureg state_size = register_param1(p, STATE_POINT_SIZE); + struct ureg out = register_output(p, VERT_RESULT_PSIZ); + emit_op1(p, OPCODE_MOV, out, WRITEMASK_X, state_size); +} + +static void build_tnl_program( struct tnl_program *p ) +{ /* Emit the program, starting with modelviewproject: + */ + build_hpos(p); + + /* Lighting calculations: + */ + if (p->state->fragprog_inputs_read & (FRAG_BIT_COL0|FRAG_BIT_COL1)) { + if (p->state->light_global_enabled) + build_lighting(p); + else { + if (p->state->fragprog_inputs_read & FRAG_BIT_COL0) + emit_passthrough(p, VERT_ATTRIB_COLOR0, VERT_RESULT_COL0); + + if (p->state->fragprog_inputs_read & FRAG_BIT_COL1) + emit_passthrough(p, VERT_ATTRIB_COLOR1, VERT_RESULT_COL1); + } + } + + if ((p->state->fragprog_inputs_read & FRAG_BIT_FOGC) || + p->state->fog_mode != FOG_NONE) + build_fog(p); + + if (p->state->fragprog_inputs_read & FRAG_BITS_TEX_ANY) + build_texture_transform(p); + + if (p->state->point_attenuated) + build_pointsize(p); +#if 0 + else + constant_pointsize(p); +#endif + + /* Finish up: + */ + emit_op1(p, OPCODE_END, undef, 0, undef); + + /* Disassemble: + */ + if (DISASSEM) { + _mesa_printf ("\n"); + } +} + + +static void +create_new_program( const struct state_key *key, + struct gl_vertex_program *program, + GLuint max_temps) +{ + struct tnl_program p; + + _mesa_memset(&p, 0, sizeof(p)); + p.state = key; + p.program = program; + p.eye_position = undef; + p.eye_position_normalized = undef; + p.eye_normal = undef; + p.identity = undef; + p.temp_in_use = 0; + + if (max_temps >= sizeof(int) * 8) + p.temp_reserved = 0; + else + p.temp_reserved = ~((1<Base.Instructions = _mesa_alloc_instructions(MAX_INSN); + p.program->Base.String = NULL; + p.program->Base.NumInstructions = + p.program->Base.NumTemporaries = + p.program->Base.NumParameters = + p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0; + p.program->Base.Parameters = _mesa_new_parameter_list(); + p.program->Base.InputsRead = 0; + p.program->Base.OutputsWritten = 0; + + build_tnl_program( &p ); +} + + +/** + * Return a vertex program which implements the current fixed-function + * transform/lighting/texgen operations. + * XXX move this into core mesa (main/) + */ +struct gl_vertex_program * +_mesa_get_fixed_func_vertex_program(GLcontext *ctx) +{ + struct gl_vertex_program *prog; + struct state_key *key; + + /* Grab all the relevent state and put it in a single structure: + */ + key = make_state_key(ctx); + + /* Look for an already-prepared program for this state: + */ + prog = (struct gl_vertex_program *) + _mesa_search_program_cache(ctx->VertexProgram.Cache, key, sizeof(*key)); + + if (!prog) { + /* OK, we'll have to build a new one */ + if (0) + _mesa_printf("Build new TNL program\n"); + + prog = (struct gl_vertex_program *) + ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); + + create_new_program( key, prog, + ctx->Const.VertexProgram.MaxTemps ); + +#if 0 + if (ctx->Driver.ProgramStringNotify) + ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB, + &prog->Base ); +#endif + _mesa_program_cache_insert(ctx, ctx->VertexProgram.Cache, + key, sizeof(*key), &prog->Base); + } + else { + /* use cached program */ + _mesa_free(key); + } + + return prog; +} diff --git a/src/mesa/main/ffvertex_prog.h b/src/mesa/main/ffvertex_prog.h new file mode 100644 index 0000000000..74cafc65c1 --- /dev/null +++ b/src/mesa/main/ffvertex_prog.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef FFVERTEX_PROG_H +#define FFVERTEX_PROG_H + + +struct gl_vertex_program * +_mesa_get_fixed_func_vertex_program(GLcontext *ctx); + + + +#endif /* FFVERTEX_PROG_H */ -- cgit v1.2.3 From 4990695f1b197b270a4e3d98dd6bcd2bb4e89b4a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:12:37 -0600 Subject: fix some breakage from lifting vbo/tnl code --- src/mesa/drivers/x11/xm_api.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index ffe8361d26..08c98eab48 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -1551,6 +1551,9 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) xmesa_init_driver_functions(v, &functions); st_init_driver_functions(&functions); + /* override st's function */ + functions.UpdateState = xmesa_update_state; + /* functions.NewRenderbuffer = xmesa_new_renderbuffer; */ @@ -1594,12 +1597,13 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) /* Initialize the software rasterizer and helper modules. */ - if (!_swrast_CreateContext( mesaCtx ) || + if (!_swrast_CreateContext( mesaCtx ) #if 0 - !_vbo_CreateContext( mesaCtx ) || + || !_vbo_CreateContext( mesaCtx ) || !_tnl_CreateContext( mesaCtx ) || + !_swsetup_CreateContext( mesaCtx ) #endif - !_swsetup_CreateContext( mesaCtx )) { + ) { _mesa_free_context_data(&c->mesa); _mesa_free(c); return NULL; @@ -1613,12 +1617,14 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) /* swrast setup */ xmesa_register_swrast_functions( mesaCtx ); - _swsetup_Wakeup(mesaCtx); st_create_context( mesaCtx, xmesa_create_softpipe( c ) ); + _swsetup_CreateContext( mesaCtx ); + _swsetup_Wakeup(mesaCtx); + /* override these functions, as if the xlib driver were derived from * the softpipe driver. */ -- cgit v1.2.3 From 8d9afa76eb090ff58ca9a8a7a86a0b23ffc56857 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:17:32 -0600 Subject: Use ffvertex_prog.c code instead of t_vp_build.c code. --- src/mesa/main/ffvertex_prog.h | 2 ++ src/mesa/sources | 1 + src/mesa/state_tracker/st_draw.c | 4 ---- src/mesa/tnl/t_context.c | 4 ++++ src/mesa/tnl/t_vp_build.c | 6 +++++- src/mesa/tnl/t_vp_build.h | 4 ++++ 6 files changed, 16 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.h b/src/mesa/main/ffvertex_prog.h index 74cafc65c1..38dc5fbb8d 100644 --- a/src/mesa/main/ffvertex_prog.h +++ b/src/mesa/main/ffvertex_prog.h @@ -30,6 +30,8 @@ #define FFVERTEX_PROG_H +#include "main/mtypes.h" + struct gl_vertex_program * _mesa_get_fixed_func_vertex_program(GLcontext *ctx); diff --git a/src/mesa/sources b/src/mesa/sources index 727872378e..845a5ff556 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -27,6 +27,7 @@ MAIN_SOURCES = \ main/extensions.c \ main/fbobject.c \ main/feedback.c \ + main/ffvertex_prog.c \ main/fog.c \ main/framebuffer.c \ main/get.c \ diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 065e157bc6..c3f33a447e 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -36,8 +36,6 @@ #include "vbo/vbo.h" #include "vbo/vbo_context.h" -#include "tnl/t_vp_build.h" - #include "st_atom.h" #include "st_cache.h" #include "st_context.h" @@ -527,8 +525,6 @@ void st_init_draw( struct st_context *st ) assert(vbo); assert(vbo->draw_prims); vbo->draw_prims = st_draw_vbo; - - _tnl_ProgramCacheInit( ctx ); } diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c index 3b8dd18bbb..60770a91c2 100644 --- a/src/mesa/tnl/t_context.c +++ b/src/mesa/tnl/t_context.c @@ -61,7 +61,9 @@ _tnl_CreateContext( GLcontext *ctx ) /* Initialize tnl state. */ if (ctx->VertexProgram._MaintainTnlProgram) { +#if 0 _tnl_ProgramCacheInit( ctx ); +#endif _tnl_install_pipeline( ctx, _tnl_vp_pipeline ); } else { _tnl_install_pipeline( ctx, _tnl_default_pipeline ); @@ -90,8 +92,10 @@ _tnl_DestroyContext( GLcontext *ctx ) _tnl_destroy_pipeline( ctx ); +#if 0 if (ctx->VertexProgram._MaintainTnlProgram) _tnl_ProgramCacheDestroy( ctx ); +#endif FREE(tnl); ctx->swtnl_context = NULL; diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index 2b7d8eebe0..215d6653a3 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -33,6 +33,7 @@ #include "glheader.h" #include "macros.h" #include "enums.h" +#include "main/ffvertex_prog.h" #include "shader/program.h" #include "shader/prog_instruction.h" #include "shader/prog_parameter.h" @@ -41,7 +42,7 @@ #include "t_context.h" /* NOTE: very light dependency on this */ #include "t_vp_build.h" - +#if 0 struct state_key { unsigned light_global_enabled:1; unsigned light_local_viewer:1; @@ -1605,6 +1606,7 @@ _mesa_get_fixed_func_vertex_program(GLcontext *ctx) return prog; } +#endif void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ) @@ -1627,6 +1629,7 @@ void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ) } } +#if 0 void _tnl_ProgramCacheInit( GLcontext *ctx ) { TNLcontext *tnl = TNL_CONTEXT(ctx); @@ -1655,3 +1658,4 @@ void _tnl_ProgramCacheDestroy( GLcontext *ctx ) FREE(tnl->vp_cache->items); FREE(tnl->vp_cache); } +#endif diff --git a/src/mesa/tnl/t_vp_build.h b/src/mesa/tnl/t_vp_build.h index 034701d8c4..adcd8f1662 100644 --- a/src/mesa/tnl/t_vp_build.h +++ b/src/mesa/tnl/t_vp_build.h @@ -37,12 +37,16 @@ _NEW_FOG | \ _NEW_POINT) +#if 0 extern struct gl_vertex_program * _mesa_get_fixed_func_vertex_program(GLcontext *ctx); +#endif extern void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ); +#if 0 extern void _tnl_ProgramCacheInit( GLcontext *ctx ); extern void _tnl_ProgramCacheDestroy( GLcontext *ctx ); +#endif #endif -- cgit v1.2.3 From 68ab379be09de775244bb787f0d30e562fc21038 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:27:47 -0600 Subject: more flags for MaintainTnlProgram case, update #includes --- src/mesa/main/state.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 0e4cf047c8..2f88ec615a 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -62,6 +62,7 @@ #if FEATURE_EXT_framebuffer_object #include "fbobject.h" #endif +#include "ffvertex_prog.h" #include "framebuffer.h" #include "hint.h" #include "histogram.h" @@ -92,7 +93,6 @@ #include "shader/program.h" #include "texenvprogram.h" #endif -#include "tnl/t_vp_build.h" #if FEATURE_ARB_shader_objects #include "shaders.h" #endif @@ -1202,7 +1202,9 @@ _mesa_update_state_locked( GLcontext *ctx ) prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR); } if (ctx->VertexProgram._MaintainTnlProgram) { - prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _NEW_LIGHT); + prog_flags |= (_NEW_TEXTURE | _NEW_TEXTURE_MATRIX | + _NEW_TRANSFORM | _NEW_POINT | + _NEW_FOG | _NEW_LIGHT); } if (new_state & prog_flags) update_program( ctx ); -- cgit v1.2.3 From d4dfe3e0b43cc2bab1f15175e6e296702733e1af Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:30:56 -0600 Subject: No longer need st_update_tnl atom. --- src/mesa/sources | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 845a5ff556..c983d802d4 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -205,7 +205,6 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_clip.c \ state_tracker/st_atom_constbuf.c \ state_tracker/st_atom_depth.c \ - state_tracker/st_atom_fixedfunction.c \ state_tracker/st_atom_framebuffer.c \ state_tracker/st_atom_pixeltransfer.c \ state_tracker/st_atom_sampler.c \ -- cgit v1.2.3 From 8db4acc5547370761a9a489c947e9621adc8f945 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:31:25 -0600 Subject: No longer need st_update_tnl atom --- src/mesa/state_tracker/st_atom.c | 1 - src/mesa/state_tracker/st_atom.h | 1 - 2 files changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index d967b9b978..0797ea615e 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -49,7 +49,6 @@ static const struct st_tracked_state *atoms[] = &st_update_depth_stencil, &st_update_clip, - &st_update_tnl, &st_update_shader, &st_update_rasterizer, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index a7caca5459..63ce35958a 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -48,7 +48,6 @@ const struct st_tracked_state st_update_framebuffer; const struct st_tracked_state st_update_clip; const struct st_tracked_state st_update_clear_color; const struct st_tracked_state st_update_depth_stencil; -const struct st_tracked_state st_update_tnl; const struct st_tracked_state st_update_shader; const struct st_tracked_state st_update_rasterizer; const struct st_tracked_state st_update_polygon_stipple; -- cgit v1.2.3 From a35a2fc0d3aaa795a0088bfc390b98a033fc28de Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:34:09 -0600 Subject: remove dead program cache code --- src/mesa/tnl/t_context.c | 8 - src/mesa/tnl/t_context.h | 17 - src/mesa/tnl/t_vp_build.c | 1604 +-------------------------------------------- src/mesa/tnl/t_vp_build.h | 10 - 4 files changed, 1 insertion(+), 1638 deletions(-) (limited to 'src') diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c index 60770a91c2..b87452d2eb 100644 --- a/src/mesa/tnl/t_context.c +++ b/src/mesa/tnl/t_context.c @@ -61,9 +61,6 @@ _tnl_CreateContext( GLcontext *ctx ) /* Initialize tnl state. */ if (ctx->VertexProgram._MaintainTnlProgram) { -#if 0 - _tnl_ProgramCacheInit( ctx ); -#endif _tnl_install_pipeline( ctx, _tnl_vp_pipeline ); } else { _tnl_install_pipeline( ctx, _tnl_default_pipeline ); @@ -92,11 +89,6 @@ _tnl_DestroyContext( GLcontext *ctx ) _tnl_destroy_pipeline( ctx ); -#if 0 - if (ctx->VertexProgram._MaintainTnlProgram) - _tnl_ProgramCacheDestroy( ctx ); -#endif - FREE(tnl); ctx->swtnl_context = NULL; } diff --git a/src/mesa/tnl/t_context.h b/src/mesa/tnl/t_context.h index baf283ef0f..0a6ce04614 100644 --- a/src/mesa/tnl/t_context.h +++ b/src/mesa/tnl/t_context.h @@ -385,19 +385,6 @@ struct tnl_clipspace }; -struct tnl_cache_item { - GLuint hash; - void *key; - void *data; - struct tnl_cache_item *next; -}; - -struct tnl_cache { - struct tnl_cache_item **items; - GLuint size, n_items; -}; - - struct tnl_device_driver { /*** @@ -549,10 +536,6 @@ typedef struct GLubyte *block[VERT_ATTRIB_MAX]; GLuint nr_blocks; - /* Cache of fixed-function-replacing vertex programs: - */ - struct tnl_cache *vp_cache; - } TNLcontext; diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index 215d6653a3..249bccb443 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -31,1584 +31,13 @@ #include "glheader.h" -#include "macros.h" -#include "enums.h" #include "main/ffvertex_prog.h" -#include "shader/program.h" -#include "shader/prog_instruction.h" -#include "shader/prog_parameter.h" -#include "shader/prog_print.h" -#include "shader/prog_statevars.h" -#include "t_context.h" /* NOTE: very light dependency on this */ #include "t_vp_build.h" -#if 0 -struct state_key { - unsigned light_global_enabled:1; - unsigned light_local_viewer:1; - unsigned light_twoside:1; - unsigned light_color_material:1; - unsigned light_color_material_mask:12; - unsigned light_material_mask:12; - - unsigned normalize:1; - unsigned rescale_normals:1; - unsigned fog_source_is_depth:1; - unsigned tnl_do_vertex_fog:1; - unsigned separate_specular:1; - unsigned fog_mode:2; - unsigned point_attenuated:1; - unsigned texture_enabled_global:1; - unsigned fragprog_inputs_read:12; - - struct { - unsigned light_enabled:1; - unsigned light_eyepos3_is_zero:1; - unsigned light_spotcutoff_is_180:1; - unsigned light_attenuated:1; - unsigned texunit_really_enabled:1; - unsigned texmat_enabled:1; - unsigned texgen_enabled:4; - unsigned texgen_mode0:4; - unsigned texgen_mode1:4; - unsigned texgen_mode2:4; - unsigned texgen_mode3:4; - } unit[8]; -}; - - - -#define FOG_NONE 0 -#define FOG_LINEAR 1 -#define FOG_EXP 2 -#define FOG_EXP2 3 - -static GLuint translate_fog_mode( GLenum mode ) -{ - switch (mode) { - case GL_LINEAR: return FOG_LINEAR; - case GL_EXP: return FOG_EXP; - case GL_EXP2: return FOG_EXP2; - default: return FOG_NONE; - } -} - -#define TXG_NONE 0 -#define TXG_OBJ_LINEAR 1 -#define TXG_EYE_LINEAR 2 -#define TXG_SPHERE_MAP 3 -#define TXG_REFLECTION_MAP 4 -#define TXG_NORMAL_MAP 5 - -static GLuint translate_texgen( GLboolean enabled, GLenum mode ) -{ - if (!enabled) - return TXG_NONE; - - switch (mode) { - case GL_OBJECT_LINEAR: return TXG_OBJ_LINEAR; - case GL_EYE_LINEAR: return TXG_EYE_LINEAR; - case GL_SPHERE_MAP: return TXG_SPHERE_MAP; - case GL_REFLECTION_MAP_NV: return TXG_REFLECTION_MAP; - case GL_NORMAL_MAP_NV: return TXG_NORMAL_MAP; - default: return TXG_NONE; - } -} - /** - * Returns bitmask of flags indicating which materials are set per-vertex - * in the current VB. - * XXX get these from the VBO... + * XXX This should go away someday, but still referenced by some drivers... */ -static GLbitfield -tnl_get_per_vertex_materials(GLcontext *ctx) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *VB = &tnl->vb; - GLuint i; - GLbitfield mask = 0x0; - - for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) - if (VB->AttribPtr[i] && VB->AttribPtr[i]->stride) - mask |= 1 << (i - _TNL_FIRST_MAT); - - return mask; -} - -/** - * Should fog be computed per-vertex? - */ -static GLbitfield -tnl_get_per_vertex_fog(GLcontext *ctx) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - return tnl->_DoVertexFog; -} - - -static struct state_key *make_state_key( GLcontext *ctx ) -{ - const struct gl_fragment_program *fp; - struct state_key *key = CALLOC_STRUCT(state_key); - GLuint i; - - fp = ctx->FragmentProgram._Current; - - /* This now relies on texenvprogram.c being active: - */ - assert(fp); - - key->fragprog_inputs_read = fp->Base.InputsRead; - - if (ctx->RenderMode == GL_FEEDBACK) { - /* make sure the vertprog emits color and tex0 */ - key->fragprog_inputs_read |= (FRAG_BIT_COL0 | FRAG_BIT_TEX0); - } - - key->separate_specular = (ctx->Light.Model.ColorControl == - GL_SEPARATE_SPECULAR_COLOR); - - if (ctx->Light.Enabled) { - key->light_global_enabled = 1; - - if (ctx->Light.Model.LocalViewer) - key->light_local_viewer = 1; - - if (ctx->Light.Model.TwoSide) - key->light_twoside = 1; - - if (ctx->Light.ColorMaterialEnabled) { - key->light_color_material = 1; - key->light_color_material_mask = ctx->Light.ColorMaterialBitmask; - } - - key->light_material_mask = tnl_get_per_vertex_materials(ctx); - - for (i = 0; i < MAX_LIGHTS; i++) { - struct gl_light *light = &ctx->Light.Light[i]; - - if (light->Enabled) { - key->unit[i].light_enabled = 1; - - if (light->EyePosition[3] == 0.0) - key->unit[i].light_eyepos3_is_zero = 1; - - if (light->SpotCutoff == 180.0) - key->unit[i].light_spotcutoff_is_180 = 1; - - if (light->ConstantAttenuation != 1.0 || - light->LinearAttenuation != 0.0 || - light->QuadraticAttenuation != 0.0) - key->unit[i].light_attenuated = 1; - } - } - } - - if (ctx->Transform.Normalize) - key->normalize = 1; - - if (ctx->Transform.RescaleNormals) - key->rescale_normals = 1; - - key->fog_mode = translate_fog_mode(fp->FogOption); - - if (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT) - key->fog_source_is_depth = 1; - - key->tnl_do_vertex_fog = tnl_get_per_vertex_fog(ctx); - - if (ctx->Point._Attenuated) - key->point_attenuated = 1; - - if (ctx->Texture._TexGenEnabled || - ctx->Texture._TexMatEnabled || - ctx->Texture._EnabledUnits) - key->texture_enabled_global = 1; - - for (i = 0; i < MAX_TEXTURE_UNITS; i++) { - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i]; - - if (texUnit->_ReallyEnabled) - key->unit[i].texunit_really_enabled = 1; - - if (ctx->Texture._TexMatEnabled & ENABLE_TEXMAT(i)) - key->unit[i].texmat_enabled = 1; - - if (texUnit->TexGenEnabled) { - key->unit[i].texgen_enabled = 1; - - key->unit[i].texgen_mode0 = - translate_texgen( texUnit->TexGenEnabled & (1<<0), - texUnit->GenModeS ); - key->unit[i].texgen_mode1 = - translate_texgen( texUnit->TexGenEnabled & (1<<1), - texUnit->GenModeT ); - key->unit[i].texgen_mode2 = - translate_texgen( texUnit->TexGenEnabled & (1<<2), - texUnit->GenModeR ); - key->unit[i].texgen_mode3 = - translate_texgen( texUnit->TexGenEnabled & (1<<3), - texUnit->GenModeQ ); - } - } - - return key; -} - - - -/* Very useful debugging tool - produces annotated listing of - * generated program with line/function references for each - * instruction back into this file: - */ -#define DISASSEM (MESA_VERBOSE&VERBOSE_DISASSEM) - -/* Should be tunable by the driver - do we want to do matrix - * multiplications with DP4's or with MUL/MAD's? SSE works better - * with the latter, drivers may differ. - */ -#define PREFER_DP4 0 - -#define MAX_INSN 256 - -/* Use uregs to represent registers internally, translate to Mesa's - * expected formats on emit. - * - * NOTE: These are passed by value extensively in this file rather - * than as usual by pointer reference. If this disturbs you, try - * remembering they are just 32bits in size. - * - * GCC is smart enough to deal with these dword-sized structures in - * much the same way as if I had defined them as dwords and was using - * macros to access and set the fields. This is much nicer and easier - * to evolve. - */ -struct ureg { - GLuint file:4; - GLint idx:8; /* relative addressing may be negative */ - GLuint negate:1; - GLuint swz:12; - GLuint pad:7; -}; - - -struct tnl_program { - const struct state_key *state; - struct gl_vertex_program *program; - - GLuint temp_in_use; - GLuint temp_reserved; - - struct ureg eye_position; - struct ureg eye_position_normalized; - struct ureg eye_normal; - struct ureg identity; - - GLuint materials; - GLuint color_materials; -}; - - -static const struct ureg undef = { - PROGRAM_UNDEFINED, - ~0, - 0, - 0, - 0 -}; - -/* Local shorthand: - */ -#define X SWIZZLE_X -#define Y SWIZZLE_Y -#define Z SWIZZLE_Z -#define W SWIZZLE_W - - -/* Construct a ureg: - */ -static struct ureg make_ureg(GLuint file, GLint idx) -{ - struct ureg reg; - reg.file = file; - reg.idx = idx; - reg.negate = 0; - reg.swz = SWIZZLE_NOOP; - reg.pad = 0; - return reg; -} - - - -static struct ureg negate( struct ureg reg ) -{ - reg.negate ^= 1; - return reg; -} - - -static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w ) -{ - reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x), - GET_SWZ(reg.swz, y), - GET_SWZ(reg.swz, z), - GET_SWZ(reg.swz, w)); - - return reg; -} - -static struct ureg swizzle1( struct ureg reg, int x ) -{ - return swizzle(reg, x, x, x, x); -} - -static struct ureg get_temp( struct tnl_program *p ) -{ - int bit = _mesa_ffs( ~p->temp_in_use ); - if (!bit) { - _mesa_problem(NULL, "%s: out of temporaries\n", __FILE__); - _mesa_exit(1); - } - - if ((GLuint) bit > p->program->Base.NumTemporaries) - p->program->Base.NumTemporaries = bit; - - p->temp_in_use |= 1<<(bit-1); - return make_ureg(PROGRAM_TEMPORARY, bit-1); -} - -static struct ureg reserve_temp( struct tnl_program *p ) -{ - struct ureg temp = get_temp( p ); - p->temp_reserved |= 1<temp_in_use &= ~(1<temp_in_use |= p->temp_reserved; /* can't release reserved temps */ - } -} - -static void release_temps( struct tnl_program *p ) -{ - p->temp_in_use = p->temp_reserved; -} - - - -static struct ureg register_input( struct tnl_program *p, GLuint input ) -{ - p->program->Base.InputsRead |= (1<program->Base.OutputsWritten |= (1<program->Base.Parameters, values, 4, - &swizzle ); - ASSERT(swizzle == SWIZZLE_NOOP); - return make_ureg(PROGRAM_STATE_VAR, idx); -} - -#define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1) -#define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0) -#define register_const2f(p, s0, s1) register_const4f(p, s0, s1, 0, 1) -#define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1) - -static GLboolean is_undef( struct ureg reg ) -{ - return reg.file == PROGRAM_UNDEFINED; -} - -static struct ureg get_identity_param( struct tnl_program *p ) -{ - if (is_undef(p->identity)) - p->identity = register_const4f(p, 0,0,0,1); - - return p->identity; -} - -static struct ureg register_param5(struct tnl_program *p, - GLint s0, - GLint s1, - GLint s2, - GLint s3, - GLint s4) -{ - gl_state_index tokens[STATE_LENGTH]; - GLint idx; - tokens[0] = s0; - tokens[1] = s1; - tokens[2] = s2; - tokens[3] = s3; - tokens[4] = s4; - idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens ); - return make_ureg(PROGRAM_STATE_VAR, idx); -} - - -#define register_param1(p,s0) register_param5(p,s0,0,0,0,0) -#define register_param2(p,s0,s1) register_param5(p,s0,s1,0,0,0) -#define register_param3(p,s0,s1,s2) register_param5(p,s0,s1,s2,0,0) -#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0) - - -static void register_matrix_param5( struct tnl_program *p, - GLint s0, /* modelview, projection, etc */ - GLint s1, /* texture matrix number */ - GLint s2, /* first row */ - GLint s3, /* last row */ - GLint s4, /* inverse, transpose, etc */ - struct ureg *matrix ) -{ - GLint i; - - /* This is a bit sad as the support is there to pull the whole - * matrix out in one go: - */ - for (i = 0; i <= s3 - s2; i++) - matrix[i] = register_param5( p, s0, s1, i, i, s4 ); -} - - -static void emit_arg( struct prog_src_register *src, - struct ureg reg ) -{ - src->File = reg.file; - src->Index = reg.idx; - src->Swizzle = reg.swz; - src->NegateBase = reg.negate ? NEGATE_XYZW : 0; - src->Abs = 0; - src->NegateAbs = 0; - src->RelAddr = 0; -} - -static void emit_dst( struct prog_dst_register *dst, - struct ureg reg, GLuint mask ) -{ - dst->File = reg.file; - dst->Index = reg.idx; - /* allow zero as a shorthand for xyzw */ - dst->WriteMask = mask ? mask : WRITEMASK_XYZW; - dst->CondMask = COND_TR; /* always pass cond test */ - dst->CondSwizzle = SWIZZLE_NOOP; - dst->CondSrc = 0; - dst->pad = 0; -} - -static void debug_insn( struct prog_instruction *inst, const char *fn, - GLuint line ) -{ - if (DISASSEM) { - static const char *last_fn; - - if (fn != last_fn) { - last_fn = fn; - _mesa_printf("%s:\n", fn); - } - - _mesa_printf("%d:\t", line); - _mesa_print_instruction(inst); - } -} - - -static void emit_op3fn(struct tnl_program *p, - enum prog_opcode op, - struct ureg dest, - GLuint mask, - struct ureg src0, - struct ureg src1, - struct ureg src2, - const char *fn, - GLuint line) -{ - GLuint nr = p->program->Base.NumInstructions++; - struct prog_instruction *inst = &p->program->Base.Instructions[nr]; - - if (p->program->Base.NumInstructions > MAX_INSN) { - _mesa_problem(0, "Out of instructions in emit_op3fn\n"); - return; - } - - inst->Opcode = (enum prog_opcode) op; - inst->StringPos = 0; - inst->Data = 0; - - emit_arg( &inst->SrcReg[0], src0 ); - emit_arg( &inst->SrcReg[1], src1 ); - emit_arg( &inst->SrcReg[2], src2 ); - - emit_dst( &inst->DstReg, dest, mask ); - - debug_insn(inst, fn, line); -} - - -#define emit_op3(p, op, dst, mask, src0, src1, src2) \ - emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__) - -#define emit_op2(p, op, dst, mask, src0, src1) \ - emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__) - -#define emit_op1(p, op, dst, mask, src0) \ - emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__) - - -static struct ureg make_temp( struct tnl_program *p, struct ureg reg ) -{ - if (reg.file == PROGRAM_TEMPORARY && - !(p->temp_reserved & (1<eye_position)) { - struct ureg pos = register_input( p, VERT_ATTRIB_POS ); - struct ureg modelview[4]; - - p->eye_position = reserve_temp(p); - - if (PREFER_DP4) { - register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3, - 0, modelview ); - - emit_matrix_transform_vec4(p, p->eye_position, modelview, pos); - } - else { - register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3, - STATE_MATRIX_TRANSPOSE, modelview ); - - emit_transpose_matrix_transform_vec4(p, p->eye_position, modelview, pos); - } - } - - return p->eye_position; -} - - -static struct ureg get_eye_position_normalized( struct tnl_program *p ) -{ - if (is_undef(p->eye_position_normalized)) { - struct ureg eye = get_eye_position(p); - p->eye_position_normalized = reserve_temp(p); - emit_normalize_vec3(p, p->eye_position_normalized, eye); - } - - return p->eye_position_normalized; -} - - -static struct ureg get_eye_normal( struct tnl_program *p ) -{ - if (is_undef(p->eye_normal)) { - struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL ); - struct ureg mvinv[3]; - - register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 2, - STATE_MATRIX_INVTRANS, mvinv ); - - p->eye_normal = reserve_temp(p); - - /* Transform to eye space: - */ - emit_matrix_transform_vec3( p, p->eye_normal, mvinv, normal ); - - /* Normalize/Rescale: - */ - if (p->state->normalize) { - emit_normalize_vec3( p, p->eye_normal, p->eye_normal ); - } - else if (p->state->rescale_normals) { - struct ureg rescale = register_param2(p, STATE_INTERNAL, - STATE_NORMAL_SCALE); - - emit_op2( p, OPCODE_MUL, p->eye_normal, 0, p->eye_normal, - swizzle1(rescale, X)); - } - } - - return p->eye_normal; -} - - - -static void build_hpos( struct tnl_program *p ) -{ - struct ureg pos = register_input( p, VERT_ATTRIB_POS ); - struct ureg hpos = register_output( p, VERT_RESULT_HPOS ); - struct ureg mvp[4]; - - if (PREFER_DP4) { - register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, - 0, mvp ); - emit_matrix_transform_vec4( p, hpos, mvp, pos ); - } - else { - register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, - STATE_MATRIX_TRANSPOSE, mvp ); - emit_transpose_matrix_transform_vec4( p, hpos, mvp, pos ); - } -} - - -static GLuint material_attrib( GLuint side, GLuint property ) -{ - return ((property - STATE_AMBIENT) * 2 + - side); -} - -/* Get a bitmask of which material values vary on a per-vertex basis. - */ -static void set_material_flags( struct tnl_program *p ) -{ - p->color_materials = 0; - p->materials = 0; - - if (p->state->light_color_material) { - p->materials = - p->color_materials = p->state->light_color_material_mask; - } - - p->materials |= p->state->light_material_mask; -} - - -static struct ureg get_material( struct tnl_program *p, GLuint side, - GLuint property ) -{ - GLuint attrib = material_attrib(side, property); - - if (p->color_materials & (1<materials & (1<materials & SCENE_COLOR_BITS(side)) { - struct ureg lm_ambient = register_param1(p, STATE_LIGHTMODEL_AMBIENT); - struct ureg material_emission = get_material(p, side, STATE_EMISSION); - struct ureg material_ambient = get_material(p, side, STATE_AMBIENT); - struct ureg material_diffuse = get_material(p, side, STATE_DIFFUSE); - struct ureg tmp = make_temp(p, material_diffuse); - emit_op3(p, OPCODE_MAD, tmp, WRITEMASK_XYZ, lm_ambient, - material_ambient, material_emission); - return tmp; - } - else - return register_param2( p, STATE_LIGHTMODEL_SCENECOLOR, side ); -} - - -static struct ureg get_lightprod( struct tnl_program *p, GLuint light, - GLuint side, GLuint property ) -{ - GLuint attrib = material_attrib(side, property); - if (p->materials & (1<state->unit[i].light_spotcutoff_is_180) { - struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL, - STATE_SPOT_DIR_NORMALIZED, i); - struct ureg spot = get_temp(p); - struct ureg slt = get_temp(p); - - emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot_dir_norm); - emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot); - emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W)); - emit_op2(p, OPCODE_MUL, att, 0, slt, spot); - - release_temp(p, spot); - release_temp(p, slt); - } - - /* Calculate distance attenuation: - */ - if (p->state->unit[i].light_attenuated) { - - /* 1/d,d,d,1/d */ - emit_op1(p, OPCODE_RCP, dist, WRITEMASK_YZ, dist); - /* 1,d,d*d,1/d */ - emit_op2(p, OPCODE_MUL, dist, WRITEMASK_XZ, dist, swizzle1(dist,Y)); - /* 1/dist-atten */ - emit_op2(p, OPCODE_DP3, dist, 0, attenuation, dist); - - if (!p->state->unit[i].light_spotcutoff_is_180) { - /* dist-atten */ - emit_op1(p, OPCODE_RCP, dist, 0, dist); - /* spot-atten * dist-atten */ - emit_op2(p, OPCODE_MUL, att, 0, dist, att); - } else { - /* dist-atten */ - emit_op1(p, OPCODE_RCP, att, 0, dist); - } - } - - return att; -} - - - - - -/* Need to add some addtional parameters to allow lighting in object - * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye - * space lighting. - */ -static void build_lighting( struct tnl_program *p ) -{ - const GLboolean twoside = p->state->light_twoside; - const GLboolean separate = p->state->separate_specular; - GLuint nr_lights = 0, count = 0; - struct ureg normal = get_eye_normal(p); - struct ureg lit = get_temp(p); - struct ureg dots = get_temp(p); - struct ureg _col0 = undef, _col1 = undef; - struct ureg _bfc0 = undef, _bfc1 = undef; - GLuint i; - - for (i = 0; i < MAX_LIGHTS; i++) - if (p->state->unit[i].light_enabled) - nr_lights++; - - set_material_flags(p); - - { - struct ureg shininess = get_material(p, 0, STATE_SHININESS); - emit_op1(p, OPCODE_MOV, dots, WRITEMASK_W, swizzle1(shininess,X)); - release_temp(p, shininess); - - _col0 = make_temp(p, get_scenecolor(p, 0)); - if (separate) - _col1 = make_temp(p, get_identity_param(p)); - else - _col1 = _col0; - - } - - if (twoside) { - struct ureg shininess = get_material(p, 1, STATE_SHININESS); - emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z, - negate(swizzle1(shininess,X))); - release_temp(p, shininess); - - _bfc0 = make_temp(p, get_scenecolor(p, 1)); - if (separate) - _bfc1 = make_temp(p, get_identity_param(p)); - else - _bfc1 = _bfc0; - } - - - /* If no lights, still need to emit the scenecolor. - */ - { - struct ureg res0 = register_output( p, VERT_RESULT_COL0 ); - emit_op1(p, OPCODE_MOV, res0, 0, _col0); - } - - if (separate) { - struct ureg res1 = register_output( p, VERT_RESULT_COL1 ); - emit_op1(p, OPCODE_MOV, res1, 0, _col1); - } - - if (twoside) { - struct ureg res0 = register_output( p, VERT_RESULT_BFC0 ); - emit_op1(p, OPCODE_MOV, res0, 0, _bfc0); - } - - if (twoside && separate) { - struct ureg res1 = register_output( p, VERT_RESULT_BFC1 ); - emit_op1(p, OPCODE_MOV, res1, 0, _bfc1); - } - - if (nr_lights == 0) { - release_temps(p); - return; - } - - - for (i = 0; i < MAX_LIGHTS; i++) { - if (p->state->unit[i].light_enabled) { - struct ureg half = undef; - struct ureg att = undef, VPpli = undef; - - count++; - - if (p->state->unit[i].light_eyepos3_is_zero) { - /* Can used precomputed constants in this case. - * Attenuation never applies to infinite lights. - */ - VPpli = register_param3(p, STATE_LIGHT, i, - STATE_POSITION_NORMALIZED); - if (p->state->light_local_viewer) { - struct ureg eye_hat = get_eye_position_normalized(p); - half = get_temp(p); - emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); - emit_normalize_vec3(p, half, half); - } else { - half = register_param3(p, STATE_LIGHT, i, STATE_HALF_VECTOR); - } - } - else { - struct ureg Ppli = register_param3(p, STATE_LIGHT, i, - STATE_POSITION); - struct ureg V = get_eye_position(p); - struct ureg dist = get_temp(p); - - VPpli = get_temp(p); - half = get_temp(p); - - /* Calulate VPpli vector - */ - emit_op2(p, OPCODE_SUB, VPpli, 0, Ppli, V); - - /* Normalize VPpli. The dist value also used in - * attenuation below. - */ - emit_op2(p, OPCODE_DP3, dist, 0, VPpli, VPpli); - emit_op1(p, OPCODE_RSQ, dist, 0, dist); - emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist); - - - /* Calculate attenuation: - */ - if (!p->state->unit[i].light_spotcutoff_is_180 || - p->state->unit[i].light_attenuated) { - att = calculate_light_attenuation(p, i, VPpli, dist); - } - - - /* Calculate viewer direction, or use infinite viewer: - */ - if (p->state->light_local_viewer) { - struct ureg eye_hat = get_eye_position_normalized(p); - emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); - } - else { - struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z); - emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir); - } - - emit_normalize_vec3(p, half, half); - - release_temp(p, dist); - } - - /* Calculate dot products: - */ - emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); - emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half); - - - /* Front face lighting: - */ - { - struct ureg ambient = get_lightprod(p, i, 0, STATE_AMBIENT); - struct ureg diffuse = get_lightprod(p, i, 0, STATE_DIFFUSE); - struct ureg specular = get_lightprod(p, i, 0, STATE_SPECULAR); - struct ureg res0, res1; - GLuint mask0, mask1; - - emit_op1(p, OPCODE_LIT, lit, 0, dots); - - if (!is_undef(att)) - emit_op2(p, OPCODE_MUL, lit, 0, lit, att); - - - if (count == nr_lights) { - if (separate) { - mask0 = WRITEMASK_XYZ; - mask1 = WRITEMASK_XYZ; - res0 = register_output( p, VERT_RESULT_COL0 ); - res1 = register_output( p, VERT_RESULT_COL1 ); - } - else { - mask0 = 0; - mask1 = WRITEMASK_XYZ; - res0 = _col0; - res1 = register_output( p, VERT_RESULT_COL0 ); - } - } else { - mask0 = 0; - mask1 = 0; - res0 = _col0; - res1 = _col1; - } - - emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0); - emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0); - emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1); - - release_temp(p, ambient); - release_temp(p, diffuse); - release_temp(p, specular); - } - - /* Back face lighting: - */ - if (twoside) { - struct ureg ambient = get_lightprod(p, i, 1, STATE_AMBIENT); - struct ureg diffuse = get_lightprod(p, i, 1, STATE_DIFFUSE); - struct ureg specular = get_lightprod(p, i, 1, STATE_SPECULAR); - struct ureg res0, res1; - GLuint mask0, mask1; - - emit_op1(p, OPCODE_LIT, lit, 0, negate(swizzle(dots,X,Y,W,Z))); - - if (!is_undef(att)) - emit_op2(p, OPCODE_MUL, lit, 0, lit, att); - - if (count == nr_lights) { - if (separate) { - mask0 = WRITEMASK_XYZ; - mask1 = WRITEMASK_XYZ; - res0 = register_output( p, VERT_RESULT_BFC0 ); - res1 = register_output( p, VERT_RESULT_BFC1 ); - } - else { - mask0 = 0; - mask1 = WRITEMASK_XYZ; - res0 = _bfc0; - res1 = register_output( p, VERT_RESULT_BFC0 ); - } - } else { - res0 = _bfc0; - res1 = _bfc1; - mask0 = 0; - mask1 = 0; - } - - emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0); - emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0); - emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1); - - release_temp(p, ambient); - release_temp(p, diffuse); - release_temp(p, specular); - } - - release_temp(p, half); - release_temp(p, VPpli); - release_temp(p, att); - } - } - - release_temps( p ); -} - - -static void build_fog( struct tnl_program *p ) -{ - struct ureg fog = register_output(p, VERT_RESULT_FOGC); - struct ureg input; - - if (p->state->fog_source_is_depth) { - input = swizzle1(get_eye_position(p), Z); - } - else { - input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X); - } - - if (p->state->fog_mode && p->state->tnl_do_vertex_fog) { - struct ureg params = register_param2(p, STATE_INTERNAL, - STATE_FOG_PARAMS_OPTIMIZED); - struct ureg tmp = get_temp(p); - GLboolean useabs = (p->state->fog_mode != FOG_EXP2); - - if (useabs) { - emit_op1(p, OPCODE_ABS, tmp, 0, input); - } - - switch (p->state->fog_mode) { - case FOG_LINEAR: { - struct ureg id = get_identity_param(p); - emit_op3(p, OPCODE_MAD, tmp, 0, useabs ? tmp : input, - swizzle1(params,X), swizzle1(params,Y)); - emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */ - emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W)); - break; - } - case FOG_EXP: - emit_op2(p, OPCODE_MUL, tmp, 0, useabs ? tmp : input, - swizzle1(params,Z)); - emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp)); - break; - case FOG_EXP2: - emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W)); - emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp); - emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp)); - break; - } - - release_temp(p, tmp); - } - else { - /* results = incoming fog coords (compute fog per-fragment later) - * - * KW: Is it really necessary to do anything in this case? - * BP: Yes, we always need to compute the absolute value, unless - * we want to push that down into the fragment program... - */ - GLboolean useabs = GL_TRUE; - emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, WRITEMASK_X, input); - } -} - -static void build_reflect_texgen( struct tnl_program *p, - struct ureg dest, - GLuint writemask ) -{ - struct ureg normal = get_eye_normal(p); - struct ureg eye_hat = get_eye_position_normalized(p); - struct ureg tmp = get_temp(p); - - /* n.u */ - emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); - /* 2n.u */ - emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); - /* (-2n.u)n + u */ - emit_op3(p, OPCODE_MAD, dest, writemask, negate(tmp), normal, eye_hat); - - release_temp(p, tmp); -} - -static void build_sphere_texgen( struct tnl_program *p, - struct ureg dest, - GLuint writemask ) -{ - struct ureg normal = get_eye_normal(p); - struct ureg eye_hat = get_eye_position_normalized(p); - struct ureg tmp = get_temp(p); - struct ureg half = register_scalar_const(p, .5); - struct ureg r = get_temp(p); - struct ureg inv_m = get_temp(p); - struct ureg id = get_identity_param(p); - - /* Could share the above calculations, but it would be - * a fairly odd state for someone to set (both sphere and - * reflection active for different texture coordinate - * components. Of course - if two texture units enable - * reflect and/or sphere, things start to tilt in favour - * of seperating this out: - */ - - /* n.u */ - emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); - /* 2n.u */ - emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); - /* (-2n.u)n + u */ - emit_op3(p, OPCODE_MAD, r, 0, negate(tmp), normal, eye_hat); - /* r + 0,0,1 */ - emit_op2(p, OPCODE_ADD, tmp, 0, r, swizzle(id,X,Y,W,Z)); - /* rx^2 + ry^2 + (rz+1)^2 */ - emit_op2(p, OPCODE_DP3, tmp, 0, tmp, tmp); - /* 2/m */ - emit_op1(p, OPCODE_RSQ, tmp, 0, tmp); - /* 1/m */ - emit_op2(p, OPCODE_MUL, inv_m, 0, tmp, half); - /* r/m + 1/2 */ - emit_op3(p, OPCODE_MAD, dest, writemask, r, inv_m, half); - - release_temp(p, tmp); - release_temp(p, r); - release_temp(p, inv_m); -} - - -static void build_texture_transform( struct tnl_program *p ) -{ - GLuint i, j; - - for (i = 0; i < MAX_TEXTURE_UNITS; i++) { - - if (!(p->state->fragprog_inputs_read & FRAG_BIT_TEX(i))) - continue; - - if (p->state->unit[i].texgen_enabled || - p->state->unit[i].texmat_enabled) { - - GLuint texmat_enabled = p->state->unit[i].texmat_enabled; - struct ureg out = register_output(p, VERT_RESULT_TEX0 + i); - struct ureg out_texgen = undef; - - if (p->state->unit[i].texgen_enabled) { - GLuint copy_mask = 0; - GLuint sphere_mask = 0; - GLuint reflect_mask = 0; - GLuint normal_mask = 0; - GLuint modes[4]; - - if (texmat_enabled) - out_texgen = get_temp(p); - else - out_texgen = out; - - modes[0] = p->state->unit[i].texgen_mode0; - modes[1] = p->state->unit[i].texgen_mode1; - modes[2] = p->state->unit[i].texgen_mode2; - modes[3] = p->state->unit[i].texgen_mode3; - - for (j = 0; j < 4; j++) { - switch (modes[j]) { - case TXG_OBJ_LINEAR: { - struct ureg obj = register_input(p, VERT_ATTRIB_POS); - struct ureg plane = - register_param3(p, STATE_TEXGEN, i, - STATE_TEXGEN_OBJECT_S + j); - - emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, - obj, plane ); - break; - } - case TXG_EYE_LINEAR: { - struct ureg eye = get_eye_position(p); - struct ureg plane = - register_param3(p, STATE_TEXGEN, i, - STATE_TEXGEN_EYE_S + j); - - emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, - eye, plane ); - break; - } - case TXG_SPHERE_MAP: - sphere_mask |= WRITEMASK_X << j; - break; - case TXG_REFLECTION_MAP: - reflect_mask |= WRITEMASK_X << j; - break; - case TXG_NORMAL_MAP: - normal_mask |= WRITEMASK_X << j; - break; - case TXG_NONE: - copy_mask |= WRITEMASK_X << j; - } - - } - - - if (sphere_mask) { - build_sphere_texgen(p, out_texgen, sphere_mask); - } - - if (reflect_mask) { - build_reflect_texgen(p, out_texgen, reflect_mask); - } - - if (normal_mask) { - struct ureg normal = get_eye_normal(p); - emit_op1(p, OPCODE_MOV, out_texgen, normal_mask, normal ); - } - - if (copy_mask) { - struct ureg in = register_input(p, VERT_ATTRIB_TEX0+i); - emit_op1(p, OPCODE_MOV, out_texgen, copy_mask, in ); - } - } - - if (texmat_enabled) { - struct ureg texmat[4]; - struct ureg in = (!is_undef(out_texgen) ? - out_texgen : - register_input(p, VERT_ATTRIB_TEX0+i)); - if (PREFER_DP4) { - register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3, - 0, texmat ); - emit_matrix_transform_vec4( p, out, texmat, in ); - } - else { - register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3, - STATE_MATRIX_TRANSPOSE, texmat ); - emit_transpose_matrix_transform_vec4( p, out, texmat, in ); - } - } - - release_temps(p); - } - else { - emit_passthrough(p, VERT_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i); - } - } -} - - -static void build_pointsize( struct tnl_program *p ) -{ - struct ureg eye = get_eye_position(p); - struct ureg state_size = register_param1(p, STATE_POINT_SIZE); - struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION); - struct ureg out = register_output(p, VERT_RESULT_PSIZ); - struct ureg ut = get_temp(p); - - /* dist = |eyez| */ - emit_op1(p, OPCODE_ABS, ut, WRITEMASK_Y, swizzle1(eye, Z)); - /* p1 + dist * (p2 + dist * p3); */ - emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y), - swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y)); - emit_op3(p, OPCODE_MAD, ut, WRITEMASK_X, swizzle1(ut, Y), - ut, swizzle1(state_attenuation, X)); - - /* 1 / sqrt(factor) */ - emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut ); - -#if 1 - /* out = pointSize / sqrt(factor) */ - emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size); -#else - /* not sure, might make sense to do clamping here, - but it's not done in t_vb_points neither */ - emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size); - emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y)); - emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z)); -#endif - - release_temp(p, ut); -} - -/** - * Emit constant point size. - */ -static void constant_pointsize( struct tnl_program *p ) -{ - struct ureg state_size = register_param1(p, STATE_POINT_SIZE); - struct ureg out = register_output(p, VERT_RESULT_PSIZ); - emit_op1(p, OPCODE_MOV, out, WRITEMASK_X, state_size); -} - -static void build_tnl_program( struct tnl_program *p ) -{ /* Emit the program, starting with modelviewproject: - */ - build_hpos(p); - - /* Lighting calculations: - */ - if (p->state->fragprog_inputs_read & (FRAG_BIT_COL0|FRAG_BIT_COL1)) { - if (p->state->light_global_enabled) - build_lighting(p); - else { - if (p->state->fragprog_inputs_read & FRAG_BIT_COL0) - emit_passthrough(p, VERT_ATTRIB_COLOR0, VERT_RESULT_COL0); - - if (p->state->fragprog_inputs_read & FRAG_BIT_COL1) - emit_passthrough(p, VERT_ATTRIB_COLOR1, VERT_RESULT_COL1); - } - } - - if ((p->state->fragprog_inputs_read & FRAG_BIT_FOGC) || - p->state->fog_mode != FOG_NONE) - build_fog(p); - - if (p->state->fragprog_inputs_read & FRAG_BITS_TEX_ANY) - build_texture_transform(p); - - if (p->state->point_attenuated) - build_pointsize(p); -#if 0 - else - constant_pointsize(p); -#endif - - /* Finish up: - */ - emit_op1(p, OPCODE_END, undef, 0, undef); - - /* Disassemble: - */ - if (DISASSEM) { - _mesa_printf ("\n"); - } -} - - -static void -create_new_program( const struct state_key *key, - struct gl_vertex_program *program, - GLuint max_temps) -{ - struct tnl_program p; - - _mesa_memset(&p, 0, sizeof(p)); - p.state = key; - p.program = program; - p.eye_position = undef; - p.eye_position_normalized = undef; - p.eye_normal = undef; - p.identity = undef; - p.temp_in_use = 0; - - if (max_temps >= sizeof(int) * 8) - p.temp_reserved = 0; - else - p.temp_reserved = ~((1<Base.Instructions = _mesa_alloc_instructions(MAX_INSN); - p.program->Base.String = NULL; - p.program->Base.NumInstructions = - p.program->Base.NumTemporaries = - p.program->Base.NumParameters = - p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0; - p.program->Base.Parameters = _mesa_new_parameter_list(); - p.program->Base.InputsRead = 0; - p.program->Base.OutputsWritten = 0; - - build_tnl_program( &p ); -} - -static void *search_cache( struct tnl_cache *cache, - GLuint hash, - const void *key, - GLuint keysize) -{ - struct tnl_cache_item *c; - - for (c = cache->items[hash % cache->size]; c; c = c->next) { - if (c->hash == hash && _mesa_memcmp(c->key, key, keysize) == 0) - return c->data; - } - - return NULL; -} - -static void rehash( struct tnl_cache *cache ) -{ - struct tnl_cache_item **items; - struct tnl_cache_item *c, *next; - GLuint size, i; - - size = cache->size * 3; - items = (struct tnl_cache_item**) _mesa_malloc(size * sizeof(*items)); - _mesa_memset(items, 0, size * sizeof(*items)); - - for (i = 0; i < cache->size; i++) - for (c = cache->items[i]; c; c = next) { - next = c->next; - c->next = items[c->hash % size]; - items[c->hash % size] = c; - } - - FREE(cache->items); - cache->items = items; - cache->size = size; -} - -static void cache_item( struct tnl_cache *cache, - GLuint hash, - void *key, - void *data ) -{ - struct tnl_cache_item *c = (struct tnl_cache_item*) _mesa_malloc(sizeof(*c)); - c->hash = hash; - c->key = key; - c->data = data; - - if (++cache->n_items > cache->size * 1.5) - rehash(cache); - - c->next = cache->items[hash % cache->size]; - cache->items[hash % cache->size] = c; -} - -static GLuint hash_key( struct state_key *key ) -{ - GLuint *ikey = (GLuint *)key; - GLuint hash = 0, i; - - /* I'm sure this can be improved on, but speed is important: - */ - for (i = 0; i < sizeof(*key)/sizeof(GLuint); i++) - hash ^= ikey[i]; - - return hash; -} - - -/** - * Return a vertex program which implements the current fixed-function - * transform/lighting/texgen operations. - * XXX move this into core mesa (main/) - */ -struct gl_vertex_program * -_mesa_get_fixed_func_vertex_program(GLcontext *ctx) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct gl_vertex_program *prog; - struct state_key *key; - GLuint hash; - - /* Grab all the relevent state and put it in a single structure: - */ - key = make_state_key(ctx); - hash = hash_key(key); - - /* Look for an already-prepared program for this state: - */ - prog = (struct gl_vertex_program *) - search_cache( tnl->vp_cache, hash, key, sizeof(*key) ); - - if (!prog) { - /* OK, we'll have to build a new one */ - if (0) - _mesa_printf("Build new TNL program\n"); - - prog = (struct gl_vertex_program *) - ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); - - create_new_program( key, prog, - ctx->Const.VertexProgram.MaxTemps ); - -#if 0 - if (ctx->Driver.ProgramStringNotify) - ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB, - &prog->Base ); -#endif - cache_item(tnl->vp_cache, hash, key, prog); - } - else { - /* use cached program */ - _mesa_free(key); - } - - return prog; -} -#endif - - void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ) { const struct gl_vertex_program *prev = ctx->VertexProgram._Current; @@ -1628,34 +57,3 @@ void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ) (struct gl_program *) ctx->VertexProgram._Current); } } - -#if 0 -void _tnl_ProgramCacheInit( GLcontext *ctx ) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - - tnl->vp_cache = (struct tnl_cache *) MALLOC(sizeof(*tnl->vp_cache)); - tnl->vp_cache->size = 17; - tnl->vp_cache->n_items = 0; - tnl->vp_cache->items = (struct tnl_cache_item**) - _mesa_calloc(tnl->vp_cache->size * sizeof(*tnl->vp_cache->items)); -} - -void _tnl_ProgramCacheDestroy( GLcontext *ctx ) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct tnl_cache_item *c, *next; - GLuint i; - - for (i = 0; i < tnl->vp_cache->size; i++) - for (c = tnl->vp_cache->items[i]; c; c = next) { - next = c->next; - FREE(c->key); - FREE(c->data); - FREE(c); - } - - FREE(tnl->vp_cache->items); - FREE(tnl->vp_cache); -} -#endif diff --git a/src/mesa/tnl/t_vp_build.h b/src/mesa/tnl/t_vp_build.h index adcd8f1662..efe12e41a4 100644 --- a/src/mesa/tnl/t_vp_build.h +++ b/src/mesa/tnl/t_vp_build.h @@ -37,16 +37,6 @@ _NEW_FOG | \ _NEW_POINT) -#if 0 -extern struct gl_vertex_program * -_mesa_get_fixed_func_vertex_program(GLcontext *ctx); -#endif - extern void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx ); -#if 0 -extern void _tnl_ProgramCacheInit( GLcontext *ctx ); -extern void _tnl_ProgramCacheDestroy( GLcontext *ctx ); -#endif - #endif -- cgit v1.2.3 From f4a5ea2ccb472958a4635c606e9510011bceaa3d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 12:45:32 -0600 Subject: Update texenvprogram.c code to use prog_cache.c routines. --- src/mesa/main/mtypes.h | 14 ----- src/mesa/main/texenvprogram.c | 135 +++--------------------------------------- src/mesa/main/texenvprogram.h | 11 +--- src/mesa/main/texstate.c | 4 -- 4 files changed, 9 insertions(+), 155 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index b435c29793..8e49431a8f 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -1557,17 +1557,6 @@ struct gl_texture_unit /*@}*/ }; -struct texenvprog_cache_item { - GLuint hash; - void *key; - struct gl_fragment_program *data; - struct texenvprog_cache_item *next; -}; - -struct texenvprog_cache { - struct texenvprog_cache_item **items; - GLuint size, n_items; -}; /** * Texture attribute group (GL_TEXTURE_BIT). @@ -1599,9 +1588,6 @@ struct gl_texture_attrib /** GL_EXT_shared_texture_palette */ GLboolean SharedPalette; struct gl_color_table Palette; - - /** Cached texenv fragment programs */ - struct texenvprog_cache env_fp_cache; }; diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index cf92503c34..efb3b35f6a 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -29,6 +29,7 @@ #include "macros.h" #include "enums.h" #include "shader/prog_parameter.h" +#include "shader/prog_cache.h" #include "shader/prog_instruction.h" #include "shader/prog_print.h" #include "shader/prog_statevars.h" @@ -1131,109 +1132,6 @@ create_new_program(GLcontext *ctx, struct state_key *key, } -static struct gl_fragment_program * -search_cache(const struct texenvprog_cache *cache, - GLuint hash, - const void *key, - GLuint keysize) -{ - struct texenvprog_cache_item *c; - - for (c = cache->items[hash % cache->size]; c; c = c->next) { - if (c->hash == hash && memcmp(c->key, key, keysize) == 0) - return (struct gl_fragment_program *) c->data; - } - - return NULL; -} - -static void rehash( struct texenvprog_cache *cache ) -{ - struct texenvprog_cache_item **items; - struct texenvprog_cache_item *c, *next; - GLuint size, i; - - size = cache->size * 3; - items = (struct texenvprog_cache_item**) _mesa_malloc(size * sizeof(*items)); - _mesa_memset(items, 0, size * sizeof(*items)); - - for (i = 0; i < cache->size; i++) - for (c = cache->items[i]; c; c = next) { - next = c->next; - c->next = items[c->hash % size]; - items[c->hash % size] = c; - } - - _mesa_free(cache->items); - cache->items = items; - cache->size = size; -} - -static void clear_cache( GLcontext *ctx, struct texenvprog_cache *cache ) -{ - struct texenvprog_cache_item *c, *next; - GLuint i; - - for (i = 0; i < cache->size; i++) { - for (c = cache->items[i]; c; c = next) { - next = c->next; - _mesa_free(c->key); - ctx->Driver.DeleteProgram(ctx, (struct gl_program *) c->data); - _mesa_free(c); - } - cache->items[i] = NULL; - } - - - cache->n_items = 0; -} - - -static void cache_item( GLcontext *ctx, - struct texenvprog_cache *cache, - GLuint hash, - const struct state_key *key, - void *data ) -{ - struct texenvprog_cache_item *c - = (struct texenvprog_cache_item *) MALLOC(sizeof(*c)); - c->hash = hash; - - c->key = _mesa_malloc(sizeof(*key)); - memcpy(c->key, key, sizeof(*key)); - - c->data = (struct gl_fragment_program *) data; - - if (cache->n_items > cache->size * 1.5) { - if (cache->size < 1000) - rehash(cache); - else - clear_cache(ctx, cache); - } - - cache->n_items++; - c->next = cache->items[hash % cache->size]; - cache->items[hash % cache->size] = c; -} - -static GLuint hash_key( const struct state_key *key ) -{ - GLuint *ikey = (GLuint *)key; - GLuint hash = 0, i; - - /* Make a slightly better attempt at a hash function: - */ - for (i = 0; i < sizeof(*key)/sizeof(*ikey); i++) - { - hash += ikey[i]; - hash += (hash << 10); - hash ^= (hash >> 6); - } - - return hash; -} - - /** * Return a fragment program which implements the current * fixed-function texture, fog and color-sum operations. @@ -1243,23 +1141,21 @@ _mesa_get_fixed_func_fragment_program(GLcontext *ctx) { struct gl_fragment_program *prog; struct state_key key; - GLuint hash; make_state_key(ctx, &key); - hash = hash_key(&key); - prog = search_cache(&ctx->Texture.env_fp_cache, hash, &key, sizeof(key)); + prog = (struct gl_fragment_program *) + _mesa_search_program_cache(ctx->FragmentProgram.Cache, + &key, sizeof(key)); if (!prog) { - if (0) - _mesa_printf("Building new texenv proggy for key %x\n", hash); - prog = (struct gl_fragment_program *) ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); create_new_program(ctx, &key, prog); - cache_item(ctx, &ctx->Texture.env_fp_cache, hash, &key, prog); + _mesa_program_cache_insert(ctx, ctx->FragmentProgram.Cache, + &key, sizeof(key), &prog->Base); } return prog; @@ -1297,20 +1193,3 @@ _mesa_UpdateTexEnvProgram( GLcontext *ctx ) (struct gl_program *) ctx->FragmentProgram._Current); } } - - -void _mesa_TexEnvProgramCacheInit( GLcontext *ctx ) -{ - ctx->Texture.env_fp_cache.size = 17; - ctx->Texture.env_fp_cache.n_items = 0; - ctx->Texture.env_fp_cache.items = (struct texenvprog_cache_item **) - _mesa_calloc(ctx->Texture.env_fp_cache.size * - sizeof(struct texenvprog_cache_item)); -} - - -void _mesa_TexEnvProgramCacheDestroy( GLcontext *ctx ) -{ - clear_cache(ctx, &ctx->Texture.env_fp_cache); - _mesa_free(ctx->Texture.env_fp_cache.items); -} diff --git a/src/mesa/main/texenvprogram.h b/src/mesa/main/texenvprogram.h index ac73910cb5..a7aa60cf37 100644 --- a/src/mesa/main/texenvprogram.h +++ b/src/mesa/main/texenvprogram.h @@ -1,13 +1,8 @@ -/** - * \file texenvprogram.h - * Texture state management. - */ - /* * Mesa 3-D graphics library - * Version: 5.1 + * Version: 7.1 * - * Copyright (C) 1999-2002 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -38,7 +33,5 @@ extern struct gl_fragment_program * _mesa_get_fixed_func_fragment_program(GLcontext *ctx); extern void _mesa_UpdateTexEnvProgram( GLcontext *ctx ); -extern void _mesa_TexEnvProgramCacheInit( GLcontext *ctx ); -extern void _mesa_TexEnvProgramCacheDestroy( GLcontext *ctx ); #endif diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c index c9f8a0656e..cb7da39b51 100644 --- a/src/mesa/main/texstate.c +++ b/src/mesa/main/texstate.c @@ -3210,8 +3210,6 @@ _mesa_init_texture(GLcontext *ctx) ctx->Texture.SharedPalette = GL_FALSE; _mesa_init_colortable(&ctx->Texture.Palette); - _mesa_TexEnvProgramCacheInit( ctx ); - /* Allocate proxy textures */ if (!alloc_proxy_textures( ctx )) return GL_FALSE; @@ -3239,6 +3237,4 @@ _mesa_free_texture_data(GLcontext *ctx) for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++) _mesa_free_colortable_data( &ctx->Texture.Unit[i].ColorTable ); - - _mesa_TexEnvProgramCacheDestroy( ctx ); } -- cgit v1.2.3 From ef25c496d52f4f6c45816b64b4c0999321476cd7 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:19:09 -0600 Subject: Copy of xlib driver. This will get trimmed down a lot for Gallium. --- src/mesa/pipe/xlib/fakeglx.c | 3186 ++++++++++++++++++++++++++ src/mesa/pipe/xlib/glxapi.c | 1408 ++++++++++++ src/mesa/pipe/xlib/glxapi.h | 228 ++ src/mesa/pipe/xlib/glxheader.h | 62 + src/mesa/pipe/xlib/realglx.c | 180 ++ src/mesa/pipe/xlib/realglx.h | 326 +++ src/mesa/pipe/xlib/xfonts.c | 377 +++ src/mesa/pipe/xlib/xfonts.h | 41 + src/mesa/pipe/xlib/xm_api.c | 2554 +++++++++++++++++++++ src/mesa/pipe/xlib/xm_buffer.c | 503 ++++ src/mesa/pipe/xlib/xm_dd.c | 1141 ++++++++++ src/mesa/pipe/xlib/xm_image.c | 133 ++ src/mesa/pipe/xlib/xm_image.h | 77 + src/mesa/pipe/xlib/xm_line.c | 691 ++++++ src/mesa/pipe/xlib/xm_span.c | 4815 +++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/xlib/xm_surface.c | 251 ++ src/mesa/pipe/xlib/xm_tri.c | 1694 ++++++++++++++ src/mesa/pipe/xlib/xm_winsys.c | 368 +++ src/mesa/pipe/xlib/xmesaP.h | 631 +++++ 19 files changed, 18666 insertions(+) create mode 100644 src/mesa/pipe/xlib/fakeglx.c create mode 100644 src/mesa/pipe/xlib/glxapi.c create mode 100644 src/mesa/pipe/xlib/glxapi.h create mode 100644 src/mesa/pipe/xlib/glxheader.h create mode 100644 src/mesa/pipe/xlib/realglx.c create mode 100644 src/mesa/pipe/xlib/realglx.h create mode 100644 src/mesa/pipe/xlib/xfonts.c create mode 100644 src/mesa/pipe/xlib/xfonts.h create mode 100644 src/mesa/pipe/xlib/xm_api.c create mode 100644 src/mesa/pipe/xlib/xm_buffer.c create mode 100644 src/mesa/pipe/xlib/xm_dd.c create mode 100644 src/mesa/pipe/xlib/xm_image.c create mode 100644 src/mesa/pipe/xlib/xm_image.h create mode 100644 src/mesa/pipe/xlib/xm_line.c create mode 100644 src/mesa/pipe/xlib/xm_span.c create mode 100644 src/mesa/pipe/xlib/xm_surface.c create mode 100644 src/mesa/pipe/xlib/xm_tri.c create mode 100644 src/mesa/pipe/xlib/xm_winsys.c create mode 100644 src/mesa/pipe/xlib/xmesaP.h (limited to 'src') diff --git a/src/mesa/pipe/xlib/fakeglx.c b/src/mesa/pipe/xlib/fakeglx.c new file mode 100644 index 0000000000..7a170b4d3d --- /dev/null +++ b/src/mesa/pipe/xlib/fakeglx.c @@ -0,0 +1,3186 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* + * This is an emulation of the GLX API which allows Mesa/GLX-based programs + * to run on X servers which do not have the real GLX extension. + * + * Thanks to the contributors: + * + * Initial version: Philip Brown (phil@bolthole.com) + * Better glXGetConfig() support: Armin Liebchen (liebchen@asylum.cs.utah.edu) + * Further visual-handling refinements: Wolfram Gloger + * (wmglo@Dent.MED.Uni-Muenchen.DE). + * + * Notes: + * Don't be fooled, stereo isn't supported yet. + */ + + + +#include "glxheader.h" +#include "glxapi.h" +#include "GL/xmesa.h" +#include "context.h" +#include "config.h" +#include "macros.h" +#include "imports.h" +#include "mtypes.h" +#include "version.h" +#include "xfonts.h" +#include "xmesaP.h" + +#ifdef __VMS +#define _mesa_sprintf sprintf +#endif + +/* This indicates the client-side GLX API and GLX encoder version. */ +#define CLIENT_MAJOR_VERSION 1 +#define CLIENT_MINOR_VERSION 4 /* but don't have 1.3's pbuffers, etc yet */ + +/* This indicates the server-side GLX decoder version. + * GLX 1.4 indicates OpenGL 1.3 support + */ +#define SERVER_MAJOR_VERSION 1 +#define SERVER_MINOR_VERSION 4 + +/* This is appended onto the glXGetClient/ServerString version strings. */ +#define MESA_GLX_VERSION "Mesa " MESA_VERSION_STRING + +/* Who implemented this GLX? */ +#define VENDOR "Brian Paul" + +#define EXTENSIONS \ + "GLX_MESA_set_3dfx_mode " \ + "GLX_MESA_copy_sub_buffer " \ + "GLX_MESA_pixmap_colormap " \ + "GLX_MESA_release_buffers " \ + "GLX_ARB_get_proc_address " \ + "GLX_EXT_texture_from_pixmap " \ + "GLX_EXT_visual_info " \ + "GLX_EXT_visual_rating " \ + /*"GLX_SGI_video_sync "*/ \ + "GLX_SGIX_fbconfig " \ + "GLX_SGIX_pbuffer " + +/* + * Our fake GLX context will contain a "real" GLX context and an XMesa context. + * + * Note that a pointer to a __GLXcontext is a pointer to a fake_glx_context, + * and vice versa. + * + * We really just need this structure in order to make the libGL functions + * glXGetCurrentContext(), glXGetCurrentDrawable() and glXGetCurrentDisplay() + * work correctly. + */ +struct fake_glx_context { + __GLXcontext glxContext; /* this MUST be first! */ + XMesaContext xmesaContext; +}; + + + +/**********************************************************************/ +/*** GLX Visual Code ***/ +/**********************************************************************/ + +#define DONT_CARE -1 + + +static XMesaVisual *VisualTable = NULL; +static int NumVisuals = 0; + + +/* + * This struct and some code fragments borrowed + * from Mark Kilgard's GLUT library. + */ +typedef struct _OverlayInfo { + /* Avoid 64-bit portability problems by being careful to use + longs due to the way XGetWindowProperty is specified. Note + that these parameters are passed as CARD32s over X + protocol. */ + unsigned long overlay_visual; + long transparent_type; + long value; + long layer; +} OverlayInfo; + + + +/* Macro to handle c_class vs class field name in XVisualInfo struct */ +#if defined(__cplusplus) || defined(c_plusplus) +#define CLASS c_class +#else +#define CLASS class +#endif + + + +/* + * Test if the given XVisualInfo is usable for Mesa rendering. + */ +static GLboolean +is_usable_visual( XVisualInfo *vinfo ) +{ + switch (vinfo->CLASS) { + case StaticGray: + case GrayScale: + /* Any StaticGray/GrayScale visual works in RGB or CI mode */ + return GL_TRUE; + case StaticColor: + case PseudoColor: + /* Any StaticColor/PseudoColor visual of at least 4 bits */ + if (vinfo->depth>=4) { + return GL_TRUE; + } + else { + return GL_FALSE; + } + case TrueColor: + case DirectColor: + /* Any depth of TrueColor or DirectColor works in RGB mode */ + return GL_TRUE; + default: + /* This should never happen */ + return GL_FALSE; + } +} + + + +/** + * Get an array OverlayInfo records for specified screen. + * \param dpy the display + * \param screen screen number + * \param numOverlays returns numver of OverlayInfo records + * \return pointer to OverlayInfo array, free with XFree() + */ +static OverlayInfo * +GetOverlayInfo(Display *dpy, int screen, int *numOverlays) +{ + Atom overlayVisualsAtom; + Atom actualType; + Status status; + unsigned char *ovInfo; + unsigned long sizeData, bytesLeft; + int actualFormat; + + /* + * The SERVER_OVERLAY_VISUALS property on the root window contains + * a list of overlay visuals. Get that list now. + */ + overlayVisualsAtom = XInternAtom(dpy,"SERVER_OVERLAY_VISUALS", True); + if (overlayVisualsAtom == None) { + return 0; + } + + status = XGetWindowProperty(dpy, RootWindow(dpy, screen), + overlayVisualsAtom, 0L, (long) 10000, False, + overlayVisualsAtom, &actualType, &actualFormat, + &sizeData, &bytesLeft, + &ovInfo); + + if (status != Success || actualType != overlayVisualsAtom || + actualFormat != 32 || sizeData < 4) { + /* something went wrong */ + XFree((void *) ovInfo); + *numOverlays = 0; + return NULL; + } + + *numOverlays = sizeData / 4; + return (OverlayInfo *) ovInfo; +} + + + +/** + * Return the level (overlay, normal, underlay) of a given XVisualInfo. + * Input: dpy - the X display + * vinfo - the XVisualInfo to test + * Return: level of the visual: + * 0 = normal planes + * >0 = overlay planes + * <0 = underlay planes + */ +static int +level_of_visual( Display *dpy, XVisualInfo *vinfo ) +{ + OverlayInfo *overlay_info; + int numOverlaysPerScreen, i; + + overlay_info = GetOverlayInfo(dpy, vinfo->screen, &numOverlaysPerScreen); + if (!overlay_info) { + return 0; + } + + /* search the overlay visual list for the visual ID of interest */ + for (i = 0; i < numOverlaysPerScreen; i++) { + const OverlayInfo *ov = overlay_info + i; + if (ov->overlay_visual == vinfo->visualid) { + /* found the visual */ + if (/*ov->transparent_type==1 &&*/ ov->layer!=0) { + int level = ov->layer; + XFree((void *) overlay_info); + return level; + } + else { + XFree((void *) overlay_info); + return 0; + } + } + } + + /* The visual ID was not found in the overlay list. */ + XFree((void *) overlay_info); + return 0; +} + + + + +/* + * Given an XVisualInfo and RGB, Double, and Depth buffer flags, save the + * configuration in our list of GLX visuals. + */ +static XMesaVisual +save_glx_visual( Display *dpy, XVisualInfo *vinfo, + GLboolean rgbFlag, GLboolean alphaFlag, GLboolean dbFlag, + GLboolean stereoFlag, + GLint depth_size, GLint stencil_size, + GLint accumRedSize, GLint accumGreenSize, + GLint accumBlueSize, GLint accumAlphaSize, + GLint level, GLint numAuxBuffers ) +{ + GLboolean ximageFlag = GL_TRUE; + XMesaVisual xmvis; + GLint i; + GLboolean comparePointers; + + if (dbFlag) { + /* Check if the MESA_BACK_BUFFER env var is set */ + char *backbuffer = _mesa_getenv("MESA_BACK_BUFFER"); + if (backbuffer) { + if (backbuffer[0]=='p' || backbuffer[0]=='P') { + ximageFlag = GL_FALSE; + } + else if (backbuffer[0]=='x' || backbuffer[0]=='X') { + ximageFlag = GL_TRUE; + } + else { + _mesa_warning(NULL, "Mesa: invalid value for MESA_BACK_BUFFER environment variable, using an XImage."); + } + } + } + + if (stereoFlag) { + /* stereo not supported */ + return NULL; + } + + /* Comparing IDs uses less memory but sometimes fails. */ + /* XXX revisit this after 3.0 is finished. */ + if (_mesa_getenv("MESA_GLX_VISUAL_HACK")) + comparePointers = GL_TRUE; + else + comparePointers = GL_FALSE; + + /* Force the visual to have an alpha channel */ + if (rgbFlag && _mesa_getenv("MESA_GLX_FORCE_ALPHA")) + alphaFlag = GL_TRUE; + + /* First check if a matching visual is already in the list */ + for (i=0; idisplay == dpy + && v->mesa_visual.level == level + && v->mesa_visual.numAuxBuffers == numAuxBuffers + && v->ximage_flag == ximageFlag + && v->mesa_visual.rgbMode == rgbFlag + && v->mesa_visual.doubleBufferMode == dbFlag + && v->mesa_visual.stereoMode == stereoFlag + && (v->mesa_visual.alphaBits > 0) == alphaFlag + && (v->mesa_visual.depthBits >= depth_size || depth_size == 0) + && (v->mesa_visual.stencilBits >= stencil_size || stencil_size == 0) + && (v->mesa_visual.accumRedBits >= accumRedSize || accumRedSize == 0) + && (v->mesa_visual.accumGreenBits >= accumGreenSize || accumGreenSize == 0) + && (v->mesa_visual.accumBlueBits >= accumBlueSize || accumBlueSize == 0) + && (v->mesa_visual.accumAlphaBits >= accumAlphaSize || accumAlphaSize == 0)) { + /* now either compare XVisualInfo pointers or visual IDs */ + if ((!comparePointers && v->visinfo->visualid == vinfo->visualid) + || (comparePointers && v->vishandle == vinfo)) { + return v; + } + } + } + + /* Create a new visual and add it to the list. */ + + xmvis = XMesaCreateVisual( dpy, vinfo, rgbFlag, alphaFlag, dbFlag, + stereoFlag, ximageFlag, + depth_size, stencil_size, + accumRedSize, accumBlueSize, + accumBlueSize, accumAlphaSize, 0, level, + GLX_NONE_EXT ); + if (xmvis) { + /* Save a copy of the pointer now so we can find this visual again + * if we need to search for it in find_glx_visual(). + */ + xmvis->vishandle = vinfo; + /* Allocate more space for additional visual */ + VisualTable = (XMesaVisual *) _mesa_realloc( VisualTable, + sizeof(XMesaVisual) * NumVisuals, + sizeof(XMesaVisual) * (NumVisuals + 1)); + /* add xmvis to the list */ + VisualTable[NumVisuals] = xmvis; + NumVisuals++; + /* XXX minor hack, because XMesaCreateVisual doesn't support an + * aux buffers parameter. + */ + xmvis->mesa_visual.numAuxBuffers = numAuxBuffers; + } + return xmvis; +} + + +/** + * Return the default number of bits for the Z buffer. + * If defined, use the MESA_GLX_DEPTH_BITS env var value. + * Otherwise, use the DEFAULT_SOFTWARE_DEPTH_BITS constant. + * XXX probably do the same thing for stencil, accum, etc. + */ +static GLint +default_depth_bits(void) +{ + int zBits; + const char *zEnv = _mesa_getenv("MESA_GLX_DEPTH_BITS"); + if (zEnv) + zBits = _mesa_atoi(zEnv); + else + zBits = DEFAULT_SOFTWARE_DEPTH_BITS; + return zBits; +} + +static GLint +default_alpha_bits(void) +{ + int aBits; + const char *aEnv = _mesa_getenv("MESA_GLX_ALPHA_BITS"); + if (aEnv) + aBits = _mesa_atoi(aEnv); + else + aBits = 0; + return aBits; +} + +static GLint +default_accum_bits(void) +{ + return 16; +} + + + +/* + * Create a GLX visual from a regular XVisualInfo. + * This is called when Fake GLX is given an XVisualInfo which wasn't + * returned by glXChooseVisual. Since this is the first time we're + * considering this visual we'll take a guess at reasonable values + * for depth buffer size, stencil size, accum size, etc. + * This is the best we can do with a client-side emulation of GLX. + */ +static XMesaVisual +create_glx_visual( Display *dpy, XVisualInfo *visinfo ) +{ + int vislevel; + GLint zBits = default_depth_bits(); + GLint accBits = default_accum_bits(); + GLboolean alphaFlag = default_alpha_bits() > 0; + + vislevel = level_of_visual( dpy, visinfo ); + if (vislevel) { + /* Configure this visual as a CI, single-buffered overlay */ + return save_glx_visual( dpy, visinfo, + GL_FALSE, /* rgb */ + GL_FALSE, /* alpha */ + GL_FALSE, /* double */ + GL_FALSE, /* stereo */ + 0, /* depth bits */ + 0, /* stencil bits */ + 0,0,0,0, /* accum bits */ + vislevel, /* level */ + 0 /* numAux */ + ); + } + else if (is_usable_visual( visinfo )) { + if (_mesa_getenv("MESA_GLX_FORCE_CI")) { + /* Configure this visual as a COLOR INDEX visual. */ + return save_glx_visual( dpy, visinfo, + GL_FALSE, /* rgb */ + GL_FALSE, /* alpha */ + GL_TRUE, /* double */ + GL_FALSE, /* stereo */ + zBits, + STENCIL_BITS, + 0, 0, 0, 0, /* accum bits */ + 0, /* level */ + 0 /* numAux */ + ); + } + else { + /* Configure this visual as RGB, double-buffered, depth-buffered. */ + /* This is surely wrong for some people's needs but what else */ + /* can be done? They should use glXChooseVisual(). */ + return save_glx_visual( dpy, visinfo, + GL_TRUE, /* rgb */ + alphaFlag, /* alpha */ + GL_TRUE, /* double */ + GL_FALSE, /* stereo */ + zBits, + STENCIL_BITS, + accBits, /* r */ + accBits, /* g */ + accBits, /* b */ + accBits, /* a */ + 0, /* level */ + 0 /* numAux */ + ); + } + } + else { + _mesa_warning(NULL, "Mesa: error in glXCreateContext: bad visual\n"); + return NULL; + } +} + + + +/* + * Find the GLX visual associated with an XVisualInfo. + */ +static XMesaVisual +find_glx_visual( Display *dpy, XVisualInfo *vinfo ) +{ + int i; + + /* try to match visual id */ + for (i=0;idisplay==dpy + && VisualTable[i]->visinfo->visualid == vinfo->visualid) { + return VisualTable[i]; + } + } + + /* if that fails, try to match pointers */ + for (i=0;idisplay==dpy && VisualTable[i]->vishandle==vinfo) { + return VisualTable[i]; + } + } + + return NULL; +} + + + +/** + * Return the transparent pixel value for a GLX visual. + * Input: glxvis - the glx_visual + * Return: a pixel value or -1 if no transparent pixel + */ +static int +transparent_pixel( XMesaVisual glxvis ) +{ + Display *dpy = glxvis->display; + XVisualInfo *vinfo = glxvis->visinfo; + OverlayInfo *overlay_info; + int numOverlaysPerScreen, i; + + overlay_info = GetOverlayInfo(dpy, vinfo->screen, &numOverlaysPerScreen); + if (!overlay_info) { + return -1; + } + + for (i = 0; i < numOverlaysPerScreen; i++) { + const OverlayInfo *ov = overlay_info + i; + if (ov->overlay_visual == vinfo->visualid) { + /* found it! */ + if (ov->transparent_type == 0) { + /* type 0 indicates no transparency */ + XFree((void *) overlay_info); + return -1; + } + else { + /* ov->value is the transparent pixel */ + XFree((void *) overlay_info); + return ov->value; + } + } + } + + /* The visual ID was not found in the overlay list. */ + XFree((void *) overlay_info); + return -1; +} + + + +/** + * Try to get an X visual which matches the given arguments. + */ +static XVisualInfo * +get_visual( Display *dpy, int scr, unsigned int depth, int xclass ) +{ + XVisualInfo temp, *vis; + long mask; + int n; + unsigned int default_depth; + int default_class; + + mask = VisualScreenMask | VisualDepthMask | VisualClassMask; + temp.screen = scr; + temp.depth = depth; + temp.CLASS = xclass; + + default_depth = DefaultDepth(dpy,scr); + default_class = DefaultVisual(dpy,scr)->CLASS; + + if (depth==default_depth && xclass==default_class) { + /* try to get root window's visual */ + temp.visualid = DefaultVisual(dpy,scr)->visualid; + mask |= VisualIDMask; + } + + vis = XGetVisualInfo( dpy, mask, &temp, &n ); + + /* In case bits/pixel > 24, make sure color channels are still <=8 bits. + * An SGI Infinite Reality system, for example, can have 30bpp pixels: + * 10 bits per color channel. Mesa's limited to a max of 8 bits/channel. + */ + if (vis && depth > 24 && (xclass==TrueColor || xclass==DirectColor)) { + if (_mesa_bitcount((GLuint) vis->red_mask ) <= 8 && + _mesa_bitcount((GLuint) vis->green_mask) <= 8 && + _mesa_bitcount((GLuint) vis->blue_mask ) <= 8) { + return vis; + } + else { + XFree((void *) vis); + return NULL; + } + } + + return vis; +} + + + +/* + * Retrieve the value of the given environment variable and find + * the X visual which matches it. + * Input: dpy - the display + * screen - the screen number + * varname - the name of the environment variable + * Return: an XVisualInfo pointer to NULL if error. + */ +static XVisualInfo * +get_env_visual(Display *dpy, int scr, const char *varname) +{ + char value[100], type[100]; + int depth, xclass = -1; + XVisualInfo *vis; + + if (!_mesa_getenv( varname )) { + return NULL; + } + + _mesa_strncpy( value, _mesa_getenv(varname), 100 ); + value[99] = 0; + + sscanf( value, "%s %d", type, &depth ); + + if (_mesa_strcmp(type,"TrueColor")==0) xclass = TrueColor; + else if (_mesa_strcmp(type,"DirectColor")==0) xclass = DirectColor; + else if (_mesa_strcmp(type,"PseudoColor")==0) xclass = PseudoColor; + else if (_mesa_strcmp(type,"StaticColor")==0) xclass = StaticColor; + else if (_mesa_strcmp(type,"GrayScale")==0) xclass = GrayScale; + else if (_mesa_strcmp(type,"StaticGray")==0) xclass = StaticGray; + + if (xclass>-1 && depth>0) { + vis = get_visual( dpy, scr, depth, xclass ); + if (vis) { + return vis; + } + } + + _mesa_warning(NULL, "GLX unable to find visual class=%s, depth=%d.", + type, depth); + + return NULL; +} + + + +/* + * Select an X visual which satisfies the RGBA/CI flag and minimum depth. + * Input: dpy, screen - X display and screen number + * rgba - GL_TRUE = RGBA mode, GL_FALSE = CI mode + * min_depth - minimum visual depth + * preferred_class - preferred GLX visual class or DONT_CARE + * Return: pointer to an XVisualInfo or NULL. + */ +static XVisualInfo * +choose_x_visual( Display *dpy, int screen, GLboolean rgba, int min_depth, + int preferred_class ) +{ + XVisualInfo *vis; + int xclass, visclass = 0; + int depth; + + if (rgba) { + Atom hp_cr_maps = XInternAtom(dpy, "_HP_RGB_SMOOTH_MAP_LIST", True); + /* First see if the MESA_RGB_VISUAL env var is defined */ + vis = get_env_visual( dpy, screen, "MESA_RGB_VISUAL" ); + if (vis) { + return vis; + } + /* Otherwise, search for a suitable visual */ + if (preferred_class==DONT_CARE) { + for (xclass=0;xclass<6;xclass++) { + switch (xclass) { + case 0: visclass = TrueColor; break; + case 1: visclass = DirectColor; break; + case 2: visclass = PseudoColor; break; + case 3: visclass = StaticColor; break; + case 4: visclass = GrayScale; break; + case 5: visclass = StaticGray; break; + } + if (min_depth==0) { + /* start with shallowest */ + for (depth=0;depth<=32;depth++) { + if (visclass==TrueColor && depth==8 && !hp_cr_maps) { + /* Special case: try to get 8-bit PseudoColor before */ + /* 8-bit TrueColor */ + vis = get_visual( dpy, screen, 8, PseudoColor ); + if (vis) { + return vis; + } + } + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + else { + /* start with deepest */ + for (depth=32;depth>=min_depth;depth--) { + if (visclass==TrueColor && depth==8 && !hp_cr_maps) { + /* Special case: try to get 8-bit PseudoColor before */ + /* 8-bit TrueColor */ + vis = get_visual( dpy, screen, 8, PseudoColor ); + if (vis) { + return vis; + } + } + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + } + } + else { + /* search for a specific visual class */ + switch (preferred_class) { + case GLX_TRUE_COLOR_EXT: visclass = TrueColor; break; + case GLX_DIRECT_COLOR_EXT: visclass = DirectColor; break; + case GLX_PSEUDO_COLOR_EXT: visclass = PseudoColor; break; + case GLX_STATIC_COLOR_EXT: visclass = StaticColor; break; + case GLX_GRAY_SCALE_EXT: visclass = GrayScale; break; + case GLX_STATIC_GRAY_EXT: visclass = StaticGray; break; + default: return NULL; + } + if (min_depth==0) { + /* start with shallowest */ + for (depth=0;depth<=32;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + else { + /* start with deepest */ + for (depth=32;depth>=min_depth;depth--) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + } + } + else { + /* First see if the MESA_CI_VISUAL env var is defined */ + vis = get_env_visual( dpy, screen, "MESA_CI_VISUAL" ); + if (vis) { + return vis; + } + /* Otherwise, search for a suitable visual, starting with shallowest */ + if (preferred_class==DONT_CARE) { + for (xclass=0;xclass<4;xclass++) { + switch (xclass) { + case 0: visclass = PseudoColor; break; + case 1: visclass = StaticColor; break; + case 2: visclass = GrayScale; break; + case 3: visclass = StaticGray; break; + } + /* try 8-bit up through 16-bit */ + for (depth=8;depth<=16;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + /* try min_depth up to 8-bit */ + for (depth=min_depth;depth<8;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + } + else { + /* search for a specific visual class */ + switch (preferred_class) { + case GLX_TRUE_COLOR_EXT: visclass = TrueColor; break; + case GLX_DIRECT_COLOR_EXT: visclass = DirectColor; break; + case GLX_PSEUDO_COLOR_EXT: visclass = PseudoColor; break; + case GLX_STATIC_COLOR_EXT: visclass = StaticColor; break; + case GLX_GRAY_SCALE_EXT: visclass = GrayScale; break; + case GLX_STATIC_GRAY_EXT: visclass = StaticGray; break; + default: return NULL; + } + /* try 8-bit up through 16-bit */ + for (depth=8;depth<=16;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + /* try min_depth up to 8-bit */ + for (depth=min_depth;depth<8;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + } + + /* didn't find a visual */ + return NULL; +} + + + +/* + * Find the deepest X over/underlay visual of at least min_depth. + * Input: dpy, screen - X display and screen number + * level - the over/underlay level + * trans_type - transparent pixel type: GLX_NONE_EXT, + * GLX_TRANSPARENT_RGB_EXT, GLX_TRANSPARENT_INDEX_EXT, + * or DONT_CARE + * trans_value - transparent pixel value or DONT_CARE + * min_depth - minimum visual depth + * preferred_class - preferred GLX visual class or DONT_CARE + * Return: pointer to an XVisualInfo or NULL. + */ +static XVisualInfo * +choose_x_overlay_visual( Display *dpy, int scr, GLboolean rgbFlag, + int level, int trans_type, int trans_value, + int min_depth, int preferred_class ) +{ + OverlayInfo *overlay_info; + int numOverlaysPerScreen; + int i; + XVisualInfo *deepvis; + int deepest; + + /*DEBUG int tt, tv; */ + + switch (preferred_class) { + case GLX_TRUE_COLOR_EXT: preferred_class = TrueColor; break; + case GLX_DIRECT_COLOR_EXT: preferred_class = DirectColor; break; + case GLX_PSEUDO_COLOR_EXT: preferred_class = PseudoColor; break; + case GLX_STATIC_COLOR_EXT: preferred_class = StaticColor; break; + case GLX_GRAY_SCALE_EXT: preferred_class = GrayScale; break; + case GLX_STATIC_GRAY_EXT: preferred_class = StaticGray; break; + default: preferred_class = DONT_CARE; + } + + overlay_info = GetOverlayInfo(dpy, scr, &numOverlaysPerScreen); + if (!overlay_info) { + return NULL; + } + + /* Search for the deepest overlay which satisifies all criteria. */ + deepest = min_depth; + deepvis = NULL; + + for (i = 0; i < numOverlaysPerScreen; i++) { + const OverlayInfo *ov = overlay_info + i; + XVisualInfo *vislist, vistemplate; + int count; + + if (ov->layer!=level) { + /* failed overlay level criteria */ + continue; + } + if (!(trans_type==DONT_CARE + || (trans_type==GLX_TRANSPARENT_INDEX_EXT + && ov->transparent_type>0) + || (trans_type==GLX_NONE_EXT && ov->transparent_type==0))) { + /* failed transparent pixel type criteria */ + continue; + } + if (trans_value!=DONT_CARE && trans_value!=ov->value) { + /* failed transparent pixel value criteria */ + continue; + } + + /* get XVisualInfo and check the depth */ + vistemplate.visualid = ov->overlay_visual; + vistemplate.screen = scr; + vislist = XGetVisualInfo( dpy, VisualIDMask | VisualScreenMask, + &vistemplate, &count ); + + if (count!=1) { + /* something went wrong */ + continue; + } + if (preferred_class!=DONT_CARE && preferred_class!=vislist->CLASS) { + /* wrong visual class */ + continue; + } + + /* if RGB was requested, make sure we have True/DirectColor */ + if (rgbFlag && vislist->CLASS != TrueColor + && vislist->CLASS != DirectColor) + continue; + + /* if CI was requested, make sure we have a color indexed visual */ + if (!rgbFlag + && (vislist->CLASS == TrueColor || vislist->CLASS == DirectColor)) + continue; + + if (deepvis==NULL || vislist->depth > deepest) { + /* YES! found a satisfactory visual */ + if (deepvis) { + XFree( deepvis ); + } + deepest = vislist->depth; + deepvis = vislist; + /* DEBUG tt = ov->transparent_type;*/ + /* DEBUG tv = ov->value; */ + } + } + +/*DEBUG + if (deepvis) { + printf("chose 0x%x: layer=%d depth=%d trans_type=%d trans_value=%d\n", + deepvis->visualid, level, deepvis->depth, tt, tv ); + } +*/ + return deepvis; +} + + +/**********************************************************************/ +/*** Display-related functions ***/ +/**********************************************************************/ + + +/** + * Free all XMesaVisuals which are associated with the given display. + */ +static void +destroy_visuals_on_display(Display *dpy) +{ + int i; + for (i = 0; i < NumVisuals; i++) { + if (VisualTable[i]->display == dpy) { + /* remove this visual */ + int j; + free(VisualTable[i]); + for (j = i; j < NumVisuals - 1; j++) + VisualTable[j] = VisualTable[j + 1]; + NumVisuals--; + } + } +} + + +/** + * Called from XCloseDisplay() to let us free our display-related data. + */ +static int +close_display_callback(Display *dpy, XExtCodes *codes) +{ + destroy_visuals_on_display(dpy); + xmesa_destroy_buffers_on_display(dpy); + return 0; +} + + +/** + * Look for the named extension on given display and return a pointer + * to the _XExtension data, or NULL if extension not found. + */ +static _XExtension * +lookup_extension(Display *dpy, const char *extName) +{ + _XExtension *ext; + for (ext = dpy->ext_procs; ext; ext = ext->next) { + if (ext->name && strcmp(ext->name, extName) == 0) { + return ext; + } + } + return NULL; +} + + +/** + * Whenever we're given a new Display pointer, call this function to + * register our close_display_callback function. + */ +static void +register_with_display(Display *dpy) +{ + const char *extName = "MesaGLX"; + _XExtension *ext; + + ext = lookup_extension(dpy, extName); + if (!ext) { + XExtCodes *c = XAddExtension(dpy); + ext = dpy->ext_procs; /* new extension is at head of list */ + assert(c->extension == ext->codes.extension); + ext->name = _mesa_strdup(extName); + ext->close_display = close_display_callback; + } +} + + +/**********************************************************************/ +/*** Begin Fake GLX API Functions ***/ +/**********************************************************************/ + + +/** + * Helper used by glXChooseVisual and glXChooseFBConfig. + * The fbConfig parameter must be GL_FALSE for the former and GL_TRUE for + * the later. + * In either case, the attribute list is terminated with the value 'None'. + */ +static XMesaVisual +choose_visual( Display *dpy, int screen, const int *list, GLboolean fbConfig ) +{ + const GLboolean rgbModeDefault = fbConfig; + const int *parselist; + XVisualInfo *vis; + int min_ci = 0; + int min_red=0, min_green=0, min_blue=0; + GLboolean rgb_flag = rgbModeDefault; + GLboolean alpha_flag = GL_FALSE; + GLboolean double_flag = GL_FALSE; + GLboolean stereo_flag = GL_FALSE; + GLint depth_size = 0; + GLint stencil_size = 0; + GLint accumRedSize = 0; + GLint accumGreenSize = 0; + GLint accumBlueSize = 0; + GLint accumAlphaSize = 0; + int level = 0; + int visual_type = DONT_CARE; + int trans_type = DONT_CARE; + int trans_value = DONT_CARE; + GLint caveat = DONT_CARE; + XMesaVisual xmvis = NULL; + int desiredVisualID = -1; + int numAux = 0; + + parselist = list; + + while (*parselist) { + + switch (*parselist) { + case GLX_USE_GL: + if (fbConfig) { + /* invalid token */ + return NULL; + } + else { + /* skip */ + parselist++; + } + break; + case GLX_BUFFER_SIZE: + parselist++; + min_ci = *parselist++; + break; + case GLX_LEVEL: + parselist++; + level = *parselist++; + break; + case GLX_RGBA: + if (fbConfig) { + /* invalid token */ + return NULL; + } + else { + rgb_flag = GL_TRUE; + parselist++; + } + break; + case GLX_DOUBLEBUFFER: + parselist++; + if (fbConfig) { + double_flag = *parselist++; + } + else { + double_flag = GL_TRUE; + } + break; + case GLX_STEREO: + parselist++; + if (fbConfig) { + stereo_flag = *parselist++; + } + else { + stereo_flag = GL_TRUE; + } + break; + case GLX_AUX_BUFFERS: + parselist++; + numAux = *parselist++; + if (numAux > MAX_AUX_BUFFERS) + return NULL; + break; + case GLX_RED_SIZE: + parselist++; + min_red = *parselist++; + break; + case GLX_GREEN_SIZE: + parselist++; + min_green = *parselist++; + break; + case GLX_BLUE_SIZE: + parselist++; + min_blue = *parselist++; + break; + case GLX_ALPHA_SIZE: + parselist++; + { + GLint size = *parselist++; + alpha_flag = size ? GL_TRUE : GL_FALSE; + } + break; + case GLX_DEPTH_SIZE: + parselist++; + depth_size = *parselist++; + break; + case GLX_STENCIL_SIZE: + parselist++; + stencil_size = *parselist++; + break; + case GLX_ACCUM_RED_SIZE: + parselist++; + { + GLint size = *parselist++; + accumRedSize = MAX2( accumRedSize, size ); + } + break; + case GLX_ACCUM_GREEN_SIZE: + parselist++; + { + GLint size = *parselist++; + accumGreenSize = MAX2( accumGreenSize, size ); + } + break; + case GLX_ACCUM_BLUE_SIZE: + parselist++; + { + GLint size = *parselist++; + accumBlueSize = MAX2( accumBlueSize, size ); + } + break; + case GLX_ACCUM_ALPHA_SIZE: + parselist++; + { + GLint size = *parselist++; + accumAlphaSize = MAX2( accumAlphaSize, size ); + } + break; + + /* + * GLX_EXT_visual_info extension + */ + case GLX_X_VISUAL_TYPE_EXT: + parselist++; + visual_type = *parselist++; + break; + case GLX_TRANSPARENT_TYPE_EXT: + parselist++; + trans_type = *parselist++; + break; + case GLX_TRANSPARENT_INDEX_VALUE_EXT: + parselist++; + trans_value = *parselist++; + break; + case GLX_TRANSPARENT_RED_VALUE_EXT: + case GLX_TRANSPARENT_GREEN_VALUE_EXT: + case GLX_TRANSPARENT_BLUE_VALUE_EXT: + case GLX_TRANSPARENT_ALPHA_VALUE_EXT: + /* ignore */ + parselist++; + parselist++; + break; + + /* + * GLX_EXT_visual_info extension + */ + case GLX_VISUAL_CAVEAT_EXT: + parselist++; + caveat = *parselist++; /* ignored for now */ + break; + + /* + * GLX_ARB_multisample + */ + case GLX_SAMPLE_BUFFERS_ARB: + /* ms not supported */ + return NULL; + case GLX_SAMPLES_ARB: + /* ms not supported */ + return NULL; + + /* + * FBConfig attribs. + */ + case GLX_RENDER_TYPE: + if (!fbConfig) + return NULL; + parselist++; + if (*parselist == GLX_RGBA_BIT) { + rgb_flag = GL_TRUE; + } + else if (*parselist == GLX_COLOR_INDEX_BIT) { + rgb_flag = GL_FALSE; + } + else if (*parselist == 0) { + rgb_flag = GL_TRUE; + } + parselist++; + break; + case GLX_DRAWABLE_TYPE: + if (!fbConfig) + return NULL; + parselist++; + if (*parselist & ~(GLX_WINDOW_BIT | GLX_PIXMAP_BIT | GLX_PBUFFER_BIT)) { + return NULL; /* bad bit */ + } + parselist++; + break; + case GLX_FBCONFIG_ID: + if (!fbConfig) + return NULL; + parselist++; + desiredVisualID = *parselist++; + break; + case GLX_X_RENDERABLE: + if (!fbConfig) + return NULL; + parselist += 2; + /* ignore */ + break; + +#ifdef GLX_EXT_texture_from_pixmap + case GLX_BIND_TO_TEXTURE_RGB_EXT: + parselist++; /*skip*/ + break; + case GLX_BIND_TO_TEXTURE_RGBA_EXT: + parselist++; /*skip*/ + break; + case GLX_BIND_TO_MIPMAP_TEXTURE_EXT: + parselist++; /*skip*/ + break; + case GLX_BIND_TO_TEXTURE_TARGETS_EXT: + parselist++; + if (*parselist & ~(GLX_TEXTURE_1D_BIT_EXT | + GLX_TEXTURE_2D_BIT_EXT | + GLX_TEXTURE_RECTANGLE_BIT_EXT)) { + /* invalid bit */ + return NULL; + } + break; + case GLX_Y_INVERTED_EXT: + parselist++; /*skip*/ + break; +#endif + + case None: + /* end of list */ + break; + + default: + /* undefined attribute */ + _mesa_warning(NULL, "unexpected attrib 0x%x in choose_visual()", + *parselist); + return NULL; + } + } + + (void) caveat; + + /* + * Since we're only simulating the GLX extension this function will never + * find any real GL visuals. Instead, all we can do is try to find an RGB + * or CI visual of appropriate depth. Other requested attributes such as + * double buffering, depth buffer, etc. will be associated with the X + * visual and stored in the VisualTable[]. + */ + if (desiredVisualID != -1) { + /* try to get a specific visual, by visualID */ + XVisualInfo temp; + int n; + temp.visualid = desiredVisualID; + temp.screen = screen; + vis = XGetVisualInfo(dpy, VisualIDMask | VisualScreenMask, &temp, &n); + if (vis) { + /* give the visual some useful GLX attributes */ + double_flag = GL_TRUE; + if (vis->depth > 8) + rgb_flag = GL_TRUE; + depth_size = default_depth_bits(); + stencil_size = STENCIL_BITS; + /* XXX accum??? */ + } + } + else if (level==0) { + /* normal color planes */ + if (rgb_flag) { + /* Get an RGB visual */ + int min_rgb = min_red + min_green + min_blue; + if (min_rgb>1 && min_rgb<8) { + /* a special case to be sure we can get a monochrome visual */ + min_rgb = 1; + } + vis = choose_x_visual( dpy, screen, rgb_flag, min_rgb, visual_type ); + } + else { + /* Get a color index visual */ + vis = choose_x_visual( dpy, screen, rgb_flag, min_ci, visual_type ); + accumRedSize = accumGreenSize = accumBlueSize = accumAlphaSize = 0; + } + } + else { + /* over/underlay planes */ + if (rgb_flag) { + /* rgba overlay */ + int min_rgb = min_red + min_green + min_blue; + if (min_rgb>1 && min_rgb<8) { + /* a special case to be sure we can get a monochrome visual */ + min_rgb = 1; + } + vis = choose_x_overlay_visual( dpy, screen, rgb_flag, level, + trans_type, trans_value, min_rgb, visual_type ); + } + else { + /* color index overlay */ + vis = choose_x_overlay_visual( dpy, screen, rgb_flag, level, + trans_type, trans_value, min_ci, visual_type ); + } + } + + if (vis) { + /* Note: we're not exactly obeying the glXChooseVisual rules here. + * When GLX_DEPTH_SIZE = 1 is specified we're supposed to choose the + * largest depth buffer size, which is 32bits/value. Instead, we + * return 16 to maintain performance with earlier versions of Mesa. + */ + if (depth_size > 24) + depth_size = 32; + else if (depth_size > 16) + depth_size = 24; + else if (depth_size > 0) { + depth_size = default_depth_bits(); + } + + if (!alpha_flag) { + alpha_flag = default_alpha_bits() > 0; + } + + /* we only support one size of stencil and accum buffers. */ + if (stencil_size > 0) + stencil_size = STENCIL_BITS; + if (accumRedSize > 0 || accumGreenSize > 0 || accumBlueSize > 0 || + accumAlphaSize > 0) { + accumRedSize = + accumGreenSize = + accumBlueSize = default_accum_bits(); + accumAlphaSize = alpha_flag ? accumRedSize : 0; + } + + xmvis = save_glx_visual( dpy, vis, rgb_flag, alpha_flag, double_flag, + stereo_flag, depth_size, stencil_size, + accumRedSize, accumGreenSize, + accumBlueSize, accumAlphaSize, level, numAux ); + } + + return xmvis; +} + + +static XVisualInfo * +Fake_glXChooseVisual( Display *dpy, int screen, int *list ) +{ + XMesaVisual xmvis; + + /* register ourselves as an extension on this display */ + register_with_display(dpy); + + xmvis = choose_visual(dpy, screen, list, GL_FALSE); + if (xmvis) { +#if 0 + return xmvis->vishandle; +#else + /* create a new vishandle - the cached one may be stale */ + xmvis->vishandle = (XVisualInfo *) _mesa_malloc(sizeof(XVisualInfo)); + if (xmvis->vishandle) { + _mesa_memcpy(xmvis->vishandle, xmvis->visinfo, sizeof(XVisualInfo)); + } + return xmvis->vishandle; +#endif + } + else + return NULL; +} + + +static GLXContext +Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo, + GLXContext share_list, Bool direct ) +{ + XMesaVisual xmvis; + struct fake_glx_context *glxCtx; + struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list; + + if (!dpy || !visinfo) + return 0; + + glxCtx = CALLOC_STRUCT(fake_glx_context); + if (!glxCtx) + return 0; + + /* deallocate unused windows/buffers */ +#if 0 + XMesaGarbageCollect(); +#endif + + xmvis = find_glx_visual( dpy, visinfo ); + if (!xmvis) { + /* This visual wasn't found with glXChooseVisual() */ + xmvis = create_glx_visual( dpy, visinfo ); + if (!xmvis) { + /* unusable visual */ + _mesa_free(glxCtx); + return NULL; + } + } + + glxCtx->xmesaContext = XMesaCreateContext(xmvis, + shareCtx ? shareCtx->xmesaContext : NULL); + if (!glxCtx->xmesaContext) { + _mesa_free(glxCtx); + return NULL; + } + + glxCtx->xmesaContext->direct = GL_FALSE; + glxCtx->glxContext.isDirect = GL_FALSE; + glxCtx->glxContext.currentDpy = dpy; + glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ + + assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); + + return (GLXContext) glxCtx; +} + + +/* XXX these may have to be removed due to thread-safety issues. */ +static GLXContext MakeCurrent_PrevContext = 0; +static GLXDrawable MakeCurrent_PrevDrawable = 0; +static GLXDrawable MakeCurrent_PrevReadable = 0; +static XMesaBuffer MakeCurrent_PrevDrawBuffer = 0; +static XMesaBuffer MakeCurrent_PrevReadBuffer = 0; + + +/* GLX 1.3 and later */ +static Bool +Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, + GLXDrawable read, GLXContext ctx ) +{ + struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + + if (ctx && draw && read) { + XMesaBuffer drawBuffer, readBuffer; + XMesaContext xmctx = glxCtx->xmesaContext; + + /* Find the XMesaBuffer which corresponds to the GLXDrawable 'draw' */ + if (ctx == MakeCurrent_PrevContext + && draw == MakeCurrent_PrevDrawable) { + drawBuffer = MakeCurrent_PrevDrawBuffer; + } + else { + drawBuffer = XMesaFindBuffer( dpy, draw ); + } + if (!drawBuffer) { + /* drawable must be a new window! */ + drawBuffer = XMesaCreateWindowBuffer( xmctx->xm_visual, draw ); + if (!drawBuffer) { + /* Out of memory, or context/drawable depth mismatch */ + return False; + } +#ifdef FX + FXcreateContext( xmctx->xm_visual, draw, xmctx, drawBuffer ); +#endif + } + + /* Find the XMesaBuffer which corresponds to the GLXDrawable 'read' */ + if (ctx == MakeCurrent_PrevContext + && read == MakeCurrent_PrevReadable) { + readBuffer = MakeCurrent_PrevReadBuffer; + } + else { + readBuffer = XMesaFindBuffer( dpy, read ); + } + if (!readBuffer) { + /* drawable must be a new window! */ + readBuffer = XMesaCreateWindowBuffer( xmctx->xm_visual, read ); + if (!readBuffer) { + /* Out of memory, or context/drawable depth mismatch */ + return False; + } +#ifdef FX + FXcreateContext( xmctx->xm_visual, read, xmctx, readBuffer ); +#endif + } + + MakeCurrent_PrevContext = ctx; + MakeCurrent_PrevDrawable = draw; + MakeCurrent_PrevReadable = read; + MakeCurrent_PrevDrawBuffer = drawBuffer; + MakeCurrent_PrevReadBuffer = readBuffer; + + /* Now make current! */ + if (XMesaMakeCurrent2(xmctx, drawBuffer, readBuffer)) { + ((__GLXcontext *) ctx)->currentDpy = dpy; + ((__GLXcontext *) ctx)->currentDrawable = draw; + ((__GLXcontext *) ctx)->currentReadable = read; + return True; + } + else { + return False; + } + } + else if (!ctx && !draw && !read) { + /* release current context w/out assigning new one. */ + XMesaMakeCurrent( NULL, NULL ); + MakeCurrent_PrevContext = 0; + MakeCurrent_PrevDrawable = 0; + MakeCurrent_PrevReadable = 0; + MakeCurrent_PrevDrawBuffer = 0; + MakeCurrent_PrevReadBuffer = 0; + return True; + } + else { + /* The args must either all be non-zero or all zero. + * This is an error. + */ + return False; + } +} + + +static Bool +Fake_glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx ) +{ + return Fake_glXMakeContextCurrent( dpy, drawable, drawable, ctx ); +} + + +static GLXPixmap +Fake_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ) +{ + XMesaVisual v; + XMesaBuffer b; + + v = find_glx_visual( dpy, visinfo ); + if (!v) { + v = create_glx_visual( dpy, visinfo ); + if (!v) { + /* unusable visual */ + return 0; + } + } + + b = XMesaCreatePixmapBuffer( v, pixmap, 0 ); + if (!b) { + return 0; + } + return b->frontxrb->pixmap; +} + + +/*** GLX_MESA_pixmap_colormap ***/ + +static GLXPixmap +Fake_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, + Pixmap pixmap, Colormap cmap ) +{ + XMesaVisual v; + XMesaBuffer b; + + v = find_glx_visual( dpy, visinfo ); + if (!v) { + v = create_glx_visual( dpy, visinfo ); + if (!v) { + /* unusable visual */ + return 0; + } + } + + b = XMesaCreatePixmapBuffer( v, pixmap, cmap ); + if (!b) { + return 0; + } + return b->frontxrb->pixmap; +} + + +static void +Fake_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, pixmap); + if (b) { + XMesaDestroyBuffer(b); + } + else if (_mesa_getenv("MESA_DEBUG")) { + _mesa_warning(NULL, "Mesa: glXDestroyGLXPixmap: invalid pixmap\n"); + } +} + + +static void +Fake_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, + unsigned long mask ) +{ + struct fake_glx_context *fakeSrc = (struct fake_glx_context *) src; + struct fake_glx_context *fakeDst = (struct fake_glx_context *) dst; + XMesaContext xm_src = fakeSrc->xmesaContext; + XMesaContext xm_dst = fakeDst->xmesaContext; + (void) dpy; + if (MakeCurrent_PrevContext == src) { + _mesa_Flush(); + } + _mesa_copy_context( &(xm_src->mesa), &(xm_dst->mesa), (GLuint) mask ); +} + + +static Bool +Fake_glXQueryExtension( Display *dpy, int *errorb, int *event ) +{ + /* Mesa's GLX isn't really an X extension but we try to act like one. */ + (void) dpy; + (void) errorb; + (void) event; + return True; +} + + +extern void _kw_ungrab_all( Display *dpy ); +void _kw_ungrab_all( Display *dpy ) +{ + XUngrabPointer( dpy, CurrentTime ); + XUngrabKeyboard( dpy, CurrentTime ); +} + + +static void +Fake_glXDestroyContext( Display *dpy, GLXContext ctx ) +{ + struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + (void) dpy; + MakeCurrent_PrevContext = 0; + MakeCurrent_PrevDrawable = 0; + MakeCurrent_PrevReadable = 0; + MakeCurrent_PrevDrawBuffer = 0; + MakeCurrent_PrevReadBuffer = 0; + XMesaDestroyContext( glxCtx->xmesaContext ); + XMesaGarbageCollect(); + _mesa_free(glxCtx); +} + + +static Bool +Fake_glXIsDirect( Display *dpy, GLXContext ctx ) +{ + struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + (void) dpy; + return glxCtx->xmesaContext->direct; +} + + + +static void +Fake_glXSwapBuffers( Display *dpy, GLXDrawable drawable ) +{ + XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable ); + + if (buffer) { + XMesaSwapBuffers(buffer); + } + else if (_mesa_getenv("MESA_DEBUG")) { + _mesa_warning(NULL, "glXSwapBuffers: invalid drawable 0x%x\n", + (int) drawable); + } +} + + + +/*** GLX_MESA_copy_sub_buffer ***/ + +static void +Fake_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable, + int x, int y, int width, int height ) +{ + XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable ); + if (buffer) { + XMesaCopySubBuffer(buffer, x, y, width, height); + } + else if (_mesa_getenv("MESA_DEBUG")) { + _mesa_warning(NULL, "Mesa: glXCopySubBufferMESA: invalid drawable\n"); + } +} + + +static Bool +Fake_glXQueryVersion( Display *dpy, int *maj, int *min ) +{ + (void) dpy; + /* Return GLX version, not Mesa version */ + assert(CLIENT_MAJOR_VERSION == SERVER_MAJOR_VERSION); + *maj = CLIENT_MAJOR_VERSION; + *min = MIN2( CLIENT_MINOR_VERSION, SERVER_MINOR_VERSION ); + return True; +} + + +/* + * Query the GLX attributes of the given XVisualInfo. + */ +static int +get_config( XMesaVisual xmvis, int attrib, int *value, GLboolean fbconfig ) +{ + ASSERT(xmvis); + switch(attrib) { + case GLX_USE_GL: + if (fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = (int) True; + return 0; + case GLX_BUFFER_SIZE: + *value = xmvis->visinfo->depth; + return 0; + case GLX_LEVEL: + *value = xmvis->mesa_visual.level; + return 0; + case GLX_RGBA: + if (fbconfig) + return GLX_BAD_ATTRIBUTE; + if (xmvis->mesa_visual.rgbMode) { + *value = True; + } + else { + *value = False; + } + return 0; + case GLX_DOUBLEBUFFER: + *value = (int) xmvis->mesa_visual.doubleBufferMode; + return 0; + case GLX_STEREO: + *value = (int) xmvis->mesa_visual.stereoMode; + return 0; + case GLX_AUX_BUFFERS: + *value = xmvis->mesa_visual.numAuxBuffers; + return 0; + case GLX_RED_SIZE: + *value = xmvis->mesa_visual.redBits; + return 0; + case GLX_GREEN_SIZE: + *value = xmvis->mesa_visual.greenBits; + return 0; + case GLX_BLUE_SIZE: + *value = xmvis->mesa_visual.blueBits; + return 0; + case GLX_ALPHA_SIZE: + *value = xmvis->mesa_visual.alphaBits; + return 0; + case GLX_DEPTH_SIZE: + *value = xmvis->mesa_visual.depthBits; + return 0; + case GLX_STENCIL_SIZE: + *value = xmvis->mesa_visual.stencilBits; + return 0; + case GLX_ACCUM_RED_SIZE: + *value = xmvis->mesa_visual.accumRedBits; + return 0; + case GLX_ACCUM_GREEN_SIZE: + *value = xmvis->mesa_visual.accumGreenBits; + return 0; + case GLX_ACCUM_BLUE_SIZE: + *value = xmvis->mesa_visual.accumBlueBits; + return 0; + case GLX_ACCUM_ALPHA_SIZE: + *value = xmvis->mesa_visual.accumAlphaBits; + return 0; + + /* + * GLX_EXT_visual_info extension + */ + case GLX_X_VISUAL_TYPE_EXT: + switch (xmvis->visinfo->CLASS) { + case StaticGray: *value = GLX_STATIC_GRAY_EXT; return 0; + case GrayScale: *value = GLX_GRAY_SCALE_EXT; return 0; + case StaticColor: *value = GLX_STATIC_GRAY_EXT; return 0; + case PseudoColor: *value = GLX_PSEUDO_COLOR_EXT; return 0; + case TrueColor: *value = GLX_TRUE_COLOR_EXT; return 0; + case DirectColor: *value = GLX_DIRECT_COLOR_EXT; return 0; + } + return 0; + case GLX_TRANSPARENT_TYPE_EXT: + if (xmvis->mesa_visual.level==0) { + /* normal planes */ + *value = GLX_NONE_EXT; + } + else if (xmvis->mesa_visual.level>0) { + /* overlay */ + if (xmvis->mesa_visual.rgbMode) { + *value = GLX_TRANSPARENT_RGB_EXT; + } + else { + *value = GLX_TRANSPARENT_INDEX_EXT; + } + } + else if (xmvis->mesa_visual.level<0) { + /* underlay */ + *value = GLX_NONE_EXT; + } + return 0; + case GLX_TRANSPARENT_INDEX_VALUE_EXT: + { + int pixel = transparent_pixel( xmvis ); + if (pixel>=0) { + *value = pixel; + } + /* else undefined */ + } + return 0; + case GLX_TRANSPARENT_RED_VALUE_EXT: + /* undefined */ + return 0; + case GLX_TRANSPARENT_GREEN_VALUE_EXT: + /* undefined */ + return 0; + case GLX_TRANSPARENT_BLUE_VALUE_EXT: + /* undefined */ + return 0; + case GLX_TRANSPARENT_ALPHA_VALUE_EXT: + /* undefined */ + return 0; + + /* + * GLX_EXT_visual_info extension + */ + case GLX_VISUAL_CAVEAT_EXT: + /* test for zero, just in case */ + if (xmvis->mesa_visual.visualRating > 0) + *value = xmvis->mesa_visual.visualRating; + else + *value = GLX_NONE_EXT; + return 0; + + /* + * GLX_ARB_multisample + */ + case GLX_SAMPLE_BUFFERS_ARB: + *value = 0; + return 0; + case GLX_SAMPLES_ARB: + *value = 0; + return 0; + + /* + * For FBConfigs: + */ + case GLX_SCREEN_EXT: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = xmvis->visinfo->screen; + break; + case GLX_DRAWABLE_TYPE: /*SGIX too */ + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = GLX_WINDOW_BIT | GLX_PIXMAP_BIT | GLX_PBUFFER_BIT; + break; + case GLX_RENDER_TYPE_SGIX: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + if (xmvis->mesa_visual.rgbMode) + *value = GLX_RGBA_BIT; + else + *value = GLX_COLOR_INDEX_BIT; + break; + case GLX_X_RENDERABLE_SGIX: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = True; /* XXX really? */ + break; + case GLX_FBCONFIG_ID_SGIX: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = xmvis->visinfo->visualid; + break; + case GLX_MAX_PBUFFER_WIDTH: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + /* XXX or MAX_WIDTH? */ + *value = DisplayWidth(xmvis->display, xmvis->visinfo->screen); + break; + case GLX_MAX_PBUFFER_HEIGHT: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = DisplayHeight(xmvis->display, xmvis->visinfo->screen); + break; + case GLX_MAX_PBUFFER_PIXELS: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = DisplayWidth(xmvis->display, xmvis->visinfo->screen) * + DisplayHeight(xmvis->display, xmvis->visinfo->screen); + break; + case GLX_VISUAL_ID: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = xmvis->visinfo->visualid; + break; + +#ifdef GLX_EXT_texture_from_pixmap + case GLX_BIND_TO_TEXTURE_RGB_EXT: + *value = True; /*XXX*/ + break; + case GLX_BIND_TO_TEXTURE_RGBA_EXT: + /* XXX review */ + *value = xmvis->mesa_visual.alphaBits > 0 ? True : False; + break; + case GLX_BIND_TO_MIPMAP_TEXTURE_EXT: + *value = True; /*XXX*/ + break; + case GLX_BIND_TO_TEXTURE_TARGETS_EXT: + *value = (GLX_TEXTURE_1D_BIT_EXT | + GLX_TEXTURE_2D_BIT_EXT | + GLX_TEXTURE_RECTANGLE_BIT_EXT); /*XXX*/ + break; + case GLX_Y_INVERTED_EXT: + *value = True; /*XXX*/ + break; +#endif + + default: + return GLX_BAD_ATTRIBUTE; + } + return Success; +} + + +static int +Fake_glXGetConfig( Display *dpy, XVisualInfo *visinfo, + int attrib, int *value ) +{ + XMesaVisual xmvis; + int k; + if (!dpy || !visinfo) + return GLX_BAD_ATTRIBUTE; + + xmvis = find_glx_visual( dpy, visinfo ); + if (!xmvis) { + /* this visual wasn't obtained with glXChooseVisual */ + xmvis = create_glx_visual( dpy, visinfo ); + if (!xmvis) { + /* this visual can't be used for GL rendering */ + if (attrib==GLX_USE_GL) { + *value = (int) False; + return 0; + } + else { + return GLX_BAD_VISUAL; + } + } + } + + k = get_config(xmvis, attrib, value, GL_FALSE); + return k; +} + + +static void +Fake_glXWaitGL( void ) +{ + XMesaContext xmesa = XMesaGetCurrentContext(); + XMesaFlush( xmesa ); +} + + + +static void +Fake_glXWaitX( void ) +{ + XMesaContext xmesa = XMesaGetCurrentContext(); + XMesaFlush( xmesa ); +} + + +static const char * +get_extensions( void ) +{ +#ifdef FX + const char *fx = _mesa_getenv("MESA_GLX_FX"); + if (fx && fx[0] != 'd') { + return EXTENSIONS; + } +#endif + return EXTENSIONS + 23; /* skip "GLX_MESA_set_3dfx_mode" */ +} + + + +/* GLX 1.1 and later */ +static const char * +Fake_glXQueryExtensionsString( Display *dpy, int screen ) +{ + (void) dpy; + (void) screen; + return get_extensions(); +} + + + +/* GLX 1.1 and later */ +static const char * +Fake_glXQueryServerString( Display *dpy, int screen, int name ) +{ + static char version[1000]; + _mesa_sprintf(version, "%d.%d %s", + SERVER_MAJOR_VERSION, SERVER_MINOR_VERSION, MESA_GLX_VERSION); + + (void) dpy; + (void) screen; + + switch (name) { + case GLX_EXTENSIONS: + return get_extensions(); + case GLX_VENDOR: + return VENDOR; + case GLX_VERSION: + return version; + default: + return NULL; + } +} + + + +/* GLX 1.1 and later */ +static const char * +Fake_glXGetClientString( Display *dpy, int name ) +{ + static char version[1000]; + _mesa_sprintf(version, "%d.%d %s", CLIENT_MAJOR_VERSION, + CLIENT_MINOR_VERSION, MESA_GLX_VERSION); + + (void) dpy; + + switch (name) { + case GLX_EXTENSIONS: + return get_extensions(); + case GLX_VENDOR: + return VENDOR; + case GLX_VERSION: + return version; + default: + return NULL; + } +} + + + +/* + * GLX 1.3 and later + */ + + +static int +Fake_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config, + int attribute, int *value ) +{ + XMesaVisual v = (XMesaVisual) config; + (void) dpy; + (void) config; + + if (!dpy || !config || !value) + return -1; + + return get_config(v, attribute, value, GL_TRUE); +} + + +static GLXFBConfig * +Fake_glXGetFBConfigs( Display *dpy, int screen, int *nelements ) +{ + XVisualInfo *visuals, visTemplate; + const long visMask = VisualScreenMask; + int i; + + /* Get list of all X visuals */ + visTemplate.screen = screen; + visuals = XGetVisualInfo(dpy, visMask, &visTemplate, nelements); + if (*nelements > 0) { + XMesaVisual *results; + results = (XMesaVisual *) _mesa_malloc(*nelements * sizeof(XMesaVisual)); + if (!results) { + *nelements = 0; + return NULL; + } + for (i = 0; i < *nelements; i++) { + results[i] = create_glx_visual(dpy, visuals + i); + } + return (GLXFBConfig *) results; + } + return NULL; +} + + +static GLXFBConfig * +Fake_glXChooseFBConfig( Display *dpy, int screen, + const int *attribList, int *nitems ) +{ + XMesaVisual xmvis; + + if (!attribList || !attribList[0]) { + /* return list of all configs (per GLX_SGIX_fbconfig spec) */ + return Fake_glXGetFBConfigs(dpy, screen, nitems); + } + + xmvis = choose_visual(dpy, screen, attribList, GL_TRUE); + if (xmvis) { + GLXFBConfig *config = (GLXFBConfig *) _mesa_malloc(sizeof(XMesaVisual)); + if (!config) { + *nitems = 0; + return NULL; + } + *nitems = 1; + config[0] = (GLXFBConfig) xmvis; + return (GLXFBConfig *) config; + } + else { + *nitems = 0; + return NULL; + } +} + + +static XVisualInfo * +Fake_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config ) +{ + if (dpy && config) { + XMesaVisual xmvis = (XMesaVisual) config; +#if 0 + return xmvis->vishandle; +#else + /* create a new vishandle - the cached one may be stale */ + xmvis->vishandle = (XVisualInfo *) _mesa_malloc(sizeof(XVisualInfo)); + if (xmvis->vishandle) { + _mesa_memcpy(xmvis->vishandle, xmvis->visinfo, sizeof(XVisualInfo)); + } + return xmvis->vishandle; +#endif + } + else { + return NULL; + } +} + + +static GLXWindow +Fake_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win, + const int *attribList ) +{ + XMesaVisual xmvis = (XMesaVisual) config; + XMesaBuffer xmbuf; + if (!xmvis) + return 0; + + xmbuf = XMesaCreateWindowBuffer(xmvis, win); + if (!xmbuf) + return 0; + +#ifdef FX + /* XXX this will segfault if actually called */ + FXcreateContext(xmvis, win, NULL, xmbuf); +#endif + + (void) dpy; + (void) attribList; /* Ignored in GLX 1.3 */ + + return win; /* A hack for now */ +} + + +static void +Fake_glXDestroyWindow( Display *dpy, GLXWindow window ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, (XMesaDrawable) window); + if (b) + XMesaDestroyBuffer(b); + /* don't destroy X window */ +} + + +/* XXX untested */ +static GLXPixmap +Fake_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, + const int *attribList ) +{ + XMesaVisual v = (XMesaVisual) config; + XMesaBuffer b; + const int *attr; + int target = 0, format = 0, mipmap = 0; + int value; + + if (!dpy || !config || !pixmap) + return 0; + + for (attr = attribList; *attr; attr++) { + switch (*attr) { + case GLX_TEXTURE_FORMAT_EXT: + attr++; + switch (*attr) { + case GLX_TEXTURE_FORMAT_NONE_EXT: + case GLX_TEXTURE_FORMAT_RGB_EXT: + case GLX_TEXTURE_FORMAT_RGBA_EXT: + format = *attr; + break; + default: + /* error */ + return 0; + } + break; + case GLX_TEXTURE_TARGET_EXT: + attr++; + switch (*attr) { + case GLX_TEXTURE_1D_EXT: + case GLX_TEXTURE_2D_EXT: + case GLX_TEXTURE_RECTANGLE_EXT: + target = *attr; + break; + default: + /* error */ + return 0; + } + break; + case GLX_MIPMAP_TEXTURE_EXT: + attr++; + if (*attr) + mipmap = 1; + break; + default: + /* error */ + return 0; + } + } + + if (format == GLX_TEXTURE_FORMAT_RGB_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_RGB_EXT, + &value, GL_TRUE) != Success + || !value) { + return 0; /* error! */ + } + } + else if (format == GLX_TEXTURE_FORMAT_RGBA_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_RGBA_EXT, + &value, GL_TRUE) != Success + || !value) { + return 0; /* error! */ + } + } + if (mipmap) { + if (get_config(v, GLX_BIND_TO_MIPMAP_TEXTURE_EXT, + &value, GL_TRUE) != Success + || !value) { + return 0; /* error! */ + } + } + if (target == GLX_TEXTURE_1D_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT, + &value, GL_TRUE) != Success + || (value & GLX_TEXTURE_1D_BIT_EXT) == 0) { + return 0; /* error! */ + } + } + else if (target == GLX_TEXTURE_2D_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT, + &value, GL_TRUE) != Success + || (value & GLX_TEXTURE_2D_BIT_EXT) == 0) { + return 0; /* error! */ + } + } + if (target == GLX_TEXTURE_RECTANGLE_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT, + &value, GL_TRUE) != Success + || (value & GLX_TEXTURE_RECTANGLE_BIT_EXT) == 0) { + return 0; /* error! */ + } + } + + if (format || target || mipmap) { + /* texture from pixmap */ + b = XMesaCreatePixmapTextureBuffer(v, pixmap, 0, format, target, mipmap); + } + else { + b = XMesaCreatePixmapBuffer( v, pixmap, 0 ); + } + if (!b) { + return 0; + } + + return pixmap; +} + + +static void +Fake_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, (XMesaDrawable)pixmap); + if (b) + XMesaDestroyBuffer(b); + /* don't destroy X pixmap */ +} + + +static GLXPbuffer +Fake_glXCreatePbuffer( Display *dpy, GLXFBConfig config, + const int *attribList ) +{ + XMesaVisual xmvis = (XMesaVisual) config; + XMesaBuffer xmbuf; + const int *attrib; + int width = 0, height = 0; + GLboolean useLargest = GL_FALSE, preserveContents = GL_FALSE; + + (void) dpy; + + for (attrib = attribList; *attrib; attrib++) { + switch (*attrib) { + case GLX_PBUFFER_WIDTH: + attrib++; + width = *attrib; + break; + case GLX_PBUFFER_HEIGHT: + attrib++; + height = *attrib; + break; + case GLX_PRESERVED_CONTENTS: + attrib++; + preserveContents = *attrib; /* ignored */ + break; + case GLX_LARGEST_PBUFFER: + attrib++; + useLargest = *attrib; /* ignored */ + break; + default: + return 0; + } + } + + /* not used at this time */ + (void) useLargest; + (void) preserveContents; + + if (width == 0 || height == 0) + return 0; + + xmbuf = XMesaCreatePBuffer( xmvis, 0, width, height); + /* A GLXPbuffer handle must be an X Drawable because that's what + * glXMakeCurrent takes. + */ + if (xmbuf) + return (GLXPbuffer) xmbuf->frontxrb->pixmap; + else + return 0; +} + + +static void +Fake_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, pbuf); + if (b) { + XMesaDestroyBuffer(b); + } +} + + +static void +Fake_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, + unsigned int *value ) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, draw); + if (!xmbuf) + return; + + switch (attribute) { + case GLX_WIDTH: + *value = xmbuf->mesa_buffer.Width; + break; + case GLX_HEIGHT: + *value = xmbuf->mesa_buffer.Height; + break; + case GLX_PRESERVED_CONTENTS: + *value = True; + break; + case GLX_LARGEST_PBUFFER: + *value = xmbuf->mesa_buffer.Width * xmbuf->mesa_buffer.Height; + break; + case GLX_FBCONFIG_ID: + *value = xmbuf->xm_visual->visinfo->visualid; + return; +#ifdef GLX_EXT_texture_from_pixmap + case GLX_TEXTURE_FORMAT_EXT: + *value = xmbuf->TextureFormat; + break; + case GLX_TEXTURE_TARGET_EXT: + *value = xmbuf->TextureTarget; + break; + case GLX_MIPMAP_TEXTURE_EXT: + *value = xmbuf->TextureMipmap; + break; +#endif + + default: + return; /* raise BadValue error */ + } +} + + +static GLXContext +Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config, + int renderType, GLXContext shareList, Bool direct ) +{ + struct fake_glx_context *glxCtx; + struct fake_glx_context *shareCtx = (struct fake_glx_context *) shareList; + XMesaVisual xmvis = (XMesaVisual) config; + + if (!dpy || !config || + (renderType != GLX_RGBA_TYPE && renderType != GLX_COLOR_INDEX_TYPE)) + return 0; + + glxCtx = CALLOC_STRUCT(fake_glx_context); + if (!glxCtx) + return 0; + + /* deallocate unused windows/buffers */ + XMesaGarbageCollect(); + + glxCtx->xmesaContext = XMesaCreateContext(xmvis, + shareCtx ? shareCtx->xmesaContext : NULL); + if (!glxCtx->xmesaContext) { + _mesa_free(glxCtx); + return NULL; + } + + glxCtx->xmesaContext->direct = GL_FALSE; + glxCtx->glxContext.isDirect = GL_FALSE; + glxCtx->glxContext.currentDpy = dpy; + glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ + + assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); + + return (GLXContext) glxCtx; +} + + +static int +Fake_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value ) +{ + struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + XMesaContext xmctx = glxCtx->xmesaContext; + + (void) dpy; + (void) ctx; + + switch (attribute) { + case GLX_FBCONFIG_ID: + *value = xmctx->xm_visual->visinfo->visualid; + break; + case GLX_RENDER_TYPE: + if (xmctx->xm_visual->mesa_visual.rgbMode) + *value = GLX_RGBA_BIT; + else + *value = GLX_COLOR_INDEX_BIT; + break; + case GLX_SCREEN: + *value = 0; + return Success; + default: + return GLX_BAD_ATTRIBUTE; + } + return 0; +} + + +static void +Fake_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask ) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); + if (xmbuf) + xmbuf->selectedEvents = mask; +} + + +static void +Fake_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable, + unsigned long *mask ) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); + if (xmbuf) + *mask = xmbuf->selectedEvents; + else + *mask = 0; +} + + + +/*** GLX_SGI_swap_control ***/ + +static int +Fake_glXSwapIntervalSGI(int interval) +{ + (void) interval; + return 0; +} + + + +/*** GLX_SGI_video_sync ***/ + +static unsigned int FrameCounter = 0; + +static int +Fake_glXGetVideoSyncSGI(unsigned int *count) +{ + /* this is a bogus implementation */ + *count = FrameCounter++; + return 0; +} + +static int +Fake_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) +{ + if (divisor <= 0 || remainder < 0) + return GLX_BAD_VALUE; + /* this is a bogus implementation */ + FrameCounter++; + while (FrameCounter % divisor != remainder) + FrameCounter++; + *count = FrameCounter; + return 0; +} + + + +/*** GLX_SGI_make_current_read ***/ + +static Bool +Fake_glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) +{ + return Fake_glXMakeContextCurrent( dpy, draw, read, ctx ); +} + +/* not used +static GLXDrawable +Fake_glXGetCurrentReadDrawableSGI(void) +{ + return 0; +} +*/ + + +/*** GLX_SGIX_video_source ***/ +#if defined(_VL_H) + +static GLXVideoSourceSGIX +Fake_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode) +{ + (void) dpy; + (void) screen; + (void) server; + (void) path; + (void) nodeClass; + (void) drainNode; + return 0; +} + +static void +Fake_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) +{ + (void) dpy; + (void) src; +} + +#endif + + +/*** GLX_EXT_import_context ***/ + +static void +Fake_glXFreeContextEXT(Display *dpy, GLXContext context) +{ + (void) dpy; + (void) context; +} + +static GLXContextID +Fake_glXGetContextIDEXT(const GLXContext context) +{ + (void) context; + return 0; +} + +static GLXContext +Fake_glXImportContextEXT(Display *dpy, GLXContextID contextID) +{ + (void) dpy; + (void) contextID; + return 0; +} + +static int +Fake_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value) +{ + (void) dpy; + (void) context; + (void) attribute; + (void) value; + return 0; +} + + + +/*** GLX_SGIX_fbconfig ***/ + +static int +Fake_glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value) +{ + return Fake_glXGetFBConfigAttrib(dpy, config, attribute, value); +} + +static GLXFBConfigSGIX * +Fake_glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements) +{ + return (GLXFBConfig *) Fake_glXChooseFBConfig(dpy, screen, attrib_list, nelements); +} + + +static GLXPixmap +Fake_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap) +{ + XMesaVisual xmvis = (XMesaVisual) config; + XMesaBuffer xmbuf = XMesaCreatePixmapBuffer(xmvis, pixmap, 0); + return xmbuf->frontxrb->pixmap; /* need to return an X ID */ +} + + +static GLXContext +Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct) +{ + XMesaVisual xmvis = (XMesaVisual) config; + struct fake_glx_context *glxCtx; + struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list; + + glxCtx = CALLOC_STRUCT(fake_glx_context); + if (!glxCtx) + return 0; + + /* deallocate unused windows/buffers */ + XMesaGarbageCollect(); + + glxCtx->xmesaContext = XMesaCreateContext(xmvis, + shareCtx ? shareCtx->xmesaContext : NULL); + if (!glxCtx->xmesaContext) { + _mesa_free(glxCtx); + return NULL; + } + + glxCtx->xmesaContext->direct = GL_FALSE; + glxCtx->glxContext.isDirect = GL_FALSE; + glxCtx->glxContext.currentDpy = dpy; + glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ + + assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); + + return (GLXContext) glxCtx; +} + + +static XVisualInfo * +Fake_glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config) +{ + return Fake_glXGetVisualFromFBConfig(dpy, config); +} + + +static GLXFBConfigSGIX +Fake_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) +{ + XMesaVisual xmvis = find_glx_visual(dpy, vis); + if (!xmvis) { + /* This visual wasn't found with glXChooseVisual() */ + xmvis = create_glx_visual(dpy, vis); + } + + return (GLXFBConfigSGIX) xmvis; +} + + + +/*** GLX_SGIX_pbuffer ***/ + +static GLXPbufferSGIX +Fake_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, + unsigned int width, unsigned int height, + int *attribList) +{ + XMesaVisual xmvis = (XMesaVisual) config; + XMesaBuffer xmbuf; + const int *attrib; + GLboolean useLargest = GL_FALSE, preserveContents = GL_FALSE; + + (void) dpy; + + for (attrib = attribList; attrib && *attrib; attrib++) { + switch (*attrib) { + case GLX_PRESERVED_CONTENTS_SGIX: + attrib++; + preserveContents = *attrib; /* ignored */ + break; + case GLX_LARGEST_PBUFFER_SGIX: + attrib++; + useLargest = *attrib; /* ignored */ + break; + default: + return 0; + } + } + + /* not used at this time */ + (void) useLargest; + (void) preserveContents; + + xmbuf = XMesaCreatePBuffer( xmvis, 0, width, height); + /* A GLXPbuffer handle must be an X Drawable because that's what + * glXMakeCurrent takes. + */ + return (GLXPbuffer) xmbuf->frontxrb->pixmap; +} + + +static void +Fake_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf); + if (xmbuf) { + XMesaDestroyBuffer(xmbuf); + } +} + + +static int +Fake_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value) +{ + const XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf); + + if (!xmbuf) { + /* Generate GLXBadPbufferSGIX for bad pbuffer */ + return 0; + } + + switch (attribute) { + case GLX_PRESERVED_CONTENTS_SGIX: + *value = True; + break; + case GLX_LARGEST_PBUFFER_SGIX: + *value = xmbuf->mesa_buffer.Width * xmbuf->mesa_buffer.Height; + break; + case GLX_WIDTH_SGIX: + *value = xmbuf->mesa_buffer.Width; + break; + case GLX_HEIGHT_SGIX: + *value = xmbuf->mesa_buffer.Height; + break; + case GLX_EVENT_MASK_SGIX: + *value = 0; /* XXX might be wrong */ + break; + default: + *value = 0; + } + return 0; +} + + +static void +Fake_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); + if (xmbuf) { + /* Note: we'll never generate clobber events */ + xmbuf->selectedEvents = mask; + } +} + + +static void +Fake_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); + if (xmbuf) { + *mask = xmbuf->selectedEvents; + } + else { + *mask = 0; + } +} + + + +/*** GLX_SGI_cushion ***/ + +static void +Fake_glXCushionSGI(Display *dpy, Window win, float cushion) +{ + (void) dpy; + (void) win; + (void) cushion; +} + + + +/*** GLX_SGIX_video_resize ***/ + +static int +Fake_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) window; + return 0; +} + +static int +Fake_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) x; + (void) y; + (void) w; + (void) h; + return 0; +} + +static int +Fake_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) x; + (void) y; + (void) w; + (void) h; + return 0; +} + +static int +Fake_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) dx; + (void) dy; + (void) dw; + (void) dh; + return 0; +} + +static int +Fake_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) synctype; + return 0; +} + + + +/*** GLX_SGIX_dmbuffer **/ + +#if defined(_DM_BUFFER_H_) +static Bool +Fake_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer) +{ + (void) dpy; + (void) pbuffer; + (void) params; + (void) dmbuffer; + return False; +} +#endif + + +/*** GLX_SGIX_swap_group ***/ + +static void +Fake_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member) +{ + (void) dpy; + (void) drawable; + (void) member; +} + + + +/*** GLX_SGIX_swap_barrier ***/ + +static void +Fake_glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier) +{ + (void) dpy; + (void) drawable; + (void) barrier; +} + +static Bool +Fake_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) +{ + (void) dpy; + (void) screen; + (void) max; + return False; +} + + + +/*** GLX_SUN_get_transparent_index ***/ + +static Status +Fake_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent) +{ + (void) dpy; + (void) overlay; + (void) underlay; + (void) pTransparent; + return 0; +} + + + +/*** GLX_MESA_release_buffers ***/ + +/* + * Release the depth, stencil, accum buffers attached to a GLXDrawable + * (a window or pixmap) prior to destroying the GLXDrawable. + */ +static Bool +Fake_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, d); + if (b) { + XMesaDestroyBuffer(b); + return True; + } + return False; +} + + + +/*** GLX_MESA_set_3dfx_mode ***/ + +static Bool +Fake_glXSet3DfxModeMESA( int mode ) +{ + return XMesaSetFXmode( mode ); +} + + + +/*** GLX_NV_vertex_array range ***/ +static void * +Fake_glXAllocateMemoryNV( GLsizei size, + GLfloat readFrequency, + GLfloat writeFrequency, + GLfloat priority ) +{ + (void) size; + (void) readFrequency; + (void) writeFrequency; + (void) priority; + return NULL; +} + + +static void +Fake_glXFreeMemoryNV( GLvoid *pointer ) +{ + (void) pointer; +} + + +/*** GLX_MESA_agp_offset ***/ + +static GLuint +Fake_glXGetAGPOffsetMESA( const GLvoid *pointer ) +{ + (void) pointer; + return ~0; +} + + +/*** GLX_EXT_texture_from_pixmap ***/ + +static void +Fake_glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, + const int *attrib_list) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, drawable); + if (b) + XMesaBindTexImage(dpy, b, buffer, attrib_list); +} + +static void +Fake_glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, drawable); + if (b) + XMesaReleaseTexImage(dpy, b, buffer); +} + + +/* silence warning */ +extern struct _glxapi_table *_mesa_GetGLXDispatchTable(void); + + +/** + * Create a new GLX API dispatch table with its function pointers + * initialized to point to Mesa's "fake" GLX API functions. + * Note: there's a similar function (_real_GetGLXDispatchTable) that + * returns a new dispatch table with all pointers initalized to point + * to "real" GLX functions (which understand GLX wire protocol, etc). + */ +struct _glxapi_table * +_mesa_GetGLXDispatchTable(void) +{ + static struct _glxapi_table glx; + + /* be sure our dispatch table size <= libGL's table */ + { + GLuint size = sizeof(struct _glxapi_table) / sizeof(void *); + (void) size; + assert(_glxapi_get_dispatch_table_size() >= size); + } + + /* initialize the whole table to no-ops */ + _glxapi_set_no_op_table(&glx); + + /* now initialize the table with the functions I implement */ + glx.ChooseVisual = Fake_glXChooseVisual; + glx.CopyContext = Fake_glXCopyContext; + glx.CreateContext = Fake_glXCreateContext; + glx.CreateGLXPixmap = Fake_glXCreateGLXPixmap; + glx.DestroyContext = Fake_glXDestroyContext; + glx.DestroyGLXPixmap = Fake_glXDestroyGLXPixmap; + glx.GetConfig = Fake_glXGetConfig; + /*glx.GetCurrentContext = Fake_glXGetCurrentContext;*/ + /*glx.GetCurrentDrawable = Fake_glXGetCurrentDrawable;*/ + glx.IsDirect = Fake_glXIsDirect; + glx.MakeCurrent = Fake_glXMakeCurrent; + glx.QueryExtension = Fake_glXQueryExtension; + glx.QueryVersion = Fake_glXQueryVersion; + glx.SwapBuffers = Fake_glXSwapBuffers; + glx.UseXFont = Fake_glXUseXFont; + glx.WaitGL = Fake_glXWaitGL; + glx.WaitX = Fake_glXWaitX; + + /*** GLX_VERSION_1_1 ***/ + glx.GetClientString = Fake_glXGetClientString; + glx.QueryExtensionsString = Fake_glXQueryExtensionsString; + glx.QueryServerString = Fake_glXQueryServerString; + + /*** GLX_VERSION_1_2 ***/ + /*glx.GetCurrentDisplay = Fake_glXGetCurrentDisplay;*/ + + /*** GLX_VERSION_1_3 ***/ + glx.ChooseFBConfig = Fake_glXChooseFBConfig; + glx.CreateNewContext = Fake_glXCreateNewContext; + glx.CreatePbuffer = Fake_glXCreatePbuffer; + glx.CreatePixmap = Fake_glXCreatePixmap; + glx.CreateWindow = Fake_glXCreateWindow; + glx.DestroyPbuffer = Fake_glXDestroyPbuffer; + glx.DestroyPixmap = Fake_glXDestroyPixmap; + glx.DestroyWindow = Fake_glXDestroyWindow; + /*glx.GetCurrentReadDrawable = Fake_glXGetCurrentReadDrawable;*/ + glx.GetFBConfigAttrib = Fake_glXGetFBConfigAttrib; + glx.GetFBConfigs = Fake_glXGetFBConfigs; + glx.GetSelectedEvent = Fake_glXGetSelectedEvent; + glx.GetVisualFromFBConfig = Fake_glXGetVisualFromFBConfig; + glx.MakeContextCurrent = Fake_glXMakeContextCurrent; + glx.QueryContext = Fake_glXQueryContext; + glx.QueryDrawable = Fake_glXQueryDrawable; + glx.SelectEvent = Fake_glXSelectEvent; + + /*** GLX_SGI_swap_control ***/ + glx.SwapIntervalSGI = Fake_glXSwapIntervalSGI; + + /*** GLX_SGI_video_sync ***/ + glx.GetVideoSyncSGI = Fake_glXGetVideoSyncSGI; + glx.WaitVideoSyncSGI = Fake_glXWaitVideoSyncSGI; + + /*** GLX_SGI_make_current_read ***/ + glx.MakeCurrentReadSGI = Fake_glXMakeCurrentReadSGI; + /*glx.GetCurrentReadDrawableSGI = Fake_glXGetCurrentReadDrawableSGI;*/ + +/*** GLX_SGIX_video_source ***/ +#if defined(_VL_H) + glx.CreateGLXVideoSourceSGIX = Fake_glXCreateGLXVideoSourceSGIX; + glx.DestroyGLXVideoSourceSGIX = Fake_glXDestroyGLXVideoSourceSGIX; +#endif + + /*** GLX_EXT_import_context ***/ + glx.FreeContextEXT = Fake_glXFreeContextEXT; + glx.GetContextIDEXT = Fake_glXGetContextIDEXT; + /*glx.GetCurrentDisplayEXT = Fake_glXGetCurrentDisplayEXT;*/ + glx.ImportContextEXT = Fake_glXImportContextEXT; + glx.QueryContextInfoEXT = Fake_glXQueryContextInfoEXT; + + /*** GLX_SGIX_fbconfig ***/ + glx.GetFBConfigAttribSGIX = Fake_glXGetFBConfigAttribSGIX; + glx.ChooseFBConfigSGIX = Fake_glXChooseFBConfigSGIX; + glx.CreateGLXPixmapWithConfigSGIX = Fake_glXCreateGLXPixmapWithConfigSGIX; + glx.CreateContextWithConfigSGIX = Fake_glXCreateContextWithConfigSGIX; + glx.GetVisualFromFBConfigSGIX = Fake_glXGetVisualFromFBConfigSGIX; + glx.GetFBConfigFromVisualSGIX = Fake_glXGetFBConfigFromVisualSGIX; + + /*** GLX_SGIX_pbuffer ***/ + glx.CreateGLXPbufferSGIX = Fake_glXCreateGLXPbufferSGIX; + glx.DestroyGLXPbufferSGIX = Fake_glXDestroyGLXPbufferSGIX; + glx.QueryGLXPbufferSGIX = Fake_glXQueryGLXPbufferSGIX; + glx.SelectEventSGIX = Fake_glXSelectEventSGIX; + glx.GetSelectedEventSGIX = Fake_glXGetSelectedEventSGIX; + + /*** GLX_SGI_cushion ***/ + glx.CushionSGI = Fake_glXCushionSGI; + + /*** GLX_SGIX_video_resize ***/ + glx.BindChannelToWindowSGIX = Fake_glXBindChannelToWindowSGIX; + glx.ChannelRectSGIX = Fake_glXChannelRectSGIX; + glx.QueryChannelRectSGIX = Fake_glXQueryChannelRectSGIX; + glx.QueryChannelDeltasSGIX = Fake_glXQueryChannelDeltasSGIX; + glx.ChannelRectSyncSGIX = Fake_glXChannelRectSyncSGIX; + + /*** GLX_SGIX_dmbuffer **/ +#if defined(_DM_BUFFER_H_) + glx.AssociateDMPbufferSGIX = NULL; +#endif + + /*** GLX_SGIX_swap_group ***/ + glx.JoinSwapGroupSGIX = Fake_glXJoinSwapGroupSGIX; + + /*** GLX_SGIX_swap_barrier ***/ + glx.BindSwapBarrierSGIX = Fake_glXBindSwapBarrierSGIX; + glx.QueryMaxSwapBarriersSGIX = Fake_glXQueryMaxSwapBarriersSGIX; + + /*** GLX_SUN_get_transparent_index ***/ + glx.GetTransparentIndexSUN = Fake_glXGetTransparentIndexSUN; + + /*** GLX_MESA_copy_sub_buffer ***/ + glx.CopySubBufferMESA = Fake_glXCopySubBufferMESA; + + /*** GLX_MESA_release_buffers ***/ + glx.ReleaseBuffersMESA = Fake_glXReleaseBuffersMESA; + + /*** GLX_MESA_pixmap_colormap ***/ + glx.CreateGLXPixmapMESA = Fake_glXCreateGLXPixmapMESA; + + /*** GLX_MESA_set_3dfx_mode ***/ + glx.Set3DfxModeMESA = Fake_glXSet3DfxModeMESA; + + /*** GLX_NV_vertex_array_range ***/ + glx.AllocateMemoryNV = Fake_glXAllocateMemoryNV; + glx.FreeMemoryNV = Fake_glXFreeMemoryNV; + + /*** GLX_MESA_agp_offset ***/ + glx.GetAGPOffsetMESA = Fake_glXGetAGPOffsetMESA; + + /*** GLX_EXT_texture_from_pixmap ***/ + glx.BindTexImageEXT = Fake_glXBindTexImageEXT; + glx.ReleaseTexImageEXT = Fake_glXReleaseTexImageEXT; + + return &glx; +} diff --git a/src/mesa/pipe/xlib/glxapi.c b/src/mesa/pipe/xlib/glxapi.c new file mode 100644 index 0000000000..309a0008d7 --- /dev/null +++ b/src/mesa/pipe/xlib/glxapi.c @@ -0,0 +1,1408 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* + * This is the GLX API dispatcher. Calls to the glX* functions are + * either routed to the real GLX encoders or to Mesa's pseudo-GLX functions. + * See the glxapi.h file for more details. + */ + + +#include +#include +#include +#include +#include "glheader.h" +#include "glapi.h" +#include "glxapi.h" + + +extern struct _glxapi_table *_real_GetGLXDispatchTable(void); +extern struct _glxapi_table *_mesa_GetGLXDispatchTable(void); + + +struct display_dispatch { + Display *Dpy; + struct _glxapi_table *Table; + struct display_dispatch *Next; +}; + +static struct display_dispatch *DispatchList = NULL; + + +/* Display -> Dispatch caching */ +static Display *prevDisplay = NULL; +static struct _glxapi_table *prevTable = NULL; + + +static struct _glxapi_table * +get_dispatch(Display *dpy) +{ + if (!dpy) + return NULL; + + /* search list of display/dispatch pairs for this display */ + { + const struct display_dispatch *d = DispatchList; + while (d) { + if (d->Dpy == dpy) { + prevDisplay = dpy; + prevTable = d->Table; + return d->Table; /* done! */ + } + d = d->Next; + } + } + + /* A new display, determine if we should use real GLX + * or Mesa's pseudo-GLX. + */ + { + struct _glxapi_table *t = _mesa_GetGLXDispatchTable(); + + if (t) { + struct display_dispatch *d; + d = (struct display_dispatch *) malloc(sizeof(struct display_dispatch)); + if (d) { + d->Dpy = dpy; + d->Table = t; + /* insert at head of list */ + d->Next = DispatchList; + DispatchList = d; + /* update cache */ + prevDisplay = dpy; + prevTable = t; + return t; + } + } + } + + /* If we get here that means we can't use real GLX on this display + * and the Mesa pseudo-GLX software renderer wasn't compiled in. + * Or, we ran out of memory! + */ + return NULL; +} + + +/* Don't use the GET_DISPATCH defined in glthread.h */ +#undef GET_DISPATCH + +#define GET_DISPATCH(DPY, TABLE) \ + if (DPY == prevDisplay) { \ + TABLE = prevTable; \ + } \ + else if (!DPY) { \ + TABLE = NULL; \ + } \ + else { \ + TABLE = get_dispatch(DPY); \ + } + + + + +/** + * GLX API current context. + */ +#if defined(GLX_USE_TLS) +PUBLIC __thread void * CurrentContext + __attribute__((tls_model("initial-exec"))); +#elif defined(THREADS) +static _glthread_TSD ContextTSD; /**< Per-thread context pointer */ +#else +static GLXContext CurrentContext = 0; +#endif + + +static void +SetCurrentContext(GLXContext c) +{ +#if defined(GLX_USE_TLS) + CurrentContext = c; +#elif defined(THREADS) + _glthread_SetTSD(&ContextTSD, c); +#else + CurrentContext = c; +#endif +} + + +/* + * GLX API entrypoints + */ + +/*** GLX_VERSION_1_0 ***/ + +XVisualInfo PUBLIC * +glXChooseVisual(Display *dpy, int screen, int *list) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->ChooseVisual)(dpy, screen, list); +} + + +void PUBLIC +glXCopyContext(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->CopyContext)(dpy, src, dst, mask); +} + + +GLXContext PUBLIC +glXCreateContext(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateContext)(dpy, visinfo, shareList, direct); +} + + +GLXPixmap PUBLIC +glXCreateGLXPixmap(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXPixmap)(dpy, visinfo, pixmap); +} + + +void PUBLIC +glXDestroyContext(Display *dpy, GLXContext ctx) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + if (glXGetCurrentContext() == ctx) + SetCurrentContext(NULL); + (t->DestroyContext)(dpy, ctx); +} + + +void PUBLIC +glXDestroyGLXPixmap(Display *dpy, GLXPixmap pixmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyGLXPixmap)(dpy, pixmap); +} + + +int PUBLIC +glXGetConfig(Display *dpy, XVisualInfo *visinfo, int attrib, int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return GLX_NO_EXTENSION; + return (t->GetConfig)(dpy, visinfo, attrib, value); +} + + +GLXContext PUBLIC +glXGetCurrentContext(void) +{ +#if defined(GLX_USE_TLS) + return CurrentContext; +#elif defined(THREADS) + return (GLXContext) _glthread_GetTSD(&ContextTSD); +#else + return CurrentContext; +#endif +} + + +GLXDrawable PUBLIC +glXGetCurrentDrawable(void) +{ + __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); + return gc ? gc->currentDrawable : 0; +} + + +Bool PUBLIC +glXIsDirect(Display *dpy, GLXContext ctx) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->IsDirect)(dpy, ctx); +} + + +Bool PUBLIC +glXMakeCurrent(Display *dpy, GLXDrawable drawable, GLXContext ctx) +{ + Bool b; + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) { + return False; + } + b = (*t->MakeCurrent)(dpy, drawable, ctx); + if (b) { + SetCurrentContext(ctx); + } + return b; +} + + +Bool PUBLIC +glXQueryExtension(Display *dpy, int *errorb, int *event) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->QueryExtension)(dpy, errorb, event); +} + + +Bool PUBLIC +glXQueryVersion(Display *dpy, int *maj, int *min) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->QueryVersion)(dpy, maj, min); +} + + +void PUBLIC +glXSwapBuffers(Display *dpy, GLXDrawable drawable) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->SwapBuffers)(dpy, drawable); +} + + +void PUBLIC +glXUseXFont(Font font, int first, int count, int listBase) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->UseXFont)(font, first, count, listBase); +} + + +void PUBLIC +glXWaitGL(void) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->WaitGL)(); +} + + +void PUBLIC +glXWaitX(void) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->WaitX)(); +} + + + +/*** GLX_VERSION_1_1 ***/ + +const char PUBLIC * +glXGetClientString(Display *dpy, int name) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->GetClientString)(dpy, name); +} + + +const char PUBLIC * +glXQueryExtensionsString(Display *dpy, int screen) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->QueryExtensionsString)(dpy, screen); +} + + +const char PUBLIC * +glXQueryServerString(Display *dpy, int screen, int name) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->QueryServerString)(dpy, screen, name); +} + + +/*** GLX_VERSION_1_2 ***/ + +Display PUBLIC * +glXGetCurrentDisplay(void) +{ + /* Same code as in libGL's glxext.c */ + __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); + if (NULL == gc) return NULL; + return gc->currentDpy; +} + + + +/*** GLX_VERSION_1_3 ***/ + +GLXFBConfig PUBLIC * +glXChooseFBConfig(Display *dpy, int screen, const int *attribList, int *nitems) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ChooseFBConfig)(dpy, screen, attribList, nitems); +} + + +GLXContext PUBLIC +glXCreateNewContext(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateNewContext)(dpy, config, renderType, shareList, direct); +} + + +GLXPbuffer PUBLIC +glXCreatePbuffer(Display *dpy, GLXFBConfig config, const int *attribList) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreatePbuffer)(dpy, config, attribList); +} + + +GLXPixmap PUBLIC +glXCreatePixmap(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreatePixmap)(dpy, config, pixmap, attribList); +} + + +GLXWindow PUBLIC +glXCreateWindow(Display *dpy, GLXFBConfig config, Window win, const int *attribList) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateWindow)(dpy, config, win, attribList); +} + + +void PUBLIC +glXDestroyPbuffer(Display *dpy, GLXPbuffer pbuf) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyPbuffer)(dpy, pbuf); +} + + +void PUBLIC +glXDestroyPixmap(Display *dpy, GLXPixmap pixmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyPixmap)(dpy, pixmap); +} + + +void PUBLIC +glXDestroyWindow(Display *dpy, GLXWindow window) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyWindow)(dpy, window); +} + + +GLXDrawable PUBLIC +glXGetCurrentReadDrawable(void) +{ + __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); + return gc ? gc->currentReadable : 0; +} + + +int PUBLIC +glXGetFBConfigAttrib(Display *dpy, GLXFBConfig config, int attribute, int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return GLX_NO_EXTENSION; + return (t->GetFBConfigAttrib)(dpy, config, attribute, value); +} + + +GLXFBConfig PUBLIC * +glXGetFBConfigs(Display *dpy, int screen, int *nelements) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->GetFBConfigs)(dpy, screen, nelements); +} + +void PUBLIC +glXGetSelectedEvent(Display *dpy, GLXDrawable drawable, unsigned long *mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->GetSelectedEvent)(dpy, drawable, mask); +} + + +XVisualInfo PUBLIC * +glXGetVisualFromFBConfig(Display *dpy, GLXFBConfig config) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->GetVisualFromFBConfig)(dpy, config); +} + + +Bool PUBLIC +glXMakeContextCurrent(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) +{ + Bool b; + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + b = (t->MakeContextCurrent)(dpy, draw, read, ctx); + if (b) { + SetCurrentContext(ctx); + } + return b; +} + + +int PUBLIC +glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + assert(t); + if (!t) + return 0; /* XXX correct? */ + return (t->QueryContext)(dpy, ctx, attribute, value); +} + + +void PUBLIC +glXQueryDrawable(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->QueryDrawable)(dpy, draw, attribute, value); +} + + +void PUBLIC +glXSelectEvent(Display *dpy, GLXDrawable drawable, unsigned long mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->SelectEvent)(dpy, drawable, mask); +} + + + +/*** GLX_SGI_swap_control ***/ + +int PUBLIC +glXSwapIntervalSGI(int interval) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->SwapIntervalSGI)(interval); +} + + + +/*** GLX_SGI_video_sync ***/ + +int PUBLIC +glXGetVideoSyncSGI(unsigned int *count) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t || !glXGetCurrentContext()) + return GLX_BAD_CONTEXT; + return (t->GetVideoSyncSGI)(count); +} + +int PUBLIC +glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t || !glXGetCurrentContext()) + return GLX_BAD_CONTEXT; + return (t->WaitVideoSyncSGI)(divisor, remainder, count); +} + + + +/*** GLX_SGI_make_current_read ***/ + +Bool PUBLIC +glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->MakeCurrentReadSGI)(dpy, draw, read, ctx); +} + +GLXDrawable PUBLIC +glXGetCurrentReadDrawableSGI(void) +{ + return glXGetCurrentReadDrawable(); +} + + +#if defined(_VL_H) + +GLXVideoSourceSGIX PUBLIC +glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXVideoSourceSGIX)(dpy, screen, server, path, nodeClass, drainNode); +} + +void PUBLIC +glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->DestroyGLXVideoSourceSGIX)(dpy, src); +} + +#endif + + +/*** GLX_EXT_import_context ***/ + +void PUBLIC +glXFreeContextEXT(Display *dpy, GLXContext context) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->FreeContextEXT)(dpy, context); +} + +GLXContextID PUBLIC +glXGetContextIDEXT(const GLXContext context) +{ + return ((__GLXcontext *) context)->xid; +} + +Display PUBLIC * +glXGetCurrentDisplayEXT(void) +{ + return glXGetCurrentDisplay(); +} + +GLXContext PUBLIC +glXImportContextEXT(Display *dpy, GLXContextID contextID) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ImportContextEXT)(dpy, contextID); +} + +int PUBLIC +glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute,int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; /* XXX ok? */ + return (t->QueryContextInfoEXT)(dpy, context, attribute, value); +} + + + +/*** GLX_SGIX_fbconfig ***/ + +int PUBLIC +glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->GetFBConfigAttribSGIX)(dpy, config, attribute, value); +} + +GLXFBConfigSGIX PUBLIC * +glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ChooseFBConfigSGIX)(dpy, screen, attrib_list, nelements); +} + +GLXPixmap PUBLIC +glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXPixmapWithConfigSGIX)(dpy, config, pixmap); +} + +GLXContext PUBLIC +glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateContextWithConfigSGIX)(dpy, config, render_type, share_list, direct); +} + +XVisualInfo PUBLIC * +glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->GetVisualFromFBConfigSGIX)(dpy, config); +} + +GLXFBConfigSGIX PUBLIC +glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->GetFBConfigFromVisualSGIX)(dpy, vis); +} + + + +/*** GLX_SGIX_pbuffer ***/ + +GLXPbufferSGIX PUBLIC +glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attrib_list) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXPbufferSGIX)(dpy, config, width, height, attrib_list); +} + +void PUBLIC +glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyGLXPbufferSGIX)(dpy, pbuf); +} + +int PUBLIC +glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->QueryGLXPbufferSGIX)(dpy, pbuf, attribute, value); +} + +void PUBLIC +glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->SelectEventSGIX)(dpy, drawable, mask); +} + +void PUBLIC +glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->GetSelectedEventSGIX)(dpy, drawable, mask); +} + + + +/*** GLX_SGI_cushion ***/ + +void PUBLIC +glXCushionSGI(Display *dpy, Window win, float cushion) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->CushionSGI)(dpy, win, cushion); +} + + + +/*** GLX_SGIX_video_resize ***/ + +int PUBLIC +glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->BindChannelToWindowSGIX)(dpy, screen, channel, window); +} + +int PUBLIC +glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ChannelRectSGIX)(dpy, screen, channel, x, y, w, h); +} + +int PUBLIC +glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->QueryChannelRectSGIX)(dpy, screen, channel, x, y, w, h); +} + +int PUBLIC +glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->QueryChannelDeltasSGIX)(dpy, screen, channel, dx, dy, dw, dh); +} + +int PUBLIC +glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ChannelRectSyncSGIX)(dpy, screen, channel, synctype); +} + + + +#if defined(_DM_BUFFER_H_) + +Bool PUBLIC +glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->AssociateDMPbufferSGIX)(dpy, pbuffer, params, dmbuffer); +} + +#endif + + +/*** GLX_SGIX_swap_group ***/ + +void PUBLIC +glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (*t->JoinSwapGroupSGIX)(dpy, drawable, member); +} + + +/*** GLX_SGIX_swap_barrier ***/ + +void PUBLIC +glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (*t->BindSwapBarrierSGIX)(dpy, drawable, barrier); +} + +Bool PUBLIC +glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (*t->QueryMaxSwapBarriersSGIX)(dpy, screen, max); +} + + + +/*** GLX_SUN_get_transparent_index ***/ + +Status PUBLIC +glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (*t->GetTransparentIndexSUN)(dpy, overlay, underlay, pTransparent); +} + + + +/*** GLX_MESA_copy_sub_buffer ***/ + +void PUBLIC +glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->CopySubBufferMESA)(dpy, drawable, x, y, width, height); +} + + + +/*** GLX_MESA_release_buffers ***/ + +Bool PUBLIC +glXReleaseBuffersMESA(Display *dpy, Window w) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->ReleaseBuffersMESA)(dpy, w); +} + + + +/*** GLX_MESA_pixmap_colormap ***/ + +GLXPixmap PUBLIC +glXCreateGLXPixmapMESA(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXPixmapMESA)(dpy, visinfo, pixmap, cmap); +} + + + +/*** GLX_MESA_set_3dfx_mode ***/ + +Bool PUBLIC +glXSet3DfxModeMESA(int mode) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->Set3DfxModeMESA)(mode); +} + + + +/*** GLX_NV_vertex_array_range ***/ + +void PUBLIC * +glXAllocateMemoryNV( GLsizei size, + GLfloat readFrequency, + GLfloat writeFrequency, + GLfloat priority ) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->AllocateMemoryNV)(size, readFrequency, writeFrequency, priority); +} + + +void PUBLIC +glXFreeMemoryNV( GLvoid *pointer ) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->FreeMemoryNV)(pointer); +} + + + + +/*** GLX_MESA_agp_offset */ + +GLuint PUBLIC +glXGetAGPOffsetMESA( const GLvoid *pointer ) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return ~0; + return (t->GetAGPOffsetMESA)(pointer); +} + + +/*** GLX_MESA_allocate_memory */ + +void * +glXAllocateMemoryMESA(Display *dpy, int scrn, size_t size, + float readfreq, float writefreq, float priority) +{ + /* dummy */ + return NULL; +} + +void +glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer) +{ + /* dummy */ +} + + +GLuint +glXGetMemoryOffsetMESA(Display *dpy, int scrn, const void *pointer) +{ + /* dummy */ + return 0; +} + + +/*** GLX_EXT_texture_from_pixmap */ + +void +glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, + const int *attrib_list) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (t) + t->BindTexImageEXT(dpy, drawable, buffer, attrib_list); +} + +void +glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (t) + t->ReleaseTexImageEXT(dpy, drawable, buffer); +} + + +/**********************************************************************/ +/* GLX API management functions */ +/**********************************************************************/ + + +const char * +_glxapi_get_version(void) +{ + return "1.3"; +} + + +/* + * Return array of extension strings. + */ +const char ** +_glxapi_get_extensions(void) +{ + static const char *extensions[] = { +#ifdef GLX_EXT_import_context + "GLX_EXT_import_context", +#endif +#ifdef GLX_SGI_video_sync + "GLX_SGI_video_sync", +#endif +#ifdef GLX_MESA_copy_sub_buffer + "GLX_MESA_copy_sub_buffer", +#endif +#ifdef GLX_MESA_release_buffers + "GLX_MESA_release_buffers", +#endif +#ifdef GLX_MESA_pixmap_colormap + "GLX_MESA_pixmap_colormap", +#endif +#ifdef GLX_MESA_set_3dfx_mode + "GLX_MESA_set_3dfx_mode", +#endif +#ifdef GLX_SGIX_fbconfig + "GLX_SGIX_fbconfig", +#endif +#ifdef GLX_SGIX_pbuffer + "GLX_SGIX_pbuffer", +#endif +#ifdef GLX_EXT_texture_from_pixmap + "GLX_EXT_texture_from_pixmap", +#endif + NULL + }; + return extensions; +} + + +/* + * Return size of the GLX dispatch table, in entries, not bytes. + */ +GLuint +_glxapi_get_dispatch_table_size(void) +{ + return sizeof(struct _glxapi_table) / sizeof(void *); +} + + +static int +generic_no_op_func(void) +{ + return 0; +} + + +/* + * Initialize all functions in given dispatch table to be no-ops + */ +void +_glxapi_set_no_op_table(struct _glxapi_table *t) +{ + typedef int (*nop_func)(void); + nop_func *dispatch = (nop_func *) t; + GLuint n = _glxapi_get_dispatch_table_size(); + GLuint i; + for (i = 0; i < n; i++) { + dispatch[i] = generic_no_op_func; + } +} + + +struct name_address_pair { + const char *Name; + __GLXextFuncPtr Address; +}; + +static struct name_address_pair GLX_functions[] = { + /*** GLX_VERSION_1_0 ***/ + { "glXChooseVisual", (__GLXextFuncPtr) glXChooseVisual }, + { "glXCopyContext", (__GLXextFuncPtr) glXCopyContext }, + { "glXCreateContext", (__GLXextFuncPtr) glXCreateContext }, + { "glXCreateGLXPixmap", (__GLXextFuncPtr) glXCreateGLXPixmap }, + { "glXDestroyContext", (__GLXextFuncPtr) glXDestroyContext }, + { "glXDestroyGLXPixmap", (__GLXextFuncPtr) glXDestroyGLXPixmap }, + { "glXGetConfig", (__GLXextFuncPtr) glXGetConfig }, + { "glXGetCurrentContext", (__GLXextFuncPtr) glXGetCurrentContext }, + { "glXGetCurrentDrawable", (__GLXextFuncPtr) glXGetCurrentDrawable }, + { "glXIsDirect", (__GLXextFuncPtr) glXIsDirect }, + { "glXMakeCurrent", (__GLXextFuncPtr) glXMakeCurrent }, + { "glXQueryExtension", (__GLXextFuncPtr) glXQueryExtension }, + { "glXQueryVersion", (__GLXextFuncPtr) glXQueryVersion }, + { "glXSwapBuffers", (__GLXextFuncPtr) glXSwapBuffers }, + { "glXUseXFont", (__GLXextFuncPtr) glXUseXFont }, + { "glXWaitGL", (__GLXextFuncPtr) glXWaitGL }, + { "glXWaitX", (__GLXextFuncPtr) glXWaitX }, + + /*** GLX_VERSION_1_1 ***/ + { "glXGetClientString", (__GLXextFuncPtr) glXGetClientString }, + { "glXQueryExtensionsString", (__GLXextFuncPtr) glXQueryExtensionsString }, + { "glXQueryServerString", (__GLXextFuncPtr) glXQueryServerString }, + + /*** GLX_VERSION_1_2 ***/ + { "glXGetCurrentDisplay", (__GLXextFuncPtr) glXGetCurrentDisplay }, + + /*** GLX_VERSION_1_3 ***/ + { "glXChooseFBConfig", (__GLXextFuncPtr) glXChooseFBConfig }, + { "glXCreateNewContext", (__GLXextFuncPtr) glXCreateNewContext }, + { "glXCreatePbuffer", (__GLXextFuncPtr) glXCreatePbuffer }, + { "glXCreatePixmap", (__GLXextFuncPtr) glXCreatePixmap }, + { "glXCreateWindow", (__GLXextFuncPtr) glXCreateWindow }, + { "glXDestroyPbuffer", (__GLXextFuncPtr) glXDestroyPbuffer }, + { "glXDestroyPixmap", (__GLXextFuncPtr) glXDestroyPixmap }, + { "glXDestroyWindow", (__GLXextFuncPtr) glXDestroyWindow }, + { "glXGetCurrentReadDrawable", (__GLXextFuncPtr) glXGetCurrentReadDrawable }, + { "glXGetFBConfigAttrib", (__GLXextFuncPtr) glXGetFBConfigAttrib }, + { "glXGetFBConfigs", (__GLXextFuncPtr) glXGetFBConfigs }, + { "glXGetSelectedEvent", (__GLXextFuncPtr) glXGetSelectedEvent }, + { "glXGetVisualFromFBConfig", (__GLXextFuncPtr) glXGetVisualFromFBConfig }, + { "glXMakeContextCurrent", (__GLXextFuncPtr) glXMakeContextCurrent }, + { "glXQueryContext", (__GLXextFuncPtr) glXQueryContext }, + { "glXQueryDrawable", (__GLXextFuncPtr) glXQueryDrawable }, + { "glXSelectEvent", (__GLXextFuncPtr) glXSelectEvent }, + + /*** GLX_VERSION_1_4 ***/ + { "glXGetProcAddress", (__GLXextFuncPtr) glXGetProcAddress }, + + /*** GLX_SGI_swap_control ***/ + { "glXSwapIntervalSGI", (__GLXextFuncPtr) glXSwapIntervalSGI }, + + /*** GLX_SGI_video_sync ***/ + { "glXGetVideoSyncSGI", (__GLXextFuncPtr) glXGetVideoSyncSGI }, + { "glXWaitVideoSyncSGI", (__GLXextFuncPtr) glXWaitVideoSyncSGI }, + + /*** GLX_SGI_make_current_read ***/ + { "glXMakeCurrentReadSGI", (__GLXextFuncPtr) glXMakeCurrentReadSGI }, + { "glXGetCurrentReadDrawableSGI", (__GLXextFuncPtr) glXGetCurrentReadDrawableSGI }, + + /*** GLX_SGIX_video_source ***/ +#if defined(_VL_H) + { "glXCreateGLXVideoSourceSGIX", (__GLXextFuncPtr) glXCreateGLXVideoSourceSGIX }, + { "glXDestroyGLXVideoSourceSGIX", (__GLXextFuncPtr) glXDestroyGLXVideoSourceSGIX }, +#endif + + /*** GLX_EXT_import_context ***/ + { "glXFreeContextEXT", (__GLXextFuncPtr) glXFreeContextEXT }, + { "glXGetContextIDEXT", (__GLXextFuncPtr) glXGetContextIDEXT }, + { "glXGetCurrentDisplayEXT", (__GLXextFuncPtr) glXGetCurrentDisplayEXT }, + { "glXImportContextEXT", (__GLXextFuncPtr) glXImportContextEXT }, + { "glXQueryContextInfoEXT", (__GLXextFuncPtr) glXQueryContextInfoEXT }, + + /*** GLX_SGIX_fbconfig ***/ + { "glXGetFBConfigAttribSGIX", (__GLXextFuncPtr) glXGetFBConfigAttribSGIX }, + { "glXChooseFBConfigSGIX", (__GLXextFuncPtr) glXChooseFBConfigSGIX }, + { "glXCreateGLXPixmapWithConfigSGIX", (__GLXextFuncPtr) glXCreateGLXPixmapWithConfigSGIX }, + { "glXCreateContextWithConfigSGIX", (__GLXextFuncPtr) glXCreateContextWithConfigSGIX }, + { "glXGetVisualFromFBConfigSGIX", (__GLXextFuncPtr) glXGetVisualFromFBConfigSGIX }, + { "glXGetFBConfigFromVisualSGIX", (__GLXextFuncPtr) glXGetFBConfigFromVisualSGIX }, + + /*** GLX_SGIX_pbuffer ***/ + { "glXCreateGLXPbufferSGIX", (__GLXextFuncPtr) glXCreateGLXPbufferSGIX }, + { "glXDestroyGLXPbufferSGIX", (__GLXextFuncPtr) glXDestroyGLXPbufferSGIX }, + { "glXQueryGLXPbufferSGIX", (__GLXextFuncPtr) glXQueryGLXPbufferSGIX }, + { "glXSelectEventSGIX", (__GLXextFuncPtr) glXSelectEventSGIX }, + { "glXGetSelectedEventSGIX", (__GLXextFuncPtr) glXGetSelectedEventSGIX }, + + /*** GLX_SGI_cushion ***/ + { "glXCushionSGI", (__GLXextFuncPtr) glXCushionSGI }, + + /*** GLX_SGIX_video_resize ***/ + { "glXBindChannelToWindowSGIX", (__GLXextFuncPtr) glXBindChannelToWindowSGIX }, + { "glXChannelRectSGIX", (__GLXextFuncPtr) glXChannelRectSGIX }, + { "glXQueryChannelRectSGIX", (__GLXextFuncPtr) glXQueryChannelRectSGIX }, + { "glXQueryChannelDeltasSGIX", (__GLXextFuncPtr) glXQueryChannelDeltasSGIX }, + { "glXChannelRectSyncSGIX", (__GLXextFuncPtr) glXChannelRectSyncSGIX }, + + /*** GLX_SGIX_dmbuffer **/ +#if defined(_DM_BUFFER_H_) + { "glXAssociateDMPbufferSGIX", (__GLXextFuncPtr) glXAssociateDMPbufferSGIX }, +#endif + + /*** GLX_SGIX_swap_group ***/ + { "glXJoinSwapGroupSGIX", (__GLXextFuncPtr) glXJoinSwapGroupSGIX }, + + /*** GLX_SGIX_swap_barrier ***/ + { "glXBindSwapBarrierSGIX", (__GLXextFuncPtr) glXBindSwapBarrierSGIX }, + { "glXQueryMaxSwapBarriersSGIX", (__GLXextFuncPtr) glXQueryMaxSwapBarriersSGIX }, + + /*** GLX_SUN_get_transparent_index ***/ + { "glXGetTransparentIndexSUN", (__GLXextFuncPtr) glXGetTransparentIndexSUN }, + + /*** GLX_MESA_copy_sub_buffer ***/ + { "glXCopySubBufferMESA", (__GLXextFuncPtr) glXCopySubBufferMESA }, + + /*** GLX_MESA_pixmap_colormap ***/ + { "glXCreateGLXPixmapMESA", (__GLXextFuncPtr) glXCreateGLXPixmapMESA }, + + /*** GLX_MESA_release_buffers ***/ + { "glXReleaseBuffersMESA", (__GLXextFuncPtr) glXReleaseBuffersMESA }, + + /*** GLX_MESA_set_3dfx_mode ***/ + { "glXSet3DfxModeMESA", (__GLXextFuncPtr) glXSet3DfxModeMESA }, + + /*** GLX_ARB_get_proc_address ***/ + { "glXGetProcAddressARB", (__GLXextFuncPtr) glXGetProcAddressARB }, + + /*** GLX_NV_vertex_array_range ***/ + { "glXAllocateMemoryNV", (__GLXextFuncPtr) glXAllocateMemoryNV }, + { "glXFreeMemoryNV", (__GLXextFuncPtr) glXFreeMemoryNV }, + + /*** GLX_MESA_agp_offset ***/ + { "glXGetAGPOffsetMESA", (__GLXextFuncPtr) glXGetAGPOffsetMESA }, + + /*** GLX_MESA_allocate_memory ***/ + { "glXAllocateMemoryMESA", (__GLXextFuncPtr) glXAllocateMemoryMESA }, + { "glXFreeMemoryMESA", (__GLXextFuncPtr) glXFreeMemoryMESA }, + { "glXGetMemoryOffsetMESA", (__GLXextFuncPtr) glXGetMemoryOffsetMESA }, + + /*** GLX_EXT_texture_from_pixmap ***/ + { "glXBindTexImageEXT", (__GLXextFuncPtr) glXBindTexImageEXT }, + { "glXReleaseTexImageEXT", (__GLXextFuncPtr) glXReleaseTexImageEXT }, + + { NULL, NULL } /* end of list */ +}; + + + +/* + * Return address of named glX function, or NULL if not found. + */ +__GLXextFuncPtr +_glxapi_get_proc_address(const char *funcName) +{ + GLuint i; + for (i = 0; GLX_functions[i].Name; i++) { + if (strcmp(GLX_functions[i].Name, funcName) == 0) + return GLX_functions[i].Address; + } + return NULL; +} + + + +/* + * This function does not get dispatched through the dispatch table + * since it's really a "meta" function. + */ +__GLXextFuncPtr +glXGetProcAddressARB(const GLubyte *procName) +{ + __GLXextFuncPtr f; + + f = _glxapi_get_proc_address((const char *) procName); + if (f) { + return f; + } + + f = (__GLXextFuncPtr) _glapi_get_proc_address((const char *) procName); + return f; +} + + +/* GLX 1.4 */ +void (*glXGetProcAddress(const GLubyte *procName))() +{ + return glXGetProcAddressARB(procName); +} diff --git a/src/mesa/pipe/xlib/glxapi.h b/src/mesa/pipe/xlib/glxapi.h new file mode 100644 index 0000000000..37de81e55a --- /dev/null +++ b/src/mesa/pipe/xlib/glxapi.h @@ -0,0 +1,228 @@ +/* + * Mesa 3-D graphics library + * Version: 6.3 + * + * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef _glxapi_h_ +#define _glxapi_h_ + + +#define GLX_GLXEXT_PROTOTYPES +#include "GL/glx.h" + + +/* The GLX API dispatcher (i.e. this code) is being built into stand-alone + * Mesa. We don't know anything about XFree86 or real GLX so we define a + * minimal __GLXContextRec here so some of the functions in this file can + * work properly. + */ +typedef struct __GLXcontextRec { + Display *currentDpy; + GLboolean isDirect; + GLXDrawable currentDrawable; + GLXDrawable currentReadable; + XID xid; +} __GLXcontext; + + +/* + * Almost all the GLX API functions get routed through this dispatch table. + * The exceptions are the glXGetCurrentXXX() functions. + * + * This dispatch table allows multiple GLX client-side modules to coexist. + * Specifically, a real GLX library (like SGI's or the Utah GLX) and Mesa's + * pseudo-GLX can be present at the same time. The former being used on + * GLX-enabled X servers and the later on non-GLX X servers. + * + * Red Hat has been using this since Red Hat Linux 7.0 (I think). + * This'll be a standard feature in XFree86 4.3. It basically allows one + * libGL to do both DRI-rendering and "fake GLX" rendering to X displays + * that lack the GLX extension. + */ +struct _glxapi_table { + /*** GLX_VERSION_1_0 ***/ + XVisualInfo *(*ChooseVisual)(Display *dpy, int screen, int *list); + void (*CopyContext)(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask); + GLXContext (*CreateContext)(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct); + GLXPixmap (*CreateGLXPixmap)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap); + void (*DestroyContext)(Display *dpy, GLXContext ctx); + void (*DestroyGLXPixmap)(Display *dpy, GLXPixmap pixmap); + int (*GetConfig)(Display *dpy, XVisualInfo *visinfo, int attrib, int *value); + /*GLXContext (*GetCurrentContext)(void);*/ + /*GLXDrawable (*GetCurrentDrawable)(void);*/ + Bool (*IsDirect)(Display *dpy, GLXContext ctx); + Bool (*MakeCurrent)(Display *dpy, GLXDrawable drawable, GLXContext ctx); + Bool (*QueryExtension)(Display *dpy, int *errorb, int *event); + Bool (*QueryVersion)(Display *dpy, int *maj, int *min); + void (*SwapBuffers)(Display *dpy, GLXDrawable drawable); + void (*UseXFont)(Font font, int first, int count, int listBase); + void (*WaitGL)(void); + void (*WaitX)(void); + + /*** GLX_VERSION_1_1 ***/ + const char *(*GetClientString)(Display *dpy, int name); + const char *(*QueryExtensionsString)(Display *dpy, int screen); + const char *(*QueryServerString)(Display *dpy, int screen, int name); + + /*** GLX_VERSION_1_2 ***/ + /*Display *(*GetCurrentDisplay)(void);*/ + + /*** GLX_VERSION_1_3 ***/ + GLXFBConfig *(*ChooseFBConfig)(Display *dpy, int screen, const int *attribList, int *nitems); + GLXContext (*CreateNewContext)(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct); + GLXPbuffer (*CreatePbuffer)(Display *dpy, GLXFBConfig config, const int *attribList); + GLXPixmap (*CreatePixmap)(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList); + GLXWindow (*CreateWindow)(Display *dpy, GLXFBConfig config, Window win, const int *attribList); + void (*DestroyPbuffer)(Display *dpy, GLXPbuffer pbuf); + void (*DestroyPixmap)(Display *dpy, GLXPixmap pixmap); + void (*DestroyWindow)(Display *dpy, GLXWindow window); + /*GLXDrawable (*GetCurrentReadDrawable)(void);*/ + int (*GetFBConfigAttrib)(Display *dpy, GLXFBConfig config, int attribute, int *value); + GLXFBConfig *(*GetFBConfigs)(Display *dpy, int screen, int *nelements); + void (*GetSelectedEvent)(Display *dpy, GLXDrawable drawable, unsigned long *mask); + XVisualInfo *(*GetVisualFromFBConfig)(Display *dpy, GLXFBConfig config); + Bool (*MakeContextCurrent)(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx); + int (*QueryContext)(Display *dpy, GLXContext ctx, int attribute, int *value); + void (*QueryDrawable)(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value); + void (*SelectEvent)(Display *dpy, GLXDrawable drawable, unsigned long mask); + + /*** GLX_SGI_swap_control ***/ + int (*SwapIntervalSGI)(int); + + /*** GLX_SGI_video_sync ***/ + int (*GetVideoSyncSGI)(unsigned int *count); + int (*WaitVideoSyncSGI)(int divisor, int remainder, unsigned int *count); + + /*** GLX_SGI_make_current_read ***/ + Bool (*MakeCurrentReadSGI)(Display *, GLXDrawable, GLXDrawable, GLXContext); + /*GLXDrawable (*GetCurrentReadDrawableSGI)(void);*/ + + /*** GLX_SGIX_video_source (needs video library) ***/ +#if defined(_VL_H_) + GLXVideoSourceSGIX (*CreateGLXVideoSourceSGIX)(Display *, int, VLServer, VLPath, int, VLNode); + void (*DestroyGLXVideoSourceSGIX)(Display *, GLXVideoSourceSGIX); +#else + void *CreateGLXVideoSourceSGIX; + void *DestroyGLXVideoSourceSGIX; +#endif + + /*** GLX_EXT_import_context ***/ + void (*FreeContextEXT)(Display *dpy, GLXContext context); + GLXContextID (*GetContextIDEXT)(const GLXContext context); + /*Display *(*GetCurrentDisplayEXT)(void);*/ + GLXContext (*ImportContextEXT)(Display *dpy, GLXContextID contextID); + int (*QueryContextInfoEXT)(Display *dpy, GLXContext context, int attribute,int *value); + + /*** GLX_SGIX_fbconfig ***/ + int (*GetFBConfigAttribSGIX)(Display *, GLXFBConfigSGIX, int, int *); + GLXFBConfigSGIX * (*ChooseFBConfigSGIX)(Display *, int, int *, int *); + GLXPixmap (*CreateGLXPixmapWithConfigSGIX)(Display *, GLXFBConfigSGIX, Pixmap); + GLXContext (*CreateContextWithConfigSGIX)(Display *, GLXFBConfigSGIX, int, GLXContext, Bool); + XVisualInfo * (*GetVisualFromFBConfigSGIX)(Display *, GLXFBConfigSGIX); + GLXFBConfigSGIX (*GetFBConfigFromVisualSGIX)(Display *, XVisualInfo *); + + /*** GLX_SGIX_pbuffer ***/ + GLXPbufferSGIX (*CreateGLXPbufferSGIX)(Display *, GLXFBConfigSGIX, unsigned int, unsigned int, int *); + void (*DestroyGLXPbufferSGIX)(Display *, GLXPbufferSGIX); + int (*QueryGLXPbufferSGIX)(Display *, GLXPbufferSGIX, int, unsigned int *); + void (*SelectEventSGIX)(Display *, GLXDrawable, unsigned long); + void (*GetSelectedEventSGIX)(Display *, GLXDrawable, unsigned long *); + + /*** GLX_SGI_cushion ***/ + void (*CushionSGI)(Display *, Window, float); + + /*** GLX_SGIX_video_resize ***/ + int (*BindChannelToWindowSGIX)(Display *, int, int, Window); + int (*ChannelRectSGIX)(Display *, int, int, int, int, int, int); + int (*QueryChannelRectSGIX)(Display *, int, int, int *, int *, int *, int *); + int (*QueryChannelDeltasSGIX)(Display *, int, int, int *, int *, int *, int *); + int (*ChannelRectSyncSGIX)(Display *, int, int, GLenum); + + /*** GLX_SGIX_dmbuffer (needs dmedia library) ***/ +#if defined (_DM_BUFFER_H_) + Bool (*AssociateDMPbufferSGIX)(Display *, GLXPbufferSGIX, DMparams *, DMbuffer); +#else + void *AssociciateDMPbufferSGIX; +#endif + + /*** GLX_SGIX_swap_group ***/ + void (*JoinSwapGroupSGIX)(Display *, GLXDrawable, GLXDrawable); + + /*** GLX_SGIX_swap_barrier ***/ + void (*BindSwapBarrierSGIX)(Display *, GLXDrawable, int); + Bool (*QueryMaxSwapBarriersSGIX)(Display *, int, int *); + + /*** GLX_SUN_get_transparent_index ***/ + Status (*GetTransparentIndexSUN)(Display *, Window, Window, long *); + + /*** GLX_MESA_copy_sub_buffer ***/ + void (*CopySubBufferMESA)(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height); + + /*** GLX_MESA_release_buffers ***/ + Bool (*ReleaseBuffersMESA)(Display *dpy, Window w); + + /*** GLX_MESA_pixmap_colormap ***/ + GLXPixmap (*CreateGLXPixmapMESA)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap); + + /*** GLX_MESA_set_3dfx_mode ***/ + Bool (*Set3DfxModeMESA)(int mode); + + /*** GLX_NV_vertex_array_range ***/ + void * (*AllocateMemoryNV)( GLsizei size, + GLfloat readFrequency, + GLfloat writeFrequency, + GLfloat priority ); + void (*FreeMemoryNV)( GLvoid *pointer ); + + /*** GLX_MESA_agp_offset ***/ + GLuint (*GetAGPOffsetMESA)( const GLvoid *pointer ); + + /*** GLX_EXT_texture_from_pixmap ***/ + void (*BindTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer, + const int *attrib_list); + void (*ReleaseTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer); +}; + + + +extern const char * +_glxapi_get_version(void); + + +extern const char ** +_glxapi_get_extensions(void); + + +extern GLuint +_glxapi_get_dispatch_table_size(void); + + +extern void +_glxapi_set_no_op_table(struct _glxapi_table *t); + + +extern __GLXextFuncPtr +_glxapi_get_proc_address(const char *funcName); + + +#endif diff --git a/src/mesa/pipe/xlib/glxheader.h b/src/mesa/pipe/xlib/glxheader.h new file mode 100644 index 0000000000..a402191f13 --- /dev/null +++ b/src/mesa/pipe/xlib/glxheader.h @@ -0,0 +1,62 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5.1 + * + * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef GLX_HEADER_H +#define GLX_HEADER_H + +#ifdef __VMS +#include +#endif + +#include "glheader.h" + +#ifdef XFree86Server + +# include "resource.h" +# include "windowstr.h" + +#else + +# include +# include +# include +# ifdef USE_XSHM /* was SHM */ +# include +# include +# include +# endif +# include +# include + +#endif + + + +/* this silences a compiler warning on several systems */ +struct timespec; +struct itimerspec; + + +#endif /*GLX_HEADER*/ diff --git a/src/mesa/pipe/xlib/realglx.c b/src/mesa/pipe/xlib/realglx.c new file mode 100644 index 0000000000..30adb7465b --- /dev/null +++ b/src/mesa/pipe/xlib/realglx.c @@ -0,0 +1,180 @@ + +/* + * Mesa 3-D graphics library + * Version: 5.1 + * + * Copyright (C) 1999-2002 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#include +#include +#include "realglx.h" +#include "glxapi.h" + + +struct _glxapi_table * +_real_GetGLXDispatchTable(void) +{ + static struct _glxapi_table glx; + + /* be sure our dispatch table size <= libGL's table */ + { + GLuint size = sizeof(struct _glxapi_table) / sizeof(void *); + (void) size; + assert(_glxapi_get_dispatch_table_size() >= size); + } + + /* initialize the whole table to no-ops */ + _glxapi_set_no_op_table(&glx); + + /* now initialize the table with the functions I implement */ + + /*** GLX_VERSION_1_0 ***/ + glx.ChooseVisual = _real_glXChooseVisual; + glx.CopyContext = _real_glXCopyContext; + glx.CreateContext = _real_glXCreateContext; + glx.CreateGLXPixmap = _real_glXCreateGLXPixmap; + glx.DestroyContext = _real_glXDestroyContext; + glx.DestroyGLXPixmap = _real_glXDestroyGLXPixmap; + glx.GetConfig = _real_glXGetConfig; + /*glx.GetCurrentContext = _real_glXGetCurrentContext;*/ + /*glx.GetCurrentDrawable = _real_glXGetCurrentDrawable;*/ + glx.IsDirect = _real_glXIsDirect; + glx.MakeCurrent = _real_glXMakeCurrent; + glx.QueryExtension = _real_glXQueryExtension; + glx.QueryVersion = _real_glXQueryVersion; + glx.SwapBuffers = _real_glXSwapBuffers; + glx.UseXFont = _real_glXUseXFont; + glx.WaitGL = _real_glXWaitGL; + glx.WaitX = _real_glXWaitX; + + /*** GLX_VERSION_1_1 ***/ + glx.GetClientString = _real_glXGetClientString; + glx.QueryExtensionsString = _real_glXQueryExtensionsString; + glx.QueryServerString = _real_glXQueryServerString; + + /*** GLX_VERSION_1_2 ***/ + /*glx.GetCurrentDisplay = _real_glXGetCurrentDisplay;*/ + + /*** GLX_VERSION_1_3 ***/ + glx.ChooseFBConfig = _real_glXChooseFBConfig; + glx.CreateNewContext = _real_glXCreateNewContext; + glx.CreatePbuffer = _real_glXCreatePbuffer; + glx.CreatePixmap = _real_glXCreatePixmap; + glx.CreateWindow = _real_glXCreateWindow; + glx.DestroyPbuffer = _real_glXDestroyPbuffer; + glx.DestroyPixmap = _real_glXDestroyPixmap; + glx.DestroyWindow = _real_glXDestroyWindow; + /*glx.GetCurrentReadDrawable = _real_glXGetCurrentReadDrawable;*/ + glx.GetFBConfigAttrib = _real_glXGetFBConfigAttrib; + glx.GetFBConfigs = _real_glXGetFBConfigs; + glx.GetSelectedEvent = _real_glXGetSelectedEvent; + glx.GetVisualFromFBConfig = _real_glXGetVisualFromFBConfig; + glx.MakeContextCurrent = _real_glXMakeContextCurrent; + glx.QueryContext = _real_glXQueryContext; + glx.QueryDrawable = _real_glXQueryDrawable; + glx.SelectEvent = _real_glXSelectEvent; + + /*** GLX_SGI_swap_control ***/ + glx.SwapIntervalSGI = _real_glXSwapIntervalSGI; + + /*** GLX_SGI_video_sync ***/ + glx.GetVideoSyncSGI = _real_glXGetVideoSyncSGI; + glx.WaitVideoSyncSGI = _real_glXWaitVideoSyncSGI; + + /*** GLX_SGI_make_current_read ***/ + glx.MakeCurrentReadSGI = _real_glXMakeCurrentReadSGI; + /*glx.GetCurrentReadDrawableSGI = _real_glXGetCurrentReadDrawableSGI;*/ + +#if defined(_VL_H) + /*** GLX_SGIX_video_source ***/ + glx.CreateGLXVideoSourceSGIX = _real_glXCreateGLXVideoSourceSGIX; + glx.DestroyGLXVideoSourceSGIX = _real_glXDestroyGLXVideoSourceSGIX; +#endif + + /*** GLX_EXT_import_context ***/ + glx.FreeContextEXT = _real_glXFreeContextEXT; + /*glx.GetContextIDEXT = _real_glXGetContextIDEXT;*/ + /*glx.GetCurrentDisplayEXT = _real_glXGetCurrentDisplayEXT;*/ + glx.ImportContextEXT = _real_glXImportContextEXT; + glx.QueryContextInfoEXT = _real_glXQueryContextInfoEXT; + + /*** GLX_SGIX_fbconfig ***/ + glx.GetFBConfigAttribSGIX = _real_glXGetFBConfigAttribSGIX; + glx.ChooseFBConfigSGIX = _real_glXChooseFBConfigSGIX; + glx.CreateGLXPixmapWithConfigSGIX = _real_glXCreateGLXPixmapWithConfigSGIX; + glx.CreateContextWithConfigSGIX = _real_glXCreateContextWithConfigSGIX; + glx.GetVisualFromFBConfigSGIX = _real_glXGetVisualFromFBConfigSGIX; + glx.GetFBConfigFromVisualSGIX = _real_glXGetFBConfigFromVisualSGIX; + + /*** GLX_SGIX_pbuffer ***/ + glx.CreateGLXPbufferSGIX = _real_glXCreateGLXPbufferSGIX; + glx.DestroyGLXPbufferSGIX = _real_glXDestroyGLXPbufferSGIX; + glx.QueryGLXPbufferSGIX = _real_glXQueryGLXPbufferSGIX; + glx.SelectEventSGIX = _real_glXSelectEventSGIX; + glx.GetSelectedEventSGIX = _real_glXGetSelectedEventSGIX; + + /*** GLX_SGI_cushion ***/ + glx.CushionSGI = _real_glXCushionSGI; + + /*** GLX_SGIX_video_resize ***/ + glx.BindChannelToWindowSGIX = _real_glXBindChannelToWindowSGIX; + glx.ChannelRectSGIX = _real_glXChannelRectSGIX; + glx.QueryChannelRectSGIX = _real_glXQueryChannelRectSGIX; + glx.QueryChannelDeltasSGIX = _real_glXQueryChannelDeltasSGIX; + glx.ChannelRectSyncSGIX = _real_glXChannelRectSyncSGIX; + +#if defined(_DM_BUFFER_H_) + /*** (GLX_SGIX_dmbuffer ***/ + glx.AssociateDMPbufferSGIX = NULL; +#endif + + /*** GLX_SGIX_swap_group ***/ + glx.JoinSwapGroupSGIX = _real_glXJoinSwapGroupSGIX; + + /*** GLX_SGIX_swap_barrier ***/ + glx.BindSwapBarrierSGIX = _real_glXBindSwapBarrierSGIX; + glx.QueryMaxSwapBarriersSGIX = _real_glXQueryMaxSwapBarriersSGIX; + + /*** GLX_SUN_get_transparent_index ***/ + glx.GetTransparentIndexSUN = _real_glXGetTransparentIndexSUN; + + /*** GLX_MESA_copy_sub_buffer ***/ + glx.CopySubBufferMESA = _real_glXCopySubBufferMESA; + + /*** GLX_MESA_release_buffers ***/ + glx.ReleaseBuffersMESA = _real_glXReleaseBuffersMESA; + + /*** GLX_MESA_pixmap_colormap ***/ + glx.CreateGLXPixmapMESA = _real_glXCreateGLXPixmapMESA; + + /*** GLX_MESA_set_3dfx_mode ***/ + glx.Set3DfxModeMESA = _real_glXSet3DfxModeMESA; + + /*** GLX_NV_vertex_array_range ***/ + glx.AllocateMemoryNV = _real_glXAllocateMemoryNV; + glx.FreeMemoryNV = _real_glXFreeMemoryNV; + + /*** GLX_MESA_agp_offset ***/ + glx.GetAGPOffsetMESA = _real_glXGetAGPOffsetMESA; + + return &glx; +} diff --git a/src/mesa/pipe/xlib/realglx.h b/src/mesa/pipe/xlib/realglx.h new file mode 100644 index 0000000000..150129db68 --- /dev/null +++ b/src/mesa/pipe/xlib/realglx.h @@ -0,0 +1,326 @@ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef REALGLX_H +#define REALGLX_H + + +extern struct _glxapi_table * +_real_GetGLXDispatchTable(void); + + +/* + * Basically just need these to prevent compiler warnings. + */ + + +extern XVisualInfo * +_real_glXChooseVisual( Display *dpy, int screen, int *list ); + +extern GLXContext +_real_glXCreateContext( Display *dpy, XVisualInfo *visinfo, + GLXContext share_list, Bool direct ); + +extern GLXPixmap +_real_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ); + +extern GLXPixmap +_real_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, + Pixmap pixmap, Colormap cmap ); + +extern void +_real_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap ); + +extern void +_real_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, + unsigned long mask ); + +extern Bool +_real_glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx ); + +extern Bool +_real_glXQueryExtension( Display *dpy, int *errorb, int *event ); + +extern void +_real_glXDestroyContext( Display *dpy, GLXContext ctx ); + +extern Bool +_real_glXIsDirect( Display *dpy, GLXContext ctx ); + +extern void +_real_glXSwapBuffers( Display *dpy, GLXDrawable drawable ); + +extern void +_real_glXUseXFont( Font font, int first, int count, int listbase ); + +extern Bool +_real_glXQueryVersion( Display *dpy, int *maj, int *min ); + +extern int +_real_glXGetConfig( Display *dpy, XVisualInfo *visinfo, + int attrib, int *value ); + +extern void +_real_glXWaitGL( void ); + + +extern void +_real_glXWaitX( void ); + +/* GLX 1.1 and later */ +extern const char * +_real_glXQueryExtensionsString( Display *dpy, int screen ); + +/* GLX 1.1 and later */ +extern const char * +_real_glXQueryServerString( Display *dpy, int screen, int name ); + +/* GLX 1.1 and later */ +extern const char * +_real_glXGetClientString( Display *dpy, int name ); + + +/* + * GLX 1.3 and later + */ + +extern GLXFBConfig * +_real_glXChooseFBConfig( Display *dpy, int screen, + const int *attribList, int *nitems ); + +extern int +_real_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config, + int attribute, int *value ); + +extern GLXFBConfig * +_real_glXGetFBConfigs( Display *dpy, int screen, int *nelements ); + +extern XVisualInfo * +_real_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config ); + +extern GLXWindow +_real_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win, + const int *attribList ); + +extern void +_real_glXDestroyWindow( Display *dpy, GLXWindow window ); + +extern GLXPixmap +_real_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, + const int *attribList ); + +extern void +_real_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap ); + +extern GLXPbuffer +_real_glXCreatePbuffer( Display *dpy, GLXFBConfig config, + const int *attribList ); + +extern void +_real_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf ); + +extern void +_real_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, + unsigned int *value ); + +extern GLXContext +_real_glXCreateNewContext( Display *dpy, GLXFBConfig config, + int renderType, GLXContext shareList, Bool direct ); + + +extern Bool +_real_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, + GLXDrawable read, GLXContext ctx ); + +extern int +_real_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value ); + +extern void +_real_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask ); + +extern void +_real_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable, + unsigned long *mask ); + +#ifdef GLX_SGI_swap_control +extern int +_real_glXSwapIntervalSGI(int interval); +#endif + + +#ifdef GLX_SGI_video_sync +extern int +_real_glXGetVideoSyncSGI(unsigned int *count); + +extern int +_real_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count); +#endif + + +#ifdef GLX_SGI_make_current_read +extern Bool +_real_glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx); + +extern GLXDrawable +_real_glXGetCurrentReadDrawableSGI(void); +#endif + +#if defined(_VL_H) && defined(GLX_SGIX_video_source) +extern GLXVideoSourceSGIX +_real_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode); + +extern void +_real_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src); +#endif + +#ifdef GLX_EXT_import_context +extern void +_real_glXFreeContextEXT(Display *dpy, GLXContext context); + +extern GLXContextID +_real_glXGetContextIDEXT(const GLXContext context); + +extern Display * +_real_glXGetCurrentDisplayEXT(void); + +extern GLXContext +_real_glXImportContextEXT(Display *dpy, GLXContextID contextID); + +extern int +_real_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value); +#endif + +#ifdef GLX_SGIX_fbconfig +extern int +_real_glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value); + +extern GLXFBConfigSGIX * +_real_glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements); + +extern GLXPixmap +_real_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap); + +extern GLXContext +_real_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct); + +extern XVisualInfo * +_real_glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config); + +extern GLXFBConfigSGIX +_real_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis); +#endif + +#ifdef GLX_SGIX_pbuffer +extern GLXPbufferSGIX +_real_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attrib_list); + +extern void +_real_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf); + +extern int +_real_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value); + +extern void +_real_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask); + +extern void +_real_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask); +#endif + +#ifdef GLX_SGI_cushion +extern void +_real_glXCushionSGI(Display *dpy, Window win, float cushion); +#endif + +#ifdef GLX_SGIX_video_resize +extern int +_real_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window); + +extern int +_real_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h); + +extern int +_real_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h); + +extern int +_real_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh); + +extern int +_real_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype); +#endif + +#if defined(_DM_BUFFER_H_) && defined(GLX_SGIX_dmbuffer) +extern Bool +_real_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer); +#endif + +#ifdef GLX_SGIX_swap_group +extern void +_real_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member); +#endif + +#ifdef GLX_SGIX_swap_barrier +extern void +_real_glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier); + +extern Bool +_real_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max); +#endif + +#ifdef GLX_SUN_get_transparent_index +extern Status +_real_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent); +#endif + +#ifdef GLX_MESA_release_buffers +extern Bool +_real_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d ); +#endif + +#ifdef GLX_MESA_set_3dfx_mode +extern Bool +_real_glXSet3DfxModeMESA( int mode ); +#endif + +#ifdef GLX_NV_vertex_array_range +extern void * +_real_glXAllocateMemoryNV(GLsizei size, GLfloat readfreq, GLfloat writefreq, GLfloat priority); +extern void +_real_glXFreeMemoryNV(GLvoid *pointer); +#endif + +#ifdef GLX_MESA_agp_offset +extern GLuint +_real_glXGetAGPOffsetMESA(const GLvoid *pointer); +#endif + +#ifdef GLX_MESA_copy_sub_buffer +extern void +_real_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable, + int x, int y, int width, int height ); +#endif + +#endif /* REALGLX_H */ diff --git a/src/mesa/pipe/xlib/xfonts.c b/src/mesa/pipe/xlib/xfonts.c new file mode 100644 index 0000000000..d72c600bd1 --- /dev/null +++ b/src/mesa/pipe/xlib/xfonts.c @@ -0,0 +1,377 @@ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2000 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* xfonts.c -- glXUseXFont() for Mesa written by + * Copyright (C) 1995 Thorsten.Ohl @ Physik.TH-Darmstadt.de + */ + +#ifdef __VMS +#include +#endif + +#include "glxheader.h" +#include "context.h" +#include "imports.h" +#include "xfonts.h" + + +/* Some debugging info. */ + +#ifdef DEBUG +#undef _R +#undef _G +#undef _B +#include + +int debug_xfonts = 0; + +static void +dump_char_struct(XCharStruct * ch, char *prefix) +{ + printf("%slbearing = %d, rbearing = %d, width = %d\n", + prefix, ch->lbearing, ch->rbearing, ch->width); + printf("%sascent = %d, descent = %d, attributes = %u\n", + prefix, ch->ascent, ch->descent, (unsigned int) ch->attributes); +} + +static void +dump_font_struct(XFontStruct * font) +{ + printf("ascent = %d, descent = %d\n", font->ascent, font->descent); + printf("char_or_byte2 = (%u,%u)\n", + font->min_char_or_byte2, font->max_char_or_byte2); + printf("byte1 = (%u,%u)\n", font->min_byte1, font->max_byte1); + printf("all_chars_exist = %s\n", font->all_chars_exist ? "True" : "False"); + printf("default_char = %c (\\%03o)\n", + (char) (isprint(font->default_char) ? font->default_char : ' '), + font->default_char); + dump_char_struct(&font->min_bounds, "min> "); + dump_char_struct(&font->max_bounds, "max> "); +#if 0 + for (c = font->min_char_or_byte2; c <= font->max_char_or_byte2; c++) { + char prefix[8]; + sprintf(prefix, "%d> ", c); + dump_char_struct(&font->per_char[c], prefix); + } +#endif +} + +static void +dump_bitmap(unsigned int width, unsigned int height, GLubyte * bitmap) +{ + unsigned int x, y; + + printf(" "); + for (x = 0; x < 8 * width; x++) + printf("%o", 7 - (x % 8)); + putchar('\n'); + for (y = 0; y < height; y++) { + printf("%3o:", y); + for (x = 0; x < 8 * width; x++) + putchar((bitmap[width * (height - y - 1) + x / 8] & (1 << (7 - (x % + 8)))) + ? '*' : '.'); + printf(" "); + for (x = 0; x < width; x++) + printf("0x%02x, ", bitmap[width * (height - y - 1) + x]); + putchar('\n'); + } +} +#endif /* DEBUG */ + + +/* Implementation. */ + +/* Fill a BITMAP with a character C from thew current font + in the graphics context GC. WIDTH is the width in bytes + and HEIGHT is the height in bits. + + Note that the generated bitmaps must be used with + + glPixelStorei (GL_UNPACK_SWAP_BYTES, GL_FALSE); + glPixelStorei (GL_UNPACK_LSB_FIRST, GL_FALSE); + glPixelStorei (GL_UNPACK_ROW_LENGTH, 0); + glPixelStorei (GL_UNPACK_SKIP_ROWS, 0); + glPixelStorei (GL_UNPACK_SKIP_PIXELS, 0); + glPixelStorei (GL_UNPACK_ALIGNMENT, 1); + + Possible optimizations: + + * use only one reusable pixmap with the maximum dimensions. + * draw the entire font into a single pixmap (careful with + proportional fonts!). +*/ + + +/* + * Generate OpenGL-compatible bitmap. + */ +static void +fill_bitmap(Display * dpy, Window win, GC gc, + unsigned int width, unsigned int height, + int x0, int y0, unsigned int c, GLubyte * bitmap) +{ + XImage *image; + unsigned int x, y; + Pixmap pixmap; + XChar2b char2b; + + pixmap = XCreatePixmap(dpy, win, 8 * width, height, 1); + XSetForeground(dpy, gc, 0); + XFillRectangle(dpy, pixmap, gc, 0, 0, 8 * width, height); + XSetForeground(dpy, gc, 1); + + char2b.byte1 = (c >> 8) & 0xff; + char2b.byte2 = (c & 0xff); + + XDrawString16(dpy, pixmap, gc, x0, y0, &char2b, 1); + + image = XGetImage(dpy, pixmap, 0, 0, 8 * width, height, 1, XYPixmap); + if (image) { + /* Fill the bitmap (X11 and OpenGL are upside down wrt each other). */ + for (y = 0; y < height; y++) + for (x = 0; x < 8 * width; x++) + if (XGetPixel(image, x, y)) + bitmap[width * (height - y - 1) + x / 8] |= + (1 << (7 - (x % 8))); + XDestroyImage(image); + } + + XFreePixmap(dpy, pixmap); +} + +/* + * determine if a given glyph is valid and return the + * corresponding XCharStruct. + */ +static XCharStruct * +isvalid(XFontStruct * fs, unsigned int which) +{ + unsigned int rows, pages; + unsigned int byte1 = 0, byte2 = 0; + int i, valid = 1; + + rows = fs->max_byte1 - fs->min_byte1 + 1; + pages = fs->max_char_or_byte2 - fs->min_char_or_byte2 + 1; + + if (rows == 1) { + /* "linear" fonts */ + if ((fs->min_char_or_byte2 > which) || (fs->max_char_or_byte2 < which)) + valid = 0; + } + else { + /* "matrix" fonts */ + byte2 = which & 0xff; + byte1 = which >> 8; + if ((fs->min_char_or_byte2 > byte2) || + (fs->max_char_or_byte2 < byte2) || + (fs->min_byte1 > byte1) || (fs->max_byte1 < byte1)) + valid = 0; + } + + if (valid) { + if (fs->per_char) { + if (rows == 1) { + /* "linear" fonts */ + return (fs->per_char + (which - fs->min_char_or_byte2)); + } + else { + /* "matrix" fonts */ + i = ((byte1 - fs->min_byte1) * pages) + + (byte2 - fs->min_char_or_byte2); + return (fs->per_char + i); + } + } + else { + return (&fs->min_bounds); + } + } + return (NULL); +} + + +void +Fake_glXUseXFont(Font font, int first, int count, int listbase) +{ + Display *dpy; + Window win; + Pixmap pixmap; + GC gc; + XGCValues values; + unsigned long valuemask; + XFontStruct *fs; + GLint swapbytes, lsbfirst, rowlength; + GLint skiprows, skippixels, alignment; + unsigned int max_width, max_height, max_bm_width, max_bm_height; + GLubyte *bm; + int i; + + dpy = glXGetCurrentDisplay(); + if (!dpy) + return; /* I guess glXMakeCurrent wasn't called */ + win = RootWindow(dpy, DefaultScreen(dpy)); + + fs = XQueryFont(dpy, font); + if (!fs) { + _mesa_error(NULL, GL_INVALID_VALUE, + "Couldn't get font structure information"); + return; + } + + /* Allocate a bitmap that can fit all characters. */ + max_width = fs->max_bounds.rbearing - fs->min_bounds.lbearing; + max_height = fs->max_bounds.ascent + fs->max_bounds.descent; + max_bm_width = (max_width + 7) / 8; + max_bm_height = max_height; + + bm = (GLubyte *) MALLOC((max_bm_width * max_bm_height) * sizeof(GLubyte)); + if (!bm) { + XFreeFontInfo(NULL, fs, 1); + _mesa_error(NULL, GL_OUT_OF_MEMORY, + "Couldn't allocate bitmap in glXUseXFont()"); + return; + } + +#if 0 + /* get the page info */ + pages = fs->max_char_or_byte2 - fs->min_char_or_byte2 + 1; + firstchar = (fs->min_byte1 << 8) + fs->min_char_or_byte2; + lastchar = (fs->max_byte1 << 8) + fs->max_char_or_byte2; + rows = fs->max_byte1 - fs->min_byte1 + 1; + unsigned int first_char, last_char, pages, rows; +#endif + + /* Save the current packing mode for bitmaps. */ + glGetIntegerv(GL_UNPACK_SWAP_BYTES, &swapbytes); + glGetIntegerv(GL_UNPACK_LSB_FIRST, &lsbfirst); + glGetIntegerv(GL_UNPACK_ROW_LENGTH, &rowlength); + glGetIntegerv(GL_UNPACK_SKIP_ROWS, &skiprows); + glGetIntegerv(GL_UNPACK_SKIP_PIXELS, &skippixels); + glGetIntegerv(GL_UNPACK_ALIGNMENT, &alignment); + + /* Enforce a standard packing mode which is compatible with + fill_bitmap() from above. This is actually the default mode, + except for the (non)alignment. */ + glPixelStorei(GL_UNPACK_SWAP_BYTES, GL_FALSE); + glPixelStorei(GL_UNPACK_LSB_FIRST, GL_FALSE); + glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); + glPixelStorei(GL_UNPACK_SKIP_ROWS, 0); + glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0); + glPixelStorei(GL_UNPACK_ALIGNMENT, 1); + + pixmap = XCreatePixmap(dpy, win, 10, 10, 1); + values.foreground = BlackPixel(dpy, DefaultScreen(dpy)); + values.background = WhitePixel(dpy, DefaultScreen(dpy)); + values.font = fs->fid; + valuemask = GCForeground | GCBackground | GCFont; + gc = XCreateGC(dpy, pixmap, valuemask, &values); + XFreePixmap(dpy, pixmap); + +#ifdef DEBUG + if (debug_xfonts) + dump_font_struct(fs); +#endif + + for (i = 0; i < count; i++) { + unsigned int width, height, bm_width, bm_height; + GLfloat x0, y0, dx, dy; + XCharStruct *ch; + int x, y; + unsigned int c = first + i; + int list = listbase + i; + int valid; + + /* check on index validity and get the bounds */ + ch = isvalid(fs, c); + if (!ch) { + ch = &fs->max_bounds; + valid = 0; + } + else { + valid = 1; + } + +#ifdef DEBUG + if (debug_xfonts) { + char s[7]; + sprintf(s, isprint(c) ? "%c> " : "\\%03o> ", c); + dump_char_struct(ch, s); + } +#endif + + /* glBitmap()' parameters: + straight from the glXUseXFont(3) manpage. */ + width = ch->rbearing - ch->lbearing; + height = ch->ascent + ch->descent; + x0 = -ch->lbearing; + y0 = ch->descent - 0; /* XXX used to subtract 1 here */ + /* but that caused a conformace failure */ + dx = ch->width; + dy = 0; + + /* X11's starting point. */ + x = -ch->lbearing; + y = ch->ascent; + + /* Round the width to a multiple of eight. We will use this also + for the pixmap for capturing the X11 font. This is slightly + inefficient, but it makes the OpenGL part real easy. */ + bm_width = (width + 7) / 8; + bm_height = height; + + glNewList(list, GL_COMPILE); + if (valid && (bm_width > 0) && (bm_height > 0)) { + + MEMSET(bm, '\0', bm_width * bm_height); + fill_bitmap(dpy, win, gc, bm_width, bm_height, x, y, c, bm); + + glBitmap(width, height, x0, y0, dx, dy, bm); +#ifdef DEBUG + if (debug_xfonts) { + printf("width/height = %u/%u\n", width, height); + printf("bm_width/bm_height = %u/%u\n", bm_width, bm_height); + dump_bitmap(bm_width, bm_height, bm); + } +#endif + } + else { + glBitmap(0, 0, 0.0, 0.0, dx, dy, NULL); + } + glEndList(); + } + + FREE(bm); + XFreeFontInfo(NULL, fs, 1); + XFreeGC(dpy, gc); + + /* Restore saved packing modes. */ + glPixelStorei(GL_UNPACK_SWAP_BYTES, swapbytes); + glPixelStorei(GL_UNPACK_LSB_FIRST, lsbfirst); + glPixelStorei(GL_UNPACK_ROW_LENGTH, rowlength); + glPixelStorei(GL_UNPACK_SKIP_ROWS, skiprows); + glPixelStorei(GL_UNPACK_SKIP_PIXELS, skippixels); + glPixelStorei(GL_UNPACK_ALIGNMENT, alignment); +} diff --git a/src/mesa/pipe/xlib/xfonts.h b/src/mesa/pipe/xlib/xfonts.h new file mode 100644 index 0000000000..e36f42f817 --- /dev/null +++ b/src/mesa/pipe/xlib/xfonts.h @@ -0,0 +1,41 @@ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2000 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef XFONTS_H +#define XFONTS_H + +#ifdef __VMS +#include +#endif + +#include + + +extern void Fake_glXUseXFont( Font font, int first, int count, int listbase ); + + +#endif + diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c new file mode 100644 index 0000000000..08c98eab48 --- /dev/null +++ b/src/mesa/pipe/xlib/xm_api.c @@ -0,0 +1,2554 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file xm_api.c + * + * All the XMesa* API functions. + * + * + * NOTES: + * + * The window coordinate system origin (0,0) is in the lower-left corner + * of the window. X11's window coordinate origin is in the upper-left + * corner of the window. Therefore, most drawing functions in this + * file have to flip Y coordinates. + * + * Define USE_XSHM in the Makefile with -DUSE_XSHM if you want to compile + * in support for the MIT Shared Memory extension. If enabled, when you + * use an Ximage for the back buffer in double buffered mode, the "swap" + * operation will be faster. You must also link with -lXext. + * + * Byte swapping: If the Mesa host and the X display use a different + * byte order then there's some trickiness to be aware of when using + * XImages. The byte ordering used for the XImage is that of the X + * display, not the Mesa host. + * The color-to-pixel encoding for True/DirectColor must be done + * according to the display's visual red_mask, green_mask, and blue_mask. + * If XPutPixel is used to put a pixel into an XImage then XPutPixel will + * do byte swapping if needed. If one wants to directly "poke" the pixel + * into the XImage's buffer then the pixel must be byte swapped first. In + * Mesa, when byte swapping is needed we use the PF_TRUECOLOR pixel format + * and use XPutPixel everywhere except in the implementation of + * glClear(GL_COLOR_BUFFER_BIT). We want this function to be fast so + * instead of using XPutPixel we "poke" our values after byte-swapping + * the clear pixel value if needed. + * + */ + +#ifdef __CYGWIN__ +#undef WIN32 +#undef __WIN32__ +#endif + +#include "glxheader.h" +#include "GL/xmesa.h" +#include "xmesaP.h" +#include "context.h" +#include "extensions.h" +#include "framebuffer.h" +#include "glthread.h" +#include "imports.h" +#include "macros.h" +#include "renderbuffer.h" +#include "teximage.h" +#include "swrast/swrast.h" +#include "swrast_setup/swrast_setup.h" +#include "vbo/vbo.h" +#if 0 +#include "tnl/tnl.h" +#include "tnl/t_context.h" +#include "tnl/t_pipeline.h" +#endif +#include "drivers/common/driverfuncs.h" + +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "pipe/softpipe/sp_context.h" +#include "pipe/p_defines.h" + +/** + * Global X driver lock + */ +_glthread_Mutex _xmesa_lock; + + + +/** + * Lookup tables for HPCR pixel format: + */ +static short hpcr_rgbTbl[3][256] = { +{ + 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22, 23, 23, + 24, 24, 25, 25, 26, 26, 27, 27, 28, 28, 29, 29, 30, 30, 31, 31, + 32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 37, 37, 38, 38, 39, 39, + 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, + 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, + 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, + 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, + 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, +112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, +128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, +144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, +160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, +176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, +192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, +208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, +224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239 +}, +{ + 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22, 23, 23, + 24, 24, 25, 25, 26, 26, 27, 27, 28, 28, 29, 29, 30, 30, 31, 31, + 32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 37, 37, 38, 38, 39, 39, + 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, + 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, + 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, + 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, + 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, +112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, +128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, +144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, +160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, +176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, +192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, +208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, +224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239 +}, +{ + 32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 37, 37, 38, 38, 39, 39, + 40, 40, 41, 41, 42, 42, 43, 43, 44, 44, 45, 45, 46, 46, 47, 47, + 48, 48, 49, 49, 50, 50, 51, 51, 52, 52, 53, 53, 54, 54, 55, 55, + 56, 56, 57, 57, 58, 58, 59, 59, 60, 60, 61, 61, 62, 62, 63, 63, + 64, 64, 65, 65, 66, 66, 67, 67, 68, 68, 69, 69, 70, 70, 71, 71, + 72, 72, 73, 73, 74, 74, 75, 75, 76, 76, 77, 77, 78, 78, 79, 79, + 80, 80, 81, 81, 82, 82, 83, 83, 84, 84, 85, 85, 86, 86, 87, 87, + 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, + 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, +112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, +128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, +144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, +160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, +176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, +192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, +208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223 +} +}; + + + +/**********************************************************************/ +/***** X Utility Functions *****/ +/**********************************************************************/ + + +/** + * Return the host's byte order as LSBFirst or MSBFirst ala X. + */ +#ifndef XFree86Server +static int host_byte_order( void ) +{ + int i = 1; + char *cptr = (char *) &i; + return (*cptr==1) ? LSBFirst : MSBFirst; +} +#endif + + +/** + * Check if the X Shared Memory extension is available. + * Return: 0 = not available + * 1 = shared XImage support available + * 2 = shared Pixmap support available also + */ +static int check_for_xshm( XMesaDisplay *display ) +{ +#if defined(USE_XSHM) && !defined(XFree86Server) + int major, minor, ignore; + Bool pixmaps; + + if (XQueryExtension( display, "MIT-SHM", &ignore, &ignore, &ignore )) { + if (XShmQueryVersion( display, &major, &minor, &pixmaps )==True) { + return (pixmaps==True) ? 2 : 1; + } + else { + return 0; + } + } + else { + return 0; + } +#else + /* No XSHM support */ + return 0; +#endif +} + + +/** + * Apply gamma correction to an intensity value in [0..max]. Return the + * new intensity value. + */ +static GLint +gamma_adjust( GLfloat gamma, GLint value, GLint max ) +{ + if (gamma == 1.0) { + return value; + } + else { + double x = (double) value / (double) max; + return IROUND_POS((GLfloat) max * _mesa_pow(x, 1.0F/gamma)); + } +} + + + +/** + * Return the true number of bits per pixel for XImages. + * For example, if we request a 24-bit deep visual we may actually need/get + * 32bpp XImages. This function returns the appropriate bpp. + * Input: dpy - the X display + * visinfo - desribes the visual to be used for XImages + * Return: true number of bits per pixel for XImages + */ +static int +bits_per_pixel( XMesaVisual xmv ) +{ +#ifdef XFree86Server + const int depth = xmv->nplanes; + int i; + assert(depth > 0); + for (i = 0; i < screenInfo.numPixmapFormats; i++) { + if (screenInfo.formats[i].depth == depth) + return screenInfo.formats[i].bitsPerPixel; + } + return depth; /* should never get here, but this should be safe */ +#else + XMesaDisplay *dpy = xmv->display; + XMesaVisualInfo visinfo = xmv->visinfo; + XMesaImage *img; + int bitsPerPixel; + /* Create a temporary XImage */ + img = XCreateImage( dpy, visinfo->visual, visinfo->depth, + ZPixmap, 0, /*format, offset*/ + (char*) MALLOC(8), /*data*/ + 1, 1, /*width, height*/ + 32, /*bitmap_pad*/ + 0 /*bytes_per_line*/ + ); + assert(img); + /* grab the bits/pixel value */ + bitsPerPixel = img->bits_per_pixel; + /* free the XImage */ + _mesa_free( img->data ); + img->data = NULL; + XMesaDestroyImage( img ); + return bitsPerPixel; +#endif +} + + + +/* + * Determine if a given X window ID is valid (window exists). + * Do this by calling XGetWindowAttributes() for the window and + * checking if we catch an X error. + * Input: dpy - the display + * win - the window to check for existance + * Return: GL_TRUE - window exists + * GL_FALSE - window doesn't exist + */ +#ifndef XFree86Server +static GLboolean WindowExistsFlag; + +static int window_exists_err_handler( XMesaDisplay* dpy, XErrorEvent* xerr ) +{ + (void) dpy; + if (xerr->error_code == BadWindow) { + WindowExistsFlag = GL_FALSE; + } + return 0; +} + +static GLboolean window_exists( XMesaDisplay *dpy, Window win ) +{ + XWindowAttributes wa; + int (*old_handler)( XMesaDisplay*, XErrorEvent* ); + WindowExistsFlag = GL_TRUE; + old_handler = XSetErrorHandler(window_exists_err_handler); + XGetWindowAttributes( dpy, win, &wa ); /* dummy request */ + XSetErrorHandler(old_handler); + return WindowExistsFlag; +} + +static Status +get_drawable_size( XMesaDisplay *dpy, Drawable d, GLuint *width, GLuint *height ) +{ + Window root; + Status stat; + int xpos, ypos; + unsigned int w, h, bw, depth; + stat = XGetGeometry(dpy, d, &root, &xpos, &ypos, &w, &h, &bw, &depth); + *width = w; + *height = h; + return stat; +} +#endif + + +/** + * Return the size of the window (or pixmap) that corresponds to the + * given XMesaBuffer. + * \param width returns width in pixels + * \param height returns height in pixels + */ +void +xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b, + GLuint *width, GLuint *height) +{ +#ifdef XFree86Server + *width = MIN2(b->frontxrb->drawable->width, MAX_WIDTH); + *height = MIN2(b->frontxrb->drawable->height, MAX_HEIGHT); +#else + Status stat; + + _glthread_LOCK_MUTEX(_xmesa_lock); + XSync(b->xm_visual->display, 0); /* added for Chromium */ + stat = get_drawable_size(dpy, b->frontxrb->pixmap, width, height); + _glthread_UNLOCK_MUTEX(_xmesa_lock); + + if (!stat) { + /* probably querying a window that's recently been destroyed */ + _mesa_warning(NULL, "XGetGeometry failed!\n"); + *width = *height = 1; + } +#endif +} + + + +/**********************************************************************/ +/***** Linked list of XMesaBuffers *****/ +/**********************************************************************/ + +XMesaBuffer XMesaBufferList = NULL; + + +/** + * Allocate a new XMesaBuffer object which corresponds to the given drawable. + * Note that XMesaBuffer is derived from GLframebuffer. + * The new XMesaBuffer will not have any size (Width=Height=0). + * + * \param d the corresponding X drawable (window or pixmap) + * \param type either WINDOW, PIXMAP or PBUFFER, describing d + * \param vis the buffer's visual + * \param cmap the window's colormap, if known. + * \return new XMesaBuffer or NULL if any problem + */ +static XMesaBuffer +create_xmesa_buffer(XMesaDrawable d, BufferType type, + XMesaVisual vis, XMesaColormap cmap) +{ + XMesaBuffer b; + + ASSERT(type == WINDOW || type == PIXMAP || type == PBUFFER); + + b = (XMesaBuffer) CALLOC_STRUCT(xmesa_buffer); + if (!b) + return NULL; + + b->display = vis->display; + b->xm_visual = vis; + b->type = type; + b->cmap = cmap; + + _mesa_initialize_framebuffer(&b->mesa_buffer, &vis->mesa_visual); + b->mesa_buffer.Delete = xmesa_delete_framebuffer; + + /* + * Front renderbuffer + */ + b->frontxrb = xmesa_create_renderbuffer(NULL, 0, &vis->mesa_visual, GL_FALSE); + if (!b->frontxrb) { + _mesa_free(b); + return NULL; + } + b->frontxrb->Parent = b; + b->frontxrb->drawable = d; + b->frontxrb->pixmap = (XMesaPixmap) d; + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_FRONT_LEFT, + &b->frontxrb->St.Base); + + /* + * Back renderbuffer + */ + if (vis->mesa_visual.doubleBufferMode) { + b->backxrb = xmesa_create_renderbuffer(NULL, 0, &vis->mesa_visual, GL_TRUE); + if (!b->backxrb) { + /* XXX free front xrb too */ + _mesa_free(b); + return NULL; + } + b->backxrb->Parent = b; + /* determine back buffer implementation */ + b->db_mode = vis->ximage_flag ? BACK_XIMAGE : BACK_PIXMAP; + + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_BACK_LEFT, + &b->backxrb->St.Base); + } + + /* + * Software alpha planes + */ + if (vis->mesa_visual.alphaBits > 0 + && vis->undithered_pf != PF_8A8B8G8R + && vis->undithered_pf != PF_8A8R8G8B) { + /* Visual has alpha, but pixel format doesn't support it. + * We'll use an alpha renderbuffer wrapper. + */ + b->swAlpha = GL_TRUE; + } + else { + b->swAlpha = GL_FALSE; + } + + if (vis->mesa_visual.depthBits > 0 && + vis->mesa_visual.stencilBits > 0) { + /* combined depth/stencil */ + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_DEPTH, rb); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_STENCIL, rb); + } + else { + if (vis->mesa_visual.depthBits > 0) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_DEPTH, rb); + } + + if (vis->mesa_visual.stencilBits > 0) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_STENCIL, rb); + } + } + + if (vis->mesa_visual.accumRedBits > 0) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_RGBA16); + _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_ACCUM, rb); + } + + + /* + * Other renderbuffer (depth, stencil, etc) + */ + _mesa_add_soft_renderbuffers(&b->mesa_buffer, + GL_FALSE, /* color */ + GL_FALSE, /*vis->mesa_visual.haveDepthBuffer,*/ + GL_FALSE, /* stencil */ + GL_FALSE, /* accum */ + b->swAlpha, + vis->mesa_visual.numAuxBuffers > 0 ); + + /* GLX_EXT_texture_from_pixmap */ + b->TextureTarget = 0; + b->TextureFormat = GLX_TEXTURE_FORMAT_NONE_EXT; + b->TextureMipmap = 0; + + /* insert buffer into linked list */ + b->Next = XMesaBufferList; + XMesaBufferList = b; + + return b; +} + + +/** + * Find an XMesaBuffer by matching X display and colormap but NOT matching + * the notThis buffer. + */ +XMesaBuffer +xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis) +{ + XMesaBuffer b; + for (b=XMesaBufferList; b; b=b->Next) { + if (b->display==dpy && b->cmap==cmap && b!=notThis) { + return b; + } + } + return NULL; +} + + +/** + * Remove buffer from linked list, delete if no longer referenced. + */ +static void +xmesa_free_buffer(XMesaBuffer buffer) +{ + XMesaBuffer prev = NULL, b; + + for (b = XMesaBufferList; b; b = b->Next) { + if (b == buffer) { + struct gl_framebuffer *fb = &buffer->mesa_buffer; + + /* unlink buffer from list */ + if (prev) + prev->Next = buffer->Next; + else + XMesaBufferList = buffer->Next; + + /* mark as delete pending */ + fb->DeletePending = GL_TRUE; + + /* Since the X window for the XMesaBuffer is going away, we don't + * want to dereference this pointer in the future. + */ + b->frontxrb->drawable = 0; + + /* Unreference. If count = zero we'll really delete the buffer */ + _mesa_unreference_framebuffer(&fb); + + return; + } + /* continue search */ + prev = b; + } + /* buffer not found in XMesaBufferList */ + _mesa_problem(NULL,"xmesa_free_buffer() - buffer not found\n"); +} + + +/** + * Copy X color table stuff from one XMesaBuffer to another. + */ +static void +copy_colortable_info(XMesaBuffer dst, const XMesaBuffer src) +{ + MEMCPY(dst->color_table, src->color_table, sizeof(src->color_table)); + MEMCPY(dst->pixel_to_r, src->pixel_to_r, sizeof(src->pixel_to_r)); + MEMCPY(dst->pixel_to_g, src->pixel_to_g, sizeof(src->pixel_to_g)); + MEMCPY(dst->pixel_to_b, src->pixel_to_b, sizeof(src->pixel_to_b)); + dst->num_alloced = src->num_alloced; + MEMCPY(dst->alloced_colors, src->alloced_colors, + sizeof(src->alloced_colors)); +} + + + +/**********************************************************************/ +/***** Misc Private Functions *****/ +/**********************************************************************/ + + +/** + * A replacement for XAllocColor. This function should never + * fail to allocate a color. When XAllocColor fails, we return + * the nearest matching color. If we have to allocate many colors + * this function isn't too efficient; the XQueryColors() could be + * done just once. + * Written by Michael Pichler, Brian Paul, Mark Kilgard + * Input: dpy - X display + * cmap - X colormap + * cmapSize - size of colormap + * In/Out: color - the XColor struct + * Output: exact - 1=exact color match, 0=closest match + * alloced - 1=XAlloc worked, 0=XAlloc failed + */ +static void +noFaultXAllocColor( int client, + XMesaDisplay *dpy, + XMesaColormap cmap, + int cmapSize, + XMesaColor *color, + int *exact, int *alloced ) +{ +#ifdef XFree86Server + Pixel *ppixIn; + xrgb *ctable; +#else + /* we'll try to cache ctable for better remote display performance */ + static Display *prevDisplay = NULL; + static XMesaColormap prevCmap = 0; + static int prevCmapSize = 0; + static XMesaColor *ctable = NULL; +#endif + XMesaColor subColor; + int i, bestmatch; + double mindist; /* 3*2^16^2 exceeds long int precision. */ + + (void) client; + + /* First try just using XAllocColor. */ +#ifdef XFree86Server + if (AllocColor(cmap, + &color->red, &color->green, &color->blue, + &color->pixel, + client) == Success) +#else + if (XAllocColor(dpy, cmap, color)) +#endif + { + *exact = 1; + *alloced = 1; + return; + } + + /* Alloc failed, search for closest match */ + + /* Retrieve color table entries. */ + /* XXX alloca candidate. */ +#ifdef XFree86Server + ppixIn = (Pixel *) MALLOC(cmapSize * sizeof(Pixel)); + ctable = (xrgb *) MALLOC(cmapSize * sizeof(xrgb)); + for (i = 0; i < cmapSize; i++) { + ppixIn[i] = i; + } + QueryColors(cmap, cmapSize, ppixIn, ctable); +#else + if (prevDisplay != dpy || prevCmap != cmap + || prevCmapSize != cmapSize || !ctable) { + /* free previously cached color table */ + if (ctable) + _mesa_free(ctable); + /* Get the color table from X */ + ctable = (XMesaColor *) MALLOC(cmapSize * sizeof(XMesaColor)); + assert(ctable); + for (i = 0; i < cmapSize; i++) { + ctable[i].pixel = i; + } + XQueryColors(dpy, cmap, ctable, cmapSize); + prevDisplay = dpy; + prevCmap = cmap; + prevCmapSize = cmapSize; + } +#endif + + /* Find best match. */ + bestmatch = -1; + mindist = 0.0; + for (i = 0; i < cmapSize; i++) { + double dr = 0.30 * ((double) color->red - (double) ctable[i].red); + double dg = 0.59 * ((double) color->green - (double) ctable[i].green); + double db = 0.11 * ((double) color->blue - (double) ctable[i].blue); + double dist = dr * dr + dg * dg + db * db; + if (bestmatch < 0 || dist < mindist) { + bestmatch = i; + mindist = dist; + } + } + + /* Return result. */ + subColor.red = ctable[bestmatch].red; + subColor.green = ctable[bestmatch].green; + subColor.blue = ctable[bestmatch].blue; + /* Try to allocate the closest match color. This should only + * fail if the cell is read/write. Otherwise, we're incrementing + * the cell's reference count. + */ +#ifdef XFree86Server + if (AllocColor(cmap, + &subColor.red, &subColor.green, &subColor.blue, + &subColor.pixel, + client) == Success) { +#else + if (XAllocColor(dpy, cmap, &subColor)) { +#endif + *alloced = 1; + } + else { + /* do this to work around a problem reported by Frank Ortega */ + subColor.pixel = (unsigned long) bestmatch; + subColor.red = ctable[bestmatch].red; + subColor.green = ctable[bestmatch].green; + subColor.blue = ctable[bestmatch].blue; + subColor.flags = DoRed | DoGreen | DoBlue; + *alloced = 0; + } +#ifdef XFree86Server + _mesa_free(ppixIn); + _mesa_free(ctable); +#else + /* don't free table, save it for next time */ +#endif + + *color = subColor; + *exact = 0; +} + + + +/** + * Do setup for PF_GRAYSCALE pixel format. + * Note that buffer may be NULL. + */ +static GLboolean +setup_grayscale(int client, XMesaVisual v, + XMesaBuffer buffer, XMesaColormap cmap) +{ + if (GET_VISUAL_DEPTH(v)<4 || GET_VISUAL_DEPTH(v)>16) { + return GL_FALSE; + } + + if (buffer) { + XMesaBuffer prevBuffer; + + if (!cmap) { + return GL_FALSE; + } + + prevBuffer = xmesa_find_buffer(v->display, cmap, buffer); + if (prevBuffer && + (buffer->xm_visual->mesa_visual.rgbMode == + prevBuffer->xm_visual->mesa_visual.rgbMode)) { + /* Copy colormap stuff from previous XMesaBuffer which uses same + * X colormap. Do this to avoid time spent in noFaultXAllocColor. + */ + copy_colortable_info(buffer, prevBuffer); + } + else { + /* Allocate 256 shades of gray */ + int gray; + int colorsfailed = 0; + for (gray=0;gray<256;gray++) { + GLint r = gamma_adjust( v->RedGamma, gray, 255 ); + GLint g = gamma_adjust( v->GreenGamma, gray, 255 ); + GLint b = gamma_adjust( v->BlueGamma, gray, 255 ); + int exact, alloced; + XMesaColor xcol; + xcol.red = (r << 8) | r; + xcol.green = (g << 8) | g; + xcol.blue = (b << 8) | b; + noFaultXAllocColor( client, v->display, + cmap, GET_COLORMAP_SIZE(v), + &xcol, &exact, &alloced ); + if (!exact) { + colorsfailed++; + } + if (alloced) { + assert(buffer->num_alloced<256); + buffer->alloced_colors[buffer->num_alloced] = xcol.pixel; + buffer->num_alloced++; + } + + /*OLD + assert(gray < 576); + buffer->color_table[gray*3+0] = xcol.pixel; + buffer->color_table[gray*3+1] = xcol.pixel; + buffer->color_table[gray*3+2] = xcol.pixel; + assert(xcol.pixel < 65536); + buffer->pixel_to_r[xcol.pixel] = gray * 30 / 100; + buffer->pixel_to_g[xcol.pixel] = gray * 59 / 100; + buffer->pixel_to_b[xcol.pixel] = gray * 11 / 100; + */ + buffer->color_table[gray] = xcol.pixel; + assert(xcol.pixel < 65536); + buffer->pixel_to_r[xcol.pixel] = gray; + buffer->pixel_to_g[xcol.pixel] = gray; + buffer->pixel_to_b[xcol.pixel] = gray; + } + + if (colorsfailed && _mesa_getenv("MESA_DEBUG")) { + _mesa_warning(NULL, + "Note: %d out of 256 needed colors do not match exactly.\n", + colorsfailed ); + } + } + } + + v->dithered_pf = PF_Grayscale; + v->undithered_pf = PF_Grayscale; + return GL_TRUE; +} + + + +/** + * Setup RGB rendering for a window with a PseudoColor, StaticColor, + * or 8-bit TrueColor visual visual. We try to allocate a palette of 225 + * colors (5 red, 9 green, 5 blue) and dither to approximate a 24-bit RGB + * color. While this function was originally designed just for 8-bit + * visuals, it has also proven to work from 4-bit up to 16-bit visuals. + * Dithering code contributed by Bob Mercier. + */ +static GLboolean +setup_dithered_color(int client, XMesaVisual v, + XMesaBuffer buffer, XMesaColormap cmap) +{ + if (GET_VISUAL_DEPTH(v)<4 || GET_VISUAL_DEPTH(v)>16) { + return GL_FALSE; + } + + if (buffer) { + XMesaBuffer prevBuffer; + + if (!cmap) { + return GL_FALSE; + } + + prevBuffer = xmesa_find_buffer(v->display, cmap, buffer); + if (prevBuffer && + (buffer->xm_visual->mesa_visual.rgbMode == + prevBuffer->xm_visual->mesa_visual.rgbMode)) { + /* Copy colormap stuff from previous, matching XMesaBuffer. + * Do this to avoid time spent in noFaultXAllocColor. + */ + copy_colortable_info(buffer, prevBuffer); + } + else { + /* Allocate X colors and initialize color_table[], red_table[], etc */ + int r, g, b, i; + int colorsfailed = 0; + for (r = 0; r < DITH_R; r++) { + for (g = 0; g < DITH_G; g++) { + for (b = 0; b < DITH_B; b++) { + XMesaColor xcol; + int exact, alloced; + xcol.red =gamma_adjust(v->RedGamma, r*65535/(DITH_R-1),65535); + xcol.green=gamma_adjust(v->GreenGamma, g*65535/(DITH_G-1),65535); + xcol.blue =gamma_adjust(v->BlueGamma, b*65535/(DITH_B-1),65535); + noFaultXAllocColor( client, v->display, + cmap, GET_COLORMAP_SIZE(v), + &xcol, &exact, &alloced ); + if (!exact) { + colorsfailed++; + } + if (alloced) { + assert(buffer->num_alloced<256); + buffer->alloced_colors[buffer->num_alloced] = xcol.pixel; + buffer->num_alloced++; + } + i = DITH_MIX( r, g, b ); + assert(i < 576); + buffer->color_table[i] = xcol.pixel; + assert(xcol.pixel < 65536); + buffer->pixel_to_r[xcol.pixel] = r * 255 / (DITH_R-1); + buffer->pixel_to_g[xcol.pixel] = g * 255 / (DITH_G-1); + buffer->pixel_to_b[xcol.pixel] = b * 255 / (DITH_B-1); + } + } + } + + if (colorsfailed && _mesa_getenv("MESA_DEBUG")) { + _mesa_warning(NULL, + "Note: %d out of %d needed colors do not match exactly.\n", + colorsfailed, DITH_R * DITH_G * DITH_B ); + } + } + } + + v->dithered_pf = PF_Dither; + v->undithered_pf = PF_Lookup; + return GL_TRUE; +} + + +/** + * Setup for Hewlett Packard Color Recovery 8-bit TrueColor mode. + * HPCR simulates 24-bit color fidelity with an 8-bit frame buffer. + * Special dithering tables have to be initialized. + */ +static void +setup_8bit_hpcr(XMesaVisual v) +{ + /* HP Color Recovery contributed by: Alex De Bruyn (ad@lms.be) + * To work properly, the atom _HP_RGB_SMOOTH_MAP_LIST must be defined + * on the root window AND the colormap obtainable by XGetRGBColormaps + * for that atom must be set on the window. (see also tkInitWindow) + * If that colormap is not set, the output will look stripy. + */ + + /* Setup color tables with gamma correction */ + int i; + double g; + + g = 1.0 / v->RedGamma; + for (i=0; i<256; i++) { + GLint red = IROUND_POS(255.0 * _mesa_pow( hpcr_rgbTbl[0][i]/255.0, g )); + v->hpcr_rgbTbl[0][i] = CLAMP( red, 16, 239 ); + } + + g = 1.0 / v->GreenGamma; + for (i=0; i<256; i++) { + GLint green = IROUND_POS(255.0 * _mesa_pow( hpcr_rgbTbl[1][i]/255.0, g )); + v->hpcr_rgbTbl[1][i] = CLAMP( green, 16, 239 ); + } + + g = 1.0 / v->BlueGamma; + for (i=0; i<256; i++) { + GLint blue = IROUND_POS(255.0 * _mesa_pow( hpcr_rgbTbl[2][i]/255.0, g )); + v->hpcr_rgbTbl[2][i] = CLAMP( blue, 32, 223 ); + } + v->undithered_pf = PF_HPCR; /* can't really disable dithering for now */ + v->dithered_pf = PF_HPCR; + + /* which method should I use to clear */ + /* GL_FALSE: keep the ordinary method */ + /* GL_TRUE : clear with dither pattern */ + v->hpcr_clear_flag = _mesa_getenv("MESA_HPCR_CLEAR") ? GL_TRUE : GL_FALSE; + + if (v->hpcr_clear_flag) { + v->hpcr_clear_pixmap = XMesaCreatePixmap(v->display, + DefaultRootWindow(v->display), + 16, 2, 8); +#ifndef XFree86Server + v->hpcr_clear_ximage = XGetImage(v->display, v->hpcr_clear_pixmap, + 0, 0, 16, 2, AllPlanes, ZPixmap); +#endif + } +} + + +/** + * Setup RGB rendering for a window with a True/DirectColor visual. + */ +static void +setup_truecolor(XMesaVisual v, XMesaBuffer buffer, XMesaColormap cmap) +{ + unsigned long rmask, gmask, bmask; + (void) buffer; + (void) cmap; + + /* Compute red multiplier (mask) and bit shift */ + v->rshift = 0; + rmask = GET_REDMASK(v); + while ((rmask & 1)==0) { + v->rshift++; + rmask = rmask >> 1; + } + + /* Compute green multiplier (mask) and bit shift */ + v->gshift = 0; + gmask = GET_GREENMASK(v); + while ((gmask & 1)==0) { + v->gshift++; + gmask = gmask >> 1; + } + + /* Compute blue multiplier (mask) and bit shift */ + v->bshift = 0; + bmask = GET_BLUEMASK(v); + while ((bmask & 1)==0) { + v->bshift++; + bmask = bmask >> 1; + } + + /* + * Compute component-to-pixel lookup tables and dithering kernel + */ + { + static GLubyte kernel[16] = { + 0*16, 8*16, 2*16, 10*16, + 12*16, 4*16, 14*16, 6*16, + 3*16, 11*16, 1*16, 9*16, + 15*16, 7*16, 13*16, 5*16, + }; + GLint rBits = _mesa_bitcount(rmask); + GLint gBits = _mesa_bitcount(gmask); + GLint bBits = _mesa_bitcount(bmask); + GLint maxBits; + GLuint i; + + /* convert pixel components in [0,_mask] to RGB values in [0,255] */ + for (i=0; i<=rmask; i++) + v->PixelToR[i] = (unsigned char) ((i * 255) / rmask); + for (i=0; i<=gmask; i++) + v->PixelToG[i] = (unsigned char) ((i * 255) / gmask); + for (i=0; i<=bmask; i++) + v->PixelToB[i] = (unsigned char) ((i * 255) / bmask); + + /* convert RGB values from [0,255] to pixel components */ + + for (i=0;i<256;i++) { + GLint r = gamma_adjust(v->RedGamma, i, 255); + GLint g = gamma_adjust(v->GreenGamma, i, 255); + GLint b = gamma_adjust(v->BlueGamma, i, 255); + v->RtoPixel[i] = (r >> (8-rBits)) << v->rshift; + v->GtoPixel[i] = (g >> (8-gBits)) << v->gshift; + v->BtoPixel[i] = (b >> (8-bBits)) << v->bshift; + } + /* overflow protection */ + for (i=256;i<512;i++) { + v->RtoPixel[i] = v->RtoPixel[255]; + v->GtoPixel[i] = v->GtoPixel[255]; + v->BtoPixel[i] = v->BtoPixel[255]; + } + + /* setup dithering kernel */ + maxBits = rBits; + if (gBits > maxBits) maxBits = gBits; + if (bBits > maxBits) maxBits = bBits; + for (i=0;i<16;i++) { + v->Kernel[i] = kernel[i] >> maxBits; + } + + v->undithered_pf = PF_Truecolor; + v->dithered_pf = (GET_VISUAL_DEPTH(v)<24) ? PF_Dither_True : PF_Truecolor; + } + + /* + * Now check for TrueColor visuals which we can optimize. + */ + if ( GET_REDMASK(v) ==0x0000ff + && GET_GREENMASK(v)==0x00ff00 + && GET_BLUEMASK(v) ==0xff0000 + && CHECK_BYTE_ORDER(v) + && v->BitsPerPixel==32 + && v->RedGamma==1.0 && v->GreenGamma==1.0 && v->BlueGamma==1.0) { + /* common 32 bpp config used on SGI, Sun */ + v->undithered_pf = v->dithered_pf = PF_8A8B8G8R; /* ABGR */ + } + else if (GET_REDMASK(v) == 0xff0000 + && GET_GREENMASK(v)== 0x00ff00 + && GET_BLUEMASK(v) == 0x0000ff + && CHECK_BYTE_ORDER(v) + && v->RedGamma == 1.0 && v->GreenGamma == 1.0 && v->BlueGamma == 1.0){ + if (v->BitsPerPixel==32) { + /* if 32 bpp, and visual indicates 8 bpp alpha channel */ + if (GET_VISUAL_DEPTH(v) == 32 && v->mesa_visual.alphaBits == 8) + v->undithered_pf = v->dithered_pf = PF_8A8R8G8B; /* ARGB */ + else + v->undithered_pf = v->dithered_pf = PF_8R8G8B; /* xRGB */ + } + else if (v->BitsPerPixel == 24) { + v->undithered_pf = v->dithered_pf = PF_8R8G8B24; /* RGB */ + } + } + else if (GET_REDMASK(v) ==0xf800 + && GET_GREENMASK(v)==0x07e0 + && GET_BLUEMASK(v) ==0x001f + && CHECK_BYTE_ORDER(v) + && v->BitsPerPixel==16 + && v->RedGamma==1.0 && v->GreenGamma==1.0 && v->BlueGamma==1.0) { + /* 5-6-5 RGB */ + v->undithered_pf = PF_5R6G5B; + v->dithered_pf = PF_Dither_5R6G5B; + } + else if (GET_REDMASK(v) ==0xe0 + && GET_GREENMASK(v)==0x1c + && GET_BLUEMASK(v) ==0x03 + && CHECK_FOR_HPCR(v)) { + /* 8-bit HP color recovery */ + setup_8bit_hpcr( v ); + } +} + + + +/** + * Setup RGB rendering for a window with a monochrome visual. + */ +static void +setup_monochrome( XMesaVisual v, XMesaBuffer b ) +{ + (void) b; + v->dithered_pf = v->undithered_pf = PF_1Bit; + /* if black=1 then we must flip pixel values */ + v->bitFlip = (GET_BLACK_PIXEL(v) != 0); +} + + + +/** + * When a context is bound for the first time, we can finally finish + * initializing the context's visual and buffer information. + * \param v the XMesaVisual to initialize + * \param b the XMesaBuffer to initialize (may be NULL) + * \param rgb_flag TRUE = RGBA mode, FALSE = color index mode + * \param window the window/pixmap we're rendering into + * \param cmap the colormap associated with the window/pixmap + * \return GL_TRUE=success, GL_FALSE=failure + */ +static GLboolean +initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, + GLboolean rgb_flag, XMesaDrawable window, + XMesaColormap cmap) +{ + int client = 0; + +#ifdef XFree86Server + client = (window) ? CLIENT_ID(window->id) : 0; +#endif + + ASSERT(!b || b->xm_visual == v); + + /* Save true bits/pixel */ + v->BitsPerPixel = bits_per_pixel(v); + assert(v->BitsPerPixel > 0); + + if (rgb_flag == GL_FALSE) { + /* COLOR-INDEXED WINDOW: + * Even if the visual is TrueColor or DirectColor we treat it as + * being color indexed. This is weird but might be useful to someone. + */ + v->dithered_pf = v->undithered_pf = PF_Index; + v->mesa_visual.indexBits = GET_VISUAL_DEPTH(v); + } + else { + /* RGB WINDOW: + * We support RGB rendering into almost any kind of visual. + */ + const int xclass = v->mesa_visual.visualType; + if (xclass == GLX_TRUE_COLOR || xclass == GLX_DIRECT_COLOR) { + setup_truecolor( v, b, cmap ); + } + else if (xclass == GLX_STATIC_GRAY && GET_VISUAL_DEPTH(v) == 1) { + setup_monochrome( v, b ); + } + else if (xclass == GLX_GRAY_SCALE || xclass == GLX_STATIC_GRAY) { + if (!setup_grayscale( client, v, b, cmap )) { + return GL_FALSE; + } + } + else if ((xclass == GLX_PSEUDO_COLOR || xclass == GLX_STATIC_COLOR) + && GET_VISUAL_DEPTH(v)>=4 && GET_VISUAL_DEPTH(v)<=16) { + if (!setup_dithered_color( client, v, b, cmap )) { + return GL_FALSE; + } + } + else { + _mesa_warning(NULL, "XMesa: RGB mode rendering not supported in given visual.\n"); + return GL_FALSE; + } + v->mesa_visual.indexBits = 0; + + if (_mesa_getenv("MESA_NO_DITHER")) { + v->dithered_pf = v->undithered_pf; + } + } + + + /* + * If MESA_INFO env var is set print out some debugging info + * which can help Brian figure out what's going on when a user + * reports bugs. + */ + if (_mesa_getenv("MESA_INFO")) { + _mesa_printf("X/Mesa visual = %p\n", (void *) v); + _mesa_printf("X/Mesa dithered pf = %u\n", v->dithered_pf); + _mesa_printf("X/Mesa undithered pf = %u\n", v->undithered_pf); + _mesa_printf("X/Mesa level = %d\n", v->mesa_visual.level); + _mesa_printf("X/Mesa depth = %d\n", GET_VISUAL_DEPTH(v)); + _mesa_printf("X/Mesa bits per pixel = %d\n", v->BitsPerPixel); + } + + if (b && window) { + char *data; + + /* Do window-specific initializations */ + + /* these should have been set in create_xmesa_buffer */ + ASSERT(b->frontxrb->drawable == window); + ASSERT(b->frontxrb->pixmap == (XMesaPixmap) window); + + /* Setup for single/double buffering */ + if (v->mesa_visual.doubleBufferMode) { + /* Double buffered */ + b->shm = check_for_xshm( v->display ); + } + + /* X11 graphics contexts */ +#ifdef XFree86Server + b->gc = CreateScratchGC(v->display, window->depth); +#else + b->gc = XCreateGC( v->display, window, 0, NULL ); +#endif + XMesaSetFunction( v->display, b->gc, GXcopy ); + + /* cleargc - for glClear() */ +#ifdef XFree86Server + b->cleargc = CreateScratchGC(v->display, window->depth); +#else + b->cleargc = XCreateGC( v->display, window, 0, NULL ); +#endif + XMesaSetFunction( v->display, b->cleargc, GXcopy ); + + /* + * Don't generate Graphics Expose/NoExpose events in swapbuffers(). + * Patch contributed by Michael Pichler May 15, 1995. + */ +#ifdef XFree86Server + b->swapgc = CreateScratchGC(v->display, window->depth); + { + CARD32 v[1]; + v[0] = FALSE; + dixChangeGC(NullClient, b->swapgc, GCGraphicsExposures, v, NULL); + } +#else + { + XGCValues gcvalues; + gcvalues.graphics_exposures = False; + b->swapgc = XCreateGC(v->display, window, + GCGraphicsExposures, &gcvalues); + } +#endif + XMesaSetFunction( v->display, b->swapgc, GXcopy ); + /* + * Set fill style and tile pixmap once for all for HPCR stuff + * (instead of doing it each time in clear_color_HPCR_pixmap()) + * Initialize whole stuff + * Patch contributed by Jacques Leroy March 8, 1998. + */ + if (v->hpcr_clear_flag && b->backxrb && b->backxrb->pixmap) { + int i; + for (i = 0; i < 16; i++) { + XMesaPutPixel(v->hpcr_clear_ximage, i, 0, 0); + XMesaPutPixel(v->hpcr_clear_ximage, i, 1, 0); + } + XMesaPutImage(b->display, (XMesaDrawable) v->hpcr_clear_pixmap, + b->cleargc, v->hpcr_clear_ximage, 0, 0, 0, 0, 16, 2); + XMesaSetFillStyle( v->display, b->cleargc, FillTiled); + XMesaSetTile( v->display, b->cleargc, v->hpcr_clear_pixmap ); + } + + /* Initialize the row buffer XImage for use in write_color_span() */ + data = (char*) MALLOC(MAX_WIDTH*4); +#ifdef XFree86Server + b->rowimage = XMesaCreateImage(GET_VISUAL_DEPTH(v), MAX_WIDTH, 1, data); +#else + b->rowimage = XCreateImage( v->display, + v->visinfo->visual, + v->visinfo->depth, + ZPixmap, 0, /*format, offset*/ + data, /*data*/ + MAX_WIDTH, 1, /*width, height*/ + 32, /*bitmap_pad*/ + 0 /*bytes_per_line*/ ); +#endif + if (!b->rowimage) + return GL_FALSE; + } + + return GL_TRUE; +} + + + +/* + * Convert an RGBA color to a pixel value. + */ +unsigned long +xmesa_color_to_pixel(GLcontext *ctx, + GLubyte r, GLubyte g, GLubyte b, GLubyte a, + GLuint pixelFormat) +{ + XMesaContext xmesa = XMESA_CONTEXT(ctx); + switch (pixelFormat) { + case PF_Index: + return 0; + case PF_Truecolor: + { + unsigned long p; + PACK_TRUECOLOR( p, r, g, b ); + return p; + } + case PF_8A8B8G8R: + return PACK_8A8B8G8R( r, g, b, a ); + case PF_8A8R8G8B: + return PACK_8A8R8G8B( r, g, b, a ); + case PF_8R8G8B: + /* fall through */ + case PF_8R8G8B24: + return PACK_8R8G8B( r, g, b ); + case PF_5R6G5B: + return PACK_5R6G5B( r, g, b ); + case PF_Dither: + { + DITHER_SETUP; + return DITHER( 1, 0, r, g, b ); + } + case PF_1Bit: + /* 382 = (3*255)/2 */ + return ((r+g+b) > 382) ^ xmesa->xm_visual->bitFlip; + case PF_HPCR: + return DITHER_HPCR(1, 1, r, g, b); + case PF_Lookup: + { + LOOKUP_SETUP; + return LOOKUP( r, g, b ); + } + case PF_Grayscale: + return GRAY_RGB( r, g, b ); + case PF_Dither_True: + /* fall through */ + case PF_Dither_5R6G5B: + { + unsigned long p; + PACK_TRUEDITHER(p, 1, 0, r, g, b); + return p; + } + default: + _mesa_problem(ctx, "Bad pixel format in xmesa_color_to_pixel"); + } + return 0; +} + + +#define NUM_VISUAL_TYPES 6 + +/** + * Convert an X visual type to a GLX visual type. + * + * \param visualType X visual type (i.e., \c TrueColor, \c StaticGray, etc.) + * to be converted. + * \return If \c visualType is a valid X visual type, a GLX visual type will + * be returned. Otherwise \c GLX_NONE will be returned. + * + * \note + * This code was lifted directly from lib/GL/glx/glcontextmodes.c in the + * DRI CVS tree. + */ +static GLint +xmesa_convert_from_x_visual_type( int visualType ) +{ + static const int glx_visual_types[ NUM_VISUAL_TYPES ] = { + GLX_STATIC_GRAY, GLX_GRAY_SCALE, + GLX_STATIC_COLOR, GLX_PSEUDO_COLOR, + GLX_TRUE_COLOR, GLX_DIRECT_COLOR + }; + + return ( (unsigned) visualType < NUM_VISUAL_TYPES ) + ? glx_visual_types[ visualType ] : GLX_NONE; +} + + +/**********************************************************************/ +/***** Public Functions *****/ +/**********************************************************************/ + + +/* + * Create a new X/Mesa visual. + * Input: display - X11 display + * visinfo - an XVisualInfo pointer + * rgb_flag - GL_TRUE = RGB mode, + * GL_FALSE = color index mode + * alpha_flag - alpha buffer requested? + * db_flag - GL_TRUE = double-buffered, + * GL_FALSE = single buffered + * stereo_flag - stereo visual? + * ximage_flag - GL_TRUE = use an XImage for back buffer, + * GL_FALSE = use an off-screen pixmap for back buffer + * depth_size - requested bits/depth values, or zero + * stencil_size - requested bits/stencil values, or zero + * accum_red_size - requested bits/red accum values, or zero + * accum_green_size - requested bits/green accum values, or zero + * accum_blue_size - requested bits/blue accum values, or zero + * accum_alpha_size - requested bits/alpha accum values, or zero + * num_samples - number of samples/pixel if multisampling, or zero + * level - visual level, usually 0 + * visualCaveat - ala the GLX extension, usually GLX_NONE + * Return; a new XMesaVisual or 0 if error. + */ +PUBLIC +XMesaVisual XMesaCreateVisual( XMesaDisplay *display, + XMesaVisualInfo visinfo, + GLboolean rgb_flag, + GLboolean alpha_flag, + GLboolean db_flag, + GLboolean stereo_flag, + GLboolean ximage_flag, + GLint depth_size, + GLint stencil_size, + GLint accum_red_size, + GLint accum_green_size, + GLint accum_blue_size, + GLint accum_alpha_size, + GLint num_samples, + GLint level, + GLint visualCaveat ) +{ + char *gamma; + XMesaVisual v; + GLint red_bits, green_bits, blue_bits, alpha_bits; + +#ifndef XFree86Server + /* For debugging only */ + if (_mesa_getenv("MESA_XSYNC")) { + /* This makes debugging X easier. + * In your debugger, set a breakpoint on _XError to stop when an + * X protocol error is generated. + */ + XSynchronize( display, 1 ); + } +#endif + + v = (XMesaVisual) CALLOC_STRUCT(xmesa_visual); + if (!v) { + return NULL; + } + + v->display = display; + + /* Save a copy of the XVisualInfo struct because the user may X_mesa_free() + * the struct but we may need some of the information contained in it + * at a later time. + */ +#ifndef XFree86Server + v->visinfo = (XVisualInfo *) MALLOC(sizeof(*visinfo)); + if(!v->visinfo) { + _mesa_free(v); + return NULL; + } + MEMCPY(v->visinfo, visinfo, sizeof(*visinfo)); +#endif + + /* check for MESA_GAMMA environment variable */ + gamma = _mesa_getenv("MESA_GAMMA"); + if (gamma) { + v->RedGamma = v->GreenGamma = v->BlueGamma = 0.0; + sscanf( gamma, "%f %f %f", &v->RedGamma, &v->GreenGamma, &v->BlueGamma ); + if (v->RedGamma<=0.0) v->RedGamma = 1.0; + if (v->GreenGamma<=0.0) v->GreenGamma = v->RedGamma; + if (v->BlueGamma<=0.0) v->BlueGamma = v->RedGamma; + } + else { + v->RedGamma = v->GreenGamma = v->BlueGamma = 1.0; + } + + v->ximage_flag = ximage_flag; + +#ifdef XFree86Server + /* We could calculate these values by ourselves. nplanes is either the sum + * of the red, green, and blue bits or the number index bits. + * ColormapEntries is either (1U << index_bits) or + * (1U << max(redBits, greenBits, blueBits)). + */ + assert(visinfo->nplanes > 0); + v->nplanes = visinfo->nplanes; + v->ColormapEntries = visinfo->ColormapEntries; + + v->mesa_visual.redMask = visinfo->redMask; + v->mesa_visual.greenMask = visinfo->greenMask; + v->mesa_visual.blueMask = visinfo->blueMask; + v->mesa_visual.visualID = visinfo->vid; + v->mesa_visual.screen = 0; /* FIXME: What should be done here? */ +#else + v->mesa_visual.redMask = visinfo->red_mask; + v->mesa_visual.greenMask = visinfo->green_mask; + v->mesa_visual.blueMask = visinfo->blue_mask; + v->mesa_visual.visualID = visinfo->visualid; + v->mesa_visual.screen = visinfo->screen; +#endif + +#if defined(XFree86Server) || !(defined(__cplusplus) || defined(c_plusplus)) + v->mesa_visual.visualType = xmesa_convert_from_x_visual_type(visinfo->class); +#else + v->mesa_visual.visualType = xmesa_convert_from_x_visual_type(visinfo->c_class); +#endif + + v->mesa_visual.visualRating = visualCaveat; + + if (alpha_flag) + v->mesa_visual.alphaBits = 8; + + (void) initialize_visual_and_buffer( v, NULL, rgb_flag, 0, 0 ); + + { + const int xclass = v->mesa_visual.visualType; + if (xclass == GLX_TRUE_COLOR || xclass == GLX_DIRECT_COLOR) { + red_bits = _mesa_bitcount(GET_REDMASK(v)); + green_bits = _mesa_bitcount(GET_GREENMASK(v)); + blue_bits = _mesa_bitcount(GET_BLUEMASK(v)); + } + else { + /* this is an approximation */ + int depth; + depth = GET_VISUAL_DEPTH(v); + red_bits = depth / 3; + depth -= red_bits; + green_bits = depth / 2; + depth -= green_bits; + blue_bits = depth; + alpha_bits = 0; + assert( red_bits + green_bits + blue_bits == GET_VISUAL_DEPTH(v) ); + } + alpha_bits = v->mesa_visual.alphaBits; + } + + _mesa_initialize_visual( &v->mesa_visual, + rgb_flag, db_flag, stereo_flag, + red_bits, green_bits, + blue_bits, alpha_bits, + v->mesa_visual.indexBits, + depth_size, + stencil_size, + accum_red_size, accum_green_size, + accum_blue_size, accum_alpha_size, + 0 ); + + /* XXX minor hack */ + v->mesa_visual.level = level; + return v; +} + + +PUBLIC +void XMesaDestroyVisual( XMesaVisual v ) +{ +#ifndef XFree86Server + _mesa_free(v->visinfo); +#endif + _mesa_free(v); +} + + + +/** + * Create a new XMesaContext. + * \param v the XMesaVisual + * \param share_list another XMesaContext with which to share display + * lists or NULL if no sharing is wanted. + * \return an XMesaContext or NULL if error. + */ +PUBLIC +XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) +{ + static GLboolean firstTime = GL_TRUE; + XMesaContext c; + GLcontext *mesaCtx; + struct dd_function_table functions; +#if 0 + TNLcontext *tnl; +#endif + + if (firstTime) { + _glthread_INIT_MUTEX(_xmesa_lock); + firstTime = GL_FALSE; + } + + /* Note: the XMesaContext contains a Mesa GLcontext struct (inheritance) */ + c = (XMesaContext) CALLOC_STRUCT(xmesa_context); + if (!c) + return NULL; + + mesaCtx = &(c->mesa); + + /* initialize with default driver functions, then plug in XMesa funcs */ + _mesa_init_driver_functions(&functions); + xmesa_init_driver_functions(v, &functions); + st_init_driver_functions(&functions); + + /* override st's function */ + functions.UpdateState = xmesa_update_state; + + /* + functions.NewRenderbuffer = xmesa_new_renderbuffer; + */ + + if (!_mesa_initialize_context(mesaCtx, &v->mesa_visual, + share_list ? &(share_list->mesa) : (GLcontext *) NULL, + &functions, (void *) c)) { + _mesa_free(c); + return NULL; + } + + _mesa_enable_sw_extensions(mesaCtx); + _mesa_enable_1_3_extensions(mesaCtx); + _mesa_enable_1_4_extensions(mesaCtx); + _mesa_enable_1_5_extensions(mesaCtx); + _mesa_enable_2_0_extensions(mesaCtx); +#if ENABLE_EXT_texure_compression_s3tc + if (c->Mesa_DXTn) { + _mesa_enable_extension(mesaCtx, "GL_EXT_texture_compression_s3tc"); + _mesa_enable_extension(mesaCtx, "GL_S3_s3tc"); + } + _mesa_enable_extension(mesaCtx, "GL_3DFX_texture_compression_FXT1"); +#endif +#if ENABLE_EXT_timer_query + _mesa_enable_extension(mesaCtx, "GL_EXT_timer_query"); +#endif + +#ifdef XFree86Server + /* If we're running in the X server, do bounds checking to prevent + * segfaults and server crashes! + */ + mesaCtx->Const.CheckArrayBounds = GL_TRUE; +#endif + + /* finish up xmesa context initializations */ + c->swapbytes = CHECK_BYTE_ORDER(v) ? GL_FALSE : GL_TRUE; + c->xm_visual = v; + c->xm_buffer = NULL; /* set later by XMesaMakeCurrent */ + c->display = v->display; + c->pixelformat = v->dithered_pf; /* Dithering is enabled by default */ + + /* Initialize the software rasterizer and helper modules. + */ + if (!_swrast_CreateContext( mesaCtx ) +#if 0 + || !_vbo_CreateContext( mesaCtx ) || + !_tnl_CreateContext( mesaCtx ) || + !_swsetup_CreateContext( mesaCtx ) +#endif + ) { + _mesa_free_context_data(&c->mesa); + _mesa_free(c); + return NULL; + } + +#if 0 + /* tnl setup */ + tnl = TNL_CONTEXT(mesaCtx); + tnl->Driver.RunPipeline = _tnl_run_pipeline; +#endif + + /* swrast setup */ + xmesa_register_swrast_functions( mesaCtx ); + + + st_create_context( mesaCtx, + xmesa_create_softpipe( c ) ); + + _swsetup_CreateContext( mesaCtx ); + _swsetup_Wakeup(mesaCtx); + + /* override these functions, as if the xlib driver were derived from + * the softpipe driver. + */ +#if 0 + mesaCtx->st->pipe->surface_alloc = xmesa_surface_alloc; +#endif + mesaCtx->st->pipe->is_format_supported = xmesa_is_format_supported; + mesaCtx->st->pipe->get_tile_rgba = xmesa_get_tile_rgba; + mesaCtx->st->pipe->put_tile_rgba = xmesa_put_tile_rgba; + + mesaCtx->st->haveFramebufferRegions = GL_FALSE; + + /* special pipe->clear function */ + mesaCtx->st->pipe->clear = xmesa_clear; + + return c; +} + + + +PUBLIC +void XMesaDestroyContext( XMesaContext c ) +{ + GLcontext *mesaCtx = &c->mesa; + +#ifdef FX + FXdestroyContext( XMESA_BUFFER(mesaCtx->DrawBuffer) ); +#endif + + _swsetup_DestroyContext( mesaCtx ); + _swrast_DestroyContext( mesaCtx ); +#if 0 + _tnl_DestroyContext( mesaCtx ); + _vbo_DestroyContext( mesaCtx ); +#endif + _mesa_free_context_data( mesaCtx ); + _mesa_free( c ); +} + + + +/** + * Private function for creating an XMesaBuffer which corresponds to an + * X window or pixmap. + * \param v the window's XMesaVisual + * \param w the window we're wrapping + * \return new XMesaBuffer or NULL if error + */ +PUBLIC XMesaBuffer +XMesaCreateWindowBuffer(XMesaVisual v, XMesaWindow w) +{ +#ifndef XFree86Server + XWindowAttributes attr; +#endif + XMesaBuffer b; + XMesaColormap cmap; + int depth; + + assert(v); + assert(w); + + /* Check that window depth matches visual depth */ +#ifdef XFree86Server + depth = ((XMesaDrawable)w)->depth; +#else + XGetWindowAttributes( v->display, w, &attr ); + depth = attr.depth; +#endif + if (GET_VISUAL_DEPTH(v) != depth) { + _mesa_warning(NULL, "XMesaCreateWindowBuffer: depth mismatch between visual (%d) and window (%d)!\n", + GET_VISUAL_DEPTH(v), depth); + return NULL; + } + + /* Find colormap */ +#ifdef XFree86Server + cmap = (ColormapPtr)LookupIDByType(wColormap(w), RT_COLORMAP); +#else + if (attr.colormap) { + cmap = attr.colormap; + } + else { + _mesa_warning(NULL, "Window %u has no colormap!\n", (unsigned int) w); + /* this is weird, a window w/out a colormap!? */ + /* OK, let's just allocate a new one and hope for the best */ + cmap = XCreateColormap(v->display, w, attr.visual, AllocNone); + } +#endif + + b = create_xmesa_buffer((XMesaDrawable) w, WINDOW, v, cmap); + if (!b) + return NULL; + + if (!initialize_visual_and_buffer( v, b, v->mesa_visual.rgbMode, + (XMesaDrawable) w, cmap )) { + xmesa_free_buffer(b); + return NULL; + } + + return b; +} + + + +/** + * Create a new XMesaBuffer from an X pixmap. + * + * \param v the XMesaVisual + * \param p the pixmap + * \param cmap the colormap, may be 0 if using a \c GLX_TRUE_COLOR or + * \c GLX_DIRECT_COLOR visual for the pixmap + * \returns new XMesaBuffer or NULL if error + */ +PUBLIC XMesaBuffer +XMesaCreatePixmapBuffer(XMesaVisual v, XMesaPixmap p, XMesaColormap cmap) +{ + XMesaBuffer b; + + assert(v); + + b = create_xmesa_buffer((XMesaDrawable) p, PIXMAP, v, cmap); + if (!b) + return NULL; + + if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode, + (XMesaDrawable) p, cmap)) { + xmesa_free_buffer(b); + return NULL; + } + + return b; +} + + +/** + * For GLX_EXT_texture_from_pixmap + */ +XMesaBuffer +XMesaCreatePixmapTextureBuffer(XMesaVisual v, XMesaPixmap p, + XMesaColormap cmap, + int format, int target, int mipmap) +{ + GET_CURRENT_CONTEXT(ctx); + XMesaBuffer b; + GLuint width, height; + + assert(v); + + b = create_xmesa_buffer((XMesaDrawable) p, PIXMAP, v, cmap); + if (!b) + return NULL; + + /* get pixmap size, update framebuffer/renderbuffer dims */ + xmesa_get_window_size(v->display, b, &width, &height); + _mesa_resize_framebuffer(NULL, &(b->mesa_buffer), width, height); + + if (target == 0) { + /* examine dims */ + if (ctx->Extensions.ARB_texture_non_power_of_two) { + target = GLX_TEXTURE_2D_EXT; + } + else if ( _mesa_bitcount(width) == 1 + && _mesa_bitcount(height) == 1) { + /* power of two size */ + if (height == 1) { + target = GLX_TEXTURE_1D_EXT; + } + else { + target = GLX_TEXTURE_2D_EXT; + } + } + else if (ctx->Extensions.NV_texture_rectangle) { + target = GLX_TEXTURE_RECTANGLE_EXT; + } + else { + /* non power of two textures not supported */ + XMesaDestroyBuffer(b); + return 0; + } + } + + b->TextureTarget = target; + b->TextureFormat = format; + b->TextureMipmap = mipmap; + + if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode, + (XMesaDrawable) p, cmap)) { + xmesa_free_buffer(b); + return NULL; + } + + return b; +} + + + +XMesaBuffer +XMesaCreatePBuffer(XMesaVisual v, XMesaColormap cmap, + unsigned int width, unsigned int height) +{ +#ifndef XFree86Server + XMesaWindow root; + XMesaDrawable drawable; /* X Pixmap Drawable */ + XMesaBuffer b; + + /* allocate pixmap for front buffer */ + root = RootWindow( v->display, v->visinfo->screen ); + drawable = XCreatePixmap(v->display, root, width, height, + v->visinfo->depth); + if (!drawable) + return NULL; + + b = create_xmesa_buffer(drawable, PBUFFER, v, cmap); + if (!b) + return NULL; + + if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode, + drawable, cmap)) { + xmesa_free_buffer(b); + return NULL; + } + + return b; +#else + return 0; +#endif +} + + + +/* + * Deallocate an XMesaBuffer structure and all related info. + */ +PUBLIC void +XMesaDestroyBuffer(XMesaBuffer b) +{ + xmesa_free_buffer(b); +} + + +/** + * Query the current window size and update the corresponding GLframebuffer + * and all attached renderbuffers. + * Called when: + * 1. the first time a buffer is bound to a context. + * 2. from glViewport to poll for window size changes + * 3. from the XMesaResizeBuffers() API function. + * Note: it's possible (and legal) for xmctx to be NULL. That can happen + * when resizing a buffer when no rendering context is bound. + */ +void +xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer) +{ + GLuint width, height; + xmesa_get_window_size(drawBuffer->display, drawBuffer, &width, &height); + if (drawBuffer->mesa_buffer.Width != width || + drawBuffer->mesa_buffer.Height != height) { + GLcontext *ctx = xmctx ? &xmctx->mesa : NULL; + _mesa_resize_framebuffer(ctx, &(drawBuffer->mesa_buffer), width, height); + } + drawBuffer->mesa_buffer.Initialized = GL_TRUE; /* XXX TEMPORARY? */ +} + + +/* + * Bind buffer b to context c and make c the current rendering context. + */ +GLboolean XMesaMakeCurrent( XMesaContext c, XMesaBuffer b ) +{ + return XMesaMakeCurrent2( c, b, b ); +} + + +/* + * Bind buffer b to context c and make c the current rendering context. + */ +PUBLIC +GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, + XMesaBuffer readBuffer ) +{ + if (c) { + if (!drawBuffer || !readBuffer) + return GL_FALSE; /* must specify buffers! */ + + if (&(c->mesa) == _mesa_get_current_context() + && c->mesa.DrawBuffer == &drawBuffer->mesa_buffer + && c->mesa.ReadBuffer == &readBuffer->mesa_buffer + && XMESA_BUFFER(c->mesa.DrawBuffer)->wasCurrent) { + /* same context and buffer, do nothing */ + return GL_TRUE; + } + + c->xm_buffer = drawBuffer; + +#ifdef FX + if (FXmakeCurrent( drawBuffer )) + return GL_TRUE; +#endif + + /* Call this periodically to detect when the user has begun using + * GL rendering from multiple threads. + */ + _glapi_check_multithread(); + + xmesa_check_and_update_buffer_size(c, drawBuffer); + if (readBuffer != drawBuffer) + xmesa_check_and_update_buffer_size(c, readBuffer); + + _mesa_make_current(&(c->mesa), + &drawBuffer->mesa_buffer, + &readBuffer->mesa_buffer); + + if (c->xm_visual->mesa_visual.rgbMode) { + /* + * Must recompute and set these pixel values because colormap + * can be different for different windows. + */ + c->clearpixel = xmesa_color_to_pixel( &c->mesa, + c->clearcolor[0], + c->clearcolor[1], + c->clearcolor[2], + c->clearcolor[3], + c->xm_visual->undithered_pf); + XMesaSetForeground(c->display, drawBuffer->cleargc, c->clearpixel); + } + + /* Solution to Stephane Rehel's problem with glXReleaseBuffersMESA(): */ + drawBuffer->wasCurrent = GL_TRUE; + } + else { + /* Detach */ + _mesa_make_current( NULL, NULL, NULL ); + } + return GL_TRUE; +} + + +/* + * Unbind the context c from its buffer. + */ +GLboolean XMesaUnbindContext( XMesaContext c ) +{ + /* A no-op for XFree86 integration purposes */ + return GL_TRUE; +} + + +XMesaContext XMesaGetCurrentContext( void ) +{ + GET_CURRENT_CONTEXT(ctx); + if (ctx) { + XMesaContext xmesa = XMESA_CONTEXT(ctx); + return xmesa; + } + else { + return 0; + } +} + + +XMesaBuffer XMesaGetCurrentBuffer( void ) +{ + GET_CURRENT_CONTEXT(ctx); + if (ctx) { + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + return xmbuf; + } + else { + return 0; + } +} + + +/* New in Mesa 3.1 */ +XMesaBuffer XMesaGetCurrentReadBuffer( void ) +{ + GET_CURRENT_CONTEXT(ctx); + if (ctx) { + return XMESA_BUFFER(ctx->ReadBuffer); + } + else { + return 0; + } +} + + +#ifdef XFree86Server +PUBLIC +GLboolean XMesaForceCurrent(XMesaContext c) +{ + if (c) { + _glapi_set_dispatch(c->mesa.CurrentDispatch); + + if (&(c->mesa) != _mesa_get_current_context()) { + _mesa_make_current(&c->mesa, c->mesa.DrawBuffer, c->mesa.ReadBuffer); + } + } + else { + _mesa_make_current(NULL, NULL, NULL); + } + return GL_TRUE; +} + + +PUBLIC +GLboolean XMesaLoseCurrent(XMesaContext c) +{ + (void) c; + _mesa_make_current(NULL, NULL, NULL); + return GL_TRUE; +} + + +PUBLIC +GLboolean XMesaCopyContext( XMesaContext xm_src, XMesaContext xm_dst, GLuint mask ) +{ + _mesa_copy_context(&xm_src->mesa, &xm_dst->mesa, mask); + return GL_TRUE; +} +#endif /* XFree86Server */ + + +#ifndef FX +GLboolean XMesaSetFXmode( GLint mode ) +{ + (void) mode; + return GL_FALSE; +} +#endif + + + +/* + * Copy the back buffer to the front buffer. If there's no back buffer + * this is a no-op. + */ +PUBLIC +void XMesaSwapBuffers( XMesaBuffer b ) +{ + GET_CURRENT_CONTEXT(ctx); + + if (!b->backxrb) { + /* single buffered */ + return; + } + + /* If we're swapping the buffer associated with the current context + * we have to flush any pending rendering commands first. + */ + if (ctx && ctx->DrawBuffer == &(b->mesa_buffer)) + _mesa_notifySwapBuffers(ctx); + + if (b->db_mode) { +#ifdef FX + if (FXswapBuffers(b)) + return; +#endif + if (b->backxrb->ximage) { + /* Copy Ximage (back buf) from client memory to server window */ +#if defined(USE_XSHM) && !defined(XFree86Server) + if (b->shm) { + /*_glthread_LOCK_MUTEX(_xmesa_lock);*/ + XShmPutImage( b->xm_visual->display, b->frontxrb->drawable, + b->swapgc, + b->backxrb->ximage, 0, 0, + 0, 0, b->mesa_buffer.Width, b->mesa_buffer.Height, + False ); + /*_glthread_UNLOCK_MUTEX(_xmesa_lock);*/ + } + else +#endif + { + /*_glthread_LOCK_MUTEX(_xmesa_lock);*/ + XMesaPutImage( b->xm_visual->display, b->frontxrb->drawable, + b->swapgc, + b->backxrb->ximage, 0, 0, + 0, 0, b->mesa_buffer.Width, b->mesa_buffer.Height ); + /*_glthread_UNLOCK_MUTEX(_xmesa_lock);*/ + } + } + else if (b->backxrb->pixmap) { + /* Copy pixmap (back buf) to window (front buf) on server */ + /*_glthread_LOCK_MUTEX(_xmesa_lock);*/ + XMesaCopyArea( b->xm_visual->display, + b->backxrb->pixmap, /* source drawable */ + b->frontxrb->drawable, /* dest. drawable */ + b->swapgc, + 0, 0, b->mesa_buffer.Width, b->mesa_buffer.Height, + 0, 0 /* dest region */ + ); + /*_glthread_UNLOCK_MUTEX(_xmesa_lock);*/ + } + + if (b->swAlpha) + _mesa_copy_soft_alpha_renderbuffers(ctx, &b->mesa_buffer); + } +#if !defined(XFree86Server) + XSync( b->xm_visual->display, False ); +#endif +} + + + +/* + * Copy sub-region of back buffer to front buffer + */ +void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height ) +{ + GET_CURRENT_CONTEXT(ctx); + + /* If we're swapping the buffer associated with the current context + * we have to flush any pending rendering commands first. + */ + if (ctx && ctx->DrawBuffer == &(b->mesa_buffer)) + _mesa_notifySwapBuffers(ctx); + + if (!b->backxrb) { + /* single buffered */ + return; + } + + if (b->db_mode) { + int yTop = b->mesa_buffer.Height - y - height; +#ifdef FX + if (FXswapBuffers(b)) + return; +#endif + if (b->backxrb->ximage) { + /* Copy Ximage from host's memory to server's window */ +#if defined(USE_XSHM) && !defined(XFree86Server) + if (b->shm) { + /* XXX assuming width and height aren't too large! */ + XShmPutImage( b->xm_visual->display, b->frontxrb->drawable, + b->swapgc, + b->backxrb->ximage, x, yTop, + x, yTop, width, height, False ); + /* wait for finished event??? */ + } + else +#endif + { + /* XXX assuming width and height aren't too large! */ + XMesaPutImage( b->xm_visual->display, b->frontxrb->drawable, + b->swapgc, + b->backxrb->ximage, x, yTop, + x, yTop, width, height ); + } + } + else { + /* Copy pixmap to window on server */ + XMesaCopyArea( b->xm_visual->display, + b->backxrb->pixmap, /* source drawable */ + b->frontxrb->drawable, /* dest. drawable */ + b->swapgc, + x, yTop, width, height, /* source region */ + x, yTop /* dest region */ + ); + } + } +} + + +/* + * Return a pointer to the XMesa backbuffer Pixmap or XImage. This function + * is a way to get "under the hood" of X/Mesa so one can manipulate the + * back buffer directly. + * Output: pixmap - pointer to back buffer's Pixmap, or 0 + * ximage - pointer to back buffer's XImage, or NULL + * Return: GL_TRUE = context is double buffered + * GL_FALSE = context is single buffered + */ +#ifndef XFree86Server +GLboolean XMesaGetBackBuffer( XMesaBuffer b, + XMesaPixmap *pixmap, + XMesaImage **ximage ) +{ + if (b->db_mode) { + if (pixmap) + *pixmap = b->backxrb->pixmap; + if (ximage) + *ximage = b->backxrb->ximage; + return GL_TRUE; + } + else { + *pixmap = 0; + *ximage = NULL; + return GL_FALSE; + } +} +#endif /* XFree86Server */ + + +/* + * Return the depth buffer associated with an XMesaBuffer. + * Input: b - the XMesa buffer handle + * Output: width, height - size of buffer in pixels + * bytesPerValue - bytes per depth value (2 or 4) + * buffer - pointer to depth buffer values + * Return: GL_TRUE or GL_FALSE to indicate success or failure. + */ +GLboolean XMesaGetDepthBuffer( XMesaBuffer b, GLint *width, GLint *height, + GLint *bytesPerValue, void **buffer ) +{ + struct gl_renderbuffer *rb + = b->mesa_buffer.Attachment[BUFFER_DEPTH].Renderbuffer; + if (!rb || !rb->Data) { + *width = 0; + *height = 0; + *bytesPerValue = 0; + *buffer = 0; + return GL_FALSE; + } + else { + *width = b->mesa_buffer.Width; + *height = b->mesa_buffer.Height; + *bytesPerValue = b->mesa_buffer.Visual.depthBits <= 16 + ? sizeof(GLushort) : sizeof(GLuint); + *buffer = rb->Data; + return GL_TRUE; + } +} + + +void XMesaFlush( XMesaContext c ) +{ + if (c && c->xm_visual) { +#ifdef XFree86Server + /* NOT_NEEDED */ +#else + XSync( c->xm_visual->display, False ); +#endif + } +} + + + +const char *XMesaGetString( XMesaContext c, int name ) +{ + (void) c; + if (name==XMESA_VERSION) { + return "5.0"; + } + else if (name==XMESA_EXTENSIONS) { + return ""; + } + else { + return NULL; + } +} + + + +XMesaBuffer XMesaFindBuffer( XMesaDisplay *dpy, XMesaDrawable d ) +{ + XMesaBuffer b; + for (b=XMesaBufferList; b; b=b->Next) { + if (b->frontxrb->drawable == d && b->display == dpy) { + return b; + } + } + return NULL; +} + + +/** + * Free/destroy all XMesaBuffers associated with given display. + */ +void xmesa_destroy_buffers_on_display(XMesaDisplay *dpy) +{ + XMesaBuffer b, next; + for (b = XMesaBufferList; b; b = next) { + next = b->Next; + if (b->display == dpy) { + xmesa_free_buffer(b); + } + } +} + + +/* + * Look for XMesaBuffers whose X window has been destroyed. + * Deallocate any such XMesaBuffers. + */ +void XMesaGarbageCollect( void ) +{ + XMesaBuffer b, next; + for (b=XMesaBufferList; b; b=next) { + next = b->Next; + if (b->display && b->frontxrb->drawable && b->type == WINDOW) { +#ifdef XFree86Server + /* NOT_NEEDED */ +#else + XSync(b->display, False); + if (!window_exists( b->display, b->frontxrb->drawable )) { + /* found a dead window, free the ancillary info */ + XMesaDestroyBuffer( b ); + } +#endif + } + } +} + + +unsigned long XMesaDitherColor( XMesaContext xmesa, GLint x, GLint y, + GLfloat red, GLfloat green, + GLfloat blue, GLfloat alpha ) +{ + GLcontext *ctx = &xmesa->mesa; + GLint r = (GLint) (red * 255.0F); + GLint g = (GLint) (green * 255.0F); + GLint b = (GLint) (blue * 255.0F); + GLint a = (GLint) (alpha * 255.0F); + + switch (xmesa->pixelformat) { + case PF_Index: + return 0; + case PF_Truecolor: + { + unsigned long p; + PACK_TRUECOLOR( p, r, g, b ); + return p; + } + case PF_8A8B8G8R: + return PACK_8A8B8G8R( r, g, b, a ); + case PF_8A8R8G8B: + return PACK_8A8R8G8B( r, g, b, a ); + case PF_8R8G8B: + return PACK_8R8G8B( r, g, b ); + case PF_5R6G5B: + return PACK_5R6G5B( r, g, b ); + case PF_Dither: + { + DITHER_SETUP; + return DITHER( x, y, r, g, b ); + } + case PF_1Bit: + /* 382 = (3*255)/2 */ + return ((r+g+b) > 382) ^ xmesa->xm_visual->bitFlip; + case PF_HPCR: + return DITHER_HPCR(x, y, r, g, b); + case PF_Lookup: + { + LOOKUP_SETUP; + return LOOKUP( r, g, b ); + } + case PF_Grayscale: + return GRAY_RGB( r, g, b ); + case PF_Dither_5R6G5B: + /* fall through */ + case PF_Dither_True: + { + unsigned long p; + PACK_TRUEDITHER(p, x, y, r, g, b); + return p; + } + default: + _mesa_problem(NULL, "Bad pixel format in XMesaDitherColor"); + } + return 0; +} + + +/* + * This is typically called when the window size changes and we need + * to reallocate the buffer's back/depth/stencil/accum buffers. + */ +PUBLIC void +XMesaResizeBuffers( XMesaBuffer b ) +{ + GET_CURRENT_CONTEXT(ctx); + XMesaContext xmctx = XMESA_CONTEXT(ctx); + if (!xmctx) + return; + xmesa_check_and_update_buffer_size(xmctx, b); +} + + +static GLint +xbuffer_to_renderbuffer(int buffer) +{ + assert(MAX_AUX_BUFFERS <= 4); + + switch (buffer) { + case GLX_FRONT_LEFT_EXT: + return BUFFER_FRONT_LEFT; + case GLX_FRONT_RIGHT_EXT: + return BUFFER_FRONT_RIGHT; + case GLX_BACK_LEFT_EXT: + return BUFFER_BACK_LEFT; + case GLX_BACK_RIGHT_EXT: + return BUFFER_BACK_RIGHT; + case GLX_AUX0_EXT: + return BUFFER_AUX0; + case GLX_AUX1_EXT: + return BUFFER_AUX1; + case GLX_AUX2_EXT: + return BUFFER_AUX2; + case GLX_AUX3_EXT: + return BUFFER_AUX3; + case GLX_AUX4_EXT: + case GLX_AUX5_EXT: + case GLX_AUX6_EXT: + case GLX_AUX7_EXT: + case GLX_AUX8_EXT: + case GLX_AUX9_EXT: + default: + /* BadValue error */ + return -1; + } +} + + +PUBLIC void +XMesaBindTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer, + const int *attrib_list) +{ +#if 0 + GET_CURRENT_CONTEXT(ctx); + const GLuint unit = ctx->Texture.CurrentUnit; + struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; + struct gl_texture_object *texObj; +#endif + struct gl_renderbuffer *rb; + struct xmesa_renderbuffer *xrb; + GLint b; + XMesaImage *img = NULL; + GLboolean freeImg = GL_FALSE; + + b = xbuffer_to_renderbuffer(buffer); + if (b < 0) + return; + + if (drawable->TextureFormat == GLX_TEXTURE_FORMAT_NONE_EXT) + return; /* BadMatch error */ + + rb = drawable->mesa_buffer.Attachment[b].Renderbuffer; + if (!rb) { + /* invalid buffer */ + return; + } + xrb = xmesa_renderbuffer(rb); + +#if 0 + switch (drawable->TextureTarget) { + case GLX_TEXTURE_1D_EXT: + texObj = texUnit->Current1D; + break; + case GLX_TEXTURE_2D_EXT: + texObj = texUnit->Current2D; + break; + case GLX_TEXTURE_RECTANGLE_EXT: + texObj = texUnit->CurrentRect; + break; + default: + return; /* BadMatch error */ + } +#endif + + /* + * The following is a quick and simple way to implement + * BindTexImage. The better way is to write some new FetchTexel() + * functions which would extract texels from XImages. We'd still + * need to use GetImage when texturing from a Pixmap (front buffer) + * but texturing from a back buffer (XImage) would avoid an image + * copy. + */ + + /* get XImage */ + if (xrb->pixmap) { + img = XMesaGetImage(dpy, xrb->pixmap, 0, 0, rb->Width, rb->Height, ~0L, + ZPixmap); + freeImg = GL_TRUE; + } + else if (xrb->ximage) { + img = xrb->ximage; + } + + /* store the XImage as a new texture image */ + if (img) { + GLenum format, type, intFormat; + if (img->bits_per_pixel == 32) { + format = GL_BGRA; + type = GL_UNSIGNED_BYTE; + intFormat = GL_RGBA; + } + else if (img->bits_per_pixel == 24) { + format = GL_BGR; + type = GL_UNSIGNED_BYTE; + intFormat = GL_RGB; + } + else if (img->bits_per_pixel == 16) { + format = GL_BGR; + type = GL_UNSIGNED_SHORT_5_6_5; + intFormat = GL_RGB; + } + else { + _mesa_problem(NULL, "Unexpected XImage format in XMesaBindTexImage"); + return; + } + if (drawable->TextureFormat == GLX_TEXTURE_FORMAT_RGBA_EXT) { + intFormat = GL_RGBA; + } + else if (drawable->TextureFormat == GLX_TEXTURE_FORMAT_RGB_EXT) { + intFormat = GL_RGB; + } + + _mesa_TexImage2D(GL_TEXTURE_2D, 0, intFormat, rb->Width, rb->Height, 0, + format, type, img->data); + + if (freeImg) { + XMesaDestroyImage(img); + } + } +} + + + +PUBLIC void +XMesaReleaseTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer) +{ + const GLint b = xbuffer_to_renderbuffer(buffer); + if (b < 0) + return; + + /* no-op for now */ +} + diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c new file mode 100644 index 0000000000..09356c7329 --- /dev/null +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -0,0 +1,503 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5.2 + * + * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * \file xm_buffer.h + * Framebuffer and renderbuffer-related functions. + */ + + +#include "glxheader.h" +#include "GL/xmesa.h" +#include "xmesaP.h" +#include "imports.h" +#include "framebuffer.h" +#include "renderbuffer.h" +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "state_tracker/st_context.h" + + +#if defined(USE_XSHM) && !defined(XFree86Server) +static volatile int mesaXErrorFlag = 0; + +/** + * Catches potential Xlib errors. + */ +static int +mesaHandleXError(XMesaDisplay *dpy, XErrorEvent *event) +{ + (void) dpy; + (void) event; + mesaXErrorFlag = 1; + return 0; +} + +/** + * Allocate a shared memory XImage back buffer for the given XMesaBuffer. + * Return: GL_TRUE if success, GL_FALSE if error + */ +static GLboolean +alloc_back_shm_ximage(XMesaBuffer b, GLuint width, GLuint height) +{ + /* + * We have to do a _lot_ of error checking here to be sure we can + * really use the XSHM extension. It seems different servers trigger + * errors at different points if the extension won't work. Therefore + * we have to be very careful... + */ + GC gc; + int (*old_handler)(XMesaDisplay *, XErrorEvent *); + + if (width == 0 || height == 0) { + /* this will be true the first time we're called on 'b' */ + return GL_FALSE; + } + + b->backxrb->ximage = XShmCreateImage(b->xm_visual->display, + b->xm_visual->visinfo->visual, + b->xm_visual->visinfo->depth, + ZPixmap, NULL, &b->shminfo, + width, height); + if (b->backxrb->ximage == NULL) { + _mesa_warning(NULL, "alloc_back_buffer: Shared memory error (XShmCreateImage), disabling.\n"); + b->shm = 0; + return GL_FALSE; + } + + b->shminfo.shmid = shmget(IPC_PRIVATE, b->backxrb->ximage->bytes_per_line + * b->backxrb->ximage->height, IPC_CREAT|0777); + if (b->shminfo.shmid < 0) { + _mesa_warning(NULL, "shmget failed while allocating back buffer.\n"); + XDestroyImage(b->backxrb->ximage); + b->backxrb->ximage = NULL; + _mesa_warning(NULL, "alloc_back_buffer: Shared memory error (shmget), disabling.\n"); + b->shm = 0; + return GL_FALSE; + } + + b->shminfo.shmaddr = b->backxrb->ximage->data + = (char*)shmat(b->shminfo.shmid, 0, 0); + if (b->shminfo.shmaddr == (char *) -1) { + _mesa_warning(NULL, "shmat() failed while allocating back buffer.\n"); + XDestroyImage(b->backxrb->ximage); + shmctl(b->shminfo.shmid, IPC_RMID, 0); + b->backxrb->ximage = NULL; + _mesa_warning(NULL, "alloc_back_buffer: Shared memory error (shmat), disabling.\n"); + b->shm = 0; + return GL_FALSE; + } + + b->shminfo.readOnly = False; + mesaXErrorFlag = 0; + old_handler = XSetErrorHandler(mesaHandleXError); + /* This may trigger the X protocol error we're ready to catch: */ + XShmAttach(b->xm_visual->display, &b->shminfo); + XSync(b->xm_visual->display, False); + + if (mesaXErrorFlag) { + /* we are on a remote display, this error is normal, don't print it */ + XFlush(b->xm_visual->display); + mesaXErrorFlag = 0; + XDestroyImage(b->backxrb->ximage); + shmdt(b->shminfo.shmaddr); + shmctl(b->shminfo.shmid, IPC_RMID, 0); + b->backxrb->ximage = NULL; + b->shm = 0; + (void) XSetErrorHandler(old_handler); + return GL_FALSE; + } + + shmctl(b->shminfo.shmid, IPC_RMID, 0); /* nobody else needs it */ + + /* Finally, try an XShmPutImage to be really sure the extension works */ + gc = XCreateGC(b->xm_visual->display, b->frontxrb->drawable, 0, NULL); + XShmPutImage(b->xm_visual->display, b->frontxrb->drawable, gc, + b->backxrb->ximage, 0, 0, 0, 0, 1, 1 /*one pixel*/, False); + XSync(b->xm_visual->display, False); + XFreeGC(b->xm_visual->display, gc); + (void) XSetErrorHandler(old_handler); + if (mesaXErrorFlag) { + XFlush(b->xm_visual->display); + mesaXErrorFlag = 0; + XDestroyImage(b->backxrb->ximage); + shmdt(b->shminfo.shmaddr); + shmctl(b->shminfo.shmid, IPC_RMID, 0); + b->backxrb->ximage = NULL; + b->shm = 0; + return GL_FALSE; + } + + return GL_TRUE; +} +#else +static GLboolean +alloc_back_shm_ximage(XMesaBuffer b, GLuint width, GLuint height) +{ + /* Can't compile XSHM support */ + return GL_FALSE; +} +#endif + + + +/** + * Setup an off-screen pixmap or Ximage to use as the back buffer. + * Input: b - the X/Mesa buffer + */ +static void +alloc_back_buffer(XMesaBuffer b, GLuint width, GLuint height) +{ + if (b->db_mode == BACK_XIMAGE) { + /* Deallocate the old backxrb->ximage, if any */ + if (b->backxrb->ximage) { +#if defined(USE_XSHM) && !defined(XFree86Server) + if (b->shm) { + XShmDetach(b->xm_visual->display, &b->shminfo); + XDestroyImage(b->backxrb->ximage); + shmdt(b->shminfo.shmaddr); + } + else +#endif + XMesaDestroyImage(b->backxrb->ximage); + b->backxrb->ximage = NULL; + } + + if (width == 0 || height == 0) + return; + + /* Allocate new back buffer */ + if (b->shm == 0 || !alloc_back_shm_ximage(b, width, height)) { + /* Allocate a regular XImage for the back buffer. */ +#ifdef XFree86Server + b->backxrb->ximage = XMesaCreateImage(b->xm_visual->BitsPerPixel, + width, height, NULL); +#else + b->backxrb->ximage = XCreateImage(b->xm_visual->display, + b->xm_visual->visinfo->visual, + GET_VISUAL_DEPTH(b->xm_visual), + ZPixmap, 0, /* format, offset */ + NULL, + width, height, + 8, 0); /* pad, bytes_per_line */ +#endif + if (!b->backxrb->ximage) { + _mesa_warning(NULL, "alloc_back_buffer: XCreateImage failed.\n"); + return; + } + b->backxrb->ximage->data = (char *) MALLOC(b->backxrb->ximage->height + * b->backxrb->ximage->bytes_per_line); + if (!b->backxrb->ximage->data) { + _mesa_warning(NULL, "alloc_back_buffer: MALLOC failed.\n"); + XMesaDestroyImage(b->backxrb->ximage); + b->backxrb->ximage = NULL; + } + } + b->backxrb->pixmap = None; + } + else if (b->db_mode == BACK_PIXMAP) { + /* Free the old back pixmap */ + if (b->backxrb->pixmap) { + XMesaFreePixmap(b->xm_visual->display, b->backxrb->pixmap); + b->backxrb->pixmap = 0; + } + + if (width > 0 && height > 0) { + /* Allocate new back pixmap */ + b->backxrb->pixmap = XMesaCreatePixmap(b->xm_visual->display, + b->frontxrb->drawable, + width, height, + GET_VISUAL_DEPTH(b->xm_visual)); + } + + b->backxrb->ximage = NULL; + } +} + + +static void +xmesa_delete_renderbuffer(struct gl_renderbuffer *rb) +{ + /* XXX Note: the ximage or Pixmap attached to this renderbuffer + * should probably get freed here, but that's currently done in + * XMesaDestroyBuffer(). + */ + _mesa_free(rb); +} + + +static void +finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) +{ + struct pipe_context *pipe = ctx->st->pipe; + if (!xrb->St.surface->region) { + int w = 1, h = 1; + xrb->St.surface->region = pipe->winsys->region_alloc(pipe->winsys, + 1, w, h, 0x0); + } +} + + +/** + * Reallocate renderbuffer storage for front color buffer. + * Called via gl_renderbuffer::AllocStorage() + */ +static GLboolean +xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, GLuint width, GLuint height) +{ + struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); + + /* just clear these to be sure we don't accidentally use them */ + xrb->origin1 = NULL; + xrb->origin2 = NULL; + xrb->origin3 = NULL; + xrb->origin4 = NULL; + + /* for the FLIP macro: */ + xrb->bottom = height - 1; + + rb->Width = width; + rb->Height = height; + rb->InternalFormat = internalFormat; + + if (!xrb->St.surface || !xrb->St.surface->region) + finish_surface_init(ctx, xrb); + + xrb->St.surface->width = width; + xrb->St.surface->height = height; + + return GL_TRUE; +} + + +/** + * Reallocate renderbuffer storage for back color buffer. + * Called via gl_renderbuffer::AllocStorage() + */ +static GLboolean +xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, GLuint width, GLuint height) +{ + struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); + + /* reallocate the back buffer XImage or Pixmap */ + assert(xrb->Parent); + alloc_back_buffer(xrb->Parent, width, height); + + /* same as front buffer */ + /* XXX why is this here? */ + (void) xmesa_alloc_front_storage(ctx, rb, internalFormat, width, height); + + /* plus... */ + if (xrb->ximage) { + /* Needed by PIXELADDR1 macro */ + xrb->width1 = xrb->ximage->bytes_per_line; + xrb->origin1 = (GLubyte *) xrb->ximage->data + xrb->width1 * (height - 1); + + /* Needed by PIXELADDR2 macro */ + xrb->width2 = xrb->ximage->bytes_per_line / 2; + xrb->origin2 = (GLushort *) xrb->ximage->data + xrb->width2 * (height - 1); + + /* Needed by PIXELADDR3 macro */ + xrb->width3 = xrb->ximage->bytes_per_line; + xrb->origin3 = (GLubyte *) xrb->ximage->data + xrb->width3 * (height - 1); + + /* Needed by PIXELADDR4 macro */ + xrb->width4 = xrb->ximage->width; + xrb->origin4 = (GLuint *) xrb->ximage->data + xrb->width4 * (height - 1); + } + else { + /* out of memory or buffer size is 0 x 0 */ + xrb->width1 = xrb->width2 = xrb->width3 = xrb->width4 = 0; + xrb->origin1 = NULL; + xrb->origin2 = NULL; + xrb->origin3 = NULL; + xrb->origin4 = NULL; + } + + if (!xrb->St.surface || !xrb->St.surface->region) + finish_surface_init(ctx, xrb); + + xrb->St.surface->width = width; + xrb->St.surface->height = height; + + return GL_TRUE; +} + + +/** + * Called to create the front/back color renderbuffers, not user-created + * renderbuffers. + */ +struct xmesa_renderbuffer * +xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, + GLboolean backBuffer) +{ + struct xmesa_renderbuffer *xrb = CALLOC_STRUCT(xmesa_renderbuffer); + struct pipe_context *pipe = NULL;/*ctx->st->pipe;*/ + if (xrb) { + GLuint name = 0; + GLuint pipeFormat = 0; + struct xmesa_surface *xms; + + _mesa_init_renderbuffer(&xrb->St.Base, name); + + xrb->St.Base.Delete = xmesa_delete_renderbuffer; + if (backBuffer) + xrb->St.Base.AllocStorage = xmesa_alloc_back_storage; + else + xrb->St.Base.AllocStorage = xmesa_alloc_front_storage; + + if (visual->rgbMode) { + xrb->St.Base.InternalFormat = GL_RGBA; + xrb->St.Base._BaseFormat = GL_RGBA; + xrb->St.Base.DataType = GL_UNSIGNED_BYTE; + xrb->St.Base.RedBits = visual->redBits; + xrb->St.Base.GreenBits = visual->greenBits; + xrb->St.Base.BlueBits = visual->blueBits; + xrb->St.Base.AlphaBits = visual->alphaBits; + pipeFormat = PIPE_FORMAT_U_A8_R8_G8_B8; + } + else { + xrb->St.Base.InternalFormat = GL_COLOR_INDEX; + xrb->St.Base._BaseFormat = GL_COLOR_INDEX; + xrb->St.Base.DataType = GL_UNSIGNED_INT; + xrb->St.Base.IndexBits = visual->indexBits; + } + /* only need to set Red/Green/EtcBits fields for user-created RBs */ + + xrb->St.surface = xmesa_new_color_surface(pipe, pipeFormat); + xms = (struct xmesa_surface *) xrb->St.surface; + xms->xrb = xrb; + } + return xrb; +} + + +#if 0 +struct gl_renderbuffer * +xmesa_new_renderbuffer(GLcontext *ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, GLuint width, GLuint height) +{ + struct xmesa_renderbuffer *xrb = CALLOC_STRUCT(xmesa_renderbuffer); + if (xrb) { + GLuint name = 0; + _mesa_init_renderbuffer(&xrb->St.Base, name); + + xrb->St.Base.Delete = xmesa_delete_renderbuffer; + if (backBuffer) + xrb->St.Base.AllocStorage = xmesa_alloc_back_storage; + else + xrb->St.Base.AllocStorage = xmesa_alloc_front_storage; + + if (visual->rgbMode) { + xrb->St.Base.InternalFormat = GL_RGBA; + xrb->St.Base._BaseFormat = GL_RGBA; + xrb->St.Base.DataType = GL_UNSIGNED_BYTE; + xrb->St.Base.RedBits = visual->redBits; + xrb->St.Base.GreenBits = visual->greenBits; + xrb->St.Base.BlueBits = visual->blueBits; + xrb->St.Base.AlphaBits = visual->alphaBits; + } + else { + xrb->St.Base.InternalFormat = GL_COLOR_INDEX; + xrb->St.Base._BaseFormat = GL_COLOR_INDEX; + xrb->St.Base.DataType = GL_UNSIGNED_INT; + xrb->St.Base.IndexBits = visual->indexBits; + } + /* only need to set Red/Green/EtcBits fields for user-created RBs */ + } + return xrb; +} +#endif + + +/** + * Called via gl_framebuffer::Delete() method when this buffer + * is _really_ being deleted. + */ +void +xmesa_delete_framebuffer(struct gl_framebuffer *fb) +{ + XMesaBuffer b = XMESA_BUFFER(fb); + + if (b->num_alloced > 0) { + /* If no other buffer uses this X colormap then free the colors. */ + if (!xmesa_find_buffer(b->display, b->cmap, b)) { +#ifdef XFree86Server + int client = 0; + if (b->frontxrb->drawable) + client = CLIENT_ID(b->frontxrb->drawable->id); + (void)FreeColors(b->cmap, client, + b->num_alloced, b->alloced_colors, 0); +#else + XFreeColors(b->display, b->cmap, + b->alloced_colors, b->num_alloced, 0); +#endif + } + } + + if (b->gc) + XMesaFreeGC(b->display, b->gc); + if (b->cleargc) + XMesaFreeGC(b->display, b->cleargc); + if (b->swapgc) + XMesaFreeGC(b->display, b->swapgc); + + if (fb->Visual.doubleBufferMode) { + /* free back ximage/pixmap/shmregion */ + if (b->backxrb->ximage) { +#if defined(USE_XSHM) && !defined(XFree86Server) + if (b->shm) { + XShmDetach( b->display, &b->shminfo ); + XDestroyImage( b->backxrb->ximage ); + shmdt( b->shminfo.shmaddr ); + } + else +#endif + XMesaDestroyImage( b->backxrb->ximage ); + b->backxrb->ximage = NULL; + } + if (b->backxrb->pixmap) { + XMesaFreePixmap( b->display, b->backxrb->pixmap ); + if (b->xm_visual->hpcr_clear_flag) { + XMesaFreePixmap( b->display, + b->xm_visual->hpcr_clear_pixmap ); + XMesaDestroyImage( b->xm_visual->hpcr_clear_ximage ); + } + } + } + + if (b->rowimage) { + _mesa_free( b->rowimage->data ); + b->rowimage->data = NULL; + XMesaDestroyImage( b->rowimage ); + } + + _mesa_free_framebuffer_data(fb); + _mesa_free(fb); +} diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c new file mode 100644 index 0000000000..8ae243ae66 --- /dev/null +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -0,0 +1,1141 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5.2 + * + * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * \file xm_dd.h + * General device driver functions for Xlib driver. + */ + +#include "glxheader.h" +#include "bufferobj.h" +#include "buffers.h" +#include "context.h" +#include "colormac.h" +#include "depth.h" +#include "drawpix.h" +#include "extensions.h" +#include "framebuffer.h" +#include "macros.h" +#include "image.h" +#include "imports.h" +#include "mtypes.h" +#include "state.h" +#include "texobj.h" +#include "teximage.h" +#include "texstore.h" +#include "texformat.h" +#include "xmesaP.h" +#include "swrast/swrast.h" +#include "swrast/s_context.h" +#include "swrast_setup/swrast_setup.h" +#include "tnl/tnl.h" +#include "tnl/t_context.h" + +#include "pipe/softpipe/sp_context.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_draw.h" + + +/* + * Dithering kernels and lookup tables. + */ + +const int xmesa_kernel8[DITH_DY * DITH_DX] = { + 0 * MAXC, 8 * MAXC, 2 * MAXC, 10 * MAXC, + 12 * MAXC, 4 * MAXC, 14 * MAXC, 6 * MAXC, + 3 * MAXC, 11 * MAXC, 1 * MAXC, 9 * MAXC, + 15 * MAXC, 7 * MAXC, 13 * MAXC, 5 * MAXC, +}; + +const short xmesa_HPCR_DRGB[3][2][16] = { + { + { 16, -4, 1,-11, 14, -6, 3, -9, 15, -5, 2,-10, 13, -7, 4, -8}, + {-15, 5, 0, 12,-13, 7, -2, 10,-14, 6, -1, 11,-12, 8, -3, 9} + }, + { + {-11, 15, -7, 3, -8, 14, -4, 2,-10, 16, -6, 4, -9, 13, -5, 1}, + { 12,-14, 8, -2, 9,-13, 5, -1, 11,-15, 7, -3, 10,-12, 6, 0} + }, + { + { 6,-18, 26,-14, 2,-22, 30,-10, 8,-16, 28,-12, 4,-20, 32, -8}, + { -4, 20,-24, 16, 0, 24,-28, 12, -6, 18,-26, 14, -2, 22,-30, 10} + } +}; + +const int xmesa_kernel1[16] = { + 0*47, 9*47, 4*47, 12*47, /* 47 = (255*3)/16 */ + 6*47, 2*47, 14*47, 8*47, + 10*47, 1*47, 5*47, 11*47, + 7*47, 13*47, 3*47, 15*47 +}; + + +static void +finish_or_flush( GLcontext *ctx ) +{ +#ifdef XFree86Server + /* NOT_NEEDED */ +#else + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + if (xmesa) { + _glthread_LOCK_MUTEX(_xmesa_lock); + XSync( xmesa->display, False ); + _glthread_UNLOCK_MUTEX(_xmesa_lock); + } +#endif +} + + +static void +clear_index( GLcontext *ctx, GLuint index ) +{ + if (ctx->DrawBuffer->Name == 0) { + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + xmesa->clearpixel = (unsigned long) index; + XMesaSetForeground( xmesa->display, xmbuf->cleargc, (unsigned long) index ); + } +} + + +static void +clear_color( GLcontext *ctx, const GLfloat color[4] ) +{ + if (ctx->DrawBuffer->Name == 0) { + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[0], color[0]); + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[1], color[1]); + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[2], color[2]); + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[3], color[3]); + xmesa->clearpixel = xmesa_color_to_pixel( ctx, + xmesa->clearcolor[0], + xmesa->clearcolor[1], + xmesa->clearcolor[2], + xmesa->clearcolor[3], + xmesa->xm_visual->undithered_pf ); + _glthread_LOCK_MUTEX(_xmesa_lock); + XMesaSetForeground( xmesa->display, xmbuf->cleargc, + xmesa->clearpixel ); + _glthread_UNLOCK_MUTEX(_xmesa_lock); + } +} + + + +/* Set index mask ala glIndexMask */ +static void +index_mask( GLcontext *ctx, GLuint mask ) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + /* not sure this conditional is really needed */ + if (xmbuf->backxrb && xmbuf->backxrb->pixmap) { + unsigned long m; + if (mask==0xffffffff) { + m = ((unsigned long)~0L); + } + else { + m = (unsigned long) mask; + } + XMesaSetPlaneMask( xmesa->display, xmbuf->cleargc, m ); + } +} + + +/* Implements glColorMask() */ +static void +color_mask(GLcontext *ctx, + GLboolean rmask, GLboolean gmask, GLboolean bmask, GLboolean amask) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaBuffer xmbuf; + const int xclass = xmesa->xm_visual->mesa_visual.visualType; + (void) amask; + + if (ctx->DrawBuffer->Name != 0) + return; + + xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + + if (xclass == GLX_TRUE_COLOR || xclass == GLX_DIRECT_COLOR) { + unsigned long m; + if (rmask && gmask && bmask) { + m = ((unsigned long)~0L); + } + else { + m = 0; + if (rmask) m |= GET_REDMASK(xmesa->xm_visual); + if (gmask) m |= GET_GREENMASK(xmesa->xm_visual); + if (bmask) m |= GET_BLUEMASK(xmesa->xm_visual); + } + XMesaSetPlaneMask( xmesa->display, xmbuf->cleargc, m ); + } +} + + + +/**********************************************************************/ +/*** glClear implementations ***/ +/**********************************************************************/ + + +/** + * Clear the front or back color buffer, if it's implemented with a pixmap. + */ +static void +clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLint x, GLint y, GLint width, GLint height) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + + assert(xmbuf); + assert(xrb->pixmap); + assert(xmesa); + assert(xmesa->display); + assert(xrb->pixmap); + assert(xmbuf->cleargc); + + XMesaFillRectangle( xmesa->display, xrb->pixmap, xmbuf->cleargc, + x, xrb->St.Base.Height - y - height, + width, height ); +} + + +static void +clear_8bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLint x, GLint y, GLint width, GLint height ) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLint i; + for (i = 0; i < height; i++) { + GLubyte *ptr = PIXEL_ADDR1(xrb, x, y + i); + MEMSET( ptr, xmesa->clearpixel, width ); + } +} + + +static void +clear_HPCR_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLint x, GLint y, GLint width, GLint height ) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLint i; + for (i = y; i < y + height; i++) { + GLubyte *ptr = PIXEL_ADDR1( xrb, x, i ); + int j; + const GLubyte *sptr = xmesa->xm_visual->hpcr_clear_ximage_pattern[0]; + if (i & 1) { + sptr += 16; + } + for (j = x; j < x + width; j++) { + *ptr = sptr[j&15]; + ptr++; + } + } +} + + +static void +clear_16bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLint x, GLint y, GLint width, GLint height) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLuint pixel = (GLuint) xmesa->clearpixel; + GLint i, j; + + if (xmesa->swapbytes) { + pixel = ((pixel >> 8) & 0x00ff) | ((pixel << 8) & 0xff00); + } + + for (j = 0; j < height; j++) { + GLushort *ptr2 = PIXEL_ADDR2(xrb, x, y + j); + for (i = 0; i < width; i++) { + ptr2[i] = pixel; + } + } +} + + +/* Optimized code provided by Nozomi Ytow */ +static void +clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLint x, GLint y, GLint width, GLint height) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + const GLubyte r = xmesa->clearcolor[0]; + const GLubyte g = xmesa->clearcolor[1]; + const GLubyte b = xmesa->clearcolor[2]; + + if (r == g && g == b) { + /* same value for all three components (gray) */ + GLint j; + for (j = 0; j < height; j++) { + bgr_t *ptr3 = PIXEL_ADDR3(xrb, x, y + j); + MEMSET(ptr3, r, 3 * width); + } + } + else { + /* non-gray clear color */ + GLint i, j; + for (j = 0; j < height; j++) { + bgr_t *ptr3 = PIXEL_ADDR3(xrb, x, y + j); + for (i = 0; i < width; i++) { + ptr3->r = r; + ptr3->g = g; + ptr3->b = b; + ptr3++; + } + } + } +} + + +static void +clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLint x, GLint y, GLint width, GLint height) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + register GLuint pixel = (GLuint) xmesa->clearpixel; + + if (!xrb->ximage) + return; + + if (xmesa->swapbytes) { + pixel = ((pixel >> 24) & 0x000000ff) + | ((pixel >> 8) & 0x0000ff00) + | ((pixel << 8) & 0x00ff0000) + | ((pixel << 24) & 0xff000000); + } + + if (width == xrb->St.Base.Width && height == xrb->St.Base.Height) { + /* clearing whole buffer */ + const GLuint n = xrb->St.Base.Width * xrb->St.Base.Height; + GLuint *ptr4 = (GLuint *) xrb->ximage->data; + if (pixel == 0) { + /* common case */ + _mesa_memset(ptr4, pixel, 4 * n); + } + else { + GLuint i; + for (i = 0; i < n; i++) + ptr4[i] = pixel; + } + } + else { + /* clearing scissored region */ + GLint i, j; + for (j = 0; j < height; j++) { + GLuint *ptr4 = PIXEL_ADDR4(xrb, x, y + j); + for (i = 0; i < width; i++) { + ptr4[i] = pixel; + } + } + } +} + + +static void +clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLint x, GLint y, GLint width, GLint height) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaImage *img = xrb->ximage; + GLint i, j; + + /* TODO: optimize this */ + y = YFLIP(xrb, y); + for (j = 0; j < height; j++) { + for (i = 0; i < width; i++) { + XMesaPutPixel(img, x+i, y-j, xmesa->clearpixel); + } + } +} + + + +void +xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers) +{ + if (ctx->DrawBuffer->Name == 0) { + /* this is a window system framebuffer */ + const GLuint *colorMask = (GLuint *) &ctx->Color.ColorMask; + XMesaBuffer b = XMESA_BUFFER(ctx->DrawBuffer); + const GLint x = ctx->DrawBuffer->_Xmin; + const GLint y = ctx->DrawBuffer->_Ymin; + const GLint width = ctx->DrawBuffer->_Xmax - x; + const GLint height = ctx->DrawBuffer->_Ymax - y; + + /* we can't handle color or index masking */ + if (*colorMask == 0xffffffff && ctx->Color.IndexMask == 0xffffffff) { + if (buffers & BUFFER_BIT_FRONT_LEFT) { + /* clear front color buffer */ + struct gl_renderbuffer *frontRb + = ctx->DrawBuffer->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; + if (b->frontxrb == xmesa_renderbuffer(frontRb)) { + /* renderbuffer is not wrapped - great! */ + b->frontxrb->clearFunc(ctx, b->frontxrb, x, y, width, height); + buffers &= ~BUFFER_BIT_FRONT_LEFT; + } + else { + /* we can't directly clear an alpha-wrapped color buffer */ + } + } + if (buffers & BUFFER_BIT_BACK_LEFT) { + /* clear back color buffer */ + struct gl_renderbuffer *backRb + = ctx->DrawBuffer->Attachment[BUFFER_BACK_LEFT].Renderbuffer; + if (b->backxrb == xmesa_renderbuffer(backRb)) { + /* renderbuffer is not wrapped - great! */ + b->backxrb->clearFunc(ctx, b->backxrb, x, y, width, height); + buffers &= ~BUFFER_BIT_BACK_LEFT; + } + } + } + } + if (buffers) + _swrast_Clear(ctx, buffers); +} + + +#ifndef XFree86Server +/* XXX this was never tested in the Xserver environment */ + +/** + * This function implements glDrawPixels() with an XPutImage call when + * drawing to the front buffer (X Window drawable). + * The image format must be GL_BGRA to match the PF_8R8G8B pixel format. + */ +static void +xmesa_DrawPixels_8R8G8B( GLcontext *ctx, + GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels ) +{ + const SWcontext *swrast = SWRAST_CONTEXT( ctx ); + struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[0][0]; + struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb->Wrapped); + + if (swrast->NewState) + _swrast_validate_derived( ctx ); + + if (ctx->DrawBuffer->Name == 0 && + format == GL_BGRA && + type == GL_UNSIGNED_BYTE && + (swrast->_RasterMask & ~CLIP_BIT) == 0 && /* no blend, z-test, etc */ + ctx->_ImageTransferState == 0 && /* no color tables, scale/bias, etc */ + ctx->Pixel.ZoomX == 1.0 && /* no zooming */ + ctx->Pixel.ZoomY == 1.0 && + xrb->pixmap && /* drawing to pixmap or window */ + xrb->St.Base.AlphaBits == 0) + { + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + XMesaDisplay *dpy = xmesa->xm_visual->display; + const XMesaGC gc = xmbuf->cleargc; /* effected by glColorMask */ + int dstX = x; + int dstY = y; + int w = width; + int h = height; + struct gl_pixelstore_attrib clippedUnpack = *unpack; + + ASSERT(xmesa->xm_visual->dithered_pf == PF_8R8G8B); + ASSERT(xmesa->xm_visual->undithered_pf == PF_8R8G8B); + ASSERT(dpy); + ASSERT(gc); + + if (unpack->BufferObj->Name) { + /* unpack from PBO */ + GLubyte *buf; + if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, + format, type, pixels)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glDrawPixels(invalid PBO access)"); + return; + } + buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, + GL_PIXEL_UNPACK_BUFFER_EXT, + GL_READ_ONLY_ARB, + unpack->BufferObj); + if (!buf) { + /* buffer is already mapped - that's an error */ + _mesa_error(ctx, GL_INVALID_OPERATION, + "glDrawPixels(PBO is mapped)"); + return; + } + pixels = ADD_POINTERS(buf, pixels); + } + + if (_mesa_clip_drawpixels(ctx, &dstX, &dstY, &w, &h, &clippedUnpack)) { + /* This is a little tricky since all coordinates up to now have + * been in the OpenGL bottom-to-top orientation. X is top-to-bottom + * so we have to carefully compute the Y coordinates/addresses here. + */ + int srcX = clippedUnpack.SkipPixels; + int srcY = clippedUnpack.SkipRows; + int rowLength = clippedUnpack.RowLength; + XMesaImage ximage; + MEMSET(&ximage, 0, sizeof(XMesaImage)); + ximage.width = width; + ximage.height = height; + ximage.format = ZPixmap; + ximage.data = (char *) pixels + + ((srcY + h - 1) * rowLength + srcX) * 4; + ximage.byte_order = LSBFirst; + ximage.bitmap_unit = 32; + ximage.bitmap_bit_order = LSBFirst; + ximage.bitmap_pad = 32; + ximage.depth = 24; + ximage.bytes_per_line = -rowLength * 4; /* negative to flip image */ + ximage.bits_per_pixel = 32; + /* it seems we don't need to set the ximage.red/green/blue_mask fields */ + /* flip Y axis for dest position */ + dstY = YFLIP(xrb, dstY) - h + 1; + XPutImage(dpy, xrb->pixmap, gc, &ximage, 0, 0, dstX, dstY, w, h); + } + + if (unpack->BufferObj->Name) { + ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, + unpack->BufferObj); + } + } + else { + /* software fallback */ + _swrast_DrawPixels(ctx, x, y, width, height, + format, type, unpack, pixels); + } +} + + + +/** + * This function implements glDrawPixels() with an XPutImage call when + * drawing to the front buffer (X Window drawable). The image format + * must be GL_RGB and image type must be GL_UNSIGNED_SHORT_5_6_5 to + * match the PF_5R6G5B pixel format. + */ +static void +xmesa_DrawPixels_5R6G5B( GLcontext *ctx, + GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels ) +{ + struct xmesa_renderbuffer *xrb + = xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped); + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + const SWcontext *swrast = SWRAST_CONTEXT( ctx ); + XMesaDisplay *dpy = xmesa->xm_visual->display; + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + const XMesaGC gc = xmbuf->cleargc; /* effected by glColorMask */ + + ASSERT(dpy); + ASSERT(gc); + ASSERT(xmesa->xm_visual->undithered_pf == PF_5R6G5B); + + if (swrast->NewState) + _swrast_validate_derived( ctx ); + + if (xrb->pixmap && /* drawing to pixmap or window */ + format == GL_RGB && + type == GL_UNSIGNED_SHORT_5_6_5 && + !ctx->Color.DitherFlag && /* no dithering */ + (swrast->_RasterMask & ~CLIP_BIT) == 0 && /* no blend, z-test, etc */ + ctx->_ImageTransferState == 0 && /* no color tables, scale/bias, etc */ + ctx->Pixel.ZoomX == 1.0 && /* no zooming */ + ctx->Pixel.ZoomY == 1.0) { + int dstX = x; + int dstY = y; + int w = width; + int h = height; + struct gl_pixelstore_attrib clippedUnpack = *unpack; + + if (unpack->BufferObj->Name) { + /* unpack from PBO */ + GLubyte *buf; + if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, + format, type, pixels)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glDrawPixels(invalid PBO access)"); + return; + } + buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, + GL_PIXEL_UNPACK_BUFFER_EXT, + GL_READ_ONLY_ARB, + unpack->BufferObj); + if (!buf) { + /* buffer is already mapped - that's an error */ + _mesa_error(ctx, GL_INVALID_OPERATION, + "glDrawPixels(PBO is mapped)"); + return; + } + pixels = ADD_POINTERS(buf, pixels); + } + + if (_mesa_clip_drawpixels(ctx, &dstX, &dstY, &w, &h, &clippedUnpack)) { + /* This is a little tricky since all coordinates up to now have + * been in the OpenGL bottom-to-top orientation. X is top-to-bottom + * so we have to carefully compute the Y coordinates/addresses here. + */ + int srcX = clippedUnpack.SkipPixels; + int srcY = clippedUnpack.SkipRows; + int rowLength = clippedUnpack.RowLength; + XMesaImage ximage; + MEMSET(&ximage, 0, sizeof(XMesaImage)); + ximage.width = width; + ximage.height = height; + ximage.format = ZPixmap; + ximage.data = (char *) pixels + + ((srcY + h - 1) * rowLength + srcX) * 2; + ximage.byte_order = LSBFirst; + ximage.bitmap_unit = 16; + ximage.bitmap_bit_order = LSBFirst; + ximage.bitmap_pad = 16; + ximage.depth = 16; + ximage.bytes_per_line = -rowLength * 2; /* negative to flip image */ + ximage.bits_per_pixel = 16; + /* it seems we don't need to set the ximage.red/green/blue_mask fields */ + /* flip Y axis for dest position */ + dstY = YFLIP(xrb, dstY) - h + 1; + XPutImage(dpy, xrb->pixmap, gc, &ximage, 0, 0, dstX, dstY, w, h); + } + + if (unpack->BufferObj->Name) { + ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, + unpack->BufferObj); + } + } + else { + /* software fallback */ + _swrast_DrawPixels(ctx, x, y, width, height, + format, type, unpack, pixels); + } +} + + + +/** + * Implement glCopyPixels for the front color buffer (or back buffer Pixmap) + * for the color buffer. Don't support zooming, pixel transfer, etc. + * We do support copying from one window to another, ala glXMakeCurrentRead. + */ +static void +xmesa_CopyPixels( GLcontext *ctx, + GLint srcx, GLint srcy, GLsizei width, GLsizei height, + GLint destx, GLint desty, GLenum type ) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + const SWcontext *swrast = SWRAST_CONTEXT( ctx ); + XMesaDisplay *dpy = xmesa->xm_visual->display; + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + const XMesaGC gc = xmbuf->cleargc; /* effected by glColorMask */ + struct xmesa_renderbuffer *srcXrb + = xmesa_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer->Wrapped); + struct xmesa_renderbuffer *dstXrb + = xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped); + + ASSERT(dpy); + ASSERT(gc); + + if (swrast->NewState) + _swrast_validate_derived( ctx ); + + if (ctx->Color.DrawBuffer[0] == GL_FRONT && + ctx->Pixel.ReadBuffer == GL_FRONT && + srcXrb->pixmap && + dstXrb->pixmap && + type == GL_COLOR && + (swrast->_RasterMask & ~CLIP_BIT) == 0 && /* no blend, z-test, etc */ + ctx->_ImageTransferState == 0 && /* no color tables, scale/bias, etc */ + ctx->Pixel.ZoomX == 1.0 && /* no zooming */ + ctx->Pixel.ZoomY == 1.0) { + /* Note: we don't do any special clipping work here. We could, + * but X will do it for us. + */ + srcy = YFLIP(srcXrb, srcy) - height + 1; + desty = YFLIP(dstXrb, desty) - height + 1; + XCopyArea(dpy, srcXrb->pixmap, dstXrb->pixmap, gc, + srcx, srcy, width, height, destx, desty); + } + else { + _swrast_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type ); + } +} +#endif /* XFree86Server */ + + + +/* + * Every driver should implement a GetString function in order to + * return a meaningful GL_RENDERER string. + */ +static const GLubyte * +get_string( GLcontext *ctx, GLenum name ) +{ + (void) ctx; + switch (name) { + case GL_RENDERER: +#ifdef XFree86Server + return (const GLubyte *) "Mesa GLX Indirect"; +#else + return (const GLubyte *) "Mesa X11 (softpipe)"; +#endif + case GL_VENDOR: +#ifdef XFree86Server + return (const GLubyte *) "Mesa project: www.mesa3d.org"; +#else + return NULL; +#endif + default: + return NULL; + } +} + + +/* + * We implement the glEnable function only because we care about + * dither enable/disable. + */ +static void +enable( GLcontext *ctx, GLenum pname, GLboolean state ) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + + switch (pname) { + case GL_DITHER: + if (state) + xmesa->pixelformat = xmesa->xm_visual->dithered_pf; + else + xmesa->pixelformat = xmesa->xm_visual->undithered_pf; + break; + default: + ; /* silence compiler warning */ + } +} + + +static void +clear_color_HPCR_ximage( GLcontext *ctx, const GLfloat color[4] ) +{ + int i; + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[0], color[0]); + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[1], color[1]); + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[2], color[2]); + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[3], color[3]); + + if (color[0] == 0.0 && color[1] == 0.0 && color[2] == 0.0) { + /* black is black */ + MEMSET( xmesa->xm_visual->hpcr_clear_ximage_pattern, 0x0 , + sizeof(xmesa->xm_visual->hpcr_clear_ximage_pattern)); + } + else { + /* build clear pattern */ + for (i=0; i<16; i++) { + xmesa->xm_visual->hpcr_clear_ximage_pattern[0][i] = + DITHER_HPCR(i, 0, + xmesa->clearcolor[0], + xmesa->clearcolor[1], + xmesa->clearcolor[2]); + xmesa->xm_visual->hpcr_clear_ximage_pattern[1][i] = + DITHER_HPCR(i, 1, + xmesa->clearcolor[0], + xmesa->clearcolor[1], + xmesa->clearcolor[2]); + } + } +} + + +static void +clear_color_HPCR_pixmap( GLcontext *ctx, const GLfloat color[4] ) +{ + int i; + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[0], color[0]); + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[1], color[1]); + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[2], color[2]); + CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[3], color[3]); + + if (color[0] == 0.0 && color[1] == 0.0 && color[2] == 0.0) { + /* black is black */ + for (i=0; i<16; i++) { + XMesaPutPixel(xmesa->xm_visual->hpcr_clear_ximage, i, 0, 0); + XMesaPutPixel(xmesa->xm_visual->hpcr_clear_ximage, i, 1, 0); + } + } + else { + for (i=0; i<16; i++) { + XMesaPutPixel(xmesa->xm_visual->hpcr_clear_ximage, i, 0, + DITHER_HPCR(i, 0, + xmesa->clearcolor[0], + xmesa->clearcolor[1], + xmesa->clearcolor[2])); + XMesaPutPixel(xmesa->xm_visual->hpcr_clear_ximage, i, 1, + DITHER_HPCR(i, 1, + xmesa->clearcolor[0], + xmesa->clearcolor[1], + xmesa->clearcolor[2])); + } + } + /* change tile pixmap content */ + XMesaPutImage(xmesa->display, + (XMesaDrawable)xmesa->xm_visual->hpcr_clear_pixmap, + XMESA_BUFFER(ctx->DrawBuffer)->cleargc, + xmesa->xm_visual->hpcr_clear_ximage, 0, 0, 0, 0, 16, 2); +} + + +/** + * Called when the driver should update its state, based on the new_state + * flags. + */ +void +xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + + /* Propagate statechange information to swrast and swrast_setup + * modules. The X11 driver has no internal GL-dependent state. + */ + _swrast_InvalidateState( ctx, new_state ); + _tnl_InvalidateState( ctx, new_state ); + _vbo_InvalidateState( ctx, new_state ); + _swsetup_InvalidateState( ctx, new_state ); + + st_invalidate_state( ctx, new_state ); + + + if (ctx->DrawBuffer->Name != 0) + return; + + /* + * GL_DITHER, GL_READ/DRAW_BUFFER, buffer binding state, etc. effect + * renderbuffer span/clear funcs. + */ + if (new_state & (_NEW_COLOR | _NEW_BUFFERS)) { + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + struct xmesa_renderbuffer *front_xrb, *back_xrb; + + front_xrb = xmbuf->frontxrb; + if (front_xrb) { + xmesa_set_renderbuffer_funcs(front_xrb, xmesa->pixelformat, + xmesa->xm_visual->BitsPerPixel); + front_xrb->clearFunc = clear_pixmap; + } + + back_xrb = xmbuf->backxrb; + if (back_xrb) { + xmesa_set_renderbuffer_funcs(back_xrb, xmesa->pixelformat, + xmesa->xm_visual->BitsPerPixel); + if (xmbuf->backxrb->pixmap) { + back_xrb->clearFunc = clear_pixmap; + } + else { + switch (xmesa->xm_visual->BitsPerPixel) { + case 8: + if (xmesa->xm_visual->hpcr_clear_flag) { + back_xrb->clearFunc = clear_HPCR_ximage; + } + else { + back_xrb->clearFunc = clear_8bit_ximage; + } + break; + case 16: + back_xrb->clearFunc = clear_16bit_ximage; + break; + case 24: + back_xrb->clearFunc = clear_24bit_ximage; + break; + case 32: + back_xrb->clearFunc = clear_32bit_ximage; + break; + default: + back_xrb->clearFunc = clear_nbit_ximage; + break; + } + } + } + } + + if (xmesa->xm_visual->hpcr_clear_flag) { + /* this depends on whether we're drawing to the front or back buffer */ + /* XXX FIX THIS! */ +#if 0 + if (pixmap) { + ctx->Driver.ClearColor = clear_color_HPCR_pixmap; + } + else { + ctx->Driver.ClearColor = clear_color_HPCR_ximage; + } +#else + (void) clear_color_HPCR_pixmap; + (void) clear_color_HPCR_ximage; +#endif + } +} + + + +/** + * Called via ctx->Driver.TestProxyTeximage(). Normally, we'd just use + * the _mesa_test_proxy_teximage() fallback function, but we're going to + * special-case the 3D texture case to allow textures up to 512x512x32 + * texels. + */ +static GLboolean +test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, GLenum format, GLenum type, + GLint width, GLint height, GLint depth, GLint border) +{ + if (target == GL_PROXY_TEXTURE_3D) { + /* special case for 3D textures */ + if (width * height * depth > 512 * 512 * 64 || + width < 2 * border || + (!ctx->Extensions.ARB_texture_non_power_of_two && + _mesa_bitcount(width - 2 * border) != 1) || + height < 2 * border || + (!ctx->Extensions.ARB_texture_non_power_of_two && + _mesa_bitcount(height - 2 * border) != 1) || + depth < 2 * border || + (!ctx->Extensions.ARB_texture_non_power_of_two && + _mesa_bitcount(depth - 2 * border) != 1)) { + /* Bad size, or too many texels */ + return GL_FALSE; + } + return GL_TRUE; + } + else { + /* use the fallback routine for 1D, 2D, cube and rect targets */ + return _mesa_test_proxy_teximage(ctx, target, level, internalFormat, + format, type, width, height, depth, + border); + } +} + + +/** + * In SW, we don't really compress GL_COMPRESSED_RGB[A] textures! + */ +static const struct gl_texture_format * +choose_tex_format( GLcontext *ctx, GLint internalFormat, + GLenum format, GLenum type ) +{ + switch (internalFormat) { + case GL_COMPRESSED_RGB_ARB: + return &_mesa_texformat_rgb; + case GL_COMPRESSED_RGBA_ARB: + return &_mesa_texformat_rgba; + default: + return _mesa_choose_tex_format(ctx, internalFormat, format, type); + } +} + + +/** + * Called by glViewport. + * This is a good time for us to poll the current X window size and adjust + * our renderbuffers to match the current window size. + * Remember, we have no opportunity to respond to conventional + * X Resize/StructureNotify events since the X driver has no event loop. + * Thus, we poll. + * Note that this trick isn't fool-proof. If the application never calls + * glViewport, our notion of the current window size may be incorrect. + * That problem led to the GLX_MESA_resize_buffers extension. + */ +static void +xmesa_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) +{ + XMesaContext xmctx = XMESA_CONTEXT(ctx); + XMesaBuffer xmdrawbuf = XMESA_BUFFER(ctx->WinSysDrawBuffer); + XMesaBuffer xmreadbuf = XMESA_BUFFER(ctx->WinSysReadBuffer); + xmesa_check_and_update_buffer_size(xmctx, xmdrawbuf); + xmesa_check_and_update_buffer_size(xmctx, xmreadbuf); + (void) x; + (void) y; + (void) w; + (void) h; +} + + +#if ENABLE_EXT_timer_query + +/* + * The GL_EXT_timer_query extension is not enabled for the XServer + * indirect renderer. Not sure about how/if wrapping of gettimeofday() + * is done, etc. + */ + +struct xmesa_query_object +{ + struct gl_query_object Base; + struct timeval StartTime; +}; + + +static struct gl_query_object * +xmesa_new_query_object(GLcontext *ctx, GLuint id) +{ + struct xmesa_query_object *q = CALLOC_STRUCT(xmesa_query_object); + if (q) { + q->Base.Id = id; + q->Base.Ready = GL_TRUE; + } + return &q->Base; +} + + +static void +xmesa_begin_query(GLcontext *ctx, struct gl_query_object *q) +{ + if (q->Target == GL_TIME_ELAPSED_EXT) { + struct xmesa_query_object *xq = (struct xmesa_query_object *) q; + (void) gettimeofday(&xq->StartTime, NULL); + } +} + + +/** + * Return the difference between the two given times in microseconds. + */ +#ifdef __VMS +#define suseconds_t unsigned int +#endif +static GLuint64EXT +time_diff(const struct timeval *t0, const struct timeval *t1) +{ + GLuint64EXT seconds0 = t0->tv_sec & 0xff; /* 0 .. 255 seconds */ + GLuint64EXT seconds1 = t1->tv_sec & 0xff; /* 0 .. 255 seconds */ + GLuint64EXT nanosec0 = (seconds0 * 1000000 + t0->tv_usec) * 1000; + GLuint64EXT nanosec1 = (seconds1 * 1000000 + t1->tv_usec) * 1000; + return nanosec1 - nanosec0; +} + + +static void +xmesa_end_query(GLcontext *ctx, struct gl_query_object *q) +{ + if (q->Target == GL_TIME_ELAPSED_EXT) { + struct xmesa_query_object *xq = (struct xmesa_query_object *) q; + struct timeval endTime; + (void) gettimeofday(&endTime, NULL); + /* result is in nanoseconds! */ + q->Result = time_diff(&xq->StartTime, &endTime); + } + q->Ready = GL_TRUE; +} + +#endif /* ENABLE_timer_query */ + + +/** + * Initialize the device driver function table with the functions + * we implement in this driver. + */ +void +xmesa_init_driver_functions( XMesaVisual xmvisual, + struct dd_function_table *driver ) +{ + driver->GetString = get_string; + driver->UpdateState = xmesa_update_state; + driver->GetBufferSize = NULL; /* OBSOLETE */ + driver->Flush = finish_or_flush; + driver->Finish = finish_or_flush; + driver->ClearIndex = clear_index; + driver->ClearColor = clear_color; + driver->IndexMask = index_mask; + driver->ColorMask = color_mask; + driver->Enable = enable; + driver->Clear = xmesa_clear_buffers; + driver->Viewport = xmesa_viewport; +#ifndef XFree86Server + driver->CopyPixels = xmesa_CopyPixels; + if (xmvisual->undithered_pf == PF_8R8G8B && + xmvisual->dithered_pf == PF_8R8G8B) { + driver->DrawPixels = xmesa_DrawPixels_8R8G8B; + } + else if (xmvisual->undithered_pf == PF_5R6G5B) { + driver->DrawPixels = xmesa_DrawPixels_5R6G5B; + } +#endif + driver->TestProxyTexImage = test_proxy_teximage; +#if ENABLE_EXT_texure_compression_s3tc + driver->ChooseTextureFormat = choose_tex_format; +#else + (void) choose_tex_format; +#endif + +#if ENABLE_EXT_timer_query + driver->NewQueryObject = xmesa_new_query_object; + driver->BeginQuery = xmesa_begin_query; + driver->EndQuery = xmesa_end_query; +#endif + +} + + +#define XMESA_NEW_POINT (_NEW_POINT | \ + _NEW_RENDERMODE | \ + _SWRAST_NEW_RASTERMASK) + +#define XMESA_NEW_LINE (_NEW_LINE | \ + _NEW_TEXTURE | \ + _NEW_LIGHT | \ + _NEW_DEPTH | \ + _NEW_RENDERMODE | \ + _SWRAST_NEW_RASTERMASK) + +#define XMESA_NEW_TRIANGLE (_NEW_POLYGON | \ + _NEW_TEXTURE | \ + _NEW_LIGHT | \ + _NEW_DEPTH | \ + _NEW_RENDERMODE | \ + _SWRAST_NEW_RASTERMASK) + + +/** + * Extend the software rasterizer with our line/point/triangle + * functions. + * Called during context creation only. + */ +void xmesa_register_swrast_functions( GLcontext *ctx ) +{ + SWcontext *swrast = SWRAST_CONTEXT( ctx ); + + swrast->choose_point = xmesa_choose_point; + swrast->choose_line = xmesa_choose_line; + swrast->choose_triangle = xmesa_choose_triangle; + + /* XXX these lines have no net effect. Remove??? */ + swrast->InvalidatePointMask |= XMESA_NEW_POINT; + swrast->InvalidateLineMask |= XMESA_NEW_LINE; + swrast->InvalidateTriangleMask |= XMESA_NEW_TRIANGLE; +} diff --git a/src/mesa/pipe/xlib/xm_image.c b/src/mesa/pipe/xlib/xm_image.c new file mode 100644 index 0000000000..087b4e4c3a --- /dev/null +++ b/src/mesa/pipe/xlib/xm_image.c @@ -0,0 +1,133 @@ +/************************************************************************** + +Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sub license, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial portions +of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR +ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Kevin E. Martin + * Brian Paul + */ + +#include +#include + +#include "glxheader.h" +#include "xmesaP.h" + +#ifdef XFree86Server + +#ifdef ROUNDUP +#undef ROUNDUP +#endif + +#define ROUNDUP(nbytes, pad) ((((nbytes) + ((pad)-1)) / (pad)) * ((pad)>>3)) + +XMesaImage *XMesaCreateImage(int bitsPerPixel, int width, int height, char *data) +{ + XMesaImage *image; + + image = (XMesaImage *)xalloc(sizeof(XMesaImage)); + + if (image) { + image->width = width; + image->height = height; + image->data = data; + /* Always pad to 32 bits */ + image->bytes_per_line = ROUNDUP((bitsPerPixel * width), 32); + image->bits_per_pixel = bitsPerPixel; + } + + return image; +} + +void XMesaDestroyImage(XMesaImage *image) +{ + if (image->data) + free(image->data); + xfree(image); +} + +unsigned long XMesaGetPixel(XMesaImage *image, int x, int y) +{ + CARD8 *row = (CARD8 *)(image->data + y*image->bytes_per_line); + CARD8 *i8; + CARD16 *i16; + CARD32 *i32; + switch (image->bits_per_pixel) { + case 8: + i8 = (CARD8 *)row; + return i8[x]; + break; + case 15: + case 16: + i16 = (CARD16 *)row; + return i16[x]; + break; + case 24: /* WARNING: architecture specific code */ + i8 = (CARD8 *)row; + return (((CARD32)i8[x*3]) | + (((CARD32)i8[x*3+1])<<8) | + (((CARD32)i8[x*3+2])<<16)); + break; + case 32: + i32 = (CARD32 *)row; + return i32[x]; + break; + } + return 0; +} + +#ifndef XMESA_USE_PUTPIXEL_MACRO +void XMesaPutPixel(XMesaImage *image, int x, int y, unsigned long pixel) +{ + CARD8 *row = (CARD8 *)(image->data + y*image->bytes_per_line); + CARD8 *i8; + CARD16 *i16; + CARD32 *i32; + switch (image->bits_per_pixel) { + case 8: + i8 = (CARD8 *)row; + i8[x] = (CARD8)pixel; + break; + case 15: + case 16: + i16 = (CARD16 *)row; + i16[x] = (CARD16)pixel; + break; + case 24: /* WARNING: architecture specific code */ + i8 = (CARD8 *)__row; + i8[x*3] = (CARD8)(p); + i8[x*3+1] = (CARD8)(p>>8); + i8[x*3+2] = (CARD8)(p>>16); + case 32: + i32 = (CARD32 *)row; + i32[x] = (CARD32)pixel; + break; + } +} +#endif + +#endif /* XFree86Server */ diff --git a/src/mesa/pipe/xlib/xm_image.h b/src/mesa/pipe/xlib/xm_image.h new file mode 100644 index 0000000000..2a5e0f3777 --- /dev/null +++ b/src/mesa/pipe/xlib/xm_image.h @@ -0,0 +1,77 @@ +/************************************************************************** + +Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sub license, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial portions +of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR +ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Kevin E. Martin + * Brian Paul + */ + +#ifndef _XM_IMAGE_H_ +#define _XM_IMAGE_H_ + +#define XMESA_USE_PUTPIXEL_MACRO + +extern XMesaImage *XMesaCreateImage(int bitsPerPixel, int width, int height, + char *data); +extern void XMesaDestroyImage(XMesaImage *image); +extern unsigned long XMesaGetPixel(XMesaImage *image, int x, int y); +#ifdef XMESA_USE_PUTPIXEL_MACRO +#define XMesaPutPixel(__i,__x,__y,__p) \ +{ \ + CARD8 *__row = (CARD8 *)(__i->data + __y*__i->bytes_per_line); \ + CARD8 *__i8; \ + CARD16 *__i16; \ + CARD32 *__i32; \ + switch (__i->bits_per_pixel) { \ + case 8: \ + __i8 = (CARD8 *)__row; \ + __i8[__x] = (CARD8)__p; \ + break; \ + case 15: \ + case 16: \ + __i16 = (CARD16 *)__row; \ + __i16[__x] = (CARD16)__p; \ + break; \ + case 24: /* WARNING: architecture specific code */ \ + __i8 = (CARD8 *)__row; \ + __i8[__x*3] = (CARD8)(__p); \ + __i8[__x*3+1] = (CARD8)(__p>>8); \ + __i8[__x*3+2] = (CARD8)(__p>>16); \ + break; \ + case 32: \ + __i32 = (CARD32 *)__row; \ + __i32[__x] = (CARD32)__p; \ + break; \ + } \ +} +#else +extern void XMesaPutPixel(XMesaImage *image, int x, int y, + unsigned long pixel); +#endif + +#endif /* _XM_IMAGE_H_ */ diff --git a/src/mesa/pipe/xlib/xm_line.c b/src/mesa/pipe/xlib/xm_line.c new file mode 100644 index 0000000000..deeae5019c --- /dev/null +++ b/src/mesa/pipe/xlib/xm_line.c @@ -0,0 +1,691 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* + * This file contains "accelerated" point, line, and triangle functions. + * It should be fairly easy to write new special-purpose point, line or + * triangle functions and hook them into this module. + */ + + +#include "glxheader.h" +#include "depth.h" +#include "macros.h" +#include "mtypes.h" +#include "xmesaP.h" + +/* Internal swrast includes: + */ +#include "swrast/s_depth.h" +#include "swrast/s_points.h" +#include "swrast/s_lines.h" +#include "swrast/s_context.h" + + +/**********************************************************************/ +/*** Point rendering ***/ +/**********************************************************************/ + + +/* + * Render an array of points into a pixmap, any pixel format. + */ +#if 000 +/* XXX don't use this, it doesn't dither correctly */ +static void draw_points_ANY_pixmap( GLcontext *ctx, const SWvertex *vert ) +{ + XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaDisplay *dpy = xmesa->xm_visual->display; + XMesaDrawable buffer = xmesa->xm_buffer->buffer; + XMesaGC gc = xmesa->xm_buffer->gc; + + if (xmesa->xm_visual->mesa_visual.RGBAflag) { + register int x, y; + const GLubyte *color = vert->color; + unsigned long pixel = xmesa_color_to_pixel( xmesa, + color[0], color[1], + color[2], color[3], + xmesa->pixelformat); + XMesaSetForeground( dpy, gc, pixel ); + x = (GLint) vert->win[0]; + y = YFLIP( xrb, (GLint) vert->win[1] ); + XMesaDrawPoint( dpy, buffer, gc, x, y); + } + else { + /* Color index mode */ + register int x, y; + XMesaSetForeground( dpy, gc, vert->index ); + x = (GLint) vert->win[0]; + y = YFLIP( xrb, (GLint) vert->win[1] ); + XMesaDrawPoint( dpy, buffer, gc, x, y); + } +} +#endif + + +/* Override the swrast point-selection function. Try to use one of + * our internal point functions, otherwise fall back to the standard + * swrast functions. + */ +void xmesa_choose_point( GLcontext *ctx ) +{ +#if 0 + XMesaContext xmesa = XMESA_CONTEXT(ctx); + SWcontext *swrast = SWRAST_CONTEXT(ctx); + + if (ctx->RenderMode == GL_RENDER + && ctx->Point.Size == 1.0F && !ctx->Point.SmoothFlag + && swrast->_RasterMask == 0 + && !ctx->Texture._EnabledUnits + && xmesa->xm_buffer->buffer != XIMAGE) { + swrast->Point = draw_points_ANY_pixmap; + } + else { + _swrast_choose_point( ctx ); + } +#else + _swrast_choose_point( ctx ); +#endif +} + + + +/**********************************************************************/ +/*** Line rendering ***/ +/**********************************************************************/ + + +#if CHAN_BITS == 8 + + +#define GET_XRB(XRB) struct xmesa_renderbuffer *XRB = \ + xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped) + + +/* + * Draw a flat-shaded, PF_TRUECOLOR line into an XImage. + */ +#define NAME flat_TRUECOLOR_line +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + unsigned long pixel; \ + PACK_TRUECOLOR( pixel, color[0], color[1], color[2] ); +#define CLIP_HACK 1 +#define PLOT(X,Y) XMesaPutPixel(xrb->ximage, X, YFLIP(xrb, Y), pixel ); +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, PF_8A8B8G8R line into an XImage. + */ +#define NAME flat_8A8B8G8R_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLuint pixel = PACK_8A8B8G8R(color[0], color[1], color[2], color[3]); +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) *pixelPtr = pixel; +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, PF_8A8R8G8B line into an XImage. + */ +#define NAME flat_8A8R8G8B_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLuint pixel = PACK_8A8R8G8B(color[0], color[1], color[2], color[3]); +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) *pixelPtr = pixel; +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, PF_8R8G8B line into an XImage. + */ +#define NAME flat_8R8G8B_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLuint pixel = PACK_8R8G8B( color[0], color[1], color[2] ); +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) *pixelPtr = pixel; +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, PF_8R8G8B24 line into an XImage. + */ +#define NAME flat_8R8G8B24_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; +#define PIXEL_TYPE bgr_t +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) { \ + pixelPtr->r = color[RCOMP]; \ + pixelPtr->g = color[GCOMP]; \ + pixelPtr->b = color[BCOMP]; \ +} +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, PF_5R6G5B line into an XImage. + */ +#define NAME flat_5R6G5B_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLushort pixel = PACK_5R6G5B( color[0], color[1], color[2] ); +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) *pixelPtr = pixel; +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, PF_DITHER_5R6G5B line into an XImage. + */ +#define NAME flat_DITHER_5R6G5B_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + const GLubyte *color = vert1->color; +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) PACK_TRUEDITHER( *pixelPtr, X, Y, color[0], color[1], color[2] ); +#include "swrast/s_linetemp.h" + + + + +/* + * Draw a flat-shaded, PF_DITHER 8-bit line into an XImage. + */ +#define NAME flat_DITHER8_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLint r = color[0], g = color[1], b = color[2]; \ + DITHER_SETUP; +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) *pixelPtr = DITHER(X,Y,r,g,b); +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, PF_LOOKUP 8-bit line into an XImage. + */ +#define NAME flat_LOOKUP8_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLubyte pixel; \ + LOOKUP_SETUP; \ + pixel = (GLubyte) LOOKUP( color[0], color[1], color[2] ); +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) *pixelPtr = pixel; +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, PF_HPCR line into an XImage. + */ +#define NAME flat_HPCR_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + const GLubyte *color = vert1->color; \ + GLint r = color[0], g = color[1], b = color[2]; +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) *pixelPtr = (GLubyte) DITHER_HPCR(X,Y,r,g,b); +#include "swrast/s_linetemp.h" + + + + +/* + * Draw a flat-shaded, Z-less, PF_TRUECOLOR line into an XImage. + */ +#define NAME flat_TRUECOLOR_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + const GLubyte *color = vert1->color; \ + unsigned long pixel; \ + PACK_TRUECOLOR( pixel, color[0], color[1], color[2] ); +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + XMesaPutPixel(xrb->ximage, X, YFLIP(xrb, Y), pixel); \ + } +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, Z-less, PF_8A8B8G8R line into an XImage. + */ +#define NAME flat_8A8B8G8R_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLuint pixel = PACK_8A8B8G8R(color[0], color[1], color[2], color[3]); +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + *pixelPtr = pixel; \ + } +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, Z-less, PF_8A8R8G8B line into an XImage. + */ +#define NAME flat_8A8R8G8B_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLuint pixel = PACK_8A8R8G8B(color[0], color[1], color[2], color[3]); +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + *pixelPtr = pixel; \ + } +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, Z-less, PF_8R8G8B line into an XImage. + */ +#define NAME flat_8R8G8B_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLuint pixel = PACK_8R8G8B( color[0], color[1], color[2] ); +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + *pixelPtr = pixel; \ + } +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, Z-less, PF_8R8G8B24 line into an XImage. + */ +#define NAME flat_8R8G8B24_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_TYPE bgr_t +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + pixelPtr->r = color[RCOMP]; \ + pixelPtr->g = color[GCOMP]; \ + pixelPtr->b = color[BCOMP]; \ + } +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, Z-less, PF_5R6G5B line into an XImage. + */ +#define NAME flat_5R6G5B_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLushort pixel = PACK_5R6G5B( color[0], color[1], color[2] ); +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + *pixelPtr = pixel; \ + } +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, Z-less, PF_DITHER_5R6G5B line into an XImage. + */ +#define NAME flat_DITHER_5R6G5B_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + const GLubyte *color = vert1->color; +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + PACK_TRUEDITHER(*pixelPtr, X, Y, color[0], color[1], color[2]); \ + } +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, Z-less, PF_DITHER 8-bit line into an XImage. + */ +#define NAME flat_DITHER8_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLint r = color[0], g = color[1], b = color[2]; \ + DITHER_SETUP; +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + *pixelPtr = (GLubyte) DITHER( X, Y, r, g, b); \ + } +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, Z-less, PF_LOOKUP 8-bit line into an XImage. + */ +#define NAME flat_LOOKUP8_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = vert1->color; \ + GLubyte pixel; \ + LOOKUP_SETUP; \ + pixel = (GLubyte) LOOKUP( color[0], color[1], color[2] ); +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + *pixelPtr = pixel; \ + } +#include "swrast/s_linetemp.h" + + + +/* + * Draw a flat-shaded, Z-less, PF_HPCR line into an XImage. + */ +#define NAME flat_HPCR_z_line +#define SETUP_CODE \ + GET_XRB(xrb); \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + const GLubyte *color = vert1->color; \ + GLint r = color[0], g = color[1], b = color[2]; +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) +#define CLIP_HACK 1 +#define PLOT(X,Y) \ + if (Z < *zPtr) { \ + *zPtr = Z; \ + *pixelPtr = (GLubyte) DITHER_HPCR( X, Y, r, g, b); \ + } +#include "swrast/s_linetemp.h" + + + + +#ifndef XFree86Server +/** + * Draw fast, XOR line with XDrawLine in front color buffer. + * WARNING: this isn't fully OpenGL conformant because different pixels + * will be hit versus using the other line functions. + * Don't use the code in X server GLcore module since we need a wrapper + * for the XSetLineAttributes() function call. + */ +static void +xor_line(GLcontext *ctx, const SWvertex *vert0, const SWvertex *vert1) +{ + XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaDisplay *dpy = xmesa->xm_visual->display; + XMesaGC gc = xmesa->xm_buffer->gc; + GET_XRB(xrb); + unsigned long pixel = xmesa_color_to_pixel(ctx, + vert1->color[0], vert1->color[1], + vert1->color[2], vert1->color[3], + xmesa->pixelformat); + int x0 = (GLint) vert0->attrib[FRAG_ATTRIB_WPOS][0]; + int y0 = YFLIP(xrb, (GLint) vert0->attrib[FRAG_ATTRIB_WPOS][1]); + int x1 = (GLint) vert1->attrib[FRAG_ATTRIB_WPOS][0]; + int y1 = YFLIP(xrb, (GLint) vert1->attrib[FRAG_ATTRIB_WPOS][1]); + XMesaSetForeground(dpy, gc, pixel); + XMesaSetFunction(dpy, gc, GXxor); + XSetLineAttributes(dpy, gc, (int) ctx->Line.Width, + LineSolid, CapButt, JoinMiter); + XDrawLine(dpy, xrb->pixmap, gc, x0, y0, x1, y1); + XMesaSetFunction(dpy, gc, GXcopy); /* this gc is used elsewhere */ +} +#endif /* XFree86Server */ + + +#endif /* CHAN_BITS == 8 */ + + +/** + * Return pointer to line drawing function, or NULL if we should use a + * swrast fallback. + */ +static swrast_line_func +get_line_func(GLcontext *ctx) +{ +#if CHAN_BITS == 8 + SWcontext *swrast = SWRAST_CONTEXT(ctx); + XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + const int depth = GET_VISUAL_DEPTH(xmesa->xm_visual); + const struct xmesa_renderbuffer *xrb; + + if ((ctx->DrawBuffer->_ColorDrawBufferMask[0] + & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)) == 0) + return (swrast_line_func) NULL; + if (ctx->RenderMode != GL_RENDER) return (swrast_line_func) NULL; + if (ctx->Line.SmoothFlag) return (swrast_line_func) NULL; + if (ctx->Texture._EnabledUnits) return (swrast_line_func) NULL; + if (ctx->Light.ShadeModel != GL_FLAT) return (swrast_line_func) NULL; + if (ctx->Line.StippleFlag) return (swrast_line_func) NULL; + if (swrast->_RasterMask & MULTI_DRAW_BIT) return (swrast_line_func) NULL; + if (xmbuf->swAlpha) return (swrast_line_func) NULL; + + xrb = xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped); + + if (xrb->ximage + && swrast->_RasterMask==DEPTH_BIT + && ctx->Depth.Func==GL_LESS + && ctx->Depth.Mask==GL_TRUE + && ctx->Visual.depthBits == DEFAULT_SOFTWARE_DEPTH_BITS + && ctx->Line.Width==1.0F) { + switch (xmesa->pixelformat) { + case PF_Truecolor: + return flat_TRUECOLOR_z_line; + case PF_8A8B8G8R: + return flat_8A8B8G8R_z_line; + case PF_8A8R8G8B: + return flat_8A8R8G8B_z_line; + case PF_8R8G8B: + return flat_8R8G8B_z_line; + case PF_8R8G8B24: + return flat_8R8G8B24_z_line; + case PF_5R6G5B: + return flat_5R6G5B_z_line; + case PF_Dither_5R6G5B: + return flat_DITHER_5R6G5B_z_line; + case PF_Dither: + return (depth==8) ? flat_DITHER8_z_line : (swrast_line_func) NULL; + case PF_Lookup: + return (depth==8) ? flat_LOOKUP8_z_line : (swrast_line_func) NULL; + case PF_HPCR: + return flat_HPCR_z_line; + default: + return (swrast_line_func)NULL; + } + } + if (xrb->ximage + && swrast->_RasterMask==0 + && ctx->Line.Width==1.0F) { + switch (xmesa->pixelformat) { + case PF_Truecolor: + return flat_TRUECOLOR_line; + case PF_8A8B8G8R: + return flat_8A8B8G8R_line; + case PF_8A8R8G8B: + return flat_8A8R8G8B_line; + case PF_8R8G8B: + return flat_8R8G8B_line; + case PF_8R8G8B24: + return flat_8R8G8B24_line; + case PF_5R6G5B: + return flat_5R6G5B_line; + case PF_Dither_5R6G5B: + return flat_DITHER_5R6G5B_line; + case PF_Dither: + return (depth==8) ? flat_DITHER8_line : (swrast_line_func) NULL; + case PF_Lookup: + return (depth==8) ? flat_LOOKUP8_line : (swrast_line_func) NULL; + case PF_HPCR: + return flat_HPCR_line; + default: + return (swrast_line_func)NULL; + } + } + +#ifndef XFree86Server + if (ctx->DrawBuffer->_NumColorDrawBuffers[0] == 1 + && ctx->DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT + && swrast->_RasterMask == LOGIC_OP_BIT + && ctx->Color.LogicOp == GL_XOR + && !ctx->Line.StippleFlag + && !ctx->Line.SmoothFlag) { + return xor_line; + } +#endif /* XFree86Server */ + +#endif /* CHAN_BITS == 8 */ + return (swrast_line_func) NULL; +} + + +/** + * Override for the swrast line-selection function. Try to use one + * of our internal line functions, otherwise fall back to the + * standard swrast functions. + */ +void +xmesa_choose_line(GLcontext *ctx) +{ + SWcontext *swrast = SWRAST_CONTEXT(ctx); + + if (!(swrast->Line = get_line_func( ctx ))) + _swrast_choose_line( ctx ); +} diff --git a/src/mesa/pipe/xlib/xm_span.c b/src/mesa/pipe/xlib/xm_span.c new file mode 100644 index 0000000000..ce54a18a27 --- /dev/null +++ b/src/mesa/pipe/xlib/xm_span.c @@ -0,0 +1,4815 @@ +/* + * Mesa 3-D graphics library + * Version: 6.3 + * + * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "glxheader.h" +#include "colormac.h" +#include "context.h" +#include "depth.h" +#include "drawpix.h" +#include "extensions.h" +#include "macros.h" +#include "imports.h" +#include "mtypes.h" +#include "state.h" +#include "xmesaP.h" + +#include "swrast/swrast.h" + + +/* + * The following functions are used to trap XGetImage() calls which + * generate BadMatch errors if the drawable isn't mapped. + */ + +#ifndef XFree86Server +static int caught_xgetimage_error = 0; +static int (*old_xerror_handler)( XMesaDisplay *dpy, XErrorEvent *ev ); +static unsigned long xgetimage_serial; + +/* + * This is the error handler which will be called if XGetImage fails. + */ +static int xgetimage_error_handler( XMesaDisplay *dpy, XErrorEvent *ev ) +{ + if (ev->serial==xgetimage_serial && ev->error_code==BadMatch) { + /* caught the expected error */ + caught_xgetimage_error = 0; + } + else { + /* call the original X error handler, if any. otherwise ignore */ + if (old_xerror_handler) { + (*old_xerror_handler)( dpy, ev ); + } + } + return 0; +} + + +/* + * Call this right before XGetImage to setup error trap. + */ +static void catch_xgetimage_errors( XMesaDisplay *dpy ) +{ + xgetimage_serial = NextRequest( dpy ); + old_xerror_handler = XSetErrorHandler( xgetimage_error_handler ); + caught_xgetimage_error = 0; +} + + +/* + * Call this right after XGetImage to check if an error occured. + */ +static int check_xgetimage_errors( void ) +{ + /* restore old handler */ + (void) XSetErrorHandler( old_xerror_handler ); + /* return 0=no error, 1=error caught */ + return caught_xgetimage_error; +} +#endif + + +/* + * Read a pixel from an X drawable. + */ +static unsigned long read_pixel( XMesaDisplay *dpy, + XMesaDrawable d, int x, int y ) +{ + unsigned long p; +#ifndef XFree86Server + XMesaImage *pixel = NULL; + int error; + + catch_xgetimage_errors( dpy ); + pixel = XGetImage( dpy, d, x, y, 1, 1, AllPlanes, ZPixmap ); + error = check_xgetimage_errors(); + if (pixel && !error) { + p = XMesaGetPixel( pixel, 0, 0 ); + } + else { + p = 0; + } + if (pixel) { + XMesaDestroyImage( pixel ); + } +#else + (*dpy->GetImage)(d, x, y, 1, 1, ZPixmap, ~0L, (pointer)&p); +#endif + return p; +} + + + +/* + * The Mesa library needs to be able to draw pixels in a number of ways: + * 1. RGB vs Color Index + * 2. as horizontal spans (polygons, images) vs random locations (points, + * lines) + * 3. different color per-pixel or same color for all pixels + * + * Furthermore, the X driver needs to support rendering to 3 possible + * "buffers", usually one, but sometimes two at a time: + * 1. The front buffer as an X window + * 2. The back buffer as a Pixmap + * 3. The back buffer as an XImage + * + * Finally, if the back buffer is an XImage, we can avoid using XPutPixel and + * optimize common cases such as 24-bit and 8-bit modes. + * + * By multiplication, there's at least 48 possible combinations of the above. + * + * Below are implementations of the most commonly used combinations. They are + * accessed through function pointers which get initialized here and are used + * directly from the Mesa library. The 8 function pointers directly correspond + * to the first 3 cases listed above. + * + * + * The function naming convention is: + * + * [put|get]_[mono]_[row|values]_[format]_[pixmap|ximage] + * + * New functions optimized for specific cases can be added without too much + * trouble. An example might be the 24-bit TrueColor mode 8A8R8G8B which is + * found on IBM RS/6000 X servers. + */ + + + + +/**********************************************************************/ +/*** Write COLOR SPAN functions ***/ +/**********************************************************************/ + + +#define PUT_ROW_ARGS \ + GLcontext *ctx, \ + struct gl_renderbuffer *rb, \ + GLuint n, GLint x, GLint y, \ + const void *values, const GLubyte mask[] + +#define RGB_SPAN_ARGS \ + GLcontext *ctx, \ + struct gl_renderbuffer *rb, \ + GLuint n, GLint x, GLint y, \ + const void *values, const GLubyte mask[] + + +#define GET_XRB(XRB) \ + struct xmesa_renderbuffer *XRB = xmesa_renderbuffer(rb) + + +/* + * Write a span of PF_TRUECOLOR pixels to a pixmap. + */ +static void put_row_TRUECOLOR_pixmap( PUT_ROW_ARGS ) +{ + const GLubyte (*rgba)[4] = (const GLubyte (*)[4]) values; + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GET_XRB(xrb); + XMesaDisplay *dpy = XMESA_BUFFER(ctx->DrawBuffer)->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLuint *ptr4 = (GLuint *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLuint *ptr4 = (GLuint *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLuint *ptr4 = (GLuint *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLuint *ptr4 = (GLuint *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLuint *ptr4 = (GLuint *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + y = YFLIP(xrb, y); + if (mask) { + register GLuint i; + for (i=0;iDrawBuffer)->rowimage; + register GLuint *ptr4 = (GLuint *) rowimg->data; + register GLuint pixel; + static const GLuint shift[4] = {0, 8, 16, 24}; + register GLuint i = 0; + int w = n; + while (w > 3) { + pixel = rgba[i][BCOMP] /* << shift[0]*/; + pixel |= rgba[i][GCOMP] << shift[1]; + pixel |= rgba[i++][RCOMP] << shift[2]; + pixel |= rgba[i][BCOMP] << shift[3]; + *ptr4++ = pixel; + + pixel = rgba[i][GCOMP] /* << shift[0]*/; + pixel |= rgba[i++][RCOMP] << shift[1]; + pixel |= rgba[i][BCOMP] << shift[2]; + pixel |= rgba[i][GCOMP] << shift[3]; + *ptr4++ = pixel; + + pixel = rgba[i++][RCOMP]/* << shift[0]*/; + pixel |= rgba[i][BCOMP] << shift[1]; + pixel |= rgba[i][GCOMP] << shift[2]; + pixel |= rgba[i++][RCOMP] << shift[3]; + *ptr4++ = pixel; + + w -= 4; + } + switch (w) { + case 3: + pixel = 0; + pixel |= rgba[i][BCOMP] /*<< shift[0]*/; + pixel |= rgba[i][GCOMP] << shift[1]; + pixel |= rgba[i++][RCOMP] << shift[2]; + pixel |= rgba[i][BCOMP] << shift[3]; + *ptr4++ = pixel; + pixel = 0; + pixel |= rgba[i][GCOMP] /*<< shift[0]*/; + pixel |= rgba[i++][RCOMP] << shift[1]; + pixel |= rgba[i][BCOMP] << shift[2]; + pixel |= rgba[i][GCOMP] << shift[3]; + *ptr4++ = pixel; + pixel = 0xffffff00 & *ptr4; + pixel |= rgba[i][RCOMP] /*<< shift[0]*/; + *ptr4 = pixel; + break; + case 2: + pixel = 0; + pixel |= rgba[i][BCOMP] /*<< shift[0]*/; + pixel |= rgba[i][GCOMP] << shift[1]; + pixel |= rgba[i++][RCOMP] << shift[2]; + pixel |= rgba[i][BCOMP] << shift[3]; + *ptr4++ = pixel; + pixel = 0xffff0000 & *ptr4; + pixel |= rgba[i][GCOMP] /*<< shift[0]*/; + pixel |= rgba[i][RCOMP] << shift[1]; + *ptr4 = pixel; + break; + case 1: + pixel = 0xff000000 & *ptr4; + pixel |= rgba[i][BCOMP] /*<< shift[0]*/; + pixel |= rgba[i][GCOMP] << shift[1]; + pixel |= rgba[i][RCOMP] << shift[2]; + *ptr4 = pixel; + break; + case 0: + break; + } + XMesaPutImage( dpy, buffer, gc, rowimg, 0, 0, x, y, n, 1 ); + } +} + + +/* + * Write a span of PF_8R8G8B pixels to a pixmap (no alpha). + */ +static void put_row_rgb_8R8G8B_pixmap( RGB_SPAN_ARGS ) +{ + const GLubyte (*rgb)[3] = (const GLubyte (*)[3]) values; + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GET_XRB(xrb); + XMesaDisplay *dpy = xmesa->xm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLuint *ptr4 = (GLuint *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + y = YFLIP(xrb, y); + if (mask) { + register GLuint i; + for (i=0;iDrawBuffer)->rowimage; + register GLuint *ptr4 = (GLuint *) rowimg->data; + register GLuint pixel; + static const GLuint shift[4] = {0, 8, 16, 24}; + unsigned w = n; + register GLuint i = 0; + while (w > 3) { + pixel = 0; + pixel |= rgb[i][BCOMP]/* << shift[0]*/; + pixel |= rgb[i][GCOMP] << shift[1]; + pixel |= rgb[i++][RCOMP] << shift[2]; + pixel |= rgb[i][BCOMP] <xm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLushort *ptr2 = (GLushort *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLushort *ptr2 = (GLushort *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLushort *ptr2 = (GLushort *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLushort *ptr2 = (GLushort *) rowimg->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + XDITHER_SETUP(y); + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + XDITHER_SETUP(y); + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + SETUP_1BIT; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + SETUP_1BIT; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLubyte *ptr = (GLubyte *) XMESA_BUFFER(ctx->DrawBuffer)->rowimage->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + register GLubyte *ptr = (GLubyte *) XMESA_BUFFER(ctx->DrawBuffer)->rowimage->data; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + LOOKUP_SETUP; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + LOOKUP_SETUP; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iDrawBuffer)->rowimage; + for (i=0;iximage; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;i 3) { + pixel = rgba[i][BCOMP]; + pixel |= rgba[i][GCOMP] << 8; + pixel |= rgba[i++][RCOMP] << 16; + pixel |= rgba[i][BCOMP] << 24; + *ptr4++ = pixel; + pixel = rgba[i][GCOMP]; + pixel |= rgba[i++][RCOMP] << 8; + pixel |= rgba[i][BCOMP] << 16; + pixel |= rgba[i][GCOMP] << 24; + *ptr4++ = pixel; + pixel = rgba[i++][RCOMP]; + pixel |= rgba[i][BCOMP] << 8; + pixel |= rgba[i][GCOMP] << 16; + pixel |= rgba[i++][RCOMP] << 24; + *ptr4++ = pixel; + w -= 4; + } + switch (w) { + case 0: + break; + case 1: + pixel = *ptr4 & 0xff000000; + pixel |= rgba[i][BCOMP]; + pixel |= rgba[i][GCOMP] << 8; + pixel |= rgba[i][RCOMP] << 16; + *ptr4 = pixel; + break; + case 2: + pixel = rgba[i][BCOMP]; + pixel |= rgba[i][GCOMP] << 8; + pixel |= rgba[i++][RCOMP] << 16; + pixel |= rgba[i][BCOMP] << 24; + *ptr4++ = pixel; + pixel = *ptr4 & 0xffff0000; + pixel |= rgba[i][GCOMP]; + pixel |= rgba[i][RCOMP] << 8; + *ptr4 = pixel; + break; + case 3: + pixel = rgba[i][BCOMP]; + pixel |= rgba[i][GCOMP] << 8; + pixel |= rgba[i++][RCOMP] << 16; + pixel |= rgba[i][BCOMP] << 24; + *ptr4++ = pixel; + pixel = rgba[i][GCOMP]; + pixel |= rgba[i++][RCOMP] << 8; + pixel |= rgba[i][BCOMP] << 16; + pixel |= rgba[i][GCOMP] << 24; + *ptr4++ = pixel; + pixel = *ptr4 & 0xffffff00; + pixel |= rgba[i][RCOMP]; + *ptr4 = pixel; + break; + } + } +} + + +/* + * Write a span of PF_8R8G8B-format pixels to an ximage (no alpha). + */ +static void put_row_rgb_8R8G8B_ximage( RGB_SPAN_ARGS ) +{ + const GLubyte (*rgb)[3] = (const GLubyte (*)[3]) values; + GET_XRB(xrb); + register GLuint i; + register GLuint *ptr = PIXEL_ADDR4(xrb, x, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + int yy = YFLIP(xrb, y); + XDITHER_SETUP(yy); + if (mask) { + for (i=0;iximage; + register GLuint i; + int yy = YFLIP(xrb, y); + XDITHER_SETUP(yy); + if (mask) { + for (i=0;iximage; + register GLuint i; + SETUP_1BIT; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + SETUP_1BIT; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + LOOKUP_SETUP; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + LOOKUP_SETUP; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + DITHER_SETUP; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + SETUP_1BIT; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + LOOKUP_SETUP; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;iximage; + register GLuint i; + for (i=0;iximage; + register GLuint i; + for (i=0;ir = rgba[i][RCOMP]; + ptr->g = rgba[i][GCOMP]; + ptr->b = rgba[i][BCOMP]; + } + } +} + + +/* + * Write an array of PF_5R6G5B pixels to an ximage. + */ +static void put_values_5R6G5B_ximage( PUT_VALUES_ARGS ) +{ + const GLubyte (*rgba)[4] = (const GLubyte (*)[4]) values; + GET_XRB(xrb); + register GLuint i; + for (i=0;iximage; + register GLuint i; + DITHER_SETUP; + for (i=0;iximage; + register GLuint i; + SETUP_1BIT; + for (i=0;iximage; + register GLuint i; + LOOKUP_SETUP; + for (i=0;iximage; + register GLuint i; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + const unsigned long pixel = xmesa_color_to_pixel(ctx, color[RCOMP], + color[GCOMP], color[BCOMP], color[ACOMP], xmesa->pixelformat); + register GLuint i; + XMesaSetForeground( xmesa->display, gc, pixel ); + y = YFLIP(xrb, y); + + /* New code contributed by Jeff Epler and cleaned up by Keith + * Whitwell. + */ + for (i = 0; i < n; ) { + GLuint start = i; + + /* Identify and emit contiguous rendered pixels + */ + while (i < n && (!mask || mask[i])) + i++; + + if (start < i) + XMesaFillRectangle( dpy, buffer, gc, + (int)(x+start), (int) y, + (int)(i-start), 1); + + /* Eat up non-rendered pixels + */ + while (i < n && !mask[i]) + i++; + } +} + + + +static void +put_mono_row_ci_pixmap( PUT_MONO_ROW_ARGS ) +{ + GLuint colorIndex = *((GLuint *) value); + XMesaContext xmesa = XMESA_CONTEXT(ctx); + GET_XRB(xrb); + XMesaDisplay *dpy = xmesa->xm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + XMesaSetForeground( xmesa->display, gc, colorIndex ); + y = YFLIP(xrb, y); + + for (i = 0 ; i < n ;) { + GLuint start = i; + + /* Identify and emit contiguous rendered pixels + */ + while (i < n && (!mask || mask[i])) + i++; + + if (start < i) + XMesaFillRectangle( dpy, buffer, gc, + (int)(x+start), (int) y, + (int)(i-start), 1); + + /* Eat up non-rendered pixels + */ + while (i < n && !mask[i]) + i++; + } +} + + + +/* + * Write a span of PF_TRUEDITHER pixels to a pixmap. + */ +static void put_mono_row_TRUEDITHER_pixmap( PUT_MONO_ROW_ARGS ) +{ + const GLubyte *color = (const GLubyte *) value; + GET_XRB(xrb); + XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaDisplay *dpy = xmesa->xm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; + register GLuint i; + int yy = YFLIP(xrb, y); + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; + register GLuint i; + int yy = YFLIP(xrb, y); + XDITHER_SETUP(yy); + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; + register GLuint i; + SETUP_1BIT; + y = YFLIP(xrb, y); + for (i=0;iximage; + register GLuint i; + const unsigned long pixel = xmesa_color_to_pixel(ctx, color[RCOMP], + color[GCOMP], color[BCOMP], color[ACOMP], xmesa->pixelformat); + y = YFLIP(xrb, y); + for (i=0;iximage; + register GLuint i; + y = YFLIP(xrb, y); + for (i=0;iximage; + const GLint r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; + GLuint i; + y = YFLIP(xrb, y); + for (i=0;ipixelformat); + ptr = PIXEL_ADDR4(xrb, x, y ); + for (i=0;ipixelformat); + ptr = PIXEL_ADDR4(xrb, x, y ); + for (i=0;iximage; + int yy = YFLIP(xrb, y); + register GLuint i; + XDITHER_SETUP(yy); + for (i=0;iximage; + register GLuint i; + SETUP_1BIT; + y = YFLIP(xrb, y); + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + const unsigned long pixel = xmesa_color_to_pixel(ctx, color[RCOMP], + color[GCOMP], color[BCOMP], color[ACOMP], xmesa->pixelformat); + XMesaSetForeground( xmesa->display, gc, pixel ); + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + XMesaSetForeground( xmesa->display, gc, colorIndex ); + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; + DITHER_SETUP; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; + SETUP_1BIT; + for (i=0;iximage; + register GLuint i; + const unsigned long pixel = xmesa_color_to_pixel(ctx, color[RCOMP], + color[GCOMP], color[BCOMP], color[ACOMP], xmesa->pixelformat); + for (i=0;iximage; + register GLuint i; + for (i=0;iximage; + register GLuint i; + const int r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; + for (i=0;ir = r; + ptr->g = g; + ptr->b = b; + } + } +} + + +/* + * Write an array of identical PF_DITHER pixels to an XImage. + */ +static void put_mono_values_DITHER_ximage( PUT_MONO_VALUES_ARGS ) +{ + const GLubyte *color = (const GLubyte *) value; + GET_XRB(xrb); + const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; + XMesaImage *img = xrb->ximage; + register GLuint i; + DITHER_SETUP; + for (i=0;iximage; + register GLuint i; + SETUP_1BIT; + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;iximage; + register GLuint i; + y = YFLIP(xrb, y); + if (mask) { + for (i=0;ixm_visual->display; + XMesaDrawable buffer = xrb->drawable; + XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; + register GLuint i; + for (i=0;iximage; + register GLuint i; + for (i=0;iDrawBuffer); + Window rootWin = RootWindow(xmesa->display, 0); + Window child; + GLint screenWidth = WidthOfScreen(DefaultScreenOfDisplay(xmesa->display)); + GLint dx, dy; + if (source->type == PBUFFER || source->type == PIXMAP) + return 0; + XTranslateCoordinates(xmesa->display, source->frontxrb->pixmap, rootWin, + *x, *y, &dx, &dy, &child); + if (dx >= screenWidth) { + /* totally clipped on right */ + return -1; + } + if (dx < 0) { + /* clipped on left */ + GLint clip = -dx; + if (clip >= (GLint) *n) + return -1; /* totally clipped on left */ + *x += clip; + *n -= clip; + dx = 0; + return clip; + } + if ((GLint) (dx + *n) > screenWidth) { + /* clipped on right */ + GLint clip = dx + *n - screenWidth; + *n -= clip; + } + return 0; +} +#endif + + +/* + * Read a horizontal span of color-index pixels. + */ +static void +get_row_ci(GLcontext *ctx, struct gl_renderbuffer *rb, + GLuint n, GLint x, GLint y, void *values) +{ + GLuint *index = (GLuint *) values; + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GET_XRB(xrb); + GLuint i; + + y = YFLIP(xrb, y); + + if (xrb->pixmap) { +#ifndef XFree86Server + XMesaImage *span = NULL; + int error; + int k = clip_for_xgetimage(ctx, &n, &x, &y); + if (k < 0) + return; + index += k; + + catch_xgetimage_errors( xmesa->display ); + span = XGetImage( xmesa->display, xrb->pixmap, + x, y, n, 1, AllPlanes, ZPixmap ); + error = check_xgetimage_errors(); + if (span && !error) { + for (i=0;idisplay->GetImage)(xrb->drawable, + x, y, n, 1, ZPixmap, + ~0L, (pointer)index); +#endif + } + else if (xrb->ximage) { + XMesaImage *img = xrb->ximage; + for (i=0;iDrawBuffer); + + if (xrb->pixmap) { + /* Read from Pixmap or Window */ + XMesaImage *span = NULL; + int error; +#ifdef XFree86Server + span = XMesaCreateImage(xmesa->xm_visual->BitsPerPixel, n, 1, NULL); + span->data = (char *)MALLOC(span->height * span->bytes_per_line); + error = (!span->data); + (*xmesa->display->GetImage)(xrb->drawable, + x, YFLIP(xrb, y), n, 1, ZPixmap, + ~0L, (pointer)span->data); +#else + int k; + y = YFLIP(xrb, y); + k = clip_for_xgetimage(ctx, &n, &x, &y); + if (k < 0) + return; + rgba += k; + catch_xgetimage_errors( xmesa->display ); + span = XGetImage( xmesa->display, xrb->pixmap, + x, y, n, 1, AllPlanes, ZPixmap ); + error = check_xgetimage_errors(); +#endif + if (span && !error) { + switch (xmesa->pixelformat) { + case PF_Truecolor: + case PF_Dither_True: + { + const GLubyte *pixelToR = xmesa->xm_visual->PixelToR; + const GLubyte *pixelToG = xmesa->xm_visual->PixelToG; + const GLubyte *pixelToB = xmesa->xm_visual->PixelToB; + unsigned long rMask = GET_REDMASK(xmesa->xm_visual); + unsigned long gMask = GET_GREENMASK(xmesa->xm_visual); + unsigned long bMask = GET_BLUEMASK(xmesa->xm_visual); + GLint rShift = xmesa->xm_visual->rshift; + GLint gShift = xmesa->xm_visual->gshift; + GLint bShift = xmesa->xm_visual->bshift; + GLuint i; + for (i=0;i> rShift]; + rgba[i][GCOMP] = pixelToG[(p & gMask) >> gShift]; + rgba[i][BCOMP] = pixelToB[(p & bMask) >> bShift]; + rgba[i][ACOMP] = 255; + } + } + break; + case PF_5R6G5B: + case PF_Dither_5R6G5B: + { + const GLubyte *pixelToR = xmesa->xm_visual->PixelToR; + const GLubyte *pixelToG = xmesa->xm_visual->PixelToG; + const GLubyte *pixelToB = xmesa->xm_visual->PixelToB; + GLuint i; + for (i=0;i> 8) & 0xf8); + rgba[i][GCOMP] = ((p >> 3) & 0xfc); + rgba[i][BCOMP] = ((p << 3) & 0xff); + */ + rgba[i][RCOMP] = pixelToR[p >> 11]; + rgba[i][GCOMP] = pixelToG[(p >> 5) & 0x3f]; + rgba[i][BCOMP] = pixelToB[p & 0x1f]; + rgba[i][ACOMP] = 255; + } + } + break; + case PF_8A8B8G8R: + { + const GLuint *ptr4 = (GLuint *) span->data; + GLuint i; + for (i=0;i> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ((p4 >> 16) & 0xff); + rgba[i][ACOMP] = (GLubyte) ((p4 >> 24) & 0xff); + } + } + break; + case PF_8A8R8G8B: + { + const GLuint *ptr4 = (GLuint *) span->data; + GLuint i; + for (i=0;i> 16) & 0xff); + rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); + rgba[i][ACOMP] = (GLubyte) ((p4 >> 24) & 0xff); + } + } + break; + case PF_8R8G8B: + { + const GLuint *ptr4 = (GLuint *) span->data; + GLuint i; + for (i=0;i> 16) & 0xff); + rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); + rgba[i][ACOMP] = 255; + } + } + break; + case PF_8R8G8B24: + { + const bgr_t *ptr3 = (bgr_t *) span->data; + GLuint i; + for (i=0;idata; + GLuint i; + for (i=0;ipixel_to_r; + GLubyte *gTable = source->pixel_to_g; + GLubyte *bTable = source->pixel_to_b; + if (GET_VISUAL_DEPTH(xmesa->xm_visual)==8) { + const GLubyte *ptr1 = (GLubyte *) span->data; + GLuint i; + for (i=0;ixm_visual->bitFlip; + GLuint i; + for (i=0;iximage) { + /* Read from XImage back buffer */ + switch (xmesa->pixelformat) { + case PF_Truecolor: + case PF_Dither_True: + { + const GLubyte *pixelToR = xmesa->xm_visual->PixelToR; + const GLubyte *pixelToG = xmesa->xm_visual->PixelToG; + const GLubyte *pixelToB = xmesa->xm_visual->PixelToB; + unsigned long rMask = GET_REDMASK(xmesa->xm_visual); + unsigned long gMask = GET_GREENMASK(xmesa->xm_visual); + unsigned long bMask = GET_BLUEMASK(xmesa->xm_visual); + GLint rShift = xmesa->xm_visual->rshift; + GLint gShift = xmesa->xm_visual->gshift; + GLint bShift = xmesa->xm_visual->bshift; + XMesaImage *img = xrb->ximage; + GLuint i; + y = YFLIP(xrb, y); + for (i=0;i> rShift]; + rgba[i][GCOMP] = pixelToG[(p & gMask) >> gShift]; + rgba[i][BCOMP] = pixelToB[(p & bMask) >> bShift]; + rgba[i][ACOMP] = 255; + } + } + break; + case PF_5R6G5B: + case PF_Dither_5R6G5B: + { + const GLubyte *pixelToR = xmesa->xm_visual->PixelToR; + const GLubyte *pixelToG = xmesa->xm_visual->PixelToG; + const GLubyte *pixelToB = xmesa->xm_visual->PixelToB; + const GLushort *ptr2 = PIXEL_ADDR2(xrb, x, y); + GLuint i; +#if defined(__i386__) /* word stores don't have to be on 4-byte boundaries */ + const GLuint *ptr4 = (const GLuint *) ptr2; + GLuint extraPixel = (n & 1); + n -= extraPixel; + for (i = 0; i < n; i += 2) { + const GLuint p = *ptr4++; + const GLuint p0 = p & 0xffff; + const GLuint p1 = p >> 16; + /* fast, but not quite accurate + rgba[i][RCOMP] = ((p >> 8) & 0xf8); + rgba[i][GCOMP] = ((p >> 3) & 0xfc); + rgba[i][BCOMP] = ((p << 3) & 0xff); + */ + rgba[i][RCOMP] = pixelToR[p0 >> 11]; + rgba[i][GCOMP] = pixelToG[(p0 >> 5) & 0x3f]; + rgba[i][BCOMP] = pixelToB[p0 & 0x1f]; + rgba[i][ACOMP] = 255; + rgba[i+1][RCOMP] = pixelToR[p1 >> 11]; + rgba[i+1][GCOMP] = pixelToG[(p1 >> 5) & 0x3f]; + rgba[i+1][BCOMP] = pixelToB[p1 & 0x1f]; + rgba[i+1][ACOMP] = 255; + } + if (extraPixel) { + GLushort p = ptr2[n]; + rgba[n][RCOMP] = pixelToR[p >> 11]; + rgba[n][GCOMP] = pixelToG[(p >> 5) & 0x3f]; + rgba[n][BCOMP] = pixelToB[p & 0x1f]; + rgba[n][ACOMP] = 255; + } +#else + for (i = 0; i < n; i++) { + const GLushort p = ptr2[i]; + rgba[i][RCOMP] = pixelToR[p >> 11]; + rgba[i][GCOMP] = pixelToG[(p >> 5) & 0x3f]; + rgba[i][BCOMP] = pixelToB[p & 0x1f]; + rgba[i][ACOMP] = 255; + } +#endif + } + break; + case PF_8A8B8G8R: + { + const GLuint *ptr4 = PIXEL_ADDR4(xrb, x, y); + GLuint i; + for (i=0;i> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ((p4 >> 16) & 0xff); + rgba[i][ACOMP] = (GLint) ((p4 >> 24) & 0xff); + } + } + break; + case PF_8A8R8G8B: + { + const GLuint *ptr4 = PIXEL_ADDR4(xrb, x, y); + GLuint i; + for (i=0;i> 16) & 0xff); + rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); + rgba[i][ACOMP] = (GLint) ((p4 >> 24) & 0xff); + } + } + break; + case PF_8R8G8B: + { + const GLuint *ptr4 = PIXEL_ADDR4(xrb, x, y); + GLuint i; + for (i=0;i> 16) & 0xff); + rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); + rgba[i][ACOMP] = 255; + } + } + break; + case PF_8R8G8B24: + { + const bgr_t *ptr3 = PIXEL_ADDR3(xrb, x, y); + GLuint i; + for (i=0;ipixel_to_r; + const GLubyte *gTable = source->pixel_to_g; + const GLubyte *bTable = source->pixel_to_b; + if (GET_VISUAL_DEPTH(xmesa->xm_visual)==8) { + GLubyte *ptr1 = PIXEL_ADDR1(xrb, x, y); + GLuint i; + for (i=0;iximage; + GLuint i; + y = YFLIP(xrb, y); + for (i=0;iximage; + int bitFlip = xmesa->xm_visual->bitFlip; + GLuint i; + y = YFLIP(xrb, y); + for (i=0;ipixmap) { + for (i=0;idisplay, xrb->drawable, + x[i], YFLIP(xrb, y[i]) ); + } + } + else if (xrb->ximage) { + XMesaImage *img = xrb->ximage; + for (i=0;ixm_visual->display; + XMesaBuffer source = XMESA_BUFFER(ctx->DrawBuffer); + register GLuint i; + + if (xrb->pixmap) { + XMesaDrawable buffer = xrb->drawable; + switch (xmesa->pixelformat) { + case PF_Truecolor: + case PF_Dither_True: + case PF_5R6G5B: + case PF_Dither_5R6G5B: + { + unsigned long rMask = GET_REDMASK(xmesa->xm_visual); + unsigned long gMask = GET_GREENMASK(xmesa->xm_visual); + unsigned long bMask = GET_BLUEMASK(xmesa->xm_visual); + GLubyte *pixelToR = xmesa->xm_visual->PixelToR; + GLubyte *pixelToG = xmesa->xm_visual->PixelToG; + GLubyte *pixelToB = xmesa->xm_visual->PixelToB; + GLint rShift = xmesa->xm_visual->rshift; + GLint gShift = xmesa->xm_visual->gshift; + GLint bShift = xmesa->xm_visual->bshift; + for (i=0;i> rShift]; + rgba[i][GCOMP] = pixelToG[(p & gMask) >> gShift]; + rgba[i][BCOMP] = pixelToB[(p & bMask) >> bShift]; + rgba[i][ACOMP] = 255; + } + } + break; + case PF_8A8B8G8R: + for (i=0;i> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ((p >> 16) & 0xff); + rgba[i][ACOMP] = (GLubyte) ((p >> 24) & 0xff); + } + break; + case PF_8A8R8G8B: + for (i=0;i> 16) & 0xff); + rgba[i][GCOMP] = (GLubyte) ((p >> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ( p & 0xff); + rgba[i][ACOMP] = (GLubyte) ((p >> 24) & 0xff); + } + break; + case PF_8R8G8B: + for (i=0;i> 16) & 0xff); + rgba[i][GCOMP] = (GLubyte) ((p >> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ( p & 0xff); + rgba[i][ACOMP] = 255; + } + break; + case PF_8R8G8B24: + for (i=0;i> 16) & 0xff); + rgba[i][GCOMP] = (GLubyte) ((p >> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ( p & 0xff); + rgba[i][ACOMP] = 255; + } + break; + case PF_HPCR: + for (i=0;ipixel_to_r; + GLubyte *gTable = source->pixel_to_g; + GLubyte *bTable = source->pixel_to_b; + for (i=0;ixm_visual->bitFlip; + for (i=0;iximage) { + /* Read from XImage back buffer */ + switch (xmesa->pixelformat) { + case PF_Truecolor: + case PF_Dither_True: + case PF_5R6G5B: + case PF_Dither_5R6G5B: + { + unsigned long rMask = GET_REDMASK(xmesa->xm_visual); + unsigned long gMask = GET_GREENMASK(xmesa->xm_visual); + unsigned long bMask = GET_BLUEMASK(xmesa->xm_visual); + GLubyte *pixelToR = xmesa->xm_visual->PixelToR; + GLubyte *pixelToG = xmesa->xm_visual->PixelToG; + GLubyte *pixelToB = xmesa->xm_visual->PixelToB; + GLint rShift = xmesa->xm_visual->rshift; + GLint gShift = xmesa->xm_visual->gshift; + GLint bShift = xmesa->xm_visual->bshift; + XMesaImage *img = xrb->ximage; + for (i=0;i> rShift]; + rgba[i][GCOMP] = pixelToG[(p & gMask) >> gShift]; + rgba[i][BCOMP] = pixelToB[(p & bMask) >> bShift]; + rgba[i][ACOMP] = 255; + } + } + break; + case PF_8A8B8G8R: + for (i=0;i> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ((p4 >> 16) & 0xff); + rgba[i][ACOMP] = (GLubyte) ((p4 >> 24) & 0xff); + } + break; + case PF_8A8R8G8B: + for (i=0;i> 16) & 0xff); + rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); + rgba[i][ACOMP] = (GLubyte) ((p4 >> 24) & 0xff); + } + break; + case PF_8R8G8B: + for (i=0;i> 16) & 0xff); + rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); + rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); + rgba[i][ACOMP] = 255; + } + break; + case PF_8R8G8B24: + for (i=0;ir; + rgba[i][GCOMP] = ptr3->g; + rgba[i][BCOMP] = ptr3->b; + rgba[i][ACOMP] = 255; + } + break; + case PF_HPCR: + for (i=0;ipixel_to_r; + GLubyte *gTable = source->pixel_to_g; + GLubyte *bTable = source->pixel_to_b; + XMesaImage *img = xrb->ximage; + for (i=0;iximage; + int bitFlip = xmesa->xm_visual->bitFlip; + for (i=0;ipixmap ? GL_TRUE : GL_FALSE; + struct gl_renderbuffer *rb = &xrb->St.Base; + + switch (pixelformat) { + case PF_Index: + ASSERT(rb->DataType == GL_UNSIGNED_INT); + if (pixmap) { + rb->PutRow = put_row_ci_pixmap; + rb->PutRowRGB = NULL; + rb->PutMonoRow = put_mono_row_ci_pixmap; + rb->PutValues = put_values_ci_pixmap; + rb->PutMonoValues = put_mono_values_ci_pixmap; + } + else { + rb->PutRow = put_row_ci_ximage; + rb->PutRowRGB = NULL; + rb->PutMonoRow = put_mono_row_ci_ximage; + rb->PutValues = put_values_ci_ximage; + rb->PutMonoValues = put_mono_values_ci_ximage; + } + break; + case PF_Truecolor: + if (pixmap) { + rb->PutRow = put_row_TRUECOLOR_pixmap; + rb->PutRowRGB = put_row_rgb_TRUECOLOR_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_TRUECOLOR_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; + } + else { + rb->PutRow = put_row_TRUECOLOR_ximage; + rb->PutRowRGB = put_row_rgb_TRUECOLOR_ximage; + rb->PutMonoRow = put_mono_row_ximage; + rb->PutValues = put_values_TRUECOLOR_ximage; + rb->PutMonoValues = put_mono_values_ximage; + } + break; + case PF_Dither_True: + if (pixmap) { + rb->PutRow = put_row_TRUEDITHER_pixmap; + rb->PutRowRGB = put_row_rgb_TRUEDITHER_pixmap; + rb->PutMonoRow = put_mono_row_TRUEDITHER_pixmap; + rb->PutValues = put_values_TRUEDITHER_pixmap; + rb->PutMonoValues = put_mono_values_TRUEDITHER_pixmap; + } + else { + rb->PutRow = put_row_TRUEDITHER_ximage; + rb->PutRowRGB = put_row_rgb_TRUEDITHER_ximage; + rb->PutMonoRow = put_mono_row_TRUEDITHER_ximage; + rb->PutValues = put_values_TRUEDITHER_ximage; + rb->PutMonoValues = put_mono_values_TRUEDITHER_ximage; + } + break; + case PF_8A8B8G8R: + if (pixmap) { + rb->PutRow = put_row_8A8B8G8R_pixmap; + rb->PutRowRGB = put_row_rgb_8A8B8G8R_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_8A8B8G8R_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; + } + else { + rb->PutRow = put_row_8A8B8G8R_ximage; + rb->PutRowRGB = put_row_rgb_8A8B8G8R_ximage; + rb->PutMonoRow = put_mono_row_8A8B8G8R_ximage; + rb->PutValues = put_values_8A8B8G8R_ximage; + rb->PutMonoValues = put_mono_values_8A8B8G8R_ximage; + rb->GetPointer = get_pointer_4_ximage; + } + break; + case PF_8A8R8G8B: + if (pixmap) { + rb->PutRow = put_row_8A8R8G8B_pixmap; + rb->PutRowRGB = put_row_rgb_8A8R8G8B_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_8A8R8G8B_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; + } + else { + rb->PutRow = put_row_8A8R8G8B_ximage; + rb->PutRowRGB = put_row_rgb_8A8R8G8B_ximage; + rb->PutMonoRow = put_mono_row_8A8R8G8B_ximage; + rb->PutValues = put_values_8A8R8G8B_ximage; + rb->PutMonoValues = put_mono_values_8A8R8G8B_ximage; + rb->GetPointer = get_pointer_4_ximage; + } + break; + case PF_8R8G8B: + if (pixmap) { + rb->PutRow = put_row_8R8G8B_pixmap; + rb->PutRowRGB = put_row_rgb_8R8G8B_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_8R8G8B_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; + } + else { + rb->PutRow = put_row_8R8G8B_ximage; + rb->PutRowRGB = put_row_rgb_8R8G8B_ximage; + rb->PutMonoRow = put_mono_row_8R8G8B_ximage; + rb->PutValues = put_values_8R8G8B_ximage; + rb->PutMonoValues = put_mono_values_8R8G8B_ximage; + } + break; + case PF_8R8G8B24: + if (pixmap) { + rb->PutRow = put_row_8R8G8B24_pixmap; + rb->PutRowRGB = put_row_rgb_8R8G8B24_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_8R8G8B24_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; + } + else { + rb->PutRow = put_row_8R8G8B24_ximage; + rb->PutRowRGB = put_row_rgb_8R8G8B24_ximage; + rb->PutMonoRow = put_mono_row_8R8G8B24_ximage; + rb->PutValues = put_values_8R8G8B24_ximage; + rb->PutMonoValues = put_mono_values_8R8G8B24_ximage; + } + break; + case PF_5R6G5B: + if (pixmap) { + rb->PutRow = put_row_5R6G5B_pixmap; + rb->PutRowRGB = put_row_rgb_5R6G5B_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_5R6G5B_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; + } + else { + rb->PutRow = put_row_5R6G5B_ximage; + rb->PutRowRGB = put_row_rgb_5R6G5B_ximage; + rb->PutMonoRow = put_mono_row_ximage; + rb->PutValues = put_values_5R6G5B_ximage; + rb->PutMonoValues = put_mono_values_ximage; + } + break; + case PF_Dither_5R6G5B: + if (pixmap) { + rb->PutRow = put_row_DITHER_5R6G5B_pixmap; + rb->PutRowRGB = put_row_rgb_DITHER_5R6G5B_pixmap; + rb->PutMonoRow = put_mono_row_TRUEDITHER_pixmap; + rb->PutValues = put_values_DITHER_5R6G5B_pixmap; + rb->PutMonoValues = put_mono_values_TRUEDITHER_pixmap; + } + else { + rb->PutRow = put_row_DITHER_5R6G5B_ximage; + rb->PutRowRGB = put_row_rgb_DITHER_5R6G5B_ximage; + rb->PutMonoRow = put_mono_row_DITHER_5R6G5B_ximage; + rb->PutValues = put_values_DITHER_5R6G5B_ximage; + rb->PutMonoValues = put_mono_values_DITHER_5R6G5B_ximage; + } + break; + case PF_Dither: + if (pixmap) { + rb->PutRow = put_row_DITHER_pixmap; + rb->PutRowRGB = put_row_rgb_DITHER_pixmap; + rb->PutMonoRow = put_mono_row_DITHER_pixmap; + rb->PutValues = put_values_DITHER_pixmap; + rb->PutMonoValues = put_mono_values_DITHER_pixmap; + } + else { + if (depth == 8) { + rb->PutRow = put_row_DITHER8_ximage; + rb->PutRowRGB = put_row_rgb_DITHER8_ximage; + rb->PutMonoRow = put_mono_row_DITHER8_ximage; + rb->PutValues = put_values_DITHER8_ximage; + rb->PutMonoValues = put_mono_values_DITHER8_ximage; + } + else { + rb->PutRow = put_row_DITHER_ximage; + rb->PutRowRGB = put_row_rgb_DITHER_ximage; + rb->PutMonoRow = put_mono_row_DITHER_ximage; + rb->PutValues = put_values_DITHER_ximage; + rb->PutMonoValues = put_mono_values_DITHER_ximage; + } + } + break; + case PF_1Bit: + if (pixmap) { + rb->PutRow = put_row_1BIT_pixmap; + rb->PutRowRGB = put_row_rgb_1BIT_pixmap; + rb->PutMonoRow = put_mono_row_1BIT_pixmap; + rb->PutValues = put_values_1BIT_pixmap; + rb->PutMonoValues = put_mono_values_1BIT_pixmap; + } + else { + rb->PutRow = put_row_1BIT_ximage; + rb->PutRowRGB = put_row_rgb_1BIT_ximage; + rb->PutMonoRow = put_mono_row_1BIT_ximage; + rb->PutValues = put_values_1BIT_ximage; + rb->PutMonoValues = put_mono_values_1BIT_ximage; + } + break; + case PF_HPCR: + if (pixmap) { + rb->PutRow = put_row_HPCR_pixmap; + rb->PutRowRGB = put_row_rgb_HPCR_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_HPCR_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; + } + else { + rb->PutRow = put_row_HPCR_ximage; + rb->PutRowRGB = put_row_rgb_HPCR_ximage; + rb->PutMonoRow = put_mono_row_HPCR_ximage; + rb->PutValues = put_values_HPCR_ximage; + rb->PutMonoValues = put_mono_values_HPCR_ximage; + } + break; + case PF_Lookup: + if (pixmap) { + rb->PutRow = put_row_LOOKUP_pixmap; + rb->PutRowRGB = put_row_rgb_LOOKUP_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_LOOKUP_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; + } + else { + if (depth==8) { + rb->PutRow = put_row_LOOKUP8_ximage; + rb->PutRowRGB = put_row_rgb_LOOKUP8_ximage; + rb->PutMonoRow = put_mono_row_LOOKUP8_ximage; + rb->PutValues = put_values_LOOKUP8_ximage; + rb->PutMonoValues = put_mono_values_LOOKUP8_ximage; + } + else { + rb->PutRow = put_row_LOOKUP_ximage; + rb->PutRowRGB = put_row_rgb_LOOKUP_ximage; + rb->PutMonoRow = put_mono_row_ximage; + rb->PutValues = put_values_LOOKUP_ximage; + rb->PutMonoValues = put_mono_values_ximage; + } + } + break; + case PF_Grayscale: + if (pixmap) { + rb->PutRow = put_row_GRAYSCALE_pixmap; + rb->PutRowRGB = put_row_rgb_GRAYSCALE_pixmap; + rb->PutMonoRow = put_mono_row_pixmap; + rb->PutValues = put_values_GRAYSCALE_pixmap; + rb->PutMonoValues = put_mono_values_pixmap; + } + else { + if (depth == 8) { + rb->PutRow = put_row_GRAYSCALE8_ximage; + rb->PutRowRGB = put_row_rgb_GRAYSCALE8_ximage; + rb->PutMonoRow = put_mono_row_GRAYSCALE8_ximage; + rb->PutValues = put_values_GRAYSCALE8_ximage; + rb->PutMonoValues = put_mono_values_GRAYSCALE8_ximage; + } + else { + rb->PutRow = put_row_GRAYSCALE_ximage; + rb->PutRowRGB = put_row_rgb_GRAYSCALE_ximage; + rb->PutMonoRow = put_mono_row_ximage; + rb->PutValues = put_values_GRAYSCALE_ximage; + rb->PutMonoValues = put_mono_values_ximage; + } + } + break; + default: + _mesa_problem(NULL, "Bad pixel format in xmesa_update_state (1)"); + return; + } + + + /* Get functions */ + if (pixelformat == PF_Index) { + rb->GetRow = get_row_ci; + rb->GetValues = get_values_ci; + } + else { + rb->GetRow = get_row_rgba; + rb->GetValues = get_values_rgba; + } +} + diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c new file mode 100644 index 0000000000..5533158ece --- /dev/null +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -0,0 +1,251 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * \file xm_surface.c + * Code to allow the softpipe code to write to X windows/buffers. + * This is a bit of a hack for now. We've basically got two different + * abstractions for color buffers: gl_renderbuffer and pipe_surface. + * They'll need to get merged someday... + * For now, they're separate things that point to each other. + */ + + +#include "glxheader.h" +#include "GL/xmesa.h" +#include "xmesaP.h" +#include "context.h" +#include "imports.h" +#include "macros.h" +#include "framebuffer.h" +#include "renderbuffer.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/softpipe/sp_context.h" +#include "pipe/softpipe/sp_clear.h" +#include "pipe/softpipe/sp_tile_cache.h" +#include "pipe/softpipe/sp_surface.h" +#include "state_tracker/st_context.h" + + +#define CLIP_TILE \ + do { \ + if (x + w > ps->width) \ + w = ps->width - x; \ + if (y + h > ps->height) \ + h = ps->height -y; \ + } while(0) + + +static INLINE struct xmesa_surface * +xmesa_surface(struct pipe_surface *ps) +{ + return (struct xmesa_surface *) ps; +} + + +static INLINE struct xmesa_renderbuffer * +xmesa_rb(struct pipe_surface *ps) +{ + struct xmesa_surface *xms = xmesa_surface(ps); + return xms->xrb; +} + + +#define FLIP(Y) Y = xrb->St.Base.Height - (Y) - 1; + + +void +xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, float *p) +{ + struct xmesa_surface *xms = xmesa_surface(ps); + struct xmesa_renderbuffer *xrb = xms->xrb; + + if (xrb) { + /* this is a front/back color buffer */ + GLubyte tmp[MAX_WIDTH * 4]; + GLuint i, j; + uint w0 = w; + GET_CURRENT_CONTEXT(ctx); + + CLIP_TILE; + + FLIP(y); + for (i = 0; i < h; i++) { + xrb->St.Base.GetRow(ctx, &xrb->St.Base, w, x, y - i, tmp); + for (j = 0; j < w * 4; j++) { + p[j] = UBYTE_TO_FLOAT(tmp[j]); + } + p += w0 * 4; + } + } + else { + /* other softpipe surface */ + softpipe_get_tile_rgba(pipe, ps, x, y, w, h, p); + } +} + + +void +xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const float *p) +{ + struct xmesa_surface *xms = xmesa_surface(ps); + struct xmesa_renderbuffer *xrb = xms->xrb; + + if (xrb) { + /* this is a front/back color buffer */ + GLubyte tmp[MAX_WIDTH * 4]; + GLuint i, j; + uint w0 = w; + GET_CURRENT_CONTEXT(ctx); + CLIP_TILE; + FLIP(y); + for (i = 0; i < h; i++) { + for (j = 0; j < w * 4; j++) { + UNCLAMPED_FLOAT_TO_UBYTE(tmp[j], p[j]); + } + xrb->St.Base.PutRow(ctx, &xrb->St.Base, w, x, y - i, tmp, NULL); + p += w0 * 4; + } +#if 0 /* debug: flush */ + { + XMesaContext xm = XMESA_CONTEXT(ctx); + XSync(xm->display, 0); + } +#endif + } + else { + /* other softpipe surface */ + softpipe_put_tile_rgba(pipe, ps, x, y, w, h, p); + } +} + + + +/** + * Called to create a pipe_surface for each X renderbuffer. + * Note: this is being used instead of pipe->surface_alloc() since we + * have special/unique quad read/write functions for X. + */ +struct pipe_surface * +xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) +{ + struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + + assert(pipeFormat); + + xms->surface.format = pipeFormat; + xms->surface.refcount = 1; + + /* Note, the region we allocate doesn't actually have any storage + * since we're drawing into an XImage or Pixmap. + * The region's size will get set in the xmesa_alloc_front/back_storage() + * functions. + */ + if (pipe) + xms->surface.region = pipe->winsys->region_alloc(pipe->winsys, + 1, 0, 0, 0x0); + + return &xms->surface; +} + + +/** + * Called via pipe->surface_alloc() to create new surfaces (textures, + * renderbuffers, etc. + */ +struct pipe_surface * +xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) +{ + struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + + assert(pipe); + assert(pipeFormat); + + xms->surface.format = pipeFormat; + xms->surface.refcount = 1; + + return &xms->surface; +} + + +boolean +xmesa_is_format_supported(struct pipe_context *pipe, uint format) +{ + switch( format ) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_S_R16_G16_B16_A16: + case PIPE_FORMAT_S8_Z24: + return TRUE; + }; + return FALSE; +} + + +/** + * Called via pipe->clear() + */ +void +xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) +{ + struct xmesa_renderbuffer *xrb = xmesa_rb(ps); + + /* XXX actually, we should just discard any cached tiles from this + * surface since we don't want to accidentally re-use them after clearing. + */ + pipe->flush(pipe, 0); + + { + struct softpipe_context *sp = softpipe_context(pipe); + if (ps == sp_tile_cache_get_surface(sp->cbuf_cache[0])) { + float clear[4]; + clear[0] = 0.2; /* XXX hack */ + clear[1] = 0.2; + clear[2] = 0.2; + clear[3] = 0.2; + sp_tile_cache_clear(sp->cbuf_cache[0], clear); + } + } + + if (xrb && xrb->ximage) { + /* clearing back color buffer */ + GET_CURRENT_CONTEXT(ctx); + xmesa_clear_buffers(ctx, BUFFER_BIT_BACK_LEFT); + } + else if (xrb && xrb->pixmap) { + /* clearing front color buffer */ + GET_CURRENT_CONTEXT(ctx); + xmesa_clear_buffers(ctx, BUFFER_BIT_FRONT_LEFT); + } + else { + /* clearing other buffer */ + softpipe_clear(pipe, ps, value); + } +} + diff --git a/src/mesa/pipe/xlib/xm_tri.c b/src/mesa/pipe/xlib/xm_tri.c new file mode 100644 index 0000000000..9f17083f90 --- /dev/null +++ b/src/mesa/pipe/xlib/xm_tri.c @@ -0,0 +1,1694 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5 + * + * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* + * This file contains "accelerated" triangle functions. It should be + * fairly easy to write new special-purpose triangle functions and hook + * them into this module. + */ + + +#include "glxheader.h" +#include "depth.h" +#include "macros.h" +#include "imports.h" +#include "mtypes.h" +#include "xmesaP.h" + +/* Internal swrast includes: + */ +#include "swrast/s_context.h" +#include "swrast/s_depth.h" +#include "swrast/s_triangle.h" + + +#define GET_XRB(XRB) struct xmesa_renderbuffer *XRB = \ + xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped) + + +/**********************************************************************/ +/*** Triangle rendering ***/ +/**********************************************************************/ + + +#if CHAN_BITS == 8 + +/* + * XImage, smooth, depth-buffered, PF_TRUECOLOR triangle. + */ +#define NAME smooth_TRUECOLOR_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); + +#define RENDER_SPAN( span ) \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + GLuint i; \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + unsigned long p; \ + PACK_TRUECOLOR(p, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + XMesaPutPixel(xrb->ximage, x, y, p); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } + +#include "swrast/s_tritemp.h" + + + + +/* + * XImage, smooth, depth-buffered, PF_8A8B8G8R triangle. + */ +#define NAME smooth_8A8B8G8R_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define INTERP_ALPHA 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = PACK_8A8B8G8R(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue), \ + FixedToInt(span.alpha)); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.alpha += span.alphaStep; \ + span.z += span.zStep; \ + } + +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, PF_8A8R8G8B triangle. + */ +#define NAME smooth_8A8R8G8B_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define INTERP_ALPHA 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); + +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = PACK_8A8R8G8B(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue), \ + FixedToInt(span.alpha)); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.alpha += span.alphaStep; \ + span.z += span.zStep; \ + } + +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, PF_8R8G8B triangle. + */ +#define NAME smooth_8R8G8B_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); + +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = PACK_8R8G8B(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } + +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, PF_8R8G8B24 triangle. + */ +#define NAME smooth_8R8G8B24_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) +#define PIXEL_TYPE bgr_t +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + PIXEL_TYPE *ptr = pRow + i; \ + ptr->r = FixedToInt(span.red); \ + ptr->g = FixedToInt(span.green); \ + ptr->b = FixedToInt(span.blue); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, PF_TRUEDITHER triangle. + */ +#define NAME smooth_TRUEDITHER_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + unsigned long p; \ + PACK_TRUEDITHER(p, x, y, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + XMesaPutPixel(xrb->ximage, x, y, p); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, PF_5R6G5B triangle. + */ +#define NAME smooth_5R6G5B_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = PACK_5R6G5B(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, PF_DITHER_5R6G5B triangle. + */ +#define NAME smooth_DITHER_5R6G5B_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + PACK_TRUEDITHER(pRow[i], x, y, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, 8-bit, PF_DITHER8 triangle. + */ +#define NAME smooth_DITHER8_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + XDITHER_SETUP(y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = (PIXEL_TYPE) XDITHER(x, FixedToInt(span.red),\ + FixedToInt(span.green), FixedToInt(span.blue) ); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, PF_DITHER triangle. + */ +#define NAME smooth_DITHER_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define SETUP_CODE \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + XDITHER_SETUP(y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + unsigned long p = XDITHER(x, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + XMesaPutPixel(img, x, y, p); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, 8-bit PF_LOOKUP triangle. + */ +#define NAME smooth_LOOKUP8_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + LOOKUP_SETUP; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = LOOKUP(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, depth-buffered, 8-bit PF_HPCR triangle. + */ +#define NAME smooth_HPCR_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = DITHER_HPCR(x, y, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue) ); \ + zRow[i] = z; \ + } \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, PF_TRUECOLOR triangle. + */ +#define NAME flat_TRUECOLOR_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; \ + unsigned long pixel; \ + PACK_TRUECOLOR(pixel, v2->color[0], v2->color[1], v2->color[2]); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + XMesaPutPixel(img, x, y, pixel); \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, PF_8A8B8G8R triangle. + */ +#define NAME flat_8A8B8G8R_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + GLuint p = PACK_8A8B8G8R( v2->color[0], v2->color[1],\ + v2->color[2], v2->color[3]); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = (PIXEL_TYPE) p; \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, PF_8A8R8G8B triangle. + */ +#define NAME flat_8A8R8G8B_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + GLuint p = PACK_8A8R8G8B(v2->color[0], v2->color[1], \ + v2->color[2], v2->color[3]); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = (PIXEL_TYPE) p; \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, PF_8R8G8B triangle. + */ +#define NAME flat_8R8G8B_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + GLuint p = PACK_8R8G8B( v2->color[0], v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = (PIXEL_TYPE) p; \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } + +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, PF_8R8G8B24 triangle. + */ +#define NAME flat_8R8G8B24_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) +#define PIXEL_TYPE bgr_t +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = v2->color; +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + PIXEL_TYPE *ptr = pRow + i; \ + ptr->r = color[RCOMP]; \ + ptr->g = color[GCOMP]; \ + ptr->b = color[BCOMP]; \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, PF_TRUEDITHER triangle. + */ +#define NAME flat_TRUEDITHER_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + unsigned long p; \ + PACK_TRUEDITHER(p, x, y, v2->color[0], \ + v2->color[1], v2->color[2]); \ + XMesaPutPixel(img, x, y, p); \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, PF_5R6G5B triangle. + */ +#define NAME flat_5R6G5B_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + GLushort p = PACK_5R6G5B( v2->color[0], v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = (PIXEL_TYPE) p; \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, PF_DITHER_5R6G5B triangle. + */ +#define NAME flat_DITHER_5R6G5B_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + const GLubyte *color = v2->color; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + PACK_TRUEDITHER(pRow[i], x, y, color[RCOMP], \ + color[GCOMP], color[BCOMP]); \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, 8-bit PF_DITHER triangle. + */ +#define NAME flat_DITHER8_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + FLAT_DITHER_SETUP( v2->color[0], v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + FLAT_DITHER_ROW_SETUP(YFLIP(xrb, y)); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = (PIXEL_TYPE) FLAT_DITHER(x); \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, PF_DITHER triangle. + */ +#define NAME flat_DITHER_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define SETUP_CODE \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; \ + FLAT_DITHER_SETUP( v2->color[0], v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + FLAT_DITHER_ROW_SETUP(y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + unsigned long p = FLAT_DITHER(x); \ + XMesaPutPixel(img, x, y, p); \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, 8-bit PF_HPCR triangle. + */ +#define NAME flat_HPCR_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + GLubyte r = v2->color[0]; \ + GLubyte g = v2->color[1]; \ + GLubyte b = v2->color[2]; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = (PIXEL_TYPE) DITHER_HPCR(x, y, r, g, b); \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, depth-buffered, 8-bit PF_LOOKUP triangle. + */ +#define NAME flat_LOOKUP8_z_triangle +#define INTERP_Z 1 +#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + LOOKUP_SETUP; \ + GLubyte r = v2->color[0]; \ + GLubyte g = v2->color[1]; \ + GLubyte b = v2->color[2]; \ + GLubyte p = LOOKUP(r,g,b); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + const DEPTH_TYPE z = FixedToDepth(span.z); \ + if (z < zRow[i]) { \ + pRow[i] = p; \ + zRow[i] = z; \ + } \ + span.z += span.zStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, PF_TRUECOLOR triangle. + */ +#define NAME smooth_TRUECOLOR_triangle +#define INTERP_RGB 1 +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + unsigned long p; \ + PACK_TRUECOLOR(p, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + XMesaPutPixel(img, x, y, p); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, PF_8A8B8G8R triangle. + */ +#define NAME smooth_8A8B8G8R_triangle +#define INTERP_RGB 1 +#define INTERP_ALPHA 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = PACK_8A8B8G8R(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue), \ + FixedToInt(span.alpha)); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.alpha += span.alphaStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, PF_8A8R8G8B triangle. + */ +#define NAME smooth_8A8R8G8B_triangle +#define INTERP_RGB 1 +#define INTERP_ALPHA 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = PACK_8A8R8G8B(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue), \ + FixedToInt(span.alpha)); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + span.alpha += span.alphaStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, PF_8R8G8B triangle. + */ +#define NAME smooth_8R8G8B_triangle +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = PACK_8R8G8B(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue) ); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, PF_8R8G8B triangle. + */ +#define NAME smooth_8R8G8B24_triangle +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) +#define PIXEL_TYPE bgr_t +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + PIXEL_TYPE *pixel = pRow; \ + for (i = 0; i < span.end; i++, pixel++) { \ + pixel->r = FixedToInt(span.red); \ + pixel->g = FixedToInt(span.green); \ + pixel->b = FixedToInt(span.blue); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, PF_TRUEDITHER triangle. + */ +#define NAME smooth_TRUEDITHER_triangle +#define INTERP_RGB 1 +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + unsigned long p; \ + PACK_TRUEDITHER(p, x, y, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + XMesaPutPixel(img, x, y, p ); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, PF_5R6G5B triangle. + */ +#define NAME smooth_5R6G5B_triangle +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = (PIXEL_TYPE) PACK_5R6G5B(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, PF_DITHER_5R6G5B triangle. + */ +#define NAME smooth_DITHER_5R6G5B_triangle +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + PACK_TRUEDITHER(pRow[i], x, y, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, 8-bit PF_DITHER triangle. + */ +#define NAME smooth_DITHER8_triangle +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + XDITHER_SETUP(y); \ + for (i = 0; i < span.end; i++, x++) { \ + pRow[i] = (PIXEL_TYPE) XDITHER(x, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue) ); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, PF_DITHER triangle. + */ +#define NAME smooth_DITHER_triangle +#define INTERP_RGB 1 +#define SETUP_CODE \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + XDITHER_SETUP(y); \ + for (i = 0; i < span.end; i++, x++) { \ + unsigned long p = XDITHER(x, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue) ); \ + XMesaPutPixel(img, x, y, p); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, 8-bit PF_LOOKUP triangle. + */ +#define NAME smooth_LOOKUP8_triangle +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + LOOKUP_SETUP; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = LOOKUP(FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, smooth, NON-depth-buffered, 8-bit PF_HPCR triangle. + */ +#define NAME smooth_HPCR_triangle +#define INTERP_RGB 1 +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + pRow[i] = DITHER_HPCR(x, y, FixedToInt(span.red), \ + FixedToInt(span.green), FixedToInt(span.blue)); \ + span.red += span.redStep; \ + span.green += span.greenStep; \ + span.blue += span.blueStep; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, PF_TRUECOLOR triangle. + */ +#define NAME flat_TRUECOLOR_triangle +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; \ + unsigned long pixel; \ + PACK_TRUECOLOR(pixel, v2->color[0], v2->color[1], v2->color[2]); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + XMesaPutPixel(img, x, y, pixel); \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, PF_8A8B8G8R triangle. + */ +#define NAME flat_8A8B8G8R_triangle +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + unsigned long p = PACK_8B8G8R( v2->color[0], \ + v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = (PIXEL_TYPE) p; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, PF_8A8R8G8B triangle. + */ +#define NAME flat_8A8R8G8B_triangle +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + unsigned long p = PACK_8R8G8B( v2->color[0], \ + v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = (PIXEL_TYPE) p; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, PF_8R8G8B triangle. + */ +#define NAME flat_8R8G8B_triangle +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) +#define PIXEL_TYPE GLuint +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + unsigned long p = PACK_8R8G8B( v2->color[0], \ + v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = (PIXEL_TYPE) p; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, PF_8R8G8B24 triangle. + */ +#define NAME flat_8R8G8B24_triangle +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) +#define PIXEL_TYPE bgr_t +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + const GLubyte *color = v2->color; +#define RENDER_SPAN( span ) \ + GLuint i; \ + PIXEL_TYPE *pixel = pRow; \ + for (i = 0; i < span.end; i++, pixel++) { \ + pixel->r = color[RCOMP]; \ + pixel->g = color[GCOMP]; \ + pixel->b = color[BCOMP]; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, PF_TRUEDITHER triangle. + */ +#define NAME flat_TRUEDITHER_triangle +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + unsigned long p; \ + PACK_TRUEDITHER(p, x, y, v2->color[0], \ + v2->color[1], v2->color[2] ); \ + XMesaPutPixel(img, x, y, p); \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, PF_5R6G5B triangle. + */ +#define NAME flat_5R6G5B_triangle +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + unsigned long p = PACK_5R6G5B( v2->color[0], \ + v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = (PIXEL_TYPE) p; \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, PF_DITHER_5R6G5B triangle. + */ +#define NAME flat_DITHER_5R6G5B_triangle +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) +#define PIXEL_TYPE GLushort +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + const GLubyte *color = v2->color; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + PACK_TRUEDITHER(pRow[i], x, y, color[RCOMP], \ + color[GCOMP], color[BCOMP]); \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, 8-bit PF_DITHER triangle. + */ +#define NAME flat_DITHER8_triangle +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + FLAT_DITHER_SETUP( v2->color[0], v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + FLAT_DITHER_ROW_SETUP(YFLIP(xrb, y)); \ + for (i = 0; i < span.end; i++, x++) { \ + pRow[i] = (PIXEL_TYPE) FLAT_DITHER(x); \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, PF_DITHER triangle. + */ +#define NAME flat_DITHER_triangle +#define SETUP_CODE \ + GET_XRB(xrb); \ + XMesaImage *img = xrb->ximage; \ + FLAT_DITHER_SETUP( v2->color[0], v2->color[1], v2->color[2] ); +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + FLAT_DITHER_ROW_SETUP(y); \ + for (i = 0; i < span.end; i++, x++) { \ + unsigned long p = FLAT_DITHER(x); \ + XMesaPutPixel(img, x, y, p ); \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, 8-bit PF_HPCR triangle. + */ +#define NAME flat_HPCR_triangle +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + XMesaContext xmesa = XMESA_CONTEXT(ctx); \ + GET_XRB(xrb); \ + GLubyte r = v2->color[0]; \ + GLubyte g = v2->color[1]; \ + GLubyte b = v2->color[2]; +#define RENDER_SPAN( span ) \ + GLuint i; \ + GLint x = span.x, y = YFLIP(xrb, span.y); \ + for (i = 0; i < span.end; i++, x++) { \ + pRow[i] = (PIXEL_TYPE) DITHER_HPCR(x, y, r, g, b); \ + } +#include "swrast/s_tritemp.h" + + + +/* + * XImage, flat, NON-depth-buffered, 8-bit PF_LOOKUP triangle. + */ +#define NAME flat_LOOKUP8_triangle +#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) +#define PIXEL_TYPE GLubyte +#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) +#define SETUP_CODE \ + GET_XRB(xrb); \ + LOOKUP_SETUP; \ + GLubyte r = v2->color[0]; \ + GLubyte g = v2->color[1]; \ + GLubyte b = v2->color[2]; \ + GLubyte p = LOOKUP(r,g,b); +#define RENDER_SPAN( span ) \ + GLuint i; \ + for (i = 0; i < span.end; i++) { \ + pRow[i] = (PIXEL_TYPE) p; \ + } +#include "swrast/s_tritemp.h" + + +#endif /* CHAN_BITS == 8 */ + + +#if defined(DEBUG) && CHAN_BITS == 8 +extern void _xmesa_print_triangle_func( swrast_tri_func triFunc ); +void _xmesa_print_triangle_func( swrast_tri_func triFunc ) +{ + _mesa_printf("XMesa tri func = "); + if (triFunc ==smooth_TRUECOLOR_z_triangle) + _mesa_printf("smooth_TRUECOLOR_z_triangle\n"); + else if (triFunc ==smooth_8A8B8G8R_z_triangle) + _mesa_printf("smooth_8A8B8G8R_z_triangle\n"); + else if (triFunc ==smooth_8A8R8G8B_z_triangle) + _mesa_printf("smooth_8A8R8G8B_z_triangle\n"); + else if (triFunc ==smooth_8R8G8B_z_triangle) + _mesa_printf("smooth_8R8G8B_z_triangle\n"); + else if (triFunc ==smooth_8R8G8B24_z_triangle) + _mesa_printf("smooth_8R8G8B24_z_triangle\n"); + else if (triFunc ==smooth_TRUEDITHER_z_triangle) + _mesa_printf("smooth_TRUEDITHER_z_triangle\n"); + else if (triFunc ==smooth_5R6G5B_z_triangle) + _mesa_printf("smooth_5R6G5B_z_triangle\n"); + else if (triFunc ==smooth_DITHER_5R6G5B_z_triangle) + _mesa_printf("smooth_DITHER_5R6G5B_z_triangle\n"); + else if (triFunc ==smooth_HPCR_z_triangle) + _mesa_printf("smooth_HPCR_z_triangle\n"); + else if (triFunc ==smooth_DITHER8_z_triangle) + _mesa_printf("smooth_DITHER8_z_triangle\n"); + else if (triFunc ==smooth_LOOKUP8_z_triangle) + _mesa_printf("smooth_LOOKUP8_z_triangle\n"); + else if (triFunc ==flat_TRUECOLOR_z_triangle) + _mesa_printf("flat_TRUECOLOR_z_triangle\n"); + else if (triFunc ==flat_8A8B8G8R_z_triangle) + _mesa_printf("flat_8A8B8G8R_z_triangle\n"); + else if (triFunc ==flat_8A8R8G8B_z_triangle) + _mesa_printf("flat_8A8R8G8B_z_triangle\n"); + else if (triFunc ==flat_8R8G8B_z_triangle) + _mesa_printf("flat_8R8G8B_z_triangle\n"); + else if (triFunc ==flat_8R8G8B24_z_triangle) + _mesa_printf("flat_8R8G8B24_z_triangle\n"); + else if (triFunc ==flat_TRUEDITHER_z_triangle) + _mesa_printf("flat_TRUEDITHER_z_triangle\n"); + else if (triFunc ==flat_5R6G5B_z_triangle) + _mesa_printf("flat_5R6G5B_z_triangle\n"); + else if (triFunc ==flat_DITHER_5R6G5B_z_triangle) + _mesa_printf("flat_DITHER_5R6G5B_z_triangle\n"); + else if (triFunc ==flat_HPCR_z_triangle) + _mesa_printf("flat_HPCR_z_triangle\n"); + else if (triFunc ==flat_DITHER8_z_triangle) + _mesa_printf("flat_DITHER8_z_triangle\n"); + else if (triFunc ==flat_LOOKUP8_z_triangle) + _mesa_printf("flat_LOOKUP8_z_triangle\n"); + else if (triFunc ==smooth_TRUECOLOR_triangle) + _mesa_printf("smooth_TRUECOLOR_triangle\n"); + else if (triFunc ==smooth_8A8B8G8R_triangle) + _mesa_printf("smooth_8A8B8G8R_triangle\n"); + else if (triFunc ==smooth_8A8R8G8B_triangle) + _mesa_printf("smooth_8A8R8G8B_triangle\n"); + else if (triFunc ==smooth_8R8G8B_triangle) + _mesa_printf("smooth_8R8G8B_triangle\n"); + else if (triFunc ==smooth_8R8G8B24_triangle) + _mesa_printf("smooth_8R8G8B24_triangle\n"); + else if (triFunc ==smooth_TRUEDITHER_triangle) + _mesa_printf("smooth_TRUEDITHER_triangle\n"); + else if (triFunc ==smooth_5R6G5B_triangle) + _mesa_printf("smooth_5R6G5B_triangle\n"); + else if (triFunc ==smooth_DITHER_5R6G5B_triangle) + _mesa_printf("smooth_DITHER_5R6G5B_triangle\n"); + else if (triFunc ==smooth_HPCR_triangle) + _mesa_printf("smooth_HPCR_triangle\n"); + else if (triFunc ==smooth_DITHER8_triangle) + _mesa_printf("smooth_DITHER8_triangle\n"); + else if (triFunc ==smooth_LOOKUP8_triangle) + _mesa_printf("smooth_LOOKUP8_triangle\n"); + else if (triFunc ==flat_TRUECOLOR_triangle) + _mesa_printf("flat_TRUECOLOR_triangle\n"); + else if (triFunc ==flat_TRUEDITHER_triangle) + _mesa_printf("flat_TRUEDITHER_triangle\n"); + else if (triFunc ==flat_8A8B8G8R_triangle) + _mesa_printf("flat_8A8B8G8R_triangle\n"); + else if (triFunc ==flat_8A8R8G8B_triangle) + _mesa_printf("flat_8A8R8G8B_triangle\n"); + else if (triFunc ==flat_8R8G8B_triangle) + _mesa_printf("flat_8R8G8B_triangle\n"); + else if (triFunc ==flat_8R8G8B24_triangle) + _mesa_printf("flat_8R8G8B24_triangle\n"); + else if (triFunc ==flat_5R6G5B_triangle) + _mesa_printf("flat_5R6G5B_triangle\n"); + else if (triFunc ==flat_DITHER_5R6G5B_triangle) + _mesa_printf("flat_DITHER_5R6G5B_triangle\n"); + else if (triFunc ==flat_HPCR_triangle) + _mesa_printf("flat_HPCR_triangle\n"); + else if (triFunc ==flat_DITHER8_triangle) + _mesa_printf("flat_DITHER8_triangle\n"); + else if (triFunc ==flat_LOOKUP8_triangle) + _mesa_printf("flat_LOOKUP8_triangle\n"); + else + _mesa_printf("???\n"); +} +#endif + + +#ifdef DEBUG + +/* record the current triangle function name */ +static const char *triFuncName = NULL; + +#define USE(triFunc) \ +do { \ + triFuncName = #triFunc; \ + return triFunc; \ +} while (0) + +#else + +#define USE(triFunc) return triFunc + +#endif + + +#if 0 +GLboolean xmesa_get_cbuf_details( GLcontext *ctx, + void **ptr, + GLuint *cpp, + GLint *stride, + GLuint *format ) +{ + XMesaContext xmesa = XMESA_CONTEXT(ctx); + struct gl_framebuffer *fb = ctx->DrawBuffer; + struct gl_renderbuffer *crb = fb->_ColorDrawBuffers[0][0]; + struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(crb->Wrapped); + + *ptr = crb->GetPointer(ctx, crb, 0, 0); + *stride = ((GLubyte *)crb->GetPointer(ctx, crb, 0, 1) - + (GLubyte *)crb->GetPointer(ctx, crb, 0, 0)); + + if (!ptr) + goto bad; + + switch (xmesa->pixelformat) { + case PF_8A8B8G8R: + case PF_8A8R8G8B: + *format = 1; /* whatever */ + *cpp = 4; + break; + default: + goto bad; + } + + return GL_TRUE; + + bad: + *ptr = NULL; + *stride = 0; + *format = 0; + return GL_FALSE; +} +#endif + + +/** + * Return pointer to line drawing function, or NULL if we should use a + * swrast fallback. + */ +static swrast_tri_func +get_triangle_func(GLcontext *ctx) +{ +#if CHAN_BITS == 8 + SWcontext *swrast = SWRAST_CONTEXT(ctx); + XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + const int depth = GET_VISUAL_DEPTH(xmesa->xm_visual); + const struct xmesa_renderbuffer *xrb; + +#ifdef DEBUG + triFuncName = NULL; +#endif + + /* trivial fallback tests */ + if ((ctx->DrawBuffer->_ColorDrawBufferMask[0] + & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)) == 0) + return (swrast_tri_func) NULL; + if (ctx->RenderMode != GL_RENDER) + return (swrast_tri_func) NULL; + if (ctx->Polygon.SmoothFlag) + return (swrast_tri_func) NULL; + if (ctx->Texture._EnabledUnits) + return (swrast_tri_func) NULL; + if (swrast->_RasterMask & MULTI_DRAW_BIT) + return (swrast_tri_func) NULL; + if (ctx->Polygon.CullFlag && + ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) + return (swrast_tri_func) NULL; + if (xmbuf->swAlpha) + return (swrast_tri_func) NULL; + + xrb = xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped); + + if (xrb->ximage) { + if ( ctx->Light.ShadeModel==GL_SMOOTH + && swrast->_RasterMask==DEPTH_BIT + && ctx->Depth.Func==GL_LESS + && ctx->Depth.Mask==GL_TRUE + && ctx->Visual.depthBits == DEFAULT_SOFTWARE_DEPTH_BITS + && ctx->Polygon.StippleFlag==GL_FALSE) { + switch (xmesa->pixelformat) { + case PF_Truecolor: + USE(smooth_TRUECOLOR_z_triangle); + case PF_8A8B8G8R: + USE(smooth_8A8B8G8R_z_triangle); + case PF_8A8R8G8B: + USE(smooth_8A8R8G8B_z_triangle); + case PF_8R8G8B: + USE(smooth_8R8G8B_z_triangle); + case PF_8R8G8B24: + USE(smooth_8R8G8B24_z_triangle); + case PF_Dither_True: + USE(smooth_TRUEDITHER_z_triangle); + case PF_5R6G5B: + USE(smooth_5R6G5B_z_triangle); + case PF_Dither_5R6G5B: + USE(smooth_DITHER_5R6G5B_z_triangle); + case PF_HPCR: + USE(smooth_HPCR_z_triangle); + case PF_Dither: + if (depth == 8) + USE(smooth_DITHER8_z_triangle); + else + USE(smooth_DITHER_z_triangle); + case PF_Lookup: + if (depth == 8) + USE(smooth_LOOKUP8_z_triangle); + else + return (swrast_tri_func) NULL; + default: + return (swrast_tri_func) NULL; + } + } + if ( ctx->Light.ShadeModel==GL_FLAT + && swrast->_RasterMask==DEPTH_BIT + && ctx->Depth.Func==GL_LESS + && ctx->Depth.Mask==GL_TRUE + && ctx->Visual.depthBits == DEFAULT_SOFTWARE_DEPTH_BITS + && ctx->Polygon.StippleFlag==GL_FALSE) { + switch (xmesa->pixelformat) { + case PF_Truecolor: + USE(flat_TRUECOLOR_z_triangle); + case PF_8A8B8G8R: + USE(flat_8A8B8G8R_z_triangle); + case PF_8A8R8G8B: + USE(flat_8A8R8G8B_z_triangle); + case PF_8R8G8B: + USE(flat_8R8G8B_z_triangle); + case PF_8R8G8B24: + USE(flat_8R8G8B24_z_triangle); + case PF_Dither_True: + USE(flat_TRUEDITHER_z_triangle); + case PF_5R6G5B: + USE(flat_5R6G5B_z_triangle); + case PF_Dither_5R6G5B: + USE(flat_DITHER_5R6G5B_z_triangle); + case PF_HPCR: + USE(flat_HPCR_z_triangle); + case PF_Dither: + if (depth == 8) + USE(flat_DITHER8_z_triangle); + else + USE(flat_DITHER_z_triangle); + case PF_Lookup: + if (depth == 8) + USE(flat_LOOKUP8_z_triangle); + else + return (swrast_tri_func) NULL; + default: + return (swrast_tri_func) NULL; + } + } + if ( swrast->_RasterMask==0 /* no depth test */ + && ctx->Light.ShadeModel==GL_SMOOTH + && ctx->Polygon.StippleFlag==GL_FALSE) { + switch (xmesa->pixelformat) { + case PF_Truecolor: + USE(smooth_TRUECOLOR_triangle); + case PF_8A8B8G8R: + USE(smooth_8A8B8G8R_triangle); + case PF_8A8R8G8B: + USE(smooth_8A8R8G8B_triangle); + case PF_8R8G8B: + USE(smooth_8R8G8B_triangle); + case PF_8R8G8B24: + USE(smooth_8R8G8B24_triangle); + case PF_Dither_True: + USE(smooth_TRUEDITHER_triangle); + case PF_5R6G5B: + USE(smooth_5R6G5B_triangle); + case PF_Dither_5R6G5B: + USE(smooth_DITHER_5R6G5B_triangle); + case PF_HPCR: + USE(smooth_HPCR_triangle); + case PF_Dither: + if (depth == 8) + USE(smooth_DITHER8_triangle); + else + USE(smooth_DITHER_triangle); + case PF_Lookup: + if (depth == 8) + USE(smooth_LOOKUP8_triangle); + else + return (swrast_tri_func) NULL; + default: + return (swrast_tri_func) NULL; + } + } + + if ( swrast->_RasterMask==0 /* no depth test */ + && ctx->Light.ShadeModel==GL_FLAT + && ctx->Polygon.StippleFlag==GL_FALSE) { + switch (xmesa->pixelformat) { + case PF_Truecolor: + USE(flat_TRUECOLOR_triangle); + case PF_Dither_True: + USE(flat_TRUEDITHER_triangle); + case PF_8A8B8G8R: + USE(flat_8A8B8G8R_triangle); + case PF_8A8R8G8B: + USE(flat_8A8R8G8B_triangle); + case PF_8R8G8B: + USE(flat_8R8G8B_triangle); + case PF_8R8G8B24: + USE(flat_8R8G8B24_triangle); + case PF_5R6G5B: + USE(flat_5R6G5B_triangle); + case PF_Dither_5R6G5B: + USE(flat_DITHER_5R6G5B_triangle); + case PF_HPCR: + USE(flat_HPCR_triangle); + case PF_Dither: + if (depth == 8) + USE(flat_DITHER8_triangle); + else + USE(flat_DITHER_triangle); + case PF_Lookup: + if (depth == 8) + USE(flat_LOOKUP8_triangle); + else + return (swrast_tri_func) NULL; + default: + return (swrast_tri_func) NULL; + } + } + } +#endif /* CHAN_BITS == 8 */ + + return (swrast_tri_func) NULL; +} + + +/* Override for the swrast tri-selection function. Try to use one + * of our internal tri functions, otherwise fall back to the + * standard swrast functions. + */ +void xmesa_choose_triangle( GLcontext *ctx ) +{ + SWcontext *swrast = SWRAST_CONTEXT(ctx); + + if (!(swrast->Triangle = get_triangle_func( ctx ))) + _swrast_choose_triangle( ctx ); +} + diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c new file mode 100644 index 0000000000..36805437f0 --- /dev/null +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -0,0 +1,368 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "glxheader.h" +#include "xmesaP.h" +#include "main/macros.h" + +#include "pipe/p_winsys.h" +#include "pipe/softpipe/sp_winsys.h" + + +/** + * XMesa winsys, derived from softpipe winsys. + * NOTE: there's nothing really X-specific in this winsys layer so + * we could probably lift it up somewhere. + */ +struct xm_winsys +{ + struct softpipe_winsys sws; + int foo; /* placeholder */ +}; + + +/** + * Low-level OS/window system memory buffer + */ +struct xm_buffer +{ + boolean userBuffer; /** Is this a user-space buffer? */ + int refcount; + unsigned size; + void *data; + void *mapped; +}; + + + +/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static inline struct xm_buffer * +xm_bo( struct pipe_buffer_handle *bo ) +{ + return (struct xm_buffer *) bo; +} + +static inline struct pipe_buffer_handle * +pipe_bo( struct xm_buffer *bo ) +{ + return (struct pipe_buffer_handle *) bo; +} + +/* Turn a softpipe winsys into an xm/softpipe winsys: + */ +static inline struct xm_winsys * +xm_winsys(struct softpipe_winsys *sws) +{ + return (struct xm_winsys *) sws; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void * +xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned flags) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + xm_buf->mapped = xm_buf->data; + return xm_buf->mapped; +} + +static void +xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + xm_buf->mapped = NULL; +} + +static void +xm_buffer_reference(struct pipe_winsys *pws, + struct pipe_buffer_handle **ptr, + struct pipe_buffer_handle *buf) +{ + if (*ptr) { + struct xm_buffer *oldBuf = xm_bo(*ptr); + oldBuf->refcount--; + assert(oldBuf->refcount >= 0); + if (oldBuf->refcount == 0) { + if (oldBuf->data) { + if (!oldBuf->userBuffer) + free(oldBuf->data); + oldBuf->data = NULL; + } + free(oldBuf); + } + *ptr = NULL; + } + + assert(!(*ptr)); + + if (buf) { + struct xm_buffer *newBuf = xm_bo(buf); + newBuf->refcount++; + *ptr = buf; + } +} + +static void +xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned size, const void *data ) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + assert(!xm_buf->userBuffer); + if (xm_buf->size != size) { + if (xm_buf->data) + free(xm_buf->data); + xm_buf->data = malloc(size); + xm_buf->size = size; + } + if (data) + memcpy(xm_buf->data, data, size); +} + +static void +xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned long offset, unsigned long size, const void *data) +{ + struct xm_buffer *xm_buf = xm_bo(buf); + GLubyte *b = (GLubyte *) xm_buf->data; + assert(!xm_buf->userBuffer); + assert(b); + memcpy(b + offset, data, size); +} + +static void +xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, + unsigned long offset, unsigned long size, void *data) +{ + const struct xm_buffer *xm_buf = xm_bo(buf); + const GLubyte *b = (GLubyte *) xm_buf->data; + assert(!xm_buf->userBuffer); + assert(b); + memcpy(data, b + offset, size); +} + +static void +xm_flush_frontbuffer(struct pipe_winsys *pws) +{ + /* + struct intel_context *intel = intel_pipe_winsys(sws)->intel; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelCopyBuffer(dPriv, NULL); + */ +} + +static void +xm_wait_idle(struct pipe_winsys *pws) +{ + /* no-op */ +} + +static void +xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) +{ + va_list args; + va_start( args, fmtString ); + vfprintf(stderr, fmtString, args); + va_end( args ); +} + +static const char * +xm_get_name(struct pipe_winsys *pws) +{ + return "Xlib"; +} + + +static struct pipe_buffer_handle * +xm_buffer_create(struct pipe_winsys *pws, unsigned alignment) +{ + struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); + buffer->refcount = 1; + return pipe_bo(buffer); +} + + +/** + * Create buffer which wraps user-space data. + */ +static struct pipe_buffer_handle * +xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) +{ + struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); + buffer->userBuffer = TRUE; + buffer->refcount = 1; + buffer->data = ptr; + buffer->size = bytes; + return pipe_bo(buffer); +} + + + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + + +static struct pipe_region * +xm_region_alloc(struct pipe_winsys *winsys, + unsigned cpp, unsigned width, unsigned height, unsigned flags) +{ + struct pipe_region *region = CALLOC_STRUCT(pipe_region); + const unsigned alignment = 64; + + region->cpp = cpp; + region->pitch = round_up(width, alignment / cpp); + region->height = height; + region->refcount = 1; + + assert(region->pitch > 0); + + region->buffer = winsys->buffer_create( winsys, alignment ) +; + + /* NULL data --> just allocate the space */ + winsys->buffer_data( winsys, + region->buffer, + region->pitch * cpp * height, + NULL ); + return region; +} + + +static void +xm_region_release(struct pipe_winsys *winsys, struct pipe_region **region) +{ + if (!*region) + return; + + assert((*region)->refcount > 0); + (*region)->refcount--; + + if ((*region)->refcount == 0) { + assert((*region)->map_refcount == 0); + + winsys->buffer_reference( winsys, &((*region)->buffer), NULL ); + free(*region); + } + *region = NULL; +} + + +/** + * Called via pipe->surface_alloc() to create new surfaces (textures, + * renderbuffers, etc. + */ +static struct pipe_surface * +xm_surface_alloc(struct pipe_winsys *ws, GLuint pipeFormat) +{ + struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + + assert(ws); + assert(pipeFormat); + + xms->surface.format = pipeFormat; + xms->surface.refcount = 1; +#if 0 + /* + * This is really just a softpipe surface, not an XImage/Pixmap surface. + */ + softpipe_init_surface_funcs(&xms->surface); +#endif + return &xms->surface; +} + + + + +struct xmesa_pipe_winsys +{ + struct pipe_winsys winsys; + XMesaContext xmesa; +}; + +static struct pipe_winsys * +xmesa_create_pipe_winsys( XMesaContext xmesa ) +{ + struct xmesa_pipe_winsys *xws = CALLOC_STRUCT(xmesa_pipe_winsys); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + * + * Pipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + xws->winsys.buffer_create = xm_buffer_create; + xws->winsys.user_buffer_create = xm_user_buffer_create; + xws->winsys.buffer_map = xm_buffer_map; + xws->winsys.buffer_unmap = xm_buffer_unmap; + xws->winsys.buffer_reference = xm_buffer_reference; + xws->winsys.buffer_data = xm_buffer_data; + xws->winsys.buffer_subdata = xm_buffer_subdata; + xws->winsys.buffer_get_subdata = xm_buffer_get_subdata; + + xws->winsys.region_alloc = xm_region_alloc; + xws->winsys.region_release = xm_region_release; + + xws->winsys.surface_alloc = xm_surface_alloc; + + xws->winsys.flush_frontbuffer = xm_flush_frontbuffer; + xws->winsys.wait_idle = xm_wait_idle; + xws->winsys.printf = xm_printf; + xws->winsys.get_name = xm_get_name; + xws->xmesa = xmesa; + + return &xws->winsys; +} + + +struct pipe_context * +xmesa_create_softpipe(XMesaContext xmesa) +{ + struct xm_winsys *xm_ws = CALLOC_STRUCT( xm_winsys ); + + /* Create the softpipe context: + */ + return softpipe_create( xmesa_create_pipe_winsys(xmesa), &xm_ws->sws ); +} diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h new file mode 100644 index 0000000000..5796fddb16 --- /dev/null +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -0,0 +1,631 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef XMESAP_H +#define XMESAP_H + + +#include "GL/xmesa.h" +#include "mtypes.h" +#if defined(FX) +#include "GL/fxmesa.h" +#endif +#ifdef XFree86Server +#include "xm_image.h" +#endif +#include "state_tracker/st_cb_fbo.h" +#include "pipe/softpipe/sp_context.h" +#include "pipe/softpipe/sp_surface.h" + + +extern _glthread_Mutex _xmesa_lock; + +extern XMesaBuffer XMesaBufferList; + +/* for PF_8R8G8B24 pixel format */ +typedef struct { + GLubyte b; + GLubyte g; + GLubyte r; +} bgr_t; + + +struct xmesa_renderbuffer; + + +/* Function pointer for clearing color buffers */ +typedef void (*ClearFunc)( GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLint x, GLint y, GLint width, GLint height ); + + + + +/** Framebuffer pixel formats */ +enum pixel_format { + PF_Index, /**< Color Index mode */ + PF_Truecolor, /**< TrueColor or DirectColor, any depth */ + PF_Dither_True, /**< TrueColor with dithering */ + PF_8A8R8G8B, /**< 32-bit TrueColor: 8-A, 8-R, 8-G, 8-B bits */ + PF_8A8B8G8R, /**< 32-bit TrueColor: 8-A, 8-B, 8-G, 8-R bits */ + PF_8R8G8B, /**< 32-bit TrueColor: 8-R, 8-G, 8-B bits */ + PF_8R8G8B24, /**< 24-bit TrueColor: 8-R, 8-G, 8-B bits */ + PF_5R6G5B, /**< 16-bit TrueColor: 5-R, 6-G, 5-B bits */ + PF_Dither, /**< Color-mapped RGB with dither */ + PF_Lookup, /**< Color-mapped RGB without dither */ + PF_HPCR, /**< HP Color Recovery (ad@lms.be 30/08/95) */ + PF_1Bit, /**< monochrome dithering of RGB */ + PF_Grayscale, /**< Grayscale or StaticGray */ + PF_Dither_5R6G5B /**< 16-bit dithered TrueColor: 5-R, 6-G, 5-B */ +}; + + +/** + * Visual inforation, derived from GLvisual. + * Basically corresponds to an XVisualInfo. + */ +struct xmesa_visual { + GLvisual mesa_visual; /* Device independent visual parameters */ + XMesaDisplay *display; /* The X11 display */ +#ifdef XFree86Server + GLint ColormapEntries; + GLint nplanes; +#else + XMesaVisualInfo visinfo; /* X's visual info (pointer to private copy) */ + XVisualInfo *vishandle; /* Only used in fakeglx.c */ +#endif + GLint BitsPerPixel; /* True bits per pixel for XImages */ + + GLboolean ximage_flag; /* Use XImage for back buffer (not pixmap)? */ + + enum pixel_format dithered_pf; /* Pixel format when dithering */ + enum pixel_format undithered_pf;/* Pixel format when not dithering */ + + GLfloat RedGamma; /* Gamma values, 1.0 is default */ + GLfloat GreenGamma; + GLfloat BlueGamma; + + /* For PF_TRUECOLOR */ + GLint rshift, gshift, bshift;/* Pixel color component shifts */ + GLubyte Kernel[16]; /* Dither kernel */ + unsigned long RtoPixel[512]; /* RGB to pixel conversion */ + unsigned long GtoPixel[512]; + unsigned long BtoPixel[512]; + GLubyte PixelToR[256]; /* Pixel to RGB conversion */ + GLubyte PixelToG[256]; + GLubyte PixelToB[256]; + + /* For PF_HPCR */ + short hpcr_rgbTbl[3][256]; + GLboolean hpcr_clear_flag; + GLubyte hpcr_clear_ximage_pattern[2][16]; + XMesaImage *hpcr_clear_ximage; + XMesaPixmap hpcr_clear_pixmap; + + /* For PF_1BIT */ + int bitFlip; +}; + + +/** + * Context info, dDerived from GLcontext. + * Basically corresponds to a GLXContext. + */ +struct xmesa_context { + GLcontext mesa; /* the core library context (containment) */ + XMesaVisual xm_visual; /* Describes the buffers */ + XMesaBuffer xm_buffer; /* current span/point/line/triangle buffer */ + + XMesaDisplay *display; /* == xm_visual->display */ + GLboolean swapbytes; /* Host byte order != display byte order? */ + GLboolean direct; /* Direct rendering context? */ + + enum pixel_format pixelformat; + + GLubyte clearcolor[4]; /* current clearing color */ + unsigned long clearpixel; /* current clearing pixel value */ +}; + + +/** + * Types of X/GLX drawables we might render into. + */ +typedef enum { + WINDOW, /* An X window */ + GLXWINDOW, /* GLX window */ + PIXMAP, /* GLX pixmap */ + PBUFFER /* GLX Pbuffer */ +} BufferType; + + +/** Values for db_mode: */ +/*@{*/ +#define BACK_PIXMAP 1 +#define BACK_XIMAGE 2 +/*@}*/ + + +/** + * An xmesa_renderbuffer represents the back or front color buffer. + * For the front color buffer: + * is the X window + * For the back color buffer: + * Either or will be used, never both. + * In any case, always equals . + * For stand-alone Mesa, we could merge and into one + * field. We don't do that for the server-side GLcore module because + * pixmaps and drawables are different and we'd need a bunch of casts. + */ +struct xmesa_renderbuffer +{ +#if 0 + struct gl_renderbuffer Base; /* Base class */ +#else + struct st_renderbuffer St; /**< Base class */ +#endif + + XMesaBuffer Parent; /**< The XMesaBuffer this renderbuffer belongs to */ + XMesaDrawable drawable; /* Usually the X window ID */ + XMesaPixmap pixmap; /* Back color buffer */ + XMesaImage *ximage; /* The back buffer, if not using a Pixmap */ + + GLubyte *origin1; /* used for PIXEL_ADDR1 macro */ + GLint width1; + GLushort *origin2; /* used for PIXEL_ADDR2 macro */ + GLint width2; + GLubyte *origin3; /* used for PIXEL_ADDR3 macro */ + GLint width3; + GLuint *origin4; /* used for PIXEL_ADDR4 macro */ + GLint width4; + + GLint bottom; /* used for FLIP macro, equals height - 1 */ + + ClearFunc clearFunc; + + void *pSurface; /** pipe surface */ +}; + + +/** + * Framebuffer information, derived from. + * Basically corresponds to a GLXDrawable. + */ +struct xmesa_buffer { + GLframebuffer mesa_buffer; /* depth, stencil, accum, etc buffers */ + /* This MUST BE FIRST! */ + GLboolean wasCurrent; /* was ever the current buffer? */ + XMesaVisual xm_visual; /* the X/Mesa visual */ + + XMesaDisplay *display; + BufferType type; /* window, pixmap, pbuffer or glxwindow */ + + struct xmesa_renderbuffer *frontxrb; /* front color renderbuffer */ + struct xmesa_renderbuffer *backxrb; /* back color renderbuffer */ + + XMesaColormap cmap; /* the X colormap */ + + unsigned long selectedEvents;/* for pbuffers only */ + + GLint db_mode; /* 0 = single buffered */ + /* BACK_PIXMAP = use Pixmap for back buffer */ + /* BACK_XIMAGE = use XImage for back buffer */ + GLboolean swAlpha; + + GLuint shm; /* X Shared Memory extension status: */ + /* 0 = not available */ + /* 1 = XImage support available */ + /* 2 = Pixmap support available too */ +#if defined(USE_XSHM) && !defined(XFree86Server) + XShmSegmentInfo shminfo; +#endif + + XMesaImage *rowimage; /* Used for optimized span writing */ + XMesaPixmap stipple_pixmap; /* For polygon stippling */ + XMesaGC stipple_gc; /* For polygon stippling */ + + XMesaGC gc; /* scratch GC for span, line, tri drawing */ + XMesaGC cleargc; /* GC for clearing the color buffer */ + XMesaGC swapgc; /* GC for swapping the color buffers */ + + /* The following are here instead of in the XMesaVisual + * because they depend on the window's colormap. + */ + + /* For PF_DITHER, PF_LOOKUP, PF_GRAYSCALE */ + unsigned long color_table[576]; /* RGB -> pixel value */ + + /* For PF_DITHER, PF_LOOKUP, PF_GRAYSCALE */ + GLubyte pixel_to_r[65536]; /* pixel value -> red */ + GLubyte pixel_to_g[65536]; /* pixel value -> green */ + GLubyte pixel_to_b[65536]; /* pixel value -> blue */ + + /* Used to do XAllocColor/XFreeColors accounting: */ + int num_alloced; +#if defined(XFree86Server) + Pixel alloced_colors[256]; +#else + unsigned long alloced_colors[256]; +#endif + +#if defined( FX ) + /* For 3Dfx Glide only */ + GLboolean FXisHackUsable; /* Can we render into window? */ + GLboolean FXwindowHack; /* Are we rendering into a window? */ + fxMesaContext FXctx; +#endif + + /* GLX_EXT_texture_from_pixmap */ + GLint TextureTarget; /** GLX_TEXTURE_1D_EXT, for example */ + GLint TextureFormat; /** GLX_TEXTURE_FORMAT_RGB_EXT, for example */ + GLint TextureMipmap; /** 0 or 1 */ + + struct xmesa_buffer *Next; /* Linked list pointer: */ +}; + + +/** + * If pixelformat==PF_TRUECOLOR: + */ +#define PACK_TRUECOLOR( PIXEL, R, G, B ) \ + PIXEL = xmesa->xm_visual->RtoPixel[R] \ + | xmesa->xm_visual->GtoPixel[G] \ + | xmesa->xm_visual->BtoPixel[B]; \ + + +/** + * If pixelformat==PF_TRUEDITHER: + */ +#define PACK_TRUEDITHER( PIXEL, X, Y, R, G, B ) \ +{ \ + int d = xmesa->xm_visual->Kernel[((X)&3) | (((Y)&3)<<2)]; \ + PIXEL = xmesa->xm_visual->RtoPixel[(R)+d] \ + | xmesa->xm_visual->GtoPixel[(G)+d] \ + | xmesa->xm_visual->BtoPixel[(B)+d]; \ +} + + + +/** + * If pixelformat==PF_8A8B8G8R: + */ +#define PACK_8A8B8G8R( R, G, B, A ) \ + ( ((A) << 24) | ((B) << 16) | ((G) << 8) | (R) ) + + +/** + * Like PACK_8A8B8G8R() but don't use alpha. This is usually an acceptable + * shortcut. + */ +#define PACK_8B8G8R( R, G, B ) ( ((B) << 16) | ((G) << 8) | (R) ) + + + +/** + * If pixelformat==PF_8R8G8B: + */ +#define PACK_8R8G8B( R, G, B) ( ((R) << 16) | ((G) << 8) | (B) ) + + +/** + * If pixelformat==PF_5R6G5B: + */ +#define PACK_5R6G5B( R, G, B) ( (((R) & 0xf8) << 8) | (((G) & 0xfc) << 3) | ((B) >> 3) ) + + +/** + * If pixelformat==PF_8A8R8G8B: + */ +#define PACK_8A8R8G8B( R, G, B, A ) \ + ( ((A) << 24) | ((R) << 16) | ((G) << 8) | (B) ) + + + +/** + * If pixelformat==PF_DITHER: + * + * Improved 8-bit RGB dithering code contributed by Bob Mercier + * (mercier@hollywood.cinenet.net). Thanks Bob! + */ +#ifdef DITHER666 +# define DITH_R 6 +# define DITH_G 6 +# define DITH_B 6 +# define DITH_MIX(r,g,b) (((r) * DITH_G + (g)) * DITH_B + (b)) +#else +# define DITH_R 5 +# define DITH_G 9 +# define DITH_B 5 +# define DITH_MIX(r,g,b) (((g) << 6) | ((b) << 3) | (r)) +#endif +#define DITH_DX 4 +#define DITH_DY 4 +#define DITH_N (DITH_DX * DITH_DY) + +#define _dither(C, c, d) (((unsigned)((DITH_N * (C - 1) + 1) * c + d)) >> 12) + +#define MAXC 256 +extern const int xmesa_kernel8[DITH_DY * DITH_DX]; + +/* Dither for random X,Y */ +#define DITHER_SETUP \ + int __d; \ + unsigned long *ctable = XMESA_BUFFER(ctx->DrawBuffer)->color_table; + +#define DITHER( X, Y, R, G, B ) \ + (__d = xmesa_kernel8[(((Y)&3)<<2) | ((X)&3)], \ + ctable[DITH_MIX(_dither(DITH_R, (R), __d), \ + _dither(DITH_G, (G), __d), \ + _dither(DITH_B, (B), __d))]) + +/* Dither for random X, fixed Y */ +#define XDITHER_SETUP(Y) \ + int __d; \ + unsigned long *ctable = XMESA_BUFFER(ctx->DrawBuffer)->color_table; \ + const int *kernel = &xmesa_kernel8[ ((Y)&3) << 2 ]; + +#define XDITHER( X, R, G, B ) \ + (__d = kernel[(X)&3], \ + ctable[DITH_MIX(_dither(DITH_R, (R), __d), \ + _dither(DITH_G, (G), __d), \ + _dither(DITH_B, (B), __d))]) + + + +/* + * Dithering for flat-shaded triangles. Precompute all 16 possible + * pixel values given the triangle's RGB color. Contributed by Martin Shenk. + */ +#define FLAT_DITHER_SETUP( R, G, B ) \ + GLushort ditherValues[16]; \ + { \ + unsigned long *ctable = XMESA_BUFFER(ctx->DrawBuffer)->color_table; \ + int msdr = (DITH_N*((DITH_R)-1)+1) * (R); \ + int msdg = (DITH_N*((DITH_G)-1)+1) * (G); \ + int msdb = (DITH_N*((DITH_B)-1)+1) * (B); \ + int i; \ + for (i=0;i<16;i++) { \ + int k = xmesa_kernel8[i]; \ + int j = DITH_MIX( (msdr+k)>>12, (msdg+k)>>12, (msdb+k)>>12 );\ + ditherValues[i] = (GLushort) ctable[j]; \ + } \ + } + +#define FLAT_DITHER_ROW_SETUP(Y) \ + GLushort *ditherRow = ditherValues + ( ((Y)&3) << 2); + +#define FLAT_DITHER(X) ditherRow[(X)&3] + + + +/** + * If pixelformat==PF_LOOKUP: + */ +#define _dither_lookup(C, c) (((unsigned)((DITH_N * (C - 1) + 1) * c)) >> 12) + +#define LOOKUP_SETUP \ + unsigned long *ctable = XMESA_BUFFER(ctx->DrawBuffer)->color_table + +#define LOOKUP( R, G, B ) \ + ctable[DITH_MIX(_dither_lookup(DITH_R, (R)), \ + _dither_lookup(DITH_G, (G)), \ + _dither_lookup(DITH_B, (B)))] + + +/** + * If pixelformat==PF_HPCR: + * + * HP Color Recovery dithering (ad@lms.be 30/08/95) + * HP has on it's 8-bit 700-series computers, a feature called + * 'Color Recovery'. This allows near 24-bit output (so they say). + * It is enabled by selecting the 8-bit TrueColor visual AND + * corresponding colormap (see tkInitWindow) AND doing some special + * dither. + */ +extern const short xmesa_HPCR_DRGB[3][2][16]; + +#define DITHER_HPCR( X, Y, R, G, B ) \ + ( ((xmesa->xm_visual->hpcr_rgbTbl[0][R] + xmesa_HPCR_DRGB[0][(Y)&1][(X)&15]) & 0xE0) \ + |(((xmesa->xm_visual->hpcr_rgbTbl[1][G] + xmesa_HPCR_DRGB[1][(Y)&1][(X)&15]) & 0xE0)>>3) \ + | ((xmesa->xm_visual->hpcr_rgbTbl[2][B] + xmesa_HPCR_DRGB[2][(Y)&1][(X)&15])>>6) \ + ) + + + +/** + * If pixelformat==PF_1BIT: + */ +extern const int xmesa_kernel1[16]; + +#define SETUP_1BIT int bitFlip = xmesa->xm_visual->bitFlip +#define DITHER_1BIT( X, Y, R, G, B ) \ + (( ((int)(R)+(int)(G)+(int)(B)) > xmesa_kernel1[(((Y)&3) << 2) | ((X)&3)] ) ^ bitFlip) + + + +/** + * If pixelformat==PF_GRAYSCALE: + */ +#define GRAY_RGB( R, G, B ) XMESA_BUFFER(ctx->DrawBuffer)->color_table[((R) + (G) + (B))/3] + + + +/** + * Converts a GL window Y coord to an X window Y coord: + */ +#define YFLIP(XRB, Y) ((XRB)->bottom - (Y)) + + +/** + * Return the address of a 1, 2 or 4-byte pixel in the buffer's XImage: + * X==0 is left, Y==0 is bottom. + */ +#define PIXEL_ADDR1(XRB, X, Y) \ + ( (XRB)->origin1 - (Y) * (XRB)->width1 + (X) ) + +#define PIXEL_ADDR2(XRB, X, Y) \ + ( (XRB)->origin2 - (Y) * (XRB)->width2 + (X) ) + +#define PIXEL_ADDR3(XRB, X, Y) \ + ( (bgr_t *) ( (XRB)->origin3 - (Y) * (XRB)->width3 + 3 * (X) )) + +#define PIXEL_ADDR4(XRB, X, Y) \ + ( (XRB)->origin4 - (Y) * (XRB)->width4 + (X) ) + + + +/* + * External functions: + */ + +extern struct xmesa_renderbuffer * +xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, + GLboolean backBuffer); + +extern void +xmesa_delete_framebuffer(struct gl_framebuffer *fb); + +extern XMesaBuffer +xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis); + +extern unsigned long +xmesa_color_to_pixel( GLcontext *ctx, + GLubyte r, GLubyte g, GLubyte b, GLubyte a, + GLuint pixelFormat ); + +extern void +xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b, + GLuint *width, GLuint *height); + +extern void +xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer); + +extern void +xmesa_init_driver_functions( XMesaVisual xmvisual, + struct dd_function_table *driver ); + +extern void +xmesa_update_state( GLcontext *ctx, GLbitfield new_state ); + +extern void +xmesa_set_renderbuffer_funcs(struct xmesa_renderbuffer *xrb, + enum pixel_format pixelformat, GLint depth); + +extern void +xmesa_destroy_buffers_on_display(XMesaDisplay *dpy); + + +/** + * Using a function instead of an ordinary cast is safer. + */ +static INLINE struct xmesa_renderbuffer * +xmesa_renderbuffer(struct gl_renderbuffer *rb) +{ + return (struct xmesa_renderbuffer *) rb; +} + + +/** + * Return pointer to XMesaContext corresponding to a Mesa GLcontext. + * Since we're using structure containment, it's just a cast!. + * XXX should use inlined function for better type safety. + */ +static INLINE XMesaContext +XMESA_CONTEXT(GLcontext *ctx) +{ + return (XMesaContext) ctx; +} + + +/** + * Return pointer to XMesaBuffer corresponding to a Mesa GLframebuffer. + * Since we're using structure containment, it's just a cast!. + * XXX should use inlined function for better type safety. + */ +static INLINE XMesaBuffer +XMESA_BUFFER(GLframebuffer *b) +{ + return (XMesaBuffer) b; +} + + +/* Plugged into the software rasterizer. Try to use internal + * swrast-style point, line and triangle functions. + */ +extern void xmesa_choose_point( GLcontext *ctx ); +extern void xmesa_choose_line( GLcontext *ctx ); +extern void xmesa_choose_triangle( GLcontext *ctx ); + + +extern void xmesa_register_swrast_functions( GLcontext *ctx ); + + + +#define ENABLE_EXT_texure_compression_s3tc 0 /* SW texture compression */ + +#ifdef XFree86Server +#define ENABLE_EXT_timer_query 0 +#elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L +#define ENABLE_EXT_timer_query 1 /* should have 64-bit GLuint64EXT */ +#else +#define ENABLE_EXT_timer_query 0 /* may not have 64-bit GLuint64EXT */ +#endif + + +struct pipe_surface; +struct pipe_context; + +struct xmesa_surface +{ + struct pipe_surface surface; + struct xmesa_renderbuffer *xrb; +}; + + +extern void +xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value); + +extern void +xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers); + +extern struct pipe_context * +xmesa_create_softpipe(XMesaContext xm); + +extern struct pipe_surface * +xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); + +extern struct pipe_surface * +xmesa_new_color_surface(struct pipe_context *pipe, GLuint format); + +extern boolean +xmesa_is_format_supported(struct pipe_context *pipe, uint format); + +extern void +xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, float *p); + +extern void +xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const float *p); + +#endif -- cgit v1.2.3 From 87560bc10f57fd523d3317dfa1ff98d40099eb39 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:19:31 -0600 Subject: switch to Xlib driver in pipe/xlib/ --- src/mesa/sources | 23 +++++++++++------------ 1 file changed, 11 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index c983d802d4..312e48499e 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -328,18 +328,17 @@ COMMON_DRIVER_SOURCES = \ drivers/common/driverfuncs.c X11_DRIVER_SOURCES = \ - drivers/x11/glxapi.c \ - drivers/x11/fakeglx.c \ - drivers/x11/xfonts.c \ - drivers/x11/xm_api.c \ - drivers/x11/xm_buffer.c \ - drivers/x11/xm_dd.c \ - drivers/x11/xm_glide.c \ - drivers/x11/xm_line.c \ - drivers/x11/xm_winsys.c \ - drivers/x11/xm_span.c \ - drivers/x11/xm_surface.c \ - drivers/x11/xm_tri.c + pipe/xlib/glxapi.c \ + pipe/xlib/fakeglx.c \ + pipe/xlib/xfonts.c \ + pipe/xlib/xm_api.c \ + pipe/xlib/xm_buffer.c \ + pipe/xlib/xm_dd.c \ + pipe/xlib/xm_line.c \ + pipe/xlib/xm_winsys.c \ + pipe/xlib/xm_span.c \ + pipe/xlib/xm_surface.c \ + pipe/xlib/xm_tri.c OSMESA_DRIVER_SOURCES = \ drivers/osmesa/osmesa.c -- cgit v1.2.3 From 12f3f6cb2ac012c111d046e77772acd5561ebbf6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:22:49 -0600 Subject: omit X point/line/triangle routines --- src/mesa/pipe/xlib/xm_api.c | 3 ++- src/mesa/pipe/xlib/xm_dd.c | 2 ++ src/mesa/sources | 4 +--- 3 files changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 08c98eab48..3632390364 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1615,9 +1615,10 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) tnl->Driver.RunPipeline = _tnl_run_pipeline; #endif +#if 0 /* swrast setup */ xmesa_register_swrast_functions( mesaCtx ); - +#endif st_create_context( mesaCtx, xmesa_create_softpipe( c ) ); diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 8ae243ae66..88ac0e594f 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -1121,6 +1121,7 @@ xmesa_init_driver_functions( XMesaVisual xmvisual, _SWRAST_NEW_RASTERMASK) +#if 0 /** * Extend the software rasterizer with our line/point/triangle * functions. @@ -1139,3 +1140,4 @@ void xmesa_register_swrast_functions( GLcontext *ctx ) swrast->InvalidateLineMask |= XMESA_NEW_LINE; swrast->InvalidateTriangleMask |= XMESA_NEW_TRIANGLE; } +#endif diff --git a/src/mesa/sources b/src/mesa/sources index 312e48499e..c5c07cf442 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -334,11 +334,9 @@ X11_DRIVER_SOURCES = \ pipe/xlib/xm_api.c \ pipe/xlib/xm_buffer.c \ pipe/xlib/xm_dd.c \ - pipe/xlib/xm_line.c \ pipe/xlib/xm_winsys.c \ pipe/xlib/xm_span.c \ - pipe/xlib/xm_surface.c \ - pipe/xlib/xm_tri.c + pipe/xlib/xm_surface.c OSMESA_DRIVER_SOURCES = \ drivers/osmesa/osmesa.c -- cgit v1.2.3 From 5d6b314bcf64d3949038bdd5845793ce7dd0ecd0 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:27:38 -0600 Subject: remove Draw/CopyPixels functions --- src/mesa/pipe/xlib/xm_dd.c | 284 +-------------------------------------------- 1 file changed, 5 insertions(+), 279 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 88ac0e594f..65eec028ee 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -50,8 +50,10 @@ #include "swrast/swrast.h" #include "swrast/s_context.h" #include "swrast_setup/swrast_setup.h" +#if 0 #include "tnl/tnl.h" #include "tnl/t_context.h" +#endif #include "pipe/softpipe/sp_context.h" #include "state_tracker/st_public.h" @@ -423,274 +425,6 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers) } -#ifndef XFree86Server -/* XXX this was never tested in the Xserver environment */ - -/** - * This function implements glDrawPixels() with an XPutImage call when - * drawing to the front buffer (X Window drawable). - * The image format must be GL_BGRA to match the PF_8R8G8B pixel format. - */ -static void -xmesa_DrawPixels_8R8G8B( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels ) -{ - const SWcontext *swrast = SWRAST_CONTEXT( ctx ); - struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[0][0]; - struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb->Wrapped); - - if (swrast->NewState) - _swrast_validate_derived( ctx ); - - if (ctx->DrawBuffer->Name == 0 && - format == GL_BGRA && - type == GL_UNSIGNED_BYTE && - (swrast->_RasterMask & ~CLIP_BIT) == 0 && /* no blend, z-test, etc */ - ctx->_ImageTransferState == 0 && /* no color tables, scale/bias, etc */ - ctx->Pixel.ZoomX == 1.0 && /* no zooming */ - ctx->Pixel.ZoomY == 1.0 && - xrb->pixmap && /* drawing to pixmap or window */ - xrb->St.Base.AlphaBits == 0) - { - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - XMesaDisplay *dpy = xmesa->xm_visual->display; - const XMesaGC gc = xmbuf->cleargc; /* effected by glColorMask */ - int dstX = x; - int dstY = y; - int w = width; - int h = height; - struct gl_pixelstore_attrib clippedUnpack = *unpack; - - ASSERT(xmesa->xm_visual->dithered_pf == PF_8R8G8B); - ASSERT(xmesa->xm_visual->undithered_pf == PF_8R8G8B); - ASSERT(dpy); - ASSERT(gc); - - if (unpack->BufferObj->Name) { - /* unpack from PBO */ - GLubyte *buf; - if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glDrawPixels(invalid PBO access)"); - return; - } - buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, - GL_PIXEL_UNPACK_BUFFER_EXT, - GL_READ_ONLY_ARB, - unpack->BufferObj); - if (!buf) { - /* buffer is already mapped - that's an error */ - _mesa_error(ctx, GL_INVALID_OPERATION, - "glDrawPixels(PBO is mapped)"); - return; - } - pixels = ADD_POINTERS(buf, pixels); - } - - if (_mesa_clip_drawpixels(ctx, &dstX, &dstY, &w, &h, &clippedUnpack)) { - /* This is a little tricky since all coordinates up to now have - * been in the OpenGL bottom-to-top orientation. X is top-to-bottom - * so we have to carefully compute the Y coordinates/addresses here. - */ - int srcX = clippedUnpack.SkipPixels; - int srcY = clippedUnpack.SkipRows; - int rowLength = clippedUnpack.RowLength; - XMesaImage ximage; - MEMSET(&ximage, 0, sizeof(XMesaImage)); - ximage.width = width; - ximage.height = height; - ximage.format = ZPixmap; - ximage.data = (char *) pixels - + ((srcY + h - 1) * rowLength + srcX) * 4; - ximage.byte_order = LSBFirst; - ximage.bitmap_unit = 32; - ximage.bitmap_bit_order = LSBFirst; - ximage.bitmap_pad = 32; - ximage.depth = 24; - ximage.bytes_per_line = -rowLength * 4; /* negative to flip image */ - ximage.bits_per_pixel = 32; - /* it seems we don't need to set the ximage.red/green/blue_mask fields */ - /* flip Y axis for dest position */ - dstY = YFLIP(xrb, dstY) - h + 1; - XPutImage(dpy, xrb->pixmap, gc, &ximage, 0, 0, dstX, dstY, w, h); - } - - if (unpack->BufferObj->Name) { - ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, - unpack->BufferObj); - } - } - else { - /* software fallback */ - _swrast_DrawPixels(ctx, x, y, width, height, - format, type, unpack, pixels); - } -} - - - -/** - * This function implements glDrawPixels() with an XPutImage call when - * drawing to the front buffer (X Window drawable). The image format - * must be GL_RGB and image type must be GL_UNSIGNED_SHORT_5_6_5 to - * match the PF_5R6G5B pixel format. - */ -static void -xmesa_DrawPixels_5R6G5B( GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels ) -{ - struct xmesa_renderbuffer *xrb - = xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped); - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - const SWcontext *swrast = SWRAST_CONTEXT( ctx ); - XMesaDisplay *dpy = xmesa->xm_visual->display; - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - const XMesaGC gc = xmbuf->cleargc; /* effected by glColorMask */ - - ASSERT(dpy); - ASSERT(gc); - ASSERT(xmesa->xm_visual->undithered_pf == PF_5R6G5B); - - if (swrast->NewState) - _swrast_validate_derived( ctx ); - - if (xrb->pixmap && /* drawing to pixmap or window */ - format == GL_RGB && - type == GL_UNSIGNED_SHORT_5_6_5 && - !ctx->Color.DitherFlag && /* no dithering */ - (swrast->_RasterMask & ~CLIP_BIT) == 0 && /* no blend, z-test, etc */ - ctx->_ImageTransferState == 0 && /* no color tables, scale/bias, etc */ - ctx->Pixel.ZoomX == 1.0 && /* no zooming */ - ctx->Pixel.ZoomY == 1.0) { - int dstX = x; - int dstY = y; - int w = width; - int h = height; - struct gl_pixelstore_attrib clippedUnpack = *unpack; - - if (unpack->BufferObj->Name) { - /* unpack from PBO */ - GLubyte *buf; - if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glDrawPixels(invalid PBO access)"); - return; - } - buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, - GL_PIXEL_UNPACK_BUFFER_EXT, - GL_READ_ONLY_ARB, - unpack->BufferObj); - if (!buf) { - /* buffer is already mapped - that's an error */ - _mesa_error(ctx, GL_INVALID_OPERATION, - "glDrawPixels(PBO is mapped)"); - return; - } - pixels = ADD_POINTERS(buf, pixels); - } - - if (_mesa_clip_drawpixels(ctx, &dstX, &dstY, &w, &h, &clippedUnpack)) { - /* This is a little tricky since all coordinates up to now have - * been in the OpenGL bottom-to-top orientation. X is top-to-bottom - * so we have to carefully compute the Y coordinates/addresses here. - */ - int srcX = clippedUnpack.SkipPixels; - int srcY = clippedUnpack.SkipRows; - int rowLength = clippedUnpack.RowLength; - XMesaImage ximage; - MEMSET(&ximage, 0, sizeof(XMesaImage)); - ximage.width = width; - ximage.height = height; - ximage.format = ZPixmap; - ximage.data = (char *) pixels - + ((srcY + h - 1) * rowLength + srcX) * 2; - ximage.byte_order = LSBFirst; - ximage.bitmap_unit = 16; - ximage.bitmap_bit_order = LSBFirst; - ximage.bitmap_pad = 16; - ximage.depth = 16; - ximage.bytes_per_line = -rowLength * 2; /* negative to flip image */ - ximage.bits_per_pixel = 16; - /* it seems we don't need to set the ximage.red/green/blue_mask fields */ - /* flip Y axis for dest position */ - dstY = YFLIP(xrb, dstY) - h + 1; - XPutImage(dpy, xrb->pixmap, gc, &ximage, 0, 0, dstX, dstY, w, h); - } - - if (unpack->BufferObj->Name) { - ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, - unpack->BufferObj); - } - } - else { - /* software fallback */ - _swrast_DrawPixels(ctx, x, y, width, height, - format, type, unpack, pixels); - } -} - - - -/** - * Implement glCopyPixels for the front color buffer (or back buffer Pixmap) - * for the color buffer. Don't support zooming, pixel transfer, etc. - * We do support copying from one window to another, ala glXMakeCurrentRead. - */ -static void -xmesa_CopyPixels( GLcontext *ctx, - GLint srcx, GLint srcy, GLsizei width, GLsizei height, - GLint destx, GLint desty, GLenum type ) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - const SWcontext *swrast = SWRAST_CONTEXT( ctx ); - XMesaDisplay *dpy = xmesa->xm_visual->display; - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - const XMesaGC gc = xmbuf->cleargc; /* effected by glColorMask */ - struct xmesa_renderbuffer *srcXrb - = xmesa_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer->Wrapped); - struct xmesa_renderbuffer *dstXrb - = xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped); - - ASSERT(dpy); - ASSERT(gc); - - if (swrast->NewState) - _swrast_validate_derived( ctx ); - - if (ctx->Color.DrawBuffer[0] == GL_FRONT && - ctx->Pixel.ReadBuffer == GL_FRONT && - srcXrb->pixmap && - dstXrb->pixmap && - type == GL_COLOR && - (swrast->_RasterMask & ~CLIP_BIT) == 0 && /* no blend, z-test, etc */ - ctx->_ImageTransferState == 0 && /* no color tables, scale/bias, etc */ - ctx->Pixel.ZoomX == 1.0 && /* no zooming */ - ctx->Pixel.ZoomY == 1.0) { - /* Note: we don't do any special clipping work here. We could, - * but X will do it for us. - */ - srcy = YFLIP(srcXrb, srcy) - height + 1; - desty = YFLIP(dstXrb, desty) - height + 1; - XCopyArea(dpy, srcXrb->pixmap, dstXrb->pixmap, gc, - srcx, srcy, width, height, destx, desty); - } - else { - _swrast_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type ); - } -} -#endif /* XFree86Server */ - - - /* * Every driver should implement a GetString function in order to * return a meaningful GL_RENDERER string. @@ -826,10 +560,12 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) /* Propagate statechange information to swrast and swrast_setup * modules. The X11 driver has no internal GL-dependent state. */ +#if 0 _swrast_InvalidateState( ctx, new_state ); _tnl_InvalidateState( ctx, new_state ); - _vbo_InvalidateState( ctx, new_state ); _swsetup_InvalidateState( ctx, new_state ); + _vbo_InvalidateState( ctx, new_state ); +#endif st_invalidate_state( ctx, new_state ); @@ -1076,16 +812,6 @@ xmesa_init_driver_functions( XMesaVisual xmvisual, driver->Enable = enable; driver->Clear = xmesa_clear_buffers; driver->Viewport = xmesa_viewport; -#ifndef XFree86Server - driver->CopyPixels = xmesa_CopyPixels; - if (xmvisual->undithered_pf == PF_8R8G8B && - xmvisual->dithered_pf == PF_8R8G8B) { - driver->DrawPixels = xmesa_DrawPixels_8R8G8B; - } - else if (xmvisual->undithered_pf == PF_5R6G5B) { - driver->DrawPixels = xmesa_DrawPixels_5R6G5B; - } -#endif driver->TestProxyTexImage = test_proxy_teximage; #if ENABLE_EXT_texure_compression_s3tc driver->ChooseTextureFormat = choose_tex_format; -- cgit v1.2.3 From e0e8cf1346513eb52771924e7efd8f19e935dd0a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:28:05 -0600 Subject: disable swrast/tnl stuff --- src/mesa/pipe/xlib/xm_api.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 3632390364..d13d98857a 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -73,8 +73,10 @@ #include "macros.h" #include "renderbuffer.h" #include "teximage.h" +#if 0 #include "swrast/swrast.h" #include "swrast_setup/swrast_setup.h" +#endif #include "vbo/vbo.h" #if 0 #include "tnl/tnl.h" @@ -1597,6 +1599,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) /* Initialize the software rasterizer and helper modules. */ +#if 0 if (!_swrast_CreateContext( mesaCtx ) #if 0 || !_vbo_CreateContext( mesaCtx ) || @@ -1608,6 +1611,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) _mesa_free(c); return NULL; } +#endif #if 0 /* tnl setup */ @@ -1623,8 +1627,10 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) st_create_context( mesaCtx, xmesa_create_softpipe( c ) ); +#if 0 _swsetup_CreateContext( mesaCtx ); _swsetup_Wakeup(mesaCtx); +#endif /* override these functions, as if the xlib driver were derived from * the softpipe driver. @@ -1655,8 +1661,10 @@ void XMesaDestroyContext( XMesaContext c ) FXdestroyContext( XMESA_BUFFER(mesaCtx->DrawBuffer) ); #endif +#if 0 _swsetup_DestroyContext( mesaCtx ); _swrast_DestroyContext( mesaCtx ); +#endif #if 0 _tnl_DestroyContext( mesaCtx ); _vbo_DestroyContext( mesaCtx ); -- cgit v1.2.3 From 68a38b58537280dada02f4451d56268f064e59bb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:29:36 -0600 Subject: remove dead swrast/tnl code --- src/mesa/pipe/xlib/xm_dd.c | 60 ---------------------------------------------- 1 file changed, 60 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 65eec028ee..3c5c413c73 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -47,13 +47,6 @@ #include "texstore.h" #include "texformat.h" #include "xmesaP.h" -#include "swrast/swrast.h" -#include "swrast/s_context.h" -#include "swrast_setup/swrast_setup.h" -#if 0 -#include "tnl/tnl.h" -#include "tnl/t_context.h" -#endif #include "pipe/softpipe/sp_context.h" #include "state_tracker/st_public.h" @@ -420,8 +413,6 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers) } } } - if (buffers) - _swrast_Clear(ctx, buffers); } @@ -557,16 +548,6 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); - /* Propagate statechange information to swrast and swrast_setup - * modules. The X11 driver has no internal GL-dependent state. - */ -#if 0 - _swrast_InvalidateState( ctx, new_state ); - _tnl_InvalidateState( ctx, new_state ); - _swsetup_InvalidateState( ctx, new_state ); - _vbo_InvalidateState( ctx, new_state ); -#endif - st_invalidate_state( ctx, new_state ); @@ -826,44 +807,3 @@ xmesa_init_driver_functions( XMesaVisual xmvisual, #endif } - - -#define XMESA_NEW_POINT (_NEW_POINT | \ - _NEW_RENDERMODE | \ - _SWRAST_NEW_RASTERMASK) - -#define XMESA_NEW_LINE (_NEW_LINE | \ - _NEW_TEXTURE | \ - _NEW_LIGHT | \ - _NEW_DEPTH | \ - _NEW_RENDERMODE | \ - _SWRAST_NEW_RASTERMASK) - -#define XMESA_NEW_TRIANGLE (_NEW_POLYGON | \ - _NEW_TEXTURE | \ - _NEW_LIGHT | \ - _NEW_DEPTH | \ - _NEW_RENDERMODE | \ - _SWRAST_NEW_RASTERMASK) - - -#if 0 -/** - * Extend the software rasterizer with our line/point/triangle - * functions. - * Called during context creation only. - */ -void xmesa_register_swrast_functions( GLcontext *ctx ) -{ - SWcontext *swrast = SWRAST_CONTEXT( ctx ); - - swrast->choose_point = xmesa_choose_point; - swrast->choose_line = xmesa_choose_line; - swrast->choose_triangle = xmesa_choose_triangle; - - /* XXX these lines have no net effect. Remove??? */ - swrast->InvalidatePointMask |= XMESA_NEW_POINT; - swrast->InvalidateLineMask |= XMESA_NEW_LINE; - swrast->InvalidateTriangleMask |= XMESA_NEW_TRIANGLE; -} -#endif -- cgit v1.2.3 From 751d7a74ca00df44dbefee628b2a305e68638479 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:30:20 -0600 Subject: Obsolete --- src/mesa/pipe/xlib/xm_line.c | 691 ----------------- src/mesa/pipe/xlib/xm_tri.c | 1694 ------------------------------------------ 2 files changed, 2385 deletions(-) delete mode 100644 src/mesa/pipe/xlib/xm_line.c delete mode 100644 src/mesa/pipe/xlib/xm_tri.c (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_line.c b/src/mesa/pipe/xlib/xm_line.c deleted file mode 100644 index deeae5019c..0000000000 --- a/src/mesa/pipe/xlib/xm_line.c +++ /dev/null @@ -1,691 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/* - * This file contains "accelerated" point, line, and triangle functions. - * It should be fairly easy to write new special-purpose point, line or - * triangle functions and hook them into this module. - */ - - -#include "glxheader.h" -#include "depth.h" -#include "macros.h" -#include "mtypes.h" -#include "xmesaP.h" - -/* Internal swrast includes: - */ -#include "swrast/s_depth.h" -#include "swrast/s_points.h" -#include "swrast/s_lines.h" -#include "swrast/s_context.h" - - -/**********************************************************************/ -/*** Point rendering ***/ -/**********************************************************************/ - - -/* - * Render an array of points into a pixmap, any pixel format. - */ -#if 000 -/* XXX don't use this, it doesn't dither correctly */ -static void draw_points_ANY_pixmap( GLcontext *ctx, const SWvertex *vert ) -{ - XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaDisplay *dpy = xmesa->xm_visual->display; - XMesaDrawable buffer = xmesa->xm_buffer->buffer; - XMesaGC gc = xmesa->xm_buffer->gc; - - if (xmesa->xm_visual->mesa_visual.RGBAflag) { - register int x, y; - const GLubyte *color = vert->color; - unsigned long pixel = xmesa_color_to_pixel( xmesa, - color[0], color[1], - color[2], color[3], - xmesa->pixelformat); - XMesaSetForeground( dpy, gc, pixel ); - x = (GLint) vert->win[0]; - y = YFLIP( xrb, (GLint) vert->win[1] ); - XMesaDrawPoint( dpy, buffer, gc, x, y); - } - else { - /* Color index mode */ - register int x, y; - XMesaSetForeground( dpy, gc, vert->index ); - x = (GLint) vert->win[0]; - y = YFLIP( xrb, (GLint) vert->win[1] ); - XMesaDrawPoint( dpy, buffer, gc, x, y); - } -} -#endif - - -/* Override the swrast point-selection function. Try to use one of - * our internal point functions, otherwise fall back to the standard - * swrast functions. - */ -void xmesa_choose_point( GLcontext *ctx ) -{ -#if 0 - XMesaContext xmesa = XMESA_CONTEXT(ctx); - SWcontext *swrast = SWRAST_CONTEXT(ctx); - - if (ctx->RenderMode == GL_RENDER - && ctx->Point.Size == 1.0F && !ctx->Point.SmoothFlag - && swrast->_RasterMask == 0 - && !ctx->Texture._EnabledUnits - && xmesa->xm_buffer->buffer != XIMAGE) { - swrast->Point = draw_points_ANY_pixmap; - } - else { - _swrast_choose_point( ctx ); - } -#else - _swrast_choose_point( ctx ); -#endif -} - - - -/**********************************************************************/ -/*** Line rendering ***/ -/**********************************************************************/ - - -#if CHAN_BITS == 8 - - -#define GET_XRB(XRB) struct xmesa_renderbuffer *XRB = \ - xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped) - - -/* - * Draw a flat-shaded, PF_TRUECOLOR line into an XImage. - */ -#define NAME flat_TRUECOLOR_line -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - unsigned long pixel; \ - PACK_TRUECOLOR( pixel, color[0], color[1], color[2] ); -#define CLIP_HACK 1 -#define PLOT(X,Y) XMesaPutPixel(xrb->ximage, X, YFLIP(xrb, Y), pixel ); -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, PF_8A8B8G8R line into an XImage. - */ -#define NAME flat_8A8B8G8R_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLuint pixel = PACK_8A8B8G8R(color[0], color[1], color[2], color[3]); -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) *pixelPtr = pixel; -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, PF_8A8R8G8B line into an XImage. - */ -#define NAME flat_8A8R8G8B_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLuint pixel = PACK_8A8R8G8B(color[0], color[1], color[2], color[3]); -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) *pixelPtr = pixel; -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, PF_8R8G8B line into an XImage. - */ -#define NAME flat_8R8G8B_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLuint pixel = PACK_8R8G8B( color[0], color[1], color[2] ); -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) *pixelPtr = pixel; -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, PF_8R8G8B24 line into an XImage. - */ -#define NAME flat_8R8G8B24_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; -#define PIXEL_TYPE bgr_t -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) { \ - pixelPtr->r = color[RCOMP]; \ - pixelPtr->g = color[GCOMP]; \ - pixelPtr->b = color[BCOMP]; \ -} -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, PF_5R6G5B line into an XImage. - */ -#define NAME flat_5R6G5B_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLushort pixel = PACK_5R6G5B( color[0], color[1], color[2] ); -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) *pixelPtr = pixel; -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, PF_DITHER_5R6G5B line into an XImage. - */ -#define NAME flat_DITHER_5R6G5B_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - const GLubyte *color = vert1->color; -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) PACK_TRUEDITHER( *pixelPtr, X, Y, color[0], color[1], color[2] ); -#include "swrast/s_linetemp.h" - - - - -/* - * Draw a flat-shaded, PF_DITHER 8-bit line into an XImage. - */ -#define NAME flat_DITHER8_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLint r = color[0], g = color[1], b = color[2]; \ - DITHER_SETUP; -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) *pixelPtr = DITHER(X,Y,r,g,b); -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, PF_LOOKUP 8-bit line into an XImage. - */ -#define NAME flat_LOOKUP8_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLubyte pixel; \ - LOOKUP_SETUP; \ - pixel = (GLubyte) LOOKUP( color[0], color[1], color[2] ); -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) *pixelPtr = pixel; -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, PF_HPCR line into an XImage. - */ -#define NAME flat_HPCR_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - const GLubyte *color = vert1->color; \ - GLint r = color[0], g = color[1], b = color[2]; -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) *pixelPtr = (GLubyte) DITHER_HPCR(X,Y,r,g,b); -#include "swrast/s_linetemp.h" - - - - -/* - * Draw a flat-shaded, Z-less, PF_TRUECOLOR line into an XImage. - */ -#define NAME flat_TRUECOLOR_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - const GLubyte *color = vert1->color; \ - unsigned long pixel; \ - PACK_TRUECOLOR( pixel, color[0], color[1], color[2] ); -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - XMesaPutPixel(xrb->ximage, X, YFLIP(xrb, Y), pixel); \ - } -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, Z-less, PF_8A8B8G8R line into an XImage. - */ -#define NAME flat_8A8B8G8R_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLuint pixel = PACK_8A8B8G8R(color[0], color[1], color[2], color[3]); -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - *pixelPtr = pixel; \ - } -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, Z-less, PF_8A8R8G8B line into an XImage. - */ -#define NAME flat_8A8R8G8B_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLuint pixel = PACK_8A8R8G8B(color[0], color[1], color[2], color[3]); -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - *pixelPtr = pixel; \ - } -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, Z-less, PF_8R8G8B line into an XImage. - */ -#define NAME flat_8R8G8B_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLuint pixel = PACK_8R8G8B( color[0], color[1], color[2] ); -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - *pixelPtr = pixel; \ - } -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, Z-less, PF_8R8G8B24 line into an XImage. - */ -#define NAME flat_8R8G8B24_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_TYPE bgr_t -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - pixelPtr->r = color[RCOMP]; \ - pixelPtr->g = color[GCOMP]; \ - pixelPtr->b = color[BCOMP]; \ - } -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, Z-less, PF_5R6G5B line into an XImage. - */ -#define NAME flat_5R6G5B_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLushort pixel = PACK_5R6G5B( color[0], color[1], color[2] ); -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - *pixelPtr = pixel; \ - } -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, Z-less, PF_DITHER_5R6G5B line into an XImage. - */ -#define NAME flat_DITHER_5R6G5B_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - const GLubyte *color = vert1->color; -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - PACK_TRUEDITHER(*pixelPtr, X, Y, color[0], color[1], color[2]); \ - } -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, Z-less, PF_DITHER 8-bit line into an XImage. - */ -#define NAME flat_DITHER8_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLint r = color[0], g = color[1], b = color[2]; \ - DITHER_SETUP; -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - *pixelPtr = (GLubyte) DITHER( X, Y, r, g, b); \ - } -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, Z-less, PF_LOOKUP 8-bit line into an XImage. - */ -#define NAME flat_LOOKUP8_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = vert1->color; \ - GLubyte pixel; \ - LOOKUP_SETUP; \ - pixel = (GLubyte) LOOKUP( color[0], color[1], color[2] ); -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - *pixelPtr = pixel; \ - } -#include "swrast/s_linetemp.h" - - - -/* - * Draw a flat-shaded, Z-less, PF_HPCR line into an XImage. - */ -#define NAME flat_HPCR_z_line -#define SETUP_CODE \ - GET_XRB(xrb); \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - const GLubyte *color = vert1->color; \ - GLint r = color[0], g = color[1], b = color[2]; -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X,Y) -#define CLIP_HACK 1 -#define PLOT(X,Y) \ - if (Z < *zPtr) { \ - *zPtr = Z; \ - *pixelPtr = (GLubyte) DITHER_HPCR( X, Y, r, g, b); \ - } -#include "swrast/s_linetemp.h" - - - - -#ifndef XFree86Server -/** - * Draw fast, XOR line with XDrawLine in front color buffer. - * WARNING: this isn't fully OpenGL conformant because different pixels - * will be hit versus using the other line functions. - * Don't use the code in X server GLcore module since we need a wrapper - * for the XSetLineAttributes() function call. - */ -static void -xor_line(GLcontext *ctx, const SWvertex *vert0, const SWvertex *vert1) -{ - XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaDisplay *dpy = xmesa->xm_visual->display; - XMesaGC gc = xmesa->xm_buffer->gc; - GET_XRB(xrb); - unsigned long pixel = xmesa_color_to_pixel(ctx, - vert1->color[0], vert1->color[1], - vert1->color[2], vert1->color[3], - xmesa->pixelformat); - int x0 = (GLint) vert0->attrib[FRAG_ATTRIB_WPOS][0]; - int y0 = YFLIP(xrb, (GLint) vert0->attrib[FRAG_ATTRIB_WPOS][1]); - int x1 = (GLint) vert1->attrib[FRAG_ATTRIB_WPOS][0]; - int y1 = YFLIP(xrb, (GLint) vert1->attrib[FRAG_ATTRIB_WPOS][1]); - XMesaSetForeground(dpy, gc, pixel); - XMesaSetFunction(dpy, gc, GXxor); - XSetLineAttributes(dpy, gc, (int) ctx->Line.Width, - LineSolid, CapButt, JoinMiter); - XDrawLine(dpy, xrb->pixmap, gc, x0, y0, x1, y1); - XMesaSetFunction(dpy, gc, GXcopy); /* this gc is used elsewhere */ -} -#endif /* XFree86Server */ - - -#endif /* CHAN_BITS == 8 */ - - -/** - * Return pointer to line drawing function, or NULL if we should use a - * swrast fallback. - */ -static swrast_line_func -get_line_func(GLcontext *ctx) -{ -#if CHAN_BITS == 8 - SWcontext *swrast = SWRAST_CONTEXT(ctx); - XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - const int depth = GET_VISUAL_DEPTH(xmesa->xm_visual); - const struct xmesa_renderbuffer *xrb; - - if ((ctx->DrawBuffer->_ColorDrawBufferMask[0] - & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)) == 0) - return (swrast_line_func) NULL; - if (ctx->RenderMode != GL_RENDER) return (swrast_line_func) NULL; - if (ctx->Line.SmoothFlag) return (swrast_line_func) NULL; - if (ctx->Texture._EnabledUnits) return (swrast_line_func) NULL; - if (ctx->Light.ShadeModel != GL_FLAT) return (swrast_line_func) NULL; - if (ctx->Line.StippleFlag) return (swrast_line_func) NULL; - if (swrast->_RasterMask & MULTI_DRAW_BIT) return (swrast_line_func) NULL; - if (xmbuf->swAlpha) return (swrast_line_func) NULL; - - xrb = xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped); - - if (xrb->ximage - && swrast->_RasterMask==DEPTH_BIT - && ctx->Depth.Func==GL_LESS - && ctx->Depth.Mask==GL_TRUE - && ctx->Visual.depthBits == DEFAULT_SOFTWARE_DEPTH_BITS - && ctx->Line.Width==1.0F) { - switch (xmesa->pixelformat) { - case PF_Truecolor: - return flat_TRUECOLOR_z_line; - case PF_8A8B8G8R: - return flat_8A8B8G8R_z_line; - case PF_8A8R8G8B: - return flat_8A8R8G8B_z_line; - case PF_8R8G8B: - return flat_8R8G8B_z_line; - case PF_8R8G8B24: - return flat_8R8G8B24_z_line; - case PF_5R6G5B: - return flat_5R6G5B_z_line; - case PF_Dither_5R6G5B: - return flat_DITHER_5R6G5B_z_line; - case PF_Dither: - return (depth==8) ? flat_DITHER8_z_line : (swrast_line_func) NULL; - case PF_Lookup: - return (depth==8) ? flat_LOOKUP8_z_line : (swrast_line_func) NULL; - case PF_HPCR: - return flat_HPCR_z_line; - default: - return (swrast_line_func)NULL; - } - } - if (xrb->ximage - && swrast->_RasterMask==0 - && ctx->Line.Width==1.0F) { - switch (xmesa->pixelformat) { - case PF_Truecolor: - return flat_TRUECOLOR_line; - case PF_8A8B8G8R: - return flat_8A8B8G8R_line; - case PF_8A8R8G8B: - return flat_8A8R8G8B_line; - case PF_8R8G8B: - return flat_8R8G8B_line; - case PF_8R8G8B24: - return flat_8R8G8B24_line; - case PF_5R6G5B: - return flat_5R6G5B_line; - case PF_Dither_5R6G5B: - return flat_DITHER_5R6G5B_line; - case PF_Dither: - return (depth==8) ? flat_DITHER8_line : (swrast_line_func) NULL; - case PF_Lookup: - return (depth==8) ? flat_LOOKUP8_line : (swrast_line_func) NULL; - case PF_HPCR: - return flat_HPCR_line; - default: - return (swrast_line_func)NULL; - } - } - -#ifndef XFree86Server - if (ctx->DrawBuffer->_NumColorDrawBuffers[0] == 1 - && ctx->DrawBuffer->_ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT - && swrast->_RasterMask == LOGIC_OP_BIT - && ctx->Color.LogicOp == GL_XOR - && !ctx->Line.StippleFlag - && !ctx->Line.SmoothFlag) { - return xor_line; - } -#endif /* XFree86Server */ - -#endif /* CHAN_BITS == 8 */ - return (swrast_line_func) NULL; -} - - -/** - * Override for the swrast line-selection function. Try to use one - * of our internal line functions, otherwise fall back to the - * standard swrast functions. - */ -void -xmesa_choose_line(GLcontext *ctx) -{ - SWcontext *swrast = SWRAST_CONTEXT(ctx); - - if (!(swrast->Line = get_line_func( ctx ))) - _swrast_choose_line( ctx ); -} diff --git a/src/mesa/pipe/xlib/xm_tri.c b/src/mesa/pipe/xlib/xm_tri.c deleted file mode 100644 index 9f17083f90..0000000000 --- a/src/mesa/pipe/xlib/xm_tri.c +++ /dev/null @@ -1,1694 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/* - * This file contains "accelerated" triangle functions. It should be - * fairly easy to write new special-purpose triangle functions and hook - * them into this module. - */ - - -#include "glxheader.h" -#include "depth.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" -#include "xmesaP.h" - -/* Internal swrast includes: - */ -#include "swrast/s_context.h" -#include "swrast/s_depth.h" -#include "swrast/s_triangle.h" - - -#define GET_XRB(XRB) struct xmesa_renderbuffer *XRB = \ - xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped) - - -/**********************************************************************/ -/*** Triangle rendering ***/ -/**********************************************************************/ - - -#if CHAN_BITS == 8 - -/* - * XImage, smooth, depth-buffered, PF_TRUECOLOR triangle. - */ -#define NAME smooth_TRUECOLOR_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); - -#define RENDER_SPAN( span ) \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - GLuint i; \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - unsigned long p; \ - PACK_TRUECOLOR(p, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - XMesaPutPixel(xrb->ximage, x, y, p); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } - -#include "swrast/s_tritemp.h" - - - - -/* - * XImage, smooth, depth-buffered, PF_8A8B8G8R triangle. - */ -#define NAME smooth_8A8B8G8R_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define INTERP_ALPHA 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = PACK_8A8B8G8R(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue), \ - FixedToInt(span.alpha)); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.alpha += span.alphaStep; \ - span.z += span.zStep; \ - } - -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, PF_8A8R8G8B triangle. - */ -#define NAME smooth_8A8R8G8B_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define INTERP_ALPHA 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); - -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = PACK_8A8R8G8B(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue), \ - FixedToInt(span.alpha)); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.alpha += span.alphaStep; \ - span.z += span.zStep; \ - } - -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, PF_8R8G8B triangle. - */ -#define NAME smooth_8R8G8B_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); - -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = PACK_8R8G8B(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } - -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, PF_8R8G8B24 triangle. - */ -#define NAME smooth_8R8G8B24_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) -#define PIXEL_TYPE bgr_t -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - PIXEL_TYPE *ptr = pRow + i; \ - ptr->r = FixedToInt(span.red); \ - ptr->g = FixedToInt(span.green); \ - ptr->b = FixedToInt(span.blue); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, PF_TRUEDITHER triangle. - */ -#define NAME smooth_TRUEDITHER_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - unsigned long p; \ - PACK_TRUEDITHER(p, x, y, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - XMesaPutPixel(xrb->ximage, x, y, p); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, PF_5R6G5B triangle. - */ -#define NAME smooth_5R6G5B_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = PACK_5R6G5B(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, PF_DITHER_5R6G5B triangle. - */ -#define NAME smooth_DITHER_5R6G5B_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - PACK_TRUEDITHER(pRow[i], x, y, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, 8-bit, PF_DITHER8 triangle. - */ -#define NAME smooth_DITHER8_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - XDITHER_SETUP(y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = (PIXEL_TYPE) XDITHER(x, FixedToInt(span.red),\ - FixedToInt(span.green), FixedToInt(span.blue) ); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, PF_DITHER triangle. - */ -#define NAME smooth_DITHER_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define SETUP_CODE \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - XDITHER_SETUP(y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - unsigned long p = XDITHER(x, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - XMesaPutPixel(img, x, y, p); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, 8-bit PF_LOOKUP triangle. - */ -#define NAME smooth_LOOKUP8_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - LOOKUP_SETUP; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = LOOKUP(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, depth-buffered, 8-bit PF_HPCR triangle. - */ -#define NAME smooth_HPCR_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = DITHER_HPCR(x, y, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue) ); \ - zRow[i] = z; \ - } \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, PF_TRUECOLOR triangle. - */ -#define NAME flat_TRUECOLOR_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; \ - unsigned long pixel; \ - PACK_TRUECOLOR(pixel, v2->color[0], v2->color[1], v2->color[2]); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - XMesaPutPixel(img, x, y, pixel); \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, PF_8A8B8G8R triangle. - */ -#define NAME flat_8A8B8G8R_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - GLuint p = PACK_8A8B8G8R( v2->color[0], v2->color[1],\ - v2->color[2], v2->color[3]); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = (PIXEL_TYPE) p; \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, PF_8A8R8G8B triangle. - */ -#define NAME flat_8A8R8G8B_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - GLuint p = PACK_8A8R8G8B(v2->color[0], v2->color[1], \ - v2->color[2], v2->color[3]); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = (PIXEL_TYPE) p; \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, PF_8R8G8B triangle. - */ -#define NAME flat_8R8G8B_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - GLuint p = PACK_8R8G8B( v2->color[0], v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = (PIXEL_TYPE) p; \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } - -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, PF_8R8G8B24 triangle. - */ -#define NAME flat_8R8G8B24_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) -#define PIXEL_TYPE bgr_t -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = v2->color; -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - PIXEL_TYPE *ptr = pRow + i; \ - ptr->r = color[RCOMP]; \ - ptr->g = color[GCOMP]; \ - ptr->b = color[BCOMP]; \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, PF_TRUEDITHER triangle. - */ -#define NAME flat_TRUEDITHER_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - unsigned long p; \ - PACK_TRUEDITHER(p, x, y, v2->color[0], \ - v2->color[1], v2->color[2]); \ - XMesaPutPixel(img, x, y, p); \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, PF_5R6G5B triangle. - */ -#define NAME flat_5R6G5B_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - GLushort p = PACK_5R6G5B( v2->color[0], v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = (PIXEL_TYPE) p; \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, PF_DITHER_5R6G5B triangle. - */ -#define NAME flat_DITHER_5R6G5B_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - const GLubyte *color = v2->color; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - PACK_TRUEDITHER(pRow[i], x, y, color[RCOMP], \ - color[GCOMP], color[BCOMP]); \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, 8-bit PF_DITHER triangle. - */ -#define NAME flat_DITHER8_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - FLAT_DITHER_SETUP( v2->color[0], v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - FLAT_DITHER_ROW_SETUP(YFLIP(xrb, y)); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = (PIXEL_TYPE) FLAT_DITHER(x); \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, PF_DITHER triangle. - */ -#define NAME flat_DITHER_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define SETUP_CODE \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; \ - FLAT_DITHER_SETUP( v2->color[0], v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - FLAT_DITHER_ROW_SETUP(y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - unsigned long p = FLAT_DITHER(x); \ - XMesaPutPixel(img, x, y, p); \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, 8-bit PF_HPCR triangle. - */ -#define NAME flat_HPCR_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - GLubyte r = v2->color[0]; \ - GLubyte g = v2->color[1]; \ - GLubyte b = v2->color[2]; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = (PIXEL_TYPE) DITHER_HPCR(x, y, r, g, b); \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, depth-buffered, 8-bit PF_LOOKUP triangle. - */ -#define NAME flat_LOOKUP8_z_triangle -#define INTERP_Z 1 -#define DEPTH_TYPE DEFAULT_SOFTWARE_DEPTH_TYPE -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - LOOKUP_SETUP; \ - GLubyte r = v2->color[0]; \ - GLubyte g = v2->color[1]; \ - GLubyte b = v2->color[2]; \ - GLubyte p = LOOKUP(r,g,b); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - const DEPTH_TYPE z = FixedToDepth(span.z); \ - if (z < zRow[i]) { \ - pRow[i] = p; \ - zRow[i] = z; \ - } \ - span.z += span.zStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, PF_TRUECOLOR triangle. - */ -#define NAME smooth_TRUECOLOR_triangle -#define INTERP_RGB 1 -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - unsigned long p; \ - PACK_TRUECOLOR(p, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - XMesaPutPixel(img, x, y, p); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, PF_8A8B8G8R triangle. - */ -#define NAME smooth_8A8B8G8R_triangle -#define INTERP_RGB 1 -#define INTERP_ALPHA 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = PACK_8A8B8G8R(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue), \ - FixedToInt(span.alpha)); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.alpha += span.alphaStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, PF_8A8R8G8B triangle. - */ -#define NAME smooth_8A8R8G8B_triangle -#define INTERP_RGB 1 -#define INTERP_ALPHA 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = PACK_8A8R8G8B(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue), \ - FixedToInt(span.alpha)); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - span.alpha += span.alphaStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, PF_8R8G8B triangle. - */ -#define NAME smooth_8R8G8B_triangle -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = PACK_8R8G8B(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue) ); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, PF_8R8G8B triangle. - */ -#define NAME smooth_8R8G8B24_triangle -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) -#define PIXEL_TYPE bgr_t -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - PIXEL_TYPE *pixel = pRow; \ - for (i = 0; i < span.end; i++, pixel++) { \ - pixel->r = FixedToInt(span.red); \ - pixel->g = FixedToInt(span.green); \ - pixel->b = FixedToInt(span.blue); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, PF_TRUEDITHER triangle. - */ -#define NAME smooth_TRUEDITHER_triangle -#define INTERP_RGB 1 -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - unsigned long p; \ - PACK_TRUEDITHER(p, x, y, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - XMesaPutPixel(img, x, y, p ); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, PF_5R6G5B triangle. - */ -#define NAME smooth_5R6G5B_triangle -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = (PIXEL_TYPE) PACK_5R6G5B(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, PF_DITHER_5R6G5B triangle. - */ -#define NAME smooth_DITHER_5R6G5B_triangle -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - PACK_TRUEDITHER(pRow[i], x, y, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, 8-bit PF_DITHER triangle. - */ -#define NAME smooth_DITHER8_triangle -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - XDITHER_SETUP(y); \ - for (i = 0; i < span.end; i++, x++) { \ - pRow[i] = (PIXEL_TYPE) XDITHER(x, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue) ); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, PF_DITHER triangle. - */ -#define NAME smooth_DITHER_triangle -#define INTERP_RGB 1 -#define SETUP_CODE \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - XDITHER_SETUP(y); \ - for (i = 0; i < span.end; i++, x++) { \ - unsigned long p = XDITHER(x, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue) ); \ - XMesaPutPixel(img, x, y, p); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, 8-bit PF_LOOKUP triangle. - */ -#define NAME smooth_LOOKUP8_triangle -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - LOOKUP_SETUP; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = LOOKUP(FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, smooth, NON-depth-buffered, 8-bit PF_HPCR triangle. - */ -#define NAME smooth_HPCR_triangle -#define INTERP_RGB 1 -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - pRow[i] = DITHER_HPCR(x, y, FixedToInt(span.red), \ - FixedToInt(span.green), FixedToInt(span.blue)); \ - span.red += span.redStep; \ - span.green += span.greenStep; \ - span.blue += span.blueStep; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, PF_TRUECOLOR triangle. - */ -#define NAME flat_TRUECOLOR_triangle -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; \ - unsigned long pixel; \ - PACK_TRUECOLOR(pixel, v2->color[0], v2->color[1], v2->color[2]); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - XMesaPutPixel(img, x, y, pixel); \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, PF_8A8B8G8R triangle. - */ -#define NAME flat_8A8B8G8R_triangle -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - unsigned long p = PACK_8B8G8R( v2->color[0], \ - v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = (PIXEL_TYPE) p; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, PF_8A8R8G8B triangle. - */ -#define NAME flat_8A8R8G8B_triangle -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - unsigned long p = PACK_8R8G8B( v2->color[0], \ - v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = (PIXEL_TYPE) p; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, PF_8R8G8B triangle. - */ -#define NAME flat_8R8G8B_triangle -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR4(xrb, X, Y) -#define PIXEL_TYPE GLuint -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - unsigned long p = PACK_8R8G8B( v2->color[0], \ - v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = (PIXEL_TYPE) p; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, PF_8R8G8B24 triangle. - */ -#define NAME flat_8R8G8B24_triangle -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR3(xrb, X, Y) -#define PIXEL_TYPE bgr_t -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - const GLubyte *color = v2->color; -#define RENDER_SPAN( span ) \ - GLuint i; \ - PIXEL_TYPE *pixel = pRow; \ - for (i = 0; i < span.end; i++, pixel++) { \ - pixel->r = color[RCOMP]; \ - pixel->g = color[GCOMP]; \ - pixel->b = color[BCOMP]; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, PF_TRUEDITHER triangle. - */ -#define NAME flat_TRUEDITHER_triangle -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - unsigned long p; \ - PACK_TRUEDITHER(p, x, y, v2->color[0], \ - v2->color[1], v2->color[2] ); \ - XMesaPutPixel(img, x, y, p); \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, PF_5R6G5B triangle. - */ -#define NAME flat_5R6G5B_triangle -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - unsigned long p = PACK_5R6G5B( v2->color[0], \ - v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = (PIXEL_TYPE) p; \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, PF_DITHER_5R6G5B triangle. - */ -#define NAME flat_DITHER_5R6G5B_triangle -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR2(xrb, X, Y) -#define PIXEL_TYPE GLushort -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - const GLubyte *color = v2->color; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - PACK_TRUEDITHER(pRow[i], x, y, color[RCOMP], \ - color[GCOMP], color[BCOMP]); \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, 8-bit PF_DITHER triangle. - */ -#define NAME flat_DITHER8_triangle -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - FLAT_DITHER_SETUP( v2->color[0], v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - FLAT_DITHER_ROW_SETUP(YFLIP(xrb, y)); \ - for (i = 0; i < span.end; i++, x++) { \ - pRow[i] = (PIXEL_TYPE) FLAT_DITHER(x); \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, PF_DITHER triangle. - */ -#define NAME flat_DITHER_triangle -#define SETUP_CODE \ - GET_XRB(xrb); \ - XMesaImage *img = xrb->ximage; \ - FLAT_DITHER_SETUP( v2->color[0], v2->color[1], v2->color[2] ); -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - FLAT_DITHER_ROW_SETUP(y); \ - for (i = 0; i < span.end; i++, x++) { \ - unsigned long p = FLAT_DITHER(x); \ - XMesaPutPixel(img, x, y, p ); \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, 8-bit PF_HPCR triangle. - */ -#define NAME flat_HPCR_triangle -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - XMesaContext xmesa = XMESA_CONTEXT(ctx); \ - GET_XRB(xrb); \ - GLubyte r = v2->color[0]; \ - GLubyte g = v2->color[1]; \ - GLubyte b = v2->color[2]; -#define RENDER_SPAN( span ) \ - GLuint i; \ - GLint x = span.x, y = YFLIP(xrb, span.y); \ - for (i = 0; i < span.end; i++, x++) { \ - pRow[i] = (PIXEL_TYPE) DITHER_HPCR(x, y, r, g, b); \ - } -#include "swrast/s_tritemp.h" - - - -/* - * XImage, flat, NON-depth-buffered, 8-bit PF_LOOKUP triangle. - */ -#define NAME flat_LOOKUP8_triangle -#define PIXEL_ADDRESS(X,Y) PIXEL_ADDR1(xrb, X, Y) -#define PIXEL_TYPE GLubyte -#define BYTES_PER_ROW (xrb->ximage->bytes_per_line) -#define SETUP_CODE \ - GET_XRB(xrb); \ - LOOKUP_SETUP; \ - GLubyte r = v2->color[0]; \ - GLubyte g = v2->color[1]; \ - GLubyte b = v2->color[2]; \ - GLubyte p = LOOKUP(r,g,b); -#define RENDER_SPAN( span ) \ - GLuint i; \ - for (i = 0; i < span.end; i++) { \ - pRow[i] = (PIXEL_TYPE) p; \ - } -#include "swrast/s_tritemp.h" - - -#endif /* CHAN_BITS == 8 */ - - -#if defined(DEBUG) && CHAN_BITS == 8 -extern void _xmesa_print_triangle_func( swrast_tri_func triFunc ); -void _xmesa_print_triangle_func( swrast_tri_func triFunc ) -{ - _mesa_printf("XMesa tri func = "); - if (triFunc ==smooth_TRUECOLOR_z_triangle) - _mesa_printf("smooth_TRUECOLOR_z_triangle\n"); - else if (triFunc ==smooth_8A8B8G8R_z_triangle) - _mesa_printf("smooth_8A8B8G8R_z_triangle\n"); - else if (triFunc ==smooth_8A8R8G8B_z_triangle) - _mesa_printf("smooth_8A8R8G8B_z_triangle\n"); - else if (triFunc ==smooth_8R8G8B_z_triangle) - _mesa_printf("smooth_8R8G8B_z_triangle\n"); - else if (triFunc ==smooth_8R8G8B24_z_triangle) - _mesa_printf("smooth_8R8G8B24_z_triangle\n"); - else if (triFunc ==smooth_TRUEDITHER_z_triangle) - _mesa_printf("smooth_TRUEDITHER_z_triangle\n"); - else if (triFunc ==smooth_5R6G5B_z_triangle) - _mesa_printf("smooth_5R6G5B_z_triangle\n"); - else if (triFunc ==smooth_DITHER_5R6G5B_z_triangle) - _mesa_printf("smooth_DITHER_5R6G5B_z_triangle\n"); - else if (triFunc ==smooth_HPCR_z_triangle) - _mesa_printf("smooth_HPCR_z_triangle\n"); - else if (triFunc ==smooth_DITHER8_z_triangle) - _mesa_printf("smooth_DITHER8_z_triangle\n"); - else if (triFunc ==smooth_LOOKUP8_z_triangle) - _mesa_printf("smooth_LOOKUP8_z_triangle\n"); - else if (triFunc ==flat_TRUECOLOR_z_triangle) - _mesa_printf("flat_TRUECOLOR_z_triangle\n"); - else if (triFunc ==flat_8A8B8G8R_z_triangle) - _mesa_printf("flat_8A8B8G8R_z_triangle\n"); - else if (triFunc ==flat_8A8R8G8B_z_triangle) - _mesa_printf("flat_8A8R8G8B_z_triangle\n"); - else if (triFunc ==flat_8R8G8B_z_triangle) - _mesa_printf("flat_8R8G8B_z_triangle\n"); - else if (triFunc ==flat_8R8G8B24_z_triangle) - _mesa_printf("flat_8R8G8B24_z_triangle\n"); - else if (triFunc ==flat_TRUEDITHER_z_triangle) - _mesa_printf("flat_TRUEDITHER_z_triangle\n"); - else if (triFunc ==flat_5R6G5B_z_triangle) - _mesa_printf("flat_5R6G5B_z_triangle\n"); - else if (triFunc ==flat_DITHER_5R6G5B_z_triangle) - _mesa_printf("flat_DITHER_5R6G5B_z_triangle\n"); - else if (triFunc ==flat_HPCR_z_triangle) - _mesa_printf("flat_HPCR_z_triangle\n"); - else if (triFunc ==flat_DITHER8_z_triangle) - _mesa_printf("flat_DITHER8_z_triangle\n"); - else if (triFunc ==flat_LOOKUP8_z_triangle) - _mesa_printf("flat_LOOKUP8_z_triangle\n"); - else if (triFunc ==smooth_TRUECOLOR_triangle) - _mesa_printf("smooth_TRUECOLOR_triangle\n"); - else if (triFunc ==smooth_8A8B8G8R_triangle) - _mesa_printf("smooth_8A8B8G8R_triangle\n"); - else if (triFunc ==smooth_8A8R8G8B_triangle) - _mesa_printf("smooth_8A8R8G8B_triangle\n"); - else if (triFunc ==smooth_8R8G8B_triangle) - _mesa_printf("smooth_8R8G8B_triangle\n"); - else if (triFunc ==smooth_8R8G8B24_triangle) - _mesa_printf("smooth_8R8G8B24_triangle\n"); - else if (triFunc ==smooth_TRUEDITHER_triangle) - _mesa_printf("smooth_TRUEDITHER_triangle\n"); - else if (triFunc ==smooth_5R6G5B_triangle) - _mesa_printf("smooth_5R6G5B_triangle\n"); - else if (triFunc ==smooth_DITHER_5R6G5B_triangle) - _mesa_printf("smooth_DITHER_5R6G5B_triangle\n"); - else if (triFunc ==smooth_HPCR_triangle) - _mesa_printf("smooth_HPCR_triangle\n"); - else if (triFunc ==smooth_DITHER8_triangle) - _mesa_printf("smooth_DITHER8_triangle\n"); - else if (triFunc ==smooth_LOOKUP8_triangle) - _mesa_printf("smooth_LOOKUP8_triangle\n"); - else if (triFunc ==flat_TRUECOLOR_triangle) - _mesa_printf("flat_TRUECOLOR_triangle\n"); - else if (triFunc ==flat_TRUEDITHER_triangle) - _mesa_printf("flat_TRUEDITHER_triangle\n"); - else if (triFunc ==flat_8A8B8G8R_triangle) - _mesa_printf("flat_8A8B8G8R_triangle\n"); - else if (triFunc ==flat_8A8R8G8B_triangle) - _mesa_printf("flat_8A8R8G8B_triangle\n"); - else if (triFunc ==flat_8R8G8B_triangle) - _mesa_printf("flat_8R8G8B_triangle\n"); - else if (triFunc ==flat_8R8G8B24_triangle) - _mesa_printf("flat_8R8G8B24_triangle\n"); - else if (triFunc ==flat_5R6G5B_triangle) - _mesa_printf("flat_5R6G5B_triangle\n"); - else if (triFunc ==flat_DITHER_5R6G5B_triangle) - _mesa_printf("flat_DITHER_5R6G5B_triangle\n"); - else if (triFunc ==flat_HPCR_triangle) - _mesa_printf("flat_HPCR_triangle\n"); - else if (triFunc ==flat_DITHER8_triangle) - _mesa_printf("flat_DITHER8_triangle\n"); - else if (triFunc ==flat_LOOKUP8_triangle) - _mesa_printf("flat_LOOKUP8_triangle\n"); - else - _mesa_printf("???\n"); -} -#endif - - -#ifdef DEBUG - -/* record the current triangle function name */ -static const char *triFuncName = NULL; - -#define USE(triFunc) \ -do { \ - triFuncName = #triFunc; \ - return triFunc; \ -} while (0) - -#else - -#define USE(triFunc) return triFunc - -#endif - - -#if 0 -GLboolean xmesa_get_cbuf_details( GLcontext *ctx, - void **ptr, - GLuint *cpp, - GLint *stride, - GLuint *format ) -{ - XMesaContext xmesa = XMESA_CONTEXT(ctx); - struct gl_framebuffer *fb = ctx->DrawBuffer; - struct gl_renderbuffer *crb = fb->_ColorDrawBuffers[0][0]; - struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(crb->Wrapped); - - *ptr = crb->GetPointer(ctx, crb, 0, 0); - *stride = ((GLubyte *)crb->GetPointer(ctx, crb, 0, 1) - - (GLubyte *)crb->GetPointer(ctx, crb, 0, 0)); - - if (!ptr) - goto bad; - - switch (xmesa->pixelformat) { - case PF_8A8B8G8R: - case PF_8A8R8G8B: - *format = 1; /* whatever */ - *cpp = 4; - break; - default: - goto bad; - } - - return GL_TRUE; - - bad: - *ptr = NULL; - *stride = 0; - *format = 0; - return GL_FALSE; -} -#endif - - -/** - * Return pointer to line drawing function, or NULL if we should use a - * swrast fallback. - */ -static swrast_tri_func -get_triangle_func(GLcontext *ctx) -{ -#if CHAN_BITS == 8 - SWcontext *swrast = SWRAST_CONTEXT(ctx); - XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - const int depth = GET_VISUAL_DEPTH(xmesa->xm_visual); - const struct xmesa_renderbuffer *xrb; - -#ifdef DEBUG - triFuncName = NULL; -#endif - - /* trivial fallback tests */ - if ((ctx->DrawBuffer->_ColorDrawBufferMask[0] - & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT)) == 0) - return (swrast_tri_func) NULL; - if (ctx->RenderMode != GL_RENDER) - return (swrast_tri_func) NULL; - if (ctx->Polygon.SmoothFlag) - return (swrast_tri_func) NULL; - if (ctx->Texture._EnabledUnits) - return (swrast_tri_func) NULL; - if (swrast->_RasterMask & MULTI_DRAW_BIT) - return (swrast_tri_func) NULL; - if (ctx->Polygon.CullFlag && - ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) - return (swrast_tri_func) NULL; - if (xmbuf->swAlpha) - return (swrast_tri_func) NULL; - - xrb = xmesa_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]->Wrapped); - - if (xrb->ximage) { - if ( ctx->Light.ShadeModel==GL_SMOOTH - && swrast->_RasterMask==DEPTH_BIT - && ctx->Depth.Func==GL_LESS - && ctx->Depth.Mask==GL_TRUE - && ctx->Visual.depthBits == DEFAULT_SOFTWARE_DEPTH_BITS - && ctx->Polygon.StippleFlag==GL_FALSE) { - switch (xmesa->pixelformat) { - case PF_Truecolor: - USE(smooth_TRUECOLOR_z_triangle); - case PF_8A8B8G8R: - USE(smooth_8A8B8G8R_z_triangle); - case PF_8A8R8G8B: - USE(smooth_8A8R8G8B_z_triangle); - case PF_8R8G8B: - USE(smooth_8R8G8B_z_triangle); - case PF_8R8G8B24: - USE(smooth_8R8G8B24_z_triangle); - case PF_Dither_True: - USE(smooth_TRUEDITHER_z_triangle); - case PF_5R6G5B: - USE(smooth_5R6G5B_z_triangle); - case PF_Dither_5R6G5B: - USE(smooth_DITHER_5R6G5B_z_triangle); - case PF_HPCR: - USE(smooth_HPCR_z_triangle); - case PF_Dither: - if (depth == 8) - USE(smooth_DITHER8_z_triangle); - else - USE(smooth_DITHER_z_triangle); - case PF_Lookup: - if (depth == 8) - USE(smooth_LOOKUP8_z_triangle); - else - return (swrast_tri_func) NULL; - default: - return (swrast_tri_func) NULL; - } - } - if ( ctx->Light.ShadeModel==GL_FLAT - && swrast->_RasterMask==DEPTH_BIT - && ctx->Depth.Func==GL_LESS - && ctx->Depth.Mask==GL_TRUE - && ctx->Visual.depthBits == DEFAULT_SOFTWARE_DEPTH_BITS - && ctx->Polygon.StippleFlag==GL_FALSE) { - switch (xmesa->pixelformat) { - case PF_Truecolor: - USE(flat_TRUECOLOR_z_triangle); - case PF_8A8B8G8R: - USE(flat_8A8B8G8R_z_triangle); - case PF_8A8R8G8B: - USE(flat_8A8R8G8B_z_triangle); - case PF_8R8G8B: - USE(flat_8R8G8B_z_triangle); - case PF_8R8G8B24: - USE(flat_8R8G8B24_z_triangle); - case PF_Dither_True: - USE(flat_TRUEDITHER_z_triangle); - case PF_5R6G5B: - USE(flat_5R6G5B_z_triangle); - case PF_Dither_5R6G5B: - USE(flat_DITHER_5R6G5B_z_triangle); - case PF_HPCR: - USE(flat_HPCR_z_triangle); - case PF_Dither: - if (depth == 8) - USE(flat_DITHER8_z_triangle); - else - USE(flat_DITHER_z_triangle); - case PF_Lookup: - if (depth == 8) - USE(flat_LOOKUP8_z_triangle); - else - return (swrast_tri_func) NULL; - default: - return (swrast_tri_func) NULL; - } - } - if ( swrast->_RasterMask==0 /* no depth test */ - && ctx->Light.ShadeModel==GL_SMOOTH - && ctx->Polygon.StippleFlag==GL_FALSE) { - switch (xmesa->pixelformat) { - case PF_Truecolor: - USE(smooth_TRUECOLOR_triangle); - case PF_8A8B8G8R: - USE(smooth_8A8B8G8R_triangle); - case PF_8A8R8G8B: - USE(smooth_8A8R8G8B_triangle); - case PF_8R8G8B: - USE(smooth_8R8G8B_triangle); - case PF_8R8G8B24: - USE(smooth_8R8G8B24_triangle); - case PF_Dither_True: - USE(smooth_TRUEDITHER_triangle); - case PF_5R6G5B: - USE(smooth_5R6G5B_triangle); - case PF_Dither_5R6G5B: - USE(smooth_DITHER_5R6G5B_triangle); - case PF_HPCR: - USE(smooth_HPCR_triangle); - case PF_Dither: - if (depth == 8) - USE(smooth_DITHER8_triangle); - else - USE(smooth_DITHER_triangle); - case PF_Lookup: - if (depth == 8) - USE(smooth_LOOKUP8_triangle); - else - return (swrast_tri_func) NULL; - default: - return (swrast_tri_func) NULL; - } - } - - if ( swrast->_RasterMask==0 /* no depth test */ - && ctx->Light.ShadeModel==GL_FLAT - && ctx->Polygon.StippleFlag==GL_FALSE) { - switch (xmesa->pixelformat) { - case PF_Truecolor: - USE(flat_TRUECOLOR_triangle); - case PF_Dither_True: - USE(flat_TRUEDITHER_triangle); - case PF_8A8B8G8R: - USE(flat_8A8B8G8R_triangle); - case PF_8A8R8G8B: - USE(flat_8A8R8G8B_triangle); - case PF_8R8G8B: - USE(flat_8R8G8B_triangle); - case PF_8R8G8B24: - USE(flat_8R8G8B24_triangle); - case PF_5R6G5B: - USE(flat_5R6G5B_triangle); - case PF_Dither_5R6G5B: - USE(flat_DITHER_5R6G5B_triangle); - case PF_HPCR: - USE(flat_HPCR_triangle); - case PF_Dither: - if (depth == 8) - USE(flat_DITHER8_triangle); - else - USE(flat_DITHER_triangle); - case PF_Lookup: - if (depth == 8) - USE(flat_LOOKUP8_triangle); - else - return (swrast_tri_func) NULL; - default: - return (swrast_tri_func) NULL; - } - } - } -#endif /* CHAN_BITS == 8 */ - - return (swrast_tri_func) NULL; -} - - -/* Override for the swrast tri-selection function. Try to use one - * of our internal tri functions, otherwise fall back to the - * standard swrast functions. - */ -void xmesa_choose_triangle( GLcontext *ctx ) -{ - SWcontext *swrast = SWRAST_CONTEXT(ctx); - - if (!(swrast->Triangle = get_triangle_func( ctx ))) - _swrast_choose_triangle( ctx ); -} - -- cgit v1.2.3 From 7398272d4b43a42a57fd28a68194c5f6058696b3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:32:16 -0600 Subject: remove tnl/swrast code --- src/mesa/pipe/xlib/xm_api.c | 49 -------------------------------------------- src/mesa/pipe/xlib/xm_span.c | 1 - src/mesa/pipe/xlib/xmesaP.h | 11 ---------- 3 files changed, 61 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index d13d98857a..16ec304a19 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -73,16 +73,7 @@ #include "macros.h" #include "renderbuffer.h" #include "teximage.h" -#if 0 -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#endif #include "vbo/vbo.h" -#if 0 -#include "tnl/tnl.h" -#include "tnl/t_context.h" -#include "tnl/t_pipeline.h" -#endif #include "drivers/common/driverfuncs.h" #include "state_tracker/st_public.h" @@ -1597,41 +1588,9 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) c->display = v->display; c->pixelformat = v->dithered_pf; /* Dithering is enabled by default */ - /* Initialize the software rasterizer and helper modules. - */ -#if 0 - if (!_swrast_CreateContext( mesaCtx ) -#if 0 - || !_vbo_CreateContext( mesaCtx ) || - !_tnl_CreateContext( mesaCtx ) || - !_swsetup_CreateContext( mesaCtx ) -#endif - ) { - _mesa_free_context_data(&c->mesa); - _mesa_free(c); - return NULL; - } -#endif - -#if 0 - /* tnl setup */ - tnl = TNL_CONTEXT(mesaCtx); - tnl->Driver.RunPipeline = _tnl_run_pipeline; -#endif - -#if 0 - /* swrast setup */ - xmesa_register_swrast_functions( mesaCtx ); -#endif - st_create_context( mesaCtx, xmesa_create_softpipe( c ) ); -#if 0 - _swsetup_CreateContext( mesaCtx ); - _swsetup_Wakeup(mesaCtx); -#endif - /* override these functions, as if the xlib driver were derived from * the softpipe driver. */ @@ -1661,14 +1620,6 @@ void XMesaDestroyContext( XMesaContext c ) FXdestroyContext( XMESA_BUFFER(mesaCtx->DrawBuffer) ); #endif -#if 0 - _swsetup_DestroyContext( mesaCtx ); - _swrast_DestroyContext( mesaCtx ); -#endif -#if 0 - _tnl_DestroyContext( mesaCtx ); - _vbo_DestroyContext( mesaCtx ); -#endif _mesa_free_context_data( mesaCtx ); _mesa_free( c ); } diff --git a/src/mesa/pipe/xlib/xm_span.c b/src/mesa/pipe/xlib/xm_span.c index ce54a18a27..89eeff5d9d 100644 --- a/src/mesa/pipe/xlib/xm_span.c +++ b/src/mesa/pipe/xlib/xm_span.c @@ -34,7 +34,6 @@ #include "state.h" #include "xmesaP.h" -#include "swrast/swrast.h" /* diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index 5796fddb16..b7c510280b 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -569,17 +569,6 @@ XMESA_BUFFER(GLframebuffer *b) } -/* Plugged into the software rasterizer. Try to use internal - * swrast-style point, line and triangle functions. - */ -extern void xmesa_choose_point( GLcontext *ctx ); -extern void xmesa_choose_line( GLcontext *ctx ); -extern void xmesa_choose_triangle( GLcontext *ctx ); - - -extern void xmesa_register_swrast_functions( GLcontext *ctx ); - - #define ENABLE_EXT_texure_compression_s3tc 0 /* SW texture compression */ -- cgit v1.2.3 From 4a13a550e0f0fc12072f117229230175af30de24 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:33:49 -0600 Subject: remove FX/glide stuff --- src/mesa/pipe/xlib/xm_api.c | 17 ----------------- src/mesa/pipe/xlib/xmesaP.h | 10 ---------- 2 files changed, 27 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 16ec304a19..878ee30be9 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1616,10 +1616,6 @@ void XMesaDestroyContext( XMesaContext c ) { GLcontext *mesaCtx = &c->mesa; -#ifdef FX - FXdestroyContext( XMESA_BUFFER(mesaCtx->DrawBuffer) ); -#endif - _mesa_free_context_data( mesaCtx ); _mesa_free( c ); } @@ -1879,11 +1875,6 @@ GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, c->xm_buffer = drawBuffer; -#ifdef FX - if (FXmakeCurrent( drawBuffer )) - return GL_TRUE; -#endif - /* Call this periodically to detect when the user has begun using * GL rendering from multiple threads. */ @@ -2038,10 +2029,6 @@ void XMesaSwapBuffers( XMesaBuffer b ) _mesa_notifySwapBuffers(ctx); if (b->db_mode) { -#ifdef FX - if (FXswapBuffers(b)) - return; -#endif if (b->backxrb->ximage) { /* Copy Ximage (back buf) from client memory to server window */ #if defined(USE_XSHM) && !defined(XFree86Server) @@ -2108,10 +2095,6 @@ void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height ) if (b->db_mode) { int yTop = b->mesa_buffer.Height - y - height; -#ifdef FX - if (FXswapBuffers(b)) - return; -#endif if (b->backxrb->ximage) { /* Copy Ximage from host's memory to server's window */ #if defined(USE_XSHM) && !defined(XFree86Server) diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index b7c510280b..9e0c79a139 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -29,9 +29,6 @@ #include "GL/xmesa.h" #include "mtypes.h" -#if defined(FX) -#include "GL/fxmesa.h" -#endif #ifdef XFree86Server #include "xm_image.h" #endif @@ -268,13 +265,6 @@ struct xmesa_buffer { unsigned long alloced_colors[256]; #endif -#if defined( FX ) - /* For 3Dfx Glide only */ - GLboolean FXisHackUsable; /* Can we render into window? */ - GLboolean FXwindowHack; /* Are we rendering into a window? */ - fxMesaContext FXctx; -#endif - /* GLX_EXT_texture_from_pixmap */ GLint TextureTarget; /** GLX_TEXTURE_1D_EXT, for example */ GLint TextureFormat; /** GLX_TEXTURE_FORMAT_RGB_EXT, for example */ -- cgit v1.2.3 From f36372f0b27bdb7f99c3e5615eaf156b492dbd1d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:49:46 -0600 Subject: Start removing old driver funcs like ColorMask, ClearColor, etc. --- src/mesa/pipe/xlib/xm_api.c | 3 - src/mesa/pipe/xlib/xm_dd.c | 151 ++++++++-------------------------------- src/mesa/pipe/xlib/xm_surface.c | 4 +- src/mesa/pipe/xlib/xmesaP.h | 5 +- 4 files changed, 34 insertions(+), 129 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 878ee30be9..eff84f12a7 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1523,9 +1523,6 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) XMesaContext c; GLcontext *mesaCtx; struct dd_function_table functions; -#if 0 - TNLcontext *tnl; -#endif if (firstTime) { _glthread_INIT_MUTEX(_xmesa_lock); diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 3c5c413c73..f6da1cb9f7 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -101,99 +101,10 @@ finish_or_flush( GLcontext *ctx ) _glthread_UNLOCK_MUTEX(_xmesa_lock); } #endif + abort(); } -static void -clear_index( GLcontext *ctx, GLuint index ) -{ - if (ctx->DrawBuffer->Name == 0) { - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - xmesa->clearpixel = (unsigned long) index; - XMesaSetForeground( xmesa->display, xmbuf->cleargc, (unsigned long) index ); - } -} - - -static void -clear_color( GLcontext *ctx, const GLfloat color[4] ) -{ - if (ctx->DrawBuffer->Name == 0) { - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[3], color[3]); - xmesa->clearpixel = xmesa_color_to_pixel( ctx, - xmesa->clearcolor[0], - xmesa->clearcolor[1], - xmesa->clearcolor[2], - xmesa->clearcolor[3], - xmesa->xm_visual->undithered_pf ); - _glthread_LOCK_MUTEX(_xmesa_lock); - XMesaSetForeground( xmesa->display, xmbuf->cleargc, - xmesa->clearpixel ); - _glthread_UNLOCK_MUTEX(_xmesa_lock); - } -} - - - -/* Set index mask ala glIndexMask */ -static void -index_mask( GLcontext *ctx, GLuint mask ) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - /* not sure this conditional is really needed */ - if (xmbuf->backxrb && xmbuf->backxrb->pixmap) { - unsigned long m; - if (mask==0xffffffff) { - m = ((unsigned long)~0L); - } - else { - m = (unsigned long) mask; - } - XMesaSetPlaneMask( xmesa->display, xmbuf->cleargc, m ); - } -} - - -/* Implements glColorMask() */ -static void -color_mask(GLcontext *ctx, - GLboolean rmask, GLboolean gmask, GLboolean bmask, GLboolean amask) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaBuffer xmbuf; - const int xclass = xmesa->xm_visual->mesa_visual.visualType; - (void) amask; - - if (ctx->DrawBuffer->Name != 0) - return; - - xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - - if (xclass == GLX_TRUE_COLOR || xclass == GLX_DIRECT_COLOR) { - unsigned long m; - if (rmask && gmask && bmask) { - m = ((unsigned long)~0L); - } - else { - m = 0; - if (rmask) m |= GET_REDMASK(xmesa->xm_visual); - if (gmask) m |= GET_GREENMASK(xmesa->xm_visual); - if (bmask) m |= GET_BLUEMASK(xmesa->xm_visual); - } - XMesaSetPlaneMask( xmesa->display, xmbuf->cleargc, m ); - } -} - - - /**********************************************************************/ /*** glClear implementations ***/ /**********************************************************************/ @@ -204,7 +115,7 @@ color_mask(GLcontext *ctx, */ static void clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height) + GLint x, GLint y, GLint width, GLint height, GLuint value) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); @@ -216,6 +127,8 @@ clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, assert(xrb->pixmap); assert(xmbuf->cleargc); + XMesaSetForeground( xmesa->display, xmbuf->cleargc, value ); + XMesaFillRectangle( xmesa->display, xrb->pixmap, xmbuf->cleargc, x, xrb->St.Base.Height - y - height, width, height ); @@ -224,7 +137,7 @@ clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, static void clear_8bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height ) + GLint x, GLint y, GLint width, GLint height, GLuint value ) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); GLint i; @@ -237,7 +150,7 @@ clear_8bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, static void clear_HPCR_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height ) + GLint x, GLint y, GLint width, GLint height, GLuint value ) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); GLint i; @@ -258,20 +171,19 @@ clear_HPCR_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, static void clear_16bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height) + GLint x, GLint y, GLint width, GLint height, GLuint value) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GLuint pixel = (GLuint) xmesa->clearpixel; GLint i, j; if (xmesa->swapbytes) { - pixel = ((pixel >> 8) & 0x00ff) | ((pixel << 8) & 0xff00); + value = ((value >> 8) & 0x00ff) | ((value << 8) & 0xff00); } for (j = 0; j < height; j++) { GLushort *ptr2 = PIXEL_ADDR2(xrb, x, y + j); for (i = 0; i < width; i++) { - ptr2[i] = pixel; + ptr2[i] = value; } } } @@ -280,12 +192,11 @@ clear_16bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, /* Optimized code provided by Nozomi Ytow */ static void clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height) + GLint x, GLint y, GLint width, GLint height, GLuint value) { - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - const GLubyte r = xmesa->clearcolor[0]; - const GLubyte g = xmesa->clearcolor[1]; - const GLubyte b = xmesa->clearcolor[2]; + const GLubyte r = (value ) & 0xff; + const GLubyte g = (value >> 8) & 0xff; + const GLubyte b = (value >> 16) & 0xff; if (r == g && g == b) { /* same value for all three components (gray) */ @@ -313,33 +224,32 @@ clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, static void clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height) + GLint x, GLint y, GLint width, GLint height, GLuint value) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); - register GLuint pixel = (GLuint) xmesa->clearpixel; if (!xrb->ximage) return; if (xmesa->swapbytes) { - pixel = ((pixel >> 24) & 0x000000ff) - | ((pixel >> 8) & 0x0000ff00) - | ((pixel << 8) & 0x00ff0000) - | ((pixel << 24) & 0xff000000); + value = ((value >> 24) & 0x000000ff) + | ((value >> 8) & 0x0000ff00) + | ((value << 8) & 0x00ff0000) + | ((value << 24) & 0xff000000); } if (width == xrb->St.Base.Width && height == xrb->St.Base.Height) { /* clearing whole buffer */ const GLuint n = xrb->St.Base.Width * xrb->St.Base.Height; GLuint *ptr4 = (GLuint *) xrb->ximage->data; - if (pixel == 0) { + if (value == 0) { /* common case */ - _mesa_memset(ptr4, pixel, 4 * n); + _mesa_memset(ptr4, value, 4 * n); } else { GLuint i; for (i = 0; i < n; i++) - ptr4[i] = pixel; + ptr4[i] = value; } } else { @@ -348,7 +258,7 @@ clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, for (j = 0; j < height; j++) { GLuint *ptr4 = PIXEL_ADDR4(xrb, x, y + j); for (i = 0; i < width; i++) { - ptr4[i] = pixel; + ptr4[i] = value; } } } @@ -357,9 +267,8 @@ clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, static void clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height) + GLint x, GLint y, GLint width, GLint height, GLuint value) { - const XMesaContext xmesa = XMESA_CONTEXT(ctx); XMesaImage *img = xrb->ximage; GLint i, j; @@ -367,7 +276,7 @@ clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, y = YFLIP(xrb, y); for (j = 0; j < height; j++) { for (i = 0; i < width; i++) { - XMesaPutPixel(img, x+i, y-j, xmesa->clearpixel); + XMesaPutPixel(img, x+i, y-j, value); } } } @@ -375,7 +284,7 @@ clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, void -xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers) +xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers, GLuint value) { if (ctx->DrawBuffer->Name == 0) { /* this is a window system framebuffer */ @@ -394,7 +303,7 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers) = ctx->DrawBuffer->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; if (b->frontxrb == xmesa_renderbuffer(frontRb)) { /* renderbuffer is not wrapped - great! */ - b->frontxrb->clearFunc(ctx, b->frontxrb, x, y, width, height); + b->frontxrb->clearFunc(ctx, b->frontxrb, x, y, width, height, value); buffers &= ~BUFFER_BIT_FRONT_LEFT; } else { @@ -407,7 +316,7 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers) = ctx->DrawBuffer->Attachment[BUFFER_BACK_LEFT].Renderbuffer; if (b->backxrb == xmesa_renderbuffer(backRb)) { /* renderbuffer is not wrapped - great! */ - b->backxrb->clearFunc(ctx, b->backxrb, x, y, width, height); + b->backxrb->clearFunc(ctx, b->backxrb, x, y, width, height, value); buffers &= ~BUFFER_BIT_BACK_LEFT; } } @@ -786,12 +695,10 @@ xmesa_init_driver_functions( XMesaVisual xmvisual, driver->GetBufferSize = NULL; /* OBSOLETE */ driver->Flush = finish_or_flush; driver->Finish = finish_or_flush; - driver->ClearIndex = clear_index; - driver->ClearColor = clear_color; - driver->IndexMask = index_mask; - driver->ColorMask = color_mask; driver->Enable = enable; +#if 0 driver->Clear = xmesa_clear_buffers; +#endif driver->Viewport = xmesa_viewport; driver->TestProxyTexImage = test_proxy_teximage; #if ENABLE_EXT_texure_compression_s3tc diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 5533158ece..47ed15ccca 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -236,12 +236,12 @@ xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) if (xrb && xrb->ximage) { /* clearing back color buffer */ GET_CURRENT_CONTEXT(ctx); - xmesa_clear_buffers(ctx, BUFFER_BIT_BACK_LEFT); + xmesa_clear_buffers(ctx, BUFFER_BIT_BACK_LEFT, value); } else if (xrb && xrb->pixmap) { /* clearing front color buffer */ GET_CURRENT_CONTEXT(ctx); - xmesa_clear_buffers(ctx, BUFFER_BIT_FRONT_LEFT); + xmesa_clear_buffers(ctx, BUFFER_BIT_FRONT_LEFT, value); } else { /* clearing other buffer */ diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index 9e0c79a139..9a91aebc64 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -54,7 +54,8 @@ struct xmesa_renderbuffer; /* Function pointer for clearing color buffers */ typedef void (*ClearFunc)( GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height ); + GLint x, GLint y, GLint width, GLint height, + GLuint value ); @@ -585,7 +586,7 @@ extern void xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value); extern void -xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers); +xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers, GLuint value); extern struct pipe_context * xmesa_create_softpipe(XMesaContext xm); -- cgit v1.2.3 From b345c9258b91d8a1eb86dc4c109964e060ad3ee9 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:51:15 -0600 Subject: remove timer_query, s3tc extension bits --- src/mesa/pipe/xlib/xm_api.c | 10 ----- src/mesa/pipe/xlib/xm_dd.c | 100 -------------------------------------------- 2 files changed, 110 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index eff84f12a7..3beb9033c9 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1560,16 +1560,6 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) _mesa_enable_1_4_extensions(mesaCtx); _mesa_enable_1_5_extensions(mesaCtx); _mesa_enable_2_0_extensions(mesaCtx); -#if ENABLE_EXT_texure_compression_s3tc - if (c->Mesa_DXTn) { - _mesa_enable_extension(mesaCtx, "GL_EXT_texture_compression_s3tc"); - _mesa_enable_extension(mesaCtx, "GL_S3_s3tc"); - } - _mesa_enable_extension(mesaCtx, "GL_3DFX_texture_compression_FXT1"); -#endif -#if ENABLE_EXT_timer_query - _mesa_enable_extension(mesaCtx, "GL_EXT_timer_query"); -#endif #ifdef XFree86Server /* If we're running in the X server, do bounds checking to prevent diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index f6da1cb9f7..5a8be2e773 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -568,24 +568,6 @@ test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level, } -/** - * In SW, we don't really compress GL_COMPRESSED_RGB[A] textures! - */ -static const struct gl_texture_format * -choose_tex_format( GLcontext *ctx, GLint internalFormat, - GLenum format, GLenum type ) -{ - switch (internalFormat) { - case GL_COMPRESSED_RGB_ARB: - return &_mesa_texformat_rgb; - case GL_COMPRESSED_RGBA_ARB: - return &_mesa_texformat_rgba; - default: - return _mesa_choose_tex_format(ctx, internalFormat, format, type); - } -} - - /** * Called by glViewport. * This is a good time for us to poll the current X window size and adjust @@ -612,76 +594,6 @@ xmesa_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) } -#if ENABLE_EXT_timer_query - -/* - * The GL_EXT_timer_query extension is not enabled for the XServer - * indirect renderer. Not sure about how/if wrapping of gettimeofday() - * is done, etc. - */ - -struct xmesa_query_object -{ - struct gl_query_object Base; - struct timeval StartTime; -}; - - -static struct gl_query_object * -xmesa_new_query_object(GLcontext *ctx, GLuint id) -{ - struct xmesa_query_object *q = CALLOC_STRUCT(xmesa_query_object); - if (q) { - q->Base.Id = id; - q->Base.Ready = GL_TRUE; - } - return &q->Base; -} - - -static void -xmesa_begin_query(GLcontext *ctx, struct gl_query_object *q) -{ - if (q->Target == GL_TIME_ELAPSED_EXT) { - struct xmesa_query_object *xq = (struct xmesa_query_object *) q; - (void) gettimeofday(&xq->StartTime, NULL); - } -} - - -/** - * Return the difference between the two given times in microseconds. - */ -#ifdef __VMS -#define suseconds_t unsigned int -#endif -static GLuint64EXT -time_diff(const struct timeval *t0, const struct timeval *t1) -{ - GLuint64EXT seconds0 = t0->tv_sec & 0xff; /* 0 .. 255 seconds */ - GLuint64EXT seconds1 = t1->tv_sec & 0xff; /* 0 .. 255 seconds */ - GLuint64EXT nanosec0 = (seconds0 * 1000000 + t0->tv_usec) * 1000; - GLuint64EXT nanosec1 = (seconds1 * 1000000 + t1->tv_usec) * 1000; - return nanosec1 - nanosec0; -} - - -static void -xmesa_end_query(GLcontext *ctx, struct gl_query_object *q) -{ - if (q->Target == GL_TIME_ELAPSED_EXT) { - struct xmesa_query_object *xq = (struct xmesa_query_object *) q; - struct timeval endTime; - (void) gettimeofday(&endTime, NULL); - /* result is in nanoseconds! */ - q->Result = time_diff(&xq->StartTime, &endTime); - } - q->Ready = GL_TRUE; -} - -#endif /* ENABLE_timer_query */ - - /** * Initialize the device driver function table with the functions * we implement in this driver. @@ -701,16 +613,4 @@ xmesa_init_driver_functions( XMesaVisual xmvisual, #endif driver->Viewport = xmesa_viewport; driver->TestProxyTexImage = test_proxy_teximage; -#if ENABLE_EXT_texure_compression_s3tc - driver->ChooseTextureFormat = choose_tex_format; -#else - (void) choose_tex_format; -#endif - -#if ENABLE_EXT_timer_query - driver->NewQueryObject = xmesa_new_query_object; - driver->BeginQuery = xmesa_begin_query; - driver->EndQuery = xmesa_end_query; -#endif - } -- cgit v1.2.3 From c1f11891dfc5df6ea0bcbb809f3f4ad2b81a1f6e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 14:53:55 -0600 Subject: remove get_string, test_proxy_teximage --- src/mesa/pipe/xlib/xm_dd.c | 66 ---------------------------------------------- 1 file changed, 66 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 5a8be2e773..6fd4f736bf 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -325,33 +325,6 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers, GLuint value) } -/* - * Every driver should implement a GetString function in order to - * return a meaningful GL_RENDERER string. - */ -static const GLubyte * -get_string( GLcontext *ctx, GLenum name ) -{ - (void) ctx; - switch (name) { - case GL_RENDERER: -#ifdef XFree86Server - return (const GLubyte *) "Mesa GLX Indirect"; -#else - return (const GLubyte *) "Mesa X11 (softpipe)"; -#endif - case GL_VENDOR: -#ifdef XFree86Server - return (const GLubyte *) "Mesa project: www.mesa3d.org"; -#else - return NULL; -#endif - default: - return NULL; - } -} - - /* * We implement the glEnable function only because we care about * dither enable/disable. @@ -531,43 +504,6 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) -/** - * Called via ctx->Driver.TestProxyTeximage(). Normally, we'd just use - * the _mesa_test_proxy_teximage() fallback function, but we're going to - * special-case the 3D texture case to allow textures up to 512x512x32 - * texels. - */ -static GLboolean -test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, GLenum format, GLenum type, - GLint width, GLint height, GLint depth, GLint border) -{ - if (target == GL_PROXY_TEXTURE_3D) { - /* special case for 3D textures */ - if (width * height * depth > 512 * 512 * 64 || - width < 2 * border || - (!ctx->Extensions.ARB_texture_non_power_of_two && - _mesa_bitcount(width - 2 * border) != 1) || - height < 2 * border || - (!ctx->Extensions.ARB_texture_non_power_of_two && - _mesa_bitcount(height - 2 * border) != 1) || - depth < 2 * border || - (!ctx->Extensions.ARB_texture_non_power_of_two && - _mesa_bitcount(depth - 2 * border) != 1)) { - /* Bad size, or too many texels */ - return GL_FALSE; - } - return GL_TRUE; - } - else { - /* use the fallback routine for 1D, 2D, cube and rect targets */ - return _mesa_test_proxy_teximage(ctx, target, level, internalFormat, - format, type, width, height, depth, - border); - } -} - - /** * Called by glViewport. * This is a good time for us to poll the current X window size and adjust @@ -602,7 +538,6 @@ void xmesa_init_driver_functions( XMesaVisual xmvisual, struct dd_function_table *driver ) { - driver->GetString = get_string; driver->UpdateState = xmesa_update_state; driver->GetBufferSize = NULL; /* OBSOLETE */ driver->Flush = finish_or_flush; @@ -612,5 +547,4 @@ xmesa_init_driver_functions( XMesaVisual xmvisual, driver->Clear = xmesa_clear_buffers; #endif driver->Viewport = xmesa_viewport; - driver->TestProxyTexImage = test_proxy_teximage; } -- cgit v1.2.3 From 4411614fed938ba9495f43c01de1c3099febd860 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 15:01:25 -0600 Subject: don't pass x,y,width,height to clear functions --- src/mesa/pipe/xlib/xm_dd.c | 118 +++++++++++++++----------------------------- src/mesa/pipe/xlib/xmesaP.h | 1 - 2 files changed, 39 insertions(+), 80 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 6fd4f736bf..7a9d664840 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -114,8 +114,7 @@ finish_or_flush( GLcontext *ctx ) * Clear the front or back color buffer, if it's implemented with a pixmap. */ static void -clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height, GLuint value) +clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); @@ -130,38 +129,40 @@ clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, XMesaSetForeground( xmesa->display, xmbuf->cleargc, value ); XMesaFillRectangle( xmesa->display, xrb->pixmap, xmbuf->cleargc, - x, xrb->St.Base.Height - y - height, - width, height ); + 0, 0, xrb->St.Base.Width, xrb->St.Base.Height); } static void -clear_8bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height, GLuint value ) +clear_8bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; GLint i; for (i = 0; i < height; i++) { - GLubyte *ptr = PIXEL_ADDR1(xrb, x, y + i); + GLubyte *ptr = PIXEL_ADDR1(xrb, 0, i); MEMSET( ptr, xmesa->clearpixel, width ); } } static void -clear_HPCR_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height, GLuint value ) +clear_HPCR_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) + { const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; GLint i; - for (i = y; i < y + height; i++) { - GLubyte *ptr = PIXEL_ADDR1( xrb, x, i ); + for (i = 0; i < height; i++) { + GLubyte *ptr = PIXEL_ADDR1( xrb, 0, i ); int j; const GLubyte *sptr = xmesa->xm_visual->hpcr_clear_ximage_pattern[0]; if (i & 1) { sptr += 16; } - for (j = x; j < x + width; j++) { + for (j = 0; j < width; j++) { *ptr = sptr[j&15]; ptr++; } @@ -170,10 +171,11 @@ clear_HPCR_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, static void -clear_16bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height, GLuint value) +clear_16bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; GLint i, j; if (xmesa->swapbytes) { @@ -181,7 +183,7 @@ clear_16bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, } for (j = 0; j < height; j++) { - GLushort *ptr2 = PIXEL_ADDR2(xrb, x, y + j); + GLushort *ptr2 = PIXEL_ADDR2(xrb, 0, j); for (i = 0; i < width; i++) { ptr2[i] = value; } @@ -192,8 +194,10 @@ clear_16bit_ximage( GLcontext *ctx, struct xmesa_renderbuffer *xrb, /* Optimized code provided by Nozomi Ytow */ static void clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height, GLuint value) + GLuint value) { + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; const GLubyte r = (value ) & 0xff; const GLubyte g = (value >> 8) & 0xff; const GLubyte b = (value >> 16) & 0xff; @@ -202,7 +206,7 @@ clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, /* same value for all three components (gray) */ GLint j; for (j = 0; j < height; j++) { - bgr_t *ptr3 = PIXEL_ADDR3(xrb, x, y + j); + bgr_t *ptr3 = PIXEL_ADDR3(xrb, 0, j); MEMSET(ptr3, r, 3 * width); } } @@ -210,7 +214,7 @@ clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, /* non-gray clear color */ GLint i, j; for (j = 0; j < height; j++) { - bgr_t *ptr3 = PIXEL_ADDR3(xrb, x, y + j); + bgr_t *ptr3 = PIXEL_ADDR3(xrb, 0, j); for (i = 0; i < width; i++) { ptr3->r = r; ptr3->g = g; @@ -224,9 +228,13 @@ clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, static void clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height, GLuint value) + GLuint value) { const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; + const GLuint n = width * height; + GLuint *ptr4 = (GLuint *) xrb->ximage->data; if (!xrb->ximage) return; @@ -238,45 +246,28 @@ clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, | ((value << 24) & 0xff000000); } - if (width == xrb->St.Base.Width && height == xrb->St.Base.Height) { - /* clearing whole buffer */ - const GLuint n = xrb->St.Base.Width * xrb->St.Base.Height; - GLuint *ptr4 = (GLuint *) xrb->ximage->data; - if (value == 0) { - /* common case */ - _mesa_memset(ptr4, value, 4 * n); - } - else { - GLuint i; - for (i = 0; i < n; i++) - ptr4[i] = value; - } + if (value == 0) { + /* common case */ + _mesa_memset(ptr4, value, 4 * n); } else { - /* clearing scissored region */ - GLint i, j; - for (j = 0; j < height; j++) { - GLuint *ptr4 = PIXEL_ADDR4(xrb, x, y + j); - for (i = 0; i < width; i++) { - ptr4[i] = value; - } - } + GLuint i; + for (i = 0; i < n; i++) + ptr4[i] = value; } } static void -clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height, GLuint value) +clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) { XMesaImage *img = xrb->ximage; + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; GLint i, j; - - /* TODO: optimize this */ - y = YFLIP(xrb, y); for (j = 0; j < height; j++) { for (i = 0; i < width; i++) { - XMesaPutPixel(img, x+i, y-j, value); + XMesaPutPixel(img, i, j, value); } } } @@ -290,10 +281,6 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers, GLuint value) /* this is a window system framebuffer */ const GLuint *colorMask = (GLuint *) &ctx->Color.ColorMask; XMesaBuffer b = XMESA_BUFFER(ctx->DrawBuffer); - const GLint x = ctx->DrawBuffer->_Xmin; - const GLint y = ctx->DrawBuffer->_Ymin; - const GLint width = ctx->DrawBuffer->_Xmax - x; - const GLint height = ctx->DrawBuffer->_Ymax - y; /* we can't handle color or index masking */ if (*colorMask == 0xffffffff && ctx->Color.IndexMask == 0xffffffff) { @@ -303,7 +290,7 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers, GLuint value) = ctx->DrawBuffer->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; if (b->frontxrb == xmesa_renderbuffer(frontRb)) { /* renderbuffer is not wrapped - great! */ - b->frontxrb->clearFunc(ctx, b->frontxrb, x, y, width, height, value); + b->frontxrb->clearFunc(ctx, b->frontxrb, value); buffers &= ~BUFFER_BIT_FRONT_LEFT; } else { @@ -316,7 +303,7 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers, GLuint value) = ctx->DrawBuffer->Attachment[BUFFER_BACK_LEFT].Renderbuffer; if (b->backxrb == xmesa_renderbuffer(backRb)) { /* renderbuffer is not wrapped - great! */ - b->backxrb->clearFunc(ctx, b->backxrb, x, y, width, height, value); + b->backxrb->clearFunc(ctx, b->backxrb, value); buffers &= ~BUFFER_BIT_BACK_LEFT; } } @@ -325,28 +312,6 @@ xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers, GLuint value) } -/* - * We implement the glEnable function only because we care about - * dither enable/disable. - */ -static void -enable( GLcontext *ctx, GLenum pname, GLboolean state ) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - - switch (pname) { - case GL_DITHER: - if (state) - xmesa->pixelformat = xmesa->xm_visual->dithered_pf; - else - xmesa->pixelformat = xmesa->xm_visual->undithered_pf; - break; - default: - ; /* silence compiler warning */ - } -} - - static void clear_color_HPCR_ximage( GLcontext *ctx, const GLfloat color[4] ) { @@ -539,12 +504,7 @@ xmesa_init_driver_functions( XMesaVisual xmvisual, struct dd_function_table *driver ) { driver->UpdateState = xmesa_update_state; - driver->GetBufferSize = NULL; /* OBSOLETE */ driver->Flush = finish_or_flush; driver->Finish = finish_or_flush; - driver->Enable = enable; -#if 0 - driver->Clear = xmesa_clear_buffers; -#endif driver->Viewport = xmesa_viewport; } diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index 9a91aebc64..bf4b7178f9 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -54,7 +54,6 @@ struct xmesa_renderbuffer; /* Function pointer for clearing color buffers */ typedef void (*ClearFunc)( GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLint x, GLint y, GLint width, GLint height, GLuint value ); -- cgit v1.2.3 From 017d08a5e040ee476b19d672c17090eaca7fa918 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 15:08:19 -0600 Subject: get rid of xmesa_clear_buffers() --- src/mesa/pipe/xlib/xm_dd.c | 39 --------------------------------------- src/mesa/pipe/xlib/xm_surface.c | 13 ++++--------- src/mesa/pipe/xlib/xmesaP.h | 3 --- 3 files changed, 4 insertions(+), 51 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 7a9d664840..a07f17bfea 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -273,45 +273,6 @@ clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) } - -void -xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers, GLuint value) -{ - if (ctx->DrawBuffer->Name == 0) { - /* this is a window system framebuffer */ - const GLuint *colorMask = (GLuint *) &ctx->Color.ColorMask; - XMesaBuffer b = XMESA_BUFFER(ctx->DrawBuffer); - - /* we can't handle color or index masking */ - if (*colorMask == 0xffffffff && ctx->Color.IndexMask == 0xffffffff) { - if (buffers & BUFFER_BIT_FRONT_LEFT) { - /* clear front color buffer */ - struct gl_renderbuffer *frontRb - = ctx->DrawBuffer->Attachment[BUFFER_FRONT_LEFT].Renderbuffer; - if (b->frontxrb == xmesa_renderbuffer(frontRb)) { - /* renderbuffer is not wrapped - great! */ - b->frontxrb->clearFunc(ctx, b->frontxrb, value); - buffers &= ~BUFFER_BIT_FRONT_LEFT; - } - else { - /* we can't directly clear an alpha-wrapped color buffer */ - } - } - if (buffers & BUFFER_BIT_BACK_LEFT) { - /* clear back color buffer */ - struct gl_renderbuffer *backRb - = ctx->DrawBuffer->Attachment[BUFFER_BACK_LEFT].Renderbuffer; - if (b->backxrb == xmesa_renderbuffer(backRb)) { - /* renderbuffer is not wrapped - great! */ - b->backxrb->clearFunc(ctx, b->backxrb, value); - buffers &= ~BUFFER_BIT_BACK_LEFT; - } - } - } - } -} - - static void clear_color_HPCR_ximage( GLcontext *ctx, const GLfloat color[4] ) { diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 47ed15ccca..4e3368b6b9 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -214,6 +214,7 @@ xmesa_is_format_supported(struct pipe_context *pipe, uint format) void xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) { + GET_CURRENT_CONTEXT(ctx); struct xmesa_renderbuffer *xrb = xmesa_rb(ps); /* XXX actually, we should just discard any cached tiles from this @@ -233,15 +234,9 @@ xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) } } - if (xrb && xrb->ximage) { - /* clearing back color buffer */ - GET_CURRENT_CONTEXT(ctx); - xmesa_clear_buffers(ctx, BUFFER_BIT_BACK_LEFT, value); - } - else if (xrb && xrb->pixmap) { - /* clearing front color buffer */ - GET_CURRENT_CONTEXT(ctx); - xmesa_clear_buffers(ctx, BUFFER_BIT_FRONT_LEFT, value); + if (xrb) { + /* clearing front/back color buffer */ + xrb->clearFunc(ctx, xrb, value); } else { /* clearing other buffer */ diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index bf4b7178f9..f878af974f 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -584,9 +584,6 @@ struct xmesa_surface extern void xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value); -extern void -xmesa_clear_buffers(GLcontext *ctx, GLbitfield buffers, GLuint value); - extern struct pipe_context * xmesa_create_softpipe(XMesaContext xm); -- cgit v1.2.3 From e5ff772523e73dcc8b79270d680a8de1a7bad7bc Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 15:13:29 -0600 Subject: remove HPCR support --- src/mesa/pipe/xlib/xm_api.c | 146 ----------------------- src/mesa/pipe/xlib/xm_buffer.c | 5 - src/mesa/pipe/xlib/xm_dd.c | 135 +--------------------- src/mesa/pipe/xlib/xm_span.c | 256 ----------------------------------------- src/mesa/pipe/xlib/xmesaP.h | 27 ----- 5 files changed, 1 insertion(+), 568 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 3beb9033c9..74246ddbe3 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -88,68 +88,6 @@ _glthread_Mutex _xmesa_lock; -/** - * Lookup tables for HPCR pixel format: - */ -static short hpcr_rgbTbl[3][256] = { -{ - 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22, 23, 23, - 24, 24, 25, 25, 26, 26, 27, 27, 28, 28, 29, 29, 30, 30, 31, 31, - 32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 37, 37, 38, 38, 39, 39, - 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, - 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, - 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, - 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, - 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, -112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, -128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, -144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, -160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, -176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, -192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, -208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, -224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239 -}, -{ - 16, 16, 17, 17, 18, 18, 19, 19, 20, 20, 21, 21, 22, 22, 23, 23, - 24, 24, 25, 25, 26, 26, 27, 27, 28, 28, 29, 29, 30, 30, 31, 31, - 32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 37, 37, 38, 38, 39, 39, - 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, - 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, - 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, - 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, - 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, -112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, -128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, -144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, -160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, -176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, -192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, -208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223, -224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238, 239 -}, -{ - 32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 37, 37, 38, 38, 39, 39, - 40, 40, 41, 41, 42, 42, 43, 43, 44, 44, 45, 45, 46, 46, 47, 47, - 48, 48, 49, 49, 50, 50, 51, 51, 52, 52, 53, 53, 54, 54, 55, 55, - 56, 56, 57, 57, 58, 58, 59, 59, 60, 60, 61, 61, 62, 62, 63, 63, - 64, 64, 65, 65, 66, 66, 67, 67, 68, 68, 69, 69, 70, 70, 71, 71, - 72, 72, 73, 73, 74, 74, 75, 75, 76, 76, 77, 77, 78, 78, 79, 79, - 80, 80, 81, 81, 82, 82, 83, 83, 84, 84, 85, 85, 86, 86, 87, 87, - 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, - 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, -112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, -128, 129, 130, 131, 132, 133, 134, 135, 136, 137, 138, 139, 140, 141, 142, 143, -144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, -160, 161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, -176, 177, 178, 179, 180, 181, 182, 183, 184, 185, 186, 187, 188, 189, 190, 191, -192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, -208, 209, 210, 211, 212, 213, 214, 215, 216, 217, 218, 219, 220, 221, 222, 223 -} -}; - - - /**********************************************************************/ /***** X Utility Functions *****/ /**********************************************************************/ @@ -858,62 +796,6 @@ setup_dithered_color(int client, XMesaVisual v, } -/** - * Setup for Hewlett Packard Color Recovery 8-bit TrueColor mode. - * HPCR simulates 24-bit color fidelity with an 8-bit frame buffer. - * Special dithering tables have to be initialized. - */ -static void -setup_8bit_hpcr(XMesaVisual v) -{ - /* HP Color Recovery contributed by: Alex De Bruyn (ad@lms.be) - * To work properly, the atom _HP_RGB_SMOOTH_MAP_LIST must be defined - * on the root window AND the colormap obtainable by XGetRGBColormaps - * for that atom must be set on the window. (see also tkInitWindow) - * If that colormap is not set, the output will look stripy. - */ - - /* Setup color tables with gamma correction */ - int i; - double g; - - g = 1.0 / v->RedGamma; - for (i=0; i<256; i++) { - GLint red = IROUND_POS(255.0 * _mesa_pow( hpcr_rgbTbl[0][i]/255.0, g )); - v->hpcr_rgbTbl[0][i] = CLAMP( red, 16, 239 ); - } - - g = 1.0 / v->GreenGamma; - for (i=0; i<256; i++) { - GLint green = IROUND_POS(255.0 * _mesa_pow( hpcr_rgbTbl[1][i]/255.0, g )); - v->hpcr_rgbTbl[1][i] = CLAMP( green, 16, 239 ); - } - - g = 1.0 / v->BlueGamma; - for (i=0; i<256; i++) { - GLint blue = IROUND_POS(255.0 * _mesa_pow( hpcr_rgbTbl[2][i]/255.0, g )); - v->hpcr_rgbTbl[2][i] = CLAMP( blue, 32, 223 ); - } - v->undithered_pf = PF_HPCR; /* can't really disable dithering for now */ - v->dithered_pf = PF_HPCR; - - /* which method should I use to clear */ - /* GL_FALSE: keep the ordinary method */ - /* GL_TRUE : clear with dither pattern */ - v->hpcr_clear_flag = _mesa_getenv("MESA_HPCR_CLEAR") ? GL_TRUE : GL_FALSE; - - if (v->hpcr_clear_flag) { - v->hpcr_clear_pixmap = XMesaCreatePixmap(v->display, - DefaultRootWindow(v->display), - 16, 2, 8); -#ifndef XFree86Server - v->hpcr_clear_ximage = XGetImage(v->display, v->hpcr_clear_pixmap, - 0, 0, 16, 2, AllPlanes, ZPixmap); -#endif - } -} - - /** * Setup RGB rendering for a window with a True/DirectColor visual. */ @@ -1039,13 +921,6 @@ setup_truecolor(XMesaVisual v, XMesaBuffer buffer, XMesaColormap cmap) v->undithered_pf = PF_5R6G5B; v->dithered_pf = PF_Dither_5R6G5B; } - else if (GET_REDMASK(v) ==0xe0 - && GET_GREENMASK(v)==0x1c - && GET_BLUEMASK(v) ==0x03 - && CHECK_FOR_HPCR(v)) { - /* 8-bit HP color recovery */ - setup_8bit_hpcr( v ); - } } @@ -1198,23 +1073,6 @@ initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, } #endif XMesaSetFunction( v->display, b->swapgc, GXcopy ); - /* - * Set fill style and tile pixmap once for all for HPCR stuff - * (instead of doing it each time in clear_color_HPCR_pixmap()) - * Initialize whole stuff - * Patch contributed by Jacques Leroy March 8, 1998. - */ - if (v->hpcr_clear_flag && b->backxrb && b->backxrb->pixmap) { - int i; - for (i = 0; i < 16; i++) { - XMesaPutPixel(v->hpcr_clear_ximage, i, 0, 0); - XMesaPutPixel(v->hpcr_clear_ximage, i, 1, 0); - } - XMesaPutImage(b->display, (XMesaDrawable) v->hpcr_clear_pixmap, - b->cleargc, v->hpcr_clear_ximage, 0, 0, 0, 0, 16, 2); - XMesaSetFillStyle( v->display, b->cleargc, FillTiled); - XMesaSetTile( v->display, b->cleargc, v->hpcr_clear_pixmap ); - } /* Initialize the row buffer XImage for use in write_color_span() */ data = (char*) MALLOC(MAX_WIDTH*4); @@ -1275,8 +1133,6 @@ xmesa_color_to_pixel(GLcontext *ctx, case PF_1Bit: /* 382 = (3*255)/2 */ return ((r+g+b) > 382) ^ xmesa->xm_visual->bitFlip; - case PF_HPCR: - return DITHER_HPCR(1, 1, r, g, b); case PF_Lookup: { LOOKUP_SETUP; @@ -2293,8 +2149,6 @@ unsigned long XMesaDitherColor( XMesaContext xmesa, GLint x, GLint y, case PF_1Bit: /* 382 = (3*255)/2 */ return ((r+g+b) > 382) ^ xmesa->xm_visual->bitFlip; - case PF_HPCR: - return DITHER_HPCR(x, y, r, g, b); case PF_Lookup: { LOOKUP_SETUP; diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index 09356c7329..b25ceeffdd 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -484,11 +484,6 @@ xmesa_delete_framebuffer(struct gl_framebuffer *fb) } if (b->backxrb->pixmap) { XMesaFreePixmap( b->display, b->backxrb->pixmap ); - if (b->xm_visual->hpcr_clear_flag) { - XMesaFreePixmap( b->display, - b->xm_visual->hpcr_clear_pixmap ); - XMesaDestroyImage( b->xm_visual->hpcr_clear_ximage ); - } } } diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index a07f17bfea..87d1e404ae 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -65,21 +65,6 @@ const int xmesa_kernel8[DITH_DY * DITH_DX] = { 15 * MAXC, 7 * MAXC, 13 * MAXC, 5 * MAXC, }; -const short xmesa_HPCR_DRGB[3][2][16] = { - { - { 16, -4, 1,-11, 14, -6, 3, -9, 15, -5, 2,-10, 13, -7, 4, -8}, - {-15, 5, 0, 12,-13, 7, -2, 10,-14, 6, -1, 11,-12, 8, -3, 9} - }, - { - {-11, 15, -7, 3, -8, 14, -4, 2,-10, 16, -6, 4, -9, 13, -5, 1}, - { 12,-14, 8, -2, 9,-13, 5, -1, 11,-15, 7, -3, 10,-12, 6, 0} - }, - { - { 6,-18, 26,-14, 2,-22, 30,-10, 8,-16, 28,-12, 4,-20, 32, -8}, - { -4, 20,-24, 16, 0, 24,-28, 12, -6, 18,-26, 14, -2, 22,-30, 10} - } -}; - const int xmesa_kernel1[16] = { 0*47, 9*47, 4*47, 12*47, /* 47 = (255*3)/16 */ 6*47, 2*47, 14*47, 8*47, @@ -147,29 +132,6 @@ clear_8bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) } -static void -clear_HPCR_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) - -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - GLint i; - for (i = 0; i < height; i++) { - GLubyte *ptr = PIXEL_ADDR1( xrb, 0, i ); - int j; - const GLubyte *sptr = xmesa->xm_visual->hpcr_clear_ximage_pattern[0]; - if (i & 1) { - sptr += 16; - } - for (j = 0; j < width; j++) { - *ptr = sptr[j&15]; - ptr++; - } - } -} - - static void clear_16bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) { @@ -273,80 +235,6 @@ clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) } -static void -clear_color_HPCR_ximage( GLcontext *ctx, const GLfloat color[4] ) -{ - int i; - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[3], color[3]); - - if (color[0] == 0.0 && color[1] == 0.0 && color[2] == 0.0) { - /* black is black */ - MEMSET( xmesa->xm_visual->hpcr_clear_ximage_pattern, 0x0 , - sizeof(xmesa->xm_visual->hpcr_clear_ximage_pattern)); - } - else { - /* build clear pattern */ - for (i=0; i<16; i++) { - xmesa->xm_visual->hpcr_clear_ximage_pattern[0][i] = - DITHER_HPCR(i, 0, - xmesa->clearcolor[0], - xmesa->clearcolor[1], - xmesa->clearcolor[2]); - xmesa->xm_visual->hpcr_clear_ximage_pattern[1][i] = - DITHER_HPCR(i, 1, - xmesa->clearcolor[0], - xmesa->clearcolor[1], - xmesa->clearcolor[2]); - } - } -} - - -static void -clear_color_HPCR_pixmap( GLcontext *ctx, const GLfloat color[4] ) -{ - int i; - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(xmesa->clearcolor[3], color[3]); - - if (color[0] == 0.0 && color[1] == 0.0 && color[2] == 0.0) { - /* black is black */ - for (i=0; i<16; i++) { - XMesaPutPixel(xmesa->xm_visual->hpcr_clear_ximage, i, 0, 0); - XMesaPutPixel(xmesa->xm_visual->hpcr_clear_ximage, i, 1, 0); - } - } - else { - for (i=0; i<16; i++) { - XMesaPutPixel(xmesa->xm_visual->hpcr_clear_ximage, i, 0, - DITHER_HPCR(i, 0, - xmesa->clearcolor[0], - xmesa->clearcolor[1], - xmesa->clearcolor[2])); - XMesaPutPixel(xmesa->xm_visual->hpcr_clear_ximage, i, 1, - DITHER_HPCR(i, 1, - xmesa->clearcolor[0], - xmesa->clearcolor[1], - xmesa->clearcolor[2])); - } - } - /* change tile pixmap content */ - XMesaPutImage(xmesa->display, - (XMesaDrawable)xmesa->xm_visual->hpcr_clear_pixmap, - XMESA_BUFFER(ctx->DrawBuffer)->cleargc, - xmesa->xm_visual->hpcr_clear_ximage, 0, 0, 0, 0, 16, 2); -} - - /** * Called when the driver should update its state, based on the new_state * flags. @@ -387,12 +275,7 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) else { switch (xmesa->xm_visual->BitsPerPixel) { case 8: - if (xmesa->xm_visual->hpcr_clear_flag) { - back_xrb->clearFunc = clear_HPCR_ximage; - } - else { - back_xrb->clearFunc = clear_8bit_ximage; - } + back_xrb->clearFunc = clear_8bit_ximage; break; case 16: back_xrb->clearFunc = clear_16bit_ximage; @@ -410,22 +293,6 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) } } } - - if (xmesa->xm_visual->hpcr_clear_flag) { - /* this depends on whether we're drawing to the front or back buffer */ - /* XXX FIX THIS! */ -#if 0 - if (pixmap) { - ctx->Driver.ClearColor = clear_color_HPCR_pixmap; - } - else { - ctx->Driver.ClearColor = clear_color_HPCR_ximage; - } -#else - (void) clear_color_HPCR_pixmap; - (void) clear_color_HPCR_ximage; -#endif - } } diff --git a/src/mesa/pipe/xlib/xm_span.c b/src/mesa/pipe/xlib/xm_span.c index 89eeff5d9d..e404ff49eb 100644 --- a/src/mesa/pipe/xlib/xm_span.c +++ b/src/mesa/pipe/xlib/xm_span.c @@ -988,71 +988,6 @@ static void put_row_rgb_1BIT_pixmap( RGB_SPAN_ARGS ) } -/* - * Write a span of PF_HPCR pixels to a pixmap. - */ -static void put_row_HPCR_pixmap( PUT_ROW_ARGS ) -{ - const GLubyte (*rgba)[4] = (const GLubyte (*)[4]) values; - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GET_XRB(xrb); - XMesaDisplay *dpy = xmesa->xm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLubyte *ptr = (GLubyte *) XMESA_BUFFER(ctx->DrawBuffer)->rowimage->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLubyte *ptr = (GLubyte *) XMESA_BUFFER(ctx->DrawBuffer)->rowimage->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;idata; - GLuint i; - for (i=0;iPutMonoValues = put_mono_values_1BIT_ximage; } break; - case PF_HPCR: - if (pixmap) { - rb->PutRow = put_row_HPCR_pixmap; - rb->PutRowRGB = put_row_rgb_HPCR_pixmap; - rb->PutMonoRow = put_mono_row_pixmap; - rb->PutValues = put_values_HPCR_pixmap; - rb->PutMonoValues = put_mono_values_pixmap; - } - else { - rb->PutRow = put_row_HPCR_ximage; - rb->PutRowRGB = put_row_rgb_HPCR_ximage; - rb->PutMonoRow = put_mono_row_HPCR_ximage; - rb->PutValues = put_values_HPCR_ximage; - rb->PutMonoValues = put_mono_values_HPCR_ximage; - } - break; case PF_Lookup: if (pixmap) { rb->PutRow = put_row_LOOKUP_pixmap; diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index f878af974f..195d4839a9 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -71,7 +71,6 @@ enum pixel_format { PF_5R6G5B, /**< 16-bit TrueColor: 5-R, 6-G, 5-B bits */ PF_Dither, /**< Color-mapped RGB with dither */ PF_Lookup, /**< Color-mapped RGB without dither */ - PF_HPCR, /**< HP Color Recovery (ad@lms.be 30/08/95) */ PF_1Bit, /**< monochrome dithering of RGB */ PF_Grayscale, /**< Grayscale or StaticGray */ PF_Dither_5R6G5B /**< 16-bit dithered TrueColor: 5-R, 6-G, 5-B */ @@ -113,13 +112,6 @@ struct xmesa_visual { GLubyte PixelToG[256]; GLubyte PixelToB[256]; - /* For PF_HPCR */ - short hpcr_rgbTbl[3][256]; - GLboolean hpcr_clear_flag; - GLubyte hpcr_clear_ximage_pattern[2][16]; - XMesaImage *hpcr_clear_ximage; - XMesaPixmap hpcr_clear_pixmap; - /* For PF_1BIT */ int bitFlip; }; @@ -422,25 +414,6 @@ extern const int xmesa_kernel8[DITH_DY * DITH_DX]; _dither_lookup(DITH_B, (B)))] -/** - * If pixelformat==PF_HPCR: - * - * HP Color Recovery dithering (ad@lms.be 30/08/95) - * HP has on it's 8-bit 700-series computers, a feature called - * 'Color Recovery'. This allows near 24-bit output (so they say). - * It is enabled by selecting the 8-bit TrueColor visual AND - * corresponding colormap (see tkInitWindow) AND doing some special - * dither. - */ -extern const short xmesa_HPCR_DRGB[3][2][16]; - -#define DITHER_HPCR( X, Y, R, G, B ) \ - ( ((xmesa->xm_visual->hpcr_rgbTbl[0][R] + xmesa_HPCR_DRGB[0][(Y)&1][(X)&15]) & 0xE0) \ - |(((xmesa->xm_visual->hpcr_rgbTbl[1][G] + xmesa_HPCR_DRGB[1][(Y)&1][(X)&15]) & 0xE0)>>3) \ - | ((xmesa->xm_visual->hpcr_rgbTbl[2][B] + xmesa_HPCR_DRGB[2][(Y)&1][(X)&15])>>6) \ - ) - - /** * If pixelformat==PF_1BIT: -- cgit v1.2.3 From 30fb4c43b02f4105944453e48d7a0f878746595d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 15:16:22 -0600 Subject: clean-up includes --- src/mesa/pipe/xlib/xm_buffer.c | 8 ++++---- src/mesa/pipe/xlib/xm_dd.c | 19 ++----------------- src/mesa/pipe/xlib/xm_span.c | 5 ----- src/mesa/pipe/xlib/xm_surface.c | 10 ++++------ 4 files changed, 10 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index b25ceeffdd..56b797a6a9 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -29,12 +29,12 @@ */ -#include "glxheader.h" #include "GL/xmesa.h" +#include "glxheader.h" #include "xmesaP.h" -#include "imports.h" -#include "framebuffer.h" -#include "renderbuffer.h" +#include "main/imports.h" +#include "main/framebuffer.h" +#include "main/renderbuffer.h" #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 87d1e404ae..46bd3c85d2 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -29,24 +29,9 @@ */ #include "glxheader.h" -#include "bufferobj.h" -#include "buffers.h" -#include "context.h" -#include "colormac.h" -#include "depth.h" -#include "drawpix.h" -#include "extensions.h" -#include "framebuffer.h" -#include "macros.h" -#include "image.h" -#include "imports.h" -#include "mtypes.h" -#include "state.h" -#include "texobj.h" -#include "teximage.h" -#include "texstore.h" -#include "texformat.h" #include "xmesaP.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "pipe/softpipe/sp_context.h" #include "state_tracker/st_public.h" diff --git a/src/mesa/pipe/xlib/xm_span.c b/src/mesa/pipe/xlib/xm_span.c index e404ff49eb..032f08a28e 100644 --- a/src/mesa/pipe/xlib/xm_span.c +++ b/src/mesa/pipe/xlib/xm_span.c @@ -24,14 +24,9 @@ #include "glxheader.h" #include "colormac.h" -#include "context.h" -#include "depth.h" -#include "drawpix.h" -#include "extensions.h" #include "macros.h" #include "imports.h" #include "mtypes.h" -#include "state.h" #include "xmesaP.h" diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 4e3368b6b9..e7254d0d5a 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -33,14 +33,12 @@ */ -#include "glxheader.h" #include "GL/xmesa.h" +#include "glxheader.h" #include "xmesaP.h" -#include "context.h" -#include "imports.h" -#include "macros.h" -#include "framebuffer.h" -#include "renderbuffer.h" +#include "main/context.h" +#include "main/imports.h" +#include "main/macros.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -- cgit v1.2.3 From 8d0d6f04a15f31c4491966767c547fe1ffbeb362 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 15:27:09 -0600 Subject: move surface clear functions to xm_buffer.c --- src/mesa/pipe/xlib/xm_buffer.c | 168 +++++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/xlib/xm_dd.c | 151 ++---------------------------------- 2 files changed, 174 insertions(+), 145 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index 56b797a6a9..183a7ec8f5 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -261,6 +261,146 @@ finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) } +/** + * Clear the front or back color buffer, if it's implemented with a pixmap. + */ +static void +clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + + assert(xmbuf); + assert(xrb->pixmap); + assert(xmesa); + assert(xmesa->display); + assert(xrb->pixmap); + assert(xmbuf->cleargc); + + XMesaSetForeground( xmesa->display, xmbuf->cleargc, value ); + + XMesaFillRectangle( xmesa->display, xrb->pixmap, xmbuf->cleargc, + 0, 0, xrb->St.Base.Width, xrb->St.Base.Height); +} + + +static void +clear_8bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; + GLint i; + for (i = 0; i < height; i++) { + GLubyte *ptr = PIXEL_ADDR1(xrb, 0, i); + MEMSET( ptr, xmesa->clearpixel, width ); + } +} + + +static void +clear_16bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; + GLint i, j; + + if (xmesa->swapbytes) { + value = ((value >> 8) & 0x00ff) | ((value << 8) & 0xff00); + } + + for (j = 0; j < height; j++) { + GLushort *ptr2 = PIXEL_ADDR2(xrb, 0, j); + for (i = 0; i < width; i++) { + ptr2[i] = value; + } + } +} + + +/* Optimized code provided by Nozomi Ytow */ +static void +clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLuint value) +{ + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; + const GLubyte r = (value ) & 0xff; + const GLubyte g = (value >> 8) & 0xff; + const GLubyte b = (value >> 16) & 0xff; + + if (r == g && g == b) { + /* same value for all three components (gray) */ + GLint j; + for (j = 0; j < height; j++) { + bgr_t *ptr3 = PIXEL_ADDR3(xrb, 0, j); + MEMSET(ptr3, r, 3 * width); + } + } + else { + /* non-gray clear color */ + GLint i, j; + for (j = 0; j < height; j++) { + bgr_t *ptr3 = PIXEL_ADDR3(xrb, 0, j); + for (i = 0; i < width; i++) { + ptr3->r = r; + ptr3->g = g; + ptr3->b = b; + ptr3++; + } + } + } +} + + +static void +clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, + GLuint value) +{ + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; + const GLuint n = width * height; + GLuint *ptr4 = (GLuint *) xrb->ximage->data; + + if (!xrb->ximage) + return; + + if (xmesa->swapbytes) { + value = ((value >> 24) & 0x000000ff) + | ((value >> 8) & 0x0000ff00) + | ((value << 8) & 0x00ff0000) + | ((value << 24) & 0xff000000); + } + + if (value == 0) { + /* common case */ + _mesa_memset(ptr4, value, 4 * n); + } + else { + GLuint i; + for (i = 0; i < n; i++) + ptr4[i] = value; + } +} + + +static void +clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) +{ + XMesaImage *img = xrb->ximage; + GLint width = xrb->St.Base.Width; + GLint height = xrb->St.Base.Height; + GLint i, j; + for (j = 0; j < height; j++) { + for (i = 0; i < width; i++) { + XMesaPutPixel(img, i, j, value); + } + } +} + + /** * Reallocate renderbuffer storage for front color buffer. * Called via gl_renderbuffer::AllocStorage() @@ -269,6 +409,7 @@ static GLboolean xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, GLenum internalFormat, GLuint width, GLuint height) { + const XMesaContext xmesa = XMESA_CONTEXT(ctx); struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); /* just clear these to be sure we don't accidentally use them */ @@ -287,6 +428,11 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, if (!xrb->St.surface || !xrb->St.surface->region) finish_surface_init(ctx, xrb); + xmesa_set_renderbuffer_funcs(xrb, xmesa->pixelformat, + xmesa->xm_visual->BitsPerPixel); + + xrb->clearFunc = clear_pixmap; + xrb->St.surface->width = width; xrb->St.surface->height = height; @@ -302,6 +448,7 @@ static GLboolean xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, GLenum internalFormat, GLuint width, GLuint height) { + const XMesaContext xmesa = XMESA_CONTEXT(ctx); struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); /* reallocate the back buffer XImage or Pixmap */ @@ -339,6 +486,27 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xrb->origin4 = NULL; } + xmesa_set_renderbuffer_funcs(xrb, xmesa->pixelformat, + xmesa->xm_visual->BitsPerPixel); + + switch (xmesa->xm_visual->BitsPerPixel) { + case 8: + xrb->clearFunc = clear_8bit_ximage; + break; + case 16: + xrb->clearFunc = clear_16bit_ximage; + break; + case 24: + xrb->clearFunc = clear_24bit_ximage; + break; + case 32: + xrb->clearFunc = clear_32bit_ximage; + break; + default: + xrb->clearFunc = clear_nbit_ximage; + break; + } + if (!xrb->St.surface || !xrb->St.surface->region) finish_surface_init(ctx, xrb); diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 46bd3c85d2..862a55381f 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -75,151 +75,6 @@ finish_or_flush( GLcontext *ctx ) } -/**********************************************************************/ -/*** glClear implementations ***/ -/**********************************************************************/ - - -/** - * Clear the front or back color buffer, if it's implemented with a pixmap. - */ -static void -clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - - assert(xmbuf); - assert(xrb->pixmap); - assert(xmesa); - assert(xmesa->display); - assert(xrb->pixmap); - assert(xmbuf->cleargc); - - XMesaSetForeground( xmesa->display, xmbuf->cleargc, value ); - - XMesaFillRectangle( xmesa->display, xrb->pixmap, xmbuf->cleargc, - 0, 0, xrb->St.Base.Width, xrb->St.Base.Height); -} - - -static void -clear_8bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - GLint i; - for (i = 0; i < height; i++) { - GLubyte *ptr = PIXEL_ADDR1(xrb, 0, i); - MEMSET( ptr, xmesa->clearpixel, width ); - } -} - - -static void -clear_16bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - GLint i, j; - - if (xmesa->swapbytes) { - value = ((value >> 8) & 0x00ff) | ((value << 8) & 0xff00); - } - - for (j = 0; j < height; j++) { - GLushort *ptr2 = PIXEL_ADDR2(xrb, 0, j); - for (i = 0; i < width; i++) { - ptr2[i] = value; - } - } -} - - -/* Optimized code provided by Nozomi Ytow */ -static void -clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLuint value) -{ - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - const GLubyte r = (value ) & 0xff; - const GLubyte g = (value >> 8) & 0xff; - const GLubyte b = (value >> 16) & 0xff; - - if (r == g && g == b) { - /* same value for all three components (gray) */ - GLint j; - for (j = 0; j < height; j++) { - bgr_t *ptr3 = PIXEL_ADDR3(xrb, 0, j); - MEMSET(ptr3, r, 3 * width); - } - } - else { - /* non-gray clear color */ - GLint i, j; - for (j = 0; j < height; j++) { - bgr_t *ptr3 = PIXEL_ADDR3(xrb, 0, j); - for (i = 0; i < width; i++) { - ptr3->r = r; - ptr3->g = g; - ptr3->b = b; - ptr3++; - } - } - } -} - - -static void -clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLuint value) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - const GLuint n = width * height; - GLuint *ptr4 = (GLuint *) xrb->ximage->data; - - if (!xrb->ximage) - return; - - if (xmesa->swapbytes) { - value = ((value >> 24) & 0x000000ff) - | ((value >> 8) & 0x0000ff00) - | ((value << 8) & 0x00ff0000) - | ((value << 24) & 0xff000000); - } - - if (value == 0) { - /* common case */ - _mesa_memset(ptr4, value, 4 * n); - } - else { - GLuint i; - for (i = 0; i < n; i++) - ptr4[i] = value; - } -} - - -static void -clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) -{ - XMesaImage *img = xrb->ximage; - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - GLint i, j; - for (j = 0; j < height; j++) { - for (i = 0; i < width; i++) { - XMesaPutPixel(img, i, j, value); - } - } -} - - /** * Called when the driver should update its state, based on the new_state * flags. @@ -227,7 +82,9 @@ clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) void xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) { +#if 0 const XMesaContext xmesa = XMESA_CONTEXT(ctx); +#endif st_invalidate_state( ctx, new_state ); @@ -245,13 +102,16 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) front_xrb = xmbuf->frontxrb; if (front_xrb) { +#if 0 xmesa_set_renderbuffer_funcs(front_xrb, xmesa->pixelformat, xmesa->xm_visual->BitsPerPixel); front_xrb->clearFunc = clear_pixmap; +#endif } back_xrb = xmbuf->backxrb; if (back_xrb) { +#if 0 xmesa_set_renderbuffer_funcs(back_xrb, xmesa->pixelformat, xmesa->xm_visual->BitsPerPixel); if (xmbuf->backxrb->pixmap) { @@ -276,6 +136,7 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) break; } } +#endif } } } -- cgit v1.2.3 From a778d60076f0255caff5237fb40dadfaa9e406e4 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 15:32:16 -0600 Subject: Remove xmesa_update_state(), which was called by ctx->Driver.UpdateState() --- src/mesa/pipe/xlib/xm_api.c | 7 ----- src/mesa/pipe/xlib/xm_buffer.c | 38 ----------------------- src/mesa/pipe/xlib/xm_dd.c | 69 ------------------------------------------ src/mesa/pipe/xlib/xm_span.c | 2 +- src/mesa/pipe/xlib/xmesaP.h | 3 -- 5 files changed, 1 insertion(+), 118 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 74246ddbe3..7d52010b2f 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1397,13 +1397,6 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) xmesa_init_driver_functions(v, &functions); st_init_driver_functions(&functions); - /* override st's function */ - functions.UpdateState = xmesa_update_state; - - /* - functions.NewRenderbuffer = xmesa_new_renderbuffer; - */ - if (!_mesa_initialize_context(mesaCtx, &v->mesa_visual, share_list ? &(share_list->mesa) : (GLcontext *) NULL, &functions, (void *) c)) { diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index 183a7ec8f5..8ac9fb55b6 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -566,44 +566,6 @@ xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, } -#if 0 -struct gl_renderbuffer * -xmesa_new_renderbuffer(GLcontext *ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, GLuint width, GLuint height) -{ - struct xmesa_renderbuffer *xrb = CALLOC_STRUCT(xmesa_renderbuffer); - if (xrb) { - GLuint name = 0; - _mesa_init_renderbuffer(&xrb->St.Base, name); - - xrb->St.Base.Delete = xmesa_delete_renderbuffer; - if (backBuffer) - xrb->St.Base.AllocStorage = xmesa_alloc_back_storage; - else - xrb->St.Base.AllocStorage = xmesa_alloc_front_storage; - - if (visual->rgbMode) { - xrb->St.Base.InternalFormat = GL_RGBA; - xrb->St.Base._BaseFormat = GL_RGBA; - xrb->St.Base.DataType = GL_UNSIGNED_BYTE; - xrb->St.Base.RedBits = visual->redBits; - xrb->St.Base.GreenBits = visual->greenBits; - xrb->St.Base.BlueBits = visual->blueBits; - xrb->St.Base.AlphaBits = visual->alphaBits; - } - else { - xrb->St.Base.InternalFormat = GL_COLOR_INDEX; - xrb->St.Base._BaseFormat = GL_COLOR_INDEX; - xrb->St.Base.DataType = GL_UNSIGNED_INT; - xrb->St.Base.IndexBits = visual->indexBits; - } - /* only need to set Red/Green/EtcBits fields for user-created RBs */ - } - return xrb; -} -#endif - - /** * Called via gl_framebuffer::Delete() method when this buffer * is _really_ being deleted. diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 862a55381f..a4645fcc82 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -75,74 +75,6 @@ finish_or_flush( GLcontext *ctx ) } -/** - * Called when the driver should update its state, based on the new_state - * flags. - */ -void -xmesa_update_state( GLcontext *ctx, GLbitfield new_state ) -{ -#if 0 - const XMesaContext xmesa = XMESA_CONTEXT(ctx); -#endif - - st_invalidate_state( ctx, new_state ); - - - if (ctx->DrawBuffer->Name != 0) - return; - - /* - * GL_DITHER, GL_READ/DRAW_BUFFER, buffer binding state, etc. effect - * renderbuffer span/clear funcs. - */ - if (new_state & (_NEW_COLOR | _NEW_BUFFERS)) { - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - struct xmesa_renderbuffer *front_xrb, *back_xrb; - - front_xrb = xmbuf->frontxrb; - if (front_xrb) { -#if 0 - xmesa_set_renderbuffer_funcs(front_xrb, xmesa->pixelformat, - xmesa->xm_visual->BitsPerPixel); - front_xrb->clearFunc = clear_pixmap; -#endif - } - - back_xrb = xmbuf->backxrb; - if (back_xrb) { -#if 0 - xmesa_set_renderbuffer_funcs(back_xrb, xmesa->pixelformat, - xmesa->xm_visual->BitsPerPixel); - if (xmbuf->backxrb->pixmap) { - back_xrb->clearFunc = clear_pixmap; - } - else { - switch (xmesa->xm_visual->BitsPerPixel) { - case 8: - back_xrb->clearFunc = clear_8bit_ximage; - break; - case 16: - back_xrb->clearFunc = clear_16bit_ximage; - break; - case 24: - back_xrb->clearFunc = clear_24bit_ximage; - break; - case 32: - back_xrb->clearFunc = clear_32bit_ximage; - break; - default: - back_xrb->clearFunc = clear_nbit_ximage; - break; - } - } -#endif - } - } -} - - - /** * Called by glViewport. * This is a good time for us to poll the current X window size and adjust @@ -177,7 +109,6 @@ void xmesa_init_driver_functions( XMesaVisual xmvisual, struct dd_function_table *driver ) { - driver->UpdateState = xmesa_update_state; driver->Flush = finish_or_flush; driver->Finish = finish_or_flush; driver->Viewport = xmesa_viewport; diff --git a/src/mesa/pipe/xlib/xm_span.c b/src/mesa/pipe/xlib/xm_span.c index 032f08a28e..52b5c88804 100644 --- a/src/mesa/pipe/xlib/xm_span.c +++ b/src/mesa/pipe/xlib/xm_span.c @@ -4535,7 +4535,7 @@ xmesa_set_renderbuffer_funcs(struct xmesa_renderbuffer *xrb, } break; default: - _mesa_problem(NULL, "Bad pixel format in xmesa_update_state (1)"); + _mesa_problem(NULL, "Bad pixel format in xmesa_set_renderbuffer_funcs"); return; } diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index 195d4839a9..3cfb2e958d 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -487,9 +487,6 @@ extern void xmesa_init_driver_functions( XMesaVisual xmvisual, struct dd_function_table *driver ); -extern void -xmesa_update_state( GLcontext *ctx, GLbitfield new_state ); - extern void xmesa_set_renderbuffer_funcs(struct xmesa_renderbuffer *xrb, enum pixel_format pixelformat, GLint depth); -- cgit v1.2.3 From 3023756684eb54642477802757eeb9fb5da433cc Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 15:57:57 -0600 Subject: move dither kernels/tables to xm_span.c --- src/mesa/pipe/xlib/xm_dd.c | 19 ------------------- src/mesa/pipe/xlib/xm_span.c | 20 ++++++++++++++++++++ 2 files changed, 20 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index a4645fcc82..56db798869 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -39,25 +39,6 @@ #include "state_tracker/st_draw.h" -/* - * Dithering kernels and lookup tables. - */ - -const int xmesa_kernel8[DITH_DY * DITH_DX] = { - 0 * MAXC, 8 * MAXC, 2 * MAXC, 10 * MAXC, - 12 * MAXC, 4 * MAXC, 14 * MAXC, 6 * MAXC, - 3 * MAXC, 11 * MAXC, 1 * MAXC, 9 * MAXC, - 15 * MAXC, 7 * MAXC, 13 * MAXC, 5 * MAXC, -}; - -const int xmesa_kernel1[16] = { - 0*47, 9*47, 4*47, 12*47, /* 47 = (255*3)/16 */ - 6*47, 2*47, 14*47, 8*47, - 10*47, 1*47, 5*47, 11*47, - 7*47, 13*47, 3*47, 15*47 -}; - - static void finish_or_flush( GLcontext *ctx ) { diff --git a/src/mesa/pipe/xlib/xm_span.c b/src/mesa/pipe/xlib/xm_span.c index 52b5c88804..b3dbe6a9c8 100644 --- a/src/mesa/pipe/xlib/xm_span.c +++ b/src/mesa/pipe/xlib/xm_span.c @@ -31,6 +31,26 @@ +/* + * Dithering kernels and lookup tables. + */ + +const int xmesa_kernel8[DITH_DY * DITH_DX] = { + 0 * MAXC, 8 * MAXC, 2 * MAXC, 10 * MAXC, + 12 * MAXC, 4 * MAXC, 14 * MAXC, 6 * MAXC, + 3 * MAXC, 11 * MAXC, 1 * MAXC, 9 * MAXC, + 15 * MAXC, 7 * MAXC, 13 * MAXC, 5 * MAXC, +}; + +const int xmesa_kernel1[16] = { + 0*47, 9*47, 4*47, 12*47, /* 47 = (255*3)/16 */ + 6*47, 2*47, 14*47, 8*47, + 10*47, 1*47, 5*47, 11*47, + 7*47, 13*47, 3*47, 15*47 +}; + + + /* * The following functions are used to trap XGetImage() calls which * generate BadMatch errors if the drawable isn't mapped. -- cgit v1.2.3 From 869318db646e8a1e040cc56f271936f0c94762dc Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 15:59:24 -0600 Subject: move few remaining bits of xm_dd.c into xm_api.c --- src/mesa/pipe/xlib/xm_api.c | 57 +++++++++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/xlib/xm_dd.c | 55 ------------------------------------------- src/mesa/pipe/xlib/xmesaP.h | 4 ---- 3 files changed, 57 insertions(+), 59 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 7d52010b2f..1693530fbf 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1365,6 +1365,63 @@ void XMesaDestroyVisual( XMesaVisual v ) +static void +finish_or_flush( GLcontext *ctx ) +{ +#ifdef XFree86Server + /* NOT_NEEDED */ +#else + const XMesaContext xmesa = XMESA_CONTEXT(ctx); + if (xmesa) { + _glthread_LOCK_MUTEX(_xmesa_lock); + XSync( xmesa->display, False ); + _glthread_UNLOCK_MUTEX(_xmesa_lock); + } +#endif + abort(); +} + + +/** + * Called by glViewport. + * This is a good time for us to poll the current X window size and adjust + * our renderbuffers to match the current window size. + * Remember, we have no opportunity to respond to conventional + * X Resize/StructureNotify events since the X driver has no event loop. + * Thus, we poll. + * Note that this trick isn't fool-proof. If the application never calls + * glViewport, our notion of the current window size may be incorrect. + * That problem led to the GLX_MESA_resize_buffers extension. + */ +static void +xmesa_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) +{ + XMesaContext xmctx = XMESA_CONTEXT(ctx); + XMesaBuffer xmdrawbuf = XMESA_BUFFER(ctx->WinSysDrawBuffer); + XMesaBuffer xmreadbuf = XMESA_BUFFER(ctx->WinSysReadBuffer); + xmesa_check_and_update_buffer_size(xmctx, xmdrawbuf); + xmesa_check_and_update_buffer_size(xmctx, xmreadbuf); + (void) x; + (void) y; + (void) w; + (void) h; +} + + +/** + * Initialize the device driver function table with the functions + * we implement in this driver. + */ +static void +xmesa_init_driver_functions( XMesaVisual xmvisual, + struct dd_function_table *driver ) +{ + driver->Flush = finish_or_flush; + driver->Finish = finish_or_flush; + driver->Viewport = xmesa_viewport; +} + + /** * Create a new XMesaContext. * \param v the XMesaVisual diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c index 56db798869..8a58c3598e 100644 --- a/src/mesa/pipe/xlib/xm_dd.c +++ b/src/mesa/pipe/xlib/xm_dd.c @@ -39,58 +39,3 @@ #include "state_tracker/st_draw.h" -static void -finish_or_flush( GLcontext *ctx ) -{ -#ifdef XFree86Server - /* NOT_NEEDED */ -#else - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - if (xmesa) { - _glthread_LOCK_MUTEX(_xmesa_lock); - XSync( xmesa->display, False ); - _glthread_UNLOCK_MUTEX(_xmesa_lock); - } -#endif - abort(); -} - - -/** - * Called by glViewport. - * This is a good time for us to poll the current X window size and adjust - * our renderbuffers to match the current window size. - * Remember, we have no opportunity to respond to conventional - * X Resize/StructureNotify events since the X driver has no event loop. - * Thus, we poll. - * Note that this trick isn't fool-proof. If the application never calls - * glViewport, our notion of the current window size may be incorrect. - * That problem led to the GLX_MESA_resize_buffers extension. - */ -static void -xmesa_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) -{ - XMesaContext xmctx = XMESA_CONTEXT(ctx); - XMesaBuffer xmdrawbuf = XMESA_BUFFER(ctx->WinSysDrawBuffer); - XMesaBuffer xmreadbuf = XMESA_BUFFER(ctx->WinSysReadBuffer); - xmesa_check_and_update_buffer_size(xmctx, xmdrawbuf); - xmesa_check_and_update_buffer_size(xmctx, xmreadbuf); - (void) x; - (void) y; - (void) w; - (void) h; -} - - -/** - * Initialize the device driver function table with the functions - * we implement in this driver. - */ -void -xmesa_init_driver_functions( XMesaVisual xmvisual, - struct dd_function_table *driver ) -{ - driver->Flush = finish_or_flush; - driver->Finish = finish_or_flush; - driver->Viewport = xmesa_viewport; -} diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index 3cfb2e958d..c4c78be864 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -483,10 +483,6 @@ xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b, extern void xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer); -extern void -xmesa_init_driver_functions( XMesaVisual xmvisual, - struct dd_function_table *driver ); - extern void xmesa_set_renderbuffer_funcs(struct xmesa_renderbuffer *xrb, enum pixel_format pixelformat, GLint depth); -- cgit v1.2.3 From b60f88aec62da7c12f8e93b339826967733e0b9c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 15:59:47 -0600 Subject: remove xm_dd.c from build --- src/mesa/sources | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index c5c07cf442..dec2efe8f8 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -333,7 +333,6 @@ X11_DRIVER_SOURCES = \ pipe/xlib/xfonts.c \ pipe/xlib/xm_api.c \ pipe/xlib/xm_buffer.c \ - pipe/xlib/xm_dd.c \ pipe/xlib/xm_winsys.c \ pipe/xlib/xm_span.c \ pipe/xlib/xm_surface.c -- cgit v1.2.3 From 3feca9b5fc0577fc81eef822c5e4e8ee500203af Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 16:00:14 -0600 Subject: Obsolete --- src/mesa/pipe/xlib/xm_dd.c | 41 ----------------------------------------- 1 file changed, 41 deletions(-) delete mode 100644 src/mesa/pipe/xlib/xm_dd.c (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_dd.c b/src/mesa/pipe/xlib/xm_dd.c deleted file mode 100644 index 8a58c3598e..0000000000 --- a/src/mesa/pipe/xlib/xm_dd.c +++ /dev/null @@ -1,41 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5.2 - * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/** - * \file xm_dd.h - * General device driver functions for Xlib driver. - */ - -#include "glxheader.h" -#include "xmesaP.h" -#include "main/imports.h" -#include "main/mtypes.h" - -#include "pipe/softpipe/sp_context.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_draw.h" - - -- cgit v1.2.3 From 1e3651569b3395d38e1d343f894844c043a66a6f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 16:01:18 -0600 Subject: remove some extension #defines --- src/mesa/pipe/xlib/xmesaP.h | 11 ----------- 1 file changed, 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index c4c78be864..d0e8f32430 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -526,17 +526,6 @@ XMESA_BUFFER(GLframebuffer *b) -#define ENABLE_EXT_texure_compression_s3tc 0 /* SW texture compression */ - -#ifdef XFree86Server -#define ENABLE_EXT_timer_query 0 -#elif defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L -#define ENABLE_EXT_timer_query 1 /* should have 64-bit GLuint64EXT */ -#else -#define ENABLE_EXT_timer_query 0 /* may not have 64-bit GLuint64EXT */ -#endif - - struct pipe_surface; struct pipe_context; -- cgit v1.2.3 From 1103e9c3d0c3ba636837250eea3d50956a3b1e02 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 18:19:46 -0600 Subject: update comments for pipe_context --- src/mesa/pipe/p_context.h | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 33d1878158..4f5e786334 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -31,9 +31,14 @@ #include "p_state.h" struct pipe_state_cache; + + /** - * Software pipeline rendering context. Basically a collection of - * state setting functions, plus VBO drawing entrypoint. + * Gallium rendering context. Basically: + * - state setting functions + * - VBO drawing functions + * - memory region function + * - device queries */ struct pipe_context { struct pipe_winsys *winsys; -- cgit v1.2.3 From 4c7be3faa353aace2237b1f565bd42612002b65a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 18:29:00 -0600 Subject: Clean up xmesa winsys stuff Plug xmesa_is_format_supported() into the softpipe_winsys object. --- src/mesa/pipe/xlib/xm_api.c | 2 +- src/mesa/pipe/xlib/xm_surface.c | 13 ----- src/mesa/pipe/xlib/xm_winsys.c | 111 +++++++++++++++++++--------------------- src/mesa/pipe/xlib/xmesaP.h | 3 -- 4 files changed, 55 insertions(+), 74 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 1693530fbf..a05ff358c0 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1489,8 +1489,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) */ #if 0 mesaCtx->st->pipe->surface_alloc = xmesa_surface_alloc; -#endif mesaCtx->st->pipe->is_format_supported = xmesa_is_format_supported; +#endif mesaCtx->st->pipe->get_tile_rgba = xmesa_get_tile_rgba; mesaCtx->st->pipe->put_tile_rgba = xmesa_put_tile_rgba; diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index e7254d0d5a..b7e9d7811d 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -193,19 +193,6 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) } -boolean -xmesa_is_format_supported(struct pipe_context *pipe, uint format) -{ - switch( format ) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - case PIPE_FORMAT_S_R16_G16_B16_A16: - case PIPE_FORMAT_S8_Z24: - return TRUE; - }; - return FALSE; -} - - /** * Called via pipe->clear() */ diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 36805437f0..3dd15e198d 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -41,18 +41,6 @@ #include "pipe/softpipe/sp_winsys.h" -/** - * XMesa winsys, derived from softpipe winsys. - * NOTE: there's nothing really X-specific in this winsys layer so - * we could probably lift it up somewhere. - */ -struct xm_winsys -{ - struct softpipe_winsys sws; - int foo; /* placeholder */ -}; - - /** * Low-level OS/window system memory buffer */ @@ -82,14 +70,6 @@ pipe_bo( struct xm_buffer *bo ) return (struct pipe_buffer_handle *) bo; } -/* Turn a softpipe winsys into an xm/softpipe winsys: - */ -static inline struct xm_winsys * -xm_winsys(struct softpipe_winsys *sws) -{ - return (struct xm_winsys *) sws; -} - /* Most callbacks map direcly onto dri_bufmgr operations: */ @@ -315,54 +295,71 @@ xm_surface_alloc(struct pipe_winsys *ws, GLuint pipeFormat) -struct xmesa_pipe_winsys -{ - struct pipe_winsys winsys; - XMesaContext xmesa; -}; - +/** + * Create a winsys layer. + * Nothing special for the Xlib driver so no subclassing or anything. + */ static struct pipe_winsys * -xmesa_create_pipe_winsys( XMesaContext xmesa ) +xmesa_create_pipe_winsys(void) { - struct xmesa_pipe_winsys *xws = CALLOC_STRUCT(xmesa_pipe_winsys); + struct pipe_winsys *ws = CALLOC_STRUCT(pipe_winsys); /* Fill in this struct with callbacks that pipe will need to * communicate with the window system, buffer manager, etc. - * - * Pipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate _DriBufferObject. */ - xws->winsys.buffer_create = xm_buffer_create; - xws->winsys.user_buffer_create = xm_user_buffer_create; - xws->winsys.buffer_map = xm_buffer_map; - xws->winsys.buffer_unmap = xm_buffer_unmap; - xws->winsys.buffer_reference = xm_buffer_reference; - xws->winsys.buffer_data = xm_buffer_data; - xws->winsys.buffer_subdata = xm_buffer_subdata; - xws->winsys.buffer_get_subdata = xm_buffer_get_subdata; - - xws->winsys.region_alloc = xm_region_alloc; - xws->winsys.region_release = xm_region_release; - - xws->winsys.surface_alloc = xm_surface_alloc; - - xws->winsys.flush_frontbuffer = xm_flush_frontbuffer; - xws->winsys.wait_idle = xm_wait_idle; - xws->winsys.printf = xm_printf; - xws->winsys.get_name = xm_get_name; - xws->xmesa = xmesa; - - return &xws->winsys; + ws->buffer_create = xm_buffer_create; + ws->user_buffer_create = xm_user_buffer_create; + ws->buffer_map = xm_buffer_map; + ws->buffer_unmap = xm_buffer_unmap; + ws->buffer_reference = xm_buffer_reference; + ws->buffer_data = xm_buffer_data; + ws->buffer_subdata = xm_buffer_subdata; + ws->buffer_get_subdata = xm_buffer_get_subdata; + + ws->region_alloc = xm_region_alloc; + ws->region_release = xm_region_release; + + ws->surface_alloc = xm_surface_alloc; + + ws->flush_frontbuffer = xm_flush_frontbuffer; + ws->wait_idle = xm_wait_idle; + ws->printf = xm_printf; + ws->get_name = xm_get_name; + + return ws; +} + + +static boolean +xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) +{ + switch (format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_S_R16_G16_B16_A16: + case PIPE_FORMAT_S8_Z24: + return TRUE; + default: + return FALSE; + }; +} + + +static struct softpipe_winsys * +xmesa_create_softpipe_winsys(void) +{ + struct softpipe_winsys *spws = CALLOC_STRUCT(softpipe_winsys); + if (spws) { + spws->is_format_supported = xmesa_is_format_supported; + } + return spws; } struct pipe_context * xmesa_create_softpipe(XMesaContext xmesa) { - struct xm_winsys *xm_ws = CALLOC_STRUCT( xm_winsys ); + struct pipe_winsys *pws = xmesa_create_pipe_winsys(); + struct softpipe_winsys *spws = xmesa_create_softpipe_winsys(); - /* Create the softpipe context: - */ - return softpipe_create( xmesa_create_pipe_winsys(xmesa), &xm_ws->sws ); + return softpipe_create( pws, spws ); } diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index d0e8f32430..d42b2b3fb9 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -548,9 +548,6 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); extern struct pipe_surface * xmesa_new_color_surface(struct pipe_context *pipe, GLuint format); -extern boolean -xmesa_is_format_supported(struct pipe_context *pipe, uint format); - extern void xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p); -- cgit v1.2.3 From b7611770ea2988d800c7e39cf9299a98cc5d260c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 31 Oct 2007 19:00:23 -0600 Subject: make winsys objects singletons, comments --- src/mesa/pipe/xlib/xm_api.c | 4 +++ src/mesa/pipe/xlib/xm_winsys.c | 82 ++++++++++++++++++++++++++---------------- 2 files changed, 56 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index a05ff358c0..a4874d3f8f 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -316,6 +316,10 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, _mesa_initialize_framebuffer(&b->mesa_buffer, &vis->mesa_visual); b->mesa_buffer.Delete = xmesa_delete_framebuffer; + /* + * XXX we want to create surfaces for pipe, not renderbuffers for Mesa. + */ + /* * Front renderbuffer */ diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 3dd15e198d..ef8e78f441 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -296,40 +296,53 @@ xm_surface_alloc(struct pipe_winsys *ws, GLuint pipeFormat) /** - * Create a winsys layer. + * Return pointer to a pipe_winsys object. + * For Xlib, this is a singleton object. * Nothing special for the Xlib driver so no subclassing or anything. */ static struct pipe_winsys * -xmesa_create_pipe_winsys(void) +xmesa_get_pipe_winsys(void) { - struct pipe_winsys *ws = CALLOC_STRUCT(pipe_winsys); - - /* Fill in this struct with callbacks that pipe will need to - * communicate with the window system, buffer manager, etc. - */ - ws->buffer_create = xm_buffer_create; - ws->user_buffer_create = xm_user_buffer_create; - ws->buffer_map = xm_buffer_map; - ws->buffer_unmap = xm_buffer_unmap; - ws->buffer_reference = xm_buffer_reference; - ws->buffer_data = xm_buffer_data; - ws->buffer_subdata = xm_buffer_subdata; - ws->buffer_get_subdata = xm_buffer_get_subdata; - - ws->region_alloc = xm_region_alloc; - ws->region_release = xm_region_release; - - ws->surface_alloc = xm_surface_alloc; + static struct pipe_winsys *ws = NULL; - ws->flush_frontbuffer = xm_flush_frontbuffer; - ws->wait_idle = xm_wait_idle; - ws->printf = xm_printf; - ws->get_name = xm_get_name; + if (!ws) { + ws = CALLOC_STRUCT(pipe_winsys); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + */ + ws->buffer_create = xm_buffer_create; + ws->user_buffer_create = xm_user_buffer_create; + ws->buffer_map = xm_buffer_map; + ws->buffer_unmap = xm_buffer_unmap; + ws->buffer_reference = xm_buffer_reference; + ws->buffer_data = xm_buffer_data; + ws->buffer_subdata = xm_buffer_subdata; + ws->buffer_get_subdata = xm_buffer_get_subdata; + + ws->region_alloc = xm_region_alloc; + ws->region_release = xm_region_release; + + ws->surface_alloc = xm_surface_alloc; + + ws->flush_frontbuffer = xm_flush_frontbuffer; + ws->wait_idle = xm_wait_idle; + ws->printf = xm_printf; + ws->get_name = xm_get_name; + } return ws; } +/** + * XXX this depends on the depths supported by the screen (8/16/32/etc). + * Maybe when we're about to create a context/drawable we create a new + * softpipe_winsys object that corresponds to the specified screen... + * + * Also, this query only really matters for on-screen drawables. + * For textures and FBOs we (softpipe) can support any format. + */ static boolean xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) { @@ -344,13 +357,22 @@ xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) } +/** + * Return pointer to a softpipe_winsys object. + * For Xlib, this is a singleton object. + */ static struct softpipe_winsys * -xmesa_create_softpipe_winsys(void) +xmesa_get_softpipe_winsys(void) { - struct softpipe_winsys *spws = CALLOC_STRUCT(softpipe_winsys); - if (spws) { - spws->is_format_supported = xmesa_is_format_supported; + static struct softpipe_winsys *spws = NULL; + + if (!spws) { + spws = CALLOC_STRUCT(softpipe_winsys); + if (spws) { + spws->is_format_supported = xmesa_is_format_supported; + } } + return spws; } @@ -358,8 +380,8 @@ xmesa_create_softpipe_winsys(void) struct pipe_context * xmesa_create_softpipe(XMesaContext xmesa) { - struct pipe_winsys *pws = xmesa_create_pipe_winsys(); - struct softpipe_winsys *spws = xmesa_create_softpipe_winsys(); + struct pipe_winsys *pws = xmesa_get_pipe_winsys(); + struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(); return softpipe_create( pws, spws ); } -- cgit v1.2.3 From 0d6608ee6855e5605efc3bd9fec51ada59e208d9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 10:52:31 -0600 Subject: Implement surface clearing w/out dependency on XMesa/Mesa stuff. --- src/mesa/pipe/xlib/xm_api.c | 10 +++ src/mesa/pipe/xlib/xm_buffer.c | 180 ++++--------------------------------- src/mesa/pipe/xlib/xm_surface.c | 192 ++++++++++++++++++++++++++++++++++++---- src/mesa/pipe/xlib/xm_winsys.c | 2 +- src/mesa/pipe/xlib/xmesaP.h | 47 ++++++---- 5 files changed, 235 insertions(+), 196 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index a4874d3f8f..5e04092215 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -333,6 +333,14 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, b->frontxrb->pixmap = (XMesaPixmap) d; _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_FRONT_LEFT, &b->frontxrb->St.Base); +#if 0 /* sketch... */ + { + struct pipe_surface *front_surf; + front_surf = xmesa_create_front_surface(vis, d); + } +#endif + + /* * Back renderbuffer @@ -1495,6 +1503,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) mesaCtx->st->pipe->surface_alloc = xmesa_surface_alloc; mesaCtx->st->pipe->is_format_supported = xmesa_is_format_supported; #endif + mesaCtx->st->pipe->get_tile = xmesa_get_tile; + mesaCtx->st->pipe->put_tile = xmesa_put_tile; mesaCtx->st->pipe->get_tile_rgba = xmesa_get_tile_rgba; mesaCtx->st->pipe->put_tile_rgba = xmesa_put_tile_rgba; diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index 8ac9fb55b6..cd3b498b10 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -261,146 +261,6 @@ finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) } -/** - * Clear the front or back color buffer, if it's implemented with a pixmap. - */ -static void -clear_pixmap(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); - - assert(xmbuf); - assert(xrb->pixmap); - assert(xmesa); - assert(xmesa->display); - assert(xrb->pixmap); - assert(xmbuf->cleargc); - - XMesaSetForeground( xmesa->display, xmbuf->cleargc, value ); - - XMesaFillRectangle( xmesa->display, xrb->pixmap, xmbuf->cleargc, - 0, 0, xrb->St.Base.Width, xrb->St.Base.Height); -} - - -static void -clear_8bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - GLint i; - for (i = 0; i < height; i++) { - GLubyte *ptr = PIXEL_ADDR1(xrb, 0, i); - MEMSET( ptr, xmesa->clearpixel, width ); - } -} - - -static void -clear_16bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - GLint i, j; - - if (xmesa->swapbytes) { - value = ((value >> 8) & 0x00ff) | ((value << 8) & 0xff00); - } - - for (j = 0; j < height; j++) { - GLushort *ptr2 = PIXEL_ADDR2(xrb, 0, j); - for (i = 0; i < width; i++) { - ptr2[i] = value; - } - } -} - - -/* Optimized code provided by Nozomi Ytow */ -static void -clear_24bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLuint value) -{ - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - const GLubyte r = (value ) & 0xff; - const GLubyte g = (value >> 8) & 0xff; - const GLubyte b = (value >> 16) & 0xff; - - if (r == g && g == b) { - /* same value for all three components (gray) */ - GLint j; - for (j = 0; j < height; j++) { - bgr_t *ptr3 = PIXEL_ADDR3(xrb, 0, j); - MEMSET(ptr3, r, 3 * width); - } - } - else { - /* non-gray clear color */ - GLint i, j; - for (j = 0; j < height; j++) { - bgr_t *ptr3 = PIXEL_ADDR3(xrb, 0, j); - for (i = 0; i < width; i++) { - ptr3->r = r; - ptr3->g = g; - ptr3->b = b; - ptr3++; - } - } - } -} - - -static void -clear_32bit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLuint value) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - const GLuint n = width * height; - GLuint *ptr4 = (GLuint *) xrb->ximage->data; - - if (!xrb->ximage) - return; - - if (xmesa->swapbytes) { - value = ((value >> 24) & 0x000000ff) - | ((value >> 8) & 0x0000ff00) - | ((value << 8) & 0x00ff0000) - | ((value << 24) & 0xff000000); - } - - if (value == 0) { - /* common case */ - _mesa_memset(ptr4, value, 4 * n); - } - else { - GLuint i; - for (i = 0; i < n; i++) - ptr4[i] = value; - } -} - - -static void -clear_nbit_ximage(GLcontext *ctx, struct xmesa_renderbuffer *xrb, GLuint value) -{ - XMesaImage *img = xrb->ximage; - GLint width = xrb->St.Base.Width; - GLint height = xrb->St.Base.Height; - GLint i, j; - for (j = 0; j < height; j++) { - for (i = 0; i < width; i++) { - XMesaPutPixel(img, i, j, value); - } - } -} - - /** * Reallocate renderbuffer storage for front color buffer. * Called via gl_renderbuffer::AllocStorage() @@ -411,6 +271,7 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, { const XMesaContext xmesa = XMESA_CONTEXT(ctx); struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); + struct xmesa_surface *xms = xmesa_surface(xrb->St.surface); /* just clear these to be sure we don't accidentally use them */ xrb->origin1 = NULL; @@ -431,10 +292,13 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xmesa_set_renderbuffer_funcs(xrb, xmesa->pixelformat, xmesa->xm_visual->BitsPerPixel); - xrb->clearFunc = clear_pixmap; - - xrb->St.surface->width = width; - xrb->St.surface->height = height; + /* surface info */ + xms->surface.width = width; + xms->surface.height = height; + xms->display = xmesa->display; + xms->drawable = xrb->drawable; + xms->gc = xrb->Parent->cleargc; + xms->ximage = NULL; return GL_TRUE; } @@ -450,6 +314,7 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, { const XMesaContext xmesa = XMESA_CONTEXT(ctx); struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); + struct xmesa_surface *xms = xmesa_surface(xrb->St.surface); /* reallocate the back buffer XImage or Pixmap */ assert(xrb->Parent); @@ -489,30 +354,20 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xmesa_set_renderbuffer_funcs(xrb, xmesa->pixelformat, xmesa->xm_visual->BitsPerPixel); - switch (xmesa->xm_visual->BitsPerPixel) { - case 8: - xrb->clearFunc = clear_8bit_ximage; - break; - case 16: - xrb->clearFunc = clear_16bit_ximage; - break; - case 24: - xrb->clearFunc = clear_24bit_ximage; - break; - case 32: - xrb->clearFunc = clear_32bit_ximage; - break; - default: - xrb->clearFunc = clear_nbit_ximage; - break; - } - if (!xrb->St.surface || !xrb->St.surface->region) finish_surface_init(ctx, xrb); xrb->St.surface->width = width; xrb->St.surface->height = height; + /* surface info */ + xms->surface.width = width; + xms->surface.height = height; + xms->display = xmesa->display; + xms->drawable = xrb->drawable; + xms->gc = xrb->Parent->cleargc; + xms->ximage = xrb->ximage; + return GL_TRUE; } @@ -561,6 +416,7 @@ xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, xrb->St.surface = xmesa_new_color_surface(pipe, pipeFormat); xms = (struct xmesa_surface *) xrb->St.surface; xms->xrb = xrb; + } return xrb; } diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index b7e9d7811d..a5e0f79494 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -59,13 +59,6 @@ } while(0) -static INLINE struct xmesa_surface * -xmesa_surface(struct pipe_surface *ps) -{ - return (struct xmesa_surface *) ps; -} - - static INLINE struct xmesa_renderbuffer * xmesa_rb(struct pipe_surface *ps) { @@ -77,6 +70,26 @@ xmesa_rb(struct pipe_surface *ps) #define FLIP(Y) Y = xrb->St.Base.Height - (Y) - 1; +void +xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, void *p, int dst_stride) +{ + +} + + +void +xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const void *p, int src_stride) +{ + +} + + + +/** + * XXX rewrite to stop using renderbuffer->GetRow() + */ void xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p) @@ -109,6 +122,9 @@ xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, } +/** + * XXX rewrite to stop using renderbuffer->PutRow() + */ void xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p) @@ -145,6 +161,110 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, } +static void +clear_pixmap_surface(struct pipe_context *pipe, struct pipe_surface *ps, + uint value) +{ + struct xmesa_surface *xms = xmesa_surface(ps); + assert(xms); + assert(xms->display); + assert(xms->drawable); + assert(xms->gc); + XMesaSetForeground( xms->display, xms->gc, value ); + XMesaFillRectangle( xms->display, xms->drawable, xms->gc, + 0, 0, ps->width, ps->height); +} + +static void +clear_nbit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, + uint value) +{ + struct xmesa_surface *xms = xmesa_surface(ps); + int width = xms->surface.width; + int height = xms->surface.height; + int i, j; + for (j = 0; j < height; j++) { + for (i = 0; i < width; i++) { + XMesaPutPixel(xms->ximage, i, j, value); + } + } +} + +static void +clear_8bit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, + uint value) +{ + struct xmesa_surface *xms = xmesa_surface(ps); + memset(xms->ximage->data, + value, + xms->ximage->bytes_per_line * xms->ximage->height); +} + +static void +clear_16bit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, + uint value) +{ + struct xmesa_surface *xms = xmesa_surface(ps); + const int n = xms->ximage->width * xms->ximage->height; + ushort *dst = (ushort *) xms->ximage->data; + int i; + for (i = 0; i < n; i++) { + dst[i] = value; + } +} + + +/* Optimized code provided by Nozomi Ytow */ +static void +clear_24bit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, + uint value) +{ + struct xmesa_surface *xms = xmesa_surface(ps); + const ubyte r = (value ) & 0xff; + const ubyte g = (value >> 8) & 0xff; + const ubyte b = (value >> 16) & 0xff; + + if (r == g && g == b) { + /* same value for all three components (gray) */ + memset(xms->ximage->data, r, + xms->ximage->bytes_per_line * xms->ximage->height); + } + else { + /* non-gray clear color */ + const int n = xms->ximage->width * xms->ximage->height; + int i; + bgr_t *ptr3 = (bgr_t *) xms->ximage->data; + for (i = 0; i < n; i++) { + ptr3->r = r; + ptr3->g = g; + ptr3->b = b; + ptr3++; + } + } +} + +static void +clear_32bit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, + uint value) +{ + struct xmesa_surface *xms = xmesa_surface(ps); + + if (value == 0) { + /* common case */ + memset(xms->ximage->data, value, + xms->ximage->bytes_per_line * xms->ximage->height); + } + else { + const int n = xms->ximage->width * xms->ximage->height; + uint *dst = (uint *) xms->ximage->data; + int i; + for (i = 0; i < n; i++) + dst[i] = value; + } +} + + + /** * Called to create a pipe_surface for each X renderbuffer. @@ -194,13 +314,14 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) /** - * Called via pipe->clear() + * Called via pipe->clear() to clear entire surface to a certain value. + * If the surface is not an X pixmap or XImage, pass the call to + * softpipe_clear(). */ void -xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) +xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, uint value) { - GET_CURRENT_CONTEXT(ctx); - struct xmesa_renderbuffer *xrb = xmesa_rb(ps); + struct xmesa_surface *xms = xmesa_surface(ps); /* XXX actually, we should just discard any cached tiles from this * surface since we don't want to accidentally re-use them after clearing. @@ -219,13 +340,54 @@ xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value) } } - if (xrb) { - /* clearing front/back color buffer */ - xrb->clearFunc(ctx, xrb, value); +#if 1 + (void) clear_8bit_ximage_surface; + (void) clear_24bit_ximage_surface; +#endif + + if (xms->ximage) { + /* back color buffer */ + switch (xms->surface.format) { + case PIPE_FORMAT_U_R5_G6_B5: + clear_16bit_ximage_surface(pipe, ps, value); + break; + case PIPE_FORMAT_U_A8_R8_G8_B8: + clear_32bit_ximage_surface(pipe, ps, value); + break; + default: + clear_nbit_ximage_surface(pipe, ps, value); + break; + } + } + else if (xms->drawable) { + /* front color buffer */ + clear_pixmap_surface(pipe, ps, value); } else { - /* clearing other buffer */ + /* other kind of buffer */ softpipe_clear(pipe, ps, value); } } + +/** XXX unfinished sketch... */ +struct pipe_surface * +xmesa_create_front_surface(XMesaVisual vis, Window win) +{ + struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + if (!xms) { + return NULL; + } + + xms->display = vis->display; + xms->drawable = win; + + xms->surface.format = PIPE_FORMAT_U_A8_R8_G8_B8; + xms->surface.refcount = 1; +#if 0 + xms->surface.region = pipe->winsys->region_alloc(pipe->winsys, + 1, 0, 0, 0x0); +#endif + return &xms->surface; +} + diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index ef8e78f441..e6e98ed396 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -341,7 +341,7 @@ xmesa_get_pipe_winsys(void) * softpipe_winsys object that corresponds to the specified screen... * * Also, this query only really matters for on-screen drawables. - * For textures and FBOs we (softpipe) can support any format. + * For textures and FBOs we (softpipe) can support any format.o */ static boolean xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index d42b2b3fb9..4454ab8003 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -52,12 +52,6 @@ typedef struct { struct xmesa_renderbuffer; -/* Function pointer for clearing color buffers */ -typedef void (*ClearFunc)( GLcontext *ctx, struct xmesa_renderbuffer *xrb, - GLuint value ); - - - /** Framebuffer pixel formats */ enum pixel_format { @@ -118,7 +112,7 @@ struct xmesa_visual { /** - * Context info, dDerived from GLcontext. + * Context info, derived from GLcontext. * Basically corresponds to a GLXContext. */ struct xmesa_context { @@ -168,11 +162,7 @@ typedef enum { */ struct xmesa_renderbuffer { -#if 0 - struct gl_renderbuffer Base; /* Base class */ -#else - struct st_renderbuffer St; /**< Base class */ -#endif + struct st_renderbuffer St; /**< Base class (XXX temporary?) */ XMesaBuffer Parent; /**< The XMesaBuffer this renderbuffer belongs to */ XMesaDrawable drawable; /* Usually the X window ID */ @@ -189,10 +179,6 @@ struct xmesa_renderbuffer GLint width4; GLint bottom; /* used for FLIP macro, equals height - 1 */ - - ClearFunc clearFunc; - - void *pSurface; /** pipe surface */ }; @@ -526,18 +512,30 @@ XMESA_BUFFER(GLframebuffer *b) -struct pipe_surface; struct pipe_context; struct xmesa_surface { struct pipe_surface surface; struct xmesa_renderbuffer *xrb; + XMesaDisplay *display; + BufferType type; + XMesaDrawable drawable; + XMesaImage *ximage; + XMesaGC gc; }; +/** Cast wrapper */ +static INLINE struct xmesa_surface * +xmesa_surface(struct pipe_surface *ps) +{ + return (struct xmesa_surface *) ps; +} + + extern void -xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, GLuint value); +xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, uint value); extern struct pipe_context * xmesa_create_softpipe(XMesaContext xm); @@ -548,6 +546,15 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); extern struct pipe_surface * xmesa_new_color_surface(struct pipe_context *pipe, GLuint format); + +extern void +xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, void *p, int dst_stride); + +extern void +xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const void *p, int src_stride); + extern void xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p); @@ -556,4 +563,8 @@ extern void xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p); + +extern struct pipe_surface * +xmesa_create_front_surface(XMesaVisual vis, Window win); + #endif -- cgit v1.2.3 From a33308898666e89f2443e7eb4f1d09ae792b620a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 11:23:50 -0600 Subject: implement get/put_tile() for xlib driver --- src/mesa/pipe/softpipe/sp_surface.c | 10 ++- src/mesa/pipe/softpipe/sp_surface.h | 9 +++ src/mesa/pipe/xlib/xm_surface.c | 119 ++++++++++++++++++++++++++++++++++++ 3 files changed, 132 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 3aa72ef569..ccaf38de3e 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -502,9 +502,8 @@ softpipe_get_tex_surface(struct pipe_context *pipe, /** * Move raw block of pixels from surface to user memory. */ -static void -softpipe_get_tile(struct pipe_context *pipe, - struct pipe_surface *ps, +void +softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, void *p, int dst_stride) { @@ -535,9 +534,8 @@ softpipe_get_tile(struct pipe_context *pipe, /** * Move raw block of pixels from user memory to surface. */ -static void -softpipe_put_tile(struct pipe_context *pipe, - struct pipe_surface *ps, +void +softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const void *p, int src_stride) { diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index 7e08ce7a2b..cf87e1a92c 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -45,6 +45,15 @@ softpipe_get_tex_surface(struct pipe_context *pipe, unsigned face, unsigned level, unsigned zslice); +extern void +softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, void *p, int dst_stride); + +extern void +softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride); + extern void softpipe_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index a5e0f79494..94fc2cae68 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -70,19 +70,138 @@ xmesa_rb(struct pipe_surface *ps) #define FLIP(Y) Y = xrb->St.Base.Height - (Y) - 1; +/** + * Return raw pixels from pixmap or XImage. + */ void xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, void *p, int dst_stride) { + struct xmesa_surface *xms = xmesa_surface(ps); + XMesaImage *ximage = NULL; + ubyte *dst = (ubyte *) p; + uint i; + + if (!xms->drawable && !xms->ximage) { + /* not an X surface */ + softpipe_get_tile(pipe, ps, x, y, w, h, p, dst_stride); + return; + } + + if (!xms->ximage) { + /* XImage = pixmap data */ + assert(xms->drawable); + ximage = XGetImage(xms->display, xms->drawable, x, y, w, h, + AllPlanes, ZPixmap); + x = y = 0; + } + else { + ximage = xms->ximage; + } + + switch (ps->format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + if (!dst_stride) { + dst_stride = w * 4; + } + for (i = 0; i < h; i++) { + memcpy(dst, ximage->data + y * ximage->bytes_per_line + x * 4, 4 * w); + dst += dst_stride; + } + break; + case PIPE_FORMAT_U_R5_G6_B5: + if (!dst_stride) { + dst_stride = w * 2; + } + for (i = 0; i < h; i++) { + memcpy(dst, ximage->data + y * ximage->bytes_per_line + x * 2, 4 * 2); + dst += dst_stride; + } + break; + default: + assert(0); + } + if (!xms->ximage) { + XMesaDestroyImage(ximage); + } } +/** + * Put raw pixels into pixmap or XImage. + */ void xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const void *p, int src_stride) { + struct xmesa_surface *xms = xmesa_surface(ps); + const ubyte *src = (const ubyte *) p; + XMesaImage *ximage; + if (!xms->drawable && !xms->ximage) { + /* not an X surface */ + softpipe_put_tile(pipe, ps, x, y, w, h, p, src_stride); + return; + } + + if (xms->ximage) { + /* put to ximage */ + ximage = xms->ximage; + char *dst; + int i; + + switch (ps->format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + if (!src_stride) { + src_stride = w * 4; + } + dst = ximage->data + y * ximage->bytes_per_line + x * 4; + for (i = 0; i < h; i++) { + memcpy(dst, src, w * 4); + dst += ximage->bytes_per_line; + src += src_stride; + } + break; + case PIPE_FORMAT_U_R5_G6_B5: + if (!src_stride) { + src_stride = w * 2; + } + dst = ximage->data + y * ximage->bytes_per_line + x * 2; + for (i = 0; i < h; i++) { + memcpy(dst, src, w * 2); + dst += ximage->bytes_per_line; + src += src_stride; + } + break; + default: + assert(0); + } + } + else { + /* put to pixmap/window */ + /* Create temp XImage for data */ +#ifdef XFree86Server + ximage = XMesaCreateImage(GET_VISUAL_DEPTH(v), w, h, p); +#else + XVisualInfo *visinfo = xms->xrb->Parent->xm_visual->visinfo; + ximage = XCreateImage(xms->display, + visinfo->visual, + visinfo->depth, + ZPixmap, 0, /* format, offset */ + (char *) p, /* data */ + w, h, /* width, height */ + 32, /* bitmap_pad */ + 0); /* bytes_per_line */ +#endif + + /* send XImage data to pixmap */ + XPutImage(xms->display, xms->drawable, xms->gc, + ximage, 0, 0, x, y, w, h); + /* clean-up */ + ximage->data = NULL; /* prevents freeing user data at 'p' */ + XMesaDestroyImage(ximage); + } } -- cgit v1.2.3 From ba0b1bdc332295743874c2e1fa3a2fc0528c70ac Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 12:08:01 -0600 Subject: reimplement xmesa_put_tile_rgba() w/out span funcs --- src/mesa/pipe/xlib/xm_surface.c | 99 ++++++++++++++++++++++++++++++++--------- 1 file changed, 78 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 94fc2cae68..9ca6757489 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -248,34 +248,91 @@ void xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p) { + const uint x0 = x, y0 = y; struct xmesa_surface *xms = xmesa_surface(ps); - struct xmesa_renderbuffer *xrb = xms->xrb; + XMesaImage *ximage; + uint i, j; - if (xrb) { - /* this is a front/back color buffer */ - GLubyte tmp[MAX_WIDTH * 4]; - GLuint i, j; - uint w0 = w; - GET_CURRENT_CONTEXT(ctx); - CLIP_TILE; - FLIP(y); - for (i = 0; i < h; i++) { - for (j = 0; j < w * 4; j++) { - UNCLAMPED_FLOAT_TO_UBYTE(tmp[j], p[j]); + CLIP_TILE; + + if (!xms->drawable && !xms->ximage) { + /* not an X surface */ + softpipe_put_tile_rgba(pipe, ps, x, y, w, h, p); + return; + } + + if (!xms->ximage) { + /* create temp XImage */ + char *data = (char *) malloc(w * h * 4); +#ifdef XFree86Server + ximage = XMesaCreateImage(GET_VISUAL_DEPTH(v), w, h, data); +#else + XVisualInfo *visinfo = xms->xrb->Parent->xm_visual->visinfo; + ximage = XCreateImage(xms->display, + visinfo->visual, + visinfo->depth, + ZPixmap, 0, /* format, offset */ + data, /* data */ + w, h, /* width, height */ + 32, /* bitmap_pad */ + 0); /* bytes_per_line */ +#endif + x = y = 0; + } + else { + ximage = xms->ximage; + } + + /* convert floats to ximage's format */ + switch (ps->format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + { + uint *dst + = (uint *) (ximage->data + y * ximage->bytes_per_line + x * 4); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + ubyte r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, p[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, p[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, p[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, p[3]); + dst[j] = PACK_8A8R8G8B(r, g, b, a); + p += 4; + } + dst += ximage->width; } - xrb->St.Base.PutRow(ctx, &xrb->St.Base, w, x, y - i, tmp, NULL); - p += w0 * 4; } -#if 0 /* debug: flush */ + break; + case PIPE_FORMAT_U_R5_G6_B5: { - XMesaContext xm = XMESA_CONTEXT(ctx); - XSync(xm->display, 0); + ushort *dst = + (ushort *) (ximage->data + y * ximage->bytes_per_line + x * 2); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + ubyte r, g, b; + UNCLAMPED_FLOAT_TO_UBYTE(r, p[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, p[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, p[2]); + dst[j] = PACK_5R6G5B(r, g, b); + p += 4; + } + dst += ximage->width; + } } -#endif + break; + + default: + assert(0); } - else { - /* other softpipe surface */ - softpipe_put_tile_rgba(pipe, ps, x, y, w, h, p); + + if (!xms->ximage) { + /* send XImage data to pixmap */ + XPutImage(xms->display, xms->drawable, xms->gc, + ximage, 0, 0, x0, y0, w, h); + /* clean-up */ + free(ximage->data); + ximage->data = NULL; + XMesaDestroyImage(ximage); } } -- cgit v1.2.3 From d0dde6e26c142c27408a0b52c153b571b8737d29 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 12:24:36 -0600 Subject: rewrite xmesa_get_tile_rgba() to not use span funcs --- src/mesa/pipe/xlib/xm_surface.c | 97 +++++++++++++++++++++++++++++++---------- 1 file changed, 73 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 9ca6757489..782a1f29bc 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -88,6 +88,8 @@ xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, return; } + CLIP_TILE; + if (!xms->ximage) { /* XImage = pixmap data */ assert(xms->drawable); @@ -145,6 +147,8 @@ xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, return; } + CLIP_TILE; + if (xms->ximage) { /* put to ximage */ ximage = xms->ximage; @@ -205,45 +209,90 @@ xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, } - -/** - * XXX rewrite to stop using renderbuffer->GetRow() - */ void xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p) { struct xmesa_surface *xms = xmesa_surface(ps); - struct xmesa_renderbuffer *xrb = xms->xrb; + XMesaImage *ximage = NULL; + float *pRow = p; + uint i, j; - if (xrb) { - /* this is a front/back color buffer */ - GLubyte tmp[MAX_WIDTH * 4]; - GLuint i, j; - uint w0 = w; - GET_CURRENT_CONTEXT(ctx); + if (!xms->drawable && !xms->ximage) { + /* not an X surface */ + softpipe_get_tile_rgba(pipe, ps, x, y, w, h, p); + return; + } - CLIP_TILE; + CLIP_TILE; - FLIP(y); - for (i = 0; i < h; i++) { - xrb->St.Base.GetRow(ctx, &xrb->St.Base, w, x, y - i, tmp); - for (j = 0; j < w * 4; j++) { - p[j] = UBYTE_TO_FLOAT(tmp[j]); + if (!xms->ximage) { + /* XImage = pixmap data */ + assert(xms->drawable); + ximage = XGetImage(xms->display, xms->drawable, x, y, w, h, + AllPlanes, ZPixmap); + x = y = 0; + } + else { + ximage = xms->ximage; + } + + switch (ps->format) { + case PIPE_FORMAT_U_A8_R8_G8_B8: + { + const uint *src + = (uint *) (ximage->data + y * ximage->bytes_per_line + x * 4); + for (i = 0; i < h; i++) { + float *p = pRow; + for (j = 0; j < w; j++) { + uint pix = src[j]; + ubyte r = ((pix >> 16) & 0xff); + ubyte g = ((pix >> 8) & 0xff); + ubyte b = ( pix & 0xff); + ubyte a = ((pix >> 24) & 0xff); + p[0] = UBYTE_TO_FLOAT(r); + p[1] = UBYTE_TO_FLOAT(g); + p[2] = UBYTE_TO_FLOAT(b); + p[3] = UBYTE_TO_FLOAT(a); + p += 4; + } + src += ximage->width; + pRow += 4 * w; } - p += w0 * 4; } + break; + case PIPE_FORMAT_U_R5_G6_B5: + { + ushort *src + = (ushort *) (ximage->data + y * ximage->bytes_per_line + x * 2); + for (i = 0; i < h; i++) { + float *p = pRow; + for (j = 0; j < w; j++) { + uint pix = src[j]; + ubyte r = (pix >> 8) | ((pix >> 13) & 0x7); + ubyte g = (pix >> 3) | ((pix >> 9) & 0x3); + ubyte b = ((pix & 0x1f) << 3) | ((pix >> 2) & 0x3); + p[0] = UBYTE_TO_FLOAT(r); + p[1] = UBYTE_TO_FLOAT(g); + p[2] = UBYTE_TO_FLOAT(b); + p[3] = 1.0; + p += 4; + } + src += ximage->width; + pRow += 4 * w; + } + } + break; + default: + assert(0); } - else { - /* other softpipe surface */ - softpipe_get_tile_rgba(pipe, ps, x, y, w, h, p); + + if (!xms->ximage) { + XMesaDestroyImage(ximage); } } -/** - * XXX rewrite to stop using renderbuffer->PutRow() - */ void xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p) -- cgit v1.2.3 From cd132354dd5a6b778402a2e81d54910ac6c7fb8d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 12:26:37 -0600 Subject: move CLIP_TILE, add comments --- src/mesa/pipe/xlib/xm_surface.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 782a1f29bc..79adc552e4 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -100,7 +100,8 @@ xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, else { ximage = xms->ximage; } - + + /* this could be optimized/simplified */ switch (ps->format) { case PIPE_FORMAT_U_A8_R8_G8_B8: if (!dst_stride) { @@ -155,6 +156,7 @@ xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, char *dst; int i; + /* this could be optimized/simplified */ switch (ps->format) { case PIPE_FORMAT_U_A8_R8_G8_B8: if (!src_stride) { @@ -302,14 +304,14 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, XMesaImage *ximage; uint i, j; - CLIP_TILE; - if (!xms->drawable && !xms->ximage) { /* not an X surface */ softpipe_put_tile_rgba(pipe, ps, x, y, w, h, p); return; } + CLIP_TILE; + if (!xms->ximage) { /* create temp XImage */ char *data = (char *) malloc(w * h * 4); -- cgit v1.2.3 From ef35f664ca243529577c1b5f232f9ef8575621f5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 12:36:09 -0600 Subject: fix some tile clipping bugs --- src/mesa/pipe/xlib/xm_surface.c | 33 +++++++++++++++++++++------------ 1 file changed, 21 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 79adc552e4..c2a8fab36c 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -77,6 +77,7 @@ void xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, void *p, int dst_stride) { + const uint w0 = w; struct xmesa_surface *xms = xmesa_surface(ps); XMesaImage *ximage = NULL; ubyte *dst = (ubyte *) p; @@ -105,7 +106,7 @@ xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, switch (ps->format) { case PIPE_FORMAT_U_A8_R8_G8_B8: if (!dst_stride) { - dst_stride = w * 4; + dst_stride = w0 * 4; } for (i = 0; i < h; i++) { memcpy(dst, ximage->data + y * ximage->bytes_per_line + x * 4, 4 * w); @@ -114,7 +115,7 @@ xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, break; case PIPE_FORMAT_U_R5_G6_B5: if (!dst_stride) { - dst_stride = w * 2; + dst_stride = w0 * 2; } for (i = 0; i < h; i++) { memcpy(dst, ximage->data + y * ximage->bytes_per_line + x * 2, 4 * 2); @@ -138,6 +139,7 @@ void xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const void *p, int src_stride) { + const uint w0 = w; struct xmesa_surface *xms = xmesa_surface(ps); const ubyte *src = (const ubyte *) p; XMesaImage *ximage; @@ -160,7 +162,7 @@ xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, switch (ps->format) { case PIPE_FORMAT_U_A8_R8_G8_B8: if (!src_stride) { - src_stride = w * 4; + src_stride = w0 * 4; } dst = ximage->data + y * ximage->bytes_per_line + x * 4; for (i = 0; i < h; i++) { @@ -171,7 +173,7 @@ xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, break; case PIPE_FORMAT_U_R5_G6_B5: if (!src_stride) { - src_stride = w * 2; + src_stride = w0 * 2; } dst = ximage->data + y * ximage->bytes_per_line + x * 2; for (i = 0; i < h; i++) { @@ -213,16 +215,17 @@ xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, void xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *p) + uint x, uint y, uint w, uint h, float *pixels) { + const uint w0 = w; struct xmesa_surface *xms = xmesa_surface(ps); XMesaImage *ximage = NULL; - float *pRow = p; + float *pRow = pixels; uint i, j; if (!xms->drawable && !xms->ximage) { /* not an X surface */ - softpipe_get_tile_rgba(pipe, ps, x, y, w, h, p); + softpipe_get_tile_rgba(pipe, ps, x, y, w, h, pixels); return; } @@ -259,7 +262,7 @@ xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, p += 4; } src += ximage->width; - pRow += 4 * w; + pRow += 4 * w0; } } break; @@ -281,7 +284,7 @@ xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, p += 4; } src += ximage->width; - pRow += 4 * w; + pRow += 4 * w0; } } break; @@ -297,16 +300,16 @@ xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, void xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *p) + uint x, uint y, uint w, uint h, const float *pixels) { - const uint x0 = x, y0 = y; + const uint x0 = x, y0 = y, w0 = w; struct xmesa_surface *xms = xmesa_surface(ps); XMesaImage *ximage; uint i, j; if (!xms->drawable && !xms->ximage) { /* not an X surface */ - softpipe_put_tile_rgba(pipe, ps, x, y, w, h, p); + softpipe_put_tile_rgba(pipe, ps, x, y, w, h, pixels); return; } @@ -340,7 +343,9 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, { uint *dst = (uint *) (ximage->data + y * ximage->bytes_per_line + x * 4); + const float *pRow = pixels; for (i = 0; i < h; i++) { + const float *p = pRow; for (j = 0; j < w; j++) { ubyte r, g, b, a; UNCLAMPED_FLOAT_TO_UBYTE(r, p[0]); @@ -351,6 +356,7 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, p += 4; } dst += ximage->width; + pRow += 4 * w0; } } break; @@ -358,7 +364,9 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, { ushort *dst = (ushort *) (ximage->data + y * ximage->bytes_per_line + x * 2); + const float *pRow = pixels; for (i = 0; i < h; i++) { + const float *p = pRow; for (j = 0; j < w; j++) { ubyte r, g, b; UNCLAMPED_FLOAT_TO_UBYTE(r, p[0]); @@ -368,6 +376,7 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, p += 4; } dst += ximage->width; + pRow += 4 * w0; } } break; -- cgit v1.2.3 From ea816fd6dbba878c74e1b428b3f8180020b4c214 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 12:37:13 -0600 Subject: move dither kernels to xm_surface.c --- src/mesa/pipe/xlib/xm_span.c | 21 --------------------- src/mesa/pipe/xlib/xm_surface.c | 20 ++++++++++++++++++++ 2 files changed, 20 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_span.c b/src/mesa/pipe/xlib/xm_span.c index b3dbe6a9c8..e5d3721730 100644 --- a/src/mesa/pipe/xlib/xm_span.c +++ b/src/mesa/pipe/xlib/xm_span.c @@ -30,27 +30,6 @@ #include "xmesaP.h" - -/* - * Dithering kernels and lookup tables. - */ - -const int xmesa_kernel8[DITH_DY * DITH_DX] = { - 0 * MAXC, 8 * MAXC, 2 * MAXC, 10 * MAXC, - 12 * MAXC, 4 * MAXC, 14 * MAXC, 6 * MAXC, - 3 * MAXC, 11 * MAXC, 1 * MAXC, 9 * MAXC, - 15 * MAXC, 7 * MAXC, 13 * MAXC, 5 * MAXC, -}; - -const int xmesa_kernel1[16] = { - 0*47, 9*47, 4*47, 12*47, /* 47 = (255*3)/16 */ - 6*47, 2*47, 14*47, 8*47, - 10*47, 1*47, 5*47, 11*47, - 7*47, 13*47, 3*47, 15*47 -}; - - - /* * The following functions are used to trap XGetImage() calls which * generate BadMatch errors if the drawable isn't mapped. diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index c2a8fab36c..5018067906 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -50,6 +50,26 @@ #include "state_tracker/st_context.h" +/* + * Dithering kernels and lookup tables. + */ + +const int xmesa_kernel8[DITH_DY * DITH_DX] = { + 0 * MAXC, 8 * MAXC, 2 * MAXC, 10 * MAXC, + 12 * MAXC, 4 * MAXC, 14 * MAXC, 6 * MAXC, + 3 * MAXC, 11 * MAXC, 1 * MAXC, 9 * MAXC, + 15 * MAXC, 7 * MAXC, 13 * MAXC, 5 * MAXC, +}; + +const int xmesa_kernel1[16] = { + 0*47, 9*47, 4*47, 12*47, /* 47 = (255*3)/16 */ + 6*47, 2*47, 14*47, 8*47, + 10*47, 1*47, 5*47, 11*47, + 7*47, 13*47, 3*47, 15*47 +}; + + + #define CLIP_TILE \ do { \ if (x + w > ps->width) \ -- cgit v1.2.3 From b2fabbc1970ad84074fa9c90b44c6943bdcec4d4 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 12:54:30 -0600 Subject: xget_image() wrapper to catch BadMatch errors --- src/mesa/pipe/xlib/xm_surface.c | 98 +++++++++++++++++++++++++++++++++++++++-- 1 file changed, 94 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 5018067906..1ac96513d8 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -79,6 +79,93 @@ const int xmesa_kernel1[16] = { } while(0) + +/* + * The following functions are used to trap XGetImage() calls which + * generate BadMatch errors if the drawable isn't mapped. + */ + +#ifndef XFree86Server +static int caught_xgetimage_error = 0; +static int (*old_xerror_handler)( XMesaDisplay *dpy, XErrorEvent *ev ); +static unsigned long xgetimage_serial; + +/* + * This is the error handler which will be called if XGetImage fails. + */ +static int xgetimage_error_handler( XMesaDisplay *dpy, XErrorEvent *ev ) +{ + if (ev->serial==xgetimage_serial && ev->error_code==BadMatch) { + /* caught the expected error */ + caught_xgetimage_error = 0; + } + else { + /* call the original X error handler, if any. otherwise ignore */ + if (old_xerror_handler) { + (*old_xerror_handler)( dpy, ev ); + } + } + return 0; +} + + +/* + * Call this right before XGetImage to setup error trap. + */ +static void catch_xgetimage_errors( XMesaDisplay *dpy ) +{ + xgetimage_serial = NextRequest( dpy ); + old_xerror_handler = XSetErrorHandler( xgetimage_error_handler ); + caught_xgetimage_error = 0; +} + + +/* + * Call this right after XGetImage to check if an error occured. + */ +static int check_xgetimage_errors( void ) +{ + /* restore old handler */ + (void) XSetErrorHandler( old_xerror_handler ); + /* return 0=no error, 1=error caught */ + return caught_xgetimage_error; +} +#endif + + +/** + * Wrapper for XGetImage() that catches BadMatch errors that can occur + * when the window is unmapped or the x/y/w/h extend beyond the window + * bounds. + * If build into xserver, wrap the internal GetImage method. + */ +static XMesaImage * +xget_image(XMesaDisplay *dpy, Drawable d, int x, int y, uint w, uint h) +{ +#ifdef XFree86Server + uint bpp = 4; /* XXX fix this */ + XMesaImage *ximage = (XMesaImage *) malloc(sizeof(XMesaImage)); + if (ximage) { + ximage->data = malloc(width * height * bpp); + } + (*dpy->GetImage)(d, x, y, w, h, ZPixmap, ~0L, (pointer)ximage->data); + ximage->width = w; + ximage->height = h; + ximage->bytes_per_row = w * bpp; + return ximage; +#else + int error; + XMesaImage *ximage; + catch_xgetimage_errors(dpy); + ximage = XGetImage(dpy, d, x, y, w, h, AllPlanes, ZPixmap); + error = check_xgetimage_errors(); + return ximage; +#endif +} + + + + static INLINE struct xmesa_renderbuffer * xmesa_rb(struct pipe_surface *ps) { @@ -90,6 +177,7 @@ xmesa_rb(struct pipe_surface *ps) #define FLIP(Y) Y = xrb->St.Base.Height - (Y) - 1; + /** * Return raw pixels from pixmap or XImage. */ @@ -114,8 +202,9 @@ xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, if (!xms->ximage) { /* XImage = pixmap data */ assert(xms->drawable); - ximage = XGetImage(xms->display, xms->drawable, x, y, w, h, - AllPlanes, ZPixmap); + ximage = xget_image(xms->display, xms->drawable, x, y, w, h); + if (!ximage) + return; x = y = 0; } else { @@ -254,8 +343,9 @@ xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, if (!xms->ximage) { /* XImage = pixmap data */ assert(xms->drawable); - ximage = XGetImage(xms->display, xms->drawable, x, y, w, h, - AllPlanes, ZPixmap); + ximage = xget_image(xms->display, xms->drawable, x, y, w, h); + if (!ximage) + return; x = y = 0; } else { -- cgit v1.2.3 From 1c8d064c39468c8b1ae1e56074a8e470375bd1e0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 12:55:20 -0600 Subject: remove unused funcs, macros --- src/mesa/pipe/xlib/xm_surface.c | 13 ------------- 1 file changed, 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 1ac96513d8..6250e75de8 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -165,19 +165,6 @@ xget_image(XMesaDisplay *dpy, Drawable d, int x, int y, uint w, uint h) - -static INLINE struct xmesa_renderbuffer * -xmesa_rb(struct pipe_surface *ps) -{ - struct xmesa_surface *xms = xmesa_surface(ps); - return xms->xrb; -} - - -#define FLIP(Y) Y = xrb->St.Base.Height - (Y) - 1; - - - /** * Return raw pixels from pixmap or XImage. */ -- cgit v1.2.3 From 618089555353070f73dc8ad3b52578db3bedc294 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 12:57:26 -0600 Subject: disable/omit renderbuffer span code --- src/mesa/sources | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index dec2efe8f8..8cbc1c6993 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -334,7 +334,6 @@ X11_DRIVER_SOURCES = \ pipe/xlib/xm_api.c \ pipe/xlib/xm_buffer.c \ pipe/xlib/xm_winsys.c \ - pipe/xlib/xm_span.c \ pipe/xlib/xm_surface.c OSMESA_DRIVER_SOURCES = \ -- cgit v1.2.3 From a00bd3fc2f85997d0573ea0e4773ef1e290e1128 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 12:57:48 -0600 Subject: disable/omit renderbuffer span code --- src/mesa/pipe/xlib/xm_buffer.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index cd3b498b10..b4a05ac685 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -289,8 +289,10 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, if (!xrb->St.surface || !xrb->St.surface->region) finish_surface_init(ctx, xrb); +#if 0 xmesa_set_renderbuffer_funcs(xrb, xmesa->pixelformat, xmesa->xm_visual->BitsPerPixel); +#endif /* surface info */ xms->surface.width = width; @@ -351,8 +353,10 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xrb->origin4 = NULL; } +#if 0 xmesa_set_renderbuffer_funcs(xrb, xmesa->pixelformat, xmesa->xm_visual->BitsPerPixel); +#endif if (!xrb->St.surface || !xrb->St.surface->region) finish_surface_init(ctx, xrb); -- cgit v1.2.3 From ea50025c76f20259cc6c79bceacd8945f305b738 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 13:06:29 -0600 Subject: remove xmesa_set_renderbuffer_funcs() --- src/mesa/pipe/xlib/xmesaP.h | 3 --- 1 file changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index 4454ab8003..506db859bb 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -469,9 +469,6 @@ xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b, extern void xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer); -extern void -xmesa_set_renderbuffer_funcs(struct xmesa_renderbuffer *xrb, - enum pixel_format pixelformat, GLint depth); extern void xmesa_destroy_buffers_on_display(XMesaDisplay *dpy); -- cgit v1.2.3 From c4e9bfedbc78e4e38d3ca47d6e5a756f9a07416f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 13:06:53 -0600 Subject: Obsolete file --- src/mesa/pipe/xlib/xm_span.c | 4552 ------------------------------------------ 1 file changed, 4552 deletions(-) delete mode 100644 src/mesa/pipe/xlib/xm_span.c (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_span.c b/src/mesa/pipe/xlib/xm_span.c deleted file mode 100644 index e5d3721730..0000000000 --- a/src/mesa/pipe/xlib/xm_span.c +++ /dev/null @@ -1,4552 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.3 - * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#include "glxheader.h" -#include "colormac.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" -#include "xmesaP.h" - - -/* - * The following functions are used to trap XGetImage() calls which - * generate BadMatch errors if the drawable isn't mapped. - */ - -#ifndef XFree86Server -static int caught_xgetimage_error = 0; -static int (*old_xerror_handler)( XMesaDisplay *dpy, XErrorEvent *ev ); -static unsigned long xgetimage_serial; - -/* - * This is the error handler which will be called if XGetImage fails. - */ -static int xgetimage_error_handler( XMesaDisplay *dpy, XErrorEvent *ev ) -{ - if (ev->serial==xgetimage_serial && ev->error_code==BadMatch) { - /* caught the expected error */ - caught_xgetimage_error = 0; - } - else { - /* call the original X error handler, if any. otherwise ignore */ - if (old_xerror_handler) { - (*old_xerror_handler)( dpy, ev ); - } - } - return 0; -} - - -/* - * Call this right before XGetImage to setup error trap. - */ -static void catch_xgetimage_errors( XMesaDisplay *dpy ) -{ - xgetimage_serial = NextRequest( dpy ); - old_xerror_handler = XSetErrorHandler( xgetimage_error_handler ); - caught_xgetimage_error = 0; -} - - -/* - * Call this right after XGetImage to check if an error occured. - */ -static int check_xgetimage_errors( void ) -{ - /* restore old handler */ - (void) XSetErrorHandler( old_xerror_handler ); - /* return 0=no error, 1=error caught */ - return caught_xgetimage_error; -} -#endif - - -/* - * Read a pixel from an X drawable. - */ -static unsigned long read_pixel( XMesaDisplay *dpy, - XMesaDrawable d, int x, int y ) -{ - unsigned long p; -#ifndef XFree86Server - XMesaImage *pixel = NULL; - int error; - - catch_xgetimage_errors( dpy ); - pixel = XGetImage( dpy, d, x, y, 1, 1, AllPlanes, ZPixmap ); - error = check_xgetimage_errors(); - if (pixel && !error) { - p = XMesaGetPixel( pixel, 0, 0 ); - } - else { - p = 0; - } - if (pixel) { - XMesaDestroyImage( pixel ); - } -#else - (*dpy->GetImage)(d, x, y, 1, 1, ZPixmap, ~0L, (pointer)&p); -#endif - return p; -} - - - -/* - * The Mesa library needs to be able to draw pixels in a number of ways: - * 1. RGB vs Color Index - * 2. as horizontal spans (polygons, images) vs random locations (points, - * lines) - * 3. different color per-pixel or same color for all pixels - * - * Furthermore, the X driver needs to support rendering to 3 possible - * "buffers", usually one, but sometimes two at a time: - * 1. The front buffer as an X window - * 2. The back buffer as a Pixmap - * 3. The back buffer as an XImage - * - * Finally, if the back buffer is an XImage, we can avoid using XPutPixel and - * optimize common cases such as 24-bit and 8-bit modes. - * - * By multiplication, there's at least 48 possible combinations of the above. - * - * Below are implementations of the most commonly used combinations. They are - * accessed through function pointers which get initialized here and are used - * directly from the Mesa library. The 8 function pointers directly correspond - * to the first 3 cases listed above. - * - * - * The function naming convention is: - * - * [put|get]_[mono]_[row|values]_[format]_[pixmap|ximage] - * - * New functions optimized for specific cases can be added without too much - * trouble. An example might be the 24-bit TrueColor mode 8A8R8G8B which is - * found on IBM RS/6000 X servers. - */ - - - - -/**********************************************************************/ -/*** Write COLOR SPAN functions ***/ -/**********************************************************************/ - - -#define PUT_ROW_ARGS \ - GLcontext *ctx, \ - struct gl_renderbuffer *rb, \ - GLuint n, GLint x, GLint y, \ - const void *values, const GLubyte mask[] - -#define RGB_SPAN_ARGS \ - GLcontext *ctx, \ - struct gl_renderbuffer *rb, \ - GLuint n, GLint x, GLint y, \ - const void *values, const GLubyte mask[] - - -#define GET_XRB(XRB) \ - struct xmesa_renderbuffer *XRB = xmesa_renderbuffer(rb) - - -/* - * Write a span of PF_TRUECOLOR pixels to a pixmap. - */ -static void put_row_TRUECOLOR_pixmap( PUT_ROW_ARGS ) -{ - const GLubyte (*rgba)[4] = (const GLubyte (*)[4]) values; - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GET_XRB(xrb); - XMesaDisplay *dpy = XMESA_BUFFER(ctx->DrawBuffer)->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLuint *ptr4 = (GLuint *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLuint *ptr4 = (GLuint *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLuint *ptr4 = (GLuint *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLuint *ptr4 = (GLuint *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLuint *ptr4 = (GLuint *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - y = YFLIP(xrb, y); - if (mask) { - register GLuint i; - for (i=0;iDrawBuffer)->rowimage; - register GLuint *ptr4 = (GLuint *) rowimg->data; - register GLuint pixel; - static const GLuint shift[4] = {0, 8, 16, 24}; - register GLuint i = 0; - int w = n; - while (w > 3) { - pixel = rgba[i][BCOMP] /* << shift[0]*/; - pixel |= rgba[i][GCOMP] << shift[1]; - pixel |= rgba[i++][RCOMP] << shift[2]; - pixel |= rgba[i][BCOMP] << shift[3]; - *ptr4++ = pixel; - - pixel = rgba[i][GCOMP] /* << shift[0]*/; - pixel |= rgba[i++][RCOMP] << shift[1]; - pixel |= rgba[i][BCOMP] << shift[2]; - pixel |= rgba[i][GCOMP] << shift[3]; - *ptr4++ = pixel; - - pixel = rgba[i++][RCOMP]/* << shift[0]*/; - pixel |= rgba[i][BCOMP] << shift[1]; - pixel |= rgba[i][GCOMP] << shift[2]; - pixel |= rgba[i++][RCOMP] << shift[3]; - *ptr4++ = pixel; - - w -= 4; - } - switch (w) { - case 3: - pixel = 0; - pixel |= rgba[i][BCOMP] /*<< shift[0]*/; - pixel |= rgba[i][GCOMP] << shift[1]; - pixel |= rgba[i++][RCOMP] << shift[2]; - pixel |= rgba[i][BCOMP] << shift[3]; - *ptr4++ = pixel; - pixel = 0; - pixel |= rgba[i][GCOMP] /*<< shift[0]*/; - pixel |= rgba[i++][RCOMP] << shift[1]; - pixel |= rgba[i][BCOMP] << shift[2]; - pixel |= rgba[i][GCOMP] << shift[3]; - *ptr4++ = pixel; - pixel = 0xffffff00 & *ptr4; - pixel |= rgba[i][RCOMP] /*<< shift[0]*/; - *ptr4 = pixel; - break; - case 2: - pixel = 0; - pixel |= rgba[i][BCOMP] /*<< shift[0]*/; - pixel |= rgba[i][GCOMP] << shift[1]; - pixel |= rgba[i++][RCOMP] << shift[2]; - pixel |= rgba[i][BCOMP] << shift[3]; - *ptr4++ = pixel; - pixel = 0xffff0000 & *ptr4; - pixel |= rgba[i][GCOMP] /*<< shift[0]*/; - pixel |= rgba[i][RCOMP] << shift[1]; - *ptr4 = pixel; - break; - case 1: - pixel = 0xff000000 & *ptr4; - pixel |= rgba[i][BCOMP] /*<< shift[0]*/; - pixel |= rgba[i][GCOMP] << shift[1]; - pixel |= rgba[i][RCOMP] << shift[2]; - *ptr4 = pixel; - break; - case 0: - break; - } - XMesaPutImage( dpy, buffer, gc, rowimg, 0, 0, x, y, n, 1 ); - } -} - - -/* - * Write a span of PF_8R8G8B pixels to a pixmap (no alpha). - */ -static void put_row_rgb_8R8G8B_pixmap( RGB_SPAN_ARGS ) -{ - const GLubyte (*rgb)[3] = (const GLubyte (*)[3]) values; - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GET_XRB(xrb); - XMesaDisplay *dpy = xmesa->xm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLuint *ptr4 = (GLuint *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - y = YFLIP(xrb, y); - if (mask) { - register GLuint i; - for (i=0;iDrawBuffer)->rowimage; - register GLuint *ptr4 = (GLuint *) rowimg->data; - register GLuint pixel; - static const GLuint shift[4] = {0, 8, 16, 24}; - unsigned w = n; - register GLuint i = 0; - while (w > 3) { - pixel = 0; - pixel |= rgb[i][BCOMP]/* << shift[0]*/; - pixel |= rgb[i][GCOMP] << shift[1]; - pixel |= rgb[i++][RCOMP] << shift[2]; - pixel |= rgb[i][BCOMP] <xm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLushort *ptr2 = (GLushort *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLushort *ptr2 = (GLushort *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLushort *ptr2 = (GLushort *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - register GLushort *ptr2 = (GLushort *) rowimg->data; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - XDITHER_SETUP(y); - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - XDITHER_SETUP(y); - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - SETUP_1BIT; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - SETUP_1BIT; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - LOOKUP_SETUP; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - LOOKUP_SETUP; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iDrawBuffer)->rowimage; - for (i=0;iximage; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;i 3) { - pixel = rgba[i][BCOMP]; - pixel |= rgba[i][GCOMP] << 8; - pixel |= rgba[i++][RCOMP] << 16; - pixel |= rgba[i][BCOMP] << 24; - *ptr4++ = pixel; - pixel = rgba[i][GCOMP]; - pixel |= rgba[i++][RCOMP] << 8; - pixel |= rgba[i][BCOMP] << 16; - pixel |= rgba[i][GCOMP] << 24; - *ptr4++ = pixel; - pixel = rgba[i++][RCOMP]; - pixel |= rgba[i][BCOMP] << 8; - pixel |= rgba[i][GCOMP] << 16; - pixel |= rgba[i++][RCOMP] << 24; - *ptr4++ = pixel; - w -= 4; - } - switch (w) { - case 0: - break; - case 1: - pixel = *ptr4 & 0xff000000; - pixel |= rgba[i][BCOMP]; - pixel |= rgba[i][GCOMP] << 8; - pixel |= rgba[i][RCOMP] << 16; - *ptr4 = pixel; - break; - case 2: - pixel = rgba[i][BCOMP]; - pixel |= rgba[i][GCOMP] << 8; - pixel |= rgba[i++][RCOMP] << 16; - pixel |= rgba[i][BCOMP] << 24; - *ptr4++ = pixel; - pixel = *ptr4 & 0xffff0000; - pixel |= rgba[i][GCOMP]; - pixel |= rgba[i][RCOMP] << 8; - *ptr4 = pixel; - break; - case 3: - pixel = rgba[i][BCOMP]; - pixel |= rgba[i][GCOMP] << 8; - pixel |= rgba[i++][RCOMP] << 16; - pixel |= rgba[i][BCOMP] << 24; - *ptr4++ = pixel; - pixel = rgba[i][GCOMP]; - pixel |= rgba[i++][RCOMP] << 8; - pixel |= rgba[i][BCOMP] << 16; - pixel |= rgba[i][GCOMP] << 24; - *ptr4++ = pixel; - pixel = *ptr4 & 0xffffff00; - pixel |= rgba[i][RCOMP]; - *ptr4 = pixel; - break; - } - } -} - - -/* - * Write a span of PF_8R8G8B-format pixels to an ximage (no alpha). - */ -static void put_row_rgb_8R8G8B_ximage( RGB_SPAN_ARGS ) -{ - const GLubyte (*rgb)[3] = (const GLubyte (*)[3]) values; - GET_XRB(xrb); - register GLuint i; - register GLuint *ptr = PIXEL_ADDR4(xrb, x, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - int yy = YFLIP(xrb, y); - XDITHER_SETUP(yy); - if (mask) { - for (i=0;iximage; - register GLuint i; - int yy = YFLIP(xrb, y); - XDITHER_SETUP(yy); - if (mask) { - for (i=0;iximage; - register GLuint i; - SETUP_1BIT; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - SETUP_1BIT; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - LOOKUP_SETUP; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - LOOKUP_SETUP; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - DITHER_SETUP; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - SETUP_1BIT; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - LOOKUP_SETUP; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;iximage; - register GLuint i; - for (i=0;iximage; - register GLuint i; - for (i=0;ir = rgba[i][RCOMP]; - ptr->g = rgba[i][GCOMP]; - ptr->b = rgba[i][BCOMP]; - } - } -} - - -/* - * Write an array of PF_5R6G5B pixels to an ximage. - */ -static void put_values_5R6G5B_ximage( PUT_VALUES_ARGS ) -{ - const GLubyte (*rgba)[4] = (const GLubyte (*)[4]) values; - GET_XRB(xrb); - register GLuint i; - for (i=0;iximage; - register GLuint i; - DITHER_SETUP; - for (i=0;iximage; - register GLuint i; - SETUP_1BIT; - for (i=0;iximage; - register GLuint i; - LOOKUP_SETUP; - for (i=0;iximage; - register GLuint i; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - const unsigned long pixel = xmesa_color_to_pixel(ctx, color[RCOMP], - color[GCOMP], color[BCOMP], color[ACOMP], xmesa->pixelformat); - register GLuint i; - XMesaSetForeground( xmesa->display, gc, pixel ); - y = YFLIP(xrb, y); - - /* New code contributed by Jeff Epler and cleaned up by Keith - * Whitwell. - */ - for (i = 0; i < n; ) { - GLuint start = i; - - /* Identify and emit contiguous rendered pixels - */ - while (i < n && (!mask || mask[i])) - i++; - - if (start < i) - XMesaFillRectangle( dpy, buffer, gc, - (int)(x+start), (int) y, - (int)(i-start), 1); - - /* Eat up non-rendered pixels - */ - while (i < n && !mask[i]) - i++; - } -} - - - -static void -put_mono_row_ci_pixmap( PUT_MONO_ROW_ARGS ) -{ - GLuint colorIndex = *((GLuint *) value); - XMesaContext xmesa = XMESA_CONTEXT(ctx); - GET_XRB(xrb); - XMesaDisplay *dpy = xmesa->xm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - XMesaSetForeground( xmesa->display, gc, colorIndex ); - y = YFLIP(xrb, y); - - for (i = 0 ; i < n ;) { - GLuint start = i; - - /* Identify and emit contiguous rendered pixels - */ - while (i < n && (!mask || mask[i])) - i++; - - if (start < i) - XMesaFillRectangle( dpy, buffer, gc, - (int)(x+start), (int) y, - (int)(i-start), 1); - - /* Eat up non-rendered pixels - */ - while (i < n && !mask[i]) - i++; - } -} - - - -/* - * Write a span of PF_TRUEDITHER pixels to a pixmap. - */ -static void put_mono_row_TRUEDITHER_pixmap( PUT_MONO_ROW_ARGS ) -{ - const GLubyte *color = (const GLubyte *) value; - GET_XRB(xrb); - XMesaContext xmesa = XMESA_CONTEXT(ctx); - XMesaDisplay *dpy = xmesa->xm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; - register GLuint i; - int yy = YFLIP(xrb, y); - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; - register GLuint i; - int yy = YFLIP(xrb, y); - XDITHER_SETUP(yy); - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; - register GLuint i; - SETUP_1BIT; - y = YFLIP(xrb, y); - for (i=0;iximage; - register GLuint i; - const unsigned long pixel = xmesa_color_to_pixel(ctx, color[RCOMP], - color[GCOMP], color[BCOMP], color[ACOMP], xmesa->pixelformat); - y = YFLIP(xrb, y); - for (i=0;iximage; - register GLuint i; - y = YFLIP(xrb, y); - for (i=0;iximage; - const GLint r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; - GLuint i; - y = YFLIP(xrb, y); - for (i=0;ipixelformat); - ptr = PIXEL_ADDR4(xrb, x, y ); - for (i=0;ipixelformat); - ptr = PIXEL_ADDR4(xrb, x, y ); - for (i=0;iximage; - int yy = YFLIP(xrb, y); - register GLuint i; - XDITHER_SETUP(yy); - for (i=0;iximage; - register GLuint i; - SETUP_1BIT; - y = YFLIP(xrb, y); - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - const unsigned long pixel = xmesa_color_to_pixel(ctx, color[RCOMP], - color[GCOMP], color[BCOMP], color[ACOMP], xmesa->pixelformat); - XMesaSetForeground( xmesa->display, gc, pixel ); - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - XMesaSetForeground( xmesa->display, gc, colorIndex ); - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; - DITHER_SETUP; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; - SETUP_1BIT; - for (i=0;iximage; - register GLuint i; - const unsigned long pixel = xmesa_color_to_pixel(ctx, color[RCOMP], - color[GCOMP], color[BCOMP], color[ACOMP], xmesa->pixelformat); - for (i=0;iximage; - register GLuint i; - for (i=0;iximage; - register GLuint i; - const int r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; - for (i=0;ir = r; - ptr->g = g; - ptr->b = b; - } - } -} - - -/* - * Write an array of identical PF_DITHER pixels to an XImage. - */ -static void put_mono_values_DITHER_ximage( PUT_MONO_VALUES_ARGS ) -{ - const GLubyte *color = (const GLubyte *) value; - GET_XRB(xrb); - const GLubyte r = color[RCOMP], g = color[GCOMP], b = color[BCOMP]; - XMesaImage *img = xrb->ximage; - register GLuint i; - DITHER_SETUP; - for (i=0;iximage; - register GLuint i; - SETUP_1BIT; - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;iximage; - register GLuint i; - y = YFLIP(xrb, y); - if (mask) { - for (i=0;ixm_visual->display; - XMesaDrawable buffer = xrb->drawable; - XMesaGC gc = XMESA_BUFFER(ctx->DrawBuffer)->gc; - register GLuint i; - for (i=0;iximage; - register GLuint i; - for (i=0;iDrawBuffer); - Window rootWin = RootWindow(xmesa->display, 0); - Window child; - GLint screenWidth = WidthOfScreen(DefaultScreenOfDisplay(xmesa->display)); - GLint dx, dy; - if (source->type == PBUFFER || source->type == PIXMAP) - return 0; - XTranslateCoordinates(xmesa->display, source->frontxrb->pixmap, rootWin, - *x, *y, &dx, &dy, &child); - if (dx >= screenWidth) { - /* totally clipped on right */ - return -1; - } - if (dx < 0) { - /* clipped on left */ - GLint clip = -dx; - if (clip >= (GLint) *n) - return -1; /* totally clipped on left */ - *x += clip; - *n -= clip; - dx = 0; - return clip; - } - if ((GLint) (dx + *n) > screenWidth) { - /* clipped on right */ - GLint clip = dx + *n - screenWidth; - *n -= clip; - } - return 0; -} -#endif - - -/* - * Read a horizontal span of color-index pixels. - */ -static void -get_row_ci(GLcontext *ctx, struct gl_renderbuffer *rb, - GLuint n, GLint x, GLint y, void *values) -{ - GLuint *index = (GLuint *) values; - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - GET_XRB(xrb); - GLuint i; - - y = YFLIP(xrb, y); - - if (xrb->pixmap) { -#ifndef XFree86Server - XMesaImage *span = NULL; - int error; - int k = clip_for_xgetimage(ctx, &n, &x, &y); - if (k < 0) - return; - index += k; - - catch_xgetimage_errors( xmesa->display ); - span = XGetImage( xmesa->display, xrb->pixmap, - x, y, n, 1, AllPlanes, ZPixmap ); - error = check_xgetimage_errors(); - if (span && !error) { - for (i=0;idisplay->GetImage)(xrb->drawable, - x, y, n, 1, ZPixmap, - ~0L, (pointer)index); -#endif - } - else if (xrb->ximage) { - XMesaImage *img = xrb->ximage; - for (i=0;iDrawBuffer); - - if (xrb->pixmap) { - /* Read from Pixmap or Window */ - XMesaImage *span = NULL; - int error; -#ifdef XFree86Server - span = XMesaCreateImage(xmesa->xm_visual->BitsPerPixel, n, 1, NULL); - span->data = (char *)MALLOC(span->height * span->bytes_per_line); - error = (!span->data); - (*xmesa->display->GetImage)(xrb->drawable, - x, YFLIP(xrb, y), n, 1, ZPixmap, - ~0L, (pointer)span->data); -#else - int k; - y = YFLIP(xrb, y); - k = clip_for_xgetimage(ctx, &n, &x, &y); - if (k < 0) - return; - rgba += k; - catch_xgetimage_errors( xmesa->display ); - span = XGetImage( xmesa->display, xrb->pixmap, - x, y, n, 1, AllPlanes, ZPixmap ); - error = check_xgetimage_errors(); -#endif - if (span && !error) { - switch (xmesa->pixelformat) { - case PF_Truecolor: - case PF_Dither_True: - { - const GLubyte *pixelToR = xmesa->xm_visual->PixelToR; - const GLubyte *pixelToG = xmesa->xm_visual->PixelToG; - const GLubyte *pixelToB = xmesa->xm_visual->PixelToB; - unsigned long rMask = GET_REDMASK(xmesa->xm_visual); - unsigned long gMask = GET_GREENMASK(xmesa->xm_visual); - unsigned long bMask = GET_BLUEMASK(xmesa->xm_visual); - GLint rShift = xmesa->xm_visual->rshift; - GLint gShift = xmesa->xm_visual->gshift; - GLint bShift = xmesa->xm_visual->bshift; - GLuint i; - for (i=0;i> rShift]; - rgba[i][GCOMP] = pixelToG[(p & gMask) >> gShift]; - rgba[i][BCOMP] = pixelToB[(p & bMask) >> bShift]; - rgba[i][ACOMP] = 255; - } - } - break; - case PF_5R6G5B: - case PF_Dither_5R6G5B: - { - const GLubyte *pixelToR = xmesa->xm_visual->PixelToR; - const GLubyte *pixelToG = xmesa->xm_visual->PixelToG; - const GLubyte *pixelToB = xmesa->xm_visual->PixelToB; - GLuint i; - for (i=0;i> 8) & 0xf8); - rgba[i][GCOMP] = ((p >> 3) & 0xfc); - rgba[i][BCOMP] = ((p << 3) & 0xff); - */ - rgba[i][RCOMP] = pixelToR[p >> 11]; - rgba[i][GCOMP] = pixelToG[(p >> 5) & 0x3f]; - rgba[i][BCOMP] = pixelToB[p & 0x1f]; - rgba[i][ACOMP] = 255; - } - } - break; - case PF_8A8B8G8R: - { - const GLuint *ptr4 = (GLuint *) span->data; - GLuint i; - for (i=0;i> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ((p4 >> 16) & 0xff); - rgba[i][ACOMP] = (GLubyte) ((p4 >> 24) & 0xff); - } - } - break; - case PF_8A8R8G8B: - { - const GLuint *ptr4 = (GLuint *) span->data; - GLuint i; - for (i=0;i> 16) & 0xff); - rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); - rgba[i][ACOMP] = (GLubyte) ((p4 >> 24) & 0xff); - } - } - break; - case PF_8R8G8B: - { - const GLuint *ptr4 = (GLuint *) span->data; - GLuint i; - for (i=0;i> 16) & 0xff); - rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); - rgba[i][ACOMP] = 255; - } - } - break; - case PF_8R8G8B24: - { - const bgr_t *ptr3 = (bgr_t *) span->data; - GLuint i; - for (i=0;ipixel_to_r; - GLubyte *gTable = source->pixel_to_g; - GLubyte *bTable = source->pixel_to_b; - if (GET_VISUAL_DEPTH(xmesa->xm_visual)==8) { - const GLubyte *ptr1 = (GLubyte *) span->data; - GLuint i; - for (i=0;ixm_visual->bitFlip; - GLuint i; - for (i=0;iximage) { - /* Read from XImage back buffer */ - switch (xmesa->pixelformat) { - case PF_Truecolor: - case PF_Dither_True: - { - const GLubyte *pixelToR = xmesa->xm_visual->PixelToR; - const GLubyte *pixelToG = xmesa->xm_visual->PixelToG; - const GLubyte *pixelToB = xmesa->xm_visual->PixelToB; - unsigned long rMask = GET_REDMASK(xmesa->xm_visual); - unsigned long gMask = GET_GREENMASK(xmesa->xm_visual); - unsigned long bMask = GET_BLUEMASK(xmesa->xm_visual); - GLint rShift = xmesa->xm_visual->rshift; - GLint gShift = xmesa->xm_visual->gshift; - GLint bShift = xmesa->xm_visual->bshift; - XMesaImage *img = xrb->ximage; - GLuint i; - y = YFLIP(xrb, y); - for (i=0;i> rShift]; - rgba[i][GCOMP] = pixelToG[(p & gMask) >> gShift]; - rgba[i][BCOMP] = pixelToB[(p & bMask) >> bShift]; - rgba[i][ACOMP] = 255; - } - } - break; - case PF_5R6G5B: - case PF_Dither_5R6G5B: - { - const GLubyte *pixelToR = xmesa->xm_visual->PixelToR; - const GLubyte *pixelToG = xmesa->xm_visual->PixelToG; - const GLubyte *pixelToB = xmesa->xm_visual->PixelToB; - const GLushort *ptr2 = PIXEL_ADDR2(xrb, x, y); - GLuint i; -#if defined(__i386__) /* word stores don't have to be on 4-byte boundaries */ - const GLuint *ptr4 = (const GLuint *) ptr2; - GLuint extraPixel = (n & 1); - n -= extraPixel; - for (i = 0; i < n; i += 2) { - const GLuint p = *ptr4++; - const GLuint p0 = p & 0xffff; - const GLuint p1 = p >> 16; - /* fast, but not quite accurate - rgba[i][RCOMP] = ((p >> 8) & 0xf8); - rgba[i][GCOMP] = ((p >> 3) & 0xfc); - rgba[i][BCOMP] = ((p << 3) & 0xff); - */ - rgba[i][RCOMP] = pixelToR[p0 >> 11]; - rgba[i][GCOMP] = pixelToG[(p0 >> 5) & 0x3f]; - rgba[i][BCOMP] = pixelToB[p0 & 0x1f]; - rgba[i][ACOMP] = 255; - rgba[i+1][RCOMP] = pixelToR[p1 >> 11]; - rgba[i+1][GCOMP] = pixelToG[(p1 >> 5) & 0x3f]; - rgba[i+1][BCOMP] = pixelToB[p1 & 0x1f]; - rgba[i+1][ACOMP] = 255; - } - if (extraPixel) { - GLushort p = ptr2[n]; - rgba[n][RCOMP] = pixelToR[p >> 11]; - rgba[n][GCOMP] = pixelToG[(p >> 5) & 0x3f]; - rgba[n][BCOMP] = pixelToB[p & 0x1f]; - rgba[n][ACOMP] = 255; - } -#else - for (i = 0; i < n; i++) { - const GLushort p = ptr2[i]; - rgba[i][RCOMP] = pixelToR[p >> 11]; - rgba[i][GCOMP] = pixelToG[(p >> 5) & 0x3f]; - rgba[i][BCOMP] = pixelToB[p & 0x1f]; - rgba[i][ACOMP] = 255; - } -#endif - } - break; - case PF_8A8B8G8R: - { - const GLuint *ptr4 = PIXEL_ADDR4(xrb, x, y); - GLuint i; - for (i=0;i> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ((p4 >> 16) & 0xff); - rgba[i][ACOMP] = (GLint) ((p4 >> 24) & 0xff); - } - } - break; - case PF_8A8R8G8B: - { - const GLuint *ptr4 = PIXEL_ADDR4(xrb, x, y); - GLuint i; - for (i=0;i> 16) & 0xff); - rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); - rgba[i][ACOMP] = (GLint) ((p4 >> 24) & 0xff); - } - } - break; - case PF_8R8G8B: - { - const GLuint *ptr4 = PIXEL_ADDR4(xrb, x, y); - GLuint i; - for (i=0;i> 16) & 0xff); - rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); - rgba[i][ACOMP] = 255; - } - } - break; - case PF_8R8G8B24: - { - const bgr_t *ptr3 = PIXEL_ADDR3(xrb, x, y); - GLuint i; - for (i=0;ipixel_to_r; - const GLubyte *gTable = source->pixel_to_g; - const GLubyte *bTable = source->pixel_to_b; - if (GET_VISUAL_DEPTH(xmesa->xm_visual)==8) { - GLubyte *ptr1 = PIXEL_ADDR1(xrb, x, y); - GLuint i; - for (i=0;iximage; - GLuint i; - y = YFLIP(xrb, y); - for (i=0;iximage; - int bitFlip = xmesa->xm_visual->bitFlip; - GLuint i; - y = YFLIP(xrb, y); - for (i=0;ipixmap) { - for (i=0;idisplay, xrb->drawable, - x[i], YFLIP(xrb, y[i]) ); - } - } - else if (xrb->ximage) { - XMesaImage *img = xrb->ximage; - for (i=0;ixm_visual->display; - XMesaBuffer source = XMESA_BUFFER(ctx->DrawBuffer); - register GLuint i; - - if (xrb->pixmap) { - XMesaDrawable buffer = xrb->drawable; - switch (xmesa->pixelformat) { - case PF_Truecolor: - case PF_Dither_True: - case PF_5R6G5B: - case PF_Dither_5R6G5B: - { - unsigned long rMask = GET_REDMASK(xmesa->xm_visual); - unsigned long gMask = GET_GREENMASK(xmesa->xm_visual); - unsigned long bMask = GET_BLUEMASK(xmesa->xm_visual); - GLubyte *pixelToR = xmesa->xm_visual->PixelToR; - GLubyte *pixelToG = xmesa->xm_visual->PixelToG; - GLubyte *pixelToB = xmesa->xm_visual->PixelToB; - GLint rShift = xmesa->xm_visual->rshift; - GLint gShift = xmesa->xm_visual->gshift; - GLint bShift = xmesa->xm_visual->bshift; - for (i=0;i> rShift]; - rgba[i][GCOMP] = pixelToG[(p & gMask) >> gShift]; - rgba[i][BCOMP] = pixelToB[(p & bMask) >> bShift]; - rgba[i][ACOMP] = 255; - } - } - break; - case PF_8A8B8G8R: - for (i=0;i> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ((p >> 16) & 0xff); - rgba[i][ACOMP] = (GLubyte) ((p >> 24) & 0xff); - } - break; - case PF_8A8R8G8B: - for (i=0;i> 16) & 0xff); - rgba[i][GCOMP] = (GLubyte) ((p >> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ( p & 0xff); - rgba[i][ACOMP] = (GLubyte) ((p >> 24) & 0xff); - } - break; - case PF_8R8G8B: - for (i=0;i> 16) & 0xff); - rgba[i][GCOMP] = (GLubyte) ((p >> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ( p & 0xff); - rgba[i][ACOMP] = 255; - } - break; - case PF_8R8G8B24: - for (i=0;i> 16) & 0xff); - rgba[i][GCOMP] = (GLubyte) ((p >> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ( p & 0xff); - rgba[i][ACOMP] = 255; - } - break; - case PF_Dither: - case PF_Lookup: - case PF_Grayscale: - { - GLubyte *rTable = source->pixel_to_r; - GLubyte *gTable = source->pixel_to_g; - GLubyte *bTable = source->pixel_to_b; - for (i=0;ixm_visual->bitFlip; - for (i=0;iximage) { - /* Read from XImage back buffer */ - switch (xmesa->pixelformat) { - case PF_Truecolor: - case PF_Dither_True: - case PF_5R6G5B: - case PF_Dither_5R6G5B: - { - unsigned long rMask = GET_REDMASK(xmesa->xm_visual); - unsigned long gMask = GET_GREENMASK(xmesa->xm_visual); - unsigned long bMask = GET_BLUEMASK(xmesa->xm_visual); - GLubyte *pixelToR = xmesa->xm_visual->PixelToR; - GLubyte *pixelToG = xmesa->xm_visual->PixelToG; - GLubyte *pixelToB = xmesa->xm_visual->PixelToB; - GLint rShift = xmesa->xm_visual->rshift; - GLint gShift = xmesa->xm_visual->gshift; - GLint bShift = xmesa->xm_visual->bshift; - XMesaImage *img = xrb->ximage; - for (i=0;i> rShift]; - rgba[i][GCOMP] = pixelToG[(p & gMask) >> gShift]; - rgba[i][BCOMP] = pixelToB[(p & bMask) >> bShift]; - rgba[i][ACOMP] = 255; - } - } - break; - case PF_8A8B8G8R: - for (i=0;i> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ((p4 >> 16) & 0xff); - rgba[i][ACOMP] = (GLubyte) ((p4 >> 24) & 0xff); - } - break; - case PF_8A8R8G8B: - for (i=0;i> 16) & 0xff); - rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); - rgba[i][ACOMP] = (GLubyte) ((p4 >> 24) & 0xff); - } - break; - case PF_8R8G8B: - for (i=0;i> 16) & 0xff); - rgba[i][GCOMP] = (GLubyte) ((p4 >> 8) & 0xff); - rgba[i][BCOMP] = (GLubyte) ( p4 & 0xff); - rgba[i][ACOMP] = 255; - } - break; - case PF_8R8G8B24: - for (i=0;ir; - rgba[i][GCOMP] = ptr3->g; - rgba[i][BCOMP] = ptr3->b; - rgba[i][ACOMP] = 255; - } - break; - case PF_Dither: - case PF_Lookup: - case PF_Grayscale: - { - GLubyte *rTable = source->pixel_to_r; - GLubyte *gTable = source->pixel_to_g; - GLubyte *bTable = source->pixel_to_b; - XMesaImage *img = xrb->ximage; - for (i=0;iximage; - int bitFlip = xmesa->xm_visual->bitFlip; - for (i=0;ipixmap ? GL_TRUE : GL_FALSE; - struct gl_renderbuffer *rb = &xrb->St.Base; - - switch (pixelformat) { - case PF_Index: - ASSERT(rb->DataType == GL_UNSIGNED_INT); - if (pixmap) { - rb->PutRow = put_row_ci_pixmap; - rb->PutRowRGB = NULL; - rb->PutMonoRow = put_mono_row_ci_pixmap; - rb->PutValues = put_values_ci_pixmap; - rb->PutMonoValues = put_mono_values_ci_pixmap; - } - else { - rb->PutRow = put_row_ci_ximage; - rb->PutRowRGB = NULL; - rb->PutMonoRow = put_mono_row_ci_ximage; - rb->PutValues = put_values_ci_ximage; - rb->PutMonoValues = put_mono_values_ci_ximage; - } - break; - case PF_Truecolor: - if (pixmap) { - rb->PutRow = put_row_TRUECOLOR_pixmap; - rb->PutRowRGB = put_row_rgb_TRUECOLOR_pixmap; - rb->PutMonoRow = put_mono_row_pixmap; - rb->PutValues = put_values_TRUECOLOR_pixmap; - rb->PutMonoValues = put_mono_values_pixmap; - } - else { - rb->PutRow = put_row_TRUECOLOR_ximage; - rb->PutRowRGB = put_row_rgb_TRUECOLOR_ximage; - rb->PutMonoRow = put_mono_row_ximage; - rb->PutValues = put_values_TRUECOLOR_ximage; - rb->PutMonoValues = put_mono_values_ximage; - } - break; - case PF_Dither_True: - if (pixmap) { - rb->PutRow = put_row_TRUEDITHER_pixmap; - rb->PutRowRGB = put_row_rgb_TRUEDITHER_pixmap; - rb->PutMonoRow = put_mono_row_TRUEDITHER_pixmap; - rb->PutValues = put_values_TRUEDITHER_pixmap; - rb->PutMonoValues = put_mono_values_TRUEDITHER_pixmap; - } - else { - rb->PutRow = put_row_TRUEDITHER_ximage; - rb->PutRowRGB = put_row_rgb_TRUEDITHER_ximage; - rb->PutMonoRow = put_mono_row_TRUEDITHER_ximage; - rb->PutValues = put_values_TRUEDITHER_ximage; - rb->PutMonoValues = put_mono_values_TRUEDITHER_ximage; - } - break; - case PF_8A8B8G8R: - if (pixmap) { - rb->PutRow = put_row_8A8B8G8R_pixmap; - rb->PutRowRGB = put_row_rgb_8A8B8G8R_pixmap; - rb->PutMonoRow = put_mono_row_pixmap; - rb->PutValues = put_values_8A8B8G8R_pixmap; - rb->PutMonoValues = put_mono_values_pixmap; - } - else { - rb->PutRow = put_row_8A8B8G8R_ximage; - rb->PutRowRGB = put_row_rgb_8A8B8G8R_ximage; - rb->PutMonoRow = put_mono_row_8A8B8G8R_ximage; - rb->PutValues = put_values_8A8B8G8R_ximage; - rb->PutMonoValues = put_mono_values_8A8B8G8R_ximage; - rb->GetPointer = get_pointer_4_ximage; - } - break; - case PF_8A8R8G8B: - if (pixmap) { - rb->PutRow = put_row_8A8R8G8B_pixmap; - rb->PutRowRGB = put_row_rgb_8A8R8G8B_pixmap; - rb->PutMonoRow = put_mono_row_pixmap; - rb->PutValues = put_values_8A8R8G8B_pixmap; - rb->PutMonoValues = put_mono_values_pixmap; - } - else { - rb->PutRow = put_row_8A8R8G8B_ximage; - rb->PutRowRGB = put_row_rgb_8A8R8G8B_ximage; - rb->PutMonoRow = put_mono_row_8A8R8G8B_ximage; - rb->PutValues = put_values_8A8R8G8B_ximage; - rb->PutMonoValues = put_mono_values_8A8R8G8B_ximage; - rb->GetPointer = get_pointer_4_ximage; - } - break; - case PF_8R8G8B: - if (pixmap) { - rb->PutRow = put_row_8R8G8B_pixmap; - rb->PutRowRGB = put_row_rgb_8R8G8B_pixmap; - rb->PutMonoRow = put_mono_row_pixmap; - rb->PutValues = put_values_8R8G8B_pixmap; - rb->PutMonoValues = put_mono_values_pixmap; - } - else { - rb->PutRow = put_row_8R8G8B_ximage; - rb->PutRowRGB = put_row_rgb_8R8G8B_ximage; - rb->PutMonoRow = put_mono_row_8R8G8B_ximage; - rb->PutValues = put_values_8R8G8B_ximage; - rb->PutMonoValues = put_mono_values_8R8G8B_ximage; - } - break; - case PF_8R8G8B24: - if (pixmap) { - rb->PutRow = put_row_8R8G8B24_pixmap; - rb->PutRowRGB = put_row_rgb_8R8G8B24_pixmap; - rb->PutMonoRow = put_mono_row_pixmap; - rb->PutValues = put_values_8R8G8B24_pixmap; - rb->PutMonoValues = put_mono_values_pixmap; - } - else { - rb->PutRow = put_row_8R8G8B24_ximage; - rb->PutRowRGB = put_row_rgb_8R8G8B24_ximage; - rb->PutMonoRow = put_mono_row_8R8G8B24_ximage; - rb->PutValues = put_values_8R8G8B24_ximage; - rb->PutMonoValues = put_mono_values_8R8G8B24_ximage; - } - break; - case PF_5R6G5B: - if (pixmap) { - rb->PutRow = put_row_5R6G5B_pixmap; - rb->PutRowRGB = put_row_rgb_5R6G5B_pixmap; - rb->PutMonoRow = put_mono_row_pixmap; - rb->PutValues = put_values_5R6G5B_pixmap; - rb->PutMonoValues = put_mono_values_pixmap; - } - else { - rb->PutRow = put_row_5R6G5B_ximage; - rb->PutRowRGB = put_row_rgb_5R6G5B_ximage; - rb->PutMonoRow = put_mono_row_ximage; - rb->PutValues = put_values_5R6G5B_ximage; - rb->PutMonoValues = put_mono_values_ximage; - } - break; - case PF_Dither_5R6G5B: - if (pixmap) { - rb->PutRow = put_row_DITHER_5R6G5B_pixmap; - rb->PutRowRGB = put_row_rgb_DITHER_5R6G5B_pixmap; - rb->PutMonoRow = put_mono_row_TRUEDITHER_pixmap; - rb->PutValues = put_values_DITHER_5R6G5B_pixmap; - rb->PutMonoValues = put_mono_values_TRUEDITHER_pixmap; - } - else { - rb->PutRow = put_row_DITHER_5R6G5B_ximage; - rb->PutRowRGB = put_row_rgb_DITHER_5R6G5B_ximage; - rb->PutMonoRow = put_mono_row_DITHER_5R6G5B_ximage; - rb->PutValues = put_values_DITHER_5R6G5B_ximage; - rb->PutMonoValues = put_mono_values_DITHER_5R6G5B_ximage; - } - break; - case PF_Dither: - if (pixmap) { - rb->PutRow = put_row_DITHER_pixmap; - rb->PutRowRGB = put_row_rgb_DITHER_pixmap; - rb->PutMonoRow = put_mono_row_DITHER_pixmap; - rb->PutValues = put_values_DITHER_pixmap; - rb->PutMonoValues = put_mono_values_DITHER_pixmap; - } - else { - if (depth == 8) { - rb->PutRow = put_row_DITHER8_ximage; - rb->PutRowRGB = put_row_rgb_DITHER8_ximage; - rb->PutMonoRow = put_mono_row_DITHER8_ximage; - rb->PutValues = put_values_DITHER8_ximage; - rb->PutMonoValues = put_mono_values_DITHER8_ximage; - } - else { - rb->PutRow = put_row_DITHER_ximage; - rb->PutRowRGB = put_row_rgb_DITHER_ximage; - rb->PutMonoRow = put_mono_row_DITHER_ximage; - rb->PutValues = put_values_DITHER_ximage; - rb->PutMonoValues = put_mono_values_DITHER_ximage; - } - } - break; - case PF_1Bit: - if (pixmap) { - rb->PutRow = put_row_1BIT_pixmap; - rb->PutRowRGB = put_row_rgb_1BIT_pixmap; - rb->PutMonoRow = put_mono_row_1BIT_pixmap; - rb->PutValues = put_values_1BIT_pixmap; - rb->PutMonoValues = put_mono_values_1BIT_pixmap; - } - else { - rb->PutRow = put_row_1BIT_ximage; - rb->PutRowRGB = put_row_rgb_1BIT_ximage; - rb->PutMonoRow = put_mono_row_1BIT_ximage; - rb->PutValues = put_values_1BIT_ximage; - rb->PutMonoValues = put_mono_values_1BIT_ximage; - } - break; - case PF_Lookup: - if (pixmap) { - rb->PutRow = put_row_LOOKUP_pixmap; - rb->PutRowRGB = put_row_rgb_LOOKUP_pixmap; - rb->PutMonoRow = put_mono_row_pixmap; - rb->PutValues = put_values_LOOKUP_pixmap; - rb->PutMonoValues = put_mono_values_pixmap; - } - else { - if (depth==8) { - rb->PutRow = put_row_LOOKUP8_ximage; - rb->PutRowRGB = put_row_rgb_LOOKUP8_ximage; - rb->PutMonoRow = put_mono_row_LOOKUP8_ximage; - rb->PutValues = put_values_LOOKUP8_ximage; - rb->PutMonoValues = put_mono_values_LOOKUP8_ximage; - } - else { - rb->PutRow = put_row_LOOKUP_ximage; - rb->PutRowRGB = put_row_rgb_LOOKUP_ximage; - rb->PutMonoRow = put_mono_row_ximage; - rb->PutValues = put_values_LOOKUP_ximage; - rb->PutMonoValues = put_mono_values_ximage; - } - } - break; - case PF_Grayscale: - if (pixmap) { - rb->PutRow = put_row_GRAYSCALE_pixmap; - rb->PutRowRGB = put_row_rgb_GRAYSCALE_pixmap; - rb->PutMonoRow = put_mono_row_pixmap; - rb->PutValues = put_values_GRAYSCALE_pixmap; - rb->PutMonoValues = put_mono_values_pixmap; - } - else { - if (depth == 8) { - rb->PutRow = put_row_GRAYSCALE8_ximage; - rb->PutRowRGB = put_row_rgb_GRAYSCALE8_ximage; - rb->PutMonoRow = put_mono_row_GRAYSCALE8_ximage; - rb->PutValues = put_values_GRAYSCALE8_ximage; - rb->PutMonoValues = put_mono_values_GRAYSCALE8_ximage; - } - else { - rb->PutRow = put_row_GRAYSCALE_ximage; - rb->PutRowRGB = put_row_rgb_GRAYSCALE_ximage; - rb->PutMonoRow = put_mono_row_ximage; - rb->PutValues = put_values_GRAYSCALE_ximage; - rb->PutMonoValues = put_mono_values_ximage; - } - } - break; - default: - _mesa_problem(NULL, "Bad pixel format in xmesa_set_renderbuffer_funcs"); - return; - } - - - /* Get functions */ - if (pixelformat == PF_Index) { - rb->GetRow = get_row_ci; - rb->GetValues = get_values_ci; - } - else { - rb->GetRow = get_row_rgba; - rb->GetValues = get_values_rgba; - } -} - -- cgit v1.2.3 From 54be3a7added5b6675a78f6df711d3bebc4c6cd5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 1 Nov 2007 22:19:42 +0000 Subject: Reserve the right number of dwords for hardware state, and handle the case where it fails. --- src/mesa/pipe/i915simple/i915_state_emit.c | 40 +++++++++++++++++++++++++----- 1 file changed, 34 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 4804b0caee..15eff97813 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -60,8 +60,31 @@ static unsigned translate_depth_format( unsigned zformat ) void i915_emit_hardware_state(struct i915_context *i915 ) { - BEGIN_BATCH(100, 10); + /* XXX: there must be an easier way */ + const unsigned dwords = ( 14 + + 5 + + I915_MAX_DYNAMIC + + 8 + + 2 + I915_TEX_UNITS*3 + + 2 + I915_TEX_UNITS*3 + + 2 + I915_MAX_CONSTANT*4 + + i915->current.program_len + + 6 + ) * 3/2; /* plus 50% margin */ + const unsigned relocs = ( I915_TEX_UNITS + + 2 + ) * 3/2; /* plus 50% margin */ + +#if 0 + fprintf (stderr, "i915_emit_hardware_state: %d dwords, %d relocs\n", dwords, relocs); +#endif + + if(!BEGIN_BATCH(dwords, relocs)) { + FLUSH_BATCH(); + BEGIN_BATCH(dwords, relocs); + } + /* 14 dwords, 0 relocs */ if (i915->hardware_dirty & I915_HW_INVARIENT) { OUT_BATCH(_3DSTATE_AA_CMD | @@ -111,7 +134,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(0); } - + /* 5 dwords, 0 relocs */ if (i915->hardware_dirty & I915_HW_IMMEDIATE) { OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | @@ -125,9 +148,9 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S4]); OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S5]); OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S6]); - } + } - + /* I915_MAX_DYNAMIC dwords, 0 relocs */ if (i915->hardware_dirty & I915_HW_DYNAMIC) { int i; @@ -135,7 +158,8 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(i915->current.dynamic[i]); } } - + + /* 8 dwords, 2 relocs */ if (i915->hardware_dirty & I915_HW_STATIC) { if (i915->framebuffer.cbufs[0]) { @@ -189,9 +213,9 @@ i915_emit_hardware_state(struct i915_context *i915 ) } } - #if 01 /* texture images */ + /* 2 + I915_TEX_UNITS*3 dwords, I915_TEX_UNITS relocs */ if (i915->hardware_dirty & (I915_HW_MAP | I915_HW_SAMPLER)) { /* XXX: we were refering to sampler state @@ -230,6 +254,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) #if 01 /* samplers */ + /* 2 + I915_TEX_UNITS*3 dwords, 0 relocs */ if (i915->hardware_dirty & I915_HW_SAMPLER) { if (i915->current.sampler_enable_nr) { @@ -252,6 +277,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) #endif /* constants */ + /* 2 + I915_MAX_CONSTANT*4 dwords, 0 relocs */ if (i915->hardware_dirty & I915_HW_PROGRAM) { const uint nr = i915->current.num_constants[PIPE_SHADER_FRAGMENT]; @@ -272,6 +298,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) } /* Fragment program */ + /* i915->current.program_len dwords, 0 relocs */ if (i915->hardware_dirty & I915_HW_PROGRAM) { uint i; @@ -283,6 +310,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) } /* drawing surface size */ + /* 6 dwords, 0 relocs */ { int w = i915->framebuffer.cbufs[0]->width; int h = i915->framebuffer.cbufs[0]->height; -- cgit v1.2.3 From b1ed405cfcafb59b7b65af5937bdef5768cf2578 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 13:58:38 -0600 Subject: remove old comment about tnl/vf --- src/mesa/pipe/i915simple/i915_state.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 05f8a6e1fd..d0fe7d2a52 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -626,11 +626,7 @@ static void i915_set_viewport_state( struct pipe_context *pipe, /* pass the viewport info to the draw module */ draw_set_viewport_state(i915->draw, &i915->viewport); - /* Using tnl/ and vf/ modules is temporary while getting started. - * Full pipe will have vertex shader, vertex fetch of its own. - */ i915->dirty |= I915_NEW_VIEWPORT; - } -- cgit v1.2.3 From 308bc50dc2048f28c48d68efd083c72bd501088c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 14:01:14 -0600 Subject: remove unneeded tnl stuff --- src/mesa/state_tracker/st_context.c | 7 ------- 1 file changed, 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 97e178bbe1..d2f7163d62 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -27,7 +27,6 @@ #include "main/imports.h" #include "main/extensions.h" -#include "tnl/tnl.h" #include "vbo/vbo.h" #include "st_public.h" #include "st_context.h" @@ -80,8 +79,6 @@ struct st_context *st_create_context( GLcontext *ctx, /* state tracker needs the VBO module */ _vbo_CreateContext(ctx); - /* XXX temporary */ - _tnl_CreateContext(ctx); st->draw = draw_create(); /* for selection/feedback */ @@ -117,9 +114,6 @@ struct st_context *st_create_context( GLcontext *ctx, /* XXXX This is temporary! */ _mesa_enable_sw_extensions(ctx); - /* we'll always do per-pixel fog in the fragment shader */ - _tnl_allow_vertex_fog(ctx, GL_FALSE); - return st; } @@ -131,7 +125,6 @@ void st_destroy_context( struct st_context *st ) st_destroy_draw( st ); _vbo_DestroyContext(st->ctx); - _tnl_DestroyContext(st->ctx); /* XXX temporary */ #if 0 st_destroy_cb_clear( st ); -- cgit v1.2.3 From 2072a9c6d0eec55b5fa6f8229287c8f8225c6d2e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 14:01:39 -0600 Subject: remove dead code --- src/mesa/state_tracker/st_context.c | 14 -------------- 1 file changed, 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index d2f7163d62..69942d9685 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -104,13 +104,6 @@ struct st_context *st_create_context( GLcontext *ctx, st->pixel_xfer.cache = _mesa_new_program_cache(); -#if 0 - st_init_cb_clear( st ); - st_init_cb_program( st ); - st_init_cb_drawpixels( st ); - st_init_cb_texture( st ); -#endif - /* XXXX This is temporary! */ _mesa_enable_sw_extensions(ctx); @@ -126,13 +119,6 @@ void st_destroy_context( struct st_context *st ) _vbo_DestroyContext(st->ctx); -#if 0 - st_destroy_cb_clear( st ); - st_destroy_cb_program( st ); - st_destroy_cb_drawpixels( st ); - /*st_destroy_cb_teximage( st );*/ - st_destroy_cb_texture( st ); -#endif cso_cache_delete( st->cache ); _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache); -- cgit v1.2.3 From f8549e8f4f8bdfebcb4e0a2754df59b3fe4fdff7 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 14:14:30 -0600 Subject: plug in _mesa_test_proxy_teximage, temporarily --- src/mesa/state_tracker/st_cb_texture.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index e1cf4861d0..4e6d4857f4 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1599,4 +1599,7 @@ st_init_texture_functions(struct dd_function_table *functions) functions->IsTextureResident = st_IsTextureResident; functions->TextureMemCpy = do_memcpy; + + /* XXX Temporary until we can query pipe's texture sizes */ + functions->TestProxyTexImage = _mesa_test_proxy_teximage; } -- cgit v1.2.3 From ca123a69faa377571c3b22167f7d444a26a3f776 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 14:14:53 -0600 Subject: remove _mesa_init_driver_functions() --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 29 ++++++++++++++++++----- 1 file changed, 23 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 6c880498e5..f8fe74ff1e 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -29,8 +29,7 @@ #include "glheader.h" #include "context.h" #include "extensions.h" - -#include "drivers/common/driverfuncs.h" +#include "framebuffer.h" #include "i830_dri.h" @@ -45,7 +44,7 @@ #include "pipe/p_defines.h" #include "pipe/p_context.h" -#include "drirenderbuffer.h" +/*#include "drirenderbuffer.h"*/ #include "vblank.h" #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ @@ -165,7 +164,7 @@ intelFlush(GLcontext * ctx) static void intelInitDriverFunctions(struct dd_function_table *functions) { - _mesa_init_driver_functions(functions); + memset(functions, 0, sizeof(*functions)); st_init_driver_functions(functions); } @@ -335,6 +334,24 @@ intelUnbindContext(__DRIcontextPrivate * driContextPriv) return GL_TRUE; } + +/** + * Copied/modified from drirenderbuffer.c + */ +static void +updateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv) +{ + struct gl_framebuffer *fb = (struct gl_framebuffer *) dPriv->driverPrivate; + if (fb && (dPriv->w != fb->Width || dPriv->h != fb->Height)) { + _mesa_resize_framebuffer(ctx, fb, dPriv->w, dPriv->h); + /* if the driver needs the hw lock for ResizeBuffers, the drawable + might have changed again by now */ + assert(fb->Width == dPriv->w); + assert(fb->Height == dPriv->h); + } +} + + GLboolean intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, @@ -360,10 +377,10 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, intel->intelScreen->dummyctxptr = intel; /* update GLframebuffer size to match window if needed */ - driUpdateFramebufferSize(&intel->ctx, driDrawPriv); + updateFramebufferSize(&intel->ctx, driDrawPriv); if (driReadPriv != driDrawPriv) { - driUpdateFramebufferSize(&intel->ctx, driReadPriv); + updateFramebufferSize(&intel->ctx, driReadPriv); } _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); -- cgit v1.2.3 From 15c565b018f90df80493eaa0e713f1cf3eb10a38 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 14:20:42 -0600 Subject: don't call _mesa_init_driver_functions() --- src/mesa/pipe/xlib/xm_api.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 5e04092215..7f144a5d32 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -74,7 +74,6 @@ #include "renderbuffer.h" #include "teximage.h" #include "vbo/vbo.h" -#include "drivers/common/driverfuncs.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" @@ -1462,7 +1461,11 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) mesaCtx = &(c->mesa); /* initialize with default driver functions, then plug in XMesa funcs */ +#if 0 _mesa_init_driver_functions(&functions); +#else + memset(&functions, 0, sizeof(functions)); +#endif xmesa_init_driver_functions(v, &functions); st_init_driver_functions(&functions); -- cgit v1.2.3 From 64988ff7fe4deb2d439bddf4475bec6c2fbf1b8e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 14:51:08 -0600 Subject: silence the finalize_mipmap_tree msg --- src/mesa/state_tracker/st_atom_texture.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 7970bcf2b8..f25cfd386a 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -56,8 +56,10 @@ update_textures(struct st_context *st) GLboolean flush, retval; retval = st_finalize_mipmap_tree(st->ctx, st->pipe, u, &flush); +#if 0 printf("finalize_mipmap_tree returned %d, flush = %d\n", retval, flush); +#endif mt = st_get_texobj_mipmap_tree(texObj); } -- cgit v1.2.3 From 7cafaff0ebb7c3fb7461573442aa44b354682d81 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 14:51:37 -0600 Subject: disable the driverContext assertions --- src/mesa/main/context.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index e8335e19ca..563a89e686 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1044,7 +1044,7 @@ _mesa_initialize_context(GLcontext *ctx, const struct dd_function_table *driverFunctions, void *driverContext) { - ASSERT(driverContext); + /*ASSERT(driverContext);*/ assert(driverFunctions->NewTextureObject); assert(driverFunctions->FreeTexImageData); @@ -1143,7 +1143,7 @@ _mesa_create_context(const GLvisual *visual, GLcontext *ctx; ASSERT(visual); - ASSERT(driverContext); + /*ASSERT(driverContext);*/ ctx = (GLcontext *) _mesa_calloc(sizeof(GLcontext)); if (!ctx) -- cgit v1.2.3 From 80d2658e129d097f30c84fe57e07daeb81bcc790 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 17:46:04 -0600 Subject: Sketch out new create/destroy context functions which create/wrap a Mesa context. --- src/mesa/state_tracker/st_context.c | 27 +++++++++++++++++++++++++++ src/mesa/state_tracker/st_public.h | 6 ++++++ 2 files changed, 33 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 69942d9685..1d129ad077 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "main/imports.h" +#include "main/context.h" #include "main/extensions.h" #include "vbo/vbo.h" #include "st_public.h" @@ -67,6 +68,23 @@ void st_invalidate_state(GLcontext * ctx, GLuint new_state) } +struct st_context *st_create_context2(struct pipe_context *pipe, + const GLvisual *visual, + struct st_context *share) +{ + GLcontext *ctx; + GLcontext *shareCtx = share ? share->ctx : NULL; + struct dd_function_table funcs; + + memset(&funcs, 0, sizeof(funcs)); + st_init_driver_functions(&funcs); + + ctx = _mesa_create_context(visual, shareCtx, &funcs, NULL); + + return st_create_context(ctx, pipe); +} + + struct st_context *st_create_context( GLcontext *ctx, struct pipe_context *pipe ) { @@ -111,6 +129,15 @@ struct st_context *st_create_context( GLcontext *ctx, } +void st_destroy_context2( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + _mesa_free_context_data(ctx); + st_destroy_context(st); + free(ctx); +} + + void st_destroy_context( struct st_context *st ) { draw_destroy(st->draw); diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 3191549a2f..3056b5a3e7 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -36,8 +36,14 @@ struct pipe_context; struct st_context *st_create_context( GLcontext *ctx, struct pipe_context *pipe); +struct st_context *st_create_context2(struct pipe_context *pipe, + const GLvisual *visual, + struct st_context *share); + void st_destroy_context( struct st_context *st ); +void st_destroy_context2( struct st_context *st ); + void st_invalidate_state(GLcontext * ctx, GLuint new_state); #endif -- cgit v1.2.3 From 28bed6d355e7ea3acbc4dbef0490e269d560f89e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 18:01:23 -0600 Subject: Use new state tracker context/destroy funcs with wrap Mesa context. --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 45 +++++++++++++++------- src/mesa/drivers/dri/intel_winsys/intel_context.h | 12 +++--- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 17 ++++---- 3 files changed, 49 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index f8fe74ff1e..dc6a02999e 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -177,24 +177,27 @@ intelCreateContext(const __GLcontextModes * mesaVis, __DRIcontextPrivate * driContextPriv, void *sharedContextPrivate) { - struct dd_function_table functions; - struct intel_context *intel = CALLOC_STRUCT(intel_context); +#if 0 + struct dd_function_table functions; GLcontext *ctx = &intel->ctx; - GLcontext *shareCtx = (GLcontext *) sharedContextPrivate; +#endif + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; drmI830Sarea *saPriv = intelScreen->sarea; int fthrottle_mode; GLboolean havePools; +#if 0 intelInitDriverFunctions(&functions); if (!_mesa_initialize_context(&intel->ctx, mesaVis, shareCtx, &functions, (void *) intel)) return GL_FALSE; +#endif driContextPriv->driverPrivate = intel; intel->intelScreen = intelScreen; @@ -231,10 +234,13 @@ intelCreateContext(const __GLcontextModes * mesaVis, /* Disable imaging extension until convolution is working in * teximage paths: */ +#if 0 driInitExtensions(ctx, card_extensions, /* GL_TRUE, */ GL_FALSE); +#endif +#if 0 if (intel->ctx.Mesa_DXTn) { _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); _mesa_enable_extension(ctx, "GL_S3_s3tc"); @@ -242,6 +248,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) { _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); } +#endif #ifdef DEBUG __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); @@ -275,7 +282,12 @@ intelCreateContext(const __GLcontextModes * mesaVis, } } +#if 0 st_create_context( &intel->ctx, intel->pipe ); +#else + intel->st = st_create_context2(intel->pipe, mesaVis, NULL); + intel->st->ctx->DriverCtx = intel; +#endif return GL_TRUE; } @@ -285,6 +297,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) { struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; + GLcontext *ctx = intel->st->ctx; assert(intel); /* should never be null */ if (intel) { @@ -294,7 +307,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) //intel->vtbl.destroy(intel); - release_texture_heaps = (intel->ctx.Shared->RefCount == 1); + release_texture_heaps = (ctx->Shared->RefCount == 1); intel_batchbuffer_free(intel->batch); @@ -316,10 +329,14 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) */ } +#if 0 /* free the Mesa context data */ - _mesa_free_context_data(&intel->ctx); + _mesa_free_context_data(ctx); st_destroy_context(intel->ctx.st); +#else + st_destroy_context2(intel->st); +#endif } } @@ -329,8 +346,9 @@ intelUnbindContext(__DRIcontextPrivate * driContextPriv) struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; /* XXX UnbindContext is called AFTER the new context is made current. Hopefully shouldn't be a problem ? */ - FLUSH_VERTICES((&intel->ctx), 0); - intelFlush(&intel->ctx); + GLcontext *ctx = intel->st->ctx; + FLUSH_VERTICES(ctx, 0); + intelFlush(ctx); return GL_TRUE; } @@ -338,8 +356,8 @@ intelUnbindContext(__DRIcontextPrivate * driContextPriv) /** * Copied/modified from drirenderbuffer.c */ -static void -updateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv) +void +intelUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv) { struct gl_framebuffer *fb = (struct gl_framebuffer *) dPriv->driverPrivate; if (fb && (dPriv->w != fb->Width || dPriv->h != fb->Height)) { @@ -371,23 +389,24 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, struct intel_framebuffer *intel_fb = (struct intel_framebuffer *) driDrawPriv->driverPrivate; GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; + GLcontext *ctx = intel->st->ctx; /* this is a hack so we have a valid context when the region allocation is done. Need a per-screen context? */ intel->intelScreen->dummyctxptr = intel; /* update GLframebuffer size to match window if needed */ - updateFramebufferSize(&intel->ctx, driDrawPriv); + intelUpdateFramebufferSize(ctx, driDrawPriv); if (driReadPriv != driDrawPriv) { - updateFramebufferSize(&intel->ctx, driReadPriv); + intelUpdateFramebufferSize(ctx, driReadPriv); } - _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); + _mesa_make_current(ctx, &intel_fb->Base, readFb); /* The drawbuffer won't always be updated by _mesa_make_current: */ - if (intel->ctx.DrawBuffer == &intel_fb->Base) { + if (ctx->DrawBuffer == &intel_fb->Base) { if (intel->driDrawable != driDrawPriv) { if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 76c900157a..281d68cc7e 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -44,16 +44,16 @@ struct pipe_context; struct pipe_region; struct intel_context; struct _DriBufferObject; - +struct st_context; #define INTEL_MAX_FIXUP 64 struct intel_context { - GLcontext ctx; /* the parent class */ - - struct pipe_context *pipe; + struct st_context *st; + + struct pipe_context *pipe; /**< a softpipe or i915simple context */ GLint refcount; @@ -158,7 +158,7 @@ extern void intelFlush(GLcontext * ctx); static INLINE struct intel_context * intel_context(GLcontext * ctx) { - return (struct intel_context *) ctx; + return (struct intel_context *) ctx->DriverCtx; } extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer @@ -166,5 +166,7 @@ extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer extern void intel_init_region_functions(struct pipe_context *pipe); +extern void +intelUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv); #endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index a6d0d43e4d..2ffc757469 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -30,12 +30,14 @@ #include "intel_swapbuffers.h" #include "intel_batchbuffer.h" #include "intel_reg.h" +#include "intel_winsys.h" #include "context.h" #include "utils.h" #include "drirenderbuffer.h" #include "vblank.h" #include "pipe/p_context.h" +#include "state_tracker/st_context.h" #include "state_tracker/st_cb_fbo.h" @@ -95,8 +97,8 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, /* if this drawable isn't currently bound the LOCK_HARDWARE done on the current context (which is what intelScreenContext should return) might not get a contended lock and thus cliprects not updated (tests/manywin) */ - if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); if (dPriv && dPriv->numClipRects) { @@ -210,7 +212,8 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); OUT_BATCH((sbox.y1 << 16) | sbox.x1); OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + OUT_RELOC(dri_bo(backRegion->buffer), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); ADVANCE_BATCH(); @@ -245,18 +248,18 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, void intelWindowMoved(struct intel_context *intel) { - GLcontext *ctx = &intel->ctx; + GLcontext *ctx = intel->st->ctx; __DRIdrawablePrivate *dPriv = intel->driDrawable; struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - if (!intel->ctx.DrawBuffer) { + if (!intel->st->ctx->DrawBuffer) { /* when would this happen? -BP */ assert(0); intel->numClipRects = 0; } /* Update Mesa's notion of window size */ - driUpdateFramebufferSize(ctx, dPriv); + intelUpdateFramebufferSize(ctx, dPriv); intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ { @@ -615,7 +618,7 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { struct intel_context *intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate; - GLcontext *ctx = &intel->ctx; + GLcontext *ctx = intel->st->ctx; if (ctx->Visual.doubleBufferMode) { drm_clip_rect_t rect; -- cgit v1.2.3 From d8e66aca8443c6802ecd8f1a353024ed1d0f32c3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 Nov 2007 18:37:00 -0600 Subject: Start re-working SwapBuffers. intelCopyBuffer() is now intelDisplayBuffer(): it displays the given surface in the on-screen window. Added a pipe_surface parameter to winsys->flush_frontbuffer(). Front buffer rendering/flushing actually works now. But, we should only allocate the front surface on demand... --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 79 +++++++++++----------- .../drivers/dri/intel_winsys/intel_swapbuffers.h | 5 +- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 7 +- src/mesa/pipe/p_winsys.h | 4 +- src/mesa/pipe/xlib/xm_winsys.c | 14 ++-- src/mesa/state_tracker/st_cb_flush.c | 7 +- 6 files changed, 64 insertions(+), 52 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 2ffc757469..350c3f38a2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -58,15 +58,33 @@ typedef struct drm_i915_flip { #endif +/** + * Return the pipe_surface for the given renderbuffer. + */ +static struct pipe_surface * +get_color_surface(struct intel_framebuffer *intel_fb, + GLuint bufferIndex) +{ + struct st_renderbuffer *strb + = st_renderbuffer(intel_fb->Base.Attachment[bufferIndex].Renderbuffer); + if (strb) + return strb->surface; + return NULL; +} /** - * Copy the back color buffer to the front color buffer. - * Used for SwapBuffers(). + * Display a colorbuffer surface in an X window. + * Used for SwapBuffers and flushing front buffer rendering. + * + * \param dPriv the window/drawable to display into + * \param surf the surface to display + * \param rect optional subrect of surface to display (may be NULL). */ void -intelCopyBuffer(__DRIdrawablePrivate * dPriv, - const drm_clip_rect_t * rect) +intelDisplayBuffer(__DRIdrawablePrivate * dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t * rect) { struct intel_context *intel; @@ -103,49 +121,24 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, if (dPriv && dPriv->numClipRects) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; -#if 0 - const struct pipe_region *backRegion - = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? - intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : - intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); -#endif const int backWidth = intel_fb->Base.Width; const int backHeight = intel_fb->Base.Height; const int nbox = dPriv->numClipRects; const drm_clip_rect_t *pbox = dPriv->pClipRects; const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; -#if 0 - const int srcpitch = backRegion->pitch; -#endif const int cpp = intelScreen->front.cpp; int BR13, CMD; int i; - - const struct pipe_surface *backSurf; - const struct pipe_region *backRegion; - int srcpitch; - struct st_renderbuffer *strb; - - /* blit from back color buffer if it exists, else front buffer */ - strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); - if (strb) { - backSurf = strb->surface; - } - else { - strb = st_renderbuffer(intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); - backSurf = strb->surface; - } - - backRegion = backSurf->region; - srcpitch = backRegion->pitch; + const struct pipe_region *srcRegion = surf->region; + const int srcpitch= srcRegion->pitch; ASSERT(intel_fb); ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ - ASSERT(backRegion); - ASSERT(backRegion->cpp == cpp); + ASSERT(srcRegion); + ASSERT(srcRegion->cpp == cpp); DBG(SWAP, "front pitch %d back pitch %d\n", - pitch, backRegion->pitch); + pitch, srcRegion->pitch); if (cpp == 2) { BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); @@ -164,12 +157,15 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, if (pbox->x1 > pbox->x2 || pbox->y1 > pbox->y2 || pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) + pbox->y2 > intelScreen->front.height) { + /* invalid cliprect, skip it */ continue; + } box = *pbox; if (rect) { + /* intersect cliprect with user-provided src rect */ drm_clip_rect_t rrect; rrect.x1 = dPriv->x + rect->x1; @@ -212,7 +208,7 @@ intelCopyBuffer(__DRIdrawablePrivate * dPriv, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); OUT_BATCH((sbox.y1 << 16) | sbox.x1); OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(dri_bo(backRegion->buffer), + OUT_RELOC(dri_bo(srcRegion->buffer), DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); @@ -590,10 +586,13 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ if (!intelScheduleSwap(dPriv, &missed_target)) { + struct pipe_surface *back_surf + = get_color_surface(intel_fb, BUFFER_BACK_LEFT); + driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, &missed_target); - intelCopyBuffer(dPriv, NULL); + intelDisplayBuffer(dPriv, back_surf, NULL); } intel_fb->swap_count++; @@ -621,6 +620,10 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) GLcontext *ctx = intel->st->ctx; if (ctx->Visual.doubleBufferMode) { + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + struct pipe_surface *back_surf + = get_color_surface(intel_fb, BUFFER_BACK_LEFT); + drm_clip_rect_t rect; /* fixup cliprect (driDrawable may have changed?) later */ rect.x1 = x; @@ -628,7 +631,7 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) rect.x2 = w; rect.y2 = h; _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - intelCopyBuffer(dPriv, &rect); + intelDisplayBuffer(dPriv, back_surf, &rect); } } else { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index 0065eac9b2..15ed6704d2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -62,8 +62,9 @@ struct intel_framebuffer }; -void intelCopyBuffer(__DRIdrawablePrivate * dPriv, - const drm_clip_rect_t * rect); +extern void intelDisplayBuffer(__DRIdrawablePrivate * dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t * rect); extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 4569b1e3bf..7f788b8537 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -185,12 +185,13 @@ static void intel_wait_idle( struct pipe_winsys *sws ) * we copied its contents to the real frontbuffer. Our task is easy: */ static void -intel_flush_frontbuffer( struct pipe_winsys *sws ) +intel_flush_frontbuffer( struct pipe_winsys *sws, + struct pipe_surface *surf ) { struct intel_context *intel = intel_pipe_winsys(sws)->intel; __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelCopyBuffer(dPriv, NULL); + + intelDisplayBuffer(dPriv, surf, NULL); } diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 10a2caf1a2..2d4432dbca 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -53,6 +53,7 @@ struct pipe_buffer_handle; struct pipe_region; +struct pipe_surface; /** Opaque type */ struct pipe_buffer_handle; @@ -63,7 +64,8 @@ struct pipe_winsys * Do any special operations to ensure frontbuffer contents are * displayed, eg copy fake frontbuffer. */ - void (*flush_frontbuffer)( struct pipe_winsys *sws ); + void (*flush_frontbuffer)( struct pipe_winsys *sws, + struct pipe_surface *surf ); /** Debug output */ void (*printf)( struct pipe_winsys *sws, diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index e6e98ed396..5de811a66f 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -157,14 +157,14 @@ xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, } static void -xm_flush_frontbuffer(struct pipe_winsys *pws) +xm_flush_frontbuffer(struct pipe_winsys *pws, + struct pipe_surface *surf ) { - /* - struct intel_context *intel = intel_pipe_winsys(sws)->intel; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelCopyBuffer(dPriv, NULL); - */ + /* The Xlib driver's front color surfaces are actually X Windows so + * this flush is a no-op. + * If we instead did front buffer rendering to a temporary XImage, + * this would be the place to copy the Ximage to the on-screen Window. + */ } static void diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 41f21c361c..c2c3c80b87 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -35,6 +35,7 @@ #include "main/macros.h" #include "st_context.h" #include "st_cb_flush.h" +#include "st_cb_fbo.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" @@ -59,9 +60,13 @@ static void st_flush(GLcontext *ctx) if (st->flags.frontbuffer_dirty) { + struct st_renderbuffer *strb + = st_renderbuffer(ctx->DrawBuffer->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + struct pipe_surface *front_surf = strb->surface; + /* Hook for copying "fake" frontbuffer if necessary: */ - st->pipe->winsys->flush_frontbuffer( st->pipe->winsys ); + st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf ); st->flags.frontbuffer_dirty = 0; } } -- cgit v1.2.3 From 7ff0df6c2bf11a36bc6101e361484bde57595a79 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 31 Oct 2007 06:51:19 -0400 Subject: Renaming llvmtgsi to gallivm. Taking first steps on the way to supporting fragment shaders through llvm. --- src/mesa/pipe/Makefile.template | 2 +- src/mesa/pipe/draw/draw_vertex_shader.c | 6 +- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 2 +- src/mesa/pipe/llvm/gallivm.cpp | 887 +++++++++++++++++++++++++++ src/mesa/pipe/llvm/gallivm.h | 72 +++ src/mesa/pipe/llvm/llvmtgsi.cpp | 879 -------------------------- src/mesa/pipe/llvm/llvmtgsi.h | 71 --- src/mesa/pipe/softpipe/sp_quad_fs.c | 70 +++ src/mesa/pipe/softpipe/sp_state.h | 7 + src/mesa/pipe/softpipe/sp_state_fs.c | 12 + src/mesa/sources | 2 +- src/mesa/state_tracker/st_program.c | 5 +- 12 files changed, 1056 insertions(+), 959 deletions(-) create mode 100644 src/mesa/pipe/llvm/gallivm.cpp create mode 100644 src/mesa/pipe/llvm/gallivm.h delete mode 100644 src/mesa/pipe/llvm/llvmtgsi.cpp delete mode 100644 src/mesa/pipe/llvm/llvmtgsi.h (limited to 'src') diff --git a/src/mesa/pipe/Makefile.template b/src/mesa/pipe/Makefile.template index 495bf44c07..8c2f84b328 100644 --- a/src/mesa/pipe/Makefile.template +++ b/src/mesa/pipe/Makefile.template @@ -22,7 +22,7 @@ INCLUDES = \ ##### RULES ##### .c.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + $(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ .S.o: $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 056ad007cf..7768a62000 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -39,7 +39,7 @@ #include "x86/rtasm/x86sse.h" #include "pipe/tgsi/exec/tgsi_core.h" -#include "pipe/llvm/llvmtgsi.h" +#include "pipe/llvm/gallivm.h" #define DBG_VS 0 @@ -240,8 +240,10 @@ draw_create_vertex_shader(struct draw_context *draw, #endif #ifdef MESA_LLVM vs->llvm_prog = gallivm_from_tgsi(shader->tokens); - if (!draw->engine) + draw->engine = gallivm_global_cpu_engine(); + if (!draw->engine) { draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); + } else gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); #endif diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index b340ab38fd..c42e9bbd69 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -37,7 +37,7 @@ #ifdef MESA_LLVM -#include "pipe/llvm/llvmtgsi.h" +#include "pipe/llvm/gallivm.h" #include "pipe/tgsi/exec/tgsi_core.h" #define DBG 0 diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp new file mode 100644 index 0000000000..b09a2ff8ee --- /dev/null +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -0,0 +1,887 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "gallivm.h" + +#include "instructions.h" +#include "storage.h" + +#include "pipe/p_context.h" +#include "pipe/tgsi/exec/tgsi_exec.h" +#include "pipe/tgsi/exec/tgsi_token.h" +#include "pipe/tgsi/exec/tgsi_build.h" +#include "pipe/tgsi/exec/tgsi_util.h" +#include "pipe/tgsi/exec/tgsi_parse.h" +#include "pipe/tgsi/exec/tgsi_dump.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +struct gallivm_prog { + llvm::Module *module; + void *function; + int num_consts; + int id; +}; + +struct gallivm_cpu_engine { + llvm::ExecutionEngine *engine; +}; + +using namespace llvm; +#include "llvm_base_shader.cpp" + +static int GLOBAL_ID = 0; + +static inline void AddStandardCompilePasses(PassManager &PM) { + PM.add(createVerifierPass()); // Verify that input is correct + + PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp + + //PM.add(createStripSymbolsPass(true)); + + PM.add(createRaiseAllocationsPass()); // call %malloc -> malloc inst + PM.add(createCFGSimplificationPass()); // Clean up disgusting code + PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas + PM.add(createGlobalOptimizerPass()); // Optimize out global vars + PM.add(createGlobalDCEPass()); // Remove unused fns and globs + PM.add(createIPConstantPropagationPass());// IP Constant Propagation + PM.add(createDeadArgEliminationPass()); // Dead argument elimination + PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE + PM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE + + PM.add(createPruneEHPass()); // Remove dead EH info + + PM.add(createFunctionInliningPass()); // Inline small functions + PM.add(createArgumentPromotionPass()); // Scalarize uninlined fn args + + PM.add(createTailDuplicationPass()); // Simplify cfg by copying code + PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl. + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas + PM.add(createInstructionCombiningPass()); // Combine silly seq's + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createTailCallEliminationPass()); // Eliminate tail calls + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createReassociatePass()); // Reassociate expressions + PM.add(createLoopRotatePass()); + PM.add(createLICMPass()); // Hoist loop invariants + PM.add(createLoopUnswitchPass()); // Unswitch loops. + PM.add(createLoopIndexSplitPass()); // Index split loops. + PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc + PM.add(createIndVarSimplifyPass()); // Canonicalize indvars + PM.add(createLoopUnrollPass()); // Unroll small loops + PM.add(createInstructionCombiningPass()); // Clean up after the unroller + PM.add(createGVNPass()); // Remove redundancies + PM.add(createSCCPPass()); // Constant prop with SCCP + + // Run instcombine after redundancy elimination to exploit opportunities + // opened up by them. + PM.add(createInstructionCombiningPass()); + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createDeadStoreEliminationPass()); // Delete dead stores + PM.add(createAggressiveDCEPass()); // SSA based 'Aggressive DCE' + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createSimplifyLibCallsPass()); // Library Call Optimizations + PM.add(createDeadTypeEliminationPass()); // Eliminate dead types + PM.add(createConstantMergePass()); // Merge dup global constants +} + +static void +translate_declaration(llvm::Module *module, + Storage *storage, + struct tgsi_full_declaration *decl, + struct tgsi_full_declaration *fd) +{ +} + + +static void +translate_immediate(Storage *storage, + struct tgsi_full_immediate *imm) +{ + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert( 0 ); + } + } + storage->addImmediate(vec); +} + +static inline llvm::Value * +swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, + Storage *storage) +{ + int swizzle = 0; + int start = 1000; + const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + + TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; + for (int k = 0; k < 4; ++k) { + swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; + start /= 10; + } + if (swizzle != NO_SWIZZLE) { + /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ + val = storage->shuffleVector(val, swizzle); + } + return val; +} + +static void +translate_instruction(llvm::Module *module, + Storage *storage, + Instructions *instr, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi, + unsigned instno) +{ + llvm::Value *inputs[4]; + inputs[0] = 0; + inputs[1] = 0; + inputs[2] = 0; + inputs[3] = 0; + + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + llvm::Value *val = 0; + llvm::Value *indIdx = 0; + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + indIdx = storage->extractIndex(indIdx); + } + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->constElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->inputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->tempElement(src->SrcRegister.Index); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->outputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->immediateElement(src->SrcRegister.Index); + } else { + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); + return; + } + + inputs[i] = swizzleVector(val, src, storage); + } + + /*if (inputs[0]) + instr->printVector(inputs[0]); + if (inputs[1]) + instr->printVector(inputs[1]);*/ + llvm::Value *out = 0; + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); + } + break; + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } + break; + case TGSI_OPCODE_LIT: { + out = instr->lit(inputs[0]); + } + break; + case TGSI_OPCODE_RCP: { + out = instr->rcp(inputs[0]); + } + break; + case TGSI_OPCODE_RSQ: { + out = instr->rsq(inputs[0]); + } + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP3: { + out = instr->dp3(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP4: { + out = instr->dp4(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DST: { + out = instr->dst(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MIN: { + out = instr->min(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAX: { + out = instr->max(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SLT: { + out = instr->slt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SGE: { + out = instr->sge(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SUB: { + out = instr->sub(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_LERP: { + out = instr->lerp(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: { + out = instr->frc(inputs[0]); + } + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: { + out = instr->floor(inputs[0]); + } + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: { + out = instr->ex2(inputs[0]); + } + break; + case TGSI_OPCODE_LOGBASE2: { + out = instr->lg2(inputs[0]); + } + break; + case TGSI_OPCODE_POWER: { + out = instr->pow(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_CROSSPRODUCT: { + out = instr->cross(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: { + out = instr->abs(inputs[0]); + } + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: { + out = instr->dph(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_COS: + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: { + out = instr->sgt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SIN: + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_TXP: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: { + instr->cal(inst->InstructionExtLabel.Label, + storage->outputPtr(), + storage->inputPtr(), + storage->constPtr(), + storage->tempPtr()); + return; + } + break; + case TGSI_OPCODE_RET: { + instr->end(); + return; + } + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: + break; + case TGSI_OPCODE_SCS: + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: { + instr->brk(); + return; + } + break; + case TGSI_OPCODE_IF: { + instr->ifop(inputs[0]); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: { + instr->elseop(); + storage->setCurrentBlock(instr->currentBlock()); + return; //only state update + } + break; + case TGSI_OPCODE_ENDIF: { + instr->endif(); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: { + out = instr->trunc(inputs[0]); + } + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: { + instr->beginLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_BGNSUB: { + instr->bgnSub(instno); + storage->setCurrentBlock(instr->currentBlock()); + storage->pushTemps(); + return; + } + break; + case TGSI_OPCODE_ENDLOOP2: { + instr->endLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_ENDSUB: { + instr->endSub(); + storage->setCurrentBlock(instr->currentBlock()); + storage->popArguments(); + storage->popTemps(); + return; + } + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + instr->end(); + return; + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + if (!out) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } + + /* # not sure if we need this */ + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + /*TXT( "_SAT" );*/ + break; + case TGSI_SAT_MINUS_PLUS_ONE: + /*TXT( "_SAT[-1,1]" );*/ + break; + default: + assert( 0 ); + } + + /* store results */ + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); + } + } +} + +static llvm::Module * +tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) +{ + llvm::Module *mod = createBaseShader(); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned instno = 0; + Function* shader = mod->getFunction("execute_shader"); + std::ostringstream stream; + stream << "execute_shader"; + stream << prog->id; + std::string func_name = stream.str(); + shader->setName(func_name.c_str()); + + Function::arg_iterator args = shader->arg_begin(); + Value *ptr_OUT = args++; + ptr_OUT->setName("OUT"); + Value *ptr_IN = args++; + ptr_IN->setName("IN"); + Value *ptr_CONST = args++; + ptr_CONST->setName("CONST"); + Value *ptr_TEMPS = args++; + ptr_TEMPS->setName("TEMPS"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + + tgsi_parse_init(&parse, tokens); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + Storage storage(label_entry, ptr_OUT, ptr_IN, ptr_CONST, ptr_TEMPS); + Instructions instr(mod, shader, label_entry, &storage); + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declaration(mod, &storage, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediate(&storage, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + translate_instruction(mod, &storage, &instr, + &parse.FullToken.FullInstruction, + &fi, instno); + ++instno; + break; + + default: + assert(0); + } + } + + tgsi_parse_free(&parse); + + prog->num_consts = storage.numConsts(); + return mod; +} + +/*! + Translates the TGSI tokens into LLVM format. Translated representation + is stored in the gallivm_prog and returned. + After calling this function the gallivm_prog can either be used with a custom + code generator to generate machine code for the GPU which the code generator + addresses or it can be jit compiled with gallivm_cpu_jit_compile and executed + with gallivm_prog_exec to run the module on the CPU. + */ +struct gallivm_prog * +gallivm_from_tgsi(const struct tgsi_token *tokens) +{ + std::cout << "Creating llvm from: " <id = GLOBAL_ID; + tgsi_dump(tokens, 0); + + llvm::Module *mod = tgsi_to_llvm(gallivm, tokens); + gallivm->module = mod; + gallivm_prog_dump(gallivm, 0); + + /* Run optimization passes over it */ + PassManager passes; + passes.add(new TargetData(mod)); + AddStandardCompilePasses(passes); + passes.run(*mod); + + gallivm->module = mod; + + gallivm_prog_dump(gallivm, 0); + + return gallivm; +} + +void gallivm_prog_delete(struct gallivm_prog *prog) +{ + llvm::Module *mod = static_cast(prog->module); + delete mod; + prog->module = 0; + prog->function = 0; + free(prog); +} + +typedef void (*vertex_shader_runner)(float (*ainputs)[PIPE_MAX_SHADER_INPUTS][4], + float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + float (*aconsts)[4], + int num_vertices, + int num_inputs, + int num_attribs, + int num_consts); + + +/*! + This function is used to execute the gallivm_prog in software. Before calling + this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile + function. + */ +int gallivm_prog_exec(struct gallivm_prog *prog, + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + float (*consts)[4], + int num_vertices, + int num_inputs, + int num_attribs) +{ + vertex_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + runner(inputs, dests, consts, num_vertices, num_inputs, + num_attribs, prog->num_consts); + + return 0; +} + +void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) +{ + llvm::Module *mod; + if (!prog || !prog->module) + return; + + mod = static_cast(prog->module); + + if (file_prefix) { + std::ostringstream stream; + stream << file_prefix; + stream << prog->id; + stream << ".ll"; + std::string name = stream.str(); + std::ofstream out(name.c_str()); + if (!out) { + std::cerr<<"Can't open file : "<getFunctionList(); + llvm::Module::FunctionListType::const_iterator itr; + std::cout<<"; ---------- Start shader "<id<id<(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); + cpu->engine = ee; + + llvm::Function *func = mod->getFunction("run_vertex_shader"); + prog->function = ee->getPointerToFunctionOrStub(func); + CPU = cpu; + return cpu; +} + + +/*! + This function JIT compiles the given gallivm_prog with the given cpu based execution engine. + The reference to the generated machine code entry point will be stored + in the gallivm_prog program. After executing this function one can call gallivm_prog_exec + in order to execute the gallivm_prog on the CPU. + */ +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) +{ + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = cpu->engine; + assert(ee); + ee->addModuleProvider(mp); + + llvm::Function *func = mod->getFunction("run_vertex_shader"); + prog->function = ee->getPointerToFunctionOrStub(func); +} + +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) +{ + free(cpu); +} + +struct gallivm_cpu_engine * gallivm_global_cpu_engine() +{ + return CPU; +} + +#endif /* MESA_LLVM */ + + + diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h new file mode 100644 index 0000000000..b4e98c881b --- /dev/null +++ b/src/mesa/pipe/llvm/gallivm.h @@ -0,0 +1,72 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef LLVMTGSI_H +#define LLVMTGSI_H + +#if defined __cplusplus +extern "C" { +#endif + +#include "pipe/p_state.h" + +#ifdef MESA_LLVM + +struct tgsi_token; + +struct gallivm_prog; +struct gallivm_cpu_engine; + +struct gallivm_prog *gallivm_from_tgsi(const struct tgsi_token *tokens); +void gallivm_prog_delete(struct gallivm_prog *prog); +int gallivm_prog_exec(struct gallivm_prog *prog, + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + float (*consts)[4], + int num_vertices, + int num_inputs, + int num_attribs); +void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); + + +struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); +struct gallivm_cpu_engine *gallivm_global_cpu_engine(); +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); + +#endif /* MESA_LLVM */ + +#if defined __cplusplus +} // extern "C" +#endif + +#endif diff --git a/src/mesa/pipe/llvm/llvmtgsi.cpp b/src/mesa/pipe/llvm/llvmtgsi.cpp deleted file mode 100644 index c787f3ab9f..0000000000 --- a/src/mesa/pipe/llvm/llvmtgsi.cpp +++ /dev/null @@ -1,879 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "llvmtgsi.h" - -#include "instructions.h" -#include "storage.h" - -#include "pipe/p_context.h" -#include "pipe/tgsi/exec/tgsi_exec.h" -#include "pipe/tgsi/exec/tgsi_token.h" -#include "pipe/tgsi/exec/tgsi_build.h" -#include "pipe/tgsi/exec/tgsi_util.h" -#include "pipe/tgsi/exec/tgsi_parse.h" -#include "pipe/tgsi/exec/tgsi_dump.h" - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -struct gallivm_prog { - llvm::Module *module; - void *function; - int num_consts; - int id; -}; - -struct gallivm_cpu_engine { - llvm::ExecutionEngine *engine; -}; - -using namespace llvm; -#include "llvm_base_shader.cpp" - -static int GLOBAL_ID = 0; - -static inline void AddStandardCompilePasses(PassManager &PM) { - PM.add(createVerifierPass()); // Verify that input is correct - - PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp - - //PM.add(createStripSymbolsPass(true)); - - PM.add(createRaiseAllocationsPass()); // call %malloc -> malloc inst - PM.add(createCFGSimplificationPass()); // Clean up disgusting code - PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas - PM.add(createGlobalOptimizerPass()); // Optimize out global vars - PM.add(createGlobalDCEPass()); // Remove unused fns and globs - PM.add(createIPConstantPropagationPass());// IP Constant Propagation - PM.add(createDeadArgEliminationPass()); // Dead argument elimination - PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE - PM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE - - PM.add(createPruneEHPass()); // Remove dead EH info - - PM.add(createFunctionInliningPass()); // Inline small functions - PM.add(createArgumentPromotionPass()); // Scalarize uninlined fn args - - PM.add(createTailDuplicationPass()); // Simplify cfg by copying code - PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl. - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas - PM.add(createInstructionCombiningPass()); // Combine silly seq's - PM.add(createCondPropagationPass()); // Propagate conditionals - - PM.add(createTailCallEliminationPass()); // Eliminate tail calls - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createReassociatePass()); // Reassociate expressions - PM.add(createLoopRotatePass()); - PM.add(createLICMPass()); // Hoist loop invariants - PM.add(createLoopUnswitchPass()); // Unswitch loops. - PM.add(createLoopIndexSplitPass()); // Index split loops. - PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc - PM.add(createIndVarSimplifyPass()); // Canonicalize indvars - PM.add(createLoopUnrollPass()); // Unroll small loops - PM.add(createInstructionCombiningPass()); // Clean up after the unroller - PM.add(createGVNPass()); // Remove redundancies - PM.add(createSCCPPass()); // Constant prop with SCCP - - // Run instcombine after redundancy elimination to exploit opportunities - // opened up by them. - PM.add(createInstructionCombiningPass()); - PM.add(createCondPropagationPass()); // Propagate conditionals - - PM.add(createDeadStoreEliminationPass()); // Delete dead stores - PM.add(createAggressiveDCEPass()); // SSA based 'Aggressive DCE' - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createSimplifyLibCallsPass()); // Library Call Optimizations - PM.add(createDeadTypeEliminationPass()); // Eliminate dead types - PM.add(createConstantMergePass()); // Merge dup global constants -} - -static void -translate_declaration(llvm::Module *module, - Storage *storage, - struct tgsi_full_declaration *decl, - struct tgsi_full_declaration *fd) -{ -} - - -static void -translate_immediate(Storage *storage, - struct tgsi_full_immediate *imm) -{ - float vec[4]; - int i; - for (i = 0; i < imm->Immediate.Size - 1; ++i) { - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; - break; - default: - assert( 0 ); - } - } - storage->addImmediate(vec); -} - -static inline llvm::Value * -swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, - Storage *storage) -{ - int swizzle = 0; - int start = 1000; - const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + - TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; - for (int k = 0; k < 4; ++k) { - swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; - start /= 10; - } - if (swizzle != NO_SWIZZLE) { - /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ - val = storage->shuffleVector(val, swizzle); - } - return val; -} - -static void -translate_instruction(llvm::Module *module, - Storage *storage, - Instructions *instr, - struct tgsi_full_instruction *inst, - struct tgsi_full_instruction *fi, - unsigned instno) -{ - llvm::Value *inputs[4]; - inputs[0] = 0; - inputs[1] = 0; - inputs[2] = 0; - inputs[3] = 0; - - for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - llvm::Value *val = 0; - llvm::Value *indIdx = 0; - - if (src->SrcRegister.Indirect) { - indIdx = storage->addrElement(src->SrcRegisterInd.Index); - indIdx = storage->extractIndex(indIdx); - } - if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->constElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->inputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->tempElement(src->SrcRegister.Index); - } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->outputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->immediateElement(src->SrcRegister.Index); - } else { - fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); - return; - } - - inputs[i] = swizzleVector(val, src, storage); - } - - /*if (inputs[0]) - instr->printVector(inputs[0]); - if (inputs[1]) - instr->printVector(inputs[1]);*/ - llvm::Value *out = 0; - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: { - out = instr->arl(inputs[0]); - } - break; - case TGSI_OPCODE_MOV: { - out = inputs[0]; - } - break; - case TGSI_OPCODE_LIT: { - out = instr->lit(inputs[0]); - } - break; - case TGSI_OPCODE_RCP: { - out = instr->rcp(inputs[0]); - } - break; - case TGSI_OPCODE_RSQ: { - out = instr->rsq(inputs[0]); - } - break; - case TGSI_OPCODE_EXP: - break; - case TGSI_OPCODE_LOG: - break; - case TGSI_OPCODE_MUL: { - out = instr->mul(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_ADD: { - out = instr->add(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP3: { - out = instr->dp3(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP4: { - out = instr->dp4(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DST: { - out = instr->dst(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MIN: { - out = instr->min(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAX: { - out = instr->max(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SLT: { - out = instr->slt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SGE: { - out = instr->sge(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAD: { - out = instr->madd(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SUB: { - out = instr->sub(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_LERP: { - out = instr->lerp(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_CND: - break; - case TGSI_OPCODE_CND0: - break; - case TGSI_OPCODE_DOT2ADD: - break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { - out = instr->frc(inputs[0]); - } - break; - case TGSI_OPCODE_CLAMP: - break; - case TGSI_OPCODE_FLOOR: { - out = instr->floor(inputs[0]); - } - break; - case TGSI_OPCODE_ROUND: - break; - case TGSI_OPCODE_EXPBASE2: { - out = instr->ex2(inputs[0]); - } - break; - case TGSI_OPCODE_LOGBASE2: { - out = instr->lg2(inputs[0]); - } - break; - case TGSI_OPCODE_POWER: { - out = instr->pow(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_CROSSPRODUCT: { - out = instr->cross(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; - case TGSI_OPCODE_ABS: { - out = instr->abs(inputs[0]); - } - break; - case TGSI_OPCODE_RCC: - break; - case TGSI_OPCODE_DPH: { - out = instr->dph(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_COS: - break; - case TGSI_OPCODE_DDX: - break; - case TGSI_OPCODE_DDY: - break; - case TGSI_OPCODE_KILP: - break; - case TGSI_OPCODE_PK2H: - break; - case TGSI_OPCODE_PK2US: - break; - case TGSI_OPCODE_PK4B: - break; - case TGSI_OPCODE_PK4UB: - break; - case TGSI_OPCODE_RFL: - break; - case TGSI_OPCODE_SEQ: - break; - case TGSI_OPCODE_SFL: - break; - case TGSI_OPCODE_SGT: { - out = instr->sgt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SIN: - break; - case TGSI_OPCODE_SLE: - break; - case TGSI_OPCODE_SNE: - break; - case TGSI_OPCODE_STR: - break; - case TGSI_OPCODE_TEX: - break; - case TGSI_OPCODE_TXD: - break; - case TGSI_OPCODE_TXP: - break; - case TGSI_OPCODE_UP2H: - break; - case TGSI_OPCODE_UP2US: - break; - case TGSI_OPCODE_UP4B: - break; - case TGSI_OPCODE_UP4UB: - break; - case TGSI_OPCODE_X2D: - break; - case TGSI_OPCODE_ARA: - break; - case TGSI_OPCODE_ARR: - break; - case TGSI_OPCODE_BRA: - break; - case TGSI_OPCODE_CAL: { - instr->cal(inst->InstructionExtLabel.Label, - storage->outputPtr(), - storage->inputPtr(), - storage->constPtr(), - storage->tempPtr()); - return; - } - break; - case TGSI_OPCODE_RET: { - instr->end(); - return; - } - break; - case TGSI_OPCODE_SSG: - break; - case TGSI_OPCODE_CMP: - break; - case TGSI_OPCODE_SCS: - break; - case TGSI_OPCODE_TXB: - break; - case TGSI_OPCODE_NRM: - break; - case TGSI_OPCODE_DIV: - break; - case TGSI_OPCODE_DP2: - break; - case TGSI_OPCODE_TXL: - break; - case TGSI_OPCODE_BRK: { - instr->brk(); - return; - } - break; - case TGSI_OPCODE_IF: { - instr->ifop(inputs[0]); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_LOOP: - break; - case TGSI_OPCODE_REP: - break; - case TGSI_OPCODE_ELSE: { - instr->elseop(); - storage->setCurrentBlock(instr->currentBlock()); - return; //only state update - } - break; - case TGSI_OPCODE_ENDIF: { - instr->endif(); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_ENDLOOP: - break; - case TGSI_OPCODE_ENDREP: - break; - case TGSI_OPCODE_PUSHA: - break; - case TGSI_OPCODE_POPA: - break; - case TGSI_OPCODE_CEIL: - break; - case TGSI_OPCODE_I2F: - break; - case TGSI_OPCODE_NOT: - break; - case TGSI_OPCODE_TRUNC: { - out = instr->trunc(inputs[0]); - } - break; - case TGSI_OPCODE_SHL: - break; - case TGSI_OPCODE_SHR: - break; - case TGSI_OPCODE_AND: - break; - case TGSI_OPCODE_OR: - break; - case TGSI_OPCODE_MOD: - break; - case TGSI_OPCODE_XOR: - break; - case TGSI_OPCODE_SAD: - break; - case TGSI_OPCODE_TXF: - break; - case TGSI_OPCODE_TXQ: - break; - case TGSI_OPCODE_CONT: - break; - case TGSI_OPCODE_EMIT: - break; - case TGSI_OPCODE_ENDPRIM: - break; - case TGSI_OPCODE_BGNLOOP2: { - instr->beginLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_BGNSUB: { - instr->bgnSub(instno); - storage->setCurrentBlock(instr->currentBlock()); - storage->pushTemps(); - return; - } - break; - case TGSI_OPCODE_ENDLOOP2: { - instr->endLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_ENDSUB: { - instr->endSub(); - storage->setCurrentBlock(instr->currentBlock()); - storage->popArguments(); - storage->popTemps(); - return; - } - break; - case TGSI_OPCODE_NOISE1: - break; - case TGSI_OPCODE_NOISE2: - break; - case TGSI_OPCODE_NOISE3: - break; - case TGSI_OPCODE_NOISE4: - break; - case TGSI_OPCODE_NOP: - break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; - case TGSI_OPCODE_NRM4: - break; - case TGSI_OPCODE_CALLNZ: - break; - case TGSI_OPCODE_IFC: - break; - case TGSI_OPCODE_BREAKC: - break; - case TGSI_OPCODE_KIL: - break; - case TGSI_OPCODE_END: - instr->end(); - return; - break; - default: - fprintf(stderr, "ERROR: Unknown opcode %d\n", - inst->Instruction.Opcode); - assert(0); - break; - } - - if (!out) { - fprintf(stderr, "ERROR: unsupported opcode %d\n", - inst->Instruction.Opcode); - assert(!"Unsupported opcode"); - } - - /* # not sure if we need this */ - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - /*TXT( "_SAT" );*/ - break; - case TGSI_SAT_MINUS_PLUS_ONE: - /*TXT( "_SAT[-1,1]" );*/ - break; - default: - assert( 0 ); - } - - /* store results */ - for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else { - fprintf(stderr, "ERROR: unsupported LLVM destination!"); - assert(!"wrong destination"); - } - } -} - -static llvm::Module * -tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) -{ - llvm::Module *mod = createBaseShader(); - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned instno = 0; - Function* shader = mod->getFunction("execute_shader"); - std::ostringstream stream; - stream << "execute_shader"; - stream << prog->id; - std::string func_name = stream.str(); - shader->setName(func_name.c_str()); - - Function::arg_iterator args = shader->arg_begin(); - Value *ptr_OUT = args++; - ptr_OUT->setName("OUT"); - Value *ptr_IN = args++; - ptr_IN->setName("IN"); - Value *ptr_CONST = args++; - ptr_CONST->setName("CONST"); - Value *ptr_TEMPS = args++; - ptr_TEMPS->setName("TEMPS"); - - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); - - tgsi_parse_init(&parse, tokens); - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - Storage storage(label_entry, ptr_OUT, ptr_IN, ptr_CONST, ptr_TEMPS); - Instructions instr(mod, shader, label_entry, &storage); - while(!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - translate_declaration(mod, &storage, - &parse.FullToken.FullDeclaration, - &fd); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - translate_immediate(&storage, - &parse.FullToken.FullImmediate); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - translate_instruction(mod, &storage, &instr, - &parse.FullToken.FullInstruction, - &fi, instno); - ++instno; - break; - - default: - assert(0); - } - } - - tgsi_parse_free(&parse); - - prog->num_consts = storage.numConsts(); - return mod; -} - -/*! - Translates the TGSI tokens into LLVM format. Translated representation - is stored in the gallivm_prog and returned. - After calling this function the gallivm_prog can either be used with a custom - code generator to generate machine code for the GPU which the code generator - addresses or it can be jit compiled with gallivm_cpu_jit_compile and executed - with gallivm_prog_exec to run the module on the CPU. - */ -struct gallivm_prog * -gallivm_from_tgsi(const struct tgsi_token *tokens) -{ - std::cout << "Creating llvm from: " <id = GLOBAL_ID; - tgsi_dump(tokens, 0); - - llvm::Module *mod = tgsi_to_llvm(gallivm, tokens); - gallivm->module = mod; - gallivm_prog_dump(gallivm, 0); - - /* Run optimization passes over it */ - PassManager passes; - passes.add(new TargetData(mod)); - AddStandardCompilePasses(passes); - passes.run(*mod); - - gallivm->module = mod; - - gallivm_prog_dump(gallivm, 0); - - return gallivm; -} - -void gallivm_prog_delete(struct gallivm_prog *prog) -{ - llvm::Module *mod = static_cast(prog->module); - delete mod; - prog->module = 0; - prog->function = 0; - free(prog); -} - -typedef void (*vertex_shader_runner)(float (*ainputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], - float (*aconsts)[4], - int num_vertices, - int num_inputs, - int num_attribs, - int num_consts); - - -/*! - This function is used to execute the gallivm_prog in software. Before calling - this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile - function. - */ -int gallivm_prog_exec(struct gallivm_prog *prog, - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], - float (*consts)[4], - int num_vertices, - int num_inputs, - int num_attribs) -{ - vertex_shader_runner runner = reinterpret_cast(prog->function); - assert(runner); - runner(inputs, dests, consts, num_vertices, num_inputs, - num_attribs, prog->num_consts); - - return 0; -} - -void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) -{ - llvm::Module *mod; - if (!prog || !prog->module) - return; - - mod = static_cast(prog->module); - - if (file_prefix) { - std::ostringstream stream; - stream << file_prefix; - stream << prog->id; - stream << ".ll"; - std::string name = stream.str(); - std::ofstream out(name.c_str()); - if (!out) { - std::cerr<<"Can't open file : "<getFunctionList(); - llvm::Module::FunctionListType::const_iterator itr; - std::cout<<"; ---------- Start shader "<id<id<(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); - cpu->engine = ee; - - llvm::Function *func = mod->getFunction("run_vertex_shader"); - prog->function = ee->getPointerToFunctionOrStub(func); - return cpu; -} - - -/*! - This function JIT compiles the given gallivm_prog with the given cpu based execution engine. - The reference to the generated machine code entry point will be stored - in the gallivm_prog program. After executing this function one can call gallivm_prog_exec - in order to execute the gallivm_prog on the CPU. - */ -void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) -{ - llvm::Module *mod = static_cast(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = cpu->engine; - assert(ee); - ee->addModuleProvider(mp); - - llvm::Function *func = mod->getFunction("run_vertex_shader"); - prog->function = ee->getPointerToFunctionOrStub(func); -} - -void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) -{ - free(cpu); -} - -#endif /* MESA_LLVM */ - - diff --git a/src/mesa/pipe/llvm/llvmtgsi.h b/src/mesa/pipe/llvm/llvmtgsi.h deleted file mode 100644 index dadc0ea455..0000000000 --- a/src/mesa/pipe/llvm/llvmtgsi.h +++ /dev/null @@ -1,71 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef LLVMTGSI_H -#define LLVMTGSI_H - -#if defined __cplusplus -extern "C" { -#endif - -#include "pipe/p_state.h" - -#ifdef MESA_LLVM - -struct tgsi_token; - -struct gallivm_prog; -struct gallivm_cpu_engine; - -struct gallivm_prog *gallivm_from_tgsi(const struct tgsi_token *tokens); -void gallivm_prog_delete(struct gallivm_prog *prog); -int gallivm_prog_exec(struct gallivm_prog *prog, - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], - float (*consts)[4], - int num_vertices, - int num_inputs, - int num_attribs); -void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); - - -struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); -void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); -void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); - -#endif /* MESA_LLVM */ - -#if defined __cplusplus -} // extern "C" -#endif - -#endif diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index d6d7eb6489..4bc604d682 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -156,6 +156,76 @@ shade_quad( } } +#if 0 +static void +shade_quad_llvm(struct quad_stage *qs, + struct quad_header *quad) +{ + struct quad_shade_stage *qss = quad_shade_stage(qs); + struct softpipe_context *softpipe = qs->softpipe; + const float fx = (float) quad->x0; + const float fy = (float) quad->y0; + + /* Consts does not require 16 byte alignment. */ + machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; + + machine->SamplerUnits = softpipe->sampler_units; + machine->InterpCoefs = quad->coef; + + machine->Inputs[0].xyzw[0].f[0] = fx; + machine->Inputs[0].xyzw[0].f[1] = fx + 1.0f; + machine->Inputs[0].xyzw[0].f[2] = fx; + machine->Inputs[0].xyzw[0].f[3] = fx + 1.0f; + + machine->Inputs[0].xyzw[1].f[0] = fy; + machine->Inputs[0].xyzw[1].f[1] = fy; + machine->Inputs[0].xyzw[1].f[2] = fy + 1.0f; + machine->Inputs[0].xyzw[1].f[3] = fy + 1.0f; + + /* run shader */ +#if defined(__i386__) || defined(__386__) + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps, + machine->InterpCoefs ); + quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]); +#endif + ga_llvm_prog_exec(softpipe->fs->llvm_prog); + + /* store result color */ + if (qss->colorOutSlot >= 0) { + /* XXX need to handle multiple color outputs someday */ + assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] + == TGSI_SEMANTIC_COLOR); + memcpy( + quad->outputs.color, + &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); + } + + /* store result Z */ + if (qss->depthOutSlot >= 0) { + /* output[slot] is new Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; + } + } + else { + /* copy input Z (which was interpolated by the executor) to output Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; + } + } + + /* shader may cull fragments */ + if( quad->mask ) { + qs->next->run( qs->next, quad ); + } +} +#endif /** * Per-primitive (or per-begin?) setup diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 676ad06bfc..2f096a9cc9 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -35,6 +35,10 @@ #include "x86/rtasm/x86sse.h" +#ifdef MESA_LLVM +struct gallivm_prog; +#endif + /** * Softpipe fs state is derived from pipe_shader_state. */ @@ -43,6 +47,9 @@ struct sp_fragment_shader_state { #if defined(__i386__) || defined(__386__) struct x86_function sse2_program; #endif +#ifdef MESA_LLVM + struct gallivm_prog *llvm_prog; +#endif }; void * diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 351c2954f5..92b775ae51 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -33,6 +33,7 @@ #include "pipe/p_winsys.h" #include "pipe/draw/draw_context.h" #include "pipe/tgsi/exec/tgsi_core.h" +#include "pipe/llvm/gallivm.h" void * softpipe_create_fs_state(struct pipe_context *pipe, @@ -60,6 +61,17 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, } #endif +#ifdef MESA_LLVM + fprintf(stderr, "+++++++++++++++++++++++++++++++++++++++++++++++++\n"); + state->llvm_prog = gallivm_from_tgsi(state->shader.tokens); + if (!gallivm_global_cpu_engine()) { + gallivm_cpu_engine_create(state->llvm_prog); + } + else + gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); + fprintf(stderr, "+++++++++++++++++++++++++++++++++++++++++++++++++\n"); +#endif + return state; } diff --git a/src/mesa/sources b/src/mesa/sources index 8cbc1c6993..50af7d208a 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -189,7 +189,7 @@ TGSIDECO_SOURCES = \ LLVMTGSI_SOURCES = \ - pipe/llvm/llvmtgsi.cpp \ + pipe/llvm/gallivm.cpp \ pipe/llvm/storage.cpp \ pipe/llvm/instructions.cpp diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 706238cbf5..ce5384671c 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -38,7 +38,6 @@ #include "pipe/p_defines.h" #include "pipe/draw/draw_context.h" #include "pipe/tgsi/exec/tgsi_core.h" -#include "pipe/llvm/llvmtgsi.h" #include "st_context.h" #include "st_cache.h" @@ -407,9 +406,7 @@ st_translate_fragment_program(struct st_context *st, tokensOut, maxTokens); fs.tokens = tokensOut; -#ifdef MESA_LLVM - /*fs.llvm_prog = gallivm_from_tgsi(st->pipe, fs.tokens);*/ -#endif + cso = st_cached_fs_state(st, &fs); stfp->fs = cso; -- cgit v1.2.3 From 25d91c23ff834a129e537891ec3ad63197d37da5 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 31 Oct 2007 12:23:45 -0400 Subject: Add basic entry points for fragment shaders. --- src/mesa/pipe/draw/draw_vertex_shader.c | 2 +- src/mesa/pipe/llvm/gallivm.cpp | 53 ++++++++++++++++++++++++++++-- src/mesa/pipe/llvm/gallivm.h | 14 +++++++- src/mesa/pipe/llvm/llvm_entry.c | 57 +++++++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_quad_fs.c | 35 ++++++-------------- src/mesa/pipe/softpipe/sp_state_fs.c | 2 +- 6 files changed, 132 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 7768a62000..529ed288eb 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -239,7 +239,7 @@ draw_create_vertex_shader(struct draw_context *draw, } #endif #ifdef MESA_LLVM - vs->llvm_prog = gallivm_from_tgsi(shader->tokens); + vs->llvm_prog = gallivm_from_tgsi(shader->tokens, GALLIVM_VS); draw->engine = gallivm_global_cpu_engine(); if (!draw->engine) { draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index b09a2ff8ee..48c7babc45 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -73,6 +73,7 @@ struct gallivm_prog { void *function; int num_consts; int id; + enum gallivm_shader_type type; }; struct gallivm_cpu_engine { @@ -722,7 +723,7 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) with gallivm_prog_exec to run the module on the CPU. */ struct gallivm_prog * -gallivm_from_tgsi(const struct tgsi_token *tokens) +gallivm_from_tgsi(const struct tgsi_token *tokens, enum gallivm_shader_type type) { std::cout << "Creating llvm from: " <module = mod; + gallivm->type = type; gallivm_prog_dump(gallivm, 0); return gallivm; } + void gallivm_prog_delete(struct gallivm_prog *prog) { llvm::Module *mod = static_cast(prog->module); @@ -787,6 +790,28 @@ int gallivm_prog_exec(struct gallivm_prog *prog, return 0; } + +typedef int (*fragment_shader_runner)(float x, float y, + float (*dests)[32][4], + struct tgsi_interp_coef *coef, + float (*consts)[4], int num_consts, + struct tgsi_sampler *samplers, + int num_samplers); +int gallivm_fragment_shader_exec(struct gallivm_prog *prog, + float x, float y, + float (*dests)[32][4], + struct tgsi_interp_coef *coef, + float (*consts)[4], + struct tgsi_sampler *samplers, + int num_samplers) +{ + fragment_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + runner(x, y, dests, coef, consts, prog->num_consts, samplers, num_samplers); + + return 0; +} + void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) { llvm::Module *mod; @@ -829,6 +854,26 @@ void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) static struct gallivm_cpu_engine *CPU = 0; + +static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) +{ + llvm::Module *mod = prog->module; + llvm::Function *func = 0; + + switch (prog->type) { + case GALLIVM_VS: + func = mod->getFunction("run_vertex_shader"); + break; + case GALLIVM_FS: + func = mod->getFunction("run_fragment_shader"); + break; + default: + assert(!"Unknown shader type!"); + break; + } + return func; +} + /*! This function creates a CPU based execution engine for the given gallivm_prog. gallivm_cpu_engine should be used as a singleton throughout the library. Before @@ -846,7 +891,8 @@ struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); cpu->engine = ee; - llvm::Function *func = mod->getFunction("run_vertex_shader"); + llvm::Function *func = func_for_shader(prog); + prog->function = ee->getPointerToFunctionOrStub(func); CPU = cpu; return cpu; @@ -867,7 +913,7 @@ void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog assert(ee); ee->addModuleProvider(mp); - llvm::Function *func = mod->getFunction("run_vertex_shader"); + llvm::Function *func = func_for_shader(prog); prog->function = ee->getPointerToFunctionOrStub(func); } @@ -885,3 +931,4 @@ struct gallivm_cpu_engine * gallivm_global_cpu_engine() + diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h index b4e98c881b..636a585dae 100644 --- a/src/mesa/pipe/llvm/gallivm.h +++ b/src/mesa/pipe/llvm/gallivm.h @@ -46,7 +46,12 @@ struct tgsi_token; struct gallivm_prog; struct gallivm_cpu_engine; -struct gallivm_prog *gallivm_from_tgsi(const struct tgsi_token *tokens); +enum gallivm_shader_type { + GALLIVM_VS, + GALLIVM_FS +}; + +struct gallivm_prog *gallivm_from_tgsi(const struct tgsi_token *tokens, enum gallivm_shader_type type); void gallivm_prog_delete(struct gallivm_prog *prog); int gallivm_prog_exec(struct gallivm_prog *prog, float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], @@ -55,6 +60,13 @@ int gallivm_prog_exec(struct gallivm_prog *prog, int num_vertices, int num_inputs, int num_attribs); +int gallivm_fragment_shader_exec(struct gallivm_prog *prog, + float x, float y, + float (*dests)[4], + struct tgsi_interp_coef *coef, + float (*consts)[4], + struct tgsi_sampler *samplers, + int num_samplers); void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index 67066455ae..99fa64057f 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -191,3 +191,60 @@ void run_vertex_shader(float (*ainputs)[16][4], to_array(dests[i], res, num_attribs); } } + + +struct pipe_sampler_state; +struct pipe_mipmap_tree; +struct softpipe_tile_cache; + +#define NUM_CHANNELS 4 /* R,G,B,A */ +#define QUAD_SIZE 4 /* 4 pixel/quad */ + +struct tgsi_sampler +{ + const struct pipe_sampler_state *state; + struct pipe_mipmap_tree *texture; + /** Get samples for four fragments in a quad */ + void (*get_samples)(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); + void *pipe; /*XXX temporary*/ + struct softpipe_tile_cache *cache; +}; + +struct tgsi_interp_coef +{ + float a0[NUM_CHANNELS]; /* in an xyzw layout */ + float dadx[NUM_CHANNELS]; + float dady[NUM_CHANNELS]; +}; + +int run_fragment_shader(float x, float y, + float (*dests)[32][4], + struct tgsi_interp_coef *coef, + float (*consts)[4], + int num_consts, + struct tgsi_sampler *samplers, + int num_samplers) +{ + float4 inputs[4][16]; + float4 consts[32]; + float4 results[4][16]; + float4 temps[128];//MAX_PROGRAM_TEMPS + + /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", + num_vertices, num_inputs, num_attribs, num_consts);*/ + //from_array(inputs, ainputs, num_vertices, num_inputs); + from_consts(consts, aconsts, num_consts); + printf("AAAAAAAAAAAAAAAAAAAAAAA FRAGMENT SHADER %f %f\n", x, y); + for (int i = 0; i < 4; ++i) { + float4 *in = inputs[i]; + float4 *res = results[i]; + //execute_shader(res, in, consts, temps); + to_array(dests[i], res, num_attribs); + } +} + diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 4bc604d682..cd8d337044 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -91,6 +91,11 @@ shade_quad( machine->SamplerUnits = softpipe->sampler_units; machine->InterpCoefs = quad->coef; + printf("COEF = [%f %f %f %f], [%f %f %f %f], [%f %f %f %f] %p\n", + quad->coef->a0[0], quad->coef->a0[1], quad->coef->a0[2], quad->coef->a0[3], + quad->coef->dadx[0], quad->coef->dadx[1], quad->coef->dadx[2], quad->coef->dadx[3], + quad->coef->dady[0], quad->coef->dady[1], quad->coef->dady[2], quad->coef->dady[3], + quad->coef); machine->Inputs[0].xyzw[0].f[0] = fx; machine->Inputs[0].xyzw[0].f[1] = fx + 1.0f; @@ -165,33 +170,13 @@ shade_quad_llvm(struct quad_stage *qs, struct softpipe_context *softpipe = qs->softpipe; const float fx = (float) quad->x0; const float fy = (float) quad->y0; + struct gallivm_prog *llvm = qss->llvm_prog; - /* Consts does not require 16 byte alignment. */ - machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; - - machine->SamplerUnits = softpipe->sampler_units; - machine->InterpCoefs = quad->coef; - - machine->Inputs[0].xyzw[0].f[0] = fx; - machine->Inputs[0].xyzw[0].f[1] = fx + 1.0f; - machine->Inputs[0].xyzw[0].f[2] = fx; - machine->Inputs[0].xyzw[0].f[3] = fx + 1.0f; - - machine->Inputs[0].xyzw[1].f[0] = fy; - machine->Inputs[0].xyzw[1].f[1] = fy; - machine->Inputs[0].xyzw[1].f[2] = fy + 1.0f; - machine->Inputs[0].xyzw[1].f[3] = fy + 1.0f; - /* run shader */ -#if defined(__i386__) || defined(__386__) - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps, - machine->InterpCoefs ); - quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]); -#endif - ga_llvm_prog_exec(softpipe->fs->llvm_prog); + quad->mask = gallivm_fragment_shader_exec( + llvm, fx, fy, quad->coef, + softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], + qss->samplers, softpipe->sampler_units); /* store result color */ if (qss->colorOutSlot >= 0) { diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 92b775ae51..08a7b58a6f 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -63,7 +63,7 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, #ifdef MESA_LLVM fprintf(stderr, "+++++++++++++++++++++++++++++++++++++++++++++++++\n"); - state->llvm_prog = gallivm_from_tgsi(state->shader.tokens); + state->llvm_prog = gallivm_from_tgsi(state->shader.tokens, GALLIVM_FS); if (!gallivm_global_cpu_engine()) { gallivm_cpu_engine_create(state->llvm_prog); } -- cgit v1.2.3 From 2af2f7e419c1b6d796822a049f019afe3dfc6021 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 1 Nov 2007 05:53:44 -0400 Subject: Change the fragment shader signature to better match actual arguments that we need there. --- src/mesa/pipe/llvm/gallivm.cpp | 7 +- src/mesa/pipe/llvm/gallivm.h | 6 +- src/mesa/pipe/llvm/llvm_base_shader.cpp | 1117 +++++++++++++++++++++---------- src/mesa/pipe/llvm/llvm_entry.c | 32 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 19 +- 5 files changed, 819 insertions(+), 362 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index 48c7babc45..6642f15498 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -796,18 +796,19 @@ typedef int (*fragment_shader_runner)(float x, float y, struct tgsi_interp_coef *coef, float (*consts)[4], int num_consts, struct tgsi_sampler *samplers, - int num_samplers); + unsigned *sampler_units); + int gallivm_fragment_shader_exec(struct gallivm_prog *prog, float x, float y, float (*dests)[32][4], struct tgsi_interp_coef *coef, float (*consts)[4], struct tgsi_sampler *samplers, - int num_samplers) + unsigned *sampler_units) { fragment_shader_runner runner = reinterpret_cast(prog->function); assert(runner); - runner(x, y, dests, coef, consts, prog->num_consts, samplers, num_samplers); + runner(x, y, dests, coef, consts, prog->num_consts, samplers, sampler_units); return 0; } diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h index 636a585dae..aaaabf25b1 100644 --- a/src/mesa/pipe/llvm/gallivm.h +++ b/src/mesa/pipe/llvm/gallivm.h @@ -45,6 +45,8 @@ struct tgsi_token; struct gallivm_prog; struct gallivm_cpu_engine; +struct tgsi_interp_coef; +struct tgsi_sampler; enum gallivm_shader_type { GALLIVM_VS, @@ -62,11 +64,11 @@ int gallivm_prog_exec(struct gallivm_prog *prog, int num_attribs); int gallivm_fragment_shader_exec(struct gallivm_prog *prog, float x, float y, - float (*dests)[4], + float (*dests)[32][4], struct tgsi_interp_coef *coef, float (*consts)[4], struct tgsi_sampler *samplers, - int num_samplers); + unsigned *sampler_units); void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index bf95ba3846..54c7f15924 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -6,64 +6,84 @@ Module* createBaseShader() { Module* mod = new Module("Shader"); // Type Definitions - std::vectorFuncTy_0_args; - VectorType* VectorTy_3 = VectorType::get(Type::FloatTy, 4); + OpaqueType* OpaqueTy_struct_pipe_mipmap_tree = OpaqueType::get(); + mod->addTypeName("struct.pipe_mipmap_tree", OpaqueTy_struct_pipe_mipmap_tree); - ArrayType* ArrayTy_2 = ArrayType::get(VectorTy_3, 16); + OpaqueType* OpaqueTy_struct_pipe_sampler_state = OpaqueType::get(); + mod->addTypeName("struct.pipe_sampler_state", OpaqueTy_struct_pipe_sampler_state); - PointerType* PointerTy_1 = PointerType::get(ArrayTy_2); + OpaqueType* OpaqueTy_struct_softpipe_tile_cache = OpaqueType::get(); + mod->addTypeName("struct.softpipe_tile_cache", OpaqueTy_struct_softpipe_tile_cache); - FuncTy_0_args.push_back(PointerTy_1); - ArrayType* ArrayTy_6 = ArrayType::get(Type::FloatTy, 4); + std::vectorStructTy_struct_tgsi_interp_coef_fields; + ArrayType* ArrayTy_0 = ArrayType::get(Type::FloatTy, 4); - ArrayType* ArrayTy_5 = ArrayType::get(ArrayTy_6, 16); + StructTy_struct_tgsi_interp_coef_fields.push_back(ArrayTy_0); + StructTy_struct_tgsi_interp_coef_fields.push_back(ArrayTy_0); + StructTy_struct_tgsi_interp_coef_fields.push_back(ArrayTy_0); + StructType* StructTy_struct_tgsi_interp_coef = StructType::get(StructTy_struct_tgsi_interp_coef_fields, /*isPacked=*/false); + mod->addTypeName("struct.tgsi_interp_coef", StructTy_struct_tgsi_interp_coef); - PointerType* PointerTy_4 = PointerType::get(ArrayTy_5); + std::vectorStructTy_struct_tgsi_sampler_fields; + PointerType* PointerTy_1 = PointerType::get(OpaqueTy_struct_pipe_sampler_state); - FuncTy_0_args.push_back(PointerTy_4); - FuncTy_0_args.push_back(IntegerType::get(32)); - FuncTy_0_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_0_PAL = 0; - FunctionType* FuncTy_0 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_0_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_0_PAL); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_1); + PointerType* PointerTy_2 = PointerType::get(OpaqueTy_struct_pipe_mipmap_tree); - PointerType* PointerTy_7 = PointerType::get(Type::FloatTy); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_2); + std::vectorFuncTy_4_args; + PATypeHolder StructTy_struct_tgsi_sampler_fwd = OpaqueType::get(); + PointerType* PointerTy_5 = PointerType::get(StructTy_struct_tgsi_sampler_fwd); - PointerType* PointerTy_8 = PointerType::get(VectorTy_3); + FuncTy_4_args.push_back(PointerTy_5); + PointerType* PointerTy_6 = PointerType::get(Type::FloatTy); - std::vectorFuncTy_9_args; - FuncTy_9_args.push_back(PointerTy_8); - PointerType* PointerTy_10 = PointerType::get(ArrayTy_6); + FuncTy_4_args.push_back(PointerTy_6); + FuncTy_4_args.push_back(PointerTy_6); + FuncTy_4_args.push_back(PointerTy_6); + FuncTy_4_args.push_back(Type::FloatTy); + PointerType* PointerTy_7 = PointerType::get(ArrayTy_0); - FuncTy_9_args.push_back(PointerTy_10); - FuncTy_9_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_9_PAL = 0; - FunctionType* FuncTy_9 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_9_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_9_PAL); - - std::vectorFuncTy_11_args; - FuncTy_11_args.push_back(PointerTy_10); - FuncTy_11_args.push_back(PointerTy_8); - FuncTy_11_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_11_PAL = 0; - FunctionType* FuncTy_11 = FunctionType::get( + FuncTy_4_args.push_back(PointerTy_7); + ParamAttrsList *FuncTy_4_PAL = 0; + FunctionType* FuncTy_4 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_11_args, + /*Params=*/FuncTy_4_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_11_PAL); + /*ParamAttrs=*/FuncTy_4_PAL); + + PointerType* PointerTy_3 = PointerType::get(FuncTy_4); + + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_3); + PointerType* PointerTy_8 = PointerType::get(IntegerType::get(8)); + + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_8); + PointerType* PointerTy_9 = PointerType::get(OpaqueTy_struct_softpipe_tile_cache); + + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_9); + StructType* StructTy_struct_tgsi_sampler = StructType::get(StructTy_struct_tgsi_sampler_fields, /*isPacked=*/false); + mod->addTypeName("struct.tgsi_sampler", StructTy_struct_tgsi_sampler); + cast(StructTy_struct_tgsi_sampler_fwd.get())->refineAbstractTypeTo(StructTy_struct_tgsi_sampler); + StructTy_struct_tgsi_sampler = cast(StructTy_struct_tgsi_sampler_fwd.get()); + + + ArrayType* ArrayTy_10 = ArrayType::get(IntegerType::get(8), 47); + + PointerType* PointerTy_11 = PointerType::get(ArrayTy_10); std::vectorFuncTy_12_args; - FuncTy_12_args.push_back(PointerTy_4); - FuncTy_12_args.push_back(PointerTy_4); - FuncTy_12_args.push_back(PointerTy_10); - FuncTy_12_args.push_back(IntegerType::get(32)); - FuncTy_12_args.push_back(IntegerType::get(32)); + VectorType* VectorTy_15 = VectorType::get(Type::FloatTy, 4); + + ArrayType* ArrayTy_14 = ArrayType::get(VectorTy_15, 16); + + PointerType* PointerTy_13 = PointerType::get(ArrayTy_14); + + FuncTy_12_args.push_back(PointerTy_13); + ArrayType* ArrayTy_17 = ArrayType::get(ArrayTy_0, 16); + + PointerType* PointerTy_16 = PointerType::get(ArrayTy_17); + + FuncTy_12_args.push_back(PointerTy_16); FuncTy_12_args.push_back(IntegerType::get(32)); FuncTy_12_args.push_back(IntegerType::get(32)); ParamAttrsList *FuncTy_12_PAL = 0; @@ -73,23 +93,23 @@ Module* createBaseShader() { /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_12_PAL); - ArrayType* ArrayTy_14 = ArrayType::get(ArrayTy_2, 2048); - - PointerType* PointerTy_13 = PointerType::get(ArrayTy_14); - - ArrayType* ArrayTy_16 = ArrayType::get(VectorTy_3, 32); - - PointerType* PointerTy_15 = PointerType::get(ArrayTy_16); - - ArrayType* ArrayTy_18 = ArrayType::get(VectorTy_3, 128); + PointerType* PointerTy_18 = PointerType::get(VectorTy_15); - PointerType* PointerTy_17 = PointerType::get(ArrayTy_18); + std::vectorFuncTy_19_args; + FuncTy_19_args.push_back(PointerTy_18); + FuncTy_19_args.push_back(PointerTy_7); + FuncTy_19_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_19_PAL = 0; + FunctionType* FuncTy_19 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_19_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_19_PAL); std::vectorFuncTy_20_args; - FuncTy_20_args.push_back(PointerTy_8); - FuncTy_20_args.push_back(PointerTy_8); - FuncTy_20_args.push_back(PointerTy_8); - FuncTy_20_args.push_back(PointerTy_8); + FuncTy_20_args.push_back(PointerTy_7); + FuncTy_20_args.push_back(PointerTy_18); + FuncTy_20_args.push_back(IntegerType::get(32)); ParamAttrsList *FuncTy_20_PAL = 0; FunctionType* FuncTy_20 = FunctionType::get( /*Result=*/Type::VoidTy, @@ -97,52 +117,160 @@ Module* createBaseShader() { /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_20_PAL); - PointerType* PointerTy_19 = PointerType::get(FuncTy_20); + std::vectorFuncTy_21_args; + FuncTy_21_args.push_back(PointerTy_16); + FuncTy_21_args.push_back(PointerTy_16); + FuncTy_21_args.push_back(PointerTy_7); + FuncTy_21_args.push_back(IntegerType::get(32)); + FuncTy_21_args.push_back(IntegerType::get(32)); + FuncTy_21_args.push_back(IntegerType::get(32)); + FuncTy_21_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_21_PAL = 0; + FunctionType* FuncTy_21 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_21_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_21_PAL); + + ArrayType* ArrayTy_23 = ArrayType::get(ArrayTy_14, 2048); + + PointerType* PointerTy_22 = PointerType::get(ArrayTy_23); + + ArrayType* ArrayTy_25 = ArrayType::get(VectorTy_15, 32); + + PointerType* PointerTy_24 = PointerType::get(ArrayTy_25); + + ArrayType* ArrayTy_27 = ArrayType::get(VectorTy_15, 128); + + PointerType* PointerTy_26 = PointerType::get(ArrayTy_27); + + std::vectorFuncTy_29_args; + FuncTy_29_args.push_back(PointerTy_18); + FuncTy_29_args.push_back(PointerTy_18); + FuncTy_29_args.push_back(PointerTy_18); + FuncTy_29_args.push_back(PointerTy_18); + ParamAttrsList *FuncTy_29_PAL = 0; + FunctionType* FuncTy_29 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_29_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_29_PAL); + + PointerType* PointerTy_28 = PointerType::get(FuncTy_29); + + std::vectorFuncTy_30_args; + FuncTy_30_args.push_back(Type::FloatTy); + FuncTy_30_args.push_back(Type::FloatTy); + ArrayType* ArrayTy_32 = ArrayType::get(ArrayTy_0, 32); + + PointerType* PointerTy_31 = PointerType::get(ArrayTy_32); + + FuncTy_30_args.push_back(PointerTy_31); + PointerType* PointerTy_33 = PointerType::get(StructTy_struct_tgsi_interp_coef); + + FuncTy_30_args.push_back(PointerTy_33); + FuncTy_30_args.push_back(PointerTy_7); + FuncTy_30_args.push_back(IntegerType::get(32)); + FuncTy_30_args.push_back(PointerTy_5); + PointerType* PointerTy_34 = PointerType::get(IntegerType::get(32)); + + FuncTy_30_args.push_back(PointerTy_34); + ParamAttrsList *FuncTy_30_PAL = 0; + FunctionType* FuncTy_30 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_30_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_30_PAL); + + PointerType* PointerTy_35 = PointerType::get(PointerTy_33); + + ArrayType* ArrayTy_37 = ArrayType::get(ArrayTy_14, 4); + + PointerType* PointerTy_36 = PointerType::get(ArrayTy_37); + + std::vectorFuncTy_39_args; + ParamAttrsList *FuncTy_39_PAL = 0; + FunctionType* FuncTy_39 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_39_args, + /*isVarArg=*/true, + /*ParamAttrs=*/FuncTy_39_PAL); + + PointerType* PointerTy_38 = PointerType::get(FuncTy_39); // Function Declarations Function* func_from_array = new Function( - /*Type=*/FuncTy_0, + /*Type=*/FuncTy_12, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_array", mod); func_from_array->setCallingConv(CallingConv::C); Function* func_from_consts = new Function( - /*Type=*/FuncTy_9, + /*Type=*/FuncTy_19, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_consts", mod); func_from_consts->setCallingConv(CallingConv::C); Function* func_to_array = new Function( - /*Type=*/FuncTy_11, + /*Type=*/FuncTy_20, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"to_array", mod); func_to_array->setCallingConv(CallingConv::C); Function* func_run_vertex_shader = new Function( - /*Type=*/FuncTy_12, + /*Type=*/FuncTy_21, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_vertex_shader", mod); func_run_vertex_shader->setCallingConv(CallingConv::C); Function* func_execute_shader = new Function( - /*Type=*/FuncTy_20, + /*Type=*/FuncTy_29, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"execute_shader", mod); // (external, no body) func_execute_shader->setCallingConv(CallingConv::C); + Function* func_run_fragment_shader = new Function( + /*Type=*/FuncTy_30, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"run_fragment_shader", mod); + func_run_fragment_shader->setCallingConv(CallingConv::C); + + Function* func_printf = new Function( + /*Type=*/FuncTy_39, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", mod); // (external, no body) + func_printf->setCallingConv(CallingConv::C); + // Global Variable Declarations + GlobalVariable* gvar_array__str = new GlobalVariable( + /*Type=*/ArrayTy_10, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/".str", + mod); + // Constant Definitions - Constant* const_int32_21 = Constant::getNullValue(IntegerType::get(32)); - UndefValue* const_packed_22 = UndefValue::get(VectorTy_3); - ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); - ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "2", 10)); - ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "3", 10)); + Constant* const_array_40 = ConstantArray::get("AAAAAAAAAAAAAAAAAAAAAAA FRAGMENT SHADER %f %f\x0A", true); + Constant* const_int32_41 = Constant::getNullValue(IntegerType::get(32)); + UndefValue* const_packed_42 = UndefValue::get(VectorTy_15); + ConstantInt* const_int32_43 = ConstantInt::get(APInt(32, "1", 10)); + ConstantInt* const_int32_44 = ConstantInt::get(APInt(32, "2", 10)); + ConstantInt* const_int32_45 = ConstantInt::get(APInt(32, "3", 10)); + ConstantFP* const_float_46 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + std::vector const_ptr_47_indices; + const_ptr_47_indices.push_back(const_int32_41); + const_ptr_47_indices.push_back(const_int32_41); + Constant* const_ptr_47 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_47_indices[0], const_ptr_47_indices.size() ); + ConstantInt* const_int32_48 = ConstantInt::get(APInt(32, "4", 10)); + UndefValue* const_int32_49 = UndefValue::get(IntegerType::get(32)); // Global Variable Definitions + gvar_array__str->setInitializer(const_array_40); // Function Definitions @@ -164,73 +292,73 @@ Module* createBaseShader() { BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); // Block entry (label_entry) - ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_21, "cmp", label_entry); - ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_21, "cmp5", label_entry); + ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_41, "cmp", label_entry); + ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_41, "cmp5", label_entry); BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); new BranchInst(label_forbody6, label_afterfor60, int1_bothcond, label_entry); // Block forbody6 (label_forbody6) - Argument* fwdref_27 = new Argument(IntegerType::get(32)); - Argument* fwdref_28 = new Argument(IntegerType::get(32)); + Argument* fwdref_51 = new Argument(IntegerType::get(32)); + Argument* fwdref_52 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody6); int32_i_0_reg2mem_0->reserveOperandSpace(3); - int32_i_0_reg2mem_0->addIncoming(const_int32_21, label_entry); - int32_i_0_reg2mem_0->addIncoming(fwdref_27, label_forinc57); - int32_i_0_reg2mem_0->addIncoming(fwdref_28, label_forbody6); + int32_i_0_reg2mem_0->addIncoming(const_int32_41, label_entry); + int32_i_0_reg2mem_0->addIncoming(fwdref_51, label_forinc57); + int32_i_0_reg2mem_0->addIncoming(fwdref_52, label_forbody6); - Argument* fwdref_29 = new Argument(IntegerType::get(32)); + Argument* fwdref_53 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0", label_forbody6); int32_j_0_reg2mem_0->reserveOperandSpace(3); - int32_j_0_reg2mem_0->addIncoming(fwdref_29, label_forbody6); - int32_j_0_reg2mem_0->addIncoming(const_int32_21, label_forinc57); - int32_j_0_reg2mem_0->addIncoming(const_int32_21, label_entry); + int32_j_0_reg2mem_0->addIncoming(fwdref_53, label_forbody6); + int32_j_0_reg2mem_0->addIncoming(const_int32_41, label_forinc57); + int32_j_0_reg2mem_0->addIncoming(const_int32_41, label_entry); - Argument* fwdref_30 = new Argument(VectorTy_3); - PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_3, "vec.0.reg2mem.0", label_forbody6); + Argument* fwdref_54 = new Argument(VectorTy_15); + PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_15, "vec.0.reg2mem.0", label_forbody6); packed_vec_0_reg2mem_0->reserveOperandSpace(3); - packed_vec_0_reg2mem_0->addIncoming(fwdref_30, label_forbody6); - packed_vec_0_reg2mem_0->addIncoming(const_packed_22, label_entry); - packed_vec_0_reg2mem_0->addIncoming(fwdref_30, label_forinc57); + packed_vec_0_reg2mem_0->addIncoming(fwdref_54, label_forbody6); + packed_vec_0_reg2mem_0->addIncoming(const_packed_42, label_entry); + packed_vec_0_reg2mem_0->addIncoming(fwdref_54, label_forinc57); std::vector ptr_arraydecay11_indices; ptr_arraydecay11_indices.push_back(int32_i_0_reg2mem_0); ptr_arraydecay11_indices.push_back(int32_j_0_reg2mem_0); - ptr_arraydecay11_indices.push_back(const_int32_21); + ptr_arraydecay11_indices.push_back(const_int32_41); Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); - InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_21, "tmp15", label_forbody6); + InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_41, "tmp15", label_forbody6); std::vector ptr_arrayidx23_indices; ptr_arrayidx23_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx23_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx23_indices.push_back(const_int32_23); + ptr_arrayidx23_indices.push_back(const_int32_43); Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_23, "tmp26", label_forbody6); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_43, "tmp26", label_forbody6); std::vector ptr_arrayidx34_indices; ptr_arrayidx34_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx34_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx34_indices.push_back(const_int32_24); + ptr_arrayidx34_indices.push_back(const_int32_44); Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6); LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6); - InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_24, "tmp37", label_forbody6); + InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_44, "tmp37", label_forbody6); std::vector ptr_arrayidx45_indices; ptr_arrayidx45_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx45_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx45_indices.push_back(const_int32_25); + ptr_arrayidx45_indices.push_back(const_int32_45); Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6); LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6); - InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_25, "tmp48", label_forbody6); + InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_45, "tmp48", label_forbody6); std::vector ptr_arrayidx54_indices; ptr_arrayidx54_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0); Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); - StoreInst* void_31 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); - BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_23, "inc", label_forbody6); + StoreInst* void_55 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); + BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_43, "inc", label_forbody6); ICmpInst* int1_cmp59 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp59", label_forbody6); new BranchInst(label_forbody6, label_forinc57, int1_cmp59, label_forbody6); // Block forinc57 (label_forinc57) - BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_23, "inc59", label_forinc57); + BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_43, "inc59", label_forinc57); ICmpInst* int1_cmp17 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59, int32_count, "cmp17", label_forinc57); new BranchInst(label_forbody6, label_afterfor60, int1_cmp17, label_forinc57); @@ -238,80 +366,80 @@ Module* createBaseShader() { new ReturnInst(label_afterfor60); // Resolve Forward References - fwdref_28->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_28; - fwdref_30->replaceAllUsesWith(packed_tmp48); delete fwdref_30; - fwdref_29->replaceAllUsesWith(int32_inc); delete fwdref_29; - fwdref_27->replaceAllUsesWith(int32_inc59); delete fwdref_27; + fwdref_52->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_52; + fwdref_54->replaceAllUsesWith(packed_tmp48); delete fwdref_54; + fwdref_53->replaceAllUsesWith(int32_inc); delete fwdref_53; + fwdref_51->replaceAllUsesWith(int32_inc59); delete fwdref_51; } // Function: from_consts (func_from_consts) { Function::arg_iterator args = func_from_consts->arg_begin(); - Value* ptr_res_35 = args++; - ptr_res_35->setName("res"); - Value* ptr_ainputs_36 = args++; - ptr_ainputs_36->setName("ainputs"); - Value* int32_count_37 = args++; - int32_count_37->setName("count"); - - BasicBlock* label_entry_38 = new BasicBlock("entry",func_from_consts,0); + Value* ptr_res_59 = args++; + ptr_res_59->setName("res"); + Value* ptr_ainputs_60 = args++; + ptr_ainputs_60->setName("ainputs"); + Value* int32_count_61 = args++; + int32_count_61->setName("count"); + + BasicBlock* label_entry_62 = new BasicBlock("entry",func_from_consts,0); BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); - // Block entry (label_entry_38) - ICmpInst* int1_cmp_39 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_37, const_int32_21, "cmp", label_entry_38); - new BranchInst(label_forbody, label_afterfor, int1_cmp_39, label_entry_38); + // Block entry (label_entry_62) + ICmpInst* int1_cmp_63 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_61, const_int32_41, "cmp", label_entry_62); + new BranchInst(label_forbody, label_afterfor, int1_cmp_63, label_entry_62); // Block forbody (label_forbody) - Argument* fwdref_42 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_41 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); - int32_i_0_reg2mem_0_41->reserveOperandSpace(2); - int32_i_0_reg2mem_0_41->addIncoming(const_int32_21, label_entry_38); - int32_i_0_reg2mem_0_41->addIncoming(fwdref_42, label_forbody); - - Argument* fwdref_44 = new Argument(VectorTy_3); - PHINode* packed_vec_0_reg2mem_0_43 = new PHINode(VectorTy_3, "vec.0.reg2mem.0", label_forbody); - packed_vec_0_reg2mem_0_43->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_43->addIncoming(const_packed_22, label_entry_38); - packed_vec_0_reg2mem_0_43->addIncoming(fwdref_44, label_forbody); + Argument* fwdref_66 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_65 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); + int32_i_0_reg2mem_0_65->reserveOperandSpace(2); + int32_i_0_reg2mem_0_65->addIncoming(const_int32_41, label_entry_62); + int32_i_0_reg2mem_0_65->addIncoming(fwdref_66, label_forbody); + + Argument* fwdref_68 = new Argument(VectorTy_15); + PHINode* packed_vec_0_reg2mem_0_67 = new PHINode(VectorTy_15, "vec.0.reg2mem.0", label_forbody); + packed_vec_0_reg2mem_0_67->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_67->addIncoming(const_packed_42, label_entry_62); + packed_vec_0_reg2mem_0_67->addIncoming(fwdref_68, label_forbody); std::vector ptr_arraydecay_indices; - ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_41); - ptr_arraydecay_indices.push_back(const_int32_21); - Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_36, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); + ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_65); + ptr_arraydecay_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_60, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); - InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_43, float_tmp5, const_int32_21, "tmp7", label_forbody); + InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_67, float_tmp5, const_int32_41, "tmp7", label_forbody); std::vector ptr_arrayidx12_indices; - ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_41); - ptr_arrayidx12_indices.push_back(const_int32_23); - Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_36, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); - LoadInst* float_tmp13_45 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); - InsertElementInst* packed_tmp15_46 = new InsertElementInst(packed_tmp7, float_tmp13_45, const_int32_23, "tmp15", label_forbody); + ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_65); + ptr_arrayidx12_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_60, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); + LoadInst* float_tmp13_69 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); + InsertElementInst* packed_tmp15_70 = new InsertElementInst(packed_tmp7, float_tmp13_69, const_int32_43, "tmp15", label_forbody); std::vector ptr_arrayidx20_indices; - ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_41); - ptr_arrayidx20_indices.push_back(const_int32_24); - Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_36, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); + ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_65); + ptr_arrayidx20_indices.push_back(const_int32_44); + Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_60, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); - InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_46, float_tmp21, const_int32_24, "tmp23", label_forbody); + InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_70, float_tmp21, const_int32_44, "tmp23", label_forbody); std::vector ptr_arrayidx28_indices; - ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_41); - ptr_arrayidx28_indices.push_back(const_int32_25); - Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_36, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); + ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_65); + ptr_arrayidx28_indices.push_back(const_int32_45); + Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_60, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); - InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_25, "tmp31", label_forbody); - GetElementPtrInst* ptr_arrayidx34_47 = new GetElementPtrInst(ptr_res_35, int32_i_0_reg2mem_0_41, "arrayidx34", label_forbody); - StoreInst* void_48 = new StoreInst(packed_tmp31, ptr_arrayidx34_47, false, label_forbody); - BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_41, const_int32_23, "indvar.next", label_forbody); - ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_37, "exitcond", label_forbody); + InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_45, "tmp31", label_forbody); + GetElementPtrInst* ptr_arrayidx34_71 = new GetElementPtrInst(ptr_res_59, int32_i_0_reg2mem_0_65, "arrayidx34", label_forbody); + StoreInst* void_72 = new StoreInst(packed_tmp31, ptr_arrayidx34_71, false, label_forbody); + BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_65, const_int32_43, "indvar.next", label_forbody); + ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_61, "exitcond", label_forbody); new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); // Block afterfor (label_afterfor) new ReturnInst(label_afterfor); // Resolve Forward References - fwdref_44->replaceAllUsesWith(packed_tmp31); delete fwdref_44; - fwdref_42->replaceAllUsesWith(int32_indvar_next); delete fwdref_42; + fwdref_68->replaceAllUsesWith(packed_tmp31); delete fwdref_68; + fwdref_66->replaceAllUsesWith(int32_indvar_next); delete fwdref_66; } @@ -322,81 +450,81 @@ Module* createBaseShader() { ptr_dests->setName("dests"); Value* ptr_in = args++; ptr_in->setName("in"); - Value* int32_num_attribs_51 = args++; - int32_num_attribs_51->setName("num_attribs"); - - BasicBlock* label_entry_52 = new BasicBlock("entry",func_to_array,0); - BasicBlock* label_forbody_53 = new BasicBlock("forbody",func_to_array,0); - BasicBlock* label_afterfor_54 = new BasicBlock("afterfor",func_to_array,0); - - // Block entry (label_entry_52) - ICmpInst* int1_cmp_55 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_51, const_int32_21, "cmp", label_entry_52); - new BranchInst(label_forbody_53, label_afterfor_54, int1_cmp_55, label_entry_52); - - // Block forbody (label_forbody_53) - Argument* fwdref_58 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_57 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_53); - int32_i_0_reg2mem_0_57->reserveOperandSpace(2); - int32_i_0_reg2mem_0_57->addIncoming(const_int32_21, label_entry_52); - int32_i_0_reg2mem_0_57->addIncoming(fwdref_58, label_forbody_53); - - std::vector ptr_arraydecay_59_indices; - ptr_arraydecay_59_indices.push_back(int32_i_0_reg2mem_0_57); - ptr_arraydecay_59_indices.push_back(const_int32_21); - Instruction* ptr_arraydecay_59 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_59_indices.begin(), ptr_arraydecay_59_indices.end(), "arraydecay", label_forbody_53); - GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_57, "arrayidx6", label_forbody_53); - LoadInst* packed_tmp7_60 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_53); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_60, const_int32_21, "tmp11", label_forbody_53); - StoreInst* void_61 = new StoreInst(float_tmp11, ptr_arraydecay_59, false, label_forbody_53); + Value* int32_num_attribs_75 = args++; + int32_num_attribs_75->setName("num_attribs"); + + BasicBlock* label_entry_76 = new BasicBlock("entry",func_to_array,0); + BasicBlock* label_forbody_77 = new BasicBlock("forbody",func_to_array,0); + BasicBlock* label_afterfor_78 = new BasicBlock("afterfor",func_to_array,0); + + // Block entry (label_entry_76) + ICmpInst* int1_cmp_79 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_75, const_int32_41, "cmp", label_entry_76); + new BranchInst(label_forbody_77, label_afterfor_78, int1_cmp_79, label_entry_76); + + // Block forbody (label_forbody_77) + Argument* fwdref_82 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_81 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_77); + int32_i_0_reg2mem_0_81->reserveOperandSpace(2); + int32_i_0_reg2mem_0_81->addIncoming(const_int32_41, label_entry_76); + int32_i_0_reg2mem_0_81->addIncoming(fwdref_82, label_forbody_77); + + std::vector ptr_arraydecay_83_indices; + ptr_arraydecay_83_indices.push_back(int32_i_0_reg2mem_0_81); + ptr_arraydecay_83_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay_83 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_83_indices.begin(), ptr_arraydecay_83_indices.end(), "arraydecay", label_forbody_77); + GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_81, "arrayidx6", label_forbody_77); + LoadInst* packed_tmp7_84 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_77); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_84, const_int32_41, "tmp11", label_forbody_77); + StoreInst* void_85 = new StoreInst(float_tmp11, ptr_arraydecay_83, false, label_forbody_77); std::vector ptr_arrayidx13_indices; - ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_57); - ptr_arrayidx13_indices.push_back(const_int32_23); - Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_53); - ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_60, const_int32_23, "tmp15", label_forbody_53); - StoreInst* void_62 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_53); + ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_81); + ptr_arrayidx13_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_77); + ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_84, const_int32_43, "tmp15", label_forbody_77); + StoreInst* void_86 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_77); std::vector ptr_arrayidx17_indices; - ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_57); - ptr_arrayidx17_indices.push_back(const_int32_24); - Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_53); - ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_60, const_int32_24, "tmp19", label_forbody_53); - StoreInst* void_63 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_53); + ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_81); + ptr_arrayidx17_indices.push_back(const_int32_44); + Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_77); + ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_84, const_int32_44, "tmp19", label_forbody_77); + StoreInst* void_87 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_77); std::vector ptr_arrayidx21_indices; - ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_57); - ptr_arrayidx21_indices.push_back(const_int32_25); - Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_53); - ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_60, const_int32_25, "tmp23", label_forbody_53); - StoreInst* void_64 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_53); - BinaryOperator* int32_indvar_next_65 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_57, const_int32_23, "indvar.next", label_forbody_53); - ICmpInst* int1_exitcond_66 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_65, int32_num_attribs_51, "exitcond", label_forbody_53); - new BranchInst(label_afterfor_54, label_forbody_53, int1_exitcond_66, label_forbody_53); - - // Block afterfor (label_afterfor_54) - new ReturnInst(label_afterfor_54); + ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_81); + ptr_arrayidx21_indices.push_back(const_int32_45); + Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_77); + ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_84, const_int32_45, "tmp23", label_forbody_77); + StoreInst* void_88 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_77); + BinaryOperator* int32_indvar_next_89 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_81, const_int32_43, "indvar.next", label_forbody_77); + ICmpInst* int1_exitcond_90 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_89, int32_num_attribs_75, "exitcond", label_forbody_77); + new BranchInst(label_afterfor_78, label_forbody_77, int1_exitcond_90, label_forbody_77); + + // Block afterfor (label_afterfor_78) + new ReturnInst(label_afterfor_78); // Resolve Forward References - fwdref_58->replaceAllUsesWith(int32_indvar_next_65); delete fwdref_58; + fwdref_82->replaceAllUsesWith(int32_indvar_next_89); delete fwdref_82; } // Function: run_vertex_shader (func_run_vertex_shader) { Function::arg_iterator args = func_run_vertex_shader->arg_begin(); - Value* ptr_ainputs_69 = args++; - ptr_ainputs_69->setName("ainputs"); - Value* ptr_dests_70 = args++; - ptr_dests_70->setName("dests"); + Value* ptr_ainputs_93 = args++; + ptr_ainputs_93->setName("ainputs"); + Value* ptr_dests_94 = args++; + ptr_dests_94->setName("dests"); Value* ptr_aconsts = args++; ptr_aconsts->setName("aconsts"); Value* int32_num_vertices = args++; int32_num_vertices->setName("num_vertices"); Value* int32_num_inputs = args++; int32_num_inputs->setName("num_inputs"); - Value* int32_num_attribs_71 = args++; - int32_num_attribs_71->setName("num_attribs"); + Value* int32_num_attribs_95 = args++; + int32_num_attribs_95->setName("num_attribs"); Value* int32_num_consts = args++; int32_num_consts->setName("num_consts"); - BasicBlock* label_entry_72 = new BasicBlock("entry",func_run_vertex_shader,0); + BasicBlock* label_entry_96 = new BasicBlock("entry",func_run_vertex_shader,0); BasicBlock* label_forbody6_i = new BasicBlock("forbody6.i",func_run_vertex_shader,0); BasicBlock* label_forinc57_i = new BasicBlock("forinc57.i",func_run_vertex_shader,0); BasicBlock* label_from_array_exit = new BasicBlock("from_array.exit",func_run_vertex_shader,0); @@ -406,269 +534,566 @@ Module* createBaseShader() { BasicBlock* label_forbody_us = new BasicBlock("forbody.us",func_run_vertex_shader,0); BasicBlock* label_to_array_exit_us = new BasicBlock("to_array.exit.us",func_run_vertex_shader,0); BasicBlock* label_forbody_i_us = new BasicBlock("forbody.i.us",func_run_vertex_shader,0); - BasicBlock* label_forbody_73 = new BasicBlock("forbody",func_run_vertex_shader,0); - BasicBlock* label_afterfor_74 = new BasicBlock("afterfor",func_run_vertex_shader,0); - - // Block entry (label_entry_72) - AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_14, "inputs", label_entry_72); - AllocaInst* ptr_consts = new AllocaInst(ArrayTy_16, "consts", label_entry_72); - AllocaInst* ptr_results = new AllocaInst(ArrayTy_14, "results", label_entry_72); - AllocaInst* ptr_temps = new AllocaInst(ArrayTy_18, "temps", label_entry_72); - ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_21, "cmp.i", label_entry_72); - ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_21, "cmp5.i", label_entry_72); - BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_72); - new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_72); + BasicBlock* label_forbody_97 = new BasicBlock("forbody",func_run_vertex_shader,0); + BasicBlock* label_afterfor_98 = new BasicBlock("afterfor",func_run_vertex_shader,0); + + // Block entry (label_entry_96) + AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_23, "inputs", label_entry_96); + AllocaInst* ptr_consts = new AllocaInst(ArrayTy_25, "consts", label_entry_96); + AllocaInst* ptr_results = new AllocaInst(ArrayTy_23, "results", label_entry_96); + AllocaInst* ptr_temps = new AllocaInst(ArrayTy_27, "temps", label_entry_96); + ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_41, "cmp.i", label_entry_96); + ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_41, "cmp5.i", label_entry_96); + BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_96); + new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_96); // Block forbody6.i (label_forbody6_i) - Argument* fwdref_76 = new Argument(IntegerType::get(32)); - Argument* fwdref_77 = new Argument(IntegerType::get(32)); + Argument* fwdref_100 = new Argument(IntegerType::get(32)); + Argument* fwdref_101 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i_ph = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i.ph", label_forbody6_i); int32_i_0_reg2mem_0_i_ph->reserveOperandSpace(3); - int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_21, label_entry_72); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_76, label_forinc57_i); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_77, label_forbody6_i); + int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_41, label_entry_96); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_100, label_forinc57_i); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_101, label_forbody6_i); - Argument* fwdref_78 = new Argument(IntegerType::get(32)); + Argument* fwdref_102 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i); int32_j_0_reg2mem_0_i->reserveOperandSpace(3); - int32_j_0_reg2mem_0_i->addIncoming(fwdref_78, label_forbody6_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_21, label_forinc57_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_21, label_entry_72); + int32_j_0_reg2mem_0_i->addIncoming(fwdref_102, label_forbody6_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_41, label_forinc57_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_41, label_entry_96); - Argument* fwdref_79 = new Argument(VectorTy_3); - PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_3, "vec.0.reg2mem.0.i", label_forbody6_i); + Argument* fwdref_103 = new Argument(VectorTy_15); + PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_15, "vec.0.reg2mem.0.i", label_forbody6_i); packed_vec_0_reg2mem_0_i->reserveOperandSpace(3); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_79, label_forbody6_i); - packed_vec_0_reg2mem_0_i->addIncoming(const_packed_22, label_entry_72); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_79, label_forinc57_i); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_103, label_forbody6_i); + packed_vec_0_reg2mem_0_i->addIncoming(const_packed_42, label_entry_96); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_103, label_forinc57_i); std::vector ptr_arraydecay11_i_indices; ptr_arraydecay11_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arraydecay11_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arraydecay11_i_indices.push_back(const_int32_21); - Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_69, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); + ptr_arraydecay11_i_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_93, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); LoadInst* float_tmp13_i = new LoadInst(ptr_arraydecay11_i, "tmp13.i", false, label_forbody6_i); - InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_21, "tmp15.i", label_forbody6_i); + InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_41, "tmp15.i", label_forbody6_i); std::vector ptr_arrayidx23_i_indices; ptr_arrayidx23_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx23_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx23_i_indices.push_back(const_int32_23); - Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_69, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); + ptr_arrayidx23_i_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_93, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); LoadInst* float_tmp24_i = new LoadInst(ptr_arrayidx23_i, "tmp24.i", false, label_forbody6_i); - InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_23, "tmp26.i", label_forbody6_i); + InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_43, "tmp26.i", label_forbody6_i); std::vector ptr_arrayidx34_i_indices; ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx34_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx34_i_indices.push_back(const_int32_24); - Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_69, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); + ptr_arrayidx34_i_indices.push_back(const_int32_44); + Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_93, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); LoadInst* float_tmp35_i = new LoadInst(ptr_arrayidx34_i, "tmp35.i", false, label_forbody6_i); - InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_24, "tmp37.i", label_forbody6_i); + InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_44, "tmp37.i", label_forbody6_i); std::vector ptr_arrayidx45_i_indices; ptr_arrayidx45_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx45_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx45_i_indices.push_back(const_int32_25); - Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_69, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); + ptr_arrayidx45_i_indices.push_back(const_int32_45); + Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_93, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); LoadInst* float_tmp46_i = new LoadInst(ptr_arrayidx45_i, "tmp46.i", false, label_forbody6_i); - InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_25, "tmp48.i", label_forbody6_i); + InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_45, "tmp48.i", label_forbody6_i); std::vector ptr_arrayidx54_i_indices; - ptr_arrayidx54_i_indices.push_back(const_int32_21); + ptr_arrayidx54_i_indices.push_back(const_int32_41); ptr_arrayidx54_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx54_i_indices.push_back(int32_j_0_reg2mem_0_i); Instruction* ptr_arrayidx54_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx54_i_indices.begin(), ptr_arrayidx54_i_indices.end(), "arrayidx54.i", label_forbody6_i); - StoreInst* void_80 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); - BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_23, "inc.i", label_forbody6_i); + StoreInst* void_104 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); + BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_43, "inc.i", label_forbody6_i); ICmpInst* int1_cmp59_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_inputs, "cmp59.i", label_forbody6_i); new BranchInst(label_forbody6_i, label_forinc57_i, int1_cmp59_i, label_forbody6_i); // Block forinc57.i (label_forinc57_i) - BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_23, "inc59.i", label_forinc57_i); + BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_43, "inc59.i", label_forinc57_i); ICmpInst* int1_cmp17_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59_i, int32_num_vertices, "cmp17.i", label_forinc57_i); new BranchInst(label_forbody6_i, label_from_array_exit, int1_cmp17_i, label_forinc57_i); // Block from_array.exit (label_from_array_exit) - ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_21, "cmp.i4", label_from_array_exit); + ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_41, "cmp.i4", label_from_array_exit); new BranchInst(label_forbody_i15, label_forcond, int1_cmp_i4, label_from_array_exit); // Block forbody.i15 (label_forbody_i15) - Argument* fwdref_84 = new Argument(IntegerType::get(32)); + Argument* fwdref_108 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i5 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i5", label_forbody_i15); int32_i_0_reg2mem_0_i5->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i5->addIncoming(const_int32_21, label_from_array_exit); - int32_i_0_reg2mem_0_i5->addIncoming(fwdref_84, label_forbody_i15); + int32_i_0_reg2mem_0_i5->addIncoming(const_int32_41, label_from_array_exit); + int32_i_0_reg2mem_0_i5->addIncoming(fwdref_108, label_forbody_i15); - Argument* fwdref_85 = new Argument(VectorTy_3); - PHINode* packed_vec_0_reg2mem_0_i6 = new PHINode(VectorTy_3, "vec.0.reg2mem.0.i6", label_forbody_i15); + Argument* fwdref_109 = new Argument(VectorTy_15); + PHINode* packed_vec_0_reg2mem_0_i6 = new PHINode(VectorTy_15, "vec.0.reg2mem.0.i6", label_forbody_i15); packed_vec_0_reg2mem_0_i6->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_22, label_from_array_exit); - packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_85, label_forbody_i15); + packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_42, label_from_array_exit); + packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_109, label_forbody_i15); std::vector ptr_arraydecay_i7_indices; ptr_arraydecay_i7_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arraydecay_i7_indices.push_back(const_int32_21); + ptr_arraydecay_i7_indices.push_back(const_int32_41); Instruction* ptr_arraydecay_i7 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i7_indices.begin(), ptr_arraydecay_i7_indices.end(), "arraydecay.i7", label_forbody_i15); LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i7, "tmp5.i", false, label_forbody_i15); - InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_21, "tmp7.i8", label_forbody_i15); + InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_41, "tmp7.i8", label_forbody_i15); std::vector ptr_arrayidx12_i_indices; ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx12_i_indices.push_back(const_int32_23); + ptr_arrayidx12_i_indices.push_back(const_int32_43); Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i15); LoadInst* float_tmp13_i9 = new LoadInst(ptr_arrayidx12_i, "tmp13.i9", false, label_forbody_i15); - InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_23, "tmp15.i10", label_forbody_i15); + InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_43, "tmp15.i10", label_forbody_i15); std::vector ptr_arrayidx20_i_indices; ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx20_i_indices.push_back(const_int32_24); + ptr_arrayidx20_i_indices.push_back(const_int32_44); Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i15); LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i15); - InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_24, "tmp23.i11", label_forbody_i15); + InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_44, "tmp23.i11", label_forbody_i15); std::vector ptr_arrayidx28_i_indices; ptr_arrayidx28_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx28_i_indices.push_back(const_int32_25); + ptr_arrayidx28_i_indices.push_back(const_int32_45); Instruction* ptr_arrayidx28_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i_indices.begin(), ptr_arrayidx28_i_indices.end(), "arrayidx28.i", label_forbody_i15); LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i15); - InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_25, "tmp31.i", label_forbody_i15); + InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_45, "tmp31.i", label_forbody_i15); std::vector ptr_arrayidx34_i12_indices; - ptr_arrayidx34_i12_indices.push_back(const_int32_21); + ptr_arrayidx34_i12_indices.push_back(const_int32_41); ptr_arrayidx34_i12_indices.push_back(int32_i_0_reg2mem_0_i5); Instruction* ptr_arrayidx34_i12 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i12_indices.begin(), ptr_arrayidx34_i12_indices.end(), "arrayidx34.i12", label_forbody_i15); - StoreInst* void_86 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); - BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_23, "indvar.next24", label_forbody_i15); + StoreInst* void_110 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); + BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_43, "indvar.next24", label_forbody_i15); ICmpInst* int1_exitcond25 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next24, int32_num_consts, "exitcond25", label_forbody_i15); new BranchInst(label_forcond, label_forbody_i15, int1_exitcond25, label_forbody_i15); // Block forcond (label_forcond) - new BranchInst(label_forbody_preheader, label_afterfor_74, int1_cmp_i, label_forcond); + new BranchInst(label_forbody_preheader, label_afterfor_98, int1_cmp_i, label_forcond); // Block forbody.preheader (label_forbody_preheader) std::vector ptr_arraydecay17_indices; - ptr_arraydecay17_indices.push_back(const_int32_21); - ptr_arraydecay17_indices.push_back(const_int32_21); + ptr_arraydecay17_indices.push_back(const_int32_41); + ptr_arraydecay17_indices.push_back(const_int32_41); Instruction* ptr_arraydecay17 = new GetElementPtrInst(ptr_consts, ptr_arraydecay17_indices.begin(), ptr_arraydecay17_indices.end(), "arraydecay17", label_forbody_preheader); std::vector ptr_arraydecay18_indices; - ptr_arraydecay18_indices.push_back(const_int32_21); - ptr_arraydecay18_indices.push_back(const_int32_21); + ptr_arraydecay18_indices.push_back(const_int32_41); + ptr_arraydecay18_indices.push_back(const_int32_41); Instruction* ptr_arraydecay18 = new GetElementPtrInst(ptr_temps, ptr_arraydecay18_indices.begin(), ptr_arraydecay18_indices.end(), "arraydecay18", label_forbody_preheader); - ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_71, const_int32_21, "cmp.i1", label_forbody_preheader); - new BranchInst(label_forbody_us, label_forbody_73, int1_cmp_i1, label_forbody_preheader); + ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_95, const_int32_41, "cmp.i1", label_forbody_preheader); + new BranchInst(label_forbody_us, label_forbody_97, int1_cmp_i1, label_forbody_preheader); // Block forbody.us (label_forbody_us) - Argument* fwdref_90 = new Argument(IntegerType::get(32)); + Argument* fwdref_114 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.us", label_forbody_us); int32_i_0_reg2mem_0_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_us->addIncoming(const_int32_21, label_forbody_preheader); - int32_i_0_reg2mem_0_us->addIncoming(fwdref_90, label_to_array_exit_us); + int32_i_0_reg2mem_0_us->addIncoming(const_int32_41, label_forbody_preheader); + int32_i_0_reg2mem_0_us->addIncoming(fwdref_114, label_to_array_exit_us); std::vector ptr_arraydecay10_us_indices; - ptr_arraydecay10_us_indices.push_back(const_int32_21); + ptr_arraydecay10_us_indices.push_back(const_int32_41); ptr_arraydecay10_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay10_us_indices.push_back(const_int32_21); + ptr_arraydecay10_us_indices.push_back(const_int32_41); Instruction* ptr_arraydecay10_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_us_indices.begin(), ptr_arraydecay10_us_indices.end(), "arraydecay10.us", label_forbody_us); std::vector ptr_arraydecay14_us_indices; - ptr_arraydecay14_us_indices.push_back(const_int32_21); + ptr_arraydecay14_us_indices.push_back(const_int32_41); ptr_arraydecay14_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay14_us_indices.push_back(const_int32_21); + ptr_arraydecay14_us_indices.push_back(const_int32_41); Instruction* ptr_arraydecay14_us = new GetElementPtrInst(ptr_results, ptr_arraydecay14_us_indices.begin(), ptr_arraydecay14_us_indices.end(), "arraydecay14.us", label_forbody_us); - std::vector void_91_params; - void_91_params.push_back(ptr_arraydecay14_us); - void_91_params.push_back(ptr_arraydecay10_us); - void_91_params.push_back(ptr_arraydecay17); - void_91_params.push_back(ptr_arraydecay18); - CallInst* void_91 = new CallInst(func_execute_shader, void_91_params.begin(), void_91_params.end(), "", label_forbody_us); - void_91->setCallingConv(CallingConv::C); - void_91->setTailCall(false); + std::vector void_115_params; + void_115_params.push_back(ptr_arraydecay14_us); + void_115_params.push_back(ptr_arraydecay10_us); + void_115_params.push_back(ptr_arraydecay17); + void_115_params.push_back(ptr_arraydecay18); + CallInst* void_115 = new CallInst(func_execute_shader, void_115_params.begin(), void_115_params.end(), "", label_forbody_us); + void_115->setCallingConv(CallingConv::C); + void_115->setTailCall(false); new BranchInst(label_forbody_i_us, label_forbody_us); // Block to_array.exit.us (label_to_array_exit_us) - BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_23, "inc.us", label_to_array_exit_us); + BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_43, "inc.us", label_to_array_exit_us); ICmpInst* int1_cmp21_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_num_vertices, "cmp21.us", label_to_array_exit_us); - new BranchInst(label_forbody_us, label_afterfor_74, int1_cmp21_us, label_to_array_exit_us); + new BranchInst(label_forbody_us, label_afterfor_98, int1_cmp21_us, label_to_array_exit_us); // Block forbody.i.us (label_forbody_i_us) - Argument* fwdref_94 = new Argument(IntegerType::get(32)); + Argument* fwdref_118 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i2_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i2.us", label_forbody_i_us); int32_i_0_reg2mem_0_i2_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_21, label_forbody_us); - int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_94, label_forbody_i_us); + int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_41, label_forbody_us); + int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_118, label_forbody_i_us); std::vector ptr_arraydecay_i_us_indices; ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arraydecay_i_us_indices.push_back(const_int32_21); - Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_70, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); + ptr_arraydecay_i_us_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_94, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); std::vector ptr_arrayidx6_i_us_indices; - ptr_arrayidx6_i_us_indices.push_back(const_int32_21); + ptr_arrayidx6_i_us_indices.push_back(const_int32_41); ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); Instruction* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_results, ptr_arrayidx6_i_us_indices.begin(), ptr_arrayidx6_i_us_indices.end(), "arrayidx6.i.us", label_forbody_i_us); LoadInst* packed_tmp7_i_us = new LoadInst(ptr_arrayidx6_i_us, "tmp7.i.us", false, label_forbody_i_us); - ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_21, "tmp11.i.us", label_forbody_i_us); - StoreInst* void_95 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); + ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_41, "tmp11.i.us", label_forbody_i_us); + StoreInst* void_119 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx13_i_us_indices; ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx13_i_us_indices.push_back(const_int32_23); - Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_70, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_23, "tmp15.i3.us", label_forbody_i_us); - StoreInst* void_96 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); + ptr_arrayidx13_i_us_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_94, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_43, "tmp15.i3.us", label_forbody_i_us); + StoreInst* void_120 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx17_i_us_indices; ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx17_i_us_indices.push_back(const_int32_24); - Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_70, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_24, "tmp19.i.us", label_forbody_i_us); - StoreInst* void_97 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); + ptr_arrayidx17_i_us_indices.push_back(const_int32_44); + Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_94, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_44, "tmp19.i.us", label_forbody_i_us); + StoreInst* void_121 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx21_i_us_indices; ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx21_i_us_indices.push_back(const_int32_25); - Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_70, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_25, "tmp23.i.us", label_forbody_i_us); - StoreInst* void_98 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); - BinaryOperator* int32_indvar_next_99 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_23, "indvar.next", label_forbody_i_us); - ICmpInst* int1_exitcond_100 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_99, int32_num_attribs_71, "exitcond", label_forbody_i_us); - new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_100, label_forbody_i_us); - - // Block forbody (label_forbody_73) - Argument* fwdref_103 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_102 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_73); - int32_i_0_reg2mem_0_102->reserveOperandSpace(2); - int32_i_0_reg2mem_0_102->addIncoming(const_int32_21, label_forbody_preheader); - int32_i_0_reg2mem_0_102->addIncoming(fwdref_103, label_forbody_73); + ptr_arrayidx21_i_us_indices.push_back(const_int32_45); + Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_94, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_45, "tmp23.i.us", label_forbody_i_us); + StoreInst* void_122 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); + BinaryOperator* int32_indvar_next_123 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_43, "indvar.next", label_forbody_i_us); + ICmpInst* int1_exitcond_124 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_123, int32_num_attribs_95, "exitcond", label_forbody_i_us); + new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_124, label_forbody_i_us); + + // Block forbody (label_forbody_97) + Argument* fwdref_127 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_126 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_97); + int32_i_0_reg2mem_0_126->reserveOperandSpace(2); + int32_i_0_reg2mem_0_126->addIncoming(const_int32_41, label_forbody_preheader); + int32_i_0_reg2mem_0_126->addIncoming(fwdref_127, label_forbody_97); std::vector ptr_arraydecay10_indices; - ptr_arraydecay10_indices.push_back(const_int32_21); - ptr_arraydecay10_indices.push_back(int32_i_0_reg2mem_0_102); - ptr_arraydecay10_indices.push_back(const_int32_21); - Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_forbody_73); + ptr_arraydecay10_indices.push_back(const_int32_41); + ptr_arraydecay10_indices.push_back(int32_i_0_reg2mem_0_126); + ptr_arraydecay10_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_forbody_97); std::vector ptr_arraydecay14_indices; - ptr_arraydecay14_indices.push_back(const_int32_21); - ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_102); - ptr_arraydecay14_indices.push_back(const_int32_21); - Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_results, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_73); - std::vector void_104_params; - void_104_params.push_back(ptr_arraydecay14); - void_104_params.push_back(ptr_arraydecay10); - void_104_params.push_back(ptr_arraydecay17); - void_104_params.push_back(ptr_arraydecay18); - CallInst* void_104 = new CallInst(func_execute_shader, void_104_params.begin(), void_104_params.end(), "", label_forbody_73); - void_104->setCallingConv(CallingConv::C); - void_104->setTailCall(false); - BinaryOperator* int32_inc_105 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_102, const_int32_23, "inc", label_forbody_73); - ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_105, int32_num_vertices, "cmp21", label_forbody_73); - new BranchInst(label_forbody_73, label_afterfor_74, int1_cmp21, label_forbody_73); - - // Block afterfor (label_afterfor_74) - new ReturnInst(label_afterfor_74); + ptr_arraydecay14_indices.push_back(const_int32_41); + ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_126); + ptr_arraydecay14_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_results, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_97); + std::vector void_128_params; + void_128_params.push_back(ptr_arraydecay14); + void_128_params.push_back(ptr_arraydecay10); + void_128_params.push_back(ptr_arraydecay17); + void_128_params.push_back(ptr_arraydecay18); + CallInst* void_128 = new CallInst(func_execute_shader, void_128_params.begin(), void_128_params.end(), "", label_forbody_97); + void_128->setCallingConv(CallingConv::C); + void_128->setTailCall(false); + BinaryOperator* int32_inc_129 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_126, const_int32_43, "inc", label_forbody_97); + ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_129, int32_num_vertices, "cmp21", label_forbody_97); + new BranchInst(label_forbody_97, label_afterfor_98, int1_cmp21, label_forbody_97); + + // Block afterfor (label_afterfor_98) + new ReturnInst(label_afterfor_98); + + // Resolve Forward References + fwdref_101->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_101; + fwdref_103->replaceAllUsesWith(packed_tmp48_i); delete fwdref_103; + fwdref_102->replaceAllUsesWith(int32_inc_i); delete fwdref_102; + fwdref_100->replaceAllUsesWith(int32_inc59_i); delete fwdref_100; + fwdref_109->replaceAllUsesWith(packed_tmp31_i); delete fwdref_109; + fwdref_108->replaceAllUsesWith(int32_indvar_next24); delete fwdref_108; + fwdref_114->replaceAllUsesWith(int32_inc_us); delete fwdref_114; + fwdref_118->replaceAllUsesWith(int32_indvar_next_123); delete fwdref_118; + fwdref_127->replaceAllUsesWith(int32_inc_129); delete fwdref_127; + + } + + // Function: run_fragment_shader (func_run_fragment_shader) + { + Function::arg_iterator args = func_run_fragment_shader->arg_begin(); + Value* float_x = args++; + float_x->setName("x"); + Value* float_y = args++; + float_y->setName("y"); + Value* ptr_dests_132 = args++; + ptr_dests_132->setName("dests"); + Value* ptr_coef = args++; + ptr_coef->setName("coef"); + Value* ptr_aconsts_133 = args++; + ptr_aconsts_133->setName("aconsts"); + Value* int32_num_consts_134 = args++; + int32_num_consts_134->setName("num_consts"); + Value* ptr_samplers = args++; + ptr_samplers->setName("samplers"); + Value* ptr_sampler_units = args++; + ptr_sampler_units->setName("sampler_units"); + + BasicBlock* label_entry_135 = new BasicBlock("entry",func_run_fragment_shader,0); + BasicBlock* label_forbody_i = new BasicBlock("forbody.i",func_run_fragment_shader,0); + BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); + BasicBlock* label_forbody91 = new BasicBlock("forbody91",func_run_fragment_shader,0); + BasicBlock* label_afterfor112 = new BasicBlock("afterfor112",func_run_fragment_shader,0); + + // Block entry (label_entry_135) + AllocaInst* ptr_coef_addr = new AllocaInst(PointerTy_33, "coef.addr", label_entry_135); + AllocaInst* ptr_inputs_136 = new AllocaInst(ArrayTy_37, "inputs", label_entry_135); + AllocaInst* ptr_consts_137 = new AllocaInst(ArrayTy_25, "consts", label_entry_135); + AllocaInst* ptr_results_138 = new AllocaInst(ArrayTy_37, "results", label_entry_135); + AllocaInst* ptr_temps_139 = new AllocaInst(ArrayTy_27, "temps", label_entry_135); + StoreInst* void_140 = new StoreInst(ptr_coef, ptr_coef_addr, false, label_entry_135); + InsertElementInst* packed_tmp2 = new InsertElementInst(const_packed_42, float_x, const_int32_41, "tmp2", label_entry_135); + InsertElementInst* packed_tmp5 = new InsertElementInst(packed_tmp2, float_y, const_int32_43, "tmp5", label_entry_135); + BinaryOperator* float_add = BinaryOperator::create(Instruction::Add, float_x, const_float_46, "add", label_entry_135); + InsertElementInst* packed_tmp8 = new InsertElementInst(const_packed_42, float_add, const_int32_41, "tmp8", label_entry_135); + InsertElementInst* packed_tmp11 = new InsertElementInst(packed_tmp8, float_y, const_int32_43, "tmp11", label_entry_135); + BinaryOperator* float_add16 = BinaryOperator::create(Instruction::Add, float_y, const_float_46, "add16", label_entry_135); + InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp2, float_add16, const_int32_43, "tmp18", label_entry_135); + InsertElementInst* packed_tmp26_141 = new InsertElementInst(packed_tmp8, float_add16, const_int32_43, "tmp26", label_entry_135); + std::vector ptr_arraydecay27_indices; + ptr_arraydecay27_indices.push_back(const_int32_41); + ptr_arraydecay27_indices.push_back(const_int32_41); + ptr_arraydecay27_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay27 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay27_indices.begin(), ptr_arraydecay27_indices.end(), "arraydecay27", label_entry_135); + StoreInst* void_142 = new StoreInst(packed_tmp5, ptr_arraydecay27, false, label_entry_135); + std::vector ptr_arraydecay32_indices; + ptr_arraydecay32_indices.push_back(const_int32_41); + ptr_arraydecay32_indices.push_back(const_int32_43); + ptr_arraydecay32_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay32 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay32_indices.begin(), ptr_arraydecay32_indices.end(), "arraydecay32", label_entry_135); + StoreInst* void_143 = new StoreInst(packed_tmp11, ptr_arraydecay32, false, label_entry_135); + std::vector ptr_arraydecay37_indices; + ptr_arraydecay37_indices.push_back(const_int32_41); + ptr_arraydecay37_indices.push_back(const_int32_44); + ptr_arraydecay37_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay37 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay37_indices.begin(), ptr_arraydecay37_indices.end(), "arraydecay37", label_entry_135); + StoreInst* void_144 = new StoreInst(packed_tmp18, ptr_arraydecay37, false, label_entry_135); + std::vector ptr_arraydecay42_indices; + ptr_arraydecay42_indices.push_back(const_int32_41); + ptr_arraydecay42_indices.push_back(const_int32_45); + ptr_arraydecay42_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay42 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay42_indices.begin(), ptr_arraydecay42_indices.end(), "arraydecay42", label_entry_135); + StoreInst* void_145 = new StoreInst(packed_tmp26_141, ptr_arraydecay42, false, label_entry_135); + CastInst* ptr_arraydecay48 = new BitCastInst(ptr_coef_addr, PointerTy_6, "arraydecay48", label_entry_135); + LoadInst* float_tmp50 = new LoadInst(ptr_arraydecay48, "tmp50", false, label_entry_135); + CastInst* ptr_tmp53 = new BitCastInst(ptr_coef_addr, PointerTy_33, "tmp53", label_entry_135); + std::vector ptr_arrayidx56_indices; + ptr_arrayidx56_indices.push_back(const_int32_41); + ptr_arrayidx56_indices.push_back(const_int32_41); + ptr_arrayidx56_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx56 = new GetElementPtrInst(ptr_tmp53, ptr_arrayidx56_indices.begin(), ptr_arrayidx56_indices.end(), "arrayidx56", label_entry_135); + LoadInst* float_tmp57 = new LoadInst(ptr_arrayidx56, "tmp57", false, label_entry_135); + std::vector ptr_arrayidx63_indices; + ptr_arrayidx63_indices.push_back(const_int32_41); + ptr_arrayidx63_indices.push_back(const_int32_41); + ptr_arrayidx63_indices.push_back(const_int32_44); + Instruction* ptr_arrayidx63 = new GetElementPtrInst(ptr_tmp53, ptr_arrayidx63_indices.begin(), ptr_arrayidx63_indices.end(), "arrayidx63", label_entry_135); + LoadInst* float_tmp64 = new LoadInst(ptr_arrayidx63, "tmp64", false, label_entry_135); + std::vector ptr_arrayidx70_indices; + ptr_arrayidx70_indices.push_back(const_int32_41); + ptr_arrayidx70_indices.push_back(const_int32_41); + ptr_arrayidx70_indices.push_back(const_int32_45); + Instruction* ptr_arrayidx70 = new GetElementPtrInst(ptr_tmp53, ptr_arrayidx70_indices.begin(), ptr_arrayidx70_indices.end(), "arrayidx70", label_entry_135); + LoadInst* float_tmp71 = new LoadInst(ptr_arrayidx70, "tmp71", false, label_entry_135); + InsertElementInst* packed_tmp52 = new InsertElementInst(const_packed_42, float_tmp50, const_int32_41, "tmp52", label_entry_135); + InsertElementInst* packed_tmp59 = new InsertElementInst(packed_tmp52, float_tmp57, const_int32_43, "tmp59", label_entry_135); + InsertElementInst* packed_tmp66 = new InsertElementInst(packed_tmp59, float_tmp64, const_int32_44, "tmp66", label_entry_135); + InsertElementInst* packed_tmp73 = new InsertElementInst(packed_tmp66, float_tmp71, const_int32_45, "tmp73", label_entry_135); + std::vector ptr_arrayidx78_indices; + ptr_arrayidx78_indices.push_back(const_int32_41); + ptr_arrayidx78_indices.push_back(const_int32_41); + ptr_arrayidx78_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx78 = new GetElementPtrInst(ptr_inputs_136, ptr_arrayidx78_indices.begin(), ptr_arrayidx78_indices.end(), "arrayidx78", label_entry_135); + StoreInst* void_146 = new StoreInst(packed_tmp73, ptr_arrayidx78, false, label_entry_135); + InsertElementInst* packed_tmp52_1 = new InsertElementInst(packed_tmp73, float_tmp50, const_int32_41, "tmp52.1", label_entry_135); + InsertElementInst* packed_tmp59_1 = new InsertElementInst(packed_tmp52_1, float_tmp57, const_int32_43, "tmp59.1", label_entry_135); + InsertElementInst* packed_tmp66_1 = new InsertElementInst(packed_tmp59_1, float_tmp64, const_int32_44, "tmp66.1", label_entry_135); + InsertElementInst* packed_tmp73_1 = new InsertElementInst(packed_tmp66_1, float_tmp71, const_int32_45, "tmp73.1", label_entry_135); + std::vector ptr_arrayidx78_1_indices; + ptr_arrayidx78_1_indices.push_back(const_int32_41); + ptr_arrayidx78_1_indices.push_back(const_int32_43); + ptr_arrayidx78_1_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx78_1 = new GetElementPtrInst(ptr_inputs_136, ptr_arrayidx78_1_indices.begin(), ptr_arrayidx78_1_indices.end(), "arrayidx78.1", label_entry_135); + StoreInst* void_147 = new StoreInst(packed_tmp73_1, ptr_arrayidx78_1, false, label_entry_135); + InsertElementInst* packed_tmp52_2 = new InsertElementInst(packed_tmp73_1, float_tmp50, const_int32_41, "tmp52.2", label_entry_135); + InsertElementInst* packed_tmp59_2 = new InsertElementInst(packed_tmp52_2, float_tmp57, const_int32_43, "tmp59.2", label_entry_135); + InsertElementInst* packed_tmp66_2 = new InsertElementInst(packed_tmp59_2, float_tmp64, const_int32_44, "tmp66.2", label_entry_135); + InsertElementInst* packed_tmp73_2 = new InsertElementInst(packed_tmp66_2, float_tmp71, const_int32_45, "tmp73.2", label_entry_135); + std::vector ptr_arrayidx78_2_indices; + ptr_arrayidx78_2_indices.push_back(const_int32_41); + ptr_arrayidx78_2_indices.push_back(const_int32_44); + ptr_arrayidx78_2_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx78_2 = new GetElementPtrInst(ptr_inputs_136, ptr_arrayidx78_2_indices.begin(), ptr_arrayidx78_2_indices.end(), "arrayidx78.2", label_entry_135); + StoreInst* void_148 = new StoreInst(packed_tmp73_2, ptr_arrayidx78_2, false, label_entry_135); + InsertElementInst* packed_tmp52_3 = new InsertElementInst(packed_tmp73_2, float_tmp50, const_int32_41, "tmp52.3", label_entry_135); + InsertElementInst* packed_tmp59_3 = new InsertElementInst(packed_tmp52_3, float_tmp57, const_int32_43, "tmp59.3", label_entry_135); + InsertElementInst* packed_tmp66_3 = new InsertElementInst(packed_tmp59_3, float_tmp64, const_int32_44, "tmp66.3", label_entry_135); + InsertElementInst* packed_tmp73_3 = new InsertElementInst(packed_tmp66_3, float_tmp71, const_int32_45, "tmp73.3", label_entry_135); + std::vector ptr_arrayidx78_3_indices; + ptr_arrayidx78_3_indices.push_back(const_int32_41); + ptr_arrayidx78_3_indices.push_back(const_int32_45); + ptr_arrayidx78_3_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx78_3 = new GetElementPtrInst(ptr_inputs_136, ptr_arrayidx78_3_indices.begin(), ptr_arrayidx78_3_indices.end(), "arrayidx78.3", label_entry_135); + StoreInst* void_149 = new StoreInst(packed_tmp73_3, ptr_arrayidx78_3, false, label_entry_135); + ICmpInst* int1_cmp_i_150 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_134, const_int32_41, "cmp.i", label_entry_135); + new BranchInst(label_forbody_i, label_from_consts_exit, int1_cmp_i_150, label_entry_135); + + // Block forbody.i (label_forbody_i) + Argument* fwdref_152 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i); + int32_i_0_reg2mem_0_i->reserveOperandSpace(2); + int32_i_0_reg2mem_0_i->addIncoming(const_int32_41, label_entry_135); + int32_i_0_reg2mem_0_i->addIncoming(fwdref_152, label_forbody_i); + + Argument* fwdref_154 = new Argument(VectorTy_15); + PHINode* packed_vec_0_reg2mem_0_i_153 = new PHINode(VectorTy_15, "vec.0.reg2mem.0.i", label_forbody_i); + packed_vec_0_reg2mem_0_i_153->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_153->addIncoming(const_packed_42, label_entry_135); + packed_vec_0_reg2mem_0_i_153->addIncoming(fwdref_154, label_forbody_i); + + std::vector ptr_arraydecay_i_indices; + ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_i); + ptr_arraydecay_i_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_aconsts_133, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_i); + LoadInst* float_tmp5_i_155 = new LoadInst(ptr_arraydecay_i, "tmp5.i", false, label_forbody_i); + InsertElementInst* packed_tmp7_i = new InsertElementInst(packed_vec_0_reg2mem_0_i_153, float_tmp5_i_155, const_int32_41, "tmp7.i", label_forbody_i); + std::vector ptr_arrayidx12_i_156_indices; + ptr_arrayidx12_i_156_indices.push_back(int32_i_0_reg2mem_0_i); + ptr_arrayidx12_i_156_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx12_i_156 = new GetElementPtrInst(ptr_aconsts_133, ptr_arrayidx12_i_156_indices.begin(), ptr_arrayidx12_i_156_indices.end(), "arrayidx12.i", label_forbody_i); + LoadInst* float_tmp13_i_157 = new LoadInst(ptr_arrayidx12_i_156, "tmp13.i", false, label_forbody_i); + InsertElementInst* packed_tmp15_i_158 = new InsertElementInst(packed_tmp7_i, float_tmp13_i_157, const_int32_43, "tmp15.i", label_forbody_i); + std::vector ptr_arrayidx20_i_159_indices; + ptr_arrayidx20_i_159_indices.push_back(int32_i_0_reg2mem_0_i); + ptr_arrayidx20_i_159_indices.push_back(const_int32_44); + Instruction* ptr_arrayidx20_i_159 = new GetElementPtrInst(ptr_aconsts_133, ptr_arrayidx20_i_159_indices.begin(), ptr_arrayidx20_i_159_indices.end(), "arrayidx20.i", label_forbody_i); + LoadInst* float_tmp21_i_160 = new LoadInst(ptr_arrayidx20_i_159, "tmp21.i", false, label_forbody_i); + InsertElementInst* packed_tmp23_i = new InsertElementInst(packed_tmp15_i_158, float_tmp21_i_160, const_int32_44, "tmp23.i", label_forbody_i); + std::vector ptr_arrayidx28_i_161_indices; + ptr_arrayidx28_i_161_indices.push_back(int32_i_0_reg2mem_0_i); + ptr_arrayidx28_i_161_indices.push_back(const_int32_45); + Instruction* ptr_arrayidx28_i_161 = new GetElementPtrInst(ptr_aconsts_133, ptr_arrayidx28_i_161_indices.begin(), ptr_arrayidx28_i_161_indices.end(), "arrayidx28.i", label_forbody_i); + LoadInst* float_tmp29_i_162 = new LoadInst(ptr_arrayidx28_i_161, "tmp29.i", false, label_forbody_i); + InsertElementInst* packed_tmp31_i_163 = new InsertElementInst(packed_tmp23_i, float_tmp29_i_162, const_int32_45, "tmp31.i", label_forbody_i); + std::vector ptr_arrayidx34_i_164_indices; + ptr_arrayidx34_i_164_indices.push_back(const_int32_41); + ptr_arrayidx34_i_164_indices.push_back(int32_i_0_reg2mem_0_i); + Instruction* ptr_arrayidx34_i_164 = new GetElementPtrInst(ptr_consts_137, ptr_arrayidx34_i_164_indices.begin(), ptr_arrayidx34_i_164_indices.end(), "arrayidx34.i", label_forbody_i); + StoreInst* void_165 = new StoreInst(packed_tmp31_i_163, ptr_arrayidx34_i_164, false, label_forbody_i); + BinaryOperator* int32_indvar_next21 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i, const_int32_43, "indvar.next21", label_forbody_i); + ICmpInst* int1_exitcond22 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next21, int32_num_consts_134, "exitcond22", label_forbody_i); + new BranchInst(label_from_consts_exit, label_forbody_i, int1_exitcond22, label_forbody_i); + + // Block from_consts.exit (label_from_consts_exit) + CastInst* double_conv = new FPExtInst(float_x, Type::DoubleTy, "conv", label_from_consts_exit); + CastInst* double_conv86 = new FPExtInst(float_y, Type::DoubleTy, "conv86", label_from_consts_exit); + std::vector int32_call_params; + int32_call_params.push_back(const_ptr_47); + int32_call_params.push_back(double_conv); + int32_call_params.push_back(double_conv86); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_from_consts_exit); + int32_call->setCallingConv(CallingConv::C); + int32_call->setTailCall(false); + std::vector ptr_arraydecay102_indices; + ptr_arraydecay102_indices.push_back(const_int32_41); + ptr_arraydecay102_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay102 = new GetElementPtrInst(ptr_consts_137, ptr_arraydecay102_indices.begin(), ptr_arraydecay102_indices.end(), "arraydecay102", label_from_consts_exit); + std::vector ptr_arraydecay103_indices; + ptr_arraydecay103_indices.push_back(const_int32_41); + ptr_arraydecay103_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay103 = new GetElementPtrInst(ptr_temps_139, ptr_arraydecay103_indices.begin(), ptr_arraydecay103_indices.end(), "arraydecay103", label_from_consts_exit); + new BranchInst(label_forbody91, label_from_consts_exit); + + // Block forbody91 (label_forbody91) + Argument* fwdref_168 = new Argument(IntegerType::get(32)); + PHINode* int32_i87_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i87.0.reg2mem.0", label_forbody91); + int32_i87_0_reg2mem_0->reserveOperandSpace(2); + int32_i87_0_reg2mem_0->addIncoming(const_int32_41, label_from_consts_exit); + int32_i87_0_reg2mem_0->addIncoming(fwdref_168, label_forbody91); + + std::vector ptr_arraydecay95_indices; + ptr_arraydecay95_indices.push_back(const_int32_41); + ptr_arraydecay95_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arraydecay95_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay95 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay95_indices.begin(), ptr_arraydecay95_indices.end(), "arraydecay95", label_forbody91); + std::vector ptr_arraydecay99_indices; + ptr_arraydecay99_indices.push_back(const_int32_41); + ptr_arraydecay99_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arraydecay99_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay99 = new GetElementPtrInst(ptr_results_138, ptr_arraydecay99_indices.begin(), ptr_arraydecay99_indices.end(), "arraydecay99", label_forbody91); + std::vector void_169_params; + void_169_params.push_back(ptr_arraydecay99); + void_169_params.push_back(ptr_arraydecay95); + void_169_params.push_back(ptr_arraydecay102); + void_169_params.push_back(ptr_arraydecay103); + CallInst* void_169 = new CallInst(func_execute_shader, void_169_params.begin(), void_169_params.end(), "", label_forbody91); + void_169->setCallingConv(CallingConv::C); + void_169->setTailCall(false); + std::vector ptr_arraydecay_i2_indices; + ptr_arraydecay_i2_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arraydecay_i2_indices.push_back(const_int32_41); + ptr_arraydecay_i2_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay_i2 = new GetElementPtrInst(ptr_dests_132, ptr_arraydecay_i2_indices.begin(), ptr_arraydecay_i2_indices.end(), "arraydecay.i2", label_forbody91); + LoadInst* packed_tmp7_i3 = new LoadInst(ptr_arraydecay99, "tmp7.i3", false, label_forbody91); + ExtractElementInst* float_tmp11_i = new ExtractElementInst(packed_tmp7_i3, const_int32_41, "tmp11.i", label_forbody91); + StoreInst* void_170 = new StoreInst(float_tmp11_i, ptr_arraydecay_i2, false, label_forbody91); + std::vector ptr_arrayidx13_i_indices; + ptr_arrayidx13_i_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arrayidx13_i_indices.push_back(const_int32_41); + ptr_arrayidx13_i_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx13_i = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx13_i_indices.begin(), ptr_arrayidx13_i_indices.end(), "arrayidx13.i", label_forbody91); + ExtractElementInst* float_tmp15_i4 = new ExtractElementInst(packed_tmp7_i3, const_int32_43, "tmp15.i4", label_forbody91); + StoreInst* void_171 = new StoreInst(float_tmp15_i4, ptr_arrayidx13_i, false, label_forbody91); + std::vector ptr_arrayidx17_i_indices; + ptr_arrayidx17_i_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arrayidx17_i_indices.push_back(const_int32_41); + ptr_arrayidx17_i_indices.push_back(const_int32_44); + Instruction* ptr_arrayidx17_i = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx17_i_indices.begin(), ptr_arrayidx17_i_indices.end(), "arrayidx17.i", label_forbody91); + ExtractElementInst* float_tmp19_i = new ExtractElementInst(packed_tmp7_i3, const_int32_44, "tmp19.i", label_forbody91); + StoreInst* void_172 = new StoreInst(float_tmp19_i, ptr_arrayidx17_i, false, label_forbody91); + std::vector ptr_arrayidx21_i_indices; + ptr_arrayidx21_i_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arrayidx21_i_indices.push_back(const_int32_41); + ptr_arrayidx21_i_indices.push_back(const_int32_45); + Instruction* ptr_arrayidx21_i = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx21_i_indices.begin(), ptr_arrayidx21_i_indices.end(), "arrayidx21.i", label_forbody91); + ExtractElementInst* float_tmp23_i5 = new ExtractElementInst(packed_tmp7_i3, const_int32_45, "tmp23.i5", label_forbody91); + StoreInst* void_173 = new StoreInst(float_tmp23_i5, ptr_arrayidx21_i, false, label_forbody91); + std::vector ptr_arraydecay_i2_1_indices; + ptr_arraydecay_i2_1_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arraydecay_i2_1_indices.push_back(const_int32_43); + ptr_arraydecay_i2_1_indices.push_back(const_int32_41); + Instruction* ptr_arraydecay_i2_1 = new GetElementPtrInst(ptr_dests_132, ptr_arraydecay_i2_1_indices.begin(), ptr_arraydecay_i2_1_indices.end(), "arraydecay.i2.1", label_forbody91); + std::vector ptr_arrayidx6_i_1_indices; + ptr_arrayidx6_i_1_indices.push_back(const_int32_41); + ptr_arrayidx6_i_1_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arrayidx6_i_1_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx6_i_1 = new GetElementPtrInst(ptr_results_138, ptr_arrayidx6_i_1_indices.begin(), ptr_arrayidx6_i_1_indices.end(), "arrayidx6.i.1", label_forbody91); + LoadInst* packed_tmp7_i3_1 = new LoadInst(ptr_arrayidx6_i_1, "tmp7.i3.1", false, label_forbody91); + ExtractElementInst* float_tmp11_i_1 = new ExtractElementInst(packed_tmp7_i3_1, const_int32_41, "tmp11.i.1", label_forbody91); + StoreInst* void_174 = new StoreInst(float_tmp11_i_1, ptr_arraydecay_i2_1, false, label_forbody91); + std::vector ptr_arrayidx13_i_1_indices; + ptr_arrayidx13_i_1_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arrayidx13_i_1_indices.push_back(const_int32_43); + ptr_arrayidx13_i_1_indices.push_back(const_int32_43); + Instruction* ptr_arrayidx13_i_1 = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx13_i_1_indices.begin(), ptr_arrayidx13_i_1_indices.end(), "arrayidx13.i.1", label_forbody91); + ExtractElementInst* float_tmp15_i4_1 = new ExtractElementInst(packed_tmp7_i3_1, const_int32_43, "tmp15.i4.1", label_forbody91); + StoreInst* void_175 = new StoreInst(float_tmp15_i4_1, ptr_arrayidx13_i_1, false, label_forbody91); + std::vector ptr_arrayidx17_i_1_indices; + ptr_arrayidx17_i_1_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arrayidx17_i_1_indices.push_back(const_int32_43); + ptr_arrayidx17_i_1_indices.push_back(const_int32_44); + Instruction* ptr_arrayidx17_i_1 = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx17_i_1_indices.begin(), ptr_arrayidx17_i_1_indices.end(), "arrayidx17.i.1", label_forbody91); + ExtractElementInst* float_tmp19_i_1 = new ExtractElementInst(packed_tmp7_i3_1, const_int32_44, "tmp19.i.1", label_forbody91); + StoreInst* void_176 = new StoreInst(float_tmp19_i_1, ptr_arrayidx17_i_1, false, label_forbody91); + std::vector ptr_arrayidx21_i_1_indices; + ptr_arrayidx21_i_1_indices.push_back(int32_i87_0_reg2mem_0); + ptr_arrayidx21_i_1_indices.push_back(const_int32_43); + ptr_arrayidx21_i_1_indices.push_back(const_int32_45); + Instruction* ptr_arrayidx21_i_1 = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx21_i_1_indices.begin(), ptr_arrayidx21_i_1_indices.end(), "arrayidx21.i.1", label_forbody91); + ExtractElementInst* float_tmp23_i5_1 = new ExtractElementInst(packed_tmp7_i3_1, const_int32_45, "tmp23.i5.1", label_forbody91); + StoreInst* void_177 = new StoreInst(float_tmp23_i5_1, ptr_arrayidx21_i_1, false, label_forbody91); + BinaryOperator* int32_indvar_next19 = BinaryOperator::create(Instruction::Add, int32_i87_0_reg2mem_0, const_int32_43, "indvar.next19", label_forbody91); + ICmpInst* int1_exitcond20 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next19, const_int32_48, "exitcond20", label_forbody91); + new BranchInst(label_afterfor112, label_forbody91, int1_exitcond20, label_forbody91); + + // Block afterfor112 (label_afterfor112) + new ReturnInst(const_int32_49, label_afterfor112); // Resolve Forward References - fwdref_77->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_77; - fwdref_79->replaceAllUsesWith(packed_tmp48_i); delete fwdref_79; - fwdref_78->replaceAllUsesWith(int32_inc_i); delete fwdref_78; - fwdref_76->replaceAllUsesWith(int32_inc59_i); delete fwdref_76; - fwdref_85->replaceAllUsesWith(packed_tmp31_i); delete fwdref_85; - fwdref_84->replaceAllUsesWith(int32_indvar_next24); delete fwdref_84; - fwdref_90->replaceAllUsesWith(int32_inc_us); delete fwdref_90; - fwdref_94->replaceAllUsesWith(int32_indvar_next_99); delete fwdref_94; - fwdref_103->replaceAllUsesWith(int32_inc_105); delete fwdref_103; + fwdref_154->replaceAllUsesWith(packed_tmp31_i_163); delete fwdref_154; + fwdref_152->replaceAllUsesWith(int32_indvar_next21); delete fwdref_152; + fwdref_168->replaceAllUsesWith(int32_indvar_next19); delete fwdref_168; } diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index 99fa64057f..b85490e1dc 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -221,20 +221,42 @@ struct tgsi_interp_coef float dadx[NUM_CHANNELS]; float dady[NUM_CHANNELS]; }; - int run_fragment_shader(float x, float y, float (*dests)[32][4], struct tgsi_interp_coef *coef, - float (*consts)[4], + float (*aconsts)[4], int num_consts, struct tgsi_sampler *samplers, - int num_samplers) + unsigned *sampler_units) { float4 inputs[4][16]; float4 consts[32]; float4 results[4][16]; float4 temps[128];//MAX_PROGRAM_TEMPS + float4 fr1, fr2, fr3, fr4; + fr1.x = x; + fr1.y = y; + fr2.x = x + 1.f; + fr2.y = y; + fr3.x = x; + fr3.y = y + 1.f; + fr4.x = x + 1.f; + fr4.y = y + 1.f; + + inputs[0][0] = fr1; + inputs[1][0] = fr2; + inputs[2][0] = fr3; + inputs[3][0] = fr4; + + for (int i = 0; i < 4; ++i) { + float4 vec; + vec.x = coef->a0[0]; + vec.y = coef->a0[1]; + vec.z = coef->a0[2]; + vec.w = coef->a0[3]; + inputs[i][1] = vec; + } /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", num_vertices, num_inputs, num_attribs, num_consts);*/ //from_array(inputs, ainputs, num_vertices, num_inputs); @@ -243,8 +265,8 @@ int run_fragment_shader(float x, float y, for (int i = 0; i < 4; ++i) { float4 *in = inputs[i]; float4 *res = results[i]; - //execute_shader(res, in, consts, temps); - to_array(dests[i], res, num_attribs); + execute_shader(res, in, consts, temps); + to_array(dests[i], res, 2); } } diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index cd8d337044..23505c46f6 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -40,6 +40,8 @@ #include "x86/rtasm/x86sse.h" +#include "pipe/llvm/gallivm.h" + #include "sp_context.h" #include "sp_state.h" #include "sp_headers.h" @@ -54,6 +56,7 @@ struct quad_shade_stage struct tgsi_exec_machine machine; struct tgsi_exec_vector *inputs, *outputs; int colorOutSlot, depthOutSlot; + struct gallivm_prog *llvm_prog; }; @@ -161,20 +164,20 @@ shade_quad( } } -#if 0 static void shade_quad_llvm(struct quad_stage *qs, struct quad_header *quad) { struct quad_shade_stage *qss = quad_shade_stage(qs); struct softpipe_context *softpipe = qs->softpipe; + float dests[4][32][4]; const float fx = (float) quad->x0; const float fy = (float) quad->y0; struct gallivm_prog *llvm = qss->llvm_prog; quad->mask = gallivm_fragment_shader_exec( - llvm, fx, fy, quad->coef, + llvm, fx, fy, dests, quad->coef, softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], qss->samplers, softpipe->sampler_units); @@ -185,7 +188,7 @@ shade_quad_llvm(struct quad_stage *qs, == TGSI_SEMANTIC_COLOR); memcpy( quad->outputs.color, - &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], + &dests[0][qss->colorOutSlot], sizeof( quad->outputs.color ) ); } @@ -194,14 +197,14 @@ shade_quad_llvm(struct quad_stage *qs, /* output[slot] is new Z */ uint i; for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; + quad->outputs.depth[i] = dests[0][2][i]; } } else { /* copy input Z (which was interpolated by the executor) to output Z */ uint i; for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; + quad->outputs.depth[i] = dests[0][2][i]; } } @@ -210,7 +213,6 @@ shade_quad_llvm(struct quad_stage *qs, qs->next->run( qs->next, quad ); } } -#endif /** * Per-primitive (or per-begin?) setup @@ -227,6 +229,7 @@ static void shade_begin(struct quad_stage *qs) qss->samplers[i].texture = softpipe->texture[i]; } + qss->llvm_prog = softpipe->fs->llvm_prog; /* XXX only do this if the fragment shader changes... */ tgsi_exec_machine_init(&qss->machine, softpipe->fs->shader.tokens, @@ -275,7 +278,11 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) qss->stage.softpipe = softpipe; qss->stage.begin = shade_begin; +#ifdef MESA_LLVM + qss->stage.run = shade_quad_llvm; +#else qss->stage.run = shade_quad; +#endif qss->stage.destroy = shade_destroy; /* set TGSI sampler state that's constant */ -- cgit v1.2.3 From 45003b0bb920fe701304acb1599185f4dc92c89b Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 1 Nov 2007 14:00:25 -0400 Subject: Get fragment shaders working on top of LLVM. Redo the entry points, get the output propagation correctly, interpolate the inputs before feeding into llvm for now. --- src/mesa/pipe/llvm/gallivm.cpp | 175 +++- src/mesa/pipe/llvm/gallivm.h | 7 +- src/mesa/pipe/llvm/llvm_base_shader.cpp | 1486 ++++++++++++++++--------------- src/mesa/pipe/llvm/llvm_entry.c | 42 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 65 +- 5 files changed, 1003 insertions(+), 772 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index 6642f15498..b5e90b037e 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -68,12 +68,21 @@ #include #include +struct gallivm_interpolate { + int attrib; + int chan; + int type; +}; + struct gallivm_prog { llvm::Module *module; void *function; int num_consts; int id; enum gallivm_shader_type type; + + struct gallivm_interpolate interpolators[32*4]; //FIXME: this might not be enough for some shaders + int num_interp; }; struct gallivm_cpu_engine { @@ -141,12 +150,71 @@ static inline void AddStandardCompilePasses(PassManager &PM) { PM.add(createConstantMergePass()); // Merge dup global constants } +static inline void +add_interpolator(struct gallivm_prog *prog, + struct gallivm_interpolate *interp) +{ + prog->interpolators[prog->num_interp] = *interp; + ++prog->num_interp; +} static void -translate_declaration(llvm::Module *module, +translate_declaration(struct gallivm_prog *prog, + llvm::Module *module, Storage *storage, struct tgsi_full_declaration *decl, struct tgsi_full_declaration *fd) { + if (decl->Declaration.File == TGSI_FILE_INPUT) { + unsigned first, last, mask; + uint interp_method; + + assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if (first == 0) { + mask &= ~TGSI_WRITEMASK_XY; + if (mask == TGSI_WRITEMASK_NONE) { + first++; + if (first > last) { + return; + } + } + } + + interp_method = decl->Interpolation.Interpolate; + + if (mask == TGSI_WRITEMASK_XYZW) { + unsigned i, j; + + for (i = first; i <= last; i++) { + for (j = 0; j < NUM_CHANNELS; j++) { + //interp( mach, i, j ); + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } else { + unsigned i, j; + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } + } + } } @@ -686,7 +754,7 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) switch (parse.FullToken.Token.Type) { case TGSI_TOKEN_TYPE_DECLARATION: - translate_declaration(mod, &storage, + translate_declaration(prog, mod, &storage, &parse.FullToken.FullDeclaration, &fd); break; @@ -791,24 +859,83 @@ int gallivm_prog_exec(struct gallivm_prog *prog, } + +static inline void +constant_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan) +{ + unsigned i; + + for (i = 0; i < QUAD_SIZE; ++i) { + inputs[i][attrib][chan] = coefs[attrib].a0[chan]; + } +} + +static inline void +linear_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = inputs[i][0][0]; + const float y = inputs[i][0][1]; + + inputs[i][attrib][chan] = + coefs[attrib].a0[chan] + + coefs[attrib].dadx[chan] * x + + coefs[attrib].dady[chan] * y; + } +} + +static inline void +perspective_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = inputs[i][0][0]; + const float y = inputs[i][0][1]; + /* WPOS.w here is really 1/w */ + const float w = 1.0f / inputs[i][0][3]; + assert(inputs[i][0][3] != 0.0); + + inputs[i][attrib][chan] = + (coefs[attrib].a0[chan] + + coefs[attrib].dadx[chan] * x + + coefs[attrib].dady[chan] * y) * w; + } +} + typedef int (*fragment_shader_runner)(float x, float y, - float (*dests)[32][4], - struct tgsi_interp_coef *coef, - float (*consts)[4], int num_consts, - struct tgsi_sampler *samplers, - unsigned *sampler_units); + float (*dests)[16][4], + float (*inputs)[16][4], + int num_attribs, + float (*consts)[4], int num_consts, + struct tgsi_sampler *samplers, + unsigned *sampler_units); int gallivm_fragment_shader_exec(struct gallivm_prog *prog, - float x, float y, - float (*dests)[32][4], - struct tgsi_interp_coef *coef, + float fx, float fy, + float (*dests)[16][4], + float (*inputs)[16][4], float (*consts)[4], struct tgsi_sampler *samplers, unsigned *sampler_units) { fragment_shader_runner runner = reinterpret_cast(prog->function); assert(runner); - runner(x, y, dests, coef, consts, prog->num_consts, samplers, sampler_units); + + runner(fx, fy, dests, inputs, prog->num_interp, + consts, prog->num_consts, + samplers, sampler_units); return 0; } @@ -928,8 +1055,34 @@ struct gallivm_cpu_engine * gallivm_global_cpu_engine() return CPU; } +void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, + float (*inputs)[16][4], + const struct tgsi_interp_coef *coef) +{ + for (int i = 0; i < prog->num_interp; ++i) { + const gallivm_interpolate &interp = prog->interpolators[i]; + switch (interp.type) { + case TGSI_INTERPOLATE_CONSTANT: + constant_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + case TGSI_INTERPOLATE_LINEAR: + linear_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + perspective_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + default: + assert( 0 ); + } + } +} + #endif /* MESA_LLVM */ + diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h index aaaabf25b1..6a05a55db4 100644 --- a/src/mesa/pipe/llvm/gallivm.h +++ b/src/mesa/pipe/llvm/gallivm.h @@ -64,11 +64,14 @@ int gallivm_prog_exec(struct gallivm_prog *prog, int num_attribs); int gallivm_fragment_shader_exec(struct gallivm_prog *prog, float x, float y, - float (*dests)[32][4], - struct tgsi_interp_coef *coef, + float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], float (*consts)[4], struct tgsi_sampler *samplers, unsigned *sampler_units); +void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + const struct tgsi_interp_coef *coefs); void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index 54c7f15924..f141ea2da0 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -15,46 +15,39 @@ Module* createBaseShader() { OpaqueType* OpaqueTy_struct_softpipe_tile_cache = OpaqueType::get(); mod->addTypeName("struct.softpipe_tile_cache", OpaqueTy_struct_softpipe_tile_cache); - std::vectorStructTy_struct_tgsi_interp_coef_fields; - ArrayType* ArrayTy_0 = ArrayType::get(Type::FloatTy, 4); - - StructTy_struct_tgsi_interp_coef_fields.push_back(ArrayTy_0); - StructTy_struct_tgsi_interp_coef_fields.push_back(ArrayTy_0); - StructTy_struct_tgsi_interp_coef_fields.push_back(ArrayTy_0); - StructType* StructTy_struct_tgsi_interp_coef = StructType::get(StructTy_struct_tgsi_interp_coef_fields, /*isPacked=*/false); - mod->addTypeName("struct.tgsi_interp_coef", StructTy_struct_tgsi_interp_coef); - std::vectorStructTy_struct_tgsi_sampler_fields; - PointerType* PointerTy_1 = PointerType::get(OpaqueTy_struct_pipe_sampler_state); + PointerType* PointerTy_0 = PointerType::get(OpaqueTy_struct_pipe_sampler_state); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_1); - PointerType* PointerTy_2 = PointerType::get(OpaqueTy_struct_pipe_mipmap_tree); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_0); + PointerType* PointerTy_1 = PointerType::get(OpaqueTy_struct_pipe_mipmap_tree); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_2); - std::vectorFuncTy_4_args; + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_1); + std::vectorFuncTy_3_args; PATypeHolder StructTy_struct_tgsi_sampler_fwd = OpaqueType::get(); - PointerType* PointerTy_5 = PointerType::get(StructTy_struct_tgsi_sampler_fwd); + PointerType* PointerTy_4 = PointerType::get(StructTy_struct_tgsi_sampler_fwd); - FuncTy_4_args.push_back(PointerTy_5); - PointerType* PointerTy_6 = PointerType::get(Type::FloatTy); + FuncTy_3_args.push_back(PointerTy_4); + PointerType* PointerTy_5 = PointerType::get(Type::FloatTy); - FuncTy_4_args.push_back(PointerTy_6); - FuncTy_4_args.push_back(PointerTy_6); - FuncTy_4_args.push_back(PointerTy_6); - FuncTy_4_args.push_back(Type::FloatTy); - PointerType* PointerTy_7 = PointerType::get(ArrayTy_0); + FuncTy_3_args.push_back(PointerTy_5); + FuncTy_3_args.push_back(PointerTy_5); + FuncTy_3_args.push_back(PointerTy_5); + FuncTy_3_args.push_back(Type::FloatTy); + ArrayType* ArrayTy_7 = ArrayType::get(Type::FloatTy, 4); - FuncTy_4_args.push_back(PointerTy_7); - ParamAttrsList *FuncTy_4_PAL = 0; - FunctionType* FuncTy_4 = FunctionType::get( + PointerType* PointerTy_6 = PointerType::get(ArrayTy_7); + + FuncTy_3_args.push_back(PointerTy_6); + ParamAttrsList *FuncTy_3_PAL = 0; + FunctionType* FuncTy_3 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_4_args, + /*Params=*/FuncTy_3_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_4_PAL); + /*ParamAttrs=*/FuncTy_3_PAL); - PointerType* PointerTy_3 = PointerType::get(FuncTy_4); + PointerType* PointerTy_2 = PointerType::get(FuncTy_3); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_3); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_2); PointerType* PointerTy_8 = PointerType::get(IntegerType::get(8)); StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_8); @@ -67,37 +60,59 @@ Module* createBaseShader() { StructTy_struct_tgsi_sampler = cast(StructTy_struct_tgsi_sampler_fwd.get()); - ArrayType* ArrayTy_10 = ArrayType::get(IntegerType::get(8), 47); + std::vectorFuncTy_10_args; + VectorType* VectorTy_13 = VectorType::get(Type::FloatTy, 4); - PointerType* PointerTy_11 = PointerType::get(ArrayTy_10); + ArrayType* ArrayTy_12 = ArrayType::get(VectorTy_13, 16); - std::vectorFuncTy_12_args; - VectorType* VectorTy_15 = VectorType::get(Type::FloatTy, 4); + PointerType* PointerTy_11 = PointerType::get(ArrayTy_12); - ArrayType* ArrayTy_14 = ArrayType::get(VectorTy_15, 16); + FuncTy_10_args.push_back(PointerTy_11); + ArrayType* ArrayTy_15 = ArrayType::get(ArrayTy_7, 16); - PointerType* PointerTy_13 = PointerType::get(ArrayTy_14); + PointerType* PointerTy_14 = PointerType::get(ArrayTy_15); - FuncTy_12_args.push_back(PointerTy_13); - ArrayType* ArrayTy_17 = ArrayType::get(ArrayTy_0, 16); + FuncTy_10_args.push_back(PointerTy_14); + FuncTy_10_args.push_back(IntegerType::get(32)); + FuncTy_10_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_10_PAL = 0; + FunctionType* FuncTy_10 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_10_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_10_PAL); - PointerType* PointerTy_16 = PointerType::get(ArrayTy_17); + PointerType* PointerTy_16 = PointerType::get(VectorTy_13); - FuncTy_12_args.push_back(PointerTy_16); - FuncTy_12_args.push_back(IntegerType::get(32)); - FuncTy_12_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_12_PAL = 0; - FunctionType* FuncTy_12 = FunctionType::get( + std::vectorFuncTy_17_args; + FuncTy_17_args.push_back(PointerTy_16); + FuncTy_17_args.push_back(PointerTy_6); + FuncTy_17_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_17_PAL = 0; + FunctionType* FuncTy_17 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_12_args, + /*Params=*/FuncTy_17_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_12_PAL); - - PointerType* PointerTy_18 = PointerType::get(VectorTy_15); + /*ParamAttrs=*/FuncTy_17_PAL); + + std::vectorFuncTy_18_args; + FuncTy_18_args.push_back(PointerTy_6); + FuncTy_18_args.push_back(PointerTy_16); + FuncTy_18_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_18_PAL = 0; + FunctionType* FuncTy_18 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_18_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_18_PAL); std::vectorFuncTy_19_args; - FuncTy_19_args.push_back(PointerTy_18); - FuncTy_19_args.push_back(PointerTy_7); + FuncTy_19_args.push_back(PointerTy_14); + FuncTy_19_args.push_back(PointerTy_14); + FuncTy_19_args.push_back(PointerTy_6); + FuncTy_19_args.push_back(IntegerType::get(32)); + FuncTy_19_args.push_back(IntegerType::get(32)); + FuncTy_19_args.push_back(IntegerType::get(32)); FuncTy_19_args.push_back(IntegerType::get(32)); ParamAttrsList *FuncTy_19_PAL = 0; FunctionType* FuncTy_19 = FunctionType::get( @@ -106,171 +121,107 @@ Module* createBaseShader() { /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_19_PAL); - std::vectorFuncTy_20_args; - FuncTy_20_args.push_back(PointerTy_7); - FuncTy_20_args.push_back(PointerTy_18); - FuncTy_20_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_20_PAL = 0; - FunctionType* FuncTy_20 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_20_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_20_PAL); + ArrayType* ArrayTy_21 = ArrayType::get(ArrayTy_12, 2048); - std::vectorFuncTy_21_args; - FuncTy_21_args.push_back(PointerTy_16); - FuncTy_21_args.push_back(PointerTy_16); - FuncTy_21_args.push_back(PointerTy_7); - FuncTy_21_args.push_back(IntegerType::get(32)); - FuncTy_21_args.push_back(IntegerType::get(32)); - FuncTy_21_args.push_back(IntegerType::get(32)); - FuncTy_21_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_21_PAL = 0; - FunctionType* FuncTy_21 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_21_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_21_PAL); + PointerType* PointerTy_20 = PointerType::get(ArrayTy_21); - ArrayType* ArrayTy_23 = ArrayType::get(ArrayTy_14, 2048); + ArrayType* ArrayTy_23 = ArrayType::get(VectorTy_13, 32); PointerType* PointerTy_22 = PointerType::get(ArrayTy_23); - ArrayType* ArrayTy_25 = ArrayType::get(VectorTy_15, 32); + ArrayType* ArrayTy_25 = ArrayType::get(VectorTy_13, 128); PointerType* PointerTy_24 = PointerType::get(ArrayTy_25); - ArrayType* ArrayTy_27 = ArrayType::get(VectorTy_15, 128); - - PointerType* PointerTy_26 = PointerType::get(ArrayTy_27); - - std::vectorFuncTy_29_args; - FuncTy_29_args.push_back(PointerTy_18); - FuncTy_29_args.push_back(PointerTy_18); - FuncTy_29_args.push_back(PointerTy_18); - FuncTy_29_args.push_back(PointerTy_18); - ParamAttrsList *FuncTy_29_PAL = 0; - FunctionType* FuncTy_29 = FunctionType::get( + std::vectorFuncTy_27_args; + FuncTy_27_args.push_back(PointerTy_16); + FuncTy_27_args.push_back(PointerTy_16); + FuncTy_27_args.push_back(PointerTy_16); + FuncTy_27_args.push_back(PointerTy_16); + ParamAttrsList *FuncTy_27_PAL = 0; + FunctionType* FuncTy_27 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_29_args, + /*Params=*/FuncTy_27_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_29_PAL); - - PointerType* PointerTy_28 = PointerType::get(FuncTy_29); - - std::vectorFuncTy_30_args; - FuncTy_30_args.push_back(Type::FloatTy); - FuncTy_30_args.push_back(Type::FloatTy); - ArrayType* ArrayTy_32 = ArrayType::get(ArrayTy_0, 32); - - PointerType* PointerTy_31 = PointerType::get(ArrayTy_32); - - FuncTy_30_args.push_back(PointerTy_31); - PointerType* PointerTy_33 = PointerType::get(StructTy_struct_tgsi_interp_coef); - - FuncTy_30_args.push_back(PointerTy_33); - FuncTy_30_args.push_back(PointerTy_7); - FuncTy_30_args.push_back(IntegerType::get(32)); - FuncTy_30_args.push_back(PointerTy_5); - PointerType* PointerTy_34 = PointerType::get(IntegerType::get(32)); - - FuncTy_30_args.push_back(PointerTy_34); - ParamAttrsList *FuncTy_30_PAL = 0; - FunctionType* FuncTy_30 = FunctionType::get( + /*ParamAttrs=*/FuncTy_27_PAL); + + PointerType* PointerTy_26 = PointerType::get(FuncTy_27); + + std::vectorFuncTy_28_args; + FuncTy_28_args.push_back(Type::FloatTy); + FuncTy_28_args.push_back(Type::FloatTy); + FuncTy_28_args.push_back(PointerTy_14); + FuncTy_28_args.push_back(PointerTy_14); + FuncTy_28_args.push_back(IntegerType::get(32)); + FuncTy_28_args.push_back(PointerTy_6); + FuncTy_28_args.push_back(IntegerType::get(32)); + FuncTy_28_args.push_back(PointerTy_4); + PointerType* PointerTy_29 = PointerType::get(IntegerType::get(32)); + + FuncTy_28_args.push_back(PointerTy_29); + ParamAttrsList *FuncTy_28_PAL = 0; + FunctionType* FuncTy_28 = FunctionType::get( /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_30_args, + /*Params=*/FuncTy_28_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_30_PAL); - - PointerType* PointerTy_35 = PointerType::get(PointerTy_33); - - ArrayType* ArrayTy_37 = ArrayType::get(ArrayTy_14, 4); + /*ParamAttrs=*/FuncTy_28_PAL); - PointerType* PointerTy_36 = PointerType::get(ArrayTy_37); + ArrayType* ArrayTy_31 = ArrayType::get(ArrayTy_12, 4); - std::vectorFuncTy_39_args; - ParamAttrsList *FuncTy_39_PAL = 0; - FunctionType* FuncTy_39 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_39_args, - /*isVarArg=*/true, - /*ParamAttrs=*/FuncTy_39_PAL); - - PointerType* PointerTy_38 = PointerType::get(FuncTy_39); + PointerType* PointerTy_30 = PointerType::get(ArrayTy_31); // Function Declarations Function* func_from_array = new Function( - /*Type=*/FuncTy_12, + /*Type=*/FuncTy_10, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_array", mod); func_from_array->setCallingConv(CallingConv::C); Function* func_from_consts = new Function( - /*Type=*/FuncTy_19, + /*Type=*/FuncTy_17, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_consts", mod); func_from_consts->setCallingConv(CallingConv::C); Function* func_to_array = new Function( - /*Type=*/FuncTy_20, + /*Type=*/FuncTy_18, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"to_array", mod); func_to_array->setCallingConv(CallingConv::C); Function* func_run_vertex_shader = new Function( - /*Type=*/FuncTy_21, + /*Type=*/FuncTy_19, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_vertex_shader", mod); func_run_vertex_shader->setCallingConv(CallingConv::C); Function* func_execute_shader = new Function( - /*Type=*/FuncTy_29, + /*Type=*/FuncTy_27, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"execute_shader", mod); // (external, no body) func_execute_shader->setCallingConv(CallingConv::C); Function* func_run_fragment_shader = new Function( - /*Type=*/FuncTy_30, + /*Type=*/FuncTy_28, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_fragment_shader", mod); func_run_fragment_shader->setCallingConv(CallingConv::C); - Function* func_printf = new Function( - /*Type=*/FuncTy_39, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", mod); // (external, no body) - func_printf->setCallingConv(CallingConv::C); - // Global Variable Declarations - GlobalVariable* gvar_array__str = new GlobalVariable( - /*Type=*/ArrayTy_10, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/".str", - mod); - // Constant Definitions - Constant* const_array_40 = ConstantArray::get("AAAAAAAAAAAAAAAAAAAAAAA FRAGMENT SHADER %f %f\x0A", true); - Constant* const_int32_41 = Constant::getNullValue(IntegerType::get(32)); - UndefValue* const_packed_42 = UndefValue::get(VectorTy_15); - ConstantInt* const_int32_43 = ConstantInt::get(APInt(32, "1", 10)); - ConstantInt* const_int32_44 = ConstantInt::get(APInt(32, "2", 10)); - ConstantInt* const_int32_45 = ConstantInt::get(APInt(32, "3", 10)); - ConstantFP* const_float_46 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - std::vector const_ptr_47_indices; - const_ptr_47_indices.push_back(const_int32_41); - const_ptr_47_indices.push_back(const_int32_41); - Constant* const_ptr_47 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_47_indices[0], const_ptr_47_indices.size() ); - ConstantInt* const_int32_48 = ConstantInt::get(APInt(32, "4", 10)); - UndefValue* const_int32_49 = UndefValue::get(IntegerType::get(32)); + Constant* const_int32_32 = Constant::getNullValue(IntegerType::get(32)); + UndefValue* const_packed_33 = UndefValue::get(VectorTy_13); + ConstantInt* const_int32_34 = ConstantInt::get(APInt(32, "1", 10)); + ConstantInt* const_int32_35 = ConstantInt::get(APInt(32, "2", 10)); + ConstantInt* const_int32_36 = ConstantInt::get(APInt(32, "3", 10)); + ConstantInt* const_int32_37 = ConstantInt::get(APInt(32, "4", 10)); + ConstantInt* const_int32_38 = ConstantInt::get(APInt(32, "-1", 10)); // Global Variable Definitions - gvar_array__str->setInitializer(const_array_40); // Function Definitions @@ -292,73 +243,73 @@ Module* createBaseShader() { BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); // Block entry (label_entry) - ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_41, "cmp", label_entry); - ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_41, "cmp5", label_entry); + ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_32, "cmp", label_entry); + ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_32, "cmp5", label_entry); BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); new BranchInst(label_forbody6, label_afterfor60, int1_bothcond, label_entry); // Block forbody6 (label_forbody6) - Argument* fwdref_51 = new Argument(IntegerType::get(32)); - Argument* fwdref_52 = new Argument(IntegerType::get(32)); + Argument* fwdref_40 = new Argument(IntegerType::get(32)); + Argument* fwdref_41 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody6); int32_i_0_reg2mem_0->reserveOperandSpace(3); - int32_i_0_reg2mem_0->addIncoming(const_int32_41, label_entry); - int32_i_0_reg2mem_0->addIncoming(fwdref_51, label_forinc57); - int32_i_0_reg2mem_0->addIncoming(fwdref_52, label_forbody6); + int32_i_0_reg2mem_0->addIncoming(const_int32_32, label_entry); + int32_i_0_reg2mem_0->addIncoming(fwdref_40, label_forinc57); + int32_i_0_reg2mem_0->addIncoming(fwdref_41, label_forbody6); - Argument* fwdref_53 = new Argument(IntegerType::get(32)); + Argument* fwdref_42 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0", label_forbody6); int32_j_0_reg2mem_0->reserveOperandSpace(3); - int32_j_0_reg2mem_0->addIncoming(fwdref_53, label_forbody6); - int32_j_0_reg2mem_0->addIncoming(const_int32_41, label_forinc57); - int32_j_0_reg2mem_0->addIncoming(const_int32_41, label_entry); + int32_j_0_reg2mem_0->addIncoming(fwdref_42, label_forbody6); + int32_j_0_reg2mem_0->addIncoming(const_int32_32, label_forinc57); + int32_j_0_reg2mem_0->addIncoming(const_int32_32, label_entry); - Argument* fwdref_54 = new Argument(VectorTy_15); - PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_15, "vec.0.reg2mem.0", label_forbody6); + Argument* fwdref_43 = new Argument(VectorTy_13); + PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_13, "vec.0.reg2mem.0", label_forbody6); packed_vec_0_reg2mem_0->reserveOperandSpace(3); - packed_vec_0_reg2mem_0->addIncoming(fwdref_54, label_forbody6); - packed_vec_0_reg2mem_0->addIncoming(const_packed_42, label_entry); - packed_vec_0_reg2mem_0->addIncoming(fwdref_54, label_forinc57); + packed_vec_0_reg2mem_0->addIncoming(fwdref_43, label_forbody6); + packed_vec_0_reg2mem_0->addIncoming(const_packed_33, label_entry); + packed_vec_0_reg2mem_0->addIncoming(fwdref_43, label_forinc57); std::vector ptr_arraydecay11_indices; ptr_arraydecay11_indices.push_back(int32_i_0_reg2mem_0); ptr_arraydecay11_indices.push_back(int32_j_0_reg2mem_0); - ptr_arraydecay11_indices.push_back(const_int32_41); + ptr_arraydecay11_indices.push_back(const_int32_32); Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); - InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_41, "tmp15", label_forbody6); + InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_32, "tmp15", label_forbody6); std::vector ptr_arrayidx23_indices; ptr_arrayidx23_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx23_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx23_indices.push_back(const_int32_43); + ptr_arrayidx23_indices.push_back(const_int32_34); Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_43, "tmp26", label_forbody6); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_34, "tmp26", label_forbody6); std::vector ptr_arrayidx34_indices; ptr_arrayidx34_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx34_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx34_indices.push_back(const_int32_44); + ptr_arrayidx34_indices.push_back(const_int32_35); Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6); LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6); - InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_44, "tmp37", label_forbody6); + InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_35, "tmp37", label_forbody6); std::vector ptr_arrayidx45_indices; ptr_arrayidx45_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx45_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx45_indices.push_back(const_int32_45); + ptr_arrayidx45_indices.push_back(const_int32_36); Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6); LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6); - InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_45, "tmp48", label_forbody6); + InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_36, "tmp48", label_forbody6); std::vector ptr_arrayidx54_indices; ptr_arrayidx54_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0); Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); - StoreInst* void_55 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); - BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_43, "inc", label_forbody6); + StoreInst* void_44 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); + BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_34, "inc", label_forbody6); ICmpInst* int1_cmp59 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp59", label_forbody6); new BranchInst(label_forbody6, label_forinc57, int1_cmp59, label_forbody6); // Block forinc57 (label_forinc57) - BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_43, "inc59", label_forinc57); + BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_34, "inc59", label_forinc57); ICmpInst* int1_cmp17 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59, int32_count, "cmp17", label_forinc57); new BranchInst(label_forbody6, label_afterfor60, int1_cmp17, label_forinc57); @@ -366,80 +317,80 @@ Module* createBaseShader() { new ReturnInst(label_afterfor60); // Resolve Forward References - fwdref_52->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_52; - fwdref_54->replaceAllUsesWith(packed_tmp48); delete fwdref_54; - fwdref_53->replaceAllUsesWith(int32_inc); delete fwdref_53; - fwdref_51->replaceAllUsesWith(int32_inc59); delete fwdref_51; + fwdref_41->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_41; + fwdref_43->replaceAllUsesWith(packed_tmp48); delete fwdref_43; + fwdref_42->replaceAllUsesWith(int32_inc); delete fwdref_42; + fwdref_40->replaceAllUsesWith(int32_inc59); delete fwdref_40; } // Function: from_consts (func_from_consts) { Function::arg_iterator args = func_from_consts->arg_begin(); - Value* ptr_res_59 = args++; - ptr_res_59->setName("res"); - Value* ptr_ainputs_60 = args++; - ptr_ainputs_60->setName("ainputs"); - Value* int32_count_61 = args++; - int32_count_61->setName("count"); - - BasicBlock* label_entry_62 = new BasicBlock("entry",func_from_consts,0); + Value* ptr_res_48 = args++; + ptr_res_48->setName("res"); + Value* ptr_ainputs_49 = args++; + ptr_ainputs_49->setName("ainputs"); + Value* int32_count_50 = args++; + int32_count_50->setName("count"); + + BasicBlock* label_entry_51 = new BasicBlock("entry",func_from_consts,0); BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); - // Block entry (label_entry_62) - ICmpInst* int1_cmp_63 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_61, const_int32_41, "cmp", label_entry_62); - new BranchInst(label_forbody, label_afterfor, int1_cmp_63, label_entry_62); + // Block entry (label_entry_51) + ICmpInst* int1_cmp_52 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_50, const_int32_32, "cmp", label_entry_51); + new BranchInst(label_forbody, label_afterfor, int1_cmp_52, label_entry_51); // Block forbody (label_forbody) - Argument* fwdref_66 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_65 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); - int32_i_0_reg2mem_0_65->reserveOperandSpace(2); - int32_i_0_reg2mem_0_65->addIncoming(const_int32_41, label_entry_62); - int32_i_0_reg2mem_0_65->addIncoming(fwdref_66, label_forbody); - - Argument* fwdref_68 = new Argument(VectorTy_15); - PHINode* packed_vec_0_reg2mem_0_67 = new PHINode(VectorTy_15, "vec.0.reg2mem.0", label_forbody); - packed_vec_0_reg2mem_0_67->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_67->addIncoming(const_packed_42, label_entry_62); - packed_vec_0_reg2mem_0_67->addIncoming(fwdref_68, label_forbody); + Argument* fwdref_55 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_54 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); + int32_i_0_reg2mem_0_54->reserveOperandSpace(2); + int32_i_0_reg2mem_0_54->addIncoming(const_int32_32, label_entry_51); + int32_i_0_reg2mem_0_54->addIncoming(fwdref_55, label_forbody); + + Argument* fwdref_57 = new Argument(VectorTy_13); + PHINode* packed_vec_0_reg2mem_0_56 = new PHINode(VectorTy_13, "vec.0.reg2mem.0", label_forbody); + packed_vec_0_reg2mem_0_56->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_56->addIncoming(const_packed_33, label_entry_51); + packed_vec_0_reg2mem_0_56->addIncoming(fwdref_57, label_forbody); std::vector ptr_arraydecay_indices; - ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_65); - ptr_arraydecay_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_60, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); + ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_54); + ptr_arraydecay_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_49, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); - InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_67, float_tmp5, const_int32_41, "tmp7", label_forbody); + InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_56, float_tmp5, const_int32_32, "tmp7", label_forbody); std::vector ptr_arrayidx12_indices; - ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_65); - ptr_arrayidx12_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_60, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); - LoadInst* float_tmp13_69 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); - InsertElementInst* packed_tmp15_70 = new InsertElementInst(packed_tmp7, float_tmp13_69, const_int32_43, "tmp15", label_forbody); + ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_54); + ptr_arrayidx12_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_49, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); + LoadInst* float_tmp13_58 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); + InsertElementInst* packed_tmp15_59 = new InsertElementInst(packed_tmp7, float_tmp13_58, const_int32_34, "tmp15", label_forbody); std::vector ptr_arrayidx20_indices; - ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_65); - ptr_arrayidx20_indices.push_back(const_int32_44); - Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_60, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); + ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_54); + ptr_arrayidx20_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_49, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); - InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_70, float_tmp21, const_int32_44, "tmp23", label_forbody); + InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_59, float_tmp21, const_int32_35, "tmp23", label_forbody); std::vector ptr_arrayidx28_indices; - ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_65); - ptr_arrayidx28_indices.push_back(const_int32_45); - Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_60, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); + ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_54); + ptr_arrayidx28_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_49, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); - InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_45, "tmp31", label_forbody); - GetElementPtrInst* ptr_arrayidx34_71 = new GetElementPtrInst(ptr_res_59, int32_i_0_reg2mem_0_65, "arrayidx34", label_forbody); - StoreInst* void_72 = new StoreInst(packed_tmp31, ptr_arrayidx34_71, false, label_forbody); - BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_65, const_int32_43, "indvar.next", label_forbody); - ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_61, "exitcond", label_forbody); + InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_36, "tmp31", label_forbody); + GetElementPtrInst* ptr_arrayidx34_60 = new GetElementPtrInst(ptr_res_48, int32_i_0_reg2mem_0_54, "arrayidx34", label_forbody); + StoreInst* void_61 = new StoreInst(packed_tmp31, ptr_arrayidx34_60, false, label_forbody); + BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_54, const_int32_34, "indvar.next", label_forbody); + ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_50, "exitcond", label_forbody); new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); // Block afterfor (label_afterfor) new ReturnInst(label_afterfor); // Resolve Forward References - fwdref_68->replaceAllUsesWith(packed_tmp31); delete fwdref_68; - fwdref_66->replaceAllUsesWith(int32_indvar_next); delete fwdref_66; + fwdref_57->replaceAllUsesWith(packed_tmp31); delete fwdref_57; + fwdref_55->replaceAllUsesWith(int32_indvar_next); delete fwdref_55; } @@ -450,81 +401,81 @@ Module* createBaseShader() { ptr_dests->setName("dests"); Value* ptr_in = args++; ptr_in->setName("in"); - Value* int32_num_attribs_75 = args++; - int32_num_attribs_75->setName("num_attribs"); - - BasicBlock* label_entry_76 = new BasicBlock("entry",func_to_array,0); - BasicBlock* label_forbody_77 = new BasicBlock("forbody",func_to_array,0); - BasicBlock* label_afterfor_78 = new BasicBlock("afterfor",func_to_array,0); - - // Block entry (label_entry_76) - ICmpInst* int1_cmp_79 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_75, const_int32_41, "cmp", label_entry_76); - new BranchInst(label_forbody_77, label_afterfor_78, int1_cmp_79, label_entry_76); - - // Block forbody (label_forbody_77) - Argument* fwdref_82 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_81 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_77); - int32_i_0_reg2mem_0_81->reserveOperandSpace(2); - int32_i_0_reg2mem_0_81->addIncoming(const_int32_41, label_entry_76); - int32_i_0_reg2mem_0_81->addIncoming(fwdref_82, label_forbody_77); - - std::vector ptr_arraydecay_83_indices; - ptr_arraydecay_83_indices.push_back(int32_i_0_reg2mem_0_81); - ptr_arraydecay_83_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay_83 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_83_indices.begin(), ptr_arraydecay_83_indices.end(), "arraydecay", label_forbody_77); - GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_81, "arrayidx6", label_forbody_77); - LoadInst* packed_tmp7_84 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_77); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_84, const_int32_41, "tmp11", label_forbody_77); - StoreInst* void_85 = new StoreInst(float_tmp11, ptr_arraydecay_83, false, label_forbody_77); + Value* int32_num_attribs_64 = args++; + int32_num_attribs_64->setName("num_attribs"); + + BasicBlock* label_entry_65 = new BasicBlock("entry",func_to_array,0); + BasicBlock* label_forbody_66 = new BasicBlock("forbody",func_to_array,0); + BasicBlock* label_afterfor_67 = new BasicBlock("afterfor",func_to_array,0); + + // Block entry (label_entry_65) + ICmpInst* int1_cmp_68 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_64, const_int32_32, "cmp", label_entry_65); + new BranchInst(label_forbody_66, label_afterfor_67, int1_cmp_68, label_entry_65); + + // Block forbody (label_forbody_66) + Argument* fwdref_71 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_70 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_66); + int32_i_0_reg2mem_0_70->reserveOperandSpace(2); + int32_i_0_reg2mem_0_70->addIncoming(const_int32_32, label_entry_65); + int32_i_0_reg2mem_0_70->addIncoming(fwdref_71, label_forbody_66); + + std::vector ptr_arraydecay_72_indices; + ptr_arraydecay_72_indices.push_back(int32_i_0_reg2mem_0_70); + ptr_arraydecay_72_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay_72 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_72_indices.begin(), ptr_arraydecay_72_indices.end(), "arraydecay", label_forbody_66); + GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_70, "arrayidx6", label_forbody_66); + LoadInst* packed_tmp7_73 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_66); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_73, const_int32_32, "tmp11", label_forbody_66); + StoreInst* void_74 = new StoreInst(float_tmp11, ptr_arraydecay_72, false, label_forbody_66); std::vector ptr_arrayidx13_indices; - ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_81); - ptr_arrayidx13_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_77); - ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_84, const_int32_43, "tmp15", label_forbody_77); - StoreInst* void_86 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_77); + ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_70); + ptr_arrayidx13_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_66); + ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_73, const_int32_34, "tmp15", label_forbody_66); + StoreInst* void_75 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_66); std::vector ptr_arrayidx17_indices; - ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_81); - ptr_arrayidx17_indices.push_back(const_int32_44); - Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_77); - ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_84, const_int32_44, "tmp19", label_forbody_77); - StoreInst* void_87 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_77); + ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_70); + ptr_arrayidx17_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_66); + ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_73, const_int32_35, "tmp19", label_forbody_66); + StoreInst* void_76 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_66); std::vector ptr_arrayidx21_indices; - ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_81); - ptr_arrayidx21_indices.push_back(const_int32_45); - Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_77); - ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_84, const_int32_45, "tmp23", label_forbody_77); - StoreInst* void_88 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_77); - BinaryOperator* int32_indvar_next_89 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_81, const_int32_43, "indvar.next", label_forbody_77); - ICmpInst* int1_exitcond_90 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_89, int32_num_attribs_75, "exitcond", label_forbody_77); - new BranchInst(label_afterfor_78, label_forbody_77, int1_exitcond_90, label_forbody_77); - - // Block afterfor (label_afterfor_78) - new ReturnInst(label_afterfor_78); + ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_70); + ptr_arrayidx21_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_66); + ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_73, const_int32_36, "tmp23", label_forbody_66); + StoreInst* void_77 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_66); + BinaryOperator* int32_indvar_next_78 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_70, const_int32_34, "indvar.next", label_forbody_66); + ICmpInst* int1_exitcond_79 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_78, int32_num_attribs_64, "exitcond", label_forbody_66); + new BranchInst(label_afterfor_67, label_forbody_66, int1_exitcond_79, label_forbody_66); + + // Block afterfor (label_afterfor_67) + new ReturnInst(label_afterfor_67); // Resolve Forward References - fwdref_82->replaceAllUsesWith(int32_indvar_next_89); delete fwdref_82; + fwdref_71->replaceAllUsesWith(int32_indvar_next_78); delete fwdref_71; } // Function: run_vertex_shader (func_run_vertex_shader) { Function::arg_iterator args = func_run_vertex_shader->arg_begin(); - Value* ptr_ainputs_93 = args++; - ptr_ainputs_93->setName("ainputs"); - Value* ptr_dests_94 = args++; - ptr_dests_94->setName("dests"); + Value* ptr_ainputs_82 = args++; + ptr_ainputs_82->setName("ainputs"); + Value* ptr_dests_83 = args++; + ptr_dests_83->setName("dests"); Value* ptr_aconsts = args++; ptr_aconsts->setName("aconsts"); Value* int32_num_vertices = args++; int32_num_vertices->setName("num_vertices"); Value* int32_num_inputs = args++; int32_num_inputs->setName("num_inputs"); - Value* int32_num_attribs_95 = args++; - int32_num_attribs_95->setName("num_attribs"); + Value* int32_num_attribs_84 = args++; + int32_num_attribs_84->setName("num_attribs"); Value* int32_num_consts = args++; int32_num_consts->setName("num_consts"); - BasicBlock* label_entry_96 = new BasicBlock("entry",func_run_vertex_shader,0); + BasicBlock* label_entry_85 = new BasicBlock("entry",func_run_vertex_shader,0); BasicBlock* label_forbody6_i = new BasicBlock("forbody6.i",func_run_vertex_shader,0); BasicBlock* label_forinc57_i = new BasicBlock("forinc57.i",func_run_vertex_shader,0); BasicBlock* label_from_array_exit = new BasicBlock("from_array.exit",func_run_vertex_shader,0); @@ -534,269 +485,269 @@ Module* createBaseShader() { BasicBlock* label_forbody_us = new BasicBlock("forbody.us",func_run_vertex_shader,0); BasicBlock* label_to_array_exit_us = new BasicBlock("to_array.exit.us",func_run_vertex_shader,0); BasicBlock* label_forbody_i_us = new BasicBlock("forbody.i.us",func_run_vertex_shader,0); - BasicBlock* label_forbody_97 = new BasicBlock("forbody",func_run_vertex_shader,0); - BasicBlock* label_afterfor_98 = new BasicBlock("afterfor",func_run_vertex_shader,0); - - // Block entry (label_entry_96) - AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_23, "inputs", label_entry_96); - AllocaInst* ptr_consts = new AllocaInst(ArrayTy_25, "consts", label_entry_96); - AllocaInst* ptr_results = new AllocaInst(ArrayTy_23, "results", label_entry_96); - AllocaInst* ptr_temps = new AllocaInst(ArrayTy_27, "temps", label_entry_96); - ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_41, "cmp.i", label_entry_96); - ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_41, "cmp5.i", label_entry_96); - BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_96); - new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_96); + BasicBlock* label_forbody_86 = new BasicBlock("forbody",func_run_vertex_shader,0); + BasicBlock* label_afterfor_87 = new BasicBlock("afterfor",func_run_vertex_shader,0); + + // Block entry (label_entry_85) + AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_21, "inputs", label_entry_85); + AllocaInst* ptr_consts = new AllocaInst(ArrayTy_23, "consts", label_entry_85); + AllocaInst* ptr_results = new AllocaInst(ArrayTy_21, "results", label_entry_85); + AllocaInst* ptr_temps = new AllocaInst(ArrayTy_25, "temps", label_entry_85); + ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_32, "cmp.i", label_entry_85); + ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_32, "cmp5.i", label_entry_85); + BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_85); + new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_85); // Block forbody6.i (label_forbody6_i) - Argument* fwdref_100 = new Argument(IntegerType::get(32)); - Argument* fwdref_101 = new Argument(IntegerType::get(32)); + Argument* fwdref_89 = new Argument(IntegerType::get(32)); + Argument* fwdref_90 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i_ph = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i.ph", label_forbody6_i); int32_i_0_reg2mem_0_i_ph->reserveOperandSpace(3); - int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_41, label_entry_96); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_100, label_forinc57_i); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_101, label_forbody6_i); + int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_32, label_entry_85); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_89, label_forinc57_i); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_90, label_forbody6_i); - Argument* fwdref_102 = new Argument(IntegerType::get(32)); + Argument* fwdref_91 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i); int32_j_0_reg2mem_0_i->reserveOperandSpace(3); - int32_j_0_reg2mem_0_i->addIncoming(fwdref_102, label_forbody6_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_41, label_forinc57_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_41, label_entry_96); + int32_j_0_reg2mem_0_i->addIncoming(fwdref_91, label_forbody6_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_32, label_forinc57_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_32, label_entry_85); - Argument* fwdref_103 = new Argument(VectorTy_15); - PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_15, "vec.0.reg2mem.0.i", label_forbody6_i); + Argument* fwdref_92 = new Argument(VectorTy_13); + PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i", label_forbody6_i); packed_vec_0_reg2mem_0_i->reserveOperandSpace(3); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_103, label_forbody6_i); - packed_vec_0_reg2mem_0_i->addIncoming(const_packed_42, label_entry_96); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_103, label_forinc57_i); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_92, label_forbody6_i); + packed_vec_0_reg2mem_0_i->addIncoming(const_packed_33, label_entry_85); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_92, label_forinc57_i); std::vector ptr_arraydecay11_i_indices; ptr_arraydecay11_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arraydecay11_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arraydecay11_i_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_93, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); + ptr_arraydecay11_i_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_82, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); LoadInst* float_tmp13_i = new LoadInst(ptr_arraydecay11_i, "tmp13.i", false, label_forbody6_i); - InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_41, "tmp15.i", label_forbody6_i); + InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_32, "tmp15.i", label_forbody6_i); std::vector ptr_arrayidx23_i_indices; ptr_arrayidx23_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx23_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx23_i_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_93, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); + ptr_arrayidx23_i_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_82, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); LoadInst* float_tmp24_i = new LoadInst(ptr_arrayidx23_i, "tmp24.i", false, label_forbody6_i); - InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_43, "tmp26.i", label_forbody6_i); + InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_34, "tmp26.i", label_forbody6_i); std::vector ptr_arrayidx34_i_indices; ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx34_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx34_i_indices.push_back(const_int32_44); - Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_93, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); + ptr_arrayidx34_i_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_82, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); LoadInst* float_tmp35_i = new LoadInst(ptr_arrayidx34_i, "tmp35.i", false, label_forbody6_i); - InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_44, "tmp37.i", label_forbody6_i); + InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_35, "tmp37.i", label_forbody6_i); std::vector ptr_arrayidx45_i_indices; ptr_arrayidx45_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx45_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx45_i_indices.push_back(const_int32_45); - Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_93, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); + ptr_arrayidx45_i_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_82, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); LoadInst* float_tmp46_i = new LoadInst(ptr_arrayidx45_i, "tmp46.i", false, label_forbody6_i); - InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_45, "tmp48.i", label_forbody6_i); + InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_36, "tmp48.i", label_forbody6_i); std::vector ptr_arrayidx54_i_indices; - ptr_arrayidx54_i_indices.push_back(const_int32_41); + ptr_arrayidx54_i_indices.push_back(const_int32_32); ptr_arrayidx54_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx54_i_indices.push_back(int32_j_0_reg2mem_0_i); Instruction* ptr_arrayidx54_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx54_i_indices.begin(), ptr_arrayidx54_i_indices.end(), "arrayidx54.i", label_forbody6_i); - StoreInst* void_104 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); - BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_43, "inc.i", label_forbody6_i); + StoreInst* void_93 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); + BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_34, "inc.i", label_forbody6_i); ICmpInst* int1_cmp59_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_inputs, "cmp59.i", label_forbody6_i); new BranchInst(label_forbody6_i, label_forinc57_i, int1_cmp59_i, label_forbody6_i); // Block forinc57.i (label_forinc57_i) - BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_43, "inc59.i", label_forinc57_i); + BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_34, "inc59.i", label_forinc57_i); ICmpInst* int1_cmp17_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59_i, int32_num_vertices, "cmp17.i", label_forinc57_i); new BranchInst(label_forbody6_i, label_from_array_exit, int1_cmp17_i, label_forinc57_i); // Block from_array.exit (label_from_array_exit) - ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_41, "cmp.i4", label_from_array_exit); + ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_32, "cmp.i4", label_from_array_exit); new BranchInst(label_forbody_i15, label_forcond, int1_cmp_i4, label_from_array_exit); // Block forbody.i15 (label_forbody_i15) - Argument* fwdref_108 = new Argument(IntegerType::get(32)); + Argument* fwdref_97 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i5 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i5", label_forbody_i15); int32_i_0_reg2mem_0_i5->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i5->addIncoming(const_int32_41, label_from_array_exit); - int32_i_0_reg2mem_0_i5->addIncoming(fwdref_108, label_forbody_i15); + int32_i_0_reg2mem_0_i5->addIncoming(const_int32_32, label_from_array_exit); + int32_i_0_reg2mem_0_i5->addIncoming(fwdref_97, label_forbody_i15); - Argument* fwdref_109 = new Argument(VectorTy_15); - PHINode* packed_vec_0_reg2mem_0_i6 = new PHINode(VectorTy_15, "vec.0.reg2mem.0.i6", label_forbody_i15); + Argument* fwdref_98 = new Argument(VectorTy_13); + PHINode* packed_vec_0_reg2mem_0_i6 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i6", label_forbody_i15); packed_vec_0_reg2mem_0_i6->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_42, label_from_array_exit); - packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_109, label_forbody_i15); + packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_33, label_from_array_exit); + packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_98, label_forbody_i15); std::vector ptr_arraydecay_i7_indices; ptr_arraydecay_i7_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arraydecay_i7_indices.push_back(const_int32_41); + ptr_arraydecay_i7_indices.push_back(const_int32_32); Instruction* ptr_arraydecay_i7 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i7_indices.begin(), ptr_arraydecay_i7_indices.end(), "arraydecay.i7", label_forbody_i15); LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i7, "tmp5.i", false, label_forbody_i15); - InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_41, "tmp7.i8", label_forbody_i15); + InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_32, "tmp7.i8", label_forbody_i15); std::vector ptr_arrayidx12_i_indices; ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx12_i_indices.push_back(const_int32_43); + ptr_arrayidx12_i_indices.push_back(const_int32_34); Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i15); LoadInst* float_tmp13_i9 = new LoadInst(ptr_arrayidx12_i, "tmp13.i9", false, label_forbody_i15); - InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_43, "tmp15.i10", label_forbody_i15); + InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_34, "tmp15.i10", label_forbody_i15); std::vector ptr_arrayidx20_i_indices; ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx20_i_indices.push_back(const_int32_44); + ptr_arrayidx20_i_indices.push_back(const_int32_35); Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i15); LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i15); - InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_44, "tmp23.i11", label_forbody_i15); + InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_35, "tmp23.i11", label_forbody_i15); std::vector ptr_arrayidx28_i_indices; ptr_arrayidx28_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx28_i_indices.push_back(const_int32_45); + ptr_arrayidx28_i_indices.push_back(const_int32_36); Instruction* ptr_arrayidx28_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i_indices.begin(), ptr_arrayidx28_i_indices.end(), "arrayidx28.i", label_forbody_i15); LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i15); - InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_45, "tmp31.i", label_forbody_i15); + InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_36, "tmp31.i", label_forbody_i15); std::vector ptr_arrayidx34_i12_indices; - ptr_arrayidx34_i12_indices.push_back(const_int32_41); + ptr_arrayidx34_i12_indices.push_back(const_int32_32); ptr_arrayidx34_i12_indices.push_back(int32_i_0_reg2mem_0_i5); Instruction* ptr_arrayidx34_i12 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i12_indices.begin(), ptr_arrayidx34_i12_indices.end(), "arrayidx34.i12", label_forbody_i15); - StoreInst* void_110 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); - BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_43, "indvar.next24", label_forbody_i15); + StoreInst* void_99 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); + BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_34, "indvar.next24", label_forbody_i15); ICmpInst* int1_exitcond25 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next24, int32_num_consts, "exitcond25", label_forbody_i15); new BranchInst(label_forcond, label_forbody_i15, int1_exitcond25, label_forbody_i15); // Block forcond (label_forcond) - new BranchInst(label_forbody_preheader, label_afterfor_98, int1_cmp_i, label_forcond); + new BranchInst(label_forbody_preheader, label_afterfor_87, int1_cmp_i, label_forcond); // Block forbody.preheader (label_forbody_preheader) std::vector ptr_arraydecay17_indices; - ptr_arraydecay17_indices.push_back(const_int32_41); - ptr_arraydecay17_indices.push_back(const_int32_41); + ptr_arraydecay17_indices.push_back(const_int32_32); + ptr_arraydecay17_indices.push_back(const_int32_32); Instruction* ptr_arraydecay17 = new GetElementPtrInst(ptr_consts, ptr_arraydecay17_indices.begin(), ptr_arraydecay17_indices.end(), "arraydecay17", label_forbody_preheader); std::vector ptr_arraydecay18_indices; - ptr_arraydecay18_indices.push_back(const_int32_41); - ptr_arraydecay18_indices.push_back(const_int32_41); + ptr_arraydecay18_indices.push_back(const_int32_32); + ptr_arraydecay18_indices.push_back(const_int32_32); Instruction* ptr_arraydecay18 = new GetElementPtrInst(ptr_temps, ptr_arraydecay18_indices.begin(), ptr_arraydecay18_indices.end(), "arraydecay18", label_forbody_preheader); - ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_95, const_int32_41, "cmp.i1", label_forbody_preheader); - new BranchInst(label_forbody_us, label_forbody_97, int1_cmp_i1, label_forbody_preheader); + ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_84, const_int32_32, "cmp.i1", label_forbody_preheader); + new BranchInst(label_forbody_us, label_forbody_86, int1_cmp_i1, label_forbody_preheader); // Block forbody.us (label_forbody_us) - Argument* fwdref_114 = new Argument(IntegerType::get(32)); + Argument* fwdref_103 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.us", label_forbody_us); int32_i_0_reg2mem_0_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_us->addIncoming(const_int32_41, label_forbody_preheader); - int32_i_0_reg2mem_0_us->addIncoming(fwdref_114, label_to_array_exit_us); + int32_i_0_reg2mem_0_us->addIncoming(const_int32_32, label_forbody_preheader); + int32_i_0_reg2mem_0_us->addIncoming(fwdref_103, label_to_array_exit_us); std::vector ptr_arraydecay10_us_indices; - ptr_arraydecay10_us_indices.push_back(const_int32_41); + ptr_arraydecay10_us_indices.push_back(const_int32_32); ptr_arraydecay10_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay10_us_indices.push_back(const_int32_41); + ptr_arraydecay10_us_indices.push_back(const_int32_32); Instruction* ptr_arraydecay10_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_us_indices.begin(), ptr_arraydecay10_us_indices.end(), "arraydecay10.us", label_forbody_us); std::vector ptr_arraydecay14_us_indices; - ptr_arraydecay14_us_indices.push_back(const_int32_41); + ptr_arraydecay14_us_indices.push_back(const_int32_32); ptr_arraydecay14_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay14_us_indices.push_back(const_int32_41); + ptr_arraydecay14_us_indices.push_back(const_int32_32); Instruction* ptr_arraydecay14_us = new GetElementPtrInst(ptr_results, ptr_arraydecay14_us_indices.begin(), ptr_arraydecay14_us_indices.end(), "arraydecay14.us", label_forbody_us); - std::vector void_115_params; - void_115_params.push_back(ptr_arraydecay14_us); - void_115_params.push_back(ptr_arraydecay10_us); - void_115_params.push_back(ptr_arraydecay17); - void_115_params.push_back(ptr_arraydecay18); - CallInst* void_115 = new CallInst(func_execute_shader, void_115_params.begin(), void_115_params.end(), "", label_forbody_us); - void_115->setCallingConv(CallingConv::C); - void_115->setTailCall(false); + std::vector void_104_params; + void_104_params.push_back(ptr_arraydecay14_us); + void_104_params.push_back(ptr_arraydecay10_us); + void_104_params.push_back(ptr_arraydecay17); + void_104_params.push_back(ptr_arraydecay18); + CallInst* void_104 = new CallInst(func_execute_shader, void_104_params.begin(), void_104_params.end(), "", label_forbody_us); + void_104->setCallingConv(CallingConv::C); + void_104->setTailCall(false); new BranchInst(label_forbody_i_us, label_forbody_us); // Block to_array.exit.us (label_to_array_exit_us) - BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_43, "inc.us", label_to_array_exit_us); + BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_34, "inc.us", label_to_array_exit_us); ICmpInst* int1_cmp21_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_num_vertices, "cmp21.us", label_to_array_exit_us); - new BranchInst(label_forbody_us, label_afterfor_98, int1_cmp21_us, label_to_array_exit_us); + new BranchInst(label_forbody_us, label_afterfor_87, int1_cmp21_us, label_to_array_exit_us); // Block forbody.i.us (label_forbody_i_us) - Argument* fwdref_118 = new Argument(IntegerType::get(32)); + Argument* fwdref_107 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i2_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i2.us", label_forbody_i_us); int32_i_0_reg2mem_0_i2_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_41, label_forbody_us); - int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_118, label_forbody_i_us); + int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_32, label_forbody_us); + int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_107, label_forbody_i_us); std::vector ptr_arraydecay_i_us_indices; ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arraydecay_i_us_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_94, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); + ptr_arraydecay_i_us_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_83, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); std::vector ptr_arrayidx6_i_us_indices; - ptr_arrayidx6_i_us_indices.push_back(const_int32_41); + ptr_arrayidx6_i_us_indices.push_back(const_int32_32); ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); Instruction* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_results, ptr_arrayidx6_i_us_indices.begin(), ptr_arrayidx6_i_us_indices.end(), "arrayidx6.i.us", label_forbody_i_us); LoadInst* packed_tmp7_i_us = new LoadInst(ptr_arrayidx6_i_us, "tmp7.i.us", false, label_forbody_i_us); - ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_41, "tmp11.i.us", label_forbody_i_us); - StoreInst* void_119 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); + ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_32, "tmp11.i.us", label_forbody_i_us); + StoreInst* void_108 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx13_i_us_indices; ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx13_i_us_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_94, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_43, "tmp15.i3.us", label_forbody_i_us); - StoreInst* void_120 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); + ptr_arrayidx13_i_us_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_83, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_34, "tmp15.i3.us", label_forbody_i_us); + StoreInst* void_109 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx17_i_us_indices; ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx17_i_us_indices.push_back(const_int32_44); - Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_94, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_44, "tmp19.i.us", label_forbody_i_us); - StoreInst* void_121 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); + ptr_arrayidx17_i_us_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_83, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_35, "tmp19.i.us", label_forbody_i_us); + StoreInst* void_110 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx21_i_us_indices; ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx21_i_us_indices.push_back(const_int32_45); - Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_94, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_45, "tmp23.i.us", label_forbody_i_us); - StoreInst* void_122 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); - BinaryOperator* int32_indvar_next_123 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_43, "indvar.next", label_forbody_i_us); - ICmpInst* int1_exitcond_124 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_123, int32_num_attribs_95, "exitcond", label_forbody_i_us); - new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_124, label_forbody_i_us); - - // Block forbody (label_forbody_97) - Argument* fwdref_127 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_126 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_97); - int32_i_0_reg2mem_0_126->reserveOperandSpace(2); - int32_i_0_reg2mem_0_126->addIncoming(const_int32_41, label_forbody_preheader); - int32_i_0_reg2mem_0_126->addIncoming(fwdref_127, label_forbody_97); + ptr_arrayidx21_i_us_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_83, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_36, "tmp23.i.us", label_forbody_i_us); + StoreInst* void_111 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); + BinaryOperator* int32_indvar_next_112 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_34, "indvar.next", label_forbody_i_us); + ICmpInst* int1_exitcond_113 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_112, int32_num_attribs_84, "exitcond", label_forbody_i_us); + new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_113, label_forbody_i_us); + + // Block forbody (label_forbody_86) + Argument* fwdref_116 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_115 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_86); + int32_i_0_reg2mem_0_115->reserveOperandSpace(2); + int32_i_0_reg2mem_0_115->addIncoming(const_int32_32, label_forbody_preheader); + int32_i_0_reg2mem_0_115->addIncoming(fwdref_116, label_forbody_86); std::vector ptr_arraydecay10_indices; - ptr_arraydecay10_indices.push_back(const_int32_41); - ptr_arraydecay10_indices.push_back(int32_i_0_reg2mem_0_126); - ptr_arraydecay10_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_forbody_97); + ptr_arraydecay10_indices.push_back(const_int32_32); + ptr_arraydecay10_indices.push_back(int32_i_0_reg2mem_0_115); + ptr_arraydecay10_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_forbody_86); std::vector ptr_arraydecay14_indices; - ptr_arraydecay14_indices.push_back(const_int32_41); - ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_126); - ptr_arraydecay14_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_results, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_97); - std::vector void_128_params; - void_128_params.push_back(ptr_arraydecay14); - void_128_params.push_back(ptr_arraydecay10); - void_128_params.push_back(ptr_arraydecay17); - void_128_params.push_back(ptr_arraydecay18); - CallInst* void_128 = new CallInst(func_execute_shader, void_128_params.begin(), void_128_params.end(), "", label_forbody_97); - void_128->setCallingConv(CallingConv::C); - void_128->setTailCall(false); - BinaryOperator* int32_inc_129 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_126, const_int32_43, "inc", label_forbody_97); - ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_129, int32_num_vertices, "cmp21", label_forbody_97); - new BranchInst(label_forbody_97, label_afterfor_98, int1_cmp21, label_forbody_97); - - // Block afterfor (label_afterfor_98) - new ReturnInst(label_afterfor_98); + ptr_arraydecay14_indices.push_back(const_int32_32); + ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_115); + ptr_arraydecay14_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_results, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_86); + std::vector void_117_params; + void_117_params.push_back(ptr_arraydecay14); + void_117_params.push_back(ptr_arraydecay10); + void_117_params.push_back(ptr_arraydecay17); + void_117_params.push_back(ptr_arraydecay18); + CallInst* void_117 = new CallInst(func_execute_shader, void_117_params.begin(), void_117_params.end(), "", label_forbody_86); + void_117->setCallingConv(CallingConv::C); + void_117->setTailCall(false); + BinaryOperator* int32_inc_118 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_115, const_int32_34, "inc", label_forbody_86); + ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_118, int32_num_vertices, "cmp21", label_forbody_86); + new BranchInst(label_forbody_86, label_afterfor_87, int1_cmp21, label_forbody_86); + + // Block afterfor (label_afterfor_87) + new ReturnInst(label_afterfor_87); // Resolve Forward References - fwdref_101->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_101; - fwdref_103->replaceAllUsesWith(packed_tmp48_i); delete fwdref_103; - fwdref_102->replaceAllUsesWith(int32_inc_i); delete fwdref_102; - fwdref_100->replaceAllUsesWith(int32_inc59_i); delete fwdref_100; - fwdref_109->replaceAllUsesWith(packed_tmp31_i); delete fwdref_109; - fwdref_108->replaceAllUsesWith(int32_indvar_next24); delete fwdref_108; - fwdref_114->replaceAllUsesWith(int32_inc_us); delete fwdref_114; - fwdref_118->replaceAllUsesWith(int32_indvar_next_123); delete fwdref_118; - fwdref_127->replaceAllUsesWith(int32_inc_129); delete fwdref_127; + fwdref_90->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_90; + fwdref_92->replaceAllUsesWith(packed_tmp48_i); delete fwdref_92; + fwdref_91->replaceAllUsesWith(int32_inc_i); delete fwdref_91; + fwdref_89->replaceAllUsesWith(int32_inc59_i); delete fwdref_89; + fwdref_98->replaceAllUsesWith(packed_tmp31_i); delete fwdref_98; + fwdref_97->replaceAllUsesWith(int32_indvar_next24); delete fwdref_97; + fwdref_103->replaceAllUsesWith(int32_inc_us); delete fwdref_103; + fwdref_107->replaceAllUsesWith(int32_indvar_next_112); delete fwdref_107; + fwdref_116->replaceAllUsesWith(int32_inc_118); delete fwdref_116; } @@ -807,293 +758,412 @@ Module* createBaseShader() { float_x->setName("x"); Value* float_y = args++; float_y->setName("y"); - Value* ptr_dests_132 = args++; - ptr_dests_132->setName("dests"); - Value* ptr_coef = args++; - ptr_coef->setName("coef"); - Value* ptr_aconsts_133 = args++; - ptr_aconsts_133->setName("aconsts"); - Value* int32_num_consts_134 = args++; - int32_num_consts_134->setName("num_consts"); + Value* ptr_dests_121 = args++; + ptr_dests_121->setName("dests"); + Value* ptr_ainputs_122 = args++; + ptr_ainputs_122->setName("ainputs"); + Value* int32_num_inputs_123 = args++; + int32_num_inputs_123->setName("num_inputs"); + Value* ptr_aconsts_124 = args++; + ptr_aconsts_124->setName("aconsts"); + Value* int32_num_consts_125 = args++; + int32_num_consts_125->setName("num_consts"); Value* ptr_samplers = args++; ptr_samplers->setName("samplers"); Value* ptr_sampler_units = args++; ptr_sampler_units->setName("sampler_units"); - BasicBlock* label_entry_135 = new BasicBlock("entry",func_run_fragment_shader,0); - BasicBlock* label_forbody_i = new BasicBlock("forbody.i",func_run_fragment_shader,0); - BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); - BasicBlock* label_forbody91 = new BasicBlock("forbody91",func_run_fragment_shader,0); - BasicBlock* label_afterfor112 = new BasicBlock("afterfor112",func_run_fragment_shader,0); - - // Block entry (label_entry_135) - AllocaInst* ptr_coef_addr = new AllocaInst(PointerTy_33, "coef.addr", label_entry_135); - AllocaInst* ptr_inputs_136 = new AllocaInst(ArrayTy_37, "inputs", label_entry_135); - AllocaInst* ptr_consts_137 = new AllocaInst(ArrayTy_25, "consts", label_entry_135); - AllocaInst* ptr_results_138 = new AllocaInst(ArrayTy_37, "results", label_entry_135); - AllocaInst* ptr_temps_139 = new AllocaInst(ArrayTy_27, "temps", label_entry_135); - StoreInst* void_140 = new StoreInst(ptr_coef, ptr_coef_addr, false, label_entry_135); - InsertElementInst* packed_tmp2 = new InsertElementInst(const_packed_42, float_x, const_int32_41, "tmp2", label_entry_135); - InsertElementInst* packed_tmp5 = new InsertElementInst(packed_tmp2, float_y, const_int32_43, "tmp5", label_entry_135); - BinaryOperator* float_add = BinaryOperator::create(Instruction::Add, float_x, const_float_46, "add", label_entry_135); - InsertElementInst* packed_tmp8 = new InsertElementInst(const_packed_42, float_add, const_int32_41, "tmp8", label_entry_135); - InsertElementInst* packed_tmp11 = new InsertElementInst(packed_tmp8, float_y, const_int32_43, "tmp11", label_entry_135); - BinaryOperator* float_add16 = BinaryOperator::create(Instruction::Add, float_y, const_float_46, "add16", label_entry_135); - InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp2, float_add16, const_int32_43, "tmp18", label_entry_135); - InsertElementInst* packed_tmp26_141 = new InsertElementInst(packed_tmp8, float_add16, const_int32_43, "tmp26", label_entry_135); - std::vector ptr_arraydecay27_indices; - ptr_arraydecay27_indices.push_back(const_int32_41); - ptr_arraydecay27_indices.push_back(const_int32_41); - ptr_arraydecay27_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay27 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay27_indices.begin(), ptr_arraydecay27_indices.end(), "arraydecay27", label_entry_135); - StoreInst* void_142 = new StoreInst(packed_tmp5, ptr_arraydecay27, false, label_entry_135); - std::vector ptr_arraydecay32_indices; - ptr_arraydecay32_indices.push_back(const_int32_41); - ptr_arraydecay32_indices.push_back(const_int32_43); - ptr_arraydecay32_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay32 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay32_indices.begin(), ptr_arraydecay32_indices.end(), "arraydecay32", label_entry_135); - StoreInst* void_143 = new StoreInst(packed_tmp11, ptr_arraydecay32, false, label_entry_135); - std::vector ptr_arraydecay37_indices; - ptr_arraydecay37_indices.push_back(const_int32_41); - ptr_arraydecay37_indices.push_back(const_int32_44); - ptr_arraydecay37_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay37 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay37_indices.begin(), ptr_arraydecay37_indices.end(), "arraydecay37", label_entry_135); - StoreInst* void_144 = new StoreInst(packed_tmp18, ptr_arraydecay37, false, label_entry_135); - std::vector ptr_arraydecay42_indices; - ptr_arraydecay42_indices.push_back(const_int32_41); - ptr_arraydecay42_indices.push_back(const_int32_45); - ptr_arraydecay42_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay42 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay42_indices.begin(), ptr_arraydecay42_indices.end(), "arraydecay42", label_entry_135); - StoreInst* void_145 = new StoreInst(packed_tmp26_141, ptr_arraydecay42, false, label_entry_135); - CastInst* ptr_arraydecay48 = new BitCastInst(ptr_coef_addr, PointerTy_6, "arraydecay48", label_entry_135); - LoadInst* float_tmp50 = new LoadInst(ptr_arraydecay48, "tmp50", false, label_entry_135); - CastInst* ptr_tmp53 = new BitCastInst(ptr_coef_addr, PointerTy_33, "tmp53", label_entry_135); - std::vector ptr_arrayidx56_indices; - ptr_arrayidx56_indices.push_back(const_int32_41); - ptr_arrayidx56_indices.push_back(const_int32_41); - ptr_arrayidx56_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx56 = new GetElementPtrInst(ptr_tmp53, ptr_arrayidx56_indices.begin(), ptr_arrayidx56_indices.end(), "arrayidx56", label_entry_135); - LoadInst* float_tmp57 = new LoadInst(ptr_arrayidx56, "tmp57", false, label_entry_135); - std::vector ptr_arrayidx63_indices; - ptr_arrayidx63_indices.push_back(const_int32_41); - ptr_arrayidx63_indices.push_back(const_int32_41); - ptr_arrayidx63_indices.push_back(const_int32_44); - Instruction* ptr_arrayidx63 = new GetElementPtrInst(ptr_tmp53, ptr_arrayidx63_indices.begin(), ptr_arrayidx63_indices.end(), "arrayidx63", label_entry_135); - LoadInst* float_tmp64 = new LoadInst(ptr_arrayidx63, "tmp64", false, label_entry_135); - std::vector ptr_arrayidx70_indices; - ptr_arrayidx70_indices.push_back(const_int32_41); - ptr_arrayidx70_indices.push_back(const_int32_41); - ptr_arrayidx70_indices.push_back(const_int32_45); - Instruction* ptr_arrayidx70 = new GetElementPtrInst(ptr_tmp53, ptr_arrayidx70_indices.begin(), ptr_arrayidx70_indices.end(), "arrayidx70", label_entry_135); - LoadInst* float_tmp71 = new LoadInst(ptr_arrayidx70, "tmp71", false, label_entry_135); - InsertElementInst* packed_tmp52 = new InsertElementInst(const_packed_42, float_tmp50, const_int32_41, "tmp52", label_entry_135); - InsertElementInst* packed_tmp59 = new InsertElementInst(packed_tmp52, float_tmp57, const_int32_43, "tmp59", label_entry_135); - InsertElementInst* packed_tmp66 = new InsertElementInst(packed_tmp59, float_tmp64, const_int32_44, "tmp66", label_entry_135); - InsertElementInst* packed_tmp73 = new InsertElementInst(packed_tmp66, float_tmp71, const_int32_45, "tmp73", label_entry_135); - std::vector ptr_arrayidx78_indices; - ptr_arrayidx78_indices.push_back(const_int32_41); - ptr_arrayidx78_indices.push_back(const_int32_41); - ptr_arrayidx78_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx78 = new GetElementPtrInst(ptr_inputs_136, ptr_arrayidx78_indices.begin(), ptr_arrayidx78_indices.end(), "arrayidx78", label_entry_135); - StoreInst* void_146 = new StoreInst(packed_tmp73, ptr_arrayidx78, false, label_entry_135); - InsertElementInst* packed_tmp52_1 = new InsertElementInst(packed_tmp73, float_tmp50, const_int32_41, "tmp52.1", label_entry_135); - InsertElementInst* packed_tmp59_1 = new InsertElementInst(packed_tmp52_1, float_tmp57, const_int32_43, "tmp59.1", label_entry_135); - InsertElementInst* packed_tmp66_1 = new InsertElementInst(packed_tmp59_1, float_tmp64, const_int32_44, "tmp66.1", label_entry_135); - InsertElementInst* packed_tmp73_1 = new InsertElementInst(packed_tmp66_1, float_tmp71, const_int32_45, "tmp73.1", label_entry_135); - std::vector ptr_arrayidx78_1_indices; - ptr_arrayidx78_1_indices.push_back(const_int32_41); - ptr_arrayidx78_1_indices.push_back(const_int32_43); - ptr_arrayidx78_1_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx78_1 = new GetElementPtrInst(ptr_inputs_136, ptr_arrayidx78_1_indices.begin(), ptr_arrayidx78_1_indices.end(), "arrayidx78.1", label_entry_135); - StoreInst* void_147 = new StoreInst(packed_tmp73_1, ptr_arrayidx78_1, false, label_entry_135); - InsertElementInst* packed_tmp52_2 = new InsertElementInst(packed_tmp73_1, float_tmp50, const_int32_41, "tmp52.2", label_entry_135); - InsertElementInst* packed_tmp59_2 = new InsertElementInst(packed_tmp52_2, float_tmp57, const_int32_43, "tmp59.2", label_entry_135); - InsertElementInst* packed_tmp66_2 = new InsertElementInst(packed_tmp59_2, float_tmp64, const_int32_44, "tmp66.2", label_entry_135); - InsertElementInst* packed_tmp73_2 = new InsertElementInst(packed_tmp66_2, float_tmp71, const_int32_45, "tmp73.2", label_entry_135); - std::vector ptr_arrayidx78_2_indices; - ptr_arrayidx78_2_indices.push_back(const_int32_41); - ptr_arrayidx78_2_indices.push_back(const_int32_44); - ptr_arrayidx78_2_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx78_2 = new GetElementPtrInst(ptr_inputs_136, ptr_arrayidx78_2_indices.begin(), ptr_arrayidx78_2_indices.end(), "arrayidx78.2", label_entry_135); - StoreInst* void_148 = new StoreInst(packed_tmp73_2, ptr_arrayidx78_2, false, label_entry_135); - InsertElementInst* packed_tmp52_3 = new InsertElementInst(packed_tmp73_2, float_tmp50, const_int32_41, "tmp52.3", label_entry_135); - InsertElementInst* packed_tmp59_3 = new InsertElementInst(packed_tmp52_3, float_tmp57, const_int32_43, "tmp59.3", label_entry_135); - InsertElementInst* packed_tmp66_3 = new InsertElementInst(packed_tmp59_3, float_tmp64, const_int32_44, "tmp66.3", label_entry_135); - InsertElementInst* packed_tmp73_3 = new InsertElementInst(packed_tmp66_3, float_tmp71, const_int32_45, "tmp73.3", label_entry_135); - std::vector ptr_arrayidx78_3_indices; - ptr_arrayidx78_3_indices.push_back(const_int32_41); - ptr_arrayidx78_3_indices.push_back(const_int32_45); - ptr_arrayidx78_3_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx78_3 = new GetElementPtrInst(ptr_inputs_136, ptr_arrayidx78_3_indices.begin(), ptr_arrayidx78_3_indices.end(), "arrayidx78.3", label_entry_135); - StoreInst* void_149 = new StoreInst(packed_tmp73_3, ptr_arrayidx78_3, false, label_entry_135); - ICmpInst* int1_cmp_i_150 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_134, const_int32_41, "cmp.i", label_entry_135); - new BranchInst(label_forbody_i, label_from_consts_exit, int1_cmp_i_150, label_entry_135); - - // Block forbody.i (label_forbody_i) - Argument* fwdref_152 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i); - int32_i_0_reg2mem_0_i->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i->addIncoming(const_int32_41, label_entry_135); - int32_i_0_reg2mem_0_i->addIncoming(fwdref_152, label_forbody_i); - - Argument* fwdref_154 = new Argument(VectorTy_15); - PHINode* packed_vec_0_reg2mem_0_i_153 = new PHINode(VectorTy_15, "vec.0.reg2mem.0.i", label_forbody_i); - packed_vec_0_reg2mem_0_i_153->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_153->addIncoming(const_packed_42, label_entry_135); - packed_vec_0_reg2mem_0_i_153->addIncoming(fwdref_154, label_forbody_i); + BasicBlock* label_entry_126 = new BasicBlock("entry",func_run_fragment_shader,0); + BasicBlock* label_forbody6_i_127 = new BasicBlock("forbody6.i",func_run_fragment_shader,0); + BasicBlock* label_from_array_exit_128 = new BasicBlock("from_array.exit",func_run_fragment_shader,0); + BasicBlock* label_forbody_i13 = new BasicBlock("forbody.i13",func_run_fragment_shader,0); + BasicBlock* label_forbody_preheader_129 = new BasicBlock("forbody.preheader",func_run_fragment_shader,0); + BasicBlock* label_forbody_130 = new BasicBlock("forbody",func_run_fragment_shader,0); + BasicBlock* label_afterfor_131 = new BasicBlock("afterfor",func_run_fragment_shader,0); + BasicBlock* label_forbody6_i_1 = new BasicBlock("forbody6.i.1",func_run_fragment_shader,0); + BasicBlock* label_forbody6_i_2 = new BasicBlock("forbody6.i.2",func_run_fragment_shader,0); + BasicBlock* label_forbody6_i_3 = new BasicBlock("forbody6.i.3",func_run_fragment_shader,0); + + // Block entry (label_entry_126) + AllocaInst* ptr_inputs_132 = new AllocaInst(ArrayTy_31, "inputs", label_entry_126); + AllocaInst* ptr_consts_133 = new AllocaInst(ArrayTy_23, "consts", label_entry_126); + AllocaInst* ptr_results_134 = new AllocaInst(ArrayTy_31, "results", label_entry_126); + AllocaInst* ptr_temps_135 = new AllocaInst(ArrayTy_25, "temps", label_entry_126); + ICmpInst* int1_cmp5_i_136 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs_123, const_int32_32, "cmp5.i", label_entry_126); + new BranchInst(label_forbody6_i_127, label_from_array_exit_128, int1_cmp5_i_136, label_entry_126); + + // Block forbody6.i (label_forbody6_i_127) + Argument* fwdref_139 = new Argument(IntegerType::get(32)); + PHINode* int32_j_0_reg2mem_0_i_138 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i_127); + int32_j_0_reg2mem_0_i_138->reserveOperandSpace(2); + int32_j_0_reg2mem_0_i_138->addIncoming(const_int32_32, label_entry_126); + int32_j_0_reg2mem_0_i_138->addIncoming(fwdref_139, label_forbody6_i_127); + + Argument* fwdref_141 = new Argument(VectorTy_13); + PHINode* packed_vec_0_reg2mem_0_i_140 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i", label_forbody6_i_127); + packed_vec_0_reg2mem_0_i_140->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_140->addIncoming(const_packed_33, label_entry_126); + packed_vec_0_reg2mem_0_i_140->addIncoming(fwdref_141, label_forbody6_i_127); + + std::vector ptr_arraydecay11_i_142_indices; + ptr_arraydecay11_i_142_indices.push_back(const_int32_32); + ptr_arraydecay11_i_142_indices.push_back(int32_j_0_reg2mem_0_i_138); + ptr_arraydecay11_i_142_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay11_i_142 = new GetElementPtrInst(ptr_ainputs_122, ptr_arraydecay11_i_142_indices.begin(), ptr_arraydecay11_i_142_indices.end(), "arraydecay11.i", label_forbody6_i_127); + LoadInst* float_tmp13_i_143 = new LoadInst(ptr_arraydecay11_i_142, "tmp13.i", false, label_forbody6_i_127); + InsertElementInst* packed_tmp15_i_144 = new InsertElementInst(packed_vec_0_reg2mem_0_i_140, float_tmp13_i_143, const_int32_32, "tmp15.i", label_forbody6_i_127); + std::vector ptr_arrayidx23_i_145_indices; + ptr_arrayidx23_i_145_indices.push_back(const_int32_32); + ptr_arrayidx23_i_145_indices.push_back(int32_j_0_reg2mem_0_i_138); + ptr_arrayidx23_i_145_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx23_i_145 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx23_i_145_indices.begin(), ptr_arrayidx23_i_145_indices.end(), "arrayidx23.i", label_forbody6_i_127); + LoadInst* float_tmp24_i_146 = new LoadInst(ptr_arrayidx23_i_145, "tmp24.i", false, label_forbody6_i_127); + InsertElementInst* packed_tmp26_i_147 = new InsertElementInst(packed_tmp15_i_144, float_tmp24_i_146, const_int32_34, "tmp26.i", label_forbody6_i_127); + std::vector ptr_arrayidx34_i_148_indices; + ptr_arrayidx34_i_148_indices.push_back(const_int32_32); + ptr_arrayidx34_i_148_indices.push_back(int32_j_0_reg2mem_0_i_138); + ptr_arrayidx34_i_148_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx34_i_148 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx34_i_148_indices.begin(), ptr_arrayidx34_i_148_indices.end(), "arrayidx34.i", label_forbody6_i_127); + LoadInst* float_tmp35_i_149 = new LoadInst(ptr_arrayidx34_i_148, "tmp35.i", false, label_forbody6_i_127); + InsertElementInst* packed_tmp37_i_150 = new InsertElementInst(packed_tmp26_i_147, float_tmp35_i_149, const_int32_35, "tmp37.i", label_forbody6_i_127); + std::vector ptr_arrayidx45_i_151_indices; + ptr_arrayidx45_i_151_indices.push_back(const_int32_32); + ptr_arrayidx45_i_151_indices.push_back(int32_j_0_reg2mem_0_i_138); + ptr_arrayidx45_i_151_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx45_i_151 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx45_i_151_indices.begin(), ptr_arrayidx45_i_151_indices.end(), "arrayidx45.i", label_forbody6_i_127); + LoadInst* float_tmp46_i_152 = new LoadInst(ptr_arrayidx45_i_151, "tmp46.i", false, label_forbody6_i_127); + InsertElementInst* packed_tmp48_i_153 = new InsertElementInst(packed_tmp37_i_150, float_tmp46_i_152, const_int32_36, "tmp48.i", label_forbody6_i_127); + std::vector ptr_arrayidx54_i_154_indices; + ptr_arrayidx54_i_154_indices.push_back(const_int32_32); + ptr_arrayidx54_i_154_indices.push_back(const_int32_32); + ptr_arrayidx54_i_154_indices.push_back(int32_j_0_reg2mem_0_i_138); + Instruction* ptr_arrayidx54_i_154 = new GetElementPtrInst(ptr_inputs_132, ptr_arrayidx54_i_154_indices.begin(), ptr_arrayidx54_i_154_indices.end(), "arrayidx54.i", label_forbody6_i_127); + StoreInst* void_155 = new StoreInst(packed_tmp48_i_153, ptr_arrayidx54_i_154, false, label_forbody6_i_127); + BinaryOperator* int32_inc_i_156 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_138, const_int32_34, "inc.i", label_forbody6_i_127); + ICmpInst* int1_cmp59_i_157 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_156, int32_num_inputs_123, "cmp59.i", label_forbody6_i_127); + new BranchInst(label_forbody6_i_127, label_forbody6_i_1, int1_cmp59_i_157, label_forbody6_i_127); + // Block from_array.exit (label_from_array_exit_128) + ICmpInst* int1_cmp_i_159 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_125, const_int32_32, "cmp.i", label_from_array_exit_128); + new BranchInst(label_forbody_i13, label_forbody_preheader_129, int1_cmp_i_159, label_from_array_exit_128); + + // Block forbody.i13 (label_forbody_i13) + Argument* fwdref_161 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_i3 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i3", label_forbody_i13); + int32_i_0_reg2mem_0_i3->reserveOperandSpace(2); + int32_i_0_reg2mem_0_i3->addIncoming(const_int32_32, label_from_array_exit_128); + int32_i_0_reg2mem_0_i3->addIncoming(fwdref_161, label_forbody_i13); + + Argument* fwdref_162 = new Argument(VectorTy_13); + PHINode* packed_vec_0_reg2mem_0_i4 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i4", label_forbody_i13); + packed_vec_0_reg2mem_0_i4->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i4->addIncoming(const_packed_33, label_from_array_exit_128); + packed_vec_0_reg2mem_0_i4->addIncoming(fwdref_162, label_forbody_i13); + + std::vector ptr_arraydecay_i5_indices; + ptr_arraydecay_i5_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arraydecay_i5_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay_i5 = new GetElementPtrInst(ptr_aconsts_124, ptr_arraydecay_i5_indices.begin(), ptr_arraydecay_i5_indices.end(), "arraydecay.i5", label_forbody_i13); + LoadInst* float_tmp5_i_163 = new LoadInst(ptr_arraydecay_i5, "tmp5.i", false, label_forbody_i13); + InsertElementInst* packed_tmp7_i6 = new InsertElementInst(packed_vec_0_reg2mem_0_i4, float_tmp5_i_163, const_int32_32, "tmp7.i6", label_forbody_i13); + std::vector ptr_arrayidx12_i_164_indices; + ptr_arrayidx12_i_164_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arrayidx12_i_164_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx12_i_164 = new GetElementPtrInst(ptr_aconsts_124, ptr_arrayidx12_i_164_indices.begin(), ptr_arrayidx12_i_164_indices.end(), "arrayidx12.i", label_forbody_i13); + LoadInst* float_tmp13_i7 = new LoadInst(ptr_arrayidx12_i_164, "tmp13.i7", false, label_forbody_i13); + InsertElementInst* packed_tmp15_i8 = new InsertElementInst(packed_tmp7_i6, float_tmp13_i7, const_int32_34, "tmp15.i8", label_forbody_i13); + std::vector ptr_arrayidx20_i_165_indices; + ptr_arrayidx20_i_165_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arrayidx20_i_165_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx20_i_165 = new GetElementPtrInst(ptr_aconsts_124, ptr_arrayidx20_i_165_indices.begin(), ptr_arrayidx20_i_165_indices.end(), "arrayidx20.i", label_forbody_i13); + LoadInst* float_tmp21_i_166 = new LoadInst(ptr_arrayidx20_i_165, "tmp21.i", false, label_forbody_i13); + InsertElementInst* packed_tmp23_i9 = new InsertElementInst(packed_tmp15_i8, float_tmp21_i_166, const_int32_35, "tmp23.i9", label_forbody_i13); + std::vector ptr_arrayidx28_i_167_indices; + ptr_arrayidx28_i_167_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arrayidx28_i_167_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx28_i_167 = new GetElementPtrInst(ptr_aconsts_124, ptr_arrayidx28_i_167_indices.begin(), ptr_arrayidx28_i_167_indices.end(), "arrayidx28.i", label_forbody_i13); + LoadInst* float_tmp29_i_168 = new LoadInst(ptr_arrayidx28_i_167, "tmp29.i", false, label_forbody_i13); + InsertElementInst* packed_tmp31_i_169 = new InsertElementInst(packed_tmp23_i9, float_tmp29_i_168, const_int32_36, "tmp31.i", label_forbody_i13); + std::vector ptr_arrayidx34_i10_indices; + ptr_arrayidx34_i10_indices.push_back(const_int32_32); + ptr_arrayidx34_i10_indices.push_back(int32_i_0_reg2mem_0_i3); + Instruction* ptr_arrayidx34_i10 = new GetElementPtrInst(ptr_consts_133, ptr_arrayidx34_i10_indices.begin(), ptr_arrayidx34_i10_indices.end(), "arrayidx34.i10", label_forbody_i13); + StoreInst* void_170 = new StoreInst(packed_tmp31_i_169, ptr_arrayidx34_i10, false, label_forbody_i13); + BinaryOperator* int32_indvar_next22 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i3, const_int32_34, "indvar.next22", label_forbody_i13); + ICmpInst* int1_exitcond23 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next22, int32_num_consts_125, "exitcond23", label_forbody_i13); + new BranchInst(label_forbody_preheader_129, label_forbody_i13, int1_exitcond23, label_forbody_i13); + + // Block forbody.preheader (label_forbody_preheader_129) + std::vector ptr_arraydecay15_indices; + ptr_arraydecay15_indices.push_back(const_int32_32); + ptr_arraydecay15_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay15 = new GetElementPtrInst(ptr_consts_133, ptr_arraydecay15_indices.begin(), ptr_arraydecay15_indices.end(), "arraydecay15", label_forbody_preheader_129); + std::vector ptr_arraydecay16_indices; + ptr_arraydecay16_indices.push_back(const_int32_32); + ptr_arraydecay16_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay16 = new GetElementPtrInst(ptr_temps_135, ptr_arraydecay16_indices.begin(), ptr_arraydecay16_indices.end(), "arraydecay16", label_forbody_preheader_129); + new BranchInst(label_forbody_130, label_forbody_preheader_129); + + // Block forbody (label_forbody_130) + Argument* fwdref_174 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_173 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_130); + int32_i_0_reg2mem_0_173->reserveOperandSpace(2); + int32_i_0_reg2mem_0_173->addIncoming(const_int32_32, label_forbody_preheader_129); + int32_i_0_reg2mem_0_173->addIncoming(fwdref_174, label_forbody_130); + + std::vector ptr_arraydecay8_indices; + ptr_arraydecay8_indices.push_back(const_int32_32); + ptr_arraydecay8_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arraydecay8_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay8 = new GetElementPtrInst(ptr_inputs_132, ptr_arraydecay8_indices.begin(), ptr_arraydecay8_indices.end(), "arraydecay8", label_forbody_130); + std::vector ptr_arraydecay12_indices; + ptr_arraydecay12_indices.push_back(const_int32_32); + ptr_arraydecay12_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arraydecay12_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay12 = new GetElementPtrInst(ptr_results_134, ptr_arraydecay12_indices.begin(), ptr_arraydecay12_indices.end(), "arraydecay12", label_forbody_130); + std::vector void_175_params; + void_175_params.push_back(ptr_arraydecay12); + void_175_params.push_back(ptr_arraydecay8); + void_175_params.push_back(ptr_arraydecay15); + void_175_params.push_back(ptr_arraydecay16); + CallInst* void_175 = new CallInst(func_execute_shader, void_175_params.begin(), void_175_params.end(), "", label_forbody_130); + void_175->setCallingConv(CallingConv::C); + void_175->setTailCall(false); std::vector ptr_arraydecay_i_indices; - ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arraydecay_i_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_aconsts_133, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_i); - LoadInst* float_tmp5_i_155 = new LoadInst(ptr_arraydecay_i, "tmp5.i", false, label_forbody_i); - InsertElementInst* packed_tmp7_i = new InsertElementInst(packed_vec_0_reg2mem_0_i_153, float_tmp5_i_155, const_int32_41, "tmp7.i", label_forbody_i); - std::vector ptr_arrayidx12_i_156_indices; - ptr_arrayidx12_i_156_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arrayidx12_i_156_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx12_i_156 = new GetElementPtrInst(ptr_aconsts_133, ptr_arrayidx12_i_156_indices.begin(), ptr_arrayidx12_i_156_indices.end(), "arrayidx12.i", label_forbody_i); - LoadInst* float_tmp13_i_157 = new LoadInst(ptr_arrayidx12_i_156, "tmp13.i", false, label_forbody_i); - InsertElementInst* packed_tmp15_i_158 = new InsertElementInst(packed_tmp7_i, float_tmp13_i_157, const_int32_43, "tmp15.i", label_forbody_i); - std::vector ptr_arrayidx20_i_159_indices; - ptr_arrayidx20_i_159_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arrayidx20_i_159_indices.push_back(const_int32_44); - Instruction* ptr_arrayidx20_i_159 = new GetElementPtrInst(ptr_aconsts_133, ptr_arrayidx20_i_159_indices.begin(), ptr_arrayidx20_i_159_indices.end(), "arrayidx20.i", label_forbody_i); - LoadInst* float_tmp21_i_160 = new LoadInst(ptr_arrayidx20_i_159, "tmp21.i", false, label_forbody_i); - InsertElementInst* packed_tmp23_i = new InsertElementInst(packed_tmp15_i_158, float_tmp21_i_160, const_int32_44, "tmp23.i", label_forbody_i); - std::vector ptr_arrayidx28_i_161_indices; - ptr_arrayidx28_i_161_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arrayidx28_i_161_indices.push_back(const_int32_45); - Instruction* ptr_arrayidx28_i_161 = new GetElementPtrInst(ptr_aconsts_133, ptr_arrayidx28_i_161_indices.begin(), ptr_arrayidx28_i_161_indices.end(), "arrayidx28.i", label_forbody_i); - LoadInst* float_tmp29_i_162 = new LoadInst(ptr_arrayidx28_i_161, "tmp29.i", false, label_forbody_i); - InsertElementInst* packed_tmp31_i_163 = new InsertElementInst(packed_tmp23_i, float_tmp29_i_162, const_int32_45, "tmp31.i", label_forbody_i); - std::vector ptr_arrayidx34_i_164_indices; - ptr_arrayidx34_i_164_indices.push_back(const_int32_41); - ptr_arrayidx34_i_164_indices.push_back(int32_i_0_reg2mem_0_i); - Instruction* ptr_arrayidx34_i_164 = new GetElementPtrInst(ptr_consts_137, ptr_arrayidx34_i_164_indices.begin(), ptr_arrayidx34_i_164_indices.end(), "arrayidx34.i", label_forbody_i); - StoreInst* void_165 = new StoreInst(packed_tmp31_i_163, ptr_arrayidx34_i_164, false, label_forbody_i); - BinaryOperator* int32_indvar_next21 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i, const_int32_43, "indvar.next21", label_forbody_i); - ICmpInst* int1_exitcond22 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next21, int32_num_consts_134, "exitcond22", label_forbody_i); - new BranchInst(label_from_consts_exit, label_forbody_i, int1_exitcond22, label_forbody_i); - - // Block from_consts.exit (label_from_consts_exit) - CastInst* double_conv = new FPExtInst(float_x, Type::DoubleTy, "conv", label_from_consts_exit); - CastInst* double_conv86 = new FPExtInst(float_y, Type::DoubleTy, "conv86", label_from_consts_exit); - std::vector int32_call_params; - int32_call_params.push_back(const_ptr_47); - int32_call_params.push_back(double_conv); - int32_call_params.push_back(double_conv86); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_from_consts_exit); - int32_call->setCallingConv(CallingConv::C); - int32_call->setTailCall(false); - std::vector ptr_arraydecay102_indices; - ptr_arraydecay102_indices.push_back(const_int32_41); - ptr_arraydecay102_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay102 = new GetElementPtrInst(ptr_consts_137, ptr_arraydecay102_indices.begin(), ptr_arraydecay102_indices.end(), "arraydecay102", label_from_consts_exit); - std::vector ptr_arraydecay103_indices; - ptr_arraydecay103_indices.push_back(const_int32_41); - ptr_arraydecay103_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay103 = new GetElementPtrInst(ptr_temps_139, ptr_arraydecay103_indices.begin(), ptr_arraydecay103_indices.end(), "arraydecay103", label_from_consts_exit); - new BranchInst(label_forbody91, label_from_consts_exit); - - // Block forbody91 (label_forbody91) - Argument* fwdref_168 = new Argument(IntegerType::get(32)); - PHINode* int32_i87_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i87.0.reg2mem.0", label_forbody91); - int32_i87_0_reg2mem_0->reserveOperandSpace(2); - int32_i87_0_reg2mem_0->addIncoming(const_int32_41, label_from_consts_exit); - int32_i87_0_reg2mem_0->addIncoming(fwdref_168, label_forbody91); - - std::vector ptr_arraydecay95_indices; - ptr_arraydecay95_indices.push_back(const_int32_41); - ptr_arraydecay95_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arraydecay95_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay95 = new GetElementPtrInst(ptr_inputs_136, ptr_arraydecay95_indices.begin(), ptr_arraydecay95_indices.end(), "arraydecay95", label_forbody91); - std::vector ptr_arraydecay99_indices; - ptr_arraydecay99_indices.push_back(const_int32_41); - ptr_arraydecay99_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arraydecay99_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay99 = new GetElementPtrInst(ptr_results_138, ptr_arraydecay99_indices.begin(), ptr_arraydecay99_indices.end(), "arraydecay99", label_forbody91); - std::vector void_169_params; - void_169_params.push_back(ptr_arraydecay99); - void_169_params.push_back(ptr_arraydecay95); - void_169_params.push_back(ptr_arraydecay102); - void_169_params.push_back(ptr_arraydecay103); - CallInst* void_169 = new CallInst(func_execute_shader, void_169_params.begin(), void_169_params.end(), "", label_forbody91); - void_169->setCallingConv(CallingConv::C); - void_169->setTailCall(false); - std::vector ptr_arraydecay_i2_indices; - ptr_arraydecay_i2_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arraydecay_i2_indices.push_back(const_int32_41); - ptr_arraydecay_i2_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay_i2 = new GetElementPtrInst(ptr_dests_132, ptr_arraydecay_i2_indices.begin(), ptr_arraydecay_i2_indices.end(), "arraydecay.i2", label_forbody91); - LoadInst* packed_tmp7_i3 = new LoadInst(ptr_arraydecay99, "tmp7.i3", false, label_forbody91); - ExtractElementInst* float_tmp11_i = new ExtractElementInst(packed_tmp7_i3, const_int32_41, "tmp11.i", label_forbody91); - StoreInst* void_170 = new StoreInst(float_tmp11_i, ptr_arraydecay_i2, false, label_forbody91); + ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arraydecay_i_indices.push_back(const_int32_32); + ptr_arraydecay_i_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_dests_121, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_130); + LoadInst* packed_tmp7_i = new LoadInst(ptr_arraydecay12, "tmp7.i", false, label_forbody_130); + ExtractElementInst* float_tmp11_i = new ExtractElementInst(packed_tmp7_i, const_int32_32, "tmp11.i", label_forbody_130); + StoreInst* void_176 = new StoreInst(float_tmp11_i, ptr_arraydecay_i, false, label_forbody_130); std::vector ptr_arrayidx13_i_indices; - ptr_arrayidx13_i_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arrayidx13_i_indices.push_back(const_int32_41); - ptr_arrayidx13_i_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx13_i = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx13_i_indices.begin(), ptr_arrayidx13_i_indices.end(), "arrayidx13.i", label_forbody91); - ExtractElementInst* float_tmp15_i4 = new ExtractElementInst(packed_tmp7_i3, const_int32_43, "tmp15.i4", label_forbody91); - StoreInst* void_171 = new StoreInst(float_tmp15_i4, ptr_arrayidx13_i, false, label_forbody91); + ptr_arrayidx13_i_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arrayidx13_i_indices.push_back(const_int32_32); + ptr_arrayidx13_i_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx13_i = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx13_i_indices.begin(), ptr_arrayidx13_i_indices.end(), "arrayidx13.i", label_forbody_130); + ExtractElementInst* float_tmp15_i2 = new ExtractElementInst(packed_tmp7_i, const_int32_34, "tmp15.i2", label_forbody_130); + StoreInst* void_177 = new StoreInst(float_tmp15_i2, ptr_arrayidx13_i, false, label_forbody_130); std::vector ptr_arrayidx17_i_indices; - ptr_arrayidx17_i_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arrayidx17_i_indices.push_back(const_int32_41); - ptr_arrayidx17_i_indices.push_back(const_int32_44); - Instruction* ptr_arrayidx17_i = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx17_i_indices.begin(), ptr_arrayidx17_i_indices.end(), "arrayidx17.i", label_forbody91); - ExtractElementInst* float_tmp19_i = new ExtractElementInst(packed_tmp7_i3, const_int32_44, "tmp19.i", label_forbody91); - StoreInst* void_172 = new StoreInst(float_tmp19_i, ptr_arrayidx17_i, false, label_forbody91); + ptr_arrayidx17_i_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arrayidx17_i_indices.push_back(const_int32_32); + ptr_arrayidx17_i_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx17_i = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx17_i_indices.begin(), ptr_arrayidx17_i_indices.end(), "arrayidx17.i", label_forbody_130); + ExtractElementInst* float_tmp19_i = new ExtractElementInst(packed_tmp7_i, const_int32_35, "tmp19.i", label_forbody_130); + StoreInst* void_178 = new StoreInst(float_tmp19_i, ptr_arrayidx17_i, false, label_forbody_130); std::vector ptr_arrayidx21_i_indices; - ptr_arrayidx21_i_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arrayidx21_i_indices.push_back(const_int32_41); - ptr_arrayidx21_i_indices.push_back(const_int32_45); - Instruction* ptr_arrayidx21_i = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx21_i_indices.begin(), ptr_arrayidx21_i_indices.end(), "arrayidx21.i", label_forbody91); - ExtractElementInst* float_tmp23_i5 = new ExtractElementInst(packed_tmp7_i3, const_int32_45, "tmp23.i5", label_forbody91); - StoreInst* void_173 = new StoreInst(float_tmp23_i5, ptr_arrayidx21_i, false, label_forbody91); - std::vector ptr_arraydecay_i2_1_indices; - ptr_arraydecay_i2_1_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arraydecay_i2_1_indices.push_back(const_int32_43); - ptr_arraydecay_i2_1_indices.push_back(const_int32_41); - Instruction* ptr_arraydecay_i2_1 = new GetElementPtrInst(ptr_dests_132, ptr_arraydecay_i2_1_indices.begin(), ptr_arraydecay_i2_1_indices.end(), "arraydecay.i2.1", label_forbody91); + ptr_arrayidx21_i_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arrayidx21_i_indices.push_back(const_int32_32); + ptr_arrayidx21_i_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx21_i = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx21_i_indices.begin(), ptr_arrayidx21_i_indices.end(), "arrayidx21.i", label_forbody_130); + ExtractElementInst* float_tmp23_i = new ExtractElementInst(packed_tmp7_i, const_int32_36, "tmp23.i", label_forbody_130); + StoreInst* void_179 = new StoreInst(float_tmp23_i, ptr_arrayidx21_i, false, label_forbody_130); + std::vector ptr_arraydecay_i_1_indices; + ptr_arraydecay_i_1_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arraydecay_i_1_indices.push_back(const_int32_34); + ptr_arraydecay_i_1_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay_i_1 = new GetElementPtrInst(ptr_dests_121, ptr_arraydecay_i_1_indices.begin(), ptr_arraydecay_i_1_indices.end(), "arraydecay.i.1", label_forbody_130); std::vector ptr_arrayidx6_i_1_indices; - ptr_arrayidx6_i_1_indices.push_back(const_int32_41); - ptr_arrayidx6_i_1_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arrayidx6_i_1_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx6_i_1 = new GetElementPtrInst(ptr_results_138, ptr_arrayidx6_i_1_indices.begin(), ptr_arrayidx6_i_1_indices.end(), "arrayidx6.i.1", label_forbody91); - LoadInst* packed_tmp7_i3_1 = new LoadInst(ptr_arrayidx6_i_1, "tmp7.i3.1", false, label_forbody91); - ExtractElementInst* float_tmp11_i_1 = new ExtractElementInst(packed_tmp7_i3_1, const_int32_41, "tmp11.i.1", label_forbody91); - StoreInst* void_174 = new StoreInst(float_tmp11_i_1, ptr_arraydecay_i2_1, false, label_forbody91); + ptr_arrayidx6_i_1_indices.push_back(const_int32_32); + ptr_arrayidx6_i_1_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arrayidx6_i_1_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx6_i_1 = new GetElementPtrInst(ptr_results_134, ptr_arrayidx6_i_1_indices.begin(), ptr_arrayidx6_i_1_indices.end(), "arrayidx6.i.1", label_forbody_130); + LoadInst* packed_tmp7_i_1 = new LoadInst(ptr_arrayidx6_i_1, "tmp7.i.1", false, label_forbody_130); + ExtractElementInst* float_tmp11_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_32, "tmp11.i.1", label_forbody_130); + StoreInst* void_180 = new StoreInst(float_tmp11_i_1, ptr_arraydecay_i_1, false, label_forbody_130); std::vector ptr_arrayidx13_i_1_indices; - ptr_arrayidx13_i_1_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arrayidx13_i_1_indices.push_back(const_int32_43); - ptr_arrayidx13_i_1_indices.push_back(const_int32_43); - Instruction* ptr_arrayidx13_i_1 = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx13_i_1_indices.begin(), ptr_arrayidx13_i_1_indices.end(), "arrayidx13.i.1", label_forbody91); - ExtractElementInst* float_tmp15_i4_1 = new ExtractElementInst(packed_tmp7_i3_1, const_int32_43, "tmp15.i4.1", label_forbody91); - StoreInst* void_175 = new StoreInst(float_tmp15_i4_1, ptr_arrayidx13_i_1, false, label_forbody91); + ptr_arrayidx13_i_1_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arrayidx13_i_1_indices.push_back(const_int32_34); + ptr_arrayidx13_i_1_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx13_i_1 = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx13_i_1_indices.begin(), ptr_arrayidx13_i_1_indices.end(), "arrayidx13.i.1", label_forbody_130); + ExtractElementInst* float_tmp15_i2_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_34, "tmp15.i2.1", label_forbody_130); + StoreInst* void_181 = new StoreInst(float_tmp15_i2_1, ptr_arrayidx13_i_1, false, label_forbody_130); std::vector ptr_arrayidx17_i_1_indices; - ptr_arrayidx17_i_1_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arrayidx17_i_1_indices.push_back(const_int32_43); - ptr_arrayidx17_i_1_indices.push_back(const_int32_44); - Instruction* ptr_arrayidx17_i_1 = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx17_i_1_indices.begin(), ptr_arrayidx17_i_1_indices.end(), "arrayidx17.i.1", label_forbody91); - ExtractElementInst* float_tmp19_i_1 = new ExtractElementInst(packed_tmp7_i3_1, const_int32_44, "tmp19.i.1", label_forbody91); - StoreInst* void_176 = new StoreInst(float_tmp19_i_1, ptr_arrayidx17_i_1, false, label_forbody91); + ptr_arrayidx17_i_1_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arrayidx17_i_1_indices.push_back(const_int32_34); + ptr_arrayidx17_i_1_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx17_i_1 = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx17_i_1_indices.begin(), ptr_arrayidx17_i_1_indices.end(), "arrayidx17.i.1", label_forbody_130); + ExtractElementInst* float_tmp19_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_35, "tmp19.i.1", label_forbody_130); + StoreInst* void_182 = new StoreInst(float_tmp19_i_1, ptr_arrayidx17_i_1, false, label_forbody_130); std::vector ptr_arrayidx21_i_1_indices; - ptr_arrayidx21_i_1_indices.push_back(int32_i87_0_reg2mem_0); - ptr_arrayidx21_i_1_indices.push_back(const_int32_43); - ptr_arrayidx21_i_1_indices.push_back(const_int32_45); - Instruction* ptr_arrayidx21_i_1 = new GetElementPtrInst(ptr_dests_132, ptr_arrayidx21_i_1_indices.begin(), ptr_arrayidx21_i_1_indices.end(), "arrayidx21.i.1", label_forbody91); - ExtractElementInst* float_tmp23_i5_1 = new ExtractElementInst(packed_tmp7_i3_1, const_int32_45, "tmp23.i5.1", label_forbody91); - StoreInst* void_177 = new StoreInst(float_tmp23_i5_1, ptr_arrayidx21_i_1, false, label_forbody91); - BinaryOperator* int32_indvar_next19 = BinaryOperator::create(Instruction::Add, int32_i87_0_reg2mem_0, const_int32_43, "indvar.next19", label_forbody91); - ICmpInst* int1_exitcond20 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next19, const_int32_48, "exitcond20", label_forbody91); - new BranchInst(label_afterfor112, label_forbody91, int1_exitcond20, label_forbody91); - - // Block afterfor112 (label_afterfor112) - new ReturnInst(const_int32_49, label_afterfor112); + ptr_arrayidx21_i_1_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arrayidx21_i_1_indices.push_back(const_int32_34); + ptr_arrayidx21_i_1_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx21_i_1 = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx21_i_1_indices.begin(), ptr_arrayidx21_i_1_indices.end(), "arrayidx21.i.1", label_forbody_130); + ExtractElementInst* float_tmp23_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_36, "tmp23.i.1", label_forbody_130); + StoreInst* void_183 = new StoreInst(float_tmp23_i_1, ptr_arrayidx21_i_1, false, label_forbody_130); + BinaryOperator* int32_indvar_next20 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_173, const_int32_34, "indvar.next20", label_forbody_130); + ICmpInst* int1_exitcond21 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next20, const_int32_37, "exitcond21", label_forbody_130); + new BranchInst(label_afterfor_131, label_forbody_130, int1_exitcond21, label_forbody_130); + + // Block afterfor (label_afterfor_131) + new ReturnInst(const_int32_38, label_afterfor_131); + + // Block forbody6.i.1 (label_forbody6_i_1) + Argument* fwdref_186 = new Argument(IntegerType::get(32)); + PHINode* int32_j_0_reg2mem_0_i_1 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.1", label_forbody6_i_1); + int32_j_0_reg2mem_0_i_1->reserveOperandSpace(2); + int32_j_0_reg2mem_0_i_1->addIncoming(const_int32_32, label_forbody6_i_127); + int32_j_0_reg2mem_0_i_1->addIncoming(fwdref_186, label_forbody6_i_1); + + Argument* fwdref_187 = new Argument(VectorTy_13); + PHINode* packed_vec_0_reg2mem_0_i_1 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i.1", label_forbody6_i_1); + packed_vec_0_reg2mem_0_i_1->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_1->addIncoming(packed_tmp48_i_153, label_forbody6_i_127); + packed_vec_0_reg2mem_0_i_1->addIncoming(fwdref_187, label_forbody6_i_1); + + std::vector ptr_arraydecay11_i_1_indices; + ptr_arraydecay11_i_1_indices.push_back(const_int32_34); + ptr_arraydecay11_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); + ptr_arraydecay11_i_1_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay11_i_1 = new GetElementPtrInst(ptr_ainputs_122, ptr_arraydecay11_i_1_indices.begin(), ptr_arraydecay11_i_1_indices.end(), "arraydecay11.i.1", label_forbody6_i_1); + LoadInst* float_tmp13_i_1 = new LoadInst(ptr_arraydecay11_i_1, "tmp13.i.1", false, label_forbody6_i_1); + InsertElementInst* packed_tmp15_i_1 = new InsertElementInst(packed_vec_0_reg2mem_0_i_1, float_tmp13_i_1, const_int32_32, "tmp15.i.1", label_forbody6_i_1); + std::vector ptr_arrayidx23_i_1_indices; + ptr_arrayidx23_i_1_indices.push_back(const_int32_34); + ptr_arrayidx23_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); + ptr_arrayidx23_i_1_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx23_i_1 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx23_i_1_indices.begin(), ptr_arrayidx23_i_1_indices.end(), "arrayidx23.i.1", label_forbody6_i_1); + LoadInst* float_tmp24_i_1 = new LoadInst(ptr_arrayidx23_i_1, "tmp24.i.1", false, label_forbody6_i_1); + InsertElementInst* packed_tmp26_i_1 = new InsertElementInst(packed_tmp15_i_1, float_tmp24_i_1, const_int32_34, "tmp26.i.1", label_forbody6_i_1); + std::vector ptr_arrayidx34_i_1_indices; + ptr_arrayidx34_i_1_indices.push_back(const_int32_34); + ptr_arrayidx34_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); + ptr_arrayidx34_i_1_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx34_i_1 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx34_i_1_indices.begin(), ptr_arrayidx34_i_1_indices.end(), "arrayidx34.i.1", label_forbody6_i_1); + LoadInst* float_tmp35_i_1 = new LoadInst(ptr_arrayidx34_i_1, "tmp35.i.1", false, label_forbody6_i_1); + InsertElementInst* packed_tmp37_i_1 = new InsertElementInst(packed_tmp26_i_1, float_tmp35_i_1, const_int32_35, "tmp37.i.1", label_forbody6_i_1); + std::vector ptr_arrayidx45_i_1_indices; + ptr_arrayidx45_i_1_indices.push_back(const_int32_34); + ptr_arrayidx45_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); + ptr_arrayidx45_i_1_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx45_i_1 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx45_i_1_indices.begin(), ptr_arrayidx45_i_1_indices.end(), "arrayidx45.i.1", label_forbody6_i_1); + LoadInst* float_tmp46_i_1 = new LoadInst(ptr_arrayidx45_i_1, "tmp46.i.1", false, label_forbody6_i_1); + InsertElementInst* packed_tmp48_i_1 = new InsertElementInst(packed_tmp37_i_1, float_tmp46_i_1, const_int32_36, "tmp48.i.1", label_forbody6_i_1); + std::vector ptr_arrayidx54_i_1_indices; + ptr_arrayidx54_i_1_indices.push_back(const_int32_32); + ptr_arrayidx54_i_1_indices.push_back(const_int32_34); + ptr_arrayidx54_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); + Instruction* ptr_arrayidx54_i_1 = new GetElementPtrInst(ptr_inputs_132, ptr_arrayidx54_i_1_indices.begin(), ptr_arrayidx54_i_1_indices.end(), "arrayidx54.i.1", label_forbody6_i_1); + StoreInst* void_188 = new StoreInst(packed_tmp48_i_1, ptr_arrayidx54_i_1, false, label_forbody6_i_1); + BinaryOperator* int32_inc_i_1 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_1, const_int32_34, "inc.i.1", label_forbody6_i_1); + ICmpInst* int1_cmp59_i_1 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_1, int32_num_inputs_123, "cmp59.i.1", label_forbody6_i_1); + new BranchInst(label_forbody6_i_1, label_forbody6_i_2, int1_cmp59_i_1, label_forbody6_i_1); + + // Block forbody6.i.2 (label_forbody6_i_2) + Argument* fwdref_190 = new Argument(IntegerType::get(32)); + PHINode* int32_j_0_reg2mem_0_i_2 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.2", label_forbody6_i_2); + int32_j_0_reg2mem_0_i_2->reserveOperandSpace(2); + int32_j_0_reg2mem_0_i_2->addIncoming(const_int32_32, label_forbody6_i_1); + int32_j_0_reg2mem_0_i_2->addIncoming(fwdref_190, label_forbody6_i_2); + + Argument* fwdref_191 = new Argument(VectorTy_13); + PHINode* packed_vec_0_reg2mem_0_i_2 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i.2", label_forbody6_i_2); + packed_vec_0_reg2mem_0_i_2->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_2->addIncoming(packed_tmp48_i_1, label_forbody6_i_1); + packed_vec_0_reg2mem_0_i_2->addIncoming(fwdref_191, label_forbody6_i_2); + + std::vector ptr_arraydecay11_i_2_indices; + ptr_arraydecay11_i_2_indices.push_back(const_int32_35); + ptr_arraydecay11_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); + ptr_arraydecay11_i_2_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay11_i_2 = new GetElementPtrInst(ptr_ainputs_122, ptr_arraydecay11_i_2_indices.begin(), ptr_arraydecay11_i_2_indices.end(), "arraydecay11.i.2", label_forbody6_i_2); + LoadInst* float_tmp13_i_2 = new LoadInst(ptr_arraydecay11_i_2, "tmp13.i.2", false, label_forbody6_i_2); + InsertElementInst* packed_tmp15_i_2 = new InsertElementInst(packed_vec_0_reg2mem_0_i_2, float_tmp13_i_2, const_int32_32, "tmp15.i.2", label_forbody6_i_2); + std::vector ptr_arrayidx23_i_2_indices; + ptr_arrayidx23_i_2_indices.push_back(const_int32_35); + ptr_arrayidx23_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); + ptr_arrayidx23_i_2_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx23_i_2 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx23_i_2_indices.begin(), ptr_arrayidx23_i_2_indices.end(), "arrayidx23.i.2", label_forbody6_i_2); + LoadInst* float_tmp24_i_2 = new LoadInst(ptr_arrayidx23_i_2, "tmp24.i.2", false, label_forbody6_i_2); + InsertElementInst* packed_tmp26_i_2 = new InsertElementInst(packed_tmp15_i_2, float_tmp24_i_2, const_int32_34, "tmp26.i.2", label_forbody6_i_2); + std::vector ptr_arrayidx34_i_2_indices; + ptr_arrayidx34_i_2_indices.push_back(const_int32_35); + ptr_arrayidx34_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); + ptr_arrayidx34_i_2_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx34_i_2 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx34_i_2_indices.begin(), ptr_arrayidx34_i_2_indices.end(), "arrayidx34.i.2", label_forbody6_i_2); + LoadInst* float_tmp35_i_2 = new LoadInst(ptr_arrayidx34_i_2, "tmp35.i.2", false, label_forbody6_i_2); + InsertElementInst* packed_tmp37_i_2 = new InsertElementInst(packed_tmp26_i_2, float_tmp35_i_2, const_int32_35, "tmp37.i.2", label_forbody6_i_2); + std::vector ptr_arrayidx45_i_2_indices; + ptr_arrayidx45_i_2_indices.push_back(const_int32_35); + ptr_arrayidx45_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); + ptr_arrayidx45_i_2_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx45_i_2 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx45_i_2_indices.begin(), ptr_arrayidx45_i_2_indices.end(), "arrayidx45.i.2", label_forbody6_i_2); + LoadInst* float_tmp46_i_2 = new LoadInst(ptr_arrayidx45_i_2, "tmp46.i.2", false, label_forbody6_i_2); + InsertElementInst* packed_tmp48_i_2 = new InsertElementInst(packed_tmp37_i_2, float_tmp46_i_2, const_int32_36, "tmp48.i.2", label_forbody6_i_2); + std::vector ptr_arrayidx54_i_2_indices; + ptr_arrayidx54_i_2_indices.push_back(const_int32_32); + ptr_arrayidx54_i_2_indices.push_back(const_int32_35); + ptr_arrayidx54_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); + Instruction* ptr_arrayidx54_i_2 = new GetElementPtrInst(ptr_inputs_132, ptr_arrayidx54_i_2_indices.begin(), ptr_arrayidx54_i_2_indices.end(), "arrayidx54.i.2", label_forbody6_i_2); + StoreInst* void_192 = new StoreInst(packed_tmp48_i_2, ptr_arrayidx54_i_2, false, label_forbody6_i_2); + BinaryOperator* int32_inc_i_2 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_2, const_int32_34, "inc.i.2", label_forbody6_i_2); + ICmpInst* int1_cmp59_i_2 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_2, int32_num_inputs_123, "cmp59.i.2", label_forbody6_i_2); + new BranchInst(label_forbody6_i_2, label_forbody6_i_3, int1_cmp59_i_2, label_forbody6_i_2); + + // Block forbody6.i.3 (label_forbody6_i_3) + Argument* fwdref_194 = new Argument(IntegerType::get(32)); + PHINode* int32_j_0_reg2mem_0_i_3 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.3", label_forbody6_i_3); + int32_j_0_reg2mem_0_i_3->reserveOperandSpace(2); + int32_j_0_reg2mem_0_i_3->addIncoming(const_int32_32, label_forbody6_i_2); + int32_j_0_reg2mem_0_i_3->addIncoming(fwdref_194, label_forbody6_i_3); + + Argument* fwdref_195 = new Argument(VectorTy_13); + PHINode* packed_vec_0_reg2mem_0_i_3 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i.3", label_forbody6_i_3); + packed_vec_0_reg2mem_0_i_3->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_3->addIncoming(packed_tmp48_i_2, label_forbody6_i_2); + packed_vec_0_reg2mem_0_i_3->addIncoming(fwdref_195, label_forbody6_i_3); + + std::vector ptr_arraydecay11_i_3_indices; + ptr_arraydecay11_i_3_indices.push_back(const_int32_36); + ptr_arraydecay11_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); + ptr_arraydecay11_i_3_indices.push_back(const_int32_32); + Instruction* ptr_arraydecay11_i_3 = new GetElementPtrInst(ptr_ainputs_122, ptr_arraydecay11_i_3_indices.begin(), ptr_arraydecay11_i_3_indices.end(), "arraydecay11.i.3", label_forbody6_i_3); + LoadInst* float_tmp13_i_3 = new LoadInst(ptr_arraydecay11_i_3, "tmp13.i.3", false, label_forbody6_i_3); + InsertElementInst* packed_tmp15_i_3 = new InsertElementInst(packed_vec_0_reg2mem_0_i_3, float_tmp13_i_3, const_int32_32, "tmp15.i.3", label_forbody6_i_3); + std::vector ptr_arrayidx23_i_3_indices; + ptr_arrayidx23_i_3_indices.push_back(const_int32_36); + ptr_arrayidx23_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); + ptr_arrayidx23_i_3_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx23_i_3 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx23_i_3_indices.begin(), ptr_arrayidx23_i_3_indices.end(), "arrayidx23.i.3", label_forbody6_i_3); + LoadInst* float_tmp24_i_3 = new LoadInst(ptr_arrayidx23_i_3, "tmp24.i.3", false, label_forbody6_i_3); + InsertElementInst* packed_tmp26_i_3 = new InsertElementInst(packed_tmp15_i_3, float_tmp24_i_3, const_int32_34, "tmp26.i.3", label_forbody6_i_3); + std::vector ptr_arrayidx34_i_3_indices; + ptr_arrayidx34_i_3_indices.push_back(const_int32_36); + ptr_arrayidx34_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); + ptr_arrayidx34_i_3_indices.push_back(const_int32_35); + Instruction* ptr_arrayidx34_i_3 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx34_i_3_indices.begin(), ptr_arrayidx34_i_3_indices.end(), "arrayidx34.i.3", label_forbody6_i_3); + LoadInst* float_tmp35_i_3 = new LoadInst(ptr_arrayidx34_i_3, "tmp35.i.3", false, label_forbody6_i_3); + InsertElementInst* packed_tmp37_i_3 = new InsertElementInst(packed_tmp26_i_3, float_tmp35_i_3, const_int32_35, "tmp37.i.3", label_forbody6_i_3); + std::vector ptr_arrayidx45_i_3_indices; + ptr_arrayidx45_i_3_indices.push_back(const_int32_36); + ptr_arrayidx45_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); + ptr_arrayidx45_i_3_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx45_i_3 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx45_i_3_indices.begin(), ptr_arrayidx45_i_3_indices.end(), "arrayidx45.i.3", label_forbody6_i_3); + LoadInst* float_tmp46_i_3 = new LoadInst(ptr_arrayidx45_i_3, "tmp46.i.3", false, label_forbody6_i_3); + InsertElementInst* packed_tmp48_i_3 = new InsertElementInst(packed_tmp37_i_3, float_tmp46_i_3, const_int32_36, "tmp48.i.3", label_forbody6_i_3); + std::vector ptr_arrayidx54_i_3_indices; + ptr_arrayidx54_i_3_indices.push_back(const_int32_32); + ptr_arrayidx54_i_3_indices.push_back(const_int32_36); + ptr_arrayidx54_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); + Instruction* ptr_arrayidx54_i_3 = new GetElementPtrInst(ptr_inputs_132, ptr_arrayidx54_i_3_indices.begin(), ptr_arrayidx54_i_3_indices.end(), "arrayidx54.i.3", label_forbody6_i_3); + StoreInst* void_196 = new StoreInst(packed_tmp48_i_3, ptr_arrayidx54_i_3, false, label_forbody6_i_3); + BinaryOperator* int32_inc_i_3 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_3, const_int32_34, "inc.i.3", label_forbody6_i_3); + ICmpInst* int1_cmp59_i_3 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_3, int32_num_inputs_123, "cmp59.i.3", label_forbody6_i_3); + new BranchInst(label_forbody6_i_3, label_from_array_exit_128, int1_cmp59_i_3, label_forbody6_i_3); // Resolve Forward References - fwdref_154->replaceAllUsesWith(packed_tmp31_i_163); delete fwdref_154; - fwdref_152->replaceAllUsesWith(int32_indvar_next21); delete fwdref_152; - fwdref_168->replaceAllUsesWith(int32_indvar_next19); delete fwdref_168; + fwdref_195->replaceAllUsesWith(packed_tmp48_i_3); delete fwdref_195; + fwdref_194->replaceAllUsesWith(int32_inc_i_3); delete fwdref_194; + fwdref_141->replaceAllUsesWith(packed_tmp48_i_153); delete fwdref_141; + fwdref_139->replaceAllUsesWith(int32_inc_i_156); delete fwdref_139; + fwdref_162->replaceAllUsesWith(packed_tmp31_i_169); delete fwdref_162; + fwdref_161->replaceAllUsesWith(int32_indvar_next22); delete fwdref_161; + fwdref_174->replaceAllUsesWith(int32_indvar_next20); delete fwdref_174; + fwdref_187->replaceAllUsesWith(packed_tmp48_i_1); delete fwdref_187; + fwdref_186->replaceAllUsesWith(int32_inc_i_1); delete fwdref_186; + fwdref_191->replaceAllUsesWith(packed_tmp48_i_2); delete fwdref_191; + fwdref_190->replaceAllUsesWith(int32_inc_i_2); delete fwdref_190; } diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index b85490e1dc..2459d14cb8 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -215,15 +215,10 @@ struct tgsi_sampler struct softpipe_tile_cache *cache; }; -struct tgsi_interp_coef -{ - float a0[NUM_CHANNELS]; /* in an xyzw layout */ - float dadx[NUM_CHANNELS]; - float dady[NUM_CHANNELS]; -}; int run_fragment_shader(float x, float y, - float (*dests)[32][4], - struct tgsi_interp_coef *coef, + float (*dests)[16][4], + float (*ainputs)[16][4], + int num_inputs, float (*aconsts)[4], int num_consts, struct tgsi_sampler *samplers, @@ -233,40 +228,17 @@ int run_fragment_shader(float x, float y, float4 consts[32]; float4 results[4][16]; float4 temps[128];//MAX_PROGRAM_TEMPS + int kilmask = 0; - float4 fr1, fr2, fr3, fr4; - fr1.x = x; - fr1.y = y; - fr2.x = x + 1.f; - fr2.y = y; - fr3.x = x; - fr3.y = y + 1.f; - fr4.x = x + 1.f; - fr4.y = y + 1.f; - - inputs[0][0] = fr1; - inputs[1][0] = fr2; - inputs[2][0] = fr3; - inputs[3][0] = fr4; - - for (int i = 0; i < 4; ++i) { - float4 vec; - vec.x = coef->a0[0]; - vec.y = coef->a0[1]; - vec.z = coef->a0[2]; - vec.w = coef->a0[3]; - inputs[i][1] = vec; - } - /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", - num_vertices, num_inputs, num_attribs, num_consts);*/ - //from_array(inputs, ainputs, num_vertices, num_inputs); + from_array(inputs, ainputs, 4, num_inputs); from_consts(consts, aconsts, num_consts); - printf("AAAAAAAAAAAAAAAAAAAAAAA FRAGMENT SHADER %f %f\n", x, y); + //printf("AAAAAAAAAAAAAAAAAAAAAAA FRAGMENT SHADER %f %f\n", x, y); for (int i = 0; i < 4; ++i) { float4 *in = inputs[i]; float4 *res = results[i]; execute_shader(res, in, consts, temps); to_array(dests[i], res, 2); } + return ~kilmask; } diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 23505c46f6..f51054ef2b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -94,11 +94,6 @@ shade_quad( machine->SamplerUnits = softpipe->sampler_units; machine->InterpCoefs = quad->coef; - printf("COEF = [%f %f %f %f], [%f %f %f %f], [%f %f %f %f] %p\n", - quad->coef->a0[0], quad->coef->a0[1], quad->coef->a0[2], quad->coef->a0[3], - quad->coef->dadx[0], quad->coef->dadx[1], quad->coef->dadx[2], quad->coef->dadx[3], - quad->coef->dady[0], quad->coef->dady[1], quad->coef->dady[2], quad->coef->dady[3], - quad->coef); machine->Inputs[0].xyzw[0].f[0] = fx; machine->Inputs[0].xyzw[0].f[1] = fx + 1.0f; @@ -170,26 +165,59 @@ shade_quad_llvm(struct quad_stage *qs, { struct quad_shade_stage *qss = quad_shade_stage(qs); struct softpipe_context *softpipe = qs->softpipe; - float dests[4][32][4]; + float dests[4][16][4]; const float fx = (float) quad->x0; const float fy = (float) quad->y0; struct gallivm_prog *llvm = qss->llvm_prog; + float inputs[4][16][4]; + memset(inputs, 0, sizeof(inputs)); + + inputs[0][0][0] = fx; + inputs[1][0][0] = fx + 1.0f; + inputs[2][0][0] = fx; + inputs[3][0][0] = fx + 1.0f; + + inputs[0][0][1] = fy; + inputs[1][0][1] = fy; + inputs[2][0][1] = fy + 1.0f; + inputs[3][0][1] = fy + 1.0f; + printf("MASK = %d\n", quad->mask); + gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef); + for (int i = 0; i < 4; ++i) { + for (int j = 0; j < 2; ++j) { + printf("IN(%d,%d) [%f %f %f %f]\n", i, j, + inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]); + } + } + /*quad->mask &=*/ + gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs, + softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], + qss->samplers, softpipe->sampler_units); - quad->mask = gallivm_fragment_shader_exec( - llvm, fx, fy, dests, quad->coef, - softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], - qss->samplers, softpipe->sampler_units); + printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n", + dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], + dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]); /* store result color */ if (qss->colorOutSlot >= 0) { + unsigned i; /* XXX need to handle multiple color outputs someday */ assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] == TGSI_SEMANTIC_COLOR); - memcpy( - quad->outputs.color, - &dests[0][qss->colorOutSlot], - sizeof( quad->outputs.color ) ); + for (i = 0; i < QUAD_SIZE; ++i) { + quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0]; + quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1]; + quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2]; + quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3]; + } + } + for (int i = 0; i < QUAD_SIZE; ++i) { + printf("Q%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, + quad->outputs.color[0][i], + quad->outputs.color[1][i], + quad->outputs.color[2][i], + quad->outputs.color[3][i]); } /* store result Z */ @@ -197,16 +225,21 @@ shade_quad_llvm(struct quad_stage *qs, /* output[slot] is new Z */ uint i; for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = dests[0][2][i]; + quad->outputs.depth[i] = dests[i][0][2]; } } else { /* copy input Z (which was interpolated by the executor) to output Z */ uint i; for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = dests[0][2][i]; + quad->outputs.depth[i] = inputs[i][0][2]; } } + printf("D [%f, %f, %f, %f] mask = %d\n", + quad->outputs.depth[0], + quad->outputs.depth[1], + quad->outputs.depth[2], + quad->outputs.depth[3], quad->mask); /* shader may cull fragments */ if( quad->mask ) { -- cgit v1.2.3 From cf363ba30746ee0fd46b97986ea9fd753e093039 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 2 Nov 2007 07:02:15 -0400 Subject: Add debugging ifdefs to make it less verbose --- src/mesa/pipe/softpipe/sp_quad_fs.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index f51054ef2b..3d9dad0296 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -159,6 +159,7 @@ shade_quad( } } +#define DLLVM 0 static void shade_quad_llvm(struct quad_stage *qs, struct quad_header *quad) @@ -181,23 +182,28 @@ shade_quad_llvm(struct quad_stage *qs, inputs[1][0][1] = fy; inputs[2][0][1] = fy + 1.0f; inputs[3][0][1] = fy + 1.0f; +#if DLLVM printf("MASK = %d\n", quad->mask); +#endif gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef); +#if DLLVM for (int i = 0; i < 4; ++i) { for (int j = 0; j < 2; ++j) { printf("IN(%d,%d) [%f %f %f %f]\n", i, j, inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]); } } +#endif /*quad->mask &=*/ gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs, softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], qss->samplers, softpipe->sampler_units); - +#if DLLVM printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n", dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]); +#endif /* store result color */ if (qss->colorOutSlot >= 0) { @@ -212,6 +218,7 @@ shade_quad_llvm(struct quad_stage *qs, quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3]; } } +#if DLLVM for (int i = 0; i < QUAD_SIZE; ++i) { printf("Q%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, quad->outputs.color[0][i], @@ -219,6 +226,7 @@ shade_quad_llvm(struct quad_stage *qs, quad->outputs.color[2][i], quad->outputs.color[3][i]); } +#endif /* store result Z */ if (qss->depthOutSlot >= 0) { @@ -235,11 +243,13 @@ shade_quad_llvm(struct quad_stage *qs, quad->outputs.depth[i] = inputs[i][0][2]; } } +#if DLLVM printf("D [%f, %f, %f, %f] mask = %d\n", quad->outputs.depth[0], quad->outputs.depth[1], quad->outputs.depth[2], quad->outputs.depth[3], quad->mask); +#endif /* shader may cull fragments */ if( quad->mask ) { -- cgit v1.2.3 From 5c7bfb06e087ce4162590359ad75d1fca98f3549 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 2 Nov 2007 11:47:09 -0400 Subject: Implement COS and CMP opcode. There's some weird rounding issue with COS that I can't figure out. --- src/mesa/pipe/llvm/gallivm.cpp | 8 +- src/mesa/pipe/llvm/gallivm_builtins.cpp | 372 ++++++++++++++++++++++++++------ src/mesa/pipe/llvm/instructions.cpp | 35 ++- src/mesa/pipe/llvm/instructions.h | 2 + src/mesa/pipe/llvm/llvm_builtins.c | 28 ++- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- 6 files changed, 379 insertions(+), 68 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index b5e90b037e..a737b5631f 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -424,7 +424,9 @@ translate_instruction(llvm::Module *module, out = instr->dph(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_COS: + case TGSI_OPCODE_COS: { + out = instr->cos(inputs[0]); + } break; case TGSI_OPCODE_DDX: break; @@ -496,7 +498,9 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_SSG: break; - case TGSI_OPCODE_CMP: + case TGSI_OPCODE_CMP: { + out = instr->cmp(inputs[0], inputs[1], inputs[2]); + } break; case TGSI_OPCODE_SCS: break; diff --git a/src/mesa/pipe/llvm/gallivm_builtins.cpp b/src/mesa/pipe/llvm/gallivm_builtins.cpp index 4acbedc566..b23af91550 100644 --- a/src/mesa/pipe/llvm/gallivm_builtins.cpp +++ b/src/mesa/pipe/llvm/gallivm_builtins.cpp @@ -6,77 +6,192 @@ Module* createGallivmBuiltins(Module *mod) { mod->setModuleIdentifier("shader"); // Type Definitions -std::vectorFuncTy_0_args; -FuncTy_0_args.push_back(Type::FloatTy); -FuncTy_0_args.push_back(Type::FloatTy); -ParamAttrsList *FuncTy_0_PAL = 0; -FunctionType* FuncTy_0 = FunctionType::get( +ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 25); + +PointerType* PointerTy_1 = PointerType::get(ArrayTy_0); + +std::vectorFuncTy_2_args; +FuncTy_2_args.push_back(Type::FloatTy); +FuncTy_2_args.push_back(Type::FloatTy); +ParamAttrsList *FuncTy_2_PAL = 0; +FunctionType* FuncTy_2 = FunctionType::get( /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_0_args, + /*Params=*/FuncTy_2_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_0_PAL); + /*ParamAttrs=*/FuncTy_2_PAL); + +PointerType* PointerTy_3 = PointerType::get(FuncTy_2); -PointerType* PointerTy_1 = PointerType::get(FuncTy_0); +VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); -VectorType* VectorTy_2 = VectorType::get(Type::FloatTy, 4); +std::vectorFuncTy_5_args; +FuncTy_5_args.push_back(VectorTy_4); +ParamAttrsList *FuncTy_5_PAL = 0; +FunctionType* FuncTy_5 = FunctionType::get( + /*Result=*/VectorTy_4, + /*Params=*/FuncTy_5_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_5_PAL); + +std::vectorFuncTy_6_args; +FuncTy_6_args.push_back(VectorTy_4); +FuncTy_6_args.push_back(VectorTy_4); +FuncTy_6_args.push_back(VectorTy_4); +ParamAttrsList *FuncTy_6_PAL = 0; +FunctionType* FuncTy_6 = FunctionType::get( + /*Result=*/VectorTy_4, + /*Params=*/FuncTy_6_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_6_PAL); -std::vectorFuncTy_3_args; -FuncTy_3_args.push_back(VectorTy_2); -ParamAttrsList *FuncTy_3_PAL = 0; -FunctionType* FuncTy_3 = FunctionType::get( - /*Result=*/VectorTy_2, - /*Params=*/FuncTy_3_args, +VectorType* VectorTy_7 = VectorType::get(IntegerType::get(32), 4); + +std::vectorFuncTy_9_args; +ParamAttrsList *FuncTy_9_PAL = 0; +FunctionType* FuncTy_9 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_9_args, + /*isVarArg=*/true, + /*ParamAttrs=*/FuncTy_9_PAL); + +PointerType* PointerTy_8 = PointerType::get(FuncTy_9); + +PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8)); + +std::vectorFuncTy_12_args; +FuncTy_12_args.push_back(Type::FloatTy); +ParamAttrsList *FuncTy_12_PAL = 0; +FunctionType* FuncTy_12 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_12_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_3_PAL); + /*ParamAttrs=*/FuncTy_12_PAL); + +PointerType* PointerTy_11 = PointerType::get(FuncTy_12); // Function Declarations Function* func_approx = new Function( - /*Type=*/FuncTy_0, + /*Type=*/FuncTy_2, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"approx", mod); func_approx->setCallingConv(CallingConv::C); Function* func_powf = new Function( - /*Type=*/FuncTy_0, + /*Type=*/FuncTy_2, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"powf", mod); // (external, no body) func_powf->setCallingConv(CallingConv::C); Function* func_lit = new Function( - /*Type=*/FuncTy_3, + /*Type=*/FuncTy_5, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"lit", mod); func_lit->setCallingConv(CallingConv::C); +Function* func_cmp = new Function( + /*Type=*/FuncTy_6, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"cmp", mod); +func_cmp->setCallingConv(CallingConv::C); + +Function* func_vcos = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"vcos", mod); +func_vcos->setCallingConv(CallingConv::C); + +Function* func_printf = new Function( + /*Type=*/FuncTy_9, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", mod); // (external, no body) +func_printf->setCallingConv(CallingConv::C); + +Function* func_cosf = new Function( + /*Type=*/FuncTy_12, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"cosf", mod); // (external, no body) +func_cosf->setCallingConv(CallingConv::C); + // Global Variable Declarations +GlobalVariable* gvar_array__str = new GlobalVariable( +/*Type=*/ArrayTy_0, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str", +mod); + +GlobalVariable* gvar_array__str1 = new GlobalVariable( +/*Type=*/ArrayTy_0, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str1", +mod); + // Constant Definitions -ConstantFP* const_float_4 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); -ConstantFP* const_float_5 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); -Constant* const_float_6 = Constant::getNullValue(Type::FloatTy); -Constant* const_int32_7 = Constant::getNullValue(IntegerType::get(32)); -std::vector const_packed_8_elems; -ConstantFP* const_float_9 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); -const_packed_8_elems.push_back(const_float_9); -UndefValue* const_float_10 = UndefValue::get(Type::FloatTy); -const_packed_8_elems.push_back(const_float_10); -const_packed_8_elems.push_back(const_float_10); -const_packed_8_elems.push_back(const_float_9); -Constant* const_packed_8 = ConstantVector::get(VectorTy_2, const_packed_8_elems); -ConstantInt* const_int32_11 = ConstantInt::get(APInt(32, "1", 10)); -ConstantInt* const_int32_12 = ConstantInt::get(APInt(32, "3", 10)); -ConstantInt* const_int32_13 = ConstantInt::get(APInt(32, "2", 10)); -std::vector const_packed_14_elems; -const_packed_14_elems.push_back(const_float_9); -const_packed_14_elems.push_back(const_float_6); -const_packed_14_elems.push_back(const_float_6); -const_packed_14_elems.push_back(const_float_9); -Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elems); +Constant* const_array_13 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); +Constant* const_array_14 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); +ConstantFP* const_float_15 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); +ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); +Constant* const_float_17 = Constant::getNullValue(Type::FloatTy); +Constant* const_int32_18 = Constant::getNullValue(IntegerType::get(32)); +std::vector const_packed_19_elems; +ConstantFP* const_float_20 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); +const_packed_19_elems.push_back(const_float_20); +UndefValue* const_float_21 = UndefValue::get(Type::FloatTy); +const_packed_19_elems.push_back(const_float_21); +const_packed_19_elems.push_back(const_float_21); +const_packed_19_elems.push_back(const_float_20); +Constant* const_packed_19 = ConstantVector::get(VectorTy_4, const_packed_19_elems); +ConstantInt* const_int32_22 = ConstantInt::get(APInt(32, "1", 10)); +ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "3", 10)); +ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "2", 10)); +std::vector const_packed_25_elems; +const_packed_25_elems.push_back(const_float_20); +const_packed_25_elems.push_back(const_float_17); +const_packed_25_elems.push_back(const_float_17); +const_packed_25_elems.push_back(const_float_20); +Constant* const_packed_25 = ConstantVector::get(VectorTy_4, const_packed_25_elems); +Constant* const_double_26 = Constant::getNullValue(Type::DoubleTy); +std::vector const_packed_27_elems; +const_packed_27_elems.push_back(const_int32_18); +ConstantInt* const_int32_28 = ConstantInt::get(APInt(32, "5", 10)); +const_packed_27_elems.push_back(const_int32_28); +const_packed_27_elems.push_back(const_int32_24); +const_packed_27_elems.push_back(const_int32_23); +Constant* const_packed_27 = ConstantVector::get(VectorTy_7, const_packed_27_elems); +std::vector const_packed_29_elems; +const_packed_29_elems.push_back(const_int32_18); +const_packed_29_elems.push_back(const_int32_22); +ConstantInt* const_int32_30 = ConstantInt::get(APInt(32, "6", 10)); +const_packed_29_elems.push_back(const_int32_30); +const_packed_29_elems.push_back(const_int32_23); +Constant* const_packed_29 = ConstantVector::get(VectorTy_7, const_packed_29_elems); +std::vector const_packed_31_elems; +const_packed_31_elems.push_back(const_int32_18); +const_packed_31_elems.push_back(const_int32_22); +const_packed_31_elems.push_back(const_int32_24); +ConstantInt* const_int32_32 = ConstantInt::get(APInt(32, "7", 10)); +const_packed_31_elems.push_back(const_int32_32); +Constant* const_packed_31 = ConstantVector::get(VectorTy_7, const_packed_31_elems); +std::vector const_ptr_33_indices; +const_ptr_33_indices.push_back(const_int32_18); +const_ptr_33_indices.push_back(const_int32_18); +Constant* const_ptr_33 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_33_indices[0], const_ptr_33_indices.size() ); +UndefValue* const_packed_34 = UndefValue::get(VectorTy_4); +std::vector const_ptr_35_indices; +const_ptr_35_indices.push_back(const_int32_18); +const_ptr_35_indices.push_back(const_int32_18); +Constant* const_ptr_35 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_35_indices[0], const_ptr_35_indices.size() ); // Global Variable Definitions +gvar_array__str->setInitializer(const_array_13); +gvar_array__str1->setInitializer(const_array_14); // Function Definitions @@ -91,12 +206,12 @@ Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elem BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); // Block entry (label_entry) - FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_4, "cmp", label_entry); - SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_4, float_b, "b.addr.0", label_entry); - FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_5, "cmp3", label_entry); - SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_5, float_b_addr_0, "b.addr.1", label_entry); - FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_6, "cmp7", label_entry); - SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_6, float_a, "a.addr.0", label_entry); + FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_15, "cmp", label_entry); + SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_15, float_b, "b.addr.0", label_entry); + FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_16, "cmp3", label_entry); + SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_16, float_b_addr_0, "b.addr.1", label_entry); + FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_17, "cmp7", label_entry); + SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_17, float_a, "a.addr.0", label_entry); std::vector float_call_params; float_call_params.push_back(float_a_addr_0); float_call_params.push_back(float_b_addr_1); @@ -113,36 +228,169 @@ Constant* const_packed_14 = ConstantVector::get(VectorTy_2, const_packed_14_elem Value* packed_tmp = args++; packed_tmp->setName("tmp"); - BasicBlock* label_entry_16 = new BasicBlock("entry",func_lit,0); + BasicBlock* label_entry_37 = new BasicBlock("entry",func_lit,0); BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); - // Block entry (label_entry_16) - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_7, "tmp7", label_entry_16); - FCmpInst* int1_cmp_17 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_6, "cmp", label_entry_16); - new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_17, label_entry_16); + // Block entry (label_entry_37) + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_18, "tmp7", label_entry_37); + FCmpInst* int1_cmp_38 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_17, "cmp", label_entry_37); + new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_38, label_entry_37); // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_8, float_tmp7, const_int32_11, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_11, "tmp14", label_ifthen); - ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_12, "tmp16", label_ifthen); - FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_4, "cmp.i", label_ifthen); - SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_4, float_tmp16, "b.addr.0.i", label_ifthen); - FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_5, "cmp3.i", label_ifthen); - SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_5, float_b_addr_0_i, "b.addr.1.i", label_ifthen); - FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_6, "cmp7.i", label_ifthen); - SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_6, float_tmp14, "a.addr.0.i", label_ifthen); + InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_19, float_tmp7, const_int32_22, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_22, "tmp14", label_ifthen); + ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp16", label_ifthen); + FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_15, "cmp.i", label_ifthen); + SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_15, float_tmp16, "b.addr.0.i", label_ifthen); + FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_16, "cmp3.i", label_ifthen); + SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_16, float_b_addr_0_i, "b.addr.1.i", label_ifthen); + FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_17, "cmp7.i", label_ifthen); + SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_17, float_tmp14, "a.addr.0.i", label_ifthen); std::vector float_call_i_params; float_call_i_params.push_back(float_a_addr_0_i); float_call_i_params.push_back(float_b_addr_1_i); CallInst* float_call_i = new CallInst(func_powf, float_call_i_params.begin(), float_call_i_params.end(), "call.i", label_ifthen); float_call_i->setCallingConv(CallingConv::C); float_call_i->setTailCall(true); - InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_13, "tmp18", label_ifthen); + InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_24, "tmp18", label_ifthen); new ReturnInst(packed_tmp18, label_ifthen); // Block UnifiedReturnBlock (label_UnifiedReturnBlock) - new ReturnInst(const_packed_14, label_UnifiedReturnBlock); + new ReturnInst(const_packed_25, label_UnifiedReturnBlock); + +} + +// Function: cmp (func_cmp) +{ + Function::arg_iterator args = func_cmp->arg_begin(); + Value* packed_tmp0 = args++; + packed_tmp0->setName("tmp0"); + Value* packed_tmp1 = args++; + packed_tmp1->setName("tmp1"); + Value* packed_tmp2 = args++; + packed_tmp2->setName("tmp2"); + + BasicBlock* label_entry_42 = new BasicBlock("entry",func_cmp,0); + BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); + BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); + BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); + BasicBlock* label_cond_cont34 = new BasicBlock("cond.cont34",func_cmp,0); + BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); + BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); + + // Block entry (label_entry_42) + ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_18, "tmp3", label_entry_42); + CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_42); + FCmpInst* int1_cmp_43 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_26, "cmp", label_entry_42); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_22, "tmp11", label_entry_42); + CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_42); + FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_26, "cmp13", label_entry_42); + SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_43, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_42); + new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_42); + + // Block cond.?14 (label_cond__14) + ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_27, "tmp233", label_cond__14); + ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp254", label_cond__14); + CastInst* double_conv265 = new FPExtInst(float_tmp254, Type::DoubleTy, "conv265", label_cond__14); + FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_26, "cmp276", label_cond__14); + new BranchInst(label_cond__28, label_cond_cont34, int1_cmp276, label_cond__14); + + // Block cond.cont20 (label_cond_cont20) + ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_27, "tmp23", label_cond_cont20); + ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp25", label_cond_cont20); + CastInst* double_conv26 = new FPExtInst(float_tmp25, Type::DoubleTy, "conv26", label_cond_cont20); + FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_26, "cmp27", label_cond_cont20); + new BranchInst(label_cond__28, label_cond_cont34, int1_cmp27, label_cond_cont20); + + // Block cond.?28 (label_cond__28) + PHINode* packed_tmp23_reg2mem_0 = new PHINode(VectorTy_4, "tmp23.reg2mem.0", label_cond__28); + packed_tmp23_reg2mem_0->reserveOperandSpace(2); + packed_tmp23_reg2mem_0->addIncoming(packed_tmp233, label_cond__14); + packed_tmp23_reg2mem_0->addIncoming(packed_tmp23, label_cond_cont20); + + ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_29, "tmp378", label_cond__28); + ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp399", label_cond__28); + CastInst* double_conv4010 = new FPExtInst(float_tmp399, Type::DoubleTy, "conv4010", label_cond__28); + FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_26, "cmp4111", label_cond__28); + new BranchInst(label_cond__42, label_cond_cont48, int1_cmp4111, label_cond__28); + + // Block cond.cont34 (label_cond_cont34) + PHINode* packed_tmp23_reg2mem_1 = new PHINode(VectorTy_4, "tmp23.reg2mem.1", label_cond_cont34); + packed_tmp23_reg2mem_1->reserveOperandSpace(2); + packed_tmp23_reg2mem_1->addIncoming(packed_tmp233, label_cond__14); + packed_tmp23_reg2mem_1->addIncoming(packed_tmp23, label_cond_cont20); + + ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_29, "tmp37", label_cond_cont34); + ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp39", label_cond_cont34); + CastInst* double_conv40 = new FPExtInst(float_tmp39, Type::DoubleTy, "conv40", label_cond_cont34); + FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_26, "cmp41", label_cond_cont34); + new BranchInst(label_cond__42, label_cond_cont48, int1_cmp41, label_cond_cont34); + + // Block cond.?42 (label_cond__42) + PHINode* packed_tmp37_reg2mem_0 = new PHINode(VectorTy_4, "tmp37.reg2mem.0", label_cond__42); + packed_tmp37_reg2mem_0->reserveOperandSpace(2); + packed_tmp37_reg2mem_0->addIncoming(packed_tmp378, label_cond__28); + packed_tmp37_reg2mem_0->addIncoming(packed_tmp37, label_cond_cont34); + + ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_31, "tmp5113", label_cond__42); + new ReturnInst(packed_tmp5113, label_cond__42); + + // Block cond.cont48 (label_cond_cont48) + PHINode* packed_tmp37_reg2mem_1 = new PHINode(VectorTy_4, "tmp37.reg2mem.1", label_cond_cont48); + packed_tmp37_reg2mem_1->reserveOperandSpace(2); + packed_tmp37_reg2mem_1->addIncoming(packed_tmp378, label_cond__28); + packed_tmp37_reg2mem_1->addIncoming(packed_tmp37, label_cond_cont34); + + ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_31, "tmp51", label_cond_cont48); + new ReturnInst(packed_tmp51, label_cond_cont48); + +} + +// Function: vcos (func_vcos) +{ + Function::arg_iterator args = func_vcos->arg_begin(); + Value* packed_val = args++; + packed_val->setName("val"); + + BasicBlock* label_entry_51 = new BasicBlock("entry",func_vcos,0); + + // Block entry (label_entry_51) + ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_18, "tmp1", label_entry_51); + CastInst* double_conv_52 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_51); + ExtractElementInst* float_tmp3_53 = new ExtractElementInst(packed_val, const_int32_22, "tmp3", label_entry_51); + CastInst* double_conv4 = new FPExtInst(float_tmp3_53, Type::DoubleTy, "conv4", label_entry_51); + ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_val, const_int32_24, "tmp6", label_entry_51); + CastInst* double_conv7 = new FPExtInst(float_tmp6, Type::DoubleTy, "conv7", label_entry_51); + ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_23, "tmp9", label_entry_51); + CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_51); + std::vector int32_call_params; + int32_call_params.push_back(const_ptr_33); + int32_call_params.push_back(double_conv_52); + int32_call_params.push_back(double_conv4); + int32_call_params.push_back(double_conv7); + int32_call_params.push_back(double_conv10); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_51); + int32_call->setCallingConv(CallingConv::C); + int32_call->setTailCall(true); + CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_51); + float_call13->setCallingConv(CallingConv::C); + float_call13->setTailCall(true); + InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_34, float_call13, const_int32_18, "tmp15", label_entry_51); + InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call13, const_int32_22, "tmp20", label_entry_51); + InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call13, const_int32_24, "tmp25", label_entry_51); + InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call13, const_int32_23, "tmp30", label_entry_51); + CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_51); + std::vector int32_call43_params; + int32_call43_params.push_back(const_ptr_35); + int32_call43_params.push_back(double_conv33); + int32_call43_params.push_back(double_conv33); + int32_call43_params.push_back(double_conv33); + int32_call43_params.push_back(double_conv33); + CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_51); + int32_call43->setCallingConv(CallingConv::C); + int32_call43->setTailCall(true); + new ReturnInst(packed_tmp30, label_entry_51); } diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index aa6e108598..8e74ab3e94 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -87,7 +87,7 @@ llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, Value *mulRes = mul(in1, in2); return add(mulRes, in3); } - + llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2) { return m_builder.CreateMul(in1, in2, name("mul")); @@ -830,7 +830,38 @@ std::vector Instructions::extractVector(llvm::Value *vec) return elems; } -#endif //MESA_LLVM +llvm::Value * Instructions::cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3) +{ + llvm::Function *func = m_mod->getFunction("cmp"); + assert(func); + std::vector params; + params.push_back(in1); + params.push_back(in2); + params.push_back(in3); + CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), name("cmpres")); + call->setTailCall(false); + return call; +} +llvm::Value * Instructions::cos(llvm::Value *in) +{ +#if 0 + llvm::Function *func = m_mod->getFunction("vcos"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("cosres")); + call->setTailCall(false); + return call; +#else + std::vector elems = extractVector(in); + Function *func = m_mod->getFunction("cosf"); + assert(func); + CallInst *cos = m_builder.CreateCall(func, elems[0], name("cosres")); + cos->setCallingConv(CallingConv::C); + cos->setTailCall(true); + return vectorFromVals(cos, cos, cos, cos); +#endif +} +#endif //MESA_LLVM diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 83d07e2f2e..d33f61a9c1 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -64,6 +64,8 @@ public: void brk(); void cal(int label, llvm::Value *out, llvm::Value *in, llvm::Value *cst, llvm::Value *tmp); + llvm::Value *cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); + llvm::Value *cos(llvm::Value *in); llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index 9d3d6141a6..a570e0e089 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -43,7 +43,7 @@ inline float approx(float a, float b) return powf(a, b); } -float4 lit(float4 tmp) +inline float4 lit(float4 tmp) { float4 result; result.x = 1.0; @@ -57,3 +57,29 @@ float4 lit(float4 tmp) } return result; } + +inline float4 cmp(float4 tmp0, float4 tmp1, float4 tmp2) +{ + float4 result; + + result.x = (tmp0.x < 0.0) ? tmp1.x : tmp2.x; + result.y = (tmp0.y < 0.0) ? tmp1.y : tmp2.y; + result.z = (tmp0.z < 0.0) ? tmp1.z : tmp2.z; + result.w = (tmp0.w < 0.0) ? tmp1.w : tmp2.w; + + return result; +} + +extern float cosf(float val); + +inline float4 vcos(float4 val) +{ + float4 result; + printf("VEC IN is %f %f %f %f\n", val.x, val.y, val.z, val.w); + result.x = cosf(val.x); + result.y = cosf(val.x); + result.z = cosf(val.x); + result.w = cosf(val.x); + printf("VEC OUT is %f %f %f %f\n", result.x, result.y, result.z, result.w); + return result; +} diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 3d9dad0296..2cecfb268a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -220,7 +220,7 @@ shade_quad_llvm(struct quad_stage *qs, } #if DLLVM for (int i = 0; i < QUAD_SIZE; ++i) { - printf("Q%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, + printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, quad->outputs.color[0][i], quad->outputs.color[1][i], quad->outputs.color[2][i], -- cgit v1.2.3 From e0e91e7ceb50f0e23311788559a8547dd24c7a80 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 2 Nov 2007 12:05:00 -0400 Subject: Implement scs opcode --- src/mesa/pipe/llvm/gallivm.cpp | 4 +++- src/mesa/pipe/llvm/gallivm_builtins.cpp | 34 +++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.cpp | 11 +++++++++++ src/mesa/pipe/llvm/instructions.h | 1 + src/mesa/pipe/llvm/llvm_builtins.c | 10 ++++++++++ 5 files changed, 59 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index a737b5631f..7e91f8c556 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -502,7 +502,9 @@ translate_instruction(llvm::Module *module, out = instr->cmp(inputs[0], inputs[1], inputs[2]); } break; - case TGSI_OPCODE_SCS: + case TGSI_OPCODE_SCS: { + out = instr->scs(inputs[0]); + } break; case TGSI_OPCODE_TXB: break; diff --git a/src/mesa/pipe/llvm/gallivm_builtins.cpp b/src/mesa/pipe/llvm/gallivm_builtins.cpp index b23af91550..b06629265a 100644 --- a/src/mesa/pipe/llvm/gallivm_builtins.cpp +++ b/src/mesa/pipe/llvm/gallivm_builtins.cpp @@ -114,6 +114,18 @@ Function* func_cosf = new Function( /*Name=*/"cosf", mod); // (external, no body) func_cosf->setCallingConv(CallingConv::C); +Function* func_scs = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"scs", mod); +func_scs->setCallingConv(CallingConv::C); + +Function* func_sinf = new Function( + /*Type=*/FuncTy_12, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"sinf", mod); // (external, no body) +func_sinf->setCallingConv(CallingConv::C); + // Global Variable Declarations @@ -394,6 +406,28 @@ gvar_array__str1->setInitializer(const_array_14); } +// Function: scs (func_scs) +{ + Function::arg_iterator args = func_scs->arg_begin(); + Value* packed_val_55 = args++; + packed_val_55->setName("val"); + + BasicBlock* label_entry_56 = new BasicBlock("entry",func_scs,0); + + // Block entry (label_entry_56) + ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_55, const_int32_18, "tmp2", label_entry_56); + CallInst* float_call_57 = new CallInst(func_cosf, float_tmp2, "call", label_entry_56); + float_call_57->setCallingConv(CallingConv::C); + float_call_57->setTailCall(true); + InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_34, float_call_57, const_int32_18, "tmp5", label_entry_56); + CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_56); + float_call7->setCallingConv(CallingConv::C); + float_call7->setTailCall(true); + InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_22, "tmp9", label_entry_56); + new ReturnInst(packed_tmp9, label_entry_56); + +} + return mod; } diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 8e74ab3e94..c4a1b2d5c1 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -864,4 +864,15 @@ llvm::Value * Instructions::cos(llvm::Value *in) #endif } +llvm::Value * Instructions::scs(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("scs"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("scsres")); + call->setTailCall(false); + return call; +} + #endif //MESA_LLVM + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index d33f61a9c1..95c845e862 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -92,6 +92,7 @@ public: llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); llvm::Value *rcp(llvm::Value *in); llvm::Value *rsq(llvm::Value *in); + llvm::Value *scs(llvm::Value *in); llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index a570e0e089..ca15995557 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -71,6 +71,7 @@ inline float4 cmp(float4 tmp0, float4 tmp1, float4 tmp2) } extern float cosf(float val); +extern float sinf(float val); inline float4 vcos(float4 val) { @@ -83,3 +84,12 @@ inline float4 vcos(float4 val) printf("VEC OUT is %f %f %f %f\n", result.x, result.y, result.z, result.w); return result; } + +inline float4 scs(float4 val) +{ + float4 result; + float tmp = val.x; + result.x = cosf(tmp); + result.y = sinf(tmp); + return result; +} -- cgit v1.2.3 From a2debc2704b9126d92d947c0407a0fbd709ab932 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 2 Nov 2007 12:09:23 -0400 Subject: Implement sin opcode. Seems to have similar rounding border problems as cos. --- src/mesa/pipe/llvm/gallivm.cpp | 4 +++- src/mesa/pipe/llvm/gallivm_builtins.cpp | 27 +++++++++++++++++++++++++++ src/mesa/pipe/llvm/instructions.cpp | 10 ++++++++++ src/mesa/pipe/llvm/instructions.h | 1 + src/mesa/pipe/llvm/llvm_builtins.c | 13 +++++++++++++ 5 files changed, 54 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index 7e91f8c556..bd8bfac208 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -452,7 +452,9 @@ translate_instruction(llvm::Module *module, out = instr->sgt(inputs[0], inputs[1]); } break; - case TGSI_OPCODE_SIN: + case TGSI_OPCODE_SIN: { + out = instr->sin(inputs[0]); + } break; case TGSI_OPCODE_SLE: break; diff --git a/src/mesa/pipe/llvm/gallivm_builtins.cpp b/src/mesa/pipe/llvm/gallivm_builtins.cpp index b06629265a..da1e6ae1de 100644 --- a/src/mesa/pipe/llvm/gallivm_builtins.cpp +++ b/src/mesa/pipe/llvm/gallivm_builtins.cpp @@ -126,6 +126,12 @@ Function* func_sinf = new Function( /*Name=*/"sinf", mod); // (external, no body) func_sinf->setCallingConv(CallingConv::C); +Function* func_vsin = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"vsin", mod); +func_vsin->setCallingConv(CallingConv::C); + // Global Variable Declarations @@ -428,6 +434,27 @@ gvar_array__str1->setInitializer(const_array_14); } +// Function: vsin (func_vsin) +{ + Function::arg_iterator args = func_vsin->arg_begin(); + Value* packed_val_59 = args++; + packed_val_59->setName("val"); + + BasicBlock* label_entry_60 = new BasicBlock("entry",func_vsin,0); + + // Block entry (label_entry_60) + ExtractElementInst* float_tmp2_61 = new ExtractElementInst(packed_val_59, const_int32_18, "tmp2", label_entry_60); + CallInst* float_call_62 = new CallInst(func_sinf, float_tmp2_61, "call", label_entry_60); + float_call_62->setCallingConv(CallingConv::C); + float_call_62->setTailCall(true); + InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_34, float_call_62, const_int32_18, "tmp6", label_entry_60); + InsertElementInst* packed_tmp9_63 = new InsertElementInst(packed_tmp6, float_call_62, const_int32_22, "tmp9", label_entry_60); + InsertElementInst* packed_tmp12_64 = new InsertElementInst(packed_tmp9_63, float_call_62, const_int32_24, "tmp12", label_entry_60); + InsertElementInst* packed_tmp15_65 = new InsertElementInst(packed_tmp12_64, float_call_62, const_int32_23, "tmp15", label_entry_60); + new ReturnInst(packed_tmp15_65, label_entry_60); + +} + return mod; } diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index c4a1b2d5c1..232dd9cd5d 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -874,5 +874,15 @@ llvm::Value * Instructions::scs(llvm::Value *in) return call; } + +llvm::Value * Instructions::sin(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("vsin"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("sinres")); + call->setTailCall(false); + return call; +} #endif //MESA_LLVM diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index 95c845e862..e9bfc9d740 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -95,6 +95,7 @@ public: llvm::Value *scs(llvm::Value *in); llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sin(llvm::Value *in); llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); llvm::Value *trunc(llvm::Value *in); diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index ca15995557..517aa2e84b 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -93,3 +93,16 @@ inline float4 scs(float4 val) result.y = sinf(tmp); return result; } + + +inline float4 vsin(float4 val) +{ + float4 result; + float tmp = val.x; + float res = sinf(tmp); + result.x = res; + result.y = res; + result.z = res; + result.w = res; + return result; +} -- cgit v1.2.3 From 16fe6a0f2d53e63884dcecaf8ba61e0105a92c99 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 2 Nov 2007 12:18:02 -0400 Subject: Cleanups. Remove some debugging output and try to make sure that Mesa compiles when configured without LLVM --- src/mesa/pipe/softpipe/sp_quad_fs.c | 11 ++++++++--- src/mesa/pipe/softpipe/sp_state_fs.c | 2 -- 2 files changed, 8 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 2cecfb268a..17fb10ea72 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -40,7 +40,9 @@ #include "x86/rtasm/x86sse.h" +#ifdef MESA_LLVM #include "pipe/llvm/gallivm.h" +#endif #include "sp_context.h" #include "sp_state.h" @@ -56,7 +58,9 @@ struct quad_shade_stage struct tgsi_exec_machine machine; struct tgsi_exec_vector *inputs, *outputs; int colorOutSlot, depthOutSlot; +#ifdef MESA_LLVM struct gallivm_prog *llvm_prog; +#endif }; @@ -120,9 +124,6 @@ shade_quad( else #endif { -#ifdef MESA_LLVM - /*ga_llvm_prog_exec(softpipe->fs->llvm_prog);*/ -#endif quad->mask &= tgsi_exec_machine_run( machine ); } @@ -159,6 +160,7 @@ shade_quad( } } +#ifdef MESA_LLVM #define DLLVM 0 static void shade_quad_llvm(struct quad_stage *qs, @@ -256,6 +258,7 @@ shade_quad_llvm(struct quad_stage *qs, qs->next->run( qs->next, quad ); } } +#endif /*MESA_LLVM*/ /** * Per-primitive (or per-begin?) setup @@ -272,7 +275,9 @@ static void shade_begin(struct quad_stage *qs) qss->samplers[i].texture = softpipe->texture[i]; } +#ifdef MESA_LLVM qss->llvm_prog = softpipe->fs->llvm_prog; +#endif /* XXX only do this if the fragment shader changes... */ tgsi_exec_machine_init(&qss->machine, softpipe->fs->shader.tokens, diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 08a7b58a6f..ba564b16e6 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -62,14 +62,12 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, #endif #ifdef MESA_LLVM - fprintf(stderr, "+++++++++++++++++++++++++++++++++++++++++++++++++\n"); state->llvm_prog = gallivm_from_tgsi(state->shader.tokens, GALLIVM_FS); if (!gallivm_global_cpu_engine()) { gallivm_cpu_engine_create(state->llvm_prog); } else gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); - fprintf(stderr, "+++++++++++++++++++++++++++++++++++++++++++++++++\n"); #endif return state; -- cgit v1.2.3 From 3c393b8df302493c4f48a750bfd7bd1c6aadbabb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 1 Nov 2007 18:14:44 +0000 Subject: Render primitives using indirect vertices in a vertex buffer. --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 74 ++++++++++++++++--------------- 1 file changed, 39 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index dc6e091681..6f497879ad 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -110,7 +110,7 @@ static boolean check_space( struct vbuf_stage *vbuf ) { if (overflow( vbuf->vertex_map, vbuf->vertex_ptr, - 4 * vbuf->vertex_size, + vbuf->vertex_size, VBUF_SIZE )) return FALSE; @@ -274,12 +274,11 @@ static void vbuf_draw( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); struct i915_context *i915 = vbuf->i915; - struct pipe_winsys *winsys = i915->pipe.winsys; unsigned nr = vbuf->nr_elements; unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ unsigned hwprim; unsigned i; - char *ptr; + unsigned *ptr; switch(vbuf->prim) { case PIPE_PRIM_POINTS: @@ -295,40 +294,20 @@ static void vbuf_draw( struct draw_stage *stage ) assert(0); return; } - + assert(vbuf->vertex_ptr - vbuf->vertex_map == vbuf->nr_vertices * vertex_size / 4); - /* FIXME: handle failure */ - if(!vbuf->buf) - vbuf->buf = winsys->buffer_create(winsys, 64); - - winsys->buffer_data(winsys, vbuf->buf, 8 + nr * vertex_size, NULL); - ptr = winsys->buffer_map(winsys, vbuf->buf, PIPE_BUFFER_FLAG_WRITE); - *(unsigned *)ptr = _3DPRIMITIVE | - hwprim | - ((4 + vertex_size * nr)/4 - 2); - ptr += 4; - for (i = 0; i < nr; i++) { - memcpy(ptr, - (char*)vbuf->vertex_map + vbuf->element_map[i]*vertex_size, - vertex_size ); - ptr += vertex_size; - } - *(unsigned *)ptr = MI_BATCH_BUFFER_END; - ptr += 4; - winsys->buffer_unmap(winsys, vbuf->buf); - if (i915->dirty) i915_update_derived( i915 ); if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); - ptr = BEGIN_BATCH( 2, 1 ); + ptr = BEGIN_BATCH( 4 + (nr + 1)/2, 1 ); #if 1 assert(ptr); #else - /* XXX: below is bogues as ptr always nonzero except in fatal errors */ + /* XXX: below is bogus as ptr always nonzero except in fatal errors */ if (ptr == 0) { FLUSH_BATCH(); @@ -344,20 +323,35 @@ static void vbuf_draw( struct draw_stage *stage ) } } #endif - - /* chain the vertex buffer in the batch buffer */ - OUT_BATCH(MI_BATCH_BUFFER_START - | (2 << 6) /* GTT-mapped memory */); + + /* FIXME: don't do this every time */ + OUT_BATCH( _3DSTATE_LOAD_STATE_IMMEDIATE_1 | + I1_LOAD_S(0) | + I1_LOAD_S(1) | + (1)); OUT_RELOC( vbuf->buf, I915_BUFFER_ACCESS_READ, 0 ); - /* FIXME: we need to flush here since control after chained buffers returns - * directly to the ring buffer */ - FLUSH_BATCH(); + OUT_BATCH( ((vertex_size/4) << 24) | /* vertex size in dwords */ + ((vertex_size/4) << 16) ); /* vertex pitch in dwords */ + OUT_BATCH( _3DPRIMITIVE | + PRIM_INDIRECT | + hwprim | + PRIM_INDIRECT_ELTS | + nr ); + for (i = 0; i + 1 < nr; i += 2) { + OUT_BATCH( vbuf->element_map[i] | + (vbuf->element_map[i + 1] << 16) ); + } + if (i < nr) { + OUT_BATCH( vbuf->element_map[i] ); + } } static void vbuf_flush_elements( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); + struct i915_context *i915 = vbuf->i915; + struct pipe_winsys *winsys = i915->pipe.winsys; if (vbuf->nr_elements) { #if 0 @@ -373,7 +367,8 @@ static void vbuf_flush_elements( struct draw_stage *stage ) vbuf->nr_elements = 0; - vbuf->vertex_ptr = vbuf->vertex_map; + winsys->buffer_unmap(winsys, vbuf->buf); + vbuf->nr_vertices = 0; /* Reset vertex ids? Actually, want to not do that unless our @@ -385,6 +380,15 @@ static void vbuf_flush_elements( struct draw_stage *stage ) draw_vertex_cache_reset_vertex_ids( vbuf->i915->draw ); } + /* FIXME: handle failure */ + if(!vbuf->buf) + vbuf->buf = winsys->buffer_create(winsys, 64); + winsys->buffer_data(winsys, vbuf->buf, VBUF_SIZE, NULL); + vbuf->vertex_map = winsys->buffer_map(winsys, + vbuf->buf, + PIPE_BUFFER_FLAG_WRITE ); + vbuf->vertex_ptr = vbuf->vertex_map; + stage->tri = vbuf_first_tri; stage->line = vbuf_first_line; stage->point = vbuf_first_point; @@ -430,7 +434,7 @@ struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) /* FIXME: free this memory on takedown */ vbuf->element_map = malloc( IBUF_SIZE ); - vbuf->vertex_map = malloc( VBUF_SIZE ); + vbuf->vertex_map = NULL; vbuf->vertex_ptr = vbuf->vertex_map; -- cgit v1.2.3 From 5c1606a2b3e951c32f028e0b328e6c06e9424e28 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 3 Nov 2007 01:08:15 +0000 Subject: Detail i915 winsys interface comments. --- src/mesa/pipe/i915simple/i915_winsys.h | 58 +++++++++++++++++++++++----------- 1 file changed, 40 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 544763644c..947c5a334d 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -25,15 +25,17 @@ * **************************************************************************/ -#ifndef I915_WINSYS_H -#define I915_WINSYS_H - - -/* This is the interface that softpipe requires any window system - * hosting it to implement. This is the only include file in softpipe +/** + * \file + * This is the interface that i915simple requires any window system + * hosting it to implement. This is the only include file in i915simple * which is public. + * */ +#ifndef I915_WINSYS_H +#define I915_WINSYS_H + /* Pipe drivers are (meant to be!) independent of both GL and the * window system. The window system provides a buffer manager and a @@ -48,31 +50,51 @@ struct pipe_buffer_handle; struct pipe_winsys; + +/** + * Additional winsys interface for i915simple. + * + * It is an over-simple batchbuffer mechanism. Will want to improve the + * performance of this, perhaps based on the cmdstream stuff. It + * would be pretty impossible to implement swz on top of this + * interface. + * + * Will also need additions/changes to implement static/dynamic + * indirect state. + */ struct i915_winsys { - /* An over-simple batchbuffer mechanism. Will want to improve the - * performance of this, perhaps based on the cmdstream stuff. It - * would be pretty impossible to implement swz on top of this - * interface. - * - * Will also need additions/changes to implement static/dynamic - * indirect state. + /** + * Reserve space on batch buffer. + * + * Returns a null pointer if there is insufficient space in the batch buffer + * to hold the requested number of dwords and relocations. + * + * The number of dwords should also include the number of relocations. */ unsigned *(*batch_start)( struct i915_winsys *sws, unsigned dwords, unsigned relocs ); + void (*batch_dword)( struct i915_winsys *sws, unsigned dword ); + + /** + * Emit a relocation to a buffer. + * + * Used not only when the buffer addresses are not pinned, but also to + * ensure refered buffers will not be destroyed until the current batch + * buffer execution is finished. + * + * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and + * I915_BUFFER_ACCESS_READ macros. + */ void (*batch_reloc)( struct i915_winsys *sws, struct pipe_buffer_handle *buf, unsigned access_flags, unsigned delta ); + void (*batch_flush)( struct i915_winsys *sws ); - -#if 0 - void (*batch_chain)( struct i915_winsys *sws, - struct pipe_buffer_handle *buf ): -#endif }; #define I915_BUFFER_ACCESS_WRITE 0x1 -- cgit v1.2.3 From 5fa6ea68586e906a984291dd4c20f664924157eb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 3 Nov 2007 01:31:47 +0000 Subject: Do something sensible when failed to reserve space in the batch buffer. --- src/mesa/pipe/i915simple/i915_blit.c | 10 ++++++++-- src/mesa/pipe/i915simple/i915_flush.c | 5 ++++- src/mesa/pipe/i915simple/i915_prim_emit.c | 7 ++----- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 11 ++--------- src/mesa/pipe/i915simple/i915_state_emit.c | 2 +- 5 files changed, 17 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/mesa/pipe/i915simple/i915_blit.c index 0a8a1542d7..a1f953ebbe 100644 --- a/src/mesa/pipe/i915simple/i915_blit.c +++ b/src/mesa/pipe/i915simple/i915_blit.c @@ -69,7 +69,10 @@ i915_fill_blit(struct i915_context *i915, // __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); - BEGIN_BATCH(6, 1); + if (!BEGIN_BATCH(6, 1)) { + FLUSH_BATCH(); + assert(BEGIN_BATCH(6, 1)); + } OUT_BATCH(CMD); OUT_BATCH(BR13); OUT_BATCH((y << 16) | x); @@ -140,7 +143,10 @@ i915_copy_blit( struct i915_context *i915, assert (dst_pitch > 0 && src_pitch > 0); - BEGIN_BATCH(8, 2); + if (!BEGIN_BATCH(8, 2)) { + FLUSH_BATCH(); + assert(BEGIN_BATCH(8, 2)); + } OUT_BATCH(CMD); OUT_BATCH(BR13); OUT_BATCH((dst_y << 16) | dst_x); diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c index 9a31342cbd..9c2adf8763 100644 --- a/src/mesa/pipe/i915simple/i915_flush.c +++ b/src/mesa/pipe/i915simple/i915_flush.c @@ -56,7 +56,10 @@ static void i915_flush( struct pipe_context *pipe, if (flags & PIPE_FLUSH_TEXTURE_CACHE) flush |= FLUSH_MAP_CACHE; - BEGIN_BATCH( 1, 0 ); + if (!BEGIN_BATCH(1, 0)) { + FLUSH_BATCH(); + assert(BEGIN_BATCH(1, 0)); + } OUT_BATCH( flush ); ADVANCE_BATCH(); } diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index 40e33e6e09..0a8be79ae4 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -123,7 +123,6 @@ emit_prim( struct draw_stage *stage, { struct i915_context *i915 = setup_stage(stage)->i915; unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ - unsigned *ptr; unsigned i; assert(vertex_size >= 12); /* never smaller than 12 bytes */ @@ -134,8 +133,7 @@ emit_prim( struct draw_stage *stage, if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); - ptr = BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 ); - if (ptr == 0) { + if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) { FLUSH_BATCH(); /* Make sure state is re-emitted after a flush: @@ -143,8 +141,7 @@ emit_prim( struct draw_stage *stage, i915_update_derived( i915 ); i915_emit_hardware_state( i915 ); - ptr = BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 ); - if (ptr == 0) { + if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) { assert(0); return; } diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 6f497879ad..3632adce79 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -303,12 +303,7 @@ static void vbuf_draw( struct draw_stage *stage ) if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); - ptr = BEGIN_BATCH( 4 + (nr + 1)/2, 1 ); -#if 1 - assert(ptr); -#else - /* XXX: below is bogus as ptr always nonzero except in fatal errors */ - if (ptr == 0) { + if (!BEGIN_BATCH( 4 + (nr + 1)/2, 1 )) { FLUSH_BATCH(); /* Make sure state is re-emitted after a flush: @@ -316,13 +311,11 @@ static void vbuf_draw( struct draw_stage *stage ) i915_update_derived( i915 ); i915_emit_hardware_state( i915 ); - ptr = BEGIN_BATCH( 2, 1 ); - if (ptr == 0) { + if (!BEGIN_BATCH( 4 + (nr + 1)/2, 1 )) { assert(0); return; } } -#endif /* FIXME: don't do this every time */ OUT_BATCH( _3DSTATE_LOAD_STATE_IMMEDIATE_1 | diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 15eff97813..29fe9d9718 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -81,7 +81,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) if(!BEGIN_BATCH(dwords, relocs)) { FLUSH_BATCH(); - BEGIN_BATCH(dwords, relocs); + assert(BEGIN_BATCH(dwords, relocs)); } /* 14 dwords, 0 relocs */ -- cgit v1.2.3 From 7a7899a2476592e846b908a557a738a49fa9a948 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 3 Nov 2007 02:00:28 +0000 Subject: Allow batch buffer debugging without calling i915_winsys->batch_start so often. --- src/mesa/pipe/i915simple/i915_batch.h | 10 +++++----- src/mesa/pipe/i915simple/i915_context.c | 2 +- src/mesa/pipe/i915simple/i915_debug.c | 12 ++++++++---- src/mesa/pipe/i915simple/i915_debug.h | 4 +--- 4 files changed, 15 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h index d0b0946cdf..fb88cd6db0 100644 --- a/src/mesa/pipe/i915simple/i915_batch.h +++ b/src/mesa/pipe/i915simple/i915_batch.h @@ -34,7 +34,7 @@ #define BATCH_LOCALS #define BEGIN_BATCH( dwords, relocs ) \ - i915->winsys->batch_start( i915->winsys, dwords, relocs ) + (i915->batch_start = i915->winsys->batch_start( i915->winsys, dwords, relocs )) #define OUT_BATCH( dword ) \ i915->winsys->batch_dword( i915->winsys, dword ) @@ -44,10 +44,10 @@ #define ADVANCE_BATCH() -#define FLUSH_BATCH() do { \ - if (0) i915_dump_batchbuffer( i915, i915->batch_start, BEGIN_BATCH(0, 0) ); \ - i915->winsys->batch_flush( i915->winsys ); \ - i915->batch_start = BEGIN_BATCH(0, 0); \ +#define FLUSH_BATCH() do { \ + if (0) i915_dump_batchbuffer( i915 ); \ + i915->winsys->batch_flush( i915->winsys ); \ + i915->batch_start = NULL; \ i915->hardware_dirty = ~0; \ } while (0) diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index fc878c175c..2c36a194c7 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -347,7 +347,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, /* Batch stream debugging is a bit hacked up at the moment: */ - i915->batch_start = BEGIN_BATCH(0, 0); + i915->batch_start = NULL; /* * XXX we could plug GL selection/feedback into the drawing pipeline diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index d142194d84..d07d2f1fa9 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -822,20 +822,24 @@ static boolean i915_debug_packet( struct debug_stream *stream ) void -i915_dump_batchbuffer( struct i915_context *i915, - unsigned *start, - unsigned *end ) +i915_dump_batchbuffer( struct i915_context *i915 ) { struct debug_stream stream; + unsigned *start = i915->batch_start; + unsigned *end = i915->winsys->batch_start( i915->winsys, 0, 0 ); unsigned bytes = (end - start) * 4; boolean done = FALSE; - stream.offset = 0; stream.ptr = (char *)start; stream.print_addresses = 0; stream.winsys = i915->pipe.winsys; + if (!start || !end) { + stream.winsys->printf( stream.winsys, "\n\nBATCH: ???\n"); + return; + } + stream.winsys->printf( stream.winsys, "\n\nBATCH: (%d)\n", bytes / 4); while (!done && diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h index 0ea131171e..4c3aa64b42 100644 --- a/src/mesa/pipe/i915simple/i915_debug.h +++ b/src/mesa/pipe/i915simple/i915_debug.h @@ -79,9 +79,7 @@ void i915_print_ureg(const char *msg, unsigned ureg); #endif -void i915_dump_batchbuffer( struct i915_context *i915, - unsigned *start, - unsigned *end ); +void i915_dump_batchbuffer( struct i915_context *i915 ); -- cgit v1.2.3 From fda91cfa4b7b7868172a563da49cb0d7ba6cf5e0 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 4 Nov 2007 16:38:36 -0700 Subject: Check if the user/texenvprogram is just a pass-through program and skip program concatenation. --- src/mesa/state_tracker/st_cb_drawpixels.c | 39 +++++++++++++++++++++++++++---- 1 file changed, 35 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index ffaa34d7d3..fc58035d00 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -57,6 +57,30 @@ #include "shader/prog_instruction.h" +/** + * Check if the given program is: + * 0: MOVE result.color, fragment.color; + * 1: END; + */ +static GLboolean +is_passthrough_program(const struct gl_fragment_program *prog) +{ + if (prog->Base.NumInstructions == 2) { + const struct prog_instruction *inst = prog->Base.Instructions; + if (inst[0].Opcode == OPCODE_MOV && + inst[1].Opcode == OPCODE_END && + inst[0].DstReg.File == PROGRAM_OUTPUT && + inst[0].DstReg.Index == FRAG_RESULT_COLR && + inst[0].DstReg.WriteMask == WRITEMASK_XYZW && + inst[0].SrcReg[0].File == PROGRAM_INPUT && + inst[0].SrcReg[0].Index == FRAG_ATTRIB_COL0 && + inst[0].SrcReg[0].Swizzle == SWIZZLE_XYZW) { + return GL_TRUE; + } + } + return GL_FALSE; +} + /** * Make fragment program for glBitmap: @@ -215,14 +239,21 @@ combined_drawpix_fragment_program(GLcontext *ctx) /* Concatenate the pixel transfer program with the current user- * defined program. */ - stfp = (struct st_fragment_program *) - _mesa_combine_programs(ctx, - &st->pixel_xfer.program->Base.Base, - &st->fp->Base.Base); + if (is_passthrough_program(&st->fp->Base)) { + stfp = (struct st_fragment_program *) + _mesa_clone_program(ctx, &st->pixel_xfer.program->Base.Base); + } + else { + stfp = (struct st_fragment_program *) + _mesa_combine_programs(ctx, + &st->pixel_xfer.program->Base.Base, + &st->fp->Base.Base); + } #if 0 { struct gl_program *p = &stfp->Base.Base; + printf("Combined DrawPixels program:\n"); _mesa_print_program(p); printf("InputsRead: 0x%x\n", p->InputsRead); printf("OutputsWritten: 0x%x\n", p->OutputsWritten); -- cgit v1.2.3 From 17d044ec019039e1470004a213584c014ba30c62 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 10:58:38 -0600 Subject: more clean-up in intelDisplayBuffer() --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 28 +++++++++++----------- 1 file changed, 14 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 350c3f38a2..a366e52517 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -86,9 +86,9 @@ intelDisplayBuffer(__DRIdrawablePrivate * dPriv, struct pipe_surface *surf, const drm_clip_rect_t * rect) { - struct intel_context *intel; - const intelScreenPrivate *intelScreen; + const intelScreenPrivate *intelScreen + = (intelScreenPrivate *) dPriv->driScreenPriv->private; DBG(SWAP, "%s\n", __FUNCTION__); @@ -98,8 +98,6 @@ intelDisplayBuffer(__DRIdrawablePrivate * dPriv, if (!intel) return; - intelScreen = intel->intelScreen; - if (intel->last_swap_fence) { driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); driFenceUnReference(intel->last_swap_fence); @@ -120,9 +118,8 @@ intelDisplayBuffer(__DRIdrawablePrivate * dPriv, if (dPriv && dPriv->numClipRects) { - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - const int backWidth = intel_fb->Base.Width; - const int backHeight = intel_fb->Base.Height; + const int srcWidth = surf->width; + const int srcHeight = surf->height; const int nbox = dPriv->numClipRects; const drm_clip_rect_t *pbox = dPriv->pClipRects; const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; @@ -132,12 +129,10 @@ intelDisplayBuffer(__DRIdrawablePrivate * dPriv, const struct pipe_region *srcRegion = surf->region; const int srcpitch= srcRegion->pitch; - ASSERT(intel_fb); - ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ ASSERT(srcRegion); ASSERT(srcRegion->cpp == cpp); - DBG(SWAP, "front pitch %d back pitch %d\n", + DBG(SWAP, "screen pitch %d src surface pitch %d\n", pitch, srcRegion->pitch); if (cpp == 2) { @@ -186,10 +181,10 @@ intelDisplayBuffer(__DRIdrawablePrivate * dPriv, } /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > backWidth) - box.x2 = backWidth + box.x1; - if (box.y2 - box.y1 > backHeight) - box.y2 = backHeight + box.y1; + if (box.x2 - box.x1 > srcWidth) + box.x2 = srcWidth + box.x1; + if (box.y2 - box.y1 > srcHeight) + box.y2 = srcHeight + box.y1; DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", box.x1, box.x2, box.y1, box.y2); @@ -611,6 +606,11 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) } } + +/** + * Called via glXCopySubBufferMESA() to copy a subrect of the back + * buffer to the front buffer/screen. + */ void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) { -- cgit v1.2.3 From 00a133ad19a64e15c78314157a1a7102529441b3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 11:27:23 -0600 Subject: remove some dead code, clean-ups --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 25 ++++------------------ 1 file changed, 4 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index a366e52517..986a6f7112 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -243,12 +243,6 @@ intelWindowMoved(struct intel_context *intel) __DRIdrawablePrivate *dPriv = intel->driDrawable; struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - if (!intel->st->ctx->DrawBuffer) { - /* when would this happen? -BP */ - assert(0); - intel->numClipRects = 0; - } - /* Update Mesa's notion of window size */ intelUpdateFramebufferSize(ctx, dPriv); intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ @@ -257,10 +251,12 @@ intelWindowMoved(struct intel_context *intel) drmI830Sarea *sarea = intel->sarea; drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; - drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, .y1 = sarea->pipeA_y, + drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, + .y1 = sarea->pipeA_y, .x2 = sarea->pipeA_x + sarea->pipeA_w, .y2 = sarea->pipeA_y + sarea->pipeA_h }; - drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, .y1 = sarea->pipeB_y, + drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, + .y1 = sarea->pipeB_y, .x2 = sarea->pipeB_x + sarea->pipeB_w, .y2 = sarea->pipeB_y + sarea->pipeB_h }; GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); @@ -326,10 +322,6 @@ intelWindowMoved(struct intel_context *intel) } intel_fb->pf_active = pf_active; -#if 0 - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); -#endif /* Update vblank info */ @@ -368,15 +360,6 @@ intelWindowMoved(struct intel_context *intel) } } - /* This will be picked up by looking at the dirty state flags: - */ - - /* Update hardware scissor */ -// ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, -// ctx->Scissor.Width, ctx->Scissor.Height); - - /* Re-calculate viewport related state */ -// ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far ); } -- cgit v1.2.3 From 381b68d80413dba7e88f03057f720c0fe2b0ac2e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 11:27:39 -0600 Subject: comment about renderbuffers vs. surfaces --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index e6bd87a4ea..be31774904 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -295,6 +295,11 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); + /* + * XXX Create pipe_surfaces for front/back buffers, + * hand them to state tracker to create a framebuffer object. + */ + { /* fake frontbuffer */ /* XXX allocation should only happen in the unusual case -- cgit v1.2.3 From 09771f8c708ceab5956a8caf1483c5d3d30f9020 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 11:28:02 -0600 Subject: #include clean-ups, fixes --- src/mesa/drivers/dri/intel_winsys/intel_context.h | 3 --- src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c | 1 + src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c | 1 + 3 files changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 281d68cc7e..0686162ebe 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -32,12 +32,9 @@ #include "mtypes.h" #include "drm.h" -#include "mm.h" -#include "texmem.h" #include "intel_screen.h" #include "i915_drm.h" -#include "i830_common.h" struct pipe_context; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index e07fadb278..022fb66927 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -38,6 +38,7 @@ #include "intel_batchbuffer.h" #include "intel_winsys.h" +#include "pipe/p_util.h" #include "pipe/i915simple/i915_winsys.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 7f788b8537..9c3fd56d8d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -42,6 +42,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_defines.h" #include "pipe/p_state.h" +#include "pipe/p_util.h" -- cgit v1.2.3 From f6427e35572f857567f4459ad0501babbecef2f7 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 11:31:39 -0600 Subject: remove extern decls for non-existant functions --- src/mesa/drivers/dri/intel_winsys/intel_context.h | 7 ------- 1 file changed, 7 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 0686162ebe..762335a004 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -143,9 +143,6 @@ extern int __intel_debug; * intel_context.c: */ -extern void intelGetLock(struct intel_context *intel, GLuint flags); - -extern void intelFinish(GLcontext * ctx); extern void intelFlush(GLcontext * ctx); /*====================================================================== @@ -158,10 +155,6 @@ intel_context(GLcontext * ctx) return (struct intel_context *) ctx->DriverCtx; } -extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer - *rb); - -extern void intel_init_region_functions(struct pipe_context *pipe); extern void intelUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv); -- cgit v1.2.3 From 4019277f09448a0f7ffb7dd620e9bc5613f9b758 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 11:41:32 -0600 Subject: public st_flush() --- src/mesa/state_tracker/st_cb_flush.c | 33 +++++++++++++++++++++++---------- src/mesa/state_tracker/st_public.h | 2 ++ 2 files changed, 25 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index c2c3c80b87..fe920a44ab 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -36,14 +36,15 @@ #include "st_context.h" #include "st_cb_flush.h" #include "st_cb_fbo.h" +#include "st_public.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -static void st_flush(GLcontext *ctx) +void st_flush( struct st_context *st ) { - struct st_context *st = ctx->st; + GLframebuffer *fb = st->ctx->DrawBuffer; /* If there has been no rendering to the frontbuffer, consider * short-circuiting this, or perhaps pass an "optional" flag down @@ -55,13 +56,12 @@ static void st_flush(GLcontext *ctx) /* XXX: temporary hack. This flag should only be set if we do any * rendering to the front buffer. */ - st->flags.frontbuffer_dirty = (ctx->DrawBuffer->_ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT); - + st->flags.frontbuffer_dirty + = (fb->_ColorDrawBufferMask[0] & BUFFER_BIT_FRONT_LEFT); if (st->flags.frontbuffer_dirty) { struct st_renderbuffer *strb - = st_renderbuffer(ctx->DrawBuffer->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); struct pipe_surface *front_surf = strb->surface; /* Hook for copying "fake" frontbuffer if necessary: @@ -71,17 +71,30 @@ static void st_flush(GLcontext *ctx) } } -static void st_finish(GLcontext *ctx) + +/** + * Called via ctx->Driver.Flush() + */ +static void st_Flush(GLcontext *ctx) +{ + st_flush(ctx->st); +} + + +/** + * Called via ctx->Driver.Finish() + */ +static void st_Finish(GLcontext *ctx) { struct st_context *st = ctx->st; - st_flush( ctx ); + st_flush( st ); st->pipe->winsys->wait_idle( st->pipe->winsys ); } void st_init_flush_functions(struct dd_function_table *functions) { - functions->Flush = st_flush; - functions->Finish = st_finish; + functions->Flush = st_Flush; + functions->Finish = st_Finish; } diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 3056b5a3e7..1cc6e6e40e 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -46,4 +46,6 @@ void st_destroy_context2( struct st_context *st ); void st_invalidate_state(GLcontext * ctx, GLuint new_state); +void st_flush( struct st_context *st ); + #endif -- cgit v1.2.3 From 08f88cbbc9a5da5667d02d5eeafac79d2de1ee95 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 11:43:33 -0600 Subject: call FLUSH_VERTICES() in st_flush() --- src/mesa/state_tracker/st_cb_flush.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index fe920a44ab..5af391d8ed 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -46,6 +46,8 @@ void st_flush( struct st_context *st ) { GLframebuffer *fb = st->ctx->DrawBuffer; + FLUSH_VERTICES(st->ctx, 0); + /* If there has been no rendering to the frontbuffer, consider * short-circuiting this, or perhaps pass an "optional" flag down * to the driver so that it can make the decision. -- cgit v1.2.3 From 156e490699144bc6f3a8706a3a1b2d1f3c35f029 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 11:44:07 -0600 Subject: replace IntelFlush(), FLUSH_VERTICES(), etc. with st_flush() call --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 20 +------------------- src/mesa/drivers/dri/intel_winsys/intel_context.h | 6 ------ .../drivers/dri/intel_winsys/intel_swapbuffers.c | 2 +- 3 files changed, 2 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index dc6a02999e..b0c1a608fa 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -147,20 +147,6 @@ static const struct dri_debug_control debug_control[] = { #endif -void -intelFlush(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - - /* Hmm: - */ - intel->pipe->flush( intel->pipe, 0 ); -} - - - - - static void intelInitDriverFunctions(struct dd_function_table *functions) { @@ -344,11 +330,7 @@ GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv) { struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; - /* XXX UnbindContext is called AFTER the new context is made current. - Hopefully shouldn't be a problem ? */ - GLcontext *ctx = intel->st->ctx; - FLUSH_VERTICES(ctx, 0); - intelFlush(ctx); + st_flush(intel->st); return GL_TRUE; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 762335a004..788d444d00 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -139,12 +139,6 @@ extern int __intel_debug; #define PCI_CHIP_Q33_G 0x29D2 -/* ================================================================ - * intel_context.c: - */ - -extern void intelFlush(GLcontext * ctx); - /*====================================================================== * Inline conversion functions. * These are better-typed than the macros used previously: diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 986a6f7112..7629a0fe29 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -417,7 +417,7 @@ intelPageFlip(const __DRIdrawablePrivate * dPriv) if (intel->intelScreen->drmMinor < 9) return GL_FALSE; - intelFlush(&intel->ctx); + st_flush(intel->st); ret = 0; -- cgit v1.2.3 From 0ff447e7c4ae26a1c0ae6f92265dee4561816832 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:11:40 -0600 Subject: include context.h --- src/mesa/state_tracker/st_cb_flush.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 5af391d8ed..eb47e75b40 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -33,6 +33,7 @@ #include "main/glheader.h" #include "main/macros.h" +#include "main/context.h" #include "st_context.h" #include "st_cb_flush.h" #include "st_cb_fbo.h" -- cgit v1.2.3 From b8897d7481bb27974ccaedbe96d96fce77c1cfc8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:14:01 -0600 Subject: disable page flip code --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 30 +++++++++++++++++----- .../drivers/dri/intel_winsys/intel_swapbuffers.h | 3 +++ 2 files changed, 27 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 7629a0fe29..aadd6f5665 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -262,11 +262,13 @@ intelWindowMoved(struct intel_context *intel) GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); GLuint flags = intel_fb->vblank_flags; +#if PF GLboolean pf_active; GLint pf_planes; - +#endif /* Update page flipping info */ +#if PF pf_planes = 0; if (areaA > 0) @@ -274,7 +276,9 @@ intelWindowMoved(struct intel_context *intel) if (areaB > 0) pf_planes |= 2; +#endif +#if PF intel_fb->pf_current_page = (intel->sarea->pf_current_page >> (intel_fb->pf_planes & 0x2)) & 0x3; @@ -322,6 +326,7 @@ intelWindowMoved(struct intel_context *intel) } intel_fb->pf_active = pf_active; +#endif /* Update vblank info */ @@ -334,14 +339,16 @@ intelWindowMoved(struct intel_context *intel) if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { drmVBlank vbl; +#if PF int i; - +#endif vbl.request.type = DRM_VBLANK_ABSOLUTE; if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { vbl.request.type |= DRM_VBLANK_SECONDARY; } +#if PF for (i = 0; i < intel_fb->pf_num_pages; i++) { if ((intel_fb->vbl_waited - intel_fb->vbl_pending[i]) <= (1<<23)) continue; @@ -349,14 +356,16 @@ intelWindowMoved(struct intel_context *intel) vbl.request.sequence = intel_fb->vbl_pending[i]; drmWaitVBlank(intel->driFd, &vbl); } - +#endif intel_fb->vblank_flags = flags; driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); intel_fb->vbl_waited = intel_fb->vbl_seq; +#if PF for (i = 0; i < intel_fb->pf_num_pages; i++) { intel_fb->vbl_pending[i] = intel_fb->vbl_waited; } +#endif } } @@ -468,15 +477,21 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); struct intel_context *intel = intelScreenContext(dPriv->driScreenPriv->private); +#if PF const intelScreenPrivate *intelScreen = intel->intelScreen; +#endif unsigned int target; drm_i915_vblank_swap_t swap; GLboolean ret; /* XXX: Scheduled buffer swaps don't work with private back buffers yet */ if (1 || !intel_fb->vblank_flags || - (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || - intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) + (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) +#if PF + || + intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6) +#endif +) return GL_FALSE; swap.seqtype = DRM_VBLANK_ABSOLUTE; @@ -498,6 +513,7 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) intel_batchbuffer_flush(intel->batch); +#if PF if ( intel_fb->pf_active ) { swap.seqtype |= DRM_VBLANK_FLIP; @@ -505,6 +521,7 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) (intel_fb->pf_planes & 0x2)) & 0x3) + 1) % intel_fb->pf_num_pages; } +#endif if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, sizeof(swap))) { @@ -530,12 +547,13 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) ret = GL_TRUE; } else { +#if PF if (swap.seqtype & DRM_VBLANK_FLIP) { intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> (intel_fb->pf_planes & 0x2)) & 0x3) % intel_fb->pf_num_pages; } - +#endif ret = GL_FALSE; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index 15ed6704d2..b7343db5e2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -39,12 +39,15 @@ struct intel_framebuffer { struct gl_framebuffer Base; +#define PF 0 +#if PF /* Drawable page flipping state */ GLboolean pf_active; GLuint pf_seq; GLint pf_planes; GLint pf_current_page; GLint pf_num_pages; +#endif /* VBI */ -- cgit v1.2.3 From 29feee2c02a7a558a5448434904991c1dd6eb19f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:17:13 -0600 Subject: remove dead pageflip code --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 230 +-------------------- .../drivers/dri/intel_winsys/intel_swapbuffers.h | 12 -- 2 files changed, 2 insertions(+), 240 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index aadd6f5665..ac4ae80cfb 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -41,22 +41,6 @@ #include "state_tracker/st_cb_fbo.h" -/* This block can be removed when libdrm >= 2.3.1 is required */ - -#ifndef DRM_VBLANK_FLIP - -#define DRM_VBLANK_FLIP 0x8000000 - -typedef struct drm_i915_flip { - int pipes; -} drm_i915_flip_t; - -#undef DRM_IOCTL_I915_FLIP -#define DRM_IOCTL_I915_FLIP DRM_IOW(DRM_COMMAND_BASE + DRM_I915_FLIP, \ - drm_i915_flip_t) - -#endif - /** * Return the pipe_surface for the given renderbuffer. @@ -262,71 +246,6 @@ intelWindowMoved(struct intel_context *intel) GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); GLuint flags = intel_fb->vblank_flags; -#if PF - GLboolean pf_active; - GLint pf_planes; -#endif - /* Update page flipping info - */ -#if PF - pf_planes = 0; - - if (areaA > 0) - pf_planes |= 1; - - if (areaB > 0) - pf_planes |= 2; -#endif - -#if PF - intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_planes & 0x2)) & 0x3; - - intel_fb->pf_num_pages = 2 /*intel->intelScreen->third.handle ? 3 : 2*/; - - pf_active = pf_planes && (pf_planes & intel->sarea->pf_active) == pf_planes; - - if (pf_active != intel_fb->pf_active) - DBG(LOCK, "%s - Page flipping %sactive\n", - __progname, pf_active ? "" : "in"); - - if (pf_active) { - /* Sync pages between planes if we're flipping on both at the same time */ - if (pf_planes == 0x3 && pf_planes != intel_fb->pf_planes && - (intel->sarea->pf_current_page & 0x3) != - (((intel->sarea->pf_current_page) >> 2) & 0x3)) { - drm_i915_flip_t flip; - - if (intel_fb->pf_current_page == - (intel->sarea->pf_current_page & 0x3)) { - /* XXX: This is ugly, but emitting two flips 'in a row' can cause - * lockups for unknown reasons. - */ - intel->sarea->pf_current_page = - intel->sarea->pf_current_page & 0x3; - intel->sarea->pf_current_page |= - ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % - intel_fb->pf_num_pages) << 2; - - flip.pipes = 0x2; - } else { - intel->sarea->pf_current_page = - intel->sarea->pf_current_page & (0x3 << 2); - intel->sarea->pf_current_page |= - (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % - intel_fb->pf_num_pages; - - flip.pipes = 0x1; - } - - drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); - } - - intel_fb->pf_planes = pf_planes; - } - - intel_fb->pf_active = pf_active; -#endif /* Update vblank info */ @@ -339,33 +258,17 @@ intelWindowMoved(struct intel_context *intel) if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { drmVBlank vbl; -#if PF - int i; -#endif + vbl.request.type = DRM_VBLANK_ABSOLUTE; if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { vbl.request.type |= DRM_VBLANK_SECONDARY; } -#if PF - for (i = 0; i < intel_fb->pf_num_pages; i++) { - if ((intel_fb->vbl_waited - intel_fb->vbl_pending[i]) <= (1<<23)) - continue; - - vbl.request.sequence = intel_fb->vbl_pending[i]; - drmWaitVBlank(intel->driFd, &vbl); - } -#endif intel_fb->vblank_flags = flags; driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); intel_fb->vbl_waited = intel_fb->vbl_seq; -#if PF - for (i = 0; i < intel_fb->pf_num_pages; i++) { - intel_fb->vbl_pending[i] = intel_fb->vbl_waited; - } -#endif } } @@ -373,103 +276,6 @@ intelWindowMoved(struct intel_context *intel) - - -/* Emit wait for pending flips */ -#if 0 -void -intel_wait_flips(struct intel_context *intel, GLuint batch_flags) -{ - struct intel_framebuffer *intel_fb = - (struct intel_framebuffer *) intel->ctx.DrawBuffer; - struct intel_renderbuffer *intel_rb = - intel_get_renderbuffer(&intel_fb->Base, - intel_fb->Base._ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : - BUFFER_BACK_LEFT); - - if (intel_fb->Base.Name == 0 && intel_rb->pf_pending == intel_fb->pf_seq) { - GLint pf_planes = intel_fb->pf_planes; - BATCH_LOCALS; - - /* Wait for pending flips to take effect */ - BEGIN_BATCH(2, batch_flags); - OUT_BATCH(pf_planes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) - : 0); - OUT_BATCH(pf_planes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) - : 0); - ADVANCE_BATCH(); - - intel_rb->pf_pending--; - } -} -#endif - -#if 0 -/* Flip the front & back buffers - */ -static GLboolean -intelPageFlip(const __DRIdrawablePrivate * dPriv) -{ - struct intel_context *intel; - int ret; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - - DBG(SWAP, "%s\n", __FUNCTION__); - - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate; - - if (intel->intelScreen->drmMinor < 9) - return GL_FALSE; - - st_flush(intel->st); - - ret = 0; - - LOCK_HARDWARE(intel); - - if (dPriv->numClipRects && intel_fb->pf_active) { - drm_i915_flip_t flip; - - flip.pipes = intel_fb->pf_planes; - - ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); - } - - UNLOCK_HARDWARE(intel); - - if (ret || !intel_fb->pf_active) - return GL_FALSE; - - if (!dPriv->numClipRects) { - usleep(10000); /* throttle invisible client 10ms */ - } - - intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_planes & 0x2)) & 0x3; - - if (dPriv->numClipRects != 0) { - intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending = - intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->pf_pending = - ++intel_fb->pf_seq; - } - -#if 0 - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, &intel_fb->Base); -#endif - - DBG(SWAP, "%s: success\n", __FUNCTION__); - - return GL_TRUE; -} -#endif - - static GLboolean intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) { @@ -477,21 +283,13 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); struct intel_context *intel = intelScreenContext(dPriv->driScreenPriv->private); -#if PF - const intelScreenPrivate *intelScreen = intel->intelScreen; -#endif unsigned int target; drm_i915_vblank_swap_t swap; GLboolean ret; /* XXX: Scheduled buffer swaps don't work with private back buffers yet */ if (1 || !intel_fb->vblank_flags || - (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) -#if PF - || - intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6) -#endif -) + (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) return GL_FALSE; swap.seqtype = DRM_VBLANK_ABSOLUTE; @@ -513,16 +311,6 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) intel_batchbuffer_flush(intel->batch); -#if PF - if ( intel_fb->pf_active ) { - swap.seqtype |= DRM_VBLANK_FLIP; - - intel_fb->pf_current_page = (((intel->sarea->pf_current_page >> - (intel_fb->pf_planes & 0x2)) & 0x3) + 1) % - intel_fb->pf_num_pages; - } -#endif - if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, sizeof(swap))) { intel_fb->vbl_seq = swap.sequence; @@ -538,22 +326,8 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) intel_fb->vbl_seq; #endif - if (swap.seqtype & DRM_VBLANK_FLIP) { -#if 0 - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); -#endif - } - ret = GL_TRUE; } else { -#if PF - if (swap.seqtype & DRM_VBLANK_FLIP) { - intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> - (intel_fb->pf_planes & 0x2)) & 0x3) % - intel_fb->pf_num_pages; - } -#endif ret = GL_FALSE; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index b7343db5e2..aae4e7c61f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -39,16 +39,6 @@ struct intel_framebuffer { struct gl_framebuffer Base; -#define PF 0 -#if PF - /* Drawable page flipping state */ - GLboolean pf_active; - GLuint pf_seq; - GLint pf_planes; - GLint pf_current_page; - GLint pf_num_pages; -#endif - /* VBI */ GLuint vbl_seq; @@ -69,8 +59,6 @@ extern void intelDisplayBuffer(__DRIdrawablePrivate * dPriv, struct pipe_surface *surf, const drm_clip_rect_t * rect); -extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); - extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); extern void intelWindowMoved(struct intel_context *intel); -- cgit v1.2.3 From a5df4239c2a1b91e36e7d1191c19a078e6b74a22 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:22:43 -0600 Subject: added check for fb==NULL --- src/mesa/state_tracker/st_cb_flush.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index eb47e75b40..819957a1ee 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -55,7 +55,9 @@ void st_flush( struct st_context *st ) */ st->pipe->flush( st->pipe, 0 ); - + if (!fb) + return; + /* XXX: temporary hack. This flag should only be set if we do any * rendering to the front buffer. */ -- cgit v1.2.3 From 488326b0b9c559511e3282f4dee6027db6fcae2f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:26:16 -0600 Subject: disable vblank/sync code --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 ++ src/mesa/drivers/dri/intel_winsys/intel_lock.c | 2 ++ src/mesa/drivers/dri/intel_winsys/intel_screen.c | 2 ++ src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 16 +++++++++++++++- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h | 3 +++ 5 files changed, 24 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index b0c1a608fa..ea732715a6 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -392,6 +392,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, if (intel->driDrawable != driDrawPriv) { if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) { +#if VBL int i; intel_fb->vblank_flags = driGetDefaultVBlankFlags(&intel->optionCache); @@ -404,6 +405,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, for (i = 0; i < 2; i++) { intel_fb->vbl_pending[i] = intel_fb->vbl_seq; } +#endif } } } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 38f0a4e2d7..059393f8b7 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -127,6 +127,7 @@ void LOCK_HARDWARE( struct intel_context *intel ) curbuf = 0; /* current draw buf: 0 = front, 1 = back */ +#if VBL if (intel_fb && intel_fb->vblank_flags && !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && (intel_fb->vbl_waited - intel_fb->vbl_pending[curbuf]) > (1<<23)) { @@ -142,6 +143,7 @@ void LOCK_HARDWARE( struct intel_context *intel ) drmWaitVBlank(intel->driFd, &vbl); intel_fb->vbl_waited = vbl.reply.sequence; } +#endif DRM_CAS(intel->driHwLock, intel->hHWContext, (DRM_LOCK_HELD|intel->hHWContext), __ret); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index be31774904..68f20556b4 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -363,6 +363,7 @@ intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) } intel_fb = dPriv->driverPrivate; +#if VBL sInfo->swap_count = intel_fb->swap_count; sInfo->swap_ust = intel_fb->swap_ust; sInfo->swap_missed_count = intel_fb->swap_missed_count; @@ -370,6 +371,7 @@ intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0) ? driCalculateSwapUsage(dPriv, 0, intel_fb->swap_missed_ust) : 0.0; +#endif return 0; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index ac4ae80cfb..5d20a4b940 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -231,6 +231,7 @@ intelWindowMoved(struct intel_context *intel) intelUpdateFramebufferSize(ctx, dPriv); intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ +#if VBL { drmI830Sarea *sarea = intel->sarea; drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, @@ -271,11 +272,12 @@ intelWindowMoved(struct intel_context *intel) } } - +#endif } +#if VBL static GLboolean intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) { @@ -335,6 +337,8 @@ intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) return ret; } +#endif + void intelSwapBuffers(__DRIdrawablePrivate * dPriv) @@ -355,6 +359,7 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ +#if VBL if (!intelScheduleSwap(dPriv, &missed_target)) { struct pipe_surface *back_surf = get_color_surface(intel_fb, BUFFER_BACK_LEFT); @@ -364,7 +369,15 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) intelDisplayBuffer(dPriv, back_surf, NULL); } +#else + { + struct pipe_surface *back_surf + = get_color_surface(intel_fb, BUFFER_BACK_LEFT); + intelDisplayBuffer(dPriv, back_surf, NULL); + } +#endif +#if VBL intel_fb->swap_count++; (*dri_interface->getUST) (&ust); if (missed_target) { @@ -373,6 +386,7 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) } intel_fb->swap_ust = ust; +#endif } } else { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index aae4e7c61f..138728862a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -39,6 +39,8 @@ struct intel_framebuffer { struct gl_framebuffer Base; +#define VBL 0 +#if VBL /* VBI */ GLuint vbl_seq; @@ -52,6 +54,7 @@ struct intel_framebuffer GLuint swap_missed_count; GLuint vbl_pending[3]; /**< [number of color buffers] */ +#endif }; -- cgit v1.2.3 From 2beb872e3bac49c9e132cbab0548d4276531c123 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:28:39 -0600 Subject: remove disabled vblank/sync code --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 14 --- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 18 --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 9 -- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 128 --------------------- .../drivers/dri/intel_winsys/intel_swapbuffers.h | 16 --- 5 files changed, 185 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index ea732715a6..9ed19325b7 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -392,20 +392,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, if (intel->driDrawable != driDrawPriv) { if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) { -#if VBL - int i; - - intel_fb->vblank_flags = driGetDefaultVBlankFlags(&intel->optionCache); - - (*dri_interface->getUST) (&intel_fb->swap_ust); - driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags, - &intel_fb->vbl_seq); - intel_fb->vbl_waited = intel_fb->vbl_seq; - - for (i = 0; i < 2; i++) { - intel_fb->vbl_pending[i] = intel_fb->vbl_seq; - } -#endif } } } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 059393f8b7..90ade2679c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -127,24 +127,6 @@ void LOCK_HARDWARE( struct intel_context *intel ) curbuf = 0; /* current draw buf: 0 = front, 1 = back */ -#if VBL - if (intel_fb && intel_fb->vblank_flags && - !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && - (intel_fb->vbl_waited - intel_fb->vbl_pending[curbuf]) > (1<<23)) { - drmVBlank vbl; - - vbl.request.type = DRM_VBLANK_ABSOLUTE; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - vbl.request.type |= DRM_VBLANK_SECONDARY; - } - - vbl.request.sequence = intel_fb->vbl_pending[curbuf]; - drmWaitVBlank(intel->driFd, &vbl); - intel_fb->vbl_waited = vbl.reply.sequence; - } -#endif - DRM_CAS(intel->driHwLock, intel->hHWContext, (DRM_LOCK_HELD|intel->hHWContext), __ret); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 68f20556b4..5dd4f94343 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -363,15 +363,6 @@ intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) } intel_fb = dPriv->driverPrivate; -#if VBL - sInfo->swap_count = intel_fb->swap_count; - sInfo->swap_ust = intel_fb->swap_ust; - sInfo->swap_missed_count = intel_fb->swap_missed_count; - - sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0) - ? driCalculateSwapUsage(dPriv, 0, intel_fb->swap_missed_ust) - : 0.0; -#endif return 0; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 5d20a4b940..1deb138adf 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -230,116 +230,10 @@ intelWindowMoved(struct intel_context *intel) /* Update Mesa's notion of window size */ intelUpdateFramebufferSize(ctx, dPriv); intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ - -#if VBL - { - drmI830Sarea *sarea = intel->sarea; - drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, - .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; - drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, - .y1 = sarea->pipeA_y, - .x2 = sarea->pipeA_x + sarea->pipeA_w, - .y2 = sarea->pipeA_y + sarea->pipeA_h }; - drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, - .y1 = sarea->pipeB_y, - .x2 = sarea->pipeB_x + sarea->pipeB_w, - .y2 = sarea->pipeB_y + sarea->pipeB_h }; - GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); - GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); - GLuint flags = intel_fb->vblank_flags; - - /* Update vblank info - */ - if (areaB > areaA || (areaA == areaB && areaB > 0)) { - flags = intel_fb->vblank_flags | VBLANK_FLAG_SECONDARY; - } else { - flags = intel_fb->vblank_flags & ~VBLANK_FLAG_SECONDARY; - } - - if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && - !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { - drmVBlank vbl; - - vbl.request.type = DRM_VBLANK_ABSOLUTE; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - vbl.request.type |= DRM_VBLANK_SECONDARY; - } - - intel_fb->vblank_flags = flags; - driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); - intel_fb->vbl_waited = intel_fb->vbl_seq; - - } - } -#endif } -#if VBL -static GLboolean -intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) -{ - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); - struct intel_context *intel = - intelScreenContext(dPriv->driScreenPriv->private); - unsigned int target; - drm_i915_vblank_swap_t swap; - GLboolean ret; - - /* XXX: Scheduled buffer swaps don't work with private back buffers yet */ - if (1 || !intel_fb->vblank_flags || - (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) - return GL_FALSE; - - swap.seqtype = DRM_VBLANK_ABSOLUTE; - - if (intel_fb->vblank_flags & VBLANK_FLAG_SYNC) { - swap.seqtype |= DRM_VBLANK_NEXTONMISS; - } else if (interval == 0) { - return GL_FALSE; - } - - swap.drawable = dPriv->hHWDrawable; - target = swap.sequence = intel_fb->vbl_seq + interval; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - swap.seqtype |= DRM_VBLANK_SECONDARY; - } - - LOCK_HARDWARE(intel); - - intel_batchbuffer_flush(intel->batch); - - if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, - sizeof(swap))) { - intel_fb->vbl_seq = swap.sequence; - swap.sequence -= target; - *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); - -#if 1 - intel_fb->vbl_pending[1] = intel_fb->vbl_pending[0] = intel_fb->vbl_seq; -#else - intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending = - intel_get_renderbuffer(&intel_fb->Base, - BUFFER_FRONT_LEFT)->vbl_pending = - intel_fb->vbl_seq; -#endif - - ret = GL_TRUE; - } else { - ret = GL_FALSE; - } - - UNLOCK_HARDWARE(intel); - - return ret; -} -#endif - - void intelSwapBuffers(__DRIdrawablePrivate * dPriv) { @@ -359,34 +253,12 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ -#if VBL - if (!intelScheduleSwap(dPriv, &missed_target)) { - struct pipe_surface *back_surf - = get_color_surface(intel_fb, BUFFER_BACK_LEFT); - - driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, - &missed_target); - - intelDisplayBuffer(dPriv, back_surf, NULL); - } -#else { struct pipe_surface *back_surf = get_color_surface(intel_fb, BUFFER_BACK_LEFT); intelDisplayBuffer(dPriv, back_surf, NULL); } -#endif - -#if VBL - intel_fb->swap_count++; - (*dri_interface->getUST) (&ust); - if (missed_target) { - intel_fb->swap_missed_count++; - intel_fb->swap_missed_ust = ust - intel_fb->swap_ust; - } - intel_fb->swap_ust = ust; -#endif } } else { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index 138728862a..9b3152f5d2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -39,22 +39,6 @@ struct intel_framebuffer { struct gl_framebuffer Base; -#define VBL 0 -#if VBL - /* VBI - */ - GLuint vbl_seq; - GLuint vblank_flags; - GLuint vbl_waited; - - int64_t swap_ust; - int64_t swap_missed_ust; - - GLuint swap_count; - GLuint swap_missed_count; - - GLuint vbl_pending[3]; /**< [number of color buffers] */ -#endif }; -- cgit v1.2.3 From d083cbb4db3801f82f6bbcf0a6306013498b5a3a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:31:39 -0600 Subject: remove useless code --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 10 ---------- 1 file changed, 10 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 9ed19325b7..0d33ee2f0f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -386,16 +386,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, _mesa_make_current(ctx, &intel_fb->Base, readFb); - /* The drawbuffer won't always be updated by _mesa_make_current: - */ - if (ctx->DrawBuffer == &intel_fb->Base) { - - if (intel->driDrawable != driDrawPriv) { - if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) { - } - } - } - if ((intel->driDrawable != driDrawPriv) || (intel->lastStamp != driDrawPriv->lastStamp)) { intel->driDrawable = driDrawPriv; -- cgit v1.2.3 From d04938331a15a94682ee7909de0f073314171cde Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:31:56 -0600 Subject: simplify intelSwapBuffers() --- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 14 +++----------- 1 file changed, 3 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 1deb138adf..ceebc10496 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -239,26 +239,18 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) { if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel; if (ctx == NULL) return; - intel = intel_context(ctx); - if (ctx->Visual.doubleBufferMode) { - GLboolean missed_target; struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - int64_t ust; + struct pipe_surface *back_surf + = get_color_surface(intel_fb, BUFFER_BACK_LEFT); _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - { - struct pipe_surface *back_surf - = get_color_surface(intel_fb, BUFFER_BACK_LEFT); - intelDisplayBuffer(dPriv, back_surf, NULL); - } - + intelDisplayBuffer(dPriv, back_surf, NULL); } } else { -- cgit v1.2.3 From 0452eb9086ff24508099deae9f98885f918c76db Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:33:33 -0600 Subject: s/intelDisplayBuffer/intelDisplaySurface --- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 10 +++++----- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h | 6 +++--- src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c | 2 +- 3 files changed, 9 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index ceebc10496..a642d53216 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -66,9 +66,9 @@ get_color_surface(struct intel_framebuffer *intel_fb, * \param rect optional subrect of surface to display (may be NULL). */ void -intelDisplayBuffer(__DRIdrawablePrivate * dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t * rect) +intelDisplaySurface(__DRIdrawablePrivate * dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t * rect) { struct intel_context *intel; const intelScreenPrivate *intelScreen @@ -250,7 +250,7 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - intelDisplayBuffer(dPriv, back_surf, NULL); + intelDisplaySurface(dPriv, back_surf, NULL); } } else { @@ -284,7 +284,7 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) rect.x2 = w; rect.y2 = h; _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - intelDisplayBuffer(dPriv, back_surf, &rect); + intelDisplaySurface(dPriv, back_surf, &rect); } } else { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index 9b3152f5d2..ed4a704787 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -42,9 +42,9 @@ struct intel_framebuffer }; -extern void intelDisplayBuffer(__DRIdrawablePrivate * dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t * rect); +extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t * rect); extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 9c3fd56d8d..af05a3398c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -192,7 +192,7 @@ intel_flush_frontbuffer( struct pipe_winsys *sws, struct intel_context *intel = intel_pipe_winsys(sws)->intel; __DRIdrawablePrivate *dPriv = intel->driDrawable; - intelDisplayBuffer(dPriv, surf, NULL); + intelDisplaySurface(dPriv, surf, NULL); } -- cgit v1.2.3 From a50d2feadd345c492a9a4710145f19df05e45edc Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:44:25 -0600 Subject: remove some struct decls --- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index ed4a704787..8d77ffce85 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -29,8 +29,6 @@ #define INTEL_BUFFERS_H -struct intel_context; -struct intel_framebuffer; /** * Intel framebuffer, derived from gl_framebuffer. -- cgit v1.2.3 From ca96ee421a40b8ab640a717baf12542869f1264a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:44:42 -0600 Subject: remove intel->pipe field --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 11 ++++++----- src/mesa/drivers/dri/intel_winsys/intel_context.h | 2 -- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 2 +- 3 files changed, 7 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 0d33ee2f0f..5433ae4842 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -175,6 +175,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, drmI830Sarea *saPriv = intelScreen->sarea; int fthrottle_mode; GLboolean havePools; + struct pipe_context *pipe; #if 0 intelInitDriverFunctions(&functions); @@ -245,7 +246,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, * Pipe-related setup */ if (!getenv("INTEL_HW")) { - intel->pipe = intel_create_softpipe( intel ); + pipe = intel_create_softpipe( intel ); } else { switch (intel->intelScreen->deviceID) { @@ -257,21 +258,21 @@ intelCreateContext(const __GLcontextModes * mesaVis, case PCI_CHIP_Q35_G: case PCI_CHIP_I915_G: case PCI_CHIP_I915_GM: - intel->pipe = intel_create_i915simple( intel ); + pipe = intel_create_i915simple( intel ); break; default: _mesa_printf("Unknown PCIID %x in %s, using software driver\n", intel->intelScreen->deviceID, __FUNCTION__); - intel->pipe = intel_create_softpipe( intel ); + pipe = intel_create_softpipe( intel ); break; } } #if 0 - st_create_context( &intel->ctx, intel->pipe ); + st_create_context( &intel->ctx, pipe ); #else - intel->st = st_create_context2(intel->pipe, mesaVis, NULL); + intel->st = st_create_context2(pipe, mesaVis, NULL); intel->st->ctx->DriverCtx = intel; #endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 788d444d00..78d549145b 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -50,8 +50,6 @@ struct intel_context { struct st_context *st; - struct pipe_context *pipe; /**< a softpipe or i915simple context */ - GLint refcount; struct _DriFenceObject *last_swap_fence; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 5dd4f94343..16e3a226e2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -383,7 +383,7 @@ intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, return; if (stObj->mt) - st_miptree_release(intel->pipe, &stObj->mt); + st_miptree_release(intel->st->pipe, &stObj->mt); stObj->imageOverride = GL_TRUE; stObj->depthOverride = depth; -- cgit v1.2.3 From cf30d377ac9b4242e5ba8d617e6d594105db0709 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:46:19 -0600 Subject: remove refcount, fboRect fields --- src/mesa/drivers/dri/intel_winsys/intel_context.h | 5 ----- 1 file changed, 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 78d549145b..e9c14b6a5b 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -50,24 +50,19 @@ struct intel_context { struct st_context *st; - GLint refcount; - struct _DriFenceObject *last_swap_fence; struct _DriFenceObject *first_swap_fence; struct intel_batchbuffer *batch; - GLboolean locked; char *prevLockFile; int prevLockLine; - /* These refer to the current drawing buffer: */ GLuint numClipRects; /**< cliprects for drawing */ drm_clip_rect_t *pClipRects; - drm_clip_rect_t fboRect; /**< cliprect for rendering */ GLuint irqsEmitted; drm_i915_irq_wait_t iw; -- cgit v1.2.3 From 798658cc81ca0d0ecb6d5c0aa9539e4d0996ec6d Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:48:39 -0600 Subject: remove INTEL_FIREVERTICES --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 +- src/mesa/drivers/dri/intel_winsys/intel_context.h | 3 --- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 2 +- 3 files changed, 2 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 5433ae4842..3898d2e4af 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -290,7 +290,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) if (intel) { GLboolean release_texture_heaps; - INTEL_FIREVERTICES(intel); + st_flush(intel->st); //intel->vtbl.destroy(intel); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index e9c14b6a5b..efaf22e2f1 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -91,9 +91,6 @@ void UNLOCK_HARDWARE( struct intel_context *intel ); extern char *__progname; -/* Will become a call into state_tracker: - */ -#define INTEL_FIREVERTICES(intel) /* ================================================================ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 90ade2679c..01d02f9a2a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -93,7 +93,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) */ intel->numClipRects = 0; - INTEL_FIREVERTICES(intel); + st_flush(intel->st); if (intel->batch->map != intel->batch->ptr) intel_batchbuffer_flush(intel->batch); -- cgit v1.2.3 From e8f1006081f9cf638d9f3869da24f6c96d227eba Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 12:53:46 -0600 Subject: remove dead code, unnecessary #includes, etc --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 14 +------------- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 11 +---------- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 2 -- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 3 --- 4 files changed, 2 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 3898d2e4af..188e973de0 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -44,8 +44,6 @@ #include "pipe/p_defines.h" #include "pipe/p_context.h" -/*#include "drirenderbuffer.h"*/ -#include "vblank.h" #include "utils.h" #include "xmlpool.h" /* for symbolic values of enum-type options */ @@ -327,6 +325,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) } } + GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv) { @@ -358,14 +357,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv) { - -#if 0 - if (driDrawPriv) { - fprintf(stderr, "x %d, y %d, width %d, height %d\n", - driDrawPriv->x, driDrawPriv->y, driDrawPriv->w, driDrawPriv->h); - } -#endif - if (driContextPriv) { struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; @@ -393,7 +384,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, intelWindowMoved(intel); intel->lastStamp = driDrawPriv->lastStamp; } - } else { _mesa_make_current(NULL, NULL, NULL); @@ -413,8 +403,6 @@ struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) GET_CURRENT_CONTEXT(ctx); if (ctx == NULL) { -/* _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; */ /* need a context for the first time makecurrent is called (for hw lock when allocating priv buffers) */ if (intelScreen->dummyctxptr == NULL) { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 01d02f9a2a..9636fa998d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -28,7 +28,7 @@ #include "glheader.h" #include "context.h" -#include "extensions.h" +#include "state_tracker/st_public.h" #include "intel_screen.h" #include "intel_context.h" @@ -38,15 +38,6 @@ #include "i830_dri.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" - -#include "drirenderbuffer.h" -#include "vblank.h" -#include "utils.h" -#include "xmlpool.h" /* for symbolic values of enum-type options */ - - _glthread_DECLARE_STATIC_MUTEX( lockMutex ); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 16e3a226e2..1fa6b57378 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -28,9 +28,7 @@ #include "glheader.h" #include "context.h" #include "framebuffer.h" -#include "matrix.h" #include "renderbuffer.h" -#include "simple_list.h" #include "utils.h" #include "vblank.h" #include "xmlpool.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index a642d53216..f6e7e16000 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -32,9 +32,6 @@ #include "intel_reg.h" #include "intel_winsys.h" #include "context.h" -#include "utils.h" -#include "drirenderbuffer.h" -#include "vblank.h" #include "pipe/p_context.h" #include "state_tracker/st_context.h" -- cgit v1.2.3 From 35924a34e5a405bc84459229abb477b2cb01409b Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 13:06:40 -0600 Subject: remove old mesa context creation code --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 37 +---------------------- 1 file changed, 1 insertion(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 188e973de0..da79f18c0c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -145,16 +145,6 @@ static const struct dri_debug_control debug_control[] = { #endif -static void -intelInitDriverFunctions(struct dd_function_table *functions) -{ - memset(functions, 0, sizeof(*functions)); - st_init_driver_functions(functions); -} - - - - GLboolean intelCreateContext(const __GLcontextModes * mesaVis, @@ -162,11 +152,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, void *sharedContextPrivate) { struct intel_context *intel = CALLOC_STRUCT(intel_context); -#if 0 - struct dd_function_table functions; - GLcontext *ctx = &intel->ctx; - GLcontext *shareCtx = (GLcontext *) sharedContextPrivate; -#endif __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; @@ -175,15 +160,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, GLboolean havePools; struct pipe_context *pipe; -#if 0 - intelInitDriverFunctions(&functions); - - if (!_mesa_initialize_context(&intel->ctx, - mesaVis, shareCtx, - &functions, (void *) intel)) - return GL_FALSE; -#endif - driContextPriv->driverPrivate = intel; intel->intelScreen = intelScreen; intel->driScreen = sPriv; @@ -267,12 +243,8 @@ intelCreateContext(const __GLcontextModes * mesaVis, } } -#if 0 - st_create_context( &intel->ctx, pipe ); -#else intel->st = st_create_context2(pipe, mesaVis, NULL); - intel->st->ctx->DriverCtx = intel; -#endif + intel->st->ctx->DriverCtx = intel; /* hope to get rid of this... */ return GL_TRUE; } @@ -314,14 +286,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) */ } -#if 0 - /* free the Mesa context data */ - _mesa_free_context_data(ctx); - - st_destroy_context(intel->ctx.st); -#else st_destroy_context2(intel->st); -#endif } } -- cgit v1.2.3 From 070aad3a9d8433fb29f687f500ed6ba2336ab72f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 13:08:20 -0600 Subject: remove mesa extension code --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 100 ---------------------- 1 file changed, 100 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index da79f18c0c..dd646a432a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -28,7 +28,6 @@ #include "glheader.h" #include "context.h" -#include "extensions.h" #include "framebuffer.h" #include "i830_dri.h" @@ -48,90 +47,10 @@ #include "xmlpool.h" /* for symbolic values of enum-type options */ - - #ifdef DEBUG int __intel_debug = 0; #endif -#define need_GL_ARB_multisample -#define need_GL_ARB_point_parameters -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_buffer_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_window_pos -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_equation_separate -#define need_GL_EXT_blend_func_separate -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_cull_vertex -#define need_GL_EXT_fog_coord -#define need_GL_EXT_framebuffer_object -#define need_GL_EXT_multi_draw_arrays -#define need_GL_EXT_secondary_color -#define need_GL_NV_vertex_program -#include "extension_helper.h" - - -#define DRIVER_DATE "20070731" - - - -/** - * Extension strings exported by the intel driver. - * - * \note - * It appears that ARB_texture_env_crossbar has "disappeared" compared to the - * old i830-specific driver. - */ -const struct dri_extension card_extensions[] = { - {"GL_ARB_multisample", GL_ARB_multisample_functions}, - {"GL_ARB_multitexture", NULL}, - {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, - {"GL_ARB_texture_border_clamp", NULL}, - {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, - {"GL_ARB_texture_cube_map", NULL}, - {"GL_ARB_texture_env_add", NULL}, - {"GL_ARB_texture_env_combine", NULL}, - {"GL_ARB_texture_env_dot3", NULL}, - {"GL_ARB_texture_mirrored_repeat", NULL}, - {"GL_ARB_texture_rectangle", NULL}, - {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, - {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, - {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, - {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", - GL_EXT_blend_equation_separate_functions}, - {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, - {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_subtract", NULL}, - {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, - {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, -#if 1 /* XXX FBO temporary? */ - {"GL_EXT_packed_depth_stencil", NULL}, -#endif - {"GL_EXT_pixel_buffer_object", NULL}, - {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, - {"GL_EXT_stencil_wrap", NULL}, - {"GL_EXT_texture_edge_clamp", NULL}, - {"GL_EXT_texture_env_combine", NULL}, - {"GL_EXT_texture_env_dot3", NULL}, - {"GL_EXT_texture_filter_anisotropic", NULL}, - {"GL_EXT_texture_lod_bias", NULL}, - {"GL_3DFX_texture_compression_FXT1", NULL}, - {"GL_APPLE_client_storage", NULL}, - {"GL_MESA_pack_invert", NULL}, - {"GL_MESA_ycbcr_texture", NULL}, - {"GL_NV_blend_square", NULL}, - {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, - {"GL_NV_vertex_program1_1", NULL}, -/* { "GL_SGIS_generate_mipmap", NULL }, */ - {NULL, NULL} -}; - #ifdef DEBUG @@ -192,25 +111,6 @@ intelCreateContext(const __GLcontextModes * mesaVis, intel->last_swap_fence = NULL; intel->first_swap_fence = NULL; - /* Disable imaging extension until convolution is working in - * teximage paths: - */ -#if 0 - driInitExtensions(ctx, card_extensions, -/* GL_TRUE, */ - GL_FALSE); -#endif - -#if 0 - if (intel->ctx.Mesa_DXTn) { - _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); - _mesa_enable_extension(ctx, "GL_S3_s3tc"); - } - else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) { - _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); - } -#endif - #ifdef DEBUG __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); #endif -- cgit v1.2.3 From 5ff1ff2b551821ccd90e85c4b8e7f32f60177adc Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 13:10:31 -0600 Subject: move intelScreenContext() to intel_swapbuffers.c and make static --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 24 ---------------------- src/mesa/drivers/dri/intel_winsys/intel_screen.h | 2 -- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 22 ++++++++++++++++++++ 3 files changed, 22 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index dd646a432a..f0e654c0be 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -256,27 +256,3 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, return GL_TRUE; } - - - -struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) -{ - /* - * This should probably change to have the screen allocate a dummy - * context at screen creation. For now just use the current context. - */ - - GET_CURRENT_CONTEXT(ctx); - if (ctx == NULL) { - /* need a context for the first time makecurrent is called (for hw lock - when allocating priv buffers) */ - if (intelScreen->dummyctxptr == NULL) { - _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; - } - return intelScreen->dummyctxptr; - } - - return intel_context(ctx); -} - diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index 549587a6f2..12a5206200 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -97,8 +97,6 @@ extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, unsigned numBufs, unsigned checkDelayed); -extern struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen); - extern GLboolean intelCreatePools(intelScreenPrivate *intelScreen); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index f6e7e16000..a80b5a4198 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -38,6 +38,28 @@ #include "state_tracker/st_cb_fbo.h" +static struct intel_context * +intelScreenContext(intelScreenPrivate *intelScreen) +{ + /* + * This should probably change to have the screen allocate a dummy + * context at screen creation. For now just use the current context. + */ + + GET_CURRENT_CONTEXT(ctx); + if (ctx == NULL) { + /* need a context for the first time makecurrent is called (for hw lock + when allocating priv buffers) */ + if (intelScreen->dummyctxptr == NULL) { + _mesa_problem(NULL, "No current context in intelScreenContext\n"); + return NULL; + } + return intelScreen->dummyctxptr; + } + + return intel_context(ctx); +} + /** * Return the pipe_surface for the given renderbuffer. -- cgit v1.2.3 From 7edf6304687a05f34e3f1e471451f54e7c8a5d47 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 13:16:03 -0600 Subject: more dead code removal, clean-up --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 17 +++-------------- 1 file changed, 3 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index f0e654c0be..25e11ebaf9 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -149,6 +149,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, return GL_TRUE; } + void intelDestroyContext(__DRIcontextPrivate * driContextPriv) { @@ -158,14 +159,8 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) assert(intel); /* should never be null */ if (intel) { - GLboolean release_texture_heaps; - st_flush(intel->st); - //intel->vtbl.destroy(intel); - - release_texture_heaps = (ctx->Shared->RefCount == 1); - intel_batchbuffer_free(intel->batch); if (intel->last_swap_fence) { @@ -179,13 +174,6 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) intel->first_swap_fence = NULL; } - - if (release_texture_heaps) { - /* This share group is about to go away, free our private - * texture object data. - */ - } - st_destroy_context2(intel->st); } } @@ -194,7 +182,8 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv) { - struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; + struct intel_context *intel + = (struct intel_context *) driContextPriv->driverPrivate; st_flush(intel->st); return GL_TRUE; } -- cgit v1.2.3 From 794e03d3926614036d285f2033d8708c7667814a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 13:22:22 -0600 Subject: Introduce st_framebuffer type and st_create_framebuffer(), st_make_current() --- src/mesa/state_tracker/st_context.c | 23 +++++++++++++++++++++++ src/mesa/state_tracker/st_context.h | 10 ++++++++++ src/mesa/state_tracker/st_public.h | 8 ++++++++ 3 files changed, 41 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 1d129ad077..db933b86e4 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -155,6 +155,29 @@ void st_destroy_context( struct st_context *st ) } +struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ) +{ + struct st_framebuffer *stfb + = CALLOC_STRUCT(st_framebuffer); + if (stfb) { + } + return stfb; +} + + +void st_make_current(struct st_context *st, + struct st_framebuffer *draw, + struct st_framebuffer *read) +{ + if (st) { + _mesa_make_current(st->ctx, &draw->Base, &read->Base); + } + else { + _mesa_make_current(NULL, NULL, NULL); + } +} + + void st_init_driver_functions(struct dd_function_table *functions) { diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 7ab69df7c7..9b232827f4 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -167,6 +167,16 @@ static INLINE struct st_context *st_context(GLcontext *ctx) } +/** + * Wrapper for GLframebuffer, nothing extra for now. + * This is an opaque type to the outside world. + */ +struct st_framebuffer +{ + GLframebuffer Base; +}; + + extern void st_init_driver_functions(struct dd_function_table *functions); diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 1cc6e6e40e..94d1cdfde4 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -31,6 +31,8 @@ #include "mtypes.h" struct st_context; +struct st_framebuffer; + struct pipe_context; struct st_context *st_create_context( GLcontext *ctx, @@ -44,6 +46,12 @@ void st_destroy_context( struct st_context *st ); void st_destroy_context2( struct st_context *st ); +struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ); + +void st_make_current(struct st_context *st, + struct st_framebuffer *draw, + struct st_framebuffer *read); + void st_invalidate_state(GLcontext * ctx, GLuint new_state); void st_flush( struct st_context *st ); -- cgit v1.2.3 From 01b71b8f7a77b3eb5ccc009342c1ebdc7b3dda80 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 13:25:13 -0600 Subject: new file for st_framebuffer stuff --- src/mesa/state_tracker/st_framebuffer.c | 103 ++++++++++++++++++++++++++++++++ 1 file changed, 103 insertions(+) create mode 100644 src/mesa/state_tracker/st_framebuffer.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c new file mode 100644 index 0000000000..3997681f36 --- /dev/null +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -0,0 +1,103 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "main/imports.h" +#include "main/context.h" +#include "st_public.h" +#include "st_context.h" + + +struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ) +{ + struct st_framebuffer *stfb + = CALLOC_STRUCT(st_framebuffer); + if (stfb) { + +#if 0 + /* from intel driver... */ + + GLboolean swStencil = (mesaVis->stencilBits > 0 && + mesaVis->depthBits != 24); + GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); + + struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); + if (!intel_fb) + return GL_FALSE; + + _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); + + /* + * XXX Create pipe_surfaces for front/back buffers, + * hand them to state tracker to create a framebuffer object. + */ + + { + /* fake frontbuffer */ + /* XXX allocation should only happen in the unusual case + it's actually needed */ + struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, rb); + } + + if (mesaVis->doubleBufferMode) { + struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, rb); + } + + if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { + /* combined depth/stencil buffer */ + struct gl_renderbuffer *depthStencilRb + = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + /* note: bind RB to two attachment points */ + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, depthStencilRb); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL,depthStencilRb); + } + else if (mesaVis->depthBits == 16) { + /* just 16-bit depth buffer, no hw stencil */ + struct gl_renderbuffer *depthRb + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); + _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, depthRb); + } + + + /* now add any/all software-based renderbuffers we may need */ + _mesa_add_soft_renderbuffers(&intel_fb->Base, + GL_FALSE, /* never sw color */ + GL_FALSE, /* never sw depth */ + swStencil, mesaVis->accumRedBits > 0, + GL_FALSE, /* never sw alpha */ + GL_FALSE /* never sw aux */ ); + driDrawPriv->driverPrivate = (void *) intel_fb; + +#endif + + } + return stfb; +} + + -- cgit v1.2.3 From bffed01e272b9a848aa7980ae72ae291d2677d05 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 2 Nov 2007 13:25:44 -0600 Subject: move st_create_framebuffer() to new st_framebuffer.c file --- src/mesa/sources | 1 + src/mesa/state_tracker/st_context.c | 10 ---------- 2 files changed, 1 insertion(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 50af7d208a..5322cd8867 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -232,6 +232,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_debug.c \ state_tracker/st_draw.c \ state_tracker/st_format.c \ + state_tracker/st_framebuffer.c \ state_tracker/st_mesa_to_tgsi.c \ state_tracker/st_program.c \ state_tracker/st_mipmap_tree.c diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index db933b86e4..fbce19b58e 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -155,16 +155,6 @@ void st_destroy_context( struct st_context *st ) } -struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ) -{ - struct st_framebuffer *stfb - = CALLOC_STRUCT(st_framebuffer); - if (stfb) { - } - return stfb; -} - - void st_make_current(struct st_context *st, struct st_framebuffer *draw, struct st_framebuffer *read) -- cgit v1.2.3 From e49efdab4a05d41f4b2f0bf989cdf036ab25879c Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 09:39:12 -0600 Subject: st_create_framebuffer() working --- src/mesa/state_tracker/st_framebuffer.c | 47 ++++++++++++--------------------- 1 file changed, 17 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 3997681f36..18669712c2 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -28,8 +28,11 @@ #include "main/imports.h" #include "main/context.h" +#include "main/framebuffer.h" +#include "main/renderbuffer.h" #include "st_public.h" #include "st_context.h" +#include "st_cb_fbo.h" struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ) @@ -37,64 +40,48 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ) struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer); if (stfb) { + GLboolean swStencil = (visual->stencilBits > 0 && + visual->depthBits != 24); + GLenum rgbFormat = (visual->redBits == 5 ? GL_RGB5 : GL_RGBA8); -#if 0 - /* from intel driver... */ - - GLboolean swStencil = (mesaVis->stencilBits > 0 && - mesaVis->depthBits != 24); - GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); - - struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); - if (!intel_fb) - return GL_FALSE; - - _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); - - /* - * XXX Create pipe_surfaces for front/back buffers, - * hand them to state tracker to create a framebuffer object. - */ + _mesa_initialize_framebuffer(&stfb->Base, visual); { /* fake frontbuffer */ /* XXX allocation should only happen in the unusual case it's actually needed */ struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, rb); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb); } - if (mesaVis->doubleBufferMode) { + if (visual->doubleBufferMode) { struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, rb); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); } - if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { + if (visual->depthBits == 24 && visual->stencilBits == 8) { /* combined depth/stencil buffer */ struct gl_renderbuffer *depthStencilRb = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); /* note: bind RB to two attachment points */ - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, depthStencilRb); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL,depthStencilRb); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL,depthStencilRb); } - else if (mesaVis->depthBits == 16) { + else if (visual->depthBits == 16) { /* just 16-bit depth buffer, no hw stencil */ struct gl_renderbuffer *depthRb = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, depthRb); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } /* now add any/all software-based renderbuffers we may need */ - _mesa_add_soft_renderbuffers(&intel_fb->Base, + _mesa_add_soft_renderbuffers(&stfb->Base, GL_FALSE, /* never sw color */ GL_FALSE, /* never sw depth */ - swStencil, mesaVis->accumRedBits > 0, + swStencil, visual->accumRedBits > 0, GL_FALSE, /* never sw alpha */ GL_FALSE /* never sw aux */ ); - driDrawPriv->driverPrivate = (void *) intel_fb; - -#endif } return stfb; -- cgit v1.2.3 From 2085cc9ac7cc2efb681f8ac824686390e83e2475 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 09:39:31 -0600 Subject: use st_create_framebuffer() --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 1fa6b57378..ec94f3bbe1 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -41,6 +41,7 @@ #include "dri_bufpool.h" #include "pipe/p_context.h" +#include "state_tracker/st_public.h" #include "state_tracker/st_cb_fbo.h" @@ -283,6 +284,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, return GL_FALSE; /* not implemented */ } else { +#if 0 GLboolean swStencil = (mesaVis->stencilBits > 0 && mesaVis->depthBits != 24); GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); @@ -335,6 +337,13 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, GL_FALSE, /* never sw alpha */ GL_FALSE /* never sw aux */ ); driDrawPriv->driverPrivate = (void *) intel_fb; +#else + struct st_framebuffer *stfb; + + stfb = st_create_framebuffer(mesaVis); + + driDrawPriv->driverPrivate = (void *) stfb; +#endif return GL_TRUE; } -- cgit v1.2.3 From c4a9c49cdb6740da7a7bda8804ffe04fe04c0b0e Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 09:52:27 -0600 Subject: restore dummy card_extensions extensions array, needed for dispatch setup --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 25e11ebaf9..35e90f967e 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -53,6 +53,20 @@ int __intel_debug = 0; + +/** + * Extension strings exported by the intel driver. + * + * \note + * It appears that ARB_texture_env_crossbar has "disappeared" compared to the + * old i830-specific driver. + */ +const struct dri_extension card_extensions[] = { + {NULL, NULL} +}; + + + #ifdef DEBUG static const struct dri_debug_control debug_control[] = { {"ioctl", DEBUG_IOCTL}, -- cgit v1.2.3 From ecb41279d6c9f7fd0b2dc44309fd71fad5db33a9 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 10:04:24 -0600 Subject: use st_make_current() --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 35e90f967e..721f73bcd9 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -228,9 +228,10 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, if (driContextPriv) { struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; - struct intel_framebuffer *intel_fb = - (struct intel_framebuffer *) driDrawPriv->driverPrivate; - GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; + struct st_framebuffer *draw_fb + = (struct st_framebuffer *) driDrawPriv->driverPrivate; + struct st_framebuffer *read_fb + = (struct st_framebuffer *) driReadPriv->driverPrivate; GLcontext *ctx = intel->st->ctx; /* this is a hack so we have a valid context when the region allocation @@ -244,7 +245,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, intelUpdateFramebufferSize(ctx, driReadPriv); } - _mesa_make_current(ctx, &intel_fb->Base, readFb); + st_make_current(intel->st, draw_fb, read_fb); if ((intel->driDrawable != driDrawPriv) || (intel->lastStamp != driDrawPriv->lastStamp)) { @@ -254,7 +255,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, } } else { - _mesa_make_current(NULL, NULL, NULL); + st_make_current(NULL, NULL, NULL); } return GL_TRUE; -- cgit v1.2.3 From 3d14b2c01e12823877ae9270c7a79da65681f25c Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 10:19:10 -0600 Subject: added st_resize_framebuffer() --- src/mesa/state_tracker/st_framebuffer.c | 14 ++++++++++++++ src/mesa/state_tracker/st_public.h | 3 +++ 2 files changed, 17 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 18669712c2..99638c5d5f 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -88,3 +88,17 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ) } +void st_resize_framebuffer( struct st_framebuffer *stfb, + GLuint width, GLuint height ) +{ + if (stfb->Base.Width != width || stfb->Base.Height != height) { + GET_CURRENT_CONTEXT(ctx); + if (ctx) { + _mesa_resize_framebuffer(ctx, &stfb->Base, width, height); + + assert(stfb->Base.Width == width); + assert(stfb->Base.Height == height); + } + } +} + diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 94d1cdfde4..74a94d1abc 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -48,6 +48,9 @@ void st_destroy_context2( struct st_context *st ); struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ); +void st_resize_framebuffer( struct st_framebuffer *, + GLuint width, GLuint height ); + void st_make_current(struct st_context *st, struct st_framebuffer *draw, struct st_framebuffer *read); -- cgit v1.2.3 From e9d2156edfba222d6aa56d94c041fc2017c240fd Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 10:19:38 -0600 Subject: Remove intelUpdateFramebufferSize(), use st_resize_framebuffer(). --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 38 +++++----------------- src/mesa/drivers/dri/intel_winsys/intel_context.h | 3 -- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 9 ++--- 3 files changed, 14 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 721f73bcd9..8cc285276a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -169,7 +169,6 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) { struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; - GLcontext *ctx = intel->st->ctx; assert(intel); /* should never be null */ if (intel) { @@ -203,55 +202,36 @@ intelUnbindContext(__DRIcontextPrivate * driContextPriv) } -/** - * Copied/modified from drirenderbuffer.c - */ -void -intelUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv) -{ - struct gl_framebuffer *fb = (struct gl_framebuffer *) dPriv->driverPrivate; - if (fb && (dPriv->w != fb->Width || dPriv->h != fb->Height)) { - _mesa_resize_framebuffer(ctx, fb, dPriv->w, dPriv->h); - /* if the driver needs the hw lock for ResizeBuffers, the drawable - might have changed again by now */ - assert(fb->Width == dPriv->w); - assert(fb->Height == dPriv->h); - } -} - - GLboolean intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv) { if (driContextPriv) { - struct intel_context *intel = - (struct intel_context *) driContextPriv->driverPrivate; + struct intel_context *intel + = (struct intel_context *) driContextPriv->driverPrivate; struct st_framebuffer *draw_fb = (struct st_framebuffer *) driDrawPriv->driverPrivate; struct st_framebuffer *read_fb = (struct st_framebuffer *) driReadPriv->driverPrivate; - GLcontext *ctx = intel->st->ctx; /* this is a hack so we have a valid context when the region allocation is done. Need a per-screen context? */ intel->intelScreen->dummyctxptr = intel; - /* update GLframebuffer size to match window if needed */ - intelUpdateFramebufferSize(ctx, driDrawPriv); + st_make_current(intel->st, draw_fb, read_fb); + /* update size of Mesa framebuffer(s) to match window */ + st_resize_framebuffer(draw_fb, driDrawPriv->w, driDrawPriv->h); if (driReadPriv != driDrawPriv) { - intelUpdateFramebufferSize(ctx, driReadPriv); + st_resize_framebuffer(read_fb, driReadPriv->w, driReadPriv->h); } - st_make_current(intel->st, draw_fb, read_fb); - if ((intel->driDrawable != driDrawPriv) || (intel->lastStamp != driDrawPriv->lastStamp)) { - intel->driDrawable = driDrawPriv; - intelWindowMoved(intel); - intel->lastStamp = driDrawPriv->lastStamp; + intel->driDrawable = driDrawPriv; + intelWindowMoved(intel); + intel->lastStamp = driDrawPriv->lastStamp; } } else { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index efaf22e2f1..8f3ce026b6 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -140,7 +140,4 @@ intel_context(GLcontext * ctx) } -extern void -intelUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv); - #endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index a80b5a4198..387e3bc72e 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -34,6 +34,7 @@ #include "context.h" #include "pipe/p_context.h" +#include "state_tracker/st_public.h" #include "state_tracker/st_context.h" #include "state_tracker/st_cb_fbo.h" @@ -237,17 +238,17 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, /** * This will be called whenever the currently bound window is moved/resized. - * XXX: actually, it seems to NOT be called when the window is only moved (BP). */ void intelWindowMoved(struct intel_context *intel) { - GLcontext *ctx = intel->st->ctx; __DRIdrawablePrivate *dPriv = intel->driDrawable; struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + struct st_framebuffer *stfb + = (struct st_framebuffer *) dPriv->driverPrivate; + + st_resize_framebuffer(stfb, dPriv->w, dPriv->h); - /* Update Mesa's notion of window size */ - intelUpdateFramebufferSize(ctx, dPriv); intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ } -- cgit v1.2.3 From ed8774e920706acb3445471833709dcaa691cb0c Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 10:25:59 -0600 Subject: set stfb->Base.Initialized --- src/mesa/state_tracker/st_framebuffer.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 99638c5d5f..f454499fd8 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -83,6 +83,9 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ) GL_FALSE, /* never sw alpha */ GL_FALSE /* never sw aux */ ); + + stfb->Base.Initialized = GL_TRUE; + } return stfb; } -- cgit v1.2.3 From cb62b64e33278f599fca62d928433f6b6989317e Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 10:26:19 -0600 Subject: simplify the window moved/resized code a bit --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 12 ++++++++++++ src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 3 --- 2 files changed, 12 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 8cc285276a..2b581a6b0f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -221,11 +221,13 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, st_make_current(intel->st, draw_fb, read_fb); +#if 0 /* update size of Mesa framebuffer(s) to match window */ st_resize_framebuffer(draw_fb, driDrawPriv->w, driDrawPriv->h); if (driReadPriv != driDrawPriv) { st_resize_framebuffer(read_fb, driReadPriv->w, driReadPriv->h); } +#endif if ((intel->driDrawable != driDrawPriv) || (intel->lastStamp != driDrawPriv->lastStamp)) { @@ -233,6 +235,16 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, intelWindowMoved(intel); intel->lastStamp = driDrawPriv->lastStamp; } + +#if 1 + /* The size of the draw buffer will have been updated above. + * If the readbuffer is a different window, check/update its size now. + */ + if (driReadPriv != driDrawPriv) { + st_resize_framebuffer(read_fb, driReadPriv->w, driReadPriv->h); + } +#endif + } else { st_make_current(NULL, NULL, NULL); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 387e3bc72e..5d04815da8 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -243,13 +243,10 @@ void intelWindowMoved(struct intel_context *intel) { __DRIdrawablePrivate *dPriv = intel->driDrawable; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; struct st_framebuffer *stfb = (struct st_framebuffer *) dPriv->driverPrivate; st_resize_framebuffer(stfb, dPriv->w, dPriv->h); - - intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ } -- cgit v1.2.3 From 02a1c8f5690a82dc77db7af381d1b6e4aea67e29 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 10:35:39 -0600 Subject: pass __DRIdrawablePrivate to intelWindowMoved(), rather than context ptr --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 4 ++-- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 14 +++----------- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h | 3 ++- 3 files changed, 7 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 2b581a6b0f..443276cfc0 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -232,7 +232,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, if ((intel->driDrawable != driDrawPriv) || (intel->lastStamp != driDrawPriv->lastStamp)) { intel->driDrawable = driDrawPriv; - intelWindowMoved(intel); + intelWindowMoved(driDrawPriv); intel->lastStamp = driDrawPriv->lastStamp; } @@ -241,7 +241,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, * If the readbuffer is a different window, check/update its size now. */ if (driReadPriv != driDrawPriv) { - st_resize_framebuffer(read_fb, driReadPriv->w, driReadPriv->h); + intelWindowMoved(driReadPriv); } #endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 5d04815da8..2db3a29db2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -222,16 +222,10 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, UNLOCK_HARDWARE(intel); - /* XXX this is bogus. The context here may not even be bound to this drawable! */ if (intel->lastStamp != dPriv->lastStamp) { - GET_CURRENT_CONTEXT(currctx); - struct intel_context *intelcurrent = intel_context(currctx); - if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { - intelWindowMoved(intel); - intel->lastStamp = dPriv->lastStamp; - } + intelWindowMoved(dPriv); + intel->lastStamp = dPriv->lastStamp; } - } @@ -240,12 +234,10 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, * This will be called whenever the currently bound window is moved/resized. */ void -intelWindowMoved(struct intel_context *intel) +intelWindowMoved(__DRIdrawablePrivate *dPriv) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; struct st_framebuffer *stfb = (struct st_framebuffer *) dPriv->driverPrivate; - st_resize_framebuffer(stfb, dPriv->w, dPriv->h); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index 8d77ffce85..7bea6e6d40 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -46,6 +46,7 @@ extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); -extern void intelWindowMoved(struct intel_context *intel); +extern void intelWindowMoved(__DRIdrawablePrivate *dPriv); + #endif /* INTEL_BUFFERS_H */ -- cgit v1.2.3 From fc73b216065325fc4ce2c500be5a5fa2072db0b3 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 3 Nov 2007 10:41:03 -0600 Subject: rename intelWindowMoved() to intelUpdateWindowSize() --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 14 ++------------ src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 4 ++-- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h | 2 +- 3 files changed, 5 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 443276cfc0..8c29a51885 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -221,29 +221,19 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, st_make_current(intel->st, draw_fb, read_fb); -#if 0 - /* update size of Mesa framebuffer(s) to match window */ - st_resize_framebuffer(draw_fb, driDrawPriv->w, driDrawPriv->h); - if (driReadPriv != driDrawPriv) { - st_resize_framebuffer(read_fb, driReadPriv->w, driReadPriv->h); - } -#endif - if ((intel->driDrawable != driDrawPriv) || (intel->lastStamp != driDrawPriv->lastStamp)) { intel->driDrawable = driDrawPriv; - intelWindowMoved(driDrawPriv); + intelUpdateWindowSize(driDrawPriv); intel->lastStamp = driDrawPriv->lastStamp; } -#if 1 /* The size of the draw buffer will have been updated above. * If the readbuffer is a different window, check/update its size now. */ if (driReadPriv != driDrawPriv) { - intelWindowMoved(driReadPriv); + intelUpdateWindowSize(driReadPriv); } -#endif } else { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 2db3a29db2..aef3b87bc8 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -223,7 +223,7 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, UNLOCK_HARDWARE(intel); if (intel->lastStamp != dPriv->lastStamp) { - intelWindowMoved(dPriv); + intelUpdateWindowSize(dPriv); intel->lastStamp = dPriv->lastStamp; } } @@ -234,7 +234,7 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, * This will be called whenever the currently bound window is moved/resized. */ void -intelWindowMoved(__DRIdrawablePrivate *dPriv) +intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) { struct st_framebuffer *stfb = (struct st_framebuffer *) dPriv->driverPrivate; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index 7bea6e6d40..0d99a771db 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -46,7 +46,7 @@ extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); -extern void intelWindowMoved(__DRIdrawablePrivate *dPriv); +extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); #endif /* INTEL_BUFFERS_H */ -- cgit v1.2.3 From 44dfed15e0c5113d42d781e3937a2f619284454a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 08:41:45 -0700 Subject: added st_unreference_framebuffer() --- src/mesa/state_tracker/st_framebuffer.c | 7 +++++++ src/mesa/state_tracker/st_public.h | 4 +++- 2 files changed, 10 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index f454499fd8..54ae175e60 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -105,3 +105,10 @@ void st_resize_framebuffer( struct st_framebuffer *stfb, } } + +void st_unreference_framebuffer( struct st_framebuffer **stfb ) +{ + _mesa_unreference_framebuffer((struct gl_framebuffer **) stfb); +} + + diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 74a94d1abc..abdfcca80f 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -48,9 +48,11 @@ void st_destroy_context2( struct st_context *st ); struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ); -void st_resize_framebuffer( struct st_framebuffer *, +void st_resize_framebuffer( struct st_framebuffer *stfb, GLuint width, GLuint height ); +void st_unreference_framebuffer( struct st_framebuffer **stfb ); + void st_make_current(struct st_context *st, struct st_framebuffer *draw, struct st_framebuffer *read); -- cgit v1.2.3 From 131330994aed8ff63360a9fb54d498a3d3641e5f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 08:42:18 -0700 Subject: use st_unreference_framebuffer() --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index ec94f3bbe1..bcbb93ecb8 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -352,7 +352,8 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, static void intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) { - _mesa_unreference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate))); + st_unreference_framebuffer((struct st_framebuffer **) + (&(driDrawPriv->driverPrivate))); } -- cgit v1.2.3 From 4e0b38958d5227a20653fb37da8f7621afe6402c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 08:47:48 -0700 Subject: remove mesa includes --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 4 ---- src/mesa/drivers/dri/intel_winsys/intel_context.h | 2 +- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 2 -- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 4 ---- src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c | 3 +-- 5 files changed, 2 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 8c29a51885..292c509d19 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -26,10 +26,6 @@ **************************************************************************/ -#include "glheader.h" -#include "context.h" -#include "framebuffer.h" - #include "i830_dri.h" #include "intel_screen.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 8f3ce026b6..03c7fd2630 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -106,7 +106,7 @@ extern int __intel_debug; #define DBG(flag, ...) do { \ if (__intel_debug & (DEBUG_##flag)) \ - _mesa_printf(__VA_ARGS__); \ + printf(__VA_ARGS__); \ } while(0) #else diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 9636fa998d..af9b53a52b 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -26,8 +26,6 @@ **************************************************************************/ -#include "glheader.h" -#include "context.h" #include "state_tracker/st_public.h" #include "intel_screen.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index bcbb93ecb8..423f126439 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -25,10 +25,6 @@ * **************************************************************************/ -#include "glheader.h" -#include "context.h" -#include "framebuffer.h" -#include "renderbuffer.h" #include "utils.h" #include "vblank.h" #include "xmlpool.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c index f1358fef1a..5407973fba 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c @@ -29,12 +29,11 @@ * Authors: Keith Whitwell */ -#include "imports.h" #include "intel_context.h" #include "intel_winsys.h" #include "pipe/softpipe/sp_winsys.h" #include "pipe/p_defines.h" - +#include "pipe/p_util.h" struct intel_softpipe_winsys { -- cgit v1.2.3 From 5d8e3833801d161e47b05f3eed13f0584f156d2e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 08:58:45 -0700 Subject: remove cliprect code --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 34 +++++----------------- 1 file changed, 8 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 6fe9ecc539..9f55c7ece7 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -174,14 +174,14 @@ static void intel_batch_ioctl(struct intel_context *intel, GLuint start_offset, GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) + GLboolean allow_unlock) { drmI830BatchBuffer batch; batch.start = start_offset; batch.used = used; - batch.cliprects = intel->pClipRects; - batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + batch.cliprects = NULL; /* unused */ + batch.num_cliprects = 0; batch.DR1 = 0; batch.DR4 = 0; /* still need this ? */ @@ -206,7 +206,7 @@ intel_batch_ioctl(struct intel_context *intel, static void do_flush_locked(struct intel_batchbuffer *batch, GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) + GLboolean allow_unlock) { GLuint *ptr; GLuint i; @@ -236,17 +236,9 @@ do_flush_locked(struct intel_batchbuffer *batch, driBOUnmap(batch->buffer); batch->map = NULL; - /* Throw away non-effective packets. Won't work once we have - * hardware contexts which would preserve statechanges beyond a - * single buffer. - */ - - if (!(intel->numClipRects == 0 && !ignore_cliprects)) { - intel_batch_ioctl(batch->intel, - driBOOffset(batch->buffer), - used, ignore_cliprects, allow_unlock); - } - + intel_batch_ioctl(batch->intel, + driBOOffset(batch->buffer), + used, allow_unlock); /* * Kernel fencing. The flags tells the kernel that we've @@ -279,15 +271,6 @@ do_flush_locked(struct intel_batchbuffer *batch, driBOFence(r->buf, fo); } } - - if (intel->numClipRects == 0 && !ignore_cliprects) { - if (allow_unlock) { - UNLOCK_HARDWARE(intel); - sched_yield(); - LOCK_HARDWARE(intel); - } -// intel->vtbl.lost_hardware(intel); - } } @@ -328,8 +311,7 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch) if (!was_locked) LOCK_HARDWARE(intel); - do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), - GL_FALSE); + do_flush_locked(batch, used, GL_FALSE); if (!was_locked) UNLOCK_HARDWARE(intel); -- cgit v1.2.3 From 1cb81470b78f22708e9b83a40acb470744575ff3 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:00:18 -0700 Subject: remove unused var --- src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 9f55c7ece7..6b7772ddb2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -210,7 +210,6 @@ do_flush_locked(struct intel_batchbuffer *batch, { GLuint *ptr; GLuint i; - struct intel_context *intel = batch->intel; unsigned fenceFlags; struct _DriFenceObject *fo; -- cgit v1.2.3 From bac76b71f1fed705fb5ee68bec88b82ad74de920 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:02:54 -0700 Subject: remove intel_context cliprect code --- src/mesa/drivers/dri/intel_winsys/intel_context.h | 5 ---- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 33 +---------------------- 2 files changed, 1 insertion(+), 37 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 03c7fd2630..3a61a5610f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -59,11 +59,6 @@ struct intel_context char *prevLockFile; int prevLockLine; - /* These refer to the current drawing buffer: - */ - GLuint numClipRects; /**< cliprects for drawing */ - drm_clip_rect_t *pClipRects; - GLuint irqsEmitted; drm_i915_irq_wait_t iw; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index af9b53a52b..eee1ccc6ea 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -65,40 +65,9 @@ intelContendedLock(struct intel_context *intel, GLuint flags) intelUpdateScreenRotation(sPriv, sarea); } - -#if 0 - if (sarea->width != intel->width || - sarea->height != intel->height || - sarea->rotation != intel->current_rotation) { - int numClipRects = intel->numClipRects; - - /* - * FIXME: Really only need to do this when drawing to a - * common back- or front buffer. - */ - - /* - * This will essentially drop the outstanding batchbuffer on the floor. - */ - intel->numClipRects = 0; - - st_flush(intel->st); - - if (intel->batch->map != intel->batch->ptr) - intel_batchbuffer_flush(intel->batch); - - intel->numClipRects = numClipRects; - - /* force window update */ - intel->lastStamp = 0; - - intel->width = sarea->width; - intel->height = sarea->height; - intel->current_rotation = sarea->rotation; - } -#endif } + /* Lock the hardware and validate our state. */ void LOCK_HARDWARE( struct intel_context *intel ) -- cgit v1.2.3 From c1933ed23fa72e0d5bffd156c5c52558cc6b3190 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:04:37 -0700 Subject: remove dead code in intelCreateBuffer() --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 61 +----------------------- 1 file changed, 1 insertion(+), 60 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 423f126439..0472ea9fe1 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -280,67 +280,8 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, return GL_FALSE; /* not implemented */ } else { -#if 0 - GLboolean swStencil = (mesaVis->stencilBits > 0 && - mesaVis->depthBits != 24); - GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); - - struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); - if (!intel_fb) - return GL_FALSE; - - _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); - - /* - * XXX Create pipe_surfaces for front/back buffers, - * hand them to state tracker to create a framebuffer object. - */ - - { - /* fake frontbuffer */ - /* XXX allocation should only happen in the unusual case - it's actually needed */ - struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, rb); - } - - if (mesaVis->doubleBufferMode) { - struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, rb); - } - - if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { - /* combined depth/stencil buffer */ - struct gl_renderbuffer *depthStencilRb - = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); - /* note: bind RB to two attachment points */ - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, depthStencilRb); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL,depthStencilRb); - } - else if (mesaVis->depthBits == 16) { - /* just 16-bit depth buffer, no hw stencil */ - struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, depthRb); - } - - - /* now add any/all software-based renderbuffers we may need */ - _mesa_add_soft_renderbuffers(&intel_fb->Base, - GL_FALSE, /* never sw color */ - GL_FALSE, /* never sw depth */ - swStencil, mesaVis->accumRedBits > 0, - GL_FALSE, /* never sw alpha */ - GL_FALSE /* never sw aux */ ); - driDrawPriv->driverPrivate = (void *) intel_fb; -#else - struct st_framebuffer *stfb; - - stfb = st_create_framebuffer(mesaVis); - + struct st_framebuffer *stfb = st_create_framebuffer(mesaVis); driDrawPriv->driverPrivate = (void *) stfb; -#endif - return GL_TRUE; } } -- cgit v1.2.3 From 32f05c35d2814269dfd72c020c06e2cbcba68ad2 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:07:35 -0700 Subject: move intelCopySubBuffer() decl --- src/mesa/drivers/dri/intel_winsys/intel_screen.h | 5 ----- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h | 3 +++ 2 files changed, 3 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index 12a5206200..4dd10b08cb 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -87,11 +87,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv); -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h); - extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, unsigned long bufSize, unsigned numBufs, diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index 0d99a771db..b8a4ac97fb 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -46,6 +46,9 @@ extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); +extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, + int x, int y, int w, int h); + extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); -- cgit v1.2.3 From 1f17d845ff7a221a01de721aed81754175ab0a54 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:09:11 -0700 Subject: remove unused intel_fb decls, code --- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 5 ----- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 4 ---- 2 files changed, 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index eee1ccc6ea..5962b9259d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -73,16 +73,11 @@ intelContendedLock(struct intel_context *intel, GLuint flags) void LOCK_HARDWARE( struct intel_context *intel ) { char __ret=0; - struct intel_framebuffer *intel_fb = NULL; int curbuf; _glthread_LOCK_MUTEX(lockMutex); assert(!intel->locked); - if (intel->driDrawable) { - intel_fb = intel->driDrawable->driverPrivate; - } - curbuf = 0; /* current draw buf: 0 = front, 1 = back */ DRM_CAS(intel->driHwLock, intel->hHWContext, diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 0472ea9fe1..cfa61fa67f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -300,15 +300,11 @@ intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) static int intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) { - struct intel_framebuffer *intel_fb; - if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) || (sInfo == NULL)) { return -1; } - intel_fb = dPriv->driverPrivate; - return 0; } -- cgit v1.2.3 From 2edc87eb3d7283274c3d0714e90078736f8d985e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:24:03 -0700 Subject: clean-up/re-org of intel_framebuffer code --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 11 ++++---- src/mesa/drivers/dri/intel_winsys/intel_context.h | 32 +++++++++++++++++++--- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 20 +++++++++++--- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 8 +++--- .../drivers/dri/intel_winsys/intel_swapbuffers.h | 17 ++---------- 5 files changed, 57 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 292c509d19..ac2e71d6dc 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -206,16 +206,17 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, if (driContextPriv) { struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; - struct st_framebuffer *draw_fb - = (struct st_framebuffer *) driDrawPriv->driverPrivate; - struct st_framebuffer *read_fb - = (struct st_framebuffer *) driReadPriv->driverPrivate; + struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv); + struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv); + + assert(draw_fb->stfb); + assert(read_fb->stfb); /* this is a hack so we have a valid context when the region allocation is done. Need a per-screen context? */ intel->intelScreen->dummyctxptr = intel; - st_make_current(intel->st, draw_fb, read_fb); + st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); if ((intel->driDrawable != driDrawPriv) || (intel->lastStamp != driDrawPriv->lastStamp)) { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 3a61a5610f..b2c5bfc9a2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -46,6 +46,9 @@ struct st_context; #define INTEL_MAX_FIXUP 64 +/** + * Intel rendering context, contains a state tracker and intel-specific info. + */ struct intel_context { struct st_context *st; @@ -79,6 +82,22 @@ struct intel_context driOptionCache optionCache; }; + + +/** + * Intel framebuffer. + */ +struct intel_framebuffer +{ + struct st_framebuffer *stfb; + + /* other fields TBD */ + int other; +}; + + + + /* These are functions now: */ void LOCK_HARDWARE( struct intel_context *intel ); @@ -124,10 +143,7 @@ extern int __intel_debug; #define PCI_CHIP_Q33_G 0x29D2 -/*====================================================================== - * Inline conversion functions. - * These are better-typed than the macros used previously: - */ +/** Cast wrapper */ static INLINE struct intel_context * intel_context(GLcontext * ctx) { @@ -135,4 +151,12 @@ intel_context(GLcontext * ctx) } +/** Cast wrapper */ +static INLINE struct intel_framebuffer * +intel_framebuffer(__DRIdrawablePrivate * driDrawPriv) +{ + return (struct intel_framebuffer *) driDrawPriv->driverPrivate; +} + + #endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index cfa61fa67f..c058c2d214 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -29,6 +29,7 @@ #include "vblank.h" #include "xmlpool.h" +#include "intel_context.h" #include "intel_screen.h" #include "intel_batchbuffer.h" #include "intel_swapbuffers.h" @@ -280,8 +281,17 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, return GL_FALSE; /* not implemented */ } else { - struct st_framebuffer *stfb = st_create_framebuffer(mesaVis); - driDrawPriv->driverPrivate = (void *) stfb; + struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); + if (!intelfb) + return GL_FALSE; + + intelfb->stfb = st_create_framebuffer(mesaVis); + if (!intelfb->stfb) { + free(intelfb); + return GL_FALSE; + } + + driDrawPriv->driverPrivate = (void *) intelfb; return GL_TRUE; } } @@ -289,8 +299,10 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, static void intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) { - st_unreference_framebuffer((struct st_framebuffer **) - (&(driDrawPriv->driverPrivate))); + struct intel_framebuffer *intelfb = intel_framebuffer(driDrawPriv); + assert(intelfb->stfb); + st_unreference_framebuffer(&intelfb->stfb); + free(intelfb); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index aef3b87bc8..7083e813b0 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -70,7 +70,7 @@ get_color_surface(struct intel_framebuffer *intel_fb, GLuint bufferIndex) { struct st_renderbuffer *strb - = st_renderbuffer(intel_fb->Base.Attachment[bufferIndex].Renderbuffer); + = st_renderbuffer(intel_fb->stfb->Base.Attachment[bufferIndex].Renderbuffer); if (strb) return strb->surface; return NULL; @@ -236,9 +236,9 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) { - struct st_framebuffer *stfb - = (struct st_framebuffer *) dPriv->driverPrivate; - st_resize_framebuffer(stfb, dPriv->w, dPriv->h); + struct intel_framebuffer *intelfb = intel_framebuffer(dPriv); + assert(intelfb->stfb); + st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index b8a4ac97fb..fca006b767 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -25,19 +25,8 @@ * **************************************************************************/ -#ifndef INTEL_BUFFERS_H -#define INTEL_BUFFERS_H - - - -/** - * Intel framebuffer, derived from gl_framebuffer. - */ -struct intel_framebuffer -{ - struct gl_framebuffer Base; - -}; +#ifndef INTEL_SWAPBUFFERS_H +#define INTEL_SWAPBUFFERS_H extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, @@ -52,4 +41,4 @@ extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); -#endif /* INTEL_BUFFERS_H */ +#endif /* INTEL_SWAPBUFFERS_H */ -- cgit v1.2.3 From 29c1fdd0ce57a7a69e4d90eaacff7102b69bca9c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:28:28 -0700 Subject: better use of intel_context() cast wrappers --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 9 +++------ src/mesa/drivers/dri/intel_winsys/intel_context.h | 10 +++++++++- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 7 +++---- 3 files changed, 15 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index ac2e71d6dc..c1169c621f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -163,8 +163,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, void intelDestroyContext(__DRIcontextPrivate * driContextPriv) { - struct intel_context *intel = - (struct intel_context *) driContextPriv->driverPrivate; + struct intel_context *intel = intel_context(driContextPriv); assert(intel); /* should never be null */ if (intel) { @@ -191,8 +190,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv) { - struct intel_context *intel - = (struct intel_context *) driContextPriv->driverPrivate; + struct intel_context *intel = intel_context(driContextPriv); st_flush(intel->st); return GL_TRUE; } @@ -204,8 +202,7 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driReadPriv) { if (driContextPriv) { - struct intel_context *intel - = (struct intel_context *) driContextPriv->driverPrivate; + struct intel_context *intel = intel_context(driContextPriv); struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv); struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index b2c5bfc9a2..32fb37dd94 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -145,12 +145,20 @@ extern int __intel_debug; /** Cast wrapper */ static INLINE struct intel_context * -intel_context(GLcontext * ctx) +intel_context_mesa(GLcontext * ctx) { return (struct intel_context *) ctx->DriverCtx; } +/** Cast wrapper */ +static INLINE struct intel_context * +intel_context(__DRIcontextPrivate *driContextPriv) +{ + return (struct intel_context *) driContextPriv->driverPrivate; +} + + /** Cast wrapper */ static INLINE struct intel_framebuffer * intel_framebuffer(__DRIdrawablePrivate * driDrawPriv) diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 7083e813b0..b1cbebb944 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -58,7 +58,7 @@ intelScreenContext(intelScreenPrivate *intelScreen) return intelScreen->dummyctxptr; } - return intel_context(ctx); + return intel_context_mesa(ctx); } @@ -117,7 +117,7 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, /* if this drawable isn't currently bound the LOCK_HARDWARE done on the current context (which is what intelScreenContext should return) might not get a contended lock and thus cliprects not updated (tests/manywin) */ - if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) + if (intel_context(dPriv->driContextPriv) != intel) DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); @@ -277,8 +277,7 @@ void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) { if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - struct intel_context *intel = - (struct intel_context *) dPriv->driContextPriv->driverPrivate; + struct intel_context *intel = intel_context(dPriv->driContextPriv); GLcontext *ctx = intel->st->ctx; if (ctx->Visual.doubleBufferMode) { -- cgit v1.2.3 From f221ea658b8aabefb419ad19826906a3afa3e806 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:35:31 -0700 Subject: comments, tweaks --- src/mesa/state_tracker/st_context.c | 10 ++++++++-- src/mesa/state_tracker/st_public.h | 2 +- 2 files changed, 9 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index fbce19b58e..a8cc618d77 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -68,8 +68,11 @@ void st_invalidate_state(GLcontext * ctx, GLuint new_state) } +/* + * XXX rename after above func is removed. + */ struct st_context *st_create_context2(struct pipe_context *pipe, - const GLvisual *visual, + const __GLcontextModes *visual, struct st_context *share) { GLcontext *ctx; @@ -129,6 +132,9 @@ struct st_context *st_create_context( GLcontext *ctx, } +/* + * XXX rename after below func is removed. + */ void st_destroy_context2( struct st_context *st ) { GLcontext *ctx = st->ctx; @@ -151,7 +157,7 @@ void st_destroy_context( struct st_context *st ) _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache); st->pipe->destroy( st->pipe ); - FREE( st ); + free( st ); } diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index abdfcca80f..93239e66d6 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -39,7 +39,7 @@ struct st_context *st_create_context( GLcontext *ctx, struct pipe_context *pipe); struct st_context *st_create_context2(struct pipe_context *pipe, - const GLvisual *visual, + const __GLcontextModes *visual, struct st_context *share); void st_destroy_context( struct st_context *st ); -- cgit v1.2.3 From 616bf3556939d94852b1dfe73f92ef94b10782c4 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:52:14 -0700 Subject: added st_get_framebuffer_surface() and ST_SURFACE_x tokens --- src/mesa/state_tracker/st_framebuffer.c | 20 ++++++++++++++++++++ src/mesa/state_tracker/st_public.h | 10 ++++++++++ 2 files changed, 30 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 54ae175e60..326773c505 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -112,3 +112,23 @@ void st_unreference_framebuffer( struct st_framebuffer **stfb ) } + +/** + * Return the pipe_surface for the given renderbuffer. + */ +struct pipe_surface * +st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex) +{ + struct st_renderbuffer *strb; + + assert(surfIndex <= ST_SURFACE_BACK_RIGHT); + + /* sanity checks, ST tokens should match Mesa tokens */ + assert(ST_SURFACE_FRONT_LEFT == BUFFER_FRONT_LEFT); + assert(ST_SURFACE_BACK_RIGHT == BUFFER_BACK_RIGHT); + + strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer); + if (strb) + return strb->surface; + return NULL; +} diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 93239e66d6..acbf54d5ab 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -30,6 +30,13 @@ #include "mtypes.h" + +#define ST_SURFACE_FRONT_LEFT 0 +#define ST_SURFACE_BACK_LEFT 1 +#define ST_SURFACE_FRONT_RIGHT 2 +#define ST_SURFACE_BACK_RIGHT 3 + + struct st_context; struct st_framebuffer; @@ -51,6 +58,9 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ); void st_resize_framebuffer( struct st_framebuffer *stfb, GLuint width, GLuint height ); +struct pipe_surface *st_get_framebuffer_surface(struct st_framebuffer *stfb, + uint surfIndex); + void st_unreference_framebuffer( struct st_framebuffer **stfb ); void st_make_current(struct st_context *st, -- cgit v1.2.3 From 9fc03fed01e3988f6cf07bb9c041328232b7bacf Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:53:02 -0700 Subject: replace get_color_surface() with st_get_framebuffer_surface() --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 1 + src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 19 ++----------------- 2 files changed, 3 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index c1169c621f..565e377669 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -192,6 +192,7 @@ intelUnbindContext(__DRIcontextPrivate * driContextPriv) { struct intel_context *intel = intel_context(driContextPriv); st_flush(intel->st); + /* XXX make_current(NULL)? */ return GL_TRUE; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index b1cbebb944..0d5ceec1e5 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -62,21 +62,6 @@ intelScreenContext(intelScreenPrivate *intelScreen) } -/** - * Return the pipe_surface for the given renderbuffer. - */ -static struct pipe_surface * -get_color_surface(struct intel_framebuffer *intel_fb, - GLuint bufferIndex) -{ - struct st_renderbuffer *strb - = st_renderbuffer(intel_fb->stfb->Base.Attachment[bufferIndex].Renderbuffer); - if (strb) - return strb->surface; - return NULL; -} - - /** * Display a colorbuffer surface in an X window. * Used for SwapBuffers and flushing front buffer rendering. @@ -255,7 +240,7 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) if (ctx->Visual.doubleBufferMode) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; struct pipe_surface *back_surf - = get_color_surface(intel_fb, BUFFER_BACK_LEFT); + = st_get_framebuffer_surface(intel_fb->stfb, ST_SURFACE_BACK_LEFT); _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ @@ -283,7 +268,7 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) if (ctx->Visual.doubleBufferMode) { struct intel_framebuffer *intel_fb = dPriv->driverPrivate; struct pipe_surface *back_surf - = get_color_surface(intel_fb, BUFFER_BACK_LEFT); + = st_get_framebuffer_surface(intel_fb->stfb, ST_SURFACE_BACK_LEFT); drm_clip_rect_t rect; /* fixup cliprect (driDrawable may have changed?) later */ -- cgit v1.2.3 From 74866737db739bef8b7593d5060e4f80d91ff33a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 09:59:34 -0700 Subject: added st_notify_swapbuffers() --- src/mesa/state_tracker/st_framebuffer.c | 17 +++++++++++++++++ src/mesa/state_tracker/st_public.h | 2 ++ 2 files changed, 19 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 326773c505..2e7687d9c9 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -132,3 +132,20 @@ st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex) return strb->surface; return NULL; } + + +/** + * This function is to be called prior to SwapBuffers on the given + * framebuffer. It checks if the current context is bound to the framebuffer + * and flushes rendering if needed. + */ +void +st_notify_swapbuffers(struct st_framebuffer *stfb) +{ + GET_CURRENT_CONTEXT(ctx); + + if (ctx && ctx->DrawBuffer == &stfb->Base) { + st_flush(ctx->st); + } +} + diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index acbf54d5ab..8162ceeffa 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -71,4 +71,6 @@ void st_invalidate_state(GLcontext * ctx, GLuint new_state); void st_flush( struct st_context *st ); +void st_notify_swapbuffers(struct st_framebuffer *stfb); + #endif -- cgit v1.2.3 From 81c1993daa57ef0fc1fc6993aa832f3b4005154b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 10:01:51 -0700 Subject: rewrite/simplify intelSwapBuffers() --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 25 ++++++++-------------- 1 file changed, 9 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 0d5ceec1e5..bc1ad04395 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -231,25 +231,18 @@ intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) void intelSwapBuffers(__DRIdrawablePrivate * dPriv) { - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - GET_CURRENT_CONTEXT(ctx); - - if (ctx == NULL) - return; + struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); + struct pipe_surface *back_surf; - if (ctx->Visual.doubleBufferMode) { - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - struct pipe_surface *back_surf - = st_get_framebuffer_surface(intel_fb->stfb, ST_SURFACE_BACK_LEFT); + assert(intel_fb); + assert(intel_fb->stfb); - _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ + st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, NULL); - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { + intelDisplaySurface(dPriv, back_surf, NULL); } } -- cgit v1.2.3 From 413688415ca14b8b58922c99d450507a012cd42d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 10:04:23 -0700 Subject: rewrite/simplify intelCopySubBuffer() --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 42 ++++++++++------------ 1 file changed, 18 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index bc1ad04395..b085be49d7 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -237,11 +237,10 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) assert(intel_fb); assert(intel_fb->stfb); - st_notify_swapbuffers(intel_fb->stfb); - back_surf = st_get_framebuffer_surface(intel_fb->stfb, ST_SURFACE_BACK_LEFT); if (back_surf) { + st_notify_swapbuffers(intel_fb->stfb); intelDisplaySurface(dPriv, back_surf, NULL); } } @@ -254,27 +253,22 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) { - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - struct intel_context *intel = intel_context(dPriv->driContextPriv); - GLcontext *ctx = intel->st->ctx; - - if (ctx->Visual.doubleBufferMode) { - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - struct pipe_surface *back_surf - = st_get_framebuffer_surface(intel_fb->stfb, ST_SURFACE_BACK_LEFT); - - drm_clip_rect_t rect; - /* fixup cliprect (driDrawable may have changed?) later */ - rect.x1 = x; - rect.y1 = y; - rect.x2 = w; - rect.y2 = h; - _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - intelDisplaySurface(dPriv, back_surf, &rect); - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); + struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); + struct pipe_surface *back_surf; + + assert(intel_fb); + assert(intel_fb->stfb); + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { + drm_clip_rect_t rect; + rect.x1 = x; + rect.y1 = y; + rect.x2 = w; + rect.y2 = h; + + st_notify_swapbuffers(intel_fb->stfb); + intelDisplaySurface(dPriv, back_surf, &rect); } } -- cgit v1.2.3 From 897428c4265de7212ee1fc345e8689b775b287e8 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 10:14:24 -0700 Subject: comments, clean-ups, assertions --- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index b085be49d7..72ddfec40e 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -39,6 +39,7 @@ #include "state_tracker/st_cb_fbo.h" +/** XXX temporary - want to get rid of this */ static struct intel_context * intelScreenContext(intelScreenPrivate *intelScreen) { @@ -71,9 +72,9 @@ intelScreenContext(intelScreenPrivate *intelScreen) * \param rect optional subrect of surface to display (may be NULL). */ void -intelDisplaySurface(__DRIdrawablePrivate * dPriv, +intelDisplaySurface(__DRIdrawablePrivate *dPriv, struct pipe_surface *surf, - const drm_clip_rect_t * rect) + const drm_clip_rect_t *rect) { struct intel_context *intel; const intelScreenPrivate *intelScreen @@ -100,8 +101,9 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, */ LOCK_HARDWARE(intel); /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - current context (which is what intelScreenContext should return) might - not get a contended lock and thus cliprects not updated (tests/manywin) */ + * current context (which is what intelScreenContext should return) might + * not get a contended lock and thus cliprects not updated (tests/manywin) + */ if (intel_context(dPriv->driContextPriv) != intel) DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); @@ -113,10 +115,10 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, const drm_clip_rect_t *pbox = dPriv->pClipRects; const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; const int cpp = intelScreen->front.cpp; + const struct pipe_region *srcRegion = surf->region; + const int srcpitch = srcRegion->pitch; int BR13, CMD; int i; - const struct pipe_region *srcRegion = surf->region; - const int srcpitch= srcRegion->pitch; ASSERT(srcRegion); ASSERT(srcRegion->cpp == cpp); @@ -181,6 +183,10 @@ intelDisplaySurface(__DRIdrawablePrivate * dPriv, sbox.x1 = box.x1 - dPriv->x; sbox.y1 = box.y1 - dPriv->y; + assert(box.x1 < box.x2); + assert(box.y1 < box.y2); + + /* XXX this could be done with pipe->region_copy() */ BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); OUT_BATCH(CMD); OUT_BATCH(BR13); -- cgit v1.2.3 From 6a95fe87350ed504cdf47dea2331f679db1ba3b4 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 10:16:28 -0700 Subject: add missing free() in intelDestroyContext --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 565e377669..9e6f95abb0 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -153,7 +153,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, } } - intel->st = st_create_context2(pipe, mesaVis, NULL); + intel->st = st_create_context2(pipe, mesaVis, NULL); intel->st->ctx->DriverCtx = intel; /* hope to get rid of this... */ return GL_TRUE; @@ -183,6 +183,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) } st_destroy_context2(intel->st); + free(intel); } } -- cgit v1.2.3 From f8f1c4a02205da3fb95d68049c9a9cfd6b0c751d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 11:09:19 -0700 Subject: remove unused intelScreen->texPool --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 2 -- src/mesa/drivers/dri/intel_winsys/intel_screen.h | 1 - 2 files changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index c058c2d214..332dc46461 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -176,8 +176,6 @@ intelCreatePools(intelScreenPrivate *intelScreen) if (!intelScreen->staticPool) return GL_FALSE; - intelScreen->texPool = intelScreen->regionPool; - intelScreen->batchPool = driBatchPoolInit(sPriv->fd, DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT | diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index 4dd10b08cb..09e4971f89 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -64,7 +64,6 @@ struct intel_screen */ driOptionCache optionCache; struct _DriBufferPool *batchPool; - struct _DriBufferPool *texPool; struct _DriBufferPool *regionPool; struct _DriBufferPool *staticPool; GLboolean havePools; -- cgit v1.2.3 From 1bbc86e6514a80cba6d48b8dc8615e7dc1be7afa Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 11:17:40 -0700 Subject: fix typo, formatting --- src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h index 6d4d05e0bb..620fafca48 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h @@ -55,7 +55,7 @@ void intel_batchbuffer_reset(struct intel_batchbuffer *batch); /* Unlike bmBufferData, this currently requires the buffer be mapped. - * Consider it a convenience function wrapping multple + * Consider it a convenience function wrapping multiple * intel_buffer_dword() calls. */ void intel_batchbuffer_data(struct intel_batchbuffer *batch, @@ -113,7 +113,7 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch, #define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) #define OUT_RELOC(buf,flags,mask,delta) do { \ - assert((delta) >= 0); \ + assert((delta) >= 0); \ intel_batchbuffer_emit_reloc(intel->batch, buf, flags, mask, delta); \ } while (0) -- cgit v1.2.3 From cd360b7d60721883390215ac33623636a222cde2 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 11:35:43 -0700 Subject: comments --- src/mesa/drivers/dri/intel_winsys/intel_screen.h | 4 ++-- src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index 09e4971f89..c0e7ff8c5d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -58,14 +58,14 @@ struct intel_screen __DRIscreenPrivate *driScrnPriv; drmI830Sarea *sarea; - /** * Configuration cache with default values for all contexts */ driOptionCache optionCache; + struct _DriBufferPool *batchPool; struct _DriBufferPool *regionPool; - struct _DriBufferPool *staticPool; + struct _DriBufferPool *staticPool; /** for the X screen/framebuffer */ GLboolean havePools; struct intel_context *dummyctxptr; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index 022fb66927..4fb8f02daa 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -43,7 +43,7 @@ struct intel_i915_winsys { - struct i915_winsys winsys; + struct i915_winsys winsys; /**< batch buffer funcs */ struct intel_context *intel; }; -- cgit v1.2.3 From 736baf22e98a28ed714c31650d1cee78b5b381f4 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 11:38:46 -0700 Subject: remove driScrnPriv field from intel_screen --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 +- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 5 ++--- src/mesa/drivers/dri/intel_winsys/intel_screen.h | 3 +-- 3 files changed, 4 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 9e6f95abb0..91ac077a48 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -102,7 +102,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, * memory pools */ DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - havePools = intelCreatePools(intelScreen); + havePools = intelCreatePools(sPriv); DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); if (!havePools) return GL_FALSE; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 332dc46461..646c4ec911 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -157,10 +157,10 @@ intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) GLboolean -intelCreatePools(intelScreenPrivate *intelScreen) +intelCreatePools(__DRIscreenPrivate * sPriv) { unsigned batchPoolSize = 1024*1024; - __DRIscreenPrivate * sPriv = intelScreen->driScrnPriv; + intelScreenPrivate *intelScreen = sPriv->private; if (intelScreen->havePools) return GL_TRUE; @@ -221,7 +221,6 @@ intelInitDriver(__DRIscreenPrivate * sPriv) driParseOptionInfo(&intelScreen->optionCache, __driConfigOptions, __driNConfigOptions); - intelScreen->driScrnPriv = sPriv; sPriv->private = (void *) intelScreen; intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index c0e7ff8c5d..aab745ea4d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -55,7 +55,6 @@ struct intel_screen int deviceID; int drmMinor; - __DRIscreenPrivate *driScrnPriv; drmI830Sarea *sarea; /** @@ -92,7 +91,7 @@ extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, unsigned checkDelayed); extern GLboolean -intelCreatePools(intelScreenPrivate *intelScreen); +intelCreatePools(__DRIscreenPrivate *sPriv); extern GLboolean intelCreateContext(const __GLcontextModes * mesaVis, -- cgit v1.2.3 From adf3761fbd193e4467b0637340709e57bdae0111 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 11:48:54 -0700 Subject: remove intelScreenPrivate typedef and use cast wrapper function to be like intel_context and intel_framebuffer --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 +- src/mesa/drivers/dri/intel_winsys/intel_context.h | 2 +- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 2 +- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 12 ++++++------ src/mesa/drivers/dri/intel_winsys/intel_screen.h | 12 +++++++++++- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 5 ++--- 6 files changed, 22 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 91ac077a48..b7bfc39f89 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -83,7 +83,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, struct intel_context *intel = CALLOC_STRUCT(intel_context); __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + struct intel_screen *intelScreen = intel_screen(sPriv); drmI830Sarea *saPriv = intelScreen->sarea; int fthrottle_mode; GLboolean havePools; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 32fb37dd94..9d24ca124a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -71,7 +71,7 @@ struct intel_context __DRIdrawablePrivate *driDrawable; __DRIscreenPrivate *driScreen; - intelScreenPrivate *intelScreen; + struct intel_screen *intelScreen; drmI830Sarea *sarea; GLuint lastStamp; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 5962b9259d..74a7a55bd0 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -45,7 +45,7 @@ intelContendedLock(struct intel_context *intel, GLuint flags) { __DRIdrawablePrivate *dPriv = intel->driDrawable; __DRIscreenPrivate *sPriv = intel->driScreen; - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + struct intel_screen *intelScreen = intel_screen(sPriv); drmI830Sarea *sarea = intel->sarea; drmGetLock(intel->driFd, intel->hHWContext, flags); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 646c4ec911..feaeb181f4 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -64,7 +64,7 @@ PUBLIC const char __driConfigOptions[] = static void -intelPrintDRIInfo(intelScreenPrivate * intelScreen, +intelPrintDRIInfo(struct intel_screen * intelScreen, __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) { fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", @@ -110,7 +110,7 @@ intelPrintSAREA(const drmI830Sarea * sarea) void intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) { - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + struct intel_screen *intelScreen = intel_screen(sPriv); if (intelScreen->front.map) { drmUnmap(intelScreen->front.map, intelScreen->front.size); @@ -160,7 +160,7 @@ GLboolean intelCreatePools(__DRIscreenPrivate * sPriv) { unsigned batchPoolSize = 1024*1024; - intelScreenPrivate *intelScreen = sPriv->private; + struct intel_screen *intelScreen = intel_screen(sPriv); if (intelScreen->havePools) return GL_TRUE; @@ -198,7 +198,7 @@ intelCreatePools(__DRIscreenPrivate * sPriv) static GLboolean intelInitDriver(__DRIscreenPrivate * sPriv) { - intelScreenPrivate *intelScreen; + struct intel_screen *intelScreen; I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = @@ -213,7 +213,7 @@ intelInitDriver(__DRIscreenPrivate * sPriv) } /* Allocate the private area */ - intelScreen = (intelScreenPrivate *) CALLOC(sizeof(intelScreenPrivate)); + intelScreen = CALLOC_STRUCT(intel_screen); if (!intelScreen) return GL_FALSE; @@ -252,7 +252,7 @@ intelInitDriver(__DRIscreenPrivate * sPriv) static void intelDestroyScreen(__DRIscreenPrivate * sPriv) { - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; + struct intel_screen *intelScreen = intel_screen(sPriv); // intelUnmapScreenRegions(intelScreen); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index aab745ea4d..b0d1e8e36a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -33,6 +33,8 @@ #include "xmlconfig.h" #include "dri_bufpool.h" +#include "pipe/p_compiler.h" + struct intel_screen { @@ -70,7 +72,15 @@ struct intel_screen struct intel_context *dummyctxptr; }; -typedef struct intel_screen intelScreenPrivate; + + +/** cast wrapper */ +static INLINE struct intel_screen * +intel_screen(__DRIscreenPrivate *sPriv) +{ + return (struct intel_screen *) sPriv->private; +} + extern void intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 72ddfec40e..f3bbbe874d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -41,7 +41,7 @@ /** XXX temporary - want to get rid of this */ static struct intel_context * -intelScreenContext(intelScreenPrivate *intelScreen) +intelScreenContext(struct intel_screen *intelScreen) { /* * This should probably change to have the screen allocate a dummy @@ -77,8 +77,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, const drm_clip_rect_t *rect) { struct intel_context *intel; - const intelScreenPrivate *intelScreen - = (intelScreenPrivate *) dPriv->driScreenPriv->private; + const struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); DBG(SWAP, "%s\n", __FUNCTION__); -- cgit v1.2.3 From b17675d70976bd40ff58ea49226292f49cffbe13 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 11:53:56 -0700 Subject: Move driBatchPoolInit() into new intel_batchpool.h header. Note that intel_batchpool.[ch] have no intel-specific dependencies at this point. Maybe rename files for re-use in the future. --- .../drivers/dri/intel_winsys/intel_batchpool.h | 37 ++++++++++++++++++++++ 1 file changed, 37 insertions(+) create mode 100644 src/mesa/drivers/dri/intel_winsys/intel_batchpool.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.h b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.h new file mode 100644 index 0000000000..f6a95723bc --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BATCHPOOL_H +#define INTEL_BATCHPOOL_H + +extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, + unsigned long bufSize, + unsigned numBufs, + unsigned checkDelayed); + + +#endif /* INTEL_BATCHPOOL_H */ -- cgit v1.2.3 From 6ad4656cc475b170f419627ff6c4a43ebab09b7c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 11:54:36 -0700 Subject: Move driBatchPoolInit() into new intel_batchpool.h header. Note that intel_batchpool.[ch] have no intel-specific dependencies at this poi Maybe rename files for re-use in the future. --- src/mesa/drivers/dri/intel_winsys/intel_batchpool.c | 8 +++++++- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 1 + src/mesa/drivers/dri/intel_winsys/intel_screen.h | 4 ---- 3 files changed, 8 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c index 2503b8a62a..a81a463baa 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c @@ -29,6 +29,11 @@ * Authors: Thomas Hellström */ +/** + * XXX NOTE: there are no intel dependencies in this file. + * Rename to dri_batchpool.c? + */ + #include #include #include @@ -36,7 +41,8 @@ #include "glthread.h" #include "dri_bufpool.h" #include "dri_bufmgr.h" -#include "intel_screen.h" +#include "intel_batchpool.h" + typedef struct { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index feaeb181f4..c14e84ebd6 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -32,6 +32,7 @@ #include "intel_context.h" #include "intel_screen.h" #include "intel_batchbuffer.h" +#include "intel_batchpool.h" #include "intel_swapbuffers.h" #include "i830_dri.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index b0d1e8e36a..87a6cb21d1 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -95,10 +95,6 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv); -extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, - unsigned long bufSize, - unsigned numBufs, - unsigned checkDelayed); extern GLboolean intelCreatePools(__DRIscreenPrivate *sPriv); -- cgit v1.2.3 From 3f80be3cb5dcd1378316c0330446a72a2095d9ca Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 11:59:09 -0700 Subject: move intel_context_mesa() into intel_swapbuffers.c (the only place it's used now) --- src/mesa/drivers/dri/intel_winsys/intel_context.h | 8 -------- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 9 +++++++++ 2 files changed, 9 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 9d24ca124a..5aa07d4ed1 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -143,14 +143,6 @@ extern int __intel_debug; #define PCI_CHIP_Q33_G 0x29D2 -/** Cast wrapper */ -static INLINE struct intel_context * -intel_context_mesa(GLcontext * ctx) -{ - return (struct intel_context *) ctx->DriverCtx; -} - - /** Cast wrapper */ static INLINE struct intel_context * intel_context(__DRIcontextPrivate *driContextPriv) diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index f3bbbe874d..8a35a71669 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -39,6 +39,15 @@ #include "state_tracker/st_cb_fbo.h" + +/** Cast wrapper */ +static INLINE struct intel_context * +intel_context_mesa(GLcontext * ctx) +{ + return (struct intel_context *) ctx->DriverCtx; +} + + /** XXX temporary - want to get rid of this */ static struct intel_context * intelScreenContext(struct intel_screen *intelScreen) -- cgit v1.2.3 From e4bad270998e0b97bd97e8e38b08826ed45e2f5e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 13:18:47 -0700 Subject: remove some GL types, mesa includes --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 +- src/mesa/drivers/dri/intel_winsys/intel_context.h | 9 +++------ src/mesa/drivers/dri/intel_winsys/intel_lock.c | 2 +- src/mesa/drivers/dri/intel_winsys/intel_screen.h | 2 +- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h | 3 +++ 5 files changed, 9 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index b7bfc39f89..2a29c2ab27 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -35,7 +35,7 @@ #include "intel_batchbuffer.h" #include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" +#include "state_tracker/st_context.h" /* XXX temporary */ #include "pipe/p_defines.h" #include "pipe/p_context.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 5aa07d4ed1..15d849d52b 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -29,10 +29,7 @@ #define INTELCONTEXT_INC - -#include "mtypes.h" #include "drm.h" - #include "intel_screen.h" #include "i915_drm.h" @@ -58,11 +55,11 @@ struct intel_context struct intel_batchbuffer *batch; - GLboolean locked; + boolean locked; char *prevLockFile; int prevLockLine; - GLuint irqsEmitted; + uint irqsEmitted; drm_i915_irq_wait_t iw; drm_context_t hHWContext; @@ -74,7 +71,7 @@ struct intel_context struct intel_screen *intelScreen; drmI830Sarea *sarea; - GLuint lastStamp; + uint lastStamp; /** * Configuration cache diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 74a7a55bd0..65d3eb3b3a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -41,7 +41,7 @@ _glthread_DECLARE_STATIC_MUTEX( lockMutex ); static void -intelContendedLock(struct intel_context *intel, GLuint flags) +intelContendedLock(struct intel_context *intel, uint flags) { __DRIdrawablePrivate *dPriv = intel->driDrawable; __DRIscreenPrivate *sPriv = intel->driScreen; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index 87a6cb21d1..1d24ff7ff7 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -67,7 +67,7 @@ struct intel_screen struct _DriBufferPool *batchPool; struct _DriBufferPool *regionPool; struct _DriBufferPool *staticPool; /** for the X screen/framebuffer */ - GLboolean havePools; + boolean havePools; struct intel_context *dummyctxptr; }; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h index fca006b767..7ae5fd15a5 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h @@ -29,6 +29,9 @@ #define INTEL_SWAPBUFFERS_H +struct pipe_surface; + + extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, struct pipe_surface *surf, const drm_clip_rect_t * rect); -- cgit v1.2.3 From f4d51d8923db9fd71e5f9fe965769625bd0d1240 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 13:37:12 -0700 Subject: asst clean-ups --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 1 - src/mesa/drivers/dri/intel_winsys/intel_context.h | 4 ++-- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 8 -------- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 10 +++++----- src/mesa/drivers/dri/intel_winsys/intel_winsys.h | 4 ++-- 5 files changed, 9 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 2a29c2ab27..607d316e36 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -40,7 +40,6 @@ #include "pipe/p_context.h" #include "utils.h" -#include "xmlpool.h" /* for symbolic values of enum-type options */ #ifdef DEBUG diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index 15d849d52b..eac5cd8b4c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -25,8 +25,8 @@ * **************************************************************************/ -#ifndef INTELCONTEXT_INC -#define INTELCONTEXT_INC +#ifndef INTEL_CONTEXT_H +#define INTEL_CONTEXT_H #include "drm.h" diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 65d3eb3b3a..55606dba8b 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -27,12 +27,7 @@ #include "state_tracker/st_public.h" - -#include "intel_screen.h" #include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_swapbuffers.h" - #include "i830_dri.h" @@ -73,13 +68,10 @@ intelContendedLock(struct intel_context *intel, uint flags) void LOCK_HARDWARE( struct intel_context *intel ) { char __ret=0; - int curbuf; _glthread_LOCK_MUTEX(lockMutex); assert(!intel->locked); - curbuf = 0; /* current draw buf: 0 = front, 1 = back */ - DRM_CAS(intel->driHwLock, intel->hHWContext, (DRM_LOCK_HELD|intel->hHWContext), __ret); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index c14e84ebd6..fa5c19ecd7 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -52,14 +52,14 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_FORCE_S3TC_ENABLE(false) DRI_CONF_ALLOW_LARGE_TEXTURES(1) DRI_CONF_SECTION_END DRI_CONF_END; - const GLuint __driNConfigOptions = 4; + +const GLuint __driNConfigOptions = 4; #ifdef USE_NEW_INTERFACE - static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; +static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; #endif /*USE_NEW_INTERFACE */ - extern const struct dri_extension card_extensions[]; - +extern const struct dri_extension card_extensions[]; @@ -255,7 +255,7 @@ intelDestroyScreen(__DRIscreenPrivate * sPriv) { struct intel_screen *intelScreen = intel_screen(sPriv); -// intelUnmapScreenRegions(intelScreen); + /* intelUnmapScreenRegions(intelScreen); */ if (intelScreen->havePools) { driPoolTakeDown(intelScreen->regionPool); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h index 2c8aa20c2a..78f5cb6519 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h @@ -25,8 +25,8 @@ * **************************************************************************/ -#ifndef INTEL_PIPE_H -#define INTEL_PIPE_H +#ifndef INTEL_WINSYS_H +#define INTEL_WINSYS_H struct intel_context; struct pipe_context; -- cgit v1.2.3 From 9a563d5e696a7c8fc09f7da5a0d33a9675b00e4c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 15:42:55 -0700 Subject: no-op glCopyPixels if width or height is zero --- src/mesa/main/drawpix.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c index e4be0d496e..ae9c7e29a1 100644 --- a/src/mesa/main/drawpix.c +++ b/src/mesa/main/drawpix.c @@ -239,7 +239,7 @@ _mesa_CopyPixels( GLint srcx, GLint srcy, GLsizei width, GLsizei height, return; } - if (!ctx->Current.RasterPosValid) { + if (!ctx->Current.RasterPosValid || width ==0 || height == 0) { return; } -- cgit v1.2.3 From e39f1b4cbea6d41dc09430bd3d811cb7bbdea5f8 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 15:59:55 -0700 Subject: Update xlib driver to use newer state tracker context/framebuffer functions. XMesaContext has an st_context * which contains a mesa context. --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 2 +- src/mesa/pipe/xlib/fakeglx.c | 17 ++- src/mesa/pipe/xlib/xm_api.c | 160 ++++++++++------------- src/mesa/pipe/xlib/xmesaP.h | 31 ++++- src/mesa/state_tracker/st_context.c | 7 + src/mesa/state_tracker/st_context.h | 3 +- src/mesa/state_tracker/st_framebuffer.c | 84 ++++++------ src/mesa/state_tracker/st_public.h | 9 +- 8 files changed, 170 insertions(+), 143 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index fa5c19ecd7..899988340e 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -283,7 +283,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, if (!intelfb) return GL_FALSE; - intelfb->stfb = st_create_framebuffer(mesaVis); + intelfb->stfb = st_create_framebuffer(mesaVis, GL_TRUE, (void*) intelfb); if (!intelfb->stfb) { free(intelfb); return GL_FALSE; diff --git a/src/mesa/pipe/xlib/fakeglx.c b/src/mesa/pipe/xlib/fakeglx.c index 7a170b4d3d..d7e7b292d1 100644 --- a/src/mesa/pipe/xlib/fakeglx.c +++ b/src/mesa/pipe/xlib/fakeglx.c @@ -51,6 +51,9 @@ #include "version.h" #include "xfonts.h" #include "xmesaP.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_public.h" + #ifdef __VMS #define _mesa_sprintf sprintf @@ -1617,7 +1620,7 @@ Fake_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, if (MakeCurrent_PrevContext == src) { _mesa_Flush(); } - _mesa_copy_context( &(xm_src->mesa), &(xm_dst->mesa), (GLuint) mask ); + st_copy_context_state( xm_src->st, xm_dst->st, (GLuint) mask ); } @@ -2383,16 +2386,16 @@ Fake_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, switch (attribute) { case GLX_WIDTH: - *value = xmbuf->mesa_buffer.Width; + *value = xmesa_buffer_width(xmbuf); break; case GLX_HEIGHT: - *value = xmbuf->mesa_buffer.Height; + *value = xmesa_buffer_width(xmbuf); break; case GLX_PRESERVED_CONTENTS: *value = True; break; case GLX_LARGEST_PBUFFER: - *value = xmbuf->mesa_buffer.Width * xmbuf->mesa_buffer.Height; + *value = xmesa_buffer_width(xmbuf) * xmesa_buffer_height(xmbuf); break; case GLX_FBCONFIG_ID: *value = xmbuf->xm_visual->visinfo->visualid; @@ -2762,13 +2765,13 @@ Fake_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, un *value = True; break; case GLX_LARGEST_PBUFFER_SGIX: - *value = xmbuf->mesa_buffer.Width * xmbuf->mesa_buffer.Height; + *value = xmesa_buffer_width(xmbuf) * xmesa_buffer_height(xmbuf); break; case GLX_WIDTH_SGIX: - *value = xmbuf->mesa_buffer.Width; + *value = xmesa_buffer_width(xmbuf); break; case GLX_HEIGHT_SGIX: - *value = xmbuf->mesa_buffer.Height; + *value = xmesa_buffer_height(xmbuf); break; case GLX_EVENT_MASK_SGIX: *value = 0; /* XXX might be wrong */ diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 7f144a5d32..abd0b4f292 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -300,6 +300,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, XMesaVisual vis, XMesaColormap cmap) { XMesaBuffer b; + GLframebuffer *fb; ASSERT(type == WINDOW || type == PIXMAP || type == PBUFFER); @@ -312,8 +313,13 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, b->type = type; b->cmap = cmap; - _mesa_initialize_framebuffer(&b->mesa_buffer, &vis->mesa_visual); - b->mesa_buffer.Delete = xmesa_delete_framebuffer; + /* + * Create framebuffer, but we'll plug in our own renderbuffers below. + */ + b->stfb = st_create_framebuffer(&vis->mesa_visual, GL_FALSE, (void *) b); + fb = &b->stfb->Base; + + fb->Delete = xmesa_delete_framebuffer; /* * XXX we want to create surfaces for pipe, not renderbuffers for Mesa. @@ -330,8 +336,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, b->frontxrb->Parent = b; b->frontxrb->drawable = d; b->frontxrb->pixmap = (XMesaPixmap) d; - _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_FRONT_LEFT, - &b->frontxrb->St.Base); + _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &b->frontxrb->St.Base); #if 0 /* sketch... */ { struct pipe_surface *front_surf; @@ -339,8 +344,6 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, } #endif - - /* * Back renderbuffer */ @@ -355,8 +358,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, /* determine back buffer implementation */ b->db_mode = vis->ximage_flag ? BACK_XIMAGE : BACK_PIXMAP; - _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_BACK_LEFT, - &b->backxrb->St.Base); + _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &b->backxrb->St.Base); } /* @@ -379,34 +381,34 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, /* combined depth/stencil */ struct gl_renderbuffer *rb = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); - _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_DEPTH, rb); - _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_STENCIL, rb); + _mesa_add_renderbuffer(fb, BUFFER_DEPTH, rb); + _mesa_add_renderbuffer(fb, BUFFER_STENCIL, rb); } else { if (vis->mesa_visual.depthBits > 0) { struct gl_renderbuffer *rb = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32); - _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_DEPTH, rb); + _mesa_add_renderbuffer(fb, BUFFER_DEPTH, rb); } if (vis->mesa_visual.stencilBits > 0) { struct gl_renderbuffer *rb = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT); - _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_STENCIL, rb); + _mesa_add_renderbuffer(fb, BUFFER_STENCIL, rb); } } if (vis->mesa_visual.accumRedBits > 0) { struct gl_renderbuffer *rb = st_new_renderbuffer_fb(GL_RGBA16); - _mesa_add_renderbuffer(&b->mesa_buffer, BUFFER_ACCUM, rb); + _mesa_add_renderbuffer(fb, BUFFER_ACCUM, rb); } /* * Other renderbuffer (depth, stencil, etc) */ - _mesa_add_soft_renderbuffers(&b->mesa_buffer, + _mesa_add_soft_renderbuffers(fb, GL_FALSE, /* color */ GL_FALSE, /*vis->mesa_visual.haveDepthBuffer,*/ GL_FALSE, /* stencil */ @@ -454,7 +456,7 @@ xmesa_free_buffer(XMesaBuffer buffer) for (b = XMesaBufferList; b; b = b->Next) { if (b == buffer) { - struct gl_framebuffer *fb = &buffer->mesa_buffer; + struct gl_framebuffer *fb = &buffer->stfb->Base; /* unlink buffer from list */ if (prev) @@ -1112,11 +1114,12 @@ initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, * Convert an RGBA color to a pixel value. */ unsigned long -xmesa_color_to_pixel(GLcontext *ctx, +xmesa_color_to_pixel(XMesaContext xmesa, GLubyte r, GLubyte g, GLubyte b, GLubyte a, GLuint pixelFormat) { - XMesaContext xmesa = XMESA_CONTEXT(ctx); + GLcontext *ctx = xmesa->st->ctx; + switch (pixelFormat) { case PF_Index: return 0; @@ -1444,9 +1447,11 @@ PUBLIC XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) { static GLboolean firstTime = GL_TRUE; + struct pipe_context *pipe; XMesaContext c; GLcontext *mesaCtx; - struct dd_function_table functions; + + (void) xmesa_init_driver_functions; if (firstTime) { _glthread_INIT_MUTEX(_xmesa_lock); @@ -1458,29 +1463,20 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) if (!c) return NULL; - mesaCtx = &(c->mesa); - - /* initialize with default driver functions, then plug in XMesa funcs */ -#if 0 - _mesa_init_driver_functions(&functions); -#else - memset(&functions, 0, sizeof(functions)); -#endif - xmesa_init_driver_functions(v, &functions); - st_init_driver_functions(&functions); + pipe = xmesa_create_softpipe( c ); - if (!_mesa_initialize_context(mesaCtx, &v->mesa_visual, - share_list ? &(share_list->mesa) : (GLcontext *) NULL, - &functions, (void *) c)) { - _mesa_free(c); - return NULL; - } + c->st = st_create_context2(pipe, &v->mesa_visual, + share_list ? share_list->st : NULL); + mesaCtx = c->st->ctx; + c->st->ctx->DriverCtx = c; +#if 00 _mesa_enable_sw_extensions(mesaCtx); _mesa_enable_1_3_extensions(mesaCtx); _mesa_enable_1_4_extensions(mesaCtx); _mesa_enable_1_5_extensions(mesaCtx); _mesa_enable_2_0_extensions(mesaCtx); +#endif #ifdef XFree86Server /* If we're running in the X server, do bounds checking to prevent @@ -1496,25 +1492,18 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) c->display = v->display; c->pixelformat = v->dithered_pf; /* Dithering is enabled by default */ - st_create_context( mesaCtx, - xmesa_create_softpipe( c ) ); - /* override these functions, as if the xlib driver were derived from * the softpipe driver. */ -#if 0 - mesaCtx->st->pipe->surface_alloc = xmesa_surface_alloc; - mesaCtx->st->pipe->is_format_supported = xmesa_is_format_supported; -#endif - mesaCtx->st->pipe->get_tile = xmesa_get_tile; - mesaCtx->st->pipe->put_tile = xmesa_put_tile; - mesaCtx->st->pipe->get_tile_rgba = xmesa_get_tile_rgba; - mesaCtx->st->pipe->put_tile_rgba = xmesa_put_tile_rgba; + pipe->get_tile = xmesa_get_tile; + pipe->put_tile = xmesa_put_tile; + pipe->get_tile_rgba = xmesa_get_tile_rgba; + pipe->put_tile_rgba = xmesa_put_tile_rgba; - mesaCtx->st->haveFramebufferRegions = GL_FALSE; + c->st->haveFramebufferRegions = GL_FALSE; /* special pipe->clear function */ - mesaCtx->st->pipe->clear = xmesa_clear; + pipe->clear = xmesa_clear; return c; } @@ -1524,10 +1513,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) PUBLIC void XMesaDestroyContext( XMesaContext c ) { - GLcontext *mesaCtx = &c->mesa; - - _mesa_free_context_data( mesaCtx ); - _mesa_free( c ); + st_destroy_context2(c->st); + _mesa_free(c); } @@ -1645,7 +1632,7 @@ XMesaCreatePixmapTextureBuffer(XMesaVisual v, XMesaPixmap p, /* get pixmap size, update framebuffer/renderbuffer dims */ xmesa_get_window_size(v->display, b, &width, &height); - _mesa_resize_framebuffer(NULL, &(b->mesa_buffer), width, height); + _mesa_resize_framebuffer(NULL, &(b->stfb->Base), width, height); if (target == 0) { /* examine dims */ @@ -1746,12 +1733,7 @@ xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer) { GLuint width, height; xmesa_get_window_size(drawBuffer->display, drawBuffer, &width, &height); - if (drawBuffer->mesa_buffer.Width != width || - drawBuffer->mesa_buffer.Height != height) { - GLcontext *ctx = xmctx ? &xmctx->mesa : NULL; - _mesa_resize_framebuffer(ctx, &(drawBuffer->mesa_buffer), width, height); - } - drawBuffer->mesa_buffer.Initialized = GL_TRUE; /* XXX TEMPORARY? */ + st_resize_framebuffer(drawBuffer->stfb, width, height); } @@ -1775,6 +1757,8 @@ GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, if (!drawBuffer || !readBuffer) return GL_FALSE; /* must specify buffers! */ +#if 0 + /* XXX restore this optimization */ if (&(c->mesa) == _mesa_get_current_context() && c->mesa.DrawBuffer == &drawBuffer->mesa_buffer && c->mesa.ReadBuffer == &readBuffer->mesa_buffer @@ -1782,6 +1766,7 @@ GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, /* same context and buffer, do nothing */ return GL_TRUE; } +#endif c->xm_buffer = drawBuffer; @@ -1794,16 +1779,15 @@ GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, if (readBuffer != drawBuffer) xmesa_check_and_update_buffer_size(c, readBuffer); - _mesa_make_current(&(c->mesa), - &drawBuffer->mesa_buffer, - &readBuffer->mesa_buffer); + st_make_current(c->st, drawBuffer->stfb, readBuffer->stfb); +#if 0 if (c->xm_visual->mesa_visual.rgbMode) { /* * Must recompute and set these pixel values because colormap * can be different for different windows. */ - c->clearpixel = xmesa_color_to_pixel( &c->mesa, + c->clearpixel = xmesa_color_to_pixel( c, c->clearcolor[0], c->clearcolor[1], c->clearcolor[2], @@ -1811,13 +1795,14 @@ GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, c->xm_visual->undithered_pf); XMesaSetForeground(c->display, drawBuffer->cleargc, c->clearpixel); } +#endif /* Solution to Stephane Rehel's problem with glXReleaseBuffersMESA(): */ drawBuffer->wasCurrent = GL_TRUE; } else { /* Detach */ - _mesa_make_current( NULL, NULL, NULL ); + st_make_current( NULL, NULL, NULL ); } return GL_TRUE; } @@ -1925,8 +1910,6 @@ GLboolean XMesaSetFXmode( GLint mode ) PUBLIC void XMesaSwapBuffers( XMesaBuffer b ) { - GET_CURRENT_CONTEXT(ctx); - if (!b->backxrb) { /* single buffered */ return; @@ -1935,8 +1918,7 @@ void XMesaSwapBuffers( XMesaBuffer b ) /* If we're swapping the buffer associated with the current context * we have to flush any pending rendering commands first. */ - if (ctx && ctx->DrawBuffer == &(b->mesa_buffer)) - _mesa_notifySwapBuffers(ctx); + st_notify_swapbuffers(b->stfb); if (b->db_mode) { if (b->backxrb->ximage) { @@ -1947,7 +1929,7 @@ void XMesaSwapBuffers( XMesaBuffer b ) XShmPutImage( b->xm_visual->display, b->frontxrb->drawable, b->swapgc, b->backxrb->ximage, 0, 0, - 0, 0, b->mesa_buffer.Width, b->mesa_buffer.Height, + 0, 0, xmesa_buffer_width(b), xmesa_buffer_height(b), False ); /*_glthread_UNLOCK_MUTEX(_xmesa_lock);*/ } @@ -1958,7 +1940,8 @@ void XMesaSwapBuffers( XMesaBuffer b ) XMesaPutImage( b->xm_visual->display, b->frontxrb->drawable, b->swapgc, b->backxrb->ximage, 0, 0, - 0, 0, b->mesa_buffer.Width, b->mesa_buffer.Height ); + 0, 0, + xmesa_buffer_width(b), xmesa_buffer_height(b)); /*_glthread_UNLOCK_MUTEX(_xmesa_lock);*/ } } @@ -1969,14 +1952,16 @@ void XMesaSwapBuffers( XMesaBuffer b ) b->backxrb->pixmap, /* source drawable */ b->frontxrb->drawable, /* dest. drawable */ b->swapgc, - 0, 0, b->mesa_buffer.Width, b->mesa_buffer.Height, + 0, 0, xmesa_buffer_width(b), xmesa_buffer_height(b), 0, 0 /* dest region */ ); /*_glthread_UNLOCK_MUTEX(_xmesa_lock);*/ } - if (b->swAlpha) - _mesa_copy_soft_alpha_renderbuffers(ctx, &b->mesa_buffer); + if (b->swAlpha) { + GET_CURRENT_CONTEXT(ctx); + _mesa_copy_soft_alpha_renderbuffers(ctx, &b->stfb->Base); + } } #if !defined(XFree86Server) XSync( b->xm_visual->display, False ); @@ -1990,21 +1975,18 @@ void XMesaSwapBuffers( XMesaBuffer b ) */ void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height ) { - GET_CURRENT_CONTEXT(ctx); - - /* If we're swapping the buffer associated with the current context - * we have to flush any pending rendering commands first. - */ - if (ctx && ctx->DrawBuffer == &(b->mesa_buffer)) - _mesa_notifySwapBuffers(ctx); - if (!b->backxrb) { /* single buffered */ return; } + /* If we're swapping the buffer associated with the current context + * we have to flush any pending rendering commands first. + */ + st_notify_swapbuffers(b->stfb); + if (b->db_mode) { - int yTop = b->mesa_buffer.Height - y - height; + int yTop = xmesa_buffer_height(b) - y - height; if (b->backxrb->ximage) { /* Copy Ximage from host's memory to server's window */ #if defined(USE_XSHM) && !defined(XFree86Server) @@ -2082,7 +2064,7 @@ GLboolean XMesaGetDepthBuffer( XMesaBuffer b, GLint *width, GLint *height, GLint *bytesPerValue, void **buffer ) { struct gl_renderbuffer *rb - = b->mesa_buffer.Attachment[BUFFER_DEPTH].Renderbuffer; + = b->stfb->Base.Attachment[BUFFER_DEPTH].Renderbuffer; if (!rb || !rb->Data) { *width = 0; *height = 0; @@ -2091,9 +2073,9 @@ GLboolean XMesaGetDepthBuffer( XMesaBuffer b, GLint *width, GLint *height, return GL_FALSE; } else { - *width = b->mesa_buffer.Width; - *height = b->mesa_buffer.Height; - *bytesPerValue = b->mesa_buffer.Visual.depthBits <= 16 + *width = xmesa_buffer_width(b); + *height = xmesa_buffer_height(b); + *bytesPerValue = b->stfb->Base.Visual.depthBits <= 16 ? sizeof(GLushort) : sizeof(GLuint); *buffer = rb->Data; return GL_TRUE; @@ -2103,11 +2085,11 @@ GLboolean XMesaGetDepthBuffer( XMesaBuffer b, GLint *width, GLint *height, void XMesaFlush( XMesaContext c ) { - if (c && c->xm_visual) { + if (c && c->display) { #ifdef XFree86Server /* NOT_NEEDED */ #else - XSync( c->xm_visual->display, False ); + XSync( c->display, False ); #endif } } @@ -2185,7 +2167,7 @@ unsigned long XMesaDitherColor( XMesaContext xmesa, GLint x, GLint y, GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha ) { - GLcontext *ctx = &xmesa->mesa; + GLcontext *ctx = xmesa->st->ctx; GLint r = (GLint) (red * 255.0F); GLint g = (GLint) (green * 255.0F); GLint b = (GLint) (blue * 255.0F); @@ -2311,7 +2293,7 @@ XMesaBindTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer, if (drawable->TextureFormat == GLX_TEXTURE_FORMAT_NONE_EXT) return; /* BadMatch error */ - rb = drawable->mesa_buffer.Attachment[b].Renderbuffer; + rb = drawable->stfb->Base.Attachment[b].Renderbuffer; if (!rb) { /* invalid buffer */ return; diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index 506db859bb..f1ca2a8fe8 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -33,6 +33,8 @@ #include "xm_image.h" #endif #include "state_tracker/st_cb_fbo.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_public.h" #include "pipe/softpipe/sp_context.h" #include "pipe/softpipe/sp_surface.h" @@ -116,7 +118,8 @@ struct xmesa_visual { * Basically corresponds to a GLXContext. */ struct xmesa_context { - GLcontext mesa; /* the core library context (containment) */ + struct st_context *st; + XMesaVisual xm_visual; /* Describes the buffers */ XMesaBuffer xm_buffer; /* current span/point/line/triangle buffer */ @@ -187,8 +190,8 @@ struct xmesa_renderbuffer * Basically corresponds to a GLXDrawable. */ struct xmesa_buffer { - GLframebuffer mesa_buffer; /* depth, stencil, accum, etc buffers */ - /* This MUST BE FIRST! */ + struct st_framebuffer *stfb; + GLboolean wasCurrent; /* was ever the current buffer? */ XMesaVisual xm_visual; /* the X/Mesa visual */ @@ -458,7 +461,7 @@ extern XMesaBuffer xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis); extern unsigned long -xmesa_color_to_pixel( GLcontext *ctx, +xmesa_color_to_pixel( XMesaContext xmesa, GLubyte r, GLubyte g, GLubyte b, GLubyte a, GLuint pixelFormat ); @@ -492,7 +495,7 @@ xmesa_renderbuffer(struct gl_renderbuffer *rb) static INLINE XMesaContext XMESA_CONTEXT(GLcontext *ctx) { - return (XMesaContext) ctx; + return (XMesaContext) ctx->DriverCtx; } @@ -502,9 +505,10 @@ XMESA_CONTEXT(GLcontext *ctx) * XXX should use inlined function for better type safety. */ static INLINE XMesaBuffer -XMESA_BUFFER(GLframebuffer *b) +XMESA_BUFFER(GLframebuffer *fb) { - return (XMesaBuffer) b; + struct st_framebuffer *stfb = (struct st_framebuffer *) fb; + return (XMesaBuffer) st_framebuffer_private(stfb); } @@ -564,4 +568,17 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, extern struct pipe_surface * xmesa_create_front_surface(XMesaVisual vis, Window win); +static INLINE GLuint +xmesa_buffer_width(XMesaBuffer b) +{ + return b->stfb->Base.Width; +} + +static INLINE GLuint +xmesa_buffer_height(XMesaBuffer b) +{ + return b->stfb->Base.Height; +} + + #endif diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index a8cc618d77..138f18aa99 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -174,6 +174,13 @@ void st_make_current(struct st_context *st, } +void st_copy_context_state(struct st_context *dst, + struct st_context *src, + uint mask) +{ + _mesa_copy_context(dst->ctx, src->ctx, mask); +} + void st_init_driver_functions(struct dd_function_table *functions) { diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 9b232827f4..30407768b9 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -168,12 +168,13 @@ static INLINE struct st_context *st_context(GLcontext *ctx) /** - * Wrapper for GLframebuffer, nothing extra for now. + * Wrapper for GLframebuffer. * This is an opaque type to the outside world. */ struct st_framebuffer { GLframebuffer Base; + void *Private; }; diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 2e7687d9c9..ec120ba39a 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -35,7 +35,9 @@ #include "st_cb_fbo.h" -struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ) +struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, + boolean createRenderbuffers, + void *private) { struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer); @@ -46,46 +48,48 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ) _mesa_initialize_framebuffer(&stfb->Base, visual); - { - /* fake frontbuffer */ - /* XXX allocation should only happen in the unusual case - it's actually needed */ - struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb); + if (createRenderbuffers) { + { + /* fake frontbuffer */ + /* XXX allocation should only happen in the unusual case + it's actually needed */ + struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb); + } + + if (visual->doubleBufferMode) { + struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); + } + + if (visual->depthBits == 24 && visual->stencilBits == 8) { + /* combined depth/stencil buffer */ + struct gl_renderbuffer *depthStencilRb + = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + /* note: bind RB to two attachment points */ + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL,depthStencilRb); + } + else if (visual->depthBits == 16) { + /* just 16-bit depth buffer, no hw stencil */ + struct gl_renderbuffer *depthRb + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); + } + + + /* now add any/all software-based renderbuffers we may need */ + _mesa_add_soft_renderbuffers(&stfb->Base, + GL_FALSE, /* never sw color */ + GL_FALSE, /* never sw depth */ + swStencil, visual->accumRedBits > 0, + GL_FALSE, /* never sw alpha */ + GL_FALSE /* never sw aux */ ); } - if (visual->doubleBufferMode) { - struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); - } - - if (visual->depthBits == 24 && visual->stencilBits == 8) { - /* combined depth/stencil buffer */ - struct gl_renderbuffer *depthStencilRb - = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); - /* note: bind RB to two attachment points */ - _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL,depthStencilRb); - } - else if (visual->depthBits == 16) { - /* just 16-bit depth buffer, no hw stencil */ - struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); - } - - - /* now add any/all software-based renderbuffers we may need */ - _mesa_add_soft_renderbuffers(&stfb->Base, - GL_FALSE, /* never sw color */ - GL_FALSE, /* never sw depth */ - swStencil, visual->accumRedBits > 0, - GL_FALSE, /* never sw alpha */ - GL_FALSE /* never sw aux */ ); - - stfb->Base.Initialized = GL_TRUE; + stfb->Private = private; } return stfb; } @@ -149,3 +153,9 @@ st_notify_swapbuffers(struct st_framebuffer *stfb) } } + +void *st_framebuffer_private( struct st_framebuffer *stfb ) +{ + return stfb->Private; +} + diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 8162ceeffa..c94e8d32c8 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -53,7 +53,12 @@ void st_destroy_context( struct st_context *st ); void st_destroy_context2( struct st_context *st ); -struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual ); +void st_copy_context_state(struct st_context *dst, struct st_context *src, + uint mask); + +struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, + GLboolean createRenderbuffers, + void *privateData); void st_resize_framebuffer( struct st_framebuffer *stfb, GLuint width, GLuint height ); @@ -61,6 +66,8 @@ void st_resize_framebuffer( struct st_framebuffer *stfb, struct pipe_surface *st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex); +void *st_framebuffer_private( struct st_framebuffer *stfb ); + void st_unreference_framebuffer( struct st_framebuffer **stfb ); void st_make_current(struct st_context *st, -- cgit v1.2.3 From 91564eedcc5e98e28d749267ac81ffd4082b4147 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 16:15:43 -0700 Subject: Remove some temporary state tracker context/framebuffer_create functions. --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 4 +- src/mesa/pipe/xlib/xm_api.c | 6 +-- src/mesa/state_tracker/st_context.c | 56 ++++++++++------------- src/mesa/state_tracker/st_public.h | 12 ++--- 4 files changed, 34 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 607d316e36..85ccb1bc2a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -152,7 +152,7 @@ intelCreateContext(const __GLcontextModes * mesaVis, } } - intel->st = st_create_context2(pipe, mesaVis, NULL); + intel->st = st_create_context(pipe, mesaVis, NULL); intel->st->ctx->DriverCtx = intel; /* hope to get rid of this... */ return GL_TRUE; @@ -181,7 +181,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) intel->first_swap_fence = NULL; } - st_destroy_context2(intel->st); + st_destroy_context(intel->st); free(intel); } } diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index abd0b4f292..ae96a866d7 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1465,8 +1465,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) pipe = xmesa_create_softpipe( c ); - c->st = st_create_context2(pipe, &v->mesa_visual, - share_list ? share_list->st : NULL); + c->st = st_create_context(pipe, &v->mesa_visual, + share_list ? share_list->st : NULL); mesaCtx = c->st->ctx; c->st->ctx->DriverCtx = c; @@ -1513,7 +1513,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) PUBLIC void XMesaDestroyContext( XMesaContext c ) { - st_destroy_context2(c->st); + st_destroy_context(c->st); _mesa_free(c); } diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 138f18aa99..88fbaeeb7a 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -68,28 +68,8 @@ void st_invalidate_state(GLcontext * ctx, GLuint new_state) } -/* - * XXX rename after above func is removed. - */ -struct st_context *st_create_context2(struct pipe_context *pipe, - const __GLcontextModes *visual, - struct st_context *share) -{ - GLcontext *ctx; - GLcontext *shareCtx = share ? share->ctx : NULL; - struct dd_function_table funcs; - - memset(&funcs, 0, sizeof(funcs)); - st_init_driver_functions(&funcs); - - ctx = _mesa_create_context(visual, shareCtx, &funcs, NULL); - - return st_create_context(ctx, pipe); -} - - -struct st_context *st_create_context( GLcontext *ctx, - struct pipe_context *pipe ) +static struct st_context * +st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) { struct st_context *st = CALLOC_STRUCT( st_context ); @@ -132,19 +112,24 @@ struct st_context *st_create_context( GLcontext *ctx, } -/* - * XXX rename after below func is removed. - */ -void st_destroy_context2( struct st_context *st ) +struct st_context *st_create_context(struct pipe_context *pipe, + const __GLcontextModes *visual, + struct st_context *share) { - GLcontext *ctx = st->ctx; - _mesa_free_context_data(ctx); - st_destroy_context(st); - free(ctx); + GLcontext *ctx; + GLcontext *shareCtx = share ? share->ctx : NULL; + struct dd_function_table funcs; + + memset(&funcs, 0, sizeof(funcs)); + st_init_driver_functions(&funcs); + + ctx = _mesa_create_context(visual, shareCtx, &funcs, NULL); + + return st_create_context_priv(ctx, pipe); } -void st_destroy_context( struct st_context *st ) +static void st_destroy_context_priv( struct st_context *st ) { draw_destroy(st->draw); st_destroy_atoms( st ); @@ -161,6 +146,15 @@ void st_destroy_context( struct st_context *st ) } +void st_destroy_context( struct st_context *st ) +{ + GLcontext *ctx = st->ctx; + _mesa_free_context_data(ctx); + st_destroy_context_priv(st); + free(ctx); +} + + void st_make_current(struct st_context *st, struct st_framebuffer *draw, struct st_framebuffer *read) diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index c94e8d32c8..b21a8d3886 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -39,20 +39,16 @@ struct st_context; struct st_framebuffer; - struct pipe_context; +struct pipe_surface; -struct st_context *st_create_context( GLcontext *ctx, - struct pipe_context *pipe); -struct st_context *st_create_context2(struct pipe_context *pipe, - const __GLcontextModes *visual, - struct st_context *share); +struct st_context *st_create_context(struct pipe_context *pipe, + const __GLcontextModes *visual, + struct st_context *share); void st_destroy_context( struct st_context *st ); -void st_destroy_context2( struct st_context *st ); - void st_copy_context_state(struct st_context *dst, struct st_context *src, uint mask); -- cgit v1.2.3 From 9b0f71e37f2ed129997b247d58f8d5e9004f4935 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 16:34:47 -0700 Subject: move st_invalidate_state() prototype to st_context.h --- src/mesa/state_tracker/st_context.h | 2 ++ src/mesa/state_tracker/st_public.h | 2 -- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 30407768b9..82e133911f 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -180,6 +180,8 @@ struct st_framebuffer extern void st_init_driver_functions(struct dd_function_table *functions); +void st_invalidate_state(GLcontext * ctx, GLuint new_state); + #define Y_0_TOP 1 diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index b21a8d3886..d81cf60503 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -70,8 +70,6 @@ void st_make_current(struct st_context *st, struct st_framebuffer *draw, struct st_framebuffer *read); -void st_invalidate_state(GLcontext * ctx, GLuint new_state); - void st_flush( struct st_context *st ); void st_notify_swapbuffers(struct st_framebuffer *stfb); -- cgit v1.2.3 From ced9c0f8c4246c055aee85f428af705f232702bc Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 16:38:07 -0700 Subject: don't include mtypes.h in st_public.h --- src/mesa/state_tracker/st_framebuffer.c | 2 +- src/mesa/state_tracker/st_public.h | 6 +++--- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index ec120ba39a..b43b9b7b9b 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -96,7 +96,7 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, void st_resize_framebuffer( struct st_framebuffer *stfb, - GLuint width, GLuint height ) + uint width, uint height ) { if (stfb->Base.Width != width || stfb->Base.Height != height) { GET_CURRENT_CONTEXT(ctx); diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index d81cf60503..9e36e1e6e5 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -28,7 +28,7 @@ #ifndef ST_PUBLIC_H #define ST_PUBLIC_H -#include "mtypes.h" +#include "pipe/p_compiler.h" #define ST_SURFACE_FRONT_LEFT 0 @@ -53,11 +53,11 @@ void st_copy_context_state(struct st_context *dst, struct st_context *src, uint mask); struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, - GLboolean createRenderbuffers, + boolean createRenderbuffers, void *privateData); void st_resize_framebuffer( struct st_framebuffer *stfb, - GLuint width, GLuint height ); + uint width, uint height ); struct pipe_surface *st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex); -- cgit v1.2.3 From cadaa4330e9472be07d07fc994dfb9c5107202d7 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 16:45:49 -0700 Subject: include mtypes.h (temporary) --- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 55606dba8b..738e6844a7 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -26,6 +26,7 @@ **************************************************************************/ +#include "main/mtypes.h" /* XXX try to remove */ #include "state_tracker/st_public.h" #include "intel_context.h" #include "i830_dri.h" -- cgit v1.2.3 From d31d93f4776fd19738a607eda337f9d3e88c5c93 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 17:10:12 -0700 Subject: change includes, minor simplification in intelDisplaySurface() --- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 8a35a71669..8e0f1113b8 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -85,14 +85,14 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, struct pipe_surface *surf, const drm_clip_rect_t *rect) { + struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); struct intel_context *intel; - const struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); DBG(SWAP, "%s\n", __FUNCTION__); assert(dPriv); - intel = intelScreenContext(dPriv->driScreenPriv->private); + intel = intelScreenContext(intelScreen); if (!intel) return; -- cgit v1.2.3 From 01e716553001d57462e75aa7d76d05df92da8e87 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 17:15:12 -0700 Subject: clean ups --- src/mesa/drivers/dri/intel_winsys/intel_lock.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c index 738e6844a7..70aa7ea5f4 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_lock.c @@ -26,7 +26,9 @@ **************************************************************************/ -#include "main/mtypes.h" /* XXX try to remove */ +#include "main/glheader.h" +#include "glapi/glthread.h" +#include #include "state_tracker/st_public.h" #include "intel_context.h" #include "i830_dri.h" @@ -68,7 +70,7 @@ intelContendedLock(struct intel_context *intel, uint flags) */ void LOCK_HARDWARE( struct intel_context *intel ) { - char __ret=0; + char __ret = 0; _glthread_LOCK_MUTEX(lockMutex); assert(!intel->locked); @@ -85,8 +87,8 @@ void LOCK_HARDWARE( struct intel_context *intel ) } - /* Unlock the hardware using the global current context - */ +/* Unlock the hardware using the global current context + */ void UNLOCK_HARDWARE( struct intel_context *intel ) { assert(intel->locked); -- cgit v1.2.3 From c6499a741c99394e81d1d86ffd066f3d9749875c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 5 Nov 2007 18:04:30 -0700 Subject: Determine GL extensions/limits by making pipe queries. The state tracker calls pipe->get_param() to determine the GL limits and which OpenGL extensions are supported. This is an initial implementation that'll probably change... --- src/mesa/main/extensions.c | 2 +- src/mesa/pipe/i915simple/i915_context.c | 20 ++++ src/mesa/pipe/p_defines.h | 17 +++ src/mesa/pipe/softpipe/sp_context.c | 20 ++++ src/mesa/sources | 1 + src/mesa/state_tracker/st_context.c | 8 +- src/mesa/state_tracker/st_extensions.c | 186 ++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_extensions.h | 38 +++++++ 8 files changed, 288 insertions(+), 4 deletions(-) create mode 100644 src/mesa/state_tracker/st_extensions.c create mode 100644 src/mesa/state_tracker/st_extensions.h (limited to 'src') diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c index 80dce56c0c..e5279e7f3e 100644 --- a/src/mesa/main/extensions.c +++ b/src/mesa/main/extensions.c @@ -86,7 +86,7 @@ static const struct { { OFF, "GL_EXT_blend_logic_op", F(EXT_blend_logic_op) }, { OFF, "GL_EXT_blend_minmax", F(EXT_blend_minmax) }, { OFF, "GL_EXT_blend_subtract", F(EXT_blend_subtract) }, - { ON, "GL_EXT_clip_volume_hint", F(EXT_clip_volume_hint) }, + { OFF, "GL_EXT_clip_volume_hint", F(EXT_clip_volume_hint) }, { OFF, "GL_EXT_cull_vertex", F(EXT_cull_vertex) }, { ON, "GL_EXT_compiled_vertex_array", F(EXT_compiled_vertex_array) }, { OFF, "GL_EXT_convolution", F(EXT_convolution) }, diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 2c36a194c7..6de1e68f73 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -142,6 +142,26 @@ static int i915_get_param(struct pipe_context *pipe, int param) { switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 0; default: return 0; } diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 8982428636..ef79716ed9 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -225,4 +225,21 @@ #define PIPE_SPRITE_COORD_UPPER_LEFT 1 #define PIPE_SPRITE_COORD_LOWER_LEFT 2 + +/** + * Implementation capabilities/limits + * Passed to pipe->get_param() + * XXX this will need some fine tuning... + */ +#define PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS 1 +#define PIPE_CAP_NPOT_TEXTURES 2 +#define PIPE_CAP_TWO_SIDED_STENCIL 3 +#define PIPE_CAP_GLSL 4 /* XXX need something better */ +#define PIPE_CAP_S3TC 5 +#define PIPE_CAP_ANISOTROPIC_FILTER 6 +#define PIPE_CAP_POINT_SPRITE 7 +#define PIPE_CAP_MAX_RENDER_TARGETS 8 +#define PIPE_CAP_OCCLUSION_QUERY 9 +#define PIPE_CAP_TEXTURE_SHADOW_MAP 10 + #endif diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 46f591e425..effecda87e 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -277,6 +277,26 @@ static const char *softpipe_get_vendor( struct pipe_context *pipe ) static int softpipe_get_param(struct pipe_context *pipe, int param) { switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 1; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 1; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 1; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; default: return 0; } diff --git a/src/mesa/sources b/src/mesa/sources index 5322cd8867..0b9b5fca3a 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -231,6 +231,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_context.c \ state_tracker/st_debug.c \ state_tracker/st_draw.c \ + state_tracker/st_extensions.c \ state_tracker/st_format.c \ state_tracker/st_framebuffer.c \ state_tracker/st_mesa_to_tgsi.c \ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 88fbaeeb7a..3810729847 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -45,6 +45,7 @@ #include "st_cb_strings.h" #include "st_atom.h" #include "st_draw.h" +#include "st_extensions.h" #include "st_program.h" #include "pipe/p_context.h" #include "pipe/draw/draw_context.h" @@ -105,8 +106,9 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st->pixel_xfer.cache = _mesa_new_program_cache(); - /* XXXX This is temporary! */ - _mesa_enable_sw_extensions(ctx); + /* GL limits and extensions */ + st_init_limits(st); + st_init_extensions(st); return st; } diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c new file mode 100644 index 0000000000..3f56b537bb --- /dev/null +++ b/src/mesa/state_tracker/st_extensions.c @@ -0,0 +1,186 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "main/imports.h" +#include "main/context.h" +#include "main/extensions.h" +#include "main/macros.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + +#include "st_context.h" +#include "st_extensions.h" + + +/* + * Compute floor(log_base_2(n)). + * If n < 0 return -1. + */ +static int +logbase2( int n ) +{ + GLint i = 1; + GLint log2 = 0; + + if (n < 0) + return -1; + + if (n == 0) + return 0; + + while ( n > i ) { + i *= 2; + log2++; + } + if (i != n) { + return log2 - 1; + } + else { + return log2; + } +} + + +void st_init_limits(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + GLcontext *ctx = st->ctx; + uint w, h, d; + + ctx->Const.MaxTextureImageUnits + = ctx->Const.MaxTextureCoordUnits + = pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS); + + pipe->max_texture_size(pipe, PIPE_TEXTURE_2D, &w, &h, &d); + ctx->Const.MaxTextureLevels = logbase2(w) + 1; + ctx->Const.MaxTextureRectSize = w; + + pipe->max_texture_size(pipe, PIPE_TEXTURE_3D, &w, &h, &d); + ctx->Const.Max3DTextureLevels = logbase2(d) + 1; + + pipe->max_texture_size(pipe, PIPE_TEXTURE_CUBE, &w, &h, &d); + ctx->Const.MaxCubeTextureLevels = logbase2(w) + 1; + + ctx->Const.MaxDrawBuffers = MAX2(1, pipe->get_param(pipe, PIPE_CAP_MAX_RENDER_TARGETS)); + +} + + +/** + * XXX this needs careful review + */ +void st_init_extensions(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + GLcontext *ctx = st->ctx; + + /* + * Extensions that are supported by all Gallium drivers: + */ + ctx->Extensions.ARB_fragment_program = GL_TRUE; + ctx->Extensions.ARB_texture_cube_map = GL_TRUE; + ctx->Extensions.ARB_texture_env_combine = GL_TRUE; + ctx->Extensions.ARB_texture_env_crossbar = GL_TRUE; + ctx->Extensions.ARB_texture_env_dot3 = GL_TRUE; + ctx->Extensions.ARB_vertex_program = GL_TRUE; + ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE; + + ctx->Extensions.EXT_blend_color = GL_TRUE; + ctx->Extensions.EXT_blend_equation_separate = GL_TRUE; + ctx->Extensions.EXT_blend_func_separate = GL_TRUE; + ctx->Extensions.EXT_blend_logic_op = GL_TRUE; + ctx->Extensions.EXT_blend_minmax = GL_TRUE; + ctx->Extensions.EXT_blend_subtract = GL_TRUE; + ctx->Extensions.EXT_framebuffer_object = GL_TRUE; + ctx->Extensions.EXT_fog_coord = GL_TRUE; + ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE; + ctx->Extensions.EXT_pixel_buffer_object = GL_TRUE; + ctx->Extensions.EXT_point_parameters = GL_TRUE; + ctx->Extensions.EXT_secondary_color = GL_TRUE; + ctx->Extensions.EXT_stencil_wrap = GL_TRUE; + ctx->Extensions.EXT_texture_env_add = GL_TRUE; + ctx->Extensions.EXT_texture_env_combine = GL_TRUE; + ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE; + ctx->Extensions.EXT_texture_lod_bias = GL_TRUE; + + ctx->Extensions.NV_blend_square = GL_TRUE; + ctx->Extensions.NV_texgen_reflection = GL_TRUE; + + + /* + * Extensions that depend on the driver/hardware: + */ + if (pipe->get_param(pipe, PIPE_CAP_MAX_RENDER_TARGETS) > 1) { + ctx->Extensions.ARB_draw_buffers = GL_TRUE; + } + + if (pipe->get_param(pipe, PIPE_CAP_GLSL) > 1) { + ctx->Extensions.ARB_fragment_shader = GL_TRUE; + ctx->Extensions.ARB_vertex_shader = GL_TRUE; + ctx->Extensions.ARB_shader_objects = GL_TRUE; + ctx->Extensions.ARB_shading_language_100 = GL_TRUE; + ctx->Extensions.ARB_shading_language_120 = GL_TRUE; + } + + if (pipe->get_param(pipe, PIPE_CAP_NPOT_TEXTURES)) { + ctx->Extensions.ARB_texture_non_power_of_two = GL_TRUE; + ctx->Extensions.NV_texture_rectangle = GL_TRUE; + } + + if (pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS) > 1) { + ctx->Extensions.ARB_multitexture = GL_TRUE; + } + + if (pipe->get_param(pipe, PIPE_CAP_TWO_SIDED_STENCIL) > 1) { + ctx->Extensions.ATI_separate_stencil = GL_TRUE; + } + + if (pipe->get_param(pipe, PIPE_CAP_S3TC) > 1) { + ctx->Extensions.ARB_texture_compression = GL_TRUE; + ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE; + } + + if (pipe->get_param(pipe, PIPE_CAP_ANISOTROPIC_FILTER)) { + ctx->Extensions.EXT_texture_filter_anisotropic = GL_TRUE; + } + + if (pipe->get_param(pipe, PIPE_CAP_POINT_SPRITE)) { + ctx->Extensions.ARB_point_sprite = GL_TRUE; + } + + if (pipe->get_param(pipe, PIPE_CAP_OCCLUSION_QUERY) > 1) { + ctx->Extensions.ARB_occlusion_query = GL_TRUE; + } + + if (pipe->get_param(pipe, PIPE_CAP_TEXTURE_SHADOW_MAP) > 1) { + ctx->Extensions.ARB_depth_texture = GL_TRUE; + ctx->Extensions.ARB_shadow = GL_TRUE; + /*ctx->Extensions.ARB_shadow_ambient = GL_TRUE;*/ + } + +} diff --git a/src/mesa/state_tracker/st_extensions.h b/src/mesa/state_tracker/st_extensions.h new file mode 100644 index 0000000000..2994f16dd3 --- /dev/null +++ b/src/mesa/state_tracker/st_extensions.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_EXTENSIONS_H +#define ST_EXTENSIONS_H + + +extern void st_init_limits(struct st_context *st); + +extern void st_init_extensions(struct st_context *st); + + +#endif /* ST_EXTENSIONS_H */ -- cgit v1.2.3 From 0ab2c84ce95cea5c72aa4e9680862bf82d31259d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 6 Nov 2007 10:16:34 +0100 Subject: Fix non-x86 build. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index bc20d9ca05..a5c6e1a33a 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -29,6 +29,8 @@ #include "tgsi_core.h" #include "x86/rtasm/x86sse.h" +#ifdef USE_X86_ASM + #define DUMP_SSE 0 #if DUMP_SSE @@ -2325,3 +2327,5 @@ tgsi_emit_sse2_fs( return 1; } + +#endif /* USE_X86_ASM */ -- cgit v1.2.3 From 4f79dbd5aa6b8a955ac9dcc43cf0852acd89a320 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 6 Nov 2007 10:17:59 +0100 Subject: Convert format bitfields to shifts and masks. The memory layout of bitfields depends on the ABI. --- src/mesa/pipe/p_format.h | 63 ++++++++++++++++---------------------- src/mesa/state_tracker/st_format.c | 51 +++++++++++++++--------------- 2 files changed, 52 insertions(+), 62 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index 89d83a44f5..d478a6c58d 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -42,11 +42,10 @@ #define PIPE_FORMAT_LAYOUT_RGBAZS 0 #define PIPE_FORMAT_LAYOUT_YCBCR 1 -struct pipe_format_header +static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ { - uint layout : 2; /**< PIPE_FORMAT_LAYOUT_ */ - uint padding : 30; -}; + return f & 0x3; +} /** * RGBAZS Format Layout. @@ -75,16 +74,14 @@ struct pipe_format_header #define PIPE_FORMAT_TYPE_SSCALED 5 /** - * In a nutshell, this bitfield contains instructions how to unpack - * a given format to a full-blown 4-component vector suitable for FS. * Because the destination vector is assumed to be RGBA FLOAT, we * need to know how to swizzle and expand components from the source * vector. - * Let's take U_A1_R5_G5_B5 as an example. SwizzleX is A, sizeX + * Let's take U_A1_R5_G5_B5 as an example. X swizzle is A, X size * is 1 bit and type is UNORM. So we take the most significant bit * from source vector, convert 0 to 0.0 and 1 to 1.0 and save it * in the last component of the destination RGBA component. - * Next, swizzleY is R, sizeY is 5 and type is UNORM. We normalize + * Next, Y swizzle is R, Y size is 5 and type is UNORM. We normalize * those 5 bits into [0.0; 1.0] range and put it into second * component of the destination vector. Rinse and repeat for * components Z and W. @@ -93,21 +90,23 @@ struct pipe_format_header * If any swizzle is 0 or 1, the corresponding destination component * should be filled with 0.0 and 1.0, respectively. */ -struct pipe_format_rgbazs +typedef uint pipe_format_rgbazs_t; + +static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) { - uint layout : 2; /**< PIPE_FORMAT_LAYOUT_RGBAZS */ - uint swizzleX : 3; /**< PIPE_FORMAT_COMP_ */ - uint swizzleY : 3; /**< PIPE_FORMAT_COMP_ */ - uint swizzleZ : 3; /**< PIPE_FORMAT_COMP_ */ - uint swizzleW : 3; /**< PIPE_FORMAT_COMP_ */ - uint sizeX : 3; /**< Size of X - 1 */ - uint sizeY : 3; /**< Size of Y - 1 */ - uint sizeZ : 3; /**< Size of Z - 1 */ - uint sizeW : 3; /**< Size of W - 1 */ - uint exp8 : 2; /**< Scale size by 8 ^ exp8 */ - uint type : 3; /**< PIPE_FORMAT_TYPE_ */ - uint padding : 1; -}; + return (f >> shift) & mask; +} + +#define pf_swizzle_x(f) pf_get(f, 2, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_y(f) pf_get(f, 5, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_z(f) pf_get(f, 8, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_w(f) pf_get(f, 11, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_size_x(f) pf_get(f, 14, 0x7) /**< Size of X - 1 */ +#define pf_size_y(f) pf_get(f, 17, 0x7) /**< Size of Y - 1 */ +#define pf_size_z(f) pf_get(f, 20, 0x7) /**< Size of Z - 1 */ +#define pf_size_w(f) pf_get(f, 23, 0x7) /**< Size of W - 1 */ +#define pf_exp8(f) pf_get(f, 26, 0x3) /**< Scale size by 8 ^ exp8 */ +#define pf_type(f) pf_get(f, 28, 0xf) /**< PIPE_FORMAT_TYPE_ */ /** * Helper macro to encode the above structure into a 32-bit value. @@ -179,15 +178,10 @@ struct pipe_format_rgbazs */ /** - * This bitfields is simple. It only contains a flag that indicates whether the - * format is reversed or not. + * This only contains a flag that indicates whether the format is reversed or + * not. */ -struct pipe_format_ycbcr -{ - uint layout : 2; /**< PIPE_FORMAT_LAYOUT_YCBCR */ - uint reversed : 1; - uint padding : 29; -}; +typedef uint pipe_format_ycbcr_t; /** * Helper macro to encode the above structure into a 32-bit value. @@ -196,13 +190,10 @@ struct pipe_format_ycbcr (PIPE_FORMAT_LAYOUT_YCBCR << 0) |\ ((REV) << 2) ) -union pipe_format +static INLINE uint pf_rev(pipe_format_ycbcr_t f) { - uint value32; - struct pipe_format_header header; - struct pipe_format_rgbazs rgbazs; - struct pipe_format_ycbcr ycbcr; -}; + return (f >> 2) & 0x1; +} /** * Texture/surface image formats (preliminary) diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index d5e8581b21..e9839e2936 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -45,32 +45,32 @@ static GLuint format_bits( - struct pipe_format_rgbazs info, + pipe_format_rgbazs_t info, GLuint comp ) { GLuint size; - if (info.swizzleX == comp) { - size = info.sizeX; + if (pf_swizzle_x(info) == comp) { + size = pf_size_x(info); } - else if (info.swizzleY == comp) { - size = info.sizeY; + else if (pf_swizzle_y(info) == comp) { + size = pf_size_y(info); } - else if (info.swizzleZ == comp) { - size = info.sizeZ; + else if (pf_swizzle_z(info) == comp) { + size = pf_size_z(info); } - else if (info.swizzleW == comp) { - size = info.sizeW; + else if (pf_swizzle_w(info) == comp) { + size = pf_size_w(info); } else { size = 0; } - return size << (info.exp8 * 3); + return size << (pf_exp8(info) * 3); } static GLuint format_max_bits( - struct pipe_format_rgbazs info ) + pipe_format_rgbazs_t info ) { GLuint size = format_bits( info, PIPE_FORMAT_COMP_R ); @@ -84,7 +84,7 @@ format_max_bits( static GLuint format_size( - struct pipe_format_rgbazs info ) + pipe_format_rgbazs_t info ) { return format_bits( info, PIPE_FORMAT_COMP_R ) + @@ -103,13 +103,10 @@ st_get_format_info( GLuint format, struct pipe_format_info *pinfo ) { - union pipe_format fmt; + if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { + pipe_format_rgbazs_t info; - fmt.value32 = format; - if (fmt.header.layout == PIPE_FORMAT_LAYOUT_RGBAZS) { - struct pipe_format_rgbazs info; - - info = fmt.rgbazs; + info = format; #if 0 printf( @@ -129,20 +126,20 @@ st_get_format_info( size = format_max_bits( info ); if (size == 8) { - if (info.type == PIPE_FORMAT_TYPE_UNORM) + if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM) pinfo->datatype = GL_UNSIGNED_BYTE; else pinfo->datatype = GL_BYTE; } else if (size == 16) { - if (info.type == PIPE_FORMAT_TYPE_UNORM) + if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM) pinfo->datatype = GL_UNSIGNED_SHORT; else pinfo->datatype = GL_SHORT; } else { assert( size <= 32 ); - if (info.type == PIPE_FORMAT_TYPE_UNORM) + if (pf_type(info) == PIPE_FORMAT_TYPE_UNORM) pinfo->datatype = GL_UNSIGNED_INT; else pinfo->datatype = GL_INT; @@ -161,8 +158,10 @@ st_get_format_info( pinfo->size = format_size( info ) / 8; /* Luminance & Intensity bits */ - if( info.swizzleX == PIPE_FORMAT_COMP_R && info.swizzleY == PIPE_FORMAT_COMP_R && info.swizzleZ == PIPE_FORMAT_COMP_R ) { - if( info.swizzleW == PIPE_FORMAT_COMP_R ) { + if( pf_swizzle_x(info) == PIPE_FORMAT_COMP_R && + pf_swizzle_y(info) == PIPE_FORMAT_COMP_R && + pf_swizzle_z(info) == PIPE_FORMAT_COMP_R ) { + if( pf_swizzle_w(info) == PIPE_FORMAT_COMP_R ) { pinfo->luminance_bits = 0; pinfo->intensity_bits = pinfo->red_bits; } @@ -190,11 +189,11 @@ st_get_format_info( } } else { - struct pipe_format_ycbcr info; + pipe_format_ycbcr_t info; - assert( fmt.header.layout == PIPE_FORMAT_LAYOUT_YCBCR ); + assert( pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR ); - info = fmt.ycbcr; + info = format; /* TODO */ assert( 0 ); -- cgit v1.2.3 From fa1a66d7fc4fd7854de7958a48e4992edd154489 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 5 Nov 2007 18:04:35 +0000 Subject: Supply buffer usage hints to winsys. Winsys driver needs some hints in order to allocate the appropriate kind of memory for the buffer. --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 6 ++++-- src/mesa/drivers/x11/xm_winsys.c | 5 +++-- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 4 +++- src/mesa/pipe/p_defines.h | 14 +++++++++----- src/mesa/pipe/p_winsys.h | 11 +++++++++-- src/mesa/pipe/xlib/xm_winsys.c | 5 +++-- src/mesa/state_tracker/st_atom_constbuf.c | 3 ++- src/mesa/state_tracker/st_cb_bufferobjects.c | 20 +++++++++++++++++++- src/mesa/state_tracker/st_cb_rasterpos.c | 7 +++++-- src/mesa/state_tracker/st_draw.c | 4 +++- 10 files changed, 60 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index af05a3398c..cc76a40a5a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -107,7 +107,8 @@ intel_buffer_reference(struct pipe_winsys *sws, */ static void intel_buffer_data(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, - unsigned size, const void *data ) + unsigned size, const void *data, + unsigned usage ) { struct intel_context *intel = intel_pipe_winsys(sws)->intel; @@ -229,7 +230,8 @@ intel_i915_region_alloc(struct pipe_winsys *winsys, winsys->buffer_data( winsys, region->buffer, pitch * cpp * height, - NULL ); + NULL, + PIPE_BUFFER_USAGE_PIXEL ); return region; } diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index 36805437f0..f863cdbc15 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -140,7 +140,7 @@ xm_buffer_reference(struct pipe_winsys *pws, static void xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, - unsigned size, const void *data ) + unsigned size, const void *data, unsigned usage) { struct xm_buffer *xm_buf = xm_bo(buf); assert(!xm_buf->userBuffer); @@ -265,7 +265,8 @@ xm_region_alloc(struct pipe_winsys *winsys, winsys->buffer_data( winsys, region->buffer, region->pitch * cpp * height, - NULL ); + NULL, + PIPE_BUFFER_USAGE_PIXEL ); return region; } diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 3632adce79..75ca6d6e5e 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -376,7 +376,9 @@ static void vbuf_flush_elements( struct draw_stage *stage ) /* FIXME: handle failure */ if(!vbuf->buf) vbuf->buf = winsys->buffer_create(winsys, 64); - winsys->buffer_data(winsys, vbuf->buf, VBUF_SIZE, NULL); + winsys->buffer_data( winsys, vbuf->buf, + VBUF_SIZE, NULL, + PIPE_BUFFER_USAGE_VERTEX ); vbuf->vertex_map = winsys->buffer_map(winsys, vbuf->buf, PIPE_BUFFER_FLAG_WRITE ); diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index ef79716ed9..119ea1bd77 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -168,15 +168,19 @@ /** - * Buffer flags + * Buffer access flags */ #define PIPE_BUFFER_FLAG_READ 0x1 #define PIPE_BUFFER_FLAG_WRITE 0x2 -#define PIPE_BUFFER_USE_TEXTURE 0x1 -#define PIPE_BUFFER_USE_VERTEX_BUFFER 0x2 -#define PIPE_BUFFER_USE_INDEX_BUFFER 0x4 -#define PIPE_BUFFER_USE_RENDER_TARGET 0x8 + +/** + * Buffer usage flags + */ +#define PIPE_BUFFER_USAGE_PIXEL 0x1 +#define PIPE_BUFFER_USAGE_VERTEX 0x2 +#define PIPE_BUFFER_USAGE_INDEX 0x4 +#define PIPE_BUFFER_USAGE_CONSTANT 0x8 /** diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 2d4432dbca..298b555651 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -116,10 +116,17 @@ struct pipe_winsys struct pipe_buffer_handle **ptr, struct pipe_buffer_handle *buf ); - /** Create the data store of a buffer and optionally initialize it */ + /** + * Create the data store of a buffer and optionally initialize it. + * + * usage is a bitmask of PIPE_BUFFER_USAGE_PIXEL/VERTEX/INDEX/CONSTANT. This + * usage argument is only an optimization hint, not a guarantee, therefore + * proper behavior must be observed in all circumstances. + */ void (*buffer_data)(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, - unsigned size, const void *data ); + unsigned size, const void *data, + unsigned usage); /** Modify some or all of the data contained in a buffer's data store */ void (*buffer_subdata)(struct pipe_winsys *sws, diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 5de811a66f..45ece8ef55 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -120,7 +120,7 @@ xm_buffer_reference(struct pipe_winsys *pws, static void xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, - unsigned size, const void *data ) + unsigned size, const void *data, unsigned usage ) { struct xm_buffer *xm_buf = xm_bo(buf); assert(!xm_buf->userBuffer); @@ -245,7 +245,8 @@ xm_region_alloc(struct pipe_winsys *winsys, winsys->buffer_data( winsys, region->buffer, region->pitch * cpp * height, - NULL ); + NULL, + PIPE_BUFFER_USAGE_PIXEL ); return region; } diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index 7da7136676..446250c226 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -80,7 +80,8 @@ void st_upload_constants( struct st_context *st, } /* load Mesa constants into the constant buffer */ - ws->buffer_data(ws, cbuf->buffer, paramBytes, params->ParameterValues); + ws->buffer_data(ws, cbuf->buffer, paramBytes, params->ParameterValues, + PIPE_BUFFER_USAGE_CONSTANT); cbuf->size = paramBytes; diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index a593bd74d1..99e1eb3c7a 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -104,11 +104,29 @@ st_bufferobj_data(GLcontext *ctx, { struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); + unsigned buffer_usage; st_obj->Base.Size = size; st_obj->Base.Usage = usage; + + switch(target) { + case GL_PIXEL_PACK_BUFFER_ARB: + case GL_PIXEL_UNPACK_BUFFER_ARB: + buffer_usage = PIPE_BUFFER_USAGE_PIXEL; + break; + case GL_ARRAY_BUFFER_ARB: + buffer_usage = PIPE_BUFFER_USAGE_VERTEX; + break; + case GL_ELEMENT_ARRAY_BUFFER_ARB: + buffer_usage = PIPE_BUFFER_USAGE_INDEX; + break; + default: + buffer_usage = 0; + } - pipe->winsys->buffer_data( pipe->winsys, st_obj->buffer, size, data ); + pipe->winsys->buffer_data( pipe->winsys, st_obj->buffer, + size, data, + buffer_usage ); } diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 40807fc05a..5b97c1ee34 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -282,7 +282,9 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) const unsigned size = sizeof(ctx->Current.Attrib); const void *data = ctx->Current.Attrib; /* colors, texcoords, etc */ - pipe->winsys->buffer_data(pipe->winsys, buf, size, data); + pipe->winsys->buffer_data(pipe->winsys, buf, + size, data, + PIPE_BUFFER_USAGE_VERTEX); /* position */ pipe->winsys->buffer_subdata(pipe->winsys, buf, 0, /* offset */ @@ -301,7 +303,8 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) fb_buf.start_offset = 0; pipe->winsys->buffer_data(pipe->winsys, fb_buf.buffer, fb_buf.size, - NULL); /* data */ + NULL, /* data */ + PIPE_BUFFER_USAGE_VERTEX); if (pipe->set_feedback_buffer) pipe->set_feedback_buffer(pipe, 0, &fb_buf); } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index c3f33a447e..3e2ed6cada 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -289,7 +289,9 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, /* XXX create one-time */ vbuf = pipe->winsys->buffer_create(pipe->winsys, 32); - pipe->winsys->buffer_data(pipe->winsys, vbuf, vertex_bytes, verts); + pipe->winsys->buffer_data(pipe->winsys, vbuf, + vertex_bytes, verts, + PIPE_BUFFER_USAGE_VERTEX); /* tell pipe about the vertex buffer */ vbuffer.buffer = vbuf; -- cgit v1.2.3 From a06dcfd75b1591bc1db568798f82be721dcc807e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 6 Nov 2007 14:43:20 +0000 Subject: Allow custom pipe driver buffer usage flags. --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 2 +- src/mesa/pipe/i915simple/i915_winsys.h | 5 +++++ src/mesa/pipe/p_defines.h | 11 ++++++----- 3 files changed, 12 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 75ca6d6e5e..086ad04d1e 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -378,7 +378,7 @@ static void vbuf_flush_elements( struct draw_stage *stage ) vbuf->buf = winsys->buffer_create(winsys, 64); winsys->buffer_data( winsys, vbuf->buf, VBUF_SIZE, NULL, - PIPE_BUFFER_USAGE_VERTEX ); + I915_BUFFER_USAGE_LIT_VERTEX ); vbuf->vertex_map = winsys->buffer_map(winsys, vbuf->buf, PIPE_BUFFER_FLAG_WRITE ); diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 947c5a334d..cbb851a932 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -37,6 +37,9 @@ #define I915_WINSYS_H +#include "pipe/p_defines.h" + + /* Pipe drivers are (meant to be!) independent of both GL and the * window system. The window system provides a buffer manager and a * set of additional hooks for things like command buffer submission, @@ -100,6 +103,8 @@ struct i915_winsys { #define I915_BUFFER_ACCESS_WRITE 0x1 #define I915_BUFFER_ACCESS_READ 0x2 +#define I915_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) + struct pipe_context *i915_create( struct pipe_winsys *, struct i915_winsys *, diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 119ea1bd77..cc0232cdcd 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -177,11 +177,12 @@ /** * Buffer usage flags */ -#define PIPE_BUFFER_USAGE_PIXEL 0x1 -#define PIPE_BUFFER_USAGE_VERTEX 0x2 -#define PIPE_BUFFER_USAGE_INDEX 0x4 -#define PIPE_BUFFER_USAGE_CONSTANT 0x8 - +#define PIPE_BUFFER_USAGE_PIXEL (1 << 0) +#define PIPE_BUFFER_USAGE_VERTEX (1 << 1) +#define PIPE_BUFFER_USAGE_INDEX (1 << 2) +#define PIPE_BUFFER_USAGE_CONSTANT (1 << 3) +/** Pipe driver custam usage flags should be greater or equal to this value */ +#define PIPE_BUFFER_USAGE_CUSTOM (1 << 16) /** * Flush types: -- cgit v1.2.3 From 24ac9c30ebfd2edabdd21bfc9cf4e9db21cd10df Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 08:53:25 -0700 Subject: check for __i386__ or __386__ instead of USE_X86_ASM --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index a5c6e1a33a..3a5c1da180 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -29,7 +29,7 @@ #include "tgsi_core.h" #include "x86/rtasm/x86sse.h" -#ifdef USE_X86_ASM +#if defined(__i386__) || defined(__386__) #define DUMP_SSE 0 @@ -2328,4 +2328,4 @@ tgsi_emit_sse2_fs( return 1; } -#endif /* USE_X86_ASM */ +#endif /* i386 */ -- cgit v1.2.3 From e2feb80a90f3e9300c70a4a4eb3e966131f5c313 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 09:41:45 -0700 Subject: Remove pipe->max_texture_size(), use get_param() instead. Also, in st_init_limits(), clamp driver's values against Mesa's internal limits. --- src/mesa/pipe/failover/fo_context.c | 2 +- src/mesa/pipe/i915simple/i915_context.c | 37 +++------------- src/mesa/pipe/p_context.h | 6 --- src/mesa/pipe/p_defines.h | 5 ++- src/mesa/pipe/softpipe/sp_context.c | 41 +++-------------- src/mesa/state_tracker/st_cb_drawpixels.c | 8 ++-- src/mesa/state_tracker/st_extensions.c | 73 +++++++++++++++---------------- 7 files changed, 57 insertions(+), 115 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index aa5d0885e6..e18b349ef3 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -117,9 +117,9 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.winsys = hw->winsys; failover->pipe.destroy = failover_destroy; failover->pipe.is_format_supported = hw->is_format_supported; - failover->pipe.max_texture_size = hw->max_texture_size; failover->pipe.get_name = hw->get_name; failover->pipe.get_vendor = hw->get_vendor; + failover->pipe.get_param = hw->get_param; failover->pipe.draw_arrays = failover_draw_arrays; failover->pipe.draw_elements = failover_draw_elements; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 6de1e68f73..50503a843a 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -108,36 +108,6 @@ i915_is_format_supported( struct pipe_context *pipe, } -/** - * We might want to return max texture levels instead... - */ -static void -i915_max_texture_size(struct pipe_context *pipe, unsigned textureType, - unsigned *maxWidth, unsigned *maxHeight, unsigned *maxDepth) -{ - switch (textureType) { - case PIPE_TEXTURE_1D: - *maxWidth = 2048; - break; - case PIPE_TEXTURE_2D: - *maxWidth = - *maxHeight = 2048; - break; - case PIPE_TEXTURE_3D: - *maxWidth = - *maxHeight = - *maxDepth = 256; - break; - case PIPE_TEXTURE_CUBE: - *maxWidth = - *maxHeight = 2048; - break; - default: - assert(0); - } -} - - static int i915_get_param(struct pipe_context *pipe, int param) { @@ -162,6 +132,12 @@ i915_get_param(struct pipe_context *pipe, int param) return 0; case PIPE_CAP_TEXTURE_SHADOW_MAP: return 0; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 11; /* max 1024x1024 */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 11; /* max 1024x1024 */ default: return 0; } @@ -324,7 +300,6 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.destroy = i915_destroy; i915->pipe.is_format_supported = i915_is_format_supported; - i915->pipe.max_texture_size = i915_max_texture_size; i915->pipe.get_param = i915_get_param; i915->pipe.clear = i915_clear; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 4f5e786334..ddc7acc747 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -51,12 +51,6 @@ struct pipe_context { boolean (*is_format_supported)( struct pipe_context *pipe, uint format ); - void (*max_texture_size)(struct pipe_context *pipe, - unsigned textureType, /* PIPE_TEXTURE_x */ - unsigned *maxWidth, - unsigned *maxHeight, - unsigned *maxDepth); - const char *(*get_name)( struct pipe_context *pipe ); const char *(*get_vendor)( struct pipe_context *pipe ); diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index cc0232cdcd..ca9929bfee 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -245,6 +245,9 @@ #define PIPE_CAP_POINT_SPRITE 7 #define PIPE_CAP_MAX_RENDER_TARGETS 8 #define PIPE_CAP_OCCLUSION_QUERY 9 -#define PIPE_CAP_TEXTURE_SHADOW_MAP 10 +#define PIPE_CAP_TEXTURE_SHADOW_MAP 10 +#define PIPE_CAP_MAX_TEXTURE_2D_LEVELS 11 +#define PIPE_CAP_MAX_TEXTURE_3D_LEVELS 12 +#define PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS 13 #endif diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index effecda87e..954884e7e9 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -83,39 +83,6 @@ softpipe_is_format_supported( struct pipe_context *pipe, } -/** XXX remove these? */ -#define MAX_TEXTURE_LEVELS 11 -#define MAX_TEXTURE_RECT_SIZE 2048 -#define MAX_3D_TEXTURE_LEVELS 8 - -static void -softpipe_max_texture_size(struct pipe_context *pipe, unsigned textureType, - unsigned *maxWidth, unsigned *maxHeight, - unsigned *maxDepth) -{ - switch (textureType) { - case PIPE_TEXTURE_1D: - *maxWidth = 1 << (MAX_TEXTURE_LEVELS - 1); - break; - case PIPE_TEXTURE_2D: - *maxWidth = - *maxHeight = 1 << (MAX_TEXTURE_LEVELS - 1); - break; - case PIPE_TEXTURE_3D: - *maxWidth = - *maxHeight = - *maxDepth = 1 << (MAX_3D_TEXTURE_LEVELS - 1); - break; - case PIPE_TEXTURE_CUBE: - *maxWidth = - *maxHeight = MAX_TEXTURE_RECT_SIZE; - break; - default: - assert(0); - } -} - - /** * Map any drawing surfaces which aren't already mapped */ @@ -297,6 +264,12 @@ static int softpipe_get_param(struct pipe_context *pipe, int param) return 1; case PIPE_CAP_TEXTURE_SHADOW_MAP: return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 12; /* max 2Kx2K */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 12; /* max 2Kx2K */ default: return 0; } @@ -321,7 +294,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, /* queries */ softpipe->pipe.is_format_supported = softpipe_is_format_supported; - softpipe->pipe.max_texture_size = softpipe_max_texture_size; + //softpipe->pipe.max_texture_size = softpipe_max_texture_size; softpipe->pipe.get_param = softpipe_get_param; /* state setters */ diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index fc58035d00..df41d5ce5b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -703,15 +703,15 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, const GLuint unit = 0; struct pipe_context *pipe = ctx->st->pipe; GLfloat x0, y0, x1, y1; - GLuint maxWidth, maxHeight; + GLuint maxSize; /* limit checks */ /* XXX if DrawPixels image is larger than max texture size, break * it up into chunks. */ - pipe->max_texture_size(pipe, PIPE_TEXTURE_2D, &maxWidth, &maxHeight, NULL); - assert(width <= maxWidth); - assert(height <= maxHeight); + maxSize = 1 << (pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1); + assert(width <= maxSize); + assert(height <= maxSize); /* setup state: just scissor */ { diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 3f56b537bb..0707c489a6 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -37,57 +37,54 @@ #include "st_extensions.h" -/* - * Compute floor(log_base_2(n)). - * If n < 0 return -1. - */ -static int -logbase2( int n ) +static int min(int a, int b) { - GLint i = 1; - GLint log2 = 0; - - if (n < 0) - return -1; - - if (n == 0) - return 0; + return (a < b) ? a : b; +} - while ( n > i ) { - i *= 2; - log2++; - } - if (i != n) { - return log2 - 1; - } - else { - return log2; - } +static int clamp(int a, int min, int max) +{ + if (a < min) + return min; + else if (a > max) + return max; + else + return a; } +/** + * Query driver to get implementation limits. + * Note that we have to limit/clamp against Mesa's internal limits too. + */ void st_init_limits(struct st_context *st) { struct pipe_context *pipe = st->pipe; - GLcontext *ctx = st->ctx; - uint w, h, d; + struct gl_constants *c = &st->ctx->Const; + + c->MaxTextureLevels + = min(pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_2D_LEVELS), + MAX_TEXTURE_LEVELS); - ctx->Const.MaxTextureImageUnits - = ctx->Const.MaxTextureCoordUnits - = pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS); + c->Max3DTextureLevels + = min(pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_3D_LEVELS), + MAX_3D_TEXTURE_LEVELS); - pipe->max_texture_size(pipe, PIPE_TEXTURE_2D, &w, &h, &d); - ctx->Const.MaxTextureLevels = logbase2(w) + 1; - ctx->Const.MaxTextureRectSize = w; + c->MaxCubeTextureLevels + = min(pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS), + MAX_CUBE_TEXTURE_LEVELS); - pipe->max_texture_size(pipe, PIPE_TEXTURE_3D, &w, &h, &d); - ctx->Const.Max3DTextureLevels = logbase2(d) + 1; + c->MaxTextureRectSize + = min(1 << (c->MaxTextureLevels - 1), MAX_TEXTURE_RECT_SIZE); - pipe->max_texture_size(pipe, PIPE_TEXTURE_CUBE, &w, &h, &d); - ctx->Const.MaxCubeTextureLevels = logbase2(w) + 1; - - ctx->Const.MaxDrawBuffers = MAX2(1, pipe->get_param(pipe, PIPE_CAP_MAX_RENDER_TARGETS)); + c->MaxTextureImageUnits + = c->MaxTextureCoordUnits + = min(pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS), + MAX_TEXTURE_IMAGE_UNITS); + c->MaxDrawBuffers + = clamp(pipe->get_param(pipe, PIPE_CAP_MAX_RENDER_TARGETS), + 1, MAX_DRAW_BUFFERS); } -- cgit v1.2.3 From 0800342296fb9eeb6bac8f24965441dff0e71812 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 10:05:01 -0700 Subject: Remove last of core Mesa dependencies in intel_swapbuffers.c Use the "dummyContext" pointer (for now) instead of GET_CURRENT_CONTEXT(). --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 10 +++-- src/mesa/drivers/dri/intel_winsys/intel_screen.h | 6 ++- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 45 +++------------------- 3 files changed, 17 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 85ccb1bc2a..ed6448646a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -181,6 +181,9 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) intel->first_swap_fence = NULL; } + if (intel->intelScreen->dummyContext == intel) + intel->intelScreen->dummyContext = NULL; + st_destroy_context(intel->st); free(intel); } @@ -210,9 +213,10 @@ intelMakeCurrent(__DRIcontextPrivate * driContextPriv, assert(draw_fb->stfb); assert(read_fb->stfb); - /* this is a hack so we have a valid context when the region allocation - is done. Need a per-screen context? */ - intel->intelScreen->dummyctxptr = intel; + /* This is for situations in which we need a rendering context but + * there may not be any currently bound. + */ + intel->intelScreen->dummyContext = intel; st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index 1d24ff7ff7..3bcbc22963 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -69,7 +69,11 @@ struct intel_screen struct _DriBufferPool *staticPool; /** for the X screen/framebuffer */ boolean havePools; - struct intel_context *dummyctxptr; + /** + * Temporary(?) context to use for SwapBuffers or other situations in + * which we need a rendering context, but none is currently bound. + */ + struct intel_context *dummyContext; }; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 8e0f1113b8..9fcb2dac27 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -31,7 +31,6 @@ #include "intel_batchbuffer.h" #include "intel_reg.h" #include "intel_winsys.h" -#include "context.h" #include "pipe/p_context.h" #include "state_tracker/st_public.h" @@ -39,39 +38,6 @@ #include "state_tracker/st_cb_fbo.h" - -/** Cast wrapper */ -static INLINE struct intel_context * -intel_context_mesa(GLcontext * ctx) -{ - return (struct intel_context *) ctx->DriverCtx; -} - - -/** XXX temporary - want to get rid of this */ -static struct intel_context * -intelScreenContext(struct intel_screen *intelScreen) -{ - /* - * This should probably change to have the screen allocate a dummy - * context at screen creation. For now just use the current context. - */ - - GET_CURRENT_CONTEXT(ctx); - if (ctx == NULL) { - /* need a context for the first time makecurrent is called (for hw lock - when allocating priv buffers) */ - if (intelScreen->dummyctxptr == NULL) { - _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; - } - return intelScreen->dummyctxptr; - } - - return intel_context_mesa(ctx); -} - - /** * Display a colorbuffer surface in an X window. * Used for SwapBuffers and flushing front buffer rendering. @@ -86,18 +52,17 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, const drm_clip_rect_t *rect) { struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - struct intel_context *intel; + struct intel_context *intel = intelScreen->dummyContext; DBG(SWAP, "%s\n", __FUNCTION__); - assert(dPriv); - - intel = intelScreenContext(intelScreen); - if (!intel) + if (!intel) { + /* XXX this is where some kind of extra/meta context could be useful */ return; + } if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); driFenceUnReference(intel->last_swap_fence); intel->last_swap_fence = NULL; } -- cgit v1.2.3 From c80d17b68dfbee091b43588f19acc2e8bf1e1522 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 10:07:31 -0700 Subject: don't include mtypes.h --- .../drivers/dri/intel_winsys/intel_batchbuffer.h | 29 +++++++++++++++++++++- 1 file changed, 28 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h index 620fafca48..43a9a9fa79 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h @@ -1,7 +1,34 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #ifndef INTEL_BATCHBUFFER_H #define INTEL_BATCHBUFFER_H -#include "mtypes.h" +#include "main/glheader.h" #include "dri_bufmgr.h" struct intel_context; -- cgit v1.2.3 From eef6783537c3bf3a454b2c56b1cf71d0c73e45ee Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 10:10:50 -0700 Subject: Remove use of Mesa/GL types/functions (s/GLuint/uint/ etc) --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 39 ++++++++++---------- .../drivers/dri/intel_winsys/intel_batchbuffer.h | 41 +++++++++++----------- 2 files changed, 39 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 6b7772ddb2..fc9010f58a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -26,7 +26,6 @@ **************************************************************************/ #include -#include "main/imports.c" #include "intel_batchbuffer.h" #include "intel_context.h" #include "intel_screen.h" @@ -72,13 +71,13 @@ */ static void -intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) +intel_dump_batchbuffer(uint offset, uint * ptr, uint count) { int i; - _mesa_printf("\n\n\nSTART BATCH (%d dwords):\n", count / 4); + printf("\n\n\nSTART BATCH (%d dwords):\n", count / 4); for (i = 0; i < count / 4; i += 1) - _mesa_printf("\t0x%08x\n", ptr[i]); - _mesa_printf("END BATCH\n\n\n"); + printf("\t0x%08x\n", ptr[i]); + printf("END BATCH\n\n\n"); } void @@ -172,9 +171,9 @@ intel_batchbuffer_free(struct intel_batchbuffer *batch) static void intel_batch_ioctl(struct intel_context *intel, - GLuint start_offset, - GLuint used, - GLboolean allow_unlock) + uint start_offset, + uint used, + boolean allow_unlock) { drmI830BatchBuffer batch; @@ -192,7 +191,7 @@ intel_batch_ioctl(struct intel_context *intel, if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, sizeof(batch))) { - _mesa_printf("DRM_I830_BATCHBUFFER: %d\n", -errno); + printf("DRM_I830_BATCHBUFFER: %d\n", -errno); UNLOCK_HARDWARE(intel); exit(1); } @@ -205,11 +204,11 @@ intel_batch_ioctl(struct intel_context *intel, */ static void do_flush_locked(struct intel_batchbuffer *batch, - GLuint used, - GLboolean allow_unlock) + uint used, + boolean allow_unlock) { - GLuint *ptr; - GLuint i; + uint *ptr; + uint i; unsigned fenceFlags; struct _DriFenceObject *fo; @@ -219,7 +218,7 @@ do_flush_locked(struct intel_batchbuffer *batch, * whole task should be done internally by the memory manager, and * that dma buffers probably need to be pinned within agp space. */ - ptr = (GLuint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, + ptr = (uint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, DRM_BO_HINT_ALLOW_UNFENCED_MAP); @@ -277,8 +276,8 @@ struct _DriFenceObject * intel_batchbuffer_flush(struct intel_batchbuffer *batch) { struct intel_context *intel = batch->intel; - GLuint used = batch->ptr - batch->map; - GLboolean was_locked = intel->locked; + uint used = batch->ptr - batch->map; + boolean was_locked = intel->locked; if (used == 0) return batch->last_fence; @@ -333,10 +332,10 @@ intel_batchbuffer_finish(struct intel_batchbuffer *batch) /* This is the only way buffers get added to the validate list. */ -GLboolean +boolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, struct _DriBufferObject *buffer, - GLuint flags, GLuint mask, GLuint delta) + uint flags, uint mask, uint delta) { assert(batch->nr_relocs < MAX_RELOCS); @@ -348,7 +347,7 @@ intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, r->buf = buffer; r->offset = batch->ptr - batch->map; r->delta = delta; - *(GLuint *) batch->ptr = 0x12345678; + *(uint *) batch->ptr = 0x12345678; } batch->ptr += 4; @@ -359,7 +358,7 @@ intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, void intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags) + const void *data, uint bytes, uint flags) { assert((bytes & 3) == 0); intel_batchbuffer_require_space(batch, bytes, flags); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h index 43a9a9fa79..82feafa21f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h @@ -28,7 +28,7 @@ #ifndef INTEL_BATCHBUFFER_H #define INTEL_BATCHBUFFER_H -#include "main/glheader.h" +#include "pipe/p_compiler.h" #include "dri_bufmgr.h" struct intel_context; @@ -44,8 +44,8 @@ struct intel_context; struct buffer_reloc { struct _DriBufferObject *buf; - GLuint offset; - GLuint delta; /* not needed? */ + uint offset; + uint delta; /* not needed? */ }; struct intel_batchbuffer @@ -55,20 +55,19 @@ struct intel_batchbuffer struct _DriBufferObject *buffer; struct _DriFenceObject *last_fence; - GLuint flags; + uint flags; drmBOList list; - GLuint list_count; - GLubyte *map; - GLubyte *ptr; + uint list_count; + ubyte *map; + ubyte *ptr; struct buffer_reloc reloc[MAX_RELOCS]; - GLuint nr_relocs; - GLuint size; + uint nr_relocs; + uint size; }; -struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context - *intel); +struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context *intel); void intel_batchbuffer_free(struct intel_batchbuffer *batch); @@ -86,22 +85,22 @@ void intel_batchbuffer_reset(struct intel_batchbuffer *batch); * intel_buffer_dword() calls. */ void intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags); + const void *data, uint bytes, uint flags); void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, - GLuint bytes); + uint bytes); -GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, - struct _DriBufferObject *buffer, - GLuint flags, - GLuint mask, GLuint offset); +boolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, + struct _DriBufferObject *buffer, + uint flags, + uint mask, uint offset); /* Inline functions - might actually be better off with these * non-inlined. Certainly better off switching all command packets to * be passed as structs rather than dwords, but that's a little bit of * work... */ -static INLINE GLuint +static INLINE uint intel_batchbuffer_space(struct intel_batchbuffer *batch) { return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); @@ -109,17 +108,17 @@ intel_batchbuffer_space(struct intel_batchbuffer *batch) static INLINE void -intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) +intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, uint dword) { assert(batch->map); assert(intel_batchbuffer_space(batch) >= 4); - *(GLuint *) (batch->ptr) = dword; + *(uint *) (batch->ptr) = dword; batch->ptr += 4; } static INLINE void intel_batchbuffer_require_space(struct intel_batchbuffer *batch, - GLuint sz, GLuint flags) + uint sz, uint flags) { assert(sz < batch->size - 8); if (intel_batchbuffer_space(batch) < sz || -- cgit v1.2.3 From 7b358e8ea9d894470097824b3dda3d60ef232abe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 6 Nov 2007 19:00:23 +0100 Subject: Fix build error: dereferencing pointer to incomplete type --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 086ad04d1e..5499315926 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -41,6 +41,7 @@ #include "pipe/draw/draw_private.h" #include "pipe/draw/draw_vertex.h" #include "pipe/p_util.h" +#include "pipe/p_winsys.h" #include "softpipe/sp_context.h" #include "softpipe/sp_headers.h" -- cgit v1.2.3 From 26eb608a352ec017b534579e1c81c2d1fedf3d1b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 6 Nov 2007 17:38:47 +0000 Subject: Handle the corner case of 24bit depth buffer with 0bit stencil buffer. --- src/mesa/state_tracker/st_framebuffer.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index b43b9b7b9b..4833d10322 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -43,6 +43,7 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, = CALLOC_STRUCT(st_framebuffer); if (stfb) { GLboolean swStencil = (visual->stencilBits > 0 && + visual->stencilBits != 8 && visual->depthBits != 24); GLenum rgbFormat = (visual->redBits == 5 ? GL_RGB5 : GL_RGBA8); @@ -62,13 +63,14 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); } - if (visual->depthBits == 24 && visual->stencilBits == 8) { + if (visual->depthBits == 24) { /* combined depth/stencil buffer */ struct gl_renderbuffer *depthStencilRb = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); /* note: bind RB to two attachment points */ _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL,depthStencilRb); + if(visual->stencilBits == 8) + _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL,depthStencilRb); } else if (visual->depthBits == 16) { /* just 16-bit depth buffer, no hw stencil */ @@ -76,7 +78,8 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } - + else + assert(0); /* now add any/all software-based renderbuffers we may need */ _mesa_add_soft_renderbuffers(&stfb->Base, -- cgit v1.2.3 From 2e784848e68c0c5252631e6f2a6b021fe008f53a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 13:39:02 -0700 Subject: Call xmesa_init_driver_functions() to install xmesa_viewport() function. Fixes failed assertion in glxgears. --- src/mesa/pipe/xlib/xm_api.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index ae96a866d7..4140bf23d0 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1427,8 +1427,7 @@ xmesa_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) * we implement in this driver. */ static void -xmesa_init_driver_functions( XMesaVisual xmvisual, - struct dd_function_table *driver ) +xmesa_init_driver_functions(struct dd_function_table *driver) { driver->Flush = finish_or_flush; driver->Finish = finish_or_flush; @@ -1451,8 +1450,6 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) XMesaContext c; GLcontext *mesaCtx; - (void) xmesa_init_driver_functions; - if (firstTime) { _glthread_INIT_MUTEX(_xmesa_lock); firstTime = GL_FALSE; @@ -1485,6 +1482,8 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) mesaCtx->Const.CheckArrayBounds = GL_TRUE; #endif + xmesa_init_driver_functions(&mesaCtx->Driver); + /* finish up xmesa context initializations */ c->swapbytes = CHECK_BYTE_ORDER(v) ? GL_FALSE : GL_TRUE; c->xm_visual = v; -- cgit v1.2.3 From 0862df21545df4cac56f0c182e8818573131f951 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 13:42:37 -0700 Subject: remove debug abort() call --- src/mesa/pipe/xlib/xm_api.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 4140bf23d0..67ef797994 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1392,7 +1392,6 @@ finish_or_flush( GLcontext *ctx ) _glthread_UNLOCK_MUTEX(_xmesa_lock); } #endif - abort(); } -- cgit v1.2.3 From fa44b74f4ec1a51fcbe656c5da94b0635defa992 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 13:52:57 -0700 Subject: define CLIP_TILE as in sp_surface.c --- src/mesa/pipe/xlib/xm_surface.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 6250e75de8..9969cc728d 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -72,6 +72,10 @@ const int xmesa_kernel1[16] = { #define CLIP_TILE \ do { \ + if (x >= ps->width) \ + return; \ + if (y >= ps->height) \ + return; \ if (x + w > ps->width) \ w = ps->width - x; \ if (y + h > ps->height) \ -- cgit v1.2.3 From da3994988b95d145a98cd9130d719b36c0940406 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 13:53:32 -0700 Subject: init luminance/intensity_bits to zero in st_get_format_info() --- src/mesa/state_tracker/st_format.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index e9839e2936..ddba650941 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -153,6 +153,8 @@ st_get_format_info( pinfo->alpha_bits = format_bits( info, PIPE_FORMAT_COMP_A ); pinfo->depth_bits = format_bits( info, PIPE_FORMAT_COMP_Z ); pinfo->stencil_bits = format_bits( info, PIPE_FORMAT_COMP_S ); + pinfo->luminance_bits = 0; + pinfo->intensity_bits = 0; /* Format size */ pinfo->size = format_size( info ) / 8; @@ -162,12 +164,10 @@ st_get_format_info( pf_swizzle_y(info) == PIPE_FORMAT_COMP_R && pf_swizzle_z(info) == PIPE_FORMAT_COMP_R ) { if( pf_swizzle_w(info) == PIPE_FORMAT_COMP_R ) { - pinfo->luminance_bits = 0; pinfo->intensity_bits = pinfo->red_bits; } else { pinfo->luminance_bits = pinfo->red_bits; - pinfo->intensity_bits = 0; } pinfo->red_bits = 0; } -- cgit v1.2.3 From ea286d4df270897ca2a8f9e5e41b82cea419bdae Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 13:58:25 -0700 Subject: need to call pipe->flush() in finish_or_flush() --- src/mesa/pipe/xlib/xm_api.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 67ef797994..2d07f17bfe 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1386,6 +1386,7 @@ finish_or_flush( GLcontext *ctx ) /* NOT_NEEDED */ #else const XMesaContext xmesa = XMESA_CONTEXT(ctx); + ctx->st->pipe->flush(ctx->st->pipe, 0); if (xmesa) { _glthread_LOCK_MUTEX(_xmesa_lock); XSync( xmesa->display, False ); -- cgit v1.2.3 From 5e24e3c4266779704fc30737ac5e005ba71fd797 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 10:14:53 -0700 Subject: code clean-up --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 39 +++++++--------------- 1 file changed, 12 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index fc9010f58a..89d72c9448 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -80,10 +80,10 @@ intel_dump_batchbuffer(uint offset, uint * ptr, uint count) printf("END BATCH\n\n\n"); } + void intel_batchbuffer_reset(struct intel_batchbuffer *batch) { - int i; if (batch->map) { @@ -91,11 +91,9 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) batch->map = NULL; } - /* * Get a new, free batchbuffer. */ - batch->size = BATCH_SZ; driBOData(batch->buffer, batch->size, NULL, 0); @@ -104,7 +102,6 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) /* * Unreference buffers previously on the relocation list. */ - for (i = 0; i < batch->nr_relocs; i++) { struct buffer_reloc *r = &batch->reloc[i]; driBOUnReference(r->buf); @@ -119,7 +116,6 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) * while it's on the list. */ - driBOAddListItem(&batch->list, batch->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE); @@ -129,6 +125,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) batch->ptr = batch->map; } + /*====================================================================== * Public functions */ @@ -148,6 +145,7 @@ intel_batchbuffer_alloc(struct intel_context *intel) return batch; } + void intel_batchbuffer_free(struct intel_batchbuffer *batch) { @@ -167,13 +165,9 @@ intel_batchbuffer_free(struct intel_batchbuffer *batch) } - - static void intel_batch_ioctl(struct intel_context *intel, - uint start_offset, - uint used, - boolean allow_unlock) + uint start_offset, uint used, boolean allow_unlock) { drmI830BatchBuffer batch; @@ -198,18 +192,14 @@ intel_batch_ioctl(struct intel_context *intel, } - - /* TODO: Push this whole function into bufmgr. */ static void do_flush_locked(struct intel_batchbuffer *batch, - uint used, - boolean allow_unlock) + uint used, boolean allow_unlock) { uint *ptr; - uint i; - unsigned fenceFlags; + uint i, fenceFlags; struct _DriFenceObject *fo; driBOValidateList(batch->intel->driFd, &batch->list); @@ -221,7 +211,6 @@ do_flush_locked(struct intel_batchbuffer *batch, ptr = (uint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, DRM_BO_HINT_ALLOW_UNFENCED_MAP); - for (i = 0; i < batch->nr_relocs; i++) { struct buffer_reloc *r = &batch->reloc[i]; @@ -242,26 +231,22 @@ do_flush_locked(struct intel_batchbuffer *batch, * Kernel fencing. The flags tells the kernel that we've * programmed an MI_FLUSH. */ - fenceFlags = DRM_I915_FENCE_FLAG_FLUSHED; - fo = driFenceBuffers(batch->intel->driFd, - "Batch fence", fenceFlags); + fo = driFenceBuffers(batch->intel->driFd, "Batch fence", fenceFlags); /* * User space fencing. */ - driBOFence(batch->buffer, fo); if (driFenceType(fo) == DRM_FENCE_TYPE_EXE) { - /* * Oops. We only validated a batch buffer. This means we * didn't do any proper rendering. Discard this fence object. */ - driFenceUnReference(fo); - } else { + } + else { driFenceUnReference(batch->last_fence); batch->last_fence = fo; for (i = 0; i < batch->nr_relocs; i++) { @@ -277,12 +262,12 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch) { struct intel_context *intel = batch->intel; uint used = batch->ptr - batch->map; - boolean was_locked = intel->locked; + const boolean was_locked = intel->locked; if (used == 0) return batch->last_fence; -#define MI_FLUSH ((0<<29)|(4<<23)) +#define MI_FLUSH ((0 << 29) | (4 << 23)) /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a * performance drain that we would like to avoid. @@ -320,6 +305,7 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch) return batch->last_fence; } + void intel_batchbuffer_finish(struct intel_batchbuffer *batch) { @@ -355,7 +341,6 @@ intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, } - void intel_batchbuffer_data(struct intel_batchbuffer *batch, const void *data, uint bytes, uint flags) -- cgit v1.2.3 From 4cf56a418399c142335e0cab3fe3b8e54bfc703c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 10:19:27 -0700 Subject: implement context sharing code --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 16 ++++++++++------ 1 file changed, 10 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index ed6448646a..26420c848a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -35,7 +35,7 @@ #include "intel_batchbuffer.h" #include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" /* XXX temporary */ +#include "state_tracker/st_context.h" #include "pipe/p_defines.h" #include "pipe/p_context.h" @@ -75,18 +75,22 @@ static const struct dri_debug_control debug_control[] = { GLboolean -intelCreateContext(const __GLcontextModes * mesaVis, +intelCreateContext(const __GLcontextModes * visual, __DRIcontextPrivate * driContextPriv, void *sharedContextPrivate) { struct intel_context *intel = CALLOC_STRUCT(intel_context); - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; struct intel_screen *intelScreen = intel_screen(sPriv); drmI830Sarea *saPriv = intelScreen->sarea; int fthrottle_mode; GLboolean havePools; struct pipe_context *pipe; + struct st_context *st_share = NULL; + + if (sharedContextPrivate) { + st_share = ((struct intel_context *) sharedContextPrivate)->st; + } driContextPriv->driverPrivate = intel; intel->intelScreen = intelScreen; @@ -144,15 +148,15 @@ intelCreateContext(const __GLcontextModes * mesaVis, pipe = intel_create_i915simple( intel ); break; default: - _mesa_printf("Unknown PCIID %x in %s, using software driver\n", - intel->intelScreen->deviceID, __FUNCTION__); + fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", + intel->intelScreen->deviceID, __FUNCTION__); pipe = intel_create_softpipe( intel ); break; } } - intel->st = st_create_context(pipe, mesaVis, NULL); + intel->st = st_create_context(pipe, visual, st_share); intel->st->ctx->DriverCtx = intel; /* hope to get rid of this... */ return GL_TRUE; -- cgit v1.2.3 From 37624458e4e35569d9ca1d0452114a4c976682ec Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 10:25:56 -0700 Subject: remove more dependencies on Mesa types/functions --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 35 ++++++++++++------------ src/mesa/drivers/dri/intel_winsys/intel_screen.h | 10 +++---- 2 files changed, 23 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 899988340e..334803e264 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -53,7 +53,7 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_ALLOW_LARGE_TEXTURES(1) DRI_CONF_SECTION_END DRI_CONF_END; -const GLuint __driNConfigOptions = 4; +const uint __driNConfigOptions = 4; #ifdef USE_NEW_INTERFACE static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; @@ -74,6 +74,7 @@ intelPrintDRIInfo(struct intel_screen * intelScreen, fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); } + #if 0 static void intelPrintSAREA(const drmI830Sarea * sarea) @@ -103,7 +104,6 @@ intelPrintSAREA(const drmI830Sarea * sarea) #endif - /** * Use the information in the sarea to update the screen parameters * related to screen rotation. Needs to be called locked. @@ -138,7 +138,7 @@ intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) sarea->front_handle, intelScreen->front.size, (drmAddress *) & intelScreen->front.map) != 0) { - _mesa_problem(NULL, "drmMap(frontbuffer) failed!"); + fprintf(stderr, "drmMap(frontbuffer) failed!\n"); return; } @@ -156,8 +156,7 @@ intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) } - -GLboolean +boolean intelCreatePools(__DRIscreenPrivate * sPriv) { unsigned batchPoolSize = 1024*1024; @@ -196,7 +195,7 @@ intelCreatePools(__DRIscreenPrivate * sPriv) } -static GLboolean +static boolean intelInitDriver(__DRIscreenPrivate * sPriv) { struct intel_screen *intelScreen; @@ -270,10 +269,10 @@ intelDestroyScreen(__DRIscreenPrivate * sPriv) /** * This is called when we need to set up GL rendering to a new X window. */ -static GLboolean +static boolean intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, __DRIdrawablePrivate * driDrawPriv, - const __GLcontextModes * mesaVis, GLboolean isPixmap) + const __GLcontextModes * visual, boolean isPixmap) { if (isPixmap) { return GL_FALSE; /* not implemented */ @@ -283,7 +282,7 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, if (!intelfb) return GL_FALSE; - intelfb->stfb = st_create_framebuffer(mesaVis, GL_TRUE, (void*) intelfb); + intelfb->stfb = st_create_framebuffer(visual, GL_TRUE, (void*) intelfb); if (!intelfb->stfb) { free(intelfb); return GL_FALSE; @@ -320,8 +319,8 @@ intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) static void -intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, - unsigned long long offset, GLint depth, GLuint pitch) +intelSetTexOffset(__DRIcontext *pDRICtx, int texname, + unsigned long long offset, int depth, uint pitch) { abort(); #if 0 @@ -368,7 +367,7 @@ static const struct __DriverAPIRec intelAPI = { static __GLcontextModes * intelFillInModes(unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, GLboolean have_back_buffer) + unsigned stencil_bits, boolean have_back_buffer) { __GLcontextModes *modes; __GLcontextModes *m; @@ -499,11 +498,13 @@ __driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, (dri_priv->cpp == 2) ? 16 : 24, (dri_priv->cpp == 2) ? 0 : 8, 1); - /* Calling driInitExtensions here, with a NULL context pointer, does not actually - * enable the extensions. It just makes sure that all the dispatch offsets for all - * the extensions that *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is called, but we can't - * enable the extensions until we have a context pointer. + /* Calling driInitExtensions here, with a NULL context pointer, + * does not actually enable the extensions. It just makes sure + * that all the dispatch offsets for all the extensions that + * *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create + * is called, but we can't enable the extensions until we have a + * context pointer. * * Hello chicken. Hello egg. How are you two today? */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index 3bcbc22963..f0446fe7b3 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -92,19 +92,19 @@ intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); -extern GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); +extern boolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); -extern GLboolean +extern boolean intelMakeCurrent(__DRIcontextPrivate * driContextPriv, __DRIdrawablePrivate * driDrawPriv, __DRIdrawablePrivate * driReadPriv); -extern GLboolean +extern boolean intelCreatePools(__DRIscreenPrivate *sPriv); -extern GLboolean -intelCreateContext(const __GLcontextModes * mesaVis, +extern boolean +intelCreateContext(const __GLcontextModes * visual, __DRIcontextPrivate * driContextPriv, void *sharedContextPrivate); -- cgit v1.2.3 From db01ff6a19fb2365d4185321654bdfa09dae653f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 14:39:53 -0700 Subject: document flags param to region_alloc() --- src/mesa/pipe/p_winsys.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 298b555651..5bc6e6475a 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -72,6 +72,9 @@ struct pipe_winsys const char *, ... ); + /** + * flags is bitmask of PIPE_SURFACE_FLAG_RENDER, PIPE_SURFACE_FLAG_TEXTURE + */ struct pipe_region *(*region_alloc)(struct pipe_winsys *ws, unsigned cpp, unsigned width, unsigned height, unsigned flags); -- cgit v1.2.3 From d35cb1a121c272bf5be29fc62d59f4f71724201a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 14:59:17 -0700 Subject: remove dead code in softpipe_is_format_supported() --- src/mesa/pipe/softpipe/sp_context.c | 28 ++-------------------------- 1 file changed, 2 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 954884e7e9..09c5a7152d 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -51,35 +51,11 @@ * parameter or another function. */ static boolean -softpipe_is_format_supported( struct pipe_context *pipe, - uint format ) +softpipe_is_format_supported( struct pipe_context *pipe, uint format ) { -#if 0 - /* XXX: This is broken -- rewrite if still needed. */ - static const unsigned supported[] = { - PIPE_FORMAT_U_R8_G8_B8_A8, - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_U_L8, - PIPE_FORMAT_U_A8, - PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_L8_A8, - PIPE_FORMAT_S_R16_G16_B16_A16, - PIPE_FORMAT_YCBCR, - PIPE_FORMAT_YCBCR_REV, - PIPE_FORMAT_U_Z16, - PIPE_FORMAT_U_Z32, - PIPE_FORMAT_F_Z32, - PIPE_FORMAT_S8_Z24, - PIPE_FORMAT_U_S8 - }; - - *numFormats = sizeof(supported)/sizeof(supported[0]); - return supported; -#else struct softpipe_context *softpipe = softpipe_context( pipe ); + /* ask winsys if the format is supported */ return softpipe->winsys->is_format_supported( softpipe->winsys, format ); -#endif } -- cgit v1.2.3 From 5fb6ebf85d778aa68df96ccf71fcaba8cf691b32 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 15:15:49 -0700 Subject: disable finish_or_flush() for now --- src/mesa/pipe/xlib/xm_api.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 2d07f17bfe..0e303d597a 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1383,7 +1383,7 @@ static void finish_or_flush( GLcontext *ctx ) { #ifdef XFree86Server - /* NOT_NEEDED */ + /* NOT_NEEDED */ #else const XMesaContext xmesa = XMESA_CONTEXT(ctx); ctx->st->pipe->flush(ctx->st->pipe, 0); @@ -1429,8 +1429,12 @@ xmesa_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) static void xmesa_init_driver_functions(struct dd_function_table *driver) { +#if 0 /* not needed for now */ driver->Flush = finish_or_flush; driver->Finish = finish_or_flush; +#else + (void) finish_or_flush; +#endif driver->Viewport = xmesa_viewport; } -- cgit v1.2.3 From 54b43da7e9ae08c9078ba2ca4f8c242dd6aa3bff Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 15:16:22 -0700 Subject: rename xmesa_create_softpipe() xmesa_create_context() --- src/mesa/pipe/xlib/xm_winsys.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 45ece8ef55..5763a4e53f 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -379,7 +379,7 @@ xmesa_get_softpipe_winsys(void) struct pipe_context * -xmesa_create_softpipe(XMesaContext xmesa) +xmesa_create_context(XMesaContext xmesa) { struct pipe_winsys *pws = xmesa_get_pipe_winsys(); struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(); -- cgit v1.2.3 From a151ad21bb5f5b52f1c1f28f2f09d5c5f5915028 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 15:16:42 -0700 Subject: rename xmesa_create_softpipe() xmesa_create_context() --- src/mesa/pipe/xlib/xm_api.c | 2 +- src/mesa/pipe/xlib/xmesaP.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 0e303d597a..939dfe745b 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1464,7 +1464,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) if (!c) return NULL; - pipe = xmesa_create_softpipe( c ); + pipe = xmesa_create_context( c ); c->st = st_create_context(pipe, &v->mesa_visual, share_list ? share_list->st : NULL); diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index f1ca2a8fe8..c70dc67608 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -539,7 +539,7 @@ extern void xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, uint value); extern struct pipe_context * -xmesa_create_softpipe(XMesaContext xm); +xmesa_create_context(XMesaContext xm); extern struct pipe_surface * xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); -- cgit v1.2.3 From 3470d819fd7e3d3dd259d6fb2d4b963a514f0520 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 15:17:20 -0700 Subject: realloc surface if format changes, remove dead code --- src/mesa/state_tracker/st_cb_fbo.c | 39 +++++++++++++------------------------- 1 file changed, 13 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 1ffd1a30b8..7c18f380cd 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -61,7 +61,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); - const GLuint pipeFormat + const uint pipeFormat = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE); struct pipe_format_info info; GLuint cpp; @@ -85,6 +85,12 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, cpp = info.size; + if (strb->surface && strb->surface->format != pipeFormat) { + /* need to change surface types, free this surface */ + pipe_surface_reference(&strb->surface, NULL); + assert(strb->surface == NULL); + } + if (!strb->surface) { strb->surface = pipe->winsys->surface_alloc(pipe->winsys, pipeFormat); assert(strb->surface); @@ -187,28 +193,10 @@ st_new_renderbuffer(GLcontext *ctx, GLuint name) } -#if 000 -struct gl_renderbuffer * -st_new_renderbuffer_fb(struct pipe_region *region, GLuint width, GLuint height) -{ - struct st_renderbuffer *strb = CALLOC_STRUCT(st_renderbuffer); - if (!strb) - return; - - _mesa_init_renderbuffer(&strb->Base, name); - strb->Base.Delete = st_renderbuffer_delete; - strb->Base.AllocStorage = st_renderbuffer_alloc_storage; - strb->Base.GetPointer = null_get_pointer; - strb->Base.Width = width; - strb->Base.Heigth = height; - - strb->region = region; - - return &strb->Base; -} - -#else - +/** + * Allocate a renderbuffer for a an on-screen window (not a user-created + * renderbuffer). The window system code determines the internal format. + */ struct gl_renderbuffer * st_new_renderbuffer_fb(GLenum intFormat) { @@ -221,7 +209,7 @@ st_new_renderbuffer_fb(GLenum intFormat) } _mesa_init_renderbuffer(&strb->Base, 0); - strb->Base.ClassID = 0x42; /* XXX temp */ + strb->Base.ClassID = 0x4242; /* just a unique value */ strb->Base.InternalFormat = intFormat; switch (intFormat) { @@ -251,12 +239,11 @@ st_new_renderbuffer_fb(GLenum intFormat) strb->Base.AllocStorage = st_renderbuffer_alloc_storage; strb->Base.GetPointer = null_get_pointer; - /* surface is allocate in alloc_renderbuffer_storage() */ + /* surface is allocated in st_renderbuffer_alloc_storage() */ strb->surface = NULL; return &strb->Base; } -#endif -- cgit v1.2.3 From 2b0d1b6eda6b2a6c2df66a1e1a0dfd3aa762ece5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 15:19:36 -0700 Subject: Update xmesa_is_format_supported() --- src/mesa/pipe/xlib/xm_winsys.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 5763a4e53f..35e485390a 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -347,10 +347,17 @@ xmesa_get_pipe_winsys(void) static boolean xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) { + /* Any format supported by softpipe can be listed here. + * This query is not used for allocating window-system color buffers + * (which would depend on the screen depth/bpp). + */ switch (format) { case PIPE_FORMAT_U_A8_R8_G8_B8: case PIPE_FORMAT_S_R16_G16_B16_A16: case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_U_S8: + case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_U_Z32: return TRUE; default: return FALSE; -- cgit v1.2.3 From a895910d8813a56d8fe3861e97c9b9a2a6b01589 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 16:40:05 -0700 Subject: disable PIPE_FORMAT_U_Z32, doesn't work ATM --- src/mesa/pipe/xlib/xm_winsys.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 35e485390a..f7e55ed8f7 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -357,7 +357,7 @@ xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) case PIPE_FORMAT_S8_Z24: case PIPE_FORMAT_U_S8: case PIPE_FORMAT_U_Z16: - case PIPE_FORMAT_U_Z32: + /*case PIPE_FORMAT_U_Z32:*/ return TRUE; default: return FALSE; -- cgit v1.2.3 From b0e7da86abba5fc9040332a4b8ca81628a343956 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 17:24:45 -0700 Subject: restore dispatch/extension code (without, dynamic entrypoints aren't added) --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 58 +++++++++++++++++++++++ 1 file changed, 58 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 26420c848a..f565e2ccf8 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -47,6 +47,23 @@ int __intel_debug = 0; #endif +#define need_GL_ARB_multisample +#define need_GL_ARB_point_parameters +#define need_GL_ARB_texture_compression +#define need_GL_ARB_vertex_buffer_object +#define need_GL_ARB_vertex_program +#define need_GL_ARB_window_pos +#define need_GL_EXT_blend_color +#define need_GL_EXT_blend_equation_separate +#define need_GL_EXT_blend_func_separate +#define need_GL_EXT_blend_minmax +#define need_GL_EXT_cull_vertex +#define need_GL_EXT_fog_coord +#define need_GL_EXT_framebuffer_object +#define need_GL_EXT_multi_draw_arrays +#define need_GL_EXT_secondary_color +#define need_GL_NV_vertex_program +#include "extension_helper.h" /** @@ -57,6 +74,47 @@ int __intel_debug = 0; * old i830-specific driver. */ const struct dri_extension card_extensions[] = { + {"GL_ARB_multisample", GL_ARB_multisample_functions}, + {"GL_ARB_multitexture", NULL}, + {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_texture_border_clamp", NULL}, + {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, + {"GL_ARB_texture_cube_map", NULL}, + {"GL_ARB_texture_env_add", NULL}, + {"GL_ARB_texture_env_combine", NULL}, + {"GL_ARB_texture_env_dot3", NULL}, + {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_rectangle", NULL}, + {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, + {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, + {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, + {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, + {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_subtract", NULL}, + {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, + {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, + {"GL_EXT_packed_depth_stencil", NULL}, + {"GL_EXT_pixel_buffer_object", NULL}, + {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, + {"GL_EXT_stencil_wrap", NULL}, + {"GL_EXT_texture_edge_clamp", NULL}, + {"GL_EXT_texture_env_combine", NULL}, + {"GL_EXT_texture_env_dot3", NULL}, + {"GL_EXT_texture_filter_anisotropic", NULL}, + {"GL_EXT_texture_lod_bias", NULL}, + {"GL_3DFX_texture_compression_FXT1", NULL}, + {"GL_APPLE_client_storage", NULL}, + {"GL_MESA_pack_invert", NULL}, + {"GL_MESA_ycbcr_texture", NULL}, + {"GL_NV_blend_square", NULL}, + {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, + {"GL_NV_vertex_program1_1", NULL}, + {"GL_SGIS_generate_mipmap", NULL }, {NULL, NULL} }; -- cgit v1.2.3 From 85c7683f1f5f1d0d9e95eb59270705a2975fa437 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 Nov 2007 17:34:32 -0700 Subject: Improved logic for setting up depth/stencil buffers in st_create_framebuffer() --- src/mesa/state_tracker/st_framebuffer.c | 55 ++++++++++++++++++++++++--------- 1 file changed, 41 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 4833d10322..b04dcdb79d 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -42,9 +42,6 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer); if (stfb) { - GLboolean swStencil = (visual->stencilBits > 0 && - visual->stencilBits != 8 && - visual->depthBits != 24); GLenum rgbFormat = (visual->redBits == 5 ? GL_RGB5 : GL_RGBA8); _mesa_initialize_framebuffer(&stfb->Base, visual); @@ -63,29 +60,59 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); } - if (visual->depthBits == 24) { + if (visual->depthBits == 24 && visual->stencilBits == 8) { /* combined depth/stencil buffer */ struct gl_renderbuffer *depthStencilRb = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); /* note: bind RB to two attachment points */ _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); - if(visual->stencilBits == 8) - _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL,depthStencilRb); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb); } - else if (visual->depthBits == 16) { - /* just 16-bit depth buffer, no hw stencil */ - struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); + else { + /* separate depth and/or stencil */ + + if (visual->depthBits == 32) { + /* 32-bit depth buffer */ + struct gl_renderbuffer *depthRb + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); + } + else if (visual->depthBits == 24) { + /* 24-bit depth buffer, ignore stencil bits */ + struct gl_renderbuffer *depthRb + = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); + } + else if (visual->depthBits > 0) { + /* 16-bit depth buffer */ + struct gl_renderbuffer *depthRb + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); + } + + if (visual->stencilBits > 0) { + /* 8-bit stencil */ + struct gl_renderbuffer *stencilRb + = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb); + } } - else - assert(0); + +#if 0 + if (visual->accumRedBits > 0) { + /* 16-bit/channel accum */ + struct gl_renderbuffer *accumRb + = st_new_renderbuffer_fb(GL_RGBA16); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb); + } +#endif /* now add any/all software-based renderbuffers we may need */ _mesa_add_soft_renderbuffers(&stfb->Base, GL_FALSE, /* never sw color */ GL_FALSE, /* never sw depth */ - swStencil, visual->accumRedBits > 0, + GL_FALSE, /* stencil */ + visual->accumRedBits > 0, GL_FALSE, /* never sw alpha */ GL_FALSE /* never sw aux */ ); } -- cgit v1.2.3 From f39a520892259bc3ff13b47423fb86cd7714c70a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 6 Nov 2007 19:16:40 +0000 Subject: Track hardware vertex buffer state changes. --- src/mesa/pipe/i915simple/i915_context.h | 6 +++- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 30 ++++++++---------- src/mesa/pipe/i915simple/i915_state_emit.c | 18 ++++++++--- src/mesa/pipe/i915simple/i915_state_immediate.c | 41 +++++++++++++++++++++++++ 4 files changed, 72 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 5c748ea896..ee430ebc90 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -109,7 +109,7 @@ struct i915_state /** Describes the current hardware vertex layout */ struct vertex_info vertex_info; - + unsigned id; /* track lost context events */ }; @@ -185,6 +185,9 @@ struct i915_context unsigned *batch_start; + /** Vertex buffer */ + struct pipe_buffer_handle *vbo; + struct i915_state current; unsigned hardware_dirty; @@ -211,6 +214,7 @@ struct i915_context #define I915_NEW_SAMPLER 0x400 #define I915_NEW_TEXTURE 0x800 #define I915_NEW_CONSTANTS 0x1000 +#define I915_NEW_VBO 0x2000 /* Driver's internally generated state flags: diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 5499315926..caae6c6f85 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -82,8 +82,6 @@ struct vbuf_stage { ushort *element_map; unsigned nr_elements; - struct pipe_buffer_handle *buf; - unsigned prim; struct i915_context *i915; @@ -279,7 +277,6 @@ static void vbuf_draw( struct draw_stage *stage ) unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ unsigned hwprim; unsigned i; - unsigned *ptr; switch(vbuf->prim) { case PIPE_PRIM_POINTS: @@ -304,7 +301,7 @@ static void vbuf_draw( struct draw_stage *stage ) if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); - if (!BEGIN_BATCH( 4 + (nr + 1)/2, 1 )) { + if (!BEGIN_BATCH( 1 + (nr + 1)/2, 1 )) { FLUSH_BATCH(); /* Make sure state is re-emitted after a flush: @@ -312,20 +309,12 @@ static void vbuf_draw( struct draw_stage *stage ) i915_update_derived( i915 ); i915_emit_hardware_state( i915 ); - if (!BEGIN_BATCH( 4 + (nr + 1)/2, 1 )) { + if (!BEGIN_BATCH( 1 + (nr + 1)/2, 1 )) { assert(0); return; } } - /* FIXME: don't do this every time */ - OUT_BATCH( _3DSTATE_LOAD_STATE_IMMEDIATE_1 | - I1_LOAD_S(0) | - I1_LOAD_S(1) | - (1)); - OUT_RELOC( vbuf->buf, I915_BUFFER_ACCESS_READ, 0 ); - OUT_BATCH( ((vertex_size/4) << 24) | /* vertex size in dwords */ - ((vertex_size/4) << 16) ); /* vertex pitch in dwords */ OUT_BATCH( _3DPRIMITIVE | PRIM_INDIRECT | hwprim | @@ -361,7 +350,7 @@ static void vbuf_flush_elements( struct draw_stage *stage ) vbuf->nr_elements = 0; - winsys->buffer_unmap(winsys, vbuf->buf); + winsys->buffer_unmap(winsys, i915->vbo); vbuf->nr_vertices = 0; @@ -375,13 +364,16 @@ static void vbuf_flush_elements( struct draw_stage *stage ) } /* FIXME: handle failure */ - if(!vbuf->buf) - vbuf->buf = winsys->buffer_create(winsys, 64); - winsys->buffer_data( winsys, vbuf->buf, + if(!i915->vbo) + i915->vbo = winsys->buffer_create(winsys, 64); + winsys->buffer_data( winsys, i915->vbo, VBUF_SIZE, NULL, I915_BUFFER_USAGE_LIT_VERTEX ); + + i915->dirty |= I915_NEW_VBO; + vbuf->vertex_map = winsys->buffer_map(winsys, - vbuf->buf, + i915->vbo, PIPE_BUFFER_FLAG_WRITE ); vbuf->vertex_ptr = vbuf->vertex_map; @@ -394,7 +386,9 @@ static void vbuf_flush_elements( struct draw_stage *stage ) static void vbuf_begin( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage(stage); + struct i915_context *i915 = vbuf->i915; + assert(!i915->dirty); vbuf->vertex_size = vbuf->i915->current.vertex_info.size * 4; } diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 29fe9d9718..ee44e526ca 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -62,7 +62,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) { /* XXX: there must be an easier way */ const unsigned dwords = ( 14 + - 5 + + 7 + I915_MAX_DYNAMIC + 8 + 2 + I915_TEX_UNITS*3 + @@ -72,7 +72,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) 6 ) * 3/2; /* plus 50% margin */ const unsigned relocs = ( I915_TEX_UNITS + - 2 + 3 ) * 3/2; /* plus 50% margin */ #if 0 @@ -134,16 +134,26 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(0); } - /* 5 dwords, 0 relocs */ + /* 7 dwords, 1 relocs */ if (i915->hardware_dirty & I915_HW_IMMEDIATE) { OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | + I1_LOAD_S(0) | + I1_LOAD_S(1) | I1_LOAD_S(2) | I1_LOAD_S(4) | I1_LOAD_S(5) | I1_LOAD_S(6) | - (3)); + (5)); + if(i915->vbo) + OUT_RELOC(i915->vbo, + I915_BUFFER_ACCESS_READ, + i915->current.immediate[I915_IMMEDIATE_S0]); + else + /* FIXME: we should not do this */ + OUT_BATCH(0); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S1]); OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S2]); OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S4]); OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S5]); diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 479d72a87f..d830bb78e8 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -45,6 +45,46 @@ +/*********************************************************************** + * S0,S1: Vertex buffer state. + */ +static void upload_S0S1(struct i915_context *i915) +{ + unsigned LIS0, LIS1; + + /* INTEL_NEW_VBO */ + /* TODO: re-use vertex buffers here? */ + LIS0 = 0; + + /* INTEL_NEW_VERTEX_SIZE -- do this where the vertex size is calculated! + */ + { + unsigned vertex_size = i915->current.vertex_info.size; + + LIS1 = ((vertex_size << 24) | + (vertex_size << 16)); + } + + /* INTEL_NEW_VBO */ + /* TODO: use a vertex generation number to track vbo changes */ + if (1 || + i915->current.immediate[I915_IMMEDIATE_S0] != LIS0 || + i915->current.immediate[I915_IMMEDIATE_S1] != LIS1) + { + i915->current.immediate[I915_IMMEDIATE_S0] = LIS0; + i915->current.immediate[I915_IMMEDIATE_S1] = LIS1; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + +const struct i915_tracked_state i915_upload_S0S1 = { + .dirty = I915_NEW_VBO | I915_NEW_VERTEX_FORMAT, + .update = upload_S0S1 +}; + + + + /*********************************************************************** * S4: Vertex format, rasterization state */ @@ -166,6 +206,7 @@ const struct i915_tracked_state i915_upload_S7 = { static const struct i915_tracked_state *atoms[] = { + &i915_upload_S0S1, &i915_upload_S2S4, &i915_upload_S5, &i915_upload_S6, -- cgit v1.2.3 From 3922baede207c64ce07ec2ac19ffab04f7035483 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 Nov 2007 11:04:02 +0000 Subject: Reuse the vertex buffer until the vertex size changes. --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 74 ++++++++++++++++--------------- 1 file changed, 39 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index caae6c6f85..c33cc2198c 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -56,6 +56,7 @@ static void vbuf_flush_elements( struct draw_stage *stage ); +static void vbuf_flush_vertices( struct draw_stage *stage ); #define VBUF_SIZE (64*1024) @@ -105,18 +106,16 @@ boolean overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) } -static boolean check_space( struct vbuf_stage *vbuf ) +static void check_space( struct vbuf_stage *vbuf ) { if (overflow( vbuf->vertex_map, vbuf->vertex_ptr, vbuf->vertex_size, VBUF_SIZE )) - return FALSE; + vbuf_flush_vertices(&vbuf->stage); if (vbuf->nr_elements + 4 > IBUF_SIZE / sizeof(ushort) ) - return FALSE; - - return TRUE; + vbuf_flush_elements(&vbuf->stage); } @@ -188,8 +187,7 @@ static void vbuf_tri( struct draw_stage *stage, struct vbuf_stage *vbuf = vbuf_stage( stage ); unsigned i; - if (!check_space( vbuf )) - vbuf_flush_elements( stage ); + check_space( vbuf ); for (i = 0; i < 3; i++) { if (prim->v[i]->vertex_id == 0xffff) @@ -206,8 +204,7 @@ static void vbuf_line(struct draw_stage *stage, struct vbuf_stage *vbuf = vbuf_stage( stage ); unsigned i; - if (!check_space( vbuf )) - vbuf_flush_elements( stage ); + check_space( vbuf ); for (i = 0; i < 2; i++) { if (prim->v[i]->vertex_id == 0xffff) @@ -223,8 +220,7 @@ static void vbuf_point(struct draw_stage *stage, { struct vbuf_stage *vbuf = vbuf_stage( stage ); - if (!check_space( vbuf )) - vbuf_flush_elements( stage ); + check_space( vbuf ); if (prim->v[0]->vertex_id == 0xffff) emit_vertex( vbuf, prim->v[0] ); @@ -269,7 +265,7 @@ static void vbuf_first_point( struct draw_stage *stage, } -static void vbuf_draw( struct draw_stage *stage ) +static void vbuf_flush_elements( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); struct i915_context *i915 = vbuf->i915; @@ -278,6 +274,9 @@ static void vbuf_draw( struct draw_stage *stage ) unsigned hwprim; unsigned i; + if(!nr) + return; + switch(vbuf->prim) { case PIPE_PRIM_POINTS: hwprim = PRIM3D_POINTLIST; @@ -295,6 +294,8 @@ static void vbuf_draw( struct draw_stage *stage ) assert(vbuf->vertex_ptr - vbuf->vertex_map == vbuf->nr_vertices * vertex_size / 4); + assert((i915->dirty & ~I915_NEW_VBO) == 0); + if (i915->dirty) i915_update_derived( i915 ); @@ -327,34 +328,31 @@ static void vbuf_draw( struct draw_stage *stage ) if (i < nr) { OUT_BATCH( vbuf->element_map[i] ); } + + vbuf->nr_elements = 0; } -static void vbuf_flush_elements( struct draw_stage *stage ) + +/** + * Flush vertex buffer. + */ +static void vbuf_flush_vertices( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); struct i915_context *i915 = vbuf->i915; struct pipe_winsys *winsys = i915->pipe.winsys; - if (vbuf->nr_elements) { -#if 0 - fprintf(stderr, "%s (%d elts, %d verts)\n", - __FUNCTION__, - vbuf->nr_elements, - vbuf->nr_vertices); -#endif - - /* Draw now or add to list of primitives??? - */ - vbuf_draw( stage ); + if(vbuf->nr_vertices) { + + vbuf_flush_elements(stage); - vbuf->nr_elements = 0; - winsys->buffer_unmap(winsys, i915->vbo); - + vbuf->nr_vertices = 0; - - /* Reset vertex ids? Actually, want to not do that unless our + + /** + * XXX: Reset vertex ids? Actually, want to not do that unless our * vertex buffer is full. Would like separate * flush-on-index-full and flush-on-vb-full, but may raise * issues uploading vertices if the hardware wants to flush when @@ -362,7 +360,7 @@ static void vbuf_flush_elements( struct draw_stage *stage ) */ draw_vertex_cache_reset_vertex_ids( vbuf->i915->draw ); } - + /* FIXME: handle failure */ if(!i915->vbo) i915->vbo = winsys->buffer_create(winsys, 64); @@ -376,20 +374,22 @@ static void vbuf_flush_elements( struct draw_stage *stage ) i915->vbo, PIPE_BUFFER_FLAG_WRITE ); vbuf->vertex_ptr = vbuf->vertex_map; - - stage->tri = vbuf_first_tri; - stage->line = vbuf_first_line; - stage->point = vbuf_first_point; } + static void vbuf_begin( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage(stage); struct i915_context *i915 = vbuf->i915; + unsigned vertex_size = i915->current.vertex_info.size * 4; assert(!i915->dirty); - vbuf->vertex_size = vbuf->i915->current.vertex_info.size * 4; + + if(vbuf->vertex_size != vertex_size) + vbuf_flush_vertices(stage); + + vbuf->vertex_size = vertex_size; } @@ -398,6 +398,10 @@ static void vbuf_end( struct draw_stage *stage ) /* Overkill. */ vbuf_flush_elements( stage ); + + stage->point = vbuf_first_point; + stage->line = vbuf_first_line; + stage->tri = vbuf_first_tri; } -- cgit v1.2.3 From c28fdf309607ec2994ef9a1109931a8389854300 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 Nov 2007 12:08:19 +0000 Subject: Use a consistent number to identify undefined vertices. --- src/mesa/pipe/draw/draw_clip.c | 2 +- src/mesa/pipe/draw/draw_private.h | 6 +++++- src/mesa/pipe/draw/draw_vertex_cache.c | 4 ++-- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 27 +++++++++++++++++---------- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 6 +++--- 5 files changed, 28 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 222022d6c1..bc62d422a4 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -102,7 +102,7 @@ static void interp( const struct clipper *clip, dst->clipmask = 0; dst->edgeflag = 0; dst->pad = 0; - dst->vertex_id = 0; + dst->vertex_id = UNDEFINED_VERTEX_ID; } /* Clip coordinates: interpolate normally diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index f52ff0bd44..bdc3a6b9e7 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -52,6 +52,7 @@ struct gallivm_prog; struct gallivm_cpu_engine; + /** * Basic vertex info. * Carry some useful information around with the vertices in the prim pipe. @@ -67,6 +68,9 @@ struct vertex_header { float data[][4]; /* Note variable size */ }; +/* NOTE: It should match vertex_id size above */ +#define UNDEFINED_VERTEX_ID 0xffff + /* XXX This is too large */ #define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) @@ -299,7 +303,7 @@ dup_vert( struct draw_stage *stage, { struct vertex_header *tmp = stage->tmp[idx]; memcpy(tmp, vert, stage->draw->vertex_info.size * sizeof(float) ); - tmp->vertex_id = ~0; + tmp->vertex_id = UNDEFINED_VERTEX_ID; return tmp; } diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index 511f371ac8..6689907ddf 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -88,7 +88,7 @@ static struct vertex_header *get_vertex( struct draw_context *draw, draw->vcache.vertex[slot]->clipmask = 0; draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ draw->vcache.vertex[slot]->pad = 0; - draw->vcache.vertex[slot]->vertex_id = ~0; + draw->vcache.vertex[slot]->vertex_id = UNDEFINED_VERTEX_ID; } @@ -130,7 +130,7 @@ void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) unsigned i; for (i = 0; i < Elements(draw->vcache.vertex); i++) - draw->vcache.vertex[i]->vertex_id = ~0; + draw->vcache.vertex[i]->vertex_id = UNDEFINED_VERTEX_ID; } diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index c33cc2198c..8881d16c88 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -98,15 +98,16 @@ static INLINE struct vbuf_stage *vbuf_stage( struct draw_stage *stage ) } -static inline -boolean overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) +static INLINE boolean +overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) { unsigned long used = (char *)ptr - (char *)map; return (used + bytes) > bufsz; } -static void check_space( struct vbuf_stage *vbuf ) +static INLINE void +check_space( struct vbuf_stage *vbuf ) { if (overflow( vbuf->vertex_map, vbuf->vertex_ptr, @@ -125,7 +126,7 @@ static void check_space( struct vbuf_stage *vbuf ) * have a couple of slots at the beginning (1-dword header, 4-dword * clip pos) that we ignore here. */ -static inline void +static INLINE void emit_vertex( struct vbuf_stage *vbuf, struct vertex_header *vertex ) { @@ -137,6 +138,15 @@ emit_vertex( struct vbuf_stage *vbuf, // fprintf(stderr, "emit vertex %d to %p\n", // vbuf->nr_vertices, vbuf->vertex_ptr); + if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { + if(vertex->vertex_id < vbuf->nr_vertices) + return; + else + fprintf(stderr, "Bad vertex id 0x%04x (>= 0x%04x)\n", + vertex->vertex_id, vbuf->nr_vertices); + return; + } + vertex->vertex_id = vbuf->nr_vertices++; for (i = 0; i < vinfo->num_attribs; i++) { @@ -190,8 +200,7 @@ static void vbuf_tri( struct draw_stage *stage, check_space( vbuf ); for (i = 0; i < 3; i++) { - if (prim->v[i]->vertex_id == 0xffff) - emit_vertex( vbuf, prim->v[i] ); + emit_vertex( vbuf, prim->v[i] ); vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; } @@ -207,8 +216,7 @@ static void vbuf_line(struct draw_stage *stage, check_space( vbuf ); for (i = 0; i < 2; i++) { - if (prim->v[i]->vertex_id == 0xffff) - emit_vertex( vbuf, prim->v[i] ); + emit_vertex( vbuf, prim->v[i] ); vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; } @@ -222,8 +230,7 @@ static void vbuf_point(struct draw_stage *stage, check_space( vbuf ); - if (prim->v[0]->vertex_id == 0xffff) - emit_vertex( vbuf, prim->v[0] ); + emit_vertex( vbuf, prim->v[0] ); vbuf->element_map[vbuf->nr_elements++] = prim->v[0]->vertex_id; } diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index ddf662eead..7cb3da6feb 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -140,7 +140,7 @@ static void vbuf_tri( struct draw_stage *stage, vbuf_flush_elements( stage ); for (i = 0; i < 3; i++) { - if (prim->v[i]->vertex_id == 0xffff) + if (prim->v[i]->vertex_id == UNDEFINED_VERTEX_ID) emit_vertex( vbuf, prim->v[i] ); vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; @@ -158,7 +158,7 @@ static void vbuf_line(struct draw_stage *stage, vbuf_flush_elements( stage ); for (i = 0; i < 2; i++) { - if (prim->v[i]->vertex_id == 0xffff) + if (prim->v[i]->vertex_id == UNDEFINED_VERTEX_ID) emit_vertex( vbuf, prim->v[i] ); vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; @@ -174,7 +174,7 @@ static void vbuf_point(struct draw_stage *stage, if (!check_space( vbuf )) vbuf_flush_elements( stage ); - if (prim->v[0]->vertex_id == 0xffff) + if (prim->v[0]->vertex_id == UNDEFINED_VERTEX_ID) emit_vertex( vbuf, prim->v[0] ); vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[0]->vertex_id; -- cgit v1.2.3 From 3e22180fc893bb09bf6b990bc4e858fd85f522ab Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 Nov 2007 13:07:20 +0000 Subject: Reset temporary vertices ids. --- src/mesa/pipe/draw/draw_context.c | 27 +++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_private.h | 10 ++++++++++ src/mesa/pipe/i915simple/i915_prim_vbuf.c | 4 +++- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 2 +- 4 files changed, 41 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 80317dd6cf..0de8bed529 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -243,6 +243,18 @@ void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) } } +/** + * Reset the verticies ids of this and subsequent stages. + */ +void draw_reset_tmps( struct draw_stage *stage ) +{ + unsigned i; + + if (stage->tmp) + for (i = 0; i < stage->nr_tmps; i++) + stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; +} + void draw_free_tmps( struct draw_stage *stage ) { if (stage->tmp) { @@ -251,9 +263,24 @@ void draw_free_tmps( struct draw_stage *stage ) } } + boolean draw_use_sse(struct draw_context *draw) { return (boolean) draw->use_sse; } +void draw_reset_vertex_ids(struct draw_context *draw) +{ + struct draw_stage *stage = draw->pipeline.first; + + while (stage) { + if (stage->reset_tmps) + stage->reset_tmps(stage); + else + draw_reset_tmps(stage); + stage = stage->next; + } + + draw_vertex_cache_reset_vertex_ids(draw); +} diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index bdc3a6b9e7..dc4057bff3 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -116,6 +116,13 @@ struct draw_stage void (*end)( struct draw_stage * ); + /** + * Reset temporary vertices ids in this stage + * + * draw_free_tmps will be called instead if null. + */ + void (*reset_tmps)( struct draw_stage * ); + void (*reset_stipple_counter)( struct draw_stage * ); }; @@ -254,8 +261,11 @@ extern struct draw_stage *draw_validate_stage( struct draw_context *context ); extern void draw_free_tmps( struct draw_stage *stage ); +extern void draw_reset_tmps( struct draw_stage *stage ); extern void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ); +extern void draw_reset_vertex_ids( struct draw_context *draw ); + extern int draw_vertex_cache_check_space( struct draw_context *draw, unsigned nr_verts ); diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 8881d16c88..a78c2b0a6c 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -290,9 +290,11 @@ static void vbuf_flush_elements( struct draw_stage *stage ) break; case PIPE_PRIM_LINES: hwprim = PRIM3D_LINELIST; + assert(nr % 2 == 0); break; case PIPE_PRIM_TRIANGLES: hwprim = PRIM3D_TRILIST; + assert(nr % 3 == 0); break; default: assert(0); @@ -365,7 +367,7 @@ static void vbuf_flush_vertices( struct draw_stage *stage ) * issues uploading vertices if the hardware wants to flush when * we flush. */ - draw_vertex_cache_reset_vertex_ids( vbuf->i915->draw ); + draw_reset_vertex_ids( vbuf->i915->draw ); } /* FIXME: handle failure */ diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index 7cb3da6feb..59cf1c4eb7 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -243,7 +243,7 @@ static void vbuf_flush_elements( struct draw_stage *stage ) * issues uploading vertices if the hardware wants to flush when * we flush. */ - draw_vertex_cache_reset_vertex_ids( vbuf->draw_context ); + draw_reset_vertex_ids( vbuf->draw_context ); } stage->tri = vbuf_first_tri; -- cgit v1.2.3 From 52236661653169140d07a500facd65185b6b3666 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 Nov 2007 13:21:01 +0000 Subject: Check the right ammount of free space in vertex buffer. --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index a78c2b0a6c..35174c64a0 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -59,7 +59,7 @@ static void vbuf_flush_elements( struct draw_stage *stage ); static void vbuf_flush_vertices( struct draw_stage *stage ); -#define VBUF_SIZE (64*1024) +#define VBUF_SIZE (128*1024) #define IBUF_SIZE (16*1024) @@ -107,15 +107,15 @@ overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) static INLINE void -check_space( struct vbuf_stage *vbuf ) +check_space( struct vbuf_stage *vbuf, unsigned nr ) { if (overflow( vbuf->vertex_map, vbuf->vertex_ptr, - vbuf->vertex_size, + nr*vbuf->vertex_size, VBUF_SIZE )) vbuf_flush_vertices(&vbuf->stage); - if (vbuf->nr_elements + 4 > IBUF_SIZE / sizeof(ushort) ) + if (vbuf->nr_elements + nr > IBUF_SIZE / sizeof(ushort) ) vbuf_flush_elements(&vbuf->stage); } @@ -197,7 +197,7 @@ static void vbuf_tri( struct draw_stage *stage, struct vbuf_stage *vbuf = vbuf_stage( stage ); unsigned i; - check_space( vbuf ); + check_space( vbuf, 3 ); for (i = 0; i < 3; i++) { emit_vertex( vbuf, prim->v[i] ); @@ -213,7 +213,7 @@ static void vbuf_line(struct draw_stage *stage, struct vbuf_stage *vbuf = vbuf_stage( stage ); unsigned i; - check_space( vbuf ); + check_space( vbuf, 2 ); for (i = 0; i < 2; i++) { emit_vertex( vbuf, prim->v[i] ); @@ -228,7 +228,7 @@ static void vbuf_point(struct draw_stage *stage, { struct vbuf_stage *vbuf = vbuf_stage( stage ); - check_space( vbuf ); + check_space( vbuf, 1 ); emit_vertex( vbuf, prim->v[0] ); @@ -435,6 +435,8 @@ struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) vbuf->stage.end = vbuf_end; vbuf->stage.reset_stipple_counter = reset_stipple_counter; + assert(IBUF_SIZE < UNDEFINED_VERTEX_ID); + /* FIXME: free this memory on takedown */ vbuf->element_map = malloc( IBUF_SIZE ); vbuf->vertex_map = NULL; -- cgit v1.2.3 From 7d1a04e499564212a2a9aace12b05f424a357d3f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 08:05:09 -0700 Subject: Add winsys->surface_release() to complement winsys->surface_alloc(). pipe_surface now has a pointer to the winsys which create/owns the surface. This allows clean surface deallocation w/out a rendering context. --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 17 +++++++++++++++ src/mesa/pipe/p_inlines.h | 16 +++++---------- src/mesa/pipe/p_state.h | 3 +++ src/mesa/pipe/p_winsys.h | 2 ++ src/mesa/pipe/xlib/xm_api.c | 11 +++------- src/mesa/pipe/xlib/xm_buffer.c | 24 ++++++++-------------- src/mesa/pipe/xlib/xm_surface.c | 8 ++++---- src/mesa/pipe/xlib/xm_winsys.c | 15 +++++++++++++- src/mesa/pipe/xlib/xmesaP.h | 11 +++++----- src/mesa/state_tracker/st_cb_fbo.c | 17 ++++----------- 10 files changed, 66 insertions(+), 58 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index cc76a40a5a..56a600aebe 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -264,11 +264,27 @@ intel_i915_surface_alloc(struct pipe_winsys *winsys, unsigned format) if (surf) { surf->format = format; surf->refcount = 1; + surf->winsys = winsys; } return surf; } +static void +intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + surf->refcount--; + if (surf->refcount == 0) { + if (surf->region) + winsys->region_release(winsys, &surf->region); + free(surf); + } + *s = NULL; +} + + + static void intel_printf( struct pipe_winsys *sws, const char *fmtString, ... ) { @@ -315,6 +331,7 @@ intel_create_pipe_winsys( struct intel_context *intel ) iws->winsys.region_release = intel_i915_region_release; iws->winsys.surface_alloc = intel_i915_surface_alloc; + iws->winsys.surface_release = intel_i915_surface_release; return &iws->winsys; } diff --git a/src/mesa/pipe/p_inlines.h b/src/mesa/pipe/p_inlines.h index ea666fa20a..2418d016e1 100644 --- a/src/mesa/pipe/p_inlines.h +++ b/src/mesa/pipe/p_inlines.h @@ -29,7 +29,8 @@ #define P_INLINES_H #include "p_context.h" -//#include "p_util.h" +#include "p_winsys.h" + /** * Set 'ptr' to point to 'region' and update reference counting. @@ -68,16 +69,9 @@ pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) { assert(ptr); if (*ptr) { - /* unreference the old thing */ - struct pipe_surface *oldSurf = *ptr; - assert(oldSurf->refcount > 0); - oldSurf->refcount--; - if (oldSurf->refcount == 0) { - /* free the old region */ - pipe_region_reference(&oldSurf->region, NULL); - FREE( oldSurf ); - } - *ptr = NULL; + struct pipe_winsys *winsys = (*ptr)->winsys; + winsys->surface_release(winsys, ptr); + assert(!*ptr); } if (surf) { /* reference the new thing */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 0e1a038c2e..848c32701f 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -60,6 +60,8 @@ struct pipe_surface; /* opaque type */ struct pipe_buffer_handle; +struct pipe_winsys; + /*** *** State objects @@ -281,6 +283,7 @@ struct pipe_surface unsigned width, height; unsigned offset; /**< offset from start of region, in bytes */ unsigned refcount; + struct pipe_winsys *winsys; /**< winsys which owns/created the surface */ }; diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 5bc6e6475a..e4c888ad33 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -86,6 +86,8 @@ struct pipe_winsys struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws, unsigned format); + void (*surface_release)(struct pipe_winsys *ws, struct pipe_surface **s); + /** * The buffer manager is modeled after the dri_bufmgr interface, which * in turn is modeled after the ARB_vertex_buffer_object extension, diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 939dfe745b..ec889ca34f 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -301,6 +301,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, { XMesaBuffer b; GLframebuffer *fb; + struct pipe_winsys *winsys = xmesa_get_pipe_winsys(); ASSERT(type == WINDOW || type == PIXMAP || type == PBUFFER); @@ -328,7 +329,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, /* * Front renderbuffer */ - b->frontxrb = xmesa_create_renderbuffer(NULL, 0, &vis->mesa_visual, GL_FALSE); + b->frontxrb = xmesa_create_renderbuffer(winsys, 0, &vis->mesa_visual, GL_FALSE); if (!b->frontxrb) { _mesa_free(b); return NULL; @@ -337,18 +338,12 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, b->frontxrb->drawable = d; b->frontxrb->pixmap = (XMesaPixmap) d; _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &b->frontxrb->St.Base); -#if 0 /* sketch... */ - { - struct pipe_surface *front_surf; - front_surf = xmesa_create_front_surface(vis, d); - } -#endif /* * Back renderbuffer */ if (vis->mesa_visual.doubleBufferMode) { - b->backxrb = xmesa_create_renderbuffer(NULL, 0, &vis->mesa_visual, GL_TRUE); + b->backxrb = xmesa_create_renderbuffer(winsys, 0, &vis->mesa_visual, GL_TRUE); if (!b->backxrb) { /* XXX free front xrb too */ _mesa_free(b); diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index b4a05ac685..7dc85bf2eb 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -241,6 +241,12 @@ alloc_back_buffer(XMesaBuffer b, GLuint width, GLuint height) static void xmesa_delete_renderbuffer(struct gl_renderbuffer *rb) { + struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); + if (xrb->St.surface) { + struct pipe_winsys *ws = xrb->St.surface->winsys; + ws->surface_release(ws, &xrb->St.surface); + } + /* XXX Note: the ximage or Pixmap attached to this renderbuffer * should probably get freed here, but that's currently done in * XMesaDestroyBuffer(). @@ -289,11 +295,6 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, if (!xrb->St.surface || !xrb->St.surface->region) finish_surface_init(ctx, xrb); -#if 0 - xmesa_set_renderbuffer_funcs(xrb, xmesa->pixelformat, - xmesa->xm_visual->BitsPerPixel); -#endif - /* surface info */ xms->surface.width = width; xms->surface.height = height; @@ -353,11 +354,6 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xrb->origin4 = NULL; } -#if 0 - xmesa_set_renderbuffer_funcs(xrb, xmesa->pixelformat, - xmesa->xm_visual->BitsPerPixel); -#endif - if (!xrb->St.surface || !xrb->St.surface->region) finish_surface_init(ctx, xrb); @@ -381,11 +377,11 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, * renderbuffers. */ struct xmesa_renderbuffer * -xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, +xmesa_create_renderbuffer(struct pipe_winsys *winsys, + GLuint name, const GLvisual *visual, GLboolean backBuffer) { struct xmesa_renderbuffer *xrb = CALLOC_STRUCT(xmesa_renderbuffer); - struct pipe_context *pipe = NULL;/*ctx->st->pipe;*/ if (xrb) { GLuint name = 0; GLuint pipeFormat = 0; @@ -416,11 +412,9 @@ xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, xrb->St.Base.IndexBits = visual->indexBits; } /* only need to set Red/Green/EtcBits fields for user-created RBs */ - - xrb->St.surface = xmesa_new_color_surface(pipe, pipeFormat); + xrb->St.surface = xmesa_new_color_surface(winsys, pipeFormat); xms = (struct xmesa_surface *) xrb->St.surface; xms->xrb = xrb; - } return xrb; } diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 9969cc728d..43e5050747 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -609,7 +609,7 @@ clear_32bit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, * have special/unique quad read/write functions for X. */ struct pipe_surface * -xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) +xmesa_new_color_surface(struct pipe_winsys *winsys, GLuint pipeFormat) { struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); @@ -617,15 +617,14 @@ xmesa_new_color_surface(struct pipe_context *pipe, GLuint pipeFormat) xms->surface.format = pipeFormat; xms->surface.refcount = 1; + xms->surface.winsys = winsys; /* Note, the region we allocate doesn't actually have any storage * since we're drawing into an XImage or Pixmap. * The region's size will get set in the xmesa_alloc_front/back_storage() * functions. */ - if (pipe) - xms->surface.region = pipe->winsys->region_alloc(pipe->winsys, - 1, 0, 0, 0x0); + xms->surface.region = winsys->region_alloc(winsys, 1, 1, 1, 0x0); return &xms->surface; } @@ -645,6 +644,7 @@ xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) xms->surface.format = pipeFormat; xms->surface.refcount = 1; + xms->surface.winsys = pipe->winsys; return &xms->surface; } diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index f7e55ed8f7..b73fcab68d 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -284,6 +284,7 @@ xm_surface_alloc(struct pipe_winsys *ws, GLuint pipeFormat) xms->surface.format = pipeFormat; xms->surface.refcount = 1; + xms->surface.winsys = ws; #if 0 /* * This is really just a softpipe surface, not an XImage/Pixmap surface. @@ -295,13 +296,24 @@ xm_surface_alloc(struct pipe_winsys *ws, GLuint pipeFormat) +static void +xm_surface_release(struct pipe_winsys *ws, struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + if (surf->region) + winsys->region_release(winsys, &surf->region); + free(surf); + *s = NULL; +} + + /** * Return pointer to a pipe_winsys object. * For Xlib, this is a singleton object. * Nothing special for the Xlib driver so no subclassing or anything. */ -static struct pipe_winsys * +struct pipe_winsys * xmesa_get_pipe_winsys(void) { static struct pipe_winsys *ws = NULL; @@ -325,6 +337,7 @@ xmesa_get_pipe_winsys(void) ws->region_release = xm_region_release; ws->surface_alloc = xm_surface_alloc; + ws->surface_release = xm_surface_release; ws->flush_frontbuffer = xm_flush_frontbuffer; ws->wait_idle = xm_wait_idle; diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index c70dc67608..ba0ccdbcec 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -51,10 +51,6 @@ typedef struct { } bgr_t; -struct xmesa_renderbuffer; - - - /** Framebuffer pixel formats */ enum pixel_format { PF_Index, /**< Color Index mode */ @@ -451,7 +447,8 @@ extern const int xmesa_kernel1[16]; */ extern struct xmesa_renderbuffer * -xmesa_create_renderbuffer(GLcontext *ctx, GLuint name, const GLvisual *visual, +xmesa_create_renderbuffer(struct pipe_winsys *winsys, + GLuint name, const GLvisual *visual, GLboolean backBuffer); extern void @@ -545,8 +542,10 @@ extern struct pipe_surface * xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); extern struct pipe_surface * -xmesa_new_color_surface(struct pipe_context *pipe, GLuint format); +xmesa_new_color_surface(struct pipe_winsys *winsys, GLuint format); +extern struct pipe_winsys * +xmesa_get_pipe_winsys(void); extern void xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 7c18f380cd..0d23f7eec4 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -129,19 +129,10 @@ static void st_renderbuffer_delete(struct gl_renderbuffer *rb) { struct st_renderbuffer *strb = st_renderbuffer(rb); - GET_CURRENT_CONTEXT(ctx); - if (ctx) { - struct pipe_context *pipe = ctx->st->pipe; - ASSERT(strb); - if (strb && strb->surface) { - if (strb->surface->region) { - pipe->winsys->region_release(pipe->winsys, &strb->surface->region); - } - free(strb->surface); - } - } - else { - _mesa_warning(NULL, "st_renderbuffer_delete() called, but no current context"); + ASSERT(strb); + if (strb->surface) { + struct pipe_winsys *ws = strb->surface->winsys; + ws->surface_release(ws, &strb->surface); } free(strb); } -- cgit v1.2.3 From 6c4447e6a8ecfcfb094ab2ad8b7371dfa2b7d7bb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 08:18:34 -0700 Subject: fix typos, re-enable some extensions to get back to GL 2.1 level --- src/mesa/state_tracker/st_extensions.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 0707c489a6..4355eaa5f9 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -99,11 +99,16 @@ void st_init_extensions(struct st_context *st) /* * Extensions that are supported by all Gallium drivers: */ + ctx->Extensions.ARB_multisample = GL_TRUE; /* API support */ ctx->Extensions.ARB_fragment_program = GL_TRUE; + ctx->Extensions.ARB_texture_border_clamp = GL_TRUE; /* XXX temp */ + ctx->Extensions.ARB_texture_compression = GL_TRUE; /* API support only */ ctx->Extensions.ARB_texture_cube_map = GL_TRUE; ctx->Extensions.ARB_texture_env_combine = GL_TRUE; ctx->Extensions.ARB_texture_env_crossbar = GL_TRUE; ctx->Extensions.ARB_texture_env_dot3 = GL_TRUE; + ctx->Extensions.ARB_texture_mirrored_repeat = GL_TRUE; /* XXX temp */ + ctx->Extensions.ARB_vertex_program = GL_TRUE; ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE; @@ -119,24 +124,27 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.EXT_pixel_buffer_object = GL_TRUE; ctx->Extensions.EXT_point_parameters = GL_TRUE; ctx->Extensions.EXT_secondary_color = GL_TRUE; + ctx->Extensions.EXT_shadow_funcs = GL_TRUE; /* XXX temp */ ctx->Extensions.EXT_stencil_wrap = GL_TRUE; ctx->Extensions.EXT_texture_env_add = GL_TRUE; ctx->Extensions.EXT_texture_env_combine = GL_TRUE; ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE; ctx->Extensions.EXT_texture_lod_bias = GL_TRUE; + ctx->Extensions.EXT_texture_sRGB = GL_TRUE; /* XXX temp */ ctx->Extensions.NV_blend_square = GL_TRUE; ctx->Extensions.NV_texgen_reflection = GL_TRUE; + ctx->Extensions.SGIS_generate_mipmap = GL_TRUE; /* XXX temp */ /* * Extensions that depend on the driver/hardware: */ - if (pipe->get_param(pipe, PIPE_CAP_MAX_RENDER_TARGETS) > 1) { + if (pipe->get_param(pipe, PIPE_CAP_MAX_RENDER_TARGETS) > 0) { ctx->Extensions.ARB_draw_buffers = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_GLSL) > 1) { + if (pipe->get_param(pipe, PIPE_CAP_GLSL)) { ctx->Extensions.ARB_fragment_shader = GL_TRUE; ctx->Extensions.ARB_vertex_shader = GL_TRUE; ctx->Extensions.ARB_shader_objects = GL_TRUE; @@ -153,12 +161,11 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.ARB_multitexture = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_TWO_SIDED_STENCIL) > 1) { + if (pipe->get_param(pipe, PIPE_CAP_TWO_SIDED_STENCIL)) { ctx->Extensions.ATI_separate_stencil = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_S3TC) > 1) { - ctx->Extensions.ARB_texture_compression = GL_TRUE; + if (pipe->get_param(pipe, PIPE_CAP_S3TC)) { ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE; } @@ -170,11 +177,11 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.ARB_point_sprite = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_OCCLUSION_QUERY) > 1) { + if (pipe->get_param(pipe, PIPE_CAP_OCCLUSION_QUERY)) { ctx->Extensions.ARB_occlusion_query = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_TEXTURE_SHADOW_MAP) > 1) { + if (pipe->get_param(pipe, PIPE_CAP_TEXTURE_SHADOW_MAP)) { ctx->Extensions.ARB_depth_texture = GL_TRUE; ctx->Extensions.ARB_shadow = GL_TRUE; /*ctx->Extensions.ARB_shadow_ambient = GL_TRUE;*/ -- cgit v1.2.3 From 601a9ea9a79603763651db8dd93351691594b444 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 08:18:50 -0700 Subject: init glsl functions in st_init_driver_functions() --- src/mesa/state_tracker/st_context.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 3810729847..a4ec3721be 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -29,6 +29,7 @@ #include "main/context.h" #include "main/extensions.h" #include "vbo/vbo.h" +#include "drivers/common/driverfuncs.h" #include "st_public.h" #include "st_context.h" #include "st_cb_accum.h" @@ -180,6 +181,8 @@ void st_copy_context_state(struct st_context *dst, void st_init_driver_functions(struct dd_function_table *functions) { + _mesa_init_glsl_driver_functions(functions); + st_init_accum_functions(functions); st_init_bufferobject_functions(functions); st_init_clear_functions(functions); -- cgit v1.2.3 From c3e7bb4127b74db0037d5a25340793273f5c3183 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 08:21:14 -0700 Subject: fix bad varname --- src/mesa/pipe/xlib/xm_winsys.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index b73fcab68d..d737e984d1 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -297,7 +297,7 @@ xm_surface_alloc(struct pipe_winsys *ws, GLuint pipeFormat) static void -xm_surface_release(struct pipe_winsys *ws, struct pipe_surface **s) +xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) { struct pipe_surface *surf = *s; if (surf->region) -- cgit v1.2.3 From 5d9021554ee78699933ddf79e5231cded6f9c641 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 08:57:56 -0700 Subject: redo xm_surface_release() to match i915 --- src/mesa/pipe/xlib/xm_winsys.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index d737e984d1..506923784c 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -300,9 +300,12 @@ static void xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) { struct pipe_surface *surf = *s; - if (surf->region) - winsys->region_release(winsys, &surf->region); - free(surf); + surf->refcount--; + if (surf->refcount == 0) { + if (surf->region) + winsys->region_release(winsys, &surf->region); + free(surf); + } *s = NULL; } -- cgit v1.2.3 From 182e861eeae479c85191fcd6ccb41bf10f43507b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 09:53:25 -0700 Subject: Disable the call to driBOReference() in i915_update_texture() It doesn't seem to be needed and disabling it fixes a big memory leak with some programs like xdemos/wincopy.c --- src/mesa/pipe/i915simple/i915_state_sampler.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 8dec6781cd..95da30af45 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -234,10 +234,14 @@ i915_update_texture(struct i915_context *i915, uint unit, } #endif - + /* this reference does not seem to be needed. In fact, when it's enabled + * we leak a lot of memory (try xdemos/wincopy). + */ +#if 0 { /*struct pipe_buffer_handle *p =*/ driBOReference(mt->region->buffer); } +#endif #if 0 i915->state.tex_buffer[unit] = driBOReference(intelObj->mt->region-> -- cgit v1.2.3 From 7e884c6f86621ca07ed31fc7ee3f4d891f0873da Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 09:54:02 -0700 Subject: disable/remove dead code --- src/mesa/pipe/i915simple/i915_state_sampler.c | 11 +++-------- 1 file changed, 3 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 95da30af45..65a4d3e199 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -25,13 +25,6 @@ * **************************************************************************/ -//#include "mtypes.h" -//#include "enums.h" -//#include "texformat.h" -//#include "macros.h" -//#include "dri_bufmgr.h" - - #include "pipe/p_context.h" #include "pipe/p_state.h" #include "pipe/p_util.h" @@ -40,7 +33,7 @@ #include "i915_context.h" #include "i915_reg.h" #include "i915_state.h" -//#include "i915_cache.h" + static uint bitcount(uint k) @@ -55,6 +48,7 @@ bitcount(uint k) } +#if 0 static boolean is_power_of_two_texture(const struct pipe_mipmap_tree *mt) { @@ -66,6 +60,7 @@ is_power_of_two_texture(const struct pipe_mipmap_tree *mt) else return 0; } +#endif /** -- cgit v1.2.3 From 187b631b6b3c504fa334e33f4b1af433b6232bac Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 14:40:38 -0700 Subject: Float->uint conversion for PIPE_FORMAT_U_Z32 resulted in overflow in depth_value(). Special-case it. --- src/mesa/state_tracker/st_cb_clear.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 2c79e2890d..cb7e43fb8e 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -86,7 +86,11 @@ depth_value(GLuint pipeFormat, GLfloat value) val = (GLuint) (value * 0xffffff); break; case PIPE_FORMAT_U_Z32: - val = (GLuint) (value * 0xffffffff); + /* special-case to avoid overflow */ + if (value == 1.0) + val = 0xffffffff; + else + val = (GLuint) (value * 0xffffffff); break; case PIPE_FORMAT_S8_Z24: /*case PIPE_FORMAT_Z24_S8:*/ -- cgit v1.2.3 From fc6172bc9abfc24836c6701f3bfcd791907ca9a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 14:41:01 -0700 Subject: better depth buffer selection --- src/mesa/pipe/xlib/xm_api.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index ec889ca34f..e456ea2fd4 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -380,11 +380,21 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, _mesa_add_renderbuffer(fb, BUFFER_STENCIL, rb); } else { - if (vis->mesa_visual.depthBits > 0) { + if (vis->mesa_visual.depthBits > 24) { struct gl_renderbuffer *rb = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32); _mesa_add_renderbuffer(fb, BUFFER_DEPTH, rb); } + else if (vis->mesa_visual.depthBits > 16) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + _mesa_add_renderbuffer(fb, BUFFER_DEPTH, rb); + } + else if (vis->mesa_visual.depthBits > 1) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); + _mesa_add_renderbuffer(fb, BUFFER_DEPTH, rb); + } if (vis->mesa_visual.stencilBits > 0) { struct gl_renderbuffer *rb -- cgit v1.2.3 From 7a9c129a0a171bf0b52f477cfca72591c21ac09a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 14:41:14 -0700 Subject: re-enable PIPE_FORMAT_U_Z32 support --- src/mesa/pipe/xlib/xm_winsys.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 506923784c..cae53c70db 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -373,7 +373,7 @@ xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) case PIPE_FORMAT_S8_Z24: case PIPE_FORMAT_U_S8: case PIPE_FORMAT_U_Z16: - /*case PIPE_FORMAT_U_Z32:*/ + case PIPE_FORMAT_U_Z32: return TRUE; default: return FALSE; -- cgit v1.2.3 From 237429aa3b8c2025e11aaa23d04e7107cf6a838f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 10:34:18 -0700 Subject: more dead code removal --- src/mesa/pipe/i915simple/i915_state_sampler.c | 64 +-------------------------- 1 file changed, 1 insertion(+), 63 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 65a4d3e199..0991e6ac0d 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -35,34 +35,6 @@ #include "i915_state.h" -static uint -bitcount(uint k) -{ - uint count = 0; - while (k) { - if (k & 1) - count++; - k = k >> 1; - } - return count; -} - - -#if 0 -static boolean -is_power_of_two_texture(const struct pipe_mipmap_tree *mt) -{ - if (bitcount(mt->width0) == 1 && - bitcount(mt->height0) == 1 && - bitcount(mt->depth0) == 1) { - return 1; - } - else - return 0; -} -#endif - - /** * Compute i915 texture sampling state. * @@ -124,6 +96,7 @@ static void update_sampler(struct i915_context *i915, state[1] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT); } + void i915_update_samplers( struct i915_context *i915 ) { uint unit; @@ -151,8 +124,6 @@ void i915_update_samplers( struct i915_context *i915 ) } - - static uint translate_texture_format(uint pipeFormat) { @@ -204,10 +175,6 @@ translate_texture_format(uint pipeFormat) } -#define I915_TEXREG_MS3 1 -#define I915_TEXREG_MS4 2 - - static void i915_update_texture(struct i915_context *i915, uint unit, uint state[6]) @@ -222,40 +189,12 @@ i915_update_texture(struct i915_context *i915, uint unit, assert(height); assert(depth); -#if 0 - if (i915->state.tex_buffer[unit] != NULL) { - driBOUnReference(i915->state.tex_buffer[unit]); - i915->state.tex_buffer[unit] = NULL; - } -#endif - - /* this reference does not seem to be needed. In fact, when it's enabled - * we leak a lot of memory (try xdemos/wincopy). - */ -#if 0 - { - /*struct pipe_buffer_handle *p =*/ driBOReference(mt->region->buffer); - } -#endif - -#if 0 - i915->state.tex_buffer[unit] = driBOReference(intelObj->mt->region-> - buffer); - i915->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt, - 0, intelObj-> - firstLevel); -#endif - format = translate_texture_format(mt->format); pitch = mt->pitch * mt->cpp; assert(format); assert(pitch); - /* - printf("texture format = 0x%x\n", format); - */ - /* MS3 state */ state[0] = (((height - 1) << MS3_HEIGHT_SHIFT) @@ -272,7 +211,6 @@ i915_update_texture(struct i915_context *i915, uint unit, } - void i915_update_textures(struct i915_context *i915) { -- cgit v1.2.3 From f40f45ceea7a827059b7b533f2160f98774c8a77 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 10:39:30 -0700 Subject: var renaming: s/sws/winsys/ --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 41 +++++++++++----------- 1 file changed, 20 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 56a600aebe..02c72e9d7f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -56,15 +56,15 @@ struct intel_pipe_winsys { /* Turn a pipe winsys into an intel/pipe winsys: */ static inline struct intel_pipe_winsys * -intel_pipe_winsys( struct pipe_winsys *sws ) +intel_pipe_winsys( struct pipe_winsys *winsys ) { - return (struct intel_pipe_winsys *)sws; + return (struct intel_pipe_winsys *)winsys; } /* Most callbacks map direcly onto dri_bufmgr operations: */ -static void *intel_buffer_map(struct pipe_winsys *sws, +static void *intel_buffer_map(struct pipe_winsys *winsys, struct pipe_buffer_handle *buf, unsigned flags ) { @@ -79,7 +79,7 @@ static void *intel_buffer_map(struct pipe_winsys *sws, return driBOMap( dri_bo(buf), drm_flags, 0 ); } -static void intel_buffer_unmap(struct pipe_winsys *sws, +static void intel_buffer_unmap(struct pipe_winsys *winsys, struct pipe_buffer_handle *buf) { driBOUnmap( dri_bo(buf) ); @@ -87,7 +87,7 @@ static void intel_buffer_unmap(struct pipe_winsys *sws, static void -intel_buffer_reference(struct pipe_winsys *sws, +intel_buffer_reference(struct pipe_winsys *winsys, struct pipe_buffer_handle **ptr, struct pipe_buffer_handle *buf) { @@ -105,19 +105,19 @@ intel_buffer_reference(struct pipe_winsys *sws, /* Grabs the hardware lock! */ -static void intel_buffer_data(struct pipe_winsys *sws, +static void intel_buffer_data(struct pipe_winsys *winsys, struct pipe_buffer_handle *buf, unsigned size, const void *data, unsigned usage ) { - struct intel_context *intel = intel_pipe_winsys(sws)->intel; + struct intel_context *intel = intel_pipe_winsys(winsys)->intel; LOCK_HARDWARE( intel ); driBOData( dri_bo(buf), size, data, 0 ); UNLOCK_HARDWARE( intel ); } -static void intel_buffer_subdata(struct pipe_winsys *sws, +static void intel_buffer_subdata(struct pipe_winsys *winsys, struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, @@ -126,7 +126,7 @@ static void intel_buffer_subdata(struct pipe_winsys *sws, driBOSubData( dri_bo(buf), offset, size, data ); } -static void intel_buffer_get_subdata(struct pipe_winsys *sws, +static void intel_buffer_get_subdata(struct pipe_winsys *winsys, struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, @@ -139,10 +139,10 @@ static void intel_buffer_get_subdata(struct pipe_winsys *sws, * for all buffers. */ static struct pipe_buffer_handle * -intel_buffer_create(struct pipe_winsys *sws, +intel_buffer_create(struct pipe_winsys *winsys, unsigned alignment) { - struct intel_context *intel = intel_pipe_winsys(sws)->intel; + struct intel_context *intel = intel_pipe_winsys(winsys)->intel; struct _DriBufferObject *buffer; LOCK_HARDWARE( intel ); @@ -155,9 +155,9 @@ intel_buffer_create(struct pipe_winsys *sws, static struct pipe_buffer_handle * -intel_user_buffer_create(struct pipe_winsys *sws, void *ptr, unsigned bytes) +intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) { - struct intel_context *intel = intel_pipe_winsys(sws)->intel; + struct intel_context *intel = intel_pipe_winsys(winsys)->intel; struct _DriBufferObject *buffer; LOCK_HARDWARE( intel ); @@ -169,9 +169,9 @@ intel_user_buffer_create(struct pipe_winsys *sws, void *ptr, unsigned bytes) } -static void intel_wait_idle( struct pipe_winsys *sws ) +static void intel_wait_idle( struct pipe_winsys *winsys ) { - struct intel_context *intel = intel_pipe_winsys(sws)->intel; + struct intel_context *intel = intel_pipe_winsys(winsys)->intel; if (intel->batch->last_fence) { driFenceFinish(intel->batch->last_fence, @@ -187,10 +187,10 @@ static void intel_wait_idle( struct pipe_winsys *sws ) * we copied its contents to the real frontbuffer. Our task is easy: */ static void -intel_flush_frontbuffer( struct pipe_winsys *sws, +intel_flush_frontbuffer( struct pipe_winsys *winsys, struct pipe_surface *surf ) { - struct intel_context *intel = intel_pipe_winsys(sws)->intel; + struct intel_context *intel = intel_pipe_winsys(winsys)->intel; __DRIdrawablePrivate *dPriv = intel->driDrawable; intelDisplaySurface(dPriv, surf, NULL); @@ -249,8 +249,7 @@ intel_i915_region_release(struct pipe_winsys *winsys, if ((*region)->refcount == 0) { assert((*region)->map_refcount == 0); - winsys->buffer_reference( winsys, - &((*region)->buffer), NULL ); + winsys->buffer_reference( winsys, &((*region)->buffer), NULL ); free(*region); } *region = NULL; @@ -286,7 +285,7 @@ intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) static void -intel_printf( struct pipe_winsys *sws, const char *fmtString, ... ) +intel_printf( struct pipe_winsys *winsys, const char *fmtString, ... ) { va_list args; va_start( args, fmtString ); @@ -295,7 +294,7 @@ intel_printf( struct pipe_winsys *sws, const char *fmtString, ... ) } static const char * -intel_get_name( struct pipe_winsys *sws ) +intel_get_name( struct pipe_winsys *winsys ) { return "Intel/DRI/ttm"; } -- cgit v1.2.3 From 2611703b363ced1ad267d120ed89677299a36d9c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 11:44:48 -0700 Subject: Pass winsys as a parameter to intel_create_i915simple(), intel_create_softpipe() --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 8 +++++--- src/mesa/drivers/dri/intel_winsys/intel_winsys.h | 6 ++++-- src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c | 11 +++++++---- src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c | 9 ++++++--- 4 files changed, 22 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index f565e2ccf8..31de96dc84 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -144,6 +144,7 @@ intelCreateContext(const __GLcontextModes * visual, int fthrottle_mode; GLboolean havePools; struct pipe_context *pipe; + struct pipe_winsys *winsys; struct st_context *st_share = NULL; if (sharedContextPrivate) { @@ -186,12 +187,13 @@ intelCreateContext(const __GLcontextModes * visual, __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); #endif + winsys = intel_create_pipe_winsys( intel ); /* * Pipe-related setup */ if (!getenv("INTEL_HW")) { - pipe = intel_create_softpipe( intel ); + pipe = intel_create_softpipe( intel, winsys ); } else { switch (intel->intelScreen->deviceID) { @@ -203,13 +205,13 @@ intelCreateContext(const __GLcontextModes * visual, case PCI_CHIP_Q35_G: case PCI_CHIP_I915_G: case PCI_CHIP_I915_GM: - pipe = intel_create_i915simple( intel ); + pipe = intel_create_i915simple( intel, winsys ); break; default: fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", intel->intelScreen->deviceID, __FUNCTION__); - pipe = intel_create_softpipe( intel ); + pipe = intel_create_softpipe( intel, winsys ); break; } } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h index 78f5cb6519..d7de57227b 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h @@ -38,10 +38,12 @@ struct pipe_winsys * intel_create_pipe_winsys( struct intel_context *intel ); struct pipe_context * -intel_create_softpipe( struct intel_context *intel ); +intel_create_softpipe( struct intel_context *intel, + struct pipe_winsys *winsys ); struct pipe_context * -intel_create_i915simple( struct intel_context *intel ); +intel_create_i915simple( struct intel_context *intel, + struct pipe_winsys *winsys ); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index 4fb8f02daa..d6ba6024b2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -120,9 +120,12 @@ static void intel_i915_batch_flush( struct i915_winsys *sws ) } - +/** + * Create i915 hardware rendering context. + */ struct pipe_context * -intel_create_i915simple( struct intel_context *intel ) +intel_create_i915simple( struct intel_context *intel, + struct pipe_winsys *winsys ) { struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); @@ -137,7 +140,7 @@ intel_create_i915simple( struct intel_context *intel ) /* Create the i915simple context: */ - return i915_create( intel_create_pipe_winsys(intel), - &iws->winsys, + return i915_create( winsys, + &iws->winsys, intel->intelScreen->deviceID ); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c index 5407973fba..7a93546bc2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c @@ -58,8 +58,12 @@ intel_is_format_supported(struct softpipe_winsys *sws, uint format) } +/** + * Create rendering context which uses software rendering. + */ struct pipe_context * -intel_create_softpipe( struct intel_context *intel ) +intel_create_softpipe( struct intel_context *intel, + struct pipe_winsys *winsys ) { struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); @@ -71,6 +75,5 @@ intel_create_softpipe( struct intel_context *intel ) /* Create the softpipe context: */ - return softpipe_create( intel_create_pipe_winsys( intel ), - &isws->sws ); + return softpipe_create( winsys, &isws->sws ); } -- cgit v1.2.3 From c712d092a178096d8d95e58bd1bd20a33045cd7b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 11:59:04 -0700 Subject: remove DriverCtx=intel assigment/hack --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 31de96dc84..88551bc3a3 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -35,7 +35,6 @@ #include "intel_batchbuffer.h" #include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" #include "pipe/p_defines.h" #include "pipe/p_context.h" @@ -217,7 +216,6 @@ intelCreateContext(const __GLcontextModes * visual, } intel->st = st_create_context(pipe, visual, st_share); - intel->st->ctx->DriverCtx = intel; /* hope to get rid of this... */ return GL_TRUE; } -- cgit v1.2.3 From 0852cf8611dd77b564db1df9b61e035d9ea4b280 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 12:06:50 -0700 Subject: just reorder a few members of pipe_winsys, update comments --- src/mesa/pipe/p_winsys.h | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index e4c888ad33..9dbac87b4b 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -60,6 +60,12 @@ struct pipe_buffer_handle; struct pipe_winsys { + /** Returns name of this winsys interface */ + const char *(*get_name)( struct pipe_winsys *sws ); + + /** Wait for any buffered rendering to finish */ + void (*wait_idle)( struct pipe_winsys *sws ); + /** * Do any special operations to ensure frontbuffer contents are * displayed, eg copy fake frontbuffer. @@ -147,13 +153,6 @@ struct pipe_winsys unsigned long size, void *data); - - /** Wait for any hw swapbuffers, etc. to finish */ - void (*wait_idle)( struct pipe_winsys *sws ); - - /** Queries */ - const char *(*get_name)( struct pipe_winsys *sws ); - }; -- cgit v1.2.3 From 51653cd551aeed9b1e5b3edd17172c8906d989e5 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 15:22:07 -0700 Subject: replace magic number 3 with DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW in driFenceFinish() call --- src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 89d72c9448..784ed2210a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -151,7 +151,8 @@ intel_batchbuffer_free(struct intel_batchbuffer *batch) { if (batch->last_fence) { driFenceFinish(batch->last_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, + GL_FALSE); driFenceUnReference(batch->last_fence); batch->last_fence = NULL; } @@ -311,7 +312,9 @@ intel_batchbuffer_finish(struct intel_batchbuffer *batch) { struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); driFenceReference(fence); - driFenceFinish(fence, 3, GL_FALSE); + driFenceFinish(fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, + GL_FALSE); driFenceUnReference(fence); } -- cgit v1.2.3 From ed6e591270a583956c6407ae6536c3110a86397a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 15:22:31 -0700 Subject: cosmetic fix to driFenceFinish() call --- src/mesa/drivers/dri/intel_winsys/intel_batchpool.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c index a81a463baa..33b56817f6 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c @@ -288,7 +288,7 @@ static int pool_waitIdle(struct _DriBufferPool *pool, void *private, int lazy) { BBuf *buf = (BBuf *) private; - driFenceFinish(buf->fence, 0, lazy); + driFenceFinish(buf->fence, 0x0, lazy); return 0; } -- cgit v1.2.3 From 03cfeb31af7834c2b2701ad25ec39f8375df6c96 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 15:57:19 -0700 Subject: enable GL_ARB/NV_point_sprite together --- src/mesa/state_tracker/st_extensions.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 4355eaa5f9..fc003d4776 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -175,6 +175,7 @@ void st_init_extensions(struct st_context *st) if (pipe->get_param(pipe, PIPE_CAP_POINT_SPRITE)) { ctx->Extensions.ARB_point_sprite = GL_TRUE; + ctx->Extensions.NV_point_sprite = GL_TRUE; } if (pipe->get_param(pipe, PIPE_CAP_OCCLUSION_QUERY)) { -- cgit v1.2.3 From 10c62bf0683437672c83339138a6802d56aeca8f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 16:07:17 -0700 Subject: Remove context dependencies in winsys layer. The winsys object is now per-screen and shared by multiple contexts. The regionPool is now part of the i915 winsys layer. The winsys wait_idle() and flush_frontbuffer() funcs will get more attention... --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 11 +++-- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 12 ++--- src/mesa/drivers/dri/intel_winsys/intel_screen.h | 3 +- src/mesa/drivers/dri/intel_winsys/intel_winsys.h | 5 ++- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 52 +++++++++++----------- src/mesa/pipe/p_context.h | 2 + src/mesa/pipe/p_winsys.h | 5 ++- src/mesa/pipe/xlib/xm_winsys.c | 2 +- src/mesa/state_tracker/st_cb_flush.c | 5 ++- 9 files changed, 51 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 88551bc3a3..47be72b233 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -143,7 +143,6 @@ intelCreateContext(const __GLcontextModes * visual, int fthrottle_mode; GLboolean havePools; struct pipe_context *pipe; - struct pipe_winsys *winsys; struct st_context *st_share = NULL; if (sharedContextPrivate) { @@ -186,13 +185,11 @@ intelCreateContext(const __GLcontextModes * visual, __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); #endif - winsys = intel_create_pipe_winsys( intel ); - /* * Pipe-related setup */ if (!getenv("INTEL_HW")) { - pipe = intel_create_softpipe( intel, winsys ); + pipe = intel_create_softpipe( intel, intelScreen->winsys ); } else { switch (intel->intelScreen->deviceID) { @@ -204,17 +201,19 @@ intelCreateContext(const __GLcontextModes * visual, case PCI_CHIP_Q35_G: case PCI_CHIP_I915_G: case PCI_CHIP_I915_GM: - pipe = intel_create_i915simple( intel, winsys ); + pipe = intel_create_i915simple( intel, intelScreen->winsys ); break; default: fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", intel->intelScreen->deviceID, __FUNCTION__); - pipe = intel_create_softpipe( intel, winsys ); + pipe = intel_create_softpipe( intel, intelScreen->winsys ); break; } } + pipe->private = intel; + intel->st = st_create_context(pipe, visual, st_share); return GL_TRUE; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 334803e264..01460e5be3 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -34,6 +34,7 @@ #include "intel_batchbuffer.h" #include "intel_batchpool.h" #include "intel_swapbuffers.h" +#include "intel_winsys.h" #include "i830_dri.h" #include "dri_bufpool.h" @@ -165,17 +166,11 @@ intelCreatePools(__DRIscreenPrivate * sPriv) if (intelScreen->havePools) return GL_TRUE; - batchPoolSize /= BATCH_SZ; - intelScreen->regionPool = driDRMPoolInit(sPriv->fd); - - if (!intelScreen->regionPool) - return GL_FALSE; - intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); - if (!intelScreen->staticPool) return GL_FALSE; + batchPoolSize /= BATCH_SZ; intelScreen->batchPool = driBatchPoolInit(sPriv->fd, DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT | @@ -245,6 +240,8 @@ intelInitDriver(__DRIscreenPrivate * sPriv) (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); } + intelScreen->winsys = intel_create_pipe_winsys(sPriv->fd); + return GL_TRUE; } @@ -257,7 +254,6 @@ intelDestroyScreen(__DRIscreenPrivate * sPriv) /* intelUnmapScreenRegions(intelScreen); */ if (intelScreen->havePools) { - driPoolTakeDown(intelScreen->regionPool); driPoolTakeDown(intelScreen->staticPool); driPoolTakeDown(intelScreen->batchPool); } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h index f0446fe7b3..3396f9e564 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.h @@ -65,7 +65,6 @@ struct intel_screen driOptionCache optionCache; struct _DriBufferPool *batchPool; - struct _DriBufferPool *regionPool; struct _DriBufferPool *staticPool; /** for the X screen/framebuffer */ boolean havePools; @@ -74,6 +73,8 @@ struct intel_screen * which we need a rendering context, but none is currently bound. */ struct intel_context *dummyContext; + + struct pipe_winsys *winsys; }; diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h index d7de57227b..89e63e0a79 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h @@ -35,7 +35,10 @@ struct pipe_buffer_handle; struct _DriBufferObject; struct pipe_winsys * -intel_create_pipe_winsys( struct intel_context *intel ); +intel_create_pipe_winsys( int fd ); + +void +intel_destroy_pipe_winsys( struct pipe_winsys *winsys ); struct pipe_context * intel_create_softpipe( struct intel_context *intel, diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 02c72e9d7f..16fff77af4 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -48,7 +48,7 @@ struct intel_pipe_winsys { struct pipe_winsys winsys; - struct intel_context *intel; + struct _DriBufferPool *regionPool; }; @@ -110,11 +110,7 @@ static void intel_buffer_data(struct pipe_winsys *winsys, unsigned size, const void *data, unsigned usage ) { - struct intel_context *intel = intel_pipe_winsys(winsys)->intel; - - LOCK_HARDWARE( intel ); driBOData( dri_bo(buf), size, data, 0 ); - UNLOCK_HARDWARE( intel ); } static void intel_buffer_subdata(struct pipe_winsys *winsys, @@ -142,14 +138,10 @@ static struct pipe_buffer_handle * intel_buffer_create(struct pipe_winsys *winsys, unsigned alignment) { - struct intel_context *intel = intel_pipe_winsys(winsys)->intel; struct _DriBufferObject *buffer; - - LOCK_HARDWARE( intel ); - driGenBuffers( intel->intelScreen->regionPool, + struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + driGenBuffers( iws->regionPool, "pipe buffer", 1, &buffer, alignment, 0, 0 ); - UNLOCK_HARDWARE( intel ); - return pipe_bo(buffer); } @@ -157,21 +149,18 @@ intel_buffer_create(struct pipe_winsys *winsys, static struct pipe_buffer_handle * intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) { - struct intel_context *intel = intel_pipe_winsys(winsys)->intel; struct _DriBufferObject *buffer; - - LOCK_HARDWARE( intel ); - driGenUserBuffer( intel->intelScreen->regionPool, + struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + driGenUserBuffer( iws->regionPool, "pipe user buffer", &buffer, ptr, bytes); - UNLOCK_HARDWARE( intel ); - return pipe_bo(buffer); } -static void intel_wait_idle( struct pipe_winsys *winsys ) +static void +intel_wait_idle( struct pipe_winsys *winsys, void *context_private ) { - struct intel_context *intel = intel_pipe_winsys(winsys)->intel; + struct intel_context *intel = (struct intel_context *) context_private; if (intel->batch->last_fence) { driFenceFinish(intel->batch->last_fence, @@ -188,9 +177,10 @@ static void intel_wait_idle( struct pipe_winsys *winsys ) */ static void intel_flush_frontbuffer( struct pipe_winsys *winsys, - struct pipe_surface *surf ) + struct pipe_surface *surf, + void *context_private) { - struct intel_context *intel = intel_pipe_winsys(winsys)->intel; + struct intel_context *intel = (struct intel_context *) context_private; __DRIdrawablePrivate *dPriv = intel->driDrawable; intelDisplaySurface(dPriv, surf, NULL); @@ -301,7 +291,7 @@ intel_get_name( struct pipe_winsys *winsys ) struct pipe_winsys * -intel_create_pipe_winsys( struct intel_context *intel ) +intel_create_pipe_winsys( int fd ) { struct intel_pipe_winsys *iws = CALLOC_STRUCT( intel_pipe_winsys ); @@ -324,13 +314,25 @@ intel_create_pipe_winsys( struct intel_context *intel ) iws->winsys.wait_idle = intel_wait_idle; iws->winsys.printf = intel_printf; iws->winsys.get_name = intel_get_name; - iws->intel = intel; - iws->winsys.region_alloc = intel_i915_region_alloc; iws->winsys.region_release = intel_i915_region_release; - iws->winsys.surface_alloc = intel_i915_surface_alloc; iws->winsys.surface_release = intel_i915_surface_release; + if (fd) + iws->regionPool = driDRMPoolInit(fd); + return &iws->winsys; } + + +void +intel_destroy_pipe_winsys( struct pipe_winsys *winsys ) +{ + struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + if (iws->regionPool) { + driPoolTakeDown(iws->regionPool); + } + free(iws); +} + diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index ddc7acc747..3962c0fa6d 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -43,6 +43,8 @@ struct pipe_state_cache; struct pipe_context { struct pipe_winsys *winsys; + void *private; /** context private data (for DRI for example) */ + void (*destroy)( struct pipe_context * ); /* diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 9dbac87b4b..ee10e30559 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -64,14 +64,15 @@ struct pipe_winsys const char *(*get_name)( struct pipe_winsys *sws ); /** Wait for any buffered rendering to finish */ - void (*wait_idle)( struct pipe_winsys *sws ); + void (*wait_idle)( struct pipe_winsys *sws, void *context_private ); /** * Do any special operations to ensure frontbuffer contents are * displayed, eg copy fake frontbuffer. */ void (*flush_frontbuffer)( struct pipe_winsys *sws, - struct pipe_surface *surf ); + struct pipe_surface *surf, + void *context_private ); /** Debug output */ void (*printf)( struct pipe_winsys *sws, diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index cae53c70db..ea6b06a93c 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -168,7 +168,7 @@ xm_flush_frontbuffer(struct pipe_winsys *pws, } static void -xm_wait_idle(struct pipe_winsys *pws) +xm_wait_idle(struct pipe_winsys *pws, void *context_private) { /* no-op */ } diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 819957a1ee..6354306e75 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -71,7 +71,8 @@ void st_flush( struct st_context *st ) /* Hook for copying "fake" frontbuffer if necessary: */ - st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf ); + st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf, + st->pipe->private ); st->flags.frontbuffer_dirty = 0; } } @@ -94,7 +95,7 @@ static void st_Finish(GLcontext *ctx) struct st_context *st = ctx->st; st_flush( st ); - st->pipe->winsys->wait_idle( st->pipe->winsys ); + st->pipe->winsys->wait_idle( st->pipe->winsys, st->pipe->private ); } -- cgit v1.2.3 From ae44a81d1bd40852a7cea9b8025dfa3821adc785 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 16:59:37 -0700 Subject: New PIPE_FLUSH_WAIT flag for pipe->flush(). The state tracker doesn't have to directly call winsys->wait_idle() anymore. glFlush and glFinish both go through pipe->flush() now. --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 4 ++-- src/mesa/pipe/i915simple/i915_flush.c | 6 +++++- src/mesa/pipe/p_defines.h | 3 ++- src/mesa/state_tracker/st_cb_flush.c | 11 ++++------- src/mesa/state_tracker/st_framebuffer.c | 2 +- src/mesa/state_tracker/st_public.h | 2 +- 6 files changed, 15 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 47be72b233..480350492f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -227,7 +227,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) assert(intel); /* should never be null */ if (intel) { - st_flush(intel->st); + st_flush(intel->st, PIPE_FLUSH_WAIT); intel_batchbuffer_free(intel->batch); @@ -255,7 +255,7 @@ GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv) { struct intel_context *intel = intel_context(driContextPriv); - st_flush(intel->st); + st_flush(intel->st, 0x0); /* XXX make_current(NULL)? */ return GL_TRUE; } diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c index 9c2adf8763..5a80ed5e2f 100644 --- a/src/mesa/pipe/i915simple/i915_flush.c +++ b/src/mesa/pipe/i915simple/i915_flush.c @@ -47,7 +47,7 @@ static void i915_flush( struct pipe_context *pipe, /* Do we need to emit an MI_FLUSH command to flush the hardware * caches? */ - if (flags) { + if (flags & (PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE)) { unsigned flush = MI_FLUSH; if (!(flags & PIPE_FLUSH_RENDER_CACHE)) @@ -67,6 +67,10 @@ static void i915_flush( struct pipe_context *pipe, /* If there are no flags, just flush pending commands to hardware: */ FLUSH_BATCH(); + + if (flags & PIPE_FLUSH_WAIT) { + i915->pipe.winsys->wait_idle(i915->pipe.winsys, i915->pipe.private); + } } diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index ca9929bfee..6b5881b64d 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -187,8 +187,9 @@ /** * Flush types: */ -#define PIPE_FLUSH_RENDER_CACHE 0x1 +#define PIPE_FLUSH_RENDER_CACHE 0x1 #define PIPE_FLUSH_TEXTURE_CACHE 0x2 +#define PIPE_FLUSH_WAIT 0x4 /** diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 6354306e75..39a9f29bca 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -43,7 +43,7 @@ #include "pipe/p_winsys.h" -void st_flush( struct st_context *st ) +void st_flush( struct st_context *st, uint pipeFlushFlags ) { GLframebuffer *fb = st->ctx->DrawBuffer; @@ -53,7 +53,7 @@ void st_flush( struct st_context *st ) * short-circuiting this, or perhaps pass an "optional" flag down * to the driver so that it can make the decision. */ - st->pipe->flush( st->pipe, 0 ); + st->pipe->flush( st->pipe, pipeFlushFlags ); if (!fb) return; @@ -83,7 +83,7 @@ void st_flush( struct st_context *st ) */ static void st_Flush(GLcontext *ctx) { - st_flush(ctx->st); + st_flush(ctx->st, 0x0); } @@ -92,10 +92,7 @@ static void st_Flush(GLcontext *ctx) */ static void st_Finish(GLcontext *ctx) { - struct st_context *st = ctx->st; - - st_flush( st ); - st->pipe->winsys->wait_idle( st->pipe->winsys, st->pipe->private ); + st_flush(ctx->st, PIPE_FLUSH_WAIT); } diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index b04dcdb79d..4ae2837f0a 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -179,7 +179,7 @@ st_notify_swapbuffers(struct st_framebuffer *stfb) GET_CURRENT_CONTEXT(ctx); if (ctx && ctx->DrawBuffer == &stfb->Base) { - st_flush(ctx->st); + st_flush(ctx->st, 0x0); } } diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 9e36e1e6e5..408e1927e5 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -70,7 +70,7 @@ void st_make_current(struct st_context *st, struct st_framebuffer *draw, struct st_framebuffer *read); -void st_flush( struct st_context *st ); +void st_flush( struct st_context *st, uint pipeFlushFlags ); void st_notify_swapbuffers(struct st_framebuffer *stfb); -- cgit v1.2.3 From 3fe055ec9218a8745d9a53f75f4abd4760fcce28 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 17:16:47 -0700 Subject: Remove winsys->wait_idle(), subsumed by pipe->flush(). Connect intel_i915_batch_finish() into i915_winsys, just like intel_i915_batch_flush(). Call i915_winsys->batch_finish() in response to pipe->flush(PIPE_FLUSH_WAIT). Now all the batchbuffer/fence code is in one place and a little cleaner. --- src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c | 8 ++++++++ src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c | 15 --------------- src/mesa/pipe/i915simple/i915_flush.c | 2 +- src/mesa/pipe/i915simple/i915_winsys.h | 1 + src/mesa/pipe/p_winsys.h | 3 --- src/mesa/pipe/xlib/xm_winsys.c | 6 ------ 6 files changed, 10 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index d6ba6024b2..8e0eea4392 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -120,6 +120,13 @@ static void intel_i915_batch_flush( struct i915_winsys *sws ) } +static void intel_i915_batch_finish( struct i915_winsys *sws ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + intel_batchbuffer_finish( intel->batch ); +} + + /** * Create i915 hardware rendering context. */ @@ -136,6 +143,7 @@ intel_create_i915simple( struct intel_context *intel, iws->winsys.batch_dword = intel_i915_batch_dword; iws->winsys.batch_reloc = intel_i915_batch_reloc; iws->winsys.batch_flush = intel_i915_batch_flush; + iws->winsys.batch_finish = intel_i915_batch_finish; iws->intel = intel; /* Create the i915simple context: diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 16fff77af4..ae02f98a78 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -157,20 +157,6 @@ intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) } -static void -intel_wait_idle( struct pipe_winsys *winsys, void *context_private ) -{ - struct intel_context *intel = (struct intel_context *) context_private; - - if (intel->batch->last_fence) { - driFenceFinish(intel->batch->last_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); - driFenceUnReference(intel->batch->last_fence); - intel->batch->last_fence = NULL; - } -} - - /* The state tracker (should!) keep track of whether the fake * frontbuffer has been touched by any rendering since the last time * we copied its contents to the real frontbuffer. Our task is easy: @@ -311,7 +297,6 @@ intel_create_pipe_winsys( int fd ) iws->winsys.buffer_subdata = intel_buffer_subdata; iws->winsys.buffer_get_subdata = intel_buffer_get_subdata; iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; - iws->winsys.wait_idle = intel_wait_idle; iws->winsys.printf = intel_printf; iws->winsys.get_name = intel_get_name; iws->winsys.region_alloc = intel_i915_region_alloc; diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c index 5a80ed5e2f..51cf9960aa 100644 --- a/src/mesa/pipe/i915simple/i915_flush.c +++ b/src/mesa/pipe/i915simple/i915_flush.c @@ -69,7 +69,7 @@ static void i915_flush( struct pipe_context *pipe, FLUSH_BATCH(); if (flags & PIPE_FLUSH_WAIT) { - i915->pipe.winsys->wait_idle(i915->pipe.winsys, i915->pipe.private); + i915->winsys->batch_finish(i915->winsys); } } diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index cbb851a932..2c0f335d34 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -98,6 +98,7 @@ struct i915_winsys { unsigned delta ); void (*batch_flush)( struct i915_winsys *sws ); + void (*batch_finish)( struct i915_winsys *sws ); }; #define I915_BUFFER_ACCESS_WRITE 0x1 diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index ee10e30559..41d522e11e 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -63,9 +63,6 @@ struct pipe_winsys /** Returns name of this winsys interface */ const char *(*get_name)( struct pipe_winsys *sws ); - /** Wait for any buffered rendering to finish */ - void (*wait_idle)( struct pipe_winsys *sws, void *context_private ); - /** * Do any special operations to ensure frontbuffer contents are * displayed, eg copy fake frontbuffer. diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index ea6b06a93c..df90a807d4 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -167,11 +167,6 @@ xm_flush_frontbuffer(struct pipe_winsys *pws, */ } -static void -xm_wait_idle(struct pipe_winsys *pws, void *context_private) -{ - /* no-op */ -} static void xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) @@ -343,7 +338,6 @@ xmesa_get_pipe_winsys(void) ws->surface_release = xm_surface_release; ws->flush_frontbuffer = xm_flush_frontbuffer; - ws->wait_idle = xm_wait_idle; ws->printf = xm_printf; ws->get_name = xm_get_name; } -- cgit v1.2.3 From 74df1f4adbbc4fe6027265625cd248c3add154f5 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 17:17:32 -0700 Subject: update flush/fence comments for i915_flush() --- src/mesa/pipe/i915simple/i915_flush.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c index 51cf9960aa..3c2069b827 100644 --- a/src/mesa/pipe/i915simple/i915_flush.c +++ b/src/mesa/pipe/i915simple/i915_flush.c @@ -35,9 +35,9 @@ #include "i915_reg.h" #include "i915_batch.h" -/* There will be actual work to do here. In future we may want a - * fence-like interface instead of finish, and perhaps flush will take - * flags to indicate what type of flush is required. + +/** + * In future we may want a fence-like interface instead of finish. */ static void i915_flush( struct pipe_context *pipe, unsigned flags ) -- cgit v1.2.3 From 6400658d7154f33b8fe2e7d638ebb317ac6bfff8 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 7 Nov 2007 17:29:14 -0700 Subject: fix warning --- src/mesa/pipe/xlib/xm_winsys.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index df90a807d4..b842cf766e 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -158,7 +158,8 @@ xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, static void xm_flush_frontbuffer(struct pipe_winsys *pws, - struct pipe_surface *surf ) + struct pipe_surface *surf, + void *context_private) { /* The Xlib driver's front color surfaces are actually X Windows so * this flush is a no-op. -- cgit v1.2.3 From 3871d2882a5e3ae8aafe3206603e095cc90761b5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 8 Nov 2007 13:10:18 +0100 Subject: Rename struct field 'private' to 'priv'. This broke the LLVM build because 'private' is a C++ keyword. --- src/mesa/drivers/dri/intel_winsys/intel_context.c | 2 +- src/mesa/pipe/p_context.h | 2 +- src/mesa/state_tracker/st_cb_flush.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c index 480350492f..c033f2a592 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.c @@ -212,7 +212,7 @@ intelCreateContext(const __GLcontextModes * visual, } } - pipe->private = intel; + pipe->priv = intel; intel->st = st_create_context(pipe, visual, st_share); diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 3962c0fa6d..48356bd0e6 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -43,7 +43,7 @@ struct pipe_state_cache; struct pipe_context { struct pipe_winsys *winsys; - void *private; /** context private data (for DRI for example) */ + void *priv; /** context private data (for DRI for example) */ void (*destroy)( struct pipe_context * ); diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 39a9f29bca..95149a3200 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -72,7 +72,7 @@ void st_flush( struct st_context *st, uint pipeFlushFlags ) /* Hook for copying "fake" frontbuffer if necessary: */ st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf, - st->pipe->private ); + st->pipe->priv ); st->flags.frontbuffer_dirty = 0; } } -- cgit v1.2.3 From 95128c1d4c88238a79ead6e36215a646f83bbdd3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 07:58:53 -0700 Subject: check for fence==NULL in intel_batchbuffer_finish(), fixes glxinfo crash on exit --- src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 784ed2210a..49e04d81ec 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -311,11 +311,13 @@ void intel_batchbuffer_finish(struct intel_batchbuffer *batch) { struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); - driFenceReference(fence); - driFenceFinish(fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, - GL_FALSE); - driFenceUnReference(fence); + if (fence) { + driFenceReference(fence); + driFenceFinish(fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, + GL_FALSE); + driFenceUnReference(fence); + } } -- cgit v1.2.3 From 44254b92480115e5c8a2d5cf78f99195c03701eb Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 08:22:42 -0700 Subject: Added pipe->get_paramf() to query float limits. So far max point size, line width, texture anistopy and lod bias. --- src/mesa/pipe/failover/fo_context.c | 1 + src/mesa/pipe/i915simple/i915_context.c | 27 +++++++++++++++++++++++++++ src/mesa/pipe/p_context.h | 1 + src/mesa/pipe/p_defines.h | 7 +++++++ src/mesa/pipe/softpipe/sp_context.c | 26 +++++++++++++++++++++++++- src/mesa/state_tracker/st_extensions.c | 21 +++++++++++++++++++++ 6 files changed, 82 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index e18b349ef3..c5fab73fb1 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -120,6 +120,7 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.get_name = hw->get_name; failover->pipe.get_vendor = hw->get_vendor; failover->pipe.get_param = hw->get_param; + failover->pipe.get_paramf = hw->get_paramf; failover->pipe.draw_arrays = failover_draw_arrays; failover->pipe.draw_elements = failover_draw_elements; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 50503a843a..11be13705f 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -144,6 +144,32 @@ i915_get_param(struct pipe_context *pipe, int param) } +static float +i915_get_paramf(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 7.5; + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 0.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; + + default: + return 0; + } +} + + static void i915_destroy( struct pipe_context *pipe ) { struct i915_context *i915 = i915_context( pipe ); @@ -301,6 +327,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.destroy = i915_destroy; i915->pipe.is_format_supported = i915_is_format_supported; i915->pipe.get_param = i915_get_param; + i915->pipe.get_paramf = i915_get_paramf; i915->pipe.clear = i915_clear; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 48356bd0e6..8bed958feb 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -58,6 +58,7 @@ struct pipe_context { const char *(*get_vendor)( struct pipe_context *pipe ); int (*get_param)( struct pipe_context *pipe, int param ); + float (*get_paramf)( struct pipe_context *pipe, int param ); /* diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 6b5881b64d..a853605486 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -250,5 +250,12 @@ #define PIPE_CAP_MAX_TEXTURE_2D_LEVELS 11 #define PIPE_CAP_MAX_TEXTURE_3D_LEVELS 12 #define PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS 13 +#define PIPE_CAP_MAX_LINE_WIDTH 14 +#define PIPE_CAP_MAX_LINE_WIDTH_AA 15 +#define PIPE_CAP_MAX_POINT_WIDTH 16 +#define PIPE_CAP_MAX_POINT_WIDTH_AA 17 +#define PIPE_CAP_MAX_TEXTURE_ANISOTROPY 18 +#define PIPE_CAP_MAX_TEXTURE_LOD_BIAS 19 + #endif diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 09c5a7152d..be4da0ec64 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -251,6 +251,30 @@ static int softpipe_get_param(struct pipe_context *pipe, int param) } } +static float softpipe_get_paramf(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 0.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; /* arbitrary */ + + default: + return 0; + } +} + struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, struct softpipe_winsys *softpipe_winsys ) { @@ -270,8 +294,8 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, /* queries */ softpipe->pipe.is_format_supported = softpipe_is_format_supported; - //softpipe->pipe.max_texture_size = softpipe_max_texture_size; softpipe->pipe.get_param = softpipe_get_param; + softpipe->pipe.get_paramf = softpipe_get_paramf; /* state setters */ softpipe->pipe.create_alpha_test_state = softpipe_create_alpha_test_state; diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index fc003d4776..97578e5002 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -42,6 +42,11 @@ static int min(int a, int b) return (a < b) ? a : b; } +static int max(int a, int b) +{ + return (a > b) ? a : b; +} + static int clamp(int a, int min, int max) { if (a < min) @@ -85,6 +90,22 @@ void st_init_limits(struct st_context *st) c->MaxDrawBuffers = clamp(pipe->get_param(pipe, PIPE_CAP_MAX_RENDER_TARGETS), 1, MAX_DRAW_BUFFERS); + + c->MaxLineWidth + = max(1.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_LINE_WIDTH)); + c->MaxLineWidthAA + = max(1.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_LINE_WIDTH_AA)); + + c->MaxPointSize + = max(1.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_POINT_WIDTH)); + c->MaxPointSizeAA + = max(1.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_POINT_WIDTH_AA)); + + c->MaxTextureMaxAnisotropy + = max(2.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_TEXTURE_ANISOTROPY)); + + c->MaxTextureLodBias + = pipe->get_paramf(pipe, PIPE_CAP_MAX_TEXTURE_LOD_BIAS); } -- cgit v1.2.3 From e2593bb1632cee65af9ef5fe1f9b67928caa25a0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 08:52:36 -0700 Subject: remove unneeded #include --- src/mesa/pipe/i915simple/i915_context.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 11be13705f..9f11fe5385 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -25,7 +25,6 @@ * **************************************************************************/ -//#include "main/imports.h" /* CALLOC */ #include "i915_context.h" #include "i915_winsys.h" #include "i915_state.h" -- cgit v1.2.3 From 98c539d337fec8b4e21d9788b52ff551462e90d6 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 09:14:43 -0700 Subject: in check_end_texture_render(), test for presence of a renderbuffer and texobj --- src/mesa/main/fbobject.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 6f7effcce7..963e35d678 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -913,8 +913,7 @@ check_end_texture_render(GLcontext *ctx, struct gl_framebuffer *fb) GLuint i; for (i = 0; i < BUFFER_COUNT; i++) { struct gl_renderbuffer_attachment *att = fb->Attachment + i; - struct gl_texture_object *texObj = att->Texture; - if (texObj) { + if (att->Texture && att->Renderbuffer) { ctx->Driver.FinishRenderTexture(ctx, att); } } -- cgit v1.2.3 From 2c1184626879cb349bcdc87eebb82da87f1a9f85 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 09:16:14 -0700 Subject: Fix some issues hit when rendering to a depth-only renderbuffer (shadowtex). Added 16bpp case to translate_depth_format(). Added framebuffer_size() to determine framebuffer size. --- src/mesa/pipe/i915simple/i915_state_emit.c | 55 ++++++++++++++++++++++++++---- 1 file changed, 48 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index ee44e526ca..d793e92a14 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -50,8 +50,44 @@ static unsigned translate_format( unsigned format ) static unsigned translate_depth_format( unsigned zformat ) { - assert(zformat == PIPE_FORMAT_S8_Z24); - return DEPTH_FRMT_24_FIXED_8_OTHER; + switch (zformat) { + case PIPE_FORMAT_S8_Z24: + return DEPTH_FRMT_24_FIXED_8_OTHER; + case PIPE_FORMAT_U_Z16: + return DEPTH_FRMT_16_FIXED; + default: + assert(0); + return 0; + } +} + + +/** + * Examine framebuffer state to determine width, height. + */ +static boolean +framebuffer_size(const struct pipe_framebuffer_state *fb, + uint *width, uint *height) +{ + if (fb->cbufs[0]) { + *width = fb->cbufs[0]->width; + *height = fb->cbufs[0]->height; + return TRUE; + } + else if (fb->zbuf) { + *width = fb->zbuf->width; + *height = fb->zbuf->height; + return TRUE; + } + else if (fb->sbuf) { + *width = fb->sbuf->width; + *height = fb->sbuf->height; + return TRUE; + } + else { + *width = *height = 0; + return FALSE; + } } @@ -207,9 +243,14 @@ i915_emit_hardware_state(struct i915_context *i915 ) } { - unsigned cformat = translate_format( i915->framebuffer.cbufs[0]->format ); - unsigned zformat = 0; + unsigned cformat, zformat = 0; + if (i915->framebuffer.cbufs[0]) + cformat = i915->framebuffer.cbufs[0]->format; + else + cformat = PIPE_FORMAT_U_A8_R8_G8_B8; /* arbitrary */ + cformat = translate_format(cformat); + if (i915->framebuffer.zbuf) zformat = translate_depth_format( i915->framebuffer.zbuf->format ); @@ -322,8 +363,9 @@ i915_emit_hardware_state(struct i915_context *i915 ) /* drawing surface size */ /* 6 dwords, 0 relocs */ { - int w = i915->framebuffer.cbufs[0]->width; - int h = i915->framebuffer.cbufs[0]->height; + uint w, h; + boolean k = framebuffer_size(&i915->framebuffer, &w, &h); + assert(k); OUT_BATCH(_3DSTATE_DRAW_RECT_CMD); OUT_BATCH(0); @@ -336,4 +378,3 @@ i915_emit_hardware_state(struct i915_context *i915 ) i915->hardware_dirty = 0; } - -- cgit v1.2.3 From b18763141d651ae1112e80bb7ef3e813c8ce1d18 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 09:31:57 -0700 Subject: Replace gl_stencil_func_to_sp() and gl_depth_func_to_sp() with st_compare_func_to_pipe() --- src/mesa/state_tracker/st_atom.h | 2 ++ src/mesa/state_tracker/st_atom_depth.c | 49 ++++++++++------------------------ 2 files changed, 16 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 63ce35958a..3c0db0db09 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -62,4 +62,6 @@ const struct st_tracked_state st_update_alpha_test; const struct st_tracked_state st_update_pixel_transfer; +uint st_compare_func_to_pipe(GLenum func); + #endif diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index caf51f17ac..f0f360e6dd 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -41,10 +41,10 @@ /** - * Convert GLenum stencil func tokens to pipe tokens. + * Convert an OpenGL compare mode to a pipe tokens. */ -static GLuint -gl_stencil_func_to_sp(GLenum func) +GLuint +st_compare_func_to_pipe(GLenum func) { /* Same values, just biased */ assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER); @@ -65,7 +65,7 @@ gl_stencil_func_to_sp(GLenum func) * Convert GLenum stencil op tokens to pipe tokens. */ static GLuint -gl_stencil_op_to_sp(GLenum func) +gl_stencil_op_to_pipe(GLenum func) { switch (func) { case GL_KEEP: @@ -85,32 +85,11 @@ gl_stencil_op_to_sp(GLenum func) case GL_INVERT: return PIPE_STENCIL_OP_INVERT; default: - assert("invalid GL token in gl_stencil_op_to_sp()" == NULL); + assert("invalid GL token in gl_stencil_op_to_pipe()" == NULL); return 0; } } -/** - * Convert GLenum depth func tokens to pipe tokens. - */ -static GLuint -gl_depth_func_to_sp(GLenum func) -{ - /* Same values, just biased */ - assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER); - assert(PIPE_FUNC_LESS == GL_LESS - GL_NEVER); - assert(PIPE_FUNC_EQUAL == GL_EQUAL - GL_NEVER); - assert(PIPE_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); - assert(PIPE_FUNC_GREATER == GL_GREATER - GL_NEVER); - assert(PIPE_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); - assert(PIPE_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); - assert(PIPE_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); - assert(func >= GL_NEVER); - assert(func <= GL_ALWAYS); - return func - GL_NEVER; -} - - static void update_depth_stencil(struct st_context *st) { @@ -121,7 +100,7 @@ update_depth_stencil(struct st_context *st) depth_stencil.depth.enabled = st->ctx->Depth.Test; depth_stencil.depth.writemask = st->ctx->Depth.Mask; - depth_stencil.depth.func = gl_depth_func_to_sp(st->ctx->Depth.Func); + depth_stencil.depth.func = st_compare_func_to_pipe(st->ctx->Depth.Func); depth_stencil.depth.clear = st->ctx->Depth.Clear; if (st->ctx->Query.CurrentOcclusionObject && @@ -130,19 +109,19 @@ update_depth_stencil(struct st_context *st) if (st->ctx->Stencil.Enabled) { depth_stencil.stencil.front_enabled = 1; - depth_stencil.stencil.front_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[0]); - depth_stencil.stencil.front_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[0]); - depth_stencil.stencil.front_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[0]); - depth_stencil.stencil.front_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[0]); + depth_stencil.stencil.front_func = st_compare_func_to_pipe(st->ctx->Stencil.Function[0]); + depth_stencil.stencil.front_fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[0]); + depth_stencil.stencil.front_zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[0]); + depth_stencil.stencil.front_zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[0]); depth_stencil.stencil.ref_value[0] = st->ctx->Stencil.Ref[0] & 0xff; depth_stencil.stencil.value_mask[0] = st->ctx->Stencil.ValueMask[0] & 0xff; depth_stencil.stencil.write_mask[0] = st->ctx->Stencil.WriteMask[0] & 0xff; if (st->ctx->Stencil.TestTwoSide) { depth_stencil.stencil.back_enabled = 1; - depth_stencil.stencil.back_func = gl_stencil_func_to_sp(st->ctx->Stencil.Function[1]); - depth_stencil.stencil.back_fail_op = gl_stencil_op_to_sp(st->ctx->Stencil.FailFunc[1]); - depth_stencil.stencil.back_zfail_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZFailFunc[1]); - depth_stencil.stencil.back_zpass_op = gl_stencil_op_to_sp(st->ctx->Stencil.ZPassFunc[1]); + depth_stencil.stencil.back_func = st_compare_func_to_pipe(st->ctx->Stencil.Function[1]); + depth_stencil.stencil.back_fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[1]); + depth_stencil.stencil.back_zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[1]); + depth_stencil.stencil.back_zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[1]); depth_stencil.stencil.ref_value[1] = st->ctx->Stencil.Ref[1] & 0xff; depth_stencil.stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1] & 0xff; depth_stencil.stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1] & 0xff; -- cgit v1.2.3 From a736670219c67a45483f58b07ab9d55095efdbbd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 09:33:13 -0700 Subject: set sampler state for shadow test --- src/mesa/state_tracker/st_atom_sampler.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index bdc6024477..67a9159069 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -153,6 +153,14 @@ update_samplers(struct st_context *st) sampler.max_anisotropy = texobj->MaxAnisotropy; + /* only care about ARB_shadow, not SGI shadow */ + if (texobj->CompareMode == GL_COMPARE_R_TO_TEXTURE) { + sampler.compare = 1; + sampler.compare_mode = PIPE_TEX_COMPARE_R_TO_TEXTURE; + sampler.compare_func + = st_compare_func_to_pipe(texobj->CompareFunc); + } + /* XXX more sampler state here */ } -- cgit v1.2.3 From 8ce6d6b97e6e90451e7372fe3d94e7e31b0ea569 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 09:33:33 -0700 Subject: move EXT_shadow_funcs assignment --- src/mesa/state_tracker/st_extensions.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 97578e5002..0157bdd6b3 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -145,7 +145,6 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.EXT_pixel_buffer_object = GL_TRUE; ctx->Extensions.EXT_point_parameters = GL_TRUE; ctx->Extensions.EXT_secondary_color = GL_TRUE; - ctx->Extensions.EXT_shadow_funcs = GL_TRUE; /* XXX temp */ ctx->Extensions.EXT_stencil_wrap = GL_TRUE; ctx->Extensions.EXT_texture_env_add = GL_TRUE; ctx->Extensions.EXT_texture_env_combine = GL_TRUE; @@ -206,6 +205,7 @@ void st_init_extensions(struct st_context *st) if (pipe->get_param(pipe, PIPE_CAP_TEXTURE_SHADOW_MAP)) { ctx->Extensions.ARB_depth_texture = GL_TRUE; ctx->Extensions.ARB_shadow = GL_TRUE; + ctx->Extensions.EXT_shadow_funcs = GL_TRUE; /*ctx->Extensions.ARB_shadow_ambient = GL_TRUE;*/ } -- cgit v1.2.3 From dca71c40ea6ae70c6af380653df5b1073dfc9475 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 10:07:57 -0700 Subject: remove translate_compare_func(), use i915_translate_compare_func() --- src/mesa/pipe/i915simple/i915_state.c | 20 +------------------- 1 file changed, 1 insertion(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index d0fe7d2a52..bfd2f8894e 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -90,24 +90,6 @@ static unsigned translate_mip_filter( unsigned filter ) } } -static unsigned translate_compare_func(unsigned func) -{ - switch (func) { - case PIPE_FUNC_NEVER: - case PIPE_FUNC_LESS: - case PIPE_FUNC_EQUAL: - case PIPE_FUNC_LEQUAL: - case PIPE_FUNC_GREATER: - case PIPE_FUNC_NOTEQUAL: - case PIPE_FUNC_GEQUAL: - case PIPE_FUNC_ALWAYS: - return 0; - default: - assert(0); - return 0; - } -} - /* None of this state is actually used for anything yet. */ @@ -247,7 +229,7 @@ i915_create_sampler_state(struct pipe_context *pipe, if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { cso->state[0] |= (SS2_SHADOW_ENABLE | - translate_compare_func(sampler->compare_func)); + i915_translate_compare_func(sampler->compare_func)); minFilt = FILTER_4X4_FLAT; magFilt = FILTER_4X4_FLAT; -- cgit v1.2.3 From 548eddc964c99e24f45ce88cd69f3d9af5aeab1a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 10:08:24 -0700 Subject: enable PIPE_CAP_TEXTURE_SHADOW_MAP (but not quite working yet) --- src/mesa/pipe/i915simple/i915_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 9f11fe5385..4a5b6bec61 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -130,7 +130,7 @@ i915_get_param(struct pipe_context *pipe, int param) case PIPE_CAP_OCCLUSION_QUERY: return 0; case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 0; + return 1; case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: return 11; /* max 1024x1024 */ case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: -- cgit v1.2.3 From 89b7b187dda26f7c7cc5e80360d49fa32f89b6fd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 10:10:35 -0700 Subject: tweak anisotropic filtering code --- src/mesa/pipe/i915simple/i915_context.c | 2 +- src/mesa/pipe/i915simple/i915_state.c | 3 +++ 2 files changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 4a5b6bec61..e43274dc66 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -158,7 +158,7 @@ i915_get_paramf(struct pipe_context *pipe, int param) return 255.0; case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 0.0; + return 4.0; case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: return 16.0; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index bfd2f8894e..70b8195bf1 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -212,6 +212,9 @@ i915_create_sampler_state(struct pipe_context *pipe, if (sampler->max_anisotropy > 1.0) { minFilt = FILTER_ANISOTROPIC; magFilt = FILTER_ANISOTROPIC; + if (sampler->max_anisotropy > 2.0) { + cso->state[0] |= SS2_MAX_ANISO_4; + } } else { minFilt = translate_img_filter( sampler->min_img_filter ); -- cgit v1.2.3 From 02f7f46fa15c7d31d774c638684d4f5b81e360ec Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 11:38:50 -0700 Subject: new init_renderbuffer_bits() helper --- src/mesa/state_tracker/st_cb_fbo.c | 38 ++++++++++++++++++++++++-------------- 1 file changed, 24 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 0d23f7eec4..f58d532b2a 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -52,24 +52,15 @@ /** - * gl_renderbuffer::AllocStorage() + * Compute the renderbuffer's Red/Green/EtcBit fields from the pipe format. */ -static GLboolean -st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, - GLuint width, GLuint height) +static int +init_renderbuffer_bits(struct st_renderbuffer *strb, uint pipeFormat) { - struct pipe_context *pipe = ctx->st->pipe; - struct st_renderbuffer *strb = st_renderbuffer(rb); - const uint pipeFormat - = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE); struct pipe_format_info info; - GLuint cpp; - GLbitfield flags = PIPE_SURFACE_FLAG_RENDER; /* want to render to surface */ if (!st_get_format_info( pipeFormat, &info )) { assert( 0 ); - return GL_FALSE; } strb->Base._ActualFormat = info.base_format; @@ -81,9 +72,26 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, strb->Base.StencilBits = info.stencil_bits; strb->Base.DataType = st_format_datatype(pipeFormat); - assert(strb->Base.DataType); + return info.size; +} - cpp = info.size; + +/** + * gl_renderbuffer::AllocStorage() + */ +static GLboolean +st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, + GLenum internalFormat, + GLuint width, GLuint height) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_renderbuffer *strb = st_renderbuffer(rb); + const uint pipeFormat + = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE); + GLuint cpp; + GLbitfield flags = PIPE_SURFACE_FLAG_RENDER; /* want to render to surface */ + + cpp = init_renderbuffer_bits(strb, pipeFormat); if (strb->surface && strb->surface->format != pipeFormat) { /* need to change surface types, free this surface */ @@ -305,6 +313,8 @@ st_render_texture(GLcontext *ctx, att->Zoffset); assert(strb->surface); + init_renderbuffer_bits(strb, mt->format); + /* printf("RENDER TO TEXTURE obj=%p mt=%p surf=%p %d x %d\n", att->Texture, mt, strb->surface, rb->Width, rb->Height); -- cgit v1.2.3 From fa63d6d32c93a5d8cd88c0c7d4a8ac4c1c1c1a9d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 11:39:42 -0700 Subject: check for divide by W flag when translating TGSI_OPCODE_TEX (fixes projected textures) --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 7838031c17..b867b87e74 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -831,7 +831,13 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_TEX: - emit_tex(p, inst, T0_TEXLD); + if (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide + == TGSI_EXTSWIZZLE_W) { + emit_tex(p, inst, T0_TEXLDP); + } + else { + emit_tex(p, inst, T0_TEXLD); + } break; case TGSI_OPCODE_TXB: -- cgit v1.2.3 From ae9fe0f981377cb25bc3fe6f23a6ee7e3b73d0c2 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 11:44:12 -0700 Subject: Remove TGSI_OPCODE_TXP Use TGSI_OPCODE_TEX with ExtDivide=TGSI_EXTSWIZZLE_W instead. --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 4 ---- src/mesa/pipe/tgsi/exec/tgsi_token.h | 10 +++++----- 2 files changed, 5 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index b867b87e74..d74da85216 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -844,10 +844,6 @@ i915_translate_instruction(struct i915_fp_compile *p, emit_tex(p, inst, T0_TEXLDB); break; - case TGSI_OPCODE_TXP: - emit_tex(p, inst, T0_TEXLDP); - break; - case TGSI_OPCODE_XPD: /* Cross product: * result.x = src0.y * src1.z - src0.z * src1.y; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h index 8d5992facb..2b922c7aef 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_token.h @@ -265,7 +265,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_SUB */ #define TGSI_OPCODE_TEX 52 #define TGSI_OPCODE_TXD 53 -#define TGSI_OPCODE_TXP 132 +/* TGSI_OPCODE_TXP */ #define TGSI_OPCODE_UP2H 54 #define TGSI_OPCODE_UP2US 55 #define TGSI_OPCODE_UP4B 56 @@ -909,7 +909,7 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_TEXKILL */ /* TGSI_OPCODE_TEXLD */ #define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB -#define TGSI_OPCODE_TEXLDP TGSI_OPCODE_TXP +/* TGSI_OPCODE_TEXLDP */ /* CMP - use TGSI_OPCODE_CND0 */ #define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A @@ -1100,10 +1100,10 @@ struct tgsi_immediate_float32 /* TGSI_OPCODE_MOVA */ /* TGSI_OPCODE_LOGP */ -#define TGSI_OPCODE_KIL 133 /* unpredicated kill */ -#define TGSI_OPCODE_END 134 /* aka HALT */ +#define TGSI_OPCODE_KIL 132 /* unpredicated kill */ +#define TGSI_OPCODE_END 133 /* aka HALT */ -#define TGSI_OPCODE_LAST 135 +#define TGSI_OPCODE_LAST 134 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ -- cgit v1.2.3 From cee1d1328aadc501935f60f32cc2c8f8199ff57c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 13:17:43 -0700 Subject: fix bad width/height code in softpipe_clear() --- src/mesa/pipe/softpipe/sp_clear.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 078b6a7964..87f850b6fd 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -51,9 +51,6 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, softpipe_update_derived(softpipe); /* not needed?? */ - w = softpipe->framebuffer.cbufs[0]->width; - h = softpipe->framebuffer.cbufs[0]->height; - /* Use the X coord to trick region_fill() into filling at an offset * from the start of the region. Perhaps pipe_region should have the * 'offset' field, not pipe_surface??? @@ -61,6 +58,8 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, assert(ps->offset % ps->region->cpp == 0); x = ps->offset / ps->region->cpp; y = 0; + w = ps->width; + h = ps->height; assert(w <= ps->region->pitch); assert(h <= ps->region->height); -- cgit v1.2.3 From c0b27149458c1c70b8664cdedb2be842229f4359 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 14:40:29 -0700 Subject: simplify depth_value(), return proper value for Z16 format --- src/mesa/state_tracker/st_cb_clear.c | 20 +++++++------------- 1 file changed, 7 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index cb7e43fb8e..3e591170da 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -77,30 +77,24 @@ color_value(GLuint pipeFormat, const GLfloat color[4]) } -static GLuint +static uint depth_value(GLuint pipeFormat, GLfloat value) { - GLuint val; switch (pipeFormat) { case PIPE_FORMAT_U_Z16: - val = (GLuint) (value * 0xffffff); - break; + return (uint) (value * 0xffff); case PIPE_FORMAT_U_Z32: /* special-case to avoid overflow */ if (value == 1.0) - val = 0xffffffff; + return 0xffffffff; else - val = (GLuint) (value * 0xffffffff); - break; + return (uint) (value * 0xffffffff); case PIPE_FORMAT_S8_Z24: - /*case PIPE_FORMAT_Z24_S8:*/ - val = (GLuint) (value * 0xffffff); - break; + return (uint) (value * 0xffffff); default: - val = 0; assert(0); + return 0; } - return val; } @@ -480,7 +474,7 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) } else { /* simple clear of whole buffer */ - GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); + uint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } } -- cgit v1.2.3 From a7be1c5ac25fe86c4b217625976af1eb37e48a25 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 15:00:11 -0700 Subject: use pipe_surface_reference() in sp_tile_cache_set_surface() --- src/mesa/pipe/softpipe/sp_tile_cache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 9542ec3765..19c06323e1 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -126,7 +126,7 @@ void sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, struct pipe_surface *ps) { - tc->surface = ps; + pipe_surface_reference(&tc->surface, ps); } -- cgit v1.2.3 From 3d8c05f7320151898dd224c1daaf3118e1f7ea34 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 15:01:13 -0700 Subject: Implement shadow comparisons. --- src/mesa/pipe/softpipe/sp_tex_sample.c | 68 +++++++++++++++++++++++++++++++++- 1 file changed, 67 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index e3e607d27c..92958400fc 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -553,6 +553,50 @@ get_texel(struct tgsi_sampler *sampler, } +/** + * Compare texcoord 'p' (aka R) against texture value 'rgba[0]' + * When we sampled the depth texture, the depth value was put into all + * RGBA channels. We look at the red channel here. + */ +static INLINE void +shadow_compare(uint compare_func, + float rgba[NUM_CHANNELS][QUAD_SIZE], + const float p[QUAD_SIZE], + uint j) +{ + int k; + switch (compare_func) { + case PIPE_FUNC_LESS: + k = p[j] < rgba[0][j]; + break; + case PIPE_FUNC_LEQUAL: + k = p[j] <= rgba[0][j]; + break; + case PIPE_FUNC_GREATER: + k = p[j] > rgba[0][j]; + break; + case PIPE_FUNC_GEQUAL: + k = p[j] >= rgba[0][j]; + break; + case PIPE_FUNC_EQUAL: + k = p[j] == rgba[0][j]; + break; + case PIPE_FUNC_NOTEQUAL: + k = p[j] != rgba[0][j]; + break; + case PIPE_FUNC_ALWAYS: + k = 1; + break; + case PIPE_FUNC_NEVER: + k = 0; + break; + default: + assert(0); + } + + rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k; +} + /** * Common code for sampling 1D/2D/cube textures. @@ -567,6 +611,7 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, float rgba[NUM_CHANNELS][QUAD_SIZE], const unsigned faces[4]) { + const uint compare_func = sampler->state->compare_func; unsigned level0, level1, j, imgFilter; int width, height; float levelBlend; @@ -590,6 +635,9 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); get_texel(sampler, faces[j], level0, x, y, 0, rgba, j); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(compare_func, rgba, p, j); + } if (level0 != level1) { /* get texels from second mipmap level and blend */ @@ -598,6 +646,10 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, x = x / 2; y = y / 2; get_texel(sampler, faces[j], level1, x, y, 0, rgba2, j); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ + shadow_compare(compare_func, rgba2, p, j); + } + for (c = 0; c < NUM_CHANNELS; c++) { rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); } @@ -614,6 +666,13 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, get_texel(sampler, faces[j], level0, x1, y0, 0, tx, 1); get_texel(sampler, faces[j], level0, x0, y1, 0, tx, 2); get_texel(sampler, faces[j], level0, x1, y1, 0, tx, 3); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(compare_func, tx, p, 0); + shadow_compare(compare_func, tx, p, 1); + shadow_compare(compare_func, tx, p, 2); + shadow_compare(compare_func, tx, p, 3); + } + for (c = 0; c < 4; c++) { rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); } @@ -629,6 +688,13 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, get_texel(sampler, faces[j], level1, x1, y0, 0, tx, 1); get_texel(sampler, faces[j], level1, x0, y1, 0, tx, 2); get_texel(sampler, faces[j], level1, x1, y1, 0, tx, 3); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ + shadow_compare(compare_func, tx, p, 0); + shadow_compare(compare_func, tx, p, 1); + shadow_compare(compare_func, tx, p, 2); + shadow_compare(compare_func, tx, p, 3); + } + for (c = 0; c < 4; c++) { rgba2[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); @@ -669,7 +735,7 @@ sp_get_samples_2d(struct tgsi_sampler *sampler, float rgba[NUM_CHANNELS][QUAD_SIZE]) { static const unsigned faces[4] = {0, 0, 0, 0}; - sp_get_samples_2d_common(sampler, s, t, NULL, lodbias, rgba, faces); + sp_get_samples_2d_common(sampler, s, t, p, lodbias, rgba, faces); } -- cgit v1.2.3 From f66fed57e65c518c751dc71ca26439ac76313ce9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 15:02:42 -0700 Subject: For TGSI_TEXTURE_2D/RECT, need to provide 3 coords (STP) in case shadow compare mode is enabled. --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 3494b21599..5c2fecb3e3 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1253,12 +1253,14 @@ exec_tex(struct tgsi_exec_machine *mach, FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { case TGSI_EXTSWIZZLE_W: - FETCH(&r[2], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[2] ); - micro_div( &r[1], &r[1], &r[2] ); + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); break; case TGSI_EXTSWIZZLE_ONE: @@ -1269,15 +1271,15 @@ exec_tex(struct tgsi_exec_machine *mach, } if (biasLod) { - FETCH(&r[2], 0, CHAN_W); - lodBias = r[2].f[0]; + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; } else lodBias = 0.0; fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], NULL, lodBias, - &r[0], &r[1], &r[2], &r[3]); + &r[0], &r[1], &r[2], lodBias, /* inputs */ + &r[0], &r[1], &r[2], &r[3]); /* outputs */ break; case TGSI_TEXTURE_3D: -- cgit v1.2.3 From 03484d0aaefce84bc304c9d87f28edd77e10844a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 16:25:17 -0700 Subject: comments, remove unneeded #includes --- src/mesa/pipe/draw/draw_prim.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index cf8943b40d..94f6df56df 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -30,12 +30,9 @@ * Keith Whitwell */ -#include "pipe/p_util.h" #include "draw_private.h" #include "draw_context.h" -#include "draw_prim.h" -#include "pipe/tgsi/exec/tgsi_core.h" #define RP_NONE 0 @@ -150,7 +147,10 @@ static struct prim_header *get_queued_prim( struct draw_context *draw, - +/** + * Add a point to the primitive queue. + * \param i0 index into user's vertex arrays + */ static void do_point( struct draw_context *draw, unsigned i0 ) { @@ -163,6 +163,11 @@ static void do_point( struct draw_context *draw, } +/** + * Add a line to the primitive queue. + * \param i0 index into user's vertex arrays + * \param i1 index into user's vertex arrays + */ static void do_line( struct draw_context *draw, boolean reset_stipple, unsigned i0, @@ -177,6 +182,9 @@ static void do_line( struct draw_context *draw, prim->v[1] = draw->vcache.get_vertex( draw, i1 ); } +/** + * Add a triangle to the primitive queue. + */ static void do_triangle( struct draw_context *draw, unsigned i0, unsigned i1, -- cgit v1.2.3 From 03f4a487dd53f7d76830665e98fdbc9313d7c9b2 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 16:26:55 -0700 Subject: remove unneeded #includes --- src/mesa/pipe/draw/draw_debug.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_debug.c b/src/mesa/pipe/draw/draw_debug.c index 246e20fb30..d6220b5f62 100644 --- a/src/mesa/pipe/draw/draw_debug.c +++ b/src/mesa/pipe/draw/draw_debug.c @@ -30,10 +30,8 @@ * Keith Whitwell */ -#include "pipe/p_util.h" #include "draw_private.h" #include "draw_context.h" -#include "draw_prim.h" -- cgit v1.2.3 From 48863cdd0a912f34d54bdc312485241a970a9fba Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 16:32:24 -0700 Subject: new comments --- src/mesa/pipe/draw/draw_private.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index dc4057bff3..89d292901e 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -224,8 +224,8 @@ struct draw_context */ struct { struct { - unsigned elt; - struct vertex_header *dest; + unsigned elt; /**< index into the user's vertex arrays */ + struct vertex_header *dest; /**< points into vcache.vertex[] array */ } queue[VS_QUEUE_LENGTH]; unsigned queue_nr; } vs; -- cgit v1.2.3 From 1574611faf9d9106cf167d277238b8c2662beabb Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 16:36:00 -0700 Subject: Fix minor logic error in get_queued_prim() which caused flushing when there was still room in the prim queue. --- src/mesa/pipe/draw/draw_prim.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 94f6df56df..baae6e98c5 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -60,6 +60,10 @@ static void draw_prim_queue_flush( struct draw_context *draw ) struct draw_stage *first = draw->pipeline.first; unsigned i; + if (0) + printf("Flushing with %d prims, %d verts\n", + draw->pq.queue_nr, draw->vs.queue_nr); + /* Make sure all vertices are available: */ if (draw->vs.queue_nr) @@ -137,11 +141,13 @@ static struct prim_header *get_queued_prim( struct draw_context *draw, // fprintf(stderr, "v"); draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE_INVALIDATE ); } - else if (draw->pq.queue_nr + 1 >= PRIM_QUEUE_LENGTH) { + else if (draw->pq.queue_nr == PRIM_QUEUE_LENGTH) { // fprintf(stderr, "p"); draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); } + assert(draw->pq.queue_nr < PRIM_QUEUE_LENGTH); + return &draw->pq.queue[draw->pq.queue_nr++]; } -- cgit v1.2.3 From ec0c5e59c2c3470cc59d25383854d12a76dca70a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 16:37:25 -0700 Subject: Remove unneeded header, draw_prim.h --- src/mesa/pipe/draw/draw_prim.h | 11 ----------- 1 file changed, 11 deletions(-) delete mode 100644 src/mesa/pipe/draw/draw_prim.h (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.h b/src/mesa/pipe/draw/draw_prim.h deleted file mode 100644 index 07cd3e2c9f..0000000000 --- a/src/mesa/pipe/draw/draw_prim.h +++ /dev/null @@ -1,11 +0,0 @@ - - -#ifndef DRAW_PRIM_H -#define DRAW_PRIM_H - - -void draw_invalidate_vcache( struct draw_context *draw ); - - - -#endif /* DRAW_PRIM_H */ -- cgit v1.2.3 From 990fe4c0bf735206c3cc7346d84adc782595bc3a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 16:55:29 -0700 Subject: remove unneeded #includes, added assertion in draw_compute_vertex_size() --- src/mesa/pipe/draw/draw_vertex.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index ce76cb043c..dea26a3d0a 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -34,11 +34,7 @@ */ -#include "pipe/p_defines.h" -#include "pipe/p_util.h" - #include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_context.h" #include "pipe/draw/draw_vertex.h" @@ -86,6 +82,8 @@ draw_compute_vertex_size(struct vertex_info *vinfo) assert(0); } } + + assert(vinfo->size * 4 <= MAX_VERTEX_SIZE); } -- cgit v1.2.3 From 64469863212dcc41995c473032856096c4af12b3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 8 Nov 2007 17:07:12 -0700 Subject: Reorganize user-space vertex fields in draw_context into 'user' group. This sub-struct collects the incoming user-provided data/pointers in one place. Ex: draw->mapped_vbuffer becomes draw->user.vbuffer, etc. --- src/mesa/pipe/draw/draw_context.c | 8 ++++---- src/mesa/pipe/draw/draw_feedback.c | 8 ++++---- src/mesa/pipe/draw/draw_prim.c | 2 +- src/mesa/pipe/draw/draw_private.h | 30 +++++++++++++++++----------- src/mesa/pipe/draw/draw_vertex_cache.c | 19 ++++++++++++------ src/mesa/pipe/draw/draw_vertex_fetch.c | 2 +- src/mesa/pipe/draw/draw_vertex_shader.c | 4 ++-- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 4 ++-- 8 files changed, 45 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 0de8bed529..4f6392605e 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -196,7 +196,7 @@ draw_set_mapped_vertex_buffer(struct draw_context *draw, { draw_flush( draw ); - draw->mapped_vbuffer[attr] = buffer; + draw->user.vbuffer[attr] = buffer; } @@ -206,7 +206,7 @@ draw_set_mapped_constant_buffer(struct draw_context *draw, { draw_flush( draw ); - draw->mapped_constants = buffer; + draw->user.constants = buffer; } @@ -217,8 +217,8 @@ draw_set_mapped_feedback_buffer(struct draw_context *draw, uint index, draw_flush( draw ); assert(index < PIPE_MAX_FEEDBACK_ATTRIBS); - draw->mapped_feedback_buffer[index] = buffer; - draw->mapped_feedback_buffer_size[index] = size; /* in bytes */ + draw->user.feedback_buffer[index] = buffer; + draw->user.feedback_buffer_size[index] = size; /* in bytes */ } diff --git a/src/mesa/pipe/draw/draw_feedback.c b/src/mesa/pipe/draw/draw_feedback.c index ee54db0ad5..b9906e5b4b 100644 --- a/src/mesa/pipe/draw/draw_feedback.c +++ b/src/mesa/pipe/draw/draw_feedback.c @@ -128,20 +128,20 @@ static void feedback_begin( struct draw_stage *stage ) vertex_size += feedback->size[i]; } /* compute max number of vertices we can feedback */ - fs->max_vert_emit = stage->draw->mapped_feedback_buffer_size[0] + fs->max_vert_emit = stage->draw->user.feedback_buffer_size[0] / sizeof(float) / vertex_size; - fs->dest[0] = (float *) stage->draw->mapped_feedback_buffer[0]; + fs->dest[0] = (float *) stage->draw->user.feedback_buffer[0]; } else { uint i; uint max = ~0; for (i = 0; i < feedback->num_attribs; i++) { - uint n = stage->draw->mapped_feedback_buffer_size[i] + uint n = stage->draw->user.feedback_buffer_size[i] / sizeof(float) / feedback->size[i]; if (n < max) max = n; - fs->dest[i] = (float *) stage->draw->mapped_feedback_buffer[i]; + fs->dest[i] = (float *) stage->draw->user.feedback_buffer[i]; } fs->max_vert_emit = max; } diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index baae6e98c5..e4a65b9f24 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -64,7 +64,7 @@ static void draw_prim_queue_flush( struct draw_context *draw ) printf("Flushing with %d prims, %d verts\n", draw->pq.queue_nr, draw->vs.queue_nr); - /* Make sure all vertices are available: + /* Make sure all vertices are available/shaded: */ if (draw->vs.queue_nr) draw_vertex_shader_queue_flush(draw); diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 89d292901e..03b48aa93c 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -178,17 +178,23 @@ struct draw_context struct pipe_vertex_buffer feedback_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element feedback_element[PIPE_ATTRIB_MAX]; - /** The mapped vertex element/index buffer */ - const void *mapped_elts; - unsigned eltSize; /**< bytes per index (0, 1, 2 or 4) */ - /** The mapped vertex arrays */ - const void *mapped_vbuffer[PIPE_ATTRIB_MAX]; - /** The mapped constant buffers (for vertex shader) */ - const void *mapped_constants; - - /** The mapped vertex element/index buffer */ - void *mapped_feedback_buffer[PIPE_MAX_FEEDBACK_ATTRIBS]; - uint mapped_feedback_buffer_size[PIPE_MAX_FEEDBACK_ATTRIBS]; /* in bytes */ + /* user-space vertex data, buffers */ + struct { + /** vertex element/index buffer (ex: glDrawElements) */ + const void *elts; + /** bytes per index (0, 1, 2 or 4) */ + unsigned eltSize; + + /** vertex arrays */ + const void *vbuffer[PIPE_ATTRIB_MAX]; + + /** constant buffer (for vertex shader) */ + const void *constants; + + /** The vertex feedback buffer */ + void *feedback_buffer[PIPE_MAX_FEEDBACK_ATTRIBS]; + uint feedback_buffer_size[PIPE_MAX_FEEDBACK_ATTRIBS]; /* in bytes */ + } user; /* Clip derived state: */ @@ -216,6 +222,7 @@ struct draw_context struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; unsigned overflow; + /** To find space in the vertex cache: */ struct vertex_header *(*get_vertex)( struct draw_context *draw, unsigned i ); } vcache; @@ -233,7 +240,6 @@ struct draw_context /* Prim pipeline queue: */ struct { - /* Need to queue up primitives until their vertices have been * transformed by a vs queue flush. */ diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index 6689907ddf..29993f14d2 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -51,8 +51,15 @@ void draw_vertex_cache_invalidate( struct draw_context *draw ) } -/* Check if vertex is in cache, otherwise add it. It won't go through +/** + * Check if vertex is in cache, otherwise add it. It won't go through * VS yet, not until there is a flush operation or the VS queue fills up. + * + * Note that cache entries are basically just two pointers: the first + * an index into the user's vertex arrays, the second a location in + * the vertex shader cache for the post-transformed vertex. + * + * \return pointer to location of (post-transformed) vertex header in the cache */ static struct vertex_header *get_vertex( struct draw_context *draw, unsigned i ) @@ -104,7 +111,7 @@ static struct vertex_header *get_vertex( struct draw_context *draw, static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, unsigned i ) { - const unsigned *elts = (const unsigned *) draw->mapped_elts; + const unsigned *elts = (const unsigned *) draw->user.elts; return get_vertex( draw, elts[i] ); } @@ -112,7 +119,7 @@ static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, unsigned i ) { - const ushort *elts = (const ushort *) draw->mapped_elts; + const ushort *elts = (const ushort *) draw->user.elts; return get_vertex( draw, elts[i] ); } @@ -120,7 +127,7 @@ static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, unsigned i ) { - const ubyte *elts = (const ubyte *) draw->mapped_elts; + const ubyte *elts = (const ubyte *) draw->user.elts; return get_vertex( draw, elts[i] ); } @@ -189,7 +196,7 @@ draw_set_mapped_element_buffer( struct draw_context *draw, default: assert(0); } - draw->mapped_elts = elements; - draw->eltSize = eltSize; + draw->user.elts = elements; + draw->user.eltSize = eltSize; } diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 5a7e6febe9..eca10e89fb 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -103,7 +103,7 @@ void draw_vertex_fetch( struct draw_context *draw, unsigned buf = draw->vertex_element[attr].vertex_buffer_index; const void *src - = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] + = (const void *) ((const ubyte *) draw->user.vbuffer[buf] + draw->vertex_buffer[buf].buffer_offset + draw->vertex_element[attr].src_offset + elts[j] * draw->vertex_buffer[buf].pitch); diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 529ed288eb..e8801addac 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -93,7 +93,7 @@ run_vertex_program(struct draw_context *draw, == TGSI_SEMANTIC_POSITION); /* Consts does not require 16 byte alignment. */ - machine->Consts = (float (*)[4]) draw->mapped_constants; + machine->Consts = (float (*)[4]) draw->user.constants; machine->Inputs = ALIGN16_ASSIGN(inputs); machine->Outputs = ALIGN16_ASSIGN(outputs); @@ -179,8 +179,8 @@ run_vertex_program(struct draw_context *draw, /** + * Run the vertex shader on all vertices in the vertex queue. * Called by the draw module when the vertx cache needs to be flushed. - * This involves running the vertex shader. */ void draw_vertex_shader_queue_flush( struct draw_context *draw ) { diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index c42e9bbd69..4f48db610f 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -83,7 +83,7 @@ void vertex_fetch(struct draw_context *draw, unsigned buf = draw->vertex_element[attr].vertex_buffer_index; const void *src - = (const void *) ((const ubyte *) draw->mapped_vbuffer[buf] + = (const void *) ((const ubyte *) draw->user.vbuffer[buf] + draw->vertex_buffer[buf].buffer_offset + draw->vertex_element[attr].src_offset + elt * draw->vertex_buffer[buf].pitch); @@ -124,7 +124,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) struct vertex_header *dests[VS_QUEUE_LENGTH]; float inputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; float outputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; - float (*consts)[4] = (float (*)[4]) draw->mapped_constants; + float (*consts)[4] = (float (*)[4]) draw->user.constants; struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; -- cgit v1.2.3 From 33cf20bb458ed736d4404d4e3340a117c7d07f63 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Fri, 9 Nov 2007 09:21:40 +0000 Subject: remove duplicate init func --- src/mesa/vf/vf_sse.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/vf/vf_sse.c b/src/mesa/vf/vf_sse.c index a5d143461b..3ce76e2b04 100644 --- a/src/mesa/vf/vf_sse.c +++ b/src/mesa/vf/vf_sse.c @@ -345,8 +345,6 @@ static GLboolean build_vertex_emit( struct x86_program *p ) struct x86_reg vp1 = x86_make_reg(file_XMM, 2); GLubyte *fixup, *label; - x86_init_func(&p->func); - /* Push a few regs? */ x86_push(&p->func, countEBP); -- cgit v1.2.3 From 50a56cd755d17e4b754fc019badbda0f113c3065 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 9 Nov 2007 07:54:28 -0700 Subject: minor simplifcation in vertex_fetch() --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 9 +-------- 1 file changed, 1 insertion(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 4f48db610f..3e005a76f3 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -87,14 +87,7 @@ void vertex_fetch(struct draw_context *draw, + draw->vertex_buffer[buf].buffer_offset + draw->vertex_element[attr].src_offset + elt * draw->vertex_buffer[buf].pitch); - float p[4]; - - fetch_attrib4(src, draw->vertex_element[attr].src_format, p); - - inputs[attr][0] = p[0]; /*X*/ - inputs[attr][1] = p[1]; /*Y*/ - inputs[attr][2] = p[2]; /*Z*/ - inputs[attr][3] = p[3]; /*W*/ + fetch_attrib4(src, draw->vertex_element[attr].src_format, inputs[attr]); } } -- cgit v1.2.3 From 193c85ec7a1aec44eebc67c6224fb6ecbb4607a5 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 9 Nov 2007 07:54:46 -0700 Subject: comments, assertions --- src/mesa/pipe/draw/draw_context.c | 8 ++++++-- src/mesa/pipe/draw/draw_private.h | 4 ++-- 2 files changed, 8 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 4f6392605e..e7997180b5 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -126,7 +126,8 @@ void draw_set_rasterizer_state( struct draw_context *draw, /** - * Plug in the primitive rendering/rasterization stage. + * Plug in the primitive rendering/rasterization stage (which is the last + * stage in the drawing pipeline). * This is provided by the device driver. */ void draw_set_rasterize_stage( struct draw_context *draw, @@ -230,6 +231,8 @@ draw_set_mapped_feedback_buffer(struct draw_context *draw, uint index, */ void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) { + assert(!stage->tmp); + stage->nr_tmps = nr; if (nr) { @@ -244,7 +247,7 @@ void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) } /** - * Reset the verticies ids of this and subsequent stages. + * Reset the vertex ids for the stage's temp verts. */ void draw_reset_tmps( struct draw_stage *stage ) { @@ -260,6 +263,7 @@ void draw_free_tmps( struct draw_stage *stage ) if (stage->tmp) { FREE( stage->tmp[0] ); FREE( stage->tmp ); + stage->tmp = NULL; } } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 03b48aa93c..09acf69623 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -100,7 +100,7 @@ struct draw_stage struct draw_stage *next; /**< next stage in pipeline */ - struct vertex_header **tmp; + struct vertex_header **tmp; /**< temp vert storage, such as for clipping */ unsigned nr_tmps; void (*begin)( struct draw_stage * ); @@ -117,7 +117,7 @@ struct draw_stage void (*end)( struct draw_stage * ); /** - * Reset temporary vertices ids in this stage + * Reset temporary vertex ids in this stage * * draw_free_tmps will be called instead if null. */ -- cgit v1.2.3 From 90dd0cb822f2fe14258c786e5c37da69472b7d17 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 15 Nov 2007 17:07:50 +0000 Subject: Include fences in the i915simple winsys interface. --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 14 ----- .../drivers/dri/intel_winsys/intel_batchbuffer.h | 2 - src/mesa/drivers/dri/intel_winsys/intel_winsys.h | 17 ++++++ .../drivers/dri/intel_winsys/intel_winsys_i915.c | 60 +++++++++++++++++++--- src/mesa/pipe/i915simple/i915_batch.h | 10 ++-- src/mesa/pipe/i915simple/i915_context.h | 3 ++ src/mesa/pipe/i915simple/i915_flush.c | 3 +- src/mesa/pipe/i915simple/i915_winsys.h | 18 ++++++- 8 files changed, 96 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index 49e04d81ec..ed223977e2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -307,20 +307,6 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch) } -void -intel_batchbuffer_finish(struct intel_batchbuffer *batch) -{ - struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); - if (fence) { - driFenceReference(fence); - driFenceFinish(fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, - GL_FALSE); - driFenceUnReference(fence); - } -} - - /* This is the only way buffers get added to the validate list. */ boolean diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h index 82feafa21f..2c943e68e5 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h @@ -72,8 +72,6 @@ struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context *intel); void intel_batchbuffer_free(struct intel_batchbuffer *batch); -void intel_batchbuffer_finish(struct intel_batchbuffer *batch); - struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer *batch); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h index 89e63e0a79..f944cd23c3 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h @@ -32,7 +32,9 @@ struct intel_context; struct pipe_context; struct pipe_winsys; struct pipe_buffer_handle; +struct pipe_fence; struct _DriBufferObject; +struct _DriFenceObject; struct pipe_winsys * intel_create_pipe_winsys( int fd ); @@ -66,5 +68,20 @@ pipe_bo( struct _DriBufferObject *bo ) } +/* Turn the pipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static INLINE struct _DriFenceObject * +dri_fo( struct pipe_fence *bo ) +{ + return (struct _DriFenceObject *)bo; +} + +static INLINE struct pipe_fence * +pipe_fo( struct _DriFenceObject *bo ) +{ + return (struct pipe_fence *)bo; +} + #endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index 8e0eea4392..7713d6e48d 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -111,19 +111,62 @@ static void intel_i915_batch_reloc( struct i915_winsys *sws, -static void intel_i915_batch_flush( struct i915_winsys *sws ) +static struct pipe_fence * +intel_i915_batch_flush( struct i915_winsys *sws ) { struct intel_context *intel = intel_i915_winsys(sws)->intel; - intel_batchbuffer_flush( intel->batch ); -// if (0) intel_i915_batch_wait_idle( sws ); + return pipe_fo(intel_batchbuffer_flush( intel->batch )); } -static void intel_i915_batch_finish( struct i915_winsys *sws ) +static void +intel_i915_fence_reference( struct i915_winsys *sws, + struct pipe_fence **dst_fence, + struct pipe_fence *src_fence ) { - struct intel_context *intel = intel_i915_winsys(sws)->intel; - intel_batchbuffer_finish( intel->batch ); + struct _DriFenceObject **dri_dst_fence = (struct _DriFenceObject **)dst_fence; + struct _DriFenceObject *dri_src_fence = (struct _DriFenceObject *)dst_fence; + + if(dri_src_fence) + driFenceReference(dri_src_fence); + + if(*dri_dst_fence) + driFenceUnReference(*dri_dst_fence); + + *dri_dst_fence = dri_src_fence; +} + + +static int +intel_i915_fence_is_signalled( struct i915_winsys *sws, + struct pipe_fence *fence ) +{ + struct _DriFenceObject *dri_fence = dri_fo(fence); + int ret = 1; + if (fence) { + driFenceReference(dri_fence); + ret = driFenceSignaled(dri_fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW); + driFenceUnReference(dri_fence); + } + return ret; +} + + +static int +intel_i915_fence_wait( struct i915_winsys *sws, + struct pipe_fence *fence ) +{ + struct _DriFenceObject *dri_fence = dri_fo(fence); + if (fence) { + driFenceReference(dri_fence); + driFenceFinish(dri_fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, + GL_FALSE); + driFenceUnReference(dri_fence); + } + return 1; } @@ -143,7 +186,10 @@ intel_create_i915simple( struct intel_context *intel, iws->winsys.batch_dword = intel_i915_batch_dword; iws->winsys.batch_reloc = intel_i915_batch_reloc; iws->winsys.batch_flush = intel_i915_batch_flush; - iws->winsys.batch_finish = intel_i915_batch_finish; + iws->winsys.fence_reference = intel_i915_fence_reference; + iws->winsys.fence_is_signalled = intel_i915_fence_is_signalled; + iws->winsys.fence_wait = intel_i915_fence_wait; + iws->intel = intel; /* Create the i915simple context: diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h index fb88cd6db0..99b03c37db 100644 --- a/src/mesa/pipe/i915simple/i915_batch.h +++ b/src/mesa/pipe/i915simple/i915_batch.h @@ -44,11 +44,11 @@ #define ADVANCE_BATCH() -#define FLUSH_BATCH() do { \ - if (0) i915_dump_batchbuffer( i915 ); \ - i915->winsys->batch_flush( i915->winsys ); \ - i915->batch_start = NULL; \ - i915->hardware_dirty = ~0; \ +#define FLUSH_BATCH() do { \ + if (0) i915_dump_batchbuffer( i915 ); \ + i915->last_fence = i915->winsys->batch_flush( i915->winsys ); \ + i915->batch_start = NULL; \ + i915->hardware_dirty = ~0; \ } while (0) #endif diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index ee430ebc90..488682f852 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -80,6 +80,7 @@ +struct pipe_fence; struct i915_cache_context; /* Use to calculate differences between state emitted to hardware and @@ -184,6 +185,8 @@ struct i915_context unsigned dirty; unsigned *batch_start; + + struct pipe_fence *last_fence; /** Vertex buffer */ struct pipe_buffer_handle *vbo; diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c index 3c2069b827..1044cb1bdb 100644 --- a/src/mesa/pipe/i915simple/i915_flush.c +++ b/src/mesa/pipe/i915simple/i915_flush.c @@ -69,7 +69,8 @@ static void i915_flush( struct pipe_context *pipe, FLUSH_BATCH(); if (flags & PIPE_FLUSH_WAIT) { - i915->winsys->batch_finish(i915->winsys); + if( i915->last_fence ) + i915->winsys->fence_wait(i915->winsys, i915->last_fence); } } diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 2c0f335d34..03457c1d61 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -52,6 +52,7 @@ struct pipe_buffer_handle; struct pipe_winsys; +struct pipe_fence; /** @@ -97,8 +98,21 @@ struct i915_winsys { unsigned access_flags, unsigned delta ); - void (*batch_flush)( struct i915_winsys *sws ); - void (*batch_finish)( struct i915_winsys *sws ); + struct pipe_fence *(*batch_flush)( struct i915_winsys *sws ); + + + /* Fence + */ + void (*fence_reference)( struct i915_winsys *sws, + struct pipe_fence **dst_fence, + struct pipe_fence *src_fence ); + + int (*fence_is_signalled)( struct i915_winsys *sws, + struct pipe_fence *fence ); + + int (*fence_wait)( struct i915_winsys *sws, + struct pipe_fence *fence ); + }; #define I915_BUFFER_ACCESS_WRITE 0x1 -- cgit v1.2.3 From f361edae5d495254f4bc27473bf24b1ac7d5af06 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 16 Nov 2007 09:31:25 -0700 Subject: added a clip-bypass flag to rasterizer state --- src/mesa/pipe/draw/draw_validate.c | 1 + src/mesa/pipe/p_state.h | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index fdeb1be8ca..4e8f986b27 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -85,6 +85,7 @@ static void validate_begin( struct draw_stage *stage ) /* Clip stage */ + if (!draw->rasterizer->bypass_clipping) { draw->pipeline.clip->next = next; next = draw->pipeline.clip; diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 848c32701f..570f44e24e 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -92,6 +92,7 @@ struct pipe_rasterizer_state unsigned line_stipple_enable:1; unsigned line_stipple_factor:8; /**< [1..256] actually */ unsigned line_stipple_pattern:16; + unsigned bypass_clipping:1; float line_width; float point_size; /**< used when no per-vertex size */ -- cgit v1.2.3 From 95f066fc54ed351a93350b99b584987b88524e71 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 16 Nov 2007 10:39:08 -0700 Subject: note problem with fragment coord Y orientation --- src/mesa/pipe/softpipe/sp_quad_fs.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 17fb10ea72..ed14dac18e 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -104,6 +104,10 @@ shade_quad( machine->Inputs[0].xyzw[0].f[2] = fx; machine->Inputs[0].xyzw[0].f[3] = fx + 1.0f; + /* XXX for OpenGL we need to invert the Y pos here (y=0=top). + * but that'll mess up linear/perspective interpolation of other + * attributes... + */ machine->Inputs[0].xyzw[1].f[0] = fy; machine->Inputs[0].xyzw[1].f[1] = fy; machine->Inputs[0].xyzw[1].f[2] = fy + 1.0f; -- cgit v1.2.3 From 6a1154bab0d296b60f889bdc13254568f4051104 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 16 Nov 2007 17:12:22 -0700 Subject: adjustments so st_feedback_draw_vbo() can be used for glRasterPos --- src/mesa/state_tracker/st_draw.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 3e2ed6cada..3b6d829145 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -318,7 +318,7 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, /** * Set the (private) draw module's post-transformed vertex format when in - * GL_SELECT or GL_FEEDBACK mode. + * GL_SELECT or GL_FEEDBACK mode or for glRasterPos. */ static void set_feedback_vertex_format(GLcontext *ctx) @@ -328,7 +328,14 @@ set_feedback_vertex_format(GLcontext *ctx) enum interp_mode interp[PIPE_MAX_SHADER_OUTPUTS]; GLuint n, i; - if (ctx->RenderMode == GL_FEEDBACK) { + if (ctx->RenderMode == GL_SELECT) { + assert(ctx->RenderMode == GL_SELECT); + n = 1; + attrs[0] = FORMAT_4F; + interp[0] = INTERP_NONE; + } + else { + /* GL_FEEDBACK, or glRasterPos */ /* emit all attribs (pos, color, texcoord) as GLfloat[4] */ n = st->state.vs->state.num_outputs; for (i = 0; i < n; i++) { @@ -336,19 +343,14 @@ set_feedback_vertex_format(GLcontext *ctx) interp[i] = INTERP_NONE; } } - else { - assert(ctx->RenderMode == GL_SELECT); - n = 1; - attrs[0] = FORMAT_4F; - interp[0] = INTERP_NONE; - } draw_set_vertex_attributes(st->draw, attrs, interp, n); } /** - * Called by VBO to draw arrays when in selection or feedback mode. + * Called by VBO to draw arrays when in selection or feedback mode and + * to implement glRasterPos. * This is very much like the normal draw_vbo() function above. * Look at code refactoring some day. * Might move this into the failover module some day. @@ -373,8 +375,6 @@ st_feedback_draw_vbo(GLcontext *ctx, GLuint attr, i; ubyte *mapped_constants; - assert(ctx->RenderMode == GL_SELECT || - ctx->RenderMode == GL_FEEDBACK); assert(draw); st_validate_state(ctx->st); -- cgit v1.2.3 From aa880bdfa05d8ff2486ef8266f93dea983b7c6fd Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 16 Nov 2007 17:13:01 -0700 Subject: Reimplement glRasterPos using the private 'draw' module. --- src/mesa/state_tracker/st_cb_rasterpos.c | 440 +++++++++++-------------------- src/mesa/state_tracker/st_context.h | 3 +- 2 files changed, 160 insertions(+), 283 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 5b97c1ee34..5279cb1cd4 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -25,353 +25,229 @@ * **************************************************************************/ - /* - * Authors: - * Brian Paul - */ +/** + * glRasterPos implementation. Basically render a GL_POINT with our + * private draw module. Plug in a special "rasterpos" stage at the end + * of the 'draw' pipeline to capture the results and update the current + * raster pos attributes. + * + * Authors: + * Brian Paul + */ + #include "main/imports.h" -#include "main/feedback.h" #include "main/macros.h" #include "st_context.h" #include "st_atom.h" -#include "st_cache.h" #include "st_draw.h" -#include "st_program.h" #include "st_cb_rasterpos.h" #include "st_draw.h" -#include "st_format.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_winsys.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_private.h" #include "shader/prog_instruction.h" +#include "vbo/vbo.h" -static void -setup_vertex_attribs(GLcontext *ctx) +/** + * Our special drawing pipeline stage (replaces rasterization). + */ +struct rastpos_stage { - struct pipe_context *pipe = ctx->st->pipe; - const struct cso_vertex_shader *vs = ctx->st->state.vs; - const struct st_vertex_program *stvp = ctx->st->vp; - uint slot; - - /* all attributes come from the default attribute buffer */ - { - struct pipe_vertex_buffer vbuffer; - vbuffer.buffer = ctx->st->default_attrib_buffer; - vbuffer.buffer_offset = 0; - vbuffer.pitch = 0; /* must be zero! */ - vbuffer.max_index = 1; - pipe->set_vertex_buffer(pipe, 0, &vbuffer); - } + struct draw_stage stage; /**< Base class */ + GLcontext *ctx; /**< Rendering context */ - for (slot = 0; slot < vs->state.num_inputs; slot++) { - struct pipe_vertex_element velement; - const GLuint attr = stvp->index_to_input[slot]; + /* vertex attrib info we can setup once and re-use */ + struct gl_client_array array[VERT_ATTRIB_MAX]; + const struct gl_client_array *arrays[VERT_ATTRIB_MAX]; + struct _mesa_prim prim; +}; - velement.src_offset = attr * 4 * sizeof(GLfloat); - velement.vertex_buffer_index = 0; - velement.dst_offset = 0; - velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - pipe->set_vertex_element(pipe, slot, &velement); - } -} +static INLINE struct rastpos_stage * +rastpos_stage( struct draw_stage *stage ) +{ + return (struct rastpos_stage *) stage; +} static void -setup_feedback(GLcontext *ctx) +rastpos_begin( struct draw_stage *stage ) { - struct pipe_context *pipe = ctx->st->pipe; - const struct pipe_shader_state *vs = &ctx->st->state.vs->state; - struct pipe_feedback_state feedback; - uint i; - - memset(&feedback, 0, sizeof(feedback)); - feedback.enabled = 1; - feedback.interleaved = 1; - feedback.discard = 1; - feedback.num_attribs = 0; - - /* feedback all results from vertex shader */ - for (i = 0; i < vs->num_outputs; i++) { - feedback.attrib[feedback.num_attribs] = i; - feedback.size[feedback.num_attribs] = 4; - feedback.num_attribs++; - } - - if (pipe->set_feedback_state) - pipe->set_feedback_state(pipe, &feedback); + /* no-op */ } +static void +rastpos_end( struct draw_stage *stage ) +{ + /* no-op */ +} - - - -/** - * Clip a point against the view volume. - * - * \param v vertex vector describing the point to clip. - * - * \return zero if outside view volume, or one if inside. - */ -static GLuint -viewclip_point( const GLfloat v[] ) +static void +rastpos_reset_stipple_counter( struct draw_stage *stage ) { - if ( v[0] > v[3] || v[0] < -v[3] - || v[1] > v[3] || v[1] < -v[3] - || v[2] > v[3] || v[2] < -v[3] ) { - return 0; - } - else { - return 1; - } + /* no-op */ } +static void +rastpos_tri( struct draw_stage *stage, struct prim_header *prim ) +{ + /* should never get here */ + assert(0); +} -/** - * Clip a point against the far/near Z clipping planes. - * - * \param v vertex vector describing the point to clip. - * - * \return zero if outside view volume, or one if inside. - */ -static GLuint -viewclip_point_z( const GLfloat v[] ) +static void +rastpos_line( struct draw_stage *stage, struct prim_header *prim ) { - if (v[2] > v[3] || v[2] < -v[3] ) { - return 0; - } - else { - return 1; - } + /* should never get here */ + assert(0); } /** - * Clip a point against the user clipping planes. - * - * \param ctx GL context. - * \param v vertex vector describing the point to clip. - * - * \return zero if the point was clipped, or one otherwise. + * Update a raster pos attribute from the vertex result if it's present, + * else copy the current attrib. */ -static GLuint -userclip_point( GLcontext *ctx, const GLfloat v[] ) +static void +update_attrib(GLcontext *ctx, const GLuint *outputMapping, + const struct vertex_header *vert, + GLfloat *dest, + GLuint result, GLuint defaultAttrib) { - GLuint p; - - for (p = 0; p < ctx->Const.MaxClipPlanes; p++) { - if (ctx->Transform.ClipPlanesEnabled & (1 << p)) { - GLfloat dot = v[0] * ctx->Transform._ClipUserPlane[p][0] - + v[1] * ctx->Transform._ClipUserPlane[p][1] - + v[2] * ctx->Transform._ClipUserPlane[p][2] - + v[3] * ctx->Transform._ClipUserPlane[p][3]; - if (dot < 0.0F) { - return 0; - } - } - } - - return 1; + const GLfloat *src; + const GLuint k = outputMapping[result]; + if (k != ~0) + src = vert->data[k]; + else + src = ctx->Current.Attrib[defaultAttrib]; + COPY_4V(dest, src); } /** - * Update the current raster position. - * Do clip testing, etc. here. + * Normally, this function would render a GL_POINT. */ static void -update_rasterpos(GLcontext *ctx, - const float clipPos[4], - const float color0[4], - const float color1[4], - const float *fog, - const float *tex) +rastpos_point(struct draw_stage *stage, struct prim_header *prim) { - uint i; - float d, ndc[3]; - - /* clip to view volume */ - if (ctx->Transform.RasterPositionUnclipped) { - /* GL_IBM_rasterpos_clip: only clip against Z */ - if (viewclip_point_z(clipPos) == 0) { - ctx->Current.RasterPosValid = GL_FALSE; - return; - } - } - else if (viewclip_point(clipPos) == 0) { - /* Normal OpenGL behaviour */ - ctx->Current.RasterPosValid = GL_FALSE; - return; - } - - /* clip to user clipping planes */ - if (ctx->Transform.ClipPlanesEnabled && !userclip_point(ctx, clipPos)) { - ctx->Current.RasterPosValid = GL_FALSE; - return; - } + struct rastpos_stage *rs = rastpos_stage(stage); + GLcontext *ctx = rs->ctx; + struct st_context *st = ctx->st; + const GLfloat height = ctx->DrawBuffer->Height; + const GLuint *outputMapping = st->vertex_result_to_slot; + const GLfloat *pos; + GLuint i; + + /* if we get here, we didn't get clipped */ + ctx->Current.RasterPosValid = GL_TRUE; + /* update raster pos */ + pos = prim->v[0]->data[0]; + ctx->Current.RasterPos[0] = pos[0]; + ctx->Current.RasterPos[1] = height - 1 - pos[1]; + ctx->Current.RasterPos[2] = pos[2]; + ctx->Current.RasterPos[3] = pos[3]; - /* - * update current raster position - */ - /* ndc = clip / W */ - d = (clipPos[3] == 0.0F) ? 1.0F : 1.0F / clipPos[3]; - ndc[0] = clipPos[0] * d; - ndc[1] = clipPos[1] * d; - ndc[2] = clipPos[2] * d; - /* wincoord = viewport_mapping(ndc) */ - ctx->Current.RasterPos[0] = (ndc[0] * ctx->Viewport._WindowMap.m[MAT_SX] - + ctx->Viewport._WindowMap.m[MAT_TX]); - ctx->Current.RasterPos[1] = (ndc[1] * ctx->Viewport._WindowMap.m[MAT_SY] - + ctx->Viewport._WindowMap.m[MAT_TY]); - ctx->Current.RasterPos[2] = (ndc[2] * ctx->Viewport._WindowMap.m[MAT_SZ] - + ctx->Viewport._WindowMap.m[MAT_TZ]) - / ctx->DrawBuffer->_DepthMaxF; - ctx->Current.RasterPos[3] = clipPos[3]; - - /* compute raster distance */ -#if 0 - if (ctx->Fog.FogCoordinateSource == GL_FOG_COORDINATE_EXT) - ctx->Current.RasterDistance = ctx->Current.Attrib[VERT_ATTRIB_FOG][0]; - else { - /* XXX we don't have an eye coord! */ - ctx->Current.RasterDistance = - SQRTF( eye[0]*eye[0] + eye[1]*eye[1] + eye[2]*eye[2] ); - } -#else - ctx->Current.RasterDistance = fog[0]; -#endif - - /* colors and texcoords */ - COPY_4FV(ctx->Current.RasterColor, color0); - COPY_4FV(ctx->Current.RasterSecondaryColor, color1); - for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { - COPY_4FV(ctx->Current.RasterTexCoords + i, tex + i *4); - } + /* update other raster attribs */ + update_attrib(ctx, outputMapping, prim->v[0], + ctx->Current.RasterColor, + VERT_RESULT_COL0, VERT_ATTRIB_COLOR0); - ctx->Current.RasterPosValid = GL_TRUE; + update_attrib(ctx, outputMapping, prim->v[0], + ctx->Current.RasterSecondaryColor, + VERT_RESULT_COL1, VERT_ATTRIB_COLOR1); - if (ctx->RenderMode == GL_SELECT) { - _mesa_update_hitflag( ctx, ctx->Current.RasterPos[2] ); + for (i = 0; i < MAX_TEXTURE_UNITS; i++) { + update_attrib(ctx, outputMapping, prim->v[0], + ctx->Current.RasterTexCoords[i], + VERT_RESULT_TEX0 + i, VERT_ATTRIB_TEX0 + i); } } +/** + * Create rasterpos "drawing" stage. + */ +static struct rastpos_stage * +new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw) +{ + struct rastpos_stage *rs = CALLOC_STRUCT(rastpos_stage); + + rs->stage.draw = draw; + rs->stage.next = NULL; + rs->stage.begin = rastpos_begin; + rs->stage.point = rastpos_point; + rs->stage.line = rastpos_line; + rs->stage.tri = rastpos_tri; + rs->stage.end = rastpos_end; + rs->stage.reset_stipple_counter = rastpos_reset_stipple_counter; + rs->ctx = ctx; + + return rs; +} + static void st_RasterPos(GLcontext *ctx, const GLfloat v[4]) { - const struct st_context *st = ctx->st; - struct pipe_context *pipe = st->pipe; - float *buf_map; - struct pipe_feedback_buffer fb_buf; + struct st_context *st = ctx->st; + struct draw_context *draw = st->draw; + struct rastpos_stage *rs; - st_validate_state(ctx->st); - - /* setup vertex buffers */ - setup_vertex_attribs(ctx); - - /* - * Load the default attribute buffer with current attribs. - */ - { - struct pipe_buffer_handle *buf = st->default_attrib_buffer; - const unsigned size = sizeof(ctx->Current.Attrib); - const void *data = ctx->Current.Attrib; - /* colors, texcoords, etc */ - pipe->winsys->buffer_data(pipe->winsys, buf, - size, data, - PIPE_BUFFER_USAGE_VERTEX); - /* position */ - pipe->winsys->buffer_subdata(pipe->winsys, buf, - 0, /* offset */ - 4 * sizeof(float), /* size */ - v); /* data */ + if (st->rastpos_stage) { + /* get rastpos stage info */ + rs = rastpos_stage(st->rastpos_stage); } - - - /* setup feedback state */ - setup_feedback(ctx); - - /* setup vertex feedback buffer */ - { - fb_buf.size = 8 * 4 * sizeof(float); - fb_buf.buffer = pipe->winsys->buffer_create(pipe->winsys, 0); - fb_buf.start_offset = 0; - pipe->winsys->buffer_data(pipe->winsys, fb_buf.buffer, - fb_buf.size, - NULL, /* data */ - PIPE_BUFFER_USAGE_VERTEX); - if (pipe->set_feedback_buffer) - pipe->set_feedback_buffer(pipe, 0, &fb_buf); - } - - - /* draw a point */ - pipe->draw_arrays(pipe, GL_POINTS, 0, 1); - - /* get feedback */ - buf_map = (float *) pipe->winsys->buffer_map(pipe->winsys, fb_buf.buffer, - PIPE_BUFFER_FLAG_READ); - - /* extract values and update rasterpos state */ - { - const GLuint *outputMapping = st->vertex_result_to_slot; - const float *pos, *color0, *color1, *fog, *tex0; - float *buf = buf_map; - - assert(outputMapping[VERT_RESULT_HPOS] != ~0); - pos = buf; - buf += 4; - - if (outputMapping[VERT_RESULT_COL0] != ~0) { - color0 = buf; - buf += 4; - } - else { - color0 = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; + else { + /* create rastpos draw stage */ + GLuint i; + + rs = new_draw_rastpos_stage(ctx, draw); + st->rastpos_stage = &rs->stage; + + /* one-time init */ + for (i = 0; i < VERT_ATTRIB_MAX; i++) { + rs->array[i].Size = 4; + rs->array[i].Type = GL_FLOAT; + rs->array[i].Stride = 0; + rs->array[i].StrideB = 0; + rs->array[i].Ptr = (GLubyte *) ctx->Current.Attrib[i]; + rs->array[i].Enabled = GL_TRUE; + rs->array[i].Normalized = GL_TRUE; + rs->array[i].BufferObj = NULL; + rs->arrays[i] = &rs->array[i]; } - if (outputMapping[VERT_RESULT_COL1] != ~0) { - color1 = buf; - buf += 4; - } - else { - color1 = ctx->Current.Attrib[VERT_ATTRIB_COLOR1]; - } + rs->prim.mode = GL_POINTS; + rs->prim.indexed = 0; + rs->prim.begin = 1; + rs->prim.end = 1; + rs->prim.weak = 0; + rs->prim.start = 0; + rs->prim.count = 1; + } - if (outputMapping[VERT_RESULT_FOGC] != ~0) { - fog = buf; - buf += 4; - } - else { - fog = ctx->Current.Attrib[VERT_ATTRIB_FOG]; - } + /* plug our rastpos stage into the draw module */ + draw_set_rasterize_stage(st->draw, st->rastpos_stage); - if (outputMapping[VERT_RESULT_TEX0] != ~0) { - tex0 = buf; - buf += 4; - } - else { - tex0 = ctx->Current.Attrib[VERT_ATTRIB_TEX0]; - } + /* make sure everything's up to date */ + st_validate_state(ctx->st); - update_rasterpos(ctx, pos, color0, color1, fog, tex0); - } + /* This will get set only if rastpos_point(), above, gets called */ + ctx->Current.RasterPosValid = GL_FALSE; - /* free vertex feedback buffer */ - pipe->winsys->buffer_unmap(pipe->winsys, fb_buf.buffer); - pipe->winsys->buffer_reference(pipe->winsys, &fb_buf.buffer, NULL); + /* All vertex attribs but position were previously initialized above. + * Just plug in position pointer now. + */ + rs->array[0].Ptr = (GLubyte *) v; - /* restore pipe state */ - if (pipe->set_feedback_state) - pipe->set_feedback_state(pipe, &st->state.feedback); + /* draw the point */ + st_feedback_draw_vbo(ctx, rs->arrays, &rs->prim, 1, NULL, 0, 1); } + void st_init_rasterpos_functions(struct dd_function_table *functions) { functions->RasterPos = st_RasterPos; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 82e133911f..a6045230a0 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -68,9 +68,10 @@ struct st_context struct pipe_context *pipe; - struct draw_context *draw; /**< For selection/feedback */ + struct draw_context *draw; /**< For selection/feedback/rastpos only */ struct draw_stage *feedback_stage; /**< For GL_FEEDBACK rendermode */ struct draw_stage *selection_stage; /**< For GL_SELECT rendermode */ + struct draw_stage *rastpos_stage; /**< For glRasterPos */ /* Some state is contained in constant objects. * Other state is just parameter values. -- cgit v1.2.3 From 34a00276c7b2ee8ab88a56905352023f8a435a53 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 16 Nov 2007 17:13:26 -0700 Subject: more convenient debug code --- src/mesa/pipe/draw/draw_vertex_fetch.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index eca10e89fb..b510a4dbba 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -37,6 +37,10 @@ #include "pipe/tgsi/exec/tgsi_core.h" + +#define DBG 0 + + /** * Fetch a float[4] vertex attribute from memory, doing format/type * conversion as needed. @@ -96,7 +100,9 @@ void draw_vertex_fetch( struct draw_context *draw, for (j = 0; j < count; j++) { uint attr; - /*printf("fetch vertex %u: \n", j);*/ +#if DBG + printf("fetch vertex %u: \n", j); +#endif /* loop over vertex attributes (vertex shader inputs) */ for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { @@ -111,7 +117,9 @@ void draw_vertex_fetch( struct draw_context *draw, fetch_attrib4(src, draw->vertex_element[attr].src_format, p); - /*printf(" %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]);*/ +#if DBG + printf(" %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]); +#endif /* Transform to AoS xxxx/yyyy/zzzz/wwww representation: */ -- cgit v1.2.3 From dec60d33b2570cf2bdce72a00a1539ee93133f91 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 16 Nov 2007 17:36:35 +0000 Subject: Proper fence reference counting. --- .../drivers/dri/intel_winsys/intel_winsys_i915.c | 30 +++++++++++----------- src/mesa/pipe/i915simple/i915_batch.h | 2 +- src/mesa/pipe/i915simple/i915_context.c | 2 ++ src/mesa/pipe/i915simple/i915_winsys.h | 9 ++++++- 4 files changed, 26 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index 7713d6e48d..eea8c0be18 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -111,12 +111,19 @@ static void intel_i915_batch_reloc( struct i915_winsys *sws, -static struct pipe_fence * -intel_i915_batch_flush( struct i915_winsys *sws ) +static void +intel_i915_batch_flush( struct i915_winsys *sws, + struct pipe_fence **fence ) { struct intel_context *intel = intel_i915_winsys(sws)->intel; - - return pipe_fo(intel_batchbuffer_flush( intel->batch )); + struct pipe_fence *tmp_fence; + + tmp_fence = pipe_fo(intel_batchbuffer_flush( intel->batch )); + + /* this also increases the fence reference count, which is not done inside + * intel_batchbuffer_flush call above + */ + sws->fence_reference(sws, fence, tmp_fence); } @@ -143,14 +150,10 @@ intel_i915_fence_is_signalled( struct i915_winsys *sws, struct pipe_fence *fence ) { struct _DriFenceObject *dri_fence = dri_fo(fence); - int ret = 1; - if (fence) { - driFenceReference(dri_fence); - ret = driFenceSignaled(dri_fence, + if (fence) + return driFenceSignaled(dri_fence, DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW); - driFenceUnReference(dri_fence); - } - return ret; + return 1; } @@ -159,13 +162,10 @@ intel_i915_fence_wait( struct i915_winsys *sws, struct pipe_fence *fence ) { struct _DriFenceObject *dri_fence = dri_fo(fence); - if (fence) { - driFenceReference(dri_fence); + if (fence) driFenceFinish(dri_fence, DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); - driFenceUnReference(dri_fence); - } return 1; } diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h index 99b03c37db..603d193f62 100644 --- a/src/mesa/pipe/i915simple/i915_batch.h +++ b/src/mesa/pipe/i915simple/i915_batch.h @@ -46,7 +46,7 @@ #define FLUSH_BATCH() do { \ if (0) i915_dump_batchbuffer( i915 ); \ - i915->last_fence = i915->winsys->batch_flush( i915->winsys ); \ + i915->winsys->batch_flush( i915->winsys, &i915->last_fence ); \ i915->batch_start = NULL; \ i915->hardware_dirty = ~0; \ } while (0) diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index e43274dc66..f5d770ce0d 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -175,6 +175,8 @@ static void i915_destroy( struct pipe_context *pipe ) draw_destroy( i915->draw ); + i915->winsys->fence_reference( i915->winsys, &i915->last_fence, NULL ); + free( i915 ); } diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 03457c1d61..386ed745d3 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -98,7 +98,14 @@ struct i915_winsys { unsigned access_flags, unsigned delta ); - struct pipe_fence *(*batch_flush)( struct i915_winsys *sws ); + /** + * Flush the batch buffer. + * + * Fence argument must point to NULL or to a previous fence, and the caller + * must call fence_reference when done with the fence. + */ + void (*batch_flush)( struct i915_winsys *sws, + struct pipe_fence **fence ); /* Fence -- cgit v1.2.3 From ca7f68a7cf25a51f382bba8c42d8c6ab7db57b5d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 17 Nov 2007 15:06:01 +0000 Subject: Fix typo --- src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index eea8c0be18..d73b309fd4 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -133,7 +133,7 @@ intel_i915_fence_reference( struct i915_winsys *sws, struct pipe_fence *src_fence ) { struct _DriFenceObject **dri_dst_fence = (struct _DriFenceObject **)dst_fence; - struct _DriFenceObject *dri_src_fence = (struct _DriFenceObject *)dst_fence; + struct _DriFenceObject *dri_src_fence = (struct _DriFenceObject *)src_fence; if(dri_src_fence) driFenceReference(dri_src_fence); -- cgit v1.2.3 From 7f718f047676e88b660618784f256a96f7e8ed58 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 18 Nov 2007 18:20:20 +0000 Subject: Implement early depth test. Early depth test is enabled when depth test is enabled and alpha test is disabled and fragment shader does not write depth. The early-z is implemented by moving the depth test stage just before the fragment shader stage and prepending it with an earlyz stage, introduced with this commit. The earlyz stage prepares the quad->outputs.depth for the following depth test stage by interpolating Z position, just as the fragment shader would do. --- src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_context.c | 2 + src/mesa/pipe/softpipe/sp_context.h | 1 + src/mesa/pipe/softpipe/sp_quad.c | 70 ++++++++++++++-------- src/mesa/pipe/softpipe/sp_quad.h | 1 + src/mesa/pipe/softpipe/sp_quad_earlyz.c | 100 ++++++++++++++++++++++++++++++++ 6 files changed, 150 insertions(+), 25 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_quad_earlyz.c (limited to 'src') diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 2cbd9e5b51..59628531cc 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -18,6 +18,7 @@ DRIVER_SOURCES = \ sp_quad_colormask.c \ sp_quad_coverage.c \ sp_quad_depth_test.c \ + sp_quad_earlyz.c \ sp_quad_fs.c \ sp_quad_occlusion.c \ sp_quad_output.c \ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index be4da0ec64..d5e68c189d 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -161,6 +161,7 @@ static void softpipe_destroy( struct pipe_context *pipe ) draw_destroy( softpipe->draw ); softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); + softpipe->quad.earlyz->destroy( softpipe->quad.earlyz ); softpipe->quad.shade->destroy( softpipe->quad.shade ); softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); @@ -369,6 +370,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, /* setup quad rendering stages */ softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); + softpipe->quad.earlyz = sp_quad_earlyz_stage(softpipe); softpipe->quad.shade = sp_quad_shade_stage(softpipe); softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index a411969bc0..872766101d 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -134,6 +134,7 @@ struct softpipe_context { /** Software quad rendering pipeline */ struct { struct quad_stage *polygon_stipple; + struct quad_stage *earlyz; struct quad_stage *shade; struct quad_stage *alpha_test; struct quad_stage *stencil_test; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 429497e9b2..5a0df6de9d 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -28,24 +28,52 @@ #include "sp_context.h" +#include "sp_state.h" +#include "pipe/tgsi/exec/tgsi_token.h" +static void +sp_push_quad_first( + struct softpipe_context *sp, + struct quad_stage *quad ) +{ + quad->next = sp->quad.first; + sp->quad.first = quad; +} + +static void +sp_build_depth_stencil( + struct softpipe_context *sp ) +{ + if (sp->depth_stencil->stencil.front_enabled || + sp->depth_stencil->stencil.back_enabled) { + sp_push_quad_first( sp, sp->quad.stencil_test ); + } + else if (sp->depth_stencil->depth.enabled && + sp->framebuffer.zbuf) { + sp_push_quad_first( sp, sp->quad.depth_test ); + } +} void sp_build_quad_pipeline(struct softpipe_context *sp) { + boolean early_depth_test = + sp->depth_stencil->depth.enabled && + sp->framebuffer.zbuf && + !sp->alpha_test->enabled && + sp->fs->shader.output_semantic_name[0] != TGSI_SEMANTIC_POSITION; + /* build up the pipeline in reverse order... */ sp->quad.first = sp->quad.output; if (sp->blend->colormask != 0xf) { - sp->quad.colormask->next = sp->quad.first; - sp->quad.first = sp->quad.colormask; + sp_push_quad_first( sp, sp->quad.colormask ); } if (sp->blend->blend_enable || sp->blend->logicop_enable) { - sp->quad.blend->next = sp->quad.first; - sp->quad.first = sp->quad.blend; + sp_push_quad_first( sp, sp->quad.blend ); } if (sp->framebuffer.num_cbufs == 1) { @@ -54,46 +82,38 @@ sp_build_quad_pipeline(struct softpipe_context *sp) } else { /* insert bufloop stage */ - sp->quad.bufloop->next = sp->quad.first; - sp->quad.first = sp->quad.bufloop; + sp_push_quad_first( sp, sp->quad.bufloop ); } if (sp->depth_stencil->depth.occlusion_count) { - sp->quad.occlusion->next = sp->quad.first; - sp->quad.first = sp->quad.occlusion; + sp_push_quad_first( sp, sp->quad.occlusion ); } if (sp->rasterizer->poly_smooth || sp->rasterizer->line_smooth || sp->rasterizer->point_smooth) { - sp->quad.coverage->next = sp->quad.first; - sp->quad.first = sp->quad.coverage; + sp_push_quad_first( sp, sp->quad.coverage ); } - if ( sp->depth_stencil->stencil.front_enabled - || sp->depth_stencil->stencil.back_enabled) { - sp->quad.stencil_test->next = sp->quad.first; - sp->quad.first = sp->quad.stencil_test; - } - else if (sp->depth_stencil->depth.enabled && - sp->framebuffer.zbuf) { - sp->quad.depth_test->next = sp->quad.first; - sp->quad.first = sp->quad.depth_test; + if (!early_depth_test) { + sp_build_depth_stencil( sp ); } if (sp->alpha_test->enabled) { - sp->quad.alpha_test->next = sp->quad.first; - sp->quad.first = sp->quad.alpha_test; + sp_push_quad_first( sp, sp->quad.alpha_test ); } /* XXX always enable shader? */ if (1) { - sp->quad.shade->next = sp->quad.first; - sp->quad.first = sp->quad.shade; + sp_push_quad_first( sp, sp->quad.shade ); + } + + if (early_depth_test) { + sp_build_depth_stencil( sp ); + sp_push_quad_first( sp, sp->quad.earlyz ); } if (sp->rasterizer->poly_stipple_enable) { - sp->quad.polygon_stipple->next = sp->quad.first; - sp->quad.first = sp->quad.polygon_stipple; + sp_push_quad_first( sp, sp->quad.polygon_stipple ); } } diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h index 534541122b..f1e0281764 100644 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ b/src/mesa/pipe/softpipe/sp_quad.h @@ -51,6 +51,7 @@ struct quad_stage { struct quad_stage *sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_earlyz_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_quad_earlyz.c b/src/mesa/pipe/softpipe/sp_quad_earlyz.c new file mode 100644 index 0000000000..7e0b37519c --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_quad_earlyz.c @@ -0,0 +1,100 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Quad early-z testing + */ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" +#include "sp_tile_cache.h" + +static void +earlyz_quad( + struct quad_stage *qs, + struct quad_header *quad ) +{ + uint i; + const float fx = (float) quad->x0; + const float fy = (float) quad->y0; + float xy[4][2]; + + xy[0][0] = fx; + xy[1][0] = fx + 1.0f; + xy[2][0] = fx; + xy[3][0] = fx + 1.0f; + + xy[0][1] = fy; + xy[1][1] = fy; + xy[2][1] = fy + 1.0f; + xy[3][1] = fy + 1.0f; + + for (i = 0; i < QUAD_SIZE; i++) { + quad->outputs.depth[i] = + quad->coef[0].a0[2] + + quad->coef[0].dadx[2] * xy[i][0] + + quad->coef[0].dady[2] * xy[i][1]; + } + + if (qs->next) { + qs->next->run( qs->next, quad ); + } +} + +static void +earlyz_begin( + struct quad_stage *qs ) +{ + if (qs->next) { + qs->next->begin( qs->next ); + } +} + +static void +earlyz_destroy( + struct quad_stage *qs ) +{ + FREE( qs ); +} + +struct quad_stage * +sp_quad_earlyz_stage( + struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT( quad_stage ); + + stage->softpipe = softpipe; + stage->begin = earlyz_begin; + stage->run = earlyz_quad; + stage->destroy = earlyz_destroy; + + return stage; +} -- cgit v1.2.3 From 0204cbb4f17ef24a3c1ae1e426d5da3dd3744f92 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 19 Nov 2007 18:15:21 -0700 Subject: optimize linear_interpolation(), perspective_interpolation() functions --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 43 ++++++++++++++++--------------------- 1 file changed, 18 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 5c2fecb3e3..ea6c5021b3 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1346,17 +1346,15 @@ linear_interpolation( unsigned attrib, unsigned chan ) { - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = mach->Inputs[0].xyzw[0].f[i]; - const float y = mach->Inputs[0].xyzw[1].f[i]; - - mach->Inputs[attrib].xyzw[chan].f[i] = - mach->InterpCoefs[attrib].a0[chan] + - mach->InterpCoefs[attrib].dadx[chan] * x + - mach->InterpCoefs[attrib].dady[chan] * y; - } + const float x = mach->Inputs[0].xyzw[0].f[0]; + const float y = mach->Inputs[0].xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + mach->Inputs[attrib].xyzw[chan].f[0] = a0; + mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; + mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; + mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; } static void @@ -1365,20 +1363,15 @@ perspective_interpolation( unsigned attrib, unsigned chan ) { - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = mach->Inputs[0].xyzw[0].f[i]; - const float y = mach->Inputs[0].xyzw[1].f[i]; - /* WPOS.w here is really 1/w */ - const float w = 1.0f / mach->Inputs[0].xyzw[3].f[i]; - assert(mach->Inputs[0].xyzw[3].f[i] != 0.0); - - mach->Inputs[attrib].xyzw[chan].f[i] = - (mach->InterpCoefs[attrib].a0[chan] + - mach->InterpCoefs[attrib].dadx[chan] * x + - mach->InterpCoefs[attrib].dady[chan] * y) * w; - } + const float x = mach->Inputs[0].xyzw[0].f[0]; + const float y = mach->Inputs[0].xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + mach->Inputs[attrib].xyzw[chan].f[0] = a0 / mach->Inputs[0].xyzw[3].f[0]; + mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / mach->Inputs[0].xyzw[3].f[1]; + mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / mach->Inputs[0].xyzw[3].f[2]; + mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / mach->Inputs[0].xyzw[3].f[3]; } -- cgit v1.2.3 From c4f9fbb57f6e941a3b896b52c69665f7dced2350 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 19 Nov 2007 18:16:07 -0700 Subject: optimize earlyz_quad(), add comments, remove unneeded #includes --- src/mesa/pipe/softpipe/sp_quad_earlyz.c | 34 +++++++++++++-------------------- 1 file changed, 13 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_earlyz.c b/src/mesa/pipe/softpipe/sp_quad_earlyz.c index 7e0b37519c..89fab2dd35 100644 --- a/src/mesa/pipe/softpipe/sp_quad_earlyz.c +++ b/src/mesa/pipe/softpipe/sp_quad_earlyz.c @@ -31,38 +31,30 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" -#include "sp_context.h" #include "sp_headers.h" -#include "sp_surface.h" #include "sp_quad.h" -#include "sp_tile_cache.h" + +/** + * All this stage does is compute the quad's Z values (which is normally + * done by the shading stage). + * The next stage will do the actual depth test. + */ static void earlyz_quad( struct quad_stage *qs, struct quad_header *quad ) { - uint i; const float fx = (float) quad->x0; const float fy = (float) quad->y0; - float xy[4][2]; + const float dzdx = quad->coef[0].dadx[2]; + const float dzdy = quad->coef[0].dady[2]; + const float z0 = quad->coef[0].a0[2] + dzdx * fx + dzdy * fy; - xy[0][0] = fx; - xy[1][0] = fx + 1.0f; - xy[2][0] = fx; - xy[3][0] = fx + 1.0f; - - xy[0][1] = fy; - xy[1][1] = fy; - xy[2][1] = fy + 1.0f; - xy[3][1] = fy + 1.0f; - - for (i = 0; i < QUAD_SIZE; i++) { - quad->outputs.depth[i] = - quad->coef[0].a0[2] + - quad->coef[0].dadx[2] * xy[i][0] + - quad->coef[0].dady[2] * xy[i][1]; - } + quad->outputs.depth[0] = z0; + quad->outputs.depth[1] = z0 + dzdx; + quad->outputs.depth[2] = z0 + dzdy; + quad->outputs.depth[3] = z0 + dzdx + dzdy; if (qs->next) { qs->next->run( qs->next, quad ); -- cgit v1.2.3 From 0191570f024ba787799ca2bccd46549a8af74aa9 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 20 Nov 2007 08:30:10 -0700 Subject: initial support for PIPE_FORMAT_Z24_S8 --- src/mesa/pipe/p_format.h | 2 ++ src/mesa/state_tracker/st_cb_clear.c | 8 ++++---- src/mesa/state_tracker/st_cb_readpixels.c | 9 +++++++++ src/mesa/state_tracker/st_format.c | 9 +++++++++ 4 files changed, 24 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index d478a6c58d..f42f987c4e 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -171,6 +171,7 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) #define _PIPE_FORMAT_RRRG _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G ) #define _PIPE_FORMAT_Z000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_ZS00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) /** @@ -222,6 +223,7 @@ static INLINE uint pf_rev(pipe_format_ycbcr_t f) #define PIPE_FORMAT_U_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< uint Z/depth */ #define PIPE_FORMAT_F_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) /**< float Z/depth */ #define PIPE_FORMAT_S8_Z24 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil + 24-bit Z */ +#define PIPE_FORMAT_Z24_S8 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 24-bit Z + 8-bit stencil */ #define PIPE_FORMAT_U_S8 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil */ #define PIPE_FORMAT_R64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) #define PIPE_FORMAT_R64G64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 3e591170da..219a5afcbd 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -91,6 +91,8 @@ depth_value(GLuint pipeFormat, GLfloat value) return (uint) (value * 0xffffffff); case PIPE_FORMAT_S8_Z24: return (uint) (value * 0xffffff); + case PIPE_FORMAT_Z24_S8: + return ((uint) (value * 0xffffff)) << 8; default: assert(0); return 0; @@ -103,7 +105,7 @@ is_depth_stencil_format(GLuint pipeFormat) { switch (pipeFormat) { case PIPE_FORMAT_S8_Z24: - /*case PIPE_FORMAT_Z24_S8:*/ + case PIPE_FORMAT_Z24_S8: return GL_TRUE; default: return GL_FALSE; @@ -521,11 +523,9 @@ clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) case PIPE_FORMAT_S8_Z24: clearValue |= ctx->Stencil.Clear << 24; break; -#if 0 case PIPE_FORMAT_Z24_S8: - clearValue = (clearValue << 8) | clearVal; + clearValue |= clearValue | ctx->Stencil.Clear; break; -#endif default: assert(0); } diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index c347a0b688..2e7c01672a 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -100,6 +100,15 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } } break; + case PIPE_FORMAT_Z24_S8: + { + const uint *src = (uint *) stmap + srcY * ps->region->pitch + x; + GLint k; + for (k = 0; k < width; k++) { + values[k] = src[k] & 0xff; + } + } + break; default: assert(0); } diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index ddba650941..c6b5bc968f 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -320,6 +320,9 @@ default_depth_format( if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) { return PIPE_FORMAT_S8_Z24; } + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24_S8 )) { + return PIPE_FORMAT_Z24_S8; + } return PIPE_FORMAT_NONE; } @@ -486,6 +489,8 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_DEPTH_COMPONENT24: if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) return PIPE_FORMAT_S8_Z24; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24_S8 )) + return PIPE_FORMAT_Z24_S8; /* fall-through */ case GL_DEPTH_COMPONENT32: if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_Z32 )) @@ -503,12 +508,16 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, return PIPE_FORMAT_U_S8; if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) return PIPE_FORMAT_S8_Z24; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24_S8 )) + return PIPE_FORMAT_Z24_S8; return PIPE_FORMAT_NONE; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) return PIPE_FORMAT_S8_Z24; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24_S8 )) + return PIPE_FORMAT_Z24_S8; return PIPE_FORMAT_NONE; default: -- cgit v1.2.3 From 5a6017d496ccce94d7e3cf9a6cfe1db886dcc767 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 20 Nov 2007 08:36:06 -0700 Subject: add PIPE_FORMAT_Z24_S8 support to softpipe patsh --- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 24 +++++++++++++++++ src/mesa/pipe/softpipe/sp_quad_stencil.c | 16 ++++++++++++ src/mesa/pipe/softpipe/sp_surface.c | 40 +++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_tile_cache.c | 13 ++++++++++ 4 files changed, 93 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index f7e93af784..3318189621 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -119,6 +119,21 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } } break; + case PIPE_FORMAT_Z24_S8: + { + float scale = (float) ((1 << 24) - 1); + + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); + } + + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + bzzzz[j] = tile->data.depth32[y][x] >> 8; + } + } + break; default: assert(0); } @@ -210,6 +225,15 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) tile->data.depth32[y][x] = s8z24; } break; + case PIPE_FORMAT_Z24_S8: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + uint z24s8 = tile->data.depth32[y][x]; + z24s8 = (z24s8 & 0xff) | (bzzzz[j] << 24); + tile->data.depth32[y][x] = z24s8; + } + break; default: assert(0); } diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 831ad8bad0..0149b20f48 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -241,6 +241,13 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) stencilVals[j] = tile->data.depth32[y][x] >> 24; } break; + case PIPE_FORMAT_Z24_S8: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + stencilVals[j] = tile->data.depth32[y][x] & 0xff; + } + break; case PIPE_FORMAT_U_S8: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); @@ -300,6 +307,15 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) tile->data.depth32[y][x] = s8z24; } break; + case PIPE_FORMAT_Z24_S8: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + uint z24s8 = tile->data.depth32[y][x]; + z24s8 = (z24s8 & 0xffffff00) | stencilVals[j]; + tile->data.depth32[y][x] = z24s8; + } + break; case PIPE_FORMAT_U_S8: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index ccaf38de3e..b7c9d4f004 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -461,6 +461,40 @@ s8z24_get_tile(struct pipe_surface *ps, } +/*** PIPE_FORMAT_Z24_S8 ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +z24s8_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const uint *src + = ((const uint *) (ps->region->map + ps->offset)) + + y * ps->region->pitch + x; + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_Z24_S8); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = (float) (scale * (src[j] >> 8)); + } + src += ps->region->pitch; + p += 4 * w0; + } +} + + /** * Called via pipe->get_tex_surface() * XXX is this in the right place? @@ -601,6 +635,9 @@ softpipe_get_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_S8_Z24: s8z24_get_tile(ps, x, y, w, h, p); break; + case PIPE_FORMAT_Z24_S8: + z24s8_get_tile(ps, x, y, w, h, p); + break; default: assert(0); } @@ -645,6 +682,9 @@ softpipe_put_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_S8_Z24: /*s8z24_put_tile(ps, x, y, w, h, p);*/ break; + case PIPE_FORMAT_Z24_S8: + /*z24s8_put_tile(ps, x, y, w, h, p);*/ + break; default: assert(0); } diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 19c06323e1..ea0c8b8f91 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -166,6 +166,7 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, return; is_depth_stencil = (ps->format == PIPE_FORMAT_S8_Z24 || + ps->format == PIPE_FORMAT_Z24_S8 || ps->format == PIPE_FORMAT_U_Z16 || ps->format == PIPE_FORMAT_U_Z32 || ps->format == PIPE_FORMAT_U_S8); @@ -203,6 +204,7 @@ sp_get_cached_tile(struct softpipe_context *softpipe, struct pipe_surface *ps = tc->surface; boolean is_depth_stencil = (ps->format == PIPE_FORMAT_S8_Z24 || + ps->format == PIPE_FORMAT_Z24_S8 || ps->format == PIPE_FORMAT_U_Z16 || ps->format == PIPE_FORMAT_U_Z32 || ps->format == PIPE_FORMAT_U_S8); @@ -268,6 +270,17 @@ sp_get_cached_tile(struct softpipe_context *softpipe, } } break; + case PIPE_FORMAT_Z24_S8: + { + uint clear_val = ((uint) (tc->clear_value[0] * 0xffffff)) << 8; + clear_val |= ((uint) tc->clear_value[1]) & 0xff; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.depth32[i][j] = clear_val; + } + } + } + break; case PIPE_FORMAT_U_S8: { ubyte clear_val = (uint) tc->clear_value[0]; -- cgit v1.2.3 From fbe68bf6b286056bb03f44907a078918d04cbdfd Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 21 Nov 2007 15:40:20 -0700 Subject: Simplify draw module's vertex_info. No longer store the vertex header and clip pos info in the draw module's vertex_info. The vertex_info just describes the data[] elements. This simplifies the code in several places. --- src/mesa/pipe/draw/draw_clip.c | 14 +++----------- src/mesa/pipe/draw/draw_flatshade.c | 5 +++-- src/mesa/pipe/draw/draw_vertex.c | 26 ++++++++++++-------------- src/mesa/pipe/draw/draw_vertex.h | 1 - src/mesa/pipe/draw/draw_vertex_shader.c | 4 +--- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 4 +--- 6 files changed, 20 insertions(+), 34 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index bc62d422a4..e4c257a0ee 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -128,12 +128,8 @@ static void interp( const struct clipper *clip, /* Other attributes * Note: start at 1 to skip winpos (data[0]) since we just computed * it above. - * Subtract two from nr_attrs since the first two attribs (always - * VF_ATTRIB_VERTEX_HEADER and VF_ATTRIB_CLIP_POS, see - * draw_set_vertex_attributes()) are in the vertex_header struct, - * not in the data[] array. */ - for (j = 1; j < nr_attrs - 2; j++) { + for (j = 1; j < nr_attrs; j++) { interp_attr(dst->data[j], t, in->data[j], out->data[j]); } } @@ -352,12 +348,8 @@ do_clip_line( struct draw_stage *stage, static void clip_begin( struct draw_stage *stage ) { - /* sanity checks. If these fail, review the clip/interp code! */ - assert(stage->draw->vertex_info.num_attribs >= 3); -#if 0 - assert(stage->draw->vertex_info.slot_to_attrib[0] == TGSI_ATTRIB_VERTEX_HEADER); - assert(stage->draw->vertex_info.slot_to_attrib[1] == TGSI_ATTRIB_CLIP_POS); -#endif + /* should always have position, at least */ + assert(stage->draw->vertex_info.num_attribs >= 1); stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 3b22c01b34..d46e53f2be 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -60,8 +60,9 @@ static INLINE void copy_colors( struct draw_stage *stage, uint i; /* Look for constant/flat attribs and duplicate from src to dst vertex */ - for (i = 1; i < num_attribs - 2; i++) { - if (interp[i + 2] == INTERP_CONSTANT) { + /* skip attrib[0] which is vert pos */ + for (i = 1; i < num_attribs; i++) { + if (interp[i] == INTERP_CONSTANT) { copy_attr( i, dst, src ); } } diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index dea26a3d0a..983ed71ec0 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -75,7 +75,6 @@ draw_compute_vertex_size(struct vertex_info *vinfo) vinfo->size += 3; break; case FORMAT_4F: - case FORMAT_4F_VIEWPORT: vinfo->size += 4; break; default: @@ -99,27 +98,26 @@ draw_set_vertex_attributes( struct draw_context *draw, struct vertex_info *vinfo = &draw->vertex_info; unsigned i; -#if 0 - assert(slot_to_vf_attr[0] == TGSI_ATTRIB_POS); -#endif + assert(interps[0] == INTERP_LINEAR); /* should be vert pos */ - memset(vinfo, 0, sizeof(*vinfo)); + assert(nr_attrs <= PIPE_MAX_SHADER_OUTPUTS); - /* - * First three attribs are always the same: header, clip pos, winpos + /* Note that draw-module vertices will consist of the attributes passed + * to this function, plus a header/prefix containing the vertex header + * flags and GLfloat[4] clip pos. */ - emit_vertex_attr(vinfo, FORMAT_1F, INTERP_NONE); - emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR); - emit_vertex_attr(vinfo, FORMAT_4F_VIEWPORT, INTERP_LINEAR); - /* - * Remaining attribs (color, texcoords, etc) - */ - for (i = 1; i < nr_attrs; i++) { + memset(vinfo, 0, sizeof(*vinfo)); + + /* copy attrib info */ + for (i = 0; i < nr_attrs; i++) { emit_vertex_attr(vinfo, FORMAT_4F, interps[i]); } draw_compute_vertex_size(vinfo); + + /* add extra words for vertex header (uint), clip pos (float[4]) */ + vinfo->size += 5; } diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index a1fa7aae5a..d9b5e7c8c0 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -46,7 +46,6 @@ enum attrib_format { FORMAT_2F, FORMAT_3F, FORMAT_4F, - FORMAT_4F_VIEWPORT, FORMAT_4UB }; diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index e8801addac..52fb2d8596 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -158,10 +158,8 @@ run_vertex_program(struct draw_context *draw, #endif /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. - * Skip 0 since we just did it above. - * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. */ - for (slot = 1; slot < draw->vertex_info.num_attribs - 2; slot++) { + for (slot = 1; slot < draw->vertex_info.num_attribs; slot++) { vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 3e005a76f3..53a1776ffc 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -173,10 +173,8 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. - * Skip 0 since we just did it above. - * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. */ - for (slot = 1; slot < draw->vertex_info.num_attribs - 2; slot++) { + for (slot = 1; slot < draw->vertex_info.num_attribs; slot++) { vOut->data[slot][0] = dests[slot][0]; vOut->data[slot][1] = dests[slot][1]; vOut->data[slot][2] = dests[slot][2]; -- cgit v1.2.3 From 9f0b5bba707d6c36896b4b8afad4e6b459da5e99 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 21 Nov 2007 16:00:57 -0700 Subject: Replace draw_set_vertex_attributes() with simpler draw_set_vertex_info(). Just pass in the vertex_info object and make a copy of it. --- src/mesa/pipe/draw/draw_vertex.c | 35 ++++++++++++--------------- src/mesa/pipe/draw/draw_vertex.h | 6 ++--- src/mesa/pipe/i915simple/i915_state_derived.c | 7 ++---- src/mesa/pipe/softpipe/sp_state_derived.c | 5 +--- src/mesa/state_tracker/st_draw.c | 21 ++++++++-------- 5 files changed, 31 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index 983ed71ec0..89f54ff186 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -87,37 +87,34 @@ draw_compute_vertex_size(struct vertex_info *vinfo) /** - * Tell the drawing module about the layout of post-transformation vertices + * Tell the drawing module about the contents of post-transformation vertices. + * Note that the vertex attribute format info isn't used by 'draw'; all + * attributes are handled as float[4]. But when the driver emits vertices + * it'll use that info. + * We _do_ care about the number of attributes and their interpolation modes. */ void -draw_set_vertex_attributes( struct draw_context *draw, - const uint *slot_to_vf_attr, - const enum interp_mode *interps, - unsigned nr_attrs ) +draw_set_vertex_info( struct draw_context *draw, + const struct vertex_info *info) { - struct vertex_info *vinfo = &draw->vertex_info; - unsigned i; + assert(info->interp_mode[0] == INTERP_LINEAR); /* should be vert pos */ - assert(interps[0] == INTERP_LINEAR); /* should be vert pos */ - - assert(nr_attrs <= PIPE_MAX_SHADER_OUTPUTS); + assert(info->num_attribs <= PIPE_MAX_SHADER_OUTPUTS); /* Note that draw-module vertices will consist of the attributes passed * to this function, plus a header/prefix containing the vertex header * flags and GLfloat[4] clip pos. */ - memset(vinfo, 0, sizeof(*vinfo)); - - /* copy attrib info */ - for (i = 0; i < nr_attrs; i++) { - emit_vertex_attr(vinfo, FORMAT_4F, interps[i]); - } + memcpy(&draw->vertex_info, info, sizeof(*info)); - draw_compute_vertex_size(vinfo); + draw_compute_vertex_size(&draw->vertex_info); - /* add extra words for vertex header (uint), clip pos (float[4]) */ - vinfo->size += 5; + /* Need to know vertex size (in words) for vertex copying elsewhere. + * Four words per attribute, plus vertex header (uint) and clip + * position (float[4]). + */ + draw->vertex_info.size = draw->vertex_info.num_attribs * 4 + 5; } diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index d9b5e7c8c0..8bb328affa 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -92,10 +92,8 @@ draw_emit_vertex_attr(struct vertex_info *vinfo, } -extern void draw_set_vertex_attributes( struct draw_context *draw, - const uint *attrs, - const enum interp_mode *interps, - unsigned nr_attrs ); +extern void draw_set_vertex_info( struct draw_context *draw, + const struct vertex_info *info); extern void draw_set_twoside_attributes(struct draw_context *draw, uint front0, uint back0, diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index ed1521fcce..688c0c5798 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -132,11 +132,8 @@ static void calculate_vertex_layout( struct i915_context *i915 ) /* If the attributes have changed, tell the draw module about the new * vertex layout. We'll also update the hardware vertex format info. */ - draw_set_vertex_attributes( i915->draw, - NULL,/*vinfo.slot_to_attrib,*/ - vinfo.interp_mode, - vinfo.num_attribs); - + draw_set_vertex_info( i915->draw, &vinfo); + draw_set_twoside_attributes(i915->draw, front0, back0, front1, back1); diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 81f70f0f24..7e5efbfde3 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -156,10 +156,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) if (1/*vinfo->attr_mask != softpipe->attr_mask*/) { /*softpipe->attr_mask = vinfo->attr_mask;*/ - draw_set_vertex_attributes( softpipe->draw, - NULL,/*vinfo->slot_to_attrib,*/ - vinfo->interp_mode, - vinfo->num_attribs); + draw_set_vertex_info( softpipe->draw, vinfo); draw_set_twoside_attributes(softpipe->draw, front0, back0, front1, back1); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 3b6d829145..61ff034550 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -324,27 +324,26 @@ static void set_feedback_vertex_format(GLcontext *ctx) { struct st_context *st = ctx->st; - uint attrs[PIPE_MAX_SHADER_OUTPUTS]; - enum interp_mode interp[PIPE_MAX_SHADER_OUTPUTS]; - GLuint n, i; + struct vertex_info vinfo; + GLuint i; if (ctx->RenderMode == GL_SELECT) { assert(ctx->RenderMode == GL_SELECT); - n = 1; - attrs[0] = FORMAT_4F; - interp[0] = INTERP_NONE; + vinfo.num_attribs = 1; + vinfo.format[0] = FORMAT_4F; + vinfo.interp_mode[0] = INTERP_NONE; } else { /* GL_FEEDBACK, or glRasterPos */ /* emit all attribs (pos, color, texcoord) as GLfloat[4] */ - n = st->state.vs->state.num_outputs; - for (i = 0; i < n; i++) { - attrs[i] = FORMAT_4F; - interp[i] = INTERP_NONE; + vinfo.num_attribs = st->state.vs->state.num_outputs; + for (i = 0; i < vinfo.num_attribs; i++) { + vinfo.format[i] = FORMAT_4F; + vinfo.interp_mode[i] = INTERP_LINEAR; } } - draw_set_vertex_attributes(st->draw, attrs, interp, n); + draw_set_vertex_info(st->draw, &vinfo); } -- cgit v1.2.3 From 2112191d452bd76d99ca48f8da17bb49eca595aa Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 21 Nov 2007 16:03:16 -0700 Subject: more simplification, clean-up in draw_set_vertex_info() --- src/mesa/pipe/draw/draw_vertex.c | 8 -------- 1 file changed, 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index 89f54ff186..a1926d951a 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -98,18 +98,10 @@ draw_set_vertex_info( struct draw_context *draw, const struct vertex_info *info) { assert(info->interp_mode[0] == INTERP_LINEAR); /* should be vert pos */ - assert(info->num_attribs <= PIPE_MAX_SHADER_OUTPUTS); - /* Note that draw-module vertices will consist of the attributes passed - * to this function, plus a header/prefix containing the vertex header - * flags and GLfloat[4] clip pos. - */ - memcpy(&draw->vertex_info, info, sizeof(*info)); - draw_compute_vertex_size(&draw->vertex_info); - /* Need to know vertex size (in words) for vertex copying elsewhere. * Four words per attribute, plus vertex header (uint) and clip * position (float[4]). -- cgit v1.2.3 From 4541ee5343df7c3ca937e088a85ec3f62970d318 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 22 Nov 2007 10:56:09 +0000 Subject: Remove fences from the i915simple winsys interface. Fences will be part of the pipe winsys interface, so remove this to avoid merge conflicts later on. This reverts commit ca7f68a7cf25a51f382bba8c42d8c6ab7db57b5d. This reverts commit dec60d33b2570cf2bdce72a00a1539ee93133f91. This reverts commit 90dd0cb822f2fe14258c786e5c37da69472b7d17. --- .../drivers/dri/intel_winsys/intel_batchbuffer.c | 14 +++++ .../drivers/dri/intel_winsys/intel_batchbuffer.h | 2 + src/mesa/drivers/dri/intel_winsys/intel_winsys.h | 17 ------ .../drivers/dri/intel_winsys/intel_winsys_i915.c | 60 +++------------------- src/mesa/pipe/i915simple/i915_batch.h | 10 ++-- src/mesa/pipe/i915simple/i915_context.c | 2 - src/mesa/pipe/i915simple/i915_context.h | 3 -- src/mesa/pipe/i915simple/i915_flush.c | 3 +- src/mesa/pipe/i915simple/i915_winsys.h | 25 +-------- 9 files changed, 31 insertions(+), 105 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c index ed223977e2..49e04d81ec 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c @@ -307,6 +307,20 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch) } +void +intel_batchbuffer_finish(struct intel_batchbuffer *batch) +{ + struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); + if (fence) { + driFenceReference(fence); + driFenceFinish(fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, + GL_FALSE); + driFenceUnReference(fence); + } +} + + /* This is the only way buffers get added to the validate list. */ boolean diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h index 2c943e68e5..82feafa21f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h @@ -72,6 +72,8 @@ struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context *intel); void intel_batchbuffer_free(struct intel_batchbuffer *batch); +void intel_batchbuffer_finish(struct intel_batchbuffer *batch); + struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer *batch); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h index f944cd23c3..89e63e0a79 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h @@ -32,9 +32,7 @@ struct intel_context; struct pipe_context; struct pipe_winsys; struct pipe_buffer_handle; -struct pipe_fence; struct _DriBufferObject; -struct _DriFenceObject; struct pipe_winsys * intel_create_pipe_winsys( int fd ); @@ -68,20 +66,5 @@ pipe_bo( struct _DriBufferObject *bo ) } -/* Turn the pipe opaque buffer pointer into a dri_bufmgr opaque - * buffer pointer... - */ -static INLINE struct _DriFenceObject * -dri_fo( struct pipe_fence *bo ) -{ - return (struct _DriFenceObject *)bo; -} - -static INLINE struct pipe_fence * -pipe_fo( struct _DriFenceObject *bo ) -{ - return (struct pipe_fence *)bo; -} - #endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index d73b309fd4..8e0eea4392 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -111,62 +111,19 @@ static void intel_i915_batch_reloc( struct i915_winsys *sws, -static void -intel_i915_batch_flush( struct i915_winsys *sws, - struct pipe_fence **fence ) +static void intel_i915_batch_flush( struct i915_winsys *sws ) { struct intel_context *intel = intel_i915_winsys(sws)->intel; - struct pipe_fence *tmp_fence; - - tmp_fence = pipe_fo(intel_batchbuffer_flush( intel->batch )); - - /* this also increases the fence reference count, which is not done inside - * intel_batchbuffer_flush call above - */ - sws->fence_reference(sws, fence, tmp_fence); -} - -static void -intel_i915_fence_reference( struct i915_winsys *sws, - struct pipe_fence **dst_fence, - struct pipe_fence *src_fence ) -{ - struct _DriFenceObject **dri_dst_fence = (struct _DriFenceObject **)dst_fence; - struct _DriFenceObject *dri_src_fence = (struct _DriFenceObject *)src_fence; - - if(dri_src_fence) - driFenceReference(dri_src_fence); - - if(*dri_dst_fence) - driFenceUnReference(*dri_dst_fence); - - *dri_dst_fence = dri_src_fence; + intel_batchbuffer_flush( intel->batch ); +// if (0) intel_i915_batch_wait_idle( sws ); } -static int -intel_i915_fence_is_signalled( struct i915_winsys *sws, - struct pipe_fence *fence ) +static void intel_i915_batch_finish( struct i915_winsys *sws ) { - struct _DriFenceObject *dri_fence = dri_fo(fence); - if (fence) - return driFenceSignaled(dri_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW); - return 1; -} - - -static int -intel_i915_fence_wait( struct i915_winsys *sws, - struct pipe_fence *fence ) -{ - struct _DriFenceObject *dri_fence = dri_fo(fence); - if (fence) - driFenceFinish(dri_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, - GL_FALSE); - return 1; + struct intel_context *intel = intel_i915_winsys(sws)->intel; + intel_batchbuffer_finish( intel->batch ); } @@ -186,10 +143,7 @@ intel_create_i915simple( struct intel_context *intel, iws->winsys.batch_dword = intel_i915_batch_dword; iws->winsys.batch_reloc = intel_i915_batch_reloc; iws->winsys.batch_flush = intel_i915_batch_flush; - iws->winsys.fence_reference = intel_i915_fence_reference; - iws->winsys.fence_is_signalled = intel_i915_fence_is_signalled; - iws->winsys.fence_wait = intel_i915_fence_wait; - + iws->winsys.batch_finish = intel_i915_batch_finish; iws->intel = intel; /* Create the i915simple context: diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h index 603d193f62..fb88cd6db0 100644 --- a/src/mesa/pipe/i915simple/i915_batch.h +++ b/src/mesa/pipe/i915simple/i915_batch.h @@ -44,11 +44,11 @@ #define ADVANCE_BATCH() -#define FLUSH_BATCH() do { \ - if (0) i915_dump_batchbuffer( i915 ); \ - i915->winsys->batch_flush( i915->winsys, &i915->last_fence ); \ - i915->batch_start = NULL; \ - i915->hardware_dirty = ~0; \ +#define FLUSH_BATCH() do { \ + if (0) i915_dump_batchbuffer( i915 ); \ + i915->winsys->batch_flush( i915->winsys ); \ + i915->batch_start = NULL; \ + i915->hardware_dirty = ~0; \ } while (0) #endif diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index f5d770ce0d..e43274dc66 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -175,8 +175,6 @@ static void i915_destroy( struct pipe_context *pipe ) draw_destroy( i915->draw ); - i915->winsys->fence_reference( i915->winsys, &i915->last_fence, NULL ); - free( i915 ); } diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 488682f852..ee430ebc90 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -80,7 +80,6 @@ -struct pipe_fence; struct i915_cache_context; /* Use to calculate differences between state emitted to hardware and @@ -185,8 +184,6 @@ struct i915_context unsigned dirty; unsigned *batch_start; - - struct pipe_fence *last_fence; /** Vertex buffer */ struct pipe_buffer_handle *vbo; diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c index 1044cb1bdb..3c2069b827 100644 --- a/src/mesa/pipe/i915simple/i915_flush.c +++ b/src/mesa/pipe/i915simple/i915_flush.c @@ -69,8 +69,7 @@ static void i915_flush( struct pipe_context *pipe, FLUSH_BATCH(); if (flags & PIPE_FLUSH_WAIT) { - if( i915->last_fence ) - i915->winsys->fence_wait(i915->winsys, i915->last_fence); + i915->winsys->batch_finish(i915->winsys); } } diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 386ed745d3..2c0f335d34 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -52,7 +52,6 @@ struct pipe_buffer_handle; struct pipe_winsys; -struct pipe_fence; /** @@ -98,28 +97,8 @@ struct i915_winsys { unsigned access_flags, unsigned delta ); - /** - * Flush the batch buffer. - * - * Fence argument must point to NULL or to a previous fence, and the caller - * must call fence_reference when done with the fence. - */ - void (*batch_flush)( struct i915_winsys *sws, - struct pipe_fence **fence ); - - - /* Fence - */ - void (*fence_reference)( struct i915_winsys *sws, - struct pipe_fence **dst_fence, - struct pipe_fence *src_fence ); - - int (*fence_is_signalled)( struct i915_winsys *sws, - struct pipe_fence *fence ); - - int (*fence_wait)( struct i915_winsys *sws, - struct pipe_fence *fence ); - + void (*batch_flush)( struct i915_winsys *sws ); + void (*batch_finish)( struct i915_winsys *sws ); }; #define I915_BUFFER_ACCESS_WRITE 0x1 -- cgit v1.2.3 From 5961732c1b59403b4e736fa354a64d4a0e5d8af2 Mon Sep 17 00:00:00 2001 From: Michal Date: Sat, 17 Nov 2007 14:26:24 +0000 Subject: Make it compile under Win32. --- src/mesa/pipe/i915simple/i915_blit.c | 8 +++--- src/mesa/pipe/i915simple/i915_context.c | 4 +-- src/mesa/pipe/i915simple/i915_debug.c | 9 +++---- src/mesa/pipe/i915simple/i915_debug.h | 4 +-- src/mesa/pipe/i915simple/i915_debug_fp.c | 2 +- src/mesa/pipe/i915simple/i915_fpc_emit.c | 6 ++--- src/mesa/pipe/i915simple/i915_fpc_translate.c | 30 +++++++++++----------- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 4 +-- src/mesa/pipe/i915simple/i915_regions.c | 14 +++++----- src/mesa/pipe/i915simple/i915_state.c | 28 ++++++++++---------- src/mesa/pipe/i915simple/i915_state_dynamic.c | 34 ++++++++++++------------- src/mesa/pipe/i915simple/i915_state_immediate.c | 24 ++++++++--------- src/mesa/pipe/i915simple/i915_surface.c | 2 +- src/mesa/pipe/i915simple/i915_tex_layout.c | 10 ++++---- 14 files changed, 89 insertions(+), 90 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/mesa/pipe/i915simple/i915_blit.c index a1f953ebbe..6e95313a0d 100644 --- a/src/mesa/pipe/i915simple/i915_blit.c +++ b/src/mesa/pipe/i915simple/i915_blit.c @@ -47,7 +47,7 @@ i915_fill_blit(struct i915_context *i915, unsigned BR13, CMD; BATCH_LOCALS; - dst_pitch *= cpp; + dst_pitch *= (short) cpp; switch (cpp) { case 1: @@ -102,14 +102,14 @@ i915_copy_blit( struct i915_context *i915, BATCH_LOCALS; - DBG(i915, + I915_DBG(i915, "%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", __FUNCTION__, src_buffer, src_pitch, src_offset, src_x, src_y, dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); - src_pitch *= cpp; - dst_pitch *= cpp; + src_pitch *= (short) cpp; + dst_pitch *= (short) cpp; switch (cpp) { case 1: diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index e43274dc66..a0ed3032b1 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -175,7 +175,7 @@ static void i915_destroy( struct pipe_context *pipe ) draw_destroy( i915->draw ); - free( i915 ); + FREE( i915 ); } @@ -341,7 +341,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, */ i915->draw = draw_create(); assert(i915->draw); - if (getenv("I915_VBUF")) { + if (GETENV("I915_VBUF")) { draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915)); } else { diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index d07d2f1fa9..001f695e2b 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -498,7 +498,7 @@ static boolean debug_map_state( struct debug_stream *stream, unsigned len ) { unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - int j = 0; + unsigned j = 0; PRINTF("%s (%d dwords):\n", name, len); PRINTF("\t0x%08x\n", ptr[j++]); @@ -550,7 +550,7 @@ static boolean debug_sampler_state( struct debug_stream *stream, unsigned len ) { unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - int j = 0; + unsigned j = 0; PRINTF("%s (%d dwords):\n", name, len); PRINTF("\t0x%08x\n", ptr[j++]); @@ -827,7 +827,7 @@ i915_dump_batchbuffer( struct i915_context *i915 ) struct debug_stream stream; unsigned *start = i915->batch_start; unsigned *end = i915->winsys->batch_start( i915->winsys, 0, 0 ); - unsigned bytes = (end - start) * 4; + unsigned long bytes = (unsigned long) (end - start) * 4; boolean done = FALSE; stream.offset = 0; @@ -843,8 +843,7 @@ i915_dump_batchbuffer( struct i915_context *i915 ) stream.winsys->printf( stream.winsys, "\n\nBATCH: (%d)\n", bytes / 4); while (!done && - stream.offset < bytes && - stream.offset >= 0) + stream.offset < bytes) { if (!i915_debug_packet( &stream )) break; diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h index 4c3aa64b42..356c751a5a 100644 --- a/src/mesa/pipe/i915simple/i915_debug.h +++ b/src/mesa/pipe/i915simple/i915_debug.h @@ -69,12 +69,12 @@ void i915_print_ureg(const char *msg, unsigned ureg); #ifdef DEBUG #include "pipe/p_winsys.h" -#define DBG( i915, ... ) do { \ +#define I915_DBG( i915, ... ) do { \ if ((i915)->debug & FILE_DEBUG_FLAG) \ (i915)->pipe.winsys->printf( (i915)->pipe.winsys, __VA_ARGS__ ); \ } while(0) #else -#define DBG( i915, ... ) \ +#define I915_DBG( i915, ... ) \ (void)i915 #endif diff --git a/src/mesa/pipe/i915simple/i915_debug_fp.c b/src/mesa/pipe/i915simple/i915_debug_fp.c index 87fc3b2f9a..ec6b0cbf19 100644 --- a/src/mesa/pipe/i915simple/i915_debug_fp.c +++ b/src/mesa/pipe/i915simple/i915_debug_fp.c @@ -327,7 +327,7 @@ i915_disassemble_program(struct debug_stream *stream, const unsigned * program, unsigned sz) { unsigned size = program[0] & 0x1ff; - int i; + unsigned i; PRINTF("\t\tBEGIN\n"); diff --git a/src/mesa/pipe/i915simple/i915_fpc_emit.c b/src/mesa/pipe/i915simple/i915_fpc_emit.c index c8d36435d9..74924ff0a1 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_emit.c +++ b/src/mesa/pipe/i915simple/i915_fpc_emit.c @@ -235,7 +235,7 @@ uint i915_emit_texld( struct i915_fp_compile *p, uint i915_emit_const1f(struct i915_fp_compile * p, float c0) { - int reg, idx; + unsigned reg, idx; if (c0 == 0.0) return swizzle(UREG(REG_TYPE_R, 0), ZERO, ZERO, ZERO, ZERO); @@ -264,7 +264,7 @@ i915_emit_const1f(struct i915_fp_compile * p, float c0) uint i915_emit_const2f(struct i915_fp_compile * p, float c0, float c1) { - int reg, idx; + unsigned reg, idx; if (c0 == 0.0) return swizzle(i915_emit_const1f(p, c1), ZERO, X, Z, W); @@ -302,7 +302,7 @@ uint i915_emit_const4f(struct i915_fp_compile * p, float c0, float c1, float c2, float c3) { - int reg; + unsigned reg; for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { if (p->constant_flags[reg] == 0xf && diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index d74da85216..f9673e5424 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -69,16 +69,16 @@ static unsigned passthrough[] = /* 1, -1/3!, 1/5!, -1/7! */ static const float sin_constants[4] = { 1.0, - -1.0 / (3 * 2 * 1), - 1.0 / (5 * 4 * 3 * 2 * 1), - -1.0 / (7 * 6 * 5 * 4 * 3 * 2 * 1) + -1.0f / (3 * 2 * 1), + 1.0f / (5 * 4 * 3 * 2 * 1), + -1.0f / (7 * 6 * 5 * 4 * 3 * 2 * 1) }; /* 1, -1/2!, 1/4!, -1/6! */ static const float cos_constants[4] = { 1.0, - -1.0 / (2 * 1), - 1.0 / (4 * 3 * 2 * 1), - -1.0 / (6 * 5 * 4 * 3 * 2 * 1) + -1.0f / (2 * 1), + 1.0f / (4 * 3 * 2 * 1), + -1.0f / (6 * 5 * 4 * 3 * 2 * 1) }; @@ -102,7 +102,7 @@ i915_use_passthrough_shader(struct i915_context *i915) { fprintf(stderr, "**** Using i915 pass-through fragment shader\n"); - i915->current.program = (uint *) malloc(sizeof(passthrough)); + i915->current.program = (uint *) MALLOC(sizeof(passthrough)); if (i915->current.program) { memcpy(i915->current.program, passthrough, sizeof(passthrough)); i915->current.program_len = Elements(passthrough); @@ -167,7 +167,7 @@ src_vector(struct i915_fp_compile *p, switch (sem_name) { case TGSI_SEMANTIC_POSITION: - printf("SKIP SEM POS\n"); + fprintf(stderr, "SKIP SEM POS\n"); /* assert(p->wpos_tex != -1); src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); @@ -430,7 +430,7 @@ i915_translate_instruction(struct i915_fp_compile *p, i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - src0, i915_emit_const1f(p, 1.0 / (M_PI * 2)), 0); + src0, i915_emit_const1f(p, 1.0f / (M_PI * 2.0f)), 0); i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); @@ -439,7 +439,7 @@ i915_translate_instruction(struct i915_fp_compile *p, i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - tmp, i915_emit_const1f(p, (M_PI * 2)), 0); + tmp, i915_emit_const1f(p, (M_PI * 2.0f)), 0); /* * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 @@ -986,8 +986,8 @@ i915_init_compile(struct i915_context *i915, static void i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) { - uint program_size = p->csr - p->program; - uint decl_size = p->decl - p->declarations; + unsigned long program_size = (unsigned long) (p->csr - p->program); + unsigned long decl_size = (unsigned long) (p->decl - p->declarations); if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT) i915_program_error(p, "Exceeded max nr indirect texture lookups"); @@ -1003,7 +1003,7 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) /* free old program, if present */ if (i915->current.program) { - free(i915->current.program); + FREE(i915->current.program); i915->current.program_len = 0; } @@ -1028,7 +1028,7 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) /* Copy compilation results to fragment program struct: */ i915->current.program - = (uint *) malloc((program_size + decl_size) * sizeof(uint)); + = (uint *) MALLOC((program_size + decl_size) * sizeof(uint)); if (i915->current.program) { i915->current.program_len = program_size + decl_size; @@ -1049,7 +1049,7 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) /* Release the compilation struct: */ - free(p); + FREE(p); } diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 35174c64a0..1c8c6a37ba 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -101,7 +101,7 @@ static INLINE struct vbuf_stage *vbuf_stage( struct draw_stage *stage ) static INLINE boolean overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) { - unsigned long used = (char *)ptr - (char *)map; + unsigned long used = (unsigned long) ((char *)ptr - (char *)map); return (used + bytes) > bufsz; } @@ -438,7 +438,7 @@ struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) assert(IBUF_SIZE < UNDEFINED_VERTEX_ID); /* FIXME: free this memory on takedown */ - vbuf->element_map = malloc( IBUF_SIZE ); + vbuf->element_map = MALLOC( IBUF_SIZE ); vbuf->vertex_map = NULL; vbuf->vertex_ptr = vbuf->vertex_map; diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index 0410446326..e8c4c92bc8 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -163,9 +163,9 @@ i915_region_copy(struct pipe_context *pipe, else { i915_copy_blit( i915_context(pipe), dst->cpp, - src->pitch, src->buffer, src_offset, - dst->pitch, dst->buffer, dst_offset, - srcx, srcy, dstx, dsty, width, height ); + (short) src->pitch, src->buffer, src_offset, + (short) dst->pitch, dst->buffer, dst_offset, + (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); } } @@ -204,7 +204,7 @@ i915_region_fill(struct pipe_context *pipe, ushort *row = (ushort *) get_pointer(dst, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) - row[j] = value; + row[j] = (ushort) value; row += dst->pitch; } } @@ -226,10 +226,10 @@ i915_region_fill(struct pipe_context *pipe, else { i915_fill_blit( i915_context(pipe), dst->cpp, - dst->pitch, + (short) dst->pitch, dst->buffer, dst_offset, - dstx, dsty, - width, height, + (short) dstx, (short) dsty, + (short) width, (short) height, value ); } } diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 70b8195bf1..468d0ce91b 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -97,7 +97,7 @@ static void * i915_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - struct i915_blend_state *cso_data = calloc(1, sizeof(struct i915_blend_state)); + struct i915_blend_state *cso_data = CALLOC_STRUCT( i915_blend_state ); { unsigned eqRGB = blend->rgb_func; @@ -182,7 +182,7 @@ static void i915_bind_blend_state(struct pipe_context *pipe, static void i915_delete_blend_state(struct pipe_context *pipe, void *blend) { - free(blend); + FREE(blend); } static void i915_set_blend_color( struct pipe_context *pipe, @@ -199,15 +199,15 @@ static void * i915_create_sampler_state(struct pipe_context *pipe, const struct pipe_sampler_state *sampler) { - struct i915_sampler_state *cso = calloc(1, sizeof(struct i915_sampler_state)); - cso->templ = sampler; - + struct i915_sampler_state *cso = CALLOC_STRUCT( i915_sampler_state ); const unsigned ws = sampler->wrap_s; const unsigned wt = sampler->wrap_t; const unsigned wr = sampler->wrap_r; unsigned minFilt, magFilt; unsigned mipFilt; + cso->templ = sampler; + mipFilt = translate_mip_filter(sampler->min_mip_filter); if (sampler->max_anisotropy > 1.0) { minFilt = FILTER_ANISOTROPIC; @@ -222,7 +222,7 @@ i915_create_sampler_state(struct pipe_context *pipe, } { - int b = sampler->lod_bias * 16.0; + int b = (int) (sampler->lod_bias * 16.0); b = CLAMP(b, -256, 255); cso->state[0] |= ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); } @@ -274,7 +274,7 @@ static void i915_bind_sampler_state(struct pipe_context *pipe, static void i915_delete_sampler_state(struct pipe_context *pipe, void *sampler) { - free(sampler); + FREE(sampler); } @@ -286,7 +286,7 @@ static void * i915_create_depth_stencil_state(struct pipe_context *pipe, const struct pipe_depth_stencil_state *depth_stencil) { - struct i915_depth_stencil_state *cso = calloc(1, sizeof(struct i915_depth_stencil_state)); + struct i915_depth_stencil_state *cso = CALLOC_STRUCT( i915_depth_stencil_state ); { int testmask = depth_stencil->stencil.value_mask[0] & 0xff; @@ -379,7 +379,7 @@ static void i915_bind_depth_stencil_state(struct pipe_context *pipe, static void i915_delete_depth_stencil_state(struct pipe_context *pipe, void *depth_stencil) { - free(depth_stencil); + FREE(depth_stencil); } @@ -387,7 +387,7 @@ static void * i915_create_alpha_test_state(struct pipe_context *pipe, const struct pipe_alpha_test_state *alpha_test) { - struct i915_alpha_test_state *cso = calloc(1, sizeof(struct i915_alpha_test_state)); + struct i915_alpha_test_state *cso = CALLOC_STRUCT( i915_alpha_test_state ); if (alpha_test->enabled) { int test = i915_translate_compare_func(alpha_test->func); @@ -413,7 +413,7 @@ static void i915_bind_alpha_test_state(struct pipe_context *pipe, static void i915_delete_alpha_test_state(struct pipe_context *pipe, void *alpha) { - free(alpha); + FREE(alpha); } static void i915_set_scissor_state( struct pipe_context *pipe, @@ -619,7 +619,7 @@ static void * i915_create_rasterizer_state(struct pipe_context *pipe, const struct pipe_rasterizer_state *rasterizer) { - struct i915_rasterizer_state *cso = calloc(1, sizeof(struct i915_rasterizer_state)); + struct i915_rasterizer_state *cso = CALLOC_STRUCT( i915_rasterizer_state ); cso->templ = rasterizer; cso->color_interp = rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; @@ -671,7 +671,7 @@ i915_create_rasterizer_state(struct pipe_context *pipe, S4_FLATSHADE_SPECULAR); } - cso->LIS7 = rasterizer->offset_units; /* probably incorrect */ + cso->LIS7 = fui( rasterizer->offset_units ); return cso; @@ -693,7 +693,7 @@ static void i915_bind_rasterizer_state( struct pipe_context *pipe, static void i915_delete_rasterizer_state(struct pipe_context *pipe, void *setup) { - free(setup); + FREE(setup); } static void i915_set_vertex_buffer( struct pipe_context *pipe, diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index 845873baa0..08fa513de4 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -50,7 +50,7 @@ static inline void set_dynamic_indirect( struct i915_context *i915, const unsigned *src, unsigned dwords ) { - int i; + unsigned i; for (i = 0; i < dwords; i++) i915->current.dynamic[offset + i] = src[i]; @@ -80,8 +80,8 @@ static void upload_MODES4( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_MODES4 = { - .dirty = I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, - .update = upload_MODES4 + I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, + upload_MODES4 }; @@ -99,8 +99,8 @@ static void upload_BFO( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_BFO = { - .dirty = I915_NEW_DEPTH_STENCIL, - .update = upload_BFO + I915_NEW_DEPTH_STENCIL, + upload_BFO }; @@ -133,8 +133,8 @@ static void upload_BLENDCOLOR( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_BLENDCOLOR = { - .dirty = I915_NEW_BLEND, - .update = upload_BLENDCOLOR + I915_NEW_BLEND, + upload_BLENDCOLOR }; /*********************************************************************** @@ -153,8 +153,8 @@ static void upload_IAB( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_IAB = { - .dirty = I915_NEW_BLEND, - .update = upload_IAB + I915_NEW_BLEND, + upload_IAB }; @@ -172,8 +172,8 @@ static void upload_DEPTHSCALE( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_DEPTHSCALE = { - .dirty = I915_NEW_RASTERIZER, - .update = upload_DEPTHSCALE + I915_NEW_RASTERIZER, + upload_DEPTHSCALE }; @@ -230,8 +230,8 @@ static void upload_STIPPLE( struct i915_context *i915 ) const struct i915_tracked_state i915_upload_STIPPLE = { - .dirty = I915_NEW_RASTERIZER | I915_NEW_STIPPLE, - .update = upload_STIPPLE + I915_NEW_RASTERIZER | I915_NEW_STIPPLE, + upload_STIPPLE }; @@ -248,8 +248,8 @@ static void upload_SCISSOR_ENABLE( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_SCISSOR_ENABLE = { - .dirty = I915_NEW_RASTERIZER, - .update = upload_SCISSOR_ENABLE + I915_NEW_RASTERIZER, + upload_SCISSOR_ENABLE }; @@ -274,8 +274,8 @@ static void upload_SCISSOR_RECT( struct i915_context *i915 ) const struct i915_tracked_state i915_upload_SCISSOR_RECT = { - .dirty = I915_NEW_SCISSOR, - .update = upload_SCISSOR_RECT + I915_NEW_SCISSOR, + upload_SCISSOR_RECT }; diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index d830bb78e8..da2402c018 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -78,8 +78,8 @@ static void upload_S0S1(struct i915_context *i915) } const struct i915_tracked_state i915_upload_S0S1 = { - .dirty = I915_NEW_VBO | I915_NEW_VERTEX_FORMAT, - .update = upload_S0S1 + I915_NEW_VBO | I915_NEW_VERTEX_FORMAT, + upload_S0S1 }; @@ -115,8 +115,8 @@ static void upload_S2S4(struct i915_context *i915) const struct i915_tracked_state i915_upload_S2S4 = { - .dirty = I915_NEW_RASTERIZER | I915_NEW_VERTEX_FORMAT, - .update = upload_S2S4 + I915_NEW_RASTERIZER | I915_NEW_VERTEX_FORMAT, + upload_S2S4 }; @@ -147,8 +147,8 @@ static void upload_S5( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_S5 = { - .dirty = (I915_NEW_DEPTH_STENCIL | I915_NEW_BLEND | I915_NEW_RASTERIZER), - .update = upload_S5 + (I915_NEW_DEPTH_STENCIL | I915_NEW_BLEND | I915_NEW_RASTERIZER), + upload_S5 }; @@ -178,8 +178,8 @@ static void upload_S6( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_S6 = { - .dirty = I915_NEW_ALPHA_TEST | I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, - .update = upload_S6 + I915_NEW_ALPHA_TEST | I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, + upload_S6 }; @@ -187,11 +187,11 @@ const struct i915_tracked_state i915_upload_S6 = { */ static void upload_S7( struct i915_context *i915 ) { - float LIS7; + unsigned LIS7; /* I915_NEW_RASTERIZER */ - LIS7 = i915->rasterizer->LIS7; /* probably incorrect */ + LIS7 = i915->rasterizer->LIS7; if (LIS7 != i915->current.immediate[I915_IMMEDIATE_S7]) { i915->current.immediate[I915_IMMEDIATE_S7] = LIS7; @@ -200,8 +200,8 @@ static void upload_S7( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_S7 = { - .dirty = I915_NEW_RASTERIZER, - .update = upload_S7 + I915_NEW_RASTERIZER, + upload_S7 }; diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index afe15fb525..bec7ddb731 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -81,7 +81,7 @@ i915_get_tile_rgba(struct pipe_context *pipe, break; case PIPE_FORMAT_S8_Z24: { - const float scale = 1.0 / (float) 0xffffff; + const float scale = 1.0f / (float) 0xffffff; for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { diff --git a/src/mesa/pipe/i915simple/i915_tex_layout.c b/src/mesa/pipe/i915simple/i915_tex_layout.c index 39c3cf2f71..cb372a7170 100644 --- a/src/mesa/pipe/i915simple/i915_tex_layout.c +++ b/src/mesa/pipe/i915simple/i915_tex_layout.c @@ -72,14 +72,14 @@ i915_miptree_set_level_info(struct pipe_mipmap_tree *mt, /* Not sure when this would happen, but anyway: */ if (mt->level[level].image_offset) { - free(mt->level[level].image_offset); + FREE(mt->level[level].image_offset); mt->level[level].image_offset = NULL; } assert(nr_images); assert(!mt->level[level].image_offset); - mt->level[level].image_offset = (unsigned *) malloc(nr_images * sizeof(unsigned)); + mt->level[level].image_offset = (unsigned *) MALLOC(nr_images * sizeof(unsigned)); mt->level[level].image_offset[0] = 0; } @@ -186,7 +186,7 @@ static const int step_offsets[6][2] = { boolean i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) { - int level; + unsigned level; switch (mt->target) { case PIPE_TEXTURE_CUBE: { @@ -309,7 +309,7 @@ i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) boolean i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) { - int level; + unsigned level; switch (mt->target) { case PIPE_TEXTURE_CUBE:{ @@ -417,7 +417,7 @@ i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) unsigned nr_images = mt->target == PIPE_TEXTURE_3D ? depth : 6; int x = 0; int y = 0; - int q, j; + unsigned q, j; i915_miptree_set_level_info(mt, level, nr_images, 0, mt->total_height, -- cgit v1.2.3 From 0e31e24659a1e691bdfa213fcd073bbfaa4ed6e9 Mon Sep 17 00:00:00 2001 From: Michal Date: Sun, 18 Nov 2007 13:20:57 +0000 Subject: Fix compatibility issues between gcc and msvc. --- src/mesa/pipe/draw/draw_prim.c | 2 +- src/mesa/pipe/draw/draw_vertex_fetch.c | 7 + src/mesa/pipe/i915simple/i915_debug.c | 481 ++++++++++++++------------ src/mesa/pipe/i915simple/i915_debug.h | 38 +- src/mesa/pipe/i915simple/i915_debug_fp.c | 99 +++--- src/mesa/pipe/i915simple/i915_fpc_translate.c | 15 +- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 6 +- src/mesa/pipe/i915simple/i915_state_dynamic.c | 2 +- src/mesa/pipe/i915simple/i915_strings.c | 2 +- src/mesa/pipe/tgsi/tgsi_platform.h | 35 +- 10 files changed, 387 insertions(+), 300 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index e4a65b9f24..f8fc23b510 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -61,7 +61,7 @@ static void draw_prim_queue_flush( struct draw_context *draw ) unsigned i; if (0) - printf("Flushing with %d prims, %d verts\n", + fprintf(stdout,"Flushing with %d prims, %d verts\n", draw->pq.queue_nr, draw->vs.queue_nr); /* Make sure all vertices are available/shaded: diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index b510a4dbba..88fa80dddc 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -80,6 +80,13 @@ fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) attrib[0] = (float) ((int *) ptr)[0]; break; + case PIPE_FORMAT_U_A8_R8_G8_B8: + attrib[0] = (float) ((unsigned char *) ptr)[2] / 255.0f; + attrib[1] = (float) ((unsigned char *) ptr)[1] / 255.0f; + attrib[2] = (float) ((unsigned char *) ptr)[0] / 255.0f; + attrib[3] = (float) ((unsigned char *) ptr)[3] / 255.0f; + break; + default: assert(0); } diff --git a/src/mesa/pipe/i915simple/i915_debug.c b/src/mesa/pipe/i915simple/i915_debug.c index 001f695e2b..94db44e1aa 100644 --- a/src/mesa/pipe/i915simple/i915_debug.c +++ b/src/mesa/pipe/i915simple/i915_debug.c @@ -33,7 +33,21 @@ #include "i915_debug.h" #include "pipe/p_winsys.h" -#define PRINTF( ... ) (stream)->winsys->printf( (stream)->winsys, __VA_ARGS__ ) + +static void +PRINTF( + struct debug_stream *stream, + const char *fmt, + ... ) +{ + va_list args; + char buffer[256]; + + va_start( args, fmt ); + vsprintf( buffer, fmt, args ); + stream->winsys->printf( stream->winsys, buffer ); + va_end( args ); +} static boolean debug( struct debug_stream *stream, const char *name, unsigned len ) @@ -42,19 +56,19 @@ static boolean debug( struct debug_stream *stream, const char *name, unsigned le unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); if (len == 0) { - PRINTF("Error - zero length packet (0x%08x)\n", stream->ptr[0]); + PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]); assert(0); return FALSE; } if (stream->print_addresses) - PRINTF("%08x: ", stream->offset); + PRINTF(stream, "%08x: ", stream->offset); - PRINTF("%s (%d dwords):\n", name, len); + PRINTF(stream, "%s (%d dwords):\n", name, len); for (i = 0; i < len; i++) - PRINTF("\t0x%08x\n", ptr[i]); - PRINTF("\n"); + PRINTF(stream, "\t0x%08x\n", ptr[i]); + PRINTF(stream, "\n"); stream->offset += len * sizeof(unsigned); @@ -91,17 +105,17 @@ static boolean debug_prim( struct debug_stream *stream, const char *name, - PRINTF("%s %s (%d dwords):\n", name, prim, len); - PRINTF("\t0x%08x\n", ptr[0]); + PRINTF(stream, "%s %s (%d dwords):\n", name, prim, len); + PRINTF(stream, "\t0x%08x\n", ptr[0]); for (i = 1; i < len; i++) { if (dump_floats) - PRINTF("\t0x%08x // %f\n", ptr[i], *(float *)&ptr[i]); + PRINTF(stream, "\t0x%08x // %f\n", ptr[i], *(float *)&ptr[i]); else - PRINTF("\t0x%08x\n", ptr[i]); + PRINTF(stream, "\t0x%08x\n", ptr[i]); } - PRINTF("\n"); + PRINTF(stream, "\n"); stream->offset += len * sizeof(unsigned); @@ -116,15 +130,15 @@ static boolean debug_program( struct debug_stream *stream, const char *name, uns unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); if (len == 0) { - PRINTF("Error - zero length packet (0x%08x)\n", stream->ptr[0]); + PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]); assert(0); return FALSE; } if (stream->print_addresses) - PRINTF("%08x: ", stream->offset); + PRINTF(stream, "%08x: ", stream->offset); - PRINTF("%s (%d dwords):\n", name, len); + PRINTF(stream, "%s (%d dwords):\n", name, len); i915_disassemble_program( stream, ptr, len ); stream->offset += len * sizeof(unsigned); @@ -138,17 +152,17 @@ static boolean debug_chain( struct debug_stream *stream, const char *name, unsig unsigned old_offset = stream->offset + len * sizeof(unsigned); unsigned i; - PRINTF("%s (%d dwords):\n", name, len); + PRINTF(stream, "%s (%d dwords):\n", name, len); for (i = 0; i < len; i++) - PRINTF("\t0x%08x\n", ptr[i]); + PRINTF(stream, "\t0x%08x\n", ptr[i]); stream->offset = ptr[1] & ~0x3; if (stream->offset < old_offset) - PRINTF("\n... skipping backwards from 0x%x --> 0x%x ...\n\n", + PRINTF(stream, "\n... skipping backwards from 0x%x --> 0x%x ...\n\n", old_offset, stream->offset ); else - PRINTF("\n... skipping from 0x%x --> 0x%x ...\n\n", + PRINTF(stream, "\n... skipping from 0x%x --> 0x%x ...\n\n", old_offset, stream->offset ); @@ -168,23 +182,38 @@ static boolean debug_variable_length_prim( struct debug_stream *stream ) len = 1+(i+2)/2; - PRINTF("3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); + PRINTF(stream, "3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); for (i = 0; i < len; i++) - PRINTF("\t0x%08x\n", ptr[i]); - PRINTF("\n"); + PRINTF(stream, "\t0x%08x\n", ptr[i]); + PRINTF(stream, "\n"); stream->offset += len * sizeof(unsigned); return TRUE; } -#define BITS( dw, hi, lo, ... ) \ -do { \ - unsigned himask = ~0UL >> (31 - (hi)); \ - PRINTF("\t\t "); \ - PRINTF(__VA_ARGS__); \ - PRINTF(": 0x%x\n", ((dw) & himask) >> (lo)); \ -} while (0) +static void +BITS( + struct debug_stream *stream, + unsigned dw, + unsigned hi, + unsigned lo, + const char *fmt, + ... ) +{ + va_list args; + char buffer[256]; + unsigned himask = ~0UL >> (31 - (hi)); + + PRINTF(stream, "\t\t "); + + va_start( args, fmt ); + vsprintf( buffer, fmt, args ); + stream->winsys->printf( stream->winsys, buffer ); + va_end( args ); + + PRINTF(stream, ": 0x%x\n", ((dw) & himask) >> (lo)); +} #define MBZ( dw, hi, lo) do { \ unsigned x = (dw) >> (lo); \ @@ -194,14 +223,28 @@ do { \ assert ((x & himask & ~lomask) == 0); \ } while (0) -#define FLAG( dw, bit, ... ) \ -do { \ - if (((dw) >> (bit)) & 1) { \ - PRINTF("\t\t "); \ - PRINTF(__VA_ARGS__); \ - PRINTF("\n"); \ - } \ -} while (0) +static void +FLAG( + struct debug_stream *stream, + unsigned dw, + unsigned bit, + const char *fmt, + ... ) +{ + if (((dw) >> (bit)) & 1) { + va_list args; + char buffer[256]; + + PRINTF(stream, "\t\t "); + + va_start( args, fmt ); + vsprintf( buffer, fmt, args ); + stream->winsys->printf( stream->winsys, buffer ); + va_end( args ); + + PRINTF(stream, "\n"); + } +} static boolean debug_load_immediate( struct debug_stream *stream, const char *name, @@ -211,95 +254,95 @@ static boolean debug_load_immediate( struct debug_stream *stream, unsigned bits = (ptr[0] >> 4) & 0xff; unsigned j = 0; - PRINTF("%s (%d dwords, flags: %x):\n", name, len, bits); - PRINTF("\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords, flags: %x):\n", name, len, bits); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); if (bits & (1<<0)) { - PRINTF("\t LIS0: 0x%08x\n", ptr[j]); - PRINTF("\t vb address: 0x%08x\n", (ptr[j] & ~0x3)); - BITS(ptr[j], 0, 0, "vb invalidate disable"); + PRINTF(stream, "\t LIS0: 0x%08x\n", ptr[j]); + PRINTF(stream, "\t vb address: 0x%08x\n", (ptr[j] & ~0x3)); + BITS(stream, ptr[j], 0, 0, "vb invalidate disable"); j++; } if (bits & (1<<1)) { - PRINTF("\t LIS1: 0x%08x\n", ptr[j]); - BITS(ptr[j], 29, 24, "vb dword width"); - BITS(ptr[j], 21, 16, "vb dword pitch"); - BITS(ptr[j], 15, 0, "vb max index"); + PRINTF(stream, "\t LIS1: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 29, 24, "vb dword width"); + BITS(stream, ptr[j], 21, 16, "vb dword pitch"); + BITS(stream, ptr[j], 15, 0, "vb max index"); j++; } if (bits & (1<<2)) { int i; - PRINTF("\t LIS2: 0x%08x\n", ptr[j]); + PRINTF(stream, "\t LIS2: 0x%08x\n", ptr[j]); for (i = 0; i < 8; i++) { unsigned tc = (ptr[j] >> (i * 4)) & 0xf; if (tc != 0xf) - BITS(tc, 3, 0, "tex coord %d", i); + BITS(stream, tc, 3, 0, "tex coord %d", i); } j++; } if (bits & (1<<3)) { - PRINTF("\t LIS3: 0x%08x\n", ptr[j]); + PRINTF(stream, "\t LIS3: 0x%08x\n", ptr[j]); j++; } if (bits & (1<<4)) { - PRINTF("\t LIS4: 0x%08x\n", ptr[j]); - BITS(ptr[j], 31, 23, "point width"); - BITS(ptr[j], 22, 19, "line width"); - FLAG(ptr[j], 18, "alpha flatshade"); - FLAG(ptr[j], 17, "fog flatshade"); - FLAG(ptr[j], 16, "spec flatshade"); - FLAG(ptr[j], 15, "rgb flatshade"); - BITS(ptr[j], 14, 13, "cull mode"); - FLAG(ptr[j], 12, "vfmt: point width"); - FLAG(ptr[j], 11, "vfmt: specular/fog"); - FLAG(ptr[j], 10, "vfmt: rgba"); - FLAG(ptr[j], 9, "vfmt: depth offset"); - BITS(ptr[j], 8, 6, "vfmt: position (2==xyzw)"); - FLAG(ptr[j], 5, "force dflt diffuse"); - FLAG(ptr[j], 4, "force dflt specular"); - FLAG(ptr[j], 3, "local depth offset enable"); - FLAG(ptr[j], 2, "vfmt: fp32 fog coord"); - FLAG(ptr[j], 1, "sprite point"); - FLAG(ptr[j], 0, "antialiasing"); + PRINTF(stream, "\t LIS4: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 23, "point width"); + BITS(stream, ptr[j], 22, 19, "line width"); + FLAG(stream, ptr[j], 18, "alpha flatshade"); + FLAG(stream, ptr[j], 17, "fog flatshade"); + FLAG(stream, ptr[j], 16, "spec flatshade"); + FLAG(stream, ptr[j], 15, "rgb flatshade"); + BITS(stream, ptr[j], 14, 13, "cull mode"); + FLAG(stream, ptr[j], 12, "vfmt: point width"); + FLAG(stream, ptr[j], 11, "vfmt: specular/fog"); + FLAG(stream, ptr[j], 10, "vfmt: rgba"); + FLAG(stream, ptr[j], 9, "vfmt: depth offset"); + BITS(stream, ptr[j], 8, 6, "vfmt: position (2==xyzw)"); + FLAG(stream, ptr[j], 5, "force dflt diffuse"); + FLAG(stream, ptr[j], 4, "force dflt specular"); + FLAG(stream, ptr[j], 3, "local depth offset enable"); + FLAG(stream, ptr[j], 2, "vfmt: fp32 fog coord"); + FLAG(stream, ptr[j], 1, "sprite point"); + FLAG(stream, ptr[j], 0, "antialiasing"); j++; } if (bits & (1<<5)) { - PRINTF("\t LIS5: 0x%08x\n", ptr[j]); - BITS(ptr[j], 31, 28, "rgba write disables"); - FLAG(ptr[j], 27, "force dflt point width"); - FLAG(ptr[j], 26, "last pixel enable"); - FLAG(ptr[j], 25, "global z offset enable"); - FLAG(ptr[j], 24, "fog enable"); - BITS(ptr[j], 23, 16, "stencil ref"); - BITS(ptr[j], 15, 13, "stencil test"); - BITS(ptr[j], 12, 10, "stencil fail op"); - BITS(ptr[j], 9, 7, "stencil pass z fail op"); - BITS(ptr[j], 6, 4, "stencil pass z pass op"); - FLAG(ptr[j], 3, "stencil write enable"); - FLAG(ptr[j], 2, "stencil test enable"); - FLAG(ptr[j], 1, "color dither enable"); - FLAG(ptr[j], 0, "logiop enable"); + PRINTF(stream, "\t LIS5: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 28, "rgba write disables"); + FLAG(stream, ptr[j], 27, "force dflt point width"); + FLAG(stream, ptr[j], 26, "last pixel enable"); + FLAG(stream, ptr[j], 25, "global z offset enable"); + FLAG(stream, ptr[j], 24, "fog enable"); + BITS(stream, ptr[j], 23, 16, "stencil ref"); + BITS(stream, ptr[j], 15, 13, "stencil test"); + BITS(stream, ptr[j], 12, 10, "stencil fail op"); + BITS(stream, ptr[j], 9, 7, "stencil pass z fail op"); + BITS(stream, ptr[j], 6, 4, "stencil pass z pass op"); + FLAG(stream, ptr[j], 3, "stencil write enable"); + FLAG(stream, ptr[j], 2, "stencil test enable"); + FLAG(stream, ptr[j], 1, "color dither enable"); + FLAG(stream, ptr[j], 0, "logiop enable"); j++; } if (bits & (1<<6)) { - PRINTF("\t LIS6: 0x%08x\n", ptr[j]); - FLAG(ptr[j], 31, "alpha test enable"); - BITS(ptr[j], 30, 28, "alpha func"); - BITS(ptr[j], 27, 20, "alpha ref"); - FLAG(ptr[j], 19, "depth test enable"); - BITS(ptr[j], 18, 16, "depth func"); - FLAG(ptr[j], 15, "blend enable"); - BITS(ptr[j], 14, 12, "blend func"); - BITS(ptr[j], 11, 8, "blend src factor"); - BITS(ptr[j], 7, 4, "blend dst factor"); - FLAG(ptr[j], 3, "depth write enable"); - FLAG(ptr[j], 2, "color write enable"); - BITS(ptr[j], 1, 0, "provoking vertex"); + PRINTF(stream, "\t LIS6: 0x%08x\n", ptr[j]); + FLAG(stream, ptr[j], 31, "alpha test enable"); + BITS(stream, ptr[j], 30, 28, "alpha func"); + BITS(stream, ptr[j], 27, 20, "alpha ref"); + FLAG(stream, ptr[j], 19, "depth test enable"); + BITS(stream, ptr[j], 18, 16, "depth func"); + FLAG(stream, ptr[j], 15, "blend enable"); + BITS(stream, ptr[j], 14, 12, "blend func"); + BITS(stream, ptr[j], 11, 8, "blend src factor"); + BITS(stream, ptr[j], 7, 4, "blend dst factor"); + FLAG(stream, ptr[j], 3, "depth write enable"); + FLAG(stream, ptr[j], 2, "color write enable"); + BITS(stream, ptr[j], 1, 0, "provoking vertex"); j++; } - PRINTF("\n"); + PRINTF(stream, "\n"); assert(j == len); @@ -318,34 +361,34 @@ static boolean debug_load_indirect( struct debug_stream *stream, unsigned bits = (ptr[0] >> 8) & 0x3f; unsigned i, j = 0; - PRINTF("%s (%d dwords):\n", name, len); - PRINTF("\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); for (i = 0; i < 6; i++) { if (bits & (1<ptr + stream->offset); int j = 0; - PRINTF("%s (%d dwords):\n", name, len); - PRINTF("\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); BR13(stream, ptr[j++]); BR22(stream, ptr[j++]); @@ -458,8 +501,8 @@ static boolean debug_color_blit( struct debug_stream *stream, unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; - PRINTF("%s (%d dwords):\n", name, len); - PRINTF("\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); BR13(stream, ptr[j++]); BR22(stream, ptr[j++]); @@ -479,13 +522,13 @@ static boolean debug_modes4( struct debug_stream *stream, unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; - PRINTF("%s (%d dwords):\n", name, len); - PRINTF("\t0x%08x\n", ptr[j]); - BITS(ptr[j], 21, 18, "logicop func"); - FLAG(ptr[j], 17, "stencil test mask modify-enable"); - FLAG(ptr[j], 16, "stencil write mask modify-enable"); - BITS(ptr[j], 15, 8, "stencil test mask"); - BITS(ptr[j], 7, 0, "stencil write mask"); + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 21, 18, "logicop func"); + FLAG(stream, ptr[j], 17, "stencil test mask modify-enable"); + FLAG(stream, ptr[j], 16, "stencil write mask modify-enable"); + BITS(stream, ptr[j], 15, 8, "stencil test mask"); + BITS(stream, ptr[j], 7, 0, "stencil write mask"); j++; stream->offset += len * sizeof(unsigned); @@ -500,42 +543,42 @@ static boolean debug_map_state( struct debug_stream *stream, unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); unsigned j = 0; - PRINTF("%s (%d dwords):\n", name, len); - PRINTF("\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); { - PRINTF("\t0x%08x\n", ptr[j]); - BITS(ptr[j], 15, 0, "map mask"); + PRINTF(stream, "\t0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 15, 0, "map mask"); j++; } while (j < len) { { - PRINTF("\t TMn.0: 0x%08x\n", ptr[j]); - PRINTF("\t map address: 0x%08x\n", (ptr[j] & ~0x3)); - FLAG(ptr[j], 1, "vertical line stride"); - FLAG(ptr[j], 0, "vertical line stride offset"); + PRINTF(stream, "\t TMn.0: 0x%08x\n", ptr[j]); + PRINTF(stream, "\t map address: 0x%08x\n", (ptr[j] & ~0x3)); + FLAG(stream, ptr[j], 1, "vertical line stride"); + FLAG(stream, ptr[j], 0, "vertical line stride offset"); j++; } { - PRINTF("\t TMn.1: 0x%08x\n", ptr[j]); - BITS(ptr[j], 31, 21, "height"); - BITS(ptr[j], 20, 10, "width"); - BITS(ptr[j], 9, 7, "surface format"); - BITS(ptr[j], 6, 3, "texel format"); - FLAG(ptr[j], 2, "use fence regs"); - FLAG(ptr[j], 1, "tiled surface"); - FLAG(ptr[j], 0, "tile walk ymajor"); + PRINTF(stream, "\t TMn.1: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 21, "height"); + BITS(stream, ptr[j], 20, 10, "width"); + BITS(stream, ptr[j], 9, 7, "surface format"); + BITS(stream, ptr[j], 6, 3, "texel format"); + FLAG(stream, ptr[j], 2, "use fence regs"); + FLAG(stream, ptr[j], 1, "tiled surface"); + FLAG(stream, ptr[j], 0, "tile walk ymajor"); j++; } { - PRINTF("\t TMn.2: 0x%08x\n", ptr[j]); - BITS(ptr[j], 31, 21, "dword pitch"); - BITS(ptr[j], 20, 15, "cube face enables"); - BITS(ptr[j], 14, 9, "max lod"); - FLAG(ptr[j], 8, "mip layout right"); - BITS(ptr[j], 7, 0, "depth"); + PRINTF(stream, "\t TMn.2: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 21, "dword pitch"); + BITS(stream, ptr[j], 20, 15, "cube face enables"); + BITS(stream, ptr[j], 14, 9, "max lod"); + FLAG(stream, ptr[j], 8, "mip layout right"); + BITS(stream, ptr[j], 7, 0, "depth"); j++; } } @@ -552,50 +595,50 @@ static boolean debug_sampler_state( struct debug_stream *stream, unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); unsigned j = 0; - PRINTF("%s (%d dwords):\n", name, len); - PRINTF("\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); { - PRINTF("\t0x%08x\n", ptr[j]); - BITS(ptr[j], 15, 0, "sampler mask"); + PRINTF(stream, "\t0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 15, 0, "sampler mask"); j++; } while (j < len) { { - PRINTF("\t TSn.0: 0x%08x\n", ptr[j]); - FLAG(ptr[j], 31, "reverse gamma"); - FLAG(ptr[j], 30, "planar to packed"); - FLAG(ptr[j], 29, "yuv->rgb"); - BITS(ptr[j], 28, 27, "chromakey index"); - BITS(ptr[j], 26, 22, "base mip level"); - BITS(ptr[j], 21, 20, "mip mode filter"); - BITS(ptr[j], 19, 17, "mag mode filter"); - BITS(ptr[j], 16, 14, "min mode filter"); - BITS(ptr[j], 13, 5, "lod bias (s4.4)"); - FLAG(ptr[j], 4, "shadow enable"); - FLAG(ptr[j], 3, "max-aniso-4"); - BITS(ptr[j], 2, 0, "shadow func"); + PRINTF(stream, "\t TSn.0: 0x%08x\n", ptr[j]); + FLAG(stream, ptr[j], 31, "reverse gamma"); + FLAG(stream, ptr[j], 30, "planar to packed"); + FLAG(stream, ptr[j], 29, "yuv->rgb"); + BITS(stream, ptr[j], 28, 27, "chromakey index"); + BITS(stream, ptr[j], 26, 22, "base mip level"); + BITS(stream, ptr[j], 21, 20, "mip mode filter"); + BITS(stream, ptr[j], 19, 17, "mag mode filter"); + BITS(stream, ptr[j], 16, 14, "min mode filter"); + BITS(stream, ptr[j], 13, 5, "lod bias (s4.4)"); + FLAG(stream, ptr[j], 4, "shadow enable"); + FLAG(stream, ptr[j], 3, "max-aniso-4"); + BITS(stream, ptr[j], 2, 0, "shadow func"); j++; } { - PRINTF("\t TSn.1: 0x%08x\n", ptr[j]); - BITS(ptr[j], 31, 24, "min lod"); + PRINTF(stream, "\t TSn.1: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 24, "min lod"); MBZ( ptr[j], 23, 18 ); - FLAG(ptr[j], 17, "kill pixel enable"); - FLAG(ptr[j], 16, "keyed tex filter mode"); - FLAG(ptr[j], 15, "chromakey enable"); - BITS(ptr[j], 14, 12, "tcx wrap mode"); - BITS(ptr[j], 11, 9, "tcy wrap mode"); - BITS(ptr[j], 8, 6, "tcz wrap mode"); - FLAG(ptr[j], 5, "normalized coords"); - BITS(ptr[j], 4, 1, "map (surface) index"); - FLAG(ptr[j], 0, "EAST deinterlacer enable"); + FLAG(stream, ptr[j], 17, "kill pixel enable"); + FLAG(stream, ptr[j], 16, "keyed tex filter mode"); + FLAG(stream, ptr[j], 15, "chromakey enable"); + BITS(stream, ptr[j], 14, 12, "tcx wrap mode"); + BITS(stream, ptr[j], 11, 9, "tcy wrap mode"); + BITS(stream, ptr[j], 8, 6, "tcz wrap mode"); + FLAG(stream, ptr[j], 5, "normalized coords"); + BITS(stream, ptr[j], 4, 1, "map (surface) index"); + FLAG(stream, ptr[j], 0, "EAST deinterlacer enable"); j++; } { - PRINTF("\t TSn.2: 0x%08x (default color)\n", ptr[j]); + PRINTF(stream, "\t TSn.2: 0x%08x (default color)\n", ptr[j]); j++; } } @@ -612,26 +655,26 @@ static boolean debug_dest_vars( struct debug_stream *stream, unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; - PRINTF("%s (%d dwords):\n", name, len); - PRINTF("\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); { - PRINTF("\t0x%08x\n", ptr[j]); - FLAG(ptr[j], 31, "early classic ztest"); - FLAG(ptr[j], 30, "opengl tex default color"); - FLAG(ptr[j], 29, "bypass iz"); - FLAG(ptr[j], 28, "lod preclamp"); - BITS(ptr[j], 27, 26, "dither pattern"); - FLAG(ptr[j], 25, "linear gamma blend"); - FLAG(ptr[j], 24, "debug dither"); - BITS(ptr[j], 23, 20, "dstorg x"); - BITS(ptr[j], 19, 16, "dstorg y"); + PRINTF(stream, "\t0x%08x\n", ptr[j]); + FLAG(stream, ptr[j], 31, "early classic ztest"); + FLAG(stream, ptr[j], 30, "opengl tex default color"); + FLAG(stream, ptr[j], 29, "bypass iz"); + FLAG(stream, ptr[j], 28, "lod preclamp"); + BITS(stream, ptr[j], 27, 26, "dither pattern"); + FLAG(stream, ptr[j], 25, "linear gamma blend"); + FLAG(stream, ptr[j], 24, "debug dither"); + BITS(stream, ptr[j], 23, 20, "dstorg x"); + BITS(stream, ptr[j], 19, 16, "dstorg y"); MBZ (ptr[j], 15, 15 ); - BITS(ptr[j], 14, 12, "422 write select"); - BITS(ptr[j], 11, 8, "cbuf format"); - BITS(ptr[j], 3, 2, "zbuf format"); - FLAG(ptr[j], 1, "vert line stride"); - FLAG(ptr[j], 1, "vert line stride offset"); + BITS(stream, ptr[j], 14, 12, "422 write select"); + BITS(stream, ptr[j], 11, 8, "cbuf format"); + BITS(stream, ptr[j], 3, 2, "zbuf format"); + FLAG(stream, ptr[j], 1, "vert line stride"); + FLAG(stream, ptr[j], 1, "vert line stride offset"); j++; } @@ -647,23 +690,23 @@ static boolean debug_buf_info( struct debug_stream *stream, unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); int j = 0; - PRINTF("%s (%d dwords):\n", name, len); - PRINTF("\t0x%08x\n", ptr[j++]); + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); { - PRINTF("\t0x%08x\n", ptr[j]); - BITS(ptr[j], 28, 28, "aux buffer id"); - BITS(ptr[j], 27, 24, "buffer id (7=depth, 3=back)"); - FLAG(ptr[j], 23, "use fence regs"); - FLAG(ptr[j], 22, "tiled surface"); - FLAG(ptr[j], 21, "tile walk ymajor"); + PRINTF(stream, "\t0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 28, 28, "aux buffer id"); + BITS(stream, ptr[j], 27, 24, "buffer id (7=depth, 3=back)"); + FLAG(stream, ptr[j], 23, "use fence regs"); + FLAG(stream, ptr[j], 22, "tiled surface"); + FLAG(stream, ptr[j], 21, "tile walk ymajor"); MBZ (ptr[j], 20, 14); - BITS(ptr[j], 13, 2, "dword pitch"); + BITS(stream, ptr[j], 13, 2, "dword pitch"); MBZ (ptr[j], 2, 0); j++; } - PRINTF("\t0x%08x -- buffer base address\n", ptr[j++]); + PRINTF(stream, "\t0x%08x -- buffer base address\n", ptr[j++]); stream->offset += len * sizeof(unsigned); assert(j == len); diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h index 356c751a5a..63f7195626 100644 --- a/src/mesa/pipe/i915simple/i915_debug.h +++ b/src/mesa/pipe/i915simple/i915_debug.h @@ -31,6 +31,8 @@ #ifndef I915_DEBUG_H #define I915_DEBUG_H +#include + struct i915_context; struct debug_stream @@ -68,14 +70,38 @@ void i915_print_ureg(const char *msg, unsigned ureg); #define DEBUG_WINSYS 0x4000 #ifdef DEBUG + #include "pipe/p_winsys.h" -#define I915_DBG( i915, ... ) do { \ - if ((i915)->debug & FILE_DEBUG_FLAG) \ - (i915)->pipe.winsys->printf( (i915)->pipe.winsys, __VA_ARGS__ ); \ -} while(0) + +static void +I915_DBG( + struct i915_context *i915, + const char *fmt, + ... ) +{ + if ((i915)->debug & FILE_DEBUG_FLAG) { + va_list args; + char buffer[256]; + + va_start( args, fmt ); + vsprintf( buffer, fmt, args ); + i915->pipe.winsys->printf( i915->pipe.winsys, buffer ); + va_end( args ); + } +} + #else -#define I915_DBG( i915, ... ) \ - (void)i915 + +static void +I915_DBG( + struct i915_context *i915, + const char *fmt, + ... ) +{ + (void) i915; + (void) fmt; +} + #endif diff --git a/src/mesa/pipe/i915simple/i915_debug_fp.c b/src/mesa/pipe/i915simple/i915_debug_fp.c index ec6b0cbf19..ebfdb3d93c 100644 --- a/src/mesa/pipe/i915simple/i915_debug_fp.c +++ b/src/mesa/pipe/i915simple/i915_debug_fp.c @@ -32,11 +32,20 @@ #include "pipe/p_util.h" +static void +PRINTF( + struct debug_stream *stream, + const char *fmt, + ... ) +{ + va_list args; + char buffer[256]; - -#define PRINTF( ... ) (stream)->winsys->printf( (stream)->winsys, __VA_ARGS__ ) - - + va_start( args, fmt ); + vsprintf( buffer, fmt, args ); + stream->winsys->printf( stream->winsys, buffer ); + va_end( args ); +} static const char *opcodes[0x20] = { @@ -129,27 +138,27 @@ print_reg_type_nr(struct debug_stream *stream, unsigned type, unsigned nr) case REG_TYPE_T: switch (nr) { case T_DIFFUSE: - PRINTF("T_DIFFUSE"); + PRINTF(stream, "T_DIFFUSE"); return; case T_SPECULAR: - PRINTF("T_SPECULAR"); + PRINTF(stream, "T_SPECULAR"); return; case T_FOG_W: - PRINTF("T_FOG_W"); + PRINTF(stream, "T_FOG_W"); return; default: - PRINTF("T_TEX%d", nr); + PRINTF(stream, "T_TEX%d", nr); return; } case REG_TYPE_OC: if (nr == 0) { - PRINTF("oC"); + PRINTF(stream, "oC"); return; } break; case REG_TYPE_OD: if (nr == 0) { - PRINTF("oD"); + PRINTF(stream, "oD"); return; } break; @@ -157,7 +166,7 @@ print_reg_type_nr(struct debug_stream *stream, unsigned type, unsigned nr) break; } - PRINTF("%s[%d]", regname[type], nr); + PRINTF(stream, "%s[%d]", regname[type], nr); } #define REG_SWIZZLE_MASK 0x7777 @@ -178,33 +187,33 @@ print_reg_neg_swizzle(struct debug_stream *stream, unsigned reg) (reg & REG_NEGATE_MASK) == 0) return; - PRINTF("."); + PRINTF(stream, "."); for (i = 3; i >= 0; i--) { if (reg & (1 << ((i * 4) + 3))) - PRINTF("-"); + PRINTF(stream, "-"); switch ((reg >> (i * 4)) & 0x7) { case 0: - PRINTF("x"); + PRINTF(stream, "x"); break; case 1: - PRINTF("y"); + PRINTF(stream, "y"); break; case 2: - PRINTF("z"); + PRINTF(stream, "z"); break; case 3: - PRINTF("w"); + PRINTF(stream, "w"); break; case 4: - PRINTF("0"); + PRINTF(stream, "0"); break; case 5: - PRINTF("1"); + PRINTF(stream, "1"); break; default: - PRINTF("?"); + PRINTF(stream, "?"); break; } } @@ -229,15 +238,15 @@ print_dest_reg(struct debug_stream *stream, unsigned dword) print_reg_type_nr(stream, type, nr); if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) return; - PRINTF("."); + PRINTF(stream, "."); if (dword & A0_DEST_CHANNEL_X) - PRINTF("x"); + PRINTF(stream, "x"); if (dword & A0_DEST_CHANNEL_Y) - PRINTF("y"); + PRINTF(stream, "y"); if (dword & A0_DEST_CHANNEL_Z) - PRINTF("z"); + PRINTF(stream, "z"); if (dword & A0_DEST_CHANNEL_W) - PRINTF("w"); + PRINTF(stream, "w"); } @@ -253,29 +262,29 @@ print_arith_op(struct debug_stream *stream, if (opcode != A0_NOP) { print_dest_reg(stream, program[0]); if (program[0] & A0_DEST_SATURATE) - PRINTF(" = SATURATE "); + PRINTF(stream, " = SATURATE "); else - PRINTF(" = "); + PRINTF(stream, " = "); } - PRINTF("%s ", opcodes[opcode]); + PRINTF(stream, "%s ", opcodes[opcode]); print_src_reg(stream, GET_SRC0_REG(program[0], program[1])); if (args[opcode] == 1) { - PRINTF("\n"); + PRINTF(stream, "\n"); return; } - PRINTF(", "); + PRINTF(stream, ", "); print_src_reg(stream, GET_SRC1_REG(program[1], program[2])); if (args[opcode] == 2) { - PRINTF("\n"); + PRINTF(stream, "\n"); return; } - PRINTF(", "); + PRINTF(stream, ", "); print_src_reg(stream, GET_SRC2_REG(program[2])); - PRINTF("\n"); + PRINTF(stream, "\n"); return; } @@ -285,40 +294,40 @@ print_tex_op(struct debug_stream *stream, unsigned opcode, const unsigned * program) { print_dest_reg(stream, program[0] | A0_DEST_CHANNEL_ALL); - PRINTF(" = "); + PRINTF(stream, " = "); - PRINTF("%s ", opcodes[opcode]); + PRINTF(stream, "%s ", opcodes[opcode]); - PRINTF("S[%d],", program[0] & T0_SAMPLER_NR_MASK); + PRINTF(stream, "S[%d],", program[0] & T0_SAMPLER_NR_MASK); print_reg_type_nr(stream, (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & REG_TYPE_MASK, (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); - PRINTF("\n"); + PRINTF(stream, "\n"); } static void print_texkil_op(struct debug_stream *stream, unsigned opcode, const unsigned * program) { - PRINTF("TEXKIL "); + PRINTF(stream, "TEXKIL "); print_reg_type_nr(stream, (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & REG_TYPE_MASK, (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); - PRINTF("\n"); + PRINTF(stream, "\n"); } static void print_dcl_op(struct debug_stream *stream, unsigned opcode, const unsigned * program) { - PRINTF("%s ", opcodes[opcode]); + PRINTF(stream, "%s ", opcodes[opcode]); print_dest_reg(stream, program[0] | A0_DEST_CHANNEL_ALL); - PRINTF("\n"); + PRINTF(stream, "\n"); } @@ -329,7 +338,7 @@ i915_disassemble_program(struct debug_stream *stream, unsigned size = program[0] & 0x1ff; unsigned i; - PRINTF("\t\tBEGIN\n"); + PRINTF(stream, "\t\tBEGIN\n"); assert(size + 2 == sz); @@ -337,7 +346,7 @@ i915_disassemble_program(struct debug_stream *stream, for (i = 1; i < sz; i += 3, program += 3) { unsigned opcode = program[0] & (0x1f << 24); - PRINTF("\t\t"); + PRINTF(stream, "\t\t"); if ((int) opcode >= A0_NOP && opcode <= A0_SLT) print_arith_op(stream, opcode >> 24, program); @@ -348,10 +357,10 @@ i915_disassemble_program(struct debug_stream *stream, else if (opcode == D0_DCL) print_dcl_op(stream, opcode >> 24, program); else - PRINTF("Unknown opcode 0x%x\n", opcode); + PRINTF(stream, "Unknown opcode 0x%x\n", opcode); } - PRINTF("\t\tEND\n\n"); + PRINTF(stream, "\t\tEND\n\n"); } diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index f9673e5424..0382aa26a9 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -117,11 +117,13 @@ void i915_program_error(struct i915_fp_compile *p, const char *msg, ...) { va_list args; + char buffer[1024]; fprintf(stderr, "i915_program_error: "); va_start( args, msg ); - vfprintf( stderr, msg, args ); + vsprintf( buffer, msg, args ); va_end( args ); + fprintf(stderr, buffer); fprintf(stderr, "\n"); p->error = 1; @@ -381,6 +383,9 @@ emit_simple_arith(struct i915_fp_compile *p, arg3 ); } +#ifndef M_PI +#define M_PI 3.14159265358979323846 +#endif /* * Translate TGSI instruction to i915 instruction. @@ -430,7 +435,7 @@ i915_translate_instruction(struct i915_fp_compile *p, i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - src0, i915_emit_const1f(p, 1.0f / (M_PI * 2.0f)), 0); + src0, i915_emit_const1f(p, 1.0f / (float) (M_PI * 2.0)), 0); i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); @@ -439,7 +444,7 @@ i915_translate_instruction(struct i915_fp_compile *p, i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - tmp, i915_emit_const1f(p, (M_PI * 2.0f)), 0); + tmp, i915_emit_const1f(p, (float) (M_PI * 2.0)), 0); /* * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 @@ -772,7 +777,7 @@ i915_translate_instruction(struct i915_fp_compile *p, i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - src0, i915_emit_const1f(p, 1.0 / (M_PI * 2)), 0); + src0, i915_emit_const1f(p, 1.0f / (float) (M_PI * 2.0)), 0); i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); @@ -781,7 +786,7 @@ i915_translate_instruction(struct i915_fp_compile *p, i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, - tmp, i915_emit_const1f(p, (M_PI * 2)), 0); + tmp, i915_emit_const1f(p, (float) (M_PI * 2.0)), 0); /* * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 1c8c6a37ba..736a08fb09 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -202,7 +202,7 @@ static void vbuf_tri( struct draw_stage *stage, for (i = 0; i < 3; i++) { emit_vertex( vbuf, prim->v[i] ); - vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; + vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; } } @@ -218,7 +218,7 @@ static void vbuf_line(struct draw_stage *stage, for (i = 0; i < 2; i++) { emit_vertex( vbuf, prim->v[i] ); - vbuf->element_map[vbuf->nr_elements++] = prim->v[i]->vertex_id; + vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; } } @@ -232,7 +232,7 @@ static void vbuf_point(struct draw_stage *stage, emit_vertex( vbuf, prim->v[0] ); - vbuf->element_map[vbuf->nr_elements++] = prim->v[0]->vertex_id; + vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[0]->vertex_id; } diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c index 08fa513de4..8cfbdddd19 100644 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ b/src/mesa/pipe/i915simple/i915_state_dynamic.c @@ -45,7 +45,7 @@ * (active) state every time a 4kb boundary is crossed. */ -static inline void set_dynamic_indirect( struct i915_context *i915, +static INLINE void set_dynamic_indirect( struct i915_context *i915, unsigned offset, const unsigned *src, unsigned dwords ) diff --git a/src/mesa/pipe/i915simple/i915_strings.c b/src/mesa/pipe/i915simple/i915_strings.c index a9b0bfc7c2..c713bf7208 100644 --- a/src/mesa/pipe/i915simple/i915_strings.c +++ b/src/mesa/pipe/i915simple/i915_strings.c @@ -70,7 +70,7 @@ static const char *i915_get_name( struct pipe_context *pipe ) break; } - snprintf(buffer, sizeof(buffer), "pipe/i915 (chipset: %s)", chipset); + sprintf(buffer, "pipe/i915 (chipset: %s)", chipset); return buffer; } diff --git a/src/mesa/pipe/tgsi/tgsi_platform.h b/src/mesa/pipe/tgsi/tgsi_platform.h index b98a1f21be..e7a381b201 100644 --- a/src/mesa/pipe/tgsi/tgsi_platform.h +++ b/src/mesa/pipe/tgsi/tgsi_platform.h @@ -1,19 +1,16 @@ -#if !defined TGSI_PLATFORM_H -#define TGSI_PLATFORM_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -#include "imports.h" -#include "mtypes.h" -#include "prog_instruction.h" -#include "program.h" -#include "pipe/p_compiler.h" - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_PLATFORM_H - +#if !defined TGSI_PLATFORM_H +#define TGSI_PLATFORM_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +#include "pipe/p_compiler.h" +#include "pipe/p_util.h" + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_PLATFORM_H + -- cgit v1.2.3 From c5841425433f003af76f03435de719c40635005a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 19 Nov 2007 17:23:55 +0000 Subject: Fix build errors. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 1 + src/mesa/state_tracker/st_mesa_to_tgsi.h | 4 ++++ 2 files changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index ab7aa504a1..bc57868af2 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -34,6 +34,7 @@ #include "tgsi_platform.h" #include "pipe/tgsi/exec/tgsi_core.h" #include "st_mesa_to_tgsi.h" +#include "shader/prog_instruction.h" #include "shader/prog_parameter.h" #define TGSI_DEBUG 0 diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index 941a75ab05..f2c9773107 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -29,11 +29,15 @@ #ifndef ST_MESA_TO_TGSI_H #define ST_MESA_TO_TGSI_H +#include "GL/gl.h" + + #if defined __cplusplus extern "C" { #endif struct tgsi_token; +struct gl_program; GLboolean tgsi_translate_mesa_program( -- cgit v1.2.3 From 369ff9786d88d813fb8cd07607b5c1088399a702 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 19 Nov 2007 22:01:32 +0000 Subject: The right include was mtypes.h. --- src/mesa/state_tracker/st_mesa_to_tgsi.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index f2c9773107..4cd4b96a58 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -29,7 +29,7 @@ #ifndef ST_MESA_TO_TGSI_H #define ST_MESA_TO_TGSI_H -#include "GL/gl.h" +#include "mtypes.h" #if defined __cplusplus -- cgit v1.2.3 From 45f658f172b4a3fe6e5190fdba5c00e7f332845c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 19 Nov 2007 19:22:30 +0000 Subject: Fix build. --- src/mesa/pipe/i915simple/i915_debug.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h index 63f7195626..13197f3595 100644 --- a/src/mesa/pipe/i915simple/i915_debug.h +++ b/src/mesa/pipe/i915simple/i915_debug.h @@ -69,7 +69,7 @@ void i915_print_ureg(const char *msg, unsigned ureg); #define DEBUG_SURFACE 0x2000 #define DEBUG_WINSYS 0x4000 -#ifdef DEBUG +#if defined(DEBUG) && defined(FILE_DEBUG_FLAG) #include "pipe/p_winsys.h" -- cgit v1.2.3 From 1a8daf0627dde44aaa7c40786782618d4d5a6a36 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 20 Nov 2007 13:26:00 +0000 Subject: New vertex buffer stage. --- src/mesa/pipe/draw/draw_vbuf.c | 399 +++++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_vbuf.h | 106 +++++++++++ src/mesa/sources | 1 + 3 files changed, 506 insertions(+) create mode 100644 src/mesa/pipe/draw/draw_vbuf.c create mode 100644 src/mesa/pipe/draw/draw_vbuf.h (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c new file mode 100644 index 0000000000..d00cdec56c --- /dev/null +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -0,0 +1,399 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Vertex buffer drawing stage. + * + * \author José Fonseca + * \author Keith Whitwell + */ + + +#include + +#include "pipe/draw/draw_vbuf.h" +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_vertex.h" +#include "pipe/p_util.h" + + +/** + * Vertex buffer emit stage. + */ +struct vbuf_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct vbuf_render *render; + + /** Vertex size in bytes */ + unsigned vertex_size; + + /* FIXME: we have no guarantee that 'unsigned' is 32bit */ + + /** Vertices in hardware format */ + unsigned *vertices; + unsigned *vertex_ptr; + unsigned max_vertices; + unsigned nr_vertices; + + /** Indices */ + ushort *indices; + unsigned max_indices; + unsigned nr_indices; + + /** Pipe primitive */ + unsigned prim; +}; + + +/** + * Basically a cast wrapper. + */ +static INLINE struct vbuf_stage * +vbuf_stage( struct draw_stage *stage ) +{ + assert(stage); + return (struct vbuf_stage *)stage; +} + + +static void vbuf_flush_indices( struct draw_stage *stage ); +static void vbuf_flush_vertices( struct draw_stage *stage, + unsigned new_vertex_size ); + + +static INLINE boolean +overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) +{ + unsigned long used = (unsigned long) ((char *)ptr - (char *)map); + return (used + bytes) > bufsz; +} + + +static INLINE void +check_space( struct vbuf_stage *vbuf, unsigned nr ) +{ + if (vbuf->nr_vertices + nr > vbuf->max_vertices ) + vbuf_flush_vertices(&vbuf->stage, vbuf->vertex_size ); + + if (vbuf->nr_indices + nr > vbuf->max_indices ) + vbuf_flush_indices(&vbuf->stage); +} + + +/** + * Extract the needed fields from vertex_header and emit i915 dwords. + * Recall that the vertices are constructed by the 'draw' module and + * have a couple of slots at the beginning (1-dword header, 4-dword + * clip pos) that we ignore here. + */ +static INLINE void +emit_vertex( struct vbuf_stage *vbuf, + struct vertex_header *vertex ) +{ + const struct vertex_info *vinfo = vbuf->render->get_vertex_info(vbuf->render); + + uint i; + uint count = 0; /* for debug/sanity */ + +// fprintf(stderr, "emit vertex %d to %p\n", +// vbuf->nr_vertices, vbuf->vertex_ptr); + + if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { + if(vertex->vertex_id < vbuf->nr_vertices) + return; + else + fprintf(stderr, "Bad vertex id 0x%04x (>= 0x%04x)\n", + vertex->vertex_id, vbuf->nr_vertices); + return; + } + + vertex->vertex_id = vbuf->nr_vertices++; + + for (i = 0; i < vinfo->num_attribs; i++) { + switch (vinfo->format[i]) { + case FORMAT_OMIT: + /* no-op */ + break; + case FORMAT_1F: + *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); + count++; + break; + case FORMAT_2F: + *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); + count += 2; + break; + case FORMAT_3F: + *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][2]); + count += 3; + break; + case FORMAT_4F: + *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][2]); + *vbuf->vertex_ptr++ = fui(vertex->data[i][3]); + count += 4; + break; + case FORMAT_4UB: + *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[i][2] ), + float_to_ubyte( vertex->data[i][1] ), + float_to_ubyte( vertex->data[i][0] ), + float_to_ubyte( vertex->data[i][3] )); + count += 1; + break; + default: + assert(0); + } + } + assert(count == vinfo->size); +} + + +static void +vbuf_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + check_space( vbuf, 3 ); + + for (i = 0; i < 3; i++) { + emit_vertex( vbuf, prim->v[i] ); + + vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; + } +} + + +static void +vbuf_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + check_space( vbuf, 2 ); + + for (i = 0; i < 2; i++) { + emit_vertex( vbuf, prim->v[i] ); + + vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; + } +} + + +static void +vbuf_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + check_space( vbuf, 1 ); + + emit_vertex( vbuf, prim->v[0] ); + + vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[0]->vertex_id; +} + + +static void +vbuf_first_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( stage ); + stage->tri = vbuf_tri; + stage->tri( stage, prim ); + vbuf->prim = PIPE_PRIM_TRIANGLES; + vbuf->render->set_primitive(vbuf->render, PIPE_PRIM_TRIANGLES); +} + + +static void +vbuf_first_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( stage ); + stage->line = vbuf_line; + stage->line( stage, prim ); + vbuf->prim = PIPE_PRIM_LINES; + vbuf->render->set_primitive(vbuf->render, PIPE_PRIM_LINES); +} + + +static void +vbuf_first_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( stage ); + stage->point = vbuf_point; + stage->point( stage, prim ); + vbuf->prim = PIPE_PRIM_POINTS; + vbuf->render->set_primitive(vbuf->render, PIPE_PRIM_POINTS); +} + + +static void +vbuf_flush_indices( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + if(!vbuf->nr_indices) + return; + + assert(vbuf->vertex_ptr - vbuf->vertices == + vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); + + switch(vbuf->prim) { + case PIPE_PRIM_POINTS: + break; + case PIPE_PRIM_LINES: + assert(vbuf->nr_indices % 2 == 0); + break; + case PIPE_PRIM_TRIANGLES: + assert(vbuf->nr_indices % 3 == 0); + break; + default: + assert(0); + } + + vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); + + vbuf->nr_indices = 0; +} + + +/** + * Flush existing vertex buffer and allocate a new one. + * + * XXX: We separate flush-on-index-full and flush-on-vb-full, but may + * raise issues uploading vertices if the hardware wants to flush when + * we flush. + */ +static void +vbuf_flush_vertices( struct draw_stage *stage, + unsigned new_vertex_size ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + if(vbuf->vertices) { + vbuf_flush_indices(stage); + + /* Reset temporary vertices ids */ + if(vbuf->nr_vertices) + draw_reset_vertex_ids( vbuf->stage.draw ); + + /* Free the vertex buffer */ + vbuf->render->release_vertices(vbuf->render, + vbuf->vertices, + vbuf->vertex_size, + vbuf->nr_vertices); + vbuf->nr_vertices = 0; + vbuf->vertex_ptr = vbuf->vertices = NULL; + + } + + assert(!vbuf->nr_indices); + + /* Allocate a new vertex buffer */ + vbuf->vertex_size = new_vertex_size; + vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; + vbuf->vertices = vbuf->render->allocate_vertices(vbuf->render, + vbuf->vertex_size, + vbuf->max_vertices) ; + vbuf->vertex_ptr = vbuf->vertices; +} + + +static void +vbuf_begin( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage(stage); + const struct vertex_info *vinfo = vbuf->render->get_vertex_info(vbuf->render); + unsigned vertex_size = vinfo->size * sizeof(float); + + if(vbuf->vertex_size != vertex_size) + vbuf_flush_vertices(&vbuf->stage, vertex_size); +} + + +static void +vbuf_end( struct draw_stage *stage ) +{ + /* XXX: Overkill */ + vbuf_flush_indices( stage ); + + stage->point = vbuf_first_point; + stage->line = vbuf_first_line; + stage->tri = vbuf_first_tri; +} + + +static void +vbuf_reset_stipple_counter( struct draw_stage *stage ) +{ +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *draw_vbuf_stage( struct draw_context *draw, + struct vbuf_render *render ) +{ + struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); + + vbuf->stage.draw = draw; + vbuf->stage.begin = vbuf_begin; + vbuf->stage.point = vbuf_first_point; + vbuf->stage.line = vbuf_first_line; + vbuf->stage.tri = vbuf_first_tri; + vbuf->stage.end = vbuf_end; + vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter; + + vbuf->render = render; + + assert(render->max_indices < UNDEFINED_VERTEX_ID); + vbuf->max_indices = render->max_indices; + /* FIXME: free this memory on takedown */ + vbuf->indices = MALLOC( vbuf->max_indices ); + + vbuf->vertices = NULL; + vbuf->vertex_ptr = vbuf->vertices; + + return &vbuf->stage; +} diff --git a/src/mesa/pipe/draw/draw_vbuf.h b/src/mesa/pipe/draw/draw_vbuf.h new file mode 100644 index 0000000000..43aa740f64 --- /dev/null +++ b/src/mesa/pipe/draw/draw_vbuf.h @@ -0,0 +1,106 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Vertex buffer drawing stage. + * + * \author Keith Whitwell + * \author José Fonseca + */ + +#ifndef DRAW_VBUF_H_ +#define DRAW_VBUF_H_ + + +#include "pipe/p_util.h" + + +struct draw_context; +struct vertex_info; + + +/** + * Interface for hardware vertex buffer rendering. + */ +struct vbuf_render { + + /** + * Driver limits. May be tuned lower to improve cache hits on + * index list. + */ + unsigned max_indices; + unsigned max_vertex_buffer_bytes; + + /** + * Get the hardware vertex format. + * + * XXX: have this in draw_context instead? + */ + const struct vertex_info *(*get_vertex_info)( struct vbuf_render * ); + + /** + * Request a destination for vertices. + * Hardware renderers will use ttm memory, others will just malloc + * something. + */ + void *(*allocate_vertices)( struct vbuf_render *, + ushort vertex_size, + ushort nr_vertices ); + + /** + * Notify the renderer of the current primitive when it changes. + * Prim is restricted to TRIANGLES, LINES and POINTS. + */ + void (*set_primitive)( struct vbuf_render *, unsigned prim ); + + /** + * DrawElements, note indices are ushort: + */ + void (*draw)( struct vbuf_render *, + const ushort *indices, + unsigned nr_indices ); + + /** + * Called when vbuf is done with this set of vertices: + */ + void (*release_vertices)( struct vbuf_render *, + void *vertices, + unsigned vertex_size, + unsigned vertices_used ); + + void (*destroy)( struct vbuf_render * ); +}; + + + +struct draw_stage * +draw_vbuf_stage( struct draw_context *draw, + struct vbuf_render *render ); + + +#endif /*DRAW_VBUF_H_*/ diff --git a/src/mesa/sources b/src/mesa/sources index 0b9b5fca3a..ea6840f3d9 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -169,6 +169,7 @@ DRAW_SOURCES = \ pipe/draw/draw_twoside.c \ pipe/draw/draw_unfilled.c \ pipe/draw/draw_validate.c \ + pipe/draw/draw_vbuf.c \ pipe/draw/draw_vertex.c \ pipe/draw/draw_vertex_cache.c \ pipe/draw/draw_vertex_fetch.c \ -- cgit v1.2.3 From 9924f208cf3f45424b6464d2cca9698da206816e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 20 Nov 2007 13:26:34 +0000 Subject: Use the new vertex buffer draw stage. --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 423 ++++++++---------------------- 1 file changed, 116 insertions(+), 307 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 736a08fb09..571ad40595 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -38,16 +38,12 @@ */ -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vertex.h" +#include + +#include "pipe/draw/draw_vbuf.h" #include "pipe/p_util.h" #include "pipe/p_winsys.h" -#include "softpipe/sp_context.h" -#include "softpipe/sp_headers.h" -#include "softpipe/sp_quad.h" -#include "softpipe/sp_prim_setup.h" - #include "i915_context.h" #include "i915_reg.h" #include "i915_winsys.h" @@ -55,253 +51,99 @@ #include "i915_state.h" -static void vbuf_flush_elements( struct draw_stage *stage ); -static void vbuf_flush_vertices( struct draw_stage *stage ); - - -#define VBUF_SIZE (128*1024) -#define IBUF_SIZE (16*1024) - - /** - * Vertex buffer emit stage. + * Primitive renderer for i915. */ -struct vbuf_stage { - struct draw_stage stage; /**< This must be first (base class) */ +struct i915_vbuf_render { + struct vbuf_render base; + + struct i915_context *i915; /** Vertex size in bytes */ unsigned vertex_size; - /* FIXME: we have no guarantee that 'unsigned' is 32bit */ - - /** Vertices in hardware format */ - unsigned *vertex_map; - unsigned *vertex_ptr; - unsigned max_vertices; - unsigned nr_vertices; - - ushort *element_map; - unsigned nr_elements; - - unsigned prim; - - struct i915_context *i915; + /** Hardware primitive */ + unsigned hwprim; }; /** * Basically a cast wrapper. */ -static INLINE struct vbuf_stage *vbuf_stage( struct draw_stage *stage ) -{ - return (struct vbuf_stage *)stage; -} - - -static INLINE boolean -overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) -{ - unsigned long used = (unsigned long) ((char *)ptr - (char *)map); - return (used + bytes) > bufsz; -} - - -static INLINE void -check_space( struct vbuf_stage *vbuf, unsigned nr ) -{ - if (overflow( vbuf->vertex_map, - vbuf->vertex_ptr, - nr*vbuf->vertex_size, - VBUF_SIZE )) - vbuf_flush_vertices(&vbuf->stage); - - if (vbuf->nr_elements + nr > IBUF_SIZE / sizeof(ushort) ) - vbuf_flush_elements(&vbuf->stage); -} - - -/** - * Extract the needed fields from vertex_header and emit i915 dwords. - * Recall that the vertices are constructed by the 'draw' module and - * have a couple of slots at the beginning (1-dword header, 4-dword - * clip pos) that we ignore here. - */ -static INLINE void -emit_vertex( struct vbuf_stage *vbuf, - struct vertex_header *vertex ) +static INLINE struct i915_vbuf_render * +i915_vbuf_render( struct vbuf_render *render ) { - struct i915_context *i915 = vbuf->i915; - const struct vertex_info *vinfo = &i915->current.vertex_info; - uint i; - uint count = 0; /* for debug/sanity */ - -// fprintf(stderr, "emit vertex %d to %p\n", -// vbuf->nr_vertices, vbuf->vertex_ptr); - - if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { - if(vertex->vertex_id < vbuf->nr_vertices) - return; - else - fprintf(stderr, "Bad vertex id 0x%04x (>= 0x%04x)\n", - vertex->vertex_id, vbuf->nr_vertices); - return; - } - - vertex->vertex_id = vbuf->nr_vertices++; - - for (i = 0; i < vinfo->num_attribs; i++) { - switch (vinfo->format[i]) { - case FORMAT_OMIT: - /* no-op */ - break; - case FORMAT_1F: - *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); - count++; - break; - case FORMAT_2F: - *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); - count += 2; - break; - case FORMAT_3F: - *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][2]); - count += 3; - break; - case FORMAT_4F: - *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][2]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][3]); - count += 4; - break; - case FORMAT_4UB: - *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[i][2] ), - float_to_ubyte( vertex->data[i][1] ), - float_to_ubyte( vertex->data[i][0] ), - float_to_ubyte( vertex->data[i][3] )); - count += 1; - break; - default: - assert(0); - } - } - assert(count == vinfo->size); + assert(render); + return (struct i915_vbuf_render *)render; } -static void vbuf_tri( struct draw_stage *stage, - struct prim_header *prim ) +static const struct vertex_info * +i915_vbuf_render_get_vertex_info( struct vbuf_render *render ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - check_space( vbuf, 3 ); - - for (i = 0; i < 3; i++) { - emit_vertex( vbuf, prim->v[i] ); - - vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; - } + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + return &i915->current.vertex_info; } -static void vbuf_line(struct draw_stage *stage, - struct prim_header *prim) +static void * +i915_vbuf_render_allocate_vertices( struct vbuf_render *render, + ushort vertex_size, + ushort nr_vertices ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - check_space( vbuf, 2 ); - - for (i = 0; i < 2; i++) { - emit_vertex( vbuf, prim->v[i] ); - - vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; - } -} - - -static void vbuf_point(struct draw_stage *stage, - struct prim_header *prim) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - check_space( vbuf, 1 ); + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + struct pipe_winsys *winsys = i915->pipe.winsys; + size_t size = (size_t)vertex_size * (size_t)nr_vertices; - emit_vertex( vbuf, prim->v[0] ); + /* FIXME: handle failure */ + assert(!i915->vbo); + i915->vbo = winsys->buffer_create(winsys, 64, 0, 0); + winsys->buffer_data( winsys, i915->vbo, + size, NULL, + I915_BUFFER_USAGE_LIT_VERTEX ); - vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[0]->vertex_id; -} - - -static void vbuf_first_tri( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_elements( stage ); - stage->tri = vbuf_tri; - stage->tri( stage, prim ); - vbuf->prim = PIPE_PRIM_TRIANGLES; -} - - -static void vbuf_first_line( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_elements( stage ); - stage->line = vbuf_line; - stage->line( stage, prim ); - vbuf->prim = PIPE_PRIM_LINES; -} - - -static void vbuf_first_point( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_elements( stage ); - stage->point = vbuf_point; - stage->point( stage, prim ); - vbuf->prim = PIPE_PRIM_POINTS; + i915->dirty |= I915_NEW_VBO; + + return winsys->buffer_map(winsys, + i915->vbo, + PIPE_BUFFER_FLAG_WRITE ); } -static void vbuf_flush_elements( struct draw_stage *stage ) +static void +i915_vbuf_render_set_primitive( struct vbuf_render *render, + unsigned prim ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - struct i915_context *i915 = vbuf->i915; - unsigned nr = vbuf->nr_elements; - unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ - unsigned hwprim; - unsigned i; + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); - if(!nr) - return; - - switch(vbuf->prim) { + switch(prim) { case PIPE_PRIM_POINTS: - hwprim = PRIM3D_POINTLIST; + i915_render->hwprim = PRIM3D_POINTLIST; break; case PIPE_PRIM_LINES: - hwprim = PRIM3D_LINELIST; - assert(nr % 2 == 0); + i915_render->hwprim = PRIM3D_LINELIST; break; case PIPE_PRIM_TRIANGLES: - hwprim = PRIM3D_TRILIST; - assert(nr % 3 == 0); + i915_render->hwprim = PRIM3D_TRILIST; break; default: assert(0); - return; } - - assert(vbuf->vertex_ptr - vbuf->vertex_map == vbuf->nr_vertices * vertex_size / 4); +} + + +static void +i915_vbuf_render_draw( struct vbuf_render *render, + const ushort *indices, + unsigned nr_indices ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + unsigned i; + + assert(nr_indices); assert((i915->dirty & ~I915_NEW_VBO) == 0); @@ -311,7 +153,7 @@ static void vbuf_flush_elements( struct draw_stage *stage ) if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); - if (!BEGIN_BATCH( 1 + (nr + 1)/2, 1 )) { + if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { FLUSH_BATCH(); /* Make sure state is re-emitted after a flush: @@ -319,7 +161,7 @@ static void vbuf_flush_elements( struct draw_stage *stage ) i915_update_derived( i915 ); i915_emit_hardware_state( i915 ); - if (!BEGIN_BATCH( 1 + (nr + 1)/2, 1 )) { + if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { assert(0); return; } @@ -327,95 +169,68 @@ static void vbuf_flush_elements( struct draw_stage *stage ) OUT_BATCH( _3DPRIMITIVE | PRIM_INDIRECT | - hwprim | + i915_render->hwprim | PRIM_INDIRECT_ELTS | - nr ); - for (i = 0; i + 1 < nr; i += 2) { - OUT_BATCH( vbuf->element_map[i] | - (vbuf->element_map[i + 1] << 16) ); + nr_indices ); + for (i = 0; i + 1 < nr_indices; i += 2) { + OUT_BATCH( indices[i] | + (indices[i + 1] << 16) ); } - if (i < nr) { - OUT_BATCH( vbuf->element_map[i] ); + if (i < nr_indices) { + OUT_BATCH( indices[i] ); } - - vbuf->nr_elements = 0; } - -/** - * Flush vertex buffer. - */ -static void vbuf_flush_vertices( struct draw_stage *stage ) +static void +i915_vbuf_render_release_vertices( struct vbuf_render *render, + void *vertices, + unsigned vertex_size, + unsigned vertices_used ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - struct i915_context *i915 = vbuf->i915; + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; struct pipe_winsys *winsys = i915->pipe.winsys; - if(vbuf->nr_vertices) { - - vbuf_flush_elements(stage); - - winsys->buffer_unmap(winsys, i915->vbo); - - vbuf->nr_vertices = 0; - - /** - * XXX: Reset vertex ids? Actually, want to not do that unless our - * vertex buffer is full. Would like separate - * flush-on-index-full and flush-on-vb-full, but may raise - * issues uploading vertices if the hardware wants to flush when - * we flush. - */ - draw_reset_vertex_ids( vbuf->i915->draw ); - } - - /* FIXME: handle failure */ - if(!i915->vbo) - i915->vbo = winsys->buffer_create(winsys, 64); - winsys->buffer_data( winsys, i915->vbo, - VBUF_SIZE, NULL, - I915_BUFFER_USAGE_LIT_VERTEX ); - - i915->dirty |= I915_NEW_VBO; - - vbuf->vertex_map = winsys->buffer_map(winsys, - i915->vbo, - PIPE_BUFFER_FLAG_WRITE ); - vbuf->vertex_ptr = vbuf->vertex_map; + assert(i915->vbo); + winsys->buffer_unmap(winsys, i915->vbo); + winsys->buffer_reference(winsys, &i915->vbo, NULL); } - -static void vbuf_begin( struct draw_stage *stage ) +static void +i915_vbuf_render_destroy( struct vbuf_render *render ) { - struct vbuf_stage *vbuf = vbuf_stage(stage); - struct i915_context *i915 = vbuf->i915; - unsigned vertex_size = i915->current.vertex_info.size * 4; - - assert(!i915->dirty); - - if(vbuf->vertex_size != vertex_size) - vbuf_flush_vertices(stage); - - vbuf->vertex_size = vertex_size; + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + free(i915_render); } -static void vbuf_end( struct draw_stage *stage ) +/** + * Create a new primitive render. + */ +static struct vbuf_render * +i915_vbuf_render_create( struct i915_context *i915 ) { - /* Overkill. + struct i915_vbuf_render *i915_render = CALLOC_STRUCT(i915_vbuf_render); + + i915_render->i915 = i915; + + i915_render->base.max_vertex_buffer_bytes = 128*1024; + + /* NOTE: it must be such that state and vertices indices fit in a single + * batch buffer. */ - vbuf_flush_elements( stage ); + i915_render->base.max_indices = 16*1024; - stage->point = vbuf_first_point; - stage->line = vbuf_first_line; - stage->tri = vbuf_first_tri; -} - - -static void reset_stipple_counter( struct draw_stage *stage ) -{ + i915_render->base.get_vertex_info = i915_vbuf_render_get_vertex_info; + i915_render->base.allocate_vertices = i915_vbuf_render_allocate_vertices; + i915_render->base.set_primitive = i915_vbuf_render_set_primitive; + i915_render->base.draw = i915_vbuf_render_draw; + i915_render->base.release_vertices = i915_vbuf_render_release_vertices; + i915_render->base.destroy = i915_vbuf_render_destroy; + + return &i915_render->base; } @@ -424,24 +239,18 @@ static void reset_stipple_counter( struct draw_stage *stage ) */ struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) { - struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); - - vbuf->i915 = i915; - vbuf->stage.draw = i915->draw; - vbuf->stage.begin = vbuf_begin; - vbuf->stage.point = vbuf_first_point; - vbuf->stage.line = vbuf_first_line; - vbuf->stage.tri = vbuf_first_tri; - vbuf->stage.end = vbuf_end; - vbuf->stage.reset_stipple_counter = reset_stipple_counter; - - assert(IBUF_SIZE < UNDEFINED_VERTEX_ID); - - /* FIXME: free this memory on takedown */ - vbuf->element_map = MALLOC( IBUF_SIZE ); - vbuf->vertex_map = NULL; + struct vbuf_render *render; + struct draw_stage *stage; - vbuf->vertex_ptr = vbuf->vertex_map; + render = i915_vbuf_render_create(i915); + if(!render) + return NULL; - return &vbuf->stage; + stage = draw_vbuf_stage( i915->draw, render ); + if(!stage) { + render->destroy(render); + return NULL; + } + + return stage; } -- cgit v1.2.3 From 44519be0f5cd70d767c8e29317ebe33a7fb9903e Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 23 Nov 2007 13:27:20 +0000 Subject: gallium: back out winsys interface changes --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 571ad40595..08ac5b672c 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -99,7 +99,7 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, /* FIXME: handle failure */ assert(!i915->vbo); - i915->vbo = winsys->buffer_create(winsys, 64, 0, 0); + i915->vbo = winsys->buffer_create(winsys, 64); winsys->buffer_data( winsys, i915->vbo, size, NULL, I915_BUFFER_USAGE_LIT_VERTEX ); -- cgit v1.2.3 From abd5e8e41d54f7f491f91af9354f19c8d24d3572 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 23 Nov 2007 13:28:16 +0000 Subject: gallium: reorg tgsi directories. --- src/mesa/pipe/draw/draw_private.h | 2 +- src/mesa/pipe/draw/draw_vertex_fetch.c | 3 +- src/mesa/pipe/draw/draw_vertex_shader.c | 2 +- src/mesa/pipe/draw/draw_wide_prims.c | 2 +- src/mesa/pipe/i915simple/i915_fpc_translate.c | 4 +- src/mesa/pipe/i915simple/i915_state_derived.c | 2 +- src/mesa/pipe/p_shader_tokens.h | 805 +++++++++++++ src/mesa/pipe/softpipe/sp_headers.h | 2 +- src/mesa/pipe/softpipe/sp_quad.c | 4 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 1 + src/mesa/pipe/softpipe/sp_state_derived.c | 6 +- src/mesa/pipe/softpipe/sp_state_fs.c | 2 +- src/mesa/pipe/tgsi/deco/Makefile | 3 - src/mesa/pipe/tgsi/deco/deco_caps.c | 10 - src/mesa/pipe/tgsi/deco/deco_caps.h | 112 -- src/mesa/pipe/tgsi/deco/tgsi_deco.h | 8 - src/mesa/pipe/tgsi/exec/tgsi_build.c | 1368 ---------------------- src/mesa/pipe/tgsi/exec/tgsi_build.h | 320 ------ src/mesa/pipe/tgsi/exec/tgsi_core.h | 13 - src/mesa/pipe/tgsi/exec/tgsi_dump.c | 1456 ------------------------ src/mesa/pipe/tgsi/exec/tgsi_dump.h | 22 - src/mesa/pipe/tgsi/exec/tgsi_exec.c | 7 +- src/mesa/pipe/tgsi/exec/tgsi_parse.c | 316 ------ src/mesa/pipe/tgsi/exec/tgsi_parse.h | 121 -- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 9 +- src/mesa/pipe/tgsi/exec/tgsi_token.h | 1494 ------------------------- src/mesa/pipe/tgsi/exec/tgsi_util.c | 270 ----- src/mesa/pipe/tgsi/exec/tgsi_util.h | 70 -- src/mesa/pipe/tgsi/tgsi_platform.h | 16 - src/mesa/pipe/tgsi/util/tgsi_build.c | 1370 +++++++++++++++++++++++ src/mesa/pipe/tgsi/util/tgsi_build.h | 320 ++++++ src/mesa/pipe/tgsi/util/tgsi_dump.c | 1459 ++++++++++++++++++++++++ src/mesa/pipe/tgsi/util/tgsi_dump.h | 22 + src/mesa/pipe/tgsi/util/tgsi_parse.c | 318 ++++++ src/mesa/pipe/tgsi/util/tgsi_parse.h | 121 ++ src/mesa/pipe/tgsi/util/tgsi_util.c | 273 +++++ src/mesa/pipe/tgsi/util/tgsi_util.h | 70 ++ src/mesa/sources | 15 +- src/mesa/state_tracker/st_atom_shader.c | 2 +- src/mesa/state_tracker/st_debug.c | 3 +- src/mesa/state_tracker/st_mesa_to_tgsi.c | 8 +- src/mesa/state_tracker/st_program.c | 2 +- src/mesa/state_tracker/st_program.h | 2 +- 43 files changed, 4801 insertions(+), 5634 deletions(-) create mode 100644 src/mesa/pipe/p_shader_tokens.h delete mode 100644 src/mesa/pipe/tgsi/deco/Makefile delete mode 100644 src/mesa/pipe/tgsi/deco/deco_caps.c delete mode 100644 src/mesa/pipe/tgsi/deco/deco_caps.h delete mode 100644 src/mesa/pipe/tgsi/deco/tgsi_deco.h delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_build.c delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_build.h delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_core.h delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_dump.c delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_dump.h delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_parse.c delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_parse.h delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_token.h delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_util.c delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_util.h delete mode 100644 src/mesa/pipe/tgsi/tgsi_platform.h create mode 100644 src/mesa/pipe/tgsi/util/tgsi_build.c create mode 100644 src/mesa/pipe/tgsi/util/tgsi_build.h create mode 100644 src/mesa/pipe/tgsi/util/tgsi_dump.c create mode 100644 src/mesa/pipe/tgsi/util/tgsi_dump.h create mode 100644 src/mesa/pipe/tgsi/util/tgsi_parse.c create mode 100644 src/mesa/pipe/tgsi/util/tgsi_parse.h create mode 100644 src/mesa/pipe/tgsi/util/tgsi_util.c create mode 100644 src/mesa/pipe/tgsi/util/tgsi_util.h (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 09acf69623..53d7451113 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -47,7 +47,7 @@ #include "draw_vertex.h" #include "x86/rtasm/x86sse.h" -#include "pipe/tgsi/exec/tgsi_core.h" +#include "pipe/tgsi/exec/tgsi_exec.h" struct gallivm_prog; diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 88fa80dddc..5510b3674e 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -31,12 +31,11 @@ */ #include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" #include "draw_private.h" #include "draw_context.h" #include "draw_vertex.h" -#include "pipe/tgsi/exec/tgsi_core.h" - #define DBG 0 diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 52fb2d8596..eef71a7a4a 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -32,13 +32,13 @@ */ #include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" #include "draw_private.h" #include "draw_context.h" #include "draw_vertex.h" #include "x86/rtasm/x86sse.h" -#include "pipe/tgsi/exec/tgsi_core.h" #include "pipe/llvm/gallivm.h" diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index f3b4478f9a..494a2bc619 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -30,10 +30,10 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" #include "draw_private.h" - struct wide_stage { struct draw_stage stage; diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 0382aa26a9..1cd554250c 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -32,8 +32,8 @@ #include "i915_context.h" #include "i915_fpc.h" -#include "pipe/tgsi/exec/tgsi_token.h" -#include "pipe/tgsi/exec/tgsi_parse.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" #include "pipe/draw/draw_vertex.h" diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 688c0c5798..be73769cf2 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -33,7 +33,7 @@ #include "i915_state.h" #include "i915_reg.h" #include "i915_fpc.h" -#include "pipe/tgsi/exec/tgsi_token.h" +#include "pipe/p_shader_tokens.h" /** diff --git a/src/mesa/pipe/p_shader_tokens.h b/src/mesa/pipe/p_shader_tokens.h new file mode 100644 index 0000000000..e5922b439f --- /dev/null +++ b/src/mesa/pipe/p_shader_tokens.h @@ -0,0 +1,805 @@ +#if !defined TGSI_TOKEN_H +#define TGSI_TOKEN_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_version +{ + unsigned MajorVersion : 8; + unsigned MinorVersion : 8; + unsigned Padding : 16; +}; + +struct tgsi_header +{ + unsigned HeaderSize : 8; + unsigned BodySize : 24; +}; + +#define TGSI_PROCESSOR_FRAGMENT 0 +#define TGSI_PROCESSOR_VERTEX 1 +#define TGSI_PROCESSOR_GEOMETRY 2 + +struct tgsi_processor +{ + unsigned Processor : 4; /* TGSI_PROCESSOR_ */ + unsigned Padding : 28; +}; + +#define TGSI_TOKEN_TYPE_DECLARATION 0 +#define TGSI_TOKEN_TYPE_IMMEDIATE 1 +#define TGSI_TOKEN_TYPE_INSTRUCTION 2 + +struct tgsi_token +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_ */ + unsigned Size : 8; /* UINT */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ +}; + +#define TGSI_FILE_NULL 0 +#define TGSI_FILE_CONSTANT 1 +#define TGSI_FILE_INPUT 2 +#define TGSI_FILE_OUTPUT 3 +#define TGSI_FILE_TEMPORARY 4 +#define TGSI_FILE_SAMPLER 5 +#define TGSI_FILE_ADDRESS 6 +#define TGSI_FILE_IMMEDIATE 7 + +#define TGSI_DECLARE_RANGE 0 +#define TGSI_DECLARE_MASK 1 + +#define TGSI_WRITEMASK_NONE 0x00 +#define TGSI_WRITEMASK_X 0x01 +#define TGSI_WRITEMASK_Y 0x02 +#define TGSI_WRITEMASK_XY 0x03 +#define TGSI_WRITEMASK_Z 0x04 +#define TGSI_WRITEMASK_XZ 0x05 +#define TGSI_WRITEMASK_YZ 0x06 +#define TGSI_WRITEMASK_XYZ 0x07 +#define TGSI_WRITEMASK_W 0x08 +#define TGSI_WRITEMASK_XW 0x09 +#define TGSI_WRITEMASK_YW 0x0A +#define TGSI_WRITEMASK_XYW 0x0B +#define TGSI_WRITEMASK_ZW 0x0C +#define TGSI_WRITEMASK_XZW 0x0D +#define TGSI_WRITEMASK_YZW 0x0E +#define TGSI_WRITEMASK_XYZW 0x0F + +struct tgsi_declaration +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ + unsigned Size : 8; /* UINT */ + unsigned File : 4; /* one of TGSI_FILE_x */ + unsigned Declare : 4; /* one of TGSI_DECLARE_x */ + unsigned UsageMask : 4; /* bitmask of TGSI_WRITEMASK_x flags */ + unsigned Interpolate : 1; /* BOOL, any interpolation info? */ + unsigned Semantic : 1; /* BOOL, any semantic info? */ + unsigned Padding : 5; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_declaration_range +{ + unsigned First : 16; /* UINT */ + unsigned Last : 16; /* UINT */ +}; + +struct tgsi_declaration_mask +{ + unsigned Mask : 32; /* UINT */ +}; + +#define TGSI_INTERPOLATE_CONSTANT 0 +#define TGSI_INTERPOLATE_LINEAR 1 +#define TGSI_INTERPOLATE_PERSPECTIVE 2 + +struct tgsi_declaration_interpolation +{ + unsigned Interpolate : 4; /* TGSI_INTERPOLATE_ */ + unsigned Padding : 28; +}; + +#define TGSI_SEMANTIC_POSITION 0 +#define TGSI_SEMANTIC_COLOR 1 +#define TGSI_SEMANTIC_BCOLOR 2 /**< back-face color */ +#define TGSI_SEMANTIC_FOG 3 +#define TGSI_SEMANTIC_PSIZE 4 +#define TGSI_SEMANTIC_GENERIC 5 +#define TGSI_SEMANTIC_COUNT 6 /**< number of semantic values */ + +struct tgsi_declaration_semantic +{ + unsigned SemanticName : 8; /* one of TGSI_SEMANTIC_ */ + unsigned SemanticIndex : 16; /* UINT */ + unsigned Padding : 8; +}; + +#define TGSI_IMM_FLOAT32 0 + +struct tgsi_immediate +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_IMMEDIATE */ + unsigned Size : 8; /* UINT */ + unsigned DataType : 4; /* TGSI_IMM_ */ + unsigned Padding : 15; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_immediate_float32 +{ + float Float; +}; + +/* + * GL_NV_vertex_program + */ +#define TGSI_OPCODE_ARL 0 +#define TGSI_OPCODE_MOV 1 +#define TGSI_OPCODE_LIT 2 +#define TGSI_OPCODE_RCP 3 +#define TGSI_OPCODE_RSQ 4 +#define TGSI_OPCODE_EXP 5 +#define TGSI_OPCODE_LOG 6 +#define TGSI_OPCODE_MUL 7 +#define TGSI_OPCODE_ADD 8 +#define TGSI_OPCODE_DP3 9 +#define TGSI_OPCODE_DP4 10 +#define TGSI_OPCODE_DST 11 +#define TGSI_OPCODE_MIN 12 +#define TGSI_OPCODE_MAX 13 +#define TGSI_OPCODE_SLT 14 +#define TGSI_OPCODE_SGE 15 +#define TGSI_OPCODE_MAD 16 + +/* + * GL_ATI_fragment_shader + */ +#define TGSI_OPCODE_SUB 17 +#define TGSI_OPCODE_DOT3 TGSI_OPCODE_DP3 +#define TGSI_OPCODE_DOT4 TGSI_OPCODE_DP4 +#define TGSI_OPCODE_LERP 18 +#define TGSI_OPCODE_CND 19 +#define TGSI_OPCODE_CND0 20 +#define TGSI_OPCODE_DOT2ADD 21 + +/* + * GL_EXT_vertex_shader + */ +#define TGSI_OPCODE_INDEX 22 +#define TGSI_OPCODE_NEGATE 23 +#define TGSI_OPCODE_MADD TGSI_OPCODE_MAD +#define TGSI_OPCODE_FRAC 24 +#define TGSI_OPCODE_SETGE TGSI_OPCODE_SGE +#define TGSI_OPCODE_SETLT TGSI_OPCODE_SLT +#define TGSI_OPCODE_CLAMP 25 +#define TGSI_OPCODE_FLOOR 26 +#define TGSI_OPCODE_ROUND 27 +#define TGSI_OPCODE_EXPBASE2 28 +#define TGSI_OPCODE_LOGBASE2 29 +#define TGSI_OPCODE_POWER 30 +#define TGSI_OPCODE_RECIP TGSI_OPCODE_RCP +#define TGSI_OPCODE_RECIPSQRT TGSI_OPCODE_RSQ +#define TGSI_OPCODE_CROSSPRODUCT 31 +#define TGSI_OPCODE_MULTIPLYMATRIX 32 + +/* + * GL_NV_vertex_program1_1 + */ +#define TGSI_OPCODE_ABS 33 +#define TGSI_OPCODE_RCC 34 +#define TGSI_OPCODE_DPH 35 + +/* + * GL_NV_fragment_program + */ +#define TGSI_OPCODE_COS 36 +#define TGSI_OPCODE_DDX 37 +#define TGSI_OPCODE_DDY 38 +#define TGSI_OPCODE_EX2 TGSI_OPCODE_EXPBASE2 +#define TGSI_OPCODE_FLR TGSI_OPCODE_FLOOR +#define TGSI_OPCODE_FRC TGSI_OPCODE_FRAC +#define TGSI_OPCODE_KILP 39 /* predicated kill */ +#define TGSI_OPCODE_LG2 TGSI_OPCODE_LOGBASE2 +#define TGSI_OPCODE_LRP TGSI_OPCODE_LERP +#define TGSI_OPCODE_PK2H 40 +#define TGSI_OPCODE_PK2US 41 +#define TGSI_OPCODE_PK4B 42 +#define TGSI_OPCODE_PK4UB 43 +#define TGSI_OPCODE_POW TGSI_OPCODE_POWER +#define TGSI_OPCODE_RFL 44 +#define TGSI_OPCODE_SEQ 45 +#define TGSI_OPCODE_SFL 46 +#define TGSI_OPCODE_SGT 47 +#define TGSI_OPCODE_SIN 48 +#define TGSI_OPCODE_SLE 49 +#define TGSI_OPCODE_SNE 50 +#define TGSI_OPCODE_STR 51 +#define TGSI_OPCODE_TEX 52 +#define TGSI_OPCODE_TXD 53 +#define TGSI_OPCODE_UP2H 54 +#define TGSI_OPCODE_UP2US 55 +#define TGSI_OPCODE_UP4B 56 +#define TGSI_OPCODE_UP4UB 57 +#define TGSI_OPCODE_X2D 58 + +/* + * GL_NV_vertex_program2 + */ +#define TGSI_OPCODE_ARA 59 +#define TGSI_OPCODE_ARR 60 +#define TGSI_OPCODE_BRA 61 +#define TGSI_OPCODE_CAL 62 +#define TGSI_OPCODE_RET 63 +#define TGSI_OPCODE_SSG 64 + +/* + * GL_ARB_vertex_program + */ +#define TGSI_OPCODE_SWZ TGSI_OPCODE_MOV +#define TGSI_OPCODE_XPD TGSI_OPCODE_CROSSPRODUCT + +/* + * GL_ARB_fragment_program + */ +#define TGSI_OPCODE_CMP 65 +#define TGSI_OPCODE_SCS 66 +#define TGSI_OPCODE_TXB 67 + +/* + * GL_NV_fragment_program_option + */ +/* No new opcode */ + +/* + * GL_NV_fragment_program2 + */ +#define TGSI_OPCODE_NRM 68 +#define TGSI_OPCODE_DIV 69 +#define TGSI_OPCODE_DP2 70 +#define TGSI_OPCODE_DP2A TGSI_OPCODE_DOT2ADD +#define TGSI_OPCODE_TXL 71 +#define TGSI_OPCODE_BRK 72 +#define TGSI_OPCODE_IF 73 +#define TGSI_OPCODE_LOOP 74 +#define TGSI_OPCODE_REP 75 +#define TGSI_OPCODE_ELSE 76 +#define TGSI_OPCODE_ENDIF 77 +#define TGSI_OPCODE_ENDLOOP 78 +#define TGSI_OPCODE_ENDREP 79 + +/* + * GL_NV_vertex_program2_option + */ + +/* + * GL_NV_vertex_program3 + */ +#define TGSI_OPCODE_PUSHA 80 +#define TGSI_OPCODE_POPA 81 + +/* + * GL_NV_gpu_program4 + */ +#define TGSI_OPCODE_CEIL 82 +#define TGSI_OPCODE_I2F 83 +#define TGSI_OPCODE_NOT 84 +#define TGSI_OPCODE_TRUNC 85 +#define TGSI_OPCODE_SHL 86 +#define TGSI_OPCODE_SHR 87 +#define TGSI_OPCODE_AND 88 +#define TGSI_OPCODE_OR 89 +#define TGSI_OPCODE_MOD 90 +#define TGSI_OPCODE_XOR 91 +#define TGSI_OPCODE_SAD 92 +#define TGSI_OPCODE_TXF 93 +#define TGSI_OPCODE_TXQ 94 +#define TGSI_OPCODE_CONT 95 + +/* + * GL_NV_vertex_program4 + */ +/* Same as GL_NV_gpu_program4 */ + +/* + * GL_NV_fragment_program4 + */ +/* Same as GL_NV_gpu_program4 */ + +/* + * GL_NV_geometry_program4 + */ +/* Same as GL_NV_gpu_program4 */ +#define TGSI_OPCODE_EMIT 96 +#define TGSI_OPCODE_ENDPRIM 97 + +/* + * GLSL + */ +#define TGSI_OPCODE_BGNLOOP2 98 +#define TGSI_OPCODE_BGNSUB 99 +#define TGSI_OPCODE_ENDLOOP2 100 +#define TGSI_OPCODE_ENDSUB 101 +#define TGSI_OPCODE_INT TGSI_OPCODE_TRUNC +#define TGSI_OPCODE_NOISE1 102 +#define TGSI_OPCODE_NOISE2 103 +#define TGSI_OPCODE_NOISE3 104 +#define TGSI_OPCODE_NOISE4 105 +#define TGSI_OPCODE_NOP 106 + +/* + * ps_1_1 + */ +#define TGSI_OPCODE_TEXCOORD TGSI_OPCODE_NOP +#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL +#define TGSI_OPCODE_TEXBEM 107 +#define TGSI_OPCODE_TEXBEML 108 +#define TGSI_OPCODE_TEXREG2AR 109 +#define TGSI_OPCODE_TEXM3X2PAD 110 +#define TGSI_OPCODE_TEXM3X2TEX 111 +#define TGSI_OPCODE_TEXM3X3PAD 112 +#define TGSI_OPCODE_TEXM3X3TEX 113 +#define TGSI_OPCODE_TEXM3X3SPEC 114 +#define TGSI_OPCODE_TEXM3X3VSPEC 115 + +/* + * ps_1_2 + */ +#define TGSI_OPCODE_TEXREG2GB 116 +#define TGSI_OPCODE_TEXREG2RGB 117 +#define TGSI_OPCODE_TEXDP3TEX 118 +#define TGSI_OPCODE_TEXDP3 119 +#define TGSI_OPCODE_TEXM3X3 120 +/* CMP - use TGSI_OPCODE_CND0 */ + +/* + * ps_1_3 + */ +#define TGSI_OPCODE_TEXM3X2DEPTH 121 +/* CMP - use TGSI_OPCODE_CND0 */ + +/* + * ps_1_4 + */ +#define TGSI_OPCODE_TEXCRD TGSI_OPCODE_TEXCOORD +#define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX +#define TGSI_OPCODE_TEXDEPTH 122 +#define TGSI_OPCODE_BEM 123 + +/* + * ps_2_0 + */ +#define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX +#define TGSI_OPCODE_M4X3 124 +#define TGSI_OPCODE_M3X4 125 +#define TGSI_OPCODE_M3X3 126 +#define TGSI_OPCODE_M3X2 127 +#define TGSI_OPCODE_CRS TGSI_OPCODE_XPD +#define TGSI_OPCODE_NRM4 128 +#define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS +#define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB +#define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A + +/* + * ps_2_x + */ +#define TGSI_OPCODE_CALL TGSI_OPCODE_CAL +#define TGSI_OPCODE_CALLNZ 129 +#define TGSI_OPCODE_IFC 130 +#define TGSI_OPCODE_BREAK TGSI_OPCODE_BRK +#define TGSI_OPCODE_BREAKC 131 +#define TGSI_OPCODE_DSX TGSI_OPCODE_DDX +#define TGSI_OPCODE_DSY TGSI_OPCODE_DDY +#define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD + +/* + * vs_1_1 + */ +#define TGSI_OPCODE_EXPP TGSI_OPCODE_EXP +#define TGSI_OPCODE_LOGP TGSI_OPCODE_LG2 + +/* + * vs_2_0 + */ +#define TGSI_OPCODE_SGN TGSI_OPCODE_SSG +#define TGSI_OPCODE_MOVA TGSI_OPCODE_ARR + +/* + * vs_2_x + */ + +#define TGSI_OPCODE_KIL 132 /* unpredicated kill */ +#define TGSI_OPCODE_END 133 /* aka HALT */ + +#define TGSI_OPCODE_LAST 134 + +#define TGSI_SAT_NONE 0 /* do not saturate */ +#define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ +#define TGSI_SAT_MINUS_PLUS_ONE 2 /* clamp to [-1,1] */ + +/* + * Opcode is the operation code to execute. A given operation defines the + * semantics how the source registers (if any) are interpreted and what is + * written to the destination registers (if any) as a result of execution. + * + * NumDstRegs and NumSrcRegs is the number of destination and source registers, + * respectively. For a given operation code, those numbers are fixed and are + * present here only for convenience. + * + * If Extended is TRUE, it is now executed. + * + * Saturate controls how are final results in destination registers modified. + */ + +struct tgsi_instruction +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_INSTRUCTION */ + unsigned Size : 8; /* UINT */ + unsigned Opcode : 8; /* TGSI_OPCODE_ */ + unsigned Saturate : 2; /* TGSI_SAT_ */ + unsigned NumDstRegs : 2; /* UINT */ + unsigned NumSrcRegs : 4; /* UINT */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows. + * + * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow. + * + * Then, tgsi_instruction::NumSrcRegs of tgsi_src_register follow. + * + * tgsi_instruction::Size contains the total number of words that make the + * instruction, including the instruction word. + */ + +#define TGSI_INSTRUCTION_EXT_TYPE_NV 0 +#define TGSI_INSTRUCTION_EXT_TYPE_LABEL 1 +#define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE 2 +#define TGSI_INSTRUCTION_EXT_TYPE_PREDICATE 3 + +struct tgsi_instruction_ext +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_NV, it should + * be cast to tgsi_instruction_ext_nv. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_LABEL, it + * should be cast to tgsi_instruction_ext_label. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it + * should be cast to tgsi_instruction_ext_texture. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_PREDICATE, it + * should be cast to tgsi_instruction_ext_predicate. + * + * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext + * follows. + */ + +#define TGSI_PRECISION_DEFAULT 0 +#define TGSI_PRECISION_FLOAT32 1 +#define TGSI_PRECISION_FLOAT16 2 +#define TGSI_PRECISION_FIXED12 3 + +#define TGSI_CC_GT 0 +#define TGSI_CC_EQ 1 +#define TGSI_CC_LT 2 +#define TGSI_CC_UN 3 +#define TGSI_CC_GE 4 +#define TGSI_CC_LE 5 +#define TGSI_CC_NE 6 +#define TGSI_CC_TR 7 +#define TGSI_CC_FL 8 + +#define TGSI_SWIZZLE_X 0 +#define TGSI_SWIZZLE_Y 1 +#define TGSI_SWIZZLE_Z 2 +#define TGSI_SWIZZLE_W 3 + +/* + * Precision controls the precision at which the operation should be executed. + * + * CondDstUpdate enables condition code register writes. When this field is + * TRUE, CondDstIndex specifies the index of the condition code register to + * update. + * + * CondFlowEnable enables conditional execution of the operation. When this + * field is TRUE, CondFlowIndex specifies the index of the condition code + * register to test against CondMask with component swizzle controled by + * CondSwizzleX, CondSwizzleY, CondSwizzleZ and CondSwizzleW. If the test fails, + * the operation is not executed. + */ + +struct tgsi_instruction_ext_nv +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_NV */ + unsigned Precision : 4; /* TGSI_PRECISION_ */ + unsigned CondDstIndex : 4; /* UINT */ + unsigned CondFlowIndex : 4; /* UINT */ + unsigned CondMask : 4; /* TGSI_CC_ */ + unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned CondDstUpdate : 1; /* BOOL */ + unsigned CondFlowEnable : 1; /* BOOL */ + unsigned Padding : 1; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_instruction_ext_label +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ + unsigned Label : 24; /* UINT */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ +}; + +#define TGSI_TEXTURE_UNKNOWN 0 +#define TGSI_TEXTURE_1D 1 +#define TGSI_TEXTURE_2D 2 +#define TGSI_TEXTURE_3D 3 +#define TGSI_TEXTURE_CUBE 4 +#define TGSI_TEXTURE_RECT 5 +#define TGSI_TEXTURE_SHADOW1D 6 +#define TGSI_TEXTURE_SHADOW2D 7 +#define TGSI_TEXTURE_SHADOWRECT 8 + +struct tgsi_instruction_ext_texture +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */ + unsigned Texture : 8; /* TGSI_TEXTURE_ */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_instruction_ext_predicate +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ + unsigned PredDstIndex : 4; /* UINT */ + unsigned PredWriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * File specifies the register array to access. + * + * Index specifies the element number of a register in the register file. + * + * If Indirect is TRUE, Index should be offset by the X component of a source + * register that follows. The register can be now fetched into local storage + * for further processing. + * + * If Negate is TRUE, all components of the fetched register are negated. + * + * The fetched register components are swizzled according to SwizzleX, SwizzleY, + * SwizzleZ and SwizzleW. + * + * If Extended is TRUE, any further modifications to the source register are + * made to this temporary storage. + */ + +struct tgsi_src_register +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned Negate : 1; /* BOOL */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_src_register::Extended is TRUE, tgsi_src_register_ext follows. + * + * Then, if tgsi_src_register::Indirect is TRUE, another tgsi_src_register + * follows. + * + * Then, if tgsi_src_register::Dimension is TRUE, tgsi_dimension follows. + */ + +#define TGSI_SRC_REGISTER_EXT_TYPE_SWZ 0 +#define TGSI_SRC_REGISTER_EXT_TYPE_MOD 1 + +struct tgsi_src_register_ext +{ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_SWZ, + * it should be cast to tgsi_src_register_ext_extswz. + * + * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_MOD, + * it should be cast to tgsi_src_register_ext_mod. + * + * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext + * follows. + */ + +#define TGSI_EXTSWIZZLE_X TGSI_SWIZZLE_X +#define TGSI_EXTSWIZZLE_Y TGSI_SWIZZLE_Y +#define TGSI_EXTSWIZZLE_Z TGSI_SWIZZLE_Z +#define TGSI_EXTSWIZZLE_W TGSI_SWIZZLE_W +#define TGSI_EXTSWIZZLE_ZERO 4 +#define TGSI_EXTSWIZZLE_ONE 5 + +/* + * ExtSwizzleX, ExtSwizzleY, ExtSwizzleZ and ExtSwizzleW swizzle the source + * register in an extended manner. + * + * NegateX, NegateY, NegateZ and NegateW negate individual components of the + * source register. + * + * ExtDivide specifies which component is used to divide all components of the + * source register. + */ + +struct tgsi_src_register_ext_swz +{ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */ + unsigned ExtSwizzleX : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleY : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleZ : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleW : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned NegateX : 1; /* BOOL */ + unsigned NegateY : 1; /* BOOL */ + unsigned NegateZ : 1; /* BOOL */ + unsigned NegateW : 1; /* BOOL */ + unsigned ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ +}; + +/** + * Extra src register modifiers + * + * If Complement is TRUE, the source register is modified by subtracting it + * from 1.0. + * + * If Bias is TRUE, the source register is modified by subtracting 0.5 from it. + * + * If Scale2X is TRUE, the source register is modified by multiplying it by 2.0. + * + * If Absolute is TRUE, the source register is modified by removing the sign. + * + * If Negate is TRUE, the source register is modified by negating it. + */ + +struct tgsi_src_register_ext_mod +{ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */ + unsigned Complement : 1; /* BOOL */ + unsigned Bias : 1; /* BOOL */ + unsigned Scale2X : 1; /* BOOL */ + unsigned Absolute : 1; /* BOOL */ + unsigned Negate : 1; /* BOOL */ + unsigned Padding : 22; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_dimension +{ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + unsigned Padding : 13; + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_dst_register +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Padding : 5; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_dst_register::Extended is TRUE, tgsi_dst_register_ext follows. + * + * Then, if tgsi_dst_register::Indirect is TRUE, tgsi_src_register follows. + */ + +#define TGSI_DST_REGISTER_EXT_TYPE_CONDCODE 0 +#define TGSI_DST_REGISTER_EXT_TYPE_MODULATE 1 +#define TGSI_DST_REGISTER_EXT_TYPE_PREDICATE 2 + +struct tgsi_dst_register_ext +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ +}; + +/** + * Extra destination register modifiers + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_CONDCODE, + * it should be cast to tgsi_dst_register_ext_condcode. + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_MODULATE, + * it should be cast to tgsi_dst_register_ext_modulate. + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_PREDICATE, + * it should be cast to tgsi_dst_register_ext_predicate. + * + * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext + * follows. + */ +struct tgsi_dst_register_ext_concode +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ + unsigned CondMask : 4; /* TGSI_CC_ */ + unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSrcIndex : 4; /* UINT */ + unsigned Padding : 11; + unsigned Extended : 1; /* BOOL */ +}; + +#define TGSI_MODULATE_1X 0 +#define TGSI_MODULATE_2X 1 +#define TGSI_MODULATE_4X 2 +#define TGSI_MODULATE_8X 3 +#define TGSI_MODULATE_HALF 4 +#define TGSI_MODULATE_QUARTER 5 +#define TGSI_MODULATE_EIGHTH 6 + +struct tgsi_dst_register_ext_modulate +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */ + unsigned Modulate : 4; /* TGSI_MODULATE_ */ + unsigned Padding : 23; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * Currently, the following constraints apply. + * + * - PredSwizzleXYZW is either set to identity or replicate. + * - PredSrcIndex is 0. + */ + +struct tgsi_dst_register_ext_predicate +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */ + unsigned PredSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSrcIndex : 4; /* UINT */ + unsigned Negate : 1; /* BOOL */ + unsigned Padding : 14; + unsigned Extended : 1; /* BOOL */ +}; + + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_TOKEN_H + diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index e23742f803..b9f2b2205a 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -31,7 +31,7 @@ #ifndef SP_HEADERS_H #define SP_HEADERS_H -#include "../tgsi/exec/tgsi_core.h" +#include "pipe/tgsi/exec/tgsi_exec.h" #define PRIM_POINT 1 #define PRIM_LINE 2 diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 5a0df6de9d..13fb883ef0 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -27,9 +27,9 @@ #include "sp_context.h" - +#include "sp_headers.h" #include "sp_state.h" -#include "pipe/tgsi/exec/tgsi_token.h" +#include "pipe/p_shader_tokens.h" static void sp_push_quad_first( diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index ed14dac18e..24c8a44c47 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -37,6 +37,7 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" #include "x86/rtasm/x86sse.h" diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 7e5efbfde3..33caab9372 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -26,15 +26,13 @@ **************************************************************************/ #include "pipe/p_util.h" - +#include "pipe/p_shader_tokens.h" #include "pipe/draw/draw_context.h" #include "pipe/draw/draw_vertex.h" - +#include "sp_headers.h" #include "sp_context.h" #include "sp_state.h" -#include "pipe/tgsi/exec/tgsi_token.h" - /** * Determine which post-transform / pre-rasterization vertex attributes diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index ba564b16e6..912f42d568 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -32,7 +32,7 @@ #include "pipe/p_util.h" #include "pipe/p_winsys.h" #include "pipe/draw/draw_context.h" -#include "pipe/tgsi/exec/tgsi_core.h" +#include "pipe/p_shader_tokens.h" #include "pipe/llvm/gallivm.h" diff --git a/src/mesa/pipe/tgsi/deco/Makefile b/src/mesa/pipe/tgsi/deco/Makefile deleted file mode 100644 index eb8b14e0e8..0000000000 --- a/src/mesa/pipe/tgsi/deco/Makefile +++ /dev/null @@ -1,3 +0,0 @@ -default: - cd ../../.. ; make - diff --git a/src/mesa/pipe/tgsi/deco/deco_caps.c b/src/mesa/pipe/tgsi/deco/deco_caps.c deleted file mode 100644 index 66df05a14b..0000000000 --- a/src/mesa/pipe/tgsi/deco/deco_caps.c +++ /dev/null @@ -1,10 +0,0 @@ -#include "tgsi_platform.h" -#include "tgsi_deco.h" - -void -tgsi_deco_caps_init( - struct tgsi_deco_caps *caps ) -{ - memset( caps, 0, sizeof( *caps ) ); -} - diff --git a/src/mesa/pipe/tgsi/deco/deco_caps.h b/src/mesa/pipe/tgsi/deco/deco_caps.h deleted file mode 100644 index 40ca1c96f3..0000000000 --- a/src/mesa/pipe/tgsi/deco/deco_caps.h +++ /dev/null @@ -1,112 +0,0 @@ -#if !defined DECO_CAPS_H -#define DECO_CAPS_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_deco_caps -{ - /* - * Predicates (D3D9-specific). - * - * Constraints: - * 1. Token tgsi_dst_register_ext_predicate must not be used. - * 2. Token tgsi_instruction_ext_predicate must not be used. - */ - unsigned Predicates : 1; - - /* - * Destination register post-modulate. - * - * Constraints: - * 1. Field tgsi_dst_register_ext_modulate::Modulate - * must be set to TGSI_MODULATE_1X. - */ - unsigned DstModulate : 1; - - /* - * Condition codes (NVIDIA-specific). - * - * Constraints: - * 1. Token tgsi_dst_register_ext_concode must not be used. - * 2. Field tgsi_instruction_ext_nv::CondDstUpdate must be set to FALSE. - * 3. Field tgsi_instruction_ext_nv::CondFlowEnable must be set to FALSE. - */ - unsigned ConCodes : 1; - - /* - * Source register invert. - * - * Constraints: - * 1. Field tgsi_src_register_ext_mod::Complement must be set to FALSE. - */ - unsigned SrcInvert : 1; - - /* - * Source register bias. - * - * Constraints: - * 1. Field tgsi_src_register_ext_mod::Bias must be set to FALSE. - */ - unsigned SrcBias : 1; - - /* - * Source register scale by 2. - * - * Constraints: - * 1. Field tgsi_src_register_ext_mod::Scale2X must be set to FALSE. - */ - unsigned SrcScale : 1; - - /* - * Source register absolute. - * - * Constraints: - * 1. Field tgsi_src_register_ext_mod::Absolute must be set to FALSE. - */ - unsigned SrcAbsolute : 1; - - /* - * Source register force sign. - * - * Constraints: - * 1. Fields tgsi_src_register_ext_mod::Absolute and - * tgsi_src_register_ext_mod::Negate must not be both set to TRUE - * at the same time. - */ - unsigned SrcForceSign : 1; - - /* - * Source register divide. - * - * Constraints: - * 1. Field tgsi_src_register_ext_swz::ExtDivide - * must be set to TGSI_EXTSWIZZLE_ONE. - */ - unsigned SrcDivide : 1; - - /* - * Source register extended swizzle. - * - * Constraints: - * 1. Field tgsi_src_register_ext_swz::ExtSwizzleX/Y/Z/W - * must be set to TGSI_EXTSWIZZLE_X/Y/Z/W. - * 2. Fields tgsi_src_register_ext_swz::NegateX/Y/Z/W - * must all be set to the same value. - */ - unsigned SrcExtSwizzle : 1; - - unsigned Padding : 22; -}; - -void -tgsi_deco_caps_init( - struct tgsi_deco_caps *caps ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined DECO_CAPS_H - diff --git a/src/mesa/pipe/tgsi/deco/tgsi_deco.h b/src/mesa/pipe/tgsi/deco/tgsi_deco.h deleted file mode 100644 index 3560e455f3..0000000000 --- a/src/mesa/pipe/tgsi/deco/tgsi_deco.h +++ /dev/null @@ -1,8 +0,0 @@ -#if !defined TGSI_DECO_H -#define TGSI_DECO_H - -#include "../exec/tgsi_core.h" -#include "deco_caps.h" - -#endif // !defined TGSI_DECO_H - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.c b/src/mesa/pipe/tgsi/exec/tgsi_build.c deleted file mode 100644 index 78f648aae2..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_build.c +++ /dev/null @@ -1,1368 +0,0 @@ -#include "tgsi_platform.h" -#include "tgsi_core.h" - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ) -{ - struct tgsi_version version; - - version.MajorVersion = 1; - version.MinorVersion = 1; - version.Padding = 0; - - return version; -} - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ) -{ - struct tgsi_header header; - - header.HeaderSize = 1; - header.BodySize = 0; - - return header; -} - -static void -header_headersize_grow( struct tgsi_header *header ) -{ - assert( header->HeaderSize < 0xFF ); - assert( header->BodySize == 0 ); - - header->HeaderSize++; -} - -static void -header_bodysize_grow( struct tgsi_header *header ) -{ - assert( header->BodySize < 0xFFFFFF ); - - header->BodySize++; -} - -struct tgsi_processor -tgsi_default_processor( void ) -{ - struct tgsi_processor processor; - - processor.Processor = TGSI_PROCESSOR_FRAGMENT; - processor.Padding = 0; - - return processor; -} - -struct tgsi_processor -tgsi_build_processor( - unsigned type, - struct tgsi_header *header ) -{ - struct tgsi_processor processor; - - processor = tgsi_default_processor(); - processor.Processor = type; - - header_headersize_grow( header ); - - return processor; -} - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ) -{ - struct tgsi_declaration declaration; - - declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; - declaration.Size = 1; - declaration.File = TGSI_FILE_NULL; - declaration.Declare = TGSI_DECLARE_RANGE; - declaration.UsageMask = TGSI_WRITEMASK_XYZW; - declaration.Interpolate = 0; - declaration.Semantic = 0; - declaration.Padding = 0; - declaration.Extended = 0; - - return declaration; -} - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned declare, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ) -{ - struct tgsi_declaration declaration; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( declare <= TGSI_DECLARE_MASK ); - - declaration = tgsi_default_declaration(); - declaration.File = file; - declaration.Declare = declare; - declaration.UsageMask = usage_mask; - declaration.Interpolate = interpolate; - declaration.Semantic = semantic; - - header_bodysize_grow( header ); - - return declaration; -} - -static void -declaration_grow( - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - assert( declaration->Size < 0xFF ); - - declaration->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ) -{ - struct tgsi_full_declaration full_declaration; - - full_declaration.Declaration = tgsi_default_declaration(); - full_declaration.Interpolation = tgsi_default_declaration_interpolation(); - full_declaration.Semantic = tgsi_default_declaration_semantic(); - - return full_declaration; -} - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0; - struct tgsi_declaration *declaration; - - if( maxsize <= size ) - return 0; - declaration = (struct tgsi_declaration *) &tokens[size]; - size++; - - *declaration = tgsi_build_declaration( - full_decl->Declaration.File, - full_decl->Declaration.Declare, - full_decl->Declaration.UsageMask, - full_decl->Declaration.Interpolate, - full_decl->Declaration.Semantic, - header ); - - switch( full_decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - { - struct tgsi_declaration_range *dr; - - if( maxsize <= size ) - return 0; - dr = (struct tgsi_declaration_range *) &tokens[size]; - size++; - - *dr = tgsi_build_declaration_range( - full_decl->u.DeclarationRange.First, - full_decl->u.DeclarationRange.Last, - declaration, - header ); - break; - } - - case TGSI_DECLARE_MASK: - { - struct tgsi_declaration_mask *dm; - - if( maxsize <= size ) - return 0; - dm = (struct tgsi_declaration_mask *) &tokens[size]; - size++; - - *dm = tgsi_build_declaration_mask( - full_decl->u.DeclarationMask.Mask, - declaration, - header ); - break; - } - - default: - assert( 0 ); - } - - if( full_decl->Declaration.Interpolate ) { - struct tgsi_declaration_interpolation *di; - - if( maxsize <= size ) - return 0; - di = (struct tgsi_declaration_interpolation *) &tokens[size]; - size++; - - *di = tgsi_build_declaration_interpolation( - full_decl->Interpolation.Interpolate, - declaration, - header ); - } - - if( full_decl->Declaration.Semantic ) { - struct tgsi_declaration_semantic *ds; - - if( maxsize <= size ) - return 0; - ds = (struct tgsi_declaration_semantic *) &tokens[size]; - size++; - - *ds = tgsi_build_declaration_semantic( - full_decl->Semantic.SemanticName, - full_decl->Semantic.SemanticIndex, - declaration, - header ); - } - - return size; -} - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_range declaration_range; - - assert( last >= first ); - assert( last <= 0xFFFF ); - - declaration_range.First = first; - declaration_range.Last = last; - - declaration_grow( declaration, header ); - - return declaration_range; -} - -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_mask declaration_mask; - - declaration_mask.Mask = mask; - - declaration_grow( declaration, header ); - - return declaration_mask; -} - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ) -{ - struct tgsi_declaration_interpolation di; - - di.Interpolate = TGSI_INTERPOLATE_CONSTANT; - di.Padding = 0; - - return di; -} - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_interpolation di; - - assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); - - di = tgsi_default_declaration_interpolation(); - di.Interpolate = interpolate; - - declaration_grow( declaration, header ); - - return di; -} - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ) -{ - struct tgsi_declaration_semantic ds; - - ds.SemanticName = TGSI_SEMANTIC_POSITION; - ds.SemanticIndex = 0; - ds.Padding = 0; - - return ds; -} - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_semantic ds; - - assert( semantic_name <= TGSI_SEMANTIC_COUNT ); - assert( semantic_index <= 0xFFFF ); - - ds = tgsi_default_declaration_semantic(); - ds.SemanticName = semantic_name; - ds.SemanticIndex = semantic_index; - - declaration_grow( declaration, header ); - - return ds; -} - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ) -{ - struct tgsi_immediate immediate; - - immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; - immediate.Size = 1; - immediate.DataType = TGSI_IMM_FLOAT32; - immediate.Padding = 0; - immediate.Extended = 0; - - return immediate; -} - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ) -{ - struct tgsi_immediate immediate; - - immediate = tgsi_default_immediate(); - - header_bodysize_grow( header ); - - return immediate; -} - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ) -{ - struct tgsi_full_immediate fullimm; - - fullimm.Immediate = tgsi_default_immediate(); - fullimm.u.Pointer = (void *) 0; - - return fullimm; -} - -static void -immediate_grow( - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - assert( immediate->Size < 0xFF ); - - immediate->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - float value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - struct tgsi_immediate_float32 immediate_float32; - - immediate_float32.Float = value; - - immediate_grow( immediate, header ); - - return immediate_float32; -} - -unsigned -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0, i; - struct tgsi_immediate *immediate; - - if( maxsize <= size ) - return 0; - immediate = (struct tgsi_immediate *) &tokens[size]; - size++; - - *immediate = tgsi_build_immediate( header ); - - for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { - struct tgsi_immediate_float32 *if32; - - if( maxsize <= size ) - return 0; - if32 = (struct tgsi_immediate_float32 *) &tokens[size]; - size++; - - *if32 = tgsi_build_immediate_float32( - full_imm->u.ImmediateFloat32[i].Float, - immediate, - header ); - } - - return size; -} - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ) -{ - struct tgsi_instruction instruction; - - instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; - instruction.Size = 1; - instruction.Opcode = TGSI_OPCODE_MOV; - instruction.Saturate = TGSI_SAT_NONE; - instruction.NumDstRegs = 1; - instruction.NumSrcRegs = 1; - instruction.Padding = 0; - instruction.Extended = 0; - - return instruction; -} - -struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ) -{ - struct tgsi_instruction instruction; - - assert (opcode <= TGSI_OPCODE_LAST); - assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); - assert (num_dst_regs <= 3); - assert (num_src_regs <= 15); - - instruction = tgsi_default_instruction(); - instruction.Opcode = opcode; - instruction.Saturate = saturate; - instruction.NumDstRegs = num_dst_regs; - instruction.NumSrcRegs = num_src_regs; - - header_bodysize_grow( header ); - - return instruction; -} - -static void -instruction_grow( - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - assert (instruction->Size < 0xFF); - - instruction->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ) -{ - struct tgsi_full_instruction full_instruction; - unsigned i; - - full_instruction.Instruction = tgsi_default_instruction(); - full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); - full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); - full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); - for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { - full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); - } - for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { - full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); - } - - return full_instruction; -} - -unsigned -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0; - unsigned i; - struct tgsi_instruction *instruction; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - instruction = (struct tgsi_instruction *) &tokens[size]; - size++; - - *instruction = tgsi_build_instruction( - full_inst->Instruction.Opcode, - full_inst->Instruction.Saturate, - full_inst->Instruction.NumDstRegs, - full_inst->Instruction.NumSrcRegs, - header ); - prev_token = (struct tgsi_token *) instruction; - - if( tgsi_compare_instruction_ext_nv( - full_inst->InstructionExtNv, - tgsi_default_instruction_ext_nv() ) ) { - struct tgsi_instruction_ext_nv *instruction_ext_nv; - - if( maxsize <= size ) - return 0; - instruction_ext_nv = - (struct tgsi_instruction_ext_nv *) &tokens[size]; - size++; - - *instruction_ext_nv = tgsi_build_instruction_ext_nv( - full_inst->InstructionExtNv.Precision, - full_inst->InstructionExtNv.CondDstIndex, - full_inst->InstructionExtNv.CondFlowIndex, - full_inst->InstructionExtNv.CondMask, - full_inst->InstructionExtNv.CondSwizzleX, - full_inst->InstructionExtNv.CondSwizzleY, - full_inst->InstructionExtNv.CondSwizzleZ, - full_inst->InstructionExtNv.CondSwizzleW, - full_inst->InstructionExtNv.CondDstUpdate, - full_inst->InstructionExtNv.CondFlowEnable, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_nv; - } - - if( tgsi_compare_instruction_ext_label( - full_inst->InstructionExtLabel, - tgsi_default_instruction_ext_label() ) ) { - struct tgsi_instruction_ext_label *instruction_ext_label; - - if( maxsize <= size ) - return 0; - instruction_ext_label = - (struct tgsi_instruction_ext_label *) &tokens[size]; - size++; - - *instruction_ext_label = tgsi_build_instruction_ext_label( - full_inst->InstructionExtLabel.Label, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_label; - } - - if( tgsi_compare_instruction_ext_texture( - full_inst->InstructionExtTexture, - tgsi_default_instruction_ext_texture() ) ) { - struct tgsi_instruction_ext_texture *instruction_ext_texture; - - if( maxsize <= size ) - return 0; - instruction_ext_texture = - (struct tgsi_instruction_ext_texture *) &tokens[size]; - size++; - - *instruction_ext_texture = tgsi_build_instruction_ext_texture( - full_inst->InstructionExtTexture.Texture, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_texture; - } - - for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { - const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; - struct tgsi_dst_register *dst_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - dst_register = (struct tgsi_dst_register *) &tokens[size]; - size++; - - *dst_register = tgsi_build_dst_register( - reg->DstRegister.File, - reg->DstRegister.WriteMask, - reg->DstRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register; - - if( tgsi_compare_dst_register_ext_concode( - reg->DstRegisterExtConcode, - tgsi_default_dst_register_ext_concode() ) ) { - struct tgsi_dst_register_ext_concode *dst_register_ext_concode; - - if( maxsize <= size ) - return 0; - dst_register_ext_concode = - (struct tgsi_dst_register_ext_concode *) &tokens[size]; - size++; - - *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( - reg->DstRegisterExtConcode.CondMask, - reg->DstRegisterExtConcode.CondSwizzleX, - reg->DstRegisterExtConcode.CondSwizzleY, - reg->DstRegisterExtConcode.CondSwizzleZ, - reg->DstRegisterExtConcode.CondSwizzleW, - reg->DstRegisterExtConcode.CondSrcIndex, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_concode; - } - - if( tgsi_compare_dst_register_ext_modulate( - reg->DstRegisterExtModulate, - tgsi_default_dst_register_ext_modulate() ) ) { - struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; - - if( maxsize <= size ) - return 0; - dst_register_ext_modulate = - (struct tgsi_dst_register_ext_modulate *) &tokens[size]; - size++; - - *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( - reg->DstRegisterExtModulate.Modulate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_modulate; - } - } - - for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { - const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; - struct tgsi_src_register *src_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - src_register = (struct tgsi_src_register *) &tokens[size]; - size++; - - *src_register = tgsi_build_src_register( - reg->SrcRegister.File, - reg->SrcRegister.SwizzleX, - reg->SrcRegister.SwizzleY, - reg->SrcRegister.SwizzleZ, - reg->SrcRegister.SwizzleW, - reg->SrcRegister.Negate, - reg->SrcRegister.Indirect, - reg->SrcRegister.Dimension, - reg->SrcRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register; - - if( tgsi_compare_src_register_ext_swz( - reg->SrcRegisterExtSwz, - tgsi_default_src_register_ext_swz() ) ) { - struct tgsi_src_register_ext_swz *src_register_ext_swz; - - if( maxsize <= size ) - return 0; - src_register_ext_swz = - (struct tgsi_src_register_ext_swz *) &tokens[size]; - size++; - - *src_register_ext_swz = tgsi_build_src_register_ext_swz( - reg->SrcRegisterExtSwz.ExtSwizzleX, - reg->SrcRegisterExtSwz.ExtSwizzleY, - reg->SrcRegisterExtSwz.ExtSwizzleZ, - reg->SrcRegisterExtSwz.ExtSwizzleW, - reg->SrcRegisterExtSwz.NegateX, - reg->SrcRegisterExtSwz.NegateY, - reg->SrcRegisterExtSwz.NegateZ, - reg->SrcRegisterExtSwz.NegateW, - reg->SrcRegisterExtSwz.ExtDivide, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_swz; - } - - if( tgsi_compare_src_register_ext_mod( - reg->SrcRegisterExtMod, - tgsi_default_src_register_ext_mod() ) ) { - struct tgsi_src_register_ext_mod *src_register_ext_mod; - - if( maxsize <= size ) - return 0; - src_register_ext_mod = - (struct tgsi_src_register_ext_mod *) &tokens[size]; - size++; - - *src_register_ext_mod = tgsi_build_src_register_ext_mod( - reg->SrcRegisterExtMod.Complement, - reg->SrcRegisterExtMod.Bias, - reg->SrcRegisterExtMod.Scale2X, - reg->SrcRegisterExtMod.Absolute, - reg->SrcRegisterExtMod.Negate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_mod; - } - - if( reg->SrcRegister.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterInd.File, - reg->SrcRegisterInd.SwizzleX, - reg->SrcRegisterInd.SwizzleY, - reg->SrcRegisterInd.SwizzleZ, - reg->SrcRegisterInd.SwizzleW, - reg->SrcRegisterInd.Negate, - reg->SrcRegisterInd.Indirect, - reg->SrcRegisterInd.Dimension, - reg->SrcRegisterInd.Index, - instruction, - header ); - } - - if( reg->SrcRegister.Dimension ) { - struct tgsi_dimension *dim; - - assert( !reg->SrcRegisterDim.Dimension ); - - if( maxsize <= size ) - return 0; - dim = (struct tgsi_dimension *) &tokens[size]; - size++; - - *dim = tgsi_build_dimension( - reg->SrcRegisterDim.Indirect, - reg->SrcRegisterDim.Index, - instruction, - header ); - - if( reg->SrcRegisterDim.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterDimInd.File, - reg->SrcRegisterDimInd.SwizzleX, - reg->SrcRegisterDimInd.SwizzleY, - reg->SrcRegisterDimInd.SwizzleZ, - reg->SrcRegisterDimInd.SwizzleW, - reg->SrcRegisterDimInd.Negate, - reg->SrcRegisterDimInd.Indirect, - reg->SrcRegisterDimInd.Dimension, - reg->SrcRegisterDimInd.Index, - instruction, - header ); - } - } - } - - return size; -} - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; - instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; - instruction_ext_nv.CondDstIndex = 0; - instruction_ext_nv.CondFlowIndex = 0; - instruction_ext_nv.CondMask = TGSI_CC_TR; - instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; - instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; - instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; - instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; - instruction_ext_nv.CondDstUpdate = 0; - instruction_ext_nv.CondFlowEnable = 0; - instruction_ext_nv.Padding = 0; - instruction_ext_nv.Extended = 0; - - return instruction_ext_nv; -} - -union token_u32 -{ - unsigned u32; -}; - -unsigned -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - unsigned precision, - unsigned cond_dst_index, - unsigned cond_flow_index, - unsigned cond_mask, - unsigned cond_swizzle_x, - unsigned cond_swizzle_y, - unsigned cond_swizzle_z, - unsigned cond_swizzle_w, - unsigned cond_dst_update, - unsigned cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv = tgsi_default_instruction_ext_nv(); - instruction_ext_nv.Precision = precision; - instruction_ext_nv.CondDstIndex = cond_dst_index; - instruction_ext_nv.CondFlowIndex = cond_flow_index; - instruction_ext_nv.CondMask = cond_mask; - instruction_ext_nv.CondSwizzleX = cond_swizzle_x; - instruction_ext_nv.CondSwizzleY = cond_swizzle_y; - instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; - instruction_ext_nv.CondSwizzleW = cond_swizzle_w; - instruction_ext_nv.CondDstUpdate = cond_dst_update; - instruction_ext_nv.CondFlowEnable = cond_flow_update; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_nv; -} - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; - instruction_ext_label.Label = 0; - instruction_ext_label.Padding = 0; - instruction_ext_label.Extended = 0; - - return instruction_ext_label; -} - -unsigned -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - unsigned label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label = tgsi_default_instruction_ext_label(); - instruction_ext_label.Label = label; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_label; -} - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; - instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; - instruction_ext_texture.Padding = 0; - instruction_ext_texture.Extended = 0; - - return instruction_ext_texture; -} - -unsigned -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - unsigned texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture = tgsi_default_instruction_ext_texture(); - instruction_ext_texture.Texture = texture; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_texture; -} - -struct tgsi_src_register -tgsi_default_src_register( void ) -{ - struct tgsi_src_register src_register; - - src_register.File = TGSI_FILE_NULL; - src_register.SwizzleX = TGSI_SWIZZLE_X; - src_register.SwizzleY = TGSI_SWIZZLE_Y; - src_register.SwizzleZ = TGSI_SWIZZLE_Z; - src_register.SwizzleW = TGSI_SWIZZLE_W; - src_register.Negate = 0; - src_register.Indirect = 0; - src_register.Dimension = 0; - src_register.Index = 0; - src_register.Extended = 0; - - return src_register; -} - -struct tgsi_src_register -tgsi_build_src_register( - unsigned file, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - unsigned negate, - unsigned indirect, - unsigned dimension, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register src_register; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( swizzle_x <= TGSI_SWIZZLE_W ); - assert( swizzle_y <= TGSI_SWIZZLE_W ); - assert( swizzle_z <= TGSI_SWIZZLE_W ); - assert( swizzle_w <= TGSI_SWIZZLE_W ); - assert( negate <= 1 ); - assert( index >= -0x8000 && index <= 0x7FFF ); - - src_register = tgsi_default_src_register(); - src_register.File = file; - src_register.SwizzleX = swizzle_x; - src_register.SwizzleY = swizzle_y; - src_register.SwizzleZ = swizzle_z; - src_register.SwizzleW = swizzle_w; - src_register.Negate = negate; - src_register.Indirect = indirect; - src_register.Dimension = dimension; - src_register.Index = index; - - instruction_grow( instruction, header ); - - return src_register; -} - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ) -{ - struct tgsi_full_src_register full_src_register; - - full_src_register.SrcRegister = tgsi_default_src_register(); - full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); - full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); - full_src_register.SrcRegisterInd = tgsi_default_src_register(); - full_src_register.SrcRegisterDim = tgsi_default_dimension(); - full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); - - return full_src_register; -} - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; - src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; - src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; - src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; - src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; - src_register_ext_swz.NegateX = 0; - src_register_ext_swz.NegateY = 0; - src_register_ext_swz.NegateZ = 0; - src_register_ext_swz.NegateW = 0; - src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; - src_register_ext_swz.Padding = 0; - src_register_ext_swz.Extended = 0; - - return src_register_ext_swz; -} - -unsigned -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - unsigned ext_swizzle_x, - unsigned ext_swizzle_y, - unsigned ext_swizzle_z, - unsigned ext_swizzle_w, - unsigned negate_x, - unsigned negate_y, - unsigned negate_z, - unsigned negate_w, - unsigned ext_divide, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE ); - assert( negate_x <= 1 ); - assert( negate_y <= 1 ); - assert( negate_z <= 1 ); - assert( negate_w <= 1 ); - assert( ext_divide <= TGSI_EXTSWIZZLE_ONE ); - - src_register_ext_swz = tgsi_default_src_register_ext_swz(); - src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; - src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; - src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; - src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; - src_register_ext_swz.NegateX = negate_x; - src_register_ext_swz.NegateY = negate_y; - src_register_ext_swz.NegateZ = negate_z; - src_register_ext_swz.NegateW = negate_w; - src_register_ext_swz.ExtDivide = ext_divide; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_swz; -} - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; - src_register_ext_mod.Complement = 0; - src_register_ext_mod.Bias = 0; - src_register_ext_mod.Scale2X = 0; - src_register_ext_mod.Absolute = 0; - src_register_ext_mod.Negate = 0; - src_register_ext_mod.Padding = 0; - src_register_ext_mod.Extended = 0; - - return src_register_ext_mod; -} - -unsigned -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - unsigned complement, - unsigned bias, - unsigned scale_2x, - unsigned absolute, - unsigned negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - assert( complement <= 1 ); - assert( bias <= 1 ); - assert( scale_2x <= 1 ); - assert( absolute <= 1 ); - assert( negate <= 1 ); - - src_register_ext_mod = tgsi_default_src_register_ext_mod(); - src_register_ext_mod.Complement = complement; - src_register_ext_mod.Bias = bias; - src_register_ext_mod.Scale2X = scale_2x; - src_register_ext_mod.Absolute = absolute; - src_register_ext_mod.Negate = negate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_mod; -} - -struct tgsi_dimension -tgsi_default_dimension( void ) -{ - struct tgsi_dimension dimension; - - dimension.Indirect = 0; - dimension.Dimension = 0; - dimension.Padding = 0; - dimension.Index = 0; - dimension.Extended = 0; - - return dimension; -} - -struct tgsi_dimension -tgsi_build_dimension( - unsigned indirect, - unsigned index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dimension dimension; - - dimension = tgsi_default_dimension(); - dimension.Indirect = indirect; - dimension.Index = index; - - instruction_grow( instruction, header ); - - return dimension; -} - -struct tgsi_dst_register -tgsi_default_dst_register( void ) -{ - struct tgsi_dst_register dst_register; - - dst_register.File = TGSI_FILE_NULL; - dst_register.WriteMask = TGSI_WRITEMASK_XYZW; - dst_register.Indirect = 0; - dst_register.Dimension = 0; - dst_register.Index = 0; - dst_register.Padding = 0; - dst_register.Extended = 0; - - return dst_register; -} - -struct tgsi_dst_register -tgsi_build_dst_register( - unsigned file, - unsigned mask, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register dst_register; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( mask <= TGSI_WRITEMASK_XYZW ); - assert( index >= -32768 && index <= 32767 ); - - dst_register = tgsi_default_dst_register(); - dst_register.File = file; - dst_register.WriteMask = mask; - dst_register.Index = index; - - instruction_grow( instruction, header ); - - return dst_register; -} - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ) -{ - struct tgsi_full_dst_register full_dst_register; - - full_dst_register.DstRegister = tgsi_default_dst_register(); - full_dst_register.DstRegisterExtConcode = - tgsi_default_dst_register_ext_concode(); - full_dst_register.DstRegisterExtModulate = - tgsi_default_dst_register_ext_modulate(); - - return full_dst_register; -} - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; - dst_register_ext_concode.CondMask = TGSI_CC_TR; - dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; - dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; - dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; - dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; - dst_register_ext_concode.CondSrcIndex = 0; - dst_register_ext_concode.Padding = 0; - dst_register_ext_concode.Extended = 0; - - return dst_register_ext_concode; -} - -unsigned -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - unsigned cc, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - int index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - assert( cc <= TGSI_CC_FL ); - assert( swizzle_x <= TGSI_SWIZZLE_W ); - assert( swizzle_y <= TGSI_SWIZZLE_W ); - assert( swizzle_z <= TGSI_SWIZZLE_W ); - assert( swizzle_w <= TGSI_SWIZZLE_W ); - assert( index >= -32768 && index <= 32767 ); - - dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); - dst_register_ext_concode.CondMask = cc; - dst_register_ext_concode.CondSwizzleX = swizzle_x; - dst_register_ext_concode.CondSwizzleY = swizzle_y; - dst_register_ext_concode.CondSwizzleZ = swizzle_z; - dst_register_ext_concode.CondSwizzleW = swizzle_w; - dst_register_ext_concode.CondSrcIndex = index; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_concode; -} - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; - dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; - dst_register_ext_modulate.Padding = 0; - dst_register_ext_modulate.Extended = 0; - - return dst_register_ext_modulate; -} - -unsigned -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - unsigned modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - assert( modulate <= TGSI_MODULATE_EIGHTH ); - - dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); - dst_register_ext_modulate.Modulate = modulate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_modulate; -} - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_build.h b/src/mesa/pipe/tgsi/exec/tgsi_build.h deleted file mode 100644 index 116c78abf3..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_build.h +++ /dev/null @@ -1,320 +0,0 @@ -#if !defined TGSI_BUILD_H -#define TGSI_BUILD_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ); - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ); - -struct tgsi_processor -tgsi_default_processor( void ); - -struct tgsi_processor -tgsi_build_processor( - unsigned processor, - struct tgsi_header *header ); - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ); - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned declare, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ); - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ); - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ); - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ); - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ); - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ); - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ); - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - float value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ); - -unsigned -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ); - -struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ); - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ); - -unsigned -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ); - -unsigned -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ); - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - unsigned precision, - unsigned cond_dst_index, - unsigned cond_flow_index, - unsigned cond_mask, - unsigned cond_swizzle_x, - unsigned cond_swizzle_y, - unsigned cond_swizzle_z, - unsigned cond_swizzle_w, - unsigned cond_dst_update, - unsigned cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ); - -unsigned -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ); - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - unsigned label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ); - -unsigned -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ); - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - unsigned texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register -tgsi_default_src_register( void ); - -struct tgsi_src_register -tgsi_build_src_register( - unsigned file, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - unsigned negate, - unsigned indirect, - unsigned dimension, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ); - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ); - -unsigned -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ); - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - unsigned ext_swizzle_x, - unsigned ext_swizzle_y, - unsigned ext_swizzle_z, - unsigned ext_swizzle_w, - unsigned negate_x, - unsigned negate_y, - unsigned negate_z, - unsigned negate_w, - unsigned ext_divide, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ); - -unsigned -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ); - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - unsigned complement, - unsigned bias, - unsigned scale_2x, - unsigned absolute, - unsigned negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dimension -tgsi_default_dimension( void ); - -struct tgsi_dimension -tgsi_build_dimension( - unsigned indirect, - unsigned index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register -tgsi_default_dst_register( void ); - -struct tgsi_dst_register -tgsi_build_dst_register( - unsigned file, - unsigned mask, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ); - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ); - -unsigned -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ); - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - unsigned cc, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - int index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ); - -unsigned -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ); - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - unsigned modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_BUILD_H - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_core.h b/src/mesa/pipe/tgsi/exec/tgsi_core.h deleted file mode 100644 index 30ad801b6e..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_core.h +++ /dev/null @@ -1,13 +0,0 @@ -#if !defined TGSI_CORE_H -#define TGSI_CORE_H - -#include "tgsi_token.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" -#include "tgsi_exec.h" -#include "tgsi_dump.h" -#include "tgsi_util.h" -#include "tgsi_sse2.h" - -#endif // !defined TGSI_CORE_H - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.c b/src/mesa/pipe/tgsi/exec/tgsi_dump.c deleted file mode 100644 index 1df71efff7..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.c +++ /dev/null @@ -1,1456 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "tgsi_platform.h" -#include "tgsi_core.h" - -struct text_dump -{ - FILE *file; - unsigned tabs; -}; - -static void -text_dump_str( - struct text_dump *dump, - const char *str ) -{ - unsigned i; - size_t len = strlen( str ); - - for( i = 0; i < len; i++ ) { - fprintf( dump->file, "%c", str[i] ); - - if( str[i] == '\n' ) { - unsigned i; - - for( i = 0; i < dump->tabs; i++ ) { - fprintf( dump->file, " " ); - } - } - } -} - -static void -text_dump_chr( - struct text_dump *dump, - const char chr ) -{ - char str[2]; - - str[0] = chr; - str[1] = '\0'; - text_dump_str( dump, str ); -} - -static void -text_dump_uix( - struct text_dump *dump, - const unsigned ui ) -{ - char str[36]; - - sprintf( str, "0x%x", ui ); - text_dump_str( dump, str ); -} - -static void -text_dump_uid( - struct text_dump *dump, - const unsigned ui ) -{ - char str[16]; - - sprintf( str, "%u", ui ); - text_dump_str( dump, str ); -} - -static void -text_dump_sid( - struct text_dump *dump, - const int si ) -{ - char str[16]; - - sprintf( str, "%d", si ); - text_dump_str( dump, str ); -} - -static void -text_dump_flt( - struct text_dump *dump, - const float f ) -{ - char str[48]; - - sprintf( str, "%10.4f", f ); - text_dump_str( dump, str ); -} - -static void -text_dump_enum( - struct text_dump *dump, - const unsigned e, - const char **enums, - const unsigned enums_count ) -{ - if( e >= enums_count ) { - text_dump_uid( dump, e ); - } - else { - text_dump_str( dump, enums[e] ); - } -} - -static void -text_dump_tab( - struct text_dump *dump ) -{ - dump->tabs++; -} - -static void -text_dump_untab( - struct text_dump *dump ) -{ - assert( dump->tabs > 0 ); - - --dump->tabs; -} - -#define TXT(S) text_dump_str( dump, S ) -#define CHR(C) text_dump_chr( dump, C ) -#define UIX(I) text_dump_uix( dump, I ) -#define UID(I) text_dump_uid( dump, I ) -#define SID(I) text_dump_sid( dump, I ) -#define FLT(F) text_dump_flt( dump, F ) -#define TAB() text_dump_tab( dump ) -#define UNT() text_dump_untab( dump ) -#define ENM(E,ENUMS) text_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) - -static const char *TGSI_PROCESSOR_TYPES[] = -{ - "PROCESSOR_FRAGMENT", - "PROCESSOR_VERTEX", - "PROCESSOR_GEOMETRY" -}; - -static const char *TGSI_PROCESSOR_TYPES_SHORT[] = -{ - "FRAG", - "VERT", - "GEOM" -}; - -static const char *TGSI_TOKEN_TYPES[] = -{ - "TOKEN_TYPE_DECLARATION", - "TOKEN_TYPE_IMMEDIATE", - "TOKEN_TYPE_INSTRUCTION" -}; - -static const char *TGSI_FILES[] = -{ - "FILE_NULL", - "FILE_CONSTANT", - "FILE_INPUT", - "FILE_OUTPUT", - "FILE_TEMPORARY", - "FILE_SAMPLER", - "FILE_ADDRESS", - "FILE_IMMEDIATE" -}; - -static const char *TGSI_FILES_SHORT[] = -{ - "NULL", - "CONST", - "IN", - "OUT", - "TEMP", - "SAMP", - "ADDR", - "IMM" -}; - -static const char *TGSI_DECLARES[] = -{ - "DECLARE_RANGE", - "DECLARE_MASK" -}; - -static const char *TGSI_INTERPOLATES[] = -{ - "INTERPOLATE_CONSTANT", - "INTERPOLATE_LINEAR", - "INTERPOLATE_PERSPECTIVE", - "INTERPOLATE_ATTRIB" -}; - -static const char *TGSI_INTERPOLATES_SHORT[] = -{ - "CONSTANT", - "LINEAR", - "PERSPECTIVE", - "ATTRIB" -}; - -static const char *TGSI_SEMANTICS[] = -{ - "SEMANTIC_POSITION", - "SEMANTIC_COLOR", - "SEMANTIC_BCOLOR", - "SEMANTIC_FOG", - "SEMANTIC_PSIZE", - "SEMANTIC_GENERIC," -}; - -static const char *TGSI_SEMANTICS_SHORT[] = -{ - "POSITION", - "COLOR", - "BCOLOR", - "FOG", - "PSIZE", - "GENERIC", -}; - -static const char *TGSI_IMMS[] = -{ - "IMM_FLOAT32" -}; - -static const char *TGSI_IMMS_SHORT[] = -{ - "FLT32" -}; - -static const char *TGSI_OPCODES[] = -{ - "OPCODE_ARL", - "OPCODE_MOV", - "OPCODE_LIT", - "OPCODE_RCP", - "OPCODE_RSQ", - "OPCODE_EXP", - "OPCODE_LOG", - "OPCODE_MUL", - "OPCODE_ADD", - "OPCODE_DP3", - "OPCODE_DP4", - "OPCODE_DST", - "OPCODE_MIN", - "OPCODE_MAX", - "OPCODE_SLT", - "OPCODE_SGE", - "OPCODE_MAD", - "OPCODE_SUB", - "OPCODE_LERP", - "OPCODE_CND", - "OPCODE_CND0", - "OPCODE_DOT2ADD", - "OPCODE_INDEX", - "OPCODE_NEGATE", - "OPCODE_FRAC", - "OPCODE_CLAMP", - "OPCODE_FLOOR", - "OPCODE_ROUND", - "OPCODE_EXPBASE2", - "OPCODE_LOGBASE2", - "OPCODE_POWER", - "OPCODE_CROSSPRODUCT", - "OPCODE_MULTIPLYMATRIX", - "OPCODE_ABS", - "OPCODE_RCC", - "OPCODE_DPH", - "OPCODE_COS", - "OPCODE_DDX", - "OPCODE_DDY", - "OPCODE_KILP", - "OPCODE_PK2H", - "OPCODE_PK2US", - "OPCODE_PK4B", - "OPCODE_PK4UB", - "OPCODE_RFL", - "OPCODE_SEQ", - "OPCODE_SFL", - "OPCODE_SGT", - "OPCODE_SIN", - "OPCODE_SLE", - "OPCODE_SNE", - "OPCODE_STR", - "OPCODE_TEX", - "OPCODE_TXD", - "OPCODE_UP2H", - "OPCODE_UP2US", - "OPCODE_UP4B", - "OPCODE_UP4UB", - "OPCODE_X2D", - "OPCODE_ARA", - "OPCODE_ARR", - "OPCODE_BRA", - "OPCODE_CAL", - "OPCODE_RET", - "OPCODE_SSG", - "OPCODE_CMP", - "OPCODE_SCS", - "OPCODE_TXB", - "OPCODE_NRM", - "OPCODE_DIV", - "OPCODE_DP2", - "OPCODE_TXL", - "OPCODE_BRK", - "OPCODE_IF", - "OPCODE_LOOP", - "OPCODE_REP", - "OPCODE_ELSE", - "OPCODE_ENDIF", - "OPCODE_ENDLOOP", - "OPCODE_ENDREP", - "OPCODE_PUSHA", - "OPCODE_POPA", - "OPCODE_CEIL", - "OPCODE_I2F", - "OPCODE_NOT", - "OPCODE_TRUNC", - "OPCODE_SHL", - "OPCODE_SHR", - "OPCODE_AND", - "OPCODE_OR", - "OPCODE_MOD", - "OPCODE_XOR", - "OPCODE_SAD", - "OPCODE_TXF", - "OPCODE_TXQ", - "OPCODE_CONT", - "OPCODE_EMIT", - "OPCODE_ENDPRIM", - "OPCODE_BGNLOOP2", - "OPCODE_BGNSUB", - "OPCODE_ENDLOOP2", - "OPCODE_ENDSUB", - "OPCODE_NOISE1", - "OPCODE_NOISE2", - "OPCODE_NOISE3", - "OPCODE_NOISE4", - "OPCODE_NOP", - "OPCODE_TEXBEM", - "OPCODE_TEXBEML", - "OPCODE_TEXREG2AR", - "OPCODE_TEXM3X2PAD", - "OPCODE_TEXM3X2TEX", - "OPCODE_TEXM3X3PAD", - "OPCODE_TEXM3X3TEX", - "OPCODE_TEXM3X3SPEC", - "OPCODE_TEXM3X3VSPEC", - "OPCODE_TEXREG2GB", - "OPCODE_TEXREG2RGB", - "OPCODE_TEXDP3TEX", - "OPCODE_TEXDP3", - "OPCODE_TEXM3X3", - "OPCODE_TEXM3X2DEPTH", - "OPCODE_TEXDEPTH", - "OPCODE_BEM", - "OPCODE_M4X3", - "OPCODE_M3X4", - "OPCODE_M3X3", - "OPCODE_M3X2", - "OPCODE_NRM4", - "OPCODE_CALLNZ", - "OPCODE_IFC", - "OPCODE_BREAKC", - "OPCODE_TXP", - "OPCODE_KIL", - "OPCODE_END" -}; - -static const char *TGSI_OPCODES_SHORT[] = -{ - "ARL", - "MOV", - "LIT", - "RCP", - "RSQ", - "EXP", - "LOG", - "MUL", - "ADD", - "DP3", - "DP4", - "DST", - "MIN", - "MAX", - "SLT", - "SGE", - "MAD", - "SUB", - "LERP", - "CND", - "CND0", - "DOT2ADD", - "INDEX", - "NEGATE", - "FRAC", - "CLAMP", - "FLOOR", - "ROUND", - "EXPBASE2", - "LOGBASE2", - "POWER", - "CROSSPRODUCT", - "MULTIPLYMATRIX", - "ABS", - "RCC", - "DPH", - "COS", - "DDX", - "DDY", - "KILP", - "PK2H", - "PK2US", - "PK4B", - "PK4UB", - "RFL", - "SEQ", - "SFL", - "SGT", - "SIN", - "SLE", - "SNE", - "STR", - "TEX", - "TXD", - "UP2H", - "UP2US", - "UP4B", - "UP4UB", - "X2D", - "ARA", - "ARR", - "BRA", - "CAL", - "RET", - "SSG", - "CMP", - "SCS", - "TXB", - "NRM", - "DIV", - "DP2", - "TXL", - "BRK", - "IF", - "LOOP", - "REP", - "ELSE", - "ENDIF", - "ENDLOOP", - "ENDREP", - "PUSHA", - "POPA", - "CEIL", - "I2F", - "NOT", - "TRUNC", - "SHL", - "SHR", - "AND", - "OR", - "MOD", - "XOR", - "SAD", - "TXF", - "TXQ", - "CONT", - "EMIT", - "ENDPRIM", - "BGNLOOP2", - "BGNSUB", - "ENDLOOP2", - "ENDSUB", - "NOISE1", - "NOISE2", - "NOISE3", - "NOISE4", - "NOP", - "TEXBEM", - "TEXBEML", - "TEXREG2AR", - "TEXM3X2PAD", - "TEXM3X2TEX", - "TEXM3X3PAD", - "TEXM3X3TEX", - "TEXM3X3SPEC", - "TEXM3X3VSPEC", - "TEXREG2GB", - "TEXREG2RGB", - "TEXDP3TEX", - "TEXDP3", - "TEXM3X3", - "TEXM3X2DEPTH", - "TEXDEPTH", - "BEM", - "M4X3", - "M3X4", - "M3X3", - "M3X2", - "NRM4", - "CALLNZ", - "IFC", - "BREAKC", - "TXP", - "KIL", - "END" -}; - -static const char *TGSI_SATS[] = -{ - "SAT_NONE", - "SAT_ZERO_ONE", - "SAT_MINUS_PLUS_ONE" -}; - -static const char *TGSI_INSTRUCTION_EXTS[] = -{ - "INSTRUCTION_EXT_TYPE_NV", - "INSTRUCTION_EXT_TYPE_LABEL", - "INSTRUCTION_EXT_TYPE_TEXTURE" -}; - -static const char *TGSI_PRECISIONS[] = -{ - "PRECISION_DEFAULT", - "TGSI_PRECISION_FLOAT32", - "TGSI_PRECISION_FLOAT16", - "TGSI_PRECISION_FIXED12" -}; - -static const char *TGSI_CCS[] = -{ - "CC_GT", - "CC_EQ", - "CC_LT", - "CC_UN", - "CC_GE", - "CC_LE", - "CC_NE", - "CC_TR", - "CC_FL" -}; - -static const char *TGSI_SWIZZLES[] = -{ - "SWIZZLE_X", - "SWIZZLE_Y", - "SWIZZLE_Z", - "SWIZZLE_W" -}; - -static const char *TGSI_SWIZZLES_SHORT[] = -{ - "x", - "y", - "z", - "w" -}; - -static const char *TGSI_TEXTURES[] = -{ - "TEXTURE_UNKNOWN", - "TEXTURE_1D", - "TEXTURE_2D", - "TEXTURE_3D", - "TEXTURE_CUBE", - "TEXTURE_RECT", - "TEXTURE_SHADOW1D", - "TEXTURE_SHADOW2D", - "TEXTURE_SHADOWRECT" -}; - -static const char *TGSI_SRC_REGISTER_EXTS[] = -{ - "SRC_REGISTER_EXT_TYPE_SWZ", - "SRC_REGISTER_EXT_TYPE_MOD" -}; - -static const char *TGSI_EXTSWIZZLES[] = -{ - "EXTSWIZZLE_X", - "EXTSWIZZLE_Y", - "EXTSWIZZLE_Z", - "EXTSWIZZLE_W", - "EXTSWIZZLE_ZERO", - "EXTSWIZZLE_ONE" -}; - -static const char *TGSI_EXTSWIZZLES_SHORT[] = -{ - "x", - "y", - "z", - "w", - "0", - "1" -}; - -static const char *TGSI_WRITEMASKS[] = -{ - "0", - "WRITEMASK_X", - "WRITEMASK_Y", - "WRITEMASK_XY", - "WRITEMASK_Z", - "WRITEMASK_XZ", - "WRITEMASK_YZ", - "WRITEMASK_XYZ", - "WRITEMASK_W", - "WRITEMASK_XW", - "WRITEMASK_YW", - "WRITEMASK_XYW", - "WRITEMASK_ZW", - "WRITEMASK_XZW", - "WRITEMASK_YZW", - "WRITEMASK_XYZW" -}; - -static const char *TGSI_DST_REGISTER_EXTS[] = -{ - "DST_REGISTER_EXT_TYPE_CONDCODE", - "DST_REGISTER_EXT_TYPE_MODULATE" -}; - -static const char *TGSI_MODULATES[] = -{ - "MODULATE_1X", - "MODULATE_2X", - "MODULATE_4X", - "MODULATE_8X", - "MODULATE_HALF", - "MODULATE_QUARTER", - "MODULATE_EIGHTH" -}; - -static void -dump_declaration_short( - struct text_dump *dump, - struct tgsi_full_declaration *decl ) -{ - TXT( "\nDCL " ); - ENM( decl->Declaration.File, TGSI_FILES_SHORT ); - - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - CHR( '[' ); - UID( decl->u.DeclarationRange.First ); - if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { - TXT( ".." ); - UID( decl->u.DeclarationRange.Last ); - } - CHR( ']' ); - break; - default: - assert( 0 ); - } - - if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { - CHR( '.' ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'x' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'w' ); - } - } - - if( decl->Declaration.Interpolate ) { - TXT( ", " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); - } - - if( decl->Declaration.Semantic ) { - TXT( ", " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); - CHR( '[' ); - UID( decl->Semantic.SemanticIndex ); - CHR( ']' ); - } -} - -static void -dump_declaration_verbose( - struct text_dump *dump, - struct tgsi_full_declaration *decl, - unsigned ignored, - unsigned deflt, - struct tgsi_full_declaration *fd ) -{ - TXT( "\nFile : " ); - ENM( decl->Declaration.File, TGSI_FILES ); - TXT( "\nDeclare : " ); - ENM( decl->Declaration.Declare, TGSI_DECLARES ); - if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { - TXT( "\nUsageMask : " ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'X' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'Y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'Z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'W' ); - } - } - if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { - TXT( "\nInterpolate: " ); - UID( decl->Declaration.Interpolate ); - } - if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { - TXT( "\nSemantic : " ); - UID( decl->Declaration.Semantic ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Declaration.Padding ); - } - - CHR( '\n' ); - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - TXT( "\nFirst: " ); - UID( decl->u.DeclarationRange.First ); - TXT( "\nLast : " ); - UID( decl->u.DeclarationRange.Last ); - break; - - case TGSI_DECLARE_MASK: - TXT( "\nMask: " ); - UIX( decl->u.DeclarationMask.Mask ); - break; - - default: - assert( 0 ); - } - - if( decl->Declaration.Interpolate ) { - CHR( '\n' ); - TXT( "\nInterpolate: " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Interpolation.Padding ); - } - } - - if( decl->Declaration.Semantic ) { - CHR( '\n' ); - TXT( "\nSemanticName : " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); - TXT( "\nSemanticIndex: " ); - UID( decl->Semantic.SemanticIndex ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Semantic.Padding ); - } - } -} - -static void -dump_immediate_short( - struct text_dump *dump, - struct tgsi_full_immediate *imm ) -{ - unsigned i; - - TXT( "\nIMM " ); - ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); - - TXT( " { " ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - - if( i < imm->Immediate.Size - 2 ) { - TXT( ", " ); - } - } - TXT( " }" ); -} - -static void -dump_immediate_verbose( - struct text_dump *dump, - struct tgsi_full_immediate *imm, - unsigned ignored ) -{ - unsigned i; - - TXT( "\nDataType : " ); - ENM( imm->Immediate.DataType, TGSI_IMMS ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( imm->Immediate.Padding ); - } - - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - CHR( '\n' ); - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - TXT( "\nFloat: " ); - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - } -} - -static void -dump_instruction_short( - struct text_dump *dump, - struct tgsi_full_instruction *inst, - unsigned instno ) -{ - unsigned i; - boolean first_reg = TRUE; - - CHR( '\n' ); - UID( instno ); - CHR( ':' ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); - - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - TXT( "_SAT" ); - break; - case TGSI_SAT_MINUS_PLUS_ONE: - TXT( "_SAT[-1,1]" ); - break; - default: - assert( 0 ); - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if( !first_reg ) { - CHR( ',' ); - } - CHR( ' ' ); - - ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); - SID( dst->DstRegister.Index ); - CHR( ']' ); - - if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { - CHR( '.' ); - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { - CHR( 'x' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { - CHR( 'y' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { - CHR( 'z' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { - CHR( 'w' ); - } - } - - first_reg = FALSE; - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - - if( !first_reg ) { - CHR( ',' ); - } - CHR( ' ' ); - - if( src->SrcRegisterExtMod.Negate ) { - CHR( '-' ); - } - if( src->SrcRegisterExtMod.Absolute ) { - CHR( '|' ); - } - if( src->SrcRegister.Negate ) { - CHR( '-' ); - } - - ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); - SID( src->SrcRegister.Index ); - CHR( ']' ); - - if (src->SrcRegister.Extended) { - if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || - src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || - src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || - src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { - CHR( '.' ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); - } - } - else if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || - src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || - src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || - src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { - CHR( '.' ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); - } - - if( src->SrcRegisterExtMod.Absolute ) { - CHR( '|' ); - } - - first_reg = FALSE; - } - - switch( inst->Instruction.Opcode ) { - case TGSI_OPCODE_IF: - case TGSI_OPCODE_ELSE: - case TGSI_OPCODE_BGNLOOP2: - case TGSI_OPCODE_ENDLOOP2: - case TGSI_OPCODE_CAL: - TXT( " :" ); - UID( inst->InstructionExtLabel.Label ); - break; - } -} - -static void -dump_instruction_verbose( - struct text_dump *dump, - struct tgsi_full_instruction *inst, - unsigned ignored, - unsigned deflt, - struct tgsi_full_instruction *fi ) -{ - unsigned i; - - TXT( "\nOpcode : " ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES ); - if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { - TXT( "\nSaturate : " ); - ENM( inst->Instruction.Saturate, TGSI_SATS ); - } - if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { - TXT( "\nNumDstRegs : " ); - UID( inst->Instruction.NumDstRegs ); - } - if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { - TXT( "\nNumSrcRegs : " ); - UID( inst->Instruction.NumSrcRegs ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->Instruction.Padding ); - } - - if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { - TXT( "\nPrecision : " ); - ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); - } - if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { - TXT( "\nCondDstIndex : " ); - UID( inst->InstructionExtNv.CondDstIndex ); - } - if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { - TXT( "\nCondFlowIndex : " ); - UID( inst->InstructionExtNv.CondFlowIndex ); - } - if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { - TXT( "\nCondMask : " ); - ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { - TXT( "\nCondSwizzleX : " ); - ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { - TXT( "\nCondSwizzleY : " ); - ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ : " ); - ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { - TXT( "\nCondSwizzleW : " ); - ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { - TXT( "\nCondDstUpdate : " ); - UID( inst->InstructionExtNv.CondDstUpdate ); - } - if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { - TXT( "\nCondFlowEnable: " ); - UID( inst->InstructionExtNv.CondFlowEnable ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtNv.Padding ); - if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { - TXT( "\nExtended : " ); - UID( inst->InstructionExtNv.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { - TXT( "\nLabel : " ); - UID( inst->InstructionExtLabel.Label ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtLabel.Padding ); - if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtLabel.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { - TXT( "\nTexture : " ); - ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtTexture.Padding ); - if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtTexture.Extended ); - } - } - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : " ); - ENM( dst->DstRegister.File, TGSI_FILES ); - if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { - TXT( "\nWriteMask: " ); - ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); - } - if( ignored ) { - if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( dst->DstRegister.Indirect ); - } - if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( dst->DstRegister.Dimension ); - } - } - if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { - TXT( "\nIndex : " ); - SID( dst->DstRegister.Index ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegister.Padding ); - if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegister.Extended ); - } - } - - if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { - TXT( "\nCondMask : " ); - ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { - TXT( "\nCondSwizzleX: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { - TXT( "\nCondSwizzleY: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { - TXT( "\nCondSwizzleW: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { - TXT( "\nCondSrcIndex: " ); - UID( dst->DstRegisterExtConcode.CondSrcIndex ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtConcode.Padding ); - if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegisterExtConcode.Extended ); - } - } - } - - if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { - TXT( "\nModulate: " ); - ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtModulate.Padding ); - if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { - TXT( "\nExtended: " ); - UID( dst->DstRegisterExtModulate.Extended ); - } - } - } - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : "); - ENM( src->SrcRegister.File, TGSI_FILES ); - if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { - TXT( "\nSwizzleX : " ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { - TXT( "\nSwizzleY : " ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { - TXT( "\nSwizzleZ : " ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { - TXT( "\nSwizzleW : " ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegister.Negate ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( src->SrcRegister.Indirect ); - } - if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( src->SrcRegister.Dimension ); - } - } - if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { - TXT( "\nIndex : " ); - SID( src->SrcRegister.Index ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegister.Extended ); - } - } - - if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { - TXT( "\nExtSwizzleX: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { - TXT( "\nExtSwizzleY: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { - TXT( "\nExtSwizzleZ: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { - TXT( "\nExtSwizzleW: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { - TXT( "\nNegateX : " ); - UID( src->SrcRegisterExtSwz.NegateX ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { - TXT( "\nNegateY : " ); - UID( src->SrcRegisterExtSwz.NegateY ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { - TXT( "\nNegateZ : " ); - UID( src->SrcRegisterExtSwz.NegateZ ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { - TXT( "\nNegateW : " ); - UID( src->SrcRegisterExtSwz.NegateW ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { - TXT( "\nExtDivide : " ); - ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtSwz.Padding ); - if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtSwz.Extended ); - } - } - } - - if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { - TXT( "\nComplement: " ); - UID( src->SrcRegisterExtMod.Complement ); - } - if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { - TXT( "\nBias : " ); - UID( src->SrcRegisterExtMod.Bias ); - } - if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { - TXT( "\nScale2X : " ); - UID( src->SrcRegisterExtMod.Scale2X ); - } - if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { - TXT( "\nAbsolute : " ); - UID( src->SrcRegisterExtMod.Absolute ); - } - if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegisterExtMod.Negate ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtMod.Padding ); - if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtMod.Extended ); - } - } - } - } -} - -void -tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ) -{ - struct text_dump _dump; - struct text_dump *dump = &_dump; - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned verbose = flags & TGSI_DUMP_VERBOSE; - unsigned ignored = !(flags & TGSI_DUMP_NO_IGNORED); - unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); - unsigned instno = 0; - - { -#if 0 - static unsigned counter = 0; - char buffer[64]; - - sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); - dump->file = fopen( buffer, "wt" ); -#else - dump->file = stderr; -#endif - dump->tabs = 0; - } - - /* sanity check */ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); - - tgsi_parse_init( &parse, tokens ); - - TXT( "tgsi-dump begin -----------------" ); - - CHR( '\n' ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); - CHR( ' ' ); - UID( parse.FullVersion.Version.MajorVersion ); - CHR( '.' ); - UID( parse.FullVersion.Version.MinorVersion ); - - if( verbose ) { - TXT( "\nMajorVersion: " ); - UID( parse.FullVersion.Version.MajorVersion ); - TXT( "\nMinorVersion: " ); - UID( parse.FullVersion.Version.MinorVersion ); - CHR( '\n' ); - - TXT( "\nHeaderSize: " ); - UID( parse.FullHeader.Header.HeaderSize ); - TXT( "\nBodySize : " ); - UID( parse.FullHeader.Header.BodySize ); - TXT( "\nProcessor : " ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); - CHR( '\n' ); - } - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_short( - dump, - &parse.FullToken.FullDeclaration ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_short( - dump, - &parse.FullToken.FullImmediate ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_short( - dump, - &parse.FullToken.FullInstruction, - instno ); - instno++; - break; - - default: - assert( 0 ); - } - - if( verbose ) { - TXT( "\nType : " ); - ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); - if( ignored ) { - TXT( "\nSize : " ); - UID( parse.FullToken.Token.Size ); - if( deflt || parse.FullToken.Token.Extended ) { - TXT( "\nExtended : " ); - UID( parse.FullToken.Token.Extended ); - } - } - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_verbose( - dump, - &parse.FullToken.FullDeclaration, - ignored, - deflt, - &fd ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_verbose( - dump, - &parse.FullToken.FullImmediate, - ignored ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_verbose( - dump, - &parse.FullToken.FullInstruction, - ignored, - deflt, - &fi ); - break; - - default: - assert( 0 ); - } - - CHR( '\n' ); - } - } - - TXT( "\ntgsi-dump end -------------------\n" ); - - tgsi_parse_free( &parse ); -} - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_dump.h b/src/mesa/pipe/tgsi/exec/tgsi_dump.h deleted file mode 100644 index 70860c0885..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_dump.h +++ /dev/null @@ -1,22 +0,0 @@ -#if !defined TGSI_DUMP_H -#define TGSI_DUMP_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -#define TGSI_DUMP_VERBOSE 1 -#define TGSI_DUMP_NO_IGNORED 2 -#define TGSI_DUMP_NO_DEFAULT 4 - -void -tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_DUMP_H - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index ea6c5021b3..dd11dd58b7 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -46,14 +46,17 @@ * * * Authors: - * Michael Krol + * Michal Krol * Brian Paul */ #include "pipe/p_compiler.h" #include "pipe/p_state.h" #include "pipe/p_util.h" -#include "tgsi_core.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/util/tgsi_util.h" +#include "tgsi_exec.h" #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 diff --git a/src/mesa/pipe/tgsi/exec/tgsi_parse.c b/src/mesa/pipe/tgsi/exec/tgsi_parse.c deleted file mode 100644 index aee71decb3..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_parse.c +++ /dev/null @@ -1,316 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "tgsi_platform.h" -#include "tgsi_core.h" - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ) -{ - full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; -} - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ) -{ - if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { - FREE( full_token->FullImmediate.u.Pointer ); - } -} - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ) -{ - ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; - if( ctx->FullVersion.Version.MajorVersion > 1 ) { - return TGSI_PARSE_ERROR; - } - - ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; - if( ctx->FullHeader.Header.HeaderSize >= 2 ) { - ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; - } - else { - ctx->FullHeader.Processor = tgsi_default_processor(); - } - - ctx->Tokens = tokens; - ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; - - tgsi_full_token_init( &ctx->FullToken ); - - return TGSI_PARSE_OK; -} - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ) -{ - tgsi_full_token_free( &ctx->FullToken ); -} - -boolean -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ) -{ - return ctx->Position >= - 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; -} - -static void -next_token( - struct tgsi_parse_context *ctx, - void *token ) -{ - assert( !tgsi_parse_end_of_tokens( ctx ) ); - - *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; -} - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ) -{ - struct tgsi_token token; - unsigned i; - - tgsi_full_token_free( &ctx->FullToken ); - tgsi_full_token_init( &ctx->FullToken ); - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; - - *decl = tgsi_default_full_declaration(); - decl->Declaration = *(struct tgsi_declaration *) &token; - - switch( decl->Declaration.Type ) { - case TGSI_DECLARE_RANGE: - next_token( ctx, &decl->u.DeclarationRange ); - break; - - case TGSI_DECLARE_MASK: - next_token( ctx, &decl->u.DeclarationMask ); - break; - - default: - assert (0); - } - - if( decl->Declaration.Interpolate ) { - next_token( ctx, &decl->Interpolation ); - } - - if( decl->Declaration.Semantic ) { - next_token( ctx, &decl->Semantic ); - } - - break; - } - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; - - *imm = tgsi_default_full_immediate(); - imm->Immediate = *(struct tgsi_immediate *) &token; - - assert( !imm->Immediate.Extended ); - - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - imm->u.Pointer = MALLOC( - sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - next_token( ctx, &imm->u.ImmediateFloat32[i] ); - } - break; - - default: - assert( 0 ); - } - - break; - } - - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; - unsigned extended; - - *inst = tgsi_default_full_instruction(); - inst->Instruction = *(struct tgsi_instruction *) &token; - - extended = inst->Instruction.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_INSTRUCTION_EXT_TYPE_NV: - inst->InstructionExtNv = - *(struct tgsi_instruction_ext_nv *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_LABEL: - inst->InstructionExtLabel = - *(struct tgsi_instruction_ext_label *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: - inst->InstructionExtTexture = - *(struct tgsi_instruction_ext_texture *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); - assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); - - extended = inst->FullDstRegisters[i].DstRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: - inst->FullDstRegisters[i].DstRegisterExtConcode = - *(struct tgsi_dst_register_ext_concode *) &token; - break; - - case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: - inst->FullDstRegisters[i].DstRegisterExtModulate = - *(struct tgsi_dst_register_ext_modulate *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - } - - assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); - - extended = inst->FullSrcRegisters[i].SrcRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: - inst->FullSrcRegisters[i].SrcRegisterExtSwz = - *(struct tgsi_src_register_ext_swz *) &token; - break; - - case TGSI_SRC_REGISTER_EXT_TYPE_MOD: - inst->FullSrcRegisters[i].SrcRegisterExtMod = - *(struct tgsi_src_register_ext_mod *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - - if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); - - /* - * No support for multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); - - if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - } - } - - break; - } - - default: - assert( 0 ); - } -} - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_parse.h b/src/mesa/pipe/tgsi/exec/tgsi_parse.h deleted file mode 100644 index 9372da8d5d..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_parse.h +++ /dev/null @@ -1,121 +0,0 @@ -#if !defined TGSI_PARSE_H -#define TGSI_PARSE_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_full_version -{ - struct tgsi_version Version; -}; - -struct tgsi_full_header -{ - struct tgsi_header Header; - struct tgsi_processor Processor; -}; - -struct tgsi_full_dst_register -{ - struct tgsi_dst_register DstRegister; - struct tgsi_dst_register_ext_concode DstRegisterExtConcode; - struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; -}; - -struct tgsi_full_src_register -{ - struct tgsi_src_register SrcRegister; - struct tgsi_src_register_ext_swz SrcRegisterExtSwz; - struct tgsi_src_register_ext_mod SrcRegisterExtMod; - struct tgsi_src_register SrcRegisterInd; - struct tgsi_dimension SrcRegisterDim; - struct tgsi_src_register SrcRegisterDimInd; -}; - -struct tgsi_full_declaration -{ - struct tgsi_declaration Declaration; - union - { - struct tgsi_declaration_range DeclarationRange; - struct tgsi_declaration_mask DeclarationMask; - } u; - struct tgsi_declaration_interpolation Interpolation; - struct tgsi_declaration_semantic Semantic; -}; - -struct tgsi_full_immediate -{ - struct tgsi_immediate Immediate; - union - { - void *Pointer; - struct tgsi_immediate_float32 *ImmediateFloat32; - } u; -}; - -#define TGSI_FULL_MAX_DST_REGISTERS 2 -#define TGSI_FULL_MAX_SRC_REGISTERS 3 - -struct tgsi_full_instruction -{ - struct tgsi_instruction Instruction; - struct tgsi_instruction_ext_nv InstructionExtNv; - struct tgsi_instruction_ext_label InstructionExtLabel; - struct tgsi_instruction_ext_texture InstructionExtTexture; - struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; - struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; -}; - -union tgsi_full_token -{ - struct tgsi_token Token; - struct tgsi_full_declaration FullDeclaration; - struct tgsi_full_immediate FullImmediate; - struct tgsi_full_instruction FullInstruction; -}; - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ); - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ); - -struct tgsi_parse_context -{ - const struct tgsi_token *Tokens; - unsigned Position; - struct tgsi_full_version FullVersion; - struct tgsi_full_header FullHeader; - union tgsi_full_token FullToken; -}; - -#define TGSI_PARSE_OK 0 -#define TGSI_PARSE_ERROR 1 - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ); - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ); - -boolean -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ); - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_PARSE_H - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 3a5c1da180..e403cfaaf3 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -25,8 +25,13 @@ * **************************************************************************/ -#include "tgsi_platform.h" -#include "tgsi_core.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/util/tgsi_util.h" +#include "tgsi_exec.h" +#include "tgsi_sse2.h" + #include "x86/rtasm/x86sse.h" #if defined(__i386__) || defined(__386__) diff --git a/src/mesa/pipe/tgsi/exec/tgsi_token.h b/src/mesa/pipe/tgsi/exec/tgsi_token.h deleted file mode 100644 index 2b922c7aef..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_token.h +++ /dev/null @@ -1,1494 +0,0 @@ -#if !defined TGSI_TOKEN_H -#define TGSI_TOKEN_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_version -{ - unsigned MajorVersion : 8; - unsigned MinorVersion : 8; - unsigned Padding : 16; -}; - -struct tgsi_header -{ - unsigned HeaderSize : 8; - unsigned BodySize : 24; -}; - -#define TGSI_PROCESSOR_FRAGMENT 0 -#define TGSI_PROCESSOR_VERTEX 1 -#define TGSI_PROCESSOR_GEOMETRY 2 - -struct tgsi_processor -{ - unsigned Processor : 4; /* TGSI_PROCESSOR_ */ - unsigned Padding : 28; -}; - -#define TGSI_TOKEN_TYPE_DECLARATION 0 -#define TGSI_TOKEN_TYPE_IMMEDIATE 1 -#define TGSI_TOKEN_TYPE_INSTRUCTION 2 - -struct tgsi_token -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_ */ - unsigned Size : 8; /* UINT */ - unsigned Padding : 19; - unsigned Extended : 1; /* BOOL */ -}; - -#define TGSI_FILE_NULL 0 -#define TGSI_FILE_CONSTANT 1 -#define TGSI_FILE_INPUT 2 -#define TGSI_FILE_OUTPUT 3 -#define TGSI_FILE_TEMPORARY 4 -#define TGSI_FILE_SAMPLER 5 -#define TGSI_FILE_ADDRESS 6 -#define TGSI_FILE_IMMEDIATE 7 - -#define TGSI_DECLARE_RANGE 0 -#define TGSI_DECLARE_MASK 1 - -#define TGSI_WRITEMASK_NONE 0x00 -#define TGSI_WRITEMASK_X 0x01 -#define TGSI_WRITEMASK_Y 0x02 -#define TGSI_WRITEMASK_XY 0x03 -#define TGSI_WRITEMASK_Z 0x04 -#define TGSI_WRITEMASK_XZ 0x05 -#define TGSI_WRITEMASK_YZ 0x06 -#define TGSI_WRITEMASK_XYZ 0x07 -#define TGSI_WRITEMASK_W 0x08 -#define TGSI_WRITEMASK_XW 0x09 -#define TGSI_WRITEMASK_YW 0x0A -#define TGSI_WRITEMASK_XYW 0x0B -#define TGSI_WRITEMASK_ZW 0x0C -#define TGSI_WRITEMASK_XZW 0x0D -#define TGSI_WRITEMASK_YZW 0x0E -#define TGSI_WRITEMASK_XYZW 0x0F - -struct tgsi_declaration -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ - unsigned Size : 8; /* UINT */ - unsigned File : 4; /* one of TGSI_FILE_x */ - unsigned Declare : 4; /* one of TGSI_DECLARE_x */ - unsigned UsageMask : 4; /* bitmask of TGSI_WRITEMASK_x flags */ - unsigned Interpolate : 1; /* BOOL, any interpolation info? */ - unsigned Semantic : 1; /* BOOL, any semantic info? */ - unsigned Padding : 5; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_declaration_range -{ - unsigned First : 16; /* UINT */ - unsigned Last : 16; /* UINT */ -}; - -struct tgsi_declaration_mask -{ - unsigned Mask : 32; /* UINT */ -}; - -#define TGSI_INTERPOLATE_CONSTANT 0 -#define TGSI_INTERPOLATE_LINEAR 1 -#define TGSI_INTERPOLATE_PERSPECTIVE 2 - -struct tgsi_declaration_interpolation -{ - unsigned Interpolate : 4; /* TGSI_INTERPOLATE_ */ - unsigned Padding : 28; -}; - -#define TGSI_SEMANTIC_POSITION 0 -#define TGSI_SEMANTIC_COLOR 1 -#define TGSI_SEMANTIC_BCOLOR 2 /**< back-face color */ -#define TGSI_SEMANTIC_FOG 3 -#define TGSI_SEMANTIC_PSIZE 4 -#define TGSI_SEMANTIC_GENERIC 5 -#define TGSI_SEMANTIC_COUNT 6 /**< number of semantic values */ - -struct tgsi_declaration_semantic -{ - unsigned SemanticName : 8; /* one of TGSI_SEMANTIC_ */ - unsigned SemanticIndex : 16; /* UINT */ - unsigned Padding : 8; -}; - -#define TGSI_IMM_FLOAT32 0 - -struct tgsi_immediate -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_IMMEDIATE */ - unsigned Size : 8; /* UINT */ - unsigned DataType : 4; /* TGSI_IMM_ */ - unsigned Padding : 15; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_immediate_float32 -{ - float Float; -}; - -/* - * GL_NV_vertex_program - */ -#define TGSI_OPCODE_ARL 0 -#define TGSI_OPCODE_MOV 1 -#define TGSI_OPCODE_LIT 2 -#define TGSI_OPCODE_RCP 3 -#define TGSI_OPCODE_RSQ 4 -#define TGSI_OPCODE_EXP 5 -#define TGSI_OPCODE_LOG 6 -#define TGSI_OPCODE_MUL 7 -#define TGSI_OPCODE_ADD 8 -#define TGSI_OPCODE_DP3 9 -#define TGSI_OPCODE_DP4 10 -#define TGSI_OPCODE_DST 11 -#define TGSI_OPCODE_MIN 12 -#define TGSI_OPCODE_MAX 13 -#define TGSI_OPCODE_SLT 14 -#define TGSI_OPCODE_SGE 15 -#define TGSI_OPCODE_MAD 16 - -/* - * GL_ATI_fragment_shader - */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_MUL */ -#define TGSI_OPCODE_SUB 17 -#define TGSI_OPCODE_DOT3 TGSI_OPCODE_DP3 -#define TGSI_OPCODE_DOT4 TGSI_OPCODE_DP4 -/* TGSI_OPCODE_MAD */ -#define TGSI_OPCODE_LERP 18 -#define TGSI_OPCODE_CND 19 -#define TGSI_OPCODE_CND0 20 -#define TGSI_OPCODE_DOT2ADD 21 - -/* - * GL_EXT_vertex_shader - */ -#define TGSI_OPCODE_INDEX 22 -#define TGSI_OPCODE_NEGATE 23 -/* TGSI_OPCODE_DOT3 */ -/* TGSI_OPCODE_DOT4 */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_MADD TGSI_OPCODE_MAD -#define TGSI_OPCODE_FRAC 24 -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -#define TGSI_OPCODE_SETGE TGSI_OPCODE_SGE -#define TGSI_OPCODE_SETLT TGSI_OPCODE_SLT -#define TGSI_OPCODE_CLAMP 25 -#define TGSI_OPCODE_FLOOR 26 -#define TGSI_OPCODE_ROUND 27 -#define TGSI_OPCODE_EXPBASE2 28 -#define TGSI_OPCODE_LOGBASE2 29 -#define TGSI_OPCODE_POWER 30 -#define TGSI_OPCODE_RECIP TGSI_OPCODE_RCP -#define TGSI_OPCODE_RECIPSQRT TGSI_OPCODE_RSQ -/* TGSI_OPCODE_SUB */ -#define TGSI_OPCODE_CROSSPRODUCT 31 -#define TGSI_OPCODE_MULTIPLYMATRIX 32 -/* TGSI_OPCODE_MOV */ - -/* - * GL_NV_vertex_program1_1 - */ -/* TGSI_OPCODE_ARL */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_LIT */ -#define TGSI_OPCODE_ABS 33 -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_LOG */ -#define TGSI_OPCODE_RCC 34 -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -#define TGSI_OPCODE_DPH 35 -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ - -/* - * GL_NV_fragment_program - */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_COS 36 -#define TGSI_OPCODE_DDX 37 -#define TGSI_OPCODE_DDY 38 -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DST */ -#define TGSI_OPCODE_EX2 TGSI_OPCODE_EXPBASE2 -#define TGSI_OPCODE_FLR TGSI_OPCODE_FLOOR -#define TGSI_OPCODE_FRC TGSI_OPCODE_FRAC -#define TGSI_OPCODE_KILP 39 /* predicated kill */ -#define TGSI_OPCODE_LG2 TGSI_OPCODE_LOGBASE2 -/* TGSI_OPCODE_LIT */ -#define TGSI_OPCODE_LRP TGSI_OPCODE_LERP -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -#define TGSI_OPCODE_PK2H 40 -#define TGSI_OPCODE_PK2US 41 -#define TGSI_OPCODE_PK4B 42 -#define TGSI_OPCODE_PK4UB 43 -#define TGSI_OPCODE_POW TGSI_OPCODE_POWER -/* TGSI_OPCODE_RCP */ -#define TGSI_OPCODE_RFL 44 -/* TGSI_OPCODE_RSQ */ -#define TGSI_OPCODE_SEQ 45 -#define TGSI_OPCODE_SFL 46 -/* TGSI_OPCODE_SGE */ -#define TGSI_OPCODE_SGT 47 -#define TGSI_OPCODE_SIN 48 -#define TGSI_OPCODE_SLE 49 -/* TGSI_OPCODE_SLT */ -#define TGSI_OPCODE_SNE 50 -#define TGSI_OPCODE_STR 51 -/* TGSI_OPCODE_SUB */ -#define TGSI_OPCODE_TEX 52 -#define TGSI_OPCODE_TXD 53 -/* TGSI_OPCODE_TXP */ -#define TGSI_OPCODE_UP2H 54 -#define TGSI_OPCODE_UP2US 55 -#define TGSI_OPCODE_UP4B 56 -#define TGSI_OPCODE_UP4UB 57 -#define TGSI_OPCODE_X2D 58 - -/* - * GL_NV_vertex_program2 - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_ARA 59 -/* TGSI_OPCODE_ARL */ -#define TGSI_OPCODE_ARR 60 -#define TGSI_OPCODE_BRA 61 -#define TGSI_OPCODE_CAL 62 -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCC */ -/* TGSI_OPCODE_RCP */ -#define TGSI_OPCODE_RET 63 -/* TGSI_OPCODE_RSQNV - use TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SNE */ -#define TGSI_OPCODE_SSG 64 -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_SUB */ - -/* - * GL_ARB_vertex_program - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_ARL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -#define TGSI_OPCODE_SWZ TGSI_OPCODE_MOV -#define TGSI_OPCODE_XPD TGSI_OPCODE_CROSSPRODUCT - -/* - * GL_ARB_fragment_program - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_CMP 65 -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -#define TGSI_OPCODE_SCS 66 -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXP */ -#define TGSI_OPCODE_TXB 67 -/* TGSI_OPCODE_KIL */ - -/* - * GL_NV_fragment_program_option - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_DDX */ -/* TGSI_OPCODE_DDY */ -/* TGSI_OPCODE_PK2H */ -/* TGSI_OPCODE_PK2US */ -/* TGSI_OPCODE_PK4B */ -/* TGSI_OPCODE_PK4UB */ -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SCS */ -/* TGSI_OPCODE_UP2H */ -/* TGSI_OPCODE_UP2US */ -/* TGSI_OPCODE_UP4B */ -/* TGSI_OPCODE_UP4UB */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_RFL */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_CMP */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_X2D */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXP */ -/* TGSI_OPCODE_TXB */ -/* TGSI_OPCODE_KIL */ -/* TGSI_OPCODE_TXD */ - -/* - * GL_NV_fragment_program2 - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_DDX */ -/* TGSI_OPCODE_DDY */ -/* TGSI_OPCODE_PK2H */ -/* TGSI_OPCODE_PK2US */ -/* TGSI_OPCODE_PK4B */ -/* TGSI_OPCODE_PK4UB */ -#define TGSI_OPCODE_NRM 68 -#define TGSI_OPCODE_DIV 69 -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SCS */ -/* TGSI_OPCODE_UP2H */ -/* TGSI_OPCODE_UP2US */ -/* TGSI_OPCODE_UP4B */ -/* TGSI_OPCODE_UP4UB */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_RFL */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -#define TGSI_OPCODE_DP2 70 -/* TGSI_OPCODE_CMP */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_X2D */ -#define TGSI_OPCODE_DP2A TGSI_OPCODE_DOT2ADD -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXP */ -/* TGSI_OPCODE_TXB */ -#define TGSI_OPCODE_TXL 71 -/* TGSI_OPCODE_KIL */ -/* TGSI_OPCODE_TXD */ -/* TGSI_OPCODE_CAL */ -/* TGSI_OPCODE_RET */ -#define TGSI_OPCODE_BRK 72 -#define TGSI_OPCODE_IF 73 -#define TGSI_OPCODE_LOOP 74 -#define TGSI_OPCODE_REP 75 -#define TGSI_OPCODE_ELSE 76 -#define TGSI_OPCODE_ENDIF 77 -#define TGSI_OPCODE_ENDLOOP 78 -#define TGSI_OPCODE_ENDREP 79 - -/* - * GL_NV_vertex_program2_option - */ -/* TGSI_OPCODE_ARL */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_SSG */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_RCC */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_ARR */ -/* TGSI_OPCODE_ARA */ -/* TGSI_OPCODE_BRA */ -/* TGSI_OPCODE_CAL */ -/* TGSI_OPCODE_RET */ - -/* - * GL_NV_vertex_program3 - */ -/* TGSI_OPCODE_ARL */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_SSG */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_RCC */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_ARR */ -/* TGSI_OPCODE_ARA */ -/* TGSI_OPCODE_BRA */ -/* TGSI_OPCODE_CAL */ -/* TGSI_OPCODE_RET */ -#define TGSI_OPCODE_PUSHA 80 -#define TGSI_OPCODE_POPA 81 -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXP */ -/* TGSI_OPCODE_TXB */ -/* TGSI_OPCODE_TXL */ - -/* - * GL_NV_gpu_program4 - */ -/* TGSI_OPCODE_ABS */ -#define TGSI_OPCODE_CEIL 82 -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -#define TGSI_OPCODE_I2F 83 -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_MOV */ -#define TGSI_OPCODE_NOT 84 -/* TGSI_OPCODE_NRM */ -/* TGSI_OPCODE_PK2H */ -/* TGSI_OPCODE_PK2US */ -/* TGSI_OPCODE_PK4B */ -/* TGSI_OPCODE_PK4UB */ -/* TGSI_OPCODE_ROUND */ -/* TGSI_OPCODE_SSG */ -#define TGSI_OPCODE_TRUNC 85 -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_RCC */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SCS */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_UP2H */ -/* TGSI_OPCODE_UP2US */ -/* TGSI_OPCODE_UP4B */ -/* TGSI_OPCODE_UP4UB */ -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_DIV */ -#define TGSI_OPCODE_SHL 86 -#define TGSI_OPCODE_SHR 87 -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_AND 88 -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_DPH */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MUL */ -#define TGSI_OPCODE_OR 89 -/* TGSI_OPCODE_RFL */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SFL */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_STR */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_XPD */ -/* TGSI_OPCODE_DP2 */ -#define TGSI_OPCODE_MOD 90 -#define TGSI_OPCODE_XOR 91 -/* TGSI_OPCODE_CMP */ -/* TGSI_OPCODE_DP2A */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_MAD */ -#define TGSI_OPCODE_SAD 92 -/* TGSI_OPCODE_X2D */ -/* TGSI_OPCODE_SWZ */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXB */ -#define TGSI_OPCODE_TXF 93 -/* TGSI_OPCODE_TXL */ -/* TGSI_OPCODE_TXP */ -#define TGSI_OPCODE_TXQ 94 -/* TGSI_OPCODE_TXD */ -/* TGSI_OPCODE_CAL */ -/* TGSI_OPCODE_RET */ -/* TGSI_OPCODE_BRK */ -#define TGSI_OPCODE_CONT 95 -/* TGSI_OPCODE_IF */ -/* TGSI_OPCODE_REP */ -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -/* TGSI_OPCODE_ENDREP */ - -/* - * GL_NV_vertex_program4 - */ -/* Same as GL_NV_gpu_program4 */ - -/* - * GL_NV_fragment_program4 - */ -/* Same as GL_NV_gpu_program4 */ -/* TGSI_OPCODE_KIL */ -/* TGSI_OPCODE_DDX */ -/* TGSI_OPCODE_DDY */ - -/* - * GL_NV_geometry_program4 - */ -/* Same as GL_NV_gpu_program4 */ -#define TGSI_OPCODE_EMIT 96 -#define TGSI_OPCODE_ENDPRIM 97 - -/* - * GLSL - */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_ADD */ -#define TGSI_OPCODE_BGNLOOP2 98 -#define TGSI_OPCODE_BGNSUB 99 -/* TGSI_OPCODE_BRA */ -/* TGSI_OPCODE_BRK */ -/* TGSI_OPCODE_CONT */ -/* TGSI_OPCODE_COS */ -/* TGSI_OPCODE_DDX */ -/* TGSI_OPCODE_DDY */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -#define TGSI_OPCODE_ENDLOOP2 100 -#define TGSI_OPCODE_ENDSUB 101 -/* TGSI_OPCODE_EX2 */ -/* TGSI_OPCODE_EXP */ -/* TGSI_OPCODE_FLR */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_IF */ -#define TGSI_OPCODE_INT TGSI_OPCODE_TRUNC -/* TGSI_OPCODE_KIL */ -/* TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LOG */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_MUL */ -#define TGSI_OPCODE_NOISE1 102 -#define TGSI_OPCODE_NOISE2 103 -#define TGSI_OPCODE_NOISE3 104 -#define TGSI_OPCODE_NOISE4 105 -#define TGSI_OPCODE_NOP 106 -/* TGSI_OPCODE_POW */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SIN */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SNE */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TXB */ -/* TGSI_OPCODE_TXD */ -/* TGSI_OPCODE_TXL */ -/* TGSI_OPCODE_TXP */ -/* TGSI_OPCODE_XPD */ - -/* - * ps_1_1 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_LRP */ -#define TGSI_OPCODE_TEXCOORD TGSI_OPCODE_NOP -#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL -/* TGSI_OPCODE_TEX */ -#define TGSI_OPCODE_TEXBEM 107 -#define TGSI_OPCODE_TEXBEML 108 -#define TGSI_OPCODE_TEXREG2AR 109 -#define TGSI_OPCODE_TEXM3X2PAD 110 -#define TGSI_OPCODE_TEXM3X2TEX 111 -#define TGSI_OPCODE_TEXM3X3PAD 112 -#define TGSI_OPCODE_TEXM3X3TEX 113 -#define TGSI_OPCODE_TEXM3X3SPEC 114 -#define TGSI_OPCODE_TEXM3X3VSPEC 115 -/* TGSI_OPCODE_CND */ - -/* - * ps_1_2 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_TEXCOORD */ -/* TGSI_OPCODE_TEXKILL */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TEXBEM */ -/* TGSI_OPCODE_TEXBEML */ -/* TGSI_OPCODE_TEXREG2AR */ -#define TGSI_OPCODE_TEXREG2GB 116 -/* TGSI_OPCODE_TEXM3X2PAD */ -/* TGSI_OPCODE_TEXM3X2TEX */ -/* TGSI_OPCODE_TEXM3X3PAD */ -/* TGSI_OPCODE_TEXM3X3TEX */ -/* TGSI_OPCODE_TEXM3X3SPEC */ -/* TGSI_OPCODE_TEXM3X3VSPEC */ -/* TGSI_OPCODE_CND */ -#define TGSI_OPCODE_TEXREG2RGB 117 -#define TGSI_OPCODE_TEXDP3TEX 118 -#define TGSI_OPCODE_TEXDP3 119 -#define TGSI_OPCODE_TEXM3X3 120 -/* CMP - use TGSI_OPCODE_CND0 */ - -/* - * ps_1_3 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_TEXCOORD */ -/* TGSI_OPCODE_TEXKILL */ -/* TGSI_OPCODE_TEX */ -/* TGSI_OPCODE_TEXBEM */ -/* TGSI_OPCODE_TEXBEML */ -/* TGSI_OPCODE_TEXREG2AR */ -/* TGSI_OPCODE_TEXREG2GB */ -/* TGSI_OPCODE_TEXM3X2PAD */ -/* TGSI_OPCODE_TEXM3X2TEX */ -/* TGSI_OPCODE_TEXM3X3PAD */ -/* TGSI_OPCODE_TEXM3X3TEX */ -/* TGSI_OPCODE_TEXM3X3SPEC */ -/* TGSI_OPCODE_TEXM3X3VSPEC */ -/* TGSI_OPCODE_CND */ -/* TGSI_OPCODE_TEXREG2RGB */ -/* TGSI_OPCODE_TEXDP3TEX */ -#define TGSI_OPCODE_TEXM3X2DEPTH 121 -/* TGSI_OPCODE_TEXDP3 */ -/* TGSI_OPCODE_TEXM3X3 */ -/* CMP - use TGSI_OPCODE_CND0 */ - -/* - * ps_1_4 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_LRP */ -#define TGSI_OPCODE_TEXCRD TGSI_OPCODE_TEXCOORD -/* TGSI_OPCODE_TEXKILL */ -#define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX -/* TGSI_OPCODE_CND */ -#define TGSI_OPCODE_TEXDEPTH 122 -/* CMP - use TGSI_OPCODE_CND0 */ -#define TGSI_OPCODE_BEM 123 - -/* - * ps_2_0 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_FRC */ -#define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX -#define TGSI_OPCODE_M4X3 124 -#define TGSI_OPCODE_M3X4 125 -#define TGSI_OPCODE_M3X3 126 -#define TGSI_OPCODE_M3X2 127 -/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ -#define TGSI_OPCODE_CRS TGSI_OPCODE_XPD -/* TGSI_OPCODE_ABS */ -#define TGSI_OPCODE_NRM4 128 -#define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS -/* TGSI_OPCODE_TEXKILL */ -/* TGSI_OPCODE_TEXLD */ -#define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB -/* TGSI_OPCODE_TEXLDP */ -/* CMP - use TGSI_OPCODE_CND0 */ -#define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A - -/* - * ps_2_x - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SNE */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_M4X4 */ -/* TGSI_OPCODE_M4X3 */ -/* TGSI_OPCODE_M3X4 */ -/* TGSI_OPCODE_M3X3 */ -/* TGSI_OPCODE_M3X2 */ -#define TGSI_OPCODE_CALL TGSI_OPCODE_CAL -#define TGSI_OPCODE_CALLNZ 129 -/* TGSI_OPCODE_RET */ -/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ -/* TGSI_OPCODE_CRS */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_NRM4 */ -/* TGSI_OPCODE_SINCOS */ -/* TGSI_OPCODE_REP */ -/* TGSI_OPCODE_ENDREP */ -/* TGSI_OPCODE_IF */ -#define TGSI_OPCODE_IFC 130 -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -#define TGSI_OPCODE_BREAK TGSI_OPCODE_BRK -#define TGSI_OPCODE_BREAKC 131 -/* TGSI_OPCODE_TEXKILL */ -/* TGSI_OPCODE_TEXLD */ -/* TGSI_OPCODE_TEXLDB */ -/* CMP - use TGSI_OPCODE_CND0 */ -/* TGSI_OPCODE_DP2ADD */ -#define TGSI_OPCODE_DSX TGSI_OPCODE_DDX -#define TGSI_OPCODE_DSY TGSI_OPCODE_DDY -#define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD -/* TGSI_OPCODE_TEXLDP */ - -/* - * vs_1_1 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_M4X4 */ -/* TGSI_OPCODE_M4X3 */ -/* TGSI_OPCODE_M3X4 */ -/* TGSI_OPCODE_M3X3 */ -/* TGSI_OPCODE_M3X2 */ -#define TGSI_OPCODE_EXPP TGSI_OPCODE_EXP -#define TGSI_OPCODE_LOGP TGSI_OPCODE_LG2 - -/* - * vs_2_0 - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_M4X4 */ -/* TGSI_OPCODE_M4X3 */ -/* TGSI_OPCODE_M3X4 */ -/* TGSI_OPCODE_M3X3 */ -/* TGSI_OPCODE_M3X2 */ -/* TGSI_OPCODE_CALL */ -/* TGSI_OPCODE_CALLNZ */ -/* TGSI_OPCODE_LOOP */ -/* TGSI_OPCODE_RET */ -/* TGSI_OPCODE_ENDLOOP */ -/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ -/* TGSI_OPCODE_CRS */ -#define TGSI_OPCODE_SGN TGSI_OPCODE_SSG -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_NRM4 */ -/* TGSI_OPCODE_SINCOS */ -/* TGSI_OPCODE_REP */ -/* TGSI_OPCODE_ENDREP */ -/* TGSI_OPCODE_IF */ -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -#define TGSI_OPCODE_MOVA TGSI_OPCODE_ARR -/* TGSI_OPCODE_LOGP */ - -/* - * vs_2_x - */ -/* TGSI_OPCODE_NOP */ -/* TGSI_OPCODE_MOV */ -/* TGSI_OPCODE_ADD */ -/* TGSI_OPCODE_SUB */ -/* TGSI_OPCODE_MAD */ -/* TGSI_OPCODE_MUL */ -/* TGSI_OPCODE_RCP */ -/* TGSI_OPCODE_RSQ */ /* XXX: takes ABS */ -/* TGSI_OPCODE_DP3 */ -/* TGSI_OPCODE_DP4 */ -/* TGSI_OPCODE_MIN */ -/* TGSI_OPCODE_MAX */ -/* TGSI_OPCODE_SLT */ -/* TGSI_OPCODE_SGE */ -/* TGSI_OPCODE_SGT */ -/* TGSI_OPCODE_SLE */ -/* TGSI_OPCODE_SEQ */ -/* TGSI_OPCODE_SNE */ -/* EXP - use TGSI_OPCODE_EX2 */ -/* LOG - use TGSI_OPCODE_LG2 */ -/* TGSI_OPCODE_LIT */ -/* TGSI_OPCODE_DST */ -/* TGSI_OPCODE_LRP */ -/* TGSI_OPCODE_FRC */ -/* TGSI_OPCODE_M4X4 */ -/* TGSI_OPCODE_M4X3 */ -/* TGSI_OPCODE_M3X4 */ -/* TGSI_OPCODE_M3X3 */ -/* TGSI_OPCODE_M3X2 */ -/* TGSI_OPCODE_CALL */ -/* TGSI_OPCODE_CALLNZ */ -/* TGSI_OPCODE_LOOP */ -/* TGSI_OPCODE_RET */ -/* TGSI_OPCODE_ENDLOOP */ -/* TGSI_OPCODE_POW */ /* XXX: takes ABS */ -/* TGSI_OPCODE_CRS */ -/* TGSI_OPCODE_SGN */ -/* TGSI_OPCODE_ABS */ -/* TGSI_OPCODE_NRM4 */ -/* TGSI_OPCODE_SINCOS */ -/* TGSI_OPCODE_REP */ -/* TGSI_OPCODE_ENDREP */ -/* TGSI_OPCODE_IF */ -/* TGSI_OPCODE_IFC */ -/* TGSI_OPCODE_ELSE */ -/* TGSI_OPCODE_ENDIF */ -/* TGSI_OPCODE_BREAK */ -/* TGSI_OPCODE_BREAKC */ -/* TGSI_OPCODE_MOVA */ -/* TGSI_OPCODE_LOGP */ - -#define TGSI_OPCODE_KIL 132 /* unpredicated kill */ -#define TGSI_OPCODE_END 133 /* aka HALT */ - -#define TGSI_OPCODE_LAST 134 - -#define TGSI_SAT_NONE 0 /* do not saturate */ -#define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ -#define TGSI_SAT_MINUS_PLUS_ONE 2 /* clamp to [-1,1] */ - -/* - * Opcode is the operation code to execute. A given operation defines the - * semantics how the source registers (if any) are interpreted and what is - * written to the destination registers (if any) as a result of execution. - * - * NumDstRegs and NumSrcRegs is the number of destination and source registers, - * respectively. For a given operation code, those numbers are fixed and are - * present here only for convenience. - * - * If Extended is TRUE, it is now executed. - * - * Saturate controls how are final results in destination registers modified. - */ - -struct tgsi_instruction -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_INSTRUCTION */ - unsigned Size : 8; /* UINT */ - unsigned Opcode : 8; /* TGSI_OPCODE_ */ - unsigned Saturate : 2; /* TGSI_SAT_ */ - unsigned NumDstRegs : 2; /* UINT */ - unsigned NumSrcRegs : 4; /* UINT */ - unsigned Padding : 3; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows. - * - * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow. - * - * Then, tgsi_instruction::NumSrcRegs of tgsi_src_register follow. - * - * tgsi_instruction::Size contains the total number of words that make the - * instruction, including the instruction word. - */ - -#define TGSI_INSTRUCTION_EXT_TYPE_NV 0 -#define TGSI_INSTRUCTION_EXT_TYPE_LABEL 1 -#define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE 2 -#define TGSI_INSTRUCTION_EXT_TYPE_PREDICATE 3 - -struct tgsi_instruction_ext -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_ */ - unsigned Padding : 27; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_NV, it should - * be cast to tgsi_instruction_ext_nv. - * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_LABEL, it - * should be cast to tgsi_instruction_ext_label. - * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it - * should be cast to tgsi_instruction_ext_texture. - * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_PREDICATE, it - * should be cast to tgsi_instruction_ext_predicate. - * - * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext - * follows. - */ - -#define TGSI_PRECISION_DEFAULT 0 -#define TGSI_PRECISION_FLOAT32 1 -#define TGSI_PRECISION_FLOAT16 2 -#define TGSI_PRECISION_FIXED12 3 - -#define TGSI_CC_GT 0 -#define TGSI_CC_EQ 1 -#define TGSI_CC_LT 2 -#define TGSI_CC_UN 3 -#define TGSI_CC_GE 4 -#define TGSI_CC_LE 5 -#define TGSI_CC_NE 6 -#define TGSI_CC_TR 7 -#define TGSI_CC_FL 8 - -#define TGSI_SWIZZLE_X 0 -#define TGSI_SWIZZLE_Y 1 -#define TGSI_SWIZZLE_Z 2 -#define TGSI_SWIZZLE_W 3 - -/* - * Precision controls the precision at which the operation should be executed. - * - * CondDstUpdate enables condition code register writes. When this field is - * TRUE, CondDstIndex specifies the index of the condition code register to - * update. - * - * CondFlowEnable enables conditional execution of the operation. When this - * field is TRUE, CondFlowIndex specifies the index of the condition code - * register to test against CondMask with component swizzle controled by - * CondSwizzleX, CondSwizzleY, CondSwizzleZ and CondSwizzleW. If the test fails, - * the operation is not executed. - */ - -struct tgsi_instruction_ext_nv -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_NV */ - unsigned Precision : 4; /* TGSI_PRECISION_ */ - unsigned CondDstIndex : 4; /* UINT */ - unsigned CondFlowIndex : 4; /* UINT */ - unsigned CondMask : 4; /* TGSI_CC_ */ - unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned CondDstUpdate : 1; /* BOOL */ - unsigned CondFlowEnable : 1; /* BOOL */ - unsigned Padding : 1; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_instruction_ext_label -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ - unsigned Label : 24; /* UINT */ - unsigned Padding : 3; - unsigned Extended : 1; /* BOOL */ -}; - -#define TGSI_TEXTURE_UNKNOWN 0 -#define TGSI_TEXTURE_1D 1 -#define TGSI_TEXTURE_2D 2 -#define TGSI_TEXTURE_3D 3 -#define TGSI_TEXTURE_CUBE 4 -#define TGSI_TEXTURE_RECT 5 -#define TGSI_TEXTURE_SHADOW1D 6 -#define TGSI_TEXTURE_SHADOW2D 7 -#define TGSI_TEXTURE_SHADOWRECT 8 - -struct tgsi_instruction_ext_texture -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */ - unsigned Texture : 8; /* TGSI_TEXTURE_ */ - unsigned Padding : 19; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_instruction_ext_predicate -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ - unsigned PredDstIndex : 4; /* UINT */ - unsigned PredWriteMask : 4; /* TGSI_WRITEMASK_ */ - unsigned Padding : 19; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * File specifies the register array to access. - * - * Index specifies the element number of a register in the register file. - * - * If Indirect is TRUE, Index should be offset by the X component of a source - * register that follows. The register can be now fetched into local storage - * for further processing. - * - * If Negate is TRUE, all components of the fetched register are negated. - * - * The fetched register components are swizzled according to SwizzleX, SwizzleY, - * SwizzleZ and SwizzleW. - * - * If Extended is TRUE, any further modifications to the source register are - * made to this temporary storage. - */ - -struct tgsi_src_register -{ - unsigned File : 4; /* TGSI_FILE_ */ - unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned Negate : 1; /* BOOL */ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - int Index : 16; /* SINT */ - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_src_register::Extended is TRUE, tgsi_src_register_ext follows. - * - * Then, if tgsi_src_register::Indirect is TRUE, another tgsi_src_register - * follows. - * - * Then, if tgsi_src_register::Dimension is TRUE, tgsi_dimension follows. - */ - -#define TGSI_SRC_REGISTER_EXT_TYPE_SWZ 0 -#define TGSI_SRC_REGISTER_EXT_TYPE_MOD 1 - -struct tgsi_src_register_ext -{ - unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_ */ - unsigned Padding : 27; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_SWZ, - * it should be cast to tgsi_src_register_ext_extswz. - * - * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_MOD, - * it should be cast to tgsi_src_register_ext_mod. - * - * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext - * follows. - */ - -#define TGSI_EXTSWIZZLE_X TGSI_SWIZZLE_X -#define TGSI_EXTSWIZZLE_Y TGSI_SWIZZLE_Y -#define TGSI_EXTSWIZZLE_Z TGSI_SWIZZLE_Z -#define TGSI_EXTSWIZZLE_W TGSI_SWIZZLE_W -#define TGSI_EXTSWIZZLE_ZERO 4 -#define TGSI_EXTSWIZZLE_ONE 5 - -/* - * ExtSwizzleX, ExtSwizzleY, ExtSwizzleZ and ExtSwizzleW swizzle the source - * register in an extended manner. - * - * NegateX, NegateY, NegateZ and NegateW negate individual components of the - * source register. - * - * ExtDivide specifies which component is used to divide all components of the - * source register. - */ - -struct tgsi_src_register_ext_swz -{ - unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */ - unsigned ExtSwizzleX : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned ExtSwizzleY : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned ExtSwizzleZ : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned ExtSwizzleW : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned NegateX : 1; /* BOOL */ - unsigned NegateY : 1; /* BOOL */ - unsigned NegateZ : 1; /* BOOL */ - unsigned NegateW : 1; /* BOOL */ - unsigned ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned Padding : 3; - unsigned Extended : 1; /* BOOL */ -}; - -/** - * Extra src register modifiers - * - * If Complement is TRUE, the source register is modified by subtracting it - * from 1.0. - * - * If Bias is TRUE, the source register is modified by subtracting 0.5 from it. - * - * If Scale2X is TRUE, the source register is modified by multiplying it by 2.0. - * - * If Absolute is TRUE, the source register is modified by removing the sign. - * - * If Negate is TRUE, the source register is modified by negating it. - */ - -struct tgsi_src_register_ext_mod -{ - unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */ - unsigned Complement : 1; /* BOOL */ - unsigned Bias : 1; /* BOOL */ - unsigned Scale2X : 1; /* BOOL */ - unsigned Absolute : 1; /* BOOL */ - unsigned Negate : 1; /* BOOL */ - unsigned Padding : 22; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_dimension -{ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - unsigned Padding : 13; - int Index : 16; /* SINT */ - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_dst_register -{ - unsigned File : 4; /* TGSI_FILE_ */ - unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - int Index : 16; /* SINT */ - unsigned Padding : 5; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_dst_register::Extended is TRUE, tgsi_dst_register_ext follows. - * - * Then, if tgsi_dst_register::Indirect is TRUE, tgsi_src_register follows. - */ - -#define TGSI_DST_REGISTER_EXT_TYPE_CONDCODE 0 -#define TGSI_DST_REGISTER_EXT_TYPE_MODULATE 1 -#define TGSI_DST_REGISTER_EXT_TYPE_PREDICATE 2 - -struct tgsi_dst_register_ext -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_ */ - unsigned Padding : 27; - unsigned Extended : 1; /* BOOL */ -}; - -/** - * Extra destination register modifiers - * - * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_CONDCODE, - * it should be cast to tgsi_dst_register_ext_condcode. - * - * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_MODULATE, - * it should be cast to tgsi_dst_register_ext_modulate. - * - * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_PREDICATE, - * it should be cast to tgsi_dst_register_ext_predicate. - * - * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext - * follows. - */ -struct tgsi_dst_register_ext_concode -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ - unsigned CondMask : 4; /* TGSI_CC_ */ - unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSrcIndex : 4; /* UINT */ - unsigned Padding : 11; - unsigned Extended : 1; /* BOOL */ -}; - -#define TGSI_MODULATE_1X 0 -#define TGSI_MODULATE_2X 1 -#define TGSI_MODULATE_4X 2 -#define TGSI_MODULATE_8X 3 -#define TGSI_MODULATE_HALF 4 -#define TGSI_MODULATE_QUARTER 5 -#define TGSI_MODULATE_EIGHTH 6 - -struct tgsi_dst_register_ext_modulate -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */ - unsigned Modulate : 4; /* TGSI_MODULATE_ */ - unsigned Padding : 23; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * Currently, the following constraints apply. - * - * - PredSwizzleXYZW is either set to identity or replicate. - * - PredSrcIndex is 0. - */ - -struct tgsi_dst_register_ext_predicate -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */ - unsigned PredSwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSrcIndex : 4; /* UINT */ - unsigned Negate : 1; /* BOOL */ - unsigned Padding : 14; - unsigned Extended : 1; /* BOOL */ -}; - - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_TOKEN_H - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_util.c b/src/mesa/pipe/tgsi/exec/tgsi_util.c deleted file mode 100644 index d1bc935526..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_util.c +++ /dev/null @@ -1,270 +0,0 @@ -#include "tgsi_platform.h" -#include "tgsi_core.h" - -union pointer_hack -{ - void *pointer; - unsigned long long uint64; -}; - -void * -tgsi_align_128bit( - void *unaligned ) -{ - union pointer_hack ph; - - ph.uint64 = 0; - ph.pointer = unaligned; - ph.uint64 = (ph.uint64 + 15) & ~15; - return ph.pointer; -} - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->SwizzleX; - case 1: - return reg->SwizzleY; - case 2: - return reg->SwizzleZ; - case 3: - return reg->SwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->ExtSwizzleX; - case 1: - return reg->ExtSwizzleY; - case 2: - return reg->ExtSwizzleZ; - case 3: - return reg->ExtSwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned swizzle; - - /* - * First, calculate the extended swizzle for a given channel. This will give - * us either a channel index into the simple swizzle or a constant 1 or 0. - */ - swizzle = tgsi_util_get_src_register_extswizzle( - ®->SrcRegisterExtSwz, - component ); - - assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); - assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); - assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); - assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); - - /* - * Second, calculate the simple swizzle for the unswizzled channel index. - * Leave the constants intact, they are not affected by the simple swizzle. - */ - if( swizzle <= TGSI_SWIZZLE_W ) { - swizzle = tgsi_util_get_src_register_swizzle( - ®->SrcRegister, - component ); - } - - return swizzle; -} - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - unsigned swizzle, - unsigned component ) -{ - switch( component ) { - case 0: - reg->SwizzleX = swizzle; - break; - case 1: - reg->SwizzleY = swizzle; - break; - case 2: - reg->SwizzleZ = swizzle; - break; - case 3: - reg->SwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - unsigned swizzle, - unsigned component ) -{ - switch( component ) { - case 0: - reg->ExtSwizzleX = swizzle; - break; - case 1: - reg->ExtSwizzleY = swizzle; - break; - case 2: - reg->ExtSwizzleZ = swizzle; - break; - case 3: - reg->ExtSwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->NegateX; - case 1: - return reg->NegateY; - case 2: - return reg->NegateZ; - case 3: - return reg->NegateW; - default: - assert( 0 ); - } - return 0; -} - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ) -{ - switch( component ) { - case 0: - reg->NegateX = negate; - break; - case 1: - reg->NegateY = negate; - break; - case 2: - reg->NegateZ = negate; - break; - case 3: - reg->NegateW = negate; - break; - default: - assert( 0 ); - } -} - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned sign_mode; - - if( reg->SrcRegisterExtMod.Absolute ) { - /* Consider only the post-abs negation. */ - - if( reg->SrcRegisterExtMod.Negate ) { - sign_mode = TGSI_UTIL_SIGN_SET; - } - else { - sign_mode = TGSI_UTIL_SIGN_CLEAR; - } - } - else { - /* Accumulate the three negations. */ - - unsigned negate; - - negate = reg->SrcRegister.Negate; - if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { - negate = !negate; - } - if( reg->SrcRegisterExtMod.Negate ) { - negate = !negate; - } - - if( negate ) { - sign_mode = TGSI_UTIL_SIGN_TOGGLE; - } - else { - sign_mode = TGSI_UTIL_SIGN_KEEP; - } - } - - return sign_mode; -} - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - unsigned sign_mode ) -{ - reg->SrcRegisterExtSwz.NegateX = 0; - reg->SrcRegisterExtSwz.NegateY = 0; - reg->SrcRegisterExtSwz.NegateZ = 0; - reg->SrcRegisterExtSwz.NegateW = 0; - - switch (sign_mode) - { - case TGSI_UTIL_SIGN_CLEAR: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_SET: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 1; - break; - - case TGSI_UTIL_SIGN_TOGGLE: - reg->SrcRegister.Negate = 1; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_KEEP: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - default: - assert( 0 ); - } -} - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_util.h b/src/mesa/pipe/tgsi/exec/tgsi_util.h deleted file mode 100644 index ef14446f0e..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_util.h +++ /dev/null @@ -1,70 +0,0 @@ -#if !defined TGSI_UTIL_H -#define TGSI_UTIL_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -void * -tgsi_align_128bit( - void *unaligned ); - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component); - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - unsigned swizzle, - unsigned component ); - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - unsigned swizzle, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ); - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ); - -#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ -#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ -#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ -#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - unsigned sign_mode ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_UTIL_H - diff --git a/src/mesa/pipe/tgsi/tgsi_platform.h b/src/mesa/pipe/tgsi/tgsi_platform.h deleted file mode 100644 index e7a381b201..0000000000 --- a/src/mesa/pipe/tgsi/tgsi_platform.h +++ /dev/null @@ -1,16 +0,0 @@ -#if !defined TGSI_PLATFORM_H -#define TGSI_PLATFORM_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -#include "pipe/p_compiler.h" -#include "pipe/p_util.h" - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_PLATFORM_H - diff --git a/src/mesa/pipe/tgsi/util/tgsi_build.c b/src/mesa/pipe/tgsi/util/tgsi_build.c new file mode 100644 index 0000000000..19b2aad921 --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_build.c @@ -0,0 +1,1370 @@ +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_build.h" +#include "tgsi_parse.h" + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ) +{ + struct tgsi_version version; + + version.MajorVersion = 1; + version.MinorVersion = 1; + version.Padding = 0; + + return version; +} + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ) +{ + struct tgsi_header header; + + header.HeaderSize = 1; + header.BodySize = 0; + + return header; +} + +static void +header_headersize_grow( struct tgsi_header *header ) +{ + assert( header->HeaderSize < 0xFF ); + assert( header->BodySize == 0 ); + + header->HeaderSize++; +} + +static void +header_bodysize_grow( struct tgsi_header *header ) +{ + assert( header->BodySize < 0xFFFFFF ); + + header->BodySize++; +} + +struct tgsi_processor +tgsi_default_processor( void ) +{ + struct tgsi_processor processor; + + processor.Processor = TGSI_PROCESSOR_FRAGMENT; + processor.Padding = 0; + + return processor; +} + +struct tgsi_processor +tgsi_build_processor( + unsigned type, + struct tgsi_header *header ) +{ + struct tgsi_processor processor; + + processor = tgsi_default_processor(); + processor.Processor = type; + + header_headersize_grow( header ); + + return processor; +} + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ) +{ + struct tgsi_declaration declaration; + + declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; + declaration.Size = 1; + declaration.File = TGSI_FILE_NULL; + declaration.Declare = TGSI_DECLARE_RANGE; + declaration.UsageMask = TGSI_WRITEMASK_XYZW; + declaration.Interpolate = 0; + declaration.Semantic = 0; + declaration.Padding = 0; + declaration.Extended = 0; + + return declaration; +} + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned declare, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ) +{ + struct tgsi_declaration declaration; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( declare <= TGSI_DECLARE_MASK ); + + declaration = tgsi_default_declaration(); + declaration.File = file; + declaration.Declare = declare; + declaration.UsageMask = usage_mask; + declaration.Interpolate = interpolate; + declaration.Semantic = semantic; + + header_bodysize_grow( header ); + + return declaration; +} + +static void +declaration_grow( + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + assert( declaration->Size < 0xFF ); + + declaration->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ) +{ + struct tgsi_full_declaration full_declaration; + + full_declaration.Declaration = tgsi_default_declaration(); + full_declaration.Interpolation = tgsi_default_declaration_interpolation(); + full_declaration.Semantic = tgsi_default_declaration_semantic(); + + return full_declaration; +} + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0; + struct tgsi_declaration *declaration; + + if( maxsize <= size ) + return 0; + declaration = (struct tgsi_declaration *) &tokens[size]; + size++; + + *declaration = tgsi_build_declaration( + full_decl->Declaration.File, + full_decl->Declaration.Declare, + full_decl->Declaration.UsageMask, + full_decl->Declaration.Interpolate, + full_decl->Declaration.Semantic, + header ); + + switch( full_decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + { + struct tgsi_declaration_range *dr; + + if( maxsize <= size ) + return 0; + dr = (struct tgsi_declaration_range *) &tokens[size]; + size++; + + *dr = tgsi_build_declaration_range( + full_decl->u.DeclarationRange.First, + full_decl->u.DeclarationRange.Last, + declaration, + header ); + break; + } + + case TGSI_DECLARE_MASK: + { + struct tgsi_declaration_mask *dm; + + if( maxsize <= size ) + return 0; + dm = (struct tgsi_declaration_mask *) &tokens[size]; + size++; + + *dm = tgsi_build_declaration_mask( + full_decl->u.DeclarationMask.Mask, + declaration, + header ); + break; + } + + default: + assert( 0 ); + } + + if( full_decl->Declaration.Interpolate ) { + struct tgsi_declaration_interpolation *di; + + if( maxsize <= size ) + return 0; + di = (struct tgsi_declaration_interpolation *) &tokens[size]; + size++; + + *di = tgsi_build_declaration_interpolation( + full_decl->Interpolation.Interpolate, + declaration, + header ); + } + + if( full_decl->Declaration.Semantic ) { + struct tgsi_declaration_semantic *ds; + + if( maxsize <= size ) + return 0; + ds = (struct tgsi_declaration_semantic *) &tokens[size]; + size++; + + *ds = tgsi_build_declaration_semantic( + full_decl->Semantic.SemanticName, + full_decl->Semantic.SemanticIndex, + declaration, + header ); + } + + return size; +} + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_range declaration_range; + + assert( last >= first ); + assert( last <= 0xFFFF ); + + declaration_range.First = first; + declaration_range.Last = last; + + declaration_grow( declaration, header ); + + return declaration_range; +} + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + unsigned mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_mask declaration_mask; + + declaration_mask.Mask = mask; + + declaration_grow( declaration, header ); + + return declaration_mask; +} + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ) +{ + struct tgsi_declaration_interpolation di; + + di.Interpolate = TGSI_INTERPOLATE_CONSTANT; + di.Padding = 0; + + return di; +} + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + unsigned interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_interpolation di; + + assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); + + di = tgsi_default_declaration_interpolation(); + di.Interpolate = interpolate; + + declaration_grow( declaration, header ); + + return di; +} + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ) +{ + struct tgsi_declaration_semantic ds; + + ds.SemanticName = TGSI_SEMANTIC_POSITION; + ds.SemanticIndex = 0; + ds.Padding = 0; + + return ds; +} + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_semantic ds; + + assert( semantic_name <= TGSI_SEMANTIC_COUNT ); + assert( semantic_index <= 0xFFFF ); + + ds = tgsi_default_declaration_semantic(); + ds.SemanticName = semantic_name; + ds.SemanticIndex = semantic_index; + + declaration_grow( declaration, header ); + + return ds; +} + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ) +{ + struct tgsi_immediate immediate; + + immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + immediate.Size = 1; + immediate.DataType = TGSI_IMM_FLOAT32; + immediate.Padding = 0; + immediate.Extended = 0; + + return immediate; +} + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ) +{ + struct tgsi_immediate immediate; + + immediate = tgsi_default_immediate(); + + header_bodysize_grow( header ); + + return immediate; +} + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ) +{ + struct tgsi_full_immediate fullimm; + + fullimm.Immediate = tgsi_default_immediate(); + fullimm.u.Pointer = (void *) 0; + + return fullimm; +} + +static void +immediate_grow( + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + assert( immediate->Size < 0xFF ); + + immediate->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + float value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + struct tgsi_immediate_float32 immediate_float32; + + immediate_float32.Float = value; + + immediate_grow( immediate, header ); + + return immediate_float32; +} + +unsigned +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0, i; + struct tgsi_immediate *immediate; + + if( maxsize <= size ) + return 0; + immediate = (struct tgsi_immediate *) &tokens[size]; + size++; + + *immediate = tgsi_build_immediate( header ); + + for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { + struct tgsi_immediate_float32 *if32; + + if( maxsize <= size ) + return 0; + if32 = (struct tgsi_immediate_float32 *) &tokens[size]; + size++; + + *if32 = tgsi_build_immediate_float32( + full_imm->u.ImmediateFloat32[i].Float, + immediate, + header ); + } + + return size; +} + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ) +{ + struct tgsi_instruction instruction; + + instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; + instruction.Size = 1; + instruction.Opcode = TGSI_OPCODE_MOV; + instruction.Saturate = TGSI_SAT_NONE; + instruction.NumDstRegs = 1; + instruction.NumSrcRegs = 1; + instruction.Padding = 0; + instruction.Extended = 0; + + return instruction; +} + +struct tgsi_instruction +tgsi_build_instruction( + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header ) +{ + struct tgsi_instruction instruction; + + assert (opcode <= TGSI_OPCODE_LAST); + assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); + assert (num_dst_regs <= 3); + assert (num_src_regs <= 15); + + instruction = tgsi_default_instruction(); + instruction.Opcode = opcode; + instruction.Saturate = saturate; + instruction.NumDstRegs = num_dst_regs; + instruction.NumSrcRegs = num_src_regs; + + header_bodysize_grow( header ); + + return instruction; +} + +static void +instruction_grow( + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + assert (instruction->Size < 0xFF); + + instruction->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ) +{ + struct tgsi_full_instruction full_instruction; + unsigned i; + + full_instruction.Instruction = tgsi_default_instruction(); + full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); + full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); + full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); + for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { + full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); + } + for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { + full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); + } + + return full_instruction; +} + +unsigned +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0; + unsigned i; + struct tgsi_instruction *instruction; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + instruction = (struct tgsi_instruction *) &tokens[size]; + size++; + + *instruction = tgsi_build_instruction( + full_inst->Instruction.Opcode, + full_inst->Instruction.Saturate, + full_inst->Instruction.NumDstRegs, + full_inst->Instruction.NumSrcRegs, + header ); + prev_token = (struct tgsi_token *) instruction; + + if( tgsi_compare_instruction_ext_nv( + full_inst->InstructionExtNv, + tgsi_default_instruction_ext_nv() ) ) { + struct tgsi_instruction_ext_nv *instruction_ext_nv; + + if( maxsize <= size ) + return 0; + instruction_ext_nv = + (struct tgsi_instruction_ext_nv *) &tokens[size]; + size++; + + *instruction_ext_nv = tgsi_build_instruction_ext_nv( + full_inst->InstructionExtNv.Precision, + full_inst->InstructionExtNv.CondDstIndex, + full_inst->InstructionExtNv.CondFlowIndex, + full_inst->InstructionExtNv.CondMask, + full_inst->InstructionExtNv.CondSwizzleX, + full_inst->InstructionExtNv.CondSwizzleY, + full_inst->InstructionExtNv.CondSwizzleZ, + full_inst->InstructionExtNv.CondSwizzleW, + full_inst->InstructionExtNv.CondDstUpdate, + full_inst->InstructionExtNv.CondFlowEnable, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_nv; + } + + if( tgsi_compare_instruction_ext_label( + full_inst->InstructionExtLabel, + tgsi_default_instruction_ext_label() ) ) { + struct tgsi_instruction_ext_label *instruction_ext_label; + + if( maxsize <= size ) + return 0; + instruction_ext_label = + (struct tgsi_instruction_ext_label *) &tokens[size]; + size++; + + *instruction_ext_label = tgsi_build_instruction_ext_label( + full_inst->InstructionExtLabel.Label, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_label; + } + + if( tgsi_compare_instruction_ext_texture( + full_inst->InstructionExtTexture, + tgsi_default_instruction_ext_texture() ) ) { + struct tgsi_instruction_ext_texture *instruction_ext_texture; + + if( maxsize <= size ) + return 0; + instruction_ext_texture = + (struct tgsi_instruction_ext_texture *) &tokens[size]; + size++; + + *instruction_ext_texture = tgsi_build_instruction_ext_texture( + full_inst->InstructionExtTexture.Texture, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_texture; + } + + for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { + const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; + struct tgsi_dst_register *dst_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + dst_register = (struct tgsi_dst_register *) &tokens[size]; + size++; + + *dst_register = tgsi_build_dst_register( + reg->DstRegister.File, + reg->DstRegister.WriteMask, + reg->DstRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register; + + if( tgsi_compare_dst_register_ext_concode( + reg->DstRegisterExtConcode, + tgsi_default_dst_register_ext_concode() ) ) { + struct tgsi_dst_register_ext_concode *dst_register_ext_concode; + + if( maxsize <= size ) + return 0; + dst_register_ext_concode = + (struct tgsi_dst_register_ext_concode *) &tokens[size]; + size++; + + *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( + reg->DstRegisterExtConcode.CondMask, + reg->DstRegisterExtConcode.CondSwizzleX, + reg->DstRegisterExtConcode.CondSwizzleY, + reg->DstRegisterExtConcode.CondSwizzleZ, + reg->DstRegisterExtConcode.CondSwizzleW, + reg->DstRegisterExtConcode.CondSrcIndex, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_concode; + } + + if( tgsi_compare_dst_register_ext_modulate( + reg->DstRegisterExtModulate, + tgsi_default_dst_register_ext_modulate() ) ) { + struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; + + if( maxsize <= size ) + return 0; + dst_register_ext_modulate = + (struct tgsi_dst_register_ext_modulate *) &tokens[size]; + size++; + + *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( + reg->DstRegisterExtModulate.Modulate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_modulate; + } + } + + for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { + const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; + struct tgsi_src_register *src_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + src_register = (struct tgsi_src_register *) &tokens[size]; + size++; + + *src_register = tgsi_build_src_register( + reg->SrcRegister.File, + reg->SrcRegister.SwizzleX, + reg->SrcRegister.SwizzleY, + reg->SrcRegister.SwizzleZ, + reg->SrcRegister.SwizzleW, + reg->SrcRegister.Negate, + reg->SrcRegister.Indirect, + reg->SrcRegister.Dimension, + reg->SrcRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register; + + if( tgsi_compare_src_register_ext_swz( + reg->SrcRegisterExtSwz, + tgsi_default_src_register_ext_swz() ) ) { + struct tgsi_src_register_ext_swz *src_register_ext_swz; + + if( maxsize <= size ) + return 0; + src_register_ext_swz = + (struct tgsi_src_register_ext_swz *) &tokens[size]; + size++; + + *src_register_ext_swz = tgsi_build_src_register_ext_swz( + reg->SrcRegisterExtSwz.ExtSwizzleX, + reg->SrcRegisterExtSwz.ExtSwizzleY, + reg->SrcRegisterExtSwz.ExtSwizzleZ, + reg->SrcRegisterExtSwz.ExtSwizzleW, + reg->SrcRegisterExtSwz.NegateX, + reg->SrcRegisterExtSwz.NegateY, + reg->SrcRegisterExtSwz.NegateZ, + reg->SrcRegisterExtSwz.NegateW, + reg->SrcRegisterExtSwz.ExtDivide, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_swz; + } + + if( tgsi_compare_src_register_ext_mod( + reg->SrcRegisterExtMod, + tgsi_default_src_register_ext_mod() ) ) { + struct tgsi_src_register_ext_mod *src_register_ext_mod; + + if( maxsize <= size ) + return 0; + src_register_ext_mod = + (struct tgsi_src_register_ext_mod *) &tokens[size]; + size++; + + *src_register_ext_mod = tgsi_build_src_register_ext_mod( + reg->SrcRegisterExtMod.Complement, + reg->SrcRegisterExtMod.Bias, + reg->SrcRegisterExtMod.Scale2X, + reg->SrcRegisterExtMod.Absolute, + reg->SrcRegisterExtMod.Negate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_mod; + } + + if( reg->SrcRegister.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterInd.File, + reg->SrcRegisterInd.SwizzleX, + reg->SrcRegisterInd.SwizzleY, + reg->SrcRegisterInd.SwizzleZ, + reg->SrcRegisterInd.SwizzleW, + reg->SrcRegisterInd.Negate, + reg->SrcRegisterInd.Indirect, + reg->SrcRegisterInd.Dimension, + reg->SrcRegisterInd.Index, + instruction, + header ); + } + + if( reg->SrcRegister.Dimension ) { + struct tgsi_dimension *dim; + + assert( !reg->SrcRegisterDim.Dimension ); + + if( maxsize <= size ) + return 0; + dim = (struct tgsi_dimension *) &tokens[size]; + size++; + + *dim = tgsi_build_dimension( + reg->SrcRegisterDim.Indirect, + reg->SrcRegisterDim.Index, + instruction, + header ); + + if( reg->SrcRegisterDim.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterDimInd.File, + reg->SrcRegisterDimInd.SwizzleX, + reg->SrcRegisterDimInd.SwizzleY, + reg->SrcRegisterDimInd.SwizzleZ, + reg->SrcRegisterDimInd.SwizzleW, + reg->SrcRegisterDimInd.Negate, + reg->SrcRegisterDimInd.Indirect, + reg->SrcRegisterDimInd.Dimension, + reg->SrcRegisterDimInd.Index, + instruction, + header ); + } + } + } + + return size; +} + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; + instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; + instruction_ext_nv.CondDstIndex = 0; + instruction_ext_nv.CondFlowIndex = 0; + instruction_ext_nv.CondMask = TGSI_CC_TR; + instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; + instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; + instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; + instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; + instruction_ext_nv.CondDstUpdate = 0; + instruction_ext_nv.CondFlowEnable = 0; + instruction_ext_nv.Padding = 0; + instruction_ext_nv.Extended = 0; + + return instruction_ext_nv; +} + +union token_u32 +{ + unsigned u32; +}; + +unsigned +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv = tgsi_default_instruction_ext_nv(); + instruction_ext_nv.Precision = precision; + instruction_ext_nv.CondDstIndex = cond_dst_index; + instruction_ext_nv.CondFlowIndex = cond_flow_index; + instruction_ext_nv.CondMask = cond_mask; + instruction_ext_nv.CondSwizzleX = cond_swizzle_x; + instruction_ext_nv.CondSwizzleY = cond_swizzle_y; + instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; + instruction_ext_nv.CondSwizzleW = cond_swizzle_w; + instruction_ext_nv.CondDstUpdate = cond_dst_update; + instruction_ext_nv.CondFlowEnable = cond_flow_update; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_nv; +} + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; + instruction_ext_label.Label = 0; + instruction_ext_label.Padding = 0; + instruction_ext_label.Extended = 0; + + return instruction_ext_label; +} + +unsigned +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + unsigned label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label = tgsi_default_instruction_ext_label(); + instruction_ext_label.Label = label; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_label; +} + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; + instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; + instruction_ext_texture.Padding = 0; + instruction_ext_texture.Extended = 0; + + return instruction_ext_texture; +} + +unsigned +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + unsigned texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture = tgsi_default_instruction_ext_texture(); + instruction_ext_texture.Texture = texture; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_texture; +} + +struct tgsi_src_register +tgsi_default_src_register( void ) +{ + struct tgsi_src_register src_register; + + src_register.File = TGSI_FILE_NULL; + src_register.SwizzleX = TGSI_SWIZZLE_X; + src_register.SwizzleY = TGSI_SWIZZLE_Y; + src_register.SwizzleZ = TGSI_SWIZZLE_Z; + src_register.SwizzleW = TGSI_SWIZZLE_W; + src_register.Negate = 0; + src_register.Indirect = 0; + src_register.Dimension = 0; + src_register.Index = 0; + src_register.Extended = 0; + + return src_register; +} + +struct tgsi_src_register +tgsi_build_src_register( + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register src_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( negate <= 1 ); + assert( index >= -0x8000 && index <= 0x7FFF ); + + src_register = tgsi_default_src_register(); + src_register.File = file; + src_register.SwizzleX = swizzle_x; + src_register.SwizzleY = swizzle_y; + src_register.SwizzleZ = swizzle_z; + src_register.SwizzleW = swizzle_w; + src_register.Negate = negate; + src_register.Indirect = indirect; + src_register.Dimension = dimension; + src_register.Index = index; + + instruction_grow( instruction, header ); + + return src_register; +} + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ) +{ + struct tgsi_full_src_register full_src_register; + + full_src_register.SrcRegister = tgsi_default_src_register(); + full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); + full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); + full_src_register.SrcRegisterInd = tgsi_default_src_register(); + full_src_register.SrcRegisterDim = tgsi_default_dimension(); + full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); + + return full_src_register; +} + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; + src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; + src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; + src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; + src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; + src_register_ext_swz.NegateX = 0; + src_register_ext_swz.NegateY = 0; + src_register_ext_swz.NegateZ = 0; + src_register_ext_swz.NegateW = 0; + src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; + src_register_ext_swz.Padding = 0; + src_register_ext_swz.Extended = 0; + + return src_register_ext_swz; +} + +unsigned +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + unsigned ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE ); + assert( negate_x <= 1 ); + assert( negate_y <= 1 ); + assert( negate_z <= 1 ); + assert( negate_w <= 1 ); + assert( ext_divide <= TGSI_EXTSWIZZLE_ONE ); + + src_register_ext_swz = tgsi_default_src_register_ext_swz(); + src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; + src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; + src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; + src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; + src_register_ext_swz.NegateX = negate_x; + src_register_ext_swz.NegateY = negate_y; + src_register_ext_swz.NegateZ = negate_z; + src_register_ext_swz.NegateW = negate_w; + src_register_ext_swz.ExtDivide = ext_divide; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_swz; +} + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; + src_register_ext_mod.Complement = 0; + src_register_ext_mod.Bias = 0; + src_register_ext_mod.Scale2X = 0; + src_register_ext_mod.Absolute = 0; + src_register_ext_mod.Negate = 0; + src_register_ext_mod.Padding = 0; + src_register_ext_mod.Extended = 0; + + return src_register_ext_mod; +} + +unsigned +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + assert( complement <= 1 ); + assert( bias <= 1 ); + assert( scale_2x <= 1 ); + assert( absolute <= 1 ); + assert( negate <= 1 ); + + src_register_ext_mod = tgsi_default_src_register_ext_mod(); + src_register_ext_mod.Complement = complement; + src_register_ext_mod.Bias = bias; + src_register_ext_mod.Scale2X = scale_2x; + src_register_ext_mod.Absolute = absolute; + src_register_ext_mod.Negate = negate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_mod; +} + +struct tgsi_dimension +tgsi_default_dimension( void ) +{ + struct tgsi_dimension dimension; + + dimension.Indirect = 0; + dimension.Dimension = 0; + dimension.Padding = 0; + dimension.Index = 0; + dimension.Extended = 0; + + return dimension; +} + +struct tgsi_dimension +tgsi_build_dimension( + unsigned indirect, + unsigned index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dimension dimension; + + dimension = tgsi_default_dimension(); + dimension.Indirect = indirect; + dimension.Index = index; + + instruction_grow( instruction, header ); + + return dimension; +} + +struct tgsi_dst_register +tgsi_default_dst_register( void ) +{ + struct tgsi_dst_register dst_register; + + dst_register.File = TGSI_FILE_NULL; + dst_register.WriteMask = TGSI_WRITEMASK_XYZW; + dst_register.Indirect = 0; + dst_register.Dimension = 0; + dst_register.Index = 0; + dst_register.Padding = 0; + dst_register.Extended = 0; + + return dst_register; +} + +struct tgsi_dst_register +tgsi_build_dst_register( + unsigned file, + unsigned mask, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register dst_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( mask <= TGSI_WRITEMASK_XYZW ); + assert( index >= -32768 && index <= 32767 ); + + dst_register = tgsi_default_dst_register(); + dst_register.File = file; + dst_register.WriteMask = mask; + dst_register.Index = index; + + instruction_grow( instruction, header ); + + return dst_register; +} + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ) +{ + struct tgsi_full_dst_register full_dst_register; + + full_dst_register.DstRegister = tgsi_default_dst_register(); + full_dst_register.DstRegisterExtConcode = + tgsi_default_dst_register_ext_concode(); + full_dst_register.DstRegisterExtModulate = + tgsi_default_dst_register_ext_modulate(); + + return full_dst_register; +} + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; + dst_register_ext_concode.CondMask = TGSI_CC_TR; + dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; + dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; + dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; + dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; + dst_register_ext_concode.CondSrcIndex = 0; + dst_register_ext_concode.Padding = 0; + dst_register_ext_concode.Extended = 0; + + return dst_register_ext_concode; +} + +unsigned +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + assert( cc <= TGSI_CC_FL ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( index >= -32768 && index <= 32767 ); + + dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); + dst_register_ext_concode.CondMask = cc; + dst_register_ext_concode.CondSwizzleX = swizzle_x; + dst_register_ext_concode.CondSwizzleY = swizzle_y; + dst_register_ext_concode.CondSwizzleZ = swizzle_z; + dst_register_ext_concode.CondSwizzleW = swizzle_w; + dst_register_ext_concode.CondSrcIndex = index; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_concode; +} + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; + dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; + dst_register_ext_modulate.Padding = 0; + dst_register_ext_modulate.Extended = 0; + + return dst_register_ext_modulate; +} + +unsigned +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + unsigned modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + assert( modulate <= TGSI_MODULATE_EIGHTH ); + + dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); + dst_register_ext_modulate.Modulate = modulate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_modulate; +} + diff --git a/src/mesa/pipe/tgsi/util/tgsi_build.h b/src/mesa/pipe/tgsi/util/tgsi_build.h new file mode 100644 index 0000000000..116c78abf3 --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_build.h @@ -0,0 +1,320 @@ +#if !defined TGSI_BUILD_H +#define TGSI_BUILD_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ); + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ); + +struct tgsi_processor +tgsi_default_processor( void ); + +struct tgsi_processor +tgsi_build_processor( + unsigned processor, + struct tgsi_header *header ); + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ); + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned declare, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ); + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ); + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + unsigned mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ); + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + unsigned interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ); + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ); + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ); + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ); + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + float value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ); + +unsigned +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ); + +struct tgsi_instruction +tgsi_build_instruction( + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header ); + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ); + +unsigned +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ); + +unsigned +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ); + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ); + +unsigned +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ); + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + unsigned label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ); + +unsigned +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ); + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + unsigned texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register +tgsi_default_src_register( void ); + +struct tgsi_src_register +tgsi_build_src_register( + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ); + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ); + +unsigned +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ); + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + unsigned ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ); + +unsigned +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ); + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dimension +tgsi_default_dimension( void ); + +struct tgsi_dimension +tgsi_build_dimension( + unsigned indirect, + unsigned index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register +tgsi_default_dst_register( void ); + +struct tgsi_dst_register +tgsi_build_dst_register( + unsigned file, + unsigned mask, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ); + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ); + +unsigned +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ); + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ); + +unsigned +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ); + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + unsigned modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_BUILD_H + diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.c b/src/mesa/pipe/tgsi/util/tgsi_dump.c new file mode 100644 index 0000000000..982d5ce796 --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_dump.c @@ -0,0 +1,1459 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_dump.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" + +struct text_dump +{ + FILE *file; + unsigned tabs; +}; + +static void +text_dump_str( + struct text_dump *dump, + const char *str ) +{ + unsigned i; + size_t len = strlen( str ); + + for( i = 0; i < len; i++ ) { + fprintf( dump->file, "%c", str[i] ); + + if( str[i] == '\n' ) { + unsigned i; + + for( i = 0; i < dump->tabs; i++ ) { + fprintf( dump->file, " " ); + } + } + } +} + +static void +text_dump_chr( + struct text_dump *dump, + const char chr ) +{ + char str[2]; + + str[0] = chr; + str[1] = '\0'; + text_dump_str( dump, str ); +} + +static void +text_dump_uix( + struct text_dump *dump, + const unsigned ui ) +{ + char str[36]; + + sprintf( str, "0x%x", ui ); + text_dump_str( dump, str ); +} + +static void +text_dump_uid( + struct text_dump *dump, + const unsigned ui ) +{ + char str[16]; + + sprintf( str, "%u", ui ); + text_dump_str( dump, str ); +} + +static void +text_dump_sid( + struct text_dump *dump, + const int si ) +{ + char str[16]; + + sprintf( str, "%d", si ); + text_dump_str( dump, str ); +} + +static void +text_dump_flt( + struct text_dump *dump, + const float f ) +{ + char str[48]; + + sprintf( str, "%10.4f", f ); + text_dump_str( dump, str ); +} + +static void +text_dump_enum( + struct text_dump *dump, + const unsigned e, + const char **enums, + const unsigned enums_count ) +{ + if( e >= enums_count ) { + text_dump_uid( dump, e ); + } + else { + text_dump_str( dump, enums[e] ); + } +} + +static void +text_dump_tab( + struct text_dump *dump ) +{ + dump->tabs++; +} + +static void +text_dump_untab( + struct text_dump *dump ) +{ + assert( dump->tabs > 0 ); + + --dump->tabs; +} + +#define TXT(S) text_dump_str( dump, S ) +#define CHR(C) text_dump_chr( dump, C ) +#define UIX(I) text_dump_uix( dump, I ) +#define UID(I) text_dump_uid( dump, I ) +#define SID(I) text_dump_sid( dump, I ) +#define FLT(F) text_dump_flt( dump, F ) +#define TAB() text_dump_tab( dump ) +#define UNT() text_dump_untab( dump ) +#define ENM(E,ENUMS) text_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) + +static const char *TGSI_PROCESSOR_TYPES[] = +{ + "PROCESSOR_FRAGMENT", + "PROCESSOR_VERTEX", + "PROCESSOR_GEOMETRY" +}; + +static const char *TGSI_PROCESSOR_TYPES_SHORT[] = +{ + "FRAG", + "VERT", + "GEOM" +}; + +static const char *TGSI_TOKEN_TYPES[] = +{ + "TOKEN_TYPE_DECLARATION", + "TOKEN_TYPE_IMMEDIATE", + "TOKEN_TYPE_INSTRUCTION" +}; + +static const char *TGSI_FILES[] = +{ + "FILE_NULL", + "FILE_CONSTANT", + "FILE_INPUT", + "FILE_OUTPUT", + "FILE_TEMPORARY", + "FILE_SAMPLER", + "FILE_ADDRESS", + "FILE_IMMEDIATE" +}; + +static const char *TGSI_FILES_SHORT[] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + +static const char *TGSI_DECLARES[] = +{ + "DECLARE_RANGE", + "DECLARE_MASK" +}; + +static const char *TGSI_INTERPOLATES[] = +{ + "INTERPOLATE_CONSTANT", + "INTERPOLATE_LINEAR", + "INTERPOLATE_PERSPECTIVE", + "INTERPOLATE_ATTRIB" +}; + +static const char *TGSI_INTERPOLATES_SHORT[] = +{ + "CONSTANT", + "LINEAR", + "PERSPECTIVE", + "ATTRIB" +}; + +static const char *TGSI_SEMANTICS[] = +{ + "SEMANTIC_POSITION", + "SEMANTIC_COLOR", + "SEMANTIC_BCOLOR", + "SEMANTIC_FOG", + "SEMANTIC_PSIZE", + "SEMANTIC_GENERIC," +}; + +static const char *TGSI_SEMANTICS_SHORT[] = +{ + "POSITION", + "COLOR", + "BCOLOR", + "FOG", + "PSIZE", + "GENERIC", +}; + +static const char *TGSI_IMMS[] = +{ + "IMM_FLOAT32" +}; + +static const char *TGSI_IMMS_SHORT[] = +{ + "FLT32" +}; + +static const char *TGSI_OPCODES[] = +{ + "OPCODE_ARL", + "OPCODE_MOV", + "OPCODE_LIT", + "OPCODE_RCP", + "OPCODE_RSQ", + "OPCODE_EXP", + "OPCODE_LOG", + "OPCODE_MUL", + "OPCODE_ADD", + "OPCODE_DP3", + "OPCODE_DP4", + "OPCODE_DST", + "OPCODE_MIN", + "OPCODE_MAX", + "OPCODE_SLT", + "OPCODE_SGE", + "OPCODE_MAD", + "OPCODE_SUB", + "OPCODE_LERP", + "OPCODE_CND", + "OPCODE_CND0", + "OPCODE_DOT2ADD", + "OPCODE_INDEX", + "OPCODE_NEGATE", + "OPCODE_FRAC", + "OPCODE_CLAMP", + "OPCODE_FLOOR", + "OPCODE_ROUND", + "OPCODE_EXPBASE2", + "OPCODE_LOGBASE2", + "OPCODE_POWER", + "OPCODE_CROSSPRODUCT", + "OPCODE_MULTIPLYMATRIX", + "OPCODE_ABS", + "OPCODE_RCC", + "OPCODE_DPH", + "OPCODE_COS", + "OPCODE_DDX", + "OPCODE_DDY", + "OPCODE_KILP", + "OPCODE_PK2H", + "OPCODE_PK2US", + "OPCODE_PK4B", + "OPCODE_PK4UB", + "OPCODE_RFL", + "OPCODE_SEQ", + "OPCODE_SFL", + "OPCODE_SGT", + "OPCODE_SIN", + "OPCODE_SLE", + "OPCODE_SNE", + "OPCODE_STR", + "OPCODE_TEX", + "OPCODE_TXD", + "OPCODE_UP2H", + "OPCODE_UP2US", + "OPCODE_UP4B", + "OPCODE_UP4UB", + "OPCODE_X2D", + "OPCODE_ARA", + "OPCODE_ARR", + "OPCODE_BRA", + "OPCODE_CAL", + "OPCODE_RET", + "OPCODE_SSG", + "OPCODE_CMP", + "OPCODE_SCS", + "OPCODE_TXB", + "OPCODE_NRM", + "OPCODE_DIV", + "OPCODE_DP2", + "OPCODE_TXL", + "OPCODE_BRK", + "OPCODE_IF", + "OPCODE_LOOP", + "OPCODE_REP", + "OPCODE_ELSE", + "OPCODE_ENDIF", + "OPCODE_ENDLOOP", + "OPCODE_ENDREP", + "OPCODE_PUSHA", + "OPCODE_POPA", + "OPCODE_CEIL", + "OPCODE_I2F", + "OPCODE_NOT", + "OPCODE_TRUNC", + "OPCODE_SHL", + "OPCODE_SHR", + "OPCODE_AND", + "OPCODE_OR", + "OPCODE_MOD", + "OPCODE_XOR", + "OPCODE_SAD", + "OPCODE_TXF", + "OPCODE_TXQ", + "OPCODE_CONT", + "OPCODE_EMIT", + "OPCODE_ENDPRIM", + "OPCODE_BGNLOOP2", + "OPCODE_BGNSUB", + "OPCODE_ENDLOOP2", + "OPCODE_ENDSUB", + "OPCODE_NOISE1", + "OPCODE_NOISE2", + "OPCODE_NOISE3", + "OPCODE_NOISE4", + "OPCODE_NOP", + "OPCODE_TEXBEM", + "OPCODE_TEXBEML", + "OPCODE_TEXREG2AR", + "OPCODE_TEXM3X2PAD", + "OPCODE_TEXM3X2TEX", + "OPCODE_TEXM3X3PAD", + "OPCODE_TEXM3X3TEX", + "OPCODE_TEXM3X3SPEC", + "OPCODE_TEXM3X3VSPEC", + "OPCODE_TEXREG2GB", + "OPCODE_TEXREG2RGB", + "OPCODE_TEXDP3TEX", + "OPCODE_TEXDP3", + "OPCODE_TEXM3X3", + "OPCODE_TEXM3X2DEPTH", + "OPCODE_TEXDEPTH", + "OPCODE_BEM", + "OPCODE_M4X3", + "OPCODE_M3X4", + "OPCODE_M3X3", + "OPCODE_M3X2", + "OPCODE_NRM4", + "OPCODE_CALLNZ", + "OPCODE_IFC", + "OPCODE_BREAKC", + "OPCODE_TXP", + "OPCODE_KIL", + "OPCODE_END" +}; + +static const char *TGSI_OPCODES_SHORT[] = +{ + "ARL", + "MOV", + "LIT", + "RCP", + "RSQ", + "EXP", + "LOG", + "MUL", + "ADD", + "DP3", + "DP4", + "DST", + "MIN", + "MAX", + "SLT", + "SGE", + "MAD", + "SUB", + "LERP", + "CND", + "CND0", + "DOT2ADD", + "INDEX", + "NEGATE", + "FRAC", + "CLAMP", + "FLOOR", + "ROUND", + "EXPBASE2", + "LOGBASE2", + "POWER", + "CROSSPRODUCT", + "MULTIPLYMATRIX", + "ABS", + "RCC", + "DPH", + "COS", + "DDX", + "DDY", + "KILP", + "PK2H", + "PK2US", + "PK4B", + "PK4UB", + "RFL", + "SEQ", + "SFL", + "SGT", + "SIN", + "SLE", + "SNE", + "STR", + "TEX", + "TXD", + "UP2H", + "UP2US", + "UP4B", + "UP4UB", + "X2D", + "ARA", + "ARR", + "BRA", + "CAL", + "RET", + "SSG", + "CMP", + "SCS", + "TXB", + "NRM", + "DIV", + "DP2", + "TXL", + "BRK", + "IF", + "LOOP", + "REP", + "ELSE", + "ENDIF", + "ENDLOOP", + "ENDREP", + "PUSHA", + "POPA", + "CEIL", + "I2F", + "NOT", + "TRUNC", + "SHL", + "SHR", + "AND", + "OR", + "MOD", + "XOR", + "SAD", + "TXF", + "TXQ", + "CONT", + "EMIT", + "ENDPRIM", + "BGNLOOP2", + "BGNSUB", + "ENDLOOP2", + "ENDSUB", + "NOISE1", + "NOISE2", + "NOISE3", + "NOISE4", + "NOP", + "TEXBEM", + "TEXBEML", + "TEXREG2AR", + "TEXM3X2PAD", + "TEXM3X2TEX", + "TEXM3X3PAD", + "TEXM3X3TEX", + "TEXM3X3SPEC", + "TEXM3X3VSPEC", + "TEXREG2GB", + "TEXREG2RGB", + "TEXDP3TEX", + "TEXDP3", + "TEXM3X3", + "TEXM3X2DEPTH", + "TEXDEPTH", + "BEM", + "M4X3", + "M3X4", + "M3X3", + "M3X2", + "NRM4", + "CALLNZ", + "IFC", + "BREAKC", + "TXP", + "KIL", + "END" +}; + +static const char *TGSI_SATS[] = +{ + "SAT_NONE", + "SAT_ZERO_ONE", + "SAT_MINUS_PLUS_ONE" +}; + +static const char *TGSI_INSTRUCTION_EXTS[] = +{ + "INSTRUCTION_EXT_TYPE_NV", + "INSTRUCTION_EXT_TYPE_LABEL", + "INSTRUCTION_EXT_TYPE_TEXTURE" +}; + +static const char *TGSI_PRECISIONS[] = +{ + "PRECISION_DEFAULT", + "TGSI_PRECISION_FLOAT32", + "TGSI_PRECISION_FLOAT16", + "TGSI_PRECISION_FIXED12" +}; + +static const char *TGSI_CCS[] = +{ + "CC_GT", + "CC_EQ", + "CC_LT", + "CC_UN", + "CC_GE", + "CC_LE", + "CC_NE", + "CC_TR", + "CC_FL" +}; + +static const char *TGSI_SWIZZLES[] = +{ + "SWIZZLE_X", + "SWIZZLE_Y", + "SWIZZLE_Z", + "SWIZZLE_W" +}; + +static const char *TGSI_SWIZZLES_SHORT[] = +{ + "x", + "y", + "z", + "w" +}; + +static const char *TGSI_TEXTURES[] = +{ + "TEXTURE_UNKNOWN", + "TEXTURE_1D", + "TEXTURE_2D", + "TEXTURE_3D", + "TEXTURE_CUBE", + "TEXTURE_RECT", + "TEXTURE_SHADOW1D", + "TEXTURE_SHADOW2D", + "TEXTURE_SHADOWRECT" +}; + +static const char *TGSI_SRC_REGISTER_EXTS[] = +{ + "SRC_REGISTER_EXT_TYPE_SWZ", + "SRC_REGISTER_EXT_TYPE_MOD" +}; + +static const char *TGSI_EXTSWIZZLES[] = +{ + "EXTSWIZZLE_X", + "EXTSWIZZLE_Y", + "EXTSWIZZLE_Z", + "EXTSWIZZLE_W", + "EXTSWIZZLE_ZERO", + "EXTSWIZZLE_ONE" +}; + +static const char *TGSI_EXTSWIZZLES_SHORT[] = +{ + "x", + "y", + "z", + "w", + "0", + "1" +}; + +static const char *TGSI_WRITEMASKS[] = +{ + "0", + "WRITEMASK_X", + "WRITEMASK_Y", + "WRITEMASK_XY", + "WRITEMASK_Z", + "WRITEMASK_XZ", + "WRITEMASK_YZ", + "WRITEMASK_XYZ", + "WRITEMASK_W", + "WRITEMASK_XW", + "WRITEMASK_YW", + "WRITEMASK_XYW", + "WRITEMASK_ZW", + "WRITEMASK_XZW", + "WRITEMASK_YZW", + "WRITEMASK_XYZW" +}; + +static const char *TGSI_DST_REGISTER_EXTS[] = +{ + "DST_REGISTER_EXT_TYPE_CONDCODE", + "DST_REGISTER_EXT_TYPE_MODULATE" +}; + +static const char *TGSI_MODULATES[] = +{ + "MODULATE_1X", + "MODULATE_2X", + "MODULATE_4X", + "MODULATE_8X", + "MODULATE_HALF", + "MODULATE_QUARTER", + "MODULATE_EIGHTH" +}; + +static void +dump_declaration_short( + struct text_dump *dump, + struct tgsi_full_declaration *decl ) +{ + TXT( "\nDCL " ); + ENM( decl->Declaration.File, TGSI_FILES_SHORT ); + + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + CHR( '[' ); + UID( decl->u.DeclarationRange.First ); + if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { + TXT( ".." ); + UID( decl->u.DeclarationRange.Last ); + } + CHR( ']' ); + break; + default: + assert( 0 ); + } + + if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + + if( decl->Declaration.Interpolate ) { + TXT( ", " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); + } + + if( decl->Declaration.Semantic ) { + TXT( ", " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); + CHR( '[' ); + UID( decl->Semantic.SemanticIndex ); + CHR( ']' ); + } +} + +static void +dump_declaration_verbose( + struct text_dump *dump, + struct tgsi_full_declaration *decl, + unsigned ignored, + unsigned deflt, + struct tgsi_full_declaration *fd ) +{ + TXT( "\nFile : " ); + ENM( decl->Declaration.File, TGSI_FILES ); + TXT( "\nDeclare : " ); + ENM( decl->Declaration.Declare, TGSI_DECLARES ); + if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { + TXT( "\nUsageMask : " ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'X' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'Y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'Z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'W' ); + } + } + if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { + TXT( "\nInterpolate: " ); + UID( decl->Declaration.Interpolate ); + } + if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { + TXT( "\nSemantic : " ); + UID( decl->Declaration.Semantic ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Declaration.Padding ); + } + + CHR( '\n' ); + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + TXT( "\nFirst: " ); + UID( decl->u.DeclarationRange.First ); + TXT( "\nLast : " ); + UID( decl->u.DeclarationRange.Last ); + break; + + case TGSI_DECLARE_MASK: + TXT( "\nMask: " ); + UIX( decl->u.DeclarationMask.Mask ); + break; + + default: + assert( 0 ); + } + + if( decl->Declaration.Interpolate ) { + CHR( '\n' ); + TXT( "\nInterpolate: " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Interpolation.Padding ); + } + } + + if( decl->Declaration.Semantic ) { + CHR( '\n' ); + TXT( "\nSemanticName : " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); + TXT( "\nSemanticIndex: " ); + UID( decl->Semantic.SemanticIndex ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Semantic.Padding ); + } + } +} + +static void +dump_immediate_short( + struct text_dump *dump, + struct tgsi_full_immediate *imm ) +{ + unsigned i; + + TXT( "\nIMM " ); + ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); + + TXT( " { " ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + + if( i < imm->Immediate.Size - 2 ) { + TXT( ", " ); + } + } + TXT( " }" ); +} + +static void +dump_immediate_verbose( + struct text_dump *dump, + struct tgsi_full_immediate *imm, + unsigned ignored ) +{ + unsigned i; + + TXT( "\nDataType : " ); + ENM( imm->Immediate.DataType, TGSI_IMMS ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( imm->Immediate.Padding ); + } + + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + CHR( '\n' ); + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + TXT( "\nFloat: " ); + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + } +} + +static void +dump_instruction_short( + struct text_dump *dump, + struct tgsi_full_instruction *inst, + unsigned instno ) +{ + unsigned i; + boolean first_reg = TRUE; + + CHR( '\n' ); + UID( instno ); + CHR( ':' ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); + + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + TXT( "_SAT" ); + break; + case TGSI_SAT_MINUS_PLUS_ONE: + TXT( "_SAT[-1,1]" ); + break; + default: + assert( 0 ); + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( dst->DstRegister.Index ); + CHR( ']' ); + + if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + + first_reg = FALSE; + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + if( src->SrcRegisterExtMod.Negate ) { + CHR( '-' ); + } + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + if( src->SrcRegister.Negate ) { + CHR( '-' ); + } + + ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( src->SrcRegister.Index ); + CHR( ']' ); + + if (src->SrcRegister.Extended) { + if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { + CHR( '.' ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); + } + } + else if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { + CHR( '.' ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); + } + + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + + first_reg = FALSE; + } + + switch( inst->Instruction.Opcode ) { + case TGSI_OPCODE_IF: + case TGSI_OPCODE_ELSE: + case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_CAL: + TXT( " :" ); + UID( inst->InstructionExtLabel.Label ); + break; + } +} + +static void +dump_instruction_verbose( + struct text_dump *dump, + struct tgsi_full_instruction *inst, + unsigned ignored, + unsigned deflt, + struct tgsi_full_instruction *fi ) +{ + unsigned i; + + TXT( "\nOpcode : " ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES ); + if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { + TXT( "\nSaturate : " ); + ENM( inst->Instruction.Saturate, TGSI_SATS ); + } + if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { + TXT( "\nNumDstRegs : " ); + UID( inst->Instruction.NumDstRegs ); + } + if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { + TXT( "\nNumSrcRegs : " ); + UID( inst->Instruction.NumSrcRegs ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->Instruction.Padding ); + } + + if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { + TXT( "\nPrecision : " ); + ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); + } + if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { + TXT( "\nCondDstIndex : " ); + UID( inst->InstructionExtNv.CondDstIndex ); + } + if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { + TXT( "\nCondFlowIndex : " ); + UID( inst->InstructionExtNv.CondFlowIndex ); + } + if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { + TXT( "\nCondMask : " ); + ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { + TXT( "\nCondSwizzleX : " ); + ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { + TXT( "\nCondSwizzleY : " ); + ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ : " ); + ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { + TXT( "\nCondSwizzleW : " ); + ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { + TXT( "\nCondDstUpdate : " ); + UID( inst->InstructionExtNv.CondDstUpdate ); + } + if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { + TXT( "\nCondFlowEnable: " ); + UID( inst->InstructionExtNv.CondFlowEnable ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtNv.Padding ); + if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { + TXT( "\nExtended : " ); + UID( inst->InstructionExtNv.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { + TXT( "\nLabel : " ); + UID( inst->InstructionExtLabel.Label ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtLabel.Padding ); + if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtLabel.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { + TXT( "\nTexture : " ); + ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtTexture.Padding ); + if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtTexture.Extended ); + } + } + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : " ); + ENM( dst->DstRegister.File, TGSI_FILES ); + if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { + TXT( "\nWriteMask: " ); + ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); + } + if( ignored ) { + if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( dst->DstRegister.Indirect ); + } + if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( dst->DstRegister.Dimension ); + } + } + if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { + TXT( "\nIndex : " ); + SID( dst->DstRegister.Index ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegister.Padding ); + if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegister.Extended ); + } + } + + if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { + TXT( "\nCondMask : " ); + ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { + TXT( "\nCondSwizzleX: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { + TXT( "\nCondSwizzleY: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { + TXT( "\nCondSwizzleW: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { + TXT( "\nCondSrcIndex: " ); + UID( dst->DstRegisterExtConcode.CondSrcIndex ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtConcode.Padding ); + if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegisterExtConcode.Extended ); + } + } + } + + if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { + TXT( "\nModulate: " ); + ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtModulate.Padding ); + if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { + TXT( "\nExtended: " ); + UID( dst->DstRegisterExtModulate.Extended ); + } + } + } + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : "); + ENM( src->SrcRegister.File, TGSI_FILES ); + if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { + TXT( "\nSwizzleX : " ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { + TXT( "\nSwizzleY : " ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { + TXT( "\nSwizzleZ : " ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { + TXT( "\nSwizzleW : " ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegister.Negate ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( src->SrcRegister.Indirect ); + } + if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( src->SrcRegister.Dimension ); + } + } + if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { + TXT( "\nIndex : " ); + SID( src->SrcRegister.Index ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegister.Extended ); + } + } + + if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { + TXT( "\nExtSwizzleX: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { + TXT( "\nExtSwizzleY: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { + TXT( "\nExtSwizzleZ: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { + TXT( "\nExtSwizzleW: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { + TXT( "\nNegateX : " ); + UID( src->SrcRegisterExtSwz.NegateX ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { + TXT( "\nNegateY : " ); + UID( src->SrcRegisterExtSwz.NegateY ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { + TXT( "\nNegateZ : " ); + UID( src->SrcRegisterExtSwz.NegateZ ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { + TXT( "\nNegateW : " ); + UID( src->SrcRegisterExtSwz.NegateW ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { + TXT( "\nExtDivide : " ); + ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtSwz.Padding ); + if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtSwz.Extended ); + } + } + } + + if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { + TXT( "\nComplement: " ); + UID( src->SrcRegisterExtMod.Complement ); + } + if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { + TXT( "\nBias : " ); + UID( src->SrcRegisterExtMod.Bias ); + } + if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { + TXT( "\nScale2X : " ); + UID( src->SrcRegisterExtMod.Scale2X ); + } + if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { + TXT( "\nAbsolute : " ); + UID( src->SrcRegisterExtMod.Absolute ); + } + if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegisterExtMod.Negate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtMod.Padding ); + if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtMod.Extended ); + } + } + } + } +} + +void +tgsi_dump( + const struct tgsi_token *tokens, + unsigned flags ) +{ + struct text_dump _dump; + struct text_dump *dump = &_dump; + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned verbose = flags & TGSI_DUMP_VERBOSE; + unsigned ignored = !(flags & TGSI_DUMP_NO_IGNORED); + unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); + unsigned instno = 0; + + { +#if 0 + static unsigned counter = 0; + char buffer[64]; + + sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); + dump->file = fopen( buffer, "wt" ); +#else + dump->file = stderr; +#endif + dump->tabs = 0; + } + + /* sanity check */ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); + + tgsi_parse_init( &parse, tokens ); + + TXT( "tgsi-dump begin -----------------" ); + + CHR( '\n' ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); + CHR( ' ' ); + UID( parse.FullVersion.Version.MajorVersion ); + CHR( '.' ); + UID( parse.FullVersion.Version.MinorVersion ); + + if( verbose ) { + TXT( "\nMajorVersion: " ); + UID( parse.FullVersion.Version.MajorVersion ); + TXT( "\nMinorVersion: " ); + UID( parse.FullVersion.Version.MinorVersion ); + CHR( '\n' ); + + TXT( "\nHeaderSize: " ); + UID( parse.FullHeader.Header.HeaderSize ); + TXT( "\nBodySize : " ); + UID( parse.FullHeader.Header.BodySize ); + TXT( "\nProcessor : " ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + CHR( '\n' ); + } + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_short( + dump, + &parse.FullToken.FullDeclaration ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_short( + dump, + &parse.FullToken.FullImmediate ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_short( + dump, + &parse.FullToken.FullInstruction, + instno ); + instno++; + break; + + default: + assert( 0 ); + } + + if( verbose ) { + TXT( "\nType : " ); + ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); + if( ignored ) { + TXT( "\nSize : " ); + UID( parse.FullToken.Token.Size ); + if( deflt || parse.FullToken.Token.Extended ) { + TXT( "\nExtended : " ); + UID( parse.FullToken.Token.Extended ); + } + } + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_verbose( + dump, + &parse.FullToken.FullDeclaration, + ignored, + deflt, + &fd ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_verbose( + dump, + &parse.FullToken.FullImmediate, + ignored ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_verbose( + dump, + &parse.FullToken.FullInstruction, + ignored, + deflt, + &fi ); + break; + + default: + assert( 0 ); + } + + CHR( '\n' ); + } + } + + TXT( "\ntgsi-dump end -------------------\n" ); + + tgsi_parse_free( &parse ); +} + diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.h b/src/mesa/pipe/tgsi/util/tgsi_dump.h new file mode 100644 index 0000000000..70860c0885 --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_dump.h @@ -0,0 +1,22 @@ +#if !defined TGSI_DUMP_H +#define TGSI_DUMP_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +#define TGSI_DUMP_VERBOSE 1 +#define TGSI_DUMP_NO_IGNORED 2 +#define TGSI_DUMP_NO_DEFAULT 4 + +void +tgsi_dump( + const struct tgsi_token *tokens, + unsigned flags ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_DUMP_H + diff --git a/src/mesa/pipe/tgsi/util/tgsi_parse.c b/src/mesa/pipe/tgsi/util/tgsi_parse.c new file mode 100644 index 0000000000..f0f8d44ac2 --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_parse.c @@ -0,0 +1,318 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ) +{ + full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; +} + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ) +{ + if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { + FREE( full_token->FullImmediate.u.Pointer ); + } +} + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ) +{ + ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; + if( ctx->FullVersion.Version.MajorVersion > 1 ) { + return TGSI_PARSE_ERROR; + } + + ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; + if( ctx->FullHeader.Header.HeaderSize >= 2 ) { + ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; + } + else { + ctx->FullHeader.Processor = tgsi_default_processor(); + } + + ctx->Tokens = tokens; + ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; + + tgsi_full_token_init( &ctx->FullToken ); + + return TGSI_PARSE_OK; +} + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ) +{ + tgsi_full_token_free( &ctx->FullToken ); +} + +boolean +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ) +{ + return ctx->Position >= + 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; +} + +static void +next_token( + struct tgsi_parse_context *ctx, + void *token ) +{ + assert( !tgsi_parse_end_of_tokens( ctx ) ); + + *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; +} + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ) +{ + struct tgsi_token token; + unsigned i; + + tgsi_full_token_free( &ctx->FullToken ); + tgsi_full_token_init( &ctx->FullToken ); + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; + + *decl = tgsi_default_full_declaration(); + decl->Declaration = *(struct tgsi_declaration *) &token; + + switch( decl->Declaration.Type ) { + case TGSI_DECLARE_RANGE: + next_token( ctx, &decl->u.DeclarationRange ); + break; + + case TGSI_DECLARE_MASK: + next_token( ctx, &decl->u.DeclarationMask ); + break; + + default: + assert (0); + } + + if( decl->Declaration.Interpolate ) { + next_token( ctx, &decl->Interpolation ); + } + + if( decl->Declaration.Semantic ) { + next_token( ctx, &decl->Semantic ); + } + + break; + } + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; + + *imm = tgsi_default_full_immediate(); + imm->Immediate = *(struct tgsi_immediate *) &token; + + assert( !imm->Immediate.Extended ); + + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + imm->u.Pointer = MALLOC( + sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + next_token( ctx, &imm->u.ImmediateFloat32[i] ); + } + break; + + default: + assert( 0 ); + } + + break; + } + + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; + unsigned extended; + + *inst = tgsi_default_full_instruction(); + inst->Instruction = *(struct tgsi_instruction *) &token; + + extended = inst->Instruction.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_INSTRUCTION_EXT_TYPE_NV: + inst->InstructionExtNv = + *(struct tgsi_instruction_ext_nv *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_LABEL: + inst->InstructionExtLabel = + *(struct tgsi_instruction_ext_label *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: + inst->InstructionExtTexture = + *(struct tgsi_instruction_ext_texture *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); + assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); + + extended = inst->FullDstRegisters[i].DstRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: + inst->FullDstRegisters[i].DstRegisterExtConcode = + *(struct tgsi_dst_register_ext_concode *) &token; + break; + + case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: + inst->FullDstRegisters[i].DstRegisterExtModulate = + *(struct tgsi_dst_register_ext_modulate *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + } + + assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); + + extended = inst->FullSrcRegisters[i].SrcRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: + inst->FullSrcRegisters[i].SrcRegisterExtSwz = + *(struct tgsi_src_register_ext_swz *) &token; + break; + + case TGSI_SRC_REGISTER_EXT_TYPE_MOD: + inst->FullSrcRegisters[i].SrcRegisterExtMod = + *(struct tgsi_src_register_ext_mod *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + + if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); + + /* + * No support for multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); + + if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + } + } + + break; + } + + default: + assert( 0 ); + } +} + diff --git a/src/mesa/pipe/tgsi/util/tgsi_parse.h b/src/mesa/pipe/tgsi/util/tgsi_parse.h new file mode 100644 index 0000000000..9372da8d5d --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_parse.h @@ -0,0 +1,121 @@ +#if !defined TGSI_PARSE_H +#define TGSI_PARSE_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_full_version +{ + struct tgsi_version Version; +}; + +struct tgsi_full_header +{ + struct tgsi_header Header; + struct tgsi_processor Processor; +}; + +struct tgsi_full_dst_register +{ + struct tgsi_dst_register DstRegister; + struct tgsi_dst_register_ext_concode DstRegisterExtConcode; + struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; +}; + +struct tgsi_full_src_register +{ + struct tgsi_src_register SrcRegister; + struct tgsi_src_register_ext_swz SrcRegisterExtSwz; + struct tgsi_src_register_ext_mod SrcRegisterExtMod; + struct tgsi_src_register SrcRegisterInd; + struct tgsi_dimension SrcRegisterDim; + struct tgsi_src_register SrcRegisterDimInd; +}; + +struct tgsi_full_declaration +{ + struct tgsi_declaration Declaration; + union + { + struct tgsi_declaration_range DeclarationRange; + struct tgsi_declaration_mask DeclarationMask; + } u; + struct tgsi_declaration_interpolation Interpolation; + struct tgsi_declaration_semantic Semantic; +}; + +struct tgsi_full_immediate +{ + struct tgsi_immediate Immediate; + union + { + void *Pointer; + struct tgsi_immediate_float32 *ImmediateFloat32; + } u; +}; + +#define TGSI_FULL_MAX_DST_REGISTERS 2 +#define TGSI_FULL_MAX_SRC_REGISTERS 3 + +struct tgsi_full_instruction +{ + struct tgsi_instruction Instruction; + struct tgsi_instruction_ext_nv InstructionExtNv; + struct tgsi_instruction_ext_label InstructionExtLabel; + struct tgsi_instruction_ext_texture InstructionExtTexture; + struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; + struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; +}; + +union tgsi_full_token +{ + struct tgsi_token Token; + struct tgsi_full_declaration FullDeclaration; + struct tgsi_full_immediate FullImmediate; + struct tgsi_full_instruction FullInstruction; +}; + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ); + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ); + +struct tgsi_parse_context +{ + const struct tgsi_token *Tokens; + unsigned Position; + struct tgsi_full_version FullVersion; + struct tgsi_full_header FullHeader; + union tgsi_full_token FullToken; +}; + +#define TGSI_PARSE_OK 0 +#define TGSI_PARSE_ERROR 1 + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ); + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ); + +boolean +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ); + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_PARSE_H + diff --git a/src/mesa/pipe/tgsi/util/tgsi_util.c b/src/mesa/pipe/tgsi/util/tgsi_util.c new file mode 100644 index 0000000000..1e76b0f133 --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_util.c @@ -0,0 +1,273 @@ +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" +#include "tgsi_util.h" + +union pointer_hack +{ + void *pointer; + unsigned long long uint64; +}; + +void * +tgsi_align_128bit( + void *unaligned ) +{ + union pointer_hack ph; + + ph.uint64 = 0; + ph.pointer = unaligned; + ph.uint64 = (ph.uint64 + 15) & ~15; + return ph.pointer; +} + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->SwizzleX; + case 1: + return reg->SwizzleY; + case 2: + return reg->SwizzleZ; + case 3: + return reg->SwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->ExtSwizzleX; + case 1: + return reg->ExtSwizzleY; + case 2: + return reg->ExtSwizzleZ; + case 3: + return reg->ExtSwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned swizzle; + + /* + * First, calculate the extended swizzle for a given channel. This will give + * us either a channel index into the simple swizzle or a constant 1 or 0. + */ + swizzle = tgsi_util_get_src_register_extswizzle( + ®->SrcRegisterExtSwz, + component ); + + assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); + assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); + assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); + assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); + + /* + * Second, calculate the simple swizzle for the unswizzled channel index. + * Leave the constants intact, they are not affected by the simple swizzle. + */ + if( swizzle <= TGSI_SWIZZLE_W ) { + swizzle = tgsi_util_get_src_register_swizzle( + ®->SrcRegister, + component ); + } + + return swizzle; +} + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + unsigned swizzle, + unsigned component ) +{ + switch( component ) { + case 0: + reg->SwizzleX = swizzle; + break; + case 1: + reg->SwizzleY = swizzle; + break; + case 2: + reg->SwizzleZ = swizzle; + break; + case 3: + reg->SwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + unsigned swizzle, + unsigned component ) +{ + switch( component ) { + case 0: + reg->ExtSwizzleX = swizzle; + break; + case 1: + reg->ExtSwizzleY = swizzle; + break; + case 2: + reg->ExtSwizzleZ = swizzle; + break; + case 3: + reg->ExtSwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->NegateX; + case 1: + return reg->NegateY; + case 2: + return reg->NegateZ; + case 3: + return reg->NegateW; + default: + assert( 0 ); + } + return 0; +} + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ) +{ + switch( component ) { + case 0: + reg->NegateX = negate; + break; + case 1: + reg->NegateY = negate; + break; + case 2: + reg->NegateZ = negate; + break; + case 3: + reg->NegateW = negate; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned sign_mode; + + if( reg->SrcRegisterExtMod.Absolute ) { + /* Consider only the post-abs negation. */ + + if( reg->SrcRegisterExtMod.Negate ) { + sign_mode = TGSI_UTIL_SIGN_SET; + } + else { + sign_mode = TGSI_UTIL_SIGN_CLEAR; + } + } + else { + /* Accumulate the three negations. */ + + unsigned negate; + + negate = reg->SrcRegister.Negate; + if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { + negate = !negate; + } + if( reg->SrcRegisterExtMod.Negate ) { + negate = !negate; + } + + if( negate ) { + sign_mode = TGSI_UTIL_SIGN_TOGGLE; + } + else { + sign_mode = TGSI_UTIL_SIGN_KEEP; + } + } + + return sign_mode; +} + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + unsigned sign_mode ) +{ + reg->SrcRegisterExtSwz.NegateX = 0; + reg->SrcRegisterExtSwz.NegateY = 0; + reg->SrcRegisterExtSwz.NegateZ = 0; + reg->SrcRegisterExtSwz.NegateW = 0; + + switch (sign_mode) + { + case TGSI_UTIL_SIGN_CLEAR: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_SET: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 1; + break; + + case TGSI_UTIL_SIGN_TOGGLE: + reg->SrcRegister.Negate = 1; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_KEEP: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + default: + assert( 0 ); + } +} + diff --git a/src/mesa/pipe/tgsi/util/tgsi_util.h b/src/mesa/pipe/tgsi/util/tgsi_util.h new file mode 100644 index 0000000000..ef14446f0e --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_util.h @@ -0,0 +1,70 @@ +#if !defined TGSI_UTIL_H +#define TGSI_UTIL_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +void * +tgsi_align_128bit( + void *unaligned ); + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component); + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + unsigned swizzle, + unsigned component ); + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + unsigned swizzle, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ); + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ); + +#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ +#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ +#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ +#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + unsigned sign_mode ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_UTIL_H + diff --git a/src/mesa/sources b/src/mesa/sources index ea6840f3d9..2df60d1996 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -178,15 +178,14 @@ DRAW_SOURCES = \ pipe/draw/draw_wide_prims.c TGSIEXEC_SOURCES = \ - pipe/tgsi/exec/tgsi_build.c \ - pipe/tgsi/exec/tgsi_dump.c \ pipe/tgsi/exec/tgsi_exec.c \ - pipe/tgsi/exec/tgsi_parse.c \ - pipe/tgsi/exec/tgsi_sse2.c \ - pipe/tgsi/exec/tgsi_util.c + pipe/tgsi/exec/tgsi_sse2.c -TGSIDECO_SOURCES = \ - pipe/tgsi/deco/deco_caps.c +TGSIUTIL_SOURCES = \ + pipe/tgsi/util/tgsi_build.c \ + pipe/tgsi/util/tgsi_dump.c \ + pipe/tgsi/util/tgsi_parse.c \ + pipe/tgsi/util/tgsi_util.c LLVMTGSI_SOURCES = \ @@ -385,7 +384,7 @@ SOLO_SOURCES = \ $(VF_SOURCES) \ $(DRAW_SOURCES) \ $(TGSIEXEC_SOURCES) \ - $(TGSIDECO_SOURCES) \ + $(TGSIUTIL_SOURCES) \ $(STATECACHE_SOURCES) \ $(STATETRACKER_SOURCES) \ $(TNL_SOURCES) \ diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index d4976941f9..4ec10badad 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -41,7 +41,7 @@ #include "main/mtypes.h" #include "pipe/p_context.h" -#include "pipe/tgsi/exec/tgsi_core.h" +#include "pipe/p_shader_tokens.h" #include "st_context.h" #include "st_cache.h" diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index e00684ac03..cffd66751d 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -30,7 +30,8 @@ #include "shader/prog_print.h" #include "pipe/p_state.h" -#include "pipe/tgsi/exec/tgsi_core.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_dump.h" #include "pipe/cso_cache/cso_cache.h" diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index bc57868af2..b392edf16d 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -30,9 +30,11 @@ * Michal Krol */ - -#include "tgsi_platform.h" -#include "pipe/tgsi/exec/tgsi_core.h" +#include "pipe/p_compiler.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/util/tgsi_build.h" +#include "pipe/tgsi/util/tgsi_util.h" #include "st_mesa_to_tgsi.h" #include "shader/prog_instruction.h" #include "shader/prog_parameter.h" diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index ce5384671c..4e8bdc820e 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -36,8 +36,8 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" #include "pipe/draw/draw_context.h" -#include "pipe/tgsi/exec/tgsi_core.h" #include "st_context.h" #include "st_cache.h" diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 51877af8da..049f9f659f 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -35,7 +35,7 @@ #define ST_PROGRAM_H #include "mtypes.h" -#include "pipe/tgsi/exec/tgsi_token.h" +#include "pipe/p_shader_tokens.h" #include "x86/rtasm/x86sse.h" -- cgit v1.2.3 From b3dee0185ec0202d454075de4e4011e3ac093df8 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 23 Nov 2007 13:48:12 +0000 Subject: gallium: remove sp_headers.h references. --- src/mesa/pipe/softpipe/sp_quad.c | 1 - src/mesa/pipe/softpipe/sp_state_derived.c | 1 - 2 files changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 13fb883ef0..e0327c4cf9 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -27,7 +27,6 @@ #include "sp_context.h" -#include "sp_headers.h" #include "sp_state.h" #include "pipe/p_shader_tokens.h" diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 33caab9372..c4f1a0a01a 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -29,7 +29,6 @@ #include "pipe/p_shader_tokens.h" #include "pipe/draw/draw_context.h" #include "pipe/draw/draw_vertex.h" -#include "sp_headers.h" #include "sp_context.h" #include "sp_state.h" -- cgit v1.2.3 From 7043db677f457ae9a46f2585a5ef52bf69a4e8ea Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 24 Nov 2007 16:02:31 +0000 Subject: Cleanup PIPE_FORMAT names. Add a function that builds a display name of a given format token. --- src/mesa/pipe/p_format.h | 277 ++++++++++++++++++++++++------------- src/mesa/state_tracker/st_format.c | 14 +- 2 files changed, 189 insertions(+), 102 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index f42f987c4e..70eb124709 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -97,16 +97,18 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) return (f >> shift) & mask; } -#define pf_swizzle_x(f) pf_get(f, 2, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_y(f) pf_get(f, 5, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_z(f) pf_get(f, 8, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_w(f) pf_get(f, 11, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_size_x(f) pf_get(f, 14, 0x7) /**< Size of X - 1 */ -#define pf_size_y(f) pf_get(f, 17, 0x7) /**< Size of Y - 1 */ -#define pf_size_z(f) pf_get(f, 20, 0x7) /**< Size of Z - 1 */ -#define pf_size_w(f) pf_get(f, 23, 0x7) /**< Size of W - 1 */ -#define pf_exp8(f) pf_get(f, 26, 0x3) /**< Scale size by 8 ^ exp8 */ -#define pf_type(f) pf_get(f, 28, 0xf) /**< PIPE_FORMAT_TYPE_ */ +#define pf_swizzle_x(f) pf_get(f, 2, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_y(f) pf_get(f, 5, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_z(f) pf_get(f, 8, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_w(f) pf_get(f, 11, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_xyzw(f,i) pf_get(f, 2+((i)*3), 0x7) +#define pf_size_x(f) pf_get(f, 14, 0x7) /**< Size of X */ +#define pf_size_y(f) pf_get(f, 17, 0x7) /**< Size of Y */ +#define pf_size_z(f) pf_get(f, 20, 0x7) /**< Size of Z */ +#define pf_size_w(f) pf_get(f, 23, 0x7) /**< Size of W */ +#define pf_size_xyzw(f,i) pf_get(f, 14+((i)*3), 0x7) +#define pf_exp8(f) pf_get(f, 26, 0x3) /**< Scale size by 8 ^ exp8 */ +#define pf_type(f) pf_get(f, 28, 0xf) /**< PIPE_FORMAT_TYPE_ */ /** * Helper macro to encode the above structure into a 32-bit value. @@ -127,31 +129,19 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) #define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 0) | ((SWZY) << 3) | ((SWZZ) << 6) | ((SWZW) << 9)) /** - * Shorthand macro for RGBAZS layout with uniform component sizes in 1-bit units. + * Shorthand macro for RGBAZS layout with component sizes in 1-bit units. */ -#define _PIPE_FORMAT_RGBAZS_1U( SWZ, SIZE, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 0, TYPE ) - -/** - * Shorthand macro for RGBAZS layout with non-uniform component sizes in 1-bit units. - */ -#define _PIPE_FORMAT_RGBAZS_1N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ +#define _PIPE_FORMAT_RGBAZS_1( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE ) /** - * Shorthand macro for RGBAZS layout with uniform component sizes in 8-bit units. - */ -#define _PIPE_FORMAT_RGBAZS_8U( SWZ, SIZE, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZE, SIZE, SIZE, SIZE, 1, TYPE ) - -/** - * Shorthand macro for RGBAZS layout with non-uniform component sizes in 8-bit units. + * Shorthand macro for RGBAZS layout with component sizes in 8-bit units. */ -#define _PIPE_FORMAT_RGBAZS_8N( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ +#define _PIPE_FORMAT_RGBAZS_8( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) /** - * Shorthand macro for RGBAZS layout with non-uniform component sizes in 64-bit units. + * Shorthand macro for RGBAZS layout with component sizes in 64-bit units. */ #define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 2, TYPE ) @@ -207,83 +197,178 @@ static INLINE uint pf_rev(pipe_format_ycbcr_t f) * z24s8, compressed textures, ycbcr, etc that won't fit that model. */ -#define PIPE_FORMAT_NONE _PIPE_FORMAT_RGBAZS_1U( _PIPE_FORMAT_0000, 0, PIPE_FORMAT_TYPE_UNKNOWN ) /**< unstructured */ -#define PIPE_FORMAT_U_A8_R8_G8_B8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_ARGB, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte[4] ARGB */ -#define PIPE_FORMAT_U_A1_R5_G5_B5 _PIPE_FORMAT_RGBAZS_1N( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGBA */ -#define PIPE_FORMAT_U_A4_R4_G4_B4 _PIPE_FORMAT_RGBAZS_1U( _PIPE_FORMAT_ARGB, 4, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGBA */ -#define PIPE_FORMAT_U_R5_G6_B5 _PIPE_FORMAT_RGBAZS_1N( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 16-bit packed RGB */ -#define PIPE_FORMAT_U_L8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRR1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte luminance */ -#define PIPE_FORMAT_U_A8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_000R, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha */ -#define PIPE_FORMAT_U_I8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRRR, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte intensity */ -#define PIPE_FORMAT_U_A8_L8 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RRRG, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha, luminance */ -#define PIPE_FORMAT_S_R16_G16_B16_A16 _PIPE_FORMAT_RGBAZS_8U( _PIPE_FORMAT_RGBA, 2, PIPE_FORMAT_TYPE_SNORM ) /**< signed 16-bit RGBA (accum) */ +#define PIPE_FORMAT_NONE _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_0000, 0, 0, 0, 0, PIPE_FORMAT_TYPE_UNKNOWN ) +#define PIPE_FORMAT_A8R8G8B8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_A1R5G5B5_UNORM _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_A4R4G4B4_UNORM _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R5G6B5_UNORM _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_U_L8 _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte luminance */ +#define PIPE_FORMAT_U_A8 _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha */ +#define PIPE_FORMAT_U_I8 _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte intensity */ +#define PIPE_FORMAT_U_A8_L8 _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha, luminance */ #define PIPE_FORMAT_YCBCR _PIPE_FORMAT_YCBCR( 0 ) #define PIPE_FORMAT_YCBCR_REV _PIPE_FORMAT_YCBCR( 1 ) -#define PIPE_FORMAT_U_Z16 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< ushort Z/depth */ -#define PIPE_FORMAT_U_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< uint Z/depth */ -#define PIPE_FORMAT_F_Z32 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) /**< float Z/depth */ -#define PIPE_FORMAT_S8_Z24 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil + 24-bit Z */ -#define PIPE_FORMAT_Z24_S8 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 24-bit Z + 8-bit stencil */ -#define PIPE_FORMAT_U_S8 _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) /**< 8-bit stencil */ +#define PIPE_FORMAT_Z16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_Z32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_Z32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_S8Z24_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_Z24S8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_S8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) #define PIPE_FORMAT_R64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) #define PIPE_FORMAT_R64G64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) #define PIPE_FORMAT_R64G64B64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ) #define PIPE_FORMAT_R64G64B64A64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32G32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32G32B32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32G32B32A32_FLOAT _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32G32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32G32B32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32G32B32A32_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32G32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32G32B32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32G32B32A32_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32G32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32G32B32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32G32B32A32_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R32G32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R32G32B32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R32G32B32A32_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16G16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16G16B16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16G16B16A16_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16G16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16G16B16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16G16B16A16_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16G16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16G16B16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16G16B16A16_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16G16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16G16B16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16G16B16A16_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8G8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8G8B8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8G8B8A8_UNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8G8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8G8B8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8G8B8A8_USCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8G8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8G8B8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8G8B8A8_SNORM _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8G8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8G8B8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8G8B8A8_SSCALED _PIPE_FORMAT_RGBAZS_8N( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32G32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32G32B32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32G32B32A32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ) +#define PIPE_FORMAT_R32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32G32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32G32B32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32G32B32A32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R32_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32G32_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32G32B32_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32G32B32A32_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R32_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32G32_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32G32B32_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32G32B32A32_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R32_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32G32_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32G32B32_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R32G32B32A32_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16G16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16G16B16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16G16B16A16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R16_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16G16_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16G16B16_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16G16B16A16_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R16_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16G16_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16G16B16_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16G16B16A16_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R16_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16G16_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16G16B16_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R16G16B16A16_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8G8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8G8B8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8G8B8A8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_R8_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8G8_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8G8B8_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8G8B8A8_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ) +#define PIPE_FORMAT_R8_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8G8_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8G8B8_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8G8B8A8_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ) +#define PIPE_FORMAT_R8_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8G8_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8G8B8_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ) +#define PIPE_FORMAT_R8G8B8A8_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) -/* Duplicated formats: +/** + * Duplicated formats: */ +#define PIPE_FORMAT_U_A8_R8_G8_B8 PIPE_FORMAT_A8R8G8B8_UNORM +#define PIPE_FORMAT_U_A1_R5_G5_B5 PIPE_FORMAT_A1R5G5B5_UNORM +#define PIPE_FORMAT_U_A4_R4_G4_B4 PIPE_FORMAT_A4R4G4B4_UNORM #define PIPE_FORMAT_U_R8_G8_B8_A8 PIPE_FORMAT_R8G8B8A8_UNORM +#define PIPE_FORMAT_U_R5_G6_B5 PIPE_FORMAT_R5G6B5_UNORM +#define PIPE_FORMAT_S_R16_G16_B16_A16 PIPE_FORMAT_R16G16B16A16_SNORM +#define PIPE_FORMAT_U_Z16 PIPE_FORMAT_Z16_UNORM +#define PIPE_FORMAT_U_Z32 PIPE_FORMAT_Z32_UNORM +#define PIPE_FORMAT_F_Z32 PIPE_FORMAT_Z32_FLOAT +#define PIPE_FORMAT_S8_Z24 PIPE_FORMAT_S8Z24_UNORM +#define PIPE_FORMAT_Z24_S8 PIPE_FORMAT_Z24S8_UNORM +#define PIPE_FORMAT_U_S8 PIPE_FORMAT_S8_UNORM + +/** + * Builds pipe format name from format token. + */ +static INLINE char *pf_sprint_name( char *str, uint format ) +{ + strcpy( str, "PIPE_FORMAT_" ); + switch (pf_layout( format )) { + case PIPE_FORMAT_LAYOUT_RGBAZS: { + pipe_format_rgbazs_t rgbazs = (pipe_format_rgbazs_t) format; + uint i; + uint scale = 1 << (pf_exp8( rgbazs ) * 3); + + for (i = 0; i < 4; i++) { + uint size = pf_size_xyzw( rgbazs, i ); + + if (size == 0) { + break; + } + switch (pf_swizzle_xyzw( rgbazs, i )) { + case PIPE_FORMAT_COMP_R: + strcat( str, "R" ); + break; + case PIPE_FORMAT_COMP_G: + strcat( str, "G" ); + break; + case PIPE_FORMAT_COMP_B: + strcat( str, "B" ); + break; + case PIPE_FORMAT_COMP_A: + strcat( str, "A" ); + break; + case PIPE_FORMAT_COMP_0: + strcat( str, "0" ); + break; + case PIPE_FORMAT_COMP_1: + strcat( str, "1" ); + break; + case PIPE_FORMAT_COMP_Z: + strcat( str, "Z" ); + break; + case PIPE_FORMAT_COMP_S: + strcat( str, "S" ); + break; + } + sprintf( &str[strlen( str )], "%u", size * scale ); + } + if (i != 0) { + strcat( str, "_" ); + } + switch (pf_type( rgbazs )) { + case PIPE_FORMAT_TYPE_UNKNOWN: + strcat( str, "NONE" ); + break; + case PIPE_FORMAT_TYPE_FLOAT: + strcat( str, "FLOAT" ); + break; + case PIPE_FORMAT_TYPE_UNORM: + strcat( str, "UNORM" ); + break; + case PIPE_FORMAT_TYPE_SNORM: + strcat( str, "SNORM" ); + break; + case PIPE_FORMAT_TYPE_USCALED: + strcat( str, "USCALED" ); + break; + case PIPE_FORMAT_TYPE_SSCALED: + strcat( str, "SSCALED" ); + break; + } + } + break; + case PIPE_FORMAT_LAYOUT_YCBCR: { + pipe_format_ycbcr_t ycbcr = (pipe_format_ycbcr_t) format; + + strcat( str, "YCBCR" ); + if (pf_rev( ycbcr )) { + strcat( str, "_REV" ); + } + } + break; + } + return str; +} #endif diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index c6b5bc968f..6d056dbe83 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -109,12 +109,14 @@ st_get_format_info( info = format; #if 0 - printf( - "PIPE_FORMAT: X(%u), Y(%u), Z(%u), W(%u)\n", - info.sizeX, - info.sizeY, - info.sizeZ, - info.sizeW ); + { + char fmtname[256]; + + pf_sprint_name( fmtname, format ); + printf( + "%s\n", + fmtname ); + } #endif /* Data type */ -- cgit v1.2.3 From 11a80160fd60d1eb1541b49128c659526a5d8ac8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 28 Nov 2007 19:04:54 +0100 Subject: Move dimensions from struct pipe_region to struct pipe_surface. --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 6 +- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 30 +-- src/mesa/pipe/failover/fo_context.c | 6 +- src/mesa/pipe/i915simple/i915_clear.c | 9 +- src/mesa/pipe/i915simple/i915_regions.c | 172 --------------- src/mesa/pipe/i915simple/i915_state_emit.c | 25 +-- src/mesa/pipe/i915simple/i915_surface.c | 185 +++++++++++++++- src/mesa/pipe/p_context.h | 48 ++-- src/mesa/pipe/p_state.h | 5 +- src/mesa/pipe/p_winsys.h | 6 +- src/mesa/pipe/softpipe/sp_clear.c | 16 +- src/mesa/pipe/softpipe/sp_region.c | 186 ---------------- src/mesa/pipe/softpipe/sp_surface.c | 244 ++++++++++++++++++--- src/mesa/pipe/xlib/xm_buffer.c | 5 +- src/mesa/pipe/xlib/xm_surface.c | 2 +- src/mesa/pipe/xlib/xm_winsys.c | 22 +- src/mesa/state_tracker/st_cb_drawpixels.c | 50 ++--- src/mesa/state_tracker/st_cb_fbo.c | 8 +- src/mesa/state_tracker/st_cb_readpixels.c | 6 +- src/mesa/state_tracker/st_cb_texture.c | 35 +-- src/mesa/state_tracker/st_mipmap_tree.c | 54 +++-- 21 files changed, 550 insertions(+), 570 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 9fcb2dac27..f96209d1b0 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -89,15 +89,15 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; const int cpp = intelScreen->front.cpp; const struct pipe_region *srcRegion = surf->region; - const int srcpitch = srcRegion->pitch; + const int srcpitch = surf->pitch; int BR13, CMD; int i; ASSERT(srcRegion); - ASSERT(srcRegion->cpp == cpp); + ASSERT(surf->cpp == cpp); DBG(SWAP, "screen pitch %d src surface pitch %d\n", - pitch, srcRegion->pitch); + pitch, surf->pitch); if (cpp == 2) { BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index ae02f98a78..7b3aa99482 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -173,14 +173,10 @@ intel_flush_frontbuffer( struct pipe_winsys *winsys, } -static struct pipe_region * -intel_i915_region_alloc(struct pipe_winsys *winsys, - unsigned cpp, unsigned width, - unsigned height, unsigned flags) +static unsigned +intel_i915_surface_pitch(struct pipe_winsys *winsys, + unsigned cpp, unsigned width, unsigned flags) { - struct pipe_region *region = calloc(sizeof(*region), 1); - const unsigned alignment = 64; - /* Choose a pitch to match hardware requirements - requires 64 byte * alignment of render targets. * @@ -188,24 +184,29 @@ intel_i915_region_alloc(struct pipe_winsys *winsys, * clearly want to be able to render to textures under some * circumstances, but maybe not always a requirement. */ - unsigned pitch; /* XXX is the pitch different for textures vs. drawables? */ if (flags & PIPE_SURFACE_FLAG_TEXTURE) /* or PIPE_SURFACE_FLAG_RENDER? */ - pitch = ((cpp * width + 63) & ~63) / cpp; + return ((cpp * width + 63) & ~63) / cpp; else - pitch = ((cpp * width + 63) & ~63) / cpp; + return ((cpp * width + 63) & ~63) / cpp; +} + + +static struct pipe_region * +intel_i915_region_alloc(struct pipe_winsys *winsys, + unsigned size, unsigned flags) +{ + struct pipe_region *region = calloc(sizeof(*region), 1); + const unsigned alignment = 64; - region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ region->refcount = 1; region->buffer = winsys->buffer_create( winsys, alignment ); winsys->buffer_data( winsys, region->buffer, - pitch * cpp * height, + size, NULL, PIPE_BUFFER_USAGE_PIXEL ); @@ -301,6 +302,7 @@ intel_create_pipe_winsys( int fd ) iws->winsys.get_name = intel_get_name; iws->winsys.region_alloc = intel_i915_region_alloc; iws->winsys.region_release = intel_i915_region_release; + iws->winsys.surface_pitch = intel_i915_surface_pitch; iws->winsys.surface_alloc = intel_i915_surface_alloc; iws->winsys.surface_release = intel_i915_surface_release; diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index c5fab73fb1..0cc9cab408 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -142,9 +142,9 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.region_map = hw->region_map; failover->pipe.region_unmap = hw->region_unmap; - failover->pipe.region_data = hw->region_data; - failover->pipe.region_copy = hw->region_copy; - failover->pipe.region_fill = hw->region_fill; + failover->pipe.surface_data = hw->surface_data; + failover->pipe.surface_copy = hw->surface_copy; + failover->pipe.surface_fill = hw->surface_fill; failover->pipe.mipmap_tree_layout = hw->mipmap_tree_layout; failover->pipe.flush = hw->flush; diff --git a/src/mesa/pipe/i915simple/i915_clear.c b/src/mesa/pipe/i915simple/i915_clear.c index e8087df5a4..cde69daacc 100644 --- a/src/mesa/pipe/i915simple/i915_clear.c +++ b/src/mesa/pipe/i915simple/i915_clear.c @@ -43,12 +43,5 @@ void i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { - int x, y, w, h; - - x = 0; - y = 0; - w = ps->width; - h = ps->height; - - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue); + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); } diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c index e8c4c92bc8..82fdec83d0 100644 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ b/src/mesa/pipe/i915simple/i915_regions.c @@ -34,7 +34,6 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "i915_context.h" -#include "i915_blit.h" @@ -70,181 +69,10 @@ i915_region_unmap(struct pipe_context *pipe, struct pipe_region *region) } -/* - * XXX Move this into core Mesa? - */ -static void -_mesa_copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte * src, - unsigned src_pitch, - unsigned src_x, - unsigned src_y) -{ - unsigned i; - - dst_pitch *= cpp; - src_pitch *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * dst_pitch; - width *= cpp; - - if (width == dst_pitch && width == src_pitch) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; - } - } -} - - -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -static void -i915_region_data(struct pipe_context *pipe, - struct pipe_region *dst, - unsigned dst_offset, - unsigned dstx, unsigned dsty, - const void *src, unsigned src_pitch, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - pipe->region_unmap(pipe, dst); -} - - -/* Assumes all values are within bounds -- no checking at this level - - * do it higher up if required. - */ -static void -i915_region_copy(struct pipe_context *pipe, - struct pipe_region *dst, - unsigned dst_offset, - unsigned dstx, unsigned dsty, - struct pipe_region *src, - unsigned src_offset, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - assert( dst != src ); - assert( dst->cpp == src->cpp ); - - if (0) { - _mesa_copy_rect(pipe->region_map(pipe, dst) + dst_offset, - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - pipe->region_map(pipe, src) + src_offset, - src->pitch, - srcx, srcy); - - pipe->region_unmap(pipe, src); - pipe->region_unmap(pipe, dst); - } - else { - i915_copy_blit( i915_context(pipe), - dst->cpp, - (short) src->pitch, src->buffer, src_offset, - (short) dst->pitch, dst->buffer, dst_offset, - (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); - } -} - -/* Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static ubyte * -get_pointer(struct pipe_region *dst, unsigned x, unsigned y) -{ - return dst->map + (y * dst->pitch + x) * dst->cpp; -} - - -static void -i915_region_fill(struct pipe_context *pipe, - struct pipe_region *dst, - unsigned dst_offset, - unsigned dstx, unsigned dsty, - unsigned width, unsigned height, unsigned value) -{ - if (0) { - unsigned i, j; - - (void)pipe->region_map(pipe, dst); - - switch (dst->cpp) { - case 1: { - ubyte *row = get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; - case 2: { - ushort *row = (ushort *) get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; - case 4: { - unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - break; - default: - assert(0); - break; - } - } - else { - i915_fill_blit( i915_context(pipe), - dst->cpp, - (short) dst->pitch, - dst->buffer, dst_offset, - (short) dstx, (short) dsty, - (short) width, (short) height, - value ); - } -} - - - - - void i915_init_region_functions(struct i915_context *i915) { i915->pipe.region_map = i915_region_map; i915->pipe.region_unmap = i915_region_unmap; - i915->pipe.region_data = i915_region_data; - i915->pipe.region_copy = i915_region_copy; - i915->pipe.region_fill = i915_region_fill; } diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index d793e92a14..900a91d896 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -208,10 +208,11 @@ i915_emit_hardware_state(struct i915_context *i915 ) /* 8 dwords, 2 relocs */ if (i915->hardware_dirty & I915_HW_STATIC) { - if (i915->framebuffer.cbufs[0]) { - struct pipe_region *cbuf_region = i915->framebuffer.cbufs[0]->region; - unsigned pitch = (cbuf_region->pitch * - cbuf_region->cpp); + struct pipe_surface *cbuf_surface = i915->framebuffer.cbufs[0]; + struct pipe_surface *depth_surface = i915->framebuffer.zbuf; + + if (cbuf_surface) { + unsigned pitch = (cbuf_surface->pitch * cbuf_surface->cpp); OUT_BATCH(_3DSTATE_BUF_INFO_CMD); @@ -219,17 +220,15 @@ i915_emit_hardware_state(struct i915_context *i915 ) BUF_3D_PITCH(pitch) | /* pitch in bytes */ BUF_3D_USE_FENCE); - OUT_RELOC(cbuf_region->buffer, + OUT_RELOC(cbuf_surface->region->buffer, I915_BUFFER_ACCESS_WRITE, 0); } /* What happens if no zbuf?? */ - if (i915->framebuffer.zbuf) { - struct pipe_region *depth_region = i915->framebuffer.zbuf->region; - unsigned zpitch = (depth_region->pitch * - depth_region->cpp); + if (depth_surface) { + unsigned zpitch = (depth_surface->pitch * depth_surface->cpp); OUT_BATCH(_3DSTATE_BUF_INFO_CMD); @@ -237,7 +236,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) BUF_3D_PITCH(zpitch) | /* pitch in bytes */ BUF_3D_USE_FENCE); - OUT_RELOC(depth_region->buffer, + OUT_RELOC(depth_surface->region->buffer, I915_BUFFER_ACCESS_WRITE, 0); } @@ -245,13 +244,13 @@ i915_emit_hardware_state(struct i915_context *i915 ) { unsigned cformat, zformat = 0; - if (i915->framebuffer.cbufs[0]) - cformat = i915->framebuffer.cbufs[0]->format; + if (cbuf_surface) + cformat = cbuf_surface->format; else cformat = PIPE_FORMAT_U_A8_R8_G8_B8; /* arbitrary */ cformat = translate_format(cformat); - if (i915->framebuffer.zbuf) + if (depth_surface) zformat = translate_depth_format( i915->framebuffer.zbuf->format ); OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index bec7ddb731..e4a5de00d7 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "i915_context.h" +#include "i915_blit.h" #include "i915_state.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" @@ -57,7 +58,7 @@ i915_get_tile_rgba(struct pipe_context *pipe, { const unsigned *src = ((const unsigned *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -75,7 +76,7 @@ i915_get_tile_rgba(struct pipe_context *pipe, pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); pRow += 4; } - src += ps->region->pitch; + src += ps->pitch; p += w0 * 4; } break; @@ -92,7 +93,7 @@ i915_get_tile_rgba(struct pipe_context *pipe, pRow[3] = (pixel & 0xffffff) * scale; pRow += 4; } - src += ps->region->pitch; + src += ps->pitch; p += w0 * 4; } } @@ -122,7 +123,7 @@ i915_get_tile(struct pipe_context *pipe, uint x, uint y, uint w, uint h, void *p, int dst_stride) { - const uint cpp = ps->region->cpp; + const uint cpp = ps->cpp; const uint w0 = w; const ubyte *pSrc; ubyte *pDest; @@ -136,13 +137,13 @@ i915_get_tile(struct pipe_context *pipe, dst_stride = w0 * cpp; } - pSrc = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; + pSrc = ps->region->map + ps->offset + (y * ps->pitch + x) * cpp; pDest = (ubyte *) p; for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w0 * cpp); pDest += dst_stride; - pSrc += ps->region->pitch * cpp; + pSrc += ps->pitch * cpp; } } @@ -156,7 +157,7 @@ i915_put_tile(struct pipe_context *pipe, uint x, uint y, uint w, uint h, const void *p, int src_stride) { - const uint cpp = ps->region->cpp; + const uint cpp = ps->cpp; const uint w0 = w; const ubyte *pSrc; ubyte *pDest; @@ -171,11 +172,11 @@ i915_put_tile(struct pipe_context *pipe, } pSrc = (const ubyte *) p; - pDest = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; + pDest = ps->region->map + ps->offset + (y * ps->pitch + x) * cpp; for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w0 * cpp); - pDest += ps->region->pitch * cpp; + pDest += ps->pitch * cpp; pSrc += src_stride; } } @@ -210,14 +211,177 @@ i915_get_tex_surface(struct pipe_context *pipe, assert(ps->format); assert(ps->refcount); pipe_region_reference(&ps->region, mt->region); + ps->cpp = mt->cpp; ps->width = mt->level[level].width; ps->height = mt->level[level].height; + ps->pitch = mt->pitch; ps->offset = offset; } return ps; } +/* + * XXX Move this into core Mesa? + */ +static void +_mesa_copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + unsigned src_pitch, + unsigned src_x, + unsigned src_y) +{ + unsigned i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * dst_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } +} + + +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +static void +i915_surface_data(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + const void *src, unsigned src_pitch, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + _mesa_copy_rect(pipe->region_map(pipe, dst->region) + dst->offset, + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + pipe->region_unmap(pipe, dst->region); +} + + +/* Assumes all values are within bounds -- no checking at this level - + * do it higher up if required. + */ +static void +i915_surface_copy(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + assert( dst != src ); + assert( dst->cpp == src->cpp ); + + if (0) { + _mesa_copy_rect(pipe->region_map(pipe, dst->region) + dst->offset, + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe->region_map(pipe, src->region) + src->offset, + src->pitch, + srcx, srcy); + + pipe->region_unmap(pipe, src->region); + pipe->region_unmap(pipe, dst->region); + } + else { + i915_copy_blit( i915_context(pipe), + dst->cpp, + (short) src->pitch, src->region->buffer, src->offset, + (short) dst->pitch, dst->region->buffer, dst->offset, + (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); + } +} + +/* Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +static ubyte * +get_pointer(struct pipe_surface *dst, unsigned x, unsigned y) +{ + return dst->region->map + (y * dst->pitch + x) * dst->cpp; +} + + +static void +i915_surface_fill(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) +{ + if (0) { + unsigned i, j; + + (void)pipe->region_map(pipe, dst->region); + + switch (dst->cpp) { + case 1: { + ubyte *row = get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: { + ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (ushort) value; + row += dst->pitch; + } + } + break; + case 4: { + unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + default: + assert(0); + break; + } + } + else { + i915_fill_blit( i915_context(pipe), + dst->cpp, + (short) dst->pitch, + dst->region->buffer, dst->offset, + (short) dstx, (short) dsty, + (short) width, (short) height, + value ); + } +} + + void i915_init_surface_functions(struct i915_context *i915) { @@ -226,4 +390,7 @@ i915_init_surface_functions(struct i915_context *i915) i915->pipe.put_tile = i915_put_tile; i915->pipe.get_tile_rgba = i915_get_tile_rgba; i915->pipe.put_tile_rgba = i915_put_tile_rgba; + i915->pipe.surface_data = i915_surface_data; + i915->pipe.surface_copy = i915_surface_copy; + i915->pipe.surface_fill = i915_surface_fill; } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 8bed958feb..e145b22f2f 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -208,30 +208,30 @@ struct pipe_context { void (*region_unmap)(struct pipe_context *pipe, struct pipe_region *r); - void (*region_data)(struct pipe_context *pipe, - struct pipe_region *dest, - unsigned dest_offset, - unsigned destx, unsigned desty, - const void *src, unsigned src_stride, - unsigned srcx, unsigned srcy, - unsigned width, unsigned height); - - void (*region_copy)(struct pipe_context *pipe, - struct pipe_region *dest, - unsigned dest_offset, - unsigned destx, unsigned desty, - struct pipe_region *src, /* don't make this const - - need to map/unmap */ - unsigned src_offset, - unsigned srcx, unsigned srcy, - unsigned width, unsigned height); - - void (*region_fill)(struct pipe_context *pipe, - struct pipe_region *dst, - unsigned dst_offset, - unsigned dstx, unsigned dsty, - unsigned width, unsigned height, - unsigned value); + + /* + * Surface functions + */ + void (*surface_data)(struct pipe_context *pipe, + struct pipe_surface *dest, + unsigned destx, unsigned desty, + const void *src, unsigned src_stride, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height); + + void (*surface_copy)(struct pipe_context *pipe, + struct pipe_surface *dest, + unsigned destx, unsigned desty, + struct pipe_surface *src, /* don't make this const - + need to map/unmap */ + unsigned srcx, unsigned srcy, + unsigned width, unsigned height); + + void (*surface_fill)(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, + unsigned value); /* diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 570f44e24e..642734aeb8 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -265,9 +265,6 @@ struct pipe_region struct pipe_buffer_handle *buffer; /**< driver private buffer handle */ unsigned refcount; /**< Reference count for region */ - unsigned cpp; /**< bytes per pixel */ - unsigned pitch; /**< in pixels */ - unsigned height; /**< in pixels */ ubyte *map; /**< only non-NULL when region is actually mapped */ unsigned map_refcount; /**< Reference count for mapping */ }; @@ -281,7 +278,9 @@ struct pipe_surface { struct pipe_region *region; unsigned format; /**< PIPE_FORMAT_x */ + unsigned cpp; /**< bytes per pixel */ unsigned width, height; + unsigned pitch; /**< in pixels */ unsigned offset; /**< offset from start of region, in bytes */ unsigned refcount; struct pipe_winsys *winsys; /**< winsys which owns/created the surface */ diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 41d522e11e..5adca1d6fd 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -80,12 +80,14 @@ struct pipe_winsys * flags is bitmask of PIPE_SURFACE_FLAG_RENDER, PIPE_SURFACE_FLAG_TEXTURE */ struct pipe_region *(*region_alloc)(struct pipe_winsys *ws, - unsigned cpp, unsigned width, - unsigned height, unsigned flags); + unsigned size, unsigned flags); void (*region_release)(struct pipe_winsys *ws, struct pipe_region **r); + unsigned (*surface_pitch)(struct pipe_winsys *ws, unsigned cpp, + unsigned with, unsigned flags); + /** allocate a new surface (no context dependency) */ struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws, unsigned format); diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 87f850b6fd..a4276362b9 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -47,23 +47,9 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); - unsigned x, y, w, h; softpipe_update_derived(softpipe); /* not needed?? */ - /* Use the X coord to trick region_fill() into filling at an offset - * from the start of the region. Perhaps pipe_region should have the - * 'offset' field, not pipe_surface??? - */ - assert(ps->offset % ps->region->cpp == 0); - x = ps->offset / ps->region->cpp; - y = 0; - w = ps->width; - h = ps->height; - - assert(w <= ps->region->pitch); - assert(h <= ps->region->height); - if (ps == sp_tile_cache_get_surface(softpipe->zbuf_cache)) { float clear[4]; clear[0] = 1.0; /* XXX hack */ @@ -78,7 +64,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, sp_tile_cache_clear(softpipe->cbuf_cache[0], clear); } - pipe->region_fill(pipe, ps->region, 0, x, y, w, h, clearValue); + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); #if 0 diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c index fef63ef2f6..58dc6bb96e 100644 --- a/src/mesa/pipe/softpipe/sp_region.c +++ b/src/mesa/pipe/softpipe/sp_region.c @@ -70,196 +70,10 @@ sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) } - -/** - * Copy 2D rect from one place to another. - * Position and sizes are in pixels. - */ -static void -copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte * src, - unsigned src_pitch, - unsigned src_x, - unsigned src_y) -{ - unsigned i; - - dst_pitch *= cpp; - src_pitch *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * src_pitch; - width *= cpp; - - if (width == dst_pitch && width == src_pitch) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; - } - } -} - - -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -static void -sp_region_data(struct pipe_context *pipe, - struct pipe_region *dst, - unsigned dst_offset, - unsigned dstx, unsigned dsty, - const void *src, unsigned src_pitch, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - copy_rect(pipe->region_map(pipe, dst) + dst_offset, - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - pipe->region_unmap(pipe, dst); -} - -/* Assumes all values are within bounds -- no checking at this level - - * do it higher up if required. - */ -static void -sp_region_copy(struct pipe_context *pipe, - struct pipe_region *dst, - unsigned dst_offset, - unsigned dstx, unsigned dsty, - struct pipe_region *src, - unsigned src_offset, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - ubyte *src_map, *dst_map; - assert( dst->cpp == src->cpp ); - - dst_map = pipe->region_map(pipe, dst); - src_map = pipe->region_map(pipe, src); - copy_rect(dst_map + dst_offset, - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - src_map + src_offset, - src->pitch, - srcx, srcy); - - pipe->region_unmap(pipe, src); - pipe->region_unmap(pipe, dst); -} - - -static ubyte * -get_pointer(struct pipe_region *dst, unsigned x, unsigned y) -{ - return dst->map + (y * dst->pitch + x) * dst->cpp; -} - - -#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8)) - - -/** - * Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static void -sp_region_fill(struct pipe_context *pipe, - struct pipe_region *dst, - unsigned dst_offset, - unsigned dstx, unsigned dsty, - unsigned width, unsigned height, unsigned value) -{ - unsigned i, j; - - assert(dst->pitch > 0); - assert(width <= dst->pitch); - - (void)pipe->region_map(pipe, dst); - - switch (dst->cpp) { - case 1: - { - ubyte *row = get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; - case 2: - { - ushort *row = (ushort *) get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; - case 4: - { - unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - break; - case 8: - { - /* expand the 4-byte clear value to an 8-byte value */ - ushort *row = (ushort *) get_pointer(dst, dstx, dsty); - ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); - ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); - ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); - ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) { - row[j*4+0] = val0; - row[j*4+1] = val1; - row[j*4+2] = val2; - row[j*4+3] = val3; - } - row += dst->pitch * 4; - } - } - break; - default: - assert(0); - break; - } - - pipe->region_unmap( pipe, dst ); -} - - - - - void sp_init_region_functions(struct softpipe_context *sp) { sp->pipe.region_map = sp_region_map; sp->pipe.region_unmap = sp_region_unmap; - sp->pipe.region_data = sp_region_data; - sp->pipe.region_copy = sp_region_copy; - sp->pipe.region_fill = sp_region_fill; } diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index b7c9d4f004..2ddf3ab99c 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -77,7 +77,7 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, { const unsigned *src = ((const unsigned *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -95,7 +95,7 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); pRow += 4; } - src += ps->region->pitch; + src += ps->pitch; p += w0 * 4; } } @@ -108,7 +108,7 @@ a8r8g8b8_put_tile(struct pipe_surface *ps, { unsigned *dst = ((unsigned *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -127,7 +127,7 @@ a8r8g8b8_put_tile(struct pipe_surface *ps, dst[j] = (a << 24) | (r << 16) | (g << 8) | b; pRow += 4; } - dst += ps->region->pitch; + dst += ps->pitch; p += w0 * 4; } } @@ -141,7 +141,7 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, { const ushort *src = ((const ushort *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; unsigned i, j; assert(ps->format == PIPE_FORMAT_U_A1_R5_G5_B5); @@ -155,7 +155,7 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, p[3] = ((pixel >> 15) ) * 1.0f; p += 4; } - src += ps->region->pitch; + src += ps->pitch; } } @@ -172,7 +172,7 @@ z16_get_tile(struct pipe_surface *ps, { const ushort *src = ((const ushort *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; const float scale = 1.0f / 65535.0f; unsigned i, j; unsigned w0 = w; @@ -189,7 +189,7 @@ z16_get_tile(struct pipe_surface *ps, pRow[j * 4 + 2] = pRow[j * 4 + 3] = src[j] * scale; } - src += ps->region->pitch; + src += ps->pitch; p += 4 * w0; } } @@ -205,7 +205,7 @@ l8_get_tile(struct pipe_surface *ps, { const ubyte *src = ((const ubyte *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -222,7 +222,7 @@ l8_get_tile(struct pipe_surface *ps, pRow[3] = 1.0; pRow += 4; } - src += ps->region->pitch; + src += ps->pitch; p += w0 * 4; } } @@ -236,7 +236,7 @@ a8_get_tile(struct pipe_surface *ps, { const ubyte *src = ((const ubyte *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -253,7 +253,7 @@ a8_get_tile(struct pipe_surface *ps, pRow[3] = UBYTE_TO_FLOAT(src[j]); pRow += 4; } - src += ps->region->pitch; + src += ps->pitch; p += w0 * 4; } } @@ -267,7 +267,7 @@ r16g16b16a16_get_tile(struct pipe_surface *ps, { const short *src = ((const short *) (ps->region->map + ps->offset)) - + (y * ps->region->pitch + x) * 4; + + (y * ps->pitch + x) * 4; unsigned i, j; unsigned w0 = w; @@ -286,7 +286,7 @@ r16g16b16a16_get_tile(struct pipe_surface *ps, pRow += 4; pixel += 4; } - src += ps->region->pitch * 4; + src += ps->pitch * 4; p += w0 * 4; } } @@ -299,7 +299,7 @@ r16g16b16a16_put_tile(struct pipe_surface *ps, { short *dst = ((short *) (ps->region->map + ps->offset)) - + (y * ps->region->pitch + x) * 4; + + (y * ps->pitch + x) * 4; unsigned i, j; unsigned w0 = w; @@ -321,7 +321,7 @@ r16g16b16a16_put_tile(struct pipe_surface *ps, dst[j*4+3] = a; pRow += 4; } - dst += ps->region->pitch * 4; + dst += ps->pitch * 4; p += w0 * 4; } } @@ -336,7 +336,7 @@ i8_get_tile(struct pipe_surface *ps, { const ubyte *src = ((const ubyte *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -353,7 +353,7 @@ i8_get_tile(struct pipe_surface *ps, pRow[3] = UBYTE_TO_FLOAT(src[j]); pRow += 4; } - src += ps->region->pitch; + src += ps->pitch; p += w0 * 4; } } @@ -367,7 +367,7 @@ a8_l8_get_tile(struct pipe_surface *ps, { const ushort *src = ((const ushort *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -385,7 +385,7 @@ a8_l8_get_tile(struct pipe_surface *ps, pRow[3] = UBYTE_TO_FLOAT(p >> 8); pRow += 4; } - src += ps->region->pitch; + src += ps->pitch; p += w0 * 4; } } @@ -404,7 +404,7 @@ z32_get_tile(struct pipe_surface *ps, { const uint *src = ((const uint *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; const double scale = 1.0 / (double) 0xffffffff; unsigned i, j; unsigned w0 = w; @@ -421,7 +421,7 @@ z32_get_tile(struct pipe_surface *ps, pRow[j * 4 + 2] = pRow[j * 4 + 3] = (float) (scale * src[j]); } - src += ps->region->pitch; + src += ps->pitch; p += 4 * w0; } } @@ -438,7 +438,7 @@ s8z24_get_tile(struct pipe_surface *ps, { const uint *src = ((const uint *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; const double scale = 1.0 / ((1 << 24) - 1); unsigned i, j; unsigned w0 = w; @@ -455,7 +455,7 @@ s8z24_get_tile(struct pipe_surface *ps, pRow[j * 4 + 2] = pRow[j * 4 + 3] = (float) (scale * (src[j] & 0xffffff)); } - src += ps->region->pitch; + src += ps->pitch; p += 4 * w0; } } @@ -472,7 +472,7 @@ z24s8_get_tile(struct pipe_surface *ps, { const uint *src = ((const uint *) (ps->region->map + ps->offset)) - + y * ps->region->pitch + x; + + y * ps->pitch + x; const double scale = 1.0 / ((1 << 24) - 1); unsigned i, j; unsigned w0 = w; @@ -489,7 +489,7 @@ z24s8_get_tile(struct pipe_surface *ps, pRow[j * 4 + 2] = pRow[j * 4 + 3] = (float) (scale * (src[j] >> 8)); } - src += ps->region->pitch; + src += ps->pitch; p += 4 * w0; } } @@ -525,8 +525,10 @@ softpipe_get_tex_surface(struct pipe_context *pipe, assert(ps->format); assert(ps->refcount); pipe_region_reference(&ps->region, mt->region); + ps->cpp = mt->cpp; ps->width = mt->level[level].width; ps->height = mt->level[level].height; + ps->pitch = mt->pitch; ps->offset = offset; } return ps; @@ -541,7 +543,7 @@ softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, void *p, int dst_stride) { - const uint cpp = ps->region->cpp; + const uint cpp = ps->cpp; const ubyte *pSrc; ubyte *pDest; uint i; @@ -554,13 +556,13 @@ softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, CLIP_TILE; - pSrc = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; + pSrc = ps->region->map + ps->offset + (y * ps->pitch + x) * cpp; pDest = (ubyte *) p; for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w * cpp); pDest += dst_stride; - pSrc += ps->region->pitch * cpp; + pSrc += ps->pitch * cpp; } } @@ -573,7 +575,7 @@ softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const void *p, int src_stride) { - const uint cpp = ps->region->cpp; + const uint cpp = ps->cpp; const ubyte *pSrc; ubyte *pDest; uint i; @@ -587,11 +589,11 @@ softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, CLIP_TILE; pSrc = (const ubyte *) p; - pDest = ps->region->map + ps->offset + (y * ps->region->pitch + x) * cpp; + pDest = ps->region->map + ps->offset + (y * ps->pitch + x) * cpp; for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w * cpp); - pDest += ps->region->pitch * cpp; + pDest += ps->pitch * cpp; pSrc += src_stride; } } @@ -691,6 +693,180 @@ softpipe_put_tile_rgba(struct pipe_context *pipe, } +/** + * Copy 2D rect from one place to another. + * Position and sizes are in pixels. + */ +static void +copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + unsigned src_pitch, + unsigned src_x, + unsigned src_y) +{ + unsigned i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * src_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } +} + + +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +static void +sp_surface_data(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + const void *src, unsigned src_pitch, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + copy_rect(pipe->region_map(pipe, dst->region) + dst->offset, + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + pipe->region_unmap(pipe, dst->region); +} + +/* Assumes all values are within bounds -- no checking at this level - + * do it higher up if required. + */ +static void +sp_surface_copy(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + ubyte *src_map, *dst_map; + assert( dst->cpp == src->cpp ); + + dst_map = pipe->region_map(pipe, dst->region); + src_map = pipe->region_map(pipe, src->region); + copy_rect(dst_map + dst->offset, + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + src_map + src->offset, + src->pitch, + srcx, srcy); + + pipe->region_unmap(pipe, src->region); + pipe->region_unmap(pipe, dst->region); +} + + +static ubyte * +get_pointer(struct pipe_surface *dst, unsigned x, unsigned y) +{ + return dst->region->map + (y * dst->pitch + x) * dst->cpp; +} + + +#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8)) + + +/** + * Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +static void +sp_surface_fill(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) +{ + unsigned i, j; + + assert(dst->pitch > 0); + assert(width <= dst->pitch); + + (void)pipe->region_map(pipe, dst->region); + + switch (dst->cpp) { + case 1: + { + ubyte *row = get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: + { + ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (ushort) value; + row += dst->pitch; + } + } + break; + case 4: + { + unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + case 8: + { + /* expand the 4-byte clear value to an 8-byte value */ + ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); + ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); + ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); + ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) { + row[j*4+0] = val0; + row[j*4+1] = val1; + row[j*4+2] = val2; + row[j*4+3] = val3; + } + row += dst->pitch * 4; + } + } + break; + default: + assert(0); + break; + } + + pipe->region_unmap( pipe, dst->region ); +} + void sp_init_surface_functions(struct softpipe_context *sp) @@ -700,4 +876,8 @@ sp_init_surface_functions(struct softpipe_context *sp) sp->pipe.get_tile_rgba = softpipe_get_tile_rgba; sp->pipe.put_tile_rgba = softpipe_put_tile_rgba; + + sp->pipe.surface_data = sp_surface_data; + sp->pipe.surface_copy = sp_surface_copy; + sp->pipe.surface_fill = sp_surface_fill; } diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index 7dc85bf2eb..71a2003e6d 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -260,9 +260,8 @@ finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) { struct pipe_context *pipe = ctx->st->pipe; if (!xrb->St.surface->region) { - int w = 1, h = 1; - xrb->St.surface->region = pipe->winsys->region_alloc(pipe->winsys, - 1, w, h, 0x0); + xrb->St.surface->region = pipe->winsys->region_alloc(pipe->winsys, 1, + 0x0); } } diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 43e5050747..2394563095 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -624,7 +624,7 @@ xmesa_new_color_surface(struct pipe_winsys *winsys, GLuint pipeFormat) * The region's size will get set in the xmesa_alloc_front/back_storage() * functions. */ - xms->surface.region = winsys->region_alloc(winsys, 1, 1, 1, 0x0); + xms->surface.region = winsys->region_alloc(winsys, 1, 0x0); return &xms->surface; } diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index b842cf766e..99816a811d 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -220,27 +220,30 @@ round_up(unsigned n, unsigned multiple) } +static unsigned +xm_surface_pitch(struct pipe_winsys *winsys, unsigned cpp, unsigned width, + unsigned flags) +{ + return round_up(width, 64 / cpp); +} + + static struct pipe_region * -xm_region_alloc(struct pipe_winsys *winsys, - unsigned cpp, unsigned width, unsigned height, unsigned flags) +xm_region_alloc(struct pipe_winsys *winsys, unsigned size, unsigned flags) { struct pipe_region *region = CALLOC_STRUCT(pipe_region); const unsigned alignment = 64; - region->cpp = cpp; - region->pitch = round_up(width, alignment / cpp); - region->height = height; region->refcount = 1; - assert(region->pitch > 0); + assert(size > 0); - region->buffer = winsys->buffer_create( winsys, alignment ) -; + region->buffer = winsys->buffer_create( winsys, alignment ); /* NULL data --> just allocate the space */ winsys->buffer_data( winsys, region->buffer, - region->pitch * cpp * height, + size, NULL, PIPE_BUFFER_USAGE_PIXEL ); return region; @@ -335,6 +338,7 @@ xmesa_get_pipe_winsys(void) ws->region_alloc = xm_region_alloc; ws->region_release = xm_region_release; + ws->surface_pitch = xm_surface_pitch; ws->surface_alloc = xm_surface_alloc; ws->surface_release = xm_surface_release; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index df41d5ce5b..39c0cf6b32 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -455,10 +455,6 @@ alloc_mipmap_tree(struct st_context *st, cpp = st_sizeof_format(pipeFormat); - /* allocate texture region/storage */ - mt->region = st->pipe->winsys->region_alloc(st->pipe->winsys, - cpp, width, height, flags); - mt->target = PIPE_TEXTURE_2D; mt->internal_format = GL_RGBA; mt->format = pipeFormat; @@ -469,7 +465,10 @@ alloc_mipmap_tree(struct st_context *st, mt->depth0 = 1; mt->cpp = cpp; mt->compressed = 0; - mt->pitch = mt->region->pitch; + mt->pitch = st->pipe->winsys->surface_pitch(st->pipe->winsys, cpp, width, + flags); + mt->region = st->pipe->winsys->region_alloc(st->pipe->winsys, + mt->pitch * cpp * height, flags); mt->depth_pitch = 0; mt->total_height = height; mt->level[0].level_offset = 0; @@ -524,7 +523,7 @@ make_mipmap_tree(struct st_context *st, { static const GLuint dstImageOffsets = 0; GLboolean success; - GLuint pitch = mt->region->pitch; + GLuint pitch = mt->pitch; GLubyte *dest; const GLbitfield imageTransferStateSave = ctx->_ImageTransferState; @@ -569,7 +568,7 @@ make_mipmap_tree(struct st_context *st, mt->depth0 = 1; mt->cpp = cpp; mt->compressed = 0; - mt->pitch = mt->region->pitch; + mt->pitch = mt->pitch; mt->depth_pitch = 0; mt->total_height = height; mt->level[0].level_offset = 0; @@ -952,13 +951,13 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, switch (ps->format) { case PIPE_FORMAT_U_S8: { - ubyte *dest = stmap + spanY * ps->region->pitch + spanX; + ubyte *dest = stmap + spanY * ps->pitch + spanX; memcpy(dest, values, spanWidth); } break; case PIPE_FORMAT_S8_Z24: { - uint *dest = (uint *) stmap + spanY * ps->region->pitch + spanX; + uint *dest = (uint *) stmap + spanY * ps->pitch + spanX; GLint k; for (k = 0; k < spanWidth; k++) { uint p = dest[k]; @@ -1054,7 +1053,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, { struct pipe_context *pipe = ctx->st->pipe; const uint flags = PIPE_SURFACE_FLAG_TEXTURE; - uint format = 0, cpp, comp, pitch; + uint format = 0, cpp, comp; ubyte *dest; struct pipe_mipmap_tree *mt; int row, col; @@ -1091,9 +1090,9 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, /* allocate texture region/storage */ + mt->pitch = pipe->winsys->surface_pitch(pipe->winsys, cpp, width, flags); mt->region = pipe->winsys->region_alloc(pipe->winsys, - cpp, width, height, flags); - pitch = mt->region->pitch; + mt->pitch * cpp * height, flags); /* map texture region */ dest = pipe->region_map(pipe, mt->region); @@ -1110,7 +1109,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, for (row = 0; row < height; row++) { const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); - ubyte *destRow = dest + row * pitch * cpp; + ubyte *destRow = dest + row * mt->pitch * cpp; if (unpack->LsbFirst) { /* Lsb first */ @@ -1172,7 +1171,6 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, mt->depth0 = 1; mt->cpp = cpp; mt->compressed = 0; - mt->pitch = mt->region->pitch; mt->depth_pitch = 0; mt->total_height = height; mt->level[0].level_offset = 0; @@ -1256,7 +1254,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, y = ctx->DrawBuffer->Height - y - 1; } - dst = drawMap + (y * psDraw->region->pitch + dstx) * psDraw->region->cpp; + dst = drawMap + (y * psDraw->pitch + dstx) * psDraw->cpp; src = buffer + i * width; switch (psDraw->format) { @@ -1297,6 +1295,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, struct st_vertex_program *stvp; struct st_fragment_program *stfp; struct pipe_surface *psRead; + struct pipe_surface *psTex; struct pipe_mipmap_tree *mt; GLfloat *color; uint format; @@ -1332,6 +1331,8 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, if (!mt) return; + psTex = pipe->get_tex_surface(pipe, mt, 0, 0, 0); + if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { srcy = ctx->DrawBuffer->Height - srcy - height; } @@ -1342,21 +1343,16 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, */ if (st->haveFramebufferRegions) { /* copy source framebuffer region into mipmap/texture */ - pipe->region_copy(pipe, - mt->region, /* dest */ - 0, /* dest_offset */ - 0, 0, /* destx/y */ - psRead->region, - 0, /* src_offset */ - srcx, srcy, width, height); + pipe->surface_copy(pipe, + psTex, /* dest */ + 0, 0, /* destx/y */ + psRead, + srcx, srcy, width, height); } else { /* alternate path using get/put_tile() */ - struct pipe_surface *psTex; GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - psTex = pipe->get_tex_surface(pipe, mt, 0, 0, 0); - (void) pipe->region_map(pipe, psRead->region); (void) pipe->region_map(pipe, psTex->region); @@ -1366,17 +1362,15 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, pipe->region_unmap(pipe, psRead->region); pipe->region_unmap(pipe, psTex->region); - pipe_surface_reference(&psTex, NULL); - free(buf); } - /* draw textured quad */ draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, mt, stvp, stfp, color, GL_TRUE); + pipe_surface_reference(&psTex, NULL); free_mipmap_tree(st->pipe, mt); } diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index f58d532b2a..80c92e8b7a 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -104,6 +104,9 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, assert(strb->surface); if (!strb->surface) return GL_FALSE; + strb->surface->cpp = cpp; + strb->surface->pitch = pipe->winsys->surface_pitch(pipe->winsys, cpp, + width, flags); } /* free old region */ @@ -115,8 +118,9 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe->winsys->region_release(pipe->winsys, &strb->surface->region); } - strb->surface->region = pipe->winsys->region_alloc(pipe->winsys, cpp, - width, height, flags); + strb->surface->region = pipe->winsys->region_alloc(pipe->winsys, + strb->surface->pitch * + cpp * height, flags); if (!strb->surface->region) return GL_FALSE; /* out of memory, try s/w buffer? */ diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 2e7c01672a..83fe480af8 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -87,13 +87,13 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, switch (ps->format) { case PIPE_FORMAT_U_S8: { - const ubyte *src = stmap + srcY * ps->region->pitch + x; + const ubyte *src = stmap + srcY * ps->pitch + x; memcpy(values, src, width); } break; case PIPE_FORMAT_S8_Z24: { - const uint *src = (uint *) stmap + srcY * ps->region->pitch + x; + const uint *src = (uint *) stmap + srcY * ps->pitch + x; GLint k; for (k = 0; k < width; k++) { values[k] = src[k] >> 24; @@ -102,7 +102,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, break; case PIPE_FORMAT_Z24_S8: { - const uint *src = (uint *) stmap + srcY * ps->region->pitch + x; + const uint *src = (uint *) stmap + srcY * ps->pitch + x; GLint k; for (k = 0; k < width; k++) { values[k] = src[k] & 0xff; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 4e6d4857f4..461705119f 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -44,6 +44,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" #define DBG if (0) printf @@ -1150,7 +1151,6 @@ do_copy_texsubimage(GLcontext *ctx, struct st_renderbuffer *strb; struct pipe_context *pipe = ctx->st->pipe; struct pipe_region *src_region, *dest_region; - uint dest_offset, src_offset; uint dest_format, src_format; (void) texImage; @@ -1185,12 +1185,13 @@ do_copy_texsubimage(GLcontext *ctx, ctx->_ImageTransferState == 0x0 && src_region && dest_region && - src_region->cpp == dest_region->cpp) { + strb->surface->cpp == stImage->mt->cpp) { /* do blit-style copy */ - src_offset = 0; - dest_offset = st_miptree_image_offset(stImage->mt, - stImage->face, - stImage->level); + struct pipe_surface *dest_surface = pipe->get_tex_surface(pipe, + stImage->mt, + stImage->face, + stImage->level, + destZ); /* XXX may need to invert image depending on window * vs. user-created FBO @@ -1213,18 +1214,18 @@ do_copy_texsubimage(GLcontext *ctx, GL_COPY); /* ? */ #else - pipe->region_copy(pipe, - /* dest */ - dest_region, - dest_offset, - destX, destY, - /* src */ - src_region, - src_offset, - srcX, srcY, - /* size */ - width, height); + pipe->surface_copy(pipe, + /* dest */ + dest_surface, + destX, destY, + /* src */ + strb->surface, + srcX, srcY, + /* size */ + width, height); #endif + + pipe_surface_reference(&dest_surface, NULL); } else { fallback_copy_texsubimage(ctx, target, level, diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c index d1db590bee..6ccf33105f 100644 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ b/src/mesa/state_tracker/st_mipmap_tree.c @@ -31,6 +31,8 @@ #include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/p_winsys.h" @@ -87,10 +89,9 @@ st_miptree_create(struct pipe_context *pipe, ok = pipe->mipmap_tree_layout(pipe, mt); if (ok) { - /* note: it's OK to pass 'pitch' as 'width' here: */ - mt->region = pipe->winsys->region_alloc(pipe->winsys, mt->cpp, mt->pitch, - mt->total_height, flags); - mt->pitch = mt->region->pitch; /*XXX NEW */ + mt->region = pipe->winsys->region_alloc(pipe->winsys, + mt->pitch * mt->cpp * + mt->total_height, flags); } if (!mt->region) { @@ -266,24 +267,27 @@ st_miptree_image_data(struct pipe_context *pipe, GLuint src_row_pitch, GLuint src_image_pitch) { GLuint depth = dst->level[level].depth; - GLuint dst_offset = st_miptree_image_offset(dst, face, level); - const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); GLuint i; GLuint height = 0; const GLubyte *srcUB = src; + struct pipe_surface *dst_surface; DBG("%s\n", __FUNCTION__); for (i = 0; i < depth; i++) { height = dst->level[level].height; if(dst->compressed) height /= 4; - pipe->region_data(pipe, dst->region, - dst_offset + dst_depth_offset[i], /* dst_offset */ - 0, 0, /* dstx, dsty */ - srcUB, - src_row_pitch, - 0, 0, /* source x, y */ - dst->level[level].width, height); /* width, height */ + + dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); + + pipe->surface_data(pipe, dst_surface, + 0, 0, /* dstx, dsty */ + srcUB, + src_row_pitch, + 0, 0, /* source x, y */ + dst->level[level].width, height); /* width, height */ + + pipe_surface_reference(&dst_surface, NULL); srcUB += src_image_pitch * dst->cpp; } @@ -300,21 +304,25 @@ st_miptree_image_copy(struct pipe_context *pipe, GLuint width = src->level[level].width; GLuint height = src->level[level].height; GLuint depth = src->level[level].depth; - GLuint dst_offset = st_miptree_image_offset(dst, face, level); - GLuint src_offset = st_miptree_image_offset(src, face, level); - const GLuint *dst_depth_offset = st_miptree_depth_offsets(dst, level); - const GLuint *src_depth_offset = st_miptree_depth_offsets(src, level); + struct pipe_surface *src_surface; + struct pipe_surface *dst_surface; GLuint i; if (dst->compressed) height /= 4; for (i = 0; i < depth; i++) { - pipe->region_copy(pipe, - dst->region, dst_offset + dst_depth_offset[i], - 0, 0, /* destX, Y */ - src->region, src_offset + src_depth_offset[i], - 0, 0, /* srcX, Y */ - width, height); + dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); + src_surface = pipe->get_tex_surface(pipe, src, face, level, i); + + pipe->surface_copy(pipe, + dst_surface, + 0, 0, /* destX, Y */ + src_surface, + 0, 0, /* srcX, Y */ + width, height); + + pipe_surface_reference(&dst_surface, NULL); + pipe_surface_reference(&src_surface, NULL); } } -- cgit v1.2.3 From 88b715b049efa81f4021d642a963b58113a875cf Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 2 Dec 2007 15:23:51 -0700 Subject: Initial Cell driver infrastructure. No real code yet. Just stand-ins and make/build infrastructure. --- src/mesa/Makefile | 9 ++++- src/mesa/pipe/Makefile | 26 ++++++++++-- src/mesa/pipe/cell/Makefile | 12 ++++++ src/mesa/pipe/cell/common.h | 13 ++++++ src/mesa/pipe/cell/ppu/Makefile | 39 ++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_context.c | 76 +++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_context.h | 46 +++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_surface.c | 9 +++++ src/mesa/pipe/cell/spu/Makefile | 42 +++++++++++++++++++ src/mesa/pipe/cell/spu/main.c | 29 +++++++++++++ src/mesa/pipe/cell/spu/tri.c | 9 +++++ src/mesa/pipe/cell/spu/tri.h | 4 ++ 12 files changed, 308 insertions(+), 6 deletions(-) create mode 100644 src/mesa/pipe/cell/Makefile create mode 100644 src/mesa/pipe/cell/common.h create mode 100644 src/mesa/pipe/cell/ppu/Makefile create mode 100644 src/mesa/pipe/cell/ppu/cell_context.c create mode 100644 src/mesa/pipe/cell/ppu/cell_context.h create mode 100644 src/mesa/pipe/cell/ppu/cell_surface.c create mode 100644 src/mesa/pipe/cell/spu/Makefile create mode 100644 src/mesa/pipe/cell/spu/main.c create mode 100644 src/mesa/pipe/cell/spu/tri.c create mode 100644 src/mesa/pipe/cell/spu/tri.h (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 521cbeb606..6f0877d335 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -13,6 +13,11 @@ GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) SOFTPIPE_LIB = $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a +ifeq ($(CONFIG_NAME), linux-cell) +CELL_LIB = $(TOP)/src/mesa/pipe/cell/ppu/libcell.a +endif + + .SUFFIXES : .cpp .c.o: @@ -112,12 +117,12 @@ stand-alone: depend subdirs $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) $(TOP)/$(LIB_DIR)/$ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) $(CELL_LIB) @ $(TOP)/bin/mklib -o $(GL_LIB) \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ - $(SOFTPIPE_LIB) $(GL_LIB_DEPS) + $(SOFTPIPE_LIB) $(CELL_LIB)$(GL_LIB_DEPS) # Make the OSMesa library $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECTS) diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile index d448f3f5a6..a9dc393417 100644 --- a/src/mesa/pipe/Makefile +++ b/src/mesa/pipe/Makefile @@ -1,7 +1,25 @@ -default: - cd softpipe ; make - cd i915simple ; make - cd failover ; make + +TOP = ../../.. +include $(TOP)/configs/current + + +ifeq ($(CONFIG_NAME), linux-cell) +CELL_DIR = cell +endif + +SUBDIRS = softpipe i915simple failover $(CELL_DIR) + + +default: subdirs + + +subdirs: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + clean: rm -f `find . -name \*.[oa]` \ No newline at end of file diff --git a/src/mesa/pipe/cell/Makefile b/src/mesa/pipe/cell/Makefile new file mode 100644 index 0000000000..47aef7b05f --- /dev/null +++ b/src/mesa/pipe/cell/Makefile @@ -0,0 +1,12 @@ +# Cell Gallium driver Makefile + + +default: + ( cd spu ; make ) + ( cd ppu ; make ) + + + +clean: + ( cd spu ; make clean ) + ( cd ppu ; make clean ) diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h new file mode 100644 index 0000000000..c4bad4194a --- /dev/null +++ b/src/mesa/pipe/cell/common.h @@ -0,0 +1,13 @@ + +#ifndef CELL_COMMON_H +#define CELL_COMMON_H + + +struct init_info +{ + int foo; + int bar; +}; + + +#endif /* CELL_COMMON_H */ diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile new file mode 100644 index 0000000000..ede341abca --- /dev/null +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -0,0 +1,39 @@ +# Gallium3D Cell driver: PPU code + +# This makefile builds the g3dcell.a library which gets pulled into +# the main libGL.so library + + +TOP = ../../../../.. +include $(TOP)/configs/linux-cell + + +#PROG = gl4 + +CELL_LIB = libcell.a + +SPU_CODE_MODULE = ../spu/g3d_spu.a + +OBJECTS = cell_context.o cell_surface.o + +INCLUDE_DIRS = -I$(TOP)/src/mesa + + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + + + +default: $(CELL_LIB) + + +$(CELL_LIB): $(OBJECTS) $(SPU_CODE_MODULE) + ar -ru $(CELL_LIB) $(OBJECTS) $(SPU_CODE_MODULE) + +#$(PROG): $(PPU_OBJECTS) +# $(CC) -o $(PROG) $(PPU_OBJECTS) $(SPU_CODE_MODULE) $(PPU_LFLAGS) + + + +clean: + rm -f *.o $(CELL_LIB) diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c new file mode 100644 index 0000000000..a8f6cba2fa --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Authors + * Brian Paul + */ + + +#include +#include +#include +#include "pipe/cell/ppu/cell_context.h" +#include "pipe/cell/common.h" + +#define NUM_SPUS 6 + + +extern spe_program_handle_t g3d_spu; + +static speid_t speid[NUM_SPUS]; +static struct init_info inits[NUM_SPUS]; + + +static void +start_spus(void) +{ + int i; + + for (i = 0; i < NUM_SPUS; i++) { + inits[i].foo = i; + inits[i].bar = i * 10; + + speid[i] = spe_create_thread(0, /* gid */ + &g3d_spu, /* spe program handle */ + &inits[i], /* argp */ + NULL, /* envp */ + -1, /* mask */ + 0 ); /* flags */ + } +} + + + +void cell_create_context(void) +{ + printf("cell_create_context\n"); + + start_spus(); + + /* TODO: do something with the SPUs! */ +} diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h new file mode 100644 index 0000000000..b4d93d1414 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -0,0 +1,46 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_CONTEXT_H +#define CELL_CONTEXT_H + + +#include "pipe/p_context.h" + + + +struct cell_context +{ + struct pipe_context pipe; + + int spu_info; +}; + + + +#endif /* CELL_CONTEXT_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c new file mode 100644 index 0000000000..2c4b6e640b --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -0,0 +1,9 @@ + +#include + + +void cell_create_surface(void) +{ + printf("cell_create_surface\n"); + +} diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile new file mode 100644 index 0000000000..00f931e1c1 --- /dev/null +++ b/src/mesa/pipe/cell/spu/Makefile @@ -0,0 +1,42 @@ +# Gallium3D Cell driver: PPU code + +# This makefile builds the g3d_spu.a file that's linked into the +# PPU code/library. + + +TOP = ../../../../.. +include $(TOP)/configs/linux-cell + + +PROG = g3d + +PROG_SPU = $(PROG)_spu +PROG_SPU_A = $(PROG)_spu.a +PROG_SPU_EMBED_O = $(PROG)_spu-embed.o + + +SPU_OBJECTS = main.o tri.o + + +# The .a file will be linked into the main/PPU executable +default: $(PROG_SPU_A) + +$(PROG_SPU_A): $(PROG_SPU_EMBED_O) + $(SPU_AR) $(SPU_AR_FLAGS) $(PROG_SPU_A) $(PROG_SPU_EMBED_O) + +$(PROG_SPU_EMBED_O): $(PROG_SPU) + $(SPU_EMBED) $(SPU_EMBED_FLAGS) $(PROG_SPU) $(PROG_SPU) $(PROG_SPU_EMBED_O) + +$(PROG_SPU): $(SPU_OBJECTS) + $(SPU_CC) -o $(PROG_SPU) $(SPU_OBJECTS) $(SPU_LFLAGS) + + +main.o: main.c + $(SPU_CC) $(SPU_CFLAGS) -c main.c + +tri.o: tri.c + $(SPU_CC) $(SPU_CFLAGS) -c tri.c + + +clean: + rm -f *.o *.a *.d $(PROG_SPU) diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c new file mode 100644 index 0000000000..e8d5fdccbf --- /dev/null +++ b/src/mesa/pipe/cell/spu/main.c @@ -0,0 +1,29 @@ +/* main.c for cell SPU code */ + + +#include +#include +#include + +#include "tri.h" +#include "pipe/cell/common.h" + + +static struct init_info init; + + +int +main(unsigned long long speid, + unsigned long long argp, + unsigned long long envp) +{ + int tag = 0; + + mfc_get(&init, (unsigned int) argp, sizeof(struct init_info), tag, 0, 0); + + printf("Enter spu main(): init.foo=%d\n", init.foo); + + draw_triangle(0, 1, 2); + + return 0; +} diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c new file mode 100644 index 0000000000..949c3b4c8e --- /dev/null +++ b/src/mesa/pipe/cell/spu/tri.c @@ -0,0 +1,9 @@ + +#include "tri.h" + +void +draw_triangle(int v1, int v2, int v3) +{ + + +} diff --git a/src/mesa/pipe/cell/spu/tri.h b/src/mesa/pipe/cell/spu/tri.h new file mode 100644 index 0000000000..6a915de60d --- /dev/null +++ b/src/mesa/pipe/cell/spu/tri.h @@ -0,0 +1,4 @@ + + +extern void +draw_triangle(int v1, int v2, int v3); -- cgit v1.2.3 From fa3fc81f2c23f6ae7e4962cebee4356ca5e4383d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 5 Dec 2007 14:47:27 -0700 Subject: added B8G8R8A8 format --- src/mesa/pipe/p_format.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index 70eb124709..de0eb9539f 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -154,6 +154,7 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) #define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) #define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) #define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) +#define _PIPE_FORMAT_BGRA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_A ) #define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R ) #define _PIPE_FORMAT_RRR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) @@ -199,6 +200,7 @@ static INLINE uint pf_rev(pipe_format_ycbcr_t f) #define PIPE_FORMAT_NONE _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_0000, 0, 0, 0, 0, PIPE_FORMAT_TYPE_UNKNOWN ) #define PIPE_FORMAT_A8R8G8B8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) +#define PIPE_FORMAT_B8G8R8A8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) #define PIPE_FORMAT_A1R5G5B5_UNORM _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ) #define PIPE_FORMAT_A4R4G4B4_UNORM _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ) #define PIPE_FORMAT_R5G6B5_UNORM _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ) @@ -275,6 +277,7 @@ static INLINE uint pf_rev(pipe_format_ycbcr_t f) * Duplicated formats: */ #define PIPE_FORMAT_U_A8_R8_G8_B8 PIPE_FORMAT_A8R8G8B8_UNORM +#define PIPE_FORMAT_U_B8_G8_R8_A8 PIPE_FORMAT_B8G8R8A8_UNORM #define PIPE_FORMAT_U_A1_R5_G5_B5 PIPE_FORMAT_A1R5G5B5_UNORM #define PIPE_FORMAT_U_A4_R4_G4_B4 PIPE_FORMAT_A4R4G4B4_UNORM #define PIPE_FORMAT_U_R8_G8_B8_A8 PIPE_FORMAT_R8G8B8A8_UNORM -- cgit v1.2.3 From 04516cfcaf15589768c973231d8c317e851a78cb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 5 Dec 2007 14:48:09 -0700 Subject: added tile funcs for B8G8R8A8 format --- src/mesa/pipe/softpipe/sp_surface.c | 70 +++++++++++++++++++++++++++++++++++++ 1 file changed, 70 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 2ddf3ab99c..a6ab404603 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -133,6 +133,70 @@ a8r8g8b8_put_tile(struct pipe_surface *ps, } +/*** PIPE_FORMAT_U_B8_G8_R8_A8 ***/ + +static void +b8g8r8a8_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const unsigned *src + = ((const unsigned *) (ps->region->map + ps->offset)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_B8_G8_R8_A8); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const unsigned pixel = src[j]; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +static void +b8g8r8a8_put_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + unsigned *dst + = ((unsigned *) (ps->region->map + ps->offset)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_B8_G8_R8_A8); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + dst[j] = (b << 24) | (g << 16) | (r << 8) | a; + pRow += 4; + } + dst += ps->pitch; + p += w0 * 4; + } +} + + /*** PIPE_FORMAT_U_A1_R5_G5_B5 ***/ static void @@ -610,6 +674,9 @@ softpipe_get_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_U_A8_R8_G8_B8: a8r8g8b8_get_tile(ps, x, y, w, h, p); break; + case PIPE_FORMAT_U_B8_G8_R8_A8: + b8g8r8a8_get_tile(ps, x, y, w, h, p); + break; case PIPE_FORMAT_U_A1_R5_G5_B5: a1r5g5b5_get_tile(ps, x, y, w, h, p); break; @@ -657,6 +724,9 @@ softpipe_put_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_U_A8_R8_G8_B8: a8r8g8b8_put_tile(ps, x, y, w, h, p); break; + case PIPE_FORMAT_U_B8_G8_R8_A8: + b8g8r8a8_put_tile(ps, x, y, w, h, p); + break; case PIPE_FORMAT_U_A1_R5_G5_B5: /*a1r5g5b5_put_tile(ps, x, y, w, h, p);*/ break; -- cgit v1.2.3 From 9df0a6dd9c43be0ee5c300e161a20b2f98acef2d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 5 Dec 2007 14:51:08 -0700 Subject: added B8G8R8A8 support and improved pixel format selection when doing remote display to X server of different endianness. --- src/mesa/pipe/xlib/xm_api.c | 10 +++++--- src/mesa/pipe/xlib/xm_buffer.c | 30 ++++++++++++++++++++++-- src/mesa/pipe/xlib/xm_surface.c | 51 +++++++++++++++++++++++++++++++++++++++-- src/mesa/pipe/xlib/xmesaP.h | 6 ++++- 4 files changed, 89 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index e456ea2fd4..f87d72d0c8 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -329,7 +329,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, /* * Front renderbuffer */ - b->frontxrb = xmesa_create_renderbuffer(winsys, 0, &vis->mesa_visual, GL_FALSE); + b->frontxrb = xmesa_create_renderbuffer(winsys, 0, vis, GL_FALSE); if (!b->frontxrb) { _mesa_free(b); return NULL; @@ -343,7 +343,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, * Back renderbuffer */ if (vis->mesa_visual.doubleBufferMode) { - b->backxrb = xmesa_create_renderbuffer(winsys, 0, &vis->mesa_visual, GL_TRUE); + b->backxrb = xmesa_create_renderbuffer(winsys, 0, vis, GL_TRUE); if (!b->backxrb) { /* XXX free front xrb too */ _mesa_free(b); @@ -365,7 +365,11 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, /* Visual has alpha, but pixel format doesn't support it. * We'll use an alpha renderbuffer wrapper. */ - b->swAlpha = GL_TRUE; +#if 0 + b->swAlpha = GL_TRUE; /* don't do any renderbuffer wrapping */ +#else + b->swAlpha = GL_FALSE; +#endif } else { b->swAlpha = GL_FALSE; diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index 71a2003e6d..449e90184b 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -163,6 +163,16 @@ alloc_back_shm_ximage(XMesaBuffer b, GLuint width, GLuint height) #endif +/** + * \return LSBFirst or MSBFirst + */ +static int host_byte_order( void ) +{ + int i = 1; + char *cptr = (char *) &i; + return (*cptr==1) ? LSBFirst : MSBFirst; +} + /** * Setup an off-screen pixmap or Ximage to use as the back buffer. @@ -377,9 +387,11 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, */ struct xmesa_renderbuffer * xmesa_create_renderbuffer(struct pipe_winsys *winsys, - GLuint name, const GLvisual *visual, + GLuint name, XMesaVisual xmvis, GLboolean backBuffer) { + const GLvisual *visual = &xmvis->mesa_visual; + int byteOrder = ImageByteOrder(xmvis->display); /* LSBFirst or MSBFirst */ struct xmesa_renderbuffer *xrb = CALLOC_STRUCT(xmesa_renderbuffer); if (xrb) { GLuint name = 0; @@ -402,7 +414,21 @@ xmesa_create_renderbuffer(struct pipe_winsys *winsys, xrb->St.Base.GreenBits = visual->greenBits; xrb->St.Base.BlueBits = visual->blueBits; xrb->St.Base.AlphaBits = visual->alphaBits; - pipeFormat = PIPE_FORMAT_U_A8_R8_G8_B8; + if (visual->redMask == 0xff0000 && + visual->greenMask == 0x00ff00 && + visual->blueMask == 0x0000ff) { + if (host_byte_order() != byteOrder) { + pipeFormat = PIPE_FORMAT_U_B8_G8_R8_A8; + /*printf("Using format B8_G8_R8_A8 (LE dpy)\n");*/ + } + else { + pipeFormat = PIPE_FORMAT_U_A8_R8_G8_B8; + /*printf("Using format A8_R8_G8_B8 (BE dpy)\n");*/ + } + } + else { + assert(0); + } } else { xrb->St.Base.InternalFormat = GL_COLOR_INDEX; diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 2394563095..58ca2e7fb3 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -256,7 +256,7 @@ xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, /* put to ximage */ ximage = xms->ximage; char *dst; - int i; + uint i; /* this could be optimized/simplified */ switch (ps->format) { @@ -353,7 +353,7 @@ xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, for (j = 0; j < w; j++) { uint pix = src[j]; ubyte r = ((pix >> 16) & 0xff); - ubyte g = ((pix >> 8) & 0xff); + ubyte g = ((pix >> 8) & 0xff); ubyte b = ( pix & 0xff); ubyte a = ((pix >> 24) & 0xff); p[0] = UBYTE_TO_FLOAT(r); @@ -367,6 +367,29 @@ xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, } } break; + case PIPE_FORMAT_U_B8_G8_R8_A8: + { + const uint *src + = (uint *) (ximage->data + y * ximage->bytes_per_line + x * 4); + for (i = 0; i < h; i++) { + float *p = pRow; + for (j = 0; j < w; j++) { + uint pix = src[j]; + ubyte r = ((pix >> 8) & 0xff); + ubyte g = ((pix >> 16) & 0xff); + ubyte b = ((pix >> 24) & 0xff); + ubyte a = ( pix & 0xff); + p[0] = UBYTE_TO_FLOAT(r); + p[1] = UBYTE_TO_FLOAT(g); + p[2] = UBYTE_TO_FLOAT(b); + p[3] = UBYTE_TO_FLOAT(a); + p += 4; + } + src += ximage->width; + pRow += 4 * w0; + } + } + break; case PIPE_FORMAT_U_R5_G6_B5: { ushort *src @@ -390,6 +413,7 @@ xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, } break; default: + fprintf(stderr, "Bad format in xmesa_get_tile_rgba()\n"); assert(0); } @@ -461,6 +485,27 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, } } break; + case PIPE_FORMAT_U_B8_G8_R8_A8: + { + uint *dst + = (uint *) (ximage->data + y * ximage->bytes_per_line + x * 4); + const float *pRow = pixels; + for (i = 0; i < h; i++) { + const float *p = pRow; + for (j = 0; j < w; j++) { + ubyte r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, p[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, p[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, p[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, p[3]); + dst[j] = PACK_8B8G8R8A(r, g, b, a); + p += 4; + } + dst += ximage->width; + pRow += 4 * w0; + } + } + break; case PIPE_FORMAT_U_R5_G6_B5: { ushort *dst = @@ -483,6 +528,7 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, break; default: + fprintf(stderr, "Bad format in xmesa_put_tile_rgba()\n"); assert(0); } @@ -689,6 +735,7 @@ xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, uint value) clear_16bit_ximage_surface(pipe, ps, value); break; case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_U_B8_G8_R8_A8: clear_32bit_ximage_surface(pipe, ps, value); break; default: diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index ba0ccdbcec..a067dfd3e3 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -286,6 +286,10 @@ struct xmesa_buffer { */ #define PACK_8B8G8R( R, G, B ) ( ((B) << 16) | ((G) << 8) | (R) ) +#define PACK_8B8G8R8A( R, G, B, A ) \ + ( ((B) << 24) | ((G) << 16) | ((R) << 8) | (A) ) + + /** @@ -448,7 +452,7 @@ extern const int xmesa_kernel1[16]; extern struct xmesa_renderbuffer * xmesa_create_renderbuffer(struct pipe_winsys *winsys, - GLuint name, const GLvisual *visual, + GLuint name, XMesaVisual xmvis, GLboolean backBuffer); extern void -- cgit v1.2.3 From 3eadbe64bf027b7e90167d4980e880c8a01c889f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 5 Dec 2007 14:53:19 -0700 Subject: Fix endianness bug in _mesa_texstore_argb8888() On big-endian, storing in _mesa_texformat_argb8888 format produced wrong results. Also, clean-up nearby code to match. --- src/mesa/main/texstore.c | 28 +++++++++++++++------------- 1 file changed, 15 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index 3b5151ed17..30be65525e 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.1 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -1443,7 +1443,6 @@ _mesa_texstore_argb8888(TEXSTORE_PARAMS) (baseInternalFormat == GL_RGBA || baseInternalFormat == GL_RGB) && srcType == GL_UNSIGNED_BYTE) { - int img, row, col; for (img = 0; img < srcDepth; img++) { const GLint srcRowStride = _mesa_image_row_stride(srcPacking, @@ -1455,11 +1454,12 @@ _mesa_texstore_argb8888(TEXSTORE_PARAMS) + dstYoffset * dstRowStride + dstXoffset * dstFormat->TexelBytes; for (row = 0; row < srcHeight; row++) { + GLuint *d4 = (GLuint *) dstRow; for (col = 0; col < srcWidth; col++) { - dstRow[col * 4 + 0] = srcRow[col * 3 + BCOMP]; - dstRow[col * 4 + 1] = srcRow[col * 3 + GCOMP]; - dstRow[col * 4 + 2] = srcRow[col * 3 + RCOMP]; - dstRow[col * 4 + 3] = 0xff; + d4[col] = ((0xff << 24) | + (srcRow[col * 3 + RCOMP] << 16) | + (srcRow[col * 3 + GCOMP] << 8) | + (srcRow[col * 3 + BCOMP] << 0)); } dstRow += dstRowStride; srcRow += srcRowStride; @@ -1471,7 +1471,9 @@ _mesa_texstore_argb8888(TEXSTORE_PARAMS) dstFormat == &_mesa_texformat_argb8888 && srcFormat == GL_RGBA && baseInternalFormat == GL_RGBA && - (srcType == GL_UNSIGNED_BYTE && littleEndian)) { + srcType == GL_UNSIGNED_BYTE && + littleEndian) { + /* same as above case, but src data has alpha too */ GLint img, row, col; /* For some reason, streaming copies to write-combined regions * are extremely sensitive to the characteristics of how the @@ -1488,13 +1490,13 @@ _mesa_texstore_argb8888(TEXSTORE_PARAMS) + dstImageOffsets[dstZoffset + img] * dstFormat->TexelBytes + dstYoffset * dstRowStride + dstXoffset * dstFormat->TexelBytes; - for (row = 0; row < srcHeight; row++) { + GLuint *d4 = (GLuint *) dstRow; for (col = 0; col < srcWidth; col++) { - *(GLuint *)(dstRow + col * 4) = (srcRow[col * 4 + RCOMP] << 16 | - srcRow[col * 4 + GCOMP] << 8 | - srcRow[col * 4 + BCOMP] << 0 | - srcRow[col * 4 + ACOMP] << 24); + d4[col] = ((srcRow[col * 4 + ACOMP] << 24) | + (srcRow[col * 4 + RCOMP] << 16) | + (srcRow[col * 4 + GCOMP] << 8) | + (srcRow[col * 4 + BCOMP] << 0)); } dstRow += dstRowStride; srcRow += srcRowStride; -- cgit v1.2.3 From 146483d5412e14d6be63a9b9116ff683956ee294 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 5 Dec 2007 14:54:19 -0700 Subject: added PIPE_FORMAT_U_B8_G8_R8_A8 case in color_value() --- src/mesa/state_tracker/st_cb_clear.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 219a5afcbd..8ac6699776 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -69,9 +69,12 @@ color_value(GLuint pipeFormat, const GLfloat color[4]) return (r << 24) | (g << 16) | (b << 8) | a; case PIPE_FORMAT_U_A8_R8_G8_B8: return (a << 24) | (r << 16) | (g << 8) | b; + case PIPE_FORMAT_U_B8_G8_R8_A8: + return (b << 24) | (g << 16) | (r << 8) | a; case PIPE_FORMAT_U_R5_G6_B5: return ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); default: + assert(0); return 0; } } -- cgit v1.2.3 From ce30f0550f02b28219c84851d71e0713aa315ee3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 5 Dec 2007 14:54:29 -0700 Subject: silence warning --- src/mesa/state_tracker/st_draw.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 61ff034550..b4c0d0cdd6 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -83,6 +83,7 @@ pipe_vertex_format(GLenum type, GLuint size) default: assert(0); } + return 0; /* silence compiler warning */ } -- cgit v1.2.3 From 59356268187470c5fda9e9a1a7058607f938fb3b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 5 Dec 2007 14:57:19 -0700 Subject: Add dependency checking for Cell driver sources. --- src/mesa/Makefile | 5 +++-- src/mesa/pipe/cell/ppu/Makefile | 22 ++++++++++++++++++++-- src/mesa/pipe/cell/spu/Makefile | 20 ++++++++++++++++++-- 3 files changed, 41 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 6f0877d335..6f090d9b3d 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -15,6 +15,7 @@ SOFTPIPE_LIB = $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a ifeq ($(CONFIG_NAME), linux-cell) CELL_LIB = $(TOP)/src/mesa/pipe/cell/ppu/libcell.a +CELL_LIB_SPU = $(TOP)/src/mesa/pipe/cell/spu/g3d_spu.a endif @@ -117,12 +118,12 @@ stand-alone: depend subdirs $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) $(TOP)/$(LIB_DIR)/$ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) $(CELL_LIB) +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) @ $(TOP)/bin/mklib -o $(GL_LIB) \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ - $(SOFTPIPE_LIB) $(CELL_LIB)$(GL_LIB_DEPS) + $(SOFTPIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) # Make the OSMesa library $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECTS) diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index ede341abca..c987d54e02 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -14,7 +14,12 @@ CELL_LIB = libcell.a SPU_CODE_MODULE = ../spu/g3d_spu.a -OBJECTS = cell_context.o cell_surface.o + +SOURCES = \ + cell_context.c \ + cell_surface.c + +OBJECTS = $(SOURCES:.c=.o) \ INCLUDE_DIRS = -I$(TOP)/src/mesa @@ -28,7 +33,8 @@ default: $(CELL_LIB) $(CELL_LIB): $(OBJECTS) $(SPU_CODE_MODULE) - ar -ru $(CELL_LIB) $(OBJECTS) $(SPU_CODE_MODULE) +# ar -ru $(CELL_LIB) $(OBJECTS) $(SPU_CODE_MODULE) + ar -ru $(CELL_LIB) $(OBJECTS) #$(PROG): $(PPU_OBJECTS) # $(CC) -o $(PROG) $(PPU_OBJECTS) $(SPU_CODE_MODULE) $(PPU_LFLAGS) @@ -37,3 +43,15 @@ $(CELL_LIB): $(OBJECTS) $(SPU_CODE_MODULE) clean: rm -f *.o $(CELL_LIB) + + + +depend: $(SOURCES) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(SOURCES) 2> /dev/null + +include depend + + + diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index 00f931e1c1..700a6c61cd 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -1,4 +1,4 @@ -# Gallium3D Cell driver: PPU code +# Gallium3D Cell driver: SPU code # This makefile builds the g3d_spu.a file that's linked into the # PPU code/library. @@ -15,7 +15,13 @@ PROG_SPU_A = $(PROG)_spu.a PROG_SPU_EMBED_O = $(PROG)_spu-embed.o -SPU_OBJECTS = main.o tri.o +SOURCES = \ + main.c \ + tri.c + +SPU_OBJECTS = $(SOURCES:.c=.o) \ + +INCLUDE_DIRS = -I$(TOP)/src/mesa # The .a file will be linked into the main/PPU executable @@ -40,3 +46,13 @@ tri.o: tri.c clean: rm -f *.o *.a *.d $(PROG_SPU) + + + +depend: $(SOURCES) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(SOURCES) 2> /dev/null + +include depend + -- cgit v1.2.3 From 753db0d8407147393a7b0622ae3fa28f68d0353d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 30 Nov 2007 20:48:03 +0100 Subject: Hide texture layout details from the state tracker. pipe->get_tex_surface() has to be used for access to texture image data. --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 4 +- src/mesa/pipe/failover/fo_context.c | 3 +- src/mesa/pipe/failover/fo_context.h | 2 +- src/mesa/pipe/failover/fo_state.c | 2 +- src/mesa/pipe/i915simple/Makefile | 2 +- src/mesa/pipe/i915simple/i915_context.c | 9 +- src/mesa/pipe/i915simple/i915_context.h | 30 +- src/mesa/pipe/i915simple/i915_state.c | 4 +- src/mesa/pipe/i915simple/i915_state_sampler.c | 29 +- src/mesa/pipe/i915simple/i915_surface.c | 25 +- src/mesa/pipe/i915simple/i915_tex_layout.c | 474 -------------------- src/mesa/pipe/i915simple/i915_tex_layout.h | 16 - src/mesa/pipe/i915simple/i915_texture.c | 539 +++++++++++++++++++++++ src/mesa/pipe/i915simple/i915_texture.h | 16 + src/mesa/pipe/llvm/llvm_entry.c | 2 - src/mesa/pipe/p_context.h | 11 +- src/mesa/pipe/p_inlines.h | 20 + src/mesa/pipe/p_state.h | 42 +- src/mesa/pipe/softpipe/Makefile | 2 +- src/mesa/pipe/softpipe/sp_context.c | 17 +- src/mesa/pipe/softpipe/sp_context.h | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/pipe/softpipe/sp_state.h | 31 +- src/mesa/pipe/softpipe/sp_state_fs.c | 2 + src/mesa/pipe/softpipe/sp_state_sampler.c | 4 +- src/mesa/pipe/softpipe/sp_surface.c | 25 +- src/mesa/pipe/softpipe/sp_surface.h | 2 +- src/mesa/pipe/softpipe/sp_tex_layout.c | 361 --------------- src/mesa/pipe/softpipe/sp_tex_layout.h | 16 - src/mesa/pipe/softpipe/sp_tex_sample.c | 18 +- src/mesa/pipe/softpipe/sp_texture.c | 425 ++++++++++++++++++ src/mesa/pipe/softpipe/sp_texture.h | 18 + src/mesa/pipe/softpipe/sp_tile_cache.c | 4 +- src/mesa/pipe/softpipe/sp_tile_cache.h | 2 +- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 2 +- src/mesa/sources | 2 +- src/mesa/state_tracker/st_atom_texture.c | 14 +- src/mesa/state_tracker/st_cb_drawpixels.c | 214 +++------ src/mesa/state_tracker/st_cb_fbo.c | 24 +- src/mesa/state_tracker/st_cb_texture.c | 427 +++++++++--------- src/mesa/state_tracker/st_cb_texture.h | 10 +- src/mesa/state_tracker/st_context.h | 22 +- src/mesa/state_tracker/st_mipmap_tree.c | 328 -------------- src/mesa/state_tracker/st_mipmap_tree.h | 118 ----- src/mesa/state_tracker/st_texture.c | 276 ++++++++++++ src/mesa/state_tracker/st_texture.h | 109 +++++ 46 files changed, 1877 insertions(+), 1830 deletions(-) delete mode 100644 src/mesa/pipe/i915simple/i915_tex_layout.c delete mode 100644 src/mesa/pipe/i915simple/i915_tex_layout.h create mode 100644 src/mesa/pipe/i915simple/i915_texture.c create mode 100644 src/mesa/pipe/i915simple/i915_texture.h delete mode 100644 src/mesa/pipe/softpipe/sp_tex_layout.c delete mode 100644 src/mesa/pipe/softpipe/sp_tex_layout.h create mode 100644 src/mesa/pipe/softpipe/sp_texture.c create mode 100644 src/mesa/pipe/softpipe/sp_texture.h delete mode 100644 src/mesa/state_tracker/st_mipmap_tree.c delete mode 100644 src/mesa/state_tracker/st_mipmap_tree.h create mode 100644 src/mesa/state_tracker/st_texture.c create mode 100644 src/mesa/state_tracker/st_texture.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 01460e5be3..1b520f7135 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -328,8 +328,8 @@ intelSetTexOffset(__DRIcontext *pDRICtx, int texname, if (!stObj) return; - if (stObj->mt) - st_miptree_release(intel->st->pipe, &stObj->mt); + if (stObj->pt) + st->pipe->texture_release(intel->st->pipe, &stObj->pt); stObj->imageOverride = GL_TRUE; stObj->depthOverride = depth; diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index 0cc9cab408..a25563d451 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -145,7 +145,8 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.surface_data = hw->surface_data; failover->pipe.surface_copy = hw->surface_copy; failover->pipe.surface_fill = hw->surface_fill; - failover->pipe.mipmap_tree_layout = hw->mipmap_tree_layout; + failover->pipe.texture_create = hw->texture_create; + failover->pipe.texture_release = hw->texture_release; failover->pipe.flush = hw->flush; failover->dirty = 0; diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index 759b53ccbe..7cf18c9ec1 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -85,7 +85,7 @@ struct failover_context { struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; uint sampler_units[PIPE_MAX_SAMPLERS]; - struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; + struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 2cd1a50b20..fd6137ba66 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -402,7 +402,7 @@ failover_delete_sampler_state(struct pipe_context *pipe, void *sampler) static void failover_set_texture_state(struct pipe_context *pipe, unsigned unit, - struct pipe_mipmap_tree *texture) + struct pipe_texture *texture) { struct failover_context *failover = failover_context(pipe); diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 391a084915..1223b386a3 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -22,7 +22,7 @@ DRIVER_SOURCES = \ i915_strings.c \ i915_prim_emit.c \ i915_prim_vbuf.c \ - i915_tex_layout.c \ + i915_texture.c \ i915_fpc_emit.c \ i915_fpc_translate.c \ i915_surface.c diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index a0ed3032b1..94649231cf 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -29,7 +29,7 @@ #include "i915_winsys.h" #include "i915_state.h" #include "i915_batch.h" -#include "i915_tex_layout.h" +#include "i915_texture.h" #include "i915_reg.h" #include "pipe/draw/draw_context.h" @@ -357,11 +357,8 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pci_id = pci_id; i915->flags.is_i945 = is_i945; - if (i915->flags.is_i945) - i915->pipe.mipmap_tree_layout = i945_miptree_layout; - else - i915->pipe.mipmap_tree_layout = i915_miptree_layout; - + i915->pipe.texture_create = i915_texture_create; + i915->pipe.texture_release = i915_texture_release; i915->dirty = ~0; i915->hardware_dirty = ~0; diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index ee430ebc90..8ed3465be2 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -150,6 +150,34 @@ struct i915_alpha_test_state { unsigned LIS6; }; +struct i915_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_region *region; +}; + struct i915_context { struct pipe_context pipe; @@ -174,7 +202,7 @@ struct i915_context struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; uint sampler_units[PIPE_MAX_SAMPLERS]; - struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; + struct i915_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 468d0ce91b..038fd623ea 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -525,11 +525,11 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, static void i915_set_texture_state(struct pipe_context *pipe, unsigned unit, - struct pipe_mipmap_tree *texture) + struct pipe_texture *texture) { struct i915_context *i915 = i915_context(pipe); - i915->texture[unit] = texture; /* ptr, not struct */ + i915->texture[unit] = (struct i915_texture*)texture; /* ptr, not struct */ i915->dirty |= I915_NEW_TEXTURE; } diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 0991e6ac0d..1816d9abdb 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -46,9 +46,11 @@ static void update_sampler(struct i915_context *i915, uint unit, const struct i915_sampler_state *sampler, - const struct pipe_mipmap_tree *mt, + const struct i915_texture *tex, unsigned state[3] ) { + const struct pipe_texture *pt = &tex->base; + /* Need to do this after updating the maps, which call the * intel_finalize_mipmap_tree and hence can update firstLevel: */ @@ -56,8 +58,8 @@ static void update_sampler(struct i915_context *i915, state[1] = sampler->state[1]; state[2] = sampler->state[2]; - if (mt->format == PIPE_FORMAT_YCBCR || - mt->format == PIPE_FORMAT_YCBCR_REV) + if (pt->format == PIPE_FORMAT_YCBCR || + pt->format == PIPE_FORMAT_YCBCR_REV) state[0] |= SS2_COLORSPACE_CONVERSION; /* 3D textures don't seem to respect the border color. @@ -75,7 +77,7 @@ static void update_sampler(struct i915_context *i915, const unsigned ws = sampler->templ->wrap_s; const unsigned wt = sampler->templ->wrap_t; const unsigned wr = sampler->templ->wrap_r; - if (mt->target == PIPE_TEXTURE_3D && + if (pt->target == PIPE_TEXTURE_3D && (sampler->templ->min_img_filter != PIPE_TEX_FILTER_NEAREST || sampler->templ->mag_img_filter != PIPE_TEX_FILTER_NEAREST) && (ws == PIPE_TEX_WRAP_CLAMP || @@ -105,13 +107,13 @@ void i915_update_samplers( struct i915_context *i915 ) i915->current.sampler_enable_flags = 0x0; for (unit = 0; unit < I915_TEX_UNITS; unit++) { - /* determine unit enable/disable by looking for a bound mipmap tree */ + /* determine unit enable/disable by looking for a bound texture */ /* could also examine the fragment program? */ if (i915->texture[unit]) { update_sampler( i915, unit, i915->sampler[unit], /* sampler state */ - i915->texture[unit], /* mipmap tree */ + i915->texture[unit], /* texture */ i915->current.sampler[unit] /* the result */ ); @@ -179,18 +181,19 @@ static void i915_update_texture(struct i915_context *i915, uint unit, uint state[6]) { - const struct pipe_mipmap_tree *mt = i915->texture[unit]; + const struct i915_texture *tex = i915->texture[unit]; + const struct pipe_texture *pt = &tex->base; uint format, pitch; - const uint width = mt->width0, height = mt->height0, depth = mt->depth0; - const uint num_levels = mt->last_level - mt->first_level; + const uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; + const uint num_levels = pt->last_level - pt->first_level; - assert(mt); + assert(tex); assert(width); assert(height); assert(depth); - format = translate_texture_format(mt->format); - pitch = mt->pitch * mt->cpp; + format = translate_texture_format(pt->format); + pitch = tex->pitch * pt->cpp; assert(format); assert(pitch); @@ -217,7 +220,7 @@ i915_update_textures(struct i915_context *i915) uint unit; for (unit = 0; unit < I915_TEX_UNITS; unit++) { - /* determine unit enable/disable by looking for a bound mipmap tree */ + /* determine unit enable/disable by looking for a bound texture */ /* could also examine the fragment program? */ if (i915->texture[unit]) { i915_update_texture(i915, unit, i915->current.texbuffer[unit]); diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index e4a5de00d7..385202507d 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -187,34 +187,35 @@ i915_put_tile(struct pipe_context *pipe, */ static struct pipe_surface * i915_get_tex_surface(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt, + struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice) { + struct i915_texture *tex = (struct i915_texture *)pt; struct pipe_surface *ps; unsigned offset; /* in bytes */ - offset = mt->level[level].level_offset; + offset = tex->level_offset[level]; - if (mt->target == PIPE_TEXTURE_CUBE) { - offset += mt->level[level].image_offset[face] * mt->cpp; + if (pt->target == PIPE_TEXTURE_CUBE) { + offset += tex->image_offset[level][face] * pt->cpp; } - else if (mt->target == PIPE_TEXTURE_3D) { - offset += mt->level[level].image_offset[zslice] * mt->cpp; + else if (pt->target == PIPE_TEXTURE_3D) { + offset += tex->image_offset[level][zslice] * pt->cpp; } else { assert(face == 0); assert(zslice == 0); } - ps = pipe->winsys->surface_alloc(pipe->winsys, mt->format); + ps = pipe->winsys->surface_alloc(pipe->winsys, pt->format); if (ps) { assert(ps->format); assert(ps->refcount); - pipe_region_reference(&ps->region, mt->region); - ps->cpp = mt->cpp; - ps->width = mt->level[level].width; - ps->height = mt->level[level].height; - ps->pitch = mt->pitch; + pipe_region_reference(&ps->region, tex->region); + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = tex->pitch; ps->offset = offset; } return ps; diff --git a/src/mesa/pipe/i915simple/i915_tex_layout.c b/src/mesa/pipe/i915simple/i915_tex_layout.c deleted file mode 100644 index cb372a7170..0000000000 --- a/src/mesa/pipe/i915simple/i915_tex_layout.c +++ /dev/null @@ -1,474 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Michel Dänzer - */ - -#include "pipe/p_state.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" - -#include "i915_tex_layout.h" -#include "i915_debug.h" - - -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - -static int align(int value, int alignment) -{ - return (value + alignment - 1) & ~(alignment - 1); -} - - -static void -i915_miptree_set_level_info(struct pipe_mipmap_tree *mt, - unsigned level, - unsigned nr_images, - unsigned x, unsigned y, unsigned w, unsigned h, unsigned d) -{ - assert(level < PIPE_MAX_TEXTURE_LEVELS); - - mt->level[level].width = w; - mt->level[level].height = h; - mt->level[level].depth = d; - mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; - mt->level[level].nr_images = nr_images; - - /* - DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - level, w, h, d, x, y, mt->level[level].level_offset); - */ - - /* Not sure when this would happen, but anyway: - */ - if (mt->level[level].image_offset) { - FREE(mt->level[level].image_offset); - mt->level[level].image_offset = NULL; - } - - assert(nr_images); - assert(!mt->level[level].image_offset); - - mt->level[level].image_offset = (unsigned *) MALLOC(nr_images * sizeof(unsigned)); - mt->level[level].image_offset[0] = 0; -} - - -static void -i915_miptree_set_image_offset(struct pipe_mipmap_tree *mt, - unsigned level, unsigned img, unsigned x, unsigned y) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < mt->level[level].nr_images); - - mt->level[level].image_offset[img] = (x + y * mt->pitch); - - /* - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); - */ -} - - -static void -i945_miptree_layout_2d( struct pipe_mipmap_tree *mt ) -{ - int align_h = 2, align_w = 4; - unsigned level; - unsigned x = 0; - unsigned y = 0; - unsigned width = mt->width0; - unsigned height = mt->height0; - - mt->pitch = mt->width0; - - /* May need to adjust pitch to accomodate the placement of - * the 2nd mipmap. This occurs when the alignment - * constraints of mipmap placement push the right edge of the - * 2nd mipmap out past the width of its parent. - */ - if (mt->first_level != mt->last_level) { - unsigned mip1_width = align(minify(mt->width0), align_w) - + minify(minify(mt->width0)); - - if (mip1_width > mt->width0) - mt->pitch = mip1_width; - } - - /* Pitch must be a whole number of dwords, even though we - * express it in texels. - */ - mt->pitch = align(mt->pitch * mt->cpp, 4) / mt->cpp; - mt->total_height = 0; - - for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { - unsigned img_height; - - i915_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); - - if (mt->compressed) - img_height = MAX2(1, height/4); - else - img_height = align(height, align_h); - - - /* Because the images are packed better, the final offset - * might not be the maximal one: - */ - mt->total_height = MAX2(mt->total_height, y + img_height); - - /* Layout_below: step right after second mipmap. - */ - if (level == mt->first_level + 1) { - x += align(width, align_w); - } - else { - y += img_height; - } - - width = minify(width); - height = minify(height); - } -} - - -static const int initial_offsets[6][2] = { - {0, 0}, - {0, 2}, - {1, 0}, - {1, 2}, - {1, 1}, - {1, 3} -}; - -static const int step_offsets[6][2] = { - {0, 2}, - {0, 2}, - {-1, 2}, - {-1, 2}, - {-1, 1}, - {-1, 1} -}; - - -boolean -i915_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) -{ - unsigned level; - - switch (mt->target) { - case PIPE_TEXTURE_CUBE: { - const unsigned dim = mt->width0; - unsigned face; - unsigned lvlWidth = mt->width0, lvlHeight = mt->height0; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* double pitch for cube layouts */ - mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; - mt->total_height = dim * 4; - - for (level = mt->first_level; level <= mt->last_level; level++) { - i915_miptree_set_level_info(mt, level, 6, - 0, 0, - /*OLD: mt->pitch, mt->total_height,*/ - lvlWidth, lvlHeight, - 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * dim; - unsigned y = initial_offsets[face][1] * dim; - unsigned d = dim; - - for (level = mt->first_level; level <= mt->last_level; level++) { - i915_miptree_set_image_offset(mt, level, face, x, y); - d >>= 1; - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - } - } - break; - } - case PIPE_TEXTURE_3D:{ - unsigned width = mt->width0; - unsigned height = mt->height0; - unsigned depth = mt->depth0; - unsigned stack_height = 0; - - /* Calculate the size of a single slice. - */ - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - - /* XXX: hardware expects/requires 9 levels at minimum. - */ - for (level = mt->first_level; level <= MAX2(8, mt->last_level); - level++) { - i915_miptree_set_level_info(mt, level, depth, 0, mt->total_height, - width, height, depth); - - - stack_height += MAX2(2, height); - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - - /* Fixup depth image_offsets: - */ - depth = mt->depth0; - for (level = mt->first_level; level <= mt->last_level; level++) { - unsigned i; - for (i = 0; i < depth; i++) - i915_miptree_set_image_offset(mt, level, i, - 0, i * stack_height); - - depth = minify(depth); - } - - - /* Multiply slice size by texture depth for total size. It's - * remarkable how wasteful of memory the i915 texture layouts - * are. They are largely fixed in the i945. - */ - mt->total_height = stack_height * mt->depth0; - break; - } - - default:{ - unsigned width = mt->width0; - unsigned height = mt->height0; - unsigned img_height; - - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - mt->total_height = 0; - - for (level = mt->first_level; level <= mt->last_level; level++) { - i915_miptree_set_level_info(mt, level, 1, - 0, mt->total_height, - width, height, 1); - - if (mt->compressed) - img_height = MAX2(1, height / 4); - else - img_height = (MAX2(2, height) + 1) & ~1; - - mt->total_height += img_height; - - width = minify(width); - height = minify(height); - } - break; - } - } - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); - */ - - return TRUE; -} - - -boolean -i945_miptree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) -{ - unsigned level; - - switch (mt->target) { - case PIPE_TEXTURE_CUBE:{ - const unsigned dim = mt->width0; - unsigned face; - unsigned lvlWidth = mt->width0, lvlHeight = mt->height0; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* Depending on the size of the largest images, pitch can be - * determined either by the old-style packing of cubemap faces, - * or the final row of 4x4, 2x2 and 1x1 faces below this. - */ - if (dim > 32) - mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; - else - mt->pitch = 14 * 8; - - mt->total_height = dim * 4 + 4; - - /* Set all the levels to effectively occupy the whole rectangular region. - */ - for (level = mt->first_level; level <= mt->last_level; level++) { - i915_miptree_set_level_info(mt, level, 6, - 0, 0, - lvlWidth, lvlHeight, 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - - for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * dim; - unsigned y = initial_offsets[face][1] * dim; - unsigned d = dim; - - if (dim == 4 && face >= 4) { - y = mt->total_height - 4; - x = (face - 4) * 8; - } - else if (dim < 4 && (face > 0 || mt->first_level > 0)) { - y = mt->total_height - 4; - x = face * 8; - } - - for (level = mt->first_level; level <= mt->last_level; level++) { - i915_miptree_set_image_offset(mt, level, face, x, y); - - d >>= 1; - - switch (d) { - case 4: - switch (face) { - case PIPE_TEX_FACE_POS_X: - case PIPE_TEX_FACE_NEG_X: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - case PIPE_TEX_FACE_POS_Y: - case PIPE_TEX_FACE_NEG_Y: - y += 12; - x -= 8; - break; - case PIPE_TEX_FACE_POS_Z: - case PIPE_TEX_FACE_NEG_Z: - y = mt->total_height - 4; - x = (face - 4) * 8; - break; - } - - case 2: - y = mt->total_height - 4; - x = 16 + face * 8; - break; - - case 1: - x += 48; - break; - - default: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - } - } - } - break; - } - case PIPE_TEXTURE_3D:{ - unsigned width = mt->width0; - unsigned height = mt->height0; - unsigned depth = mt->depth0; - unsigned pack_x_pitch, pack_x_nr; - unsigned pack_y_pitch; - unsigned level; - - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - mt->total_height = 0; - - pack_y_pitch = MAX2(mt->height0, 2); - pack_x_pitch = mt->pitch; - pack_x_nr = 1; - - for (level = mt->first_level; level <= mt->last_level; level++) { - unsigned nr_images = mt->target == PIPE_TEXTURE_3D ? depth : 6; - int x = 0; - int y = 0; - unsigned q, j; - - i915_miptree_set_level_info(mt, level, nr_images, - 0, mt->total_height, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - i915_miptree_set_image_offset(mt, level, q, x, y); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - mt->total_height += y; - - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= mt->pitch); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - } - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - break; - } - - case PIPE_TEXTURE_1D: - case PIPE_TEXTURE_2D: -// case PIPE_TEXTURE_RECTANGLE: - i945_miptree_layout_2d(mt); - break; - default: - assert(0); - return FALSE; - } - - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); - */ - - return TRUE; -} - diff --git a/src/mesa/pipe/i915simple/i915_tex_layout.h b/src/mesa/pipe/i915simple/i915_tex_layout.h deleted file mode 100644 index e033786381..0000000000 --- a/src/mesa/pipe/i915simple/i915_tex_layout.h +++ /dev/null @@ -1,16 +0,0 @@ - -#ifndef I915_TEX_LAYOUT_H -#define I915_TEX_LAYOUT_H - -struct pipe_context; -struct pipe_mipmap_tree; - - -extern boolean -i915_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); - -extern boolean -i945_miptree_layout(struct pipe_context *, struct pipe_mipmap_tree *); - - -#endif /* I915_TEX_LAYOUT_H */ diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c new file mode 100644 index 0000000000..3bfa806d9e --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -0,0 +1,539 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "pipe/p_state.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "i915_context.h" +#include "i915_texture.h" +#include "i915_debug.h" + + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + +static int align(int value, int alignment) +{ + return (value + alignment - 1) & ~(alignment - 1); +} + + +static void +i915_miptree_set_level_info(struct i915_texture *tex, + unsigned level, + unsigned nr_images, + unsigned x, unsigned y, unsigned w, unsigned h, unsigned d) +{ + struct pipe_texture *pt = &tex->base; + + assert(level < PIPE_MAX_TEXTURE_LEVELS); + + pt->width[level] = w; + pt->height[level] = h; + pt->depth[level] = d; + + tex->level_offset[level] = (x + y * tex->pitch) * pt->cpp; + tex->nr_images[level] = nr_images; + + /* + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, tex->level_offset[level]); + */ + + /* Not sure when this would happen, but anyway: + */ + if (tex->image_offset[level]) { + FREE(tex->image_offset[level]); + tex->image_offset[level] = NULL; + } + + assert(nr_images); + assert(!tex->image_offset[level]); + + tex->image_offset[level] = (unsigned *) MALLOC(nr_images * sizeof(unsigned)); + tex->image_offset[level][0] = 0; +} + + +static void +i915_miptree_set_image_offset(struct i915_texture *tex, + unsigned level, unsigned img, unsigned x, unsigned y) +{ + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + + assert(img < tex->nr_images[level]); + + tex->image_offset[level][img] = (x + y * tex->pitch); + + /* + DBG("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, tex->image_offset[level][img]); + */ +} + + +static void +i945_miptree_layout_2d( struct i915_texture *tex ) +{ + struct pipe_texture *pt = &tex->base; + int align_h = 2, align_w = 4; + unsigned level; + unsigned x = 0; + unsigned y = 0; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + + tex->pitch = pt->width[0]; + + /* May need to adjust pitch to accomodate the placement of + * the 2nd mipmap. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap out past the width of its parent. + */ + if (pt->first_level != pt->last_level) { + unsigned mip1_width = align(minify(pt->width[0]), align_w) + + minify(minify(pt->width[0])); + + if (mip1_width > pt->width[0]) + tex->pitch = mip1_width; + } + + /* Pitch must be a whole number of dwords, even though we + * express it in texels. + */ + tex->pitch = align(tex->pitch * pt->cpp, 4) / pt->cpp; + tex->total_height = 0; + + for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + unsigned img_height; + + i915_miptree_set_level_info(tex, level, 1, x, y, width, height, 1); + + if (pt->compressed) + img_height = MAX2(1, height/4); + else + img_height = align(height, align_h); + + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + tex->total_height = MAX2(tex->total_height, y + img_height); + + /* Layout_below: step right after second mipmap. + */ + if (level == pt->first_level + 1) { + x += align(width, align_w); + } + else { + y += img_height; + } + + width = minify(width); + height = minify(height); + } +} + + +static const int initial_offsets[6][2] = { + {0, 0}, + {0, 2}, + {1, 0}, + {1, 2}, + {1, 1}, + {1, 3} +}; + +static const int step_offsets[6][2] = { + {0, 2}, + {0, 2}, + {-1, 2}, + {-1, 2}, + {-1, 1}, + {-1, 1} +}; + + +static boolean +i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned level; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: { + const unsigned dim = pt->width[0]; + unsigned face; + unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* double pitch for cube layouts */ + tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; + tex->total_height = dim * 4; + + for (level = pt->first_level; level <= pt->last_level; level++) { + i915_miptree_set_level_info(tex, level, 6, + 0, 0, + /*OLD: tex->pitch, tex->total_height,*/ + lvlWidth, lvlHeight, + 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + for (face = 0; face < 6; face++) { + unsigned x = initial_offsets[face][0] * dim; + unsigned y = initial_offsets[face][1] * dim; + unsigned d = dim; + + for (level = pt->first_level; level <= pt->last_level; level++) { + i915_miptree_set_image_offset(tex, level, face, x, y); + d >>= 1; + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + } + } + break; + } + case PIPE_TEXTURE_3D:{ + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + unsigned stack_height = 0; + + /* Calculate the size of a single slice. + */ + tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; + + /* XXX: hardware expects/requires 9 levels at minimum. + */ + for (level = pt->first_level; level <= MAX2(8, pt->last_level); + level++) { + i915_miptree_set_level_info(tex, level, depth, 0, tex->total_height, + width, height, depth); + + + stack_height += MAX2(2, height); + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + + /* Fixup depth image_offsets: + */ + depth = pt->depth[0]; + for (level = pt->first_level; level <= pt->last_level; level++) { + unsigned i; + for (i = 0; i < depth; i++) + i915_miptree_set_image_offset(tex, level, i, + 0, i * stack_height); + + depth = minify(depth); + } + + + /* Multiply slice size by texture depth for total size. It's + * remarkable how wasteful of memory the i915 texture layouts + * are. They are largely fixed in the i945. + */ + tex->total_height = stack_height * pt->depth[0]; + break; + } + + default:{ + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned img_height; + + tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; + tex->total_height = 0; + + for (level = pt->first_level; level <= pt->last_level; level++) { + i915_miptree_set_level_info(tex, level, 1, + 0, tex->total_height, + width, height, 1); + + if (pt->compressed) + img_height = MAX2(1, height / 4); + else + img_height = (MAX2(2, height) + 1) & ~1; + + tex->total_height += img_height; + + width = minify(width); + height = minify(height); + } + break; + } + } + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + tex->pitch, + tex->total_height, pt->cpp, tex->pitch * tex->total_height * pt->cpp); + */ + + return TRUE; +} + + +static boolean +i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned level; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE:{ + const unsigned dim = pt->width[0]; + unsigned face; + unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* Depending on the size of the largest images, pitch can be + * determined either by the old-style packing of cubemap faces, + * or the final row of 4x4, 2x2 and 1x1 faces below this. + */ + if (dim > 32) + tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; + else + tex->pitch = 14 * 8; + + tex->total_height = dim * 4 + 4; + + /* Set all the levels to effectively occupy the whole rectangular region. + */ + for (level = pt->first_level; level <= pt->last_level; level++) { + i915_miptree_set_level_info(tex, level, 6, + 0, 0, + lvlWidth, lvlHeight, 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + + for (face = 0; face < 6; face++) { + unsigned x = initial_offsets[face][0] * dim; + unsigned y = initial_offsets[face][1] * dim; + unsigned d = dim; + + if (dim == 4 && face >= 4) { + y = tex->total_height - 4; + x = (face - 4) * 8; + } + else if (dim < 4 && (face > 0 || pt->first_level > 0)) { + y = tex->total_height - 4; + x = face * 8; + } + + for (level = pt->first_level; level <= pt->last_level; level++) { + i915_miptree_set_image_offset(tex, level, face, x, y); + + d >>= 1; + + switch (d) { + case 4: + switch (face) { + case PIPE_TEX_FACE_POS_X: + case PIPE_TEX_FACE_NEG_X: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + case PIPE_TEX_FACE_POS_Y: + case PIPE_TEX_FACE_NEG_Y: + y += 12; + x -= 8; + break; + case PIPE_TEX_FACE_POS_Z: + case PIPE_TEX_FACE_NEG_Z: + y = tex->total_height - 4; + x = (face - 4) * 8; + break; + } + + case 2: + y = tex->total_height - 4; + x = 16 + face * 8; + break; + + case 1: + x += 48; + break; + + default: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + } + } + } + break; + } + case PIPE_TEXTURE_3D:{ + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + unsigned pack_x_pitch, pack_x_nr; + unsigned pack_y_pitch; + unsigned level; + + tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; + tex->total_height = 0; + + pack_y_pitch = MAX2(pt->height[0], 2); + pack_x_pitch = tex->pitch; + pack_x_nr = 1; + + for (level = pt->first_level; level <= pt->last_level; level++) { + unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; + int x = 0; + int y = 0; + unsigned q, j; + + i915_miptree_set_level_info(tex, level, nr_images, + 0, tex->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + i915_miptree_set_image_offset(tex, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + tex->total_height += y; + + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= tex->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + } + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + break; + } + + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_2D: +// case PIPE_TEXTURE_RECTANGLE: + i945_miptree_layout_2d(tex); + break; + default: + assert(0); + return FALSE; + } + + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + tex->pitch, + tex->total_height, pt->cpp, tex->pitch * tex->total_height * pt->cpp); + */ + + return TRUE; +} + +void +i915_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) +{ + struct i915_texture *tex = REALLOC(*pt, sizeof(struct pipe_texture), + sizeof(struct i915_texture)); + + if (tex) { + struct i915_context *i915 = i915_context(pipe); + + memset(&tex->base + 1, 0, + sizeof(struct i915_texture) - sizeof(struct pipe_texture)); + + if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : + i915_miptree_layout(pipe, tex)) { + tex->region = pipe->winsys->region_alloc(pipe->winsys, + tex->pitch * tex->base.cpp * + tex->total_height, + PIPE_SURFACE_FLAG_TEXTURE); + } + + if (!tex->region) { + FREE(tex); + tex = NULL; + } + } + + *pt = &tex->base; +} + +void +i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct i915_texture *tex = (struct i915_texture *)*pt; + uint i; + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); + */ + + pipe->winsys->region_release(pipe->winsys, &tex->region); + + for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) + if (tex->image_offset[i]) + free(tex->image_offset[i]); + + free(tex); + } + *pt = NULL; +} diff --git a/src/mesa/pipe/i915simple/i915_texture.h b/src/mesa/pipe/i915simple/i915_texture.h new file mode 100644 index 0000000000..84a0502e81 --- /dev/null +++ b/src/mesa/pipe/i915simple/i915_texture.h @@ -0,0 +1,16 @@ + +#ifndef I915_TEXTURE_H +#define I915_TEXTURE_H + +struct pipe_context; +struct pipe_texture; + + +extern void +i915_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); + +extern void +i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); + + +#endif /* I915_TEXTURE_H */ diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index 2459d14cb8..fe32e7810d 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -194,7 +194,6 @@ void run_vertex_shader(float (*ainputs)[16][4], struct pipe_sampler_state; -struct pipe_mipmap_tree; struct softpipe_tile_cache; #define NUM_CHANNELS 4 /* R,G,B,A */ @@ -203,7 +202,6 @@ struct softpipe_tile_cache; struct tgsi_sampler { const struct pipe_sampler_state *state; - struct pipe_mipmap_tree *texture; /** Get samples for four fragments in a quad */ void (*get_samples)(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index e145b22f2f..5033209323 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -155,7 +155,7 @@ struct pipe_context { void (*set_texture_state)( struct pipe_context *, unsigned unit, - struct pipe_mipmap_tree * ); + struct pipe_texture * ); void (*set_viewport_state)( struct pipe_context *, const struct pipe_viewport_state * ); @@ -180,7 +180,7 @@ struct pipe_context { /** Get a surface which is a "view" into a texture */ struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, - struct pipe_mipmap_tree *texture, + struct pipe_texture *texture, unsigned face, unsigned level, unsigned zslice); @@ -237,8 +237,11 @@ struct pipe_context { /* * Texture functions */ - boolean (*mipmap_tree_layout)( struct pipe_context *pipe, - struct pipe_mipmap_tree *mt ); + void (*texture_create)(struct pipe_context *pipe, + struct pipe_texture **pt); + + void (*texture_release)(struct pipe_context *pipe, + struct pipe_texture **pt); /* Flush rendering: diff --git a/src/mesa/pipe/p_inlines.h b/src/mesa/pipe/p_inlines.h index 2418d016e1..c04d46dddd 100644 --- a/src/mesa/pipe/p_inlines.h +++ b/src/mesa/pipe/p_inlines.h @@ -80,4 +80,24 @@ pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) } } + +/** + * \sa pipe_region_reference + */ +static INLINE void +pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, + struct pipe_texture *pt) +{ + assert(ptr); + if (*ptr) { + pipe->texture_release(pipe, ptr); + assert(!*ptr); + } + if (pt) { + /* reference the new thing */ + pt->refcount++; + *ptr = pt; + } +} + #endif /* P_INLINES_H */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 642734aeb8..077a8f5a06 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -288,27 +288,11 @@ struct pipe_surface /** - * Describes the location of each texture image within a texture region. + * Texture. Represents one or several texture images on one or several mipmap + * levels. */ -struct pipe_mipmap_level -{ - unsigned level_offset; - unsigned width; - unsigned height; - unsigned depth; - unsigned nr_images; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset; /**< array [depth] of offsets */ -}; - -struct pipe_mipmap_tree -{ +struct pipe_texture +{ /* Effectively the key: */ unsigned target; /* XXX convert to PIPE_TEXTURE_x */ @@ -318,25 +302,13 @@ struct pipe_mipmap_tree unsigned first_level; unsigned last_level; - unsigned width0, height0, depth0; /**< Level zero image dimensions */ + unsigned width[PIPE_MAX_TEXTURE_LEVELS]; + unsigned height[PIPE_MAX_TEXTURE_LEVELS]; + unsigned depth[PIPE_MAX_TEXTURE_LEVELS]; unsigned cpp; unsigned compressed:1; - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - /* Includes image offset tables: - */ - struct pipe_mipmap_level level[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_region *region; - /* These are also refcounted: */ unsigned refcount; diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 59628531cc..9978884c9b 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -34,7 +34,7 @@ DRIVER_SOURCES = \ sp_state_rasterizer.c \ sp_state_surface.c \ sp_state_vertex.c \ - sp_tex_layout.c \ + sp_texture.c \ sp_tex_sample.c \ sp_tile_cache.c \ sp_surface.c diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index d5e68c189d..7a9fccce9a 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -40,7 +40,7 @@ #include "sp_state.h" #include "sp_surface.h" #include "sp_tile_cache.h" -#include "sp_tex_layout.h" +#include "sp_texture.h" #include "sp_winsys.h" @@ -98,9 +98,9 @@ softpipe_map_texture_surfaces(struct softpipe_context *sp) uint i; for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - struct pipe_mipmap_tree *mt = sp->texture[i]; - if (mt) { - pipe->region_map(pipe, mt->region); + struct softpipe_texture *spt = sp->texture[i]; + if (spt) { + pipe->region_map(pipe, spt->region); } } } @@ -146,9 +146,9 @@ softpipe_unmap_texture_surfaces(struct softpipe_context *sp) struct pipe_context *pipe = &sp->pipe; uint i; for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - struct pipe_mipmap_tree *mt = sp->texture[i]; - if (mt) { - pipe->region_unmap(pipe, mt->region); + struct softpipe_texture *spt = sp->texture[i]; + if (spt) { + pipe->region_unmap(pipe, spt->region); } } } @@ -351,7 +351,8 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.get_vendor = softpipe_get_vendor; /* textures */ - softpipe->pipe.mipmap_tree_layout = softpipe_mipmap_tree_layout; + softpipe->pipe.texture_create = softpipe_texture_create; + softpipe->pipe.texture_release = softpipe_texture_release; softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; /* diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 872766101d..d4763a98c6 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -89,7 +89,7 @@ struct softpipe_context { struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; + struct softpipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 24c8a44c47..7184fcda52 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -277,7 +277,7 @@ static void shade_begin(struct quad_stage *qs) /* set TGSI sampler state that varies */ for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { qss->samplers[i].state = softpipe->sampler[i]; - qss->samplers[i].texture = softpipe->texture[i]; + qss->samplers[i].texture = &softpipe->texture[i]->base; } #ifdef MESA_LLVM diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 2f096a9cc9..a543735b52 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -52,6 +52,35 @@ struct sp_fragment_shader_state { #endif }; +struct softpipe_texture +{ + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_region *region; +}; + void * softpipe_create_alpha_test_state(struct pipe_context *, const struct pipe_alpha_test_state *); @@ -125,7 +154,7 @@ void softpipe_set_scissor_state( struct pipe_context *, void softpipe_set_texture_state( struct pipe_context *, unsigned unit, - struct pipe_mipmap_tree * ); + struct pipe_texture * ); void softpipe_set_viewport_state( struct pipe_context *, const struct pipe_viewport_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 912f42d568..a360b4f02b 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -34,6 +34,8 @@ #include "pipe/draw/draw_context.h" #include "pipe/p_shader_tokens.h" #include "pipe/llvm/gallivm.h" +#include "pipe/tgsi/util/tgsi_dump.h" +#include "pipe/tgsi/exec/tgsi_sse2.h" void * softpipe_create_fs_state(struct pipe_context *pipe, diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 246a7d6eda..e71b9159e3 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -68,12 +68,12 @@ softpipe_delete_sampler_state(struct pipe_context *pipe, void softpipe_set_texture_state(struct pipe_context *pipe, unsigned unit, - struct pipe_mipmap_tree *texture) + struct pipe_texture *texture) { struct softpipe_context *softpipe = softpipe_context(pipe); assert(unit < PIPE_MAX_SAMPLERS); - softpipe->texture[unit] = texture; /* ptr, not struct */ + softpipe->texture[unit] = (struct softpipe_texture *)texture; /* ptr, not struct */ sp_tile_cache_set_texture(softpipe->tex_cache[unit], texture); diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index a6ab404603..c41bbc59b9 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -565,34 +565,35 @@ z24s8_get_tile(struct pipe_surface *ps, */ struct pipe_surface * softpipe_get_tex_surface(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt, + struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice) { + struct softpipe_texture *spt = (struct softpipe_texture *)pt; struct pipe_surface *ps; unsigned offset; /* in bytes */ - offset = mt->level[level].level_offset; + offset = spt->level_offset[level]; - if (mt->target == PIPE_TEXTURE_CUBE) { - offset += mt->level[level].image_offset[face] * mt->cpp; + if (pt->target == PIPE_TEXTURE_CUBE) { + offset += spt->image_offset[level][face] * pt->cpp; } - else if (mt->target == PIPE_TEXTURE_3D) { - offset += mt->level[level].image_offset[zslice] * mt->cpp; + else if (pt->target == PIPE_TEXTURE_3D) { + offset += spt->image_offset[level][zslice] * pt->cpp; } else { assert(face == 0); assert(zslice == 0); } - ps = pipe->winsys->surface_alloc(pipe->winsys, mt->format); + ps = pipe->winsys->surface_alloc(pipe->winsys, pt->format); if (ps) { assert(ps->format); assert(ps->refcount); - pipe_region_reference(&ps->region, mt->region); - ps->cpp = mt->cpp; - ps->width = mt->level[level].width; - ps->height = mt->level[level].height; - ps->pitch = mt->pitch; + pipe_region_reference(&ps->region, spt->region); + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = spt->pitch; ps->offset = offset; } return ps; diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index cf87e1a92c..b652e7598e 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -41,7 +41,7 @@ struct softpipe_tile_cache; extern struct pipe_surface * softpipe_get_tex_surface(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt, + struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice); diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.c b/src/mesa/pipe/softpipe/sp_tex_layout.c deleted file mode 100644 index 8156b00301..0000000000 --- a/src/mesa/pipe/softpipe/sp_tex_layout.c +++ /dev/null @@ -1,361 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Michel Dänzer - */ - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "sp_tex_layout.h" - - -/* At the moment, just make softpipe use the same layout for its - * textures as the i945. Softpipe needs some sort of texture layout, - * this one was handy. May be worthwhile to simplify this code a - * little. - */ - -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - -static int align(int value, int alignment) -{ - return (value + alignment - 1) & ~(alignment - 1); -} - - -static void -sp_miptree_set_level_info(struct pipe_mipmap_tree *mt, - unsigned level, - unsigned nr_images, - unsigned x, unsigned y, unsigned w, unsigned h, unsigned d) -{ - assert(level < PIPE_MAX_TEXTURE_LEVELS); - - mt->level[level].width = w; - mt->level[level].height = h; - mt->level[level].depth = d; - mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; - mt->level[level].nr_images = nr_images; - - /* - DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - level, w, h, d, x, y, mt->level[level].level_offset); - */ - - /* Not sure when this would happen, but anyway: - */ - if (mt->level[level].image_offset) { - FREE( mt->level[level].image_offset ); - mt->level[level].image_offset = NULL; - } - - assert(nr_images); - assert(!mt->level[level].image_offset); - - mt->level[level].image_offset = (unsigned *) MALLOC( nr_images * sizeof(unsigned) ); - mt->level[level].image_offset[0] = 0; -} - - -static void -sp_miptree_set_image_offset(struct pipe_mipmap_tree *mt, - unsigned level, unsigned img, unsigned x, unsigned y) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < mt->level[level].nr_images); - - mt->level[level].image_offset[img] = (x + y * mt->pitch); - - /* - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); - */ -} - - -static void -sp_miptree_layout_2d( struct pipe_mipmap_tree *mt ) -{ - int align_h = 2, align_w = 4; - unsigned level; - unsigned x = 0; - unsigned y = 0; - unsigned width = mt->width0; - unsigned height = mt->height0; - - mt->pitch = mt->width0; - /* XXX FIX THIS: - * we use alignment=64 bytes in sp_region_alloc(). If we change - * that, change this too. - */ - if (mt->pitch < 16) - mt->pitch = 16; - - /* May need to adjust pitch to accomodate the placement of - * the 2nd mipmap. This occurs when the alignment - * constraints of mipmap placement push the right edge of the - * 2nd mipmap out past the width of its parent. - */ - if (mt->first_level != mt->last_level) { - unsigned mip1_width = align(minify(mt->width0), align_w) - + minify(minify(mt->width0)); - - if (mip1_width > mt->width0) - mt->pitch = mip1_width; - } - - /* Pitch must be a whole number of dwords, even though we - * express it in texels. - */ - mt->pitch = align(mt->pitch * mt->cpp, 4) / mt->cpp; - mt->total_height = 0; - - for ( level = mt->first_level ; level <= mt->last_level ; level++ ) { - unsigned img_height; - - sp_miptree_set_level_info(mt, level, 1, x, y, width, height, 1); - - if (mt->compressed) - img_height = MAX2(1, height/4); - else - img_height = align(height, align_h); - - - /* Because the images are packed better, the final offset - * might not be the maximal one: - */ - mt->total_height = MAX2(mt->total_height, y + img_height); - - /* Layout_below: step right after second mipmap. - */ - if (level == mt->first_level + 1) { - x += align(width, align_w); - } - else { - y += img_height; - } - - width = minify(width); - height = minify(height); - } -} - - -static const int initial_offsets[6][2] = { - {0, 0}, - {0, 2}, - {1, 0}, - {1, 2}, - {1, 1}, - {1, 3} -}; - -static const int step_offsets[6][2] = { - {0, 2}, - {0, 2}, - {-1, 2}, - {-1, 2}, - {-1, 1}, - {-1, 1} -}; - - - -boolean -softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct pipe_mipmap_tree * mt) -{ - unsigned level; - - switch (mt->target) { - case PIPE_TEXTURE_CUBE:{ - const unsigned dim = mt->width0; - unsigned face; - unsigned lvlWidth = mt->width0, lvlHeight = mt->height0; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* Depending on the size of the largest images, pitch can be - * determined either by the old-style packing of cubemap faces, - * or the final row of 4x4, 2x2 and 1x1 faces below this. - */ - if (dim > 32) - mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; - else - mt->pitch = 14 * 8; - - mt->total_height = dim * 4 + 4; - - /* Set all the levels to effectively occupy the whole rectangular region. - */ - for (level = mt->first_level; level <= mt->last_level; level++) { - sp_miptree_set_level_info(mt, level, 6, - 0, 0, - lvlWidth, lvlHeight, 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - - for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * dim; - unsigned y = initial_offsets[face][1] * dim; - unsigned d = dim; - - if (dim == 4 && face >= 4) { - y = mt->total_height - 4; - x = (face - 4) * 8; - } - else if (dim < 4 && (face > 0 || mt->first_level > 0)) { - y = mt->total_height - 4; - x = face * 8; - } - - for (level = mt->first_level; level <= mt->last_level; level++) { - sp_miptree_set_image_offset(mt, level, face, x, y); - - d >>= 1; - - switch (d) { - case 4: - switch (face) { - case PIPE_TEX_FACE_POS_X: - case PIPE_TEX_FACE_NEG_X: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - case PIPE_TEX_FACE_POS_Y: - case PIPE_TEX_FACE_NEG_Y: - y += 12; - x -= 8; - break; - case PIPE_TEX_FACE_POS_Z: - case PIPE_TEX_FACE_NEG_Z: - y = mt->total_height - 4; - x = (face - 4) * 8; - break; - } - - case 2: - y = mt->total_height - 4; - x = 16 + face * 8; - break; - - case 1: - x += 48; - break; - - default: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - } - } - } - break; - } - case PIPE_TEXTURE_3D:{ - unsigned width = mt->width0; - unsigned height = mt->height0; - unsigned depth = mt->depth0; - unsigned pack_x_pitch, pack_x_nr; - unsigned pack_y_pitch; - unsigned level; - - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - mt->total_height = 0; - - pack_y_pitch = MAX2(mt->height0, 2); - pack_x_pitch = mt->pitch; - pack_x_nr = 1; - - for (level = mt->first_level; level <= mt->last_level; level++) { - unsigned nr_images = mt->target == PIPE_TEXTURE_3D ? depth : 6; - int x = 0; - int y = 0; - unsigned q, j; - - sp_miptree_set_level_info(mt, level, nr_images, - 0, mt->total_height, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - sp_miptree_set_image_offset(mt, level, q, x, y); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - mt->total_height += y; - - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= mt->pitch); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - } - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - break; - } - - case PIPE_TEXTURE_1D: - case PIPE_TEXTURE_2D: -// case PIPE_TEXTURE_RECTANGLE: - sp_miptree_layout_2d(mt); - break; - default: - assert(0); - break; - } - - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); - */ - - return TRUE; -} - diff --git a/src/mesa/pipe/softpipe/sp_tex_layout.h b/src/mesa/pipe/softpipe/sp_tex_layout.h deleted file mode 100644 index ea19c13b23..0000000000 --- a/src/mesa/pipe/softpipe/sp_tex_layout.h +++ /dev/null @@ -1,16 +0,0 @@ -#ifndef SP_TEX_LAYOUT_H -#define SP_TEX_LAYOUT_H - - -struct pipe_context; -struct pipe_mipmap_tree; - - -extern boolean -softpipe_mipmap_tree_layout(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt); - - -#endif /* SP_TEX_LAYOUT_H */ - - diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 92958400fc..9e48ed0cd2 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -422,7 +422,7 @@ compute_lambda(struct tgsi_sampler *sampler, dsdy = FABSF(dsdy); rho = MAX2(dsdx, dsdy); if (sampler->state->normalized_coords) - rho *= sampler->texture->width0; + rho *= sampler->texture->width[0]; } if (t) { float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; @@ -432,7 +432,7 @@ compute_lambda(struct tgsi_sampler *sampler, dtdy = FABSF(dtdy); max = MAX2(dtdx, dtdy); if (sampler->state->normalized_coords) - max *= sampler->texture->height0; + max *= sampler->texture->height[0]; rho = MAX2(rho, max); } if (p) { @@ -443,7 +443,7 @@ compute_lambda(struct tgsi_sampler *sampler, dpdy = FABSF(dpdy); max = MAX2(dpdx, dpdy); if (sampler->state->normalized_coords) - max *= sampler->texture->depth0; + max *= sampler->texture->depth[0]; rho = MAX2(rho, max); } @@ -620,8 +620,8 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, &level0, &level1, &levelBlend, &imgFilter); if (sampler->state->normalized_coords) { - width = sampler->texture->level[level0].width; - height = sampler->texture->level[level0].height; + width = sampler->texture->width[level0]; + height = sampler->texture->height[level0]; } else { width = height = 1; @@ -757,9 +757,9 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, &level0, &level1, &levelBlend, &imgFilter); if (sampler->state->normalized_coords) { - width = sampler->texture->level[level0].width; - height = sampler->texture->level[level0].height; - depth = sampler->texture->level[level0].depth; + width = sampler->texture->width[level0]; + height = sampler->texture->height[level0]; + depth = sampler->texture->depth[level0]; } else { width = height = depth = 1; @@ -883,7 +883,7 @@ sp_get_samples_cube(struct tgsi_sampler *sampler, /** * Called via tgsi_sampler::get_samples() * Use the sampler's state setting to get a filtered RGBA value - * from the sampler's texture (mipmap tree). + * from the sampler's texture. * * XXX we can implement many versions of this function, each * tightly coded for a specific combination of sampler state diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c new file mode 100644 index 0000000000..2f9a1e9837 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -0,0 +1,425 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_texture.h" + + +/* At the moment, just make softpipe use the same layout for its + * textures as the i945. Softpipe needs some sort of texture layout, + * this one was handy. May be worthwhile to simplify this code a + * little. + */ + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + +static int align(int value, int alignment) +{ + return (value + alignment - 1) & ~(alignment - 1); +} + + +static void +sp_miptree_set_level_info(struct softpipe_texture *spt, + unsigned level, + unsigned nr_images, + unsigned x, unsigned y, unsigned w, unsigned h, + unsigned d) +{ + struct pipe_texture *pt = &spt->base; + + assert(level < PIPE_MAX_TEXTURE_LEVELS); + + pt->width[level] = w; + pt->height[level] = h; + pt->depth[level] = d; + + spt->nr_images[level] = nr_images; + spt->level_offset[level] = (x + y * spt->pitch) * pt->cpp; + + /* + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, spt->level_offset[level]); + */ + + /* Not sure when this would happen, but anyway: + */ + if (spt->image_offset[level]) { + FREE( spt->image_offset[level] ); + spt->image_offset[level] = NULL; + } + + assert(nr_images); + assert(!spt->image_offset[level]); + + spt->image_offset[level] = (unsigned *) MALLOC( nr_images * sizeof(unsigned) ); + spt->image_offset[level][0] = 0; +} + + +static void +sp_miptree_set_image_offset(struct softpipe_texture *spt, + unsigned level, unsigned img, unsigned x, unsigned y) +{ + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + + assert(img < spt->nr_images[level]); + + spt->image_offset[level][img] = (x + y * spt->pitch); + + /* + DBG("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, spt->image_offset[level][img]); + */ +} + + +static void +sp_miptree_layout_2d( struct softpipe_texture *spt ) +{ + struct pipe_texture *pt = &spt->base; + int align_h = 2, align_w = 4; + unsigned level; + unsigned x = 0; + unsigned y = 0; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + + spt->pitch = pt->width[0]; + /* XXX FIX THIS: + * we use alignment=64 bytes in sp_region_alloc(). If we change + * that, change this too. + */ + if (spt->pitch < 16) + spt->pitch = 16; + + /* May need to adjust pitch to accomodate the placement of + * the 2nd mipmap. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap out past the width of its parent. + */ + if (pt->first_level != pt->last_level) { + unsigned mip1_width = align(minify(pt->width[0]), align_w) + + minify(minify(pt->width[0])); + + if (mip1_width > pt->width[0]) + spt->pitch = mip1_width; + } + + /* Pitch must be a whole number of dwords, even though we + * express it in texels. + */ + spt->pitch = align(spt->pitch * pt->cpp, 4) / pt->cpp; + spt->total_height = 0; + + for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + unsigned img_height; + + sp_miptree_set_level_info(spt, level, 1, x, y, width, height, 1); + + if (pt->compressed) + img_height = MAX2(1, height/4); + else + img_height = align(height, align_h); + + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + spt->total_height = MAX2(spt->total_height, y + img_height); + + /* Layout_below: step right after second mipmap. + */ + if (level == pt->first_level + 1) { + x += align(width, align_w); + } + else { + y += img_height; + } + + width = minify(width); + height = minify(height); + } +} + + +static const int initial_offsets[6][2] = { + {0, 0}, + {0, 2}, + {1, 0}, + {1, 2}, + {1, 1}, + {1, 3} +}; + +static const int step_offsets[6][2] = { + {0, 2}, + {0, 2}, + {-1, 2}, + {-1, 2}, + {-1, 1}, + {-1, 1} +}; + + + +static boolean +softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct softpipe_texture * spt) +{ + struct pipe_texture *pt = &spt->base; + unsigned level; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE:{ + const unsigned dim = pt->width[0]; + unsigned face; + unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* Depending on the size of the largest images, pitch can be + * determined either by the old-style packing of cubemap faces, + * or the final row of 4x4, 2x2 and 1x1 faces below this. + */ + if (dim > 32) + spt->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; + else + spt->pitch = 14 * 8; + + spt->total_height = dim * 4 + 4; + + /* Set all the levels to effectively occupy the whole rectangular region. + */ + for (level = pt->first_level; level <= pt->last_level; level++) { + sp_miptree_set_level_info(spt, level, 6, + 0, 0, + lvlWidth, lvlHeight, 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + + for (face = 0; face < 6; face++) { + unsigned x = initial_offsets[face][0] * dim; + unsigned y = initial_offsets[face][1] * dim; + unsigned d = dim; + + if (dim == 4 && face >= 4) { + y = spt->total_height - 4; + x = (face - 4) * 8; + } + else if (dim < 4 && (face > 0 || pt->first_level > 0)) { + y = spt->total_height - 4; + x = face * 8; + } + + for (level = pt->first_level; level <= pt->last_level; level++) { + sp_miptree_set_image_offset(spt, level, face, x, y); + + d >>= 1; + + switch (d) { + case 4: + switch (face) { + case PIPE_TEX_FACE_POS_X: + case PIPE_TEX_FACE_NEG_X: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + case PIPE_TEX_FACE_POS_Y: + case PIPE_TEX_FACE_NEG_Y: + y += 12; + x -= 8; + break; + case PIPE_TEX_FACE_POS_Z: + case PIPE_TEX_FACE_NEG_Z: + y = spt->total_height - 4; + x = (face - 4) * 8; + break; + } + + case 2: + y = spt->total_height - 4; + x = 16 + face * 8; + break; + + case 1: + x += 48; + break; + + default: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + } + } + } + break; + } + case PIPE_TEXTURE_3D:{ + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + unsigned pack_x_pitch, pack_x_nr; + unsigned pack_y_pitch; + unsigned level; + + spt->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; + spt->total_height = 0; + + pack_y_pitch = MAX2(pt->height[0], 2); + pack_x_pitch = spt->pitch; + pack_x_nr = 1; + + for (level = pt->first_level; level <= pt->last_level; level++) { + unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; + int x = 0; + int y = 0; + unsigned q, j; + + sp_miptree_set_level_info(spt, level, nr_images, + 0, spt->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + sp_miptree_set_image_offset(spt, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + spt->total_height += y; + + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= spt->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + } + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + break; + } + + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_2D: +// case PIPE_TEXTURE_RECTANGLE: + sp_miptree_layout_2d(spt); + break; + default: + assert(0); + break; + } + + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + spt->pitch, + spt->total_height, pt->cpp, spt->pitch * spt->total_height * pt->cpp); + */ + + return TRUE; +} + +void +softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) +{ + struct softpipe_texture *spt = REALLOC(*pt, sizeof(struct pipe_texture), + sizeof(struct softpipe_texture)); + + if (spt) { + memset(&spt->base + 1, 0, + sizeof(struct softpipe_texture) - sizeof(struct pipe_texture)); + + if (softpipe_mipmap_tree_layout(pipe, spt)) { + spt->region = pipe->winsys->region_alloc(pipe->winsys, + spt->pitch * (*pt)->cpp * + spt->total_height, + PIPE_SURFACE_FLAG_TEXTURE); + } + + if (!spt->region) { + FREE(spt); + spt = NULL; + } + } + + *pt = &spt->base; +} + +void +softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct softpipe_texture *spt = (struct softpipe_texture *)*pt; + uint i; + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); + */ + + pipe->winsys->region_release(pipe->winsys, &spt->region); + + for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) + if (spt->image_offset[i]) + free(spt->image_offset[i]); + + free(spt); + } + *pt = NULL; +} diff --git a/src/mesa/pipe/softpipe/sp_texture.h b/src/mesa/pipe/softpipe/sp_texture.h new file mode 100644 index 0000000000..2aca57bd1d --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_texture.h @@ -0,0 +1,18 @@ +#ifndef SP_TEXTURE_H +#define SP_TEXTURE_H + + +struct pipe_context; +struct pipe_texture; + + +extern void +softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); + +extern void +softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); + + +#endif /* SP_TEXTURE */ + + diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index ea0c8b8f91..62ee6a27c9 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -51,7 +51,7 @@ struct softpipe_tile_cache { struct pipe_surface *surface; /**< the surface we're caching */ - struct pipe_mipmap_tree *texture; /**< if caching a texture */ + struct pipe_texture *texture; /**< if caching a texture */ struct softpipe_cached_tile entries[NUM_ENTRIES]; uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; float clear_value[4]; @@ -139,7 +139,7 @@ sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) void sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, - struct pipe_mipmap_tree *texture) + struct pipe_texture *texture) { uint i; diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index e66fec2e20..9967aa5044 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -71,7 +71,7 @@ sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); extern void sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, - struct pipe_mipmap_tree *texture); + struct pipe_texture *texture); extern void sp_flush_tile_cache(struct softpipe_context *softpipe, diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 1d497e97fb..2c62b30f15 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -76,7 +76,7 @@ struct softpipe_tile_cache; /**< Opaque to TGSI */ struct tgsi_sampler { const struct pipe_sampler_state *state; - struct pipe_mipmap_tree *texture; + struct pipe_texture *texture; /** Get samples for four fragments in a quad */ void (*get_samples)(struct tgsi_sampler *sampler, const float s[QUAD_SIZE], diff --git a/src/mesa/sources b/src/mesa/sources index 2df60d1996..596d3b89d5 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -236,7 +236,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_framebuffer.c \ state_tracker/st_mesa_to_tgsi.c \ state_tracker/st_program.c \ - state_tracker/st_mipmap_tree.c + state_tracker/st_texture.c SHADER_SOURCES = \ shader/arbprogparse.c \ diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index f25cfd386a..c4e5af02d5 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -51,24 +51,24 @@ update_textures(struct st_context *st) for (u = 0; u < st->ctx->Const.MaxTextureImageUnits; u++) { struct gl_texture_object *texObj = st->ctx->Texture.Unit[u]._Current; - struct pipe_mipmap_tree *mt; + struct pipe_texture *pt; if (texObj) { GLboolean flush, retval; - retval = st_finalize_mipmap_tree(st->ctx, st->pipe, u, &flush); + retval = st_finalize_texture(st->ctx, st->pipe, u, &flush); #if 0 - printf("finalize_mipmap_tree returned %d, flush = %d\n", + printf("finalize_texture returned %d, flush = %d\n", retval, flush); #endif - mt = st_get_texobj_mipmap_tree(texObj); + pt = st_get_texobj_texture(texObj); } else { - mt = NULL; + pt = NULL; } - st->state.texture[u] = mt; - st->pipe->set_texture_state(st->pipe, u, mt); + st->state.texture[u] = pt; + st->pipe->set_texture_state(st->pipe, u, pt); } } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 39c0cf6b32..c28ad15b29 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -33,6 +33,7 @@ #include "main/imports.h" #include "main/image.h" #include "main/macros.h" +#include "main/texformat.h" #include "shader/program.h" #include "shader/prog_parameter.h" #include "shader/prog_print.h" @@ -50,6 +51,7 @@ #include "st_draw.h" #include "st_format.h" #include "st_mesa_to_tgsi.h" +#include "st_texture.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" @@ -440,63 +442,20 @@ _mesa_base_format(GLenum format) } - -static struct pipe_mipmap_tree * -alloc_mipmap_tree(struct st_context *st, - GLsizei width, GLsizei height, uint pipeFormat) -{ - const GLbitfield flags = PIPE_SURFACE_FLAG_TEXTURE; - struct pipe_mipmap_tree *mt; - GLuint cpp; - - mt = CALLOC_STRUCT(pipe_mipmap_tree); - if (!mt) - return NULL; - - cpp = st_sizeof_format(pipeFormat); - - mt->target = PIPE_TEXTURE_2D; - mt->internal_format = GL_RGBA; - mt->format = pipeFormat; - mt->first_level = 0; - mt->last_level = 0; - mt->width0 = width; - mt->height0 = height; - mt->depth0 = 1; - mt->cpp = cpp; - mt->compressed = 0; - mt->pitch = st->pipe->winsys->surface_pitch(st->pipe->winsys, cpp, width, - flags); - mt->region = st->pipe->winsys->region_alloc(st->pipe->winsys, - mt->pitch * cpp * height, flags); - mt->depth_pitch = 0; - mt->total_height = height; - mt->level[0].level_offset = 0; - mt->level[0].width = width; - mt->level[0].height = height; - mt->level[0].depth = 1; - mt->level[0].nr_images = 1; - mt->level[0].image_offset = NULL; - mt->refcount = 1; - - return mt; -} - - /** - * Make mipmap tree containing an image for glDrawPixels image. + * Make texture containing an image for glDrawPixels image. * If 'pixels' is NULL, leave the texture image data undefined. */ -static struct pipe_mipmap_tree * -make_mipmap_tree(struct st_context *st, - GLsizei width, GLsizei height, GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels) +static struct pipe_texture * +make_texture(struct st_context *st, + GLsizei width, GLsizei height, GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels) { GLcontext *ctx = st->ctx; struct pipe_context *pipe = st->pipe; const struct gl_texture_format *mformat; - struct pipe_mipmap_tree *mt; + struct pipe_texture *pt; GLuint pipeFormat, cpp; GLenum baseFormat; @@ -509,29 +468,33 @@ make_mipmap_tree(struct st_context *st, assert(pipeFormat); cpp = st_sizeof_format(pipeFormat); - mt = alloc_mipmap_tree(st, width, height, pipeFormat); - if (!mt) + pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, baseFormat, 0, 0, + width, height, 1, 0); + if (!pt) return NULL; if (unpack->BufferObj && unpack->BufferObj->Name) { /* - mt->region = buffer_object_region(unpack->BufferObj); + pt->region = buffer_object_region(unpack->BufferObj); */ printf("st_DrawPixels (sourcing from PBO not implemented yet)\n"); } { + struct pipe_surface *surface; static const GLuint dstImageOffsets = 0; GLboolean success; - GLuint pitch = mt->pitch; GLubyte *dest; const GLbitfield imageTransferStateSave = ctx->_ImageTransferState; /* we'll do pixel transfer in a fragment shader */ ctx->_ImageTransferState = 0x0; + surface = pipe->get_tex_surface(pipe, pt, 0, 0, 0); + /* map texture region */ - dest = pipe->region_map(pipe, mt->region); + (void) pipe->region_map(pipe, surface->region); + dest = surface->region->map + surface->offset; /* Put image into texture region. * Note that the image is actually going to be upside down in @@ -542,7 +505,7 @@ make_mipmap_tree(struct st_context *st, mformat, /* gl_texture_format */ dest, /* dest */ 0, 0, 0, /* dstX/Y/Zoffset */ - pitch * cpp, /* dstRowStride, bytes */ + surface->pitch * cpp, /* dstRowStride, bytes */ &dstImageOffsets, /* dstImageOffsets */ width, height, 1, /* size */ format, type, /* src format/type */ @@ -550,44 +513,15 @@ make_mipmap_tree(struct st_context *st, unpack); /* unmap */ - pipe->region_unmap(pipe, mt->region); + pipe->region_unmap(pipe, surface->region); + pipe_surface_reference(&surface, NULL); assert(success); /* restore */ ctx->_ImageTransferState = imageTransferStateSave; } -#if 0 - mt->target = PIPE_TEXTURE_2D; - mt->internal_format = GL_RGBA; - mt->format = pipeFormat; - mt->first_level = 0; - mt->last_level = 0; - mt->width0 = width; - mt->height0 = height; - mt->depth0 = 1; - mt->cpp = cpp; - mt->compressed = 0; - mt->pitch = mt->pitch; - mt->depth_pitch = 0; - mt->total_height = height; - mt->level[0].level_offset = 0; - mt->level[0].width = width; - mt->level[0].height = height; - mt->level[0].depth = 1; - mt->level[0].nr_images = 1; - mt->level[0].image_offset = NULL; - mt->refcount = 1; -#endif - return mt; -} - - -static void -free_mipmap_tree(struct pipe_context *pipe, struct pipe_mipmap_tree *mt) -{ - pipe->winsys->region_release(pipe->winsys, &mt->region); - free(mt); + return pt; } @@ -693,7 +627,7 @@ static void draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, GLsizei width, GLsizei height, GLfloat zoomX, GLfloat zoomY, - struct pipe_mipmap_tree *mt, + struct pipe_texture *pt, struct st_vertex_program *stvp, struct st_fragment_program *stfp, const GLfloat *color, @@ -761,9 +695,9 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->set_viewport_state(pipe, &vp); } - /* mipmap tree state: */ + /* texture state: */ { - pipe->set_texture_state(pipe, unit, mt); + pipe->set_texture_state(pipe, unit, pt); } /* Compute window coords (y=0=bottom) with pixel zoom. @@ -1025,14 +959,13 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, any_pixel_transfer_ops(st) || !compatible_formats(format, type, ps->format)) { /* textured quad */ - struct pipe_mipmap_tree *mt - = make_mipmap_tree(ctx->st, width, height, format, type, - unpack, pixels); - if (mt) { + struct pipe_texture *pt + = make_texture(ctx->st, width, height, format, type, unpack, pixels); + if (pt) { draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, - mt, stvp, stfp, color, GL_FALSE); - free_mipmap_tree(st->pipe, mt); + pt, stvp, stfp, color, GL_FALSE); + st->pipe->texture_release(st->pipe, &pt); } } else { @@ -1046,26 +979,29 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, /** * Create a texture which represents a bitmap image. */ -static struct pipe_mipmap_tree * +static struct pipe_texture * make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap) { struct pipe_context *pipe = ctx->st->pipe; - const uint flags = PIPE_SURFACE_FLAG_TEXTURE; + struct pipe_surface *surface; uint format = 0, cpp, comp; + GLenum internal_format; ubyte *dest; - struct pipe_mipmap_tree *mt; + struct pipe_texture *pt; int row, col; /* find a texture format we know */ if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8 )) { format = PIPE_FORMAT_U_I8; + internal_format = GL_INTENSITY8; cpp = 1; comp = 0; } else if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) { format = PIPE_FORMAT_U_A8_R8_G8_B8; + internal_format = GL_RGBA8; cpp = 4; comp = 3; /* alpha channel */ /*XXX little-endian dependency */ } @@ -1075,31 +1011,25 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, } /** - * Create a mipmap tree. + * Create a texture. */ - mt = CALLOC_STRUCT(pipe_mipmap_tree); - if (!mt) + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, internal_format, + 0, 0, width, height, 1, 0); + if (!pt) return NULL; if (unpack->BufferObj && unpack->BufferObj->Name) { /* - mt->region = buffer_object_region(unpack->BufferObj); + pt->region = buffer_object_region(unpack->BufferObj); */ printf("st_Bitmap (sourcing from PBO not implemented yet)\n"); } - - /* allocate texture region/storage */ - mt->pitch = pipe->winsys->surface_pitch(pipe->winsys, cpp, width, flags); - mt->region = pipe->winsys->region_alloc(pipe->winsys, - mt->pitch * cpp * height, flags); + surface = pipe->get_tex_surface(pipe, pt, 0, 0, 0); /* map texture region */ - dest = pipe->region_map(pipe, mt->region); - if (!dest) { - printf("st_Bitmap region_map() failed!?!"); - return NULL; - } + (void) pipe->region_map(pipe, surface->region); + dest = surface->region->map + surface->offset; /* Put image into texture region. * Note that the image is actually going to be upside down in @@ -1109,7 +1039,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, for (row = 0; row < height; row++) { const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); - ubyte *destRow = dest + row * mt->pitch * cpp; + ubyte *destRow = dest + row * surface->pitch * cpp; if (unpack->LsbFirst) { /* Lsb first */ @@ -1158,30 +1088,13 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, } /* row */ - /* unmap */ - pipe->region_unmap(pipe, mt->region); - - mt->target = PIPE_TEXTURE_2D; - mt->internal_format = GL_RGBA; - mt->format = format; - mt->first_level = 0; - mt->last_level = 0; - mt->width0 = width; - mt->height0 = height; - mt->depth0 = 1; - mt->cpp = cpp; - mt->compressed = 0; - mt->depth_pitch = 0; - mt->total_height = height; - mt->level[0].level_offset = 0; - mt->level[0].width = width; - mt->level[0].height = height; - mt->level[0].depth = 1; - mt->level[0].nr_images = 1; - mt->level[0].image_offset = NULL; - mt->refcount = 1; - - return mt; + /* Release surface */ + pipe->region_unmap(pipe, surface->region); + pipe_surface_reference(&surface, NULL); + + pt->format = format; + + return pt; } @@ -1193,21 +1106,21 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct st_fragment_program *stfp; struct st_vertex_program *stvp; struct st_context *st = ctx->st; - struct pipe_mipmap_tree *mt; + struct pipe_texture *pt; stvp = make_vertex_shader(ctx->st, GL_TRUE); stfp = combined_bitmap_fragment_program(ctx); st_validate_state(st); - mt = make_bitmap_texture(ctx, width, height, unpack, bitmap); - if (mt) { + pt = make_bitmap_texture(ctx, width, height, unpack, bitmap); + if (pt) { draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, 1.0, 1.0, - mt, stvp, stfp, + pt, stvp, stfp, ctx->Current.RasterColor, GL_FALSE); - free_mipmap_tree(st->pipe, mt); + st->pipe->texture_release(st->pipe, &pt); } } @@ -1296,7 +1209,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, struct st_fragment_program *stfp; struct pipe_surface *psRead; struct pipe_surface *psTex; - struct pipe_mipmap_tree *mt; + struct pipe_texture *pt; GLfloat *color; uint format; @@ -1327,11 +1240,12 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, psRead = rbRead->surface; format = psRead->format; - mt = alloc_mipmap_tree(ctx->st, width, height, format); - if (!mt) + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, + rbRead->Base.InternalFormat, 0, 0, width, height, 1, 0); + if (!pt) return; - psTex = pipe->get_tex_surface(pipe, mt, 0, 0, 0); + psTex = pipe->get_tex_surface(pipe, pt, 0, 0, 0); if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { srcy = ctx->DrawBuffer->Height - srcy - height; @@ -1368,10 +1282,10 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, /* draw textured quad */ draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, - mt, stvp, stfp, color, GL_TRUE); + pt, stvp, stfp, color, GL_TRUE); pipe_surface_reference(&psTex, NULL); - free_mipmap_tree(st->pipe, mt); + st->pipe->texture_release(st->pipe, &pt); } diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 80c92e8b7a..43681b7f8a 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -286,7 +286,7 @@ st_render_texture(GLcontext *ctx, struct st_renderbuffer *strb; struct gl_renderbuffer *rb; struct pipe_context *pipe = st->pipe; - struct pipe_mipmap_tree *mt; + struct pipe_texture *pt; assert(!att->Renderbuffer); @@ -302,26 +302,26 @@ st_render_texture(GLcontext *ctx, rb->AllocStorage = NULL; /* should not get called */ strb = st_renderbuffer(rb); - /* get the mipmap tree for the texture */ - mt = st_get_texobj_mipmap_tree(att->Texture); - assert(mt); - assert(mt->level[att->TextureLevel].width); + /* get the texture for the texture object */ + pt = st_get_texobj_texture(att->Texture); + assert(pt); + assert(pt->width[att->TextureLevel]); - rb->Width = mt->level[att->TextureLevel].width; - rb->Height = mt->level[att->TextureLevel].height; + rb->Width = pt->width[att->TextureLevel]; + rb->Height = pt->height[att->TextureLevel]; - /* the renderbuffer's surface is inside the mipmap_tree: */ - strb->surface = pipe->get_tex_surface(pipe, mt, + /* the renderbuffer's surface is inside the texture */ + strb->surface = pipe->get_tex_surface(pipe, pt, att->CubeMapFace, att->TextureLevel, att->Zoffset); assert(strb->surface); - init_renderbuffer_bits(strb, mt->format); + init_renderbuffer_bits(strb, pt->format); /* - printf("RENDER TO TEXTURE obj=%p mt=%p surf=%p %d x %d\n", - att->Texture, mt, strb->surface, rb->Width, rb->Height); + printf("RENDER TO TEXTURE obj=%p pt=%p surf=%p %d x %d\n", + att->Texture, pt, strb->surface, rb->Width, rb->Height); */ /* Invalidate buffer state so that the pipe's framebuffer state diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 461705119f..f45d7e4275 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -40,7 +40,7 @@ #include "state_tracker/st_cb_fbo.h" #include "state_tracker/st_cb_texture.h" #include "state_tracker/st_format.h" -#include "state_tracker/st_mipmap_tree.h" +#include "state_tracker/st_texture.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -54,8 +54,7 @@ struct st_texture_object { struct gl_texture_object base; /* The "parent" object */ - /* The mipmap tree must include at least these levels once - * validated: + /* The texture must include at least these levels once validated: */ GLuint firstLevel; GLuint lastLevel; @@ -67,7 +66,7 @@ struct st_texture_object /* On validation any active images held in main memory or in other * regions will be copied to this region and the old storage freed. */ - struct pipe_mipmap_tree *mt; + struct pipe_texture *pt; GLboolean imageOverride; GLint depthOverride; @@ -76,24 +75,6 @@ struct st_texture_object -struct st_texture_image -{ - struct gl_texture_image base; - - /* These aren't stored in gl_texture_image - */ - GLuint level; - GLuint face; - - /* If stImage->mt != NULL, image data is stored here. - * Else if stImage->base.Data != NULL, image is stored there. - * Else there is no image data. - */ - struct pipe_mipmap_tree *mt; -}; - - - static INLINE struct st_texture_object * st_texture_object(struct gl_texture_object *obj) @@ -108,11 +89,11 @@ st_texture_image(struct gl_texture_image *img) } -struct pipe_mipmap_tree * -st_get_texobj_mipmap_tree(struct gl_texture_object *texObj) +struct pipe_texture * +st_get_texobj_texture(struct gl_texture_object *texObj) { struct st_texture_object *stObj = st_texture_object(texObj); - return stObj->mt; + return stObj->pt; } @@ -174,9 +155,9 @@ st_IsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) struct st_texture_object *stObj = st_texture_object(texObj); return - stObj->mt && - stObj->mt->region && - intel_is_region_resident(intel, stObj->mt->region); + stObj->pt && + stObj->pt->region && + intel_is_region_resident(intel, stObj->pt->region); #endif return 1; } @@ -207,11 +188,10 @@ static void st_DeleteTextureObject(GLcontext *ctx, struct gl_texture_object *texObj) { - struct pipe_context *pipe = ctx->st->pipe; struct st_texture_object *stObj = st_texture_object(texObj); - if (stObj->mt) - st_miptree_release(pipe, &stObj->mt); + if (stObj->pt) + ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); _mesa_delete_texture_object(ctx, texObj); } @@ -220,13 +200,12 @@ st_DeleteTextureObject(GLcontext *ctx, static void st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) { - struct pipe_context *pipe = ctx->st->pipe; struct st_texture_image *stImage = st_texture_image(texImage); DBG("%s\n", __FUNCTION__); - if (stImage->mt) { - st_miptree_release(pipe, &stImage->mt); + if (stImage->pt) { + ctx->st->pipe->texture_release(ctx->st->pipe, &stImage->pt); } if (texImage->Data) { @@ -287,10 +266,10 @@ do_memcpy(void *dest, const void *src, size_t n) } -/* Functions to store texture images. Where possible, mipmap_tree's +/* Functions to store texture images. Where possible, textures * will be created or further instantiated with image data, otherwise * images will be stored in malloc'd memory. A validation step is - * required to pull those images into a mipmap tree, or otherwise + * required to pull those images into a texture, or otherwise * decide a fallback is required. */ @@ -313,17 +292,16 @@ logbase2(int n) /* Otherwise, store it in memory if (Border != 0) or (any dimension == * 1). * - * Otherwise, if max_level >= level >= min_level, create tree with - * space for textures from min_level down to max_level. + * Otherwise, if max_level >= level >= min_level, create texture with + * space for images from min_level down to max_level. * - * Otherwise, create tree with space for textures from (level - * 0)..(1x1). Consider pruning this tree at a validation if the - * saving is worth it. + * Otherwise, create texture with space for images from (level 0)..(1x1). + * Consider pruning this texture at a validation if the saving is worth it. */ static void -guess_and_alloc_mipmap_tree(struct pipe_context *pipe, - struct st_texture_object *stObj, - struct st_texture_image *stImage) +guess_and_alloc_texture(struct st_context *st, + struct st_texture_object *stObj, + struct st_texture_image *stImage) { GLuint firstLevel; GLuint lastLevel; @@ -382,23 +360,20 @@ guess_and_alloc_mipmap_tree(struct pipe_context *pipe, lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); } - assert(!stObj->mt); + assert(!stObj->pt); if (stImage->base.IsCompressed) comp_byte = compressed_num_bytes(stImage->base.TexFormat->MesaFormat); - stObj->mt = st_miptree_create(pipe, + stObj->pt = st_texture_create(st, gl_target_to_pipe(stObj->base.Target), - stImage->base.InternalFormat, + st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat), + stImage->base.InternalFormat, firstLevel, lastLevel, width, height, depth, - stImage->base.TexFormat->TexelBytes, comp_byte); - stObj->mt->format - = st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat); - DBG("%s - success\n", __FUNCTION__); } @@ -482,11 +457,11 @@ try_pbo_upload(GLcontext *ctx, else src_stride = width; - dst_offset = st_miptree_image_offset(stImage->mt, + dst_offset = st_texture_image_offset(stImage->pt, stImage->face, stImage->level); - dst_stride = stImage->mt->pitch; + dst_stride = stImage->pt->pitch; { struct _DriBufferObject *src_buffer = @@ -495,11 +470,11 @@ try_pbo_upload(GLcontext *ctx, /* Temporary hack: cast to _DriBufferObject: */ struct _DriBufferObject *dst_buffer = - (struct _DriBufferObject *)stImage->mt->region->buffer; + (struct _DriBufferObject *)stImage->pt->region->buffer; intelEmitCopyBlit(intel, - stImage->mt->cpp, + stImage->pt->cpp, src_stride, src_buffer, src_offset, dst_stride, dst_buffer, dst_offset, 0, 0, 0, 0, width, height, @@ -540,7 +515,6 @@ st_TexImage(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage, GLsizei imageSize, int compressed) { - struct pipe_context *pipe = ctx->st->pipe; struct st_texture_object *stObj = st_texture_object(texObj); struct st_texture_image *stImage = st_texture_image(texImage); GLint postConvWidth = width; @@ -589,58 +563,58 @@ st_TexImage(GLcontext * ctx, /* Release the reference to a potentially orphaned buffer. * Release any old malloced memory. */ - if (stImage->mt) { - st_miptree_release(pipe, &stImage->mt); + if (stImage->pt) { + ctx->st->pipe->texture_release(ctx->st->pipe, &stImage->pt); assert(!texImage->Data); } else if (texImage->Data) { _mesa_align_free(texImage->Data); } - /* If this is the only texture image in the tree, could call + /* If this is the only texture image in the texture, could call * bmBufferData with NULL data to free the old block and avoid * waiting on any outstanding fences. */ - if (stObj->mt && - stObj->mt->first_level == level && - stObj->mt->last_level == level && - stObj->mt->target != PIPE_TEXTURE_CUBE && - !st_miptree_match_image(stObj->mt, &stImage->base, + if (stObj->pt && + stObj->pt->first_level == level && + stObj->pt->last_level == level && + stObj->pt->target != PIPE_TEXTURE_CUBE && + !st_texture_match_image(stObj->pt, &stImage->base, stImage->face, stImage->level)) { DBG("release it\n"); - st_miptree_release(pipe, &stObj->mt); - assert(!stObj->mt); + ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); + assert(!stObj->pt); } - if (!stObj->mt) { - guess_and_alloc_mipmap_tree(pipe, stObj, stImage); - if (!stObj->mt) { - DBG("guess_and_alloc_mipmap_tree: failed\n"); + if (!stObj->pt) { + guess_and_alloc_texture(ctx->st, stObj, stImage); + if (!stObj->pt) { + DBG("guess_and_alloc_texture: failed\n"); } } - assert(!stImage->mt); + assert(!stImage->pt); - if (stObj->mt && - st_miptree_match_image(stObj->mt, &stImage->base, + if (stObj->pt && + st_texture_match_image(stObj->pt, &stImage->base, stImage->face, stImage->level)) { - st_miptree_reference(&stImage->mt, stObj->mt); - assert(stImage->mt); + pipe_texture_reference(ctx->st->pipe, &stImage->pt, stObj->pt); + assert(stImage->pt); } - if (!stImage->mt) - DBG("XXX: Image did not fit into tree - storing in local memory!\n"); + if (!stImage->pt) + DBG("XXX: Image did not fit into texture - storing in local memory!\n"); #if 0 /* XXX FIX when st_buffer_objects are in place */ /* PBO fastpaths: */ if (dims <= 2 && - stImage->mt && + stImage->pt && intel_buffer_object(unpack->BufferObj) && check_pbo_format(internalFormat, format, - type, stImage->base.TexFormat)) { + type, texImage->TexFormat)) { DBG("trying pbo upload\n"); @@ -650,9 +624,9 @@ st_TexImage(GLcontext * ctx, * performance (in particular when pipe_region_cow() is * required). */ - if (stObj->mt == stImage->mt && - stObj->mt->first_level == level && - stObj->mt->last_level == level) { + if (stObj->pt == stImage->pt && + stObj->pt->first_level == level && + stObj->pt->last_level == level) { if (try_pbo_zcopy(intel, stImage, unpack, internalFormat, @@ -683,7 +657,7 @@ st_TexImage(GLcontext * ctx, /* intelCopyTexImage calls this function with pixels == NULL, with - * the expectation that the mipmap tree will be set up but nothing + * the expectation that the texture will be set up but nothing * more will be done. This is where those calls return: */ if (compressed) { @@ -698,13 +672,9 @@ st_TexImage(GLcontext * ctx, if (!pixels) return; - if (stImage->mt) { - texImage->Data = st_miptree_image_map(pipe, - stImage->mt, - stImage->face, - stImage->level, - &dstRowStride, - stImage->base.ImageOffsets); + if (stImage->pt) { + texImage->Data = st_texture_image_map(ctx->st, stImage, 0); + dstRowStride = stImage->surface->pitch * stImage->surface->cpp; } else { /* Allocate regular memory and store the image there temporarily. */ @@ -732,22 +702,36 @@ st_TexImage(GLcontext * ctx, if (compressed) { memcpy(texImage->Data, pixels, imageSize); } - else if (!texImage->TexFormat->StoreImage(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, - 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, unpack)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + else { + GLuint srcImageStride = _mesa_image_image_stride(unpack, width, height, + format, type); + int i; + + for (i = 0; i++ < depth;) { + if (!texImage->TexFormat->StoreImage(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + 0, 0, 0, /* dstX/Y/Zoffset */ + dstRowStride, + texImage->ImageOffsets, + width, height, 1, + format, type, pixels, unpack)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + } + + if (stImage->pt && i < depth) { + st_texture_image_unmap(ctx->st, stImage); + texImage->Data = st_texture_image_map(ctx->st, stImage, i); + pixels += srcImageStride; + } + } } _mesa_unmap_teximage_pbo(ctx, unpack); - if (stImage->mt) { - st_miptree_image_unmap(pipe, stImage->mt); + if (stImage->pt) { + st_texture_image_unmap(ctx->st, stImage); texImage->Data = NULL; } @@ -839,49 +823,58 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, /* struct intel_context *intel = intel_context(ctx); */ - struct pipe_context *pipe = ctx->st->pipe; struct st_texture_image *stImage = st_texture_image(texImage); + GLuint dstImageStride = _mesa_image_image_stride(&ctx->Pack, texImage->Width, + texImage->Height, format, + type); + GLuint depth; + int i; /* Map */ - if (stImage->mt) { + if (stImage->pt) { /* Image is stored in hardware format in a buffer managed by the * kernel. Need to explicitly map and unmap it. */ - stImage->base.Data = - st_miptree_image_map(pipe, - stImage->mt, - stImage->face, - stImage->level, - &stImage->base.RowStride, - stImage->base.ImageOffsets); - stImage->base.RowStride /= stImage->mt->cpp; + texImage->Data = st_texture_image_map(ctx->st, stImage, 0); + texImage->RowStride = stImage->surface->pitch; } else { /* Otherwise, the image should actually be stored in - * stImage->base.Data. This is pretty confusing for + * texImage->Data. This is pretty confusing for * everybody, I'd much prefer to separate the two functions of * texImage->Data - storage for texture images in main memory * and access (ie mappings) of images. In other words, we'd * create a new texImage->Map field and leave Data simply for * storage. */ - assert(stImage->base.Data); + assert(texImage->Data); } + depth = texImage->Depth; + texImage->Depth = 1; - if (compressed) { - _mesa_get_compressed_teximage(ctx, target, level, pixels, - texObj, texImage); - } else { - _mesa_get_teximage(ctx, target, level, format, type, pixels, - texObj, texImage); + for (i = 0; i++ < depth;) { + if (compressed) { + _mesa_get_compressed_teximage(ctx, target, level, pixels, + texObj, texImage); + } else { + _mesa_get_teximage(ctx, target, level, format, type, pixels, + texObj, texImage); + } + + if (stImage->pt && i < depth) { + st_texture_image_unmap(ctx->st, stImage); + texImage->Data = st_texture_image_map(ctx->st, stImage, i); + pixels += dstImageStride; + } } - + + texImage->Depth = depth; /* Unmap */ - if (stImage->mt) { - st_miptree_image_unmap(pipe, stImage->mt); - stImage->base.Data = NULL; + if (stImage->pt) { + st_texture_image_unmap(ctx->st, stImage); + texImage->Data = NULL; } } @@ -921,9 +914,11 @@ st_TexSubimage(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - struct pipe_context *pipe = ctx->st->pipe; struct st_texture_image *stImage = st_texture_image(texImage); GLuint dstRowStride; + GLuint srcImageStride = _mesa_image_image_stride(packing, width, height, + format, type); + int i; DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), @@ -938,25 +933,28 @@ st_TexSubimage(GLcontext * ctx, /* Map buffer if necessary. Need to lock to prevent other contexts * from uploading the buffer under us. */ - if (stImage->mt) - texImage->Data = st_miptree_image_map(pipe, - stImage->mt, - stImage->face, - stImage->level, - &dstRowStride, - texImage->ImageOffsets); - - assert(dstRowStride); - - if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); + if (stImage->pt) { + texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset); + dstRowStride = stImage->surface->pitch * stImage->surface->cpp; + } + + for (i = 0; i++ < depth;) { + if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, + texImage->TexFormat, + texImage->Data, + xoffset, yoffset, 0, + dstRowStride, + texImage->ImageOffsets, + width, height, 1, + format, type, pixels, packing)) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); + } + + if (stImage->pt && i < depth) { + st_texture_image_unmap(ctx->st, stImage); + texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i); + pixels += srcImageStride; + } } #if 0 @@ -970,8 +968,8 @@ st_TexSubimage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, packing); - if (stImage->mt) { - st_miptree_image_unmap(pipe, stImage->mt); + if (stImage->pt) { + st_texture_image_unmap(ctx->st, stImage); texImage->Data = NULL; } } @@ -1074,7 +1072,7 @@ fallback_copy_texsubimage(GLcontext *ctx, { struct pipe_context *pipe = ctx->st->pipe; const uint face = texture_face(target); - struct pipe_mipmap_tree *mt = stImage->mt; + struct pipe_texture *pt = stImage->pt; struct pipe_surface *src_surf, *dest_surf; GLfloat *data; GLint row, yStep; @@ -1090,7 +1088,7 @@ fallback_copy_texsubimage(GLcontext *ctx, src_surf = strb->surface; - dest_surf = pipe->get_tex_surface(pipe, mt, + dest_surf = pipe->get_tex_surface(pipe, pt, face, level, destZ); (void) pipe->region_map(pipe, dest_surf->region); @@ -1150,7 +1148,7 @@ do_copy_texsubimage(GLcontext *ctx, struct gl_framebuffer *fb = ctx->ReadBuffer; struct st_renderbuffer *strb; struct pipe_context *pipe = ctx->st->pipe; - struct pipe_region *src_region, *dest_region; + struct pipe_surface *dest_surface; uint dest_format, src_format; (void) texImage; @@ -1169,29 +1167,24 @@ do_copy_texsubimage(GLcontext *ctx, assert(strb); assert(strb->surface); - assert(stImage->mt); + assert(stImage->pt); if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { srcY = strb->Base.Height - srcY - height; } src_format = strb->surface->format; - dest_format = stImage->mt->format; + dest_format = stImage->pt->format; - src_region = strb->surface->region; - dest_region = stImage->mt->region; + dest_surface = pipe->get_tex_surface(pipe, stImage->pt, stImage->face, + stImage->level, destZ); if (src_format == dest_format && ctx->_ImageTransferState == 0x0 && - src_region && - dest_region && - strb->surface->cpp == stImage->mt->cpp) { + strb->surface->region && + dest_surface->region && + strb->surface->cpp == stImage->pt->cpp) { /* do blit-style copy */ - struct pipe_surface *dest_surface = pipe->get_tex_surface(pipe, - stImage->mt, - stImage->face, - stImage->level, - destZ); /* XXX may need to invert image depending on window * vs. user-created FBO @@ -1203,12 +1196,12 @@ do_copy_texsubimage(GLcontext *ctx, * worth it: */ intelEmitCopyBlit(intel, - stImage->mt->cpp, + stImage->pt->cpp, -src->pitch, src->buffer, src->height * src->pitch * src->cpp, - stImage->mt->pitch, - stImage->mt->region->buffer, + stImage->pt->pitch, + stImage->pt->region->buffer, dest_offset, x, y + height, dstx, dsty, width, height, GL_COPY); /* ? */ @@ -1224,8 +1217,6 @@ do_copy_texsubimage(GLcontext *ctx, /* size */ width, height); #endif - - pipe_surface_reference(&dest_surface, NULL); } else { fallback_copy_texsubimage(ctx, target, level, @@ -1234,6 +1225,7 @@ do_copy_texsubimage(GLcontext *ctx, srcX, srcY, width, height); } + pipe_surface_reference(&dest_surface, NULL); #if 0 /* GL_SGIS_generate_mipmap -- this can be accelerated now. @@ -1267,7 +1259,7 @@ st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, goto fail; #endif - /* Setup or redefine the texture object, mipmap tree and texture + /* Setup or redefine the texture object, texture and texture * image. Don't populate yet. */ ctx->Driver.TexImage1D(ctx, target, level, internalFormat, @@ -1299,7 +1291,7 @@ st_CopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, goto fail; #endif - /* Setup or redefine the texture object, mipmap tree and texture + /* Setup or redefine the texture object, texture and texture * image. Don't populate yet. */ ctx->Driver.TexImage2D(ctx, target, level, internalFormat, @@ -1407,28 +1399,28 @@ calculate_first_last_level(struct st_texture_object *stObj) static void -copy_image_data_to_tree(struct pipe_context *pipe, - struct st_texture_object *stObj, - struct st_texture_image *stImage) +copy_image_data_to_texture(struct st_context *st, + struct st_texture_object *stObj, + struct st_texture_image *stImage) { - if (stImage->mt) { + if (stImage->pt) { /* Copy potentially with the blitter: */ - st_miptree_image_copy(pipe, - stObj->mt, /* dest miptree */ + st_texture_image_copy(st->pipe, + stObj->pt, /* dest texture */ stImage->face, stImage->level, - stImage->mt /* src miptree */ + stImage->pt /* src texture */ ); - st_miptree_release(pipe, &stImage->mt); + st->pipe->texture_release(st->pipe, &stImage->pt); } else { assert(stImage->base.Data != NULL); /* More straightforward upload. */ - st_miptree_image_data(pipe, - stObj->mt, + st_texture_image_data(st->pipe, + stObj->pt, stImage->face, stImage->level, stImage->base.Data, @@ -1439,16 +1431,16 @@ copy_image_data_to_tree(struct pipe_context *pipe, stImage->base.Data = NULL; } - st_miptree_reference(&stImage->mt, stObj->mt); + pipe_texture_reference(st->pipe, &stImage->pt, stObj->pt); } /* */ GLboolean -st_finalize_mipmap_tree(GLcontext *ctx, - struct pipe_context *pipe, GLuint unit, - GLboolean *needFlush) +st_finalize_texture(GLcontext *ctx, + struct pipe_context *pipe, GLuint unit, + GLboolean *needFlush) { struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; struct st_texture_object *stObj = st_texture_object(tObj); @@ -1465,7 +1457,7 @@ st_finalize_mipmap_tree(GLcontext *ctx, */ assert(stObj->base._Complete); - /* What levels must the tree include at a minimum? + /* What levels must the texture include at a minimum? */ calculate_first_last_level(stObj); firstImage = @@ -1474,27 +1466,27 @@ st_finalize_mipmap_tree(GLcontext *ctx, /* Fallback case: */ if (firstImage->base.Border) { - if (stObj->mt) { - st_miptree_release(pipe, &stObj->mt); + if (stObj->pt) { + ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); } return GL_FALSE; } - /* If both firstImage and stObj have a tree which can contain + /* If both firstImage and stObj point to a texture which can contain * all active images, favour firstImage. Note that because of the * completeness requirement, we know that the image dimensions * will match. */ - if (firstImage->mt && - firstImage->mt != stObj->mt && - firstImage->mt->first_level <= stObj->firstLevel && - firstImage->mt->last_level >= stObj->lastLevel) { + if (firstImage->pt && + firstImage->pt != stObj->pt && + firstImage->pt->first_level <= stObj->firstLevel && + firstImage->pt->last_level >= stObj->lastLevel) { - if (stObj->mt) - st_miptree_release(pipe, &stObj->mt); + if (stObj->pt) + ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); - st_miptree_reference(&stObj->mt, firstImage->mt); + pipe_texture_reference(ctx->st->pipe, &stObj->pt, firstImage->pt); } if (firstImage->base.IsCompressed) { @@ -1505,48 +1497,45 @@ st_finalize_mipmap_tree(GLcontext *ctx, cpp = firstImage->base.TexFormat->TexelBytes; } - /* Check tree can hold all active levels. Check tree matches + /* Check texture can hold all active levels. Check texture matches * target, imageFormat, etc. * * XXX: For some layouts (eg i945?), the test might have to be - * first_level == firstLevel, as the tree isn't valid except at the + * first_level == firstLevel, as the texture isn't valid except at the * original start level. Hope to get around this by * programming minLod, maxLod, baseLevel into the hardware and - * leaving the tree alone. + * leaving the texture alone. */ - if (stObj->mt && - (stObj->mt->target != gl_target_to_pipe(stObj->base.Target) || - stObj->mt->internal_format != firstImage->base.InternalFormat || - stObj->mt->first_level != stObj->firstLevel || - stObj->mt->last_level != stObj->lastLevel || - stObj->mt->width0 != firstImage->base.Width || - stObj->mt->height0 != firstImage->base.Height || - stObj->mt->depth0 != firstImage->base.Depth || - stObj->mt->cpp != cpp || - stObj->mt->compressed != firstImage->base.IsCompressed)) { - st_miptree_release(pipe, &stObj->mt); + if (stObj->pt && + (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || + stObj->pt->internal_format != firstImage->base.InternalFormat || + stObj->pt->first_level != stObj->firstLevel || + stObj->pt->last_level != stObj->lastLevel || + stObj->pt->width[0] != firstImage->base.Width || + stObj->pt->height[0] != firstImage->base.Height || + stObj->pt->depth[0] != firstImage->base.Depth || + stObj->pt->cpp != cpp || + stObj->pt->compressed != firstImage->base.IsCompressed)) { + ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); } - /* May need to create a new tree: + /* May need to create a new texture: */ - if (!stObj->mt) { - stObj->mt = st_miptree_create(pipe, + if (!stObj->pt) { + stObj->pt = st_texture_create(ctx->st, gl_target_to_pipe(stObj->base.Target), - firstImage->base.InternalFormat, + st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat), + firstImage->base.InternalFormat, stObj->firstLevel, stObj->lastLevel, firstImage->base.Width, firstImage->base.Height, firstImage->base.Depth, - cpp, comp_byte); - - stObj->mt->format - = st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat); } - /* Pull in any images not in the object's tree: + /* Pull in any images not in the object's texture: */ nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; for (face = 0; face < nr_faces; face++) { @@ -1554,10 +1543,10 @@ st_finalize_mipmap_tree(GLcontext *ctx, struct st_texture_image *stImage = st_texture_image(stObj->base.Image[face][i]); - /* Need to import images in main memory or held in other trees. + /* Need to import images in main memory or held in other textures. */ - if (stObj->mt != stImage->mt) { - copy_image_data_to_tree(pipe, stObj, stImage); + if (stObj->pt != stImage->pt) { + copy_image_data_to_texture(ctx->st, stObj, stImage); *needFlush = GL_TRUE; } } diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h index 7a1867dc58..7f8082b029 100644 --- a/src/mesa/state_tracker/st_cb_texture.h +++ b/src/mesa/state_tracker/st_cb_texture.h @@ -2,14 +2,14 @@ #define ST_CB_TEXTURE_H -extern struct pipe_mipmap_tree * -st_get_texobj_mipmap_tree(struct gl_texture_object *texObj); +extern struct pipe_texture * +st_get_texobj_texture(struct gl_texture_object *texObj); extern GLboolean -st_finalize_mipmap_tree(GLcontext *ctx, - struct pipe_context *pipe, GLuint unit, - GLboolean *needFlush); +st_finalize_texture(GLcontext *ctx, + struct pipe_context *pipe, GLuint unit, + GLboolean *needFlush); extern void diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index a6045230a0..db97014c5a 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -36,7 +36,6 @@ struct st_context; struct st_region; struct st_texture_object; -struct st_texture_image; struct st_fragment_program; struct draw_context; struct draw_stage; @@ -61,6 +60,25 @@ struct st_tracked_state { +struct st_texture_image +{ + struct gl_texture_image base; + + /* These aren't stored in gl_texture_image + */ + GLuint level; + GLuint face; + + /* If stImage->pt != NULL, image data is stored here. + * Else if stImage->base.Data != NULL, image is stored there. + * Else there is no image data. + */ + struct pipe_texture *pt; + + struct pipe_surface *surface; +}; + + struct st_context { @@ -91,7 +109,7 @@ struct st_context struct pipe_constant_buffer constants[2]; struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; - struct pipe_mipmap_tree *texture[PIPE_MAX_SAMPLERS]; + struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_viewport_state viewport; diff --git a/src/mesa/state_tracker/st_mipmap_tree.c b/src/mesa/state_tracker/st_mipmap_tree.c deleted file mode 100644 index 6ccf33105f..0000000000 --- a/src/mesa/state_tracker/st_mipmap_tree.c +++ /dev/null @@ -1,328 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "st_mipmap_tree.h" -#include "enums.h" - -#include "pipe/p_state.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" - - -#define DBG if(0) printf - -#if 0 -static GLenum -target_to_target(GLenum target) -{ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - return GL_TEXTURE_CUBE_MAP_ARB; - default: - return target; - } -} -#endif - -struct pipe_mipmap_tree * -st_miptree_create(struct pipe_context *pipe, - unsigned target, - GLenum internal_format, - GLuint first_level, - GLuint last_level, - GLuint width0, - GLuint height0, - GLuint depth0, GLuint cpp, GLuint compress_byte) -{ - GLboolean ok; - struct pipe_mipmap_tree *mt = calloc(sizeof(*mt), 1); - GLbitfield flags = 0x0; - - assert(target <= PIPE_TEXTURE_CUBE); - - DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), - _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); - - mt->target = target; - mt->internal_format = internal_format; - mt->first_level = first_level; - mt->last_level = last_level; - mt->width0 = width0; - mt->height0 = height0; - mt->depth0 = depth0; - mt->cpp = compress_byte ? compress_byte : cpp; - mt->compressed = compress_byte ? 1 : 0; - mt->refcount = 1; - - ok = pipe->mipmap_tree_layout(pipe, mt); - if (ok) { - mt->region = pipe->winsys->region_alloc(pipe->winsys, - mt->pitch * mt->cpp * - mt->total_height, flags); - } - - if (!mt->region) { - free(mt); - return NULL; - } - - return mt; -} - - -void -st_miptree_reference(struct pipe_mipmap_tree **dst, - struct pipe_mipmap_tree *src) -{ - src->refcount++; - *dst = src; - DBG("%s %p refcount now %d\n", __FUNCTION__, (void *) src, src->refcount); -} - -void -st_miptree_release(struct pipe_context *pipe, - struct pipe_mipmap_tree **mt) -{ - if (!*mt) - return; - - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *mt, (*mt)->refcount - 1); - if (--(*mt)->refcount <= 0) { - GLuint i; - - DBG("%s deleting %p\n", __FUNCTION__, (void *) *mt); - - pipe->winsys->region_release(pipe->winsys, &((*mt)->region)); - - for (i = 0; i < MAX_TEXTURE_LEVELS; i++) - if ((*mt)->level[i].image_offset) - free((*mt)->level[i].image_offset); - - free(*mt); - } - *mt = NULL; -} - - - - -/* Can the image be pulled into a unified mipmap tree. This mirrors - * the completeness test in a lot of ways. - * - * Not sure whether I want to pass gl_texture_image here. - */ -GLboolean -st_miptree_match_image(struct pipe_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level) -{ - /* Images with borders are never pulled into mipmap trees. - */ - if (image->Border) - return GL_FALSE; - - if (image->InternalFormat != mt->internal_format || - image->IsCompressed != mt->compressed) - return GL_FALSE; - - /* Test image dimensions against the base level image adjusted for - * minification. This will also catch images not present in the - * tree, changed targets, etc. - */ - if (image->Width != mt->level[level].width || - image->Height != mt->level[level].height || - image->Depth != mt->level[level].depth) - return GL_FALSE; - - return GL_TRUE; -} - - -/* Although we use the image_offset[] array to store relative offsets - * to cube faces, Mesa doesn't know anything about this and expects - * each cube face to be treated as a separate image. - * - * These functions present that view to mesa: - */ -const GLuint * -st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level) -{ - static const GLuint zero = 0; - - if (mt->target != PIPE_TEXTURE_3D || mt->level[level].nr_images == 1) - return &zero; - else - return mt->level[level].image_offset; -} - - -/** - * Return the offset to the given mipmap texture image within the - * texture memory buffer, in bytes. - */ -GLuint -st_miptree_image_offset(const struct pipe_mipmap_tree * mt, - GLuint face, GLuint level) -{ - if (mt->target == PIPE_TEXTURE_CUBE) - return (mt->level[level].level_offset + - mt->level[level].image_offset[face] * mt->cpp); - else - return mt->level[level].level_offset; -} - - -GLuint -st_miptree_texel_offset(const struct pipe_mipmap_tree * mt, - GLuint face, GLuint level, - GLuint col, GLuint row, GLuint img) -{ - GLuint imgOffset = st_miptree_image_offset(mt, face, level); - - return imgOffset + row * (mt->pitch + col) * mt->cpp; -} - - - -/** - * Map a teximage in a mipmap tree. - * \param row_stride returns row stride in bytes - * \param image_stride returns image stride in bytes (for 3D textures). - * \return address of mapping - */ -GLubyte * -st_miptree_image_map(struct pipe_context *pipe, - struct pipe_mipmap_tree * mt, - GLuint face, - GLuint level, - GLuint * row_stride, GLuint * image_offsets) -{ - GLubyte *ptr; - DBG("%s \n", __FUNCTION__); - - if (row_stride) - *row_stride = mt->pitch * mt->cpp; - - if (image_offsets) - memcpy(image_offsets, mt->level[level].image_offset, - mt->level[level].depth * sizeof(GLuint)); - - ptr = pipe->region_map(pipe, mt->region); - - return ptr + st_miptree_image_offset(mt, face, level); -} - -void -st_miptree_image_unmap(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt) -{ - DBG("%s\n", __FUNCTION__); - pipe->region_unmap(pipe, mt->region); -} - - - -/* Upload data for a particular image. - */ -void -st_miptree_image_data(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, - GLuint level, - void *src, - GLuint src_row_pitch, GLuint src_image_pitch) -{ - GLuint depth = dst->level[level].depth; - GLuint i; - GLuint height = 0; - const GLubyte *srcUB = src; - struct pipe_surface *dst_surface; - - DBG("%s\n", __FUNCTION__); - for (i = 0; i < depth; i++) { - height = dst->level[level].height; - if(dst->compressed) - height /= 4; - - dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); - - pipe->surface_data(pipe, dst_surface, - 0, 0, /* dstx, dsty */ - srcUB, - src_row_pitch, - 0, 0, /* source x, y */ - dst->level[level].width, height); /* width, height */ - - pipe_surface_reference(&dst_surface, NULL); - - srcUB += src_image_pitch * dst->cpp; - } -} - -/* Copy mipmap image between trees - */ -void -st_miptree_image_copy(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, GLuint level, - struct pipe_mipmap_tree *src) -{ - GLuint width = src->level[level].width; - GLuint height = src->level[level].height; - GLuint depth = src->level[level].depth; - struct pipe_surface *src_surface; - struct pipe_surface *dst_surface; - GLuint i; - - if (dst->compressed) - height /= 4; - for (i = 0; i < depth; i++) { - dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); - src_surface = pipe->get_tex_surface(pipe, src, face, level, i); - - pipe->surface_copy(pipe, - dst_surface, - 0, 0, /* destX, Y */ - src_surface, - 0, 0, /* srcX, Y */ - width, height); - - pipe_surface_reference(&dst_surface, NULL); - pipe_surface_reference(&src_surface, NULL); - } - -} diff --git a/src/mesa/state_tracker/st_mipmap_tree.h b/src/mesa/state_tracker/st_mipmap_tree.h deleted file mode 100644 index 3e7fd7fa0c..0000000000 --- a/src/mesa/state_tracker/st_mipmap_tree.h +++ /dev/null @@ -1,118 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef ST_MIPMAP_TREE_H -#define ST_MIPMAP_TREE_H - - -#include "main/mtypes.h" - -struct pipe_context; -struct pipe_mipmap_tree; -struct pipe_region; - - -extern struct pipe_mipmap_tree * -st_miptree_create(struct pipe_context *pipe, - GLenum target, - GLenum internal_format, - GLuint first_level, - GLuint last_level, - GLuint width0, - GLuint height0, - GLuint depth0, - GLuint cpp, - GLuint compress_byte); - -extern void -st_miptree_reference(struct pipe_mipmap_tree **dst, - struct pipe_mipmap_tree *src); - -extern void -st_miptree_release(struct pipe_context *pipe, struct pipe_mipmap_tree **mt); - - -/* Check if an image fits an existing mipmap tree layout - */ -extern GLboolean -st_miptree_match_image(struct pipe_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level); - -/* Return a pointer to an image within a tree. Return image stride as - * well. - */ -extern GLubyte * -st_miptree_image_map(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt, - GLuint face, GLuint level, - GLuint * row_stride, GLuint * image_stride); - -extern void -st_miptree_image_unmap(struct pipe_context *pipe, - struct pipe_mipmap_tree *mt); - - -/* Return pointers to each 2d slice within an image. Indexed by depth - * value. - */ -extern const GLuint * -st_miptree_depth_offsets(struct pipe_mipmap_tree *mt, GLuint level); - - -/* Return the linear offset of an image relative to the start of the - * tree: - */ -extern GLuint -st_miptree_image_offset(const struct pipe_mipmap_tree *mt, - GLuint face, GLuint level); - -extern GLuint -st_miptree_texel_offset(const struct pipe_mipmap_tree * mt, - GLuint face, GLuint level, - GLuint col, GLuint row, GLuint img); - - -/* Upload an image into a tree - */ -extern void -st_miptree_image_data(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, GLuint level, void *src, - GLuint src_row_pitch, GLuint src_image_pitch); - - -/* Copy an image between two trees - */ -extern void -st_miptree_image_copy(struct pipe_context *pipe, - struct pipe_mipmap_tree *dst, - GLuint face, GLuint level, - struct pipe_mipmap_tree *src); - - -#endif diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c new file mode 100644 index 0000000000..a5582c31c0 --- /dev/null +++ b/src/mesa/state_tracker/st_texture.c @@ -0,0 +1,276 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "st_context.h" +#include "st_format.h" +#include "st_texture.h" +#include "enums.h" + +#include "pipe/p_state.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" + + +#define DBG if(0) printf + +#if 0 +static GLenum +target_to_target(GLenum target) +{ + switch (target) { + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + return GL_TEXTURE_CUBE_MAP_ARB; + default: + return target; + } +} +#endif + +struct pipe_texture * +st_texture_create(struct st_context *st, + unsigned target, + unsigned format, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, + GLuint compress_byte) +{ + struct pipe_texture *pt = CALLOC_STRUCT(pipe_texture); + + assert(target <= PIPE_TEXTURE_CUBE); + + DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, + _mesa_lookup_enum_by_nr(target), + _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); + + if (!pt) + return NULL; + + assert(format); + + pt->target = target; + pt->format = format; + pt->internal_format = internal_format; + pt->first_level = first_level; + pt->last_level = last_level; + pt->width[0] = width0; + pt->height[0] = height0; + pt->depth[0] = depth0; + pt->compressed = compress_byte ? 1 : 0; + pt->cpp = pt->compressed ? compress_byte : st_sizeof_format(format); + pt->refcount = 1; + + st->pipe->texture_create(st->pipe, &pt); + + return pt; +} + + + + +/* Can the image be pulled into a unified mipmap texture. This mirrors + * the completeness test in a lot of ways. + * + * Not sure whether I want to pass gl_texture_image here. + */ +GLboolean +st_texture_match_image(struct pipe_texture *pt, + struct gl_texture_image *image, + GLuint face, GLuint level) +{ + /* Images with borders are never pulled into mipmap textures. + */ + if (image->Border) + return GL_FALSE; + + if (image->InternalFormat != pt->internal_format || + image->IsCompressed != pt->compressed) + return GL_FALSE; + + /* Test image dimensions against the base level image adjusted for + * minification. This will also catch images not present in the + * texture, changed targets, etc. + */ + if (image->Width != pt->width[level] || + image->Height != pt->height[level] || + image->Depth != pt->depth[level]) + return GL_FALSE; + + return GL_TRUE; +} + + +#if 000 +/* Although we use the image_offset[] array to store relative offsets + * to cube faces, Mesa doesn't know anything about this and expects + * each cube face to be treated as a separate image. + * + * These functions present that view to mesa: + */ +const GLuint * +st_texture_depth_offsets(struct pipe_texture *pt, GLuint level) +{ + static const GLuint zero = 0; + + if (pt->target != PIPE_TEXTURE_3D || pt->level[level].nr_images == 1) + return &zero; + else + return pt->level[level].image_offset; +} + + +/** + * Return the offset to the given mipmap texture image within the + * texture memory buffer, in bytes. + */ +GLuint +st_texture_image_offset(const struct pipe_texture * pt, + GLuint face, GLuint level) +{ + if (pt->target == PIPE_TEXTURE_CUBE) + return (pt->level[level].level_offset + + pt->level[level].image_offset[face] * pt->cpp); + else + return pt->level[level].level_offset; +} +#endif + + +/** + * Map a teximage in a mipmap texture. + * \param row_stride returns row stride in bytes + * \param image_stride returns image stride in bytes (for 3D textures). + * \return address of mapping + */ +GLubyte * +st_texture_image_map(struct st_context *st, struct st_texture_image *stImage, + GLuint zoffset) +{ + struct pipe_texture *pt = stImage->pt; + DBG("%s \n", __FUNCTION__); + + stImage->surface = st->pipe->get_tex_surface(st->pipe, pt, stImage->face, + stImage->level, zoffset); + + (void) st->pipe->region_map(st->pipe, stImage->surface->region); + + return stImage->surface->region->map + stImage->surface->offset; +} + +void +st_texture_image_unmap(struct st_context *st, struct st_texture_image *stImage) +{ + DBG("%s\n", __FUNCTION__); + + st->pipe->region_unmap(st->pipe, stImage->surface->region); + + pipe_surface_reference(&stImage->surface, NULL); +} + + + +/* Upload data for a particular image. + */ +void +st_texture_image_data(struct pipe_context *pipe, + struct pipe_texture *dst, + GLuint face, + GLuint level, + void *src, + GLuint src_row_pitch, GLuint src_image_pitch) +{ + GLuint depth = dst->depth[level]; + GLuint i; + GLuint height = 0; + const GLubyte *srcUB = src; + struct pipe_surface *dst_surface; + + DBG("%s\n", __FUNCTION__); + for (i = 0; i < depth; i++) { + height = dst->height[level]; + if(dst->compressed) + height /= 4; + + dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); + + pipe->surface_data(pipe, dst_surface, + 0, 0, /* dstx, dsty */ + srcUB, + src_row_pitch, + 0, 0, /* source x, y */ + dst->width[level], height); /* width, height */ + + pipe_surface_reference(&dst_surface, NULL); + + srcUB += src_image_pitch * dst->cpp; + } +} + +/* Copy mipmap image between textures + */ +void +st_texture_image_copy(struct pipe_context *pipe, + struct pipe_texture *dst, + GLuint face, GLuint level, + struct pipe_texture *src) +{ + GLuint width = src->width[level]; + GLuint height = src->height[level]; + GLuint depth = src->depth[level]; + struct pipe_surface *src_surface; + struct pipe_surface *dst_surface; + GLuint i; + + if (dst->compressed) + height /= 4; + for (i = 0; i < depth; i++) { + dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); + src_surface = pipe->get_tex_surface(pipe, src, face, level, i); + + pipe->surface_copy(pipe, + dst_surface, + 0, 0, /* destX, Y */ + src_surface, + 0, 0, /* srcX, Y */ + width, height); + + pipe_surface_reference(&dst_surface, NULL); + pipe_surface_reference(&src_surface, NULL); + } + +} diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h new file mode 100644 index 0000000000..b25e3f3f3b --- /dev/null +++ b/src/mesa/state_tracker/st_texture.h @@ -0,0 +1,109 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ST_TEXTURE_H +#define ST_TEXTURE_H + + +#include "main/mtypes.h" + +struct pipe_context; +struct pipe_texture; +struct pipe_region; + + +extern struct pipe_texture * +st_texture_create(struct st_context *st, + unsigned target, + unsigned format, + GLenum internal_format, + GLuint first_level, + GLuint last_level, + GLuint width0, + GLuint height0, + GLuint depth0, + GLuint compress_byte); + + +/* Check if an image fits an existing texture + */ +extern GLboolean +st_texture_match_image(struct pipe_texture *pt, + struct gl_texture_image *image, + GLuint face, GLuint level); + +/* Return a pointer to an image within a texture. Return image stride as + * well. + */ +extern GLubyte * +st_texture_image_map(struct st_context *st, + struct st_texture_image *stImage, + GLuint zoffset); + +extern void +st_texture_image_unmap(struct st_context *st, + struct st_texture_image *stImage); + + +/* Return pointers to each 2d slice within an image. Indexed by depth + * value. + */ +extern const GLuint * +st_texture_depth_offsets(struct pipe_texture *pt, GLuint level); + + +/* Return the linear offset of an image relative to the start of its region: + */ +extern GLuint +st_texture_image_offset(const struct pipe_texture *pt, + GLuint face, GLuint level); + +extern GLuint +st_texture_texel_offset(const struct pipe_texture * pt, + GLuint face, GLuint level, + GLuint col, GLuint row, GLuint img); + + +/* Upload an image into a texture + */ +extern void +st_texture_image_data(struct pipe_context *pipe, + struct pipe_texture *dst, + GLuint face, GLuint level, void *src, + GLuint src_row_pitch, GLuint src_image_pitch); + + +/* Copy an image between two textures + */ +extern void +st_texture_image_copy(struct pipe_context *pipe, + struct pipe_texture *dst, + GLuint face, GLuint level, + struct pipe_texture *src); + + +#endif -- cgit v1.2.3 From d492b388c96d98825ce9c6a9567ad83b3dee14b9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 6 Dec 2007 11:25:40 +0100 Subject: st_mesa_format_to_pipe_format: Handle MESA_FORMAT_ARGB4444. --- src/mesa/state_tracker/st_format.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 6d056dbe83..b6b9375967 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -256,6 +256,8 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) case MESA_FORMAT_ARGB8888_REV: case MESA_FORMAT_ARGB8888: return PIPE_FORMAT_U_A8_R8_G8_B8; + case MESA_FORMAT_ARGB4444: + return PIPE_FORMAT_U_A4_R4_G4_B4; case MESA_FORMAT_AL88: return PIPE_FORMAT_U_A8_L8; case MESA_FORMAT_A8: -- cgit v1.2.3 From 987d59bb83e9e08192563e5f1b52949c5511053c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 6 Dec 2007 11:30:34 +0100 Subject: Remove remnants of 'intel' from active state tracker code. --- src/mesa/state_tracker/st_cb_texture.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index f45d7e4275..acc97df2e7 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -656,7 +656,7 @@ st_TexImage(GLcontext * ctx, #endif - /* intelCopyTexImage calls this function with pixels == NULL, with + /* st_CopyTexImage calls this function with pixels == NULL, with * the expectation that the texture will be set up but nothing * more will be done. This is where those calls return: */ @@ -820,9 +820,6 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, struct gl_texture_object *texObj, struct gl_texture_image *texImage, int compressed) { - /* - struct intel_context *intel = intel_context(ctx); - */ struct st_texture_image *stImage = st_texture_image(texImage); GLuint dstImageStride = _mesa_image_image_stride(&ctx->Pack, texImage->Width, texImage->Height, format, @@ -947,7 +944,7 @@ st_TexSubimage(GLcontext * ctx, texImage->ImageOffsets, width, height, 1, format, type, pixels, packing)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); + _mesa_error(ctx, GL_OUT_OF_MEMORY, "st_TexSubImage"); } if (stImage->pt && i < depth) { -- cgit v1.2.3 From b859cdf6f191b4d8b56537c8dc30082a7e2d94b3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 7 Dec 2007 12:30:35 +0100 Subject: Eliminate struct pipe_region. Directly use struct pipe_buffer_handle for storage and struct pipe_surface for (un)mapping. --- .../drivers/dri/intel_winsys/intel_swapbuffers.c | 7 +- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 51 ++----------- src/mesa/pipe/failover/fo_context.c | 2 - src/mesa/pipe/i915simple/Makefile | 1 - src/mesa/pipe/i915simple/i915_context.c | 1 - src/mesa/pipe/i915simple/i915_context.h | 6 +- src/mesa/pipe/i915simple/i915_regions.c | 78 -------------------- src/mesa/pipe/i915simple/i915_state_emit.c | 6 +- src/mesa/pipe/i915simple/i915_surface.c | 35 ++++----- src/mesa/pipe/i915simple/i915_texture.c | 17 +++-- src/mesa/pipe/p_context.h | 10 +-- src/mesa/pipe/p_inlines.h | 53 +++++++------- src/mesa/pipe/p_state.h | 22 ++---- src/mesa/pipe/softpipe/Makefile | 1 - src/mesa/pipe/softpipe/sp_clear.c | 4 +- src/mesa/pipe/softpipe/sp_context.c | 84 ++++++---------------- src/mesa/pipe/softpipe/sp_draw_arrays.c | 2 - src/mesa/pipe/softpipe/sp_region.c | 79 -------------------- src/mesa/pipe/softpipe/sp_region.h | 40 ----------- src/mesa/pipe/softpipe/sp_state.h | 2 +- src/mesa/pipe/softpipe/sp_state_surface.c | 28 ++++---- src/mesa/pipe/softpipe/sp_surface.c | 61 ++++++++-------- src/mesa/pipe/softpipe/sp_texture.c | 18 +++-- src/mesa/pipe/softpipe/sp_tile_cache.c | 24 ++++++- src/mesa/pipe/softpipe/sp_tile_cache.h | 6 +- src/mesa/pipe/xlib/xm_api.c | 2 +- src/mesa/pipe/xlib/xm_buffer.c | 9 ++- src/mesa/pipe/xlib/xm_surface.c | 8 +-- src/mesa/pipe/xlib/xm_winsys.c | 50 +------------ src/mesa/state_tracker/st_cb_accum.c | 33 ++++----- src/mesa/state_tracker/st_cb_drawpixels.c | 48 ++++++------- src/mesa/state_tracker/st_cb_fbo.c | 28 ++++---- src/mesa/state_tracker/st_cb_readpixels.c | 13 ++-- src/mesa/state_tracker/st_cb_texture.c | 30 ++++---- src/mesa/state_tracker/st_context.c | 2 +- src/mesa/state_tracker/st_context.h | 5 +- src/mesa/state_tracker/st_texture.c | 9 ++- src/mesa/state_tracker/st_texture.h | 3 +- 38 files changed, 275 insertions(+), 603 deletions(-) delete mode 100644 src/mesa/pipe/i915simple/i915_regions.c delete mode 100644 src/mesa/pipe/softpipe/sp_region.c delete mode 100644 src/mesa/pipe/softpipe/sp_region.h (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index f96209d1b0..454cd71f6c 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -88,12 +88,11 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, const drm_clip_rect_t *pbox = dPriv->pClipRects; const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; const int cpp = intelScreen->front.cpp; - const struct pipe_region *srcRegion = surf->region; const int srcpitch = surf->pitch; int BR13, CMD; int i; - ASSERT(srcRegion); + ASSERT(surf->buffer); ASSERT(surf->cpp == cpp); DBG(SWAP, "screen pitch %d src surface pitch %d\n", @@ -159,7 +158,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, assert(box.x1 < box.x2); assert(box.y1 < box.y2); - /* XXX this could be done with pipe->region_copy() */ + /* XXX this could be done with pipe->surface_copy() */ BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); OUT_BATCH(CMD); OUT_BATCH(BR13); @@ -171,7 +170,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); OUT_BATCH((sbox.y1 << 16) | sbox.x1); OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(dri_bo(srcRegion->buffer), + OUT_RELOC(dri_bo(surf->buffer), DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 7b3aa99482..1b71d0ac10 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -135,13 +135,12 @@ static void intel_buffer_get_subdata(struct pipe_winsys *winsys, * for all buffers. */ static struct pipe_buffer_handle * -intel_buffer_create(struct pipe_winsys *winsys, - unsigned alignment) +intel_buffer_create(struct pipe_winsys *winsys, unsigned flags) { struct _DriBufferObject *buffer; struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); driGenBuffers( iws->regionPool, - "pipe buffer", 1, &buffer, alignment, 0, 0 ); + "pipe buffer", 1, &buffer, 64, 0, 0 ); return pipe_bo(buffer); } @@ -193,46 +192,6 @@ intel_i915_surface_pitch(struct pipe_winsys *winsys, } -static struct pipe_region * -intel_i915_region_alloc(struct pipe_winsys *winsys, - unsigned size, unsigned flags) -{ - struct pipe_region *region = calloc(sizeof(*region), 1); - const unsigned alignment = 64; - - region->refcount = 1; - - region->buffer = winsys->buffer_create( winsys, alignment ); - - winsys->buffer_data( winsys, - region->buffer, - size, - NULL, - PIPE_BUFFER_USAGE_PIXEL ); - - return region; -} - -static void -intel_i915_region_release(struct pipe_winsys *winsys, - struct pipe_region **region) -{ - if (!*region) - return; - - assert((*region)->refcount > 0); - (*region)->refcount--; - - if ((*region)->refcount == 0) { - assert((*region)->map_refcount == 0); - - winsys->buffer_reference( winsys, &((*region)->buffer), NULL ); - free(*region); - } - *region = NULL; -} - - static struct pipe_surface * intel_i915_surface_alloc(struct pipe_winsys *winsys, unsigned format) { @@ -252,8 +211,8 @@ intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) struct pipe_surface *surf = *s; surf->refcount--; if (surf->refcount == 0) { - if (surf->region) - winsys->region_release(winsys, &surf->region); + if (surf->buffer) + winsys->buffer_reference(winsys, &surf->buffer, NULL); free(surf); } *s = NULL; @@ -300,8 +259,6 @@ intel_create_pipe_winsys( int fd ) iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; iws->winsys.printf = intel_printf; iws->winsys.get_name = intel_get_name; - iws->winsys.region_alloc = intel_i915_region_alloc; - iws->winsys.region_release = intel_i915_region_release; iws->winsys.surface_pitch = intel_i915_surface_pitch; iws->winsys.surface_alloc = intel_i915_surface_alloc; iws->winsys.surface_release = intel_i915_surface_release; diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index a25563d451..a1291f583b 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -140,8 +140,6 @@ struct pipe_context *failover_create( struct pipe_context *hw, #endif failover->pipe.get_tex_surface = hw->get_tex_surface; - failover->pipe.region_map = hw->region_map; - failover->pipe.region_unmap = hw->region_unmap; failover->pipe.surface_data = hw->surface_data; failover->pipe.surface_copy = hw->surface_copy; failover->pipe.surface_fill = hw->surface_fill; diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile index 1223b386a3..2f91de3afc 100644 --- a/src/mesa/pipe/i915simple/Makefile +++ b/src/mesa/pipe/i915simple/Makefile @@ -12,7 +12,6 @@ DRIVER_SOURCES = \ i915_context.c \ i915_debug.c \ i915_debug_fp.c \ - i915_regions.c \ i915_state.c \ i915_state_immediate.c \ i915_state_dynamic.c \ diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 94649231cf..07b331c528 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -348,7 +348,6 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, draw_set_rasterize_stage(i915->draw, i915_draw_render_stage(i915)); } - i915_init_region_functions(i915); i915_init_surface_functions(i915); i915_init_state_functions(i915); i915_init_flush_functions(i915); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 8ed3465be2..dbf0c885cc 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -175,7 +175,7 @@ struct i915_texture { /* The data is held here: */ - struct pipe_region *region; + struct pipe_buffer_handle *buffer; }; struct i915_context @@ -289,10 +289,10 @@ void i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, /*********************************************************************** - * i915_region.c: + * i915_surface.c: */ -void i915_init_region_functions( struct i915_context *i915 ); void i915_init_surface_functions( struct i915_context *i915 ); + void i915_init_state_functions( struct i915_context *i915 ); void i915_init_flush_functions( struct i915_context *i915 ); void i915_init_string_functions( struct i915_context *i915 ); diff --git a/src/mesa/pipe/i915simple/i915_regions.c b/src/mesa/pipe/i915simple/i915_regions.c deleted file mode 100644 index 82fdec83d0..0000000000 --- a/src/mesa/pipe/i915simple/i915_regions.c +++ /dev/null @@ -1,78 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Provide additional functionality on top of bufmgr buffers: - * - 2d semantics and blit operations (XXX: remove/simplify blits??) - * - refcounting of buffers for multiple images in a buffer. - * - refcounting of buffer mappings. - */ - -#include "pipe/p_defines.h" -#include "pipe/p_winsys.h" -#include "i915_context.h" - - - - -static ubyte * -i915_region_map(struct pipe_context *pipe, struct pipe_region *region) -{ - struct i915_context *i915 = i915_context( pipe ); - - if (!region->map_refcount++) { - region->map = i915->pipe.winsys->buffer_map( i915->pipe.winsys, - region->buffer, - PIPE_BUFFER_FLAG_WRITE | - PIPE_BUFFER_FLAG_READ); - } - - return region->map; -} - -static void -i915_region_unmap(struct pipe_context *pipe, struct pipe_region *region) -{ - struct i915_context *i915 = i915_context( pipe ); - - if (region->map_refcount > 0) { - assert(region->map); - if (!--region->map_refcount) { - i915->pipe.winsys->buffer_unmap( i915->pipe.winsys, - region->buffer ); - region->map = NULL; - } - } -} - - -void -i915_init_region_functions(struct i915_context *i915) -{ - i915->pipe.region_map = i915_region_map; - i915->pipe.region_unmap = i915_region_unmap; -} - diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 900a91d896..eda40d7c76 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -220,7 +220,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) BUF_3D_PITCH(pitch) | /* pitch in bytes */ BUF_3D_USE_FENCE); - OUT_RELOC(cbuf_surface->region->buffer, + OUT_RELOC(cbuf_surface->buffer, I915_BUFFER_ACCESS_WRITE, 0); } @@ -236,7 +236,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) BUF_3D_PITCH(zpitch) | /* pitch in bytes */ BUF_3D_USE_FENCE); - OUT_RELOC(depth_surface->region->buffer, + OUT_RELOC(depth_surface->buffer, I915_BUFFER_ACCESS_WRITE, 0); } @@ -284,7 +284,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) for (unit = 0; unit < I915_TEX_UNITS; unit++) { if (enabled & (1 << unit)) { struct pipe_buffer_handle *buf = - i915->texture[unit]->region->buffer; + i915->texture[unit]->buffer; uint offset = 0; assert(buf); diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 385202507d..ecbabe50eb 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -29,6 +29,7 @@ #include "i915_blit.h" #include "i915_state.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" @@ -57,7 +58,7 @@ i915_get_tile_rgba(struct pipe_context *pipe, uint x, uint y, uint w, uint h, float *p) { const unsigned *src - = ((const unsigned *) (ps->region->map + ps->offset)) + = ((const unsigned *) (ps->map)) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -129,7 +130,7 @@ i915_get_tile(struct pipe_context *pipe, ubyte *pDest; uint i; - assert(ps->region->map); + assert(ps->map); CLIP_TILE; @@ -137,7 +138,7 @@ i915_get_tile(struct pipe_context *pipe, dst_stride = w0 * cpp; } - pSrc = ps->region->map + ps->offset + (y * ps->pitch + x) * cpp; + pSrc = ps->map + (y * ps->pitch + x) * cpp; pDest = (ubyte *) p; for (i = 0; i < h; i++) { @@ -163,7 +164,7 @@ i915_put_tile(struct pipe_context *pipe, ubyte *pDest; uint i; - assert(ps->region->map); + assert(ps->map); CLIP_TILE; @@ -172,7 +173,7 @@ i915_put_tile(struct pipe_context *pipe, } pSrc = (const ubyte *) p; - pDest = ps->region->map + ps->offset + (y * ps->pitch + x) * cpp; + pDest = ps->map + (y * ps->pitch + x) * cpp; for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w0 * cpp); @@ -211,7 +212,7 @@ i915_get_tex_surface(struct pipe_context *pipe, if (ps) { assert(ps->format); assert(ps->refcount); - pipe_region_reference(&ps->region, tex->region); + pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); ps->cpp = pt->cpp; ps->width = pt->width[level]; ps->height = pt->height[level]; @@ -274,12 +275,12 @@ i915_surface_data(struct pipe_context *pipe, const void *src, unsigned src_pitch, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - _mesa_copy_rect(pipe->region_map(pipe, dst->region) + dst->offset, + _mesa_copy_rect(pipe_surface_map(dst), dst->cpp, dst->pitch, dstx, dsty, width, height, src, src_pitch, srcx, srcy); - pipe->region_unmap(pipe, dst->region); + pipe_surface_unmap(dst); } @@ -297,23 +298,23 @@ i915_surface_copy(struct pipe_context *pipe, assert( dst->cpp == src->cpp ); if (0) { - _mesa_copy_rect(pipe->region_map(pipe, dst->region) + dst->offset, + _mesa_copy_rect(pipe_surface_map(dst), dst->cpp, dst->pitch, dstx, dsty, width, height, - pipe->region_map(pipe, src->region) + src->offset, + pipe_surface_map(src), src->pitch, srcx, srcy); - pipe->region_unmap(pipe, src->region); - pipe->region_unmap(pipe, dst->region); + pipe_surface_unmap(src); + pipe_surface_unmap(dst); } else { i915_copy_blit( i915_context(pipe), dst->cpp, - (short) src->pitch, src->region->buffer, src->offset, - (short) dst->pitch, dst->region->buffer, dst->offset, + (short) src->pitch, src->buffer, src->offset, + (short) dst->pitch, dst->buffer, dst->offset, (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); } } @@ -324,7 +325,7 @@ i915_surface_copy(struct pipe_context *pipe, static ubyte * get_pointer(struct pipe_surface *dst, unsigned x, unsigned y) { - return dst->region->map + (y * dst->pitch + x) * dst->cpp; + return dst->map + (y * dst->pitch + x) * dst->cpp; } @@ -337,7 +338,7 @@ i915_surface_fill(struct pipe_context *pipe, if (0) { unsigned i, j; - (void)pipe->region_map(pipe, dst->region); + (void)pipe_surface_map(dst); switch (dst->cpp) { case 1: { @@ -375,7 +376,7 @@ i915_surface_fill(struct pipe_context *pipe, i915_fill_blit( i915_context(pipe), dst->cpp, (short) dst->pitch, - dst->region->buffer, dst->offset, + dst->buffer, dst->offset, (short) dstx, (short) dsty, (short) width, (short) height, value ); diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c index 3bfa806d9e..1ca2815dfb 100644 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -33,6 +33,7 @@ #include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" #include "pipe/p_util.h" #include "pipe/p_winsys.h" @@ -494,13 +495,17 @@ i915_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : i915_miptree_layout(pipe, tex)) { - tex->region = pipe->winsys->region_alloc(pipe->winsys, - tex->pitch * tex->base.cpp * - tex->total_height, - PIPE_SURFACE_FLAG_TEXTURE); + tex->buffer = pipe->winsys->buffer_create(pipe->winsys, + PIPE_SURFACE_FLAG_TEXTURE); + + if (tex->buffer) + pipe->winsys->buffer_data(pipe->winsys, tex->buffer, + tex->pitch * tex->base.cpp * + tex->total_height, NULL, + PIPE_BUFFER_USAGE_PIXEL); } - if (!tex->region) { + if (!tex->buffer) { FREE(tex); tex = NULL; } @@ -527,7 +532,7 @@ i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); */ - pipe->winsys->region_release(pipe->winsys, &tex->region); + pipe->winsys->buffer_reference(pipe->winsys, &tex->buffer, NULL); for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) if (tex->image_offset[i]) diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 5033209323..2420d02213 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -37,7 +37,7 @@ struct pipe_state_cache; * Gallium rendering context. Basically: * - state setting functions * - VBO drawing functions - * - memory region function + * - surface functions * - device queries */ struct pipe_context { @@ -201,14 +201,6 @@ struct pipe_context { uint x, uint y, uint w, uint h, const float *p); - /* - * Memory region functions - */ - ubyte *(*region_map)(struct pipe_context *pipe, struct pipe_region *r); - - void (*region_unmap)(struct pipe_context *pipe, struct pipe_region *r); - - /* * Surface functions */ diff --git a/src/mesa/pipe/p_inlines.h b/src/mesa/pipe/p_inlines.h index c04d46dddd..1697284b04 100644 --- a/src/mesa/pipe/p_inlines.h +++ b/src/mesa/pipe/p_inlines.h @@ -29,40 +29,42 @@ #define P_INLINES_H #include "p_context.h" +#include "p_defines.h" #include "p_winsys.h" -/** - * Set 'ptr' to point to 'region' and update reference counting. - * The old thing pointed to, if any, will be unreferenced first. - * 'region' may be NULL. - */ +static INLINE ubyte * +pipe_surface_map(struct pipe_surface *surface) +{ + if (!surface->map_refcount++) { + surface->map = surface->winsys->buffer_map( surface->winsys, + surface->buffer, + PIPE_BUFFER_FLAG_WRITE | + PIPE_BUFFER_FLAG_READ ) + + surface->offset; + } + + return surface->map; +} + static INLINE void -pipe_region_reference(struct pipe_region **ptr, struct pipe_region *region) +pipe_surface_unmap(struct pipe_surface *surface) { - assert(ptr); - if (*ptr) { - /* unreference the old thing */ - struct pipe_region *oldReg = *ptr; - assert(oldReg->refcount > 0); - oldReg->refcount--; - if (oldReg->refcount == 0) { - /* free the old region */ - assert(oldReg->map_refcount == 0); - /* XXX dereference the region->buffer */ - FREE( oldReg ); + if (surface->map_refcount > 0) { + assert(surface->map); + if (!--surface->map_refcount) { + surface->winsys->buffer_unmap( surface->winsys, + surface->buffer ); + surface->map = NULL; } - *ptr = NULL; - } - if (region) { - /* reference the new thing */ - region->refcount++; - *ptr = region; } } + /** - * \sa pipe_region_reference + * Set 'ptr' to point to 'surf' and update reference counting. + * The old thing pointed to, if any, will be unreferenced first. + * 'surf' may be NULL. */ static INLINE void pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) @@ -82,7 +84,7 @@ pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) /** - * \sa pipe_region_reference + * \sa pipe_surface_reference */ static INLINE void pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, @@ -100,4 +102,5 @@ pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, } } + #endif /* P_INLINES_H */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 077a8f5a06..44dec9b773 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -256,32 +256,20 @@ struct pipe_sampler_state }; -/*** - *** Resource Objects - ***/ - -struct pipe_region -{ - struct pipe_buffer_handle *buffer; /**< driver private buffer handle */ - - unsigned refcount; /**< Reference count for region */ - ubyte *map; /**< only non-NULL when region is actually mapped */ - unsigned map_refcount; /**< Reference count for mapping */ -}; - - /** - * 2D surface. This is basically a view into a pipe_region (memory buffer). + * 2D surface. This is basically a view into a memory buffer. * May be a renderbuffer, texture mipmap level, etc. */ struct pipe_surface { - struct pipe_region *region; + struct pipe_buffer_handle *buffer; /**< driver private buffer handle */ + ubyte *map; /**< only non-NULL when surface is actually mapped */ + unsigned map_refcount; /**< Reference count for mapping */ unsigned format; /**< PIPE_FORMAT_x */ unsigned cpp; /**< bytes per pixel */ unsigned width, height; unsigned pitch; /**< in pixels */ - unsigned offset; /**< offset from start of region, in bytes */ + unsigned offset; /**< offset from start of buffer, in bytes */ unsigned refcount; struct pipe_winsys *winsys; /**< winsys which owns/created the surface */ }; diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 9978884c9b..647cc05373 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -24,7 +24,6 @@ DRIVER_SOURCES = \ sp_quad_output.c \ sp_quad_stencil.c \ sp_quad_stipple.c \ - sp_region.c \ sp_state_blend.c \ sp_state_clip.c \ sp_state_derived.c \ diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index a4276362b9..496b38fd5f 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -50,12 +50,12 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, softpipe_update_derived(softpipe); /* not needed?? */ - if (ps == sp_tile_cache_get_surface(softpipe->zbuf_cache)) { + if (ps == sp_tile_cache_get_surface(softpipe, softpipe->zbuf_cache)) { float clear[4]; clear[0] = 1.0; /* XXX hack */ sp_tile_cache_clear(softpipe->zbuf_cache, clear); } - else if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[0])) { + else if (ps == sp_tile_cache_get_surface(softpipe, softpipe->cbuf_cache[0])) { float clear[4]; clear[0] = 0.2f; /* XXX hack */ clear[1] = 0.2f; /* XXX hack */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 7a9fccce9a..f7f0316cf2 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -31,12 +31,12 @@ #include "pipe/draw/draw_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" #include "pipe/p_util.h" #include "sp_clear.h" #include "sp_context.h" #include "sp_flush.h" #include "sp_prim_setup.h" -#include "sp_region.h" #include "sp_state.h" #include "sp_surface.h" #include "sp_tile_cache.h" @@ -65,44 +65,22 @@ softpipe_is_format_supported( struct pipe_context *pipe, uint format ) void softpipe_map_surfaces(struct softpipe_context *sp) { - struct pipe_context *pipe = &sp->pipe; + struct pipe_surface *ps; unsigned i; for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - struct pipe_surface *ps = sp->framebuffer.cbufs[i]; - if (ps->region && !ps->region->map) { - pipe->region_map(pipe, ps->region); - } + ps = sp->framebuffer.cbufs[i]; + if (ps->buffer) + pipe_surface_map(ps); } - if (sp->framebuffer.zbuf) { - struct pipe_surface *ps = sp->framebuffer.zbuf; - if (ps->region && !ps->region->map) { - pipe->region_map(pipe, ps->region); - } - } - - if (sp->framebuffer.sbuf) { - struct pipe_surface *ps = sp->framebuffer.sbuf; - if (ps->region && !ps->region->map) { - pipe->region_map(pipe, ps->region); - } - } -} - + ps = sp->framebuffer.zbuf; + if (ps && ps->buffer) + pipe_surface_map(ps); -void -softpipe_map_texture_surfaces(struct softpipe_context *sp) -{ - struct pipe_context *pipe = &sp->pipe; - uint i; - - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - struct softpipe_texture *spt = sp->texture[i]; - if (spt) { - pipe->region_map(pipe, spt->region); - } - } + ps = sp->framebuffer.sbuf; + if (ps && ps->buffer) + pipe_surface_map(ps); } @@ -112,7 +90,7 @@ softpipe_map_texture_surfaces(struct softpipe_context *sp) void softpipe_unmap_surfaces(struct softpipe_context *sp) { - struct pipe_context *pipe = &sp->pipe; + struct pipe_surface *ps; uint i; for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) @@ -121,36 +99,18 @@ softpipe_unmap_surfaces(struct softpipe_context *sp) sp_flush_tile_cache(sp, sp->sbuf_cache); for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - struct pipe_surface *ps = sp->framebuffer.cbufs[i]; - if (ps->region) - pipe->region_unmap(pipe, ps->region); + ps = sp->framebuffer.cbufs[i]; + if (ps->map) + pipe_surface_unmap(ps); } - if (sp->framebuffer.zbuf) { - struct pipe_surface *ps = sp->framebuffer.zbuf; - if (ps->region) - pipe->region_unmap(pipe, ps->region); - } - - if (sp->framebuffer.sbuf && sp->framebuffer.sbuf != sp->framebuffer.zbuf) { - struct pipe_surface *ps = sp->framebuffer.sbuf; - if (ps->region) - pipe->region_unmap(pipe, ps->region); - } -} - + ps = sp->framebuffer.zbuf; + if (ps && ps->map) + pipe_surface_unmap(ps); -void -softpipe_unmap_texture_surfaces(struct softpipe_context *sp) -{ - struct pipe_context *pipe = &sp->pipe; - uint i; - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - struct softpipe_texture *spt = sp->texture[i]; - if (spt) { - pipe->region_unmap(pipe, spt->region); - } - } + ps = sp->framebuffer.sbuf; + if (ps && ps->map) + pipe_surface_unmap(ps); } @@ -403,8 +363,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, draw_set_rasterize_stage(softpipe->draw, softpipe->setup); } - - sp_init_region_functions(softpipe); sp_init_surface_functions(softpipe); return &softpipe->pipe; diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 8a82cdfe1a..93eb68405d 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -112,7 +112,6 @@ softpipe_draw_elements(struct pipe_context *pipe, softpipe_update_derived( sp ); softpipe_map_surfaces(sp); - softpipe_map_texture_surfaces(sp); softpipe_map_constant_buffers(sp); /* @@ -184,7 +183,6 @@ softpipe_draw_elements(struct pipe_context *pipe, /* Note: leave drawing surfaces mapped */ - softpipe_unmap_texture_surfaces(sp); softpipe_unmap_constant_buffers(sp); return TRUE; diff --git a/src/mesa/pipe/softpipe/sp_region.c b/src/mesa/pipe/softpipe/sp_region.c deleted file mode 100644 index 58dc6bb96e..0000000000 --- a/src/mesa/pipe/softpipe/sp_region.c +++ /dev/null @@ -1,79 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Provide additional functionality on top of bufmgr buffers: - * - 2d semantics and blit operations (XXX: remove/simplify blits??) - * - refcounting of buffers for multiple images in a buffer. - * - refcounting of buffer mappings. - */ - -#include "sp_context.h" -#include "sp_region.h" -#include "pipe/p_util.h" -#include "pipe/p_winsys.h" -#include "pipe/p_defines.h" - - - -static ubyte * -sp_region_map(struct pipe_context *pipe, struct pipe_region *region) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - - if (!region->map_refcount++) { - region->map = sp->pipe.winsys->buffer_map( sp->pipe.winsys, - region->buffer, - PIPE_BUFFER_FLAG_WRITE | - PIPE_BUFFER_FLAG_READ); - } - - return region->map; -} - -static void -sp_region_unmap(struct pipe_context *pipe, struct pipe_region *region) -{ - struct softpipe_context *sp = softpipe_context( pipe ); - - if (region->map_refcount > 0) { - assert(region->map); - if (!--region->map_refcount) { - sp->pipe.winsys->buffer_unmap( sp->pipe.winsys, - region->buffer ); - region->map = NULL; - } - } -} - - -void -sp_init_region_functions(struct softpipe_context *sp) -{ - sp->pipe.region_map = sp_region_map; - sp->pipe.region_unmap = sp_region_unmap; -} - diff --git a/src/mesa/pipe/softpipe/sp_region.h b/src/mesa/pipe/softpipe/sp_region.h deleted file mode 100644 index 432746b27f..0000000000 --- a/src/mesa/pipe/softpipe/sp_region.h +++ /dev/null @@ -1,40 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef SP_REGION_H -#define SP_REGION_H - - -struct softpipe_context; - - -extern void -sp_init_region_functions(struct softpipe_context *sp); - - -#endif /* SP_REGION_H */ diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index a543735b52..daf9955ca5 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -78,7 +78,7 @@ struct softpipe_texture /* The data is held here: */ - struct pipe_region *region; + struct pipe_buffer_handle *buffer; }; void * diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index d4e0bd1e15..9470ef485f 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -27,6 +27,8 @@ /* Authors: Keith Whitwell */ +#include "p_inlines.h" + #include "sp_context.h" #include "sp_state.h" #include "sp_surface.h" @@ -54,17 +56,17 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, sp_flush_tile_cache(sp, sp->cbuf_cache[i]); /* unmap old */ ps = sp->framebuffer.cbufs[i]; - if (ps && ps->region) - pipe->region_unmap(pipe, ps->region); + if (ps && ps->map) + pipe_surface_unmap(ps); /* map new */ ps = fb->cbufs[i]; if (ps) - pipe->region_map(pipe, ps->region); + pipe_surface_map(ps); /* assign new */ sp->framebuffer.cbufs[i] = fb->cbufs[i]; /* update cache */ - sp_tile_cache_set_surface(sp->cbuf_cache[i], ps); + sp_tile_cache_set_surface(sp, sp->cbuf_cache[i], ps); } } @@ -76,8 +78,8 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, sp_flush_tile_cache(sp, sp->zbuf_cache); /* unmap old */ ps = sp->framebuffer.zbuf; - if (ps && ps->region) - pipe->region_unmap(pipe, ps->region); + if (ps && ps->map) + pipe_surface_unmap(ps); if (sp->framebuffer.sbuf == sp->framebuffer.zbuf) { /* combined z/stencil */ sp->framebuffer.sbuf = NULL; @@ -85,12 +87,12 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* map new */ ps = fb->zbuf; if (ps) - pipe->region_map(pipe, ps->region); + pipe_surface_map(ps); /* assign new */ sp->framebuffer.zbuf = fb->zbuf; /* update cache */ - sp_tile_cache_set_surface(sp->zbuf_cache, ps); + sp_tile_cache_set_surface(sp, sp->zbuf_cache, ps); } /* XXX combined depth/stencil here */ @@ -101,12 +103,12 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, sp_flush_tile_cache(sp, sp->sbuf_cache_sep); /* unmap old */ ps = sp->framebuffer.sbuf; - if (ps && ps->region) - pipe->region_unmap(pipe, ps->region); + if (ps && ps->map) + pipe_surface_unmap(ps); /* map new */ ps = fb->sbuf; if (ps && fb->sbuf != fb->zbuf) - pipe->region_map(pipe, ps->region); + pipe_surface_map(ps); /* assign new */ sp->framebuffer.sbuf = fb->sbuf; @@ -114,12 +116,12 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, if (fb->sbuf != fb->zbuf) { /* separate stencil buf */ sp->sbuf_cache = sp->sbuf_cache_sep; - sp_tile_cache_set_surface(sp->sbuf_cache, ps); + sp_tile_cache_set_surface(sp, sp->sbuf_cache, ps); } else { /* combined depth/stencil */ sp->sbuf_cache = sp->zbuf_cache; - sp_tile_cache_set_surface(sp->sbuf_cache, ps); + sp_tile_cache_set_surface(sp, sp->sbuf_cache, ps); } } diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index c41bbc59b9..c61e0842fc 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -76,7 +76,7 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const unsigned *src - = ((const unsigned *) (ps->region->map + ps->offset)) + = ((const unsigned *) (ps->map)) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -107,7 +107,7 @@ a8r8g8b8_put_tile(struct pipe_surface *ps, const float *p) { unsigned *dst - = ((unsigned *) (ps->region->map + ps->offset)) + = ((unsigned *) (ps->map)) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -140,7 +140,7 @@ b8g8r8a8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const unsigned *src - = ((const unsigned *) (ps->region->map + ps->offset)) + = ((const unsigned *) (ps->map)) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -171,7 +171,7 @@ b8g8r8a8_put_tile(struct pipe_surface *ps, const float *p) { unsigned *dst - = ((unsigned *) (ps->region->map + ps->offset)) + = ((unsigned *) (ps->map)) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -204,7 +204,7 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ushort *src - = ((const ushort *) (ps->region->map + ps->offset)) + = ((const ushort *) (ps->map)) + y * ps->pitch + x; unsigned i, j; @@ -235,7 +235,7 @@ z16_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ushort *src - = ((const ushort *) (ps->region->map + ps->offset)) + = ((const ushort *) (ps->map)) + y * ps->pitch + x; const float scale = 1.0f / 65535.0f; unsigned i, j; @@ -268,7 +268,7 @@ l8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ubyte *src - = ((const ubyte *) (ps->region->map + ps->offset)) + = ((const ubyte *) (ps->map)) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -299,7 +299,7 @@ a8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ubyte *src - = ((const ubyte *) (ps->region->map + ps->offset)) + = ((const ubyte *) (ps->map)) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -330,7 +330,7 @@ r16g16b16a16_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const short *src - = ((const short *) (ps->region->map + ps->offset)) + = ((const short *) (ps->map)) + (y * ps->pitch + x) * 4; unsigned i, j; unsigned w0 = w; @@ -362,7 +362,7 @@ r16g16b16a16_put_tile(struct pipe_surface *ps, const float *p) { short *dst - = ((short *) (ps->region->map + ps->offset)) + = ((short *) (ps->map)) + (y * ps->pitch + x) * 4; unsigned i, j; unsigned w0 = w; @@ -399,7 +399,7 @@ i8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ubyte *src - = ((const ubyte *) (ps->region->map + ps->offset)) + = ((const ubyte *) (ps->map)) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -430,7 +430,7 @@ a8_l8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ushort *src - = ((const ushort *) (ps->region->map + ps->offset)) + = ((const ushort *) (ps->map)) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -467,7 +467,7 @@ z32_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const uint *src - = ((const uint *) (ps->region->map + ps->offset)) + = ((const uint *) (ps->map)) + y * ps->pitch + x; const double scale = 1.0 / (double) 0xffffffff; unsigned i, j; @@ -501,7 +501,7 @@ s8z24_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const uint *src - = ((const uint *) (ps->region->map + ps->offset)) + = ((const uint *) (ps->map)) + y * ps->pitch + x; const double scale = 1.0 / ((1 << 24) - 1); unsigned i, j; @@ -535,7 +535,7 @@ z24s8_get_tile(struct pipe_surface *ps, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const uint *src - = ((const uint *) (ps->region->map + ps->offset)) + = ((const uint *) (ps->map)) + y * ps->pitch + x; const double scale = 1.0 / ((1 << 24) - 1); unsigned i, j; @@ -589,7 +589,7 @@ softpipe_get_tex_surface(struct pipe_context *pipe, if (ps) { assert(ps->format); assert(ps->refcount); - pipe_region_reference(&ps->region, spt->region); + pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); ps->cpp = pt->cpp; ps->width = pt->width[level]; ps->height = pt->height[level]; @@ -613,7 +613,7 @@ softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, ubyte *pDest; uint i; - assert(ps->region->map); + assert(ps->map); if (dst_stride == 0) { dst_stride = w * cpp; @@ -621,7 +621,7 @@ softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, CLIP_TILE; - pSrc = ps->region->map + ps->offset + (y * ps->pitch + x) * cpp; + pSrc = ps->map + (y * ps->pitch + x) * cpp; pDest = (ubyte *) p; for (i = 0; i < h; i++) { @@ -645,7 +645,7 @@ softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, ubyte *pDest; uint i; - assert(ps->region->map); + assert(ps->map); if (src_stride == 0) { src_stride = w * cpp; @@ -654,7 +654,7 @@ softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, CLIP_TILE; pSrc = (const ubyte *) p; - pDest = ps->region->map + ps->offset + (y * ps->pitch + x) * cpp; + pDest = ps->map + (y * ps->pitch + x) * cpp; for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w * cpp); @@ -817,12 +817,12 @@ sp_surface_data(struct pipe_context *pipe, const void *src, unsigned src_pitch, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - copy_rect(pipe->region_map(pipe, dst->region) + dst->offset, + copy_rect(pipe_surface_map(dst), dst->cpp, dst->pitch, dstx, dsty, width, height, src, src_pitch, srcx, srcy); - pipe->region_unmap(pipe, dst->region); + pipe_surface_unmap(dst); } /* Assumes all values are within bounds -- no checking at this level - @@ -835,29 +835,26 @@ sp_surface_copy(struct pipe_context *pipe, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - ubyte *src_map, *dst_map; assert( dst->cpp == src->cpp ); - dst_map = pipe->region_map(pipe, dst->region); - src_map = pipe->region_map(pipe, src->region); - copy_rect(dst_map + dst->offset, + copy_rect(pipe_surface_map(dst), dst->cpp, dst->pitch, dstx, dsty, width, height, - src_map + src->offset, + pipe_surface_map(src), src->pitch, srcx, srcy); - pipe->region_unmap(pipe, src->region); - pipe->region_unmap(pipe, dst->region); + pipe_surface_unmap(src); + pipe_surface_unmap(dst); } static ubyte * get_pointer(struct pipe_surface *dst, unsigned x, unsigned y) { - return dst->region->map + (y * dst->pitch + x) * dst->cpp; + return dst->map + (y * dst->pitch + x) * dst->cpp; } @@ -879,7 +876,7 @@ sp_surface_fill(struct pipe_context *pipe, assert(dst->pitch > 0); assert(width <= dst->pitch); - (void)pipe->region_map(pipe, dst->region); + (void)pipe_surface_map(dst); switch (dst->cpp) { case 1: @@ -935,7 +932,7 @@ sp_surface_fill(struct pipe_context *pipe, break; } - pipe->region_unmap( pipe, dst->region ); + pipe_surface_unmap( dst ); } diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index 2f9a1e9837..53486f9bba 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -32,6 +32,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" #include "pipe/p_util.h" #include "pipe/p_winsys.h" @@ -380,13 +381,18 @@ softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) sizeof(struct softpipe_texture) - sizeof(struct pipe_texture)); if (softpipe_mipmap_tree_layout(pipe, spt)) { - spt->region = pipe->winsys->region_alloc(pipe->winsys, - spt->pitch * (*pt)->cpp * - spt->total_height, - PIPE_SURFACE_FLAG_TEXTURE); + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, + PIPE_SURFACE_FLAG_TEXTURE); + + if (spt->buffer) { + pipe->winsys->buffer_data(pipe->winsys, spt->buffer, + spt->pitch * (*pt)->cpp * + spt->total_height, NULL, + PIPE_BUFFER_USAGE_PIXEL); + } } - if (!spt->region) { + if (!spt->buffer) { FREE(spt); spt = NULL; } @@ -413,7 +419,7 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); */ - pipe->winsys->region_release(pipe->winsys, &spt->region); + pipe->winsys->buffer_reference(pipe->winsys, &spt->buffer, NULL); for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) if (spt->image_offset[i]) diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 62ee6a27c9..08cd39cc55 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -123,16 +123,24 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc) void -sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, +sp_tile_cache_set_surface(struct softpipe_context *sp, + struct softpipe_tile_cache *tc, struct pipe_surface *ps) { + if (tc->surface && tc->surface->map) + pipe_surface_unmap(tc->surface); + pipe_surface_reference(&tc->surface, ps); } struct pipe_surface * -sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) +sp_tile_cache_get_surface(struct softpipe_context *sp, + struct softpipe_tile_cache *tc) { + if (tc->surface && !tc->surface->map) + pipe_surface_map(tc->surface); + return tc->surface; } @@ -162,7 +170,7 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, boolean is_depth_stencil; int inuse = 0, pos; - if (!ps || !ps->region || !ps->region->map) + if (!ps || !ps->buffer) return; is_depth_stencil = (ps->format == PIPE_FORMAT_S8_Z24 || @@ -367,6 +375,16 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, struct pipe_surface *ps = pipe->get_tex_surface(pipe, tc->texture, face, level, z); + if (ps != tc->surface) { + if (tc->surface && tc->surface->map) + pipe_surface_unmap(tc->surface); + + pipe_surface_reference(&tc->surface, ps); + + if (!tc->surface->map) + pipe_surface_map(tc->surface); + } + pipe->get_tile_rgba(pipe, ps, tile_x, tile_y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index 9967aa5044..de5ff2c498 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -63,11 +63,13 @@ extern void sp_destroy_tile_cache(struct softpipe_tile_cache *tc); extern void -sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, +sp_tile_cache_set_surface(struct softpipe_context *sp, + struct softpipe_tile_cache *tc, struct pipe_surface *sps); extern struct pipe_surface * -sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); +sp_tile_cache_get_surface(struct softpipe_context *sp, + struct softpipe_tile_cache *tc); extern void sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index f87d72d0c8..542b4aca49 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1512,7 +1512,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) pipe->get_tile_rgba = xmesa_get_tile_rgba; pipe->put_tile_rgba = xmesa_put_tile_rgba; - c->st->haveFramebufferRegions = GL_FALSE; + c->st->haveFramebufferSurfaces = GL_FALSE; /* special pipe->clear function */ pipe->clear = xmesa_clear; diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index 449e90184b..6fc2f4ba63 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -269,9 +269,8 @@ static void finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) { struct pipe_context *pipe = ctx->st->pipe; - if (!xrb->St.surface->region) { - xrb->St.surface->region = pipe->winsys->region_alloc(pipe->winsys, 1, - 0x0); + if (!xrb->St.surface->buffer) { + xrb->St.surface->buffer = pipe->winsys->buffer_create(pipe->winsys, 0x0); } } @@ -301,7 +300,7 @@ xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, rb->Height = height; rb->InternalFormat = internalFormat; - if (!xrb->St.surface || !xrb->St.surface->region) + if (!xrb->St.surface || !xrb->St.surface->buffer) finish_surface_init(ctx, xrb); /* surface info */ @@ -363,7 +362,7 @@ xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, xrb->origin4 = NULL; } - if (!xrb->St.surface || !xrb->St.surface->region) + if (!xrb->St.surface || !xrb->St.surface->buffer) finish_surface_init(ctx, xrb); xrb->St.surface->width = width; diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 58ca2e7fb3..4a54b5c7e1 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -665,12 +665,12 @@ xmesa_new_color_surface(struct pipe_winsys *winsys, GLuint pipeFormat) xms->surface.refcount = 1; xms->surface.winsys = winsys; - /* Note, the region we allocate doesn't actually have any storage + /* Note, the buffer we allocate doesn't actually have any storage * since we're drawing into an XImage or Pixmap. - * The region's size will get set in the xmesa_alloc_front/back_storage() + * The surface's size will get set in the xmesa_alloc_front/back_storage() * functions. */ - xms->surface.region = winsys->region_alloc(winsys, 1, 0x0); + xms->surface.buffer = winsys->buffer_create(winsys, 0x0); return &xms->surface; } @@ -713,7 +713,7 @@ xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, uint value) { struct softpipe_context *sp = softpipe_context(pipe); - if (ps == sp_tile_cache_get_surface(sp->cbuf_cache[0])) { + if (ps == sp_tile_cache_get_surface(sp, sp->cbuf_cache[0])) { float clear[4]; clear[0] = 0.2; /* XXX hack */ clear[1] = 0.2; diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 99816a811d..295174d4bb 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -186,7 +186,7 @@ xm_get_name(struct pipe_winsys *pws) static struct pipe_buffer_handle * -xm_buffer_create(struct pipe_winsys *pws, unsigned alignment) +xm_buffer_create(struct pipe_winsys *pws, unsigned flags) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); buffer->refcount = 1; @@ -228,47 +228,6 @@ xm_surface_pitch(struct pipe_winsys *winsys, unsigned cpp, unsigned width, } -static struct pipe_region * -xm_region_alloc(struct pipe_winsys *winsys, unsigned size, unsigned flags) -{ - struct pipe_region *region = CALLOC_STRUCT(pipe_region); - const unsigned alignment = 64; - - region->refcount = 1; - - assert(size > 0); - - region->buffer = winsys->buffer_create( winsys, alignment ); - - /* NULL data --> just allocate the space */ - winsys->buffer_data( winsys, - region->buffer, - size, - NULL, - PIPE_BUFFER_USAGE_PIXEL ); - return region; -} - - -static void -xm_region_release(struct pipe_winsys *winsys, struct pipe_region **region) -{ - if (!*region) - return; - - assert((*region)->refcount > 0); - (*region)->refcount--; - - if ((*region)->refcount == 0) { - assert((*region)->map_refcount == 0); - - winsys->buffer_reference( winsys, &((*region)->buffer), NULL ); - free(*region); - } - *region = NULL; -} - - /** * Called via pipe->surface_alloc() to create new surfaces (textures, * renderbuffers, etc. @@ -301,8 +260,8 @@ xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) struct pipe_surface *surf = *s; surf->refcount--; if (surf->refcount == 0) { - if (surf->region) - winsys->region_release(winsys, &surf->region); + if (surf->buffer) + winsys->buffer_reference(winsys, &surf->buffer, NULL); free(surf); } *s = NULL; @@ -335,9 +294,6 @@ xmesa_get_pipe_winsys(void) ws->buffer_subdata = xm_buffer_subdata; ws->buffer_get_subdata = xm_buffer_get_subdata; - ws->region_alloc = xm_region_alloc; - ws->region_release = xm_region_release; - ws->surface_pitch = xm_surface_pitch; ws->surface_alloc = xm_surface_alloc; ws->surface_release = xm_surface_release; diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 7a245b0ed6..c8d9cba12f 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -42,6 +42,7 @@ #include "st_format.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" /** @@ -68,7 +69,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) GLfloat *accBuf; GLint i; - (void) pipe->region_map(pipe, acc_ps->region); + (void) pipe_surface_map(acc_ps); accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); @@ -83,7 +84,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) free(accBuf); - pipe->region_unmap(pipe, acc_ps->region); + pipe_surface_unmap(acc_ps); } @@ -99,7 +100,7 @@ accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - (void) pipe->region_map(pipe, acc_ps->region); + (void) pipe_surface_map(acc_ps); pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); @@ -111,7 +112,7 @@ accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, free(accBuf); - pipe->region_unmap(pipe, acc_ps->region); + pipe_surface_unmap(acc_ps); } @@ -128,8 +129,8 @@ accum_accum(struct pipe_context *pipe, GLfloat value, colorBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - colorMap = pipe->region_map(pipe, color_ps->region); - accMap = pipe->region_map(pipe, acc_ps->region); + colorMap = pipe_surface_map(color_ps); + accMap = pipe_surface_map(acc_ps); pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, colorBuf); pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); @@ -143,8 +144,8 @@ accum_accum(struct pipe_context *pipe, GLfloat value, free(colorBuf); free(accBuf); - pipe->region_unmap(pipe, color_ps->region); - pipe->region_unmap(pipe, acc_ps->region); + pipe_surface_unmap(color_ps); + pipe_surface_unmap(acc_ps); } @@ -159,8 +160,8 @@ accum_load(struct pipe_context *pipe, GLfloat value, buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - (void) pipe->region_map(pipe, color_ps->region); - (void) pipe->region_map(pipe, acc_ps->region); + (void) pipe_surface_map(color_ps); + (void) pipe_surface_map(acc_ps); pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, buf); @@ -172,8 +173,8 @@ accum_load(struct pipe_context *pipe, GLfloat value, free(buf); - pipe->region_unmap(pipe, color_ps->region); - pipe->region_unmap(pipe, acc_ps->region); + pipe_surface_unmap(color_ps); + pipe_surface_unmap(acc_ps); } @@ -190,8 +191,8 @@ accum_return(GLcontext *ctx, GLfloat value, abuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - (void) pipe->region_map(pipe, color_ps->region); - (void) pipe->region_map(pipe, acc_ps->region); + (void) pipe_surface_map(color_ps); + (void) pipe_surface_map(acc_ps); pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); @@ -218,8 +219,8 @@ accum_return(GLcontext *ctx, GLfloat value, if (cbuf) free(cbuf); - pipe->region_unmap(pipe, color_ps->region); - pipe->region_unmap(pipe, acc_ps->region); + pipe_surface_unmap(color_ps); + pipe_surface_unmap(acc_ps); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c28ad15b29..a61daffb20 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -492,11 +492,10 @@ make_texture(struct st_context *st, surface = pipe->get_tex_surface(pipe, pt, 0, 0, 0); - /* map texture region */ - (void) pipe->region_map(pipe, surface->region); - dest = surface->region->map + surface->offset; + /* map texture surface */ + dest = pipe_surface_map(surface); - /* Put image into texture region. + /* Put image into texture surface. * Note that the image is actually going to be upside down in * the texture. We deal with that with texcoords. */ @@ -513,7 +512,7 @@ make_texture(struct st_context *st, unpack); /* unmap */ - pipe->region_unmap(pipe, surface->region); + pipe_surface_unmap(surface); pipe_surface_reference(&surface, NULL); assert(success); @@ -852,7 +851,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, pipe->flush(pipe, 0); /* map the stencil buffer */ - stmap = pipe->region_map(pipe, ps->region); + stmap = pipe_surface_map(ps); /* if width > MAX_WIDTH, have to process image in chunks */ skipPixels = 0; @@ -909,7 +908,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } /* unmap the stencil buffer */ - pipe->region_unmap(pipe, ps->region); + pipe_surface_unmap(ps); } @@ -1027,11 +1026,10 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, surface = pipe->get_tex_surface(pipe, pt, 0, 0, 0); - /* map texture region */ - (void) pipe->region_map(pipe, surface->region); - dest = surface->region->map + surface->offset; + /* map texture surface */ + dest = pipe_surface_map(surface); - /* Put image into texture region. + /* Put image into texture surface. * Note that the image is actually going to be upside down in * the texture. We deal with that with texcoords. */ @@ -1089,7 +1087,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, } /* row */ /* Release surface */ - pipe->region_unmap(pipe, surface->region); + pipe_surface_unmap(surface); pipe_surface_reference(&surface, NULL); pt->format = format; @@ -1130,8 +1128,6 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, GLsizei width, GLsizei height, GLint dstx, GLint dsty) { - struct st_context *st = ctx->st; - struct pipe_context *pipe = st->pipe; struct st_renderbuffer *rbRead = st_renderbuffer(ctx->ReadBuffer->_StencilBuffer); struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer); struct pipe_surface *psRead = rbRead->surface; @@ -1147,8 +1143,8 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, } /* map the stencil buffers */ - readMap = pipe->region_map(pipe, psRead->region); - drawMap = pipe->region_map(pipe, psDraw->region); + readMap = pipe_surface_map(psRead); + drawMap = pipe_surface_map(psDraw); /* this will do stencil pixel transfer ops */ st_read_stencil_pixels(ctx, srcx, srcy, width, height, GL_UNSIGNED_BYTE, @@ -1192,8 +1188,8 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, free(buffer); /* unmap the stencil buffers */ - pipe->region_unmap(pipe, psRead->region); - pipe->region_unmap(pipe, psDraw->region); + pipe_surface_unmap(psRead); + pipe_surface_unmap(psDraw); } @@ -1252,11 +1248,11 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, } /* For some drivers (like Xlib) it's not possible to treat the - * front/back color buffers as regions (they're XImages and Pixmaps). - * So, this var tells us if we can use region_copy here... + * front/back color buffers as surfaces (they're XImages and Pixmaps). + * So, this var tells us if we can use surface_copy here... */ - if (st->haveFramebufferRegions) { - /* copy source framebuffer region into mipmap/texture */ + if (st->haveFramebufferSurfaces) { + /* copy source framebuffer surface into mipmap/texture */ pipe->surface_copy(pipe, psTex, /* dest */ 0, 0, /* destx/y */ @@ -1267,14 +1263,14 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, /* alternate path using get/put_tile() */ GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - (void) pipe->region_map(pipe, psRead->region); - (void) pipe->region_map(pipe, psTex->region); + (void) pipe_surface_map(psRead); + (void) pipe_surface_map(psTex); pipe->get_tile_rgba(pipe, psRead, srcx, srcy, width, height, buf); pipe->put_tile_rgba(pipe, psTex, 0, 0, width, height, buf); - pipe->region_unmap(pipe, psRead->region); - pipe->region_unmap(pipe, psTex->region); + pipe_surface_unmap(psRead); + pipe_surface_unmap(psTex); free(buf); } diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 43681b7f8a..6b9023c410 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -109,22 +109,22 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, width, flags); } - /* free old region */ - if (strb->surface->region) { - /* loop here since mapping is refcounted */ - struct pipe_region *r = strb->surface->region; - while (r->map) - pipe->region_unmap(pipe, r); - pipe->winsys->region_release(pipe->winsys, &strb->surface->region); - } - - strb->surface->region = pipe->winsys->region_alloc(pipe->winsys, - strb->surface->pitch * - cpp * height, flags); - if (!strb->surface->region) + /* loop here since mapping is refcounted */ + while (strb->surface->map) + pipe_surface_unmap(strb->surface); + if (strb->surface->buffer) + pipe->winsys->buffer_reference(pipe->winsys, &strb->surface->buffer, + NULL); + + strb->surface->buffer = pipe->winsys->buffer_create(pipe->winsys, flags); + if (!strb->surface->buffer) return GL_FALSE; /* out of memory, try s/w buffer? */ - ASSERT(strb->surface->region->buffer); + pipe->winsys->buffer_data(pipe->winsys, strb->surface->buffer, + strb->surface->pitch * cpp * height, NULL, + PIPE_BUFFER_USAGE_PIXEL); + + ASSERT(strb->surface->buffer); ASSERT(strb->surface->format); strb->Base.Width = strb->surface->width = width; diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 83fe480af8..e2243e7de4 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -39,6 +39,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" #include "st_context.h" #include "st_cb_readpixels.h" #include "st_cb_fbo.h" @@ -56,8 +57,6 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, const struct gl_pixelstore_attrib *packing, GLvoid *pixels) { - struct st_context *st = ctx->st; - struct pipe_context *pipe = st->pipe; struct gl_framebuffer *fb = ctx->ReadBuffer; struct st_renderbuffer *strb = st_renderbuffer(fb->_StencilBuffer); struct pipe_surface *ps = strb->surface; @@ -65,7 +64,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, GLint j; /* map the stencil buffer */ - stmap = pipe->region_map(pipe, ps->region); + stmap = pipe_surface_map(ps); /* width should never be > MAX_WIDTH since we did clipping earlier */ ASSERT(width <= MAX_WIDTH); @@ -122,7 +121,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, /* unmap the stencil buffer */ - pipe->region_unmap(pipe, ps->region); + pipe_surface_unmap(ps); } @@ -151,7 +150,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, /* Do all needed clipping here, so that we can forget about it later */ if (!_mesa_clip_readpixels(ctx, &x, &y, &width, &height, &clippedPacking)) { - /* The ReadPixels region is totally outside the window bounds */ + /* The ReadPixels surface is totally outside the window bounds */ return; } @@ -180,7 +179,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (!strb) return; - pipe->region_map(pipe, strb->surface->region); + pipe_surface_map(strb->surface); if (format == GL_RGBA && type == GL_FLOAT) { /* write tile(row) directly into user's buffer */ @@ -231,7 +230,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } } - pipe->region_unmap(pipe, strb->surface->region); + pipe_surface_unmap(strb->surface); } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index acc97df2e7..1fcef746c1 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -64,7 +64,7 @@ struct st_texture_object GLuint textureOffset; /* On validation any active images held in main memory or in other - * regions will be copied to this region and the old storage freed. + * textures will be copied to this texture and the old storage freed. */ struct pipe_texture *pt; @@ -721,7 +721,7 @@ st_TexImage(GLcontext * ctx, } if (stImage->pt && i < depth) { - st_texture_image_unmap(ctx->st, stImage); + st_texture_image_unmap(stImage); texImage->Data = st_texture_image_map(ctx->st, stImage, i); pixels += srcImageStride; } @@ -731,7 +731,7 @@ st_TexImage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, unpack); if (stImage->pt) { - st_texture_image_unmap(ctx->st, stImage); + st_texture_image_unmap(stImage); texImage->Data = NULL; } @@ -860,7 +860,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, } if (stImage->pt && i < depth) { - st_texture_image_unmap(ctx->st, stImage); + st_texture_image_unmap(stImage); texImage->Data = st_texture_image_map(ctx->st, stImage, i); pixels += dstImageStride; } @@ -870,7 +870,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, /* Unmap */ if (stImage->pt) { - st_texture_image_unmap(ctx->st, stImage); + st_texture_image_unmap(stImage); texImage->Data = NULL; } } @@ -948,7 +948,7 @@ st_TexSubimage(GLcontext * ctx, } if (stImage->pt && i < depth) { - st_texture_image_unmap(ctx->st, stImage); + st_texture_image_unmap(stImage); texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i); pixels += srcImageStride; } @@ -966,7 +966,7 @@ st_TexSubimage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, packing); if (stImage->pt) { - st_texture_image_unmap(ctx->st, stImage); + st_texture_image_unmap(stImage); texImage->Data = NULL; } } @@ -1051,8 +1051,8 @@ texture_face(GLenum target) /** - * Do a CopyTexSubImage operation by mapping the source region and - * dest region and using get_tile()/put_tile() to access the pixels/texels. + * Do a CopyTexSubImage operation by mapping the source surface and + * dest surface and using get_tile()/put_tile() to access the pixels/texels. * * Note: srcY=0=TOP of renderbuffer */ @@ -1088,8 +1088,8 @@ fallback_copy_texsubimage(GLcontext *ctx, dest_surf = pipe->get_tex_surface(pipe, pt, face, level, destZ); - (void) pipe->region_map(pipe, dest_surf->region); - (void) pipe->region_map(pipe, src_surf->region); + (void) pipe_surface_map(dest_surf); + (void) pipe_surface_map(src_surf); /* buffer for one row */ data = (GLfloat *) malloc(width * 4 * sizeof(GLfloat)); @@ -1110,8 +1110,8 @@ fallback_copy_texsubimage(GLcontext *ctx, } - (void) pipe->region_unmap(pipe, dest_surf->region); - (void) pipe->region_unmap(pipe, src_surf->region); + (void) pipe_surface_unmap(dest_surf); + (void) pipe_surface_unmap(src_surf); free(data); } @@ -1178,8 +1178,8 @@ do_copy_texsubimage(GLcontext *ctx, if (src_format == dest_format && ctx->_ImageTransferState == 0x0 && - strb->surface->region && - dest_surface->region && + strb->surface->buffer && + dest_surface->buffer && strb->surface->cpp == stImage->pt->cpp) { /* do blit-style copy */ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index a4ec3721be..abde6d64b0 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -103,7 +103,7 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st->ctx->VertexProgram._MaintainTnlProgram = GL_TRUE; - st->haveFramebufferRegions = GL_TRUE; + st->haveFramebufferSurfaces = GL_TRUE; st->pixel_xfer.cache = _mesa_new_program_cache(); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index db97014c5a..c31b76c63f 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -34,7 +34,6 @@ struct st_context; -struct st_region; struct st_texture_object; struct st_fragment_program; struct draw_context; @@ -131,11 +130,11 @@ struct st_context char vendor[100]; char renderer[100]; - /** Can we access the front/back color buffers as pipe_regions? + /** Can we access the front/back color buffers as pipe_surfaces? * We can't with the Xlib driver... * This is a hack that should be fixed someday. */ - GLboolean haveFramebufferRegions; + GLboolean haveFramebufferSurfaces; /* State to be validated: */ diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index a5582c31c0..1ec4514873 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -33,6 +33,7 @@ #include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" @@ -187,17 +188,15 @@ st_texture_image_map(struct st_context *st, struct st_texture_image *stImage, stImage->surface = st->pipe->get_tex_surface(st->pipe, pt, stImage->face, stImage->level, zoffset); - (void) st->pipe->region_map(st->pipe, stImage->surface->region); - - return stImage->surface->region->map + stImage->surface->offset; + return pipe_surface_map(stImage->surface); } void -st_texture_image_unmap(struct st_context *st, struct st_texture_image *stImage) +st_texture_image_unmap(struct st_texture_image *stImage) { DBG("%s\n", __FUNCTION__); - st->pipe->region_unmap(st->pipe, stImage->surface->region); + pipe_surface_unmap(stImage->surface); pipe_surface_reference(&stImage->surface, NULL); } diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index b25e3f3f3b..f49d91a069 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -65,8 +65,7 @@ st_texture_image_map(struct st_context *st, GLuint zoffset); extern void -st_texture_image_unmap(struct st_context *st, - struct st_texture_image *stImage); +st_texture_image_unmap(struct st_texture_image *stImage); /* Return pointers to each 2d slice within an image. Indexed by depth -- cgit v1.2.3 From b9300f16c40e1fecf6237ca06888e42da85c55b5 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 07:48:33 -0700 Subject: silence void ptr arithmetic warning --- src/mesa/pipe/p_inlines.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_inlines.h b/src/mesa/pipe/p_inlines.h index 1697284b04..e7303c45c1 100644 --- a/src/mesa/pipe/p_inlines.h +++ b/src/mesa/pipe/p_inlines.h @@ -37,11 +37,12 @@ static INLINE ubyte * pipe_surface_map(struct pipe_surface *surface) { if (!surface->map_refcount++) { - surface->map = surface->winsys->buffer_map( surface->winsys, + surface->map + = (ubyte *) surface->winsys->buffer_map( surface->winsys, surface->buffer, PIPE_BUFFER_FLAG_WRITE | PIPE_BUFFER_FLAG_READ ) - + surface->offset; + + surface->offset; } return surface->map; -- cgit v1.2.3 From 5823f99d0fbc054b87aeb1bc15d413d3eadd27a8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 07:52:35 -0700 Subject: fix void pointer arithmetic warnings --- src/mesa/state_tracker/st_cb_texture.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 1fcef746c1..14f8d38def 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -706,6 +706,7 @@ st_TexImage(GLcontext * ctx, GLuint srcImageStride = _mesa_image_image_stride(unpack, width, height, format, type); int i; + const GLubyte *src = (const GLubyte *) pixels; for (i = 0; i++ < depth;) { if (!texImage->TexFormat->StoreImage(ctx, dims, @@ -716,14 +717,14 @@ st_TexImage(GLcontext * ctx, dstRowStride, texImage->ImageOffsets, width, height, 1, - format, type, pixels, unpack)) { + format, type, src, unpack)) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); } if (stImage->pt && i < depth) { st_texture_image_unmap(stImage); texImage->Data = st_texture_image_map(ctx->st, stImage, i); - pixels += srcImageStride; + src += srcImageStride; } } } @@ -826,6 +827,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, type); GLuint depth; int i; + GLubyte *dest; /* Map */ if (stImage->pt) { @@ -850,19 +852,21 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, depth = texImage->Depth; texImage->Depth = 1; + dest = (GLubyte *) pixels; + for (i = 0; i++ < depth;) { if (compressed) { - _mesa_get_compressed_teximage(ctx, target, level, pixels, + _mesa_get_compressed_teximage(ctx, target, level, dest, texObj, texImage); } else { - _mesa_get_teximage(ctx, target, level, format, type, pixels, + _mesa_get_teximage(ctx, target, level, format, type, dest, texObj, texImage); } if (stImage->pt && i < depth) { st_texture_image_unmap(stImage); texImage->Data = st_texture_image_map(ctx->st, stImage, i); - pixels += dstImageStride; + dest += dstImageStride; } } @@ -916,6 +920,7 @@ st_TexSubimage(GLcontext * ctx, GLuint srcImageStride = _mesa_image_image_stride(packing, width, height, format, type); int i; + const GLubyte *src; DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), @@ -935,6 +940,8 @@ st_TexSubimage(GLcontext * ctx, dstRowStride = stImage->surface->pitch * stImage->surface->cpp; } + src = (const GLubyte *) pixels; + for (i = 0; i++ < depth;) { if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, texImage->TexFormat, @@ -943,14 +950,14 @@ st_TexSubimage(GLcontext * ctx, dstRowStride, texImage->ImageOffsets, width, height, 1, - format, type, pixels, packing)) { + format, type, src, packing)) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "st_TexSubImage"); } if (stImage->pt && i < depth) { st_texture_image_unmap(stImage); texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i); - pixels += srcImageStride; + src += srcImageStride; } } -- cgit v1.2.3 From ee6b39cfe602bd88a88bedc5af672832d9157105 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 07:53:06 -0700 Subject: move surface pitch calculation so it always gets updated --- src/mesa/state_tracker/st_cb_fbo.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 6b9023c410..0ee5f45aae 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -105,10 +105,11 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, if (!strb->surface) return GL_FALSE; strb->surface->cpp = cpp; - strb->surface->pitch = pipe->winsys->surface_pitch(pipe->winsys, cpp, - width, flags); } + strb->surface->pitch = pipe->winsys->surface_pitch(pipe->winsys, cpp, + width, flags); + /* loop here since mapping is refcounted */ while (strb->surface->map) pipe_surface_unmap(strb->surface); -- cgit v1.2.3 From f77ce9e4fa6430fd6f4aaf07c58ed12569dec481 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 07:53:21 -0700 Subject: added ST_SURFACE_DEPTH --- src/mesa/state_tracker/st_framebuffer.c | 2 +- src/mesa/state_tracker/st_public.h | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 4ae2837f0a..8633f431b2 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -155,7 +155,7 @@ st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex) { struct st_renderbuffer *strb; - assert(surfIndex <= ST_SURFACE_BACK_RIGHT); + assert(surfIndex <= ST_SURFACE_DEPTH); /* sanity checks, ST tokens should match Mesa tokens */ assert(ST_SURFACE_FRONT_LEFT == BUFFER_FRONT_LEFT); diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 408e1927e5..307635936b 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -35,6 +35,7 @@ #define ST_SURFACE_BACK_LEFT 1 #define ST_SURFACE_FRONT_RIGHT 2 #define ST_SURFACE_BACK_RIGHT 3 +#define ST_SURFACE_DEPTH 8 struct st_context; -- cgit v1.2.3 From 749d723287df348bd36ee43f96de6892a3eba09f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 07:57:54 -0700 Subject: Overhaul the Xlib winsys layer. Front/back color buffers are now allocated with ordinary malloc() via the winsys buffer functions. To display surfaces in SwapBuffers() or flush_frontbuffer() we create an XImage that wraps the surface, then use XPutImage to copy to the window. Shared memory transport disabled for now. --- src/mesa/pipe/xlib/fakeglx.c | 17 +- src/mesa/pipe/xlib/xm_api.c | 1282 ++++----------------------------------- src/mesa/pipe/xlib/xm_buffer.c | 20 +- src/mesa/pipe/xlib/xm_surface.c | 683 --------------------- src/mesa/pipe/xlib/xm_winsys.c | 157 +++-- src/mesa/pipe/xlib/xmesaP.h | 457 +------------- src/mesa/sources | 4 +- 7 files changed, 257 insertions(+), 2363 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/fakeglx.c b/src/mesa/pipe/xlib/fakeglx.c index d7e7b292d1..6965ee0794 100644 --- a/src/mesa/pipe/xlib/fakeglx.c +++ b/src/mesa/pipe/xlib/fakeglx.c @@ -1433,7 +1433,6 @@ Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo, return NULL; } - glxCtx->xmesaContext->direct = GL_FALSE; glxCtx->glxContext.isDirect = GL_FALSE; glxCtx->glxContext.currentDpy = dpy; glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ @@ -1565,7 +1564,7 @@ Fake_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ) if (!b) { return 0; } - return b->frontxrb->pixmap; + return b->drawable; } @@ -1591,7 +1590,7 @@ Fake_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, if (!b) { return 0; } - return b->frontxrb->pixmap; + return b->drawable; } @@ -1662,9 +1661,9 @@ Fake_glXDestroyContext( Display *dpy, GLXContext ctx ) static Bool Fake_glXIsDirect( Display *dpy, GLXContext ctx ) { - struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; (void) dpy; - return glxCtx->xmesaContext->direct; + (void) ctx; + return False; } @@ -2360,7 +2359,7 @@ Fake_glXCreatePbuffer( Display *dpy, GLXFBConfig config, * glXMakeCurrent takes. */ if (xmbuf) - return (GLXPbuffer) xmbuf->frontxrb->pixmap; + return (GLXPbuffer) xmbuf->drawable; else return 0; } @@ -2444,7 +2443,6 @@ Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config, return NULL; } - glxCtx->xmesaContext->direct = GL_FALSE; glxCtx->glxContext.isDirect = GL_FALSE; glxCtx->glxContext.currentDpy = dpy; glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ @@ -2642,7 +2640,7 @@ Fake_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixm { XMesaVisual xmvis = (XMesaVisual) config; XMesaBuffer xmbuf = XMesaCreatePixmapBuffer(xmvis, pixmap, 0); - return xmbuf->frontxrb->pixmap; /* need to return an X ID */ + return xmbuf->drawable; /* need to return an X ID */ } @@ -2667,7 +2665,6 @@ Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int re return NULL; } - glxCtx->xmesaContext->direct = GL_FALSE; glxCtx->glxContext.isDirect = GL_FALSE; glxCtx->glxContext.currentDpy = dpy; glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ @@ -2736,7 +2733,7 @@ Fake_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, /* A GLXPbuffer handle must be an X Drawable because that's what * glXMakeCurrent takes. */ - return (GLXPbuffer) xmbuf->frontxrb->pixmap; + return (GLXPbuffer) xmbuf->drawable; } diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 542b4aca49..ce7e3cb6c2 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -48,12 +48,7 @@ * according to the display's visual red_mask, green_mask, and blue_mask. * If XPutPixel is used to put a pixel into an XImage then XPutPixel will * do byte swapping if needed. If one wants to directly "poke" the pixel - * into the XImage's buffer then the pixel must be byte swapped first. In - * Mesa, when byte swapping is needed we use the PF_TRUECOLOR pixel format - * and use XPutPixel everywhere except in the implementation of - * glClear(GL_COLOR_BUFFER_BIT). We want this function to be fast so - * instead of using XPutPixel we "poke" our values after byte-swapping - * the clear pixel value if needed. + * into the XImage's buffer then the pixel must be byte swapped first. * */ @@ -77,8 +72,8 @@ #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" -#include "pipe/softpipe/sp_context.h" #include "pipe/p_defines.h" +#include "pipe/p_context.h" /** * Global X driver lock @@ -135,24 +130,6 @@ static int check_for_xshm( XMesaDisplay *display ) } -/** - * Apply gamma correction to an intensity value in [0..max]. Return the - * new intensity value. - */ -static GLint -gamma_adjust( GLfloat gamma, GLint value, GLint max ) -{ - if (gamma == 1.0) { - return value; - } - else { - double x = (double) value / (double) max; - return IROUND_POS((GLfloat) max * _mesa_pow(x, 1.0F/gamma)); - } -} - - - /** * Return the true number of bits per pixel for XImages. * For example, if we request a 24-bit deep visual we may actually need/get @@ -252,19 +229,19 @@ get_drawable_size( XMesaDisplay *dpy, Drawable d, GLuint *width, GLuint *height * \param width returns width in pixels * \param height returns height in pixels */ -void +static void xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b, GLuint *width, GLuint *height) { #ifdef XFree86Server - *width = MIN2(b->frontxrb->drawable->width, MAX_WIDTH); - *height = MIN2(b->frontxrb->drawable->height, MAX_HEIGHT); + *width = MIN2(b->drawable->width, MAX_WIDTH); + *height = MIN2(b->drawable->height, MAX_HEIGHT); #else Status stat; _glthread_LOCK_MUTEX(_xmesa_lock); XSync(b->xm_visual->display, 0); /* added for Chromium */ - stat = get_drawable_size(dpy, b->frontxrb->pixmap, width, height); + stat = get_drawable_size(dpy, b->drawable, width, height); _glthread_UNLOCK_MUTEX(_xmesa_lock); if (!stat) { @@ -301,7 +278,6 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, { XMesaBuffer b; GLframebuffer *fb; - struct pipe_winsys *winsys = xmesa_get_pipe_winsys(); ASSERT(type == WINDOW || type == PIXMAP || type == PBUFFER); @@ -309,7 +285,8 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, if (!b) return NULL; - b->display = vis->display; + b->drawable = d; + b->xm_visual = vis; b->type = type; b->cmap = cmap; @@ -317,113 +294,20 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, /* * Create framebuffer, but we'll plug in our own renderbuffers below. */ - b->stfb = st_create_framebuffer(&vis->mesa_visual, GL_FALSE, (void *) b); + b->stfb = st_create_framebuffer(&vis->mesa_visual, GL_TRUE, (void *) b); fb = &b->stfb->Base; - fb->Delete = xmesa_delete_framebuffer; - - /* - * XXX we want to create surfaces for pipe, not renderbuffers for Mesa. - */ - - /* - * Front renderbuffer - */ - b->frontxrb = xmesa_create_renderbuffer(winsys, 0, vis, GL_FALSE); - if (!b->frontxrb) { - _mesa_free(b); - return NULL; - } - b->frontxrb->Parent = b; - b->frontxrb->drawable = d; - b->frontxrb->pixmap = (XMesaPixmap) d; - _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &b->frontxrb->St.Base); - - /* - * Back renderbuffer - */ - if (vis->mesa_visual.doubleBufferMode) { - b->backxrb = xmesa_create_renderbuffer(winsys, 0, vis, GL_TRUE); - if (!b->backxrb) { - /* XXX free front xrb too */ - _mesa_free(b); - return NULL; - } - b->backxrb->Parent = b; - /* determine back buffer implementation */ - b->db_mode = vis->ximage_flag ? BACK_XIMAGE : BACK_PIXMAP; - - _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &b->backxrb->St.Base); - } - - /* - * Software alpha planes - */ - if (vis->mesa_visual.alphaBits > 0 - && vis->undithered_pf != PF_8A8B8G8R - && vis->undithered_pf != PF_8A8R8G8B) { - /* Visual has alpha, but pixel format doesn't support it. - * We'll use an alpha renderbuffer wrapper. - */ -#if 0 - b->swAlpha = GL_TRUE; /* don't do any renderbuffer wrapping */ -#else - b->swAlpha = GL_FALSE; -#endif - } - else { - b->swAlpha = GL_FALSE; - } - - if (vis->mesa_visual.depthBits > 0 && - vis->mesa_visual.stencilBits > 0) { - /* combined depth/stencil */ - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, rb); - _mesa_add_renderbuffer(fb, BUFFER_STENCIL, rb); - } - else { - if (vis->mesa_visual.depthBits > 24) { - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32); - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, rb); - } - else if (vis->mesa_visual.depthBits > 16) { - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, rb); - } - else if (vis->mesa_visual.depthBits > 1) { - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); - _mesa_add_renderbuffer(fb, BUFFER_DEPTH, rb); - } - - if (vis->mesa_visual.stencilBits > 0) { - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT); - _mesa_add_renderbuffer(fb, BUFFER_STENCIL, rb); - } - } - - if (vis->mesa_visual.accumRedBits > 0) { - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(GL_RGBA16); - _mesa_add_renderbuffer(fb, BUFFER_ACCUM, rb); - } - - /* - * Other renderbuffer (depth, stencil, etc) + * Create scratch XImage for xmesa_display_surface() */ - _mesa_add_soft_renderbuffers(fb, - GL_FALSE, /* color */ - GL_FALSE, /*vis->mesa_visual.haveDepthBuffer,*/ - GL_FALSE, /* stencil */ - GL_FALSE, /* accum */ - b->swAlpha, - vis->mesa_visual.numAuxBuffers > 0 ); + b->tempImage = XCreateImage(vis->display, + vis->visinfo->visual, + vis->visinfo->depth, + ZPixmap, 0, /* format, offset */ + NULL, /* data */ + 0, 0, /* size */ + 32, /* bitmap_pad */ + 0); /* bytes_per_line */ /* GLX_EXT_texture_from_pixmap */ b->TextureTarget = 0; @@ -446,8 +330,10 @@ XMesaBuffer xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis) { XMesaBuffer b; - for (b=XMesaBufferList; b; b=b->Next) { - if (b->display==dpy && b->cmap==cmap && b!=notThis) { + for (b = XMesaBufferList; b; b = b->Next) { + if (b->xm_visual->display == dpy && + b->cmap == cmap && + b != notThis) { return b; } } @@ -479,7 +365,10 @@ xmesa_free_buffer(XMesaBuffer buffer) /* Since the X window for the XMesaBuffer is going away, we don't * want to dereference this pointer in the future. */ - b->frontxrb->drawable = 0; + b->drawable = 0; + + buffer->tempImage->data = NULL; + XDestroyImage(buffer->tempImage); /* Unreference. If count = zero we'll really delete the buffer */ _mesa_unreference_framebuffer(&fb); @@ -494,22 +383,6 @@ xmesa_free_buffer(XMesaBuffer buffer) } -/** - * Copy X color table stuff from one XMesaBuffer to another. - */ -static void -copy_colortable_info(XMesaBuffer dst, const XMesaBuffer src) -{ - MEMCPY(dst->color_table, src->color_table, sizeof(src->color_table)); - MEMCPY(dst->pixel_to_r, src->pixel_to_r, sizeof(src->pixel_to_r)); - MEMCPY(dst->pixel_to_g, src->pixel_to_g, sizeof(src->pixel_to_g)); - MEMCPY(dst->pixel_to_b, src->pixel_to_b, sizeof(src->pixel_to_b)); - dst->num_alloced = src->num_alloced; - MEMCPY(dst->alloced_colors, src->alloced_colors, - sizeof(src->alloced_colors)); -} - - /**********************************************************************/ /***** Misc Private Functions *****/ @@ -517,450 +390,49 @@ copy_colortable_info(XMesaBuffer dst, const XMesaBuffer src) /** - * A replacement for XAllocColor. This function should never - * fail to allocate a color. When XAllocColor fails, we return - * the nearest matching color. If we have to allocate many colors - * this function isn't too efficient; the XQueryColors() could be - * done just once. - * Written by Michael Pichler, Brian Paul, Mark Kilgard - * Input: dpy - X display - * cmap - X colormap - * cmapSize - size of colormap - * In/Out: color - the XColor struct - * Output: exact - 1=exact color match, 0=closest match - * alloced - 1=XAlloc worked, 0=XAlloc failed - */ -static void -noFaultXAllocColor( int client, - XMesaDisplay *dpy, - XMesaColormap cmap, - int cmapSize, - XMesaColor *color, - int *exact, int *alloced ) -{ -#ifdef XFree86Server - Pixel *ppixIn; - xrgb *ctable; -#else - /* we'll try to cache ctable for better remote display performance */ - static Display *prevDisplay = NULL; - static XMesaColormap prevCmap = 0; - static int prevCmapSize = 0; - static XMesaColor *ctable = NULL; -#endif - XMesaColor subColor; - int i, bestmatch; - double mindist; /* 3*2^16^2 exceeds long int precision. */ - - (void) client; - - /* First try just using XAllocColor. */ -#ifdef XFree86Server - if (AllocColor(cmap, - &color->red, &color->green, &color->blue, - &color->pixel, - client) == Success) -#else - if (XAllocColor(dpy, cmap, color)) -#endif - { - *exact = 1; - *alloced = 1; - return; - } - - /* Alloc failed, search for closest match */ - - /* Retrieve color table entries. */ - /* XXX alloca candidate. */ -#ifdef XFree86Server - ppixIn = (Pixel *) MALLOC(cmapSize * sizeof(Pixel)); - ctable = (xrgb *) MALLOC(cmapSize * sizeof(xrgb)); - for (i = 0; i < cmapSize; i++) { - ppixIn[i] = i; - } - QueryColors(cmap, cmapSize, ppixIn, ctable); -#else - if (prevDisplay != dpy || prevCmap != cmap - || prevCmapSize != cmapSize || !ctable) { - /* free previously cached color table */ - if (ctable) - _mesa_free(ctable); - /* Get the color table from X */ - ctable = (XMesaColor *) MALLOC(cmapSize * sizeof(XMesaColor)); - assert(ctable); - for (i = 0; i < cmapSize; i++) { - ctable[i].pixel = i; - } - XQueryColors(dpy, cmap, ctable, cmapSize); - prevDisplay = dpy; - prevCmap = cmap; - prevCmapSize = cmapSize; - } -#endif - - /* Find best match. */ - bestmatch = -1; - mindist = 0.0; - for (i = 0; i < cmapSize; i++) { - double dr = 0.30 * ((double) color->red - (double) ctable[i].red); - double dg = 0.59 * ((double) color->green - (double) ctable[i].green); - double db = 0.11 * ((double) color->blue - (double) ctable[i].blue); - double dist = dr * dr + dg * dg + db * db; - if (bestmatch < 0 || dist < mindist) { - bestmatch = i; - mindist = dist; - } - } - - /* Return result. */ - subColor.red = ctable[bestmatch].red; - subColor.green = ctable[bestmatch].green; - subColor.blue = ctable[bestmatch].blue; - /* Try to allocate the closest match color. This should only - * fail if the cell is read/write. Otherwise, we're incrementing - * the cell's reference count. - */ -#ifdef XFree86Server - if (AllocColor(cmap, - &subColor.red, &subColor.green, &subColor.blue, - &subColor.pixel, - client) == Success) { -#else - if (XAllocColor(dpy, cmap, &subColor)) { -#endif - *alloced = 1; - } - else { - /* do this to work around a problem reported by Frank Ortega */ - subColor.pixel = (unsigned long) bestmatch; - subColor.red = ctable[bestmatch].red; - subColor.green = ctable[bestmatch].green; - subColor.blue = ctable[bestmatch].blue; - subColor.flags = DoRed | DoGreen | DoBlue; - *alloced = 0; - } -#ifdef XFree86Server - _mesa_free(ppixIn); - _mesa_free(ctable); -#else - /* don't free table, save it for next time */ -#endif - - *color = subColor; - *exact = 0; -} - - - -/** - * Do setup for PF_GRAYSCALE pixel format. - * Note that buffer may be NULL. - */ -static GLboolean -setup_grayscale(int client, XMesaVisual v, - XMesaBuffer buffer, XMesaColormap cmap) -{ - if (GET_VISUAL_DEPTH(v)<4 || GET_VISUAL_DEPTH(v)>16) { - return GL_FALSE; - } - - if (buffer) { - XMesaBuffer prevBuffer; - - if (!cmap) { - return GL_FALSE; - } - - prevBuffer = xmesa_find_buffer(v->display, cmap, buffer); - if (prevBuffer && - (buffer->xm_visual->mesa_visual.rgbMode == - prevBuffer->xm_visual->mesa_visual.rgbMode)) { - /* Copy colormap stuff from previous XMesaBuffer which uses same - * X colormap. Do this to avoid time spent in noFaultXAllocColor. - */ - copy_colortable_info(buffer, prevBuffer); - } - else { - /* Allocate 256 shades of gray */ - int gray; - int colorsfailed = 0; - for (gray=0;gray<256;gray++) { - GLint r = gamma_adjust( v->RedGamma, gray, 255 ); - GLint g = gamma_adjust( v->GreenGamma, gray, 255 ); - GLint b = gamma_adjust( v->BlueGamma, gray, 255 ); - int exact, alloced; - XMesaColor xcol; - xcol.red = (r << 8) | r; - xcol.green = (g << 8) | g; - xcol.blue = (b << 8) | b; - noFaultXAllocColor( client, v->display, - cmap, GET_COLORMAP_SIZE(v), - &xcol, &exact, &alloced ); - if (!exact) { - colorsfailed++; - } - if (alloced) { - assert(buffer->num_alloced<256); - buffer->alloced_colors[buffer->num_alloced] = xcol.pixel; - buffer->num_alloced++; - } - - /*OLD - assert(gray < 576); - buffer->color_table[gray*3+0] = xcol.pixel; - buffer->color_table[gray*3+1] = xcol.pixel; - buffer->color_table[gray*3+2] = xcol.pixel; - assert(xcol.pixel < 65536); - buffer->pixel_to_r[xcol.pixel] = gray * 30 / 100; - buffer->pixel_to_g[xcol.pixel] = gray * 59 / 100; - buffer->pixel_to_b[xcol.pixel] = gray * 11 / 100; - */ - buffer->color_table[gray] = xcol.pixel; - assert(xcol.pixel < 65536); - buffer->pixel_to_r[xcol.pixel] = gray; - buffer->pixel_to_g[xcol.pixel] = gray; - buffer->pixel_to_b[xcol.pixel] = gray; - } - - if (colorsfailed && _mesa_getenv("MESA_DEBUG")) { - _mesa_warning(NULL, - "Note: %d out of 256 needed colors do not match exactly.\n", - colorsfailed ); - } - } - } - - v->dithered_pf = PF_Grayscale; - v->undithered_pf = PF_Grayscale; - return GL_TRUE; -} - - - -/** - * Setup RGB rendering for a window with a PseudoColor, StaticColor, - * or 8-bit TrueColor visual visual. We try to allocate a palette of 225 - * colors (5 red, 9 green, 5 blue) and dither to approximate a 24-bit RGB - * color. While this function was originally designed just for 8-bit - * visuals, it has also proven to work from 4-bit up to 16-bit visuals. - * Dithering code contributed by Bob Mercier. - */ -static GLboolean -setup_dithered_color(int client, XMesaVisual v, - XMesaBuffer buffer, XMesaColormap cmap) -{ - if (GET_VISUAL_DEPTH(v)<4 || GET_VISUAL_DEPTH(v)>16) { - return GL_FALSE; - } - - if (buffer) { - XMesaBuffer prevBuffer; - - if (!cmap) { - return GL_FALSE; - } - - prevBuffer = xmesa_find_buffer(v->display, cmap, buffer); - if (prevBuffer && - (buffer->xm_visual->mesa_visual.rgbMode == - prevBuffer->xm_visual->mesa_visual.rgbMode)) { - /* Copy colormap stuff from previous, matching XMesaBuffer. - * Do this to avoid time spent in noFaultXAllocColor. - */ - copy_colortable_info(buffer, prevBuffer); - } - else { - /* Allocate X colors and initialize color_table[], red_table[], etc */ - int r, g, b, i; - int colorsfailed = 0; - for (r = 0; r < DITH_R; r++) { - for (g = 0; g < DITH_G; g++) { - for (b = 0; b < DITH_B; b++) { - XMesaColor xcol; - int exact, alloced; - xcol.red =gamma_adjust(v->RedGamma, r*65535/(DITH_R-1),65535); - xcol.green=gamma_adjust(v->GreenGamma, g*65535/(DITH_G-1),65535); - xcol.blue =gamma_adjust(v->BlueGamma, b*65535/(DITH_B-1),65535); - noFaultXAllocColor( client, v->display, - cmap, GET_COLORMAP_SIZE(v), - &xcol, &exact, &alloced ); - if (!exact) { - colorsfailed++; - } - if (alloced) { - assert(buffer->num_alloced<256); - buffer->alloced_colors[buffer->num_alloced] = xcol.pixel; - buffer->num_alloced++; - } - i = DITH_MIX( r, g, b ); - assert(i < 576); - buffer->color_table[i] = xcol.pixel; - assert(xcol.pixel < 65536); - buffer->pixel_to_r[xcol.pixel] = r * 255 / (DITH_R-1); - buffer->pixel_to_g[xcol.pixel] = g * 255 / (DITH_G-1); - buffer->pixel_to_b[xcol.pixel] = b * 255 / (DITH_B-1); - } - } - } - - if (colorsfailed && _mesa_getenv("MESA_DEBUG")) { - _mesa_warning(NULL, - "Note: %d out of %d needed colors do not match exactly.\n", - colorsfailed, DITH_R * DITH_G * DITH_B ); - } - } - } - - v->dithered_pf = PF_Dither; - v->undithered_pf = PF_Lookup; - return GL_TRUE; -} - - -/** - * Setup RGB rendering for a window with a True/DirectColor visual. + * Choose the pixel format for the given visual. + * This will tell the gallium driver how to pack pixel data into + * drawing surfaces. */ -static void -setup_truecolor(XMesaVisual v, XMesaBuffer buffer, XMesaColormap cmap) +static GLuint +choose_pixel_format(XMesaVisual v) { - unsigned long rmask, gmask, bmask; - (void) buffer; - (void) cmap; - - /* Compute red multiplier (mask) and bit shift */ - v->rshift = 0; - rmask = GET_REDMASK(v); - while ((rmask & 1)==0) { - v->rshift++; - rmask = rmask >> 1; - } - - /* Compute green multiplier (mask) and bit shift */ - v->gshift = 0; - gmask = GET_GREENMASK(v); - while ((gmask & 1)==0) { - v->gshift++; - gmask = gmask >> 1; - } - - /* Compute blue multiplier (mask) and bit shift */ - v->bshift = 0; - bmask = GET_BLUEMASK(v); - while ((bmask & 1)==0) { - v->bshift++; - bmask = bmask >> 1; - } - - /* - * Compute component-to-pixel lookup tables and dithering kernel - */ - { - static GLubyte kernel[16] = { - 0*16, 8*16, 2*16, 10*16, - 12*16, 4*16, 14*16, 6*16, - 3*16, 11*16, 1*16, 9*16, - 15*16, 7*16, 13*16, 5*16, - }; - GLint rBits = _mesa_bitcount(rmask); - GLint gBits = _mesa_bitcount(gmask); - GLint bBits = _mesa_bitcount(bmask); - GLint maxBits; - GLuint i; - - /* convert pixel components in [0,_mask] to RGB values in [0,255] */ - for (i=0; i<=rmask; i++) - v->PixelToR[i] = (unsigned char) ((i * 255) / rmask); - for (i=0; i<=gmask; i++) - v->PixelToG[i] = (unsigned char) ((i * 255) / gmask); - for (i=0; i<=bmask; i++) - v->PixelToB[i] = (unsigned char) ((i * 255) / bmask); - - /* convert RGB values from [0,255] to pixel components */ - - for (i=0;i<256;i++) { - GLint r = gamma_adjust(v->RedGamma, i, 255); - GLint g = gamma_adjust(v->GreenGamma, i, 255); - GLint b = gamma_adjust(v->BlueGamma, i, 255); - v->RtoPixel[i] = (r >> (8-rBits)) << v->rshift; - v->GtoPixel[i] = (g >> (8-gBits)) << v->gshift; - v->BtoPixel[i] = (b >> (8-bBits)) << v->bshift; - } - /* overflow protection */ - for (i=256;i<512;i++) { - v->RtoPixel[i] = v->RtoPixel[255]; - v->GtoPixel[i] = v->GtoPixel[255]; - v->BtoPixel[i] = v->BtoPixel[255]; - } - - /* setup dithering kernel */ - maxBits = rBits; - if (gBits > maxBits) maxBits = gBits; - if (bBits > maxBits) maxBits = bBits; - for (i=0;i<16;i++) { - v->Kernel[i] = kernel[i] >> maxBits; - } - - v->undithered_pf = PF_Truecolor; - v->dithered_pf = (GET_VISUAL_DEPTH(v)<24) ? PF_Dither_True : PF_Truecolor; - } - - /* - * Now check for TrueColor visuals which we can optimize. - */ - if ( GET_REDMASK(v) ==0x0000ff - && GET_GREENMASK(v)==0x00ff00 - && GET_BLUEMASK(v) ==0xff0000 + if ( GET_REDMASK(v) == 0x0000ff + && GET_GREENMASK(v) == 0x00ff00 + && GET_BLUEMASK(v) == 0xff0000 && CHECK_BYTE_ORDER(v) - && v->BitsPerPixel==32 - && v->RedGamma==1.0 && v->GreenGamma==1.0 && v->BlueGamma==1.0) { + && v->BitsPerPixel == 32) { /* common 32 bpp config used on SGI, Sun */ - v->undithered_pf = v->dithered_pf = PF_8A8B8G8R; /* ABGR */ - } - else if (GET_REDMASK(v) == 0xff0000 - && GET_GREENMASK(v)== 0x00ff00 - && GET_BLUEMASK(v) == 0x0000ff - && CHECK_BYTE_ORDER(v) - && v->RedGamma == 1.0 && v->GreenGamma == 1.0 && v->BlueGamma == 1.0){ - if (v->BitsPerPixel==32) { - /* if 32 bpp, and visual indicates 8 bpp alpha channel */ - if (GET_VISUAL_DEPTH(v) == 32 && v->mesa_visual.alphaBits == 8) - v->undithered_pf = v->dithered_pf = PF_8A8R8G8B; /* ARGB */ - else - v->undithered_pf = v->dithered_pf = PF_8R8G8B; /* xRGB */ + assert(0); + return 0 /* PIXEL_FORMAT_U_A8_B8_G8_A8 */; + } + else if ( GET_REDMASK(v) == 0xff0000 + && GET_GREENMASK(v) == 0x00ff00 + && GET_BLUEMASK(v) == 0x0000ff + && CHECK_BYTE_ORDER(v)) { + if (v->BitsPerPixel == 32) { + return PIPE_FORMAT_U_A8_R8_G8_B8; } else if (v->BitsPerPixel == 24) { - v->undithered_pf = v->dithered_pf = PF_8R8G8B24; /* RGB */ + /* not supported yet */ + abort(); + return 0; } } - else if (GET_REDMASK(v) ==0xf800 - && GET_GREENMASK(v)==0x07e0 - && GET_BLUEMASK(v) ==0x001f - && CHECK_BYTE_ORDER(v) - && v->BitsPerPixel==16 - && v->RedGamma==1.0 && v->GreenGamma==1.0 && v->BlueGamma==1.0) { + else if ( GET_REDMASK(v) == 0xf800 + && GET_GREENMASK(v) == 0x07e0 + && GET_BLUEMASK(v) == 0x001f + && CHECK_BYTE_ORDER(v) + && v->BitsPerPixel == 16) { /* 5-6-5 RGB */ - v->undithered_pf = PF_5R6G5B; - v->dithered_pf = PF_Dither_5R6G5B; + return PIPE_FORMAT_U_R5_G6_B5; } -} - - -/** - * Setup RGB rendering for a window with a monochrome visual. - */ -static void -setup_monochrome( XMesaVisual v, XMesaBuffer b ) -{ - (void) b; - v->dithered_pf = v->undithered_pf = PF_1Bit; - /* if black=1 then we must flip pixel values */ - v->bitFlip = (GET_BLACK_PIXEL(v) != 0); + assert(0); + return 0; } - /** * When a context is bound for the first time, we can finally finish * initializing the context's visual and buffer information. @@ -976,10 +448,8 @@ initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, GLboolean rgb_flag, XMesaDrawable window, XMesaColormap cmap) { - int client = 0; - #ifdef XFree86Server - client = (window) ? CLIENT_ID(window->id) : 0; + int client = (window) ? CLIENT_ID(window->id) : 0; #endif ASSERT(!b || b->xm_visual == v); @@ -989,47 +459,22 @@ initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, assert(v->BitsPerPixel > 0); if (rgb_flag == GL_FALSE) { - /* COLOR-INDEXED WINDOW: - * Even if the visual is TrueColor or DirectColor we treat it as - * being color indexed. This is weird but might be useful to someone. - */ - v->dithered_pf = v->undithered_pf = PF_Index; - v->mesa_visual.indexBits = GET_VISUAL_DEPTH(v); + /* COLOR-INDEXED WINDOW: not supported*/ + return GL_FALSE; } else { /* RGB WINDOW: * We support RGB rendering into almost any kind of visual. */ const int xclass = v->mesa_visual.visualType; - if (xclass == GLX_TRUE_COLOR || xclass == GLX_DIRECT_COLOR) { - setup_truecolor( v, b, cmap ); - } - else if (xclass == GLX_STATIC_GRAY && GET_VISUAL_DEPTH(v) == 1) { - setup_monochrome( v, b ); - } - else if (xclass == GLX_GRAY_SCALE || xclass == GLX_STATIC_GRAY) { - if (!setup_grayscale( client, v, b, cmap )) { - return GL_FALSE; - } - } - else if ((xclass == GLX_PSEUDO_COLOR || xclass == GLX_STATIC_COLOR) - && GET_VISUAL_DEPTH(v)>=4 && GET_VISUAL_DEPTH(v)<=16) { - if (!setup_dithered_color( client, v, b, cmap )) { - return GL_FALSE; - } - } - else { - _mesa_warning(NULL, "XMesa: RGB mode rendering not supported in given visual.\n"); + if (xclass != GLX_TRUE_COLOR && xclass == !GLX_DIRECT_COLOR) { + _mesa_warning(NULL, + "XMesa: RGB mode rendering not supported in given visual.\n"); return GL_FALSE; } v->mesa_visual.indexBits = 0; - - if (_mesa_getenv("MESA_NO_DITHER")) { - v->dithered_pf = v->undithered_pf; - } } - /* * If MESA_INFO env var is set print out some debugging info * which can help Brian figure out what's going on when a user @@ -1037,21 +482,14 @@ initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, */ if (_mesa_getenv("MESA_INFO")) { _mesa_printf("X/Mesa visual = %p\n", (void *) v); - _mesa_printf("X/Mesa dithered pf = %u\n", v->dithered_pf); - _mesa_printf("X/Mesa undithered pf = %u\n", v->undithered_pf); _mesa_printf("X/Mesa level = %d\n", v->mesa_visual.level); _mesa_printf("X/Mesa depth = %d\n", GET_VISUAL_DEPTH(v)); _mesa_printf("X/Mesa bits per pixel = %d\n", v->BitsPerPixel); } if (b && window) { - char *data; - - /* Do window-specific initializations */ - /* these should have been set in create_xmesa_buffer */ - ASSERT(b->frontxrb->drawable == window); - ASSERT(b->frontxrb->pixmap == (XMesaPixmap) window); + ASSERT(b->drawable == window); /* Setup for single/double buffering */ if (v->mesa_visual.doubleBufferMode) { @@ -1059,59 +497,13 @@ initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, b->shm = check_for_xshm( v->display ); } - /* X11 graphics contexts */ + /* X11 graphics context */ #ifdef XFree86Server b->gc = CreateScratchGC(v->display, window->depth); #else b->gc = XCreateGC( v->display, window, 0, NULL ); #endif XMesaSetFunction( v->display, b->gc, GXcopy ); - - /* cleargc - for glClear() */ -#ifdef XFree86Server - b->cleargc = CreateScratchGC(v->display, window->depth); -#else - b->cleargc = XCreateGC( v->display, window, 0, NULL ); -#endif - XMesaSetFunction( v->display, b->cleargc, GXcopy ); - - /* - * Don't generate Graphics Expose/NoExpose events in swapbuffers(). - * Patch contributed by Michael Pichler May 15, 1995. - */ -#ifdef XFree86Server - b->swapgc = CreateScratchGC(v->display, window->depth); - { - CARD32 v[1]; - v[0] = FALSE; - dixChangeGC(NullClient, b->swapgc, GCGraphicsExposures, v, NULL); - } -#else - { - XGCValues gcvalues; - gcvalues.graphics_exposures = False; - b->swapgc = XCreateGC(v->display, window, - GCGraphicsExposures, &gcvalues); - } -#endif - XMesaSetFunction( v->display, b->swapgc, GXcopy ); - - /* Initialize the row buffer XImage for use in write_color_span() */ - data = (char*) MALLOC(MAX_WIDTH*4); -#ifdef XFree86Server - b->rowimage = XMesaCreateImage(GET_VISUAL_DEPTH(v), MAX_WIDTH, 1, data); -#else - b->rowimage = XCreateImage( v->display, - v->visinfo->visual, - v->visinfo->depth, - ZPixmap, 0, /*format, offset*/ - data, /*data*/ - MAX_WIDTH, 1, /*width, height*/ - 32, /*bitmap_pad*/ - 0 /*bytes_per_line*/ ); -#endif - if (!b->rowimage) - return GL_FALSE; } return GL_TRUE; @@ -1119,65 +511,6 @@ initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, -/* - * Convert an RGBA color to a pixel value. - */ -unsigned long -xmesa_color_to_pixel(XMesaContext xmesa, - GLubyte r, GLubyte g, GLubyte b, GLubyte a, - GLuint pixelFormat) -{ - GLcontext *ctx = xmesa->st->ctx; - - switch (pixelFormat) { - case PF_Index: - return 0; - case PF_Truecolor: - { - unsigned long p; - PACK_TRUECOLOR( p, r, g, b ); - return p; - } - case PF_8A8B8G8R: - return PACK_8A8B8G8R( r, g, b, a ); - case PF_8A8R8G8B: - return PACK_8A8R8G8B( r, g, b, a ); - case PF_8R8G8B: - /* fall through */ - case PF_8R8G8B24: - return PACK_8R8G8B( r, g, b ); - case PF_5R6G5B: - return PACK_5R6G5B( r, g, b ); - case PF_Dither: - { - DITHER_SETUP; - return DITHER( 1, 0, r, g, b ); - } - case PF_1Bit: - /* 382 = (3*255)/2 */ - return ((r+g+b) > 382) ^ xmesa->xm_visual->bitFlip; - case PF_Lookup: - { - LOOKUP_SETUP; - return LOOKUP( r, g, b ); - } - case PF_Grayscale: - return GRAY_RGB( r, g, b ); - case PF_Dither_True: - /* fall through */ - case PF_Dither_5R6G5B: - { - unsigned long p; - PACK_TRUEDITHER(p, 1, 0, r, g, b); - return p; - } - default: - _mesa_problem(ctx, "Bad pixel format in xmesa_color_to_pixel"); - } - return 0; -} - - #define NUM_VISUAL_TYPES 6 /** @@ -1252,7 +585,6 @@ XMesaVisual XMesaCreateVisual( XMesaDisplay *display, GLint level, GLint visualCaveat ) { - char *gamma; XMesaVisual v; GLint red_bits, green_bits, blue_bits, alpha_bits; @@ -1287,19 +619,6 @@ XMesaVisual XMesaCreateVisual( XMesaDisplay *display, MEMCPY(v->visinfo, visinfo, sizeof(*visinfo)); #endif - /* check for MESA_GAMMA environment variable */ - gamma = _mesa_getenv("MESA_GAMMA"); - if (gamma) { - v->RedGamma = v->GreenGamma = v->BlueGamma = 0.0; - sscanf( gamma, "%f %f %f", &v->RedGamma, &v->GreenGamma, &v->BlueGamma ); - if (v->RedGamma<=0.0) v->RedGamma = 1.0; - if (v->GreenGamma<=0.0) v->GreenGamma = v->RedGamma; - if (v->BlueGamma<=0.0) v->BlueGamma = v->RedGamma; - } - else { - v->RedGamma = v->GreenGamma = v->BlueGamma = 1.0; - } - v->ximage_flag = ximage_flag; #ifdef XFree86Server @@ -1388,66 +707,6 @@ void XMesaDestroyVisual( XMesaVisual v ) -static void -finish_or_flush( GLcontext *ctx ) -{ -#ifdef XFree86Server - /* NOT_NEEDED */ -#else - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - ctx->st->pipe->flush(ctx->st->pipe, 0); - if (xmesa) { - _glthread_LOCK_MUTEX(_xmesa_lock); - XSync( xmesa->display, False ); - _glthread_UNLOCK_MUTEX(_xmesa_lock); - } -#endif -} - - -/** - * Called by glViewport. - * This is a good time for us to poll the current X window size and adjust - * our renderbuffers to match the current window size. - * Remember, we have no opportunity to respond to conventional - * X Resize/StructureNotify events since the X driver has no event loop. - * Thus, we poll. - * Note that this trick isn't fool-proof. If the application never calls - * glViewport, our notion of the current window size may be incorrect. - * That problem led to the GLX_MESA_resize_buffers extension. - */ -static void -xmesa_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) -{ - XMesaContext xmctx = XMESA_CONTEXT(ctx); - XMesaBuffer xmdrawbuf = XMESA_BUFFER(ctx->WinSysDrawBuffer); - XMesaBuffer xmreadbuf = XMESA_BUFFER(ctx->WinSysReadBuffer); - xmesa_check_and_update_buffer_size(xmctx, xmdrawbuf); - xmesa_check_and_update_buffer_size(xmctx, xmreadbuf); - (void) x; - (void) y; - (void) w; - (void) h; -} - - -/** - * Initialize the device driver function table with the functions - * we implement in this driver. - */ -static void -xmesa_init_driver_functions(struct dd_function_table *driver) -{ -#if 0 /* not needed for now */ - driver->Flush = finish_or_flush; - driver->Finish = finish_or_flush; -#else - (void) finish_or_flush; -#endif - driver->Viewport = xmesa_viewport; -} - - /** * Create a new XMesaContext. * \param v the XMesaVisual @@ -1462,6 +721,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) struct pipe_context *pipe; XMesaContext c; GLcontext *mesaCtx; + uint pf; if (firstTime) { _glthread_INIT_MUTEX(_xmesa_lock); @@ -1473,7 +733,9 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) if (!c) return NULL; - pipe = xmesa_create_context( c ); + pf = choose_pixel_format(v); + + pipe = xmesa_create_pipe_context( c, pf ); c->st = st_create_context(pipe, &v->mesa_visual, share_list ? share_list->st : NULL); @@ -1495,27 +757,11 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) mesaCtx->Const.CheckArrayBounds = GL_TRUE; #endif - xmesa_init_driver_functions(&mesaCtx->Driver); - /* finish up xmesa context initializations */ - c->swapbytes = CHECK_BYTE_ORDER(v) ? GL_FALSE : GL_TRUE; c->xm_visual = v; c->xm_buffer = NULL; /* set later by XMesaMakeCurrent */ - c->display = v->display; - c->pixelformat = v->dithered_pf; /* Dithering is enabled by default */ - - /* override these functions, as if the xlib driver were derived from - * the softpipe driver. - */ - pipe->get_tile = xmesa_get_tile; - pipe->put_tile = xmesa_put_tile; - pipe->get_tile_rgba = xmesa_get_tile_rgba; - pipe->put_tile_rgba = xmesa_put_tile_rgba; - c->st->haveFramebufferSurfaces = GL_FALSE; - - /* special pipe->clear function */ - pipe->clear = xmesa_clear; + c->st->haveFramebufferSurfaces = GL_TRUE; return c; } @@ -1735,8 +981,8 @@ XMesaDestroyBuffer(XMesaBuffer b) * and all attached renderbuffers. * Called when: * 1. the first time a buffer is bound to a context. - * 2. from glViewport to poll for window size changes - * 3. from the XMesaResizeBuffers() API function. + * 2. from the XMesaResizeBuffers() API function. + * 3. SwapBuffers. XXX probabaly from xm_flush_frontbuffer() too... * Note: it's possible (and legal) for xmctx to be NULL. That can happen * when resizing a buffer when no rendering context is bound. */ @@ -1744,7 +990,7 @@ void xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer) { GLuint width, height; - xmesa_get_window_size(drawBuffer->display, drawBuffer, &width, &height); + xmesa_get_window_size(drawBuffer->xm_visual->display, drawBuffer, &width, &height); st_resize_framebuffer(drawBuffer->stfb, width, height); } @@ -1774,7 +1020,7 @@ GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, if (&(c->mesa) == _mesa_get_current_context() && c->mesa.DrawBuffer == &drawBuffer->mesa_buffer && c->mesa.ReadBuffer == &readBuffer->mesa_buffer - && XMESA_BUFFER(c->mesa.DrawBuffer)->wasCurrent) { + && xmesa_buffer(c->mesa.DrawBuffer)->wasCurrent) { /* same context and buffer, do nothing */ return GL_TRUE; } @@ -1793,22 +1039,6 @@ GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, st_make_current(c->st, drawBuffer->stfb, readBuffer->stfb); -#if 0 - if (c->xm_visual->mesa_visual.rgbMode) { - /* - * Must recompute and set these pixel values because colormap - * can be different for different windows. - */ - c->clearpixel = xmesa_color_to_pixel( c, - c->clearcolor[0], - c->clearcolor[1], - c->clearcolor[2], - c->clearcolor[3], - c->xm_visual->undithered_pf); - XMesaSetForeground(c->display, drawBuffer->cleargc, c->clearpixel); - } -#endif - /* Solution to Stephane Rehel's problem with glXReleaseBuffersMESA(): */ drawBuffer->wasCurrent = GL_TRUE; } @@ -1834,7 +1064,7 @@ XMesaContext XMesaGetCurrentContext( void ) { GET_CURRENT_CONTEXT(ctx); if (ctx) { - XMesaContext xmesa = XMESA_CONTEXT(ctx); + XMesaContext xmesa = xmesa_context(ctx); return xmesa; } else { @@ -1847,7 +1077,7 @@ XMesaBuffer XMesaGetCurrentBuffer( void ) { GET_CURRENT_CONTEXT(ctx); if (ctx) { - XMesaBuffer xmbuf = XMESA_BUFFER(ctx->DrawBuffer); + XMesaBuffer xmbuf = xmesa_buffer(ctx->DrawBuffer); return xmbuf; } else { @@ -1861,7 +1091,7 @@ XMesaBuffer XMesaGetCurrentReadBuffer( void ) { GET_CURRENT_CONTEXT(ctx); if (ctx) { - return XMESA_BUFFER(ctx->ReadBuffer); + return xmesa_buffer(ctx->ReadBuffer); } else { return 0; @@ -1922,62 +1152,19 @@ GLboolean XMesaSetFXmode( GLint mode ) PUBLIC void XMesaSwapBuffers( XMesaBuffer b ) { - if (!b->backxrb) { - /* single buffered */ - return; - } + struct pipe_surface *surf; /* If we're swapping the buffer associated with the current context * we have to flush any pending rendering commands first. */ st_notify_swapbuffers(b->stfb); - if (b->db_mode) { - if (b->backxrb->ximage) { - /* Copy Ximage (back buf) from client memory to server window */ -#if defined(USE_XSHM) && !defined(XFree86Server) - if (b->shm) { - /*_glthread_LOCK_MUTEX(_xmesa_lock);*/ - XShmPutImage( b->xm_visual->display, b->frontxrb->drawable, - b->swapgc, - b->backxrb->ximage, 0, 0, - 0, 0, xmesa_buffer_width(b), xmesa_buffer_height(b), - False ); - /*_glthread_UNLOCK_MUTEX(_xmesa_lock);*/ - } - else -#endif - { - /*_glthread_LOCK_MUTEX(_xmesa_lock);*/ - XMesaPutImage( b->xm_visual->display, b->frontxrb->drawable, - b->swapgc, - b->backxrb->ximage, 0, 0, - 0, 0, - xmesa_buffer_width(b), xmesa_buffer_height(b)); - /*_glthread_UNLOCK_MUTEX(_xmesa_lock);*/ - } - } - else if (b->backxrb->pixmap) { - /* Copy pixmap (back buf) to window (front buf) on server */ - /*_glthread_LOCK_MUTEX(_xmesa_lock);*/ - XMesaCopyArea( b->xm_visual->display, - b->backxrb->pixmap, /* source drawable */ - b->frontxrb->drawable, /* dest. drawable */ - b->swapgc, - 0, 0, xmesa_buffer_width(b), xmesa_buffer_height(b), - 0, 0 /* dest region */ - ); - /*_glthread_UNLOCK_MUTEX(_xmesa_lock);*/ - } - - if (b->swAlpha) { - GET_CURRENT_CONTEXT(ctx); - _mesa_copy_soft_alpha_renderbuffers(ctx, &b->stfb->Base); - } + surf = st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT); + if (surf) { + xmesa_display_surface(b, surf); } -#if !defined(XFree86Server) - XSync( b->xm_visual->display, False ); -#endif + + xmesa_check_and_update_buffer_size(NULL, b); } @@ -1987,82 +1174,22 @@ void XMesaSwapBuffers( XMesaBuffer b ) */ void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height ) { - if (!b->backxrb) { - /* single buffered */ - return; - } + struct pipe_surface *surf_front + = st_get_framebuffer_surface(b->stfb, ST_SURFACE_FRONT_LEFT); + struct pipe_surface *surf_back + = st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT); + struct pipe_context *pipe = NULL; /* XXX fix */ - /* If we're swapping the buffer associated with the current context - * we have to flush any pending rendering commands first. - */ - st_notify_swapbuffers(b->stfb); + if (!surf_front || !surf_back) + return; - if (b->db_mode) { - int yTop = xmesa_buffer_height(b) - y - height; - if (b->backxrb->ximage) { - /* Copy Ximage from host's memory to server's window */ -#if defined(USE_XSHM) && !defined(XFree86Server) - if (b->shm) { - /* XXX assuming width and height aren't too large! */ - XShmPutImage( b->xm_visual->display, b->frontxrb->drawable, - b->swapgc, - b->backxrb->ximage, x, yTop, - x, yTop, width, height, False ); - /* wait for finished event??? */ - } - else -#endif - { - /* XXX assuming width and height aren't too large! */ - XMesaPutImage( b->xm_visual->display, b->frontxrb->drawable, - b->swapgc, - b->backxrb->ximage, x, yTop, - x, yTop, width, height ); - } - } - else { - /* Copy pixmap to window on server */ - XMesaCopyArea( b->xm_visual->display, - b->backxrb->pixmap, /* source drawable */ - b->frontxrb->drawable, /* dest. drawable */ - b->swapgc, - x, yTop, width, height, /* source region */ - x, yTop /* dest region */ - ); - } - } + pipe->surface_copy(pipe, + surf_front, x, y, /* dest */ + surf_back, x, y, /* src */ + width, height); } -/* - * Return a pointer to the XMesa backbuffer Pixmap or XImage. This function - * is a way to get "under the hood" of X/Mesa so one can manipulate the - * back buffer directly. - * Output: pixmap - pointer to back buffer's Pixmap, or 0 - * ximage - pointer to back buffer's XImage, or NULL - * Return: GL_TRUE = context is double buffered - * GL_FALSE = context is single buffered - */ -#ifndef XFree86Server -GLboolean XMesaGetBackBuffer( XMesaBuffer b, - XMesaPixmap *pixmap, - XMesaImage **ximage ) -{ - if (b->db_mode) { - if (pixmap) - *pixmap = b->backxrb->pixmap; - if (ximage) - *ximage = b->backxrb->ximage; - return GL_TRUE; - } - else { - *pixmap = 0; - *ximage = NULL; - return GL_FALSE; - } -} -#endif /* XFree86Server */ - /* * Return the depth buffer associated with an XMesaBuffer. @@ -2075,33 +1202,32 @@ GLboolean XMesaGetBackBuffer( XMesaBuffer b, GLboolean XMesaGetDepthBuffer( XMesaBuffer b, GLint *width, GLint *height, GLint *bytesPerValue, void **buffer ) { - struct gl_renderbuffer *rb - = b->stfb->Base.Attachment[BUFFER_DEPTH].Renderbuffer; - if (!rb || !rb->Data) { + struct pipe_surface *surf + = st_get_framebuffer_surface(b->stfb, ST_SURFACE_DEPTH); + if (surf) { + *width = surf->width; + *height = surf->pitch; + *bytesPerValue = surf->cpp; + *buffer = surf->map; + return GL_TRUE; + } + else { *width = 0; *height = 0; *bytesPerValue = 0; *buffer = 0; return GL_FALSE; } - else { - *width = xmesa_buffer_width(b); - *height = xmesa_buffer_height(b); - *bytesPerValue = b->stfb->Base.Visual.depthBits <= 16 - ? sizeof(GLushort) : sizeof(GLuint); - *buffer = rb->Data; - return GL_TRUE; - } } void XMesaFlush( XMesaContext c ) { - if (c && c->display) { + if (c && c->xm_visual->display) { #ifdef XFree86Server /* NOT_NEEDED */ #else - XSync( c->display, False ); + XSync( c->xm_visual->display, False ); #endif } } @@ -2128,7 +1254,7 @@ XMesaBuffer XMesaFindBuffer( XMesaDisplay *dpy, XMesaDrawable d ) { XMesaBuffer b; for (b=XMesaBufferList; b; b=b->Next) { - if (b->frontxrb->drawable == d && b->display == dpy) { + if (b->drawable == d && b->xm_visual->display == dpy) { return b; } } @@ -2144,7 +1270,7 @@ void xmesa_destroy_buffers_on_display(XMesaDisplay *dpy) XMesaBuffer b, next; for (b = XMesaBufferList; b; b = next) { next = b->Next; - if (b->display == dpy) { + if (b->xm_visual->display == dpy) { xmesa_free_buffer(b); } } @@ -2160,12 +1286,15 @@ void XMesaGarbageCollect( void ) XMesaBuffer b, next; for (b=XMesaBufferList; b; b=next) { next = b->Next; - if (b->display && b->frontxrb->drawable && b->type == WINDOW) { + if (b->xm_visual && + b->xm_visual->display && + b->drawable && + b->type == WINDOW) { #ifdef XFree86Server /* NOT_NEEDED */ #else - XSync(b->display, False); - if (!window_exists( b->display, b->frontxrb->drawable )) { + XSync(b->xm_visual->display, False); + if (!window_exists( b->xm_visual->display, b->drawable )) { /* found a dead window, free the ancillary info */ XMesaDestroyBuffer( b ); } @@ -2179,55 +1308,7 @@ unsigned long XMesaDitherColor( XMesaContext xmesa, GLint x, GLint y, GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha ) { - GLcontext *ctx = xmesa->st->ctx; - GLint r = (GLint) (red * 255.0F); - GLint g = (GLint) (green * 255.0F); - GLint b = (GLint) (blue * 255.0F); - GLint a = (GLint) (alpha * 255.0F); - - switch (xmesa->pixelformat) { - case PF_Index: - return 0; - case PF_Truecolor: - { - unsigned long p; - PACK_TRUECOLOR( p, r, g, b ); - return p; - } - case PF_8A8B8G8R: - return PACK_8A8B8G8R( r, g, b, a ); - case PF_8A8R8G8B: - return PACK_8A8R8G8B( r, g, b, a ); - case PF_8R8G8B: - return PACK_8R8G8B( r, g, b ); - case PF_5R6G5B: - return PACK_5R6G5B( r, g, b ); - case PF_Dither: - { - DITHER_SETUP; - return DITHER( x, y, r, g, b ); - } - case PF_1Bit: - /* 382 = (3*255)/2 */ - return ((r+g+b) > 382) ^ xmesa->xm_visual->bitFlip; - case PF_Lookup: - { - LOOKUP_SETUP; - return LOOKUP( r, g, b ); - } - case PF_Grayscale: - return GRAY_RGB( r, g, b ); - case PF_Dither_5R6G5B: - /* fall through */ - case PF_Dither_True: - { - unsigned long p; - PACK_TRUEDITHER(p, x, y, r, g, b); - return p; - } - default: - _mesa_problem(NULL, "Bad pixel format in XMesaDitherColor"); - } + /* no longer supported */ return 0; } @@ -2240,149 +1321,19 @@ PUBLIC void XMesaResizeBuffers( XMesaBuffer b ) { GET_CURRENT_CONTEXT(ctx); - XMesaContext xmctx = XMESA_CONTEXT(ctx); + XMesaContext xmctx = xmesa_context(ctx); if (!xmctx) return; xmesa_check_and_update_buffer_size(xmctx, b); } -static GLint -xbuffer_to_renderbuffer(int buffer) -{ - assert(MAX_AUX_BUFFERS <= 4); - - switch (buffer) { - case GLX_FRONT_LEFT_EXT: - return BUFFER_FRONT_LEFT; - case GLX_FRONT_RIGHT_EXT: - return BUFFER_FRONT_RIGHT; - case GLX_BACK_LEFT_EXT: - return BUFFER_BACK_LEFT; - case GLX_BACK_RIGHT_EXT: - return BUFFER_BACK_RIGHT; - case GLX_AUX0_EXT: - return BUFFER_AUX0; - case GLX_AUX1_EXT: - return BUFFER_AUX1; - case GLX_AUX2_EXT: - return BUFFER_AUX2; - case GLX_AUX3_EXT: - return BUFFER_AUX3; - case GLX_AUX4_EXT: - case GLX_AUX5_EXT: - case GLX_AUX6_EXT: - case GLX_AUX7_EXT: - case GLX_AUX8_EXT: - case GLX_AUX9_EXT: - default: - /* BadValue error */ - return -1; - } -} PUBLIC void XMesaBindTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer, const int *attrib_list) { -#if 0 - GET_CURRENT_CONTEXT(ctx); - const GLuint unit = ctx->Texture.CurrentUnit; - struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *texObj; -#endif - struct gl_renderbuffer *rb; - struct xmesa_renderbuffer *xrb; - GLint b; - XMesaImage *img = NULL; - GLboolean freeImg = GL_FALSE; - - b = xbuffer_to_renderbuffer(buffer); - if (b < 0) - return; - - if (drawable->TextureFormat == GLX_TEXTURE_FORMAT_NONE_EXT) - return; /* BadMatch error */ - - rb = drawable->stfb->Base.Attachment[b].Renderbuffer; - if (!rb) { - /* invalid buffer */ - return; - } - xrb = xmesa_renderbuffer(rb); - -#if 0 - switch (drawable->TextureTarget) { - case GLX_TEXTURE_1D_EXT: - texObj = texUnit->Current1D; - break; - case GLX_TEXTURE_2D_EXT: - texObj = texUnit->Current2D; - break; - case GLX_TEXTURE_RECTANGLE_EXT: - texObj = texUnit->CurrentRect; - break; - default: - return; /* BadMatch error */ - } -#endif - - /* - * The following is a quick and simple way to implement - * BindTexImage. The better way is to write some new FetchTexel() - * functions which would extract texels from XImages. We'd still - * need to use GetImage when texturing from a Pixmap (front buffer) - * but texturing from a back buffer (XImage) would avoid an image - * copy. - */ - - /* get XImage */ - if (xrb->pixmap) { - img = XMesaGetImage(dpy, xrb->pixmap, 0, 0, rb->Width, rb->Height, ~0L, - ZPixmap); - freeImg = GL_TRUE; - } - else if (xrb->ximage) { - img = xrb->ximage; - } - - /* store the XImage as a new texture image */ - if (img) { - GLenum format, type, intFormat; - if (img->bits_per_pixel == 32) { - format = GL_BGRA; - type = GL_UNSIGNED_BYTE; - intFormat = GL_RGBA; - } - else if (img->bits_per_pixel == 24) { - format = GL_BGR; - type = GL_UNSIGNED_BYTE; - intFormat = GL_RGB; - } - else if (img->bits_per_pixel == 16) { - format = GL_BGR; - type = GL_UNSIGNED_SHORT_5_6_5; - intFormat = GL_RGB; - } - else { - _mesa_problem(NULL, "Unexpected XImage format in XMesaBindTexImage"); - return; - } - if (drawable->TextureFormat == GLX_TEXTURE_FORMAT_RGBA_EXT) { - intFormat = GL_RGBA; - } - else if (drawable->TextureFormat == GLX_TEXTURE_FORMAT_RGB_EXT) { - intFormat = GL_RGB; - } - - _mesa_TexImage2D(GL_TEXTURE_2D, 0, intFormat, rb->Width, rb->Height, 0, - format, type, img->data); - - if (freeImg) { - XMesaDestroyImage(img); - } - } } @@ -2390,10 +1341,5 @@ XMesaBindTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer, PUBLIC void XMesaReleaseTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer) { - const GLint b = xbuffer_to_renderbuffer(buffer); - if (b < 0) - return; - - /* no-op for now */ } diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c index 6fc2f4ba63..e49eafbb09 100644 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ b/src/mesa/pipe/xlib/xm_buffer.c @@ -41,7 +41,6 @@ #include "state_tracker/st_context.h" -#if defined(USE_XSHM) && !defined(XFree86Server) static volatile int mesaXErrorFlag = 0; /** @@ -458,8 +457,8 @@ xmesa_delete_framebuffer(struct gl_framebuffer *fb) if (!xmesa_find_buffer(b->display, b->cmap, b)) { #ifdef XFree86Server int client = 0; - if (b->frontxrb->drawable) - client = CLIENT_ID(b->frontxrb->drawable->id); + if (b->drawable) + client = CLIENT_ID(b->drawable->id); (void)FreeColors(b->cmap, client, b->num_alloced, b->alloced_colors, 0); #else @@ -478,21 +477,6 @@ xmesa_delete_framebuffer(struct gl_framebuffer *fb) if (fb->Visual.doubleBufferMode) { /* free back ximage/pixmap/shmregion */ - if (b->backxrb->ximage) { -#if defined(USE_XSHM) && !defined(XFree86Server) - if (b->shm) { - XShmDetach( b->display, &b->shminfo ); - XDestroyImage( b->backxrb->ximage ); - shmdt( b->shminfo.shmaddr ); - } - else -#endif - XMesaDestroyImage( b->backxrb->ximage ); - b->backxrb->ximage = NULL; - } - if (b->backxrb->pixmap) { - XMesaFreePixmap( b->display, b->backxrb->pixmap ); - } } if (b->rowimage) { diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c index 4a54b5c7e1..6f5c7ac6ec 100644 --- a/src/mesa/pipe/xlib/xm_surface.c +++ b/src/mesa/pipe/xlib/xm_surface.c @@ -70,689 +70,6 @@ const int xmesa_kernel1[16] = { -#define CLIP_TILE \ - do { \ - if (x >= ps->width) \ - return; \ - if (y >= ps->height) \ - return; \ - if (x + w > ps->width) \ - w = ps->width - x; \ - if (y + h > ps->height) \ - h = ps->height -y; \ - } while(0) - - - -/* - * The following functions are used to trap XGetImage() calls which - * generate BadMatch errors if the drawable isn't mapped. - */ - -#ifndef XFree86Server -static int caught_xgetimage_error = 0; -static int (*old_xerror_handler)( XMesaDisplay *dpy, XErrorEvent *ev ); -static unsigned long xgetimage_serial; - -/* - * This is the error handler which will be called if XGetImage fails. - */ -static int xgetimage_error_handler( XMesaDisplay *dpy, XErrorEvent *ev ) -{ - if (ev->serial==xgetimage_serial && ev->error_code==BadMatch) { - /* caught the expected error */ - caught_xgetimage_error = 0; - } - else { - /* call the original X error handler, if any. otherwise ignore */ - if (old_xerror_handler) { - (*old_xerror_handler)( dpy, ev ); - } - } - return 0; -} - - -/* - * Call this right before XGetImage to setup error trap. - */ -static void catch_xgetimage_errors( XMesaDisplay *dpy ) -{ - xgetimage_serial = NextRequest( dpy ); - old_xerror_handler = XSetErrorHandler( xgetimage_error_handler ); - caught_xgetimage_error = 0; -} - - -/* - * Call this right after XGetImage to check if an error occured. - */ -static int check_xgetimage_errors( void ) -{ - /* restore old handler */ - (void) XSetErrorHandler( old_xerror_handler ); - /* return 0=no error, 1=error caught */ - return caught_xgetimage_error; -} -#endif - - -/** - * Wrapper for XGetImage() that catches BadMatch errors that can occur - * when the window is unmapped or the x/y/w/h extend beyond the window - * bounds. - * If build into xserver, wrap the internal GetImage method. - */ -static XMesaImage * -xget_image(XMesaDisplay *dpy, Drawable d, int x, int y, uint w, uint h) -{ -#ifdef XFree86Server - uint bpp = 4; /* XXX fix this */ - XMesaImage *ximage = (XMesaImage *) malloc(sizeof(XMesaImage)); - if (ximage) { - ximage->data = malloc(width * height * bpp); - } - (*dpy->GetImage)(d, x, y, w, h, ZPixmap, ~0L, (pointer)ximage->data); - ximage->width = w; - ximage->height = h; - ximage->bytes_per_row = w * bpp; - return ximage; -#else - int error; - XMesaImage *ximage; - catch_xgetimage_errors(dpy); - ximage = XGetImage(dpy, d, x, y, w, h, AllPlanes, ZPixmap); - error = check_xgetimage_errors(); - return ximage; -#endif -} - - - -/** - * Return raw pixels from pixmap or XImage. - */ -void -xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, void *p, int dst_stride) -{ - const uint w0 = w; - struct xmesa_surface *xms = xmesa_surface(ps); - XMesaImage *ximage = NULL; - ubyte *dst = (ubyte *) p; - uint i; - - if (!xms->drawable && !xms->ximage) { - /* not an X surface */ - softpipe_get_tile(pipe, ps, x, y, w, h, p, dst_stride); - return; - } - - CLIP_TILE; - - if (!xms->ximage) { - /* XImage = pixmap data */ - assert(xms->drawable); - ximage = xget_image(xms->display, xms->drawable, x, y, w, h); - if (!ximage) - return; - x = y = 0; - } - else { - ximage = xms->ximage; - } - - /* this could be optimized/simplified */ - switch (ps->format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - if (!dst_stride) { - dst_stride = w0 * 4; - } - for (i = 0; i < h; i++) { - memcpy(dst, ximage->data + y * ximage->bytes_per_line + x * 4, 4 * w); - dst += dst_stride; - } - break; - case PIPE_FORMAT_U_R5_G6_B5: - if (!dst_stride) { - dst_stride = w0 * 2; - } - for (i = 0; i < h; i++) { - memcpy(dst, ximage->data + y * ximage->bytes_per_line + x * 2, 4 * 2); - dst += dst_stride; - } - break; - default: - assert(0); - } - - if (!xms->ximage) { - XMesaDestroyImage(ximage); - } -} - - -/** - * Put raw pixels into pixmap or XImage. - */ -void -xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const void *p, int src_stride) -{ - const uint w0 = w; - struct xmesa_surface *xms = xmesa_surface(ps); - const ubyte *src = (const ubyte *) p; - XMesaImage *ximage; - - if (!xms->drawable && !xms->ximage) { - /* not an X surface */ - softpipe_put_tile(pipe, ps, x, y, w, h, p, src_stride); - return; - } - - CLIP_TILE; - - if (xms->ximage) { - /* put to ximage */ - ximage = xms->ximage; - char *dst; - uint i; - - /* this could be optimized/simplified */ - switch (ps->format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - if (!src_stride) { - src_stride = w0 * 4; - } - dst = ximage->data + y * ximage->bytes_per_line + x * 4; - for (i = 0; i < h; i++) { - memcpy(dst, src, w * 4); - dst += ximage->bytes_per_line; - src += src_stride; - } - break; - case PIPE_FORMAT_U_R5_G6_B5: - if (!src_stride) { - src_stride = w0 * 2; - } - dst = ximage->data + y * ximage->bytes_per_line + x * 2; - for (i = 0; i < h; i++) { - memcpy(dst, src, w * 2); - dst += ximage->bytes_per_line; - src += src_stride; - } - break; - default: - assert(0); - } - } - else { - /* put to pixmap/window */ - /* Create temp XImage for data */ -#ifdef XFree86Server - ximage = XMesaCreateImage(GET_VISUAL_DEPTH(v), w, h, p); -#else - XVisualInfo *visinfo = xms->xrb->Parent->xm_visual->visinfo; - ximage = XCreateImage(xms->display, - visinfo->visual, - visinfo->depth, - ZPixmap, 0, /* format, offset */ - (char *) p, /* data */ - w, h, /* width, height */ - 32, /* bitmap_pad */ - 0); /* bytes_per_line */ -#endif - - /* send XImage data to pixmap */ - XPutImage(xms->display, xms->drawable, xms->gc, - ximage, 0, 0, x, y, w, h); - /* clean-up */ - ximage->data = NULL; /* prevents freeing user data at 'p' */ - XMesaDestroyImage(ximage); - } -} - - -void -xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *pixels) -{ - const uint w0 = w; - struct xmesa_surface *xms = xmesa_surface(ps); - XMesaImage *ximage = NULL; - float *pRow = pixels; - uint i, j; - - if (!xms->drawable && !xms->ximage) { - /* not an X surface */ - softpipe_get_tile_rgba(pipe, ps, x, y, w, h, pixels); - return; - } - - CLIP_TILE; - - if (!xms->ximage) { - /* XImage = pixmap data */ - assert(xms->drawable); - ximage = xget_image(xms->display, xms->drawable, x, y, w, h); - if (!ximage) - return; - x = y = 0; - } - else { - ximage = xms->ximage; - } - - switch (ps->format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - { - const uint *src - = (uint *) (ximage->data + y * ximage->bytes_per_line + x * 4); - for (i = 0; i < h; i++) { - float *p = pRow; - for (j = 0; j < w; j++) { - uint pix = src[j]; - ubyte r = ((pix >> 16) & 0xff); - ubyte g = ((pix >> 8) & 0xff); - ubyte b = ( pix & 0xff); - ubyte a = ((pix >> 24) & 0xff); - p[0] = UBYTE_TO_FLOAT(r); - p[1] = UBYTE_TO_FLOAT(g); - p[2] = UBYTE_TO_FLOAT(b); - p[3] = UBYTE_TO_FLOAT(a); - p += 4; - } - src += ximage->width; - pRow += 4 * w0; - } - } - break; - case PIPE_FORMAT_U_B8_G8_R8_A8: - { - const uint *src - = (uint *) (ximage->data + y * ximage->bytes_per_line + x * 4); - for (i = 0; i < h; i++) { - float *p = pRow; - for (j = 0; j < w; j++) { - uint pix = src[j]; - ubyte r = ((pix >> 8) & 0xff); - ubyte g = ((pix >> 16) & 0xff); - ubyte b = ((pix >> 24) & 0xff); - ubyte a = ( pix & 0xff); - p[0] = UBYTE_TO_FLOAT(r); - p[1] = UBYTE_TO_FLOAT(g); - p[2] = UBYTE_TO_FLOAT(b); - p[3] = UBYTE_TO_FLOAT(a); - p += 4; - } - src += ximage->width; - pRow += 4 * w0; - } - } - break; - case PIPE_FORMAT_U_R5_G6_B5: - { - ushort *src - = (ushort *) (ximage->data + y * ximage->bytes_per_line + x * 2); - for (i = 0; i < h; i++) { - float *p = pRow; - for (j = 0; j < w; j++) { - uint pix = src[j]; - ubyte r = (pix >> 8) | ((pix >> 13) & 0x7); - ubyte g = (pix >> 3) | ((pix >> 9) & 0x3); - ubyte b = ((pix & 0x1f) << 3) | ((pix >> 2) & 0x3); - p[0] = UBYTE_TO_FLOAT(r); - p[1] = UBYTE_TO_FLOAT(g); - p[2] = UBYTE_TO_FLOAT(b); - p[3] = 1.0; - p += 4; - } - src += ximage->width; - pRow += 4 * w0; - } - } - break; - default: - fprintf(stderr, "Bad format in xmesa_get_tile_rgba()\n"); - assert(0); - } - - if (!xms->ximage) { - XMesaDestroyImage(ximage); - } -} - - -void -xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *pixels) -{ - const uint x0 = x, y0 = y, w0 = w; - struct xmesa_surface *xms = xmesa_surface(ps); - XMesaImage *ximage; - uint i, j; - - if (!xms->drawable && !xms->ximage) { - /* not an X surface */ - softpipe_put_tile_rgba(pipe, ps, x, y, w, h, pixels); - return; - } - - CLIP_TILE; - - if (!xms->ximage) { - /* create temp XImage */ - char *data = (char *) malloc(w * h * 4); -#ifdef XFree86Server - ximage = XMesaCreateImage(GET_VISUAL_DEPTH(v), w, h, data); -#else - XVisualInfo *visinfo = xms->xrb->Parent->xm_visual->visinfo; - ximage = XCreateImage(xms->display, - visinfo->visual, - visinfo->depth, - ZPixmap, 0, /* format, offset */ - data, /* data */ - w, h, /* width, height */ - 32, /* bitmap_pad */ - 0); /* bytes_per_line */ -#endif - x = y = 0; - } - else { - ximage = xms->ximage; - } - - /* convert floats to ximage's format */ - switch (ps->format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - { - uint *dst - = (uint *) (ximage->data + y * ximage->bytes_per_line + x * 4); - const float *pRow = pixels; - for (i = 0; i < h; i++) { - const float *p = pRow; - for (j = 0; j < w; j++) { - ubyte r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, p[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, p[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, p[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, p[3]); - dst[j] = PACK_8A8R8G8B(r, g, b, a); - p += 4; - } - dst += ximage->width; - pRow += 4 * w0; - } - } - break; - case PIPE_FORMAT_U_B8_G8_R8_A8: - { - uint *dst - = (uint *) (ximage->data + y * ximage->bytes_per_line + x * 4); - const float *pRow = pixels; - for (i = 0; i < h; i++) { - const float *p = pRow; - for (j = 0; j < w; j++) { - ubyte r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, p[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, p[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, p[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, p[3]); - dst[j] = PACK_8B8G8R8A(r, g, b, a); - p += 4; - } - dst += ximage->width; - pRow += 4 * w0; - } - } - break; - case PIPE_FORMAT_U_R5_G6_B5: - { - ushort *dst = - (ushort *) (ximage->data + y * ximage->bytes_per_line + x * 2); - const float *pRow = pixels; - for (i = 0; i < h; i++) { - const float *p = pRow; - for (j = 0; j < w; j++) { - ubyte r, g, b; - UNCLAMPED_FLOAT_TO_UBYTE(r, p[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, p[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, p[2]); - dst[j] = PACK_5R6G5B(r, g, b); - p += 4; - } - dst += ximage->width; - pRow += 4 * w0; - } - } - break; - - default: - fprintf(stderr, "Bad format in xmesa_put_tile_rgba()\n"); - assert(0); - } - - if (!xms->ximage) { - /* send XImage data to pixmap */ - XPutImage(xms->display, xms->drawable, xms->gc, - ximage, 0, 0, x0, y0, w, h); - /* clean-up */ - free(ximage->data); - ximage->data = NULL; - XMesaDestroyImage(ximage); - } -} - - -static void -clear_pixmap_surface(struct pipe_context *pipe, struct pipe_surface *ps, - uint value) -{ - struct xmesa_surface *xms = xmesa_surface(ps); - assert(xms); - assert(xms->display); - assert(xms->drawable); - assert(xms->gc); - XMesaSetForeground( xms->display, xms->gc, value ); - XMesaFillRectangle( xms->display, xms->drawable, xms->gc, - 0, 0, ps->width, ps->height); -} - -static void -clear_nbit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, - uint value) -{ - struct xmesa_surface *xms = xmesa_surface(ps); - int width = xms->surface.width; - int height = xms->surface.height; - int i, j; - for (j = 0; j < height; j++) { - for (i = 0; i < width; i++) { - XMesaPutPixel(xms->ximage, i, j, value); - } - } -} - -static void -clear_8bit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, - uint value) -{ - struct xmesa_surface *xms = xmesa_surface(ps); - memset(xms->ximage->data, - value, - xms->ximage->bytes_per_line * xms->ximage->height); -} - -static void -clear_16bit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, - uint value) -{ - struct xmesa_surface *xms = xmesa_surface(ps); - const int n = xms->ximage->width * xms->ximage->height; - ushort *dst = (ushort *) xms->ximage->data; - int i; - for (i = 0; i < n; i++) { - dst[i] = value; - } -} - - -/* Optimized code provided by Nozomi Ytow */ -static void -clear_24bit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, - uint value) -{ - struct xmesa_surface *xms = xmesa_surface(ps); - const ubyte r = (value ) & 0xff; - const ubyte g = (value >> 8) & 0xff; - const ubyte b = (value >> 16) & 0xff; - - if (r == g && g == b) { - /* same value for all three components (gray) */ - memset(xms->ximage->data, r, - xms->ximage->bytes_per_line * xms->ximage->height); - } - else { - /* non-gray clear color */ - const int n = xms->ximage->width * xms->ximage->height; - int i; - bgr_t *ptr3 = (bgr_t *) xms->ximage->data; - for (i = 0; i < n; i++) { - ptr3->r = r; - ptr3->g = g; - ptr3->b = b; - ptr3++; - } - } -} - -static void -clear_32bit_ximage_surface(struct pipe_context *pipe, struct pipe_surface *ps, - uint value) -{ - struct xmesa_surface *xms = xmesa_surface(ps); - - if (value == 0) { - /* common case */ - memset(xms->ximage->data, value, - xms->ximage->bytes_per_line * xms->ximage->height); - } - else { - const int n = xms->ximage->width * xms->ximage->height; - uint *dst = (uint *) xms->ximage->data; - int i; - for (i = 0; i < n; i++) - dst[i] = value; - } -} - - - - -/** - * Called to create a pipe_surface for each X renderbuffer. - * Note: this is being used instead of pipe->surface_alloc() since we - * have special/unique quad read/write functions for X. - */ -struct pipe_surface * -xmesa_new_color_surface(struct pipe_winsys *winsys, GLuint pipeFormat) -{ - struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); - - assert(pipeFormat); - - xms->surface.format = pipeFormat; - xms->surface.refcount = 1; - xms->surface.winsys = winsys; - - /* Note, the buffer we allocate doesn't actually have any storage - * since we're drawing into an XImage or Pixmap. - * The surface's size will get set in the xmesa_alloc_front/back_storage() - * functions. - */ - xms->surface.buffer = winsys->buffer_create(winsys, 0x0); - - return &xms->surface; -} - - -/** - * Called via pipe->surface_alloc() to create new surfaces (textures, - * renderbuffers, etc. - */ -struct pipe_surface * -xmesa_surface_alloc(struct pipe_context *pipe, GLuint pipeFormat) -{ - struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); - - assert(pipe); - assert(pipeFormat); - - xms->surface.format = pipeFormat; - xms->surface.refcount = 1; - xms->surface.winsys = pipe->winsys; - - return &xms->surface; -} - - -/** - * Called via pipe->clear() to clear entire surface to a certain value. - * If the surface is not an X pixmap or XImage, pass the call to - * softpipe_clear(). - */ -void -xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, uint value) -{ - struct xmesa_surface *xms = xmesa_surface(ps); - - /* XXX actually, we should just discard any cached tiles from this - * surface since we don't want to accidentally re-use them after clearing. - */ - pipe->flush(pipe, 0); - - { - struct softpipe_context *sp = softpipe_context(pipe); - if (ps == sp_tile_cache_get_surface(sp, sp->cbuf_cache[0])) { - float clear[4]; - clear[0] = 0.2; /* XXX hack */ - clear[1] = 0.2; - clear[2] = 0.2; - clear[3] = 0.2; - sp_tile_cache_clear(sp->cbuf_cache[0], clear); - } - } - -#if 1 - (void) clear_8bit_ximage_surface; - (void) clear_24bit_ximage_surface; -#endif - - if (xms->ximage) { - /* back color buffer */ - switch (xms->surface.format) { - case PIPE_FORMAT_U_R5_G6_B5: - clear_16bit_ximage_surface(pipe, ps, value); - break; - case PIPE_FORMAT_U_A8_R8_G8_B8: - case PIPE_FORMAT_U_B8_G8_R8_A8: - clear_32bit_ximage_surface(pipe, ps, value); - break; - default: - clear_nbit_ximage_surface(pipe, ps, value); - break; - } - } - else if (xms->drawable) { - /* front color buffer */ - clear_pixmap_surface(pipe, ps, value); - } - else { - /* other kind of buffer */ - softpipe_clear(pipe, ps, value); - } -} - /** XXX unfinished sketch... */ struct pipe_surface * diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 295174d4bb..b9dd3e3b6e 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -38,6 +38,8 @@ #include "main/macros.h" #include "pipe/p_winsys.h" +#include "pipe/p_format.h" +#include "pipe/p_context.h" #include "pipe/softpipe/sp_winsys.h" @@ -54,17 +56,52 @@ struct xm_buffer }; +struct xmesa_surface +{ + struct pipe_surface surface; + /* no extra fields for now */ +}; + + +/** + * Derived from softpipe_winsys. + * We just need one extra field which indicates the pixel format to use for + * drawing surfaces so that we're compatible with the XVisual/window format. + */ +struct xmesa_softpipe_winsys +{ + struct softpipe_winsys spws; + uint pixelformat; +}; + + -/* Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque +/** Cast wrapper */ +static INLINE struct xmesa_surface * +xmesa_surface(struct pipe_surface *ps) +{ + assert(0); + return (struct xmesa_surface *) ps; +} + +/** cast wrapper */ +static INLINE struct xmesa_softpipe_winsys * +xmesa_softpipe_winsys(struct softpipe_winsys *spws) +{ + return (struct xmesa_softpipe_winsys *) spws; +} + +/** + * Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque * buffer pointer... */ -static inline struct xm_buffer * +static INLINE struct xm_buffer * xm_bo( struct pipe_buffer_handle *bo ) { return (struct xm_buffer *) bo; } -static inline struct pipe_buffer_handle * +static INLINE struct pipe_buffer_handle * pipe_bo( struct xm_buffer *bo ) { return (struct pipe_buffer_handle *) bo; @@ -156,6 +193,33 @@ xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, memcpy(data, b + offset, size); } + +/** + * Display/copy the image in the surface into the X window specified + * by the XMesaBuffer. + */ +void +xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) +{ + XImage *ximage = b->tempImage; + struct xm_buffer *xm_buf = xm_bo(surf->buffer); + + /* check that the XImage has been previously initialized */ + assert(ximage->format); + assert(ximage->bitmap_unit); + + /* update XImage's fields */ + ximage->width = surf->width; + ximage->height = surf->height; + ximage->bytes_per_line = surf->pitch * (ximage->bits_per_pixel / 8); + ximage->data = xm_buf->data; + + /* display image in Window */ + XPutImage(b->xm_visual->display, b->drawable, b->gc, + ximage, 0, 0, 0, 0, surf->width, surf->height); +} + + static void xm_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surf, @@ -166,9 +230,13 @@ xm_flush_frontbuffer(struct pipe_winsys *pws, * If we instead did front buffer rendering to a temporary XImage, * this would be the place to copy the Ximage to the on-screen Window. */ + XMesaContext xmctx = (XMesaContext) context_private; + + xmesa_display_surface(xmctx->xm_buffer, surf); } + static void xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) { @@ -178,6 +246,7 @@ xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) va_end( args ); } + static const char * xm_get_name(struct pipe_winsys *pws) { @@ -219,7 +288,6 @@ round_up(unsigned n, unsigned multiple) return (n + multiple - 1) & ~(multiple - 1); } - static unsigned xm_surface_pitch(struct pipe_winsys *winsys, unsigned cpp, unsigned width, unsigned flags) @@ -243,12 +311,7 @@ xm_surface_alloc(struct pipe_winsys *ws, GLuint pipeFormat) xms->surface.format = pipeFormat; xms->surface.refcount = 1; xms->surface.winsys = ws; -#if 0 - /* - * This is really just a softpipe surface, not an XImage/Pixmap surface. - */ - softpipe_init_surface_funcs(&xms->surface); -#endif + return &xms->surface; } @@ -261,7 +324,7 @@ xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) surf->refcount--; if (surf->refcount == 0) { if (surf->buffer) - winsys->buffer_reference(winsys, &surf->buffer, NULL); + winsys->buffer_reference(winsys, &surf->buffer, NULL); free(surf); } *s = NULL; @@ -274,7 +337,7 @@ xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) * For Xlib, this is a singleton object. * Nothing special for the Xlib driver so no subclassing or anything. */ -struct pipe_winsys * +static struct pipe_winsys * xmesa_get_pipe_winsys(void) { static struct pipe_winsys *ws = NULL; @@ -308,59 +371,65 @@ xmesa_get_pipe_winsys(void) /** - * XXX this depends on the depths supported by the screen (8/16/32/etc). - * Maybe when we're about to create a context/drawable we create a new - * softpipe_winsys object that corresponds to the specified screen... + * The winsys being queried will have been created at glXCreateContext + * time, with a pixel format corresponding to the context's visual. * - * Also, this query only really matters for on-screen drawables. - * For textures and FBOs we (softpipe) can support any format.o + * XXX we should pass a flag indicating if the format is going to be + * use for a drawing surface vs. a texture. In the later case, we + * can support any format. */ static boolean xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) { - /* Any format supported by softpipe can be listed here. - * This query is not used for allocating window-system color buffers - * (which would depend on the screen depth/bpp). - */ - switch (format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - case PIPE_FORMAT_S_R16_G16_B16_A16: - case PIPE_FORMAT_S8_Z24: - case PIPE_FORMAT_U_S8: - case PIPE_FORMAT_U_Z16: - case PIPE_FORMAT_U_Z32: + struct xmesa_softpipe_winsys *xmws = xmesa_softpipe_winsys(sws); + + if (format == xmws->pixelformat) { return TRUE; - default: - return FALSE; - }; + } + else { + /* non-color / window surface format */ + switch (format) { + case PIPE_FORMAT_S_R16_G16_B16_A16: + case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_U_S8: + case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_U_Z32: + return TRUE; + default: + return FALSE; + }; + } } /** * Return pointer to a softpipe_winsys object. - * For Xlib, this is a singleton object. */ static struct softpipe_winsys * -xmesa_get_softpipe_winsys(void) +xmesa_get_softpipe_winsys(uint pixelformat) { - static struct softpipe_winsys *spws = NULL; + struct xmesa_softpipe_winsys *xmws + = CALLOC_STRUCT(xmesa_softpipe_winsys); + if (!xmws) + return NULL; - if (!spws) { - spws = CALLOC_STRUCT(softpipe_winsys); - if (spws) { - spws->is_format_supported = xmesa_is_format_supported; - } - } + xmws->spws.is_format_supported = xmesa_is_format_supported; + xmws->pixelformat = pixelformat; - return spws; + return &xmws->spws; } struct pipe_context * -xmesa_create_context(XMesaContext xmesa) +xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) { struct pipe_winsys *pws = xmesa_get_pipe_winsys(); - struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(); + struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(pixelformat); + struct pipe_context *pipe; - return softpipe_create( pws, spws ); + pipe = softpipe_create( pws, spws ); + if (pipe) + pipe->priv = xmesa; + + return pipe; } diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index a067dfd3e3..e8c2a597a6 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -32,42 +32,15 @@ #ifdef XFree86Server #include "xm_image.h" #endif -#include "state_tracker/st_cb_fbo.h" + #include "state_tracker/st_context.h" #include "state_tracker/st_public.h" -#include "pipe/softpipe/sp_context.h" -#include "pipe/softpipe/sp_surface.h" extern _glthread_Mutex _xmesa_lock; extern XMesaBuffer XMesaBufferList; -/* for PF_8R8G8B24 pixel format */ -typedef struct { - GLubyte b; - GLubyte g; - GLubyte r; -} bgr_t; - - -/** Framebuffer pixel formats */ -enum pixel_format { - PF_Index, /**< Color Index mode */ - PF_Truecolor, /**< TrueColor or DirectColor, any depth */ - PF_Dither_True, /**< TrueColor with dithering */ - PF_8A8R8G8B, /**< 32-bit TrueColor: 8-A, 8-R, 8-G, 8-B bits */ - PF_8A8B8G8R, /**< 32-bit TrueColor: 8-A, 8-B, 8-G, 8-R bits */ - PF_8R8G8B, /**< 32-bit TrueColor: 8-R, 8-G, 8-B bits */ - PF_8R8G8B24, /**< 24-bit TrueColor: 8-R, 8-G, 8-B bits */ - PF_5R6G5B, /**< 16-bit TrueColor: 5-R, 6-G, 5-B bits */ - PF_Dither, /**< Color-mapped RGB with dither */ - PF_Lookup, /**< Color-mapped RGB without dither */ - PF_1Bit, /**< monochrome dithering of RGB */ - PF_Grayscale, /**< Grayscale or StaticGray */ - PF_Dither_5R6G5B /**< 16-bit dithered TrueColor: 5-R, 6-G, 5-B */ -}; - /** * Visual inforation, derived from GLvisual. @@ -86,47 +59,17 @@ struct xmesa_visual { GLint BitsPerPixel; /* True bits per pixel for XImages */ GLboolean ximage_flag; /* Use XImage for back buffer (not pixmap)? */ - - enum pixel_format dithered_pf; /* Pixel format when dithering */ - enum pixel_format undithered_pf;/* Pixel format when not dithering */ - - GLfloat RedGamma; /* Gamma values, 1.0 is default */ - GLfloat GreenGamma; - GLfloat BlueGamma; - - /* For PF_TRUECOLOR */ - GLint rshift, gshift, bshift;/* Pixel color component shifts */ - GLubyte Kernel[16]; /* Dither kernel */ - unsigned long RtoPixel[512]; /* RGB to pixel conversion */ - unsigned long GtoPixel[512]; - unsigned long BtoPixel[512]; - GLubyte PixelToR[256]; /* Pixel to RGB conversion */ - GLubyte PixelToG[256]; - GLubyte PixelToB[256]; - - /* For PF_1BIT */ - int bitFlip; }; /** - * Context info, derived from GLcontext. + * Context info, derived from st_context. * Basically corresponds to a GLXContext. */ struct xmesa_context { struct st_context *st; - - XMesaVisual xm_visual; /* Describes the buffers */ - XMesaBuffer xm_buffer; /* current span/point/line/triangle buffer */ - - XMesaDisplay *display; /* == xm_visual->display */ - GLboolean swapbytes; /* Host byte order != display byte order? */ - GLboolean direct; /* Direct rendering context? */ - - enum pixel_format pixelformat; - - GLubyte clearcolor[4]; /* current clearing color */ - unsigned long clearpixel; /* current clearing pixel value */ + XMesaVisual xm_visual; /** pixel format info */ + XMesaBuffer xm_buffer; /** current drawbuffer */ }; @@ -141,46 +84,6 @@ typedef enum { } BufferType; -/** Values for db_mode: */ -/*@{*/ -#define BACK_PIXMAP 1 -#define BACK_XIMAGE 2 -/*@}*/ - - -/** - * An xmesa_renderbuffer represents the back or front color buffer. - * For the front color buffer: - * is the X window - * For the back color buffer: - * Either or will be used, never both. - * In any case, always equals . - * For stand-alone Mesa, we could merge and into one - * field. We don't do that for the server-side GLcore module because - * pixmaps and drawables are different and we'd need a bunch of casts. - */ -struct xmesa_renderbuffer -{ - struct st_renderbuffer St; /**< Base class (XXX temporary?) */ - - XMesaBuffer Parent; /**< The XMesaBuffer this renderbuffer belongs to */ - XMesaDrawable drawable; /* Usually the X window ID */ - XMesaPixmap pixmap; /* Back color buffer */ - XMesaImage *ximage; /* The back buffer, if not using a Pixmap */ - - GLubyte *origin1; /* used for PIXEL_ADDR1 macro */ - GLint width1; - GLushort *origin2; /* used for PIXEL_ADDR2 macro */ - GLint width2; - GLubyte *origin3; /* used for PIXEL_ADDR3 macro */ - GLint width3; - GLuint *origin4; /* used for PIXEL_ADDR4 macro */ - GLint width4; - - GLint bottom; /* used for FLIP macro, equals height - 1 */ -}; - - /** * Framebuffer information, derived from. * Basically corresponds to a GLXDrawable. @@ -190,22 +93,13 @@ struct xmesa_buffer { GLboolean wasCurrent; /* was ever the current buffer? */ XMesaVisual xm_visual; /* the X/Mesa visual */ - - XMesaDisplay *display; - BufferType type; /* window, pixmap, pbuffer or glxwindow */ - - struct xmesa_renderbuffer *frontxrb; /* front color renderbuffer */ - struct xmesa_renderbuffer *backxrb; /* back color renderbuffer */ - + XMesaDrawable drawable; /* Usually the X window ID */ XMesaColormap cmap; /* the X colormap */ + BufferType type; /* window, pixmap, pbuffer or glxwindow */ + XMesaImage *tempImage; unsigned long selectedEvents;/* for pbuffers only */ - GLint db_mode; /* 0 = single buffered */ - /* BACK_PIXMAP = use Pixmap for back buffer */ - /* BACK_XIMAGE = use XImage for back buffer */ - GLboolean swAlpha; - GLuint shm; /* X Shared Memory extension status: */ /* 0 = not available */ /* 1 = XImage support available */ @@ -214,33 +108,7 @@ struct xmesa_buffer { XShmSegmentInfo shminfo; #endif - XMesaImage *rowimage; /* Used for optimized span writing */ - XMesaPixmap stipple_pixmap; /* For polygon stippling */ - XMesaGC stipple_gc; /* For polygon stippling */ - XMesaGC gc; /* scratch GC for span, line, tri drawing */ - XMesaGC cleargc; /* GC for clearing the color buffer */ - XMesaGC swapgc; /* GC for swapping the color buffers */ - - /* The following are here instead of in the XMesaVisual - * because they depend on the window's colormap. - */ - - /* For PF_DITHER, PF_LOOKUP, PF_GRAYSCALE */ - unsigned long color_table[576]; /* RGB -> pixel value */ - - /* For PF_DITHER, PF_LOOKUP, PF_GRAYSCALE */ - GLubyte pixel_to_r[65536]; /* pixel value -> red */ - GLubyte pixel_to_g[65536]; /* pixel value -> green */ - GLubyte pixel_to_b[65536]; /* pixel value -> blue */ - - /* Used to do XAllocColor/XFreeColors accounting: */ - int num_alloced; -#if defined(XFree86Server) - Pixel alloced_colors[256]; -#else - unsigned long alloced_colors[256]; -#endif /* GLX_EXT_texture_from_pixmap */ GLint TextureTarget; /** GLX_TEXTURE_1D_EXT, for example */ @@ -251,325 +119,38 @@ struct xmesa_buffer { }; -/** - * If pixelformat==PF_TRUECOLOR: - */ -#define PACK_TRUECOLOR( PIXEL, R, G, B ) \ - PIXEL = xmesa->xm_visual->RtoPixel[R] \ - | xmesa->xm_visual->GtoPixel[G] \ - | xmesa->xm_visual->BtoPixel[B]; \ - - -/** - * If pixelformat==PF_TRUEDITHER: - */ -#define PACK_TRUEDITHER( PIXEL, X, Y, R, G, B ) \ -{ \ - int d = xmesa->xm_visual->Kernel[((X)&3) | (((Y)&3)<<2)]; \ - PIXEL = xmesa->xm_visual->RtoPixel[(R)+d] \ - | xmesa->xm_visual->GtoPixel[(G)+d] \ - | xmesa->xm_visual->BtoPixel[(B)+d]; \ -} - - - -/** - * If pixelformat==PF_8A8B8G8R: - */ -#define PACK_8A8B8G8R( R, G, B, A ) \ - ( ((A) << 24) | ((B) << 16) | ((G) << 8) | (R) ) - -/** - * Like PACK_8A8B8G8R() but don't use alpha. This is usually an acceptable - * shortcut. - */ -#define PACK_8B8G8R( R, G, B ) ( ((B) << 16) | ((G) << 8) | (R) ) - -#define PACK_8B8G8R8A( R, G, B, A ) \ - ( ((B) << 24) | ((G) << 16) | ((R) << 8) | (A) ) - - - - -/** - * If pixelformat==PF_8R8G8B: - */ -#define PACK_8R8G8B( R, G, B) ( ((R) << 16) | ((G) << 8) | (B) ) - - -/** - * If pixelformat==PF_5R6G5B: - */ -#define PACK_5R6G5B( R, G, B) ( (((R) & 0xf8) << 8) | (((G) & 0xfc) << 3) | ((B) >> 3) ) - - -/** - * If pixelformat==PF_8A8R8G8B: - */ -#define PACK_8A8R8G8B( R, G, B, A ) \ - ( ((A) << 24) | ((R) << 16) | ((G) << 8) | (B) ) - - - -/** - * If pixelformat==PF_DITHER: - * - * Improved 8-bit RGB dithering code contributed by Bob Mercier - * (mercier@hollywood.cinenet.net). Thanks Bob! - */ -#ifdef DITHER666 -# define DITH_R 6 -# define DITH_G 6 -# define DITH_B 6 -# define DITH_MIX(r,g,b) (((r) * DITH_G + (g)) * DITH_B + (b)) -#else -# define DITH_R 5 -# define DITH_G 9 -# define DITH_B 5 -# define DITH_MIX(r,g,b) (((g) << 6) | ((b) << 3) | (r)) -#endif -#define DITH_DX 4 -#define DITH_DY 4 -#define DITH_N (DITH_DX * DITH_DY) - -#define _dither(C, c, d) (((unsigned)((DITH_N * (C - 1) + 1) * c + d)) >> 12) - -#define MAXC 256 -extern const int xmesa_kernel8[DITH_DY * DITH_DX]; - -/* Dither for random X,Y */ -#define DITHER_SETUP \ - int __d; \ - unsigned long *ctable = XMESA_BUFFER(ctx->DrawBuffer)->color_table; - -#define DITHER( X, Y, R, G, B ) \ - (__d = xmesa_kernel8[(((Y)&3)<<2) | ((X)&3)], \ - ctable[DITH_MIX(_dither(DITH_R, (R), __d), \ - _dither(DITH_G, (G), __d), \ - _dither(DITH_B, (B), __d))]) - -/* Dither for random X, fixed Y */ -#define XDITHER_SETUP(Y) \ - int __d; \ - unsigned long *ctable = XMESA_BUFFER(ctx->DrawBuffer)->color_table; \ - const int *kernel = &xmesa_kernel8[ ((Y)&3) << 2 ]; - -#define XDITHER( X, R, G, B ) \ - (__d = kernel[(X)&3], \ - ctable[DITH_MIX(_dither(DITH_R, (R), __d), \ - _dither(DITH_G, (G), __d), \ - _dither(DITH_B, (B), __d))]) - - - -/* - * Dithering for flat-shaded triangles. Precompute all 16 possible - * pixel values given the triangle's RGB color. Contributed by Martin Shenk. - */ -#define FLAT_DITHER_SETUP( R, G, B ) \ - GLushort ditherValues[16]; \ - { \ - unsigned long *ctable = XMESA_BUFFER(ctx->DrawBuffer)->color_table; \ - int msdr = (DITH_N*((DITH_R)-1)+1) * (R); \ - int msdg = (DITH_N*((DITH_G)-1)+1) * (G); \ - int msdb = (DITH_N*((DITH_B)-1)+1) * (B); \ - int i; \ - for (i=0;i<16;i++) { \ - int k = xmesa_kernel8[i]; \ - int j = DITH_MIX( (msdr+k)>>12, (msdg+k)>>12, (msdb+k)>>12 );\ - ditherValues[i] = (GLushort) ctable[j]; \ - } \ - } - -#define FLAT_DITHER_ROW_SETUP(Y) \ - GLushort *ditherRow = ditherValues + ( ((Y)&3) << 2); - -#define FLAT_DITHER(X) ditherRow[(X)&3] - - - -/** - * If pixelformat==PF_LOOKUP: - */ -#define _dither_lookup(C, c) (((unsigned)((DITH_N * (C - 1) + 1) * c)) >> 12) - -#define LOOKUP_SETUP \ - unsigned long *ctable = XMESA_BUFFER(ctx->DrawBuffer)->color_table - -#define LOOKUP( R, G, B ) \ - ctable[DITH_MIX(_dither_lookup(DITH_R, (R)), \ - _dither_lookup(DITH_G, (G)), \ - _dither_lookup(DITH_B, (B)))] - - - -/** - * If pixelformat==PF_1BIT: - */ -extern const int xmesa_kernel1[16]; - -#define SETUP_1BIT int bitFlip = xmesa->xm_visual->bitFlip -#define DITHER_1BIT( X, Y, R, G, B ) \ - (( ((int)(R)+(int)(G)+(int)(B)) > xmesa_kernel1[(((Y)&3) << 2) | ((X)&3)] ) ^ bitFlip) - - - -/** - * If pixelformat==PF_GRAYSCALE: - */ -#define GRAY_RGB( R, G, B ) XMESA_BUFFER(ctx->DrawBuffer)->color_table[((R) + (G) + (B))/3] - - - -/** - * Converts a GL window Y coord to an X window Y coord: - */ -#define YFLIP(XRB, Y) ((XRB)->bottom - (Y)) - - -/** - * Return the address of a 1, 2 or 4-byte pixel in the buffer's XImage: - * X==0 is left, Y==0 is bottom. - */ -#define PIXEL_ADDR1(XRB, X, Y) \ - ( (XRB)->origin1 - (Y) * (XRB)->width1 + (X) ) - -#define PIXEL_ADDR2(XRB, X, Y) \ - ( (XRB)->origin2 - (Y) * (XRB)->width2 + (X) ) - -#define PIXEL_ADDR3(XRB, X, Y) \ - ( (bgr_t *) ( (XRB)->origin3 - (Y) * (XRB)->width3 + 3 * (X) )) - -#define PIXEL_ADDR4(XRB, X, Y) \ - ( (XRB)->origin4 - (Y) * (XRB)->width4 + (X) ) - - - -/* - * External functions: - */ - -extern struct xmesa_renderbuffer * -xmesa_create_renderbuffer(struct pipe_winsys *winsys, - GLuint name, XMesaVisual xmvis, - GLboolean backBuffer); - -extern void -xmesa_delete_framebuffer(struct gl_framebuffer *fb); - -extern XMesaBuffer -xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis); - -extern unsigned long -xmesa_color_to_pixel( XMesaContext xmesa, - GLubyte r, GLubyte g, GLubyte b, GLubyte a, - GLuint pixelFormat ); - -extern void -xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b, - GLuint *width, GLuint *height); - -extern void -xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer); - - -extern void -xmesa_destroy_buffers_on_display(XMesaDisplay *dpy); - - -/** - * Using a function instead of an ordinary cast is safer. - */ -static INLINE struct xmesa_renderbuffer * -xmesa_renderbuffer(struct gl_renderbuffer *rb) -{ - return (struct xmesa_renderbuffer *) rb; -} - - -/** - * Return pointer to XMesaContext corresponding to a Mesa GLcontext. - * Since we're using structure containment, it's just a cast!. - * XXX should use inlined function for better type safety. - */ +/** cast wrapper */ static INLINE XMesaContext -XMESA_CONTEXT(GLcontext *ctx) +xmesa_context(GLcontext *ctx) { return (XMesaContext) ctx->DriverCtx; } -/** - * Return pointer to XMesaBuffer corresponding to a Mesa GLframebuffer. - * Since we're using structure containment, it's just a cast!. - * XXX should use inlined function for better type safety. - */ +/** cast wrapper */ static INLINE XMesaBuffer -XMESA_BUFFER(GLframebuffer *fb) +xmesa_buffer(GLframebuffer *fb) { struct st_framebuffer *stfb = (struct st_framebuffer *) fb; return (XMesaBuffer) st_framebuffer_private(stfb); } - -struct pipe_context; - -struct xmesa_surface -{ - struct pipe_surface surface; - struct xmesa_renderbuffer *xrb; - XMesaDisplay *display; - BufferType type; - XMesaDrawable drawable; - XMesaImage *ximage; - XMesaGC gc; -}; - - -/** Cast wrapper */ -static INLINE struct xmesa_surface * -xmesa_surface(struct pipe_surface *ps) -{ - return (struct xmesa_surface *) ps; -} - - extern void -xmesa_clear(struct pipe_context *pipe, struct pipe_surface *ps, uint value); - -extern struct pipe_context * -xmesa_create_context(XMesaContext xm); - -extern struct pipe_surface * -xmesa_surface_alloc(struct pipe_context *pipe, GLuint format); - -extern struct pipe_surface * -xmesa_new_color_surface(struct pipe_winsys *winsys, GLuint format); - -extern struct pipe_winsys * -xmesa_get_pipe_winsys(void); - -extern void -xmesa_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, void *p, int dst_stride); +xmesa_delete_framebuffer(struct gl_framebuffer *fb); -extern void -xmesa_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const void *p, int src_stride); +extern XMesaBuffer +xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis); extern void -xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *p); +xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer); extern void -xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *p); - +xmesa_destroy_buffers_on_display(XMesaDisplay *dpy); -extern struct pipe_surface * -xmesa_create_front_surface(XMesaVisual vis, Window win); +extern struct pipe_context * +xmesa_create_pipe_context(XMesaContext xm, uint pixelformat); static INLINE GLuint xmesa_buffer_width(XMesaBuffer b) @@ -583,5 +164,7 @@ xmesa_buffer_height(XMesaBuffer b) return b->stfb->Base.Height; } +extern void +xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf); #endif diff --git a/src/mesa/sources b/src/mesa/sources index 596d3b89d5..8286e558d6 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -334,9 +334,7 @@ X11_DRIVER_SOURCES = \ pipe/xlib/fakeglx.c \ pipe/xlib/xfonts.c \ pipe/xlib/xm_api.c \ - pipe/xlib/xm_buffer.c \ - pipe/xlib/xm_winsys.c \ - pipe/xlib/xm_surface.c + pipe/xlib/xm_winsys.c OSMESA_DRIVER_SOURCES = \ drivers/osmesa/osmesa.c -- cgit v1.2.3 From e920f367ede4e275c25c5fa1929c943af3baef46 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 07:58:31 -0700 Subject: Remove obsolete xlib driver files --- src/mesa/pipe/xlib/xm_buffer.c | 490 ---------------------------------------- src/mesa/pipe/xlib/xm_surface.c | 94 -------- 2 files changed, 584 deletions(-) delete mode 100644 src/mesa/pipe/xlib/xm_buffer.c delete mode 100644 src/mesa/pipe/xlib/xm_surface.c (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_buffer.c b/src/mesa/pipe/xlib/xm_buffer.c deleted file mode 100644 index e49eafbb09..0000000000 --- a/src/mesa/pipe/xlib/xm_buffer.c +++ /dev/null @@ -1,490 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5.2 - * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/** - * \file xm_buffer.h - * Framebuffer and renderbuffer-related functions. - */ - - -#include "GL/xmesa.h" -#include "glxheader.h" -#include "xmesaP.h" -#include "main/imports.h" -#include "main/framebuffer.h" -#include "main/renderbuffer.h" -#include "pipe/p_state.h" -#include "pipe/p_defines.h" -#include "pipe/p_winsys.h" -#include "state_tracker/st_context.h" - - -static volatile int mesaXErrorFlag = 0; - -/** - * Catches potential Xlib errors. - */ -static int -mesaHandleXError(XMesaDisplay *dpy, XErrorEvent *event) -{ - (void) dpy; - (void) event; - mesaXErrorFlag = 1; - return 0; -} - -/** - * Allocate a shared memory XImage back buffer for the given XMesaBuffer. - * Return: GL_TRUE if success, GL_FALSE if error - */ -static GLboolean -alloc_back_shm_ximage(XMesaBuffer b, GLuint width, GLuint height) -{ - /* - * We have to do a _lot_ of error checking here to be sure we can - * really use the XSHM extension. It seems different servers trigger - * errors at different points if the extension won't work. Therefore - * we have to be very careful... - */ - GC gc; - int (*old_handler)(XMesaDisplay *, XErrorEvent *); - - if (width == 0 || height == 0) { - /* this will be true the first time we're called on 'b' */ - return GL_FALSE; - } - - b->backxrb->ximage = XShmCreateImage(b->xm_visual->display, - b->xm_visual->visinfo->visual, - b->xm_visual->visinfo->depth, - ZPixmap, NULL, &b->shminfo, - width, height); - if (b->backxrb->ximage == NULL) { - _mesa_warning(NULL, "alloc_back_buffer: Shared memory error (XShmCreateImage), disabling.\n"); - b->shm = 0; - return GL_FALSE; - } - - b->shminfo.shmid = shmget(IPC_PRIVATE, b->backxrb->ximage->bytes_per_line - * b->backxrb->ximage->height, IPC_CREAT|0777); - if (b->shminfo.shmid < 0) { - _mesa_warning(NULL, "shmget failed while allocating back buffer.\n"); - XDestroyImage(b->backxrb->ximage); - b->backxrb->ximage = NULL; - _mesa_warning(NULL, "alloc_back_buffer: Shared memory error (shmget), disabling.\n"); - b->shm = 0; - return GL_FALSE; - } - - b->shminfo.shmaddr = b->backxrb->ximage->data - = (char*)shmat(b->shminfo.shmid, 0, 0); - if (b->shminfo.shmaddr == (char *) -1) { - _mesa_warning(NULL, "shmat() failed while allocating back buffer.\n"); - XDestroyImage(b->backxrb->ximage); - shmctl(b->shminfo.shmid, IPC_RMID, 0); - b->backxrb->ximage = NULL; - _mesa_warning(NULL, "alloc_back_buffer: Shared memory error (shmat), disabling.\n"); - b->shm = 0; - return GL_FALSE; - } - - b->shminfo.readOnly = False; - mesaXErrorFlag = 0; - old_handler = XSetErrorHandler(mesaHandleXError); - /* This may trigger the X protocol error we're ready to catch: */ - XShmAttach(b->xm_visual->display, &b->shminfo); - XSync(b->xm_visual->display, False); - - if (mesaXErrorFlag) { - /* we are on a remote display, this error is normal, don't print it */ - XFlush(b->xm_visual->display); - mesaXErrorFlag = 0; - XDestroyImage(b->backxrb->ximage); - shmdt(b->shminfo.shmaddr); - shmctl(b->shminfo.shmid, IPC_RMID, 0); - b->backxrb->ximage = NULL; - b->shm = 0; - (void) XSetErrorHandler(old_handler); - return GL_FALSE; - } - - shmctl(b->shminfo.shmid, IPC_RMID, 0); /* nobody else needs it */ - - /* Finally, try an XShmPutImage to be really sure the extension works */ - gc = XCreateGC(b->xm_visual->display, b->frontxrb->drawable, 0, NULL); - XShmPutImage(b->xm_visual->display, b->frontxrb->drawable, gc, - b->backxrb->ximage, 0, 0, 0, 0, 1, 1 /*one pixel*/, False); - XSync(b->xm_visual->display, False); - XFreeGC(b->xm_visual->display, gc); - (void) XSetErrorHandler(old_handler); - if (mesaXErrorFlag) { - XFlush(b->xm_visual->display); - mesaXErrorFlag = 0; - XDestroyImage(b->backxrb->ximage); - shmdt(b->shminfo.shmaddr); - shmctl(b->shminfo.shmid, IPC_RMID, 0); - b->backxrb->ximage = NULL; - b->shm = 0; - return GL_FALSE; - } - - return GL_TRUE; -} -#else -static GLboolean -alloc_back_shm_ximage(XMesaBuffer b, GLuint width, GLuint height) -{ - /* Can't compile XSHM support */ - return GL_FALSE; -} -#endif - - -/** - * \return LSBFirst or MSBFirst - */ -static int host_byte_order( void ) -{ - int i = 1; - char *cptr = (char *) &i; - return (*cptr==1) ? LSBFirst : MSBFirst; -} - - -/** - * Setup an off-screen pixmap or Ximage to use as the back buffer. - * Input: b - the X/Mesa buffer - */ -static void -alloc_back_buffer(XMesaBuffer b, GLuint width, GLuint height) -{ - if (b->db_mode == BACK_XIMAGE) { - /* Deallocate the old backxrb->ximage, if any */ - if (b->backxrb->ximage) { -#if defined(USE_XSHM) && !defined(XFree86Server) - if (b->shm) { - XShmDetach(b->xm_visual->display, &b->shminfo); - XDestroyImage(b->backxrb->ximage); - shmdt(b->shminfo.shmaddr); - } - else -#endif - XMesaDestroyImage(b->backxrb->ximage); - b->backxrb->ximage = NULL; - } - - if (width == 0 || height == 0) - return; - - /* Allocate new back buffer */ - if (b->shm == 0 || !alloc_back_shm_ximage(b, width, height)) { - /* Allocate a regular XImage for the back buffer. */ -#ifdef XFree86Server - b->backxrb->ximage = XMesaCreateImage(b->xm_visual->BitsPerPixel, - width, height, NULL); -#else - b->backxrb->ximage = XCreateImage(b->xm_visual->display, - b->xm_visual->visinfo->visual, - GET_VISUAL_DEPTH(b->xm_visual), - ZPixmap, 0, /* format, offset */ - NULL, - width, height, - 8, 0); /* pad, bytes_per_line */ -#endif - if (!b->backxrb->ximage) { - _mesa_warning(NULL, "alloc_back_buffer: XCreateImage failed.\n"); - return; - } - b->backxrb->ximage->data = (char *) MALLOC(b->backxrb->ximage->height - * b->backxrb->ximage->bytes_per_line); - if (!b->backxrb->ximage->data) { - _mesa_warning(NULL, "alloc_back_buffer: MALLOC failed.\n"); - XMesaDestroyImage(b->backxrb->ximage); - b->backxrb->ximage = NULL; - } - } - b->backxrb->pixmap = None; - } - else if (b->db_mode == BACK_PIXMAP) { - /* Free the old back pixmap */ - if (b->backxrb->pixmap) { - XMesaFreePixmap(b->xm_visual->display, b->backxrb->pixmap); - b->backxrb->pixmap = 0; - } - - if (width > 0 && height > 0) { - /* Allocate new back pixmap */ - b->backxrb->pixmap = XMesaCreatePixmap(b->xm_visual->display, - b->frontxrb->drawable, - width, height, - GET_VISUAL_DEPTH(b->xm_visual)); - } - - b->backxrb->ximage = NULL; - } -} - - -static void -xmesa_delete_renderbuffer(struct gl_renderbuffer *rb) -{ - struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); - if (xrb->St.surface) { - struct pipe_winsys *ws = xrb->St.surface->winsys; - ws->surface_release(ws, &xrb->St.surface); - } - - /* XXX Note: the ximage or Pixmap attached to this renderbuffer - * should probably get freed here, but that's currently done in - * XMesaDestroyBuffer(). - */ - _mesa_free(rb); -} - - -static void -finish_surface_init(GLcontext *ctx, struct xmesa_renderbuffer *xrb) -{ - struct pipe_context *pipe = ctx->st->pipe; - if (!xrb->St.surface->buffer) { - xrb->St.surface->buffer = pipe->winsys->buffer_create(pipe->winsys, 0x0); - } -} - - -/** - * Reallocate renderbuffer storage for front color buffer. - * Called via gl_renderbuffer::AllocStorage() - */ -static GLboolean -xmesa_alloc_front_storage(GLcontext *ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, GLuint width, GLuint height) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); - struct xmesa_surface *xms = xmesa_surface(xrb->St.surface); - - /* just clear these to be sure we don't accidentally use them */ - xrb->origin1 = NULL; - xrb->origin2 = NULL; - xrb->origin3 = NULL; - xrb->origin4 = NULL; - - /* for the FLIP macro: */ - xrb->bottom = height - 1; - - rb->Width = width; - rb->Height = height; - rb->InternalFormat = internalFormat; - - if (!xrb->St.surface || !xrb->St.surface->buffer) - finish_surface_init(ctx, xrb); - - /* surface info */ - xms->surface.width = width; - xms->surface.height = height; - xms->display = xmesa->display; - xms->drawable = xrb->drawable; - xms->gc = xrb->Parent->cleargc; - xms->ximage = NULL; - - return GL_TRUE; -} - - -/** - * Reallocate renderbuffer storage for back color buffer. - * Called via gl_renderbuffer::AllocStorage() - */ -static GLboolean -xmesa_alloc_back_storage(GLcontext *ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, GLuint width, GLuint height) -{ - const XMesaContext xmesa = XMESA_CONTEXT(ctx); - struct xmesa_renderbuffer *xrb = xmesa_renderbuffer(rb); - struct xmesa_surface *xms = xmesa_surface(xrb->St.surface); - - /* reallocate the back buffer XImage or Pixmap */ - assert(xrb->Parent); - alloc_back_buffer(xrb->Parent, width, height); - - /* same as front buffer */ - /* XXX why is this here? */ - (void) xmesa_alloc_front_storage(ctx, rb, internalFormat, width, height); - - /* plus... */ - if (xrb->ximage) { - /* Needed by PIXELADDR1 macro */ - xrb->width1 = xrb->ximage->bytes_per_line; - xrb->origin1 = (GLubyte *) xrb->ximage->data + xrb->width1 * (height - 1); - - /* Needed by PIXELADDR2 macro */ - xrb->width2 = xrb->ximage->bytes_per_line / 2; - xrb->origin2 = (GLushort *) xrb->ximage->data + xrb->width2 * (height - 1); - - /* Needed by PIXELADDR3 macro */ - xrb->width3 = xrb->ximage->bytes_per_line; - xrb->origin3 = (GLubyte *) xrb->ximage->data + xrb->width3 * (height - 1); - - /* Needed by PIXELADDR4 macro */ - xrb->width4 = xrb->ximage->width; - xrb->origin4 = (GLuint *) xrb->ximage->data + xrb->width4 * (height - 1); - } - else { - /* out of memory or buffer size is 0 x 0 */ - xrb->width1 = xrb->width2 = xrb->width3 = xrb->width4 = 0; - xrb->origin1 = NULL; - xrb->origin2 = NULL; - xrb->origin3 = NULL; - xrb->origin4 = NULL; - } - - if (!xrb->St.surface || !xrb->St.surface->buffer) - finish_surface_init(ctx, xrb); - - xrb->St.surface->width = width; - xrb->St.surface->height = height; - - /* surface info */ - xms->surface.width = width; - xms->surface.height = height; - xms->display = xmesa->display; - xms->drawable = xrb->drawable; - xms->gc = xrb->Parent->cleargc; - xms->ximage = xrb->ximage; - - return GL_TRUE; -} - - -/** - * Called to create the front/back color renderbuffers, not user-created - * renderbuffers. - */ -struct xmesa_renderbuffer * -xmesa_create_renderbuffer(struct pipe_winsys *winsys, - GLuint name, XMesaVisual xmvis, - GLboolean backBuffer) -{ - const GLvisual *visual = &xmvis->mesa_visual; - int byteOrder = ImageByteOrder(xmvis->display); /* LSBFirst or MSBFirst */ - struct xmesa_renderbuffer *xrb = CALLOC_STRUCT(xmesa_renderbuffer); - if (xrb) { - GLuint name = 0; - GLuint pipeFormat = 0; - struct xmesa_surface *xms; - - _mesa_init_renderbuffer(&xrb->St.Base, name); - - xrb->St.Base.Delete = xmesa_delete_renderbuffer; - if (backBuffer) - xrb->St.Base.AllocStorage = xmesa_alloc_back_storage; - else - xrb->St.Base.AllocStorage = xmesa_alloc_front_storage; - - if (visual->rgbMode) { - xrb->St.Base.InternalFormat = GL_RGBA; - xrb->St.Base._BaseFormat = GL_RGBA; - xrb->St.Base.DataType = GL_UNSIGNED_BYTE; - xrb->St.Base.RedBits = visual->redBits; - xrb->St.Base.GreenBits = visual->greenBits; - xrb->St.Base.BlueBits = visual->blueBits; - xrb->St.Base.AlphaBits = visual->alphaBits; - if (visual->redMask == 0xff0000 && - visual->greenMask == 0x00ff00 && - visual->blueMask == 0x0000ff) { - if (host_byte_order() != byteOrder) { - pipeFormat = PIPE_FORMAT_U_B8_G8_R8_A8; - /*printf("Using format B8_G8_R8_A8 (LE dpy)\n");*/ - } - else { - pipeFormat = PIPE_FORMAT_U_A8_R8_G8_B8; - /*printf("Using format A8_R8_G8_B8 (BE dpy)\n");*/ - } - } - else { - assert(0); - } - } - else { - xrb->St.Base.InternalFormat = GL_COLOR_INDEX; - xrb->St.Base._BaseFormat = GL_COLOR_INDEX; - xrb->St.Base.DataType = GL_UNSIGNED_INT; - xrb->St.Base.IndexBits = visual->indexBits; - } - /* only need to set Red/Green/EtcBits fields for user-created RBs */ - xrb->St.surface = xmesa_new_color_surface(winsys, pipeFormat); - xms = (struct xmesa_surface *) xrb->St.surface; - xms->xrb = xrb; - } - return xrb; -} - - -/** - * Called via gl_framebuffer::Delete() method when this buffer - * is _really_ being deleted. - */ -void -xmesa_delete_framebuffer(struct gl_framebuffer *fb) -{ - XMesaBuffer b = XMESA_BUFFER(fb); - - if (b->num_alloced > 0) { - /* If no other buffer uses this X colormap then free the colors. */ - if (!xmesa_find_buffer(b->display, b->cmap, b)) { -#ifdef XFree86Server - int client = 0; - if (b->drawable) - client = CLIENT_ID(b->drawable->id); - (void)FreeColors(b->cmap, client, - b->num_alloced, b->alloced_colors, 0); -#else - XFreeColors(b->display, b->cmap, - b->alloced_colors, b->num_alloced, 0); -#endif - } - } - - if (b->gc) - XMesaFreeGC(b->display, b->gc); - if (b->cleargc) - XMesaFreeGC(b->display, b->cleargc); - if (b->swapgc) - XMesaFreeGC(b->display, b->swapgc); - - if (fb->Visual.doubleBufferMode) { - /* free back ximage/pixmap/shmregion */ - } - - if (b->rowimage) { - _mesa_free( b->rowimage->data ); - b->rowimage->data = NULL; - XMesaDestroyImage( b->rowimage ); - } - - _mesa_free_framebuffer_data(fb); - _mesa_free(fb); -} diff --git a/src/mesa/pipe/xlib/xm_surface.c b/src/mesa/pipe/xlib/xm_surface.c deleted file mode 100644 index 6f5c7ac6ec..0000000000 --- a/src/mesa/pipe/xlib/xm_surface.c +++ /dev/null @@ -1,94 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 7.1 - * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/** - * \file xm_surface.c - * Code to allow the softpipe code to write to X windows/buffers. - * This is a bit of a hack for now. We've basically got two different - * abstractions for color buffers: gl_renderbuffer and pipe_surface. - * They'll need to get merged someday... - * For now, they're separate things that point to each other. - */ - - -#include "GL/xmesa.h" -#include "glxheader.h" -#include "xmesaP.h" -#include "main/context.h" -#include "main/imports.h" -#include "main/macros.h" - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_winsys.h" -#include "pipe/softpipe/sp_context.h" -#include "pipe/softpipe/sp_clear.h" -#include "pipe/softpipe/sp_tile_cache.h" -#include "pipe/softpipe/sp_surface.h" -#include "state_tracker/st_context.h" - - -/* - * Dithering kernels and lookup tables. - */ - -const int xmesa_kernel8[DITH_DY * DITH_DX] = { - 0 * MAXC, 8 * MAXC, 2 * MAXC, 10 * MAXC, - 12 * MAXC, 4 * MAXC, 14 * MAXC, 6 * MAXC, - 3 * MAXC, 11 * MAXC, 1 * MAXC, 9 * MAXC, - 15 * MAXC, 7 * MAXC, 13 * MAXC, 5 * MAXC, -}; - -const int xmesa_kernel1[16] = { - 0*47, 9*47, 4*47, 12*47, /* 47 = (255*3)/16 */ - 6*47, 2*47, 14*47, 8*47, - 10*47, 1*47, 5*47, 11*47, - 7*47, 13*47, 3*47, 15*47 -}; - - - - -/** XXX unfinished sketch... */ -struct pipe_surface * -xmesa_create_front_surface(XMesaVisual vis, Window win) -{ - struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); - if (!xms) { - return NULL; - } - - xms->display = vis->display; - xms->drawable = win; - - xms->surface.format = PIPE_FORMAT_U_A8_R8_G8_B8; - xms->surface.refcount = 1; -#if 0 - xms->surface.region = pipe->winsys->region_alloc(pipe->winsys, - 1, 0, 0, 0x0); -#endif - return &xms->surface; -} - -- cgit v1.2.3 From 9772bb7f71cf114f463e4a9383202846da071363 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 08:24:33 -0700 Subject: remove stray ; --- src/mesa/pipe/xlib/xm_winsys.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index b9dd3e3b6e..e8cb36eced 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -397,7 +397,7 @@ xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) return TRUE; default: return FALSE; - }; + } } } -- cgit v1.2.3 From 0d1669f5303e37980170169687e1dfa5f74c4c00 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 08:24:56 -0700 Subject: byteswapping fixes in choose_pixel_format() --- src/mesa/pipe/xlib/xm_api.c | 23 +++++++++++++---------- 1 file changed, 13 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index ce7e3cb6c2..5cb9ca0d36 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -400,23 +400,25 @@ choose_pixel_format(XMesaVisual v) if ( GET_REDMASK(v) == 0x0000ff && GET_GREENMASK(v) == 0x00ff00 && GET_BLUEMASK(v) == 0xff0000 - && CHECK_BYTE_ORDER(v) && v->BitsPerPixel == 32) { - /* common 32 bpp config used on SGI, Sun */ - assert(0); - return 0 /* PIXEL_FORMAT_U_A8_B8_G8_A8 */; + if (CHECK_BYTE_ORDER(v)) { + /* no byteswapping needed */ + return 0 /* PIXEL_FORMAT_U_A8_B8_G8_R8 */; + } + else { + return PIPE_FORMAT_U_R8_G8_B8_A8; + } } else if ( GET_REDMASK(v) == 0xff0000 && GET_GREENMASK(v) == 0x00ff00 && GET_BLUEMASK(v) == 0x0000ff - && CHECK_BYTE_ORDER(v)) { - if (v->BitsPerPixel == 32) { + && v->BitsPerPixel == 32) { + if (CHECK_BYTE_ORDER(v)) { + /* no byteswapping needed */ return PIPE_FORMAT_U_A8_R8_G8_B8; } - else if (v->BitsPerPixel == 24) { - /* not supported yet */ - abort(); - return 0; + else { + return PIPE_FORMAT_U_B8_G8_R8_A8; } } else if ( GET_REDMASK(v) == 0xf800 @@ -734,6 +736,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) return NULL; pf = choose_pixel_format(v); + assert(pf); pipe = xmesa_create_pipe_context( c, pf ); -- cgit v1.2.3 From da3dfe58b2c2dda38eb7c276a4ae94a4d6a1abc7 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 08:28:59 -0700 Subject: Add PIPE_FORMAT_U_B8_G8_R8_A8 in default_rgba_format(). Also, rewrite/simplify default_rgba_format() and default_depth_format(). --- src/mesa/state_tracker/st_format.c | 50 +++++++++++++++++++++----------------- 1 file changed, 28 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index b6b9375967..f93f7f48d9 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -275,20 +275,23 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) } /** - * Search list of formats for first RGBA format. + * Find an RGBA format supported by the context/winsys. */ static GLuint default_rgba_format( struct pipe_context *pipe ) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_R8_G8_B8_A8 )) { - return PIPE_FORMAT_U_R8_G8_B8_A8; - } - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) { - return PIPE_FORMAT_U_A8_R8_G8_B8; - } - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_R5_G6_B5 )) { - return PIPE_FORMAT_U_R5_G6_B5; + static const uint colorFormats[] = { + PIPE_FORMAT_U_R8_G8_B8_A8, + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_B8_G8_R8_A8, + PIPE_FORMAT_U_R5_G6_B5 + }; + uint i; + for (i = 0; i < Elements(colorFormats); i++) { + if (pipe->is_format_supported( pipe, colorFormats[i] )) { + return colorFormats[i]; + } } return PIPE_FORMAT_NONE; } @@ -309,27 +312,28 @@ default_deep_rgba_format( /** - * Search list of formats for first depth/Z format. + * Find an Z format supported by the context/winsys. */ static GLuint default_depth_format( struct pipe_context *pipe ) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_Z16 )) { - return PIPE_FORMAT_U_Z16; - } - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_Z32 )) { - return PIPE_FORMAT_U_Z32; - } - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) { - return PIPE_FORMAT_S8_Z24; - } - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24_S8 )) { - return PIPE_FORMAT_Z24_S8; + static const uint zFormats[] = { + PIPE_FORMAT_U_Z16, + PIPE_FORMAT_U_Z32, + PIPE_FORMAT_S8_Z24, + PIPE_FORMAT_Z24_S8 + }; + uint i; + for (i = 0; i < Elements(zFormats); i++) { + if (pipe->is_format_supported( pipe, zFormats[i] )) { + return zFormats[i]; + } } return PIPE_FORMAT_NONE; } + /** * Choose the PIPE_FORMAT_ to use for storing a texture image based * on the user's internalFormat, format and type parameters. @@ -541,7 +545,7 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, */ const struct gl_texture_format * st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, - GLenum format, GLenum type) + GLenum format, GLenum type) { #if 0 struct intel_context *intel = intel_context(ctx); @@ -550,6 +554,8 @@ st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, const GLboolean do32bpt = 1; #endif + (void) ctx; + switch (internalFormat) { case 4: case GL_RGBA: -- cgit v1.2.3 From 94a3cb078db2e09716604923c02e3a975ae318ae Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 10:18:34 -0700 Subject: minor code movement --- src/mesa/pipe/softpipe/sp_context.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index f7f0316cf2..696b991334 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -255,6 +255,8 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, /* queries */ softpipe->pipe.is_format_supported = softpipe_is_format_supported; + softpipe->pipe.get_name = softpipe_get_name; + softpipe->pipe.get_vendor = softpipe_get_vendor; softpipe->pipe.get_param = softpipe_get_param; softpipe->pipe.get_paramf = softpipe_get_paramf; @@ -307,9 +309,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.end_query = softpipe_end_query; softpipe->pipe.wait_query = softpipe_wait_query; - softpipe->pipe.get_name = softpipe_get_name; - softpipe->pipe.get_vendor = softpipe_get_vendor; - /* textures */ softpipe->pipe.texture_create = softpipe_texture_create; softpipe->pipe.texture_release = softpipe_texture_release; -- cgit v1.2.3 From 4c90dc7027a42e046662d52c7300f0eb68a6cf3f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 10:21:56 -0700 Subject: Move struct softpipe_texture definition into sp_texture.h Also, added softpipe_texture() cast wrapper. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 1 + src/mesa/pipe/softpipe/sp_state.h | 29 ----------------------- src/mesa/pipe/softpipe/sp_state_sampler.c | 3 ++- src/mesa/pipe/softpipe/sp_surface.c | 3 ++- src/mesa/pipe/softpipe/sp_texture.c | 2 +- src/mesa/pipe/softpipe/sp_texture.h | 39 +++++++++++++++++++++++++++++++ 6 files changed, 45 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 7184fcda52..1aba54d12a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -49,6 +49,7 @@ #include "sp_state.h" #include "sp_headers.h" #include "sp_quad.h" +#include "sp_texture.h" #include "sp_tex_sample.h" diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index daf9955ca5..ea9d2e62be 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -52,35 +52,6 @@ struct sp_fragment_shader_state { #endif }; -struct softpipe_texture -{ - struct pipe_texture base; - - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ - unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer_handle *buffer; -}; - void * softpipe_create_alpha_test_state(struct pipe_context *, const struct pipe_alpha_test_state *); diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index e71b9159e3..173901f04e 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -32,6 +32,7 @@ #include "pipe/p_util.h" #include "sp_context.h" #include "sp_state.h" +#include "sp_texture.h" #include "sp_tile_cache.h" @@ -73,7 +74,7 @@ softpipe_set_texture_state(struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); assert(unit < PIPE_MAX_SAMPLERS); - softpipe->texture[unit] = (struct softpipe_texture *)texture; /* ptr, not struct */ + softpipe->texture[unit] = softpipe_texture(texture); /* ptr, not struct */ sp_tile_cache_set_texture(softpipe->tex_cache[unit], texture); diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index c61e0842fc..16f0209eee 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -32,6 +32,7 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_surface.h" +#include "sp_texture.h" #include "sp_tile_cache.h" /** @@ -568,7 +569,7 @@ softpipe_get_tex_surface(struct pipe_context *pipe, struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice) { - struct softpipe_texture *spt = (struct softpipe_texture *)pt; + struct softpipe_texture *spt = softpipe_texture(pt); struct pipe_surface *ps; unsigned offset; /* in bytes */ diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index 53486f9bba..e8cdd67435 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -412,7 +412,7 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); */ if (--(*pt)->refcount <= 0) { - struct softpipe_texture *spt = (struct softpipe_texture *)*pt; + struct softpipe_texture *spt = softpipe_texture(*pt); uint i; /* diff --git a/src/mesa/pipe/softpipe/sp_texture.h b/src/mesa/pipe/softpipe/sp_texture.h index 2aca57bd1d..732064d986 100644 --- a/src/mesa/pipe/softpipe/sp_texture.h +++ b/src/mesa/pipe/softpipe/sp_texture.h @@ -6,6 +6,45 @@ struct pipe_context; struct pipe_texture; +struct softpipe_texture +{ + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer_handle *buffer; +}; + + +/** cast wrapper */ +static INLINE struct softpipe_texture * +softpipe_texture(struct pipe_texture *pt) +{ + return (struct softpipe_texture *) pt; +} + + + extern void softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); -- cgit v1.2.3 From e6b5cf342e8a66869d25b68d5a63f3d3600b8aad Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 13:08:28 -0700 Subject: add some whitespace --- src/mesa/pipe/softpipe/sp_context.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 696b991334..54b2076b4a 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -264,21 +264,27 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.create_alpha_test_state = softpipe_create_alpha_test_state; softpipe->pipe.bind_alpha_test_state = softpipe_bind_alpha_test_state; softpipe->pipe.delete_alpha_test_state = softpipe_delete_alpha_test_state; + softpipe->pipe.create_blend_state = softpipe_create_blend_state; softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; + softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; softpipe->pipe.bind_sampler_state = softpipe_bind_sampler_state; softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; + softpipe->pipe.create_depth_stencil_state = softpipe_create_depth_stencil_state; softpipe->pipe.bind_depth_stencil_state = softpipe_bind_depth_stencil_state; softpipe->pipe.delete_depth_stencil_state = softpipe_delete_depth_stencil_state; + softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; + softpipe->pipe.create_fs_state = softpipe_create_fs_state; softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; + softpipe->pipe.create_vs_state = softpipe_create_vs_state; softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; -- cgit v1.2.3 From dd87c43a44871e6cf7c3328120c50447bd69c26c Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 13:19:00 -0700 Subject: don't dereference pt after realloc - fixes valgrind error --- src/mesa/pipe/softpipe/sp_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index e8cdd67435..2288c343bf 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -386,7 +386,7 @@ softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) if (spt->buffer) { pipe->winsys->buffer_data(pipe->winsys, spt->buffer, - spt->pitch * (*pt)->cpp * + spt->pitch * spt->base.cpp * spt->total_height, NULL, PIPE_BUFFER_USAGE_PIXEL); } -- cgit v1.2.3 From 57df347bb8a119e1d00fe808e30a0623e0ae8563 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 15:32:13 -0700 Subject: Some tile cache improvements, clean-ups. Code for optimized clears (per-tile flag indicating clear status) in place but not enabled yet. --- src/mesa/pipe/softpipe/sp_clear.c | 4 +- src/mesa/pipe/softpipe/sp_state_surface.c | 8 +- src/mesa/pipe/softpipe/sp_tile_cache.c | 330 +++++++++++++++++------------- src/mesa/pipe/softpipe/sp_tile_cache.h | 12 +- 4 files changed, 206 insertions(+), 148 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 496b38fd5f..a4276362b9 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -50,12 +50,12 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, softpipe_update_derived(softpipe); /* not needed?? */ - if (ps == sp_tile_cache_get_surface(softpipe, softpipe->zbuf_cache)) { + if (ps == sp_tile_cache_get_surface(softpipe->zbuf_cache)) { float clear[4]; clear[0] = 1.0; /* XXX hack */ sp_tile_cache_clear(softpipe->zbuf_cache, clear); } - else if (ps == sp_tile_cache_get_surface(softpipe, softpipe->cbuf_cache[0])) { + else if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[0])) { float clear[4]; clear[0] = 0.2f; /* XXX hack */ clear[1] = 0.2f; /* XXX hack */ diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index 9470ef485f..30bedc74bc 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -66,7 +66,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, sp->framebuffer.cbufs[i] = fb->cbufs[i]; /* update cache */ - sp_tile_cache_set_surface(sp, sp->cbuf_cache[i], ps); + sp_tile_cache_set_surface(sp->cbuf_cache[i], ps); } } @@ -92,7 +92,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, sp->framebuffer.zbuf = fb->zbuf; /* update cache */ - sp_tile_cache_set_surface(sp, sp->zbuf_cache, ps); + sp_tile_cache_set_surface(sp->zbuf_cache, ps); } /* XXX combined depth/stencil here */ @@ -116,12 +116,12 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, if (fb->sbuf != fb->zbuf) { /* separate stencil buf */ sp->sbuf_cache = sp->sbuf_cache_sep; - sp_tile_cache_set_surface(sp, sp->sbuf_cache, ps); + sp_tile_cache_set_surface(sp->sbuf_cache, ps); } else { /* combined depth/stencil */ sp->sbuf_cache = sp->zbuf_cache; - sp_tile_cache_set_surface(sp, sp->sbuf_cache, ps); + sp_tile_cache_set_surface(sp->sbuf_cache, ps); } } diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 08cd39cc55..92e43d63bc 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -38,9 +38,7 @@ #include "sp_surface.h" #include "sp_tile_cache.h" -#define CLEAR_OPTIMIZATION 0 - -#define NUM_ENTRIES 20 +#define NUM_ENTRIES 30 /** XXX move these */ @@ -69,7 +67,7 @@ struct softpipe_tile_cache -static uint +static INLINE uint is_clear_flag_set(const uint *bitvec, int x, int y) { int pos, bit; @@ -82,7 +80,7 @@ is_clear_flag_set(const uint *bitvec, int x, int y) } -static void +static INLINE void clear_clear_flag(uint *bitvec, int x, int y) { int pos; @@ -123,24 +121,22 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc) void -sp_tile_cache_set_surface(struct softpipe_context *sp, - struct softpipe_tile_cache *tc, +sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, struct pipe_surface *ps) { if (tc->surface && tc->surface->map) pipe_surface_unmap(tc->surface); pipe_surface_reference(&tc->surface, ps); + + if (!ps->map) + pipe_surface_map(ps); } struct pipe_surface * -sp_tile_cache_get_surface(struct softpipe_context *sp, - struct softpipe_tile_cache *tc) +sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) { - if (tc->surface && !tc->surface->map) - pipe_surface_map(tc->surface); - return tc->surface; } @@ -161,46 +157,200 @@ sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, } +static INLINE boolean +is_depth_stencil_surface(struct pipe_surface *ps) +{ + return (ps && + (ps->format == PIPE_FORMAT_S8_Z24 || + ps->format == PIPE_FORMAT_U_Z16 || + ps->format == PIPE_FORMAT_U_Z32 || + ps->format == PIPE_FORMAT_U_S8)); +} + + +/** + * Set pixels in a tile to the given clear color/value. + */ +static void +clear_tile(struct softpipe_cached_tile *tile, uint format, + const float clear_value[4]) +{ + uint i, j; + + switch (format) { + case PIPE_FORMAT_U_Z16: + { + ushort clear_val = (ushort) (clear_value[0] * 0xffff); + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.depth16[i][j] = clear_val; + } + } + } + break; + case PIPE_FORMAT_U_Z32: + { + uint clear_val = (uint) (clear_value[0] * 0xffffffff); + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.depth32[i][j] = clear_val; + } + } + } + break; + case PIPE_FORMAT_S8_Z24: + { + uint clear_val = (uint) (clear_value[0] * 0xffffff); + clear_val |= ((uint) clear_value[1]) << 24; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.depth32[i][j] = clear_val; + } + } + } + break; + case PIPE_FORMAT_U_S8: + { + ubyte clear_val = (uint) clear_value[0]; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.stencil8[i][j] = clear_val; + } + } + } + break; + default: + /* color */ + if (clear_value[0] == 0.0 && + clear_value[1] == 0.0 && + clear_value[2] == 0.0 && + clear_value[3] == 0.0) { + memset(tile->data.color, 0, sizeof(tile->data.color)); + } + else { + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.color[i][j][0] = clear_value[0]; + tile->data.color[i][j][1] = clear_value[1]; + tile->data.color[i][j][2] = clear_value[2]; + tile->data.color[i][j][3] = clear_value[3]; + } + } + } + } +} + + +static INLINE void +put_tile(struct pipe_context *pipe, + struct pipe_surface *ps, + struct softpipe_cached_tile *tile) +{ + assert(tile->x >= 0); + assert(tile->x % TILE_SIZE == 0); + assert(tile->y >= 0); + assert(tile->y % TILE_SIZE == 0); + + if (is_depth_stencil_surface(ps)) { + pipe->put_tile(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); + } + else { + pipe->put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } +} + + +static INLINE void +get_tile(struct pipe_context *pipe, + struct pipe_surface *ps, + struct softpipe_cached_tile *tile) +{ + assert(tile->x >= 0); + assert(tile->x % TILE_SIZE == 0); + assert(tile->y >= 0); + assert(tile->y % TILE_SIZE == 0); + + if (is_depth_stencil_surface(ps)) { + pipe->get_tile(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); + } + else { + pipe->get_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } +} + + void sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc) { struct pipe_context *pipe = &softpipe->pipe; struct pipe_surface *ps = tc->surface; - boolean is_depth_stencil; int inuse = 0, pos; if (!ps || !ps->buffer) return; - is_depth_stencil = (ps->format == PIPE_FORMAT_S8_Z24 || - ps->format == PIPE_FORMAT_Z24_S8 || - ps->format == PIPE_FORMAT_U_Z16 || - ps->format == PIPE_FORMAT_U_Z32 || - ps->format == PIPE_FORMAT_U_S8); - for (pos = 0; pos < NUM_ENTRIES; pos++) { struct softpipe_cached_tile *tile = tc->entries + pos; if (tile->x >= 0) { - if (is_depth_stencil) { - pipe->put_tile(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); - } - else { - pipe->put_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - } - + put_tile(pipe, ps, tile); tile->x = tile->y = -1; /* mark as empty */ inuse++; } } - /* +#if TILE_CLEAR_OPTIMIZATION + sp_tile_cache_flush_clear(&softpipe->pipe, tc); +#endif +#if 0 printf("flushed tiles in use: %d\n", inuse); - */ +#endif +} + + +/** + * Actually clear the tiles which were flagged as being cleared. + */ +void +sp_tile_cache_flush_clear(struct pipe_context *pipe, + struct softpipe_tile_cache *tc) +{ + struct pipe_surface *ps = tc->surface; + const uint w = tc->surface->width; + const uint h = tc->surface->height; + uint x, y; + struct softpipe_cached_tile tile; + uint numCleared = 0; + + /* clear one tile to the clear value */ + clear_tile(&tile, ps->format, tc->clear_value); + + /* push the tile to all positions marked as clear */ + for (y = 0; y < h; y += TILE_SIZE) { + for (x = 0; x < w; x += TILE_SIZE) { + if (is_clear_flag_set(tc->clear_flags, x, y)) { + tile.x = x; + tile.y = y; + put_tile(pipe, ps, &tile); + + /* do this? */ + clear_clear_flag(tc->clear_flags, x, y); + + numCleared++; + } + } + } +#if 0 + printf("num cleared: %u\n", numCleared); +#endif } @@ -210,12 +360,6 @@ sp_get_cached_tile(struct softpipe_context *softpipe, { struct pipe_context *pipe = &softpipe->pipe; struct pipe_surface *ps = tc->surface; - boolean is_depth_stencil - = (ps->format == PIPE_FORMAT_S8_Z24 || - ps->format == PIPE_FORMAT_Z24_S8 || - ps->format == PIPE_FORMAT_U_Z16 || - ps->format == PIPE_FORMAT_U_Z32 || - ps->format == PIPE_FORMAT_U_S8); /* tile pos in framebuffer: */ const int tile_x = x & ~(TILE_SIZE - 1); @@ -230,104 +374,21 @@ sp_get_cached_tile(struct softpipe_context *softpipe, if (tile->x != -1) { /* put dirty tile back in framebuffer */ - if (is_depth_stencil) { - pipe->put_tile(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0 /*STRIDE*/); - } - else { - pipe->put_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - } + put_tile(pipe, ps, tile); } + tile->x = tile_x; + tile->y = tile_y; + if (is_clear_flag_set(tc->clear_flags, x, y)) { /* don't get tile from framebuffer, just clear it */ - uint i, j; - /* XXX these loops could be optimized */ - switch (ps->format) { - case PIPE_FORMAT_U_Z16: - { - ushort clear_val = (ushort) (tc->clear_value[0] * 0xffff); - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.depth16[i][j] = clear_val; - } - } - } - break; - case PIPE_FORMAT_U_Z32: - { - uint clear_val = (uint) (tc->clear_value[0] * 0xffffffff); - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.depth32[i][j] = clear_val; - } - } - } - break; - case PIPE_FORMAT_S8_Z24: - { - uint clear_val = (uint) (tc->clear_value[0] * 0xffffff); - clear_val |= ((uint) tc->clear_value[1]) << 24; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.depth32[i][j] = clear_val; - } - } - } - break; - case PIPE_FORMAT_Z24_S8: - { - uint clear_val = ((uint) (tc->clear_value[0] * 0xffffff)) << 8; - clear_val |= ((uint) tc->clear_value[1]) & 0xff; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.depth32[i][j] = clear_val; - } - } - } - break; - case PIPE_FORMAT_U_S8: - { - ubyte clear_val = (uint) tc->clear_value[0]; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.stencil8[i][j] = clear_val; - } - } - } - break; - default: - /* color */ - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.color[i][j][0] = tc->clear_value[0]; - tile->data.color[i][j][1] = tc->clear_value[1]; - tile->data.color[i][j][2] = tc->clear_value[2]; - tile->data.color[i][j][3] = tc->clear_value[3]; - } - } - } + clear_tile(tile, ps->format, tc->clear_value); clear_clear_flag(tc->clear_flags, x, y); } else { - /* get new tile from framebuffer */ - if (is_depth_stencil) { - pipe->get_tile(pipe, ps, - tile_x, tile_y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); - } - else { - pipe->get_tile_rgba(pipe, ps, - tile_x, tile_y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - } + /* get new tile data from surface */ + get_tile(pipe, ps, tile); } - - tile->x = tile_x; - tile->y = tile_y; } return tile; @@ -375,20 +436,13 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, struct pipe_surface *ps = pipe->get_tex_surface(pipe, tc->texture, face, level, z); - if (ps != tc->surface) { - if (tc->surface && tc->surface->map) - pipe_surface_unmap(tc->surface); - - pipe_surface_reference(&tc->surface, ps); - - if (!tc->surface->map) - pipe_surface_map(tc->surface); - } + pipe_surface_map(ps); pipe->get_tile_rgba(pipe, ps, tile_x, tile_y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); + pipe_surface_unmap(ps); pipe_surface_reference(&ps, NULL); tile->x = tile_x; @@ -415,7 +469,7 @@ sp_tile_cache_clear(struct softpipe_tile_cache *tc, const float value[4]) tc->clear_value[2] = value[2]; tc->clear_value[3] = value[3]; -#if CLEAR_OPTIMIZATION +#if TILE_CLEAR_OPTIMIZATION memset(tc->clear_flags, 255, sizeof(tc->clear_flags)); #else memset(tc->clear_flags, 0, sizeof(tc->clear_flags)); diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index de5ff2c498..b2a0e011e9 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -28,6 +28,8 @@ #ifndef SP_TILE_CACHE_H #define SP_TILE_CACHE_H +#define TILE_CLEAR_OPTIMIZATION 0 + #include "pipe/p_compiler.h" @@ -63,13 +65,11 @@ extern void sp_destroy_tile_cache(struct softpipe_tile_cache *tc); extern void -sp_tile_cache_set_surface(struct softpipe_context *sp, - struct softpipe_tile_cache *tc, +sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, struct pipe_surface *sps); extern struct pipe_surface * -sp_tile_cache_get_surface(struct softpipe_context *sp, - struct softpipe_tile_cache *tc); +sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); extern void sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, @@ -79,6 +79,10 @@ extern void sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc); +extern void +sp_tile_cache_flush_clear(struct pipe_context *pipe, + struct softpipe_tile_cache *tc); + extern void sp_tile_cache_clear(struct softpipe_tile_cache *tc, const float value[4]); -- cgit v1.2.3 From c69c7a6054ad038f9da85c1e0d0af39094ae443e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 15:44:58 -0700 Subject: Convert the clearValue back to floats for sp_tile_cache_clear() It might be nice if sp->clear() took the color/value as floats... --- src/mesa/pipe/softpipe/sp_clear.c | 36 ++++++++++++++++++++++++++++++++---- 1 file changed, 32 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index a4276362b9..2b24ea63ab 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -57,10 +57,38 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, } else if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[0])) { float clear[4]; - clear[0] = 0.2f; /* XXX hack */ - clear[1] = 0.2f; /* XXX hack */ - clear[2] = 0.2f; /* XXX hack */ - clear[3] = 0.2f; /* XXX hack */ + /* XXX it sure would be nice if the clear color was passed to + * this function as float[4].... + */ + uint r, g, b, a; + switch (ps->format) { + case PIPE_FORMAT_U_R8_G8_B8_A8: + r = (clearValue >> 24) & 0xff; + g = (clearValue >> 16) & 0xff; + g = (clearValue >> 8) & 0xff; + a = (clearValue ) & 0xff; + break; + case PIPE_FORMAT_U_A8_R8_G8_B8: + r = (clearValue >> 16) & 0xff; + g = (clearValue >> 8) & 0xff; + b = (clearValue ) & 0xff; + a = (clearValue >> 24) & 0xff; + break; + case PIPE_FORMAT_U_B8_G8_R8_A8: + r = (clearValue >> 8) & 0xff; + g = (clearValue >> 16) & 0xff; + b = (clearValue >> 24) & 0xff; + a = (clearValue ) & 0xff; + break; + default: + assert(0); + } + + clear[0] = r / 255.0; + clear[1] = g / 255.0; + clear[2] = b / 255.0; + clear[3] = a / 255.0; + sp_tile_cache_clear(softpipe->cbuf_cache[0], clear); } -- cgit v1.2.3 From 872caf6089e325cf3a0cee877f5119cac560edac Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 15:46:10 -0700 Subject: Make sp_tile_cache_flush_clear() static. Also, misc clean-ups, comments. --- src/mesa/pipe/softpipe/sp_tile_cache.c | 80 ++++++++++++++++++++-------------- src/mesa/pipe/softpipe/sp_tile_cache.h | 4 -- 2 files changed, 48 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 92e43d63bc..93930b18f2 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -124,8 +124,10 @@ void sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, struct pipe_surface *ps) { - if (tc->surface && tc->surface->map) + if (tc->surface && tc->surface->map) { + assert(tc->surface != ps); pipe_surface_unmap(tc->surface); + } pipe_surface_reference(&tc->surface, ps); @@ -287,39 +289,10 @@ get_tile(struct pipe_context *pipe, } -void -sp_flush_tile_cache(struct softpipe_context *softpipe, - struct softpipe_tile_cache *tc) -{ - struct pipe_context *pipe = &softpipe->pipe; - struct pipe_surface *ps = tc->surface; - int inuse = 0, pos; - - if (!ps || !ps->buffer) - return; - - for (pos = 0; pos < NUM_ENTRIES; pos++) { - struct softpipe_cached_tile *tile = tc->entries + pos; - if (tile->x >= 0) { - put_tile(pipe, ps, tile); - tile->x = tile->y = -1; /* mark as empty */ - inuse++; - } - } - -#if TILE_CLEAR_OPTIMIZATION - sp_tile_cache_flush_clear(&softpipe->pipe, tc); -#endif -#if 0 - printf("flushed tiles in use: %d\n", inuse); -#endif -} - - /** - * Actually clear the tiles which were flagged as being cleared. + * Actually clear the tiles which were flagged as being in a clear state. */ -void +static void sp_tile_cache_flush_clear(struct pipe_context *pipe, struct softpipe_tile_cache *tc) { @@ -354,6 +327,47 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, } +/** + * Flush the tile cache: write all dirty tiles back to the surface. + * any tiles "flagged" as cleared will be "really" cleared. + */ +void +sp_flush_tile_cache(struct softpipe_context *softpipe, + struct softpipe_tile_cache *tc) +{ + struct pipe_context *pipe = &softpipe->pipe; + struct pipe_surface *ps = tc->surface; + int inuse = 0, pos; + + if (!ps || !ps->buffer) + return; + + if (!ps->map) + pipe_surface_map(ps); + + for (pos = 0; pos < NUM_ENTRIES; pos++) { + struct softpipe_cached_tile *tile = tc->entries + pos; + if (tile->x >= 0) { + put_tile(pipe, ps, tile); + tile->x = tile->y = -1; /* mark as empty */ + inuse++; + } + } + +#if TILE_CLEAR_OPTIMIZATION + sp_tile_cache_flush_clear(&softpipe->pipe, tc); +#endif + +#if 0 + printf("flushed tiles in use: %d\n", inuse); +#endif +} + + +/** + * Get a tile from the cache. + * \param x, y position of tile, in pixels + */ struct softpipe_cached_tile * sp_get_cached_tile(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc, int x, int y) @@ -470,8 +484,10 @@ sp_tile_cache_clear(struct softpipe_tile_cache *tc, const float value[4]) tc->clear_value[3] = value[3]; #if TILE_CLEAR_OPTIMIZATION + /* set flags to indicate all the tiles are cleared */ memset(tc->clear_flags, 255, sizeof(tc->clear_flags)); #else + /* disable the optimization */ memset(tc->clear_flags, 0, sizeof(tc->clear_flags)); #endif } diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index b2a0e011e9..12d5815db2 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -79,10 +79,6 @@ extern void sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc); -extern void -sp_tile_cache_flush_clear(struct pipe_context *pipe, - struct softpipe_tile_cache *tc); - extern void sp_tile_cache_clear(struct softpipe_tile_cache *tc, const float value[4]); -- cgit v1.2.3 From 099bf3ffe932f4a8c1ddae064bce901555af35ce Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 15:47:39 -0700 Subject: Enable the tile clear "optimization" Not really faster yet because we're using the float-valued put/get_tile() functions so we're doing quite a bit of float/int conversion. That can be fixed though... --- src/mesa/pipe/softpipe/sp_tile_cache.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index 12d5815db2..c4f2e0f184 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -28,7 +28,7 @@ #ifndef SP_TILE_CACHE_H #define SP_TILE_CACHE_H -#define TILE_CLEAR_OPTIMIZATION 0 +#define TILE_CLEAR_OPTIMIZATION 1 #include "pipe/p_compiler.h" -- cgit v1.2.3 From 2067eed9d30bb5b260920a5650655579c1988202 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 16:01:31 -0700 Subject: Get rid of "duplicate" formats. For example, replace PIPE_FORMAT_U_A8_R8_G8_B8 with PIPE_FORMAT_A8R8G8B8_UNORM --- .../dri/intel_winsys/intel_winsys_softpipe.c | 6 +- src/mesa/pipe/draw/draw_vertex_fetch.c | 2 +- src/mesa/pipe/i915simple/i915_context.c | 24 ++++---- src/mesa/pipe/i915simple/i915_state_emit.c | 10 ++-- src/mesa/pipe/i915simple/i915_state_sampler.c | 12 ++-- src/mesa/pipe/i915simple/i915_surface.c | 4 +- src/mesa/pipe/p_format.h | 12 ---- src/mesa/pipe/softpipe/sp_clear.c | 6 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 16 ++--- src/mesa/pipe/softpipe/sp_quad_stencil.c | 8 +-- src/mesa/pipe/softpipe/sp_surface.c | 70 +++++++++++----------- src/mesa/pipe/softpipe/sp_tile_cache.c | 12 ++-- src/mesa/pipe/xlib/xm_api.c | 8 +-- src/mesa/pipe/xlib/xm_winsys.c | 8 +-- 14 files changed, 93 insertions(+), 105 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c index 7a93546bc2..f9ed6253f1 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c @@ -48,9 +48,9 @@ static boolean intel_is_format_supported(struct softpipe_winsys *sws, uint format) { switch(format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - case PIPE_FORMAT_U_R5_G6_B5: - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: return TRUE; default: return FALSE; diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 5510b3674e..77df5c6427 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -79,7 +79,7 @@ fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) attrib[0] = (float) ((int *) ptr)[0]; break; - case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_A8R8G8B8_UNORM: attrib[0] = (float) ((unsigned char *) ptr)[2] / 255.0f; attrib[1] = (float) ((unsigned char *) ptr)[1] / 255.0f; attrib[2] = (float) ((unsigned char *) ptr)[0] / 255.0f; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 07b331c528..cfce116920 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -50,32 +50,32 @@ i915_is_format_supported( struct pipe_context *pipe, #if 0 /* XXX: This is broken -- rewrite if still needed. */ static const unsigned tex_supported[] = { - PIPE_FORMAT_U_R8_G8_B8_A8, - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_R8G8B8A8_UNORM, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, PIPE_FORMAT_U_L8, PIPE_FORMAT_U_A8, PIPE_FORMAT_U_I8, PIPE_FORMAT_U_L8_A8, PIPE_FORMAT_YCBCR, PIPE_FORMAT_YCBCR_REV, - PIPE_FORMAT_S8_Z24, + PIPE_FORMAT_S8Z24_UNORM, }; /* Actually a lot more than this - add later: */ static const unsigned render_supported[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, }; /* */ static const unsigned z_stencil_supported[] = { - PIPE_FORMAT_U_Z16, - PIPE_FORMAT_U_Z32, - PIPE_FORMAT_S8_Z24, + PIPE_FORMAT_Z16_UNORM, + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_S8Z24_UNORM, }; switch (type) { @@ -97,9 +97,9 @@ i915_is_format_supported( struct pipe_context *pipe, } #else switch( format ) { - case PIPE_FORMAT_U_A8_R8_G8_B8: - case PIPE_FORMAT_U_R5_G6_B5: - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: return TRUE; }; return FALSE; diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index eda40d7c76..1e0db91024 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -38,9 +38,9 @@ static unsigned translate_format( unsigned format ) { switch (format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_A8R8G8B8_UNORM: return COLOR_BUF_ARGB8888; - case PIPE_FORMAT_U_R5_G6_B5: + case PIPE_FORMAT_R5G6B5_UNORM: return COLOR_BUF_RGB565; default: assert(0); @@ -51,9 +51,9 @@ static unsigned translate_format( unsigned format ) static unsigned translate_depth_format( unsigned zformat ) { switch (zformat) { - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: return DEPTH_FRMT_24_FIXED_8_OTHER; - case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_Z16_UNORM: return DEPTH_FRMT_16_FIXED; default: assert(0); @@ -247,7 +247,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) if (cbuf_surface) cformat = cbuf_surface->format; else - cformat = PIPE_FORMAT_U_A8_R8_G8_B8; /* arbitrary */ + cformat = PIPE_FORMAT_A8R8G8B8_UNORM; /* arbitrary */ cformat = translate_format(cformat); if (depth_surface) diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 1816d9abdb..dd2d51d9f6 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -138,13 +138,13 @@ translate_texture_format(uint pipeFormat) return MAPSURF_8BIT | MT_8BIT_A8; case PIPE_FORMAT_U_A8_L8: return MAPSURF_16BIT | MT_16BIT_AY88; - case PIPE_FORMAT_U_R5_G6_B5: + case PIPE_FORMAT_R5G6B5_UNORM: return MAPSURF_16BIT | MT_16BIT_RGB565; - case PIPE_FORMAT_U_A1_R5_G5_B5: + case PIPE_FORMAT_A1R5G5B5_UNORM: return MAPSURF_16BIT | MT_16BIT_ARGB1555; - case PIPE_FORMAT_U_A4_R4_G4_B4: + case PIPE_FORMAT_A4R4G4B4_UNORM: return MAPSURF_16BIT | MT_16BIT_ARGB4444; - case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_A8R8G8B8_UNORM: return MAPSURF_32BIT | MT_32BIT_ARGB8888; case PIPE_FORMAT_YCBCR_REV: return (MAPSURF_422 | MT_422_YCRCB_NORMAL); @@ -155,7 +155,7 @@ translate_texture_format(uint pipeFormat) case PIPE_FORMAT_RGBA_FXT1: return (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1); #endif - case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_Z16_UNORM: return (MAPSURF_16BIT | MT_16BIT_L16); #if 0 case PIPE_FORMAT_RGBA_DXT1: @@ -166,7 +166,7 @@ translate_texture_format(uint pipeFormat) case PIPE_FORMAT_RGBA_DXT5: return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); #endif - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: return (MAPSURF_32BIT | MT_32BIT_xL824); default: fprintf(stderr, "i915: translate_texture_format() bad image format %x\n", diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index ecbabe50eb..ea48c1bd8a 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -66,7 +66,7 @@ i915_get_tile_rgba(struct pipe_context *pipe, CLIP_TILE; switch (ps->format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_A8R8G8B8_UNORM: for (i = 0; i < h; i++) { float *pRow = p; for (j = 0; j < w; j++) { @@ -81,7 +81,7 @@ i915_get_tile_rgba(struct pipe_context *pipe, p += w0 * 4; } break; - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: { const float scale = 1.0f / (float) 0xffffff; for (i = 0; i < h; i++) { diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index de0eb9539f..86728f77a2 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -276,18 +276,6 @@ static INLINE uint pf_rev(pipe_format_ycbcr_t f) /** * Duplicated formats: */ -#define PIPE_FORMAT_U_A8_R8_G8_B8 PIPE_FORMAT_A8R8G8B8_UNORM -#define PIPE_FORMAT_U_B8_G8_R8_A8 PIPE_FORMAT_B8G8R8A8_UNORM -#define PIPE_FORMAT_U_A1_R5_G5_B5 PIPE_FORMAT_A1R5G5B5_UNORM -#define PIPE_FORMAT_U_A4_R4_G4_B4 PIPE_FORMAT_A4R4G4B4_UNORM -#define PIPE_FORMAT_U_R8_G8_B8_A8 PIPE_FORMAT_R8G8B8A8_UNORM -#define PIPE_FORMAT_U_R5_G6_B5 PIPE_FORMAT_R5G6B5_UNORM -#define PIPE_FORMAT_S_R16_G16_B16_A16 PIPE_FORMAT_R16G16B16A16_SNORM -#define PIPE_FORMAT_U_Z16 PIPE_FORMAT_Z16_UNORM -#define PIPE_FORMAT_U_Z32 PIPE_FORMAT_Z32_UNORM -#define PIPE_FORMAT_F_Z32 PIPE_FORMAT_Z32_FLOAT -#define PIPE_FORMAT_S8_Z24 PIPE_FORMAT_S8Z24_UNORM -#define PIPE_FORMAT_Z24_S8 PIPE_FORMAT_Z24S8_UNORM #define PIPE_FORMAT_U_S8 PIPE_FORMAT_S8_UNORM /** diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 2b24ea63ab..a6352e0616 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -62,19 +62,19 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, */ uint r, g, b, a; switch (ps->format) { - case PIPE_FORMAT_U_R8_G8_B8_A8: + case PIPE_FORMAT_R8G8B8A8_UNORM: r = (clearValue >> 24) & 0xff; g = (clearValue >> 16) & 0xff; g = (clearValue >> 8) & 0xff; a = (clearValue ) & 0xff; break; - case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_A8R8G8B8_UNORM: r = (clearValue >> 16) & 0xff; g = (clearValue >> 8) & 0xff; b = (clearValue ) & 0xff; a = (clearValue >> 24) & 0xff; break; - case PIPE_FORMAT_U_B8_G8_R8_A8: + case PIPE_FORMAT_B8G8R8A8_UNORM: r = (clearValue >> 8) & 0xff; g = (clearValue >> 16) & 0xff; b = (clearValue >> 24) & 0xff; diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 3318189621..00128fa528 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -74,7 +74,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) * Also, get the zbuffer values (bzzzz) from the cached tile. */ switch (format) { - case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_Z16_UNORM: { float scale = 65535.0; @@ -89,7 +89,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } } break; - case PIPE_FORMAT_U_Z32: + case PIPE_FORMAT_Z32_UNORM: { double scale = (double) (uint) ~0UL; @@ -104,7 +104,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } } break; - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: { float scale = (float) ((1 << 24) - 1); @@ -119,7 +119,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } } break; - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_Z24S8_UNORM: { float scale = (float) ((1 << 24) - 1); @@ -202,21 +202,21 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) /* put updated Z values back into cached tile */ switch (format) { - case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_Z16_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); tile->data.depth16[y][x] = (ushort) bzzzz[j]; } break; - case PIPE_FORMAT_U_Z32: + case PIPE_FORMAT_Z32_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); tile->data.depth32[y][x] = bzzzz[j]; } break; - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); @@ -225,7 +225,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) tile->data.depth32[y][x] = s8z24; } break; - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_Z24S8_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 0149b20f48..b8c199204d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -234,14 +234,14 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) /* get stencil values from cached tile */ switch (ps->format) { - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); stencilVals[j] = tile->data.depth32[y][x] >> 24; } break; - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_Z24S8_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); @@ -298,7 +298,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) /* put new stencil values into cached tile */ switch (ps->format) { - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); @@ -307,7 +307,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) tile->data.depth32[y][x] = s8z24; } break; - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_Z24S8_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 16f0209eee..cdbffe5400 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -70,7 +70,7 @@ #endif -/*** PIPE_FORMAT_U_A8_R8_G8_B8 ***/ +/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ static void a8r8g8b8_get_tile(struct pipe_surface *ps, @@ -82,7 +82,7 @@ a8r8g8b8_get_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); + assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); CLIP_TILE; @@ -113,7 +113,7 @@ a8r8g8b8_put_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_U_A8_R8_G8_B8); + assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); CLIP_TILE; @@ -134,7 +134,7 @@ a8r8g8b8_put_tile(struct pipe_surface *ps, } -/*** PIPE_FORMAT_U_B8_G8_R8_A8 ***/ +/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ static void b8g8r8a8_get_tile(struct pipe_surface *ps, @@ -146,7 +146,7 @@ b8g8r8a8_get_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_U_B8_G8_R8_A8); + assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); CLIP_TILE; @@ -177,7 +177,7 @@ b8g8r8a8_put_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_U_B8_G8_R8_A8); + assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); CLIP_TILE; @@ -198,7 +198,7 @@ b8g8r8a8_put_tile(struct pipe_surface *ps, } -/*** PIPE_FORMAT_U_A1_R5_G5_B5 ***/ +/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ static void a1r5g5b5_get_tile(struct pipe_surface *ps, @@ -209,7 +209,7 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, + y * ps->pitch + x; unsigned i, j; - assert(ps->format == PIPE_FORMAT_U_A1_R5_G5_B5); + assert(ps->format == PIPE_FORMAT_A1R5G5B5_UNORM); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { @@ -226,7 +226,7 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, -/*** PIPE_FORMAT_U_Z16 ***/ +/*** PIPE_FORMAT_Z16_UNORM ***/ /** * Return each Z value as four floats in [0,1]. @@ -242,7 +242,7 @@ z16_get_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_U_Z16); + assert(ps->format == PIPE_FORMAT_Z16_UNORM); CLIP_TILE; @@ -324,7 +324,7 @@ a8_get_tile(struct pipe_surface *ps, } -/*** PIPE_FORMAT_S_R16_G16_B16_A16 ***/ +/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ static void r16g16b16a16_get_tile(struct pipe_surface *ps, @@ -336,7 +336,7 @@ r16g16b16a16_get_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_S_R16_G16_B16_A16); + assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); CLIP_TILE; @@ -368,7 +368,7 @@ r16g16b16a16_put_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_S_R16_G16_B16_A16); + assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); CLIP_TILE; @@ -458,7 +458,7 @@ a8_l8_get_tile(struct pipe_surface *ps, -/*** PIPE_FORMAT_U_Z32 ***/ +/*** PIPE_FORMAT_Z32_UNORM ***/ /** * Return each Z value as four floats in [0,1]. @@ -474,7 +474,7 @@ z32_get_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_U_Z16); + assert(ps->format == PIPE_FORMAT_Z16_UNORM); CLIP_TILE; @@ -492,7 +492,7 @@ z32_get_tile(struct pipe_surface *ps, } -/*** PIPE_FORMAT_S8_Z24 ***/ +/*** PIPE_FORMAT_S8Z24_UNORM ***/ /** * Return Z component as four float in [0,1]. Stencil part ignored. @@ -508,7 +508,7 @@ s8z24_get_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_S8_Z24); + assert(ps->format == PIPE_FORMAT_S8Z24_UNORM); CLIP_TILE; @@ -526,7 +526,7 @@ s8z24_get_tile(struct pipe_surface *ps, } -/*** PIPE_FORMAT_Z24_S8 ***/ +/*** PIPE_FORMAT_Z24S8_UNORM ***/ /** * Return Z component as four float in [0,1]. Stencil part ignored. @@ -542,7 +542,7 @@ z24s8_get_tile(struct pipe_surface *ps, unsigned i, j; unsigned w0 = w; - assert(ps->format == PIPE_FORMAT_Z24_S8); + assert(ps->format == PIPE_FORMAT_Z24S8_UNORM); CLIP_TILE; @@ -673,13 +673,13 @@ softpipe_get_tile_rgba(struct pipe_context *pipe, float *p) { switch (ps->format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_A8R8G8B8_UNORM: a8r8g8b8_get_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_U_B8_G8_R8_A8: + case PIPE_FORMAT_B8G8R8A8_UNORM: b8g8r8a8_get_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_U_A1_R5_G5_B5: + case PIPE_FORMAT_A1R5G5B5_UNORM: a1r5g5b5_get_tile(ps, x, y, w, h, p); break; case PIPE_FORMAT_U_L8: @@ -694,19 +694,19 @@ softpipe_get_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_U_A8_L8: a8_l8_get_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_S_R16_G16_B16_A16: + case PIPE_FORMAT_R16G16B16A16_SNORM: r16g16b16a16_get_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_Z16_UNORM: z16_get_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_U_Z32: + case PIPE_FORMAT_Z32_UNORM: z32_get_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: s8z24_get_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_Z24S8_UNORM: z24s8_get_tile(ps, x, y, w, h, p); break; default: @@ -723,13 +723,13 @@ softpipe_put_tile_rgba(struct pipe_context *pipe, const float *p) { switch (ps->format) { - case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_A8R8G8B8_UNORM: a8r8g8b8_put_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_U_B8_G8_R8_A8: + case PIPE_FORMAT_B8G8R8A8_UNORM: b8g8r8a8_put_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_U_A1_R5_G5_B5: + case PIPE_FORMAT_A1R5G5B5_UNORM: /*a1r5g5b5_put_tile(ps, x, y, w, h, p);*/ break; case PIPE_FORMAT_U_L8: @@ -744,19 +744,19 @@ softpipe_put_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_U_A8_L8: /*a8_l8_put_tile(ps, x, y, w, h, p);*/ break; - case PIPE_FORMAT_S_R16_G16_B16_A16: + case PIPE_FORMAT_R16G16B16A16_SNORM: r16g16b16a16_put_tile(ps, x, y, w, h, p); break; - case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_Z16_UNORM: /*z16_put_tile(ps, x, y, w, h, p);*/ break; - case PIPE_FORMAT_U_Z32: + case PIPE_FORMAT_Z32_UNORM: /*z32_put_tile(ps, x, y, w, h, p);*/ break; - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: /*s8z24_put_tile(ps, x, y, w, h, p);*/ break; - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_Z24S8_UNORM: /*z24s8_put_tile(ps, x, y, w, h, p);*/ break; default: diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 93930b18f2..be5dd5c289 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -163,9 +163,9 @@ static INLINE boolean is_depth_stencil_surface(struct pipe_surface *ps) { return (ps && - (ps->format == PIPE_FORMAT_S8_Z24 || - ps->format == PIPE_FORMAT_U_Z16 || - ps->format == PIPE_FORMAT_U_Z32 || + (ps->format == PIPE_FORMAT_S8Z24_UNORM || + ps->format == PIPE_FORMAT_Z16_UNORM || + ps->format == PIPE_FORMAT_Z32_UNORM || ps->format == PIPE_FORMAT_U_S8)); } @@ -180,7 +180,7 @@ clear_tile(struct softpipe_cached_tile *tile, uint format, uint i, j; switch (format) { - case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_Z16_UNORM: { ushort clear_val = (ushort) (clear_value[0] * 0xffff); for (i = 0; i < TILE_SIZE; i++) { @@ -190,7 +190,7 @@ clear_tile(struct softpipe_cached_tile *tile, uint format, } } break; - case PIPE_FORMAT_U_Z32: + case PIPE_FORMAT_Z32_UNORM: { uint clear_val = (uint) (clear_value[0] * 0xffffffff); for (i = 0; i < TILE_SIZE; i++) { @@ -200,7 +200,7 @@ clear_tile(struct softpipe_cached_tile *tile, uint format, } } break; - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: { uint clear_val = (uint) (clear_value[0] * 0xffffff); clear_val |= ((uint) clear_value[1]) << 24; diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 5cb9ca0d36..9c7713790b 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -406,7 +406,7 @@ choose_pixel_format(XMesaVisual v) return 0 /* PIXEL_FORMAT_U_A8_B8_G8_R8 */; } else { - return PIPE_FORMAT_U_R8_G8_B8_A8; + return PIPE_FORMAT_R8G8B8A8_UNORM; } } else if ( GET_REDMASK(v) == 0xff0000 @@ -415,10 +415,10 @@ choose_pixel_format(XMesaVisual v) && v->BitsPerPixel == 32) { if (CHECK_BYTE_ORDER(v)) { /* no byteswapping needed */ - return PIPE_FORMAT_U_A8_R8_G8_B8; + return PIPE_FORMAT_A8R8G8B8_UNORM; } else { - return PIPE_FORMAT_U_B8_G8_R8_A8; + return PIPE_FORMAT_B8G8R8A8_UNORM; } } else if ( GET_REDMASK(v) == 0xf800 @@ -427,7 +427,7 @@ choose_pixel_format(XMesaVisual v) && CHECK_BYTE_ORDER(v) && v->BitsPerPixel == 16) { /* 5-6-5 RGB */ - return PIPE_FORMAT_U_R5_G6_B5; + return PIPE_FORMAT_R5G6B5_UNORM; } assert(0); diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index e8cb36eced..408797c688 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -389,11 +389,11 @@ xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) else { /* non-color / window surface format */ switch (format) { - case PIPE_FORMAT_S_R16_G16_B16_A16: - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_R16G16B16A16_SNORM: + case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_U_S8: - case PIPE_FORMAT_U_Z16: - case PIPE_FORMAT_U_Z32: + case PIPE_FORMAT_Z16_UNORM: + case PIPE_FORMAT_Z32_UNORM: return TRUE; default: return FALSE; -- cgit v1.2.3 From d3e05111c8c8f87db7f577eb7096d65479a7e481 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 16:15:14 -0700 Subject: Replace "duplicate" formats --- src/mesa/state_tracker/st_cb_clear.c | 24 ++++----- src/mesa/state_tracker/st_cb_drawpixels.c | 22 ++++---- src/mesa/state_tracker/st_cb_readpixels.c | 4 +- src/mesa/state_tracker/st_format.c | 88 +++++++++++++++---------------- 4 files changed, 69 insertions(+), 69 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 8ac6699776..1f0d1e6426 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -65,13 +65,13 @@ color_value(GLuint pipeFormat, const GLfloat color[4]) UNCLAMPED_FLOAT_TO_UBYTE(a, color[3]); switch (pipeFormat) { - case PIPE_FORMAT_U_R8_G8_B8_A8: + case PIPE_FORMAT_R8G8B8A8_UNORM: return (r << 24) | (g << 16) | (b << 8) | a; - case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_A8R8G8B8_UNORM: return (a << 24) | (r << 16) | (g << 8) | b; - case PIPE_FORMAT_U_B8_G8_R8_A8: + case PIPE_FORMAT_B8G8R8A8_UNORM: return (b << 24) | (g << 16) | (r << 8) | a; - case PIPE_FORMAT_U_R5_G6_B5: + case PIPE_FORMAT_R5G6B5_UNORM: return ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); default: assert(0); @@ -84,17 +84,17 @@ static uint depth_value(GLuint pipeFormat, GLfloat value) { switch (pipeFormat) { - case PIPE_FORMAT_U_Z16: + case PIPE_FORMAT_Z16_UNORM: return (uint) (value * 0xffff); - case PIPE_FORMAT_U_Z32: + case PIPE_FORMAT_Z32_UNORM: /* special-case to avoid overflow */ if (value == 1.0) return 0xffffffff; else return (uint) (value * 0xffffffff); - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: return (uint) (value * 0xffffff); - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_Z24S8_UNORM: return ((uint) (value * 0xffffff)) << 8; default: assert(0); @@ -107,8 +107,8 @@ static GLboolean is_depth_stencil_format(GLuint pipeFormat) { switch (pipeFormat) { - case PIPE_FORMAT_S8_Z24: - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: return GL_TRUE; default: return GL_FALSE; @@ -523,10 +523,10 @@ clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); switch (strb->surface->format) { - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: clearValue |= ctx->Stencil.Clear << 24; break; - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_Z24S8_UNORM: clearValue |= clearValue | ctx->Stencil.Clear; break; default: diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index a61daffb20..464cf5ced3 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -734,42 +734,42 @@ compatible_formats(GLenum format, GLenum type, GLuint pipeFormat) static const GLuint one = 1; GLubyte littleEndian = *((GLubyte *) &one); - if (pipeFormat == PIPE_FORMAT_U_R8_G8_B8_A8 && + if (pipeFormat == PIPE_FORMAT_R8G8B8A8_UNORM && format == GL_RGBA && type == GL_UNSIGNED_BYTE && !littleEndian) { return GL_TRUE; } - else if (pipeFormat == PIPE_FORMAT_U_R8_G8_B8_A8 && + else if (pipeFormat == PIPE_FORMAT_R8G8B8A8_UNORM && format == GL_ABGR_EXT && type == GL_UNSIGNED_BYTE && littleEndian) { return GL_TRUE; } - else if (pipeFormat == PIPE_FORMAT_U_A8_R8_G8_B8 && + else if (pipeFormat == PIPE_FORMAT_A8R8G8B8_UNORM && format == GL_BGRA && type == GL_UNSIGNED_BYTE && littleEndian) { return GL_TRUE; } - else if (pipeFormat == PIPE_FORMAT_U_R5_G6_B5 && + else if (pipeFormat == PIPE_FORMAT_R5G6B5_UNORM && format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { /* endian don't care */ return GL_TRUE; } - else if (pipeFormat == PIPE_FORMAT_U_R5_G6_B5 && + else if (pipeFormat == PIPE_FORMAT_R5G6B5_UNORM && format == GL_BGR && type == GL_UNSIGNED_SHORT_5_6_5_REV) { /* endian don't care */ return GL_TRUE; } - else if (pipeFormat == PIPE_FORMAT_U_S8 && + else if (pipeFormat == PIPE_FORMAT_S8_UNORM && format == GL_STENCIL_INDEX && type == GL_UNSIGNED_BYTE) { return GL_TRUE; } - else if (pipeFormat == PIPE_FORMAT_U_Z32 && + else if (pipeFormat == PIPE_FORMAT_Z32_UNORM && format == GL_DEPTH_COMPONENT && type == GL_UNSIGNED_INT) { return GL_TRUE; @@ -888,7 +888,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, memcpy(dest, values, spanWidth); } break; - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: { uint *dest = (uint *) stmap + spanY * ps->pitch + spanX; GLint k; @@ -998,8 +998,8 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, cpp = 1; comp = 0; } - else if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) { - format = PIPE_FORMAT_U_A8_R8_G8_B8; + else if (pipe->is_format_supported( pipe, PIPE_FORMAT_A8R8G8B8_UNORM )) { + format = PIPE_FORMAT_A8R8G8B8_UNORM; internal_format = GL_RGBA8; cpp = 4; comp = 3; /* alpha channel */ /*XXX little-endian dependency */ @@ -1167,7 +1167,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, src = buffer + i * width; switch (psDraw->format) { - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: { uint *dst4 = (uint *) dst; int j; diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index e2243e7de4..96829fcfa0 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -90,7 +90,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, memcpy(values, src, width); } break; - case PIPE_FORMAT_S8_Z24: + case PIPE_FORMAT_S8Z24_UNORM: { const uint *src = (uint *) stmap + srcY * ps->pitch + x; GLint k; @@ -99,7 +99,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } } break; - case PIPE_FORMAT_Z24_S8: + case PIPE_FORMAT_Z24S8_UNORM: { const uint *src = (uint *) stmap + srcY * ps->pitch + x; GLint k; diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index f93f7f48d9..9e249a47a6 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -120,7 +120,7 @@ st_get_format_info( #endif /* Data type */ - if (format == PIPE_FORMAT_U_A1_R5_G5_B5 || format == PIPE_FORMAT_U_R5_G6_B5) { + if (format == PIPE_FORMAT_A1R5G5B5_UNORM || format == PIPE_FORMAT_R5G6B5_UNORM) { pinfo->datatype = GL_UNSIGNED_SHORT; } else { @@ -255,9 +255,9 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) /* fix this */ case MESA_FORMAT_ARGB8888_REV: case MESA_FORMAT_ARGB8888: - return PIPE_FORMAT_U_A8_R8_G8_B8; + return PIPE_FORMAT_A8R8G8B8_UNORM; case MESA_FORMAT_ARGB4444: - return PIPE_FORMAT_U_A4_R4_G4_B4; + return PIPE_FORMAT_A4R4G4B4_UNORM; case MESA_FORMAT_AL88: return PIPE_FORMAT_U_A8_L8; case MESA_FORMAT_A8: @@ -267,7 +267,7 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) case MESA_FORMAT_I8: return PIPE_FORMAT_U_I8; case MESA_FORMAT_Z16: - return PIPE_FORMAT_U_Z16; + return PIPE_FORMAT_Z16_UNORM; default: assert(0); return 0; @@ -282,10 +282,10 @@ default_rgba_format( struct pipe_context *pipe ) { static const uint colorFormats[] = { - PIPE_FORMAT_U_R8_G8_B8_A8, - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_B8_G8_R8_A8, - PIPE_FORMAT_U_R5_G6_B5 + PIPE_FORMAT_R8G8B8A8_UNORM, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_B8G8R8A8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM }; uint i; for (i = 0; i < Elements(colorFormats); i++) { @@ -304,8 +304,8 @@ static GLuint default_deep_rgba_format( struct pipe_context *pipe ) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S_R16_G16_B16_A16 )) { - return PIPE_FORMAT_S_R16_G16_B16_A16; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_R16G16B16A16_SNORM )) { + return PIPE_FORMAT_R16G16B16A16_SNORM; } return PIPE_FORMAT_NONE; } @@ -319,10 +319,10 @@ default_depth_format( struct pipe_context *pipe ) { static const uint zFormats[] = { - PIPE_FORMAT_U_Z16, - PIPE_FORMAT_U_Z32, - PIPE_FORMAT_S8_Z24, - PIPE_FORMAT_Z24_S8 + PIPE_FORMAT_Z16_UNORM, + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_Z24S8_UNORM }; uint i; for (i = 0; i < Elements(zFormats); i++) { @@ -358,16 +358,16 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_COMPRESSED_RGBA: if (format == GL_BGRA) { if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_R8_G8_B8 )) - return PIPE_FORMAT_U_A8_R8_G8_B8; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A8R8G8B8_UNORM )) + return PIPE_FORMAT_A8R8G8B8_UNORM; } else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A4_R4_G4_B4 )) - return PIPE_FORMAT_U_A4_R4_G4_B4; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A4R4G4B4_UNORM )) + return PIPE_FORMAT_A4R4G4B4_UNORM; } else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) - return PIPE_FORMAT_U_A1_R5_G5_B5; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM )) + return PIPE_FORMAT_A1R5G5B5_UNORM; } } return default_rgba_format( pipe ); @@ -376,8 +376,8 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_RGB: case GL_COMPRESSED_RGB: if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_R5_G6_B5 )) - return PIPE_FORMAT_U_R5_G6_B5; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_R5G6B5_UNORM )) + return PIPE_FORMAT_R5G6B5_UNORM; } return default_rgba_format( pipe ); @@ -390,13 +390,13 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_RGBA4: case GL_RGBA2: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A4_R4_G4_B4 )) - return PIPE_FORMAT_U_A4_R4_G4_B4; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A4R4G4B4_UNORM )) + return PIPE_FORMAT_A4R4G4B4_UNORM; return default_rgba_format( pipe ); case GL_RGB5_A1: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) - return PIPE_FORMAT_U_A1_R5_G5_B5; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM )) + return PIPE_FORMAT_A1R5G5B5_UNORM; return default_rgba_format( pipe ); case GL_RGB8: @@ -408,8 +408,8 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A1_R5_G5_B5 )) - return PIPE_FORMAT_U_A1_R5_G5_B5; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM )) + return PIPE_FORMAT_A1R5G5B5_UNORM; return default_rgba_format( pipe ); case GL_ALPHA: @@ -491,18 +491,18 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, #endif case GL_DEPTH_COMPONENT16: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_Z16 )) - return PIPE_FORMAT_U_Z16; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z16_UNORM )) + return PIPE_FORMAT_Z16_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT24: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) - return PIPE_FORMAT_S8_Z24; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24_S8 )) - return PIPE_FORMAT_Z24_S8; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM )) + return PIPE_FORMAT_S8Z24_UNORM; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM )) + return PIPE_FORMAT_Z24S8_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT32: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_Z32 )) - return PIPE_FORMAT_U_Z32; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z32_UNORM )) + return PIPE_FORMAT_Z32_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT: return default_depth_format( pipe ); @@ -514,18 +514,18 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_STENCIL_INDEX16_EXT: if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_S8 )) return PIPE_FORMAT_U_S8; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) - return PIPE_FORMAT_S8_Z24; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24_S8 )) - return PIPE_FORMAT_Z24_S8; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM )) + return PIPE_FORMAT_S8Z24_UNORM; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM )) + return PIPE_FORMAT_Z24S8_UNORM; return PIPE_FORMAT_NONE; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8_Z24 )) - return PIPE_FORMAT_S8_Z24; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24_S8 )) - return PIPE_FORMAT_Z24_S8; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM )) + return PIPE_FORMAT_S8Z24_UNORM; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM )) + return PIPE_FORMAT_Z24S8_UNORM; return PIPE_FORMAT_NONE; default: -- cgit v1.2.3 From 6070a0eb2ba9ba29b861153a10e91c5b463a2ffc Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 16:22:39 -0700 Subject: include tgsi_dump.h --- src/mesa/state_tracker/st_program.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 4e8bdc820e..e64bf14d56 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -38,6 +38,7 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" #include "pipe/draw/draw_context.h" +#include "pipe/tgsi/util/tgsi_dump.h" #include "st_context.h" #include "st_cache.h" -- cgit v1.2.3 From 54fc80ab31f89520d3119196bfa9c6332b35fe2f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 16:46:30 -0700 Subject: Define PIPE_FORMAT_ tokens as an enum set, rather than #defines. This makes debugging a _lot_ easier. In gdb, "print format" used to display 613570600, now you see PIPE_FORMAT_A8R8G8B8_UNORM. --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 2 +- src/mesa/pipe/draw/draw_vertex_fetch.c | 2 +- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 2 +- src/mesa/pipe/i915simple/i915_context.c | 5 +- src/mesa/pipe/i915simple/i915_state_emit.c | 4 +- src/mesa/pipe/i915simple/i915_state_sampler.c | 2 +- src/mesa/pipe/p_context.h | 2 +- src/mesa/pipe/p_format.h | 154 +++++++++++---------- src/mesa/pipe/p_state.h | 7 +- src/mesa/pipe/p_winsys.h | 4 +- src/mesa/pipe/softpipe/sp_context.c | 3 +- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 2 +- src/mesa/pipe/softpipe/sp_tile_cache.c | 3 +- src/mesa/pipe/xlib/xm_winsys.c | 5 +- src/mesa/state_tracker/st_cb_clear.c | 6 +- src/mesa/state_tracker/st_cb_drawpixels.c | 5 +- src/mesa/state_tracker/st_cb_fbo.c | 5 +- src/mesa/state_tracker/st_format.c | 14 +- src/mesa/state_tracker/st_format.h | 18 ++- 19 files changed, 127 insertions(+), 118 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 1b71d0ac10..1799e9b901 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -193,7 +193,7 @@ intel_i915_surface_pitch(struct pipe_winsys *winsys, static struct pipe_surface * -intel_i915_surface_alloc(struct pipe_winsys *winsys, unsigned format) +intel_i915_surface_alloc(struct pipe_winsys *winsys, enum pipe_format format) { struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); if (surf) { diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 77df5c6427..e0759c2e9a 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -46,7 +46,7 @@ * XXX this might be a temporary thing. */ static void -fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) +fetch_attrib4(const void *ptr, enum pipe_format format, float attrib[4]) { /* defaults */ attrib[1] = 0.0; diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 53a1776ffc..29de437595 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -43,7 +43,7 @@ #define DBG 0 static INLINE void -fetch_attrib4(const void *ptr, unsigned format, float attrib[4]) +fetch_attrib4(const void *ptr, enum pipe_format format, float attrib[4]) { /* defaults */ attrib[1] = 0.0; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index cfce116920..b915a67790 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -45,7 +45,7 @@ */ static boolean i915_is_format_supported( struct pipe_context *pipe, - uint format ) + enum pipe_format format ) { #if 0 /* XXX: This is broken -- rewrite if still needed. */ @@ -101,8 +101,9 @@ i915_is_format_supported( struct pipe_context *pipe, case PIPE_FORMAT_R5G6B5_UNORM: case PIPE_FORMAT_S8Z24_UNORM: return TRUE; + default: + return FALSE; }; - return FALSE; #endif } diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 1e0db91024..09bf1fa2d6 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -35,7 +35,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" -static unsigned translate_format( unsigned format ) +static unsigned translate_format( enum pipe_format format ) { switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: @@ -48,7 +48,7 @@ static unsigned translate_format( unsigned format ) } } -static unsigned translate_depth_format( unsigned zformat ) +static unsigned translate_depth_format( enum pipe_format zformat ) { switch (zformat) { case PIPE_FORMAT_S8Z24_UNORM: diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index dd2d51d9f6..59408b6ba0 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -127,7 +127,7 @@ void i915_update_samplers( struct i915_context *i915 ) static uint -translate_texture_format(uint pipeFormat) +translate_texture_format(enum pipe_format pipeFormat) { switch (pipeFormat) { case PIPE_FORMAT_U_L8: diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 2420d02213..b3a2122ade 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -51,7 +51,7 @@ struct pipe_context { * Queries */ boolean (*is_format_supported)( struct pipe_context *pipe, - uint format ); + enum pipe_format format ); const char *(*get_name)( struct pipe_context *pipe ); diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index 86728f77a2..8f11bfab76 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -198,86 +198,90 @@ static INLINE uint pf_rev(pipe_format_ycbcr_t f) * z24s8, compressed textures, ycbcr, etc that won't fit that model. */ -#define PIPE_FORMAT_NONE _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_0000, 0, 0, 0, 0, PIPE_FORMAT_TYPE_UNKNOWN ) -#define PIPE_FORMAT_A8R8G8B8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_B8G8R8A8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_A1R5G5B5_UNORM _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_A4R4G4B4_UNORM _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R5G6B5_UNORM _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_U_L8 _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte luminance */ -#define PIPE_FORMAT_U_A8 _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha */ -#define PIPE_FORMAT_U_I8 _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte intensity */ -#define PIPE_FORMAT_U_A8_L8 _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) /**< ubyte alpha, luminance */ -#define PIPE_FORMAT_YCBCR _PIPE_FORMAT_YCBCR( 0 ) -#define PIPE_FORMAT_YCBCR_REV _PIPE_FORMAT_YCBCR( 1 ) -#define PIPE_FORMAT_Z16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_Z32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_Z32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_S8Z24_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_Z24S8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_S8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R64G64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R64G64B64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R64G64B64A64_FLOAT _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32G32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32G32B32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32G32B32A32_FLOAT _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ) -#define PIPE_FORMAT_R32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32G32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32G32B32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32G32B32A32_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R32_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32G32_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32G32B32_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32G32B32A32_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R32_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32G32_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32G32B32_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32G32B32A32_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R32_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R32G32_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R32G32B32_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R32G32B32A32_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16G16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16G16B16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16G16B16A16_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R16_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16G16_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16G16B16_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16G16B16A16_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R16_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16G16_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16G16B16_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16G16B16A16_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R16_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16G16_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16G16B16_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R16G16B16A16_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8G8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8G8B8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8G8B8A8_UNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ) -#define PIPE_FORMAT_R8_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8G8_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8G8B8_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8G8B8A8_USCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ) -#define PIPE_FORMAT_R8_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8G8_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8G8B8_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8G8B8A8_SNORM _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ) -#define PIPE_FORMAT_R8_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8G8_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8G8B8_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ) -#define PIPE_FORMAT_R8G8B8A8_SSCALED _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) +enum pipe_format { + PIPE_FORMAT_NONE = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_0000, 0, 0, 0, 0, PIPE_FORMAT_TYPE_UNKNOWN ), + PIPE_FORMAT_A8R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_B8G8R8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_A1R5G5B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_A4R4G4B4_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_U_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ + PIPE_FORMAT_U_A8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ + PIPE_FORMAT_U_I8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ + PIPE_FORMAT_U_A8_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha, luminance */ + PIPE_FORMAT_YCBCR = _PIPE_FORMAT_YCBCR( 0 ), + PIPE_FORMAT_YCBCR_REV = _PIPE_FORMAT_YCBCR( 1 ), + PIPE_FORMAT_Z16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_Z32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_Z32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_S8Z24_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_Z24S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte stencil */ + PIPE_FORMAT_R64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R64G64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R64G64B64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R64G64B64A64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32G32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32G32B32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32G32B32A32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32G32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32G32B32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32G32B32A32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32G32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32G32B32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32G32B32A32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32G32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32G32B32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32G32B32A32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32G32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32G32B32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32G32B32A32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16G16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16G16B16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16G16B16A16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16G16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16G16B16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16G16B16A16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16G16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16G16B16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16G16B16A16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16G16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16G16B16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16G16B16A16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8G8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8G8B8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8G8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8G8B8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8G8B8A8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8G8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8G8B8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8G8B8A8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8B8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8B8A8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) +}; + /** - * Duplicated formats: + * XXX should remove this, but S8_UNORM is a poor name */ #define PIPE_FORMAT_U_S8 PIPE_FORMAT_S8_UNORM + /** * Builds pipe format name from format token. */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 44dec9b773..6db9bbc953 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -39,6 +39,7 @@ #define PIPE_STATE_H #include "p_compiler.h" +#include "p_format.h" /** * Implementation limits @@ -265,7 +266,7 @@ struct pipe_surface struct pipe_buffer_handle *buffer; /**< driver private buffer handle */ ubyte *map; /**< only non-NULL when surface is actually mapped */ unsigned map_refcount; /**< Reference count for mapping */ - unsigned format; /**< PIPE_FORMAT_x */ + enum pipe_format format; /**< PIPE_FORMAT_x */ unsigned cpp; /**< bytes per pixel */ unsigned width, height; unsigned pitch; /**< in pixels */ @@ -286,7 +287,7 @@ struct pipe_texture unsigned target; /* XXX convert to PIPE_TEXTURE_x */ unsigned internal_format; /* XXX convert to PIPE_FORMAT_x */ - unsigned format; /**< PIPE_FORMAT_x */ + enum pipe_format format; /**< PIPE_FORMAT_x */ unsigned first_level; unsigned last_level; @@ -331,7 +332,7 @@ struct pipe_vertex_element unsigned vertex_buffer_index:5; unsigned dst_offset:8; - unsigned src_format; /**< PIPE_FORMAT_* */ + enum pipe_format src_format; /**< PIPE_FORMAT_* */ }; diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 5adca1d6fd..438e8bdb63 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -29,6 +29,8 @@ #define P_WINSYS_H +#include "p_format.h" + /** * \file * This is the interface that Gallium3D requires any window system @@ -90,7 +92,7 @@ struct pipe_winsys /** allocate a new surface (no context dependency) */ struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws, - unsigned format); + enum pipe_format format); void (*surface_release)(struct pipe_winsys *ws, struct pipe_surface **s); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 54b2076b4a..bdfd6228ef 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -51,7 +51,8 @@ * parameter or another function. */ static boolean -softpipe_is_format_supported( struct pipe_context *pipe, uint format ) +softpipe_is_format_supported( struct pipe_context *pipe, + enum pipe_format format ) { struct softpipe_context *softpipe = softpipe_context( pipe ); /* ask winsys if the format is supported */ diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 00128fa528..93ea1196f9 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -54,7 +54,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; struct pipe_surface *ps = softpipe->framebuffer.zbuf; - const uint format = ps->format; + const enum pipe_format format = ps->format; unsigned bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ unsigned qzzzz[QUAD_SIZE]; /**< Z values from the quad */ unsigned zmask = 0; diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index be5dd5c289..602b78ebe6 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -174,7 +174,8 @@ is_depth_stencil_surface(struct pipe_surface *ps) * Set pixels in a tile to the given clear color/value. */ static void -clear_tile(struct softpipe_cached_tile *tile, uint format, +clear_tile(struct softpipe_cached_tile *tile, + enum pipe_format format, const float clear_value[4]) { uint i, j; diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 408797c688..976884ad52 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -301,7 +301,7 @@ xm_surface_pitch(struct pipe_winsys *winsys, unsigned cpp, unsigned width, * renderbuffers, etc. */ static struct pipe_surface * -xm_surface_alloc(struct pipe_winsys *ws, GLuint pipeFormat) +xm_surface_alloc(struct pipe_winsys *ws, enum pipe_format pipeFormat) { struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); @@ -379,7 +379,8 @@ xmesa_get_pipe_winsys(void) * can support any format. */ static boolean -xmesa_is_format_supported(struct softpipe_winsys *sws, uint format) +xmesa_is_format_supported(struct softpipe_winsys *sws, + enum pipe_format format) { struct xmesa_softpipe_winsys *xmws = xmesa_softpipe_winsys(sws); diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 1f0d1e6426..b4b2429a2a 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -55,7 +55,7 @@ static GLuint -color_value(GLuint pipeFormat, const GLfloat color[4]) +color_value(enum pipe_format pipeFormat, const GLfloat color[4]) { GLubyte r, g, b, a; @@ -81,7 +81,7 @@ color_value(GLuint pipeFormat, const GLfloat color[4]) static uint -depth_value(GLuint pipeFormat, GLfloat value) +depth_value(enum pipe_format pipeFormat, GLfloat value) { switch (pipeFormat) { case PIPE_FORMAT_Z16_UNORM: @@ -104,7 +104,7 @@ depth_value(GLuint pipeFormat, GLfloat value) static GLboolean -is_depth_stencil_format(GLuint pipeFormat) +is_depth_stencil_format(enum pipe_format pipeFormat) { switch (pipeFormat) { case PIPE_FORMAT_S8Z24_UNORM: diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 464cf5ced3..c0e41dbeec 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -456,7 +456,8 @@ make_texture(struct st_context *st, struct pipe_context *pipe = st->pipe; const struct gl_texture_format *mformat; struct pipe_texture *pt; - GLuint pipeFormat, cpp; + enum pipe_format pipeFormat; + GLuint cpp; GLenum baseFormat; baseFormat = _mesa_base_format(format); @@ -729,7 +730,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, * XXX probably move this to a re-usable place. */ static GLboolean -compatible_formats(GLenum format, GLenum type, GLuint pipeFormat) +compatible_formats(GLenum format, GLenum type, enum pipe_format pipeFormat) { static const GLuint one = 1; GLubyte littleEndian = *((GLubyte *) &one); diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 0ee5f45aae..cbda56b5c3 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -55,7 +55,8 @@ * Compute the renderbuffer's Red/Green/EtcBit fields from the pipe format. */ static int -init_renderbuffer_bits(struct st_renderbuffer *strb, uint pipeFormat) +init_renderbuffer_bits(struct st_renderbuffer *strb, + enum pipe_format pipeFormat) { struct pipe_format_info info; @@ -86,7 +87,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); - const uint pipeFormat + const enum pipe_format pipeFormat = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE); GLuint cpp; GLbitfield flags = PIPE_SURFACE_FLAG_RENDER; /* want to render to surface */ diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 9e249a47a6..8d39e1bec6 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -99,9 +99,7 @@ format_size( * XXX temporary here */ GLboolean -st_get_format_info( - GLuint format, - struct pipe_format_info *pinfo ) +st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) { if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { pipe_format_rgbazs_t info; @@ -222,10 +220,10 @@ st_get_format_info( * Return bytes per pixel for the given format. */ GLuint -st_sizeof_format(GLuint pipeFormat) +st_sizeof_format(enum pipe_format format) { struct pipe_format_info info; - if (!st_get_format_info( pipeFormat, &info )) { + if (!st_get_format_info( format, &info )) { assert( 0 ); return 0; } @@ -237,10 +235,10 @@ st_sizeof_format(GLuint pipeFormat) * Return bytes per pixel for the given format. */ GLenum -st_format_datatype(GLuint pipeFormat) +st_format_datatype(enum pipe_format format) { struct pipe_format_info info; - if (!st_get_format_info( pipeFormat, &info )) { + if (!st_get_format_info( format, &info )) { assert( 0 ); return 0; } @@ -248,7 +246,7 @@ st_format_datatype(GLuint pipeFormat) } -GLuint +enum pipe_format st_mesa_format_to_pipe_format(GLuint mesaFormat) { switch (mesaFormat) { diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index 23abc36edf..6ccf5536f9 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -26,13 +26,13 @@ **************************************************************************/ -#ifndef ST_CB_TEXIMAGE_H -#define ST_CB_TEXIMAGE_H +#ifndef ST_FORMAT_H +#define ST_FORMAT_H struct pipe_format_info { - GLuint format; + enum pipe_format format; GLenum base_format; GLenum datatype; GLubyte red_bits; @@ -47,21 +47,19 @@ struct pipe_format_info }; -extern GLboolean -st_get_format_info( - GLuint format, - struct pipe_format_info *pinfo ); +GLboolean +st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo); extern GLuint -st_sizeof_format(GLuint pipeFormat); +st_sizeof_format(enum pipe_format format); extern GLenum -st_format_datatype(GLuint pipeFormat); +st_format_datatype(enum pipe_format format); -extern GLuint +extern enum pipe_format st_mesa_format_to_pipe_format(GLuint mesaFormat); -- cgit v1.2.3 From 2d2bb35f96de6a2c498a2309a4f924acd256c317 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 17:11:30 -0700 Subject: Move _mesa_init_glsl_driver_functions() into shader_api.c This allows making a bunch of functions static, and removes a state tracker dependency on driverfuncs.c --- src/mesa/drivers/common/driverfuncs.c | 37 -------- src/mesa/drivers/common/driverfuncs.h | 4 - src/mesa/shader/shader_api.c | 153 +++++++++++++++++++++------------- src/mesa/shader/shader_api.h | 103 +---------------------- src/mesa/state_tracker/st_context.c | 2 +- 5 files changed, 100 insertions(+), 199 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c index 96e5037fa5..ea0cc51d4a 100644 --- a/src/mesa/drivers/common/driverfuncs.c +++ b/src/mesa/drivers/common/driverfuncs.c @@ -259,43 +259,6 @@ _mesa_init_driver_functions(struct dd_function_table *driver) } -/** - * Plug in Mesa's GLSL functions. - */ -void -_mesa_init_glsl_driver_functions(struct dd_function_table *driver) -{ - driver->AttachShader = _mesa_attach_shader; - driver->BindAttribLocation = _mesa_bind_attrib_location; - driver->CompileShader = _mesa_compile_shader; - driver->CreateProgram = _mesa_create_program; - driver->CreateShader = _mesa_create_shader; - driver->DeleteProgram2 = _mesa_delete_program2; - driver->DeleteShader = _mesa_delete_shader; - driver->DetachShader = _mesa_detach_shader; - driver->GetActiveAttrib = _mesa_get_active_attrib; - driver->GetActiveUniform = _mesa_get_active_uniform; - driver->GetAttachedShaders = _mesa_get_attached_shaders; - driver->GetAttribLocation = _mesa_get_attrib_location; - driver->GetHandle = _mesa_get_handle; - driver->GetProgramiv = _mesa_get_programiv; - driver->GetProgramInfoLog = _mesa_get_program_info_log; - driver->GetShaderiv = _mesa_get_shaderiv; - driver->GetShaderInfoLog = _mesa_get_shader_info_log; - driver->GetShaderSource = _mesa_get_shader_source; - driver->GetUniformfv = _mesa_get_uniformfv; - driver->GetUniformLocation = _mesa_get_uniform_location; - driver->IsProgram = _mesa_is_program; - driver->IsShader = _mesa_is_shader; - driver->LinkProgram = _mesa_link_program; - driver->ShaderSource = _mesa_shader_source; - driver->Uniform = _mesa_uniform; - driver->UniformMatrix = _mesa_uniform_matrix; - driver->UseProgram = _mesa_use_program; - driver->ValidateProgram = _mesa_validate_program; -} - - /** * Call the ctx->Driver.* state functions with current values to initialize * driver state. diff --git a/src/mesa/drivers/common/driverfuncs.h b/src/mesa/drivers/common/driverfuncs.h index 6ed23c4520..4c90ed12f6 100644 --- a/src/mesa/drivers/common/driverfuncs.h +++ b/src/mesa/drivers/common/driverfuncs.h @@ -30,10 +30,6 @@ extern void _mesa_init_driver_functions(struct dd_function_table *driver); -extern void -_mesa_init_glsl_driver_functions(struct dd_function_table *driver); - - extern void _mesa_init_driver_state(GLcontext *ctx); diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index bc5ecdaa15..4470a87af0 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -52,7 +52,7 @@ /** * Allocate a new gl_shader_program object, initialize it. */ -struct gl_shader_program * +static struct gl_shader_program * _mesa_new_shader_program(GLcontext *ctx, GLuint name) { struct gl_shader_program *shProg; @@ -371,7 +371,7 @@ copy_string(GLchar *dst, GLsizei maxLength, GLsizei *length, const GLchar *src) /** * Called via ctx->Driver.AttachShader() */ -void +static void _mesa_attach_shader(GLcontext *ctx, GLuint program, GLuint shader) { struct gl_shader_program *shProg @@ -410,7 +410,38 @@ _mesa_attach_shader(GLcontext *ctx, GLuint program, GLuint shader) } -void +static GLint +_mesa_get_attrib_location(GLcontext *ctx, GLuint program, + const GLchar *name) +{ + struct gl_shader_program *shProg + = _mesa_lookup_shader_program(ctx, program); + + if (!shProg) { + _mesa_error(ctx, GL_INVALID_VALUE, "glGetAttribLocation"); + return -1; + } + + if (!shProg->LinkStatus) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glGetAttribLocation(program not linked)"); + return -1; + } + + if (!name) + return -1; + + if (shProg->Attributes) { + GLint i = _mesa_lookup_parameter_index(shProg->Attributes, -1, name); + if (i >= 0) { + return shProg->Attributes->Parameters[i].StateIndexes[0]; + } + } + return -1; +} + + +static void _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, const GLchar *name) { @@ -456,7 +487,7 @@ _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, } -GLuint +static GLuint _mesa_create_shader(GLcontext *ctx, GLenum type) { struct gl_shader *sh; @@ -480,7 +511,7 @@ _mesa_create_shader(GLcontext *ctx, GLenum type) } -GLuint +static GLuint _mesa_create_program(GLcontext *ctx) { GLuint name; @@ -501,7 +532,7 @@ _mesa_create_program(GLcontext *ctx) * Named w/ "2" to indicate OpenGL 2.x vs GL_ARB_fragment_programs's * DeleteProgramARB. */ -void +static void _mesa_delete_program2(GLcontext *ctx, GLuint name) { /* @@ -527,7 +558,7 @@ _mesa_delete_program2(GLcontext *ctx, GLuint name) } -void +static void _mesa_delete_shader(GLcontext *ctx, GLuint shader) { struct gl_shader *sh = _mesa_lookup_shader(ctx, shader); @@ -542,7 +573,7 @@ _mesa_delete_shader(GLcontext *ctx, GLuint shader) } -void +static void _mesa_detach_shader(GLcontext *ctx, GLuint program, GLuint shader) { struct gl_shader_program *shProg @@ -602,7 +633,7 @@ _mesa_detach_shader(GLcontext *ctx, GLuint program, GLuint shader) } -void +static void _mesa_get_active_attrib(GLcontext *ctx, GLuint program, GLuint index, GLsizei maxLength, GLsizei *length, GLint *size, GLenum *type, GLchar *nameOut) @@ -637,7 +668,7 @@ _mesa_get_active_attrib(GLcontext *ctx, GLuint program, GLuint index, /** * Called via ctx->Driver.GetActiveUniform(). */ -void +static void _mesa_get_active_uniform(GLcontext *ctx, GLuint program, GLuint index, GLsizei maxLength, GLsizei *length, GLint *size, GLenum *type, GLchar *nameOut) @@ -681,7 +712,7 @@ _mesa_get_active_uniform(GLcontext *ctx, GLuint program, GLuint index, /** * Called via ctx->Driver.GetAttachedShaders(). */ -void +static void _mesa_get_attached_shaders(GLcontext *ctx, GLuint program, GLsizei maxCount, GLsizei *count, GLuint *obj) { @@ -701,38 +732,7 @@ _mesa_get_attached_shaders(GLcontext *ctx, GLuint program, GLsizei maxCount, } -GLint -_mesa_get_attrib_location(GLcontext *ctx, GLuint program, - const GLchar *name) -{ - struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); - - if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, "glGetAttribLocation"); - return -1; - } - - if (!shProg->LinkStatus) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glGetAttribLocation(program not linked)"); - return -1; - } - - if (!name) - return -1; - - if (shProg->Attributes) { - GLint i = _mesa_lookup_parameter_index(shProg->Attributes, -1, name); - if (i >= 0) { - return shProg->Attributes->Parameters[i].StateIndexes[0]; - } - } - return -1; -} - - -GLuint +static GLuint _mesa_get_handle(GLcontext *ctx, GLenum pname) { #if 0 @@ -756,7 +756,7 @@ _mesa_get_handle(GLcontext *ctx, GLenum pname) } -void +static void _mesa_get_programiv(GLcontext *ctx, GLuint program, GLenum pname, GLint *params) { @@ -810,7 +810,7 @@ _mesa_get_programiv(GLcontext *ctx, GLuint program, } -void +static void _mesa_get_shaderiv(GLcontext *ctx, GLuint name, GLenum pname, GLint *params) { struct gl_shader *shader = _mesa_lookup_shader(ctx, name); @@ -843,7 +843,7 @@ _mesa_get_shaderiv(GLcontext *ctx, GLuint name, GLenum pname, GLint *params) } -void +static void _mesa_get_program_info_log(GLcontext *ctx, GLuint program, GLsizei bufSize, GLsizei *length, GLchar *infoLog) { @@ -857,7 +857,7 @@ _mesa_get_program_info_log(GLcontext *ctx, GLuint program, GLsizei bufSize, } -void +static void _mesa_get_shader_info_log(GLcontext *ctx, GLuint shader, GLsizei bufSize, GLsizei *length, GLchar *infoLog) { @@ -873,7 +873,7 @@ _mesa_get_shader_info_log(GLcontext *ctx, GLuint shader, GLsizei bufSize, /** * Called via ctx->Driver.GetShaderSource(). */ -void +static void _mesa_get_shader_source(GLcontext *ctx, GLuint shader, GLsizei maxLength, GLsizei *length, GLchar *sourceOut) { @@ -889,7 +889,7 @@ _mesa_get_shader_source(GLcontext *ctx, GLuint shader, GLsizei maxLength, /** * Called via ctx->Driver.GetUniformfv(). */ -void +static void _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, GLfloat *params) { @@ -915,7 +915,7 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, /** * Called via ctx->Driver.GetUniformLocation(). */ -GLint +static GLint _mesa_get_uniform_location(GLcontext *ctx, GLuint program, const GLchar *name) { struct gl_shader_program *shProg @@ -940,7 +940,7 @@ _mesa_get_uniform_location(GLcontext *ctx, GLuint program, const GLchar *name) } -GLboolean +static GLboolean _mesa_is_program(GLcontext *ctx, GLuint name) { struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, name); @@ -948,7 +948,7 @@ _mesa_is_program(GLcontext *ctx, GLuint name) } -GLboolean +static GLboolean _mesa_is_shader(GLcontext *ctx, GLuint name) { struct gl_shader *shader = _mesa_lookup_shader(ctx, name); @@ -960,7 +960,7 @@ _mesa_is_shader(GLcontext *ctx, GLuint name) /** * Called via ctx->Driver.ShaderSource() */ -void +static void _mesa_shader_source(GLcontext *ctx, GLuint shader, const GLchar *source) { struct gl_shader *sh = _mesa_lookup_shader(ctx, shader); @@ -981,7 +981,7 @@ _mesa_shader_source(GLcontext *ctx, GLuint shader, const GLchar *source) /** * Called via ctx->Driver.CompileShader() */ -void +static void _mesa_compile_shader(GLcontext *ctx, GLuint shaderObj) { struct gl_shader *sh = _mesa_lookup_shader(ctx, shaderObj); @@ -998,7 +998,7 @@ _mesa_compile_shader(GLcontext *ctx, GLuint shaderObj) /** * Called via ctx->Driver.LinkProgram() */ -void +static void _mesa_link_program(GLcontext *ctx, GLuint program) { struct gl_shader_program *shProg; @@ -1070,7 +1070,7 @@ update_textures_used(struct gl_program *prog) /** * Called via ctx->Driver.Uniform(). */ -void +static void _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, const GLvoid *values, GLenum type) { @@ -1182,7 +1182,7 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, /** * Called by ctx->Driver.UniformMatrix(). */ -void +static void _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, GLenum matrixType, GLint location, GLsizei count, GLboolean transpose, const GLfloat *values) @@ -1230,7 +1230,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, } -void +static void _mesa_validate_program(GLcontext *ctx, GLuint program) { struct gl_shader_program *shProg; @@ -1256,3 +1256,40 @@ _mesa_validate_program(GLcontext *ctx, GLuint program) image units allowed. */ } + + +/** + * Plug in Mesa's GLSL functions into the device driver function table. + */ +void +_mesa_init_glsl_driver_functions(struct dd_function_table *driver) +{ + driver->AttachShader = _mesa_attach_shader; + driver->BindAttribLocation = _mesa_bind_attrib_location; + driver->CompileShader = _mesa_compile_shader; + driver->CreateProgram = _mesa_create_program; + driver->CreateShader = _mesa_create_shader; + driver->DeleteProgram2 = _mesa_delete_program2; + driver->DeleteShader = _mesa_delete_shader; + driver->DetachShader = _mesa_detach_shader; + driver->GetActiveAttrib = _mesa_get_active_attrib; + driver->GetActiveUniform = _mesa_get_active_uniform; + driver->GetAttachedShaders = _mesa_get_attached_shaders; + driver->GetAttribLocation = _mesa_get_attrib_location; + driver->GetHandle = _mesa_get_handle; + driver->GetProgramiv = _mesa_get_programiv; + driver->GetProgramInfoLog = _mesa_get_program_info_log; + driver->GetShaderiv = _mesa_get_shaderiv; + driver->GetShaderInfoLog = _mesa_get_shader_info_log; + driver->GetShaderSource = _mesa_get_shader_source; + driver->GetUniformfv = _mesa_get_uniformfv; + driver->GetUniformLocation = _mesa_get_uniform_location; + driver->IsProgram = _mesa_is_program; + driver->IsShader = _mesa_is_shader; + driver->LinkProgram = _mesa_link_program; + driver->ShaderSource = _mesa_shader_source; + driver->Uniform = _mesa_uniform; + driver->UniformMatrix = _mesa_uniform_matrix; + driver->UseProgram = _mesa_use_program; + driver->ValidateProgram = _mesa_validate_program; +} diff --git a/src/mesa/shader/shader_api.h b/src/mesa/shader/shader_api.h index 27e5870d70..5521c585b5 100644 --- a/src/mesa/shader/shader_api.h +++ b/src/mesa/shader/shader_api.h @@ -41,9 +41,10 @@ _mesa_init_shader_state(GLcontext * ctx); extern void _mesa_free_shader_state(GLcontext *ctx); +/* extern struct gl_shader_program * _mesa_new_shader_program(GLcontext *ctx, GLuint name); - +*/ extern void _mesa_clear_shader_program_data(GLcontext *ctx, struct gl_shader_program *shProg); @@ -78,108 +79,12 @@ extern struct gl_shader * _mesa_lookup_shader(GLcontext *ctx, GLuint name); -/** - * API/Driver functions - */ - -extern void -_mesa_attach_shader(GLcontext *ctx, GLuint program, GLuint shader); - -extern void -_mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, - const GLchar *name); - -extern void -_mesa_compile_shader(GLcontext *ctx, GLuint shaderObj); - -extern GLuint -_mesa_create_shader(GLcontext *ctx, GLenum type); - -extern GLuint -_mesa_create_program(GLcontext *ctx); - -extern void -_mesa_delete_program2(GLcontext *ctx, GLuint name); - -extern void -_mesa_delete_shader(GLcontext *ctx, GLuint shader); - -extern void -_mesa_detach_shader(GLcontext *ctx, GLuint program, GLuint shader); - -extern void -_mesa_get_active_attrib(GLcontext *ctx, GLuint program, GLuint index, - GLsizei maxLength, GLsizei *length, GLint *size, - GLenum *type, GLchar *name); - -extern void -_mesa_get_active_uniform(GLcontext *ctx, GLuint program, GLuint index, - GLsizei maxLength, GLsizei *length, GLint *size, - GLenum *type, GLchar *name); - -extern void -_mesa_get_attached_shaders(GLcontext *ctx, GLuint program, GLsizei maxCount, - GLsizei *count, GLuint *obj); - -extern GLint -_mesa_get_attrib_location(GLcontext *ctx, GLuint program, - const GLchar *name); - -extern GLuint -_mesa_get_handle(GLcontext *ctx, GLenum pname); - -extern void -_mesa_get_programiv(GLcontext *ctx, GLuint program, - GLenum pname, GLint *params); - -extern void -_mesa_get_program_info_log(GLcontext *ctx, GLuint program, GLsizei bufSize, - GLsizei *length, GLchar *infoLog); - -extern void -_mesa_get_shaderiv(GLcontext *ctx, GLuint shader, GLenum pname, GLint *params); - -extern void -_mesa_get_shader_info_log(GLcontext *ctx, GLuint shader, GLsizei bufSize, - GLsizei *length, GLchar *infoLog); - -extern void -_mesa_get_shader_source(GLcontext *ctx, GLuint shader, GLsizei maxLength, - GLsizei *length, GLchar *sourceOut); - -extern void -_mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, - GLfloat *params); - -extern GLint -_mesa_get_uniform_location(GLcontext *ctx, GLuint program, const GLchar *name); - -extern GLboolean -_mesa_is_program(GLcontext *ctx, GLuint name); - -extern GLboolean -_mesa_is_shader(GLcontext *ctx, GLuint name); - -extern void -_mesa_link_program(GLcontext *ctx, GLuint program); - -extern void -_mesa_shader_source(GLcontext *ctx, GLuint shader, const GLchar *source); - -extern void -_mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, - const GLvoid *values, GLenum type); - -void -_mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, - GLenum matrixType, GLint location, GLsizei count, - GLboolean transpose, const GLfloat *values); - extern void _mesa_use_program(GLcontext *ctx, GLuint program); + extern void -_mesa_validate_program(GLcontext *ctx, GLuint program); +_mesa_init_glsl_driver_functions(struct dd_function_table *driver); #endif /* SHADER_API_H */ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index abde6d64b0..524e06fb00 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -29,7 +29,7 @@ #include "main/context.h" #include "main/extensions.h" #include "vbo/vbo.h" -#include "drivers/common/driverfuncs.h" +#include "shader/shader_api.h" #include "st_public.h" #include "st_context.h" #include "st_cb_accum.h" -- cgit v1.2.3 From 40e9c57d318c1d4a1e75fe0c88374182050d7f48 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 17:14:02 -0700 Subject: Disable swrast, swrast_setup and driverfuncs in the build. The gallium/state tracker no longer uses them and this speeds up the build a bit. --- src/mesa/sources | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 8286e558d6..2f16af3aa4 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -80,7 +80,7 @@ MATH_SOURCES = \ math/m_vector.c \ math/m_xform.c -SWRAST_SOURCES = \ +__SWRAST_SOURCES = \ swrast/s_aaline.c \ swrast/s_aatriangle.c \ swrast/s_accum.c \ @@ -111,7 +111,7 @@ SWRAST_SOURCES = \ swrast/s_triangle.c \ swrast/s_zoom.c -SWRAST_SETUP_SOURCES = \ +__SWRAST_SETUP_SOURCES = \ swrast_setup/ss_context.c \ swrast_setup/ss_triangle.c @@ -326,7 +326,7 @@ SPARC_SOURCES = \ SPARC_API = \ sparc/glapi_sparc.S -COMMON_DRIVER_SOURCES = \ +__COMMON_DRIVER_SOURCES = \ drivers/common/driverfuncs.c X11_DRIVER_SOURCES = \ -- cgit v1.2.3 From c885775cae8feabe5431ba7867fac99332f5ee8e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 17:24:42 -0700 Subject: New vbo_set_draw_func() to keep vbo context opaque to state tracker and tnl module. --- src/mesa/state_tracker/st_cb_feedback.c | 8 +++----- src/mesa/state_tracker/st_draw.c | 6 +----- src/mesa/tnl/t_context.c | 3 +++ src/mesa/vbo/vbo.h | 3 +++ src/mesa/vbo/vbo_context.c | 12 ++++++++++++ 5 files changed, 22 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index a9fd2579a2..ea775b9452 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -43,7 +43,6 @@ #include "main/macros.h" #include "vbo/vbo.h" -#include "vbo/vbo_context.h" #include "st_context.h" #include "st_atom.h" @@ -281,26 +280,25 @@ static void st_RenderMode(GLcontext *ctx, GLenum newMode ) { struct st_context *st = ctx->st; - struct vbo_context *vbo = (struct vbo_context *) ctx->swtnl_im; struct draw_context *draw = st->draw; if (newMode == GL_RENDER) { /* restore normal VBO draw function */ - vbo->draw_prims = st_draw_vbo; + vbo_set_draw_func(ctx, st_draw_vbo); } else if (newMode == GL_SELECT) { if (!st->selection_stage) st->selection_stage = draw_glselect_stage(ctx, draw); draw_set_rasterize_stage(draw, st->selection_stage); /* Plug in new vbo draw function */ - vbo->draw_prims = st_feedback_draw_vbo; + vbo_set_draw_func(ctx, st_feedback_draw_vbo); } else { if (!st->feedback_stage) st->feedback_stage = draw_glfeedback_stage(ctx, draw); draw_set_rasterize_stage(draw, st->feedback_stage); /* Plug in new vbo draw function */ - vbo->draw_prims = st_feedback_draw_vbo; + vbo_set_draw_func(ctx, st_feedback_draw_vbo); /* need to generate/use a vertex program that emits pos/color/tex */ st->dirty.st |= ST_NEW_VERTEX_PROGRAM; } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index b4c0d0cdd6..32dcd73c46 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -34,7 +34,6 @@ #include "main/image.h" #include "vbo/vbo.h" -#include "vbo/vbo_context.h" #include "st_atom.h" #include "st_cache.h" @@ -519,14 +518,11 @@ st_feedback_draw_vbo(GLcontext *ctx, void st_init_draw( struct st_context *st ) { GLcontext *ctx = st->ctx; - struct vbo_context *vbo = (struct vbo_context *) ctx->swtnl_im; /* actually, not used here, but elsewhere */ create_default_attribs_buffer(st); - assert(vbo); - assert(vbo->draw_prims); - vbo->draw_prims = st_draw_vbo; + vbo_set_draw_func(ctx, st_draw_vbo); } diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c index b87452d2eb..0ace5c2d6f 100644 --- a/src/mesa/tnl/t_context.c +++ b/src/mesa/tnl/t_context.c @@ -78,6 +78,9 @@ _tnl_CreateContext( GLcontext *ctx ) tnl->nr_blocks = 0; + /* plug in the VBO drawing function */ + vbo_set_draw_func(ctx, _tnl_draw_prims); + return GL_TRUE; } diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h index 4c51b44cda..79d33d09c1 100644 --- a/src/mesa/vbo/vbo.h +++ b/src/mesa/vbo/vbo.h @@ -117,4 +117,7 @@ void vbo_rebase_prims( GLcontext *ctx, void vbo_use_buffer_objects(GLcontext *ctx); +void vbo_set_draw_func(GLcontext *ctx, vbo_draw_func func); + + #endif diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c index ad4556c500..60d0b76ab6 100644 --- a/src/mesa/vbo/vbo_context.c +++ b/src/mesa/vbo/vbo_context.c @@ -31,6 +31,7 @@ #include "vbo.h" #include "vbo_context.h" +#if 0 /* Reach out and grab this to use as the default: */ extern void _tnl_draw_prims( GLcontext *ctx, @@ -40,6 +41,7 @@ extern void _tnl_draw_prims( GLcontext *ctx, const struct _mesa_index_buffer *ib, GLuint min_index, GLuint max_index ); +#endif @@ -214,7 +216,9 @@ GLboolean _vbo_CreateContext( GLcontext *ctx ) /* By default: */ +#if 0 /* dead - see vbo_set_draw_func() */ vbo->draw_prims = _tnl_draw_prims; +#endif /* Hook our functions into exec and compile dispatch tables. These * will pretty much be permanently installed, which means that the @@ -245,3 +249,11 @@ void _vbo_DestroyContext( GLcontext *ctx ) FREE(vbo_context(ctx)); ctx->swtnl_im = NULL; } + + +void vbo_set_draw_func(GLcontext *ctx, vbo_draw_func func) +{ + struct vbo_context *vbo = vbo_context(ctx); + vbo->draw_prims = func; +} + -- cgit v1.2.3 From 51537e0b6efec61e10db522ba2c0c66b407d7f26 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 17:26:06 -0700 Subject: Disable, remove tnl module from the Gallium build since no longer used. --- src/mesa/sources | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 2f16af3aa4..32e1805eca 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -115,7 +115,7 @@ __SWRAST_SETUP_SOURCES = \ swrast_setup/ss_context.c \ swrast_setup/ss_triangle.c -TNL_SOURCES = \ +__TNL_SOURCES = \ tnl/t_context.c \ tnl/t_pipeline.c \ tnl/t_draw.c \ -- cgit v1.2.3 From 127a493fe2a2f7fd8515b278185322dfacdad357 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 20:13:10 -0700 Subject: Tile clearing optimizations. Clear using int values rather than floats when possible. Better performance now. --- src/mesa/pipe/softpipe/sp_clear.c | 42 +----- src/mesa/pipe/softpipe/sp_tile_cache.c | 252 ++++++++++++++++++--------------- src/mesa/pipe/softpipe/sp_tile_cache.h | 4 +- 3 files changed, 147 insertions(+), 151 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index a6352e0616..bfe4a29144 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -51,49 +51,15 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, softpipe_update_derived(softpipe); /* not needed?? */ if (ps == sp_tile_cache_get_surface(softpipe->zbuf_cache)) { - float clear[4]; - clear[0] = 1.0; /* XXX hack */ - sp_tile_cache_clear(softpipe->zbuf_cache, clear); + sp_tile_cache_clear(softpipe->zbuf_cache, clearValue); } else if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[0])) { - float clear[4]; - /* XXX it sure would be nice if the clear color was passed to - * this function as float[4].... - */ - uint r, g, b, a; - switch (ps->format) { - case PIPE_FORMAT_R8G8B8A8_UNORM: - r = (clearValue >> 24) & 0xff; - g = (clearValue >> 16) & 0xff; - g = (clearValue >> 8) & 0xff; - a = (clearValue ) & 0xff; - break; - case PIPE_FORMAT_A8R8G8B8_UNORM: - r = (clearValue >> 16) & 0xff; - g = (clearValue >> 8) & 0xff; - b = (clearValue ) & 0xff; - a = (clearValue >> 24) & 0xff; - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - r = (clearValue >> 8) & 0xff; - g = (clearValue >> 16) & 0xff; - b = (clearValue >> 24) & 0xff; - a = (clearValue ) & 0xff; - break; - default: - assert(0); - } - - clear[0] = r / 255.0; - clear[1] = g / 255.0; - clear[2] = b / 255.0; - clear[3] = a / 255.0; - - sp_tile_cache_clear(softpipe->cbuf_cache[0], clear); + sp_tile_cache_clear(softpipe->cbuf_cache[0], clearValue); } +#if !TILE_CLEAR_OPTIMIZATION pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); - +#endif #if 0 sp_clear_tile_cache(ps, clearValue); diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 602b78ebe6..13774e6c8f 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -52,7 +52,9 @@ struct softpipe_tile_cache struct pipe_texture *texture; /**< if caching a texture */ struct softpipe_cached_tile entries[NUM_ENTRIES]; uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; - float clear_value[4]; + float clear_color[4]; + uint clear_val; + boolean depth_stencil; /** Is the surface a depth/stencil format? */ }; @@ -67,6 +69,9 @@ struct softpipe_tile_cache +/** + * Is the tile at (x,y) in cleared state? + */ static INLINE uint is_clear_flag_set(const uint *bitvec, int x, int y) { @@ -80,6 +85,9 @@ is_clear_flag_set(const uint *bitvec, int x, int y) } +/** + * Mark the tile at (x,y) as not cleared. + */ static INLINE void clear_clear_flag(uint *bitvec, int x, int y) { @@ -120,10 +128,15 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc) } +/** + * Specify the surface to cache. + */ void sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, struct pipe_surface *ps) { + assert(!tc->texture); + if (tc->surface && tc->surface->map) { assert(tc->surface != ps); pipe_surface_unmap(tc->surface); @@ -133,9 +146,19 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, if (!ps->map) pipe_surface_map(ps); + + if (ps) { + tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || + ps->format == PIPE_FORMAT_Z16_UNORM || + ps->format == PIPE_FORMAT_Z32_UNORM || + ps->format == PIPE_FORMAT_U_S8); + } } +/** + * Return the surface being cached. + */ struct pipe_surface * sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) { @@ -143,12 +166,17 @@ sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) } +/** + * Specify the texture to cache. + */ void sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, struct pipe_texture *texture) { uint i; + assert(!tc->surface); + tc->texture = texture; /* mark as entries as invalid/empty */ @@ -159,84 +187,71 @@ sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, } -static INLINE boolean -is_depth_stencil_surface(struct pipe_surface *ps) +/** + * Set pixels in a tile to the given clear color/value, float. + */ +static void +clear_tile_rgba(struct softpipe_cached_tile *tile, + enum pipe_format format, + const float clear_value[4]) { - return (ps && - (ps->format == PIPE_FORMAT_S8Z24_UNORM || - ps->format == PIPE_FORMAT_Z16_UNORM || - ps->format == PIPE_FORMAT_Z32_UNORM || - ps->format == PIPE_FORMAT_U_S8)); + if (clear_value[0] == 0.0 && + clear_value[1] == 0.0 && + clear_value[2] == 0.0 && + clear_value[3] == 0.0) { + memset(tile->data.color, 0, sizeof(tile->data.color)); + } + else { + uint i, j; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.color[i][j][0] = clear_value[0]; + tile->data.color[i][j][1] = clear_value[1]; + tile->data.color[i][j][2] = clear_value[2]; + tile->data.color[i][j][3] = clear_value[3]; + } + } + } } /** - * Set pixels in a tile to the given clear color/value. + * Set a tile to a solid value/color. */ static void clear_tile(struct softpipe_cached_tile *tile, enum pipe_format format, - const float clear_value[4]) + uint clear_value) { uint i, j; switch (format) { - case PIPE_FORMAT_Z16_UNORM: - { - ushort clear_val = (ushort) (clear_value[0] * 0xffff); - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.depth16[i][j] = clear_val; - } - } - } - break; - case PIPE_FORMAT_Z32_UNORM: - { - uint clear_val = (uint) (clear_value[0] * 0xffffffff); - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.depth32[i][j] = clear_val; - } - } - } + case PIPE_FORMAT_U_S8: + /* 8 bpp */ + memset(tile->data.any, 0, TILE_SIZE * TILE_SIZE); break; - case PIPE_FORMAT_S8Z24_UNORM: - { - uint clear_val = (uint) (clear_value[0] * 0xffffff); - clear_val |= ((uint) clear_value[1]) << 24; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.depth32[i][j] = clear_val; - } - } + case PIPE_FORMAT_Z16_UNORM: + /* 16 bpp */ + if (clear_value == 0) { + memset(tile->data.any, 0, 2 * TILE_SIZE * TILE_SIZE); } - break; - case PIPE_FORMAT_U_S8: - { - ubyte clear_val = (uint) clear_value[0]; + else { for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { - tile->data.stencil8[i][j] = clear_val; + tile->data.depth16[i][j] = clear_value; } } } break; default: - /* color */ - if (clear_value[0] == 0.0 && - clear_value[1] == 0.0 && - clear_value[2] == 0.0 && - clear_value[3] == 0.0) { - memset(tile->data.color, 0, sizeof(tile->data.color)); + /* 32 bpp */ + if (clear_value == 0) { + memset(tile->data.any, 0, 4 * TILE_SIZE * TILE_SIZE); } else { for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { - tile->data.color[i][j][0] = clear_value[0]; - tile->data.color[i][j][1] = clear_value[1]; - tile->data.color[i][j][2] = clear_value[2]; - tile->data.color[i][j][3] = clear_value[3]; + tile->data.color32[i][j] = clear_value; } } } @@ -244,52 +259,6 @@ clear_tile(struct softpipe_cached_tile *tile, } -static INLINE void -put_tile(struct pipe_context *pipe, - struct pipe_surface *ps, - struct softpipe_cached_tile *tile) -{ - assert(tile->x >= 0); - assert(tile->x % TILE_SIZE == 0); - assert(tile->y >= 0); - assert(tile->y % TILE_SIZE == 0); - - if (is_depth_stencil_surface(ps)) { - pipe->put_tile(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); - } - else { - pipe->put_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - } -} - - -static INLINE void -get_tile(struct pipe_context *pipe, - struct pipe_surface *ps, - struct softpipe_cached_tile *tile) -{ - assert(tile->x >= 0); - assert(tile->x % TILE_SIZE == 0); - assert(tile->y >= 0); - assert(tile->y % TILE_SIZE == 0); - - if (is_depth_stencil_surface(ps)) { - pipe->get_tile(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); - } - else { - pipe->get_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - } -} - - /** * Actually clear the tiles which were flagged as being in a clear state. */ @@ -305,15 +274,15 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, uint numCleared = 0; /* clear one tile to the clear value */ - clear_tile(&tile, ps->format, tc->clear_value); + clear_tile(&tile, ps->format, tc->clear_val); /* push the tile to all positions marked as clear */ for (y = 0; y < h; y += TILE_SIZE) { for (x = 0; x < w; x += TILE_SIZE) { if (is_clear_flag_set(tc->clear_flags, x, y)) { - tile.x = x; - tile.y = y; - put_tile(pipe, ps, &tile); + pipe->put_tile(pipe, ps, + x, y, TILE_SIZE, TILE_SIZE, + tile.data.color32, 0/*STRIDE*/); /* do this? */ clear_clear_flag(tc->clear_flags, x, y); @@ -349,7 +318,16 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, for (pos = 0; pos < NUM_ENTRIES; pos++) { struct softpipe_cached_tile *tile = tc->entries + pos; if (tile->x >= 0) { - put_tile(pipe, ps, tile); + if (tc->depth_stencil) { + pipe->put_tile(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); + } + else { + pipe->put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } tile->x = tile->y = -1; /* mark as empty */ inuse++; } @@ -389,7 +367,16 @@ sp_get_cached_tile(struct softpipe_context *softpipe, if (tile->x != -1) { /* put dirty tile back in framebuffer */ - put_tile(pipe, ps, tile); + if (tc->depth_stencil) { + pipe->put_tile(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); + } + else { + pipe->put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } } tile->x = tile_x; @@ -397,12 +384,26 @@ sp_get_cached_tile(struct softpipe_context *softpipe, if (is_clear_flag_set(tc->clear_flags, x, y)) { /* don't get tile from framebuffer, just clear it */ - clear_tile(tile, ps->format, tc->clear_value); + if (tc->depth_stencil) { + clear_tile(tile, ps->format, tc->clear_val); + } + else { + clear_tile_rgba(tile, ps->format, tc->clear_color); + } clear_clear_flag(tc->clear_flags, x, y); } else { /* get new tile data from surface */ - get_tile(pipe, ps, tile); + if (tc->depth_stencil) { + pipe->get_tile(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); + } + else { + pipe->get_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } } } @@ -477,12 +478,39 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, * Save the color and set a 'clearflag' for each tile of the screen. */ void -sp_tile_cache_clear(struct softpipe_tile_cache *tc, const float value[4]) +sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue) { - tc->clear_value[0] = value[0]; - tc->clear_value[1] = value[1]; - tc->clear_value[2] = value[2]; - tc->clear_value[3] = value[3]; + uint r, g, b, a; + + tc->clear_val = clearValue; + + switch (tc->surface->format) { + case PIPE_FORMAT_R8G8B8A8_UNORM: + r = (clearValue >> 24) & 0xff; + g = (clearValue >> 16) & 0xff; + b = (clearValue >> 8) & 0xff; + a = (clearValue ) & 0xff; + break; + case PIPE_FORMAT_A8R8G8B8_UNORM: + r = (clearValue >> 16) & 0xff; + g = (clearValue >> 8) & 0xff; + b = (clearValue ) & 0xff; + a = (clearValue >> 24) & 0xff; + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + r = (clearValue >> 8) & 0xff; + g = (clearValue >> 16) & 0xff; + b = (clearValue >> 24) & 0xff; + a = (clearValue ) & 0xff; + break; + default: + r = g = b = a = 0; + } + + tc->clear_color[0] = r / 255.0; + tc->clear_color[1] = g / 255.0; + tc->clear_color[2] = b / 255.0; + tc->clear_color[3] = a / 255.0; #if TILE_CLEAR_OPTIMIZATION /* set flags to indicate all the tiles are cleared */ diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index c4f2e0f184..91fb2795a2 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -51,9 +51,11 @@ struct softpipe_cached_tile int z, face, level; /**< Extra texture indexes */ union { float color[TILE_SIZE][TILE_SIZE][4]; + uint color32[TILE_SIZE][TILE_SIZE]; uint depth32[TILE_SIZE][TILE_SIZE]; ushort depth16[TILE_SIZE][TILE_SIZE]; ubyte stencil8[TILE_SIZE][TILE_SIZE]; + ubyte any[1]; } data; }; @@ -80,7 +82,7 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc); extern void -sp_tile_cache_clear(struct softpipe_tile_cache *tc, const float value[4]); +sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue); extern struct softpipe_cached_tile * sp_get_cached_tile(struct softpipe_context *softpipe, -- cgit v1.2.3 From d1d9156831d390eaffefbceb8b56916b21fb6dbc Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 20:24:24 -0700 Subject: clean-ups --- src/mesa/pipe/softpipe/sp_surface.c | 20 ++++++-------------- 1 file changed, 6 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index cdbffe5400..55b8b85ef2 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -30,10 +30,9 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "sp_context.h" -#include "sp_state.h" #include "sp_surface.h" #include "sp_texture.h" -#include "sp_tile_cache.h" + /** * Softpipe surface functions. @@ -48,14 +47,6 @@ -#if 0 -#define CLIP_TILE \ - do { \ - assert(x + w <= ps->width); \ - assert(y + h <= ps->height); \ - } while(0) - -#else #define CLIP_TILE \ do { \ if (x >= ps->width) \ @@ -65,9 +56,8 @@ if (x + w > ps->width) \ w = ps->width - x; \ if (y + h > ps->height) \ - h = ps->height -y; \ + h = ps->height - y; \ } while(0) -#endif /*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ @@ -611,6 +601,7 @@ softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, { const uint cpp = ps->cpp; const ubyte *pSrc; + const uint src_stride = ps->pitch * cpp; ubyte *pDest; uint i; @@ -628,7 +619,7 @@ softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w * cpp); pDest += dst_stride; - pSrc += ps->pitch * cpp; + pSrc += src_stride; } } @@ -643,6 +634,7 @@ softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, { const uint cpp = ps->cpp; const ubyte *pSrc; + const uint dst_stride = ps->pitch * cpp; ubyte *pDest; uint i; @@ -659,7 +651,7 @@ softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w * cpp); - pDest += ps->pitch * cpp; + pDest += dst_stride; pSrc += src_stride; } } -- cgit v1.2.3 From 5d4b53b3afcb6014200faad8712bbb13af30469e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 20:31:26 -0700 Subject: code re-org in softpipe_clear() --- src/mesa/pipe/softpipe/sp_clear.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index bfe4a29144..f32e834ac0 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -47,19 +47,28 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct softpipe_context *softpipe = softpipe_context(pipe); + uint i; +#if 0 softpipe_update_derived(softpipe); /* not needed?? */ +#endif +#if TILE_CLEAR_OPTIMIZATION if (ps == sp_tile_cache_get_surface(softpipe->zbuf_cache)) { sp_tile_cache_clear(softpipe->zbuf_cache, clearValue); + return; } - else if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[0])) { - sp_tile_cache_clear(softpipe->cbuf_cache[0], clearValue); + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[i])) { + sp_tile_cache_clear(softpipe->cbuf_cache[i], clearValue); + return; + } } +#endif -#if !TILE_CLEAR_OPTIMIZATION + /* non-cached surface */ pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); -#endif #if 0 sp_clear_tile_cache(ps, clearValue); -- cgit v1.2.3 From a110789f3ecc883fbd0f39758c40a6adc5118aff Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 20:58:01 -0700 Subject: shorten loops over color bufs --- src/mesa/pipe/softpipe/sp_clear.c | 2 +- src/mesa/pipe/softpipe/sp_context.c | 2 +- src/mesa/pipe/softpipe/sp_flush.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index f32e834ac0..5b3857145d 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -59,7 +59,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, return; } - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[i])) { sp_tile_cache_clear(softpipe->cbuf_cache[i], clearValue); return; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index bdfd6228ef..df4e0cbd5a 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -94,7 +94,7 @@ softpipe_unmap_surfaces(struct softpipe_context *sp) struct pipe_surface *ps; uint i; - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + for (i = 0; i < sp->framebuffer.num_cbufs; i++) sp_flush_tile_cache(sp, sp->cbuf_cache[i]); sp_flush_tile_cache(sp, sp->zbuf_cache); sp_flush_tile_cache(sp, sp->sbuf_cache); diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index 1010924bf6..47b11803ce 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -55,7 +55,7 @@ softpipe_flush( struct pipe_context *pipe, * - flush the render cache */ - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) if (softpipe->cbuf_cache[i]) sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]); -- cgit v1.2.3 From f83d4e7bde28d6f73a0de96781da506ddb338714 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 7 Dec 2007 21:21:58 -0700 Subject: Try to reduce the frequency of calls to pipe->get_tex_surface() Save the surface info in the tile cache and re-use whenever possible. --- src/mesa/pipe/softpipe/sp_tile_cache.c | 41 ++++++++++++++++++++++++++-------- 1 file changed, 32 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 13774e6c8f..fadd169f5d 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -55,6 +55,9 @@ struct softpipe_tile_cache float clear_color[4]; uint clear_val; boolean depth_stencil; /** Is the surface a depth/stencil format? */ + + struct pipe_surface *tex_surf; + int tex_face, tex_level, tex_z; }; @@ -179,11 +182,17 @@ sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, tc->texture = texture; + if (tc->tex_surf && tc->tex_surf->map) + pipe_surface_unmap(tc->tex_surf); + pipe_surface_reference(&tc->tex_surf, NULL); + /* mark as entries as invalid/empty */ /* XXX we should try to avoid this when the teximage hasn't changed */ for (i = 0; i < NUM_ENTRIES; i++) { tc->entries[i].x = -1; } + + tc->tex_face = -1; /* any invalid value here */ } @@ -448,19 +457,33 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, z != tile->z || face != tile->face || level != tile->level) { - /* XXX this call is a bit heavier than we'd like: */ - struct pipe_surface *ps - = pipe->get_tex_surface(pipe, tc->texture, face, level, z); + /* cache miss */ - pipe_surface_map(ps); + /* check if we need to get a new surface */ + if (!tc->tex_surf || + tc->tex_face != face || + tc->tex_level != level || + tc->tex_z != z) { + /* get new surface (view into texture) */ + struct pipe_surface *ps; - pipe->get_tile_rgba(pipe, ps, - tile_x, tile_y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + if (tc->tex_surf && tc->tex_surf->map) + pipe_surface_unmap(tc->tex_surf); - pipe_surface_unmap(ps); - pipe_surface_reference(&ps, NULL); + ps = pipe->get_tex_surface(pipe, tc->texture, face, level, z); + pipe_surface_reference(&tc->tex_surf, ps); + pipe_surface_map(ps); + + tc->tex_face = face; + tc->tex_level = level; + tc->tex_z = z; + } + + /* get tile from the surface (view into texture) */ + pipe->get_tile_rgba(pipe, tc->tex_surf, + tile_x, tile_y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); tile->x = tile_x; tile->y = tile_y; tile->z = z; -- cgit v1.2.3 From cd1eefee8404ae69ea5b604971b8be78abf588e6 Mon Sep 17 00:00:00 2001 From: keithw Date: Mon, 19 Nov 2007 20:05:38 +0000 Subject: add fence interfaces and buffer create flags to pipe_winsys --- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 2 +- src/mesa/pipe/p_defines.h | 8 +++-- src/mesa/pipe/p_winsys.h | 44 ++++++++++++++++++++-------- src/mesa/state_tracker/st_atom_constbuf.c | 2 +- src/mesa/state_tracker/st_cb_bufferobjects.c | 2 +- src/mesa/state_tracker/st_draw.c | 4 +-- 6 files changed, 42 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 08ac5b672c..571ad40595 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -99,7 +99,7 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, /* FIXME: handle failure */ assert(!i915->vbo); - i915->vbo = winsys->buffer_create(winsys, 64); + i915->vbo = winsys->buffer_create(winsys, 64, 0, 0); winsys->buffer_data( winsys, i915->vbo, size, NULL, I915_BUFFER_USAGE_LIT_VERTEX ); diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index a853605486..8dce3aba90 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -170,8 +170,12 @@ /** * Buffer access flags */ -#define PIPE_BUFFER_FLAG_READ 0x1 -#define PIPE_BUFFER_FLAG_WRITE 0x2 +#define PIPE_BUFFER_FLAG_READ 0x1 +#define PIPE_BUFFER_FLAG_WRITE 0x2 +#define PIPE_BUFFER_FLAG_MEM_LOCAL 0x4 +#define PIPE_BUFFER_FLAG_CACHED 0x8 +#define PIPE_BUFFER_FLAG_CUSTOM (1<<16) + /** diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 438e8bdb63..f46807995d 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -59,6 +59,7 @@ struct pipe_surface; /** Opaque type */ struct pipe_buffer_handle; +struct pipe_fence_handle; struct pipe_winsys { @@ -104,8 +105,10 @@ struct pipe_winsys * systems must then implement that interface (rather than the * other way around...). */ - struct pipe_buffer_handle *(*buffer_create)(struct pipe_winsys *sws, - unsigned alignment ); + struct pipe_buffer_handle *(*buffer_create)( struct pipe_winsys *sws, + unsigned alignment, + unsigned flags, + unsigned hint ); /** Create a buffer that wraps user-space data */ struct pipe_buffer_handle *(*user_buffer_create)(struct pipe_winsys *sws, @@ -136,24 +139,39 @@ struct pipe_winsys * usage argument is only an optimization hint, not a guarantee, therefore * proper behavior must be observed in all circumstances. */ - void (*buffer_data)(struct pipe_winsys *sws, + int (*buffer_data)(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, unsigned size, const void *data, unsigned usage); /** Modify some or all of the data contained in a buffer's data store */ - void (*buffer_subdata)(struct pipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data); + int (*buffer_subdata)(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data); /** Query some or all of the data contained in a buffer's data store */ - void (*buffer_get_subdata)(struct pipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data); + int (*buffer_get_subdata)(struct pipe_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data); + + + void (*fence_reference)( struct pipe_winsys *sws, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence ); + + int (*fence_signalled)( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ); + + + int (*fence_finish)( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ); + }; diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index 446250c226..57f5ec68d2 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -70,7 +70,7 @@ void st_upload_constants( struct st_context *st, _mesa_load_state_parameters(st->ctx, params); if (!cbuf->buffer) - cbuf->buffer = ws->buffer_create(ws, 1); + cbuf->buffer = ws->buffer_create(ws, 1, 0, 0); if (0) { diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index 99e1eb3c7a..872248cdb5 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -63,7 +63,7 @@ st_bufferobj_alloc(GLcontext *ctx, GLuint name, GLenum target) _mesa_initialize_buffer_object(&st_obj->Base, name, target); - st_obj->buffer = st->pipe->winsys->buffer_create( st->pipe->winsys, 32 ); + st_obj->buffer = st->pipe->winsys->buffer_create( st->pipe->winsys, 32, 0, 0 ); return &st_obj->Base; } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 32dcd73c46..0a5725ae6a 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -95,7 +95,7 @@ static void create_default_attribs_buffer(struct st_context *st) { struct pipe_context *pipe = st->pipe; - st->default_attrib_buffer = pipe->winsys->buffer_create( pipe->winsys, 32 ); + st->default_attrib_buffer = pipe->winsys->buffer_create( pipe->winsys, 32, 0, 0 ); } @@ -288,7 +288,7 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, } /* XXX create one-time */ - vbuf = pipe->winsys->buffer_create(pipe->winsys, 32); + vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); pipe->winsys->buffer_data(pipe->winsys, vbuf, vertex_bytes, verts, PIPE_BUFFER_USAGE_VERTEX); -- cgit v1.2.3 From 79ad82477adab2c234370623c48c35316e94ad02 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 20 Nov 2007 14:27:52 +0000 Subject: Document return types. --- src/mesa/pipe/p_winsys.h | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index f46807995d..2aac403c20 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -138,20 +138,30 @@ struct pipe_winsys * usage is a bitmask of PIPE_BUFFER_USAGE_PIXEL/VERTEX/INDEX/CONSTANT. This * usage argument is only an optimization hint, not a guarantee, therefore * proper behavior must be observed in all circumstances. + * + * Returns zero on success. */ int (*buffer_data)(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, unsigned size, const void *data, unsigned usage); - /** Modify some or all of the data contained in a buffer's data store */ + /** + * Modify some or all of the data contained in a buffer's data store. + * + * Returns zero on success. + */ int (*buffer_subdata)(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, const void *data); - /** Query some or all of the data contained in a buffer's data store */ + /** + * Query some or all of the data contained in a buffer's data store. + * + * Returns zero on success. + */ int (*buffer_get_subdata)(struct pipe_winsys *sws, struct pipe_buffer_handle *buf, unsigned long offset, -- cgit v1.2.3 From e87efe74073938e6dea5f30654f6afb931da1e2c Mon Sep 17 00:00:00 2001 From: keithw Date: Wed, 21 Nov 2007 11:04:20 +0000 Subject: [pipe]: add Offset() util macro --- src/mesa/pipe/p_util.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 7897bc90e4..4002e2659f 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -114,6 +114,7 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) #define Elements(x) sizeof(x)/sizeof(*(x)) +#define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER)) /** * Return a pointer aligned to next multiple of 16 bytes. -- cgit v1.2.3 From 6e876d1d32e1b08ded161bfbd7abf2d3fb422fa0 Mon Sep 17 00:00:00 2001 From: keithw Date: Wed, 21 Nov 2007 17:12:39 +0000 Subject: temporary _mesa_printf symbol --- src/mesa/pipe/p_util.h | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 4002e2659f..16e57699aa 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -301,4 +301,8 @@ static INLINE float LOG2(float val) #define CEILF(x) ((float) ceil(x)) #endif +/* Convenient... + */ +extern void _mesa_printf(const char *str, ...); + #endif -- cgit v1.2.3 From 74fe189b2e1d8f661a1678e65d21da788b3b4435 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 23 Nov 2007 10:35:53 +0000 Subject: gallium: Comment on winsys's fence functions. Fix comment layout for doxygen. --- src/mesa/pipe/p_winsys.h | 34 +++++++++++++++++++++++----------- 1 file changed, 23 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 2aac403c20..f2be7655f4 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -42,6 +42,13 @@ /** Opaque type for a buffer */ struct pipe_buffer_handle; +/** Opaque type */ +struct pipe_fence_handle; + +struct pipe_region; +struct pipe_surface; + + /** * Gallium3D drivers are (meant to be!) independent of both GL and the * window system. The window system provides a buffer manager and a @@ -52,15 +59,6 @@ struct pipe_buffer_handle; * driver and the hardware driver about the format of command buffers, * etc. */ - - -struct pipe_region; -struct pipe_surface; - -/** Opaque type */ -struct pipe_buffer_handle; -struct pipe_fence_handle; - struct pipe_winsys { /** Returns name of this winsys interface */ @@ -97,6 +95,7 @@ struct pipe_winsys void (*surface_release)(struct pipe_winsys *ws, struct pipe_surface **s); + /** * The buffer manager is modeled after the dri_bufmgr interface, which * in turn is modeled after the ARB_vertex_buffer_object extension, @@ -115,7 +114,6 @@ struct pipe_winsys void *ptr, unsigned bytes); - /** * Map the entire data store of a buffer object into the client's address. * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. @@ -169,15 +167,29 @@ struct pipe_winsys void *data); + /** Set ptr = buf, with reference counting */ void (*fence_reference)( struct pipe_winsys *sws, struct pipe_fence_handle **ptr, struct pipe_fence_handle *fence ); + /** + * Checks whether the fence has been signalled. + * + * The meaning of flag is pipe-driver specific. + * + * Returns zero if it has. + */ int (*fence_signalled)( struct pipe_winsys *sws, struct pipe_fence_handle *fence, unsigned flag ); - + /** + * Wait for the fence to finish. + * + * The meaning of flag is pipe-driver specific. + * + * Returns zero on success. + */ int (*fence_finish)( struct pipe_winsys *sws, struct pipe_fence_handle *fence, unsigned flag ); -- cgit v1.2.3 From d75454840672f462de933724daae24a839aac48e Mon Sep 17 00:00:00 2001 From: Michal Date: Fri, 23 Nov 2007 11:30:51 +0000 Subject: gallium: add draw_stage::destroy(). --- src/mesa/pipe/draw/draw_clip.c | 8 ++++++++ src/mesa/pipe/draw/draw_context.c | 10 ++++++++++ src/mesa/pipe/draw/draw_cull.c | 9 +++++++++ src/mesa/pipe/draw/draw_feedback.c | 7 +++++++ src/mesa/pipe/draw/draw_flatshade.c | 8 ++++++++ src/mesa/pipe/draw/draw_linestipple.c | 9 ++++++++- src/mesa/pipe/draw/draw_offset.c | 8 ++++++++ src/mesa/pipe/draw/draw_private.h | 2 ++ src/mesa/pipe/draw/draw_twoside.c | 8 ++++++++ src/mesa/pipe/draw/draw_unfilled.c | 8 ++++++++ src/mesa/pipe/draw/draw_validate.c | 5 +++++ src/mesa/pipe/draw/draw_vbuf.c | 11 ++++++++++- src/mesa/pipe/draw/draw_wide_prims.c | 10 +++++++++- src/mesa/pipe/i915simple/i915_prim_emit.c | 6 ++++++ src/mesa/pipe/softpipe/sp_prim_setup.c | 7 +++++++ src/mesa/pipe/softpipe/sp_prim_vbuf.c | 11 +++++++++++ src/mesa/pipe/tgsi/exec/tgsi_exec.c | 8 ++++---- 17 files changed, 128 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index e4c257a0ee..c50376f11f 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -415,6 +415,13 @@ static void clip_reset_stipple_counter( struct draw_stage *stage ) } +static void clip_destroy( struct draw_stage *stage ) +{ + draw_free_tmps( stage ); + FREE( stage ); +} + + /** * Allocate a new clipper stage. * \return pointer to new stage object @@ -432,6 +439,7 @@ struct draw_stage *draw_clip_stage( struct draw_context *draw ) clipper->stage.tri = clip_tri; clipper->stage.end = clip_end; clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; + clipper->stage.destroy = clip_destroy; clipper->plane = draw->plane; diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index e7997180b5..c7ad4c3d68 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -92,6 +92,16 @@ struct draw_context *draw_create( void ) void draw_destroy( struct draw_context *draw ) { + draw->pipeline.wide->destroy( draw->pipeline.wide ); + draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); + draw->pipeline.twoside->destroy( draw->pipeline.twoside ); + draw->pipeline.offset->destroy( draw->pipeline.offset ); + draw->pipeline.clip->destroy( draw->pipeline.clip ); + draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); + draw->pipeline.cull->destroy( draw->pipeline.cull ); + draw->pipeline.feedback->destroy( draw->pipeline.feedback ); + draw->pipeline.validate->destroy( draw->pipeline.validate ); + draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); FREE( draw->vcache.vertex[0] ); /* Frees all the vertices. */ FREE( draw ); } diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index f898834ba5..9bd53f45f2 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -116,6 +116,14 @@ static void cull_reset_stipple_counter( struct draw_stage *stage ) stage->next->reset_stipple_counter( stage->next ); } + +static void cull_destroy( struct draw_stage *stage ) +{ + draw_free_tmps( stage ); + FREE( stage ); +} + + /** * Create a new polygon culling stage. */ @@ -133,6 +141,7 @@ struct draw_stage *draw_cull_stage( struct draw_context *draw ) cull->stage.tri = cull_tri; cull->stage.end = cull_end; cull->stage.reset_stipple_counter = cull_reset_stipple_counter; + cull->stage.destroy = cull_destroy; return &cull->stage; } diff --git a/src/mesa/pipe/draw/draw_feedback.c b/src/mesa/pipe/draw/draw_feedback.c index b9906e5b4b..aea6a8184c 100644 --- a/src/mesa/pipe/draw/draw_feedback.c +++ b/src/mesa/pipe/draw/draw_feedback.c @@ -224,6 +224,12 @@ static void feedback_reset_stipple_counter( struct draw_stage *stage ) } +static void feedback_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + /** * Create feedback drawing stage. */ @@ -239,6 +245,7 @@ struct draw_stage *draw_feedback_stage( struct draw_context *draw ) feedback->stage.tri = feedback_tri; feedback->stage.end = feedback_end; feedback->stage.reset_stipple_counter = feedback_reset_stipple_counter; + feedback->stage.destroy = feedback_destroy; return &feedback->stage; } diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index d46e53f2be..d7551e7948 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -127,6 +127,13 @@ static void flatshade_reset_stipple_counter( struct draw_stage *stage ) } +static void flatshade_destroy( struct draw_stage *stage ) +{ + draw_free_tmps( stage ); + FREE( stage ); +} + + /** * Create flatshading drawing stage. */ @@ -144,6 +151,7 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) flatshade->tri = flatshade_tri; flatshade->end = flatshade_end; flatshade->reset_stipple_counter = flatshade_reset_stipple_counter; + flatshade->destroy = flatshade_destroy; return flatshade; } diff --git a/src/mesa/pipe/draw/draw_linestipple.c b/src/mesa/pipe/draw/draw_linestipple.c index 1fac1ebe66..5f0db99b23 100644 --- a/src/mesa/pipe/draw/draw_linestipple.c +++ b/src/mesa/pipe/draw/draw_linestipple.c @@ -241,12 +241,18 @@ static void stipple_begin( struct clip_pipe_stage *stage ) } - static void stipple_end( struct clip_pipe_stage *stage ) { stage->next->end( stage->next ); } + +static void stipple_destroy( struct clip_pipe_stage *stage ) +{ + FREE( stage ); +} + + struct clip_pipe_stage *clip_pipe_stipple( struct clip_pipeline *pipe ) { struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); @@ -261,6 +267,7 @@ struct clip_pipe_stage *clip_pipe_stipple( struct clip_pipeline *pipe ) stipple->stage.tri = clip_passthrough_tri; stipple->stage.reset_tmps = clip_pipe_reset_tmps; stipple->stage.end = stipple_end; + stipple->stage.destroy = stipple_destroy; return &stipple->stage; } diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index 6acc7cbcd2..f8a01db3dd 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -151,6 +151,13 @@ static void offset_reset_stipple_counter( struct draw_stage *stage ) } +static void offset_destroy( struct draw_stage *stage ) +{ + draw_free_tmps( stage ); + FREE( stage ); +} + + /** * Create polygon offset drawing stage. */ @@ -168,6 +175,7 @@ struct draw_stage *draw_offset_stage( struct draw_context *draw ) offset->stage.tri = offset_tri; offset->stage.end = offset_end; offset->stage.reset_stipple_counter = offset_reset_stipple_counter; + offset->stage.destroy = offset_destroy; return &offset->stage; } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 53d7451113..ca5ca7b3c9 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -124,6 +124,8 @@ struct draw_stage void (*reset_tmps)( struct draw_stage * ); void (*reset_stipple_counter)( struct draw_stage * ); + + void (*destroy)( struct draw_stage * ); }; diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index d987b00598..00b4ee45cd 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -146,6 +146,13 @@ static void twoside_reset_stipple_counter( struct draw_stage *stage ) } +static void twoside_destroy( struct draw_stage *stage ) +{ + draw_free_tmps( stage ); + FREE( stage ); +} + + /** * Create twoside pipeline stage. */ @@ -163,6 +170,7 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) twoside->stage.tri = twoside_tri; twoside->stage.end = twoside_end; twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; + twoside->stage.destroy = twoside_destroy; return &twoside->stage; } diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index 6cab086a45..786826b33c 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -168,6 +168,13 @@ static void unfilled_reset_stipple_counter( struct draw_stage *stage ) } +static void unfilled_destroy( struct draw_stage *stage ) +{ + draw_free_tmps( stage ); + FREE( stage ); +} + + /** * Create unfilled triangle stage. */ @@ -186,6 +193,7 @@ struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) unfilled->stage.tri = unfilled_tri; unfilled->stage.end = unfilled_end; unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; + unfilled->stage.destroy = unfilled_destroy; return &unfilled->stage; } diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index 4e8f986b27..8ce4a926e2 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -110,6 +110,10 @@ static void validate_begin( struct draw_stage *stage ) } +static void validate_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} /** @@ -127,6 +131,7 @@ struct draw_stage *draw_validate_stage( struct draw_context *draw ) stage->tri = NULL; stage->end = NULL; stage->reset_stipple_counter = NULL; + stage->destroy = validate_destroy; return stage; } diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index d00cdec56c..d010aaba07 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -369,6 +369,15 @@ vbuf_reset_stipple_counter( struct draw_stage *stage ) } +static void vbuf_destroy( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + FREE( vbuf->indices ); + FREE( stage ); +} + + /** * Create a new primitive vbuf/render stage. */ @@ -384,12 +393,12 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, vbuf->stage.tri = vbuf_first_tri; vbuf->stage.end = vbuf_end; vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter; + vbuf->stage.destroy = vbuf_destroy; vbuf->render = render; assert(render->max_indices < UNDEFINED_VERTEX_ID); vbuf->max_indices = render->max_indices; - /* FIXME: free this memory on takedown */ vbuf->indices = MALLOC( vbuf->max_indices ); vbuf->vertices = NULL; diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 494a2bc619..a56c9b8893 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -315,7 +315,6 @@ static void wide_begin( struct draw_stage *stage ) } - static void wide_end( struct draw_stage *stage ) { stage->next->end( stage->next ); @@ -327,6 +326,14 @@ static void draw_reset_stipple_counter( struct draw_stage *stage ) stage->next->reset_stipple_counter( stage->next ); } + +static void wide_destroy( struct draw_stage *stage ) +{ + draw_free_tmps( stage ); + FREE( stage ); +} + + struct draw_stage *draw_wide_stage( struct draw_context *draw ) { struct wide_stage *wide = CALLOC_STRUCT(wide_stage); @@ -341,6 +348,7 @@ struct draw_stage *draw_wide_stage( struct draw_context *draw ) wide->stage.tri = passthrough_tri; wide->stage.end = wide_end; wide->stage.reset_stipple_counter = draw_reset_stipple_counter; + wide->stage.destroy = wide_destroy; return &wide->stage; } diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index 0a8be79ae4..c50a591589 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -194,6 +194,11 @@ static void reset_stipple_counter( struct draw_stage *stage ) { } +static void render_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + /** * Create a new primitive setup/render stage. This gets plugged into @@ -211,6 +216,7 @@ struct draw_stage *i915_draw_render_stage( struct i915_context *i915 ) setup->stage.tri = setup_tri; setup->stage.end = setup_end; setup->stage.reset_stipple_counter = reset_stipple_counter; + setup->stage.destroy = render_destroy; return &setup->stage; } diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 9f5a43a827..6980564bdc 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -1124,6 +1124,12 @@ static void reset_stipple_counter( struct draw_stage *stage ) } +static void render_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + /** * Create a new primitive setup/render stage. */ @@ -1139,6 +1145,7 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) setup->stage.tri = setup_tri; setup->stage.end = setup_end; setup->stage.reset_stipple_counter = reset_stipple_counter; + setup->stage.destroy = render_destroy; setup->quad.coef = setup->coef; diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index 59cf1c4eb7..2cfdeb5809 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -275,6 +275,16 @@ static void reset_stipple_counter( struct draw_stage *stage ) } +static void vbuf_destroy( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + FREE( vbuf->element_map ); + FREE( vbuf->vertex_map ); + FREE( stage ); +} + + /** * Create a new primitive vbuf/render stage. */ @@ -290,6 +300,7 @@ struct draw_stage *sp_draw_vbuf_stage( struct draw_context *draw_context, vbuf->stage.tri = vbuf_first_tri; vbuf->stage.end = vbuf_end; vbuf->stage.reset_stipple_counter = reset_stipple_counter; + vbuf->stage.destroy = vbuf_destroy; vbuf->pipe = pipe; vbuf->draw = draw; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index dd11dd58b7..d036e9d383 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1808,7 +1808,7 @@ exec_instruction( micro_sub( &r[2], &r[2], &r[5] ); if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &r[2], 0, CHAN_X ); + STORE( &r[2], 0, CHAN_X ); } FETCH(&r[2], 1, CHAN_X); @@ -1821,7 +1821,7 @@ exec_instruction( micro_sub( &r[3], &r[3], &r[1] ); if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - STORE( &r[3], 0, CHAN_Y ); + STORE( &r[3], 0, CHAN_Y ); } micro_mul( &r[5], &r[5], &r[4] ); @@ -1829,11 +1829,11 @@ exec_instruction( micro_sub( &r[5], &r[5], &r[0] ); if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - STORE( &r[5], 0, CHAN_Z ); + STORE( &r[5], 0, CHAN_Z ); } if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); } break; -- cgit v1.2.3 From 73a05942766cee4e3cc200725e9760c5a9b67bb4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 Nov 2007 12:02:50 +0000 Subject: gallium: support the full range of possible vertex types --- src/mesa/state_tracker/st_draw.c | 152 +++++++++++++++++++++++++++++++++------ 1 file changed, 129 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 0a5725ae6a..c6fea1f7bc 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -50,37 +50,141 @@ #include "pipe/draw/draw_context.h" +static GLuint double_types[4] = { + PIPE_FORMAT_R64_FLOAT, + PIPE_FORMAT_R64G64_FLOAT, + PIPE_FORMAT_R64G64B64_FLOAT, + PIPE_FORMAT_R64G64B64A64_FLOAT +}; + +static GLuint float_types[4] = { + PIPE_FORMAT_R32_FLOAT, + PIPE_FORMAT_R32G32_FLOAT, + PIPE_FORMAT_R32G32B32_FLOAT, + PIPE_FORMAT_R32G32B32A32_FLOAT +}; + +static GLuint uint_types_norm[4] = { + PIPE_FORMAT_R32_UNORM, + PIPE_FORMAT_R32G32_UNORM, + PIPE_FORMAT_R32G32B32_UNORM, + PIPE_FORMAT_R32G32B32A32_UNORM +}; + +static GLuint uint_types_scale[4] = { + PIPE_FORMAT_R32_USCALED, + PIPE_FORMAT_R32G32_USCALED, + PIPE_FORMAT_R32G32B32_USCALED, + PIPE_FORMAT_R32G32B32A32_USCALED +}; + +static GLuint int_types_norm[4] = { + PIPE_FORMAT_R32_SNORM, + PIPE_FORMAT_R32G32_SNORM, + PIPE_FORMAT_R32G32B32_SNORM, + PIPE_FORMAT_R32G32B32A32_SNORM +}; + +static GLuint int_types_scale[4] = { + PIPE_FORMAT_R32_SSCALED, + PIPE_FORMAT_R32G32_SSCALED, + PIPE_FORMAT_R32G32B32_SSCALED, + PIPE_FORMAT_R32G32B32A32_SSCALED +}; + +static GLuint ushort_types_norm[4] = { + PIPE_FORMAT_R16_UNORM, + PIPE_FORMAT_R16G16_UNORM, + PIPE_FORMAT_R16G16B16_UNORM, + PIPE_FORMAT_R16G16B16A16_UNORM +}; + +static GLuint ushort_types_scale[4] = { + PIPE_FORMAT_R16_USCALED, + PIPE_FORMAT_R16G16_USCALED, + PIPE_FORMAT_R16G16B16_USCALED, + PIPE_FORMAT_R16G16B16A16_USCALED +}; + +static GLuint short_types_norm[4] = { + PIPE_FORMAT_R16_SNORM, + PIPE_FORMAT_R16G16_SNORM, + PIPE_FORMAT_R16G16B16_SNORM, + PIPE_FORMAT_R16G16B16A16_SNORM +}; + +static GLuint short_types_scale[4] = { + PIPE_FORMAT_R16_SSCALED, + PIPE_FORMAT_R16G16_SSCALED, + PIPE_FORMAT_R16G16B16_SSCALED, + PIPE_FORMAT_R16G16B16A16_SSCALED +}; + +static GLuint ubyte_types_norm[4] = { + PIPE_FORMAT_R8_UNORM, + PIPE_FORMAT_R8G8_UNORM, + PIPE_FORMAT_R8G8B8_UNORM, + PIPE_FORMAT_R8G8B8A8_UNORM +}; + +static GLuint ubyte_types_scale[4] = { + PIPE_FORMAT_R8_USCALED, + PIPE_FORMAT_R8G8_USCALED, + PIPE_FORMAT_R8G8B8_USCALED, + PIPE_FORMAT_R8G8B8A8_USCALED +}; + +static GLuint byte_types_norm[4] = { + PIPE_FORMAT_R8_SNORM, + PIPE_FORMAT_R8G8_SNORM, + PIPE_FORMAT_R8G8B8_SNORM, + PIPE_FORMAT_R8G8B8A8_SNORM +}; + +static GLuint byte_types_scale[4] = { + PIPE_FORMAT_R8_SSCALED, + PIPE_FORMAT_R8G8_SSCALED, + PIPE_FORMAT_R8G8B8_SSCALED, + PIPE_FORMAT_R8G8B8A8_SSCALED +}; + + /** * Return a PIPE_FORMAT_x for the given GL datatype and size. */ static GLuint -pipe_vertex_format(GLenum type, GLuint size) +pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized) { - static const GLuint float_fmts[4] = { - PIPE_FORMAT_R32_FLOAT, - PIPE_FORMAT_R32G32_FLOAT, - PIPE_FORMAT_R32G32B32_FLOAT, - PIPE_FORMAT_R32G32B32A32_FLOAT, - }; - static const GLuint int_fmts[4] = { - PIPE_FORMAT_R32_SSCALED, - PIPE_FORMAT_R32G32_SSCALED, - PIPE_FORMAT_R32G32B32_SSCALED, - PIPE_FORMAT_R32G32B32A32_SSCALED, - }; - assert(type >= GL_BYTE); assert(type <= GL_DOUBLE); assert(size >= 1); assert(size <= 4); - switch (type) { - case GL_FLOAT: - return float_fmts[size - 1]; - case GL_INT: - return int_fmts[size - 1]; - default: - assert(0); + if (normalized) { + switch (type) { + case GL_DOUBLE: return double_types[size-1]; + case GL_FLOAT: return float_types[size-1]; + case GL_INT: return int_types_norm[size-1]; + case GL_SHORT: return short_types_norm[size-1]; + case GL_BYTE: return byte_types_norm[size-1]; + case GL_UNSIGNED_INT: return uint_types_norm[size-1]; + case GL_UNSIGNED_SHORT: return ushort_types_norm[size-1]; + case GL_UNSIGNED_BYTE: return ubyte_types_norm[size-1]; + default: assert(0); return 0; + } + } + else { + switch (type) { + case GL_DOUBLE: return double_types[size-1]; + case GL_FLOAT: return float_types[size-1]; + case GL_INT: return int_types_scale[size-1]; + case GL_SHORT: return short_types_scale[size-1]; + case GL_BYTE: return byte_types_scale[size-1]; + case GL_UNSIGNED_INT: return uint_types_scale[size-1]; + case GL_UNSIGNED_SHORT: return ushort_types_scale[size-1]; + case GL_UNSIGNED_BYTE: return ubyte_types_scale[size-1]; + default: assert(0); return 0; + } } return 0; /* silence compiler warning */ } @@ -181,7 +285,8 @@ st_draw_vbo(GLcontext *ctx, velement.vertex_buffer_index = attr; velement.dst_offset = 0; /* need this? */ velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, - arrays[mesaAttr]->Size); + arrays[mesaAttr]->Size, + arrays[mesaAttr]->Normalized); assert(velement.src_format); /* tell pipe about this attribute */ @@ -438,7 +543,8 @@ st_feedback_draw_vbo(GLcontext *ctx, velement.vertex_buffer_index = attr; velement.dst_offset = 0; /* need this? */ velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, - arrays[mesaAttr]->Size); + arrays[mesaAttr]->Size, + arrays[mesaAttr]->Normalized); assert(velement.src_format); /* tell draw about this attribute */ -- cgit v1.2.3 From da47ad07ef8cf69a74f4414f7957e0749fa6f7ad Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 Nov 2007 12:03:45 +0000 Subject: gallium: add support for aliased 4ub format --- src/mesa/pipe/draw/draw_vertex_fetch.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index e0759c2e9a..b51cff59f5 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -80,6 +80,8 @@ fetch_attrib4(const void *ptr, enum pipe_format format, float attrib[4]) break; case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_R8G8B8A8_UNORM: attrib[0] = (float) ((unsigned char *) ptr)[2] / 255.0f; attrib[1] = (float) ((unsigned char *) ptr)[1] / 255.0f; attrib[2] = (float) ((unsigned char *) ptr)[0] / 255.0f; -- cgit v1.2.3 From bbecce52f8f9db729810ed70ba1c4112d45cc583 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 Nov 2007 14:44:54 +0000 Subject: gallium: initialized draw element buffer to NULL --- src/mesa/pipe/draw/draw_context.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index c7ad4c3d68..d438c0be10 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -85,6 +85,7 @@ struct draw_context *draw_create( void ) draw->prim = ~0; /* != any of PIPE_PRIM_x */ draw_vertex_cache_invalidate( draw ); + draw_set_mapped_element_buffer( draw, 0, NULL ); return draw; } -- cgit v1.2.3 From 708582f6d5bec57716a894a18491078c40448dea Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 Nov 2007 14:50:26 +0000 Subject: gallium: fix ordering of arguments to pack_ui32_float4 --- src/mesa/pipe/p_util.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 16e57699aa..3c5e98453c 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -205,8 +205,8 @@ static INLINE unsigned char float_to_ubyte( float f ) static INLINE unsigned pack_ui32_float4( float a, float b, - float d, - float c ) + float c, + float d ) { return pack_ub4( float_to_ubyte(a), float_to_ubyte(b), -- cgit v1.2.3 From b47836225185c8d2e2ccf3e191230304709d0a57 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 23 Nov 2007 17:22:54 +0000 Subject: gallium: Reusable pipe buffer library. This is an optional library that winsys drivers can use to simplify memory management, and help meet the winsys interface requirements. --- src/mesa/pipe/Makefile | 5 +- src/mesa/pipe/pipebuffer/Makefile | 27 +++ src/mesa/pipe/pipebuffer/linked_list.h | 91 ++++++++ src/mesa/pipe/pipebuffer/pb_buffer.c | 52 +++++ src/mesa/pipe/pipebuffer/pb_buffer.h | 175 ++++++++++++++ src/mesa/pipe/pipebuffer/pb_buffer_client.c | 123 ++++++++++ src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 340 ++++++++++++++++++++++++++++ src/mesa/pipe/pipebuffer/pb_buffer_fenced.h | 138 +++++++++++ src/mesa/pipe/pipebuffer/pb_buffer_handle.c | 182 +++++++++++++++ src/mesa/pipe/pipebuffer/pb_buffer_handle.h | 120 ++++++++++ src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 132 +++++++++++ src/mesa/pipe/pipebuffer/pb_buffer_null.c | 98 ++++++++ src/mesa/pipe/pipebuffer/pb_bufmgr.h | 114 ++++++++++ src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c | 128 +++++++++++ src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 265 ++++++++++++++++++++++ src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 279 +++++++++++++++++++++++ 16 files changed, 2266 insertions(+), 3 deletions(-) create mode 100644 src/mesa/pipe/pipebuffer/Makefile create mode 100644 src/mesa/pipe/pipebuffer/linked_list.h create mode 100644 src/mesa/pipe/pipebuffer/pb_buffer.c create mode 100644 src/mesa/pipe/pipebuffer/pb_buffer.h create mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_client.c create mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_fenced.c create mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_fenced.h create mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_handle.c create mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_handle.h create mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_malloc.c create mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_null.c create mode 100644 src/mesa/pipe/pipebuffer/pb_bufmgr.h create mode 100644 src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c create mode 100644 src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c create mode 100644 src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c (limited to 'src') diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile index a9dc393417..2bf7318d59 100644 --- a/src/mesa/pipe/Makefile +++ b/src/mesa/pipe/Makefile @@ -1,4 +1,3 @@ - TOP = ../../.. include $(TOP)/configs/current @@ -7,7 +6,7 @@ ifeq ($(CONFIG_NAME), linux-cell) CELL_DIR = cell endif -SUBDIRS = softpipe i915simple failover $(CELL_DIR) +SUBDIRS = softpipe i915simple failover pipebuffer $(CELL_DIR) default: subdirs @@ -22,4 +21,4 @@ subdirs: clean: - rm -f `find . -name \*.[oa]` \ No newline at end of file + rm -f `find . -name \*.[oa]` diff --git a/src/mesa/pipe/pipebuffer/Makefile b/src/mesa/pipe/pipebuffer/Makefile new file mode 100644 index 0000000000..061d8a060f --- /dev/null +++ b/src/mesa/pipe/pipebuffer/Makefile @@ -0,0 +1,27 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = pipebuffer + + +DRIVER_SOURCES = \ + pb_buffer.c \ + pb_buffer_client.c \ + pb_buffer_handle.c \ + pb_buffer_fenced.c \ + pb_buffer_malloc.c \ + pb_buffer_null.c \ + pb_bufmgr_fenced.c \ + pb_bufmgr_mm.c \ + pb_bufmgr_pool.c + +C_SOURCES = \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../Makefile.template + +symlinks: + diff --git a/src/mesa/pipe/pipebuffer/linked_list.h b/src/mesa/pipe/pipebuffer/linked_list.h new file mode 100644 index 0000000000..e99817fb13 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/linked_list.h @@ -0,0 +1,91 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND. USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + **************************************************************************/ + +/** + * \file + * List macros heavily inspired by the Linux kernel + * list handling. No list looping yet. + * + * Is not threadsafe, so common operations need to + * be protected using an external mutex. + */ + +#ifndef LINKED_LIST_H_ +#define LINKED_LIST_H_ + + +#include + + +struct list_head +{ + struct list_head *prev; + struct list_head *next; +}; + + +#define LIST_INITHEAD(__item) \ + do { \ + (__item)->prev = (__item); \ + (__item)->next = (__item); \ + } while (0) + +#define LIST_ADD(__item, __list) \ + do { \ + (__item)->prev = (__list); \ + (__item)->next = (__list)->next; \ + (__list)->next->prev = (__item); \ + (__list)->next = (__item); \ + } while (0) + +#define LIST_ADDTAIL(__item, __list) \ + do { \ + (__item)->next = (__list); \ + (__item)->prev = (__list)->prev; \ + (__list)->prev->next = (__item); \ + (__list)->prev = (__item); \ + } while(0) + +#define LIST_DEL(__item) \ + do { \ + (__item)->prev->next = (__item)->next; \ + (__item)->next->prev = (__item)->prev; \ + } while(0) + +#define LIST_DELINIT(__item) \ + do { \ + (__item)->prev->next = (__item)->next; \ + (__item)->next->prev = (__item)->prev; \ + (__item)->next = (__item); \ + (__item)->prev = (__item); \ + } while(0) + +#define LIST_ENTRY(__type, __item, __field) \ + ((__type *)(((char *)(__item)) - offsetof(__type, __field))) + + +#endif /*LINKED_LIST_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.c b/src/mesa/pipe/pipebuffer/pb_buffer.c new file mode 100644 index 0000000000..99c960b697 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_buffer.c @@ -0,0 +1,52 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer implementation. + * + * \author José Fonseca + */ + + +#include "pb_buffer.h" + + +void +buffer_reference(struct pipe_buffer **dst, + struct pipe_buffer *src) +{ + if(*dst != src) { + if (src) + src->vtbl->reference(src); + + if (*dst) + (*dst)->vtbl->release(*dst); + + *dst = src; + } +} diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.h b/src/mesa/pipe/pipebuffer/pb_buffer.h new file mode 100644 index 0000000000..0523531395 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_buffer.h @@ -0,0 +1,175 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Generic code for buffers. + * + * Behind a pipe buffle handle there can be DMA buffers, client (or user) + * buffers, regular malloced buffers, etc. This file provides an abstract base + * buffer handle that allows the driver to cope with all those kinds of buffers + * in a more flexible way. + * + * There is no obligation of a winsys driver to use this library. And a pipe + * driver should be completly agnostic about it. + * + * \author José Fonseca + */ + +#ifndef PB_BUFFER_H_ +#define PB_BUFFER_H_ + + +#include +#include + + +struct pipe_buffer_vtbl; + + +/** + * Base class for all pipe buffers. + */ +struct pipe_buffer +{ + /** + * Pointer to the virtual function table. + * + * Avoid accessing this table directly. Use the inline functions below + * instead to avoid mistakes. + */ + const struct pipe_buffer_vtbl *vtbl; +}; + + +/** + * Virtual function table for the buffer storage operations. + * + * Note that creation is not done through this table. + */ +struct pipe_buffer_vtbl +{ + /** + * Add a reference to the buffer. + * + * This method can be a no-op for buffers that don't need reference + * counting. + */ + void (*reference)( struct pipe_buffer *buf ); + + /** + * Release a reference to this buffer and destroy it. + */ + void (*release)( struct pipe_buffer *buf ); + + /** + * Map the entire data store of a buffer object into the client's address. + * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. + */ + void *(*map)( struct pipe_buffer *buf, + unsigned flags ); + + void (*unmap)( struct pipe_buffer *buf ); + + /** + * Get the base buffer and the offset. + * + * A buffer can be subdivided in smaller buffers. This method should return + * the underlaying buffer, and the relative offset. + * + * Buffers without an underlaying base buffer should return themselves, with + * a zero offset. + * + * Note that this will increase the reference count of the base buffer. + */ + void (*get_base_buffer)( struct pipe_buffer *buf, + struct pipe_buffer **base_buf, + unsigned *offset ); +}; + + +/** *dst = src with reference counting */ +void +buffer_reference(struct pipe_buffer **dst, + struct pipe_buffer *src); + + +static inline void +buffer_release(struct pipe_buffer *buf) +{ + assert(buf); + buf->vtbl->release(buf); +} + + +static inline void * +buffer_map(struct pipe_buffer *buf, + unsigned flags) +{ + assert(buf); + return buf->vtbl->map(buf, flags); +} + + +static inline void +buffer_unmap(struct pipe_buffer *buf) +{ + assert(buf); + buf->vtbl->unmap(buf); +} + + +static inline void +buffer_get_base_buffer( struct pipe_buffer *buf, + struct pipe_buffer **base_buf, + unsigned *offset ) +{ + buf->vtbl->get_base_buffer(buf, base_buf, offset); +} + + +/** Placeholder for empty buffers. */ +extern struct pipe_buffer null_buffer; + + +/** + * Client buffers (also designated as user buffers) are just for convenience + * of the state tracker, so that it can masquerade its own data as a buffer. + */ +struct pipe_buffer * +client_buffer_create(void *data); + + +/** + * Malloc-based buffer to store data that can't be used by the graphics + * hardware. + */ +struct pipe_buffer * +malloc_buffer_create(unsigned size); + + +#endif /*PB_BUFFER_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_client.c b/src/mesa/pipe/pipebuffer/pb_buffer_client.c new file mode 100644 index 0000000000..bb7f5a9a94 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_buffer_client.c @@ -0,0 +1,123 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of client buffer (also designated as "user buffers"), which + * are just state-tracker owned data masqueraded as buffers. + * + * \author José Fonseca + */ + + +#include "pb_buffer.h" + + +struct client_buffer +{ + struct pipe_buffer base; + void *data; +}; + + +extern const struct pipe_buffer_vtbl client_buffer_vtbl; + + +static inline struct client_buffer * +client_buffer(struct pipe_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &client_buffer_vtbl); + return (struct client_buffer *)buf; +} + + +static void +client_buffer_reference(struct pipe_buffer *buf) +{ + /* No-op */ +} + + +static void +client_buffer_release(struct pipe_buffer *buf) +{ + assert(buf); + free(buf); +} + + +static void * +client_buffer_map(struct pipe_buffer *buf, + unsigned flags) +{ + return client_buffer(buf)->data; +} + + +static void +client_buffer_unmap(struct pipe_buffer *buf) +{ + /* No-op */ +} + + +static void +client_buffer_get_base_buffer(struct pipe_buffer *buf, + struct pipe_buffer **base_buf, + unsigned *offset) +{ + *base_buf = buf; + *offset = 0; +} + + +const struct pipe_buffer_vtbl +client_buffer_vtbl = { + client_buffer_reference, + client_buffer_release, + client_buffer_map, + client_buffer_unmap, + client_buffer_get_base_buffer +}; + + +struct pipe_buffer * +client_buffer_create(void *data) +{ + struct client_buffer *buf; + + buf = (struct client_buffer *)malloc(sizeof(struct client_buffer)); + if(!buf) + return NULL; + + buf->base.vtbl = &client_buffer_vtbl; + + buf->data = data; + + return &buf->base; +} diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c new file mode 100644 index 0000000000..c5a06c545e --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -0,0 +1,340 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of fenced buffers. + * + * \author José Fonseca + * \author Thomas Hellström + */ + + +#include +#include +#include + +#include "main/imports.h" +#include "glapi/glthread.h" +#include "linked_list.h" + +#include "p_winsys.h" + +#include "pb_buffer.h" +#include "pb_buffer_fenced.h" + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct fenced_buffer_list +{ + _glthread_Mutex mutex; + + struct pipe_winsys *winsys; + + size_t numDelayed; + size_t checkDelayed; + + struct list_head delayed; +}; + + +/** + * Wrapper around a pipe buffer which adds fencing and reference counting. + */ +struct fenced_buffer +{ + struct pipe_buffer base; + + struct pipe_buffer *buffer; + + unsigned refcount; + struct pipe_fence_handle *fence; + + struct list_head head; + struct fenced_buffer_list *list; +}; + + +static inline struct fenced_buffer * +fenced_buffer(struct pipe_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &fenced_buffer_vtbl); + return (struct fenced_buffer *)buf; +} + + +static void +_fenced_buffer_destroy(struct fenced_buffer *fenced_buf) +{ + buffer_release(fenced_buf->buffer); + free(fenced_buf); +} + + +static void +_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait) +{ + struct pipe_winsys *winsys = fenced_list->winsys; + struct fenced_buffer *fenced_buf; + struct list_head *list, *prev; + int signaled = -1; + int i; + + list = fenced_list->delayed.next; + + if (fenced_list->numDelayed > 3) { + for (i = 0; i < fenced_list->numDelayed; i += 3) { + list = list->next; + } + } + + prev = list->prev; + for (; list != &fenced_list->delayed; list = prev, prev = list->prev) { + + fenced_buf = LIST_ENTRY(struct fenced_buffer, list, head); + + if (signaled != 0) { + if (wait) { + signaled = winsys->fence_finish(winsys, fenced_buf->fence, 0); + } + else { + signaled = winsys->fence_signalled(winsys, fenced_buf->fence, 0); + } + } + + if (signaled != 0) + /* XXX: we are assuming that buffers are freed in the same order they + * are fenced which may not always be true... + */ + break; + + winsys->fence_reference(winsys, &fenced_buf->fence, NULL); + + LIST_DEL(list); + fenced_list->numDelayed--; + + _fenced_buffer_destroy(fenced_buf); + } +} + + +static void +fenced_buffer_reference(struct pipe_buffer *buf) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + + _glthread_LOCK_MUTEX(fenced_list->mutex); + fenced_buf->refcount++; + _glthread_UNLOCK_MUTEX(fenced_list->mutex); +} + + +static void +fenced_buffer_release(struct pipe_buffer *buf) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + + _glthread_LOCK_MUTEX(fenced_list->mutex); + + fenced_buf->refcount--; + if(!fenced_buf->refcount) { + if (fenced_buf->fence) { + LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); + fenced_list->numDelayed++; + } + else { + _fenced_buffer_destroy(fenced_buf); + } + + if ((fenced_list->numDelayed % fenced_list->checkDelayed) == 0) + _fenced_buffer_list_check_free(fenced_list, 0); + } + + _glthread_UNLOCK_MUTEX(fenced_list->mutex); +} + + +static void * +fenced_buffer_map(struct pipe_buffer *buf, + unsigned flags) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + return buffer_map(fenced_buf->buffer, flags); +} + + +static void +fenced_buffer_unmap(struct pipe_buffer *buf) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + buffer_unmap(fenced_buf->buffer); +} + + +static void +fenced_buffer_get_base_buffer(struct pipe_buffer *buf, + struct pipe_buffer **base_buf, + unsigned *offset) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + buffer_get_base_buffer(fenced_buf->buffer, base_buf, offset); +} + + +const struct pipe_buffer_vtbl +fenced_buffer_vtbl = { + fenced_buffer_reference, + fenced_buffer_release, + fenced_buffer_map, + fenced_buffer_unmap, + fenced_buffer_get_base_buffer +}; + + +struct pipe_buffer * +fenced_buffer_create(struct fenced_buffer_list *fenced_list, + struct pipe_buffer *buffer) +{ + struct fenced_buffer *buf; + + buf = (struct fenced_buffer *)calloc(1, sizeof(struct fenced_buffer)); + if(!buf) + return NULL; + + buf->base.vtbl = &fenced_buffer_vtbl; + buf->buffer = buffer; + buf->refcount = 1; + buf->list = fenced_list; + + return &buf->base; +} + + +void +buffer_fence(struct pipe_buffer *buf, + struct pipe_fence_handle *fence) +{ + if(buf->vtbl == &fenced_buffer_vtbl) { + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + struct pipe_winsys *winsys = fenced_list->winsys; + + _glthread_LOCK_MUTEX(fenced_list->mutex); + winsys->fence_reference(winsys, &fenced_buf->fence, fence); + _glthread_UNLOCK_MUTEX(fenced_list->mutex); + } + else + assert(0); +} + + +void +buffer_unfence(struct pipe_buffer *buf) +{ + buffer_fence(buf, NULL); +} + + +int +buffer_finish(struct pipe_buffer *buf, + unsigned flag) +{ + if(buf->vtbl == &fenced_buffer_vtbl) { + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + if(fenced_buf->fence) { + struct fenced_buffer_list *fenced_list = fenced_buf->list; + struct pipe_winsys *winsys = fenced_list->winsys; + return winsys->fence_finish(winsys, fenced_buf->fence, flag); + } + } + + return 0; +} + + +struct fenced_buffer_list * +fenced_buffer_list_create(struct pipe_winsys *winsys) +{ + struct fenced_buffer_list *fenced_list; + + fenced_list = (struct fenced_buffer_list *)calloc(1, sizeof(*fenced_list)); + if (!fenced_list) + return NULL; + + fenced_list->winsys = winsys; + + LIST_INITHEAD(&fenced_list->delayed); + + fenced_list->numDelayed = 0; + + /* TODO: don't hard code this */ + fenced_list->checkDelayed = 5; + + _glthread_INIT_MUTEX(fenced_list->mutex); + + return fenced_list; +} + + +void +fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait) +{ + _glthread_LOCK_MUTEX(fenced_list->mutex); + _fenced_buffer_list_check_free(fenced_list, wait); + _glthread_UNLOCK_MUTEX(fenced_list->mutex); +} + + +void +fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) +{ + _glthread_LOCK_MUTEX(fenced_list->mutex); + + /* Wait on outstanding fences */ + while (fenced_list->numDelayed) { + _glthread_UNLOCK_MUTEX(fenced_list->mutex); + sched_yield(); + _fenced_buffer_list_check_free(fenced_list, GL_TRUE); + _glthread_LOCK_MUTEX(fenced_list->mutex); + } + + _glthread_UNLOCK_MUTEX(fenced_list->mutex); + + free(fenced_list); +} + + diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h new file mode 100644 index 0000000000..d12428e216 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h @@ -0,0 +1,138 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer fencing. + * + * "Fenced buffers" is actually a misnomer. They should be referred as + * "fenceable buffers", i.e, buffers that can be fenced, but I couldn't find + * the word "fenceable" in the dictionary. + * + * A "fenced buffer" is a decorator around a normal buffer, which adds two + * special properties: + * - the ability for the destruction to be delayed by a fence; + * - reference counting. + * + * Usually DMA buffers have a life-time that will extend the life-time of its + * handle. The end-of-life is dictated by the fence signalling. + * + * Between the handle's destruction, and the fence signalling, the buffer is + * stored in a fenced buffer list. + * + * \author José Fonseca + */ + +#ifndef PB_BUFFER_FENCED_H_ +#define PB_BUFFER_FENCED_H_ + + +#include + + +struct pipe_winsys; +struct pipe_buffer; +struct pipe_fence_handle; + + +/** + * List of buffers which are awaiting fence signalling. + */ +struct fenced_buffer_list; + + +/** + * The fenced buffer's virtual function table. + * + * NOTE: Made public for debugging purposes. + */ +extern const struct pipe_buffer_vtbl fenced_buffer_vtbl; + + +/** + * Create a fenced buffer list. + * + * See also fenced_bufmgr_create for a more convenient way to use this. + */ +struct fenced_buffer_list * +fenced_buffer_list_create(struct pipe_winsys *winsys); + + +/** + * Walk the fenced buffer list to check and free signalled buffers. + */ +void +fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait); + +void +fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list); + + +/** + * Wrap a buffer in a fenced buffer. + * + * NOTE: this will not increase the buffer reference count. + */ +struct pipe_buffer * +fenced_buffer_create(struct fenced_buffer_list *fenced, + struct pipe_buffer *buffer); + + +/** + * Set a buffer's fence. + * + * NOTE: Although it takes a generic pipe buffer argument, it will fail + * on everything but buffers returned by fenced_buffer_create. + */ +void +buffer_fence(struct pipe_buffer *buf, + struct pipe_fence_handle *fence); + + +/** + * Remove the buffer's fence. + * + * NOTE: Although it takes a generic pipe buffer argument, it will fail + * on everything but buffers returned by fenced_buffer_create. + */ +void +buffer_unfence(struct pipe_buffer *buf); + + +/** + * Wait for the buffer fence to signal. + * + * See also pipe_winsys::fence_finish(). + */ +int +buffer_finish(struct pipe_buffer *buf, + unsigned flag); + + + +#endif /*PB_BUFFER_FENCED_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_handle.c b/src/mesa/pipe/pipebuffer/pb_buffer_handle.c new file mode 100644 index 0000000000..5a5eaee7ac --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_buffer_handle.c @@ -0,0 +1,182 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Drop-in implementation of the winsys driver functions for buffer handles. + * + * \author José Fonseca + */ + + +#include +#include + +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" + +#include "pb_buffer.h" +#include "pb_buffer_handle.h" + + +static struct pipe_buffer_handle * +buffer_handle_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned flags, + unsigned hint) +{ + struct pipe_buffer_handle *handle; + + handle = (struct pipe_buffer_handle *)malloc(sizeof(struct pipe_buffer_handle)); + if(!handle) + return NULL; + + handle->refcount = 1; + handle->alignment = alignment; + handle->flags = flags; + handle->hint = hint; + + handle->buf = &null_buffer; + + return handle; +} + + +static struct pipe_buffer_handle * +buffer_handle_create_user(struct pipe_winsys *winsys, + void *data, unsigned size) +{ + struct pipe_buffer_handle *handle; + struct pipe_buffer *buf; + + handle = buffer_handle_create(winsys, 1, 0, 0); + if(!handle) + return NULL; + + buf = client_buffer_create(data); + if(!buf) { + free(handle); + return NULL; + } + + buffer_handle_data(handle, buf); + + return handle; +} + + +static void * +buffer_handle_map(struct pipe_winsys *winsys, + struct pipe_buffer_handle *handle, + unsigned flags) +{ + return buffer_map(handle->buf, flags); +} + + +static void +buffer_handle_unmap(struct pipe_winsys *winsys, + struct pipe_buffer_handle *handle) +{ + buffer_unmap(handle->buf); +} + + +static void +buffer_handle_reference(struct pipe_winsys *winsys, + struct pipe_buffer_handle **dst, + struct pipe_buffer_handle *src) +{ + /* XXX: should this be thread safe? */ + + if (src) { + src->refcount++; + } + + if (*dst) { + (*dst)->refcount--; + if ((*dst)->refcount == 0) { + buffer_release((*dst)->buf); + free(*dst); + } + } + + *dst = src; +} + + +static int +buffer_handle_subdata(struct pipe_winsys *winsys, + struct pipe_buffer_handle *handle, + unsigned long offset, + unsigned long size, + const void *data) +{ + void *map; + assert(handle); + assert(data); + map = buffer_handle_map(winsys, handle, PIPE_BUFFER_FLAG_WRITE); + if(map) { + memcpy((char *)map + offset, data, size); + buffer_handle_unmap(winsys, handle); + return 0; + } + return -1; +} + + +static int +buffer_handle_get_subdata(struct pipe_winsys *winsys, + struct pipe_buffer_handle *handle, + unsigned long offset, + unsigned long size, + void *data) +{ + void *map; + assert(handle); + assert(data); + map = buffer_handle_map(winsys, handle, PIPE_BUFFER_FLAG_READ); + if(map) { + memcpy(data, (char *)map + offset, size); + buffer_handle_unmap(winsys, handle); + return 0; + } + return -1; +} + + +void +buffer_handle_init_winsys(struct pipe_winsys *winsys) +{ + winsys->buffer_create = buffer_handle_create; + winsys->user_buffer_create = buffer_handle_create_user; + winsys->buffer_map = buffer_handle_map; + winsys->buffer_unmap = buffer_handle_unmap; + winsys->buffer_reference = buffer_handle_reference; + winsys->buffer_subdata = buffer_handle_subdata; + winsys->buffer_get_subdata = buffer_handle_get_subdata; +} diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_handle.h b/src/mesa/pipe/pipebuffer/pb_buffer_handle.h new file mode 100644 index 0000000000..076eec2fdc --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_buffer_handle.h @@ -0,0 +1,120 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer handle interface. + * + * \author José Fonseca + */ + +#ifndef PB_BUFFER_HANDLE_H_ +#define PB_BUFFER_HANDLE_H_ + + +#include + +#include "pb_buffer.h" + + +/** + * Buffer handle. + * + * The buffer handle and the buffer data storage are separate entities. This + * is modelled after ARB_vertex_buffer_object, which is the interface that + * Gallium requires. See p_winsys.h for more information. + */ +struct pipe_buffer_handle +{ + /** Reference count */ + unsigned refcount; + + /** Allocation characteristics */ + unsigned alignment; + unsigned flags; + unsigned hint; + + /** + * The actual buffer. + * + * It should never be NULL. Use null_buffer instead. + */ + struct pipe_buffer *buf; +}; + + +/** + * Set buffer storage. + * + * NOTE: this will not increase the buffer reference count. + */ +static inline void +buffer_handle_data(struct pipe_buffer_handle *handle, + struct pipe_buffer *buf) +{ + assert(handle); + assert(handle->buf); + buffer_release(handle->buf); + assert(buf); + handle->buf = buf; +} + + +static inline void +buffer_handle_clear(struct pipe_buffer_handle *handle) +{ + buffer_handle_data(handle, &null_buffer); +} + + +static inline int +buffer_handle_has_data(struct pipe_buffer_handle *handle) { + assert(handle); + assert(handle->buf); + return handle->buf != &null_buffer; +} + + +/** + * Fill in the pipe_winsys' buffer-related callbacks. + * + * Specifically, the fullfilled functions are: + * - buffer_create + * - user_buffer_create + * - buffer_map + * - buffer_unmap + * - buffer_reference + * - buffer_subdata + * - buffer_get_subdata + * + * NOTE: buffer_data is left untouched. + */ +void +buffer_handle_init_winsys(struct pipe_winsys *winsys); + + +#endif /*PB_BUFFER_HANDLE_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c new file mode 100644 index 0000000000..65ad51e1e7 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c @@ -0,0 +1,132 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of malloc-based buffers to store data that can't be processed + * by the hardware. + * + * \author José Fonseca + */ + + +#include +#include + +#include "pb_buffer.h" + + +struct malloc_buffer +{ + struct pipe_buffer base; + void *data; +}; + + +extern const struct pipe_buffer_vtbl malloc_buffer_vtbl; + +static inline struct malloc_buffer * +malloc_buffer(struct pipe_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &malloc_buffer_vtbl); + return (struct malloc_buffer *)buf; +} + + +static void +malloc_buffer_reference(struct pipe_buffer *buf) +{ + /* no-op */ +} + + +static void +malloc_buffer_release(struct pipe_buffer *buf) +{ + free(malloc_buffer(buf)->data); + free(buf); +} + + +static void * +malloc_buffer_map(struct pipe_buffer *buf, + unsigned flags) +{ + return malloc_buffer(buf)->data; +} + + +static void +malloc_buffer_unmap(struct pipe_buffer *buf) +{ + /* No-op */ +} + + +static void +malloc_buffer_get_base_buffer(struct pipe_buffer *buf, + struct pipe_buffer **base_buf, + unsigned *offset) +{ + *base_buf = buf; + *offset = 0; +} + + +const struct pipe_buffer_vtbl +malloc_buffer_vtbl = { + malloc_buffer_reference, + malloc_buffer_release, + malloc_buffer_map, + malloc_buffer_unmap, + malloc_buffer_get_base_buffer +}; + + +struct pipe_buffer * +malloc_buffer_create(unsigned size) +{ + struct malloc_buffer *buf; + + /* TODO: accept an alignment parameter */ + /* TODO: do a single allocation */ + + buf = (struct malloc_buffer *)malloc(sizeof(struct malloc_buffer)); + if(!buf) + return NULL; + + buf->base.vtbl = &malloc_buffer_vtbl; + + buf->data = malloc(size); + if(!buf->data) { + free(buf); + return NULL; + } + + return &buf->base; +} diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_null.c b/src/mesa/pipe/pipebuffer/pb_buffer_null.c new file mode 100644 index 0000000000..a356c6b2d5 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_buffer_null.c @@ -0,0 +1,98 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Null buffer implementation. + * + * We have a special null buffer object so that we can safely call buffer + * operations without having to check whether the buffer pointer is null or not. + * + * \author José Fonseca + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" + +#include "pb_buffer.h" + + +static void +null_buffer_reference(struct pipe_buffer *buf) +{ + /* No-op */ +} + + +static void +null_buffer_release(struct pipe_buffer *buf) +{ + /* No-op */ +} + + +static void * +null_buffer_map(struct pipe_buffer *buf, + unsigned flags) +{ + assert(0); + return NULL; +} + + +static void +null_buffer_unmap(struct pipe_buffer *buf) +{ + assert(0); +} + + +static void +null_buffer_get_base_buffer(struct pipe_buffer *buf, + struct pipe_buffer **base_buf, + unsigned *offset) +{ + *base_buf = buf; + *offset = 0; +} + + +const struct pipe_buffer_vtbl +pipe_buffer_vtbl = { + null_buffer_reference, + null_buffer_release, + null_buffer_map, + null_buffer_unmap, + null_buffer_get_base_buffer +}; + + +struct pipe_buffer +null_buffer = { + &pipe_buffer_vtbl +}; diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr.h b/src/mesa/pipe/pipebuffer/pb_bufmgr.h new file mode 100644 index 0000000000..0e6c3a8d37 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr.h @@ -0,0 +1,114 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer management. + * + * A buffer manager does only one basic thing: it creates buffers. Actually, + * "buffer factory" would probably a more accurate description. + * + * You can chain buffer managers so that you can have a finer grained memory + * management and pooling. + * + * For example, for a simple batch buffer manager you would chain: + * - the native buffer manager, which provides DMA memory from the graphics + * memory space; + * - the pool buffer manager, which keep around a pool of equally sized buffers + * to avoid latency associated with the native buffer manager; + * - the fenced buffer manager, which will delay buffer destruction until the + * the moment the card finishing processing it. + * + * \author José Fonseca + */ + +#ifndef PB_BUFMGR_H_ +#define PB_BUFMGR_H_ + + +#include + + +struct pipe_buffer; +struct pipe_winsys; + + +/** + * Abstract base class for all buffer managers. + */ +struct buffer_manager +{ + /* XXX: we will likely need more allocation flags */ + struct pipe_buffer * + (*create_buffer)( struct buffer_manager *mgr, + size_t size); + + void + (*destroy)( struct buffer_manager *mgr ); +}; + + +/** + * Static buffer pool manager. + * + * Manages the allocation of equally sized buffers. It does so by allocating + * a single big buffer and divide it equally sized buffers. + * + * It is meant to manage the allocation of batch buffer pools. + */ +struct buffer_manager * +pool_bufmgr_create(struct buffer_manager *provider, + size_t n, size_t size); + + +/** + * Wraper around the old memory manager. + * + * It managers buffers of different sizes. It does so by allocating a buffer + * with the size of the heap, and then using the old mm memory manager to manage + * that heap. + */ +struct buffer_manager * +mm_bufmgr_create(struct buffer_manager *provider, + size_t size, size_t align2); + + +/** + * Fenced buffer manager. + * + * This manager is just meant for convenience. It wraps the buffers returned + * by another manager in fenced buffers, so that + * + * NOTE: the buffer manager that provides the buffers will be destroyed + * at the same time. + */ +struct buffer_manager * +fenced_bufmgr_create(struct buffer_manager *provider, + struct pipe_winsys *winsys); + + +#endif /*PB_BUFMGR_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c new file mode 100644 index 0000000000..defd8e4df7 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c @@ -0,0 +1,128 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * \file + * A buffer manager that wraps buffers in fenced buffers. + * + * \author José Fonseca + */ + + +#include +#include + +#include "pb_buffer.h" +#include "pb_buffer_fenced.h" +#include "pb_bufmgr.h" + + +struct fenced_buffer_manager +{ + struct buffer_manager base; + + struct buffer_manager *provider; + + struct fenced_buffer_list *fenced_list; +}; + + +static inline struct fenced_buffer_manager * +fenced_buffer_manager(struct buffer_manager *mgr) +{ + assert(mgr); + return (struct fenced_buffer_manager *)mgr; +} + + +static struct pipe_buffer * +fenced_bufmgr_create_buffer(struct buffer_manager *mgr, size_t size) +{ + struct fenced_buffer_manager *fenced_mgr = fenced_buffer_manager(mgr); + struct pipe_buffer *buf; + struct pipe_buffer *fenced_buf; + + /* check for free buffers before allocating new ones */ + fenced_buffer_list_check_free(fenced_mgr->fenced_list, 0); + + buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size); + if(!buf) { + /* try harder to get a buffer */ + fenced_buffer_list_check_free(fenced_mgr->fenced_list, 1); + + buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size); + if(!buf) { + /* give up */ + return NULL; + } + } + + fenced_buf = fenced_buffer_create(fenced_mgr->fenced_list, buf); + if(!fenced_buf) { + buffer_release(buf); + } + + return fenced_buf; +} + + +static void +fenced_bufmgr_destroy(struct buffer_manager *mgr) +{ + struct fenced_buffer_manager *fenced_mgr = fenced_buffer_manager(mgr); + + fenced_buffer_list_destroy(fenced_mgr->fenced_list); + + fenced_mgr->provider->destroy(fenced_mgr->provider); + + free(fenced_mgr); +} + + +struct buffer_manager * +fenced_bufmgr_create(struct buffer_manager *provider, + struct pipe_winsys *winsys) +{ + struct fenced_buffer_manager *fenced_mgr; + + fenced_mgr = (struct fenced_buffer_manager *)calloc(1, sizeof(*fenced_mgr)); + if (!fenced_mgr) + return NULL; + + fenced_mgr->base.destroy = fenced_bufmgr_destroy; + fenced_mgr->base.create_buffer = fenced_bufmgr_create_buffer; + + fenced_mgr->provider = provider; + fenced_mgr->fenced_list = fenced_buffer_list_create(winsys); + if(!fenced_mgr->fenced_list) { + free(fenced_mgr); + return NULL; + } + + return &fenced_mgr->base; +} diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c new file mode 100644 index 0000000000..8e6dcf1116 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -0,0 +1,265 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer manager using the old texture memory manager. + * + * \author José Fonseca + */ + + +#include +#include +#include + +#include "main/imports.h" +#include "glapi/glthread.h" +#include "main/mm.h" +#include "linked_list.h" + +#include "p_defines.h" +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct mm_buffer_manager +{ + struct buffer_manager base; + + _glthread_Mutex mutex; + + size_t size; + struct mem_block *heap; + + size_t align2; + + struct pipe_buffer *buffer; + void *map; +}; + + +static inline struct mm_buffer_manager * +mm_buffer_manager(struct buffer_manager *mgr) +{ + assert(mgr); + return (struct mm_buffer_manager *)mgr; +} + + +struct mm_buffer +{ + struct pipe_buffer base; + + struct mm_buffer_manager *mgr; + + struct mem_block *block; +}; + + +static inline struct mm_buffer * +mm_buffer(struct pipe_buffer *buf) +{ + assert(buf); + return (struct mm_buffer *)buf; +} + + +static void +mm_buffer_reference(struct pipe_buffer *buf) +{ + /* No-op */ +} + + +static void +mm_buffer_release(struct pipe_buffer *buf) +{ + struct mm_buffer *mm_buf = mm_buffer(buf); + struct mm_buffer_manager *mm = mm_buf->mgr; + + _glthread_LOCK_MUTEX(mm->mutex); + mmFreeMem(mm_buf->block); + free(buf); + _glthread_UNLOCK_MUTEX(mm->mutex); +} + + +static void * +mm_buffer_map(struct pipe_buffer *buf, + unsigned flags) +{ + struct mm_buffer *mm_buf = mm_buffer(buf); + struct mm_buffer_manager *mm = mm_buf->mgr; + + return (unsigned char *) mm->map + mm_buf->block->ofs; +} + + +static void +mm_buffer_unmap(struct pipe_buffer *buf) +{ + /* No-op */ +} + + +static void +mm_buffer_get_base_buffer(struct pipe_buffer *buf, + struct pipe_buffer **base_buf, + unsigned *offset) +{ + struct mm_buffer *mm_buf = mm_buffer(buf); + struct mm_buffer_manager *mm = mm_buf->mgr; + buffer_get_base_buffer(mm->buffer, base_buf, offset); + *offset += mm_buf->block->ofs; +} + + +static const struct pipe_buffer_vtbl +mm_buffer_vtbl = { + mm_buffer_reference, + mm_buffer_release, + mm_buffer_map, + mm_buffer_unmap, + mm_buffer_get_base_buffer +}; + + +static struct pipe_buffer * +mm_bufmgr_create_buffer(struct buffer_manager *mgr, + size_t size) +{ + struct mm_buffer_manager *mm = mm_buffer_manager(mgr); + struct mm_buffer *mm_buf; + + _glthread_LOCK_MUTEX(mm->mutex); + + mm_buf = (struct mm_buffer *)malloc(sizeof(*mm_buf)); + if (!mm_buf) { + _glthread_UNLOCK_MUTEX(mm->mutex); + return NULL; + } + + mm_buf->base.vtbl = &mm_buffer_vtbl; + + mm_buf->mgr = mm; + + mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); + if(!mm_buf->block) { + fprintf(stderr, "warning: heap full\n"); +#if 0 + mmDumpMemInfo(mm->heap); +#endif + + mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); + if(!mm_buf->block) { + assert(0); + free(mm_buf); + _glthread_UNLOCK_MUTEX(mm->mutex); + return NULL; + } + } + + /* Some sanity checks */ + assert(0 <= mm_buf->block->ofs && mm_buf->block->ofs < mm->size); + assert(size <= mm_buf->block->size && mm_buf->block->ofs + mm_buf->block->size <= mm->size); + + _glthread_UNLOCK_MUTEX(mm->mutex); + return SUPER(mm_buf); +} + + +static void +mm_bufmgr_destroy(struct buffer_manager *mgr) +{ + struct mm_buffer_manager *mm = mm_buffer_manager(mgr); + + _glthread_LOCK_MUTEX(mm->mutex); + + mmDestroy(mm->heap); + + buffer_unmap(mm->buffer); + buffer_release(mm->buffer); + + _glthread_UNLOCK_MUTEX(mm->mutex); + + free(mgr); +} + + +struct buffer_manager * +mm_bufmgr_create(struct buffer_manager *provider, + size_t size, size_t align2) +{ + struct mm_buffer_manager *mm; + + mm = (struct mm_buffer_manager *)calloc(1, sizeof(*mm)); + if (!mm) + return NULL; + + assert(provider); + assert(provider->create_buffer); + mm->base.create_buffer = mm_bufmgr_create_buffer; + mm->base.destroy = mm_bufmgr_destroy; + + mm->size = size; + mm->align2 = align2; /* 64-byte alignment */ + + _glthread_INIT_MUTEX(mm->mutex); + + mm->buffer = provider->create_buffer(provider, size); + if (!mm->buffer) + goto failure; + + mm->map = buffer_map(mm->buffer, + PIPE_BUFFER_FLAG_READ | PIPE_BUFFER_FLAG_WRITE ); + if(!mm->map) + goto failure; + + mm->heap = mmInit(0, size); + if (!mm->heap) + goto failure; + + return SUPER(mm); + +failure: +if(mm->heap) + mmDestroy(mm->heap); + if(mm->map) + buffer_unmap(mm->buffer); + if(mm->buffer) + buffer_release(mm->buffer); + if(mm) + free(mm); + return NULL; +} diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c new file mode 100644 index 0000000000..ee6fa62500 --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -0,0 +1,279 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * \file + * Batch buffer pool management. + * + * \author José Fonseca + * \author Thomas Hellström + */ + + +#include +#include +#include + +#include "main/imports.h" +#include "glapi/glthread.h" +#include "linked_list.h" + +#include "p_defines.h" + +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct pool_buffer_manager +{ + struct buffer_manager base; + + _glthread_Mutex mutex; + + size_t bufSize; + + size_t numFree; + size_t numTot; + + struct list_head free; + + struct pipe_buffer *buffer; + void *map; + + struct pool_buffer *bufs; +}; + + +static inline struct pool_buffer_manager * +pool_buffer_manager(struct buffer_manager *mgr) +{ + assert(mgr); + return (struct pool_buffer_manager *)mgr; +} + + +struct pool_buffer +{ + struct pipe_buffer base; + + struct pool_buffer_manager *mgr; + + struct list_head head; + + size_t start; +}; + + +static inline struct pool_buffer * +pool_buffer(struct pipe_buffer *buf) +{ + assert(buf); + return (struct pool_buffer *)buf; +} + + +static void +pool_buffer_reference(struct pipe_buffer *buf) +{ + /* No-op */ +} + + +static void +pool_buffer_release(struct pipe_buffer *buf) +{ + struct pool_buffer *pool_buf = pool_buffer(buf); + struct pool_buffer_manager *pool = pool_buf->mgr; + + _glthread_LOCK_MUTEX(pool->mutex); + LIST_ADD(&pool_buf->head, &pool->free); + pool->numFree++; + _glthread_UNLOCK_MUTEX(pool->mutex); +} + + +static void * +pool_buffer_map(struct pipe_buffer *buf, unsigned flags) +{ + struct pool_buffer *pool_buf = pool_buffer(buf); + struct pool_buffer_manager *pool = pool_buf->mgr; + void *map; + + _glthread_LOCK_MUTEX(pool->mutex); + map = (unsigned char *) pool->map + pool_buf->start; + _glthread_UNLOCK_MUTEX(pool->mutex); + return map; +} + + +static void +pool_buffer_unmap(struct pipe_buffer *buf) +{ + /* No-op */ +} + + +static void +pool_buffer_get_base_buffer(struct pipe_buffer *buf, + struct pipe_buffer **base_buf, + unsigned *offset) +{ + struct pool_buffer *pool_buf = pool_buffer(buf); + struct pool_buffer_manager *pool = pool_buf->mgr; + buffer_get_base_buffer(pool->buffer, base_buf, offset); + *offset += pool_buf->start; +} + + +static const struct pipe_buffer_vtbl +pool_buffer_vtbl = { + pool_buffer_reference, + pool_buffer_release, + pool_buffer_map, + pool_buffer_unmap, + pool_buffer_get_base_buffer +}; + + +static struct pipe_buffer * +pool_bufmgr_create_buffer(struct buffer_manager *mgr, size_t size) +{ + struct pool_buffer_manager *pool = pool_buffer_manager(mgr); + struct pool_buffer *pool_buf; + struct list_head *item; + + assert(size == pool->bufSize); + + _glthread_LOCK_MUTEX(pool->mutex); + + if (pool->numFree == 0) { + _glthread_UNLOCK_MUTEX(pool->mutex); + fprintf(stderr, "warning: out of fixed size buffer objects\n"); + return NULL; + } + + item = pool->free.next; + + if (item == &pool->free) { + _glthread_UNLOCK_MUTEX(pool->mutex); + fprintf(stderr, "error: fixed size buffer pool corruption\n"); + return NULL; + } + + LIST_DEL(item); + --pool->numFree; + + _glthread_UNLOCK_MUTEX(pool->mutex); + pool_buf = LIST_ENTRY(struct pool_buffer, item, head); + return SUPER(pool_buf); +} + + +static void +pool_bufmgr_destroy(struct buffer_manager *mgr) +{ + struct pool_buffer_manager *pool = pool_buffer_manager(mgr); + _glthread_LOCK_MUTEX(pool->mutex); + + free(pool->bufs); + + buffer_unmap(pool->buffer); + buffer_release(pool->buffer); + + _glthread_UNLOCK_MUTEX(pool->mutex); + + free(mgr); +} + + +struct buffer_manager * +pool_bufmgr_create(struct buffer_manager *provider, + size_t numBufs, + size_t bufSize) +{ + struct pool_buffer_manager *pool; + struct pool_buffer *pool_buf; + int i; + + pool = (struct pool_buffer_manager *)calloc(1, sizeof(*pool)); + if (!pool) + return NULL; + + pool->base.destroy = pool_bufmgr_destroy; + pool->base.create_buffer = pool_bufmgr_create_buffer; + + LIST_INITHEAD(&pool->free); + + pool->numTot = numBufs; + pool->numFree = numBufs; + pool->bufSize = bufSize; + + _glthread_INIT_MUTEX(pool->mutex); + + pool->buffer = provider->create_buffer(provider, numBufs*bufSize); + if (!pool->buffer) + goto failure; + + pool->map = buffer_map(pool->buffer, + PIPE_BUFFER_FLAG_READ | + PIPE_BUFFER_FLAG_WRITE ); + if(!pool->map) + goto failure; + + pool->bufs = (struct pool_buffer *) malloc(numBufs * sizeof(*pool->bufs)); + if (!pool->bufs) + goto failure; + + pool_buf = pool->bufs; + for (i = 0; i < numBufs; ++i) { + pool_buf->base.vtbl = &pool_buffer_vtbl; + pool_buf->mgr = pool; + pool_buf->start = i * bufSize; + LIST_ADDTAIL(&pool_buf->head, &pool->free); + pool_buf++; + } + + return SUPER(pool); + +failure: + if(pool->bufs) + free(pool->bufs); + if(pool->map) + buffer_unmap(pool->buffer); + if(pool->buffer) + buffer_release(pool->buffer); + if(pool) + free(pool); + return NULL; +} -- cgit v1.2.3 From 273f6e769897032b15a8739394669fd11f76f145 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 25 Nov 2007 20:23:10 +0000 Subject: gallium: Remove buffer_unfence and buffer_finish. These interfaces have been deemed useless by Keith. --- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 24 ------------------------ src/mesa/pipe/pipebuffer/pb_buffer_fenced.h | 21 --------------------- 2 files changed, 45 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c index c5a06c545e..e69f88a323 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -261,30 +261,6 @@ buffer_fence(struct pipe_buffer *buf, } -void -buffer_unfence(struct pipe_buffer *buf) -{ - buffer_fence(buf, NULL); -} - - -int -buffer_finish(struct pipe_buffer *buf, - unsigned flag) -{ - if(buf->vtbl == &fenced_buffer_vtbl) { - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - if(fenced_buf->fence) { - struct fenced_buffer_list *fenced_list = fenced_buf->list; - struct pipe_winsys *winsys = fenced_list->winsys; - return winsys->fence_finish(winsys, fenced_buf->fence, flag); - } - } - - return 0; -} - - struct fenced_buffer_list * fenced_buffer_list_create(struct pipe_winsys *winsys) { diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h index d12428e216..07e42a67f8 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h @@ -114,25 +114,4 @@ buffer_fence(struct pipe_buffer *buf, struct pipe_fence_handle *fence); -/** - * Remove the buffer's fence. - * - * NOTE: Although it takes a generic pipe buffer argument, it will fail - * on everything but buffers returned by fenced_buffer_create. - */ -void -buffer_unfence(struct pipe_buffer *buf); - - -/** - * Wait for the buffer fence to signal. - * - * See also pipe_winsys::fence_finish(). - */ -int -buffer_finish(struct pipe_buffer *buf, - unsigned flag); - - - #endif /*PB_BUFFER_FENCED_H_*/ -- cgit v1.2.3 From 6cdcebe8017b56e2f823dd3f6d805e55056f7e9b Mon Sep 17 00:00:00 2001 From: Michal Date: Wed, 28 Nov 2007 18:10:31 +0000 Subject: Allow tgsi_dump print-out to stderr or a string buffer. Handle source register complement modifier in exec and dump. --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 4 + src/mesa/pipe/tgsi/util/tgsi_dump.c | 308 ++++++++++++++++++++++++------------ src/mesa/pipe/tgsi/util/tgsi_dump.h | 8 +- 3 files changed, 220 insertions(+), 100 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index d036e9d383..ab83f27c1b 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1066,6 +1066,10 @@ fetch_source( case TGSI_UTIL_SIGN_KEEP: break; } + + if (reg->SrcRegisterExtMod.Complement) { + micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan ); + } } static void diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.c b/src/mesa/pipe/tgsi/util/tgsi_dump.c index 982d5ce796..3f4d930dec 100644 --- a/src/mesa/pipe/tgsi/util/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/util/tgsi_dump.c @@ -31,129 +31,196 @@ #include "tgsi_parse.h" #include "tgsi_build.h" -struct text_dump +struct gen_dump { - FILE *file; unsigned tabs; + void (* write)( + struct gen_dump *dump, + const void *data, + unsigned size ); +}; + +struct text_dump +{ + struct gen_dump base; + char *text; + unsigned length; + unsigned capacity; }; static void -text_dump_str( - struct text_dump *dump, - const char *str ) +_text_dump_write( + struct gen_dump *dump, + const void *data, + unsigned size ) { - unsigned i; - size_t len = strlen( str ); + struct text_dump *td = (struct text_dump *) dump; + unsigned new_length = td->length + size; + + if( new_length >= td->capacity ) { + unsigned new_capacity = td->capacity; + + do { + if( new_capacity == 0 ) { + new_capacity = 256; + } + else { + new_capacity *= 2; + } + } while( new_length >= new_capacity ); + td->text = (char *) REALLOC( + td->text, + td->capacity, + new_capacity ); + td->capacity = new_capacity; + } + memcpy( + &td->text[td->length], + data, + size ); + td->length = new_length; + td->text[td->length] = '\0'; +} + +struct file_dump +{ + struct gen_dump base; + FILE *file; +}; + +static void +_file_dump_write( + struct gen_dump *dump, + const void *data, + unsigned size ) +{ + struct file_dump *fd = (struct file_dump *) dump; + +#if 0 + fwrite( data, 1, size, fd->file ); +#else + { + unsigned i; - for( i = 0; i < len; i++ ) { - fprintf( dump->file, "%c", str[i] ); + for (i = 0; i < size; i++ ) { + fprintf( fd->file, "%c", ((const char *) data)[i] ); + } + } +#endif +} - if( str[i] == '\n' ) { +static void +gen_dump_str( + struct gen_dump *dump, + const char *str ) +{ + unsigned i; + size_t len = strlen( str ); + + for (i = 0; i < len; i++) { + dump->write( dump, &str[i], 1 ); + if (str[i] == '\n') { unsigned i; - for( i = 0; i < dump->tabs; i++ ) { - fprintf( dump->file, " " ); + for (i = 0; i < dump->tabs; i++) { + dump->write( dump, " ", 4 ); } } } } static void -text_dump_chr( - struct text_dump *dump, - const char chr ) +gen_dump_chr( + struct gen_dump *dump, + const char chr ) { - char str[2]; - - str[0] = chr; - str[1] = '\0'; - text_dump_str( dump, str ); + dump->write( dump, &chr, 1 ); } static void -text_dump_uix( - struct text_dump *dump, - const unsigned ui ) +gen_dump_uix( + struct gen_dump *dump, + const unsigned ui ) { - char str[36]; + char str[36]; sprintf( str, "0x%x", ui ); - text_dump_str( dump, str ); + gen_dump_str( dump, str ); } static void -text_dump_uid( - struct text_dump *dump, - const unsigned ui ) +gen_dump_uid( + struct gen_dump *dump, + const unsigned ui ) { - char str[16]; + char str[16]; sprintf( str, "%u", ui ); - text_dump_str( dump, str ); + gen_dump_str( dump, str ); } static void -text_dump_sid( - struct text_dump *dump, - const int si ) +gen_dump_sid( + struct gen_dump *dump, + const int si ) { - char str[16]; + char str[16]; sprintf( str, "%d", si ); - text_dump_str( dump, str ); + gen_dump_str( dump, str ); } static void -text_dump_flt( - struct text_dump *dump, - const float f ) +gen_dump_flt( + struct gen_dump *dump, + const float flt ) { - char str[48]; + char str[48]; - sprintf( str, "%10.4f", f ); - text_dump_str( dump, str ); + sprintf( str, "%10.4f", flt ); + gen_dump_str( dump, str ); } static void -text_dump_enum( - struct text_dump *dump, - const unsigned e, - const char **enums, - const unsigned enums_count ) +gen_dump_enum( + struct gen_dump *dump, + const unsigned e, + const char **enums, + const unsigned enums_count ) { - if( e >= enums_count ) { - text_dump_uid( dump, e ); + if (e >= enums_count) { + gen_dump_uid( dump, e ); } else { - text_dump_str( dump, enums[e] ); + gen_dump_str( dump, enums[e] ); } } static void -text_dump_tab( - struct text_dump *dump ) +gen_dump_tab( + struct gen_dump *dump ) { - dump->tabs++; + ++dump->tabs; } static void -text_dump_untab( - struct text_dump *dump ) +gen_dump_untab( + struct gen_dump *dump ) { assert( dump->tabs > 0 ); --dump->tabs; } -#define TXT(S) text_dump_str( dump, S ) -#define CHR(C) text_dump_chr( dump, C ) -#define UIX(I) text_dump_uix( dump, I ) -#define UID(I) text_dump_uid( dump, I ) -#define SID(I) text_dump_sid( dump, I ) -#define FLT(F) text_dump_flt( dump, F ) -#define TAB() text_dump_tab( dump ) -#define UNT() text_dump_untab( dump ) -#define ENM(E,ENUMS) text_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) +#define TXT(S) gen_dump_str( dump, S ) +#define CHR(C) gen_dump_chr( dump, C ) +#define UIX(I) gen_dump_uix( dump, I ) +#define UID(I) gen_dump_uid( dump, I ) +#define SID(I) gen_dump_sid( dump, I ) +#define FLT(F) gen_dump_flt( dump, F ) +#define TAB() gen_dump_tab( dump ) +#define UNT() gen_dump_untab( dump ) +#define ENM(E,ENUMS) gen_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) static const char *TGSI_PROCESSOR_TYPES[] = { @@ -659,8 +726,8 @@ static const char *TGSI_MODULATES[] = static void dump_declaration_short( - struct text_dump *dump, - struct tgsi_full_declaration *decl ) + struct gen_dump *dump, + struct tgsi_full_declaration *decl ) { TXT( "\nDCL " ); ENM( decl->Declaration.File, TGSI_FILES_SHORT ); @@ -711,11 +778,11 @@ dump_declaration_short( static void dump_declaration_verbose( - struct text_dump *dump, - struct tgsi_full_declaration *decl, - unsigned ignored, - unsigned deflt, - struct tgsi_full_declaration *fd ) + struct gen_dump *dump, + struct tgsi_full_declaration *decl, + unsigned ignored, + unsigned deflt, + struct tgsi_full_declaration *fd ) { TXT( "\nFile : " ); ENM( decl->Declaration.File, TGSI_FILES ); @@ -792,7 +859,7 @@ dump_declaration_verbose( static void dump_immediate_short( - struct text_dump *dump, + struct gen_dump *dump, struct tgsi_full_immediate *imm ) { unsigned i; @@ -820,9 +887,9 @@ dump_immediate_short( static void dump_immediate_verbose( - struct text_dump *dump, + struct gen_dump *dump, struct tgsi_full_immediate *imm, - unsigned ignored ) + unsigned ignored ) { unsigned i; @@ -849,12 +916,12 @@ dump_immediate_verbose( static void dump_instruction_short( - struct text_dump *dump, - struct tgsi_full_instruction *inst, - unsigned instno ) + struct gen_dump *dump, + struct tgsi_full_instruction *inst, + unsigned instno ) { unsigned i; - boolean first_reg = TRUE; + boolean first_reg = TRUE; CHR( '\n' ); UID( instno ); @@ -915,6 +982,9 @@ dump_instruction_short( } CHR( ' ' ); + if( src->SrcRegisterExtMod.Complement ) { + TXT( "(1 - " ); + } if( src->SrcRegisterExtMod.Negate ) { CHR( '-' ); } @@ -957,6 +1027,9 @@ dump_instruction_short( if( src->SrcRegisterExtMod.Absolute ) { CHR( '|' ); } + if( src->SrcRegisterExtMod.Complement ) { + CHR( ')' ); + } first_reg = FALSE; } @@ -975,11 +1048,11 @@ dump_instruction_short( static void dump_instruction_verbose( - struct text_dump *dump, - struct tgsi_full_instruction *inst, - unsigned ignored, - unsigned deflt, - struct tgsi_full_instruction *fi ) + struct gen_dump *dump, + struct tgsi_full_instruction *inst, + unsigned ignored, + unsigned deflt, + struct tgsi_full_instruction *fi ) { unsigned i; @@ -1317,13 +1390,12 @@ dump_instruction_verbose( } } -void -tgsi_dump( +static void +dump_gen( + struct gen_dump *dump, const struct tgsi_token *tokens, - unsigned flags ) + unsigned flags ) { - struct text_dump _dump; - struct text_dump *dump = &_dump; struct tgsi_parse_context parse; struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; @@ -1332,18 +1404,7 @@ tgsi_dump( unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); unsigned instno = 0; - { -#if 0 - static unsigned counter = 0; - char buffer[64]; - - sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); - dump->file = fopen( buffer, "wt" ); -#else - dump->file = stderr; -#endif - dump->tabs = 0; - } + dump->tabs = 0; /* sanity check */ assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); @@ -1457,3 +1518,52 @@ tgsi_dump( tgsi_parse_free( &parse ); } +void +tgsi_dump( + const struct tgsi_token *tokens, + unsigned flags ) +{ + struct file_dump dump; + + dump.base.write = _file_dump_write; +#if 0 + { + static unsigned counter = 0; + char buffer[64]; + sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); + dump.file = fopen( buffer, "wt" ); + } +#else + dump.file = stderr; +#endif + + dump_gen( + &dump.base, + tokens, + flags ); + +#if 0 + fclose( dump.file ); +#endif +} + +void +tgsi_dump_str( + char **str, + const struct tgsi_token *tokens, + unsigned flags ) +{ + struct text_dump dump; + + dump.base.write = _text_dump_write; + dump.text = NULL; + dump.length = 0; + dump.capacity = 0; + + dump_gen( + &dump.base, + tokens, + flags ); + + *str = dump.text; +} diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.h b/src/mesa/pipe/tgsi/util/tgsi_dump.h index 70860c0885..1adc9db251 100644 --- a/src/mesa/pipe/tgsi/util/tgsi_dump.h +++ b/src/mesa/pipe/tgsi/util/tgsi_dump.h @@ -12,7 +12,13 @@ extern "C" { void tgsi_dump( const struct tgsi_token *tokens, - unsigned flags ); + unsigned flags ); + +void +tgsi_dump_str( + char **str, + const struct tgsi_token *tokens, + unsigned flags ); #if defined __cplusplus } // extern "C" -- cgit v1.2.3 From de7e51ff9854605cd0bbbe5829d6d987ea0a5fca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 5 Dec 2007 15:28:39 +0000 Subject: gallium: use SSE by default --- src/mesa/pipe/draw/draw_context.c | 2 +- src/mesa/pipe/softpipe/sp_context.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index d438c0be10..6792a06a4e 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -42,7 +42,7 @@ struct draw_context *draw_create( void ) struct draw_context *draw = CALLOC_STRUCT( draw_context ); #if defined(__i386__) || defined(__386__) - draw->use_sse = GETENV( "GALLIUM_SSE" ) != NULL; + draw->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; #else draw->use_sse = FALSE; #endif diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index df4e0cbd5a..809b165f45 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -244,7 +244,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, uint i; #if defined(__i386__) || defined(__386__) - softpipe->use_sse = GETENV( "GALLIUM_SSE" ) != NULL; + softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; #else softpipe->use_sse = FALSE; #endif -- cgit v1.2.3 From f8f9580a2a1c89af1dc0e169b62440053d9d7e81 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 9 Dec 2007 18:26:26 +0000 Subject: Adapt for winsys interface changes. --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 38 +++++++++++++--------- src/mesa/drivers/x11/xm_winsys.c | 7 ++-- src/mesa/pipe/draw/draw_vertex_fetch.c | 1 - src/mesa/pipe/i915simple/i915_texture.c | 3 +- src/mesa/pipe/softpipe/sp_texture.c | 4 ++- src/mesa/pipe/xlib/xm_winsys.c | 14 +++++--- src/mesa/state_tracker/st_cb_fbo.c | 2 +- 7 files changed, 43 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 1799e9b901..86ea86a58f 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -105,42 +105,48 @@ intel_buffer_reference(struct pipe_winsys *winsys, /* Grabs the hardware lock! */ -static void intel_buffer_data(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, - unsigned size, const void *data, - unsigned usage ) +static int intel_buffer_data(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned size, const void *data, + unsigned usage ) { driBOData( dri_bo(buf), size, data, 0 ); + return 0; } -static void intel_buffer_subdata(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) +static int intel_buffer_subdata(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) { driBOSubData( dri_bo(buf), offset, size, data ); + return 0; } -static void intel_buffer_get_subdata(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) +static int intel_buffer_get_subdata(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) { driBOGetSubData( dri_bo(buf), offset, size, data ); + return 0; } /* Pipe has no concept of pools. We choose the tex/region pool * for all buffers. */ static struct pipe_buffer_handle * -intel_buffer_create(struct pipe_winsys *winsys, unsigned flags) +intel_buffer_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned flags, + unsigned hint ) { struct _DriBufferObject *buffer; struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); driGenBuffers( iws->regionPool, - "pipe buffer", 1, &buffer, 64, 0, 0 ); + "pipe buffer", 1, &buffer, alignment, flags, hint ); return pipe_bo(buffer); } diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index f863cdbc15..dafbe96a1e 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -210,7 +210,10 @@ xm_get_name(struct pipe_winsys *pws) static struct pipe_buffer_handle * -xm_buffer_create(struct pipe_winsys *pws, unsigned alignment) +xm_buffer_create(struct pipe_winsys *pws, + unsigned alignment, + unsigned flags, + unsigned hint) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); buffer->refcount = 1; @@ -258,7 +261,7 @@ xm_region_alloc(struct pipe_winsys *winsys, assert(region->pitch > 0); - region->buffer = winsys->buffer_create( winsys, alignment ) + region->buffer = winsys->buffer_create( winsys, alignment, 0, 0 ) ; /* NULL data --> just allocate the space */ diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index b51cff59f5..7d983ebd29 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -80,7 +80,6 @@ fetch_attrib4(const void *ptr, enum pipe_format format, float attrib[4]) break; case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_U_A8_R8_G8_B8: case PIPE_FORMAT_R8G8B8A8_UNORM: attrib[0] = (float) ((unsigned char *) ptr)[2] / 255.0f; attrib[1] = (float) ((unsigned char *) ptr)[1] / 255.0f; diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c index 1ca2815dfb..59e8db8a95 100644 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -496,7 +496,8 @@ i915_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : i915_miptree_layout(pipe, tex)) { tex->buffer = pipe->winsys->buffer_create(pipe->winsys, - PIPE_SURFACE_FLAG_TEXTURE); + PIPE_SURFACE_FLAG_TEXTURE, + 0, 0); if (tex->buffer) pipe->winsys->buffer_data(pipe->winsys, tex->buffer, diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index 2288c343bf..cfe9628184 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -382,7 +382,9 @@ softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) if (softpipe_mipmap_tree_layout(pipe, spt)) { spt->buffer = pipe->winsys->buffer_create(pipe->winsys, - PIPE_SURFACE_FLAG_TEXTURE); + 32, + PIPE_SURFACE_FLAG_TEXTURE, + 0); if (spt->buffer) { pipe->winsys->buffer_data(pipe->winsys, spt->buffer, diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 976884ad52..c347d1c2a3 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -155,7 +155,7 @@ xm_buffer_reference(struct pipe_winsys *pws, } } -static void +static int xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, unsigned size, const void *data, unsigned usage ) { @@ -169,9 +169,10 @@ xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, } if (data) memcpy(xm_buf->data, data, size); + return 0; } -static void +static int xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, const void *data) { @@ -180,9 +181,10 @@ xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, assert(!xm_buf->userBuffer); assert(b); memcpy(b + offset, data, size); + return 0; } -static void +static int xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, unsigned long offset, unsigned long size, void *data) { @@ -191,6 +193,7 @@ xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, assert(!xm_buf->userBuffer); assert(b); memcpy(data, b + offset, size); + return 0; } @@ -255,7 +258,10 @@ xm_get_name(struct pipe_winsys *pws) static struct pipe_buffer_handle * -xm_buffer_create(struct pipe_winsys *pws, unsigned flags) +xm_buffer_create(struct pipe_winsys *pws, + unsigned alignment, + unsigned flags, + unsigned hints) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); buffer->refcount = 1; diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index cbda56b5c3..10396c3feb 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -118,7 +118,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe->winsys->buffer_reference(pipe->winsys, &strb->surface->buffer, NULL); - strb->surface->buffer = pipe->winsys->buffer_create(pipe->winsys, flags); + strb->surface->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, flags, 0); if (!strb->surface->buffer) return GL_FALSE; /* out of memory, try s/w buffer? */ -- cgit v1.2.3 From 04e88f469cf6c338ba04640738865b59e160c3d4 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 9 Dec 2007 20:45:20 -0700 Subject: added null ptr check in draw_destroy() for rasterizer stage --- src/mesa/pipe/draw/draw_context.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 6792a06a4e..33727e6547 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -102,7 +102,8 @@ void draw_destroy( struct draw_context *draw ) draw->pipeline.cull->destroy( draw->pipeline.cull ); draw->pipeline.feedback->destroy( draw->pipeline.feedback ); draw->pipeline.validate->destroy( draw->pipeline.validate ); - draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); + if (draw->pipeline.rasterize) + draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); FREE( draw->vcache.vertex[0] ); /* Frees all the vertices. */ FREE( draw ); } -- cgit v1.2.3 From 98848382731b71a51e4a80f1820f009e70f06a25 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 9 Dec 2007 21:10:39 -0700 Subject: Fix looping for multi-color buffer writing. --- src/mesa/pipe/softpipe/sp_context.h | 2 +- src/mesa/pipe/softpipe/sp_quad.c | 2 +- src/mesa/pipe/softpipe/sp_quad_blend.c | 6 ++++-- src/mesa/pipe/softpipe/sp_quad_bufloop.c | 4 +--- src/mesa/pipe/softpipe/sp_quad_colormask.c | 3 ++- src/mesa/pipe/softpipe/sp_quad_output.c | 3 ++- 6 files changed, 11 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index d4763a98c6..b97cdc52c6 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -154,7 +154,7 @@ struct softpipe_context { struct draw_stage *setup; struct draw_stage *vbuf; - struct pipe_surface *cbuf; /**< current color buffer (one of cbufs) */ + uint current_cbuf; /**< current color buffer being written to */ struct softpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS]; struct softpipe_tile_cache *zbuf_cache; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index e0327c4cf9..6330465a8b 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -77,7 +77,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) if (sp->framebuffer.num_cbufs == 1) { /* the usual case: write to exactly one colorbuf */ - sp->cbuf = sp->framebuffer.cbufs[0]; + sp->current_cbuf = 0; } else { /* insert bufloop stage */ diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 696e252af1..76a0873fc5 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -107,7 +107,8 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) uint *dst4 = (uint *) dst; uint *res4 = (uint *) res; struct softpipe_cached_tile * - tile = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[0], + tile = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[softpipe->current_cbuf], quad->x0, quad->y0); uint i, j; @@ -224,7 +225,8 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) static const float one[4] = { 1, 1, 1, 1 }; float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[0], + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[softpipe->current_cbuf], quad->x0, quad->y0); uint i, j; diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c index aac70e2b04..e704b4043c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -30,7 +30,7 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { /* set current cbuffer */ - softpipe->cbuf = softpipe->framebuffer.cbufs[i]; + softpipe->current_cbuf = i; /* pass blended quad to next stage */ qs->next->run(qs->next, quad); @@ -38,8 +38,6 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) /* restore quad's colors for next buffer */ memcpy(quad->outputs.color, tmp, sizeof(tmp)); } - - softpipe->cbuf = NULL; /* prevent accidental use */ } diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index c585aa3edd..f5e0a32d1b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -50,7 +50,8 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) float dest[4][QUAD_SIZE]; struct softpipe_cached_tile *tile = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[0], quad->x0, quad->y0); + softpipe->cbuf_cache[softpipe->current_cbuf], + quad->x0, quad->y0); uint i, j; /* get/swizzle dest colors */ diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index cebfec18f7..f04a550e3d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -43,7 +43,8 @@ output_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[0], + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[softpipe->current_cbuf], quad->x0, quad->y0); /* in-tile pos: */ const int itx = quad->x0 % TILE_SIZE; -- cgit v1.2.3 From 63e2ab83f6f3955c230bf71662136b34ff46dfe2 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 9 Dec 2007 21:44:26 -0700 Subject: use quadColor local var --- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 13 +- src/mesa/pipe/softpipe/sp_quad_blend.c | 194 ++++++++++++++-------------- src/mesa/pipe/softpipe/sp_quad_colormask.c | 9 +- src/mesa/pipe/softpipe/sp_quad_coverage.c | 3 +- src/mesa/pipe/softpipe/sp_quad_output.c | 3 +- 5 files changed, 114 insertions(+), 108 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index 6a9cf29e47..7b56bceba2 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -16,6 +16,7 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) struct softpipe_context *softpipe = qs->softpipe; const float ref = softpipe->alpha_test->ref; unsigned passMask = 0x0, j; + const float *aaaa = quad->outputs.color[3]; switch (softpipe->alpha_test->func) { case PIPE_FUNC_NEVER: @@ -27,42 +28,42 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) * passMask = (quad->outputs.color[3] <= vec4(ref)); */ for (j = 0; j < QUAD_SIZE; j++) { - if (quad->outputs.color[3][j] < ref) { + if (aaaa[j] < ref) { passMask |= (1 << j); } } break; case PIPE_FUNC_EQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->outputs.color[3][j] == ref) { + if (aaaa[j] == ref) { passMask |= (1 << j); } } break; case PIPE_FUNC_LEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->outputs.color[3][j] <= ref) { + if (aaaa[j] <= ref) { passMask |= (1 << j); } } break; case PIPE_FUNC_GREATER: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->outputs.color[3][j] > ref) { + if (aaaa[j] > ref) { passMask |= (1 << j); } } break; case PIPE_FUNC_NOTEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->outputs.color[3][j] != ref) { + if (aaaa[j] != ref) { passMask |= (1 << j); } } break; case PIPE_FUNC_GEQUAL: for (j = 0; j < QUAD_SIZE; j++) { - if (quad->outputs.color[3][j] >= ref) { + if (aaaa[j] >= ref) { passMask |= (1 << j); } } diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 76a0873fc5..1843e20684 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -110,6 +110,7 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) tile = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[softpipe->current_cbuf], quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color; uint i, j; /* get/swizzle dest colors */ @@ -128,10 +129,10 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) UNCLAMPED_FLOAT_TO_UBYTE(dst[j][2], dest[j][2]); /* P2 */ UNCLAMPED_FLOAT_TO_UBYTE(dst[j][3], dest[j][3]); /* P3 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quad->outputs.color[j][0]); /* P0 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quad->outputs.color[j][1]); /* P1 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quad->outputs.color[j][2]); /* P2 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quad->outputs.color[j][3]); /* P3 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quadColor[j][0]); /* P0 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quadColor[j][1]); /* P1 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quadColor[j][2]); /* P2 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quadColor[j][3]); /* P3 */ } switch (softpipe->blend->logicop_func) { @@ -204,10 +205,10 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) } for (j = 0; j < 4; j++) { - quad->outputs.color[j][0] = UBYTE_TO_FLOAT(res[j][0]); - quad->outputs.color[j][1] = UBYTE_TO_FLOAT(res[j][1]); - quad->outputs.color[j][2] = UBYTE_TO_FLOAT(res[j][2]); - quad->outputs.color[j][3] = UBYTE_TO_FLOAT(res[j][3]); + quadColor[j][0] = UBYTE_TO_FLOAT(res[j][0]); + quadColor[j][1] = UBYTE_TO_FLOAT(res[j][1]); + quadColor[j][2] = UBYTE_TO_FLOAT(res[j][2]); + quadColor[j][3] = UBYTE_TO_FLOAT(res[j][3]); } /* pass quad to next stage */ @@ -228,6 +229,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[softpipe->current_cbuf], quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color; uint i, j; if (softpipe->blend->logicop_enable) { @@ -249,39 +251,39 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) */ switch (softpipe->blend->rgb_src_factor) { case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[0], quad->outputs.color[0]); /* R */ - VEC4_COPY(source[1], quad->outputs.color[1]); /* G */ - VEC4_COPY(source[2], quad->outputs.color[2]); /* B */ + VEC4_COPY(source[0], quadColor[0]); /* R */ + VEC4_COPY(source[1], quadColor[1]); /* G */ + VEC4_COPY(source[2], quadColor[2]); /* B */ break; case PIPE_BLENDFACTOR_SRC_COLOR: - VEC4_MUL(source[0], quad->outputs.color[0], quad->outputs.color[0]); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], quad->outputs.color[1]); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], quad->outputs.color[2]); /* B */ + VEC4_MUL(source[0], quadColor[0], quadColor[0]); /* R */ + VEC4_MUL(source[1], quadColor[1], quadColor[1]); /* G */ + VEC4_MUL(source[2], quadColor[2], quadColor[2]); /* B */ break; case PIPE_BLENDFACTOR_SRC_ALPHA: { - const float *alpha = quad->outputs.color[3]; - VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ + const float *alpha = quadColor[3]; + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ } break; case PIPE_BLENDFACTOR_DST_COLOR: - VEC4_MUL(source[0], quad->outputs.color[0], dest[0]); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], dest[1]); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], dest[2]); /* B */ + VEC4_MUL(source[0], quadColor[0], dest[0]); /* R */ + VEC4_MUL(source[1], quadColor[1], dest[1]); /* G */ + VEC4_MUL(source[2], quadColor[2], dest[2]); /* B */ break; case PIPE_BLENDFACTOR_DST_ALPHA: { const float *alpha = dest[3]; - VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ } break; case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: { - const float *alpha = quad->outputs.color[3]; + const float *alpha = quadColor[3]; float diff[4]; VEC4_SUB(diff, one, dest[3]); VEC4_MIN(source[0], alpha, diff); /* R */ @@ -293,20 +295,20 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) { float comp[4]; VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ - VEC4_MUL(source[0], quad->outputs.color[0], comp); /* R */ + VEC4_MUL(source[0], quadColor[0], comp); /* R */ VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ - VEC4_MUL(source[1], quad->outputs.color[1], comp); /* G */ + VEC4_MUL(source[1], quadColor[1], comp); /* G */ VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ - VEC4_MUL(source[2], quad->outputs.color[2], comp); /* B */ + VEC4_MUL(source[2], quadColor[2], comp); /* B */ } break; case PIPE_BLENDFACTOR_CONST_ALPHA: { float alpha[4]; VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quad->outputs.color[0], alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], alpha); /* B */ + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ } break; case PIPE_BLENDFACTOR_SRC1_COLOR: @@ -323,41 +325,41 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_INV_SRC_COLOR: { float inv_comp[4]; - VEC4_SUB(inv_comp, one, quad->outputs.color[0]); /* R */ - VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, quad->outputs.color[1]); /* G */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, quad->outputs.color[2]); /* B */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); /* B */ + VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ } break; case PIPE_BLENDFACTOR_INV_SRC_ALPHA: { float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, quad->outputs.color[3]); - VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ + VEC4_SUB(inv_alpha, one, quadColor[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ } break; case PIPE_BLENDFACTOR_INV_DST_ALPHA: { float inv_alpha[4]; VEC4_SUB(inv_alpha, one, dest[3]); - VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ } break; case PIPE_BLENDFACTOR_INV_DST_COLOR: { float inv_comp[4]; VEC4_SUB(inv_comp, one, dest[0]); /* R */ - VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); /* R */ + VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ VEC4_SUB(inv_comp, one, dest[1]); /* G */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); /* G */ + VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ VEC4_SUB(inv_comp, one, dest[2]); /* B */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); /* B */ + VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ } break; case PIPE_BLENDFACTOR_INV_CONST_COLOR: @@ -365,22 +367,22 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) float inv_comp[4]; /* R */ VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); - VEC4_MUL(source[0], quad->outputs.color[0], inv_comp); + VEC4_MUL(source[0], quadColor[0], inv_comp); /* G */ VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); - VEC4_MUL(source[1], quad->outputs.color[1], inv_comp); + VEC4_MUL(source[1], quadColor[1], inv_comp); /* B */ VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); - VEC4_MUL(source[2], quad->outputs.color[2], inv_comp); + VEC4_MUL(source[2], quadColor[2], inv_comp); } break; case PIPE_BLENDFACTOR_INV_CONST_ALPHA: { float inv_alpha[4]; VEC4_SCALAR(inv_alpha, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quad->outputs.color[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quad->outputs.color[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quad->outputs.color[2], inv_alpha); /* B */ + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ } break; case PIPE_BLENDFACTOR_INV_SRC1_COLOR: @@ -398,24 +400,24 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) */ switch (softpipe->blend->alpha_src_factor) { case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[3], quad->outputs.color[3]); /* A */ + VEC4_COPY(source[3], quadColor[3]); /* A */ break; case PIPE_BLENDFACTOR_SRC_COLOR: /* fall-through */ case PIPE_BLENDFACTOR_SRC_ALPHA: { - const float *alpha = quad->outputs.color[3]; - VEC4_MUL(source[3], quad->outputs.color[3], alpha); /* A */ + const float *alpha = quadColor[3]; + VEC4_MUL(source[3], quadColor[3], alpha); /* A */ } break; case PIPE_BLENDFACTOR_DST_COLOR: /* fall-through */ case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(source[3], quad->outputs.color[3], dest[3]); /* A */ + VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */ break; case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: { - const float *alpha = quad->outputs.color[3]; + const float *alpha = quadColor[3]; float diff[4]; VEC4_SUB(diff, one, dest[3]); VEC4_MIN(source[3], alpha, diff); /* A */ @@ -427,7 +429,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) { float comp[4]; VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(source[3], quad->outputs.color[3], comp); /* A */ + VEC4_MUL(source[3], quadColor[3], comp); /* A */ } break; case PIPE_BLENDFACTOR_ZERO: @@ -438,8 +440,8 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_INV_SRC_ALPHA: { float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, quad->outputs.color[3]); - VEC4_MUL(source[3], quad->outputs.color[3], inv_alpha); /* A */ + VEC4_SUB(inv_alpha, one, quadColor[3]); + VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ } break; case PIPE_BLENDFACTOR_INV_DST_COLOR: @@ -448,7 +450,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) { float inv_alpha[4]; VEC4_SUB(inv_alpha, one, dest[3]); - VEC4_MUL(source[3], quad->outputs.color[3], inv_alpha); /* A */ + VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ } break; case PIPE_BLENDFACTOR_INV_CONST_COLOR: @@ -458,7 +460,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) float inv_comp[4]; /* A */ VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(source[3], quad->outputs.color[3], inv_comp); + VEC4_MUL(source[3], quadColor[3], inv_comp); } break; default: @@ -474,14 +476,14 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) /* dest = dest * 1 NO-OP, leave dest as-is */ break; case PIPE_BLENDFACTOR_SRC_COLOR: - VEC4_MUL(dest[0], dest[0], quad->outputs.color[0]); /* R */ - VEC4_MUL(dest[1], dest[1], quad->outputs.color[1]); /* G */ - VEC4_MUL(dest[2], dest[2], quad->outputs.color[2]); /* B */ + VEC4_MUL(dest[0], dest[0], quadColor[0]); /* R */ + VEC4_MUL(dest[1], dest[1], quadColor[1]); /* G */ + VEC4_MUL(dest[2], dest[2], quadColor[2]); /* B */ break; case PIPE_BLENDFACTOR_SRC_ALPHA: - VEC4_MUL(dest[0], dest[0], quad->outputs.color[3]); /* R * A */ - VEC4_MUL(dest[1], dest[1], quad->outputs.color[3]); /* G * A */ - VEC4_MUL(dest[2], dest[2], quad->outputs.color[3]); /* B * A */ + VEC4_MUL(dest[0], dest[0], quadColor[3]); /* R * A */ + VEC4_MUL(dest[1], dest[1], quadColor[3]); /* G * A */ + VEC4_MUL(dest[2], dest[2], quadColor[3]); /* B * A */ break; case PIPE_BLENDFACTOR_DST_ALPHA: VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */ @@ -529,18 +531,18 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_INV_SRC_COLOR: { float inv_comp[4]; - VEC4_SUB(inv_comp, one, quad->outputs.color[0]); /* R */ + VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ - VEC4_SUB(inv_comp, one, quad->outputs.color[1]); /* G */ + VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ - VEC4_SUB(inv_comp, one, quad->outputs.color[2]); /* B */ + VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ } break; case PIPE_BLENDFACTOR_INV_SRC_ALPHA: { float one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); + VEC4_SUB(one_minus_alpha, one, quadColor[3]); VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ @@ -549,7 +551,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_INV_DST_ALPHA: { float inv_comp[4]; - VEC4_SUB(inv_comp, one, quad->outputs.color[3]); /* A */ + VEC4_SUB(inv_comp, one, quadColor[3]); /* A */ VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ @@ -608,7 +610,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_SRC_COLOR: /* fall-through */ case PIPE_BLENDFACTOR_SRC_ALPHA: - VEC4_MUL(dest[3], dest[3], quad->outputs.color[3]); /* A * A */ + VEC4_MUL(dest[3], dest[3], quadColor[3]); /* A * A */ break; case PIPE_BLENDFACTOR_DST_COLOR: /* fall-through */ @@ -635,7 +637,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_INV_SRC_ALPHA: { float one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quad->outputs.color[3]); + VEC4_SUB(one_minus_alpha, one, quadColor[3]); VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ } break; @@ -666,29 +668,29 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) */ switch (softpipe->blend->rgb_func) { case PIPE_BLEND_ADD: - VEC4_ADD(quad->outputs.color[0], source[0], dest[0]); /* R */ - VEC4_ADD(quad->outputs.color[1], source[1], dest[1]); /* G */ - VEC4_ADD(quad->outputs.color[2], source[2], dest[2]); /* B */ + VEC4_ADD(quadColor[0], source[0], dest[0]); /* R */ + VEC4_ADD(quadColor[1], source[1], dest[1]); /* G */ + VEC4_ADD(quadColor[2], source[2], dest[2]); /* B */ break; case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quad->outputs.color[0], source[0], dest[0]); /* R */ - VEC4_SUB(quad->outputs.color[1], source[1], dest[1]); /* G */ - VEC4_SUB(quad->outputs.color[2], source[2], dest[2]); /* B */ + VEC4_SUB(quadColor[0], source[0], dest[0]); /* R */ + VEC4_SUB(quadColor[1], source[1], dest[1]); /* G */ + VEC4_SUB(quadColor[2], source[2], dest[2]); /* B */ break; case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quad->outputs.color[0], dest[0], source[0]); /* R */ - VEC4_SUB(quad->outputs.color[1], dest[1], source[1]); /* G */ - VEC4_SUB(quad->outputs.color[2], dest[2], source[2]); /* B */ + VEC4_SUB(quadColor[0], dest[0], source[0]); /* R */ + VEC4_SUB(quadColor[1], dest[1], source[1]); /* G */ + VEC4_SUB(quadColor[2], dest[2], source[2]); /* B */ break; case PIPE_BLEND_MIN: - VEC4_MIN(quad->outputs.color[0], source[0], dest[0]); /* R */ - VEC4_MIN(quad->outputs.color[1], source[1], dest[1]); /* G */ - VEC4_MIN(quad->outputs.color[2], source[2], dest[2]); /* B */ + VEC4_MIN(quadColor[0], source[0], dest[0]); /* R */ + VEC4_MIN(quadColor[1], source[1], dest[1]); /* G */ + VEC4_MIN(quadColor[2], source[2], dest[2]); /* B */ break; case PIPE_BLEND_MAX: - VEC4_MAX(quad->outputs.color[0], source[0], dest[0]); /* R */ - VEC4_MAX(quad->outputs.color[1], source[1], dest[1]); /* G */ - VEC4_MAX(quad->outputs.color[2], source[2], dest[2]); /* B */ + VEC4_MAX(quadColor[0], source[0], dest[0]); /* R */ + VEC4_MAX(quadColor[1], source[1], dest[1]); /* G */ + VEC4_MAX(quadColor[2], source[2], dest[2]); /* B */ break; default: assert(0); @@ -699,19 +701,19 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) */ switch (softpipe->blend->alpha_func) { case PIPE_BLEND_ADD: - VEC4_ADD(quad->outputs.color[3], source[3], dest[3]); /* A */ + VEC4_ADD(quadColor[3], source[3], dest[3]); /* A */ break; case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quad->outputs.color[3], source[3], dest[3]); /* A */ + VEC4_SUB(quadColor[3], source[3], dest[3]); /* A */ break; case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quad->outputs.color[3], dest[3], source[3]); /* A */ + VEC4_SUB(quadColor[3], dest[3], source[3]); /* A */ break; case PIPE_BLEND_MIN: - VEC4_MIN(quad->outputs.color[3], source[3], dest[3]); /* A */ + VEC4_MIN(quadColor[3], source[3], dest[3]); /* A */ break; case PIPE_BLEND_MAX: - VEC4_MAX(quad->outputs.color[3], source[3], dest[3]); /* A */ + VEC4_MAX(quadColor[3], source[3], dest[3]); /* A */ break; default: abort(); diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index f5e0a32d1b..867cc0725a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -52,6 +52,7 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) = sp_get_cached_tile(softpipe, softpipe->cbuf_cache[softpipe->current_cbuf], quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color; uint i, j; /* get/swizzle dest colors */ @@ -65,19 +66,19 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) /* R */ if (!(softpipe->blend->colormask & PIPE_MASK_R)) - COPY_4V(quad->outputs.color[0], dest[0]); + COPY_4V(quadColor[0], dest[0]); /* G */ if (!(softpipe->blend->colormask & PIPE_MASK_G)) - COPY_4V(quad->outputs.color[1], dest[1]); + COPY_4V(quadColor[1], dest[1]); /* B */ if (!(softpipe->blend->colormask & PIPE_MASK_B)) - COPY_4V(quad->outputs.color[2], dest[2]); + COPY_4V(quadColor[2], dest[2]); /* A */ if (!(softpipe->blend->colormask & PIPE_MASK_A)) - COPY_4V(quad->outputs.color[3], dest[3]); + COPY_4V(quadColor[3], dest[3]); /* pass quad to next stage */ qs->next->run(qs->next, quad); diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c index 9dfad7c580..cca2b9f79b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -50,11 +50,12 @@ coverage_quad(struct quad_stage *qs, struct quad_header *quad) if ((softpipe->rasterizer->poly_smooth && quad->prim == PRIM_TRI) || (softpipe->rasterizer->line_smooth && quad->prim == PRIM_LINE) || (softpipe->rasterizer->point_smooth && quad->prim == PRIM_POINT)) { + float (*quadColor)[4] = quad->outputs.color; unsigned j; for (j = 0; j < QUAD_SIZE; j++) { assert(quad->coverage[j] >= 0.0); assert(quad->coverage[j] <= 1.0); - quad->outputs.color[3][j] *= quad->coverage[j]; + quadColor[3][j] *= quad->coverage[j]; } } diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index f04a550e3d..bfd7baa946 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -49,6 +49,7 @@ output_quad(struct quad_stage *qs, struct quad_header *quad) /* in-tile pos: */ const int itx = quad->x0 % TILE_SIZE; const int ity = quad->y0 % TILE_SIZE; + float (*quadColor)[4] = quad->outputs.color; int i, j; /* get/swizzle dest colors */ @@ -57,7 +58,7 @@ output_quad(struct quad_stage *qs, struct quad_header *quad) int x = itx + (j & 1); int y = ity + (j >> 1); for (i = 0; i < 4; i++) { /* loop over color chans */ - tile->data.color[y][x][i] = quad->outputs.color[i][j]; + tile->data.color[y][x][i] = quadColor[i][j]; } } } -- cgit v1.2.3 From 913703d4b822a794853e566d4ab2bfef01ca5099 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Mon, 10 Dec 2007 09:11:57 +0100 Subject: Remove stray references to struct pipe_region. --- src/mesa/drivers/dri/intel_winsys/intel_context.h | 1 - src/mesa/pipe/p_winsys.h | 7 ------- src/mesa/state_tracker/st_texture.h | 1 - 3 files changed, 9 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h index eac5cd8b4c..b01370c049 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_context.h @@ -35,7 +35,6 @@ struct pipe_context; -struct pipe_region; struct intel_context; struct _DriBufferObject; struct st_context; diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index f2be7655f4..1418af6918 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -45,7 +45,6 @@ struct pipe_buffer_handle; /** Opaque type */ struct pipe_fence_handle; -struct pipe_region; struct pipe_surface; @@ -80,12 +79,6 @@ struct pipe_winsys /** * flags is bitmask of PIPE_SURFACE_FLAG_RENDER, PIPE_SURFACE_FLAG_TEXTURE */ - struct pipe_region *(*region_alloc)(struct pipe_winsys *ws, - unsigned size, unsigned flags); - - void (*region_release)(struct pipe_winsys *ws, struct pipe_region **r); - - unsigned (*surface_pitch)(struct pipe_winsys *ws, unsigned cpp, unsigned with, unsigned flags); diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index f49d91a069..2be53abf3a 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -33,7 +33,6 @@ struct pipe_context; struct pipe_texture; -struct pipe_region; extern struct pipe_texture * -- cgit v1.2.3 From f7195b32323d50e9541a5c5cb3143e83d05bd8d8 Mon Sep 17 00:00:00 2001 From: Ben Skeggs Date: Mon, 10 Dec 2007 21:07:53 +1100 Subject: Correct user VBO size. --- src/mesa/state_tracker/st_draw.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index c6fea1f7bc..24bb4823bc 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -266,9 +266,14 @@ st_draw_vbo(GLcontext *ctx, } else { /* attribute data is in user-space memory, not a VBO */ - uint bytes = (arrays[mesaAttr]->Size - * _mesa_sizeof_type(arrays[mesaAttr]->Type) - * (max_index + 1)); + uint bytes; + + if (!arrays[mesaAttr]->StrideB) { + bytes = arrays[mesaAttr]->Size + * _mesa_sizeof_type(arrays[mesaAttr]->Type); + } else { + bytes = arrays[mesaAttr]->StrideB * (max_index + 1); + } /* wrap user data */ vbuffer[attr].buffer -- cgit v1.2.3 From 41ce1f8fb30c9a918d85d4016ce82fe2a3d2c72c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Mon, 10 Dec 2007 16:34:25 +0100 Subject: Link fewer common objects into Gallium winsys layers. These are useless or even harmful due to referencing symbols no longer available in the Gallium build. --- src/mesa/drivers/dri/Makefile.template | 12 +++++++----- src/mesa/drivers/dri/intel_winsys/Makefile | 2 +- 2 files changed, 8 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/Makefile.template b/src/mesa/drivers/dri/Makefile.template index 37bf0a4977..3e7e527a98 100644 --- a/src/mesa/drivers/dri/Makefile.template +++ b/src/mesa/drivers/dri/Makefile.template @@ -2,14 +2,16 @@ MESA_MODULES = $(TOP)/src/mesa/libmesa.a -COMMON_SOURCES = \ - ../../common/driverfuncs.c \ +COMMON_GALLIUM_SOURCES = \ ../common/utils.c \ - ../common/texmem.c \ ../common/vblank.c \ ../common/dri_util.c \ - ../common/xmlconfig.c \ - ../common/drirenderbuffer.c + ../common/xmlconfig.c + +COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \ + ../../common/driverfuncs.c \ + ../common/texmem.c \ + ../common/drirenderbuffer.c COMMON_BM_SOURCES = \ ../common/dri_bufmgr.c \ diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile index 46b9541d7d..9ae0f01325 100644 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ b/src/mesa/drivers/dri/intel_winsys/Makefile @@ -22,7 +22,7 @@ DRIVER_SOURCES = \ intel_batchpool.c C_SOURCES = \ - $(COMMON_SOURCES) \ + $(COMMON_GALLIUM_SOURCES) \ $(COMMON_BM_SOURCES) \ $(DRIVER_SOURCES) -- cgit v1.2.3 From 37484a387495f1241e5e1220243d6355f5333aca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Mon, 10 Dec 2007 17:31:52 +0100 Subject: Fix up some confusion wrt winsys->buffer_create alignment / flags parameters. intel_winsys works again. --- src/mesa/pipe/i915simple/i915_texture.c | 4 +--- src/mesa/pipe/softpipe/sp_texture.c | 5 +---- src/mesa/state_tracker/st_cb_fbo.c | 2 +- 3 files changed, 3 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c index 59e8db8a95..fefd105adf 100644 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -495,9 +495,7 @@ i915_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : i915_miptree_layout(pipe, tex)) { - tex->buffer = pipe->winsys->buffer_create(pipe->winsys, - PIPE_SURFACE_FLAG_TEXTURE, - 0, 0); + tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, 0, 0); if (tex->buffer) pipe->winsys->buffer_data(pipe->winsys, tex->buffer, diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index cfe9628184..2dd1add6f7 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -381,10 +381,7 @@ softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) sizeof(struct softpipe_texture) - sizeof(struct pipe_texture)); if (softpipe_mipmap_tree_layout(pipe, spt)) { - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, - 32, - PIPE_SURFACE_FLAG_TEXTURE, - 0); + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); if (spt->buffer) { pipe->winsys->buffer_data(pipe->winsys, spt->buffer, diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 10396c3feb..36d25a576d 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -118,7 +118,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe->winsys->buffer_reference(pipe->winsys, &strb->surface->buffer, NULL); - strb->surface->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, flags, 0); + strb->surface->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); if (!strb->surface->buffer) return GL_FALSE; /* out of memory, try s/w buffer? */ -- cgit v1.2.3 From 1b20b2d53b0aa4c81291e1ffd888b59fb9f1fb02 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Dec 2007 13:41:19 -0700 Subject: don't use get/put_tile_rgba() for accum buffer accesses --- src/mesa/state_tracker/st_cb_accum.c | 117 +++++++++++++++++++++++++++++------ 1 file changed, 98 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index c8d9cba12f..ea0b1187fc 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -45,6 +45,10 @@ #include "pipe/p_inlines.h" +#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ + us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) + + /** * For hardware that supports deep color buffers, we could accelerate * most/all the accum operations with blending/texturing. @@ -55,7 +59,6 @@ void st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { - struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *acc_strb = st_renderbuffer(rb); struct pipe_surface *acc_ps = acc_strb->surface; const GLint xpos = ctx->DrawBuffer->_Xmin; @@ -66,28 +69,104 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) const GLfloat g = ctx->Accum.ClearColor[1]; const GLfloat b = ctx->Accum.ClearColor[2]; const GLfloat a = ctx->Accum.ClearColor[3]; - GLfloat *accBuf; - GLint i; (void) pipe_surface_map(acc_ps); - accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - - for (i = 0; i < width * height; i++) { - accBuf[i * 4 + 0] = r; - accBuf[i * 4 + 1] = g; - accBuf[i * 4 + 2] = b; - accBuf[i * 4 + 3] = a; + switch (acc_ps->format) { + case PIPE_FORMAT_R16G16B16A16_SNORM: + { + const short sr = (short) (32767 * r); + const short sg = (short) (32767 * g); + const short sb = (short) (32767 * b); + const short sa = (short) (32767 * a); + short *acc = ((short *) acc_ps->map) + + (ypos * acc_ps->pitch + xpos) * 4; + int i, j; + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) { + acc[j*4+0] = sr; + acc[j*4+1] = sg; + acc[j*4+2] = sb; + acc[j*4+3] = sa; + } + acc += acc_ps->pitch * 4; + } + } + break; + default: + assert(0); } - pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe_surface_unmap(acc_ps); +} - free(accBuf); - pipe_surface_unmap(acc_ps); +/** Get block of values from accum buffer, converting to float */ +static void +get_accum_tile(struct pipe_context *pipe, + struct pipe_surface *acc_surf, + int xpos, int ypos, int width, int height, + float *buf) +{ + switch (acc_surf->format) { + case PIPE_FORMAT_R16G16B16A16_SNORM: + { + const short *acc = ((const short *) acc_surf->map) + + (ypos * acc_surf->pitch + xpos) * 4; + int i, j; + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) { + buf[j*4+0] = SHORT_TO_FLOAT(acc[j*4+0]); + buf[j*4+1] = SHORT_TO_FLOAT(acc[j*4+1]); + buf[j*4+2] = SHORT_TO_FLOAT(acc[j*4+2]); + buf[j*4+3] = SHORT_TO_FLOAT(acc[j*4+3]); + } + acc += acc_surf->pitch * 4; + buf += width * 4; + } + } + break; + default: + assert(0); + } } +/** Put block of values into accum buffer, converting from float */ +static void +put_accum_tile(struct pipe_context *pipe, + struct pipe_surface *acc_surf, + int xpos, int ypos, int width, int height, + const float *buf) +{ + switch (acc_surf->format) { + case PIPE_FORMAT_R16G16B16A16_SNORM: + { + short *acc = ((short *) acc_surf->map) + + (ypos * acc_surf->pitch + xpos) * 4; + int i, j; + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) { + short r, g, b, a; + UNCLAMPED_FLOAT_TO_SHORT(r, buf[j*4+0]); + UNCLAMPED_FLOAT_TO_SHORT(g, buf[j*4+1]); + UNCLAMPED_FLOAT_TO_SHORT(b, buf[j*4+2]); + UNCLAMPED_FLOAT_TO_SHORT(a, buf[j*4+3]); + acc[j*4+0] = r; + acc[j*4+1] = g; + acc[j*4+2] = b; + acc[j*4+3] = a; + } + acc += acc_surf->pitch * 4; + buf += width * 4; + } + } + break; + default: + assert(0); + } +} + /** For ADD/MULT */ static void @@ -102,13 +181,13 @@ accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, (void) pipe_surface_map(acc_ps); - pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + get_accum_tile(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] * scale + bias; } - pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + put_accum_tile(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(accBuf); @@ -133,13 +212,13 @@ accum_accum(struct pipe_context *pipe, GLfloat value, accMap = pipe_surface_map(acc_ps); pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, colorBuf); - pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + get_accum_tile(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] + colorBuf[i] * value; } - pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + put_accum_tile(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(colorBuf); free(accBuf); @@ -169,7 +248,7 @@ accum_load(struct pipe_context *pipe, GLfloat value, buf[i] = buf[i] * value; } - pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, buf); + put_accum_tile(pipe, acc_ps, xpos, ypos, width, height, buf); free(buf); @@ -194,7 +273,7 @@ accum_return(GLcontext *ctx, GLfloat value, (void) pipe_surface_map(color_ps); (void) pipe_surface_map(acc_ps); - pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); + get_accum_tile(pipe, acc_ps, xpos, ypos, width, height, abuf); if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) { cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); -- cgit v1.2.3 From 4ccd3124f611b33c062c4b0004616171934f1908 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Dec 2007 13:42:34 -0700 Subject: implement i915_put_tile_rgba() - temporary code though --- src/mesa/pipe/i915simple/i915_surface.c | 34 +++++++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index ea48c1bd8a..bd6fd32704 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -52,6 +52,7 @@ * Note: this is exactly like a8r8g8b8_get_tile() in sp_surface.c * Share it someday. */ +/** XXX this will go away eventually */ static void i915_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, @@ -105,13 +106,42 @@ i915_get_tile_rgba(struct pipe_context *pipe, } +/** XXX this will go away eventually */ static void i915_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p) { - /* TODO */ - assert(0); + unsigned *dst + = ((unsigned *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); + + CLIP_TILE; + + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + dst[j] = (a << 24) | (r << 16) | (g << 8) | b; + pRow += 4; + } + dst += ps->pitch; + p += w0 * 4; + } + break; + default: + assert(0); + } } -- cgit v1.2.3 From f26936b35253b697f1ccb5c2898a8607564bdcfe Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Dec 2007 13:45:24 -0700 Subject: disable assertion in sp_tile_cache_set_surface() for now --- src/mesa/pipe/softpipe/sp_tile_cache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index fadd169f5d..25c6dd4d17 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -141,7 +141,7 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, assert(!tc->texture); if (tc->surface && tc->surface->map) { - assert(tc->surface != ps); + /*assert(tc->surface != ps);*/ pipe_surface_unmap(tc->surface); } -- cgit v1.2.3 From 4f58d9af9addb1506a1b2abc7dd8012147772b78 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Dec 2007 13:48:09 -0700 Subject: Add 'type' parameter to is_format_supported() to specify texture vs. drawing surface, etc. Additional types may be added in the future. --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 2 +- src/mesa/pipe/i915simple/i915_context.c | 77 ++++++++---------- src/mesa/pipe/p_context.h | 3 +- src/mesa/pipe/p_defines.h | 7 +- src/mesa/pipe/softpipe/sp_context.c | 24 ++++-- src/mesa/pipe/softpipe/sp_winsys.h | 16 ++-- src/mesa/pipe/xlib/xm_winsys.c | 26 +----- src/mesa/state_tracker/st_cb_drawpixels.c | 4 +- src/mesa/state_tracker/st_cb_fbo.c | 11 ++- src/mesa/state_tracker/st_cb_fbo.h | 3 +- src/mesa/state_tracker/st_format.c | 95 +++++++++++----------- src/mesa/state_tracker/st_format.h | 2 +- src/mesa/state_tracker/st_framebuffer.c | 29 +++---- 13 files changed, 142 insertions(+), 157 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 86ea86a58f..9c643dd0ba 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -191,7 +191,7 @@ intel_i915_surface_pitch(struct pipe_winsys *winsys, */ /* XXX is the pitch different for textures vs. drawables? */ - if (flags & PIPE_SURFACE_FLAG_TEXTURE) /* or PIPE_SURFACE_FLAG_RENDER? */ + if (1/*flags & PIPE_SURFACE_FLAG_TEXTURE*/) /* or PIPE_SURFACE_FLAG_RENDER? */ return ((cpp * width + 63) & ~63) / cpp; else return ((cpp * width + 63) & ~63) / cpp; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index b915a67790..f505ff6ae6 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -39,72 +39,61 @@ /** - * Query format support. - * If we find texture and drawable support differs, add a selector - * parameter or another function. + * Query format support for creating a texture, drawing surface, etc. + * \param format the format to test + * \param type one of PIPE_TEXTURE, PIPE_SURFACE, PIPE_SCREEN_SURFACE */ static boolean i915_is_format_supported( struct pipe_context *pipe, - enum pipe_format format ) + enum pipe_format format, uint type ) { -#if 0 - /* XXX: This is broken -- rewrite if still needed. */ - static const unsigned tex_supported[] = { + static const enum pipe_format tex_supported[] = { PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_R5G6B5_UNORM, PIPE_FORMAT_U_L8, PIPE_FORMAT_U_A8, PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_L8_A8, + PIPE_FORMAT_U_A8_L8, PIPE_FORMAT_YCBCR, PIPE_FORMAT_YCBCR_REV, PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_NONE /* list terminator */ }; - - - /* Actually a lot more than this - add later: - */ - static const unsigned render_supported[] = { + static const enum pipe_format surface_supported[] = { PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_R5G6B5_UNORM, - }; - - /* - */ - static const unsigned z_stencil_supported[] = { - PIPE_FORMAT_Z16_UNORM, - PIPE_FORMAT_Z32_UNORM, PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_R16G16B16A16_SNORM, + PIPE_FORMAT_NONE /* list terminator */ + }; + static const enum pipe_format screen_surface_supported[] = { + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_NONE /* list terminator */ }; + const enum pipe_format *list; + uint i; switch (type) { - case PIPE_RENDER_FORMAT: - *numFormats = Elements(render_supported); - return render_supported; - - case PIPE_TEX_FORMAT: - *numFormats = Elements(tex_supported); - return render_supported; - - case PIPE_Z_STENCIL_FORMAT: - *numFormats = Elements(render_supported); - return render_supported; - + case PIPE_TEXTURE: + list = tex_supported; + break; + case PIPE_SURFACE: + list = surface_supported; + break; + case PIPE_SCREEN_SURFACE: + list = screen_surface_supported; + break; default: - *numFormats = 0; - return NULL; + assert(0); } -#else - switch( format ) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - return TRUE; - default: - return FALSE; - }; -#endif + + for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) { + if (list[i] == format) + return TRUE; + } + + return FALSE; } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index b3a2122ade..00379fbacf 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -50,8 +50,9 @@ struct pipe_context { /* * Queries */ + /** type is one of PIPE_SURFACE, PIPE_TEXTURE, etc. */ boolean (*is_format_supported)( struct pipe_context *pipe, - enum pipe_format format ); + enum pipe_format format, uint type ); const char *(*get_name)( struct pipe_context *pipe ); diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 8dce3aba90..d3afef95b4 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -161,10 +161,11 @@ #define PIPE_TEX_FACE_MAX 6 /** - * Surface flags + * Surfaces, textures, etc. (others may be added) */ -#define PIPE_SURFACE_FLAG_TEXTURE 0x1 -#define PIPE_SURFACE_FLAG_RENDER 0x2 +#define PIPE_TEXTURE 1 +#define PIPE_SURFACE 2 /**< user-created surfaces */ +#define PIPE_SCREEN_SURFACE 3 /**< On-screen front/back colorbuffer */ /** diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 809b165f45..8b8e04c2f9 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -46,17 +46,29 @@ /** - * Query format support. - * If we find texture and drawable support differs, add a selector - * parameter or another function. + * Query format support for creating a texture, drawing surface, etc. + * \param format the format to test + * \param type one of PIPE_TEXTURE, PIPE_SURFACE, PIPE_SCREEN_SURFACE */ static boolean softpipe_is_format_supported( struct pipe_context *pipe, - enum pipe_format format ) + enum pipe_format format, uint type ) { struct softpipe_context *softpipe = softpipe_context( pipe ); - /* ask winsys if the format is supported */ - return softpipe->winsys->is_format_supported( softpipe->winsys, format ); + + switch (type) { + case PIPE_TEXTURE: + /* softpipe supports all texture formats */ + return TRUE; + case PIPE_SURFACE: + /* softpipe supports all (off-screen) surface formats */ + return TRUE; + case PIPE_SCREEN_SURFACE: + return softpipe->winsys->is_format_supported( softpipe->winsys, format ); + default: + assert(0); + return FALSE; + } } diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h index 1912e59b9f..cbf64ebb85 100644 --- a/src/mesa/pipe/softpipe/sp_winsys.h +++ b/src/mesa/pipe/softpipe/sp_winsys.h @@ -25,21 +25,23 @@ * **************************************************************************/ +/* This is the interface that softpipe requires any window system + * hosting it to implement. This is the only include file in softpipe + * which is public. + */ + + #ifndef SP_WINSYS_H #define SP_WINSYS_H -#include "pipe/p_compiler.h" // for boolean +#include "pipe/p_compiler.h" /* for boolean */ -/* This is the interface that softpipe requires any window system - * hosting it to implement. This is the only include file in softpipe - * which is public. - */ - struct softpipe_winsys { + /** test if the given format is supported for front/back color bufs */ boolean (*is_format_supported)( struct softpipe_winsys *sws, - uint format ); + enum pipe_format format ); }; diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index c347d1c2a3..68f9c39116 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -71,7 +71,7 @@ struct xmesa_surface struct xmesa_softpipe_winsys { struct softpipe_winsys spws; - uint pixelformat; + enum pipe_format pixelformat; }; @@ -377,35 +377,17 @@ xmesa_get_pipe_winsys(void) /** + * Called via softpipe_winsys->is_format_supported(). + * This function is only called to test formats for front/back color surfaces. * The winsys being queried will have been created at glXCreateContext * time, with a pixel format corresponding to the context's visual. - * - * XXX we should pass a flag indicating if the format is going to be - * use for a drawing surface vs. a texture. In the later case, we - * can support any format. */ static boolean xmesa_is_format_supported(struct softpipe_winsys *sws, enum pipe_format format) { struct xmesa_softpipe_winsys *xmws = xmesa_softpipe_winsys(sws); - - if (format == xmws->pixelformat) { - return TRUE; - } - else { - /* non-color / window surface format */ - switch (format) { - case PIPE_FORMAT_R16G16B16A16_SNORM: - case PIPE_FORMAT_S8Z24_UNORM: - case PIPE_FORMAT_U_S8: - case PIPE_FORMAT_Z16_UNORM: - case PIPE_FORMAT_Z32_UNORM: - return TRUE; - default: - return FALSE; - } - } + return (format == xmws->pixelformat); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c0e41dbeec..0179000353 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -993,13 +993,13 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, int row, col; /* find a texture format we know */ - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8 )) { + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8, PIPE_TEXTURE )) { format = PIPE_FORMAT_U_I8; internal_format = GL_INTENSITY8; cpp = 1; comp = 0; } - else if (pipe->is_format_supported( pipe, PIPE_FORMAT_A8R8G8B8_UNORM )) { + else if (pipe->is_format_supported( pipe, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_TEXTURE )) { format = PIPE_FORMAT_A8R8G8B8_UNORM; internal_format = GL_RGBA8; cpp = 4; diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 36d25a576d..6e9e7e3a24 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -87,12 +87,14 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); + uint type = strb->screenSurface ? PIPE_SCREEN_SURFACE : PIPE_SURFACE; const enum pipe_format pipeFormat - = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE); + = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE, type); GLuint cpp; - GLbitfield flags = PIPE_SURFACE_FLAG_RENDER; /* want to render to surface */ + GLbitfield flags = 0x0; /* XXX needed? */ cpp = init_renderbuffer_bits(strb, pipeFormat); + assert(cpp); if (strb->surface && strb->surface->format != pipeFormat) { /* need to change surface types, free this surface */ @@ -201,9 +203,11 @@ st_new_renderbuffer(GLcontext *ctx, GLuint name) /** * Allocate a renderbuffer for a an on-screen window (not a user-created * renderbuffer). The window system code determines the internal format. + * \param screenSurface indicates if the renderbuffer is a front/back color + * buffer that'll be displayed/copied to the screen */ struct gl_renderbuffer * -st_new_renderbuffer_fb(GLenum intFormat) +st_new_renderbuffer_fb(GLenum intFormat, GLboolean screenSurface) { struct st_renderbuffer *strb; @@ -216,6 +220,7 @@ st_new_renderbuffer_fb(GLenum intFormat) _mesa_init_renderbuffer(&strb->Base, 0); strb->Base.ClassID = 0x4242; /* just a unique value */ strb->Base.InternalFormat = intFormat; + strb->screenSurface = screenSurface; switch (intFormat) { case GL_RGB5: diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index 2280441db5..bd85bfc549 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -39,6 +39,7 @@ struct st_renderbuffer { struct gl_renderbuffer Base; struct pipe_surface *surface; + GLboolean screenSurface; /**< A front/back colorbuffer? */ }; @@ -50,7 +51,7 @@ st_renderbuffer(struct gl_renderbuffer *rb) extern struct gl_renderbuffer * -st_new_renderbuffer_fb(GLenum intFormat); +st_new_renderbuffer_fb(GLenum intFormat, GLboolean screen_surface); extern void diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 8d39e1bec6..c292a975f3 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -276,10 +276,9 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) * Find an RGBA format supported by the context/winsys. */ static GLuint -default_rgba_format( - struct pipe_context *pipe ) +default_rgba_format(struct pipe_context *pipe, uint type) { - static const uint colorFormats[] = { + static const enum pipe_format colorFormats[] = { PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_B8G8R8A8_UNORM, @@ -287,7 +286,7 @@ default_rgba_format( }; uint i; for (i = 0; i < Elements(colorFormats); i++) { - if (pipe->is_format_supported( pipe, colorFormats[i] )) { + if (pipe->is_format_supported( pipe, colorFormats[i], type )) { return colorFormats[i]; } } @@ -299,10 +298,9 @@ default_rgba_format( * Search list of formats for first RGBA format with >8 bits/channel. */ static GLuint -default_deep_rgba_format( - struct pipe_context *pipe ) +default_deep_rgba_format(struct pipe_context *pipe, uint type) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_R16G16B16A16_SNORM )) { + if (pipe->is_format_supported(pipe, PIPE_FORMAT_R16G16B16A16_SNORM, type)) { return PIPE_FORMAT_R16G16B16A16_SNORM; } return PIPE_FORMAT_NONE; @@ -313,10 +311,9 @@ default_deep_rgba_format( * Find an Z format supported by the context/winsys. */ static GLuint -default_depth_format( - struct pipe_context *pipe ) +default_depth_format(struct pipe_context *pipe, uint type) { - static const uint zFormats[] = { + static const enum pipe_format zFormats[] = { PIPE_FORMAT_Z16_UNORM, PIPE_FORMAT_Z32_UNORM, PIPE_FORMAT_S8Z24_UNORM, @@ -324,7 +321,7 @@ default_depth_format( }; uint i; for (i = 0; i < Elements(zFormats); i++) { - if (pipe->is_format_supported( pipe, zFormats[i] )) { + if (pipe->is_format_supported( pipe, zFormats[i], type )) { return zFormats[i]; } } @@ -348,67 +345,71 @@ default_depth_format( */ GLuint st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, - GLenum format, GLenum type) + GLenum format, GLenum type, uint surfType) { + assert(surfType == PIPE_TEXTURE || + surfType == PIPE_SURFACE || + surfType == PIPE_SCREEN_SURFACE); + switch (internalFormat) { case 4: case GL_RGBA: case GL_COMPRESSED_RGBA: if (format == GL_BGRA) { if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A8R8G8B8_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A8R8G8B8_UNORM, surfType )) return PIPE_FORMAT_A8R8G8B8_UNORM; } else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A4R4G4B4_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A4R4G4B4_UNORM, surfType )) return PIPE_FORMAT_A4R4G4B4_UNORM; } else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) return PIPE_FORMAT_A1R5G5B5_UNORM; } } - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case 3: case GL_RGB: case GL_COMPRESSED_RGB: if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_R5G6B5_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_R5G6B5_UNORM, surfType )) return PIPE_FORMAT_R5G6B5_UNORM; } - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case GL_RGBA16: - return default_deep_rgba_format( pipe ); + return default_deep_rgba_format( pipe, surfType ); case GL_RGBA4: case GL_RGBA2: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A4R4G4B4_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A4R4G4B4_UNORM, surfType )) return PIPE_FORMAT_A4R4G4B4_UNORM; - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case GL_RGB5_A1: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) return PIPE_FORMAT_A1R5G5B5_UNORM; - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) return PIPE_FORMAT_A1R5G5B5_UNORM; - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case GL_ALPHA: case GL_ALPHA4: @@ -416,9 +417,9 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8, surfType )) return PIPE_FORMAT_U_A8; - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case 1: case GL_LUMINANCE: @@ -427,9 +428,9 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8, surfType )) return PIPE_FORMAT_U_A8; - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case 2: case GL_LUMINANCE_ALPHA: @@ -440,9 +441,9 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_L8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_L8, surfType )) return PIPE_FORMAT_U_A8_L8; - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case GL_INTENSITY: case GL_INTENSITY4: @@ -450,17 +451,17 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8, surfType )) return PIPE_FORMAT_U_I8; - return default_rgba_format( pipe ); + return default_rgba_format( pipe, surfType ); case GL_YCBCR_MESA: if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_YCBCR )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_YCBCR, surfType )) return PIPE_FORMAT_YCBCR; } else { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_YCBCR_REV )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_YCBCR_REV, surfType )) return PIPE_FORMAT_YCBCR_REV; } return PIPE_FORMAT_NONE; @@ -489,40 +490,40 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, #endif case GL_DEPTH_COMPONENT16: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z16_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z16_UNORM, surfType )) return PIPE_FORMAT_Z16_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT24: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM, surfType )) return PIPE_FORMAT_S8Z24_UNORM; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM, surfType )) return PIPE_FORMAT_Z24S8_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT32: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z32_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z32_UNORM, surfType )) return PIPE_FORMAT_Z32_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT: - return default_depth_format( pipe ); + return default_depth_format( pipe, surfType ); case GL_STENCIL_INDEX: case GL_STENCIL_INDEX1_EXT: case GL_STENCIL_INDEX4_EXT: case GL_STENCIL_INDEX8_EXT: case GL_STENCIL_INDEX16_EXT: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_S8 )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_S8, surfType )) return PIPE_FORMAT_U_S8; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM, surfType )) return PIPE_FORMAT_S8Z24_UNORM; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM, surfType )) return PIPE_FORMAT_Z24S8_UNORM; return PIPE_FORMAT_NONE; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM, surfType )) return PIPE_FORMAT_S8Z24_UNORM; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM, surfType )) return PIPE_FORMAT_Z24S8_UNORM; return PIPE_FORMAT_NONE; diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index 6ccf5536f9..ebff7c5b91 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -65,7 +65,7 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat); extern GLuint st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, - GLenum format, GLenum type); + GLenum format, GLenum type, uint surfType); extern const struct gl_texture_format * diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 8633f431b2..454306b874 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -51,19 +51,21 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, /* fake frontbuffer */ /* XXX allocation should only happen in the unusual case it's actually needed */ - struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(rgbFormat, GL_TRUE); _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb); } if (visual->doubleBufferMode) { - struct gl_renderbuffer *rb = st_new_renderbuffer_fb(rgbFormat); + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(rgbFormat, GL_TRUE); _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); } if (visual->depthBits == 24 && visual->stencilBits == 8) { /* combined depth/stencil buffer */ struct gl_renderbuffer *depthStencilRb - = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT, GL_FALSE); /* note: bind RB to two attachment points */ _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb); @@ -74,47 +76,36 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, if (visual->depthBits == 32) { /* 32-bit depth buffer */ struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32); + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32, GL_FALSE); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } else if (visual->depthBits == 24) { /* 24-bit depth buffer, ignore stencil bits */ struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); + = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT, GL_FALSE); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } else if (visual->depthBits > 0) { /* 16-bit depth buffer */ struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); + = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16, GL_FALSE); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } if (visual->stencilBits > 0) { /* 8-bit stencil */ struct gl_renderbuffer *stencilRb - = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT); + = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT, GL_FALSE); _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb); } } -#if 0 if (visual->accumRedBits > 0) { /* 16-bit/channel accum */ struct gl_renderbuffer *accumRb - = st_new_renderbuffer_fb(GL_RGBA16); + = st_new_renderbuffer_fb(GL_RGBA16, GL_FALSE); _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb); } -#endif - - /* now add any/all software-based renderbuffers we may need */ - _mesa_add_soft_renderbuffers(&stfb->Base, - GL_FALSE, /* never sw color */ - GL_FALSE, /* never sw depth */ - GL_FALSE, /* stencil */ - visual->accumRedBits > 0, - GL_FALSE, /* never sw alpha */ - GL_FALSE /* never sw aux */ ); } stfb->Base.Initialized = GL_TRUE; -- cgit v1.2.3 From e248f940506a678acc0cad1c925c0b11cca09672 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Dec 2007 14:19:35 -0700 Subject: remove ~ backup files --- src/mesa/pipe/cell/spu/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index 700a6c61cd..8e606dd1a2 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -45,7 +45,7 @@ tri.o: tri.c clean: - rm -f *.o *.a *.d $(PROG_SPU) + rm -f *~ *.o *.a *.d $(PROG_SPU) -- cgit v1.2.3 From e53303ba3b69c2c82cefd58e90d06132c2af2bb7 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Dec 2007 14:25:30 -0700 Subject: Cell driver state-setter functions, basic tile get/put, glClear. The state setting code was mostly just copied from the softpipe driver. The SPUs can now get/put framebuffer tiles from/to main memory and clear them to a given color. Lots of debug code in effect. Tiled framebuffer is displayed in X window via the xmwinsys layer. To enable Cell driver, export GALLIUM_CELL=1 --- src/mesa/pipe/cell/common.h | 89 ++++++++- src/mesa/pipe/cell/ppu/Makefile | 15 +- src/mesa/pipe/cell/ppu/cell_context.c | 237 +++++++++++++++++++++--- src/mesa/pipe/cell/ppu/cell_context.h | 49 ++++- src/mesa/pipe/cell/ppu/cell_flush.c | 54 ++++++ src/mesa/pipe/cell/ppu/cell_flush.h | 35 ++++ src/mesa/pipe/cell/ppu/cell_spu.c | 223 +++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_spu.h | 81 +++++++++ src/mesa/pipe/cell/ppu/cell_state.h | 117 ++++++++++++ src/mesa/pipe/cell/ppu/cell_state_blend.c | 128 +++++++++++++ src/mesa/pipe/cell/ppu/cell_state_clip.c | 84 +++++++++ src/mesa/pipe/cell/ppu/cell_state_fs.c | 195 ++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_state_rasterizer.c | 65 +++++++ src/mesa/pipe/cell/ppu/cell_state_sampler.c | 100 ++++++++++ src/mesa/pipe/cell/ppu/cell_state_surface.c | 99 ++++++++++ src/mesa/pipe/cell/ppu/cell_state_vertex.c | 63 +++++++ src/mesa/pipe/cell/ppu/cell_surface.c | 99 +++++++++- src/mesa/pipe/cell/ppu/cell_surface.h | 49 +++++ src/mesa/pipe/cell/ppu/cell_winsys.c | 40 ++++ src/mesa/pipe/cell/ppu/cell_winsys.h | 50 +++++ src/mesa/pipe/cell/spu/main.c | 243 ++++++++++++++++++++++++- src/mesa/pipe/xlib/xm_winsys.c | 132 ++++++++++++-- 22 files changed, 2197 insertions(+), 50 deletions(-) create mode 100644 src/mesa/pipe/cell/ppu/cell_flush.c create mode 100644 src/mesa/pipe/cell/ppu/cell_flush.h create mode 100644 src/mesa/pipe/cell/ppu/cell_spu.c create mode 100644 src/mesa/pipe/cell/ppu/cell_spu.h create mode 100644 src/mesa/pipe/cell/ppu/cell_state.h create mode 100644 src/mesa/pipe/cell/ppu/cell_state_blend.c create mode 100644 src/mesa/pipe/cell/ppu/cell_state_clip.c create mode 100644 src/mesa/pipe/cell/ppu/cell_state_fs.c create mode 100644 src/mesa/pipe/cell/ppu/cell_state_rasterizer.c create mode 100644 src/mesa/pipe/cell/ppu/cell_state_sampler.c create mode 100644 src/mesa/pipe/cell/ppu/cell_state_surface.c create mode 100644 src/mesa/pipe/cell/ppu/cell_state_vertex.c create mode 100644 src/mesa/pipe/cell/ppu/cell_surface.h create mode 100644 src/mesa/pipe/cell/ppu/cell_winsys.c create mode 100644 src/mesa/pipe/cell/ppu/cell_winsys.h (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index c4bad4194a..da7de78803 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -1,13 +1,94 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Types and tokens which are common to the SPU and PPU code. + */ + #ifndef CELL_COMMON_H #define CELL_COMMON_H +#include "pipe/p_util.h" + + +#define ALIGN16 __attribute__( (aligned( 16 )) ) + +#define ASSERT_ALIGN16(ptr) \ + assert((((unsigned long) (ptr)) & 0xf) == 0); + + + +#define TILE_SIZE 32 + + +#define CELL_CMD_EXIT 1 +#define CELL_CMD_FRAMEBUFFER 2 +#define CELL_CMD_CLEAR_TILES 3 +#define CELL_CMD_INVERT_TILES 4 +#define CELL_CMD_FINISH 5 + -struct init_info +/** + * Tell SPUs about the framebuffer size, location + */ +struct cell_command_framebuffer { - int foo; - int bar; -}; + void *start; + int width, height; + unsigned format; +} ALIGN16; + + +/** + * Clear framebuffer tiles to given value/color. + */ +struct cell_command_clear_tiles +{ + uint value; +} ALIGN16; + + +/** XXX unions don't seem to work */ +struct cell_command +{ + struct cell_command_framebuffer fb; + struct cell_command_clear_tiles clear; +} ALIGN16; + + +struct cell_init_info +{ + unsigned id; + unsigned num_spus; + struct cell_command *cmd; +} ALIGN16; + + #endif /* CELL_COMMON_H */ diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index c987d54e02..f597784d65 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -17,7 +17,18 @@ SPU_CODE_MODULE = ../spu/g3d_spu.a SOURCES = \ cell_context.c \ - cell_surface.c + cell_flush.c \ + cell_state_blend.c \ + cell_state_clip.c \ + cell_state_fs.c \ + cell_state_rasterizer.c \ + cell_state_sampler.c \ + cell_state_surface.c \ + cell_state_vertex.c \ + cell_spu.c \ + cell_surface.c \ + cell_winsys.c + OBJECTS = $(SOURCES:.c=.o) \ @@ -42,7 +53,7 @@ $(CELL_LIB): $(OBJECTS) $(SPU_CODE_MODULE) clean: - rm -f *.o $(CELL_LIB) + rm -f *.o *~ $(CELL_LIB) diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index a8f6cba2fa..68543ecf30 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -32,45 +32,232 @@ #include -#include -#include -#include "pipe/cell/ppu/cell_context.h" + +#include "pipe/p_defines.h" +#include "pipe/p_format.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" #include "pipe/cell/common.h" +#include "pipe/draw/draw_context.h" +#include "cell_context.h" +#include "cell_flush.h" +#include "cell_state.h" +#include "cell_surface.h" +#include "cell_spu.h" -#define NUM_SPUS 6 -extern spe_program_handle_t g3d_spu; +static boolean +cell_is_format_supported( struct pipe_context *pipe, + enum pipe_format format, uint type ) +{ + struct cell_context *cell = cell_context( pipe ); -static speid_t speid[NUM_SPUS]; -static struct init_info inits[NUM_SPUS]; + switch (type) { + case PIPE_TEXTURE: + /* cell supports all texture formats, XXX for now anyway */ + return TRUE; + case PIPE_SURFACE: + /* cell supports all (off-screen) surface formats, XXX for now */ + return TRUE; + case PIPE_SCREEN_SURFACE: + return format == cell->winsys->preferredFormat; + default: + assert(0); + return FALSE; + } +} -static void -start_spus(void) +static int cell_get_param(struct pipe_context *pipe, int param) { - int i; - - for (i = 0; i < NUM_SPUS; i++) { - inits[i].foo = i; - inits[i].bar = i * 10; - - speid[i] = spe_create_thread(0, /* gid */ - &g3d_spu, /* spe program handle */ - &inits[i], /* argp */ - NULL, /* envp */ - -1, /* mask */ - 0 ); /* flags */ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 1; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 1; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 1; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 12; /* max 2Kx2K */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 12; /* max 2Kx2K */ + default: + return 0; + } +} + +static float cell_get_paramf(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 0.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; /* arbitrary */ + + default: + return 0; } } +static const char * +cell_get_name( struct pipe_context *pipe ) +{ + return "Cell"; +} + +static const char * +cell_get_vendor( struct pipe_context *pipe ) +{ + return "Tungsten Graphics, Inc."; +} + + + +static void +cell_destroy_context( struct pipe_context *pipe ) +{ + struct cell_context *cell = cell_context(pipe); + + cell_spu_exit(cell); + wait_spus(cell->num_spus); + + free(cell); +} + + -void cell_create_context(void) + +struct pipe_context * +cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) { - printf("cell_create_context\n"); + struct cell_context *cell; + + cell = CALLOC_STRUCT(cell_context); + if (!cell) + return NULL; + + cell->winsys = cws; + cell->pipe.winsys = winsys; + cell->pipe.destroy = cell_destroy_context; + + /* queries */ + cell->pipe.is_format_supported = cell_is_format_supported; + cell->pipe.get_name = cell_get_name; + cell->pipe.get_vendor = cell_get_vendor; + cell->pipe.get_param = cell_get_param; + cell->pipe.get_paramf = cell_get_paramf; + + + /* state setters */ + cell->pipe.create_alpha_test_state = cell_create_alpha_test_state; + cell->pipe.bind_alpha_test_state = cell_bind_alpha_test_state; + cell->pipe.delete_alpha_test_state = cell_delete_alpha_test_state; + + cell->pipe.create_blend_state = cell_create_blend_state; + cell->pipe.bind_blend_state = cell_bind_blend_state; + cell->pipe.delete_blend_state = cell_delete_blend_state; + + cell->pipe.create_sampler_state = cell_create_sampler_state; + cell->pipe.bind_sampler_state = cell_bind_sampler_state; + cell->pipe.delete_sampler_state = cell_delete_sampler_state; + + cell->pipe.create_depth_stencil_state = cell_create_depth_stencil_state; + cell->pipe.bind_depth_stencil_state = cell_bind_depth_stencil_state; + cell->pipe.delete_depth_stencil_state = cell_delete_depth_stencil_state; + + cell->pipe.create_rasterizer_state = cell_create_rasterizer_state; + cell->pipe.bind_rasterizer_state = cell_bind_rasterizer_state; + cell->pipe.delete_rasterizer_state = cell_delete_rasterizer_state; + + cell->pipe.create_fs_state = cell_create_fs_state; + cell->pipe.bind_fs_state = cell_bind_fs_state; + cell->pipe.delete_fs_state = cell_delete_fs_state; + + cell->pipe.create_vs_state = cell_create_vs_state; + cell->pipe.bind_vs_state = cell_bind_vs_state; + cell->pipe.delete_vs_state = cell_delete_vs_state; + + cell->pipe.set_blend_color = cell_set_blend_color; + cell->pipe.set_clip_state = cell_set_clip_state; + cell->pipe.set_clear_color_state = cell_set_clear_color_state; + cell->pipe.set_constant_buffer = cell_set_constant_buffer; +#if 0 + cell->pipe.set_feedback_state = cell_set_feedback_state; +#endif + + cell->pipe.set_framebuffer_state = cell_set_framebuffer_state; + + cell->pipe.set_polygon_stipple = cell_set_polygon_stipple; + cell->pipe.set_sampler_units = cell_set_sampler_units; + cell->pipe.set_scissor_state = cell_set_scissor_state; + cell->pipe.set_texture_state = cell_set_texture_state; + cell->pipe.set_viewport_state = cell_set_viewport_state; + + cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; + cell->pipe.set_vertex_element = cell_set_vertex_element; +#if 0 + cell->pipe.set_feedback_buffer = cell_set_feedback_buffer; + + cell->pipe.draw_arrays = cell_draw_arrays; + cell->pipe.draw_elements = cell_draw_elements; +#endif + + cell->pipe.clear = cell_clear_surface; + cell->pipe.flush = cell_flush; + +#if 0 + cell->pipe.begin_query = cell_begin_query; + cell->pipe.end_query = cell_end_query; + cell->pipe.wait_query = cell_wait_query; + + /* textures */ + cell->pipe.mipmap_tree_layout = cell_mipmap_tree_layout; + cell->pipe.get_tex_surface = cell_get_tex_surface; +#endif + + + cell->draw = draw_create(); + + /* + * SPU stuff + */ + cell->num_spus = 6; /* XXX >6 seems to fail */ + + cell_start_spus(cell->num_spus); - start_spus(); +#if 0 + test_spus(cell); + wait_spus(); +#endif - /* TODO: do something with the SPUs! */ + return &cell->pipe; } diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index b4d93d1414..a64692081d 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -31,16 +31,61 @@ #include "pipe/p_context.h" - +#include "cell_winsys.h" struct cell_context { struct pipe_context pipe; - int spu_info; + struct cell_winsys *winsys; + + const struct pipe_alpha_test_state *alpha_test; + const struct pipe_blend_state *blend; + const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct pipe_depth_stencil_state *depth_stencil; + const struct pipe_rasterizer_state *rasterizer; + + struct pipe_blend_color blend_color; + struct pipe_clear_color_state clear_color; + struct pipe_clip_state clip; + struct pipe_constant_buffer constants[2]; + struct pipe_feedback_state feedback; + struct pipe_framebuffer_state framebuffer; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct softpipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + uint sampler_units[PIPE_MAX_SAMPLERS]; + uint dirty; + + /** The primitive drawing context */ + struct draw_context *draw; + struct draw_stage *setup; + struct draw_stage *vbuf; + + + uint num_spus; + + }; +static INLINE struct cell_context * +cell_context(struct pipe_context *pipe) +{ + return (struct cell_context *) pipe; +} + + +extern struct pipe_context * +cell_create_context(struct pipe_winsys *ws, struct cell_winsys *cws); + + + + + #endif /* CELL_CONTEXT_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c new file mode 100644 index 0000000000..e844d13f06 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -0,0 +1,54 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include + +#include "cell_context.h" +#include "cell_flush.h" +#include "cell_spu.h" + + +void +cell_flush(struct pipe_context *pipe, unsigned flags) +{ + struct cell_context *cell = cell_context(pipe); + uint i; + + printf("%s\n", __FUNCTION__); + + /* Send CMD_FINISH to all SPUs */ + for (i = 0; i < cell->num_spus; i++) { + send_mbox_message(control_ps_area[i], CELL_CMD_FINISH); + } + + /* Wait for ack */ + for (i = 0; i < cell->num_spus; i++) { + uint k = wait_mbox_message(control_ps_area[i]); + assert(k == CELL_CMD_FINISH); + } +} diff --git a/src/mesa/pipe/cell/ppu/cell_flush.h b/src/mesa/pipe/cell/ppu/cell_flush.h new file mode 100644 index 0000000000..cf1a104f97 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_flush.h @@ -0,0 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_FLUSH +#define CELL_FLUSH + +extern void +cell_flush(struct pipe_context *pipe, unsigned flags); + +#endif diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c new file mode 100644 index 0000000000..ed56250ff1 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -0,0 +1,223 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include + +#include "cell_spu.h" +#include "pipe/p_format.h" +#include "pipe/p_state.h" +#include "pipe/cell/common.h" + + +/* +helpful headers: +/opt/ibm/cell-sdk/prototype/src/include/ppu/cbe_mfc.h +*/ + + +/** + * SPU/SPE handles, etc + */ +speid_t speid[MAX_SPUS]; +spe_spu_control_area_t *control_ps_area[MAX_SPUS]; + + +/** + * Data sent to SPUs + */ +struct cell_init_info inits[MAX_SPUS] ALIGN16; +struct cell_command command[MAX_SPUS] ALIGN16; + + +/** + * Write a 1-word message to the given SPE mailbox. + */ +void +send_mbox_message(spe_spu_control_area_t *ca, unsigned int msg) +{ + while (_spe_in_mbox_status(ca) < 1) + ; + _spe_in_mbox_write(ca, msg); +} + + +/** + * Wait for a 1-word message to arrive in given mailbox. + */ +uint +wait_mbox_message(spe_spu_control_area_t *ca) +{ + uint k; + while (_spe_out_mbox_status(ca) < 1) + ; + k = _spe_out_mbox_read(ca); + return k; +} + + +/** + * Create the SPU threads + */ +void +cell_start_spus(uint num_spus) +{ + uint i; + + assert((sizeof(struct cell_command) & 0xf) == 0); + ASSERT_ALIGN16(&command[0]); + ASSERT_ALIGN16(&command[1]); + + assert((sizeof(struct cell_init_info) & 0xf) == 0); + ASSERT_ALIGN16(&inits[0]); + ASSERT_ALIGN16(&inits[1]); + + /* XXX do we need to create a gid with spe_create_group()? */ + + for (i = 0; i < num_spus; i++) { + inits[i].id = i; + inits[i].num_spus = num_spus; + inits[i].cmd = &command[i]; + + speid[i] = spe_create_thread(0, /* gid */ + &g3d_spu, /* spe program handle */ + &inits[i], /* argp */ + NULL, /* envp */ + -1, /* mask */ + SPE_MAP_PS/*0*/ ); /* flags */ + + control_ps_area[i] = spe_get_ps_area(speid[i], SPE_CONTROL_AREA); + assert(control_ps_area[i]); + } +} + + +/** wait for all SPUs to finish working */ +/** XXX temporary */ +void +finish_all(uint num_spus) +{ + uint i; + + for (i = 0; i < num_spus; i++) { + send_mbox_message(control_ps_area[i], CELL_CMD_FINISH); + } + for (i = 0; i < num_spus; i++) { + /* wait for mbox message */ + unsigned k; + while (_spe_out_mbox_status(control_ps_area[i]) < 1) + ; + k = _spe_out_mbox_read(control_ps_area[i]); + assert(k == CELL_CMD_FINISH); + } +} + + +/** + ** Send test commands (XXX temporary) + **/ +void +test_spus(struct cell_context *cell) +{ + uint i; + struct pipe_surface *surf = cell->framebuffer.cbufs[0]; + + printf("PPU: sleep(2)\n\n\n"); + sleep(2); + + for (i = 0; i < cell->num_spus; i++) { + command[i].fb.start = surf->map; + command[i].fb.width = surf->width; + command[i].fb.height = surf->height; + command[i].fb.format = PIPE_FORMAT_A8R8G8B8_UNORM; + send_mbox_message(control_ps_area[i], CELL_CMD_FRAMEBUFFER); + } + + for (i = 0; i < cell->num_spus; i++) { + command[i].clear.value = 0xff880044; /* XXX */ + send_mbox_message(control_ps_area[i], CELL_CMD_CLEAR_TILES); + } + + finish_all(cell->num_spus); + + { + uint *b = (uint*) surf->map; + printf("PPU: Clear results: 0x%x 0x%x 0x%x 0x%x\n", + b[0], b[1000], b[2000], b[3000]); + } + + for (i = 0; i < cell->num_spus; i++) { + send_mbox_message(control_ps_area[i], CELL_CMD_INVERT_TILES); + } + + finish_all(cell->num_spus); + + { + uint *b = (uint*) surf->map; + printf("PPU: Inverted results: 0x%x 0x%x 0x%x 0x%x\n", + b[0], b[1000], b[2000], b[3000]); + } + + + + for (i = 0; i < cell->num_spus; i++) { + send_mbox_message(control_ps_area[i], CELL_CMD_EXIT); + } +} + + +/** + * Wait for all SPUs to exit/return. + */ +void +wait_spus(uint num_spus) +{ + int i, status; + + for (i = 0; i < num_spus; i++) { + spe_wait( speid[i], &status, 1 ); + } +} + + +/** + * Tell all the SPUs to stop/exit. + */ +void +cell_spu_exit(struct cell_context *cell) +{ + uint i; + int status; + + for (i = 0; i < cell->num_spus; i++) { + send_mbox_message(control_ps_area[i], CELL_CMD_EXIT); + } + + for (i = 0; i < cell->num_spus; i++) { + spe_wait( speid[i], &status, 1 ); + } +} diff --git a/src/mesa/pipe/cell/ppu/cell_spu.h b/src/mesa/pipe/cell/ppu/cell_spu.h new file mode 100644 index 0000000000..dcbc72573f --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_spu.h @@ -0,0 +1,81 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_SPU +#define CELL_SPU + + +#include +#include +#include "pipe/cell/common.h" + +#include "cell_context.h" + + +#define MAX_SPUS 7 + +/** + * SPU/SPE handles, etc + */ +extern spe_program_handle_t g3d_spu; +extern speid_t speid[MAX_SPUS]; +extern spe_spu_control_area_t *control_ps_area[MAX_SPUS]; + + +/** + * Data sent to SPUs + */ +extern struct cell_init_info inits[MAX_SPUS] ALIGN16; +extern struct cell_command command[MAX_SPUS] ALIGN16; + + +void +send_mbox_message(spe_spu_control_area_t *ca, unsigned int msg); + +uint +wait_mbox_message(spe_spu_control_area_t *ca); + + +void +cell_start_spus(uint num_spus); + + +void +finish_all(uint num_spus); + +void +test_spus(struct cell_context *cell); + + +void +wait_spus(uint num_spus); + +void +cell_spu_exit(struct cell_context *cell); + + +#endif /* CELL_SPU */ diff --git a/src/mesa/pipe/cell/ppu/cell_state.h b/src/mesa/pipe/cell/ppu/cell_state.h new file mode 100644 index 0000000000..033767d29b --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state.h @@ -0,0 +1,117 @@ + + +#ifndef CELL_STATE_H +#define CELL_STATE_H + + +#define CELL_NEW_VIEWPORT 0x1 +#define CELL_NEW_RASTERIZER 0x2 +#define CELL_NEW_FS 0x4 +#define CELL_NEW_BLEND 0x8 +#define CELL_NEW_CLIP 0x10 +#define CELL_NEW_SCISSOR 0x20 +#define CELL_NEW_STIPPLE 0x40 +#define CELL_NEW_FRAMEBUFFER 0x80 +#define CELL_NEW_ALPHA_TEST 0x100 +#define CELL_NEW_DEPTH_STENCIL 0x200 +#define CELL_NEW_SAMPLER 0x400 +#define CELL_NEW_TEXTURE 0x800 +#define CELL_NEW_VERTEX 0x1000 +#define CELL_NEW_VS 0x2000 +#define CELL_NEW_CONSTANTS 0x4000 + + + +extern void +cell_set_framebuffer_state( struct pipe_context *, + const struct pipe_framebuffer_state * ); + + +extern void * +cell_create_alpha_test_state(struct pipe_context *, + const struct pipe_alpha_test_state *); +extern void +cell_bind_alpha_test_state(struct pipe_context *, void *); +extern void +cell_delete_alpha_test_state(struct pipe_context *, void *); + +extern void * +cell_create_blend_state(struct pipe_context *, const struct pipe_blend_state *); +extern void cell_bind_blend_state(struct pipe_context *, void *); +extern void cell_delete_blend_state(struct pipe_context *, void *); + +extern void cell_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ); + + +void * +cell_create_sampler_state(struct pipe_context *, + const struct pipe_sampler_state *); + +extern void +cell_bind_sampler_state(struct pipe_context *, unsigned, void *); + +extern void +cell_delete_sampler_state(struct pipe_context *, void *); + + +extern void * +cell_create_depth_stencil_state(struct pipe_context *, + const struct pipe_depth_stencil_state *); + +extern void +cell_bind_depth_stencil_state(struct pipe_context *, void *); + +extern void +cell_delete_depth_stencil_state(struct pipe_context *, void *); + + +void *cell_create_fs_state(struct pipe_context *, + const struct pipe_shader_state *); +void cell_bind_fs_state(struct pipe_context *, void *); +void cell_delete_fs_state(struct pipe_context *, void *); +void *cell_create_vs_state(struct pipe_context *, + const struct pipe_shader_state *); +void cell_bind_vs_state(struct pipe_context *, void *); +void cell_delete_vs_state(struct pipe_context *, void *); + + +void * +cell_create_rasterizer_state(struct pipe_context *, + const struct pipe_rasterizer_state *); +void cell_bind_rasterizer_state(struct pipe_context *, void *); +void cell_delete_rasterizer_state(struct pipe_context *, void *); + + +void cell_set_clip_state( struct pipe_context *, + const struct pipe_clip_state * ); + +void cell_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf); + +void cell_set_polygon_stipple( struct pipe_context *, + const struct pipe_poly_stipple * ); + +void cell_set_sampler_units( struct pipe_context *, + uint numSamplers, const uint *units ); + +void cell_set_scissor_state( struct pipe_context *, + const struct pipe_scissor_state * ); + +void cell_set_texture_state( struct pipe_context *, + unsigned unit, struct pipe_texture * ); + +void cell_set_vertex_element(struct pipe_context *, + unsigned index, + const struct pipe_vertex_element *); + +void cell_set_vertex_buffer(struct pipe_context *, + unsigned index, + const struct pipe_vertex_buffer *); + +void cell_set_viewport_state( struct pipe_context *, + const struct pipe_viewport_state * ); + + +#endif diff --git a/src/mesa/pipe/cell/ppu/cell_state_blend.c b/src/mesa/pipe/cell/ppu/cell_state_blend.c new file mode 100644 index 0000000000..e807463d90 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_blend.c @@ -0,0 +1,128 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "cell_context.h" +#include "cell_state.h" + +void * +cell_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + struct pipe_blend_state *state = MALLOC( sizeof(struct pipe_blend_state) ); + memcpy(state, blend, sizeof(struct pipe_blend_state)); + return state; +} + +void cell_bind_blend_state( struct pipe_context *pipe, + void *blend ) +{ + struct cell_context *cell = cell_context(pipe); + + cell->blend = (const struct pipe_blend_state *)blend; + + cell->dirty |= CELL_NEW_BLEND; +} + +void cell_delete_blend_state(struct pipe_context *pipe, + void *blend) +{ + FREE( blend ); +} + + +void cell_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct cell_context *cell = cell_context(pipe); + + cell->blend_color = *blend_color; + + cell->dirty |= CELL_NEW_BLEND; +} + + +/** XXX move someday? Or consolidate all these simple state setters + * into one file. + */ + +void * +cell_create_alpha_test_state(struct pipe_context *pipe, + const struct pipe_alpha_test_state *alpha) +{ + struct pipe_alpha_test_state *state = MALLOC( sizeof(struct pipe_alpha_test_state) ); + memcpy(state, alpha, sizeof(struct pipe_alpha_test_state)); + return state; +} + +void +cell_bind_alpha_test_state(struct pipe_context *pipe, + void *alpha) +{ + struct cell_context *cell = cell_context(pipe); + + cell->alpha_test = (const struct pipe_alpha_test_state *)alpha; + + cell->dirty |= CELL_NEW_ALPHA_TEST; +} + +void +cell_delete_alpha_test_state(struct pipe_context *pipe, + void *alpha) +{ + FREE( alpha ); +} + +void * +cell_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *depth_stencil) +{ + struct pipe_depth_stencil_state *state = + MALLOC( sizeof(struct pipe_depth_stencil_state) ); + memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_state)); + return state; +} + +void +cell_bind_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct cell_context *cell = cell_context(pipe); + + cell->depth_stencil = (const struct pipe_depth_stencil_state *)depth_stencil; + + cell->dirty |= CELL_NEW_DEPTH_STENCIL; +} + +void +cell_delete_depth_stencil_state(struct pipe_context *pipe, void *depth) +{ + FREE( depth ); +} diff --git a/src/mesa/pipe/cell/ppu/cell_state_clip.c b/src/mesa/pipe/cell/ppu/cell_state_clip.c new file mode 100644 index 0000000000..4f43665941 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_clip.c @@ -0,0 +1,84 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "cell_context.h" +#include "cell_state.h" +#include "pipe/draw/draw_context.h" + + +void cell_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct cell_context *cell = cell_context(pipe); + + /* pass the clip state to the draw module */ + draw_set_clip_state(cell->draw, clip); +} + + + +/* Called when driver state tracker notices changes to the viewport + * matrix: + */ +void cell_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct cell_context *cell = cell_context(pipe); + + cell->viewport = *viewport; /* struct copy */ + cell->dirty |= CELL_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + draw_set_viewport_state(cell->draw, viewport); + + /* Using tnl/ and vf/ modules is temporary while getting started. + * Full pipe will have vertex shader, vertex fetch of its own. + */ +} + + +void cell_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct cell_context *cell = cell_context(pipe); + + memcpy( &cell->scissor, scissor, sizeof(*scissor) ); + cell->dirty |= CELL_NEW_SCISSOR; +} + + +void cell_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ + struct cell_context *cell = cell_context(pipe); + + memcpy( &cell->poly_stipple, stipple, sizeof(*stipple) ); + cell->dirty |= CELL_NEW_STIPPLE; +} diff --git a/src/mesa/pipe/cell/ppu/cell_state_fs.c b/src/mesa/pipe/cell/ppu/cell_state_fs.c new file mode 100644 index 0000000000..910f210cdc --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_fs.c @@ -0,0 +1,195 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#if 0 +#include "pipe/draw/draw_context.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/llvm/gallivm.h" +#include "pipe/tgsi/util/tgsi_dump.h" +#include "pipe/tgsi/exec/tgsi_sse2.h" +#endif + +#include "cell_context.h" +#include "cell_state.h" + + +void * cell_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct cell_context *cell = cell_context(pipe); + + return malloc(5); /* XXX temp */ + +#if 0 + /* Decide whether we'll be codegenerating this shader and if so do + * that now. + */ + + struct sp_fragment_shader_state *state = MALLOC( sizeof(struct sp_fragment_shader_state) ); + state->shader = *templ; + + if( cell->dump_fs ) { + tgsi_dump( + state->shader.tokens, + 0 ); + } + +#if defined(__i386__) || defined(__386__) + if (cell->use_sse) { + x86_init_func( &state->sse2_program ); + tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); + } +#endif + +#ifdef MESA_LLVM + state->llvm_prog = gallivm_from_tgsi(state->shader.tokens, GALLIVM_FS); + if (!gallivm_global_cpu_engine()) { + gallivm_cpu_engine_create(state->llvm_prog); + } + else + gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); +#endif + return state; +#endif +} + +void cell_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct cell_context *cell = cell_context(pipe); +#if 0 + cell->fs = (struct sp_fragment_shader_state *) fs; + + cell->dirty |= SP_NEW_FS; +#endif +} + +void cell_delete_fs_state(struct pipe_context *pipe, + void *shader) +{ +#if 0 + struct sp_fragment_shader_state *state = shader; + +#if defined(__i386__) || defined(__386__) + x86_release_func( &state->sse2_program ); +#endif + + FREE( state ); +#endif +} + + +void * +cell_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + return malloc(5); /* XXX */ +#if 0 + struct cell_context *cell = cell_context(pipe); + struct sp_vertex_shader_state *state; + + state = MALLOC( sizeof(struct sp_vertex_shader_state) ); + if (state == NULL ) { + return NULL; + } + + state->state = MALLOC( sizeof(struct pipe_shader_state) ); + if (state->state == NULL) { + FREE( state ); + return NULL; + } + memcpy( state->state, templ, sizeof(struct pipe_shader_state) ); + + state->draw_data = draw_create_vertex_shader(cell->draw, + state->state); + if (state->draw_data == NULL) { + FREE( state->state ); + FREE( state ); + return NULL; + } + + return state; +#endif +} + + + +void +cell_bind_vs_state(struct pipe_context *pipe, void *vs) +{ +#if 0 + struct cell_context *cell = cell_context(pipe); + + cell->vs = (const struct sp_vertex_shader_state *)vs; + + draw_bind_vertex_shader(cell->draw, cell->vs->draw_data); + + cell->dirty |= SP_NEW_VS; +#endif +} + +void +cell_delete_vs_state(struct pipe_context *pipe, void *vs) +{ +#if 0 + struct cell_context *cell = cell_context(pipe); + + struct sp_vertex_shader_state *state = + (struct sp_vertex_shader_state *)vs; + + draw_delete_vertex_shader(cell->draw, state->draw_data); + FREE( state->state ); + FREE( state ); +#endif +} + + + +void cell_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ +#if 0 + struct cell_context *cell = cell_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + /* note: reference counting */ + ws->buffer_reference(ws, + &cell->constants[shader].buffer, + buf->buffer); + cell->constants[shader].size = buf->size; + + cell->dirty |= SP_NEW_CONSTANTS; +#endif +} + + diff --git a/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c b/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c new file mode 100644 index 0000000000..11e7de7309 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c @@ -0,0 +1,65 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/draw/draw_context.h" +#include "cell_context.h" +#include "cell_state.h" + + +void * +cell_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup) +{ + struct pipe_rasterizer_state *state = + MALLOC( sizeof(struct pipe_rasterizer_state) ); + memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); + return state; +} + + +void +cell_bind_rasterizer_state(struct pipe_context *pipe, void *setup) +{ + struct cell_context *cell = cell_context(pipe); + + /* pass-through to draw module */ + draw_set_rasterizer_state(cell->draw, setup); + + cell->rasterizer = (struct pipe_rasterizer_state *)setup; + + cell->dirty |= CELL_NEW_RASTERIZER; +} + +void +cell_delete_rasterizer_state(struct pipe_context *pipe, void *rasterizer) +{ + FREE( rasterizer ); +} + + diff --git a/src/mesa/pipe/cell/ppu/cell_state_sampler.c b/src/mesa/pipe/cell/ppu/cell_state_sampler.c new file mode 100644 index 0000000000..c2a180ed30 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_sampler.c @@ -0,0 +1,100 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "cell_context.h" +#include "cell_state.h" +#if 0 +#include "cell_texture.h" +#include "cell_tile_cache.h" +#endif + + +void * +cell_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + struct pipe_sampler_state *state = MALLOC( sizeof(struct pipe_sampler_state) ); + memcpy(state, sampler, sizeof(struct pipe_sampler_state)); + return state; +} + +void +cell_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct cell_context *cell = cell_context(pipe); + + assert(unit < PIPE_MAX_SAMPLERS); + cell->sampler[unit] = (struct pipe_sampler_state *)sampler; + + cell->dirty |= CELL_NEW_SAMPLER; +} + + +void +cell_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + FREE( sampler ); +} + + +void +cell_set_texture_state(struct pipe_context *pipe, + unsigned unit, + struct pipe_texture *texture) +{ + struct cell_context *cell = cell_context(pipe); + + assert(unit < PIPE_MAX_SAMPLERS); + +#if 0 + cell->texture[unit] = cell_texture(texture); /* ptr, not struct */ + cell_tile_cache_set_texture(cell->tex_cache[unit], texture); +#endif + + cell->dirty |= CELL_NEW_TEXTURE; +} + + +void +cell_set_sampler_units(struct pipe_context *pipe, + uint num_samplers, const uint *units ) +{ + struct cell_context *cell = cell_context(pipe); + uint i; + for (i = 0; i < num_samplers; i++) + cell->sampler_units[i] = units[i]; + cell->dirty |= CELL_NEW_SAMPLER; +} + + diff --git a/src/mesa/pipe/cell/ppu/cell_state_surface.c b/src/mesa/pipe/cell/ppu/cell_state_surface.c new file mode 100644 index 0000000000..f7330caf5e --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_surface.c @@ -0,0 +1,99 @@ + + +#include "cell_context.h" +#include "cell_state.h" + +void +cell_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct cell_context *cell = cell_context(pipe); + + cell->framebuffer = *fb; + + cell->dirty |= CELL_NEW_FRAMEBUFFER; + +#if 0 + struct pipe_surface *ps; + uint i; + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + /* check if changing cbuf */ + if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) { + /* flush old */ + sp_flush_tile_cache(sp, sp->cbuf_cache[i]); + /* unmap old */ + ps = sp->framebuffer.cbufs[i]; + if (ps && ps->map) + pipe_surface_unmap(ps); + /* map new */ + ps = fb->cbufs[i]; + if (ps) + pipe_surface_map(ps); + /* assign new */ + sp->framebuffer.cbufs[i] = fb->cbufs[i]; + + /* update cache */ + sp_tile_cache_set_surface(sp, sp->cbuf_cache[i], ps); + } + } + + sp->framebuffer.num_cbufs = fb->num_cbufs; + + /* zbuf changing? */ + if (sp->framebuffer.zbuf != fb->zbuf) { + /* flush old */ + sp_flush_tile_cache(sp, sp->zbuf_cache); + /* unmap old */ + ps = sp->framebuffer.zbuf; + if (ps && ps->map) + pipe_surface_unmap(ps); + if (sp->framebuffer.sbuf == sp->framebuffer.zbuf) { + /* combined z/stencil */ + sp->framebuffer.sbuf = NULL; + } + /* map new */ + ps = fb->zbuf; + if (ps) + pipe_surface_map(ps); + /* assign new */ + sp->framebuffer.zbuf = fb->zbuf; + + /* update cache */ + sp_tile_cache_set_surface(sp, sp->zbuf_cache, ps); + } + + /* XXX combined depth/stencil here */ + + /* sbuf changing? */ + if (sp->framebuffer.sbuf != fb->sbuf) { + /* flush old */ + sp_flush_tile_cache(sp, sp->sbuf_cache_sep); + /* unmap old */ + ps = sp->framebuffer.sbuf; + if (ps && ps->map) + pipe_surface_unmap(ps); + /* map new */ + ps = fb->sbuf; + if (ps && fb->sbuf != fb->zbuf) + pipe_surface_map(ps); + /* assign new */ + sp->framebuffer.sbuf = fb->sbuf; + + /* update cache */ + if (fb->sbuf != fb->zbuf) { + /* separate stencil buf */ + sp->sbuf_cache = sp->sbuf_cache_sep; + sp_tile_cache_set_surface(sp, sp->sbuf_cache, ps); + } + else { + /* combined depth/stencil */ + sp->sbuf_cache = sp->zbuf_cache; + sp_tile_cache_set_surface(sp, sp->sbuf_cache, ps); + } + } + + sp->dirty |= SP_NEW_FRAMEBUFFER; +#endif +} + diff --git a/src/mesa/pipe/cell/ppu/cell_state_vertex.c b/src/mesa/pipe/cell/ppu/cell_state_vertex.c new file mode 100644 index 0000000000..0f01e920f9 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_vertex.c @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + + +#include "cell_context.h" +#include "cell_state.h" + +#include "pipe/draw/draw_context.h" + + +void +cell_set_vertex_element(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_element *attrib) +{ + struct cell_context *cell = cell_context(pipe); + assert(index < PIPE_ATTRIB_MAX); + cell->vertex_element[index] = *attrib; /* struct copy */ + cell->dirty |= CELL_NEW_VERTEX; + + draw_set_vertex_element(cell->draw, index, attrib); +} + + +void +cell_set_vertex_buffer(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_buffer *buffer) +{ + struct cell_context *cell = cell_context(pipe); + assert(index < PIPE_ATTRIB_MAX); + cell->vertex_buffer[index] = *buffer; /* struct copy */ + cell->dirty |= CELL_NEW_VERTEX; + + draw_set_vertex_buffer(cell->draw, index, buffer); +} diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 2c4b6e640b..1692960244 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -1,9 +1,106 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Authors + * Brian Paul + */ #include +#include +#include +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/cell/common.h" +#include "cell_context.h" +#include "cell_surface.h" +#include "cell_spu.h" + -void cell_create_surface(void) +struct pipe_surface * +cell_create_surface(int width, int height) { +#if 0 + /* XXX total hack */ + struct pipe_surface *ps = CALLOC_STRUCT(pipe_surface); + printf("cell_create_surface\n"); + ps->width = width; + ps->height = height; + + ps->region = CALLOC_STRUCT(pipe_region); + ps->region->map = align_malloc(width * height * 4, 16); + return ps; +#endif + return NULL; +} + + + +void +cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + struct cell_context *cell = cell_context(pipe); + uint i; + + printf("%s 0x%08x\n", __FUNCTION__, clearValue); + + { + char s[100]; + pf_sprint_name(s, ps->format); + printf("format = %s\n", s); + } + + if (!ps->map) + pipe_surface_map(ps); + + for (i = 0; i < cell->num_spus; i++) { + command[i].fb.start = ps->map; + command[i].fb.width = ps->width; + command[i].fb.height = ps->height; + command[i].fb.format = ps->format; + send_mbox_message(control_ps_area[i], CELL_CMD_FRAMEBUFFER); + } + + for (i = 0; i < cell->num_spus; i++) { + command[i].clear.value = clearValue | (i << 21); + send_mbox_message(control_ps_area[i], CELL_CMD_CLEAR_TILES); + } +} + + +void +cell_set_clear_color_state(struct pipe_context *pipe, + const struct pipe_clear_color_state *clear) +{ + struct cell_context *cell = cell_context(pipe); + + cell->clear_color = *clear; /* struct copy */ } diff --git a/src/mesa/pipe/cell/ppu/cell_surface.h b/src/mesa/pipe/cell/ppu/cell_surface.h new file mode 100644 index 0000000000..8b42ba02d5 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_surface.h @@ -0,0 +1,49 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_SURFACE +#define CELL_SURFACE + + +#include "pipe/p_state.h" + + +extern struct pipe_surface * +cell_create_surface(int width, int height); + +extern void +cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + + +extern void +cell_set_clear_color_state(struct pipe_context *pipe, + const struct pipe_clear_color_state *clear); + + +#endif /* CELL_SURFACE */ diff --git a/src/mesa/pipe/cell/ppu/cell_winsys.c b/src/mesa/pipe/cell/ppu/cell_winsys.c new file mode 100644 index 0000000000..ebabce3c8f --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_winsys.c @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "cell_winsys.h" + + +struct cell_winsys * +cell_get_winsys(uint format) +{ + struct cell_winsys *cws = CALLOC_STRUCT(cell_winsys); + if (cws) + cws->preferredFormat = format; + return cws; +} diff --git a/src/mesa/pipe/cell/ppu/cell_winsys.h b/src/mesa/pipe/cell/ppu/cell_winsys.h new file mode 100644 index 0000000000..ae2af5696b --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_winsys.h @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_WINSYS_H +#define CELL_WINSYS_H + +#include "pipe/p_compiler.h" + + +/** + * Very simple winsys at this time. + * Will probably eventually add SPU control info. + */ +struct cell_winsys +{ + uint preferredFormat; +}; + + +extern struct cell_winsys * +cell_get_winsys(uint format); + + + +#endif diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index e8d5fdccbf..226d81b4ca 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -1,15 +1,241 @@ -/* main.c for cell SPU code */ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/* main() for Cell SPU code */ #include +#include #include #include #include "tri.h" #include "pipe/cell/common.h" +/* +helpful headers: +/usr/lib/gcc/spu/4.1.1/include/spu_mfcio.h +/opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h +*/ + +static struct cell_init_info init; + +struct framebuffer { + void *start; + uint width, height; + uint width_tiles, height_tiles; /**< width and height in tiles */ +}; +static struct framebuffer fb; + + +static int DefaultTag = 1; + + + +static inline void +wait_on_mask(unsigned tag) +{ + mfc_write_tag_mask( tag ); + mfc_read_tag_status_any(); +} + + + +static void +get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile) +{ + uint offset = ty * fb->width_tiles + tx; + uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; + ubyte *src = (ubyte *) fb->start + offset * bytesPerTile; + int tag = DefaultTag; + + assert(tx < fb->width_tiles); + assert(ty < fb->height_tiles); + ASSERT_ALIGN16(tile); + /* + printf("get_tile: dest: %p src: 0x%x size: %d\n", + tile, (unsigned int) src, bytesPerTile); + */ + mfc_get(tile, /* dest in local memory */ + (unsigned int) src, /* src in main memory */ + bytesPerTile, + tag, + 0, /* tid */ + 0 /* rid */); +} + +static void +put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile) +{ + uint offset = ty * fb->width_tiles + tx; + uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; + ubyte *dst = (ubyte *) fb->start + offset * bytesPerTile; + int tag = DefaultTag; + + assert(tx < fb->width_tiles); + assert(ty < fb->height_tiles); + ASSERT_ALIGN16(tile); + /* + printf("put_tile: src: %p dst: 0x%x size: %d\n", + tile, (unsigned int) dst, bytesPerTile); + */ + mfc_put((void *) tile, /* src in local memory */ + (unsigned int) dst, /* dst in main mory */ + bytesPerTile, + tag, + 0, /* tid */ + 0 /* rid */); +} + + + +static void +clear_tiles(const struct cell_command_clear_tiles *clear) +{ + uint num_tiles = fb.width_tiles * fb.height_tiles; + uint i; + uint tile[TILE_SIZE * TILE_SIZE] ALIGN16; + + for (i = 0; i < TILE_SIZE * TILE_SIZE; i++) + tile[i] = clear->value; + + printf("SPU: %s num=%d w=%d h=%d\n", + __FUNCTION__, num_tiles, fb.width_tiles, fb.height_tiles); + for (i = init.id; i < num_tiles; i += init.num_spus) { + uint tx = i % fb.width_tiles; + uint ty = i / fb.width_tiles; + put_tile(&fb, tx, ty, tile); + /* XXX we don't want this here, but it fixes bad tile results */ + wait_on_mask(1 << DefaultTag); + } +} + + +/** Invert all pixels in all tiles */ +static void +invert_tiles(void) +{ + uint num_tiles = fb.width_tiles * fb.height_tiles; + uint i, j; + uint tile[TILE_SIZE * TILE_SIZE] ALIGN16; + + for (i = init.id; i < num_tiles; i += init.num_spus) { + uint tx = i % fb.width_tiles; + uint ty = i / fb.width_tiles; + + get_tile(&fb, tx, ty, tile); + wait_on_mask(1 << DefaultTag); + + for (j = 0; j < TILE_SIZE * TILE_SIZE; j++) { + tile[j] = ~tile[j]; + } + + put_tile(&fb, tx, ty, tile); + } +} + + +struct cell_command cmd ALIGN16; + + +/** + * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. + */ +static void +main_loop(void) +{ + int exitFlag = 0; + printf("SPU %u: Enter main loop\n", init.id); + + assert((sizeof(struct cell_command) & 0xf) == 0); + ASSERT_ALIGN16(&cmd); + + while (!exitFlag) { + unsigned opcode; + int tag = 0; + + printf("SPU %u: Wait for cmd...\n", init.id); + + /* read/wait from mailbox */ + opcode = (unsigned int) spu_read_in_mbox(); + + printf("SPU %u: got cmd %u\n", init.id, opcode); + + /* command payload */ + mfc_get(&cmd, /* dest */ + (unsigned int) init.cmd, /* src */ + sizeof(struct cell_command), /* bytes */ + tag, + 0, /* tid */ + 0 /* rid */); + wait_on_mask( 1 << tag ); + + switch (opcode) { + case CELL_CMD_EXIT: + printf("SPU %u: EXIT\n", init.id); + exitFlag = 1; + break; + case CELL_CMD_FRAMEBUFFER: + printf("SPU %u: FRAMEBUFFER: %d x %d at %p\n", init.id, + cmd.fb.width, + cmd.fb.height, + cmd.fb.start); + fb.width = cmd.fb.width; + fb.height = cmd.fb.height; + fb.width_tiles = fb.width / TILE_SIZE; + fb.height_tiles = fb.height / TILE_SIZE; + fb.start = cmd.fb.start; + break; + case CELL_CMD_CLEAR_TILES: + printf("SPU %u: CLEAR to 0x%08x\n", init.id, cmd.clear.value); + clear_tiles(&cmd.clear); + break; + case CELL_CMD_INVERT_TILES: + printf("SPU %u: INVERT_TILES\n", init.id); + invert_tiles(); + break; + case CELL_CMD_FINISH: + printf("SPU %u: FINISH\n", init.id); + /* wait for all outstanding DMAs to finish */ + mfc_write_tag_mask(~0); + mfc_read_tag_status_all(); + /* send mbox message to PPU */ + spu_write_out_mbox(CELL_CMD_FINISH); + break; + default: + printf("Bad opcode!\n"); + } + + } + + printf("SPU %u: Exit main loop\n", init.id); +} -static struct init_info init; int @@ -19,11 +245,18 @@ main(unsigned long long speid, { int tag = 0; - mfc_get(&init, (unsigned int) argp, sizeof(struct init_info), tag, 0, 0); + (void) speid; + (void) envp; - printf("Enter spu main(): init.foo=%d\n", init.foo); + mfc_get(&init, /* dest */ + (unsigned int) argp, /* src */ + sizeof(struct cell_init_info), /* bytes */ + tag, + 0, /* tid */ + 0 /* rid */); + wait_on_mask( 1 << tag ); - draw_triangle(0, 1, 2); + main_loop(); return 0; } diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 68f9c39116..ec171e3d9a 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -42,6 +42,52 @@ #include "pipe/p_context.h" #include "pipe/softpipe/sp_winsys.h" +#ifdef GALLIUM_CELL +#include "pipe/cell/ppu/cell_context.h" +#include "pipe/cell/ppu/cell_winsys.h" +#endif + + +/** XXX from Mesa core */ +static void * +align_malloc(size_t bytes, unsigned long alignment) +{ +#if defined(HAVE_POSIX_MEMALIGN) + void *mem; + + (void) posix_memalign(& mem, alignment, bytes); + return mem; +#else + uintptr_t ptr, buf; + + assert( alignment > 0 ); + + ptr = (uintptr_t) malloc(bytes + alignment + sizeof(void *)); + if (!ptr) + return NULL; + + buf = (ptr + alignment + sizeof(void *)) & ~(uintptr_t)(alignment - 1); + *(uintptr_t *)(buf - sizeof(void *)) = ptr; + + return (void *) buf; +#endif /* defined(HAVE_POSIX_MEMALIGN) */ +} + + +/** XXX from Mesa core */ +static void +align_free(void *ptr) +{ +#if defined(HAVE_POSIX_MEMALIGN) + free(ptr); +#else + void **cubbyHole = (void **) ((char *) ptr - sizeof(void *)); + void *realAddr = *cubbyHole; + free(realAddr); +#endif /* defined(HAVE_POSIX_MEMALIGN) */ +} + + /** * Low-level OS/window system memory buffer @@ -59,7 +105,8 @@ struct xm_buffer struct xmesa_surface { struct pipe_surface surface; - /* no extra fields for now */ + + int tileSize; }; @@ -138,7 +185,7 @@ xm_buffer_reference(struct pipe_winsys *pws, if (oldBuf->refcount == 0) { if (oldBuf->data) { if (!oldBuf->userBuffer) - free(oldBuf->data); + align_free(oldBuf->data); oldBuf->data = NULL; } free(oldBuf); @@ -163,8 +210,8 @@ xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, assert(!xm_buf->userBuffer); if (xm_buf->size != size) { if (xm_buf->data) - free(xm_buf->data); - xm_buf->data = malloc(size); + align_free(xm_buf->data); + xm_buf->data = align_malloc(size, 16); xm_buf->size = size; } if (data) @@ -197,6 +244,45 @@ xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, } +/** + * Display a surface that's in a tiled configuration. That is, all the + * pixels for a TILE_SIZExTILE_SIZE block are contiguous in memory. + */ +static void +xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) +{ + XImage *ximage = b->tempImage; + struct xm_buffer *xm_buf = xm_bo(surf->buffer); + const int TILE_SIZE = 32; + uint x, y; + + /* check that the XImage has been previously initialized */ + assert(ximage->format); + assert(ximage->bitmap_unit); + + /* update XImage's fields */ + ximage->width = TILE_SIZE; + ximage->height = TILE_SIZE; + ximage->bytes_per_line = TILE_SIZE * 4; + + for (y = 0; y < surf->height; y += TILE_SIZE) { + for (x = 0; x < surf->width; x += TILE_SIZE) { + int dx = x; + int dy = y; + int tx = x / TILE_SIZE; + int ty = y / TILE_SIZE; + int offset = ty * (surf->width / TILE_SIZE) + tx; + offset *= 4 * TILE_SIZE * TILE_SIZE; + + ximage->data = (char *) xm_buf->data + offset; + + XPutImage(b->xm_visual->display, b->drawable, b->gc, + ximage, 0, 0, dx, dy, TILE_SIZE, TILE_SIZE); + } + } +} + + /** * Display/copy the image in the surface into the X window specified * by the XMesaBuffer. @@ -206,6 +292,13 @@ xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) { XImage *ximage = b->tempImage; struct xm_buffer *xm_buf = xm_bo(surf->buffer); + const struct xmesa_surface *xm_surf + = xmesa_surface((struct pipe_surface *) surf); + + if (xm_surf->tileSize) { + xmesa_display_surface_tiled(b, surf); + return; + } /* check that the XImage has been previously initialized */ assert(ximage->format); @@ -234,7 +327,6 @@ xm_flush_frontbuffer(struct pipe_winsys *pws, * this would be the place to copy the Ximage to the on-screen Window. */ XMesaContext xmctx = (XMesaContext) context_private; - xmesa_display_surface(xmctx->xm_buffer, surf); } @@ -318,6 +410,12 @@ xm_surface_alloc(struct pipe_winsys *ws, enum pipe_format pipeFormat) xms->surface.refcount = 1; xms->surface.winsys = ws; +#ifdef GALLIUM_CELL + if (getenv("GALLIUM_CELL")) { + xms->tileSize = 32; /** probably temporary */ + } +#endif + return &xms->surface; } @@ -413,12 +511,24 @@ struct pipe_context * xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) { struct pipe_winsys *pws = xmesa_get_pipe_winsys(); - struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(pixelformat); struct pipe_context *pipe; - pipe = softpipe_create( pws, spws ); - if (pipe) - pipe->priv = xmesa; - - return pipe; +#ifdef GALLIUM_CELL + if (getenv("GALLIUM_CELL")) { + struct cell_winsys *cws = cell_get_winsys(pixelformat); + pipe = cell_create_context(pws, cws); + if (pipe) + pipe->priv = xmesa; + return pipe; + } + else +#endif + { + struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(pixelformat); + pipe = softpipe_create( pws, spws ); + if (pipe) + pipe->priv = xmesa; + + return pipe; + } } -- cgit v1.2.3 From 7c306afdaad534cc4b474f07b4970bbf3ea46ff9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 10 Dec 2007 22:02:04 +0000 Subject: gallium: remove unnecessary guards on qs->next --- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 3 +-- src/mesa/pipe/softpipe/sp_quad_blend.c | 3 +-- src/mesa/pipe/softpipe/sp_quad_bufloop.c | 3 +-- src/mesa/pipe/softpipe/sp_quad_colormask.c | 3 +-- src/mesa/pipe/softpipe/sp_quad_coverage.c | 3 +-- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 3 +-- src/mesa/pipe/softpipe/sp_quad_earlyz.c | 8 ++------ src/mesa/pipe/softpipe/sp_quad_fs.c | 3 +-- src/mesa/pipe/softpipe/sp_quad_occlusion.c | 6 ++---- src/mesa/pipe/softpipe/sp_quad_output.c | 3 +-- src/mesa/pipe/softpipe/sp_quad_stencil.c | 3 +-- src/mesa/pipe/softpipe/sp_quad_stipple.c | 3 +-- 12 files changed, 14 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index 7b56bceba2..d056abe98d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -84,8 +84,7 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) static void alpha_test_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c index 1843e20684..17f3ecd0b8 100644 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ b/src/mesa/pipe/softpipe/sp_quad_blend.c @@ -726,8 +726,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) static void blend_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c index e704b4043c..2ae4e22a7d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ b/src/mesa/pipe/softpipe/sp_quad_bufloop.c @@ -43,8 +43,7 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) static void cbuf_loop_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c index 867cc0725a..1f09d900ca 100644 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ b/src/mesa/pipe/softpipe/sp_quad_colormask.c @@ -87,8 +87,7 @@ colormask_quad(struct quad_stage *qs, struct quad_header *quad) static void colormask_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c index cca2b9f79b..b3d3fae22f 100644 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ b/src/mesa/pipe/softpipe/sp_quad_coverage.c @@ -65,8 +65,7 @@ coverage_quad(struct quad_stage *qs, struct quad_header *quad) static void coverage_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 93ea1196f9..1b8a2960af 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -253,8 +253,7 @@ depth_test_quad(struct quad_stage *qs, struct quad_header *quad) static void depth_test_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_earlyz.c b/src/mesa/pipe/softpipe/sp_quad_earlyz.c index 89fab2dd35..3abd1f1fb9 100644 --- a/src/mesa/pipe/softpipe/sp_quad_earlyz.c +++ b/src/mesa/pipe/softpipe/sp_quad_earlyz.c @@ -56,18 +56,14 @@ earlyz_quad( quad->outputs.depth[2] = z0 + dzdy; quad->outputs.depth[3] = z0 + dzdx + dzdy; - if (qs->next) { - qs->next->run( qs->next, quad ); - } + qs->next->run( qs->next, quad ); } static void earlyz_begin( struct quad_stage *qs ) { - if (qs->next) { - qs->next->begin( qs->next ); - } + qs->next->begin( qs->next ); } static void diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 1aba54d12a..75576a9bde 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -304,8 +304,7 @@ static void shade_begin(struct quad_stage *qs) } } - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c index 028d30c92f..d65fdbdab7 100644 --- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c +++ b/src/mesa/pipe/softpipe/sp_quad_occlusion.c @@ -52,15 +52,13 @@ occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad) occ->count += (quad->mask >> 2) & 1; occ->count += (quad->mask >> 3) & 1; - if (quad->mask) - qs->next->run(qs->next, quad); + qs->next->run(qs->next, quad); } static void occlusion_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c index bfd7baa946..cfe8f11808 100644 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ b/src/mesa/pipe/softpipe/sp_quad_output.c @@ -67,8 +67,7 @@ output_quad(struct quad_stage *qs, struct quad_header *quad) static void output_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + assert(qs->next == NULL); } diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index b8c199204d..3f3eca078b 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -334,8 +334,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) static void stencil_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index fcbbf00c7d..04d95989c4 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -60,8 +60,7 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) static void stipple_begin(struct quad_stage *qs) { - if (qs->next) - qs->next->begin(qs->next); + qs->next->begin(qs->next); } -- cgit v1.2.3 From f2feb2e4c7304b1c57af2f77f00766cb529822da Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 10 Dec 2007 22:27:04 +0000 Subject: gallium: disable assert(0) in xmesa_surface() --- src/mesa/pipe/xlib/xm_winsys.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index ec171e3d9a..c372545cf8 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -127,7 +127,7 @@ struct xmesa_softpipe_winsys static INLINE struct xmesa_surface * xmesa_surface(struct pipe_surface *ps) { - assert(0); +// assert(0); return (struct xmesa_surface *) ps; } -- cgit v1.2.3 From 0a79c65073db7dbeb9a83e40a4516edcb4f25ee5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 10 Dec 2007 22:38:44 +0000 Subject: gallium: use hardcoded cliptest code for fixed planes --- src/mesa/pipe/draw/draw_vertex_shader.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index eef71a7a4a..5294d38f47 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -51,7 +51,18 @@ compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) unsigned mask = 0; unsigned i; - for (i = 0; i < nr; i++) { + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { if (dot4(clip, plane[i]) < 0) mask |= (1< Date: Mon, 10 Dec 2007 16:18:24 -0700 Subject: minor clean-ups, comments --- src/mesa/pipe/softpipe/sp_prim_setup.c | 35 ++++++++++++++++++---------------- 1 file changed, 19 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 6980564bdc..fc96f92af1 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -47,11 +47,11 @@ * Triangle edge info */ struct edge { - float dx; /**< X(v1) - X(v0), used only during setup */ - float dy; /**< Y(v1) - Y(v0), used only during setup */ + float dx; /**< X(v1) - X(v0), used only during setup */ + float dy; /**< Y(v1) - Y(v0), used only during setup */ float dxdy; /**< dx/dy */ - float sx, sy; /**< first sample point coord */ - int lines; /**< number of lines on this edge */ + float sx, sy; /**< first sample point coord */ + int lines; /**< number of lines on this edge */ }; @@ -178,10 +178,9 @@ static INLINE int block( int x ) * this is pretty nasty... may need to rework flush_spans again to * fix it, if possible. */ -static unsigned calculate_mask( struct setup_stage *setup, - int x ) +static unsigned calculate_mask( struct setup_stage *setup, int x ) { - unsigned mask = 0; + unsigned mask = 0x0; if (x >= setup->span.left[0] && x < setup->span.right[0]) mask |= MASK_TOP_LEFT; @@ -207,18 +206,21 @@ static void flush_spans( struct setup_stage *setup ) int minleft, maxright; int x; - switch (setup->span.y_flags) { - case 3: + switch (setup->span.y_flags) { + case 0x3: + /* both odd and even lines written (both quad rows) */ minleft = MIN2(setup->span.left[0], setup->span.left[1]); maxright = MAX2(setup->span.right[0], setup->span.right[1]); break; - case 1: + case 0x1: + /* only even line written (quad top row) */ minleft = setup->span.left[0]; maxright = setup->span.right[0]; break; - case 2: + case 0x2: + /* only odd line written (quad bottom row) */ minleft = setup->span.left[1]; maxright = setup->span.right[1]; break; @@ -227,12 +229,12 @@ static void flush_spans( struct setup_stage *setup ) return; } - - for (x = block(minleft); x <= block(maxright); ) - { + /* XXX this loop could be moved into the above switch cases and + * calculate_mask() could be simplified a bit... + */ + for (x = block(minleft); x <= block(maxright); x += 2) { emit_quad( setup, x, setup->span.y, calculate_mask( setup, x ) ); - x += 2; } setup->span.y = 0; @@ -593,7 +595,8 @@ static void subtriangle( struct setup_stage *setup, setup->span.y = block(_y); } - setup->span.left[_y&1] = left;setup->span.right[_y&1] = right; + setup->span.left[_y&1] = left; + setup->span.right[_y&1] = right; setup->span.y_flags |= 1<<(_y&1); } } -- cgit v1.2.3 From c9f9c5098042e4f200efc3d25447a5a8b7e69b36 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Dec 2007 17:25:54 -0700 Subject: XXX comments about some hard-coded values that need to be fixed --- src/mesa/state_tracker/st_cb_fbo.c | 1 + src/mesa/state_tracker/st_draw.c | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 6e9e7e3a24..ef6aec6e88 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -120,6 +120,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe->winsys->buffer_reference(pipe->winsys, &strb->surface->buffer, NULL); + /* XXX don't hard-code magic 32 here */ strb->surface->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); if (!strb->surface->buffer) return GL_FALSE; /* out of memory, try s/w buffer? */ diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 24bb4823bc..274ae86a3e 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -199,6 +199,7 @@ static void create_default_attribs_buffer(struct st_context *st) { struct pipe_context *pipe = st->pipe; + /* XXX don't hardcode magic 32 here */ st->default_attrib_buffer = pipe->winsys->buffer_create( pipe->winsys, 32, 0, 0 ); } -- cgit v1.2.3 From 7d1894c6558c5cc7f503142cda11b8a12ea24e65 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Dec 2007 17:32:08 -0700 Subject: Cell: first triangle. This is a feeble first step, but it works. The cell_clear_surface() function has been hijacked to set up a "draw triangle" command and send it to all the SPUs. The Gallium softpipe triangle code was copied to the SPU module and modified. Only the progs/trivial/clear.c program runs. --- src/mesa/pipe/cell/common.h | 12 +- src/mesa/pipe/cell/ppu/cell_spu.c | 14 - src/mesa/pipe/cell/ppu/cell_surface.c | 29 +- src/mesa/pipe/cell/spu/main.c | 66 +-- src/mesa/pipe/cell/spu/tri.c | 818 +++++++++++++++++++++++++++++++++- src/mesa/pipe/cell/spu/tri.h | 50 ++- src/mesa/pipe/xlib/xm_winsys.c | 4 +- 7 files changed, 935 insertions(+), 58 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index da7de78803..5e5abad8c4 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -49,7 +49,7 @@ #define CELL_CMD_EXIT 1 #define CELL_CMD_FRAMEBUFFER 2 #define CELL_CMD_CLEAR_TILES 3 -#define CELL_CMD_INVERT_TILES 4 +#define CELL_CMD_TRIANGLE 4 #define CELL_CMD_FINISH 5 @@ -73,11 +73,21 @@ struct cell_command_clear_tiles } ALIGN16; +struct cell_command_triangle +{ + float x0, y0; + float x1, y1; + float x2, y2; + uint color; +} ALIGN16; + + /** XXX unions don't seem to work */ struct cell_command { struct cell_command_framebuffer fb; struct cell_command_clear_tiles clear; + struct cell_command_triangle tri; } ALIGN16; diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index ed56250ff1..4f111cd219 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -170,20 +170,6 @@ test_spus(struct cell_context *cell) b[0], b[1000], b[2000], b[3000]); } - for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(control_ps_area[i], CELL_CMD_INVERT_TILES); - } - - finish_all(cell->num_spus); - - { - uint *b = (uint*) surf->map; - printf("PPU: Inverted results: 0x%x 0x%x 0x%x 0x%x\n", - b[0], b[1000], b[2000], b[3000]); - } - - - for (i = 0; i < cell->num_spus; i++) { send_mbox_message(control_ps_area[i], CELL_CMD_EXIT); } diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 1692960244..66e4b5108b 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -70,14 +70,6 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, struct cell_context *cell = cell_context(pipe); uint i; - printf("%s 0x%08x\n", __FUNCTION__, clearValue); - - { - char s[100]; - pf_sprint_name(s, ps->format); - printf("format = %s\n", s); - } - if (!ps->map) pipe_surface_map(ps); @@ -90,9 +82,30 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, } for (i = 0; i < cell->num_spus; i++) { + /* XXX clear color varies per-SPU for debugging */ command[i].clear.value = clearValue | (i << 21); send_mbox_message(control_ps_area[i], CELL_CMD_CLEAR_TILES); } + +#if 1 + /* XXX Draw a test triangle over the cleared surface */ + for (i = 0; i < cell->num_spus; i++) { + /* Same triangle data for all SPUs, of course: */ + command[i].tri.x0 = 20.0; + command[i].tri.y0 = ps->height - 20; + + command[i].tri.x1 = ps->width - 20.0; + command[i].tri.y1 = ps->height - 20; + + command[i].tri.x2 = ps->width / 2; + command[i].tri.y2 = 20.0; + + /* XXX color varies per SPU */ + command[i].tri.color = 0xffff00 | ((i*40)<<24); /* yellow */ + + send_mbox_message(control_ps_area[i], CELL_CMD_TRIANGLE); + } +#endif } diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 226d81b4ca..f2c6d31d3c 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -34,6 +34,7 @@ #include #include +#include "main.h" #include "tri.h" #include "pipe/cell/common.h" @@ -43,21 +44,15 @@ helpful headers: /opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h */ -static struct cell_init_info init; +struct cell_init_info init; -struct framebuffer { - void *start; - uint width, height; - uint width_tiles, height_tiles; /**< width and height in tiles */ -}; -static struct framebuffer fb; +struct framebuffer fb; +int DefaultTag; -static int DefaultTag = 1; - -static inline void +void wait_on_mask(unsigned tag) { mfc_write_tag_mask( tag ); @@ -66,7 +61,7 @@ wait_on_mask(unsigned tag) -static void +void get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile) { uint offset = ty * fb->width_tiles + tx; @@ -89,7 +84,7 @@ get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile) 0 /* rid */); } -static void +void put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile) { uint offset = ty * fb->width_tiles + tx; @@ -136,32 +131,29 @@ clear_tiles(const struct cell_command_clear_tiles *clear) } -/** Invert all pixels in all tiles */ static void -invert_tiles(void) +triangle(const struct cell_command_triangle *tri) { uint num_tiles = fb.width_tiles * fb.height_tiles; - uint i, j; - uint tile[TILE_SIZE * TILE_SIZE] ALIGN16; + struct prim_header prim; + uint i; + + prim.v[0].data[0][0] = tri->x0; + prim.v[0].data[0][1] = tri->y0; + prim.v[1].data[0][0] = tri->x1; + prim.v[1].data[0][1] = tri->y1; + prim.v[2].data[0][0] = tri->x2; + prim.v[2].data[0][1] = tri->y2; + prim.color = tri->color; for (i = init.id; i < num_tiles; i += init.num_spus) { uint tx = i % fb.width_tiles; uint ty = i / fb.width_tiles; - - get_tile(&fb, tx, ty, tile); - wait_on_mask(1 << DefaultTag); - - for (j = 0; j < TILE_SIZE * TILE_SIZE; j++) { - tile[j] = ~tile[j]; - } - - put_tile(&fb, tx, ty, tile); + draw_triangle(&prim, tx, ty); } } -struct cell_command cmd ALIGN16; - /** * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. @@ -169,7 +161,9 @@ struct cell_command cmd ALIGN16; static void main_loop(void) { + struct cell_command cmd ALIGN16; int exitFlag = 0; + printf("SPU %u: Enter main loop\n", init.id); assert((sizeof(struct cell_command) & 0xf) == 0); @@ -207,17 +201,23 @@ main_loop(void) cmd.fb.start); fb.width = cmd.fb.width; fb.height = cmd.fb.height; - fb.width_tiles = fb.width / TILE_SIZE; - fb.height_tiles = fb.height / TILE_SIZE; + fb.width_tiles = (fb.width + TILE_SIZE - 1) / TILE_SIZE; + fb.height_tiles = (fb.height + TILE_SIZE - 1) / TILE_SIZE; + printf("SPU %u: %u x %u tiles\n", + init.id, fb.width_tiles, fb.height_tiles); fb.start = cmd.fb.start; break; case CELL_CMD_CLEAR_TILES: printf("SPU %u: CLEAR to 0x%08x\n", init.id, cmd.clear.value); clear_tiles(&cmd.clear); break; - case CELL_CMD_INVERT_TILES: - printf("SPU %u: INVERT_TILES\n", init.id); - invert_tiles(); + case CELL_CMD_TRIANGLE: + printf("SPU %u: TRIANGLE (%g,%g) (%g,%g) (%g,%g)\n", + init.id, + cmd.tri.x0, cmd.tri.y0, + cmd.tri.x1, cmd.tri.y1, + cmd.tri.x2, cmd.tri.y2); + triangle(&cmd.tri); break; case CELL_CMD_FINISH: printf("SPU %u: FINISH\n", init.id); @@ -245,6 +245,8 @@ main(unsigned long long speid, { int tag = 0; + DefaultTag = 1; + (void) speid; (void) envp; diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 949c3b4c8e..793bf55743 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -1,9 +1,825 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +/** + * Triangle rendering within a tile. + */ + + +#if 0 +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "sp_prim_setup.h" +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_vertex.h" +#include "pipe/p_util.h" +#endif + + +#include "pipe/p_compiler.h" +#include "pipe/p_util.h" +#include "main.h" #include "tri.h" + +#if 1 + +/* XXX fix this */ +#undef CEILF +#define CEILF(X) ((float) (int) ((X) + 0.99999)) + + +#define QUAD_TOP_LEFT 0 +#define QUAD_TOP_RIGHT 1 +#define QUAD_BOTTOM_LEFT 2 +#define QUAD_BOTTOM_RIGHT 3 +#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) +#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) +#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) +#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) +#define MASK_ALL 0xf + +static int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; + +static uint tile[TILE_SIZE][TILE_SIZE] ALIGN16; + +#endif + + +#define DEBUG_VERTS 0 + +/** + * Triangle edge info + */ +struct edge { + float dx; /**< X(v1) - X(v0), used only during setup */ + float dy; /**< Y(v1) - Y(v0), used only during setup */ + float dxdy; /**< dx/dy */ + float sx, sy; /**< first sample point coord */ + int lines; /**< number of lines on this edge */ +}; + + +/** + * Triangle setup info (derived from draw_stage). + * Also used for line drawing (taking some liberties). + */ +struct setup_stage { +#if 0 + struct draw_stage stage; /**< This must be first (base class) */ + + struct softpipe_context *softpipe; +#endif + + /* Vertices are just an array of floats making up each attribute in + * turn. Currently fixed at 4 floats, but should change in time. + * Codegen will help cope with this. + */ + const struct vertex_header *vmax; + const struct vertex_header *vmid; + const struct vertex_header *vmin; + const struct vertex_header *vprovoke; + + struct edge ebot; + struct edge etop; + struct edge emaj; + + float oneoverarea; + +#if 0 + struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; +#endif +#if 0 + struct quad_header quad; +#endif +#if 1 + uint color; +#endif + + struct { + int left[2]; /**< [0] = row0, [1] = row1 */ + int right[2]; + int y; + unsigned y_flags; + unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ + } span; +}; + + +#if 0 +/** + * Basically a cast wrapper. + */ +static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) +{ + return (struct setup_stage *)stage; +} +#endif + +#if 0 +/** + * Clip setup->quad against the scissor/surface bounds. + */ +static INLINE void +quad_clip(struct setup_stage *setup) +{ + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + + if (setup->quad.x0 >= maxx || + setup->quad.y0 >= maxy || + setup->quad.x0 + 1 < minx || + setup->quad.y0 + 1 < miny) { + /* totally clipped */ + setup->quad.mask = 0x0; + return; + } + if (setup->quad.x0 < minx) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + if (setup->quad.y0 < miny) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + if (setup->quad.x0 == maxx - 1) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + if (setup->quad.y0 == maxy - 1) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); +} +#endif + +#if 0 +/** + * Emit a quad (pass to next stage) with clipping. + */ +static INLINE void +clip_emit_quad(struct setup_stage *setup) +{ + quad_clip(setup); + if (setup->quad.mask) { + struct softpipe_context *sp = setup->softpipe; + sp->quad.first->run(sp->quad.first, &setup->quad); + } +} +#endif + +/** + * Emit a quad (pass to next stage). No clipping is done. + */ +static INLINE void +emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) +{ +#if 0 + struct softpipe_context *sp = setup->softpipe; + setup->quad.x0 = x; + setup->quad.y0 = y; + setup->quad.mask = mask; + sp->quad.first->run(sp->quad.first, &setup->quad); +#else + /* Cell: "write" quad fragments to the tile by setting prim color */ + int ix = x - cliprect_minx; + int iy = y - cliprect_miny; + if (mask & MASK_TOP_LEFT) + tile[iy][ix] = setup->color; + if (mask & MASK_TOP_RIGHT) + tile[iy][ix+1] = setup->color; + if (mask & MASK_BOTTOM_LEFT) + tile[iy+1][ix] = setup->color; + if (mask & MASK_BOTTOM_RIGHT) + tile[iy+1][ix+1] = setup->color; +#endif +} + + +/** + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. + */ +static INLINE int block( int x ) +{ + return x & ~1; +} + + +/** + * Compute mask which indicates which pixels in the 2x2 quad are actually inside + * the triangle's bounds. + * + * this is pretty nasty... may need to rework flush_spans again to + * fix it, if possible. + */ +static unsigned calculate_mask( struct setup_stage *setup, int x ) +{ + unsigned mask = 0x0; + + if (x >= setup->span.left[0] && x < setup->span.right[0]) + mask |= MASK_TOP_LEFT; + + if (x >= setup->span.left[1] && x < setup->span.right[1]) + mask |= MASK_BOTTOM_LEFT; + + if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) + mask |= MASK_TOP_RIGHT; + + if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) + mask |= MASK_BOTTOM_RIGHT; + + return mask; +} + + +/** + * Render a horizontal span of quads + */ +static void flush_spans( struct setup_stage *setup ) +{ + int minleft, maxright; + int x; + + switch (setup->span.y_flags) { + case 0x3: + /* both odd and even lines written (both quad rows) */ + minleft = MIN2(setup->span.left[0], setup->span.left[1]); + maxright = MAX2(setup->span.right[0], setup->span.right[1]); + break; + + case 0x1: + /* only even line written (quad top row) */ + minleft = setup->span.left[0]; + maxright = setup->span.right[0]; + break; + + case 0x2: + /* only odd line written (quad bottom row) */ + minleft = setup->span.left[1]; + maxright = setup->span.right[1]; + break; + + default: + return; + } + + /* XXX this loop could be moved into the above switch cases and + * calculate_mask() could be simplified a bit... + */ + for (x = block(minleft); x <= block(maxright); x += 2) { + emit_quad( setup, x, setup->span.y, + calculate_mask( setup, x ) ); + } + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +} + +#if DEBUG_VERTS +static void print_vertex(const struct setup_stage *setup, + const struct vertex_header *v) +{ + int i; + fprintf(stderr, "Vertex: (%p)\n", v); + for (i = 0; i < setup->quad.nr_attrs; i++) { + fprintf(stderr, " %d: %f %f %f %f\n", i, + v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); + } +} +#endif + +static boolean setup_sort_vertices( struct setup_stage *setup, + const struct prim_header *prim ) +{ +#if 0 + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + const struct vertex_header *v2 = prim->v[2]; +#else + const struct vertex_header *v0 = &prim->v[0]; + const struct vertex_header *v1 = &prim->v[1]; + const struct vertex_header *v2 = &prim->v[2]; +#endif + +#if DEBUG_VERTS + fprintf(stderr, "Triangle:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); + print_vertex(setup, v2); +#endif + + setup->vprovoke = v2; + + /* determine bottom to top order of vertices */ + { + float y0 = v0->data[0][1]; + float y1 = v1->data[0][1]; + float y2 = v2->data[0][1]; + if (y0 <= y1) { + if (y1 <= y2) { + /* y0<=y1<=y2 */ + setup->vmin = v0; + setup->vmid = v1; + setup->vmax = v2; + } + else if (y2 <= y0) { + /* y2<=y0<=y1 */ + setup->vmin = v2; + setup->vmid = v0; + setup->vmax = v1; + } + else { + /* y0<=y2<=y1 */ + setup->vmin = v0; + setup->vmid = v2; + setup->vmax = v1; + } + } + else { + if (y0 <= y2) { + /* y1<=y0<=y2 */ + setup->vmin = v1; + setup->vmid = v0; + setup->vmax = v2; + } + else if (y2 <= y1) { + /* y2<=y1<=y0 */ + setup->vmin = v2; + setup->vmid = v1; + setup->vmax = v0; + } + else { + /* y1<=y2<=y0 */ + setup->vmin = v1; + setup->vmid = v2; + setup->vmax = v0; + } + } + } + + setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; + setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; + setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; + + /* + * Compute triangle's area. Use 1/area to compute partial + * derivatives of attributes later. + * + * The area will be the same as prim->det, but the sign may be + * different depending on how the vertices get sorted above. + * + * To determine whether the primitive is front or back facing we + * use the prim->det value because its sign is correct. + */ + { + const float area = (setup->emaj.dx * setup->ebot.dy - + setup->ebot.dx * setup->emaj.dy); + + setup->oneoverarea = 1.0f / area; + /* + _mesa_printf("%s one-over-area %f area %f det %f\n", + __FUNCTION__, setup->oneoverarea, area, prim->det ); + */ + } + +#if 0 + /* We need to know if this is a front or back-facing triangle for: + * - the GLSL gl_FrontFacing fragment attribute (bool) + * - two-sided stencil test + */ + setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); +#endif + + return TRUE; +} + + +#if 0 +/** + * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + * The value value comes from vertex->data[slot][i]. + * The result will be put into setup->coef[slot].a0[i]. + * \param slot which attribute slot + * \param i which component of the slot (0..3) + */ +static void const_coeff( struct setup_stage *setup, + unsigned slot, + unsigned i ) +{ + assert(slot < PIPE_MAX_SHADER_INPUTS); + assert(i <= 3); + + setup->coef[slot].dadx[i] = 0; + setup->coef[slot].dady[i] = 0; + + /* need provoking vertex info! + */ + setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; +} +#endif + + +#if 0 +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a triangle. + */ +static void tri_linear_coeff( struct setup_stage *setup, + unsigned slot, + unsigned i) +{ + float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; + float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + assert(slot < PIPE_MAX_SHADER_INPUTS); + assert(i <= 3); + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); + + /* + _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", + slot, "xyzw"[i], + setup->coef[slot].a0[i], + setup->coef[slot].dadx[i], + setup->coef[slot].dady[i]); + */ +} +#endif + + +#if 0 +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a triangle. + * We basically multiply the vertex value by 1/w before computing + * the plane coefficients (a0, dadx, dady). + * Later, when we compute the value at a particular fragment position we'll + * divide the interpolated value by the interpolated W at that fragment. + */ +static void tri_persp_coeff( struct setup_stage *setup, + unsigned slot, + unsigned i ) +{ + /* premultiply by 1/w: + */ + float mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; + float mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; + float maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; + + float botda = mida - mina; + float majda = maxa - mina; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + /* + printf("tri persp %d,%d: %f %f %f\n", slot, i, + setup->vmin->data[slot][i], + setup->vmid->data[slot][i], + setup->vmax->data[slot][i] + ); + */ + + assert(slot < PIPE_MAX_SHADER_INPUTS); + assert(i <= 3); + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + setup->coef[slot].a0[i] = (mina - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); +} +#endif + + +#if 0 +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. + */ +static void setup_tri_coefficients( struct setup_stage *setup ) +{ + const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + unsigned slot, j; + + /* z and w are done by linear interpolation: + */ + tri_linear_coeff(setup, 0, 2); + tri_linear_coeff(setup, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + switch (interp[slot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + tri_linear_coeff(setup, slot, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + tri_persp_coeff(setup, slot, j); + break; + + default: + /* invalid interp mode */ + assert(0); + } + } +} +#endif + + +static void setup_tri_edges( struct setup_stage *setup ) +{ + float vmin_x = setup->vmin->data[0][0] + 0.5f; + float vmid_x = setup->vmid->data[0][0] + 0.5f; + + float vmin_y = setup->vmin->data[0][1] - 0.5f; + float vmid_y = setup->vmid->data[0][1] - 0.5f; + float vmax_y = setup->vmax->data[0][1] - 0.5f; + + setup->emaj.sy = CEILF(vmin_y); + setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); + setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; + setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; + + setup->etop.sy = CEILF(vmid_y); + setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); + setup->etop.dxdy = setup->etop.dx / setup->etop.dy; + setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; + + setup->ebot.sy = CEILF(vmin_y); + setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); + setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; + setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; +} + + +/** + * Render the upper or lower half of a triangle. + * Scissoring/cliprect is applied here too. + */ +static void subtriangle( struct setup_stage *setup, + struct edge *eleft, + struct edge *eright, + unsigned lines ) +{ +#if 0 + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; +#else + const int minx = cliprect_minx; + const int maxx = cliprect_maxx; + const int miny = cliprect_miny; + const int maxy = cliprect_maxy; +#endif + int y, start_y, finish_y; + int sy = (int)eleft->sy; + + assert((int)eleft->sy == (int) eright->sy); + + /* clip top/bottom */ + start_y = sy; + finish_y = sy + lines; + + if (start_y < miny) + start_y = miny; + + if (finish_y > maxy) + finish_y = maxy; + + start_y -= sy; + finish_y -= sy; + + /* + _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + */ + + for (y = start_y; y < finish_y; y++) { + + /* avoid accumulating adds as floats don't have the precision to + * accurately iterate large triangle edges that way. luckily we + * can just multiply these days. + * + * this is all drowned out by the attribute interpolation anyway. + */ + int left = (int)(eleft->sx + y * eleft->dxdy); + int right = (int)(eright->sx + y * eright->dxdy); + + /* clip left/right */ + if (left < minx) + left = minx; + if (right > maxx) + right = maxx; + + if (left < right) { + int _y = sy + y; + if (block(_y) != setup->span.y) { + flush_spans(setup); + setup->span.y = block(_y); + } + + setup->span.left[_y&1] = left; + setup->span.right[_y&1] = right; + setup->span.y_flags |= 1<<(_y&1); + } + } + + + /* save the values so that emaj can be restarted: + */ + eleft->sx += lines * eleft->dxdy; + eright->sx += lines * eright->dxdy; + eleft->sy += lines; + eright->sy += lines; +} + + +/** + * Do setup for triangle rasterization, then render the triangle. + */ +static void setup_tri( +#if 0 + struct draw_stage *stage, +#endif + struct prim_header *prim ) +{ +#if 0 + struct setup_stage *setup = setup_stage( stage ); +#else + struct setup_stage ss; + struct setup_stage *setup = &ss; + ss.color = prim->color; +#endif + + /* + _mesa_printf("%s\n", __FUNCTION__ ); + */ + + setup_sort_vertices( setup, prim ); +#if 0 + setup_tri_coefficients( setup ); +#endif + setup_tri_edges( setup ); + +#if 0 + setup->quad.prim = PRIM_TRI; +#endif + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; + /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ + + /* init_constant_attribs( setup ); */ + + if (setup->oneoverarea < 0.0) { + /* emaj on left: + */ + subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); + subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); + } + else { + /* emaj on right: + */ + subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); + subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); + } + + flush_spans( setup ); +} + + + + +#if 0 +static void setup_begin( struct draw_stage *stage ) +{ + struct setup_stage *setup = setup_stage(stage); + struct softpipe_context *sp = setup->softpipe; + + setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; + + sp->quad.first->begin(sp->quad.first); +} +#endif + +#if 0 +static void setup_end( struct draw_stage *stage ) +{ +} +#endif + + +#if 0 +static void reset_stipple_counter( struct draw_stage *stage ) +{ + struct setup_stage *setup = setup_stage(stage); + setup->softpipe->line_stipple_counter = 0; +} +#endif + +#if 0 +static void render_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} +#endif + + +#if 0 +/** + * Create a new primitive setup/render stage. + */ +struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) +{ + struct setup_stage *setup = CALLOC_STRUCT(setup_stage); + + setup->softpipe = softpipe; + setup->stage.draw = softpipe->draw; + setup->stage.begin = setup_begin; + setup->stage.point = setup_point; + setup->stage.line = setup_line; + setup->stage.tri = setup_tri; + setup->stage.end = setup_end; + setup->stage.reset_stipple_counter = reset_stipple_counter; + setup->stage.destroy = render_destroy; + + setup->quad.coef = setup->coef; + + return &setup->stage; +} +#endif + + void -draw_triangle(int v1, int v2, int v3) +draw_triangle(struct prim_header *tri, uint tx, uint ty) { + /* set clipping bounds to tile bounds */ + cliprect_minx = tx * TILE_SIZE; + cliprect_miny = ty * TILE_SIZE; + cliprect_maxx = (tx + 1) * TILE_SIZE; + cliprect_maxy = (ty + 1) * TILE_SIZE; + + get_tile(&fb, tx, ty, (uint *) tile); + wait_on_mask(1 << DefaultTag); + setup_tri(tri); + put_tile(&fb, tx, ty, (uint *) tile); + wait_on_mask(1 << DefaultTag); } diff --git a/src/mesa/pipe/cell/spu/tri.h b/src/mesa/pipe/cell/spu/tri.h index 6a915de60d..dc66ad0f47 100644 --- a/src/mesa/pipe/cell/spu/tri.h +++ b/src/mesa/pipe/cell/spu/tri.h @@ -1,4 +1,52 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef TRI_H +#define TRI_H + + +/** + * Simplified types taken from other parts of Gallium + */ + +struct vertex_header { + float data[2][4]; /* pos and color */ +}; + + +struct prim_header { + struct vertex_header v[3]; + uint color; +}; extern void -draw_triangle(int v1, int v2, int v3); +draw_triangle(struct prim_header *tri, uint tx, uint ty); + + +#endif /* TRI_H */ diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index c372545cf8..b090d8927c 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -211,6 +211,7 @@ xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, if (xm_buf->size != size) { if (xm_buf->data) align_free(xm_buf->data); + /* align to 16-byte multiple for Cell */ xm_buf->data = align_malloc(size, 16); xm_buf->size = size; } @@ -254,6 +255,7 @@ xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) XImage *ximage = b->tempImage; struct xm_buffer *xm_buf = xm_bo(surf->buffer); const int TILE_SIZE = 32; + const uint tilesPerRow = (surf->width + TILE_SIZE - 1) / TILE_SIZE; uint x, y; /* check that the XImage has been previously initialized */ @@ -271,7 +273,7 @@ xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) int dy = y; int tx = x / TILE_SIZE; int ty = y / TILE_SIZE; - int offset = ty * (surf->width / TILE_SIZE) + tx; + int offset = ty * tilesPerRow + tx; offset *= 4 * TILE_SIZE * TILE_SIZE; ximage->data = (char *) xm_buf->data + offset; -- cgit v1.2.3 From 609538f57c93c6b6166777a329d80c46fef86f0b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 11 Dec 2007 00:46:44 +0000 Subject: Add inline funtion to comput format size based on code in st_format.c. Including state_tracker/st_format.h from pipe drivers is not an option since it uses GL* types and pipe/p_util.h will clash with main/imports.h. --- src/mesa/pipe/p_format.h | 47 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 47 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index 8f11bfab76..95a50c18af 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -366,4 +366,51 @@ static INLINE char *pf_sprint_name( char *str, uint format ) return str; } +static INLINE uint pf_get_component_bits( uint format, uint comp ) +{ + uint size; + + if (pf_swizzle_x(format) == comp) { + size = pf_size_x(format); + } + else if (pf_swizzle_y(format) == comp) { + size = pf_size_y(format); + } + else if (pf_swizzle_z(format) == comp) { + size = pf_size_z(format); + } + else if (pf_swizzle_w(format) == comp) { + size = pf_size_w(format); + } + else { + size = 0; + } + return size << (pf_exp8(format) * 3); +} + +static INLINE uint pf_get_bits( uint format ) +{ + if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { + return + pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_S ); + } + else { + assert( pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR ); + + /* TODO */ + assert( 0 ); + return 0; + } +} + +static INLINE uint pf_get_size( uint format ) { + assert(pf_get_bits(format) % 8 == 0); + return pf_get_bits(format) / 8; +} + #endif -- cgit v1.2.3 From 12363674e5aa39b780020339038186b7715bd4b2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 11 Dec 2007 01:14:38 +0000 Subject: Add surface storage allocation function to winsys interface. --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 72 +++++++++++++++------- src/mesa/pipe/i915simple/i915_surface.c | 3 +- src/mesa/pipe/p_winsys.h | 18 +++--- src/mesa/pipe/softpipe/sp_surface.c | 4 +- src/mesa/pipe/xlib/xm_winsys.c | 41 +++++++++--- src/mesa/state_tracker/st_cb_fbo.c | 36 ++++------- 6 files changed, 108 insertions(+), 66 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 9c643dd0ba..c7b519d95b 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -178,32 +178,11 @@ intel_flush_frontbuffer( struct pipe_winsys *winsys, } -static unsigned -intel_i915_surface_pitch(struct pipe_winsys *winsys, - unsigned cpp, unsigned width, unsigned flags) -{ - /* Choose a pitch to match hardware requirements - requires 64 byte - * alignment of render targets. - * - * XXX: is this ok for textures?? - * clearly want to be able to render to textures under some - * circumstances, but maybe not always a requirement. - */ - - /* XXX is the pitch different for textures vs. drawables? */ - if (1/*flags & PIPE_SURFACE_FLAG_TEXTURE*/) /* or PIPE_SURFACE_FLAG_RENDER? */ - return ((cpp * width + 63) & ~63) / cpp; - else - return ((cpp * width + 63) & ~63) / cpp; -} - - static struct pipe_surface * -intel_i915_surface_alloc(struct pipe_winsys *winsys, enum pipe_format format) +intel_i915_surface_alloc(struct pipe_winsys *winsys) { struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); if (surf) { - surf->format = format; surf->refcount = 1; surf->winsys = winsys; } @@ -211,6 +190,53 @@ intel_i915_surface_alloc(struct pipe_winsys *winsys, enum pipe_format format) } +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + +/** + * Copied from xm_winsys.c + */ +static int +intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags) +{ + const unsigned alignment = 64; + int ret; + + surf->width = width; + surf->height = height; + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = round_up(width, alignment / surf->cpp); + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, 0, 0); + if(!surf->buffer) + return -1; + + ret = winsys->buffer_data(winsys, + surf->buffer, + surf->pitch * surf->cpp * height, + NULL, + 0); + if(ret) { + winsys->buffer_reference(winsys, &surf->buffer, NULL); + return ret; + } + + return 0; +} + + static void intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) { @@ -265,8 +291,8 @@ intel_create_pipe_winsys( int fd ) iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; iws->winsys.printf = intel_printf; iws->winsys.get_name = intel_get_name; - iws->winsys.surface_pitch = intel_i915_surface_pitch; iws->winsys.surface_alloc = intel_i915_surface_alloc; + iws->winsys.surface_alloc_storage = intel_i915_surface_alloc_storage; iws->winsys.surface_release = intel_i915_surface_release; if (fd) diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index bd6fd32704..f93a75b0f0 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -238,11 +238,12 @@ i915_get_tex_surface(struct pipe_context *pipe, assert(zslice == 0); } - ps = pipe->winsys->surface_alloc(pipe->winsys, pt->format); + ps = pipe->winsys->surface_alloc(pipe->winsys); if (ps) { assert(ps->format); assert(ps->refcount); pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; ps->height = pt->height[level]; diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 1418af6918..aa9362ec0b 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -76,16 +76,16 @@ struct pipe_winsys const char *, ... ); - /** - * flags is bitmask of PIPE_SURFACE_FLAG_RENDER, PIPE_SURFACE_FLAG_TEXTURE - */ - unsigned (*surface_pitch)(struct pipe_winsys *ws, unsigned cpp, - unsigned with, unsigned flags); - /** allocate a new surface (no context dependency) */ - struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws, - enum pipe_format format); - + struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws); + + /** allocate storage for a pipe_surface */ + int (*surface_alloc_storage)(struct pipe_winsys *ws, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags); + void (*surface_release)(struct pipe_winsys *ws, struct pipe_surface **s); diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 55b8b85ef2..3ef3db9f1f 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -576,11 +576,11 @@ softpipe_get_tex_surface(struct pipe_context *pipe, assert(zslice == 0); } - ps = pipe->winsys->surface_alloc(pipe->winsys, pt->format); + ps = pipe->winsys->surface_alloc(pipe->winsys); if (ps) { - assert(ps->format); assert(ps->refcount); pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); + ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; ps->height = pt->height[level]; diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index b090d8927c..ad31f4498e 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -388,11 +388,38 @@ round_up(unsigned n, unsigned multiple) return (n + multiple - 1) & ~(multiple - 1); } -static unsigned -xm_surface_pitch(struct pipe_winsys *winsys, unsigned cpp, unsigned width, - unsigned flags) +static int +xm_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags) { - return round_up(width, 64 / cpp); + const unsigned alignment = 64; + int ret; + + surf->width = width; + surf->height = height; + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = round_up(width, alignment / surf->cpp); + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, 0, 0); + if(!surf->buffer) + return -1; + + ret = winsys->buffer_data(winsys, + surf->buffer, + surf->pitch * surf->cpp * height, + NULL, + 0); + if(ret) { + winsys->buffer_reference(winsys, &surf->buffer, NULL); + return ret; + } + + return 0; } @@ -401,14 +428,12 @@ xm_surface_pitch(struct pipe_winsys *winsys, unsigned cpp, unsigned width, * renderbuffers, etc. */ static struct pipe_surface * -xm_surface_alloc(struct pipe_winsys *ws, enum pipe_format pipeFormat) +xm_surface_alloc(struct pipe_winsys *ws) { struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); assert(ws); - assert(pipeFormat); - xms->surface.format = pipeFormat; xms->surface.refcount = 1; xms->surface.winsys = ws; @@ -463,8 +488,8 @@ xmesa_get_pipe_winsys(void) ws->buffer_subdata = xm_buffer_subdata; ws->buffer_get_subdata = xm_buffer_get_subdata; - ws->surface_pitch = xm_surface_pitch; ws->surface_alloc = xm_surface_alloc; + ws->surface_alloc_storage = xm_surface_alloc_storage; ws->surface_release = xm_surface_release; ws->flush_frontbuffer = xm_flush_frontbuffer; diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index ef6aec6e88..047de412e3 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -90,29 +90,15 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, uint type = strb->screenSurface ? PIPE_SCREEN_SURFACE : PIPE_SURFACE; const enum pipe_format pipeFormat = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE, type); - GLuint cpp; GLbitfield flags = 0x0; /* XXX needed? */ - cpp = init_renderbuffer_bits(strb, pipeFormat); - assert(cpp); - - if (strb->surface && strb->surface->format != pipeFormat) { - /* need to change surface types, free this surface */ - pipe_surface_reference(&strb->surface, NULL); - assert(strb->surface == NULL); - } - if (!strb->surface) { - strb->surface = pipe->winsys->surface_alloc(pipe->winsys, pipeFormat); + strb->surface = pipe->winsys->surface_alloc(pipe->winsys); assert(strb->surface); if (!strb->surface) return GL_FALSE; - strb->surface->cpp = cpp; } - strb->surface->pitch = pipe->winsys->surface_pitch(pipe->winsys, cpp, - width, flags); - /* loop here since mapping is refcounted */ while (strb->surface->map) pipe_surface_unmap(strb->surface); @@ -120,20 +106,24 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe->winsys->buffer_reference(pipe->winsys, &strb->surface->buffer, NULL); - /* XXX don't hard-code magic 32 here */ - strb->surface->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); + pipe->winsys->surface_alloc_storage(pipe->winsys, + strb->surface, + width, + height, + pipeFormat, + flags); if (!strb->surface->buffer) return GL_FALSE; /* out of memory, try s/w buffer? */ - pipe->winsys->buffer_data(pipe->winsys, strb->surface->buffer, - strb->surface->pitch * cpp * height, NULL, - PIPE_BUFFER_USAGE_PIXEL); - ASSERT(strb->surface->buffer); ASSERT(strb->surface->format); + ASSERT(strb->surface->cpp); + ASSERT(strb->surface->width == width); + ASSERT(strb->surface->height == height); + ASSERT(strb->surface->pitch); - strb->Base.Width = strb->surface->width = width; - strb->Base.Height = strb->surface->height = height; + strb->Base.Width = width; + strb->Base.Height = height; return GL_TRUE; } -- cgit v1.2.3 From c474f1fb9088528af998168717783b52e5c2f0a2 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 11 Dec 2007 07:19:11 -0500 Subject: Port i965 driver to Gallium3D. This is a squashed commit of i965 branch on ssh://people.freedesktop.org/~zack/mesa Because of the porting the branch often didn't compile so squashing it makes more sense. The port is still far from complete. --- src/mesa/Makefile | 8 +- src/mesa/pipe/Makefile | 2 +- src/mesa/pipe/i965simple/Makefile | 75 ++ src/mesa/pipe/i965simple/brw_batch.h | 57 + src/mesa/pipe/i965simple/brw_blit.c | 217 ++++ src/mesa/pipe/i965simple/brw_blit.h | 32 + src/mesa/pipe/i965simple/brw_cc.c | 268 +++++ src/mesa/pipe/i965simple/brw_clip.c | 201 ++++ src/mesa/pipe/i965simple/brw_clip.h | 170 +++ src/mesa/pipe/i965simple/brw_clip_line.c | 245 ++++ src/mesa/pipe/i965simple/brw_clip_point.c | 47 + src/mesa/pipe/i965simple/brw_clip_state.c | 92 ++ src/mesa/pipe/i965simple/brw_clip_tri.c | 566 ++++++++++ src/mesa/pipe/i965simple/brw_clip_unfilled.c | 477 ++++++++ src/mesa/pipe/i965simple/brw_clip_util.c | 351 ++++++ src/mesa/pipe/i965simple/brw_context.c | 287 +++++ src/mesa/pipe/i965simple/brw_context.h | 699 ++++++++++++ src/mesa/pipe/i965simple/brw_curbe.c | 369 ++++++ src/mesa/pipe/i965simple/brw_defines.h | 852 ++++++++++++++ src/mesa/pipe/i965simple/brw_draw.c | 238 ++++ src/mesa/pipe/i965simple/brw_draw.h | 56 + src/mesa/pipe/i965simple/brw_draw_upload.c | 306 +++++ src/mesa/pipe/i965simple/brw_eu.c | 130 +++ src/mesa/pipe/i965simple/brw_eu.h | 877 +++++++++++++++ src/mesa/pipe/i965simple/brw_eu_debug.c | 88 ++ src/mesa/pipe/i965simple/brw_eu_emit.c | 1080 ++++++++++++++++++ src/mesa/pipe/i965simple/brw_eu_util.c | 126 +++ src/mesa/pipe/i965simple/brw_flush.c | 84 ++ src/mesa/pipe/i965simple/brw_gs.c | 196 ++++ src/mesa/pipe/i965simple/brw_gs.h | 75 ++ src/mesa/pipe/i965simple/brw_gs_emit.c | 148 +++ src/mesa/pipe/i965simple/brw_gs_state.c | 88 ++ src/mesa/pipe/i965simple/brw_metaops.c | 538 +++++++++ src/mesa/pipe/i965simple/brw_misc_state.c | 425 +++++++ src/mesa/pipe/i965simple/brw_program.c | 130 +++ src/mesa/pipe/i965simple/brw_reg.h | 76 ++ src/mesa/pipe/i965simple/brw_regions.c | 39 + src/mesa/pipe/i965simple/brw_regions.h | 5 + src/mesa/pipe/i965simple/brw_sf.c | 194 ++++ src/mesa/pipe/i965simple/brw_sf.h | 111 ++ src/mesa/pipe/i965simple/brw_sf_emit.c | 696 ++++++++++++ src/mesa/pipe/i965simple/brw_sf_state.c | 236 ++++ src/mesa/pipe/i965simple/brw_state.c | 466 ++++++++ src/mesa/pipe/i965simple/brw_state.h | 157 +++ src/mesa/pipe/i965simple/brw_state_batch.c | 113 ++ src/mesa/pipe/i965simple/brw_state_cache.c | 442 ++++++++ src/mesa/pipe/i965simple/brw_state_pool.c | 139 +++ src/mesa/pipe/i965simple/brw_state_upload.c | 231 ++++ src/mesa/pipe/i965simple/brw_strings.c | 72 ++ src/mesa/pipe/i965simple/brw_structs.h | 1348 ++++++++++++++++++++++ src/mesa/pipe/i965simple/brw_surface.c | 397 +++++++ src/mesa/pipe/i965simple/brw_tex.c | 65 ++ src/mesa/pipe/i965simple/brw_tex_layout.c | 359 ++++++ src/mesa/pipe/i965simple/brw_tex_layout.h | 15 + src/mesa/pipe/i965simple/brw_urb.c | 197 ++++ src/mesa/pipe/i965simple/brw_util.c | 104 ++ src/mesa/pipe/i965simple/brw_util.h | 43 + src/mesa/pipe/i965simple/brw_vs.c | 130 +++ src/mesa/pipe/i965simple/brw_vs.h | 82 ++ src/mesa/pipe/i965simple/brw_vs_constval.c | 223 ++++ src/mesa/pipe/i965simple/brw_vs_emit.c | 1243 +++++++++++++++++++++ src/mesa/pipe/i965simple/brw_vs_state.c | 102 ++ src/mesa/pipe/i965simple/brw_vtbl.c | 149 +++ src/mesa/pipe/i965simple/brw_winsys.h | 192 ++++ src/mesa/pipe/i965simple/brw_wm.c | 338 ++++++ src/mesa/pipe/i965simple/brw_wm.h | 274 +++++ src/mesa/pipe/i965simple/brw_wm_debug.c | 172 +++ src/mesa/pipe/i965simple/brw_wm_emit.c | 1289 +++++++++++++++++++++ src/mesa/pipe/i965simple/brw_wm_fp.c | 1007 +++++++++++++++++ src/mesa/pipe/i965simple/brw_wm_glsl.c | 1356 +++++++++++++++++++++++ src/mesa/pipe/i965simple/brw_wm_iz.c | 214 ++++ src/mesa/pipe/i965simple/brw_wm_pass0.c | 466 ++++++++ src/mesa/pipe/i965simple/brw_wm_pass1.c | 277 +++++ src/mesa/pipe/i965simple/brw_wm_pass2.c | 335 ++++++ src/mesa/pipe/i965simple/brw_wm_sampler_state.c | 275 +++++ src/mesa/pipe/i965simple/brw_wm_state.c | 202 ++++ src/mesa/pipe/i965simple/brw_wm_surface_state.c | 294 +++++ src/mesa/pipe/xlib/brw_aub.c | 392 +++++++ src/mesa/pipe/xlib/brw_aub.h | 114 ++ src/mesa/pipe/xlib/xm_api.c | 18 +- src/mesa/pipe/xlib/xm_winsys.c | 12 +- src/mesa/pipe/xlib/xmesaP.h | 6 + src/mesa/sources | 4 +- 83 files changed, 24260 insertions(+), 11 deletions(-) create mode 100644 src/mesa/pipe/i965simple/Makefile create mode 100644 src/mesa/pipe/i965simple/brw_batch.h create mode 100644 src/mesa/pipe/i965simple/brw_blit.c create mode 100644 src/mesa/pipe/i965simple/brw_blit.h create mode 100644 src/mesa/pipe/i965simple/brw_cc.c create mode 100644 src/mesa/pipe/i965simple/brw_clip.c create mode 100644 src/mesa/pipe/i965simple/brw_clip.h create mode 100644 src/mesa/pipe/i965simple/brw_clip_line.c create mode 100644 src/mesa/pipe/i965simple/brw_clip_point.c create mode 100644 src/mesa/pipe/i965simple/brw_clip_state.c create mode 100644 src/mesa/pipe/i965simple/brw_clip_tri.c create mode 100644 src/mesa/pipe/i965simple/brw_clip_unfilled.c create mode 100644 src/mesa/pipe/i965simple/brw_clip_util.c create mode 100644 src/mesa/pipe/i965simple/brw_context.c create mode 100644 src/mesa/pipe/i965simple/brw_context.h create mode 100644 src/mesa/pipe/i965simple/brw_curbe.c create mode 100644 src/mesa/pipe/i965simple/brw_defines.h create mode 100644 src/mesa/pipe/i965simple/brw_draw.c create mode 100644 src/mesa/pipe/i965simple/brw_draw.h create mode 100644 src/mesa/pipe/i965simple/brw_draw_upload.c create mode 100644 src/mesa/pipe/i965simple/brw_eu.c create mode 100644 src/mesa/pipe/i965simple/brw_eu.h create mode 100644 src/mesa/pipe/i965simple/brw_eu_debug.c create mode 100644 src/mesa/pipe/i965simple/brw_eu_emit.c create mode 100644 src/mesa/pipe/i965simple/brw_eu_util.c create mode 100644 src/mesa/pipe/i965simple/brw_flush.c create mode 100644 src/mesa/pipe/i965simple/brw_gs.c create mode 100644 src/mesa/pipe/i965simple/brw_gs.h create mode 100644 src/mesa/pipe/i965simple/brw_gs_emit.c create mode 100644 src/mesa/pipe/i965simple/brw_gs_state.c create mode 100644 src/mesa/pipe/i965simple/brw_metaops.c create mode 100644 src/mesa/pipe/i965simple/brw_misc_state.c create mode 100644 src/mesa/pipe/i965simple/brw_program.c create mode 100644 src/mesa/pipe/i965simple/brw_reg.h create mode 100644 src/mesa/pipe/i965simple/brw_regions.c create mode 100644 src/mesa/pipe/i965simple/brw_regions.h create mode 100644 src/mesa/pipe/i965simple/brw_sf.c create mode 100644 src/mesa/pipe/i965simple/brw_sf.h create mode 100644 src/mesa/pipe/i965simple/brw_sf_emit.c create mode 100644 src/mesa/pipe/i965simple/brw_sf_state.c create mode 100644 src/mesa/pipe/i965simple/brw_state.c create mode 100644 src/mesa/pipe/i965simple/brw_state.h create mode 100644 src/mesa/pipe/i965simple/brw_state_batch.c create mode 100644 src/mesa/pipe/i965simple/brw_state_cache.c create mode 100644 src/mesa/pipe/i965simple/brw_state_pool.c create mode 100644 src/mesa/pipe/i965simple/brw_state_upload.c create mode 100644 src/mesa/pipe/i965simple/brw_strings.c create mode 100644 src/mesa/pipe/i965simple/brw_structs.h create mode 100644 src/mesa/pipe/i965simple/brw_surface.c create mode 100644 src/mesa/pipe/i965simple/brw_tex.c create mode 100644 src/mesa/pipe/i965simple/brw_tex_layout.c create mode 100644 src/mesa/pipe/i965simple/brw_tex_layout.h create mode 100644 src/mesa/pipe/i965simple/brw_urb.c create mode 100644 src/mesa/pipe/i965simple/brw_util.c create mode 100644 src/mesa/pipe/i965simple/brw_util.h create mode 100644 src/mesa/pipe/i965simple/brw_vs.c create mode 100644 src/mesa/pipe/i965simple/brw_vs.h create mode 100644 src/mesa/pipe/i965simple/brw_vs_constval.c create mode 100644 src/mesa/pipe/i965simple/brw_vs_emit.c create mode 100644 src/mesa/pipe/i965simple/brw_vs_state.c create mode 100644 src/mesa/pipe/i965simple/brw_vtbl.c create mode 100644 src/mesa/pipe/i965simple/brw_winsys.h create mode 100644 src/mesa/pipe/i965simple/brw_wm.c create mode 100644 src/mesa/pipe/i965simple/brw_wm.h create mode 100644 src/mesa/pipe/i965simple/brw_wm_debug.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_emit.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_fp.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_glsl.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_iz.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_pass0.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_pass1.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_pass2.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_sampler_state.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_state.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_surface_state.c create mode 100644 src/mesa/pipe/xlib/brw_aub.c create mode 100644 src/mesa/pipe/xlib/brw_aub.h (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 6f090d9b3d..dd8e2c3762 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -11,7 +11,9 @@ GL_MINOR = 5 GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) -SOFTPIPE_LIB = $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a +PIPE_LIB = \ + $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a \ + $(TOP)/src/mesa/pipe/i965simple/libi965simple.a ifeq ($(CONFIG_NAME), linux-cell) CELL_LIB = $(TOP)/src/mesa/pipe/cell/ppu/libcell.a @@ -118,12 +120,12 @@ stand-alone: depend subdirs $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) $(TOP)/$(LIB_DIR)/$ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(SOFTPIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) @ $(TOP)/bin/mklib -o $(GL_LIB) \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ - $(SOFTPIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) + $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) # Make the OSMesa library $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECTS) diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile index 2bf7318d59..c0345a9cb5 100644 --- a/src/mesa/pipe/Makefile +++ b/src/mesa/pipe/Makefile @@ -6,7 +6,7 @@ ifeq ($(CONFIG_NAME), linux-cell) CELL_DIR = cell endif -SUBDIRS = softpipe i915simple failover pipebuffer $(CELL_DIR) +SUBDIRS = softpipe i915simple i965simple failover pipebuffer $(CELL_DIR) default: subdirs diff --git a/src/mesa/pipe/i965simple/Makefile b/src/mesa/pipe/i965simple/Makefile new file mode 100644 index 0000000000..460bc52c35 --- /dev/null +++ b/src/mesa/pipe/i965simple/Makefile @@ -0,0 +1,75 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = i965simple + +DRIVER_SOURCES = \ + brw_blit.c \ + brw_flush.c \ + brw_strings.c \ + brw_surface.c \ + brw_cc.c \ + brw_clip.c \ + brw_clip_line.c \ + brw_clip_point.c \ + brw_clip_state.c \ + brw_clip_tri.c \ + brw_clip_util.c \ + brw_context.c \ + brw_curbe.c \ + brw_draw.c \ + brw_draw_upload.c \ + brw_eu.c \ + brw_eu_debug.c \ + brw_eu_emit.c \ + brw_eu_util.c \ + brw_gs.c \ + brw_gs_emit.c \ + brw_gs_state.c \ + brw_misc_state.c \ + brw_program.c \ + brw_regions.c \ + brw_sf.c \ + brw_sf_emit.c \ + brw_sf_state.c \ + brw_state.c \ + brw_state_batch.c \ + brw_state_cache.c \ + brw_state_pool.c \ + brw_state_upload.c \ + brw_tex.c \ + brw_tex_layout.c \ + brw_urb.c \ + brw_util.c \ + brw_vs.c \ + brw_vs_constval.c \ + brw_vs_emit.c \ + brw_vs_state.c \ + brw_vtbl.c \ + brw_wm.c \ + brw_wm_debug.c \ + brw_wm_emit.c \ + brw_wm_fp.c \ + brw_wm_iz.c \ + brw_wm_glsl.c \ + brw_wm_pass0.c \ + brw_wm_pass1.c \ + brw_wm_pass2.c \ + brw_wm_sampler_state.c \ + brw_wm_state.c \ + brw_wm_surface_state.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(COMMON_BM_SOURCES) \ + $(MINIGLX_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +DRIVER_DEFINES = -I. + +include ../Makefile.template + +symlinks: diff --git a/src/mesa/pipe/i965simple/brw_batch.h b/src/mesa/pipe/i965simple/brw_batch.h new file mode 100644 index 0000000000..7c778f360b --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_batch.h @@ -0,0 +1,57 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef BRW_BATCH_H +#define BRW_BATCH_H + +#include "brw_winsys.h" + +#define BATCH_LOCALS + +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 + +#define BEGIN_BATCH( dwords, relocs ) \ + (brw->batch_start = brw->winsys->batch_start(brw->winsys, dwords, relocs)) + +#define OUT_BATCH( dword ) \ + brw->winsys->batch_dword(brw->winsys, dword) + +#define OUT_RELOC( buf, flags, delta ) \ + brw->winsys->batch_reloc(brw->winsys, buf, flags, delta) + +#define ADVANCE_BATCH() + +/* XXX: this is bogus - need proper handling for out-of-memory in batchbuffer. + */ +#define FLUSH_BATCH(fence) do { \ + brw->winsys->batch_flush(brw->winsys, fence); \ + brw->batch_start = NULL; \ + brw->hardware_dirty = ~0; \ +} while (0) + +#endif diff --git a/src/mesa/pipe/i965simple/brw_blit.c b/src/mesa/pipe/i965simple/brw_blit.c new file mode 100644 index 0000000000..4692129e40 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_blit.c @@ -0,0 +1,217 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include +#include + +#include "brw_batch.h" +#include "brw_blit.h" +#include "brw_context.h" +#include "brw_reg.h" + +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +#define FILE_DEBUG_FLAG DEBUG_BLIT + +void brw_fill_blit(struct brw_context *brw, + unsigned cpp, + short dst_pitch, + struct pipe_buffer_handle *dst_buffer, + unsigned dst_offset, + boolean dst_tiled, + short x, short y, + short w, short h, + unsigned color) +{ + unsigned BR13, CMD; + BATCH_LOCALS; + + dst_pitch *= cpp; + + switch(cpp) { + case 1: + case 2: + case 3: + BR13 = (0xF0 << 16) | (1<<24); + CMD = XY_COLOR_BLT_CMD; + break; + case 4: + BR13 = (0xF0 << 16) | (1<<24) | (1<<25); + CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; + break; + default: + return; + } + + if (dst_tiled) { + CMD |= XY_DST_TILED; + dst_pitch /= 4; + } + + BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH( CMD ); + OUT_BATCH( dst_pitch | BR13 ); + OUT_BATCH( (y << 16) | x ); + OUT_BATCH( ((y+h) << 16) | (x+w) ); + OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, dst_offset ); + OUT_BATCH( color ); + ADVANCE_BATCH(); +} + +static unsigned translate_raster_op(unsigned logicop) +{ + switch(logicop) { + case PIPE_LOGICOP_CLEAR: return 0x00; + case PIPE_LOGICOP_AND: return 0x88; + case PIPE_LOGICOP_AND_REVERSE: return 0x44; + case PIPE_LOGICOP_COPY: return 0xCC; + case PIPE_LOGICOP_AND_INVERTED: return 0x22; + case PIPE_LOGICOP_NOOP: return 0xAA; + case PIPE_LOGICOP_XOR: return 0x66; + case PIPE_LOGICOP_OR: return 0xEE; + case PIPE_LOGICOP_NOR: return 0x11; + case PIPE_LOGICOP_EQUIV: return 0x99; + case PIPE_LOGICOP_INVERT: return 0x55; + case PIPE_LOGICOP_OR_REVERSE: return 0xDD; + case PIPE_LOGICOP_COPY_INVERTED: return 0x33; + case PIPE_LOGICOP_OR_INVERTED: return 0xBB; + case PIPE_LOGICOP_NAND: return 0x77; + case PIPE_LOGICOP_SET: return 0xFF; + default: return 0; + } +} + + +/* Copy BitBlt + */ +void brw_copy_blit(struct brw_context *brw, + unsigned cpp, + short src_pitch, + struct pipe_buffer_handle *src_buffer, + unsigned src_offset, + boolean src_tiled, + short dst_pitch, + struct pipe_buffer_handle *dst_buffer, + unsigned dst_offset, + boolean dst_tiled, + short src_x, short src_y, + short dst_x, short dst_y, + short w, short h, + unsigned logic_op) +{ + unsigned CMD, BR13; + int dst_y2 = dst_y + h; + int dst_x2 = dst_x + w; + BATCH_LOCALS; + + + DBG("%s src:buf(%d)/%d %d,%d dst:buf(%d)/%d %d,%d sz:%dx%d op:%d\n", + __FUNCTION__, + src_buffer, src_pitch, src_x, src_y, + dst_buffer, dst_pitch, dst_x, dst_y, + w,h,logic_op); + + assert( logic_op - PIPE_LOGICOP_CLEAR >= 0 ); + assert( logic_op - PIPE_LOGICOP_CLEAR < 0x10 ); + + src_pitch *= cpp; + dst_pitch *= cpp; + + switch(cpp) { + case 1: + case 2: + case 3: + BR13 = (translate_raster_op(logic_op) << 16) | (1<<24); + CMD = XY_SRC_COPY_BLT_CMD; + break; + case 4: + BR13 = (translate_raster_op(logic_op) << 16) | (1<<24) | + (1<<25); + CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; + break; + default: + return; + } + + if (src_tiled) { + CMD |= XY_SRC_TILED; + src_pitch /= 4; + } + + if (dst_tiled) { + CMD |= XY_DST_TILED; + dst_pitch /= 4; + } + + if (dst_y2 < dst_y || + dst_x2 < dst_x) { + return; + } + + dst_pitch &= 0xffff; + src_pitch &= 0xffff; + + /* Initial y values don't seem to work with negative pitches. If + * we adjust the offsets manually (below), it seems to work fine. + * + * On the other hand, if we always adjust, the hardware doesn't + * know which blit directions to use, so overlapping copypixels get + * the wrong result. + */ + if (dst_pitch > 0 && src_pitch > 0) { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH( CMD ); + OUT_BATCH( dst_pitch | BR13 ); + OUT_BATCH( (dst_y << 16) | dst_x ); + OUT_BATCH( (dst_y2 << 16) | dst_x2 ); + OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, + dst_offset ); + OUT_BATCH( (src_y << 16) | src_x ); + OUT_BATCH( src_pitch ); + OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ, + src_offset ); + ADVANCE_BATCH(); + } + else { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH( CMD ); + OUT_BATCH( (dst_pitch & 0xffff) | BR13 ); + OUT_BATCH( (0 << 16) | dst_x ); + OUT_BATCH( (h << 16) | dst_x2 ); + OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, + dst_offset + dst_y * dst_pitch ); + OUT_BATCH( (src_pitch & 0xffff) ); + OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ, + src_offset + src_y * src_pitch ); + ADVANCE_BATCH(); + } +} + + + diff --git a/src/mesa/pipe/i965simple/brw_blit.h b/src/mesa/pipe/i965simple/brw_blit.h new file mode 100644 index 0000000000..371a135375 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_blit.h @@ -0,0 +1,32 @@ +#ifndef BRW_BLIT_H +#define BRW_BLIT_H + +#include "pipe/p_compiler.h" + +struct pipe_buffer_handle; +struct brw_context; + +void brw_fill_blit(struct brw_context *intel, + unsigned cpp, + short dst_pitch, + struct pipe_buffer_handle *dst_buffer, + unsigned dst_offset, + boolean dst_tiled, + short x, short y, + short w, short h, + unsigned color); +void brw_copy_blit(struct brw_context *intel, + unsigned cpp, + short src_pitch, + struct pipe_buffer_handle *src_buffer, + unsigned src_offset, + boolean src_tiled, + short dst_pitch, + struct pipe_buffer_handle *dst_buffer, + unsigned dst_offset, + boolean dst_tiled, + short src_x, short src_y, + short dst_x, short dst_y, + short w, short h, + unsigned logic_op); +#endif diff --git a/src/mesa/pipe/i965simple/brw_cc.c b/src/mesa/pipe/i965simple/brw_cc.c new file mode 100644 index 0000000000..fc7fdba53f --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_cc.c @@ -0,0 +1,268 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_util.h" + + +static int brw_translate_compare_func(int func) +{ + switch(func) { + case PIPE_FUNC_NEVER: + return BRW_COMPAREFUNCTION_NEVER; + case PIPE_FUNC_LESS: + return BRW_COMPAREFUNCTION_LESS; + case PIPE_FUNC_LEQUAL: + return BRW_COMPAREFUNCTION_LEQUAL; + case PIPE_FUNC_GREATER: + return BRW_COMPAREFUNCTION_GREATER; + case PIPE_FUNC_GEQUAL: + return BRW_COMPAREFUNCTION_GEQUAL; + case PIPE_FUNC_NOTEQUAL: + return BRW_COMPAREFUNCTION_NOTEQUAL; + case PIPE_FUNC_EQUAL: + return BRW_COMPAREFUNCTION_EQUAL; + case PIPE_FUNC_ALWAYS: + return BRW_COMPAREFUNCTION_ALWAYS; + } + + fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); + return BRW_COMPAREFUNCTION_ALWAYS; +} + +static int brw_translate_stencil_op(int op) +{ + switch(op) { + case PIPE_STENCIL_OP_KEEP: + return BRW_STENCILOP_KEEP; + case PIPE_STENCIL_OP_ZERO: + return BRW_STENCILOP_ZERO; + case PIPE_STENCIL_OP_REPLACE: + return BRW_STENCILOP_REPLACE; + case PIPE_STENCIL_OP_INCR: + return BRW_STENCILOP_INCRSAT; + case PIPE_STENCIL_OP_DECR: + return BRW_STENCILOP_DECRSAT; + case PIPE_STENCIL_OP_INCR_WRAP: + return BRW_STENCILOP_INCR; + case PIPE_STENCIL_OP_DECR_WRAP: + return BRW_STENCILOP_DECR; + case PIPE_STENCIL_OP_INVERT: + return BRW_STENCILOP_INVERT; + default: + return BRW_STENCILOP_ZERO; + } +} + + +static int brw_translate_logic_op(int opcode) +{ + switch(opcode) { + case PIPE_LOGICOP_CLEAR: + return BRW_LOGICOPFUNCTION_CLEAR; + case PIPE_LOGICOP_AND: + return BRW_LOGICOPFUNCTION_AND; + case PIPE_LOGICOP_AND_REVERSE: + return BRW_LOGICOPFUNCTION_AND_REVERSE; + case PIPE_LOGICOP_COPY: + return BRW_LOGICOPFUNCTION_COPY; + case PIPE_LOGICOP_COPY_INVERTED: + return BRW_LOGICOPFUNCTION_COPY_INVERTED; + case PIPE_LOGICOP_AND_INVERTED: + return BRW_LOGICOPFUNCTION_AND_INVERTED; + case PIPE_LOGICOP_NOOP: + return BRW_LOGICOPFUNCTION_NOOP; + case PIPE_LOGICOP_XOR: + return BRW_LOGICOPFUNCTION_XOR; + case PIPE_LOGICOP_OR: + return BRW_LOGICOPFUNCTION_OR; + case PIPE_LOGICOP_OR_INVERTED: + return BRW_LOGICOPFUNCTION_OR_INVERTED; + case PIPE_LOGICOP_NOR: + return BRW_LOGICOPFUNCTION_NOR; + case PIPE_LOGICOP_EQUIV: + return BRW_LOGICOPFUNCTION_EQUIV; + case PIPE_LOGICOP_INVERT: + return BRW_LOGICOPFUNCTION_INVERT; + case PIPE_LOGICOP_OR_REVERSE: + return BRW_LOGICOPFUNCTION_OR_REVERSE; + case PIPE_LOGICOP_NAND: + return BRW_LOGICOPFUNCTION_NAND; + case PIPE_LOGICOP_SET: + return BRW_LOGICOPFUNCTION_SET; + default: + return BRW_LOGICOPFUNCTION_SET; + } +} + + +static void upload_cc_vp( struct brw_context *brw ) +{ + struct brw_cc_viewport ccv; + + memset(&ccv, 0, sizeof(ccv)); + + ccv.min_depth = 0.0; + ccv.max_depth = 1.0; + + brw->cc.vp_gs_offset = brw_cache_data( &brw->cache[BRW_CC_VP], &ccv ); +} + +const struct brw_tracked_state brw_cc_vp = { + .dirty = { + .brw = BRW_NEW_CONTEXT, + .cache = 0 + }, + .update = upload_cc_vp +}; + + +static void upload_cc_unit( struct brw_context *brw ) +{ + struct brw_cc_unit_state cc; + + memset(&cc, 0, sizeof(cc)); + + /* BRW_NEW_DEPTH_STENCIL */ + if (brw->attribs.DepthStencil->stencil.front_enabled) { + cc.cc0.stencil_enable = brw->attribs.DepthStencil->stencil.front_enabled; + cc.cc0.stencil_func = brw_translate_compare_func(brw->attribs.DepthStencil->stencil.front_func); + cc.cc0.stencil_fail_op = brw_translate_stencil_op(brw->attribs.DepthStencil->stencil.front_fail_op); + cc.cc0.stencil_pass_depth_fail_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil.front_zfail_op); + cc.cc0.stencil_pass_depth_pass_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil.front_zpass_op); + cc.cc1.stencil_ref = brw->attribs.DepthStencil->stencil.ref_value[0]; + cc.cc1.stencil_write_mask = brw->attribs.DepthStencil->stencil.write_mask[0]; + cc.cc1.stencil_test_mask = brw->attribs.DepthStencil->stencil.value_mask[0]; + + if (brw->attribs.DepthStencil->stencil.back_enabled) { + cc.cc0.bf_stencil_enable = brw->attribs.DepthStencil->stencil.back_enabled; + cc.cc0.bf_stencil_func = brw_translate_compare_func( + brw->attribs.DepthStencil->stencil.back_func); + cc.cc0.bf_stencil_fail_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil.back_fail_op); + cc.cc0.bf_stencil_pass_depth_fail_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil.back_zfail_op); + cc.cc0.bf_stencil_pass_depth_pass_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil.back_zpass_op); + cc.cc1.bf_stencil_ref = brw->attribs.DepthStencil->stencil.ref_value[1]; + cc.cc2.bf_stencil_write_mask = brw->attribs.DepthStencil->stencil.write_mask[1]; + cc.cc2.bf_stencil_test_mask = brw->attribs.DepthStencil->stencil.value_mask[1]; + } + + /* Not really sure about this: + */ + if (brw->attribs.DepthStencil->stencil.write_mask[0] || + (brw->attribs.DepthStencil->stencil.back_enabled && + brw->attribs.DepthStencil->stencil.write_mask[1])) + cc.cc0.stencil_write_enable = 1; + } + + /* BRW_NEW_BLEND */ + if (brw->attribs.Blend->logicop_enable) { + cc.cc2.logicop_enable = 1; + cc.cc5.logicop_func = brw_translate_logic_op( brw->attribs.Blend->logicop_func ); + } + else if (brw->attribs.Blend->blend_enable) { + int eqRGB = brw->attribs.Blend->rgb_func; + int eqA = brw->attribs.Blend->alpha_func; + int srcRGB = brw->attribs.Blend->rgb_src_factor; + int dstRGB = brw->attribs.Blend->rgb_dst_factor; + int srcA = brw->attribs.Blend->alpha_src_factor; + int dstA = brw->attribs.Blend->alpha_dst_factor; + + if (eqRGB == PIPE_BLEND_MIN || eqRGB == PIPE_BLEND_MAX) { + srcRGB = dstRGB = PIPE_BLENDFACTOR_ONE; + } + + if (eqA == PIPE_BLEND_MIN || eqA == PIPE_BLEND_MAX) { + srcA = dstA = PIPE_BLENDFACTOR_ONE; + } + + cc.cc6.dest_blend_factor = brw_translate_blend_factor(dstRGB); + cc.cc6.src_blend_factor = brw_translate_blend_factor(srcRGB); + cc.cc6.blend_function = brw_translate_blend_equation( eqRGB ); + + cc.cc5.ia_dest_blend_factor = brw_translate_blend_factor(dstA); + cc.cc5.ia_src_blend_factor = brw_translate_blend_factor(srcA); + cc.cc5.ia_blend_function = brw_translate_blend_equation( eqA ); + + cc.cc3.blend_enable = 1; + cc.cc3.ia_blend_enable = (srcA != srcRGB || + dstA != dstRGB || + eqA != eqRGB); + } + + /* BRW_NEW_ALPHATEST + */ + if (brw->attribs.AlphaTest->enabled) { + cc.cc3.alpha_test = 1; + cc.cc3.alpha_test_func = brw_translate_compare_func(brw->attribs.AlphaTest->func); + + UNCLAMPED_FLOAT_TO_UBYTE(cc.cc7.alpha_ref.ub[0], brw->attribs.AlphaTest->ref); + + cc.cc3.alpha_test_format = BRW_ALPHATEST_FORMAT_UNORM8; + } + + if (brw->attribs.Blend->dither) { + cc.cc5.dither_enable = 1; + cc.cc6.y_dither_offset = 0; + cc.cc6.x_dither_offset = 0; + } + + if (brw->attribs.DepthStencil->depth.enabled) { + cc.cc2.depth_test = brw->attribs.DepthStencil->depth.enabled; + cc.cc2.depth_test_function = brw_translate_compare_func(brw->attribs.DepthStencil->depth.func); + cc.cc2.depth_write_enable = brw->attribs.DepthStencil->depth.writemask; + } + + /* CACHE_NEW_CC_VP */ + cc.cc4.cc_viewport_state_offset = brw->cc.vp_gs_offset >> 5; + + if (BRW_DEBUG & DEBUG_STATS) + cc.cc5.statistics_enable = 1; + + brw->cc.state_gs_offset = brw_cache_data( &brw->cache[BRW_CC_UNIT], &cc ); +} + +const struct brw_tracked_state brw_cc_unit = { + .dirty = { + .brw = BRW_NEW_DEPTH_STENCIL | BRW_NEW_BLEND | BRW_NEW_ALPHA_TEST, + .cache = CACHE_NEW_CC_VP + }, + .update = upload_cc_unit +}; + diff --git a/src/mesa/pipe/i965simple/brw_clip.c b/src/mesa/pipe/i965simple/brw_clip.c new file mode 100644 index 0000000000..fc86e0a446 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_clip.c @@ -0,0 +1,201 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_state.h" +#include "brw_clip.h" + +#define FRONT_UNFILLED_BIT 0x1 +#define BACK_UNFILLED_BIT 0x2 + + +static void compile_clip_prog( struct brw_context *brw, + struct brw_clip_prog_key *key ) +{ + struct brw_clip_compile c; + const unsigned *program; + unsigned program_size; + unsigned delta; + unsigned i; + + memset(&c, 0, sizeof(c)); + + /* Begin the compilation: + */ + brw_init_compile(&c.func); + + c.func.single_program_flow = 1; + + c.key = *key; + + + /* Need to locate the two positions present in vertex + header. + * These are currently hardcoded: + */ + c.header_position_offset = ATTR_SIZE; + + for (i = 0, delta = REG_SIZE; i < PIPE_MAX_SHADER_OUTPUTS; i++) + if (c.key.attrs & (1<primitive) { + case PIPE_PRIM_TRIANGLES: +#if 0 + if (key->do_unfilled) + brw_emit_unfilled_clip( &c ); + else +#endif + brw_emit_tri_clip( &c ); + break; + case PIPE_PRIM_LINES: + brw_emit_line_clip( &c ); + break; + case PIPE_PRIM_POINTS: + brw_emit_point_clip( &c ); + break; + default: + assert(0); + return; + } + + + + /* get the program + */ + program = brw_get_program(&c.func, &program_size); + + /* Upload + */ + brw->clip.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_CLIP_PROG], + &c.key, + sizeof(c.key), + program, + program_size, + &c.prog_data, + &brw->clip.prog_data ); +} + + +static boolean search_cache( struct brw_context *brw, + struct brw_clip_prog_key *key ) +{ + return brw_search_cache(&brw->cache[BRW_CLIP_PROG], + key, sizeof(*key), + &brw->clip.prog_data, + &brw->clip.prog_gs_offset); +} + + + + +/* Calculate interpolants for triangle and line rasterization. + */ +static void upload_clip_prog(struct brw_context *brw) +{ + struct brw_clip_prog_key key; + + memset(&key, 0, sizeof(key)); + + /* Populate the key: + */ + /* BRW_NEW_REDUCED_PRIMITIVE */ + key.primitive = brw->reduced_primitive; + /* CACHE_NEW_VS_PROG */ + key.attrs = brw->vs.prog_data->outputs_written; + /* BRW_NEW_RASTER */ + key.do_flat_shading = (brw->attribs.Raster->flatshade); + /* BRW_NEW_CLIP */ + key.nr_userclip = brw->attribs.Clip.nr; /* XXX */ + key.clip_mode = BRW_CLIPMODE_NORMAL; + + if (key.primitive == PIPE_PRIM_TRIANGLES) { + if (brw->attribs.Raster->cull_mode == PIPE_WINDING_BOTH) + key.clip_mode = BRW_CLIPMODE_REJECT_ALL; + else { + if (brw->attribs.Raster->fill_cw != PIPE_POLYGON_MODE_FILL || + brw->attribs.Raster->fill_ccw != PIPE_POLYGON_MODE_FILL) + key.do_unfilled = 1; + + /* Most cases the fixed function units will handle. Cases where + * one or more polygon faces are unfilled will require help: + */ + if (key.do_unfilled) { + key.clip_mode = BRW_CLIPMODE_CLIP_NON_REJECTED; + + if (brw->attribs.Raster->offset_cw || + brw->attribs.Raster->offset_ccw) { + key.offset_units = brw->attribs.Raster->offset_units; + key.offset_factor = brw->attribs.Raster->offset_scale; + } + key.fill_ccw = brw->attribs.Raster->fill_ccw; + key.fill_cw = brw->attribs.Raster->fill_cw; + key.offset_ccw = brw->attribs.Raster->offset_ccw; + key.offset_cw = brw->attribs.Raster->offset_cw; + if (brw->attribs.Raster->light_twoside && + key.fill_cw != CLIP_CULL) + key.copy_bfc_cw = 1; + } + } + } + + if (!search_cache(brw, &key)) + compile_clip_prog( brw, &key ); +} + +const struct brw_tracked_state brw_clip_prog = { + .dirty = { + .brw = (BRW_NEW_RASTERIZER | + BRW_NEW_CLIP | + BRW_NEW_REDUCED_PRIMITIVE), + .cache = CACHE_NEW_VS_PROG + }, + .update = upload_clip_prog +}; diff --git a/src/mesa/pipe/i965simple/brw_clip.h b/src/mesa/pipe/i965simple/brw_clip.h new file mode 100644 index 0000000000..a89d08b791 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_clip.h @@ -0,0 +1,170 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#ifndef BRW_CLIP_H +#define BRW_CLIP_H + + +#include "brw_context.h" +#include "brw_eu.h" + +#define MAX_VERTS (3+6+6) + +/* Note that if unfilled primitives are being emitted, we have to fix + * up polygon offset and flatshading at this point: + */ +struct brw_clip_prog_key { + unsigned attrs:32; + unsigned primitive:4; + unsigned nr_userclip:3; + unsigned do_flat_shading:1; + unsigned do_unfilled:1; + unsigned fill_cw:2; /* includes cull information */ + unsigned fill_ccw:2; /* includes cull information */ + unsigned offset_cw:1; + unsigned offset_ccw:1; + unsigned pad0:17; + + unsigned copy_bfc_cw:1; + unsigned copy_bfc_ccw:1; + unsigned clip_mode:3; + unsigned pad1:27; + + float offset_factor; + float offset_units; +}; + + +#define CLIP_LINE 0 +#define CLIP_POINT 1 +#define CLIP_FILL 2 +#define CLIP_CULL 3 + + +#define PRIM_MASK (0x1f) + +struct brw_clip_compile { + struct brw_compile func; + struct brw_clip_prog_key key; + struct brw_clip_prog_data prog_data; + + struct { + struct brw_reg R0; + struct brw_reg vertex[MAX_VERTS]; + + struct brw_reg t; + struct brw_reg t0, t1; + struct brw_reg dp0, dp1; + + struct brw_reg dpPrev; + struct brw_reg dp; + struct brw_reg loopcount; + struct brw_reg nr_verts; + struct brw_reg planemask; + + struct brw_reg inlist; + struct brw_reg outlist; + struct brw_reg freelist; + + struct brw_reg dir; + struct brw_reg tmp0, tmp1; + struct brw_reg offset; + + struct brw_reg fixed_planes; + struct brw_reg plane_equation; + } reg; + + /* 3 different ways of expressing vertex size: + */ + unsigned nr_attrs; + unsigned nr_regs; + unsigned nr_bytes; + + unsigned first_tmp; + unsigned last_tmp; + + boolean need_direction; + + unsigned last_mrf; + + unsigned header_position_offset; + unsigned offset[PIPE_ATTRIB_MAX]; +}; + +#define ATTR_SIZE (4*4) + +/* Points are only culled, so no need for a clip routine, however it + * works out easier to have a dummy one. + */ +void brw_emit_unfilled_clip( struct brw_clip_compile *c ); +void brw_emit_tri_clip( struct brw_clip_compile *c ); +void brw_emit_line_clip( struct brw_clip_compile *c ); +void brw_emit_point_clip( struct brw_clip_compile *c ); + +/* brw_clip_tri.c, for use by the unfilled clip routine: + */ +void brw_clip_tri_init_vertices( struct brw_clip_compile *c ); +void brw_clip_tri_flat_shade( struct brw_clip_compile *c ); +void brw_clip_tri( struct brw_clip_compile *c ); +void brw_clip_tri_emit_polygon( struct brw_clip_compile *c ); +void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, + unsigned nr_verts ); + + +/* Utils: + */ + +void brw_clip_interp_vertex( struct brw_clip_compile *c, + struct brw_indirect dest_ptr, + struct brw_indirect v0_ptr, /* from */ + struct brw_indirect v1_ptr, /* to */ + struct brw_reg t0, + boolean force_edgeflag ); + +void brw_clip_init_planes( struct brw_clip_compile *c ); + +void brw_clip_emit_vue(struct brw_clip_compile *c, + struct brw_indirect vert, + boolean allocate, + boolean eot, + unsigned header); + +void brw_clip_kill_thread(struct brw_clip_compile *c); + +struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c ); +struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c ); + +void brw_clip_copy_colors( struct brw_clip_compile *c, + unsigned to, unsigned from ); + +void brw_clip_init_clipmask( struct brw_clip_compile *c ); + +#endif diff --git a/src/mesa/pipe/i965simple/brw_clip_line.c b/src/mesa/pipe/i965simple/brw_clip_line.c new file mode 100644 index 0000000000..75d9e5fcda --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_clip_line.c @@ -0,0 +1,245 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + + +static void brw_clip_line_alloc_regs( struct brw_clip_compile *c ) +{ + unsigned i = 0,j; + + /* Register usage is static, precompute here: + */ + c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; + + if (c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec4_grf(i, 0); + i += (6 + c->key.nr_userclip + 1) / 2; + + c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2; + } + else + c->prog_data.curb_read_length = 0; + + + /* Payload vertices plus space for more generated vertices: + */ + for (j = 0; j < 4; j++) { + c->reg.vertex[j] = brw_vec4_grf(i, 0); + i += c->nr_regs; + } + + c->reg.t = brw_vec1_grf(i, 0); + c->reg.t0 = brw_vec1_grf(i, 1); + c->reg.t1 = brw_vec1_grf(i, 2); + c->reg.planemask = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD); + c->reg.plane_equation = brw_vec4_grf(i, 4); + i++; + + c->reg.dp0 = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */ + c->reg.dp1 = brw_vec1_grf(i, 4); + i++; + + if (!c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec8_grf(i, 0); + i++; + } + + + c->first_tmp = i; + c->last_tmp = i; + + c->prog_data.urb_read_length = c->nr_regs; /* ? */ + c->prog_data.total_grf = i; +} + + + +/* Line clipping, more or less following the following algorithm: + * + * for (p=0;p t1) t1 = t; + * } else { + * float t = dp0 / (dp0 - dp1); + * if (t > t0) t0 = t; + * } + * + * if (t0 + t1 >= 1.0) + * return; + * } + * } + * + * interp( ctx, newvtx0, vtx0, vtx1, t0 ); + * interp( ctx, newvtx1, vtx1, vtx0, t1 ); + * + */ +static void clip_and_emit_line( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_indirect vtx0 = brw_indirect(0, 0); + struct brw_indirect vtx1 = brw_indirect(1, 0); + struct brw_indirect newvtx0 = brw_indirect(2, 0); + struct brw_indirect newvtx1 = brw_indirect(3, 0); + struct brw_indirect plane_ptr = brw_indirect(4, 0); + struct brw_instruction *plane_loop; + struct brw_instruction *plane_active; + struct brw_instruction *is_negative; + struct brw_instruction *is_neg2; + struct brw_instruction *not_culled; + struct brw_reg v1_null_ud = retype(vec1(brw_null_reg()), BRW_REGISTER_TYPE_UD); + + brw_MOV(p, get_addr_reg(vtx0), brw_address(c->reg.vertex[0])); + brw_MOV(p, get_addr_reg(vtx1), brw_address(c->reg.vertex[1])); + brw_MOV(p, get_addr_reg(newvtx0), brw_address(c->reg.vertex[2])); + brw_MOV(p, get_addr_reg(newvtx1), brw_address(c->reg.vertex[3])); + brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c)); + + /* Note: init t0, t1 together: + */ + brw_MOV(p, vec2(c->reg.t0), brw_imm_f(0)); + + brw_clip_init_planes(c); + brw_clip_init_clipmask(c); + + /* -ve rhw workaround */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), + brw_imm_ud(1<<20)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(0x3f)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + plane_loop = brw_DO(p, BRW_EXECUTE_1); + { + /* if (planemask & 1) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, v1_null_ud, c->reg.planemask, brw_imm_ud(1)); + + plane_active = brw_IF(p, BRW_EXECUTE_1); + { + if (c->key.nr_userclip) + brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0)); + else + brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0)); + +#if 0 + /* dp = DP4(vtx->position, plane) + */ + brw_DP4(p, vec4(c->reg.dp0), deref_4f(vtx0, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); + + /* if (IS_NEGATIVE(dp1)) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, vec4(c->reg.dp1), deref_4f(vtx1, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); +#else + #warning "disabled" +#endif + is_negative = brw_IF(p, BRW_EXECUTE_1); + { + brw_ADD(p, c->reg.t, c->reg.dp1, negate(c->reg.dp0)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp1); + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t1 ); + brw_MOV(p, c->reg.t1, c->reg.t); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + is_negative = brw_ELSE(p, is_negative); + { + /* Coming back in. We know that both cannot be negative + * because the line would have been culled in that case. + */ + + /* If both are positive, do nothing */ + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.dp0, brw_imm_f(0.0)); + is_neg2 = brw_IF(p, BRW_EXECUTE_1); + { + brw_ADD(p, c->reg.t, c->reg.dp0, negate(c->reg.dp1)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp0); + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t0 ); + brw_MOV(p, c->reg.t0, c->reg.t); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + brw_ENDIF(p, is_neg2); + } + brw_ENDIF(p, is_negative); + } + brw_ENDIF(p, plane_active); + + /* plane_ptr++; + */ + brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c)); + + /* while (planemask>>=1) != 0 + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1)); + } + brw_WHILE(p, plane_loop); + + brw_ADD(p, c->reg.t, c->reg.t0, c->reg.t1); + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.t, brw_imm_f(1.0)); + not_culled = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_interp_vertex(c, newvtx0, vtx0, vtx1, c->reg.t0, FALSE); + brw_clip_interp_vertex(c, newvtx1, vtx1, vtx0, c->reg.t1, FALSE); + + brw_clip_emit_vue(c, newvtx0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START); + brw_clip_emit_vue(c, newvtx1, 0, 1, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END); + } + brw_ENDIF(p, not_culled); + brw_clip_kill_thread(c); +} + + + +void brw_emit_line_clip( struct brw_clip_compile *c ) +{ + brw_clip_line_alloc_regs(c); + + if (c->key.do_flat_shading) + brw_clip_copy_colors(c, 0, 1); + + clip_and_emit_line(c); +} diff --git a/src/mesa/pipe/i965simple/brw_clip_point.c b/src/mesa/pipe/i965simple/brw_clip_point.c new file mode 100644 index 0000000000..6fce7210d1 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_clip_point.c @@ -0,0 +1,47 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + +/* Point clipping, nothing to do? + */ +void brw_emit_point_clip( struct brw_clip_compile *c ) +{ + /* Send an empty message to kill the thread: + */ + brw_clip_tri_alloc_regs(c, 0); + brw_clip_kill_thread(c); +} diff --git a/src/mesa/pipe/i965simple/brw_clip_state.c b/src/mesa/pipe/i965simple/brw_clip_state.c new file mode 100644 index 0000000000..51a4666a0b --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_clip_state.c @@ -0,0 +1,92 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + + + +static void upload_clip_unit( struct brw_context *brw ) +{ + struct brw_clip_unit_state clip; + + memset(&clip, 0, sizeof(clip)); + + /* CACHE_NEW_CLIP_PROG */ + clip.thread0.grf_reg_count = + ALIGN(brw->clip.prog_data->total_grf, 16) / 16 - 1; + clip.thread0.kernel_start_pointer = brw->clip.prog_gs_offset >> 6; + clip.thread3.urb_entry_read_length = brw->clip.prog_data->urb_read_length; + clip.thread3.const_urb_entry_read_length = brw->clip.prog_data->curb_read_length; + clip.clip5.clip_mode = brw->clip.prog_data->clip_mode; + + /* BRW_NEW_CURBE_OFFSETS */ + clip.thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2; + + /* BRW_NEW_URB_FENCE */ + clip.thread4.nr_urb_entries = brw->urb.nr_clip_entries; + clip.thread4.urb_entry_allocation_size = brw->urb.vsize - 1; + clip.thread4.max_threads = 1; /* 2 threads */ + + if (BRW_DEBUG & DEBUG_STATS) + clip.thread4.stats_enable = 1; + + /* CONSTANT */ + clip.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + clip.thread1.single_program_flow = 1; + clip.thread3.dispatch_grf_start_reg = 1; + clip.thread3.urb_entry_read_offset = 0; + clip.clip5.userclip_enable_flags = 0x7f; + clip.clip5.userclip_must_clip = 1; + clip.clip5.guard_band_enable = 0; + clip.clip5.viewport_z_clip_enable = 1; + clip.clip5.viewport_xy_clip_enable = 1; + clip.clip5.vertex_position_space = BRW_CLIP_NDCSPACE; + clip.clip5.api_mode = BRW_CLIP_API_OGL; + clip.clip6.clipper_viewport_state_ptr = 0; + clip.viewport_xmin = -1; + clip.viewport_xmax = 1; + clip.viewport_ymin = -1; + clip.viewport_ymax = 1; + + brw->clip.state_gs_offset = brw_cache_data( &brw->cache[BRW_CLIP_UNIT], &clip ); +} + + +const struct brw_tracked_state brw_clip_unit = { + .dirty = { + .brw = (BRW_NEW_CURBE_OFFSETS | + BRW_NEW_URB_FENCE), + .cache = CACHE_NEW_CLIP_PROG + }, + .update = upload_clip_unit +}; diff --git a/src/mesa/pipe/i965simple/brw_clip_tri.c b/src/mesa/pipe/i965simple/brw_clip_tri.c new file mode 100644 index 0000000000..c5da7b825e --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_clip_tri.c @@ -0,0 +1,566 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + +static struct brw_reg get_tmp( struct brw_clip_compile *c ) +{ + struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0); + + if (++c->last_tmp > c->prog_data.total_grf) + c->prog_data.total_grf = c->last_tmp; + + return tmp; +} + +static void release_tmps( struct brw_clip_compile *c ) +{ + c->last_tmp = c->first_tmp; +} + + +void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, + unsigned nr_verts ) +{ + unsigned i = 0,j; + + /* Register usage is static, precompute here: + */ + c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; + + if (c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec4_grf(i, 0); + i += (6 + c->key.nr_userclip + 1) / 2; + + c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2; + } + else + c->prog_data.curb_read_length = 0; + + + /* Payload vertices plus space for more generated vertices: + */ + for (j = 0; j < nr_verts; j++) { + c->reg.vertex[j] = brw_vec4_grf(i, 0); + i += c->nr_regs; + } + + if (c->nr_attrs & 1) { + for (j = 0; j < 3; j++) { + unsigned delta = c->nr_attrs*16 + 32; + brw_MOV(&c->func, byte_offset(c->reg.vertex[j], delta), brw_imm_f(0)); + } + } + + c->reg.t = brw_vec1_grf(i, 0); + c->reg.loopcount = retype(brw_vec1_grf(i, 1), BRW_REGISTER_TYPE_UD); + c->reg.nr_verts = retype(brw_vec1_grf(i, 2), BRW_REGISTER_TYPE_UD); + c->reg.planemask = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD); + c->reg.plane_equation = brw_vec4_grf(i, 4); + i++; + + c->reg.dpPrev = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */ + c->reg.dp = brw_vec1_grf(i, 4); + i++; + + c->reg.inlist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); + i++; + + c->reg.outlist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); + i++; + + c->reg.freelist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); + i++; + + if (!c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec8_grf(i, 0); + i++; + } + + if (c->key.do_unfilled) { + c->reg.dir = brw_vec4_grf(i, 0); + c->reg.offset = brw_vec4_grf(i, 4); + i++; + c->reg.tmp0 = brw_vec4_grf(i, 0); + c->reg.tmp1 = brw_vec4_grf(i, 4); + i++; + } + + c->first_tmp = i; + c->last_tmp = i; + + c->prog_data.urb_read_length = c->nr_regs; /* ? */ + c->prog_data.total_grf = i; +} + + + +void brw_clip_tri_init_vertices( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */ + struct brw_instruction *is_rev; + + /* Initial list of indices for incoming vertexes: + */ + brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_EQ, + tmp0, + brw_imm_ud(_3DPRIM_TRISTRIP_REVERSE)); + + /* XXX: Is there an easier way to do this? Need to reverse every + * second tristrip element: Can ignore sometimes? + */ + is_rev = brw_IF(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[1]) ); + brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[0]) ); + if (c->need_direction) + brw_MOV(p, c->reg.dir, brw_imm_f(-1)); + } + is_rev = brw_ELSE(p, is_rev); + { + brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[0]) ); + brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[1]) ); + if (c->need_direction) + brw_MOV(p, c->reg.dir, brw_imm_f(1)); + } + brw_ENDIF(p, is_rev); + + brw_MOV(p, get_element(c->reg.inlist, 2), brw_address(c->reg.vertex[2]) ); + brw_MOV(p, brw_vec8_grf(c->reg.outlist.nr, 0), brw_imm_f(0)); + brw_MOV(p, c->reg.nr_verts, brw_imm_ud(3)); +} + + + +void brw_clip_tri_flat_shade( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *is_poly; + struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */ + + brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_EQ, + tmp0, + brw_imm_ud(_3DPRIM_POLYGON)); + + is_poly = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_copy_colors(c, 1, 0); + brw_clip_copy_colors(c, 2, 0); + } + is_poly = brw_ELSE(p, is_poly); + { + brw_clip_copy_colors(c, 0, 2); + brw_clip_copy_colors(c, 1, 2); + } + brw_ENDIF(p, is_poly); +} + + + +/* Use mesa's clipping algorithms, translated to GEN4 assembly. + */ +void brw_clip_tri( struct brw_clip_compile *c ) +{ +#if 0 + struct brw_compile *p = &c->func; + struct brw_indirect vtx = brw_indirect(0, 0); + struct brw_indirect vtxPrev = brw_indirect(1, 0); + struct brw_indirect vtxOut = brw_indirect(2, 0); + struct brw_indirect plane_ptr = brw_indirect(3, 0); + struct brw_indirect inlist_ptr = brw_indirect(4, 0); + struct brw_indirect outlist_ptr = brw_indirect(5, 0); + struct brw_indirect freelist_ptr = brw_indirect(6, 0); + struct brw_instruction *plane_loop; + struct brw_instruction *plane_active; + struct brw_instruction *vertex_loop; + struct brw_instruction *next_test; + struct brw_instruction *prev_test; + + brw_MOV(p, get_addr_reg(vtxPrev), brw_address(c->reg.vertex[2]) ); + brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c)); + brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist)); + brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist)); + + brw_MOV(p, get_addr_reg(freelist_ptr), brw_address(c->reg.vertex[3]) ); + + plane_loop = brw_DO(p, BRW_EXECUTE_1); + { + /* if (planemask & 1) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, vec1(brw_null_reg()), c->reg.planemask, brw_imm_ud(1)); + + plane_active = brw_IF(p, BRW_EXECUTE_1); + { + /* vtxOut = freelist_ptr++ + */ + brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(freelist_ptr) ); + brw_ADD(p, get_addr_reg(freelist_ptr), get_addr_reg(freelist_ptr), brw_imm_uw(c->nr_regs * REG_SIZE)); + + if (c->key.nr_userclip) + brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0)); + else + brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0)); + + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, c->reg.nr_verts, brw_imm_ud(0)); + + vertex_loop = brw_DO(p, BRW_EXECUTE_1); + { + /* vtx = *input_ptr; + */ + brw_MOV(p, get_addr_reg(vtx), deref_1uw(inlist_ptr, 0)); + + /* IS_NEGATIVE(prev) */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, vec4(c->reg.dpPrev), deref_4f(vtxPrev, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); + prev_test = brw_IF(p, BRW_EXECUTE_1); + { + /* IS_POSITIVE(next) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_GE); + brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); + next_test = brw_IF(p, BRW_EXECUTE_1); + { + + /* Coming back in. + */ + brw_ADD(p, c->reg.t, c->reg.dpPrev, negate(c->reg.dp)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dpPrev); + + /* If (vtxOut == 0) vtxOut = vtxPrev + */ + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) ); + brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtxPrev) ); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_clip_interp_vertex(c, vtxOut, vtxPrev, vtx, c->reg.t, FALSE); + + /* *outlist_ptr++ = vtxOut; + * nr_verts++; + * vtxOut = 0; + */ + brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut)); + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); + brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); + brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) ); + } + brw_ENDIF(p, next_test); + + } + prev_test = brw_ELSE(p, prev_test); + { + /* *outlist_ptr++ = vtxPrev; + * nr_verts++; + */ + brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxPrev)); + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); + brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); + + /* IS_NEGATIVE(next) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); + next_test = brw_IF(p, BRW_EXECUTE_1); + { + /* Going out of bounds. Avoid division by zero as we + * know dp != dpPrev from DIFFERENT_SIGNS, above. + */ + brw_ADD(p, c->reg.t, c->reg.dp, negate(c->reg.dpPrev)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp); + + /* If (vtxOut == 0) vtxOut = vtx + */ + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) ); + brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtx) ); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_clip_interp_vertex(c, vtxOut, vtx, vtxPrev, c->reg.t, TRUE); + + /* *outlist_ptr++ = vtxOut; + * nr_verts++; + * vtxOut = 0; + */ + brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut)); + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); + brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); + brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) ); + } + brw_ENDIF(p, next_test); + } + brw_ENDIF(p, prev_test); + + /* vtxPrev = vtx; + * inlist_ptr++; + */ + brw_MOV(p, get_addr_reg(vtxPrev), get_addr_reg(vtx)); + brw_ADD(p, get_addr_reg(inlist_ptr), get_addr_reg(inlist_ptr), brw_imm_uw(sizeof(short))); + + /* while (--loopcount != 0) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, vertex_loop); + + /* vtxPrev = *(outlist_ptr-1) OR: outlist[nr_verts-1] + * inlist = outlist + * inlist_ptr = &inlist[0] + * outlist_ptr = &outlist[0] + */ + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_w(-2)); + brw_MOV(p, get_addr_reg(vtxPrev), deref_1uw(outlist_ptr, 0)); + brw_MOV(p, brw_vec8_grf(c->reg.inlist.nr, 0), brw_vec8_grf(c->reg.outlist.nr, 0)); + brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist)); + brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist)); + } + brw_ENDIF(p, plane_active); + + /* plane_ptr++; + */ + brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c)); + + /* nr_verts >= 3 + */ + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_GE, + c->reg.nr_verts, + brw_imm_ud(3)); + + /* && (planemask>>=1) != 0 + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1)); + } + brw_WHILE(p, plane_loop); +#else + #warning "disabled" +#endif +} + + + +void brw_clip_tri_emit_polygon(struct brw_clip_compile *c) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *loop, *if_insn; + + /* for (loopcount = nr_verts-2; loopcount > 0; loopcount--) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_G); + brw_ADD(p, + c->reg.loopcount, + c->reg.nr_verts, + brw_imm_d(-2)); + + if_insn = brw_IF(p, BRW_EXECUTE_1); + { + struct brw_indirect v0 = brw_indirect(0, 0); + struct brw_indirect vptr = brw_indirect(1, 0); + + brw_MOV(p, get_addr_reg(vptr), brw_address(c->reg.inlist)); + brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); + + brw_clip_emit_vue(c, v0, 1, 0, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_START)); + + brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2)); + brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_TRIFAN << 2)); + + brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2)); + brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); + + brw_clip_emit_vue(c, v0, 0, 1, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_END)); + } + brw_ENDIF(p, if_insn); +} + +static void do_clip_tri( struct brw_clip_compile *c ) +{ + brw_clip_init_planes(c); + + brw_clip_tri(c); +} + + +static void maybe_do_clip_tri( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *do_clip; + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0)); + do_clip = brw_IF(p, BRW_EXECUTE_1); + { + do_clip_tri(c); + } + brw_ENDIF(p, do_clip); +} + +static void brw_clip_test( struct brw_clip_compile *c ) +{ +#if 0 + struct brw_reg t = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + struct brw_reg t1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + struct brw_reg t2 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + struct brw_reg t3 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + + struct brw_reg v0 = get_tmp(c); + struct brw_reg v1 = get_tmp(c); + struct brw_reg v2 = get_tmp(c); + + struct brw_indirect vt0 = brw_indirect(0, 0); + struct brw_indirect vt1 = brw_indirect(1, 0); + struct brw_indirect vt2 = brw_indirect(2, 0); + + struct brw_compile *p = &c->func; + + brw_MOV(p, get_addr_reg(vt0), brw_address(c->reg.vertex[0])); + brw_MOV(p, get_addr_reg(vt1), brw_address(c->reg.vertex[1])); + brw_MOV(p, get_addr_reg(vt2), brw_address(c->reg.vertex[2])); + brw_MOV(p, v0, deref_4f(vt0, c->offset[VERT_RESULT_HPOS])); + brw_MOV(p, v1, deref_4f(vt1, c->offset[VERT_RESULT_HPOS])); + brw_MOV(p, v2, deref_4f(vt2, c->offset[VERT_RESULT_HPOS])); + + /* test nearz, xmin, ymin plane */ + brw_CMP(p, t1, BRW_CONDITIONAL_LE, negate(v0), get_element(v0, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t2, BRW_CONDITIONAL_LE, negate(v1), get_element(v1, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t3, BRW_CONDITIONAL_LE, negate(v2), get_element(v2, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_XOR(p, t, t1, t2); + brw_XOR(p, t1, t2, t3); + brw_OR(p, t, t, t1); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 0), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<5))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 1), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<3))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 2), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<1))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + /* test farz, xmax, ymax plane */ + brw_CMP(p, t1, BRW_CONDITIONAL_L, v0, get_element(v0, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t2, BRW_CONDITIONAL_L, v1, get_element(v1, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t3, BRW_CONDITIONAL_L, v2, get_element(v2, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_XOR(p, t, t1, t2); + brw_XOR(p, t1, t2, t3); + brw_OR(p, t, t, t1); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 0), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<4))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 1), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<2))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 2), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<0))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + release_tmps(c); +#else + #warning "disabled" +#endif +} + + +void brw_emit_tri_clip( struct brw_clip_compile *c ) +{ + struct brw_instruction *neg_rhw; + struct brw_compile *p = &c->func; + brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6); + brw_clip_tri_init_vertices(c); + brw_clip_init_clipmask(c); + + /* if -ve rhw workaround bit is set, + do cliptest */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), + brw_imm_ud(1<<20)); + neg_rhw = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_test(c); + } + brw_ENDIF(p, neg_rhw); + + /* Can't push into do_clip_tri because with polygon (or quad) + * flatshading, need to apply the flatshade here because we don't + * respect the PV when converting to trifan for emit: + */ + if (c->key.do_flat_shading) + brw_clip_tri_flat_shade(c); + + if (c->key.clip_mode == BRW_CLIPMODE_NORMAL) + do_clip_tri(c); + else + maybe_do_clip_tri(c); + + brw_clip_tri_emit_polygon(c); + + /* Send an empty message to kill the thread: + */ + brw_clip_kill_thread(c); +} diff --git a/src/mesa/pipe/i965simple/brw_clip_unfilled.c b/src/mesa/pipe/i965simple/brw_clip_unfilled.c new file mode 100644 index 0000000000..b774a76dd6 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_clip_unfilled.c @@ -0,0 +1,477 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + + +/* This is performed against the original triangles, so no indirection + * required: +BZZZT! + */ +static void compute_tri_direction( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg e = c->reg.tmp0; + struct brw_reg f = c->reg.tmp1; + struct brw_reg v0 = byte_offset(c->reg.vertex[0], c->offset[VERT_RESULT_HPOS]); + struct brw_reg v1 = byte_offset(c->reg.vertex[1], c->offset[VERT_RESULT_HPOS]); + struct brw_reg v2 = byte_offset(c->reg.vertex[2], c->offset[VERT_RESULT_HPOS]); + + + /* Calculate the vectors of two edges of the triangle: + */ + brw_ADD(p, e, v0, negate(v2)); + brw_ADD(p, f, v1, negate(v2)); + + /* Take their crossproduct: + */ + brw_set_access_mode(p, BRW_ALIGN_16); + brw_MUL(p, vec4(brw_null_reg()), brw_swizzle(e, 1,2,0,3), brw_swizzle(f,2,0,1,3)); + brw_MAC(p, vec4(e), negate(brw_swizzle(e, 2,0,1,3)), brw_swizzle(f,1,2,0,3)); + brw_set_access_mode(p, BRW_ALIGN_1); + + brw_MUL(p, c->reg.dir, c->reg.dir, vec4(e)); +} + + +static void cull_direction( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *ccw; + unsigned conditional; + + assert (!(c->key.fill_ccw == CLIP_CULL && + c->key.fill_cw == CLIP_CULL)); + + if (c->key.fill_ccw == CLIP_CULL) + conditional = BRW_CONDITIONAL_GE; + else + conditional = BRW_CONDITIONAL_L; + + brw_CMP(p, + vec1(brw_null_reg()), + conditional, + get_element(c->reg.dir, 2), + brw_imm_f(0)); + + ccw = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_kill_thread(c); + } + brw_ENDIF(p, ccw); +} + + + +static void copy_bfc( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *ccw; + unsigned conditional; + + /* Do we have any colors to copy? + */ + if (!(c->offset[VERT_RESULT_COL0] && c->offset[VERT_RESULT_BFC0]) && + !(c->offset[VERT_RESULT_COL1] && c->offset[VERT_RESULT_BFC1])) + return; + + /* In some wierd degnerate cases we can end up testing the + * direction twice, once for culling and once for bfc copying. Oh + * well, that's what you get for setting wierd GL state. + */ + if (c->key.copy_bfc_ccw) + conditional = BRW_CONDITIONAL_GE; + else + conditional = BRW_CONDITIONAL_L; + + brw_CMP(p, + vec1(brw_null_reg()), + conditional, + get_element(c->reg.dir, 2), + brw_imm_f(0)); + + ccw = brw_IF(p, BRW_EXECUTE_1); + { + unsigned i; + + for (i = 0; i < 3; i++) { + if (c->offset[VERT_RESULT_COL0] && c->offset[VERT_RESULT_BFC0]) + brw_MOV(p, + byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_COL0]), + byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_BFC0])); + + if (c->offset[VERT_RESULT_COL1] && c->offset[VERT_RESULT_BFC1]) + brw_MOV(p, + byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_COL1]), + byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_BFC1])); + } + } + brw_ENDIF(p, ccw); +} + + + + +/* + float iz = 1.0 / dir.z; + float ac = dir.x * iz; + float bc = dir.y * iz; + offset = ctx->Polygon.OffsetUnits * DEPTH_SCALE; + offset += MAX2( abs(ac), abs(bc) ) * ctx->Polygon.OffsetFactor; + offset *= MRD; +*/ +static void compute_offset( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg off = c->reg.offset; + struct brw_reg dir = c->reg.dir; + + brw_math_invert(p, get_element(off, 2), get_element(dir, 2)); + brw_MUL(p, vec2(off), dir, get_element(off, 2)); + + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_GE, + brw_abs(get_element(off, 0)), + brw_abs(get_element(off, 1))); + + brw_SEL(p, vec1(off), brw_abs(get_element(off, 0)), brw_abs(get_element(off, 1))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_MUL(p, vec1(off), off, brw_imm_f(c->key.offset_factor)); + brw_ADD(p, vec1(off), off, brw_imm_f(c->key.offset_units)); +} + + +static void merge_edgeflags( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *is_poly; + struct brw_reg tmp0 = get_element_ud(c->reg.tmp0, 0); + + brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_EQ, + tmp0, + brw_imm_ud(_3DPRIM_POLYGON)); + + /* Get away with using reg.vertex because we know that this is not + * a _3DPRIM_TRISTRIP_REVERSE: + */ + is_poly = brw_IF(p, BRW_EXECUTE_1); + { + brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ); + brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<8)); + brw_MOV(p, byte_offset(c->reg.vertex[0], c->offset[VERT_RESULT_EDGE]), brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ); + brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<9)); + brw_MOV(p, byte_offset(c->reg.vertex[2], c->offset[VERT_RESULT_EDGE]), brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + brw_ENDIF(p, is_poly); +} + + + +static void apply_one_offset( struct brw_clip_compile *c, + struct brw_indirect vert ) +{ + struct brw_compile *p = &c->func; + struct brw_reg pos = deref_4f(vert, c->offset[VERT_RESULT_HPOS]); + struct brw_reg z = get_element(pos, 2); + + brw_ADD(p, z, z, vec1(c->reg.offset)); +} + + + +/*********************************************************************** + * Output clipped polygon as an unfilled primitive: + */ +static void emit_lines(struct brw_clip_compile *c, + boolean do_offset) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *loop; + struct brw_instruction *draw_edge; + struct brw_indirect v0 = brw_indirect(0, 0); + struct brw_indirect v1 = brw_indirect(1, 0); + struct brw_indirect v0ptr = brw_indirect(2, 0); + struct brw_indirect v1ptr = brw_indirect(3, 0); + + /* Need a seperate loop for offset: + */ + if (do_offset) { + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); + brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); + + apply_one_offset(c, v0); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_G); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); + } + + /* v1ptr = &inlist[nr_verts] + * *v1ptr = v0 + */ + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); + brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v0ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW)); + brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v1ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW)); + brw_MOV(p, deref_1uw(v1ptr, 0), deref_1uw(v0ptr, 0)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); + brw_MOV(p, get_addr_reg(v1), deref_1uw(v0ptr, 2)); + brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); + + /* draw edge if edgeflag != 0 */ + brw_CMP(p, + vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, + deref_1f(v0, c->offset[VERT_RESULT_EDGE]), + brw_imm_f(0)); + draw_edge = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START); + brw_clip_emit_vue(c, v1, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END); + } + brw_ENDIF(p, draw_edge); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); +} + + + +static void emit_points(struct brw_clip_compile *c, + boolean do_offset ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *loop; + struct brw_instruction *draw_point; + + struct brw_indirect v0 = brw_indirect(0, 0); + struct brw_indirect v0ptr = brw_indirect(2, 0); + + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); + brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); + + /* draw if edgeflag != 0 + */ + brw_CMP(p, + vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, + deref_1f(v0, c->offset[VERT_RESULT_EDGE]), + brw_imm_f(0)); + draw_point = brw_IF(p, BRW_EXECUTE_1); + { + if (do_offset) + apply_one_offset(c, v0); + + brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END); + } + brw_ENDIF(p, draw_point); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); +} + + + + + + + +static void emit_primitives( struct brw_clip_compile *c, + unsigned mode, + boolean do_offset ) +{ + switch (mode) { + case CLIP_FILL: + brw_clip_tri_emit_polygon(c); + break; + + case CLIP_LINE: + emit_lines(c, do_offset); + break; + + case CLIP_POINT: + emit_points(c, do_offset); + break; + + case CLIP_CULL: + assert(0); + break; + } +} + + + +static void emit_unfilled_primitives( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *ccw; + + /* Direction culling has already been done. + */ + if (c->key.fill_ccw != c->key.fill_cw && + c->key.fill_ccw != CLIP_CULL && + c->key.fill_cw != CLIP_CULL) + { + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_GE, + get_element(c->reg.dir, 2), + brw_imm_f(0)); + + ccw = brw_IF(p, BRW_EXECUTE_1); + { + emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw); + } + ccw = brw_ELSE(p, ccw); + { + emit_primitives(c, c->key.fill_cw, c->key.offset_cw); + } + brw_ENDIF(p, ccw); + } + else if (c->key.fill_cw != CLIP_CULL) { + emit_primitives(c, c->key.fill_cw, c->key.offset_cw); + } + else if (c->key.fill_ccw != CLIP_CULL) { + emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw); + } +} + + + + +static void check_nr_verts( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *if_insn; + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.nr_verts, brw_imm_d(3)); + if_insn = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_kill_thread(c); + } + brw_ENDIF(p, if_insn); +} + + +void brw_emit_unfilled_clip( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *do_clip; + + + c->need_direction = ((c->key.offset_ccw || c->key.offset_cw) || + (c->key.fill_ccw != c->key.fill_cw) || + c->key.fill_ccw == CLIP_CULL || + c->key.fill_cw == CLIP_CULL || + c->key.copy_bfc_cw || + c->key.copy_bfc_ccw); + + brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6); + brw_clip_tri_init_vertices(c); + + assert(c->offset[VERT_RESULT_EDGE]); + + if (c->key.fill_ccw == CLIP_CULL && + c->key.fill_cw == CLIP_CULL) { + brw_clip_kill_thread(c); + return; + } + + merge_edgeflags(c); + + /* Need to use the inlist indirection here: + */ + if (c->need_direction) + compute_tri_direction(c); + + if (c->key.fill_ccw == CLIP_CULL || + c->key.fill_cw == CLIP_CULL) + cull_direction(c); + + if (c->key.offset_ccw || + c->key.offset_cw) + compute_offset(c); + + if (c->key.copy_bfc_ccw || + c->key.copy_bfc_cw) + copy_bfc(c); + + /* Need to do this whether we clip or not: + */ + if (c->key.do_flat_shading) + brw_clip_tri_flat_shade(c); + + brw_clip_init_clipmask(c); + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0)); + do_clip = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_init_planes(c); + brw_clip_tri(c); + check_nr_verts(c); + } + brw_ENDIF(p, do_clip); + + emit_unfilled_primitives(c); + brw_clip_kill_thread(c); +} + + + diff --git a/src/mesa/pipe/i965simple/brw_clip_util.c b/src/mesa/pipe/i965simple/brw_clip_util.c new file mode 100644 index 0000000000..6d58ceafff --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_clip_util.c @@ -0,0 +1,351 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + + + + +static struct brw_reg get_tmp( struct brw_clip_compile *c ) +{ + struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0); + + if (++c->last_tmp > c->prog_data.total_grf) + c->prog_data.total_grf = c->last_tmp; + + return tmp; +} + +static void release_tmp( struct brw_clip_compile *c, struct brw_reg tmp ) +{ + if (tmp.nr == c->last_tmp-1) + c->last_tmp--; +} + + +static struct brw_reg make_plane_ud(unsigned x, unsigned y, unsigned z, unsigned w) +{ + return brw_imm_ud((w<<24) | (z<<16) | (y<<8) | x); +} + + +void brw_clip_init_planes( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + + if (!c->key.nr_userclip) { + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 0), make_plane_ud( 0, 0, 0xff, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 1), make_plane_ud( 0, 0, 1, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 2), make_plane_ud( 0, 0xff, 0, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 3), make_plane_ud( 0, 1, 0, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 4), make_plane_ud(0xff, 0, 0, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 5), make_plane_ud( 1, 0, 0, 1)); + } +} + + + +#define W 3 + +/* Project 'pos' to screen space (or back again), overwrite with results: + */ +static void brw_clip_project_position(struct brw_clip_compile *c, struct brw_reg pos ) +{ + struct brw_compile *p = &c->func; + + /* calc rhw + */ + brw_math_invert(p, get_element(pos, W), get_element(pos, W)); + + /* value.xyz *= value.rhw + */ + brw_set_access_mode(p, BRW_ALIGN_16); + brw_MUL(p, brw_writemask(pos, TGSI_WRITEMASK_XYZ), pos, brw_swizzle1(pos, W)); + brw_set_access_mode(p, BRW_ALIGN_1); +} + + +static void brw_clip_project_vertex( struct brw_clip_compile *c, + struct brw_indirect vert_addr ) +{ +#if 0 + struct brw_compile *p = &c->func; + struct brw_reg tmp = get_tmp(c); + + /* Fixup position. Extract from the original vertex and re-project + * to screen space: + */ + brw_MOV(p, tmp, deref_4f(vert_addr, c->offset[VERT_RESULT_HPOS])); + brw_clip_project_position(c, tmp); + brw_MOV(p, deref_4f(vert_addr, c->header_position_offset), tmp); + + release_tmp(c, tmp); +#else + #warning "disabled" +#endif +} + + + + +/* Interpolate between two vertices and put the result into a0.0. + * Increment a0.0 accordingly. + */ +void brw_clip_interp_vertex( struct brw_clip_compile *c, + struct brw_indirect dest_ptr, + struct brw_indirect v0_ptr, /* from */ + struct brw_indirect v1_ptr, /* to */ + struct brw_reg t0, + boolean force_edgeflag) +{ +#if 0 + struct brw_compile *p = &c->func; + struct brw_reg tmp = get_tmp(c); + unsigned i; + + /* Just copy the vertex header: + */ + brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1); + + /* Iterate over each attribute (could be done in pairs?) + */ + for (i = 0; i < c->nr_attrs; i++) { + unsigned delta = i*16 + 32; + + if (delta == c->offset[VERT_RESULT_EDGE]) { + if (force_edgeflag) + brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(1)); + else + brw_MOV(p, deref_4f(dest_ptr, delta), deref_4f(v0_ptr, delta)); + } + else { + /* Interpolate: + * + * New = attr0 + t*attr1 - t*attr0 + */ + brw_MUL(p, + vec4(brw_null_reg()), + deref_4f(v1_ptr, delta), + t0); + + brw_MAC(p, + tmp, + negate(deref_4f(v0_ptr, delta)), + t0); + + brw_ADD(p, + deref_4f(dest_ptr, delta), + deref_4f(v0_ptr, delta), + tmp); + } + } + + if (i & 1) { + unsigned delta = i*16 + 32; + brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0)); + } + + release_tmp(c, tmp); + + /* Recreate the projected (NDC) coordinate in the new vertex + * header: + */ + brw_clip_project_vertex(c, dest_ptr ); +#else + #warning "disabled" +#endif +} + + + + +#define MAX_MRF 16 + +void brw_clip_emit_vue(struct brw_clip_compile *c, + struct brw_indirect vert, + boolean allocate, + boolean eot, + unsigned header) +{ + struct brw_compile *p = &c->func; + unsigned start = c->last_mrf; + + assert(!(allocate && eot)); + + /* Cycle through mrf regs - probably futile as we have to wait for + * the allocation response anyway. Also, the order this function + * is invoked doesn't correspond to the order the instructions will + * be executed, so it won't have any effect in many cases. + */ +#if 0 + if (start + c->nr_regs + 1 >= MAX_MRF) + start = 0; + + c->last_mrf = start + c->nr_regs + 1; +#endif + + /* Copy the vertex from vertn into m1..mN+1: + */ + brw_copy_from_indirect(p, brw_message_reg(start+1), vert, c->nr_regs); + + /* Overwrite PrimType and PrimStart in the message header, for + * each vertex in turn: + */ + brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header)); + + + /* Send each vertex as a seperate write to the urb. This + * is different to the concept in brw_sf_emit.c, where + * subsequent writes are used to build up a single urb + * entry. Each of these writes instantiates a seperate + * urb entry - (I think... what about 'allocate'?) + */ + brw_urb_WRITE(p, + allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), + start, + c->reg.R0, + allocate, + 1, /* used */ + c->nr_regs + 1, /* msg length */ + allocate ? 1 : 0, /* response_length */ + eot, /* eot */ + 1, /* writes_complete */ + 0, /* urb offset */ + BRW_URB_SWIZZLE_NONE); +} + + + +void brw_clip_kill_thread(struct brw_clip_compile *c) +{ + struct brw_compile *p = &c->func; + + /* Send an empty message to kill the thread and release any + * allocated urb entry: + */ + brw_urb_WRITE(p, + retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), + 0, + c->reg.R0, + 0, /* allocate */ + 0, /* used */ + 0, /* msg len */ + 0, /* response len */ + 1, /* eot */ + 1, /* writes complete */ + 0, + BRW_URB_SWIZZLE_NONE); +} + + + + +struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c ) +{ + return brw_address(c->reg.fixed_planes); +} + + +struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c ) +{ + if (c->key.nr_userclip) { + return brw_imm_uw(16); + } + else { + return brw_imm_uw(4); + } +} + + +/* If flatshading, distribute color from provoking vertex prior to + * clipping. + */ +void brw_clip_copy_colors( struct brw_clip_compile *c, + unsigned to, unsigned from ) +{ +#if 0 + struct brw_compile *p = &c->func; + + if (c->offset[VERT_RESULT_COL0]) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL0]), + byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL0])); + + if (c->offset[VERT_RESULT_COL1]) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL1]), + byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL1])); + + if (c->offset[VERT_RESULT_BFC0]) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC0]), + byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC0])); + + if (c->offset[VERT_RESULT_BFC1]) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC1]), + byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC1])); +#else + #warning "disabled" +#endif +} + + + +void brw_clip_init_clipmask( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg incoming = get_element_ud(c->reg.R0, 2); + + /* Shift so that lowest outcode bit is rightmost: + */ + brw_SHR(p, c->reg.planemask, incoming, brw_imm_ud(26)); + + if (c->key.nr_userclip) { + struct brw_reg tmp = retype(vec1(get_tmp(c)), BRW_REGISTER_TYPE_UD); + + /* Rearrange userclip outcodes so that they come directly after + * the fixed plane bits. + */ + brw_AND(p, tmp, incoming, brw_imm_ud(0x3f<<14)); + brw_SHR(p, tmp, tmp, brw_imm_ud(8)); + brw_OR(p, c->reg.planemask, c->reg.planemask, tmp); + + release_tmp(c, tmp); + } +} + diff --git a/src/mesa/pipe/i965simple/brw_context.c b/src/mesa/pipe/i965simple/brw_context.c new file mode 100644 index 0000000000..dc1987cec8 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_context.c @@ -0,0 +1,287 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_draw.h" +#include "brw_vs.h" +#include "brw_tex_layout.h" +#include "brw_winsys.h" + +#include "pipe/p_winsys.h" +#include "pipe/p_context.h" +#include "pipe/p_util.h" + +/*************************************** + * Mesa's Driver Functions + ***************************************/ + +#ifndef BRW_DEBUG +int BRW_DEBUG = (0); +#endif + +static void brw_destroy(struct pipe_context *pipe) +{ + struct brw_context *brw = brw_context(pipe); + + FREE(brw); +} + +static void brw_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + int x, y, w, h; + /* FIXME: corny... */ + + x = 0; + y = 0; + w = ps->width; + h = ps->height; + + pipe->surface_fill(pipe, ps, x, y, w, h, clearValue); +} + + +static int +brw_get_param(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 11; /* max 1024x1024 */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 11; /* max 1024x1024 */ + default: + return 0; + } +} + + +static float +brw_get_paramf(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 7.5; + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 4.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; + + default: + return 0; + } +} + +static boolean +brw_is_format_supported( struct pipe_context *pipe, + uint format ) +{ +#if 0 + /* XXX: This is broken -- rewrite if still needed. */ + static const unsigned tex_supported[] = { + PIPE_FORMAT_U_R8_G8_B8_A8, + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_U_L8, + PIPE_FORMAT_U_A8, + PIPE_FORMAT_U_I8, + PIPE_FORMAT_U_L8_A8, + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + PIPE_FORMAT_S8_Z24, + }; + + + /* Actually a lot more than this - add later: + */ + static const unsigned render_supported[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + }; + + /* + */ + static const unsigned z_stencil_supported[] = { + PIPE_FORMAT_U_Z16, + PIPE_FORMAT_U_Z32, + PIPE_FORMAT_S8_Z24, + }; + + switch (type) { + case PIPE_RENDER_FORMAT: + *numFormats = Elements(render_supported); + return render_supported; + + case PIPE_TEX_FORMAT: + *numFormats = Elements(tex_supported); + return render_supported; + + case PIPE_Z_STENCIL_FORMAT: + *numFormats = Elements(render_supported); + return render_supported; + + default: + *numFormats = 0; + return NULL; + } +#else + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + return TRUE; + }; + return FALSE; +#endif +} + +static void +brw_begin_query(struct pipe_context *pipe, struct pipe_query_object *q) +{ + /* should never be called */ + assert(0); +} + + +static void +brw_end_query(struct pipe_context *pipe, struct pipe_query_object *q) +{ + /* should never be called */ + assert(0); +} + + +static boolean +brw_draw_elements(struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) +{ + return FALSE; +} + +static boolean brw_draw_arrays(struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) +{ + return brw_draw_elements(pipe, NULL, 0, prim, start, count); +} + + +struct pipe_context *brw_create(struct pipe_winsys *pipe_winsys, + struct brw_winsys *brw_winsys, + unsigned pci_id) +{ + struct brw_context *brw; + + pipe_winsys->printf(pipe_winsys, + "%s: creating brw_context with pci id 0x%x\n", + __FUNCTION__, pci_id); + + brw = CALLOC_STRUCT(brw_context); + if (brw == NULL) + return NULL; + + brw->winsys = brw_winsys; + brw->pipe.winsys = pipe_winsys; + + brw->pipe.destroy = brw_destroy; + brw->pipe.is_format_supported = brw_is_format_supported; + brw->pipe.get_param = brw_get_param; + brw->pipe.get_paramf = brw_get_paramf; + + brw->pipe.clear = brw_clear; + + brw->pipe.begin_query = brw_begin_query; + brw->pipe.end_query = brw_end_query; + + brw->pipe.draw_arrays = brw_draw_arrays; + brw->pipe.draw_elements = brw_draw_elements; + + brw_init_surface_functions(brw); + brw_init_state_functions(brw); + brw_init_flush_functions(brw); + brw_init_string_functions(brw); + + brw->pci_id = pci_id; + + brw->pipe.texture_create = brw_texture_create; + brw->pipe.texture_release = brw_texture_release; + + brw->dirty = ~0; + brw->hardware_dirty = ~0; + + /* Batch stream debugging is a bit hacked up at the moment: + */ + brw->batch_start = NULL; + + + brw_init_state( brw ); + + memset(&brw->wm.bind, ~0, sizeof(brw->wm.bind)); + + brw_draw_init( brw ); + + + return &brw->pipe; +} + diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h new file mode 100644 index 0000000000..00df46e704 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -0,0 +1,699 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRWCONTEXT_INC +#define BRWCONTEXT_INC + + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "brw_structs.h" +#include "brw_winsys.h" + + +/* Glossary: + * + * URB - uniform resource buffer. A mid-sized buffer which is + * partitioned between the fixed function units and used for passing + * values (vertices, primitives, constants) between them. + * + * CURBE - constant URB entry. An urb region (entry) used to hold + * constant values which the fixed function units can be instructed to + * preload into the GRF when spawining a thread. + * + * VUE - vertex URB entry. An urb entry holding a vertex and usually + * a vertex header. The header contains control information and + * things like primitive type, Begin/end flags and clip codes. + * + * PUE - primitive URB entry. An urb entry produced by the setup (SF) + * unit holding rasterization and interpolation parameters. + * + * GRF - general register file. One of several register files + * addressable by programmed threads. The inputs (r0, payload, curbe, + * urb) of the thread are preloaded to this area before the thread is + * spawned. The registers are individually 8 dwords wide and suitable + * for general usage. Registers holding thread input values are not + * special and may be overwritten. + * + * MRF - message register file. Threads communicate (and terminate) + * by sending messages. Message parameters are placed in contigous + * MRF registers. All program output is via these messages. URB + * entries are populated by sending a message to the shared URB + * function containing the new data, together with a control word, + * often an unmodified copy of R0. + * + * R0 - GRF register 0. Typically holds control information used when + * sending messages to other threads. + * + * EU or GEN4 EU: The name of the programmable subsystem of the + * i965 hardware. Threads are executed by the EU, the registers + * described above are part of the EU architecture. + * + * Fixed function units: + * + * CS - Command streamer. Notional first unit, little software + * interaction. Holds the URB entries used for constant data, ie the + * CURBEs. + * + * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of + * this unit is responsible for pulling vertices out of vertex buffers + * in vram and injecting them into the processing pipe as VUEs. If + * enabled, it first passes them to a VS thread which is a good place + * for the driver to implement any active vertex shader. + * + * GS - Geometry Shader. This corresponds to a new DX10 concept. If + * enabled, incoming strips etc are passed to GS threads in individual + * line/triangle/point units. The GS thread may perform arbitary + * computation and emit whatever primtives with whatever vertices it + * chooses. This makes GS an excellent place to implement GL's + * unfilled polygon modes, though of course it is capable of much + * more. Additionally, GS is used to translate away primitives not + * handled by latter units, including Quads and Lineloops. + * + * CS - Clipper. Mesa's clipping algorithms are imported to run on + * this unit. The fixed function part performs cliptesting against + * the 6 fixed clipplanes and makes descisions on whether or not the + * incoming primitive needs to be passed to a thread for clipping. + * User clip planes are handled via cooperation with the VS thread. + * + * SF - Strips Fans or Setup: Triangles are prepared for + * rasterization. Interpolation coefficients are calculated. + * Flatshading and two-side lighting usually performed here. + * + * WM - Windower. Interpolation of vertex attributes performed here. + * Fragment shader implemented here. SIMD aspects of EU taken full + * advantage of, as pixels are processed in blocks of 16. + * + * CC - Color Calculator. No EU threads associated with this unit. + * Handles blending and (presumably) depth and stencil testing. + */ + +#define BRW_FALLBACK_TEXTURE 0x1 +#define BRW_MAX_CURBE (32*16) + +struct brw_context; +struct brw_winsys; + + +/* Raised when we receive new state across the pipe interface: + */ +#define BRW_NEW_VIEWPORT 0x1 +#define BRW_NEW_RASTERIZER 0x2 +#define BRW_NEW_FS 0x4 +#define BRW_NEW_BLEND 0x8 +#define BRW_NEW_CLIP 0x10 +#define BRW_NEW_SCISSOR 0x20 +#define BRW_NEW_STIPPLE 0x40 +#define BRW_NEW_FRAMEBUFFER 0x80 +#define BRW_NEW_ALPHA_TEST 0x100 +#define BRW_NEW_DEPTH_STENCIL 0x200 +#define BRW_NEW_SAMPLER 0x400 +#define BRW_NEW_TEXTURE 0x800 +#define BRW_NEW_CONSTANTS 0x1000 +#define BRW_NEW_VBO 0x2000 +#define BRW_NEW_VS 0x4000 + +/* Raised for other internal events: + */ +#define BRW_NEW_URB_FENCE 0x10000 +#define BRW_NEW_INPUT_DIMENSIONS 0x20000 +#define BRW_NEW_CURBE_OFFSETS 0x40000 +#define BRW_NEW_REDUCED_PRIMITIVE 0x80000 +#define BRW_NEW_PRIMITIVE 0x100000 +#define BRW_NEW_CONTEXT 0x200000 +#define BRW_NEW_WM_INPUT_DIMENSIONS 0x400000 +#define BRW_NEW_INPUT_VARYING 0x800000 +#define BRW_NEW_PSP 0x1000000 + +#define ALIGN(value, alignment) ((value + alignment - 1) & ~(alignment - 1)) +extern int BRW_DEBUG; + +#define DEBUG_TEXTURE 0x1 +#define DEBUG_STATE 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_PRIMS 0x8 +#define DEBUG_VERTS 0x10 +#define DEBUG_FALLBACKS 0x20 +#define DEBUG_VERBOSE 0x40 +#define DEBUG_DRI 0x80 +#define DEBUG_DMA 0x100 +#define DEBUG_SANITY 0x200 +#define DEBUG_SYNC 0x400 +#define DEBUG_SLEEP 0x800 +#define DEBUG_PIXEL 0x1000 +#define DEBUG_STATS 0x2000 +#define DEBUG_TILE 0x4000 +#define DEBUG_SINGLE_THREAD 0x8000 +#define DEBUG_WM 0x10000 +#define DEBUG_URB 0x20000 +#define DEBUG_VS 0x40000 +#define DEBUG_BATCH 0x80000 +#define DEBUG_BUFMGR 0x100000 +#define DEBUG_BLIT 0x200000 +#define DEBUG_REGION 0x400000 +#define DEBUG_MIPTREE 0x800000 + +#define DBG(...) do { \ + if (BRW_DEBUG & FILE_DEBUG_FLAG) \ + brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ +} while(0) + +#define PRINT(...) do { \ + brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ +} while(0) + +struct brw_state_flags { + unsigned cache; + unsigned brw; +}; + +struct brw_vertex_program { + struct pipe_shader_state program; + unsigned id; + unsigned param_state; /* flags indicating state tracked by params */ +}; + + + +struct brw_fragment_program { + struct pipe_shader_state program; + unsigned id; + unsigned param_state; /* flags indicating state tracked by params */ + boolean UsesKill; + boolean ComputesDepth; +}; + + +struct brw_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer_handle *buffer; +}; + +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ +struct brw_wm_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + + unsigned first_curbe_grf; + unsigned total_grf; + unsigned total_scratch; + + unsigned nr_params; + boolean error; + + /* Pointer to tracked values (only valid once + * _mesa_load_state_parameters has been called at runtime). + */ + const float *param[BRW_MAX_CURBE]; +}; + +struct brw_sf_prog_data { + unsigned urb_read_length; + unsigned total_grf; + + /* Each vertex may have upto 12 attributes, 4 components each, + * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 + * rows. + * + * Actually we use 4 for each, so call it 12 rows. + */ + unsigned urb_entry_size; +}; + +struct brw_clip_prog_data { + unsigned curb_read_length; /* user planes? */ + unsigned clip_mode; + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_gs_prog_data { + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_vs_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + unsigned total_grf; + unsigned outputs_written; + + unsigned inputs_read; + + /* Used for calculating urb partitions: + */ + unsigned urb_entry_size; +}; + + +/* Size == 0 if output either not written, or always [0,0,0,1] + */ +struct brw_vs_ouput_sizes { + ubyte output_size[PIPE_MAX_SHADER_OUTPUTS]; +}; + + +#define BRW_MAX_TEX_UNIT 8 +#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 + +/* Create a fixed sized struct for caching binding tables: + */ +struct brw_surface_binding_table { + unsigned surf_ss_offset[BRW_WM_MAX_SURF]; +}; + + +struct brw_cache; + +struct brw_mem_pool { + struct pipe_buffer_handle *buffer; + + unsigned size; + unsigned offset; /* offset of first free byte */ + + struct brw_context *brw; +}; + +struct brw_cache_item { + unsigned hash; + unsigned key_size; /* for variable-sized keys */ + const void *key; + + unsigned offset; /* offset within pool's buffer */ + unsigned data_size; + + struct brw_cache_item *next; +}; + + + +struct brw_cache { + unsigned id; + + const char *name; + + struct brw_context *brw; + struct brw_mem_pool *pool; + + struct brw_cache_item **items; + unsigned size, n_items; + + unsigned key_size; /* for fixed-size keys */ + unsigned aux_size; + + unsigned last_addr; /* offset of active item */ +}; + + + + +/* Considered adding a member to this struct to document which flags + * an update might raise so that ordering of the state atoms can be + * checked or derived at runtime. Dropped the idea in favor of having + * a debug mode where the state is monitored for flags which are + * raised that have already been tested against. + */ +struct brw_tracked_state { + struct brw_state_flags dirty; + void (*update)( struct brw_context *brw ); + void (*emit_reloc)( struct brw_context *brw ); + boolean always_update; +}; + + +/* Flags for brw->state.cache. + */ +#define CACHE_NEW_CC_VP (1< 32. Wouldn't life + * be easier if C allowed arrays of packed elements? + */ +#define ATTRIB_BIT_DWORDS ((PIPE_ATTRIB_MAX+31)/32) + +struct brw_vertex_element { + struct brw_vertex_element_state vep; + + unsigned index; + unsigned element_size; + unsigned count; + unsigned vbo_rebase_offset; +}; + + + +struct brw_vertex_info { + unsigned varying; /* varying:1[PIPE_ATTRIB_MAX] */ + unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_ATTRIB_MAX] */ +}; + + + + + +struct brw_context +{ + struct pipe_context pipe; + struct brw_winsys *winsys; + + unsigned primitive; + unsigned reduced_primitive; + + boolean emit_state_always; + + struct { + struct brw_state_flags dirty; + struct brw_tracked_state **atoms; + unsigned nr_atoms; + } state; + + + struct { + const struct pipe_alpha_test_state *AlphaTest; + const struct pipe_blend_state *Blend; + const struct pipe_depth_stencil_state *DepthStencil; + const struct pipe_poly_stipple *PolygonStipple; + const struct pipe_rasterizer_state *Raster; + const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; + const struct brw_vertex_program *VertexProgram; + const struct brw_fragment_program *FragmentProgram; + + struct pipe_clip_state Clip; + struct pipe_blend_color BlendColor; + struct pipe_scissor_state Scissor; + struct pipe_viewport_state Viewport; + struct pipe_framebuffer_state FrameBuffer; + struct pipe_clear_color_state ClearColor; + + const struct pipe_constant_buffer *Constants[2]; + const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; + } attribs; + + struct brw_mem_pool pool[BRW_MAX_POOL]; + struct brw_cache cache[BRW_MAX_CACHE]; + struct brw_cached_batch_item *cached_batch_items; + + struct { + + /* Arrays with buffer objects to copy non-bufferobj arrays into + * for upload: + */ + struct pipe_vertex_buffer vbo_array[PIPE_ATTRIB_MAX]; + + struct brw_vertex_element inputs[PIPE_ATTRIB_MAX]; + +#define BRW_NR_UPLOAD_BUFS 17 +#define BRW_UPLOAD_INIT_SIZE (128*1024) + + struct { + struct pipe_buffer_handle *vbo[BRW_NR_UPLOAD_BUFS]; + unsigned buf; + unsigned offset; + unsigned size; + unsigned wrap; + } upload; + + /* Summary of size and varying of active arrays, so we can check + * for changes to this state: + */ + struct brw_vertex_info info; + int last_vb; + } vb; + + + unsigned *batch_start; + unsigned hardware_dirty; + unsigned dirty; + unsigned pci_id; + + /* BRW_NEW_URB_ALLOCATIONS: + */ + struct { + unsigned vsize; /* vertex size plus header in urb registers */ + unsigned csize; /* constant buffer size in urb registers */ + unsigned sfsize; /* setup data size in urb registers */ + + boolean constrained; + + unsigned nr_vs_entries; + unsigned nr_gs_entries; + unsigned nr_clip_entries; + unsigned nr_sf_entries; + unsigned nr_cs_entries; + +/* unsigned vs_size; */ +/* unsigned gs_size; */ +/* unsigned clip_size; */ +/* unsigned sf_size; */ +/* unsigned cs_size; */ + + unsigned vs_start; + unsigned gs_start; + unsigned clip_start; + unsigned sf_start; + unsigned cs_start; + } urb; + + + /* BRW_NEW_CURBE_OFFSETS: + */ + struct { + unsigned wm_start; + unsigned wm_size; + unsigned clip_start; + unsigned clip_size; + unsigned vs_start; + unsigned vs_size; + unsigned total_size; + + /* Dynamic tracker which changes to reflect the state referenced + * by active fp and vp program parameters: + */ + struct brw_tracked_state tracked_state; + + unsigned gs_offset; + + float *last_buf; + unsigned last_bufsz; + } curbe; + + struct { + struct brw_vs_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } vs; + + struct { + struct brw_gs_prog_data *prog_data; + + boolean prog_active; + unsigned prog_gs_offset; + unsigned state_gs_offset; + } gs; + + struct { + struct brw_clip_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } clip; + + + struct { + struct brw_sf_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } sf; + + struct { + struct brw_wm_prog_data *prog_data; + struct brw_wm_compile *compile_data; + + /* Input sizes, calculated from active vertex program: + */ + unsigned input_size_masks[4]; + + + /** + * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER + * cache + */ + struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; + + unsigned render_surf; + unsigned nr_surfaces; + + unsigned max_threads; + struct pipe_buffer_handle *scratch_buffer; + unsigned scratch_buffer_size; + + unsigned sampler_count; + unsigned sampler_gs_offset; + + struct brw_surface_binding_table bind; + unsigned bind_ss_offset; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } wm; + + + struct { + unsigned vp_gs_offset; + unsigned state_gs_offset; + } cc; + + + /* Used to give every program string a unique id + */ + unsigned program_id; +}; + + +#define BRW_PACKCOLOR8888(r,g,b,a) ((r<<24) | (g<<16) | (b<<8) | a) + + +/*====================================================================== + * brw_vtbl.c + */ +void brw_do_flush( struct brw_context *brw, + unsigned flags ); + + +/*====================================================================== + * brw_state.c + */ +void brw_validate_state(struct brw_context *brw); +void brw_init_state(struct brw_context *brw); +void brw_destroy_state(struct brw_context *brw); + + +/*====================================================================== + * brw_tex.c + */ +void brwUpdateTextureState( struct brw_context *brw ); +void brw_FrameBufferTexInit( struct brw_context *brw ); +void brw_FrameBufferTexDestroy( struct brw_context *brw ); + + +/* brw_urb.c + */ +void brw_upload_urb_fence(struct brw_context *brw); + +void brw_upload_constant_buffer_state(struct brw_context *brw); +void brw_upload_polygon_stipple(struct brw_context *brw); +void brw_upload_line_stipple(struct brw_context *brw); + +void brw_init_surface_functions(struct brw_context *brw); +void brw_init_state_functions(struct brw_context *brw); +void brw_init_flush_functions(struct brw_context *brw); +void brw_init_string_functions(struct brw_context *brw); + +/*====================================================================== + * Inline conversion functions. These are better-typed than the + * macros used previously: + */ +static inline struct brw_context * +brw_context( struct pipe_context *ctx ) +{ + return (struct brw_context *)ctx; +} + +#endif + diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c new file mode 100644 index 0000000000..0894e82d56 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -0,0 +1,369 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_state.h" +#include "brw_util.h" + +#define FILE_DEBUG_FLAG DEBUG_FALLBACKS + +/* Partition the CURBE between the various users of constant values: + */ +static void calculate_curbe_offsets( struct brw_context *brw ) +{ + /* CACHE_NEW_WM_PROG */ + unsigned nr_fp_regs = (brw->wm.prog_data->nr_params + 15) / 16; + + /* BRW_NEW_VERTEX_PROGRAM */ + struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->attribs.VertexProgram; + unsigned nr_vp_regs = (vp->program.num_inputs * 4 + 15) / 16; + unsigned nr_clip_regs = 0; + unsigned total_regs; + +#if 0 + /* BRW_NEW_CLIP ? */ + if (brw->attribs.Transform->ClipPlanesEnabled) { + unsigned nr_planes = 6 + brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled); + nr_clip_regs = (nr_planes * 4 + 15) / 16; + } +#endif + + + total_regs = nr_fp_regs + nr_vp_regs + nr_clip_regs; + + /* This can happen - what to do? Probably rather than falling + * back, the best thing to do is emit programs which code the + * constants as immediate values. Could do this either as a static + * cap on WM and VS, or adaptively. + * + * Unfortunately, this is currently dependent on the results of the + * program generation process (in the case of wm), so this would + * introduce the need to re-generate programs in the event of a + * curbe allocation failure. + */ + /* Max size is 32 - just large enough to + * hold the 128 parameters allowed by + * the fragment and vertex program + * api's. It's not clear what happens + * when both VP and FP want to use 128 + * parameters, though. + */ + assert(total_regs <= 32); + + /* Lazy resize: + */ + if (nr_fp_regs > brw->curbe.wm_size || + nr_vp_regs > brw->curbe.vs_size || + nr_clip_regs != brw->curbe.clip_size || + (total_regs < brw->curbe.total_size / 4 && + brw->curbe.total_size > 16)) { + + unsigned reg = 0; + + /* Calculate a new layout: + */ + reg = 0; + brw->curbe.wm_start = reg; + brw->curbe.wm_size = nr_fp_regs; reg += nr_fp_regs; + brw->curbe.clip_start = reg; + brw->curbe.clip_size = nr_clip_regs; reg += nr_clip_regs; + brw->curbe.vs_start = reg; + brw->curbe.vs_size = nr_vp_regs; reg += nr_vp_regs; + brw->curbe.total_size = reg; + +#if 0 + if (0) + DBG("curbe wm %d+%d clip %d+%d vs %d+%d\n", + brw->curbe.wm_start, + brw->curbe.wm_size, + brw->curbe.clip_start, + brw->curbe.clip_size, + brw->curbe.vs_start, + brw->curbe.vs_size ); +#endif + + brw->state.dirty.brw |= BRW_NEW_CURBE_OFFSETS; + } +} + + +const struct brw_tracked_state brw_curbe_offsets = { + .dirty = { + .brw = (BRW_NEW_CLIP | + BRW_NEW_VS), + .cache = CACHE_NEW_WM_PROG + }, + .update = calculate_curbe_offsets +}; + + + +/* Define the number of curbes within CS's urb allocation. Multiple + * urb entries -> multiple curbes. These will be used by + * fixed-function hardware in a double-buffering scheme to avoid a + * pipeline stall each time the contents of the curbe is changed. + */ +void brw_upload_constant_buffer_state(struct brw_context *brw) +{ + struct brw_constant_buffer_state cbs; + memset(&cbs, 0, sizeof(cbs)); + + /* It appears that this is the state packet for the CS unit, ie. the + * urb entries detailed here are housed in the CS range from the + * URB_FENCE command. + */ + cbs.header.opcode = CMD_CONST_BUFFER_STATE; + cbs.header.length = sizeof(cbs)/4 - 2; + + /* BRW_NEW_URB_FENCE */ + cbs.bits0.nr_urb_entries = brw->urb.nr_cs_entries; + cbs.bits0.urb_entry_size = brw->urb.csize - 1; + + assert(brw->urb.nr_cs_entries); + BRW_CACHED_BATCH_STRUCT(brw, &cbs); +} + +const struct brw_tracked_state brw_constant_buffer_state = { + .dirty = { + .brw = BRW_NEW_URB_FENCE, + .cache = 0 + }, + .update = brw_upload_constant_buffer_state +}; + + +static float fixed_plane[6][4] = { + { 0, 0, -1, 1 }, + { 0, 0, 1, 1 }, + { 0, -1, 0, 1 }, + { 0, 1, 0, 1 }, + {-1, 0, 0, 1 }, + { 1, 0, 0, 1 } +}; + +#if 0 +/* Upload a new set of constants. Too much variability to go into the + * cache mechanism, but maybe would benefit from a comparison against + * the current uploaded set of constants. + */ +static void upload_constant_buffer(struct brw_context *brw) +{ + GLcontext *ctx = &brw->intel.ctx; + struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->vertex_program; + struct brw_fragment_program *fp = (struct brw_fragment_program *)brw->fragment_program; + struct brw_mem_pool *pool = &brw->pool[BRW_GS_POOL]; + unsigned sz = brw->curbe.total_size; + unsigned bufsz = sz * 16 * sizeof(float); + float *buf; + unsigned i; + + /* Update our own dependency flags. This works because this + * function will also be called whenever fp or vp changes. + */ + brw->curbe.tracked_state.dirty.mesa = (_NEW_TRANSFORM|_NEW_PROJECTION); + brw->curbe.tracked_state.dirty.mesa |= vp->param_state; + brw->curbe.tracked_state.dirty.mesa |= fp->param_state; + + if (sz == 0) { + struct brw_constant_buffer cb; + cb.header.opcode = CMD_CONST_BUFFER; + cb.header.length = sizeof(cb)/4 - 2; + cb.header.valid = 0; + cb.bits0.buffer_length = 0; + cb.bits0.buffer_address = 0; + BRW_BATCH_STRUCT(brw, &cb); + + if (brw->curbe.last_buf) { + free(brw->curbe.last_buf); + brw->curbe.last_buf = NULL; + brw->curbe.last_bufsz = 0; + } + + return; + } + + buf = (float *)malloc(bufsz); + + memset(buf, 0, bufsz); + + if (brw->curbe.wm_size) { + unsigned offset = brw->curbe.wm_start * 16; + + _mesa_load_state_parameters(ctx, fp->program.Base.Parameters); + + for (i = 0; i < brw->wm.prog_data->nr_params; i++) + buf[offset + i] = brw->wm.prog_data->param[i][0]; + } + + + /* The clipplanes are actually delivered to both CLIP and VS units. + * VS uses them to calculate the outcode bitmasks. + */ + if (brw->curbe.clip_size) { + unsigned offset = brw->curbe.clip_start * 16; + unsigned j; + + /* If any planes are going this way, send them all this way: + */ + for (i = 0; i < 6; i++) { + buf[offset + i * 4 + 0] = fixed_plane[i][0]; + buf[offset + i * 4 + 1] = fixed_plane[i][1]; + buf[offset + i * 4 + 2] = fixed_plane[i][2]; + buf[offset + i * 4 + 3] = fixed_plane[i][3]; + } + + /* Clip planes: _NEW_TRANSFORM plus _NEW_PROJECTION to get to + * clip-space: + */ + assert(MAX_CLIP_PLANES == 6); + for (j = 0; j < MAX_CLIP_PLANES; j++) { + if (brw->attribs.Transform->ClipPlanesEnabled & (1<attribs.Transform->_ClipUserPlane[j][0]; + buf[offset + i * 4 + 1] = brw->attribs.Transform->_ClipUserPlane[j][1]; + buf[offset + i * 4 + 2] = brw->attribs.Transform->_ClipUserPlane[j][2]; + buf[offset + i * 4 + 3] = brw->attribs.Transform->_ClipUserPlane[j][3]; + i++; + } + } + } + + + if (brw->curbe.vs_size) { + unsigned offset = brw->curbe.vs_start * 16; + unsigned nr = vp->program.Base.Parameters->NumParameters; + + _mesa_load_state_parameters(ctx, vp->program.Base.Parameters); + + for (i = 0; i < nr; i++) { + buf[offset + i * 4 + 0] = vp->program.Base.Parameters->ParameterValues[i][0]; + buf[offset + i * 4 + 1] = vp->program.Base.Parameters->ParameterValues[i][1]; + buf[offset + i * 4 + 2] = vp->program.Base.Parameters->ParameterValues[i][2]; + buf[offset + i * 4 + 3] = vp->program.Base.Parameters->ParameterValues[i][3]; + } + } + + if (0) { + for (i = 0; i < sz*16; i+=4) + _mesa_printf("curbe %d.%d: %f %f %f %f\n", i/8, i&4, + buf[i+0], buf[i+1], buf[i+2], buf[i+3]); + + _mesa_printf("last_buf %p buf %p sz %d/%d cmp %d\n", + brw->curbe.last_buf, buf, + bufsz, brw->curbe.last_bufsz, + brw->curbe.last_buf ? memcmp(buf, brw->curbe.last_buf, bufsz) : -1); + } + + if (brw->curbe.last_buf && + bufsz == brw->curbe.last_bufsz && + memcmp(buf, brw->curbe.last_buf, bufsz) == 0) { + free(buf); +/* return; */ + } + else { + if (brw->curbe.last_buf) + free(brw->curbe.last_buf); + brw->curbe.last_buf = buf; + brw->curbe.last_bufsz = bufsz; + + + if (!brw_pool_alloc(pool, + bufsz, + 1 << 6, + &brw->curbe.gs_offset)) { + _mesa_printf("out of GS memory for curbe\n"); + assert(0); + return; + } + + + /* Copy data to the buffer: + */ + dri_bo_subdata(pool->buffer, brw->curbe.gs_offset, bufsz, buf); + } + + /* TODO: only emit the constant_buffer packet when necessary, ie: + - contents have changed + - offset has changed + - hw requirements due to other packets emitted. + */ + { + struct brw_constant_buffer cb; + + memset(&cb, 0, sizeof(cb)); + + cb.header.opcode = CMD_CONST_BUFFER; + cb.header.length = sizeof(cb)/4 - 2; + cb.header.valid = 1; + cb.bits0.buffer_length = sz - 1; + cb.bits0.buffer_address = brw->curbe.gs_offset >> 6; + + /* Because this provokes an action (ie copy the constants into the + * URB), it shouldn't be shortcircuited if identical to the + * previous time - because eg. the urb destination may have + * changed, or the urb contents different to last time. + * + * Note that the data referred to is actually copied internally, + * not just used in place according to passed pointer. + * + * It appears that the CS unit takes care of using each available + * URB entry (Const URB Entry == CURBE) in turn, and issuing + * flushes as necessary when doublebuffering of CURBEs isn't + * possible. + */ +/* intel_batchbuffer_align(brw->intel.batch, 64, sizeof(cb)); */ + BRW_BATCH_STRUCT(brw, &cb); +/* intel_batchbuffer_align(brw->intel.batch, 64, 0); */ + } +} + +/* This tracked state is unique in that the state it monitors varies + * dynamically depending on the parameters tracked by the fragment and + * vertex programs. This is the template used as a starting point, + * each context will maintain a copy of this internally and update as + * required. + */ +const struct brw_tracked_state brw_constant_buffer = { + .dirty = { + .mesa = (_NEW_TRANSFORM|_NEW_PROJECTION), /* plus fp and vp flags */ + .brw = (BRW_NEW_FRAGMENT_PROGRAM | + BRW_NEW_VERTEX_PROGRAM | + BRW_NEW_URB_FENCE | /* Implicit - hardware requires this, not used above */ + BRW_NEW_PSP | /* Implicit - hardware requires this, not used above */ + BRW_NEW_CURBE_OFFSETS), + .cache = (CACHE_NEW_WM_PROG) + }, + .update = upload_constant_buffer +}; + +#endif diff --git a/src/mesa/pipe/i965simple/brw_defines.h b/src/mesa/pipe/i965simple/brw_defines.h new file mode 100644 index 0000000000..9379a397f6 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_defines.h @@ -0,0 +1,852 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_DEFINES_H +#define BRW_DEFINES_H + +/* + */ +#define MI_NOOP 0x00 +#define MI_USER_INTERRUPT 0x02 +#define MI_WAIT_FOR_EVENT 0x03 +#define MI_FLUSH 0x04 +#define MI_REPORT_HEAD 0x07 +#define MI_ARB_ON_OFF 0x08 +#define MI_BATCH_BUFFER_END 0x0A +#define MI_OVERLAY_FLIP 0x11 +#define MI_LOAD_SCAN_LINES_INCL 0x12 +#define MI_LOAD_SCAN_LINES_EXCL 0x13 +#define MI_DISPLAY_BUFFER_INFO 0x14 +#define MI_SET_CONTEXT 0x18 +#define MI_STORE_DATA_IMM 0x20 +#define MI_STORE_DATA_INDEX 0x21 +#define MI_LOAD_REGISTER_IMM 0x22 +#define MI_STORE_REGISTER_MEM 0x24 +#define MI_BATCH_BUFFER_START 0x31 + +#define MI_SYNCHRONOUS_FLIP 0x0 +#define MI_ASYNCHRONOUS_FLIP 0x1 + +#define MI_BUFFER_SECURE 0x0 +#define MI_BUFFER_NONSECURE 0x1 + +#define MI_ARBITRATE_AT_CHAIN_POINTS 0x0 +#define MI_ARBITRATE_BETWEEN_INSTS 0x1 +#define MI_NO_ARBITRATION 0x3 + +#define MI_CONDITION_CODE_WAIT_DISABLED 0x0 +#define MI_CONDITION_CODE_WAIT_0 0x1 +#define MI_CONDITION_CODE_WAIT_1 0x2 +#define MI_CONDITION_CODE_WAIT_2 0x3 +#define MI_CONDITION_CODE_WAIT_3 0x4 +#define MI_CONDITION_CODE_WAIT_4 0x5 + +#define MI_DISPLAY_PIPE_A 0x0 +#define MI_DISPLAY_PIPE_B 0x1 + +#define MI_DISPLAY_PLANE_A 0x0 +#define MI_DISPLAY_PLANE_B 0x1 +#define MI_DISPLAY_PLANE_C 0x2 + +#define MI_STANDARD_FLIP 0x0 +#define MI_ENQUEUE_FLIP_PERFORM_BASE_FRAME_NUMBER_LOAD 0x1 +#define MI_ENQUEUE_FLIP_TARGET_FRAME_NUMBER_RELATIVE 0x2 +#define MI_ENQUEUE_FLIP_ABSOLUTE_TARGET_FRAME_NUMBER 0x3 + +#define MI_PHYSICAL_ADDRESS 0x0 +#define MI_VIRTUAL_ADDRESS 0x1 + +#define MI_BUFFER_MEMORY_MAIN 0x0 +#define MI_BUFFER_MEMORY_GTT 0x2 +#define MI_BUFFER_MEMORY_PER_PROCESS_GTT 0x3 + +#define MI_FLIP_CONTINUE 0x0 +#define MI_FLIP_ON 0x1 +#define MI_FLIP_OFF 0x2 + +#define MI_UNTRUSTED_REGISTER_SPACE 0x0 +#define MI_TRUSTED_REGISTER_SPACE 0x1 + +/* 3D state: + */ +#define _3DOP_3DSTATE_PIPELINED 0x0 +#define _3DOP_3DSTATE_NONPIPELINED 0x1 +#define _3DOP_3DCONTROL 0x2 +#define _3DOP_3DPRIMITIVE 0x3 + +#define _3DSTATE_PIPELINED_POINTERS 0x00 +#define _3DSTATE_BINDING_TABLE_POINTERS 0x01 +#define _3DSTATE_VERTEX_BUFFERS 0x08 +#define _3DSTATE_VERTEX_ELEMENTS 0x09 +#define _3DSTATE_INDEX_BUFFER 0x0A +#define _3DSTATE_VF_STATISTICS 0x0B +#define _3DSTATE_DRAWING_RECTANGLE 0x00 +#define _3DSTATE_CONSTANT_COLOR 0x01 +#define _3DSTATE_SAMPLER_PALETTE_LOAD 0x02 +#define _3DSTATE_CHROMA_KEY 0x04 +#define _3DSTATE_DEPTH_BUFFER 0x05 +#define _3DSTATE_POLY_STIPPLE_OFFSET 0x06 +#define _3DSTATE_POLY_STIPPLE_PATTERN 0x07 +#define _3DSTATE_LINE_STIPPLE 0x08 +#define _3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP 0x09 +#define _3DCONTROL 0x00 +#define _3DPRIMITIVE 0x00 + +#define PIPE_CONTROL_NOWRITE 0x00 +#define PIPE_CONTROL_WRITEIMMEDIATE 0x01 +#define PIPE_CONTROL_WRITEDEPTH 0x02 +#define PIPE_CONTROL_WRITETIMESTAMP 0x03 + +#define PIPE_CONTROL_GTTWRITE_PROCESS_LOCAL 0x00 +#define PIPE_CONTROL_GTTWRITE_GLOBAL 0x01 + +#define _3DPRIM_POINTLIST 0x01 +#define _3DPRIM_LINELIST 0x02 +#define _3DPRIM_LINESTRIP 0x03 +#define _3DPRIM_TRILIST 0x04 +#define _3DPRIM_TRISTRIP 0x05 +#define _3DPRIM_TRIFAN 0x06 +#define _3DPRIM_QUADLIST 0x07 +#define _3DPRIM_QUADSTRIP 0x08 +#define _3DPRIM_LINELIST_ADJ 0x09 +#define _3DPRIM_LINESTRIP_ADJ 0x0A +#define _3DPRIM_TRILIST_ADJ 0x0B +#define _3DPRIM_TRISTRIP_ADJ 0x0C +#define _3DPRIM_TRISTRIP_REVERSE 0x0D +#define _3DPRIM_POLYGON 0x0E +#define _3DPRIM_RECTLIST 0x0F +#define _3DPRIM_LINELOOP 0x10 +#define _3DPRIM_POINTLIST_BF 0x11 +#define _3DPRIM_LINESTRIP_CONT 0x12 +#define _3DPRIM_LINESTRIP_BF 0x13 +#define _3DPRIM_LINESTRIP_CONT_BF 0x14 +#define _3DPRIM_TRIFAN_NOSTIPPLE 0x15 + +#define _3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL 0 +#define _3DPRIM_VERTEXBUFFER_ACCESS_RANDOM 1 + +#define BRW_ANISORATIO_2 0 +#define BRW_ANISORATIO_4 1 +#define BRW_ANISORATIO_6 2 +#define BRW_ANISORATIO_8 3 +#define BRW_ANISORATIO_10 4 +#define BRW_ANISORATIO_12 5 +#define BRW_ANISORATIO_14 6 +#define BRW_ANISORATIO_16 7 + +#define BRW_BLENDFACTOR_ONE 0x1 +#define BRW_BLENDFACTOR_SRC_COLOR 0x2 +#define BRW_BLENDFACTOR_SRC_ALPHA 0x3 +#define BRW_BLENDFACTOR_DST_ALPHA 0x4 +#define BRW_BLENDFACTOR_DST_COLOR 0x5 +#define BRW_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 +#define BRW_BLENDFACTOR_CONST_COLOR 0x7 +#define BRW_BLENDFACTOR_CONST_ALPHA 0x8 +#define BRW_BLENDFACTOR_SRC1_COLOR 0x9 +#define BRW_BLENDFACTOR_SRC1_ALPHA 0x0A +#define BRW_BLENDFACTOR_ZERO 0x11 +#define BRW_BLENDFACTOR_INV_SRC_COLOR 0x12 +#define BRW_BLENDFACTOR_INV_SRC_ALPHA 0x13 +#define BRW_BLENDFACTOR_INV_DST_ALPHA 0x14 +#define BRW_BLENDFACTOR_INV_DST_COLOR 0x15 +#define BRW_BLENDFACTOR_INV_CONST_COLOR 0x17 +#define BRW_BLENDFACTOR_INV_CONST_ALPHA 0x18 +#define BRW_BLENDFACTOR_INV_SRC1_COLOR 0x19 +#define BRW_BLENDFACTOR_INV_SRC1_ALPHA 0x1A + +#define BRW_BLENDFUNCTION_ADD 0 +#define BRW_BLENDFUNCTION_SUBTRACT 1 +#define BRW_BLENDFUNCTION_REVERSE_SUBTRACT 2 +#define BRW_BLENDFUNCTION_MIN 3 +#define BRW_BLENDFUNCTION_MAX 4 + +#define BRW_ALPHATEST_FORMAT_UNORM8 0 +#define BRW_ALPHATEST_FORMAT_FLOAT32 1 + +#define BRW_CHROMAKEY_KILL_ON_ANY_MATCH 0 +#define BRW_CHROMAKEY_REPLACE_BLACK 1 + +#define BRW_CLIP_API_OGL 0 +#define BRW_CLIP_API_DX 1 + +#define BRW_CLIPMODE_NORMAL 0 +#define BRW_CLIPMODE_CLIP_ALL 1 +#define BRW_CLIPMODE_CLIP_NON_REJECTED 2 +#define BRW_CLIPMODE_REJECT_ALL 3 +#define BRW_CLIPMODE_ACCEPT_ALL 4 + +#define BRW_CLIP_NDCSPACE 0 +#define BRW_CLIP_SCREENSPACE 1 + +#define BRW_COMPAREFUNCTION_ALWAYS 0 +#define BRW_COMPAREFUNCTION_NEVER 1 +#define BRW_COMPAREFUNCTION_LESS 2 +#define BRW_COMPAREFUNCTION_EQUAL 3 +#define BRW_COMPAREFUNCTION_LEQUAL 4 +#define BRW_COMPAREFUNCTION_GREATER 5 +#define BRW_COMPAREFUNCTION_NOTEQUAL 6 +#define BRW_COMPAREFUNCTION_GEQUAL 7 + +#define BRW_COVERAGE_PIXELS_HALF 0 +#define BRW_COVERAGE_PIXELS_1 1 +#define BRW_COVERAGE_PIXELS_2 2 +#define BRW_COVERAGE_PIXELS_4 3 + +#define BRW_CULLMODE_BOTH 0 +#define BRW_CULLMODE_NONE 1 +#define BRW_CULLMODE_FRONT 2 +#define BRW_CULLMODE_BACK 3 + +#define BRW_DEFAULTCOLOR_R8G8B8A8_UNORM 0 +#define BRW_DEFAULTCOLOR_R32G32B32A32_FLOAT 1 + +#define BRW_DEPTHFORMAT_D32_FLOAT_S8X24_UINT 0 +#define BRW_DEPTHFORMAT_D32_FLOAT 1 +#define BRW_DEPTHFORMAT_D24_UNORM_S8_UINT 2 +#define BRW_DEPTHFORMAT_D16_UNORM 5 + +#define BRW_FLOATING_POINT_IEEE_754 0 +#define BRW_FLOATING_POINT_NON_IEEE_754 1 + +#define BRW_FRONTWINDING_CW 0 +#define BRW_FRONTWINDING_CCW 1 + +#define BRW_SPRITE_POINT_ENABLE 16 + +#define BRW_INDEX_BYTE 0 +#define BRW_INDEX_WORD 1 +#define BRW_INDEX_DWORD 2 + +#define BRW_LOGICOPFUNCTION_CLEAR 0 +#define BRW_LOGICOPFUNCTION_NOR 1 +#define BRW_LOGICOPFUNCTION_AND_INVERTED 2 +#define BRW_LOGICOPFUNCTION_COPY_INVERTED 3 +#define BRW_LOGICOPFUNCTION_AND_REVERSE 4 +#define BRW_LOGICOPFUNCTION_INVERT 5 +#define BRW_LOGICOPFUNCTION_XOR 6 +#define BRW_LOGICOPFUNCTION_NAND 7 +#define BRW_LOGICOPFUNCTION_AND 8 +#define BRW_LOGICOPFUNCTION_EQUIV 9 +#define BRW_LOGICOPFUNCTION_NOOP 10 +#define BRW_LOGICOPFUNCTION_OR_INVERTED 11 +#define BRW_LOGICOPFUNCTION_COPY 12 +#define BRW_LOGICOPFUNCTION_OR_REVERSE 13 +#define BRW_LOGICOPFUNCTION_OR 14 +#define BRW_LOGICOPFUNCTION_SET 15 + +#define BRW_MAPFILTER_NEAREST 0x0 +#define BRW_MAPFILTER_LINEAR 0x1 +#define BRW_MAPFILTER_ANISOTROPIC 0x2 + +#define BRW_MIPFILTER_NONE 0 +#define BRW_MIPFILTER_NEAREST 1 +#define BRW_MIPFILTER_LINEAR 3 + +#define BRW_POLYGON_FRONT_FACING 0 +#define BRW_POLYGON_BACK_FACING 1 + +#define BRW_PREFILTER_ALWAYS 0x0 +#define BRW_PREFILTER_NEVER 0x1 +#define BRW_PREFILTER_LESS 0x2 +#define BRW_PREFILTER_EQUAL 0x3 +#define BRW_PREFILTER_LEQUAL 0x4 +#define BRW_PREFILTER_GREATER 0x5 +#define BRW_PREFILTER_NOTEQUAL 0x6 +#define BRW_PREFILTER_GEQUAL 0x7 + +#define BRW_PROVOKING_VERTEX_0 0 +#define BRW_PROVOKING_VERTEX_1 1 +#define BRW_PROVOKING_VERTEX_2 2 + +#define BRW_RASTRULE_UPPER_LEFT 0 +#define BRW_RASTRULE_UPPER_RIGHT 1 + +#define BRW_RENDERTARGET_CLAMPRANGE_UNORM 0 +#define BRW_RENDERTARGET_CLAMPRANGE_SNORM 1 +#define BRW_RENDERTARGET_CLAMPRANGE_FORMAT 2 + +#define BRW_STENCILOP_KEEP 0 +#define BRW_STENCILOP_ZERO 1 +#define BRW_STENCILOP_REPLACE 2 +#define BRW_STENCILOP_INCRSAT 3 +#define BRW_STENCILOP_DECRSAT 4 +#define BRW_STENCILOP_INCR 5 +#define BRW_STENCILOP_DECR 6 +#define BRW_STENCILOP_INVERT 7 + +#define BRW_SURFACE_MIPMAPLAYOUT_BELOW 0 +#define BRW_SURFACE_MIPMAPLAYOUT_RIGHT 1 + +#define BRW_SURFACEFORMAT_R32G32B32A32_FLOAT 0x000 +#define BRW_SURFACEFORMAT_R32G32B32A32_SINT 0x001 +#define BRW_SURFACEFORMAT_R32G32B32A32_UINT 0x002 +#define BRW_SURFACEFORMAT_R32G32B32A32_UNORM 0x003 +#define BRW_SURFACEFORMAT_R32G32B32A32_SNORM 0x004 +#define BRW_SURFACEFORMAT_R64G64_FLOAT 0x005 +#define BRW_SURFACEFORMAT_R32G32B32X32_FLOAT 0x006 +#define BRW_SURFACEFORMAT_R32G32B32A32_SSCALED 0x007 +#define BRW_SURFACEFORMAT_R32G32B32A32_USCALED 0x008 +#define BRW_SURFACEFORMAT_R32G32B32_FLOAT 0x040 +#define BRW_SURFACEFORMAT_R32G32B32_SINT 0x041 +#define BRW_SURFACEFORMAT_R32G32B32_UINT 0x042 +#define BRW_SURFACEFORMAT_R32G32B32_UNORM 0x043 +#define BRW_SURFACEFORMAT_R32G32B32_SNORM 0x044 +#define BRW_SURFACEFORMAT_R32G32B32_SSCALED 0x045 +#define BRW_SURFACEFORMAT_R32G32B32_USCALED 0x046 +#define BRW_SURFACEFORMAT_R16G16B16A16_UNORM 0x080 +#define BRW_SURFACEFORMAT_R16G16B16A16_SNORM 0x081 +#define BRW_SURFACEFORMAT_R16G16B16A16_SINT 0x082 +#define BRW_SURFACEFORMAT_R16G16B16A16_UINT 0x083 +#define BRW_SURFACEFORMAT_R16G16B16A16_FLOAT 0x084 +#define BRW_SURFACEFORMAT_R32G32_FLOAT 0x085 +#define BRW_SURFACEFORMAT_R32G32_SINT 0x086 +#define BRW_SURFACEFORMAT_R32G32_UINT 0x087 +#define BRW_SURFACEFORMAT_R32_FLOAT_X8X24_TYPELESS 0x088 +#define BRW_SURFACEFORMAT_X32_TYPELESS_G8X24_UINT 0x089 +#define BRW_SURFACEFORMAT_L32A32_FLOAT 0x08A +#define BRW_SURFACEFORMAT_R32G32_UNORM 0x08B +#define BRW_SURFACEFORMAT_R32G32_SNORM 0x08C +#define BRW_SURFACEFORMAT_R64_FLOAT 0x08D +#define BRW_SURFACEFORMAT_R16G16B16X16_UNORM 0x08E +#define BRW_SURFACEFORMAT_R16G16B16X16_FLOAT 0x08F +#define BRW_SURFACEFORMAT_A32X32_FLOAT 0x090 +#define BRW_SURFACEFORMAT_L32X32_FLOAT 0x091 +#define BRW_SURFACEFORMAT_I32X32_FLOAT 0x092 +#define BRW_SURFACEFORMAT_R16G16B16A16_SSCALED 0x093 +#define BRW_SURFACEFORMAT_R16G16B16A16_USCALED 0x094 +#define BRW_SURFACEFORMAT_R32G32_SSCALED 0x095 +#define BRW_SURFACEFORMAT_R32G32_USCALED 0x096 +#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM 0x0C0 +#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB 0x0C1 +#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM 0x0C2 +#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM_SRGB 0x0C3 +#define BRW_SURFACEFORMAT_R10G10B10A2_UINT 0x0C4 +#define BRW_SURFACEFORMAT_R10G10B10_SNORM_A2_UNORM 0x0C5 +#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM 0x0C7 +#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB 0x0C8 +#define BRW_SURFACEFORMAT_R8G8B8A8_SNORM 0x0C9 +#define BRW_SURFACEFORMAT_R8G8B8A8_SINT 0x0CA +#define BRW_SURFACEFORMAT_R8G8B8A8_UINT 0x0CB +#define BRW_SURFACEFORMAT_R16G16_UNORM 0x0CC +#define BRW_SURFACEFORMAT_R16G16_SNORM 0x0CD +#define BRW_SURFACEFORMAT_R16G16_SINT 0x0CE +#define BRW_SURFACEFORMAT_R16G16_UINT 0x0CF +#define BRW_SURFACEFORMAT_R16G16_FLOAT 0x0D0 +#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM 0x0D1 +#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM_SRGB 0x0D2 +#define BRW_SURFACEFORMAT_R11G11B10_FLOAT 0x0D3 +#define BRW_SURFACEFORMAT_R32_SINT 0x0D6 +#define BRW_SURFACEFORMAT_R32_UINT 0x0D7 +#define BRW_SURFACEFORMAT_R32_FLOAT 0x0D8 +#define BRW_SURFACEFORMAT_R24_UNORM_X8_TYPELESS 0x0D9 +#define BRW_SURFACEFORMAT_X24_TYPELESS_G8_UINT 0x0DA +#define BRW_SURFACEFORMAT_L16A16_UNORM 0x0DF +#define BRW_SURFACEFORMAT_I24X8_UNORM 0x0E0 +#define BRW_SURFACEFORMAT_L24X8_UNORM 0x0E1 +#define BRW_SURFACEFORMAT_A24X8_UNORM 0x0E2 +#define BRW_SURFACEFORMAT_I32_FLOAT 0x0E3 +#define BRW_SURFACEFORMAT_L32_FLOAT 0x0E4 +#define BRW_SURFACEFORMAT_A32_FLOAT 0x0E5 +#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM 0x0E9 +#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM_SRGB 0x0EA +#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM 0x0EB +#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM_SRGB 0x0EC +#define BRW_SURFACEFORMAT_R9G9B9E5_SHAREDEXP 0x0ED +#define BRW_SURFACEFORMAT_B10G10R10X2_UNORM 0x0EE +#define BRW_SURFACEFORMAT_L16A16_FLOAT 0x0F0 +#define BRW_SURFACEFORMAT_R32_UNORM 0x0F1 +#define BRW_SURFACEFORMAT_R32_SNORM 0x0F2 +#define BRW_SURFACEFORMAT_R10G10B10X2_USCALED 0x0F3 +#define BRW_SURFACEFORMAT_R8G8B8A8_SSCALED 0x0F4 +#define BRW_SURFACEFORMAT_R8G8B8A8_USCALED 0x0F5 +#define BRW_SURFACEFORMAT_R16G16_SSCALED 0x0F6 +#define BRW_SURFACEFORMAT_R16G16_USCALED 0x0F7 +#define BRW_SURFACEFORMAT_R32_SSCALED 0x0F8 +#define BRW_SURFACEFORMAT_R32_USCALED 0x0F9 +#define BRW_SURFACEFORMAT_B5G6R5_UNORM 0x100 +#define BRW_SURFACEFORMAT_B5G6R5_UNORM_SRGB 0x101 +#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM 0x102 +#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM_SRGB 0x103 +#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM 0x104 +#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM_SRGB 0x105 +#define BRW_SURFACEFORMAT_R8G8_UNORM 0x106 +#define BRW_SURFACEFORMAT_R8G8_SNORM 0x107 +#define BRW_SURFACEFORMAT_R8G8_SINT 0x108 +#define BRW_SURFACEFORMAT_R8G8_UINT 0x109 +#define BRW_SURFACEFORMAT_R16_UNORM 0x10A +#define BRW_SURFACEFORMAT_R16_SNORM 0x10B +#define BRW_SURFACEFORMAT_R16_SINT 0x10C +#define BRW_SURFACEFORMAT_R16_UINT 0x10D +#define BRW_SURFACEFORMAT_R16_FLOAT 0x10E +#define BRW_SURFACEFORMAT_I16_UNORM 0x111 +#define BRW_SURFACEFORMAT_L16_UNORM 0x112 +#define BRW_SURFACEFORMAT_A16_UNORM 0x113 +#define BRW_SURFACEFORMAT_L8A8_UNORM 0x114 +#define BRW_SURFACEFORMAT_I16_FLOAT 0x115 +#define BRW_SURFACEFORMAT_L16_FLOAT 0x116 +#define BRW_SURFACEFORMAT_A16_FLOAT 0x117 +#define BRW_SURFACEFORMAT_R5G5_SNORM_B6_UNORM 0x119 +#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM 0x11A +#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM_SRGB 0x11B +#define BRW_SURFACEFORMAT_R8G8_SSCALED 0x11C +#define BRW_SURFACEFORMAT_R8G8_USCALED 0x11D +#define BRW_SURFACEFORMAT_R16_SSCALED 0x11E +#define BRW_SURFACEFORMAT_R16_USCALED 0x11F +#define BRW_SURFACEFORMAT_R8_UNORM 0x140 +#define BRW_SURFACEFORMAT_R8_SNORM 0x141 +#define BRW_SURFACEFORMAT_R8_SINT 0x142 +#define BRW_SURFACEFORMAT_R8_UINT 0x143 +#define BRW_SURFACEFORMAT_A8_UNORM 0x144 +#define BRW_SURFACEFORMAT_I8_UNORM 0x145 +#define BRW_SURFACEFORMAT_L8_UNORM 0x146 +#define BRW_SURFACEFORMAT_P4A4_UNORM 0x147 +#define BRW_SURFACEFORMAT_A4P4_UNORM 0x148 +#define BRW_SURFACEFORMAT_R8_SSCALED 0x149 +#define BRW_SURFACEFORMAT_R8_USCALED 0x14A +#define BRW_SURFACEFORMAT_R1_UINT 0x181 +#define BRW_SURFACEFORMAT_YCRCB_NORMAL 0x182 +#define BRW_SURFACEFORMAT_YCRCB_SWAPUVY 0x183 +#define BRW_SURFACEFORMAT_BC1_UNORM 0x186 +#define BRW_SURFACEFORMAT_BC2_UNORM 0x187 +#define BRW_SURFACEFORMAT_BC3_UNORM 0x188 +#define BRW_SURFACEFORMAT_BC4_UNORM 0x189 +#define BRW_SURFACEFORMAT_BC5_UNORM 0x18A +#define BRW_SURFACEFORMAT_BC1_UNORM_SRGB 0x18B +#define BRW_SURFACEFORMAT_BC2_UNORM_SRGB 0x18C +#define BRW_SURFACEFORMAT_BC3_UNORM_SRGB 0x18D +#define BRW_SURFACEFORMAT_MONO8 0x18E +#define BRW_SURFACEFORMAT_YCRCB_SWAPUV 0x18F +#define BRW_SURFACEFORMAT_YCRCB_SWAPY 0x190 +#define BRW_SURFACEFORMAT_DXT1_RGB 0x191 +#define BRW_SURFACEFORMAT_FXT1 0x192 +#define BRW_SURFACEFORMAT_R8G8B8_UNORM 0x193 +#define BRW_SURFACEFORMAT_R8G8B8_SNORM 0x194 +#define BRW_SURFACEFORMAT_R8G8B8_SSCALED 0x195 +#define BRW_SURFACEFORMAT_R8G8B8_USCALED 0x196 +#define BRW_SURFACEFORMAT_R64G64B64A64_FLOAT 0x197 +#define BRW_SURFACEFORMAT_R64G64B64_FLOAT 0x198 +#define BRW_SURFACEFORMAT_BC4_SNORM 0x199 +#define BRW_SURFACEFORMAT_BC5_SNORM 0x19A +#define BRW_SURFACEFORMAT_R16G16B16_UNORM 0x19C +#define BRW_SURFACEFORMAT_R16G16B16_SNORM 0x19D +#define BRW_SURFACEFORMAT_R16G16B16_SSCALED 0x19E +#define BRW_SURFACEFORMAT_R16G16B16_USCALED 0x19F + +#define BRW_SURFACERETURNFORMAT_FLOAT32 0 +#define BRW_SURFACERETURNFORMAT_S1 1 + +#define BRW_SURFACE_1D 0 +#define BRW_SURFACE_2D 1 +#define BRW_SURFACE_3D 2 +#define BRW_SURFACE_CUBE 3 +#define BRW_SURFACE_BUFFER 4 +#define BRW_SURFACE_NULL 7 + +#define BRW_TEXCOORDMODE_WRAP 0 +#define BRW_TEXCOORDMODE_MIRROR 1 +#define BRW_TEXCOORDMODE_CLAMP 2 +#define BRW_TEXCOORDMODE_CUBE 3 +#define BRW_TEXCOORDMODE_CLAMP_BORDER 4 +#define BRW_TEXCOORDMODE_MIRROR_ONCE 5 + +#define BRW_THREAD_PRIORITY_NORMAL 0 +#define BRW_THREAD_PRIORITY_HIGH 1 + +#define BRW_TILEWALK_XMAJOR 0 +#define BRW_TILEWALK_YMAJOR 1 + +#define BRW_VERTEX_SUBPIXEL_PRECISION_8BITS 0 +#define BRW_VERTEX_SUBPIXEL_PRECISION_4BITS 1 + +#define BRW_VERTEXBUFFER_ACCESS_VERTEXDATA 0 +#define BRW_VERTEXBUFFER_ACCESS_INSTANCEDATA 1 + +#define BRW_VFCOMPONENT_NOSTORE 0 +#define BRW_VFCOMPONENT_STORE_SRC 1 +#define BRW_VFCOMPONENT_STORE_0 2 +#define BRW_VFCOMPONENT_STORE_1_FLT 3 +#define BRW_VFCOMPONENT_STORE_1_INT 4 +#define BRW_VFCOMPONENT_STORE_VID 5 +#define BRW_VFCOMPONENT_STORE_IID 6 +#define BRW_VFCOMPONENT_STORE_PID 7 + + + +/* Execution Unit (EU) defines + */ + +#define BRW_ALIGN_1 0 +#define BRW_ALIGN_16 1 + +#define BRW_ADDRESS_DIRECT 0 +#define BRW_ADDRESS_REGISTER_INDIRECT_REGISTER 1 + +#define BRW_CHANNEL_X 0 +#define BRW_CHANNEL_Y 1 +#define BRW_CHANNEL_Z 2 +#define BRW_CHANNEL_W 3 + +#define BRW_COMPRESSION_NONE 0 +#define BRW_COMPRESSION_2NDHALF 1 +#define BRW_COMPRESSION_COMPRESSED 2 + +#define BRW_CONDITIONAL_NONE 0 +#define BRW_CONDITIONAL_Z 1 +#define BRW_CONDITIONAL_NZ 2 +#define BRW_CONDITIONAL_EQ 1 /* Z */ +#define BRW_CONDITIONAL_NEQ 2 /* NZ */ +#define BRW_CONDITIONAL_G 3 +#define BRW_CONDITIONAL_GE 4 +#define BRW_CONDITIONAL_L 5 +#define BRW_CONDITIONAL_LE 6 +#define BRW_CONDITIONAL_C 7 +#define BRW_CONDITIONAL_O 8 + +#define BRW_DEBUG_NONE 0 +#define BRW_DEBUG_BREAKPOINT 1 + +#define BRW_DEPENDENCY_NORMAL 0 +#define BRW_DEPENDENCY_NOTCLEARED 1 +#define BRW_DEPENDENCY_NOTCHECKED 2 +#define BRW_DEPENDENCY_DISABLE 3 + +#define BRW_EXECUTE_1 0 +#define BRW_EXECUTE_2 1 +#define BRW_EXECUTE_4 2 +#define BRW_EXECUTE_8 3 +#define BRW_EXECUTE_16 4 +#define BRW_EXECUTE_32 5 + +#define BRW_HORIZONTAL_STRIDE_0 0 +#define BRW_HORIZONTAL_STRIDE_1 1 +#define BRW_HORIZONTAL_STRIDE_2 2 +#define BRW_HORIZONTAL_STRIDE_4 3 + +#define BRW_INSTRUCTION_NORMAL 0 +#define BRW_INSTRUCTION_SATURATE 1 + +#define BRW_MASK_ENABLE 0 +#define BRW_MASK_DISABLE 1 + +#define BRW_OPCODE_MOV 1 +#define BRW_OPCODE_SEL 2 +#define BRW_OPCODE_NOT 4 +#define BRW_OPCODE_AND 5 +#define BRW_OPCODE_OR 6 +#define BRW_OPCODE_XOR 7 +#define BRW_OPCODE_SHR 8 +#define BRW_OPCODE_SHL 9 +#define BRW_OPCODE_RSR 10 +#define BRW_OPCODE_RSL 11 +#define BRW_OPCODE_ASR 12 +#define BRW_OPCODE_CMP 16 +#define BRW_OPCODE_JMPI 32 +#define BRW_OPCODE_IF 34 +#define BRW_OPCODE_IFF 35 +#define BRW_OPCODE_ELSE 36 +#define BRW_OPCODE_ENDIF 37 +#define BRW_OPCODE_DO 38 +#define BRW_OPCODE_WHILE 39 +#define BRW_OPCODE_BREAK 40 +#define BRW_OPCODE_CONTINUE 41 +#define BRW_OPCODE_HALT 42 +#define BRW_OPCODE_MSAVE 44 +#define BRW_OPCODE_MRESTORE 45 +#define BRW_OPCODE_PUSH 46 +#define BRW_OPCODE_POP 47 +#define BRW_OPCODE_WAIT 48 +#define BRW_OPCODE_SEND 49 +#define BRW_OPCODE_ADD 64 +#define BRW_OPCODE_MUL 65 +#define BRW_OPCODE_AVG 66 +#define BRW_OPCODE_FRC 67 +#define BRW_OPCODE_RNDU 68 +#define BRW_OPCODE_RNDD 69 +#define BRW_OPCODE_RNDE 70 +#define BRW_OPCODE_RNDZ 71 +#define BRW_OPCODE_MAC 72 +#define BRW_OPCODE_MACH 73 +#define BRW_OPCODE_LZD 74 +#define BRW_OPCODE_SAD2 80 +#define BRW_OPCODE_SADA2 81 +#define BRW_OPCODE_DP4 84 +#define BRW_OPCODE_DPH 85 +#define BRW_OPCODE_DP3 86 +#define BRW_OPCODE_DP2 87 +#define BRW_OPCODE_DPA2 88 +#define BRW_OPCODE_LINE 89 +#define BRW_OPCODE_NOP 126 + +#define BRW_PREDICATE_NONE 0 +#define BRW_PREDICATE_NORMAL 1 +#define BRW_PREDICATE_ALIGN1_ANYV 2 +#define BRW_PREDICATE_ALIGN1_ALLV 3 +#define BRW_PREDICATE_ALIGN1_ANY2H 4 +#define BRW_PREDICATE_ALIGN1_ALL2H 5 +#define BRW_PREDICATE_ALIGN1_ANY4H 6 +#define BRW_PREDICATE_ALIGN1_ALL4H 7 +#define BRW_PREDICATE_ALIGN1_ANY8H 8 +#define BRW_PREDICATE_ALIGN1_ALL8H 9 +#define BRW_PREDICATE_ALIGN1_ANY16H 10 +#define BRW_PREDICATE_ALIGN1_ALL16H 11 +#define BRW_PREDICATE_ALIGN16_REPLICATE_X 2 +#define BRW_PREDICATE_ALIGN16_REPLICATE_Y 3 +#define BRW_PREDICATE_ALIGN16_REPLICATE_Z 4 +#define BRW_PREDICATE_ALIGN16_REPLICATE_W 5 +#define BRW_PREDICATE_ALIGN16_ANY4H 6 +#define BRW_PREDICATE_ALIGN16_ALL4H 7 + +#define BRW_ARCHITECTURE_REGISTER_FILE 0 +#define BRW_GENERAL_REGISTER_FILE 1 +#define BRW_MESSAGE_REGISTER_FILE 2 +#define BRW_IMMEDIATE_VALUE 3 + +#define BRW_REGISTER_TYPE_UD 0 +#define BRW_REGISTER_TYPE_D 1 +#define BRW_REGISTER_TYPE_UW 2 +#define BRW_REGISTER_TYPE_W 3 +#define BRW_REGISTER_TYPE_UB 4 +#define BRW_REGISTER_TYPE_B 5 +#define BRW_REGISTER_TYPE_VF 5 /* packed float vector, immediates only? */ +#define BRW_REGISTER_TYPE_HF 6 +#define BRW_REGISTER_TYPE_V 6 /* packed int vector, immediates only, uword dest only */ +#define BRW_REGISTER_TYPE_F 7 + +#define BRW_ARF_NULL 0x00 +#define BRW_ARF_ADDRESS 0x10 +#define BRW_ARF_ACCUMULATOR 0x20 +#define BRW_ARF_FLAG 0x30 +#define BRW_ARF_MASK 0x40 +#define BRW_ARF_MASK_STACK 0x50 +#define BRW_ARF_MASK_STACK_DEPTH 0x60 +#define BRW_ARF_STATE 0x70 +#define BRW_ARF_CONTROL 0x80 +#define BRW_ARF_NOTIFICATION_COUNT 0x90 +#define BRW_ARF_IP 0xA0 + +#define BRW_AMASK 0 +#define BRW_IMASK 1 +#define BRW_LMASK 2 +#define BRW_CMASK 3 + + + +#define BRW_THREAD_NORMAL 0 +#define BRW_THREAD_ATOMIC 1 +#define BRW_THREAD_SWITCH 2 + +#define BRW_VERTICAL_STRIDE_0 0 +#define BRW_VERTICAL_STRIDE_1 1 +#define BRW_VERTICAL_STRIDE_2 2 +#define BRW_VERTICAL_STRIDE_4 3 +#define BRW_VERTICAL_STRIDE_8 4 +#define BRW_VERTICAL_STRIDE_16 5 +#define BRW_VERTICAL_STRIDE_32 6 +#define BRW_VERTICAL_STRIDE_64 7 +#define BRW_VERTICAL_STRIDE_128 8 +#define BRW_VERTICAL_STRIDE_256 9 +#define BRW_VERTICAL_STRIDE_ONE_DIMENSIONAL 0xF + +#define BRW_WIDTH_1 0 +#define BRW_WIDTH_2 1 +#define BRW_WIDTH_4 2 +#define BRW_WIDTH_8 3 +#define BRW_WIDTH_16 4 + +#define BRW_STATELESS_BUFFER_BOUNDARY_1K 0 +#define BRW_STATELESS_BUFFER_BOUNDARY_2K 1 +#define BRW_STATELESS_BUFFER_BOUNDARY_4K 2 +#define BRW_STATELESS_BUFFER_BOUNDARY_8K 3 +#define BRW_STATELESS_BUFFER_BOUNDARY_16K 4 +#define BRW_STATELESS_BUFFER_BOUNDARY_32K 5 +#define BRW_STATELESS_BUFFER_BOUNDARY_64K 6 +#define BRW_STATELESS_BUFFER_BOUNDARY_128K 7 +#define BRW_STATELESS_BUFFER_BOUNDARY_256K 8 +#define BRW_STATELESS_BUFFER_BOUNDARY_512K 9 +#define BRW_STATELESS_BUFFER_BOUNDARY_1M 10 +#define BRW_STATELESS_BUFFER_BOUNDARY_2M 11 + +#define BRW_POLYGON_FACING_FRONT 0 +#define BRW_POLYGON_FACING_BACK 1 + +#define BRW_MESSAGE_TARGET_NULL 0 +#define BRW_MESSAGE_TARGET_MATH 1 +#define BRW_MESSAGE_TARGET_SAMPLER 2 +#define BRW_MESSAGE_TARGET_GATEWAY 3 +#define BRW_MESSAGE_TARGET_DATAPORT_READ 4 +#define BRW_MESSAGE_TARGET_DATAPORT_WRITE 5 +#define BRW_MESSAGE_TARGET_URB 6 +#define BRW_MESSAGE_TARGET_THREAD_SPAWNER 7 + +#define BRW_SAMPLER_RETURN_FORMAT_FLOAT32 0 +#define BRW_SAMPLER_RETURN_FORMAT_UINT32 2 +#define BRW_SAMPLER_RETURN_FORMAT_SINT32 3 + +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS 0 +#define BRW_SAMPLER_MESSAGE_SIMD8_KILLPIX 1 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD 1 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD 1 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS 2 +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_RESINFO 2 +#define BRW_SAMPLER_MESSAGE_SIMD8_RESINFO 2 +#define BRW_SAMPLER_MESSAGE_SIMD16_RESINFO 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_LD 3 +#define BRW_SAMPLER_MESSAGE_SIMD8_LD 3 +#define BRW_SAMPLER_MESSAGE_SIMD16_LD 3 + +#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW 0 +#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDHIGH 1 +#define BRW_DATAPORT_OWORD_BLOCK_2_OWORDS 2 +#define BRW_DATAPORT_OWORD_BLOCK_4_OWORDS 3 +#define BRW_DATAPORT_OWORD_BLOCK_8_OWORDS 4 + +#define BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD 0 +#define BRW_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS 2 + +#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS 2 +#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS 3 + +#define BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ 0 +#define BRW_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 1 +#define BRW_DATAPORT_READ_MESSAGE_DWORD_BLOCK_READ 2 +#define BRW_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 3 + +#define BRW_DATAPORT_READ_TARGET_DATA_CACHE 0 +#define BRW_DATAPORT_READ_TARGET_RENDER_CACHE 1 +#define BRW_DATAPORT_READ_TARGET_SAMPLER_CACHE 2 + +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE 0 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED 1 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01 2 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23 3 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01 4 + +#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE 0 +#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE 1 +#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_BLOCK_WRITE 2 +#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE 3 +#define BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE 4 +#define BRW_DATAPORT_WRITE_MESSAGE_STREAMED_VERTEX_BUFFER_WRITE 5 +#define BRW_DATAPORT_WRITE_MESSAGE_FLUSH_RENDER_CACHE 7 + +#define BRW_MATH_FUNCTION_INV 1 +#define BRW_MATH_FUNCTION_LOG 2 +#define BRW_MATH_FUNCTION_EXP 3 +#define BRW_MATH_FUNCTION_SQRT 4 +#define BRW_MATH_FUNCTION_RSQ 5 +#define BRW_MATH_FUNCTION_SIN 6 /* was 7 */ +#define BRW_MATH_FUNCTION_COS 7 /* was 8 */ +#define BRW_MATH_FUNCTION_SINCOS 8 /* was 6 */ +#define BRW_MATH_FUNCTION_TAN 9 +#define BRW_MATH_FUNCTION_POW 10 +#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER 11 +#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT 12 +#define BRW_MATH_FUNCTION_INT_DIV_REMAINDER 13 + +#define BRW_MATH_INTEGER_UNSIGNED 0 +#define BRW_MATH_INTEGER_SIGNED 1 + +#define BRW_MATH_PRECISION_FULL 0 +#define BRW_MATH_PRECISION_PARTIAL 1 + +#define BRW_MATH_SATURATE_NONE 0 +#define BRW_MATH_SATURATE_SATURATE 1 + +#define BRW_MATH_DATA_VECTOR 0 +#define BRW_MATH_DATA_SCALAR 1 + +#define BRW_URB_OPCODE_WRITE 0 + +#define BRW_URB_SWIZZLE_NONE 0 +#define BRW_URB_SWIZZLE_INTERLEAVE 1 +#define BRW_URB_SWIZZLE_TRANSPOSE 2 + +#define BRW_SCRATCH_SPACE_SIZE_1K 0 +#define BRW_SCRATCH_SPACE_SIZE_2K 1 +#define BRW_SCRATCH_SPACE_SIZE_4K 2 +#define BRW_SCRATCH_SPACE_SIZE_8K 3 +#define BRW_SCRATCH_SPACE_SIZE_16K 4 +#define BRW_SCRATCH_SPACE_SIZE_32K 5 +#define BRW_SCRATCH_SPACE_SIZE_64K 6 +#define BRW_SCRATCH_SPACE_SIZE_128K 7 +#define BRW_SCRATCH_SPACE_SIZE_256K 8 +#define BRW_SCRATCH_SPACE_SIZE_512K 9 +#define BRW_SCRATCH_SPACE_SIZE_1M 10 +#define BRW_SCRATCH_SPACE_SIZE_2M 11 + + + + +#define CMD_URB_FENCE 0x6000 +#define CMD_CONST_BUFFER_STATE 0x6001 +#define CMD_CONST_BUFFER 0x6002 + +#define CMD_STATE_BASE_ADDRESS 0x6101 +#define CMD_STATE_INSN_POINTER 0x6102 +#define CMD_PIPELINE_SELECT 0x6104 + +#define CMD_PIPELINED_STATE_POINTERS 0x7800 +#define CMD_BINDING_TABLE_PTRS 0x7801 +#define CMD_VERTEX_BUFFER 0x7808 +#define CMD_VERTEX_ELEMENT 0x7809 +#define CMD_INDEX_BUFFER 0x780a +#define CMD_VF_STATISTICS 0x780b + +#define CMD_DRAW_RECT 0x7900 +#define CMD_BLEND_CONSTANT_COLOR 0x7901 +#define CMD_CHROMA_KEY 0x7904 +#define CMD_DEPTH_BUFFER 0x7905 +#define CMD_POLY_STIPPLE_OFFSET 0x7906 +#define CMD_POLY_STIPPLE_PATTERN 0x7907 +#define CMD_LINE_STIPPLE_PATTERN 0x7908 +#define CMD_GLOBAL_DEPTH_OFFSET_CLAMP 0x7909 + +#define CMD_PIPE_CONTROL 0x7a00 + +#define CMD_3D_PRIM 0x7b00 + +#define CMD_MI_FLUSH 0x0200 + + +/* Various values from the R0 vertex header: + */ +#define R02_PRIM_END 0x1 +#define R02_PRIM_START 0x2 + + + +#endif diff --git a/src/mesa/pipe/i965simple/brw_draw.c b/src/mesa/pipe/i965simple/brw_draw.c new file mode 100644 index 0000000000..8b92f11280 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_draw.c @@ -0,0 +1,238 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include + +#include "brw_batch.h" +#include "brw_draw.h" +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_state.h" + +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = { + _3DPRIM_POINTLIST, + _3DPRIM_LINELIST, + _3DPRIM_LINELOOP, + _3DPRIM_LINESTRIP, + _3DPRIM_TRILIST, + _3DPRIM_TRISTRIP, + _3DPRIM_TRIFAN, + _3DPRIM_QUADLIST, + _3DPRIM_QUADSTRIP, + _3DPRIM_POLYGON +}; + + +static const int reduced_prim[PIPE_PRIM_POLYGON+1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES +}; + + +/* When the primitive changes, set a state bit and re-validate. Not + * the nicest and would rather deal with this by having all the + * programs be immune to the active primitive (ie. cope with all + * possibilities). That may not be realistic however. + */ +static void brw_set_prim(struct brw_context *brw, int prim) +{ + PRINT("PRIM: %d\n", prim); + + /* Slight optimization to avoid the GS program when not needed: + */ + if (prim == PIPE_PRIM_QUAD_STRIP && + brw->attribs.Raster->flatshade && + brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_FILL && + brw->attribs.Raster->fill_ccw == PIPE_POLYGON_MODE_FILL) + prim = PIPE_PRIM_TRIANGLE_STRIP; + + if (prim != brw->primitive) { + brw->primitive = prim; + brw->state.dirty.brw |= BRW_NEW_PRIMITIVE; + + if (reduced_prim[prim] != brw->reduced_primitive) { + brw->reduced_primitive = reduced_prim[prim]; + brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE; + } + + brw_validate_state(brw); + } + +} + + +static unsigned trim(int prim, unsigned length) +{ + if (prim == PIPE_PRIM_QUAD_STRIP) + return length > 3 ? (length - length % 2) : 0; + else if (prim == PIPE_PRIM_QUADS) + return length - length % 4; + else + return length; +} + + + +static boolean brw_emit_prim( struct brw_context *brw, + boolean indexed, + unsigned start, + unsigned count ) + +{ + struct brw_3d_primitive prim_packet; + + if (BRW_DEBUG & DEBUG_PRIMS) + PRINT("PRIM: %d %d %d\n", brw->primitive, start, count); + + prim_packet.header.opcode = CMD_3D_PRIM; + prim_packet.header.length = sizeof(prim_packet)/4 - 2; + prim_packet.header.pad = 0; + prim_packet.header.topology = hw_prim[brw->primitive]; + prim_packet.header.indexed = indexed; + + prim_packet.verts_per_instance = trim(brw->primitive, count); + prim_packet.start_vert_location = start; + prim_packet.instance_count = 1; + prim_packet.start_instance_location = 0; + prim_packet.base_vert_location = 0; + + if (prim_packet.verts_per_instance == 0) + return TRUE; + + return brw_batchbuffer_data( brw->winsys, + &prim_packet, + sizeof(prim_packet) ); +} + + +/* May fail if out of video memory for texture or vbo upload, or on + * fallback conditions. + */ +static boolean brw_try_draw_elements( struct pipe_context *pipe, + struct pipe_buffer_handle *index_buffer, + unsigned index_size, + unsigned mode, + unsigned start, + unsigned count ) +{ + struct brw_context *brw = brw_context(pipe); + boolean retval = FALSE; + + /* Set the first primitive ahead of validate_state: + */ + brw_set_prim(brw, mode); + + /* Upload index, vertex data: + */ + if (!brw_upload_indices( brw, index_buffer, index_size, start, count )) + return FALSE; + + if (!brw_upload_vertex_elements( brw )) + return FALSE; + + /* XXX: Need to separate validate and upload of state. + */ + if (brw->state.dirty.brw) + brw_validate_state( brw ); + + if (brw_emit_prim(brw, TRUE, start, count)) + return FALSE; + + return TRUE; +} + + + +static boolean brw_draw_elements( struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned mode, + unsigned start, + unsigned count ) +{ + if (!brw_try_draw_elements( pipe, + indexBuffer, + indexSize, + mode, start, count )) + { + /* flush ? */ + + if (!brw_try_draw_elements( pipe, + indexBuffer, + indexSize, + mode, start, + count )) { + assert(0); + return FALSE; + } + } + + return TRUE; +} + + + +static boolean brw_draw_arrays( struct pipe_context *pipe, + unsigned mode, + unsigned start, + unsigned count ) +{ + if (!brw_try_draw_elements( pipe, NULL, 0, mode, start, count )) { + /* flush ? */ + + if (!brw_try_draw_elements( pipe, NULL, 0, mode, start, count )) { + assert(0); + return FALSE; + } + } + + return TRUE; +} + + + +void brw_draw_init( struct brw_context *brw ) +{ + brw->pipe.draw_arrays = brw_draw_arrays; + brw->pipe.draw_elements = brw_draw_elements; +} + +void brw_draw_destroy( struct brw_context *brw ) +{ +} + diff --git a/src/mesa/pipe/i965simple/brw_draw.h b/src/mesa/pipe/i965simple/brw_draw.h new file mode 100644 index 0000000000..ebac8365c8 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_draw.h @@ -0,0 +1,56 @@ + /************************************************************************** + * + * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef BRW_DRAW_H +#define BRW_DRAW_H + +#include "pipe/p_context.h" + +struct brw_context; + + + +void brw_draw_init( struct brw_context *brw ); +void brw_draw_destroy( struct brw_context *brw ); + + +boolean brw_upload_vertices( struct brw_context *brw, + unsigned min_index, + unsigned max_index ); + +boolean brw_upload_indices(struct brw_context *brw, + const struct pipe_buffer_handle *index_buffer, + int ib_size, int start, int count); + +boolean brw_upload_vertex_buffers( struct brw_context *brw ); +boolean brw_upload_vertex_elements( struct brw_context *brw ); + +unsigned brw_translate_surface_format( unsigned id ); + + + +#endif diff --git a/src/mesa/pipe/i965simple/brw_draw_upload.c b/src/mesa/pipe/i965simple/brw_draw_upload.c new file mode 100644 index 0000000000..186a6274fa --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_draw_upload.c @@ -0,0 +1,306 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include + +#include "brw_batch.h" +#include "brw_draw.h" +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_state.h" + +struct brw_array_state { + union header_union header; + + struct { + union { + struct { + unsigned pitch:11; + unsigned pad:15; + unsigned access_type:1; + unsigned vb_index:5; + } bits; + unsigned dword; + } vb0; + + struct pipe_buffer_handle *buffer; + unsigned offset; + + unsigned max_index; + unsigned instance_data_step_rate; + + } vb[BRW_VBP_MAX]; +}; + + + +unsigned brw_translate_surface_format( unsigned id ) +{ + switch (id) { + case PIPE_FORMAT_R64_FLOAT: + return BRW_SURFACEFORMAT_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return BRW_SURFACEFORMAT_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return BRW_SURFACEFORMAT_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return BRW_SURFACEFORMAT_R64G64B64A64_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + return BRW_SURFACEFORMAT_R32_FLOAT; + case PIPE_FORMAT_R32G32_FLOAT: + return BRW_SURFACEFORMAT_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return BRW_SURFACEFORMAT_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return BRW_SURFACEFORMAT_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return BRW_SURFACEFORMAT_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return BRW_SURFACEFORMAT_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return BRW_SURFACEFORMAT_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return BRW_SURFACEFORMAT_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return BRW_SURFACEFORMAT_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return BRW_SURFACEFORMAT_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return BRW_SURFACEFORMAT_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return BRW_SURFACEFORMAT_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return BRW_SURFACEFORMAT_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return BRW_SURFACEFORMAT_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return BRW_SURFACEFORMAT_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return BRW_SURFACEFORMAT_R32G32B32A32_SNORM; + + case PIPE_FORMAT_R32_SSCALED: + return BRW_SURFACEFORMAT_R32_SSCALED; + case PIPE_FORMAT_R32G32_SSCALED: + return BRW_SURFACEFORMAT_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return BRW_SURFACEFORMAT_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return BRW_SURFACEFORMAT_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return BRW_SURFACEFORMAT_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return BRW_SURFACEFORMAT_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return BRW_SURFACEFORMAT_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return BRW_SURFACEFORMAT_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return BRW_SURFACEFORMAT_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return BRW_SURFACEFORMAT_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return BRW_SURFACEFORMAT_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return BRW_SURFACEFORMAT_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return BRW_SURFACEFORMAT_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return BRW_SURFACEFORMAT_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return BRW_SURFACEFORMAT_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return BRW_SURFACEFORMAT_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return BRW_SURFACEFORMAT_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return BRW_SURFACEFORMAT_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return BRW_SURFACEFORMAT_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return BRW_SURFACEFORMAT_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return BRW_SURFACEFORMAT_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return BRW_SURFACEFORMAT_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return BRW_SURFACEFORMAT_R8G8B8_UNORM; + case PIPE_FORMAT_R8G8B8A8_UNORM: + return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return BRW_SURFACEFORMAT_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return BRW_SURFACEFORMAT_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return BRW_SURFACEFORMAT_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return BRW_SURFACEFORMAT_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return BRW_SURFACEFORMAT_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return BRW_SURFACEFORMAT_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return BRW_SURFACEFORMAT_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return BRW_SURFACEFORMAT_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return BRW_SURFACEFORMAT_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return BRW_SURFACEFORMAT_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return BRW_SURFACEFORMAT_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return BRW_SURFACEFORMAT_R8G8B8A8_SSCALED; + + default: + assert(0); + return 0; + } +} + +static unsigned get_index_type(int type) +{ + switch (type) { + case 1: return BRW_INDEX_BYTE; + case 2: return BRW_INDEX_WORD; + case 4: return BRW_INDEX_DWORD; + default: assert(0); return 0; + } +} + + +boolean brw_upload_vertex_buffers( struct brw_context *brw ) +{ + struct brw_array_state vbp; + unsigned i; + int nr_enabled = brw->vb.last_vb + 1; + + memset(&vbp, 0, sizeof(vbp)); + + /* This is a hardware limit: + */ + if (nr_enabled >= BRW_VEP_MAX) + return FALSE; + + for (i = 0; i < nr_enabled; i++) + { + vbp.vb[i].vb0.bits.pitch = brw->vb.vbo_array[i].pitch; + vbp.vb[i].vb0.bits.pad = 0; + vbp.vb[i].vb0.bits.access_type = BRW_VERTEXBUFFER_ACCESS_VERTEXDATA; + vbp.vb[i].vb0.bits.vb_index = i; + vbp.vb[i].offset = brw->vb.vbo_array[i].buffer_offset; + vbp.vb[i].buffer = brw->vb.vbo_array[i].buffer; + vbp.vb[i].max_index = brw->vb.vbo_array[i].max_index; + } + + + vbp.header.bits.length = (1 + nr_enabled * 4) - 2; + vbp.header.bits.opcode = CMD_VERTEX_BUFFER; + + BEGIN_BATCH(vbp.header.bits.length+2, 0); + OUT_BATCH( vbp.header.dword ); + + for (i = 0; i < nr_enabled; i++) { + OUT_BATCH( vbp.vb[i].vb0.dword ); + OUT_RELOC( vbp.vb[i].buffer, PIPE_BUFFER_FLAG_READ, + vbp.vb[i].offset); + OUT_BATCH( vbp.vb[i].max_index ); + OUT_BATCH( vbp.vb[i].instance_data_step_rate ); + } + ADVANCE_BATCH(); + return TRUE; +} + + + +boolean brw_upload_vertex_elements( struct brw_context *brw ) +{ + struct brw_vertex_element_packet vep; + + unsigned i; + unsigned nr_enabled = brw->attribs.VertexProgram->program.num_inputs; + + memset(&vep, 0, sizeof(vep)); + + for (i = 0; i < nr_enabled; i++) { + struct brw_vertex_element *input = &brw->vb.inputs[i]; + + switch (brw->vb.vbo_array[input->vep.ve0.vertex_buffer_index].pitch) { + case 0: input->vep.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_0; + case 1: input->vep.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; + case 2: input->vep.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; + case 3: input->vep.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; + break; + } + vep.ve[i] = input->vep; + } + + + vep.header.length = (1 + nr_enabled * sizeof(vep.ve[0])/4) - 2; + vep.header.opcode = CMD_VERTEX_ELEMENT; + brw_cached_batch_struct(brw, &vep, 4 + nr_enabled * sizeof(vep.ve[0])); + + return TRUE; +} + +boolean brw_upload_indices( struct brw_context *brw, + const struct pipe_buffer_handle *index_buffer, + int ib_size, int start, int count) +{ + /* Emit the indexbuffer packet: + */ + { + struct brw_indexbuffer ib; + + memset(&ib, 0, sizeof(ib)); + + ib.header.bits.opcode = CMD_INDEX_BUFFER; + ib.header.bits.length = sizeof(ib)/4 - 2; + ib.header.bits.index_format = get_index_type(ib_size); + ib.header.bits.cut_index_enable = 0; + + + BEGIN_BATCH(4, 0); + OUT_BATCH( ib.header.dword ); + OUT_RELOC( index_buffer, PIPE_BUFFER_FLAG_READ, start); + OUT_RELOC( index_buffer, PIPE_BUFFER_FLAG_READ, start + count); + OUT_BATCH( 0 ); + ADVANCE_BATCH(); + } + return TRUE; +} diff --git a/src/mesa/pipe/i965simple/brw_eu.c b/src/mesa/pipe/i965simple/brw_eu.c new file mode 100644 index 0000000000..e2002d1821 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_eu.c @@ -0,0 +1,130 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" + + + +/* How does predicate control work when execution_size != 8? Do I + * need to test/set for 0xffff when execution_size is 16? + */ +void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value ) +{ + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + if (value != 0xff) { + if (value != p->flag_value) { + brw_push_insn_state(p); + brw_MOV(p, brw_flag_reg(), brw_imm_uw(value)); + p->flag_value = value; + brw_pop_insn_state(p); + } + + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + } +} + +void brw_set_predicate_control( struct brw_compile *p, unsigned pc ) +{ + p->current->header.predicate_control = pc; +} + +void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional ) +{ + p->current->header.destreg__conditonalmod = conditional; +} + +void brw_set_access_mode( struct brw_compile *p, unsigned access_mode ) +{ + p->current->header.access_mode = access_mode; +} + +void brw_set_compression_control( struct brw_compile *p, boolean compression_control ) +{ + p->current->header.compression_control = compression_control; +} + +void brw_set_mask_control( struct brw_compile *p, unsigned value ) +{ + p->current->header.mask_control = value; +} + +void brw_set_saturate( struct brw_compile *p, unsigned value ) +{ + p->current->header.saturate = value; +} + +void brw_push_insn_state( struct brw_compile *p ) +{ + assert(p->current != &p->stack[BRW_EU_MAX_INSN_STACK-1]); + memcpy(p->current+1, p->current, sizeof(struct brw_instruction)); + p->current++; +} + +void brw_pop_insn_state( struct brw_compile *p ) +{ + assert(p->current != p->stack); + p->current--; +} + + +/*********************************************************************** + */ +void brw_init_compile( struct brw_compile *p ) +{ + p->nr_insn = 0; + p->current = p->stack; + memset(p->current, 0, sizeof(p->current[0])); + + /* Some defaults? + */ + brw_set_mask_control(p, BRW_MASK_ENABLE); /* what does this do? */ + brw_set_saturate(p, 0); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_predicate_control_flag_value(p, 0xff); +} + + +const unsigned *brw_get_program( struct brw_compile *p, + unsigned *sz ) +{ + unsigned i; + + for (i = 0; i < 8; i++) + brw_NOP(p); + + *sz = p->nr_insn * sizeof(struct brw_instruction); + return (const unsigned *)p->store; +} + diff --git a/src/mesa/pipe/i965simple/brw_eu.h b/src/mesa/pipe/i965simple/brw_eu.h new file mode 100644 index 0000000000..111edb1506 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_eu.h @@ -0,0 +1,877 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_EU_H +#define BRW_EU_H + +#include "brw_structs.h" +#include "brw_defines.h" + +#include "pipe/p_compiler.h" +#include "pipe/p_shader_tokens.h" + +#define BRW_SWIZZLE4(a,b,c,d) (((a)<<0) | ((b)<<2) | ((c)<<4) | ((d)<<6)) +#define BRW_GET_SWZ(swz, idx) (((swz) >> ((idx)*2)) & 0x3) + +#define BRW_SWIZZLE_NOOP BRW_SWIZZLE4(0,1,2,3) +#define BRW_SWIZZLE_XYZW BRW_SWIZZLE4(0,1,2,3) +#define BRW_SWIZZLE_XXXX BRW_SWIZZLE4(0,0,0,0) +#define BRW_SWIZZLE_XYXY BRW_SWIZZLE4(0,1,0,1) + + +#define REG_SIZE (8*4) + + +/* These aren't hardware structs, just something useful for us to pass around: + * + * Align1 operation has a lot of control over input ranges. Used in + * WM programs to implement shaders decomposed into "channel serial" + * or "structure of array" form: + */ +struct brw_reg +{ + unsigned type:4; + unsigned file:2; + unsigned nr:8; + unsigned subnr:5; /* :1 in align16 */ + unsigned negate:1; /* source only */ + unsigned abs:1; /* source only */ + unsigned vstride:4; /* source only */ + unsigned width:3; /* src only, align1 only */ + unsigned hstride:2; /* src only, align1 only */ + unsigned address_mode:1; /* relative addressing, hopefully! */ + unsigned pad0:1; + + union { + struct { + unsigned swizzle:8; /* src only, align16 only */ + unsigned writemask:4; /* dest only, align16 only */ + int indirect_offset:10; /* relative addressing offset */ + unsigned pad1:10; /* two dwords total */ + } bits; + + float f; + int d; + unsigned ud; + } dw1; +}; + + +struct brw_indirect { + unsigned addr_subnr:4; + int addr_offset:10; + unsigned pad:18; +}; + + +#define BRW_EU_MAX_INSN_STACK 5 +#define BRW_EU_MAX_INSN 1200 + +struct brw_compile { + struct brw_instruction store[BRW_EU_MAX_INSN]; + unsigned nr_insn; + + /* Allow clients to push/pop instruction state: + */ + struct brw_instruction stack[BRW_EU_MAX_INSN_STACK]; + struct brw_instruction *current; + + unsigned flag_value; + boolean single_program_flow; +}; + + + +static __inline int type_sz( unsigned type ) +{ + switch( type ) { + case BRW_REGISTER_TYPE_UD: + case BRW_REGISTER_TYPE_D: + case BRW_REGISTER_TYPE_F: + return 4; + case BRW_REGISTER_TYPE_HF: + case BRW_REGISTER_TYPE_UW: + case BRW_REGISTER_TYPE_W: + return 2; + case BRW_REGISTER_TYPE_UB: + case BRW_REGISTER_TYPE_B: + return 1; + default: + return 0; + } +} + +static __inline struct brw_reg brw_reg( unsigned file, + unsigned nr, + unsigned subnr, + unsigned type, + unsigned vstride, + unsigned width, + unsigned hstride, + unsigned swizzle, + unsigned writemask) +{ + + struct brw_reg reg; + reg.type = type; + reg.file = file; + reg.nr = nr; + reg.subnr = subnr * type_sz(type); + reg.negate = 0; + reg.abs = 0; + reg.vstride = vstride; + reg.width = width; + reg.hstride = hstride; + reg.address_mode = BRW_ADDRESS_DIRECT; + reg.pad0 = 0; + + /* Could do better: If the reg is r5.3<0;1,0>, we probably want to + * set swizzle and writemask to W, as the lower bits of subnr will + * be lost when converted to align16. This is probably too much to + * keep track of as you'd want it adjusted by suboffset(), etc. + * Perhaps fix up when converting to align16? + */ + reg.dw1.bits.swizzle = swizzle; + reg.dw1.bits.writemask = writemask; + reg.dw1.bits.indirect_offset = 0; + reg.dw1.bits.pad1 = 0; + return reg; +} + +static __inline struct brw_reg brw_vec16_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_16, + BRW_WIDTH_16, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYZW, + TGSI_WRITEMASK_XYZW); +} + +static __inline struct brw_reg brw_vec8_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_8, + BRW_WIDTH_8, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYZW, + TGSI_WRITEMASK_XYZW); +} + + +static __inline struct brw_reg brw_vec4_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_4, + BRW_WIDTH_4, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYZW, + TGSI_WRITEMASK_XYZW); +} + + +static __inline struct brw_reg brw_vec2_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_2, + BRW_WIDTH_2, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYXY, + TGSI_WRITEMASK_XY); +} + +static __inline struct brw_reg brw_vec1_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_0, + BRW_WIDTH_1, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XXXX, + TGSI_WRITEMASK_X); +} + + +static __inline struct brw_reg retype( struct brw_reg reg, + unsigned type ) +{ + reg.type = type; + return reg; +} + +static __inline struct brw_reg suboffset( struct brw_reg reg, + unsigned delta ) +{ + reg.subnr += delta * type_sz(reg.type); + return reg; +} + + +static __inline struct brw_reg offset( struct brw_reg reg, + unsigned delta ) +{ + reg.nr += delta; + return reg; +} + + +static __inline struct brw_reg byte_offset( struct brw_reg reg, + unsigned bytes ) +{ + unsigned newoffset = reg.nr * REG_SIZE + reg.subnr + bytes; + reg.nr = newoffset / REG_SIZE; + reg.subnr = newoffset % REG_SIZE; + return reg; +} + + +static __inline struct brw_reg brw_uw16_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return suboffset(retype(brw_vec16_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); +} + +static __inline struct brw_reg brw_uw8_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return suboffset(retype(brw_vec8_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); +} + +static __inline struct brw_reg brw_uw1_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return suboffset(retype(brw_vec1_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); +} + +static __inline struct brw_reg brw_imm_reg( unsigned type ) +{ + return brw_reg( BRW_IMMEDIATE_VALUE, + 0, + 0, + type, + BRW_VERTICAL_STRIDE_0, + BRW_WIDTH_1, + BRW_HORIZONTAL_STRIDE_0, + 0, + 0); +} + +static __inline struct brw_reg brw_imm_f( float f ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_F); + imm.dw1.f = f; + return imm; +} + +static __inline struct brw_reg brw_imm_d( int d ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_D); + imm.dw1.d = d; + return imm; +} + +static __inline struct brw_reg brw_imm_ud( unsigned ud ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UD); + imm.dw1.ud = ud; + return imm; +} + +static __inline struct brw_reg brw_imm_uw( ushort uw ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UW); + imm.dw1.ud = uw; + return imm; +} + +static __inline struct brw_reg brw_imm_w( short w ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_W); + imm.dw1.d = w; + return imm; +} + +/* brw_imm_b and brw_imm_ub aren't supported by hardware - the type + * numbers alias with _V and _VF below: + */ + +/* Vector of eight signed half-byte values: + */ +static __inline struct brw_reg brw_imm_v( unsigned v ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_V); + imm.vstride = BRW_VERTICAL_STRIDE_0; + imm.width = BRW_WIDTH_8; + imm.hstride = BRW_HORIZONTAL_STRIDE_1; + imm.dw1.ud = v; + return imm; +} + +/* Vector of four 8-bit float values: + */ +static __inline struct brw_reg brw_imm_vf( unsigned v ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); + imm.vstride = BRW_VERTICAL_STRIDE_0; + imm.width = BRW_WIDTH_4; + imm.hstride = BRW_HORIZONTAL_STRIDE_1; + imm.dw1.ud = v; + return imm; +} + +#define VF_ZERO 0x0 +#define VF_ONE 0x30 +#define VF_NEG (1<<7) + +static __inline struct brw_reg brw_imm_vf4( unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); + imm.vstride = BRW_VERTICAL_STRIDE_0; + imm.width = BRW_WIDTH_4; + imm.hstride = BRW_HORIZONTAL_STRIDE_1; + imm.dw1.ud = ((v0 << 0) | + (v1 << 8) | + (v2 << 16) | + (v3 << 24)); + return imm; +} + + +static __inline struct brw_reg brw_address( struct brw_reg reg ) +{ + return brw_imm_uw(reg.nr * REG_SIZE + reg.subnr); +} + + +static __inline struct brw_reg brw_vec1_grf( unsigned nr, + unsigned subnr ) +{ + return brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +static __inline struct brw_reg brw_vec8_grf( unsigned nr, + unsigned subnr ) +{ + return brw_vec8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +static __inline struct brw_reg brw_vec4_grf( unsigned nr, + unsigned subnr ) +{ + return brw_vec4_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + + +static __inline struct brw_reg brw_vec2_grf( unsigned nr, + unsigned subnr ) +{ + return brw_vec2_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +static __inline struct brw_reg brw_uw8_grf( unsigned nr, + unsigned subnr ) +{ + return brw_uw8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +static __inline struct brw_reg brw_null_reg( void ) +{ + return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_NULL, + 0); +} + +static __inline struct brw_reg brw_address_reg( unsigned subnr ) +{ + return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_ADDRESS, + subnr); +} + +/* If/else instructions break in align16 mode if writemask & swizzle + * aren't xyzw. This goes against the convention for other scalar + * regs: + */ +static __inline struct brw_reg brw_ip_reg( void ) +{ + return brw_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_IP, + 0, + BRW_REGISTER_TYPE_UD, + BRW_VERTICAL_STRIDE_4, /* ? */ + BRW_WIDTH_1, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XYZW, /* NOTE! */ + TGSI_WRITEMASK_XYZW); /* NOTE! */ +} + +static __inline struct brw_reg brw_acc_reg( void ) +{ + return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_ACCUMULATOR, + 0); +} + + +static __inline struct brw_reg brw_flag_reg( void ) +{ + return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_FLAG, + 0); +} + + +static __inline struct brw_reg brw_mask_reg( unsigned subnr ) +{ + return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_MASK, + subnr); +} + +static __inline struct brw_reg brw_message_reg( unsigned nr ) +{ + return brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, + nr, + 0); +} + + + + +/* This is almost always called with a numeric constant argument, so + * make things easy to evaluate at compile time: + */ +static __inline unsigned cvt( unsigned val ) +{ + switch (val) { + case 0: return 0; + case 1: return 1; + case 2: return 2; + case 4: return 3; + case 8: return 4; + case 16: return 5; + case 32: return 6; + } + return 0; +} + +static __inline struct brw_reg stride( struct brw_reg reg, + unsigned vstride, + unsigned width, + unsigned hstride ) +{ + + reg.vstride = cvt(vstride); + reg.width = cvt(width) - 1; + reg.hstride = cvt(hstride); + return reg; +} + +static __inline struct brw_reg vec16( struct brw_reg reg ) +{ + return stride(reg, 16,16,1); +} + +static __inline struct brw_reg vec8( struct brw_reg reg ) +{ + return stride(reg, 8,8,1); +} + +static __inline struct brw_reg vec4( struct brw_reg reg ) +{ + return stride(reg, 4,4,1); +} + +static __inline struct brw_reg vec2( struct brw_reg reg ) +{ + return stride(reg, 2,2,1); +} + +static __inline struct brw_reg vec1( struct brw_reg reg ) +{ + return stride(reg, 0,1,0); +} + +static __inline struct brw_reg get_element( struct brw_reg reg, unsigned elt ) +{ + return vec1(suboffset(reg, elt)); +} + +static __inline struct brw_reg get_element_ud( struct brw_reg reg, unsigned elt ) +{ + return vec1(suboffset(retype(reg, BRW_REGISTER_TYPE_UD), elt)); +} + + +static __inline struct brw_reg brw_swizzle( struct brw_reg reg, + unsigned x, + unsigned y, + unsigned z, + unsigned w) +{ + reg.dw1.bits.swizzle = BRW_SWIZZLE4(BRW_GET_SWZ(reg.dw1.bits.swizzle, x), + BRW_GET_SWZ(reg.dw1.bits.swizzle, y), + BRW_GET_SWZ(reg.dw1.bits.swizzle, z), + BRW_GET_SWZ(reg.dw1.bits.swizzle, w)); + return reg; +} + + +static __inline struct brw_reg brw_swizzle1( struct brw_reg reg, + unsigned x ) +{ + return brw_swizzle(reg, x, x, x, x); +} + +static __inline struct brw_reg brw_writemask( struct brw_reg reg, + unsigned mask ) +{ + reg.dw1.bits.writemask &= mask; + return reg; +} + +static __inline struct brw_reg brw_set_writemask( struct brw_reg reg, + unsigned mask ) +{ + reg.dw1.bits.writemask = mask; + return reg; +} + +static __inline struct brw_reg negate( struct brw_reg reg ) +{ + reg.negate ^= 1; + return reg; +} + +static __inline struct brw_reg brw_abs( struct brw_reg reg ) +{ + reg.abs = 1; + return reg; +} + +/*********************************************************************** + */ +static __inline struct brw_reg brw_vec4_indirect( unsigned subnr, + int offset ) +{ + struct brw_reg reg = brw_vec4_grf(0, 0); + reg.subnr = subnr; + reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; + reg.dw1.bits.indirect_offset = offset; + return reg; +} + +static __inline struct brw_reg brw_vec1_indirect( unsigned subnr, + int offset ) +{ + struct brw_reg reg = brw_vec1_grf(0, 0); + reg.subnr = subnr; + reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; + reg.dw1.bits.indirect_offset = offset; + return reg; +} + +static __inline struct brw_reg deref_4f(struct brw_indirect ptr, int offset) +{ + return brw_vec4_indirect(ptr.addr_subnr, ptr.addr_offset + offset); +} + +static __inline struct brw_reg deref_1f(struct brw_indirect ptr, int offset) +{ + return brw_vec1_indirect(ptr.addr_subnr, ptr.addr_offset + offset); +} + +static __inline struct brw_reg deref_4b(struct brw_indirect ptr, int offset) +{ + return retype(deref_4f(ptr, offset), BRW_REGISTER_TYPE_B); +} + +static __inline struct brw_reg deref_1uw(struct brw_indirect ptr, int offset) +{ + return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UW); +} + +static __inline struct brw_reg deref_1ud(struct brw_indirect ptr, int offset) +{ + return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UD); +} + +static __inline struct brw_reg get_addr_reg(struct brw_indirect ptr) +{ + return brw_address_reg(ptr.addr_subnr); +} + +static __inline struct brw_indirect brw_indirect_offset( struct brw_indirect ptr, int offset ) +{ + ptr.addr_offset += offset; + return ptr; +} + +static __inline struct brw_indirect brw_indirect( unsigned addr_subnr, int offset ) +{ + struct brw_indirect ptr; + ptr.addr_subnr = addr_subnr; + ptr.addr_offset = offset; + ptr.pad = 0; + return ptr; +} + +static __inline struct brw_instruction *current_insn( struct brw_compile *p) +{ + return &p->store[p->nr_insn]; +} + +void brw_pop_insn_state( struct brw_compile *p ); +void brw_push_insn_state( struct brw_compile *p ); +void brw_set_mask_control( struct brw_compile *p, unsigned value ); +void brw_set_saturate( struct brw_compile *p, unsigned value ); +void brw_set_access_mode( struct brw_compile *p, unsigned access_mode ); +void brw_set_compression_control( struct brw_compile *p, boolean control ); +void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value ); +void brw_set_predicate_control( struct brw_compile *p, unsigned pc ); +void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional ); + +void brw_init_compile( struct brw_compile *p ); +const unsigned *brw_get_program( struct brw_compile *p, unsigned *sz ); + + +/* Helpers for regular instructions: + */ +#define ALU1(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0); + +#define ALU2(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0, \ + struct brw_reg src1); + +ALU1(MOV) +ALU2(SEL) +ALU1(NOT) +ALU2(AND) +ALU2(OR) +ALU2(XOR) +ALU2(SHR) +ALU2(SHL) +ALU2(RSR) +ALU2(RSL) +ALU2(ASR) +ALU2(JMPI) +ALU2(ADD) +ALU2(MUL) +ALU1(FRC) +ALU1(RNDD) +ALU2(MAC) +ALU2(MACH) +ALU1(LZD) +ALU2(DP4) +ALU2(DPH) +ALU2(DP3) +ALU2(DP2) +ALU2(LINE) + +#undef ALU1 +#undef ALU2 + + + +/* Helpers for SEND instruction: + */ +void brw_urb_WRITE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + boolean allocate, + boolean used, + unsigned msg_length, + unsigned response_length, + boolean eot, + boolean writes_complete, + unsigned offset, + unsigned swizzle); + +void brw_fb_WRITE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + unsigned binding_table_index, + unsigned msg_length, + unsigned response_length, + boolean eot); + +void brw_SAMPLE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + unsigned binding_table_index, + unsigned sampler, + unsigned writemask, + unsigned msg_type, + unsigned response_length, + unsigned msg_length, + boolean eot); + +void brw_math_16( struct brw_compile *p, + struct brw_reg dest, + unsigned function, + unsigned saturate, + unsigned msg_reg_nr, + struct brw_reg src, + unsigned precision ); + +void brw_math( struct brw_compile *p, + struct brw_reg dest, + unsigned function, + unsigned saturate, + unsigned msg_reg_nr, + struct brw_reg src, + unsigned data_type, + unsigned precision ); + +void brw_dp_READ_16( struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + unsigned scratch_offset ); + +void brw_dp_WRITE_16( struct brw_compile *p, + struct brw_reg src, + unsigned msg_reg_nr, + unsigned scratch_offset ); + +/* If/else/endif. Works by manipulating the execution flags on each + * channel. + */ +struct brw_instruction *brw_IF(struct brw_compile *p, + unsigned execute_size); + +struct brw_instruction *brw_ELSE(struct brw_compile *p, + struct brw_instruction *if_insn); + +void brw_ENDIF(struct brw_compile *p, + struct brw_instruction *if_or_else_insn); + + +/* DO/WHILE loops: + */ +struct brw_instruction *brw_DO(struct brw_compile *p, + unsigned execute_size); + +struct brw_instruction *brw_WHILE(struct brw_compile *p, + struct brw_instruction *patch_insn); + +struct brw_instruction *brw_BREAK(struct brw_compile *p); +struct brw_instruction *brw_CONT(struct brw_compile *p); +/* Forward jumps: + */ +void brw_land_fwd_jump(struct brw_compile *p, + struct brw_instruction *jmp_insn); + + + +void brw_NOP(struct brw_compile *p); + +/* Special case: there is never a destination, execution size will be + * taken from src0: + */ +void brw_CMP(struct brw_compile *p, + struct brw_reg dest, + unsigned conditional, + struct brw_reg src0, + struct brw_reg src1); + +void brw_print_reg( struct brw_reg reg ); + + +/*********************************************************************** + * brw_eu_util.c: + */ + +void brw_copy_indirect_to_indirect(struct brw_compile *p, + struct brw_indirect dst_ptr, + struct brw_indirect src_ptr, + unsigned count); + +void brw_copy_from_indirect(struct brw_compile *p, + struct brw_reg dst, + struct brw_indirect ptr, + unsigned count); + +void brw_copy4(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + unsigned count); + +void brw_copy8(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + unsigned count); + +void brw_math_invert( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src); + +void brw_set_src1( struct brw_instruction *insn, + struct brw_reg reg ); +#endif diff --git a/src/mesa/pipe/i965simple/brw_eu_debug.c b/src/mesa/pipe/i965simple/brw_eu_debug.c new file mode 100644 index 0000000000..be692f6502 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_eu_debug.c @@ -0,0 +1,88 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_eu.h" + +void brw_print_reg( struct brw_reg hwreg ) +{ + static const char *file[] = { + "arf", + "grf", + "msg", + "imm" + }; + + static const char *type[] = { + "ud", + "d", + "uw", + "w", + "ub", + "vf", + "hf", + "f" + }; + + _mesa_printf("%s%s", + hwreg.abs ? "abs/" : "", + hwreg.negate ? "-" : ""); + + if (hwreg.file == BRW_GENERAL_REGISTER_FILE && + hwreg.nr % 2 == 0 && + hwreg.subnr == 0 && + hwreg.vstride == BRW_VERTICAL_STRIDE_8 && + hwreg.width == BRW_WIDTH_8 && + hwreg.hstride == BRW_HORIZONTAL_STRIDE_1 && + hwreg.type == BRW_REGISTER_TYPE_F) { + _mesa_printf("vec%d", hwreg.nr); + } + else if (hwreg.file == BRW_GENERAL_REGISTER_FILE && + hwreg.vstride == BRW_VERTICAL_STRIDE_0 && + hwreg.width == BRW_WIDTH_1 && + hwreg.hstride == BRW_HORIZONTAL_STRIDE_0 && + hwreg.type == BRW_REGISTER_TYPE_F) { + _mesa_printf("scl%d.%d", hwreg.nr, hwreg.subnr / 4); + } + else { + _mesa_printf("%s%d.%d<%d;%d,%d>:%s", + file[hwreg.file], + hwreg.nr, + hwreg.subnr / type_sz(hwreg.type), + hwreg.vstride ? (1<<(hwreg.vstride-1)) : 0, + 1< + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" + + + + +/*********************************************************************** + * Internal helper for constructing instructions + */ + +static void guess_execution_size( struct brw_instruction *insn, + struct brw_reg reg ) +{ + if (reg.width == BRW_WIDTH_8 && + insn->header.compression_control == BRW_COMPRESSION_COMPRESSED) + insn->header.execution_size = BRW_EXECUTE_16; + else + insn->header.execution_size = reg.width; /* note - definitions are compatible */ +} + + +static void brw_set_dest( struct brw_instruction *insn, + struct brw_reg dest ) +{ + insn->bits1.da1.dest_reg_file = dest.file; + insn->bits1.da1.dest_reg_type = dest.type; + insn->bits1.da1.dest_address_mode = dest.address_mode; + + if (dest.address_mode == BRW_ADDRESS_DIRECT) { + insn->bits1.da1.dest_reg_nr = dest.nr; + + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits1.da1.dest_subreg_nr = dest.subnr; + insn->bits1.da1.dest_horiz_stride = BRW_HORIZONTAL_STRIDE_1; + } + else { + insn->bits1.da16.dest_subreg_nr = dest.subnr / 16; + insn->bits1.da16.dest_writemask = dest.dw1.bits.writemask; + } + } + else { + insn->bits1.ia1.dest_subreg_nr = dest.subnr; + + /* These are different sizes in align1 vs align16: + */ + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits1.ia1.dest_indirect_offset = dest.dw1.bits.indirect_offset; + insn->bits1.ia1.dest_horiz_stride = BRW_HORIZONTAL_STRIDE_1; + } + else { + insn->bits1.ia16.dest_indirect_offset = dest.dw1.bits.indirect_offset; + } + } + + /* NEW: Set the execution size based on dest.width and + * insn->compression_control: + */ + guess_execution_size(insn, dest); +} + +static void brw_set_src0( struct brw_instruction *insn, + struct brw_reg reg ) +{ + assert(reg.file != BRW_MESSAGE_REGISTER_FILE); + + insn->bits1.da1.src0_reg_file = reg.file; + insn->bits1.da1.src0_reg_type = reg.type; + insn->bits2.da1.src0_abs = reg.abs; + insn->bits2.da1.src0_negate = reg.negate; + insn->bits2.da1.src0_address_mode = reg.address_mode; + + if (reg.file == BRW_IMMEDIATE_VALUE) { + insn->bits3.ud = reg.dw1.ud; + + /* Required to set some fields in src1 as well: + */ + insn->bits1.da1.src1_reg_file = 0; /* arf */ + insn->bits1.da1.src1_reg_type = reg.type; + } + else + { + if (reg.address_mode == BRW_ADDRESS_DIRECT) { + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits2.da1.src0_subreg_nr = reg.subnr; + insn->bits2.da1.src0_reg_nr = reg.nr; + } + else { + insn->bits2.da16.src0_subreg_nr = reg.subnr / 16; + insn->bits2.da16.src0_reg_nr = reg.nr; + } + } + else { + insn->bits2.ia1.src0_subreg_nr = reg.subnr; + + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits2.ia1.src0_indirect_offset = reg.dw1.bits.indirect_offset; + } + else { + insn->bits2.ia16.src0_subreg_nr = reg.dw1.bits.indirect_offset; + } + } + + if (insn->header.access_mode == BRW_ALIGN_1) { + if (reg.width == BRW_WIDTH_1 && + insn->header.execution_size == BRW_EXECUTE_1) { + insn->bits2.da1.src0_horiz_stride = BRW_HORIZONTAL_STRIDE_0; + insn->bits2.da1.src0_width = BRW_WIDTH_1; + insn->bits2.da1.src0_vert_stride = BRW_VERTICAL_STRIDE_0; + } + else { + insn->bits2.da1.src0_horiz_stride = reg.hstride; + insn->bits2.da1.src0_width = reg.width; + insn->bits2.da1.src0_vert_stride = reg.vstride; + } + } + else { + insn->bits2.da16.src0_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); + insn->bits2.da16.src0_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); + insn->bits2.da16.src0_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); + insn->bits2.da16.src0_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); + + /* This is an oddity of the fact we're using the same + * descriptions for registers in align_16 as align_1: + */ + if (reg.vstride == BRW_VERTICAL_STRIDE_8) + insn->bits2.da16.src0_vert_stride = BRW_VERTICAL_STRIDE_4; + else + insn->bits2.da16.src0_vert_stride = reg.vstride; + } + } +} + + +void brw_set_src1( struct brw_instruction *insn, + struct brw_reg reg ) +{ + assert(reg.file != BRW_MESSAGE_REGISTER_FILE); + + insn->bits1.da1.src1_reg_file = reg.file; + insn->bits1.da1.src1_reg_type = reg.type; + insn->bits3.da1.src1_abs = reg.abs; + insn->bits3.da1.src1_negate = reg.negate; + + /* Only src1 can be immediate in two-argument instructions. + */ + assert(insn->bits1.da1.src0_reg_file != BRW_IMMEDIATE_VALUE); + + if (reg.file == BRW_IMMEDIATE_VALUE) { + insn->bits3.ud = reg.dw1.ud; + } + else { + /* This is a hardware restriction, which may or may not be lifted + * in the future: + */ + assert (reg.address_mode == BRW_ADDRESS_DIRECT); + //assert (reg.file == BRW_GENERAL_REGISTER_FILE); + + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits3.da1.src1_subreg_nr = reg.subnr; + insn->bits3.da1.src1_reg_nr = reg.nr; + } + else { + insn->bits3.da16.src1_subreg_nr = reg.subnr / 16; + insn->bits3.da16.src1_reg_nr = reg.nr; + } + + if (insn->header.access_mode == BRW_ALIGN_1) { + if (reg.width == BRW_WIDTH_1 && + insn->header.execution_size == BRW_EXECUTE_1) { + insn->bits3.da1.src1_horiz_stride = BRW_HORIZONTAL_STRIDE_0; + insn->bits3.da1.src1_width = BRW_WIDTH_1; + insn->bits3.da1.src1_vert_stride = BRW_VERTICAL_STRIDE_0; + } + else { + insn->bits3.da1.src1_horiz_stride = reg.hstride; + insn->bits3.da1.src1_width = reg.width; + insn->bits3.da1.src1_vert_stride = reg.vstride; + } + } + else { + insn->bits3.da16.src1_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); + insn->bits3.da16.src1_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); + insn->bits3.da16.src1_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); + insn->bits3.da16.src1_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); + + /* This is an oddity of the fact we're using the same + * descriptions for registers in align_16 as align_1: + */ + if (reg.vstride == BRW_VERTICAL_STRIDE_8) + insn->bits3.da16.src1_vert_stride = BRW_VERTICAL_STRIDE_4; + else + insn->bits3.da16.src1_vert_stride = reg.vstride; + } + } +} + + + +static void brw_set_math_message( struct brw_instruction *insn, + unsigned msg_length, + unsigned response_length, + unsigned function, + unsigned integer_type, + boolean low_precision, + boolean saturate, + unsigned dataType ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.math.function = function; + insn->bits3.math.int_type = integer_type; + insn->bits3.math.precision = low_precision; + insn->bits3.math.saturate = saturate; + insn->bits3.math.data_type = dataType; + insn->bits3.math.response_length = response_length; + insn->bits3.math.msg_length = msg_length; + insn->bits3.math.msg_target = BRW_MESSAGE_TARGET_MATH; + insn->bits3.math.end_of_thread = 0; +} + +static void brw_set_urb_message( struct brw_instruction *insn, + boolean allocate, + boolean used, + unsigned msg_length, + unsigned response_length, + boolean end_of_thread, + boolean complete, + unsigned offset, + unsigned swizzle_control ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.urb.opcode = 0; /* ? */ + insn->bits3.urb.offset = offset; + insn->bits3.urb.swizzle_control = swizzle_control; + insn->bits3.urb.allocate = allocate; + insn->bits3.urb.used = used; /* ? */ + insn->bits3.urb.complete = complete; + insn->bits3.urb.response_length = response_length; + insn->bits3.urb.msg_length = msg_length; + insn->bits3.urb.msg_target = BRW_MESSAGE_TARGET_URB; + insn->bits3.urb.end_of_thread = end_of_thread; +} + +static void brw_set_dp_write_message( struct brw_instruction *insn, + unsigned binding_table_index, + unsigned msg_control, + unsigned msg_type, + unsigned msg_length, + unsigned pixel_scoreboard_clear, + unsigned response_length, + unsigned end_of_thread ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.dp_write.binding_table_index = binding_table_index; + insn->bits3.dp_write.msg_control = msg_control; + insn->bits3.dp_write.pixel_scoreboard_clear = pixel_scoreboard_clear; + insn->bits3.dp_write.msg_type = msg_type; + insn->bits3.dp_write.send_commit_msg = 0; + insn->bits3.dp_write.response_length = response_length; + insn->bits3.dp_write.msg_length = msg_length; + insn->bits3.dp_write.msg_target = BRW_MESSAGE_TARGET_DATAPORT_WRITE; + insn->bits3.urb.end_of_thread = end_of_thread; +} + +static void brw_set_dp_read_message( struct brw_instruction *insn, + unsigned binding_table_index, + unsigned msg_control, + unsigned msg_type, + unsigned target_cache, + unsigned msg_length, + unsigned response_length, + unsigned end_of_thread ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.dp_read.binding_table_index = binding_table_index; + insn->bits3.dp_read.msg_control = msg_control; + insn->bits3.dp_read.msg_type = msg_type; + insn->bits3.dp_read.target_cache = target_cache; + insn->bits3.dp_read.response_length = response_length; + insn->bits3.dp_read.msg_length = msg_length; + insn->bits3.dp_read.msg_target = BRW_MESSAGE_TARGET_DATAPORT_READ; + insn->bits3.dp_read.end_of_thread = end_of_thread; +} + +static void brw_set_sampler_message( struct brw_instruction *insn, + unsigned binding_table_index, + unsigned sampler, + unsigned msg_type, + unsigned response_length, + unsigned msg_length, + boolean eot) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.sampler.binding_table_index = binding_table_index; + insn->bits3.sampler.sampler = sampler; + insn->bits3.sampler.msg_type = msg_type; + insn->bits3.sampler.return_format = BRW_SAMPLER_RETURN_FORMAT_FLOAT32; + insn->bits3.sampler.response_length = response_length; + insn->bits3.sampler.msg_length = msg_length; + insn->bits3.sampler.end_of_thread = eot; + insn->bits3.sampler.msg_target = BRW_MESSAGE_TARGET_SAMPLER; +} + + + +static struct brw_instruction *next_insn( struct brw_compile *p, + unsigned opcode ) +{ + struct brw_instruction *insn; + + assert(p->nr_insn + 1 < BRW_EU_MAX_INSN); + + insn = &p->store[p->nr_insn++]; + memcpy(insn, p->current, sizeof(*insn)); + + /* Reset this one-shot flag: + */ + + if (p->current->header.destreg__conditonalmod) { + p->current->header.destreg__conditonalmod = 0; + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + } + + insn->header.opcode = opcode; + return insn; +} + + +static struct brw_instruction *brw_alu1( struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src ) +{ + struct brw_instruction *insn = next_insn(p, opcode); + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + return insn; +} + +static struct brw_instruction *brw_alu2(struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1 ) +{ + struct brw_instruction *insn = next_insn(p, opcode); + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, src1); + return insn; +} + + +/*********************************************************************** + * Convenience routines. + */ +#define ALU1(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0) \ +{ \ + return brw_alu1(p, BRW_OPCODE_##OP, dest, src0); \ +} + +#define ALU2(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0, \ + struct brw_reg src1) \ +{ \ + return brw_alu2(p, BRW_OPCODE_##OP, dest, src0, src1); \ +} + + +ALU1(MOV) +ALU2(SEL) +ALU1(NOT) +ALU2(AND) +ALU2(OR) +ALU2(XOR) +ALU2(SHR) +ALU2(SHL) +ALU2(RSR) +ALU2(RSL) +ALU2(ASR) +ALU2(ADD) +ALU2(MUL) +ALU1(FRC) +ALU1(RNDD) +ALU2(MAC) +ALU2(MACH) +ALU1(LZD) +ALU2(DP4) +ALU2(DPH) +ALU2(DP3) +ALU2(DP2) +ALU2(LINE) + + + + +void brw_NOP(struct brw_compile *p) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_NOP); + brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src1(insn, brw_imm_ud(0x0)); +} + + + + + +/*********************************************************************** + * Comparisons, if/else/endif + */ + +struct brw_instruction *brw_JMPI(struct brw_compile *p, + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1) +{ + struct brw_instruction *insn = brw_alu2(p, BRW_OPCODE_JMPI, dest, src0, src1); + + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + return insn; +} + +/* EU takes the value from the flag register and pushes it onto some + * sort of a stack (presumably merging with any flag value already on + * the stack). Within an if block, the flags at the top of the stack + * control execution on each channel of the unit, eg. on each of the + * 16 pixel values in our wm programs. + * + * When the matching 'else' instruction is reached (presumably by + * countdown of the instruction count patched in by our ELSE/ENDIF + * functions), the relevent flags are inverted. + * + * When the matching 'endif' instruction is reached, the flags are + * popped off. If the stack is now empty, normal execution resumes. + * + * No attempt is made to deal with stack overflow (14 elements?). + */ +struct brw_instruction *brw_IF(struct brw_compile *p, unsigned execute_size) +{ + struct brw_instruction *insn; + + if (p->single_program_flow) { + assert(execute_size == BRW_EXECUTE_1); + + insn = next_insn(p, BRW_OPCODE_ADD); + insn->header.predicate_inverse = 1; + } else { + insn = next_insn(p, BRW_OPCODE_IF); + } + + /* Override the defaults for this instruction: + */ + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.execution_size = execute_size; + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.predicate_control = BRW_PREDICATE_NORMAL; + insn->header.mask_control = BRW_MASK_ENABLE; + + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + return insn; +} + + +struct brw_instruction *brw_ELSE(struct brw_compile *p, + struct brw_instruction *if_insn) +{ + struct brw_instruction *insn; + + if (p->single_program_flow) { + insn = next_insn(p, BRW_OPCODE_ADD); + } else { + insn = next_insn(p, BRW_OPCODE_ELSE); + } + + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = if_insn->header.execution_size; + insn->header.mask_control = BRW_MASK_ENABLE; + + /* Patch the if instruction to point at this instruction. + */ + if (p->single_program_flow) { + assert(if_insn->header.opcode == BRW_OPCODE_ADD); + + if_insn->bits3.ud = (insn - if_insn + 1) * 16; + } else { + assert(if_insn->header.opcode == BRW_OPCODE_IF); + + if_insn->bits3.if_else.jump_count = insn - if_insn; + if_insn->bits3.if_else.pop_count = 1; + if_insn->bits3.if_else.pad0 = 0; + } + + return insn; +} + +void brw_ENDIF(struct brw_compile *p, + struct brw_instruction *patch_insn) +{ + if (p->single_program_flow) { + /* In single program flow mode, there's no need to execute an ENDIF, + * since we don't need to do any stack operations, and if we're executing + * currently, we want to just continue executing. + */ + struct brw_instruction *next = &p->store[p->nr_insn]; + + assert(patch_insn->header.opcode == BRW_OPCODE_ADD); + + patch_insn->bits3.ud = (next - patch_insn) * 16; + } else { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_ENDIF); + + brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = patch_insn->header.execution_size; + insn->header.mask_control = BRW_MASK_ENABLE; + + assert(patch_insn->bits3.if_else.jump_count == 0); + + /* Patch the if or else instructions to point at this or the next + * instruction respectively. + */ + if (patch_insn->header.opcode == BRW_OPCODE_IF) { + /* Automagically turn it into an IFF: + */ + patch_insn->header.opcode = BRW_OPCODE_IFF; + patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1; + patch_insn->bits3.if_else.pop_count = 0; + patch_insn->bits3.if_else.pad0 = 0; + } else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) { + patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1; + patch_insn->bits3.if_else.pop_count = 1; + patch_insn->bits3.if_else.pad0 = 0; + } else { + assert(0); + } + + /* Also pop item off the stack in the endif instruction: + */ + insn->bits3.if_else.jump_count = 0; + insn->bits3.if_else.pop_count = 1; + insn->bits3.if_else.pad0 = 0; + } +} + +struct brw_instruction *brw_BREAK(struct brw_compile *p) +{ + struct brw_instruction *insn; + insn = next_insn(p, BRW_OPCODE_BREAK); + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = BRW_EXECUTE_8; + insn->header.mask_control = BRW_MASK_DISABLE; + insn->bits3.if_else.pad0 = 0; + return insn; +} + +struct brw_instruction *brw_CONT(struct brw_compile *p) +{ + struct brw_instruction *insn; + insn = next_insn(p, BRW_OPCODE_CONTINUE); + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = BRW_EXECUTE_8; + insn->header.mask_control = BRW_MASK_DISABLE; + insn->bits3.if_else.pad0 = 0; + return insn; +} + +/* DO/WHILE loop: + */ +struct brw_instruction *brw_DO(struct brw_compile *p, unsigned execute_size) +{ + if (p->single_program_flow) { + return &p->store[p->nr_insn]; + } else { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_DO); + + /* Override the defaults for this instruction: + */ + brw_set_dest(insn, brw_null_reg()); + brw_set_src0(insn, brw_null_reg()); + brw_set_src1(insn, brw_null_reg()); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = execute_size; + insn->header.predicate_control = BRW_PREDICATE_NONE; + /* insn->header.mask_control = BRW_MASK_ENABLE; */ + insn->header.mask_control = BRW_MASK_DISABLE; + + return insn; + } +} + + + +struct brw_instruction *brw_WHILE(struct brw_compile *p, + struct brw_instruction *do_insn) +{ + struct brw_instruction *insn; + + if (p->single_program_flow) + insn = next_insn(p, BRW_OPCODE_ADD); + else + insn = next_insn(p, BRW_OPCODE_WHILE); + + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + + if (p->single_program_flow) { + insn->header.execution_size = BRW_EXECUTE_1; + + insn->bits3.d = (do_insn - insn) * 16; + } else { + insn->header.execution_size = do_insn->header.execution_size; + + assert(do_insn->header.opcode == BRW_OPCODE_DO); + insn->bits3.if_else.jump_count = do_insn - insn; + insn->bits3.if_else.pop_count = 0; + insn->bits3.if_else.pad0 = 0; + } + +/* insn->header.mask_control = BRW_MASK_ENABLE; */ + + insn->header.mask_control = BRW_MASK_DISABLE; + p->current->header.predicate_control = BRW_PREDICATE_NONE; + return insn; +} + + +/* FORWARD JUMPS: + */ +void brw_land_fwd_jump(struct brw_compile *p, + struct brw_instruction *jmp_insn) +{ + struct brw_instruction *landing = &p->store[p->nr_insn]; + + assert(jmp_insn->header.opcode == BRW_OPCODE_JMPI); + assert(jmp_insn->bits1.da1.src1_reg_file = BRW_IMMEDIATE_VALUE); + + jmp_insn->bits3.ud = (landing - jmp_insn) - 1; +} + + + +/* To integrate with the above, it makes sense that the comparison + * instruction should populate the flag register. It might be simpler + * just to use the flag reg for most WM tasks? + */ +void brw_CMP(struct brw_compile *p, + struct brw_reg dest, + unsigned conditional, + struct brw_reg src0, + struct brw_reg src1) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_CMP); + + insn->header.destreg__conditonalmod = conditional; + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, src1); + +/* guess_execution_size(insn, src0); */ + + + /* Make it so that future instructions will use the computed flag + * value until brw_set_predicate_control_flag_value() is called + * again. + */ + if (dest.file == BRW_ARCHITECTURE_REGISTER_FILE && + dest.nr == 0) { + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + p->flag_value = 0xff; + } +} + + + +/*********************************************************************** + * Helpers for the various SEND message types: + */ + +/* Invert 8 values + */ +void brw_math( struct brw_compile *p, + struct brw_reg dest, + unsigned function, + unsigned saturate, + unsigned msg_reg_nr, + struct brw_reg src, + unsigned data_type, + unsigned precision ) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + unsigned msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1; + unsigned response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1; + + /* Example code doesn't set predicate_control for send + * instructions. + */ + insn->header.predicate_control = 0; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + brw_set_math_message(insn, + msg_length, response_length, + function, + BRW_MATH_INTEGER_UNSIGNED, + precision, + saturate, + data_type); +} + +/* Use 2 send instructions to invert 16 elements + */ +void brw_math_16( struct brw_compile *p, + struct brw_reg dest, + unsigned function, + unsigned saturate, + unsigned msg_reg_nr, + struct brw_reg src, + unsigned precision ) +{ + struct brw_instruction *insn; + unsigned msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1; + unsigned response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1; + + /* First instruction: + */ + brw_push_insn_state(p); + brw_set_predicate_control_flag_value(p, 0xff); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + insn = next_insn(p, BRW_OPCODE_SEND); + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + brw_set_math_message(insn, + msg_length, response_length, + function, + BRW_MATH_INTEGER_UNSIGNED, + precision, + saturate, + BRW_MATH_DATA_VECTOR); + + /* Second instruction: + */ + insn = next_insn(p, BRW_OPCODE_SEND); + insn->header.compression_control = BRW_COMPRESSION_2NDHALF; + insn->header.destreg__conditonalmod = msg_reg_nr+1; + + brw_set_dest(insn, offset(dest,1)); + brw_set_src0(insn, src); + brw_set_math_message(insn, + msg_length, response_length, + function, + BRW_MATH_INTEGER_UNSIGNED, + precision, + saturate, + BRW_MATH_DATA_VECTOR); + + brw_pop_insn_state(p); +} + + + + +void brw_dp_WRITE_16( struct brw_compile *p, + struct brw_reg src, + unsigned msg_reg_nr, + unsigned scratch_offset ) +{ + { + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + brw_MOV(p, + retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D), + brw_imm_d(scratch_offset)); + + brw_pop_insn_state(p); + } + + { + unsigned msg_length = 3; + struct brw_reg dest = retype(brw_null_reg(), BRW_REGISTER_TYPE_UW); + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + + brw_set_dp_write_message(insn, + 255, /* bti */ + BRW_DATAPORT_OWORD_BLOCK_4_OWORDS, /* msg_control */ + BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE, /* msg_type */ + msg_length, + 0, /* pixel scoreboard */ + 0, /* response_length */ + 0); /* eot */ + } + +} + + +void brw_dp_READ_16( struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + unsigned scratch_offset ) +{ + { + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_mask_control(p, BRW_MASK_DISABLE); + + brw_MOV(p, + retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D), + brw_imm_d(scratch_offset)); + + brw_pop_insn_state(p); + } + + { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); /* UW? */ + brw_set_src0(insn, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW)); + + brw_set_dp_read_message(insn, + 255, /* bti */ + 3, /* msg_control */ + BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ + 1, /* target cache */ + 1, /* msg_length */ + 2, /* response_length */ + 0); /* eot */ + } +} + + +void brw_fb_WRITE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + unsigned binding_table_index, + unsigned msg_length, + unsigned response_length, + boolean eot) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_dp_write_message(insn, + binding_table_index, + BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE, /* msg_control */ + BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE, /* msg_type */ + msg_length, + 1, /* pixel scoreboard */ + response_length, + eot); +} + + + +void brw_SAMPLE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + unsigned binding_table_index, + unsigned sampler, + unsigned writemask, + unsigned msg_type, + unsigned response_length, + unsigned msg_length, + boolean eot) +{ + boolean need_stall = 0; + + if(writemask == 0) { +/* _mesa_printf("%s: zero writemask??\n", __FUNCTION__); */ + return; + } + + /* Hardware doesn't do destination dependency checking on send + * instructions properly. Add a workaround which generates the + * dependency by other means. In practice it seems like this bug + * only crops up for texture samples, and only where registers are + * written by the send and then written again later without being + * read in between. Luckily for us, we already track that + * information and use it to modify the writemask for the + * instruction, so that is a guide for whether a workaround is + * needed. + */ + if (writemask != TGSI_WRITEMASK_XYZW) { + unsigned dst_offset = 0; + unsigned i, newmask = 0, len = 0; + + for (i = 0; i < 4; i++) { + if (writemask & (1<header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_sampler_message(insn, + binding_table_index, + sampler, + msg_type, + response_length, + msg_length, + eot); + } + + if (need_stall) + { + struct brw_reg reg = vec8(offset(dest, response_length-1)); + + /* mov (8) r9.0<1>:f r9.0<8;8,1>:f { Align1 } + */ + brw_push_insn_state(p); + brw_set_compression_control(p, FALSE); + brw_MOV(p, reg, reg); + brw_pop_insn_state(p); + } + +} + +/* All these variables are pretty confusing - we might be better off + * using bitmasks and macros for this, in the old style. Or perhaps + * just having the caller instantiate the fields in dword3 itself. + */ +void brw_urb_WRITE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + boolean allocate, + boolean used, + unsigned msg_length, + unsigned response_length, + boolean eot, + boolean writes_complete, + unsigned offset, + unsigned swizzle) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + assert(msg_length < 16); + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, brw_imm_d(0)); + + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_urb_message(insn, + allocate, + used, + msg_length, + response_length, + eot, + writes_complete, + offset, + swizzle); +} + diff --git a/src/mesa/pipe/i965simple/brw_eu_util.c b/src/mesa/pipe/i965simple/brw_eu_util.c new file mode 100644 index 0000000000..3a65b141f0 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_eu_util.c @@ -0,0 +1,126 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" + + +void brw_math_invert( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src) +{ + brw_math( p, + dst, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 0, + src, + BRW_MATH_PRECISION_FULL, + BRW_MATH_DATA_VECTOR ); +} + + + +void brw_copy4(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + unsigned count) +{ + unsigned i; + + dst = vec4(dst); + src = vec4(src); + + for (i = 0; i < count; i++) + { + unsigned delta = i*32; + brw_MOV(p, byte_offset(dst, delta), byte_offset(src, delta)); + brw_MOV(p, byte_offset(dst, delta+16), byte_offset(src, delta+16)); + } +} + + +void brw_copy8(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + unsigned count) +{ + unsigned i; + + dst = vec8(dst); + src = vec8(src); + + for (i = 0; i < count; i++) + { + unsigned delta = i*32; + brw_MOV(p, byte_offset(dst, delta), byte_offset(src, delta)); + } +} + + +void brw_copy_indirect_to_indirect(struct brw_compile *p, + struct brw_indirect dst_ptr, + struct brw_indirect src_ptr, + unsigned count) +{ + unsigned i; + + for (i = 0; i < count; i++) + { + unsigned delta = i*32; + brw_MOV(p, deref_4f(dst_ptr, delta), deref_4f(src_ptr, delta)); + brw_MOV(p, deref_4f(dst_ptr, delta+16), deref_4f(src_ptr, delta+16)); + } +} + + +void brw_copy_from_indirect(struct brw_compile *p, + struct brw_reg dst, + struct brw_indirect ptr, + unsigned count) +{ + unsigned i; + + dst = vec4(dst); + + for (i = 0; i < count; i++) + { + unsigned delta = i*32; + brw_MOV(p, byte_offset(dst, delta), deref_4f(ptr, delta)); + brw_MOV(p, byte_offset(dst, delta+16), deref_4f(ptr, delta+16)); + } +} + + + + diff --git a/src/mesa/pipe/i965simple/brw_flush.c b/src/mesa/pipe/i965simple/brw_flush.c new file mode 100644 index 0000000000..1f11c5f164 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_flush.c @@ -0,0 +1,84 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_batch.h" + + +/** + * In future we may want a fence-like interface instead of finish. + */ +static void brw_flush( struct pipe_context *pipe, + unsigned flags ) +{ + struct brw_context *brw = brw_context(pipe); + struct pipe_fence_handle *fence; + + /* Do we need to emit an MI_FLUSH command to flush the hardware + * caches? + */ + if (flags & (PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE)) { + unsigned flush = CMD_MI_FLUSH; + +#if 0 + if (!(flags & PIPE_FLUSH_RENDER_CACHE)) + flush |= INHIBIT_FLUSH_RENDER_CACHE; + + if (flags & PIPE_FLUSH_TEXTURE_CACHE) + flush |= FLUSH_MAP_CACHE; +#endif + + if (!BEGIN_BATCH(1, 0)) { + FLUSH_BATCH( &fence ); + assert(BEGIN_BATCH(1, 0)); + } + OUT_BATCH( flush ); + ADVANCE_BATCH(); + } + + /* If there are no flags, just flush pending commands to hardware: + */ + FLUSH_BATCH( &fence ); + + if (flags & PIPE_FLUSH_WAIT) { +// brw->winsys->wait_fence(brw->winsys, fence); + } +} + + + +void brw_init_flush_functions( struct brw_context *brw ) +{ + brw->pipe.flush = brw_flush; +} diff --git a/src/mesa/pipe/i965simple/brw_gs.c b/src/mesa/pipe/i965simple/brw_gs.c new file mode 100644 index 0000000000..de60868ccc --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_gs.c @@ -0,0 +1,196 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_state.h" +#include "brw_gs.h" + + + +static void compile_gs_prog( struct brw_context *brw, + struct brw_gs_prog_key *key ) +{ + struct brw_gs_compile c; + const unsigned *program; + unsigned program_size; + + memset(&c, 0, sizeof(c)); + + c.key = *key; + + /* Need to locate the two positions present in vertex + header. + * These are currently hardcoded: + */ + c.nr_attrs = brw_count_bits(c.key.attrs); + c.nr_regs = (c.nr_attrs + 1) / 2 + 1; /* are vertices packed, or reg-aligned? */ + c.nr_bytes = c.nr_regs * REG_SIZE; + + + /* Begin the compilation: + */ + brw_init_compile(&c.func); + + c.func.single_program_flow = 1; + + /* For some reason the thread is spawned with only 4 channels + * unmasked. + */ + brw_set_mask_control(&c.func, BRW_MASK_DISABLE); + + + /* Note that primitives which don't require a GS program have + * already been weeded out by this stage: + */ + switch (key->primitive) { + case PIPE_PRIM_QUADS: + brw_gs_quads( &c ); + break; + case PIPE_PRIM_QUAD_STRIP: + brw_gs_quad_strip( &c ); + break; + case PIPE_PRIM_LINE_LOOP: + brw_gs_lines( &c ); + break; + case PIPE_PRIM_LINES: + if (key->hint_gs_always) + brw_gs_lines( &c ); + else { + return; + } + break; + case PIPE_PRIM_TRIANGLES: + if (key->hint_gs_always) + brw_gs_tris( &c ); + else { + return; + } + break; + case PIPE_PRIM_POINTS: + if (key->hint_gs_always) + brw_gs_points( &c ); + else { + return; + } + break; + default: + return; + } + + /* get the program + */ + program = brw_get_program(&c.func, &program_size); + + /* Upload + */ + brw->gs.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_GS_PROG], + &c.key, + sizeof(c.key), + program, + program_size, + &c.prog_data, + &brw->gs.prog_data ); +} + + +static boolean search_cache( struct brw_context *brw, + struct brw_gs_prog_key *key ) +{ + return brw_search_cache(&brw->cache[BRW_GS_PROG], + key, sizeof(*key), + &brw->gs.prog_data, + &brw->gs.prog_gs_offset); +} + + +static const int gs_prim[PIPE_PRIM_POLYGON+1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINE_LOOP, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_QUADS, + PIPE_PRIM_QUAD_STRIP, + PIPE_PRIM_TRIANGLES +}; + +static void populate_key( struct brw_context *brw, + struct brw_gs_prog_key *key ) +{ + memset(key, 0, sizeof(*key)); + + /* CACHE_NEW_VS_PROG */ + key->attrs = brw->vs.prog_data->outputs_written; + + /* BRW_NEW_PRIMITIVE */ + key->primitive = gs_prim[brw->primitive]; + + key->hint_gs_always = 0; /* debug code? */ + + key->need_gs_prog = (key->hint_gs_always || + brw->primitive == PIPE_PRIM_QUADS || + brw->primitive == PIPE_PRIM_QUAD_STRIP || + brw->primitive == PIPE_PRIM_LINE_LOOP); +} + +/* Calculate interpolants for triangle and line rasterization. + */ +static void upload_gs_prog( struct brw_context *brw ) +{ + struct brw_gs_prog_key key; + + /* Populate the key: + */ + populate_key(brw, &key); + + if (brw->gs.prog_active != key.need_gs_prog) { + brw->state.dirty.cache |= CACHE_NEW_GS_PROG; + brw->gs.prog_active = key.need_gs_prog; + } + + if (brw->gs.prog_active) { + if (!search_cache(brw, &key)) + compile_gs_prog( brw, &key ); + } +} + + +const struct brw_tracked_state brw_gs_prog = { + .dirty = { + .brw = BRW_NEW_PRIMITIVE, + .cache = CACHE_NEW_VS_PROG + }, + .update = upload_gs_prog +}; diff --git a/src/mesa/pipe/i965simple/brw_gs.h b/src/mesa/pipe/i965simple/brw_gs.h new file mode 100644 index 0000000000..f09141c6aa --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_gs.h @@ -0,0 +1,75 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_GS_H +#define BRW_GS_H + + +#include "brw_context.h" +#include "brw_eu.h" + +#define MAX_GS_VERTS (4) + +struct brw_gs_prog_key { + unsigned attrs:32; + unsigned primitive:4; + unsigned hint_gs_always:1; + unsigned need_gs_prog:1; + unsigned pad:26; +}; + +struct brw_gs_compile { + struct brw_compile func; + struct brw_gs_prog_key key; + struct brw_gs_prog_data prog_data; + + struct { + struct brw_reg R0; + struct brw_reg vertex[MAX_GS_VERTS]; + } reg; + + /* 3 different ways of expressing vertex size: + */ + unsigned nr_attrs; + unsigned nr_regs; + unsigned nr_bytes; +}; + +#define ATTR_SIZE (4*4) + +void brw_gs_quads( struct brw_gs_compile *c ); +void brw_gs_quad_strip( struct brw_gs_compile *c ); +void brw_gs_tris( struct brw_gs_compile *c ); +void brw_gs_lines( struct brw_gs_compile *c ); +void brw_gs_points( struct brw_gs_compile *c ); + +#endif diff --git a/src/mesa/pipe/i965simple/brw_gs_emit.c b/src/mesa/pipe/i965simple/brw_gs_emit.c new file mode 100644 index 0000000000..c3cc90b10f --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_gs_emit.c @@ -0,0 +1,148 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_gs.h" + +static void brw_gs_alloc_regs( struct brw_gs_compile *c, + unsigned nr_verts ) +{ + unsigned i = 0,j; + + /* Register usage is static, precompute here: + */ + c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; + + /* Payload vertices plus space for more generated vertices: + */ + for (j = 0; j < nr_verts; j++) { + c->reg.vertex[j] = brw_vec4_grf(i, 0); + i += c->nr_regs; + } + + c->prog_data.urb_read_length = c->nr_regs; + c->prog_data.total_grf = i; +} + + +static void brw_gs_emit_vue(struct brw_gs_compile *c, + struct brw_reg vert, + boolean last, + unsigned header) +{ + struct brw_compile *p = &c->func; + boolean allocate = !last; + + /* Overwrite PrimType and PrimStart in the message header, for + * each vertex in turn: + */ + brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header)); + + /* Copy the vertex from vertn into m1..mN+1: + */ + brw_copy8(p, brw_message_reg(1), vert, c->nr_regs); + + /* Send each vertex as a seperate write to the urb. This is + * different to the concept in brw_sf_emit.c, where subsequent + * writes are used to build up a single urb entry. Each of these + * writes instantiates a seperate urb entry, and a new one must be + * allocated each time. + */ + brw_urb_WRITE(p, + allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), + 0, + c->reg.R0, + allocate, + 1, /* used */ + c->nr_regs + 1, /* msg length */ + allocate ? 1 : 0, /* response length */ + allocate ? 0 : 1, /* eot */ + 1, /* writes_complete */ + 0, /* urb offset */ + BRW_URB_SWIZZLE_NONE); +} + + + +void brw_gs_quads( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 4); + + /* Use polygons for correct edgeflag behaviour. Note that vertex 3 + * is the PV for quads, but vertex 0 for polygons: + */ + brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); +} + +void brw_gs_quad_strip( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 4); + + brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); +} + +void brw_gs_tris( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 3); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_TRILIST << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_TRILIST << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_TRILIST << 2) | R02_PRIM_END)); +} + +void brw_gs_lines( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 2); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_END)); +} + +void brw_gs_points( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 1); + brw_gs_emit_vue(c, c->reg.vertex[0], 1, ((_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END)); +} + + + + + + + + diff --git a/src/mesa/pipe/i965simple/brw_gs_state.c b/src/mesa/pipe/i965simple/brw_gs_state.c new file mode 100644 index 0000000000..8e62eb4bd7 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_gs_state.c @@ -0,0 +1,88 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + + + +static void upload_gs_unit( struct brw_context *brw ) +{ + struct brw_gs_unit_state gs; + + memset(&gs, 0, sizeof(gs)); + + /* CACHE_NEW_GS_PROG */ + if (brw->gs.prog_active) { + gs.thread0.grf_reg_count = + ALIGN(brw->gs.prog_data->total_grf, 16) / 16 - 1; + gs.thread0.kernel_start_pointer = brw->gs.prog_gs_offset >> 6; + gs.thread3.urb_entry_read_length = brw->gs.prog_data->urb_read_length; + } + else { + gs.thread0.grf_reg_count = 0; + gs.thread0.kernel_start_pointer = 0; + gs.thread3.urb_entry_read_length = 1; + } + + /* BRW_NEW_URB_FENCE */ + gs.thread4.nr_urb_entries = brw->urb.nr_gs_entries; + gs.thread4.urb_entry_allocation_size = brw->urb.vsize - 1; + + gs.thread4.max_threads = 0; /* Hardware requirement */ + + if (BRW_DEBUG & DEBUG_STATS) + gs.thread4.stats_enable = 1; + + /* CONSTANT */ + gs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + gs.thread1.single_program_flow = 1; + gs.thread3.dispatch_grf_start_reg = 1; + gs.thread3.const_urb_entry_read_offset = 0; + gs.thread3.const_urb_entry_read_length = 0; + gs.thread3.urb_entry_read_offset = 0; + + + brw->gs.state_gs_offset = brw_cache_data( &brw->cache[BRW_GS_UNIT], &gs ); +} + + +const struct brw_tracked_state brw_gs_unit = { + .dirty = { + .brw = (BRW_NEW_CURBE_OFFSETS | + BRW_NEW_URB_FENCE), + .cache = CACHE_NEW_GS_PROG + }, + .update = upload_gs_unit +}; diff --git a/src/mesa/pipe/i965simple/brw_metaops.c b/src/mesa/pipe/i965simple/brw_metaops.c new file mode 100644 index 0000000000..09f5d28961 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_metaops.c @@ -0,0 +1,538 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + * frame buffer texture by Gary Wong + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_draw.h" + +#if 0 + +#define INIT(brw, STRUCT, ATTRIB) \ +do { \ + brw->attribs.ATTRIB = &ctx->ATTRIB; \ +} while (0) + +#define DUP(brw, STRUCT, ATTRIB) \ +do { \ + brw->metaops.attribs.ATTRIB = MALLOC_STRUCT(STRUCT); \ + memcpy(brw->metaops.attribs.ATTRIB, \ + brw->attribs.ATTRIB, \ + sizeof(struct STRUCT)); \ +} while (0) + + +#define INSTALL(brw, ATTRIB, STATE) \ +do { \ + brw->attribs.ATTRIB = brw->metaops.attribs.ATTRIB; \ + brw->state.dirty.mesa |= STATE; \ +} while (0) + +#define RESTORE(brw, ATTRIB, STATE) \ +do { \ + brw->attribs.ATTRIB = &brw->intel.ctx.ATTRIB; \ + brw->state.dirty.mesa |= STATE; \ +} while (0) + +static void init_attribs( struct brw_context *brw ) +{ + DUP(brw, gl_colorbuffer_attrib, Color); + DUP(brw, gl_depthbuffer_attrib, Depth); + DUP(brw, gl_fog_attrib, Fog); + DUP(brw, gl_hint_attrib, Hint); + DUP(brw, gl_light_attrib, Light); + DUP(brw, gl_line_attrib, Line); + DUP(brw, gl_point_attrib, Point); + DUP(brw, gl_polygon_attrib, Polygon); + DUP(brw, gl_scissor_attrib, Scissor); + DUP(brw, gl_stencil_attrib, Stencil); + DUP(brw, gl_texture_attrib, Texture); + DUP(brw, gl_transform_attrib, Transform); + DUP(brw, gl_viewport_attrib, Viewport); + DUP(brw, gl_vertex_program_state, VertexProgram); + DUP(brw, gl_fragment_program_state, FragmentProgram); +} + +static void install_attribs( struct brw_context *brw ) +{ + INSTALL(brw, Color, _NEW_COLOR); + INSTALL(brw, Depth, _NEW_DEPTH); + INSTALL(brw, Fog, _NEW_FOG); + INSTALL(brw, Hint, _NEW_HINT); + INSTALL(brw, Light, _NEW_LIGHT); + INSTALL(brw, Line, _NEW_LINE); + INSTALL(brw, Point, _NEW_POINT); + INSTALL(brw, Polygon, _NEW_POLYGON); + INSTALL(brw, Scissor, _NEW_SCISSOR); + INSTALL(brw, Stencil, _NEW_STENCIL); + INSTALL(brw, Texture, _NEW_TEXTURE); + INSTALL(brw, Transform, _NEW_TRANSFORM); + INSTALL(brw, Viewport, _NEW_VIEWPORT); + INSTALL(brw, VertexProgram, _NEW_PROGRAM); + INSTALL(brw, FragmentProgram, _NEW_PROGRAM); +} + +static void restore_attribs( struct brw_context *brw ) +{ + RESTORE(brw, Color, _NEW_COLOR); + RESTORE(brw, Depth, _NEW_DEPTH); + RESTORE(brw, Fog, _NEW_FOG); + RESTORE(brw, Hint, _NEW_HINT); + RESTORE(brw, Light, _NEW_LIGHT); + RESTORE(brw, Line, _NEW_LINE); + RESTORE(brw, Point, _NEW_POINT); + RESTORE(brw, Polygon, _NEW_POLYGON); + RESTORE(brw, Scissor, _NEW_SCISSOR); + RESTORE(brw, Stencil, _NEW_STENCIL); + RESTORE(brw, Texture, _NEW_TEXTURE); + RESTORE(brw, Transform, _NEW_TRANSFORM); + RESTORE(brw, Viewport, _NEW_VIEWPORT); + RESTORE(brw, VertexProgram, _NEW_PROGRAM); + RESTORE(brw, FragmentProgram, _NEW_PROGRAM); +} + + +static const char *vp_prog = + "!!ARBvp1.0\n" + "MOV result.color, vertex.color;\n" + "MOV result.position, vertex.position;\n" + "END\n"; + +static const char *fp_prog = + "!!ARBfp1.0\n" + "MOV result.color, fragment.color;\n" + "END\n"; + +static const char *fp_tex_prog = + "!!ARBfp1.0\n" + "TEMP a;\n" + "ADD a, fragment.position, program.local[0];\n" + "MUL a, a, program.local[1];\n" + "TEX result.color, a, texture[0], 2D;\n" + "MOV result.depth.z, fragment.position;\n" + "END\n"; + +/* Derived values of importance: + * + * FragmentProgram->_Current + * VertexProgram->_Enabled + * brw->vertex_program + * DrawBuffer->_ColorDrawBufferMask[0] + * + * + * More if drawpixels-through-texture is added. + */ +static void init_metaops_state( struct brw_context *brw ) +{ + GLcontext *ctx = &brw->intel.ctx; + + brw->metaops.vbo = ctx->Driver.NewBufferObject(ctx, 1, GL_ARRAY_BUFFER_ARB); + + ctx->Driver.BufferData(ctx, + GL_ARRAY_BUFFER_ARB, + 4096, + NULL, + GL_DYNAMIC_DRAW_ARB, + brw->metaops.vbo); + + brw->metaops.fp = (struct gl_fragment_program *) + ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 1 ); + + brw->metaops.fp_tex = (struct gl_fragment_program *) + ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 1 ); + + brw->metaops.vp = (struct gl_vertex_program *) + ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 1 ); + + _mesa_parse_arb_fragment_program(ctx, GL_FRAGMENT_PROGRAM_ARB, + fp_prog, strlen(fp_prog), + brw->metaops.fp); + + _mesa_parse_arb_fragment_program(ctx, GL_FRAGMENT_PROGRAM_ARB, + fp_tex_prog, strlen(fp_tex_prog), + brw->metaops.fp_tex); + + _mesa_parse_arb_vertex_program(ctx, GL_VERTEX_PROGRAM_ARB, + vp_prog, strlen(vp_prog), + brw->metaops.vp); + + brw->metaops.attribs.VertexProgram->Current = brw->metaops.vp; + brw->metaops.attribs.VertexProgram->_Enabled = TRUE; + + brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp; +} + +static void meta_flat_shade( struct intel_context *intel ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + brw->metaops.attribs.Raster->flatshade = TRUE; + brw->state.dirty.mesa |= _NEW_LIGHT; +} + + +static void meta_no_stencil_write( struct intel_context *intel ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + brw->metaops.attribs.Stencil->Enabled = FALSE; + brw->metaops.attribs.Stencil->WriteMask[0] = FALSE; + brw->state.dirty.mesa |= _NEW_STENCIL; +} + +static void meta_no_depth_write( struct intel_context *intel ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + brw->metaops.attribs.Depth->Test = FALSE; + brw->metaops.attribs.Depth->Mask = FALSE; + brw->state.dirty.mesa |= _NEW_DEPTH; +} + + +static void meta_depth_replace( struct intel_context *intel ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, TRUE ) + * ctx->Driver.DepthMask( ctx, TRUE ) + */ + brw->metaops.attribs.Depth->Test = TRUE; + brw->metaops.attribs.Depth->Mask = TRUE; + brw->state.dirty.mesa |= _NEW_DEPTH; + + /* ctx->Driver.DepthFunc( ctx, GL_ALWAYS ) + */ + brw->metaops.attribs.Depth->Func = GL_ALWAYS; + + brw->state.dirty.mesa |= _NEW_DEPTH; +} + + +static void meta_stencil_replace( struct intel_context *intel, + unsigned s_mask, + unsigned s_clear) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + brw->metaops.attribs.Stencil->Enabled = TRUE; + brw->metaops.attribs.Stencil->WriteMask[0] = s_mask; + brw->metaops.attribs.Stencil->ValueMask[0] = 0xff; + brw->metaops.attribs.Stencil->Ref[0] = s_clear; + brw->metaops.attribs.Stencil->Function[0] = GL_ALWAYS; + brw->metaops.attribs.Stencil->FailFunc[0] = GL_REPLACE; + brw->metaops.attribs.Stencil->ZPassFunc[0] = GL_REPLACE; + brw->metaops.attribs.Stencil->ZFailFunc[0] = GL_REPLACE; + brw->state.dirty.mesa |= _NEW_STENCIL; +} + + +static void meta_color_mask( struct intel_context *intel, boolean state ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + if (state) + COPY_4V(brw->metaops.attribs.Color->ColorMask, + brw->intel.ctx.Color.ColorMask); + else + ASSIGN_4V(brw->metaops.attribs.Color->ColorMask, 0, 0, 0, 0); + + brw->state.dirty.mesa |= _NEW_COLOR; +} + +static void meta_no_texture( struct intel_context *intel ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp; + + brw->metaops.attribs.Texture->CurrentUnit = 0; + brw->metaops.attribs.Texture->_EnabledUnits = 0; + brw->metaops.attribs.Texture->_EnabledCoordUnits = 0; + brw->metaops.attribs.Texture->Unit[ 0 ].Enabled = 0; + brw->metaops.attribs.Texture->Unit[ 0 ]._ReallyEnabled = 0; + + brw->state.dirty.mesa |= _NEW_TEXTURE | _NEW_PROGRAM; +} + +static void meta_texture_blend_replace(struct intel_context *intel) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + brw->metaops.attribs.Texture->CurrentUnit = 0; + brw->metaops.attribs.Texture->_EnabledUnits = 1; + brw->metaops.attribs.Texture->_EnabledCoordUnits = 1; + brw->metaops.attribs.Texture->Unit[ 0 ].Enabled = TEXTURE_2D_BIT; + brw->metaops.attribs.Texture->Unit[ 0 ]._ReallyEnabled = TEXTURE_2D_BIT; + brw->metaops.attribs.Texture->Unit[ 0 ].Current2D = + intel->frame_buffer_texobj; + brw->metaops.attribs.Texture->Unit[ 0 ]._Current = + intel->frame_buffer_texobj; + + brw->state.dirty.mesa |= _NEW_TEXTURE | _NEW_PROGRAM; +} + +static void meta_import_pixel_state(struct intel_context *intel) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + RESTORE(brw, Color, _NEW_COLOR); + RESTORE(brw, Depth, _NEW_DEPTH); + RESTORE(brw, Fog, _NEW_FOG); + RESTORE(brw, Scissor, _NEW_SCISSOR); + RESTORE(brw, Stencil, _NEW_STENCIL); + RESTORE(brw, Texture, _NEW_TEXTURE); + RESTORE(brw, FragmentProgram, _NEW_PROGRAM); +} + +static void meta_frame_buffer_texture( struct intel_context *intel, + int xoff, int yoff ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + struct intel_region *region = intel_drawbuf_region( intel ); + + INSTALL(brw, FragmentProgram, _NEW_PROGRAM); + + brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp_tex; + /* This is unfortunate, but seems to be necessary, since later on we + will end up calling _mesa_load_state_parameters to lookup the + local params (below), and that will want to look in ctx.FragmentProgram + instead of brw->attribs.FragmentProgram. */ + intel->ctx.FragmentProgram.Current = brw->metaops.fp_tex; + + brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 0 ] = xoff; + brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 1 ] = yoff; + brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 2 ] = 0.0; + brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 3 ] = 0.0; + brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 0 ] = + 1.0 / region->pitch; + brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 1 ] = + -1.0 / region->height; + brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 2 ] = 0.0; + brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 3 ] = 1.0; + + brw->state.dirty.mesa |= _NEW_PROGRAM; +} + + +static void meta_draw_region( struct intel_context *intel, + struct intel_region *draw_region, + struct intel_region *depth_region ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + if (!brw->metaops.saved_draw_region) { + brw->metaops.saved_draw_region = brw->state.draw_region; + brw->metaops.saved_depth_region = brw->state.depth_region; + } + + brw->state.draw_region = draw_region; + brw->state.depth_region = depth_region; + + brw->state.dirty.mesa |= _NEW_BUFFERS; +} + + +static void meta_draw_quad(struct intel_context *intel, + float x0, float x1, + float y0, float y1, + float z, + ubyte red, ubyte green, + ubyte blue, ubyte alpha, + float s0, float s1, + float t0, float t1) +{ + GLcontext *ctx = &intel->ctx; + struct brw_context *brw = brw_context(&intel->ctx); + struct gl_client_array pos_array; + struct gl_client_array color_array; + struct gl_client_array *attribs[PIPE_ATTRIB_MAX]; + struct _mesa_prim prim[1]; + float pos[4][3]; + ubyte color[4]; + + ctx->Driver.BufferData(ctx, + GL_ARRAY_BUFFER_ARB, + sizeof(pos) + sizeof(color), + NULL, + GL_DYNAMIC_DRAW_ARB, + brw->metaops.vbo); + + pos[0][0] = x0; + pos[0][1] = y0; + pos[0][2] = z; + + pos[1][0] = x1; + pos[1][1] = y0; + pos[1][2] = z; + + pos[2][0] = x1; + pos[2][1] = y1; + pos[2][2] = z; + + pos[3][0] = x0; + pos[3][1] = y1; + pos[3][2] = z; + + + ctx->Driver.BufferSubData(ctx, + GL_ARRAY_BUFFER_ARB, + 0, + sizeof(pos), + pos, + brw->metaops.vbo); + + color[0] = red; + color[1] = green; + color[2] = blue; + color[3] = alpha; + + ctx->Driver.BufferSubData(ctx, + GL_ARRAY_BUFFER_ARB, + sizeof(pos), + sizeof(color), + color, + brw->metaops.vbo); + + /* Ignoring texture coords. + */ + + memset(attribs, 0, PIPE_ATTRIB_MAX * sizeof(*attribs)); + + attribs[VERT_ATTRIB_POS] = &pos_array; + attribs[VERT_ATTRIB_POS]->Ptr = 0; + attribs[VERT_ATTRIB_POS]->Type = GL_FLOAT; + attribs[VERT_ATTRIB_POS]->Enabled = 1; + attribs[VERT_ATTRIB_POS]->Size = 3; + attribs[VERT_ATTRIB_POS]->StrideB = 3 * sizeof(float); + attribs[VERT_ATTRIB_POS]->Stride = 3 * sizeof(float); + attribs[VERT_ATTRIB_POS]->_MaxElement = 4; + attribs[VERT_ATTRIB_POS]->Normalized = 0; + attribs[VERT_ATTRIB_POS]->BufferObj = brw->metaops.vbo; + + attribs[VERT_ATTRIB_COLOR0] = &color_array; + attribs[VERT_ATTRIB_COLOR0]->Ptr = (const ubyte *)sizeof(pos); + attribs[VERT_ATTRIB_COLOR0]->Type = GL_UNSIGNED_BYTE; + attribs[VERT_ATTRIB_COLOR0]->Enabled = 1; + attribs[VERT_ATTRIB_COLOR0]->Size = 4; + attribs[VERT_ATTRIB_COLOR0]->StrideB = 0; + attribs[VERT_ATTRIB_COLOR0]->Stride = 0; + attribs[VERT_ATTRIB_COLOR0]->_MaxElement = 1; + attribs[VERT_ATTRIB_COLOR0]->Normalized = 1; + attribs[VERT_ATTRIB_COLOR0]->BufferObj = brw->metaops.vbo; + + /* Just ignoring texture coordinates for now. + */ + + memset(prim, 0, sizeof(*prim)); + + prim[0].mode = GL_TRIANGLE_FAN; + prim[0].begin = 1; + prim[0].end = 1; + prim[0].weak = 0; + prim[0].pad = 0; + prim[0].start = 0; + prim[0].count = 4; + + brw_draw_prims(&brw->intel.ctx, + (const struct gl_client_array **)attribs, + prim, 1, + NULL, + 0, + 3 ); +} + + +static void install_meta_state( struct intel_context *intel ) +{ + GLcontext *ctx = &intel->ctx; + struct brw_context *brw = brw_context(ctx); + + if (!brw->metaops.vbo) { + init_metaops_state(brw); + } + + install_attribs(brw); + + meta_no_texture(&brw->intel); + meta_flat_shade(&brw->intel); + brw->metaops.restore_draw_mask = ctx->DrawBuffer->_ColorDrawBufferMask[0]; + brw->metaops.restore_fp = ctx->FragmentProgram.Current; + + /* This works without adjusting refcounts. Fix later? + */ + brw->metaops.saved_draw_region = brw->state.draw_region; + brw->metaops.saved_depth_region = brw->state.depth_region; + brw->metaops.active = 1; + + brw->state.dirty.brw |= BRW_NEW_METAOPS; +} + +static void leave_meta_state( struct intel_context *intel ) +{ + GLcontext *ctx = &intel->ctx; + struct brw_context *brw = brw_context(ctx); + + restore_attribs(brw); + + ctx->DrawBuffer->_ColorDrawBufferMask[0] = brw->metaops.restore_draw_mask; + ctx->FragmentProgram.Current = brw->metaops.restore_fp; + + brw->state.draw_region = brw->metaops.saved_draw_region; + brw->state.depth_region = brw->metaops.saved_depth_region; + brw->metaops.saved_draw_region = NULL; + brw->metaops.saved_depth_region = NULL; + brw->metaops.active = 0; + + brw->state.dirty.mesa |= _NEW_BUFFERS; + brw->state.dirty.brw |= BRW_NEW_METAOPS; +} + + + +void brw_init_metaops( struct brw_context *brw ) +{ + init_attribs(brw); +} + +void brw_destroy_metaops( struct brw_context *brw ) +{ + GLcontext *ctx = &brw->intel.ctx; + + if (brw->metaops.vbo) + ctx->Driver.DeleteBuffer( ctx, brw->metaops.vbo ); + +/* ctx->Driver.DeleteProgram( ctx, brw->metaops.fp ); */ +/* ctx->Driver.DeleteProgram( ctx, brw->metaops.fp_tex ); */ +/* ctx->Driver.DeleteProgram( ctx, brw->metaops.vp ); */ +} +#endif diff --git a/src/mesa/pipe/i965simple/brw_misc_state.c b/src/mesa/pipe/i965simple/brw_misc_state.c new file mode 100644 index 0000000000..0750502334 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_misc_state.c @@ -0,0 +1,425 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_batch.h" +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + + + + + +/*********************************************************************** + * Blend color + */ + +void brw_upload_blend_constant_color(struct brw_context *brw) +{ + struct brw_blend_constant_color bcc; + + memset(&bcc, 0, sizeof(bcc)); + bcc.header.opcode = CMD_BLEND_CONSTANT_COLOR; + bcc.header.length = sizeof(bcc)/4-2; + bcc.blend_constant_color[0] = brw->attribs.BlendColor.color[0]; + bcc.blend_constant_color[1] = brw->attribs.BlendColor.color[1]; + bcc.blend_constant_color[2] = brw->attribs.BlendColor.color[2]; + bcc.blend_constant_color[3] = brw->attribs.BlendColor.color[3]; + + BRW_CACHED_BATCH_STRUCT(brw, &bcc); +} + + +/** + * Upload the binding table pointers, which point each stage's array of surface + * state pointers. + * + * The binding table pointers are relative to the surface state base address, + * which is the BRW_SS_POOL cache buffer. + */ +static void upload_binding_table_pointers(struct brw_context *brw) +{ + struct brw_binding_table_pointers btp; + memset(&btp, 0, sizeof(btp)); + + btp.header.opcode = CMD_BINDING_TABLE_PTRS; + btp.header.length = sizeof(btp)/4 - 2; + btp.vs = 0; + btp.gs = 0; + btp.clp = 0; + btp.sf = 0; + btp.wm = brw->wm.bind_ss_offset; + + BRW_CACHED_BATCH_STRUCT(brw, &btp); +} + +const struct brw_tracked_state brw_binding_table_pointers = { + .dirty = { + .brw = 0, + .cache = CACHE_NEW_SURF_BIND + }, + .update = upload_binding_table_pointers, +}; + + +/** + * Upload pointers to the per-stage state. + * + * The state pointers in this packet are all relative to the general state + * base address set by CMD_STATE_BASE_ADDRESS, which is the BRW_GS_POOL buffer. + */ +static void upload_pipelined_state_pointers(struct brw_context *brw ) +{ + struct brw_pipelined_state_pointers psp; + memset(&psp, 0, sizeof(psp)); + + psp.header.opcode = CMD_PIPELINED_STATE_POINTERS; + psp.header.length = sizeof(psp)/4 - 2; + + psp.vs.offset = brw->vs.state_gs_offset >> 5; + psp.sf.offset = brw->sf.state_gs_offset >> 5; + psp.wm.offset = brw->wm.state_gs_offset >> 5; + psp.cc.offset = brw->cc.state_gs_offset >> 5; + + /* GS gets turned on and off regularly. Need to re-emit URB fence + * after this occurs. + */ + if (brw->gs.prog_active) { + psp.gs.offset = brw->gs.state_gs_offset >> 5; + psp.gs.enable = 1; + } + + { + psp.clp.offset = brw->clip.state_gs_offset >> 5; + psp.clp.enable = 1; + } + + + if (BRW_CACHED_BATCH_STRUCT(brw, &psp)) + brw->state.dirty.brw |= BRW_NEW_PSP; +} + +const struct brw_tracked_state brw_pipelined_state_pointers = { + .dirty = { + .brw = 0, + .cache = (CACHE_NEW_VS_UNIT | + CACHE_NEW_GS_UNIT | + CACHE_NEW_GS_PROG | + CACHE_NEW_CLIP_UNIT | + CACHE_NEW_SF_UNIT | + CACHE_NEW_WM_UNIT | + CACHE_NEW_CC_UNIT) + }, + .update = upload_pipelined_state_pointers +}; + +static void upload_psp_urb_cbs(struct brw_context *brw ) +{ + upload_pipelined_state_pointers(brw); + brw_upload_urb_fence(brw); + brw_upload_constant_buffer_state(brw); +} + + +const struct brw_tracked_state brw_psp_urb_cbs = { + .dirty = { + .brw = BRW_NEW_URB_FENCE, + .cache = (CACHE_NEW_VS_UNIT | + CACHE_NEW_GS_UNIT | + CACHE_NEW_GS_PROG | + CACHE_NEW_CLIP_UNIT | + CACHE_NEW_SF_UNIT | + CACHE_NEW_WM_UNIT | + CACHE_NEW_CC_UNIT) + }, + .update = upload_psp_urb_cbs +}; + +/** + * Upload the depthbuffer offset and format. + * + * We have to do this per state validation as we need to emit the relocation + * in the batch buffer. + */ +static void upload_depthbuffer(struct brw_context *brw) +{ + struct pipe_surface *depth_surface = brw->attribs.FrameBuffer.zbuf; + + BEGIN_BATCH(5, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (5 - 2)); + if (depth_surface == NULL) { + OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) | + (BRW_SURFACE_NULL << 29)); + OUT_BATCH(0); + OUT_BATCH(0); + OUT_BATCH(0); + } else { + unsigned int format; + + switch (depth_surface->cpp) { + case 2: + format = BRW_DEPTHFORMAT_D16_UNORM; + break; + case 4: + if (depth_surface->format == PIPE_FORMAT_Z32_FLOAT) + format = BRW_DEPTHFORMAT_D32_FLOAT; + else + format = BRW_DEPTHFORMAT_D24_UNORM_S8_UINT; + break; + default: + assert(0); + return; + } + + OUT_BATCH(((depth_surface->pitch * depth_surface->cpp) - 1) | + (format << 18) | + (BRW_TILEWALK_YMAJOR << 26) | +#if 0 + (depth_surface->region->tiled << 27) | +#endif + (BRW_SURFACE_2D << 29)); + OUT_RELOC(depth_surface->buffer, + PIPE_BUFFER_FLAG_READ | PIPE_BUFFER_FLAG_WRITE, 0); + OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) | + ((depth_surface->pitch - 1) << 6) | + ((depth_surface->height - 1) << 19)); + OUT_BATCH(0); + } + ADVANCE_BATCH(); +} + +const struct brw_tracked_state brw_depthbuffer = { + .dirty = { + .brw = 0, + .cache = 0 + }, + .update = upload_depthbuffer, +}; + + + +/*********************************************************************** + * Polygon stipple offset packet + */ + +static void upload_polygon_stipple_offset(struct brw_context *brw) +{ + struct brw_polygon_stipple_offset bpso; + + memset(&bpso, 0, sizeof(bpso)); + bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET; + bpso.header.length = sizeof(bpso)/4-2; + + bpso.bits0.x_offset = 0; + bpso.bits0.y_offset = 0; + + BRW_CACHED_BATCH_STRUCT(brw, &bpso); +} + +/*********************************************************************** + * Polygon stipple packet + */ + +void brw_upload_polygon_stipple(struct brw_context *brw) +{ + struct brw_polygon_stipple bps; + unsigned i; + + /*update the offset at the same time it will always be 0*/ + upload_polygon_stipple_offset(brw); + memset(&bps, 0, sizeof(bps)); + bps.header.opcode = CMD_POLY_STIPPLE_PATTERN; + bps.header.length = sizeof(bps)/4-2; + + for (i = 0; i < 32; i++) + bps.stipple[i] = brw->attribs.PolygonStipple->stipple[31 - i]; /* invert */ + + BRW_CACHED_BATCH_STRUCT(brw, &bps); +} + +/*********************************************************************** + * Line stipple packet + */ + +void brw_upload_line_stipple(struct brw_context *brw) +{ + struct brw_line_stipple bls; + float tmp; + int tmpi; + + memset(&bls, 0, sizeof(bls)); + bls.header.opcode = CMD_LINE_STIPPLE_PATTERN; + bls.header.length = sizeof(bls)/4 - 2; + + bls.bits0.pattern = brw->attribs.Raster->line_stipple_pattern; + bls.bits1.repeat_count = brw->attribs.Raster->line_stipple_factor; + + tmp = 1.0 / (float) brw->attribs.Raster->line_stipple_factor; + tmpi = tmp * (1<<13); + + + bls.bits1.inverse_repeat_count = tmpi; + + BRW_CACHED_BATCH_STRUCT(brw, &bls); +} + + +/*********************************************************************** + * Misc constant state packets + */ + +static void upload_pipe_control(struct brw_context *brw) +{ + struct brw_pipe_control pc; + + return; + + memset(&pc, 0, sizeof(pc)); + + pc.header.opcode = CMD_PIPE_CONTROL; + pc.header.length = sizeof(pc)/4 - 2; + pc.header.post_sync_operation = PIPE_CONTROL_NOWRITE; + + pc.header.instruction_state_cache_flush_enable = 1; + + pc.bits1.dest_addr_type = PIPE_CONTROL_GTTWRITE_GLOBAL; + + BRW_BATCH_STRUCT(brw, &pc); +} + +const struct brw_tracked_state brw_pipe_control = { + .dirty = { + .brw = BRW_NEW_CONTEXT, + .cache = 0 + }, + .update = upload_pipe_control +}; + + +/*********************************************************************** + * Misc invarient state packets + */ + +static void upload_invarient_state( struct brw_context *brw ) +{ + { + /* 0x61040000 Pipeline Select */ + /* PipelineSelect : 0 */ + struct brw_pipeline_select ps; + + memset(&ps, 0, sizeof(ps)); + ps.header.opcode = CMD_PIPELINE_SELECT; + ps.header.pipeline_select = 0; + BRW_BATCH_STRUCT(brw, &ps); + } + + { + struct brw_global_depth_offset_clamp gdo; + memset(&gdo, 0, sizeof(gdo)); + + /* Disable depth offset clamping. + */ + gdo.header.opcode = CMD_GLOBAL_DEPTH_OFFSET_CLAMP; + gdo.header.length = sizeof(gdo)/4 - 2; + gdo.depth_offset_clamp = 0.0; + + BRW_BATCH_STRUCT(brw, &gdo); + } + + + /* 0x61020000 State Instruction Pointer */ + { + struct brw_system_instruction_pointer sip; + memset(&sip, 0, sizeof(sip)); + + sip.header.opcode = CMD_STATE_INSN_POINTER; + sip.header.length = 0; + sip.bits0.pad = 0; + sip.bits0.system_instruction_pointer = 0; + BRW_BATCH_STRUCT(brw, &sip); + } + + + { + struct brw_vf_statistics vfs; + memset(&vfs, 0, sizeof(vfs)); + + vfs.opcode = CMD_VF_STATISTICS; + if (BRW_DEBUG & DEBUG_STATS) + vfs.statistics_enable = 1; + + BRW_BATCH_STRUCT(brw, &vfs); + } +} + +const struct brw_tracked_state brw_invarient_state = { + .dirty = { + .brw = BRW_NEW_CONTEXT, + .cache = 0 + }, + .update = upload_invarient_state +}; + +/** + * Define the base addresses which some state is referenced from. + * + * This allows us to avoid having to emit relocations in many places for + * cached state, and instead emit pointers inside of large, mostly-static + * state pools. This comes at the expense of memory, and more expensive cache + * misses. + */ +static void upload_state_base_address( struct brw_context *brw ) +{ + /* Output the structure (brw_state_base_address) directly to the + * batchbuffer, so we can emit relocations inline. + */ + BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2)); + OUT_RELOC(brw->pool[BRW_GS_POOL].buffer, + PIPE_BUFFER_FLAG_READ, + 1); /* General state base address */ + OUT_RELOC(brw->pool[BRW_SS_POOL].buffer, + PIPE_BUFFER_FLAG_READ, + 1); /* Surface state base address */ + OUT_BATCH(1); /* Indirect object base address */ + OUT_BATCH(1); /* General state upper bound */ + OUT_BATCH(1); /* Indirect object upper bound */ + ADVANCE_BATCH(); +} + + +const struct brw_tracked_state brw_state_base_address = { + .dirty = { + .brw = BRW_NEW_CONTEXT, + .cache = 0 + }, + .update = upload_state_base_address +}; diff --git a/src/mesa/pipe/i965simple/brw_program.c b/src/mesa/pipe/i965simple/brw_program.c new file mode 100644 index 0000000000..c4640a2b06 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_program.c @@ -0,0 +1,130 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_context.h" +#include "brw_util.h" + +#if 0 +static void brwBindProgram( GLcontext *ctx, + int target, + struct gl_program *prog ) +{ + struct brw_context *brw = brw_context(ctx); + + switch (target) { + case GL_VERTEX_PROGRAM_ARB: + brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM; + break; + case GL_FRAGMENT_PROGRAM_ARB: + brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM; + break; + } +} + +static struct gl_program *brwNewProgram( GLcontext *ctx, + int target, + unsigned id ) +{ + struct brw_context *brw = brw_context(ctx); + + switch (target) { + case GL_VERTEX_PROGRAM_ARB: { + struct brw_vertex_program *prog = CALLOC_STRUCT(brw_vertex_program); + if (prog) { + prog->id = brw->program_id++; + + return _mesa_init_vertex_program( ctx, &prog->program, + target, id ); + } + else + return NULL; + } + + case GL_FRAGMENT_PROGRAM_ARB: { + struct brw_fragment_program *prog = CALLOC_STRUCT(brw_fragment_program); + if (prog) { + prog->id = brw->program_id++; + + return _mesa_init_fragment_program( ctx, &prog->program, + target, id ); + } + else + return NULL; + } + + default: + return _mesa_new_program(ctx, target, id); + } +} + +static void brwDeleteProgram( GLcontext *ctx, + struct gl_program *prog ) +{ + + _mesa_delete_program( ctx, prog ); +} + + +static boolean brwIsProgramNative( GLcontext *ctx, + int target, + struct gl_program *prog ) +{ + return TRUE; +} + +static void brwProgramStringNotify( GLcontext *ctx, + int target, + struct gl_program *prog ) +{ + if (target == GL_FRAGMENT_PROGRAM_ARB) { + struct brw_context *brw = brw_context(ctx); + struct brw_fragment_program *p = (struct brw_fragment_program *)prog; + struct brw_fragment_program *fp = (struct brw_fragment_program *)brw->fragment_program; + if (p == fp) + brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM; + p->id = brw->program_id++; + p->param_state = p->program.Base.Parameters->StateFlags; + } + else if (target == GL_VERTEX_PROGRAM_ARB) { + struct brw_context *brw = brw_context(ctx); + struct brw_vertex_program *p = (struct brw_vertex_program *)prog; + struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->vertex_program; + if (p == vp) + brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM; + p->id = brw->program_id++; + p->param_state = p->program.Base.Parameters->StateFlags; + + /* Also tell tnl about it: + */ + _tnl_program_string(ctx, target, prog); + } +} +#endif diff --git a/src/mesa/pipe/i965simple/brw_reg.h b/src/mesa/pipe/i965simple/brw_reg.h new file mode 100644 index 0000000000..9e885c3b3b --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_reg.h @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#define CMD_MI (0x0 << 29) +#define CMD_2D (0x2 << 29) +#define CMD_3D (0x3 << 29) + +#define MI_BATCH_BUFFER_END (CMD_MI | 0xA << 23) + +/* Stalls command execution waiting for the given events to have occurred. */ +#define MI_WAIT_FOR_EVENT (CMD_MI | (0x3 << 23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +/* Primitive dispatch on 830-945 */ +#define _3DPRIMITIVE (CMD_3D | (0x1f << 24)) +#define PRIM_INDIRECT (1<<23) +#define PRIM_INLINE (0<<23) +#define PRIM_INDIRECT_SEQUENTIAL (0<<17) +#define PRIM_INDIRECT_ELTS (1<<17) + +#define PRIM3D_TRILIST (0x0<<18) +#define PRIM3D_TRISTRIP (0x1<<18) +#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) +#define PRIM3D_TRIFAN (0x3<<18) +#define PRIM3D_POLY (0x4<<18) +#define PRIM3D_LINELIST (0x5<<18) +#define PRIM3D_LINESTRIP (0x6<<18) +#define PRIM3D_RECTLIST (0x7<<18) +#define PRIM3D_POINTLIST (0x8<<18) +#define PRIM3D_DIB (0x9<<18) +#define PRIM3D_MASK (0x1f<<18) + +#define XY_SETUP_BLT_CMD (CMD_2D | (0x01 << 22) | 6) + +#define XY_COLOR_BLT_CMD (CMD_2D | (0x50 << 22) | 4) + +#define XY_SRC_COPY_BLT_CMD (CMD_2D | (0x53 << 22) | 6) + +/* BR00 */ +#define XY_BLT_WRITE_ALPHA (1 << 21) +#define XY_BLT_WRITE_RGB (1 << 20) +#define XY_SRC_TILED (1 << 15) +#define XY_DST_TILED (1 << 11) + +/* BR13 */ +#define BR13_565 (0x1 << 24) +#define BR13_8888 (0x3 << 24) + +#define FENCE_LINEAR 0 +#define FENCE_XMAJOR 1 +#define FENCE_YMAJOR 2 diff --git a/src/mesa/pipe/i965simple/brw_regions.c b/src/mesa/pipe/i965simple/brw_regions.c new file mode 100644 index 0000000000..32ad988279 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_regions.c @@ -0,0 +1,39 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Provide additional functionality on top of bufmgr buffers: + * - 2d semantics and blit operations (XXX: remove/simplify blits??) + * - refcounting of buffers for multiple images in a buffer. + * - refcounting of buffer mappings. + */ + +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "brw_context.h" +#include "brw_blit.h" + + diff --git a/src/mesa/pipe/i965simple/brw_regions.h b/src/mesa/pipe/i965simple/brw_regions.h new file mode 100644 index 0000000000..1eebd1d5ee --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_regions.h @@ -0,0 +1,5 @@ +#ifndef BRW_REGIONS_H +#define BRW_REGIONS_H + + +#endif diff --git a/src/mesa/pipe/i965simple/brw_sf.c b/src/mesa/pipe/i965simple/brw_sf.c new file mode 100644 index 0000000000..f009ff37d9 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_sf.c @@ -0,0 +1,194 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_sf.h" +#include "brw_state.h" + +#if 0 +#define DO_SETUP_BITS ((1<<(FRAG_ATTRIB_MAX)) - 1) + +static void compile_sf_prog( struct brw_context *brw, + struct brw_sf_prog_key *key ) +{ + struct brw_sf_compile c; + const unsigned *program; + unsigned program_size; + unsigned i, idx; + + memset(&c, 0, sizeof(c)); + + /* Begin the compilation: + */ + brw_init_compile(&c.func); + + c.key = *key; + c.nr_attrs = brw_count_bits(c.key.attrs); + c.nr_attr_regs = (c.nr_attrs+1)/2; + c.nr_setup_attrs = brw_count_bits(c.key.attrs & DO_SETUP_BITS); + c.nr_setup_regs = (c.nr_setup_attrs+1)/2; + + c.prog_data.urb_read_length = c.nr_attr_regs; + c.prog_data.urb_entry_size = c.nr_setup_regs * 2; + + /* Construct map from attribute number to position in the vertex. + */ + for (i = idx = 0; i < VERT_RESULT_MAX; i++) + if (c.key.attrs & (1<= VERT_RESULT_TEX0 && i <= VERT_RESULT_TEX7) { + c.point_attrs[i].CoordReplace = + brw->attribs.Point->CoordReplace[i - VERT_RESULT_TEX0]; + } else + c.point_attrs[i].CoordReplace = FALSE; + idx++; + } + + /* Which primitive? Or all three? + */ + switch (key->primitive) { + case SF_TRIANGLES: + c.nr_verts = 3; + brw_emit_tri_setup( &c ); + break; + case SF_LINES: + c.nr_verts = 2; + brw_emit_line_setup( &c ); + break; + case SF_POINTS: + c.nr_verts = 1; + if (key->do_point_sprite) + brw_emit_point_sprite_setup( &c ); + else + brw_emit_point_setup( &c ); + break; + case SF_UNFILLED_TRIS: + c.nr_verts = 3; + brw_emit_anyprim_setup( &c ); + break; + default: + assert(0); + return; + } + + + /* get the program + */ + program = brw_get_program(&c.func, &program_size); + + /* Upload + */ + brw->sf.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_SF_PROG], + &c.key, + sizeof(c.key), + program, + program_size, + &c.prog_data, + &brw->sf.prog_data ); +} + + +static boolean search_cache( struct brw_context *brw, + struct brw_sf_prog_key *key ) +{ + return brw_search_cache(&brw->cache[BRW_SF_PROG], + key, sizeof(*key), + &brw->sf.prog_data, + &brw->sf.prog_gs_offset); +} + + +/* Calculate interpolants for triangle and line rasterization. + */ +static void upload_sf_prog( struct brw_context *brw ) +{ + struct brw_sf_prog_key key; + + memset(&key, 0, sizeof(key)); + + /* Populate the key, noting state dependencies: + */ + /* CACHE_NEW_VS_PROG */ + key.attrs = brw->vs.prog_data->outputs_written; + + /* BRW_NEW_REDUCED_PRIMITIVE */ + switch (brw->reduced_primitive) { + case PIPE_PRIM_TRIANGLES: + /* NOTE: We just use the edgeflag attribute as an indicator that + * unfilled triangles are active. We don't actually do the + * edgeflag testing here, it is already done in the clip + * program. + */ + if (key.attrs & (1<attribs.Point->PointSprite; + key.SpriteOrigin = brw->attribs.Point->SpriteOrigin; + /* BRW_NEW_RASTER */ + key.do_flat_shading = (brw->attribs.Raster->flatshade); + key.do_twoside_color = (brw->attribs.Light->Enabled && brw->attribs.Light->Model.TwoSide); + + /* _NEW_POLYGON */ + if (key.do_twoside_color) + key.frontface_ccw = (brw->attribs.Polygon->FrontFace == GL_CCW); + + + if (!search_cache(brw, &key)) + compile_sf_prog( brw, &key ); +} + + +const struct brw_tracked_state brw_sf_prog = { + .dirty = { + .brw = (BRW_NEW_RASTER | + BRW_NEW_REDUCED_PRIMITIVE), + .cache = CACHE_NEW_VS_PROG + }, + .update = upload_sf_prog +}; + +#endif diff --git a/src/mesa/pipe/i965simple/brw_sf.h b/src/mesa/pipe/i965simple/brw_sf.h new file mode 100644 index 0000000000..d04388325d --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_sf.h @@ -0,0 +1,111 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_SF_H +#define BRW_SF_H + +#include "brw_context.h" +#include "brw_eu.h" + + +#define SF_POINTS 0 +#define SF_LINES 1 +#define SF_TRIANGLES 2 +#define SF_UNFILLED_TRIS 3 + +struct brw_sf_prog_key { + unsigned attrs:32; + unsigned primitive:2; + unsigned do_twoside_color:1; + unsigned do_flat_shading:1; + unsigned frontface_ccw:1; + unsigned do_point_sprite:1; + unsigned pad:10; + int SpriteOrigin; +}; + +struct brw_sf_point_tex { + boolean CoordReplace; +}; + +struct brw_sf_compile { + struct brw_compile func; + struct brw_sf_prog_key key; + struct brw_sf_prog_data prog_data; + + struct brw_reg pv; + struct brw_reg det; + struct brw_reg dx0; + struct brw_reg dx2; + struct brw_reg dy0; + struct brw_reg dy2; + + /* z and 1/w passed in seperately: + */ + struct brw_reg z[3]; + struct brw_reg inv_w[3]; + + /* The vertices: + */ + struct brw_reg vert[3]; + + /* Temporaries, allocated after last vertex reg. + */ + struct brw_reg inv_det; + struct brw_reg a1_sub_a0; + struct brw_reg a2_sub_a0; + struct brw_reg tmp; + + struct brw_reg m1Cx; + struct brw_reg m2Cy; + struct brw_reg m3C0; + + unsigned nr_verts; + unsigned nr_attrs; + unsigned nr_attr_regs; + unsigned nr_setup_attrs; + unsigned nr_setup_regs; +#if 0 + ubyte attr_to_idx[VERT_RESULT_MAX]; + ubyte idx_to_attr[VERT_RESULT_MAX]; + struct brw_sf_point_tex point_attrs[VERT_RESULT_MAX]; +#endif +}; + + +void brw_emit_tri_setup( struct brw_sf_compile *c ); +void brw_emit_line_setup( struct brw_sf_compile *c ); +void brw_emit_point_setup( struct brw_sf_compile *c ); +void brw_emit_point_sprite_setup( struct brw_sf_compile *c ); +void brw_emit_anyprim_setup( struct brw_sf_compile *c ); + +#endif diff --git a/src/mesa/pipe/i965simple/brw_sf_emit.c b/src/mesa/pipe/i965simple/brw_sf_emit.c new file mode 100644 index 0000000000..93f23171f2 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_sf_emit.c @@ -0,0 +1,696 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_sf.h" + +#if 0 +static struct brw_reg get_vert_attr(struct brw_sf_compile *c, + struct brw_reg vert, + unsigned attr) +{ + unsigned off = c->attr_to_idx[attr] / 2; + unsigned sub = c->attr_to_idx[attr] % 2; + + return brw_vec4_grf(vert.nr + off, sub * 4); +} + +static boolean have_attr(struct brw_sf_compile *c, + unsigned attr) +{ + return (c->key.attrs & (1<func; + unsigned i; + + for (i = 0; i < 2; i++) { + if (have_attr(c, VERT_RESULT_COL0+i) && + have_attr(c, VERT_RESULT_BFC0+i)) + brw_MOV(p, + get_vert_attr(c, vert, VERT_RESULT_COL0+i), + get_vert_attr(c, vert, VERT_RESULT_BFC0+i)); + } +} + + +static void do_twoside_color( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *if_insn; + unsigned backface_conditional = c->key.frontface_ccw ? BRW_CONDITIONAL_G : BRW_CONDITIONAL_L; + + /* Already done in clip program: + */ + if (c->key.primitive == SF_UNFILLED_TRIS) + return; + + /* XXX: What happens if BFC isn't present? This could only happen + * for user-supplied vertex programs, as t_vp_build.c always does + * the right thing. + */ + if (!(have_attr(c, VERT_RESULT_COL0) && have_attr(c, VERT_RESULT_BFC0)) && + !(have_attr(c, VERT_RESULT_COL1) && have_attr(c, VERT_RESULT_BFC1))) + return; + + /* Need to use BRW_EXECUTE_4 and also do an 4-wide compare in order + * to get all channels active inside the IF. In the clipping code + * we run with NoMask, so it's not an option and we can use + * BRW_EXECUTE_1 for all comparisions. + */ + brw_push_insn_state(p); + brw_CMP(p, vec4(brw_null_reg()), backface_conditional, c->det, brw_imm_f(0)); + if_insn = brw_IF(p, BRW_EXECUTE_4); + { + switch (c->nr_verts) { + case 3: copy_bfc(c, c->vert[2]); + case 2: copy_bfc(c, c->vert[1]); + case 1: copy_bfc(c, c->vert[0]); + } + } + brw_ENDIF(p, if_insn); + brw_pop_insn_state(p); +} + + + +/*********************************************************************** + * Flat shading + */ + +#define VERT_RESULT_COLOR_BITS ((1<func; + unsigned i; + + for (i = VERT_RESULT_COL0; i <= VERT_RESULT_COL1; i++) { + if (have_attr(c,i)) + brw_MOV(p, + get_vert_attr(c, dst, i), + get_vert_attr(c, src, i)); + } +} + + + +/* Need to use a computed jump to copy flatshaded attributes as the + * vertices are ordered according to y-coordinate before reaching this + * point, so the PV could be anywhere. + */ +static void do_flatshade_triangle( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg ip = brw_ip_reg(); + unsigned nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS); + if (!nr) + return; + + /* Already done in clip program: + */ + if (c->key.primitive == SF_UNFILLED_TRIS) + return; + + brw_push_insn_state(p); + + brw_MUL(p, c->pv, c->pv, brw_imm_ud(nr*2+1)); + brw_JMPI(p, ip, ip, c->pv); + + copy_colors(c, c->vert[1], c->vert[0]); + copy_colors(c, c->vert[2], c->vert[0]); + brw_JMPI(p, ip, ip, brw_imm_ud(nr*4+1)); + + copy_colors(c, c->vert[0], c->vert[1]); + copy_colors(c, c->vert[2], c->vert[1]); + brw_JMPI(p, ip, ip, brw_imm_ud(nr*2)); + + copy_colors(c, c->vert[0], c->vert[2]); + copy_colors(c, c->vert[1], c->vert[2]); + + brw_pop_insn_state(p); +} + + +static void do_flatshade_line( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg ip = brw_ip_reg(); + unsigned nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS); + + if (!nr) + return; + + /* Already done in clip program: + */ + if (c->key.primitive == SF_UNFILLED_TRIS) + return; + + brw_push_insn_state(p); + + brw_MUL(p, c->pv, c->pv, brw_imm_ud(nr+1)); + brw_JMPI(p, ip, ip, c->pv); + copy_colors(c, c->vert[1], c->vert[0]); + + brw_JMPI(p, ip, ip, brw_imm_ud(nr)); + copy_colors(c, c->vert[0], c->vert[1]); + + brw_pop_insn_state(p); +} + + + +/*********************************************************************** + * Triangle setup. + */ + + +static void alloc_regs( struct brw_sf_compile *c ) +{ + unsigned reg, i; + + /* Values computed by fixed function unit: + */ + c->pv = retype(brw_vec1_grf(1, 1), BRW_REGISTER_TYPE_UD); + c->det = brw_vec1_grf(1, 2); + c->dx0 = brw_vec1_grf(1, 3); + c->dx2 = brw_vec1_grf(1, 4); + c->dy0 = brw_vec1_grf(1, 5); + c->dy2 = brw_vec1_grf(1, 6); + + /* z and 1/w passed in seperately: + */ + c->z[0] = brw_vec1_grf(2, 0); + c->inv_w[0] = brw_vec1_grf(2, 1); + c->z[1] = brw_vec1_grf(2, 2); + c->inv_w[1] = brw_vec1_grf(2, 3); + c->z[2] = brw_vec1_grf(2, 4); + c->inv_w[2] = brw_vec1_grf(2, 5); + + /* The vertices: + */ + reg = 3; + for (i = 0; i < c->nr_verts; i++) { + c->vert[i] = brw_vec8_grf(reg, 0); + reg += c->nr_attr_regs; + } + + /* Temporaries, allocated after last vertex reg. + */ + c->inv_det = brw_vec1_grf(reg, 0); reg++; + c->a1_sub_a0 = brw_vec8_grf(reg, 0); reg++; + c->a2_sub_a0 = brw_vec8_grf(reg, 0); reg++; + c->tmp = brw_vec8_grf(reg, 0); reg++; + + /* Note grf allocation: + */ + c->prog_data.total_grf = reg; + + + /* Outputs of this program - interpolation coefficients for + * rasterization: + */ + c->m1Cx = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 1, 0); + c->m2Cy = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 2, 0); + c->m3C0 = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 3, 0); +} + + +static void copy_z_inv_w( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned i; + + brw_push_insn_state(p); + + /* Copy both scalars with a single MOV: + */ + for (i = 0; i < c->nr_verts; i++) + brw_MOV(p, vec2(suboffset(c->vert[i], 2)), vec2(c->z[i])); + + brw_pop_insn_state(p); +} + + +static void invert_det( struct brw_sf_compile *c) +{ + /* Looks like we invert all 8 elements just to get 1/det in + * position 2 !?! + */ + brw_math(&c->func, + c->inv_det, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 0, + c->det, + BRW_MATH_DATA_SCALAR, + BRW_MATH_PRECISION_FULL); + +} + +#define NON_PERPECTIVE_ATTRS (FRAG_BIT_WPOS | \ + FRAG_BIT_COL0 | \ + FRAG_BIT_COL1) + +static boolean calculate_masks( struct brw_sf_compile *c, + unsigned reg, + ushort *pc, + ushort *pc_persp, + ushort *pc_linear) +{ + boolean is_last_attr = (reg == c->nr_setup_regs - 1); + unsigned persp_mask = c->key.attrs & ~NON_PERPECTIVE_ATTRS; + unsigned linear_mask; + + if (c->key.do_flat_shading) + linear_mask = c->key.attrs & ~(FRAG_BIT_COL0|FRAG_BIT_COL1); + else + linear_mask = c->key.attrs; + + *pc_persp = 0; + *pc_linear = 0; + *pc = 0xf; + + if (persp_mask & (1 << c->idx_to_attr[reg*2])) + *pc_persp = 0xf; + + if (linear_mask & (1 << c->idx_to_attr[reg*2])) + *pc_linear = 0xf; + + /* Maybe only processs one attribute on the final round: + */ + if (reg*2+1 < c->nr_setup_attrs) { + *pc |= 0xf0; + + if (persp_mask & (1 << c->idx_to_attr[reg*2+1])) + *pc_persp |= 0xf0; + + if (linear_mask & (1 << c->idx_to_attr[reg*2+1])) + *pc_linear |= 0xf0; + } + + return is_last_attr; +} + + + +void brw_emit_tri_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned i; + + c->nr_verts = 3; + alloc_regs(c); + invert_det(c); + copy_z_inv_w(c); + + if (c->key.do_twoside_color) + do_twoside_color(c); + + if (c->key.do_flat_shading) + do_flatshade_triangle(c); + + + for (i = 0; i < c->nr_setup_regs; i++) + { + /* Pair of incoming attributes: + */ + struct brw_reg a0 = offset(c->vert[0], i); + struct brw_reg a1 = offset(c->vert[1], i); + struct brw_reg a2 = offset(c->vert[2], i); + ushort pc, pc_persp, pc_linear; + boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + brw_MUL(p, a1, a1, c->inv_w[1]); + brw_MUL(p, a2, a2, c->inv_w[2]); + } + + + /* Calculate coefficients for interpolated values: + */ + if (pc_linear) + { + brw_set_predicate_control_flag_value(p, pc_linear); + + brw_ADD(p, c->a1_sub_a0, a1, negate(a0)); + brw_ADD(p, c->a2_sub_a0, a2, negate(a0)); + + /* calculate dA/dx + */ + brw_MUL(p, brw_null_reg(), c->a1_sub_a0, c->dy2); + brw_MAC(p, c->tmp, c->a2_sub_a0, negate(c->dy0)); + brw_MUL(p, c->m1Cx, c->tmp, c->inv_det); + + /* calculate dA/dy + */ + brw_MUL(p, brw_null_reg(), c->a2_sub_a0, c->dx0); + brw_MAC(p, c->tmp, c->a1_sub_a0, negate(c->dx2)); + brw_MUL(p, c->m2Cy, c->tmp, c->inv_det); + } + + { + brw_set_predicate_control_flag_value(p, pc); + /* start point for interpolation + */ + brw_MOV(p, c->m3C0, a0); + + /* Copy m0..m3 to URB. m0 is implicitly copied from r0 in + * the send instruction: + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), /* r0, will be copied to m0 */ + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* offset */ + BRW_URB_SWIZZLE_TRANSPOSE); /* XXX: Swizzle control "SF to windower" */ + } + } +} + + + +void brw_emit_line_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned i; + + + c->nr_verts = 2; + alloc_regs(c); + invert_det(c); + copy_z_inv_w(c); + + if (c->key.do_flat_shading) + do_flatshade_line(c); + + for (i = 0; i < c->nr_setup_regs; i++) + { + /* Pair of incoming attributes: + */ + struct brw_reg a0 = offset(c->vert[0], i); + struct brw_reg a1 = offset(c->vert[1], i); + ushort pc, pc_persp, pc_linear; + boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + brw_MUL(p, a1, a1, c->inv_w[1]); + } + + /* Calculate coefficients for position, color: + */ + if (pc_linear) { + brw_set_predicate_control_flag_value(p, pc_linear); + + brw_ADD(p, c->a1_sub_a0, a1, negate(a0)); + + brw_MUL(p, c->tmp, c->a1_sub_a0, c->dx0); + brw_MUL(p, c->m1Cx, c->tmp, c->inv_det); + + brw_MUL(p, c->tmp, c->a1_sub_a0, c->dy0); + brw_MUL(p, c->m2Cy, c->tmp, c->inv_det); + } + + { + brw_set_predicate_control_flag_value(p, pc); + + /* start point for interpolation + */ + brw_MOV(p, c->m3C0, a0); + + /* Copy m0..m3 to URB. + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* urb destination offset */ + BRW_URB_SWIZZLE_TRANSPOSE); + } + } +} + +void brw_emit_point_sprite_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned i; + + c->nr_verts = 1; + alloc_regs(c); + copy_z_inv_w(c); + for (i = 0; i < c->nr_setup_regs; i++) + { + struct brw_sf_point_tex *tex = &c->point_attrs[c->idx_to_attr[2*i]]; + struct brw_reg a0 = offset(c->vert[0], i); + ushort pc, pc_persp, pc_linear; + boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + if (!tex->CoordReplace) { + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + } + } + + if (tex->CoordReplace) { + /* Caculate 1.0/PointWidth */ + brw_math(&c->func, + c->tmp, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 0, + c->dx0, + BRW_MATH_DATA_SCALAR, + BRW_MATH_PRECISION_FULL); + + if (c->key.SpriteOrigin == GL_UPPER_LEFT) { + brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]); + brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0)); + brw_MUL(p, c->m2Cy, c->tmp, negate(c->inv_w[0])); + brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0)); + } else { + brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]); + brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0)); + brw_MUL(p, c->m2Cy, c->tmp, c->inv_w[0]); + brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0)); + } + } else { + brw_MOV(p, c->m1Cx, brw_imm_ud(0)); + brw_MOV(p, c->m2Cy, brw_imm_ud(0)); + } + + { + brw_set_predicate_control_flag_value(p, pc); + if (tex->CoordReplace) { + if (c->key.SpriteOrigin == GL_UPPER_LEFT) { + brw_MUL(p, c->m3C0, c->inv_w[0], brw_imm_f(1.0)); + brw_MOV(p, vec1(suboffset(c->m3C0, 0)), brw_imm_f(0.0)); + } + else + brw_MOV(p, c->m3C0, brw_imm_f(0.0)); + } else { + brw_MOV(p, c->m3C0, a0); /* constant value */ + } + + /* Copy m0..m3 to URB. + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* urb destination offset */ + BRW_URB_SWIZZLE_TRANSPOSE); + } + } +} + +/* Points setup - several simplifications as all attributes are + * constant across the face of the point (point sprites excluded!) + */ +void brw_emit_point_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned i; + + c->nr_verts = 1; + alloc_regs(c); + copy_z_inv_w(c); + + brw_MOV(p, c->m1Cx, brw_imm_ud(0)); /* zero - move out of loop */ + brw_MOV(p, c->m2Cy, brw_imm_ud(0)); /* zero - move out of loop */ + + for (i = 0; i < c->nr_setup_regs; i++) + { + struct brw_reg a0 = offset(c->vert[0], i); + ushort pc, pc_persp, pc_linear; + boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + /* This seems odd as the values are all constant, but the + * fragment shader will be expecting it: + */ + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + } + + + /* The delta values are always zero, just send the starting + * coordinate. Again, this is to fit in with the interpolation + * code in the fragment shader. + */ + { + brw_set_predicate_control_flag_value(p, pc); + + brw_MOV(p, c->m3C0, a0); /* constant value */ + + /* Copy m0..m3 to URB. + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* urb destination offset */ + BRW_URB_SWIZZLE_TRANSPOSE); + } + } +} + +void brw_emit_anyprim_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg ip = brw_ip_reg(); + struct brw_reg payload_prim = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0); + struct brw_reg payload_attr = get_element_ud(brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0), 0); + struct brw_reg primmask; + struct brw_instruction *jmp; + struct brw_reg v1_null_ud = vec1(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD)); + + alloc_regs(c); + + primmask = retype(get_element(c->tmp, 0), BRW_REGISTER_TYPE_UD); + + brw_MOV(p, primmask, brw_imm_ud(1)); + brw_SHL(p, primmask, primmask, payload_prim); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); + brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_TRILIST) | + (1<<_3DPRIM_TRISTRIP) | + (1<<_3DPRIM_TRIFAN) | + (1<<_3DPRIM_TRISTRIP_REVERSE) | + (1<<_3DPRIM_POLYGON) | + (1<<_3DPRIM_RECTLIST) | + (1<<_3DPRIM_TRIFAN_NOSTIPPLE))); + jmp = brw_JMPI(p, ip, ip, brw_imm_w(0)); + { + brw_push_insn_state(p); + brw_emit_tri_setup( c ); + brw_pop_insn_state(p); + /* note - thread killed in subroutine */ + } + brw_land_fwd_jump(p, jmp); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); + brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_LINELIST) | + (1<<_3DPRIM_LINESTRIP) | + (1<<_3DPRIM_LINELOOP) | + (1<<_3DPRIM_LINESTRIP_CONT) | + (1<<_3DPRIM_LINESTRIP_BF) | + (1<<_3DPRIM_LINESTRIP_CONT_BF))); + jmp = brw_JMPI(p, ip, ip, brw_imm_w(0)); + { + brw_push_insn_state(p); + brw_emit_line_setup( c ); + brw_pop_insn_state(p); + /* note - thread killed in subroutine */ + } + brw_land_fwd_jump(p, jmp); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); + brw_AND(p, v1_null_ud, payload_attr, brw_imm_ud(1< + */ + + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + +#if 0 +static void upload_sf_vp(struct brw_context *brw) +{ + GLcontext *ctx = &brw->intel.ctx; + struct brw_sf_viewport sfv; + struct intel_renderbuffer *irb = + intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); + float y_scale, y_bias; + int x, y, w, h, x1, x2, y1, y2; + int draw_h = ctx->DrawBuffer->Height; + + memset(&sfv, 0, sizeof(sfv)); + + if (ctx->DrawBuffer->Name) { + /* User-created FBO */ + if (irb && !irb->RenderToTexture) { + y_scale = -1.0; + y_bias = draw_h; + } else { + y_scale = 1.0; + y_bias = 0; + } + } else { + if (brw->intel.driDrawable != NULL) { + y_scale = -1.0; + y_bias = draw_h; + } else { + y_scale = 1.0; + y_bias = 0; + } + } + + /* _NEW_VIEWPORT, BRW_NEW_METAOPS */ + + if (!brw->metaops.active) { + const float *v = brw->intel.ctx.Viewport._WindowMap.m; + + sfv.viewport.m00 = v[MAT_SX]; + sfv.viewport.m11 = v[MAT_SY] * y_scale; + sfv.viewport.m22 = v[MAT_SZ] * brw->intel.depth_scale; + sfv.viewport.m30 = v[MAT_TX]; + sfv.viewport.m31 = v[MAT_TY] * y_scale + y_bias; + sfv.viewport.m32 = v[MAT_TZ] * brw->intel.depth_scale; + } else { + sfv.viewport.m00 = 1; + sfv.viewport.m11 = - 1; + sfv.viewport.m22 = 1; + sfv.viewport.m30 = 0; + sfv.viewport.m31 = brw->intel.driDrawable->h; + sfv.viewport.m32 = 0; + } + + /* _NEW_SCISSOR */ + x = brw->attribs.Scissor->X; + y = brw->attribs.Scissor->Y; + w = brw->attribs.Scissor->Width; + h = brw->attribs.Scissor->Height; + + if (ctx->DrawBuffer->Name == 0) { + x1 = x; + y1 = draw_h - (y + h); + x2 = x + w - 1; + y2 = y1 + h - 1; + } else { + /* FBO has non-inverted coords. */ + x1 = x; + y1 = y; + x2 = x + w - 1; + y2 = y + h - 1; + } + + sfv.scissor.xmin = CLAMP(x1, 0, ctx->DrawBuffer->Width - 1); + sfv.scissor.xmax = CLAMP(y1, 0, ctx->DrawBuffer->Height - 1); + sfv.scissor.ymin = CLAMP(x2, 0, ctx->DrawBuffer->Width - 1); + sfv.scissor.ymax = CLAMP(y2, 0, ctx->DrawBuffer->Height - 1); + + brw->sf.vp_gs_offset = brw_cache_data( &brw->cache[BRW_SF_VP], &sfv ); +} + +const struct brw_tracked_state brw_sf_vp = { + .dirty = { + .mesa = (_NEW_VIEWPORT | + _NEW_SCISSOR), + .brw = BRW_NEW_METAOPS, + .cache = 0 + }, + .update = upload_sf_vp +}; + +static void upload_sf_unit( struct brw_context *brw ) +{ + struct brw_sf_unit_state sf; + memset(&sf, 0, sizeof(sf)); + + /* CACHE_NEW_SF_PROG */ + sf.thread0.grf_reg_count = ALIGN(brw->sf.prog_data->total_grf, 16) / 16 - 1; + sf.thread0.kernel_start_pointer = brw->sf.prog_gs_offset >> 6; + sf.thread3.urb_entry_read_length = brw->sf.prog_data->urb_read_length; + + sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + sf.thread3.dispatch_grf_start_reg = 3; + sf.thread3.urb_entry_read_offset = 1; + + /* BRW_NEW_URB_FENCE */ + sf.thread4.nr_urb_entries = brw->urb.nr_sf_entries; + sf.thread4.urb_entry_allocation_size = brw->urb.sfsize - 1; + sf.thread4.max_threads = MIN2(12, brw->urb.nr_sf_entries / 2) - 1; + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + sf.thread4.max_threads = 0; + + if (BRW_DEBUG & DEBUG_STATS) + sf.thread4.stats_enable = 1; + + /* CACHE_NEW_SF_VP */ + sf.sf5.sf_viewport_state_offset = brw->sf.vp_gs_offset >> 5; + + sf.sf5.viewport_transform = 1; + + /* _NEW_SCISSOR */ + if (brw->attribs.Scissor->Enabled) + sf.sf6.scissor = 1; + + /* _NEW_POLYGON */ + if (brw->attribs.Polygon->FrontFace == GL_CCW) + sf.sf5.front_winding = BRW_FRONTWINDING_CCW; + else + sf.sf5.front_winding = BRW_FRONTWINDING_CW; + + if (brw->attribs.Polygon->CullFlag) { + switch (brw->attribs.Polygon->CullFaceMode) { + case GL_FRONT: + sf.sf6.cull_mode = BRW_CULLMODE_FRONT; + break; + case GL_BACK: + sf.sf6.cull_mode = BRW_CULLMODE_BACK; + break; + case GL_FRONT_AND_BACK: + sf.sf6.cull_mode = BRW_CULLMODE_BOTH; + break; + default: + assert(0); + break; + } + } + else + sf.sf6.cull_mode = BRW_CULLMODE_NONE; + + + /* _NEW_LINE */ + /* XXX use ctx->Const.Min/MaxLineWidth here */ + sf.sf6.line_width = CLAMP(brw->attribs.Line->Width, 1.0, 5.0) * (1<<1); + + sf.sf6.line_endcap_aa_region_width = 1; + if (brw->attribs.Line->SmoothFlag) + sf.sf6.aa_enable = 1; + else if (sf.sf6.line_width <= 0x2) + sf.sf6.line_width = 0; + + /* _NEW_POINT */ + sf.sf6.point_rast_rule = 1; /* opengl conventions */ + /* XXX clamp max depends on AA vs. non-AA */ + + sf.sf7.sprite_point = brw->attribs.Point->PointSprite; + sf.sf7.point_size = CLAMP(brw->attribs.Point->Size, 1.0, 255.0) * (1<<3); + sf.sf7.use_point_size_state = !brw->attribs.Point->_Attenuated; + + /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons: + */ + sf.sf7.trifan_pv = 2; + sf.sf7.linestrip_pv = 1; + sf.sf7.tristrip_pv = 2; + sf.sf7.line_last_pixel_enable = 0; + + /* Set bias for OpenGL rasterization rules: + */ + sf.sf6.dest_org_vbias = 0x8; + sf.sf6.dest_org_hbias = 0x8; + + brw->sf.state_gs_offset = brw_cache_data( &brw->cache[BRW_SF_UNIT], &sf ); +} + + +const struct brw_tracked_state brw_sf_unit = { + .dirty = { + .mesa = (_NEW_POLYGON | + _NEW_LINE | + _NEW_POINT | + _NEW_SCISSOR), + .brw = (BRW_NEW_URB_FENCE | + BRW_NEW_METAOPS), + .cache = (CACHE_NEW_SF_VP | + CACHE_NEW_SF_PROG) + }, + .update = upload_sf_unit +}; + + +#endif diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c new file mode 100644 index 0000000000..b50f23c1a2 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -0,0 +1,466 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Zack Rusin + * Keith Whitwell + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_dump.h" + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_state.h" +#include "brw_draw.h" + + +#define DUP( TYPE, VAL ) \ +do { \ + struct TYPE *x = malloc(sizeof(*x)); \ + memcpy(x, VAL, sizeof(*x) ); \ + return x; \ +} while (0) + +/************************************************************************ + * Blend + */ +static void * +brw_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + DUP( pipe_blend_state, blend ); +} + +static void brw_bind_blend_state(struct pipe_context *pipe, + void *blend) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Blend = (struct pipe_blend_state*)blend; + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + + +static void brw_delete_blend_state(struct pipe_context *pipe, void *blend) +{ + free(blend); +} + +static void brw_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.BlendColor = *blend_color; + + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + +/************************************************************************ + * Sampler + */ + +static void * +brw_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + DUP( pipe_sampler_state, sampler ); +} + +static void brw_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Samplers[unit] = sampler; + brw->state.dirty.brw |= BRW_NEW_SAMPLER; +} + +static void brw_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + free(sampler); +} + + +static void brw_set_sampler_units(struct pipe_context *pipe, + uint numSamplers, const uint *units) +{ +} + +/************************************************************************ + * Depth stencil + */ + +static void * +brw_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_state *depth_stencil) +{ + DUP( pipe_depth_stencil_state, depth_stencil ); +} + +static void brw_bind_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.DepthStencil = (const struct pipe_depth_stencil_state *)depth_stencil; + + brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; +} + +static void brw_delete_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + free(depth_stencil); +} + +/************************************************************************ + * Alpha test + */ +static void * +brw_create_alpha_test_state(struct pipe_context *pipe, + const struct pipe_alpha_test_state *alpha_test) +{ + DUP(pipe_alpha_test_state, alpha_test); +} + +static void brw_bind_alpha_test_state(struct pipe_context *pipe, + void *alpha) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.AlphaTest = (const struct pipe_alpha_test_state*)alpha; + + brw->state.dirty.brw |= BRW_NEW_ALPHA_TEST; +} + +static void brw_delete_alpha_test_state(struct pipe_context *pipe, + void *alpha) +{ + free(alpha); +} + +/************************************************************************ + * Scissor + */ +static void brw_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct brw_context *brw = brw_context(pipe); + + memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) ); + brw->state.dirty.brw |= BRW_NEW_SCISSOR; +} + + +/************************************************************************ + * Stipple + */ + +static void brw_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ +} + + +/************************************************************************ + * Fragment shader + */ + +static void * brw_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); + + /* XXX: Do I have to duplicate the tokens as well?? + */ + brw_fp->program = *shader; + + return (void *)brw_fp; +} + +static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader; + brw->state.dirty.brw |= BRW_NEW_FS; +} + +static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) +{ + FREE(shader); +} + + +/************************************************************************ + * Vertex shader and other TNL state + */ + +static void *brw_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); + + /* XXX: Do I have to duplicate the tokens as well?? + */ + brw_vp->program = *shader; + + tgsi_dump(shader->tokens, 0); + + return (void *)brw_vp; +} + +static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; + brw->state.dirty.brw |= BRW_NEW_VS; + + printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); +} + +static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) +{ + FREE(shader); +} + + +static void brw_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Clip = *clip; +} + + +static void brw_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Viewport = *viewport; /* struct copy */ + brw->state.dirty.brw |= BRW_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + //draw_set_viewport_state(brw->draw, viewport); +} + + +static void brw_set_vertex_buffer( struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_buffer *buffer ) +{ + struct brw_context *brw = brw_context(pipe); + brw->vb.vbo_array[index] = *buffer; + if (index > brw->vb.last_vb) + brw->vb.last_vb = index; + assert(brw->vb.last_vb < BRW_VEP_MAX); +} + +static void brw_set_vertex_element(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_element *element) +{ + /* flush ? */ + struct brw_context *brw = brw_context(pipe); + + assert(index < PIPE_ATTRIB_MAX); + struct brw_vertex_element el; + memset(&el, 0, sizeof(struct brw_vertex_element)); + + /* do we need those anymore?*/ + el.index = index; +#if 0 + /*FIXME*/ + el.element_size = 0; + el.count = 0; + el.vbo_rebase_offset = 0; +#endif + + el.vep.ve0.src_offset = element->src_offset; + el.vep.ve0.src_format = brw_translate_surface_format(element->src_format); + el.vep.ve0.valid = 1; + el.vep.ve0.vertex_buffer_index = element->vertex_buffer_index; + + el.vep.ve1.dst_offset = element->dst_offset; + el.vep.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; + el.vep.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; + el.vep.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; + el.vep.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; + /*can we count of brw->vb.vbo_array[element->vertex_buffer_index] + * being initialized ok to actually compute vbcomponent's + * correctly? */ + + brw->vb.inputs[index] = el; +} + + + +/************************************************************************ + * Constant buffers + */ + +static void brw_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct brw_context *brw = brw_context(pipe); + + assert(buf == 0 || index == 0); + + brw->attribs.Constants[shader] = buf; + brw->state.dirty.brw |= BRW_NEW_CONSTANTS; +} + + +/************************************************************************ + * Texture surfaces + */ + + +static void brw_set_texture_state(struct pipe_context *pipe, + unsigned unit, + struct pipe_texture *texture) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Texture[unit] = (struct brw_texture*)texture; /* ptr, not struct */ + + brw->state.dirty.brw |= BRW_NEW_TEXTURE; +} + + +/************************************************************************ + * Render targets, etc + */ + +static void brw_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FrameBuffer = *fb; /* struct copy */ + + brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER; +} + + + +static void brw_set_clear_color_state(struct pipe_context *pipe, + const struct pipe_clear_color_state *clear) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.ClearColor = *clear; /* struct copy */ +} + +/************************************************************************ + * Rasterizer state + */ + +static void * +brw_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *rasterizer) +{ + DUP(pipe_rasterizer_state, rasterizer); +} + +static void brw_bind_rasterizer_state( struct pipe_context *pipe, + void *setup ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Raster = (struct pipe_rasterizer_state *)setup; + + /* Also pass-through to draw module: + */ + //draw_set_rasterizer_state(brw->draw, setup); + + brw->state.dirty.brw |= BRW_NEW_RASTERIZER; +} + +static void brw_delete_rasterizer_state(struct pipe_context *pipe, + void *setup) +{ + free(setup); +} + + + +void +brw_init_state_functions( struct brw_context *brw ) +{ + brw->pipe.create_alpha_test_state = brw_create_alpha_test_state; + brw->pipe.bind_alpha_test_state = brw_bind_alpha_test_state; + brw->pipe.delete_alpha_test_state = brw_delete_alpha_test_state; + + brw->pipe.create_blend_state = brw_create_blend_state; + brw->pipe.bind_blend_state = brw_bind_blend_state; + brw->pipe.delete_blend_state = brw_delete_blend_state; + + brw->pipe.create_sampler_state = brw_create_sampler_state; + brw->pipe.bind_sampler_state = brw_bind_sampler_state; + brw->pipe.delete_sampler_state = brw_delete_sampler_state; + + brw->pipe.create_depth_stencil_state = brw_create_depth_stencil_state; + brw->pipe.bind_depth_stencil_state = brw_bind_depth_stencil_state; + brw->pipe.delete_depth_stencil_state = brw_delete_depth_stencil_state; + + brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; + brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; + brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state; + brw->pipe.create_fs_state = brw_create_fs_state; + brw->pipe.bind_fs_state = brw_bind_fs_state; + brw->pipe.delete_fs_state = brw_delete_fs_state; + brw->pipe.create_vs_state = brw_create_vs_state; + brw->pipe.bind_vs_state = brw_bind_vs_state; + brw->pipe.delete_vs_state = brw_delete_vs_state; + + brw->pipe.set_blend_color = brw_set_blend_color; + brw->pipe.set_clip_state = brw_set_clip_state; + brw->pipe.set_clear_color_state = brw_set_clear_color_state; + brw->pipe.set_constant_buffer = brw_set_constant_buffer; + brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; + +// brw->pipe.set_feedback_state = brw_set_feedback_state; +// brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; + + brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; + brw->pipe.set_sampler_units = brw_set_sampler_units; + brw->pipe.set_scissor_state = brw_set_scissor_state; + brw->pipe.set_texture_state = brw_set_texture_state; + brw->pipe.set_viewport_state = brw_set_viewport_state; + brw->pipe.set_vertex_buffer = brw_set_vertex_buffer; + brw->pipe.set_vertex_element = brw_set_vertex_element; +} diff --git a/src/mesa/pipe/i965simple/brw_state.h b/src/mesa/pipe/i965simple/brw_state.h new file mode 100644 index 0000000000..4dabfe8082 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_state.h @@ -0,0 +1,157 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_STATE_H +#define BRW_STATE_H + +#include "brw_context.h" +#include "brw_winsys.h" + + +const struct brw_tracked_state brw_blend_constant_color; +const struct brw_tracked_state brw_cc_unit; +const struct brw_tracked_state brw_cc_vp; +const struct brw_tracked_state brw_clip_prog; +const struct brw_tracked_state brw_clip_unit; +const struct brw_tracked_state brw_constant_buffer_state; +const struct brw_tracked_state brw_constant_buffer; +const struct brw_tracked_state brw_curbe_offsets; +const struct brw_tracked_state brw_invarient_state; +const struct brw_tracked_state brw_gs_prog; +const struct brw_tracked_state brw_gs_unit; +const struct brw_tracked_state brw_drawing_rect; +const struct brw_tracked_state brw_line_stipple; +const struct brw_tracked_state brw_pipelined_state_pointers; +const struct brw_tracked_state brw_binding_table_pointers; +const struct brw_tracked_state brw_depthbuffer; +const struct brw_tracked_state brw_polygon_stipple_offset; +const struct brw_tracked_state brw_polygon_stipple; +const struct brw_tracked_state brw_program_parameters; +const struct brw_tracked_state brw_recalculate_urb_fence; +const struct brw_tracked_state brw_sf_prog; +const struct brw_tracked_state brw_sf_unit; +const struct brw_tracked_state brw_sf_vp; +const struct brw_tracked_state brw_state_base_address; +const struct brw_tracked_state brw_urb_fence; +const struct brw_tracked_state brw_vertex_state; +const struct brw_tracked_state brw_vs_prog; +const struct brw_tracked_state brw_vs_unit; +const struct brw_tracked_state brw_wm_input_sizes; +const struct brw_tracked_state brw_wm_prog; +const struct brw_tracked_state brw_wm_samplers; +const struct brw_tracked_state brw_wm_surfaces; +const struct brw_tracked_state brw_wm_unit; + +const struct brw_tracked_state brw_psp_urb_cbs; + +const struct brw_tracked_state brw_active_vertprog; +const struct brw_tracked_state brw_tnl_vertprog; +const struct brw_tracked_state brw_pipe_control; + +const struct brw_tracked_state brw_clear_surface_cache; +const struct brw_tracked_state brw_clear_batch_cache; + +/*********************************************************************** + * brw_state_cache.c + */ +unsigned brw_cache_data(struct brw_cache *cache, + const void *data ); + +unsigned brw_cache_data_sz(struct brw_cache *cache, + const void *data, + unsigned data_sz); + +unsigned brw_upload_cache( struct brw_cache *cache, + const void *key, + unsigned key_sz, + const void *data, + unsigned data_sz, + const void *aux, + void *aux_return ); + +boolean brw_search_cache( struct brw_cache *cache, + const void *key, + unsigned key_size, + void *aux_return, + unsigned *offset_return); + +void brw_init_caches( struct brw_context *brw ); +void brw_destroy_caches( struct brw_context *brw ); + +static inline struct pipe_buffer_handle *brw_cache_buffer(struct brw_context *brw, + enum brw_cache_id id) +{ + return brw->cache[id].pool->buffer; +} + +/*********************************************************************** + * brw_state_batch.c + */ +#define BRW_BATCH_STRUCT(brw, s) brw_batchbuffer_data( brw->winsys, (s), sizeof(*(s))) +#define BRW_CACHED_BATCH_STRUCT(brw, s) brw_cached_batch_struct( brw, (s), sizeof(*(s)) ) + +boolean brw_cached_batch_struct( struct brw_context *brw, + const void *data, + unsigned sz ); + +void brw_destroy_batch_cache( struct brw_context *brw ); + + +/*********************************************************************** + * brw_state_pool.c + */ +void brw_init_pools( struct brw_context *brw ); +void brw_destroy_pools( struct brw_context *brw ); + +boolean brw_pool_alloc( struct brw_mem_pool *pool, + unsigned size, + unsigned alignment, + unsigned *offset_return); + +void brw_pool_fence( struct brw_context *brw, + struct brw_mem_pool *pool, + unsigned fence ); + + +void brw_pool_check_wrap( struct brw_context *brw, + struct brw_mem_pool *pool ); + +void brw_clear_all_caches( struct brw_context *brw ); +void brw_invalidate_pools( struct brw_context *brw ); +void brw_clear_batch_cache_flush( struct brw_context *brw ); + +void brw_upload_cc_unit(struct brw_context *brw); +void brw_upload_clip_prog(struct brw_context *brw); +void brw_upload_blend_constant_color(struct brw_context *brw); +void brw_upload_wm_samplers(struct brw_context *brw); + +#endif diff --git a/src/mesa/pipe/i965simple/brw_state_batch.c b/src/mesa/pipe/i965simple/brw_state_batch.c new file mode 100644 index 0000000000..35db76b594 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_state_batch.c @@ -0,0 +1,113 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_state.h" +#include "brw_winsys.h" + +#include "pipe/p_util.h" + +/* A facility similar to the data caching code above, which aims to + * prevent identical commands being issued repeatedly. + */ +boolean brw_cached_batch_struct( struct brw_context *brw, + const void *data, + unsigned sz ) +{ + struct brw_cached_batch_item *item = brw->cached_batch_items; + struct header *newheader = (struct header *)data; + + if (brw->emit_state_always) { + brw_batchbuffer_data(brw->winsys, data, sz); + return TRUE; + } + + while (item) { + if (item->header->opcode == newheader->opcode) { + if (item->sz == sz && memcmp(item->header, newheader, sz) == 0) + return FALSE; + if (item->sz != sz) { + FREE(item->header); + item->header = MALLOC(sz); + item->sz = sz; + } + goto emit; + } + item = item->next; + } + + assert(!item); + item = CALLOC_STRUCT(brw_cached_batch_item); + item->header = MALLOC(sz); + item->sz = sz; + item->next = brw->cached_batch_items; + brw->cached_batch_items = item; + +emit: + memcpy(item->header, newheader, sz); + brw_batchbuffer_data(brw->winsys, data, sz); + return TRUE; +} + +static void clear_batch_cache( struct brw_context *brw ) +{ + struct brw_cached_batch_item *item = brw->cached_batch_items; + + while (item) { + struct brw_cached_batch_item *next = item->next; + free((void *)item->header); + free(item); + item = next; + } + + brw->cached_batch_items = NULL; + + + brw_clear_all_caches(brw); + + brw_invalidate_pools(brw); +} + +void brw_clear_batch_cache_flush( struct brw_context *brw ) +{ + clear_batch_cache(brw); + +/* brw_do_flush(brw, BRW_FLUSH_STATE_CACHE|BRW_FLUSH_READ_CACHE); */ + + brw->state.dirty.brw |= ~0; + brw->state.dirty.cache |= ~0; +} + + + +void brw_destroy_batch_cache( struct brw_context *brw ) +{ + clear_batch_cache(brw); +} diff --git a/src/mesa/pipe/i965simple/brw_state_cache.c b/src/mesa/pipe/i965simple/brw_state_cache.c new file mode 100644 index 0000000000..13e262d2e5 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_state_cache.c @@ -0,0 +1,442 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_state.h" + +#include "brw_wm.h" +#include "brw_vs.h" +#include "brw_clip.h" +#include "brw_sf.h" +#include "brw_gs.h" + +#include "pipe/p_util.h" + + + +/*********************************************************************** + * Check cache for uploaded version of struct, else upload new one. + * Fail when memory is exhausted. + * + * XXX: FIXME: Currently search is so slow it would be quicker to + * regenerate the data every time... + */ + +static unsigned hash_key( const void *key, unsigned key_size ) +{ + unsigned *ikey = (unsigned *)key; + unsigned hash = 0, i; + + assert(key_size % 4 == 0); + + /* I'm sure this can be improved on: + */ + for (i = 0; i < key_size/4; i++) + hash ^= ikey[i]; + + return hash; +} + +static struct brw_cache_item *search_cache( struct brw_cache *cache, + unsigned hash, + const void *key, + unsigned key_size) +{ + struct brw_cache_item *c; + + for (c = cache->items[hash % cache->size]; c; c = c->next) { + if (c->hash == hash && + c->key_size == key_size && + memcmp(c->key, key, key_size) == 0) + return c; + } + + return NULL; +} + + +static void rehash( struct brw_cache *cache ) +{ + struct brw_cache_item **items; + struct brw_cache_item *c, *next; + unsigned size, i; + + size = cache->size * 3; + items = (struct brw_cache_item**) MALLOC(size * sizeof(*items)); + memset(items, 0, size * sizeof(*items)); + + for (i = 0; i < cache->size; i++) + for (c = cache->items[i]; c; c = next) { + next = c->next; + c->next = items[c->hash % size]; + items[c->hash % size] = c; + } + + FREE(cache->items); + cache->items = items; + cache->size = size; +} + + +boolean brw_search_cache( struct brw_cache *cache, + const void *key, + unsigned key_size, + void *aux_return, + unsigned *offset_return) +{ + struct brw_cache_item *item; + unsigned addr = 0; + unsigned hash = hash_key(key, key_size); + + item = search_cache(cache, hash, key, key_size); + + if (item) { + if (aux_return) + *(void **)aux_return = (void *)((char *)item->key + item->key_size); + + *offset_return = addr = item->offset; + } + + if (item == NULL || addr != cache->last_addr) { + cache->brw->state.dirty.cache |= 1<id; + cache->last_addr = addr; + } + + return item != NULL; +} + +unsigned brw_upload_cache( struct brw_cache *cache, + const void *key, + unsigned key_size, + const void *data, + unsigned data_size, + const void *aux, + void *aux_return ) +{ + unsigned offset; + struct brw_cache_item *item = CALLOC_STRUCT(brw_cache_item); + unsigned hash = hash_key(key, key_size); + void *tmp = MALLOC(key_size + cache->aux_size); + + if (!brw_pool_alloc(cache->pool, data_size, 1 << 6, &offset)) { + /* Should not be possible: + */ + printf("brw_pool_alloc failed\n"); + exit(1); + } + + memcpy(tmp, key, key_size); + + if (cache->aux_size) + memcpy(tmp+key_size, aux, cache->aux_size); + + item->key = tmp; + item->hash = hash; + item->key_size = key_size; + item->offset = offset; + item->data_size = data_size; + + if (++cache->n_items > cache->size * 1.5) + rehash(cache); + + hash %= cache->size; + item->next = cache->items[hash]; + cache->items[hash] = item; + + if (aux_return) { + assert(cache->aux_size); + *(void **)aux_return = (void *)((char *)item->key + item->key_size); + } + + if (BRW_DEBUG & DEBUG_STATE) + printf("upload %s: %d bytes to pool buffer %p offset %x\n", + cache->name, data_size, + cache->pool->buffer, + offset); + + /* Copy data to the buffer: + */ + cache->brw->winsys->buffer_subdata_typed(cache->brw->winsys, + cache->pool->buffer, + offset, + data_size, + data, + cache->id); + + cache->brw->state.dirty.cache |= 1<id; + cache->last_addr = offset; + + return offset; +} + +/* This doesn't really work with aux data. Use search/upload instead + */ +unsigned brw_cache_data_sz(struct brw_cache *cache, + const void *data, + unsigned data_size) +{ + unsigned addr; + + if (!brw_search_cache(cache, data, data_size, NULL, &addr)) { + addr = brw_upload_cache(cache, + data, data_size, + data, data_size, + NULL, NULL); + } + + return addr; +} + +unsigned brw_cache_data(struct brw_cache *cache, + const void *data) +{ + return brw_cache_data_sz(cache, data, cache->key_size); +} + +enum pool_type { + DW_SURFACE_STATE, + DW_GENERAL_STATE +}; + +static void brw_init_cache( struct brw_context *brw, + const char *name, + unsigned id, + unsigned key_size, + unsigned aux_size, + enum pool_type pool_type) +{ + struct brw_cache *cache = &brw->cache[id]; + cache->brw = brw; + cache->id = id; + cache->name = name; + cache->items = NULL; + + cache->size = 7; + cache->n_items = 0; + cache->items = (struct brw_cache_item **) + CALLOC(cache->size, sizeof(struct brw_cache_item)); + + + cache->key_size = key_size; + cache->aux_size = aux_size; + switch (pool_type) { + case DW_GENERAL_STATE: cache->pool = &brw->pool[BRW_GS_POOL]; break; + case DW_SURFACE_STATE: cache->pool = &brw->pool[BRW_SS_POOL]; break; + default: assert(0); break; + } +} + +void brw_init_caches( struct brw_context *brw ) +{ + + brw_init_cache(brw, + "CC_VP", + BRW_CC_VP, + sizeof(struct brw_cc_viewport), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "CC_UNIT", + BRW_CC_UNIT, + sizeof(struct brw_cc_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "WM_PROG", + BRW_WM_PROG, + sizeof(struct brw_wm_prog_key), + sizeof(struct brw_wm_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SAMPLER_DEFAULT_COLOR", + BRW_SAMPLER_DEFAULT_COLOR, + sizeof(struct brw_sampler_default_color), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SAMPLER", + BRW_SAMPLER, + 0, /* variable key/data size */ + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "WM_UNIT", + BRW_WM_UNIT, + sizeof(struct brw_wm_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SF_PROG", + BRW_SF_PROG, + sizeof(struct brw_sf_prog_key), + sizeof(struct brw_sf_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SF_VP", + BRW_SF_VP, + sizeof(struct brw_sf_viewport), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SF_UNIT", + BRW_SF_UNIT, + sizeof(struct brw_sf_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "VS_UNIT", + BRW_VS_UNIT, + sizeof(struct brw_vs_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "VS_PROG", + BRW_VS_PROG, + sizeof(struct brw_vs_prog_key), + sizeof(struct brw_vs_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "CLIP_UNIT", + BRW_CLIP_UNIT, + sizeof(struct brw_clip_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "CLIP_PROG", + BRW_CLIP_PROG, + sizeof(struct brw_clip_prog_key), + sizeof(struct brw_clip_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "GS_UNIT", + BRW_GS_UNIT, + sizeof(struct brw_gs_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "GS_PROG", + BRW_GS_PROG, + sizeof(struct brw_gs_prog_key), + sizeof(struct brw_gs_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SS_SURFACE", + BRW_SS_SURFACE, + sizeof(struct brw_surface_state), + 0, + DW_SURFACE_STATE); + + brw_init_cache(brw, + "SS_SURF_BIND", + BRW_SS_SURF_BIND, + sizeof(struct brw_surface_binding_table), + 0, + DW_SURFACE_STATE); +} + + +/* When we lose hardware context, need to invalidate the surface cache + * as these structs must be explicitly re-uploaded. They are subject + * to fixup by the memory manager as they contain absolute agp + * offsets, so we need to ensure there is a fresh version of the + * struct available to receive the fixup. + * + * XXX: Need to ensure that there aren't two versions of a surface or + * bufferobj with different backing data active in the same buffer at + * once? Otherwise the cache could confuse them. Maybe better not to + * cache at all? + * + * --> Isn't this the same as saying need to ensure batch is flushed + * before new data is uploaded to an existing buffer? We + * already try to make sure of that. + */ +static void clear_cache( struct brw_cache *cache ) +{ + struct brw_cache_item *c, *next; + unsigned i; + + for (i = 0; i < cache->size; i++) { + for (c = cache->items[i]; c; c = next) { + next = c->next; + free((void *)c->key); + free(c); + } + cache->items[i] = NULL; + } + + cache->n_items = 0; +} + +void brw_clear_all_caches( struct brw_context *brw ) +{ + int i; + + if (BRW_DEBUG & DEBUG_STATE) + fprintf(stderr, "%s\n", __FUNCTION__); + + for (i = 0; i < BRW_MAX_CACHE; i++) + clear_cache(&brw->cache[i]); + + if (brw->curbe.last_buf) { + FREE(brw->curbe.last_buf); + brw->curbe.last_buf = NULL; + } + + brw->state.dirty.brw |= ~0; + brw->state.dirty.cache |= ~0; +} + + + + + +void brw_destroy_caches( struct brw_context *brw ) +{ + unsigned i; + + for (i = 0; i < BRW_MAX_CACHE; i++) + clear_cache(&brw->cache[i]); +} diff --git a/src/mesa/pipe/i965simple/brw_state_pool.c b/src/mesa/pipe/i965simple/brw_state_pool.c new file mode 100644 index 0000000000..a490049024 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_state_pool.c @@ -0,0 +1,139 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +/** @file brw_state_pool.c + * Implements the state pool allocator. + * + * For the 965, we create two state pools for state cache entries. Objects + * will be allocated into the pools depending on which state base address + * their pointer is relative to in other 965 state. + * + * The state pools are relatively simple: As objects are allocated, increment + * the offset to allocate space. When the pool is "full" (rather, close to + * full), we reset the pool and reset the state cache entries that point into + * the pool. + */ + +#include "pipe/p_winsys.h" +#include "brw_context.h" +#include "brw_state.h" + +boolean brw_pool_alloc( struct brw_mem_pool *pool, + unsigned size, + unsigned align, + unsigned *offset_return) +{ + unsigned fixup = ALIGN(pool->offset, align) - pool->offset; + + size = ALIGN(size, 4); + + if (pool->offset + fixup + size >= pool->size) { + printf("%s failed\n", __FUNCTION__); + assert(0); + exit(0); + } + + pool->offset += fixup; + *offset_return = pool->offset; + pool->offset += size; + + return TRUE; +} + +static +void brw_invalidate_pool( struct brw_mem_pool *pool ) +{ + if (BRW_DEBUG & DEBUG_STATE) + printf("\n\n\n %s \n\n\n", __FUNCTION__); + + pool->offset = 0; + + brw_clear_all_caches(pool->brw); +} + + +static void brw_init_pool( struct brw_context *brw, + unsigned pool_id, + unsigned size ) +{ + struct brw_mem_pool *pool = &brw->pool[pool_id]; + + pool->size = size; + pool->brw = brw; + + pool->buffer = brw->pipe.winsys->buffer_create(brw->pipe.winsys, + 4096, 0, 0); + + brw->pipe.winsys->buffer_data(brw->pipe.winsys, + pool->buffer, + size, + NULL, + 0 /* DRM_BO_FLAG_MEM_TT */); +} + +static void brw_destroy_pool( struct brw_context *brw, + unsigned pool_id ) +{ + struct brw_mem_pool *pool = &brw->pool[pool_id]; + + pool->brw->pipe.winsys->buffer_reference( pool->brw->pipe.winsys, + &pool->buffer, + NULL ); +} + + +void brw_pool_check_wrap( struct brw_context *brw, + struct brw_mem_pool *pool ) +{ + if (pool->offset > (pool->size * 3) / 4) { + brw->state.dirty.brw |= BRW_NEW_CONTEXT; + } + +} + +void brw_init_pools( struct brw_context *brw ) +{ + brw_init_pool(brw, BRW_GS_POOL, 0x80000); + brw_init_pool(brw, BRW_SS_POOL, 0x80000); +} + +void brw_destroy_pools( struct brw_context *brw ) +{ + brw_destroy_pool(brw, BRW_GS_POOL); + brw_destroy_pool(brw, BRW_SS_POOL); +} + + +void brw_invalidate_pools( struct brw_context *brw ) +{ + brw_invalidate_pool(&brw->pool[BRW_GS_POOL]); + brw_invalidate_pool(&brw->pool[BRW_SS_POOL]); +} diff --git a/src/mesa/pipe/i965simple/brw_state_upload.c b/src/mesa/pipe/i965simple/brw_state_upload.c new file mode 100644 index 0000000000..1ca7484958 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_state_upload.c @@ -0,0 +1,231 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_state.h" + +#include "pipe/p_util.h" + +/* This is used to initialize brw->state.atoms[]. We could use this + * list directly except for a single atom, brw_constant_buffer, which + * has a .dirty value which changes according to the parameters of the + * current fragment and vertex programs, and so cannot be a static + * value. + */ +const struct brw_tracked_state *atoms[] = +{ + &brw_wm_input_sizes, + &brw_vs_prog, + &brw_gs_prog, + &brw_clip_prog, + &brw_sf_prog, + &brw_wm_prog, + + /* Once all the programs are done, we know how large urb entry + * sizes need to be and can decide if we need to change the urb + * layout. + */ + &brw_curbe_offsets, + &brw_recalculate_urb_fence, + + + &brw_cc_vp, + &brw_cc_unit, + + &brw_wm_surfaces, /* must do before samplers */ + &brw_wm_samplers, + + &brw_wm_unit, + &brw_sf_vp, + &brw_sf_unit, + &brw_vs_unit, /* always required, enabled or not */ + &brw_clip_unit, + &brw_gs_unit, + + /* Command packets: + */ + &brw_invarient_state, + &brw_state_base_address, + &brw_pipe_control, + + &brw_binding_table_pointers, + &brw_blend_constant_color, + + &brw_drawing_rect, + &brw_depthbuffer, + + &brw_polygon_stipple, + &brw_polygon_stipple_offset, + + &brw_line_stipple, + + &brw_psp_urb_cbs, + + + NULL, /* brw_constant_buffer */ +}; + + +void brw_init_state( struct brw_context *brw ) +{ + unsigned i; + + brw_init_pools(brw); + brw_init_caches(brw); + + brw->state.atoms = MALLOC(sizeof(atoms)); + brw->state.nr_atoms = sizeof(atoms)/sizeof(*atoms); + memcpy(brw->state.atoms, atoms, sizeof(atoms)); + + /* Patch in a pointer to the dynamic state atom: + */ + for (i = 0; i < brw->state.nr_atoms; i++) + if (brw->state.atoms[i] == NULL) + brw->state.atoms[i] = &brw->curbe.tracked_state; + + memcpy(&brw->curbe.tracked_state, + &brw_constant_buffer, + sizeof(brw_constant_buffer)); + + brw->state.dirty.brw = ~0; + brw->emit_state_always = 0; + + +} + + +void brw_destroy_state( struct brw_context *brw ) +{ + if (brw->state.atoms) { + FREE(brw->state.atoms); + brw->state.atoms = NULL; + } + + brw_destroy_caches(brw); + brw_destroy_batch_cache(brw); + brw_destroy_pools(brw); +} + +/*********************************************************************** + */ + +static boolean check_state( const struct brw_state_flags *a, + const struct brw_state_flags *b ) +{ + return ((a->brw & b->brw) || + (a->cache & b->cache)); +} + +static void accumulate_state( struct brw_state_flags *a, + const struct brw_state_flags *b ) +{ + a->brw |= b->brw; + a->cache |= b->cache; +} + + +static void xor_states( struct brw_state_flags *result, + const struct brw_state_flags *a, + const struct brw_state_flags *b ) +{ + result->brw = a->brw ^ b->brw; + result->cache = a->cache ^ b->cache; +} + + +/*********************************************************************** + * Emit all state: + */ +void brw_validate_state( struct brw_context *brw ) +{ + struct brw_state_flags *state = &brw->state.dirty; + unsigned i; + + if (brw->emit_state_always) + state->brw |= ~0; + + if (state->cache == 0 && + state->brw == 0) + return; + + if (brw->state.dirty.brw & BRW_NEW_CONTEXT) + brw_clear_batch_cache_flush(brw); + + if (BRW_DEBUG) { + /* Debug version which enforces various sanity checks on the + * state flags which are generated and checked to help ensure + * state atoms are ordered correctly in the list. + */ + struct brw_state_flags examined, prev; + memset(&examined, 0, sizeof(examined)); + prev = *state; + + for (i = 0; i < brw->state.nr_atoms; i++) { + const struct brw_tracked_state *atom = brw->state.atoms[i]; + struct brw_state_flags generated; + + assert(atom->dirty.brw || + atom->dirty.cache); + assert(atom->update); + + if (check_state(state, &atom->dirty) || atom->always_update) { + atom->update( brw ); + +/* emit_foo(brw); */ + } + if (atom->emit_reloc != NULL) + atom->emit_reloc(brw); + + accumulate_state(&examined, &atom->dirty); + + /* generated = (prev ^ state) + * if (examined & generated) + * fail; + */ + xor_states(&generated, &prev, state); + assert(!check_state(&examined, &generated)); + prev = *state; + } + } + else { + for (i = 0; i < Elements(atoms); i++) { + const struct brw_tracked_state *atom = brw->state.atoms[i]; + + if (check_state(state, &atom->dirty) || atom->always_update) + atom->update( brw ); + if (atom->emit_reloc != NULL) + atom->emit_reloc(brw); + } + } + + memset(state, 0, sizeof(*state)); +} diff --git a/src/mesa/pipe/i965simple/brw_strings.c b/src/mesa/pipe/i965simple/brw_strings.c new file mode 100644 index 0000000000..29a41ed1e9 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_strings.c @@ -0,0 +1,72 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "brw_context.h" +#include "brw_reg.h" + + +static const char *brw_get_vendor( struct pipe_context *pipe ) +{ + return "Tungsten Graphics, Inc."; +} + + +static const char *brw_get_name( struct pipe_context *pipe ) +{ + static char buffer[128]; + const char *chipset; + + switch (brw_context(pipe)->pci_id) { + case PCI_CHIP_I965_Q: + chipset = "Intel(R) 965Q"; + break; + case PCI_CHIP_I965_G: + case PCI_CHIP_I965_G_1: + chipset = "Intel(R) 965G"; + break; + case PCI_CHIP_I965_GM: + chipset = "Intel(R) 965GM"; + break; + case PCI_CHIP_I965_GME: + chipset = "Intel(R) 965GME/GLE"; + break; + default: + chipset = "unknown"; + break; + } + + sprintf(buffer, "pipe/i965 (chipset: %s)", chipset); + return buffer; +} + + +void +brw_init_string_functions(struct brw_context *brw) +{ + brw->pipe.get_name = brw_get_name; + brw->pipe.get_vendor = brw_get_vendor; +} diff --git a/src/mesa/pipe/i965simple/brw_structs.h b/src/mesa/pipe/i965simple/brw_structs.h new file mode 100644 index 0000000000..bbb087e95d --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_structs.h @@ -0,0 +1,1348 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_STRUCTS_H +#define BRW_STRUCTS_H + +#include "pipe/p_compiler.h" + +/* Command packets: + */ +struct header +{ + unsigned length:16; + unsigned opcode:16; +}; + + +union header_union +{ + struct header bits; + unsigned dword; +}; + +struct brw_3d_control +{ + struct + { + unsigned length:8; + unsigned notify_enable:1; + unsigned pad:3; + unsigned wc_flush_enable:1; + unsigned depth_stall_enable:1; + unsigned operation:2; + unsigned opcode:16; + } header; + + struct + { + unsigned pad:2; + unsigned dest_addr_type:1; + unsigned dest_addr:29; + } dest; + + unsigned dword2; + unsigned dword3; +}; + + +struct brw_3d_primitive +{ + struct + { + unsigned length:8; + unsigned pad:2; + unsigned topology:5; + unsigned indexed:1; + unsigned opcode:16; + } header; + + unsigned verts_per_instance; + unsigned start_vert_location; + unsigned instance_count; + unsigned start_instance_location; + unsigned base_vert_location; +}; + +/* These seem to be passed around as function args, so it works out + * better to keep them as #defines: + */ +#define BRW_FLUSH_READ_CACHE 0x1 +#define BRW_FLUSH_STATE_CACHE 0x2 +#define BRW_INHIBIT_FLUSH_RENDER_CACHE 0x4 +#define BRW_FLUSH_SNAPSHOT_COUNTERS 0x8 + +struct brw_mi_flush +{ + unsigned flags:4; + unsigned pad:12; + unsigned opcode:16; +}; + +struct brw_vf_statistics +{ + unsigned statistics_enable:1; + unsigned pad:15; + unsigned opcode:16; +}; + + + +struct brw_binding_table_pointers +{ + struct header header; + unsigned vs; + unsigned gs; + unsigned clp; + unsigned sf; + unsigned wm; +}; + + +struct brw_blend_constant_color +{ + struct header header; + float blend_constant_color[4]; +}; + + +struct brw_depthbuffer +{ + union header_union header; + + union { + struct { + unsigned pitch:18; + unsigned format:3; + unsigned pad:4; + unsigned depth_offset_disable:1; + unsigned tile_walk:1; + unsigned tiled_surface:1; + unsigned pad2:1; + unsigned surface_type:3; + } bits; + unsigned dword; + } dword1; + + unsigned dword2_base_addr; + + union { + struct { + unsigned pad:1; + unsigned mipmap_layout:1; + unsigned lod:4; + unsigned width:13; + unsigned height:13; + } bits; + unsigned dword; + } dword3; + + union { + struct { + unsigned pad:12; + unsigned min_array_element:9; + unsigned depth:11; + } bits; + unsigned dword; + } dword4; +}; + +struct brw_drawrect +{ + struct header header; + unsigned xmin:16; + unsigned ymin:16; + unsigned xmax:16; + unsigned ymax:16; + unsigned xorg:16; + unsigned yorg:16; +}; + + + + +struct brw_global_depth_offset_clamp +{ + struct header header; + float depth_offset_clamp; +}; + +struct brw_indexbuffer +{ + union { + struct + { + unsigned length:8; + unsigned index_format:2; + unsigned cut_index_enable:1; + unsigned pad:5; + unsigned opcode:16; + } bits; + unsigned dword; + + } header; + + unsigned buffer_start; + unsigned buffer_end; +}; + + +struct brw_line_stipple +{ + struct header header; + + struct + { + unsigned pattern:16; + unsigned pad:16; + } bits0; + + struct + { + unsigned repeat_count:9; + unsigned pad:7; + unsigned inverse_repeat_count:16; + } bits1; +}; + + +struct brw_pipelined_state_pointers +{ + struct header header; + + struct { + unsigned pad:5; + unsigned offset:27; + } vs; + + struct + { + unsigned enable:1; + unsigned pad:4; + unsigned offset:27; + } gs; + + struct + { + unsigned enable:1; + unsigned pad:4; + unsigned offset:27; + } clp; + + struct + { + unsigned pad:5; + unsigned offset:27; + } sf; + + struct + { + unsigned pad:5; + unsigned offset:27; + } wm; + + struct + { + unsigned pad:5; + unsigned offset:27; /* KW: check me! */ + } cc; +}; + + +struct brw_polygon_stipple_offset +{ + struct header header; + + struct { + unsigned y_offset:5; + unsigned pad:3; + unsigned x_offset:5; + unsigned pad0:19; + } bits0; +}; + + + +struct brw_polygon_stipple +{ + struct header header; + unsigned stipple[32]; +}; + + + +struct brw_pipeline_select +{ + struct + { + unsigned pipeline_select:1; + unsigned pad:15; + unsigned opcode:16; + } header; +}; + + +struct brw_pipe_control +{ + struct + { + unsigned length:8; + unsigned notify_enable:1; + unsigned pad:2; + unsigned instruction_state_cache_flush_enable:1; + unsigned write_cache_flush_enable:1; + unsigned depth_stall_enable:1; + unsigned post_sync_operation:2; + + unsigned opcode:16; + } header; + + struct + { + unsigned pad:2; + unsigned dest_addr_type:1; + unsigned dest_addr:29; + } bits1; + + unsigned data0; + unsigned data1; +}; + + +struct brw_urb_fence +{ + struct + { + unsigned length:8; + unsigned vs_realloc:1; + unsigned gs_realloc:1; + unsigned clp_realloc:1; + unsigned sf_realloc:1; + unsigned vfe_realloc:1; + unsigned cs_realloc:1; + unsigned pad:2; + unsigned opcode:16; + } header; + + struct + { + unsigned vs_fence:10; + unsigned gs_fence:10; + unsigned clp_fence:10; + unsigned pad:2; + } bits0; + + struct + { + unsigned sf_fence:10; + unsigned vf_fence:10; + unsigned cs_fence:10; + unsigned pad:2; + } bits1; +}; + +struct brw_constant_buffer_state /* previously brw_command_streamer */ +{ + struct header header; + + struct + { + unsigned nr_urb_entries:3; + unsigned pad:1; + unsigned urb_entry_size:5; + unsigned pad0:23; + } bits0; +}; + +struct brw_constant_buffer +{ + struct + { + unsigned length:8; + unsigned valid:1; + unsigned pad:7; + unsigned opcode:16; + } header; + + struct + { + unsigned buffer_length:6; + unsigned buffer_address:26; + } bits0; +}; + +struct brw_state_base_address +{ + struct header header; + + struct + { + unsigned modify_enable:1; + unsigned pad:4; + unsigned general_state_address:27; + } bits0; + + struct + { + unsigned modify_enable:1; + unsigned pad:4; + unsigned surface_state_address:27; + } bits1; + + struct + { + unsigned modify_enable:1; + unsigned pad:4; + unsigned indirect_object_state_address:27; + } bits2; + + struct + { + unsigned modify_enable:1; + unsigned pad:11; + unsigned general_state_upper_bound:20; + } bits3; + + struct + { + unsigned modify_enable:1; + unsigned pad:11; + unsigned indirect_object_state_upper_bound:20; + } bits4; +}; + +struct brw_state_prefetch +{ + struct header header; + + struct + { + unsigned prefetch_count:3; + unsigned pad:3; + unsigned prefetch_pointer:26; + } bits0; +}; + +struct brw_system_instruction_pointer +{ + struct header header; + + struct + { + unsigned pad:4; + unsigned system_instruction_pointer:28; + } bits0; +}; + + + + +/* State structs for the various fixed function units: + */ + + +struct thread0 +{ + unsigned pad0:1; + unsigned grf_reg_count:3; + unsigned pad1:2; + unsigned kernel_start_pointer:26; +}; + +struct thread1 +{ + unsigned ext_halt_exception_enable:1; + unsigned sw_exception_enable:1; + unsigned mask_stack_exception_enable:1; + unsigned timeout_exception_enable:1; + unsigned illegal_op_exception_enable:1; + unsigned pad0:3; + unsigned depth_coef_urb_read_offset:6; /* WM only */ + unsigned pad1:2; + unsigned floating_point_mode:1; + unsigned thread_priority:1; + unsigned binding_table_entry_count:8; + unsigned pad3:5; + unsigned single_program_flow:1; +}; + +struct thread2 +{ + unsigned per_thread_scratch_space:4; + unsigned pad0:6; + unsigned scratch_space_base_pointer:22; +}; + + +struct thread3 +{ + unsigned dispatch_grf_start_reg:4; + unsigned urb_entry_read_offset:6; + unsigned pad0:1; + unsigned urb_entry_read_length:6; + unsigned pad1:1; + unsigned const_urb_entry_read_offset:6; + unsigned pad2:1; + unsigned const_urb_entry_read_length:6; + unsigned pad3:1; +}; + + + +struct brw_clip_unit_state +{ + struct thread0 thread0; + struct + { + unsigned pad0:7; + unsigned sw_exception_enable:1; + unsigned pad1:3; + unsigned mask_stack_exception_enable:1; + unsigned pad2:1; + unsigned illegal_op_exception_enable:1; + unsigned pad3:2; + unsigned floating_point_mode:1; + unsigned thread_priority:1; + unsigned binding_table_entry_count:8; + unsigned pad4:5; + unsigned single_program_flow:1; + } thread1; + + struct thread2 thread2; + struct thread3 thread3; + + struct + { + unsigned pad0:9; + unsigned gs_output_stats:1; /* not always */ + unsigned stats_enable:1; + unsigned nr_urb_entries:7; + unsigned pad1:1; + unsigned urb_entry_allocation_size:5; + unsigned pad2:1; + unsigned max_threads:1; /* may be less */ + unsigned pad3:6; + } thread4; + + struct + { + unsigned pad0:13; + unsigned clip_mode:3; + unsigned userclip_enable_flags:8; + unsigned userclip_must_clip:1; + unsigned pad1:1; + unsigned guard_band_enable:1; + unsigned viewport_z_clip_enable:1; + unsigned viewport_xy_clip_enable:1; + unsigned vertex_position_space:1; + unsigned api_mode:1; + unsigned pad2:1; + } clip5; + + struct + { + unsigned pad0:5; + unsigned clipper_viewport_state_ptr:27; + } clip6; + + + float viewport_xmin; + float viewport_xmax; + float viewport_ymin; + float viewport_ymax; +}; + + + +struct brw_cc_unit_state +{ + struct + { + unsigned pad0:3; + unsigned bf_stencil_pass_depth_pass_op:3; + unsigned bf_stencil_pass_depth_fail_op:3; + unsigned bf_stencil_fail_op:3; + unsigned bf_stencil_func:3; + unsigned bf_stencil_enable:1; + unsigned pad1:2; + unsigned stencil_write_enable:1; + unsigned stencil_pass_depth_pass_op:3; + unsigned stencil_pass_depth_fail_op:3; + unsigned stencil_fail_op:3; + unsigned stencil_func:3; + unsigned stencil_enable:1; + } cc0; + + + struct + { + unsigned bf_stencil_ref:8; + unsigned stencil_write_mask:8; + unsigned stencil_test_mask:8; + unsigned stencil_ref:8; + } cc1; + + + struct + { + unsigned logicop_enable:1; + unsigned pad0:10; + unsigned depth_write_enable:1; + unsigned depth_test_function:3; + unsigned depth_test:1; + unsigned bf_stencil_write_mask:8; + unsigned bf_stencil_test_mask:8; + } cc2; + + + struct + { + unsigned pad0:8; + unsigned alpha_test_func:3; + unsigned alpha_test:1; + unsigned blend_enable:1; + unsigned ia_blend_enable:1; + unsigned pad1:1; + unsigned alpha_test_format:1; + unsigned pad2:16; + } cc3; + + struct + { + unsigned pad0:5; + unsigned cc_viewport_state_offset:27; + } cc4; + + struct + { + unsigned pad0:2; + unsigned ia_dest_blend_factor:5; + unsigned ia_src_blend_factor:5; + unsigned ia_blend_function:3; + unsigned statistics_enable:1; + unsigned logicop_func:4; + unsigned pad1:11; + unsigned dither_enable:1; + } cc5; + + struct + { + unsigned clamp_post_alpha_blend:1; + unsigned clamp_pre_alpha_blend:1; + unsigned clamp_range:2; + unsigned pad0:11; + unsigned y_dither_offset:2; + unsigned x_dither_offset:2; + unsigned dest_blend_factor:5; + unsigned src_blend_factor:5; + unsigned blend_function:3; + } cc6; + + struct { + union { + float f; + ubyte ub[4]; + } alpha_ref; + } cc7; +}; + + + +struct brw_sf_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct + { + unsigned pad0:10; + unsigned stats_enable:1; + unsigned nr_urb_entries:7; + unsigned pad1:1; + unsigned urb_entry_allocation_size:5; + unsigned pad2:1; + unsigned max_threads:6; + unsigned pad3:1; + } thread4; + + struct + { + unsigned front_winding:1; + unsigned viewport_transform:1; + unsigned pad0:3; + unsigned sf_viewport_state_offset:27; + } sf5; + + struct + { + unsigned pad0:9; + unsigned dest_org_vbias:4; + unsigned dest_org_hbias:4; + unsigned scissor:1; + unsigned disable_2x2_trifilter:1; + unsigned disable_zero_pix_trifilter:1; + unsigned point_rast_rule:2; + unsigned line_endcap_aa_region_width:2; + unsigned line_width:4; + unsigned fast_scissor_disable:1; + unsigned cull_mode:2; + unsigned aa_enable:1; + } sf6; + + struct + { + unsigned point_size:11; + unsigned use_point_size_state:1; + unsigned subpixel_precision:1; + unsigned sprite_point:1; + unsigned pad0:11; + unsigned trifan_pv:2; + unsigned linestrip_pv:2; + unsigned tristrip_pv:2; + unsigned line_last_pixel_enable:1; + } sf7; + +}; + + +struct brw_gs_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct + { + unsigned pad0:10; + unsigned stats_enable:1; + unsigned nr_urb_entries:7; + unsigned pad1:1; + unsigned urb_entry_allocation_size:5; + unsigned pad2:1; + unsigned max_threads:1; + unsigned pad3:6; + } thread4; + + struct + { + unsigned sampler_count:3; + unsigned pad0:2; + unsigned sampler_state_pointer:27; + } gs5; + + + struct + { + unsigned max_vp_index:4; + unsigned pad0:26; + unsigned reorder_enable:1; + unsigned pad1:1; + } gs6; +}; + + +struct brw_vs_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct + { + unsigned pad0:10; + unsigned stats_enable:1; + unsigned nr_urb_entries:7; + unsigned pad1:1; + unsigned urb_entry_allocation_size:5; + unsigned pad2:1; + unsigned max_threads:4; + unsigned pad3:3; + } thread4; + + struct + { + unsigned sampler_count:3; + unsigned pad0:2; + unsigned sampler_state_pointer:27; + } vs5; + + struct + { + unsigned vs_enable:1; + unsigned vert_cache_disable:1; + unsigned pad0:30; + } vs6; +}; + + +struct brw_wm_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct { + unsigned stats_enable:1; + unsigned pad0:1; + unsigned sampler_count:3; + unsigned sampler_state_pointer:27; + } wm4; + + struct + { + unsigned enable_8_pix:1; + unsigned enable_16_pix:1; + unsigned enable_32_pix:1; + unsigned pad0:7; + unsigned legacy_global_depth_bias:1; + unsigned line_stipple:1; + unsigned depth_offset:1; + unsigned polygon_stipple:1; + unsigned line_aa_region_width:2; + unsigned line_endcap_aa_region_width:2; + unsigned early_depth_test:1; + unsigned thread_dispatch_enable:1; + unsigned program_uses_depth:1; + unsigned program_computes_depth:1; + unsigned program_uses_killpixel:1; + unsigned legacy_line_rast: 1; + unsigned pad1:1; + unsigned max_threads:6; + unsigned pad2:1; + } wm5; + + float global_depth_offset_constant; + float global_depth_offset_scale; +}; + +struct brw_sampler_default_color { + float color[4]; +}; + +struct brw_sampler_state +{ + + struct + { + unsigned shadow_function:3; + unsigned lod_bias:11; + unsigned min_filter:3; + unsigned mag_filter:3; + unsigned mip_filter:2; + unsigned base_level:5; + unsigned pad:1; + unsigned lod_preclamp:1; + unsigned default_color_mode:1; + unsigned pad0:1; + unsigned disable:1; + } ss0; + + struct + { + unsigned r_wrap_mode:3; + unsigned t_wrap_mode:3; + unsigned s_wrap_mode:3; + unsigned pad:3; + unsigned max_lod:10; + unsigned min_lod:10; + } ss1; + + + struct + { + unsigned pad:5; + unsigned default_color_pointer:27; + } ss2; + + struct + { + unsigned pad:19; + unsigned max_aniso:3; + unsigned chroma_key_mode:1; + unsigned chroma_key_index:2; + unsigned chroma_key_enable:1; + unsigned monochrome_filter_width:3; + unsigned monochrome_filter_height:3; + } ss3; +}; + + +struct brw_clipper_viewport +{ + float xmin; + float xmax; + float ymin; + float ymax; +}; + +struct brw_cc_viewport +{ + float min_depth; + float max_depth; +}; + +struct brw_sf_viewport +{ + struct { + float m00; + float m11; + float m22; + float m30; + float m31; + float m32; + } viewport; + + struct { + short xmin; + short ymin; + short xmax; + short ymax; + } scissor; +}; + +/* Documented in the subsystem/shared-functions/sampler chapter... + */ +struct brw_surface_state +{ + struct { + unsigned cube_pos_z:1; + unsigned cube_neg_z:1; + unsigned cube_pos_y:1; + unsigned cube_neg_y:1; + unsigned cube_pos_x:1; + unsigned cube_neg_x:1; + unsigned pad:4; + unsigned mipmap_layout_mode:1; + unsigned vert_line_stride_ofs:1; + unsigned vert_line_stride:1; + unsigned color_blend:1; + unsigned writedisable_blue:1; + unsigned writedisable_green:1; + unsigned writedisable_red:1; + unsigned writedisable_alpha:1; + unsigned surface_format:9; + unsigned data_return_format:1; + unsigned pad0:1; + unsigned surface_type:3; + } ss0; + + struct { + unsigned base_addr; + } ss1; + + struct { + unsigned pad:2; + unsigned mip_count:4; + unsigned width:13; + unsigned height:13; + } ss2; + + struct { + unsigned tile_walk:1; + unsigned tiled_surface:1; + unsigned pad:1; + unsigned pitch:18; + unsigned depth:11; + } ss3; + + struct { + unsigned pad:19; + unsigned min_array_elt:9; + unsigned min_lod:4; + } ss4; +}; + + + +struct brw_vertex_buffer_state +{ + struct { + unsigned pitch:11; + unsigned pad:15; + unsigned access_type:1; + unsigned vb_index:5; + } vb0; + + unsigned start_addr; + unsigned max_index; +#if 1 + unsigned instance_data_step_rate; /* not included for sequential/random vertices? */ +#endif +}; + +#define BRW_VBP_MAX 17 + +struct brw_vb_array_state { + struct header header; + struct brw_vertex_buffer_state vb[BRW_VBP_MAX]; +}; + + +struct brw_vertex_element_state +{ + struct + { + unsigned src_offset:11; + unsigned pad:5; + unsigned src_format:9; + unsigned pad0:1; + unsigned valid:1; + unsigned vertex_buffer_index:5; + } ve0; + + struct + { + unsigned dst_offset:8; + unsigned pad:8; + unsigned vfcomponent3:4; + unsigned vfcomponent2:4; + unsigned vfcomponent1:4; + unsigned vfcomponent0:4; + } ve1; +}; + +#define BRW_VEP_MAX 18 + +struct brw_vertex_element_packet { + struct header header; + struct brw_vertex_element_state ve[BRW_VEP_MAX]; /* note: less than _TNL_ATTRIB_MAX */ +}; + + +struct brw_urb_immediate { + unsigned opcode:4; + unsigned offset:6; + unsigned swizzle_control:2; + unsigned pad:1; + unsigned allocate:1; + unsigned used:1; + unsigned complete:1; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; +}; + +/* Instruction format for the execution units: + */ + +struct brw_instruction +{ + struct + { + unsigned opcode:7; + unsigned pad:1; + unsigned access_mode:1; + unsigned mask_control:1; + unsigned dependency_control:2; + unsigned compression_control:2; + unsigned thread_control:2; + unsigned predicate_control:4; + unsigned predicate_inverse:1; + unsigned execution_size:3; + unsigned destreg__conditonalmod:4; /* destreg - send, conditionalmod - others */ + unsigned pad0:2; + unsigned debug_control:1; + unsigned saturate:1; + } header; + + union { + struct + { + unsigned dest_reg_file:2; + unsigned dest_reg_type:3; + unsigned src0_reg_file:2; + unsigned src0_reg_type:3; + unsigned src1_reg_file:2; + unsigned src1_reg_type:3; + unsigned pad:1; + unsigned dest_subreg_nr:5; + unsigned dest_reg_nr:8; + unsigned dest_horiz_stride:2; + unsigned dest_address_mode:1; + } da1; + + struct + { + unsigned dest_reg_file:2; + unsigned dest_reg_type:3; + unsigned src0_reg_file:2; + unsigned src0_reg_type:3; + unsigned pad:6; + int dest_indirect_offset:10; /* offset against the deref'd address reg */ + unsigned dest_subreg_nr:3; /* subnr for the address reg a0.x */ + unsigned dest_horiz_stride:2; + unsigned dest_address_mode:1; + } ia1; + + struct + { + unsigned dest_reg_file:2; + unsigned dest_reg_type:3; + unsigned src0_reg_file:2; + unsigned src0_reg_type:3; + unsigned src1_reg_file:2; + unsigned src1_reg_type:3; + unsigned pad0:1; + unsigned dest_writemask:4; + unsigned dest_subreg_nr:1; + unsigned dest_reg_nr:8; + unsigned pad1:2; + unsigned dest_address_mode:1; + } da16; + + struct + { + unsigned dest_reg_file:2; + unsigned dest_reg_type:3; + unsigned src0_reg_file:2; + unsigned src0_reg_type:3; + unsigned pad0:6; + unsigned dest_writemask:4; + int dest_indirect_offset:6; + unsigned dest_subreg_nr:3; + unsigned pad1:2; + unsigned dest_address_mode:1; + } ia16; + } bits1; + + + union { + struct + { + unsigned src0_subreg_nr:5; + unsigned src0_reg_nr:8; + unsigned src0_abs:1; + unsigned src0_negate:1; + unsigned src0_address_mode:1; + unsigned src0_horiz_stride:2; + unsigned src0_width:3; + unsigned src0_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad:6; + } da1; + + struct + { + int src0_indirect_offset:10; + unsigned src0_subreg_nr:3; + unsigned src0_abs:1; + unsigned src0_negate:1; + unsigned src0_address_mode:1; + unsigned src0_horiz_stride:2; + unsigned src0_width:3; + unsigned src0_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad:6; + } ia1; + + struct + { + unsigned src0_swz_x:2; + unsigned src0_swz_y:2; + unsigned src0_subreg_nr:1; + unsigned src0_reg_nr:8; + unsigned src0_abs:1; + unsigned src0_negate:1; + unsigned src0_address_mode:1; + unsigned src0_swz_z:2; + unsigned src0_swz_w:2; + unsigned pad0:1; + unsigned src0_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad1:6; + } da16; + + struct + { + unsigned src0_swz_x:2; + unsigned src0_swz_y:2; + int src0_indirect_offset:6; + unsigned src0_subreg_nr:3; + unsigned src0_abs:1; + unsigned src0_negate:1; + unsigned src0_address_mode:1; + unsigned src0_swz_z:2; + unsigned src0_swz_w:2; + unsigned pad0:1; + unsigned src0_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad1:6; + } ia16; + + } bits2; + + union + { + struct + { + unsigned src1_subreg_nr:5; + unsigned src1_reg_nr:8; + unsigned src1_abs:1; + unsigned src1_negate:1; + unsigned pad:1; + unsigned src1_horiz_stride:2; + unsigned src1_width:3; + unsigned src1_vert_stride:4; + unsigned pad0:7; + } da1; + + struct + { + unsigned src1_swz_x:2; + unsigned src1_swz_y:2; + unsigned src1_subreg_nr:1; + unsigned src1_reg_nr:8; + unsigned src1_abs:1; + unsigned src1_negate:1; + unsigned pad0:1; + unsigned src1_swz_z:2; + unsigned src1_swz_w:2; + unsigned pad1:1; + unsigned src1_vert_stride:4; + unsigned pad2:7; + } da16; + + struct + { + int src1_indirect_offset:10; + unsigned src1_subreg_nr:3; + unsigned src1_abs:1; + unsigned src1_negate:1; + unsigned pad0:1; + unsigned src1_horiz_stride:2; + unsigned src1_width:3; + unsigned src1_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad1:6; + } ia1; + + struct + { + unsigned src1_swz_x:2; + unsigned src1_swz_y:2; + int src1_indirect_offset:6; + unsigned src1_subreg_nr:3; + unsigned src1_abs:1; + unsigned src1_negate:1; + unsigned pad0:1; + unsigned src1_swz_z:2; + unsigned src1_swz_w:2; + unsigned pad1:1; + unsigned src1_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad2:6; + } ia16; + + + struct + { + int jump_count:16; /* note: signed */ + unsigned pop_count:4; + unsigned pad0:12; + } if_else; + + struct { + unsigned function:4; + unsigned int_type:1; + unsigned precision:1; + unsigned saturate:1; + unsigned data_type:1; + unsigned pad0:8; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } math; + + struct { + unsigned binding_table_index:8; + unsigned sampler:4; + unsigned return_format:2; + unsigned msg_type:2; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } sampler; + + struct brw_urb_immediate urb; + + struct { + unsigned binding_table_index:8; + unsigned msg_control:4; + unsigned msg_type:2; + unsigned target_cache:2; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } dp_read; + + struct { + unsigned binding_table_index:8; + unsigned msg_control:3; + unsigned pixel_scoreboard_clear:1; + unsigned msg_type:3; + unsigned send_commit_msg:1; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } dp_write; + + struct { + unsigned pad:16; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } generic; + + int d; + unsigned ud; + } bits3; +}; + + +#endif diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c new file mode 100644 index 0000000000..b5a0318047 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -0,0 +1,397 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "brw_blit.h" +#include "brw_context.h" +#include "brw_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" + + +#define CLIP_TILE \ + do { \ + if (x >= ps->width) \ + return; \ + if (y >= ps->height) \ + return; \ + if (x + w > ps->width) \ + w = ps->width - x; \ + if (y + h > ps->height) \ + h = ps->height -y; \ + } while(0) + + +/** + * Note: this is exactly like a8r8g8b8_get_tile() in sp_surface.c + * Share it someday. + */ +static void +brw_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, float *p) +{ + const unsigned *src + = ((const unsigned *) (ps->map + ps->offset)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + CLIP_TILE; + + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const unsigned pixel = src[j]; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } + break; + case PIPE_FORMAT_S8Z24_UNORM: + { + const float scale = 1.0f / (float) 0xffffff; + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const unsigned pixel = src[j]; + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (pixel & 0xffffff) * scale; + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } + } + break; + default: + assert(0); + } +} + + +static void +brw_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, const float *p) +{ + /* TODO */ + assert(0); +} + + +/* + * XXX note: same as code in sp_surface.c + */ +static void +brw_get_tile(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride) +{ + const uint cpp = ps->cpp; + const uint w0 = w; + const ubyte *pSrc; + ubyte *pDest; + uint i; + + assert(ps->map); + + CLIP_TILE; + + if (dst_stride == 0) { + dst_stride = w0 * cpp; + } + + pSrc = ps->map + ps->offset + (y * ps->pitch + x) * cpp; + pDest = (ubyte *) p; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w0 * cpp); + pDest += dst_stride; + pSrc += ps->pitch * cpp; + } +} + + +/* + * XXX note: same as code in sp_surface.c + */ +static void +brw_put_tile(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride) +{ + const uint cpp = ps->cpp; + const uint w0 = w; + const ubyte *pSrc; + ubyte *pDest; + uint i; + + assert(ps->map); + + CLIP_TILE; + + if (src_stride == 0) { + src_stride = w0 * cpp; + } + + pSrc = (const ubyte *) p; + pDest = ps->map + ps->offset + (y * ps->pitch + x) * cpp; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w0 * cpp); + pDest += ps->pitch * cpp; + pSrc += src_stride; + } +} + + +/* + * XXX note: same as code in sp_surface.c + */ +static struct pipe_surface * +brw_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct brw_texture *tex = (struct brw_texture *)pt; + struct pipe_surface *ps; + unsigned offset; /* in bytes */ + + offset = tex->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE) { + offset += tex->image_offset[level][face] * pt->cpp; + } + else if (pt->target == PIPE_TEXTURE_3D) { + offset += tex->image_offset[level][zslice] * pt->cpp; + } + else { + assert(face == 0); + assert(zslice == 0); + } + + ps = pipe->winsys->surface_alloc(pipe->winsys); + if (ps) { + assert(ps->format); + assert(ps->refcount); + pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = tex->pitch; + ps->offset = offset; + } + return ps; +} + +/* + * XXX Move this into core Mesa? + */ +static void +_mesa_copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + unsigned src_pitch, + unsigned src_x, + unsigned src_y) +{ + unsigned i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * dst_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } +} + +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +static void +brw_surface_data(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + const void *src, unsigned src_pitch, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + _mesa_copy_rect(pipe_surface_map(dst) + dst->offset, + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + pipe_surface_unmap(dst); +} + + +/* Assumes all values are within bounds -- no checking at this level - + * do it higher up if required. + */ +static void +brw_surface_copy(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + assert(dst != src); + assert(dst->cpp == src->cpp); + + if (0) { + _mesa_copy_rect(pipe_surface_map(dst) + dst->offset, + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe_surface_map(src) + src->offset, + src->pitch, + srcx, srcy); + + pipe_surface_unmap(src); + pipe_surface_unmap(dst); + } + else { + brw_copy_blit(brw_context(pipe), + dst->cpp, + (short) src->pitch, src->buffer, src->offset, FALSE, + (short) dst->pitch, dst->buffer, dst->offset, FALSE, + (short) srcx, (short) srcy, (short) dstx, (short) dsty, + (short) width, (short) height, PIPE_LOGICOP_COPY); + } +} + +/* Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +static ubyte * +get_pointer(struct pipe_surface *dst, unsigned x, unsigned y) +{ + return dst->map + (y * dst->pitch + x) * dst->cpp; +} + + +static void +brw_surface_fill(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) +{ + if (0) { + unsigned i, j; + + (void)pipe_surface_map(dst); + + switch (dst->cpp) { + case 1: { + ubyte *row = get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: { + ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (ushort) value; + row += dst->pitch; + } + } + break; + case 4: { + unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + default: + assert(0); + break; + } + } + else { + brw_fill_blit(brw_context(pipe), + dst->cpp, + (short) dst->pitch, + dst->buffer, dst->offset, FALSE, + (short) dstx, (short) dsty, + (short) width, (short) height, + value); + } +} + +void +brw_init_surface_functions(struct brw_context *brw) +{ + brw->pipe.get_tex_surface = brw_get_tex_surface; + brw->pipe.get_tile = brw_get_tile; + brw->pipe.put_tile = brw_put_tile; + brw->pipe.get_tile_rgba = brw_get_tile_rgba; + brw->pipe.put_tile_rgba = brw_put_tile_rgba; + + brw->pipe.surface_data = brw_surface_data; + brw->pipe.surface_copy = brw_surface_copy; + brw->pipe.surface_fill = brw_surface_fill; +} diff --git a/src/mesa/pipe/i965simple/brw_tex.c b/src/mesa/pipe/i965simple/brw_tex.c new file mode 100644 index 0000000000..e9ff67c3f4 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_tex.c @@ -0,0 +1,65 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_defines.h" + +#if 0 +void brw_FrameBufferTexInit( struct brw_context *brw ) +{ + struct intel_context *intel = &brw->intel; + GLcontext *ctx = &intel->ctx; + struct intel_region *region = intel->intelScreen->front_region; + struct gl_texture_object *obj; + struct gl_texture_image *img; + + intel->frame_buffer_texobj = obj = + ctx->Driver.NewTextureObject( ctx, (unsigned) -1, GL_TEXTURE_2D ); + + obj->MinFilter = GL_NEAREST; + obj->MagFilter = GL_NEAREST; + + img = ctx->Driver.NewTextureImage( ctx ); + + _mesa_init_teximage_fields( ctx, GL_TEXTURE_2D, img, + region->pitch, region->height, 1, 0, + region->cpp == 4 ? GL_RGBA : GL_RGB ); + + _mesa_set_tex_image( obj, GL_TEXTURE_2D, 0, img ); +} + +void brw_FrameBufferTexDestroy( struct brw_context *brw ) +{ + brw->intel.ctx.Driver.DeleteTexture( &brw->intel.ctx, + brw->intel.frame_buffer_texobj ); +} +#endif diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.c b/src/mesa/pipe/i965simple/brw_tex_layout.c new file mode 100644 index 0000000000..b9514be0c2 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_tex_layout.c @@ -0,0 +1,359 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +/* Code to layout images in a mipmap tree for i965. + */ + +#include "brw_tex_layout.h" + +#include "pipe/p_state.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "brw_context.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +#if 0 +unsigned intel_compressed_alignment(unsigned internalFormat) +{ + unsigned alignment = 4; + + switch (internalFormat) { + case GL_COMPRESSED_RGB_FXT1_3DFX: + case GL_COMPRESSED_RGBA_FXT1_3DFX: + alignment = 8; + break; + + default: + break; + } + + return alignment; +} +#endif + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + + +static boolean brw_miptree_layout(struct pipe_context *, struct brw_texture *); + +static void intel_miptree_set_image_offset(struct brw_texture *tex, + unsigned level, + unsigned img, + unsigned x, unsigned y) +{ + struct pipe_texture *pt = &tex->base; + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + assert(img < tex->nr_images[level]); + + tex->image_offset[level][img] = (x + y * tex->pitch) * pt->cpp; +} + +static void intel_miptree_set_level_info(struct brw_texture *tex, + unsigned level, + unsigned nr_images, + unsigned x, unsigned y, + unsigned w, unsigned h, unsigned d) +{ + struct pipe_texture *pt = &tex->base; + + assert(level < PIPE_MAX_TEXTURE_LEVELS); + + pt->width[level] = w; + pt->height[level] = h; + pt->depth[level] = d; + + tex->level_offset[level] = (x + y * tex->pitch) * pt->cpp; + tex->nr_images[level] = nr_images; + + /* + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, tex->level_offset[level]); + */ + + /* Not sure when this would happen, but anyway: + */ + if (tex->image_offset[level]) { + FREE(tex->image_offset[level]); + tex->image_offset[level] = NULL; + } + + assert(nr_images); + assert(!tex->image_offset[level]); + + tex->image_offset[level] = (unsigned *) MALLOC(nr_images * sizeof(unsigned)); + tex->image_offset[level][0] = 0; +} + +static void i945_miptree_layout_2d(struct brw_texture *tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned align_h = 2, align_w = 4; + unsigned level; + unsigned x = 0; + unsigned y = 0; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + + tex->pitch = pt->width[0]; + +#if 0 + if (pt->compressed) { + align_w = intel_compressed_alignment(pt->internal_format); + tex->pitch = ALIGN(pt->width[0], align_w); + } +#endif + + /* May need to adjust pitch to accomodate the placement of + * the 2nd mipmap. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap out past the width of its parent. + */ + if (pt->first_level != pt->last_level) { + unsigned mip1_width; + + if (pt->compressed) { + mip1_width = ALIGN(minify(pt->width[0]), align_w) + + ALIGN(minify(minify(pt->width[0])), align_w); + } else { + mip1_width = ALIGN(minify(pt->width[0]), align_w) + + minify(minify(pt->width[0])); + } + + if (mip1_width > tex->pitch) { + tex->pitch = mip1_width; + } + } + + /* Pitch must be a whole number of dwords, even though we + * express it in texels. + */ + tex->pitch = ALIGN(tex->pitch * pt->cpp, 4) / pt->cpp; + tex->total_height = 0; + + for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + unsigned img_height; + + intel_miptree_set_level_info(tex, level, 1, x, y, width, + height, 1); + + if (pt->compressed) + img_height = MAX2(1, height/4); + else + img_height = ALIGN(height, align_h); + + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + tex->total_height = MAX2(tex->total_height, y + img_height); + + /* Layout_below: step right after second mipmap. + */ + if (level == pt->first_level + 1) { + x += ALIGN(width, align_w); + } + else { + y += img_height; + } + + width = minify(width); + height = minify(height); + } +} + +static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture *tex) +{ + struct pipe_texture *pt = &tex->base; + /* XXX: these vary depending on image format: + */ +/* int align_w = 4; */ + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: + case PIPE_TEXTURE_3D: { + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + unsigned pack_x_pitch, pack_x_nr; + unsigned pack_y_pitch; + unsigned level; + unsigned align_h = 2; + unsigned align_w = 4; + + tex->total_height = 0; +#if 0 + if (pt->compressed) { + align_w = intel_compressed_alignment(pt->internal_format); + pt->pitch = ALIGN(width, align_w); + pack_y_pitch = (height + 3) / 4; + } else +#endif + { + tex->pitch = ALIGN(pt->width[0] * pt->cpp, 4) / pt->cpp; + pack_y_pitch = ALIGN(pt->height[0], align_h); + } + + pack_x_pitch = tex->pitch; + pack_x_nr = 1; + + for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; + int x = 0; + int y = 0; + int q, j; + + intel_miptree_set_level_info(tex, level, nr_images, + 0, tex->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + intel_miptree_set_image_offset(tex, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + tex->total_height += y; + width = minify(width); + height = minify(height); + depth = minify(depth); + + if (pt->compressed) { + pack_y_pitch = (height + 3) / 4; + + if (pack_x_pitch > ALIGN(width, align_w)) { + pack_x_pitch = ALIGN(width, align_w); + pack_x_nr <<= 1; + } + } else { + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= tex->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + pack_y_pitch = ALIGN(pack_y_pitch, align_h); + } + } + + } + break; + } + + default: + i945_miptree_layout_2d(tex); + break; + } +#if 0 + PRINT("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + pt->pitch, + pt->total_height, + pt->cpp, + pt->pitch * pt->total_height * pt->cpp ); +#endif + + return TRUE; +} + +void +brw_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) +{ + struct brw_texture *tex = REALLOC(*pt, sizeof(struct pipe_texture), + sizeof(struct brw_texture)); + + if (tex) { + struct brw_context *brw = brw_context(pipe); + + memset(&tex->base + 1, 0, + sizeof(struct brw_texture) - sizeof(struct pipe_texture)); + + if (brw_miptree_layout(pipe, tex)) { + tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, 0, 0); + + if (tex->buffer) + pipe->winsys->buffer_data(pipe->winsys, tex->buffer, + tex->pitch * tex->base.cpp * + tex->total_height, NULL, + PIPE_BUFFER_USAGE_PIXEL); + } + + if (!tex->buffer) { + FREE(tex); + tex = NULL; + } + } + + *pt = &tex->base; +} + +void +brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct brw_texture *tex = (struct brw_texture *)*pt; + uint i; + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); + */ + + pipe->winsys->buffer_reference(pipe->winsys, &tex->buffer, NULL); + + for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) + if (tex->image_offset[i]) + free(tex->image_offset[i]); + + free(tex); + } + *pt = NULL; +} diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.h b/src/mesa/pipe/i965simple/brw_tex_layout.h new file mode 100644 index 0000000000..15e275058a --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_tex_layout.h @@ -0,0 +1,15 @@ +#ifndef BRW_TEX_LAYOUT_H +#define BRW_TEX_LAYOUT_H + +#include "pipe/p_compiler.h" + +struct pipe_context; +struct pipe_texture; + +extern void +brw_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); + +extern void +brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); + +#endif diff --git a/src/mesa/pipe/i965simple/brw_urb.c b/src/mesa/pipe/i965simple/brw_urb.c new file mode 100644 index 0000000000..64255b67c4 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_urb.c @@ -0,0 +1,197 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + +#define VS 0 +#define GS 1 +#define CLP 2 +#define SF 3 +#define CS 4 + +/* XXX: Are the min_entry_size numbers useful? + * XXX: Verify min_nr_entries, esp for VS. + * XXX: Verify SF min_entry_size. + */ +static const struct { + unsigned min_nr_entries; + unsigned preferred_nr_entries; + unsigned min_entry_size; + unsigned max_entry_size; +} limits[CS+1] = { + { 8, 32, 1, 5 }, /* vs */ + { 4, 8, 1, 5 }, /* gs */ + { 6, 8, 1, 5 }, /* clp */ + { 1, 8, 1, 12 }, /* sf */ + { 1, 4, 1, 32 } /* cs */ +}; + + +static boolean check_urb_layout( struct brw_context *brw ) +{ + brw->urb.vs_start = 0; + brw->urb.gs_start = brw->urb.nr_vs_entries * brw->urb.vsize; + brw->urb.clip_start = brw->urb.gs_start + brw->urb.nr_gs_entries * brw->urb.vsize; + brw->urb.sf_start = brw->urb.clip_start + brw->urb.nr_clip_entries * brw->urb.vsize; + brw->urb.cs_start = brw->urb.sf_start + brw->urb.nr_sf_entries * brw->urb.sfsize; + + return brw->urb.cs_start + brw->urb.nr_cs_entries * brw->urb.csize <= 256; +} + +/* Most minimal update, forces re-emit of URB fence packet after GS + * unit turned on/off. + */ +static void recalculate_urb_fence( struct brw_context *brw ) +{ + unsigned csize = brw->curbe.total_size; + unsigned vsize = brw->vs.prog_data->urb_entry_size; + unsigned sfsize = brw->sf.prog_data->urb_entry_size; + + if (csize < limits[CS].min_entry_size) + csize = limits[CS].min_entry_size; + + if (vsize < limits[VS].min_entry_size) + vsize = limits[VS].min_entry_size; + + if (sfsize < limits[SF].min_entry_size) + sfsize = limits[SF].min_entry_size; + + if (brw->urb.vsize < vsize || + brw->urb.sfsize < sfsize || + brw->urb.csize < csize || + (brw->urb.constrained && (brw->urb.vsize > brw->urb.vsize || + brw->urb.sfsize > brw->urb.sfsize || + brw->urb.csize > brw->urb.csize))) { + + + brw->urb.csize = csize; + brw->urb.sfsize = sfsize; + brw->urb.vsize = vsize; + + brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries; + brw->urb.nr_gs_entries = limits[GS].preferred_nr_entries; + brw->urb.nr_clip_entries = limits[CLP].preferred_nr_entries; + brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries; + brw->urb.nr_cs_entries = limits[CS].preferred_nr_entries; + + if (!check_urb_layout(brw)) { + brw->urb.nr_vs_entries = limits[VS].min_nr_entries; + brw->urb.nr_gs_entries = limits[GS].min_nr_entries; + brw->urb.nr_clip_entries = limits[CLP].min_nr_entries; + brw->urb.nr_sf_entries = limits[SF].min_nr_entries; + brw->urb.nr_cs_entries = limits[CS].min_nr_entries; + + brw->urb.constrained = 1; + + if (!check_urb_layout(brw)) { + /* This is impossible, given the maximal sizes of urb + * entries and the values for minimum nr of entries + * provided above. + */ + fprintf(stderr, "couldn't calculate URB layout!\n"); + exit(1); + } + + if (BRW_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS)) + printf("URB CONSTRAINED\n"); + } + else + brw->urb.constrained = 0; + + if (BRW_DEBUG & DEBUG_URB) + printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n", + brw->urb.vs_start, + brw->urb.gs_start, + brw->urb.clip_start, + brw->urb.sf_start, + brw->urb.cs_start, + 256); + + brw->state.dirty.brw |= BRW_NEW_URB_FENCE; + } +} + + +const struct brw_tracked_state brw_recalculate_urb_fence = { + .dirty = { + .brw = BRW_NEW_CURBE_OFFSETS, + .cache = (CACHE_NEW_VS_PROG | + CACHE_NEW_SF_PROG) + }, + .update = recalculate_urb_fence +}; + + + + + +void brw_upload_urb_fence(struct brw_context *brw) +{ + struct brw_urb_fence uf; + memset(&uf, 0, sizeof(uf)); + + uf.header.opcode = CMD_URB_FENCE; + uf.header.length = sizeof(uf)/4-2; + uf.header.vs_realloc = 1; + uf.header.gs_realloc = 1; + uf.header.clp_realloc = 1; + uf.header.sf_realloc = 1; + uf.header.vfe_realloc = 1; + uf.header.cs_realloc = 1; + + /* The ordering below is correct, not the layout in the + * instruction. + * + * There are 256 urb reg pairs in total. + */ + uf.bits0.vs_fence = brw->urb.gs_start; + uf.bits0.gs_fence = brw->urb.clip_start; + uf.bits0.clp_fence = brw->urb.sf_start; + uf.bits1.sf_fence = brw->urb.cs_start; + uf.bits1.cs_fence = 256; + + BRW_BATCH_STRUCT(brw, &uf); +} + + +#if 0 +const struct brw_tracked_state brw_urb_fence = { + .dirty = { + .mesa = 0, + .brw = BRW_NEW_URB_FENCE | BRW_NEW_PSP, + .cache = 0 + }, + .update = brw_upload_urb_fence +}; +#endif diff --git a/src/mesa/pipe/i965simple/brw_util.c b/src/mesa/pipe/i965simple/brw_util.c new file mode 100644 index 0000000000..42391d7c8c --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_util.c @@ -0,0 +1,104 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_util.h" +#include "brw_defines.h" + +#include "pipe/p_defines.h" + +unsigned brw_count_bits( unsigned val ) +{ + unsigned i; + for (i = 0; val ; val >>= 1) + if (val & 1) + i++; + return i; +} + + +unsigned brw_translate_blend_equation( int mode ) +{ + switch (mode) { + case PIPE_BLEND_ADD: + return BRW_BLENDFUNCTION_ADD; + case PIPE_BLEND_MIN: + return BRW_BLENDFUNCTION_MIN; + case PIPE_BLEND_MAX: + return BRW_BLENDFUNCTION_MAX; + case PIPE_BLEND_SUBTRACT: + return BRW_BLENDFUNCTION_SUBTRACT; + case PIPE_BLEND_REVERSE_SUBTRACT: + return BRW_BLENDFUNCTION_REVERSE_SUBTRACT; + default: + assert(0); + return BRW_BLENDFUNCTION_ADD; + } +} + +unsigned brw_translate_blend_factor( int factor ) +{ + switch(factor) { + case PIPE_BLENDFACTOR_ZERO: + return BRW_BLENDFACTOR_ZERO; + case PIPE_BLENDFACTOR_SRC_ALPHA: + return BRW_BLENDFACTOR_SRC_ALPHA; + case PIPE_BLENDFACTOR_ONE: + return BRW_BLENDFACTOR_ONE; + case PIPE_BLENDFACTOR_SRC_COLOR: + return BRW_BLENDFACTOR_SRC_COLOR; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + return BRW_BLENDFACTOR_INV_SRC_COLOR; + case PIPE_BLENDFACTOR_DST_COLOR: + return BRW_BLENDFACTOR_DST_COLOR; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + return BRW_BLENDFACTOR_INV_DST_COLOR; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + return BRW_BLENDFACTOR_INV_SRC_ALPHA; + case PIPE_BLENDFACTOR_DST_ALPHA: + return BRW_BLENDFACTOR_DST_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + return BRW_BLENDFACTOR_INV_DST_ALPHA; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + return BRW_BLENDFACTOR_SRC_ALPHA_SATURATE; + case PIPE_BLENDFACTOR_CONST_COLOR: + return BRW_BLENDFACTOR_CONST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + return BRW_BLENDFACTOR_INV_CONST_COLOR; + case PIPE_BLENDFACTOR_CONST_ALPHA: + return BRW_BLENDFACTOR_CONST_ALPHA; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + return BRW_BLENDFACTOR_INV_CONST_ALPHA; + default: + assert(0); + return BRW_BLENDFACTOR_ZERO; + } +} diff --git a/src/mesa/pipe/i965simple/brw_util.h b/src/mesa/pipe/i965simple/brw_util.h new file mode 100644 index 0000000000..d60e5934db --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_util.h @@ -0,0 +1,43 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_UTIL_H +#define BRW_UTIL_H + +#include "pipe/p_state.h" + +extern unsigned brw_count_bits( unsigned val ); +extern unsigned brw_translate_blend_factor( int factor ); +extern unsigned brw_translate_blend_equation( int mode ); + + +#endif diff --git a/src/mesa/pipe/i965simple/brw_vs.c b/src/mesa/pipe/i965simple/brw_vs.c new file mode 100644 index 0000000000..33c6624214 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_vs.c @@ -0,0 +1,130 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_vs.h" +#include "brw_util.h" +#include "brw_state.h" + + +static void do_vs_prog( struct brw_context *brw, + const struct brw_vertex_program *vp, + struct brw_vs_prog_key *key ) +{ + unsigned program_size; + const unsigned *program; + struct brw_vs_compile c; + + memset(&c, 0, sizeof(c)); + memcpy(&c.key, key, sizeof(*key)); + + brw_init_compile(&c.func); + c.vp = vp; + + c.prog_data.outputs_written = vp->program.num_outputs; + c.prog_data.inputs_read = vp->program.num_inputs; + +#if 0 + if (c.key.copy_edgeflag) { + c.prog_data.outputs_written |= 1<vs.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_VS_PROG], + &c.key, + sizeof(c.key), + program, + program_size, + &c.prog_data, + &brw->vs.prog_data); +} + + +static void brw_upload_vs_prog( struct brw_context *brw ) +{ + struct brw_vs_prog_key key; + const struct brw_vertex_program *vp = brw->attribs.VertexProgram; + + assert(vp); + + memset(&key, 0, sizeof(key)); + + /* Just upload the program verbatim for now. Always send it all + * the inputs it asks for, whether they are varying or not. + */ + key.program_string_id = vp->id; + key.nr_userclip = brw->attribs.Clip.nr; + key.copy_edgeflag = (brw->attribs.Raster->fill_cw != PIPE_POLYGON_MODE_FILL || + brw->attribs.Raster->fill_ccw != PIPE_POLYGON_MODE_FILL); + +#if 0 + /* BRW_NEW_METAOPS + */ + if (brw->metaops.active) + key.know_w_is_one = 1; +#endif + + /* Make an early check for the key. + */ + if (brw_search_cache(&brw->cache[BRW_VS_PROG], + &key, sizeof(key), + &brw->vs.prog_data, + &brw->vs.prog_gs_offset)) + return; + + do_vs_prog(brw, vp, &key); +} + + +/* See brw_vs.c: + */ +const struct brw_tracked_state brw_vs_prog = { + .dirty = { +#if 0 + .brw = BRW_NEW_VS | BRW_NEW_METAOPS, +#endif + .brw = BRW_NEW_VS, + .cache = 0 + }, + .update = brw_upload_vs_prog +}; diff --git a/src/mesa/pipe/i965simple/brw_vs.h b/src/mesa/pipe/i965simple/brw_vs.h new file mode 100644 index 0000000000..0e58f043b0 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_vs.h @@ -0,0 +1,82 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_VS_H +#define BRW_VS_H + + +#include "brw_context.h" +#include "brw_eu.h" + + +struct brw_vs_prog_key { + unsigned program_string_id; + unsigned nr_userclip:4; + unsigned copy_edgeflag:1; + unsigned know_w_is_one:1; + unsigned pad:26; +}; + + +struct brw_vs_compile { + struct brw_compile func; + struct brw_vs_prog_key key; + struct brw_vs_prog_data prog_data; + + struct brw_vertex_program *vp; + + unsigned nr_inputs; + + unsigned first_output; + unsigned nr_outputs; + + unsigned first_tmp; + unsigned last_tmp; + + struct brw_reg r0; + struct brw_reg r1; + struct brw_reg regs[12][128]; + struct brw_reg tmp; + struct brw_reg stack; + + struct { + boolean used_in_src; + struct brw_reg reg; + } output_regs[128]; + + struct brw_reg userplane[6]; + +}; + +void brw_vs_emit( struct brw_vs_compile *c ); + +#endif diff --git a/src/mesa/pipe/i965simple/brw_vs_constval.c b/src/mesa/pipe/i965simple/brw_vs_constval.c new file mode 100644 index 0000000000..de43e72c1d --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_vs_constval.c @@ -0,0 +1,223 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_vs.h" + +#if 0 +/* Component is active if it may diverge from [0,0,0,1]. Undef values + * are promoted to [0,0,0,1] for the purposes of this analysis. + */ +struct tracker { + boolean twoside; + ubyte active[PROGRAM_OUTPUT+1][128]; + unsigned size_masks[4]; +}; + + +static void set_active_component( struct tracker *t, + unsigned file, + unsigned index, + ubyte active ) +{ + switch (file) { + case PROGRAM_TEMPORARY: + case PROGRAM_INPUT: + case PROGRAM_OUTPUT: + t->active[file][index] |= active; + + default: + break; + } +} + +static void set_active( struct tracker *t, + struct prog_dst_register dst, + unsigned active ) +{ + set_active_component( t, dst.File, dst.Index, active & dst.WriteMask ); +} + + +static ubyte get_active_component( struct tracker *t, + unsigned file, + unsigned index, + unsigned component, + ubyte swz ) +{ + switch (swz) { + case SWIZZLE_ZERO: + return component < 3 ? 0 : (1<active[file][index] & (1<active[PROGRAM_OUTPUT][idx]; + if (active & (1<<3)) return 4; + if (active & (1<<2)) return 3; + if (active & (1<<1)) return 2; + if (active & (1<<0)) return 1; + return 0; +} + +/* Note the potential copying that occurs in the setup program: + */ +static void calc_sizes( struct tracker *t ) +{ + unsigned i; + + if (t->twoside) { + t->active[PROGRAM_OUTPUT][VERT_RESULT_COL0] |= + t->active[PROGRAM_OUTPUT][VERT_RESULT_BFC0]; + + t->active[PROGRAM_OUTPUT][VERT_RESULT_COL1] |= + t->active[PROGRAM_OUTPUT][VERT_RESULT_BFC1]; + } + + for (i = 0; i < FRAG_ATTRIB_MAX; i++) { + switch (get_output_size(t, i)) { + case 4: t->size_masks[4-1] |= 1<size_masks[3-1] |= 1<size_masks[2-1] |= 1<size_masks[1-1] |= 1<vb.info.sizes[attr/16]; + unsigned sizes_bits = (sizes_dword>>((attr%16)*2)) & 0x3; + return sizes_bits + 1; +/* return brw->vb.inputs[attr].glarray->Size; */ +} + +/* Calculate sizes of vertex program outputs. Size is the largest + * component index which might vary from [0,0,0,1] + */ +static void calc_wm_input_sizes( struct brw_context *brw ) +{ + /* BRW_NEW_VERTEX_PROGRAM */ + struct brw_vertex_program *vp = + (struct brw_vertex_program *)brw->vertex_program; + /* BRW_NEW_INPUT_DIMENSIONS */ + struct tracker t; + unsigned insn; + unsigned i; + + memset(&t, 0, sizeof(t)); + + /* _NEW_LIGHT */ + if (brw->attribs.Light->Model.TwoSide) + t.twoside = 1; + + for (i = 0; i < PIPE_ATTRIB_MAX; i++) + if (vp->program.Base.InputsRead & (1<program.Base.NumInstructions; insn++) { + struct prog_instruction *inst = &vp->program.Base.Instructions[insn]; + + switch (inst->Opcode) { + case OPCODE_ARL: + break; + + case OPCODE_MOV: + set_active(&t, inst->DstReg, get_active(&t, inst->SrcReg[0])); + break; + + default: + set_active(&t, inst->DstReg, 0xf); + break; + } + } + + calc_sizes(&t); + + if (memcmp(brw->wm.input_size_masks, t.size_masks, sizeof(t.size_masks)) != 0) { + memcpy(brw->wm.input_size_masks, t.size_masks, sizeof(t.size_masks)); + brw->state.dirty.brw |= BRW_NEW_WM_INPUT_DIMENSIONS; + } +} + +const struct brw_tracked_state brw_wm_input_sizes = { + .dirty = { + .mesa = _NEW_LIGHT, + .brw = BRW_NEW_VERTEX_PROGRAM | BRW_NEW_INPUT_DIMENSIONS, + .cache = 0 + }, + .update = calc_wm_input_sizes +}; +#endif diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c new file mode 100644 index 0000000000..4121eae632 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -0,0 +1,1243 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_context.h" +#include "brw_vs.h" + +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" + +struct brw_prog_info { + unsigned num_temps; + unsigned num_addrs; + + unsigned writes_psize; + + unsigned pos_idx; + unsigned result_edge_idx; + unsigned edge_flag_idx; + unsigned psize_idx; +}; + +/* Do things as simply as possible. Allocate and populate all regs + * ahead of time. + */ +static void brw_vs_alloc_regs( struct brw_vs_compile *c, + struct brw_prog_info *info ) +{ + unsigned i, reg = 0, mrf; + unsigned nr_params; + + /* r0 -- reserved as usual + */ + c->r0 = brw_vec8_grf(reg, 0); reg++; + + /* User clip planes from curbe: + */ + if (c->key.nr_userclip) { + for (i = 0; i < c->key.nr_userclip; i++) { + c->userplane[i] = stride( brw_vec4_grf(reg+3+i/2, (i%2) * 4), 0, 4, 1); + } + + /* Deal with curbe alignment: + */ + reg += ((6+c->key.nr_userclip+3)/4)*2; + } + + /* Vertex program parameters from curbe: + */ + nr_params = c->vp->program.num_inputs; /*FIXME: i think this is wrong... */ + for (i = 0; i < nr_params; i++) { + c->regs[TGSI_FILE_INPUT][i] = stride( brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1); + } + reg += (nr_params+1)/2; + + c->prog_data.curb_read_length = reg - 1; + + + + /* Allocate input regs: + */ + c->nr_inputs = 0; + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (c->prog_data.inputs_read & (1<nr_inputs++; + c->regs[TGSI_FILE_INPUT][i] = brw_vec8_grf(reg, 0); + reg++; + } + } + + + /* Allocate outputs: TODO: could organize the non-position outputs + * to go straight into message regs. + */ + c->nr_outputs = 0; + c->first_output = reg; + mrf = 4; + for (i = 0; i < c->vp->program.num_outputs; i++) { + if (c->prog_data.outputs_written & (1<nr_outputs++; +#if 0 + if (i == VERT_RESULT_HPOS) { + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); + reg++; + } + else if (i == VERT_RESULT_PSIZ) { + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); + reg++; + mrf++; /* just a placeholder? XXX fix later stages & remove this */ + } + else { + c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf); + mrf++; + } +#else + /* for now stuff everything in grf */ + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); + reg++; +#endif + } + } + + /* Allocate program temporaries: + */ + for (i = 0; i < info->num_temps; i++) { + c->regs[TGSI_FILE_TEMPORARY][i] = brw_vec8_grf(reg, 0); + reg++; + } + + /* Address reg(s). Don't try to use the internal address reg until + * deref time. + */ + for (i = 0; i < info->num_addrs; i++) { + c->regs[TGSI_FILE_ADDRESS][i] = brw_reg(BRW_GENERAL_REGISTER_FILE, + reg, + 0, + BRW_REGISTER_TYPE_D, + BRW_VERTICAL_STRIDE_8, + BRW_WIDTH_8, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XXXX, + TGSI_WRITEMASK_X); + reg++; + } + + for (i = 0; i < 128; i++) { + if (c->output_regs[i].used_in_src) { + c->output_regs[i].reg = brw_vec8_grf(reg, 0); + reg++; + } + } + + c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg, 0); + reg += 2; + + + /* Some opcodes need an internal temporary: + */ + c->first_tmp = reg; + c->last_tmp = reg; /* for allocation purposes */ + + /* Each input reg holds data from two vertices. The + * urb_read_length is the number of registers read from *each* + * vertex urb, so is half the amount: + */ + c->prog_data.urb_read_length = (c->nr_inputs+1)/2; + + c->prog_data.urb_entry_size = (c->nr_outputs+2+3)/4; + c->prog_data.total_grf = reg; +} + + +static struct brw_reg get_tmp( struct brw_vs_compile *c ) +{ + struct brw_reg tmp = brw_vec8_grf(c->last_tmp, 0); + + if (++c->last_tmp > c->prog_data.total_grf) + c->prog_data.total_grf = c->last_tmp; + + return tmp; +} + +static void release_tmp( struct brw_vs_compile *c, struct brw_reg tmp ) +{ + if (tmp.nr == c->last_tmp-1) + c->last_tmp--; +} + +static void release_tmps( struct brw_vs_compile *c ) +{ + c->last_tmp = c->first_tmp; +} + + +static void unalias1( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + void (*func)( struct brw_vs_compile *, + struct brw_reg, + struct brw_reg )) +{ + if (dst.file == arg0.file && dst.nr == arg0.nr) { + struct brw_compile *p = &c->func; + struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask); + func(c, tmp, arg0); + brw_MOV(p, dst, tmp); + } + else { + func(c, dst, arg0); + } +} + +static void unalias2( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + void (*func)( struct brw_vs_compile *, + struct brw_reg, + struct brw_reg, + struct brw_reg )) +{ + if ((dst.file == arg0.file && dst.nr == arg0.nr) || + (dst.file == arg1.file && dst.nr == arg1.nr)) { + struct brw_compile *p = &c->func; + struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask); + func(c, tmp, arg0, arg1); + brw_MOV(p, dst, tmp); + } + else { + func(c, dst, arg0, arg1); + } +} + +static void emit_sop( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + unsigned cond) +{ + brw_push_insn_state(p); + brw_CMP(p, brw_null_reg(), cond, arg0, arg1); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_MOV(p, dst, brw_imm_f(1.0f)); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, brw_imm_f(0.0f)); + brw_pop_insn_state(p); +} + +static void emit_seq( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_EQ); +} + +static void emit_sne( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_NEQ); +} +static void emit_slt( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_L); +} + +static void emit_sle( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_LE); +} + +static void emit_sgt( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_G); +} + +static void emit_sge( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_GE); +} + +static void emit_max( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1); + brw_SEL(p, dst, arg1, arg0); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +} + +static void emit_min( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1); + brw_SEL(p, dst, arg0, arg1); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +} + + +static void emit_math1( struct brw_vs_compile *c, + unsigned function, + struct brw_reg dst, + struct brw_reg arg0, + unsigned precision) +{ + /* There are various odd behaviours with SEND on the simulator. In + * addition there are documented issues with the fact that the GEN4 + * processor doesn't do dependency control properly on SEND + * results. So, on balance, this kludge to get around failures + * with writemasked math results looks like it might be necessary + * whether that turns out to be a simulator bug or not: + */ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + boolean need_tmp = (dst.dw1.bits.writemask != 0xf || + dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_math(p, + tmp, + function, + BRW_MATH_SATURATE_NONE, + 2, + arg0, + BRW_MATH_DATA_SCALAR, + precision); + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + +static void emit_math2( struct brw_vs_compile *c, + unsigned function, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + unsigned precision) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + boolean need_tmp = (dst.dw1.bits.writemask != 0xf || + dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_MOV(p, brw_message_reg(3), arg1); + + brw_math(p, + tmp, + function, + BRW_MATH_SATURATE_NONE, + 2, + arg0, + BRW_MATH_DATA_SCALAR, + precision); + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + + + +static void emit_exp_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_X) { + struct brw_reg tmp = get_tmp(c); + struct brw_reg tmp_d = retype(tmp, BRW_REGISTER_TYPE_D); + + /* tmp_d = floor(arg0.x) */ + brw_RNDD(p, tmp_d, brw_swizzle1(arg0, 0)); + + /* result[0] = 2.0 ^ tmp */ + + /* Adjust exponent for floating point: + * exp += 127 + */ + brw_ADD(p, brw_writemask(tmp_d, TGSI_WRITEMASK_X), tmp_d, brw_imm_d(127)); + + /* Install exponent and sign. + * Excess drops off the edge: + */ + brw_SHL(p, brw_writemask(retype(dst, BRW_REGISTER_TYPE_D), TGSI_WRITEMASK_X), + tmp_d, brw_imm_d(23)); + + release_tmp(c, tmp); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Y) { + /* result[1] = arg0.x - floor(arg0.x) */ + brw_FRC(p, brw_writemask(dst, TGSI_WRITEMASK_Y), brw_swizzle1(arg0, 0)); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) { + /* As with the LOG instruction, we might be better off just + * doing a taylor expansion here, seeing as we have to do all + * the prep work. + * + * If mathbox partial precision is too low, consider also: + * result[3] = result[0] * EXP(result[1]) + */ + emit_math1(c, + BRW_MATH_FUNCTION_EXP, + brw_writemask(dst, TGSI_WRITEMASK_Z), + brw_swizzle1(arg0, 0), + BRW_MATH_PRECISION_PARTIAL); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) { + /* result[3] = 1.0; */ + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_W), brw_imm_f(1)); + } +} + + +static void emit_log_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + struct brw_reg tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD); + struct brw_reg arg0_ud = retype(arg0, BRW_REGISTER_TYPE_UD); + boolean need_tmp = (dst.dw1.bits.writemask != 0xf || + dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) { + tmp = get_tmp(c); + tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD); + } + + /* Perform mant = frexpf(fabsf(x), &exp), adjust exp and mnt + * according to spec: + * + * These almost look likey they could be joined up, but not really + * practical: + * + * result[0].f = (x.i & ((1<<31)-1) >> 23) - 127 + * result[1].i = (x.i & ((1<<23)-1) + (127<<23) + */ + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_XZ) { + brw_AND(p, + brw_writemask(tmp_ud, TGSI_WRITEMASK_X), + brw_swizzle1(arg0_ud, 0), + brw_imm_ud((1U<<31)-1)); + + brw_SHR(p, + brw_writemask(tmp_ud, TGSI_WRITEMASK_X), + tmp_ud, + brw_imm_ud(23)); + + brw_ADD(p, + brw_writemask(tmp, TGSI_WRITEMASK_X), + retype(tmp_ud, BRW_REGISTER_TYPE_D), /* does it matter? */ + brw_imm_d(-127)); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_YZ) { + brw_AND(p, + brw_writemask(tmp_ud, TGSI_WRITEMASK_Y), + brw_swizzle1(arg0_ud, 0), + brw_imm_ud((1<<23)-1)); + + brw_OR(p, + brw_writemask(tmp_ud, TGSI_WRITEMASK_Y), + tmp_ud, + brw_imm_ud(127<<23)); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) { + /* result[2] = result[0] + LOG2(result[1]); */ + + /* Why bother? The above is just a hint how to do this with a + * taylor series. Maybe we *should* use a taylor series as by + * the time all the above has been done it's almost certainly + * quicker than calling the mathbox, even with low precision. + * + * Options are: + * - result[0] + mathbox.LOG2(result[1]) + * - mathbox.LOG2(arg0.x) + * - result[0] + inline_taylor_approx(result[1]) + */ + emit_math1(c, + BRW_MATH_FUNCTION_LOG, + brw_writemask(tmp, TGSI_WRITEMASK_Z), + brw_swizzle1(tmp, 1), + BRW_MATH_PRECISION_FULL); + + brw_ADD(p, + brw_writemask(tmp, TGSI_WRITEMASK_Z), + brw_swizzle1(tmp, 2), + brw_swizzle1(tmp, 0)); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) { + /* result[3] = 1.0; */ + brw_MOV(p, brw_writemask(tmp, TGSI_WRITEMASK_W), brw_imm_f(1)); + } + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + + + + +/* Need to unalias - consider swizzles: r0 = DST r0.xxxx r1 + */ +static void emit_dst_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1) +{ + struct brw_compile *p = &c->func; + + /* There must be a better way to do this: + */ + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_X) + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_X), brw_imm_f(1.0)); + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Y) + brw_MUL(p, brw_writemask(dst, TGSI_WRITEMASK_Y), arg0, arg1); + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_Z), arg0); + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_W), arg1); +} + +static void emit_xpd( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg t, + struct brw_reg u) +{ + brw_MUL(p, brw_null_reg(), brw_swizzle(t, 1,2,0,3), brw_swizzle(u,2,0,1,3)); + brw_MAC(p, dst, negate(brw_swizzle(t, 2,0,1,3)), brw_swizzle(u,1,2,0,3)); +} + + + +static void emit_lit_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *if_insn; + struct brw_reg tmp = dst; + boolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_YZ), brw_imm_f(0)); + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_XW), brw_imm_f(1)); + + /* Need to use BRW_EXECUTE_8 and also do an 8-wide compare in order + * to get all channels active inside the IF. In the clipping code + * we run with NoMask, so it's not an option and we can use + * BRW_EXECUTE_1 for all comparisions. + */ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,0), brw_imm_f(0)); + if_insn = brw_IF(p, BRW_EXECUTE_8); + { + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_Y), brw_swizzle1(arg0,0)); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,1), brw_imm_f(0)); + brw_MOV(p, brw_writemask(tmp, TGSI_WRITEMASK_Z), brw_swizzle1(arg0,1)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + emit_math2(c, + BRW_MATH_FUNCTION_POW, + brw_writemask(dst, TGSI_WRITEMASK_Z), + brw_swizzle1(tmp, 2), + brw_swizzle1(arg0, 3), + BRW_MATH_PRECISION_PARTIAL); + } + + brw_ENDIF(p, if_insn); +} + + + + + +/* TODO: relative addressing! + */ +static struct brw_reg get_reg( struct brw_vs_compile *c, + unsigned file, + unsigned index ) +{ + switch (file) { + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_INPUT: + case TGSI_FILE_OUTPUT: + assert(c->regs[file][index].nr != 0); + return c->regs[file][index]; + case TGSI_FILE_CONSTANT: + assert(c->regs[TGSI_FILE_CONSTANT][index].nr != 0); + return c->regs[TGSI_FILE_CONSTANT][index]; + case TGSI_FILE_ADDRESS: + assert(index == 0); + return c->regs[file][index]; + + case TGSI_FILE_NULL: /* undef values */ + return brw_null_reg(); + + default: + assert(0); + return brw_null_reg(); + } +} + + + +static struct brw_reg deref( struct brw_vs_compile *c, + struct brw_reg arg, + int offset) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = vec4(get_tmp(c)); + struct brw_reg vp_address = retype(vec1(get_reg(c, TGSI_FILE_ADDRESS, 0)), BRW_REGISTER_TYPE_UW); + unsigned byte_offset = arg.nr * 32 + arg.subnr + offset * 16; + struct brw_reg indirect = brw_vec4_indirect(0,0); + + { + brw_push_insn_state(p); + brw_set_access_mode(p, BRW_ALIGN_1); + + /* This is pretty clunky - load the address register twice and + * fetch each 4-dword value in turn. There must be a way to do + * this in a single pass, but I couldn't get it to work. + */ + brw_ADD(p, brw_address_reg(0), vp_address, brw_imm_d(byte_offset)); + brw_MOV(p, tmp, indirect); + + brw_ADD(p, brw_address_reg(0), suboffset(vp_address, 8), brw_imm_d(byte_offset)); + brw_MOV(p, suboffset(tmp, 4), indirect); + + brw_pop_insn_state(p); + } + + return vec8(tmp); +} + + +static void emit_arl( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + boolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_RNDD(p, tmp, arg0); + brw_MUL(p, dst, tmp, brw_imm_d(16)); + + if (need_tmp) + release_tmp(c, tmp); +} + + +/* Will return mangled results for SWZ op. The emit_swz() function + * ignores this result and recalculates taking extended swizzles into + * account. + */ +static struct brw_reg get_arg( struct brw_vs_compile *c, + struct tgsi_src_register *src ) +{ + struct brw_reg reg; + + if (src->File == TGSI_FILE_NULL) + return brw_null_reg(); + +#if 0 + if (src->RelAddr) + reg = deref(c, c->regs[PROGRAM_STATE_VAR][0], src->Index); + else +#endif + reg = get_reg(c, src->File, src->Index); + + /* Convert 3-bit swizzle to 2-bit. + */ + reg.dw1.bits.swizzle = BRW_SWIZZLE4(src->SwizzleX, + src->SwizzleY, + src->SwizzleZ, + src->SwizzleW); + + /* Note this is ok for non-swizzle instructions: + */ + reg.negate = src->Negate ? 1 : 0; + + return reg; +} + + +static struct brw_reg get_dst( struct brw_vs_compile *c, + struct tgsi_dst_register dst ) +{ + struct brw_reg reg = get_reg(c, dst.File, dst.Index); + + reg.dw1.bits.writemask = dst.WriteMask; + + return reg; +} + + + + +static void emit_swz( struct brw_vs_compile *c, + struct brw_reg dst, + struct tgsi_src_register src ) +{ + struct brw_compile *p = &c->func; + unsigned zeros_mask = 0; + unsigned ones_mask = 0; + unsigned src_mask = 0; + ubyte src_swz[4]; + boolean need_tmp = (src.Negate && + dst.file != BRW_GENERAL_REGISTER_FILE); + struct brw_reg tmp = dst; + unsigned i; + + if (need_tmp) + tmp = get_tmp(c); + + for (i = 0; i < 4; i++) { + if (dst.dw1.bits.writemask & (1<regs[PROGRAM_STATE_VAR][0], src.Index); + else +#endif + arg0 = get_reg(c, src.File, src.Index); + + arg0 = brw_swizzle(arg0, + src_swz[0], src_swz[1], + src_swz[2], src_swz[3]); + + brw_MOV(p, brw_writemask(tmp, src_mask), arg0); + } + + if (zeros_mask) + brw_MOV(p, brw_writemask(tmp, zeros_mask), brw_imm_f(0)); + + if (ones_mask) + brw_MOV(p, brw_writemask(tmp, ones_mask), brw_imm_f(1)); + + if (src.Negate) + brw_MOV(p, brw_writemask(tmp, src.Negate), negate(tmp)); + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + + + +/* Post-vertex-program processing. Send the results to the URB. + */ +static void emit_vertex_write( struct brw_vs_compile *c, struct brw_prog_info *info) +{ + struct brw_compile *p = &c->func; + struct brw_reg m0 = brw_message_reg(0); + struct brw_reg pos = c->regs[TGSI_FILE_OUTPUT][info->pos_idx]; + struct brw_reg ndc; + + if (c->key.copy_edgeflag) { + brw_MOV(p, + get_reg(c, TGSI_FILE_OUTPUT, info->result_edge_idx), + get_reg(c, TGSI_FILE_INPUT, info->edge_flag_idx)); + } + + + /* Build ndc coords? TODO: Shortcircuit when w is known to be one. + */ + if (!c->key.know_w_is_one) { + ndc = get_tmp(c); + emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL); + brw_MUL(p, brw_writemask(ndc, TGSI_WRITEMASK_XYZ), pos, ndc); + } + else { + ndc = pos; + } + + /* This includes the workaround for -ve rhw, so is no longer an + * optional step: + */ + if (info->writes_psize || + c->key.nr_userclip || + !c->key.know_w_is_one) + { + struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + unsigned i; + + brw_MOV(p, header1, brw_imm_ud(0)); + + brw_set_access_mode(p, BRW_ALIGN_16); + + if (info->writes_psize) { + struct brw_reg psiz = c->regs[TGSI_FILE_OUTPUT][info->psize_idx]; + brw_MUL(p, brw_writemask(header1, TGSI_WRITEMASK_W), + brw_swizzle1(psiz, 0), brw_imm_f(1<<11)); + brw_AND(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, + brw_imm_ud(0x7ff<<8)); + } + + + for (i = 0; i < c->key.nr_userclip; i++) { + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, brw_null_reg(), pos, c->userplane[i]); + brw_OR(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, brw_imm_ud(1<key.know_w_is_one) { + brw_CMP(p, + vec8(brw_null_reg()), + BRW_CONDITIONAL_L, + brw_swizzle1(ndc, 3), + brw_imm_f(0)); + + brw_OR(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, brw_imm_ud(1<<6)); + brw_MOV(p, ndc, brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + + brw_set_access_mode(p, BRW_ALIGN_1); /* why? */ + brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), header1); + brw_set_access_mode(p, BRW_ALIGN_16); + + release_tmp(c, header1); + } + else { + brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), brw_imm_ud(0)); + } + + + /* Emit the (interleaved) headers for the two vertices - an 8-reg + * of zeros followed by two sets of NDC coordinates: + */ + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, offset(m0, 2), ndc); + brw_MOV(p, offset(m0, 3), pos); + + + brw_urb_WRITE(p, + brw_null_reg(), /* dest */ + 0, /* starting mrf reg nr */ + c->r0, /* src */ + 0, /* allocate */ + 1, /* used */ + c->nr_outputs + 3, /* msg len */ + 0, /* response len */ + 1, /* eot */ + 1, /* writes complete */ + 0, /* urb destination offset */ + BRW_URB_SWIZZLE_INTERLEAVE); + +} + +static void +post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst ) +{ + struct tgsi_parse_context parse; + const struct tgsi_token *tokens = c->vp->program.tokens; + struct brw_instruction *brw_inst1, *brw_inst2; + const struct tgsi_full_instruction *inst1, *inst2; + int offset; + tgsi_parse_init(&parse, tokens); + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + if (parse.FullToken.Token.Type == TGSI_TOKEN_TYPE_INSTRUCTION) { +#if 0 + inst1 = &parse.FullToken.FullInstruction; + brw_inst1 = inst1->Data; + switch (inst1->Opcode) { + case TGSI_OPCODE_CAL: + case TGSI_OPCODE_BRA: + target_insn = inst1->BranchTarget; + inst2 = &c->vp->program.Base.Instructions[target_insn]; + brw_inst2 = inst2->Data; + offset = brw_inst2 - brw_inst1; + brw_set_src1(brw_inst1, brw_imm_d(offset*16)); + break; + case TGSI_OPCODE_END: + offset = end_inst - brw_inst1; + brw_set_src1(brw_inst1, brw_imm_d(offset*16)); + break; + default: + break; + } +#endif + } + } + tgsi_parse_free(&parse); +} + +/* Emit the fragment program instructions here. + */ +void brw_vs_emit(struct brw_vs_compile *c) +{ +#define MAX_IFSN 32 + struct brw_compile *p = &c->func; + unsigned insn, if_insn = 0; + struct brw_instruction *end_inst; + struct brw_instruction *if_inst[MAX_IFSN]; + struct brw_indirect stack_index = brw_indirect(0, 0); + struct tgsi_parse_context parse; + const struct tgsi_token *tokens = c->vp->program.tokens; + struct brw_prog_info prog_info; + unsigned allocated_registers = 0; + memset(&prog_info, 0, sizeof(struct brw_prog_info)); + + unsigned index; + unsigned file; + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_access_mode(p, BRW_ALIGN_16); + +#if 0 + tgsi_parse_init(&parse, tokens); + /* Message registers can't be read, so copy the output into GRF register + if they are used in source registers */ + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + unsigned i; + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_INSTRUCTION: { + const struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction; + struct prog_src_register *src = &inst->SrcReg[i]; + unsigned index = src->Index; + unsigned file = src->File; + if (file == TGSI_FILE_OUTPUT) + c->output_regs[index].used_in_src = TRUE; + } + break; + default: + /* nothing */ + break; + } + } +#endif + + tgsi_parse_init(&parse, tokens); + + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: { + struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; + /* FIXME: fill in brw_prog_info based on declarations here */ + } + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: { + + } + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: { + if (!allocated_registers) { + /* first instruction after declerations. + * know that we know used vars allocate + * registers */ + brw_vs_alloc_regs(c, &prog_info); + brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); + allocated_registers = 1; + } + } + break; + } + } +#if 0 + struct prog_instruction *inst = &c->vp->program.Base.Instructions[insn]; + struct brw_reg args[3], dst; + unsigned i; + + /* Get argument regs. SWZ is special and does this itself. + */ + inst->Data = &p->store[p->nr_insn]; + if (inst->Opcode != TGSI_OPCODE_SWZ) + for (i = 0; i < 3; i++) { + struct prog_src_register *src = &inst->SrcReg[i]; + index = src->Index; + file = src->File; + if (file == TGSI_FILE_OUTPUT&&c->output_regs[index].used_in_src) + args[i] = c->output_regs[index].reg; + else + args[i] = get_arg(c, src); + } + + /* Get dest regs. Note that it is possible for a reg to be both + * dst and arg, given the static allocation of registers. So + * care needs to be taken emitting multi-operation instructions. + */ + index = inst->DstReg.Index; + file = inst->DstReg.File; + if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src) + dst = c->output_regs[index].reg; + else + dst = get_dst(c, inst->DstReg); + + switch (inst->Opcode) { + case TGSI_OPCODE_ABS: + brw_MOV(p, dst, brw_abs(args[0])); + break; + case TGSI_OPCODE_ADD: + brw_ADD(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DP3: + brw_DP3(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DP4: + brw_DP4(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DPH: + brw_DPH(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DST: + unalias2(c, dst, args[0], args[1], emit_dst_noalias); + break; + case TGSI_OPCODE_EXP: + unalias1(c, dst, args[0], emit_exp_noalias); + break; + case TGSI_OPCODE_EX2: + emit_math1(c, BRW_MATH_FUNCTION_EXP, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_ARL: + emit_arl(c, dst, args[0]); + break; + case TGSI_OPCODE_FLR: + brw_RNDD(p, dst, args[0]); + break; + case TGSI_OPCODE_FRC: + brw_FRC(p, dst, args[0]); + break; + case TGSI_OPCODE_LOG: + unalias1(c, dst, args[0], emit_log_noalias); + break; + case TGSI_OPCODE_LG2: + emit_math1(c, BRW_MATH_FUNCTION_LOG, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_LIT: + unalias1(c, dst, args[0], emit_lit_noalias); + break; + case TGSI_OPCODE_MAD: + brw_MOV(p, brw_acc_reg(), args[2]); + brw_MAC(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MAX: + emit_max(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MIN: + emit_min(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MOV: + brw_MOV(p, dst, args[0]); + break; + case TGSI_OPCODE_MUL: + brw_MUL(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_POW: + emit_math2(c, BRW_MATH_FUNCTION_POW, dst, args[0], args[1], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_RCP: + emit_math1(c, BRW_MATH_FUNCTION_INV, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_RSQ: + emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + + case TGSI_OPCODE_SEQ: + emit_seq(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SNE: + emit_sne(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SGE: + emit_sge(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SGT: + emit_sgt(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SLT: + emit_slt(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SLE: + emit_sle(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SUB: + brw_ADD(p, dst, args[0], negate(args[1])); + break; + case TGSI_OPCODE_SWZ: + /* The args[0] value can't be used here as it won't have + * correctly encoded the full swizzle: + */ + emit_swz(c, dst, inst->SrcReg[0] ); + break; + case TGSI_OPCODE_XPD: + emit_xpd(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_IF: + assert(if_insn < MAX_IFSN); + if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8); + break; + case TGSI_OPCODE_ELSE: + if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]); + break; + case TGSI_OPCODE_ENDIF: + assert(if_insn > 0); + brw_ENDIF(p, if_inst[--if_insn]); + break; + case TGSI_OPCODE_BRA: + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + brw_set_predicate_control_flag_value(p, 0xff); + break; + case TGSI_OPCODE_CAL: + brw_set_access_mode(p, BRW_ALIGN_1); + brw_ADD(p, deref_1uw(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(4)); + inst->Data = &p->store[p->nr_insn]; + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + break; + case TGSI_OPCODE_RET: + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(-4)); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, brw_ip_reg(), deref_1uw(stack_index, 0)); + brw_set_access_mode(p, BRW_ALIGN_16); + case TGSI_OPCODE_END: + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + break; + case TGSI_OPCODE_BGNSUB: + case TGSI_OPCODE_ENDSUB: + break; + default: + printf("Unsupport opcode %d in vertex shader\n", inst->Opcode); + break; + } + + if (inst->DstReg.File == TGSI_FILE_OUTPUT + && inst->DstReg.Index != VERT_RESULT_HPOS + && c->output_regs[inst->DstReg.Index].used_in_src) + brw_MOV(p, get_dst(c, inst->DstReg), dst); + + release_tmps(c); + } +#endif + + end_inst = &p->store[p->nr_insn]; + emit_vertex_write(c, &prog_info); + post_vs_emit(c, end_inst); + tgsi_parse_free(&parse); +} diff --git a/src/mesa/pipe/i965simple/brw_vs_state.c b/src/mesa/pipe/i965simple/brw_vs_state.c new file mode 100644 index 0000000000..7d6fb383b9 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_vs_state.c @@ -0,0 +1,102 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + +#include "pipe/p_util.h" + +static void upload_vs_unit( struct brw_context *brw ) +{ + struct brw_vs_unit_state vs; + + memset(&vs, 0, sizeof(vs)); + + /* CACHE_NEW_VS_PROG */ + vs.thread0.kernel_start_pointer = brw->vs.prog_gs_offset >> 6; + vs.thread0.grf_reg_count = ALIGN(brw->vs.prog_data->total_grf, 16) / 16 - 1; + vs.thread3.urb_entry_read_length = brw->vs.prog_data->urb_read_length; + vs.thread3.const_urb_entry_read_length = brw->vs.prog_data->curb_read_length; + vs.thread3.dispatch_grf_start_reg = 1; + + + /* BRW_NEW_URB_FENCE */ + vs.thread4.nr_urb_entries = brw->urb.nr_vs_entries; + vs.thread4.urb_entry_allocation_size = brw->urb.vsize - 1; + vs.thread4.max_threads = MIN2( + MAX2(0, (brw->urb.nr_vs_entries - 6) / 2 - 1), + 15); + + + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + vs.thread4.max_threads = 0; + + /* BRW_NEW_CURBE_OFFSETS, _NEW_TRANSFORM */ + if (0 /*brw->attribs.Clip->ClipPlanesEnabled*/) { + /* Note that we read in the userclip planes as well, hence + * clip_start: + */ + vs.thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2; + } + else { + vs.thread3.const_urb_entry_read_offset = brw->curbe.vs_start * 2; + } + + vs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + vs.thread3.urb_entry_read_offset = 0; + + /* No samplers for ARB_vp programs: + */ + vs.vs5.sampler_count = 0; + + if (BRW_DEBUG & DEBUG_STATS) + vs.thread4.stats_enable = 1; + + /* Vertex program always enabled: + */ + vs.vs6.vs_enable = 1; + + brw->vs.state_gs_offset = brw_cache_data( &brw->cache[BRW_VS_UNIT], &vs ); +} + + +const struct brw_tracked_state brw_vs_unit = { + .dirty = { + .brw = (BRW_NEW_CLIP | + BRW_NEW_CURBE_OFFSETS | + BRW_NEW_URB_FENCE), + .cache = CACHE_NEW_VS_PROG + }, + .update = upload_vs_unit +}; diff --git a/src/mesa/pipe/i965simple/brw_vtbl.c b/src/mesa/pipe/i965simple/brw_vtbl.c new file mode 100644 index 0000000000..6dc3bd838b --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_vtbl.c @@ -0,0 +1,149 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_state.h" + +#include "brw_draw.h" +#include "brw_state.h" +#include "brw_vs.h" +#include + +#if 0 +/* called from intelDestroyContext() + */ +static void brw_destroy_context( struct intel_context *intel ) +{ + GLcontext *ctx = &intel->ctx; + struct brw_context *brw = brw_context(&intel->ctx); + + brw_destroy_metaops(brw); + brw_destroy_state(brw); + brw_draw_destroy( brw ); + + brw_ProgramCacheDestroy( ctx ); + brw_FrameBufferTexDestroy( brw ); +} + +/* called from intelDrawBuffer() + */ +static void brw_set_draw_region( struct intel_context *intel, + struct intel_region *draw_region, + struct intel_region *depth_region) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + intel_region_release(&brw->state.draw_region); + intel_region_release(&brw->state.depth_region); + intel_region_reference(&brw->state.draw_region, draw_region); + intel_region_reference(&brw->state.depth_region, depth_region); +} + + +/* called from intelFlushBatchLocked + */ +static void brw_lost_hardware( struct intel_context *intel ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + /* Note that we effectively lose the context after this. + * + * Setting this flag provokes a state buffer wrap and also flushes + * the hardware caches. + */ + brw->state.dirty.brw |= BRW_NEW_CONTEXT; + + /* Which means there shouldn't be any commands already queued: + */ + assert(intel->batch->ptr == intel->batch->map); + + brw->state.dirty.mesa |= ~0; + brw->state.dirty.brw |= ~0; + brw->state.dirty.cache |= ~0; +} + +static void brw_note_fence( struct intel_context *intel, + unsigned fence ) +{ + brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_FENCE; +} + +static void brw_note_unlock( struct intel_context *intel ) +{ + struct brw_context *brw = brw_context(&intel->ctx); + + brw_pool_check_wrap(brw, &brw->pool[BRW_GS_POOL]); + brw_pool_check_wrap(brw, &brw->pool[BRW_SS_POOL]); + + brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_LOCK; +} + + +void brw_do_flush( struct brw_context *brw, + unsigned flags ) +{ + struct brw_mi_flush flush; + memset(&flush, 0, sizeof(flush)); + flush.opcode = CMD_MI_FLUSH; + flush.flags = flags; + BRW_BATCH_STRUCT(brw, &flush); +} + + +static void brw_emit_flush( struct intel_context *intel, + unsigned unused ) +{ + brw_do_flush(brw_context(&intel->ctx), + BRW_FLUSH_STATE_CACHE|BRW_FLUSH_READ_CACHE); +} + + +/* called from intelWaitForIdle() and intelFlush() + * + * For now, just flush everything. Could be smarter later. + */ +static unsigned brw_flush_cmd( void ) +{ + struct brw_mi_flush flush; + flush.opcode = CMD_MI_FLUSH; + flush.pad = 0; + flush.flags = BRW_FLUSH_READ_CACHE | BRW_FLUSH_STATE_CACHE; + return *(unsigned *)&flush; +} + +static void brw_invalidate_state( struct intel_context *intel, unsigned new_state ) +{ + /* nothing */ +} +#endif diff --git a/src/mesa/pipe/i965simple/brw_winsys.h b/src/mesa/pipe/i965simple/brw_winsys.h new file mode 100644 index 0000000000..cc0a210a9f --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_winsys.h @@ -0,0 +1,192 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * This is the interface that i965simple requires any window system + * hosting it to implement. This is the only include file in i965simple + * which is public. + * + */ + +#ifndef BRW_WINSYS_H +#define BRW_WINSYS_H + + +#include "pipe/p_defines.h" + + +/* Pipe drivers are (meant to be!) independent of both GL and the + * window system. The window system provides a buffer manager and a + * set of additional hooks for things like command buffer submission, + * etc. + * + * There clearly has to be some agreement between the window system + * driver and the hardware driver about the format of command buffers, + * etc. + */ + +struct pipe_buffer_handle; +struct pipe_fence_handle; +struct pipe_winsys; + +/* The pipe driver currently understands the following chipsets: + */ +#define PCI_CHIP_I965_G 0x29A2 +#define PCI_CHIP_I965_Q 0x2992 +#define PCI_CHIP_I965_G_1 0x2982 +#define PCI_CHIP_I965_GM 0x2A02 +#define PCI_CHIP_I965_GME 0x2A12 + + +/* These are the names of all the state caches managed by the driver. + * + * When data is uploaded to a buffer with buffer_subdata, we use the + * special version of that function below so that information about + * what type of data this is can be passed to the winsys backend. + * That in turn allows the correct flags to be set in the aub file + * dump to allow human-readable file dumps later on. + */ + +enum brw_cache_id { + BRW_CC_VP, + BRW_CC_UNIT, + BRW_WM_PROG, + BRW_SAMPLER_DEFAULT_COLOR, + BRW_SAMPLER, + BRW_WM_UNIT, + BRW_SF_PROG, + BRW_SF_VP, + BRW_SF_UNIT, + BRW_VS_UNIT, + BRW_VS_PROG, + BRW_GS_UNIT, + BRW_GS_PROG, + BRW_CLIP_VP, + BRW_CLIP_UNIT, + BRW_CLIP_PROG, + BRW_SS_SURFACE, + BRW_SS_SURF_BIND, + + BRW_MAX_CACHE +}; + +/** + * Additional winsys interface for i965simple. + * + * It is an over-simple batchbuffer mechanism. Will want to improve the + * performance of this, perhaps based on the cmdstream stuff. It + * would be pretty impossible to implement swz on top of this + * interface. + * + * Will also need additions/changes to implement static/dynamic + * indirect state. + */ +struct brw_winsys { + + /** + * Reserve space on batch buffer. + * + * Returns a null pointer if there is insufficient space in the batch buffer + * to hold the requested number of dwords and relocations. + * + * The number of dwords should also include the number of relocations. + */ + unsigned *(*batch_start)(struct brw_winsys *sws, + unsigned dwords, + unsigned relocs); + + void (*batch_dword)(struct brw_winsys *sws, + unsigned dword); + + /** + * Emit a relocation to a buffer. + * + * Used not only when the buffer addresses are not pinned, but also to + * ensure refered buffers will not be destroyed until the current batch + * buffer execution is finished. + * + * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and + * I915_BUFFER_ACCESS_READ macros. + */ + void (*batch_reloc)(struct brw_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned access_flags, + unsigned delta); + + + /* Not used yet, but really want this: + */ + void (*batch_end)( struct brw_winsys *sws ); + + /** + * Flush the batch buffer. + * + * Fence argument must point to NULL or to a previous fence, and the caller + * must call fence_reference when done with the fence. + */ + void (*batch_flush)(struct brw_winsys *sws, + struct pipe_fence_handle **fence); + + + /* A version of buffer_subdata that includes information for the + * simulator: + */ + void (*buffer_subdata_typed)(struct brw_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data, + unsigned data_type); + + +}; + +#define BRW_BUFFER_ACCESS_WRITE 0x1 +#define BRW_BUFFER_ACCESS_READ 0x2 + +#define BRW_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) + + +struct pipe_context *brw_create(struct pipe_winsys *, + struct brw_winsys *, + unsigned pci_id); + +static inline boolean brw_batchbuffer_data(struct brw_winsys *winsys, + const void *data, + unsigned bytes) +{ + static const unsigned incr = sizeof(unsigned); + int i; + const unsigned *udata = (const unsigned*)(data); + unsigned size = bytes/incr; + for (i = 0; i < size; ++i) { + winsys->batch_dword(winsys, udata[i]); + } + return (i == size); +} +#endif diff --git a/src/mesa/pipe/i965simple/brw_wm.c b/src/mesa/pipe/i965simple/brw_wm.c new file mode 100644 index 0000000000..65271f22fd --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm.c @@ -0,0 +1,338 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_util.h" +#include "brw_wm.h" +#include "brw_state.h" + +unsigned brw_wm_nr_args( unsigned opcode ) +{ + switch (opcode) { + + case WM_PIXELXY: + case TGSI_OPCODE_ABS: + case TGSI_OPCODE_FLR: + case TGSI_OPCODE_FRC: + case TGSI_OPCODE_MOV: + case TGSI_OPCODE_COS: + case TGSI_OPCODE_EX2: + case TGSI_OPCODE_LG2: + case TGSI_OPCODE_RCP: + case TGSI_OPCODE_RSQ: + case TGSI_OPCODE_SIN: + case TGSI_OPCODE_SCS: + case TGSI_OPCODE_TEX: + case TGSI_OPCODE_TXB: + case TGSI_OPCODE_TXD: + case TGSI_OPCODE_KIL: + case TGSI_OPCODE_LIT: + case WM_CINTERP: + case WM_WPOSXY: + return 1; + + case TGSI_OPCODE_POW: + case TGSI_OPCODE_SUB: + case TGSI_OPCODE_SGE: + case TGSI_OPCODE_SGT: + case TGSI_OPCODE_SLE: + case TGSI_OPCODE_SLT: + case TGSI_OPCODE_SEQ: + case TGSI_OPCODE_SNE: + case TGSI_OPCODE_ADD: + case TGSI_OPCODE_MAX: + case TGSI_OPCODE_MIN: + case TGSI_OPCODE_MUL: + case TGSI_OPCODE_XPD: + case TGSI_OPCODE_DP3: + case TGSI_OPCODE_DP4: + case TGSI_OPCODE_DPH: + case TGSI_OPCODE_DST: + case WM_LINTERP: + case WM_DELTAXY: + case WM_PIXELW: + return 2; + + case WM_FB_WRITE: + case WM_PINTERP: + case TGSI_OPCODE_MAD: + case TGSI_OPCODE_CMP: + case TGSI_OPCODE_LRP: + return 3; + + default: + return 0; + } +} + + +unsigned brw_wm_is_scalar_result( unsigned opcode ) +{ + switch (opcode) { + case TGSI_OPCODE_COS: + case TGSI_OPCODE_EX2: + case TGSI_OPCODE_LG2: + case TGSI_OPCODE_POW: + case TGSI_OPCODE_RCP: + case TGSI_OPCODE_RSQ: + case TGSI_OPCODE_SIN: + case TGSI_OPCODE_DP3: + case TGSI_OPCODE_DP4: + case TGSI_OPCODE_DPH: + case TGSI_OPCODE_DST: + return 1; + + default: + return 0; + } +} + + +static void do_wm_prog( struct brw_context *brw, + struct brw_fragment_program *fp, + struct brw_wm_prog_key *key) +{ + struct brw_wm_compile *c; + const unsigned *program; + unsigned program_size; + + c = brw->wm.compile_data; + if (c == NULL) { + brw->wm.compile_data = calloc(1, sizeof(*brw->wm.compile_data)); + c = brw->wm.compile_data; + } else { + memset(c, 0, sizeof(*brw->wm.compile_data)); + } + memcpy(&c->key, key, sizeof(*key)); + + c->fp = fp; + fprintf(stderr, "XXXXXXXX FP\n"); + +#if 0 + c->env_param = brw->intel.ctx.FragmentProgram.Parameters; + + if (brw_wm_is_glsl(&c->fp->program)) { + brw_wm_glsl_emit(c); + } else + { + /* Augment fragment program. Add instructions for pre- and + * post-fragment-program tasks such as interpolation and fogging. + */ + brw_wm_pass_fp(c); + + /* Translate to intermediate representation. Build register usage + * chains. + */ + brw_wm_pass0(c); + + /* Dead code removal. + */ + brw_wm_pass1(c); + + /* Register allocation. + */ + c->grf_limit = BRW_WM_MAX_GRF/2; + + /* This is where we start emitting gen4 code: + */ + brw_init_compile(&c->func); + + brw_wm_pass2(c); + + c->prog_data.total_grf = c->max_wm_grf; + if (c->last_scratch) { + c->prog_data.total_scratch = + c->last_scratch + 0x40; + } else { + c->prog_data.total_scratch = 0; + } + + /* Emit GEN4 code. + */ + brw_wm_emit(c); + } + /* get the program + */ + program = brw_get_program(&c->func, &program_size); + + /* + */ + brw->wm.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_WM_PROG], + &c->key, + sizeof(c->key), + program, + program_size, + &c->prog_data, + &brw->wm.prog_data ); +#endif +} + + + +static void brw_wm_populate_key( struct brw_context *brw, + struct brw_wm_prog_key *key ) +{ + /* BRW_NEW_FRAGMENT_PROGRAM */ + struct brw_fragment_program *fp = + (struct brw_fragment_program *)brw->attribs.FragmentProgram; + unsigned lookup = 0; + unsigned line_aa; + unsigned i; + + memset(key, 0, sizeof(*key)); + + /* Build the index for table lookup + */ + /* _NEW_COLOR */ + if (fp->UsesKill || + brw->attribs.AlphaTest->enabled) + lookup |= IZ_PS_KILL_ALPHATEST_BIT; + + if (fp->ComputesDepth) + lookup |= IZ_PS_COMPUTES_DEPTH_BIT; + + /* _NEW_DEPTH */ + if (brw->attribs.DepthStencil->depth.enabled) + lookup |= IZ_DEPTH_TEST_ENABLE_BIT; + + if (brw->attribs.DepthStencil->depth.enabled && + brw->attribs.DepthStencil->depth.writemask) /* ?? */ + lookup |= IZ_DEPTH_WRITE_ENABLE_BIT; + + /* _NEW_STENCIL */ + if (brw->attribs.DepthStencil->stencil.front_enabled) { + lookup |= IZ_STENCIL_TEST_ENABLE_BIT; + + if (brw->attribs.DepthStencil->stencil.write_mask[0] || + (brw->attribs.DepthStencil->stencil.back_enabled && + brw->attribs.DepthStencil->stencil.write_mask[1])) + lookup |= IZ_STENCIL_WRITE_ENABLE_BIT; + } + + /* XXX: when should this be disabled? + */ + if (1) + lookup |= IZ_EARLY_DEPTH_TEST_BIT; + + + line_aa = AA_NEVER; + + /* _NEW_LINE, _NEW_POLYGON, BRW_NEW_REDUCED_PRIMITIVE */ + if (brw->attribs.Raster->line_smooth) { + if (brw->reduced_primitive == PIPE_PRIM_LINES) { + line_aa = AA_ALWAYS; + } + else if (brw->reduced_primitive == PIPE_PRIM_TRIANGLES) { + if (brw->attribs.Raster->fill_ccw == PIPE_POLYGON_MODE_LINE) { + line_aa = AA_SOMETIMES; + + if (brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_LINE || + (brw->attribs.Raster->cull_mode == PIPE_WINDING_CW)) + line_aa = AA_ALWAYS; + } + else if (brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_LINE) { + line_aa = AA_SOMETIMES; + + if (brw->attribs.Raster->cull_mode == PIPE_WINDING_CCW) + line_aa = AA_ALWAYS; + } + } + } + + brw_wm_lookup_iz(line_aa, + lookup, + key); + + +#if 0 + /* BRW_NEW_WM_INPUT_DIMENSIONS */ + key->projtex_mask = brw->wm.input_size_masks[4-1] >> (FRAG_ATTRIB_TEX0 - FRAG_ATTRIB_WPOS); +#endif + + /* _NEW_LIGHT */ + key->flat_shade = (brw->attribs.Raster->flatshade); + + /* _NEW_TEXTURE */ + for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { + const struct pipe_sampler_state *unit = brw->attribs.Samplers[i]; + + if (unit) { + + if (unit->compare && + unit->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + key->shadowtex_mask |= 1<Image[0][t->BaseLevel]->InternalFormat == GL_YCBCR_MESA) + key->yuvtex_mask |= 1<program_string_id = fp->id; + +} + + +static void brw_upload_wm_prog( struct brw_context *brw ) +{ + struct brw_wm_prog_key key; + struct brw_fragment_program *fp = (struct brw_fragment_program *) + brw->attribs.FragmentProgram; + + brw_wm_populate_key(brw, &key); + + /* Make an early check for the key. + */ + if (brw_search_cache(&brw->cache[BRW_WM_PROG], + &key, sizeof(key), + &brw->wm.prog_data, + &brw->wm.prog_gs_offset)) + return; + + do_wm_prog(brw, fp, &key); +} + + +const struct brw_tracked_state brw_wm_prog = { + .dirty = { + .brw = (BRW_NEW_FS | + BRW_NEW_WM_INPUT_DIMENSIONS | + BRW_NEW_REDUCED_PRIMITIVE), + .cache = 0 + }, + .update = brw_upload_wm_prog +}; + diff --git a/src/mesa/pipe/i965simple/brw_wm.h b/src/mesa/pipe/i965simple/brw_wm.h new file mode 100644 index 0000000000..a394e25da3 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm.h @@ -0,0 +1,274 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_WM_H +#define BRW_WM_H + + +#include "brw_context.h" +#include "brw_eu.h" + +/* A big lookup table is used to figure out which and how many + * additional regs will inserted before the main payload in the WM + * program execution. These mainly relate to depth and stencil + * processing and the early-depth-test optimization. + */ +#define IZ_PS_KILL_ALPHATEST_BIT 0x1 +#define IZ_PS_COMPUTES_DEPTH_BIT 0x2 +#define IZ_DEPTH_WRITE_ENABLE_BIT 0x4 +#define IZ_DEPTH_TEST_ENABLE_BIT 0x8 +#define IZ_STENCIL_WRITE_ENABLE_BIT 0x10 +#define IZ_STENCIL_TEST_ENABLE_BIT 0x20 +#define IZ_EARLY_DEPTH_TEST_BIT 0x40 +#define IZ_BIT_MAX 0x80 + +#define AA_NEVER 0 +#define AA_SOMETIMES 1 +#define AA_ALWAYS 2 + +struct brw_wm_prog_key { + unsigned source_depth_reg:3; + unsigned aa_dest_stencil_reg:3; + unsigned dest_depth_reg:3; + unsigned nr_depth_regs:3; + unsigned projtex_mask:8; + unsigned shadowtex_mask:8; + unsigned computes_depth:1; /* could be derived from program string */ + unsigned source_depth_to_render_target:1; + unsigned flat_shade:1; + unsigned runtime_check_aads_emit:1; + + unsigned yuvtex_mask:8; + unsigned pad1:24; + + unsigned program_string_id:32; +}; + + +/* A bit of a glossary: + * + * brw_wm_value: A computed value or program input. Values are + * constant, they are created once and are never modified. When a + * fragment program register is written or overwritten, new values are + * created fresh, preserving the rule that values are constant. + * + * brw_wm_ref: A reference to a value. Wherever a value used is by an + * instruction or as a program output, that is tracked with an + * instance of this struct. All references to a value occur after it + * is created. After the last reference, a value is dead and can be + * discarded. + * + * brw_wm_grf: Represents a physical hardware register. May be either + * empty or hold a value. Register allocation is the process of + * assigning values to grf registers. This occurs in pass2 and the + * brw_wm_grf struct is not used before that. + * + * Fragment program registers: These are time-varying constructs that + * are hard to reason about and which we translate away in pass0. A + * single fragment program register element (eg. temp[0].x) will be + * translated to one or more brw_wm_value structs, one for each time + * that temp[0].x is written to during the program. + */ + + + +/* Used in pass2 to track register allocation. + */ +struct brw_wm_grf { + struct brw_wm_value *value; + unsigned nextuse; +}; + +struct brw_wm_value { + struct brw_reg hw_reg; /* emitted to this reg, may not always be there */ + struct brw_wm_ref *lastuse; + struct brw_wm_grf *resident; + unsigned contributes_to_output:1; + unsigned spill_slot:16; /* if non-zero, spill immediately after calculation */ +}; + +struct brw_wm_ref { + struct brw_reg hw_reg; /* nr filled in in pass2, everything else, pass0 */ + struct brw_wm_value *value; + struct brw_wm_ref *prevuse; + unsigned unspill_reg:7; /* unspill to reg */ + unsigned emitted:1; + unsigned insn:24; +}; + +struct brw_wm_constref { + const struct brw_wm_ref *ref; + float constval; +}; + + +struct brw_wm_instruction { + struct brw_wm_value *dst[4]; + struct brw_wm_ref *src[3][4]; + unsigned opcode:8; + unsigned saturate:1; + unsigned writemask:4; + unsigned tex_unit:4; /* texture unit for TEX, TXD, TXP instructions */ + unsigned tex_idx:3; /* TEXTURE_1D,2D,3D,CUBE,RECT_INDEX source target */ +}; + +#define PROGRAM_INTERNAL_PARAM +#define MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS 1024 /* 72 for GL_ARB_f_p */ +#define BRW_WM_MAX_INSN (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + PIPE_ATTRIB_MAX + 3) +#define BRW_WM_MAX_GRF 128 /* hardware limit */ +#define BRW_WM_MAX_VREG (BRW_WM_MAX_INSN * 4) +#define BRW_WM_MAX_REF (BRW_WM_MAX_INSN * 12) +#define BRW_WM_MAX_PARAM 256 +#define BRW_WM_MAX_CONST 256 +#define BRW_WM_MAX_KILLS MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS + + + +/* New opcodes to track internal operations required for WM unit. + * These are added early so that the registers used can be tracked, + * freed and reused like those of other instructions. + */ +#define WM_PIXELXY (TGSI_OPCODE_LAST) +#define WM_DELTAXY (TGSI_OPCODE_LAST + 1) +#define WM_PIXELW (TGSI_OPCODE_LAST + 2) +#define WM_LINTERP (TGSI_OPCODE_LAST + 3) +#define WM_PINTERP (TGSI_OPCODE_LAST + 4) +#define WM_CINTERP (TGSI_OPCODE_LAST + 5) +#define WM_WPOSXY (TGSI_OPCODE_LAST + 6) +#define WM_FB_WRITE (TGSI_OPCODE_LAST + 7) +#define MAX_WM_OPCODE (TGSI_OPCODE_LAST + 8) + +#define PAYLOAD_DEPTH (PIPE_ATTRIB_MAX) + +struct brw_wm_compile { + struct brw_compile func; + struct brw_wm_prog_key key; + struct brw_wm_prog_data prog_data; + + struct brw_fragment_program *fp; + + float (*env_param)[4]; + + enum { + START, + PASS2_DONE + } state; + + /* Initial pass - translate fp instructions to fp instructions, + * simplifying and adding instructions for interpolation and + * framebuffer writes. + */ + const struct pipe_shader_state *prog_instructions; + unsigned nr_fp_insns; + unsigned fp_temp; + unsigned fp_interp_emitted; + unsigned fp_deriv_emitted; + + struct tgsi_src_register pixel_xy; + struct tgsi_src_register delta_xy; + struct tgsi_src_register pixel_w; + + + struct brw_wm_value vreg[BRW_WM_MAX_VREG]; + unsigned nr_vreg; + + struct brw_wm_value creg[BRW_WM_MAX_PARAM]; + unsigned nr_creg; + + struct { + struct brw_wm_value depth[4]; /* includes r0/r1 */ + struct brw_wm_value input_interp[PIPE_ATTRIB_MAX]; + } payload; + + + const struct brw_wm_ref *pass0_fp_reg[16][256][4]; + + struct brw_wm_ref undef_ref; + struct brw_wm_value undef_value; + + struct brw_wm_ref refs[BRW_WM_MAX_REF]; + unsigned nr_refs; + + struct brw_wm_instruction instruction[BRW_WM_MAX_INSN]; + unsigned nr_insns; + + struct brw_wm_constref constref[BRW_WM_MAX_CONST]; + unsigned nr_constrefs; + + struct brw_wm_grf pass2_grf[BRW_WM_MAX_GRF/2]; + + unsigned grf_limit; + unsigned max_wm_grf; + unsigned last_scratch; + + struct { + boolean inited; + struct brw_reg reg; + } wm_regs[16][256][4]; + struct brw_reg stack; + struct brw_reg emit_mask_reg; + unsigned reg_index; + unsigned tmp_index; +}; + + +unsigned brw_wm_nr_args( unsigned opcode ); +unsigned brw_wm_is_scalar_result( unsigned opcode ); + +void brw_wm_pass_fp( struct brw_wm_compile *c ); +void brw_wm_pass0( struct brw_wm_compile *c ); +void brw_wm_pass1( struct brw_wm_compile *c ); +void brw_wm_pass2( struct brw_wm_compile *c ); +void brw_wm_emit( struct brw_wm_compile *c ); + +void brw_wm_print_value( struct brw_wm_compile *c, + struct brw_wm_value *value ); + +void brw_wm_print_ref( struct brw_wm_compile *c, + struct brw_wm_ref *ref ); + +void brw_wm_print_insn( struct brw_wm_compile *c, + struct brw_wm_instruction *inst ); + +void brw_wm_print_program( struct brw_wm_compile *c, + const char *stage ); + +void brw_wm_lookup_iz( unsigned line_aa, + unsigned lookup, + struct brw_wm_prog_key *key ); + +#if 0 +boolean brw_wm_is_glsl(struct gl_fragment_program *fp); +void brw_wm_glsl_emit(struct brw_wm_compile *c); +#endif + +#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_debug.c b/src/mesa/pipe/i965simple/brw_wm_debug.c new file mode 100644 index 0000000000..b31a608901 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_debug.c @@ -0,0 +1,172 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_wm.h" + +#if 0 +void brw_wm_print_value( struct brw_wm_compile *c, + struct brw_wm_value *value ) +{ + assert(value); + if (c->state >= PASS2_DONE) + brw_print_reg(value->hw_reg); + else if( value == &c->undef_value ) + _mesa_printf("undef"); + else if( value - c->vreg >= 0 && + value - c->vreg < BRW_WM_MAX_VREG) + _mesa_printf("r%d", value - c->vreg); + else if (value - c->creg >= 0 && + value - c->creg < BRW_WM_MAX_PARAM) + _mesa_printf("c%d", value - c->creg); + else if (value - c->payload.input_interp >= 0 && + value - c->payload.input_interp < FRAG_ATTRIB_MAX) + _mesa_printf("i%d", value - c->payload.input_interp); + else if (value - c->payload.depth >= 0 && + value - c->payload.depth < FRAG_ATTRIB_MAX) + _mesa_printf("d%d", value - c->payload.depth); + else + _mesa_printf("?"); +} + +void brw_wm_print_ref( struct brw_wm_compile *c, + struct brw_wm_ref *ref ) +{ + struct brw_reg hw_reg = ref->hw_reg; + + if (ref->unspill_reg) + _mesa_printf("UNSPILL(%x)/", ref->value->spill_slot); + + if (c->state >= PASS2_DONE) + brw_print_reg(ref->hw_reg); + else { + _mesa_printf("%s", hw_reg.negate ? "-" : ""); + _mesa_printf("%s", hw_reg.abs ? "abs/" : ""); + brw_wm_print_value(c, ref->value); + if ((hw_reg.nr&1) || hw_reg.subnr) { + _mesa_printf("->%d.%d", (hw_reg.nr&1), hw_reg.subnr); + } + } +} + +void brw_wm_print_insn( struct brw_wm_compile *c, + struct brw_wm_instruction *inst ) +{ + unsigned i, arg; + unsigned nr_args = brw_wm_nr_args(inst->opcode); + + _mesa_printf("["); + for (i = 0; i < 4; i++) { + if (inst->dst[i]) { + brw_wm_print_value(c, inst->dst[i]); + if (inst->dst[i]->spill_slot) + _mesa_printf("/SPILL(%x)",inst->dst[i]->spill_slot); + } + else + _mesa_printf("#"); + if (i < 3) + _mesa_printf(","); + } + _mesa_printf("]"); + + if (inst->writemask != WRITEMASK_XYZW) + _mesa_printf(".%s%s%s%s", + GET_BIT(inst->writemask, 0) ? "x" : "", + GET_BIT(inst->writemask, 1) ? "y" : "", + GET_BIT(inst->writemask, 2) ? "z" : "", + GET_BIT(inst->writemask, 3) ? "w" : ""); + + switch (inst->opcode) { + case WM_PIXELXY: + _mesa_printf(" = PIXELXY"); + break; + case WM_DELTAXY: + _mesa_printf(" = DELTAXY"); + break; + case WM_PIXELW: + _mesa_printf(" = PIXELW"); + break; + case WM_WPOSXY: + _mesa_printf(" = WPOSXY"); + break; + case WM_PINTERP: + _mesa_printf(" = PINTERP"); + break; + case WM_LINTERP: + _mesa_printf(" = LINTERP"); + break; + case WM_CINTERP: + _mesa_printf(" = CINTERP"); + break; + case WM_FB_WRITE: + _mesa_printf(" = FB_WRITE"); + break; + default: + _mesa_printf(" = %s", _mesa_opcode_string(inst->opcode)); + break; + } + + if (inst->saturate) + _mesa_printf("_SAT"); + + for (arg = 0; arg < nr_args; arg++) { + + _mesa_printf(" ["); + + for (i = 0; i < 4; i++) { + if (inst->src[arg][i]) { + brw_wm_print_ref(c, inst->src[arg][i]); + } + else + _mesa_printf("%%"); + + if (i < 3) + _mesa_printf(","); + else + _mesa_printf("]"); + } + } + _mesa_printf("\n"); +} + +void brw_wm_print_program( struct brw_wm_compile *c, + const char *stage ) +{ + unsigned insn; + + _mesa_printf("\n\n\n%s:\n", stage); + for (insn = 0; insn < c->nr_insns; insn++) + brw_wm_print_insn(c, &c->instruction[insn]); + _mesa_printf("\n\n\n"); +} + +#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_emit.c b/src/mesa/pipe/i965simple/brw_wm_emit.c new file mode 100644 index 0000000000..75bc4cd419 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_emit.c @@ -0,0 +1,1289 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_wm.h" + +#define SATURATE (1<<5) + +/* Not quite sure how correct this is - need to understand horiz + * vs. vertical strides a little better. + */ +static __inline struct brw_reg sechalf( struct brw_reg reg ) +{ + if (reg.vstride) + reg.nr++; + return reg; +} + +/* Payload R0: + * + * R0.0 -- pixel mask, one bit for each of 4 pixels in 4 tiles, + * corresponding to each of the 16 execution channels. + * R0.1..8 -- ? + * R1.0 -- triangle vertex 0.X + * R1.1 -- triangle vertex 0.Y + * R1.2 -- tile 0 x,y coords (2 packed uwords) + * R1.3 -- tile 1 x,y coords (2 packed uwords) + * R1.4 -- tile 2 x,y coords (2 packed uwords) + * R1.5 -- tile 3 x,y coords (2 packed uwords) + * R1.6 -- ? + * R1.7 -- ? + * R1.8 -- ? + */ + + +static void emit_pixel_xy(struct brw_compile *p, + const struct brw_reg *dst, + unsigned mask, + const struct brw_reg *arg0) +{ + struct brw_reg r1 = brw_vec1_grf(1, 0); + struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW); + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + /* Calculate pixel centers by adding 1 or 0 to each of the + * micro-tile coordinates passed in r1. + */ + if (mask & TGSI_WRITEMASK_X) { + brw_ADD(p, + vec16(retype(dst[0], BRW_REGISTER_TYPE_UW)), + stride(suboffset(r1_uw, 4), 2, 4, 0), + brw_imm_v(0x10101010)); + } + + if (mask & TGSI_WRITEMASK_Y) { + brw_ADD(p, + vec16(retype(dst[1], BRW_REGISTER_TYPE_UW)), + stride(suboffset(r1_uw,5), 2, 4, 0), + brw_imm_v(0x11001100)); + } + + brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); +} + + + +static void emit_delta_xy(struct brw_compile *p, + const struct brw_reg *dst, + unsigned mask, + const struct brw_reg *arg0, + const struct brw_reg *arg1) +{ + struct brw_reg r1 = brw_vec1_grf(1, 0); + + /* Calc delta X,Y by subtracting origin in r1 from the pixel + * centers. + */ + if (mask & TGSI_WRITEMASK_X) { + brw_ADD(p, + dst[0], + retype(arg0[0], BRW_REGISTER_TYPE_UW), + negate(r1)); + } + + if (mask & TGSI_WRITEMASK_Y) { + brw_ADD(p, + dst[1], + retype(arg0[1], BRW_REGISTER_TYPE_UW), + negate(suboffset(r1,1))); + + } +} + +static void emit_wpos_xy(struct brw_compile *p, + const struct brw_reg *dst, + unsigned mask, + const struct brw_reg *arg0) +{ + /* Calc delta X,Y by subtracting origin in r1 from the pixel + * centers. + */ + if (mask & TGSI_WRITEMASK_X) { + brw_MOV(p, + dst[0], + retype(arg0[0], BRW_REGISTER_TYPE_UW)); + } + + if (mask & TGSI_WRITEMASK_Y) { + /* TODO -- window_height - Y */ + brw_MOV(p, + dst[1], + negate(retype(arg0[1], BRW_REGISTER_TYPE_UW))); + + } +} + + +static void emit_pixel_w( struct brw_compile *p, + const struct brw_reg *dst, + unsigned mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas) +{ + /* Don't need this if all you are doing is interpolating color, for + * instance. + */ + if (mask & TGSI_WRITEMASK_W) { + struct brw_reg interp3 = brw_vec1_grf(arg0[0].nr+1, 4); + + /* Calc 1/w - just linterp wpos[3] optimized by putting the + * result straight into a message reg. + */ + brw_LINE(p, brw_null_reg(), interp3, deltas[0]); + brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), deltas[1]); + + /* Calc w */ + brw_math_16( p, dst[3], + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 2, brw_null_reg(), + BRW_MATH_PRECISION_FULL); + } +} + + + +static void emit_linterp( struct brw_compile *p, + const struct brw_reg *dst, + unsigned mask, + const struct brw_reg *arg0, + const struct brw_reg *deltas ) +{ + struct brw_reg interp[4]; + unsigned nr = arg0[0].nr; + unsigned i; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + unsigned msgLength, responseLength; + boolean shadow = (c->key.shadowtex_mask & (1<tex_unit)) ? 1 : 0; + unsigned i, nr; + unsigned emit; + + /* How many input regs are there? + */ + switch (inst->tex_idx) { + case TGSI_TEXTURE_1D: + emit = TGSI_WRITEMASK_X; + nr = 1; + break; + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + emit = TGSI_WRITEMASK_XY; + nr = 2; + break; + default: + emit = TGSI_WRITEMASK_XYZ; + nr = 3; + break; + } + + if (shadow) { + nr = 4; + emit |= TGSI_WRITEMASK_W; + } + + msgLength = 1; + + for (i = 0; i < nr; i++) { + static const unsigned swz[4] = {0,1,2,2}; + if (emit & (1<payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), + inst->tex_unit + 1, /* surface */ + inst->tex_unit, /* sampler */ + inst->writemask, + (shadow ? + BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE : + BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE), + responseLength, + msgLength, + 0); + + if (shadow) + brw_MOV(p, dst[3], brw_imm_f(1.0)); +} + + +static void emit_txb( struct brw_wm_compile *c, + const struct brw_wm_instruction *inst, + struct brw_reg *dst, + unsigned dst_flags, + struct brw_reg *arg ) +{ + struct brw_compile *p = &c->func; + unsigned msgLength; + + /* Shadow ignored for txb. + */ + switch (inst->tex_idx) { + case TGSI_TEXTURE_1D: + brw_MOV(p, brw_message_reg(2), arg[0]); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); + brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); + break; + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + brw_MOV(p, brw_message_reg(2), arg[0]); + brw_MOV(p, brw_message_reg(4), arg[1]); + brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); + break; + default: + brw_MOV(p, brw_message_reg(2), arg[0]); + brw_MOV(p, brw_message_reg(4), arg[1]); + brw_MOV(p, brw_message_reg(6), arg[2]); + break; + } + + brw_MOV(p, brw_message_reg(8), arg[3]); + msgLength = 9; + + + brw_SAMPLE(p, + retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW), + 1, + retype(c->payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), + inst->tex_unit + 1, /* surface */ + inst->tex_unit, /* sampler */ + inst->writemask, + BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS, + 8, /* responseLength */ + msgLength, + 0); + +} + + +static void emit_lit( struct brw_compile *p, + const struct brw_reg *dst, + unsigned mask, + const struct brw_reg *arg0 ) +{ + assert((mask & TGSI_WRITEMASK_XW) == 0); + + if (mask & TGSI_WRITEMASK_Y) { + brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); + brw_MOV(p, dst[1], arg0[0]); + brw_set_saturate(p, 0); + } + + if (mask & TGSI_WRITEMASK_Z) { + emit_math2(p, BRW_MATH_FUNCTION_POW, + &dst[2], + TGSI_WRITEMASK_X | (mask & SATURATE), + &arg0[1], + &arg0[3]); + } + + /* Ordinarily you'd use an iff statement to skip or shortcircuit + * some of the POW calculations above, but 16-wide iff statements + * seem to lock c1 hardware, so this is a nasty workaround: + */ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_LE, arg0[0], brw_imm_f(0)); + { + if (mask & TGSI_WRITEMASK_Y) + brw_MOV(p, dst[1], brw_imm_f(0)); + + if (mask & TGSI_WRITEMASK_Z) + brw_MOV(p, dst[2], brw_imm_f(0)); + } + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +} + + +/* Kill pixel - set execution mask to zero for those pixels which + * fail. + */ +static void emit_kil( struct brw_wm_compile *c, + struct brw_reg *arg0) +{ + struct brw_compile *p = &c->func; + struct brw_reg r0uw = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); + unsigned i; + + + /* XXX - usually won't need 4 compares! + */ + for (i = 0; i < 4; i++) { + brw_push_insn_state(p); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_GE, arg0[i], brw_imm_f(0)); + brw_set_predicate_control_flag_value(p, 0xff); + brw_AND(p, r0uw, brw_flag_reg(), r0uw); + brw_pop_insn_state(p); + } +} + +static void fire_fb_write( struct brw_wm_compile *c, + unsigned base_reg, + unsigned nr ) +{ + struct brw_compile *p = &c->func; + + /* Pass through control information: + */ +/* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ + { + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, + brw_message_reg(base_reg + 1), + brw_vec8_grf(1, 0)); + brw_pop_insn_state(p); + } + + /* Send framebuffer write message: */ +/* send (16) null.0<1>:uw m0 r0.0<8;8,1>:uw 0x85a04000:ud { Align1 EOT } */ + brw_fb_WRITE(p, + retype(vec16(brw_null_reg()), BRW_REGISTER_TYPE_UW), + base_reg, + retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), + 0, /* render surface always 0 */ + nr, + 0, + 1); +} + +static void emit_aa( struct brw_wm_compile *c, + struct brw_reg *arg1, + unsigned reg ) +{ + struct brw_compile *p = &c->func; + unsigned comp = c->key.aa_dest_stencil_reg / 2; + unsigned off = c->key.aa_dest_stencil_reg % 2; + struct brw_reg aa = offset(arg1[comp], off); + + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); /* ?? */ + brw_MOV(p, brw_message_reg(reg), aa); + brw_pop_insn_state(p); +} + + +/* Post-fragment-program processing. Send the results to the + * framebuffer. + */ +static void emit_fb_write( struct brw_wm_compile *c, + struct brw_reg *arg0, + struct brw_reg *arg1, + struct brw_reg *arg2) +{ + struct brw_compile *p = &c->func; + unsigned nr = 2; + unsigned channel; + + /* Reserve a space for AA - may not be needed: + */ + if (c->key.aa_dest_stencil_reg) + nr += 1; + + /* I don't really understand how this achieves the color interleave + * (ie RGBARGBA) in the result: [Do the saturation here] + */ + { + brw_push_insn_state(p); + + for (channel = 0; channel < 4; channel++) { + /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ + /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, + brw_message_reg(nr + channel), + arg0[channel]); + + brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); + brw_MOV(p, + brw_message_reg(nr + channel + 4), + sechalf(arg0[channel])); + } + + /* skip over the regs populated above: + */ + nr += 8; + + brw_pop_insn_state(p); + } + + if (c->key.source_depth_to_render_target) + { + if (c->key.computes_depth) + brw_MOV(p, brw_message_reg(nr), arg2[2]); + else + brw_MOV(p, brw_message_reg(nr), arg1[1]); /* ? */ + + nr += 2; + } + + if (c->key.dest_depth_reg) + { + unsigned comp = c->key.dest_depth_reg / 2; + unsigned off = c->key.dest_depth_reg % 2; + + if (off != 0) { + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, brw_message_reg(nr), arg1[comp]); + /* 2nd half? */ + brw_MOV(p, brw_message_reg(nr+1), offset(arg1[comp],1)); + brw_pop_insn_state(p); + } + else { + brw_MOV(p, brw_message_reg(nr), arg1[comp]); + } + nr += 2; + } + + + if (!c->key.runtime_check_aads_emit) { + if (c->key.aa_dest_stencil_reg) + emit_aa(c, arg1, 2); + + fire_fb_write(c, 0, nr); + } + else { + struct brw_reg v1_null_ud = vec1(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD)); + struct brw_reg ip = brw_ip_reg(); + struct brw_instruction *jmp; + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); + brw_AND(p, + v1_null_ud, + get_element_ud(brw_vec8_grf(1,0), 6), + brw_imm_ud(1<<26)); + + jmp = brw_JMPI(p, ip, ip, brw_imm_w(0)); + { + emit_aa(c, arg1, 2); + fire_fb_write(c, 0, nr); + /* note - thread killed in subroutine */ + } + brw_land_fwd_jump(p, jmp); + + /* ELSE: Shuffle up one register to fill in the hole left for AA: + */ + fire_fb_write(c, 1, nr-1); + } +} + + + + +/* Post-fragment-program processing. Send the results to the + * framebuffer. + */ +static void emit_spill( struct brw_wm_compile *c, + struct brw_reg reg, + unsigned slot ) +{ + struct brw_compile *p = &c->func; + + /* + mov (16) m2.0<1>:ud r2.0<8;8,1>:ud { Align1 Compr } + */ + brw_MOV(p, brw_message_reg(2), reg); + + /* + mov (1) r0.2<1>:d 0x00000080:d { Align1 NoMask } + send (16) null.0<1>:uw m1 r0.0<8;8,1>:uw 0x053003ff:ud { Align1 } + */ + brw_dp_WRITE_16(p, + retype(vec16(brw_vec8_grf(0, 0)), BRW_REGISTER_TYPE_UW), + 1, + slot); +} + +static void emit_unspill( struct brw_wm_compile *c, + struct brw_reg reg, + unsigned slot ) +{ + struct brw_compile *p = &c->func; + + /* Slot 0 is the undef value. + */ + if (slot == 0) { + brw_MOV(p, reg, brw_imm_f(0)); + return; + } + + /* + mov (1) r0.2<1>:d 0x000000c0:d { Align1 NoMask } + send (16) r110.0<1>:uw m1 r0.0<8;8,1>:uw 0x041243ff:ud { Align1 } + */ + + brw_dp_READ_16(p, + retype(vec16(reg), BRW_REGISTER_TYPE_UW), + 1, + slot); +} + + + +/** + * Retrieve upto 4 GEN4 register pairs for the given wm reg: + */ +static void get_argument_regs( struct brw_wm_compile *c, + struct brw_wm_ref *arg[], + struct brw_reg *regs ) +{ + unsigned i; + + for (i = 0; i < 4; i++) { + if (arg[i]) { + + if (arg[i]->unspill_reg) + emit_unspill(c, + brw_vec8_grf(arg[i]->unspill_reg, 0), + arg[i]->value->spill_slot); + + regs[i] = arg[i]->hw_reg; + } + else { + regs[i] = brw_null_reg(); + } + } +} + +static void spill_values( struct brw_wm_compile *c, + struct brw_wm_value *values, + unsigned nr ) +{ + unsigned i; + + for (i = 0; i < nr; i++) + if (values[i].spill_slot) + emit_spill(c, values[i].hw_reg, values[i].spill_slot); +} + + + +/* Emit the fragment program instructions here. + */ +void brw_wm_emit( struct brw_wm_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned insn; + + brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); + + /* Check if any of the payload regs need to be spilled: + */ + spill_values(c, c->payload.depth, 4); + spill_values(c, c->creg, c->nr_creg); + spill_values(c, c->payload.input_interp, PIPE_ATTRIB_MAX); + + + for (insn = 0; insn < c->nr_insns; insn++) { + + struct brw_wm_instruction *inst = &c->instruction[insn]; + struct brw_reg args[3][4], dst[4]; + unsigned i, dst_flags; + + /* Get argument regs: + */ + for (i = 0; i < 3; i++) + get_argument_regs(c, inst->src[i], args[i]); + + /* Get dest regs: + */ + for (i = 0; i < 4; i++) + if (inst->dst[i]) + dst[i] = inst->dst[i]->hw_reg; + else + dst[i] = brw_null_reg(); + + /* Flags + */ + dst_flags = inst->writemask; + if (inst->saturate) + dst_flags |= SATURATE; + + switch (inst->opcode) { + /* Generated instructions for calculating triangle interpolants: + */ + case WM_PIXELXY: + emit_pixel_xy(p, dst, dst_flags, args[0]); + break; + + case WM_DELTAXY: + emit_delta_xy(p, dst, dst_flags, args[0], args[1]); + break; + + case WM_WPOSXY: + emit_wpos_xy(p, dst, dst_flags, args[0]); + break; + + case WM_PIXELW: + emit_pixel_w(p, dst, dst_flags, args[0], args[1]); + break; + + case WM_LINTERP: + emit_linterp(p, dst, dst_flags, args[0], args[1]); + break; + + case WM_PINTERP: + emit_pinterp(p, dst, dst_flags, args[0], args[1], args[2]); + break; + + case WM_CINTERP: + emit_cinterp(p, dst, dst_flags, args[0]); + break; + + case WM_FB_WRITE: + emit_fb_write(c, args[0], args[1], args[2]); + break; + + /* Straightforward arithmetic: + */ + case TGSI_OPCODE_ADD: + emit_alu2(p, brw_ADD, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_FRC: + emit_alu1(p, brw_FRC, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_FLR: + emit_alu1(p, brw_RNDD, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_DP3: /* */ + emit_dp3(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_DP4: + emit_dp4(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_DPH: + emit_dph(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_LRP: /* */ + emit_lrp(p, dst, dst_flags, args[0], args[1], args[2]); + break; + + case TGSI_OPCODE_MAD: + emit_mad(p, dst, dst_flags, args[0], args[1], args[2]); + break; + + case TGSI_OPCODE_MOV: + emit_alu1(p, brw_MOV, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_MUL: + emit_alu2(p, brw_MUL, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_XPD: + emit_xpd(p, dst, dst_flags, args[0], args[1]); + break; + + /* Higher math functions: + */ + case TGSI_OPCODE_RCP: + emit_math1(p, BRW_MATH_FUNCTION_INV, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_RSQ: + emit_math1(p, BRW_MATH_FUNCTION_RSQ, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_SIN: + emit_math1(p, BRW_MATH_FUNCTION_SIN, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_COS: + emit_math1(p, BRW_MATH_FUNCTION_COS, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_EX2: + emit_math1(p, BRW_MATH_FUNCTION_EXP, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_LG2: + emit_math1(p, BRW_MATH_FUNCTION_LOG, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_SCS: + /* There is an scs math function, but it would need some + * fixup for 16-element execution. + */ + if (dst_flags & TGSI_WRITEMASK_X) + emit_math1(p, BRW_MATH_FUNCTION_COS, dst, (dst_flags&SATURATE)|TGSI_WRITEMASK_X, + args[0]); + if (dst_flags & TGSI_WRITEMASK_Y) + emit_math1(p, BRW_MATH_FUNCTION_SIN, dst+1, (dst_flags&SATURATE)|TGSI_WRITEMASK_X, + args[0]); + break; + + case TGSI_OPCODE_POW: + emit_math2(p, BRW_MATH_FUNCTION_POW, dst, dst_flags, args[0], args[1]); + break; + + /* Comparisons: + */ + case TGSI_OPCODE_CMP: + emit_cmp(p, dst, dst_flags, args[0], args[1], args[2]); + break; + + case TGSI_OPCODE_MAX: + emit_max(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_MIN: + emit_min(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_SLT: + emit_slt(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_SLE: + emit_sle(p, dst, dst_flags, args[0], args[1]); + break; + case TGSI_OPCODE_SGT: + emit_sgt(p, dst, dst_flags, args[0], args[1]); + break; + case TGSI_OPCODE_SGE: + emit_sge(p, dst, dst_flags, args[0], args[1]); + break; + case TGSI_OPCODE_SEQ: + emit_seq(p, dst, dst_flags, args[0], args[1]); + break; + case TGSI_OPCODE_SNE: + emit_sne(p, dst, dst_flags, args[0], args[1]); + break; + + case TGSI_OPCODE_LIT: + emit_lit(p, dst, dst_flags, args[0]); + break; + + /* Texturing operations: + */ + case TGSI_OPCODE_TEX: + emit_tex(c, inst, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_TXB: + emit_txb(c, inst, dst, dst_flags, args[0]); + break; + + case TGSI_OPCODE_KIL: + emit_kil(c, args[0]); + break; + + default: + printf("unsupport opcode %d in fragment program\n", + inst->opcode); + } + + for (i = 0; i < 4; i++) + if (inst->dst[i] && inst->dst[i]->spill_slot) + emit_spill(c, + inst->dst[i]->hw_reg, + inst->dst[i]->spill_slot); + } +} diff --git a/src/mesa/pipe/i965simple/brw_wm_fp.c b/src/mesa/pipe/i965simple/brw_wm_fp.c new file mode 100644 index 0000000000..20e90bc612 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_fp.c @@ -0,0 +1,1007 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_wm.h" +#include "brw_util.h" + + +#define FIRST_INTERNAL_TEMP MAX_NV_FRAGMENT_PROGRAM_TEMPS + +#define X 0 +#define Y 1 +#define Z 2 +#define W 3 + + +static const char *wm_opcode_strings[] = { + "PIXELXY", + "DELTAXY", + "PIXELW", + "LINTERP", + "PINTERP", + "CINTERP", + "WPOSXY", + "FB_WRITE" +}; + +#if 0 +static const char *wm_file_strings[] = { + "PAYLOAD" +}; +#endif + + +/*********************************************************************** + * Source regs + */ +#if 0 +static struct prog_src_register src_reg(unsigned file, unsigned idx) +{ + struct prog_src_register reg; + reg.File = file; + reg.Index = idx; + reg.Swizzle = SWIZZLE_NOOP; + reg.RelAddr = 0; + reg.NegateBase = 0; + reg.Abs = 0; + reg.NegateAbs = 0; + return reg; +} + +static struct prog_src_register src_reg_from_dst(struct prog_dst_register dst) +{ + return src_reg(dst.File, dst.Index); +} + +static struct prog_src_register src_undef( void ) +{ + return src_reg(PROGRAM_UNDEFINED, 0); +} + +static boolean src_is_undef(struct prog_src_register src) +{ + return src.File == PROGRAM_UNDEFINED; +} + +static struct prog_src_register src_swizzle( struct prog_src_register reg, int x, int y, int z, int w ) +{ + reg.Swizzle = MAKE_SWIZZLE4(x,y,z,w); + return reg; +} + +static struct prog_src_register src_swizzle1( struct prog_src_register reg, int x ) +{ + return src_swizzle(reg, x, x, x, x); +} + + +/*********************************************************************** + * Dest regs + */ + +static struct prog_dst_register dst_reg(unsigned file, unsigned idx) +{ + struct prog_dst_register reg; + reg.File = file; + reg.Index = idx; + reg.WriteMask = WRITEMASK_XYZW; + reg.CondMask = 0; + reg.CondSwizzle = 0; + reg.pad = 0; + reg.CondSrc = 0; + return reg; +} + +static struct prog_dst_register dst_mask( struct prog_dst_register reg, int mask ) +{ + reg.WriteMask &= mask; + return reg; +} + +static struct prog_dst_register dst_undef( void ) +{ + return dst_reg(PROGRAM_UNDEFINED, 0); +} + + + +static struct prog_dst_register get_temp( struct brw_wm_compile *c ) +{ + int bit = ffs( ~c->fp_temp ); + + if (!bit) { + _mesa_printf("%s: out of temporaries\n", __FILE__); + exit(1); + } + + c->fp_temp |= 1<<(bit-1); + return dst_reg(PROGRAM_TEMPORARY, FIRST_INTERNAL_TEMP+(bit-1)); +} + + +static void release_temp( struct brw_wm_compile *c, struct prog_dst_register temp ) +{ + c->fp_temp &= ~1<<(temp.Index + 1 - FIRST_INTERNAL_TEMP); +} + + +/*********************************************************************** + * Instructions + */ + +static const struct tgsi_token *get_fp_inst(struct brw_wm_compile *c) +{ + return &c->prog_instructions->tokens[c->nr_fp_insns++]; +} + +static struct prog_instruction *emit_insn(struct brw_wm_compile *c, + const struct prog_instruction *inst0) +{ + struct prog_instruction *inst = get_fp_inst(c); + *inst = *inst0; + inst->Data = (void *)inst0; + return inst; +} + +static struct prog_instruction * emit_op(struct brw_wm_compile *c, + unsigned op, + struct prog_dst_register dest, + unsigned saturate, + unsigned tex_src_unit, + unsigned tex_src_target, + struct prog_src_register src0, + struct prog_src_register src1, + struct prog_src_register src2 ) +{ + struct prog_instruction *inst = get_fp_inst(c); + + memset(inst, 0, sizeof(*inst)); + + inst->Opcode = op; + inst->DstReg = dest; + inst->SaturateMode = saturate; + inst->TexSrcUnit = tex_src_unit; + inst->TexSrcTarget = tex_src_target; + inst->SrcReg[0] = src0; + inst->SrcReg[1] = src1; + inst->SrcReg[2] = src2; + return inst; +} + + + + +/*********************************************************************** + * Special instructions for interpolation and other tasks + */ + +static struct prog_src_register get_pixel_xy( struct brw_wm_compile *c ) +{ + if (src_is_undef(c->pixel_xy)) { + struct prog_dst_register pixel_xy = get_temp(c); + struct prog_src_register payload_r0_depth = src_reg(PROGRAM_PAYLOAD, PAYLOAD_DEPTH); + + + /* Emit the out calculations, and hold onto the results. Use + * two instructions as a temporary is required. + */ + /* pixel_xy.xy = PIXELXY payload[0]; + */ + emit_op(c, + WM_PIXELXY, + dst_mask(pixel_xy, WRITEMASK_XY), + 0, 0, 0, + payload_r0_depth, + src_undef(), + src_undef()); + + c->pixel_xy = src_reg_from_dst(pixel_xy); + } + + return c->pixel_xy; +} + +static struct prog_src_register get_delta_xy( struct brw_wm_compile *c ) +{ + if (src_is_undef(c->delta_xy)) { + struct prog_dst_register delta_xy = get_temp(c); + struct prog_src_register pixel_xy = get_pixel_xy(c); + struct prog_src_register payload_r0_depth = src_reg(PROGRAM_PAYLOAD, PAYLOAD_DEPTH); + + /* deltas.xy = DELTAXY pixel_xy, payload[0] + */ + emit_op(c, + WM_DELTAXY, + dst_mask(delta_xy, WRITEMASK_XY), + 0, 0, 0, + pixel_xy, + payload_r0_depth, + src_undef()); + + c->delta_xy = src_reg_from_dst(delta_xy); + } + + return c->delta_xy; +} + +static struct prog_src_register get_pixel_w( struct brw_wm_compile *c ) +{ + if (src_is_undef(c->pixel_w)) { + struct prog_dst_register pixel_w = get_temp(c); + struct prog_src_register deltas = get_delta_xy(c); + struct prog_src_register interp_wpos = src_reg(PROGRAM_PAYLOAD, FRAG_ATTRIB_WPOS); + + + /* deltas.xyw = DELTAS2 deltas.xy, payload.interp_wpos.x + */ + emit_op(c, + WM_PIXELW, + dst_mask(pixel_w, WRITEMASK_W), + 0, 0, 0, + interp_wpos, + deltas, + src_undef()); + + + c->pixel_w = src_reg_from_dst(pixel_w); + } + + return c->pixel_w; +} + +static void emit_interp( struct brw_wm_compile *c, + unsigned idx ) +{ + struct prog_dst_register dst = dst_reg(PROGRAM_INPUT, idx); + struct prog_src_register interp = src_reg(PROGRAM_PAYLOAD, idx); + struct prog_src_register deltas = get_delta_xy(c); + struct prog_src_register arg2; + unsigned opcode; + + /* Need to use PINTERP on attributes which have been + * multiplied by 1/W in the SF program, and LINTERP on those + * which have not: + */ + switch (idx) { + case FRAG_ATTRIB_WPOS: + opcode = WM_LINTERP; + arg2 = src_undef(); + + /* Have to treat wpos.xy specially: + */ + emit_op(c, + WM_WPOSXY, + dst_mask(dst, WRITEMASK_XY), + 0, 0, 0, + get_pixel_xy(c), + src_undef(), + src_undef()); + + dst = dst_mask(dst, WRITEMASK_ZW); + + /* PROGRAM_INPUT.attr.xyzw = INTERP payload.interp[attr].x, deltas.xyw + */ + emit_op(c, + WM_LINTERP, + dst, + 0, 0, 0, + interp, + deltas, + arg2); + break; + case FRAG_ATTRIB_COL0: + case FRAG_ATTRIB_COL1: + if (c->key.flat_shade) { + emit_op(c, + WM_CINTERP, + dst, + 0, 0, 0, + interp, + src_undef(), + src_undef()); + } + else { + emit_op(c, + WM_LINTERP, + dst, + 0, 0, 0, + interp, + deltas, + src_undef()); + } + break; + default: + emit_op(c, + WM_PINTERP, + dst, + 0, 0, 0, + interp, + deltas, + get_pixel_w(c)); + break; + } + + c->fp_interp_emitted |= 1<SrcReg[0].Index; + struct prog_src_register interp = src_reg(PROGRAM_PAYLOAD, idx); + + c->fp_deriv_emitted |= 1<DstReg, + 0, 0, 0, + interp, + get_pixel_w(c), + src_undef()); +} + +static void emit_ddy( struct brw_wm_compile *c, + const struct prog_instruction *inst ) +{ + unsigned idx = inst->SrcReg[0].Index; + struct prog_src_register interp = src_reg(PROGRAM_PAYLOAD, idx); + + c->fp_deriv_emitted |= 1<DstReg, + 0, 0, 0, + interp, + get_pixel_w(c), + src_undef()); +} + +/*********************************************************************** + * Hacks to extend the program parameter and constant lists. + */ + +/* Add the fog parameters to the parameter list of the original + * program, rather than creating a new list. Doesn't really do any + * harm and it's not as if the parameter handling isn't a big hack + * anyway. + */ +static struct prog_src_register search_or_add_param5(struct brw_wm_compile *c, + int s0, + int s1, + int s2, + int s3, + int s4) +{ + struct gl_program_parameter_list *paramList = c->fp->program.Base.Parameters; + gl_state_index tokens[STATE_LENGTH]; + unsigned idx; + tokens[0] = s0; + tokens[1] = s1; + tokens[2] = s2; + tokens[3] = s3; + tokens[4] = s4; + + for (idx = 0; idx < paramList->NumParameters; idx++) { + if (paramList->Parameters[idx].Type == PROGRAM_STATE_VAR && + memcmp(paramList->Parameters[idx].StateIndexes, tokens, sizeof(tokens)) == 0) + return src_reg(PROGRAM_STATE_VAR, idx); + } + + idx = _mesa_add_state_reference( paramList, tokens ); + + /* Recalculate state dependency: + */ + c->fp->param_state = paramList->StateFlags; + + return src_reg(PROGRAM_STATE_VAR, idx); +} + + +static struct prog_src_register search_or_add_const4f( struct brw_wm_compile *c, + float s0, + float s1, + float s2, + float s3) +{ + struct gl_program_parameter_list *paramList = c->fp->program.Base.Parameters; + float values[4]; + unsigned idx; + unsigned swizzle; + + values[0] = s0; + values[1] = s1; + values[2] = s2; + values[3] = s3; + + /* Have to search, otherwise multiple compilations will each grow + * the parameter list. + */ + for (idx = 0; idx < paramList->NumParameters; idx++) { + if (paramList->Parameters[idx].Type == PROGRAM_CONSTANT && + memcmp(paramList->ParameterValues[idx], values, sizeof(values)) == 0) + + /* XXX: this mimics the mesa bug which puts all constants and + * parameters into the "PROGRAM_STATE_VAR" category: + */ + return src_reg(PROGRAM_STATE_VAR, idx); + } + + idx = _mesa_add_unnamed_constant( paramList, values, 4, &swizzle ); + /* XXX what about swizzle? */ + return src_reg(PROGRAM_STATE_VAR, idx); +} + + + +/*********************************************************************** + * Expand various instructions here to simpler forms. + */ +static void precalc_dst( struct brw_wm_compile *c, + const struct prog_instruction *inst ) +{ + struct prog_src_register src0 = inst->SrcReg[0]; + struct prog_src_register src1 = inst->SrcReg[1]; + struct prog_dst_register dst = inst->DstReg; + + if (dst.WriteMask & WRITEMASK_Y) { + /* dst.y = mul src0.y, src1.y + */ + emit_op(c, + OPCODE_MUL, + dst_mask(dst, WRITEMASK_Y), + inst->SaturateMode, 0, 0, + src0, + src1, + src_undef()); + } + + + if (dst.WriteMask & WRITEMASK_XZ) { + unsigned z = GET_SWZ(src0.Swizzle, Z); + + /* dst.xz = swz src0.1zzz + */ + emit_op(c, + OPCODE_SWZ, + dst_mask(dst, WRITEMASK_XZ), + inst->SaturateMode, 0, 0, + src_swizzle(src0, SWIZZLE_ONE, z, z, z), + src_undef(), + src_undef()); + } + if (dst.WriteMask & WRITEMASK_W) { + /* dst.w = mov src1.w + */ + emit_op(c, + OPCODE_MOV, + dst_mask(dst, WRITEMASK_W), + inst->SaturateMode, 0, 0, + src1, + src_undef(), + src_undef()); + } +} + + +static void precalc_lit( struct brw_wm_compile *c, + const struct prog_instruction *inst ) +{ + struct prog_src_register src0 = inst->SrcReg[0]; + struct prog_dst_register dst = inst->DstReg; + + if (dst.WriteMask & WRITEMASK_XW) { + /* dst.xw = swz src0.1111 + */ + emit_op(c, + OPCODE_SWZ, + dst_mask(dst, WRITEMASK_XW), + 0, 0, 0, + src_swizzle1(src0, SWIZZLE_ONE), + src_undef(), + src_undef()); + } + + + if (dst.WriteMask & WRITEMASK_YZ) { + emit_op(c, + OPCODE_LIT, + dst_mask(dst, WRITEMASK_YZ), + inst->SaturateMode, 0, 0, + src0, + src_undef(), + src_undef()); + } +} + +static void precalc_tex( struct brw_wm_compile *c, + const struct prog_instruction *inst ) +{ + struct prog_src_register coord; + struct prog_dst_register tmpcoord; + + if (inst->TexSrcTarget == TEXTURE_CUBE_INDEX) { + struct prog_instruction *out; + struct prog_dst_register tmp0 = get_temp(c); + struct prog_src_register tmp0src = src_reg_from_dst(tmp0); + struct prog_dst_register tmp1 = get_temp(c); + struct prog_src_register tmp1src = src_reg_from_dst(tmp1); + struct prog_src_register src0 = inst->SrcReg[0]; + + tmpcoord = get_temp(c); + coord = src_reg_from_dst(tmpcoord); + + out = emit_op(c, OPCODE_MOV, + tmpcoord, + 0, 0, 0, + src0, + src_undef(), + src_undef()); + out->SrcReg[0].NegateBase = 0; + out->SrcReg[0].Abs = 1; + + emit_op(c, OPCODE_MAX, + tmp0, + 0, 0, 0, + src_swizzle1(coord, X), + src_swizzle1(coord, Y), + src_undef()); + + emit_op(c, OPCODE_MAX, + tmp1, + 0, 0, 0, + tmp0src, + src_swizzle1(coord, Z), + src_undef()); + + emit_op(c, OPCODE_RCP, + tmp0, + 0, 0, 0, + tmp1src, + src_undef(), + src_undef()); + + emit_op(c, OPCODE_MUL, + tmpcoord, + 0, 0, 0, + src0, + tmp0src, + src_undef()); + + release_temp(c, tmp0); + release_temp(c, tmp1); + } else if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) { + struct prog_src_register scale = + search_or_add_param5( c, + STATE_INTERNAL, + STATE_TEXRECT_SCALE, + inst->TexSrcUnit, + 0,0 ); + + tmpcoord = get_temp(c); + + /* coord.xy = MUL inst->SrcReg[0], { 1/width, 1/height } + */ + emit_op(c, + OPCODE_MUL, + tmpcoord, + 0, 0, 0, + inst->SrcReg[0], + scale, + src_undef()); + + coord = src_reg_from_dst(tmpcoord); + } + else { + coord = inst->SrcReg[0]; + } + + /* Need to emit YUV texture conversions by hand. Probably need to + * do this here - the alternative is in brw_wm_emit.c, but the + * conversion requires allocating a temporary variable which we + * don't have the facility to do that late in the compilation. + */ + if (!(c->key.yuvtex_mask & (1<TexSrcUnit))) { + emit_op(c, + OPCODE_TEX, + inst->DstReg, + inst->SaturateMode, + inst->TexSrcUnit, + inst->TexSrcTarget, + coord, + src_undef(), + src_undef()); + } + else { + /* + CONST C0 = { -.5, -.0625, -.5, 1.164 } + CONST C1 = { 1.596, -0.813, 2.018, -.391 } + UYV = TEX ... + UYV.xyz = ADD UYV, C0 + UYV.y = MUL UYV.y, C0.w + RGB.xyz = MAD UYV.xxz, C1, UYV.y + RGB.y = MAD UYV.z, C1.w, RGB.y + */ + struct prog_dst_register dst = inst->DstReg; + struct prog_src_register src0 = inst->SrcReg[0]; + struct prog_dst_register tmp = get_temp(c); + struct prog_src_register tmpsrc = src_reg_from_dst(tmp); + struct prog_src_register C0 = search_or_add_const4f( c, -.5, -.0625, -.5, 1.164 ); + struct prog_src_register C1 = search_or_add_const4f( c, 1.596, -0.813, 2.018, -.391 ); + + /* tmp = TEX ... + */ + emit_op(c, + OPCODE_TEX, + tmp, + inst->SaturateMode, + inst->TexSrcUnit, + inst->TexSrcTarget, + src0, + src_undef(), + src_undef()); + + /* tmp.xyz = ADD TMP, C0 + */ + emit_op(c, + OPCODE_ADD, + dst_mask(tmp, WRITEMASK_XYZ), + 0, 0, 0, + tmpsrc, + C0, + src_undef()); + + /* YUV.y = MUL YUV.y, C0.w + */ + emit_op(c, + OPCODE_MUL, + dst_mask(tmp, WRITEMASK_Y), + 0, 0, 0, + tmpsrc, + src_swizzle1(C0, W), + src_undef()); + + /* RGB.xyz = MAD YUV.xxz, C1, YUV.y + */ + emit_op(c, + OPCODE_MAD, + dst_mask(dst, WRITEMASK_XYZ), + 0, 0, 0, + src_swizzle(tmpsrc, X,X,Z,Z), + C1, + src_swizzle1(tmpsrc, Y)); + + /* RGB.y = MAD YUV.z, C1.w, RGB.y + */ + emit_op(c, + OPCODE_MAD, + dst_mask(dst, WRITEMASK_Y), + 0, 0, 0, + src_swizzle1(tmpsrc, Z), + src_swizzle1(C1, W), + src_swizzle1(src_reg_from_dst(dst), Y)); + + release_temp(c, tmp); + } + + if (inst->TexSrcTarget == GL_TEXTURE_RECTANGLE_NV) + release_temp(c, tmpcoord); +} + + +static boolean projtex( struct brw_wm_compile *c, + const struct prog_instruction *inst ) +{ + struct prog_src_register src = inst->SrcReg[0]; + + /* Only try to detect the simplest cases. Could detect (later) + * cases where we are trying to emit code like RCP {1.0}, MUL x, + * {1.0}, and so on. + * + * More complex cases than this typically only arise from + * user-provided fragment programs anyway: + */ + if (inst->TexSrcTarget == TEXTURE_CUBE_INDEX) + return 0; /* ut2004 gun rendering !?! */ + else if (src.File == PROGRAM_INPUT && + GET_SWZ(src.Swizzle, W) == W && + (c->key.projtex_mask & (1<<(src.Index + FRAG_ATTRIB_WPOS - FRAG_ATTRIB_TEX0))) == 0) + return 0; + else + return 1; +} + + +static void precalc_txp( struct brw_wm_compile *c, + const struct prog_instruction *inst ) +{ + struct prog_src_register src0 = inst->SrcReg[0]; + + if (projtex(c, inst)) { + struct prog_dst_register tmp = get_temp(c); + struct prog_instruction tmp_inst; + + /* tmp0.w = RCP inst.arg[0][3] + */ + emit_op(c, + OPCODE_RCP, + dst_mask(tmp, WRITEMASK_W), + 0, 0, 0, + src_swizzle1(src0, GET_SWZ(src0.Swizzle, W)), + src_undef(), + src_undef()); + + /* tmp0.xyz = MUL inst.arg[0], tmp0.wwww + */ + emit_op(c, + OPCODE_MUL, + dst_mask(tmp, WRITEMASK_XYZ), + 0, 0, 0, + src0, + src_swizzle1(src_reg_from_dst(tmp), W), + src_undef()); + + /* dst = precalc(TEX tmp0) + */ + tmp_inst = *inst; + tmp_inst.SrcReg[0] = src_reg_from_dst(tmp); + precalc_tex(c, &tmp_inst); + + release_temp(c, tmp); + } + else + { + /* dst = precalc(TEX src0) + */ + precalc_tex(c, inst); + } +} + + + + + +/*********************************************************************** + * Add instructions to perform fog blending + */ + +static void fog_blend( struct brw_wm_compile *c, + struct prog_src_register fog_factor ) +{ + struct prog_dst_register outcolor = dst_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLR); + struct prog_src_register fogcolor = search_or_add_param5( c, STATE_FOG_COLOR, 0,0,0,0 ); + + /* color.xyz = LRP fog_factor.xxxx, output_color, fog_color */ + + emit_op(c, + OPCODE_LRP, + dst_mask(outcolor, WRITEMASK_XYZ), + 0, 0, 0, + fog_factor, + src_reg_from_dst(outcolor), + fogcolor); +} + + + +/* This one is simple - just take the interpolated fog coordinate and + * use it as the fog blend factor. + */ +static void fog_interpolated( struct brw_wm_compile *c ) +{ + struct prog_src_register fogc = src_reg(PROGRAM_INPUT, FRAG_ATTRIB_FOGC); + + if (!(c->fp_interp_emitted & (1<fp->program.FogOption) + return; + + if (1) + fog_interpolated( c ); + else { + /* TODO: per-pixel fog */ + assert(0); + } +} + +static void emit_fb_write( struct brw_wm_compile *c ) +{ + struct prog_src_register outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLR); + struct prog_src_register payload_r0_depth = src_reg(PROGRAM_PAYLOAD, PAYLOAD_DEPTH); + struct prog_src_register outdepth = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_DEPR); + + emit_op(c, + WM_FB_WRITE, + dst_mask(dst_undef(),0), + 0, 0, 0, + outcolor, + payload_r0_depth, + outdepth); +} + + + + +/*********************************************************************** + * Emit INTERP instructions ahead of first use of each attrib. + */ + +static void validate_src_regs( struct brw_wm_compile *c, + const struct prog_instruction *inst ) +{ + unsigned nr_args = brw_wm_nr_args( inst->Opcode ); + unsigned i; + + for (i = 0; i < nr_args; i++) { + if (inst->SrcReg[i].File == PROGRAM_INPUT) { + unsigned idx = inst->SrcReg[i].Index; + if (!(c->fp_interp_emitted & (1<Opcode < MAX_OPCODE) + _mesa_print_instruction(insn); + else if (insn->Opcode < MAX_WM_OPCODE) { + unsigned idx = insn->Opcode - MAX_OPCODE; + + _mesa_print_alu_instruction(insn, + wm_opcode_strings[idx], + 3); + } + else + _mesa_printf("UNKNOWN\n"); + + } +} +void brw_wm_pass_fp( struct brw_wm_compile *c ) +{ + struct brw_fragment_program *fp = c->fp; + unsigned insn; + if (BRW_DEBUG & DEBUG_WM) { + _mesa_printf("\n\n\npre-fp:\n"); + _mesa_print_program(&fp->program.Base); + _mesa_printf("\n"); + } + + c->pixel_xy = src_undef(); + c->delta_xy = src_undef(); + c->pixel_w = src_undef(); + c->nr_fp_insns = 0; + + /* Emit preamble instructions: + */ + + + for (insn = 0; insn < fp->program.Base.NumInstructions; insn++) { + const struct prog_instruction *inst = &fp->program.Base.Instructions[insn]; + struct prog_instruction *out; + + /* Check for INPUT values, emit INTERP instructions where + * necessary: + */ + validate_src_regs(c, inst); + + + switch (inst->Opcode) { + case OPCODE_SWZ: + out = emit_insn(c, inst); + out->Opcode = OPCODE_MOV; + break; + + case OPCODE_ABS: + out = emit_insn(c, inst); + out->Opcode = OPCODE_MOV; + out->SrcReg[0].NegateBase = 0; + out->SrcReg[0].Abs = 1; + break; + + case OPCODE_SUB: + out = emit_insn(c, inst); + out->Opcode = OPCODE_ADD; + out->SrcReg[1].NegateBase ^= 0xf; + break; + + case OPCODE_SCS: + out = emit_insn(c, inst); + /* This should probably be done in the parser. + */ + out->DstReg.WriteMask &= WRITEMASK_XY; + break; + + case OPCODE_DST: + precalc_dst(c, inst); + break; + + case OPCODE_LIT: + precalc_lit(c, inst); + break; + + case OPCODE_TXP: + precalc_txp(c, inst); + break; + + case OPCODE_XPD: + out = emit_insn(c, inst); + /* This should probably be done in the parser. + */ + out->DstReg.WriteMask &= WRITEMASK_XYZ; + break; + + case OPCODE_KIL: + out = emit_insn(c, inst); + /* This should probably be done in the parser. + */ + out->DstReg.WriteMask = 0; + break; + case OPCODE_DDX: + emit_ddx(c, inst); + break; + case OPCODE_DDY: + emit_ddy(c, inst); + break; + case OPCODE_END: + emit_fog(c); + emit_fb_write(c); + break; + case OPCODE_PRINT: + break; + + default: + emit_insn(c, inst); + break; + } + } + + if (BRW_DEBUG & DEBUG_WM) { + _mesa_printf("\n\n\npass_fp:\n"); + print_insns( c->prog_instructions, c->nr_fp_insns ); + _mesa_printf("\n"); + } +} +#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_glsl.c b/src/mesa/pipe/i965simple/brw_wm_glsl.c new file mode 100644 index 0000000000..90e73a605a --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_glsl.c @@ -0,0 +1,1356 @@ + +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_wm.h" + +#if 0 + +/* Only guess, need a flag in gl_fragment_program later */ +boolean brw_wm_is_glsl(struct gl_fragment_program *fp) +{ + int i; + for (i = 0; i < fp->Base.NumInstructions; i++) { + struct prog_instruction *inst = &fp->Base.Instructions[i]; + switch (inst->Opcode) { + case OPCODE_IF: + case OPCODE_INT: + case OPCODE_ENDIF: + case OPCODE_CAL: + case OPCODE_BRK: + case OPCODE_RET: + case OPCODE_DDX: + case OPCODE_DDY: + case OPCODE_BGNLOOP: + return TRUE; + default: + break; + } + } + return FALSE; +} + +static void set_reg(struct brw_wm_compile *c, int file, int index, + int component, struct brw_reg reg) +{ + c->wm_regs[file][index][component].reg = reg; + c->wm_regs[file][index][component].inited = TRUE; +} + +static int get_scalar_dst_index(struct prog_instruction *inst) +{ + int i; + for (i = 0; i < 4; i++) + if (inst->DstReg.WriteMask & (1<tmp_index--, 0); + return reg; +} + +static void release_tmps(struct brw_wm_compile *c) +{ + c->tmp_index = 127; +} + +static struct brw_reg +get_reg(struct brw_wm_compile *c, int file, int index, int component, int nr, unsigned neg, unsigned abs) +{ + struct brw_reg reg; + switch (file) { + case PROGRAM_STATE_VAR: + case PROGRAM_CONSTANT: + case PROGRAM_UNIFORM: + file = PROGRAM_STATE_VAR; + break; + case PROGRAM_UNDEFINED: + return brw_null_reg(); + default: + break; + } + + if(c->wm_regs[file][index][component].inited) + reg = c->wm_regs[file][index][component].reg; + else + reg = brw_vec8_grf(c->reg_index, 0); + + if(!c->wm_regs[file][index][component].inited) { + set_reg(c, file, index, component, reg); + c->reg_index++; + } + + if (neg & (1<< component)) { + reg = negate(reg); + } + if (abs) + reg = brw_abs(reg); + return reg; +} + +static void prealloc_reg(struct brw_wm_compile *c) +{ + int i, j; + struct brw_reg reg; + int nr_interp_regs = 0; + unsigned inputs = FRAG_BIT_WPOS | c->fp_interp_emitted | c->fp_deriv_emitted; + + for (i = 0; i < 4; i++) { + reg = (i < c->key.nr_depth_regs) + ? brw_vec8_grf(i*2, 0) : brw_vec8_grf(0, 0); + set_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, i, reg); + } + c->reg_index += 2*c->key.nr_depth_regs; + { + int nr_params = c->fp->program.Base.Parameters->NumParameters; + struct gl_program_parameter_list *plist = + c->fp->program.Base.Parameters; + int index = 0; + c->prog_data.nr_params = 4*nr_params; + for (i = 0; i < nr_params; i++) { + for (j = 0; j < 4; j++, index++) { + reg = brw_vec1_grf(c->reg_index + index/8, + index%8); + c->prog_data.param[index] = + &plist->ParameterValues[i][j]; + set_reg(c, PROGRAM_STATE_VAR, i, j, reg); + } + } + c->nr_creg = 2*((4*nr_params+15)/16); + c->reg_index += c->nr_creg; + } + for (i = 0; i < FRAG_ATTRIB_MAX; i++) { + if (inputs & (1<reg_index, 0); + for (j = 0; j < 4; j++) + set_reg(c, PROGRAM_PAYLOAD, i, j, reg); + c->reg_index += 2; + + } + } + c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; + c->prog_data.urb_read_length = nr_interp_regs * 2; + c->prog_data.curb_read_length = c->nr_creg; + c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); + c->reg_index++; + c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); + c->reg_index += 2; +} + +static struct brw_reg get_dst_reg(struct brw_wm_compile *c, + struct prog_instruction *inst, int component, int nr) +{ + return get_reg(c, inst->DstReg.File, inst->DstReg.Index, component, nr, + 0, 0); +} + +static struct brw_reg get_src_reg(struct brw_wm_compile *c, + struct prog_src_register *src, int index, int nr) +{ + int component = GET_SWZ(src->Swizzle, index); + return get_reg(c, src->File, src->Index, component, nr, + src->NegateBase, src->Abs); +} + +static void emit_abs( struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + int i; + struct brw_compile *p = &c->func; + brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); + for (i = 0; i < 4; i++) { + if (inst->DstReg.WriteMask & (1<SrcReg[0], i, 1); + brw_MOV(p, dst, brw_abs(src)); + } + } + brw_set_saturate(p, 0); +} + +static void emit_int( struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + int i; + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); + for (i = 0; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + brw_RNDD(p, dst, src); + } + } + brw_set_saturate(p, 0); +} + +static void emit_mov( struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + int i; + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); + for (i = 0; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + brw_MOV(p, dst, src); + } + } + brw_set_saturate(p, 0); +} + +static void emit_pixel_xy(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_reg r1 = brw_vec1_grf(1, 0); + struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW); + + struct brw_reg dst0, dst1; + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + + dst0 = get_dst_reg(c, inst, 0, 1); + dst1 = get_dst_reg(c, inst, 1, 1); + /* Calculate pixel centers by adding 1 or 0 to each of the + * micro-tile coordinates passed in r1. + */ + if (mask & WRITEMASK_X) { + brw_ADD(p, + vec8(retype(dst0, BRW_REGISTER_TYPE_UW)), + stride(suboffset(r1_uw, 4), 2, 4, 0), + brw_imm_v(0x10101010)); + } + + if (mask & WRITEMASK_Y) { + brw_ADD(p, + vec8(retype(dst1, BRW_REGISTER_TYPE_UW)), + stride(suboffset(r1_uw, 5), 2, 4, 0), + brw_imm_v(0x11001100)); + } + +} + +static void emit_delta_xy(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_reg r1 = brw_vec1_grf(1, 0); + struct brw_reg dst0, dst1, src0, src1; + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + + dst0 = get_dst_reg(c, inst, 0, 1); + dst1 = get_dst_reg(c, inst, 1, 1); + src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); + src1 = get_src_reg(c, &inst->SrcReg[0], 1, 1); + /* Calc delta X,Y by subtracting origin in r1 from the pixel + * centers. + */ + if (mask & WRITEMASK_X) { + brw_ADD(p, + dst0, + retype(src0, BRW_REGISTER_TYPE_UW), + negate(r1)); + } + + if (mask & WRITEMASK_Y) { + brw_ADD(p, + dst1, + retype(src1, BRW_REGISTER_TYPE_UW), + negate(suboffset(r1,1))); + + } + +} + + +static void fire_fb_write( struct brw_wm_compile *c, + unsigned base_reg, + unsigned nr ) +{ + struct brw_compile *p = &c->func; + + /* Pass through control information: + */ + /* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ + { + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ + brw_MOV(p, + brw_message_reg(base_reg + 1), + brw_vec8_grf(1, 0)); + brw_pop_insn_state(p); + } + /* Send framebuffer write message: */ + brw_fb_WRITE(p, + retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW), + base_reg, + retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), + 0, /* render surface always 0 */ + nr, + 0, + 1); +} + +static void emit_fb_write(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + int nr = 2; + int channel; + struct brw_reg src0;//, src1, src2, dst; + + /* Reserve a space for AA - may not be needed: + */ + if (c->key.aa_dest_stencil_reg) + nr += 1; + { + brw_push_insn_state(p); + for (channel = 0; channel < 4; channel++) { + src0 = get_src_reg(c, &inst->SrcReg[0], channel, 1); + /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ + /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ + brw_MOV(p, brw_message_reg(nr + channel), src0); + } + /* skip over the regs populated above: */ + nr += 8; + brw_pop_insn_state(p); + } + fire_fb_write(c, 0, nr); +} + +static void emit_pixel_w( struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + if (mask & WRITEMASK_W) { + struct brw_reg dst, src0, delta0, delta1; + struct brw_reg interp3; + + dst = get_dst_reg(c, inst, 3, 1); + src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); + delta0 = get_src_reg(c, &inst->SrcReg[1], 0, 1); + delta1 = get_src_reg(c, &inst->SrcReg[1], 1, 1); + + interp3 = brw_vec1_grf(src0.nr+1, 4); + /* Calc 1/w - just linterp wpos[3] optimized by putting the + * result straight into a message reg. + */ + brw_LINE(p, brw_null_reg(), interp3, delta0); + brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), delta1); + + /* Calc w */ + brw_math_16( p, dst, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 2, brw_null_reg(), + BRW_MATH_PRECISION_FULL); + } +} + +static void emit_linterp(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + struct brw_reg interp[4]; + struct brw_reg dst, delta0, delta1; + struct brw_reg src0; + + src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); + delta0 = get_src_reg(c, &inst->SrcReg[1], 0, 1); + delta1 = get_src_reg(c, &inst->SrcReg[1], 1, 1); + unsigned nr = src0.nr; + int i; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + unsigned mask = inst->DstReg.WriteMask; + + struct brw_reg interp[4]; + struct brw_reg dst, src0; + + src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); + unsigned nr = src0.nr; + int i; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + unsigned mask = inst->DstReg.WriteMask; + + struct brw_reg interp[4]; + struct brw_reg dst, delta0, delta1; + struct brw_reg src0, w; + + src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); + delta0 = get_src_reg(c, &inst->SrcReg[1], 0, 1); + delta1 = get_src_reg(c, &inst->SrcReg[1], 1, 1); + w = get_src_reg(c, &inst->SrcReg[2], 3, 1); + unsigned nr = src0.nr; + int i; + + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + unsigned mask = inst->DstReg.WriteMask; + for (i = 0; i < 4; i++) { + unsigned i2 = (i+2)%3; + unsigned i1 = (i+1)%3; + if (mask & (1<SrcReg[0], i2, 1)); + src1 = get_src_reg(c, &inst->SrcReg[1], i1, 1); + brw_MUL(p, brw_null_reg(), src0, src1); + src0 = get_src_reg(c, &inst->SrcReg[0], i1, 1); + src1 = get_src_reg(c, &inst->SrcReg[1], i2, 1); + brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); + brw_MAC(p, dst, src0, src1); + brw_set_saturate(p, 0); + } + } + brw_set_saturate(p, 0); +} + +static void emit_dp3(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_reg src0[3], src1[3], dst; + int i; + struct brw_compile *p = &c->func; + for (i = 0; i < 3; i++) { + src0[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); + src1[i] = get_src_reg(c, &inst->SrcReg[1], i, 1); + } + + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_MAC(p, dst, src0[2], src1[2]); + brw_set_saturate(p, 0); +} + +static void emit_dp4(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_reg src0[4], src1[4], dst; + int i; + struct brw_compile *p = &c->func; + for (i = 0; i < 4; i++) { + src0[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); + src1[i] = get_src_reg(c, &inst->SrcReg[1], i, 1); + } + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_MAC(p, brw_null_reg(), src0[2], src1[2]); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_MAC(p, dst, src0[3], src1[3]); + brw_set_saturate(p, 0); +} + +static void emit_dph(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_reg src0[4], src1[4], dst; + int i; + struct brw_compile *p = &c->func; + for (i = 0; i < 4; i++) { + src0[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); + src1[i] = get_src_reg(c, &inst->SrcReg[1], i, 1); + } + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_MAC(p, dst, src0[2], src1[2]); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_ADD(p, dst, src0[3], src1[3]); + brw_set_saturate(p, 0); +} + +static void emit_math1(struct brw_wm_compile *c, + struct prog_instruction *inst, unsigned func) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; + + src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); + brw_MOV(p, brw_message_reg(2), src0); + brw_math(p, + dst, + func, + (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + 2, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); +} + +static void emit_rcp(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_INV); +} + +static void emit_rsq(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ); +} + +static void emit_sin(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_SIN); +} + +static void emit_cos(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_COS); +} + +static void emit_ex2(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_EXP); +} + +static void emit_lg2(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_math1(c, inst, BRW_MATH_FUNCTION_LOG); +} + +static void emit_add(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, src1, dst; + unsigned mask = inst->DstReg.WriteMask; + int i; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); + brw_ADD(p, dst, src0, src1); + } + } + brw_set_saturate(p, 0); +} + +static void emit_sub(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, src1, dst; + unsigned mask = inst->DstReg.WriteMask; + int i; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); + brw_ADD(p, dst, src0, negate(src1)); + } + } + brw_set_saturate(p, 0); +} + +static void emit_mul(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, src1, dst; + unsigned mask = inst->DstReg.WriteMask; + int i; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); + brw_MUL(p, dst, src0, src1); + } + } + brw_set_saturate(p, 0); +} + +static void emit_frc(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; + unsigned mask = inst->DstReg.WriteMask; + int i; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + brw_FRC(p, dst, src0); + } + } + if (inst->SaturateMode != SATURATE_OFF) + brw_set_saturate(p, 0); +} + +static void emit_flr(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; + unsigned mask = inst->DstReg.WriteMask; + int i; + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + brw_RNDD(p, dst, src0); + } + } + brw_set_saturate(p, 0); +} + +static void emit_max(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + struct brw_reg src0, src1, dst; + int i; + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_MOV(p, dst, src0); + brw_set_saturate(p, 0); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src0, src1); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, src1); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + } + } + brw_pop_insn_state(p); +} + +static void emit_min(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + struct brw_reg src0, src1, dst; + int i; + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_MOV(p, dst, src0); + brw_set_saturate(p, 0); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src1, src0); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, src1); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + } + } + brw_pop_insn_state(p); +} + +static void emit_pow(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg dst, src0, src1; + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); + src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); + src1 = get_src_reg(c, &inst->SrcReg[1], 0, 1); + + brw_MOV(p, brw_message_reg(2), src0); + brw_MOV(p, brw_message_reg(3), src1); + + brw_math(p, + dst, + BRW_MATH_FUNCTION_POW, + (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + 2, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); +} + +static void emit_lrp(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + struct brw_reg dst, tmp1, tmp2, src0, src1, src2; + int i; + for (i = 0; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + + src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); + + if (src1.nr == dst.nr) { + tmp1 = alloc_tmp(c); + brw_MOV(p, tmp1, src1); + } else + tmp1 = src1; + + src2 = get_src_reg(c, &inst->SrcReg[2], i, 1); + if (src2.nr == dst.nr) { + tmp2 = alloc_tmp(c); + brw_MOV(p, tmp2, src2); + } else + tmp2 = src2; + + brw_ADD(p, dst, negate(src0), brw_imm_f(1.0)); + brw_MUL(p, brw_null_reg(), dst, tmp2); + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_MAC(p, dst, src0, tmp1); + brw_set_saturate(p, 0); + } + release_tmps(c); + } +} + +static void emit_kil(struct brw_wm_compile *c) +{ + struct brw_compile *p = &c->func; + struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK + brw_AND(p, depth, c->emit_mask_reg, depth); + brw_pop_insn_state(p); +} + +static void emit_mad(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + struct brw_reg dst, src0, src1, src2; + int i; + + for (i = 0; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); + src2 = get_src_reg(c, &inst->SrcReg[2], i, 1); + brw_MUL(p, dst, src0, src1); + + brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); + brw_ADD(p, dst, dst, src2); + brw_set_saturate(p, 0); + } + } +} + +static void emit_sop(struct brw_wm_compile *c, + struct prog_instruction *inst, unsigned cond) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + struct brw_reg dst, src0, src1; + int i; + + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<SrcReg[0], i, 1); + src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); + brw_CMP(p, brw_null_reg(), cond, src0, src1); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_MOV(p, dst, brw_imm_f(0.0)); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, brw_imm_f(1.0)); + } + } + brw_pop_insn_state(p); +} + +static void emit_slt(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_L); +} + +static void emit_sle(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_LE); +} + +static void emit_sgt(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_G); +} + +static void emit_sge(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_GE); +} + +static void emit_seq(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_EQ); +} + +static void emit_sne(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + emit_sop(c, inst, BRW_CONDITIONAL_NEQ); +} + +static void emit_ddx(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->DstReg.WriteMask; + struct brw_reg interp[4]; + struct brw_reg dst; + struct brw_reg src0, w; + unsigned nr, i; + src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); + w = get_src_reg(c, &inst->SrcReg[1], 3, 1); + nr = src0.nr; + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + unsigned mask = inst->DstReg.WriteMask; + struct brw_reg interp[4]; + struct brw_reg dst; + struct brw_reg src0, w; + unsigned nr, i; + + src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); + nr = src0.nr; + w = get_src_reg(c, &inst->SrcReg[1], 3, 1); + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + unsigned mask = inst->DstReg.WriteMask; + struct brw_reg src0[2], dst[2]; + + dst[0] = get_dst_reg(c, inst, 0, 1); + dst[1] = get_dst_reg(c, inst, 1, 1); + + src0[0] = get_src_reg(c, &inst->SrcReg[0], 0, 1); + src0[1] = get_src_reg(c, &inst->SrcReg[0], 1, 1); + + /* Calc delta X,Y by subtracting origin in r1 from the pixel + * centers. + */ + if (mask & WRITEMASK_X) { + brw_MOV(p, + dst[0], + retype(src0[0], BRW_REGISTER_TYPE_UW)); + } + + if (mask & WRITEMASK_Y) { + /* TODO -- window_height - Y */ + brw_MOV(p, + dst[1], + retype(src0[1], BRW_REGISTER_TYPE_UW)); + + } +} + +/* TODO + BIAS on SIMD8 not workind yet... + */ +static void emit_txb(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg dst[4], src[4], payload_reg; + unsigned i; + payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0); + for (i = 0; i < 4; i++) + dst[i] = get_dst_reg(c, inst, i, 1); + for (i = 0; i < 4; i++) + src[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); + + switch (inst->TexSrcTarget) { + case TEXTURE_1D_INDEX: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), brw_imm_f(0)); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); + break; + case TEXTURE_2D_INDEX: + case TEXTURE_RECT_INDEX: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); + break; + default: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), src[2]); + break; + } + brw_MOV(p, brw_message_reg(5), src[3]); + brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); + brw_SAMPLE(p, + retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), + 1, + retype(payload_reg, BRW_REGISTER_TYPE_UW), + inst->TexSrcUnit + 1, /* surface */ + inst->TexSrcUnit, /* sampler */ + inst->DstReg.WriteMask, + BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS, + 4, + 4, + 0); +} + +static void emit_tex(struct brw_wm_compile *c, + struct prog_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg dst[4], src[4], payload_reg; + unsigned msg_len; + unsigned i, nr; + unsigned emit; + boolean shadow = (c->key.shadowtex_mask & (1<TexSrcUnit)) ? 1 : 0; + + payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0); + + for (i = 0; i < 4; i++) + dst[i] = get_dst_reg(c, inst, i, 1); + for (i = 0; i < 4; i++) + src[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); + + + switch (inst->TexSrcTarget) { + case TEXTURE_1D_INDEX: + emit = WRITEMASK_X; + nr = 1; + break; + case TEXTURE_2D_INDEX: + case TEXTURE_RECT_INDEX: + emit = WRITEMASK_XY; + nr = 2; + break; + default: + emit = WRITEMASK_XYZ; + nr = 3; + break; + } + msg_len = 1; + + for (i = 0; i < nr; i++) { + static const unsigned swz[4] = {0,1,2,2}; + if (emit & (1<TexSrcUnit + 1, /* surface */ + inst->TexSrcUnit, /* sampler */ + inst->DstReg.WriteMask, + BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE, + 4, + shadow ? 6 : 4, + 0); + + if (shadow) + brw_MOV(p, dst[3], brw_imm_f(1.0)); +} + +static void post_wm_emit( struct brw_wm_compile *c ) +{ + unsigned nr_insns = c->fp->program.Base.NumInstructions; + unsigned insn, target_insn; + struct prog_instruction *inst1, *inst2; + struct brw_instruction *brw_inst1, *brw_inst2; + int offset; + for (insn = 0; insn < nr_insns; insn++) { + inst1 = &c->fp->program.Base.Instructions[insn]; + brw_inst1 = inst1->Data; + switch (inst1->Opcode) { + case OPCODE_CAL: + target_insn = inst1->BranchTarget; + inst2 = &c->fp->program.Base.Instructions[target_insn]; + brw_inst2 = inst2->Data; + offset = brw_inst2 - brw_inst1; + brw_set_src1(brw_inst1, brw_imm_d(offset*16)); + break; + default: + break; + } + } +} + +static void brw_wm_emit_glsl(struct brw_wm_compile *c) + +{ +#define MAX_IFSN 32 +#define MAX_LOOP_DEPTH 32 + struct brw_instruction *if_inst[MAX_IFSN], *loop_inst[MAX_LOOP_DEPTH]; + struct brw_instruction *inst0, *inst1; + int i, if_insn = 0, loop_insn = 0; + struct brw_compile *p = &c->func; + struct brw_indirect stack_index = brw_indirect(0, 0); + + brw_init_compile(&c->func); + c->reg_index = 0; + prealloc_reg(c); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); + + for (i = 0; i < c->nr_fp_insns; i++) { + struct prog_instruction *inst = &c->prog_instructions[i]; + struct prog_instruction *orig_inst; + + if ((orig_inst = inst->Data) != 0) + orig_inst->Data = current_insn(p); + + if (inst->CondUpdate) + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + else + brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); + + switch (inst->Opcode) { + case WM_PIXELXY: + emit_pixel_xy(c, inst); + break; + case WM_DELTAXY: + emit_delta_xy(c, inst); + break; + case WM_PIXELW: + emit_pixel_w(c, inst); + break; + case WM_LINTERP: + emit_linterp(c, inst); + break; + case WM_PINTERP: + emit_pinterp(c, inst); + break; + case WM_CINTERP: + emit_cinterp(c, inst); + break; + case WM_WPOSXY: + emit_wpos_xy(c, inst); + break; + case WM_FB_WRITE: + emit_fb_write(c, inst); + break; + case OPCODE_ABS: + emit_abs(c, inst); + break; + case OPCODE_ADD: + emit_add(c, inst); + break; + case OPCODE_SUB: + emit_sub(c, inst); + break; + case OPCODE_FRC: + emit_frc(c, inst); + break; + case OPCODE_FLR: + emit_flr(c, inst); + break; + case OPCODE_LRP: + emit_lrp(c, inst); + break; + case OPCODE_INT: + emit_int(c, inst); + break; + case OPCODE_MOV: + emit_mov(c, inst); + break; + case OPCODE_DP3: + emit_dp3(c, inst); + break; + case OPCODE_DP4: + emit_dp4(c, inst); + break; + case OPCODE_XPD: + emit_xpd(c, inst); + break; + case OPCODE_DPH: + emit_dph(c, inst); + break; + case OPCODE_RCP: + emit_rcp(c, inst); + break; + case OPCODE_RSQ: + emit_rsq(c, inst); + break; + case OPCODE_SIN: + emit_sin(c, inst); + break; + case OPCODE_COS: + emit_cos(c, inst); + break; + case OPCODE_EX2: + emit_ex2(c, inst); + break; + case OPCODE_LG2: + emit_lg2(c, inst); + break; + case OPCODE_MAX: + emit_max(c, inst); + break; + case OPCODE_MIN: + emit_min(c, inst); + break; + case OPCODE_DDX: + emit_ddx(c, inst); + break; + case OPCODE_DDY: + emit_ddy(c, inst); + break; + case OPCODE_SLT: + emit_slt(c, inst); + break; + case OPCODE_SLE: + emit_sle(c, inst); + break; + case OPCODE_SGT: + emit_sgt(c, inst); + break; + case OPCODE_SGE: + emit_sge(c, inst); + break; + case OPCODE_SEQ: + emit_seq(c, inst); + break; + case OPCODE_SNE: + emit_sne(c, inst); + break; + case OPCODE_MUL: + emit_mul(c, inst); + break; + case OPCODE_POW: + emit_pow(c, inst); + break; + case OPCODE_MAD: + emit_mad(c, inst); + break; + case OPCODE_TEX: + emit_tex(c, inst); + break; + case OPCODE_TXB: + emit_txb(c, inst); + break; + case OPCODE_KIL_NV: + emit_kil(c); + break; + case OPCODE_IF: + assert(if_insn < MAX_IFSN); + if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8); + break; + case OPCODE_ELSE: + if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]); + break; + case OPCODE_ENDIF: + assert(if_insn > 0); + brw_ENDIF(p, if_inst[--if_insn]); + break; + case OPCODE_BGNSUB: + case OPCODE_ENDSUB: + break; + case OPCODE_CAL: + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_ADD(p, deref_1ud(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(4)); + orig_inst = inst->Data; + orig_inst->Data = &p->store[p->nr_insn]; + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + brw_pop_insn_state(p); + break; + + case OPCODE_RET: + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(-4)); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, brw_ip_reg(), deref_1ud(stack_index, 0)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_pop_insn_state(p); + + break; + case OPCODE_BGNLOOP: + loop_inst[loop_insn++] = brw_DO(p, BRW_EXECUTE_8); + break; + case OPCODE_BRK: + brw_BREAK(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case OPCODE_CONT: + brw_CONT(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case OPCODE_ENDLOOP: + loop_insn--; + inst0 = inst1 = brw_WHILE(p, loop_inst[loop_insn]); + /* patch all the BREAK instructions from + last BEGINLOOP */ + while (inst0 > loop_inst[loop_insn]) { + inst0--; + if (inst0->header.opcode == BRW_OPCODE_BREAK) { + inst0->bits3.if_else.jump_count = inst1 - inst0 + 1; + inst0->bits3.if_else.pop_count = 0; + } else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) { + inst0->bits3.if_else.jump_count = inst1 - inst0; + inst0->bits3.if_else.pop_count = 0; + } + } + break; + default: + _mesa_printf("unsupported IR in fragment shader %d\n", + inst->Opcode); + } + if (inst->CondUpdate) + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + else + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + post_wm_emit(c); + for (i = 0; i < c->fp->program.Base.NumInstructions; i++) + c->fp->program.Base.Instructions[i].Data = NULL; +} + +void brw_wm_glsl_emit(struct brw_wm_compile *c) +{ + brw_wm_pass_fp(c); + c->tmp_index = 127; + brw_wm_emit_glsl(c); + c->prog_data.total_grf = c->reg_index; + c->prog_data.total_scratch = 0; +} +#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_iz.c b/src/mesa/pipe/i965simple/brw_wm_iz.c new file mode 100644 index 0000000000..6c5f25bf39 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_iz.c @@ -0,0 +1,214 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_wm.h" + + +#undef P /* prompted depth */ +#undef C /* computed */ +#undef N /* non-promoted? */ + +#define P 0 +#define C 1 +#define N 2 + +const struct { + unsigned mode:2; + unsigned sd_present:1; + unsigned sd_to_rt:1; + unsigned dd_present:1; + unsigned ds_present:1; +} wm_iz_table[IZ_BIT_MAX] = +{ + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 1, 1, 0, 0 }, + { C, 1, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 1, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 1, 1, 0, 0 }, + { C, 1, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 1, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 0, 0, 1 }, + { C, 0, 0, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 1, 1, 0, 1 }, + { C, 1, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 1, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 0, 0, 1 }, + { C, 0, 0, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 1, 1, 0, 1 }, + { C, 1, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 1, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 1 }, + { N, 0, 1, 0, 1 }, + { N, 0, 1, 0, 1 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { C, 0, 0, 0, 1 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 0, 1 }, + { P, 0, 0, 0, 0 }, + { C, 1, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { P, 0, 0, 0, 0 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 } +}; + +void brw_wm_lookup_iz( unsigned line_aa, + unsigned lookup, + struct brw_wm_prog_key *key ) +{ + unsigned reg = 2; + + assert (lookup < IZ_BIT_MAX); + + if (lookup & IZ_PS_COMPUTES_DEPTH_BIT) + key->computes_depth = 1; + + if (wm_iz_table[lookup].sd_present) { + key->source_depth_reg = reg; + reg += 2; + } + + if (wm_iz_table[lookup].sd_to_rt) + key->source_depth_to_render_target = 1; + + if (wm_iz_table[lookup].ds_present || line_aa != AA_NEVER) { + key->aa_dest_stencil_reg = reg; + key->runtime_check_aads_emit = (!wm_iz_table[lookup].ds_present && + line_aa == AA_SOMETIMES); + reg++; + } + + if (wm_iz_table[lookup].dd_present) { + key->dest_depth_reg = reg; + reg+=2; + } + + key->nr_depth_regs = (reg+1)/2; +} + diff --git a/src/mesa/pipe/i965simple/brw_wm_pass0.c b/src/mesa/pipe/i965simple/brw_wm_pass0.c new file mode 100644 index 0000000000..2ec3ac00c1 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_pass0.c @@ -0,0 +1,466 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_wm.h" + + +#if 0 +/*********************************************************************** + */ + +static struct brw_wm_ref *get_ref( struct brw_wm_compile *c ) +{ + assert(c->nr_refs < BRW_WM_MAX_REF); + return &c->refs[c->nr_refs++]; +} + +static struct brw_wm_value *get_value( struct brw_wm_compile *c) +{ + assert(c->nr_refs < BRW_WM_MAX_VREG); + return &c->vreg[c->nr_vreg++]; +} + +static struct brw_wm_instruction *get_instruction( struct brw_wm_compile *c ) +{ + assert(c->nr_insns < BRW_WM_MAX_INSN); + return &c->instruction[c->nr_insns++]; +} + +/*********************************************************************** + */ + +static void pass0_init_undef( struct brw_wm_compile *c) +{ + struct brw_wm_ref *ref = &c->undef_ref; + ref->value = &c->undef_value; + ref->hw_reg = brw_vec8_grf(0, 0); + ref->insn = 0; + ref->prevuse = NULL; +} + +static void pass0_set_fpreg_value( struct brw_wm_compile *c, + unsigned file, + unsigned idx, + unsigned component, + struct brw_wm_value *value ) +{ + struct brw_wm_ref *ref = get_ref(c); + ref->value = value; + ref->hw_reg = brw_vec8_grf(0, 0); + ref->insn = 0; + ref->prevuse = NULL; + c->pass0_fp_reg[file][idx][component] = ref; +} + +static void pass0_set_fpreg_ref( struct brw_wm_compile *c, + unsigned file, + unsigned idx, + unsigned component, + const struct brw_wm_ref *src_ref ) +{ + c->pass0_fp_reg[file][idx][component] = src_ref; +} + +static const struct brw_wm_ref *get_param_ref( struct brw_wm_compile *c, + const float *param_ptr ) +{ + unsigned i = c->prog_data.nr_params++; + + if (i >= BRW_WM_MAX_PARAM) { + _mesa_printf("%s: out of params\n", __FUNCTION__); + c->prog_data.error = 1; + return NULL; + } + else { + struct brw_wm_ref *ref = get_ref(c); + + c->prog_data.param[i] = param_ptr; + c->nr_creg = (i+16)/16; + + /* Push the offsets into hw_reg. These will be added to the + * real register numbers once one is allocated in pass2. + */ + ref->hw_reg = brw_vec1_grf((i&8)?1:0, i%8); + ref->value = &c->creg[i/16]; + ref->insn = 0; + ref->prevuse = NULL; + + return ref; + } +} + + +static const struct brw_wm_ref *get_const_ref( struct brw_wm_compile *c, + const float *constval ) +{ + unsigned i; + + /* Search for an existing const value matching the request: + */ + for (i = 0; i < c->nr_constrefs; i++) { + if (c->constref[i].constval == *constval) + return c->constref[i].ref; + } + + /* Else try to add a new one: + */ + if (c->nr_constrefs < BRW_WM_MAX_CONST) { + unsigned i = c->nr_constrefs++; + + /* A constant is a special type of parameter: + */ + c->constref[i].constval = *constval; + c->constref[i].ref = get_param_ref(c, constval); + + return c->constref[i].ref; + } + else { + _mesa_printf("%s: out of constrefs\n", __FUNCTION__); + c->prog_data.error = 1; + return NULL; + } +} + + +/* Lookup our internal registers + */ +static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c, + unsigned file, + unsigned idx, + unsigned component ) +{ + const struct brw_wm_ref *ref = c->pass0_fp_reg[file][idx][component]; + + if (!ref) { + switch (file) { + case PROGRAM_INPUT: + case PROGRAM_PAYLOAD: + case PROGRAM_TEMPORARY: + case PROGRAM_OUTPUT: + case PROGRAM_VARYING: + break; + + case PROGRAM_LOCAL_PARAM: + ref = get_param_ref(c, &c->fp->program.Base.LocalParams[idx][component]); + break; + + case PROGRAM_ENV_PARAM: + ref = get_param_ref(c, &c->env_param[idx][component]); + break; + + case PROGRAM_STATE_VAR: + case PROGRAM_UNIFORM: + case PROGRAM_CONSTANT: + case PROGRAM_NAMED_PARAM: { + struct gl_program_parameter_list *plist = c->fp->program.Base.Parameters; + + /* There's something really hokey about parameters parsed in + * arb programs - they all end up in here, whether they be + * state values, paramters or constants. This duplicates the + * structure above & also seems to subvert the limits set for + * each type of constant/param. + */ + switch (plist->Parameters[idx].Type) { + case PROGRAM_NAMED_PARAM: + case PROGRAM_CONSTANT: + /* These are invarient: + */ + ref = get_const_ref(c, &plist->ParameterValues[idx][component]); + break; + + case PROGRAM_STATE_VAR: + case PROGRAM_UNIFORM: + /* These may change from run to run: + */ + ref = get_param_ref(c, &plist->ParameterValues[idx][component] ); + break; + + default: + assert(0); + break; + } + break; + } + + default: + assert(0); + break; + } + + c->pass0_fp_reg[file][idx][component] = ref; + } + + if (!ref) + ref = &c->undef_ref; + + return ref; +} + + + + +/*********************************************************************** + * Straight translation to internal instruction format + */ + +static void pass0_set_dst( struct brw_wm_compile *c, + struct brw_wm_instruction *out, + const struct prog_instruction *inst, + unsigned writemask ) +{ + const struct prog_dst_register *dst = &inst->DstReg; + unsigned i; + + for (i = 0; i < 4; i++) { + if (writemask & (1<dst[i] = get_value(c); + + pass0_set_fpreg_value(c, dst->File, dst->Index, i, out->dst[i]); + } + } + + out->writemask = writemask; +} + + +static void pass0_set_dst_scalar( struct brw_wm_compile *c, + struct brw_wm_instruction *out, + const struct prog_instruction *inst, + unsigned writemask ) +{ + if (writemask) { + const struct prog_dst_register *dst = &inst->DstReg; + unsigned i; + + /* Compute only the first (X) value: + */ + out->writemask = WRITEMASK_X; + out->dst[0] = get_value(c); + + /* Update our tracking register file for all the components in + * writemask: + */ + for (i = 0; i < 4; i++) { + if (writemask & (1<File, dst->Index, i, out->dst[0]); + } + } + } + else + out->writemask = 0; +} + + + +static const struct brw_wm_ref *get_fp_src_reg_ref( struct brw_wm_compile *c, + struct prog_src_register src, + unsigned i ) +{ + unsigned component = GET_SWZ(src.Swizzle,i); + const struct brw_wm_ref *src_ref; + static const float const_zero = 0.0; + static const float const_one = 1.0; + + + if (component == SWIZZLE_ZERO) + src_ref = get_const_ref(c, &const_zero); + else if (component == SWIZZLE_ONE) + src_ref = get_const_ref(c, &const_one); + else + src_ref = pass0_get_reg(c, src.File, src.Index, component); + + return src_ref; +} + + +static struct brw_wm_ref *get_new_ref( struct brw_wm_compile *c, + struct prog_src_register src, + unsigned i, + struct brw_wm_instruction *insn) +{ + const struct brw_wm_ref *ref = get_fp_src_reg_ref(c, src, i); + struct brw_wm_ref *newref = get_ref(c); + + newref->value = ref->value; + newref->hw_reg = ref->hw_reg; + + if (insn) { + newref->insn = insn - c->instruction; + newref->prevuse = newref->value->lastuse; + newref->value->lastuse = newref; + } + + if (src.NegateBase & (1<hw_reg.negate ^= 1; + + if (src.Abs) { + newref->hw_reg.negate = 0; + newref->hw_reg.abs = 1; + } + + return newref; +} + + + +static struct brw_wm_instruction *translate_insn( struct brw_wm_compile *c, + const struct prog_instruction *inst ) +{ + struct brw_wm_instruction *out = get_instruction(c); + unsigned writemask = inst->DstReg.WriteMask; + unsigned nr_args = brw_wm_nr_args(inst->Opcode); + unsigned i, j; + + /* Copy some data out of the instruction + */ + out->opcode = inst->Opcode; + out->saturate = (inst->SaturateMode != SATURATE_OFF); + out->tex_unit = inst->TexSrcUnit; + out->tex_idx = inst->TexSrcTarget; + + /* Args: + */ + for (i = 0; i < nr_args; i++) { + for (j = 0; j < 4; j++) { + out->src[i][j] = get_new_ref(c, inst->SrcReg[i], j, out); + } + } + + /* Dst: + */ + if (brw_wm_is_scalar_result(out->opcode)) + pass0_set_dst_scalar(c, out, inst, writemask); + else + pass0_set_dst(c, out, inst, writemask); + + return out; +} + + + +/*********************************************************************** + * Optimize moves and swizzles away: + */ +static void pass0_precalc_mov( struct brw_wm_compile *c, + const struct prog_instruction *inst ) +{ + const struct prog_dst_register *dst = &inst->DstReg; + unsigned writemask = inst->DstReg.WriteMask; + unsigned i; + + /* Get the effect of a MOV by manipulating our register table: + */ + for (i = 0; i < 4; i++) { + if (writemask & (1<File, dst->Index, i, + get_new_ref(c, inst->SrcReg[0], i, NULL)); + } + } +} + + +/* Initialize payload "registers". + */ +static void pass0_init_payload( struct brw_wm_compile *c ) +{ + unsigned i; + + for (i = 0; i < 4; i++) { + unsigned j = i >= c->key.nr_depth_regs ? 0 : i; + pass0_set_fpreg_value( c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, i, + &c->payload.depth[j] ); + } + +#if 0 + /* This seems to be an alternative to the INTERP_WPOS stuff I do + * elsewhere: + */ + if (c->key.source_depth_reg) + pass0_set_fpreg_value(c, PROGRAM_INPUT, FRAG_ATTRIB_WPOS, 2, + &c->payload.depth[c->key.source_depth_reg/2]); +#endif + + for (i = 0; i < FRAG_ATTRIB_MAX; i++) + pass0_set_fpreg_value( c, PROGRAM_PAYLOAD, i, 0, + &c->payload.input_interp[i] ); +} + +/*********************************************************************** + * PASS 0 + * + * Work forwards to give each calculated value a unique number. Where + * an instruction produces duplicate values (eg DP3), all are given + * the same number. + * + * Translate away swizzling and eliminate non-saturating moves. + */ +void brw_wm_pass0( struct brw_wm_compile *c ) +{ + unsigned insn; + + c->nr_vreg = 0; + c->nr_insns = 0; + + pass0_init_undef(c); + pass0_init_payload(c); + + for (insn = 0; insn < c->nr_fp_insns; insn++) { + const struct prog_instruction *inst = &c->prog_instructions[insn]; + + + /* Optimize away moves, otherwise emit translated instruction: + */ + switch (inst->Opcode) { + case OPCODE_MOV: + case OPCODE_SWZ: + if (!inst->SaturateMode) { + pass0_precalc_mov(c, inst); + } + else { + translate_insn(c, inst); + } + break; + + + default: + translate_insn(c, inst); + break; + } + } + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_program(c, "pass0"); + } +} + +#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_pass1.c b/src/mesa/pipe/i965simple/brw_wm_pass1.c new file mode 100644 index 0000000000..e6736bd6df --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_pass1.c @@ -0,0 +1,277 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_wm.h" + +#if 0 +static unsigned get_tracked_mask(struct brw_wm_compile *c, + struct brw_wm_instruction *inst) +{ + unsigned i; + for (i = 0; i < 4; i++) { + if (inst->writemask & (1<dst[i]->contributes_to_output) { + inst->writemask &= ~(1<dst[i] = 0; + } + } + } + + return inst->writemask; +} + +/* Remove a reference from a value's usage chain. + */ +static void unlink_ref(struct brw_wm_ref *ref) +{ + struct brw_wm_value *value = ref->value; + + if (ref == value->lastuse) { + value->lastuse = ref->prevuse; + } else { + struct brw_wm_ref *i = value->lastuse; + while (i->prevuse != ref) i = i->prevuse; + i->prevuse = ref->prevuse; + } +} + +static void track_arg(struct brw_wm_compile *c, + struct brw_wm_instruction *inst, + unsigned arg, + unsigned readmask) +{ + unsigned i; + + for (i = 0; i < 4; i++) { + struct brw_wm_ref *ref = inst->src[arg][i]; + if (ref) { + if (readmask & (1<value->contributes_to_output = 1; + else { + unlink_ref(ref); + inst->src[arg][i] = NULL; + } + } + } +} + +static unsigned get_texcoord_mask( unsigned tex_idx ) +{ + switch (tex_idx) { + case TEXTURE_1D_INDEX: return WRITEMASK_X; + case TEXTURE_2D_INDEX: return WRITEMASK_XY; + case TEXTURE_3D_INDEX: return WRITEMASK_XYZ; + case TEXTURE_CUBE_INDEX: return WRITEMASK_XYZ; + case TEXTURE_RECT_INDEX: return WRITEMASK_XY; + default: return 0; + } +} + +/* Step two: Basically this is dead code elimination. + * + * Iterate backwards over instructions, noting which values + * contribute to the final result. Adjust writemasks to only + * calculate these values. + */ +void brw_wm_pass1( struct brw_wm_compile *c ) +{ + int insn; + + for (insn = c->nr_insns-1; insn >= 0; insn--) { + struct brw_wm_instruction *inst = &c->instruction[insn]; + unsigned writemask; + unsigned read0, read1, read2; + + if (inst->opcode == OPCODE_KIL) { + track_arg(c, inst, 0, WRITEMASK_XYZW); /* All args contribute to final */ + continue; + } + + if (inst->opcode == WM_FB_WRITE) { + track_arg(c, inst, 0, WRITEMASK_XYZW); + track_arg(c, inst, 1, WRITEMASK_XYZW); + if (c->key.source_depth_to_render_target && + c->key.computes_depth) + track_arg(c, inst, 2, WRITEMASK_Z); + else + track_arg(c, inst, 2, 0); + continue; + } + + /* Lookup all the registers which were written by this + * instruction and get a mask of those that contribute to the output: + */ + writemask = get_tracked_mask(c, inst); + if (!writemask) { + unsigned arg; + for (arg = 0; arg < 3; arg++) + track_arg(c, inst, arg, 0); + continue; + } + + read0 = 0; + read1 = 0; + read2 = 0; + + /* Mark all inputs which contribute to the marked outputs: + */ + switch (inst->opcode) { + case OPCODE_ABS: + case OPCODE_FLR: + case OPCODE_FRC: + case OPCODE_MOV: + read0 = writemask; + break; + + case OPCODE_SUB: + case OPCODE_SLT: + case OPCODE_SLE: + case OPCODE_SGE: + case OPCODE_SGT: + case OPCODE_SEQ: + case OPCODE_SNE: + case OPCODE_ADD: + case OPCODE_MAX: + case OPCODE_MIN: + case OPCODE_MUL: + read0 = writemask; + read1 = writemask; + break; + + case OPCODE_MAD: + case OPCODE_CMP: + case OPCODE_LRP: + read0 = writemask; + read1 = writemask; + read2 = writemask; + break; + + case OPCODE_XPD: + if (writemask & WRITEMASK_X) read0 |= WRITEMASK_YZ; + if (writemask & WRITEMASK_Y) read0 |= WRITEMASK_XZ; + if (writemask & WRITEMASK_Z) read0 |= WRITEMASK_XY; + read1 = read0; + break; + + case OPCODE_COS: + case OPCODE_EX2: + case OPCODE_LG2: + case OPCODE_RCP: + case OPCODE_RSQ: + case OPCODE_SIN: + case OPCODE_SCS: + case WM_CINTERP: + case WM_PIXELXY: + read0 = WRITEMASK_X; + break; + + case OPCODE_POW: + read0 = WRITEMASK_X; + read1 = WRITEMASK_X; + break; + + case OPCODE_TEX: + read0 = get_texcoord_mask(inst->tex_idx); + + if (c->key.shadowtex_mask & (1<tex_unit)) + read0 |= WRITEMASK_Z; + break; + + case OPCODE_TXB: + /* Shadow ignored for txb. + */ + read0 = get_texcoord_mask(inst->tex_idx) | WRITEMASK_W; + break; + + case WM_WPOSXY: + read0 = writemask & WRITEMASK_XY; + break; + + case WM_DELTAXY: + read0 = writemask & WRITEMASK_XY; + read1 = WRITEMASK_X; + break; + + case WM_PIXELW: + read0 = WRITEMASK_X; + read1 = WRITEMASK_XY; + break; + + case WM_LINTERP: + read0 = WRITEMASK_X; + read1 = WRITEMASK_XY; + break; + + case WM_PINTERP: + read0 = WRITEMASK_X; /* interpolant */ + read1 = WRITEMASK_XY; /* deltas */ + read2 = WRITEMASK_W; /* pixel w */ + break; + + case OPCODE_DP3: + read0 = WRITEMASK_XYZ; + read1 = WRITEMASK_XYZ; + break; + + case OPCODE_DPH: + read0 = WRITEMASK_XYZ; + read1 = WRITEMASK_XYZW; + break; + + case OPCODE_DP4: + read0 = WRITEMASK_XYZW; + read1 = WRITEMASK_XYZW; + break; + + case OPCODE_LIT: + read0 = WRITEMASK_XYW; + break; + + case OPCODE_SWZ: + case OPCODE_DST: + case OPCODE_TXP: + default: + break; + } + + track_arg(c, inst, 0, read0); + track_arg(c, inst, 1, read1); + track_arg(c, inst, 2, read2); + } + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_program(c, "pass1"); + } +} + +#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_pass2.c b/src/mesa/pipe/i965simple/brw_wm_pass2.c new file mode 100644 index 0000000000..b1b404706f --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_pass2.c @@ -0,0 +1,335 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_wm.h" + +#if 0 +/* Use these to force spilling so that that functionality can be + * tested with known-good examples rather than having to construct new + * tests. + */ +#define TEST_PAYLOAD_SPILLS 0 +#define TEST_DST_SPILLS 0 + +static void spill_value(struct brw_wm_compile *c, + struct brw_wm_value *value); + +static void prealloc_reg(struct brw_wm_compile *c, + struct brw_wm_value *value, + unsigned reg) +{ + if (value->lastuse) { + /* Set nextuse to zero, it will be corrected by + * update_register_usage(). + */ + c->pass2_grf[reg].value = value; + c->pass2_grf[reg].nextuse = 0; + + value->resident = &c->pass2_grf[reg]; + value->hw_reg = brw_vec8_grf(reg*2, 0); + + if (TEST_PAYLOAD_SPILLS) + spill_value(c, value); + } +} + + +/* Initialize all the register values. Do the initial setup + * calculations for interpolants. + */ +static void init_registers( struct brw_wm_compile *c ) +{ + unsigned inputs = FRAG_BIT_WPOS | c->fp_interp_emitted; + unsigned nr_interp_regs = 0; + unsigned i = 0; + unsigned j; + + for (j = 0; j < c->grf_limit; j++) + c->pass2_grf[j].nextuse = BRW_WM_MAX_INSN; + + for (j = 0; j < c->key.nr_depth_regs; j++) + prealloc_reg(c, &c->payload.depth[j], i++); + + for (j = 0; j < c->nr_creg; j++) + prealloc_reg(c, &c->creg[j], i++); + + for (j = 0; j < FRAG_ATTRIB_MAX; j++) + if (inputs & (1<payload.input_interp[j], i++); + } + + assert(nr_interp_regs >= 1); + + c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; + c->prog_data.urb_read_length = nr_interp_regs * 2; + c->prog_data.curb_read_length = c->nr_creg * 2; + + c->max_wm_grf = i * 2; +} + + +/* Update the nextuse value for each register in our file. + */ +static void update_register_usage(struct brw_wm_compile *c, + unsigned thisinsn) +{ + unsigned i; + + for (i = 1; i < c->grf_limit; i++) { + struct brw_wm_grf *grf = &c->pass2_grf[i]; + + /* Only search those which can change: + */ + if (grf->nextuse < thisinsn) { + struct brw_wm_ref *ref = grf->value->lastuse; + + /* Has last use of value been passed? + */ + if (ref->insn < thisinsn) { + grf->value->resident = 0; + grf->value = 0; + grf->nextuse = BRW_WM_MAX_INSN; + } + else { + /* Else loop through chain to update: + */ + while (ref->prevuse && ref->prevuse->insn >= thisinsn) + ref = ref->prevuse; + + grf->nextuse = ref->insn; + } + } + } +} + + +static void spill_value(struct brw_wm_compile *c, + struct brw_wm_value *value) +{ + /* Allocate a spill slot. Note that allocations start from 0x40 - + * the first slot is reserved to mean "undef" in brw_wm_emit.c + */ + if (!value->spill_slot) { + c->last_scratch += 0x40; + value->spill_slot = c->last_scratch; + } + + /* The spill will be done in brw_wm_emit.c immediately after the + * value is calculated, so we can just take this reg without any + * further work. + */ + value->resident->value = NULL; + value->resident->nextuse = BRW_WM_MAX_INSN; + value->resident = NULL; +} + + + +/* Search for contiguous region with the most distant nearest + * member. Free regs count as very distant. + * + * TODO: implement spill-to-reg so that we can rearrange discontigous + * free regs and then spill the oldest non-free regs in sequence. + * This would mean inserting instructions in this pass. + */ +static unsigned search_contiguous_regs(struct brw_wm_compile *c, + unsigned nr, + unsigned thisinsn) +{ + struct brw_wm_grf *grf = c->pass2_grf; + unsigned furthest = 0; + unsigned reg = 0; + unsigned i, j; + + /* Start search at 1: r0 is special and can't be used or spilled. + */ + for (i = 1; i < c->grf_limit && furthest < BRW_WM_MAX_INSN; i++) { + unsigned group_nextuse = BRW_WM_MAX_INSN; + + for (j = 0; j < nr; j++) { + if (grf[i+j].nextuse < group_nextuse) + group_nextuse = grf[i+j].nextuse; + } + + if (group_nextuse > furthest) { + furthest = group_nextuse; + reg = i; + } + } + + assert(furthest != thisinsn); + + /* Any non-empty regs will need to be spilled: + */ + for (j = 0; j < nr; j++) + if (grf[reg+j].value) + spill_value(c, grf[reg+j].value); + + return reg; +} + + +static void alloc_contiguous_dest(struct brw_wm_compile *c, + struct brw_wm_value *dst[], + unsigned nr, + unsigned thisinsn) +{ + unsigned reg = search_contiguous_regs(c, nr, thisinsn); + unsigned i; + + for (i = 0; i < nr; i++) { + if (!dst[i]) { + /* Need to grab a dummy value in TEX case. Don't introduce + * it into the tracking scheme. + */ + dst[i] = &c->vreg[c->nr_vreg++]; + } + else { + assert(!dst[i]->resident); + assert(c->pass2_grf[reg+i].nextuse != thisinsn); + + c->pass2_grf[reg+i].value = dst[i]; + c->pass2_grf[reg+i].nextuse = thisinsn; + + dst[i]->resident = &c->pass2_grf[reg+i]; + } + + dst[i]->hw_reg = brw_vec8_grf((reg+i)*2, 0); + } + + if ((reg+nr)*2 > c->max_wm_grf) + c->max_wm_grf = (reg+nr) * 2; +} + + +static void load_args(struct brw_wm_compile *c, + struct brw_wm_instruction *inst) +{ + unsigned thisinsn = inst - c->instruction; + unsigned i,j; + + for (i = 0; i < 3; i++) { + for (j = 0; j < 4; j++) { + struct brw_wm_ref *ref = inst->src[i][j]; + + if (ref) { + if (!ref->value->resident) { + /* Need to bring the value in from scratch space. The code for + * this will be done in brw_wm_emit.c, here we just do the + * register allocation and mark the ref as requiring a fill. + */ + unsigned reg = search_contiguous_regs(c, 1, thisinsn); + + c->pass2_grf[reg].value = ref->value; + c->pass2_grf[reg].nextuse = thisinsn; + + ref->value->resident = &c->pass2_grf[reg]; + + /* Note that a fill is required: + */ + ref->unspill_reg = reg*2; + } + + /* Adjust the hw_reg to point at the value's current location: + */ + assert(ref->value == ref->value->resident->value); + ref->hw_reg.nr += (ref->value->resident - c->pass2_grf) * 2; + } + } + } +} + + + +/* Step 3: Work forwards once again. Perform register allocations, + * taking into account instructions like TEX which require contiguous + * result registers. Where necessary spill registers to scratch space + * and reload later. + */ +void brw_wm_pass2( struct brw_wm_compile *c ) +{ + unsigned insn; + unsigned i; + + init_registers(c); + + for (insn = 0; insn < c->nr_insns; insn++) { + struct brw_wm_instruction *inst = &c->instruction[insn]; + + /* Update registers' nextuse values: + */ + update_register_usage(c, insn); + + /* May need to unspill some args. + */ + load_args(c, inst); + + /* Allocate registers to hold results: + */ + switch (inst->opcode) { + case OPCODE_TEX: + case OPCODE_TXB: + case OPCODE_TXP: + alloc_contiguous_dest(c, inst->dst, 4, insn); + break; + + default: + for (i = 0; i < 4; i++) { + if (inst->writemask & (1<dst[i]); + alloc_contiguous_dest(c, &inst->dst[i], 1, insn); + } + } + break; + } + + if (TEST_DST_SPILLS && inst->opcode != WM_PIXELXY) + for (i = 0; i < 4; i++) + if (inst->dst[i]) + spill_value(c, inst->dst[i]); + + } + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_program(c, "pass2"); + } + + c->state = PASS2_DONE; + + if (BRW_DEBUG & DEBUG_WM) { + brw_wm_print_program(c, "pass2/done"); + } +} + +#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c b/src/mesa/pipe/i965simple/brw_wm_sampler_state.c new file mode 100644 index 0000000000..4ea0dd7db0 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_sampler_state.c @@ -0,0 +1,275 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + +#include "pipe/p_util.h" + + +#define COMPAREFUNC_ALWAYS 0 +#define COMPAREFUNC_NEVER 0x1 +#define COMPAREFUNC_LESS 0x2 +#define COMPAREFUNC_EQUAL 0x3 +#define COMPAREFUNC_LEQUAL 0x4 +#define COMPAREFUNC_GREATER 0x5 +#define COMPAREFUNC_NOTEQUAL 0x6 +#define COMPAREFUNC_GEQUAL 0x7 + +/* Samplers aren't strictly wm state from the hardware's perspective, + * but that is the only situation in which we use them in this driver. + */ + +static int intel_translate_shadow_compare_func(unsigned func) +{ + switch(func) { + case PIPE_FUNC_NEVER: + return COMPAREFUNC_ALWAYS; + case PIPE_FUNC_LESS: + return COMPAREFUNC_LEQUAL; + case PIPE_FUNC_LEQUAL: + return COMPAREFUNC_LESS; + case PIPE_FUNC_GREATER: + return COMPAREFUNC_GEQUAL; + case PIPE_FUNC_GEQUAL: + return COMPAREFUNC_GREATER; + case PIPE_FUNC_NOTEQUAL: + return COMPAREFUNC_EQUAL; + case PIPE_FUNC_EQUAL: + return COMPAREFUNC_NOTEQUAL; + case PIPE_FUNC_ALWAYS: + return COMPAREFUNC_NEVER; + } + + fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); + return COMPAREFUNC_NEVER; +} + +/* The brw (and related graphics cores) do not support GL_CLAMP. The + * Intel drivers for "other operating systems" implement GL_CLAMP as + * GL_CLAMP_TO_EDGE, so the same is done here. + */ +static unsigned translate_wrap_mode( int wrap ) +{ + switch( wrap ) { + case PIPE_TEX_WRAP_REPEAT: + return BRW_TEXCOORDMODE_WRAP; + case PIPE_TEX_WRAP_CLAMP: + return BRW_TEXCOORDMODE_CLAMP; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return BRW_TEXCOORDMODE_CLAMP; /* conform likes it this way */ + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return BRW_TEXCOORDMODE_CLAMP_BORDER; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + return BRW_TEXCOORDMODE_MIRROR; + default: + return BRW_TEXCOORDMODE_WRAP; + } +} + + +static unsigned U_FIXED(float value, unsigned frac_bits) +{ + value *= (1<cache[BRW_SAMPLER_DEFAULT_COLOR], &sdc ); +} + + +/* + */ +static void brw_update_sampler_state( const struct pipe_sampler_state *pipe_sampler, + unsigned sdc_gs_offset, + struct brw_sampler_state *sampler) +{ + memset(sampler, 0, sizeof(*sampler)); + + switch (pipe_sampler->min_mip_filter) { + case PIPE_TEX_FILTER_NEAREST: + sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST; + break; + case PIPE_TEX_FILTER_LINEAR: + sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR; + break; + default: + break; + } + + switch (pipe_sampler->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + sampler->ss0.mip_filter = BRW_MIPFILTER_NONE; + break; + default: + break; + } + /* Set Anisotropy: + */ + if (pipe_sampler->max_anisotropy > 1.0) { + sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC; + sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC; + + if (pipe_sampler->max_anisotropy > 2.0) { + sampler->ss3.max_aniso = MAX2((pipe_sampler->max_anisotropy - 2) / 2, + BRW_ANISORATIO_16); + } + } + else { + switch (pipe_sampler->mag_img_filter) { + case PIPE_TEX_FILTER_NEAREST: + sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST; + break; + case PIPE_TEX_FILTER_LINEAR: + sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR; + break; + default: + break; + } + } + + sampler->ss1.s_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_s); + sampler->ss1.r_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_r); + sampler->ss1.t_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_t); + + /* Fulsim complains if I don't do this. Hardware doesn't mind: + */ +#if 0 + if (texObj->Target == GL_TEXTURE_CUBE_MAP_ARB) { + sampler->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CUBE; + sampler->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CUBE; + sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CUBE; + } +#endif + + /* Set shadow function: + */ + if (pipe_sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + /* Shadowing is "enabled" by emitting a particular sampler + * message (sample_c). So need to recompile WM program when + * shadow comparison is enabled on each/any texture unit. + */ + sampler->ss0.shadow_function = intel_translate_shadow_compare_func(pipe_sampler->compare_func); + } + + /* Set LOD bias: + */ + sampler->ss0.lod_bias = S_FIXED(CLAMP(pipe_sampler->lod_bias, -16, 15), 6); + + sampler->ss0.lod_preclamp = 1; /* OpenGL mode */ + sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */ + + /* Set BaseMipLevel, MaxLOD, MinLOD: + * + * XXX: I don't think that using firstLevel, lastLevel works, + * because we always setup the surface state as if firstLevel == + * level zero. Probably have to subtract firstLevel from each of + * these: + */ + sampler->ss0.base_level = U_FIXED(0, 1); + + sampler->ss1.max_lod = U_FIXED(MIN2(MAX2(pipe_sampler->max_lod, 0), 13), 6); + sampler->ss1.min_lod = U_FIXED(MIN2(MAX2(pipe_sampler->min_lod, 0), 13), 6); + + sampler->ss2.default_color_pointer = sdc_gs_offset >> 5; +} + + + +/* All samplers must be uploaded in a single contiguous array, which + * complicates various things. However, this is still too confusing - + * FIXME: simplify all the different new texture state flags. + */ +void brw_upload_wm_samplers(struct brw_context *brw) +{ + unsigned unit; + unsigned sampler_count = 0; + + /* _NEW_TEXTURE */ + for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { + if (brw->attribs.Samplers[unit]) { /* FIXME: correctly detect enabled ones */ + const struct pipe_sampler_state *sampler = brw->attribs.Samplers[unit]; + unsigned sdc_gs_offset = upload_default_color(brw, sampler->border_color); + + brw_update_sampler_state(sampler, + sdc_gs_offset, + &brw->wm.sampler[unit]); + + sampler_count = unit + 1; + } + } + + if (brw->wm.sampler_count != sampler_count) { + brw->wm.sampler_count = sampler_count; + brw->state.dirty.cache |= CACHE_NEW_SAMPLER; + } + + brw->wm.sampler_gs_offset = 0; + + if (brw->wm.sampler_count) + brw->wm.sampler_gs_offset = + brw_cache_data_sz(&brw->cache[BRW_SAMPLER], + brw->wm.sampler, + sizeof(struct brw_sampler_state) * brw->wm.sampler_count); +} + +#if 0 +const struct brw_tracked_state brw_wm_samplers = { + .dirty = { + .mesa = _NEW_TEXTURE, + .brw = 0, + .cache = 0 + }, + .update = upload_wm_samplers +}; +#endif + diff --git a/src/mesa/pipe/i965simple/brw_wm_state.c b/src/mesa/pipe/i965simple/brw_wm_state.c new file mode 100644 index 0000000000..b45fb2f56b --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_state.c @@ -0,0 +1,202 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_wm.h" + +/*********************************************************************** + * WM unit - fragment programs and rasterization + */ + +#if 0 +static void upload_wm_unit(struct brw_context *brw ) +{ + struct intel_context *intel = &brw->intel; + struct brw_wm_unit_state wm; + unsigned max_threads; + unsigned per_thread; + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + max_threads = 0; + else + max_threads = 31; + + + memset(&wm, 0, sizeof(wm)); + + /* CACHE_NEW_WM_PROG */ + wm.thread0.grf_reg_count = ALIGN(brw->wm.prog_data->total_grf, 16) / 16 - 1; + wm.thread0.kernel_start_pointer = brw->wm.prog_gs_offset >> 6; + wm.thread3.dispatch_grf_start_reg = brw->wm.prog_data->first_curbe_grf; + wm.thread3.urb_entry_read_length = brw->wm.prog_data->urb_read_length; + wm.thread3.const_urb_entry_read_length = brw->wm.prog_data->curb_read_length; + + wm.wm5.max_threads = max_threads; + + per_thread = ALIGN(brw->wm.prog_data->total_scratch, 1024); + assert(per_thread <= 12 * 1024); + + if (brw->wm.prog_data->total_scratch) { + unsigned total = per_thread * (max_threads + 1); + + /* Scratch space -- just have to make sure there is sufficient + * allocated for the active program and current number of threads. + */ + brw->wm.scratch_buffer_size = total; + if (brw->wm.scratch_buffer && + brw->wm.scratch_buffer_size > brw->wm.scratch_buffer->size) { + dri_bo_unreference(brw->wm.scratch_buffer); + brw->wm.scratch_buffer = NULL; + } + if (!brw->wm.scratch_buffer) { + brw->wm.scratch_buffer = dri_bo_alloc(intel->intelScreen->bufmgr, + "wm scratch", + brw->wm.scratch_buffer_size, + 4096, DRM_BO_FLAG_MEM_TT); + } + } + /* XXX: Scratch buffers are not implemented correectly. + * + * The scratch offset to be programmed into wm is relative to the general + * state base address. However, using dri_bo_alloc/dri_bo_emit_reloc (or + * the previous bmGenBuffers scheme), we get an offset relative to the + * start of framebuffer. Even before then, it was broken in other ways, + * so just fail for now if we hit that path. + */ + assert(brw->wm.prog_data->total_scratch == 0); + + /* CACHE_NEW_SURFACE */ + wm.thread1.binding_table_entry_count = brw->wm.nr_surfaces; + + /* BRW_NEW_CURBE_OFFSETS */ + wm.thread3.const_urb_entry_read_offset = brw->curbe.wm_start * 2; + + wm.thread3.urb_entry_read_offset = 0; + wm.thread1.depth_coef_urb_read_offset = 1; + wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + + /* CACHE_NEW_SAMPLER */ + wm.wm4.sampler_count = (brw->wm.sampler_count + 1) / 4; + wm.wm4.sampler_state_pointer = brw->wm.sampler_gs_offset >> 5; + + /* BRW_NEW_FRAGMENT_PROGRAM */ + { + const struct gl_fragment_program *fp = brw->fragment_program; + + if (fp->Base.InputsRead & (1<Base.OutputsWritten & (1<UsesKill || + brw->attribs.Color->AlphaEnabled) + wm.wm5.program_uses_killpixel = 1; + + if (brw_wm_is_glsl(fp)) + wm.wm5.enable_8_pix = 1; + else + wm.wm5.enable_16_pix = 1; + } + + wm.wm5.thread_dispatch_enable = 1; /* AKA: color_write */ + wm.wm5.legacy_line_rast = 0; + wm.wm5.legacy_global_depth_bias = 0; + wm.wm5.early_depth_test = 1; /* never need to disable */ + wm.wm5.line_aa_region_width = 0; + wm.wm5.line_endcap_aa_region_width = 1; + + /* _NEW_POLYGONSTIPPLE */ + if (brw->attribs.Polygon->StippleFlag) + wm.wm5.polygon_stipple = 1; + + /* _NEW_POLYGON */ + if (brw->attribs.Polygon->OffsetFill) { + wm.wm5.depth_offset = 1; + /* Something wierd going on with legacy_global_depth_bias, + * offset_constant, scaling and MRD. This value passes glean + * but gives some odd results elsewere (eg. the + * quad-offset-units test). + */ + wm.global_depth_offset_constant = brw->attribs.Polygon->OffsetUnits * 2; + + /* This is the only value that passes glean: + */ + wm.global_depth_offset_scale = brw->attribs.Polygon->OffsetFactor; + } + + /* _NEW_LINE */ + if (brw->attribs.Line->StippleFlag) { + wm.wm5.line_stipple = 1; + } + + if (BRW_DEBUG & DEBUG_STATS || intel->stats_wm) + wm.wm4.stats_enable = 1; + + brw->wm.state_gs_offset = brw_cache_data( &brw->cache[BRW_WM_UNIT], &wm ); + + if (brw->wm.prog_data->total_scratch) { + /* + dri_emit_reloc(brw->cache[BRW_WM_UNIT].pool->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + (per_thread / 1024) - 1, + brw->wm.state_gs_offset + + ((char *)&wm.thread2 - (char *)&wm), + brw->wm.scratch_buffer); + */ + } else { + wm.thread2.scratch_space_base_pointer = 0; + } +} + +const struct brw_tracked_state brw_wm_unit = { + .dirty = { + .mesa = (_NEW_POLYGON | + _NEW_POLYGONSTIPPLE | + _NEW_LINE | + _NEW_COLOR), + + .brw = (BRW_NEW_FRAGMENT_PROGRAM | + BRW_NEW_CURBE_OFFSETS | + BRW_NEW_LOCK), + + .cache = (CACHE_NEW_SURFACE | + CACHE_NEW_WM_PROG | + CACHE_NEW_SAMPLER) + }, + .update = upload_wm_unit +}; + +#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c new file mode 100644 index 0000000000..c04fd57494 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -0,0 +1,294 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + + +#if 0 +static unsigned translate_tex_target( int target ) +{ + switch (target) { + case PIPE_TEXTURE_1D: + return BRW_SURFACE_1D; + + case PIPE_TEXTURE_2D: + return BRW_SURFACE_2D; + + case PIPE_TEXTURE_3D: + return BRW_SURFACE_3D; + + case PIPE_TEXTURE_CUBE: + return BRW_SURFACE_CUBE; + + default: + assert(0); + return 0; + } +} + +static unsigned translate_tex_format( unsigned mesa_format ) +{ + switch( mesa_format ) { + case PIPE_FORMAT_U_L8: + return BRW_SURFACEFORMAT_L8_UNORM; + + case PIPE_FORMAT_U_I8: + return BRW_SURFACEFORMAT_I8_UNORM; + + case PIPE_FORMAT_U_A8: + return BRW_SURFACEFORMAT_A8_UNORM; + + case PIPE_FORMAT_U_A8_L8: + return BRW_SURFACEFORMAT_L8A8_UNORM; + + case PIPE_FORMAT_R8G8B8_UNORM: + assert(0); /* not supported for sampling */ + return BRW_SURFACEFORMAT_R8G8B8_UNORM; + + case PIPE_FORMAT_U_A8_R8_G8_B8: + return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + + case PIPE_FORMAT_RGBA8888_REV: + return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; + + case PIPE_FORMAT_U_R5_G6_B5: + return BRW_SURFACEFORMAT_B5G6R5_UNORM; + + case PIPE_FORMAT_A1R5G5B5_UNORM: + return BRW_SURFACEFORMAT_B5G5R5A1_UNORM; + + case PIPE_FORMAT_A4R4G4B4_UNORM: + return BRW_SURFACEFORMAT_B4G4R4A4_UNORM; + + case PIPE_FORMAT_YCBCR_REV: + return BRW_SURFACEFORMAT_YCRCB_NORMAL; + + case PIPE_FORMAT_YCBCR: + return BRW_SURFACEFORMAT_YCRCB_SWAPUVY; + + case PIPE_FORMAT_RGB_FXT1: + case PIPE_FORMAT_RGBA_FXT1: + return BRW_SURFACEFORMAT_FXT1; + + case PIPE_FORMAT_Z16_UNORM: + return BRW_SURFACEFORMAT_I16_UNORM; + + case PIPE_FORMAT_RGB_DXT1: + return BRW_SURFACEFORMAT_DXT1_RGB; + + case PIPE_FORMAT_RGBA_DXT1: + return BRW_SURFACEFORMAT_BC1_UNORM; + + case PIPE_FORMAT_RGBA_DXT3: + return BRW_SURFACEFORMAT_BC2_UNORM; + + case PIPE_FORMAT_RGBA_DXT5: + return BRW_SURFACEFORMAT_BC3_UNORM; + + case PIPE_FORMAT_SRGBA8: + return BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB; + case PIPE_FORMAT_SRGB_DXT1: + return BRW_SURFACEFORMAT_BC1_UNORM_SRGB; + + default: + assert(0); + return 0; + } +} + +static +void brw_update_texture_surface( GLcontext *ctx, + unsigned unit ) +{ + struct brw_context *brw = brw_context(ctx); + struct gl_texture_object *tObj = brw->attribs.Texture->Unit[unit]._Current; + struct intel_texture_object *intelObj = intel_texture_object(tObj); + struct gl_texture_image *firstImage = tObj->Image[0][intelObj->firstLevel]; + struct brw_surface_state surf; + + memset(&surf, 0, sizeof(surf)); + + surf.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW; + surf.ss0.surface_type = translate_tex_target(tObj->Target); + surf.ss0.surface_format = translate_tex_format(firstImage->TexFormat->MesaFormat); + + /* This is ok for all textures with channel width 8bit or less: + */ +/* surf.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */ + + /* Updated in emit_reloc */ + surf.ss1.base_addr = intelObj->mt->region->buffer->offset; + + surf.ss2.mip_count = intelObj->lastLevel - intelObj->firstLevel; + surf.ss2.width = firstImage->Width - 1; + surf.ss2.height = firstImage->Height - 1; + + surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; + surf.ss3.tiled_surface = intelObj->mt->region->tiled; /* always zero */ + surf.ss3.pitch = (intelObj->mt->pitch * intelObj->mt->cpp) - 1; + surf.ss3.depth = firstImage->Depth - 1; + + surf.ss4.min_lod = 0; + + if (tObj->Target == GL_TEXTURE_CUBE_MAP) { + surf.ss0.cube_pos_x = 1; + surf.ss0.cube_pos_y = 1; + surf.ss0.cube_pos_z = 1; + surf.ss0.cube_neg_x = 1; + surf.ss0.cube_neg_y = 1; + surf.ss0.cube_neg_z = 1; + } + + brw->wm.bind.surf_ss_offset[unit + 1] = + brw_cache_data( &brw->cache[BRW_SS_SURFACE], &surf ); +} + + + +#define OFFSET(TYPE, FIELD) ( (unsigned)&(((TYPE *)0)->FIELD) ) + + +static void upload_wm_surfaces(struct brw_context *brw ) +{ + GLcontext *ctx = &brw->intel.ctx; + struct intel_context *intel = &brw->intel; + unsigned i; + + { + struct brw_surface_state surf; + struct intel_region *region = brw->state.draw_region; + + memset(&surf, 0, sizeof(surf)); + + if (region != NULL) { + if (region->cpp == 4) + surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + else + surf.ss0.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM; + + surf.ss0.surface_type = BRW_SURFACE_2D; + + surf.ss2.width = region->pitch - 1; /* XXX: not really! */ + surf.ss2.height = region->height - 1; + surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; + surf.ss3.tiled_surface = region->tiled; + surf.ss3.pitch = (region->pitch * region->cpp) - 1; + } else { + surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + surf.ss0.surface_type = BRW_SURFACE_NULL; + } + + /* _NEW_COLOR */ + surf.ss0.color_blend = (!brw->attribs.Color->_LogicOpEnabled && + brw->attribs.Color->BlendEnabled); + + + surf.ss0.writedisable_red = !brw->attribs.Color->ColorMask[0]; + surf.ss0.writedisable_green = !brw->attribs.Color->ColorMask[1]; + surf.ss0.writedisable_blue = !brw->attribs.Color->ColorMask[2]; + surf.ss0.writedisable_alpha = !brw->attribs.Color->ColorMask[3]; + + brw->wm.bind.surf_ss_offset[0] = brw_cache_data( &brw->cache[BRW_SS_SURFACE], &surf ); + + brw->wm.nr_surfaces = 1; + } + + + for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { + struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i]; + + /* _NEW_TEXTURE, BRW_NEW_TEXDATA + */ + if (texUnit->_ReallyEnabled && + intel_finalize_mipmap_tree(intel, i)) { + + brw_update_texture_surface(ctx, i); + + brw->wm.nr_surfaces = i+2; + } + else if( texUnit->_ReallyEnabled && + texUnit->_Current == intel->frame_buffer_texobj ) + { + brw->wm.bind.surf_ss_offset[i+1] = brw->wm.bind.surf_ss_offset[0]; + brw->wm.nr_surfaces = i+2; + } + else { + brw->wm.bind.surf_ss_offset[i+1] = 0; + } + } + + brw->wm.bind_ss_offset = brw_cache_data( &brw->cache[BRW_SS_SURF_BIND], + &brw->wm.bind ); +} + +static void emit_reloc_wm_surfaces(struct brw_context *brw) +{ + int unit; + + if (brw->state.draw_region != NULL) { + /* Emit framebuffer relocation */ + dri_emit_reloc(brw_cache_buffer(brw, BRW_SS_SURFACE), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, + brw->wm.bind.surf_ss_offset[0] + + offsetof(struct brw_surface_state, ss1), + brw->state.draw_region->buffer); + } + + /* Emit relocations for texture buffers */ + for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { + struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[unit]; + struct gl_texture_object *tObj = texUnit->_Current; + struct intel_texture_object *intelObj = intel_texture_object(tObj); + + if (texUnit->_ReallyEnabled && intelObj->mt != NULL) { + dri_emit_reloc(brw_cache_buffer(brw, BRW_SS_SURFACE), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + 0, + brw->wm.bind.surf_ss_offset[unit + 1] + + offsetof(struct brw_surface_state, ss1), + intelObj->mt->region->buffer); + } + } +} + +const struct brw_tracked_state brw_wm_surfaces = { + .dirty = { + .mesa = _NEW_COLOR | _NEW_TEXTURE | _NEW_BUFFERS, + .brw = BRW_NEW_CONTEXT, + .cache = 0 + }, + .update = upload_wm_surfaces, + .emit_reloc = emit_reloc_wm_surfaces, +}; +#endif diff --git a/src/mesa/pipe/xlib/brw_aub.c b/src/mesa/pipe/xlib/brw_aub.c new file mode 100644 index 0000000000..1b7fc3c20e --- /dev/null +++ b/src/mesa/pipe/xlib/brw_aub.c @@ -0,0 +1,392 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_aub.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" +#include "imports.h" +//#include "intel_winsys.h" + + +struct brw_aubfile { + FILE *file; + unsigned next_free_page; +}; + + +extern char *__progname; + + +struct aub_file_header { + unsigned int instruction_type; + unsigned int pad0:16; + unsigned int minor:8; + unsigned int major:8; + unsigned char application[8*4]; + unsigned int day:8; + unsigned int month:8; + unsigned int year:16; + unsigned int timezone:8; + unsigned int second:8; + unsigned int minute:8; + unsigned int hour:8; + unsigned int comment_length:16; + unsigned int pad1:16; +}; + +struct aub_block_header { + unsigned int instruction_type; + unsigned int operation:8; + unsigned int type:8; + unsigned int address_space:8; + unsigned int pad0:8; + unsigned int general_state_type:8; + unsigned int surface_state_type:8; + unsigned int pad1:16; + unsigned int address; + unsigned int length; +}; + +struct aub_dump_bmp { + unsigned int instruction_type; + unsigned int xmin:16; + unsigned int ymin:16; + unsigned int pitch:16; + unsigned int bpp:8; + unsigned int format:8; + unsigned int xsize:16; + unsigned int ysize:16; + unsigned int addr; + unsigned int unknown; +}; + +enum bh_operation { + BH_COMMENT, + BH_DATA_WRITE, + BH_COMMAND_WRITE, + BH_MMI0_WRITE32, + BH_END_SCENE, + BH_CONFIG_MEMORY_MAP, + BH_MAX_OPERATION +}; + +enum command_write_type { + CW_HWB_RING = 1, + CW_PRIMARY_RING_A, + CW_PRIMARY_RING_B, /* XXX - disagreement with listaub! */ + CW_PRIMARY_RING_C, + CW_MAX_TYPE +}; + +enum memory_map_type { + MM_DEFAULT, + MM_DYNAMIC, + MM_MAX_TYPE +}; + +enum address_space { + ADDR_GTT, + ADDR_LOCAL, + ADDR_MAIN, + ADDR_MAX +}; + + +#define AUB_FILE_HEADER 0xe085000b +#define AUB_BLOCK_HEADER 0xe0c10003 +#define AUB_DUMP_BMP 0xe09e0004 + +/* Registers to control page table + */ +#define PGETBL_CTL 0x2020 +#define PGETBL_ENABLED 0x1 + +#define NR_GTT_ENTRIES 65536 /* 256 mb */ + +#define FAIL \ +do { \ + fprintf(stderr, "failed to write aub data at %s/%d\n", __FUNCTION__, __LINE__); \ + exit(1); \ +} while (0) + + +/* Emit the headers at the top of each aubfile. Initialize the GTT. + */ +static void init_aubfile( FILE *aub_file ) +{ + struct aub_file_header fh; + struct aub_block_header bh; + unsigned int data; + + static int nr; + + nr++; + + /* Emit the aub header: + */ + memset(&fh, 0, sizeof(fh)); + + fh.instruction_type = AUB_FILE_HEADER; + fh.minor = 0x0; + fh.major = 0x7; + memcpy(fh.application, __progname, sizeof(fh.application)); + fh.day = (nr>>24) & 0xff; + fh.month = 0x0; + fh.year = 0x0; + fh.timezone = 0x0; + fh.second = nr & 0xff; + fh.minute = (nr>>8) & 0xff; + fh.hour = (nr>>16) & 0xff; + fh.comment_length = 0x0; + + if (fwrite(&fh, sizeof(fh), 1, aub_file) < 0) + FAIL; + + /* Setup the GTT starting at main memory address zero (!): + */ + memset(&bh, 0, sizeof(bh)); + + bh.instruction_type = AUB_BLOCK_HEADER; + bh.operation = BH_MMI0_WRITE32; + bh.type = 0x0; + bh.address_space = ADDR_GTT; /* ??? */ + bh.general_state_type = 0x0; + bh.surface_state_type = 0x0; + bh.address = PGETBL_CTL; + bh.length = 0x4; + + if (fwrite(&bh, sizeof(bh), 1, aub_file) < 0) + FAIL; + + data = 0x0 | PGETBL_ENABLED; + + if (fwrite(&data, sizeof(data), 1, aub_file) < 0) + FAIL; +} + + +static void init_aub_gtt( struct brw_aubfile *aubfile, + unsigned start_offset, + unsigned size ) +{ + FILE *aub_file = aubfile->file; + struct aub_block_header bh; + unsigned int i; + + assert(start_offset + size < NR_GTT_ENTRIES * 4096); + + + memset(&bh, 0, sizeof(bh)); + + bh.instruction_type = AUB_BLOCK_HEADER; + bh.operation = BH_DATA_WRITE; + bh.type = 0x0; + bh.address_space = ADDR_MAIN; + bh.general_state_type = 0x0; + bh.surface_state_type = 0x0; + bh.address = start_offset / 4096 * 4; + bh.length = size / 4096 * 4; + + if (fwrite(&bh, sizeof(bh), 1, aub_file) < 0) + FAIL; + + for (i = 0; i < size / 4096; i++) { + unsigned data = aubfile->next_free_page | 1; + + aubfile->next_free_page += 4096; + + if (fwrite(&data, sizeof(data), 1, aub_file) < 0) + FAIL; + } + +} + +static void write_block_header( FILE *aub_file, + struct aub_block_header *bh, + const unsigned *data, + unsigned sz ) +{ + sz = (sz + 3) & ~3; + + if (fwrite(bh, sizeof(*bh), 1, aub_file) < 0) + FAIL; + + if (fwrite(data, sz, 1, aub_file) < 0) + FAIL; + + fflush(aub_file); +} + + +static void write_dump_bmp( FILE *aub_file, + struct aub_dump_bmp *db ) +{ + if (fwrite(db, sizeof(*db), 1, aub_file) < 0) + FAIL; + + fflush(aub_file); +} + + + +void brw_aub_gtt_data( struct brw_aubfile *aubfile, + unsigned offset, + const void *data, + unsigned sz, + unsigned type, + unsigned state_type ) +{ + struct aub_block_header bh; + + bh.instruction_type = AUB_BLOCK_HEADER; + bh.operation = BH_DATA_WRITE; + bh.type = type; + bh.address_space = ADDR_GTT; + bh.pad0 = 0; + + if (type == DW_GENERAL_STATE) { + bh.general_state_type = state_type; + bh.surface_state_type = 0; + } + else { + bh.general_state_type = 0; + bh.surface_state_type = state_type; + } + + bh.pad1 = 0; + bh.address = offset; + bh.length = sz; + + write_block_header(aubfile->file, &bh, data, sz); +} + + + +void brw_aub_gtt_cmds( struct brw_aubfile *aubfile, + unsigned offset, + const void *data, + unsigned sz ) +{ + struct aub_block_header bh; + unsigned type = CW_PRIMARY_RING_A; + + + bh.instruction_type = AUB_BLOCK_HEADER; + bh.operation = BH_COMMAND_WRITE; + bh.type = type; + bh.address_space = ADDR_GTT; + bh.pad0 = 0; + bh.general_state_type = 0; + bh.surface_state_type = 0; + bh.pad1 = 0; + bh.address = offset; + bh.length = sz; + + write_block_header(aubfile->file, &bh, data, sz); +} + +void brw_aub_dump_bmp( struct brw_aubfile *aubfile, + struct pipe_surface *surface, + unsigned gtt_offset ) +{ + struct aub_dump_bmp db; + unsigned format; + + if (surface->cpp == 4) + format = 0x7; + else + format = 0x3; + + db.instruction_type = AUB_DUMP_BMP; + db.xmin = 0; + db.ymin = 0; + db.format = format; + db.bpp = surface->cpp * 8; + db.pitch = surface->pitch; + db.xsize = surface->pitch; + db.ysize = surface->height; + db.addr = gtt_offset; + db.unknown = /* surface->tiled ? 0x4 : */ 0x0; + + write_dump_bmp(aubfile->file, &db); +} + + + +struct brw_aubfile *brw_aubfile_create( void ) +{ + struct brw_aubfile *aubfile = CALLOC_STRUCT(brw_aubfile); + char filename[80]; + int val; + static int i = 0; + + i++; + + if (_mesa_getenv("INTEL_AUBFILE")) { + val = snprintf(filename, sizeof(filename), "%s%d.aub", _mesa_getenv("INTEL_AUBFILE"), i%4); + _mesa_printf("--> Aub file: %s\n", filename); + aubfile->file = fopen(filename, "w"); + } + else { + val = snprintf(filename, sizeof(filename), "%s.aub", __progname); + if (val < 0 || val > sizeof(filename)) + strcpy(filename, "default.aub"); + + _mesa_printf("--> Aub file: %s\n", filename); + aubfile->file = fopen(filename, "w"); + } + + if (!aubfile->file) { + _mesa_printf("couldn't open aubfile\n"); + exit(1); + } + + init_aubfile(aubfile->file); + + /* The GTT is located starting address zero in main memory. Pages + * to populate the gtt start after this point. + */ + aubfile->next_free_page = (NR_GTT_ENTRIES * 4 + 4095) & ~4095; + + /* More or less correspond with all the agp regions mapped by the + * driver: + */ + init_aub_gtt(aubfile, 0, 4096*4); + init_aub_gtt(aubfile, AUB_BUF_START, AUB_BUF_SIZE); + + return aubfile; +} + +void brw_aub_destroy( struct brw_aubfile *aubfile ) +{ + fclose(aubfile->file); + FREE(aubfile); +} diff --git a/src/mesa/pipe/xlib/brw_aub.h b/src/mesa/pipe/xlib/brw_aub.h new file mode 100644 index 0000000000..f5c60c7be2 --- /dev/null +++ b/src/mesa/pipe/xlib/brw_aub.h @@ -0,0 +1,114 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#ifndef BRW_AUB_H +#define BRW_AUB_H + +/* We set up this region, buffers may be allocated here: + */ +#define AUB_BUF_START (4096*4) +#define AUB_BUF_SIZE (8*1024*1024) + +struct intel_context; +struct pipe_surface; + +struct brw_aubfile *brw_aubfile_create( void ); + +void brw_aub_destroy( struct brw_aubfile *aubfile ); + +void brw_aub_gtt_data( struct brw_aubfile *aubfile, + unsigned offset, + const void *data, + unsigned sz, + unsigned type, + unsigned state_type ); + +void brw_aub_gtt_cmds( struct brw_aubfile *aubfile, + unsigned offset, + const void *data, + unsigned sz ); + +void brw_aub_dump_bmp( struct brw_aubfile *aubfile, + struct pipe_surface *surface, + unsigned gtt_offset ); + + +enum data_write_type { + DW_NOTYPE, + DW_BATCH_BUFFER, + DW_BIN_BUFFER, + DW_BIN_POINTER_LIST, + DW_SLOW_STATE_BUFFER, + DW_VERTEX_BUFFER, + DW_2D_MAP, + DW_CUBE_MAP, + DW_INDIRECT_STATE_BUFFER, + DW_VOLUME_MAP, + DW_1D_MAP, + DW_CONSTANT_BUFFER, + DW_CONSTANT_URB_ENTRY, + DW_INDEX_BUFFER, + DW_GENERAL_STATE, + DW_SURFACE_STATE, + DW_MEDIA_OBJECT_INDIRECT_DATA, + DW_MAX_TYPE +}; + +enum data_write_general_state_type { + DWGS_NOTYPE, + DWGS_VERTEX_SHADER_STATE, + DWGS_GEOMETRY_SHADER_STATE , + DWGS_CLIPPER_STATE, + DWGS_STRIPS_FANS_STATE, + DWGS_WINDOWER_IZ_STATE, + DWGS_COLOR_CALC_STATE, + DWGS_CLIPPER_VIEWPORT_STATE, /* was 0x7 */ + DWGS_STRIPS_FANS_VIEWPORT_STATE, + DWGS_COLOR_CALC_VIEWPORT_STATE, /* was 0x9 */ + DWGS_SAMPLER_STATE, + DWGS_KERNEL_INSTRUCTIONS, + DWGS_SCRATCH_SPACE, + DWGS_SAMPLER_DEFAULT_COLOR, + DWGS_INTERFACE_DESCRIPTOR, + DWGS_VLD_STATE, + DWGS_VFE_STATE, + DWGS_MAX_TYPE +}; + +enum data_write_surface_state_type { + DWSS_NOTYPE, + DWSS_BINDING_TABLE_STATE, + DWSS_SURFACE_STATE, + DWSS_MAX_TYPE +}; + + +#endif diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 9c7713790b..83e2b529a4 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -75,12 +75,16 @@ #include "pipe/p_defines.h" #include "pipe/p_context.h" +#include "xm_winsys_aub.h" + /** * Global X driver lock */ _glthread_Mutex _xmesa_lock; +int xmesa_mode; + /**********************************************************************/ /***** X Utility Functions *****/ @@ -738,7 +742,14 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) pf = choose_pixel_format(v); assert(pf); - pipe = xmesa_create_pipe_context( c, pf ); + if (!getenv("XM_AUB")) { + xmesa_mode = XMESA_SOFTPIPE; + pipe = xmesa_create_pipe_context( c, pf ); + } + else { + xmesa_mode = XMESA_AUB; + pipe = xmesa_create_i965simple( xmesa_get_pipe_winsys_aub() ); + } c->st = st_create_context(pipe, &v->mesa_visual, share_list ? share_list->st : NULL); @@ -1164,7 +1175,10 @@ void XMesaSwapBuffers( XMesaBuffer b ) surf = st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT); if (surf) { - xmesa_display_surface(b, surf); + if (xmesa_mode == XMESA_AUB) + xmesa_display_aub( surf ); + else + xmesa_display_surface(b, surf); } xmesa_check_and_update_buffer_size(NULL, b); diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index ad31f4498e..a03c9835a5 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -46,6 +46,7 @@ #include "pipe/cell/ppu/cell_context.h" #include "pipe/cell/ppu/cell_winsys.h" #endif +#include "xm_winsys_aub.h" /** XXX from Mesa core */ @@ -468,12 +469,15 @@ xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) * For Xlib, this is a singleton object. * Nothing special for the Xlib driver so no subclassing or anything. */ -static struct pipe_winsys * -xmesa_get_pipe_winsys(void) +struct pipe_winsys * +xmesa_get_pipe_winsys_aub(void) { static struct pipe_winsys *ws = NULL; - if (!ws) { + if (!ws && getenv("XM_AUB")) { + ws = xmesa_create_pipe_winsys_aub(); + } + else if (!ws) { ws = CALLOC_STRUCT(pipe_winsys); /* Fill in this struct with callbacks that pipe will need to @@ -537,7 +541,7 @@ xmesa_get_softpipe_winsys(uint pixelformat) struct pipe_context * xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) { - struct pipe_winsys *pws = xmesa_get_pipe_winsys(); + struct pipe_winsys *pws = xmesa_get_pipe_winsys_aub(); struct pipe_context *pipe; #ifdef GALLIUM_CELL diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h index e8c2a597a6..fa8d1f14b9 100644 --- a/src/mesa/pipe/xlib/xmesaP.h +++ b/src/mesa/pipe/xlib/xmesaP.h @@ -41,6 +41,12 @@ extern _glthread_Mutex _xmesa_lock; extern XMesaBuffer XMesaBufferList; +/* + */ +#define XMESA_SOFTPIPE 1 +#define XMESA_AUB 2 +extern int xmesa_mode; + /** * Visual inforation, derived from GLvisual. diff --git a/src/mesa/sources b/src/mesa/sources index 32e1805eca..c34361eaae 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -334,7 +334,9 @@ X11_DRIVER_SOURCES = \ pipe/xlib/fakeglx.c \ pipe/xlib/xfonts.c \ pipe/xlib/xm_api.c \ - pipe/xlib/xm_winsys.c + pipe/xlib/xm_winsys.c \ + pipe/xlib/xm_winsys_aub.c \ + pipe/xlib/brw_aub.c OSMESA_DRIVER_SOURCES = \ drivers/osmesa/osmesa.c -- cgit v1.2.3 From 60b8900b13a591e4a761baa0d6d40bcaca0c36d8 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 11 Dec 2007 07:44:07 -0500 Subject: Hmm, missed these two files. Sorry. --- src/mesa/pipe/xlib/xm_winsys_aub.c | 618 +++++++++++++++++++++++++++++++++++++ src/mesa/pipe/xlib/xm_winsys_aub.h | 67 ++++ 2 files changed, 685 insertions(+) create mode 100644 src/mesa/pipe/xlib/xm_winsys_aub.c create mode 100644 src/mesa/pipe/xlib/xm_winsys_aub.h (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c new file mode 100644 index 0000000000..ee3c2d6181 --- /dev/null +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -0,0 +1,618 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "glxheader.h" +#include "xmesaP.h" +#include "main/macros.h" + +#include "pipe/p_winsys.h" +#include "pipe/i965simple/brw_winsys.h" +#include "brw_aub.h" +#include "xm_winsys_aub.h" + + + +struct aub_buffer { + char *data; + unsigned offset; + unsigned size; + unsigned refcount; + unsigned map_count; + boolean dump_on_unmap; + boolean userbuffer; +}; + + + +struct aub_pipe_winsys { + struct pipe_winsys winsys; + + struct brw_aubfile *aubfile; + + /* This is simple, isn't it: + */ + char *pool; + unsigned size; + unsigned used; +}; + + +/* Turn a pipe winsys into an aub/pipe winsys: + */ +static inline struct aub_pipe_winsys * +aub_pipe_winsys( struct pipe_winsys *winsys ) +{ + return (struct aub_pipe_winsys *)winsys; +} + + + +static INLINE struct aub_buffer * +aub_bo( struct pipe_buffer_handle *bo ) +{ + return (struct aub_buffer *)bo; +} + +static INLINE struct pipe_buffer_handle * +pipe_bo( struct aub_buffer *bo ) +{ + return (struct pipe_buffer_handle *)bo; +} + + + + +static void *aub_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned flags ) +{ + struct aub_buffer *sbo = aub_bo(buf); + + if (flags & PIPE_BUFFER_FLAG_WRITE) + sbo->dump_on_unmap = 1; + + sbo->map_count++; + return sbo->data; +} + +static void aub_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf) +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + struct aub_buffer *sbo = aub_bo(buf); + + sbo->map_count--; + + if (sbo->map_count == 0 && + sbo->dump_on_unmap) { + brw_aub_gtt_data( iws->aubfile, + sbo->offset, + sbo->data, + sbo->size, + 0, + 0); + } +} + + +static void +aub_buffer_reference(struct pipe_winsys *winsys, + struct pipe_buffer_handle **ptr, + struct pipe_buffer_handle *buf) +{ + if (*ptr) { + if (--(aub_bo(*ptr)->refcount) == 0) + free(*ptr); + *ptr = NULL; + } + + if (buf) { + aub_bo(buf)->refcount++; + *ptr = buf; + } +} + + +static int aub_buffer_data(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned size, const void *data, + unsigned usage ) +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + struct aub_buffer *sbo = aub_bo(buf); + + /* Could reuse buffers that are not referenced in current + * batchbuffer. Can't do that atm, so always reallocate: + */ + if (1 || sbo->size < size) { + assert(iws->used + size < iws->size); + sbo->data = iws->pool + iws->used; + sbo->offset = AUB_BUF_START + iws->used; + iws->used += size; + } + + sbo->size = size; + + if (data != NULL) { + memcpy(iws->pool, data, size); + + brw_aub_gtt_data( iws->aubfile, + sbo->offset, + sbo->data, + sbo->size, + 0, + 0 ); + } + return 0; +} + +static int aub_buffer_subdata(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data) +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + struct aub_buffer *sbo = aub_bo(buf); + + assert(sbo->size > offset + size); + memcpy(sbo->data + offset, data, size); + + brw_aub_gtt_data( iws->aubfile, + sbo->offset + offset, + sbo->data + offset, + size, + 0, + 0 ); + return 0; +} + + +void xmesa_buffer_subdata_aub(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data, + unsigned aub_type, + unsigned aub_sub_type) +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + struct aub_buffer *sbo = aub_bo(buf); + + assert(sbo->size > offset + size); + memcpy(sbo->data + offset, data, size); + + brw_aub_gtt_data( iws->aubfile, + sbo->offset + offset, + sbo->data + offset, + size, + aub_type, + aub_sub_type ); +} + +void xmesa_commands_aub(struct pipe_winsys *winsys, + unsigned *cmds, + unsigned nr_dwords) +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + brw_aub_gtt_cmds( iws->aubfile, + 0, /* ?? */ + cmds, + nr_dwords * sizeof(int) ); +} + + +static struct aub_pipe_winsys *global_winsys = NULL; + +void xmesa_display_aub( /* struct pipe_winsys *winsys, */ + struct pipe_surface *surface ) +{ +// struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + brw_aub_dump_bmp( global_winsys->aubfile, + surface, + aub_bo(surface->buffer)->offset ); +} + + + +static int aub_buffer_get_subdata(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + void *data) +{ + struct aub_buffer *sbo = aub_bo(buf); + assert(sbo->size > offset + size); + memcpy(data, sbo->data + offset, size); + return 0; +} + +/* Pipe has no concept of pools. We choose the tex/region pool + * for all buffers. + */ +static struct pipe_buffer_handle * +aub_buffer_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned flags, + unsigned hint) +{ + return pipe_bo(CALLOC_STRUCT(aub_buffer)); +} + + +static struct pipe_buffer_handle * +aub_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) +{ + struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer); + sbo->size = bytes; + sbo->userbuffer = 1; + sbo->data = ptr; + return pipe_bo(sbo); +} + + +/* The state tracker (should!) keep track of whether the fake + * frontbuffer has been touched by any rendering since the last time + * we copied its contents to the real frontbuffer. Our task is easy: + */ +static void +aub_flush_frontbuffer( struct pipe_winsys *winsys, + struct pipe_surface *surf, + void *context_private) +{ + xmesa_display_aub( surf ); +} + +static struct pipe_surface * +aub_i915_surface_alloc(struct pipe_winsys *winsys) +{ + struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); + if (surf) { + surf->refcount = 1; + surf->winsys = winsys; + } + return surf; +} + + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + +static int +aub_i915_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags) +{ + const unsigned alignment = 64; + int ret; + + surf->width = width; + surf->height = height; + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = round_up(width, alignment / surf->cpp); + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, 0, 0); + if(!surf->buffer) + return -1; + + ret = winsys->buffer_data(winsys, + surf->buffer, + surf->pitch * surf->cpp * height, + NULL, + 0); + if(ret) { + winsys->buffer_reference(winsys, &surf->buffer, NULL); + return ret; + } + + return 0; +} + +static void +aub_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + surf->refcount--; + if (surf->refcount == 0) { + if (surf->buffer) + winsys->buffer_reference(winsys, &surf->buffer, NULL); + free(surf); + } + *s = NULL; +} + + + +static void +aub_printf( struct pipe_winsys *winsys, const char *fmtString, ... ) +{ + va_list args; + va_start( args, fmtString ); + vfprintf(stderr, fmtString, args); + va_end( args ); +} + +static const char * +aub_get_name( struct pipe_winsys *winsys ) +{ + return "Aub/xlib"; +} + +struct pipe_winsys * +xmesa_create_pipe_winsys_aub( void ) +{ + struct aub_pipe_winsys *iws = CALLOC_STRUCT( aub_pipe_winsys ); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + * + * Pipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + iws->winsys.buffer_create = aub_buffer_create; + iws->winsys.user_buffer_create = aub_user_buffer_create; + iws->winsys.buffer_map = aub_buffer_map; + iws->winsys.buffer_unmap = aub_buffer_unmap; + iws->winsys.buffer_reference = aub_buffer_reference; + iws->winsys.buffer_data = aub_buffer_data; + iws->winsys.buffer_subdata = aub_buffer_subdata; + iws->winsys.buffer_get_subdata = aub_buffer_get_subdata; + iws->winsys.flush_frontbuffer = aub_flush_frontbuffer; + iws->winsys.printf = aub_printf; + iws->winsys.get_name = aub_get_name; + + iws->winsys.surface_alloc = aub_i915_surface_alloc; + iws->winsys.surface_alloc_storage = aub_i915_surface_alloc_storage; + iws->winsys.surface_release = aub_i915_surface_release; + + iws->aubfile = brw_aubfile_create(); + iws->size = AUB_BUF_SIZE; + iws->pool = malloc(AUB_BUF_SIZE); + + /* HACK: static copy of this pointer: + */ + assert(global_winsys == NULL); + global_winsys = iws; + + return &iws->winsys; +} + + +void +xmesa_destroy_pipe_winsys_aub( struct pipe_winsys *winsys ) + +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + brw_aub_destroy(iws->aubfile); + free(iws->pool); + free(iws); +} + + + + + + + +#define IWS_BATCHBUFFER_SIZE 1024 + +struct aub_brw_winsys { + struct brw_winsys winsys; /**< batch buffer funcs */ + struct aub_context *aub; + + struct pipe_winsys *pipe_winsys; + + unsigned data[IWS_BATCHBUFFER_SIZE]; + unsigned nr; + unsigned size; +}; + + +/* Turn a i965simple winsys into an aub/i965simple winsys: + */ +static inline struct aub_brw_winsys * +aub_brw_winsys( struct brw_winsys *sws ) +{ + return (struct aub_brw_winsys *)sws; +} + + +/* Simple batchbuffer interface: + */ + +static unsigned *aub_i965_batch_start( struct brw_winsys *sws, + unsigned dwords, + unsigned relocs ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + + if (iws->size < iws->nr + dwords) + return NULL; + + return (void *)1; /* not a valid pointer! */ +} + +static void aub_i965_batch_dword( struct brw_winsys *sws, + unsigned dword ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + + iws->data[iws->nr++] = dword; +} + +static void aub_i965_batch_reloc( struct brw_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned access_flags, + unsigned delta ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + + iws->data[iws->nr++] = aub_bo(buf)->offset + delta; +} + + + +static void aub_i965_batch_flush( struct brw_winsys *sws, + struct pipe_fence_handle **fence ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + assert(iws->nr <= iws->size); + + if (iws->nr) + xmesa_commands_aub( iws->pipe_winsys, + iws->data, + iws->nr ); + iws->nr = 0; +} + + + +static void aub_i965_buffer_subdata_typed(struct brw_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data, + unsigned data_type) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(winsys); + unsigned aub_type = DW_GENERAL_STATE; + unsigned aub_sub_type; + + switch (data_type) { + case BRW_CC_VP: + aub_sub_type = DWGS_COLOR_CALC_VIEWPORT_STATE; + break; + case BRW_CC_UNIT: + aub_sub_type = DWGS_COLOR_CALC_STATE; + break; + case BRW_WM_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_SAMPLER_DEFAULT_COLOR: + aub_sub_type = DWGS_SAMPLER_DEFAULT_COLOR; + break; + case BRW_SAMPLER: + aub_sub_type = DWGS_SAMPLER_STATE; + break; + case BRW_WM_UNIT: + aub_sub_type = DWGS_WINDOWER_IZ_STATE; + break; + case BRW_SF_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_SF_VP: + aub_sub_type = DWGS_STRIPS_FANS_VIEWPORT_STATE; + break; + case BRW_SF_UNIT: + aub_sub_type = DWGS_STRIPS_FANS_STATE; + break; + case BRW_VS_UNIT: + aub_sub_type = DWGS_VERTEX_SHADER_STATE; + break; + case BRW_VS_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_GS_UNIT: + aub_sub_type = DWGS_GEOMETRY_SHADER_STATE; + break; + case BRW_GS_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_CLIP_VP: + aub_sub_type = DWGS_CLIPPER_VIEWPORT_STATE; + break; + case BRW_CLIP_UNIT: + aub_sub_type = DWGS_CLIPPER_STATE; + break; + case BRW_CLIP_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_SS_SURFACE: + aub_type = DW_SURFACE_STATE; + aub_sub_type = DWSS_SURFACE_STATE; + break; + case BRW_SS_SURF_BIND: + aub_type = DW_SURFACE_STATE; + aub_sub_type = DWSS_BINDING_TABLE_STATE; + break; + } + + xmesa_buffer_subdata_aub( iws->pipe_winsys, + buf, + offset, + size, + data, + aub_type, + aub_sub_type ); +} + +/** + * Create i965 hardware rendering context. + */ +struct pipe_context * +xmesa_create_i965simple( struct pipe_winsys *winsys ) +{ + struct aub_brw_winsys *iws = CALLOC_STRUCT( aub_brw_winsys ); + + /* Fill in this struct with callbacks that i965simple will need to + * communicate with the window system, buffer manager, etc. + */ + iws->winsys.batch_start = aub_i965_batch_start; + iws->winsys.batch_dword = aub_i965_batch_dword; + iws->winsys.batch_reloc = aub_i965_batch_reloc; + iws->winsys.batch_flush = aub_i965_batch_flush; + iws->winsys.buffer_subdata_typed = aub_i965_buffer_subdata_typed; + + iws->pipe_winsys = winsys; + + iws->size = IWS_BATCHBUFFER_SIZE; + + /* Create the i965simple context: + */ + return brw_create( winsys, + &iws->winsys, + 0 ); +} diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.h b/src/mesa/pipe/xlib/xm_winsys_aub.h new file mode 100644 index 0000000000..c0fe449107 --- /dev/null +++ b/src/mesa/pipe/xlib/xm_winsys_aub.h @@ -0,0 +1,67 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef AUB_WINSYS_H +#define AUB_WINSYS_H + +struct pipe_context; +struct pipe_winsys; +struct pipe_buffer_handle; +struct pipe_surface; + +struct pipe_winsys * +xmesa_create_pipe_winsys_aub( void ); + +void +xmesa_destroy_pipe_winsys_aub( struct pipe_winsys *winsys ); + + + +struct pipe_context * +xmesa_create_i965simple( struct pipe_winsys *winsys ); + + + +void xmesa_buffer_subdata_aub(struct pipe_winsys *winsys, + struct pipe_buffer_handle *buf, + unsigned long offset, + unsigned long size, + const void *data, + unsigned aub_type, + unsigned aub_sub_type); + +void xmesa_commands_aub(struct pipe_winsys *winsys, + unsigned *cmds, + unsigned nr_dwords); + + +void xmesa_display_aub( /* struct pipe_winsys *winsys, */ + struct pipe_surface *surface ); + +struct pipe_winsys *xmesa_get_pipe_winsys_aub(void); + +#endif -- cgit v1.2.3 From 89afc929f4e18165f1043c36844150e78f300cd5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 11 Dec 2007 12:25:42 +0000 Subject: gallium: remove dead pbo zcopy code --- src/mesa/state_tracker/st_cb_texture.c | 31 ------------------------------- 1 file changed, 31 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 14f8d38def..2c93a2f3dd 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -487,18 +487,6 @@ try_pbo_upload(GLcontext *ctx, -static GLboolean -try_pbo_zcopy(GLcontext *ctx, - struct st_texture_image *stImage, - const struct gl_pixelstore_attrib *unpack, - GLint internalFormat, - GLint width, GLint height, - GLenum format, GLenum type, const void *pixels) -{ - return GL_FALSE; -} - - @@ -618,24 +606,6 @@ st_TexImage(GLcontext * ctx, DBG("trying pbo upload\n"); - /* Attempt to texture directly from PBO data (zero copy upload). - * - * Currently disable as it can lead to worse as well as better - * performance (in particular when pipe_region_cow() is - * required). - */ - if (stObj->pt == stImage->pt && - stObj->pt->first_level == level && - stObj->pt->last_level == level) { - - if (try_pbo_zcopy(intel, stImage, unpack, - internalFormat, - width, height, format, type, pixels)) { - - DBG("pbo zcopy upload succeeded\n"); - return; - } - } /* Otherwise, attempt to use the blitter for PBO image uploads. @@ -652,7 +622,6 @@ st_TexImage(GLcontext * ctx, #else (void) try_pbo_upload; (void) check_pbo_format; - (void) try_pbo_zcopy; #endif -- cgit v1.2.3 From 48731280d08bef51c406703e82986643e17b4757 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 11 Dec 2007 13:00:12 +0000 Subject: gallium: Remove feedback interfaces from pipe driver. Something similar will return when geometry shaders are added, but for now this interface is not required. --- src/mesa/pipe/draw/draw_context.c | 21 --- src/mesa/pipe/draw/draw_context.h | 6 - src/mesa/pipe/draw/draw_feedback.c | 253 ----------------------------- src/mesa/pipe/draw/draw_private.h | 9 - src/mesa/pipe/draw/draw_validate.c | 5 - src/mesa/pipe/i915simple/i915_context.c | 22 --- src/mesa/pipe/i915simple/i915_context.h | 4 - src/mesa/pipe/i915simple/i915_state.c | 30 ---- src/mesa/pipe/p_context.h | 9 - src/mesa/pipe/p_state.h | 22 --- src/mesa/pipe/softpipe/Makefile | 1 - src/mesa/pipe/softpipe/sp_context.c | 2 - src/mesa/pipe/softpipe/sp_context.h | 4 - src/mesa/pipe/softpipe/sp_draw_arrays.c | 22 --- src/mesa/pipe/softpipe/sp_state.h | 8 - src/mesa/pipe/softpipe/sp_state_feedback.c | 72 -------- src/mesa/sources | 1 - src/mesa/state_tracker/st_cb_rasterpos.c | 44 +++-- src/mesa/state_tracker/st_context.h | 1 - 19 files changed, 21 insertions(+), 515 deletions(-) delete mode 100644 src/mesa/pipe/draw/draw_feedback.c delete mode 100644 src/mesa/pipe/softpipe/sp_state_feedback.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 33727e6547..179f7ed0d6 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -55,7 +55,6 @@ struct draw_context *draw_create( void ) draw->pipeline.clip = draw_clip_stage( draw ); draw->pipeline.flatshade = draw_flatshade_stage( draw ); draw->pipeline.cull = draw_cull_stage( draw ); - draw->pipeline.feedback = draw_feedback_stage( draw ); draw->pipeline.validate = draw_validate_stage( draw ); draw->pipeline.first = draw->pipeline.validate; @@ -100,7 +99,6 @@ void draw_destroy( struct draw_context *draw ) draw->pipeline.clip->destroy( draw->pipeline.clip ); draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); draw->pipeline.cull->destroy( draw->pipeline.cull ); - draw->pipeline.feedback->destroy( draw->pipeline.feedback ); draw->pipeline.validate->destroy( draw->pipeline.validate ); if (draw->pipeline.rasterize) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); @@ -117,13 +115,6 @@ void draw_flush( struct draw_context *draw ) } -void draw_set_feedback_state( struct draw_context *draw, - const struct pipe_feedback_state *feedback ) -{ - draw_flush( draw ); - draw->feedback = *feedback; -} - /** * Register new primitive rasterization/rendering state. @@ -223,18 +214,6 @@ draw_set_mapped_constant_buffer(struct draw_context *draw, } -void -draw_set_mapped_feedback_buffer(struct draw_context *draw, uint index, - void *buffer, uint size) -{ - draw_flush( draw ); - - assert(index < PIPE_MAX_FEEDBACK_ATTRIBS); - draw->user.feedback_buffer[index] = buffer; - draw->user.feedback_buffer_size[index] = size; /* in bytes */ -} - - diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 8e2232244c..6dc6e4ce82 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -82,9 +82,6 @@ void draw_set_viewport_state( struct draw_context *draw, void draw_set_clip_state( struct draw_context *pipe, const struct pipe_clip_state *clip ); -void draw_set_feedback_state( struct draw_context *draw, - const struct pipe_feedback_state * ); - void draw_set_rasterizer_state( struct draw_context *draw, const struct pipe_rasterizer_state *raster ); @@ -120,9 +117,6 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw, void draw_set_mapped_constant_buffer(struct draw_context *draw, const void *buffer); -void -draw_set_mapped_feedback_buffer(struct draw_context *draw, uint index, - void *buffer, uint size); /*********************************************************************** diff --git a/src/mesa/pipe/draw/draw_feedback.c b/src/mesa/pipe/draw/draw_feedback.c deleted file mode 100644 index aea6a8184c..0000000000 --- a/src/mesa/pipe/draw/draw_feedback.c +++ /dev/null @@ -1,253 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Primitive/vertex feedback (and/or discard) stage. - * Used to implement transformation feedback/streaming and other things - * which require a post-transformed vertex position (such as rasterpos, - * selection and feedback modes). - * - * Authors: - * Brian Paul - */ - - -#include "pipe/p_util.h" -#include "draw_private.h" - - -struct feedback_stage { - struct draw_stage stage; /**< base class */ - uint num_prim_generated; /**< number of primitives received */ - uint num_prim_emitted; /**< number of primitives fed back */ - uint num_vert_emitted; /**< number of vertices fed back */ - uint max_vert_emit; /**< max number of verts we can emit */ - float *dest[PIPE_MAX_FEEDBACK_ATTRIBS]; /**< dests for vertex attribs */ -}; - - - -/** - * Check if there's space to store 'numVerts' in the feedback buffer(s). - */ -static boolean -check_space(const struct draw_stage *stage, uint numVerts) -{ - const struct feedback_stage *fs = (struct feedback_stage *) stage; - return fs->num_vert_emitted + numVerts <= fs->max_vert_emit; -} - - -/** - * Record the given vertex's attributes into the feedback buffer(s). - */ -static void -feedback_vertex(struct draw_stage *stage, const struct vertex_header *vertex) -{ - struct feedback_stage *fs = (struct feedback_stage *) stage; - const struct pipe_feedback_state *feedback = &stage->draw->feedback; - const uint select = feedback->interleaved ? 0 : 1; - uint i; - - /* - * Note: 'select' is either 0 or 1. By multiplying 'i' by 'select' - * we can either address output buffer 0 (for interleaving) or - * output buffer i (for non-interleaved). - */ - for (i = 0; i < feedback->num_attribs; i++) { - const uint slot = feedback->attrib[i]; - const float *src = slot ? vertex->data[slot] : vertex->clip; - const uint size = feedback->size[i]; - float *dest = fs->dest[i * select]; - - switch (size) { - case 4: - dest[3] = src[3]; - /* fall-through */ - case 3: - dest[2] = src[2]; - /* fall-through */ - case 2: - dest[1] = src[1]; - /* fall-through */ - case 1: - dest[0] = src[0]; - /* fall-through */ - default: - ; - } - fs->dest[i * select] += size; - } - - fs->num_vert_emitted++; -} - - -static void feedback_begin( struct draw_stage *stage ) -{ - struct feedback_stage *fs = (struct feedback_stage *) stage; - const struct pipe_feedback_state *feedback = &stage->draw->feedback; - - fs->num_prim_generated = 0; - fs->num_prim_emitted = 0; - fs->num_vert_emitted = 0; - - assert(feedback->enabled); - - /* Compute max_vert_emit, the max number of vertices we can emit. - * And, setup dest[] pointers. - */ - if (stage->draw->feedback.interleaved) { - uint i, vertex_size = 0; - /* compute size of each interleaved vertex, in floats */ - for (i = 0; i < feedback->num_attribs; i++) { - vertex_size += feedback->size[i]; - } - /* compute max number of vertices we can feedback */ - fs->max_vert_emit = stage->draw->user.feedback_buffer_size[0] - / sizeof(float) / vertex_size; - - fs->dest[0] = (float *) stage->draw->user.feedback_buffer[0]; - } - else { - uint i; - uint max = ~0; - for (i = 0; i < feedback->num_attribs; i++) { - uint n = stage->draw->user.feedback_buffer_size[i] - / sizeof(float) / feedback->size[i]; - if (n < max) - max = n; - fs->dest[i] = (float *) stage->draw->user.feedback_buffer[i]; - } - fs->max_vert_emit = max; - } - - if (!feedback->discard) - stage->next->begin( stage->next ); -} - - -static void feedback_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct feedback_stage *fs = (struct feedback_stage *) stage; - - fs->num_prim_generated++; - - if (stage->draw->feedback.enabled && check_space(stage, 3)) { - feedback_vertex(stage, header->v[0]); - feedback_vertex(stage, header->v[1]); - feedback_vertex(stage, header->v[2]); - fs->num_prim_emitted++; - } - - if (!stage->draw->feedback.discard) - stage->next->tri( stage->next, header ); -} - - -static void feedback_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct feedback_stage *fs = (struct feedback_stage *) stage; - - fs->num_prim_generated++; - - if (stage->draw->feedback.enabled && check_space(stage, 2)) { - feedback_vertex(stage, header->v[0]); - feedback_vertex(stage, header->v[1]); - fs->num_prim_emitted++; - } - - if (!stage->draw->feedback.discard) - stage->next->line( stage->next, header ); -} - - -static void feedback_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct feedback_stage *fs = (struct feedback_stage *) stage; - - fs->num_prim_generated++; - - if (stage->draw->feedback.enabled && check_space(stage, 1)) { - feedback_vertex(stage, header->v[0]); - fs->num_prim_emitted++; - } - - if (!stage->draw->feedback.discard) - stage->next->point( stage->next, header ); -} - - -static void feedback_end( struct draw_stage *stage ) -{ - - /* XXX Unmap the vertex feedback buffers so we can write to them */ - - - if (!stage->draw->feedback.discard) - stage->next->end( stage->next ); -} - - - -static void feedback_reset_stipple_counter( struct draw_stage *stage ) -{ - if (!stage->draw->feedback.discard) - stage->next->reset_stipple_counter( stage->next ); -} - - -static void feedback_destroy( struct draw_stage *stage ) -{ - FREE( stage ); -} - - -/** - * Create feedback drawing stage. - */ -struct draw_stage *draw_feedback_stage( struct draw_context *draw ) -{ - struct feedback_stage *feedback = CALLOC_STRUCT(feedback_stage); - - feedback->stage.draw = draw; - feedback->stage.next = NULL; - feedback->stage.begin = feedback_begin; - feedback->stage.point = feedback_point; - feedback->stage.line = feedback_line; - feedback->stage.tri = feedback_tri; - feedback->stage.end = feedback_end; - feedback->stage.reset_stipple_counter = feedback_reset_stipple_counter; - feedback->stage.destroy = feedback_destroy; - - return &feedback->stage; -} - - diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index ca5ca7b3c9..5f89ac121e 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -159,7 +159,6 @@ struct draw_context struct draw_stage *validate; /* stages (in logical order) */ - struct draw_stage *feedback; struct draw_stage *flatshade; struct draw_stage *clip; struct draw_stage *cull; @@ -172,13 +171,10 @@ struct draw_context /* pipe state that we need: */ const struct pipe_rasterizer_state *rasterizer; - struct pipe_feedback_state feedback; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; const struct draw_vertex_shader *vertex_shader; - struct pipe_vertex_buffer feedback_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element feedback_element[PIPE_ATTRIB_MAX]; /* user-space vertex data, buffers */ struct { @@ -192,10 +188,6 @@ struct draw_context /** constant buffer (for vertex shader) */ const void *constants; - - /** The vertex feedback buffer */ - void *feedback_buffer[PIPE_MAX_FEEDBACK_ATTRIBS]; - uint feedback_buffer_size[PIPE_MAX_FEEDBACK_ATTRIBS]; /* in bytes */ } user; /* Clip derived state: @@ -257,7 +249,6 @@ struct draw_context -extern struct draw_stage *draw_feedback_stage( struct draw_context *context ); extern struct draw_stage *draw_unfilled_stage( struct draw_context *context ); extern struct draw_stage *draw_twoside_stage( struct draw_context *context ); extern struct draw_stage *draw_offset_stage( struct draw_context *context ); diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index 8ce4a926e2..58cf340281 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -100,11 +100,6 @@ static void validate_begin( struct draw_stage *stage ) next = draw->pipeline.flatshade; } - if (draw->feedback.enabled || draw->feedback.discard) { - draw->pipeline.feedback->next = next; - next = draw->pipeline.feedback; - } - draw->pipeline.first = next; draw->pipeline.first->begin( draw->pipeline.first ); } diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index f505ff6ae6..d2bbeea16a 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -223,18 +223,6 @@ i915_draw_elements( struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } - /* Map feedback buffers if enabled */ - if (i915->feedback.enabled) { - const uint n = i915->feedback.interleaved ? 1 : i915->feedback.num_attribs; - for (i = 0; i < n; i++) { - void *ptr = pipe->winsys->buffer_map(pipe->winsys, - i915->feedback_buffer[i].buffer, - PIPE_BUFFER_FLAG_WRITE); - draw_set_mapped_feedback_buffer(draw, i, ptr, - i915->feedback_buffer[i].size); - } - } - draw_set_mapped_constant_buffer(draw, i915->current.constants[PIPE_SHADER_VERTEX]); @@ -256,16 +244,6 @@ i915_draw_elements( struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } - /* Unmap feedback buffers if enabled */ - if (i915->feedback.enabled) { - const uint n = i915->feedback.interleaved ? 1 : i915->feedback.num_attribs; - for (i = 0; i < n; i++) { - pipe->winsys->buffer_unmap(pipe->winsys, - i915->feedback_buffer[i].buffer); - draw_set_mapped_feedback_buffer(draw, i, NULL, 0); - } - } - return TRUE; } diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index dbf0c885cc..b9b67c4fcf 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -197,7 +197,6 @@ struct i915_context struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; - struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; @@ -206,9 +205,6 @@ struct i915_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - /** Feedback buffers */ - struct pipe_feedback_buffer feedback_buffer[PIPE_MAX_FEEDBACK_ATTRIBS]; - unsigned dirty; unsigned *batch_start; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 038fd623ea..b9f257a007 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -547,34 +547,6 @@ static void i915_set_framebuffer_state(struct pipe_context *pipe, } -static void -i915_set_feedback_state(struct pipe_context *pipe, - const struct pipe_feedback_state *feedback) -{ - struct i915_context *i915 = i915_context(pipe); - i915->feedback = *feedback; - draw_set_feedback_state(i915->draw, feedback); -} - - -static void -i915_set_feedback_buffer(struct pipe_context *pipe, - unsigned index, - const struct pipe_feedback_buffer *feedback) -{ - struct i915_context *i915 = i915_context(pipe); - - assert(index < PIPE_MAX_FEEDBACK_ATTRIBS); - - /* Need to be careful with referencing */ - pipe->winsys->buffer_reference(pipe->winsys, - &i915->feedback_buffer[index].buffer, - feedback->buffer); - i915->feedback_buffer[index].size = feedback->size; - i915->feedback_buffer[index].start_offset = feedback->start_offset; -} - - static void i915_set_clear_color_state(struct pipe_context *pipe, const struct pipe_clear_color_state *clear) @@ -751,8 +723,6 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_clear_color_state = i915_set_clear_color_state; i915->pipe.set_constant_buffer = i915_set_constant_buffer; i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; - i915->pipe.set_feedback_state = i915_set_feedback_state; - i915->pipe.set_feedback_buffer = i915_set_feedback_buffer; i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; i915->pipe.set_sampler_units = i915_set_sampler_units; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 00379fbacf..7da4992841 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -139,9 +139,6 @@ struct pipe_context { uint shader, uint index, const struct pipe_constant_buffer *buf ); - void (*set_feedback_state)( struct pipe_context *, - const struct pipe_feedback_state *); - void (*set_framebuffer_state)( struct pipe_context *, const struct pipe_framebuffer_state * ); @@ -172,12 +169,6 @@ struct pipe_context { unsigned index, const struct pipe_vertex_element * ); - /* - * Vertex feedback - */ - void (*set_feedback_buffer)(struct pipe_context *, - unsigned index, - const struct pipe_feedback_buffer *); /** Get a surface which is a "view" into a texture */ struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 6db9bbc953..50344bea78 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -103,19 +103,6 @@ struct pipe_rasterizer_state }; -/** - * Post-transform vertex feeback - */ -struct pipe_feedback_state { - uint enabled:1; /**< enable feedback? */ - uint discard:1; /**< discard primitives? */ - uint interleaved:1; /**< interleaved output? */ - uint num_attribs; - uint attrib[PIPE_MAX_FEEDBACK_ATTRIBS]; - uint size[PIPE_MAX_FEEDBACK_ATTRIBS]; -}; - - struct pipe_poly_stipple { unsigned stipple[32]; }; @@ -336,15 +323,6 @@ struct pipe_vertex_element }; -/** - * Vertex feedback buffer - */ -struct pipe_feedback_buffer { - struct pipe_buffer_handle *buffer; - unsigned size; - unsigned start_offset; -}; - /** * Hardware queries (occlusion, transform feedback, timing, etc) diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 647cc05373..5e6886a37e 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -27,7 +27,6 @@ DRIVER_SOURCES = \ sp_state_blend.c \ sp_state_clip.c \ sp_state_derived.c \ - sp_state_feedback.c \ sp_state_fs.c \ sp_state_sampler.c \ sp_state_rasterizer.c \ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 8b8e04c2f9..2eab3aaabb 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -306,7 +306,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_clip_state = softpipe_set_clip_state; softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; - softpipe->pipe.set_feedback_state = softpipe_set_feedback_state; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; softpipe->pipe.set_sampler_units = softpipe_set_sampler_units; @@ -316,7 +315,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; - softpipe->pipe.set_feedback_buffer = softpipe_set_feedback_buffer; softpipe->pipe.draw_arrays = softpipe_draw_arrays; softpipe->pipe.draw_elements = softpipe_draw_elements; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index b97cdc52c6..1c391dcd4d 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -85,7 +85,6 @@ struct softpipe_context { struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; - struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; @@ -117,9 +116,6 @@ struct softpipe_context { boolean need_w; /**< produce quad/fragment W values? */ int psize_slot; - /** Feedback buffers */ - struct pipe_feedback_buffer feedback_buffer[PIPE_MAX_FEEDBACK_ATTRIBS]; - #if 0 /* Stipple derived state: */ diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 93eb68405d..b7626f8a5f 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -138,18 +138,6 @@ softpipe_draw_elements(struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } - /* Map feedback buffers if enabled */ - if (sp->feedback.enabled) { - const uint n = sp->feedback.interleaved ? 1 : sp->feedback.num_attribs; - for (i = 0; i < n; i++) { - void *ptr = pipe->winsys->buffer_map(pipe->winsys, - sp->feedback_buffer[i].buffer, - PIPE_BUFFER_FLAG_WRITE); - draw_set_mapped_feedback_buffer(draw, i, ptr, - sp->feedback_buffer[i].size); - } - } - /* draw! */ draw_arrays(draw, mode, start, count); @@ -171,16 +159,6 @@ softpipe_draw_elements(struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } - /* Unmap feedback buffers if enabled */ - if (sp->feedback.enabled) { - const uint n = sp->feedback.interleaved ? 1 : sp->feedback.num_attribs; - for (i = 0; i < n; i++) { - pipe->winsys->buffer_unmap(pipe->winsys, - sp->feedback_buffer[i].buffer); - draw_set_mapped_feedback_buffer(draw, i, NULL, 0); - } - } - /* Note: leave drawing surfaces mapped */ softpipe_unmap_constant_buffers(sp); diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index ea9d2e62be..f434567da5 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -102,9 +102,6 @@ void softpipe_set_constant_buffer(struct pipe_context *, uint shader, uint index, const struct pipe_constant_buffer *buf); -void softpipe_set_feedback_state( struct pipe_context *, - const struct pipe_feedback_state * ); - void *softpipe_create_fs_state(struct pipe_context *, const struct pipe_shader_state *); void softpipe_bind_fs_state(struct pipe_context *, void *); @@ -138,11 +135,6 @@ void softpipe_set_vertex_buffer(struct pipe_context *, unsigned index, const struct pipe_vertex_buffer *); -void softpipe_set_feedback_buffer(struct pipe_context *, - uint index, - const struct pipe_feedback_buffer *); - - void softpipe_update_derived( struct softpipe_context *softpipe ); diff --git a/src/mesa/pipe/softpipe/sp_state_feedback.c b/src/mesa/pipe/softpipe/sp_state_feedback.c deleted file mode 100644 index 02aaf34e75..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_feedback.c +++ /dev/null @@ -1,72 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Authors: - * Brian Paul - */ - - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_surface.h" - -#include "pipe/p_winsys.h" -#include "pipe/draw/draw_context.h" - - -void -softpipe_set_feedback_state(struct pipe_context *pipe, - const struct pipe_feedback_state *feedback) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->feedback = *feedback; /* struct copy */ - /* - softpipe->dirty |= SP_NEW_FEEDBACK; - */ - - draw_set_feedback_state(softpipe->draw, feedback); -} - - -void -softpipe_set_feedback_buffer(struct pipe_context *pipe, - unsigned index, - const struct pipe_feedback_buffer *feedback) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - assert(index < PIPE_MAX_FEEDBACK_ATTRIBS); - - /* Need to be careful with referencing */ - pipe->winsys->buffer_reference(pipe->winsys, - &softpipe->feedback_buffer[index].buffer, - feedback->buffer); - softpipe->feedback_buffer[index].size = feedback->size; - softpipe->feedback_buffer[index].start_offset = feedback->start_offset; -} diff --git a/src/mesa/sources b/src/mesa/sources index c34361eaae..4a5a97b47d 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -162,7 +162,6 @@ DRAW_SOURCES = \ pipe/draw/draw_context.c\ pipe/draw/draw_cull.c \ pipe/draw/draw_debug.c \ - pipe/draw/draw_feedback.c \ pipe/draw/draw_flatshade.c \ pipe/draw/draw_offset.c \ pipe/draw/draw_prim.c \ diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 5279cb1cd4..852cff6490 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -173,6 +173,7 @@ static struct rastpos_stage * new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw) { struct rastpos_stage *rs = CALLOC_STRUCT(rastpos_stage); + GLuint i; rs->stage.draw = draw; rs->stage.next = NULL; @@ -184,6 +185,26 @@ new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw) rs->stage.reset_stipple_counter = rastpos_reset_stipple_counter; rs->ctx = ctx; + for (i = 0; i < VERT_ATTRIB_MAX; i++) { + rs->array[i].Size = 4; + rs->array[i].Type = GL_FLOAT; + rs->array[i].Stride = 0; + rs->array[i].StrideB = 0; + rs->array[i].Ptr = (GLubyte *) ctx->Current.Attrib[i]; + rs->array[i].Enabled = GL_TRUE; + rs->array[i].Normalized = GL_TRUE; + rs->array[i].BufferObj = NULL; + rs->arrays[i] = &rs->array[i]; + } + + rs->prim.mode = GL_POINTS; + rs->prim.indexed = 0; + rs->prim.begin = 1; + rs->prim.end = 1; + rs->prim.weak = 0; + rs->prim.start = 0; + rs->prim.count = 1; + return rs; } @@ -201,31 +222,8 @@ st_RasterPos(GLcontext *ctx, const GLfloat v[4]) } else { /* create rastpos draw stage */ - GLuint i; - rs = new_draw_rastpos_stage(ctx, draw); st->rastpos_stage = &rs->stage; - - /* one-time init */ - for (i = 0; i < VERT_ATTRIB_MAX; i++) { - rs->array[i].Size = 4; - rs->array[i].Type = GL_FLOAT; - rs->array[i].Stride = 0; - rs->array[i].StrideB = 0; - rs->array[i].Ptr = (GLubyte *) ctx->Current.Attrib[i]; - rs->array[i].Enabled = GL_TRUE; - rs->array[i].Normalized = GL_TRUE; - rs->array[i].BufferObj = NULL; - rs->arrays[i] = &rs->array[i]; - } - - rs->prim.mode = GL_POINTS; - rs->prim.indexed = 0; - rs->prim.begin = 1; - rs->prim.end = 1; - rs->prim.weak = 0; - rs->prim.start = 0; - rs->prim.count = 1; } /* plug our rastpos stage into the draw module */ diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index c31b76c63f..33aacdb6d1 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -106,7 +106,6 @@ struct st_context struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; - struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; -- cgit v1.2.3 From 94cadef31f9d4ee9fce1bfa66fabb0a403a6f049 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 11 Dec 2007 13:19:33 +0000 Subject: gallium: remove redundant clear_color state. --- src/mesa/pipe/failover/fo_context.h | 1 - src/mesa/pipe/failover/fo_state.c | 11 ----- src/mesa/pipe/failover/fo_state_emit.c | 3 -- src/mesa/pipe/i915simple/i915_context.h | 1 - src/mesa/pipe/i915simple/i915_state.c | 11 ----- src/mesa/pipe/i965simple/brw_context.h | 1 - src/mesa/pipe/i965simple/brw_state.c | 9 ---- src/mesa/pipe/p_context.h | 3 -- src/mesa/pipe/p_state.h | 5 --- src/mesa/pipe/softpipe/sp_context.c | 1 - src/mesa/pipe/softpipe/sp_context.h | 1 - src/mesa/pipe/softpipe/sp_state.h | 3 -- src/mesa/pipe/softpipe/sp_state_surface.c | 8 ---- src/mesa/sources | 1 - src/mesa/state_tracker/st_atom.c | 1 - src/mesa/state_tracker/st_atom.h | 1 - src/mesa/state_tracker/st_atom_clear_color.c | 63 ---------------------------- src/mesa/state_tracker/st_context.h | 1 - 18 files changed, 125 deletions(-) delete mode 100644 src/mesa/state_tracker/st_atom_clear_color.c (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index 7cf18c9ec1..2423eb4556 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -79,7 +79,6 @@ struct failover_context { const struct fo_state *vertex_shader; struct pipe_blend_color blend_color; - struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index fd6137ba66..689d2fa3c7 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -146,16 +146,6 @@ failover_set_clip_state( struct pipe_context *pipe, failover->hw->set_clip_state( failover->hw, clip ); } -static void -failover_set_clear_color_state( struct pipe_context *pipe, - const struct pipe_clear_color_state *clear_color ) -{ - struct failover_context *failover = failover_context(pipe); - - failover->clear_color = *clear_color; - failover->dirty |= FO_NEW_CLEAR_COLOR; - failover->hw->set_clear_color_state( failover->hw, clear_color ); -} static void * failover_create_depth_stencil_state(struct pipe_context *pipe, @@ -480,7 +470,6 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.set_blend_color = failover_set_blend_color; failover->pipe.set_clip_state = failover_set_clip_state; - failover->pipe.set_clear_color_state = failover_set_clear_color_state; failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; failover->pipe.set_sampler_units = failover_set_sampler_units; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index a3aff8abd2..612b0a6ca3 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -69,9 +69,6 @@ failover_state_emit( struct failover_context *failover ) if (failover->dirty & FO_NEW_CLIP) failover->sw->set_clip_state( failover->sw, &failover->clip ); - if (failover->dirty & FO_NEW_CLEAR_COLOR) - failover->sw->set_clear_color_state( failover->sw, &failover->clear_color ); - if (failover->dirty & FO_NEW_DEPTH_STENCIL) failover->sw->bind_depth_stencil_state( failover->sw, failover->depth_stencil->sw_state ); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index b9b67c4fcf..a239c8f72e 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -194,7 +194,6 @@ struct i915_context const struct pipe_shader_state *fs; struct pipe_blend_color blend_color; - struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; struct pipe_framebuffer_state framebuffer; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index b9f257a007..a8c6565a54 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -548,16 +548,6 @@ static void i915_set_framebuffer_state(struct pipe_context *pipe, -static void i915_set_clear_color_state(struct pipe_context *pipe, - const struct pipe_clear_color_state *clear) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->clear_color = *clear; /* struct copy */ -} - - - static void i915_set_clip_state( struct pipe_context *pipe, const struct pipe_clip_state *clip ) { @@ -720,7 +710,6 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_blend_color = i915_set_blend_color; i915->pipe.set_clip_state = i915_set_clip_state; - i915->pipe.set_clear_color_state = i915_set_clear_color_state; i915->pipe.set_constant_buffer = i915_set_constant_buffer; i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h index 00df46e704..53f66cd6a9 100644 --- a/src/mesa/pipe/i965simple/brw_context.h +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -475,7 +475,6 @@ struct brw_context struct pipe_scissor_state Scissor; struct pipe_viewport_state Viewport; struct pipe_framebuffer_state FrameBuffer; - struct pipe_clear_color_state ClearColor; const struct pipe_constant_buffer *Constants[2]; const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c index b50f23c1a2..7731c2e01f 100644 --- a/src/mesa/pipe/i965simple/brw_state.c +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -377,14 +377,6 @@ static void brw_set_framebuffer_state(struct pipe_context *pipe, -static void brw_set_clear_color_state(struct pipe_context *pipe, - const struct pipe_clear_color_state *clear) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.ClearColor = *clear; /* struct copy */ -} - /************************************************************************ * Rasterizer state */ @@ -449,7 +441,6 @@ brw_init_state_functions( struct brw_context *brw ) brw->pipe.set_blend_color = brw_set_blend_color; brw->pipe.set_clip_state = brw_set_clip_state; - brw->pipe.set_clear_color_state = brw_set_clear_color_state; brw->pipe.set_constant_buffer = brw_set_constant_buffer; brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 7da4992841..6b97844445 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -132,9 +132,6 @@ struct pipe_context { void (*set_clip_state)( struct pipe_context *, const struct pipe_clip_state * ); - void (*set_clear_color_state)( struct pipe_context *, - const struct pipe_clear_color_state * ); - void (*set_constant_buffer)( struct pipe_context *, uint shader, uint index, const struct pipe_constant_buffer *buf ); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 50344bea78..56d4f96ed7 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -200,11 +200,6 @@ struct pipe_blend_color { float color[4]; }; -struct pipe_clear_color_state -{ - float color[4]; -}; - struct pipe_framebuffer_state { /** multiple colorbuffers for multiple render targets */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 2eab3aaabb..43f23dc1e8 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -304,7 +304,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_blend_color = softpipe_set_blend_color; softpipe->pipe.set_clip_state = softpipe_set_clip_state; - softpipe->pipe.set_clear_color_state = softpipe_set_clear_color_state; softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 1c391dcd4d..45d15c720e 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -82,7 +82,6 @@ struct softpipe_context { const struct sp_vertex_shader_state *vs; struct pipe_blend_color blend_color; - struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; struct pipe_framebuffer_state framebuffer; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index f434567da5..80a1cba25a 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -92,9 +92,6 @@ void softpipe_set_framebuffer_state( struct pipe_context *, void softpipe_set_blend_color( struct pipe_context *pipe, const struct pipe_blend_color *blend_color ); -void softpipe_set_clear_color_state( struct pipe_context *, - const struct pipe_clear_color_state * ); - void softpipe_set_clip_state( struct pipe_context *, const struct pipe_clip_state * ); diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index 30bedc74bc..ee72aaf4c5 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -131,11 +131,3 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, -void -softpipe_set_clear_color_state(struct pipe_context *pipe, - const struct pipe_clear_color_state *clear) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->clear_color = *clear; /* struct copy */ -} diff --git a/src/mesa/sources b/src/mesa/sources index 4a5a97b47d..fb059498f9 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -200,7 +200,6 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ state_tracker/st_atom_alphatest.c \ state_tracker/st_atom_blend.c \ - state_tracker/st_atom_clear_color.c \ state_tracker/st_atom_clip.c \ state_tracker/st_atom_constbuf.c \ state_tracker/st_atom_depth.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 0797ea615e..bde81edd8c 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -45,7 +45,6 @@ static const struct st_tracked_state *atoms[] = { &st_update_framebuffer, - &st_update_clear_color, &st_update_depth_stencil, &st_update_clip, diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 3c0db0db09..0114f42ba5 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -46,7 +46,6 @@ void st_validate_state( struct st_context *st ); const struct st_tracked_state st_update_framebuffer; const struct st_tracked_state st_update_clip; -const struct st_tracked_state st_update_clear_color; const struct st_tracked_state st_update_depth_stencil; const struct st_tracked_state st_update_shader; const struct st_tracked_state st_update_rasterizer; diff --git a/src/mesa/state_tracker/st_atom_clear_color.c b/src/mesa/state_tracker/st_atom_clear_color.c deleted file mode 100644 index ce3431c5d3..0000000000 --- a/src/mesa/state_tracker/st_atom_clear_color.c +++ /dev/null @@ -1,63 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - -#include "st_context.h" -#include "st_atom.h" -#include "pipe/p_context.h" - - -static void -update_clear_color_state( struct st_context *st ) -{ - struct pipe_clear_color_state clear; - - clear.color[0] = st->ctx->Color.ClearColor[0]; - clear.color[1] = st->ctx->Color.ClearColor[1]; - clear.color[2] = st->ctx->Color.ClearColor[2]; - clear.color[3] = st->ctx->Color.ClearColor[3]; - - if (memcmp(&clear, &st->state.clear_color, sizeof(clear)) != 0) { - st->state.clear_color = clear; - st->pipe->set_clear_color_state( st->pipe, &clear ); - } -} - - -const struct st_tracked_state st_update_clear_color = { - .name = "st_update_clear_color", - .dirty = { - .mesa = _NEW_COLOR, - .st = 0, - }, - .update = update_clear_color_state -}; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 33aacdb6d1..4855961d09 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -103,7 +103,6 @@ struct st_context const struct cso_vertex_shader *vs; struct pipe_blend_color blend_color; - struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; struct pipe_framebuffer_state framebuffer; -- cgit v1.2.3 From c3af68dc5022715cc8f126b7df12f3f5248aefe7 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 11 Dec 2007 14:39:37 +0000 Subject: gallium: remove set_sampler_units interface The effect of this mapping can be acheived by the state tracker and setting up the pipe texture state pointers to incorporate its affects. --- src/mesa/pipe/cell/ppu/cell_context.c | 1 - src/mesa/pipe/cell/ppu/cell_context.h | 1 - src/mesa/pipe/cell/ppu/cell_state.h | 3 --- src/mesa/pipe/cell/ppu/cell_state_sampler.c | 20 +++++------------- src/mesa/pipe/failover/fo_context.h | 1 - src/mesa/pipe/failover/fo_state.c | 23 +++++---------------- src/mesa/pipe/failover/fo_state_emit.c | 2 +- src/mesa/pipe/i915simple/i915_context.h | 1 - src/mesa/pipe/i915simple/i915_state.c | 19 +++++------------ src/mesa/pipe/i965simple/brw_state.c | 10 ++------- src/mesa/pipe/llvm/gallivm.cpp | 8 +++----- src/mesa/pipe/llvm/gallivm.h | 3 +-- src/mesa/pipe/llvm/llvm_base_shader.cpp | 2 -- src/mesa/pipe/llvm/llvm_entry.c | 3 +-- src/mesa/pipe/p_context.h | 12 +++++------ src/mesa/pipe/softpipe/sp_context.c | 3 +-- src/mesa/pipe/softpipe/sp_context.h | 1 - src/mesa/pipe/softpipe/sp_quad_fs.c | 3 +-- src/mesa/pipe/softpipe/sp_state.h | 5 +---- src/mesa/pipe/softpipe/sp_state_sampler.c | 17 +++------------ src/mesa/pipe/tgsi/exec/tgsi_exec.c | 3 +-- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 1 - src/mesa/state_tracker/st_atom_sampler.c | 15 +------------- src/mesa/state_tracker/st_atom_texture.c | 32 +++++++++++++++++++---------- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- src/mesa/state_tracker/st_cb_texture.c | 4 ++-- src/mesa/state_tracker/st_cb_texture.h | 3 ++- src/mesa/state_tracker/st_context.h | 2 +- 28 files changed, 65 insertions(+), 137 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 68543ecf30..5534d82ccf 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -217,7 +217,6 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->pipe.set_framebuffer_state = cell_set_framebuffer_state; cell->pipe.set_polygon_stipple = cell_set_polygon_stipple; - cell->pipe.set_sampler_units = cell_set_sampler_units; cell->pipe.set_scissor_state = cell_set_scissor_state; cell->pipe.set_texture_state = cell_set_texture_state; cell->pipe.set_viewport_state = cell_set_viewport_state; diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index a64692081d..12073b93a0 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -58,7 +58,6 @@ struct cell_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - uint sampler_units[PIPE_MAX_SAMPLERS]; uint dirty; /** The primitive drawing context */ diff --git a/src/mesa/pipe/cell/ppu/cell_state.h b/src/mesa/pipe/cell/ppu/cell_state.h index 033767d29b..4bad45950b 100644 --- a/src/mesa/pipe/cell/ppu/cell_state.h +++ b/src/mesa/pipe/cell/ppu/cell_state.h @@ -93,9 +93,6 @@ void cell_set_constant_buffer(struct pipe_context *pipe, void cell_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); -void cell_set_sampler_units( struct pipe_context *, - uint numSamplers, const uint *units ); - void cell_set_scissor_state( struct pipe_context *, const struct pipe_scissor_state * ); diff --git a/src/mesa/pipe/cell/ppu/cell_state_sampler.c b/src/mesa/pipe/cell/ppu/cell_state_sampler.c index c2a180ed30..495567b16c 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_sampler.c +++ b/src/mesa/pipe/cell/ppu/cell_state_sampler.c @@ -69,32 +69,22 @@ cell_delete_sampler_state(struct pipe_context *pipe, void -cell_set_texture_state(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) +cell_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, + struct pipe_texture *texture) { struct cell_context *cell = cell_context(pipe); assert(unit < PIPE_MAX_SAMPLERS); #if 0 - cell->texture[unit] = cell_texture(texture); /* ptr, not struct */ - cell_tile_cache_set_texture(cell->tex_cache[unit], texture); + cell->texture[sampler] = cell_texture(texture); /* ptr, not struct */ + cell_tile_cache_set_texture(cell->tex_cache[sampler], texture); #endif cell->dirty |= CELL_NEW_TEXTURE; } -void -cell_set_sampler_units(struct pipe_context *pipe, - uint num_samplers, const uint *units ) -{ - struct cell_context *cell = cell_context(pipe); - uint i; - for (i = 0; i < num_samplers; i++) - cell->sampler_units[i] = units[i]; - cell->dirty |= CELL_NEW_SAMPLER; -} diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index 2423eb4556..f5eaa0b5fa 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -83,7 +83,6 @@ struct failover_context { struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - uint sampler_units[PIPE_MAX_SAMPLERS]; struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 689d2fa3c7..6b4f1517ac 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -284,18 +284,6 @@ failover_set_polygon_stipple( struct pipe_context *pipe, failover->hw->set_polygon_stipple( failover->hw, stipple ); } -static void -failover_set_sampler_units( struct pipe_context *pipe, - uint num_samplers, const uint *units ) -{ - struct failover_context *failover = failover_context(pipe); - uint i; - - for (i = 0; i < num_samplers; i++) - failover->sampler_units[i] = units[i]; - failover->dirty |= FO_NEW_SAMPLER; - failover->hw->set_sampler_units(failover->hw, num_samplers, units); -} static void * failover_create_rasterizer_state(struct pipe_context *pipe, @@ -390,16 +378,16 @@ failover_delete_sampler_state(struct pipe_context *pipe, void *sampler) static void -failover_set_texture_state(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) +failover_set_sampler_texture(struct pipe_context *pipe, + unsigned unit, + struct pipe_texture *texture) { struct failover_context *failover = failover_context(pipe); failover->texture[unit] = texture; failover->dirty |= FO_NEW_TEXTURE; failover->dirty_texture |= (1<hw->set_texture_state( failover->hw, unit, texture ); + failover->hw->set_sampler_texture( failover->hw, unit, texture ); } @@ -472,9 +460,8 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.set_clip_state = failover_set_clip_state; failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; - failover->pipe.set_sampler_units = failover_set_sampler_units; failover->pipe.set_scissor_state = failover_set_scissor_state; - failover->pipe.set_texture_state = failover_set_texture_state; + failover->pipe.set_sampler_texture = failover_set_sampler_texture; failover->pipe.set_viewport_state = failover_set_viewport_state; failover->pipe.set_vertex_buffer = failover_set_vertex_buffer; failover->pipe.set_vertex_element = failover_set_vertex_element; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index 612b0a6ca3..c99ecd4f8d 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -109,7 +109,7 @@ failover_state_emit( struct failover_context *failover ) if (failover->dirty & FO_NEW_TEXTURE) { for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { if (failover->dirty_texture & (1<sw->set_texture_state( failover->sw, i, + failover->sw->set_sampler_texture( failover->sw, i, failover->texture[i] ); } } diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index a239c8f72e..80df7f0fba 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -199,7 +199,6 @@ struct i915_context struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - uint sampler_units[PIPE_MAX_SAMPLERS]; struct i915_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index a8c6565a54..2a9a587a37 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -431,14 +431,6 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe, { } -static void i915_set_sampler_units(struct pipe_context *pipe, - uint numSamplers, const uint *units) -{ - struct i915_context *i915 = i915_context(pipe); - uint i; - for (i = 0; i < numSamplers; i++) - i915->sampler_units[i] = units[i]; -} static void * i915_create_fs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) @@ -523,13 +515,13 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, } -static void i915_set_texture_state(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) +static void i915_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, + struct pipe_texture *texture) { struct i915_context *i915 = i915_context(pipe); - i915->texture[unit] = (struct i915_texture*)texture; /* ptr, not struct */ + i915->texture[sampler] = (struct i915_texture*)texture; /* ptr, not struct */ i915->dirty |= I915_NEW_TEXTURE; } @@ -714,9 +706,8 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; - i915->pipe.set_sampler_units = i915_set_sampler_units; i915->pipe.set_scissor_state = i915_set_scissor_state; - i915->pipe.set_texture_state = i915_set_texture_state; + i915->pipe.set_sampler_texture = i915_set_sampler_texture; i915->pipe.set_viewport_state = i915_set_viewport_state; i915->pipe.set_vertex_buffer = i915_set_vertex_buffer; i915->pipe.set_vertex_element = i915_set_vertex_element; diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c index 7731c2e01f..ff4ae7999b 100644 --- a/src/mesa/pipe/i965simple/brw_state.c +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -110,11 +110,6 @@ static void brw_delete_sampler_state(struct pipe_context *pipe, } -static void brw_set_sampler_units(struct pipe_context *pipe, - uint numSamplers, const uint *units) -{ -} - /************************************************************************ * Depth stencil */ @@ -349,7 +344,7 @@ static void brw_set_constant_buffer(struct pipe_context *pipe, */ -static void brw_set_texture_state(struct pipe_context *pipe, +static void brw_set_sampler_texture(struct pipe_context *pipe, unsigned unit, struct pipe_texture *texture) { @@ -448,9 +443,8 @@ brw_init_state_functions( struct brw_context *brw ) // brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; - brw->pipe.set_sampler_units = brw_set_sampler_units; brw->pipe.set_scissor_state = brw_set_scissor_state; - brw->pipe.set_texture_state = brw_set_texture_state; + brw->pipe.set_sampler_texture = brw_set_sampler_texture; brw->pipe.set_viewport_state = brw_set_viewport_state; brw->pipe.set_vertex_buffer = brw_set_vertex_buffer; brw->pipe.set_vertex_element = brw_set_vertex_element; diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index bd8bfac208..a60440022a 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -927,23 +927,21 @@ typedef int (*fragment_shader_runner)(float x, float y, float (*inputs)[16][4], int num_attribs, float (*consts)[4], int num_consts, - struct tgsi_sampler *samplers, - unsigned *sampler_units); + struct tgsi_sampler *samplers); int gallivm_fragment_shader_exec(struct gallivm_prog *prog, float fx, float fy, float (*dests)[16][4], float (*inputs)[16][4], float (*consts)[4], - struct tgsi_sampler *samplers, - unsigned *sampler_units) + struct tgsi_sampler *samplers) { fragment_shader_runner runner = reinterpret_cast(prog->function); assert(runner); runner(fx, fy, dests, inputs, prog->num_interp, consts, prog->num_consts, - samplers, sampler_units); + samplers); return 0; } diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h index 6a05a55db4..fd9a11e5b6 100644 --- a/src/mesa/pipe/llvm/gallivm.h +++ b/src/mesa/pipe/llvm/gallivm.h @@ -67,8 +67,7 @@ int gallivm_fragment_shader_exec(struct gallivm_prog *prog, float (*dests)[PIPE_MAX_SHADER_INPUTS][4], float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], float (*consts)[4], - struct tgsi_sampler *samplers, - unsigned *sampler_units); + struct tgsi_sampler *samplers); void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], const struct tgsi_interp_coef *coefs); diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index f141ea2da0..a703ba3862 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -770,8 +770,6 @@ Module* createBaseShader() { int32_num_consts_125->setName("num_consts"); Value* ptr_samplers = args++; ptr_samplers->setName("samplers"); - Value* ptr_sampler_units = args++; - ptr_sampler_units->setName("sampler_units"); BasicBlock* label_entry_126 = new BasicBlock("entry",func_run_fragment_shader,0); BasicBlock* label_forbody6_i_127 = new BasicBlock("forbody6.i",func_run_fragment_shader,0); diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index fe32e7810d..cbe4965ef6 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -219,8 +219,7 @@ int run_fragment_shader(float x, float y, int num_inputs, float (*aconsts)[4], int num_consts, - struct tgsi_sampler *samplers, - unsigned *sampler_units) + struct tgsi_sampler *samplers) { float4 inputs[4][16]; float4 consts[32]; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 6b97844445..83b4ab07fb 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -142,15 +142,15 @@ struct pipe_context { void (*set_polygon_stipple)( struct pipe_context *, const struct pipe_poly_stipple * ); - void (*set_sampler_units)( struct pipe_context *, - uint num_samplers, const uint *units ); - void (*set_scissor_state)( struct pipe_context *, const struct pipe_scissor_state * ); - void (*set_texture_state)( struct pipe_context *, - unsigned unit, - struct pipe_texture * ); + + /* Currently a sampler is constrained to sample from a single texture: + */ + void (*set_sampler_texture)( struct pipe_context *, + unsigned sampler, + struct pipe_texture * ); void (*set_viewport_state)( struct pipe_context *, const struct pipe_viewport_state * ); diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 43f23dc1e8..7d243aaabb 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -307,9 +307,8 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; - softpipe->pipe.set_sampler_units = softpipe_set_sampler_units; softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; - softpipe->pipe.set_texture_state = softpipe_set_texture_state; + softpipe->pipe.set_sampler_texture = softpipe_set_sampler_texture; softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 45d15c720e..afdd0ec88e 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -91,7 +91,6 @@ struct softpipe_context { struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - uint sampler_units[PIPE_MAX_SAMPLERS]; unsigned dirty; /* diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 75576a9bde..9307ed233d 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -98,7 +98,6 @@ shade_quad( /* Consts does not require 16 byte alignment. */ machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; - machine->SamplerUnits = softpipe->sampler_units; machine->InterpCoefs = quad->coef; machine->Inputs[0].xyzw[0].f[0] = fx; @@ -206,7 +205,7 @@ shade_quad_llvm(struct quad_stage *qs, /*quad->mask &=*/ gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs, softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], - qss->samplers, softpipe->sampler_units); + qss->samplers); #if DLLVM printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n", dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 80a1cba25a..a3bd078a71 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -111,13 +111,10 @@ void softpipe_delete_vs_state(struct pipe_context *, void *); void softpipe_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); -void softpipe_set_sampler_units( struct pipe_context *, - uint numSamplers, const uint *units ); - void softpipe_set_scissor_state( struct pipe_context *, const struct pipe_scissor_state * ); -void softpipe_set_texture_state( struct pipe_context *, +void softpipe_set_sampler_texture( struct pipe_context *, unsigned unit, struct pipe_texture * ); diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 173901f04e..3842e71503 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -67,9 +67,9 @@ softpipe_delete_sampler_state(struct pipe_context *pipe, void -softpipe_set_texture_state(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) +softpipe_set_sampler_texture(struct pipe_context *pipe, + unsigned unit, + struct pipe_texture *texture) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -82,15 +82,4 @@ softpipe_set_texture_state(struct pipe_context *pipe, } -void -softpipe_set_sampler_units(struct pipe_context *pipe, - uint num_samplers, const uint *units ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - uint i; - for (i = 0; i < num_samplers; i++) - softpipe->sampler_units[i] = units[i]; - softpipe->dirty |= SP_NEW_SAMPLER; -} - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index ab83f27c1b..8636271a34 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1217,8 +1217,7 @@ exec_tex(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst, boolean biasLod) { - const uint sampler = inst->FullSrcRegisters[1].SrcRegister.Index; - const uint unit = mach->SamplerUnits[sampler]; + const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; union tgsi_exec_channel r[8]; uint chan_index; float lodBias; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 2c62b30f15..e7952a08e3 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -162,7 +162,6 @@ struct tgsi_exec_machine struct tgsi_exec_vector *Temps; struct tgsi_exec_vector *Addrs; - uint *SamplerUnits; struct tgsi_sampler *Samplers; float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 67a9159069..052b6dd144 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -172,26 +172,13 @@ update_samplers(struct st_context *st) st->pipe->bind_sampler_state(st->pipe, u, cso->data); } } - - - /* mapping from sampler vars to texture units */ - { - struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current; - uint sample_units[PIPE_MAX_SAMPLERS]; - uint s; - for (s = 0; s < PIPE_MAX_SAMPLERS; s++) { - sample_units[s] = fprog->Base.SamplerUnits[s]; - } - - st->pipe->set_sampler_units(st->pipe, PIPE_MAX_SAMPLERS, sample_units); - } } const struct st_tracked_state st_update_sampler = { .name = "st_update_sampler", .dirty = { - .mesa = _NEW_TEXTURE | _NEW_PROGRAM, + .mesa = _NEW_TEXTURE, .st = 0, }, .update = update_samplers diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index c4e5af02d5..fb21d29c40 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -46,20 +46,24 @@ static void update_textures(struct st_context *st) { - GLuint u; + GLuint s; - for (u = 0; u < st->ctx->Const.MaxTextureImageUnits; u++) { + /* ST_NEW_FRAGMENT_PROGRAM + */ + struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current; + + for (s = 0; s < st->ctx->Const.MaxTextureCoordUnits; s++) { + GLuint su = fprog->Base.SamplerUnits[s]; + struct gl_texture_object *texObj - = st->ctx->Texture.Unit[u]._Current; + = st->ctx->Texture.Unit[su]._Current; + struct pipe_texture *pt; + if (texObj) { GLboolean flush, retval; - retval = st_finalize_texture(st->ctx, st->pipe, u, &flush); -#if 0 - printf("finalize_texture returned %d, flush = %d\n", - retval, flush); -#endif + retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); pt = st_get_texobj_texture(texObj); } @@ -67,8 +71,14 @@ update_textures(struct st_context *st) pt = NULL; } - st->state.texture[u] = pt; - st->pipe->set_texture_state(st->pipe, u, pt); + /* XXX: need to ensure that textures are unbound/removed from + * this table before being deleted, otherwise the pointer + * comparison below could fail. + */ + if (st->state.sampler_texture[s] != pt) { + st->state.sampler_texture[s] = pt; + st->pipe->set_sampler_texture(st->pipe, s, pt); + } } } @@ -77,7 +87,7 @@ const struct st_tracked_state st_update_texture = { .name = "st_update_texture", .dirty = { .mesa = _NEW_TEXTURE, - .st = 0, + .st = ST_NEW_FRAGMENT_PROGRAM, }, .update = update_textures }; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 0179000353..5d4c443c01 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -697,7 +697,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* texture state: */ { - pipe->set_texture_state(pipe, unit, pt); + pipe->set_sampler_texture(pipe, unit, pt); } /* Compute window coords (y=0=bottom) with pixel zoom. @@ -719,7 +719,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); pipe->bind_fs_state(pipe, ctx->st->state.fs->data); pipe->bind_vs_state(pipe, ctx->st->state.vs->data); - pipe->set_texture_state(pipe, unit, ctx->st->state.texture[unit]); + pipe->set_sampler_texture(pipe, unit, ctx->st->state.sampler_texture[unit]); pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]->data); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 2c93a2f3dd..39dd21dc59 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1412,10 +1412,10 @@ copy_image_data_to_texture(struct st_context *st, */ GLboolean st_finalize_texture(GLcontext *ctx, - struct pipe_context *pipe, GLuint unit, + struct pipe_context *pipe, + struct gl_texture_object *tObj, GLboolean *needFlush) { - struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; struct st_texture_object *stObj = st_texture_object(tObj); int comp_byte = 0; int cpp; diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h index 7f8082b029..878256ec26 100644 --- a/src/mesa/state_tracker/st_cb_texture.h +++ b/src/mesa/state_tracker/st_cb_texture.h @@ -8,7 +8,8 @@ st_get_texobj_texture(struct gl_texture_object *texObj); extern GLboolean st_finalize_texture(GLcontext *ctx, - struct pipe_context *pipe, GLuint unit, + struct pipe_context *pipe, + struct gl_texture_object *tObj, GLboolean *needFlush); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 4855961d09..87646b3c71 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -106,7 +106,7 @@ struct st_context struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; struct pipe_framebuffer_state framebuffer; - struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_texture *sampler_texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_viewport_state viewport; -- cgit v1.2.3 From 8681deddd7a7e749adaf43c7df4313ea54922e62 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 5 Nov 2007 13:41:56 -0500 Subject: Rewrite argument passing to prepare for handling of the kil instruction. Pass the inputs/outputs pointer in the structure instead of infinitely expanding arguments to the functions. --- src/mesa/pipe/llvm/gallivm.cpp | 18 +- src/mesa/pipe/llvm/instructions.cpp | 20 +- src/mesa/pipe/llvm/instructions.h | 3 +- src/mesa/pipe/llvm/llvm_base_shader.cpp | 1425 ++++++++++++++++--------------- src/mesa/pipe/llvm/llvm_entry.c | 41 +- src/mesa/pipe/llvm/storage.cpp | 190 ++--- src/mesa/pipe/llvm/storage.h | 35 +- 7 files changed, 852 insertions(+), 880 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index a60440022a..65c95074fd 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -485,11 +485,7 @@ translate_instruction(llvm::Module *module, case TGSI_OPCODE_BRA: break; case TGSI_OPCODE_CAL: { - instr->cal(inst->InstructionExtLabel.Label, - storage->outputPtr(), - storage->inputPtr(), - storage->constPtr(), - storage->tempPtr()); + instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr()); return; } break; @@ -740,14 +736,8 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) shader->setName(func_name.c_str()); Function::arg_iterator args = shader->arg_begin(); - Value *ptr_OUT = args++; - ptr_OUT->setName("OUT"); - Value *ptr_IN = args++; - ptr_IN->setName("IN"); - Value *ptr_CONST = args++; - ptr_CONST->setName("CONST"); - Value *ptr_TEMPS = args++; - ptr_TEMPS->setName("TEMPS"); + Value *ptr_INPUT = args++; + ptr_INPUT->setName("input"); BasicBlock *label_entry = new BasicBlock("entry", shader, 0); @@ -755,7 +745,7 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); - Storage storage(label_entry, ptr_OUT, ptr_IN, ptr_CONST, ptr_TEMPS); + Storage storage(label_entry, ptr_INPUT); Instructions instr(mod, shader, label_entry, &storage); while(!tgsi_parse_end_of_tokens(&parse)) { tgsi_parse_token(&parse); diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 232dd9cd5d..7a70aec878 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -732,14 +732,10 @@ void Instructions::end() m_builder.CreateRetVoid(); } -void Instructions::cal(int label, llvm::Value *out, llvm::Value *in, - llvm::Value *cst, llvm::Value *temp) +void Instructions::cal(int label, llvm::Value *input) { std::vector params; - params.push_back(out); - params.push_back(in); - params.push_back(cst); - params.push_back(temp); + params.push_back(input); llvm::Function *func = findFunction(label); m_builder.CreateCall(func, params.begin(), params.end()); @@ -773,15 +769,9 @@ void Instructions::bgnSub(unsigned label) llvm::Function *func = findFunction(label); Function::arg_iterator args = func->arg_begin(); - Value *ptr_OUT = args++; - ptr_OUT->setName("OUT"); - Value *ptr_IN = args++; - ptr_IN->setName("IN"); - Value *ptr_CONST = args++; - ptr_CONST->setName("CONST"); - Value *ptr_TEMP = args++; - ptr_TEMP->setName("TEMP"); - m_storage->pushArguments(ptr_OUT, ptr_IN, ptr_CONST, ptr_TEMP); + Value *ptr_INPUT = args++; + ptr_INPUT->setName("INPUT"); + m_storage->pushArguments(ptr_INPUT); llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index e9bfc9d740..c31cc4f125 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -62,8 +62,7 @@ public: void beginLoop(); void bgnSub(unsigned); void brk(); - void cal(int label, llvm::Value *out, llvm::Value *in, - llvm::Value *cst, llvm::Value *tmp); + void cal(int label, llvm::Value *input); llvm::Value *cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); llvm::Value *cos(llvm::Value *in); llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index a703ba3862..6e7fa32807 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -4,8 +4,23 @@ Module* createBaseShader() { // Module Construction Module* mod = new Module("Shader"); + mod->setDataLayout(""); + mod->setTargetTriple("i686-apple-darwin9"); // Type Definitions + std::vectorStructTy_struct_ShaderInput_fields; + VectorType* VectorTy_1 = VectorType::get(Type::FloatTy, 4); + + PointerType* PointerTy_0 = PointerType::get(VectorTy_1); + + StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); + StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); + StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); + StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); + StructTy_struct_ShaderInput_fields.push_back(IntegerType::get(32)); + StructType* StructTy_struct_ShaderInput = StructType::get(StructTy_struct_ShaderInput_fields, /*isPacked=*/false); + mod->addTypeName("struct.ShaderInput", StructTy_struct_ShaderInput); + OpaqueType* OpaqueTy_struct_pipe_mipmap_tree = OpaqueType::get(); mod->addTypeName("struct.pipe_mipmap_tree", OpaqueTy_struct_pipe_mipmap_tree); @@ -16,77 +31,73 @@ Module* createBaseShader() { mod->addTypeName("struct.softpipe_tile_cache", OpaqueTy_struct_softpipe_tile_cache); std::vectorStructTy_struct_tgsi_sampler_fields; - PointerType* PointerTy_0 = PointerType::get(OpaqueTy_struct_pipe_sampler_state); + PointerType* PointerTy_2 = PointerType::get(OpaqueTy_struct_pipe_sampler_state); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_0); - PointerType* PointerTy_1 = PointerType::get(OpaqueTy_struct_pipe_mipmap_tree); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_2); + PointerType* PointerTy_3 = PointerType::get(OpaqueTy_struct_pipe_mipmap_tree); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_1); - std::vectorFuncTy_3_args; + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_3); + std::vectorFuncTy_5_args; PATypeHolder StructTy_struct_tgsi_sampler_fwd = OpaqueType::get(); - PointerType* PointerTy_4 = PointerType::get(StructTy_struct_tgsi_sampler_fwd); + PointerType* PointerTy_6 = PointerType::get(StructTy_struct_tgsi_sampler_fwd); - FuncTy_3_args.push_back(PointerTy_4); - PointerType* PointerTy_5 = PointerType::get(Type::FloatTy); + FuncTy_5_args.push_back(PointerTy_6); + PointerType* PointerTy_7 = PointerType::get(Type::FloatTy); - FuncTy_3_args.push_back(PointerTy_5); - FuncTy_3_args.push_back(PointerTy_5); - FuncTy_3_args.push_back(PointerTy_5); - FuncTy_3_args.push_back(Type::FloatTy); - ArrayType* ArrayTy_7 = ArrayType::get(Type::FloatTy, 4); + FuncTy_5_args.push_back(PointerTy_7); + FuncTy_5_args.push_back(PointerTy_7); + FuncTy_5_args.push_back(PointerTy_7); + FuncTy_5_args.push_back(Type::FloatTy); + ArrayType* ArrayTy_9 = ArrayType::get(Type::FloatTy, 4); - PointerType* PointerTy_6 = PointerType::get(ArrayTy_7); + PointerType* PointerTy_8 = PointerType::get(ArrayTy_9); - FuncTy_3_args.push_back(PointerTy_6); - ParamAttrsList *FuncTy_3_PAL = 0; - FunctionType* FuncTy_3 = FunctionType::get( + FuncTy_5_args.push_back(PointerTy_8); + ParamAttrsList *FuncTy_5_PAL = 0; + FunctionType* FuncTy_5 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_3_args, + /*Params=*/FuncTy_5_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_3_PAL); + /*ParamAttrs=*/FuncTy_5_PAL); - PointerType* PointerTy_2 = PointerType::get(FuncTy_3); + PointerType* PointerTy_4 = PointerType::get(FuncTy_5); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_2); - PointerType* PointerTy_8 = PointerType::get(IntegerType::get(8)); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_4); + PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8)); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_8); - PointerType* PointerTy_9 = PointerType::get(OpaqueTy_struct_softpipe_tile_cache); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_10); + PointerType* PointerTy_11 = PointerType::get(OpaqueTy_struct_softpipe_tile_cache); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_9); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_11); StructType* StructTy_struct_tgsi_sampler = StructType::get(StructTy_struct_tgsi_sampler_fields, /*isPacked=*/false); mod->addTypeName("struct.tgsi_sampler", StructTy_struct_tgsi_sampler); cast(StructTy_struct_tgsi_sampler_fwd.get())->refineAbstractTypeTo(StructTy_struct_tgsi_sampler); StructTy_struct_tgsi_sampler = cast(StructTy_struct_tgsi_sampler_fwd.get()); - std::vectorFuncTy_10_args; - VectorType* VectorTy_13 = VectorType::get(Type::FloatTy, 4); - - ArrayType* ArrayTy_12 = ArrayType::get(VectorTy_13, 16); + std::vectorFuncTy_12_args; + ArrayType* ArrayTy_14 = ArrayType::get(VectorTy_1, 16); - PointerType* PointerTy_11 = PointerType::get(ArrayTy_12); + PointerType* PointerTy_13 = PointerType::get(ArrayTy_14); - FuncTy_10_args.push_back(PointerTy_11); - ArrayType* ArrayTy_15 = ArrayType::get(ArrayTy_7, 16); + FuncTy_12_args.push_back(PointerTy_13); + ArrayType* ArrayTy_16 = ArrayType::get(ArrayTy_9, 16); - PointerType* PointerTy_14 = PointerType::get(ArrayTy_15); + PointerType* PointerTy_15 = PointerType::get(ArrayTy_16); - FuncTy_10_args.push_back(PointerTy_14); - FuncTy_10_args.push_back(IntegerType::get(32)); - FuncTy_10_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_10_PAL = 0; - FunctionType* FuncTy_10 = FunctionType::get( + FuncTy_12_args.push_back(PointerTy_15); + FuncTy_12_args.push_back(IntegerType::get(32)); + FuncTy_12_args.push_back(IntegerType::get(32)); + ParamAttrsList *FuncTy_12_PAL = 0; + FunctionType* FuncTy_12 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_10_args, + /*Params=*/FuncTy_12_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_10_PAL); - - PointerType* PointerTy_16 = PointerType::get(VectorTy_13); + /*ParamAttrs=*/FuncTy_12_PAL); std::vectorFuncTy_17_args; - FuncTy_17_args.push_back(PointerTy_16); - FuncTy_17_args.push_back(PointerTy_6); + FuncTy_17_args.push_back(PointerTy_0); + FuncTy_17_args.push_back(PointerTy_8); FuncTy_17_args.push_back(IntegerType::get(32)); ParamAttrsList *FuncTy_17_PAL = 0; FunctionType* FuncTy_17 = FunctionType::get( @@ -96,8 +107,8 @@ Module* createBaseShader() { /*ParamAttrs=*/FuncTy_17_PAL); std::vectorFuncTy_18_args; - FuncTy_18_args.push_back(PointerTy_6); - FuncTy_18_args.push_back(PointerTy_16); + FuncTy_18_args.push_back(PointerTy_8); + FuncTy_18_args.push_back(PointerTy_0); FuncTy_18_args.push_back(IntegerType::get(32)); ParamAttrsList *FuncTy_18_PAL = 0; FunctionType* FuncTy_18 = FunctionType::get( @@ -107,9 +118,9 @@ Module* createBaseShader() { /*ParamAttrs=*/FuncTy_18_PAL); std::vectorFuncTy_19_args; - FuncTy_19_args.push_back(PointerTy_14); - FuncTy_19_args.push_back(PointerTy_14); - FuncTy_19_args.push_back(PointerTy_6); + FuncTy_19_args.push_back(PointerTy_15); + FuncTy_19_args.push_back(PointerTy_15); + FuncTy_19_args.push_back(PointerTy_8); FuncTy_19_args.push_back(IntegerType::get(32)); FuncTy_19_args.push_back(IntegerType::get(32)); FuncTy_19_args.push_back(IntegerType::get(32)); @@ -121,60 +132,61 @@ Module* createBaseShader() { /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_19_PAL); - ArrayType* ArrayTy_21 = ArrayType::get(ArrayTy_12, 2048); + ArrayType* ArrayTy_21 = ArrayType::get(ArrayTy_14, 2048); PointerType* PointerTy_20 = PointerType::get(ArrayTy_21); - ArrayType* ArrayTy_23 = ArrayType::get(VectorTy_13, 32); + ArrayType* ArrayTy_23 = ArrayType::get(VectorTy_1, 32); PointerType* PointerTy_22 = PointerType::get(ArrayTy_23); - ArrayType* ArrayTy_25 = ArrayType::get(VectorTy_13, 128); + ArrayType* ArrayTy_25 = ArrayType::get(VectorTy_1, 128); PointerType* PointerTy_24 = PointerType::get(ArrayTy_25); - std::vectorFuncTy_27_args; - FuncTy_27_args.push_back(PointerTy_16); - FuncTy_27_args.push_back(PointerTy_16); - FuncTy_27_args.push_back(PointerTy_16); - FuncTy_27_args.push_back(PointerTy_16); - ParamAttrsList *FuncTy_27_PAL = 0; - FunctionType* FuncTy_27 = FunctionType::get( + PointerType* PointerTy_26 = PointerType::get(StructTy_struct_ShaderInput); + + PointerType* PointerTy_27 = PointerType::get(PointerTy_0); + + std::vectorFuncTy_29_args; + FuncTy_29_args.push_back(PointerTy_26); + ParamAttrsList *FuncTy_29_PAL = 0; + FunctionType* FuncTy_29 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_27_args, + /*Params=*/FuncTy_29_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_27_PAL); + /*ParamAttrs=*/FuncTy_29_PAL); - PointerType* PointerTy_26 = PointerType::get(FuncTy_27); + PointerType* PointerTy_28 = PointerType::get(FuncTy_29); - std::vectorFuncTy_28_args; - FuncTy_28_args.push_back(Type::FloatTy); - FuncTy_28_args.push_back(Type::FloatTy); - FuncTy_28_args.push_back(PointerTy_14); - FuncTy_28_args.push_back(PointerTy_14); - FuncTy_28_args.push_back(IntegerType::get(32)); - FuncTy_28_args.push_back(PointerTy_6); - FuncTy_28_args.push_back(IntegerType::get(32)); - FuncTy_28_args.push_back(PointerTy_4); - PointerType* PointerTy_29 = PointerType::get(IntegerType::get(32)); + std::vectorFuncTy_30_args; + FuncTy_30_args.push_back(Type::FloatTy); + FuncTy_30_args.push_back(Type::FloatTy); + FuncTy_30_args.push_back(PointerTy_15); + FuncTy_30_args.push_back(PointerTy_15); + FuncTy_30_args.push_back(IntegerType::get(32)); + FuncTy_30_args.push_back(PointerTy_8); + FuncTy_30_args.push_back(IntegerType::get(32)); + FuncTy_30_args.push_back(PointerTy_6); + PointerType* PointerTy_31 = PointerType::get(IntegerType::get(32)); - FuncTy_28_args.push_back(PointerTy_29); - ParamAttrsList *FuncTy_28_PAL = 0; - FunctionType* FuncTy_28 = FunctionType::get( + FuncTy_30_args.push_back(PointerTy_31); + ParamAttrsList *FuncTy_30_PAL = 0; + FunctionType* FuncTy_30 = FunctionType::get( /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_28_args, + /*Params=*/FuncTy_30_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_28_PAL); + /*ParamAttrs=*/FuncTy_30_PAL); - ArrayType* ArrayTy_31 = ArrayType::get(ArrayTy_12, 4); + ArrayType* ArrayTy_33 = ArrayType::get(ArrayTy_14, 4); - PointerType* PointerTy_30 = PointerType::get(ArrayTy_31); + PointerType* PointerTy_32 = PointerType::get(ArrayTy_33); // Function Declarations Function* func_from_array = new Function( - /*Type=*/FuncTy_10, + /*Type=*/FuncTy_12, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_array", mod); func_from_array->setCallingConv(CallingConv::C); @@ -198,13 +210,13 @@ Module* createBaseShader() { func_run_vertex_shader->setCallingConv(CallingConv::C); Function* func_execute_shader = new Function( - /*Type=*/FuncTy_27, + /*Type=*/FuncTy_29, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"execute_shader", mod); // (external, no body) func_execute_shader->setCallingConv(CallingConv::C); Function* func_run_fragment_shader = new Function( - /*Type=*/FuncTy_28, + /*Type=*/FuncTy_30, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_fragment_shader", mod); func_run_fragment_shader->setCallingConv(CallingConv::C); @@ -213,13 +225,13 @@ Module* createBaseShader() { // Constant Definitions - Constant* const_int32_32 = Constant::getNullValue(IntegerType::get(32)); - UndefValue* const_packed_33 = UndefValue::get(VectorTy_13); - ConstantInt* const_int32_34 = ConstantInt::get(APInt(32, "1", 10)); - ConstantInt* const_int32_35 = ConstantInt::get(APInt(32, "2", 10)); - ConstantInt* const_int32_36 = ConstantInt::get(APInt(32, "3", 10)); - ConstantInt* const_int32_37 = ConstantInt::get(APInt(32, "4", 10)); - ConstantInt* const_int32_38 = ConstantInt::get(APInt(32, "-1", 10)); + Constant* const_int32_34 = Constant::getNullValue(IntegerType::get(32)); + UndefValue* const_packed_35 = UndefValue::get(VectorTy_1); + ConstantInt* const_int32_36 = ConstantInt::get(APInt(32, "1", 10)); + ConstantInt* const_int32_37 = ConstantInt::get(APInt(32, "2", 10)); + ConstantInt* const_int32_38 = ConstantInt::get(APInt(32, "3", 10)); + ConstantInt* const_int32_39 = ConstantInt::get(APInt(32, "4", 10)); + ConstantInt* const_int32_40 = ConstantInt::get(APInt(32, "-1", 10)); // Global Variable Definitions @@ -243,73 +255,73 @@ Module* createBaseShader() { BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); // Block entry (label_entry) - ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_32, "cmp", label_entry); - ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_32, "cmp5", label_entry); + ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_34, "cmp", label_entry); + ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_34, "cmp5", label_entry); BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); new BranchInst(label_forbody6, label_afterfor60, int1_bothcond, label_entry); // Block forbody6 (label_forbody6) - Argument* fwdref_40 = new Argument(IntegerType::get(32)); - Argument* fwdref_41 = new Argument(IntegerType::get(32)); + Argument* fwdref_42 = new Argument(IntegerType::get(32)); + Argument* fwdref_43 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody6); int32_i_0_reg2mem_0->reserveOperandSpace(3); - int32_i_0_reg2mem_0->addIncoming(const_int32_32, label_entry); - int32_i_0_reg2mem_0->addIncoming(fwdref_40, label_forinc57); - int32_i_0_reg2mem_0->addIncoming(fwdref_41, label_forbody6); + int32_i_0_reg2mem_0->addIncoming(const_int32_34, label_entry); + int32_i_0_reg2mem_0->addIncoming(fwdref_42, label_forinc57); + int32_i_0_reg2mem_0->addIncoming(fwdref_43, label_forbody6); - Argument* fwdref_42 = new Argument(IntegerType::get(32)); + Argument* fwdref_44 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0", label_forbody6); int32_j_0_reg2mem_0->reserveOperandSpace(3); - int32_j_0_reg2mem_0->addIncoming(fwdref_42, label_forbody6); - int32_j_0_reg2mem_0->addIncoming(const_int32_32, label_forinc57); - int32_j_0_reg2mem_0->addIncoming(const_int32_32, label_entry); + int32_j_0_reg2mem_0->addIncoming(fwdref_44, label_forbody6); + int32_j_0_reg2mem_0->addIncoming(const_int32_34, label_forinc57); + int32_j_0_reg2mem_0->addIncoming(const_int32_34, label_entry); - Argument* fwdref_43 = new Argument(VectorTy_13); - PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_13, "vec.0.reg2mem.0", label_forbody6); + Argument* fwdref_45 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody6); packed_vec_0_reg2mem_0->reserveOperandSpace(3); - packed_vec_0_reg2mem_0->addIncoming(fwdref_43, label_forbody6); - packed_vec_0_reg2mem_0->addIncoming(const_packed_33, label_entry); - packed_vec_0_reg2mem_0->addIncoming(fwdref_43, label_forinc57); + packed_vec_0_reg2mem_0->addIncoming(fwdref_45, label_forbody6); + packed_vec_0_reg2mem_0->addIncoming(const_packed_35, label_entry); + packed_vec_0_reg2mem_0->addIncoming(fwdref_45, label_forinc57); std::vector ptr_arraydecay11_indices; ptr_arraydecay11_indices.push_back(int32_i_0_reg2mem_0); ptr_arraydecay11_indices.push_back(int32_j_0_reg2mem_0); - ptr_arraydecay11_indices.push_back(const_int32_32); + ptr_arraydecay11_indices.push_back(const_int32_34); Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); - InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_32, "tmp15", label_forbody6); + InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_34, "tmp15", label_forbody6); std::vector ptr_arrayidx23_indices; ptr_arrayidx23_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx23_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx23_indices.push_back(const_int32_34); + ptr_arrayidx23_indices.push_back(const_int32_36); Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_34, "tmp26", label_forbody6); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_36, "tmp26", label_forbody6); std::vector ptr_arrayidx34_indices; ptr_arrayidx34_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx34_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx34_indices.push_back(const_int32_35); + ptr_arrayidx34_indices.push_back(const_int32_37); Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6); LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6); - InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_35, "tmp37", label_forbody6); + InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_37, "tmp37", label_forbody6); std::vector ptr_arrayidx45_indices; ptr_arrayidx45_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx45_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx45_indices.push_back(const_int32_36); + ptr_arrayidx45_indices.push_back(const_int32_38); Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6); LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6); - InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_36, "tmp48", label_forbody6); + InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_38, "tmp48", label_forbody6); std::vector ptr_arrayidx54_indices; ptr_arrayidx54_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0); Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); - StoreInst* void_44 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); - BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_34, "inc", label_forbody6); + StoreInst* void_46 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); + BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_36, "inc", label_forbody6); ICmpInst* int1_cmp59 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp59", label_forbody6); new BranchInst(label_forbody6, label_forinc57, int1_cmp59, label_forbody6); // Block forinc57 (label_forinc57) - BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_34, "inc59", label_forinc57); + BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_36, "inc59", label_forinc57); ICmpInst* int1_cmp17 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59, int32_count, "cmp17", label_forinc57); new BranchInst(label_forbody6, label_afterfor60, int1_cmp17, label_forinc57); @@ -317,80 +329,80 @@ Module* createBaseShader() { new ReturnInst(label_afterfor60); // Resolve Forward References - fwdref_41->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_41; - fwdref_43->replaceAllUsesWith(packed_tmp48); delete fwdref_43; - fwdref_42->replaceAllUsesWith(int32_inc); delete fwdref_42; - fwdref_40->replaceAllUsesWith(int32_inc59); delete fwdref_40; + fwdref_43->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_43; + fwdref_45->replaceAllUsesWith(packed_tmp48); delete fwdref_45; + fwdref_44->replaceAllUsesWith(int32_inc); delete fwdref_44; + fwdref_42->replaceAllUsesWith(int32_inc59); delete fwdref_42; } // Function: from_consts (func_from_consts) { Function::arg_iterator args = func_from_consts->arg_begin(); - Value* ptr_res_48 = args++; - ptr_res_48->setName("res"); - Value* ptr_ainputs_49 = args++; - ptr_ainputs_49->setName("ainputs"); - Value* int32_count_50 = args++; - int32_count_50->setName("count"); - - BasicBlock* label_entry_51 = new BasicBlock("entry",func_from_consts,0); + Value* ptr_res_50 = args++; + ptr_res_50->setName("res"); + Value* ptr_ainputs_51 = args++; + ptr_ainputs_51->setName("ainputs"); + Value* int32_count_52 = args++; + int32_count_52->setName("count"); + + BasicBlock* label_entry_53 = new BasicBlock("entry",func_from_consts,0); BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); - // Block entry (label_entry_51) - ICmpInst* int1_cmp_52 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_50, const_int32_32, "cmp", label_entry_51); - new BranchInst(label_forbody, label_afterfor, int1_cmp_52, label_entry_51); + // Block entry (label_entry_53) + ICmpInst* int1_cmp_54 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_52, const_int32_34, "cmp", label_entry_53); + new BranchInst(label_forbody, label_afterfor, int1_cmp_54, label_entry_53); // Block forbody (label_forbody) - Argument* fwdref_55 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_54 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); - int32_i_0_reg2mem_0_54->reserveOperandSpace(2); - int32_i_0_reg2mem_0_54->addIncoming(const_int32_32, label_entry_51); - int32_i_0_reg2mem_0_54->addIncoming(fwdref_55, label_forbody); - - Argument* fwdref_57 = new Argument(VectorTy_13); - PHINode* packed_vec_0_reg2mem_0_56 = new PHINode(VectorTy_13, "vec.0.reg2mem.0", label_forbody); - packed_vec_0_reg2mem_0_56->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_56->addIncoming(const_packed_33, label_entry_51); - packed_vec_0_reg2mem_0_56->addIncoming(fwdref_57, label_forbody); + Argument* fwdref_57 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_56 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); + int32_i_0_reg2mem_0_56->reserveOperandSpace(2); + int32_i_0_reg2mem_0_56->addIncoming(const_int32_34, label_entry_53); + int32_i_0_reg2mem_0_56->addIncoming(fwdref_57, label_forbody); + + Argument* fwdref_59 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_58 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody); + packed_vec_0_reg2mem_0_58->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_58->addIncoming(const_packed_35, label_entry_53); + packed_vec_0_reg2mem_0_58->addIncoming(fwdref_59, label_forbody); std::vector ptr_arraydecay_indices; - ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_54); - ptr_arraydecay_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_49, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); + ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_56); + ptr_arraydecay_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_51, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); - InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_56, float_tmp5, const_int32_32, "tmp7", label_forbody); + InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_58, float_tmp5, const_int32_34, "tmp7", label_forbody); std::vector ptr_arrayidx12_indices; - ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_54); - ptr_arrayidx12_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_49, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); - LoadInst* float_tmp13_58 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); - InsertElementInst* packed_tmp15_59 = new InsertElementInst(packed_tmp7, float_tmp13_58, const_int32_34, "tmp15", label_forbody); + ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_56); + ptr_arrayidx12_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_51, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); + LoadInst* float_tmp13_60 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); + InsertElementInst* packed_tmp15_61 = new InsertElementInst(packed_tmp7, float_tmp13_60, const_int32_36, "tmp15", label_forbody); std::vector ptr_arrayidx20_indices; - ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_54); - ptr_arrayidx20_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_49, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); + ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_56); + ptr_arrayidx20_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_51, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); - InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_59, float_tmp21, const_int32_35, "tmp23", label_forbody); + InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_61, float_tmp21, const_int32_37, "tmp23", label_forbody); std::vector ptr_arrayidx28_indices; - ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_54); - ptr_arrayidx28_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_49, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); + ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_56); + ptr_arrayidx28_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_51, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); - InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_36, "tmp31", label_forbody); - GetElementPtrInst* ptr_arrayidx34_60 = new GetElementPtrInst(ptr_res_48, int32_i_0_reg2mem_0_54, "arrayidx34", label_forbody); - StoreInst* void_61 = new StoreInst(packed_tmp31, ptr_arrayidx34_60, false, label_forbody); - BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_54, const_int32_34, "indvar.next", label_forbody); - ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_50, "exitcond", label_forbody); + InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_38, "tmp31", label_forbody); + GetElementPtrInst* ptr_arrayidx34_62 = new GetElementPtrInst(ptr_res_50, int32_i_0_reg2mem_0_56, "arrayidx34", label_forbody); + StoreInst* void_63 = new StoreInst(packed_tmp31, ptr_arrayidx34_62, false, label_forbody); + BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_56, const_int32_36, "indvar.next", label_forbody); + ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_52, "exitcond", label_forbody); new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); // Block afterfor (label_afterfor) new ReturnInst(label_afterfor); // Resolve Forward References - fwdref_57->replaceAllUsesWith(packed_tmp31); delete fwdref_57; - fwdref_55->replaceAllUsesWith(int32_indvar_next); delete fwdref_55; + fwdref_59->replaceAllUsesWith(packed_tmp31); delete fwdref_59; + fwdref_57->replaceAllUsesWith(int32_indvar_next); delete fwdref_57; } @@ -401,353 +413,363 @@ Module* createBaseShader() { ptr_dests->setName("dests"); Value* ptr_in = args++; ptr_in->setName("in"); - Value* int32_num_attribs_64 = args++; - int32_num_attribs_64->setName("num_attribs"); - - BasicBlock* label_entry_65 = new BasicBlock("entry",func_to_array,0); - BasicBlock* label_forbody_66 = new BasicBlock("forbody",func_to_array,0); - BasicBlock* label_afterfor_67 = new BasicBlock("afterfor",func_to_array,0); - - // Block entry (label_entry_65) - ICmpInst* int1_cmp_68 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_64, const_int32_32, "cmp", label_entry_65); - new BranchInst(label_forbody_66, label_afterfor_67, int1_cmp_68, label_entry_65); - - // Block forbody (label_forbody_66) - Argument* fwdref_71 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_70 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_66); - int32_i_0_reg2mem_0_70->reserveOperandSpace(2); - int32_i_0_reg2mem_0_70->addIncoming(const_int32_32, label_entry_65); - int32_i_0_reg2mem_0_70->addIncoming(fwdref_71, label_forbody_66); - - std::vector ptr_arraydecay_72_indices; - ptr_arraydecay_72_indices.push_back(int32_i_0_reg2mem_0_70); - ptr_arraydecay_72_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay_72 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_72_indices.begin(), ptr_arraydecay_72_indices.end(), "arraydecay", label_forbody_66); - GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_70, "arrayidx6", label_forbody_66); - LoadInst* packed_tmp7_73 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_66); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_73, const_int32_32, "tmp11", label_forbody_66); - StoreInst* void_74 = new StoreInst(float_tmp11, ptr_arraydecay_72, false, label_forbody_66); + Value* int32_num_attribs_66 = args++; + int32_num_attribs_66->setName("num_attribs"); + + BasicBlock* label_entry_67 = new BasicBlock("entry",func_to_array,0); + BasicBlock* label_forbody_68 = new BasicBlock("forbody",func_to_array,0); + BasicBlock* label_afterfor_69 = new BasicBlock("afterfor",func_to_array,0); + + // Block entry (label_entry_67) + ICmpInst* int1_cmp_70 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_66, const_int32_34, "cmp", label_entry_67); + new BranchInst(label_forbody_68, label_afterfor_69, int1_cmp_70, label_entry_67); + + // Block forbody (label_forbody_68) + Argument* fwdref_73 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_72 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_68); + int32_i_0_reg2mem_0_72->reserveOperandSpace(2); + int32_i_0_reg2mem_0_72->addIncoming(const_int32_34, label_entry_67); + int32_i_0_reg2mem_0_72->addIncoming(fwdref_73, label_forbody_68); + + std::vector ptr_arraydecay_74_indices; + ptr_arraydecay_74_indices.push_back(int32_i_0_reg2mem_0_72); + ptr_arraydecay_74_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay_74 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_74_indices.begin(), ptr_arraydecay_74_indices.end(), "arraydecay", label_forbody_68); + GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_72, "arrayidx6", label_forbody_68); + LoadInst* packed_tmp7_75 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_68); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_75, const_int32_34, "tmp11", label_forbody_68); + StoreInst* void_76 = new StoreInst(float_tmp11, ptr_arraydecay_74, false, label_forbody_68); std::vector ptr_arrayidx13_indices; - ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_70); - ptr_arrayidx13_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_66); - ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_73, const_int32_34, "tmp15", label_forbody_66); - StoreInst* void_75 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_66); + ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_72); + ptr_arrayidx13_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_68); + ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_75, const_int32_36, "tmp15", label_forbody_68); + StoreInst* void_77 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_68); std::vector ptr_arrayidx17_indices; - ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_70); - ptr_arrayidx17_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_66); - ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_73, const_int32_35, "tmp19", label_forbody_66); - StoreInst* void_76 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_66); + ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_72); + ptr_arrayidx17_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_68); + ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_75, const_int32_37, "tmp19", label_forbody_68); + StoreInst* void_78 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_68); std::vector ptr_arrayidx21_indices; - ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_70); - ptr_arrayidx21_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_66); - ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_73, const_int32_36, "tmp23", label_forbody_66); - StoreInst* void_77 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_66); - BinaryOperator* int32_indvar_next_78 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_70, const_int32_34, "indvar.next", label_forbody_66); - ICmpInst* int1_exitcond_79 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_78, int32_num_attribs_64, "exitcond", label_forbody_66); - new BranchInst(label_afterfor_67, label_forbody_66, int1_exitcond_79, label_forbody_66); - - // Block afterfor (label_afterfor_67) - new ReturnInst(label_afterfor_67); + ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_72); + ptr_arrayidx21_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_68); + ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_75, const_int32_38, "tmp23", label_forbody_68); + StoreInst* void_79 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_68); + BinaryOperator* int32_indvar_next_80 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_72, const_int32_36, "indvar.next", label_forbody_68); + ICmpInst* int1_exitcond_81 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_80, int32_num_attribs_66, "exitcond", label_forbody_68); + new BranchInst(label_afterfor_69, label_forbody_68, int1_exitcond_81, label_forbody_68); + + // Block afterfor (label_afterfor_69) + new ReturnInst(label_afterfor_69); // Resolve Forward References - fwdref_71->replaceAllUsesWith(int32_indvar_next_78); delete fwdref_71; + fwdref_73->replaceAllUsesWith(int32_indvar_next_80); delete fwdref_73; } // Function: run_vertex_shader (func_run_vertex_shader) { Function::arg_iterator args = func_run_vertex_shader->arg_begin(); - Value* ptr_ainputs_82 = args++; - ptr_ainputs_82->setName("ainputs"); - Value* ptr_dests_83 = args++; - ptr_dests_83->setName("dests"); + Value* ptr_ainputs_84 = args++; + ptr_ainputs_84->setName("ainputs"); + Value* ptr_dests_85 = args++; + ptr_dests_85->setName("dests"); Value* ptr_aconsts = args++; ptr_aconsts->setName("aconsts"); Value* int32_num_vertices = args++; int32_num_vertices->setName("num_vertices"); Value* int32_num_inputs = args++; int32_num_inputs->setName("num_inputs"); - Value* int32_num_attribs_84 = args++; - int32_num_attribs_84->setName("num_attribs"); + Value* int32_num_attribs_86 = args++; + int32_num_attribs_86->setName("num_attribs"); Value* int32_num_consts = args++; int32_num_consts->setName("num_consts"); - BasicBlock* label_entry_85 = new BasicBlock("entry",func_run_vertex_shader,0); + BasicBlock* label_entry_87 = new BasicBlock("entry",func_run_vertex_shader,0); BasicBlock* label_forbody6_i = new BasicBlock("forbody6.i",func_run_vertex_shader,0); BasicBlock* label_forinc57_i = new BasicBlock("forinc57.i",func_run_vertex_shader,0); BasicBlock* label_from_array_exit = new BasicBlock("from_array.exit",func_run_vertex_shader,0); BasicBlock* label_forbody_i15 = new BasicBlock("forbody.i15",func_run_vertex_shader,0); - BasicBlock* label_forcond = new BasicBlock("forcond",func_run_vertex_shader,0); + BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_vertex_shader,0); BasicBlock* label_forbody_preheader = new BasicBlock("forbody.preheader",func_run_vertex_shader,0); BasicBlock* label_forbody_us = new BasicBlock("forbody.us",func_run_vertex_shader,0); BasicBlock* label_to_array_exit_us = new BasicBlock("to_array.exit.us",func_run_vertex_shader,0); BasicBlock* label_forbody_i_us = new BasicBlock("forbody.i.us",func_run_vertex_shader,0); - BasicBlock* label_forbody_86 = new BasicBlock("forbody",func_run_vertex_shader,0); - BasicBlock* label_afterfor_87 = new BasicBlock("afterfor",func_run_vertex_shader,0); - - // Block entry (label_entry_85) - AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_21, "inputs", label_entry_85); - AllocaInst* ptr_consts = new AllocaInst(ArrayTy_23, "consts", label_entry_85); - AllocaInst* ptr_results = new AllocaInst(ArrayTy_21, "results", label_entry_85); - AllocaInst* ptr_temps = new AllocaInst(ArrayTy_25, "temps", label_entry_85); - ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_32, "cmp.i", label_entry_85); - ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_32, "cmp5.i", label_entry_85); - BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_85); - new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_85); + BasicBlock* label_forbody_88 = new BasicBlock("forbody",func_run_vertex_shader,0); + BasicBlock* label_afterfor_89 = new BasicBlock("afterfor",func_run_vertex_shader,0); + + // Block entry (label_entry_87) + AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_21, "inputs", label_entry_87); + AllocaInst* ptr_consts = new AllocaInst(ArrayTy_23, "consts", label_entry_87); + AllocaInst* ptr_results = new AllocaInst(ArrayTy_21, "results", label_entry_87); + AllocaInst* ptr_temps = new AllocaInst(ArrayTy_25, "temps", label_entry_87); + AllocaInst* ptr_args = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_87); + ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_34, "cmp.i", label_entry_87); + ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_34, "cmp5.i", label_entry_87); + BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_87); + new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_87); // Block forbody6.i (label_forbody6_i) - Argument* fwdref_89 = new Argument(IntegerType::get(32)); - Argument* fwdref_90 = new Argument(IntegerType::get(32)); + Argument* fwdref_91 = new Argument(IntegerType::get(32)); + Argument* fwdref_92 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i_ph = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i.ph", label_forbody6_i); int32_i_0_reg2mem_0_i_ph->reserveOperandSpace(3); - int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_32, label_entry_85); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_89, label_forinc57_i); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_90, label_forbody6_i); + int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_34, label_entry_87); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_91, label_forinc57_i); + int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_92, label_forbody6_i); - Argument* fwdref_91 = new Argument(IntegerType::get(32)); + Argument* fwdref_93 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i); int32_j_0_reg2mem_0_i->reserveOperandSpace(3); - int32_j_0_reg2mem_0_i->addIncoming(fwdref_91, label_forbody6_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_32, label_forinc57_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_32, label_entry_85); + int32_j_0_reg2mem_0_i->addIncoming(fwdref_93, label_forbody6_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_34, label_forinc57_i); + int32_j_0_reg2mem_0_i->addIncoming(const_int32_34, label_entry_87); - Argument* fwdref_92 = new Argument(VectorTy_13); - PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i", label_forbody6_i); + Argument* fwdref_94 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody6_i); packed_vec_0_reg2mem_0_i->reserveOperandSpace(3); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_92, label_forbody6_i); - packed_vec_0_reg2mem_0_i->addIncoming(const_packed_33, label_entry_85); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_92, label_forinc57_i); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_94, label_forbody6_i); + packed_vec_0_reg2mem_0_i->addIncoming(const_packed_35, label_entry_87); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_94, label_forinc57_i); std::vector ptr_arraydecay11_i_indices; ptr_arraydecay11_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arraydecay11_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arraydecay11_i_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_82, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); + ptr_arraydecay11_i_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_84, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); LoadInst* float_tmp13_i = new LoadInst(ptr_arraydecay11_i, "tmp13.i", false, label_forbody6_i); - InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_32, "tmp15.i", label_forbody6_i); + InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_34, "tmp15.i", label_forbody6_i); std::vector ptr_arrayidx23_i_indices; ptr_arrayidx23_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx23_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx23_i_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_82, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); + ptr_arrayidx23_i_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_84, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); LoadInst* float_tmp24_i = new LoadInst(ptr_arrayidx23_i, "tmp24.i", false, label_forbody6_i); - InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_34, "tmp26.i", label_forbody6_i); + InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_36, "tmp26.i", label_forbody6_i); std::vector ptr_arrayidx34_i_indices; ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx34_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx34_i_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_82, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); + ptr_arrayidx34_i_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_84, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); LoadInst* float_tmp35_i = new LoadInst(ptr_arrayidx34_i, "tmp35.i", false, label_forbody6_i); - InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_35, "tmp37.i", label_forbody6_i); + InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_37, "tmp37.i", label_forbody6_i); std::vector ptr_arrayidx45_i_indices; ptr_arrayidx45_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx45_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx45_i_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_82, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); + ptr_arrayidx45_i_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_84, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); LoadInst* float_tmp46_i = new LoadInst(ptr_arrayidx45_i, "tmp46.i", false, label_forbody6_i); - InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_36, "tmp48.i", label_forbody6_i); + InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_38, "tmp48.i", label_forbody6_i); std::vector ptr_arrayidx54_i_indices; - ptr_arrayidx54_i_indices.push_back(const_int32_32); + ptr_arrayidx54_i_indices.push_back(const_int32_34); ptr_arrayidx54_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); ptr_arrayidx54_i_indices.push_back(int32_j_0_reg2mem_0_i); Instruction* ptr_arrayidx54_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx54_i_indices.begin(), ptr_arrayidx54_i_indices.end(), "arrayidx54.i", label_forbody6_i); - StoreInst* void_93 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); - BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_34, "inc.i", label_forbody6_i); + StoreInst* void_95 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); + BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_36, "inc.i", label_forbody6_i); ICmpInst* int1_cmp59_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_inputs, "cmp59.i", label_forbody6_i); new BranchInst(label_forbody6_i, label_forinc57_i, int1_cmp59_i, label_forbody6_i); // Block forinc57.i (label_forinc57_i) - BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_34, "inc59.i", label_forinc57_i); + BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_36, "inc59.i", label_forinc57_i); ICmpInst* int1_cmp17_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59_i, int32_num_vertices, "cmp17.i", label_forinc57_i); new BranchInst(label_forbody6_i, label_from_array_exit, int1_cmp17_i, label_forinc57_i); // Block from_array.exit (label_from_array_exit) - ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_32, "cmp.i4", label_from_array_exit); - new BranchInst(label_forbody_i15, label_forcond, int1_cmp_i4, label_from_array_exit); + ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_34, "cmp.i4", label_from_array_exit); + new BranchInst(label_forbody_i15, label_from_consts_exit, int1_cmp_i4, label_from_array_exit); // Block forbody.i15 (label_forbody_i15) - Argument* fwdref_97 = new Argument(IntegerType::get(32)); + Argument* fwdref_99 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i5 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i5", label_forbody_i15); int32_i_0_reg2mem_0_i5->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i5->addIncoming(const_int32_32, label_from_array_exit); - int32_i_0_reg2mem_0_i5->addIncoming(fwdref_97, label_forbody_i15); + int32_i_0_reg2mem_0_i5->addIncoming(const_int32_34, label_from_array_exit); + int32_i_0_reg2mem_0_i5->addIncoming(fwdref_99, label_forbody_i15); - Argument* fwdref_98 = new Argument(VectorTy_13); - PHINode* packed_vec_0_reg2mem_0_i6 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i6", label_forbody_i15); + Argument* fwdref_100 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i6 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i6", label_forbody_i15); packed_vec_0_reg2mem_0_i6->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_33, label_from_array_exit); - packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_98, label_forbody_i15); + packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_35, label_from_array_exit); + packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_100, label_forbody_i15); std::vector ptr_arraydecay_i7_indices; ptr_arraydecay_i7_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arraydecay_i7_indices.push_back(const_int32_32); + ptr_arraydecay_i7_indices.push_back(const_int32_34); Instruction* ptr_arraydecay_i7 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i7_indices.begin(), ptr_arraydecay_i7_indices.end(), "arraydecay.i7", label_forbody_i15); LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i7, "tmp5.i", false, label_forbody_i15); - InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_32, "tmp7.i8", label_forbody_i15); + InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_34, "tmp7.i8", label_forbody_i15); std::vector ptr_arrayidx12_i_indices; ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx12_i_indices.push_back(const_int32_34); + ptr_arrayidx12_i_indices.push_back(const_int32_36); Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i15); LoadInst* float_tmp13_i9 = new LoadInst(ptr_arrayidx12_i, "tmp13.i9", false, label_forbody_i15); - InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_34, "tmp15.i10", label_forbody_i15); + InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_36, "tmp15.i10", label_forbody_i15); std::vector ptr_arrayidx20_i_indices; ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx20_i_indices.push_back(const_int32_35); + ptr_arrayidx20_i_indices.push_back(const_int32_37); Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i15); LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i15); - InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_35, "tmp23.i11", label_forbody_i15); + InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_37, "tmp23.i11", label_forbody_i15); std::vector ptr_arrayidx28_i_indices; ptr_arrayidx28_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx28_i_indices.push_back(const_int32_36); + ptr_arrayidx28_i_indices.push_back(const_int32_38); Instruction* ptr_arrayidx28_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i_indices.begin(), ptr_arrayidx28_i_indices.end(), "arrayidx28.i", label_forbody_i15); LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i15); - InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_36, "tmp31.i", label_forbody_i15); + InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_38, "tmp31.i", label_forbody_i15); std::vector ptr_arrayidx34_i12_indices; - ptr_arrayidx34_i12_indices.push_back(const_int32_32); + ptr_arrayidx34_i12_indices.push_back(const_int32_34); ptr_arrayidx34_i12_indices.push_back(int32_i_0_reg2mem_0_i5); Instruction* ptr_arrayidx34_i12 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i12_indices.begin(), ptr_arrayidx34_i12_indices.end(), "arrayidx34.i12", label_forbody_i15); - StoreInst* void_99 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); - BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_34, "indvar.next24", label_forbody_i15); + StoreInst* void_101 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); + BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_36, "indvar.next24", label_forbody_i15); ICmpInst* int1_exitcond25 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next24, int32_num_consts, "exitcond25", label_forbody_i15); - new BranchInst(label_forcond, label_forbody_i15, int1_exitcond25, label_forbody_i15); - - // Block forcond (label_forcond) - new BranchInst(label_forbody_preheader, label_afterfor_87, int1_cmp_i, label_forcond); + new BranchInst(label_from_consts_exit, label_forbody_i15, int1_exitcond25, label_forbody_i15); + + // Block from_consts.exit (label_from_consts_exit) + std::vector ptr_tmp6_indices; + ptr_tmp6_indices.push_back(const_int32_34); + ptr_tmp6_indices.push_back(const_int32_38); + Instruction* ptr_tmp6 = new GetElementPtrInst(ptr_args, ptr_tmp6_indices.begin(), ptr_tmp6_indices.end(), "tmp6", label_from_consts_exit); + std::vector ptr_arraydecay7_indices; + ptr_arraydecay7_indices.push_back(const_int32_34); + ptr_arraydecay7_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay7 = new GetElementPtrInst(ptr_consts, ptr_arraydecay7_indices.begin(), ptr_arraydecay7_indices.end(), "arraydecay7", label_from_consts_exit); + StoreInst* void_103 = new StoreInst(ptr_arraydecay7, ptr_tmp6, false, label_from_consts_exit); + std::vector ptr_tmp8_indices; + ptr_tmp8_indices.push_back(const_int32_34); + ptr_tmp8_indices.push_back(const_int32_37); + Instruction* ptr_tmp8 = new GetElementPtrInst(ptr_args, ptr_tmp8_indices.begin(), ptr_tmp8_indices.end(), "tmp8", label_from_consts_exit); + std::vector ptr_arraydecay9_indices; + ptr_arraydecay9_indices.push_back(const_int32_34); + ptr_arraydecay9_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay9 = new GetElementPtrInst(ptr_temps, ptr_arraydecay9_indices.begin(), ptr_arraydecay9_indices.end(), "arraydecay9", label_from_consts_exit); + StoreInst* void_104 = new StoreInst(ptr_arraydecay9, ptr_tmp8, false, label_from_consts_exit); + new BranchInst(label_forbody_preheader, label_afterfor_89, int1_cmp_i, label_from_consts_exit); // Block forbody.preheader (label_forbody_preheader) - std::vector ptr_arraydecay17_indices; - ptr_arraydecay17_indices.push_back(const_int32_32); - ptr_arraydecay17_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay17 = new GetElementPtrInst(ptr_consts, ptr_arraydecay17_indices.begin(), ptr_arraydecay17_indices.end(), "arraydecay17", label_forbody_preheader); - std::vector ptr_arraydecay18_indices; - ptr_arraydecay18_indices.push_back(const_int32_32); - ptr_arraydecay18_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay18 = new GetElementPtrInst(ptr_temps, ptr_arraydecay18_indices.begin(), ptr_arraydecay18_indices.end(), "arraydecay18", label_forbody_preheader); - ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_84, const_int32_32, "cmp.i1", label_forbody_preheader); - new BranchInst(label_forbody_us, label_forbody_86, int1_cmp_i1, label_forbody_preheader); + std::vector ptr_tmp12_indices; + ptr_tmp12_indices.push_back(const_int32_34); + ptr_tmp12_indices.push_back(const_int32_34); + Instruction* ptr_tmp12 = new GetElementPtrInst(ptr_args, ptr_tmp12_indices.begin(), ptr_tmp12_indices.end(), "tmp12", label_forbody_preheader); + std::vector ptr_tmp16_indices; + ptr_tmp16_indices.push_back(const_int32_34); + ptr_tmp16_indices.push_back(const_int32_36); + Instruction* ptr_tmp16 = new GetElementPtrInst(ptr_args, ptr_tmp16_indices.begin(), ptr_tmp16_indices.end(), "tmp16", label_forbody_preheader); + ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_86, const_int32_34, "cmp.i1", label_forbody_preheader); + new BranchInst(label_forbody_us, label_forbody_88, int1_cmp_i1, label_forbody_preheader); // Block forbody.us (label_forbody_us) - Argument* fwdref_103 = new Argument(IntegerType::get(32)); + Argument* fwdref_107 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.us", label_forbody_us); int32_i_0_reg2mem_0_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_us->addIncoming(const_int32_32, label_forbody_preheader); - int32_i_0_reg2mem_0_us->addIncoming(fwdref_103, label_to_array_exit_us); - - std::vector ptr_arraydecay10_us_indices; - ptr_arraydecay10_us_indices.push_back(const_int32_32); - ptr_arraydecay10_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay10_us_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay10_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_us_indices.begin(), ptr_arraydecay10_us_indices.end(), "arraydecay10.us", label_forbody_us); - std::vector ptr_arraydecay14_us_indices; - ptr_arraydecay14_us_indices.push_back(const_int32_32); - ptr_arraydecay14_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay14_us_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay14_us = new GetElementPtrInst(ptr_results, ptr_arraydecay14_us_indices.begin(), ptr_arraydecay14_us_indices.end(), "arraydecay14.us", label_forbody_us); - std::vector void_104_params; - void_104_params.push_back(ptr_arraydecay14_us); - void_104_params.push_back(ptr_arraydecay10_us); - void_104_params.push_back(ptr_arraydecay17); - void_104_params.push_back(ptr_arraydecay18); - CallInst* void_104 = new CallInst(func_execute_shader, void_104_params.begin(), void_104_params.end(), "", label_forbody_us); - void_104->setCallingConv(CallingConv::C); - void_104->setTailCall(false); + int32_i_0_reg2mem_0_us->addIncoming(const_int32_34, label_forbody_preheader); + int32_i_0_reg2mem_0_us->addIncoming(fwdref_107, label_to_array_exit_us); + + std::vector ptr_arraydecay15_us_indices; + ptr_arraydecay15_us_indices.push_back(const_int32_34); + ptr_arraydecay15_us_indices.push_back(int32_i_0_reg2mem_0_us); + ptr_arraydecay15_us_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay15_us = new GetElementPtrInst(ptr_results, ptr_arraydecay15_us_indices.begin(), ptr_arraydecay15_us_indices.end(), "arraydecay15.us", label_forbody_us); + StoreInst* void_108 = new StoreInst(ptr_arraydecay15_us, ptr_tmp12, false, label_forbody_us); + std::vector ptr_arraydecay20_us_indices; + ptr_arraydecay20_us_indices.push_back(const_int32_34); + ptr_arraydecay20_us_indices.push_back(int32_i_0_reg2mem_0_us); + ptr_arraydecay20_us_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay20_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay20_us_indices.begin(), ptr_arraydecay20_us_indices.end(), "arraydecay20.us", label_forbody_us); + StoreInst* void_109 = new StoreInst(ptr_arraydecay20_us, ptr_tmp16, false, label_forbody_us); + CallInst* void_110 = new CallInst(func_execute_shader, ptr_args, "", label_forbody_us); + void_110->setCallingConv(CallingConv::C); + void_110->setTailCall(false); + LoadInst* ptr_tmp26_us = new LoadInst(ptr_tmp12, "tmp26.us", false, label_forbody_us); new BranchInst(label_forbody_i_us, label_forbody_us); // Block to_array.exit.us (label_to_array_exit_us) - BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_34, "inc.us", label_to_array_exit_us); + BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_36, "inc.us", label_to_array_exit_us); ICmpInst* int1_cmp21_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_num_vertices, "cmp21.us", label_to_array_exit_us); - new BranchInst(label_forbody_us, label_afterfor_87, int1_cmp21_us, label_to_array_exit_us); + new BranchInst(label_forbody_us, label_afterfor_89, int1_cmp21_us, label_to_array_exit_us); // Block forbody.i.us (label_forbody_i_us) - Argument* fwdref_107 = new Argument(IntegerType::get(32)); + Argument* fwdref_113 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i2_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i2.us", label_forbody_i_us); int32_i_0_reg2mem_0_i2_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_32, label_forbody_us); - int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_107, label_forbody_i_us); + int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_34, label_forbody_us); + int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_113, label_forbody_i_us); std::vector ptr_arraydecay_i_us_indices; ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arraydecay_i_us_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_83, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); - std::vector ptr_arrayidx6_i_us_indices; - ptr_arrayidx6_i_us_indices.push_back(const_int32_32); - ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arrayidx6_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - Instruction* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_results, ptr_arrayidx6_i_us_indices.begin(), ptr_arrayidx6_i_us_indices.end(), "arrayidx6.i.us", label_forbody_i_us); + ptr_arraydecay_i_us_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_85, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); + GetElementPtrInst* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_tmp26_us, int32_i_0_reg2mem_0_i2_us, "arrayidx6.i.us", label_forbody_i_us); LoadInst* packed_tmp7_i_us = new LoadInst(ptr_arrayidx6_i_us, "tmp7.i.us", false, label_forbody_i_us); - ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_32, "tmp11.i.us", label_forbody_i_us); - StoreInst* void_108 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); + ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_34, "tmp11.i.us", label_forbody_i_us); + StoreInst* void_114 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx13_i_us_indices; ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx13_i_us_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_83, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_34, "tmp15.i3.us", label_forbody_i_us); - StoreInst* void_109 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); + ptr_arrayidx13_i_us_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_85, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_36, "tmp15.i3.us", label_forbody_i_us); + StoreInst* void_115 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx17_i_us_indices; ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx17_i_us_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_83, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_35, "tmp19.i.us", label_forbody_i_us); - StoreInst* void_110 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); + ptr_arrayidx17_i_us_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_85, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_37, "tmp19.i.us", label_forbody_i_us); + StoreInst* void_116 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); std::vector ptr_arrayidx21_i_us_indices; ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_us); ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx21_i_us_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_83, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_36, "tmp23.i.us", label_forbody_i_us); - StoreInst* void_111 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); - BinaryOperator* int32_indvar_next_112 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_34, "indvar.next", label_forbody_i_us); - ICmpInst* int1_exitcond_113 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_112, int32_num_attribs_84, "exitcond", label_forbody_i_us); - new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_113, label_forbody_i_us); - - // Block forbody (label_forbody_86) - Argument* fwdref_116 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_115 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_86); - int32_i_0_reg2mem_0_115->reserveOperandSpace(2); - int32_i_0_reg2mem_0_115->addIncoming(const_int32_32, label_forbody_preheader); - int32_i_0_reg2mem_0_115->addIncoming(fwdref_116, label_forbody_86); - - std::vector ptr_arraydecay10_indices; - ptr_arraydecay10_indices.push_back(const_int32_32); - ptr_arraydecay10_indices.push_back(int32_i_0_reg2mem_0_115); - ptr_arraydecay10_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_forbody_86); - std::vector ptr_arraydecay14_indices; - ptr_arraydecay14_indices.push_back(const_int32_32); - ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_115); - ptr_arraydecay14_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_results, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_86); - std::vector void_117_params; - void_117_params.push_back(ptr_arraydecay14); - void_117_params.push_back(ptr_arraydecay10); - void_117_params.push_back(ptr_arraydecay17); - void_117_params.push_back(ptr_arraydecay18); - CallInst* void_117 = new CallInst(func_execute_shader, void_117_params.begin(), void_117_params.end(), "", label_forbody_86); - void_117->setCallingConv(CallingConv::C); - void_117->setTailCall(false); - BinaryOperator* int32_inc_118 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_115, const_int32_34, "inc", label_forbody_86); - ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_118, int32_num_vertices, "cmp21", label_forbody_86); - new BranchInst(label_forbody_86, label_afterfor_87, int1_cmp21, label_forbody_86); - - // Block afterfor (label_afterfor_87) - new ReturnInst(label_afterfor_87); + ptr_arrayidx21_i_us_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_85, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); + ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_38, "tmp23.i.us", label_forbody_i_us); + StoreInst* void_117 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); + BinaryOperator* int32_indvar_next_118 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_36, "indvar.next", label_forbody_i_us); + ICmpInst* int1_exitcond_119 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_118, int32_num_attribs_86, "exitcond", label_forbody_i_us); + new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_119, label_forbody_i_us); + + // Block forbody (label_forbody_88) + Argument* fwdref_122 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_121 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_88); + int32_i_0_reg2mem_0_121->reserveOperandSpace(2); + int32_i_0_reg2mem_0_121->addIncoming(const_int32_34, label_forbody_preheader); + int32_i_0_reg2mem_0_121->addIncoming(fwdref_122, label_forbody_88); + + std::vector ptr_arraydecay15_indices; + ptr_arraydecay15_indices.push_back(const_int32_34); + ptr_arraydecay15_indices.push_back(int32_i_0_reg2mem_0_121); + ptr_arraydecay15_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay15 = new GetElementPtrInst(ptr_results, ptr_arraydecay15_indices.begin(), ptr_arraydecay15_indices.end(), "arraydecay15", label_forbody_88); + StoreInst* void_123 = new StoreInst(ptr_arraydecay15, ptr_tmp12, false, label_forbody_88); + std::vector ptr_arraydecay20_indices; + ptr_arraydecay20_indices.push_back(const_int32_34); + ptr_arraydecay20_indices.push_back(int32_i_0_reg2mem_0_121); + ptr_arraydecay20_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay20 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay20_indices.begin(), ptr_arraydecay20_indices.end(), "arraydecay20", label_forbody_88); + StoreInst* void_124 = new StoreInst(ptr_arraydecay20, ptr_tmp16, false, label_forbody_88); + CallInst* void_125 = new CallInst(func_execute_shader, ptr_args, "", label_forbody_88); + void_125->setCallingConv(CallingConv::C); + void_125->setTailCall(false); + BinaryOperator* int32_inc_126 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_121, const_int32_36, "inc", label_forbody_88); + ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_126, int32_num_vertices, "cmp21", label_forbody_88); + new BranchInst(label_forbody_88, label_afterfor_89, int1_cmp21, label_forbody_88); + + // Block afterfor (label_afterfor_89) + new ReturnInst(label_afterfor_89); // Resolve Forward References - fwdref_90->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_90; - fwdref_92->replaceAllUsesWith(packed_tmp48_i); delete fwdref_92; - fwdref_91->replaceAllUsesWith(int32_inc_i); delete fwdref_91; - fwdref_89->replaceAllUsesWith(int32_inc59_i); delete fwdref_89; - fwdref_98->replaceAllUsesWith(packed_tmp31_i); delete fwdref_98; - fwdref_97->replaceAllUsesWith(int32_indvar_next24); delete fwdref_97; - fwdref_103->replaceAllUsesWith(int32_inc_us); delete fwdref_103; - fwdref_107->replaceAllUsesWith(int32_indvar_next_112); delete fwdref_107; - fwdref_116->replaceAllUsesWith(int32_inc_118); delete fwdref_116; + fwdref_107->replaceAllUsesWith(int32_inc_us); delete fwdref_107; + fwdref_92->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_92; + fwdref_94->replaceAllUsesWith(packed_tmp48_i); delete fwdref_94; + fwdref_93->replaceAllUsesWith(int32_inc_i); delete fwdref_93; + fwdref_91->replaceAllUsesWith(int32_inc59_i); delete fwdref_91; + fwdref_100->replaceAllUsesWith(packed_tmp31_i); delete fwdref_100; + fwdref_99->replaceAllUsesWith(int32_indvar_next24); delete fwdref_99; + fwdref_113->replaceAllUsesWith(int32_indvar_next_118); delete fwdref_113; + fwdref_122->replaceAllUsesWith(int32_inc_126); delete fwdref_122; } @@ -758,410 +780,429 @@ Module* createBaseShader() { float_x->setName("x"); Value* float_y = args++; float_y->setName("y"); - Value* ptr_dests_121 = args++; - ptr_dests_121->setName("dests"); - Value* ptr_ainputs_122 = args++; - ptr_ainputs_122->setName("ainputs"); - Value* int32_num_inputs_123 = args++; - int32_num_inputs_123->setName("num_inputs"); - Value* ptr_aconsts_124 = args++; - ptr_aconsts_124->setName("aconsts"); - Value* int32_num_consts_125 = args++; - int32_num_consts_125->setName("num_consts"); + Value* ptr_dests_129 = args++; + ptr_dests_129->setName("dests"); + Value* ptr_ainputs_130 = args++; + ptr_ainputs_130->setName("ainputs"); + Value* int32_num_inputs_131 = args++; + int32_num_inputs_131->setName("num_inputs"); + Value* ptr_aconsts_132 = args++; + ptr_aconsts_132->setName("aconsts"); + Value* int32_num_consts_133 = args++; + int32_num_consts_133->setName("num_consts"); Value* ptr_samplers = args++; ptr_samplers->setName("samplers"); - BasicBlock* label_entry_126 = new BasicBlock("entry",func_run_fragment_shader,0); - BasicBlock* label_forbody6_i_127 = new BasicBlock("forbody6.i",func_run_fragment_shader,0); - BasicBlock* label_from_array_exit_128 = new BasicBlock("from_array.exit",func_run_fragment_shader,0); + BasicBlock* label_entry_134 = new BasicBlock("entry",func_run_fragment_shader,0); + BasicBlock* label_forbody6_i_135 = new BasicBlock("forbody6.i",func_run_fragment_shader,0); + BasicBlock* label_from_array_exit_136 = new BasicBlock("from_array.exit",func_run_fragment_shader,0); BasicBlock* label_forbody_i13 = new BasicBlock("forbody.i13",func_run_fragment_shader,0); - BasicBlock* label_forbody_preheader_129 = new BasicBlock("forbody.preheader",func_run_fragment_shader,0); - BasicBlock* label_forbody_130 = new BasicBlock("forbody",func_run_fragment_shader,0); - BasicBlock* label_afterfor_131 = new BasicBlock("afterfor",func_run_fragment_shader,0); + BasicBlock* label_from_consts_exit_137 = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); + BasicBlock* label_forbody_138 = new BasicBlock("forbody",func_run_fragment_shader,0); + BasicBlock* label_afterfor_139 = new BasicBlock("afterfor",func_run_fragment_shader,0); BasicBlock* label_forbody6_i_1 = new BasicBlock("forbody6.i.1",func_run_fragment_shader,0); BasicBlock* label_forbody6_i_2 = new BasicBlock("forbody6.i.2",func_run_fragment_shader,0); BasicBlock* label_forbody6_i_3 = new BasicBlock("forbody6.i.3",func_run_fragment_shader,0); - // Block entry (label_entry_126) - AllocaInst* ptr_inputs_132 = new AllocaInst(ArrayTy_31, "inputs", label_entry_126); - AllocaInst* ptr_consts_133 = new AllocaInst(ArrayTy_23, "consts", label_entry_126); - AllocaInst* ptr_results_134 = new AllocaInst(ArrayTy_31, "results", label_entry_126); - AllocaInst* ptr_temps_135 = new AllocaInst(ArrayTy_25, "temps", label_entry_126); - ICmpInst* int1_cmp5_i_136 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs_123, const_int32_32, "cmp5.i", label_entry_126); - new BranchInst(label_forbody6_i_127, label_from_array_exit_128, int1_cmp5_i_136, label_entry_126); - - // Block forbody6.i (label_forbody6_i_127) - Argument* fwdref_139 = new Argument(IntegerType::get(32)); - PHINode* int32_j_0_reg2mem_0_i_138 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i_127); - int32_j_0_reg2mem_0_i_138->reserveOperandSpace(2); - int32_j_0_reg2mem_0_i_138->addIncoming(const_int32_32, label_entry_126); - int32_j_0_reg2mem_0_i_138->addIncoming(fwdref_139, label_forbody6_i_127); - - Argument* fwdref_141 = new Argument(VectorTy_13); - PHINode* packed_vec_0_reg2mem_0_i_140 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i", label_forbody6_i_127); - packed_vec_0_reg2mem_0_i_140->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_140->addIncoming(const_packed_33, label_entry_126); - packed_vec_0_reg2mem_0_i_140->addIncoming(fwdref_141, label_forbody6_i_127); - - std::vector ptr_arraydecay11_i_142_indices; - ptr_arraydecay11_i_142_indices.push_back(const_int32_32); - ptr_arraydecay11_i_142_indices.push_back(int32_j_0_reg2mem_0_i_138); - ptr_arraydecay11_i_142_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay11_i_142 = new GetElementPtrInst(ptr_ainputs_122, ptr_arraydecay11_i_142_indices.begin(), ptr_arraydecay11_i_142_indices.end(), "arraydecay11.i", label_forbody6_i_127); - LoadInst* float_tmp13_i_143 = new LoadInst(ptr_arraydecay11_i_142, "tmp13.i", false, label_forbody6_i_127); - InsertElementInst* packed_tmp15_i_144 = new InsertElementInst(packed_vec_0_reg2mem_0_i_140, float_tmp13_i_143, const_int32_32, "tmp15.i", label_forbody6_i_127); - std::vector ptr_arrayidx23_i_145_indices; - ptr_arrayidx23_i_145_indices.push_back(const_int32_32); - ptr_arrayidx23_i_145_indices.push_back(int32_j_0_reg2mem_0_i_138); - ptr_arrayidx23_i_145_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx23_i_145 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx23_i_145_indices.begin(), ptr_arrayidx23_i_145_indices.end(), "arrayidx23.i", label_forbody6_i_127); - LoadInst* float_tmp24_i_146 = new LoadInst(ptr_arrayidx23_i_145, "tmp24.i", false, label_forbody6_i_127); - InsertElementInst* packed_tmp26_i_147 = new InsertElementInst(packed_tmp15_i_144, float_tmp24_i_146, const_int32_34, "tmp26.i", label_forbody6_i_127); - std::vector ptr_arrayidx34_i_148_indices; - ptr_arrayidx34_i_148_indices.push_back(const_int32_32); - ptr_arrayidx34_i_148_indices.push_back(int32_j_0_reg2mem_0_i_138); - ptr_arrayidx34_i_148_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx34_i_148 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx34_i_148_indices.begin(), ptr_arrayidx34_i_148_indices.end(), "arrayidx34.i", label_forbody6_i_127); - LoadInst* float_tmp35_i_149 = new LoadInst(ptr_arrayidx34_i_148, "tmp35.i", false, label_forbody6_i_127); - InsertElementInst* packed_tmp37_i_150 = new InsertElementInst(packed_tmp26_i_147, float_tmp35_i_149, const_int32_35, "tmp37.i", label_forbody6_i_127); - std::vector ptr_arrayidx45_i_151_indices; - ptr_arrayidx45_i_151_indices.push_back(const_int32_32); - ptr_arrayidx45_i_151_indices.push_back(int32_j_0_reg2mem_0_i_138); - ptr_arrayidx45_i_151_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx45_i_151 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx45_i_151_indices.begin(), ptr_arrayidx45_i_151_indices.end(), "arrayidx45.i", label_forbody6_i_127); - LoadInst* float_tmp46_i_152 = new LoadInst(ptr_arrayidx45_i_151, "tmp46.i", false, label_forbody6_i_127); - InsertElementInst* packed_tmp48_i_153 = new InsertElementInst(packed_tmp37_i_150, float_tmp46_i_152, const_int32_36, "tmp48.i", label_forbody6_i_127); - std::vector ptr_arrayidx54_i_154_indices; - ptr_arrayidx54_i_154_indices.push_back(const_int32_32); - ptr_arrayidx54_i_154_indices.push_back(const_int32_32); - ptr_arrayidx54_i_154_indices.push_back(int32_j_0_reg2mem_0_i_138); - Instruction* ptr_arrayidx54_i_154 = new GetElementPtrInst(ptr_inputs_132, ptr_arrayidx54_i_154_indices.begin(), ptr_arrayidx54_i_154_indices.end(), "arrayidx54.i", label_forbody6_i_127); - StoreInst* void_155 = new StoreInst(packed_tmp48_i_153, ptr_arrayidx54_i_154, false, label_forbody6_i_127); - BinaryOperator* int32_inc_i_156 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_138, const_int32_34, "inc.i", label_forbody6_i_127); - ICmpInst* int1_cmp59_i_157 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_156, int32_num_inputs_123, "cmp59.i", label_forbody6_i_127); - new BranchInst(label_forbody6_i_127, label_forbody6_i_1, int1_cmp59_i_157, label_forbody6_i_127); - - // Block from_array.exit (label_from_array_exit_128) - ICmpInst* int1_cmp_i_159 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_125, const_int32_32, "cmp.i", label_from_array_exit_128); - new BranchInst(label_forbody_i13, label_forbody_preheader_129, int1_cmp_i_159, label_from_array_exit_128); + // Block entry (label_entry_134) + AllocaInst* ptr_inputs_140 = new AllocaInst(ArrayTy_33, "inputs", label_entry_134); + AllocaInst* ptr_consts_141 = new AllocaInst(ArrayTy_23, "consts", label_entry_134); + AllocaInst* ptr_results_142 = new AllocaInst(ArrayTy_33, "results", label_entry_134); + AllocaInst* ptr_temps_143 = new AllocaInst(ArrayTy_25, "temps", label_entry_134); + AllocaInst* ptr_args_144 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_134); + ICmpInst* int1_cmp5_i_145 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs_131, const_int32_34, "cmp5.i", label_entry_134); + new BranchInst(label_forbody6_i_135, label_from_array_exit_136, int1_cmp5_i_145, label_entry_134); + + // Block forbody6.i (label_forbody6_i_135) + Argument* fwdref_148 = new Argument(IntegerType::get(32)); + PHINode* int32_j_0_reg2mem_0_i_147 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i_135); + int32_j_0_reg2mem_0_i_147->reserveOperandSpace(2); + int32_j_0_reg2mem_0_i_147->addIncoming(const_int32_34, label_entry_134); + int32_j_0_reg2mem_0_i_147->addIncoming(fwdref_148, label_forbody6_i_135); + + Argument* fwdref_150 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i_149 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody6_i_135); + packed_vec_0_reg2mem_0_i_149->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_149->addIncoming(const_packed_35, label_entry_134); + packed_vec_0_reg2mem_0_i_149->addIncoming(fwdref_150, label_forbody6_i_135); + + std::vector ptr_arraydecay11_i_151_indices; + ptr_arraydecay11_i_151_indices.push_back(const_int32_34); + ptr_arraydecay11_i_151_indices.push_back(int32_j_0_reg2mem_0_i_147); + ptr_arraydecay11_i_151_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay11_i_151 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_151_indices.begin(), ptr_arraydecay11_i_151_indices.end(), "arraydecay11.i", label_forbody6_i_135); + LoadInst* float_tmp13_i_152 = new LoadInst(ptr_arraydecay11_i_151, "tmp13.i", false, label_forbody6_i_135); + InsertElementInst* packed_tmp15_i_153 = new InsertElementInst(packed_vec_0_reg2mem_0_i_149, float_tmp13_i_152, const_int32_34, "tmp15.i", label_forbody6_i_135); + std::vector ptr_arrayidx23_i_154_indices; + ptr_arrayidx23_i_154_indices.push_back(const_int32_34); + ptr_arrayidx23_i_154_indices.push_back(int32_j_0_reg2mem_0_i_147); + ptr_arrayidx23_i_154_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx23_i_154 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_154_indices.begin(), ptr_arrayidx23_i_154_indices.end(), "arrayidx23.i", label_forbody6_i_135); + LoadInst* float_tmp24_i_155 = new LoadInst(ptr_arrayidx23_i_154, "tmp24.i", false, label_forbody6_i_135); + InsertElementInst* packed_tmp26_i_156 = new InsertElementInst(packed_tmp15_i_153, float_tmp24_i_155, const_int32_36, "tmp26.i", label_forbody6_i_135); + std::vector ptr_arrayidx34_i_157_indices; + ptr_arrayidx34_i_157_indices.push_back(const_int32_34); + ptr_arrayidx34_i_157_indices.push_back(int32_j_0_reg2mem_0_i_147); + ptr_arrayidx34_i_157_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx34_i_157 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_157_indices.begin(), ptr_arrayidx34_i_157_indices.end(), "arrayidx34.i", label_forbody6_i_135); + LoadInst* float_tmp35_i_158 = new LoadInst(ptr_arrayidx34_i_157, "tmp35.i", false, label_forbody6_i_135); + InsertElementInst* packed_tmp37_i_159 = new InsertElementInst(packed_tmp26_i_156, float_tmp35_i_158, const_int32_37, "tmp37.i", label_forbody6_i_135); + std::vector ptr_arrayidx45_i_160_indices; + ptr_arrayidx45_i_160_indices.push_back(const_int32_34); + ptr_arrayidx45_i_160_indices.push_back(int32_j_0_reg2mem_0_i_147); + ptr_arrayidx45_i_160_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx45_i_160 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_160_indices.begin(), ptr_arrayidx45_i_160_indices.end(), "arrayidx45.i", label_forbody6_i_135); + LoadInst* float_tmp46_i_161 = new LoadInst(ptr_arrayidx45_i_160, "tmp46.i", false, label_forbody6_i_135); + InsertElementInst* packed_tmp48_i_162 = new InsertElementInst(packed_tmp37_i_159, float_tmp46_i_161, const_int32_38, "tmp48.i", label_forbody6_i_135); + std::vector ptr_arrayidx54_i_163_indices; + ptr_arrayidx54_i_163_indices.push_back(const_int32_34); + ptr_arrayidx54_i_163_indices.push_back(const_int32_34); + ptr_arrayidx54_i_163_indices.push_back(int32_j_0_reg2mem_0_i_147); + Instruction* ptr_arrayidx54_i_163 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_163_indices.begin(), ptr_arrayidx54_i_163_indices.end(), "arrayidx54.i", label_forbody6_i_135); + StoreInst* void_164 = new StoreInst(packed_tmp48_i_162, ptr_arrayidx54_i_163, false, label_forbody6_i_135); + BinaryOperator* int32_inc_i_165 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_147, const_int32_36, "inc.i", label_forbody6_i_135); + ICmpInst* int1_cmp59_i_166 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_165, int32_num_inputs_131, "cmp59.i", label_forbody6_i_135); + new BranchInst(label_forbody6_i_135, label_forbody6_i_1, int1_cmp59_i_166, label_forbody6_i_135); + + // Block from_array.exit (label_from_array_exit_136) + ICmpInst* int1_cmp_i_168 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_133, const_int32_34, "cmp.i", label_from_array_exit_136); + new BranchInst(label_forbody_i13, label_from_consts_exit_137, int1_cmp_i_168, label_from_array_exit_136); // Block forbody.i13 (label_forbody_i13) - Argument* fwdref_161 = new Argument(IntegerType::get(32)); + Argument* fwdref_170 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i3 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i3", label_forbody_i13); int32_i_0_reg2mem_0_i3->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i3->addIncoming(const_int32_32, label_from_array_exit_128); - int32_i_0_reg2mem_0_i3->addIncoming(fwdref_161, label_forbody_i13); + int32_i_0_reg2mem_0_i3->addIncoming(const_int32_34, label_from_array_exit_136); + int32_i_0_reg2mem_0_i3->addIncoming(fwdref_170, label_forbody_i13); - Argument* fwdref_162 = new Argument(VectorTy_13); - PHINode* packed_vec_0_reg2mem_0_i4 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i4", label_forbody_i13); + Argument* fwdref_171 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i4 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i4", label_forbody_i13); packed_vec_0_reg2mem_0_i4->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i4->addIncoming(const_packed_33, label_from_array_exit_128); - packed_vec_0_reg2mem_0_i4->addIncoming(fwdref_162, label_forbody_i13); + packed_vec_0_reg2mem_0_i4->addIncoming(const_packed_35, label_from_array_exit_136); + packed_vec_0_reg2mem_0_i4->addIncoming(fwdref_171, label_forbody_i13); std::vector ptr_arraydecay_i5_indices; ptr_arraydecay_i5_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arraydecay_i5_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay_i5 = new GetElementPtrInst(ptr_aconsts_124, ptr_arraydecay_i5_indices.begin(), ptr_arraydecay_i5_indices.end(), "arraydecay.i5", label_forbody_i13); - LoadInst* float_tmp5_i_163 = new LoadInst(ptr_arraydecay_i5, "tmp5.i", false, label_forbody_i13); - InsertElementInst* packed_tmp7_i6 = new InsertElementInst(packed_vec_0_reg2mem_0_i4, float_tmp5_i_163, const_int32_32, "tmp7.i6", label_forbody_i13); - std::vector ptr_arrayidx12_i_164_indices; - ptr_arrayidx12_i_164_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arrayidx12_i_164_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx12_i_164 = new GetElementPtrInst(ptr_aconsts_124, ptr_arrayidx12_i_164_indices.begin(), ptr_arrayidx12_i_164_indices.end(), "arrayidx12.i", label_forbody_i13); - LoadInst* float_tmp13_i7 = new LoadInst(ptr_arrayidx12_i_164, "tmp13.i7", false, label_forbody_i13); - InsertElementInst* packed_tmp15_i8 = new InsertElementInst(packed_tmp7_i6, float_tmp13_i7, const_int32_34, "tmp15.i8", label_forbody_i13); - std::vector ptr_arrayidx20_i_165_indices; - ptr_arrayidx20_i_165_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arrayidx20_i_165_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx20_i_165 = new GetElementPtrInst(ptr_aconsts_124, ptr_arrayidx20_i_165_indices.begin(), ptr_arrayidx20_i_165_indices.end(), "arrayidx20.i", label_forbody_i13); - LoadInst* float_tmp21_i_166 = new LoadInst(ptr_arrayidx20_i_165, "tmp21.i", false, label_forbody_i13); - InsertElementInst* packed_tmp23_i9 = new InsertElementInst(packed_tmp15_i8, float_tmp21_i_166, const_int32_35, "tmp23.i9", label_forbody_i13); - std::vector ptr_arrayidx28_i_167_indices; - ptr_arrayidx28_i_167_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arrayidx28_i_167_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx28_i_167 = new GetElementPtrInst(ptr_aconsts_124, ptr_arrayidx28_i_167_indices.begin(), ptr_arrayidx28_i_167_indices.end(), "arrayidx28.i", label_forbody_i13); - LoadInst* float_tmp29_i_168 = new LoadInst(ptr_arrayidx28_i_167, "tmp29.i", false, label_forbody_i13); - InsertElementInst* packed_tmp31_i_169 = new InsertElementInst(packed_tmp23_i9, float_tmp29_i_168, const_int32_36, "tmp31.i", label_forbody_i13); + ptr_arraydecay_i5_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay_i5 = new GetElementPtrInst(ptr_aconsts_132, ptr_arraydecay_i5_indices.begin(), ptr_arraydecay_i5_indices.end(), "arraydecay.i5", label_forbody_i13); + LoadInst* float_tmp5_i_172 = new LoadInst(ptr_arraydecay_i5, "tmp5.i", false, label_forbody_i13); + InsertElementInst* packed_tmp7_i6 = new InsertElementInst(packed_vec_0_reg2mem_0_i4, float_tmp5_i_172, const_int32_34, "tmp7.i6", label_forbody_i13); + std::vector ptr_arrayidx12_i_173_indices; + ptr_arrayidx12_i_173_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arrayidx12_i_173_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx12_i_173 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx12_i_173_indices.begin(), ptr_arrayidx12_i_173_indices.end(), "arrayidx12.i", label_forbody_i13); + LoadInst* float_tmp13_i7 = new LoadInst(ptr_arrayidx12_i_173, "tmp13.i7", false, label_forbody_i13); + InsertElementInst* packed_tmp15_i8 = new InsertElementInst(packed_tmp7_i6, float_tmp13_i7, const_int32_36, "tmp15.i8", label_forbody_i13); + std::vector ptr_arrayidx20_i_174_indices; + ptr_arrayidx20_i_174_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arrayidx20_i_174_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx20_i_174 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx20_i_174_indices.begin(), ptr_arrayidx20_i_174_indices.end(), "arrayidx20.i", label_forbody_i13); + LoadInst* float_tmp21_i_175 = new LoadInst(ptr_arrayidx20_i_174, "tmp21.i", false, label_forbody_i13); + InsertElementInst* packed_tmp23_i9 = new InsertElementInst(packed_tmp15_i8, float_tmp21_i_175, const_int32_37, "tmp23.i9", label_forbody_i13); + std::vector ptr_arrayidx28_i_176_indices; + ptr_arrayidx28_i_176_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arrayidx28_i_176_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx28_i_176 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx28_i_176_indices.begin(), ptr_arrayidx28_i_176_indices.end(), "arrayidx28.i", label_forbody_i13); + LoadInst* float_tmp29_i_177 = new LoadInst(ptr_arrayidx28_i_176, "tmp29.i", false, label_forbody_i13); + InsertElementInst* packed_tmp31_i_178 = new InsertElementInst(packed_tmp23_i9, float_tmp29_i_177, const_int32_38, "tmp31.i", label_forbody_i13); std::vector ptr_arrayidx34_i10_indices; - ptr_arrayidx34_i10_indices.push_back(const_int32_32); + ptr_arrayidx34_i10_indices.push_back(const_int32_34); ptr_arrayidx34_i10_indices.push_back(int32_i_0_reg2mem_0_i3); - Instruction* ptr_arrayidx34_i10 = new GetElementPtrInst(ptr_consts_133, ptr_arrayidx34_i10_indices.begin(), ptr_arrayidx34_i10_indices.end(), "arrayidx34.i10", label_forbody_i13); - StoreInst* void_170 = new StoreInst(packed_tmp31_i_169, ptr_arrayidx34_i10, false, label_forbody_i13); - BinaryOperator* int32_indvar_next22 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i3, const_int32_34, "indvar.next22", label_forbody_i13); - ICmpInst* int1_exitcond23 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next22, int32_num_consts_125, "exitcond23", label_forbody_i13); - new BranchInst(label_forbody_preheader_129, label_forbody_i13, int1_exitcond23, label_forbody_i13); - - // Block forbody.preheader (label_forbody_preheader_129) - std::vector ptr_arraydecay15_indices; - ptr_arraydecay15_indices.push_back(const_int32_32); - ptr_arraydecay15_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay15 = new GetElementPtrInst(ptr_consts_133, ptr_arraydecay15_indices.begin(), ptr_arraydecay15_indices.end(), "arraydecay15", label_forbody_preheader_129); - std::vector ptr_arraydecay16_indices; - ptr_arraydecay16_indices.push_back(const_int32_32); - ptr_arraydecay16_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay16 = new GetElementPtrInst(ptr_temps_135, ptr_arraydecay16_indices.begin(), ptr_arraydecay16_indices.end(), "arraydecay16", label_forbody_preheader_129); - new BranchInst(label_forbody_130, label_forbody_preheader_129); - - // Block forbody (label_forbody_130) - Argument* fwdref_174 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_173 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_130); - int32_i_0_reg2mem_0_173->reserveOperandSpace(2); - int32_i_0_reg2mem_0_173->addIncoming(const_int32_32, label_forbody_preheader_129); - int32_i_0_reg2mem_0_173->addIncoming(fwdref_174, label_forbody_130); - + Instruction* ptr_arrayidx34_i10 = new GetElementPtrInst(ptr_consts_141, ptr_arrayidx34_i10_indices.begin(), ptr_arrayidx34_i10_indices.end(), "arrayidx34.i10", label_forbody_i13); + StoreInst* void_179 = new StoreInst(packed_tmp31_i_178, ptr_arrayidx34_i10, false, label_forbody_i13); + BinaryOperator* int32_indvar_next22 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i3, const_int32_36, "indvar.next22", label_forbody_i13); + ICmpInst* int1_exitcond23 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next22, int32_num_consts_133, "exitcond23", label_forbody_i13); + new BranchInst(label_from_consts_exit_137, label_forbody_i13, int1_exitcond23, label_forbody_i13); + + // Block from_consts.exit (label_from_consts_exit_137) + std::vector ptr_tmp5_indices; + ptr_tmp5_indices.push_back(const_int32_34); + ptr_tmp5_indices.push_back(const_int32_38); + Instruction* ptr_tmp5 = new GetElementPtrInst(ptr_args_144, ptr_tmp5_indices.begin(), ptr_tmp5_indices.end(), "tmp5", label_from_consts_exit_137); + std::vector ptr_arraydecay6_indices; + ptr_arraydecay6_indices.push_back(const_int32_34); + ptr_arraydecay6_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay6 = new GetElementPtrInst(ptr_consts_141, ptr_arraydecay6_indices.begin(), ptr_arraydecay6_indices.end(), "arraydecay6", label_from_consts_exit_137); + StoreInst* void_181 = new StoreInst(ptr_arraydecay6, ptr_tmp5, false, label_from_consts_exit_137); + std::vector ptr_tmp7_indices; + ptr_tmp7_indices.push_back(const_int32_34); + ptr_tmp7_indices.push_back(const_int32_37); + Instruction* ptr_tmp7 = new GetElementPtrInst(ptr_args_144, ptr_tmp7_indices.begin(), ptr_tmp7_indices.end(), "tmp7", label_from_consts_exit_137); std::vector ptr_arraydecay8_indices; - ptr_arraydecay8_indices.push_back(const_int32_32); - ptr_arraydecay8_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arraydecay8_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay8 = new GetElementPtrInst(ptr_inputs_132, ptr_arraydecay8_indices.begin(), ptr_arraydecay8_indices.end(), "arraydecay8", label_forbody_130); - std::vector ptr_arraydecay12_indices; - ptr_arraydecay12_indices.push_back(const_int32_32); - ptr_arraydecay12_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arraydecay12_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay12 = new GetElementPtrInst(ptr_results_134, ptr_arraydecay12_indices.begin(), ptr_arraydecay12_indices.end(), "arraydecay12", label_forbody_130); - std::vector void_175_params; - void_175_params.push_back(ptr_arraydecay12); - void_175_params.push_back(ptr_arraydecay8); - void_175_params.push_back(ptr_arraydecay15); - void_175_params.push_back(ptr_arraydecay16); - CallInst* void_175 = new CallInst(func_execute_shader, void_175_params.begin(), void_175_params.end(), "", label_forbody_130); - void_175->setCallingConv(CallingConv::C); - void_175->setTailCall(false); + ptr_arraydecay8_indices.push_back(const_int32_34); + ptr_arraydecay8_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay8 = new GetElementPtrInst(ptr_temps_143, ptr_arraydecay8_indices.begin(), ptr_arraydecay8_indices.end(), "arraydecay8", label_from_consts_exit_137); + StoreInst* void_182 = new StoreInst(ptr_arraydecay8, ptr_tmp7, false, label_from_consts_exit_137); + std::vector ptr_tmp10_indices; + ptr_tmp10_indices.push_back(const_int32_34); + ptr_tmp10_indices.push_back(const_int32_36); + Instruction* ptr_tmp10 = new GetElementPtrInst(ptr_args_144, ptr_tmp10_indices.begin(), ptr_tmp10_indices.end(), "tmp10", label_from_consts_exit_137); + std::vector ptr_tmp14_indices; + ptr_tmp14_indices.push_back(const_int32_34); + ptr_tmp14_indices.push_back(const_int32_34); + Instruction* ptr_tmp14 = new GetElementPtrInst(ptr_args_144, ptr_tmp14_indices.begin(), ptr_tmp14_indices.end(), "tmp14", label_from_consts_exit_137); + new BranchInst(label_forbody_138, label_from_consts_exit_137); + + // Block forbody (label_forbody_138) + Argument* fwdref_185 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_184 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_138); + int32_i_0_reg2mem_0_184->reserveOperandSpace(2); + int32_i_0_reg2mem_0_184->addIncoming(const_int32_34, label_from_consts_exit_137); + int32_i_0_reg2mem_0_184->addIncoming(fwdref_185, label_forbody_138); + + std::vector ptr_arraydecay13_indices; + ptr_arraydecay13_indices.push_back(const_int32_34); + ptr_arraydecay13_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arraydecay13_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay13 = new GetElementPtrInst(ptr_inputs_140, ptr_arraydecay13_indices.begin(), ptr_arraydecay13_indices.end(), "arraydecay13", label_forbody_138); + StoreInst* void_186 = new StoreInst(ptr_arraydecay13, ptr_tmp10, false, label_forbody_138); + std::vector ptr_arraydecay18_indices; + ptr_arraydecay18_indices.push_back(const_int32_34); + ptr_arraydecay18_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arraydecay18_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay18 = new GetElementPtrInst(ptr_results_142, ptr_arraydecay18_indices.begin(), ptr_arraydecay18_indices.end(), "arraydecay18", label_forbody_138); + StoreInst* void_187 = new StoreInst(ptr_arraydecay18, ptr_tmp14, false, label_forbody_138); + CallInst* void_188 = new CallInst(func_execute_shader, ptr_args_144, "", label_forbody_138); + void_188->setCallingConv(CallingConv::C); + void_188->setTailCall(false); + LoadInst* ptr_tmp24 = new LoadInst(ptr_tmp14, "tmp24", false, label_forbody_138); std::vector ptr_arraydecay_i_indices; - ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arraydecay_i_indices.push_back(const_int32_32); - ptr_arraydecay_i_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_dests_121, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_130); - LoadInst* packed_tmp7_i = new LoadInst(ptr_arraydecay12, "tmp7.i", false, label_forbody_130); - ExtractElementInst* float_tmp11_i = new ExtractElementInst(packed_tmp7_i, const_int32_32, "tmp11.i", label_forbody_130); - StoreInst* void_176 = new StoreInst(float_tmp11_i, ptr_arraydecay_i, false, label_forbody_130); + ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arraydecay_i_indices.push_back(const_int32_34); + ptr_arraydecay_i_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_dests_129, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_138); + LoadInst* packed_tmp7_i = new LoadInst(ptr_tmp24, "tmp7.i", false, label_forbody_138); + ExtractElementInst* float_tmp11_i = new ExtractElementInst(packed_tmp7_i, const_int32_34, "tmp11.i", label_forbody_138); + StoreInst* void_189 = new StoreInst(float_tmp11_i, ptr_arraydecay_i, false, label_forbody_138); std::vector ptr_arrayidx13_i_indices; - ptr_arrayidx13_i_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arrayidx13_i_indices.push_back(const_int32_32); + ptr_arrayidx13_i_indices.push_back(int32_i_0_reg2mem_0_184); ptr_arrayidx13_i_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx13_i = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx13_i_indices.begin(), ptr_arrayidx13_i_indices.end(), "arrayidx13.i", label_forbody_130); - ExtractElementInst* float_tmp15_i2 = new ExtractElementInst(packed_tmp7_i, const_int32_34, "tmp15.i2", label_forbody_130); - StoreInst* void_177 = new StoreInst(float_tmp15_i2, ptr_arrayidx13_i, false, label_forbody_130); + ptr_arrayidx13_i_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx13_i = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx13_i_indices.begin(), ptr_arrayidx13_i_indices.end(), "arrayidx13.i", label_forbody_138); + ExtractElementInst* float_tmp15_i2 = new ExtractElementInst(packed_tmp7_i, const_int32_36, "tmp15.i2", label_forbody_138); + StoreInst* void_190 = new StoreInst(float_tmp15_i2, ptr_arrayidx13_i, false, label_forbody_138); std::vector ptr_arrayidx17_i_indices; - ptr_arrayidx17_i_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arrayidx17_i_indices.push_back(const_int32_32); - ptr_arrayidx17_i_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx17_i = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx17_i_indices.begin(), ptr_arrayidx17_i_indices.end(), "arrayidx17.i", label_forbody_130); - ExtractElementInst* float_tmp19_i = new ExtractElementInst(packed_tmp7_i, const_int32_35, "tmp19.i", label_forbody_130); - StoreInst* void_178 = new StoreInst(float_tmp19_i, ptr_arrayidx17_i, false, label_forbody_130); + ptr_arrayidx17_i_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx17_i_indices.push_back(const_int32_34); + ptr_arrayidx17_i_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx17_i = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx17_i_indices.begin(), ptr_arrayidx17_i_indices.end(), "arrayidx17.i", label_forbody_138); + ExtractElementInst* float_tmp19_i = new ExtractElementInst(packed_tmp7_i, const_int32_37, "tmp19.i", label_forbody_138); + StoreInst* void_191 = new StoreInst(float_tmp19_i, ptr_arrayidx17_i, false, label_forbody_138); std::vector ptr_arrayidx21_i_indices; - ptr_arrayidx21_i_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arrayidx21_i_indices.push_back(const_int32_32); - ptr_arrayidx21_i_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx21_i = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx21_i_indices.begin(), ptr_arrayidx21_i_indices.end(), "arrayidx21.i", label_forbody_130); - ExtractElementInst* float_tmp23_i = new ExtractElementInst(packed_tmp7_i, const_int32_36, "tmp23.i", label_forbody_130); - StoreInst* void_179 = new StoreInst(float_tmp23_i, ptr_arrayidx21_i, false, label_forbody_130); + ptr_arrayidx21_i_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx21_i_indices.push_back(const_int32_34); + ptr_arrayidx21_i_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx21_i = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx21_i_indices.begin(), ptr_arrayidx21_i_indices.end(), "arrayidx21.i", label_forbody_138); + ExtractElementInst* float_tmp23_i = new ExtractElementInst(packed_tmp7_i, const_int32_38, "tmp23.i", label_forbody_138); + StoreInst* void_192 = new StoreInst(float_tmp23_i, ptr_arrayidx21_i, false, label_forbody_138); std::vector ptr_arraydecay_i_1_indices; - ptr_arraydecay_i_1_indices.push_back(int32_i_0_reg2mem_0_173); + ptr_arraydecay_i_1_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arraydecay_i_1_indices.push_back(const_int32_36); ptr_arraydecay_i_1_indices.push_back(const_int32_34); - ptr_arraydecay_i_1_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay_i_1 = new GetElementPtrInst(ptr_dests_121, ptr_arraydecay_i_1_indices.begin(), ptr_arraydecay_i_1_indices.end(), "arraydecay.i.1", label_forbody_130); - std::vector ptr_arrayidx6_i_1_indices; - ptr_arrayidx6_i_1_indices.push_back(const_int32_32); - ptr_arrayidx6_i_1_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arrayidx6_i_1_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx6_i_1 = new GetElementPtrInst(ptr_results_134, ptr_arrayidx6_i_1_indices.begin(), ptr_arrayidx6_i_1_indices.end(), "arrayidx6.i.1", label_forbody_130); - LoadInst* packed_tmp7_i_1 = new LoadInst(ptr_arrayidx6_i_1, "tmp7.i.1", false, label_forbody_130); - ExtractElementInst* float_tmp11_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_32, "tmp11.i.1", label_forbody_130); - StoreInst* void_180 = new StoreInst(float_tmp11_i_1, ptr_arraydecay_i_1, false, label_forbody_130); + Instruction* ptr_arraydecay_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arraydecay_i_1_indices.begin(), ptr_arraydecay_i_1_indices.end(), "arraydecay.i.1", label_forbody_138); + GetElementPtrInst* ptr_arrayidx6_i_1 = new GetElementPtrInst(ptr_tmp24, const_int32_36, "arrayidx6.i.1", label_forbody_138); + LoadInst* packed_tmp7_i_1 = new LoadInst(ptr_arrayidx6_i_1, "tmp7.i.1", false, label_forbody_138); + ExtractElementInst* float_tmp11_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_34, "tmp11.i.1", label_forbody_138); + StoreInst* void_193 = new StoreInst(float_tmp11_i_1, ptr_arraydecay_i_1, false, label_forbody_138); std::vector ptr_arrayidx13_i_1_indices; - ptr_arrayidx13_i_1_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arrayidx13_i_1_indices.push_back(const_int32_34); - ptr_arrayidx13_i_1_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx13_i_1 = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx13_i_1_indices.begin(), ptr_arrayidx13_i_1_indices.end(), "arrayidx13.i.1", label_forbody_130); - ExtractElementInst* float_tmp15_i2_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_34, "tmp15.i2.1", label_forbody_130); - StoreInst* void_181 = new StoreInst(float_tmp15_i2_1, ptr_arrayidx13_i_1, false, label_forbody_130); + ptr_arrayidx13_i_1_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx13_i_1_indices.push_back(const_int32_36); + ptr_arrayidx13_i_1_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx13_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx13_i_1_indices.begin(), ptr_arrayidx13_i_1_indices.end(), "arrayidx13.i.1", label_forbody_138); + ExtractElementInst* float_tmp15_i2_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_36, "tmp15.i2.1", label_forbody_138); + StoreInst* void_194 = new StoreInst(float_tmp15_i2_1, ptr_arrayidx13_i_1, false, label_forbody_138); std::vector ptr_arrayidx17_i_1_indices; - ptr_arrayidx17_i_1_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arrayidx17_i_1_indices.push_back(const_int32_34); - ptr_arrayidx17_i_1_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx17_i_1 = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx17_i_1_indices.begin(), ptr_arrayidx17_i_1_indices.end(), "arrayidx17.i.1", label_forbody_130); - ExtractElementInst* float_tmp19_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_35, "tmp19.i.1", label_forbody_130); - StoreInst* void_182 = new StoreInst(float_tmp19_i_1, ptr_arrayidx17_i_1, false, label_forbody_130); + ptr_arrayidx17_i_1_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx17_i_1_indices.push_back(const_int32_36); + ptr_arrayidx17_i_1_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx17_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx17_i_1_indices.begin(), ptr_arrayidx17_i_1_indices.end(), "arrayidx17.i.1", label_forbody_138); + ExtractElementInst* float_tmp19_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_37, "tmp19.i.1", label_forbody_138); + StoreInst* void_195 = new StoreInst(float_tmp19_i_1, ptr_arrayidx17_i_1, false, label_forbody_138); std::vector ptr_arrayidx21_i_1_indices; - ptr_arrayidx21_i_1_indices.push_back(int32_i_0_reg2mem_0_173); - ptr_arrayidx21_i_1_indices.push_back(const_int32_34); + ptr_arrayidx21_i_1_indices.push_back(int32_i_0_reg2mem_0_184); ptr_arrayidx21_i_1_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx21_i_1 = new GetElementPtrInst(ptr_dests_121, ptr_arrayidx21_i_1_indices.begin(), ptr_arrayidx21_i_1_indices.end(), "arrayidx21.i.1", label_forbody_130); - ExtractElementInst* float_tmp23_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_36, "tmp23.i.1", label_forbody_130); - StoreInst* void_183 = new StoreInst(float_tmp23_i_1, ptr_arrayidx21_i_1, false, label_forbody_130); - BinaryOperator* int32_indvar_next20 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_173, const_int32_34, "indvar.next20", label_forbody_130); - ICmpInst* int1_exitcond21 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next20, const_int32_37, "exitcond21", label_forbody_130); - new BranchInst(label_afterfor_131, label_forbody_130, int1_exitcond21, label_forbody_130); - - // Block afterfor (label_afterfor_131) - new ReturnInst(const_int32_38, label_afterfor_131); + ptr_arrayidx21_i_1_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx21_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx21_i_1_indices.begin(), ptr_arrayidx21_i_1_indices.end(), "arrayidx21.i.1", label_forbody_138); + ExtractElementInst* float_tmp23_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_38, "tmp23.i.1", label_forbody_138); + StoreInst* void_196 = new StoreInst(float_tmp23_i_1, ptr_arrayidx21_i_1, false, label_forbody_138); + BinaryOperator* int32_indvar_next20 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_184, const_int32_36, "indvar.next20", label_forbody_138); + ICmpInst* int1_exitcond21 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next20, const_int32_39, "exitcond21", label_forbody_138); + new BranchInst(label_afterfor_139, label_forbody_138, int1_exitcond21, label_forbody_138); + + // Block afterfor (label_afterfor_139) + std::vector ptr_tmp26_indices; + ptr_tmp26_indices.push_back(const_int32_34); + ptr_tmp26_indices.push_back(const_int32_39); + Instruction* ptr_tmp26 = new GetElementPtrInst(ptr_args_144, ptr_tmp26_indices.begin(), ptr_tmp26_indices.end(), "tmp26", label_afterfor_139); + LoadInst* int32_tmp27 = new LoadInst(ptr_tmp26, "tmp27", false, label_afterfor_139); + BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_tmp27, const_int32_40, "neg", label_afterfor_139); + new ReturnInst(int32_neg, label_afterfor_139); // Block forbody6.i.1 (label_forbody6_i_1) - Argument* fwdref_186 = new Argument(IntegerType::get(32)); + Argument* fwdref_199 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i_1 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.1", label_forbody6_i_1); int32_j_0_reg2mem_0_i_1->reserveOperandSpace(2); - int32_j_0_reg2mem_0_i_1->addIncoming(const_int32_32, label_forbody6_i_127); - int32_j_0_reg2mem_0_i_1->addIncoming(fwdref_186, label_forbody6_i_1); + int32_j_0_reg2mem_0_i_1->addIncoming(const_int32_34, label_forbody6_i_135); + int32_j_0_reg2mem_0_i_1->addIncoming(fwdref_199, label_forbody6_i_1); - Argument* fwdref_187 = new Argument(VectorTy_13); - PHINode* packed_vec_0_reg2mem_0_i_1 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i.1", label_forbody6_i_1); + Argument* fwdref_200 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i_1 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i.1", label_forbody6_i_1); packed_vec_0_reg2mem_0_i_1->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_1->addIncoming(packed_tmp48_i_153, label_forbody6_i_127); - packed_vec_0_reg2mem_0_i_1->addIncoming(fwdref_187, label_forbody6_i_1); + packed_vec_0_reg2mem_0_i_1->addIncoming(packed_tmp48_i_162, label_forbody6_i_135); + packed_vec_0_reg2mem_0_i_1->addIncoming(fwdref_200, label_forbody6_i_1); std::vector ptr_arraydecay11_i_1_indices; - ptr_arraydecay11_i_1_indices.push_back(const_int32_34); + ptr_arraydecay11_i_1_indices.push_back(const_int32_36); ptr_arraydecay11_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); - ptr_arraydecay11_i_1_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay11_i_1 = new GetElementPtrInst(ptr_ainputs_122, ptr_arraydecay11_i_1_indices.begin(), ptr_arraydecay11_i_1_indices.end(), "arraydecay11.i.1", label_forbody6_i_1); + ptr_arraydecay11_i_1_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay11_i_1 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_1_indices.begin(), ptr_arraydecay11_i_1_indices.end(), "arraydecay11.i.1", label_forbody6_i_1); LoadInst* float_tmp13_i_1 = new LoadInst(ptr_arraydecay11_i_1, "tmp13.i.1", false, label_forbody6_i_1); - InsertElementInst* packed_tmp15_i_1 = new InsertElementInst(packed_vec_0_reg2mem_0_i_1, float_tmp13_i_1, const_int32_32, "tmp15.i.1", label_forbody6_i_1); + InsertElementInst* packed_tmp15_i_1 = new InsertElementInst(packed_vec_0_reg2mem_0_i_1, float_tmp13_i_1, const_int32_34, "tmp15.i.1", label_forbody6_i_1); std::vector ptr_arrayidx23_i_1_indices; - ptr_arrayidx23_i_1_indices.push_back(const_int32_34); + ptr_arrayidx23_i_1_indices.push_back(const_int32_36); ptr_arrayidx23_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); - ptr_arrayidx23_i_1_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx23_i_1 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx23_i_1_indices.begin(), ptr_arrayidx23_i_1_indices.end(), "arrayidx23.i.1", label_forbody6_i_1); + ptr_arrayidx23_i_1_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx23_i_1 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_1_indices.begin(), ptr_arrayidx23_i_1_indices.end(), "arrayidx23.i.1", label_forbody6_i_1); LoadInst* float_tmp24_i_1 = new LoadInst(ptr_arrayidx23_i_1, "tmp24.i.1", false, label_forbody6_i_1); - InsertElementInst* packed_tmp26_i_1 = new InsertElementInst(packed_tmp15_i_1, float_tmp24_i_1, const_int32_34, "tmp26.i.1", label_forbody6_i_1); + InsertElementInst* packed_tmp26_i_1 = new InsertElementInst(packed_tmp15_i_1, float_tmp24_i_1, const_int32_36, "tmp26.i.1", label_forbody6_i_1); std::vector ptr_arrayidx34_i_1_indices; - ptr_arrayidx34_i_1_indices.push_back(const_int32_34); + ptr_arrayidx34_i_1_indices.push_back(const_int32_36); ptr_arrayidx34_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); - ptr_arrayidx34_i_1_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx34_i_1 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx34_i_1_indices.begin(), ptr_arrayidx34_i_1_indices.end(), "arrayidx34.i.1", label_forbody6_i_1); + ptr_arrayidx34_i_1_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx34_i_1 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_1_indices.begin(), ptr_arrayidx34_i_1_indices.end(), "arrayidx34.i.1", label_forbody6_i_1); LoadInst* float_tmp35_i_1 = new LoadInst(ptr_arrayidx34_i_1, "tmp35.i.1", false, label_forbody6_i_1); - InsertElementInst* packed_tmp37_i_1 = new InsertElementInst(packed_tmp26_i_1, float_tmp35_i_1, const_int32_35, "tmp37.i.1", label_forbody6_i_1); + InsertElementInst* packed_tmp37_i_1 = new InsertElementInst(packed_tmp26_i_1, float_tmp35_i_1, const_int32_37, "tmp37.i.1", label_forbody6_i_1); std::vector ptr_arrayidx45_i_1_indices; - ptr_arrayidx45_i_1_indices.push_back(const_int32_34); - ptr_arrayidx45_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); ptr_arrayidx45_i_1_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx45_i_1 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx45_i_1_indices.begin(), ptr_arrayidx45_i_1_indices.end(), "arrayidx45.i.1", label_forbody6_i_1); + ptr_arrayidx45_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); + ptr_arrayidx45_i_1_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx45_i_1 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_1_indices.begin(), ptr_arrayidx45_i_1_indices.end(), "arrayidx45.i.1", label_forbody6_i_1); LoadInst* float_tmp46_i_1 = new LoadInst(ptr_arrayidx45_i_1, "tmp46.i.1", false, label_forbody6_i_1); - InsertElementInst* packed_tmp48_i_1 = new InsertElementInst(packed_tmp37_i_1, float_tmp46_i_1, const_int32_36, "tmp48.i.1", label_forbody6_i_1); + InsertElementInst* packed_tmp48_i_1 = new InsertElementInst(packed_tmp37_i_1, float_tmp46_i_1, const_int32_38, "tmp48.i.1", label_forbody6_i_1); std::vector ptr_arrayidx54_i_1_indices; - ptr_arrayidx54_i_1_indices.push_back(const_int32_32); ptr_arrayidx54_i_1_indices.push_back(const_int32_34); + ptr_arrayidx54_i_1_indices.push_back(const_int32_36); ptr_arrayidx54_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); - Instruction* ptr_arrayidx54_i_1 = new GetElementPtrInst(ptr_inputs_132, ptr_arrayidx54_i_1_indices.begin(), ptr_arrayidx54_i_1_indices.end(), "arrayidx54.i.1", label_forbody6_i_1); - StoreInst* void_188 = new StoreInst(packed_tmp48_i_1, ptr_arrayidx54_i_1, false, label_forbody6_i_1); - BinaryOperator* int32_inc_i_1 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_1, const_int32_34, "inc.i.1", label_forbody6_i_1); - ICmpInst* int1_cmp59_i_1 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_1, int32_num_inputs_123, "cmp59.i.1", label_forbody6_i_1); + Instruction* ptr_arrayidx54_i_1 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_1_indices.begin(), ptr_arrayidx54_i_1_indices.end(), "arrayidx54.i.1", label_forbody6_i_1); + StoreInst* void_201 = new StoreInst(packed_tmp48_i_1, ptr_arrayidx54_i_1, false, label_forbody6_i_1); + BinaryOperator* int32_inc_i_1 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_1, const_int32_36, "inc.i.1", label_forbody6_i_1); + ICmpInst* int1_cmp59_i_1 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_1, int32_num_inputs_131, "cmp59.i.1", label_forbody6_i_1); new BranchInst(label_forbody6_i_1, label_forbody6_i_2, int1_cmp59_i_1, label_forbody6_i_1); // Block forbody6.i.2 (label_forbody6_i_2) - Argument* fwdref_190 = new Argument(IntegerType::get(32)); + Argument* fwdref_203 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i_2 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.2", label_forbody6_i_2); int32_j_0_reg2mem_0_i_2->reserveOperandSpace(2); - int32_j_0_reg2mem_0_i_2->addIncoming(const_int32_32, label_forbody6_i_1); - int32_j_0_reg2mem_0_i_2->addIncoming(fwdref_190, label_forbody6_i_2); + int32_j_0_reg2mem_0_i_2->addIncoming(const_int32_34, label_forbody6_i_1); + int32_j_0_reg2mem_0_i_2->addIncoming(fwdref_203, label_forbody6_i_2); - Argument* fwdref_191 = new Argument(VectorTy_13); - PHINode* packed_vec_0_reg2mem_0_i_2 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i.2", label_forbody6_i_2); + Argument* fwdref_204 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i_2 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i.2", label_forbody6_i_2); packed_vec_0_reg2mem_0_i_2->reserveOperandSpace(2); packed_vec_0_reg2mem_0_i_2->addIncoming(packed_tmp48_i_1, label_forbody6_i_1); - packed_vec_0_reg2mem_0_i_2->addIncoming(fwdref_191, label_forbody6_i_2); + packed_vec_0_reg2mem_0_i_2->addIncoming(fwdref_204, label_forbody6_i_2); std::vector ptr_arraydecay11_i_2_indices; - ptr_arraydecay11_i_2_indices.push_back(const_int32_35); + ptr_arraydecay11_i_2_indices.push_back(const_int32_37); ptr_arraydecay11_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - ptr_arraydecay11_i_2_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay11_i_2 = new GetElementPtrInst(ptr_ainputs_122, ptr_arraydecay11_i_2_indices.begin(), ptr_arraydecay11_i_2_indices.end(), "arraydecay11.i.2", label_forbody6_i_2); + ptr_arraydecay11_i_2_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay11_i_2 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_2_indices.begin(), ptr_arraydecay11_i_2_indices.end(), "arraydecay11.i.2", label_forbody6_i_2); LoadInst* float_tmp13_i_2 = new LoadInst(ptr_arraydecay11_i_2, "tmp13.i.2", false, label_forbody6_i_2); - InsertElementInst* packed_tmp15_i_2 = new InsertElementInst(packed_vec_0_reg2mem_0_i_2, float_tmp13_i_2, const_int32_32, "tmp15.i.2", label_forbody6_i_2); + InsertElementInst* packed_tmp15_i_2 = new InsertElementInst(packed_vec_0_reg2mem_0_i_2, float_tmp13_i_2, const_int32_34, "tmp15.i.2", label_forbody6_i_2); std::vector ptr_arrayidx23_i_2_indices; - ptr_arrayidx23_i_2_indices.push_back(const_int32_35); + ptr_arrayidx23_i_2_indices.push_back(const_int32_37); ptr_arrayidx23_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - ptr_arrayidx23_i_2_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx23_i_2 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx23_i_2_indices.begin(), ptr_arrayidx23_i_2_indices.end(), "arrayidx23.i.2", label_forbody6_i_2); + ptr_arrayidx23_i_2_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx23_i_2 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_2_indices.begin(), ptr_arrayidx23_i_2_indices.end(), "arrayidx23.i.2", label_forbody6_i_2); LoadInst* float_tmp24_i_2 = new LoadInst(ptr_arrayidx23_i_2, "tmp24.i.2", false, label_forbody6_i_2); - InsertElementInst* packed_tmp26_i_2 = new InsertElementInst(packed_tmp15_i_2, float_tmp24_i_2, const_int32_34, "tmp26.i.2", label_forbody6_i_2); + InsertElementInst* packed_tmp26_i_2 = new InsertElementInst(packed_tmp15_i_2, float_tmp24_i_2, const_int32_36, "tmp26.i.2", label_forbody6_i_2); std::vector ptr_arrayidx34_i_2_indices; - ptr_arrayidx34_i_2_indices.push_back(const_int32_35); + ptr_arrayidx34_i_2_indices.push_back(const_int32_37); ptr_arrayidx34_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - ptr_arrayidx34_i_2_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx34_i_2 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx34_i_2_indices.begin(), ptr_arrayidx34_i_2_indices.end(), "arrayidx34.i.2", label_forbody6_i_2); + ptr_arrayidx34_i_2_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx34_i_2 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_2_indices.begin(), ptr_arrayidx34_i_2_indices.end(), "arrayidx34.i.2", label_forbody6_i_2); LoadInst* float_tmp35_i_2 = new LoadInst(ptr_arrayidx34_i_2, "tmp35.i.2", false, label_forbody6_i_2); - InsertElementInst* packed_tmp37_i_2 = new InsertElementInst(packed_tmp26_i_2, float_tmp35_i_2, const_int32_35, "tmp37.i.2", label_forbody6_i_2); + InsertElementInst* packed_tmp37_i_2 = new InsertElementInst(packed_tmp26_i_2, float_tmp35_i_2, const_int32_37, "tmp37.i.2", label_forbody6_i_2); std::vector ptr_arrayidx45_i_2_indices; - ptr_arrayidx45_i_2_indices.push_back(const_int32_35); + ptr_arrayidx45_i_2_indices.push_back(const_int32_37); ptr_arrayidx45_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - ptr_arrayidx45_i_2_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx45_i_2 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx45_i_2_indices.begin(), ptr_arrayidx45_i_2_indices.end(), "arrayidx45.i.2", label_forbody6_i_2); + ptr_arrayidx45_i_2_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx45_i_2 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_2_indices.begin(), ptr_arrayidx45_i_2_indices.end(), "arrayidx45.i.2", label_forbody6_i_2); LoadInst* float_tmp46_i_2 = new LoadInst(ptr_arrayidx45_i_2, "tmp46.i.2", false, label_forbody6_i_2); - InsertElementInst* packed_tmp48_i_2 = new InsertElementInst(packed_tmp37_i_2, float_tmp46_i_2, const_int32_36, "tmp48.i.2", label_forbody6_i_2); + InsertElementInst* packed_tmp48_i_2 = new InsertElementInst(packed_tmp37_i_2, float_tmp46_i_2, const_int32_38, "tmp48.i.2", label_forbody6_i_2); std::vector ptr_arrayidx54_i_2_indices; - ptr_arrayidx54_i_2_indices.push_back(const_int32_32); - ptr_arrayidx54_i_2_indices.push_back(const_int32_35); + ptr_arrayidx54_i_2_indices.push_back(const_int32_34); + ptr_arrayidx54_i_2_indices.push_back(const_int32_37); ptr_arrayidx54_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - Instruction* ptr_arrayidx54_i_2 = new GetElementPtrInst(ptr_inputs_132, ptr_arrayidx54_i_2_indices.begin(), ptr_arrayidx54_i_2_indices.end(), "arrayidx54.i.2", label_forbody6_i_2); - StoreInst* void_192 = new StoreInst(packed_tmp48_i_2, ptr_arrayidx54_i_2, false, label_forbody6_i_2); - BinaryOperator* int32_inc_i_2 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_2, const_int32_34, "inc.i.2", label_forbody6_i_2); - ICmpInst* int1_cmp59_i_2 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_2, int32_num_inputs_123, "cmp59.i.2", label_forbody6_i_2); + Instruction* ptr_arrayidx54_i_2 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_2_indices.begin(), ptr_arrayidx54_i_2_indices.end(), "arrayidx54.i.2", label_forbody6_i_2); + StoreInst* void_205 = new StoreInst(packed_tmp48_i_2, ptr_arrayidx54_i_2, false, label_forbody6_i_2); + BinaryOperator* int32_inc_i_2 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_2, const_int32_36, "inc.i.2", label_forbody6_i_2); + ICmpInst* int1_cmp59_i_2 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_2, int32_num_inputs_131, "cmp59.i.2", label_forbody6_i_2); new BranchInst(label_forbody6_i_2, label_forbody6_i_3, int1_cmp59_i_2, label_forbody6_i_2); // Block forbody6.i.3 (label_forbody6_i_3) - Argument* fwdref_194 = new Argument(IntegerType::get(32)); + Argument* fwdref_207 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i_3 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.3", label_forbody6_i_3); int32_j_0_reg2mem_0_i_3->reserveOperandSpace(2); - int32_j_0_reg2mem_0_i_3->addIncoming(const_int32_32, label_forbody6_i_2); - int32_j_0_reg2mem_0_i_3->addIncoming(fwdref_194, label_forbody6_i_3); + int32_j_0_reg2mem_0_i_3->addIncoming(const_int32_34, label_forbody6_i_2); + int32_j_0_reg2mem_0_i_3->addIncoming(fwdref_207, label_forbody6_i_3); - Argument* fwdref_195 = new Argument(VectorTy_13); - PHINode* packed_vec_0_reg2mem_0_i_3 = new PHINode(VectorTy_13, "vec.0.reg2mem.0.i.3", label_forbody6_i_3); + Argument* fwdref_208 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i_3 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i.3", label_forbody6_i_3); packed_vec_0_reg2mem_0_i_3->reserveOperandSpace(2); packed_vec_0_reg2mem_0_i_3->addIncoming(packed_tmp48_i_2, label_forbody6_i_2); - packed_vec_0_reg2mem_0_i_3->addIncoming(fwdref_195, label_forbody6_i_3); + packed_vec_0_reg2mem_0_i_3->addIncoming(fwdref_208, label_forbody6_i_3); std::vector ptr_arraydecay11_i_3_indices; - ptr_arraydecay11_i_3_indices.push_back(const_int32_36); + ptr_arraydecay11_i_3_indices.push_back(const_int32_38); ptr_arraydecay11_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - ptr_arraydecay11_i_3_indices.push_back(const_int32_32); - Instruction* ptr_arraydecay11_i_3 = new GetElementPtrInst(ptr_ainputs_122, ptr_arraydecay11_i_3_indices.begin(), ptr_arraydecay11_i_3_indices.end(), "arraydecay11.i.3", label_forbody6_i_3); + ptr_arraydecay11_i_3_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay11_i_3 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_3_indices.begin(), ptr_arraydecay11_i_3_indices.end(), "arraydecay11.i.3", label_forbody6_i_3); LoadInst* float_tmp13_i_3 = new LoadInst(ptr_arraydecay11_i_3, "tmp13.i.3", false, label_forbody6_i_3); - InsertElementInst* packed_tmp15_i_3 = new InsertElementInst(packed_vec_0_reg2mem_0_i_3, float_tmp13_i_3, const_int32_32, "tmp15.i.3", label_forbody6_i_3); + InsertElementInst* packed_tmp15_i_3 = new InsertElementInst(packed_vec_0_reg2mem_0_i_3, float_tmp13_i_3, const_int32_34, "tmp15.i.3", label_forbody6_i_3); std::vector ptr_arrayidx23_i_3_indices; - ptr_arrayidx23_i_3_indices.push_back(const_int32_36); + ptr_arrayidx23_i_3_indices.push_back(const_int32_38); ptr_arrayidx23_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - ptr_arrayidx23_i_3_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx23_i_3 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx23_i_3_indices.begin(), ptr_arrayidx23_i_3_indices.end(), "arrayidx23.i.3", label_forbody6_i_3); + ptr_arrayidx23_i_3_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx23_i_3 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_3_indices.begin(), ptr_arrayidx23_i_3_indices.end(), "arrayidx23.i.3", label_forbody6_i_3); LoadInst* float_tmp24_i_3 = new LoadInst(ptr_arrayidx23_i_3, "tmp24.i.3", false, label_forbody6_i_3); - InsertElementInst* packed_tmp26_i_3 = new InsertElementInst(packed_tmp15_i_3, float_tmp24_i_3, const_int32_34, "tmp26.i.3", label_forbody6_i_3); + InsertElementInst* packed_tmp26_i_3 = new InsertElementInst(packed_tmp15_i_3, float_tmp24_i_3, const_int32_36, "tmp26.i.3", label_forbody6_i_3); std::vector ptr_arrayidx34_i_3_indices; - ptr_arrayidx34_i_3_indices.push_back(const_int32_36); + ptr_arrayidx34_i_3_indices.push_back(const_int32_38); ptr_arrayidx34_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - ptr_arrayidx34_i_3_indices.push_back(const_int32_35); - Instruction* ptr_arrayidx34_i_3 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx34_i_3_indices.begin(), ptr_arrayidx34_i_3_indices.end(), "arrayidx34.i.3", label_forbody6_i_3); + ptr_arrayidx34_i_3_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx34_i_3 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_3_indices.begin(), ptr_arrayidx34_i_3_indices.end(), "arrayidx34.i.3", label_forbody6_i_3); LoadInst* float_tmp35_i_3 = new LoadInst(ptr_arrayidx34_i_3, "tmp35.i.3", false, label_forbody6_i_3); - InsertElementInst* packed_tmp37_i_3 = new InsertElementInst(packed_tmp26_i_3, float_tmp35_i_3, const_int32_35, "tmp37.i.3", label_forbody6_i_3); + InsertElementInst* packed_tmp37_i_3 = new InsertElementInst(packed_tmp26_i_3, float_tmp35_i_3, const_int32_37, "tmp37.i.3", label_forbody6_i_3); std::vector ptr_arrayidx45_i_3_indices; - ptr_arrayidx45_i_3_indices.push_back(const_int32_36); + ptr_arrayidx45_i_3_indices.push_back(const_int32_38); ptr_arrayidx45_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - ptr_arrayidx45_i_3_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx45_i_3 = new GetElementPtrInst(ptr_ainputs_122, ptr_arrayidx45_i_3_indices.begin(), ptr_arrayidx45_i_3_indices.end(), "arrayidx45.i.3", label_forbody6_i_3); + ptr_arrayidx45_i_3_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx45_i_3 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_3_indices.begin(), ptr_arrayidx45_i_3_indices.end(), "arrayidx45.i.3", label_forbody6_i_3); LoadInst* float_tmp46_i_3 = new LoadInst(ptr_arrayidx45_i_3, "tmp46.i.3", false, label_forbody6_i_3); - InsertElementInst* packed_tmp48_i_3 = new InsertElementInst(packed_tmp37_i_3, float_tmp46_i_3, const_int32_36, "tmp48.i.3", label_forbody6_i_3); + InsertElementInst* packed_tmp48_i_3 = new InsertElementInst(packed_tmp37_i_3, float_tmp46_i_3, const_int32_38, "tmp48.i.3", label_forbody6_i_3); std::vector ptr_arrayidx54_i_3_indices; - ptr_arrayidx54_i_3_indices.push_back(const_int32_32); - ptr_arrayidx54_i_3_indices.push_back(const_int32_36); + ptr_arrayidx54_i_3_indices.push_back(const_int32_34); + ptr_arrayidx54_i_3_indices.push_back(const_int32_38); ptr_arrayidx54_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - Instruction* ptr_arrayidx54_i_3 = new GetElementPtrInst(ptr_inputs_132, ptr_arrayidx54_i_3_indices.begin(), ptr_arrayidx54_i_3_indices.end(), "arrayidx54.i.3", label_forbody6_i_3); - StoreInst* void_196 = new StoreInst(packed_tmp48_i_3, ptr_arrayidx54_i_3, false, label_forbody6_i_3); - BinaryOperator* int32_inc_i_3 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_3, const_int32_34, "inc.i.3", label_forbody6_i_3); - ICmpInst* int1_cmp59_i_3 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_3, int32_num_inputs_123, "cmp59.i.3", label_forbody6_i_3); - new BranchInst(label_forbody6_i_3, label_from_array_exit_128, int1_cmp59_i_3, label_forbody6_i_3); + Instruction* ptr_arrayidx54_i_3 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_3_indices.begin(), ptr_arrayidx54_i_3_indices.end(), "arrayidx54.i.3", label_forbody6_i_3); + StoreInst* void_209 = new StoreInst(packed_tmp48_i_3, ptr_arrayidx54_i_3, false, label_forbody6_i_3); + BinaryOperator* int32_inc_i_3 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_3, const_int32_36, "inc.i.3", label_forbody6_i_3); + ICmpInst* int1_cmp59_i_3 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_3, int32_num_inputs_131, "cmp59.i.3", label_forbody6_i_3); + new BranchInst(label_forbody6_i_3, label_from_array_exit_136, int1_cmp59_i_3, label_forbody6_i_3); // Resolve Forward References - fwdref_195->replaceAllUsesWith(packed_tmp48_i_3); delete fwdref_195; - fwdref_194->replaceAllUsesWith(int32_inc_i_3); delete fwdref_194; - fwdref_141->replaceAllUsesWith(packed_tmp48_i_153); delete fwdref_141; - fwdref_139->replaceAllUsesWith(int32_inc_i_156); delete fwdref_139; - fwdref_162->replaceAllUsesWith(packed_tmp31_i_169); delete fwdref_162; - fwdref_161->replaceAllUsesWith(int32_indvar_next22); delete fwdref_161; - fwdref_174->replaceAllUsesWith(int32_indvar_next20); delete fwdref_174; - fwdref_187->replaceAllUsesWith(packed_tmp48_i_1); delete fwdref_187; - fwdref_186->replaceAllUsesWith(int32_inc_i_1); delete fwdref_186; - fwdref_191->replaceAllUsesWith(packed_tmp48_i_2); delete fwdref_191; - fwdref_190->replaceAllUsesWith(int32_inc_i_2); delete fwdref_190; + fwdref_208->replaceAllUsesWith(packed_tmp48_i_3); delete fwdref_208; + fwdref_207->replaceAllUsesWith(int32_inc_i_3); delete fwdref_207; + fwdref_150->replaceAllUsesWith(packed_tmp48_i_162); delete fwdref_150; + fwdref_148->replaceAllUsesWith(int32_inc_i_165); delete fwdref_148; + fwdref_171->replaceAllUsesWith(packed_tmp31_i_178); delete fwdref_171; + fwdref_170->replaceAllUsesWith(int32_indvar_next22); delete fwdref_170; + fwdref_185->replaceAllUsesWith(int32_indvar_next20); delete fwdref_185; + fwdref_200->replaceAllUsesWith(packed_tmp48_i_1); delete fwdref_200; + fwdref_199->replaceAllUsesWith(int32_inc_i_1); delete fwdref_199; + fwdref_204->replaceAllUsesWith(packed_tmp48_i_2); delete fwdref_204; + fwdref_203->replaceAllUsesWith(int32_inc_i_2); delete fwdref_203; } diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index cbe4965ef6..6bdb311c2b 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -1,4 +1,4 @@ -/* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -funcname=createBaseShader */ +/* clang --emit-llvm llvm_entry.c |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -funcname=createBaseShader */ /************************************************************************** * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. @@ -164,8 +164,17 @@ void to_array(float (*dests)[4], float4 *in, int num_attribs) } } -extern void execute_shader(float4 dests[16], float4 inputs[16], - float4 consts[32], float4 temps[128]); + +struct ShaderInput +{ + float4 *dests; + float4 *inputs; + float4 *temps; + float4 *consts; + int kilmask; +}; + +extern void execute_shader(struct ShaderInput *input); void run_vertex_shader(float (*ainputs)[16][4], float (*dests)[16][4], @@ -180,15 +189,18 @@ void run_vertex_shader(float (*ainputs)[16][4], float4 results[16*32*4][16]; float4 temps[128];//MAX_PROGRAM_TEMPS + struct ShaderInput args; /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", num_vertices, num_inputs, num_attribs, num_consts);*/ from_array(inputs, ainputs, num_vertices, num_inputs); from_consts(consts, aconsts, num_consts); + args.consts = consts; + args.temps = temps; for (int i = 0; i < num_vertices; ++i) { - float4 *in = inputs[i]; - float4 *res = results[i]; - execute_shader(res, in, consts, temps); - to_array(dests[i], res, num_attribs); + args.dests = results[i]; + args.inputs = inputs[i]; + execute_shader(&args); + to_array(dests[i], args.dests, num_attribs); } } @@ -213,6 +225,7 @@ struct tgsi_sampler struct softpipe_tile_cache *cache; }; + int run_fragment_shader(float x, float y, float (*dests)[16][4], float (*ainputs)[16][4], @@ -225,17 +238,19 @@ int run_fragment_shader(float x, float y, float4 consts[32]; float4 results[4][16]; float4 temps[128];//MAX_PROGRAM_TEMPS - int kilmask = 0; + struct ShaderInput args; from_array(inputs, ainputs, 4, num_inputs); from_consts(consts, aconsts, num_consts); + args.consts = consts; + args.temps = temps; //printf("AAAAAAAAAAAAAAAAAAAAAAA FRAGMENT SHADER %f %f\n", x, y); for (int i = 0; i < 4; ++i) { - float4 *in = inputs[i]; - float4 *res = results[i]; - execute_shader(res, in, consts, temps); - to_array(dests[i], res, 2); + args.inputs = inputs[i]; + args.dests = results[i]; + execute_shader(&args); + to_array(dests[i], args.dests, 2); } - return ~kilmask; + return ~args.kilmask; } diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 1715bd4de4..71045fa37f 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -46,12 +46,10 @@ using namespace llvm; -Storage::Storage(llvm::BasicBlock *block, llvm::Value *out, - llvm::Value *in, llvm::Value *consts, llvm::Value *temps) - : m_block(block), m_OUT(out), - m_IN(in), m_CONST(consts), m_TEMPS(temps), +Storage::Storage(llvm::BasicBlock *block, llvm::Value *input) + : m_block(block), + m_INPUT(input), m_addrs(32), - m_dstCache(32), m_idx(0) { m_floatVecType = VectorType::get(Type::FloatTy, 4); @@ -112,26 +110,8 @@ llvm::ConstantInt *Storage::constantInt(int idx) llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) { - GetElementPtrInst *getElem = 0; - - if (indIdx) { - getElem = new GetElementPtrInst(m_IN, - BinaryOperator::create(Instruction::Add, - indIdx, - constantInt(idx), - name("add"), - m_block), - name("input_ptr"), - m_block); - } else { - getElem = new GetElementPtrInst(m_IN, - constantInt(idx), - name("input_ptr"), - m_block); - } - - LoadInst *load = new LoadInst(getElem, name("input"), - false, m_block); + Value *val = element(InputsArg, idx, indIdx); + LoadInst *load = new LoadInst(val, name("input"), false, m_block); load->setAlignment(8); return load; @@ -141,24 +121,8 @@ llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) { m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; - GetElementPtrInst *getElem = 0; - - if (indIdx) - getElem = new GetElementPtrInst(m_CONST, - BinaryOperator::create(Instruction::Add, - indIdx, - constantInt(idx), - name("add"), - m_block), - name("const_ptr"), - m_block); - else - getElem = new GetElementPtrInst(m_CONST, - constantInt(idx), - name("const_ptr"), - m_block); - LoadInst *load = new LoadInst(getElem, name("const"), - false, m_block); + Value *elem = element(ConstsArg, idx, indIdx); + LoadInst *load = new LoadInst(elem, name("const"), false, m_block); load->setAlignment(8); return load; } @@ -175,26 +139,9 @@ llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) llvm::Value *Storage::tempElement(int idx, llvm::Value *indIdx) { - GetElementPtrInst *getElem = 0; + Value *elem = element(TempsArg, idx, indIdx); - if (indIdx) { - getElem = new GetElementPtrInst(m_TEMPS, - BinaryOperator::create(Instruction::Add, - indIdx, - constantInt(idx), - name("add"), - m_block), - name("temp_ptr"), - m_block); - } else { - getElem = new GetElementPtrInst(m_TEMPS, - constantInt(idx), - name("temp_ptr"), - m_block); - } - - LoadInst *load = new LoadInst(getElem, name("temp"), - false, m_block); + LoadInst *load = new LoadInst(elem, name("temp"), false, m_block); load->setAlignment(8); return load; @@ -208,11 +155,8 @@ void Storage::setTempElement(int idx, llvm::Value *val, int mask) templ = tempElement(idx); val = maskWrite(val, mask, templ); } - GetElementPtrInst *getElem = new GetElementPtrInst(m_TEMPS, - constantInt(idx), - name("temp_ptr"), - m_block); - StoreInst *st = new StoreInst(val, getElem, false, m_block); + Value *elem = element(TempsArg, idx); + StoreInst *st = new StoreInst(val, elem, false, m_block); st->setAlignment(8); m_tempWriteMap[idx] = true; } @@ -226,11 +170,8 @@ void Storage::setOutputElement(int dstIdx, llvm::Value *val, int mask) val = maskWrite(val, mask, templ); } - GetElementPtrInst *getElem = new GetElementPtrInst(m_OUT, - constantInt(dstIdx), - name("out_ptr"), - m_block); - StoreInst *st = new StoreInst(val, getElem, false, m_block); + Value *elem = element(DestsArg, dstIdx); + StoreInst *st = new StoreInst(val, elem, false, m_block); st->setAlignment(8); m_destWriteMap[dstIdx] = true; } @@ -310,26 +251,8 @@ void Storage::setCurrentBlock(llvm::BasicBlock *block) llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx) { - GetElementPtrInst *getElem = 0; - - if (indIdx) { - getElem = new GetElementPtrInst(m_OUT, - BinaryOperator::create(Instruction::Add, - indIdx, - constantInt(idx), - name("add"), - m_block), - name("output_ptr"), - m_block); - } else { - getElem = new GetElementPtrInst(m_OUT, - constantInt(idx), - name("output_ptr"), - m_block); - } - - LoadInst *load = new LoadInst(getElem, name("output"), - false, m_block); + Value *elem = element(DestsArg, idx, indIdx); + LoadInst *load = new LoadInst(elem, name("output"), false, m_block); load->setAlignment(8); return load; @@ -337,47 +260,19 @@ llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx) llvm::Value * Storage::inputPtr() const { - return m_IN; -} - -llvm::Value * Storage::outputPtr() const -{ - return m_OUT; -} - -llvm::Value * Storage::constPtr() const -{ - return m_CONST; + return m_INPUT; } -llvm::Value * Storage::tempPtr() const +void Storage::pushArguments(llvm::Value *input) { - return m_TEMPS; -} + m_argStack.push(m_INPUT); -void Storage::pushArguments(llvm::Value *out, llvm::Value *in, - llvm::Value *constPtr, llvm::Value *temp) -{ - Args arg; - arg.out = m_OUT; - arg.in = m_IN; - arg.cst = m_CONST; - arg.temp = m_TEMPS; - m_argStack.push(arg); - - m_OUT = out; - m_IN = in; - m_CONST = constPtr; - m_TEMPS = temp; + m_INPUT = input; } void Storage::popArguments() { - Args arg = m_argStack.top(); - m_OUT = arg.out; - m_IN = arg.in; - m_CONST = arg.cst; - m_TEMPS = arg.temp; + m_INPUT = m_argStack.top(); m_argStack.pop(); } @@ -405,4 +300,49 @@ void Storage::addImmediate(float *val) m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); } + +llvm::Value * Storage::elemPtr(Args arg) +{ + std::vector indices; + indices.push_back(constantInt(0)); + indices.push_back(constantInt(static_cast(arg))); + GetElementPtrInst *getElem = new GetElementPtrInst(m_INPUT, + indices.begin(), + indices.end(), + name("input_ptr"), + m_block); + return new LoadInst(getElem, name("input_field"), false, m_block); +} + +llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx, + llvm::Value *indIdx ) +{ + GetElementPtrInst *getElem = 0; + + if (indIdx) { + getElem = new GetElementPtrInst(ptr, + BinaryOperator::create(Instruction::Add, + indIdx, + constantInt(idx), + name("add"), + m_block), + name("field"), + m_block); + } else { + getElem = new GetElementPtrInst(ptr, + constantInt(idx), + name("field"), + m_block); + } + return getElem; +} + +llvm::Value * Storage::element(Args arg, int idx, llvm::Value *indIdx ) +{ + Value *val = elemPtr(arg); + return elemIdx(val, idx, indIdx); +} + #endif //MESA_LLVM + + diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index aa02f02f85..7f1a8bf103 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -51,13 +51,9 @@ class Storage { public: Storage(llvm::BasicBlock *block, - llvm::Value *out, - llvm::Value *in, llvm::Value *consts, llvm::Value *temps); + llvm::Value *input); llvm::Value *inputPtr() const; - llvm::Value *outputPtr() const; - llvm::Value *constPtr() const; - llvm::Value *tempPtr() const; void setCurrentBlock(llvm::BasicBlock *block); @@ -81,8 +77,7 @@ public: int numConsts() const; - void pushArguments(llvm::Value *out, llvm::Value *in, - llvm::Value *constPtr, llvm::Value *temp); + void pushArguments(llvm::Value *input); void popArguments(); void pushTemps(); void popTemps(); @@ -93,17 +88,25 @@ private: llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); const char *name(const char *prefix); + enum Args { + DestsArg = 0, + InputsArg = 1, + TempsArg = 2, + ConstsArg = 3, + KilArg = 4 + }; + llvm::Value *elemPtr(Args arg); + llvm::Value *elemIdx(llvm::Value *ptr, int idx, + llvm::Value *indIdx = 0); + llvm::Value *element(Args arg, int idx, llvm::Value *indIdx = 0); + private: llvm::BasicBlock *m_block; - llvm::Value *m_OUT; - llvm::Value *m_IN; - llvm::Value *m_CONST; - llvm::Value *m_TEMPS; + llvm::Value *m_INPUT; std::map m_constInts; std::map m_intVecs; std::vector m_addrs; - std::vector m_dstCache; std::vector m_immediates; llvm::VectorType *m_floatVecType; @@ -121,13 +124,7 @@ private: llvm::Value *m_undefIntVec; llvm::Value *m_extSwizzleVec; - struct Args { - llvm::Value *out; - llvm::Value *in; - llvm::Value *cst; - llvm::Value *temp; - }; - std::stack m_argStack; + std::stack m_argStack; std::stack > m_tempStack; }; -- cgit v1.2.3 From 1f30efb7365075de0ae119a40ab70a16bc547670 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 6 Nov 2007 06:06:04 -0500 Subject: Implement kilp and make it work --- src/mesa/pipe/llvm/gallivm.cpp | 14 +- src/mesa/pipe/llvm/gallivm_builtins.cpp | 361 +++++++++++++++++------------- src/mesa/pipe/llvm/instructions.cpp | 10 + src/mesa/pipe/llvm/instructions.h | 1 + src/mesa/pipe/llvm/llvm_base_shader.cpp | 376 ++++++++++++++++---------------- src/mesa/pipe/llvm/llvm_builtins.c | 9 +- src/mesa/pipe/llvm/llvm_entry.c | 6 + src/mesa/pipe/llvm/storage.cpp | 14 ++ src/mesa/pipe/llvm/storage.h | 2 + src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- 10 files changed, 451 insertions(+), 344 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index 65c95074fd..a1101a7ba8 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -432,7 +432,11 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_DDY: break; - case TGSI_OPCODE_KILP: + case TGSI_OPCODE_KILP: { + out = instr->kilp(inputs[0]); + storage->setKilElement(out); + return; + } break; case TGSI_OPCODE_PK2H: break; @@ -929,11 +933,9 @@ int gallivm_fragment_shader_exec(struct gallivm_prog *prog, fragment_shader_runner runner = reinterpret_cast(prog->function); assert(runner); - runner(fx, fy, dests, inputs, prog->num_interp, - consts, prog->num_consts, - samplers); - - return 0; + return runner(fx, fy, dests, inputs, prog->num_interp, + consts, prog->num_consts, + samplers); } void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) diff --git a/src/mesa/pipe/llvm/gallivm_builtins.cpp b/src/mesa/pipe/llvm/gallivm_builtins.cpp index da1e6ae1de..48693ca2ed 100644 --- a/src/mesa/pipe/llvm/gallivm_builtins.cpp +++ b/src/mesa/pipe/llvm/gallivm_builtins.cpp @@ -69,6 +69,15 @@ FunctionType* FuncTy_12 = FunctionType::get( PointerType* PointerTy_11 = PointerType::get(FuncTy_12); +std::vectorFuncTy_13_args; +FuncTy_13_args.push_back(VectorTy_4); +ParamAttrsList *FuncTy_13_PAL = 0; +FunctionType* FuncTy_13 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_13_args, + /*isVarArg=*/false, + /*ParamAttrs=*/FuncTy_13_PAL); + // Function Declarations @@ -132,6 +141,12 @@ Function* func_vsin = new Function( /*Name=*/"vsin", mod); func_vsin->setCallingConv(CallingConv::C); +Function* func_kilp = new Function( + /*Type=*/FuncTy_13, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"kilp", mod); +func_kilp->setCallingConv(CallingConv::C); + // Global Variable Declarations @@ -152,64 +167,64 @@ GlobalVariable* gvar_array__str1 = new GlobalVariable( mod); // Constant Definitions -Constant* const_array_13 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); -Constant* const_array_14 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); -ConstantFP* const_float_15 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); -ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); -Constant* const_float_17 = Constant::getNullValue(Type::FloatTy); -Constant* const_int32_18 = Constant::getNullValue(IntegerType::get(32)); -std::vector const_packed_19_elems; -ConstantFP* const_float_20 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); -const_packed_19_elems.push_back(const_float_20); -UndefValue* const_float_21 = UndefValue::get(Type::FloatTy); -const_packed_19_elems.push_back(const_float_21); -const_packed_19_elems.push_back(const_float_21); -const_packed_19_elems.push_back(const_float_20); -Constant* const_packed_19 = ConstantVector::get(VectorTy_4, const_packed_19_elems); -ConstantInt* const_int32_22 = ConstantInt::get(APInt(32, "1", 10)); -ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "3", 10)); -ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "2", 10)); -std::vector const_packed_25_elems; -const_packed_25_elems.push_back(const_float_20); -const_packed_25_elems.push_back(const_float_17); -const_packed_25_elems.push_back(const_float_17); -const_packed_25_elems.push_back(const_float_20); -Constant* const_packed_25 = ConstantVector::get(VectorTy_4, const_packed_25_elems); -Constant* const_double_26 = Constant::getNullValue(Type::DoubleTy); -std::vector const_packed_27_elems; -const_packed_27_elems.push_back(const_int32_18); -ConstantInt* const_int32_28 = ConstantInt::get(APInt(32, "5", 10)); -const_packed_27_elems.push_back(const_int32_28); -const_packed_27_elems.push_back(const_int32_24); -const_packed_27_elems.push_back(const_int32_23); -Constant* const_packed_27 = ConstantVector::get(VectorTy_7, const_packed_27_elems); -std::vector const_packed_29_elems; -const_packed_29_elems.push_back(const_int32_18); -const_packed_29_elems.push_back(const_int32_22); -ConstantInt* const_int32_30 = ConstantInt::get(APInt(32, "6", 10)); -const_packed_29_elems.push_back(const_int32_30); -const_packed_29_elems.push_back(const_int32_23); -Constant* const_packed_29 = ConstantVector::get(VectorTy_7, const_packed_29_elems); -std::vector const_packed_31_elems; -const_packed_31_elems.push_back(const_int32_18); -const_packed_31_elems.push_back(const_int32_22); -const_packed_31_elems.push_back(const_int32_24); -ConstantInt* const_int32_32 = ConstantInt::get(APInt(32, "7", 10)); -const_packed_31_elems.push_back(const_int32_32); -Constant* const_packed_31 = ConstantVector::get(VectorTy_7, const_packed_31_elems); -std::vector const_ptr_33_indices; -const_ptr_33_indices.push_back(const_int32_18); -const_ptr_33_indices.push_back(const_int32_18); -Constant* const_ptr_33 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_33_indices[0], const_ptr_33_indices.size() ); -UndefValue* const_packed_34 = UndefValue::get(VectorTy_4); -std::vector const_ptr_35_indices; -const_ptr_35_indices.push_back(const_int32_18); -const_ptr_35_indices.push_back(const_int32_18); -Constant* const_ptr_35 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_35_indices[0], const_ptr_35_indices.size() ); +Constant* const_array_14 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); +Constant* const_array_15 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); +ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); +ConstantFP* const_float_17 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); +Constant* const_float_18 = Constant::getNullValue(Type::FloatTy); +Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); +std::vector const_packed_20_elems; +ConstantFP* const_float_21 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); +const_packed_20_elems.push_back(const_float_21); +UndefValue* const_float_22 = UndefValue::get(Type::FloatTy); +const_packed_20_elems.push_back(const_float_22); +const_packed_20_elems.push_back(const_float_22); +const_packed_20_elems.push_back(const_float_21); +Constant* const_packed_20 = ConstantVector::get(VectorTy_4, const_packed_20_elems); +ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); +ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "3", 10)); +ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "2", 10)); +std::vector const_packed_26_elems; +const_packed_26_elems.push_back(const_float_21); +const_packed_26_elems.push_back(const_float_18); +const_packed_26_elems.push_back(const_float_18); +const_packed_26_elems.push_back(const_float_21); +Constant* const_packed_26 = ConstantVector::get(VectorTy_4, const_packed_26_elems); +Constant* const_double_27 = Constant::getNullValue(Type::DoubleTy); +std::vector const_packed_28_elems; +const_packed_28_elems.push_back(const_int32_19); +ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "5", 10)); +const_packed_28_elems.push_back(const_int32_29); +const_packed_28_elems.push_back(const_int32_25); +const_packed_28_elems.push_back(const_int32_24); +Constant* const_packed_28 = ConstantVector::get(VectorTy_7, const_packed_28_elems); +std::vector const_packed_30_elems; +const_packed_30_elems.push_back(const_int32_19); +const_packed_30_elems.push_back(const_int32_23); +ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "6", 10)); +const_packed_30_elems.push_back(const_int32_31); +const_packed_30_elems.push_back(const_int32_24); +Constant* const_packed_30 = ConstantVector::get(VectorTy_7, const_packed_30_elems); +std::vector const_packed_32_elems; +const_packed_32_elems.push_back(const_int32_19); +const_packed_32_elems.push_back(const_int32_23); +const_packed_32_elems.push_back(const_int32_25); +ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "7", 10)); +const_packed_32_elems.push_back(const_int32_33); +Constant* const_packed_32 = ConstantVector::get(VectorTy_7, const_packed_32_elems); +std::vector const_ptr_34_indices; +const_ptr_34_indices.push_back(const_int32_19); +const_ptr_34_indices.push_back(const_int32_19); +Constant* const_ptr_34 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_34_indices[0], const_ptr_34_indices.size() ); +UndefValue* const_packed_35 = UndefValue::get(VectorTy_4); +std::vector const_ptr_36_indices; +const_ptr_36_indices.push_back(const_int32_19); +const_ptr_36_indices.push_back(const_int32_19); +Constant* const_ptr_36 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_36_indices[0], const_ptr_36_indices.size() ); // Global Variable Definitions -gvar_array__str->setInitializer(const_array_13); -gvar_array__str1->setInitializer(const_array_14); +gvar_array__str->setInitializer(const_array_14); +gvar_array__str1->setInitializer(const_array_15); // Function Definitions @@ -224,12 +239,12 @@ gvar_array__str1->setInitializer(const_array_14); BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); // Block entry (label_entry) - FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_15, "cmp", label_entry); - SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_15, float_b, "b.addr.0", label_entry); - FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_16, "cmp3", label_entry); - SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_16, float_b_addr_0, "b.addr.1", label_entry); - FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_17, "cmp7", label_entry); - SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_17, float_a, "a.addr.0", label_entry); + FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_16, "cmp", label_entry); + SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_16, float_b, "b.addr.0", label_entry); + FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_17, "cmp3", label_entry); + SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_17, float_b_addr_0, "b.addr.1", label_entry); + FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_18, "cmp7", label_entry); + SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_18, float_a, "a.addr.0", label_entry); std::vector float_call_params; float_call_params.push_back(float_a_addr_0); float_call_params.push_back(float_b_addr_1); @@ -246,36 +261,36 @@ gvar_array__str1->setInitializer(const_array_14); Value* packed_tmp = args++; packed_tmp->setName("tmp"); - BasicBlock* label_entry_37 = new BasicBlock("entry",func_lit,0); + BasicBlock* label_entry_38 = new BasicBlock("entry",func_lit,0); BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); - // Block entry (label_entry_37) - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_18, "tmp7", label_entry_37); - FCmpInst* int1_cmp_38 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_17, "cmp", label_entry_37); - new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_38, label_entry_37); + // Block entry (label_entry_38) + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp7", label_entry_38); + FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_18, "cmp", label_entry_38); + new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_39, label_entry_38); // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_19, float_tmp7, const_int32_22, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_22, "tmp14", label_ifthen); - ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp16", label_ifthen); - FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_15, "cmp.i", label_ifthen); - SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_15, float_tmp16, "b.addr.0.i", label_ifthen); - FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_16, "cmp3.i", label_ifthen); - SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_16, float_b_addr_0_i, "b.addr.1.i", label_ifthen); - FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_17, "cmp7.i", label_ifthen); - SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_17, float_tmp14, "a.addr.0.i", label_ifthen); + InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_20, float_tmp7, const_int32_23, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp14", label_ifthen); + ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp16", label_ifthen); + FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_16, "cmp.i", label_ifthen); + SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_16, float_tmp16, "b.addr.0.i", label_ifthen); + FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_17, "cmp3.i", label_ifthen); + SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_17, float_b_addr_0_i, "b.addr.1.i", label_ifthen); + FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_18, "cmp7.i", label_ifthen); + SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_18, float_tmp14, "a.addr.0.i", label_ifthen); std::vector float_call_i_params; float_call_i_params.push_back(float_a_addr_0_i); float_call_i_params.push_back(float_b_addr_1_i); CallInst* float_call_i = new CallInst(func_powf, float_call_i_params.begin(), float_call_i_params.end(), "call.i", label_ifthen); float_call_i->setCallingConv(CallingConv::C); float_call_i->setTailCall(true); - InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_24, "tmp18", label_ifthen); + InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_25, "tmp18", label_ifthen); new ReturnInst(packed_tmp18, label_ifthen); // Block UnifiedReturnBlock (label_UnifiedReturnBlock) - new ReturnInst(const_packed_25, label_UnifiedReturnBlock); + new ReturnInst(const_packed_26, label_UnifiedReturnBlock); } @@ -289,7 +304,7 @@ gvar_array__str1->setInitializer(const_array_14); Value* packed_tmp2 = args++; packed_tmp2->setName("tmp2"); - BasicBlock* label_entry_42 = new BasicBlock("entry",func_cmp,0); + BasicBlock* label_entry_43 = new BasicBlock("entry",func_cmp,0); BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); @@ -297,28 +312,28 @@ gvar_array__str1->setInitializer(const_array_14); BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); - // Block entry (label_entry_42) - ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_18, "tmp3", label_entry_42); - CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_42); - FCmpInst* int1_cmp_43 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_26, "cmp", label_entry_42); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_22, "tmp11", label_entry_42); - CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_42); - FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_26, "cmp13", label_entry_42); - SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_43, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_42); - new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_42); + // Block entry (label_entry_43) + ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_43); + CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_43); + FCmpInst* int1_cmp_44 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_43); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_43); + CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_43); + FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_43); + SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_44, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_43); + new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_43); // Block cond.?14 (label_cond__14) - ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_27, "tmp233", label_cond__14); - ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp254", label_cond__14); + ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_28, "tmp233", label_cond__14); + ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp254", label_cond__14); CastInst* double_conv265 = new FPExtInst(float_tmp254, Type::DoubleTy, "conv265", label_cond__14); - FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_26, "cmp276", label_cond__14); + FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_27, "cmp276", label_cond__14); new BranchInst(label_cond__28, label_cond_cont34, int1_cmp276, label_cond__14); // Block cond.cont20 (label_cond_cont20) - ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_27, "tmp23", label_cond_cont20); - ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp25", label_cond_cont20); + ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_28, "tmp23", label_cond_cont20); + ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp25", label_cond_cont20); CastInst* double_conv26 = new FPExtInst(float_tmp25, Type::DoubleTy, "conv26", label_cond_cont20); - FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_26, "cmp27", label_cond_cont20); + FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_27, "cmp27", label_cond_cont20); new BranchInst(label_cond__28, label_cond_cont34, int1_cmp27, label_cond_cont20); // Block cond.?28 (label_cond__28) @@ -327,10 +342,10 @@ gvar_array__str1->setInitializer(const_array_14); packed_tmp23_reg2mem_0->addIncoming(packed_tmp233, label_cond__14); packed_tmp23_reg2mem_0->addIncoming(packed_tmp23, label_cond_cont20); - ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_29, "tmp378", label_cond__28); - ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp399", label_cond__28); + ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_30, "tmp378", label_cond__28); + ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp399", label_cond__28); CastInst* double_conv4010 = new FPExtInst(float_tmp399, Type::DoubleTy, "conv4010", label_cond__28); - FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_26, "cmp4111", label_cond__28); + FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_27, "cmp4111", label_cond__28); new BranchInst(label_cond__42, label_cond_cont48, int1_cmp4111, label_cond__28); // Block cond.cont34 (label_cond_cont34) @@ -339,10 +354,10 @@ gvar_array__str1->setInitializer(const_array_14); packed_tmp23_reg2mem_1->addIncoming(packed_tmp233, label_cond__14); packed_tmp23_reg2mem_1->addIncoming(packed_tmp23, label_cond_cont20); - ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_29, "tmp37", label_cond_cont34); - ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp39", label_cond_cont34); + ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_30, "tmp37", label_cond_cont34); + ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp39", label_cond_cont34); CastInst* double_conv40 = new FPExtInst(float_tmp39, Type::DoubleTy, "conv40", label_cond_cont34); - FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_26, "cmp41", label_cond_cont34); + FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_27, "cmp41", label_cond_cont34); new BranchInst(label_cond__42, label_cond_cont48, int1_cmp41, label_cond_cont34); // Block cond.?42 (label_cond__42) @@ -351,7 +366,7 @@ gvar_array__str1->setInitializer(const_array_14); packed_tmp37_reg2mem_0->addIncoming(packed_tmp378, label_cond__28); packed_tmp37_reg2mem_0->addIncoming(packed_tmp37, label_cond_cont34); - ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_31, "tmp5113", label_cond__42); + ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_32, "tmp5113", label_cond__42); new ReturnInst(packed_tmp5113, label_cond__42); // Block cond.cont48 (label_cond_cont48) @@ -360,7 +375,7 @@ gvar_array__str1->setInitializer(const_array_14); packed_tmp37_reg2mem_1->addIncoming(packed_tmp378, label_cond__28); packed_tmp37_reg2mem_1->addIncoming(packed_tmp37, label_cond_cont34); - ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_31, "tmp51", label_cond_cont48); + ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_32, "tmp51", label_cond_cont48); new ReturnInst(packed_tmp51, label_cond_cont48); } @@ -371,87 +386,125 @@ gvar_array__str1->setInitializer(const_array_14); Value* packed_val = args++; packed_val->setName("val"); - BasicBlock* label_entry_51 = new BasicBlock("entry",func_vcos,0); - - // Block entry (label_entry_51) - ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_18, "tmp1", label_entry_51); - CastInst* double_conv_52 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_51); - ExtractElementInst* float_tmp3_53 = new ExtractElementInst(packed_val, const_int32_22, "tmp3", label_entry_51); - CastInst* double_conv4 = new FPExtInst(float_tmp3_53, Type::DoubleTy, "conv4", label_entry_51); - ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_val, const_int32_24, "tmp6", label_entry_51); - CastInst* double_conv7 = new FPExtInst(float_tmp6, Type::DoubleTy, "conv7", label_entry_51); - ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_23, "tmp9", label_entry_51); - CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_51); + BasicBlock* label_entry_52 = new BasicBlock("entry",func_vcos,0); + + // Block entry (label_entry_52) + ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_52); + CastInst* double_conv_53 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_52); + ExtractElementInst* float_tmp3_54 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_52); + CastInst* double_conv4 = new FPExtInst(float_tmp3_54, Type::DoubleTy, "conv4", label_entry_52); + ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_52); + CastInst* double_conv7 = new FPExtInst(float_tmp6, Type::DoubleTy, "conv7", label_entry_52); + ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_52); + CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_52); std::vector int32_call_params; - int32_call_params.push_back(const_ptr_33); - int32_call_params.push_back(double_conv_52); + int32_call_params.push_back(const_ptr_34); + int32_call_params.push_back(double_conv_53); int32_call_params.push_back(double_conv4); int32_call_params.push_back(double_conv7); int32_call_params.push_back(double_conv10); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_51); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_52); int32_call->setCallingConv(CallingConv::C); int32_call->setTailCall(true); - CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_51); + CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_52); float_call13->setCallingConv(CallingConv::C); float_call13->setTailCall(true); - InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_34, float_call13, const_int32_18, "tmp15", label_entry_51); - InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call13, const_int32_22, "tmp20", label_entry_51); - InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call13, const_int32_24, "tmp25", label_entry_51); - InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call13, const_int32_23, "tmp30", label_entry_51); - CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_51); + InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_52); + InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call13, const_int32_23, "tmp20", label_entry_52); + InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call13, const_int32_25, "tmp25", label_entry_52); + InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call13, const_int32_24, "tmp30", label_entry_52); + CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_52); std::vector int32_call43_params; - int32_call43_params.push_back(const_ptr_35); + int32_call43_params.push_back(const_ptr_36); int32_call43_params.push_back(double_conv33); int32_call43_params.push_back(double_conv33); int32_call43_params.push_back(double_conv33); int32_call43_params.push_back(double_conv33); - CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_51); + CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_52); int32_call43->setCallingConv(CallingConv::C); int32_call43->setTailCall(true); - new ReturnInst(packed_tmp30, label_entry_51); + new ReturnInst(packed_tmp30, label_entry_52); } // Function: scs (func_scs) { Function::arg_iterator args = func_scs->arg_begin(); - Value* packed_val_55 = args++; - packed_val_55->setName("val"); - - BasicBlock* label_entry_56 = new BasicBlock("entry",func_scs,0); - - // Block entry (label_entry_56) - ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_55, const_int32_18, "tmp2", label_entry_56); - CallInst* float_call_57 = new CallInst(func_cosf, float_tmp2, "call", label_entry_56); - float_call_57->setCallingConv(CallingConv::C); - float_call_57->setTailCall(true); - InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_34, float_call_57, const_int32_18, "tmp5", label_entry_56); - CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_56); + Value* packed_val_56 = args++; + packed_val_56->setName("val"); + + BasicBlock* label_entry_57 = new BasicBlock("entry",func_scs,0); + + // Block entry (label_entry_57) + ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_56, const_int32_19, "tmp2", label_entry_57); + CallInst* float_call_58 = new CallInst(func_cosf, float_tmp2, "call", label_entry_57); + float_call_58->setCallingConv(CallingConv::C); + float_call_58->setTailCall(true); + InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_58, const_int32_19, "tmp5", label_entry_57); + CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_57); float_call7->setCallingConv(CallingConv::C); float_call7->setTailCall(true); - InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_22, "tmp9", label_entry_56); - new ReturnInst(packed_tmp9, label_entry_56); + InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_57); + new ReturnInst(packed_tmp9, label_entry_57); } // Function: vsin (func_vsin) { Function::arg_iterator args = func_vsin->arg_begin(); - Value* packed_val_59 = args++; - packed_val_59->setName("val"); - - BasicBlock* label_entry_60 = new BasicBlock("entry",func_vsin,0); - - // Block entry (label_entry_60) - ExtractElementInst* float_tmp2_61 = new ExtractElementInst(packed_val_59, const_int32_18, "tmp2", label_entry_60); - CallInst* float_call_62 = new CallInst(func_sinf, float_tmp2_61, "call", label_entry_60); - float_call_62->setCallingConv(CallingConv::C); - float_call_62->setTailCall(true); - InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_34, float_call_62, const_int32_18, "tmp6", label_entry_60); - InsertElementInst* packed_tmp9_63 = new InsertElementInst(packed_tmp6, float_call_62, const_int32_22, "tmp9", label_entry_60); - InsertElementInst* packed_tmp12_64 = new InsertElementInst(packed_tmp9_63, float_call_62, const_int32_24, "tmp12", label_entry_60); - InsertElementInst* packed_tmp15_65 = new InsertElementInst(packed_tmp12_64, float_call_62, const_int32_23, "tmp15", label_entry_60); - new ReturnInst(packed_tmp15_65, label_entry_60); + Value* packed_val_60 = args++; + packed_val_60->setName("val"); + + BasicBlock* label_entry_61 = new BasicBlock("entry",func_vsin,0); + + // Block entry (label_entry_61) + ExtractElementInst* float_tmp2_62 = new ExtractElementInst(packed_val_60, const_int32_19, "tmp2", label_entry_61); + CallInst* float_call_63 = new CallInst(func_sinf, float_tmp2_62, "call", label_entry_61); + float_call_63->setCallingConv(CallingConv::C); + float_call_63->setTailCall(true); + InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_63, const_int32_19, "tmp6", label_entry_61); + InsertElementInst* packed_tmp9_64 = new InsertElementInst(packed_tmp6, float_call_63, const_int32_23, "tmp9", label_entry_61); + InsertElementInst* packed_tmp12_65 = new InsertElementInst(packed_tmp9_64, float_call_63, const_int32_25, "tmp12", label_entry_61); + InsertElementInst* packed_tmp15_66 = new InsertElementInst(packed_tmp12_65, float_call_63, const_int32_24, "tmp15", label_entry_61); + new ReturnInst(packed_tmp15_66, label_entry_61); + +} + +// Function: kilp (func_kilp) +{ + Function::arg_iterator args = func_kilp->arg_begin(); + Value* packed_val_68 = args++; + packed_val_68->setName("val"); + + BasicBlock* label_entry_69 = new BasicBlock("entry",func_kilp,0); + BasicBlock* label_lor_rhs = new BasicBlock("lor_rhs",func_kilp,0); + BasicBlock* label_lor_rhs6 = new BasicBlock("lor_rhs6",func_kilp,0); + BasicBlock* label_lor_rhs13 = new BasicBlock("lor_rhs13",func_kilp,0); + BasicBlock* label_UnifiedReturnBlock_70 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); + + // Block entry (label_entry_69) + ExtractElementInst* float_tmp1_71 = new ExtractElementInst(packed_val_68, const_int32_19, "tmp1", label_entry_69); + FCmpInst* int1_cmp_72 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_71, const_float_18, "cmp", label_entry_69); + new BranchInst(label_UnifiedReturnBlock_70, label_lor_rhs, int1_cmp_72, label_entry_69); + + // Block lor_rhs (label_lor_rhs) + ExtractElementInst* float_tmp3_74 = new ExtractElementInst(packed_val_68, const_int32_23, "tmp3", label_lor_rhs); + FCmpInst* int1_cmp5 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_74, const_float_18, "cmp5", label_lor_rhs); + new BranchInst(label_UnifiedReturnBlock_70, label_lor_rhs6, int1_cmp5, label_lor_rhs); + + // Block lor_rhs6 (label_lor_rhs6) + ExtractElementInst* float_tmp8 = new ExtractElementInst(packed_val_68, const_int32_25, "tmp8", label_lor_rhs6); + FCmpInst* int1_cmp10 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp8, const_float_18, "cmp10", label_lor_rhs6); + new BranchInst(label_UnifiedReturnBlock_70, label_lor_rhs13, int1_cmp10, label_lor_rhs6); + + // Block lor_rhs13 (label_lor_rhs13) + ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_val_68, const_int32_24, "tmp15", label_lor_rhs13); + FCmpInst* int1_cmp17 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp15, const_float_18, "cmp17", label_lor_rhs13); + CastInst* int32_retval = new ZExtInst(int1_cmp17, IntegerType::get(32), "retval", label_lor_rhs13); + new ReturnInst(int32_retval, label_lor_rhs13); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock_70) + new ReturnInst(const_int32_23, label_UnifiedReturnBlock_70); } diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index 7a70aec878..c8d1992587 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -864,6 +864,15 @@ llvm::Value * Instructions::scs(llvm::Value *in) return call; } +llvm::Value * Instructions::kilp(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("kilp"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("kilpres")); + call->setTailCall(false); + return call; +} llvm::Value * Instructions::sin(llvm::Value *in) { @@ -876,3 +885,4 @@ llvm::Value * Instructions::sin(llvm::Value *in) } #endif //MESA_LLVM + diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h index c31cc4f125..9ebc17dd8e 100644 --- a/src/mesa/pipe/llvm/instructions.h +++ b/src/mesa/pipe/llvm/instructions.h @@ -79,6 +79,7 @@ public: llvm::Value *floor(llvm::Value *in); llvm::Value *frc(llvm::Value *in); void ifop(llvm::Value *in); + llvm::Value *kilp(llvm::Value *in); llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); llvm::Value *lit(llvm::Value *in); diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index 6e7fa32807..82ad6cfa26 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -810,246 +810,257 @@ Module* createBaseShader() { AllocaInst* ptr_results_142 = new AllocaInst(ArrayTy_33, "results", label_entry_134); AllocaInst* ptr_temps_143 = new AllocaInst(ArrayTy_25, "temps", label_entry_134); AllocaInst* ptr_args_144 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_134); - ICmpInst* int1_cmp5_i_145 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs_131, const_int32_34, "cmp5.i", label_entry_134); - new BranchInst(label_forbody6_i_135, label_from_array_exit_136, int1_cmp5_i_145, label_entry_134); + std::vector ptr_tmp_indices; + ptr_tmp_indices.push_back(const_int32_34); + ptr_tmp_indices.push_back(const_int32_39); + Instruction* ptr_tmp = new GetElementPtrInst(ptr_args_144, ptr_tmp_indices.begin(), ptr_tmp_indices.end(), "tmp", label_entry_134); + StoreInst* void_145 = new StoreInst(const_int32_34, ptr_tmp, false, label_entry_134); + ICmpInst* int1_cmp5_i_146 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs_131, const_int32_34, "cmp5.i", label_entry_134); + new BranchInst(label_forbody6_i_135, label_from_array_exit_136, int1_cmp5_i_146, label_entry_134); // Block forbody6.i (label_forbody6_i_135) - Argument* fwdref_148 = new Argument(IntegerType::get(32)); - PHINode* int32_j_0_reg2mem_0_i_147 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i_135); - int32_j_0_reg2mem_0_i_147->reserveOperandSpace(2); - int32_j_0_reg2mem_0_i_147->addIncoming(const_int32_34, label_entry_134); - int32_j_0_reg2mem_0_i_147->addIncoming(fwdref_148, label_forbody6_i_135); - - Argument* fwdref_150 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i_149 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody6_i_135); - packed_vec_0_reg2mem_0_i_149->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_149->addIncoming(const_packed_35, label_entry_134); - packed_vec_0_reg2mem_0_i_149->addIncoming(fwdref_150, label_forbody6_i_135); - - std::vector ptr_arraydecay11_i_151_indices; - ptr_arraydecay11_i_151_indices.push_back(const_int32_34); - ptr_arraydecay11_i_151_indices.push_back(int32_j_0_reg2mem_0_i_147); - ptr_arraydecay11_i_151_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay11_i_151 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_151_indices.begin(), ptr_arraydecay11_i_151_indices.end(), "arraydecay11.i", label_forbody6_i_135); - LoadInst* float_tmp13_i_152 = new LoadInst(ptr_arraydecay11_i_151, "tmp13.i", false, label_forbody6_i_135); - InsertElementInst* packed_tmp15_i_153 = new InsertElementInst(packed_vec_0_reg2mem_0_i_149, float_tmp13_i_152, const_int32_34, "tmp15.i", label_forbody6_i_135); - std::vector ptr_arrayidx23_i_154_indices; - ptr_arrayidx23_i_154_indices.push_back(const_int32_34); - ptr_arrayidx23_i_154_indices.push_back(int32_j_0_reg2mem_0_i_147); - ptr_arrayidx23_i_154_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx23_i_154 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_154_indices.begin(), ptr_arrayidx23_i_154_indices.end(), "arrayidx23.i", label_forbody6_i_135); - LoadInst* float_tmp24_i_155 = new LoadInst(ptr_arrayidx23_i_154, "tmp24.i", false, label_forbody6_i_135); - InsertElementInst* packed_tmp26_i_156 = new InsertElementInst(packed_tmp15_i_153, float_tmp24_i_155, const_int32_36, "tmp26.i", label_forbody6_i_135); - std::vector ptr_arrayidx34_i_157_indices; - ptr_arrayidx34_i_157_indices.push_back(const_int32_34); - ptr_arrayidx34_i_157_indices.push_back(int32_j_0_reg2mem_0_i_147); - ptr_arrayidx34_i_157_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx34_i_157 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_157_indices.begin(), ptr_arrayidx34_i_157_indices.end(), "arrayidx34.i", label_forbody6_i_135); - LoadInst* float_tmp35_i_158 = new LoadInst(ptr_arrayidx34_i_157, "tmp35.i", false, label_forbody6_i_135); - InsertElementInst* packed_tmp37_i_159 = new InsertElementInst(packed_tmp26_i_156, float_tmp35_i_158, const_int32_37, "tmp37.i", label_forbody6_i_135); - std::vector ptr_arrayidx45_i_160_indices; - ptr_arrayidx45_i_160_indices.push_back(const_int32_34); - ptr_arrayidx45_i_160_indices.push_back(int32_j_0_reg2mem_0_i_147); - ptr_arrayidx45_i_160_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx45_i_160 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_160_indices.begin(), ptr_arrayidx45_i_160_indices.end(), "arrayidx45.i", label_forbody6_i_135); - LoadInst* float_tmp46_i_161 = new LoadInst(ptr_arrayidx45_i_160, "tmp46.i", false, label_forbody6_i_135); - InsertElementInst* packed_tmp48_i_162 = new InsertElementInst(packed_tmp37_i_159, float_tmp46_i_161, const_int32_38, "tmp48.i", label_forbody6_i_135); - std::vector ptr_arrayidx54_i_163_indices; - ptr_arrayidx54_i_163_indices.push_back(const_int32_34); - ptr_arrayidx54_i_163_indices.push_back(const_int32_34); - ptr_arrayidx54_i_163_indices.push_back(int32_j_0_reg2mem_0_i_147); - Instruction* ptr_arrayidx54_i_163 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_163_indices.begin(), ptr_arrayidx54_i_163_indices.end(), "arrayidx54.i", label_forbody6_i_135); - StoreInst* void_164 = new StoreInst(packed_tmp48_i_162, ptr_arrayidx54_i_163, false, label_forbody6_i_135); - BinaryOperator* int32_inc_i_165 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_147, const_int32_36, "inc.i", label_forbody6_i_135); - ICmpInst* int1_cmp59_i_166 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_165, int32_num_inputs_131, "cmp59.i", label_forbody6_i_135); - new BranchInst(label_forbody6_i_135, label_forbody6_i_1, int1_cmp59_i_166, label_forbody6_i_135); + Argument* fwdref_149 = new Argument(IntegerType::get(32)); + PHINode* int32_j_0_reg2mem_0_i_148 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i_135); + int32_j_0_reg2mem_0_i_148->reserveOperandSpace(2); + int32_j_0_reg2mem_0_i_148->addIncoming(const_int32_34, label_entry_134); + int32_j_0_reg2mem_0_i_148->addIncoming(fwdref_149, label_forbody6_i_135); + + Argument* fwdref_151 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i_150 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody6_i_135); + packed_vec_0_reg2mem_0_i_150->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_150->addIncoming(const_packed_35, label_entry_134); + packed_vec_0_reg2mem_0_i_150->addIncoming(fwdref_151, label_forbody6_i_135); + + std::vector ptr_arraydecay11_i_152_indices; + ptr_arraydecay11_i_152_indices.push_back(const_int32_34); + ptr_arraydecay11_i_152_indices.push_back(int32_j_0_reg2mem_0_i_148); + ptr_arraydecay11_i_152_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay11_i_152 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_152_indices.begin(), ptr_arraydecay11_i_152_indices.end(), "arraydecay11.i", label_forbody6_i_135); + LoadInst* float_tmp13_i_153 = new LoadInst(ptr_arraydecay11_i_152, "tmp13.i", false, label_forbody6_i_135); + InsertElementInst* packed_tmp15_i_154 = new InsertElementInst(packed_vec_0_reg2mem_0_i_150, float_tmp13_i_153, const_int32_34, "tmp15.i", label_forbody6_i_135); + std::vector ptr_arrayidx23_i_155_indices; + ptr_arrayidx23_i_155_indices.push_back(const_int32_34); + ptr_arrayidx23_i_155_indices.push_back(int32_j_0_reg2mem_0_i_148); + ptr_arrayidx23_i_155_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx23_i_155 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_155_indices.begin(), ptr_arrayidx23_i_155_indices.end(), "arrayidx23.i", label_forbody6_i_135); + LoadInst* float_tmp24_i_156 = new LoadInst(ptr_arrayidx23_i_155, "tmp24.i", false, label_forbody6_i_135); + InsertElementInst* packed_tmp26_i_157 = new InsertElementInst(packed_tmp15_i_154, float_tmp24_i_156, const_int32_36, "tmp26.i", label_forbody6_i_135); + std::vector ptr_arrayidx34_i_158_indices; + ptr_arrayidx34_i_158_indices.push_back(const_int32_34); + ptr_arrayidx34_i_158_indices.push_back(int32_j_0_reg2mem_0_i_148); + ptr_arrayidx34_i_158_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx34_i_158 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_158_indices.begin(), ptr_arrayidx34_i_158_indices.end(), "arrayidx34.i", label_forbody6_i_135); + LoadInst* float_tmp35_i_159 = new LoadInst(ptr_arrayidx34_i_158, "tmp35.i", false, label_forbody6_i_135); + InsertElementInst* packed_tmp37_i_160 = new InsertElementInst(packed_tmp26_i_157, float_tmp35_i_159, const_int32_37, "tmp37.i", label_forbody6_i_135); + std::vector ptr_arrayidx45_i_161_indices; + ptr_arrayidx45_i_161_indices.push_back(const_int32_34); + ptr_arrayidx45_i_161_indices.push_back(int32_j_0_reg2mem_0_i_148); + ptr_arrayidx45_i_161_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx45_i_161 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_161_indices.begin(), ptr_arrayidx45_i_161_indices.end(), "arrayidx45.i", label_forbody6_i_135); + LoadInst* float_tmp46_i_162 = new LoadInst(ptr_arrayidx45_i_161, "tmp46.i", false, label_forbody6_i_135); + InsertElementInst* packed_tmp48_i_163 = new InsertElementInst(packed_tmp37_i_160, float_tmp46_i_162, const_int32_38, "tmp48.i", label_forbody6_i_135); + std::vector ptr_arrayidx54_i_164_indices; + ptr_arrayidx54_i_164_indices.push_back(const_int32_34); + ptr_arrayidx54_i_164_indices.push_back(const_int32_34); + ptr_arrayidx54_i_164_indices.push_back(int32_j_0_reg2mem_0_i_148); + Instruction* ptr_arrayidx54_i_164 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_164_indices.begin(), ptr_arrayidx54_i_164_indices.end(), "arrayidx54.i", label_forbody6_i_135); + StoreInst* void_165 = new StoreInst(packed_tmp48_i_163, ptr_arrayidx54_i_164, false, label_forbody6_i_135); + BinaryOperator* int32_inc_i_166 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_148, const_int32_36, "inc.i", label_forbody6_i_135); + ICmpInst* int1_cmp59_i_167 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_166, int32_num_inputs_131, "cmp59.i", label_forbody6_i_135); + new BranchInst(label_forbody6_i_135, label_forbody6_i_1, int1_cmp59_i_167, label_forbody6_i_135); // Block from_array.exit (label_from_array_exit_136) - ICmpInst* int1_cmp_i_168 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_133, const_int32_34, "cmp.i", label_from_array_exit_136); - new BranchInst(label_forbody_i13, label_from_consts_exit_137, int1_cmp_i_168, label_from_array_exit_136); + ICmpInst* int1_cmp_i_169 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_133, const_int32_34, "cmp.i", label_from_array_exit_136); + new BranchInst(label_forbody_i13, label_from_consts_exit_137, int1_cmp_i_169, label_from_array_exit_136); // Block forbody.i13 (label_forbody_i13) - Argument* fwdref_170 = new Argument(IntegerType::get(32)); + Argument* fwdref_171 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i3 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i3", label_forbody_i13); int32_i_0_reg2mem_0_i3->reserveOperandSpace(2); int32_i_0_reg2mem_0_i3->addIncoming(const_int32_34, label_from_array_exit_136); - int32_i_0_reg2mem_0_i3->addIncoming(fwdref_170, label_forbody_i13); + int32_i_0_reg2mem_0_i3->addIncoming(fwdref_171, label_forbody_i13); - Argument* fwdref_171 = new Argument(VectorTy_1); + Argument* fwdref_172 = new Argument(VectorTy_1); PHINode* packed_vec_0_reg2mem_0_i4 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i4", label_forbody_i13); packed_vec_0_reg2mem_0_i4->reserveOperandSpace(2); packed_vec_0_reg2mem_0_i4->addIncoming(const_packed_35, label_from_array_exit_136); - packed_vec_0_reg2mem_0_i4->addIncoming(fwdref_171, label_forbody_i13); + packed_vec_0_reg2mem_0_i4->addIncoming(fwdref_172, label_forbody_i13); std::vector ptr_arraydecay_i5_indices; ptr_arraydecay_i5_indices.push_back(int32_i_0_reg2mem_0_i3); ptr_arraydecay_i5_indices.push_back(const_int32_34); Instruction* ptr_arraydecay_i5 = new GetElementPtrInst(ptr_aconsts_132, ptr_arraydecay_i5_indices.begin(), ptr_arraydecay_i5_indices.end(), "arraydecay.i5", label_forbody_i13); - LoadInst* float_tmp5_i_172 = new LoadInst(ptr_arraydecay_i5, "tmp5.i", false, label_forbody_i13); - InsertElementInst* packed_tmp7_i6 = new InsertElementInst(packed_vec_0_reg2mem_0_i4, float_tmp5_i_172, const_int32_34, "tmp7.i6", label_forbody_i13); - std::vector ptr_arrayidx12_i_173_indices; - ptr_arrayidx12_i_173_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arrayidx12_i_173_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx12_i_173 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx12_i_173_indices.begin(), ptr_arrayidx12_i_173_indices.end(), "arrayidx12.i", label_forbody_i13); - LoadInst* float_tmp13_i7 = new LoadInst(ptr_arrayidx12_i_173, "tmp13.i7", false, label_forbody_i13); + LoadInst* float_tmp5_i_173 = new LoadInst(ptr_arraydecay_i5, "tmp5.i", false, label_forbody_i13); + InsertElementInst* packed_tmp7_i6 = new InsertElementInst(packed_vec_0_reg2mem_0_i4, float_tmp5_i_173, const_int32_34, "tmp7.i6", label_forbody_i13); + std::vector ptr_arrayidx12_i_174_indices; + ptr_arrayidx12_i_174_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arrayidx12_i_174_indices.push_back(const_int32_36); + Instruction* ptr_arrayidx12_i_174 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx12_i_174_indices.begin(), ptr_arrayidx12_i_174_indices.end(), "arrayidx12.i", label_forbody_i13); + LoadInst* float_tmp13_i7 = new LoadInst(ptr_arrayidx12_i_174, "tmp13.i7", false, label_forbody_i13); InsertElementInst* packed_tmp15_i8 = new InsertElementInst(packed_tmp7_i6, float_tmp13_i7, const_int32_36, "tmp15.i8", label_forbody_i13); - std::vector ptr_arrayidx20_i_174_indices; - ptr_arrayidx20_i_174_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arrayidx20_i_174_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx20_i_174 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx20_i_174_indices.begin(), ptr_arrayidx20_i_174_indices.end(), "arrayidx20.i", label_forbody_i13); - LoadInst* float_tmp21_i_175 = new LoadInst(ptr_arrayidx20_i_174, "tmp21.i", false, label_forbody_i13); - InsertElementInst* packed_tmp23_i9 = new InsertElementInst(packed_tmp15_i8, float_tmp21_i_175, const_int32_37, "tmp23.i9", label_forbody_i13); - std::vector ptr_arrayidx28_i_176_indices; - ptr_arrayidx28_i_176_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arrayidx28_i_176_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx28_i_176 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx28_i_176_indices.begin(), ptr_arrayidx28_i_176_indices.end(), "arrayidx28.i", label_forbody_i13); - LoadInst* float_tmp29_i_177 = new LoadInst(ptr_arrayidx28_i_176, "tmp29.i", false, label_forbody_i13); - InsertElementInst* packed_tmp31_i_178 = new InsertElementInst(packed_tmp23_i9, float_tmp29_i_177, const_int32_38, "tmp31.i", label_forbody_i13); + std::vector ptr_arrayidx20_i_175_indices; + ptr_arrayidx20_i_175_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arrayidx20_i_175_indices.push_back(const_int32_37); + Instruction* ptr_arrayidx20_i_175 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx20_i_175_indices.begin(), ptr_arrayidx20_i_175_indices.end(), "arrayidx20.i", label_forbody_i13); + LoadInst* float_tmp21_i_176 = new LoadInst(ptr_arrayidx20_i_175, "tmp21.i", false, label_forbody_i13); + InsertElementInst* packed_tmp23_i9 = new InsertElementInst(packed_tmp15_i8, float_tmp21_i_176, const_int32_37, "tmp23.i9", label_forbody_i13); + std::vector ptr_arrayidx28_i_177_indices; + ptr_arrayidx28_i_177_indices.push_back(int32_i_0_reg2mem_0_i3); + ptr_arrayidx28_i_177_indices.push_back(const_int32_38); + Instruction* ptr_arrayidx28_i_177 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx28_i_177_indices.begin(), ptr_arrayidx28_i_177_indices.end(), "arrayidx28.i", label_forbody_i13); + LoadInst* float_tmp29_i_178 = new LoadInst(ptr_arrayidx28_i_177, "tmp29.i", false, label_forbody_i13); + InsertElementInst* packed_tmp31_i_179 = new InsertElementInst(packed_tmp23_i9, float_tmp29_i_178, const_int32_38, "tmp31.i", label_forbody_i13); std::vector ptr_arrayidx34_i10_indices; ptr_arrayidx34_i10_indices.push_back(const_int32_34); ptr_arrayidx34_i10_indices.push_back(int32_i_0_reg2mem_0_i3); Instruction* ptr_arrayidx34_i10 = new GetElementPtrInst(ptr_consts_141, ptr_arrayidx34_i10_indices.begin(), ptr_arrayidx34_i10_indices.end(), "arrayidx34.i10", label_forbody_i13); - StoreInst* void_179 = new StoreInst(packed_tmp31_i_178, ptr_arrayidx34_i10, false, label_forbody_i13); - BinaryOperator* int32_indvar_next22 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i3, const_int32_36, "indvar.next22", label_forbody_i13); - ICmpInst* int1_exitcond23 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next22, int32_num_consts_133, "exitcond23", label_forbody_i13); - new BranchInst(label_from_consts_exit_137, label_forbody_i13, int1_exitcond23, label_forbody_i13); + StoreInst* void_180 = new StoreInst(packed_tmp31_i_179, ptr_arrayidx34_i10, false, label_forbody_i13); + BinaryOperator* int32_indvar_next23 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i3, const_int32_36, "indvar.next23", label_forbody_i13); + ICmpInst* int1_exitcond24 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next23, int32_num_consts_133, "exitcond24", label_forbody_i13); + new BranchInst(label_from_consts_exit_137, label_forbody_i13, int1_exitcond24, label_forbody_i13); // Block from_consts.exit (label_from_consts_exit_137) - std::vector ptr_tmp5_indices; - ptr_tmp5_indices.push_back(const_int32_34); - ptr_tmp5_indices.push_back(const_int32_38); - Instruction* ptr_tmp5 = new GetElementPtrInst(ptr_args_144, ptr_tmp5_indices.begin(), ptr_tmp5_indices.end(), "tmp5", label_from_consts_exit_137); - std::vector ptr_arraydecay6_indices; - ptr_arraydecay6_indices.push_back(const_int32_34); - ptr_arraydecay6_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay6 = new GetElementPtrInst(ptr_consts_141, ptr_arraydecay6_indices.begin(), ptr_arraydecay6_indices.end(), "arraydecay6", label_from_consts_exit_137); - StoreInst* void_181 = new StoreInst(ptr_arraydecay6, ptr_tmp5, false, label_from_consts_exit_137); - std::vector ptr_tmp7_indices; - ptr_tmp7_indices.push_back(const_int32_34); - ptr_tmp7_indices.push_back(const_int32_37); - Instruction* ptr_tmp7 = new GetElementPtrInst(ptr_args_144, ptr_tmp7_indices.begin(), ptr_tmp7_indices.end(), "tmp7", label_from_consts_exit_137); - std::vector ptr_arraydecay8_indices; - ptr_arraydecay8_indices.push_back(const_int32_34); - ptr_arraydecay8_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay8 = new GetElementPtrInst(ptr_temps_143, ptr_arraydecay8_indices.begin(), ptr_arraydecay8_indices.end(), "arraydecay8", label_from_consts_exit_137); - StoreInst* void_182 = new StoreInst(ptr_arraydecay8, ptr_tmp7, false, label_from_consts_exit_137); - std::vector ptr_tmp10_indices; - ptr_tmp10_indices.push_back(const_int32_34); - ptr_tmp10_indices.push_back(const_int32_36); - Instruction* ptr_tmp10 = new GetElementPtrInst(ptr_args_144, ptr_tmp10_indices.begin(), ptr_tmp10_indices.end(), "tmp10", label_from_consts_exit_137); - std::vector ptr_tmp14_indices; - ptr_tmp14_indices.push_back(const_int32_34); - ptr_tmp14_indices.push_back(const_int32_34); - Instruction* ptr_tmp14 = new GetElementPtrInst(ptr_args_144, ptr_tmp14_indices.begin(), ptr_tmp14_indices.end(), "tmp14", label_from_consts_exit_137); + std::vector ptr_tmp6_182_indices; + ptr_tmp6_182_indices.push_back(const_int32_34); + ptr_tmp6_182_indices.push_back(const_int32_38); + Instruction* ptr_tmp6_182 = new GetElementPtrInst(ptr_args_144, ptr_tmp6_182_indices.begin(), ptr_tmp6_182_indices.end(), "tmp6", label_from_consts_exit_137); + std::vector ptr_arraydecay7_183_indices; + ptr_arraydecay7_183_indices.push_back(const_int32_34); + ptr_arraydecay7_183_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay7_183 = new GetElementPtrInst(ptr_consts_141, ptr_arraydecay7_183_indices.begin(), ptr_arraydecay7_183_indices.end(), "arraydecay7", label_from_consts_exit_137); + StoreInst* void_184 = new StoreInst(ptr_arraydecay7_183, ptr_tmp6_182, false, label_from_consts_exit_137); + std::vector ptr_tmp8_185_indices; + ptr_tmp8_185_indices.push_back(const_int32_34); + ptr_tmp8_185_indices.push_back(const_int32_37); + Instruction* ptr_tmp8_185 = new GetElementPtrInst(ptr_args_144, ptr_tmp8_185_indices.begin(), ptr_tmp8_185_indices.end(), "tmp8", label_from_consts_exit_137); + std::vector ptr_arraydecay9_186_indices; + ptr_arraydecay9_186_indices.push_back(const_int32_34); + ptr_arraydecay9_186_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay9_186 = new GetElementPtrInst(ptr_temps_143, ptr_arraydecay9_186_indices.begin(), ptr_arraydecay9_186_indices.end(), "arraydecay9", label_from_consts_exit_137); + StoreInst* void_187 = new StoreInst(ptr_arraydecay9_186, ptr_tmp8_185, false, label_from_consts_exit_137); + std::vector ptr_tmp11_indices; + ptr_tmp11_indices.push_back(const_int32_34); + ptr_tmp11_indices.push_back(const_int32_36); + Instruction* ptr_tmp11 = new GetElementPtrInst(ptr_args_144, ptr_tmp11_indices.begin(), ptr_tmp11_indices.end(), "tmp11", label_from_consts_exit_137); + std::vector ptr_tmp15_indices; + ptr_tmp15_indices.push_back(const_int32_34); + ptr_tmp15_indices.push_back(const_int32_34); + Instruction* ptr_tmp15 = new GetElementPtrInst(ptr_args_144, ptr_tmp15_indices.begin(), ptr_tmp15_indices.end(), "tmp15", label_from_consts_exit_137); new BranchInst(label_forbody_138, label_from_consts_exit_137); // Block forbody (label_forbody_138) - Argument* fwdref_185 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_184 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_138); - int32_i_0_reg2mem_0_184->reserveOperandSpace(2); - int32_i_0_reg2mem_0_184->addIncoming(const_int32_34, label_from_consts_exit_137); - int32_i_0_reg2mem_0_184->addIncoming(fwdref_185, label_forbody_138); - - std::vector ptr_arraydecay13_indices; - ptr_arraydecay13_indices.push_back(const_int32_34); - ptr_arraydecay13_indices.push_back(int32_i_0_reg2mem_0_184); - ptr_arraydecay13_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay13 = new GetElementPtrInst(ptr_inputs_140, ptr_arraydecay13_indices.begin(), ptr_arraydecay13_indices.end(), "arraydecay13", label_forbody_138); - StoreInst* void_186 = new StoreInst(ptr_arraydecay13, ptr_tmp10, false, label_forbody_138); - std::vector ptr_arraydecay18_indices; - ptr_arraydecay18_indices.push_back(const_int32_34); - ptr_arraydecay18_indices.push_back(int32_i_0_reg2mem_0_184); - ptr_arraydecay18_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay18 = new GetElementPtrInst(ptr_results_142, ptr_arraydecay18_indices.begin(), ptr_arraydecay18_indices.end(), "arraydecay18", label_forbody_138); - StoreInst* void_187 = new StoreInst(ptr_arraydecay18, ptr_tmp14, false, label_forbody_138); - CallInst* void_188 = new CallInst(func_execute_shader, ptr_args_144, "", label_forbody_138); - void_188->setCallingConv(CallingConv::C); - void_188->setTailCall(false); - LoadInst* ptr_tmp24 = new LoadInst(ptr_tmp14, "tmp24", false, label_forbody_138); + Argument* fwdref_189 = new Argument(IntegerType::get(32)); + PHINode* int32_tmp21_rle = new PHINode(IntegerType::get(32), "tmp21.rle", label_forbody_138); + int32_tmp21_rle->reserveOperandSpace(2); + int32_tmp21_rle->addIncoming(const_int32_34, label_from_consts_exit_137); + int32_tmp21_rle->addIncoming(fwdref_189, label_forbody_138); + + Argument* fwdref_191 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_190 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_138); + int32_i_0_reg2mem_0_190->reserveOperandSpace(2); + int32_i_0_reg2mem_0_190->addIncoming(const_int32_34, label_from_consts_exit_137); + int32_i_0_reg2mem_0_190->addIncoming(fwdref_191, label_forbody_138); + + std::vector ptr_arraydecay14_indices; + ptr_arraydecay14_indices.push_back(const_int32_34); + ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_190); + ptr_arraydecay14_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_inputs_140, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_138); + StoreInst* void_192 = new StoreInst(ptr_arraydecay14, ptr_tmp11, false, label_forbody_138); + std::vector ptr_arraydecay19_indices; + ptr_arraydecay19_indices.push_back(const_int32_34); + ptr_arraydecay19_indices.push_back(int32_i_0_reg2mem_0_190); + ptr_arraydecay19_indices.push_back(const_int32_34); + Instruction* ptr_arraydecay19 = new GetElementPtrInst(ptr_results_142, ptr_arraydecay19_indices.begin(), ptr_arraydecay19_indices.end(), "arraydecay19", label_forbody_138); + StoreInst* void_193 = new StoreInst(ptr_arraydecay19, ptr_tmp15, false, label_forbody_138); + StoreInst* void_194 = new StoreInst(const_int32_34, ptr_tmp, false, label_forbody_138); + CallInst* void_195 = new CallInst(func_execute_shader, ptr_args_144, "", label_forbody_138); + void_195->setCallingConv(CallingConv::C); + void_195->setTailCall(false); + LoadInst* int32_tmp26 = new LoadInst(ptr_tmp, "tmp26", false, label_forbody_138); + BinaryOperator* int32_shl = BinaryOperator::create(Instruction::Shl, int32_tmp26, int32_i_0_reg2mem_0_190, "shl", label_forbody_138); + BinaryOperator* int32_or = BinaryOperator::create(Instruction::Or, int32_shl, int32_tmp21_rle, "or", label_forbody_138); + StoreInst* void_196 = new StoreInst(int32_or, ptr_tmp, false, label_forbody_138); + LoadInst* ptr_tmp33 = new LoadInst(ptr_tmp15, "tmp33", false, label_forbody_138); std::vector ptr_arraydecay_i_indices; - ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_190); ptr_arraydecay_i_indices.push_back(const_int32_34); ptr_arraydecay_i_indices.push_back(const_int32_34); Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_dests_129, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_138); - LoadInst* packed_tmp7_i = new LoadInst(ptr_tmp24, "tmp7.i", false, label_forbody_138); + LoadInst* packed_tmp7_i = new LoadInst(ptr_tmp33, "tmp7.i", false, label_forbody_138); ExtractElementInst* float_tmp11_i = new ExtractElementInst(packed_tmp7_i, const_int32_34, "tmp11.i", label_forbody_138); - StoreInst* void_189 = new StoreInst(float_tmp11_i, ptr_arraydecay_i, false, label_forbody_138); + StoreInst* void_197 = new StoreInst(float_tmp11_i, ptr_arraydecay_i, false, label_forbody_138); std::vector ptr_arrayidx13_i_indices; - ptr_arrayidx13_i_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx13_i_indices.push_back(int32_i_0_reg2mem_0_190); ptr_arrayidx13_i_indices.push_back(const_int32_34); ptr_arrayidx13_i_indices.push_back(const_int32_36); Instruction* ptr_arrayidx13_i = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx13_i_indices.begin(), ptr_arrayidx13_i_indices.end(), "arrayidx13.i", label_forbody_138); ExtractElementInst* float_tmp15_i2 = new ExtractElementInst(packed_tmp7_i, const_int32_36, "tmp15.i2", label_forbody_138); - StoreInst* void_190 = new StoreInst(float_tmp15_i2, ptr_arrayidx13_i, false, label_forbody_138); + StoreInst* void_198 = new StoreInst(float_tmp15_i2, ptr_arrayidx13_i, false, label_forbody_138); std::vector ptr_arrayidx17_i_indices; - ptr_arrayidx17_i_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx17_i_indices.push_back(int32_i_0_reg2mem_0_190); ptr_arrayidx17_i_indices.push_back(const_int32_34); ptr_arrayidx17_i_indices.push_back(const_int32_37); Instruction* ptr_arrayidx17_i = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx17_i_indices.begin(), ptr_arrayidx17_i_indices.end(), "arrayidx17.i", label_forbody_138); ExtractElementInst* float_tmp19_i = new ExtractElementInst(packed_tmp7_i, const_int32_37, "tmp19.i", label_forbody_138); - StoreInst* void_191 = new StoreInst(float_tmp19_i, ptr_arrayidx17_i, false, label_forbody_138); + StoreInst* void_199 = new StoreInst(float_tmp19_i, ptr_arrayidx17_i, false, label_forbody_138); std::vector ptr_arrayidx21_i_indices; - ptr_arrayidx21_i_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx21_i_indices.push_back(int32_i_0_reg2mem_0_190); ptr_arrayidx21_i_indices.push_back(const_int32_34); ptr_arrayidx21_i_indices.push_back(const_int32_38); Instruction* ptr_arrayidx21_i = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx21_i_indices.begin(), ptr_arrayidx21_i_indices.end(), "arrayidx21.i", label_forbody_138); ExtractElementInst* float_tmp23_i = new ExtractElementInst(packed_tmp7_i, const_int32_38, "tmp23.i", label_forbody_138); - StoreInst* void_192 = new StoreInst(float_tmp23_i, ptr_arrayidx21_i, false, label_forbody_138); + StoreInst* void_200 = new StoreInst(float_tmp23_i, ptr_arrayidx21_i, false, label_forbody_138); std::vector ptr_arraydecay_i_1_indices; - ptr_arraydecay_i_1_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arraydecay_i_1_indices.push_back(int32_i_0_reg2mem_0_190); ptr_arraydecay_i_1_indices.push_back(const_int32_36); ptr_arraydecay_i_1_indices.push_back(const_int32_34); Instruction* ptr_arraydecay_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arraydecay_i_1_indices.begin(), ptr_arraydecay_i_1_indices.end(), "arraydecay.i.1", label_forbody_138); - GetElementPtrInst* ptr_arrayidx6_i_1 = new GetElementPtrInst(ptr_tmp24, const_int32_36, "arrayidx6.i.1", label_forbody_138); + GetElementPtrInst* ptr_arrayidx6_i_1 = new GetElementPtrInst(ptr_tmp33, const_int32_36, "arrayidx6.i.1", label_forbody_138); LoadInst* packed_tmp7_i_1 = new LoadInst(ptr_arrayidx6_i_1, "tmp7.i.1", false, label_forbody_138); ExtractElementInst* float_tmp11_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_34, "tmp11.i.1", label_forbody_138); - StoreInst* void_193 = new StoreInst(float_tmp11_i_1, ptr_arraydecay_i_1, false, label_forbody_138); + StoreInst* void_201 = new StoreInst(float_tmp11_i_1, ptr_arraydecay_i_1, false, label_forbody_138); std::vector ptr_arrayidx13_i_1_indices; - ptr_arrayidx13_i_1_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx13_i_1_indices.push_back(int32_i_0_reg2mem_0_190); ptr_arrayidx13_i_1_indices.push_back(const_int32_36); ptr_arrayidx13_i_1_indices.push_back(const_int32_36); Instruction* ptr_arrayidx13_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx13_i_1_indices.begin(), ptr_arrayidx13_i_1_indices.end(), "arrayidx13.i.1", label_forbody_138); ExtractElementInst* float_tmp15_i2_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_36, "tmp15.i2.1", label_forbody_138); - StoreInst* void_194 = new StoreInst(float_tmp15_i2_1, ptr_arrayidx13_i_1, false, label_forbody_138); + StoreInst* void_202 = new StoreInst(float_tmp15_i2_1, ptr_arrayidx13_i_1, false, label_forbody_138); std::vector ptr_arrayidx17_i_1_indices; - ptr_arrayidx17_i_1_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx17_i_1_indices.push_back(int32_i_0_reg2mem_0_190); ptr_arrayidx17_i_1_indices.push_back(const_int32_36); ptr_arrayidx17_i_1_indices.push_back(const_int32_37); Instruction* ptr_arrayidx17_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx17_i_1_indices.begin(), ptr_arrayidx17_i_1_indices.end(), "arrayidx17.i.1", label_forbody_138); ExtractElementInst* float_tmp19_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_37, "tmp19.i.1", label_forbody_138); - StoreInst* void_195 = new StoreInst(float_tmp19_i_1, ptr_arrayidx17_i_1, false, label_forbody_138); + StoreInst* void_203 = new StoreInst(float_tmp19_i_1, ptr_arrayidx17_i_1, false, label_forbody_138); std::vector ptr_arrayidx21_i_1_indices; - ptr_arrayidx21_i_1_indices.push_back(int32_i_0_reg2mem_0_184); + ptr_arrayidx21_i_1_indices.push_back(int32_i_0_reg2mem_0_190); ptr_arrayidx21_i_1_indices.push_back(const_int32_36); ptr_arrayidx21_i_1_indices.push_back(const_int32_38); Instruction* ptr_arrayidx21_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx21_i_1_indices.begin(), ptr_arrayidx21_i_1_indices.end(), "arrayidx21.i.1", label_forbody_138); ExtractElementInst* float_tmp23_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_38, "tmp23.i.1", label_forbody_138); - StoreInst* void_196 = new StoreInst(float_tmp23_i_1, ptr_arrayidx21_i_1, false, label_forbody_138); - BinaryOperator* int32_indvar_next20 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_184, const_int32_36, "indvar.next20", label_forbody_138); - ICmpInst* int1_exitcond21 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next20, const_int32_39, "exitcond21", label_forbody_138); - new BranchInst(label_afterfor_139, label_forbody_138, int1_exitcond21, label_forbody_138); + StoreInst* void_204 = new StoreInst(float_tmp23_i_1, ptr_arrayidx21_i_1, false, label_forbody_138); + BinaryOperator* int32_indvar_next21 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_190, const_int32_36, "indvar.next21", label_forbody_138); + ICmpInst* int1_exitcond22 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next21, const_int32_39, "exitcond22", label_forbody_138); + new BranchInst(label_afterfor_139, label_forbody_138, int1_exitcond22, label_forbody_138); // Block afterfor (label_afterfor_139) - std::vector ptr_tmp26_indices; - ptr_tmp26_indices.push_back(const_int32_34); - ptr_tmp26_indices.push_back(const_int32_39); - Instruction* ptr_tmp26 = new GetElementPtrInst(ptr_args_144, ptr_tmp26_indices.begin(), ptr_tmp26_indices.end(), "tmp26", label_afterfor_139); - LoadInst* int32_tmp27 = new LoadInst(ptr_tmp26, "tmp27", false, label_afterfor_139); - BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_tmp27, const_int32_40, "neg", label_afterfor_139); + BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_or, const_int32_40, "neg", label_afterfor_139); new ReturnInst(int32_neg, label_afterfor_139); // Block forbody6.i.1 (label_forbody6_i_1) - Argument* fwdref_199 = new Argument(IntegerType::get(32)); + Argument* fwdref_207 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i_1 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.1", label_forbody6_i_1); int32_j_0_reg2mem_0_i_1->reserveOperandSpace(2); int32_j_0_reg2mem_0_i_1->addIncoming(const_int32_34, label_forbody6_i_135); - int32_j_0_reg2mem_0_i_1->addIncoming(fwdref_199, label_forbody6_i_1); + int32_j_0_reg2mem_0_i_1->addIncoming(fwdref_207, label_forbody6_i_1); - Argument* fwdref_200 = new Argument(VectorTy_1); + Argument* fwdref_208 = new Argument(VectorTy_1); PHINode* packed_vec_0_reg2mem_0_i_1 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i.1", label_forbody6_i_1); packed_vec_0_reg2mem_0_i_1->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_1->addIncoming(packed_tmp48_i_162, label_forbody6_i_135); - packed_vec_0_reg2mem_0_i_1->addIncoming(fwdref_200, label_forbody6_i_1); + packed_vec_0_reg2mem_0_i_1->addIncoming(packed_tmp48_i_163, label_forbody6_i_135); + packed_vec_0_reg2mem_0_i_1->addIncoming(fwdref_208, label_forbody6_i_1); std::vector ptr_arraydecay11_i_1_indices; ptr_arraydecay11_i_1_indices.push_back(const_int32_36); @@ -1084,23 +1095,23 @@ Module* createBaseShader() { ptr_arrayidx54_i_1_indices.push_back(const_int32_36); ptr_arrayidx54_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); Instruction* ptr_arrayidx54_i_1 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_1_indices.begin(), ptr_arrayidx54_i_1_indices.end(), "arrayidx54.i.1", label_forbody6_i_1); - StoreInst* void_201 = new StoreInst(packed_tmp48_i_1, ptr_arrayidx54_i_1, false, label_forbody6_i_1); + StoreInst* void_209 = new StoreInst(packed_tmp48_i_1, ptr_arrayidx54_i_1, false, label_forbody6_i_1); BinaryOperator* int32_inc_i_1 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_1, const_int32_36, "inc.i.1", label_forbody6_i_1); ICmpInst* int1_cmp59_i_1 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_1, int32_num_inputs_131, "cmp59.i.1", label_forbody6_i_1); new BranchInst(label_forbody6_i_1, label_forbody6_i_2, int1_cmp59_i_1, label_forbody6_i_1); // Block forbody6.i.2 (label_forbody6_i_2) - Argument* fwdref_203 = new Argument(IntegerType::get(32)); + Argument* fwdref_211 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i_2 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.2", label_forbody6_i_2); int32_j_0_reg2mem_0_i_2->reserveOperandSpace(2); int32_j_0_reg2mem_0_i_2->addIncoming(const_int32_34, label_forbody6_i_1); - int32_j_0_reg2mem_0_i_2->addIncoming(fwdref_203, label_forbody6_i_2); + int32_j_0_reg2mem_0_i_2->addIncoming(fwdref_211, label_forbody6_i_2); - Argument* fwdref_204 = new Argument(VectorTy_1); + Argument* fwdref_212 = new Argument(VectorTy_1); PHINode* packed_vec_0_reg2mem_0_i_2 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i.2", label_forbody6_i_2); packed_vec_0_reg2mem_0_i_2->reserveOperandSpace(2); packed_vec_0_reg2mem_0_i_2->addIncoming(packed_tmp48_i_1, label_forbody6_i_1); - packed_vec_0_reg2mem_0_i_2->addIncoming(fwdref_204, label_forbody6_i_2); + packed_vec_0_reg2mem_0_i_2->addIncoming(fwdref_212, label_forbody6_i_2); std::vector ptr_arraydecay11_i_2_indices; ptr_arraydecay11_i_2_indices.push_back(const_int32_37); @@ -1135,23 +1146,23 @@ Module* createBaseShader() { ptr_arrayidx54_i_2_indices.push_back(const_int32_37); ptr_arrayidx54_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); Instruction* ptr_arrayidx54_i_2 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_2_indices.begin(), ptr_arrayidx54_i_2_indices.end(), "arrayidx54.i.2", label_forbody6_i_2); - StoreInst* void_205 = new StoreInst(packed_tmp48_i_2, ptr_arrayidx54_i_2, false, label_forbody6_i_2); + StoreInst* void_213 = new StoreInst(packed_tmp48_i_2, ptr_arrayidx54_i_2, false, label_forbody6_i_2); BinaryOperator* int32_inc_i_2 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_2, const_int32_36, "inc.i.2", label_forbody6_i_2); ICmpInst* int1_cmp59_i_2 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_2, int32_num_inputs_131, "cmp59.i.2", label_forbody6_i_2); new BranchInst(label_forbody6_i_2, label_forbody6_i_3, int1_cmp59_i_2, label_forbody6_i_2); // Block forbody6.i.3 (label_forbody6_i_3) - Argument* fwdref_207 = new Argument(IntegerType::get(32)); + Argument* fwdref_215 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0_i_3 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.3", label_forbody6_i_3); int32_j_0_reg2mem_0_i_3->reserveOperandSpace(2); int32_j_0_reg2mem_0_i_3->addIncoming(const_int32_34, label_forbody6_i_2); - int32_j_0_reg2mem_0_i_3->addIncoming(fwdref_207, label_forbody6_i_3); + int32_j_0_reg2mem_0_i_3->addIncoming(fwdref_215, label_forbody6_i_3); - Argument* fwdref_208 = new Argument(VectorTy_1); + Argument* fwdref_216 = new Argument(VectorTy_1); PHINode* packed_vec_0_reg2mem_0_i_3 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i.3", label_forbody6_i_3); packed_vec_0_reg2mem_0_i_3->reserveOperandSpace(2); packed_vec_0_reg2mem_0_i_3->addIncoming(packed_tmp48_i_2, label_forbody6_i_2); - packed_vec_0_reg2mem_0_i_3->addIncoming(fwdref_208, label_forbody6_i_3); + packed_vec_0_reg2mem_0_i_3->addIncoming(fwdref_216, label_forbody6_i_3); std::vector ptr_arraydecay11_i_3_indices; ptr_arraydecay11_i_3_indices.push_back(const_int32_38); @@ -1186,23 +1197,24 @@ Module* createBaseShader() { ptr_arrayidx54_i_3_indices.push_back(const_int32_38); ptr_arrayidx54_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); Instruction* ptr_arrayidx54_i_3 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_3_indices.begin(), ptr_arrayidx54_i_3_indices.end(), "arrayidx54.i.3", label_forbody6_i_3); - StoreInst* void_209 = new StoreInst(packed_tmp48_i_3, ptr_arrayidx54_i_3, false, label_forbody6_i_3); + StoreInst* void_217 = new StoreInst(packed_tmp48_i_3, ptr_arrayidx54_i_3, false, label_forbody6_i_3); BinaryOperator* int32_inc_i_3 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_3, const_int32_36, "inc.i.3", label_forbody6_i_3); ICmpInst* int1_cmp59_i_3 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_3, int32_num_inputs_131, "cmp59.i.3", label_forbody6_i_3); new BranchInst(label_forbody6_i_3, label_from_array_exit_136, int1_cmp59_i_3, label_forbody6_i_3); // Resolve Forward References - fwdref_208->replaceAllUsesWith(packed_tmp48_i_3); delete fwdref_208; - fwdref_207->replaceAllUsesWith(int32_inc_i_3); delete fwdref_207; - fwdref_150->replaceAllUsesWith(packed_tmp48_i_162); delete fwdref_150; - fwdref_148->replaceAllUsesWith(int32_inc_i_165); delete fwdref_148; - fwdref_171->replaceAllUsesWith(packed_tmp31_i_178); delete fwdref_171; - fwdref_170->replaceAllUsesWith(int32_indvar_next22); delete fwdref_170; - fwdref_185->replaceAllUsesWith(int32_indvar_next20); delete fwdref_185; - fwdref_200->replaceAllUsesWith(packed_tmp48_i_1); delete fwdref_200; - fwdref_199->replaceAllUsesWith(int32_inc_i_1); delete fwdref_199; - fwdref_204->replaceAllUsesWith(packed_tmp48_i_2); delete fwdref_204; - fwdref_203->replaceAllUsesWith(int32_inc_i_2); delete fwdref_203; + fwdref_216->replaceAllUsesWith(packed_tmp48_i_3); delete fwdref_216; + fwdref_215->replaceAllUsesWith(int32_inc_i_3); delete fwdref_215; + fwdref_151->replaceAllUsesWith(packed_tmp48_i_163); delete fwdref_151; + fwdref_149->replaceAllUsesWith(int32_inc_i_166); delete fwdref_149; + fwdref_172->replaceAllUsesWith(packed_tmp31_i_179); delete fwdref_172; + fwdref_171->replaceAllUsesWith(int32_indvar_next23); delete fwdref_171; + fwdref_189->replaceAllUsesWith(int32_or); delete fwdref_189; + fwdref_191->replaceAllUsesWith(int32_indvar_next21); delete fwdref_191; + fwdref_208->replaceAllUsesWith(packed_tmp48_i_1); delete fwdref_208; + fwdref_207->replaceAllUsesWith(int32_inc_i_1); delete fwdref_207; + fwdref_212->replaceAllUsesWith(packed_tmp48_i_2); delete fwdref_212; + fwdref_211->replaceAllUsesWith(int32_inc_i_2); delete fwdref_211; } diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c index 517aa2e84b..4f98d754ba 100644 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ b/src/mesa/pipe/llvm/llvm_builtins.c @@ -32,7 +32,6 @@ */ typedef __attribute__(( ocu_vector_type(4) )) float float4; - extern float powf(float a, float b); inline float approx(float a, float b) @@ -106,3 +105,11 @@ inline float4 vsin(float4 val) result.w = res; return result; } + +inline int kilp(float4 val) +{ + if (val.x < 0 || val.y < 0 || val.z < 0 || val.w < 0) + return 1; + else + return 0; +} diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index 6bdb311c2b..03f7ac8f14 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -239,6 +239,8 @@ int run_fragment_shader(float x, float y, float4 results[4][16]; float4 temps[128];//MAX_PROGRAM_TEMPS struct ShaderInput args; + int mask = 0; + args.kilmask = 0; from_array(inputs, ainputs, 4, num_inputs); from_consts(consts, aconsts, num_consts); @@ -248,7 +250,11 @@ int run_fragment_shader(float x, float y, for (int i = 0; i < 4; ++i) { args.inputs = inputs[i]; args.dests = results[i]; + mask = args.kilmask; + args.kilmask = 0; execute_shader(&args); + args.kilmask = mask | (args.kilmask << i); + to_array(dests[i], args.dests, 2); } return ~args.kilmask; diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 71045fa37f..c33b9bf45d 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -343,6 +343,20 @@ llvm::Value * Storage::element(Args arg, int idx, llvm::Value *indIdx ) return elemIdx(val, idx, indIdx); } +void Storage::setKilElement(llvm::Value *val) +{ + std::vector indices; + indices.push_back(constantInt(0)); + indices.push_back(constantInt(static_cast(KilArg))); + GetElementPtrInst *elem = new GetElementPtrInst(m_INPUT, + indices.begin(), + indices.end(), + name("kil_ptr"), + m_block); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); +} + #endif //MESA_LLVM diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h index 7f1a8bf103..8574f7554e 100644 --- a/src/mesa/pipe/llvm/storage.h +++ b/src/mesa/pipe/llvm/storage.h @@ -71,6 +71,8 @@ public: llvm::Value *addrElement(int idx) const; void setAddrElement(int idx, llvm::Value *val, int mask); + void setKilElement(llvm::Value *val); + llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); llvm::Value *extractIndex(llvm::Value *vec); diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 9307ed233d..5ea07f95a4 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -202,7 +202,7 @@ shade_quad_llvm(struct quad_stage *qs, } #endif - /*quad->mask &=*/ + quad->mask &= gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs, softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], qss->samplers); -- cgit v1.2.3 From 9d6e6e86d9592afbdbaa83fc89891e6813fc3f2f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 7 Nov 2007 08:59:35 -0500 Subject: number of outputs is nr_attribs - 2 --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 29de437595..10fd33fbde 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -133,7 +133,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) gallivm_prog_exec(prog, inputs, outputs, consts, draw->vs.queue_nr, draw->vertex_shader->state->num_inputs, - draw->vertex_info.num_attribs); + draw->vertex_info.num_attribs - 2); /* store machine results */ for (int i = 0; i < draw->vs.queue_nr; ++i) { -- cgit v1.2.3 From 6dc4e6ae15676cf4acdebb9c798bfa4083ad1e14 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 7 Nov 2007 13:26:45 -0500 Subject: Redo the way we pass arguments to the llvm. simply pass aligned arrays, they should cast to vectors without any problems. also remove unnecessary memset --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 7 +- src/mesa/pipe/llvm/llvm_base_shader.cpp | 1320 +++++++++----------------- src/mesa/pipe/llvm/llvm_entry.c | 20 +- src/mesa/pipe/p_compiler.h | 2 + src/mesa/pipe/softpipe/sp_quad_fs.c | 5 +- 5 files changed, 458 insertions(+), 896 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 10fd33fbde..34a4d27684 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -115,13 +115,12 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) unsigned i; struct vertex_header *dests[VS_QUEUE_LENGTH]; - float inputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; - float outputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4]; + float inputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4] ALIGN16_ATTRIB; + float outputs[VS_QUEUE_LENGTH][PIPE_MAX_SHADER_INPUTS][4] ALIGN16_ATTRIB; float (*consts)[4] = (float (*)[4]) draw->user.constants; struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; - /* fetch the inputs */ for (i = 0; i < draw->vs.queue_nr; ++i) { unsigned elt = draw->vs.queue[i].elt; @@ -135,6 +134,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) draw->vertex_shader->state->num_inputs, draw->vertex_info.num_attribs - 2); + /* store machine results */ for (int i = 0; i < draw->vs.queue_nr; ++i) { unsigned slot; @@ -158,7 +158,6 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) vOut->clipmask = compute_clipmask(vOut->clip, draw->plane, draw->nr_planes); vOut->edgeflag = 1; - /* divide by w */ w = 1.0f / w; x *= w; diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index 82ad6cfa26..85fa389d79 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -118,8 +118,8 @@ Module* createBaseShader() { /*ParamAttrs=*/FuncTy_18_PAL); std::vectorFuncTy_19_args; - FuncTy_19_args.push_back(PointerTy_15); - FuncTy_19_args.push_back(PointerTy_15); + FuncTy_19_args.push_back(PointerTy_13); + FuncTy_19_args.push_back(PointerTy_13); FuncTy_19_args.push_back(PointerTy_8); FuncTy_19_args.push_back(IntegerType::get(32)); FuncTy_19_args.push_back(IntegerType::get(32)); @@ -132,55 +132,47 @@ Module* createBaseShader() { /*isVarArg=*/false, /*ParamAttrs=*/FuncTy_19_PAL); - ArrayType* ArrayTy_21 = ArrayType::get(ArrayTy_14, 2048); + ArrayType* ArrayTy_21 = ArrayType::get(VectorTy_1, 32); PointerType* PointerTy_20 = PointerType::get(ArrayTy_21); - ArrayType* ArrayTy_23 = ArrayType::get(VectorTy_1, 32); + ArrayType* ArrayTy_23 = ArrayType::get(VectorTy_1, 128); PointerType* PointerTy_22 = PointerType::get(ArrayTy_23); - ArrayType* ArrayTy_25 = ArrayType::get(VectorTy_1, 128); + PointerType* PointerTy_24 = PointerType::get(StructTy_struct_ShaderInput); - PointerType* PointerTy_24 = PointerType::get(ArrayTy_25); + PointerType* PointerTy_25 = PointerType::get(PointerTy_0); - PointerType* PointerTy_26 = PointerType::get(StructTy_struct_ShaderInput); - - PointerType* PointerTy_27 = PointerType::get(PointerTy_0); - - std::vectorFuncTy_29_args; - FuncTy_29_args.push_back(PointerTy_26); - ParamAttrsList *FuncTy_29_PAL = 0; - FunctionType* FuncTy_29 = FunctionType::get( + std::vectorFuncTy_27_args; + FuncTy_27_args.push_back(PointerTy_24); + ParamAttrsList *FuncTy_27_PAL = 0; + FunctionType* FuncTy_27 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_29_args, + /*Params=*/FuncTy_27_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_29_PAL); - - PointerType* PointerTy_28 = PointerType::get(FuncTy_29); - - std::vectorFuncTy_30_args; - FuncTy_30_args.push_back(Type::FloatTy); - FuncTy_30_args.push_back(Type::FloatTy); - FuncTy_30_args.push_back(PointerTy_15); - FuncTy_30_args.push_back(PointerTy_15); - FuncTy_30_args.push_back(IntegerType::get(32)); - FuncTy_30_args.push_back(PointerTy_8); - FuncTy_30_args.push_back(IntegerType::get(32)); - FuncTy_30_args.push_back(PointerTy_6); - PointerType* PointerTy_31 = PointerType::get(IntegerType::get(32)); - - FuncTy_30_args.push_back(PointerTy_31); - ParamAttrsList *FuncTy_30_PAL = 0; - FunctionType* FuncTy_30 = FunctionType::get( + /*ParamAttrs=*/FuncTy_27_PAL); + + PointerType* PointerTy_26 = PointerType::get(FuncTy_27); + + std::vectorFuncTy_28_args; + FuncTy_28_args.push_back(Type::FloatTy); + FuncTy_28_args.push_back(Type::FloatTy); + FuncTy_28_args.push_back(PointerTy_13); + FuncTy_28_args.push_back(PointerTy_13); + FuncTy_28_args.push_back(IntegerType::get(32)); + FuncTy_28_args.push_back(PointerTy_8); + FuncTy_28_args.push_back(IntegerType::get(32)); + FuncTy_28_args.push_back(PointerTy_6); + PointerType* PointerTy_29 = PointerType::get(IntegerType::get(32)); + + FuncTy_28_args.push_back(PointerTy_29); + ParamAttrsList *FuncTy_28_PAL = 0; + FunctionType* FuncTy_28 = FunctionType::get( /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_30_args, + /*Params=*/FuncTy_28_args, /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_30_PAL); - - ArrayType* ArrayTy_33 = ArrayType::get(ArrayTy_14, 4); - - PointerType* PointerTy_32 = PointerType::get(ArrayTy_33); + /*ParamAttrs=*/FuncTy_28_PAL); // Function Declarations @@ -210,13 +202,13 @@ Module* createBaseShader() { func_run_vertex_shader->setCallingConv(CallingConv::C); Function* func_execute_shader = new Function( - /*Type=*/FuncTy_29, + /*Type=*/FuncTy_27, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"execute_shader", mod); // (external, no body) func_execute_shader->setCallingConv(CallingConv::C); Function* func_run_fragment_shader = new Function( - /*Type=*/FuncTy_30, + /*Type=*/FuncTy_28, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_fragment_shader", mod); func_run_fragment_shader->setCallingConv(CallingConv::C); @@ -225,13 +217,13 @@ Module* createBaseShader() { // Constant Definitions - Constant* const_int32_34 = Constant::getNullValue(IntegerType::get(32)); - UndefValue* const_packed_35 = UndefValue::get(VectorTy_1); - ConstantInt* const_int32_36 = ConstantInt::get(APInt(32, "1", 10)); - ConstantInt* const_int32_37 = ConstantInt::get(APInt(32, "2", 10)); - ConstantInt* const_int32_38 = ConstantInt::get(APInt(32, "3", 10)); - ConstantInt* const_int32_39 = ConstantInt::get(APInt(32, "4", 10)); - ConstantInt* const_int32_40 = ConstantInt::get(APInt(32, "-1", 10)); + Constant* const_int32_30 = Constant::getNullValue(IntegerType::get(32)); + UndefValue* const_packed_31 = UndefValue::get(VectorTy_1); + ConstantInt* const_int32_32 = ConstantInt::get(APInt(32, "1", 10)); + ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "2", 10)); + ConstantInt* const_int32_34 = ConstantInt::get(APInt(32, "3", 10)); + ConstantInt* const_int32_35 = ConstantInt::get(APInt(32, "4", 10)); + ConstantInt* const_int32_36 = ConstantInt::get(APInt(32, "-1", 10)); // Global Variable Definitions @@ -255,73 +247,73 @@ Module* createBaseShader() { BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); // Block entry (label_entry) - ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_34, "cmp", label_entry); - ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_34, "cmp5", label_entry); + ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_30, "cmp", label_entry); + ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_30, "cmp5", label_entry); BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); new BranchInst(label_forbody6, label_afterfor60, int1_bothcond, label_entry); // Block forbody6 (label_forbody6) - Argument* fwdref_42 = new Argument(IntegerType::get(32)); - Argument* fwdref_43 = new Argument(IntegerType::get(32)); + Argument* fwdref_38 = new Argument(IntegerType::get(32)); + Argument* fwdref_39 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody6); int32_i_0_reg2mem_0->reserveOperandSpace(3); - int32_i_0_reg2mem_0->addIncoming(const_int32_34, label_entry); - int32_i_0_reg2mem_0->addIncoming(fwdref_42, label_forinc57); - int32_i_0_reg2mem_0->addIncoming(fwdref_43, label_forbody6); + int32_i_0_reg2mem_0->addIncoming(const_int32_30, label_entry); + int32_i_0_reg2mem_0->addIncoming(fwdref_38, label_forinc57); + int32_i_0_reg2mem_0->addIncoming(fwdref_39, label_forbody6); - Argument* fwdref_44 = new Argument(IntegerType::get(32)); + Argument* fwdref_40 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0", label_forbody6); int32_j_0_reg2mem_0->reserveOperandSpace(3); - int32_j_0_reg2mem_0->addIncoming(fwdref_44, label_forbody6); - int32_j_0_reg2mem_0->addIncoming(const_int32_34, label_forinc57); - int32_j_0_reg2mem_0->addIncoming(const_int32_34, label_entry); + int32_j_0_reg2mem_0->addIncoming(fwdref_40, label_forbody6); + int32_j_0_reg2mem_0->addIncoming(const_int32_30, label_forinc57); + int32_j_0_reg2mem_0->addIncoming(const_int32_30, label_entry); - Argument* fwdref_45 = new Argument(VectorTy_1); + Argument* fwdref_41 = new Argument(VectorTy_1); PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody6); packed_vec_0_reg2mem_0->reserveOperandSpace(3); - packed_vec_0_reg2mem_0->addIncoming(fwdref_45, label_forbody6); - packed_vec_0_reg2mem_0->addIncoming(const_packed_35, label_entry); - packed_vec_0_reg2mem_0->addIncoming(fwdref_45, label_forinc57); + packed_vec_0_reg2mem_0->addIncoming(fwdref_41, label_forbody6); + packed_vec_0_reg2mem_0->addIncoming(const_packed_31, label_entry); + packed_vec_0_reg2mem_0->addIncoming(fwdref_41, label_forinc57); std::vector ptr_arraydecay11_indices; ptr_arraydecay11_indices.push_back(int32_i_0_reg2mem_0); ptr_arraydecay11_indices.push_back(int32_j_0_reg2mem_0); - ptr_arraydecay11_indices.push_back(const_int32_34); + ptr_arraydecay11_indices.push_back(const_int32_30); Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); - InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_34, "tmp15", label_forbody6); + InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_30, "tmp15", label_forbody6); std::vector ptr_arrayidx23_indices; ptr_arrayidx23_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx23_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx23_indices.push_back(const_int32_36); + ptr_arrayidx23_indices.push_back(const_int32_32); Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_36, "tmp26", label_forbody6); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_32, "tmp26", label_forbody6); std::vector ptr_arrayidx34_indices; ptr_arrayidx34_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx34_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx34_indices.push_back(const_int32_37); + ptr_arrayidx34_indices.push_back(const_int32_33); Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6); LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6); - InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_37, "tmp37", label_forbody6); + InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_33, "tmp37", label_forbody6); std::vector ptr_arrayidx45_indices; ptr_arrayidx45_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx45_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx45_indices.push_back(const_int32_38); + ptr_arrayidx45_indices.push_back(const_int32_34); Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6); LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6); - InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_38, "tmp48", label_forbody6); + InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_34, "tmp48", label_forbody6); std::vector ptr_arrayidx54_indices; ptr_arrayidx54_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0); Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); - StoreInst* void_46 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); - BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_36, "inc", label_forbody6); + StoreInst* void_42 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); + BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_32, "inc", label_forbody6); ICmpInst* int1_cmp59 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp59", label_forbody6); new BranchInst(label_forbody6, label_forinc57, int1_cmp59, label_forbody6); // Block forinc57 (label_forinc57) - BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_36, "inc59", label_forinc57); + BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_32, "inc59", label_forinc57); ICmpInst* int1_cmp17 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59, int32_count, "cmp17", label_forinc57); new BranchInst(label_forbody6, label_afterfor60, int1_cmp17, label_forinc57); @@ -329,80 +321,80 @@ Module* createBaseShader() { new ReturnInst(label_afterfor60); // Resolve Forward References - fwdref_43->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_43; - fwdref_45->replaceAllUsesWith(packed_tmp48); delete fwdref_45; - fwdref_44->replaceAllUsesWith(int32_inc); delete fwdref_44; - fwdref_42->replaceAllUsesWith(int32_inc59); delete fwdref_42; + fwdref_39->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_39; + fwdref_41->replaceAllUsesWith(packed_tmp48); delete fwdref_41; + fwdref_40->replaceAllUsesWith(int32_inc); delete fwdref_40; + fwdref_38->replaceAllUsesWith(int32_inc59); delete fwdref_38; } // Function: from_consts (func_from_consts) { Function::arg_iterator args = func_from_consts->arg_begin(); - Value* ptr_res_50 = args++; - ptr_res_50->setName("res"); - Value* ptr_ainputs_51 = args++; - ptr_ainputs_51->setName("ainputs"); - Value* int32_count_52 = args++; - int32_count_52->setName("count"); - - BasicBlock* label_entry_53 = new BasicBlock("entry",func_from_consts,0); + Value* ptr_res_46 = args++; + ptr_res_46->setName("res"); + Value* ptr_ainputs_47 = args++; + ptr_ainputs_47->setName("ainputs"); + Value* int32_count_48 = args++; + int32_count_48->setName("count"); + + BasicBlock* label_entry_49 = new BasicBlock("entry",func_from_consts,0); BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); - // Block entry (label_entry_53) - ICmpInst* int1_cmp_54 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_52, const_int32_34, "cmp", label_entry_53); - new BranchInst(label_forbody, label_afterfor, int1_cmp_54, label_entry_53); + // Block entry (label_entry_49) + ICmpInst* int1_cmp_50 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_48, const_int32_30, "cmp", label_entry_49); + new BranchInst(label_forbody, label_afterfor, int1_cmp_50, label_entry_49); // Block forbody (label_forbody) - Argument* fwdref_57 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_56 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); - int32_i_0_reg2mem_0_56->reserveOperandSpace(2); - int32_i_0_reg2mem_0_56->addIncoming(const_int32_34, label_entry_53); - int32_i_0_reg2mem_0_56->addIncoming(fwdref_57, label_forbody); - - Argument* fwdref_59 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_58 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody); - packed_vec_0_reg2mem_0_58->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_58->addIncoming(const_packed_35, label_entry_53); - packed_vec_0_reg2mem_0_58->addIncoming(fwdref_59, label_forbody); + Argument* fwdref_53 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_52 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); + int32_i_0_reg2mem_0_52->reserveOperandSpace(2); + int32_i_0_reg2mem_0_52->addIncoming(const_int32_30, label_entry_49); + int32_i_0_reg2mem_0_52->addIncoming(fwdref_53, label_forbody); + + Argument* fwdref_55 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_54 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody); + packed_vec_0_reg2mem_0_54->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_54->addIncoming(const_packed_31, label_entry_49); + packed_vec_0_reg2mem_0_54->addIncoming(fwdref_55, label_forbody); std::vector ptr_arraydecay_indices; - ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_56); - ptr_arraydecay_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_51, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); + ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_52); + ptr_arraydecay_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_47, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); - InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_58, float_tmp5, const_int32_34, "tmp7", label_forbody); + InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_54, float_tmp5, const_int32_30, "tmp7", label_forbody); std::vector ptr_arrayidx12_indices; - ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_56); - ptr_arrayidx12_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_51, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); - LoadInst* float_tmp13_60 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); - InsertElementInst* packed_tmp15_61 = new InsertElementInst(packed_tmp7, float_tmp13_60, const_int32_36, "tmp15", label_forbody); + ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_52); + ptr_arrayidx12_indices.push_back(const_int32_32); + Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); + LoadInst* float_tmp13_56 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); + InsertElementInst* packed_tmp15_57 = new InsertElementInst(packed_tmp7, float_tmp13_56, const_int32_32, "tmp15", label_forbody); std::vector ptr_arrayidx20_indices; - ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_56); - ptr_arrayidx20_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_51, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); + ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_52); + ptr_arrayidx20_indices.push_back(const_int32_33); + Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); - InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_61, float_tmp21, const_int32_37, "tmp23", label_forbody); + InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_57, float_tmp21, const_int32_33, "tmp23", label_forbody); std::vector ptr_arrayidx28_indices; - ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_56); - ptr_arrayidx28_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_51, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); + ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_52); + ptr_arrayidx28_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); - InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_38, "tmp31", label_forbody); - GetElementPtrInst* ptr_arrayidx34_62 = new GetElementPtrInst(ptr_res_50, int32_i_0_reg2mem_0_56, "arrayidx34", label_forbody); - StoreInst* void_63 = new StoreInst(packed_tmp31, ptr_arrayidx34_62, false, label_forbody); - BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_56, const_int32_36, "indvar.next", label_forbody); - ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_52, "exitcond", label_forbody); + InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_34, "tmp31", label_forbody); + GetElementPtrInst* ptr_arrayidx34_58 = new GetElementPtrInst(ptr_res_46, int32_i_0_reg2mem_0_52, "arrayidx34", label_forbody); + StoreInst* void_59 = new StoreInst(packed_tmp31, ptr_arrayidx34_58, false, label_forbody); + BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_52, const_int32_32, "indvar.next", label_forbody); + ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_48, "exitcond", label_forbody); new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); // Block afterfor (label_afterfor) new ReturnInst(label_afterfor); // Resolve Forward References - fwdref_59->replaceAllUsesWith(packed_tmp31); delete fwdref_59; - fwdref_57->replaceAllUsesWith(int32_indvar_next); delete fwdref_57; + fwdref_55->replaceAllUsesWith(packed_tmp31); delete fwdref_55; + fwdref_53->replaceAllUsesWith(int32_indvar_next); delete fwdref_53; } @@ -413,363 +405,204 @@ Module* createBaseShader() { ptr_dests->setName("dests"); Value* ptr_in = args++; ptr_in->setName("in"); - Value* int32_num_attribs_66 = args++; - int32_num_attribs_66->setName("num_attribs"); - - BasicBlock* label_entry_67 = new BasicBlock("entry",func_to_array,0); - BasicBlock* label_forbody_68 = new BasicBlock("forbody",func_to_array,0); - BasicBlock* label_afterfor_69 = new BasicBlock("afterfor",func_to_array,0); - - // Block entry (label_entry_67) - ICmpInst* int1_cmp_70 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_66, const_int32_34, "cmp", label_entry_67); - new BranchInst(label_forbody_68, label_afterfor_69, int1_cmp_70, label_entry_67); - - // Block forbody (label_forbody_68) - Argument* fwdref_73 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_72 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_68); - int32_i_0_reg2mem_0_72->reserveOperandSpace(2); - int32_i_0_reg2mem_0_72->addIncoming(const_int32_34, label_entry_67); - int32_i_0_reg2mem_0_72->addIncoming(fwdref_73, label_forbody_68); - - std::vector ptr_arraydecay_74_indices; - ptr_arraydecay_74_indices.push_back(int32_i_0_reg2mem_0_72); - ptr_arraydecay_74_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay_74 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_74_indices.begin(), ptr_arraydecay_74_indices.end(), "arraydecay", label_forbody_68); - GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_72, "arrayidx6", label_forbody_68); - LoadInst* packed_tmp7_75 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_68); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_75, const_int32_34, "tmp11", label_forbody_68); - StoreInst* void_76 = new StoreInst(float_tmp11, ptr_arraydecay_74, false, label_forbody_68); + Value* int32_num_attribs_62 = args++; + int32_num_attribs_62->setName("num_attribs"); + + BasicBlock* label_entry_63 = new BasicBlock("entry",func_to_array,0); + BasicBlock* label_forbody_64 = new BasicBlock("forbody",func_to_array,0); + BasicBlock* label_afterfor_65 = new BasicBlock("afterfor",func_to_array,0); + + // Block entry (label_entry_63) + ICmpInst* int1_cmp_66 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_62, const_int32_30, "cmp", label_entry_63); + new BranchInst(label_forbody_64, label_afterfor_65, int1_cmp_66, label_entry_63); + + // Block forbody (label_forbody_64) + Argument* fwdref_69 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_68 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_64); + int32_i_0_reg2mem_0_68->reserveOperandSpace(2); + int32_i_0_reg2mem_0_68->addIncoming(const_int32_30, label_entry_63); + int32_i_0_reg2mem_0_68->addIncoming(fwdref_69, label_forbody_64); + + std::vector ptr_arraydecay_70_indices; + ptr_arraydecay_70_indices.push_back(int32_i_0_reg2mem_0_68); + ptr_arraydecay_70_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay_70 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_70_indices.begin(), ptr_arraydecay_70_indices.end(), "arraydecay", label_forbody_64); + GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_68, "arrayidx6", label_forbody_64); + LoadInst* packed_tmp7_71 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_64); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_71, const_int32_30, "tmp11", label_forbody_64); + StoreInst* void_72 = new StoreInst(float_tmp11, ptr_arraydecay_70, false, label_forbody_64); std::vector ptr_arrayidx13_indices; - ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_72); - ptr_arrayidx13_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_68); - ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_75, const_int32_36, "tmp15", label_forbody_68); - StoreInst* void_77 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_68); + ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_68); + ptr_arrayidx13_indices.push_back(const_int32_32); + Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_64); + ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_71, const_int32_32, "tmp15", label_forbody_64); + StoreInst* void_73 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_64); std::vector ptr_arrayidx17_indices; - ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_72); - ptr_arrayidx17_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_68); - ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_75, const_int32_37, "tmp19", label_forbody_68); - StoreInst* void_78 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_68); + ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_68); + ptr_arrayidx17_indices.push_back(const_int32_33); + Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_64); + ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_71, const_int32_33, "tmp19", label_forbody_64); + StoreInst* void_74 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_64); std::vector ptr_arrayidx21_indices; - ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_72); - ptr_arrayidx21_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_68); - ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_75, const_int32_38, "tmp23", label_forbody_68); - StoreInst* void_79 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_68); - BinaryOperator* int32_indvar_next_80 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_72, const_int32_36, "indvar.next", label_forbody_68); - ICmpInst* int1_exitcond_81 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_80, int32_num_attribs_66, "exitcond", label_forbody_68); - new BranchInst(label_afterfor_69, label_forbody_68, int1_exitcond_81, label_forbody_68); - - // Block afterfor (label_afterfor_69) - new ReturnInst(label_afterfor_69); + ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_68); + ptr_arrayidx21_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_64); + ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_71, const_int32_34, "tmp23", label_forbody_64); + StoreInst* void_75 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_64); + BinaryOperator* int32_indvar_next_76 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_68, const_int32_32, "indvar.next", label_forbody_64); + ICmpInst* int1_exitcond_77 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_76, int32_num_attribs_62, "exitcond", label_forbody_64); + new BranchInst(label_afterfor_65, label_forbody_64, int1_exitcond_77, label_forbody_64); + + // Block afterfor (label_afterfor_65) + new ReturnInst(label_afterfor_65); // Resolve Forward References - fwdref_73->replaceAllUsesWith(int32_indvar_next_80); delete fwdref_73; + fwdref_69->replaceAllUsesWith(int32_indvar_next_76); delete fwdref_69; } // Function: run_vertex_shader (func_run_vertex_shader) { Function::arg_iterator args = func_run_vertex_shader->arg_begin(); - Value* ptr_ainputs_84 = args++; - ptr_ainputs_84->setName("ainputs"); - Value* ptr_dests_85 = args++; - ptr_dests_85->setName("dests"); + Value* ptr_inputs = args++; + ptr_inputs->setName("inputs"); + Value* ptr_results = args++; + ptr_results->setName("results"); Value* ptr_aconsts = args++; ptr_aconsts->setName("aconsts"); Value* int32_num_vertices = args++; int32_num_vertices->setName("num_vertices"); Value* int32_num_inputs = args++; int32_num_inputs->setName("num_inputs"); - Value* int32_num_attribs_86 = args++; - int32_num_attribs_86->setName("num_attribs"); + Value* int32_num_attribs_80 = args++; + int32_num_attribs_80->setName("num_attribs"); Value* int32_num_consts = args++; int32_num_consts->setName("num_consts"); - BasicBlock* label_entry_87 = new BasicBlock("entry",func_run_vertex_shader,0); - BasicBlock* label_forbody6_i = new BasicBlock("forbody6.i",func_run_vertex_shader,0); - BasicBlock* label_forinc57_i = new BasicBlock("forinc57.i",func_run_vertex_shader,0); - BasicBlock* label_from_array_exit = new BasicBlock("from_array.exit",func_run_vertex_shader,0); - BasicBlock* label_forbody_i15 = new BasicBlock("forbody.i15",func_run_vertex_shader,0); + BasicBlock* label_entry_81 = new BasicBlock("entry",func_run_vertex_shader,0); + BasicBlock* label_forbody_i = new BasicBlock("forbody.i",func_run_vertex_shader,0); BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_vertex_shader,0); BasicBlock* label_forbody_preheader = new BasicBlock("forbody.preheader",func_run_vertex_shader,0); - BasicBlock* label_forbody_us = new BasicBlock("forbody.us",func_run_vertex_shader,0); - BasicBlock* label_to_array_exit_us = new BasicBlock("to_array.exit.us",func_run_vertex_shader,0); - BasicBlock* label_forbody_i_us = new BasicBlock("forbody.i.us",func_run_vertex_shader,0); - BasicBlock* label_forbody_88 = new BasicBlock("forbody",func_run_vertex_shader,0); - BasicBlock* label_afterfor_89 = new BasicBlock("afterfor",func_run_vertex_shader,0); - - // Block entry (label_entry_87) - AllocaInst* ptr_inputs = new AllocaInst(ArrayTy_21, "inputs", label_entry_87); - AllocaInst* ptr_consts = new AllocaInst(ArrayTy_23, "consts", label_entry_87); - AllocaInst* ptr_results = new AllocaInst(ArrayTy_21, "results", label_entry_87); - AllocaInst* ptr_temps = new AllocaInst(ArrayTy_25, "temps", label_entry_87); - AllocaInst* ptr_args = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_87); - ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_34, "cmp.i", label_entry_87); - ICmpInst* int1_cmp5_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs, const_int32_34, "cmp5.i", label_entry_87); - BinaryOperator* int1_bothcond_i = BinaryOperator::create(Instruction::And, int1_cmp5_i, int1_cmp_i, "bothcond.i", label_entry_87); - new BranchInst(label_forbody6_i, label_from_array_exit, int1_bothcond_i, label_entry_87); + BasicBlock* label_forbody_82 = new BasicBlock("forbody",func_run_vertex_shader,0); + BasicBlock* label_afterfor_83 = new BasicBlock("afterfor",func_run_vertex_shader,0); + + // Block entry (label_entry_81) + AllocaInst* ptr_consts = new AllocaInst(ArrayTy_21, "consts", label_entry_81); + AllocaInst* ptr_temps = new AllocaInst(ArrayTy_23, "temps", label_entry_81); + AllocaInst* ptr_args = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_81); + ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_30, "cmp.i", label_entry_81); + new BranchInst(label_forbody_i, label_from_consts_exit, int1_cmp_i, label_entry_81); + + // Block forbody.i (label_forbody_i) + Argument* fwdref_85 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i); + int32_i_0_reg2mem_0_i->reserveOperandSpace(2); + int32_i_0_reg2mem_0_i->addIncoming(const_int32_30, label_entry_81); + int32_i_0_reg2mem_0_i->addIncoming(fwdref_85, label_forbody_i); + + Argument* fwdref_86 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i); + packed_vec_0_reg2mem_0_i->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i->addIncoming(const_packed_31, label_entry_81); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_86, label_forbody_i); - // Block forbody6.i (label_forbody6_i) - Argument* fwdref_91 = new Argument(IntegerType::get(32)); - Argument* fwdref_92 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_i_ph = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i.ph", label_forbody6_i); - int32_i_0_reg2mem_0_i_ph->reserveOperandSpace(3); - int32_i_0_reg2mem_0_i_ph->addIncoming(const_int32_34, label_entry_87); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_91, label_forinc57_i); - int32_i_0_reg2mem_0_i_ph->addIncoming(fwdref_92, label_forbody6_i); - - Argument* fwdref_93 = new Argument(IntegerType::get(32)); - PHINode* int32_j_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i); - int32_j_0_reg2mem_0_i->reserveOperandSpace(3); - int32_j_0_reg2mem_0_i->addIncoming(fwdref_93, label_forbody6_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_34, label_forinc57_i); - int32_j_0_reg2mem_0_i->addIncoming(const_int32_34, label_entry_87); - - Argument* fwdref_94 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody6_i); - packed_vec_0_reg2mem_0_i->reserveOperandSpace(3); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_94, label_forbody6_i); - packed_vec_0_reg2mem_0_i->addIncoming(const_packed_35, label_entry_87); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_94, label_forinc57_i); - - std::vector ptr_arraydecay11_i_indices; - ptr_arraydecay11_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); - ptr_arraydecay11_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arraydecay11_i_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay11_i = new GetElementPtrInst(ptr_ainputs_84, ptr_arraydecay11_i_indices.begin(), ptr_arraydecay11_i_indices.end(), "arraydecay11.i", label_forbody6_i); - LoadInst* float_tmp13_i = new LoadInst(ptr_arraydecay11_i, "tmp13.i", false, label_forbody6_i); - InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp13_i, const_int32_34, "tmp15.i", label_forbody6_i); - std::vector ptr_arrayidx23_i_indices; - ptr_arrayidx23_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); - ptr_arrayidx23_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx23_i_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx23_i = new GetElementPtrInst(ptr_ainputs_84, ptr_arrayidx23_i_indices.begin(), ptr_arrayidx23_i_indices.end(), "arrayidx23.i", label_forbody6_i); - LoadInst* float_tmp24_i = new LoadInst(ptr_arrayidx23_i, "tmp24.i", false, label_forbody6_i); - InsertElementInst* packed_tmp26_i = new InsertElementInst(packed_tmp15_i, float_tmp24_i, const_int32_36, "tmp26.i", label_forbody6_i); - std::vector ptr_arrayidx34_i_indices; - ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); - ptr_arrayidx34_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx34_i_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_ainputs_84, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody6_i); - LoadInst* float_tmp35_i = new LoadInst(ptr_arrayidx34_i, "tmp35.i", false, label_forbody6_i); - InsertElementInst* packed_tmp37_i = new InsertElementInst(packed_tmp26_i, float_tmp35_i, const_int32_37, "tmp37.i", label_forbody6_i); - std::vector ptr_arrayidx45_i_indices; - ptr_arrayidx45_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); - ptr_arrayidx45_i_indices.push_back(int32_j_0_reg2mem_0_i); - ptr_arrayidx45_i_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx45_i = new GetElementPtrInst(ptr_ainputs_84, ptr_arrayidx45_i_indices.begin(), ptr_arrayidx45_i_indices.end(), "arrayidx45.i", label_forbody6_i); - LoadInst* float_tmp46_i = new LoadInst(ptr_arrayidx45_i, "tmp46.i", false, label_forbody6_i); - InsertElementInst* packed_tmp48_i = new InsertElementInst(packed_tmp37_i, float_tmp46_i, const_int32_38, "tmp48.i", label_forbody6_i); - std::vector ptr_arrayidx54_i_indices; - ptr_arrayidx54_i_indices.push_back(const_int32_34); - ptr_arrayidx54_i_indices.push_back(int32_i_0_reg2mem_0_i_ph); - ptr_arrayidx54_i_indices.push_back(int32_j_0_reg2mem_0_i); - Instruction* ptr_arrayidx54_i = new GetElementPtrInst(ptr_inputs, ptr_arrayidx54_i_indices.begin(), ptr_arrayidx54_i_indices.end(), "arrayidx54.i", label_forbody6_i); - StoreInst* void_95 = new StoreInst(packed_tmp48_i, ptr_arrayidx54_i, false, label_forbody6_i); - BinaryOperator* int32_inc_i = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i, const_int32_36, "inc.i", label_forbody6_i); - ICmpInst* int1_cmp59_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i, int32_num_inputs, "cmp59.i", label_forbody6_i); - new BranchInst(label_forbody6_i, label_forinc57_i, int1_cmp59_i, label_forbody6_i); - - // Block forinc57.i (label_forinc57_i) - BinaryOperator* int32_inc59_i = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_ph, const_int32_36, "inc59.i", label_forinc57_i); - ICmpInst* int1_cmp17_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59_i, int32_num_vertices, "cmp17.i", label_forinc57_i); - new BranchInst(label_forbody6_i, label_from_array_exit, int1_cmp17_i, label_forinc57_i); - - // Block from_array.exit (label_from_array_exit) - ICmpInst* int1_cmp_i4 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_34, "cmp.i4", label_from_array_exit); - new BranchInst(label_forbody_i15, label_from_consts_exit, int1_cmp_i4, label_from_array_exit); - - // Block forbody.i15 (label_forbody_i15) - Argument* fwdref_99 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_i5 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i5", label_forbody_i15); - int32_i_0_reg2mem_0_i5->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i5->addIncoming(const_int32_34, label_from_array_exit); - int32_i_0_reg2mem_0_i5->addIncoming(fwdref_99, label_forbody_i15); - - Argument* fwdref_100 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i6 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i6", label_forbody_i15); - packed_vec_0_reg2mem_0_i6->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i6->addIncoming(const_packed_35, label_from_array_exit); - packed_vec_0_reg2mem_0_i6->addIncoming(fwdref_100, label_forbody_i15); - - std::vector ptr_arraydecay_i7_indices; - ptr_arraydecay_i7_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arraydecay_i7_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay_i7 = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i7_indices.begin(), ptr_arraydecay_i7_indices.end(), "arraydecay.i7", label_forbody_i15); - LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i7, "tmp5.i", false, label_forbody_i15); - InsertElementInst* packed_tmp7_i8 = new InsertElementInst(packed_vec_0_reg2mem_0_i6, float_tmp5_i, const_int32_34, "tmp7.i8", label_forbody_i15); + std::vector ptr_arraydecay_i_indices; + ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_i); + ptr_arraydecay_i_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_i); + LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i, "tmp5.i", false, label_forbody_i); + InsertElementInst* packed_tmp7_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp5_i, const_int32_30, "tmp7.i", label_forbody_i); std::vector ptr_arrayidx12_i_indices; - ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx12_i_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i15); - LoadInst* float_tmp13_i9 = new LoadInst(ptr_arrayidx12_i, "tmp13.i9", false, label_forbody_i15); - InsertElementInst* packed_tmp15_i10 = new InsertElementInst(packed_tmp7_i8, float_tmp13_i9, const_int32_36, "tmp15.i10", label_forbody_i15); + ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i); + ptr_arrayidx12_i_indices.push_back(const_int32_32); + Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i); + LoadInst* float_tmp13_i = new LoadInst(ptr_arrayidx12_i, "tmp13.i", false, label_forbody_i); + InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_tmp7_i, float_tmp13_i, const_int32_32, "tmp15.i", label_forbody_i); std::vector ptr_arrayidx20_i_indices; - ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx20_i_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i15); - LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i15); - InsertElementInst* packed_tmp23_i11 = new InsertElementInst(packed_tmp15_i10, float_tmp21_i, const_int32_37, "tmp23.i11", label_forbody_i15); + ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i); + ptr_arrayidx20_i_indices.push_back(const_int32_33); + Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i); + LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i); + InsertElementInst* packed_tmp23_i = new InsertElementInst(packed_tmp15_i, float_tmp21_i, const_int32_33, "tmp23.i", label_forbody_i); std::vector ptr_arrayidx28_i_indices; - ptr_arrayidx28_i_indices.push_back(int32_i_0_reg2mem_0_i5); - ptr_arrayidx28_i_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx28_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i_indices.begin(), ptr_arrayidx28_i_indices.end(), "arrayidx28.i", label_forbody_i15); - LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i15); - InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i11, float_tmp29_i, const_int32_38, "tmp31.i", label_forbody_i15); - std::vector ptr_arrayidx34_i12_indices; - ptr_arrayidx34_i12_indices.push_back(const_int32_34); - ptr_arrayidx34_i12_indices.push_back(int32_i_0_reg2mem_0_i5); - Instruction* ptr_arrayidx34_i12 = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i12_indices.begin(), ptr_arrayidx34_i12_indices.end(), "arrayidx34.i12", label_forbody_i15); - StoreInst* void_101 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i12, false, label_forbody_i15); - BinaryOperator* int32_indvar_next24 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i5, const_int32_36, "indvar.next24", label_forbody_i15); - ICmpInst* int1_exitcond25 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next24, int32_num_consts, "exitcond25", label_forbody_i15); - new BranchInst(label_from_consts_exit, label_forbody_i15, int1_exitcond25, label_forbody_i15); + ptr_arrayidx28_i_indices.push_back(int32_i_0_reg2mem_0_i); + ptr_arrayidx28_i_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx28_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i_indices.begin(), ptr_arrayidx28_i_indices.end(), "arrayidx28.i", label_forbody_i); + LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i); + InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i, float_tmp29_i, const_int32_34, "tmp31.i", label_forbody_i); + std::vector ptr_arrayidx34_i_indices; + ptr_arrayidx34_i_indices.push_back(const_int32_30); + ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i); + Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody_i); + StoreInst* void_87 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i, false, label_forbody_i); + BinaryOperator* int32_indvar_next6 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i, const_int32_32, "indvar.next6", label_forbody_i); + ICmpInst* int1_exitcond7 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next6, int32_num_consts, "exitcond7", label_forbody_i); + new BranchInst(label_from_consts_exit, label_forbody_i, int1_exitcond7, label_forbody_i); // Block from_consts.exit (label_from_consts_exit) - std::vector ptr_tmp6_indices; - ptr_tmp6_indices.push_back(const_int32_34); - ptr_tmp6_indices.push_back(const_int32_38); - Instruction* ptr_tmp6 = new GetElementPtrInst(ptr_args, ptr_tmp6_indices.begin(), ptr_tmp6_indices.end(), "tmp6", label_from_consts_exit); - std::vector ptr_arraydecay7_indices; - ptr_arraydecay7_indices.push_back(const_int32_34); - ptr_arraydecay7_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay7 = new GetElementPtrInst(ptr_consts, ptr_arraydecay7_indices.begin(), ptr_arraydecay7_indices.end(), "arraydecay7", label_from_consts_exit); - StoreInst* void_103 = new StoreInst(ptr_arraydecay7, ptr_tmp6, false, label_from_consts_exit); - std::vector ptr_tmp8_indices; - ptr_tmp8_indices.push_back(const_int32_34); - ptr_tmp8_indices.push_back(const_int32_37); - Instruction* ptr_tmp8 = new GetElementPtrInst(ptr_args, ptr_tmp8_indices.begin(), ptr_tmp8_indices.end(), "tmp8", label_from_consts_exit); - std::vector ptr_arraydecay9_indices; - ptr_arraydecay9_indices.push_back(const_int32_34); - ptr_arraydecay9_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay9 = new GetElementPtrInst(ptr_temps, ptr_arraydecay9_indices.begin(), ptr_arraydecay9_indices.end(), "arraydecay9", label_from_consts_exit); - StoreInst* void_104 = new StoreInst(ptr_arraydecay9, ptr_tmp8, false, label_from_consts_exit); - new BranchInst(label_forbody_preheader, label_afterfor_89, int1_cmp_i, label_from_consts_exit); + std::vector ptr_tmp2_indices; + ptr_tmp2_indices.push_back(const_int32_30); + ptr_tmp2_indices.push_back(const_int32_34); + Instruction* ptr_tmp2 = new GetElementPtrInst(ptr_args, ptr_tmp2_indices.begin(), ptr_tmp2_indices.end(), "tmp2", label_from_consts_exit); + std::vector ptr_arraydecay3_indices; + ptr_arraydecay3_indices.push_back(const_int32_30); + ptr_arraydecay3_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay3 = new GetElementPtrInst(ptr_consts, ptr_arraydecay3_indices.begin(), ptr_arraydecay3_indices.end(), "arraydecay3", label_from_consts_exit); + StoreInst* void_89 = new StoreInst(ptr_arraydecay3, ptr_tmp2, false, label_from_consts_exit); + std::vector ptr_tmp4_indices; + ptr_tmp4_indices.push_back(const_int32_30); + ptr_tmp4_indices.push_back(const_int32_33); + Instruction* ptr_tmp4 = new GetElementPtrInst(ptr_args, ptr_tmp4_indices.begin(), ptr_tmp4_indices.end(), "tmp4", label_from_consts_exit); + std::vector ptr_arraydecay5_indices; + ptr_arraydecay5_indices.push_back(const_int32_30); + ptr_arraydecay5_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay5 = new GetElementPtrInst(ptr_temps, ptr_arraydecay5_indices.begin(), ptr_arraydecay5_indices.end(), "arraydecay5", label_from_consts_exit); + StoreInst* void_90 = new StoreInst(ptr_arraydecay5, ptr_tmp4, false, label_from_consts_exit); + ICmpInst* int1_cmp_91 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_30, "cmp", label_from_consts_exit); + new BranchInst(label_forbody_preheader, label_afterfor_83, int1_cmp_91, label_from_consts_exit); // Block forbody.preheader (label_forbody_preheader) + std::vector ptr_tmp8_indices; + ptr_tmp8_indices.push_back(const_int32_30); + ptr_tmp8_indices.push_back(const_int32_30); + Instruction* ptr_tmp8 = new GetElementPtrInst(ptr_args, ptr_tmp8_indices.begin(), ptr_tmp8_indices.end(), "tmp8", label_forbody_preheader); std::vector ptr_tmp12_indices; - ptr_tmp12_indices.push_back(const_int32_34); - ptr_tmp12_indices.push_back(const_int32_34); + ptr_tmp12_indices.push_back(const_int32_30); + ptr_tmp12_indices.push_back(const_int32_32); Instruction* ptr_tmp12 = new GetElementPtrInst(ptr_args, ptr_tmp12_indices.begin(), ptr_tmp12_indices.end(), "tmp12", label_forbody_preheader); - std::vector ptr_tmp16_indices; - ptr_tmp16_indices.push_back(const_int32_34); - ptr_tmp16_indices.push_back(const_int32_36); - Instruction* ptr_tmp16 = new GetElementPtrInst(ptr_args, ptr_tmp16_indices.begin(), ptr_tmp16_indices.end(), "tmp16", label_forbody_preheader); - ICmpInst* int1_cmp_i1 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_86, const_int32_34, "cmp.i1", label_forbody_preheader); - new BranchInst(label_forbody_us, label_forbody_88, int1_cmp_i1, label_forbody_preheader); - - // Block forbody.us (label_forbody_us) - Argument* fwdref_107 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.us", label_forbody_us); - int32_i_0_reg2mem_0_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_us->addIncoming(const_int32_34, label_forbody_preheader); - int32_i_0_reg2mem_0_us->addIncoming(fwdref_107, label_to_array_exit_us); - - std::vector ptr_arraydecay15_us_indices; - ptr_arraydecay15_us_indices.push_back(const_int32_34); - ptr_arraydecay15_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay15_us_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay15_us = new GetElementPtrInst(ptr_results, ptr_arraydecay15_us_indices.begin(), ptr_arraydecay15_us_indices.end(), "arraydecay15.us", label_forbody_us); - StoreInst* void_108 = new StoreInst(ptr_arraydecay15_us, ptr_tmp12, false, label_forbody_us); - std::vector ptr_arraydecay20_us_indices; - ptr_arraydecay20_us_indices.push_back(const_int32_34); - ptr_arraydecay20_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay20_us_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay20_us = new GetElementPtrInst(ptr_inputs, ptr_arraydecay20_us_indices.begin(), ptr_arraydecay20_us_indices.end(), "arraydecay20.us", label_forbody_us); - StoreInst* void_109 = new StoreInst(ptr_arraydecay20_us, ptr_tmp16, false, label_forbody_us); - CallInst* void_110 = new CallInst(func_execute_shader, ptr_args, "", label_forbody_us); - void_110->setCallingConv(CallingConv::C); - void_110->setTailCall(false); - LoadInst* ptr_tmp26_us = new LoadInst(ptr_tmp12, "tmp26.us", false, label_forbody_us); - new BranchInst(label_forbody_i_us, label_forbody_us); - - // Block to_array.exit.us (label_to_array_exit_us) - BinaryOperator* int32_inc_us = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_us, const_int32_36, "inc.us", label_to_array_exit_us); - ICmpInst* int1_cmp21_us = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_us, int32_num_vertices, "cmp21.us", label_to_array_exit_us); - new BranchInst(label_forbody_us, label_afterfor_89, int1_cmp21_us, label_to_array_exit_us); - - // Block forbody.i.us (label_forbody_i_us) - Argument* fwdref_113 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_i2_us = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i2.us", label_forbody_i_us); - int32_i_0_reg2mem_0_i2_us->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i2_us->addIncoming(const_int32_34, label_forbody_us); - int32_i_0_reg2mem_0_i2_us->addIncoming(fwdref_113, label_forbody_i_us); - - std::vector ptr_arraydecay_i_us_indices; - ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arraydecay_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arraydecay_i_us_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay_i_us = new GetElementPtrInst(ptr_dests_85, ptr_arraydecay_i_us_indices.begin(), ptr_arraydecay_i_us_indices.end(), "arraydecay.i.us", label_forbody_i_us); - GetElementPtrInst* ptr_arrayidx6_i_us = new GetElementPtrInst(ptr_tmp26_us, int32_i_0_reg2mem_0_i2_us, "arrayidx6.i.us", label_forbody_i_us); - LoadInst* packed_tmp7_i_us = new LoadInst(ptr_arrayidx6_i_us, "tmp7.i.us", false, label_forbody_i_us); - ExtractElementInst* float_tmp11_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_34, "tmp11.i.us", label_forbody_i_us); - StoreInst* void_114 = new StoreInst(float_tmp11_i_us, ptr_arraydecay_i_us, false, label_forbody_i_us); - std::vector ptr_arrayidx13_i_us_indices; - ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arrayidx13_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx13_i_us_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx13_i_us = new GetElementPtrInst(ptr_dests_85, ptr_arrayidx13_i_us_indices.begin(), ptr_arrayidx13_i_us_indices.end(), "arrayidx13.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp15_i3_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_36, "tmp15.i3.us", label_forbody_i_us); - StoreInst* void_115 = new StoreInst(float_tmp15_i3_us, ptr_arrayidx13_i_us, false, label_forbody_i_us); - std::vector ptr_arrayidx17_i_us_indices; - ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arrayidx17_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx17_i_us_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx17_i_us = new GetElementPtrInst(ptr_dests_85, ptr_arrayidx17_i_us_indices.begin(), ptr_arrayidx17_i_us_indices.end(), "arrayidx17.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp19_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_37, "tmp19.i.us", label_forbody_i_us); - StoreInst* void_116 = new StoreInst(float_tmp19_i_us, ptr_arrayidx17_i_us, false, label_forbody_i_us); - std::vector ptr_arrayidx21_i_us_indices; - ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_us); - ptr_arrayidx21_i_us_indices.push_back(int32_i_0_reg2mem_0_i2_us); - ptr_arrayidx21_i_us_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx21_i_us = new GetElementPtrInst(ptr_dests_85, ptr_arrayidx21_i_us_indices.begin(), ptr_arrayidx21_i_us_indices.end(), "arrayidx21.i.us", label_forbody_i_us); - ExtractElementInst* float_tmp23_i_us = new ExtractElementInst(packed_tmp7_i_us, const_int32_38, "tmp23.i.us", label_forbody_i_us); - StoreInst* void_117 = new StoreInst(float_tmp23_i_us, ptr_arrayidx21_i_us, false, label_forbody_i_us); - BinaryOperator* int32_indvar_next_118 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i2_us, const_int32_36, "indvar.next", label_forbody_i_us); - ICmpInst* int1_exitcond_119 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_118, int32_num_attribs_86, "exitcond", label_forbody_i_us); - new BranchInst(label_to_array_exit_us, label_forbody_i_us, int1_exitcond_119, label_forbody_i_us); - - // Block forbody (label_forbody_88) - Argument* fwdref_122 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_121 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_88); - int32_i_0_reg2mem_0_121->reserveOperandSpace(2); - int32_i_0_reg2mem_0_121->addIncoming(const_int32_34, label_forbody_preheader); - int32_i_0_reg2mem_0_121->addIncoming(fwdref_122, label_forbody_88); - - std::vector ptr_arraydecay15_indices; - ptr_arraydecay15_indices.push_back(const_int32_34); - ptr_arraydecay15_indices.push_back(int32_i_0_reg2mem_0_121); - ptr_arraydecay15_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay15 = new GetElementPtrInst(ptr_results, ptr_arraydecay15_indices.begin(), ptr_arraydecay15_indices.end(), "arraydecay15", label_forbody_88); - StoreInst* void_123 = new StoreInst(ptr_arraydecay15, ptr_tmp12, false, label_forbody_88); - std::vector ptr_arraydecay20_indices; - ptr_arraydecay20_indices.push_back(const_int32_34); - ptr_arraydecay20_indices.push_back(int32_i_0_reg2mem_0_121); - ptr_arraydecay20_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay20 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay20_indices.begin(), ptr_arraydecay20_indices.end(), "arraydecay20", label_forbody_88); - StoreInst* void_124 = new StoreInst(ptr_arraydecay20, ptr_tmp16, false, label_forbody_88); - CallInst* void_125 = new CallInst(func_execute_shader, ptr_args, "", label_forbody_88); - void_125->setCallingConv(CallingConv::C); - void_125->setTailCall(false); - BinaryOperator* int32_inc_126 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_121, const_int32_36, "inc", label_forbody_88); - ICmpInst* int1_cmp21 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_126, int32_num_vertices, "cmp21", label_forbody_88); - new BranchInst(label_forbody_88, label_afterfor_89, int1_cmp21, label_forbody_88); - - // Block afterfor (label_afterfor_89) - new ReturnInst(label_afterfor_89); + new BranchInst(label_forbody_82, label_forbody_preheader); + + // Block forbody (label_forbody_82) + Argument* fwdref_95 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_94 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_82); + int32_i_0_reg2mem_0_94->reserveOperandSpace(2); + int32_i_0_reg2mem_0_94->addIncoming(const_int32_30, label_forbody_preheader); + int32_i_0_reg2mem_0_94->addIncoming(fwdref_95, label_forbody_82); + + std::vector ptr_arraydecay11_96_indices; + ptr_arraydecay11_96_indices.push_back(int32_i_0_reg2mem_0_94); + ptr_arraydecay11_96_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay11_96 = new GetElementPtrInst(ptr_results, ptr_arraydecay11_96_indices.begin(), ptr_arraydecay11_96_indices.end(), "arraydecay11", label_forbody_82); + StoreInst* void_97 = new StoreInst(ptr_arraydecay11_96, ptr_tmp8, false, label_forbody_82); + std::vector ptr_arraydecay16_indices; + ptr_arraydecay16_indices.push_back(int32_i_0_reg2mem_0_94); + ptr_arraydecay16_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay16 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay16_indices.begin(), ptr_arraydecay16_indices.end(), "arraydecay16", label_forbody_82); + StoreInst* void_98 = new StoreInst(ptr_arraydecay16, ptr_tmp12, false, label_forbody_82); + CallInst* void_99 = new CallInst(func_execute_shader, ptr_args, "", label_forbody_82); + void_99->setCallingConv(CallingConv::C); + void_99->setTailCall(false); + BinaryOperator* int32_indvar_next_100 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_94, const_int32_32, "indvar.next", label_forbody_82); + ICmpInst* int1_exitcond_101 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_100, int32_num_vertices, "exitcond", label_forbody_82); + new BranchInst(label_afterfor_83, label_forbody_82, int1_exitcond_101, label_forbody_82); + + // Block afterfor (label_afterfor_83) + new ReturnInst(label_afterfor_83); // Resolve Forward References - fwdref_107->replaceAllUsesWith(int32_inc_us); delete fwdref_107; - fwdref_92->replaceAllUsesWith(int32_i_0_reg2mem_0_i_ph); delete fwdref_92; - fwdref_94->replaceAllUsesWith(packed_tmp48_i); delete fwdref_94; - fwdref_93->replaceAllUsesWith(int32_inc_i); delete fwdref_93; - fwdref_91->replaceAllUsesWith(int32_inc59_i); delete fwdref_91; - fwdref_100->replaceAllUsesWith(packed_tmp31_i); delete fwdref_100; - fwdref_99->replaceAllUsesWith(int32_indvar_next24); delete fwdref_99; - fwdref_113->replaceAllUsesWith(int32_indvar_next_118); delete fwdref_113; - fwdref_122->replaceAllUsesWith(int32_inc_126); delete fwdref_122; + fwdref_86->replaceAllUsesWith(packed_tmp31_i); delete fwdref_86; + fwdref_85->replaceAllUsesWith(int32_indvar_next6); delete fwdref_85; + fwdref_95->replaceAllUsesWith(int32_indvar_next_100); delete fwdref_95; } @@ -780,441 +613,180 @@ Module* createBaseShader() { float_x->setName("x"); Value* float_y = args++; float_y->setName("y"); - Value* ptr_dests_129 = args++; - ptr_dests_129->setName("dests"); - Value* ptr_ainputs_130 = args++; - ptr_ainputs_130->setName("ainputs"); - Value* int32_num_inputs_131 = args++; - int32_num_inputs_131->setName("num_inputs"); - Value* ptr_aconsts_132 = args++; - ptr_aconsts_132->setName("aconsts"); - Value* int32_num_consts_133 = args++; - int32_num_consts_133->setName("num_consts"); + Value* ptr_results_104 = args++; + ptr_results_104->setName("results"); + Value* ptr_inputs_105 = args++; + ptr_inputs_105->setName("inputs"); + Value* int32_num_inputs_106 = args++; + int32_num_inputs_106->setName("num_inputs"); + Value* ptr_aconsts_107 = args++; + ptr_aconsts_107->setName("aconsts"); + Value* int32_num_consts_108 = args++; + int32_num_consts_108->setName("num_consts"); Value* ptr_samplers = args++; ptr_samplers->setName("samplers"); - BasicBlock* label_entry_134 = new BasicBlock("entry",func_run_fragment_shader,0); - BasicBlock* label_forbody6_i_135 = new BasicBlock("forbody6.i",func_run_fragment_shader,0); - BasicBlock* label_from_array_exit_136 = new BasicBlock("from_array.exit",func_run_fragment_shader,0); - BasicBlock* label_forbody_i13 = new BasicBlock("forbody.i13",func_run_fragment_shader,0); - BasicBlock* label_from_consts_exit_137 = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); - BasicBlock* label_forbody_138 = new BasicBlock("forbody",func_run_fragment_shader,0); - BasicBlock* label_afterfor_139 = new BasicBlock("afterfor",func_run_fragment_shader,0); - BasicBlock* label_forbody6_i_1 = new BasicBlock("forbody6.i.1",func_run_fragment_shader,0); - BasicBlock* label_forbody6_i_2 = new BasicBlock("forbody6.i.2",func_run_fragment_shader,0); - BasicBlock* label_forbody6_i_3 = new BasicBlock("forbody6.i.3",func_run_fragment_shader,0); + BasicBlock* label_entry_109 = new BasicBlock("entry",func_run_fragment_shader,0); + BasicBlock* label_forbody_i_110 = new BasicBlock("forbody.i",func_run_fragment_shader,0); + BasicBlock* label_from_consts_exit_111 = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); - // Block entry (label_entry_134) - AllocaInst* ptr_inputs_140 = new AllocaInst(ArrayTy_33, "inputs", label_entry_134); - AllocaInst* ptr_consts_141 = new AllocaInst(ArrayTy_23, "consts", label_entry_134); - AllocaInst* ptr_results_142 = new AllocaInst(ArrayTy_33, "results", label_entry_134); - AllocaInst* ptr_temps_143 = new AllocaInst(ArrayTy_25, "temps", label_entry_134); - AllocaInst* ptr_args_144 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_134); + // Block entry (label_entry_109) + AllocaInst* ptr_consts_112 = new AllocaInst(ArrayTy_21, "consts", label_entry_109); + AllocaInst* ptr_temps_113 = new AllocaInst(ArrayTy_23, "temps", label_entry_109); + AllocaInst* ptr_args_114 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_109); std::vector ptr_tmp_indices; - ptr_tmp_indices.push_back(const_int32_34); - ptr_tmp_indices.push_back(const_int32_39); - Instruction* ptr_tmp = new GetElementPtrInst(ptr_args_144, ptr_tmp_indices.begin(), ptr_tmp_indices.end(), "tmp", label_entry_134); - StoreInst* void_145 = new StoreInst(const_int32_34, ptr_tmp, false, label_entry_134); - ICmpInst* int1_cmp5_i_146 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_inputs_131, const_int32_34, "cmp5.i", label_entry_134); - new BranchInst(label_forbody6_i_135, label_from_array_exit_136, int1_cmp5_i_146, label_entry_134); - - // Block forbody6.i (label_forbody6_i_135) - Argument* fwdref_149 = new Argument(IntegerType::get(32)); - PHINode* int32_j_0_reg2mem_0_i_148 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i", label_forbody6_i_135); - int32_j_0_reg2mem_0_i_148->reserveOperandSpace(2); - int32_j_0_reg2mem_0_i_148->addIncoming(const_int32_34, label_entry_134); - int32_j_0_reg2mem_0_i_148->addIncoming(fwdref_149, label_forbody6_i_135); - - Argument* fwdref_151 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i_150 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody6_i_135); - packed_vec_0_reg2mem_0_i_150->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_150->addIncoming(const_packed_35, label_entry_134); - packed_vec_0_reg2mem_0_i_150->addIncoming(fwdref_151, label_forbody6_i_135); - - std::vector ptr_arraydecay11_i_152_indices; - ptr_arraydecay11_i_152_indices.push_back(const_int32_34); - ptr_arraydecay11_i_152_indices.push_back(int32_j_0_reg2mem_0_i_148); - ptr_arraydecay11_i_152_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay11_i_152 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_152_indices.begin(), ptr_arraydecay11_i_152_indices.end(), "arraydecay11.i", label_forbody6_i_135); - LoadInst* float_tmp13_i_153 = new LoadInst(ptr_arraydecay11_i_152, "tmp13.i", false, label_forbody6_i_135); - InsertElementInst* packed_tmp15_i_154 = new InsertElementInst(packed_vec_0_reg2mem_0_i_150, float_tmp13_i_153, const_int32_34, "tmp15.i", label_forbody6_i_135); - std::vector ptr_arrayidx23_i_155_indices; - ptr_arrayidx23_i_155_indices.push_back(const_int32_34); - ptr_arrayidx23_i_155_indices.push_back(int32_j_0_reg2mem_0_i_148); - ptr_arrayidx23_i_155_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx23_i_155 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_155_indices.begin(), ptr_arrayidx23_i_155_indices.end(), "arrayidx23.i", label_forbody6_i_135); - LoadInst* float_tmp24_i_156 = new LoadInst(ptr_arrayidx23_i_155, "tmp24.i", false, label_forbody6_i_135); - InsertElementInst* packed_tmp26_i_157 = new InsertElementInst(packed_tmp15_i_154, float_tmp24_i_156, const_int32_36, "tmp26.i", label_forbody6_i_135); - std::vector ptr_arrayidx34_i_158_indices; - ptr_arrayidx34_i_158_indices.push_back(const_int32_34); - ptr_arrayidx34_i_158_indices.push_back(int32_j_0_reg2mem_0_i_148); - ptr_arrayidx34_i_158_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx34_i_158 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_158_indices.begin(), ptr_arrayidx34_i_158_indices.end(), "arrayidx34.i", label_forbody6_i_135); - LoadInst* float_tmp35_i_159 = new LoadInst(ptr_arrayidx34_i_158, "tmp35.i", false, label_forbody6_i_135); - InsertElementInst* packed_tmp37_i_160 = new InsertElementInst(packed_tmp26_i_157, float_tmp35_i_159, const_int32_37, "tmp37.i", label_forbody6_i_135); - std::vector ptr_arrayidx45_i_161_indices; - ptr_arrayidx45_i_161_indices.push_back(const_int32_34); - ptr_arrayidx45_i_161_indices.push_back(int32_j_0_reg2mem_0_i_148); - ptr_arrayidx45_i_161_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx45_i_161 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_161_indices.begin(), ptr_arrayidx45_i_161_indices.end(), "arrayidx45.i", label_forbody6_i_135); - LoadInst* float_tmp46_i_162 = new LoadInst(ptr_arrayidx45_i_161, "tmp46.i", false, label_forbody6_i_135); - InsertElementInst* packed_tmp48_i_163 = new InsertElementInst(packed_tmp37_i_160, float_tmp46_i_162, const_int32_38, "tmp48.i", label_forbody6_i_135); - std::vector ptr_arrayidx54_i_164_indices; - ptr_arrayidx54_i_164_indices.push_back(const_int32_34); - ptr_arrayidx54_i_164_indices.push_back(const_int32_34); - ptr_arrayidx54_i_164_indices.push_back(int32_j_0_reg2mem_0_i_148); - Instruction* ptr_arrayidx54_i_164 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_164_indices.begin(), ptr_arrayidx54_i_164_indices.end(), "arrayidx54.i", label_forbody6_i_135); - StoreInst* void_165 = new StoreInst(packed_tmp48_i_163, ptr_arrayidx54_i_164, false, label_forbody6_i_135); - BinaryOperator* int32_inc_i_166 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_148, const_int32_36, "inc.i", label_forbody6_i_135); - ICmpInst* int1_cmp59_i_167 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_166, int32_num_inputs_131, "cmp59.i", label_forbody6_i_135); - new BranchInst(label_forbody6_i_135, label_forbody6_i_1, int1_cmp59_i_167, label_forbody6_i_135); - - // Block from_array.exit (label_from_array_exit_136) - ICmpInst* int1_cmp_i_169 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_133, const_int32_34, "cmp.i", label_from_array_exit_136); - new BranchInst(label_forbody_i13, label_from_consts_exit_137, int1_cmp_i_169, label_from_array_exit_136); - - // Block forbody.i13 (label_forbody_i13) - Argument* fwdref_171 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_i3 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i3", label_forbody_i13); - int32_i_0_reg2mem_0_i3->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i3->addIncoming(const_int32_34, label_from_array_exit_136); - int32_i_0_reg2mem_0_i3->addIncoming(fwdref_171, label_forbody_i13); - - Argument* fwdref_172 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i4 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i4", label_forbody_i13); - packed_vec_0_reg2mem_0_i4->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i4->addIncoming(const_packed_35, label_from_array_exit_136); - packed_vec_0_reg2mem_0_i4->addIncoming(fwdref_172, label_forbody_i13); - - std::vector ptr_arraydecay_i5_indices; - ptr_arraydecay_i5_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arraydecay_i5_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay_i5 = new GetElementPtrInst(ptr_aconsts_132, ptr_arraydecay_i5_indices.begin(), ptr_arraydecay_i5_indices.end(), "arraydecay.i5", label_forbody_i13); - LoadInst* float_tmp5_i_173 = new LoadInst(ptr_arraydecay_i5, "tmp5.i", false, label_forbody_i13); - InsertElementInst* packed_tmp7_i6 = new InsertElementInst(packed_vec_0_reg2mem_0_i4, float_tmp5_i_173, const_int32_34, "tmp7.i6", label_forbody_i13); - std::vector ptr_arrayidx12_i_174_indices; - ptr_arrayidx12_i_174_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arrayidx12_i_174_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx12_i_174 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx12_i_174_indices.begin(), ptr_arrayidx12_i_174_indices.end(), "arrayidx12.i", label_forbody_i13); - LoadInst* float_tmp13_i7 = new LoadInst(ptr_arrayidx12_i_174, "tmp13.i7", false, label_forbody_i13); - InsertElementInst* packed_tmp15_i8 = new InsertElementInst(packed_tmp7_i6, float_tmp13_i7, const_int32_36, "tmp15.i8", label_forbody_i13); - std::vector ptr_arrayidx20_i_175_indices; - ptr_arrayidx20_i_175_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arrayidx20_i_175_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx20_i_175 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx20_i_175_indices.begin(), ptr_arrayidx20_i_175_indices.end(), "arrayidx20.i", label_forbody_i13); - LoadInst* float_tmp21_i_176 = new LoadInst(ptr_arrayidx20_i_175, "tmp21.i", false, label_forbody_i13); - InsertElementInst* packed_tmp23_i9 = new InsertElementInst(packed_tmp15_i8, float_tmp21_i_176, const_int32_37, "tmp23.i9", label_forbody_i13); - std::vector ptr_arrayidx28_i_177_indices; - ptr_arrayidx28_i_177_indices.push_back(int32_i_0_reg2mem_0_i3); - ptr_arrayidx28_i_177_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx28_i_177 = new GetElementPtrInst(ptr_aconsts_132, ptr_arrayidx28_i_177_indices.begin(), ptr_arrayidx28_i_177_indices.end(), "arrayidx28.i", label_forbody_i13); - LoadInst* float_tmp29_i_178 = new LoadInst(ptr_arrayidx28_i_177, "tmp29.i", false, label_forbody_i13); - InsertElementInst* packed_tmp31_i_179 = new InsertElementInst(packed_tmp23_i9, float_tmp29_i_178, const_int32_38, "tmp31.i", label_forbody_i13); - std::vector ptr_arrayidx34_i10_indices; - ptr_arrayidx34_i10_indices.push_back(const_int32_34); - ptr_arrayidx34_i10_indices.push_back(int32_i_0_reg2mem_0_i3); - Instruction* ptr_arrayidx34_i10 = new GetElementPtrInst(ptr_consts_141, ptr_arrayidx34_i10_indices.begin(), ptr_arrayidx34_i10_indices.end(), "arrayidx34.i10", label_forbody_i13); - StoreInst* void_180 = new StoreInst(packed_tmp31_i_179, ptr_arrayidx34_i10, false, label_forbody_i13); - BinaryOperator* int32_indvar_next23 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i3, const_int32_36, "indvar.next23", label_forbody_i13); - ICmpInst* int1_exitcond24 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next23, int32_num_consts_133, "exitcond24", label_forbody_i13); - new BranchInst(label_from_consts_exit_137, label_forbody_i13, int1_exitcond24, label_forbody_i13); - - // Block from_consts.exit (label_from_consts_exit_137) - std::vector ptr_tmp6_182_indices; - ptr_tmp6_182_indices.push_back(const_int32_34); - ptr_tmp6_182_indices.push_back(const_int32_38); - Instruction* ptr_tmp6_182 = new GetElementPtrInst(ptr_args_144, ptr_tmp6_182_indices.begin(), ptr_tmp6_182_indices.end(), "tmp6", label_from_consts_exit_137); - std::vector ptr_arraydecay7_183_indices; - ptr_arraydecay7_183_indices.push_back(const_int32_34); - ptr_arraydecay7_183_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay7_183 = new GetElementPtrInst(ptr_consts_141, ptr_arraydecay7_183_indices.begin(), ptr_arraydecay7_183_indices.end(), "arraydecay7", label_from_consts_exit_137); - StoreInst* void_184 = new StoreInst(ptr_arraydecay7_183, ptr_tmp6_182, false, label_from_consts_exit_137); - std::vector ptr_tmp8_185_indices; - ptr_tmp8_185_indices.push_back(const_int32_34); - ptr_tmp8_185_indices.push_back(const_int32_37); - Instruction* ptr_tmp8_185 = new GetElementPtrInst(ptr_args_144, ptr_tmp8_185_indices.begin(), ptr_tmp8_185_indices.end(), "tmp8", label_from_consts_exit_137); - std::vector ptr_arraydecay9_186_indices; - ptr_arraydecay9_186_indices.push_back(const_int32_34); - ptr_arraydecay9_186_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay9_186 = new GetElementPtrInst(ptr_temps_143, ptr_arraydecay9_186_indices.begin(), ptr_arraydecay9_186_indices.end(), "arraydecay9", label_from_consts_exit_137); - StoreInst* void_187 = new StoreInst(ptr_arraydecay9_186, ptr_tmp8_185, false, label_from_consts_exit_137); - std::vector ptr_tmp11_indices; - ptr_tmp11_indices.push_back(const_int32_34); - ptr_tmp11_indices.push_back(const_int32_36); - Instruction* ptr_tmp11 = new GetElementPtrInst(ptr_args_144, ptr_tmp11_indices.begin(), ptr_tmp11_indices.end(), "tmp11", label_from_consts_exit_137); - std::vector ptr_tmp15_indices; - ptr_tmp15_indices.push_back(const_int32_34); - ptr_tmp15_indices.push_back(const_int32_34); - Instruction* ptr_tmp15 = new GetElementPtrInst(ptr_args_144, ptr_tmp15_indices.begin(), ptr_tmp15_indices.end(), "tmp15", label_from_consts_exit_137); - new BranchInst(label_forbody_138, label_from_consts_exit_137); - - // Block forbody (label_forbody_138) - Argument* fwdref_189 = new Argument(IntegerType::get(32)); - PHINode* int32_tmp21_rle = new PHINode(IntegerType::get(32), "tmp21.rle", label_forbody_138); - int32_tmp21_rle->reserveOperandSpace(2); - int32_tmp21_rle->addIncoming(const_int32_34, label_from_consts_exit_137); - int32_tmp21_rle->addIncoming(fwdref_189, label_forbody_138); - - Argument* fwdref_191 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_190 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_138); - int32_i_0_reg2mem_0_190->reserveOperandSpace(2); - int32_i_0_reg2mem_0_190->addIncoming(const_int32_34, label_from_consts_exit_137); - int32_i_0_reg2mem_0_190->addIncoming(fwdref_191, label_forbody_138); - - std::vector ptr_arraydecay14_indices; - ptr_arraydecay14_indices.push_back(const_int32_34); - ptr_arraydecay14_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arraydecay14_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay14 = new GetElementPtrInst(ptr_inputs_140, ptr_arraydecay14_indices.begin(), ptr_arraydecay14_indices.end(), "arraydecay14", label_forbody_138); - StoreInst* void_192 = new StoreInst(ptr_arraydecay14, ptr_tmp11, false, label_forbody_138); - std::vector ptr_arraydecay19_indices; - ptr_arraydecay19_indices.push_back(const_int32_34); - ptr_arraydecay19_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arraydecay19_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay19 = new GetElementPtrInst(ptr_results_142, ptr_arraydecay19_indices.begin(), ptr_arraydecay19_indices.end(), "arraydecay19", label_forbody_138); - StoreInst* void_193 = new StoreInst(ptr_arraydecay19, ptr_tmp15, false, label_forbody_138); - StoreInst* void_194 = new StoreInst(const_int32_34, ptr_tmp, false, label_forbody_138); - CallInst* void_195 = new CallInst(func_execute_shader, ptr_args_144, "", label_forbody_138); - void_195->setCallingConv(CallingConv::C); - void_195->setTailCall(false); - LoadInst* int32_tmp26 = new LoadInst(ptr_tmp, "tmp26", false, label_forbody_138); - BinaryOperator* int32_shl = BinaryOperator::create(Instruction::Shl, int32_tmp26, int32_i_0_reg2mem_0_190, "shl", label_forbody_138); - BinaryOperator* int32_or = BinaryOperator::create(Instruction::Or, int32_shl, int32_tmp21_rle, "or", label_forbody_138); - StoreInst* void_196 = new StoreInst(int32_or, ptr_tmp, false, label_forbody_138); - LoadInst* ptr_tmp33 = new LoadInst(ptr_tmp15, "tmp33", false, label_forbody_138); - std::vector ptr_arraydecay_i_indices; - ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arraydecay_i_indices.push_back(const_int32_34); - ptr_arraydecay_i_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_dests_129, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_138); - LoadInst* packed_tmp7_i = new LoadInst(ptr_tmp33, "tmp7.i", false, label_forbody_138); - ExtractElementInst* float_tmp11_i = new ExtractElementInst(packed_tmp7_i, const_int32_34, "tmp11.i", label_forbody_138); - StoreInst* void_197 = new StoreInst(float_tmp11_i, ptr_arraydecay_i, false, label_forbody_138); - std::vector ptr_arrayidx13_i_indices; - ptr_arrayidx13_i_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arrayidx13_i_indices.push_back(const_int32_34); - ptr_arrayidx13_i_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx13_i = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx13_i_indices.begin(), ptr_arrayidx13_i_indices.end(), "arrayidx13.i", label_forbody_138); - ExtractElementInst* float_tmp15_i2 = new ExtractElementInst(packed_tmp7_i, const_int32_36, "tmp15.i2", label_forbody_138); - StoreInst* void_198 = new StoreInst(float_tmp15_i2, ptr_arrayidx13_i, false, label_forbody_138); - std::vector ptr_arrayidx17_i_indices; - ptr_arrayidx17_i_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arrayidx17_i_indices.push_back(const_int32_34); - ptr_arrayidx17_i_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx17_i = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx17_i_indices.begin(), ptr_arrayidx17_i_indices.end(), "arrayidx17.i", label_forbody_138); - ExtractElementInst* float_tmp19_i = new ExtractElementInst(packed_tmp7_i, const_int32_37, "tmp19.i", label_forbody_138); - StoreInst* void_199 = new StoreInst(float_tmp19_i, ptr_arrayidx17_i, false, label_forbody_138); - std::vector ptr_arrayidx21_i_indices; - ptr_arrayidx21_i_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arrayidx21_i_indices.push_back(const_int32_34); - ptr_arrayidx21_i_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx21_i = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx21_i_indices.begin(), ptr_arrayidx21_i_indices.end(), "arrayidx21.i", label_forbody_138); - ExtractElementInst* float_tmp23_i = new ExtractElementInst(packed_tmp7_i, const_int32_38, "tmp23.i", label_forbody_138); - StoreInst* void_200 = new StoreInst(float_tmp23_i, ptr_arrayidx21_i, false, label_forbody_138); - std::vector ptr_arraydecay_i_1_indices; - ptr_arraydecay_i_1_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arraydecay_i_1_indices.push_back(const_int32_36); - ptr_arraydecay_i_1_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arraydecay_i_1_indices.begin(), ptr_arraydecay_i_1_indices.end(), "arraydecay.i.1", label_forbody_138); - GetElementPtrInst* ptr_arrayidx6_i_1 = new GetElementPtrInst(ptr_tmp33, const_int32_36, "arrayidx6.i.1", label_forbody_138); - LoadInst* packed_tmp7_i_1 = new LoadInst(ptr_arrayidx6_i_1, "tmp7.i.1", false, label_forbody_138); - ExtractElementInst* float_tmp11_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_34, "tmp11.i.1", label_forbody_138); - StoreInst* void_201 = new StoreInst(float_tmp11_i_1, ptr_arraydecay_i_1, false, label_forbody_138); - std::vector ptr_arrayidx13_i_1_indices; - ptr_arrayidx13_i_1_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arrayidx13_i_1_indices.push_back(const_int32_36); - ptr_arrayidx13_i_1_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx13_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx13_i_1_indices.begin(), ptr_arrayidx13_i_1_indices.end(), "arrayidx13.i.1", label_forbody_138); - ExtractElementInst* float_tmp15_i2_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_36, "tmp15.i2.1", label_forbody_138); - StoreInst* void_202 = new StoreInst(float_tmp15_i2_1, ptr_arrayidx13_i_1, false, label_forbody_138); - std::vector ptr_arrayidx17_i_1_indices; - ptr_arrayidx17_i_1_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arrayidx17_i_1_indices.push_back(const_int32_36); - ptr_arrayidx17_i_1_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx17_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx17_i_1_indices.begin(), ptr_arrayidx17_i_1_indices.end(), "arrayidx17.i.1", label_forbody_138); - ExtractElementInst* float_tmp19_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_37, "tmp19.i.1", label_forbody_138); - StoreInst* void_203 = new StoreInst(float_tmp19_i_1, ptr_arrayidx17_i_1, false, label_forbody_138); - std::vector ptr_arrayidx21_i_1_indices; - ptr_arrayidx21_i_1_indices.push_back(int32_i_0_reg2mem_0_190); - ptr_arrayidx21_i_1_indices.push_back(const_int32_36); - ptr_arrayidx21_i_1_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx21_i_1 = new GetElementPtrInst(ptr_dests_129, ptr_arrayidx21_i_1_indices.begin(), ptr_arrayidx21_i_1_indices.end(), "arrayidx21.i.1", label_forbody_138); - ExtractElementInst* float_tmp23_i_1 = new ExtractElementInst(packed_tmp7_i_1, const_int32_38, "tmp23.i.1", label_forbody_138); - StoreInst* void_204 = new StoreInst(float_tmp23_i_1, ptr_arrayidx21_i_1, false, label_forbody_138); - BinaryOperator* int32_indvar_next21 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_190, const_int32_36, "indvar.next21", label_forbody_138); - ICmpInst* int1_exitcond22 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next21, const_int32_39, "exitcond22", label_forbody_138); - new BranchInst(label_afterfor_139, label_forbody_138, int1_exitcond22, label_forbody_138); - - // Block afterfor (label_afterfor_139) - BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_or, const_int32_40, "neg", label_afterfor_139); - new ReturnInst(int32_neg, label_afterfor_139); - - // Block forbody6.i.1 (label_forbody6_i_1) - Argument* fwdref_207 = new Argument(IntegerType::get(32)); - PHINode* int32_j_0_reg2mem_0_i_1 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.1", label_forbody6_i_1); - int32_j_0_reg2mem_0_i_1->reserveOperandSpace(2); - int32_j_0_reg2mem_0_i_1->addIncoming(const_int32_34, label_forbody6_i_135); - int32_j_0_reg2mem_0_i_1->addIncoming(fwdref_207, label_forbody6_i_1); - - Argument* fwdref_208 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i_1 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i.1", label_forbody6_i_1); - packed_vec_0_reg2mem_0_i_1->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_1->addIncoming(packed_tmp48_i_163, label_forbody6_i_135); - packed_vec_0_reg2mem_0_i_1->addIncoming(fwdref_208, label_forbody6_i_1); - - std::vector ptr_arraydecay11_i_1_indices; - ptr_arraydecay11_i_1_indices.push_back(const_int32_36); - ptr_arraydecay11_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); - ptr_arraydecay11_i_1_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay11_i_1 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_1_indices.begin(), ptr_arraydecay11_i_1_indices.end(), "arraydecay11.i.1", label_forbody6_i_1); - LoadInst* float_tmp13_i_1 = new LoadInst(ptr_arraydecay11_i_1, "tmp13.i.1", false, label_forbody6_i_1); - InsertElementInst* packed_tmp15_i_1 = new InsertElementInst(packed_vec_0_reg2mem_0_i_1, float_tmp13_i_1, const_int32_34, "tmp15.i.1", label_forbody6_i_1); - std::vector ptr_arrayidx23_i_1_indices; - ptr_arrayidx23_i_1_indices.push_back(const_int32_36); - ptr_arrayidx23_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); - ptr_arrayidx23_i_1_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx23_i_1 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_1_indices.begin(), ptr_arrayidx23_i_1_indices.end(), "arrayidx23.i.1", label_forbody6_i_1); - LoadInst* float_tmp24_i_1 = new LoadInst(ptr_arrayidx23_i_1, "tmp24.i.1", false, label_forbody6_i_1); - InsertElementInst* packed_tmp26_i_1 = new InsertElementInst(packed_tmp15_i_1, float_tmp24_i_1, const_int32_36, "tmp26.i.1", label_forbody6_i_1); - std::vector ptr_arrayidx34_i_1_indices; - ptr_arrayidx34_i_1_indices.push_back(const_int32_36); - ptr_arrayidx34_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); - ptr_arrayidx34_i_1_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx34_i_1 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_1_indices.begin(), ptr_arrayidx34_i_1_indices.end(), "arrayidx34.i.1", label_forbody6_i_1); - LoadInst* float_tmp35_i_1 = new LoadInst(ptr_arrayidx34_i_1, "tmp35.i.1", false, label_forbody6_i_1); - InsertElementInst* packed_tmp37_i_1 = new InsertElementInst(packed_tmp26_i_1, float_tmp35_i_1, const_int32_37, "tmp37.i.1", label_forbody6_i_1); - std::vector ptr_arrayidx45_i_1_indices; - ptr_arrayidx45_i_1_indices.push_back(const_int32_36); - ptr_arrayidx45_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); - ptr_arrayidx45_i_1_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx45_i_1 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_1_indices.begin(), ptr_arrayidx45_i_1_indices.end(), "arrayidx45.i.1", label_forbody6_i_1); - LoadInst* float_tmp46_i_1 = new LoadInst(ptr_arrayidx45_i_1, "tmp46.i.1", false, label_forbody6_i_1); - InsertElementInst* packed_tmp48_i_1 = new InsertElementInst(packed_tmp37_i_1, float_tmp46_i_1, const_int32_38, "tmp48.i.1", label_forbody6_i_1); - std::vector ptr_arrayidx54_i_1_indices; - ptr_arrayidx54_i_1_indices.push_back(const_int32_34); - ptr_arrayidx54_i_1_indices.push_back(const_int32_36); - ptr_arrayidx54_i_1_indices.push_back(int32_j_0_reg2mem_0_i_1); - Instruction* ptr_arrayidx54_i_1 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_1_indices.begin(), ptr_arrayidx54_i_1_indices.end(), "arrayidx54.i.1", label_forbody6_i_1); - StoreInst* void_209 = new StoreInst(packed_tmp48_i_1, ptr_arrayidx54_i_1, false, label_forbody6_i_1); - BinaryOperator* int32_inc_i_1 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_1, const_int32_36, "inc.i.1", label_forbody6_i_1); - ICmpInst* int1_cmp59_i_1 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_1, int32_num_inputs_131, "cmp59.i.1", label_forbody6_i_1); - new BranchInst(label_forbody6_i_1, label_forbody6_i_2, int1_cmp59_i_1, label_forbody6_i_1); - - // Block forbody6.i.2 (label_forbody6_i_2) - Argument* fwdref_211 = new Argument(IntegerType::get(32)); - PHINode* int32_j_0_reg2mem_0_i_2 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.2", label_forbody6_i_2); - int32_j_0_reg2mem_0_i_2->reserveOperandSpace(2); - int32_j_0_reg2mem_0_i_2->addIncoming(const_int32_34, label_forbody6_i_1); - int32_j_0_reg2mem_0_i_2->addIncoming(fwdref_211, label_forbody6_i_2); - - Argument* fwdref_212 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i_2 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i.2", label_forbody6_i_2); - packed_vec_0_reg2mem_0_i_2->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_2->addIncoming(packed_tmp48_i_1, label_forbody6_i_1); - packed_vec_0_reg2mem_0_i_2->addIncoming(fwdref_212, label_forbody6_i_2); - - std::vector ptr_arraydecay11_i_2_indices; - ptr_arraydecay11_i_2_indices.push_back(const_int32_37); - ptr_arraydecay11_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - ptr_arraydecay11_i_2_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay11_i_2 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_2_indices.begin(), ptr_arraydecay11_i_2_indices.end(), "arraydecay11.i.2", label_forbody6_i_2); - LoadInst* float_tmp13_i_2 = new LoadInst(ptr_arraydecay11_i_2, "tmp13.i.2", false, label_forbody6_i_2); - InsertElementInst* packed_tmp15_i_2 = new InsertElementInst(packed_vec_0_reg2mem_0_i_2, float_tmp13_i_2, const_int32_34, "tmp15.i.2", label_forbody6_i_2); - std::vector ptr_arrayidx23_i_2_indices; - ptr_arrayidx23_i_2_indices.push_back(const_int32_37); - ptr_arrayidx23_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - ptr_arrayidx23_i_2_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx23_i_2 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_2_indices.begin(), ptr_arrayidx23_i_2_indices.end(), "arrayidx23.i.2", label_forbody6_i_2); - LoadInst* float_tmp24_i_2 = new LoadInst(ptr_arrayidx23_i_2, "tmp24.i.2", false, label_forbody6_i_2); - InsertElementInst* packed_tmp26_i_2 = new InsertElementInst(packed_tmp15_i_2, float_tmp24_i_2, const_int32_36, "tmp26.i.2", label_forbody6_i_2); - std::vector ptr_arrayidx34_i_2_indices; - ptr_arrayidx34_i_2_indices.push_back(const_int32_37); - ptr_arrayidx34_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - ptr_arrayidx34_i_2_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx34_i_2 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_2_indices.begin(), ptr_arrayidx34_i_2_indices.end(), "arrayidx34.i.2", label_forbody6_i_2); - LoadInst* float_tmp35_i_2 = new LoadInst(ptr_arrayidx34_i_2, "tmp35.i.2", false, label_forbody6_i_2); - InsertElementInst* packed_tmp37_i_2 = new InsertElementInst(packed_tmp26_i_2, float_tmp35_i_2, const_int32_37, "tmp37.i.2", label_forbody6_i_2); - std::vector ptr_arrayidx45_i_2_indices; - ptr_arrayidx45_i_2_indices.push_back(const_int32_37); - ptr_arrayidx45_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - ptr_arrayidx45_i_2_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx45_i_2 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_2_indices.begin(), ptr_arrayidx45_i_2_indices.end(), "arrayidx45.i.2", label_forbody6_i_2); - LoadInst* float_tmp46_i_2 = new LoadInst(ptr_arrayidx45_i_2, "tmp46.i.2", false, label_forbody6_i_2); - InsertElementInst* packed_tmp48_i_2 = new InsertElementInst(packed_tmp37_i_2, float_tmp46_i_2, const_int32_38, "tmp48.i.2", label_forbody6_i_2); - std::vector ptr_arrayidx54_i_2_indices; - ptr_arrayidx54_i_2_indices.push_back(const_int32_34); - ptr_arrayidx54_i_2_indices.push_back(const_int32_37); - ptr_arrayidx54_i_2_indices.push_back(int32_j_0_reg2mem_0_i_2); - Instruction* ptr_arrayidx54_i_2 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_2_indices.begin(), ptr_arrayidx54_i_2_indices.end(), "arrayidx54.i.2", label_forbody6_i_2); - StoreInst* void_213 = new StoreInst(packed_tmp48_i_2, ptr_arrayidx54_i_2, false, label_forbody6_i_2); - BinaryOperator* int32_inc_i_2 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_2, const_int32_36, "inc.i.2", label_forbody6_i_2); - ICmpInst* int1_cmp59_i_2 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_2, int32_num_inputs_131, "cmp59.i.2", label_forbody6_i_2); - new BranchInst(label_forbody6_i_2, label_forbody6_i_3, int1_cmp59_i_2, label_forbody6_i_2); - - // Block forbody6.i.3 (label_forbody6_i_3) - Argument* fwdref_215 = new Argument(IntegerType::get(32)); - PHINode* int32_j_0_reg2mem_0_i_3 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0.i.3", label_forbody6_i_3); - int32_j_0_reg2mem_0_i_3->reserveOperandSpace(2); - int32_j_0_reg2mem_0_i_3->addIncoming(const_int32_34, label_forbody6_i_2); - int32_j_0_reg2mem_0_i_3->addIncoming(fwdref_215, label_forbody6_i_3); - - Argument* fwdref_216 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i_3 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i.3", label_forbody6_i_3); - packed_vec_0_reg2mem_0_i_3->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_3->addIncoming(packed_tmp48_i_2, label_forbody6_i_2); - packed_vec_0_reg2mem_0_i_3->addIncoming(fwdref_216, label_forbody6_i_3); - - std::vector ptr_arraydecay11_i_3_indices; - ptr_arraydecay11_i_3_indices.push_back(const_int32_38); - ptr_arraydecay11_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - ptr_arraydecay11_i_3_indices.push_back(const_int32_34); - Instruction* ptr_arraydecay11_i_3 = new GetElementPtrInst(ptr_ainputs_130, ptr_arraydecay11_i_3_indices.begin(), ptr_arraydecay11_i_3_indices.end(), "arraydecay11.i.3", label_forbody6_i_3); - LoadInst* float_tmp13_i_3 = new LoadInst(ptr_arraydecay11_i_3, "tmp13.i.3", false, label_forbody6_i_3); - InsertElementInst* packed_tmp15_i_3 = new InsertElementInst(packed_vec_0_reg2mem_0_i_3, float_tmp13_i_3, const_int32_34, "tmp15.i.3", label_forbody6_i_3); - std::vector ptr_arrayidx23_i_3_indices; - ptr_arrayidx23_i_3_indices.push_back(const_int32_38); - ptr_arrayidx23_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - ptr_arrayidx23_i_3_indices.push_back(const_int32_36); - Instruction* ptr_arrayidx23_i_3 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx23_i_3_indices.begin(), ptr_arrayidx23_i_3_indices.end(), "arrayidx23.i.3", label_forbody6_i_3); - LoadInst* float_tmp24_i_3 = new LoadInst(ptr_arrayidx23_i_3, "tmp24.i.3", false, label_forbody6_i_3); - InsertElementInst* packed_tmp26_i_3 = new InsertElementInst(packed_tmp15_i_3, float_tmp24_i_3, const_int32_36, "tmp26.i.3", label_forbody6_i_3); - std::vector ptr_arrayidx34_i_3_indices; - ptr_arrayidx34_i_3_indices.push_back(const_int32_38); - ptr_arrayidx34_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - ptr_arrayidx34_i_3_indices.push_back(const_int32_37); - Instruction* ptr_arrayidx34_i_3 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx34_i_3_indices.begin(), ptr_arrayidx34_i_3_indices.end(), "arrayidx34.i.3", label_forbody6_i_3); - LoadInst* float_tmp35_i_3 = new LoadInst(ptr_arrayidx34_i_3, "tmp35.i.3", false, label_forbody6_i_3); - InsertElementInst* packed_tmp37_i_3 = new InsertElementInst(packed_tmp26_i_3, float_tmp35_i_3, const_int32_37, "tmp37.i.3", label_forbody6_i_3); - std::vector ptr_arrayidx45_i_3_indices; - ptr_arrayidx45_i_3_indices.push_back(const_int32_38); - ptr_arrayidx45_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - ptr_arrayidx45_i_3_indices.push_back(const_int32_38); - Instruction* ptr_arrayidx45_i_3 = new GetElementPtrInst(ptr_ainputs_130, ptr_arrayidx45_i_3_indices.begin(), ptr_arrayidx45_i_3_indices.end(), "arrayidx45.i.3", label_forbody6_i_3); - LoadInst* float_tmp46_i_3 = new LoadInst(ptr_arrayidx45_i_3, "tmp46.i.3", false, label_forbody6_i_3); - InsertElementInst* packed_tmp48_i_3 = new InsertElementInst(packed_tmp37_i_3, float_tmp46_i_3, const_int32_38, "tmp48.i.3", label_forbody6_i_3); - std::vector ptr_arrayidx54_i_3_indices; - ptr_arrayidx54_i_3_indices.push_back(const_int32_34); - ptr_arrayidx54_i_3_indices.push_back(const_int32_38); - ptr_arrayidx54_i_3_indices.push_back(int32_j_0_reg2mem_0_i_3); - Instruction* ptr_arrayidx54_i_3 = new GetElementPtrInst(ptr_inputs_140, ptr_arrayidx54_i_3_indices.begin(), ptr_arrayidx54_i_3_indices.end(), "arrayidx54.i.3", label_forbody6_i_3); - StoreInst* void_217 = new StoreInst(packed_tmp48_i_3, ptr_arrayidx54_i_3, false, label_forbody6_i_3); - BinaryOperator* int32_inc_i_3 = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0_i_3, const_int32_36, "inc.i.3", label_forbody6_i_3); - ICmpInst* int1_cmp59_i_3 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc_i_3, int32_num_inputs_131, "cmp59.i.3", label_forbody6_i_3); - new BranchInst(label_forbody6_i_3, label_from_array_exit_136, int1_cmp59_i_3, label_forbody6_i_3); + ptr_tmp_indices.push_back(const_int32_30); + ptr_tmp_indices.push_back(const_int32_35); + Instruction* ptr_tmp = new GetElementPtrInst(ptr_args_114, ptr_tmp_indices.begin(), ptr_tmp_indices.end(), "tmp", label_entry_109); + StoreInst* void_115 = new StoreInst(const_int32_30, ptr_tmp, false, label_entry_109); + ICmpInst* int1_cmp_i_116 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_108, const_int32_30, "cmp.i", label_entry_109); + new BranchInst(label_forbody_i_110, label_from_consts_exit_111, int1_cmp_i_116, label_entry_109); + + // Block forbody.i (label_forbody_i_110) + Argument* fwdref_119 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_i_118 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i_110); + int32_i_0_reg2mem_0_i_118->reserveOperandSpace(2); + int32_i_0_reg2mem_0_i_118->addIncoming(const_int32_30, label_entry_109); + int32_i_0_reg2mem_0_i_118->addIncoming(fwdref_119, label_forbody_i_110); + + Argument* fwdref_121 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i_120 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i_110); + packed_vec_0_reg2mem_0_i_120->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_120->addIncoming(const_packed_31, label_entry_109); + packed_vec_0_reg2mem_0_i_120->addIncoming(fwdref_121, label_forbody_i_110); + + std::vector ptr_arraydecay_i_122_indices; + ptr_arraydecay_i_122_indices.push_back(int32_i_0_reg2mem_0_i_118); + ptr_arraydecay_i_122_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay_i_122 = new GetElementPtrInst(ptr_aconsts_107, ptr_arraydecay_i_122_indices.begin(), ptr_arraydecay_i_122_indices.end(), "arraydecay.i", label_forbody_i_110); + LoadInst* float_tmp5_i_123 = new LoadInst(ptr_arraydecay_i_122, "tmp5.i", false, label_forbody_i_110); + InsertElementInst* packed_tmp7_i_124 = new InsertElementInst(packed_vec_0_reg2mem_0_i_120, float_tmp5_i_123, const_int32_30, "tmp7.i", label_forbody_i_110); + std::vector ptr_arrayidx12_i_125_indices; + ptr_arrayidx12_i_125_indices.push_back(int32_i_0_reg2mem_0_i_118); + ptr_arrayidx12_i_125_indices.push_back(const_int32_32); + Instruction* ptr_arrayidx12_i_125 = new GetElementPtrInst(ptr_aconsts_107, ptr_arrayidx12_i_125_indices.begin(), ptr_arrayidx12_i_125_indices.end(), "arrayidx12.i", label_forbody_i_110); + LoadInst* float_tmp13_i_126 = new LoadInst(ptr_arrayidx12_i_125, "tmp13.i", false, label_forbody_i_110); + InsertElementInst* packed_tmp15_i_127 = new InsertElementInst(packed_tmp7_i_124, float_tmp13_i_126, const_int32_32, "tmp15.i", label_forbody_i_110); + std::vector ptr_arrayidx20_i_128_indices; + ptr_arrayidx20_i_128_indices.push_back(int32_i_0_reg2mem_0_i_118); + ptr_arrayidx20_i_128_indices.push_back(const_int32_33); + Instruction* ptr_arrayidx20_i_128 = new GetElementPtrInst(ptr_aconsts_107, ptr_arrayidx20_i_128_indices.begin(), ptr_arrayidx20_i_128_indices.end(), "arrayidx20.i", label_forbody_i_110); + LoadInst* float_tmp21_i_129 = new LoadInst(ptr_arrayidx20_i_128, "tmp21.i", false, label_forbody_i_110); + InsertElementInst* packed_tmp23_i_130 = new InsertElementInst(packed_tmp15_i_127, float_tmp21_i_129, const_int32_33, "tmp23.i", label_forbody_i_110); + std::vector ptr_arrayidx28_i_131_indices; + ptr_arrayidx28_i_131_indices.push_back(int32_i_0_reg2mem_0_i_118); + ptr_arrayidx28_i_131_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx28_i_131 = new GetElementPtrInst(ptr_aconsts_107, ptr_arrayidx28_i_131_indices.begin(), ptr_arrayidx28_i_131_indices.end(), "arrayidx28.i", label_forbody_i_110); + LoadInst* float_tmp29_i_132 = new LoadInst(ptr_arrayidx28_i_131, "tmp29.i", false, label_forbody_i_110); + InsertElementInst* packed_tmp31_i_133 = new InsertElementInst(packed_tmp23_i_130, float_tmp29_i_132, const_int32_34, "tmp31.i", label_forbody_i_110); + std::vector ptr_arrayidx34_i_134_indices; + ptr_arrayidx34_i_134_indices.push_back(const_int32_30); + ptr_arrayidx34_i_134_indices.push_back(int32_i_0_reg2mem_0_i_118); + Instruction* ptr_arrayidx34_i_134 = new GetElementPtrInst(ptr_consts_112, ptr_arrayidx34_i_134_indices.begin(), ptr_arrayidx34_i_134_indices.end(), "arrayidx34.i", label_forbody_i_110); + StoreInst* void_135 = new StoreInst(packed_tmp31_i_133, ptr_arrayidx34_i_134, false, label_forbody_i_110); + BinaryOperator* int32_indvar_next7 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_118, const_int32_32, "indvar.next7", label_forbody_i_110); + ICmpInst* int1_exitcond8 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next7, int32_num_consts_108, "exitcond8", label_forbody_i_110); + new BranchInst(label_from_consts_exit_111, label_forbody_i_110, int1_exitcond8, label_forbody_i_110); + + // Block from_consts.exit (label_from_consts_exit_111) + std::vector ptr_tmp3_indices; + ptr_tmp3_indices.push_back(const_int32_30); + ptr_tmp3_indices.push_back(const_int32_34); + Instruction* ptr_tmp3 = new GetElementPtrInst(ptr_args_114, ptr_tmp3_indices.begin(), ptr_tmp3_indices.end(), "tmp3", label_from_consts_exit_111); + std::vector ptr_arraydecay4_indices; + ptr_arraydecay4_indices.push_back(const_int32_30); + ptr_arraydecay4_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay4 = new GetElementPtrInst(ptr_consts_112, ptr_arraydecay4_indices.begin(), ptr_arraydecay4_indices.end(), "arraydecay4", label_from_consts_exit_111); + StoreInst* void_137 = new StoreInst(ptr_arraydecay4, ptr_tmp3, false, label_from_consts_exit_111); + std::vector ptr_tmp5_indices; + ptr_tmp5_indices.push_back(const_int32_30); + ptr_tmp5_indices.push_back(const_int32_33); + Instruction* ptr_tmp5 = new GetElementPtrInst(ptr_args_114, ptr_tmp5_indices.begin(), ptr_tmp5_indices.end(), "tmp5", label_from_consts_exit_111); + std::vector ptr_arraydecay6_indices; + ptr_arraydecay6_indices.push_back(const_int32_30); + ptr_arraydecay6_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay6 = new GetElementPtrInst(ptr_temps_113, ptr_arraydecay6_indices.begin(), ptr_arraydecay6_indices.end(), "arraydecay6", label_from_consts_exit_111); + StoreInst* void_138 = new StoreInst(ptr_arraydecay6, ptr_tmp5, false, label_from_consts_exit_111); + std::vector ptr_tmp8_139_indices; + ptr_tmp8_139_indices.push_back(const_int32_30); + ptr_tmp8_139_indices.push_back(const_int32_32); + Instruction* ptr_tmp8_139 = new GetElementPtrInst(ptr_args_114, ptr_tmp8_139_indices.begin(), ptr_tmp8_139_indices.end(), "tmp8", label_from_consts_exit_111); + std::vector ptr_tmp12_140_indices; + ptr_tmp12_140_indices.push_back(const_int32_30); + ptr_tmp12_140_indices.push_back(const_int32_30); + Instruction* ptr_tmp12_140 = new GetElementPtrInst(ptr_args_114, ptr_tmp12_140_indices.begin(), ptr_tmp12_140_indices.end(), "tmp12", label_from_consts_exit_111); + std::vector ptr_arraydecay11_141_indices; + ptr_arraydecay11_141_indices.push_back(const_int32_30); + ptr_arraydecay11_141_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay11_141 = new GetElementPtrInst(ptr_inputs_105, ptr_arraydecay11_141_indices.begin(), ptr_arraydecay11_141_indices.end(), "arraydecay11", label_from_consts_exit_111); + StoreInst* void_142 = new StoreInst(ptr_arraydecay11_141, ptr_tmp8_139, false, label_from_consts_exit_111); + std::vector ptr_arraydecay16_143_indices; + ptr_arraydecay16_143_indices.push_back(const_int32_30); + ptr_arraydecay16_143_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay16_143 = new GetElementPtrInst(ptr_results_104, ptr_arraydecay16_143_indices.begin(), ptr_arraydecay16_143_indices.end(), "arraydecay16", label_from_consts_exit_111); + StoreInst* void_144 = new StoreInst(ptr_arraydecay16_143, ptr_tmp12_140, false, label_from_consts_exit_111); + StoreInst* void_145 = new StoreInst(const_int32_30, ptr_tmp, false, label_from_consts_exit_111); + CallInst* void_146 = new CallInst(func_execute_shader, ptr_args_114, "", label_from_consts_exit_111); + void_146->setCallingConv(CallingConv::C); + void_146->setTailCall(false); + LoadInst* int32_tmp23 = new LoadInst(ptr_tmp, "tmp23", false, label_from_consts_exit_111); + std::vector ptr_arraydecay11_1_indices; + ptr_arraydecay11_1_indices.push_back(const_int32_32); + ptr_arraydecay11_1_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay11_1 = new GetElementPtrInst(ptr_inputs_105, ptr_arraydecay11_1_indices.begin(), ptr_arraydecay11_1_indices.end(), "arraydecay11.1", label_from_consts_exit_111); + StoreInst* void_147 = new StoreInst(ptr_arraydecay11_1, ptr_tmp8_139, false, label_from_consts_exit_111); + std::vector ptr_arraydecay16_1_indices; + ptr_arraydecay16_1_indices.push_back(const_int32_32); + ptr_arraydecay16_1_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay16_1 = new GetElementPtrInst(ptr_results_104, ptr_arraydecay16_1_indices.begin(), ptr_arraydecay16_1_indices.end(), "arraydecay16.1", label_from_consts_exit_111); + StoreInst* void_148 = new StoreInst(ptr_arraydecay16_1, ptr_tmp12_140, false, label_from_consts_exit_111); + StoreInst* void_149 = new StoreInst(const_int32_30, ptr_tmp, false, label_from_consts_exit_111); + CallInst* void_150 = new CallInst(func_execute_shader, ptr_args_114, "", label_from_consts_exit_111); + void_150->setCallingConv(CallingConv::C); + void_150->setTailCall(false); + LoadInst* int32_tmp23_1 = new LoadInst(ptr_tmp, "tmp23.1", false, label_from_consts_exit_111); + BinaryOperator* int32_shl_1 = BinaryOperator::create(Instruction::Shl, int32_tmp23_1, const_int32_32, "shl.1", label_from_consts_exit_111); + BinaryOperator* int32_or_1 = BinaryOperator::create(Instruction::Or, int32_shl_1, int32_tmp23, "or.1", label_from_consts_exit_111); + std::vector ptr_arraydecay11_2_indices; + ptr_arraydecay11_2_indices.push_back(const_int32_33); + ptr_arraydecay11_2_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay11_2 = new GetElementPtrInst(ptr_inputs_105, ptr_arraydecay11_2_indices.begin(), ptr_arraydecay11_2_indices.end(), "arraydecay11.2", label_from_consts_exit_111); + StoreInst* void_151 = new StoreInst(ptr_arraydecay11_2, ptr_tmp8_139, false, label_from_consts_exit_111); + std::vector ptr_arraydecay16_2_indices; + ptr_arraydecay16_2_indices.push_back(const_int32_33); + ptr_arraydecay16_2_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay16_2 = new GetElementPtrInst(ptr_results_104, ptr_arraydecay16_2_indices.begin(), ptr_arraydecay16_2_indices.end(), "arraydecay16.2", label_from_consts_exit_111); + StoreInst* void_152 = new StoreInst(ptr_arraydecay16_2, ptr_tmp12_140, false, label_from_consts_exit_111); + StoreInst* void_153 = new StoreInst(const_int32_30, ptr_tmp, false, label_from_consts_exit_111); + CallInst* void_154 = new CallInst(func_execute_shader, ptr_args_114, "", label_from_consts_exit_111); + void_154->setCallingConv(CallingConv::C); + void_154->setTailCall(false); + LoadInst* int32_tmp23_2 = new LoadInst(ptr_tmp, "tmp23.2", false, label_from_consts_exit_111); + BinaryOperator* int32_shl_2 = BinaryOperator::create(Instruction::Shl, int32_tmp23_2, const_int32_33, "shl.2", label_from_consts_exit_111); + BinaryOperator* int32_or_2 = BinaryOperator::create(Instruction::Or, int32_shl_2, int32_or_1, "or.2", label_from_consts_exit_111); + std::vector ptr_arraydecay11_3_indices; + ptr_arraydecay11_3_indices.push_back(const_int32_34); + ptr_arraydecay11_3_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay11_3 = new GetElementPtrInst(ptr_inputs_105, ptr_arraydecay11_3_indices.begin(), ptr_arraydecay11_3_indices.end(), "arraydecay11.3", label_from_consts_exit_111); + StoreInst* void_155 = new StoreInst(ptr_arraydecay11_3, ptr_tmp8_139, false, label_from_consts_exit_111); + std::vector ptr_arraydecay16_3_indices; + ptr_arraydecay16_3_indices.push_back(const_int32_34); + ptr_arraydecay16_3_indices.push_back(const_int32_30); + Instruction* ptr_arraydecay16_3 = new GetElementPtrInst(ptr_results_104, ptr_arraydecay16_3_indices.begin(), ptr_arraydecay16_3_indices.end(), "arraydecay16.3", label_from_consts_exit_111); + StoreInst* void_156 = new StoreInst(ptr_arraydecay16_3, ptr_tmp12_140, false, label_from_consts_exit_111); + StoreInst* void_157 = new StoreInst(const_int32_30, ptr_tmp, false, label_from_consts_exit_111); + CallInst* void_158 = new CallInst(func_execute_shader, ptr_args_114, "", label_from_consts_exit_111); + void_158->setCallingConv(CallingConv::C); + void_158->setTailCall(false); + LoadInst* int32_tmp23_3 = new LoadInst(ptr_tmp, "tmp23.3", false, label_from_consts_exit_111); + BinaryOperator* int32_shl_3 = BinaryOperator::create(Instruction::Shl, int32_tmp23_3, const_int32_34, "shl.3", label_from_consts_exit_111); + BinaryOperator* int32_or_3 = BinaryOperator::create(Instruction::Or, int32_shl_3, int32_or_2, "or.3", label_from_consts_exit_111); + BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_or_3, const_int32_36, "neg", label_from_consts_exit_111); + new ReturnInst(int32_neg, label_from_consts_exit_111); // Resolve Forward References - fwdref_216->replaceAllUsesWith(packed_tmp48_i_3); delete fwdref_216; - fwdref_215->replaceAllUsesWith(int32_inc_i_3); delete fwdref_215; - fwdref_151->replaceAllUsesWith(packed_tmp48_i_163); delete fwdref_151; - fwdref_149->replaceAllUsesWith(int32_inc_i_166); delete fwdref_149; - fwdref_172->replaceAllUsesWith(packed_tmp31_i_179); delete fwdref_172; - fwdref_171->replaceAllUsesWith(int32_indvar_next23); delete fwdref_171; - fwdref_189->replaceAllUsesWith(int32_or); delete fwdref_189; - fwdref_191->replaceAllUsesWith(int32_indvar_next21); delete fwdref_191; - fwdref_208->replaceAllUsesWith(packed_tmp48_i_1); delete fwdref_208; - fwdref_207->replaceAllUsesWith(int32_inc_i_1); delete fwdref_207; - fwdref_212->replaceAllUsesWith(packed_tmp48_i_2); delete fwdref_212; - fwdref_211->replaceAllUsesWith(int32_inc_i_2); delete fwdref_211; + fwdref_121->replaceAllUsesWith(packed_tmp31_i_133); delete fwdref_121; + fwdref_119->replaceAllUsesWith(int32_indvar_next7); delete fwdref_119; } diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index 03f7ac8f14..909bef340a 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -68,6 +68,7 @@ compute_clipmask(float4 clip, float4 (*plane), unsigned nr) return mask; } + inline void collect_results(float4 *results, struct vertex_header *vOut, float4 *planes, int nr_planes, float4 scale, float4 trans, @@ -76,7 +77,6 @@ inline void collect_results(float4 *results, struct vertex_header *vOut, /* store results */ unsigned slot; float x, y, z, w; - /* Handle attr[0] (position) specially: */ float4 res0 = results[0]; @@ -85,7 +85,6 @@ inline void collect_results(float4 *results, struct vertex_header *vOut, y = clip[1] = res0.y; z = clip[2] = res0.z; w = clip[3] = res0.w; - vOut->clipmask = compute_clipmask(res0, planes, nr_planes); vOut->edgeflag = 1; @@ -176,23 +175,20 @@ struct ShaderInput extern void execute_shader(struct ShaderInput *input); -void run_vertex_shader(float (*ainputs)[16][4], - float (*dests)[16][4], +void run_vertex_shader(float4 (*inputs)[16], + float4 (*results)[16], float (*aconsts)[4], int num_vertices, int num_inputs, int num_attribs, int num_consts) { - float4 inputs[16*32*4][16]; float4 consts[32]; - float4 results[16*32*4][16]; float4 temps[128];//MAX_PROGRAM_TEMPS struct ShaderInput args; /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", num_vertices, num_inputs, num_attribs, num_consts);*/ - from_array(inputs, ainputs, num_vertices, num_inputs); from_consts(consts, aconsts, num_consts); args.consts = consts; args.temps = temps; @@ -200,7 +196,6 @@ void run_vertex_shader(float (*ainputs)[16][4], args.dests = results[i]; args.inputs = inputs[i]; execute_shader(&args); - to_array(dests[i], args.dests, num_attribs); } } @@ -227,22 +222,19 @@ struct tgsi_sampler int run_fragment_shader(float x, float y, - float (*dests)[16][4], - float (*ainputs)[16][4], + float4 (*results)[16], + float4 (*inputs)[16], int num_inputs, float (*aconsts)[4], int num_consts, struct tgsi_sampler *samplers) { - float4 inputs[4][16]; float4 consts[32]; - float4 results[4][16]; float4 temps[128];//MAX_PROGRAM_TEMPS struct ShaderInput args; int mask = 0; args.kilmask = 0; - from_array(inputs, ainputs, 4, num_inputs); from_consts(consts, aconsts, num_consts); args.consts = consts; args.temps = temps; @@ -254,8 +246,6 @@ int run_fragment_shader(float x, float y, args.kilmask = 0; execute_shader(&args); args.kilmask = mask | (args.kilmask << i); - - to_array(dests[i], args.dests, 2); } return ~args.kilmask; } diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h index 020977b05b..4f2c9ef88a 100644 --- a/src/mesa/pipe/p_compiler.h +++ b/src/mesa/pipe/p_compiler.h @@ -73,9 +73,11 @@ typedef unsigned long long uint64; #if defined __GNUC__ #define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___aligned[SIZE] __attribute__(( aligned( 16 ) )) #define ALIGN16_ASSIGN(NAME) NAME##___aligned +#define ALIGN16_ATTRIB __attribute__(( aligned( 16 ) )) #else #define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___unaligned[SIZE + 1] #define ALIGN16_ASSIGN(NAME) align16(NAME##___unaligned) +#define ALIGN16_ATTRIB #endif diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 5ea07f95a4..251b47341a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -173,12 +173,11 @@ shade_quad_llvm(struct quad_stage *qs, { struct quad_shade_stage *qss = quad_shade_stage(qs); struct softpipe_context *softpipe = qs->softpipe; - float dests[4][16][4]; + float dests[4][16][4] ALIGN16_ATTRIB; + float inputs[4][16][4] ALIGN16_ATTRIB; const float fx = (float) quad->x0; const float fy = (float) quad->y0; struct gallivm_prog *llvm = qss->llvm_prog; - float inputs[4][16][4]; - memset(inputs, 0, sizeof(inputs)); inputs[0][0][0] = fx; inputs[1][0][0] = fx + 1.0f; -- cgit v1.2.3 From 63a9d835ce3f446fe2cf69e7623d228bcc6ee91c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 9 Nov 2007 10:08:15 -0500 Subject: Redo the fragment program examples to match vp's we just load text files instead of compiling tons of small binaries --- src/mesa/pipe/llvm/gallivm.cpp | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index a1101a7ba8..bdfe8af183 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -157,6 +157,7 @@ add_interpolator(struct gallivm_prog *prog, prog->interpolators[prog->num_interp] = *interp; ++prog->num_interp; } + static void translate_declaration(struct gallivm_prog *prog, llvm::Module *module, @@ -470,8 +471,6 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_TXD: break; - case TGSI_OPCODE_TXP: - break; case TGSI_OPCODE_UP2H: break; case TGSI_OPCODE_UP2US: @@ -1015,11 +1014,12 @@ struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) llvm::Module *mod = static_cast(prog->module); llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); + ee->DisableLazyCompilation(); cpu->engine = ee; llvm::Function *func = func_for_shader(prog); - prog->function = ee->getPointerToFunctionOrStub(func); + prog->function = ee->getPointerToFunction(func); CPU = cpu; return cpu; } @@ -1037,10 +1037,11 @@ void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); llvm::ExecutionEngine *ee = cpu->engine; assert(ee); + ee->DisableLazyCompilation(); ee->addModuleProvider(mp); llvm::Function *func = func_for_shader(prog); - prog->function = ee->getPointerToFunctionOrStub(func); + prog->function = ee->getPointerToFunction(func); } void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) -- cgit v1.2.3 From a0dacbf369b1b0496f7cd2df2f7b7ecb621a024c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 11 Dec 2007 09:44:41 -0500 Subject: Adopt llvm to some of the latest changes. --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 2 +- src/mesa/pipe/llvm/gallivm.cpp | 10 +++++----- src/mesa/pipe/llvm/storage.cpp | 2 +- 3 files changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 34a4d27684..acd61163fa 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -38,7 +38,7 @@ #ifdef MESA_LLVM #include "pipe/llvm/gallivm.h" -#include "pipe/tgsi/exec/tgsi_core.h" +#include "pipe/p_shader_tokens.h" #define DBG 0 diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index bdfe8af183..49bbf753c4 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -37,12 +37,12 @@ #include "storage.h" #include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" #include "pipe/tgsi/exec/tgsi_exec.h" -#include "pipe/tgsi/exec/tgsi_token.h" -#include "pipe/tgsi/exec/tgsi_build.h" -#include "pipe/tgsi/exec/tgsi_util.h" -#include "pipe/tgsi/exec/tgsi_parse.h" -#include "pipe/tgsi/exec/tgsi_dump.h" +#include "pipe/tgsi/util/tgsi_util.h" +#include "pipe/tgsi/util/tgsi_build.h" +#include "pipe/tgsi/util/tgsi_dump.h" #include #include diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index c33b9bf45d..08b9d29a24 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -33,7 +33,7 @@ #include "storage.h" -#include "pipe/tgsi/exec/tgsi_token.h" +#include "pipe/p_shader_tokens.h" #include #include #include -- cgit v1.2.3 From f9562ff85dd3ace60550110598687c3e728b0e6e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 08:18:47 -0700 Subject: add missing case statements for TGSI_TOKEN_TYPE_IMMEDIATE and assert(0) We're hitting this with quite a few Mesa demos. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index e403cfaaf3..8ee43f59b5 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -2245,6 +2245,11 @@ tgsi_emit_sse2( &parse.FullToken.FullInstruction ); break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* XXX implement this */ + assert(0); + break; + default: assert( 0 ); } @@ -2321,6 +2326,11 @@ tgsi_emit_sse2_fs( &parse.FullToken.FullInstruction ); break; + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* XXX implement this */ + assert(0); + break; + default: assert( 0 ); } -- cgit v1.2.3 From 6465c78049c07e4f97dd3e6494aca395ff357b0e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 08:46:13 -0700 Subject: Remove obsolete clear_color, tex sampler state functions --- src/mesa/pipe/cell/ppu/cell_context.c | 5 ----- src/mesa/pipe/cell/ppu/cell_state_sampler.c | 22 ---------------------- src/mesa/pipe/cell/ppu/cell_surface.c | 10 ---------- src/mesa/pipe/cell/ppu/cell_surface.h | 4 ---- 4 files changed, 41 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 5534d82ccf..966f355bb3 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -208,17 +208,12 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->pipe.set_blend_color = cell_set_blend_color; cell->pipe.set_clip_state = cell_set_clip_state; - cell->pipe.set_clear_color_state = cell_set_clear_color_state; cell->pipe.set_constant_buffer = cell_set_constant_buffer; -#if 0 - cell->pipe.set_feedback_state = cell_set_feedback_state; -#endif cell->pipe.set_framebuffer_state = cell_set_framebuffer_state; cell->pipe.set_polygon_stipple = cell_set_polygon_stipple; cell->pipe.set_scissor_state = cell_set_scissor_state; - cell->pipe.set_texture_state = cell_set_texture_state; cell->pipe.set_viewport_state = cell_set_viewport_state; cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; diff --git a/src/mesa/pipe/cell/ppu/cell_state_sampler.c b/src/mesa/pipe/cell/ppu/cell_state_sampler.c index 495567b16c..1e7d4f08b8 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_sampler.c +++ b/src/mesa/pipe/cell/ppu/cell_state_sampler.c @@ -66,25 +66,3 @@ cell_delete_sampler_state(struct pipe_context *pipe, { FREE( sampler ); } - - -void -cell_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, - struct pipe_texture *texture) -{ - struct cell_context *cell = cell_context(pipe); - - assert(unit < PIPE_MAX_SAMPLERS); - -#if 0 - cell->texture[sampler] = cell_texture(texture); /* ptr, not struct */ - cell_tile_cache_set_texture(cell->tex_cache[sampler], texture); -#endif - - cell->dirty |= CELL_NEW_TEXTURE; -} - - - - diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 66e4b5108b..59cd43e630 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -107,13 +107,3 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, } #endif } - - -void -cell_set_clear_color_state(struct pipe_context *pipe, - const struct pipe_clear_color_state *clear) -{ - struct cell_context *cell = cell_context(pipe); - - cell->clear_color = *clear; /* struct copy */ -} diff --git a/src/mesa/pipe/cell/ppu/cell_surface.h b/src/mesa/pipe/cell/ppu/cell_surface.h index 8b42ba02d5..e222507121 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.h +++ b/src/mesa/pipe/cell/ppu/cell_surface.h @@ -41,9 +41,5 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue); -extern void -cell_set_clear_color_state(struct pipe_context *pipe, - const struct pipe_clear_color_state *clear); - #endif /* CELL_SURFACE */ -- cgit v1.2.3 From 1bc303283e85ac15e74af1a2297166d87986e15e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 08:46:49 -0700 Subject: Remove obsolete clear_color, tex sampler state --- src/mesa/pipe/cell/ppu/cell_context.h | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 12073b93a0..96f000eef4 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -47,10 +47,8 @@ struct cell_context const struct pipe_rasterizer_state *rasterizer; struct pipe_blend_color blend_color; - struct pipe_clear_color_state clear_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; - struct pipe_feedback_state feedback; struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; -- cgit v1.2.3 From 9f0e5642d80d746aa757a1e67dec15b0fd404e5d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 08:47:13 -0700 Subject: use ALIGN16_ATTRIB from p_compiler.h --- src/mesa/pipe/cell/common.h | 14 +++++++------- src/mesa/pipe/cell/ppu/cell_spu.c | 4 ++-- src/mesa/pipe/cell/ppu/cell_spu.h | 6 +++--- src/mesa/pipe/cell/spu/main.c | 4 ++-- src/mesa/pipe/cell/spu/tri.c | 2 +- 5 files changed, 15 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 5e5abad8c4..a87417743a 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -33,11 +33,11 @@ #ifndef CELL_COMMON_H #define CELL_COMMON_H +#include "pipe/p_compiler.h" #include "pipe/p_util.h" -#define ALIGN16 __attribute__( (aligned( 16 )) ) - +/** for sanity checking */ #define ASSERT_ALIGN16(ptr) \ assert((((unsigned long) (ptr)) & 0xf) == 0); @@ -61,7 +61,7 @@ struct cell_command_framebuffer void *start; int width, height; unsigned format; -} ALIGN16; +} ALIGN16_ATTRIB; /** @@ -70,7 +70,7 @@ struct cell_command_framebuffer struct cell_command_clear_tiles { uint value; -} ALIGN16; +} ALIGN16_ATTRIB; struct cell_command_triangle @@ -79,7 +79,7 @@ struct cell_command_triangle float x1, y1; float x2, y2; uint color; -} ALIGN16; +} ALIGN16_ATTRIB; /** XXX unions don't seem to work */ @@ -88,7 +88,7 @@ struct cell_command struct cell_command_framebuffer fb; struct cell_command_clear_tiles clear; struct cell_command_triangle tri; -} ALIGN16; +} ALIGN16_ATTRIB; struct cell_init_info @@ -96,7 +96,7 @@ struct cell_init_info unsigned id; unsigned num_spus; struct cell_command *cmd; -} ALIGN16; +} ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 4f111cd219..37207c38d7 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -50,8 +50,8 @@ spe_spu_control_area_t *control_ps_area[MAX_SPUS]; /** * Data sent to SPUs */ -struct cell_init_info inits[MAX_SPUS] ALIGN16; -struct cell_command command[MAX_SPUS] ALIGN16; +struct cell_init_info inits[MAX_SPUS]; +struct cell_command command[MAX_SPUS]; /** diff --git a/src/mesa/pipe/cell/ppu/cell_spu.h b/src/mesa/pipe/cell/ppu/cell_spu.h index dcbc72573f..f030be45fc 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.h +++ b/src/mesa/pipe/cell/ppu/cell_spu.h @@ -36,7 +36,7 @@ #include "cell_context.h" -#define MAX_SPUS 7 +#define MAX_SPUS 8 /** * SPU/SPE handles, etc @@ -49,8 +49,8 @@ extern spe_spu_control_area_t *control_ps_area[MAX_SPUS]; /** * Data sent to SPUs */ -extern struct cell_init_info inits[MAX_SPUS] ALIGN16; -extern struct cell_command command[MAX_SPUS] ALIGN16; +extern struct cell_init_info inits[MAX_SPUS]; +extern struct cell_command command[MAX_SPUS]; void diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index f2c6d31d3c..83880bc906 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -114,7 +114,7 @@ clear_tiles(const struct cell_command_clear_tiles *clear) { uint num_tiles = fb.width_tiles * fb.height_tiles; uint i; - uint tile[TILE_SIZE * TILE_SIZE] ALIGN16; + uint tile[TILE_SIZE * TILE_SIZE] ALIGN16_ATTRIB; for (i = 0; i < TILE_SIZE * TILE_SIZE; i++) tile[i] = clear->value; @@ -161,7 +161,7 @@ triangle(const struct cell_command_triangle *tri) static void main_loop(void) { - struct cell_command cmd ALIGN16; + struct cell_command cmd; int exitFlag = 0; printf("SPU %u: Enter main loop\n", init.id); diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 793bf55743..60a62fdcd5 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -66,7 +66,7 @@ static int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; -static uint tile[TILE_SIZE][TILE_SIZE] ALIGN16; +static uint tile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; #endif -- cgit v1.2.3 From 5b8b542cb2f10c5c39b9db18cd95e553ad06061b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 08:58:39 -0700 Subject: Move align_malloc(), align_free() to p_util.h --- src/mesa/pipe/p_util.h | 52 ++++++++++++++++++++++++++++++++++++++++-- src/mesa/pipe/xlib/xm_winsys.c | 43 +--------------------------------- 2 files changed, 51 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 3c5e98453c..898459293e 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -30,6 +30,8 @@ #include "p_compiler.h" #include +#include + #ifdef WIN32 @@ -93,7 +95,7 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define GETENV( X ) NULL -#else // WIN32 +#else /* WIN32 */ #define MALLOC( SIZE ) malloc( SIZE ) @@ -105,10 +107,56 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define GETENV( X ) getenv( X ) -#endif // WIN32 +#endif /* WIN32 */ #define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) + + +/** + * Return memory on given byte alignment + */ +static INLINE void * +align_malloc(size_t bytes, unsigned long alignment) +{ +#if defined(HAVE_POSIX_MEMALIGN) + void *mem; + + (void) posix_memalign(& mem, alignment, bytes); + return mem; +#else + uintptr_t ptr, buf; + + assert( alignment > 0 ); + + ptr = (uintptr_t) MALLOC(bytes + alignment + sizeof(void *)); + if (!ptr) + return NULL; + + buf = (ptr + alignment + sizeof(void *)) & ~(uintptr_t)(alignment - 1); + *(uintptr_t *)(buf - sizeof(void *)) = ptr; + + return (void *) buf; +#endif /* defined(HAVE_POSIX_MEMALIGN) */ +} + +/** + * Free memory returned by align_malloc(). + */ +static INLINE void +align_free(void *ptr) +{ +#if defined(HAVE_POSIX_MEMALIGN) + FREE(ptr); +#else + void **cubbyHole = (void **) ((char *) ptr - sizeof(void *)); + void *realAddr = *cubbyHole; + FREE(realAddr); +#endif /* defined(HAVE_POSIX_MEMALIGN) */ +} + + + #define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index a03c9835a5..42c43387af 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -35,11 +35,11 @@ #include "glxheader.h" #include "xmesaP.h" -#include "main/macros.h" #include "pipe/p_winsys.h" #include "pipe/p_format.h" #include "pipe/p_context.h" +#include "pipe/p_util.h" #include "pipe/softpipe/sp_winsys.h" #ifdef GALLIUM_CELL @@ -49,47 +49,6 @@ #include "xm_winsys_aub.h" -/** XXX from Mesa core */ -static void * -align_malloc(size_t bytes, unsigned long alignment) -{ -#if defined(HAVE_POSIX_MEMALIGN) - void *mem; - - (void) posix_memalign(& mem, alignment, bytes); - return mem; -#else - uintptr_t ptr, buf; - - assert( alignment > 0 ); - - ptr = (uintptr_t) malloc(bytes + alignment + sizeof(void *)); - if (!ptr) - return NULL; - - buf = (ptr + alignment + sizeof(void *)) & ~(uintptr_t)(alignment - 1); - *(uintptr_t *)(buf - sizeof(void *)) = ptr; - - return (void *) buf; -#endif /* defined(HAVE_POSIX_MEMALIGN) */ -} - - -/** XXX from Mesa core */ -static void -align_free(void *ptr) -{ -#if defined(HAVE_POSIX_MEMALIGN) - free(ptr); -#else - void **cubbyHole = (void **) ((char *) ptr - sizeof(void *)); - void *realAddr = *cubbyHole; - free(realAddr); -#endif /* defined(HAVE_POSIX_MEMALIGN) */ -} - - - /** * Low-level OS/window system memory buffer */ -- cgit v1.2.3 From 23ca30e24bbc8390127910c64da4a321ae63ce3a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 09:09:06 -0700 Subject: change align_malloc() alignment to uint --- src/mesa/pipe/p_util.h | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 898459293e..b4d1195c9b 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -117,11 +117,10 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) * Return memory on given byte alignment */ static INLINE void * -align_malloc(size_t bytes, unsigned long alignment) +align_malloc(size_t bytes, uint alignment) { #if defined(HAVE_POSIX_MEMALIGN) void *mem; - (void) posix_memalign(& mem, alignment, bytes); return mem; #else -- cgit v1.2.3 From 4a4543f5360b5f0a3350d9db770f1325cc3a9e84 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 09:13:12 -0700 Subject: Fix align16() function to work with 32/64-bit pointers on big-endian. --- src/mesa/pipe/p_util.h | 26 +++++++++++++++++++------- 1 file changed, 19 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index b4d1195c9b..e6d284d932 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -169,13 +169,25 @@ align_free(void *ptr) static INLINE void * align16( void *unaligned ) { - union { - void *p; - uint64 u; - } pu; - pu.p = unaligned; - pu.u = (pu.u + 15) & ~15; - return pu.p; + if (sizeof(void *) == 64) { + union { + void *p; + uint64 u; + } pu; + pu.p = unaligned; + pu.u = (pu.u + 15) & ~15; + return pu.p; + } + else { + /* 32-bit pointers */ + union { + void *p; + uint u; + } pu; + pu.p = unaligned; + pu.u = (pu.u + 15) & ~15; + return pu.p; + } } -- cgit v1.2.3 From ecb5e14e8688e31446fdcbdea59d25ea628fb211 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 11 Dec 2007 15:48:30 +0000 Subject: gallium: remove clear values from depth, stencil state --- src/mesa/pipe/p_state.h | 2 -- src/mesa/state_tracker/st_atom_depth.c | 2 -- 2 files changed, 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 56d4f96ed7..a571071ea9 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -152,7 +152,6 @@ struct pipe_depth_stencil_state unsigned writemask:1; /**< allow depth buffer writes? */ unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ unsigned occlusion_count:1; /**< XXX move this elsewhere? */ - float clear; /**< Clear value in [0,1] (XXX correct place?) */ } depth; struct { unsigned front_enabled:1; @@ -168,7 +167,6 @@ struct pipe_depth_stencil_state ubyte ref_value[2]; /**< [0] = front, [1] = back */ ubyte value_mask[2]; ubyte write_mask[2]; - ubyte clear_value; } stencil; }; diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index f0f360e6dd..e785434cec 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -101,7 +101,6 @@ update_depth_stencil(struct st_context *st) depth_stencil.depth.enabled = st->ctx->Depth.Test; depth_stencil.depth.writemask = st->ctx->Depth.Mask; depth_stencil.depth.func = st_compare_func_to_pipe(st->ctx->Depth.Func); - depth_stencil.depth.clear = st->ctx->Depth.Clear; if (st->ctx->Query.CurrentOcclusionObject && st->ctx->Query.CurrentOcclusionObject->Active) @@ -126,7 +125,6 @@ update_depth_stencil(struct st_context *st) depth_stencil.stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1] & 0xff; depth_stencil.stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1] & 0xff; } - depth_stencil.stencil.clear_value = st->ctx->Stencil.Clear & 0xff; } cso = st_cached_depth_stencil_state(st, &depth_stencil); -- cgit v1.2.3 From b247ab036327d66b8b9b1aff2dbcf4520ed0284f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 11 Dec 2007 17:07:31 +0000 Subject: 965: remove duplicate brw_draw functions, make init a bit cleaner --- src/mesa/pipe/i965simple/brw_context.c | 53 ++++------------------------------ src/mesa/pipe/i965simple/brw_draw.c | 5 +--- src/mesa/pipe/i965simple/brw_draw.h | 3 +- 3 files changed, 7 insertions(+), 54 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_context.c b/src/mesa/pipe/i965simple/brw_context.c index dc1987cec8..13bed97ff2 100644 --- a/src/mesa/pipe/i965simple/brw_context.c +++ b/src/mesa/pipe/i965simple/brw_context.c @@ -195,36 +195,7 @@ brw_is_format_supported( struct pipe_context *pipe, #endif } -static void -brw_begin_query(struct pipe_context *pipe, struct pipe_query_object *q) -{ - /* should never be called */ - assert(0); -} - - -static void -brw_end_query(struct pipe_context *pipe, struct pipe_query_object *q) -{ - /* should never be called */ - assert(0); -} - - -static boolean -brw_draw_elements(struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, - unsigned indexSize, - unsigned prim, unsigned start, unsigned count) -{ - return FALSE; -} -static boolean brw_draw_arrays(struct pipe_context *pipe, - unsigned prim, unsigned start, unsigned count) -{ - return brw_draw_elements(pipe, NULL, 0, prim, start, count); -} struct pipe_context *brw_create(struct pipe_winsys *pipe_winsys, @@ -248,40 +219,26 @@ struct pipe_context *brw_create(struct pipe_winsys *pipe_winsys, brw->pipe.is_format_supported = brw_is_format_supported; brw->pipe.get_param = brw_get_param; brw->pipe.get_paramf = brw_get_paramf; - brw->pipe.clear = brw_clear; - - brw->pipe.begin_query = brw_begin_query; - brw->pipe.end_query = brw_end_query; - - brw->pipe.draw_arrays = brw_draw_arrays; - brw->pipe.draw_elements = brw_draw_elements; + brw->pipe.texture_create = brw_texture_create; + brw->pipe.texture_release = brw_texture_release; brw_init_surface_functions(brw); brw_init_state_functions(brw); brw_init_flush_functions(brw); brw_init_string_functions(brw); + brw_init_draw_functions( brw ); - brw->pci_id = pci_id; - brw->pipe.texture_create = brw_texture_create; - brw->pipe.texture_release = brw_texture_release; + brw_init_state( brw ); + brw->pci_id = pci_id; brw->dirty = ~0; brw->hardware_dirty = ~0; - - /* Batch stream debugging is a bit hacked up at the moment: - */ brw->batch_start = NULL; - - brw_init_state( brw ); - memset(&brw->wm.bind, ~0, sizeof(brw->wm.bind)); - brw_draw_init( brw ); - - return &brw->pipe; } diff --git a/src/mesa/pipe/i965simple/brw_draw.c b/src/mesa/pipe/i965simple/brw_draw.c index 8b92f11280..f443f41c6f 100644 --- a/src/mesa/pipe/i965simple/brw_draw.c +++ b/src/mesa/pipe/i965simple/brw_draw.c @@ -226,13 +226,10 @@ static boolean brw_draw_arrays( struct pipe_context *pipe, -void brw_draw_init( struct brw_context *brw ) +void brw_init_draw_functions( struct brw_context *brw ) { brw->pipe.draw_arrays = brw_draw_arrays; brw->pipe.draw_elements = brw_draw_elements; } -void brw_draw_destroy( struct brw_context *brw ) -{ -} diff --git a/src/mesa/pipe/i965simple/brw_draw.h b/src/mesa/pipe/i965simple/brw_draw.h index ebac8365c8..053f2efb9d 100644 --- a/src/mesa/pipe/i965simple/brw_draw.h +++ b/src/mesa/pipe/i965simple/brw_draw.h @@ -34,8 +34,7 @@ struct brw_context; -void brw_draw_init( struct brw_context *brw ); -void brw_draw_destroy( struct brw_context *brw ); +void brw_init_draw_functions( struct brw_context *brw ); boolean brw_upload_vertices( struct brw_context *brw, -- cgit v1.2.3 From 13699463a33c1adf44005125c488e886e074a05b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 11 Dec 2007 17:10:26 +0000 Subject: Rework gallium and mesa queries a little. Add a 'CheckQuery()' driver callback to mesa to check query completion. Make pipe_query an opaque type. Rework softpipe queries, support overlapping occlusion queries. --- src/mesa/drivers/common/driverfuncs.c | 1 + src/mesa/main/dd.h | 2 + src/mesa/main/queryobj.c | 18 +++-- src/mesa/main/queryobj.h | 3 + src/mesa/pipe/i915simple/i915_context.c | 17 ----- src/mesa/pipe/p_context.h | 20 +++++- src/mesa/pipe/p_state.h | 10 --- src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_context.c | 36 +--------- src/mesa/pipe/softpipe/sp_context.h | 6 +- src/mesa/pipe/softpipe/sp_quad_occlusion.c | 16 +++-- src/mesa/pipe/softpipe/sp_query.c | 107 +++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_query.h | 39 +++++++++++ src/mesa/state_tracker/st_cb_queryobj.c | 68 +++++++++++++----- 14 files changed, 247 insertions(+), 97 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_query.c create mode 100644 src/mesa/pipe/softpipe/sp_query.h (limited to 'src') diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c index ea0cc51d4a..33caf7dae1 100644 --- a/src/mesa/drivers/common/driverfuncs.c +++ b/src/mesa/drivers/common/driverfuncs.c @@ -224,6 +224,7 @@ _mesa_init_driver_functions(struct dd_function_table *driver) /* query objects */ driver->NewQueryObject = _mesa_new_query_object; + driver->DeleteQuery = _mesa_delete_query; driver->BeginQuery = _mesa_begin_query; driver->EndQuery = _mesa_end_query; driver->WaitQuery = _mesa_wait_query; diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index f089fcb48f..3bec3bd433 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -811,8 +811,10 @@ struct dd_function_table { */ /*@{*/ struct gl_query_object * (*NewQueryObject)(GLcontext *ctx, GLuint id); + void (*DeleteQuery)(GLcontext *ctx, struct gl_query_object *q); void (*BeginQuery)(GLcontext *ctx, struct gl_query_object *q); void (*EndQuery)(GLcontext *ctx, struct gl_query_object *q); + void (*CheckQuery)(GLcontext *ctx, struct gl_query_object *q); void (*WaitQuery)(GLcontext *ctx, struct gl_query_object *q); /*@}*/ diff --git a/src/mesa/main/queryobj.c b/src/mesa/main/queryobj.c index 688d0fc7bc..e30f5480da 100644 --- a/src/mesa/main/queryobj.c +++ b/src/mesa/main/queryobj.c @@ -94,8 +94,8 @@ _mesa_wait_query(GLcontext *ctx, struct gl_query_object *q) * Not removed from hash table here. * XXX maybe add Delete() method to gl_query_object class and call that instead */ -static void -delete_query_object(struct gl_query_object *q) +void +_mesa_delete_query(struct gl_query_object *q) { _mesa_free(q); } @@ -171,7 +171,7 @@ _mesa_DeleteQueriesARB(GLsizei n, const GLuint *ids) if (q) { ASSERT(!q->Active); /* should be caught earlier */ _mesa_HashRemove(ctx->Query.QueryObjects, ids[i]); - delete_query_object(q); + ctx->Driver.DeleteQuery(ctx, q); } } } @@ -386,6 +386,8 @@ _mesa_GetQueryObjectivARB(GLuint id, GLenum pname, GLint *params) } break; case GL_QUERY_RESULT_AVAILABLE_ARB: + if (!q->Ready) + ctx->Driver.CheckQuery( ctx, q ); *params = q->Ready; break; default: @@ -424,6 +426,8 @@ _mesa_GetQueryObjectuivARB(GLuint id, GLenum pname, GLuint *params) } break; case GL_QUERY_RESULT_AVAILABLE_ARB: + if (!q->Ready) + ctx->Driver.CheckQuery( ctx, q ); *params = q->Ready; break; default: @@ -461,6 +465,8 @@ _mesa_GetQueryObjecti64vEXT(GLuint id, GLenum pname, GLint64EXT *params) *params = q->Result; break; case GL_QUERY_RESULT_AVAILABLE_ARB: + if (!q->Ready) + ctx->Driver.CheckQuery( ctx, q ); *params = q->Ready; break; default: @@ -496,6 +502,8 @@ _mesa_GetQueryObjectui64vEXT(GLuint id, GLenum pname, GLuint64EXT *params) *params = q->Result; break; case GL_QUERY_RESULT_AVAILABLE_ARB: + if (!q->Ready) + ctx->Driver.CheckQuery( ctx, q ); *params = q->Ready; break; default: @@ -527,8 +535,8 @@ static void delete_queryobj_cb(GLuint id, void *data, void *userData) { struct gl_query_object *q= (struct gl_query_object *) data; - (void) userData; - delete_query_object(q); + GLcontext *ctx = (GLcontext *)userData; + ctx->Driver.DeleteQuery(ctx, q); } diff --git a/src/mesa/main/queryobj.h b/src/mesa/main/queryobj.h index d466aae652..c05a1f3da8 100644 --- a/src/mesa/main/queryobj.h +++ b/src/mesa/main/queryobj.h @@ -36,6 +36,9 @@ _mesa_init_query(GLcontext *ctx); extern void _mesa_free_query_data(GLcontext *ctx); +extern void +_mesa_delete_query(struct gl_query_object *q); + extern void _mesa_begin_query(GLcontext *ctx, struct gl_query_object *q); diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index d2bbeea16a..a08cf5087e 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -170,21 +170,6 @@ static void i915_destroy( struct pipe_context *pipe ) -static void -i915_begin_query(struct pipe_context *pipe, struct pipe_query_object *q) -{ - /* should never be called */ - assert(0); -} - - -static void -i915_end_query(struct pipe_context *pipe, struct pipe_query_object *q) -{ - /* should never be called */ - assert(0); -} - static boolean i915_draw_elements( struct pipe_context *pipe, @@ -298,8 +283,6 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.clear = i915_clear; - i915->pipe.begin_query = i915_begin_query; - i915->pipe.end_query = i915_end_query; i915->pipe.draw_arrays = i915_draw_arrays; i915->pipe.draw_elements = i915_draw_elements; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 83b4ab07fb..92ca7dd8e3 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -29,9 +29,13 @@ #define PIPE_CONTEXT_H #include "p_state.h" +#include struct pipe_state_cache; +/* Opaque driver handles: + */ +struct pipe_query; /** * Gallium rendering context. Basically: @@ -81,9 +85,19 @@ struct pipe_context { /** * Query objects */ - void (*begin_query)(struct pipe_context *pipe, struct pipe_query_object *q); - void (*end_query)(struct pipe_context *pipe, struct pipe_query_object *q); - void (*wait_query)(struct pipe_context *pipe, struct pipe_query_object *q); + struct pipe_query *(*create_query)( struct pipe_context *pipe, + unsigned query_type ); + + void (*destroy_query)(struct pipe_context *pipe, + struct pipe_query *q); + + void (*begin_query)(struct pipe_context *pipe, struct pipe_query *q); + void (*end_query)(struct pipe_context *pipe, struct pipe_query *q); + + boolean (*get_query_result)(struct pipe_context *pipe, + struct pipe_query *q, + boolean wait, + uint64_t *result); /* * State functions diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index a571071ea9..109913b040 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -317,14 +317,4 @@ struct pipe_vertex_element -/** - * Hardware queries (occlusion, transform feedback, timing, etc) - */ -struct pipe_query_object { - uint type:3; /**< PIPE_QUERY_x */ - uint ready:1; /**< is result ready? */ - uint64 count; -}; - - #endif diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 5e6886a37e..31438a882e 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -7,6 +7,7 @@ LIBNAME = softpipe DRIVER_SOURCES = \ sp_clear.c \ sp_flush.c \ + sp_query.c \ sp_context.c \ sp_draw_arrays.c \ sp_prim_setup.c \ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 7d243aaabb..107c8f8597 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -42,6 +42,7 @@ #include "sp_tile_cache.h" #include "sp_texture.h" #include "sp_winsys.h" +#include "sp_query.h" @@ -150,37 +151,6 @@ static void softpipe_destroy( struct pipe_context *pipe ) } -static void -softpipe_begin_query(struct pipe_context *pipe, struct pipe_query_object *q) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - assert(q->type < PIPE_QUERY_TYPES); - assert(!softpipe->queries[q->type]); - softpipe->queries[q->type] = q; -} - - -static void -softpipe_end_query(struct pipe_context *pipe, struct pipe_query_object *q) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - assert(q->type < PIPE_QUERY_TYPES); - assert(softpipe->queries[q->type]); - q->ready = 1; /* software rendering is synchronous */ - softpipe->queries[q->type] = NULL; -} - - -static void -softpipe_wait_query(struct pipe_context *pipe, struct pipe_query_object *q) -{ - /* Should never get here since we indicated that the result was - * ready in softpipe_end_query(). - */ - assert(0); -} - - static const char *softpipe_get_name( struct pipe_context *pipe ) { return "softpipe"; @@ -320,9 +290,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; - softpipe->pipe.begin_query = softpipe_begin_query; - softpipe->pipe.end_query = softpipe_end_query; - softpipe->pipe.wait_query = softpipe_wait_query; + softpipe_init_query_funcs( softpipe ); /* textures */ softpipe->pipe.texture_create = softpipe_texture_create; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index afdd0ec88e..2c038de5f7 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -93,10 +93,10 @@ struct softpipe_context { struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; unsigned dirty; - /* - * Active queries + /* Counter for occlusion queries. Note this supports overlapping + * queries. */ - struct pipe_query_object *queries[PIPE_QUERY_TYPES]; + uint64_t occlusion_count; /* * Mapped vertex buffers diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c index d65fdbdab7..54254df1f1 100644 --- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c +++ b/src/mesa/pipe/softpipe/sp_quad_occlusion.c @@ -39,18 +39,22 @@ #include "sp_surface.h" #include "sp_quad.h" +static unsigned count_bits( unsigned val ) +{ + unsigned i; + + for (i = 0; val ; val >>= 1) + i += (val & 1); + + return i; +} static void occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct pipe_query_object *occ - = softpipe->queries[PIPE_QUERY_OCCLUSION_COUNTER]; - occ->count += (quad->mask ) & 1; - occ->count += (quad->mask >> 1) & 1; - occ->count += (quad->mask >> 2) & 1; - occ->count += (quad->mask >> 3) & 1; + softpipe->occlusion_count += count_bits(quad->mask); qs->next->run(qs->next, quad); } diff --git a/src/mesa/pipe/softpipe/sp_query.c b/src/mesa/pipe/softpipe/sp_query.c new file mode 100644 index 0000000000..bf753dad98 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_query.c @@ -0,0 +1,107 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#include "pipe/draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_query.h" + +struct softpipe_query { + uint64_t start; + uint64_t end; +}; + + +static struct softpipe_query *softpipe_query( struct pipe_query *p ) +{ + return (struct softpipe_query *)p; +} + +static struct pipe_query * +softpipe_create_query(struct pipe_context *pipe, + unsigned type) +{ + assert(type == PIPE_QUERY_OCCLUSION_COUNTER); + return (struct pipe_query *)CALLOC_STRUCT( softpipe_query ); +} + + +static void +softpipe_destroy_query(struct pipe_context *pipe, struct pipe_query *q) +{ + FREE(q); +} + + +static void +softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + struct softpipe_query *sq = softpipe_query(q); + + sq->start = softpipe->occlusion_count; +} + + +static void +softpipe_end_query(struct pipe_context *pipe, struct pipe_query *q) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + struct softpipe_query *sq = softpipe_query(q); + + sq->end = softpipe->occlusion_count; +} + + +static boolean +softpipe_get_query_result(struct pipe_context *pipe, + struct pipe_query *q, + boolean wait, + uint64_t *result ) +{ + struct softpipe_query *sq = softpipe_query(q); + *result = sq->end - sq->start; + return TRUE; +} + + +void softpipe_init_query_funcs(struct softpipe_context *softpipe ) +{ + softpipe->pipe.create_query = softpipe_create_query; + softpipe->pipe.destroy_query = softpipe_destroy_query; + softpipe->pipe.begin_query = softpipe_begin_query; + softpipe->pipe.end_query = softpipe_end_query; + softpipe->pipe.get_query_result = softpipe_get_query_result; +} + + diff --git a/src/mesa/pipe/softpipe/sp_query.h b/src/mesa/pipe/softpipe/sp_query.h new file mode 100644 index 0000000000..05060a4575 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_query.h @@ -0,0 +1,39 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#ifndef SP_QUERY_H +#define SP_QUERY_H + +struct softpipe_context; +extern void softpipe_init_query_funcs(struct softpipe_context * ); + + +#endif /* SP_QUERY_H */ diff --git a/src/mesa/state_tracker/st_cb_queryobj.c b/src/mesa/state_tracker/st_cb_queryobj.c index 5b95dd7fd3..c1d0d086b4 100644 --- a/src/mesa/state_tracker/st_cb_queryobj.c +++ b/src/mesa/state_tracker/st_cb_queryobj.c @@ -47,7 +47,7 @@ struct st_query_object { struct gl_query_object base; - struct pipe_query_object pq; + struct pipe_query *pq; }; @@ -68,12 +68,28 @@ st_NewQueryObject(GLcontext *ctx, GLuint id) if (stq) { stq->base.Id = id; stq->base.Ready = GL_TRUE; + stq->pq = NULL; return &stq->base; } return NULL; } + +static void +st_DeleteQuery(GLcontext *ctx, struct gl_query_object *q) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_query_object *stq = st_query_object(q); + + if (stq->pq) { + pipe->destroy_query(pipe, stq->pq); + stq->pq = NULL; + } + + FREE(stq); +} + /** * Do glReadPixels by getting rows from the framebuffer surface with * get_tile(). Convert to requested format/type with Mesa image routines. @@ -85,25 +101,17 @@ st_BeginQuery(GLcontext *ctx, struct gl_query_object *q) struct pipe_context *pipe = ctx->st->pipe; struct st_query_object *stq = st_query_object(q); - stq->pq.count = 0; - switch (q->Target) { case GL_SAMPLES_PASSED_ARB: - stq->pq.type = PIPE_QUERY_OCCLUSION_COUNTER; - break; - case GL_PRIMITIVES_GENERATED_NV: - /* someday */ - stq->pq.type = PIPE_QUERY_PRIMITIVES_GENERATED; - break; - case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN_NV: - /* someday */ - stq->pq.type = PIPE_QUERY_PRIMITIVES_EMITTED; + if (!stq->pq) + stq->pq = pipe->create_query( pipe, PIPE_QUERY_OCCLUSION_COUNTER ); break; default: assert(0); + return; } - pipe->begin_query(pipe, &stq->pq); + pipe->begin_query(pipe, stq->pq); } @@ -113,10 +121,7 @@ st_EndQuery(GLcontext *ctx, struct gl_query_object *q) struct pipe_context *pipe = ctx->st->pipe; struct st_query_object *stq = st_query_object(q); - pipe->end_query(pipe, &stq->pq); - stq->base.Ready = stq->pq.ready; - if (stq->base.Ready) - stq->base.Result = stq->pq.count; + pipe->end_query(pipe, stq->pq); } @@ -129,17 +134,42 @@ st_WaitQuery(GLcontext *ctx, struct gl_query_object *q) /* this function should only be called if we don't have a ready result */ assert(!stq->base.Ready); - pipe->wait_query(pipe, &stq->pq); + while (!stq->base.Ready && + !pipe->get_query_result(pipe, + stq->pq, + TRUE, + &q->Result)) + { + /* nothing */ + } + q->Ready = GL_TRUE; - q->Result = stq->pq.count; } +static void +st_CheckQuery(GLcontext *ctx, struct gl_query_object *q) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_query_object *stq = st_query_object(q); + + if (!q->Ready) { + q->Ready = pipe->get_query_result(pipe, + stq->pq, + FALSE, + &q->Result); + } +} + + + void st_init_query_functions(struct dd_function_table *functions) { functions->NewQueryObject = st_NewQueryObject; + functions->DeleteQuery = st_DeleteQuery; functions->BeginQuery = st_BeginQuery; functions->EndQuery = st_EndQuery; functions->WaitQuery = st_WaitQuery; + functions->CheckQuery = st_CheckQuery; } -- cgit v1.2.3 From 88723b2fc84628c1bc1e0008b88602b85e8668be Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 11 Dec 2007 19:02:26 +0100 Subject: Remove internal_format field from struct pipe_texture. It's state tracker specific / not really necessary anyway. --- src/mesa/pipe/p_state.h | 5 ++--- src/mesa/state_tracker/st_cb_drawpixels.c | 15 ++++++--------- src/mesa/state_tracker/st_cb_texture.c | 5 ++--- src/mesa/state_tracker/st_texture.c | 6 ++---- src/mesa/state_tracker/st_texture.h | 1 - 5 files changed, 12 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 109913b040..4e42838f1d 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -264,10 +264,9 @@ struct pipe_texture { /* Effectively the key: */ - unsigned target; /* XXX convert to PIPE_TEXTURE_x */ - unsigned internal_format; /* XXX convert to PIPE_FORMAT_x */ - + unsigned target; /**< PIPE_TEXTURE_x */ enum pipe_format format; /**< PIPE_FORMAT_x */ + unsigned first_level; unsigned last_level; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 5d4c443c01..0bc48b7039 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -469,8 +469,8 @@ make_texture(struct st_context *st, assert(pipeFormat); cpp = st_sizeof_format(pipeFormat); - pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, baseFormat, 0, 0, - width, height, 1, 0); + pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, 0, width, height, + 1, 0); if (!pt) return NULL; @@ -987,7 +987,6 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, struct pipe_context *pipe = ctx->st->pipe; struct pipe_surface *surface; uint format = 0, cpp, comp; - GLenum internal_format; ubyte *dest; struct pipe_texture *pt; int row, col; @@ -995,13 +994,11 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, /* find a texture format we know */ if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8, PIPE_TEXTURE )) { format = PIPE_FORMAT_U_I8; - internal_format = GL_INTENSITY8; cpp = 1; comp = 0; } else if (pipe->is_format_supported( pipe, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_TEXTURE )) { format = PIPE_FORMAT_A8R8G8B8_UNORM; - internal_format = GL_RGBA8; cpp = 4; comp = 3; /* alpha channel */ /*XXX little-endian dependency */ } @@ -1013,8 +1010,8 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, /** * Create a texture. */ - pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, internal_format, - 0, 0, width, height, 1, 0); + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, 0, width, height, + 1, 0); if (!pt) return NULL; @@ -1237,8 +1234,8 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, psRead = rbRead->surface; format = psRead->format; - pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, - rbRead->Base.InternalFormat, 0, 0, width, height, 1, 0); + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, 0, width, height, + 1, 0); if (!pt) return; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 39dd21dc59..e813bdb47a 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -366,7 +366,6 @@ guess_and_alloc_texture(struct st_context *st, stObj->pt = st_texture_create(st, gl_target_to_pipe(stObj->base.Target), st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat), - stImage->base.InternalFormat, firstLevel, lastLevel, width, @@ -1481,7 +1480,8 @@ st_finalize_texture(GLcontext *ctx, */ if (stObj->pt && (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || - stObj->pt->internal_format != firstImage->base.InternalFormat || + stObj->pt->format != + st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat) || stObj->pt->first_level != stObj->firstLevel || stObj->pt->last_level != stObj->lastLevel || stObj->pt->width[0] != firstImage->base.Width || @@ -1499,7 +1499,6 @@ st_finalize_texture(GLcontext *ctx, stObj->pt = st_texture_create(ctx->st, gl_target_to_pipe(stObj->base.Target), st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat), - firstImage->base.InternalFormat, stObj->firstLevel, stObj->lastLevel, firstImage->base.Width, diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 1ec4514873..c7d28eeca2 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -63,7 +63,6 @@ struct pipe_texture * st_texture_create(struct st_context *st, unsigned target, unsigned format, - GLenum internal_format, GLuint first_level, GLuint last_level, GLuint width0, @@ -77,7 +76,7 @@ st_texture_create(struct st_context *st, DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), - _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); + _mesa_lookup_enum_by_nr(format), first_level, last_level); if (!pt) return NULL; @@ -86,7 +85,6 @@ st_texture_create(struct st_context *st, pt->target = target; pt->format = format; - pt->internal_format = internal_format; pt->first_level = first_level; pt->last_level = last_level; pt->width[0] = width0; @@ -119,7 +117,7 @@ st_texture_match_image(struct pipe_texture *pt, if (image->Border) return GL_FALSE; - if (image->InternalFormat != pt->internal_format || + if (st_mesa_format_to_pipe_format(image->TexFormat->MesaFormat) != pt->format || image->IsCompressed != pt->compressed) return GL_FALSE; diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 2be53abf3a..7524c219e0 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -39,7 +39,6 @@ extern struct pipe_texture * st_texture_create(struct st_context *st, unsigned target, unsigned format, - GLenum internal_format, GLuint first_level, GLuint last_level, GLuint width0, -- cgit v1.2.3 From f3789748d4b8f38bfea2f30ef93e9ff3e3888af4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 11 Dec 2007 19:07:12 +0100 Subject: softpipe: Support for PIPE_FORMAT_A4R4G4B4_UNORM and PIPE_FORMAT_R5G6B5_UNORM. The packedpixels test runs with the xlib winsys, though not all cases look correct yet. --- src/mesa/pipe/softpipe/sp_surface.c | 59 +++++++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_format.c | 4 +++ 2 files changed, 63 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 3ef3db9f1f..1dc494d6ff 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -215,6 +215,59 @@ a1r5g5b5_get_tile(struct pipe_surface *ps, } +/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ + +static void +a4r4g4b4_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + + assert(ps->format == PIPE_FORMAT_A4R4G4B4_UNORM); + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const ushort pixel = src[j]; + p[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); + p[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); + p[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); + p[3] = ((pixel >> 12) ) * (1.0f / 15.0f); + p += 4; + } + src += ps->pitch; + } +} + + +/*** PIPE_FORMAT_R5G6B5_UNORM ***/ + +static void +r5g6b5_get_tile(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + + assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const ushort pixel = src[j]; + p[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); + p[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); + p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + p[3] = 1.0f; + p += 4; + } + src += ps->pitch; + } +} + /*** PIPE_FORMAT_Z16_UNORM ***/ @@ -674,6 +727,12 @@ softpipe_get_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_A1R5G5B5_UNORM: a1r5g5b5_get_tile(ps, x, y, w, h, p); break; + case PIPE_FORMAT_A4R4G4B4_UNORM: + a4r4g4b4_get_tile(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g6b5_get_tile(ps, x, y, w, h, p); + break; case PIPE_FORMAT_U_L8: l8_get_tile(ps, x, y, w, h, p); break; diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index c292a975f3..98cc2084f0 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -254,8 +254,12 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) case MESA_FORMAT_ARGB8888_REV: case MESA_FORMAT_ARGB8888: return PIPE_FORMAT_A8R8G8B8_UNORM; + case MESA_FORMAT_ARGB1555: + return PIPE_FORMAT_A1R5G5B5_UNORM; case MESA_FORMAT_ARGB4444: return PIPE_FORMAT_A4R4G4B4_UNORM; + case MESA_FORMAT_RGB565: + return PIPE_FORMAT_R5G6B5_UNORM; case MESA_FORMAT_AL88: return PIPE_FORMAT_U_A8_L8; case MESA_FORMAT_A8: -- cgit v1.2.3 From 9a0fe23d493dc9307d76f0a35cbe4f410ee3df28 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 11 Dec 2007 12:40:58 -0800 Subject: Port to libspe2 libspe is deprecated in Cell SDK 2.1, and it is removed in Cell SDK 3.0. This patch ports the existing libspe-based code to libspe2. Some error checking (eh-hem) would be nice. --- src/mesa/pipe/cell/ppu/cell_flush.c | 4 +- src/mesa/pipe/cell/ppu/cell_spu.c | 83 +++++++++++++++++++---------------- src/mesa/pipe/cell/ppu/cell_spu.h | 14 +++--- src/mesa/pipe/cell/ppu/cell_surface.c | 6 +-- 4 files changed, 58 insertions(+), 49 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c index e844d13f06..39b72c435d 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -43,12 +43,12 @@ cell_flush(struct pipe_context *pipe, unsigned flags) /* Send CMD_FINISH to all SPUs */ for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(control_ps_area[i], CELL_CMD_FINISH); + send_mbox_message(spe_contexts[i], CELL_CMD_FINISH); } /* Wait for ack */ for (i = 0; i < cell->num_spus; i++) { - uint k = wait_mbox_message(control_ps_area[i]); + uint k = wait_mbox_message(spe_contexts[i]); assert(k == CELL_CMD_FINISH); } } diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 37207c38d7..7d41c660e4 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -27,6 +27,7 @@ #include +#include #include "cell_spu.h" #include "pipe/p_format.h" @@ -43,9 +44,8 @@ helpful headers: /** * SPU/SPE handles, etc */ -speid_t speid[MAX_SPUS]; -spe_spu_control_area_t *control_ps_area[MAX_SPUS]; - +spe_context_ptr_t spe_contexts[MAX_SPUS]; +pthread_t spe_threads[MAX_SPUS]; /** * Data sent to SPUs @@ -58,11 +58,9 @@ struct cell_command command[MAX_SPUS]; * Write a 1-word message to the given SPE mailbox. */ void -send_mbox_message(spe_spu_control_area_t *ca, unsigned int msg) +send_mbox_message(spe_context_ptr_t ctx, unsigned int msg) { - while (_spe_in_mbox_status(ca) < 1) - ; - _spe_in_mbox_write(ca, msg); + spe_in_mbox_write(ctx, &msg, 1, SPE_MBOX_ALL_BLOCKING); } @@ -70,13 +68,30 @@ send_mbox_message(spe_spu_control_area_t *ca, unsigned int msg) * Wait for a 1-word message to arrive in given mailbox. */ uint -wait_mbox_message(spe_spu_control_area_t *ca) +wait_mbox_message(spe_context_ptr_t ctx) +{ + do { + unsigned data; + int count = spe_out_mbox_read(ctx, &data, 1); + + if (count == 1) { + return data; + } + + if (count < 0) { + /* error */ ; + } + } while (1); +} + + +static void *cell_thread_function(void *arg) { - uint k; - while (_spe_out_mbox_status(ca) < 1) - ; - k = _spe_out_mbox_read(ca); - return k; + struct cell_init_info *init = (struct cell_init_info *) arg; + unsigned entry = SPE_DEFAULT_ENTRY; + + spe_context_run(spe_contexts[init->id], &entry, 0, init, NULL, NULL); + pthread_exit(NULL); } @@ -96,22 +111,17 @@ cell_start_spus(uint num_spus) ASSERT_ALIGN16(&inits[0]); ASSERT_ALIGN16(&inits[1]); - /* XXX do we need to create a gid with spe_create_group()? */ - for (i = 0; i < num_spus; i++) { inits[i].id = i; inits[i].num_spus = num_spus; inits[i].cmd = &command[i]; - speid[i] = spe_create_thread(0, /* gid */ - &g3d_spu, /* spe program handle */ - &inits[i], /* argp */ - NULL, /* envp */ - -1, /* mask */ - SPE_MAP_PS/*0*/ ); /* flags */ + spe_contexts[i] = spe_context_create(0, NULL); - control_ps_area[i] = spe_get_ps_area(speid[i], SPE_CONTROL_AREA); - assert(control_ps_area[i]); + spe_program_load(spe_contexts[i], &g3d_spu); + + pthread_create(&spe_threads[i], NULL, cell_thread_function, + & inits[i]); } } @@ -124,14 +134,15 @@ finish_all(uint num_spus) uint i; for (i = 0; i < num_spus; i++) { - send_mbox_message(control_ps_area[i], CELL_CMD_FINISH); + send_mbox_message(spe_contexts[i], CELL_CMD_FINISH); } for (i = 0; i < num_spus; i++) { /* wait for mbox message */ unsigned k; - while (_spe_out_mbox_status(control_ps_area[i]) < 1) + + while (spe_out_mbox_read(spe_contexts[i], &k, 1) < 1) ; - k = _spe_out_mbox_read(control_ps_area[i]); + assert(k == CELL_CMD_FINISH); } } @@ -154,12 +165,12 @@ test_spus(struct cell_context *cell) command[i].fb.width = surf->width; command[i].fb.height = surf->height; command[i].fb.format = PIPE_FORMAT_A8R8G8B8_UNORM; - send_mbox_message(control_ps_area[i], CELL_CMD_FRAMEBUFFER); + send_mbox_message(spe_contexts[i], CELL_CMD_FRAMEBUFFER); } for (i = 0; i < cell->num_spus; i++) { command[i].clear.value = 0xff880044; /* XXX */ - send_mbox_message(control_ps_area[i], CELL_CMD_CLEAR_TILES); + send_mbox_message(spe_contexts[i], CELL_CMD_CLEAR_TILES); } finish_all(cell->num_spus); @@ -171,7 +182,7 @@ test_spus(struct cell_context *cell) } for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(control_ps_area[i], CELL_CMD_EXIT); + send_mbox_message(spe_contexts[i], CELL_CMD_EXIT); } } @@ -182,10 +193,11 @@ test_spus(struct cell_context *cell) void wait_spus(uint num_spus) { - int i, status; + unsigned i; + void *value; for (i = 0; i < num_spus; i++) { - spe_wait( speid[i], &status, 1 ); + pthread_join(spe_threads[i], &value); } } @@ -196,14 +208,11 @@ wait_spus(uint num_spus) void cell_spu_exit(struct cell_context *cell) { - uint i; - int status; + unsigned i; for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(control_ps_area[i], CELL_CMD_EXIT); + send_mbox_message(spe_contexts[i], CELL_CMD_EXIT); } - for (i = 0; i < cell->num_spus; i++) { - spe_wait( speid[i], &status, 1 ); - } + wait_spus(cell->num_spus); } diff --git a/src/mesa/pipe/cell/ppu/cell_spu.h b/src/mesa/pipe/cell/ppu/cell_spu.h index f030be45fc..09d5e23a88 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.h +++ b/src/mesa/pipe/cell/ppu/cell_spu.h @@ -29,7 +29,7 @@ #define CELL_SPU -#include +#include #include #include "pipe/cell/common.h" @@ -42,8 +42,8 @@ * SPU/SPE handles, etc */ extern spe_program_handle_t g3d_spu; -extern speid_t speid[MAX_SPUS]; -extern spe_spu_control_area_t *control_ps_area[MAX_SPUS]; +extern spe_context_ptr_t spe_contexts[MAX_SPUS]; +extern pthread_t spe_threads[MAX_SPUS]; /** @@ -53,11 +53,11 @@ extern struct cell_init_info inits[MAX_SPUS]; extern struct cell_command command[MAX_SPUS]; -void -send_mbox_message(spe_spu_control_area_t *ca, unsigned int msg); +extern void +send_mbox_message(spe_context_ptr_t ctx, unsigned int msg); -uint -wait_mbox_message(spe_spu_control_area_t *ca); +extern uint +wait_mbox_message(spe_context_ptr_t ctx); void diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 59cd43e630..ab20adb06d 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -78,13 +78,13 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, command[i].fb.width = ps->width; command[i].fb.height = ps->height; command[i].fb.format = ps->format; - send_mbox_message(control_ps_area[i], CELL_CMD_FRAMEBUFFER); + send_mbox_message(spe_contexts[i], CELL_CMD_FRAMEBUFFER); } for (i = 0; i < cell->num_spus; i++) { /* XXX clear color varies per-SPU for debugging */ command[i].clear.value = clearValue | (i << 21); - send_mbox_message(control_ps_area[i], CELL_CMD_CLEAR_TILES); + send_mbox_message(spe_contexts[i], CELL_CMD_CLEAR_TILES); } #if 1 @@ -103,7 +103,7 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, /* XXX color varies per SPU */ command[i].tri.color = 0xffff00 | ((i*40)<<24); /* yellow */ - send_mbox_message(control_ps_area[i], CELL_CMD_TRIANGLE); + send_mbox_message(spe_contexts[i], CELL_CMD_TRIANGLE); } #endif } -- cgit v1.2.3 From 457a96e7773fb148136af76a8cf746e72c2ef722 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 13:49:48 -0700 Subject: SPU function prototypes, etc. --- src/mesa/pipe/cell/spu/main.h | 61 +++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 61 insertions(+) create mode 100644 src/mesa/pipe/cell/spu/main.h (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h new file mode 100644 index 0000000000..882be7f1c5 --- /dev/null +++ b/src/mesa/pipe/cell/spu/main.h @@ -0,0 +1,61 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef MAIN_H +#define MAIN_H + + +#include +#include +#include "pipe/cell/common.h" + + +extern struct cell_init_info init; + +struct framebuffer { + void *start; + uint width, height; + uint width_tiles, height_tiles; /**< width and height in tiles */ +}; + +extern struct framebuffer fb; + + +extern int DefaultTag; + + +void +wait_on_mask(unsigned tag); + +void +get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile); + +void +put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile); + + +#endif /* MAIN_H */ -- cgit v1.2.3 From fda387988c24fad4e0a743f16173dc3c71cbe084 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 09:44:04 -0700 Subject: Remove temporary cell_create_surface() --- src/mesa/pipe/cell/ppu/cell_surface.c | 22 ---------------------- src/mesa/pipe/cell/ppu/cell_surface.h | 12 +++++------- 2 files changed, 5 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index ab20adb06d..03b0988dc3 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -41,28 +41,6 @@ #include "cell_spu.h" - -struct pipe_surface * -cell_create_surface(int width, int height) -{ -#if 0 - /* XXX total hack */ - struct pipe_surface *ps = CALLOC_STRUCT(pipe_surface); - - printf("cell_create_surface\n"); - - ps->width = width; - ps->height = height; - - ps->region = CALLOC_STRUCT(pipe_region); - ps->region->map = align_malloc(width * height * 4, 16); - return ps; -#endif - return NULL; -} - - - void cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) diff --git a/src/mesa/pipe/cell/ppu/cell_surface.h b/src/mesa/pipe/cell/ppu/cell_surface.h index e222507121..4c69c4c89f 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.h +++ b/src/mesa/pipe/cell/ppu/cell_surface.h @@ -26,20 +26,18 @@ **************************************************************************/ -#ifndef CELL_SURFACE -#define CELL_SURFACE +#ifndef CELL_SURFACE_H +#define CELL_SURFACE_H -#include "pipe/p_state.h" +struct pipe_context; +struct pipe_surface; -extern struct pipe_surface * -cell_create_surface(int width, int height); - extern void cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue); -#endif /* CELL_SURFACE */ +#endif /* CELL_SURFACE_H */ -- cgit v1.2.3 From 9828310a1bba1c1c2dffa7ae8866b648e26c2039 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 10:28:45 -0700 Subject: Collect some global vars in a single struct. --- src/mesa/pipe/cell/common.h | 1 + src/mesa/pipe/cell/ppu/cell_flush.c | 6 +-- src/mesa/pipe/cell/ppu/cell_spu.c | 85 ++++++++++++++++++----------------- src/mesa/pipe/cell/ppu/cell_spu.h | 29 ++++++++---- src/mesa/pipe/cell/ppu/cell_surface.c | 33 +++++++------- src/mesa/pipe/cell/spu/main.c | 17 ++++--- src/mesa/pipe/cell/spu/main.h | 2 +- 7 files changed, 97 insertions(+), 76 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index a87417743a..c986b59a5d 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -91,6 +91,7 @@ struct cell_command } ALIGN16_ATTRIB; +/** This is the object passed to spe_create_thread() */ struct cell_init_info { unsigned id; diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c index 39b72c435d..b1ff0e51cd 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -26,8 +26,6 @@ **************************************************************************/ -#include - #include "cell_context.h" #include "cell_flush.h" #include "cell_spu.h" @@ -43,12 +41,12 @@ cell_flush(struct pipe_context *pipe, unsigned flags) /* Send CMD_FINISH to all SPUs */ for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(spe_contexts[i], CELL_CMD_FINISH); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FINISH); } /* Wait for ack */ for (i = 0; i < cell->num_spus; i++) { - uint k = wait_mbox_message(spe_contexts[i]); + uint k = wait_mbox_message(cell_global.spe_contexts[i]); assert(k == CELL_CMD_FINISH); } } diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 7d41c660e4..55a0d5038b 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -26,7 +26,6 @@ **************************************************************************/ -#include #include #include "cell_spu.h" @@ -41,17 +40,7 @@ helpful headers: */ -/** - * SPU/SPE handles, etc - */ -spe_context_ptr_t spe_contexts[MAX_SPUS]; -pthread_t spe_threads[MAX_SPUS]; - -/** - * Data sent to SPUs - */ -struct cell_init_info inits[MAX_SPUS]; -struct cell_command command[MAX_SPUS]; +struct cell_global_info cell_global; /** @@ -89,8 +78,15 @@ static void *cell_thread_function(void *arg) { struct cell_init_info *init = (struct cell_init_info *) arg; unsigned entry = SPE_DEFAULT_ENTRY; - - spe_context_run(spe_contexts[init->id], &entry, 0, init, NULL, NULL); + + ASSERT_ALIGN16(init); + + if (spe_context_run(cell_global.spe_contexts[init->id], &entry, 0, + init, NULL, NULL) < 0) { + fprintf(stderr, "spe_context_run() failed\n"); + exit(1); + } + pthread_exit(NULL); } @@ -103,25 +99,32 @@ cell_start_spus(uint num_spus) { uint i; - assert((sizeof(struct cell_command) & 0xf) == 0); - ASSERT_ALIGN16(&command[0]); - ASSERT_ALIGN16(&command[1]); + assert(num_spus <= MAX_SPUS); - assert((sizeof(struct cell_init_info) & 0xf) == 0); - ASSERT_ALIGN16(&inits[0]); - ASSERT_ALIGN16(&inits[1]); + ASSERT_ALIGN16(&cell_global.command[0]); + ASSERT_ALIGN16(&cell_global.command[1]); - for (i = 0; i < num_spus; i++) { - inits[i].id = i; - inits[i].num_spus = num_spus; - inits[i].cmd = &command[i]; + ASSERT_ALIGN16(&cell_global.inits[0]); + ASSERT_ALIGN16(&cell_global.inits[1]); - spe_contexts[i] = spe_context_create(0, NULL); + for (i = 0; i < num_spus; i++) { + cell_global.inits[i].id = i; + cell_global.inits[i].num_spus = num_spus; + cell_global.inits[i].cmd = &cell_global.command[i]; + + cell_global.spe_contexts[i] = spe_context_create(0, NULL); + if (!cell_global.spe_contexts[i]) { + fprintf(stderr, "spe_context_create() failed\n"); + exit(1); + } - spe_program_load(spe_contexts[i], &g3d_spu); + if (spe_program_load(cell_global.spe_contexts[i], &g3d_spu)) { + fprintf(stderr, "spe_program_load() failed\n"); + exit(1); + } - pthread_create(&spe_threads[i], NULL, cell_thread_function, - & inits[i]); + pthread_create(&cell_global.spe_threads[i], NULL, &cell_thread_function, + &cell_global.inits[i]); } } @@ -134,13 +137,13 @@ finish_all(uint num_spus) uint i; for (i = 0; i < num_spus; i++) { - send_mbox_message(spe_contexts[i], CELL_CMD_FINISH); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FINISH); } for (i = 0; i < num_spus; i++) { /* wait for mbox message */ unsigned k; - while (spe_out_mbox_read(spe_contexts[i], &k, 1) < 1) + while (spe_out_mbox_read(cell_global.spe_contexts[i], &k, 1) < 1) ; assert(k == CELL_CMD_FINISH); @@ -161,16 +164,16 @@ test_spus(struct cell_context *cell) sleep(2); for (i = 0; i < cell->num_spus; i++) { - command[i].fb.start = surf->map; - command[i].fb.width = surf->width; - command[i].fb.height = surf->height; - command[i].fb.format = PIPE_FORMAT_A8R8G8B8_UNORM; - send_mbox_message(spe_contexts[i], CELL_CMD_FRAMEBUFFER); + cell_global.command[i].fb.start = surf->map; + cell_global.command[i].fb.width = surf->width; + cell_global.command[i].fb.height = surf->height; + cell_global.command[i].fb.format = PIPE_FORMAT_A8R8G8B8_UNORM; + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FRAMEBUFFER); } for (i = 0; i < cell->num_spus; i++) { - command[i].clear.value = 0xff880044; /* XXX */ - send_mbox_message(spe_contexts[i], CELL_CMD_CLEAR_TILES); + cell_global.command[i].clear.value = 0xff880044; /* XXX */ + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_TILES); } finish_all(cell->num_spus); @@ -182,7 +185,7 @@ test_spus(struct cell_context *cell) } for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(spe_contexts[i], CELL_CMD_EXIT); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_EXIT); } } @@ -193,11 +196,11 @@ test_spus(struct cell_context *cell) void wait_spus(uint num_spus) { - unsigned i; + uint i; void *value; for (i = 0; i < num_spus; i++) { - pthread_join(spe_threads[i], &value); + pthread_join(cell_global.spe_threads[i], &value); } } @@ -211,7 +214,7 @@ cell_spu_exit(struct cell_context *cell) unsigned i; for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(spe_contexts[i], CELL_CMD_EXIT); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_EXIT); } wait_spus(cell->num_spus); diff --git a/src/mesa/pipe/cell/ppu/cell_spu.h b/src/mesa/pipe/cell/ppu/cell_spu.h index 09d5e23a88..612cb45c59 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.h +++ b/src/mesa/pipe/cell/ppu/cell_spu.h @@ -39,18 +39,29 @@ #define MAX_SPUS 8 /** - * SPU/SPE handles, etc + * Global vars, for now anyway. */ -extern spe_program_handle_t g3d_spu; -extern spe_context_ptr_t spe_contexts[MAX_SPUS]; -extern pthread_t spe_threads[MAX_SPUS]; +struct cell_global_info +{ + /** + * SPU/SPE handles, etc + */ + spe_context_ptr_t spe_contexts[MAX_SPUS]; + pthread_t spe_threads[MAX_SPUS]; + /** + * Data sent to SPUs + */ + struct cell_init_info inits[MAX_SPUS]; + struct cell_command command[MAX_SPUS]; +}; -/** - * Data sent to SPUs - */ -extern struct cell_init_info inits[MAX_SPUS]; -extern struct cell_command command[MAX_SPUS]; + +extern struct cell_global_info cell_global; + + +/** This is the handle for the actual SPE code */ +extern spe_program_handle_t g3d_spu; extern void diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 03b0988dc3..143f554b60 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -52,36 +52,39 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, pipe_surface_map(ps); for (i = 0; i < cell->num_spus; i++) { - command[i].fb.start = ps->map; - command[i].fb.width = ps->width; - command[i].fb.height = ps->height; - command[i].fb.format = ps->format; - send_mbox_message(spe_contexts[i], CELL_CMD_FRAMEBUFFER); + struct cell_command_framebuffer *fb = &cell_global.command[i].fb; + fb->start = ps->map; + fb->width = ps->width; + fb->height = ps->height; + fb->format = ps->format; + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FRAMEBUFFER); } for (i = 0; i < cell->num_spus; i++) { /* XXX clear color varies per-SPU for debugging */ - command[i].clear.value = clearValue | (i << 21); - send_mbox_message(spe_contexts[i], CELL_CMD_CLEAR_TILES); + cell_global.command[i].clear.value = clearValue | (i << 21); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_TILES); } #if 1 /* XXX Draw a test triangle over the cleared surface */ for (i = 0; i < cell->num_spus; i++) { /* Same triangle data for all SPUs, of course: */ - command[i].tri.x0 = 20.0; - command[i].tri.y0 = ps->height - 20; + struct cell_command_triangle *tri = &cell_global.command[i].tri; - command[i].tri.x1 = ps->width - 20.0; - command[i].tri.y1 = ps->height - 20; + tri->x0 = 20.0; + tri->y0 = ps->height - 20; - command[i].tri.x2 = ps->width / 2; - command[i].tri.y2 = 20.0; + tri->x1 = ps->width - 20.0; + tri->y1 = ps->height - 20; + + tri->x2 = ps->width / 2; + tri->y2 = 20.0; /* XXX color varies per SPU */ - command[i].tri.color = 0xffff00 | ((i*40)<<24); /* yellow */ + tri->color = 0xffff00 | ((i*40)<<24); /* yellow */ - send_mbox_message(spe_contexts[i], CELL_CMD_TRIANGLE); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_TRIANGLE); } #endif } diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 83880bc906..6a77e3e9fa 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -44,7 +44,7 @@ helpful headers: /opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h */ -struct cell_init_info init; +volatile struct cell_init_info init; struct framebuffer fb; @@ -238,17 +238,21 @@ main_loop(void) +/** + * SPE entrypoint. + * Note: example programs declare params as 'unsigned long long' but + * that doesn't work. + */ int -main(unsigned long long speid, - unsigned long long argp, - unsigned long long envp) +main(unsigned long speid, unsigned long argp) { int tag = 0; + (void) speid; + DefaultTag = 1; - (void) speid; - (void) envp; + printf("SPU: main() speid=%lu\n", speid); mfc_get(&init, /* dest */ (unsigned int) argp, /* src */ @@ -258,6 +262,7 @@ main(unsigned long long speid, 0 /* rid */); wait_on_mask( 1 << tag ); + main_loop(); return 0; diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index 882be7f1c5..8c2796387f 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -34,7 +34,7 @@ #include "pipe/cell/common.h" -extern struct cell_init_info init; +extern volatile struct cell_init_info init; struct framebuffer { void *start; -- cgit v1.2.3 From b2ad30d57197c2167789e4f3f5b34af6df56dde2 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 18:26:52 -0700 Subject: Cell: draw smooth-shaded triangle --- src/mesa/pipe/cell/common.h | 6 +- src/mesa/pipe/cell/ppu/cell_surface.c | 29 +++++--- src/mesa/pipe/cell/spu/main.c | 20 +++--- src/mesa/pipe/cell/spu/tri.c | 125 ++++++++++++++++++++++++---------- 4 files changed, 120 insertions(+), 60 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index c986b59a5d..f7f1e2eb41 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -75,10 +75,8 @@ struct cell_command_clear_tiles struct cell_command_triangle { - float x0, y0; - float x1, y1; - float x2, y2; - uint color; + float vert[3][4]; + float color[3][4]; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 143f554b60..185eeb26e8 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -69,20 +69,31 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, #if 1 /* XXX Draw a test triangle over the cleared surface */ for (i = 0; i < cell->num_spus; i++) { - /* Same triangle data for all SPUs, of course: */ + /* Same triangle data for all SPUs */ struct cell_command_triangle *tri = &cell_global.command[i].tri; + tri->vert[0][0] = 20.0; + tri->vert[0][1] = ps->height - 20; - tri->x0 = 20.0; - tri->y0 = ps->height - 20; + tri->vert[1][0] = ps->width - 20.0; + tri->vert[1][1] = ps->height - 20; - tri->x1 = ps->width - 20.0; - tri->y1 = ps->height - 20; + tri->vert[2][0] = ps->width / 2; + tri->vert[2][1] = 20.0; - tri->x2 = ps->width / 2; - tri->y2 = 20.0; + tri->color[0][0] = 1.0; + tri->color[0][1] = 0.0; + tri->color[0][2] = 0.0; + tri->color[0][3] = 0.0; - /* XXX color varies per SPU */ - tri->color = 0xffff00 | ((i*40)<<24); /* yellow */ + tri->color[1][0] = 0.0; + tri->color[1][1] = 1.0; + tri->color[1][2] = 0.0; + tri->color[1][3] = 0.0; + + tri->color[2][0] = 0.0; + tri->color[2][1] = 0.0; + tri->color[2][2] = 1.0; + tri->color[2][3] = 0.0; send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_TRIANGLE); } diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 6a77e3e9fa..cc5eddb0f5 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -138,13 +138,13 @@ triangle(const struct cell_command_triangle *tri) struct prim_header prim; uint i; - prim.v[0].data[0][0] = tri->x0; - prim.v[0].data[0][1] = tri->y0; - prim.v[1].data[0][0] = tri->x1; - prim.v[1].data[0][1] = tri->y1; - prim.v[2].data[0][0] = tri->x2; - prim.v[2].data[0][1] = tri->y2; - prim.color = tri->color; + COPY_4V(prim.v[0].data[0], tri->vert[0]); + COPY_4V(prim.v[1].data[0], tri->vert[1]); + COPY_4V(prim.v[2].data[0], tri->vert[2]); + + COPY_4V(prim.v[0].data[1], tri->color[0]); + COPY_4V(prim.v[1].data[1], tri->color[1]); + COPY_4V(prim.v[2].data[1], tri->color[2]); for (i = init.id; i < num_tiles; i += init.num_spus) { uint tx = i % fb.width_tiles; @@ -212,11 +212,7 @@ main_loop(void) clear_tiles(&cmd.clear); break; case CELL_CMD_TRIANGLE: - printf("SPU %u: TRIANGLE (%g,%g) (%g,%g) (%g,%g)\n", - init.id, - cmd.tri.x0, cmd.tri.y0, - cmd.tri.x1, cmd.tri.y1, - cmd.tri.x2, cmd.tri.y2); + printf("SPU %u: TRIANGLE\n", init.id); triangle(&cmd.tri); break; case CELL_CMD_FINISH: diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 60a62fdcd5..90452f14ba 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -46,6 +46,11 @@ #include "main.h" #include "tri.h" +/* +#include +#include +*/ + #if 1 @@ -64,6 +69,8 @@ #define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) #define MASK_ALL 0xf +#define PIPE_MAX_SHADER_INPUTS 8 /* XXX temp */ + static int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; static uint tile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; @@ -85,6 +92,13 @@ struct edge { }; +struct interp_coef +{ + float a0[4]; + float dadx[4]; + float dady[4]; +}; + /** * Triangle setup info (derived from draw_stage). * Also used for line drawing (taking some liberties). @@ -113,7 +127,10 @@ struct setup_stage { #if 0 struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; +#else + struct interp_coef coef[PIPE_MAX_SHADER_INPUTS]; #endif + #if 0 struct quad_header quad; #endif @@ -188,6 +205,41 @@ clip_emit_quad(struct setup_stage *setup) } #endif +/** + * Evaluate attribute coefficients (plane equations) to compute + * attribute values for the four fragments in a quad. + * Eg: four colors will be compute. + */ +static INLINE void +eval_coeff( struct setup_stage *setup, uint slot, + float x, float y, float result[4][4]) +{ + uint i; + const float *dadx = setup->coef[slot].dadx; + const float *dady = setup->coef[slot].dady; + + /* loop over XYZW comps */ + for (i = 0; i < 4; i++) { + result[QUAD_TOP_LEFT][i] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; + result[QUAD_TOP_RIGHT][i] = result[0][i] + dadx[i]; + result[QUAD_BOTTOM_LEFT][i] = result[0][i] + dady[i]; + result[QUAD_BOTTOM_RIGHT][i] = result[0][i] + dadx[i] + dady[i]; + } +} + + +static INLINE uint +pack_color(const float color[4]) +{ + uint r = (uint) (color[0] * 255.0); + uint g = (uint) (color[1] * 255.0); + uint b = (uint) (color[2] * 255.0); + uint a = (uint) (color[3] * 255.0); + uint icolor = (b << 24) | (g << 16) | (r << 8) | a; + return icolor; +} + + /** * Emit a quad (pass to next stage). No clipping is done. */ @@ -204,14 +256,18 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) /* Cell: "write" quad fragments to the tile by setting prim color */ int ix = x - cliprect_minx; int iy = y - cliprect_miny; + float colors[4][4]; + + eval_coeff(setup, 1, (float) x, (float) y, colors); + if (mask & MASK_TOP_LEFT) - tile[iy][ix] = setup->color; + tile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); if (mask & MASK_TOP_RIGHT) - tile[iy][ix+1] = setup->color; + tile[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); if (mask & MASK_BOTTOM_LEFT) - tile[iy+1][ix] = setup->color; + tile[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); if (mask & MASK_BOTTOM_RIGHT) - tile[iy+1][ix+1] = setup->color; + tile[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); #endif } @@ -445,41 +501,41 @@ static void const_coeff( struct setup_stage *setup, #endif -#if 0 /** * Compute a0, dadx and dady for a linearly interpolated coefficient, * for a triangle. */ static void tri_linear_coeff( struct setup_stage *setup, - unsigned slot, - unsigned i) + unsigned slot ) { - float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; - float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + uint i; + for (i = 0; i < 4; i++) { + float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; + float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - assert(slot < PIPE_MAX_SHADER_INPUTS); - assert(i <= 3); + assert(slot < PIPE_MAX_SHADER_INPUTS); - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; - /* calculate a0 as the value which would be sampled for the - * fragment at (0,0), taking into account that we want to sample at - * pixel centers, in other words (0.5, 0.5). - * - * this is neat but unfortunately not a good way to do things for - * triangles with very large values of dadx or dady as it will - * result in the subtraction and re-addition from a0 of a very - * large number, which means we'll end up loosing a lot of the - * fractional bits and precision from a0. the way to fix this is - * to define a0 as the sample at a pixel center somewhere near vmin - * instead - i'll switch to this later. - */ - setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); + } /* _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", @@ -489,7 +545,6 @@ static void tri_linear_coeff( struct setup_stage *setup, setup->coef[slot].dady[i]); */ } -#endif #if 0 @@ -536,13 +591,13 @@ static void tri_persp_coeff( struct setup_stage *setup, #endif -#if 0 /** * Compute the setup->coef[] array dadx, dady, a0 values. * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. */ static void setup_tri_coefficients( struct setup_stage *setup ) { +#if 0 const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; unsigned slot, j; @@ -575,8 +630,10 @@ static void setup_tri_coefficients( struct setup_stage *setup ) assert(0); } } -} +#else + tri_linear_coeff(setup, 1); /* slot 1 = color */ #endif +} static void setup_tri_edges( struct setup_stage *setup ) @@ -710,9 +767,7 @@ static void setup_tri( */ setup_sort_vertices( setup, prim ); -#if 0 setup_tri_coefficients( setup ); -#endif setup_tri_edges( setup ); #if 0 -- cgit v1.2.3 From 3d9bdaa80b9faaefcb896580717f4854e034410b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 18:58:32 -0700 Subject: include pipe/tgsi/exec/tgsi_sse2.h if needed. Silences warning. --- src/mesa/pipe/draw/draw_vertex_shader.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 5294d38f47..d34d923018 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -33,12 +33,14 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" +#if defined(__i386__) || defined(__386__) +#include "pipe/tgsi/exec/tgsi_sse2.h" +#endif #include "draw_private.h" #include "draw_context.h" #include "draw_vertex.h" #include "x86/rtasm/x86sse.h" - #include "pipe/llvm/gallivm.h" -- cgit v1.2.3 From 68caa076a94c28c1c9e62b9831d5f31c1bb008ce Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 18:58:48 -0700 Subject: trim down -I options --- src/mesa/sources | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index fb059498f9..31aeb5b451 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -427,10 +427,12 @@ INCLUDE_DIRS = \ -I$(TOP)/src/mesa/main \ -I$(TOP)/src/mesa/glapi \ -I$(TOP)/src/mesa/math \ - -I$(TOP)/src/mesa/tnl \ -I$(TOP)/src/mesa/shader \ -I$(TOP)/src/mesa/shader/grammar \ - -I$(TOP)/src/mesa/shader/slang \ + -I$(TOP)/src/mesa/shader/slang + +OLD_INCLUDE_DIRS = \ + -I$(TOP)/src/mesa/tnl \ -I$(TOP)/src/mesa/swrast \ -I$(TOP)/src/mesa/swrast_setup \ -I$(TOP)/src/mesa/pipe/tgsi -- cgit v1.2.3 From 6f467e57376e94a6da0e3d363ecaecc52ca8a3c8 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 19:16:26 -0700 Subject: trim down #includes --- src/mesa/pipe/xlib/glxapi.c | 4 ++-- src/mesa/pipe/xlib/xm_api.c | 12 +++--------- 2 files changed, 5 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/glxapi.c b/src/mesa/pipe/xlib/glxapi.c index 309a0008d7..c2ccce6f52 100644 --- a/src/mesa/pipe/xlib/glxapi.c +++ b/src/mesa/pipe/xlib/glxapi.c @@ -34,8 +34,8 @@ #include #include #include -#include "glheader.h" -#include "glapi.h" +#include "main/glheader.h" +#include "glapi/glapi.h" #include "glxapi.h" diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 83e2b529a4..4021af187c 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -60,15 +60,9 @@ #include "glxheader.h" #include "GL/xmesa.h" #include "xmesaP.h" -#include "context.h" -#include "extensions.h" -#include "framebuffer.h" -#include "glthread.h" -#include "imports.h" -#include "macros.h" -#include "renderbuffer.h" -#include "teximage.h" -#include "vbo/vbo.h" +#include "main/context.h" +#include "main/framebuffer.h" +#include "glapi/glthread.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" -- cgit v1.2.3 From 5dcc894ea2d926d4845e805298b79c514f58687d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 19:16:46 -0700 Subject: prefix some #includes --- src/mesa/main/dispatch.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/main/dispatch.c b/src/mesa/main/dispatch.c index 05c1c36c0d..c12f55a7a1 100644 --- a/src/mesa/main/dispatch.c +++ b/src/mesa/main/dispatch.c @@ -40,9 +40,9 @@ #ifndef GLX_USE_APPLEGL #include "glheader.h" -#include "glapi.h" -#include "glapitable.h" -#include "glthread.h" +#include "glapi/glapi.h" +#include "glapi/glapitable.h" +#include "glapi/glthread.h" #if !(defined(USE_X86_ASM) || defined(USE_X86_64_ASM) || defined(USE_SPARC_ASM)) @@ -87,8 +87,8 @@ #define GLAPIENTRY #endif -#include "dispatch.h" -#include "glapitemp.h" +#include "glapi/dispatch.h" +#include "glapi/glapitemp.h" #endif /* USE_X86_ASM */ -- cgit v1.2.3 From 1029c00a269d7f98ed659bb48727a17ef5dea9e5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Dec 2007 19:17:31 -0700 Subject: remove more -I paths --- src/mesa/sources | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 31aeb5b451..f889fb30c1 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -424,15 +424,14 @@ FBDEV_DRIVER_OBJECTS = $(FBDEV_DRIVER_SOURCES:.c=.o) INCLUDE_DIRS = \ -I$(TOP)/include \ -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/main \ - -I$(TOP)/src/mesa/glapi \ - -I$(TOP)/src/mesa/math \ - -I$(TOP)/src/mesa/shader \ - -I$(TOP)/src/mesa/shader/grammar \ - -I$(TOP)/src/mesa/shader/slang + -I$(TOP)/src/mesa/main OLD_INCLUDE_DIRS = \ -I$(TOP)/src/mesa/tnl \ + -I$(TOP)/src/mesa/math \ -I$(TOP)/src/mesa/swrast \ -I$(TOP)/src/mesa/swrast_setup \ + -I$(TOP)/src/mesa/shader \ + -I$(TOP)/src/mesa/shader/grammar \ + -I$(TOP)/src/mesa/shader/slang \ -I$(TOP)/src/mesa/pipe/tgsi -- cgit v1.2.3 From 94e19777c9db476e930147346fe9217bfc879b1e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 11 Dec 2007 17:46:06 -0500 Subject: i915tex: remove. it's deprecated and causes merge problems we did some small changes in the beginning of the gallium3d lifecycle in i915tex which is not in master anymore and just causes problems when doing merges. getting rid of the headache by just nuking it here --- src/mesa/drivers/dri/i915tex/Makefile | 71 -- src/mesa/drivers/dri/i915tex/i830_context.c | 104 -- src/mesa/drivers/dri/i915tex/i830_context.h | 213 ---- src/mesa/drivers/dri/i915tex/i830_metaops.c | 457 ------- src/mesa/drivers/dri/i915tex/i830_reg.h | 642 ---------- src/mesa/drivers/dri/i915tex/i830_state.c | 1106 ----------------- src/mesa/drivers/dri/i915tex/i830_tex.c | 100 -- src/mesa/drivers/dri/i915tex/i830_texblend.c | 463 ------- src/mesa/drivers/dri/i915tex/i830_texstate.c | 343 ----- src/mesa/drivers/dri/i915tex/i830_vtbl.c | 649 ---------- src/mesa/drivers/dri/i915tex/i915_context.c | 176 --- src/mesa/drivers/dri/i915tex/i915_context.h | 360 ------ src/mesa/drivers/dri/i915tex/i915_debug.c | 334 ----- src/mesa/drivers/dri/i915tex/i915_fragprog.c | 1073 ---------------- src/mesa/drivers/dri/i915tex/i915_metaops.c | 509 -------- src/mesa/drivers/dri/i915tex/i915_program.c | 515 -------- src/mesa/drivers/dri/i915tex/i915_program.h | 160 --- src/mesa/drivers/dri/i915tex/i915_reg.h | 841 ------------- src/mesa/drivers/dri/i915tex/i915_state.c | 968 --------------- src/mesa/drivers/dri/i915tex/i915_tex.c | 78 -- src/mesa/drivers/dri/i915tex/i915_tex_layout.c | 343 ----- src/mesa/drivers/dri/i915tex/i915_texstate.c | 359 ------ src/mesa/drivers/dri/i915tex/i915_vtbl.c | 562 --------- src/mesa/drivers/dri/i915tex/intel_batchbuffer.c | 340 ----- src/mesa/drivers/dri/i915tex/intel_batchbuffer.h | 124 -- src/mesa/drivers/dri/i915tex/intel_batchpool.c | 418 ------- src/mesa/drivers/dri/i915tex/intel_blit.c | 527 -------- src/mesa/drivers/dri/i915tex/intel_blit.h | 62 - .../drivers/dri/i915tex/intel_buffer_objects.c | 257 ---- .../drivers/dri/i915tex/intel_buffer_objects.h | 86 -- src/mesa/drivers/dri/i915tex/intel_buffers.c | 890 ------------- src/mesa/drivers/dri/i915tex/intel_buffers.h | 55 - src/mesa/drivers/dri/i915tex/intel_context.c | 782 ------------ src/mesa/drivers/dri/i915tex/intel_context.h | 487 -------- src/mesa/drivers/dri/i915tex/intel_depthstencil.c | 282 ----- src/mesa/drivers/dri/i915tex/intel_depthstencil.h | 14 - src/mesa/drivers/dri/i915tex/intel_fbo.c | 621 ---------- src/mesa/drivers/dri/i915tex/intel_fbo.h | 127 -- src/mesa/drivers/dri/i915tex/intel_ioctl.c | 137 -- src/mesa/drivers/dri/i915tex/intel_ioctl.h | 40 - src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c | 357 ------ src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h | 198 --- src/mesa/drivers/dri/i915tex/intel_pixel.c | 119 -- src/mesa/drivers/dri/i915tex/intel_pixel.h | 63 - src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c | 351 ------ src/mesa/drivers/dri/i915tex/intel_pixel_copy.c | 358 ------ src/mesa/drivers/dri/i915tex/intel_pixel_draw.c | 431 ------- src/mesa/drivers/dri/i915tex/intel_pixel_read.c | 327 ----- src/mesa/drivers/dri/i915tex/intel_reg.h | 88 -- src/mesa/drivers/dri/i915tex/intel_regions.c | 480 ------- src/mesa/drivers/dri/i915tex/intel_regions.h | 141 --- src/mesa/drivers/dri/i915tex/intel_render.c | 247 ---- src/mesa/drivers/dri/i915tex/intel_rotate.c | 237 ---- src/mesa/drivers/dri/i915tex/intel_rotate.h | 39 - src/mesa/drivers/dri/i915tex/intel_screen.c | 755 ----------- src/mesa/drivers/dri/i915tex/intel_screen.h | 123 -- src/mesa/drivers/dri/i915tex/intel_span.c | 401 ------ src/mesa/drivers/dri/i915tex/intel_span.h | 38 - src/mesa/drivers/dri/i915tex/intel_state.c | 261 ---- src/mesa/drivers/dri/i915tex/intel_structs.h | 132 -- src/mesa/drivers/dri/i915tex/intel_surface.c | 250 ---- src/mesa/drivers/dri/i915tex/intel_tex.c | 192 --- src/mesa/drivers/dri/i915tex/intel_tex.h | 151 --- src/mesa/drivers/dri/i915tex/intel_tex_copy.c | 293 ----- src/mesa/drivers/dri/i915tex/intel_tex_format.c | 172 --- src/mesa/drivers/dri/i915tex/intel_tex_image.c | 691 ----------- src/mesa/drivers/dri/i915tex/intel_tex_layout.c | 1 - src/mesa/drivers/dri/i915tex/intel_tex_subimage.c | 182 --- src/mesa/drivers/dri/i915tex/intel_tex_validate.c | 272 ---- src/mesa/drivers/dri/i915tex/intel_tris.c | 1164 ----------------- src/mesa/drivers/dri/i915tex/intel_tris.h | 69 -- src/mesa/drivers/dri/i915tex/server/i830_common.h | 226 ---- src/mesa/drivers/dri/i915tex/server/i830_dri.h | 63 - src/mesa/drivers/dri/i915tex/server/intel.h | 331 ----- src/mesa/drivers/dri/i915tex/server/intel_dri.c | 1306 -------------------- 75 files changed, 26184 deletions(-) delete mode 100644 src/mesa/drivers/dri/i915tex/Makefile delete mode 100644 src/mesa/drivers/dri/i915tex/i830_context.c delete mode 100644 src/mesa/drivers/dri/i915tex/i830_context.h delete mode 100644 src/mesa/drivers/dri/i915tex/i830_metaops.c delete mode 100644 src/mesa/drivers/dri/i915tex/i830_reg.h delete mode 100644 src/mesa/drivers/dri/i915tex/i830_state.c delete mode 100644 src/mesa/drivers/dri/i915tex/i830_tex.c delete mode 100644 src/mesa/drivers/dri/i915tex/i830_texblend.c delete mode 100644 src/mesa/drivers/dri/i915tex/i830_texstate.c delete mode 100644 src/mesa/drivers/dri/i915tex/i830_vtbl.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_context.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_context.h delete mode 100644 src/mesa/drivers/dri/i915tex/i915_debug.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_fragprog.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_metaops.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_program.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_program.h delete mode 100644 src/mesa/drivers/dri/i915tex/i915_reg.h delete mode 100644 src/mesa/drivers/dri/i915tex/i915_state.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_tex.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_tex_layout.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_texstate.c delete mode 100644 src/mesa/drivers/dri/i915tex/i915_vtbl.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_batchbuffer.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_batchbuffer.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_batchpool.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_blit.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_blit.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_buffer_objects.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_buffer_objects.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_buffers.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_buffers.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_context.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_context.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_depthstencil.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_depthstencil.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_fbo.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_fbo.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_ioctl.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_ioctl.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_pixel.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_pixel.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_pixel_copy.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_pixel_draw.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_pixel_read.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_reg.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_regions.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_regions.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_render.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_rotate.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_rotate.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_screen.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_screen.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_span.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_span.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_state.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_structs.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_surface.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_tex.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_tex.h delete mode 100644 src/mesa/drivers/dri/i915tex/intel_tex_copy.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_tex_format.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_tex_image.c delete mode 120000 src/mesa/drivers/dri/i915tex/intel_tex_layout.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_tex_subimage.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_tex_validate.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_tris.c delete mode 100644 src/mesa/drivers/dri/i915tex/intel_tris.h delete mode 100644 src/mesa/drivers/dri/i915tex/server/i830_common.h delete mode 100644 src/mesa/drivers/dri/i915tex/server/i830_dri.h delete mode 100644 src/mesa/drivers/dri/i915tex/server/intel.h delete mode 100644 src/mesa/drivers/dri/i915tex/server/intel_dri.c (limited to 'src') diff --git a/src/mesa/drivers/dri/i915tex/Makefile b/src/mesa/drivers/dri/i915tex/Makefile deleted file mode 100644 index 827acc0c46..0000000000 --- a/src/mesa/drivers/dri/i915tex/Makefile +++ /dev/null @@ -1,71 +0,0 @@ - -TOP = ../../../../.. -include $(TOP)/configs/current - -LIBNAME = i915tex_dri.so - -MINIGLX_SOURCES = server/intel_dri.c - -DRIVER_SOURCES = \ - i830_context.c \ - i830_metaops.c \ - i830_state.c \ - i830_texblend.c \ - i830_tex.c \ - i830_texstate.c \ - i830_vtbl.c \ - intel_render.c \ - intel_regions.c \ - intel_buffer_objects.c \ - intel_batchbuffer.c \ - intel_mipmap_tree.c \ - i915_tex_layout.c \ - intel_tex_layout.c \ - intel_tex_image.c \ - intel_tex_subimage.c \ - intel_tex_copy.c \ - intel_tex_validate.c \ - intel_tex_format.c \ - intel_tex.c \ - intel_pixel.c \ - intel_pixel_copy.c \ - intel_pixel_read.c \ - intel_pixel_draw.c \ - intel_buffers.c \ - intel_blit.c \ - i915_tex.c \ - i915_texstate.c \ - i915_context.c \ - i915_debug.c \ - i915_fragprog.c \ - i915_metaops.c \ - i915_program.c \ - i915_state.c \ - i915_vtbl.c \ - intel_context.c \ - intel_ioctl.c \ - intel_rotate.c \ - intel_screen.c \ - intel_span.c \ - intel_state.c \ - intel_surface.c \ - intel_tris.c \ - intel_fbo.c \ - intel_depthstencil.c \ - intel_batchpool.c - -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(COMMON_BM_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") - -include ../Makefile.template - -intel_tex_layout.o: ../intel/intel_tex_layout.c - -symlinks: diff --git a/src/mesa/drivers/dri/i915tex/i830_context.c b/src/mesa/drivers/dri/i915tex/i830_context.c deleted file mode 100644 index 2ff8621c42..0000000000 --- a/src/mesa/drivers/dri/i915tex/i830_context.c +++ /dev/null @@ -1,104 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i830_context.h" -#include "imports.h" -#include "texmem.h" -#include "intel_tex.h" -#include "tnl/tnl.h" -#include "tnl/t_vertex.h" -#include "tnl/t_context.h" -#include "utils.h" - -/*************************************** - * Mesa's Driver Functions - ***************************************/ - -static const struct dri_extension i830_extensions[] = { - {"GL_ARB_texture_env_crossbar", NULL}, - {NULL, NULL} -}; - - -static void -i830InitDriverFunctions(struct dd_function_table *functions) -{ - intelInitDriverFunctions(functions); - i830InitStateFuncs(functions); - i830InitTextureFuncs(functions); -} - - -GLboolean -i830CreateContext(const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate) -{ - struct dd_function_table functions; - struct i830_context *i830 = CALLOC_STRUCT(i830_context); - struct intel_context *intel = &i830->intel; - GLcontext *ctx = &intel->ctx; - if (!i830) - return GL_FALSE; - - i830InitVtbl(i830); - i830InitDriverFunctions(&functions); - - if (!intelInitContext(intel, mesaVis, driContextPriv, - sharedContextPrivate, &functions)) { - FREE(i830); - return GL_FALSE; - } - - intel->ctx.Const.MaxTextureUnits = I830_TEX_UNITS; - intel->ctx.Const.MaxTextureImageUnits = I830_TEX_UNITS; - intel->ctx.Const.MaxTextureCoordUnits = I830_TEX_UNITS; - - /* Advertise the full hardware capabilities. The new memory - * manager should cope much better with overload situations: - */ - ctx->Const.MaxTextureLevels = 12; - ctx->Const.Max3DTextureLevels = 9; - ctx->Const.MaxCubeTextureLevels = 11; - ctx->Const.MaxTextureRectSize = (1 << 11); - ctx->Const.MaxTextureUnits = I830_TEX_UNITS; - - _tnl_init_vertices(ctx, ctx->Const.MaxArrayLockSize + 12, - 18 * sizeof(GLfloat)); - - intel->verts = TNL_CONTEXT(ctx)->clipspace.vertex_buf; - - driInitExtensions(ctx, i830_extensions, GL_FALSE); - - i830InitState(i830); - i830InitMetaFuncs(i830); - - _tnl_allow_vertex_fog(ctx, 1); - _tnl_allow_pixel_fog(ctx, 0); - - return GL_TRUE; -} diff --git a/src/mesa/drivers/dri/i915tex/i830_context.h b/src/mesa/drivers/dri/i915tex/i830_context.h deleted file mode 100644 index 3d754103c0..0000000000 --- a/src/mesa/drivers/dri/i915tex/i830_context.h +++ /dev/null @@ -1,213 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef I830CONTEXT_INC -#define I830CONTEXT_INC - -#include "intel_context.h" - -#define I830_FALLBACK_TEXTURE 0x1000 -#define I830_FALLBACK_COLORMASK 0x2000 -#define I830_FALLBACK_STENCIL 0x4000 -#define I830_FALLBACK_STIPPLE 0x8000 -#define I830_FALLBACK_LOGICOP 0x10000 - -#define I830_UPLOAD_CTX 0x1 -#define I830_UPLOAD_BUFFERS 0x2 -#define I830_UPLOAD_STIPPLE 0x4 -#define I830_UPLOAD_INVARIENT 0x8 -#define I830_UPLOAD_TEX(i) (0x10<<(i)) -#define I830_UPLOAD_TEXBLEND(i) (0x100<<(i)) -#define I830_UPLOAD_TEX_ALL (0x0f0) -#define I830_UPLOAD_TEXBLEND_ALL (0xf00) - -/* State structure offsets - these will probably disappear. - */ -#define I830_DESTREG_CBUFADDR0 0 -#define I830_DESTREG_CBUFADDR1 1 -#define I830_DESTREG_DBUFADDR0 2 -#define I830_DESTREG_DBUFADDR1 3 -#define I830_DESTREG_DV0 4 -#define I830_DESTREG_DV1 5 -#define I830_DESTREG_SENABLE 6 -#define I830_DESTREG_SR0 7 -#define I830_DESTREG_SR1 8 -#define I830_DESTREG_SR2 9 -#define I830_DEST_SETUP_SIZE 10 - -#define I830_CTXREG_STATE1 0 -#define I830_CTXREG_STATE2 1 -#define I830_CTXREG_STATE3 2 -#define I830_CTXREG_STATE4 3 -#define I830_CTXREG_STATE5 4 -#define I830_CTXREG_IALPHAB 5 -#define I830_CTXREG_STENCILTST 6 -#define I830_CTXREG_ENABLES_1 7 -#define I830_CTXREG_ENABLES_2 8 -#define I830_CTXREG_AA 9 -#define I830_CTXREG_FOGCOLOR 10 -#define I830_CTXREG_BLENDCOLOR0 11 -#define I830_CTXREG_BLENDCOLOR1 12 -#define I830_CTXREG_VF 13 -#define I830_CTXREG_VF2 14 -#define I830_CTXREG_MCSB0 15 -#define I830_CTXREG_MCSB1 16 -#define I830_CTX_SETUP_SIZE 17 - -#define I830_STPREG_ST0 0 -#define I830_STPREG_ST1 1 -#define I830_STP_SETUP_SIZE 2 - -#define I830_TEXREG_TM0LI 0 /* load immediate 2 texture map n */ -#define I830_TEXREG_TM0S1 1 -#define I830_TEXREG_TM0S2 2 -#define I830_TEXREG_TM0S3 3 -#define I830_TEXREG_TM0S4 4 -#define I830_TEXREG_MCS 5 /* _3DSTATE_MAP_COORD_SETS */ -#define I830_TEXREG_CUBE 6 /* _3DSTATE_MAP_SUBE */ -#define I830_TEX_SETUP_SIZE 7 - -#define I830_TEXBLEND_SIZE 12 /* (4 args + op) * 2 + COLOR_FACTOR */ - -struct i830_texture_object -{ - struct intel_texture_object intel; - GLuint Setup[I830_TEX_SETUP_SIZE]; -}; - -#define I830_TEX_UNITS 4 - -struct i830_hw_state -{ - GLuint Ctx[I830_CTX_SETUP_SIZE]; - GLuint Buffer[I830_DEST_SETUP_SIZE]; - GLuint Stipple[I830_STP_SETUP_SIZE]; - GLuint Tex[I830_TEX_UNITS][I830_TEX_SETUP_SIZE]; - GLuint TexBlend[I830_TEX_UNITS][I830_TEXBLEND_SIZE]; - GLuint TexBlendWordsUsed[I830_TEX_UNITS]; - - struct intel_region *draw_region; - struct intel_region *depth_region; - - /* Regions aren't actually that appropriate here as the memory may - * be from a PBO or FBO. Just use the buffer id. Will have to do - * this for draw and depth for FBO's... - */ - struct _DriBufferObject *tex_buffer[I830_TEX_UNITS]; - GLuint tex_offset[I830_TEX_UNITS]; - - GLuint emitted; /* I810_UPLOAD_* */ - GLuint active; -}; - -struct i830_context -{ - struct intel_context intel; - - GLuint lodbias_tm0s3[MAX_TEXTURE_UNITS]; - DECLARE_RENDERINPUTS(last_index_bitset); - - struct i830_hw_state meta, initial, state, *current; -}; - - - - -#define I830_STATECHANGE(i830, flag) \ -do { \ - INTEL_FIREVERTICES( &i830->intel ); \ - i830->state.emitted &= ~flag; \ -} while (0) - -#define I830_ACTIVESTATE(i830, flag, mode) \ -do { \ - INTEL_FIREVERTICES( &i830->intel ); \ - if (mode) \ - i830->state.active |= flag; \ - else \ - i830->state.active &= ~flag; \ -} while (0) - -/* i830_vtbl.c - */ -extern void i830InitVtbl(struct i830_context *i830); - -extern void -i830_state_draw_region(struct intel_context *intel, - struct i830_hw_state *state, - struct intel_region *color_region, - struct intel_region *depth_region); -/* i830_context.c - */ -extern GLboolean -i830CreateContext(const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - -/* i830_tex.c, i830_texstate.c - */ -extern void i830UpdateTextureState(struct intel_context *intel); - -extern void i830InitTextureFuncs(struct dd_function_table *functions); - -/* i830_texblend.c - */ -extern GLuint i830SetTexEnvCombine(struct i830_context *i830, - const struct gl_tex_env_combine_state - *combine, GLint blendUnit, GLuint texel_op, - GLuint * state, const GLfloat * factor); - -extern void i830EmitTextureBlend(struct i830_context *i830); - - -/* i830_state.c - */ -extern void i830InitStateFuncs(struct dd_function_table *functions); - -extern void i830EmitState(struct i830_context *i830); - -extern void i830InitState(struct i830_context *i830); - -/* i830_metaops.c - */ -extern void i830InitMetaFuncs(struct i830_context *i830); - -extern void -i830RotateWindow(struct intel_context *intel, __DRIdrawablePrivate * dPriv, - GLuint srcBuf); - -/*====================================================================== - * Inline conversion functions. These are better-typed than the - * macros used previously: - */ -static INLINE struct i830_context * -i830_context(GLcontext * ctx) -{ - return (struct i830_context *) ctx; -} - -#endif diff --git a/src/mesa/drivers/dri/i915tex/i830_metaops.c b/src/mesa/drivers/dri/i915tex/i830_metaops.c deleted file mode 100644 index f76646d89d..0000000000 --- a/src/mesa/drivers/dri/i915tex/i830_metaops.c +++ /dev/null @@ -1,457 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "enums.h" -#include "mtypes.h" -#include "macros.h" -#include "utils.h" - -#include "intel_screen.h" -#include "intel_batchbuffer.h" -#include "intel_ioctl.h" -#include "intel_regions.h" - -#include "i830_context.h" -#include "i830_reg.h" - -/* A large amount of state doesn't need to be uploaded. - */ -#define ACTIVE (I830_UPLOAD_INVARIENT | \ - I830_UPLOAD_CTX | \ - I830_UPLOAD_BUFFERS | \ - I830_UPLOAD_STIPPLE | \ - I830_UPLOAD_TEXBLEND(0) | \ - I830_UPLOAD_TEX(0)) - - -#define SET_STATE( i830, STATE ) \ -do { \ - i830->current->emitted &= ~ACTIVE; \ - i830->current = &i830->STATE; \ - i830->current->emitted &= ~ACTIVE; \ -} while (0) - - -static void -set_no_stencil_write(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - - /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_FALSE ) - */ - i830->meta.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_STENCIL_TEST; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_STENCIL_WRITE; - i830->meta.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_STENCIL_TEST; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= DISABLE_STENCIL_WRITE; - - i830->meta.emitted &= ~I830_UPLOAD_CTX; -} - -static void -set_no_depth_write(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - - /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_FALSE ) - */ - i830->meta.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_DEPTH_TEST_MASK; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_DIS_DEPTH_WRITE_MASK; - i830->meta.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_DEPTH_TEST; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= DISABLE_DEPTH_WRITE; - - i830->meta.emitted &= ~I830_UPLOAD_CTX; -} - -/* Set depth unit to replace. - */ -static void -set_depth_replace(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - - /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_FALSE ) - * ctx->Driver.DepthMask( ctx, GL_TRUE ) - */ - i830->meta.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_DEPTH_TEST_MASK; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_DIS_DEPTH_WRITE_MASK; - i830->meta.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_DEPTH_TEST; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_DEPTH_WRITE; - - /* ctx->Driver.DepthFunc( ctx, GL_ALWAYS ) - */ - i830->meta.Ctx[I830_CTXREG_STATE3] &= ~DEPTH_TEST_FUNC_MASK; - i830->meta.Ctx[I830_CTXREG_STATE3] |= (ENABLE_DEPTH_TEST_FUNC | - DEPTH_TEST_FUNC - (COMPAREFUNC_ALWAYS)); - - i830->meta.emitted &= ~I830_UPLOAD_CTX; -} - - -/* Set stencil unit to replace always with the reference value. - */ -static void -set_stencil_replace(struct intel_context *intel, - GLuint s_mask, GLuint s_clear) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - - /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_TRUE ) - */ - i830->meta.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_STENCIL_TEST; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_STENCIL_WRITE; - - /* ctx->Driver.StencilMask( ctx, s_mask ) - */ - i830->meta.Ctx[I830_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK; - i830->meta.Ctx[I830_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK((s_mask & - 0xff))); - - /* ctx->Driver.StencilOp( ctx, GL_REPLACE, GL_REPLACE, GL_REPLACE ) - */ - i830->meta.Ctx[I830_CTXREG_STENCILTST] &= ~(STENCIL_OPS_MASK); - i830->meta.Ctx[I830_CTXREG_STENCILTST] |= - (ENABLE_STENCIL_PARMS | - STENCIL_FAIL_OP(STENCILOP_REPLACE) | - STENCIL_PASS_DEPTH_FAIL_OP(STENCILOP_REPLACE) | - STENCIL_PASS_DEPTH_PASS_OP(STENCILOP_REPLACE)); - - /* ctx->Driver.StencilFunc( ctx, GL_ALWAYS, s_clear, ~0 ) - */ - i830->meta.Ctx[I830_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_TEST_MASK; - i830->meta.Ctx[I830_CTXREG_STATE4] |= (ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(0xff)); - - i830->meta.Ctx[I830_CTXREG_STENCILTST] &= ~(STENCIL_REF_VALUE_MASK | - ENABLE_STENCIL_TEST_FUNC_MASK); - i830->meta.Ctx[I830_CTXREG_STENCILTST] |= - (ENABLE_STENCIL_REF_VALUE | - ENABLE_STENCIL_TEST_FUNC | - STENCIL_REF_VALUE((s_clear & 0xff)) | - STENCIL_TEST_FUNC(COMPAREFUNC_ALWAYS)); - - - - i830->meta.emitted &= ~I830_UPLOAD_CTX; -} - - -static void -set_color_mask(struct intel_context *intel, GLboolean state) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - - const GLuint mask = ((1 << WRITEMASK_RED_SHIFT) | - (1 << WRITEMASK_GREEN_SHIFT) | - (1 << WRITEMASK_BLUE_SHIFT) | - (1 << WRITEMASK_ALPHA_SHIFT)); - - i830->meta.Ctx[I830_CTXREG_ENABLES_2] &= ~mask; - - if (state) { - i830->meta.Ctx[I830_CTXREG_ENABLES_2] |= - (i830->state.Ctx[I830_CTXREG_ENABLES_2] & mask); - } - - i830->meta.emitted &= ~I830_UPLOAD_CTX; -} - -/* Installs a one-stage passthrough texture blend pipeline. Is there - * more that can be done to turn off texturing? - */ -static void -set_no_texture(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - static const struct gl_tex_env_combine_state comb = { - GL_NONE, GL_NONE, - {GL_TEXTURE, 0, 0,}, {GL_TEXTURE, 0, 0,}, - {GL_SRC_COLOR, 0, 0}, {GL_SRC_ALPHA, 0, 0}, - 0, 0, 0, 0 - }; - - i830->meta.TexBlendWordsUsed[0] = - i830SetTexEnvCombine(i830, &comb, 0, TEXBLENDARG_TEXEL0, - i830->meta.TexBlend[0], NULL); - - i830->meta.TexBlend[0][0] |= TEXOP_LAST_STAGE; - i830->meta.emitted &= ~I830_UPLOAD_TEXBLEND(0); -} - -/* Set up a single element blend stage for 'replace' texturing with no - * funny ops. - */ -static void -set_texture_blend_replace(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - static const struct gl_tex_env_combine_state comb = { - GL_REPLACE, GL_REPLACE, - {GL_TEXTURE, GL_TEXTURE, GL_TEXTURE,}, {GL_TEXTURE, GL_TEXTURE, - GL_TEXTURE,}, - {GL_SRC_COLOR, GL_SRC_COLOR, GL_SRC_COLOR}, {GL_SRC_ALPHA, GL_SRC_ALPHA, - GL_SRC_ALPHA}, - 0, 0, 1, 1 - }; - - i830->meta.TexBlendWordsUsed[0] = - i830SetTexEnvCombine(i830, &comb, 0, TEXBLENDARG_TEXEL0, - i830->meta.TexBlend[0], NULL); - - i830->meta.TexBlend[0][0] |= TEXOP_LAST_STAGE; - i830->meta.emitted &= ~I830_UPLOAD_TEXBLEND(0); - -/* fprintf(stderr, "%s: TexBlendWordsUsed[0]: %d\n", */ -/* __FUNCTION__, i830->meta.TexBlendWordsUsed[0]); */ -} - - - -/* Set up an arbitary piece of memory as a rectangular texture - * (including the front or back buffer). - */ -static GLboolean -set_tex_rect_source(struct intel_context *intel, - struct _DriBufferObject *buffer, - GLuint offset, - GLuint pitch, GLuint height, GLenum format, GLenum type) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - GLuint *setup = i830->meta.Tex[0]; - GLint numLevels = 1; - GLuint textureFormat; - GLuint cpp; - - /* A full implementation of this would do the upload through - * glTexImage2d, and get all the conversion operations at that - * point. We are restricted, but still at least have access to the - * fragment program swizzle. - */ - switch (format) { - case GL_BGRA: - switch (type) { - case GL_UNSIGNED_INT_8_8_8_8_REV: - case GL_UNSIGNED_BYTE: - textureFormat = (MAPSURF_32BIT | MT_32BIT_ARGB8888); - cpp = 4; - break; - default: - return GL_FALSE; - } - break; - case GL_RGBA: - switch (type) { - case GL_UNSIGNED_INT_8_8_8_8_REV: - case GL_UNSIGNED_BYTE: - textureFormat = (MAPSURF_32BIT | MT_32BIT_ABGR8888); - cpp = 4; - break; - default: - return GL_FALSE; - } - break; - case GL_BGR: - switch (type) { - case GL_UNSIGNED_SHORT_5_6_5_REV: - textureFormat = (MAPSURF_16BIT | MT_16BIT_RGB565); - cpp = 2; - break; - default: - return GL_FALSE; - } - break; - case GL_RGB: - switch (type) { - case GL_UNSIGNED_SHORT_5_6_5: - textureFormat = (MAPSURF_16BIT | MT_16BIT_RGB565); - cpp = 2; - break; - default: - return GL_FALSE; - } - break; - - default: - return GL_FALSE; - } - - i830->meta.tex_buffer[0] = buffer; - i830->meta.tex_offset[0] = offset; - - setup[I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 | - (LOAD_TEXTURE_MAP0 << 0) | 4); - setup[I830_TEXREG_TM0S1] = (((height - 1) << TM0S1_HEIGHT_SHIFT) | - ((pitch - 1) << TM0S1_WIDTH_SHIFT) | - textureFormat); - setup[I830_TEXREG_TM0S2] = - (((((pitch * cpp) / 4) - - 1) << TM0S2_PITCH_SHIFT) | TM0S2_CUBE_FACE_ENA_MASK); - - setup[I830_TEXREG_TM0S3] = - ((((numLevels - - 1) * - 4) << TM0S3_MIN_MIP_SHIFT) | (FILTER_NEAREST << - TM0S3_MIN_FILTER_SHIFT) | - (MIPFILTER_NONE << TM0S3_MIP_FILTER_SHIFT) | (FILTER_NEAREST << - TM0S3_MAG_FILTER_SHIFT)); - - setup[I830_TEXREG_CUBE] = (_3DSTATE_MAP_CUBE | MAP_UNIT(0)); - - setup[I830_TEXREG_MCS] = (_3DSTATE_MAP_COORD_SET_CMD | - MAP_UNIT(0) | - ENABLE_TEXCOORD_PARAMS | - TEXCOORDS_ARE_IN_TEXELUNITS | - TEXCOORDTYPE_CARTESIAN | - ENABLE_ADDR_V_CNTL | - TEXCOORD_ADDR_V_MODE(TEXCOORDMODE_WRAP) | - ENABLE_ADDR_U_CNTL | - TEXCOORD_ADDR_U_MODE(TEXCOORDMODE_WRAP)); - - i830->meta.emitted &= ~I830_UPLOAD_TEX(0); - return GL_TRUE; -} - - -static void -set_vertex_format(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - i830->meta.Ctx[I830_CTXREG_VF] = (_3DSTATE_VFT0_CMD | - VFT0_TEX_COUNT(1) | - VFT0_DIFFUSE | VFT0_XYZ); - i830->meta.Ctx[I830_CTXREG_VF2] = (_3DSTATE_VFT1_CMD | - VFT1_TEX0_FMT(TEXCOORDFMT_2D) | - VFT1_TEX1_FMT(TEXCOORDFMT_2D) | - VFT1_TEX2_FMT(TEXCOORDFMT_2D) | - VFT1_TEX3_FMT(TEXCOORDFMT_2D)); - i830->meta.emitted &= ~I830_UPLOAD_CTX; -} - - -static void -meta_import_pixel_state(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - - i830->meta.Ctx[I830_CTXREG_STATE1] = i830->state.Ctx[I830_CTXREG_STATE1]; - i830->meta.Ctx[I830_CTXREG_STATE2] = i830->state.Ctx[I830_CTXREG_STATE2]; - i830->meta.Ctx[I830_CTXREG_STATE3] = i830->state.Ctx[I830_CTXREG_STATE3]; - i830->meta.Ctx[I830_CTXREG_STATE4] = i830->state.Ctx[I830_CTXREG_STATE4]; - i830->meta.Ctx[I830_CTXREG_STATE5] = i830->state.Ctx[I830_CTXREG_STATE5]; - i830->meta.Ctx[I830_CTXREG_IALPHAB] = i830->state.Ctx[I830_CTXREG_IALPHAB]; - i830->meta.Ctx[I830_CTXREG_STENCILTST] = - i830->state.Ctx[I830_CTXREG_STENCILTST]; - i830->meta.Ctx[I830_CTXREG_ENABLES_1] = - i830->state.Ctx[I830_CTXREG_ENABLES_1]; - i830->meta.Ctx[I830_CTXREG_ENABLES_2] = - i830->state.Ctx[I830_CTXREG_ENABLES_2]; - i830->meta.Ctx[I830_CTXREG_AA] = i830->state.Ctx[I830_CTXREG_AA]; - i830->meta.Ctx[I830_CTXREG_FOGCOLOR] = - i830->state.Ctx[I830_CTXREG_FOGCOLOR]; - i830->meta.Ctx[I830_CTXREG_BLENDCOLOR0] = - i830->state.Ctx[I830_CTXREG_BLENDCOLOR0]; - i830->meta.Ctx[I830_CTXREG_BLENDCOLOR1] = - i830->state.Ctx[I830_CTXREG_BLENDCOLOR1]; - i830->meta.Ctx[I830_CTXREG_MCSB0] = i830->state.Ctx[I830_CTXREG_MCSB0]; - i830->meta.Ctx[I830_CTXREG_MCSB1] = i830->state.Ctx[I830_CTXREG_MCSB1]; - - - i830->meta.Ctx[I830_CTXREG_STATE3] &= ~CULLMODE_MASK; - i830->meta.Stipple[I830_STPREG_ST1] &= ~ST1_ENABLE; - i830->meta.emitted &= ~I830_UPLOAD_CTX; - - - i830->meta.Buffer[I830_DESTREG_SENABLE] = - i830->state.Buffer[I830_DESTREG_SENABLE]; - i830->meta.Buffer[I830_DESTREG_SR1] = i830->state.Buffer[I830_DESTREG_SR1]; - i830->meta.Buffer[I830_DESTREG_SR2] = i830->state.Buffer[I830_DESTREG_SR2]; - i830->meta.emitted &= ~I830_UPLOAD_BUFFERS; -} - - - -/* Select between front and back draw buffers. - */ -static void -meta_draw_region(struct intel_context *intel, - struct intel_region *color_region, - struct intel_region *depth_region) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - - i830_state_draw_region(intel, &i830->meta, color_region, depth_region); -} - - -/* Operations where the 3D engine is decoupled temporarily from the - * current GL state and used for other purposes than simply rendering - * incoming triangles. - */ -static void -install_meta_state(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - memcpy(&i830->meta, &i830->initial, sizeof(i830->meta)); - - i830->meta.active = ACTIVE; - i830->meta.emitted = 0; - - SET_STATE(i830, meta); - set_vertex_format(intel); - set_no_texture(intel); -} - -static void -leave_meta_state(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - intel_region_release(&i830->meta.draw_region); - intel_region_release(&i830->meta.depth_region); -/* intel_region_release(intel, &i830->meta.tex_region[0]); */ - SET_STATE(i830, state); -} - - - -void -i830InitMetaFuncs(struct i830_context *i830) -{ - i830->intel.vtbl.install_meta_state = install_meta_state; - i830->intel.vtbl.leave_meta_state = leave_meta_state; - i830->intel.vtbl.meta_no_depth_write = set_no_depth_write; - i830->intel.vtbl.meta_no_stencil_write = set_no_stencil_write; - i830->intel.vtbl.meta_stencil_replace = set_stencil_replace; - i830->intel.vtbl.meta_depth_replace = set_depth_replace; - i830->intel.vtbl.meta_color_mask = set_color_mask; - i830->intel.vtbl.meta_no_texture = set_no_texture; - i830->intel.vtbl.meta_texture_blend_replace = set_texture_blend_replace; - i830->intel.vtbl.meta_tex_rect_source = set_tex_rect_source; - i830->intel.vtbl.meta_draw_region = meta_draw_region; - i830->intel.vtbl.meta_import_pixel_state = meta_import_pixel_state; -} diff --git a/src/mesa/drivers/dri/i915tex/i830_reg.h b/src/mesa/drivers/dri/i915tex/i830_reg.h deleted file mode 100644 index 41280bca7c..0000000000 --- a/src/mesa/drivers/dri/i915tex/i830_reg.h +++ /dev/null @@ -1,642 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef _I830_REG_H_ -#define _I830_REG_H_ - - -#include "intel_reg.h" - -#define I830_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value) - -#define _3DSTATE_AA_CMD (CMD_3D | (0x06<<24)) -#define AA_LINE_ECAAR_WIDTH_ENABLE (1<<16) -#define AA_LINE_ECAAR_WIDTH_0_5 0 -#define AA_LINE_ECAAR_WIDTH_1_0 (1<<14) -#define AA_LINE_ECAAR_WIDTH_2_0 (2<<14) -#define AA_LINE_ECAAR_WIDTH_4_0 (3<<14) -#define AA_LINE_REGION_WIDTH_ENABLE (1<<8) -#define AA_LINE_REGION_WIDTH_0_5 0 -#define AA_LINE_REGION_WIDTH_1_0 (1<<6) -#define AA_LINE_REGION_WIDTH_2_0 (2<<6) -#define AA_LINE_REGION_WIDTH_4_0 (3<<6) -#define AA_LINE_ENABLE ((1<<1) | 1) -#define AA_LINE_DISABLE (1<<1) - -#define _3DSTATE_BUF_INFO_CMD (CMD_3D | (0x1d<<24) | (0x8e<<16) | 1) -/* Dword 1 */ -#define BUF_3D_ID_COLOR_BACK (0x3<<24) -#define BUF_3D_ID_DEPTH (0x7<<24) -#define BUF_3D_USE_FENCE (1<<23) -#define BUF_3D_TILED_SURFACE (1<<22) -#define BUF_3D_TILE_WALK_X 0 -#define BUF_3D_TILE_WALK_Y (1<<21) -#define BUF_3D_PITCH(x) (((x)/4)<<2) -/* Dword 2 */ -#define BUF_3D_ADDR(x) ((x) & ~0x3) - - -#define _3DSTATE_COLOR_FACTOR_CMD (CMD_3D | (0x1d<<24) | (0x1<<16)) - -#define _3DSTATE_COLOR_FACTOR_N_CMD(stage) (CMD_3D | (0x1d<<24) | \ - ((0x90+(stage))<<16)) - -#define _3DSTATE_CONST_BLEND_COLOR_CMD (CMD_3D | (0x1d<<24) | (0x88<<16)) - -#define _3DSTATE_DFLT_DIFFUSE_CMD (CMD_3D | (0x1d<<24) | (0x99<<16)) - -#define _3DSTATE_DFLT_SPEC_CMD (CMD_3D | (0x1d<<24) | (0x9a<<16)) - -#define _3DSTATE_DFLT_Z_CMD (CMD_3D | (0x1d<<24) | (0x98<<16)) - - -#define _3DSTATE_DST_BUF_VARS_CMD (CMD_3D | (0x1d<<24) | (0x85<<16)) -/* Dword 1 */ -#define DSTORG_HORT_BIAS(x) ((x)<<20) -#define DSTORG_VERT_BIAS(x) ((x)<<16) -#define COLOR_4_2_2_CHNL_WRT_ALL 0 -#define COLOR_4_2_2_CHNL_WRT_Y (1<<12) -#define COLOR_4_2_2_CHNL_WRT_CR (2<<12) -#define COLOR_4_2_2_CHNL_WRT_CB (3<<12) -#define COLOR_4_2_2_CHNL_WRT_CRCB (4<<12) -#define COLR_BUF_8BIT 0 -#define COLR_BUF_RGB555 (1<<8) -#define COLR_BUF_RGB565 (2<<8) -#define COLR_BUF_ARGB8888 (3<<8) -#define DEPTH_IS_Z 0 -#define DEPTH_IS_W (1<<6) -#define DEPTH_FRMT_16_FIXED 0 -#define DEPTH_FRMT_16_FLOAT (1<<2) -#define DEPTH_FRMT_24_FIXED_8_OTHER (2<<2) -#define DEPTH_FRMT_24_FLOAT_8_OTHER (3<<2) -#define VERT_LINE_STRIDE_1 (1<<1) -#define VERT_LINE_STRIDE_0 0 -#define VERT_LINE_STRIDE_OFS_1 1 -#define VERT_LINE_STRIDE_OFS_0 0 - - -#define _3DSTATE_DRAW_RECT_CMD (CMD_3D|(0x1d<<24)|(0x80<<16)|3) -/* Dword 1 */ -#define DRAW_RECT_DIS_DEPTH_OFS (1<<30) -#define DRAW_DITHER_OFS_X(x) ((x)<<26) -#define DRAW_DITHER_OFS_Y(x) ((x)<<24) -/* Dword 2 */ -#define DRAW_YMIN(x) ((x)<<16) -#define DRAW_XMIN(x) (x) -/* Dword 3 */ -#define DRAW_YMAX(x) ((x)<<16) -#define DRAW_XMAX(x) (x) -/* Dword 4 */ -#define DRAW_YORG(x) ((x)<<16) -#define DRAW_XORG(x) (x) - - -#define _3DSTATE_ENABLES_1_CMD (CMD_3D|(0x3<<24)) -#define ENABLE_LOGIC_OP_MASK ((1<<23)|(1<<22)) -#define ENABLE_LOGIC_OP ((1<<23)|(1<<22)) -#define DISABLE_LOGIC_OP (1<<23) -#define ENABLE_STENCIL_TEST ((1<<21)|(1<<20)) -#define DISABLE_STENCIL_TEST (1<<21) -#define ENABLE_DEPTH_BIAS ((1<<11)|(1<<10)) -#define DISABLE_DEPTH_BIAS (1<<11) -#define ENABLE_SPEC_ADD_MASK ((1<<9)|(1<<8)) -#define ENABLE_SPEC_ADD ((1<<9)|(1<<8)) -#define DISABLE_SPEC_ADD (1<<9) -#define ENABLE_DIS_FOG_MASK ((1<<7)|(1<<6)) -#define ENABLE_FOG ((1<<7)|(1<<6)) -#define DISABLE_FOG (1<<7) -#define ENABLE_DIS_ALPHA_TEST_MASK ((1<<5)|(1<<4)) -#define ENABLE_ALPHA_TEST ((1<<5)|(1<<4)) -#define DISABLE_ALPHA_TEST (1<<5) -#define ENABLE_DIS_CBLEND_MASK ((1<<3)|(1<<2)) -#define ENABLE_COLOR_BLEND ((1<<3)|(1<<2)) -#define DISABLE_COLOR_BLEND (1<<3) -#define ENABLE_DIS_DEPTH_TEST_MASK ((1<<1)|1) -#define ENABLE_DEPTH_TEST ((1<<1)|1) -#define DISABLE_DEPTH_TEST (1<<1) - -/* _3DSTATE_ENABLES_2, p138 */ -#define _3DSTATE_ENABLES_2_CMD (CMD_3D|(0x4<<24)) -#define ENABLE_STENCIL_WRITE ((1<<21)|(1<<20)) -#define DISABLE_STENCIL_WRITE (1<<21) -#define ENABLE_TEX_CACHE ((1<<17)|(1<<16)) -#define DISABLE_TEX_CACHE (1<<17) -#define ENABLE_DITHER ((1<<9)|(1<<8)) -#define DISABLE_DITHER (1<<9) -#define ENABLE_COLOR_MASK (1<<10) -#define WRITEMASK_ALPHA (1<<7) -#define WRITEMASK_ALPHA_SHIFT 7 -#define WRITEMASK_RED (1<<6) -#define WRITEMASK_RED_SHIFT 6 -#define WRITEMASK_GREEN (1<<5) -#define WRITEMASK_GREEN_SHIFT 5 -#define WRITEMASK_BLUE (1<<4) -#define WRITEMASK_BLUE_SHIFT 4 -#define WRITEMASK_MASK ((1<<4)|(1<<5)|(1<<6)|(1<<7)) -#define ENABLE_COLOR_WRITE ((1<<3)|(1<<2)) -#define DISABLE_COLOR_WRITE (1<<3) -#define ENABLE_DIS_DEPTH_WRITE_MASK 0x3 -#define ENABLE_DEPTH_WRITE ((1<<1)|1) -#define DISABLE_DEPTH_WRITE (1<<1) - -/* _3DSTATE_FOG_COLOR, p139 */ -#define _3DSTATE_FOG_COLOR_CMD (CMD_3D|(0x15<<24)) -#define FOG_COLOR_RED(x) ((x)<<16) -#define FOG_COLOR_GREEN(x) ((x)<<8) -#define FOG_COLOR_BLUE(x) (x) - -/* _3DSTATE_FOG_MODE, p140 */ -#define _3DSTATE_FOG_MODE_CMD (CMD_3D|(0x1d<<24)|(0x89<<16)|2) -/* Dword 1 */ -#define FOGFUNC_ENABLE (1<<31) -#define FOGFUNC_VERTEX 0 -#define FOGFUNC_PIXEL_EXP (1<<28) -#define FOGFUNC_PIXEL_EXP2 (2<<28) -#define FOGFUNC_PIXEL_LINEAR (3<<28) -#define FOGSRC_INDEX_Z (1<<27) -#define FOGSRC_INDEX_W ((1<<27)|(1<<25)) -#define FOG_LINEAR_CONST (1<<24) -#define FOG_CONST_1(x) ((x)<<4) -#define ENABLE_FOG_DENSITY (1<<23) -/* Dword 2 */ -#define FOG_CONST_2(x) (x) -/* Dword 3 */ -#define FOG_DENSITY(x) (x) - -/* _3DSTATE_INDEPENDENT_ALPHA_BLEND, p142 */ -#define _3DSTATE_INDPT_ALPHA_BLEND_CMD (CMD_3D|(0x0b<<24)) -#define ENABLE_INDPT_ALPHA_BLEND ((1<<23)|(1<<22)) -#define DISABLE_INDPT_ALPHA_BLEND (1<<23) -#define ALPHA_BLENDFUNC_MASK 0x3f0000 -#define ENABLE_ALPHA_BLENDFUNC (1<<21) -#define ABLENDFUNC_ADD 0 -#define ABLENDFUNC_SUB (1<<16) -#define ABLENDFUNC_RVSE_SUB (2<<16) -#define ABLENDFUNC_MIN (3<<16) -#define ABLENDFUNC_MAX (4<<16) -#define SRC_DST_ABLEND_MASK 0xfff -#define ENABLE_SRC_ABLEND_FACTOR (1<<11) -#define SRC_ABLEND_FACT(x) ((x)<<6) -#define ENABLE_DST_ABLEND_FACTOR (1<<5) -#define DST_ABLEND_FACT(x) (x) - - -/* _3DSTATE_MAP_BLEND_ARG, p152 */ -#define _3DSTATE_MAP_BLEND_ARG_CMD(stage) (CMD_3D|(0x0e<<24)|((stage)<<20)) - -#define TEXPIPE_COLOR 0 -#define TEXPIPE_ALPHA (1<<18) -#define TEXPIPE_KILL (2<<18) -#define TEXBLEND_ARG0 0 -#define TEXBLEND_ARG1 (1<<15) -#define TEXBLEND_ARG2 (2<<15) -#define TEXBLEND_ARG3 (3<<15) -#define TEXBLENDARG_MODIFY_PARMS (1<<6) -#define TEXBLENDARG_REPLICATE_ALPHA (1<<5) -#define TEXBLENDARG_INV_ARG (1<<4) -#define TEXBLENDARG_ONE 0 -#define TEXBLENDARG_FACTOR 0x01 -#define TEXBLENDARG_ACCUM 0x02 -#define TEXBLENDARG_DIFFUSE 0x03 -#define TEXBLENDARG_SPEC 0x04 -#define TEXBLENDARG_CURRENT 0x05 -#define TEXBLENDARG_TEXEL0 0x06 -#define TEXBLENDARG_TEXEL1 0x07 -#define TEXBLENDARG_TEXEL2 0x08 -#define TEXBLENDARG_TEXEL3 0x09 -#define TEXBLENDARG_FACTOR_N 0x0e - -/* _3DSTATE_MAP_BLEND_OP, p155 */ -#define _3DSTATE_MAP_BLEND_OP_CMD(stage) (CMD_3D|(0x0d<<24)|((stage)<<20)) -#if 0 -# define TEXPIPE_COLOR 0 -# define TEXPIPE_ALPHA (1<<18) -# define TEXPIPE_KILL (2<<18) -#endif -#define ENABLE_TEXOUTPUT_WRT_SEL (1<<17) -#define TEXOP_OUTPUT_CURRENT 0 -#define TEXOP_OUTPUT_ACCUM (1<<15) -#define ENABLE_TEX_CNTRL_STAGE ((1<<12)|(1<<11)) -#define DISABLE_TEX_CNTRL_STAGE (1<<12) -#define TEXOP_SCALE_SHIFT 9 -#define TEXOP_SCALE_1X (0 << TEXOP_SCALE_SHIFT) -#define TEXOP_SCALE_2X (1 << TEXOP_SCALE_SHIFT) -#define TEXOP_SCALE_4X (2 << TEXOP_SCALE_SHIFT) -#define TEXOP_MODIFY_PARMS (1<<8) -#define TEXOP_LAST_STAGE (1<<7) -#define TEXBLENDOP_KILLPIXEL 0x02 -#define TEXBLENDOP_ARG1 0x01 -#define TEXBLENDOP_ARG2 0x02 -#define TEXBLENDOP_MODULATE 0x03 -#define TEXBLENDOP_ADD 0x06 -#define TEXBLENDOP_ADDSIGNED 0x07 -#define TEXBLENDOP_BLEND 0x08 -#define TEXBLENDOP_BLEND_AND_ADD 0x09 -#define TEXBLENDOP_SUBTRACT 0x0a -#define TEXBLENDOP_DOT3 0x0b -#define TEXBLENDOP_DOT4 0x0c -#define TEXBLENDOP_MODULATE_AND_ADD 0x0d -#define TEXBLENDOP_MODULATE_2X_AND_ADD 0x0e -#define TEXBLENDOP_MODULATE_4X_AND_ADD 0x0f - -/* _3DSTATE_MAP_BUMP_TABLE, p160 TODO */ -/* _3DSTATE_MAP_COLOR_CHROMA_KEY, p161 TODO */ - -#define _3DSTATE_MAP_COORD_TRANSFORM ((3<<29)|(0x1d<<24)|(0x8c<<16)) -#define DISABLE_TEX_TRANSFORM (1<<28) -#define TEXTURE_SET(x) (x<<29) - -#define _3DSTATE_VERTEX_TRANSFORM ((3<<29)|(0x1d<<24)|(0x8b<<16)) -#define DISABLE_VIEWPORT_TRANSFORM (1<<31) -#define DISABLE_PERSPECTIVE_DIVIDE (1<<29) - - -/* _3DSTATE_MAP_COORD_SET_BINDINGS, p162 */ -#define _3DSTATE_MAP_COORD_SETBIND_CMD (CMD_3D|(0x1d<<24)|(0x02<<16)) -#define TEXBIND_MASK3 ((1<<15)|(1<<14)|(1<<13)|(1<<12)) -#define TEXBIND_MASK2 ((1<<11)|(1<<10)|(1<<9)|(1<<8)) -#define TEXBIND_MASK1 ((1<<7)|(1<<6)|(1<<5)|(1<<4)) -#define TEXBIND_MASK0 ((1<<3)|(1<<2)|(1<<1)|1) - -#define TEXBIND_SET3(x) ((x)<<12) -#define TEXBIND_SET2(x) ((x)<<8) -#define TEXBIND_SET1(x) ((x)<<4) -#define TEXBIND_SET0(x) (x) - -#define TEXCOORDSRC_KEEP 0 -#define TEXCOORDSRC_DEFAULT 0x01 -#define TEXCOORDSRC_VTXSET_0 0x08 -#define TEXCOORDSRC_VTXSET_1 0x09 -#define TEXCOORDSRC_VTXSET_2 0x0a -#define TEXCOORDSRC_VTXSET_3 0x0b -#define TEXCOORDSRC_VTXSET_4 0x0c -#define TEXCOORDSRC_VTXSET_5 0x0d -#define TEXCOORDSRC_VTXSET_6 0x0e -#define TEXCOORDSRC_VTXSET_7 0x0f - -#define MAP_UNIT(unit) ((unit)<<16) -#define MAP_UNIT_MASK (0x7<<16) - -/* _3DSTATE_MAP_COORD_SETS, p164 */ -#define _3DSTATE_MAP_COORD_SET_CMD (CMD_3D|(0x1c<<24)|(0x01<<19)) -#define ENABLE_TEXCOORD_PARAMS (1<<15) -#define TEXCOORDS_ARE_NORMAL (1<<14) -#define TEXCOORDS_ARE_IN_TEXELUNITS 0 -#define TEXCOORDTYPE_CARTESIAN 0 -#define TEXCOORDTYPE_HOMOGENEOUS (1<<11) -#define TEXCOORDTYPE_VECTOR (2<<11) -#define TEXCOORDTYPE_MASK (0x7<<11) -#define ENABLE_ADDR_V_CNTL (1<<7) -#define ENABLE_ADDR_U_CNTL (1<<3) -#define TEXCOORD_ADDR_V_MODE(x) ((x)<<4) -#define TEXCOORD_ADDR_U_MODE(x) (x) -#define TEXCOORDMODE_WRAP 0 -#define TEXCOORDMODE_MIRROR 1 -#define TEXCOORDMODE_CLAMP 2 -#define TEXCOORDMODE_WRAP_SHORTEST 3 -#define TEXCOORDMODE_CLAMP_BORDER 4 -#define TEXCOORD_ADDR_V_MASK 0x70 -#define TEXCOORD_ADDR_U_MASK 0x7 - -/* _3DSTATE_MAP_CUBE, p168 TODO */ -#define _3DSTATE_MAP_CUBE (CMD_3D|(0x1c<<24)|(0x0a<<19)) -#define CUBE_NEGX_ENABLE (1<<5) -#define CUBE_POSX_ENABLE (1<<4) -#define CUBE_NEGY_ENABLE (1<<3) -#define CUBE_POSY_ENABLE (1<<2) -#define CUBE_NEGZ_ENABLE (1<<1) -#define CUBE_POSZ_ENABLE (1<<0) - - -/* _3DSTATE_MODES_1, p190 */ -#define _3DSTATE_MODES_1_CMD (CMD_3D|(0x08<<24)) -#define BLENDFUNC_MASK 0x3f0000 -#define ENABLE_COLR_BLND_FUNC (1<<21) -#define BLENDFUNC_ADD 0 -#define BLENDFUNC_SUB (1<<16) -#define BLENDFUNC_RVRSE_SUB (2<<16) -#define BLENDFUNC_MIN (3<<16) -#define BLENDFUNC_MAX (4<<16) -#define SRC_DST_BLND_MASK 0xfff -#define ENABLE_SRC_BLND_FACTOR (1<<11) -#define ENABLE_DST_BLND_FACTOR (1<<5) -#define SRC_BLND_FACT(x) ((x)<<6) -#define DST_BLND_FACT(x) (x) - - -/* _3DSTATE_MODES_2, p192 */ -#define _3DSTATE_MODES_2_CMD (CMD_3D|(0x0f<<24)) -#define ENABLE_GLOBAL_DEPTH_BIAS (1<<22) -#define GLOBAL_DEPTH_BIAS(x) ((x)<<14) -#define ENABLE_ALPHA_TEST_FUNC (1<<13) -#define ENABLE_ALPHA_REF_VALUE (1<<8) -#define ALPHA_TEST_FUNC(x) ((x)<<9) -#define ALPHA_REF_VALUE(x) (x) - -#define ALPHA_TEST_REF_MASK 0x3fff - -/* _3DSTATE_MODES_3, p193 */ -#define _3DSTATE_MODES_3_CMD (CMD_3D|(0x02<<24)) -#define DEPTH_TEST_FUNC_MASK 0x1f0000 -#define ENABLE_DEPTH_TEST_FUNC (1<<20) -/* Uses COMPAREFUNC */ -#define DEPTH_TEST_FUNC(x) ((x)<<16) -#define ENABLE_ALPHA_SHADE_MODE (1<<11) -#define ENABLE_FOG_SHADE_MODE (1<<9) -#define ENABLE_SPEC_SHADE_MODE (1<<7) -#define ENABLE_COLOR_SHADE_MODE (1<<5) -#define ALPHA_SHADE_MODE(x) ((x)<<10) -#define FOG_SHADE_MODE(x) ((x)<<8) -#define SPEC_SHADE_MODE(x) ((x)<<6) -#define COLOR_SHADE_MODE(x) ((x)<<4) -#define CULLMODE_MASK 0xf -#define ENABLE_CULL_MODE (1<<3) -#define CULLMODE_BOTH 0 -#define CULLMODE_NONE 1 -#define CULLMODE_CW 2 -#define CULLMODE_CCW 3 - -#define SHADE_MODE_LINEAR 0 -#define SHADE_MODE_FLAT 0x1 - -/* _3DSTATE_MODES_4, p195 */ -#define _3DSTATE_MODES_4_CMD (CMD_3D|(0x16<<24)) -#define ENABLE_LOGIC_OP_FUNC (1<<23) -#define LOGIC_OP_FUNC(x) ((x)<<18) -#define LOGICOP_MASK ((1<<18)|(1<<19)|(1<<20)|(1<<21)) -#define LOGICOP_CLEAR 0 -#define LOGICOP_NOR 0x1 -#define LOGICOP_AND_INV 0x2 -#define LOGICOP_COPY_INV 0x3 -#define LOGICOP_AND_RVRSE 0x4 -#define LOGICOP_INV 0x5 -#define LOGICOP_XOR 0x6 -#define LOGICOP_NAND 0x7 -#define LOGICOP_AND 0x8 -#define LOGICOP_EQUIV 0x9 -#define LOGICOP_NOOP 0xa -#define LOGICOP_OR_INV 0xb -#define LOGICOP_COPY 0xc -#define LOGICOP_OR_RVRSE 0xd -#define LOGICOP_OR 0xe -#define LOGICOP_SET 0xf -#define MODE4_ENABLE_STENCIL_TEST_MASK ((1<<17)|(0xff00)) -#define ENABLE_STENCIL_TEST_MASK (1<<17) -#define STENCIL_TEST_MASK(x) (((x)&0xff)<<8) -#define MODE4_ENABLE_STENCIL_WRITE_MASK ((1<<16)|(0x00ff)) -#define ENABLE_STENCIL_WRITE_MASK (1<<16) -#define STENCIL_WRITE_MASK(x) ((x)&0xff) - -/* _3DSTATE_MODES_5, p196 */ -#define _3DSTATE_MODES_5_CMD (CMD_3D|(0x0c<<24)) -#define ENABLE_SPRITE_POINT_TEX (1<<23) -#define SPRITE_POINT_TEX_ON (1<<22) -#define SPRITE_POINT_TEX_OFF 0 -#define FLUSH_RENDER_CACHE (1<<18) -#define FLUSH_TEXTURE_CACHE (1<<16) -#define FIXED_LINE_WIDTH_MASK 0xfc00 -#define ENABLE_FIXED_LINE_WIDTH (1<<15) -#define FIXED_LINE_WIDTH(x) ((x)<<10) -#define FIXED_POINT_WIDTH_MASK 0x3ff -#define ENABLE_FIXED_POINT_WIDTH (1<<9) -#define FIXED_POINT_WIDTH(x) (x) - -/* _3DSTATE_RASTERIZATION_RULES, p198 */ -#define _3DSTATE_RASTER_RULES_CMD (CMD_3D|(0x07<<24)) -#define ENABLE_POINT_RASTER_RULE (1<<15) -#define OGL_POINT_RASTER_RULE (1<<13) -#define ENABLE_LINE_STRIP_PROVOKE_VRTX (1<<8) -#define ENABLE_TRI_FAN_PROVOKE_VRTX (1<<5) -#define ENABLE_TRI_STRIP_PROVOKE_VRTX (1<<2) -#define LINE_STRIP_PROVOKE_VRTX(x) ((x)<<6) -#define TRI_FAN_PROVOKE_VRTX(x) ((x)<<3) -#define TRI_STRIP_PROVOKE_VRTX(x) (x) - -/* _3DSTATE_SCISSOR_ENABLE, p200 */ -#define _3DSTATE_SCISSOR_ENABLE_CMD (CMD_3D|(0x1c<<24)|(0x10<<19)) -#define ENABLE_SCISSOR_RECT ((1<<1) | 1) -#define DISABLE_SCISSOR_RECT (1<<1) - -/* _3DSTATE_SCISSOR_RECTANGLE_0, p201 */ -#define _3DSTATE_SCISSOR_RECT_0_CMD (CMD_3D|(0x1d<<24)|(0x81<<16)|1) -/* Dword 1 */ -#define SCISSOR_RECT_0_YMIN(x) ((x)<<16) -#define SCISSOR_RECT_0_XMIN(x) (x) -/* Dword 2 */ -#define SCISSOR_RECT_0_YMAX(x) ((x)<<16) -#define SCISSOR_RECT_0_XMAX(x) (x) - -/* _3DSTATE_STENCIL_TEST, p202 */ -#define _3DSTATE_STENCIL_TEST_CMD (CMD_3D|(0x09<<24)) -#define ENABLE_STENCIL_PARMS (1<<23) -#define STENCIL_OPS_MASK (0xffc000) -#define STENCIL_FAIL_OP(x) ((x)<<20) -#define STENCIL_PASS_DEPTH_FAIL_OP(x) ((x)<<17) -#define STENCIL_PASS_DEPTH_PASS_OP(x) ((x)<<14) - -#define ENABLE_STENCIL_TEST_FUNC_MASK ((1<<13)|(1<<12)|(1<<11)|(1<<10)|(1<<9)) -#define ENABLE_STENCIL_TEST_FUNC (1<<13) -/* Uses COMPAREFUNC */ -#define STENCIL_TEST_FUNC(x) ((x)<<9) -#define STENCIL_REF_VALUE_MASK ((1<<8)|0xff) -#define ENABLE_STENCIL_REF_VALUE (1<<8) -#define STENCIL_REF_VALUE(x) (x) - -/* _3DSTATE_VERTEX_FORMAT, p204 */ -#define _3DSTATE_VFT0_CMD (CMD_3D|(0x05<<24)) -#define VFT0_POINT_WIDTH (1<<12) -#define VFT0_TEX_COUNT_MASK (7<<8) -#define VFT0_TEX_COUNT_SHIFT 8 -#define VFT0_TEX_COUNT(x) ((x)<<8) -#define VFT0_SPEC (1<<7) -#define VFT0_DIFFUSE (1<<6) -#define VFT0_DEPTH_OFFSET (1<<5) -#define VFT0_XYZ (1<<1) -#define VFT0_XYZW (2<<1) -#define VFT0_XY (3<<1) -#define VFT0_XYW (4<<1) -#define VFT0_XYZW_MASK (7<<1) - -/* _3DSTATE_VERTEX_FORMAT_2, p206 */ -#define _3DSTATE_VFT1_CMD (CMD_3D|(0x0a<<24)) -#define VFT1_TEX7_FMT(x) ((x)<<14) -#define VFT1_TEX6_FMT(x) ((x)<<12) -#define VFT1_TEX5_FMT(x) ((x)<<10) -#define VFT1_TEX4_FMT(x) ((x)<<8) -#define VFT1_TEX3_FMT(x) ((x)<<6) -#define VFT1_TEX2_FMT(x) ((x)<<4) -#define VFT1_TEX1_FMT(x) ((x)<<2) -#define VFT1_TEX0_FMT(x) (x) -#define VFT1_TEX0_MASK 3 -#define VFT1_TEX1_SHIFT 2 -#define TEXCOORDFMT_2D 0 -#define TEXCOORDFMT_3D 1 -#define TEXCOORDFMT_4D 2 -#define TEXCOORDFMT_1D 3 - -/*New stuff picked up along the way */ - -#define MLC_LOD_BIAS_MASK ((1<<7)-1) - - -/* _3DSTATE_VERTEX_TRANSFORM, p207 */ -#define _3DSTATE_VERTEX_TRANS_CMD (CMD_3D|(0x1d<<24)|(0x8b<<16)|0) -#define _3DSTATE_VERTEX_TRANS_MTX_CMD (CMD_3D|(0x1d<<24)|(0x8b<<16)|6) -/* Dword 1 */ -#define ENABLE_VIEWPORT_TRANSFORM ((1<<31)|(1<<30)) -#define DISABLE_VIEWPORT_TRANSFORM (1<<31) -#define ENABLE_PERSP_DIVIDE ((1<<29)|(1<<28)) -#define DISABLE_PERSP_DIVIDE (1<<29) -#define VRTX_TRANS_LOAD_MATRICES 0x7421 -#define VRTX_TRANS_NO_LOAD_MATRICES 0x0000 -/* Dword 2 -> 7 are matrix elements */ - -/* _3DSTATE_W_STATE, p209 */ -#define _3DSTATE_W_STATE_CMD (CMD_3D|(0x1d<<24)|(0x8d<<16)|1) -/* Dword 1 */ -#define MAGIC_W_STATE_DWORD1 0x00000008 -/* Dword 2 */ -#define WFAR_VALUE(x) (x) - - -/* Stipple command, carried over from the i810, apparently: - */ -#define _3DSTATE_STIPPLE ((0x3<<29)|(0x1d<<24)|(0x83<<16)) -#define ST1_ENABLE (1<<16) -#define ST1_MASK (0xffff) - - - -#define _3DSTATE_LOAD_STATE_IMMEDIATE_2 ((0x3<<29)|(0x1d<<24)|(0x03<<16)) -#define LOAD_TEXTURE_MAP0 (1<<11) -#define LOAD_GLOBAL_COLOR_FACTOR (1<<6) - -#define TM0S0_ADDRESS_MASK 0xfffffffc -#define TM0S0_USE_FENCE (1<<1) - -#define TM0S1_HEIGHT_SHIFT 21 -#define TM0S1_WIDTH_SHIFT 10 -#define TM0S1_PALETTE_SELECT (1<<9) -#define TM0S1_MAPSURF_FORMAT_MASK (0x7 << 6) -#define TM0S1_MAPSURF_FORMAT_SHIFT 6 -#define MAPSURF_8BIT_INDEXED (0<<6) -#define MAPSURF_8BIT (1<<6) -#define MAPSURF_16BIT (2<<6) -#define MAPSURF_32BIT (3<<6) -#define MAPSURF_411 (4<<6) -#define MAPSURF_422 (5<<6) -#define MAPSURF_COMPRESSED (6<<6) -#define MAPSURF_4BIT_INDEXED (7<<6) -#define TM0S1_MT_FORMAT_MASK (0x7 << 3) -#define TM0S1_MT_FORMAT_SHIFT 3 -#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ -#define MT_8BIT_IDX_RGB565 (0<<3) /* SURFACE_8BIT_INDEXED */ -#define MT_8BIT_IDX_ARGB1555 (1<<3) -#define MT_8BIT_IDX_ARGB4444 (2<<3) -#define MT_8BIT_IDX_AY88 (3<<3) -#define MT_8BIT_IDX_ABGR8888 (4<<3) -#define MT_8BIT_IDX_BUMP_88DVDU (5<<3) -#define MT_8BIT_IDX_BUMP_655LDVDU (6<<3) -#define MT_8BIT_IDX_ARGB8888 (7<<3) -#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ -#define MT_8BIT_L8 (1<<3) -#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ -#define MT_16BIT_ARGB1555 (1<<3) -#define MT_16BIT_ARGB4444 (2<<3) -#define MT_16BIT_AY88 (3<<3) -#define MT_16BIT_DIB_ARGB1555_8888 (4<<3) -#define MT_16BIT_BUMP_88DVDU (5<<3) -#define MT_16BIT_BUMP_655LDVDU (6<<3) -#define MT_16BIT_DIB_RGB565_8888 (7<<3) -#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ -#define MT_32BIT_ABGR8888 (1<<3) -#define MT_32BIT_XRGB8888 (2<<3) /* XXX: Guess from i915_reg.h */ -#define MT_32BIT_BUMP_XLDVDU_8888 (6<<3) -#define MT_32BIT_DIB_8888 (7<<3) -#define MT_411_YUV411 (0<<3) /* SURFACE_411 */ -#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ -#define MT_422_YCRCB_NORMAL (1<<3) -#define MT_422_YCRCB_SWAPUV (2<<3) -#define MT_422_YCRCB_SWAPUVY (3<<3) -#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ -#define MT_COMPRESS_DXT2_3 (1<<3) -#define MT_COMPRESS_DXT4_5 (2<<3) -#define MT_COMPRESS_FXT1 (3<<3) -#define TM0S1_COLORSPACE_CONVERSION (1 << 2) -#define TM0S1_TILED_SURFACE (1 << 1) -#define TM0S1_TILE_WALK (1 << 0) - -#define TM0S2_PITCH_SHIFT 21 -#define TM0S2_CUBE_FACE_ENA_SHIFT 15 -#define TM0S2_CUBE_FACE_ENA_MASK (1<<15) -#define TM0S2_MAP_FORMAT (1<<14) -#define TM0S2_VERTICAL_LINE_STRIDE (1<<13) -#define TM0S2_VERITCAL_LINE_STRIDE_OFF (1<<12) -#define TM0S2_OUTPUT_CHAN_SHIFT 10 -#define TM0S2_OUTPUT_CHAN_MASK (3<<10) - -#define TM0S3_MIP_FILTER_MASK (0x3<<30) -#define TM0S3_MIP_FILTER_SHIFT 30 -#define MIPFILTER_NONE 0 -#define MIPFILTER_NEAREST 1 -#define MIPFILTER_LINEAR 3 -#define TM0S3_MAG_FILTER_MASK (0x3<<28) -#define TM0S3_MAG_FILTER_SHIFT 28 -#define TM0S3_MIN_FILTER_MASK (0x3<<26) -#define TM0S3_MIN_FILTER_SHIFT 26 -#define FILTER_NEAREST 0 -#define FILTER_LINEAR 1 -#define FILTER_ANISOTROPIC 2 - -#define TM0S3_LOD_BIAS_SHIFT 17 -#define TM0S3_LOD_BIAS_MASK (0x1ff<<17) -#define TM0S3_MAX_MIP_SHIFT 9 -#define TM0S3_MAX_MIP_MASK (0xff<<9) -#define TM0S3_MIN_MIP_SHIFT 3 -#define TM0S3_MIN_MIP_MASK (0x3f<<3) -#define TM0S3_KILL_PIXEL (1<<2) -#define TM0S3_KEYED_FILTER (1<<1) -#define TM0S3_CHROMA_KEY (1<<0) - - -/* _3DSTATE_MAP_TEXEL_STREAM, p188 */ -#define _3DSTATE_MAP_TEX_STREAM_CMD (CMD_3D|(0x1c<<24)|(0x05<<19)) -#define DISABLE_TEX_STREAM_BUMP (1<<12) -#define ENABLE_TEX_STREAM_BUMP ((1<<12)|(1<<11)) -#define TEX_MODIFY_UNIT_0 0 -#define TEX_MODIFY_UNIT_1 (1<<8) -#define ENABLE_TEX_STREAM_COORD_SET (1<<7) -#define TEX_STREAM_COORD_SET(x) ((x)<<4) -#define ENABLE_TEX_STREAM_MAP_IDX (1<<3) -#define TEX_STREAM_MAP_IDX(x) (x) - - -#define MI_FLUSH ((0<<29)|(4<<23)) -#define FLUSH_MAP_CACHE (1<<0) - -#endif diff --git a/src/mesa/drivers/dri/i915tex/i830_state.c b/src/mesa/drivers/dri/i915tex/i830_state.c deleted file mode 100644 index 490d681429..0000000000 --- a/src/mesa/drivers/dri/i915tex/i830_state.c +++ /dev/null @@ -1,1106 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "enums.h" -#include "dd.h" - -#include "texmem.h" - -#include "drivers/common/driverfuncs.h" - -#include "intel_screen.h" -#include "intel_batchbuffer.h" -#include "intel_fbo.h" - -#include "i830_context.h" -#include "i830_reg.h" - -#define FILE_DEBUG_FLAG DEBUG_STATE - -static void -i830StencilFuncSeparate(GLcontext * ctx, GLenum face, GLenum func, GLint ref, - GLuint mask) -{ - struct i830_context *i830 = i830_context(ctx); - int test = intel_translate_compare_func(func); - - mask = mask & 0xff; - - DBG("%s : func: %s, ref : 0x%x, mask: 0x%x\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(func), ref, mask); - - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_TEST_MASK; - i830->state.Ctx[I830_CTXREG_STATE4] |= (ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(mask)); - i830->state.Ctx[I830_CTXREG_STENCILTST] &= ~(STENCIL_REF_VALUE_MASK | - ENABLE_STENCIL_TEST_FUNC_MASK); - i830->state.Ctx[I830_CTXREG_STENCILTST] |= (ENABLE_STENCIL_REF_VALUE | - ENABLE_STENCIL_TEST_FUNC | - STENCIL_REF_VALUE(ref) | - STENCIL_TEST_FUNC(test)); -} - -static void -i830StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) -{ - struct i830_context *i830 = i830_context(ctx); - - DBG("%s : mask 0x%x\n", __FUNCTION__, mask); - - mask = mask & 0xff; - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK; - i830->state.Ctx[I830_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(mask)); -} - -static void -i830StencilOpSeparate(GLcontext * ctx, GLenum face, GLenum fail, GLenum zfail, - GLenum zpass) -{ - struct i830_context *i830 = i830_context(ctx); - int fop, dfop, dpop; - - DBG("%s: fail : %s, zfail: %s, zpass : %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(fail), - _mesa_lookup_enum_by_nr(zfail), - _mesa_lookup_enum_by_nr(zpass)); - - fop = 0; - dfop = 0; - dpop = 0; - - switch (fail) { - case GL_KEEP: - fop = STENCILOP_KEEP; - break; - case GL_ZERO: - fop = STENCILOP_ZERO; - break; - case GL_REPLACE: - fop = STENCILOP_REPLACE; - break; - case GL_INCR: - fop = STENCILOP_INCRSAT; - break; - case GL_DECR: - fop = STENCILOP_DECRSAT; - break; - case GL_INCR_WRAP: - fop = STENCILOP_INCR; - break; - case GL_DECR_WRAP: - fop = STENCILOP_DECR; - break; - case GL_INVERT: - fop = STENCILOP_INVERT; - break; - default: - break; - } - switch (zfail) { - case GL_KEEP: - dfop = STENCILOP_KEEP; - break; - case GL_ZERO: - dfop = STENCILOP_ZERO; - break; - case GL_REPLACE: - dfop = STENCILOP_REPLACE; - break; - case GL_INCR: - dfop = STENCILOP_INCRSAT; - break; - case GL_DECR: - dfop = STENCILOP_DECRSAT; - break; - case GL_INCR_WRAP: - dfop = STENCILOP_INCR; - break; - case GL_DECR_WRAP: - dfop = STENCILOP_DECR; - break; - case GL_INVERT: - dfop = STENCILOP_INVERT; - break; - default: - break; - } - switch (zpass) { - case GL_KEEP: - dpop = STENCILOP_KEEP; - break; - case GL_ZERO: - dpop = STENCILOP_ZERO; - break; - case GL_REPLACE: - dpop = STENCILOP_REPLACE; - break; - case GL_INCR: - dpop = STENCILOP_INCRSAT; - break; - case GL_DECR: - dpop = STENCILOP_DECRSAT; - break; - case GL_INCR_WRAP: - dpop = STENCILOP_INCR; - break; - case GL_DECR_WRAP: - dpop = STENCILOP_DECR; - break; - case GL_INVERT: - dpop = STENCILOP_INVERT; - break; - default: - break; - } - - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_STENCILTST] &= ~(STENCIL_OPS_MASK); - i830->state.Ctx[I830_CTXREG_STENCILTST] |= (ENABLE_STENCIL_PARMS | - STENCIL_FAIL_OP(fop) | - STENCIL_PASS_DEPTH_FAIL_OP - (dfop) | - STENCIL_PASS_DEPTH_PASS_OP - (dpop)); -} - -static void -i830AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) -{ - struct i830_context *i830 = i830_context(ctx); - int test = intel_translate_compare_func(func); - GLubyte refByte; - GLuint refInt; - - UNCLAMPED_FLOAT_TO_UBYTE(refByte, ref); - refInt = (GLuint) refByte; - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_STATE2] &= ~ALPHA_TEST_REF_MASK; - i830->state.Ctx[I830_CTXREG_STATE2] |= (ENABLE_ALPHA_TEST_FUNC | - ENABLE_ALPHA_REF_VALUE | - ALPHA_TEST_FUNC(test) | - ALPHA_REF_VALUE(refInt)); -} - -/** - * Makes sure that the proper enables are set for LogicOp, Independant Alpha - * Blend, and Blending. It needs to be called from numerous places where we - * could change the LogicOp or Independant Alpha Blend without subsequent - * calls to glEnable. - * - * \todo - * This function is substantially different from the old i830-specific driver. - * I'm not sure which is correct. - */ -static void -i830EvalLogicOpBlendState(GLcontext * ctx) -{ - struct i830_context *i830 = i830_context(ctx); - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - - if (RGBA_LOGICOP_ENABLED(ctx)) { - i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~(ENABLE_COLOR_BLEND | - ENABLE_LOGIC_OP_MASK); - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= (DISABLE_COLOR_BLEND | - ENABLE_LOGIC_OP); - } - else if (ctx->Color.BlendEnabled) { - i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~(ENABLE_COLOR_BLEND | - ENABLE_LOGIC_OP_MASK); - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= (ENABLE_COLOR_BLEND | - DISABLE_LOGIC_OP); - } - else { - i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~(ENABLE_COLOR_BLEND | - ENABLE_LOGIC_OP_MASK); - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= (DISABLE_COLOR_BLEND | - DISABLE_LOGIC_OP); - } -} - -static void -i830BlendColor(GLcontext * ctx, const GLfloat color[4]) -{ - struct i830_context *i830 = i830_context(ctx); - GLubyte r, g, b, a; - - DBG("%s\n", __FUNCTION__); - - UNCLAMPED_FLOAT_TO_UBYTE(r, color[RCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(g, color[GCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(b, color[BCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(a, color[ACOMP]); - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_BLENDCOLOR1] = - (a << 24) | (r << 16) | (g << 8) | b; -} - -/** - * Sets both the blend equation (called "function" in i830 docs) and the - * blend function (called "factor" in i830 docs). This is done in a single - * function because some blend equations (i.e., \c GL_MIN and \c GL_MAX) - * change the interpretation of the blend function. - */ -static void -i830_set_blend_state(GLcontext * ctx) -{ - struct i830_context *i830 = i830_context(ctx); - int funcA; - int funcRGB; - int eqnA; - int eqnRGB; - int iab; - int s1; - - - funcRGB = - SRC_BLND_FACT(intel_translate_blend_factor(ctx->Color.BlendSrcRGB)) - | DST_BLND_FACT(intel_translate_blend_factor(ctx->Color.BlendDstRGB)); - - switch (ctx->Color.BlendEquationRGB) { - case GL_FUNC_ADD: - eqnRGB = BLENDFUNC_ADD; - break; - case GL_MIN: - eqnRGB = BLENDFUNC_MIN; - funcRGB = SRC_BLND_FACT(BLENDFACT_ONE) | DST_BLND_FACT(BLENDFACT_ONE); - break; - case GL_MAX: - eqnRGB = BLENDFUNC_MAX; - funcRGB = SRC_BLND_FACT(BLENDFACT_ONE) | DST_BLND_FACT(BLENDFACT_ONE); - break; - case GL_FUNC_SUBTRACT: - eqnRGB = BLENDFUNC_SUB; - break; - case GL_FUNC_REVERSE_SUBTRACT: - eqnRGB = BLENDFUNC_RVRSE_SUB; - break; - default: - fprintf(stderr, "[%s:%u] Invalid RGB blend equation (0x%04x).\n", - __FUNCTION__, __LINE__, ctx->Color.BlendEquationRGB); - return; - } - - - funcA = SRC_ABLEND_FACT(intel_translate_blend_factor(ctx->Color.BlendSrcA)) - | DST_ABLEND_FACT(intel_translate_blend_factor(ctx->Color.BlendDstA)); - - switch (ctx->Color.BlendEquationA) { - case GL_FUNC_ADD: - eqnA = BLENDFUNC_ADD; - break; - case GL_MIN: - eqnA = BLENDFUNC_MIN; - funcA = SRC_BLND_FACT(BLENDFACT_ONE) | DST_BLND_FACT(BLENDFACT_ONE); - break; - case GL_MAX: - eqnA = BLENDFUNC_MAX; - funcA = SRC_BLND_FACT(BLENDFACT_ONE) | DST_BLND_FACT(BLENDFACT_ONE); - break; - case GL_FUNC_SUBTRACT: - eqnA = BLENDFUNC_SUB; - break; - case GL_FUNC_REVERSE_SUBTRACT: - eqnA = BLENDFUNC_RVRSE_SUB; - break; - default: - fprintf(stderr, "[%s:%u] Invalid alpha blend equation (0x%04x).\n", - __FUNCTION__, __LINE__, ctx->Color.BlendEquationA); - return; - } - - iab = eqnA | funcA - | _3DSTATE_INDPT_ALPHA_BLEND_CMD - | ENABLE_SRC_ABLEND_FACTOR | ENABLE_DST_ABLEND_FACTOR - | ENABLE_ALPHA_BLENDFUNC; - s1 = eqnRGB | funcRGB - | _3DSTATE_MODES_1_CMD - | ENABLE_SRC_BLND_FACTOR | ENABLE_DST_BLND_FACTOR - | ENABLE_COLR_BLND_FUNC; - - if ((eqnA | funcA) != (eqnRGB | funcRGB)) - iab |= ENABLE_INDPT_ALPHA_BLEND; - else - iab |= DISABLE_INDPT_ALPHA_BLEND; - - if (iab != i830->state.Ctx[I830_CTXREG_IALPHAB] || - s1 != i830->state.Ctx[I830_CTXREG_STATE1]) { - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_IALPHAB] = iab; - i830->state.Ctx[I830_CTXREG_STATE1] = s1; - } - - /* This will catch a logicop blend equation. It will also ensure - * independant alpha blend is really in the correct state (either enabled - * or disabled) if blending is already enabled. - */ - - i830EvalLogicOpBlendState(ctx); - - if (0) { - fprintf(stderr, - "[%s:%u] STATE1: 0x%08x IALPHAB: 0x%08x blend is %sabled\n", - __FUNCTION__, __LINE__, i830->state.Ctx[I830_CTXREG_STATE1], - i830->state.Ctx[I830_CTXREG_IALPHAB], - (ctx->Color.BlendEnabled) ? "en" : "dis"); - } -} - - -static void -i830BlendEquationSeparate(GLcontext * ctx, GLenum modeRGB, GLenum modeA) -{ - DBG("%s -> %s, %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(modeRGB), - _mesa_lookup_enum_by_nr(modeA)); - - (void) modeRGB; - (void) modeA; - i830_set_blend_state(ctx); -} - - -static void -i830BlendFuncSeparate(GLcontext * ctx, GLenum sfactorRGB, - GLenum dfactorRGB, GLenum sfactorA, GLenum dfactorA) -{ - DBG("%s -> RGB(%s, %s) A(%s, %s)\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(sfactorRGB), - _mesa_lookup_enum_by_nr(dfactorRGB), - _mesa_lookup_enum_by_nr(sfactorA), - _mesa_lookup_enum_by_nr(dfactorA)); - - (void) sfactorRGB; - (void) dfactorRGB; - (void) sfactorA; - (void) dfactorA; - i830_set_blend_state(ctx); -} - - - -static void -i830DepthFunc(GLcontext * ctx, GLenum func) -{ - struct i830_context *i830 = i830_context(ctx); - int test = intel_translate_compare_func(func); - - DBG("%s\n", __FUNCTION__); - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_STATE3] &= ~DEPTH_TEST_FUNC_MASK; - i830->state.Ctx[I830_CTXREG_STATE3] |= (ENABLE_DEPTH_TEST_FUNC | - DEPTH_TEST_FUNC(test)); -} - -static void -i830DepthMask(GLcontext * ctx, GLboolean flag) -{ - struct i830_context *i830 = i830_context(ctx); - - DBG("%s flag (%d)\n", __FUNCTION__, flag); - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - - i830->state.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_DIS_DEPTH_WRITE_MASK; - - if (flag && ctx->Depth.Test) - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_DEPTH_WRITE; - else - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= DISABLE_DEPTH_WRITE; -} - -/* ============================================================= - * Polygon stipple - * - * The i830 supports a 4x4 stipple natively, GL wants 32x32. - * Fortunately stipple is usually a repeating pattern. - */ -static void -i830PolygonStipple(GLcontext * ctx, const GLubyte * mask) -{ - struct i830_context *i830 = i830_context(ctx); - const GLubyte *m = mask; - GLubyte p[4]; - int i, j, k; - int active = (ctx->Polygon.StippleFlag && - i830->intel.reduced_primitive == GL_TRIANGLES); - GLuint newMask; - - if (active) { - I830_STATECHANGE(i830, I830_UPLOAD_STIPPLE); - i830->state.Stipple[I830_STPREG_ST1] &= ~ST1_ENABLE; - } - - p[0] = mask[12] & 0xf; - p[0] |= p[0] << 4; - p[1] = mask[8] & 0xf; - p[1] |= p[1] << 4; - p[2] = mask[4] & 0xf; - p[2] |= p[2] << 4; - p[3] = mask[0] & 0xf; - p[3] |= p[3] << 4; - - for (k = 0; k < 8; k++) - for (j = 3; j >= 0; j--) - for (i = 0; i < 4; i++, m++) - if (*m != p[j]) { - i830->intel.hw_stipple = 0; - return; - } - - newMask = (((p[0] & 0xf) << 0) | - ((p[1] & 0xf) << 4) | - ((p[2] & 0xf) << 8) | ((p[3] & 0xf) << 12)); - - - if (newMask == 0xffff || newMask == 0x0) { - /* this is needed to make conform pass */ - i830->intel.hw_stipple = 0; - return; - } - - i830->state.Stipple[I830_STPREG_ST1] &= ~0xffff; - i830->state.Stipple[I830_STPREG_ST1] |= newMask; - i830->intel.hw_stipple = 1; - - if (active) - i830->state.Stipple[I830_STPREG_ST1] |= ST1_ENABLE; -} - - -/* ============================================================= - * Hardware clipping - */ -static void -i830Scissor(GLcontext * ctx, GLint x, GLint y, GLsizei w, GLsizei h) -{ - struct i830_context *i830 = i830_context(ctx); - int x1, y1, x2, y2; - - if (!ctx->DrawBuffer) - return; - - DBG("%s %d,%d %dx%d\n", __FUNCTION__, x, y, w, h); - - if (ctx->DrawBuffer->Name == 0) { - x1 = x; - y1 = ctx->DrawBuffer->Height - (y + h); - x2 = x + w - 1; - y2 = y1 + h - 1; - DBG("%s %d..%d,%d..%d (inverted)\n", __FUNCTION__, x1, x2, y1, y2); - } - else { - /* FBO - not inverted - */ - x1 = x; - y1 = y; - x2 = x + w - 1; - y2 = y + h - 1; - DBG("%s %d..%d,%d..%d (not inverted)\n", __FUNCTION__, x1, x2, y1, y2); - } - - x1 = CLAMP(x1, 0, ctx->DrawBuffer->Width - 1); - y1 = CLAMP(y1, 0, ctx->DrawBuffer->Height - 1); - x2 = CLAMP(x2, 0, ctx->DrawBuffer->Width - 1); - y2 = CLAMP(y2, 0, ctx->DrawBuffer->Height - 1); - - DBG("%s %d..%d,%d..%d (clamped)\n", __FUNCTION__, x1, x2, y1, y2); - - I830_STATECHANGE(i830, I830_UPLOAD_BUFFERS); - i830->state.Buffer[I830_DESTREG_SR1] = (y1 << 16) | (x1 & 0xffff); - i830->state.Buffer[I830_DESTREG_SR2] = (y2 << 16) | (x2 & 0xffff); -} - -static void -i830LogicOp(GLcontext * ctx, GLenum opcode) -{ - struct i830_context *i830 = i830_context(ctx); - int tmp = intel_translate_logic_op(opcode); - - DBG("%s\n", __FUNCTION__); - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_STATE4] &= ~LOGICOP_MASK; - i830->state.Ctx[I830_CTXREG_STATE4] |= LOGIC_OP_FUNC(tmp); -} - - - -static void -i830CullFaceFrontFace(GLcontext * ctx, GLenum unused) -{ - struct i830_context *i830 = i830_context(ctx); - GLuint mode; - - DBG("%s\n", __FUNCTION__); - - if (!ctx->Polygon.CullFlag) { - mode = CULLMODE_NONE; - } - else if (ctx->Polygon.CullFaceMode != GL_FRONT_AND_BACK) { - mode = CULLMODE_CW; - - if (ctx->Polygon.CullFaceMode == GL_FRONT) - mode ^= (CULLMODE_CW ^ CULLMODE_CCW); - if (ctx->Polygon.FrontFace != GL_CCW) - mode ^= (CULLMODE_CW ^ CULLMODE_CCW); - } - else { - mode = CULLMODE_BOTH; - } - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_STATE3] &= ~CULLMODE_MASK; - i830->state.Ctx[I830_CTXREG_STATE3] |= ENABLE_CULL_MODE | mode; -} - -static void -i830LineWidth(GLcontext * ctx, GLfloat widthf) -{ - struct i830_context *i830 = i830_context(ctx); - int width; - int state5; - - DBG("%s\n", __FUNCTION__); - - width = (int) (widthf * 2); - CLAMP_SELF(width, 1, 15); - - state5 = i830->state.Ctx[I830_CTXREG_STATE5] & ~FIXED_LINE_WIDTH_MASK; - state5 |= (ENABLE_FIXED_LINE_WIDTH | FIXED_LINE_WIDTH(width)); - - if (state5 != i830->state.Ctx[I830_CTXREG_STATE5]) { - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_STATE5] = state5; - } -} - -static void -i830PointSize(GLcontext * ctx, GLfloat size) -{ - struct i830_context *i830 = i830_context(ctx); - GLint point_size = (int) size; - - DBG("%s\n", __FUNCTION__); - - CLAMP_SELF(point_size, 1, 256); - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_STATE5] &= ~FIXED_POINT_WIDTH_MASK; - i830->state.Ctx[I830_CTXREG_STATE5] |= (ENABLE_FIXED_POINT_WIDTH | - FIXED_POINT_WIDTH(point_size)); -} - - -/* ============================================================= - * Color masks - */ - -static void -i830ColorMask(GLcontext * ctx, - GLboolean r, GLboolean g, GLboolean b, GLboolean a) -{ - struct i830_context *i830 = i830_context(ctx); - GLuint tmp = 0; - - DBG("%s r(%d) g(%d) b(%d) a(%d)\n", __FUNCTION__, r, g, b, a); - - tmp = ((i830->state.Ctx[I830_CTXREG_ENABLES_2] & ~WRITEMASK_MASK) | - ENABLE_COLOR_MASK | - ENABLE_COLOR_WRITE | - ((!r) << WRITEMASK_RED_SHIFT) | - ((!g) << WRITEMASK_GREEN_SHIFT) | - ((!b) << WRITEMASK_BLUE_SHIFT) | ((!a) << WRITEMASK_ALPHA_SHIFT)); - - if (tmp != i830->state.Ctx[I830_CTXREG_ENABLES_2]) { - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_ENABLES_2] = tmp; - } -} - -static void -update_specular(GLcontext * ctx) -{ - struct i830_context *i830 = i830_context(ctx); - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_SPEC_ADD_MASK; - - if (NEED_SECONDARY_COLOR(ctx)) - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_SPEC_ADD; - else - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_SPEC_ADD; -} - -static void -i830LightModelfv(GLcontext * ctx, GLenum pname, const GLfloat * param) -{ - DBG("%s\n", __FUNCTION__); - - if (pname == GL_LIGHT_MODEL_COLOR_CONTROL) { - update_specular(ctx); - } -} - -/* In Mesa 3.5 we can reliably do native flatshading. - */ -static void -i830ShadeModel(GLcontext * ctx, GLenum mode) -{ - struct i830_context *i830 = i830_context(ctx); - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - - -#define SHADE_MODE_MASK ((1<<10)|(1<<8)|(1<<6)|(1<<4)) - - i830->state.Ctx[I830_CTXREG_STATE3] &= ~SHADE_MODE_MASK; - - if (mode == GL_FLAT) { - i830->state.Ctx[I830_CTXREG_STATE3] |= - (ALPHA_SHADE_MODE(SHADE_MODE_FLAT) | FOG_SHADE_MODE(SHADE_MODE_FLAT) - | SPEC_SHADE_MODE(SHADE_MODE_FLAT) | - COLOR_SHADE_MODE(SHADE_MODE_FLAT)); - } - else { - i830->state.Ctx[I830_CTXREG_STATE3] |= - (ALPHA_SHADE_MODE(SHADE_MODE_LINEAR) | - FOG_SHADE_MODE(SHADE_MODE_LINEAR) | - SPEC_SHADE_MODE(SHADE_MODE_LINEAR) | - COLOR_SHADE_MODE(SHADE_MODE_LINEAR)); - } -} - -/* ============================================================= - * Fog - */ -static void -i830Fogfv(GLcontext * ctx, GLenum pname, const GLfloat * param) -{ - struct i830_context *i830 = i830_context(ctx); - - DBG("%s\n", __FUNCTION__); - - if (pname == GL_FOG_COLOR) { - GLuint color = (((GLubyte) (ctx->Fog.Color[0] * 255.0F) << 16) | - ((GLubyte) (ctx->Fog.Color[1] * 255.0F) << 8) | - ((GLubyte) (ctx->Fog.Color[2] * 255.0F) << 0)); - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_FOGCOLOR] = - (_3DSTATE_FOG_COLOR_CMD | color); - } -} - -/* ============================================================= - */ - -static void -i830Enable(GLcontext * ctx, GLenum cap, GLboolean state) -{ - struct i830_context *i830 = i830_context(ctx); - - switch (cap) { - case GL_LIGHTING: - case GL_COLOR_SUM: - update_specular(ctx); - break; - - case GL_ALPHA_TEST: - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_ALPHA_TEST_MASK; - if (state) - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_ALPHA_TEST; - else - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_ALPHA_TEST; - - break; - - case GL_BLEND: - i830EvalLogicOpBlendState(ctx); - break; - - case GL_COLOR_LOGIC_OP: - i830EvalLogicOpBlendState(ctx); - - /* Logicop doesn't seem to work at 16bpp: - */ - if (i830->intel.intelScreen->cpp == 2) - FALLBACK(&i830->intel, I830_FALLBACK_LOGICOP, state); - break; - - case GL_DITHER: - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_ENABLES_2] &= ~ENABLE_DITHER; - - if (state) - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_DITHER; - else - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= DISABLE_DITHER; - break; - - case GL_DEPTH_TEST: - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_DEPTH_TEST_MASK; - - if (state) - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_DEPTH_TEST; - else - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_DEPTH_TEST; - - /* Also turn off depth writes when GL_DEPTH_TEST is disabled: - */ - i830DepthMask(ctx, ctx->Depth.Mask); - break; - - case GL_SCISSOR_TEST: - I830_STATECHANGE(i830, I830_UPLOAD_BUFFERS); - - if (state) - i830->state.Buffer[I830_DESTREG_SENABLE] = - (_3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT); - else - i830->state.Buffer[I830_DESTREG_SENABLE] = - (_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); - - break; - - case GL_LINE_SMOOTH: - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - - i830->state.Ctx[I830_CTXREG_AA] &= ~AA_LINE_ENABLE; - if (state) - i830->state.Ctx[I830_CTXREG_AA] |= AA_LINE_ENABLE; - else - i830->state.Ctx[I830_CTXREG_AA] |= AA_LINE_DISABLE; - break; - - case GL_FOG: - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_DIS_FOG_MASK; - if (state) - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_FOG; - else - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_FOG; - break; - - case GL_CULL_FACE: - i830CullFaceFrontFace(ctx, 0); - break; - - case GL_TEXTURE_2D: - break; - - case GL_STENCIL_TEST: - { - GLboolean hw_stencil = GL_FALSE; - if (ctx->DrawBuffer) { - struct intel_renderbuffer *irbStencil - = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL); - hw_stencil = (irbStencil && irbStencil->region); - } - if (hw_stencil) { - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - - if (state) { - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= ENABLE_STENCIL_TEST; - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= ENABLE_STENCIL_WRITE; - } - else { - i830->state.Ctx[I830_CTXREG_ENABLES_1] &= ~ENABLE_STENCIL_TEST; - i830->state.Ctx[I830_CTXREG_ENABLES_2] &= - ~ENABLE_STENCIL_WRITE; - i830->state.Ctx[I830_CTXREG_ENABLES_1] |= DISABLE_STENCIL_TEST; - i830->state.Ctx[I830_CTXREG_ENABLES_2] |= - DISABLE_STENCIL_WRITE; - } - } - else { - FALLBACK(&i830->intel, I830_FALLBACK_STENCIL, state); - } - } - break; - - case GL_POLYGON_STIPPLE: - /* The stipple command worked on my 855GM box, but not my 845G. - * I'll do more testing later to find out exactly which hardware - * supports it. Disabled for now. - */ - if (i830->intel.hw_stipple && - i830->intel.reduced_primitive == GL_TRIANGLES) { - I830_STATECHANGE(i830, I830_UPLOAD_STIPPLE); - i830->state.Stipple[I830_STPREG_ST1] &= ~ST1_ENABLE; - if (state) - i830->state.Stipple[I830_STPREG_ST1] |= ST1_ENABLE; - } - break; - - default: - ; - } -} - - -static void -i830_init_packets(struct i830_context *i830) -{ - - /* Zero all state */ - memset(&i830->state, 0, sizeof(i830->state)); - - /* Set default blend state */ - i830->state.TexBlend[0][0] = (_3DSTATE_MAP_BLEND_OP_CMD(0) | - TEXPIPE_COLOR | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - DISABLE_TEX_CNTRL_STAGE | - TEXOP_SCALE_1X | - TEXOP_MODIFY_PARMS | - TEXOP_LAST_STAGE | TEXBLENDOP_ARG1); - i830->state.TexBlend[0][1] = (_3DSTATE_MAP_BLEND_OP_CMD(0) | - TEXPIPE_ALPHA | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - TEXOP_SCALE_1X | - TEXOP_MODIFY_PARMS | TEXBLENDOP_ARG1); - i830->state.TexBlend[0][2] = (_3DSTATE_MAP_BLEND_ARG_CMD(0) | - TEXPIPE_COLOR | - TEXBLEND_ARG1 | - TEXBLENDARG_MODIFY_PARMS | - TEXBLENDARG_DIFFUSE); - i830->state.TexBlend[0][3] = (_3DSTATE_MAP_BLEND_ARG_CMD(0) | - TEXPIPE_ALPHA | - TEXBLEND_ARG1 | - TEXBLENDARG_MODIFY_PARMS | - TEXBLENDARG_DIFFUSE); - - i830->state.TexBlendWordsUsed[0] = 4; - - - i830->state.Ctx[I830_CTXREG_VF] = 0; - i830->state.Ctx[I830_CTXREG_VF2] = 0; - - i830->state.Ctx[I830_CTXREG_AA] = (_3DSTATE_AA_CMD | - AA_LINE_ECAAR_WIDTH_ENABLE | - AA_LINE_ECAAR_WIDTH_1_0 | - AA_LINE_REGION_WIDTH_ENABLE | - AA_LINE_REGION_WIDTH_1_0 | - AA_LINE_DISABLE); - - i830->state.Ctx[I830_CTXREG_ENABLES_1] = (_3DSTATE_ENABLES_1_CMD | - DISABLE_LOGIC_OP | - DISABLE_STENCIL_TEST | - DISABLE_DEPTH_BIAS | - DISABLE_SPEC_ADD | - DISABLE_FOG | - DISABLE_ALPHA_TEST | - DISABLE_COLOR_BLEND | - DISABLE_DEPTH_TEST); - -#if 000 /* XXX all the stencil enable state is set in i830Enable(), right? */ - if (i830->intel.hw_stencil) { - i830->state.Ctx[I830_CTXREG_ENABLES_2] = (_3DSTATE_ENABLES_2_CMD | - ENABLE_STENCIL_WRITE | - ENABLE_TEX_CACHE | - ENABLE_DITHER | - ENABLE_COLOR_MASK | - /* set no color comps disabled */ - ENABLE_COLOR_WRITE | - ENABLE_DEPTH_WRITE); - } - else -#endif - { - i830->state.Ctx[I830_CTXREG_ENABLES_2] = (_3DSTATE_ENABLES_2_CMD | - DISABLE_STENCIL_WRITE | - ENABLE_TEX_CACHE | - ENABLE_DITHER | - ENABLE_COLOR_MASK | - /* set no color comps disabled */ - ENABLE_COLOR_WRITE | - ENABLE_DEPTH_WRITE); - } - - i830->state.Ctx[I830_CTXREG_STATE1] = (_3DSTATE_MODES_1_CMD | - ENABLE_COLR_BLND_FUNC | - BLENDFUNC_ADD | - ENABLE_SRC_BLND_FACTOR | - SRC_BLND_FACT(BLENDFACT_ONE) | - ENABLE_DST_BLND_FACTOR | - DST_BLND_FACT(BLENDFACT_ZERO)); - - i830->state.Ctx[I830_CTXREG_STATE2] = (_3DSTATE_MODES_2_CMD | - ENABLE_GLOBAL_DEPTH_BIAS | - GLOBAL_DEPTH_BIAS(0) | - ENABLE_ALPHA_TEST_FUNC | - ALPHA_TEST_FUNC(COMPAREFUNC_ALWAYS) - | ALPHA_REF_VALUE(0)); - - i830->state.Ctx[I830_CTXREG_STATE3] = (_3DSTATE_MODES_3_CMD | - ENABLE_DEPTH_TEST_FUNC | - DEPTH_TEST_FUNC(COMPAREFUNC_LESS) | - ENABLE_ALPHA_SHADE_MODE | - ALPHA_SHADE_MODE(SHADE_MODE_LINEAR) - | ENABLE_FOG_SHADE_MODE | - FOG_SHADE_MODE(SHADE_MODE_LINEAR) | - ENABLE_SPEC_SHADE_MODE | - SPEC_SHADE_MODE(SHADE_MODE_LINEAR) | - ENABLE_COLOR_SHADE_MODE | - COLOR_SHADE_MODE(SHADE_MODE_LINEAR) - | ENABLE_CULL_MODE | CULLMODE_NONE); - - i830->state.Ctx[I830_CTXREG_STATE4] = (_3DSTATE_MODES_4_CMD | - ENABLE_LOGIC_OP_FUNC | - LOGIC_OP_FUNC(LOGICOP_COPY) | - ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(0xff) | - ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(0xff)); - - i830->state.Ctx[I830_CTXREG_STENCILTST] = (_3DSTATE_STENCIL_TEST_CMD | - ENABLE_STENCIL_PARMS | - STENCIL_FAIL_OP(STENCILOP_KEEP) - | - STENCIL_PASS_DEPTH_FAIL_OP - (STENCILOP_KEEP) | - STENCIL_PASS_DEPTH_PASS_OP - (STENCILOP_KEEP) | - ENABLE_STENCIL_TEST_FUNC | - STENCIL_TEST_FUNC - (COMPAREFUNC_ALWAYS) | - ENABLE_STENCIL_REF_VALUE | - STENCIL_REF_VALUE(0)); - - i830->state.Ctx[I830_CTXREG_STATE5] = (_3DSTATE_MODES_5_CMD | FLUSH_TEXTURE_CACHE | ENABLE_SPRITE_POINT_TEX | SPRITE_POINT_TEX_OFF | ENABLE_FIXED_LINE_WIDTH | FIXED_LINE_WIDTH(0x2) | /* 1.0 */ - ENABLE_FIXED_POINT_WIDTH | - FIXED_POINT_WIDTH(1)); - - i830->state.Ctx[I830_CTXREG_IALPHAB] = (_3DSTATE_INDPT_ALPHA_BLEND_CMD | - DISABLE_INDPT_ALPHA_BLEND | - ENABLE_ALPHA_BLENDFUNC | - ABLENDFUNC_ADD); - - i830->state.Ctx[I830_CTXREG_FOGCOLOR] = (_3DSTATE_FOG_COLOR_CMD | - FOG_COLOR_RED(0) | - FOG_COLOR_GREEN(0) | - FOG_COLOR_BLUE(0)); - - i830->state.Ctx[I830_CTXREG_BLENDCOLOR0] = _3DSTATE_CONST_BLEND_COLOR_CMD; - i830->state.Ctx[I830_CTXREG_BLENDCOLOR1] = 0; - - i830->state.Ctx[I830_CTXREG_MCSB0] = _3DSTATE_MAP_COORD_SETBIND_CMD; - i830->state.Ctx[I830_CTXREG_MCSB1] = (TEXBIND_SET3(TEXCOORDSRC_VTXSET_3) | - TEXBIND_SET2(TEXCOORDSRC_VTXSET_2) | - TEXBIND_SET1(TEXCOORDSRC_VTXSET_1) | - TEXBIND_SET0(TEXCOORDSRC_VTXSET_0)); - - - i830->state.Stipple[I830_STPREG_ST0] = _3DSTATE_STIPPLE; - - i830->state.Buffer[I830_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD; - -#if 0 - switch (screen->fbFormat) { - case DV_PF_565: - i830->state.Buffer[I830_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - screen->fbFormat | - DEPTH_IS_Z | - DEPTH_FRMT_16_FIXED); - break; - case DV_PF_8888: - i830->state.Buffer[I830_DESTREG_DV1] = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - screen->fbFormat | - DEPTH_IS_Z | - DEPTH_FRMT_24_FIXED_8_OTHER); - break; - } -#endif - i830->state.Buffer[I830_DESTREG_SENABLE] = (_3DSTATE_SCISSOR_ENABLE_CMD | - DISABLE_SCISSOR_RECT); - i830->state.Buffer[I830_DESTREG_SR0] = _3DSTATE_SCISSOR_RECT_0_CMD; - i830->state.Buffer[I830_DESTREG_SR1] = 0; - i830->state.Buffer[I830_DESTREG_SR2] = 0; -} - - -void -i830InitStateFuncs(struct dd_function_table *functions) -{ - functions->AlphaFunc = i830AlphaFunc; - functions->BlendColor = i830BlendColor; - functions->BlendEquationSeparate = i830BlendEquationSeparate; - functions->BlendFuncSeparate = i830BlendFuncSeparate; - functions->ColorMask = i830ColorMask; - functions->CullFace = i830CullFaceFrontFace; - functions->DepthFunc = i830DepthFunc; - functions->DepthMask = i830DepthMask; - functions->Enable = i830Enable; - functions->Fogfv = i830Fogfv; - functions->FrontFace = i830CullFaceFrontFace; - functions->LightModelfv = i830LightModelfv; - functions->LineWidth = i830LineWidth; - functions->LogicOpcode = i830LogicOp; - functions->PointSize = i830PointSize; - functions->PolygonStipple = i830PolygonStipple; - functions->Scissor = i830Scissor; - functions->ShadeModel = i830ShadeModel; - functions->StencilFuncSeparate = i830StencilFuncSeparate; - functions->StencilMaskSeparate = i830StencilMaskSeparate; - functions->StencilOpSeparate = i830StencilOpSeparate; -} - -void -i830InitState(struct i830_context *i830) -{ - GLcontext *ctx = &i830->intel.ctx; - - i830_init_packets(i830); - - _mesa_init_driver_state(ctx); - - memcpy(&i830->initial, &i830->state, sizeof(i830->state)); - - i830->current = &i830->state; - i830->state.emitted = 0; - i830->state.active = (I830_UPLOAD_INVARIENT | - I830_UPLOAD_TEXBLEND(0) | - I830_UPLOAD_STIPPLE | - I830_UPLOAD_CTX | I830_UPLOAD_BUFFERS); -} diff --git a/src/mesa/drivers/dri/i915tex/i830_tex.c b/src/mesa/drivers/dri/i915tex/i830_tex.c deleted file mode 100644 index fed464d1aa..0000000000 --- a/src/mesa/drivers/dri/i915tex/i830_tex.c +++ /dev/null @@ -1,100 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "mtypes.h" -#include "imports.h" -#include "simple_list.h" -#include "enums.h" -#include "image.h" -#include "texstore.h" -#include "texformat.h" -#include "texmem.h" -#include "swrast/swrast.h" - -#include "mm.h" - -#include "intel_ioctl.h" - -#include "i830_context.h" -#include "i830_reg.h" - - - -static void -i830TexEnv(GLcontext * ctx, GLenum target, - GLenum pname, const GLfloat * param) -{ - - switch (pname) { - case GL_TEXTURE_ENV_COLOR: - case GL_TEXTURE_ENV_MODE: - case GL_COMBINE_RGB: - case GL_COMBINE_ALPHA: - case GL_SOURCE0_RGB: - case GL_SOURCE1_RGB: - case GL_SOURCE2_RGB: - case GL_SOURCE0_ALPHA: - case GL_SOURCE1_ALPHA: - case GL_SOURCE2_ALPHA: - case GL_OPERAND0_RGB: - case GL_OPERAND1_RGB: - case GL_OPERAND2_RGB: - case GL_OPERAND0_ALPHA: - case GL_OPERAND1_ALPHA: - case GL_OPERAND2_ALPHA: - case GL_RGB_SCALE: - case GL_ALPHA_SCALE: - break; - - case GL_TEXTURE_LOD_BIAS:{ - struct i830_context *i830 = i830_context(ctx); - GLuint unit = ctx->Texture.CurrentUnit; - int b = (int) ((*param) * 16.0); - if (b > 63) - b = 63; - if (b < -64) - b = -64; - I830_STATECHANGE(i830, I830_UPLOAD_TEX(unit)); - i830->lodbias_tm0s3[unit] = - ((b << TM0S3_LOD_BIAS_SHIFT) & TM0S3_LOD_BIAS_MASK); - break; - } - - default: - break; - } -} - - - - -void -i830InitTextureFuncs(struct dd_function_table *functions) -{ - functions->TexEnv = i830TexEnv; -} diff --git a/src/mesa/drivers/dri/i915tex/i830_texblend.c b/src/mesa/drivers/dri/i915tex/i830_texblend.c deleted file mode 100644 index 58f220eb7c..0000000000 --- a/src/mesa/drivers/dri/i915tex/i830_texblend.c +++ /dev/null @@ -1,463 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "macros.h" -#include "mtypes.h" -#include "simple_list.h" -#include "enums.h" -#include "texformat.h" -#include "texstore.h" - -#include "mm.h" - -#include "intel_screen.h" -#include "intel_ioctl.h" -#include "intel_tex.h" - -#include "i830_context.h" -#include "i830_reg.h" - - -/* ================================================================ - * Texture combine functions - */ -static GLuint -pass_through(GLuint * state, GLuint blendUnit) -{ - state[0] = (_3DSTATE_MAP_BLEND_OP_CMD(blendUnit) | - TEXPIPE_COLOR | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - DISABLE_TEX_CNTRL_STAGE | - TEXOP_SCALE_1X | TEXOP_MODIFY_PARMS | TEXBLENDOP_ARG1); - state[1] = (_3DSTATE_MAP_BLEND_OP_CMD(blendUnit) | - TEXPIPE_ALPHA | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - TEXOP_SCALE_1X | TEXOP_MODIFY_PARMS | TEXBLENDOP_ARG1); - state[2] = (_3DSTATE_MAP_BLEND_ARG_CMD(blendUnit) | - TEXPIPE_COLOR | - TEXBLEND_ARG1 | - TEXBLENDARG_MODIFY_PARMS | TEXBLENDARG_CURRENT); - state[3] = (_3DSTATE_MAP_BLEND_ARG_CMD(blendUnit) | - TEXPIPE_ALPHA | - TEXBLEND_ARG1 | - TEXBLENDARG_MODIFY_PARMS | TEXBLENDARG_CURRENT); - - return 4; -} - -static GLuint -emit_factor(GLuint blendUnit, GLuint * state, GLuint count, - const GLfloat * factor) -{ - GLubyte r, g, b, a; - GLuint col; - - if (0) - fprintf(stderr, "emit constant %d: %.2f %.2f %.2f %.2f\n", - blendUnit, factor[0], factor[1], factor[2], factor[3]); - - UNCLAMPED_FLOAT_TO_UBYTE(r, factor[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, factor[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, factor[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, factor[3]); - - col = ((a << 24) | (r << 16) | (g << 8) | b); - - state[count++] = _3DSTATE_COLOR_FACTOR_N_CMD(blendUnit); - state[count++] = col; - - return count; -} - - -static INLINE GLuint -GetTexelOp(GLint unit) -{ - switch (unit) { - case 0: - return TEXBLENDARG_TEXEL0; - case 1: - return TEXBLENDARG_TEXEL1; - case 2: - return TEXBLENDARG_TEXEL2; - case 3: - return TEXBLENDARG_TEXEL3; - default: - return TEXBLENDARG_TEXEL0; - } -} - - -/** - * Calculate the hardware instuctions to setup the current texture enviromnemt - * settings. Since \c gl_texture_unit::_CurrentCombine is used, both - * "classic" texture enviroments and GL_ARB_texture_env_combine type texture - * environments are treated identically. - * - * \todo - * This function should return \c GLboolean. When \c GL_FALSE is returned, - * it means that an environment is selected that the hardware cannot do. This - * is the way the Radeon and R200 drivers work. - * - * \todo - * Looking at i830_3d_regs.h, it seems the i830 can do part of - * GL_ATI_texture_env_combine3. It can handle using \c GL_ONE and - * \c GL_ZERO as combine inputs (which the code already supports). It can - * also handle the \c GL_MODULATE_ADD_ATI mode. Is it worth investigating - * partial support for the extension? - */ -GLuint -i830SetTexEnvCombine(struct i830_context * i830, - const struct gl_tex_env_combine_state * combine, - GLint blendUnit, - GLuint texel_op, GLuint * state, const GLfloat * factor) -{ - const GLuint numColorArgs = combine->_NumArgsRGB; - const GLuint numAlphaArgs = combine->_NumArgsA; - - GLuint blendop; - GLuint ablendop; - GLuint args_RGB[3]; - GLuint args_A[3]; - GLuint rgb_shift; - GLuint alpha_shift; - GLboolean need_factor = 0; - int i; - unsigned used; - static const GLuint tex_blend_rgb[3] = { - TEXPIPE_COLOR | TEXBLEND_ARG1 | TEXBLENDARG_MODIFY_PARMS, - TEXPIPE_COLOR | TEXBLEND_ARG2 | TEXBLENDARG_MODIFY_PARMS, - TEXPIPE_COLOR | TEXBLEND_ARG0 | TEXBLENDARG_MODIFY_PARMS, - }; - static const GLuint tex_blend_a[3] = { - TEXPIPE_ALPHA | TEXBLEND_ARG1 | TEXBLENDARG_MODIFY_PARMS, - TEXPIPE_ALPHA | TEXBLEND_ARG2 | TEXBLENDARG_MODIFY_PARMS, - TEXPIPE_ALPHA | TEXBLEND_ARG0 | TEXBLENDARG_MODIFY_PARMS, - }; - - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "%s\n", __FUNCTION__); - - - /* The EXT version of the DOT3 extension does not support the - * scale factor, but the ARB version (and the version in OpenGL - * 1.3) does. - */ - switch (combine->ModeRGB) { - case GL_DOT3_RGB_EXT: - alpha_shift = combine->ScaleShiftA; - rgb_shift = 0; - break; - - case GL_DOT3_RGBA_EXT: - alpha_shift = 0; - rgb_shift = 0; - break; - - default: - rgb_shift = combine->ScaleShiftRGB; - alpha_shift = combine->ScaleShiftA; - break; - } - - - switch (combine->ModeRGB) { - case GL_REPLACE: - blendop = TEXBLENDOP_ARG1; - break; - case GL_MODULATE: - blendop = TEXBLENDOP_MODULATE; - break; - case GL_ADD: - blendop = TEXBLENDOP_ADD; - break; - case GL_ADD_SIGNED: - blendop = TEXBLENDOP_ADDSIGNED; - break; - case GL_INTERPOLATE: - blendop = TEXBLENDOP_BLEND; - break; - case GL_SUBTRACT: - blendop = TEXBLENDOP_SUBTRACT; - break; - case GL_DOT3_RGB_EXT: - case GL_DOT3_RGB: - blendop = TEXBLENDOP_DOT3; - break; - case GL_DOT3_RGBA_EXT: - case GL_DOT3_RGBA: - blendop = TEXBLENDOP_DOT3; - break; - default: - return pass_through(state, blendUnit); - } - - blendop |= (rgb_shift << TEXOP_SCALE_SHIFT); - - - /* Handle RGB args */ - for (i = 0; i < 3; i++) { - switch (combine->SourceRGB[i]) { - case GL_TEXTURE: - args_RGB[i] = texel_op; - break; - case GL_TEXTURE0: - case GL_TEXTURE1: - case GL_TEXTURE2: - case GL_TEXTURE3: - args_RGB[i] = GetTexelOp(combine->SourceRGB[i] - GL_TEXTURE0); - break; - case GL_CONSTANT: - args_RGB[i] = TEXBLENDARG_FACTOR_N; - need_factor = 1; - break; - case GL_PRIMARY_COLOR: - args_RGB[i] = TEXBLENDARG_DIFFUSE; - break; - case GL_PREVIOUS: - args_RGB[i] = TEXBLENDARG_CURRENT; - break; - default: - return pass_through(state, blendUnit); - } - - switch (combine->OperandRGB[i]) { - case GL_SRC_COLOR: - args_RGB[i] |= 0; - break; - case GL_ONE_MINUS_SRC_COLOR: - args_RGB[i] |= TEXBLENDARG_INV_ARG; - break; - case GL_SRC_ALPHA: - args_RGB[i] |= TEXBLENDARG_REPLICATE_ALPHA; - break; - case GL_ONE_MINUS_SRC_ALPHA: - args_RGB[i] |= (TEXBLENDARG_REPLICATE_ALPHA | TEXBLENDARG_INV_ARG); - break; - default: - return pass_through(state, blendUnit); - } - } - - - /* Need to knobble the alpha calculations of TEXBLENDOP_DOT4 to - * match the spec. Can't use DOT3 as it won't propogate values - * into alpha as required: - * - * Note - the global factor is set up with alpha == .5, so - * the alpha part of the DOT4 calculation should be zero. - */ - if (combine->ModeRGB == GL_DOT3_RGBA_EXT || - combine->ModeRGB == GL_DOT3_RGBA) { - ablendop = TEXBLENDOP_DOT4; - args_A[0] = TEXBLENDARG_FACTOR; /* the global factor */ - args_A[1] = TEXBLENDARG_FACTOR; - args_A[2] = TEXBLENDARG_FACTOR; - } - else { - switch (combine->ModeA) { - case GL_REPLACE: - ablendop = TEXBLENDOP_ARG1; - break; - case GL_MODULATE: - ablendop = TEXBLENDOP_MODULATE; - break; - case GL_ADD: - ablendop = TEXBLENDOP_ADD; - break; - case GL_ADD_SIGNED: - ablendop = TEXBLENDOP_ADDSIGNED; - break; - case GL_INTERPOLATE: - ablendop = TEXBLENDOP_BLEND; - break; - case GL_SUBTRACT: - ablendop = TEXBLENDOP_SUBTRACT; - break; - default: - return pass_through(state, blendUnit); - } - - - ablendop |= (alpha_shift << TEXOP_SCALE_SHIFT); - - /* Handle A args */ - for (i = 0; i < 3; i++) { - switch (combine->SourceA[i]) { - case GL_TEXTURE: - args_A[i] = texel_op; - break; - case GL_TEXTURE0: - case GL_TEXTURE1: - case GL_TEXTURE2: - case GL_TEXTURE3: - args_A[i] = GetTexelOp(combine->SourceA[i] - GL_TEXTURE0); - break; - case GL_CONSTANT: - args_A[i] = TEXBLENDARG_FACTOR_N; - need_factor = 1; - break; - case GL_PRIMARY_COLOR: - args_A[i] = TEXBLENDARG_DIFFUSE; - break; - case GL_PREVIOUS: - args_A[i] = TEXBLENDARG_CURRENT; - break; - default: - return pass_through(state, blendUnit); - } - - switch (combine->OperandA[i]) { - case GL_SRC_ALPHA: - args_A[i] |= 0; - break; - case GL_ONE_MINUS_SRC_ALPHA: - args_A[i] |= TEXBLENDARG_INV_ARG; - break; - default: - return pass_through(state, blendUnit); - } - } - } - - - - /* Native Arg1 == Arg0 in GL_EXT_texture_env_combine spec */ - /* Native Arg2 == Arg1 in GL_EXT_texture_env_combine spec */ - /* Native Arg0 == Arg2 in GL_EXT_texture_env_combine spec */ - - /* When we render we need to figure out which is the last really enabled - * tex unit, and put last stage on it - */ - - - /* Build color & alpha pipelines */ - - used = 0; - state[used++] = (_3DSTATE_MAP_BLEND_OP_CMD(blendUnit) | - TEXPIPE_COLOR | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | - DISABLE_TEX_CNTRL_STAGE | TEXOP_MODIFY_PARMS | blendop); - state[used++] = (_3DSTATE_MAP_BLEND_OP_CMD(blendUnit) | - TEXPIPE_ALPHA | - ENABLE_TEXOUTPUT_WRT_SEL | - TEXOP_OUTPUT_CURRENT | TEXOP_MODIFY_PARMS | ablendop); - - for (i = 0; i < numColorArgs; i++) { - state[used++] = (_3DSTATE_MAP_BLEND_ARG_CMD(blendUnit) | - tex_blend_rgb[i] | args_RGB[i]); - } - - for (i = 0; i < numAlphaArgs; i++) { - state[used++] = (_3DSTATE_MAP_BLEND_ARG_CMD(blendUnit) | - tex_blend_a[i] | args_A[i]); - } - - - if (need_factor) - return emit_factor(blendUnit, state, used, factor); - else - return used; -} - - -static void -emit_texblend(struct i830_context *i830, GLuint unit, GLuint blendUnit, - GLboolean last_stage) -{ - struct gl_texture_unit *texUnit = &i830->intel.ctx.Texture.Unit[unit]; - GLuint tmp[I830_TEXBLEND_SIZE], tmp_sz; - - - if (0) - fprintf(stderr, "%s unit %d\n", __FUNCTION__, unit); - - /* Update i830->state.TexBlend - */ - tmp_sz = i830SetTexEnvCombine(i830, texUnit->_CurrentCombine, blendUnit, - GetTexelOp(unit), tmp, texUnit->EnvColor); - - if (last_stage) - tmp[0] |= TEXOP_LAST_STAGE; - - if (tmp_sz != i830->state.TexBlendWordsUsed[blendUnit] || - memcmp(tmp, i830->state.TexBlend[blendUnit], - tmp_sz * sizeof(GLuint))) { - - I830_STATECHANGE(i830, I830_UPLOAD_TEXBLEND(blendUnit)); - memcpy(i830->state.TexBlend[blendUnit], tmp, tmp_sz * sizeof(GLuint)); - i830->state.TexBlendWordsUsed[blendUnit] = tmp_sz; - } - - I830_ACTIVESTATE(i830, I830_UPLOAD_TEXBLEND(blendUnit), GL_TRUE); -} - -static void -emit_passthrough(struct i830_context *i830) -{ - GLuint tmp[I830_TEXBLEND_SIZE], tmp_sz; - GLuint unit = 0; - - tmp_sz = pass_through(tmp, unit); - tmp[0] |= TEXOP_LAST_STAGE; - - if (tmp_sz != i830->state.TexBlendWordsUsed[unit] || - memcmp(tmp, i830->state.TexBlend[unit], tmp_sz * sizeof(GLuint))) { - - I830_STATECHANGE(i830, I830_UPLOAD_TEXBLEND(unit)); - memcpy(i830->state.TexBlend[unit], tmp, tmp_sz * sizeof(GLuint)); - i830->state.TexBlendWordsUsed[unit] = tmp_sz; - } - - I830_ACTIVESTATE(i830, I830_UPLOAD_TEXBLEND(unit), GL_TRUE); -} - -void -i830EmitTextureBlend(struct i830_context *i830) -{ - GLcontext *ctx = &i830->intel.ctx; - GLuint unit, last_stage = 0, blendunit = 0; - - I830_ACTIVESTATE(i830, I830_UPLOAD_TEXBLEND_ALL, GL_FALSE); - - if (ctx->Texture._EnabledUnits) { - for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) - if (ctx->Texture.Unit[unit]._ReallyEnabled) - last_stage = unit; - - for (unit = 0; unit < ctx->Const.MaxTextureUnits; unit++) - if (ctx->Texture.Unit[unit]._ReallyEnabled) - emit_texblend(i830, unit, blendunit++, last_stage == unit); - } - else { - emit_passthrough(i830); - } -} diff --git a/src/mesa/drivers/dri/i915tex/i830_texstate.c b/src/mesa/drivers/dri/i915tex/i830_texstate.c deleted file mode 100644 index 0d3f053226..0000000000 --- a/src/mesa/drivers/dri/i915tex/i830_texstate.c +++ /dev/null @@ -1,343 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "mtypes.h" -#include "enums.h" -#include "texformat.h" -#include "dri_bufmgr.h" - -#include "intel_mipmap_tree.h" -#include "intel_tex.h" - -#include "i830_context.h" -#include "i830_reg.h" - - - -static GLuint -translate_texture_format(GLuint mesa_format) -{ - switch (mesa_format) { - case MESA_FORMAT_L8: - return MAPSURF_8BIT | MT_8BIT_L8; - case MESA_FORMAT_I8: - return MAPSURF_8BIT | MT_8BIT_I8; - case MESA_FORMAT_A8: - return MAPSURF_8BIT | MT_8BIT_I8; /* Kludge! */ - case MESA_FORMAT_AL88: - return MAPSURF_16BIT | MT_16BIT_AY88; - case MESA_FORMAT_RGB565: - return MAPSURF_16BIT | MT_16BIT_RGB565; - case MESA_FORMAT_ARGB1555: - return MAPSURF_16BIT | MT_16BIT_ARGB1555; - case MESA_FORMAT_ARGB4444: - return MAPSURF_16BIT | MT_16BIT_ARGB4444; - case MESA_FORMAT_ARGB8888: - return MAPSURF_32BIT | MT_32BIT_ARGB8888; - case MESA_FORMAT_YCBCR_REV: - return (MAPSURF_422 | MT_422_YCRCB_NORMAL); - case MESA_FORMAT_YCBCR: - return (MAPSURF_422 | MT_422_YCRCB_SWAPY); - case MESA_FORMAT_RGB_FXT1: - case MESA_FORMAT_RGBA_FXT1: - return (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1); - case MESA_FORMAT_RGBA_DXT1: - case MESA_FORMAT_RGB_DXT1: - return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1); - case MESA_FORMAT_RGBA_DXT3: - return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3); - case MESA_FORMAT_RGBA_DXT5: - return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); - default: - fprintf(stderr, "%s: bad image format %x\n", __FUNCTION__, mesa_format); - abort(); - return 0; - } -} - - - - -/* The i915 (and related graphics cores) do not support GL_CLAMP. The - * Intel drivers for "other operating systems" implement GL_CLAMP as - * GL_CLAMP_TO_EDGE, so the same is done here. - */ -static GLuint -translate_wrap_mode(GLenum wrap) -{ - switch (wrap) { - case GL_REPEAT: - return TEXCOORDMODE_WRAP; - case GL_CLAMP: - case GL_CLAMP_TO_EDGE: - return TEXCOORDMODE_CLAMP; /* not really correct */ - case GL_CLAMP_TO_BORDER: - return TEXCOORDMODE_CLAMP_BORDER; - case GL_MIRRORED_REPEAT: - return TEXCOORDMODE_MIRROR; - default: - return TEXCOORDMODE_WRAP; - } -} - - -/* Recalculate all state from scratch. Perhaps not the most - * efficient, but this has gotten complex enough that we need - * something which is understandable and reliable. - */ -static GLboolean -i830_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) -{ - GLcontext *ctx = &intel->ctx; - struct i830_context *i830 = i830_context(ctx); - struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; - struct intel_texture_object *intelObj = intel_texture_object(tObj); - struct gl_texture_image *firstImage; - GLuint *state = i830->state.Tex[unit], format, pitch; - - memset(state, 0, sizeof(state)); - - /*We need to refcount these. */ - - if (i830->state.tex_buffer[unit] != NULL) { - driBOUnReference(i830->state.tex_buffer[unit]); - i830->state.tex_buffer[unit] = NULL; - } - - if (!intelObj->imageOverride && !intel_finalize_mipmap_tree(intel, unit)) - return GL_FALSE; - - /* Get first image here, since intelObj->firstLevel will get set in - * the intel_finalize_mipmap_tree() call above. - */ - firstImage = tObj->Image[0][intelObj->firstLevel]; - - if (intelObj->imageOverride) { - i830->state.tex_buffer[unit] = NULL; - i830->state.tex_offset[unit] = intelObj->textureOffset; - - switch (intelObj->depthOverride) { - case 32: - format = MAPSURF_32BIT | MT_32BIT_ARGB8888; - break; - case 24: - default: - format = MAPSURF_32BIT | MT_32BIT_XRGB8888; - break; - case 16: - format = MAPSURF_16BIT | MT_16BIT_RGB565; - break; - } - - pitch = intelObj->pitchOverride; - } else { - i830->state.tex_buffer[unit] = driBOReference(intelObj->mt->region-> - buffer); - i830->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt, - 0, intelObj-> - firstLevel); - - format = translate_texture_format(firstImage->TexFormat->MesaFormat); - pitch = intelObj->mt->pitch * intelObj->mt->cpp; - } - - state[I830_TEXREG_TM0LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_2 | - (LOAD_TEXTURE_MAP0 << unit) | 4); - -/* state[I830_TEXREG_TM0S0] = (TM0S0_USE_FENCE | */ -/* t->intel.TextureOffset); */ - - - state[I830_TEXREG_TM0S1] = - (((firstImage->Height - 1) << TM0S1_HEIGHT_SHIFT) | - ((firstImage->Width - 1) << TM0S1_WIDTH_SHIFT) | format); - - state[I830_TEXREG_TM0S2] = - ((((pitch / 4) - 1) << TM0S2_PITCH_SHIFT) | TM0S2_CUBE_FACE_ENA_MASK); - - { - if (tObj->Target == GL_TEXTURE_CUBE_MAP) - state[I830_TEXREG_CUBE] = (_3DSTATE_MAP_CUBE | MAP_UNIT(unit) | - CUBE_NEGX_ENABLE | - CUBE_POSX_ENABLE | - CUBE_NEGY_ENABLE | - CUBE_POSY_ENABLE | - CUBE_NEGZ_ENABLE | CUBE_POSZ_ENABLE); - else - state[I830_TEXREG_CUBE] = (_3DSTATE_MAP_CUBE | MAP_UNIT(unit)); - } - - - - - { - GLuint minFilt, mipFilt, magFilt; - - switch (tObj->MinFilter) { - case GL_NEAREST: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_NONE; - break; - case GL_LINEAR: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_NONE; - break; - case GL_NEAREST_MIPMAP_NEAREST: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_NEAREST; - break; - case GL_LINEAR_MIPMAP_NEAREST: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_NEAREST; - break; - case GL_NEAREST_MIPMAP_LINEAR: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_LINEAR; - break; - case GL_LINEAR_MIPMAP_LINEAR: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_LINEAR; - break; - default: - return GL_FALSE; - } - - if (tObj->MaxAnisotropy > 1.0) { - minFilt = FILTER_ANISOTROPIC; - magFilt = FILTER_ANISOTROPIC; - } - else { - switch (tObj->MagFilter) { - case GL_NEAREST: - magFilt = FILTER_NEAREST; - break; - case GL_LINEAR: - magFilt = FILTER_LINEAR; - break; - default: - return GL_FALSE; - } - } - - state[I830_TEXREG_TM0S3] = i830->lodbias_tm0s3[unit]; - -#if 0 - /* YUV conversion: - */ - if (firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR || - firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR_REV) - state[I830_TEXREG_TM0S3] |= SS2_COLORSPACE_CONVERSION; -#endif - - state[I830_TEXREG_TM0S3] |= ((intelObj->lastLevel - - intelObj->firstLevel) * - 4) << TM0S3_MIN_MIP_SHIFT; - - state[I830_TEXREG_TM0S3] |= ((minFilt << TM0S3_MIN_FILTER_SHIFT) | - (mipFilt << TM0S3_MIP_FILTER_SHIFT) | - (magFilt << TM0S3_MAG_FILTER_SHIFT)); - } - - { - GLenum ws = tObj->WrapS; - GLenum wt = tObj->WrapT; - - - /* 3D textures not available on i830 - */ - if (tObj->Target == GL_TEXTURE_3D) - return GL_FALSE; - - state[I830_TEXREG_MCS] = (_3DSTATE_MAP_COORD_SET_CMD | - MAP_UNIT(unit) | - ENABLE_TEXCOORD_PARAMS | - ss3 | - ENABLE_ADDR_V_CNTL | - TEXCOORD_ADDR_V_MODE(translate_wrap_mode(wt)) - | ENABLE_ADDR_U_CNTL | - TEXCOORD_ADDR_U_MODE(translate_wrap_mode - (ws))); - } - - - state[I830_TEXREG_TM0S4] = INTEL_PACKCOLOR8888(tObj->_BorderChan[0], - tObj->_BorderChan[1], - tObj->_BorderChan[2], - tObj->_BorderChan[3]); - - - I830_ACTIVESTATE(i830, I830_UPLOAD_TEX(unit), GL_TRUE); - /* memcmp was already disabled, but definitely won't work as the - * region might now change and that wouldn't be detected: - */ - I830_STATECHANGE(i830, I830_UPLOAD_TEX(unit)); - return GL_TRUE; -} - - - - -void -i830UpdateTextureState(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - GLboolean ok = GL_TRUE; - GLuint i; - - for (i = 0; i < I830_TEX_UNITS && ok; i++) { - switch (intel->ctx.Texture.Unit[i]._ReallyEnabled) { - case TEXTURE_1D_BIT: - case TEXTURE_2D_BIT: - case TEXTURE_CUBE_BIT: - ok = i830_update_tex_unit(intel, i, TEXCOORDS_ARE_NORMAL); - break; - case TEXTURE_RECT_BIT: - ok = i830_update_tex_unit(intel, i, TEXCOORDS_ARE_IN_TEXELUNITS); - break; - case 0:{ - struct i830_context *i830 = i830_context(&intel->ctx); - if (i830->state.active & I830_UPLOAD_TEX(i)) - I830_ACTIVESTATE(i830, I830_UPLOAD_TEX(i), GL_FALSE); - - if (i830->state.tex_buffer[i] != NULL) { - driBOUnReference(i830->state.tex_buffer[i]); - i830->state.tex_buffer[i] = NULL; - } - break; - } - case TEXTURE_3D_BIT: - default: - ok = GL_FALSE; - break; - } - } - - FALLBACK(intel, I830_FALLBACK_TEXTURE, !ok); - - if (ok) - i830EmitTextureBlend(i830); -} diff --git a/src/mesa/drivers/dri/i915tex/i830_vtbl.c b/src/mesa/drivers/dri/i915tex/i830_vtbl.c deleted file mode 100644 index 5cde80a469..0000000000 --- a/src/mesa/drivers/dri/i915tex/i830_vtbl.c +++ /dev/null @@ -1,649 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "i830_context.h" -#include "i830_reg.h" -#include "intel_batchbuffer.h" -#include "intel_regions.h" -#include "tnl/t_context.h" -#include "tnl/t_vertex.h" - -#define FILE_DEBUG_FLAG DEBUG_STATE - -static GLboolean i830_check_vertex_size(struct intel_context *intel, - GLuint expected); - -#define SZ_TO_HW(sz) ((sz-2)&0x3) -#define EMIT_SZ(sz) (EMIT_1F + (sz) - 1) -#define EMIT_ATTR( ATTR, STYLE, V0 ) \ -do { \ - intel->vertex_attrs[intel->vertex_attr_count].attrib = (ATTR); \ - intel->vertex_attrs[intel->vertex_attr_count].format = (STYLE); \ - intel->vertex_attr_count++; \ - v0 |= V0; \ -} while (0) - -#define EMIT_PAD( N ) \ -do { \ - intel->vertex_attrs[intel->vertex_attr_count].attrib = 0; \ - intel->vertex_attrs[intel->vertex_attr_count].format = EMIT_PAD; \ - intel->vertex_attrs[intel->vertex_attr_count].offset = (N); \ - intel->vertex_attr_count++; \ -} while (0) - - -#define VRTX_TEX_SET_FMT(n, x) ((x)<<((n)*2)) -#define TEXBIND_SET(n, x) ((x)<<((n)*4)) - -static void -i830_render_prevalidate(struct intel_context *intel) -{ -} - -static void -i830_render_start(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - struct i830_context *i830 = i830_context(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *VB = &tnl->vb; - DECLARE_RENDERINPUTS(index_bitset); - GLuint v0 = _3DSTATE_VFT0_CMD; - GLuint v2 = _3DSTATE_VFT1_CMD; - GLuint mcsb1 = 0; - - RENDERINPUTS_COPY(index_bitset, tnl->render_inputs_bitset); - - /* Important: - */ - VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr; - intel->vertex_attr_count = 0; - - /* EMIT_ATTR's must be in order as they tell t_vertex.c how to - * build up a hardware vertex. - */ - if (RENDERINPUTS_TEST_RANGE(index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX)) { - EMIT_ATTR(_TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, VFT0_XYZW); - intel->coloroffset = 4; - } - else { - EMIT_ATTR(_TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, VFT0_XYZ); - intel->coloroffset = 3; - } - - if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_POINTSIZE)) { - EMIT_ATTR(_TNL_ATTRIB_POINTSIZE, EMIT_1F, VFT0_POINT_WIDTH); - } - - EMIT_ATTR(_TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, VFT0_DIFFUSE); - - intel->specoffset = 0; - if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_COLOR1) || - RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_FOG)) { - if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_COLOR1)) { - intel->specoffset = intel->coloroffset + 1; - EMIT_ATTR(_TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, VFT0_SPEC); - } - else - EMIT_PAD(3); - - if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_FOG)) - EMIT_ATTR(_TNL_ATTRIB_FOG, EMIT_1UB_1F, VFT0_SPEC); - else - EMIT_PAD(1); - } - - if (RENDERINPUTS_TEST_RANGE(index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX)) { - int i, count = 0; - - for (i = 0; i < I830_TEX_UNITS; i++) { - if (RENDERINPUTS_TEST(index_bitset, _TNL_ATTRIB_TEX(i))) { - GLuint sz = VB->TexCoordPtr[i]->size; - GLuint emit; - GLuint mcs = (i830->state.Tex[i][I830_TEXREG_MCS] & - ~TEXCOORDTYPE_MASK); - - switch (sz) { - case 1: - case 2: - emit = EMIT_2F; - sz = 2; - mcs |= TEXCOORDTYPE_CARTESIAN; - break; - case 3: - emit = EMIT_3F; - sz = 3; - mcs |= TEXCOORDTYPE_VECTOR; - break; - case 4: - emit = EMIT_3F_XYW; - sz = 3; - mcs |= TEXCOORDTYPE_HOMOGENEOUS; - break; - default: - continue; - }; - - - EMIT_ATTR(_TNL_ATTRIB_TEX0 + i, emit, 0); - v2 |= VRTX_TEX_SET_FMT(count, SZ_TO_HW(sz)); - mcsb1 |= (count + 8) << (i * 4); - - if (mcs != i830->state.Tex[i][I830_TEXREG_MCS]) { - I830_STATECHANGE(i830, I830_UPLOAD_TEX(i)); - i830->state.Tex[i][I830_TEXREG_MCS] = mcs; - } - - count++; - } - } - - v0 |= VFT0_TEX_COUNT(count); - } - - /* Only need to change the vertex emit code if there has been a - * statechange to a new hardware vertex format: - */ - if (v0 != i830->state.Ctx[I830_CTXREG_VF] || - v2 != i830->state.Ctx[I830_CTXREG_VF2] || - mcsb1 != i830->state.Ctx[I830_CTXREG_MCSB1] || - !RENDERINPUTS_EQUAL(index_bitset, i830->last_index_bitset)) { - int k; - - I830_STATECHANGE(i830, I830_UPLOAD_CTX); - - /* Must do this *after* statechange, so as not to affect - * buffered vertices reliant on the old state: - */ - intel->vertex_size = - _tnl_install_attrs(ctx, - intel->vertex_attrs, - intel->vertex_attr_count, - intel->ViewportMatrix.m, 0); - - intel->vertex_size >>= 2; - - i830->state.Ctx[I830_CTXREG_VF] = v0; - i830->state.Ctx[I830_CTXREG_VF2] = v2; - i830->state.Ctx[I830_CTXREG_MCSB1] = mcsb1; - RENDERINPUTS_COPY(i830->last_index_bitset, index_bitset); - - k = i830_check_vertex_size(intel, intel->vertex_size); - assert(k); - } -} - -static void -i830_reduced_primitive_state(struct intel_context *intel, GLenum rprim) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - GLuint st1 = i830->state.Stipple[I830_STPREG_ST1]; - - st1 &= ~ST1_ENABLE; - - switch (rprim) { - case GL_TRIANGLES: - if (intel->ctx.Polygon.StippleFlag && intel->hw_stipple) - st1 |= ST1_ENABLE; - break; - case GL_LINES: - case GL_POINTS: - default: - break; - } - - i830->intel.reduced_primitive = rprim; - - if (st1 != i830->state.Stipple[I830_STPREG_ST1]) { - INTEL_FIREVERTICES(intel); - - I830_STATECHANGE(i830, I830_UPLOAD_STIPPLE); - i830->state.Stipple[I830_STPREG_ST1] = st1; - } -} - -/* Pull apart the vertex format registers and figure out how large a - * vertex is supposed to be. - */ -static GLboolean -i830_check_vertex_size(struct intel_context *intel, GLuint expected) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - int vft0 = i830->current->Ctx[I830_CTXREG_VF]; - int vft1 = i830->current->Ctx[I830_CTXREG_VF2]; - int nrtex = (vft0 & VFT0_TEX_COUNT_MASK) >> VFT0_TEX_COUNT_SHIFT; - int i, sz = 0; - - switch (vft0 & VFT0_XYZW_MASK) { - case VFT0_XY: - sz = 2; - break; - case VFT0_XYZ: - sz = 3; - break; - case VFT0_XYW: - sz = 3; - break; - case VFT0_XYZW: - sz = 4; - break; - default: - fprintf(stderr, "no xyzw specified\n"); - return 0; - } - - if (vft0 & VFT0_SPEC) - sz++; - if (vft0 & VFT0_DIFFUSE) - sz++; - if (vft0 & VFT0_DEPTH_OFFSET) - sz++; - if (vft0 & VFT0_POINT_WIDTH) - sz++; - - for (i = 0; i < nrtex; i++) { - switch (vft1 & VFT1_TEX0_MASK) { - case TEXCOORDFMT_2D: - sz += 2; - break; - case TEXCOORDFMT_3D: - sz += 3; - break; - case TEXCOORDFMT_4D: - sz += 4; - break; - case TEXCOORDFMT_1D: - sz += 1; - break; - } - vft1 >>= VFT1_TEX1_SHIFT; - } - - if (sz != expected) - fprintf(stderr, "vertex size mismatch %d/%d\n", sz, expected); - - return sz == expected; -} - -static void -i830_emit_invarient_state(struct intel_context *intel) -{ - BATCH_LOCALS; - - BEGIN_BATCH(40, 0); - - OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_DFLT_SPEC_CMD); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_DFLT_Z_CMD); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_FOG_MODE_CMD); - OUT_BATCH(FOGFUNC_ENABLE | - FOG_LINEAR_CONST | FOGSRC_INDEX_Z | ENABLE_FOG_DENSITY); - OUT_BATCH(0); - OUT_BATCH(0); - - - OUT_BATCH(_3DSTATE_MAP_TEX_STREAM_CMD | - MAP_UNIT(0) | - DISABLE_TEX_STREAM_BUMP | - ENABLE_TEX_STREAM_COORD_SET | - TEX_STREAM_COORD_SET(0) | - ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(0)); - OUT_BATCH(_3DSTATE_MAP_TEX_STREAM_CMD | - MAP_UNIT(1) | - DISABLE_TEX_STREAM_BUMP | - ENABLE_TEX_STREAM_COORD_SET | - TEX_STREAM_COORD_SET(1) | - ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(1)); - OUT_BATCH(_3DSTATE_MAP_TEX_STREAM_CMD | - MAP_UNIT(2) | - DISABLE_TEX_STREAM_BUMP | - ENABLE_TEX_STREAM_COORD_SET | - TEX_STREAM_COORD_SET(2) | - ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(2)); - OUT_BATCH(_3DSTATE_MAP_TEX_STREAM_CMD | - MAP_UNIT(3) | - DISABLE_TEX_STREAM_BUMP | - ENABLE_TEX_STREAM_COORD_SET | - TEX_STREAM_COORD_SET(3) | - ENABLE_TEX_STREAM_MAP_IDX | TEX_STREAM_MAP_IDX(3)); - - OUT_BATCH(_3DSTATE_MAP_COORD_TRANSFORM); - OUT_BATCH(DISABLE_TEX_TRANSFORM | TEXTURE_SET(0)); - OUT_BATCH(_3DSTATE_MAP_COORD_TRANSFORM); - OUT_BATCH(DISABLE_TEX_TRANSFORM | TEXTURE_SET(1)); - OUT_BATCH(_3DSTATE_MAP_COORD_TRANSFORM); - OUT_BATCH(DISABLE_TEX_TRANSFORM | TEXTURE_SET(2)); - OUT_BATCH(_3DSTATE_MAP_COORD_TRANSFORM); - OUT_BATCH(DISABLE_TEX_TRANSFORM | TEXTURE_SET(3)); - - OUT_BATCH(_3DSTATE_RASTER_RULES_CMD | - ENABLE_POINT_RASTER_RULE | - OGL_POINT_RASTER_RULE | - ENABLE_LINE_STRIP_PROVOKE_VRTX | - ENABLE_TRI_FAN_PROVOKE_VRTX | - ENABLE_TRI_STRIP_PROVOKE_VRTX | - LINE_STRIP_PROVOKE_VRTX(1) | - TRI_FAN_PROVOKE_VRTX(2) | TRI_STRIP_PROVOKE_VRTX(2)); - - OUT_BATCH(_3DSTATE_VERTEX_TRANSFORM); - OUT_BATCH(DISABLE_VIEWPORT_TRANSFORM | DISABLE_PERSPECTIVE_DIVIDE); - - OUT_BATCH(_3DSTATE_W_STATE_CMD); - OUT_BATCH(MAGIC_W_STATE_DWORD1); - OUT_BATCH(0x3f800000 /* 1.0 in IEEE float */ ); - - - OUT_BATCH(_3DSTATE_COLOR_FACTOR_CMD); - OUT_BATCH(0x80808080); /* .5 required in alpha for GL_DOT3_RGBA_EXT */ - - ADVANCE_BATCH(); -} - - -#define emit( intel, state, size ) \ -do { \ - int k; \ - BEGIN_BATCH(size / sizeof(GLuint), 0); \ - for (k = 0 ; k < size / sizeof(GLuint) ; k++) { \ - if (0) _mesa_printf(" 0x%08x\n", state[k]); \ - OUT_BATCH(state[k]); \ - } \ - ADVANCE_BATCH(); \ -} while (0) - -static GLuint -get_state_size(struct i830_hw_state *state) -{ - GLuint dirty = state->active & ~state->emitted; - GLuint sz = 0; - GLuint i; - - if (dirty & I830_UPLOAD_INVARIENT) - sz += 40 * sizeof(int); - - if (dirty & I830_UPLOAD_CTX) - sz += sizeof(state->Ctx); - - if (dirty & I830_UPLOAD_BUFFERS) - sz += sizeof(state->Buffer); - - if (dirty & I830_UPLOAD_STIPPLE) - sz += sizeof(state->Stipple); - - for (i = 0; i < I830_TEX_UNITS; i++) { - if ((dirty & I830_UPLOAD_TEX(i))) - sz += sizeof(state->Tex[i]); - - if (dirty & I830_UPLOAD_TEXBLEND(i)) - sz += state->TexBlendWordsUsed[i] * 4; - } - - return sz; -} - - -/* Push the state into the sarea and/or texture memory. - */ -static void -i830_emit_state(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - struct i830_hw_state *state = i830->current; - int i; - GLuint dirty; - BATCH_LOCALS; - - /* We don't hold the lock at this point, so want to make sure that - * there won't be a buffer wrap. - * - * It might be better to talk about explicit places where - * scheduling is allowed, rather than assume that it is whenever a - * batchbuffer fills up. - */ - intel_batchbuffer_require_space(intel->batch, get_state_size(state), 0); - - /* Do this here as we may have flushed the batchbuffer above, - * causing more state to be dirty! - */ - dirty = state->active & ~state->emitted; - - if (dirty & I830_UPLOAD_INVARIENT) { - DBG("I830_UPLOAD_INVARIENT:\n"); - i830_emit_invarient_state(intel); - } - - if (dirty & I830_UPLOAD_CTX) { - DBG("I830_UPLOAD_CTX:\n"); - emit(i830, state->Ctx, sizeof(state->Ctx)); - - } - - if (dirty & I830_UPLOAD_BUFFERS) { - DBG("I830_UPLOAD_BUFFERS:\n"); - BEGIN_BATCH(I830_DEST_SETUP_SIZE + 2, 0); - if (state->draw_region->buffer) { - OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR0]); - OUT_BATCH(state->Buffer[I830_DESTREG_CBUFADDR1]); - OUT_RELOC(state->draw_region->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - state->draw_region->draw_offset); - } - if (state->depth_region) { - OUT_BATCH(state->Buffer[I830_DESTREG_DBUFADDR0]); - OUT_BATCH(state->Buffer[I830_DESTREG_DBUFADDR1]); - OUT_RELOC(state->depth_region->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - state->depth_region->draw_offset); - } - - OUT_BATCH(state->Buffer[I830_DESTREG_DV0]); - OUT_BATCH(state->Buffer[I830_DESTREG_DV1]); - OUT_BATCH(state->Buffer[I830_DESTREG_SENABLE]); - OUT_BATCH(state->Buffer[I830_DESTREG_SR0]); - OUT_BATCH(state->Buffer[I830_DESTREG_SR1]); - OUT_BATCH(state->Buffer[I830_DESTREG_SR2]); - ADVANCE_BATCH(); - } - - if (dirty & I830_UPLOAD_STIPPLE) { - DBG("I830_UPLOAD_STIPPLE:\n"); - emit(i830, state->Stipple, sizeof(state->Stipple)); - } - - for (i = 0; i < I830_TEX_UNITS; i++) { - if ((dirty & I830_UPLOAD_TEX(i))) { - DBG("I830_UPLOAD_TEX(%d):\n", i); - - BEGIN_BATCH(I830_TEX_SETUP_SIZE + 1, 0); - OUT_BATCH(state->Tex[i][I830_TEXREG_TM0LI]); - - if (state->tex_buffer[i]) { - OUT_RELOC(state->tex_buffer[i], - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, - state->tex_offset[i] | TM0S0_USE_FENCE); - } - else if (state == &i830->meta) { - assert(i == 0); - OUT_BATCH(0); - } - else { - OUT_BATCH(state->tex_offset[i]); - } - - OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S1]); - OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S2]); - OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S3]); - OUT_BATCH(state->Tex[i][I830_TEXREG_TM0S4]); - OUT_BATCH(state->Tex[i][I830_TEXREG_MCS]); - OUT_BATCH(state->Tex[i][I830_TEXREG_CUBE]); - } - - if (dirty & I830_UPLOAD_TEXBLEND(i)) { - DBG("I830_UPLOAD_TEXBLEND(%d): %d words\n", i, - state->TexBlendWordsUsed[i]); - emit(i830, state->TexBlend[i], state->TexBlendWordsUsed[i] * 4); - } - } - - state->emitted |= dirty; -} - -static void -i830_destroy_context(struct intel_context *intel) -{ - _tnl_free_vertices(&intel->ctx); -} - - -void -i830_state_draw_region(struct intel_context *intel, - struct i830_hw_state *state, - struct intel_region *color_region, - struct intel_region *depth_region) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - GLuint value; - - ASSERT(state == &i830->state || state == &i830->meta); - - if (state->draw_region != color_region) { - intel_region_release(&state->draw_region); - intel_region_reference(&state->draw_region, color_region); - } - if (state->depth_region != depth_region) { - intel_region_release(&state->depth_region); - intel_region_reference(&state->depth_region, depth_region); - } - - /* - * Set stride/cpp values - */ - if (color_region) { - state->Buffer[I830_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - state->Buffer[I830_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | - BUF_3D_PITCH(color_region->pitch * color_region->cpp) | - BUF_3D_USE_FENCE); - } - - if (depth_region) { - state->Buffer[I830_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - state->Buffer[I830_DESTREG_DBUFADDR1] = - (BUF_3D_ID_DEPTH | - BUF_3D_PITCH(depth_region->pitch * depth_region->cpp) | - BUF_3D_USE_FENCE); - } - - /* - * Compute/set I830_DESTREG_DV1 value - */ - value = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | DEPTH_IS_Z); /* .5 */ - - if (color_region && color_region->cpp == 4) { - value |= DV_PF_8888; - } - else { - value |= DV_PF_565; - } - if (depth_region && depth_region->cpp == 4) { - value |= DEPTH_FRMT_24_FIXED_8_OTHER; - } - else { - value |= DEPTH_FRMT_16_FIXED; - } - state->Buffer[I830_DESTREG_DV1] = value; - - I830_STATECHANGE(i830, I830_UPLOAD_BUFFERS); - - -} - - -static void -i830_set_draw_region(struct intel_context *intel, - struct intel_region *color_region, - struct intel_region *depth_region) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - i830_state_draw_region(intel, &i830->state, color_region, depth_region); -} - - -/* This isn't really handled at the moment. - */ -static void -i830_lost_hardware(struct intel_context *intel) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - i830->state.emitted = 0; -} - - - -static GLuint -i830_flush_cmd(void) -{ - return MI_FLUSH | FLUSH_MAP_CACHE; -} - - -static void -i830_assert_not_dirty( struct intel_context *intel ) -{ - struct i830_context *i830 = i830_context(&intel->ctx); - struct i830_hw_state *state = i830->current; - GLuint dirty = state->active & ~state->emitted; - assert(!dirty); -} - - -void -i830InitVtbl(struct i830_context *i830) -{ - i830->intel.vtbl.check_vertex_size = i830_check_vertex_size; - i830->intel.vtbl.destroy = i830_destroy_context; - i830->intel.vtbl.emit_state = i830_emit_state; - i830->intel.vtbl.lost_hardware = i830_lost_hardware; - i830->intel.vtbl.reduced_primitive_state = i830_reduced_primitive_state; - i830->intel.vtbl.set_draw_region = i830_set_draw_region; - i830->intel.vtbl.update_texture_state = i830UpdateTextureState; - i830->intel.vtbl.flush_cmd = i830_flush_cmd; - i830->intel.vtbl.render_start = i830_render_start; - i830->intel.vtbl.render_prevalidate = i830_render_prevalidate; - i830->intel.vtbl.assert_not_dirty = i830_assert_not_dirty; -} diff --git a/src/mesa/drivers/dri/i915tex/i915_context.c b/src/mesa/drivers/dri/i915tex/i915_context.c deleted file mode 100644 index b6d004b258..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_context.c +++ /dev/null @@ -1,176 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915_context.h" -#include "imports.h" -#include "intel_tex.h" -#include "intel_tris.h" -#include "tnl/t_context.h" -#include "tnl/t_pipeline.h" -#include "tnl/t_vertex.h" - -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#include "tnl/tnl.h" - -#include "utils.h" -#include "i915_reg.h" - -#include "intel_regions.h" -#include "intel_batchbuffer.h" - -/*************************************** - * Mesa's Driver Functions - ***************************************/ - -static const struct dri_extension i915_extensions[] = { - {"GL_ARB_depth_texture", NULL}, - {"GL_ARB_fragment_program", NULL}, - {"GL_ARB_shadow", NULL}, - {"GL_ARB_texture_env_crossbar", NULL}, - {"GL_ARB_texture_non_power_of_two", NULL}, - {"GL_EXT_shadow_funcs", NULL}, - /* ARB extn won't work if not enabled */ - {"GL_SGIX_depth_texture", NULL}, - {NULL, NULL} -}; - -/* Override intel default. - */ -static void -i915InvalidateState(GLcontext * ctx, GLuint new_state) -{ - _swrast_InvalidateState(ctx, new_state); - _swsetup_InvalidateState(ctx, new_state); - _vbo_InvalidateState(ctx, new_state); - _tnl_InvalidateState(ctx, new_state); - _tnl_invalidate_vertex_state(ctx, new_state); - intel_context(ctx)->NewGLState |= new_state; - - st_invalidate_state(ctx, new_state); - - /* Todo: gather state values under which tracked parameters become - * invalidated, add callbacks for things like - * ProgramLocalParameters, etc. - */ - { - struct i915_fragment_program *p = - (struct i915_fragment_program *) ctx->FragmentProgram._Current; - if (p && p->nr_params) - p->params_uptodate = 0; - } - - if (new_state & (_NEW_FOG | _NEW_HINT | _NEW_PROGRAM)) - i915_update_fog(ctx); -} - - -static void -i915InitDriverFunctions(struct dd_function_table *functions) -{ - intelInitDriverFunctions(functions); - i915InitStateFunctions(functions); - i915InitTextureFuncs(functions); - i915InitFragProgFuncs(functions); - functions->UpdateState = i915InvalidateState; -} - - - -GLboolean -i915CreateContext(const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate) -{ - struct dd_function_table functions; - struct i915_context *i915 = - (struct i915_context *) CALLOC_STRUCT(i915_context); - struct intel_context *intel = &i915->intel; - GLcontext *ctx = &intel->ctx; - - if (!i915) - return GL_FALSE; - - if (0) - _mesa_printf("\ntexmem-0-3 branch\n\n"); - - i915InitVtbl(i915); - i915InitMetaFuncs(i915); - - i915InitDriverFunctions(&functions); - - if (!intelInitContext(intel, mesaVis, driContextPriv, - sharedContextPrivate, &functions)) { - FREE(i915); - return GL_FALSE; - } - - ctx->Const.MaxTextureUnits = I915_TEX_UNITS; - ctx->Const.MaxTextureImageUnits = I915_TEX_UNITS; - ctx->Const.MaxTextureCoordUnits = I915_TEX_UNITS; - - - /* Advertise the full hardware capabilities. The new memory - * manager should cope much better with overload situations: - */ - ctx->Const.MaxTextureLevels = 12; - ctx->Const.Max3DTextureLevels = 9; - ctx->Const.MaxCubeTextureLevels = 12; - ctx->Const.MaxTextureRectSize = (1 << 11); - ctx->Const.MaxTextureUnits = I915_TEX_UNITS; - - /* GL_ARB_fragment_program limits - don't think Mesa actually - * validates programs against these, and in any case one ARB - * instruction can translate to more than one HW instruction, so - * we'll still have to check and fallback each time. - */ - ctx->Const.FragmentProgram.MaxNativeTemps = I915_MAX_TEMPORARY; - ctx->Const.FragmentProgram.MaxNativeAttribs = 11; /* 8 tex, 2 color, fog */ - ctx->Const.FragmentProgram.MaxNativeParameters = I915_MAX_CONSTANT; - ctx->Const.FragmentProgram.MaxNativeAluInstructions = I915_MAX_ALU_INSN; - ctx->Const.FragmentProgram.MaxNativeTexInstructions = I915_MAX_TEX_INSN; - ctx->Const.FragmentProgram.MaxNativeInstructions = (I915_MAX_ALU_INSN + - I915_MAX_TEX_INSN); - ctx->Const.FragmentProgram.MaxNativeTexIndirections = - I915_MAX_TEX_INDIRECT; - ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0; /* I don't think we have one */ - - ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; - ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; - - driInitExtensions(ctx, i915_extensions, GL_FALSE); - - - _tnl_init_vertices(ctx, ctx->Const.MaxArrayLockSize + 12, - 36 * sizeof(GLfloat)); - - intel->verts = TNL_CONTEXT(ctx)->clipspace.vertex_buf; - - i915InitState(i915); - - return GL_TRUE; -} diff --git a/src/mesa/drivers/dri/i915tex/i915_context.h b/src/mesa/drivers/dri/i915tex/i915_context.h deleted file mode 100644 index 3a41d66c14..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_context.h +++ /dev/null @@ -1,360 +0,0 @@ - /************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef I915CONTEXT_INC -#define I915CONTEXT_INC - -#include "intel_context.h" - -#define I915_FALLBACK_TEXTURE 0x1000 -#define I915_FALLBACK_COLORMASK 0x2000 -#define I915_FALLBACK_STENCIL 0x4000 -#define I915_FALLBACK_STIPPLE 0x8000 -#define I915_FALLBACK_PROGRAM 0x10000 -#define I915_FALLBACK_LOGICOP 0x20000 -#define I915_FALLBACK_POLYGON_SMOOTH 0x40000 -#define I915_FALLBACK_POINT_SMOOTH 0x80000 - -#define I915_UPLOAD_CTX 0x1 -#define I915_UPLOAD_BUFFERS 0x2 -#define I915_UPLOAD_STIPPLE 0x4 -#define I915_UPLOAD_PROGRAM 0x8 -#define I915_UPLOAD_CONSTANTS 0x10 -#define I915_UPLOAD_FOG 0x20 -#define I915_UPLOAD_INVARIENT 0x40 -#define I915_UPLOAD_DEFAULTS 0x80 -#define I915_UPLOAD_TEX(i) (0x00010000<<(i)) -#define I915_UPLOAD_TEX_ALL (0x00ff0000) -#define I915_UPLOAD_TEX_0_SHIFT 16 - - -/* State structure offsets - these will probably disappear. - */ -#define I915_DESTREG_CBUFADDR0 0 -#define I915_DESTREG_CBUFADDR1 1 -#define I915_DESTREG_DBUFADDR0 3 -#define I915_DESTREG_DBUFADDR1 4 -#define I915_DESTREG_DV0 6 -#define I915_DESTREG_DV1 7 -#define I915_DESTREG_SENABLE 8 -#define I915_DESTREG_SR0 9 -#define I915_DESTREG_SR1 10 -#define I915_DESTREG_SR2 11 -#define I915_DEST_SETUP_SIZE 12 - -#define I915_CTXREG_STATE4 0 -#define I915_CTXREG_LI 1 -#define I915_CTXREG_LIS2 2 -#define I915_CTXREG_LIS4 3 -#define I915_CTXREG_LIS5 4 -#define I915_CTXREG_LIS6 5 -#define I915_CTXREG_IAB 6 -#define I915_CTXREG_BLENDCOLOR0 7 -#define I915_CTXREG_BLENDCOLOR1 8 -#define I915_CTX_SETUP_SIZE 9 - -#define I915_FOGREG_COLOR 0 -#define I915_FOGREG_MODE0 1 -#define I915_FOGREG_MODE1 2 -#define I915_FOGREG_MODE2 3 -#define I915_FOGREG_MODE3 4 -#define I915_FOG_SETUP_SIZE 5 - -#define I915_STPREG_ST0 0 -#define I915_STPREG_ST1 1 -#define I915_STP_SETUP_SIZE 2 - -#define I915_TEXREG_MS3 1 -#define I915_TEXREG_MS4 2 -#define I915_TEXREG_SS2 3 -#define I915_TEXREG_SS3 4 -#define I915_TEXREG_SS4 5 -#define I915_TEX_SETUP_SIZE 6 - -#define I915_DEFREG_C0 0 -#define I915_DEFREG_C1 1 -#define I915_DEFREG_S0 2 -#define I915_DEFREG_S1 3 -#define I915_DEFREG_Z0 4 -#define I915_DEFREG_Z1 5 -#define I915_DEF_SETUP_SIZE 6 - - -#define I915_MAX_CONSTANT 32 -#define I915_CONSTANT_SIZE (2+(4*I915_MAX_CONSTANT)) - - -#define I915_PROGRAM_SIZE 192 - - -/* Hardware version of a parsed fragment program. "Derived" from the - * mesa fragment_program struct. - */ -struct i915_fragment_program -{ - struct gl_fragment_program FragProg; - - GLboolean translated; - GLboolean params_uptodate; - GLboolean on_hardware; - GLboolean error; /* If program is malformed for any reason. */ - - GLuint nr_tex_indirect; - GLuint nr_tex_insn; - GLuint nr_alu_insn; - GLuint nr_decl_insn; - - - - - /* TODO: split between the stored representation of a program and - * the state used to build that representation. - */ - GLcontext *ctx; - - GLuint declarations[I915_PROGRAM_SIZE]; - GLuint program[I915_PROGRAM_SIZE]; - - GLfloat constant[I915_MAX_CONSTANT][4]; - GLuint constant_flags[I915_MAX_CONSTANT]; - GLuint nr_constants; - - GLuint *csr; /* Cursor, points into program. - */ - - GLuint *decl; /* Cursor, points into declarations. - */ - - GLuint decl_s; /* flags for which s regs need to be decl'd */ - GLuint decl_t; /* flags for which t regs need to be decl'd */ - - GLuint temp_flag; /* Tracks temporary regs which are in - * use. - */ - - GLuint utemp_flag; /* Tracks TYPE_U temporary regs which are in - * use. - */ - - - - /* Helpers for i915_fragprog.c: - */ - GLuint wpos_tex; - GLboolean depth_written; - - struct - { - GLuint reg; /* Hardware constant idx */ - const GLfloat *values; /* Pointer to tracked values */ - } param[I915_MAX_CONSTANT]; - GLuint nr_params; - - - /* Helpers for i915_texprog.c: - */ - GLuint src_texture; /* Reg containing sampled texture color, - * else UREG_BAD. - */ - - GLuint src_previous; /* Reg containing color from previous - * stage. May need to be decl'd. - */ - - GLuint last_tex_stage; /* Number of last enabled texture unit */ - - struct vertex_buffer *VB; -}; - - - - - - - -#define I915_TEX_UNITS 8 - - -struct i915_hw_state -{ - GLuint Ctx[I915_CTX_SETUP_SIZE]; - GLuint Buffer[I915_DEST_SETUP_SIZE]; - GLuint Stipple[I915_STP_SETUP_SIZE]; - GLuint Fog[I915_FOG_SETUP_SIZE]; - GLuint Defaults[I915_DEF_SETUP_SIZE]; - GLuint Tex[I915_TEX_UNITS][I915_TEX_SETUP_SIZE]; - GLuint Constant[I915_CONSTANT_SIZE]; - GLuint ConstantSize; - GLuint Program[I915_PROGRAM_SIZE]; - GLuint ProgramSize; - - /* Region pointers for relocation: - */ - struct intel_region *draw_region; - struct intel_region *depth_region; -/* struct intel_region *tex_region[I915_TEX_UNITS]; */ - - /* Regions aren't actually that appropriate here as the memory may - * be from a PBO or FBO. Just use the buffer id. Will have to do - * this for draw and depth for FBO's... - */ - struct _DriBufferObject *tex_buffer[I915_TEX_UNITS]; - GLuint tex_offset[I915_TEX_UNITS]; - - - GLuint active; /* I915_UPLOAD_* */ - GLuint emitted; /* I915_UPLOAD_* */ -}; - -#define I915_FOG_PIXEL 2 -#define I915_FOG_VERTEX 1 -#define I915_FOG_NONE 0 - -struct i915_context -{ - struct intel_context intel; - - GLuint last_ReallyEnabled; - GLuint vertex_fog; - GLuint lodbias_ss2[MAX_TEXTURE_UNITS]; - - - struct i915_fragment_program *current_program; - - struct i915_hw_state meta, initial, state, *current; -}; - - -#define I915_STATECHANGE(i915, flag) \ -do { \ - INTEL_FIREVERTICES( &(i915)->intel ); \ - (i915)->state.emitted &= ~(flag); \ -} while (0) - -#define I915_ACTIVESTATE(i915, flag, mode) \ -do { \ - INTEL_FIREVERTICES( &(i915)->intel ); \ - if (mode) \ - (i915)->state.active |= (flag); \ - else \ - (i915)->state.active &= ~(flag); \ -} while (0) - - -/*====================================================================== - * i915_vtbl.c - */ -extern void i915InitVtbl(struct i915_context *i915); - -extern void -i915_state_draw_region(struct intel_context *intel, - struct i915_hw_state *state, - struct intel_region *color_region, - struct intel_region *depth_region); - - - -#define SZ_TO_HW(sz) ((sz-2)&0x3) -#define EMIT_SZ(sz) (EMIT_1F + (sz) - 1) -#define EMIT_ATTR( ATTR, STYLE, S4, SZ ) \ -do { \ - intel->vertex_attrs[intel->vertex_attr_count].attrib = (ATTR); \ - intel->vertex_attrs[intel->vertex_attr_count].format = (STYLE); \ - s4 |= S4; \ - intel->vertex_attr_count++; \ - offset += (SZ); \ -} while (0) - -#define EMIT_PAD( N ) \ -do { \ - intel->vertex_attrs[intel->vertex_attr_count].attrib = 0; \ - intel->vertex_attrs[intel->vertex_attr_count].format = EMIT_PAD; \ - intel->vertex_attrs[intel->vertex_attr_count].offset = (N); \ - intel->vertex_attr_count++; \ - offset += (N); \ -} while (0) - - - -/*====================================================================== - * i915_context.c - */ -extern GLboolean i915CreateContext(const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - - -/*====================================================================== - * i915_debug.c - */ -extern void i915_disassemble_program(const GLuint * program, GLuint sz); -extern void i915_print_ureg(const char *msg, GLuint ureg); - - -/*====================================================================== - * i915_state.c - */ -extern void i915InitStateFunctions(struct dd_function_table *functions); -extern void i915InitState(struct i915_context *i915); -extern void i915_update_fog(GLcontext * ctx); - - -/*====================================================================== - * i915_tex.c - */ -extern void i915UpdateTextureState(struct intel_context *intel); -extern void i915InitTextureFuncs(struct dd_function_table *functions); - -/*====================================================================== - * i915_metaops.c - */ -void i915InitMetaFuncs(struct i915_context *i915); - - -/*====================================================================== - * i915_fragprog.c - */ -extern void i915ValidateFragmentProgram(struct i915_context *i915); -extern void i915InitFragProgFuncs(struct dd_function_table *functions); - -/*====================================================================== - * Inline conversion functions. These are better-typed than the - * macros used previously: - */ -static INLINE struct i915_context * -i915_context(GLcontext * ctx) -{ - return (struct i915_context *) ctx; -} - - - -#define I915_CONTEXT(ctx) i915_context(ctx) - - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/i915_debug.c b/src/mesa/drivers/dri/i915tex/i915_debug.c deleted file mode 100644 index 974527e14c..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_debug.c +++ /dev/null @@ -1,334 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915_reg.h" -#include "i915_context.h" -#include - - -static const char *opcodes[0x20] = { - "NOP", - "ADD", - "MOV", - "MUL", - "MAD", - "DP2ADD", - "DP3", - "DP4", - "FRC", - "RCP", - "RSQ", - "EXP", - "LOG", - "CMP", - "MIN", - "MAX", - "FLR", - "MOD", - "TRC", - "SGE", - "SLT", - "TEXLD", - "TEXLDP", - "TEXLDB", - "TEXKILL", - "DCL", - "0x1a", - "0x1b", - "0x1c", - "0x1d", - "0x1e", - "0x1f", -}; - - -static const int args[0x20] = { - 0, /* 0 nop */ - 2, /* 1 add */ - 1, /* 2 mov */ - 2, /* 3 m ul */ - 3, /* 4 mad */ - 3, /* 5 dp2add */ - 2, /* 6 dp3 */ - 2, /* 7 dp4 */ - 1, /* 8 frc */ - 1, /* 9 rcp */ - 1, /* a rsq */ - 1, /* b exp */ - 1, /* c log */ - 3, /* d cmp */ - 2, /* e min */ - 2, /* f max */ - 1, /* 10 flr */ - 1, /* 11 mod */ - 1, /* 12 trc */ - 2, /* 13 sge */ - 2, /* 14 slt */ - 1, - 1, - 1, - 1, - 0, - 0, - 0, - 0, - 0, - 0, - 0, -}; - - -static const char *regname[0x8] = { - "R", - "T", - "CONST", - "S", - "OC", - "OD", - "U", - "UNKNOWN", -}; - -static void -print_reg_type_nr(GLuint type, GLuint nr) -{ - switch (type) { - case REG_TYPE_T: - switch (nr) { - case T_DIFFUSE: - fprintf(stderr, "T_DIFFUSE"); - return; - case T_SPECULAR: - fprintf(stderr, "T_SPECULAR"); - return; - case T_FOG_W: - fprintf(stderr, "T_FOG_W"); - return; - default: - fprintf(stderr, "T_TEX%d", nr); - return; - } - case REG_TYPE_OC: - if (nr == 0) { - fprintf(stderr, "oC"); - return; - } - break; - case REG_TYPE_OD: - if (nr == 0) { - fprintf(stderr, "oD"); - return; - } - break; - default: - break; - } - - fprintf(stderr, "%s[%d]", regname[type], nr); -} - -#define REG_SWIZZLE_MASK 0x7777 -#define REG_NEGATE_MASK 0x8888 - -#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ - (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ - (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ - (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) - - -static void -print_reg_neg_swizzle(GLuint reg) -{ - int i; - - if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && - (reg & REG_NEGATE_MASK) == 0) - return; - - fprintf(stderr, "."); - - for (i = 3; i >= 0; i--) { - if (reg & (1 << ((i * 4) + 3))) - fprintf(stderr, "-"); - - switch ((reg >> (i * 4)) & 0x7) { - case 0: - fprintf(stderr, "x"); - break; - case 1: - fprintf(stderr, "y"); - break; - case 2: - fprintf(stderr, "z"); - break; - case 3: - fprintf(stderr, "w"); - break; - case 4: - fprintf(stderr, "0"); - break; - case 5: - fprintf(stderr, "1"); - break; - default: - fprintf(stderr, "?"); - break; - } - } -} - - -static void -print_src_reg(GLuint dword) -{ - GLuint nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; - GLuint type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(type, nr); - print_reg_neg_swizzle(dword); -} - -void -i915_print_ureg(const char *msg, GLuint ureg) -{ - fprintf(stderr, "%s: ", msg); - print_src_reg(ureg >> 8); - fprintf(stderr, "\n"); -} - -static void -print_dest_reg(GLuint dword) -{ - GLuint nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; - GLuint type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(type, nr); - if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) - return; - fprintf(stderr, "."); - if (dword & A0_DEST_CHANNEL_X) - fprintf(stderr, "x"); - if (dword & A0_DEST_CHANNEL_Y) - fprintf(stderr, "y"); - if (dword & A0_DEST_CHANNEL_Z) - fprintf(stderr, "z"); - if (dword & A0_DEST_CHANNEL_W) - fprintf(stderr, "w"); -} - - -#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) -#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) -#define GET_SRC2_REG(r) (r) - - -static void -print_arith_op(GLuint opcode, const GLuint * program) -{ - if (opcode != A0_NOP) { - print_dest_reg(program[0]); - if (program[0] & A0_DEST_SATURATE) - fprintf(stderr, " = SATURATE "); - else - fprintf(stderr, " = "); - } - - fprintf(stderr, "%s ", opcodes[opcode]); - - print_src_reg(GET_SRC0_REG(program[0], program[1])); - if (args[opcode] == 1) { - fprintf(stderr, "\n"); - return; - } - - fprintf(stderr, ", "); - print_src_reg(GET_SRC1_REG(program[1], program[2])); - if (args[opcode] == 2) { - fprintf(stderr, "\n"); - return; - } - - fprintf(stderr, ", "); - print_src_reg(GET_SRC2_REG(program[2])); - fprintf(stderr, "\n"); - return; -} - - -static void -print_tex_op(GLuint opcode, const GLuint * program) -{ - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - fprintf(stderr, " = "); - - fprintf(stderr, "%s ", opcodes[opcode]); - - fprintf(stderr, "S[%d],", program[0] & T0_SAMPLER_NR_MASK); - - print_reg_type_nr((program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & - REG_TYPE_MASK, - (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); - fprintf(stderr, "\n"); -} - -static void -print_dcl_op(GLuint opcode, const GLuint * program) -{ - fprintf(stderr, "%s ", opcodes[opcode]); - print_dest_reg(program[0] | A0_DEST_CHANNEL_ALL); - fprintf(stderr, "\n"); -} - - -void -i915_disassemble_program(const GLuint * program, GLuint sz) -{ - GLuint size = program[0] & 0x1ff; - GLint i; - - fprintf(stderr, "BEGIN\n"); - - if (size + 2 != sz) { - fprintf(stderr, "%s: program size mismatch %d/%d\n", __FUNCTION__, - size + 2, sz); - exit(1); - } - - program++; - for (i = 1; i < sz; i += 3, program += 3) { - GLuint opcode = program[0] & (0x1f << 24); - - if ((GLint) opcode >= A0_NOP && opcode <= A0_SLT) - print_arith_op(opcode >> 24, program); - else if (opcode >= T0_TEXLD && opcode <= T0_TEXKILL) - print_tex_op(opcode >> 24, program); - else if (opcode == D0_DCL) - print_dcl_op(opcode >> 24, program); - else - fprintf(stderr, "Unknown opcode 0x%x\n", opcode); - } - - fprintf(stderr, "END\n\n"); -} diff --git a/src/mesa/drivers/dri/i915tex/i915_fragprog.c b/src/mesa/drivers/dri/i915tex/i915_fragprog.c deleted file mode 100644 index 95ec50490a..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_fragprog.c +++ /dev/null @@ -1,1073 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "macros.h" -#include "enums.h" - -#include "tnl/tnl.h" -#include "tnl/t_context.h" -#include "intel_batchbuffer.h" - -#include "i915_reg.h" -#include "i915_context.h" -#include "i915_program.h" - -#include "prog_instruction.h" -#include "prog_parameter.h" -#include "program.h" -#include "programopt.h" - - - -/* 1, -1/3!, 1/5!, -1/7! */ -static const GLfloat sin_constants[4] = { 1.0, - -1.0 / (3 * 2 * 1), - 1.0 / (5 * 4 * 3 * 2 * 1), - -1.0 / (7 * 6 * 5 * 4 * 3 * 2 * 1) -}; - -/* 1, -1/2!, 1/4!, -1/6! */ -static const GLfloat cos_constants[4] = { 1.0, - -1.0 / (2 * 1), - 1.0 / (4 * 3 * 2 * 1), - -1.0 / (6 * 5 * 4 * 3 * 2 * 1) -}; - -/** - * Retrieve a ureg for the given source register. Will emit - * constants, apply swizzling and negation as needed. - */ -static GLuint -src_vector(struct i915_fragment_program *p, - const struct prog_src_register *source, - const struct gl_fragment_program *program) -{ - GLuint src; - - switch (source->File) { - - /* Registers: - */ - case PROGRAM_TEMPORARY: - if (source->Index >= I915_MAX_TEMPORARY) { - i915_program_error(p, "Exceeded max temporary reg"); - return 0; - } - src = UREG(REG_TYPE_R, source->Index); - break; - case PROGRAM_INPUT: - switch (source->Index) { - case FRAG_ATTRIB_WPOS: - src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); - break; - case FRAG_ATTRIB_COL0: - src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); - break; - case FRAG_ATTRIB_COL1: - src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); - src = swizzle(src, X, Y, Z, ONE); - break; - case FRAG_ATTRIB_FOGC: - src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); - src = swizzle(src, W, W, W, W); - break; - case FRAG_ATTRIB_TEX0: - case FRAG_ATTRIB_TEX1: - case FRAG_ATTRIB_TEX2: - case FRAG_ATTRIB_TEX3: - case FRAG_ATTRIB_TEX4: - case FRAG_ATTRIB_TEX5: - case FRAG_ATTRIB_TEX6: - case FRAG_ATTRIB_TEX7: - src = i915_emit_decl(p, REG_TYPE_T, - T_TEX0 + (source->Index - FRAG_ATTRIB_TEX0), - D0_CHANNEL_ALL); - break; - - default: - i915_program_error(p, "Bad source->Index"); - return 0; - } - break; - - /* Various paramters and env values. All emitted to - * hardware as program constants. - */ - case PROGRAM_LOCAL_PARAM: - src = i915_emit_param4fv(p, program->Base.LocalParams[source->Index]); - break; - - case PROGRAM_ENV_PARAM: - src = - i915_emit_param4fv(p, - p->ctx->FragmentProgram.Parameters[source-> - Index]); - break; - - case PROGRAM_CONSTANT: - case PROGRAM_STATE_VAR: - case PROGRAM_NAMED_PARAM: - src = - i915_emit_param4fv(p, - program->Base.Parameters->ParameterValues[source-> - Index]); - break; - - default: - i915_program_error(p, "Bad source->File"); - return 0; - } - - src = swizzle(src, - GET_SWZ(source->Swizzle, 0), - GET_SWZ(source->Swizzle, 1), - GET_SWZ(source->Swizzle, 2), GET_SWZ(source->Swizzle, 3)); - - if (source->NegateBase) - src = negate(src, - GET_BIT(source->NegateBase, 0), - GET_BIT(source->NegateBase, 1), - GET_BIT(source->NegateBase, 2), - GET_BIT(source->NegateBase, 3)); - - return src; -} - - -static GLuint -get_result_vector(struct i915_fragment_program *p, - const struct prog_instruction *inst) -{ - switch (inst->DstReg.File) { - case PROGRAM_OUTPUT: - switch (inst->DstReg.Index) { - case FRAG_RESULT_COLR: - return UREG(REG_TYPE_OC, 0); - case FRAG_RESULT_DEPR: - p->depth_written = 1; - return UREG(REG_TYPE_OD, 0); - default: - i915_program_error(p, "Bad inst->DstReg.Index"); - return 0; - } - case PROGRAM_TEMPORARY: - return UREG(REG_TYPE_R, inst->DstReg.Index); - default: - i915_program_error(p, "Bad inst->DstReg.File"); - return 0; - } -} - -static GLuint -get_result_flags(const struct prog_instruction *inst) -{ - GLuint flags = 0; - - if (inst->SaturateMode == SATURATE_ZERO_ONE) - flags |= A0_DEST_SATURATE; - if (inst->DstReg.WriteMask & WRITEMASK_X) - flags |= A0_DEST_CHANNEL_X; - if (inst->DstReg.WriteMask & WRITEMASK_Y) - flags |= A0_DEST_CHANNEL_Y; - if (inst->DstReg.WriteMask & WRITEMASK_Z) - flags |= A0_DEST_CHANNEL_Z; - if (inst->DstReg.WriteMask & WRITEMASK_W) - flags |= A0_DEST_CHANNEL_W; - - return flags; -} - -static GLuint -translate_tex_src_target(struct i915_fragment_program *p, GLubyte bit) -{ - switch (bit) { - case TEXTURE_1D_INDEX: - return D0_SAMPLE_TYPE_2D; - case TEXTURE_2D_INDEX: - return D0_SAMPLE_TYPE_2D; - case TEXTURE_RECT_INDEX: - return D0_SAMPLE_TYPE_2D; - case TEXTURE_3D_INDEX: - return D0_SAMPLE_TYPE_VOLUME; - case TEXTURE_CUBE_INDEX: - return D0_SAMPLE_TYPE_CUBE; - default: - i915_program_error(p, "TexSrcBit"); - return 0; - } -} - -#define EMIT_TEX( OP ) \ -do { \ - GLuint dim = translate_tex_src_target( p, inst->TexSrcTarget ); \ - GLuint sampler = i915_emit_decl(p, REG_TYPE_S, \ - inst->TexSrcUnit, dim); \ - GLuint coord = src_vector( p, &inst->SrcReg[0], program); \ - /* Texel lookup */ \ - \ - i915_emit_texld( p, \ - get_result_vector( p, inst ), \ - get_result_flags( inst ), \ - sampler, \ - coord, \ - OP); \ -} while (0) - -#define EMIT_ARITH( OP, N ) \ -do { \ - i915_emit_arith( p, \ - OP, \ - get_result_vector( p, inst ), \ - get_result_flags( inst ), 0, \ - (N<1)?0:src_vector( p, &inst->SrcReg[0], program), \ - (N<2)?0:src_vector( p, &inst->SrcReg[1], program), \ - (N<3)?0:src_vector( p, &inst->SrcReg[2], program)); \ -} while (0) - -#define EMIT_1ARG_ARITH( OP ) EMIT_ARITH( OP, 1 ) -#define EMIT_2ARG_ARITH( OP ) EMIT_ARITH( OP, 2 ) -#define EMIT_3ARG_ARITH( OP ) EMIT_ARITH( OP, 3 ) - - -/* Possible concerns: - * - * SIN, COS -- could use another taylor step? - * LIT -- results seem a little different to sw mesa - * LOG -- different to mesa on negative numbers, but this is conformant. - * - * Parse failures -- Mesa doesn't currently give a good indication - * internally whether a particular program string parsed or not. This - * can lead to confusion -- hopefully we cope with it ok now. - * - */ -static void -upload_program(struct i915_fragment_program *p) -{ - const struct gl_fragment_program *program = - p->ctx->FragmentProgram._Current; - const struct prog_instruction *inst = program->Base.Instructions; - -/* _mesa_debug_fp_inst(program->Base.NumInstructions, inst); */ - - /* Is this a parse-failed program? Ensure a valid program is - * loaded, as the flagging of an error isn't sufficient to stop - * this being uploaded to hardware. - */ - if (inst[0].Opcode == OPCODE_END) { - GLuint tmp = i915_get_utemp(p); - i915_emit_arith(p, - A0_MOV, - UREG(REG_TYPE_OC, 0), - A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, ONE, ZERO, ONE, ONE), 0, 0); - return; - } - - while (1) { - GLuint src0, src1, src2, flags; - GLuint tmp = 0; - - switch (inst->Opcode) { - case OPCODE_ABS: - src0 = src_vector(p, &inst->SrcReg[0], program); - i915_emit_arith(p, - A0_MAX, - get_result_vector(p, inst), - get_result_flags(inst), 0, - src0, negate(src0, 1, 1, 1, 1), 0); - break; - - case OPCODE_ADD: - EMIT_2ARG_ARITH(A0_ADD); - break; - - case OPCODE_CMP: - src0 = src_vector(p, &inst->SrcReg[0], program); - src1 = src_vector(p, &inst->SrcReg[1], program); - src2 = src_vector(p, &inst->SrcReg[2], program); - i915_emit_arith(p, A0_CMP, get_result_vector(p, inst), get_result_flags(inst), 0, src0, src2, src1); /* NOTE: order of src2, src1 */ - break; - - case OPCODE_COS: - src0 = src_vector(p, &inst->SrcReg[0], program); - tmp = i915_get_utemp(p); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - src0, i915_emit_const1f(p, 1.0 / (M_PI)), 0); - - i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); - - /* By choosing different taylor constants, could get rid of this mul: - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, i915_emit_const1f(p, (M_PI)), 0); - - /* - * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 - * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, 1 - * t0 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 - * result = DP4 t0, cos_constants - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XY, 0, - swizzle(tmp, X, X, ONE, ONE), - swizzle(tmp, X, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XYZ, 0, - swizzle(tmp, X, Y, X, ONE), - swizzle(tmp, X, X, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XYZ, 0, - swizzle(tmp, X, X, Z, ONE), - swizzle(tmp, Z, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, inst), - get_result_flags(inst), 0, - swizzle(tmp, ONE, Z, Y, X), - i915_emit_const4fv(p, cos_constants), 0); - - break; - - case OPCODE_DP3: - EMIT_2ARG_ARITH(A0_DP3); - break; - - case OPCODE_DP4: - EMIT_2ARG_ARITH(A0_DP4); - break; - - case OPCODE_DPH: - src0 = src_vector(p, &inst->SrcReg[0], program); - src1 = src_vector(p, &inst->SrcReg[1], program); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, inst), - get_result_flags(inst), 0, - swizzle(src0, X, Y, Z, ONE), src1, 0); - break; - - case OPCODE_DST: - src0 = src_vector(p, &inst->SrcReg[0], program); - src1 = src_vector(p, &inst->SrcReg[1], program); - - /* result[0] = 1 * 1; - * result[1] = a[1] * b[1]; - * result[2] = a[2] * 1; - * result[3] = 1 * b[3]; - */ - i915_emit_arith(p, - A0_MUL, - get_result_vector(p, inst), - get_result_flags(inst), 0, - swizzle(src0, ONE, Y, Z, ONE), - swizzle(src1, ONE, Y, ONE, W), 0); - break; - - case OPCODE_EX2: - src0 = src_vector(p, &inst->SrcReg[0], program); - - i915_emit_arith(p, - A0_EXP, - get_result_vector(p, inst), - get_result_flags(inst), 0, - swizzle(src0, X, X, X, X), 0, 0); - break; - - case OPCODE_FLR: - EMIT_1ARG_ARITH(A0_FLR); - break; - - case OPCODE_FRC: - EMIT_1ARG_ARITH(A0_FRC); - break; - - case OPCODE_KIL: - src0 = src_vector(p, &inst->SrcReg[0], program); - tmp = i915_get_utemp(p); - - i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ - 0, src0, T0_TEXKILL); - break; - - case OPCODE_LG2: - src0 = src_vector(p, &inst->SrcReg[0], program); - - i915_emit_arith(p, - A0_LOG, - get_result_vector(p, inst), - get_result_flags(inst), 0, - swizzle(src0, X, X, X, X), 0, 0); - break; - - case OPCODE_LIT: - src0 = src_vector(p, &inst->SrcReg[0], program); - tmp = i915_get_utemp(p); - - /* tmp = max( a.xyzw, a.00zw ) - * XXX: Clamp tmp.w to -128..128 - * tmp.y = log(tmp.y) - * tmp.y = tmp.w * tmp.y - * tmp.y = exp(tmp.y) - * result = cmp (a.11-x1, a.1x01, a.1xy1 ) - */ - i915_emit_arith(p, A0_MAX, tmp, A0_DEST_CHANNEL_ALL, 0, - src0, swizzle(src0, ZERO, ZERO, Z, W), 0); - - i915_emit_arith(p, A0_LOG, tmp, A0_DEST_CHANNEL_Y, 0, - swizzle(tmp, Y, Y, Y, Y), 0, 0); - - i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_Y, 0, - swizzle(tmp, ZERO, Y, ZERO, ZERO), - swizzle(tmp, ZERO, W, ZERO, ZERO), 0); - - i915_emit_arith(p, A0_EXP, tmp, A0_DEST_CHANNEL_Y, 0, - swizzle(tmp, Y, Y, Y, Y), 0, 0); - - i915_emit_arith(p, A0_CMP, - get_result_vector(p, inst), - get_result_flags(inst), 0, - negate(swizzle(tmp, ONE, ONE, X, ONE), 0, 0, 1, 0), - swizzle(tmp, ONE, X, ZERO, ONE), - swizzle(tmp, ONE, X, Y, ONE)); - - break; - - case OPCODE_LRP: - src0 = src_vector(p, &inst->SrcReg[0], program); - src1 = src_vector(p, &inst->SrcReg[1], program); - src2 = src_vector(p, &inst->SrcReg[2], program); - flags = get_result_flags(inst); - tmp = i915_get_utemp(p); - - /* b*a + c*(1-a) - * - * b*a + c - ca - * - * tmp = b*a + c, - * result = (-c)*a + tmp - */ - i915_emit_arith(p, A0_MAD, tmp, - flags & A0_DEST_CHANNEL_ALL, 0, src1, src0, src2); - - i915_emit_arith(p, A0_MAD, - get_result_vector(p, inst), - flags, 0, negate(src2, 1, 1, 1, 1), src0, tmp); - break; - - case OPCODE_MAD: - EMIT_3ARG_ARITH(A0_MAD); - break; - - case OPCODE_MAX: - EMIT_2ARG_ARITH(A0_MAX); - break; - - case OPCODE_MIN: - src0 = src_vector(p, &inst->SrcReg[0], program); - src1 = src_vector(p, &inst->SrcReg[1], program); - tmp = i915_get_utemp(p); - flags = get_result_flags(inst); - - i915_emit_arith(p, - A0_MAX, - tmp, flags & A0_DEST_CHANNEL_ALL, 0, - negate(src0, 1, 1, 1, 1), - negate(src1, 1, 1, 1, 1), 0); - - i915_emit_arith(p, - A0_MOV, - get_result_vector(p, inst), - flags, 0, negate(tmp, 1, 1, 1, 1), 0, 0); - break; - - case OPCODE_MOV: - EMIT_1ARG_ARITH(A0_MOV); - break; - - case OPCODE_MUL: - EMIT_2ARG_ARITH(A0_MUL); - break; - - case OPCODE_POW: - src0 = src_vector(p, &inst->SrcReg[0], program); - src1 = src_vector(p, &inst->SrcReg[1], program); - tmp = i915_get_utemp(p); - flags = get_result_flags(inst); - - /* XXX: masking on intermediate values, here and elsewhere. - */ - i915_emit_arith(p, - A0_LOG, - tmp, A0_DEST_CHANNEL_X, 0, - swizzle(src0, X, X, X, X), 0, 0); - - i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, tmp, src1, 0); - - - i915_emit_arith(p, - A0_EXP, - get_result_vector(p, inst), - flags, 0, swizzle(tmp, X, X, X, X), 0, 0); - - break; - - case OPCODE_RCP: - src0 = src_vector(p, &inst->SrcReg[0], program); - - i915_emit_arith(p, - A0_RCP, - get_result_vector(p, inst), - get_result_flags(inst), 0, - swizzle(src0, X, X, X, X), 0, 0); - break; - - case OPCODE_RSQ: - - src0 = src_vector(p, &inst->SrcReg[0], program); - - i915_emit_arith(p, - A0_RSQ, - get_result_vector(p, inst), - get_result_flags(inst), 0, - swizzle(src0, X, X, X, X), 0, 0); - break; - - case OPCODE_SCS: - src0 = src_vector(p, &inst->SrcReg[0], program); - tmp = i915_get_utemp(p); - - /* - * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 - * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x - * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x - * scs.x = DP4 t1, sin_constants - * t1 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 - * scs.y = DP4 t1, cos_constants - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XY, 0, - swizzle(src0, X, X, ONE, ONE), - swizzle(src0, X, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X, Y, X, Y), - swizzle(tmp, X, X, ONE, ONE), 0); - - if (inst->DstReg.WriteMask & WRITEMASK_Y) { - GLuint tmp1; - - if (inst->DstReg.WriteMask & WRITEMASK_X) - tmp1 = i915_get_utemp(p); - else - tmp1 = tmp; - - i915_emit_arith(p, - A0_MUL, - tmp1, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X, Y, Y, W), - swizzle(tmp, X, Z, ONE, ONE), 0); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, inst), - A0_DEST_CHANNEL_Y, 0, - swizzle(tmp1, W, Z, Y, X), - i915_emit_const4fv(p, sin_constants), 0); - } - - if (inst->DstReg.WriteMask & WRITEMASK_X) { - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XYZ, 0, - swizzle(tmp, X, X, Z, ONE), - swizzle(tmp, Z, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, inst), - A0_DEST_CHANNEL_X, 0, - swizzle(tmp, ONE, Z, Y, X), - i915_emit_const4fv(p, cos_constants), 0); - } - break; - - case OPCODE_SGE: - EMIT_2ARG_ARITH(A0_SGE); - break; - - case OPCODE_SIN: - src0 = src_vector(p, &inst->SrcReg[0], program); - tmp = i915_get_utemp(p); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - src0, i915_emit_const1f(p, 1.0 / (M_PI)), 0); - - i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); - - /* By choosing different taylor constants, could get rid of this mul: - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, i915_emit_const1f(p, (M_PI)), 0); - - /* - * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 - * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x - * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x - * result = DP4 t1.wzyx, sin_constants - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XY, 0, - swizzle(tmp, X, X, ONE, ONE), - swizzle(tmp, X, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X, Y, X, Y), - swizzle(tmp, X, X, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X, Y, Y, W), - swizzle(tmp, X, Z, ONE, ONE), 0); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, inst), - get_result_flags(inst), 0, - swizzle(tmp, W, Z, Y, X), - i915_emit_const4fv(p, sin_constants), 0); - break; - - case OPCODE_SLT: - EMIT_2ARG_ARITH(A0_SLT); - break; - - case OPCODE_SUB: - src0 = src_vector(p, &inst->SrcReg[0], program); - src1 = src_vector(p, &inst->SrcReg[1], program); - - i915_emit_arith(p, - A0_ADD, - get_result_vector(p, inst), - get_result_flags(inst), 0, - src0, negate(src1, 1, 1, 1, 1), 0); - break; - - case OPCODE_SWZ: - EMIT_1ARG_ARITH(A0_MOV); /* extended swizzle handled natively */ - break; - - case OPCODE_TEX: - EMIT_TEX(T0_TEXLD); - break; - - case OPCODE_TXB: - EMIT_TEX(T0_TEXLDB); - break; - - case OPCODE_TXP: - EMIT_TEX(T0_TEXLDP); - break; - - case OPCODE_XPD: - /* Cross product: - * result.x = src0.y * src1.z - src0.z * src1.y; - * result.y = src0.z * src1.x - src0.x * src1.z; - * result.z = src0.x * src1.y - src0.y * src1.x; - * result.w = undef; - */ - src0 = src_vector(p, &inst->SrcReg[0], program); - src1 = src_vector(p, &inst->SrcReg[1], program); - tmp = i915_get_utemp(p); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(src0, Z, X, Y, ONE), - swizzle(src1, Y, Z, X, ONE), 0); - - i915_emit_arith(p, - A0_MAD, - get_result_vector(p, inst), - get_result_flags(inst), 0, - swizzle(src0, Y, Z, X, ONE), - swizzle(src1, Z, X, Y, ONE), - negate(tmp, 1, 1, 1, 0)); - break; - - case OPCODE_END: - return; - - default: - i915_program_error(p, "bad opcode"); - return; - } - - inst++; - i915_release_utemps(p); - } -} - -/* Rather than trying to intercept and jiggle depth writes during - * emit, just move the value into its correct position at the end of - * the program: - */ -static void -fixup_depth_write(struct i915_fragment_program *p) -{ - if (p->depth_written) { - GLuint depth = UREG(REG_TYPE_OD, 0); - - i915_emit_arith(p, - A0_MOV, - depth, A0_DEST_CHANNEL_W, 0, - swizzle(depth, X, Y, Z, Z), 0, 0); - } -} - - -static void -check_wpos(struct i915_fragment_program *p) -{ - GLuint inputs = p->FragProg.Base.InputsRead; - GLint i; - - p->wpos_tex = -1; - - for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { - if (inputs & FRAG_BIT_TEX(i)) - continue; - else if (inputs & FRAG_BIT_WPOS) { - p->wpos_tex = i; - inputs &= ~FRAG_BIT_WPOS; - } - } - - if (inputs & FRAG_BIT_WPOS) { - i915_program_error(p, "No free texcoord for wpos value"); - } -} - - -static void -translate_program(struct i915_fragment_program *p) -{ - struct i915_context *i915 = I915_CONTEXT(p->ctx); - - i915_init_program(i915, p); - check_wpos(p); - upload_program(p); - fixup_depth_write(p); - i915_fini_program(p); - - p->translated = 1; -} - - -static void -track_params(struct i915_fragment_program *p) -{ - GLint i; - - if (p->nr_params) - _mesa_load_state_parameters(p->ctx, p->FragProg.Base.Parameters); - - for (i = 0; i < p->nr_params; i++) { - GLint reg = p->param[i].reg; - COPY_4V(p->constant[reg], p->param[i].values); - } - - p->params_uptodate = 1; - p->on_hardware = 0; /* overkill */ -} - - -static void -i915BindProgram(GLcontext * ctx, GLenum target, struct gl_program *prog) -{ - if (target == GL_FRAGMENT_PROGRAM_ARB) { - struct i915_context *i915 = I915_CONTEXT(ctx); - struct i915_fragment_program *p = (struct i915_fragment_program *) prog; - - if (i915->current_program == p) - return; - - if (i915->current_program) { - i915->current_program->on_hardware = 0; - i915->current_program->params_uptodate = 0; - } - - i915->current_program = p; - - assert(p->on_hardware == 0); - assert(p->params_uptodate == 0); - - } -} - -static struct gl_program * -i915NewProgram(GLcontext * ctx, GLenum target, GLuint id) -{ - switch (target) { - case GL_VERTEX_PROGRAM_ARB: - return _mesa_init_vertex_program(ctx, CALLOC_STRUCT(gl_vertex_program), - target, id); - - case GL_FRAGMENT_PROGRAM_ARB:{ - struct i915_fragment_program *prog = - CALLOC_STRUCT(i915_fragment_program); - if (prog) { - i915_init_program(I915_CONTEXT(ctx), prog); - - return _mesa_init_fragment_program(ctx, &prog->FragProg, - target, id); - } - else - return NULL; - } - - default: - /* Just fallback: - */ - return _mesa_new_program(ctx, target, id); - } -} - -static void -i915DeleteProgram(GLcontext * ctx, struct gl_program *prog) -{ - if (prog->Target == GL_FRAGMENT_PROGRAM_ARB) { - struct i915_context *i915 = I915_CONTEXT(ctx); - struct i915_fragment_program *p = (struct i915_fragment_program *) prog; - - if (i915->current_program == p) - i915->current_program = 0; - } - - _mesa_delete_program(ctx, prog); -} - - -static GLboolean -i915IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog) -{ - if (target == GL_FRAGMENT_PROGRAM_ARB) { - struct i915_fragment_program *p = (struct i915_fragment_program *) prog; - - if (!p->translated) - translate_program(p); - - return !p->error; - } - else - return GL_TRUE; -} - -static void -i915ProgramStringNotify(GLcontext * ctx, - GLenum target, struct gl_program *prog) -{ - if (target == GL_FRAGMENT_PROGRAM_ARB) { - struct i915_fragment_program *p = (struct i915_fragment_program *) prog; - p->translated = 0; - - /* Hack: make sure fog is correctly enabled according to this - * fragment program's fog options. - */ - if (p->FragProg.FogOption) { - /* add extra instructions to do fog, then turn off FogOption field */ - _mesa_append_fog_code(ctx, &p->FragProg); - p->FragProg.FogOption = GL_NONE; - } - } - - _tnl_program_string(ctx, target, prog); -} - - -void -i915ValidateFragmentProgram(struct i915_context *i915) -{ - GLcontext *ctx = &i915->intel.ctx; - struct intel_context *intel = intel_context(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *VB = &tnl->vb; - - struct i915_fragment_program *p = - (struct i915_fragment_program *) ctx->FragmentProgram._Current; - - const GLuint inputsRead = p->FragProg.Base.InputsRead; - GLuint s4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_VFMT_MASK; - GLuint s2 = S2_TEXCOORD_NONE; - int i, offset = 0; - - if (i915->current_program != p) { - if (i915->current_program) { - i915->current_program->on_hardware = 0; - i915->current_program->params_uptodate = 0; - } - - i915->current_program = p; - } - - - /* Important: - */ - VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr; - - if (!p->translated) - translate_program(p); - - intel->vertex_attr_count = 0; - intel->wpos_offset = 0; - intel->wpos_size = 0; - intel->coloroffset = 0; - intel->specoffset = 0; - - if (inputsRead & FRAG_BITS_TEX_ANY) { - EMIT_ATTR(_TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, S4_VFMT_XYZW, 16); - } - else { - EMIT_ATTR(_TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, S4_VFMT_XYZ, 12); - } - - if (inputsRead & FRAG_BIT_COL0) { - intel->coloroffset = offset / 4; - EMIT_ATTR(_TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, S4_VFMT_COLOR, 4); - } - - if ((inputsRead & (FRAG_BIT_COL1 | FRAG_BIT_FOGC)) || - i915->vertex_fog != I915_FOG_NONE) { - - if (inputsRead & FRAG_BIT_COL1) { - intel->specoffset = offset / 4; - EMIT_ATTR(_TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, S4_VFMT_SPEC_FOG, 3); - } - else - EMIT_PAD(3); - - if ((inputsRead & FRAG_BIT_FOGC) || i915->vertex_fog != I915_FOG_NONE) - EMIT_ATTR(_TNL_ATTRIB_FOG, EMIT_1UB_1F, S4_VFMT_SPEC_FOG, 1); - else - EMIT_PAD(1); - } - - /* XXX this was disabled, but enabling this code helped fix the Glean - * tfragprog1 fog tests. - */ -#if 1 - if ((inputsRead & FRAG_BIT_FOGC) || i915->vertex_fog != I915_FOG_NONE) { - EMIT_ATTR(_TNL_ATTRIB_FOG, EMIT_1F, S4_VFMT_FOG_PARAM, 4); - } -#endif - - for (i = 0; i < p->ctx->Const.MaxTextureCoordUnits; i++) { - if (inputsRead & FRAG_BIT_TEX(i)) { - int sz = VB->TexCoordPtr[i]->size; - - s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); - s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(sz)); - - EMIT_ATTR(_TNL_ATTRIB_TEX0 + i, EMIT_SZ(sz), 0, sz * 4); - } - else if (i == p->wpos_tex) { - - /* If WPOS is required, duplicate the XYZ position data in an - * unused texture coordinate: - */ - s2 &= ~S2_TEXCOORD_FMT(i, S2_TEXCOORD_FMT0_MASK); - s2 |= S2_TEXCOORD_FMT(i, SZ_TO_HW(3)); - - intel->wpos_offset = offset; - intel->wpos_size = 3 * sizeof(GLuint); - - EMIT_PAD(intel->wpos_size); - } - } - - if (s2 != i915->state.Ctx[I915_CTXREG_LIS2] || - s4 != i915->state.Ctx[I915_CTXREG_LIS4]) { - int k; - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - - /* Must do this *after* statechange, so as not to affect - * buffered vertices reliant on the old state: - */ - intel->vertex_size = _tnl_install_attrs(&intel->ctx, - intel->vertex_attrs, - intel->vertex_attr_count, - intel->ViewportMatrix.m, 0); - - intel->vertex_size >>= 2; - - i915->state.Ctx[I915_CTXREG_LIS2] = s2; - i915->state.Ctx[I915_CTXREG_LIS4] = s4; - - k = intel->vtbl.check_vertex_size(intel, intel->vertex_size); - assert(k); - } - - if (!p->params_uptodate) - track_params(p); - - if (!p->on_hardware) - i915_upload_program(i915, p); -} - -void -i915InitFragProgFuncs(struct dd_function_table *functions) -{ - functions->BindProgram = i915BindProgram; - functions->NewProgram = i915NewProgram; - functions->DeleteProgram = i915DeleteProgram; - functions->IsProgramNative = i915IsProgramNative; - functions->ProgramStringNotify = i915ProgramStringNotify; -} diff --git a/src/mesa/drivers/dri/i915tex/i915_metaops.c b/src/mesa/drivers/dri/i915tex/i915_metaops.c deleted file mode 100644 index 3ab62bc806..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_metaops.c +++ /dev/null @@ -1,509 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "enums.h" -#include "mtypes.h" -#include "macros.h" -#include "utils.h" - -#include "intel_screen.h" -#include "intel_batchbuffer.h" -#include "intel_ioctl.h" -#include "intel_regions.h" -#include "intel_rotate.h" - -#include "i915_context.h" -#include "i915_reg.h" - -/* We touch almost everything: - */ -#define ACTIVE (I915_UPLOAD_INVARIENT | \ - I915_UPLOAD_CTX | \ - I915_UPLOAD_BUFFERS | \ - I915_UPLOAD_STIPPLE | \ - I915_UPLOAD_PROGRAM | \ - I915_UPLOAD_FOG | \ - I915_UPLOAD_TEX(0)) - -#define SET_STATE( i915, STATE ) \ -do { \ - i915->current->emitted &= ~ACTIVE; \ - i915->current = &i915->STATE; \ - i915->current->emitted &= ~ACTIVE; \ -} while (0) - - -static void -meta_no_stencil_write(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - - /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_FALSE ) - */ - i915->meta.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE); - - i915->meta.emitted &= ~I915_UPLOAD_CTX; -} - -static void -meta_no_depth_write(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - - /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_FALSE ) - */ - i915->meta.Ctx[I915_CTXREG_LIS6] &= ~(S6_DEPTH_TEST_ENABLE | - S6_DEPTH_WRITE_ENABLE); - - i915->meta.emitted &= ~I915_UPLOAD_CTX; -} - -static void -meta_depth_replace(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - - /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, GL_TRUE ) - * ctx->Driver.DepthMask( ctx, GL_TRUE ) - */ - i915->meta.Ctx[I915_CTXREG_LIS6] |= (S6_DEPTH_TEST_ENABLE | - S6_DEPTH_WRITE_ENABLE); - - /* ctx->Driver.DepthFunc( ctx, GL_REPLACE ) - */ - i915->meta.Ctx[I915_CTXREG_LIS6] &= ~S6_DEPTH_TEST_FUNC_MASK; - i915->meta.Ctx[I915_CTXREG_LIS6] |= - COMPAREFUNC_ALWAYS << S6_DEPTH_TEST_FUNC_SHIFT; - - i915->meta.emitted &= ~I915_UPLOAD_CTX; -} - - -/* Set stencil unit to replace always with the reference value. - */ -static void -meta_stencil_replace(struct intel_context *intel, - GLuint s_mask, GLuint s_clear) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - GLuint op = STENCILOP_REPLACE; - GLuint func = COMPAREFUNC_ALWAYS; - - /* ctx->Driver.Enable( ctx, GL_STENCIL_TEST, GL_TRUE ) - */ - i915->meta.Ctx[I915_CTXREG_LIS5] |= (S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE); - - /* ctx->Driver.StencilMask( ctx, s_mask ) - */ - i915->meta.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK; - - i915->meta.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(s_mask)); - - /* ctx->Driver.StencilOp( ctx, GL_REPLACE, GL_REPLACE, GL_REPLACE ) - */ - i915->meta.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_FAIL_MASK | - S5_STENCIL_PASS_Z_FAIL_MASK | - S5_STENCIL_PASS_Z_PASS_MASK); - - i915->meta.Ctx[I915_CTXREG_LIS5] |= ((op << S5_STENCIL_FAIL_SHIFT) | - (op << S5_STENCIL_PASS_Z_FAIL_SHIFT) | - (op << S5_STENCIL_PASS_Z_PASS_SHIFT)); - - - /* ctx->Driver.StencilFunc( ctx, GL_ALWAYS, s_ref, ~0 ) - */ - i915->meta.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_TEST_MASK; - i915->meta.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(0xff)); - - i915->meta.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_REF_MASK | - S5_STENCIL_TEST_FUNC_MASK); - - i915->meta.Ctx[I915_CTXREG_LIS5] |= ((s_clear << S5_STENCIL_REF_SHIFT) | - (func << S5_STENCIL_TEST_FUNC_SHIFT)); - - - i915->meta.emitted &= ~I915_UPLOAD_CTX; -} - - -static void -meta_color_mask(struct intel_context *intel, GLboolean state) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - const GLuint mask = (S5_WRITEDISABLE_RED | - S5_WRITEDISABLE_GREEN | - S5_WRITEDISABLE_BLUE | S5_WRITEDISABLE_ALPHA); - - /* Copy colormask state from "regular" hw context. - */ - if (state) { - i915->meta.Ctx[I915_CTXREG_LIS5] &= ~mask; - i915->meta.Ctx[I915_CTXREG_LIS5] |= - (i915->state.Ctx[I915_CTXREG_LIS5] & mask); - } - else - i915->meta.Ctx[I915_CTXREG_LIS5] |= mask; - - i915->meta.emitted &= ~I915_UPLOAD_CTX; -} - - - -static void -meta_import_pixel_state(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - memcpy(i915->meta.Fog, i915->state.Fog, I915_FOG_SETUP_SIZE * 4); - - i915->meta.Ctx[I915_CTXREG_LIS5] = i915->state.Ctx[I915_CTXREG_LIS5]; - i915->meta.Ctx[I915_CTXREG_LIS6] = i915->state.Ctx[I915_CTXREG_LIS6]; - i915->meta.Ctx[I915_CTXREG_STATE4] = i915->state.Ctx[I915_CTXREG_STATE4]; - i915->meta.Ctx[I915_CTXREG_BLENDCOLOR1] = - i915->state.Ctx[I915_CTXREG_BLENDCOLOR1]; - i915->meta.Ctx[I915_CTXREG_IAB] = i915->state.Ctx[I915_CTXREG_IAB]; - - i915->meta.Buffer[I915_DESTREG_SENABLE] = - i915->state.Buffer[I915_DESTREG_SENABLE]; - i915->meta.Buffer[I915_DESTREG_SR1] = i915->state.Buffer[I915_DESTREG_SR1]; - i915->meta.Buffer[I915_DESTREG_SR2] = i915->state.Buffer[I915_DESTREG_SR2]; - - i915->meta.emitted &= ~I915_UPLOAD_FOG; - i915->meta.emitted &= ~I915_UPLOAD_BUFFERS; - i915->meta.emitted &= ~I915_UPLOAD_CTX; -} - - - - -#define REG( type, nr ) (((type)<<5)|(nr)) - -#define REG_R(x) REG(REG_TYPE_R, x) -#define REG_T(x) REG(REG_TYPE_T, x) -#define REG_CONST(x) REG(REG_TYPE_CONST, x) -#define REG_S(x) REG(REG_TYPE_S, x) -#define REG_OC REG(REG_TYPE_OC, 0) -#define REG_OD REG(REG_TYPE_OD, 0) -#define REG_U(x) REG(REG_TYPE_U, x) - -#define REG_T_DIFFUSE REG(REG_TYPE_T, T_DIFFUSE) -#define REG_T_SPECULAR REG(REG_TYPE_T, T_SPECULAR) -#define REG_T_FOG_W REG(REG_TYPE_T, T_FOG_W) -#define REG_T_TEX(x) REG(REG_TYPE_T, x) - - -#define A0_DEST_REG( reg ) ( (reg) << A0_DEST_NR_SHIFT ) -#define A0_SRC0_REG( reg ) ( (reg) << A0_SRC0_NR_SHIFT ) -#define A1_SRC1_REG( reg ) ( (reg) << A1_SRC1_NR_SHIFT ) -#define A1_SRC2_REG( reg ) ( (reg) << A1_SRC2_NR_SHIFT ) -#define A2_SRC2_REG( reg ) ( (reg) << A2_SRC2_NR_SHIFT ) -#define D0_DECL_REG( reg ) ( (reg) << D0_NR_SHIFT ) -#define T0_DEST_REG( reg ) ( (reg) << T0_DEST_NR_SHIFT ) - -#define T0_SAMPLER( unit ) ((unit)<ctx); - - static const GLuint prog[] = { - _3DSTATE_PIXEL_SHADER_PROGRAM, - - /* Declare incoming diffuse color: - */ - (D0_DCL | D0_DECL_REG(REG_T_DIFFUSE) | D0_CHANNEL_ALL), - D1_MBZ, - D2_MBZ, - - /* output-color = mov(t_diffuse) - */ - (A0_MOV | - A0_DEST_REG(REG_OC) | - A0_DEST_CHANNEL_ALL | A0_SRC0_REG(REG_T_DIFFUSE)), - (A1_SRC0_XYZW), - 0, - }; - - - memcpy(i915->meta.Program, prog, sizeof(prog)); - i915->meta.ProgramSize = sizeof(prog) / sizeof(*prog); - i915->meta.Program[0] |= i915->meta.ProgramSize - 2; - i915->meta.emitted &= ~I915_UPLOAD_PROGRAM; -} - -static void -meta_texture_blend_replace(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - - static const GLuint prog[] = { - _3DSTATE_PIXEL_SHADER_PROGRAM, - - /* Declare the sampler: - */ - (D0_DCL | D0_DECL_REG(REG_S(0)) | D0_SAMPLE_TYPE_2D | D0_CHANNEL_NONE), - D1_MBZ, - D2_MBZ, - - /* Declare the interpolated texture coordinate: - */ - (D0_DCL | D0_DECL_REG(REG_T_TEX(0)) | D0_CHANNEL_ALL), - D1_MBZ, - D2_MBZ, - - /* output-color = texld(sample0, texcoord0) - */ - (T0_TEXLD | T0_DEST_REG(REG_OC) | T0_SAMPLER(0)), - T1_ADDRESS_REG(REG_TYPE_T, 0), - T2_MBZ - }; - - memcpy(i915->meta.Program, prog, sizeof(prog)); - i915->meta.ProgramSize = sizeof(prog) / sizeof(*prog); - i915->meta.Program[0] |= i915->meta.ProgramSize - 2; - i915->meta.emitted &= ~I915_UPLOAD_PROGRAM; -} - - - - - -/* Set up an arbitary piece of memory as a rectangular texture - * (including the front or back buffer). - */ -static GLboolean -meta_tex_rect_source(struct intel_context *intel, - struct _DriBufferObject *buffer, - GLuint offset, - GLuint pitch, GLuint height, GLenum format, GLenum type) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - GLuint unit = 0; - GLint numLevels = 1; - GLuint *state = i915->meta.Tex[0]; - GLuint textureFormat; - GLuint cpp; - - /* A full implementation of this would do the upload through - * glTexImage2d, and get all the conversion operations at that - * point. We are restricted, but still at least have access to the - * fragment program swizzle. - */ - switch (format) { - case GL_BGRA: - switch (type) { - case GL_UNSIGNED_INT_8_8_8_8_REV: - case GL_UNSIGNED_BYTE: - textureFormat = (MAPSURF_32BIT | MT_32BIT_ARGB8888); - cpp = 4; - break; - default: - return GL_FALSE; - } - break; - case GL_RGBA: - switch (type) { - case GL_UNSIGNED_INT_8_8_8_8_REV: - case GL_UNSIGNED_BYTE: - textureFormat = (MAPSURF_32BIT | MT_32BIT_ABGR8888); - cpp = 4; - break; - default: - return GL_FALSE; - } - break; - case GL_BGR: - switch (type) { - case GL_UNSIGNED_SHORT_5_6_5_REV: - textureFormat = (MAPSURF_16BIT | MT_16BIT_RGB565); - cpp = 2; - break; - default: - return GL_FALSE; - } - break; - case GL_RGB: - switch (type) { - case GL_UNSIGNED_SHORT_5_6_5: - textureFormat = (MAPSURF_16BIT | MT_16BIT_RGB565); - cpp = 2; - break; - default: - return GL_FALSE; - } - break; - - default: - return GL_FALSE; - } - - - if ((pitch * cpp) & 3) { - _mesa_printf("%s: texture is not dword pitch\n", __FUNCTION__); - return GL_FALSE; - } - -/* intel_region_release(&i915->meta.tex_region[0]); */ -/* intel_region_reference(&i915->meta.tex_region[0], region); */ - i915->meta.tex_buffer[0] = buffer; - i915->meta.tex_offset[0] = offset; - - state[I915_TEXREG_MS3] = (((height - 1) << MS3_HEIGHT_SHIFT) | - ((pitch - 1) << MS3_WIDTH_SHIFT) | - textureFormat | MS3_USE_FENCE_REGS); - - state[I915_TEXREG_MS4] = (((((pitch * cpp) / 4) - 1) << MS4_PITCH_SHIFT) | - MS4_CUBE_FACE_ENA_MASK | - ((((numLevels - 1) * 4)) << MS4_MAX_LOD_SHIFT)); - - state[I915_TEXREG_SS2] = ((FILTER_NEAREST << SS2_MIN_FILTER_SHIFT) | - (MIPFILTER_NONE << SS2_MIP_FILTER_SHIFT) | - (FILTER_NEAREST << SS2_MAG_FILTER_SHIFT)); - - state[I915_TEXREG_SS3] = ((TEXCOORDMODE_WRAP << SS3_TCX_ADDR_MODE_SHIFT) | - (TEXCOORDMODE_WRAP << SS3_TCY_ADDR_MODE_SHIFT) | - (TEXCOORDMODE_WRAP << SS3_TCZ_ADDR_MODE_SHIFT) | - (unit << SS3_TEXTUREMAP_INDEX_SHIFT)); - - state[I915_TEXREG_SS4] = 0; - - i915->meta.emitted &= ~I915_UPLOAD_TEX(0); - return GL_TRUE; -} - - -/** - * Set the color and depth drawing region for meta ops. - */ -static void -meta_draw_region(struct intel_context *intel, - struct intel_region *color_region, - struct intel_region *depth_region) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - i915_state_draw_region(intel, &i915->meta, color_region, depth_region); -} - - -static void -set_vertex_format(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - - i915->meta.Ctx[I915_CTXREG_LIS2] = - (S2_TEXCOORD_FMT(0, TEXCOORDFMT_2D) | - S2_TEXCOORD_FMT(1, TEXCOORDFMT_NOT_PRESENT) | - S2_TEXCOORD_FMT(2, TEXCOORDFMT_NOT_PRESENT) | - S2_TEXCOORD_FMT(3, TEXCOORDFMT_NOT_PRESENT) | - S2_TEXCOORD_FMT(4, TEXCOORDFMT_NOT_PRESENT) | - S2_TEXCOORD_FMT(5, TEXCOORDFMT_NOT_PRESENT) | - S2_TEXCOORD_FMT(6, TEXCOORDFMT_NOT_PRESENT) | - S2_TEXCOORD_FMT(7, TEXCOORDFMT_NOT_PRESENT)); - - i915->meta.Ctx[I915_CTXREG_LIS4] &= ~S4_VFMT_MASK; - - i915->meta.Ctx[I915_CTXREG_LIS4] |= (S4_VFMT_COLOR | S4_VFMT_XYZ); - - i915->meta.emitted &= ~I915_UPLOAD_CTX; -} - - - -/* Operations where the 3D engine is decoupled temporarily from the - * current GL state and used for other purposes than simply rendering - * incoming triangles. - */ -static void -install_meta_state(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - memcpy(&i915->meta, &i915->initial, sizeof(i915->meta)); - i915->meta.active = ACTIVE; - i915->meta.emitted = 0; - - SET_STATE(i915, meta); - set_vertex_format(intel); - meta_no_texture(intel); -} - -static void -leave_meta_state(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - intel_region_release(&i915->meta.draw_region); - intel_region_release(&i915->meta.depth_region); -/* intel_region_release(&i915->meta.tex_region[0]); */ - SET_STATE(i915, state); -} - - - -void -i915InitMetaFuncs(struct i915_context *i915) -{ - i915->intel.vtbl.install_meta_state = install_meta_state; - i915->intel.vtbl.leave_meta_state = leave_meta_state; - i915->intel.vtbl.meta_no_depth_write = meta_no_depth_write; - i915->intel.vtbl.meta_no_stencil_write = meta_no_stencil_write; - i915->intel.vtbl.meta_stencil_replace = meta_stencil_replace; - i915->intel.vtbl.meta_depth_replace = meta_depth_replace; - i915->intel.vtbl.meta_color_mask = meta_color_mask; - i915->intel.vtbl.meta_no_texture = meta_no_texture; - i915->intel.vtbl.meta_texture_blend_replace = meta_texture_blend_replace; - i915->intel.vtbl.meta_tex_rect_source = meta_tex_rect_source; - i915->intel.vtbl.meta_draw_region = meta_draw_region; - i915->intel.vtbl.meta_import_pixel_state = meta_import_pixel_state; -} diff --git a/src/mesa/drivers/dri/i915tex/i915_program.c b/src/mesa/drivers/dri/i915tex/i915_program.c deleted file mode 100644 index ed45e4f806..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_program.c +++ /dev/null @@ -1,515 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include - -#include "glheader.h" -#include "macros.h" -#include "enums.h" - -#include "tnl/t_context.h" -#include "intel_batchbuffer.h" - -#include "i915_reg.h" -#include "i915_context.h" -#include "i915_program.h" - - -#define A0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) -#define D0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) -#define T0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) -#define A0_SRC0( reg ) (((reg)&UREG_MASK)>>UREG_A0_SRC0_SHIFT_LEFT) -#define A1_SRC0( reg ) (((reg)&UREG_MASK)<>UREG_A1_SRC1_SHIFT_LEFT) -#define A2_SRC1( reg ) (((reg)&UREG_MASK)<>UREG_A2_SRC2_SHIFT_LEFT) - -/* These are special, and don't have swizzle/negate bits. - */ -#define T0_SAMPLER( reg ) (GET_UREG_NR(reg)<temp_flag); - if (!bit) { - fprintf(stderr, "%s: out of temporaries\n", __FILE__); - exit(1); - } - - p->temp_flag |= 1 << (bit - 1); - return UREG(REG_TYPE_R, (bit - 1)); -} - - -GLuint -i915_get_utemp(struct i915_fragment_program * p) -{ - int bit = ffs(~p->utemp_flag); - if (!bit) { - fprintf(stderr, "%s: out of temporaries\n", __FILE__); - exit(1); - } - - p->utemp_flag |= 1 << (bit - 1); - return UREG(REG_TYPE_U, (bit - 1)); -} - -void -i915_release_utemps(struct i915_fragment_program *p) -{ - p->utemp_flag = ~0x7; -} - - -GLuint -i915_emit_decl(struct i915_fragment_program *p, - GLuint type, GLuint nr, GLuint d0_flags) -{ - GLuint reg = UREG(type, nr); - - if (type == REG_TYPE_T) { - if (p->decl_t & (1 << nr)) - return reg; - - p->decl_t |= (1 << nr); - } - else if (type == REG_TYPE_S) { - if (p->decl_s & (1 << nr)) - return reg; - - p->decl_s |= (1 << nr); - } - else - return reg; - - *(p->decl++) = (D0_DCL | D0_DEST(reg) | d0_flags); - *(p->decl++) = D1_MBZ; - *(p->decl++) = D2_MBZ; - - p->nr_decl_insn++; - return reg; -} - -GLuint -i915_emit_arith(struct i915_fragment_program * p, - GLuint op, - GLuint dest, - GLuint mask, - GLuint saturate, GLuint src0, GLuint src1, GLuint src2) -{ - GLuint c[3]; - GLuint nr_const = 0; - - assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); - dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest)); - assert(dest); - - if (GET_UREG_TYPE(src0) == REG_TYPE_CONST) - c[nr_const++] = 0; - if (GET_UREG_TYPE(src1) == REG_TYPE_CONST) - c[nr_const++] = 1; - if (GET_UREG_TYPE(src2) == REG_TYPE_CONST) - c[nr_const++] = 2; - - /* Recursively call this function to MOV additional const values - * into temporary registers. Use utemp registers for this - - * currently shouldn't be possible to run out, but keep an eye on - * this. - */ - if (nr_const > 1) { - GLuint s[3], first, i, old_utemp_flag; - - s[0] = src0; - s[1] = src1; - s[2] = src2; - old_utemp_flag = p->utemp_flag; - - first = GET_UREG_NR(s[c[0]]); - for (i = 1; i < nr_const; i++) { - if (GET_UREG_NR(s[c[i]]) != first) { - GLuint tmp = i915_get_utemp(p); - - i915_emit_arith(p, A0_MOV, tmp, A0_DEST_CHANNEL_ALL, 0, - s[c[i]], 0, 0); - s[c[i]] = tmp; - } - } - - src0 = s[0]; - src1 = s[1]; - src2 = s[2]; - p->utemp_flag = old_utemp_flag; /* restore */ - } - - *(p->csr++) = (op | A0_DEST(dest) | mask | saturate | A0_SRC0(src0)); - *(p->csr++) = (A1_SRC0(src0) | A1_SRC1(src1)); - *(p->csr++) = (A2_SRC1(src1) | A2_SRC2(src2)); - - p->nr_alu_insn++; - return dest; -} - -GLuint i915_emit_texld( struct i915_fragment_program *p, - GLuint dest, - GLuint destmask, - GLuint sampler, - GLuint coord, - GLuint op ) -{ - if (coord != UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord))) { - /* No real way to work around this in the general case - need to - * allocate and declare a new temporary register (a utemp won't - * do). Will fallback for now. - */ - i915_program_error(p, "Can't (yet) swizzle TEX arguments"); - return 0; - } - - /* Don't worry about saturate as we only support - */ - if (destmask != A0_DEST_CHANNEL_ALL) { - GLuint tmp = i915_get_utemp(p); - i915_emit_texld( p, tmp, A0_DEST_CHANNEL_ALL, sampler, coord, op ); - i915_emit_arith( p, A0_MOV, dest, destmask, 0, tmp, 0, 0 ); - return dest; - } - else { - assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); - assert(dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest))); - - if (GET_UREG_TYPE(coord) != REG_TYPE_T) { - p->nr_tex_indirect++; - } - - *(p->csr++) = (op | - T0_DEST( dest ) | - T0_SAMPLER( sampler )); - - *(p->csr++) = T1_ADDRESS_REG( coord ); - *(p->csr++) = T2_MBZ; - - p->nr_tex_insn++; - return dest; - } -} - - -GLuint -i915_emit_const1f(struct i915_fragment_program * p, GLfloat c0) -{ - GLint reg, idx; - - if (c0 == 0.0) - return swizzle(UREG(REG_TYPE_R, 0), ZERO, ZERO, ZERO, ZERO); - if (c0 == 1.0) - return swizzle(UREG(REG_TYPE_R, 0), ONE, ONE, ONE, ONE); - - for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == I915_CONSTFLAG_PARAM) - continue; - for (idx = 0; idx < 4; idx++) { - if (!(p->constant_flags[reg] & (1 << idx)) || - p->constant[reg][idx] == c0) { - p->constant[reg][idx] = c0; - p->constant_flags[reg] |= 1 << idx; - if (reg + 1 > p->nr_constants) - p->nr_constants = reg + 1; - return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE); - } - } - } - - fprintf(stderr, "%s: out of constants\n", __FUNCTION__); - p->error = 1; - return 0; -} - -GLuint -i915_emit_const2f(struct i915_fragment_program * p, GLfloat c0, GLfloat c1) -{ - GLint reg, idx; - - if (c0 == 0.0) - return swizzle(i915_emit_const1f(p, c1), ZERO, X, Z, W); - if (c0 == 1.0) - return swizzle(i915_emit_const1f(p, c1), ONE, X, Z, W); - - if (c1 == 0.0) - return swizzle(i915_emit_const1f(p, c0), X, ZERO, Z, W); - if (c1 == 1.0) - return swizzle(i915_emit_const1f(p, c0), X, ONE, Z, W); - - for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == 0xf || - p->constant_flags[reg] == I915_CONSTFLAG_PARAM) - continue; - for (idx = 0; idx < 3; idx++) { - if (!(p->constant_flags[reg] & (3 << idx))) { - p->constant[reg][idx] = c0; - p->constant[reg][idx + 1] = c1; - p->constant_flags[reg] |= 3 << idx; - if (reg + 1 > p->nr_constants) - p->nr_constants = reg + 1; - return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, - ONE); - } - } - } - - fprintf(stderr, "%s: out of constants\n", __FUNCTION__); - p->error = 1; - return 0; -} - - - -GLuint -i915_emit_const4f(struct i915_fragment_program * p, - GLfloat c0, GLfloat c1, GLfloat c2, GLfloat c3) -{ - GLint reg; - - for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == 0xf && - p->constant[reg][0] == c0 && - p->constant[reg][1] == c1 && - p->constant[reg][2] == c2 && p->constant[reg][3] == c3) { - return UREG(REG_TYPE_CONST, reg); - } - else if (p->constant_flags[reg] == 0) { - p->constant[reg][0] = c0; - p->constant[reg][1] = c1; - p->constant[reg][2] = c2; - p->constant[reg][3] = c3; - p->constant_flags[reg] = 0xf; - if (reg + 1 > p->nr_constants) - p->nr_constants = reg + 1; - return UREG(REG_TYPE_CONST, reg); - } - } - - fprintf(stderr, "%s: out of constants\n", __FUNCTION__); - p->error = 1; - return 0; -} - - -GLuint -i915_emit_const4fv(struct i915_fragment_program * p, const GLfloat * c) -{ - return i915_emit_const4f(p, c[0], c[1], c[2], c[3]); -} - - -GLuint -i915_emit_param4fv(struct i915_fragment_program * p, const GLfloat * values) -{ - GLint reg, i; - - for (i = 0; i < p->nr_params; i++) { - if (p->param[i].values == values) - return UREG(REG_TYPE_CONST, p->param[i].reg); - } - - - for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == 0) { - p->constant_flags[reg] = I915_CONSTFLAG_PARAM; - i = p->nr_params++; - - p->param[i].values = values; - p->param[i].reg = reg; - p->params_uptodate = 0; - - if (reg + 1 > p->nr_constants) - p->nr_constants = reg + 1; - return UREG(REG_TYPE_CONST, reg); - } - } - - fprintf(stderr, "%s: out of constants\n", __FUNCTION__); - p->error = 1; - return 0; -} - - - -void -i915_program_error(struct i915_fragment_program *p, const char *msg) -{ - _mesa_problem(NULL, "i915_program_error: %s", msg); - p->error = 1; -} - - -void -i915_init_program(struct i915_context *i915, struct i915_fragment_program *p) -{ - GLcontext *ctx = &i915->intel.ctx; - TNLcontext *tnl = TNL_CONTEXT(ctx); - - p->translated = 0; - p->params_uptodate = 0; - p->on_hardware = 0; - p->error = 0; - - p->nr_tex_indirect = 1; /* correct? */ - p->nr_tex_insn = 0; - p->nr_alu_insn = 0; - p->nr_decl_insn = 0; - - p->ctx = ctx; - memset(p->constant_flags, 0, sizeof(p->constant_flags)); - - p->nr_constants = 0; - p->csr = p->program; - p->decl = p->declarations; - p->decl_s = 0; - p->decl_t = 0; - p->temp_flag = 0xffff000; - p->utemp_flag = ~0x7; - p->wpos_tex = -1; - p->depth_written = 0; - p->nr_params = 0; - - p->src_texture = UREG_BAD; - p->src_previous = UREG(REG_TYPE_T, T_DIFFUSE); - p->last_tex_stage = 0; - p->VB = &tnl->vb; - - *(p->decl++) = _3DSTATE_PIXEL_SHADER_PROGRAM; -} - - -void -i915_fini_program(struct i915_fragment_program *p) -{ - GLuint program_size = p->csr - p->program; - GLuint decl_size = p->decl - p->declarations; - - if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT) - i915_program_error(p, "Exceeded max nr indirect texture lookups"); - - if (p->nr_tex_insn > I915_MAX_TEX_INSN) - i915_program_error(p, "Exceeded max TEX instructions"); - - if (p->nr_alu_insn > I915_MAX_ALU_INSN) - i915_program_error(p, "Exceeded max ALU instructions"); - - if (p->nr_decl_insn > I915_MAX_DECL_INSN) - i915_program_error(p, "Exceeded max DECL instructions"); - - if (p->error) { - p->FragProg.Base.NumNativeInstructions = 0; - p->FragProg.Base.NumNativeAluInstructions = 0; - p->FragProg.Base.NumNativeTexInstructions = 0; - p->FragProg.Base.NumNativeTexIndirections = 0; - } - else { - p->FragProg.Base.NumNativeInstructions = (p->nr_alu_insn + - p->nr_tex_insn + - p->nr_decl_insn); - p->FragProg.Base.NumNativeAluInstructions = p->nr_alu_insn; - p->FragProg.Base.NumNativeTexInstructions = p->nr_tex_insn; - p->FragProg.Base.NumNativeTexIndirections = p->nr_tex_indirect; - } - - p->declarations[0] |= program_size + decl_size - 2; -} - -void -i915_upload_program(struct i915_context *i915, - struct i915_fragment_program *p) -{ - GLuint program_size = p->csr - p->program; - GLuint decl_size = p->decl - p->declarations; - - FALLBACK(&i915->intel, I915_FALLBACK_PROGRAM, p->error); - - /* Could just go straight to the batchbuffer from here: - */ - if (i915->state.ProgramSize != (program_size + decl_size) || - memcmp(i915->state.Program + decl_size, p->program, - program_size * sizeof(int)) != 0) { - I915_STATECHANGE(i915, I915_UPLOAD_PROGRAM); - memcpy(i915->state.Program, p->declarations, decl_size * sizeof(int)); - memcpy(i915->state.Program + decl_size, p->program, - program_size * sizeof(int)); - i915->state.ProgramSize = decl_size + program_size; - } - - /* Always seemed to get a failure if I used memcmp() to - * shortcircuit this state upload. Needs further investigation? - */ - if (p->nr_constants) { - GLuint nr = p->nr_constants; - - I915_ACTIVESTATE(i915, I915_UPLOAD_CONSTANTS, 1); - I915_STATECHANGE(i915, I915_UPLOAD_CONSTANTS); - - i915->state.Constant[0] = _3DSTATE_PIXEL_SHADER_CONSTANTS | ((nr) * 4); - i915->state.Constant[1] = (1 << (nr - 1)) | ((1 << (nr - 1)) - 1); - - memcpy(&i915->state.Constant[2], p->constant, 4 * sizeof(int) * (nr)); - i915->state.ConstantSize = 2 + (nr) * 4; - - if (0) { - GLuint i; - for (i = 0; i < nr; i++) { - fprintf(stderr, "const[%d]: %f %f %f %f\n", i, - p->constant[i][0], - p->constant[i][1], p->constant[i][2], p->constant[i][3]); - } - } - } - else { - I915_ACTIVESTATE(i915, I915_UPLOAD_CONSTANTS, 0); - } - - p->on_hardware = 1; -} diff --git a/src/mesa/drivers/dri/i915tex/i915_program.h b/src/mesa/drivers/dri/i915tex/i915_program.h deleted file mode 100644 index 3c12b34f16..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_program.h +++ /dev/null @@ -1,160 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef I915_PROGRAM_H -#define I915_PROGRAM_H - -#include "i915_context.h" -#include "i915_reg.h" - - - -/* Having zero and one in here makes the definition of swizzle a lot - * easier. - */ -#define UREG_TYPE_SHIFT 29 -#define UREG_NR_SHIFT 24 -#define UREG_CHANNEL_X_NEGATE_SHIFT 23 -#define UREG_CHANNEL_X_SHIFT 20 -#define UREG_CHANNEL_Y_NEGATE_SHIFT 19 -#define UREG_CHANNEL_Y_SHIFT 16 -#define UREG_CHANNEL_Z_NEGATE_SHIFT 15 -#define UREG_CHANNEL_Z_SHIFT 12 -#define UREG_CHANNEL_W_NEGATE_SHIFT 11 -#define UREG_CHANNEL_W_SHIFT 8 -#define UREG_CHANNEL_ZERO_NEGATE_MBZ 5 -#define UREG_CHANNEL_ZERO_SHIFT 4 -#define UREG_CHANNEL_ONE_NEGATE_MBZ 1 -#define UREG_CHANNEL_ONE_SHIFT 0 - -#define UREG_BAD 0xffffffff /* not a valid ureg */ - -#define X SRC_X -#define Y SRC_Y -#define Z SRC_Z -#define W SRC_W -#define ZERO SRC_ZERO -#define ONE SRC_ONE - -/* Construct a ureg: - */ -#define UREG( type, nr ) (((type)<< UREG_TYPE_SHIFT) | \ - ((nr) << UREG_NR_SHIFT) | \ - (X << UREG_CHANNEL_X_SHIFT) | \ - (Y << UREG_CHANNEL_Y_SHIFT) | \ - (Z << UREG_CHANNEL_Z_SHIFT) | \ - (W << UREG_CHANNEL_W_SHIFT) | \ - (ZERO << UREG_CHANNEL_ZERO_SHIFT) | \ - (ONE << UREG_CHANNEL_ONE_SHIFT)) - -#define GET_CHANNEL_SRC( reg, channel ) ((reg<<(channel*4)) & (0xf<<20)) -#define CHANNEL_SRC( src, channel ) (src>>(channel*4)) - -#define GET_UREG_TYPE(reg) (((reg)>>UREG_TYPE_SHIFT)®_TYPE_MASK) -#define GET_UREG_NR(reg) (((reg)>>UREG_NR_SHIFT)®_NR_MASK) - - - -#define UREG_XYZW_CHANNEL_MASK 0x00ffff00 - -/* One neat thing about the UREG representation: - */ -static INLINE int -swizzle(int reg, int x, int y, int z, int w) -{ - return ((reg & ~UREG_XYZW_CHANNEL_MASK) | - CHANNEL_SRC(GET_CHANNEL_SRC(reg, x), 0) | - CHANNEL_SRC(GET_CHANNEL_SRC(reg, y), 1) | - CHANNEL_SRC(GET_CHANNEL_SRC(reg, z), 2) | - CHANNEL_SRC(GET_CHANNEL_SRC(reg, w), 3)); -} - -/* Another neat thing about the UREG representation: - */ -static INLINE int -negate(int reg, int x, int y, int z, int w) -{ - return reg ^ (((x & 1) << UREG_CHANNEL_X_NEGATE_SHIFT) | - ((y & 1) << UREG_CHANNEL_Y_NEGATE_SHIFT) | - ((z & 1) << UREG_CHANNEL_Z_NEGATE_SHIFT) | - ((w & 1) << UREG_CHANNEL_W_NEGATE_SHIFT)); -} - - -extern GLuint i915_get_temp(struct i915_fragment_program *p); -extern GLuint i915_get_utemp(struct i915_fragment_program *p); -extern void i915_release_utemps(struct i915_fragment_program *p); - - -extern GLuint i915_emit_texld(struct i915_fragment_program *p, - GLuint dest, - GLuint destmask, - GLuint sampler, GLuint coord, GLuint op); - -extern GLuint i915_emit_arith(struct i915_fragment_program *p, - GLuint op, - GLuint dest, - GLuint mask, - GLuint saturate, - GLuint src0, GLuint src1, GLuint src2); - -extern GLuint i915_emit_decl(struct i915_fragment_program *p, - GLuint type, GLuint nr, GLuint d0_flags); - - -extern GLuint i915_emit_const1f(struct i915_fragment_program *p, GLfloat c0); - -extern GLuint i915_emit_const2f(struct i915_fragment_program *p, - GLfloat c0, GLfloat c1); - -extern GLuint i915_emit_const4fv(struct i915_fragment_program *p, - const GLfloat * c); - -extern GLuint i915_emit_const4f(struct i915_fragment_program *p, - GLfloat c0, GLfloat c1, - GLfloat c2, GLfloat c3); - - -extern GLuint i915_emit_param4fv(struct i915_fragment_program *p, - const GLfloat * values); - -extern void i915_program_error(struct i915_fragment_program *p, - const char *msg); - -extern void i915_init_program(struct i915_context *i915, - struct i915_fragment_program *p); - -extern void i915_upload_program(struct i915_context *i915, - struct i915_fragment_program *p); - -extern void i915_fini_program(struct i915_fragment_program *p); - - - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/i915_reg.h b/src/mesa/drivers/dri/i915tex/i915_reg.h deleted file mode 100644 index 34c6821405..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_reg.h +++ /dev/null @@ -1,841 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef _I915_REG_H_ -#define _I915_REG_H_ - - -#include "intel_reg.h" - -#define I915_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value) - -#define CMD_3D (0x3<<29) - -#define PRIM3D_INLINE (CMD_3D | (0x1f<<24)) -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_CLEAR_RECT (0xa<<18) -#define PRIM3D_ZONE_INIT (0xd<<18) -#define PRIM3D_MASK (0x1f<<18) - -/* p137 */ -#define _3DSTATE_AA_CMD (CMD_3D | (0x06<<24)) -#define AA_LINE_ECAAR_WIDTH_ENABLE (1<<16) -#define AA_LINE_ECAAR_WIDTH_0_5 0 -#define AA_LINE_ECAAR_WIDTH_1_0 (1<<14) -#define AA_LINE_ECAAR_WIDTH_2_0 (2<<14) -#define AA_LINE_ECAAR_WIDTH_4_0 (3<<14) -#define AA_LINE_REGION_WIDTH_ENABLE (1<<8) -#define AA_LINE_REGION_WIDTH_0_5 0 -#define AA_LINE_REGION_WIDTH_1_0 (1<<6) -#define AA_LINE_REGION_WIDTH_2_0 (2<<6) -#define AA_LINE_REGION_WIDTH_4_0 (3<<6) - -/* 3DSTATE_BACKFACE_STENCIL_OPS, p138*/ -#define _3DSTATE_BACKFACE_STENCIL_OPS (CMD_3D | (0x8<<24)) -#define BFO_ENABLE_STENCIL_REF (1<<23) -#define BFO_STENCIL_REF_SHIFT 15 -#define BFO_STENCIL_REF_MASK (0xff<<15) -#define BFO_ENABLE_STENCIL_FUNCS (1<<14) -#define BFO_STENCIL_TEST_SHIFT 11 -#define BFO_STENCIL_TEST_MASK (0x7<<11) -#define BFO_STENCIL_FAIL_SHIFT 8 -#define BFO_STENCIL_FAIL_MASK (0x7<<8) -#define BFO_STENCIL_PASS_Z_FAIL_SHIFT 5 -#define BFO_STENCIL_PASS_Z_FAIL_MASK (0x7<<5) -#define BFO_STENCIL_PASS_Z_PASS_SHIFT 2 -#define BFO_STENCIL_PASS_Z_PASS_MASK (0x7<<2) -#define BFO_ENABLE_STENCIL_TWO_SIDE (1<<1) -#define BFO_STENCIL_TWO_SIDE (1<<0) - - -/* 3DSTATE_BACKFACE_STENCIL_MASKS, p140 */ -#define _3DSTATE_BACKFACE_STENCIL_MASKS (CMD_3D | (0x9<<24)) -#define BFM_ENABLE_STENCIL_TEST_MASK (1<<17) -#define BFM_ENABLE_STENCIL_WRITE_MASK (1<<16) -#define BFM_STENCIL_TEST_MASK_SHIFT 8 -#define BFM_STENCIL_TEST_MASK_MASK (0xff<<8) -#define BFM_STENCIL_WRITE_MASK_SHIFT 0 -#define BFM_STENCIL_WRITE_MASK_MASK (0xff<<0) - - - -/* 3DSTATE_BIN_CONTROL p141 */ - -/* p143 */ -#define _3DSTATE_BUF_INFO_CMD (CMD_3D | (0x1d<<24) | (0x8e<<16) | 1) -/* Dword 1 */ -#define BUF_3D_ID_COLOR_BACK (0x3<<24) -#define BUF_3D_ID_DEPTH (0x7<<24) -#define BUF_3D_USE_FENCE (1<<23) -#define BUF_3D_TILED_SURFACE (1<<22) -#define BUF_3D_TILE_WALK_X 0 -#define BUF_3D_TILE_WALK_Y (1<<21) -#define BUF_3D_PITCH(x) (((x)/4)<<2) -/* Dword 2 */ -#define BUF_3D_ADDR(x) ((x) & ~0x3) - - -/* 3DSTATE_CHROMA_KEY */ - -/* 3DSTATE_CLEAR_PARAMETERS, p150 */ - -/* 3DSTATE_CONSTANT_BLEND_COLOR, p153 */ -#define _3DSTATE_CONST_BLEND_COLOR_CMD (CMD_3D | (0x1d<<24) | (0x88<<16)) - - - -/* 3DSTATE_COORD_SET_BINDINGS, p154 */ -#define _3DSTATE_COORD_SET_BINDINGS (CMD_3D | (0x16<<24)) -#define CSB_TCB(iunit, eunit) ((eunit)<<(iunit*3)) - -/* p156 */ -#define _3DSTATE_DFLT_DIFFUSE_CMD (CMD_3D | (0x1d<<24) | (0x99<<16)) - -/* p157 */ -#define _3DSTATE_DFLT_SPEC_CMD (CMD_3D | (0x1d<<24) | (0x9a<<16)) - -/* p158 */ -#define _3DSTATE_DFLT_Z_CMD (CMD_3D | (0x1d<<24) | (0x98<<16)) - - -/* 3DSTATE_DEPTH_OFFSET_SCALE, p159 */ -#define _3DSTATE_DEPTH_OFFSET_SCALE (CMD_3D | (0x1d<<24) | (0x97<<16)) -/* scale in dword 1 */ - - -/* 3DSTATE_DEPTH_SUBRECT_DISABLE, p160 */ -#define _3DSTATE_DEPTH_SUBRECT_DISABLE (CMD_3D | (0x1c<<24) | (0x11<<19) | 0x2) - -/* p161 */ -#define _3DSTATE_DST_BUF_VARS_CMD (CMD_3D | (0x1d<<24) | (0x85<<16)) -/* Dword 1 */ -#define TEX_DEFAULT_COLOR_OGL (0<<30) -#define TEX_DEFAULT_COLOR_D3D (1<<30) -#define ZR_EARLY_DEPTH (1<<29) -#define LOD_PRECLAMP_OGL (1<<28) -#define LOD_PRECLAMP_D3D (0<<28) -#define DITHER_FULL_ALWAYS (0<<26) -#define DITHER_FULL_ON_FB_BLEND (1<<26) -#define DITHER_CLAMPED_ALWAYS (2<<26) -#define LINEAR_GAMMA_BLEND_32BPP (1<<25) -#define DEBUG_DISABLE_ENH_DITHER (1<<24) -#define DSTORG_HORT_BIAS(x) ((x)<<20) -#define DSTORG_VERT_BIAS(x) ((x)<<16) -#define COLOR_4_2_2_CHNL_WRT_ALL 0 -#define COLOR_4_2_2_CHNL_WRT_Y (1<<12) -#define COLOR_4_2_2_CHNL_WRT_CR (2<<12) -#define COLOR_4_2_2_CHNL_WRT_CB (3<<12) -#define COLOR_4_2_2_CHNL_WRT_CRCB (4<<12) -#define COLR_BUF_8BIT 0 -#define COLR_BUF_RGB555 (1<<8) -#define COLR_BUF_RGB565 (2<<8) -#define COLR_BUF_ARGB8888 (3<<8) -#define DEPTH_FRMT_16_FIXED 0 -#define DEPTH_FRMT_16_FLOAT (1<<2) -#define DEPTH_FRMT_24_FIXED_8_OTHER (2<<2) -#define VERT_LINE_STRIDE_1 (1<<1) -#define VERT_LINE_STRIDE_0 (0<<1) -#define VERT_LINE_STRIDE_OFS_1 1 -#define VERT_LINE_STRIDE_OFS_0 0 - -/* p166 */ -#define _3DSTATE_DRAW_RECT_CMD (CMD_3D|(0x1d<<24)|(0x80<<16)|3) -/* Dword 1 */ -#define DRAW_RECT_DIS_DEPTH_OFS (1<<30) -#define DRAW_DITHER_OFS_X(x) ((x)<<26) -#define DRAW_DITHER_OFS_Y(x) ((x)<<24) -/* Dword 2 */ -#define DRAW_YMIN(x) ((x)<<16) -#define DRAW_XMIN(x) (x) -/* Dword 3 */ -#define DRAW_YMAX(x) ((x)<<16) -#define DRAW_XMAX(x) (x) -/* Dword 4 */ -#define DRAW_YORG(x) ((x)<<16) -#define DRAW_XORG(x) (x) - - -/* 3DSTATE_FILTER_COEFFICIENTS_4X4, p170 */ - -/* 3DSTATE_FILTER_COEFFICIENTS_6X5, p172 */ - - -/* _3DSTATE_FOG_COLOR, p173 */ -#define _3DSTATE_FOG_COLOR_CMD (CMD_3D|(0x15<<24)) -#define FOG_COLOR_RED(x) ((x)<<16) -#define FOG_COLOR_GREEN(x) ((x)<<8) -#define FOG_COLOR_BLUE(x) (x) - -/* _3DSTATE_FOG_MODE, p174 */ -#define _3DSTATE_FOG_MODE_CMD (CMD_3D|(0x1d<<24)|(0x89<<16)|2) -/* Dword 1 */ -#define FMC1_FOGFUNC_MODIFY_ENABLE (1<<31) -#define FMC1_FOGFUNC_VERTEX (0<<28) -#define FMC1_FOGFUNC_PIXEL_EXP (1<<28) -#define FMC1_FOGFUNC_PIXEL_EXP2 (2<<28) -#define FMC1_FOGFUNC_PIXEL_LINEAR (3<<28) -#define FMC1_FOGFUNC_MASK (3<<28) -#define FMC1_FOGINDEX_MODIFY_ENABLE (1<<27) -#define FMC1_FOGINDEX_Z (0<<25) -#define FMC1_FOGINDEX_W (1<<25) -#define FMC1_C1_C2_MODIFY_ENABLE (1<<24) -#define FMC1_DENSITY_MODIFY_ENABLE (1<<23) -#define FMC1_C1_ONE (1<<13) -#define FMC1_C1_MASK (0xffff<<4) -/* Dword 2 */ -#define FMC2_C2_ONE (1<<16) -/* Dword 3 */ -#define FMC3_D_ONE (1<<16) - - - -/* _3DSTATE_INDEPENDENT_ALPHA_BLEND, p177 */ -#define _3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD (CMD_3D|(0x0b<<24)) -#define IAB_MODIFY_ENABLE (1<<23) -#define IAB_ENABLE (1<<22) -#define IAB_MODIFY_FUNC (1<<21) -#define IAB_FUNC_SHIFT 16 -#define IAB_MODIFY_SRC_FACTOR (1<<11) -#define IAB_SRC_FACTOR_SHIFT 6 -#define IAB_SRC_FACTOR_MASK (BLENDFACT_MASK<<6) -#define IAB_MODIFY_DST_FACTOR (1<<5) -#define IAB_DST_FACTOR_SHIFT 0 -#define IAB_DST_FACTOR_MASK (BLENDFACT_MASK<<0) - - -#define BLENDFUNC_ADD 0x0 -#define BLENDFUNC_SUBTRACT 0x1 -#define BLENDFUNC_REVERSE_SUBTRACT 0x2 -#define BLENDFUNC_MIN 0x3 -#define BLENDFUNC_MAX 0x4 -#define BLENDFUNC_MASK 0x7 - -/* 3DSTATE_LOAD_INDIRECT, p180 */ - -#define _3DSTATE_LOAD_INDIRECT (CMD_3D|(0x1d<<24)|(0x7<<16)) -#define LI0_STATE_STATIC_INDIRECT (0x01<<8) -#define LI0_STATE_DYNAMIC_INDIRECT (0x02<<8) -#define LI0_STATE_SAMPLER (0x04<<8) -#define LI0_STATE_MAP (0x08<<8) -#define LI0_STATE_PROGRAM (0x10<<8) -#define LI0_STATE_CONSTANTS (0x20<<8) - -#define SIS0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define SIS0_FORCE_LOAD (1<<1) -#define SIS0_BUFFER_VALID (1<<0) -#define SIS1_BUFFER_LENGTH(x) ((x)&0xff) - -#define DIS0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define DIS0_BUFFER_RESET (1<<1) -#define DIS0_BUFFER_VALID (1<<0) - -#define SSB0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define SSB0_FORCE_LOAD (1<<1) -#define SSB0_BUFFER_VALID (1<<0) -#define SSB1_BUFFER_LENGTH(x) ((x)&0xff) - -#define MSB0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define MSB0_FORCE_LOAD (1<<1) -#define MSB0_BUFFER_VALID (1<<0) -#define MSB1_BUFFER_LENGTH(x) ((x)&0xff) - -#define PSP0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define PSP0_FORCE_LOAD (1<<1) -#define PSP0_BUFFER_VALID (1<<0) -#define PSP1_BUFFER_LENGTH(x) ((x)&0xff) - -#define PSC0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define PSC0_FORCE_LOAD (1<<1) -#define PSC0_BUFFER_VALID (1<<0) -#define PSC1_BUFFER_LENGTH(x) ((x)&0xff) - - - - - -/* _3DSTATE_RASTERIZATION_RULES */ -#define _3DSTATE_RASTER_RULES_CMD (CMD_3D|(0x07<<24)) -#define ENABLE_POINT_RASTER_RULE (1<<15) -#define OGL_POINT_RASTER_RULE (1<<13) -#define ENABLE_TEXKILL_3D_4D (1<<10) -#define TEXKILL_3D (0<<9) -#define TEXKILL_4D (1<<9) -#define ENABLE_LINE_STRIP_PROVOKE_VRTX (1<<8) -#define ENABLE_TRI_FAN_PROVOKE_VRTX (1<<5) -#define LINE_STRIP_PROVOKE_VRTX(x) ((x)<<6) -#define TRI_FAN_PROVOKE_VRTX(x) ((x)<<3) - -/* _3DSTATE_SCISSOR_ENABLE, p256 */ -#define _3DSTATE_SCISSOR_ENABLE_CMD (CMD_3D|(0x1c<<24)|(0x10<<19)) -#define ENABLE_SCISSOR_RECT ((1<<1) | 1) -#define DISABLE_SCISSOR_RECT (1<<1) - -/* _3DSTATE_SCISSOR_RECTANGLE_0, p257 */ -#define _3DSTATE_SCISSOR_RECT_0_CMD (CMD_3D|(0x1d<<24)|(0x81<<16)|1) -/* Dword 1 */ -#define SCISSOR_RECT_0_YMIN(x) ((x)<<16) -#define SCISSOR_RECT_0_XMIN(x) (x) -/* Dword 2 */ -#define SCISSOR_RECT_0_YMAX(x) ((x)<<16) -#define SCISSOR_RECT_0_XMAX(x) (x) - -/* p189 */ -#define _3DSTATE_LOAD_STATE_IMMEDIATE_1 ((0x3<<29)|(0x1d<<24)|(0x04<<16)) -#define I1_LOAD_S(n) (1<<(4+n)) - -#define S0_VB_OFFSET_MASK 0xffffffc -#define S0_AUTO_CACHE_INV_DISABLE (1<<0) - -#define S1_VERTEX_WIDTH_SHIFT 24 -#define S1_VERTEX_WIDTH_MASK (0x3f<<24) -#define S1_VERTEX_PITCH_SHIFT 16 -#define S1_VERTEX_PITCH_MASK (0x3f<<16) - -#define TEXCOORDFMT_2D 0x0 -#define TEXCOORDFMT_3D 0x1 -#define TEXCOORDFMT_4D 0x2 -#define TEXCOORDFMT_1D 0x3 -#define TEXCOORDFMT_2D_16 0x4 -#define TEXCOORDFMT_4D_16 0x5 -#define TEXCOORDFMT_NOT_PRESENT 0xf -#define S2_TEXCOORD_FMT0_MASK 0xf -#define S2_TEXCOORD_FMT1_SHIFT 4 -#define S2_TEXCOORD_FMT(unit, type) ((type)<<(unit*4)) -#define S2_TEXCOORD_NONE (~0) - -/* S3 not interesting */ - -#define S4_POINT_WIDTH_SHIFT 23 -#define S4_POINT_WIDTH_MASK (0x1ff<<23) -#define S4_LINE_WIDTH_SHIFT 19 -#define S4_LINE_WIDTH_ONE (0x2<<19) -#define S4_LINE_WIDTH_MASK (0xf<<19) -#define S4_FLATSHADE_ALPHA (1<<18) -#define S4_FLATSHADE_FOG (1<<17) -#define S4_FLATSHADE_SPECULAR (1<<16) -#define S4_FLATSHADE_COLOR (1<<15) -#define S4_CULLMODE_BOTH (0<<13) -#define S4_CULLMODE_NONE (1<<13) -#define S4_CULLMODE_CW (2<<13) -#define S4_CULLMODE_CCW (3<<13) -#define S4_CULLMODE_MASK (3<<13) -#define S4_VFMT_POINT_WIDTH (1<<12) -#define S4_VFMT_SPEC_FOG (1<<11) -#define S4_VFMT_COLOR (1<<10) -#define S4_VFMT_DEPTH_OFFSET (1<<9) -#define S4_VFMT_XYZ (1<<6) -#define S4_VFMT_XYZW (2<<6) -#define S4_VFMT_XY (3<<6) -#define S4_VFMT_XYW (4<<6) -#define S4_VFMT_XYZW_MASK (7<<6) -#define S4_FORCE_DEFAULT_DIFFUSE (1<<5) -#define S4_FORCE_DEFAULT_SPECULAR (1<<4) -#define S4_LOCAL_DEPTH_OFFSET_ENABLE (1<<3) -#define S4_VFMT_FOG_PARAM (1<<2) -#define S4_SPRITE_POINT_ENABLE (1<<1) -#define S4_LINE_ANTIALIAS_ENABLE (1<<0) - -#define S4_VFMT_MASK (S4_VFMT_POINT_WIDTH | \ - S4_VFMT_SPEC_FOG | \ - S4_VFMT_COLOR | \ - S4_VFMT_DEPTH_OFFSET | \ - S4_VFMT_XYZW_MASK | \ - S4_VFMT_FOG_PARAM) - - -#define S5_WRITEDISABLE_ALPHA (1<<31) -#define S5_WRITEDISABLE_RED (1<<30) -#define S5_WRITEDISABLE_GREEN (1<<29) -#define S5_WRITEDISABLE_BLUE (1<<28) -#define S5_WRITEDISABLE_MASK (0xf<<28) -#define S5_FORCE_DEFAULT_POINT_SIZE (1<<27) -#define S5_LAST_PIXEL_ENABLE (1<<26) -#define S5_GLOBAL_DEPTH_OFFSET_ENABLE (1<<25) -#define S5_FOG_ENABLE (1<<24) -#define S5_STENCIL_REF_SHIFT 16 -#define S5_STENCIL_REF_MASK (0xff<<16) -#define S5_STENCIL_TEST_FUNC_SHIFT 13 -#define S5_STENCIL_TEST_FUNC_MASK (0x7<<13) -#define S5_STENCIL_FAIL_SHIFT 10 -#define S5_STENCIL_FAIL_MASK (0x7<<10) -#define S5_STENCIL_PASS_Z_FAIL_SHIFT 7 -#define S5_STENCIL_PASS_Z_FAIL_MASK (0x7<<7) -#define S5_STENCIL_PASS_Z_PASS_SHIFT 4 -#define S5_STENCIL_PASS_Z_PASS_MASK (0x7<<4) -#define S5_STENCIL_WRITE_ENABLE (1<<3) -#define S5_STENCIL_TEST_ENABLE (1<<2) -#define S5_COLOR_DITHER_ENABLE (1<<1) -#define S5_LOGICOP_ENABLE (1<<0) - - -#define S6_ALPHA_TEST_ENABLE (1<<31) -#define S6_ALPHA_TEST_FUNC_SHIFT 28 -#define S6_ALPHA_TEST_FUNC_MASK (0x7<<28) -#define S6_ALPHA_REF_SHIFT 20 -#define S6_ALPHA_REF_MASK (0xff<<20) -#define S6_DEPTH_TEST_ENABLE (1<<19) -#define S6_DEPTH_TEST_FUNC_SHIFT 16 -#define S6_DEPTH_TEST_FUNC_MASK (0x7<<16) -#define S6_CBUF_BLEND_ENABLE (1<<15) -#define S6_CBUF_BLEND_FUNC_SHIFT 12 -#define S6_CBUF_BLEND_FUNC_MASK (0x7<<12) -#define S6_CBUF_SRC_BLEND_FACT_SHIFT 8 -#define S6_CBUF_SRC_BLEND_FACT_MASK (0xf<<8) -#define S6_CBUF_DST_BLEND_FACT_SHIFT 4 -#define S6_CBUF_DST_BLEND_FACT_MASK (0xf<<4) -#define S6_DEPTH_WRITE_ENABLE (1<<3) -#define S6_COLOR_WRITE_ENABLE (1<<2) -#define S6_TRISTRIP_PV_SHIFT 0 -#define S6_TRISTRIP_PV_MASK (0x3<<0) - -#define S7_DEPTH_OFFSET_CONST_MASK ~0 - -/* 3DSTATE_MAP_DEINTERLACER_PARAMETERS */ - -/* 3DSTATE_MAP_PALETTE_LOAD_32, p206 */ -#define _3DSTATE_MAP_PALETTE_LOAD_32 (CMD_3D|(0x1d<<24)|(0x8f<<16)) -/* subsequent dwords up to length (max 16) are ARGB8888 color values */ - -/* _3DSTATE_MODES_4, p218 */ -#define _3DSTATE_MODES_4_CMD (CMD_3D|(0x0d<<24)) -#define ENABLE_LOGIC_OP_FUNC (1<<23) -#define LOGIC_OP_FUNC(x) ((x)<<18) -#define LOGICOP_MASK (0xf<<18) -#define MODE4_ENABLE_STENCIL_TEST_MASK ((1<<17)|(0xff00)) -#define ENABLE_STENCIL_TEST_MASK (1<<17) -#define STENCIL_TEST_MASK(x) (((x)&0xff)<<8) -#define MODE4_ENABLE_STENCIL_WRITE_MASK ((1<<16)|(0x00ff)) -#define ENABLE_STENCIL_WRITE_MASK (1<<16) -#define STENCIL_WRITE_MASK(x) ((x)&0xff) - -/* _3DSTATE_MODES_5, p220 */ -#define _3DSTATE_MODES_5_CMD (CMD_3D|(0x0c<<24)) -#define PIPELINE_FLUSH_RENDER_CACHE (1<<18) -#define PIPELINE_FLUSH_TEXTURE_CACHE (1<<16) - - -/* p221 */ -#define _3DSTATE_PIXEL_SHADER_CONSTANTS (CMD_3D|(0x1d<<24)|(0x6<<16)) -#define PS1_REG(n) (1<<(n)) -#define PS2_CONST_X(n) (n) -#define PS3_CONST_Y(n) (n) -#define PS4_CONST_Z(n) (n) -#define PS5_CONST_W(n) (n) - -/* p222 */ - - -#define I915_MAX_TEX_INDIRECT 4 -#define I915_MAX_TEX_INSN 32 -#define I915_MAX_ALU_INSN 64 -#define I915_MAX_DECL_INSN 27 -#define I915_MAX_TEMPORARY 16 - - -/* Each instruction is 3 dwords long, though most don't require all - * this space. Maximum of 123 instructions. Smaller maxes per insn - * type. - */ -#define _3DSTATE_PIXEL_SHADER_PROGRAM (CMD_3D|(0x1d<<24)|(0x5<<16)) - -#define REG_TYPE_R 0 /* temporary regs, no need to - * dcl, must be written before - * read -- Preserved between - * phases. - */ -#define REG_TYPE_T 1 /* Interpolated values, must be - * dcl'ed before use. - * - * 0..7: texture coord, - * 8: diffuse spec, - * 9: specular color, - * 10: fog parameter in w. - */ -#define REG_TYPE_CONST 2 /* Restriction: only one const - * can be referenced per - * instruction, though it may be - * selected for multiple inputs. - * Constants not initialized - * default to zero. - */ -#define REG_TYPE_S 3 /* sampler */ -#define REG_TYPE_OC 4 /* output color (rgba) */ -#define REG_TYPE_OD 5 /* output depth (w), xyz are - * temporaries. If not written, - * interpolated depth is used? - */ -#define REG_TYPE_U 6 /* unpreserved temporaries */ -#define REG_TYPE_MASK 0x7 -#define REG_NR_MASK 0xf - - -/* REG_TYPE_T: - */ -#define T_TEX0 0 -#define T_TEX1 1 -#define T_TEX2 2 -#define T_TEX3 3 -#define T_TEX4 4 -#define T_TEX5 5 -#define T_TEX6 6 -#define T_TEX7 7 -#define T_DIFFUSE 8 -#define T_SPECULAR 9 -#define T_FOG_W 10 /* interpolated fog is in W coord */ - -/* Arithmetic instructions */ - -/* .replicate_swizzle == selection and replication of a particular - * scalar channel, ie., .xxxx, .yyyy, .zzzz or .wwww - */ -#define A0_NOP (0x0<<24) /* no operation */ -#define A0_ADD (0x1<<24) /* dst = src0 + src1 */ -#define A0_MOV (0x2<<24) /* dst = src0 */ -#define A0_MUL (0x3<<24) /* dst = src0 * src1 */ -#define A0_MAD (0x4<<24) /* dst = src0 * src1 + src2 */ -#define A0_DP2ADD (0x5<<24) /* dst.xyzw = src0.xy dot src1.xy + src2.replicate_swizzle */ -#define A0_DP3 (0x6<<24) /* dst.xyzw = src0.xyz dot src1.xyz */ -#define A0_DP4 (0x7<<24) /* dst.xyzw = src0.xyzw dot src1.xyzw */ -#define A0_FRC (0x8<<24) /* dst = src0 - floor(src0) */ -#define A0_RCP (0x9<<24) /* dst.xyzw = 1/(src0.replicate_swizzle) */ -#define A0_RSQ (0xa<<24) /* dst.xyzw = 1/(sqrt(abs(src0.replicate_swizzle))) */ -#define A0_EXP (0xb<<24) /* dst.xyzw = exp2(src0.replicate_swizzle) */ -#define A0_LOG (0xc<<24) /* dst.xyzw = log2(abs(src0.replicate_swizzle)) */ -#define A0_CMP (0xd<<24) /* dst = (src0 >= 0.0) ? src1 : src2 */ -#define A0_MIN (0xe<<24) /* dst = (src0 < src1) ? src0 : src1 */ -#define A0_MAX (0xf<<24) /* dst = (src0 >= src1) ? src0 : src1 */ -#define A0_FLR (0x10<<24) /* dst = floor(src0) */ -#define A0_MOD (0x11<<24) /* dst = src0 fmod 1.0 */ -#define A0_TRC (0x12<<24) /* dst = int(src0) */ -#define A0_SGE (0x13<<24) /* dst = src0 >= src1 ? 1.0 : 0.0 */ -#define A0_SLT (0x14<<24) /* dst = src0 < src1 ? 1.0 : 0.0 */ -#define A0_DEST_SATURATE (1<<22) -#define A0_DEST_TYPE_SHIFT 19 -/* Allow: R, OC, OD, U */ -#define A0_DEST_NR_SHIFT 14 -/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ -#define A0_DEST_CHANNEL_X (1<<10) -#define A0_DEST_CHANNEL_Y (2<<10) -#define A0_DEST_CHANNEL_Z (4<<10) -#define A0_DEST_CHANNEL_W (8<<10) -#define A0_DEST_CHANNEL_ALL (0xf<<10) -#define A0_DEST_CHANNEL_SHIFT 10 -#define A0_SRC0_TYPE_SHIFT 7 -#define A0_SRC0_NR_SHIFT 2 - -#define A0_DEST_CHANNEL_XY (A0_DEST_CHANNEL_X|A0_DEST_CHANNEL_Y) -#define A0_DEST_CHANNEL_XYZ (A0_DEST_CHANNEL_XY|A0_DEST_CHANNEL_Z) - - -#define SRC_X 0 -#define SRC_Y 1 -#define SRC_Z 2 -#define SRC_W 3 -#define SRC_ZERO 4 -#define SRC_ONE 5 - -#define A1_SRC0_CHANNEL_X_NEGATE (1<<31) -#define A1_SRC0_CHANNEL_X_SHIFT 28 -#define A1_SRC0_CHANNEL_Y_NEGATE (1<<27) -#define A1_SRC0_CHANNEL_Y_SHIFT 24 -#define A1_SRC0_CHANNEL_Z_NEGATE (1<<23) -#define A1_SRC0_CHANNEL_Z_SHIFT 20 -#define A1_SRC0_CHANNEL_W_NEGATE (1<<19) -#define A1_SRC0_CHANNEL_W_SHIFT 16 -#define A1_SRC1_TYPE_SHIFT 13 -#define A1_SRC1_NR_SHIFT 8 -#define A1_SRC1_CHANNEL_X_NEGATE (1<<7) -#define A1_SRC1_CHANNEL_X_SHIFT 4 -#define A1_SRC1_CHANNEL_Y_NEGATE (1<<3) -#define A1_SRC1_CHANNEL_Y_SHIFT 0 - -#define A2_SRC1_CHANNEL_Z_NEGATE (1<<31) -#define A2_SRC1_CHANNEL_Z_SHIFT 28 -#define A2_SRC1_CHANNEL_W_NEGATE (1<<27) -#define A2_SRC1_CHANNEL_W_SHIFT 24 -#define A2_SRC2_TYPE_SHIFT 21 -#define A2_SRC2_NR_SHIFT 16 -#define A2_SRC2_CHANNEL_X_NEGATE (1<<15) -#define A2_SRC2_CHANNEL_X_SHIFT 12 -#define A2_SRC2_CHANNEL_Y_NEGATE (1<<11) -#define A2_SRC2_CHANNEL_Y_SHIFT 8 -#define A2_SRC2_CHANNEL_Z_NEGATE (1<<7) -#define A2_SRC2_CHANNEL_Z_SHIFT 4 -#define A2_SRC2_CHANNEL_W_NEGATE (1<<3) -#define A2_SRC2_CHANNEL_W_SHIFT 0 - - - -/* Texture instructions */ -#define T0_TEXLD (0x15<<24) /* Sample texture using predeclared - * sampler and address, and output - * filtered texel data to destination - * register */ -#define T0_TEXLDP (0x16<<24) /* Same as texld but performs a - * perspective divide of the texture - * coordinate .xyz values by .w before - * sampling. */ -#define T0_TEXLDB (0x17<<24) /* Same as texld but biases the - * computed LOD by w. Only S4.6 two's - * comp is used. This implies that a - * float to fixed conversion is - * done. */ -#define T0_TEXKILL (0x18<<24) /* Does not perform a sampling - * operation. Simply kills the pixel - * if any channel of the address - * register is < 0.0. */ -#define T0_DEST_TYPE_SHIFT 19 -/* Allow: R, OC, OD, U */ -/* Note: U (unpreserved) regs do not retain their values between - * phases (cannot be used for feedback) - * - * Note: oC and OD registers can only be used as the destination of a - * texture instruction once per phase (this is an implementation - * restriction). - */ -#define T0_DEST_NR_SHIFT 14 -/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ -#define T0_SAMPLER_NR_SHIFT 0 /* This field ignored for TEXKILL */ -#define T0_SAMPLER_NR_MASK (0xf<<0) - -#define T1_ADDRESS_REG_TYPE_SHIFT 24 /* Reg to use as texture coord */ -/* Allow R, T, OC, OD -- R, OC, OD are 'dependent' reads, new program phase */ -#define T1_ADDRESS_REG_NR_SHIFT 17 -#define T2_MBZ 0 - -/* Declaration instructions */ -#define D0_DCL (0x19<<24) /* Declare a t (interpolated attrib) - * register or an s (sampler) - * register. */ -#define D0_SAMPLE_TYPE_SHIFT 22 -#define D0_SAMPLE_TYPE_2D (0x0<<22) -#define D0_SAMPLE_TYPE_CUBE (0x1<<22) -#define D0_SAMPLE_TYPE_VOLUME (0x2<<22) -#define D0_SAMPLE_TYPE_MASK (0x3<<22) - -#define D0_TYPE_SHIFT 19 -/* Allow: T, S */ -#define D0_NR_SHIFT 14 -/* Allow T: 0..10, S: 0..15 */ -#define D0_CHANNEL_X (1<<10) -#define D0_CHANNEL_Y (2<<10) -#define D0_CHANNEL_Z (4<<10) -#define D0_CHANNEL_W (8<<10) -#define D0_CHANNEL_ALL (0xf<<10) -#define D0_CHANNEL_NONE (0<<10) - -#define D0_CHANNEL_XY (D0_CHANNEL_X|D0_CHANNEL_Y) -#define D0_CHANNEL_XYZ (D0_CHANNEL_XY|D0_CHANNEL_Z) - -/* I915 Errata: Do not allow (xz), (xw), (xzw) combinations for diffuse - * or specular declarations. - * - * For T dcls, only allow: (x), (xy), (xyz), (w), (xyzw) - * - * Must be zero for S (sampler) dcls - */ -#define D1_MBZ 0 -#define D2_MBZ 0 - - - -/* p207 */ -#define _3DSTATE_MAP_STATE (CMD_3D|(0x1d<<24)|(0x0<<16)) - -#define MS1_MAPMASK_SHIFT 0 -#define MS1_MAPMASK_MASK (0x8fff<<0) - -#define MS2_UNTRUSTED_SURFACE (1<<31) -#define MS2_ADDRESS_MASK 0xfffffffc -#define MS2_VERTICAL_LINE_STRIDE (1<<1) -#define MS2_VERTICAL_OFFSET (1<<1) - -#define MS3_HEIGHT_SHIFT 21 -#define MS3_WIDTH_SHIFT 10 -#define MS3_PALETTE_SELECT (1<<9) -#define MS3_MAPSURF_FORMAT_SHIFT 7 -#define MS3_MAPSURF_FORMAT_MASK (0x7<<7) -#define MAPSURF_8BIT (1<<7) -#define MAPSURF_16BIT (2<<7) -#define MAPSURF_32BIT (3<<7) -#define MAPSURF_422 (5<<7) -#define MAPSURF_COMPRESSED (6<<7) -#define MAPSURF_4BIT_INDEXED (7<<7) -#define MS3_MT_FORMAT_MASK (0x7 << 3) -#define MS3_MT_FORMAT_SHIFT 3 -#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ -#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ -#define MT_8BIT_L8 (1<<3) -#define MT_8BIT_A8 (4<<3) -#define MT_8BIT_MONO8 (5<<3) -#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ -#define MT_16BIT_ARGB1555 (1<<3) -#define MT_16BIT_ARGB4444 (2<<3) -#define MT_16BIT_AY88 (3<<3) -#define MT_16BIT_88DVDU (5<<3) -#define MT_16BIT_BUMP_655LDVDU (6<<3) -#define MT_16BIT_I16 (7<<3) -#define MT_16BIT_L16 (8<<3) -#define MT_16BIT_A16 (9<<3) -#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ -#define MT_32BIT_ABGR8888 (1<<3) -#define MT_32BIT_XRGB8888 (2<<3) -#define MT_32BIT_XBGR8888 (3<<3) -#define MT_32BIT_QWVU8888 (4<<3) -#define MT_32BIT_AXVU8888 (5<<3) -#define MT_32BIT_LXVU8888 (6<<3) -#define MT_32BIT_XLVU8888 (7<<3) -#define MT_32BIT_ARGB2101010 (8<<3) -#define MT_32BIT_ABGR2101010 (9<<3) -#define MT_32BIT_AWVU2101010 (0xA<<3) -#define MT_32BIT_GR1616 (0xB<<3) -#define MT_32BIT_VU1616 (0xC<<3) -#define MT_32BIT_xI824 (0xD<<3) -#define MT_32BIT_xA824 (0xE<<3) -#define MT_32BIT_xL824 (0xF<<3) -#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ -#define MT_422_YCRCB_NORMAL (1<<3) -#define MT_422_YCRCB_SWAPUV (2<<3) -#define MT_422_YCRCB_SWAPUVY (3<<3) -#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ -#define MT_COMPRESS_DXT2_3 (1<<3) -#define MT_COMPRESS_DXT4_5 (2<<3) -#define MT_COMPRESS_FXT1 (3<<3) -#define MT_COMPRESS_DXT1_RGB (4<<3) -#define MS3_USE_FENCE_REGS (1<<2) -#define MS3_TILED_SURFACE (1<<1) -#define MS3_TILE_WALK (1<<0) - -#define MS4_PITCH_SHIFT 21 -#define MS4_CUBE_FACE_ENA_NEGX (1<<20) -#define MS4_CUBE_FACE_ENA_POSX (1<<19) -#define MS4_CUBE_FACE_ENA_NEGY (1<<18) -#define MS4_CUBE_FACE_ENA_POSY (1<<17) -#define MS4_CUBE_FACE_ENA_NEGZ (1<<16) -#define MS4_CUBE_FACE_ENA_POSZ (1<<15) -#define MS4_CUBE_FACE_ENA_MASK (0x3f<<15) -#define MS4_MAX_LOD_SHIFT 9 -#define MS4_MAX_LOD_MASK (0x3f<<9) -#define MS4_MIP_LAYOUT_LEGACY (0<<8) -#define MS4_MIP_LAYOUT_BELOW_LPT (0<<8) -#define MS4_MIP_LAYOUT_RIGHT_LPT (1<<8) -#define MS4_VOLUME_DEPTH_SHIFT 0 -#define MS4_VOLUME_DEPTH_MASK (0xff<<0) - -/* p244 */ -#define _3DSTATE_SAMPLER_STATE (CMD_3D|(0x1d<<24)|(0x1<<16)) - -#define SS1_MAPMASK_SHIFT 0 -#define SS1_MAPMASK_MASK (0x8fff<<0) - -#define SS2_REVERSE_GAMMA_ENABLE (1<<31) -#define SS2_PACKED_TO_PLANAR_ENABLE (1<<30) -#define SS2_COLORSPACE_CONVERSION (1<<29) -#define SS2_CHROMAKEY_SHIFT 27 -#define SS2_BASE_MIP_LEVEL_SHIFT 22 -#define SS2_BASE_MIP_LEVEL_MASK (0x1f<<22) -#define SS2_MIP_FILTER_SHIFT 20 -#define SS2_MIP_FILTER_MASK (0x3<<20) -#define MIPFILTER_NONE 0 -#define MIPFILTER_NEAREST 1 -#define MIPFILTER_LINEAR 3 -#define SS2_MAG_FILTER_SHIFT 17 -#define SS2_MAG_FILTER_MASK (0x7<<17) -#define FILTER_NEAREST 0 -#define FILTER_LINEAR 1 -#define FILTER_ANISOTROPIC 2 -#define FILTER_4X4_1 3 -#define FILTER_4X4_2 4 -#define FILTER_4X4_FLAT 5 -#define FILTER_6X5_MONO 6 /* XXX - check */ -#define SS2_MIN_FILTER_SHIFT 14 -#define SS2_MIN_FILTER_MASK (0x7<<14) -#define SS2_LOD_BIAS_SHIFT 5 -#define SS2_LOD_BIAS_ONE (0x10<<5) -#define SS2_LOD_BIAS_MASK (0x1ff<<5) -/* Shadow requires: - * MT_X8{I,L,A}24 or MT_{I,L,A}16 texture format - * FILTER_4X4_x MIN and MAG filters - */ -#define SS2_SHADOW_ENABLE (1<<4) -#define SS2_MAX_ANISO_MASK (1<<3) -#define SS2_MAX_ANISO_2 (0<<3) -#define SS2_MAX_ANISO_4 (1<<3) -#define SS2_SHADOW_FUNC_SHIFT 0 -#define SS2_SHADOW_FUNC_MASK (0x7<<0) -/* SS2_SHADOW_FUNC values: see COMPAREFUNC_* */ - -#define SS3_MIN_LOD_SHIFT 24 -#define SS3_MIN_LOD_ONE (0x10<<24) -#define SS3_MIN_LOD_MASK (0xff<<24) -#define SS3_KILL_PIXEL_ENABLE (1<<17) -#define SS3_TCX_ADDR_MODE_SHIFT 12 -#define SS3_TCX_ADDR_MODE_MASK (0x7<<12) -#define TEXCOORDMODE_WRAP 0 -#define TEXCOORDMODE_MIRROR 1 -#define TEXCOORDMODE_CLAMP_EDGE 2 -#define TEXCOORDMODE_CUBE 3 -#define TEXCOORDMODE_CLAMP_BORDER 4 -#define TEXCOORDMODE_MIRROR_ONCE 5 -#define SS3_TCY_ADDR_MODE_SHIFT 9 -#define SS3_TCY_ADDR_MODE_MASK (0x7<<9) -#define SS3_TCZ_ADDR_MODE_SHIFT 6 -#define SS3_TCZ_ADDR_MODE_MASK (0x7<<6) -#define SS3_NORMALIZED_COORDS (1<<5) -#define SS3_TEXTUREMAP_INDEX_SHIFT 1 -#define SS3_TEXTUREMAP_INDEX_MASK (0xf<<1) -#define SS3_DEINTERLACER_ENABLE (1<<0) - -#define SS4_BORDER_COLOR_MASK (~0) - -/* 3DSTATE_SPAN_STIPPLE, p258 - */ -#define _3DSTATE_STIPPLE ((0x3<<29)|(0x1d<<24)|(0x83<<16)) -#define ST1_ENABLE (1<<16) -#define ST1_MASK (0xffff) - -#define _3DSTATE_DEFAULT_Z ((0x3<<29)|(0x1d<<24)|(0x98<<16)) -#define _3DSTATE_DEFAULT_DIFFUSE ((0x3<<29)|(0x1d<<24)|(0x99<<16)) -#define _3DSTATE_DEFAULT_SPECULAR ((0x3<<29)|(0x1d<<24)|(0x9a<<16)) - - -#define MI_FLUSH ((0<<29)|(4<<23)) -#define FLUSH_MAP_CACHE (1<<0) -#define INHIBIT_FLUSH_RENDER_CACHE (1<<2) - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/i915_state.c b/src/mesa/drivers/dri/i915tex/i915_state.c deleted file mode 100644 index 1fede6a8d8..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_state.c +++ /dev/null @@ -1,968 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "enums.h" -#include "dd.h" -#include "tnl/tnl.h" -#include "tnl/t_context.h" - -#include "texmem.h" - -#include "drivers/common/driverfuncs.h" - -#include "intel_fbo.h" -#include "intel_screen.h" -#include "intel_batchbuffer.h" - -#include "i915_context.h" -#include "i915_reg.h" - -#define FILE_DEBUG_FLAG DEBUG_STATE - -static void -i915StencilFuncSeparate(GLcontext * ctx, GLenum face, GLenum func, GLint ref, - GLuint mask) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - int test = intel_translate_compare_func(func); - - mask = mask & 0xff; - - DBG("%s : func: %s, ref : 0x%x, mask: 0x%x\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(func), ref, mask); - - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_TEST_MASK; - i915->state.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(mask)); - - i915->state.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_REF_MASK | - S5_STENCIL_TEST_FUNC_MASK); - - i915->state.Ctx[I915_CTXREG_LIS5] |= ((ref << S5_STENCIL_REF_SHIFT) | - (test << - S5_STENCIL_TEST_FUNC_SHIFT)); -} - -static void -i915StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - - DBG("%s : mask 0x%x\n", __FUNCTION__, mask); - - mask = mask & 0xff; - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_STATE4] &= ~MODE4_ENABLE_STENCIL_WRITE_MASK; - i915->state.Ctx[I915_CTXREG_STATE4] |= (ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(mask)); -} - - -static void -i915StencilOpSeparate(GLcontext * ctx, GLenum face, GLenum fail, GLenum zfail, - GLenum zpass) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - int fop = intel_translate_stencil_op(fail); - int dfop = intel_translate_stencil_op(zfail); - int dpop = intel_translate_stencil_op(zpass); - - - DBG("%s: fail : %s, zfail: %s, zpass : %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(fail), - _mesa_lookup_enum_by_nr(zfail), _mesa_lookup_enum_by_nr(zpass)); - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - - i915->state.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_FAIL_MASK | - S5_STENCIL_PASS_Z_FAIL_MASK | - S5_STENCIL_PASS_Z_PASS_MASK); - - i915->state.Ctx[I915_CTXREG_LIS5] |= ((fop << S5_STENCIL_FAIL_SHIFT) | - (dfop << - S5_STENCIL_PASS_Z_FAIL_SHIFT) | - (dpop << - S5_STENCIL_PASS_Z_PASS_SHIFT)); -} - -static void -i915AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - int test = intel_translate_compare_func(func); - GLubyte refByte; - - UNCLAMPED_FLOAT_TO_UBYTE(refByte, ref); - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_LIS6] &= ~(S6_ALPHA_TEST_FUNC_MASK | - S6_ALPHA_REF_MASK); - i915->state.Ctx[I915_CTXREG_LIS6] |= ((test << S6_ALPHA_TEST_FUNC_SHIFT) | - (((GLuint) refByte) << - S6_ALPHA_REF_SHIFT)); -} - -/* This function makes sure that the proper enables are - * set for LogicOp, Independant Alpha Blend, and Blending. - * It needs to be called from numerous places where we - * could change the LogicOp or Independant Alpha Blend without subsequent - * calls to glEnable. - */ -static void -i915EvalLogicOpBlendState(GLcontext * ctx) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - - if (RGBA_LOGICOP_ENABLED(ctx)) { - i915->state.Ctx[I915_CTXREG_LIS5] |= S5_LOGICOP_ENABLE; - i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_CBUF_BLEND_ENABLE; - } - else { - i915->state.Ctx[I915_CTXREG_LIS5] &= ~S5_LOGICOP_ENABLE; - - if (ctx->Color.BlendEnabled) { - i915->state.Ctx[I915_CTXREG_LIS6] |= S6_CBUF_BLEND_ENABLE; - } - else { - i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_CBUF_BLEND_ENABLE; - } - } -} - -static void -i915BlendColor(GLcontext * ctx, const GLfloat color[4]) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - GLubyte r, g, b, a; - - DBG("%s\n", __FUNCTION__); - - UNCLAMPED_FLOAT_TO_UBYTE(r, color[RCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(g, color[GCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(b, color[BCOMP]); - UNCLAMPED_FLOAT_TO_UBYTE(a, color[ACOMP]); - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_BLENDCOLOR1] = - (a << 24) | (r << 16) | (g << 8) | b; -} - - -#define DST_BLND_FACT(f) ((f)<state.Ctx[I915_CTXREG_IAB] & - ~(IAB_SRC_FACTOR_MASK | - IAB_DST_FACTOR_MASK | - (BLENDFUNC_MASK << IAB_FUNC_SHIFT) | IAB_ENABLE)); - - GLuint lis6 = (i915->state.Ctx[I915_CTXREG_LIS6] & - ~(S6_CBUF_SRC_BLEND_FACT_MASK | - S6_CBUF_DST_BLEND_FACT_MASK | S6_CBUF_BLEND_FUNC_MASK)); - - GLuint eqRGB = ctx->Color.BlendEquationRGB; - GLuint eqA = ctx->Color.BlendEquationA; - GLuint srcRGB = ctx->Color.BlendSrcRGB; - GLuint dstRGB = ctx->Color.BlendDstRGB; - GLuint srcA = ctx->Color.BlendSrcA; - GLuint dstA = ctx->Color.BlendDstA; - - if (eqRGB == GL_MIN || eqRGB == GL_MAX) { - srcRGB = dstRGB = GL_ONE; - } - - if (eqA == GL_MIN || eqA == GL_MAX) { - srcA = dstA = GL_ONE; - } - - lis6 |= SRC_BLND_FACT(intel_translate_blend_factor(srcRGB)); - lis6 |= DST_BLND_FACT(intel_translate_blend_factor(dstRGB)); - lis6 |= translate_blend_equation(eqRGB) << S6_CBUF_BLEND_FUNC_SHIFT; - - iab |= SRC_ABLND_FACT(intel_translate_blend_factor(srcA)); - iab |= DST_ABLND_FACT(intel_translate_blend_factor(dstA)); - iab |= translate_blend_equation(eqA) << IAB_FUNC_SHIFT; - - if (srcA != srcRGB || dstA != dstRGB || eqA != eqRGB) - iab |= IAB_ENABLE; - - if (iab != i915->state.Ctx[I915_CTXREG_IAB] || - lis6 != i915->state.Ctx[I915_CTXREG_LIS6]) { - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_IAB] = iab; - i915->state.Ctx[I915_CTXREG_LIS6] = lis6; - } - - /* This will catch a logicop blend equation */ - i915EvalLogicOpBlendState(ctx); -} - - -static void -i915BlendFuncSeparate(GLcontext * ctx, GLenum srcRGB, - GLenum dstRGB, GLenum srcA, GLenum dstA) -{ - i915UpdateBlendState(ctx); -} - - -static void -i915BlendEquationSeparate(GLcontext * ctx, GLenum eqRGB, GLenum eqA) -{ - i915UpdateBlendState(ctx); -} - - -static void -i915DepthFunc(GLcontext * ctx, GLenum func) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - int test = intel_translate_compare_func(func); - - DBG("%s\n", __FUNCTION__); - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_DEPTH_TEST_FUNC_MASK; - i915->state.Ctx[I915_CTXREG_LIS6] |= test << S6_DEPTH_TEST_FUNC_SHIFT; -} - -static void -i915DepthMask(GLcontext * ctx, GLboolean flag) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - - DBG("%s flag (%d)\n", __FUNCTION__, flag); - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - - if (flag && ctx->Depth.Test) - i915->state.Ctx[I915_CTXREG_LIS6] |= S6_DEPTH_WRITE_ENABLE; - else - i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_DEPTH_WRITE_ENABLE; -} - -/* ============================================================= - * Polygon stipple - * - * The i915 supports a 4x4 stipple natively, GL wants 32x32. - * Fortunately stipple is usually a repeating pattern. - */ -static void -i915PolygonStipple(GLcontext * ctx, const GLubyte * mask) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - const GLubyte *m = mask; - GLubyte p[4]; - int i, j, k; - int active = (ctx->Polygon.StippleFlag && - i915->intel.reduced_primitive == GL_TRIANGLES); - GLuint newMask; - - if (active) { - I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE); - i915->state.Stipple[I915_STPREG_ST1] &= ~ST1_ENABLE; - } - - p[0] = mask[12] & 0xf; - p[0] |= p[0] << 4; - p[1] = mask[8] & 0xf; - p[1] |= p[1] << 4; - p[2] = mask[4] & 0xf; - p[2] |= p[2] << 4; - p[3] = mask[0] & 0xf; - p[3] |= p[3] << 4; - - for (k = 0; k < 8; k++) - for (j = 3; j >= 0; j--) - for (i = 0; i < 4; i++, m++) - if (*m != p[j]) { - i915->intel.hw_stipple = 0; - return; - } - - newMask = (((p[0] & 0xf) << 0) | - ((p[1] & 0xf) << 4) | - ((p[2] & 0xf) << 8) | ((p[3] & 0xf) << 12)); - - - if (newMask == 0xffff || newMask == 0x0) { - /* this is needed to make conform pass */ - i915->intel.hw_stipple = 0; - return; - } - - i915->state.Stipple[I915_STPREG_ST1] &= ~0xffff; - i915->state.Stipple[I915_STPREG_ST1] |= newMask; - i915->intel.hw_stipple = 1; - - if (active) - i915->state.Stipple[I915_STPREG_ST1] |= ST1_ENABLE; -} - - -/* ============================================================= - * Hardware clipping - */ -static void -i915Scissor(GLcontext * ctx, GLint x, GLint y, GLsizei w, GLsizei h) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - int x1, y1, x2, y2; - - if (!ctx->DrawBuffer) - return; - - DBG("%s %d,%d %dx%d\n", __FUNCTION__, x, y, w, h); - - if (ctx->DrawBuffer->Name == 0) { - x1 = x; - y1 = ctx->DrawBuffer->Height - (y + h); - x2 = x + w - 1; - y2 = y1 + h - 1; - DBG("%s %d..%d,%d..%d (inverted)\n", __FUNCTION__, x1, x2, y1, y2); - } - else { - /* FBO - not inverted - */ - x1 = x; - y1 = y; - x2 = x + w - 1; - y2 = y + h - 1; - DBG("%s %d..%d,%d..%d (not inverted)\n", __FUNCTION__, x1, x2, y1, y2); - } - - x1 = CLAMP(x1, 0, ctx->DrawBuffer->Width - 1); - y1 = CLAMP(y1, 0, ctx->DrawBuffer->Height - 1); - x2 = CLAMP(x2, 0, ctx->DrawBuffer->Width - 1); - y2 = CLAMP(y2, 0, ctx->DrawBuffer->Height - 1); - - DBG("%s %d..%d,%d..%d (clamped)\n", __FUNCTION__, x1, x2, y1, y2); - - I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); - i915->state.Buffer[I915_DESTREG_SR1] = (y1 << 16) | (x1 & 0xffff); - i915->state.Buffer[I915_DESTREG_SR2] = (y2 << 16) | (x2 & 0xffff); -} - -static void -i915LogicOp(GLcontext * ctx, GLenum opcode) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - int tmp = intel_translate_logic_op(opcode); - - DBG("%s\n", __FUNCTION__); - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_STATE4] &= ~LOGICOP_MASK; - i915->state.Ctx[I915_CTXREG_STATE4] |= LOGIC_OP_FUNC(tmp); -} - - - -static void -i915CullFaceFrontFace(GLcontext * ctx, GLenum unused) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - GLuint mode; - - DBG("%s %d\n", __FUNCTION__, - ctx->DrawBuffer ? ctx->DrawBuffer->Name : 0); - - if (!ctx->Polygon.CullFlag) { - mode = S4_CULLMODE_NONE; - } - else if (ctx->Polygon.CullFaceMode != GL_FRONT_AND_BACK) { - mode = S4_CULLMODE_CW; - - if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) - mode ^= (S4_CULLMODE_CW ^ S4_CULLMODE_CCW); - if (ctx->Polygon.CullFaceMode == GL_FRONT) - mode ^= (S4_CULLMODE_CW ^ S4_CULLMODE_CCW); - if (ctx->Polygon.FrontFace != GL_CCW) - mode ^= (S4_CULLMODE_CW ^ S4_CULLMODE_CCW); - } - else { - mode = S4_CULLMODE_BOTH; - } - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_LIS4] &= ~S4_CULLMODE_MASK; - i915->state.Ctx[I915_CTXREG_LIS4] |= mode; -} - -static void -i915LineWidth(GLcontext * ctx, GLfloat widthf) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - int lis4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_LINE_WIDTH_MASK; - int width; - - DBG("%s\n", __FUNCTION__); - - width = (int) (widthf * 2); - CLAMP_SELF(width, 1, 0xf); - lis4 |= width << S4_LINE_WIDTH_SHIFT; - - if (lis4 != i915->state.Ctx[I915_CTXREG_LIS4]) { - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_LIS4] = lis4; - } -} - -static void -i915PointSize(GLcontext * ctx, GLfloat size) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - int lis4 = i915->state.Ctx[I915_CTXREG_LIS4] & ~S4_POINT_WIDTH_MASK; - GLint point_size = (int) size; - - DBG("%s\n", __FUNCTION__); - - CLAMP_SELF(point_size, 1, 255); - lis4 |= point_size << S4_POINT_WIDTH_SHIFT; - - if (lis4 != i915->state.Ctx[I915_CTXREG_LIS4]) { - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_LIS4] = lis4; - } -} - - -/* ============================================================= - * Color masks - */ - -static void -i915ColorMask(GLcontext * ctx, - GLboolean r, GLboolean g, GLboolean b, GLboolean a) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - GLuint tmp = i915->state.Ctx[I915_CTXREG_LIS5] & ~S5_WRITEDISABLE_MASK; - - DBG("%s r(%d) g(%d) b(%d) a(%d)\n", __FUNCTION__, r, g, b, - a); - - if (!r) - tmp |= S5_WRITEDISABLE_RED; - if (!g) - tmp |= S5_WRITEDISABLE_GREEN; - if (!b) - tmp |= S5_WRITEDISABLE_BLUE; - if (!a) - tmp |= S5_WRITEDISABLE_ALPHA; - - if (tmp != i915->state.Ctx[I915_CTXREG_LIS5]) { - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - i915->state.Ctx[I915_CTXREG_LIS5] = tmp; - } -} - -static void -update_specular(GLcontext * ctx) -{ - /* A hack to trigger the rebuild of the fragment program. - */ - intel_context(ctx)->NewGLState |= _NEW_TEXTURE; -} - -static void -i915LightModelfv(GLcontext * ctx, GLenum pname, const GLfloat * param) -{ - DBG("%s\n", __FUNCTION__); - - if (pname == GL_LIGHT_MODEL_COLOR_CONTROL) { - update_specular(ctx); - } -} - -static void -i915ShadeModel(GLcontext * ctx, GLenum mode) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - - if (mode == GL_SMOOTH) { - i915->state.Ctx[I915_CTXREG_LIS4] &= ~(S4_FLATSHADE_ALPHA | - S4_FLATSHADE_COLOR | - S4_FLATSHADE_SPECULAR); - } - else { - i915->state.Ctx[I915_CTXREG_LIS4] |= (S4_FLATSHADE_ALPHA | - S4_FLATSHADE_COLOR | - S4_FLATSHADE_SPECULAR); - } -} - -/* ============================================================= - * Fog - */ -void -i915_update_fog(GLcontext * ctx) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - GLenum mode; - GLboolean enabled; - GLboolean try_pixel_fog; - - if (ctx->FragmentProgram._Active) { - /* Pull in static fog state from program */ - mode = ctx->FragmentProgram._Current->FogOption; - enabled = (mode != GL_NONE); - try_pixel_fog = 0; - } - else { - enabled = ctx->Fog.Enabled; - mode = ctx->Fog.Mode; -#if 0 - /* XXX - DISABLED -- Need ortho fallback */ - try_pixel_fog = (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT - && ctx->Hint.Fog == GL_NICEST); -#else - try_pixel_fog = 0; -#endif - } - - if (!enabled) { - i915->vertex_fog = I915_FOG_NONE; - } - else if (try_pixel_fog) { - I915_STATECHANGE(i915, I915_UPLOAD_FOG); - i915->state.Fog[I915_FOGREG_MODE1] &= ~FMC1_FOGFUNC_MASK; - i915->vertex_fog = I915_FOG_PIXEL; - - switch (mode) { - case GL_LINEAR: - if (ctx->Fog.End <= ctx->Fog.Start) { - /* XXX - this won't work with fragment programs. Need to - * either fallback or append fog instructions to end of - * program in the case of linear fog. - */ - printf("vertex fog!\n"); - i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_VERTEX; - i915->vertex_fog = I915_FOG_VERTEX; - } - else { - GLfloat c2 = 1.0 / (ctx->Fog.End - ctx->Fog.Start); - GLfloat c1 = ctx->Fog.End * c2; - - i915->state.Fog[I915_FOGREG_MODE1] &= ~FMC1_C1_MASK; - i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_PIXEL_LINEAR; - i915->state.Fog[I915_FOGREG_MODE1] |= - ((GLuint) (c1 * FMC1_C1_ONE)) & FMC1_C1_MASK; - - if (i915->state.Fog[I915_FOGREG_MODE1] & FMC1_FOGINDEX_Z) { - i915->state.Fog[I915_FOGREG_MODE2] - = (GLuint) (c2 * FMC2_C2_ONE); - } - else { - fi_type fi; - fi.f = c2; - i915->state.Fog[I915_FOGREG_MODE2] = fi.i; - } - } - break; - case GL_EXP: - i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_PIXEL_EXP; - break; - case GL_EXP2: - i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_PIXEL_EXP2; - break; - default: - break; - } - } - else { /* if (i915->vertex_fog != I915_FOG_VERTEX) */ - I915_STATECHANGE(i915, I915_UPLOAD_FOG); - i915->state.Fog[I915_FOGREG_MODE1] &= ~FMC1_FOGFUNC_MASK; - i915->state.Fog[I915_FOGREG_MODE1] |= FMC1_FOGFUNC_VERTEX; - i915->vertex_fog = I915_FOG_VERTEX; - } - - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - I915_ACTIVESTATE(i915, I915_UPLOAD_FOG, enabled); - if (enabled) - i915->state.Ctx[I915_CTXREG_LIS5] |= S5_FOG_ENABLE; - else - i915->state.Ctx[I915_CTXREG_LIS5] &= ~S5_FOG_ENABLE; - - /* Always enable pixel fog. Vertex fog using fog coord will conflict - * with fog code appended onto fragment program. - */ - _tnl_allow_vertex_fog( ctx, 0 ); - _tnl_allow_pixel_fog( ctx, 1 ); -} - -static void -i915Fogfv(GLcontext * ctx, GLenum pname, const GLfloat * param) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - - switch (pname) { - case GL_FOG_COORDINATE_SOURCE_EXT: - case GL_FOG_MODE: - case GL_FOG_START: - case GL_FOG_END: - break; - - case GL_FOG_DENSITY: - I915_STATECHANGE(i915, I915_UPLOAD_FOG); - - if (i915->state.Fog[I915_FOGREG_MODE1] & FMC1_FOGINDEX_Z) { - i915->state.Fog[I915_FOGREG_MODE3] = - (GLuint) (ctx->Fog.Density * FMC3_D_ONE); - } - else { - fi_type fi; - fi.f = ctx->Fog.Density; - i915->state.Fog[I915_FOGREG_MODE3] = fi.i; - } - break; - - case GL_FOG_COLOR: - I915_STATECHANGE(i915, I915_UPLOAD_FOG); - i915->state.Fog[I915_FOGREG_COLOR] = - (_3DSTATE_FOG_COLOR_CMD | - ((GLubyte) (ctx->Fog.Color[0] * 255.0F) << 16) | - ((GLubyte) (ctx->Fog.Color[1] * 255.0F) << 8) | - ((GLubyte) (ctx->Fog.Color[2] * 255.0F) << 0)); - break; - - default: - break; - } -} - -static void -i915Hint(GLcontext * ctx, GLenum target, GLenum state) -{ - switch (target) { - case GL_FOG_HINT: - break; - default: - break; - } -} - -/* ============================================================= - */ - -static void -i915Enable(GLcontext * ctx, GLenum cap, GLboolean state) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - - switch (cap) { - case GL_TEXTURE_2D: - break; - - case GL_LIGHTING: - case GL_COLOR_SUM: - update_specular(ctx); - break; - - case GL_ALPHA_TEST: - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - if (state) - i915->state.Ctx[I915_CTXREG_LIS6] |= S6_ALPHA_TEST_ENABLE; - else - i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_ALPHA_TEST_ENABLE; - break; - - case GL_BLEND: - i915EvalLogicOpBlendState(ctx); - break; - - case GL_COLOR_LOGIC_OP: - i915EvalLogicOpBlendState(ctx); - - /* Logicop doesn't seem to work at 16bpp: - */ - if (i915->intel.intelScreen->cpp == 2) /* XXX FBO fix */ - FALLBACK(&i915->intel, I915_FALLBACK_LOGICOP, state); - break; - - case GL_FRAGMENT_PROGRAM_ARB: - break; - - case GL_DITHER: - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - if (state) - i915->state.Ctx[I915_CTXREG_LIS5] |= S5_COLOR_DITHER_ENABLE; - else - i915->state.Ctx[I915_CTXREG_LIS5] &= ~S5_COLOR_DITHER_ENABLE; - break; - - case GL_DEPTH_TEST: - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - if (state) - i915->state.Ctx[I915_CTXREG_LIS6] |= S6_DEPTH_TEST_ENABLE; - else - i915->state.Ctx[I915_CTXREG_LIS6] &= ~S6_DEPTH_TEST_ENABLE; - - i915DepthMask(ctx, ctx->Depth.Mask); - break; - - case GL_SCISSOR_TEST: - I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); - if (state) - i915->state.Buffer[I915_DESTREG_SENABLE] = - (_3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT); - else - i915->state.Buffer[I915_DESTREG_SENABLE] = - (_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); - break; - - case GL_LINE_SMOOTH: - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - if (state) - i915->state.Ctx[I915_CTXREG_LIS4] |= S4_LINE_ANTIALIAS_ENABLE; - else - i915->state.Ctx[I915_CTXREG_LIS4] &= ~S4_LINE_ANTIALIAS_ENABLE; - break; - - case GL_FOG: - break; - - case GL_CULL_FACE: - i915CullFaceFrontFace(ctx, 0); - break; - - case GL_STENCIL_TEST: - { - GLboolean hw_stencil = GL_FALSE; - if (ctx->DrawBuffer) { - struct intel_renderbuffer *irbStencil - = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL); - hw_stencil = (irbStencil && irbStencil->region); - } - if (hw_stencil) { - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - if (state) - i915->state.Ctx[I915_CTXREG_LIS5] |= (S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE); - else - i915->state.Ctx[I915_CTXREG_LIS5] &= ~(S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE); - } - else { - FALLBACK(&i915->intel, I915_FALLBACK_STENCIL, state); - } - } - break; - - case GL_POLYGON_STIPPLE: - /* The stipple command worked on my 855GM box, but not my 845G. - * I'll do more testing later to find out exactly which hardware - * supports it. Disabled for now. - */ - if (i915->intel.hw_stipple && - i915->intel.reduced_primitive == GL_TRIANGLES) { - I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE); - if (state) - i915->state.Stipple[I915_STPREG_ST1] |= ST1_ENABLE; - else - i915->state.Stipple[I915_STPREG_ST1] &= ~ST1_ENABLE; - } - break; - - case GL_POLYGON_SMOOTH: - break; - - case GL_POINT_SMOOTH: - break; - - default: - ; - } -} - - -static void -i915_init_packets(struct i915_context *i915) -{ - intelScreenPrivate *screen = i915->intel.intelScreen; - - /* Zero all state */ - memset(&i915->state, 0, sizeof(i915->state)); - - - { - I915_STATECHANGE(i915, I915_UPLOAD_CTX); - /* Probably don't want to upload all this stuff every time one - * piece changes. - */ - i915->state.Ctx[I915_CTXREG_LI] = (_3DSTATE_LOAD_STATE_IMMEDIATE_1 | - I1_LOAD_S(2) | - I1_LOAD_S(4) | - I1_LOAD_S(5) | I1_LOAD_S(6) | (3)); - i915->state.Ctx[I915_CTXREG_LIS2] = 0; - i915->state.Ctx[I915_CTXREG_LIS4] = 0; - i915->state.Ctx[I915_CTXREG_LIS5] = 0; - - if (screen->cpp == 2) /* XXX FBO fix */ - i915->state.Ctx[I915_CTXREG_LIS5] |= S5_COLOR_DITHER_ENABLE; - - - i915->state.Ctx[I915_CTXREG_LIS6] = (S6_COLOR_WRITE_ENABLE | - (2 << S6_TRISTRIP_PV_SHIFT)); - - i915->state.Ctx[I915_CTXREG_STATE4] = (_3DSTATE_MODES_4_CMD | - ENABLE_LOGIC_OP_FUNC | - LOGIC_OP_FUNC(LOGICOP_COPY) | - ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(0xff) | - ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(0xff)); - - i915->state.Ctx[I915_CTXREG_IAB] = - (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | IAB_MODIFY_ENABLE | - IAB_MODIFY_FUNC | IAB_MODIFY_SRC_FACTOR | IAB_MODIFY_DST_FACTOR); - - i915->state.Ctx[I915_CTXREG_BLENDCOLOR0] = - _3DSTATE_CONST_BLEND_COLOR_CMD; - i915->state.Ctx[I915_CTXREG_BLENDCOLOR1] = 0; - - } - - { - I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE); - i915->state.Stipple[I915_STPREG_ST0] = _3DSTATE_STIPPLE; - } - - - { - I915_STATECHANGE(i915, I915_UPLOAD_FOG); - i915->state.Fog[I915_FOGREG_MODE0] = _3DSTATE_FOG_MODE_CMD; - i915->state.Fog[I915_FOGREG_MODE1] = (FMC1_FOGFUNC_MODIFY_ENABLE | - FMC1_FOGFUNC_VERTEX | - FMC1_FOGINDEX_MODIFY_ENABLE | - FMC1_FOGINDEX_W | - FMC1_C1_C2_MODIFY_ENABLE | - FMC1_DENSITY_MODIFY_ENABLE); - i915->state.Fog[I915_FOGREG_COLOR] = _3DSTATE_FOG_COLOR_CMD; - } - - - { - I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); - - i915->state.Buffer[I915_DESTREG_DV0] = _3DSTATE_DST_BUF_VARS_CMD; - - - /* scissor */ - i915->state.Buffer[I915_DESTREG_SENABLE] = - (_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); - i915->state.Buffer[I915_DESTREG_SR0] = _3DSTATE_SCISSOR_RECT_0_CMD; - i915->state.Buffer[I915_DESTREG_SR1] = 0; - i915->state.Buffer[I915_DESTREG_SR2] = 0; - } - - - /* These will be emitted every at the head of every buffer, unless - * we get hardware contexts working. - */ - i915->state.active = (I915_UPLOAD_PROGRAM | - I915_UPLOAD_STIPPLE | - I915_UPLOAD_CTX | - I915_UPLOAD_BUFFERS | I915_UPLOAD_INVARIENT); -} - -void -i915InitStateFunctions(struct dd_function_table *functions) -{ - functions->AlphaFunc = i915AlphaFunc; - functions->BlendColor = i915BlendColor; - functions->BlendEquationSeparate = i915BlendEquationSeparate; - functions->BlendFuncSeparate = i915BlendFuncSeparate; - functions->ColorMask = i915ColorMask; - functions->CullFace = i915CullFaceFrontFace; - functions->DepthFunc = i915DepthFunc; - functions->DepthMask = i915DepthMask; - functions->Enable = i915Enable; - functions->Fogfv = i915Fogfv; - functions->FrontFace = i915CullFaceFrontFace; - functions->Hint = i915Hint; - functions->LightModelfv = i915LightModelfv; - functions->LineWidth = i915LineWidth; - functions->LogicOpcode = i915LogicOp; - functions->PointSize = i915PointSize; - functions->PolygonStipple = i915PolygonStipple; - functions->Scissor = i915Scissor; - functions->ShadeModel = i915ShadeModel; - functions->StencilFuncSeparate = i915StencilFuncSeparate; - functions->StencilMaskSeparate = i915StencilMaskSeparate; - functions->StencilOpSeparate = i915StencilOpSeparate; -} - - -void -i915InitState(struct i915_context *i915) -{ - GLcontext *ctx = &i915->intel.ctx; - - i915_init_packets(i915); - - _mesa_init_driver_state(ctx); - - memcpy(&i915->initial, &i915->state, sizeof(i915->state)); - i915->current = &i915->state; -} diff --git a/src/mesa/drivers/dri/i915tex/i915_tex.c b/src/mesa/drivers/dri/i915tex/i915_tex.c deleted file mode 100644 index 59e148ca04..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_tex.c +++ /dev/null @@ -1,78 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "mtypes.h" -#include "imports.h" -#include "simple_list.h" -#include "enums.h" -#include "image.h" -#include "texstore.h" -#include "texformat.h" -#include "texmem.h" -#include "swrast/swrast.h" - -#include "mm.h" - -#include "intel_ioctl.h" - -#include "i915_context.h" -#include "i915_reg.h" - - - -static void -i915TexEnv(GLcontext * ctx, GLenum target, - GLenum pname, const GLfloat * param) -{ - struct i915_context *i915 = I915_CONTEXT(ctx); - - switch (pname) { - case GL_TEXTURE_LOD_BIAS:{ - GLuint unit = ctx->Texture.CurrentUnit; - GLint b = (int) ((*param) * 16.0); - if (b > 255) - b = 255; - if (b < -256) - b = -256; - I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit)); - i915->lodbias_ss2[unit] = - ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); - break; - } - - default: - break; - } -} - - -void -i915InitTextureFuncs(struct dd_function_table *functions) -{ - functions->TexEnv = i915TexEnv; -} diff --git a/src/mesa/drivers/dri/i915tex/i915_tex_layout.c b/src/mesa/drivers/dri/i915tex/i915_tex_layout.c deleted file mode 100644 index 7b761a7b22..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_tex_layout.c +++ /dev/null @@ -1,343 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Code to layout images in a mipmap tree for i915 and i945 - * respectively. - */ - -#include "intel_mipmap_tree.h" -#include "intel_tex_layout.h" -#include "macros.h" -#include "intel_context.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -static GLint initial_offsets[6][2] = { {0, 0}, -{0, 2}, -{1, 0}, -{1, 2}, -{1, 1}, -{1, 3} -}; - - -static GLint step_offsets[6][2] = { {0, 2}, -{0, 2}, -{-1, 2}, -{-1, 2}, -{-1, 1}, -{-1, 1} -}; - -GLboolean -i915_miptree_layout(struct intel_mipmap_tree * mt) -{ - GLint level; - - switch (mt->target) { - case GL_TEXTURE_CUBE_MAP:{ - const GLuint dim = mt->width0; - GLuint face; - GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* double pitch for cube layouts */ - mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; - mt->total_height = dim * 4; - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_level_info(mt, level, 6, - 0, 0, - /*OLD: mt->pitch, mt->total_height,*/ - lvlWidth, lvlHeight, - 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - for (face = 0; face < 6; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_image_offset(mt, level, face, x, y); - - if (d == 0) - _mesa_printf("cube mipmap %d/%d (%d..%d) is 0x0\n", - face, level, mt->first_level, mt->last_level); - - d >>= 1; - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - } - } - break; - } - case GL_TEXTURE_3D:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint stack_height = 0; - - /* Calculate the size of a single slice. - */ - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - - /* XXX: hardware expects/requires 9 levels at minimum. - */ - for (level = mt->first_level; level <= MAX2(8, mt->last_level); - level++) { - intel_miptree_set_level_info(mt, level, depth, 0, mt->total_height, - width, height, depth); - - - stack_height += MAX2(2, height); - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - - /* Fixup depth image_offsets: - */ - depth = mt->depth0; - for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint i; - for (i = 0; i < depth; i++) - intel_miptree_set_image_offset(mt, level, i, - 0, i * stack_height); - - depth = minify(depth); - } - - - /* Multiply slice size by texture depth for total size. It's - * remarkable how wasteful of memory the i915 texture layouts - * are. They are largely fixed in the i945. - */ - mt->total_height = stack_height * mt->depth0; - break; - } - - default:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint img_height; - - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - mt->total_height = 0; - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_level_info(mt, level, 1, - 0, mt->total_height, - width, height, 1); - - if (mt->compressed) - img_height = MAX2(1, height / 4); - else - img_height = (MAX2(2, height) + 1) & ~1; - - mt->total_height += img_height; - - width = minify(width); - height = minify(height); - } - break; - } - } - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); - - return GL_TRUE; -} - - -GLboolean -i945_miptree_layout(struct intel_mipmap_tree * mt) -{ - GLint level; - - switch (mt->target) { - case GL_TEXTURE_CUBE_MAP:{ - const GLuint dim = mt->width0; - GLuint face; - GLuint lvlWidth = mt->width0, lvlHeight = mt->height0; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* Depending on the size of the largest images, pitch can be - * determined either by the old-style packing of cubemap faces, - * or the final row of 4x4, 2x2 and 1x1 faces below this. - */ - if (dim > 32) - mt->pitch = ((dim * mt->cpp * 2 + 3) & ~3) / mt->cpp; - else - mt->pitch = 14 * 8; - - mt->total_height = dim * 4 + 4; - - /* Set all the levels to effectively occupy the whole rectangular region. - */ - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_level_info(mt, level, 6, - 0, 0, - lvlWidth, lvlHeight, 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - - for (face = 0; face < 6; face++) { - GLuint x = initial_offsets[face][0] * dim; - GLuint y = initial_offsets[face][1] * dim; - GLuint d = dim; - - if (dim == 4 && face >= 4) { - y = mt->total_height - 4; - x = (face - 4) * 8; - } - else if (dim < 4 && (face > 0 || mt->first_level > 0)) { - y = mt->total_height - 4; - x = face * 8; - } - - for (level = mt->first_level; level <= mt->last_level; level++) { - intel_miptree_set_image_offset(mt, level, face, x, y); - - d >>= 1; - - switch (d) { - case 4: - switch (face) { - case FACE_POS_X: - case FACE_NEG_X: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - case FACE_POS_Y: - case FACE_NEG_Y: - y += 12; - x -= 8; - break; - case FACE_POS_Z: - case FACE_NEG_Z: - y = mt->total_height - 4; - x = (face - 4) * 8; - break; - } - - case 2: - y = mt->total_height - 4; - x = 16 + face * 8; - break; - - case 1: - x += 48; - break; - - default: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - } - } - } - break; - } - case GL_TEXTURE_3D:{ - GLuint width = mt->width0; - GLuint height = mt->height0; - GLuint depth = mt->depth0; - GLuint pack_x_pitch, pack_x_nr; - GLuint pack_y_pitch; - GLuint level; - - mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp; - mt->total_height = 0; - - pack_y_pitch = MAX2(mt->height0, 2); - pack_x_pitch = mt->pitch; - pack_x_nr = 1; - - for (level = mt->first_level; level <= mt->last_level; level++) { - GLuint nr_images = mt->target == GL_TEXTURE_3D ? depth : 6; - GLint x = 0; - GLint y = 0; - GLint q, j; - - intel_miptree_set_level_info(mt, level, nr_images, - 0, mt->total_height, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - intel_miptree_set_image_offset(mt, level, q, x, y); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - mt->total_height += y; - - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= mt->pitch); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - } - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - break; - } - - case GL_TEXTURE_1D: - case GL_TEXTURE_2D: - case GL_TEXTURE_RECTANGLE_ARB: - i945_miptree_layout_2d(mt); - break; - default: - _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()"); - } - - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - mt->pitch, - mt->total_height, mt->cpp, mt->pitch * mt->total_height * mt->cpp); - - return GL_TRUE; -} diff --git a/src/mesa/drivers/dri/i915tex/i915_texstate.c b/src/mesa/drivers/dri/i915tex/i915_texstate.c deleted file mode 100644 index 3d68187cf8..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_texstate.c +++ /dev/null @@ -1,359 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "mtypes.h" -#include "enums.h" -#include "texformat.h" -#include "dri_bufmgr.h" - -#include "intel_mipmap_tree.h" -#include "intel_tex.h" - -#include "i915_context.h" -#include "i915_reg.h" - - -static GLuint -translate_texture_format(GLuint mesa_format) -{ - switch (mesa_format) { - case MESA_FORMAT_L8: - return MAPSURF_8BIT | MT_8BIT_L8; - case MESA_FORMAT_I8: - return MAPSURF_8BIT | MT_8BIT_I8; - case MESA_FORMAT_A8: - return MAPSURF_8BIT | MT_8BIT_A8; - case MESA_FORMAT_AL88: - return MAPSURF_16BIT | MT_16BIT_AY88; - case MESA_FORMAT_RGB565: - return MAPSURF_16BIT | MT_16BIT_RGB565; - case MESA_FORMAT_ARGB1555: - return MAPSURF_16BIT | MT_16BIT_ARGB1555; - case MESA_FORMAT_ARGB4444: - return MAPSURF_16BIT | MT_16BIT_ARGB4444; - case MESA_FORMAT_ARGB8888: - return MAPSURF_32BIT | MT_32BIT_ARGB8888; - case MESA_FORMAT_YCBCR_REV: - return (MAPSURF_422 | MT_422_YCRCB_NORMAL); - case MESA_FORMAT_YCBCR: - return (MAPSURF_422 | MT_422_YCRCB_SWAPY); - case MESA_FORMAT_RGB_FXT1: - case MESA_FORMAT_RGBA_FXT1: - return (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1); - case MESA_FORMAT_Z16: - return (MAPSURF_16BIT | MT_16BIT_L16); - case MESA_FORMAT_RGBA_DXT1: - case MESA_FORMAT_RGB_DXT1: - return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1); - case MESA_FORMAT_RGBA_DXT3: - return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3); - case MESA_FORMAT_RGBA_DXT5: - return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); - case MESA_FORMAT_Z24_S8: - return (MAPSURF_32BIT | MT_32BIT_xL824); - default: - fprintf(stderr, "%s: bad image format %x\n", __FUNCTION__, mesa_format); - abort(); - return 0; - } -} - - - - -/* The i915 (and related graphics cores) do not support GL_CLAMP. The - * Intel drivers for "other operating systems" implement GL_CLAMP as - * GL_CLAMP_TO_EDGE, so the same is done here. - */ -static GLuint -translate_wrap_mode(GLenum wrap) -{ - switch (wrap) { - case GL_REPEAT: - return TEXCOORDMODE_WRAP; - case GL_CLAMP: - return TEXCOORDMODE_CLAMP_EDGE; /* not quite correct */ - case GL_CLAMP_TO_EDGE: - return TEXCOORDMODE_CLAMP_EDGE; - case GL_CLAMP_TO_BORDER: - return TEXCOORDMODE_CLAMP_BORDER; - case GL_MIRRORED_REPEAT: - return TEXCOORDMODE_MIRROR; - default: - return TEXCOORDMODE_WRAP; - } -} - - - -/* Recalculate all state from scratch. Perhaps not the most - * efficient, but this has gotten complex enough that we need - * something which is understandable and reliable. - */ -static GLboolean -i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3) -{ - GLcontext *ctx = &intel->ctx; - struct i915_context *i915 = i915_context(ctx); - struct gl_texture_object *tObj = ctx->Texture.Unit[unit]._Current; - struct intel_texture_object *intelObj = intel_texture_object(tObj); - struct gl_texture_image *firstImage; - GLuint *state = i915->state.Tex[unit], format, pitch; - - memset(state, 0, sizeof(state)); - - /*We need to refcount these. */ - - if (i915->state.tex_buffer[unit] != NULL) { - driBOUnReference(i915->state.tex_buffer[unit]); - i915->state.tex_buffer[unit] = NULL; - } - - if (!intelObj->imageOverride && !intel_finalize_mipmap_tree(intel, unit)) - return GL_FALSE; - - /* Get first image here, since intelObj->firstLevel will get set in - * the intel_finalize_mipmap_tree() call above. - */ - firstImage = tObj->Image[0][intelObj->firstLevel]; - - if (intelObj->imageOverride) { - i915->state.tex_buffer[unit] = NULL; - i915->state.tex_offset[unit] = intelObj->textureOffset; - - switch (intelObj->depthOverride) { - case 32: - format = MAPSURF_32BIT | MT_32BIT_ARGB8888; - break; - case 24: - default: - format = MAPSURF_32BIT | MT_32BIT_XRGB8888; - break; - case 16: - format = MAPSURF_16BIT | MT_16BIT_RGB565; - break; - } - - pitch = intelObj->pitchOverride; - } else { - i915->state.tex_buffer[unit] = driBOReference(intelObj->mt->region-> - buffer); - i915->state.tex_offset[unit] = intel_miptree_image_offset(intelObj->mt, - 0, intelObj-> - firstLevel); - - format = translate_texture_format(firstImage->TexFormat->MesaFormat); - pitch = intelObj->mt->pitch * intelObj->mt->cpp; - } - - state[I915_TEXREG_MS3] = - (((firstImage->Height - 1) << MS3_HEIGHT_SHIFT) | - ((firstImage->Width - 1) << MS3_WIDTH_SHIFT) | format | - MS3_USE_FENCE_REGS); - - state[I915_TEXREG_MS4] = - ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | MS4_CUBE_FACE_ENA_MASK | - ((((intelObj->lastLevel - intelObj->firstLevel) * 4)) << - MS4_MAX_LOD_SHIFT) | ((firstImage->Depth - 1) << - MS4_VOLUME_DEPTH_SHIFT)); - - - { - GLuint minFilt, mipFilt, magFilt; - - switch (tObj->MinFilter) { - case GL_NEAREST: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_NONE; - break; - case GL_LINEAR: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_NONE; - break; - case GL_NEAREST_MIPMAP_NEAREST: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_NEAREST; - break; - case GL_LINEAR_MIPMAP_NEAREST: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_NEAREST; - break; - case GL_NEAREST_MIPMAP_LINEAR: - minFilt = FILTER_NEAREST; - mipFilt = MIPFILTER_LINEAR; - break; - case GL_LINEAR_MIPMAP_LINEAR: - minFilt = FILTER_LINEAR; - mipFilt = MIPFILTER_LINEAR; - break; - default: - return GL_FALSE; - } - - if (tObj->MaxAnisotropy > 1.0) { - minFilt = FILTER_ANISOTROPIC; - magFilt = FILTER_ANISOTROPIC; - } - else { - switch (tObj->MagFilter) { - case GL_NEAREST: - magFilt = FILTER_NEAREST; - break; - case GL_LINEAR: - magFilt = FILTER_LINEAR; - break; - default: - return GL_FALSE; - } - } - - state[I915_TEXREG_SS2] = i915->lodbias_ss2[unit]; - - /* YUV conversion: - */ - if (firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR || - firstImage->TexFormat->MesaFormat == MESA_FORMAT_YCBCR_REV) - state[I915_TEXREG_SS2] |= SS2_COLORSPACE_CONVERSION; - - /* Shadow: - */ - if (tObj->CompareMode == GL_COMPARE_R_TO_TEXTURE_ARB && - tObj->Target != GL_TEXTURE_3D) { - - state[I915_TEXREG_SS2] |= - (SS2_SHADOW_ENABLE | - intel_translate_compare_func(tObj->CompareFunc)); - - minFilt = FILTER_4X4_FLAT; - magFilt = FILTER_4X4_FLAT; - } - - state[I915_TEXREG_SS2] |= ((minFilt << SS2_MIN_FILTER_SHIFT) | - (mipFilt << SS2_MIP_FILTER_SHIFT) | - (magFilt << SS2_MAG_FILTER_SHIFT)); - } - - { - GLenum ws = tObj->WrapS; - GLenum wt = tObj->WrapT; - GLenum wr = tObj->WrapR; - - - /* 3D textures don't seem to respect the border color. - * Fallback if there's ever a danger that they might refer to - * it. - * - * Effectively this means fallback on 3D clamp or - * clamp_to_border. - */ - if (tObj->Target == GL_TEXTURE_3D && - (tObj->MinFilter != GL_NEAREST || - tObj->MagFilter != GL_NEAREST) && - (ws == GL_CLAMP || - wt == GL_CLAMP || - wr == GL_CLAMP || - ws == GL_CLAMP_TO_BORDER || - wt == GL_CLAMP_TO_BORDER || wr == GL_CLAMP_TO_BORDER)) - return GL_FALSE; - - - state[I915_TEXREG_SS3] = ss3; /* SS3_NORMALIZED_COORDS */ - - state[I915_TEXREG_SS3] |= - ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) | - (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | - (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT)); - - state[I915_TEXREG_SS3] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT); - } - - - state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(tObj->_BorderChan[0], - tObj->_BorderChan[1], - tObj->_BorderChan[2], - tObj->_BorderChan[3]); - - - I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(unit), GL_TRUE); - /* memcmp was already disabled, but definitely won't work as the - * region might now change and that wouldn't be detected: - */ - I915_STATECHANGE(i915, I915_UPLOAD_TEX(unit)); - - -#if 0 - DBG(TEXTURE, "state[I915_TEXREG_SS2] = 0x%x\n", state[I915_TEXREG_SS2]); - DBG(TEXTURE, "state[I915_TEXREG_SS3] = 0x%x\n", state[I915_TEXREG_SS3]); - DBG(TEXTURE, "state[I915_TEXREG_SS4] = 0x%x\n", state[I915_TEXREG_SS4]); - DBG(TEXTURE, "state[I915_TEXREG_MS2] = 0x%x\n", state[I915_TEXREG_MS2]); - DBG(TEXTURE, "state[I915_TEXREG_MS3] = 0x%x\n", state[I915_TEXREG_MS3]); - DBG(TEXTURE, "state[I915_TEXREG_MS4] = 0x%x\n", state[I915_TEXREG_MS4]); -#endif - - return GL_TRUE; -} - - - - -void -i915UpdateTextureState(struct intel_context *intel) -{ - GLboolean ok = GL_TRUE; - GLuint i; - - for (i = 0; i < I915_TEX_UNITS && ok; i++) { - switch (intel->ctx.Texture.Unit[i]._ReallyEnabled) { - case TEXTURE_1D_BIT: - case TEXTURE_2D_BIT: - case TEXTURE_CUBE_BIT: - case TEXTURE_3D_BIT: - ok = i915_update_tex_unit(intel, i, SS3_NORMALIZED_COORDS); - break; - case TEXTURE_RECT_BIT: - ok = i915_update_tex_unit(intel, i, 0); - break; - case 0:{ - struct i915_context *i915 = i915_context(&intel->ctx); - if (i915->state.active & I915_UPLOAD_TEX(i)) - I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(i), GL_FALSE); - - if (i915->state.tex_buffer[i] != NULL) { - driBOUnReference(i915->state.tex_buffer[i]); - i915->state.tex_buffer[i] = NULL; - } - - break; - } - default: - ok = GL_FALSE; - break; - } - } - - FALLBACK(intel, I915_FALLBACK_TEXTURE, !ok); -} diff --git a/src/mesa/drivers/dri/i915tex/i915_vtbl.c b/src/mesa/drivers/dri/i915tex/i915_vtbl.c deleted file mode 100644 index b836241a63..0000000000 --- a/src/mesa/drivers/dri/i915tex/i915_vtbl.c +++ /dev/null @@ -1,562 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - - -#include "glheader.h" -#include "mtypes.h" -#include "imports.h" -#include "macros.h" -#include "colormac.h" - -#include "tnl/t_context.h" -#include "tnl/t_vertex.h" - -#include "intel_batchbuffer.h" -#include "intel_tex.h" -#include "intel_regions.h" - -#include "i915_reg.h" -#include "i915_context.h" - -static void -i915_render_prevalidate(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - - i915ValidateFragmentProgram(i915); -} - -static void -i915_render_start(struct intel_context *intel) -{ -} - - -static void -i915_reduced_primitive_state(struct intel_context *intel, GLenum rprim) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - GLuint st1 = i915->state.Stipple[I915_STPREG_ST1]; - - st1 &= ~ST1_ENABLE; - - switch (rprim) { - case GL_QUADS: /* from RASTERIZE(GL_QUADS) in t_dd_tritemp.h */ - case GL_TRIANGLES: - if (intel->ctx.Polygon.StippleFlag && intel->hw_stipple) - st1 |= ST1_ENABLE; - break; - case GL_LINES: - case GL_POINTS: - default: - break; - } - - i915->intel.reduced_primitive = rprim; - - if (st1 != i915->state.Stipple[I915_STPREG_ST1]) { - INTEL_FIREVERTICES(intel); - - I915_STATECHANGE(i915, I915_UPLOAD_STIPPLE); - i915->state.Stipple[I915_STPREG_ST1] = st1; - } -} - - -/* Pull apart the vertex format registers and figure out how large a - * vertex is supposed to be. - */ -static GLboolean -i915_check_vertex_size(struct intel_context *intel, GLuint expected) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - int lis2 = i915->current->Ctx[I915_CTXREG_LIS2]; - int lis4 = i915->current->Ctx[I915_CTXREG_LIS4]; - int i, sz = 0; - - switch (lis4 & S4_VFMT_XYZW_MASK) { - case S4_VFMT_XY: - sz = 2; - break; - case S4_VFMT_XYZ: - sz = 3; - break; - case S4_VFMT_XYW: - sz = 3; - break; - case S4_VFMT_XYZW: - sz = 4; - break; - default: - fprintf(stderr, "no xyzw specified\n"); - return 0; - } - - if (lis4 & S4_VFMT_SPEC_FOG) - sz++; - if (lis4 & S4_VFMT_COLOR) - sz++; - if (lis4 & S4_VFMT_DEPTH_OFFSET) - sz++; - if (lis4 & S4_VFMT_POINT_WIDTH) - sz++; - if (lis4 & S4_VFMT_FOG_PARAM) - sz++; - - for (i = 0; i < 8; i++) { - switch (lis2 & S2_TEXCOORD_FMT0_MASK) { - case TEXCOORDFMT_2D: - sz += 2; - break; - case TEXCOORDFMT_3D: - sz += 3; - break; - case TEXCOORDFMT_4D: - sz += 4; - break; - case TEXCOORDFMT_1D: - sz += 1; - break; - case TEXCOORDFMT_2D_16: - sz += 1; - break; - case TEXCOORDFMT_4D_16: - sz += 2; - break; - case TEXCOORDFMT_NOT_PRESENT: - break; - default: - fprintf(stderr, "bad texcoord fmt %d\n", i); - return GL_FALSE; - } - lis2 >>= S2_TEXCOORD_FMT1_SHIFT; - } - - if (sz != expected) - fprintf(stderr, "vertex size mismatch %d/%d\n", sz, expected); - - return sz == expected; -} - - -static void -i915_emit_invarient_state(struct intel_context *intel) -{ - BATCH_LOCALS; - - BEGIN_BATCH(200, 0); - - OUT_BATCH(_3DSTATE_AA_CMD | - AA_LINE_ECAAR_WIDTH_ENABLE | - AA_LINE_ECAAR_WIDTH_1_0 | - AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0); - - OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_DFLT_SPEC_CMD); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_DFLT_Z_CMD); - OUT_BATCH(0); - - /* Don't support texture crossbar yet */ - OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS | - CSB_TCB(0, 0) | - CSB_TCB(1, 1) | - CSB_TCB(2, 2) | - CSB_TCB(3, 3) | - CSB_TCB(4, 4) | CSB_TCB(5, 5) | CSB_TCB(6, 6) | CSB_TCB(7, 7)); - - OUT_BATCH(_3DSTATE_RASTER_RULES_CMD | - ENABLE_POINT_RASTER_RULE | - OGL_POINT_RASTER_RULE | - ENABLE_LINE_STRIP_PROVOKE_VRTX | - ENABLE_TRI_FAN_PROVOKE_VRTX | - LINE_STRIP_PROVOKE_VRTX(1) | - TRI_FAN_PROVOKE_VRTX(2) | ENABLE_TEXKILL_3D_4D | TEXKILL_4D); - - /* Need to initialize this to zero. - */ - OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (1)); - OUT_BATCH(0); - - /* XXX: Use this */ - OUT_BATCH(_3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT); - - OUT_BATCH(_3DSTATE_SCISSOR_RECT_0_CMD); - OUT_BATCH(0); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_DEPTH_SUBRECT_DISABLE); - - OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0); /* disable indirect state */ - OUT_BATCH(0); - - - /* Don't support twosided stencil yet */ - OUT_BATCH(_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_TWO_SIDE | 0); - - ADVANCE_BATCH(); -} - - -#define emit(intel, state, size ) \ - intel_batchbuffer_data(intel->batch, state, size, 0 ) - -static GLuint -get_dirty(struct i915_hw_state *state) -{ - GLuint dirty; - - /* Workaround the multitex hang - if one texture unit state is - * modified, emit all texture units. - */ - dirty = state->active & ~state->emitted; - if (dirty & I915_UPLOAD_TEX_ALL) - state->emitted &= ~I915_UPLOAD_TEX_ALL; - dirty = state->active & ~state->emitted; - return dirty; -} - - -static GLuint -get_state_size(struct i915_hw_state *state) -{ - GLuint dirty = get_dirty(state); - GLuint i; - GLuint sz = 0; - - if (dirty & I915_UPLOAD_CTX) - sz += sizeof(state->Ctx); - - if (dirty & I915_UPLOAD_BUFFERS) - sz += sizeof(state->Buffer); - - if (dirty & I915_UPLOAD_STIPPLE) - sz += sizeof(state->Stipple); - - if (dirty & I915_UPLOAD_FOG) - sz += sizeof(state->Fog); - - if (dirty & I915_UPLOAD_TEX_ALL) { - int nr = 0; - for (i = 0; i < I915_TEX_UNITS; i++) - if (dirty & I915_UPLOAD_TEX(i)) - nr++; - - sz += (2 + nr * 3) * sizeof(GLuint) * 2; - } - - if (dirty & I915_UPLOAD_CONSTANTS) - sz += state->ConstantSize * sizeof(GLuint); - - if (dirty & I915_UPLOAD_PROGRAM) - sz += state->ProgramSize * sizeof(GLuint); - - return sz; -} - - -/* Push the state into the sarea and/or texture memory. - */ -static void -i915_emit_state(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - struct i915_hw_state *state = i915->current; - int i; - GLuint dirty; - BATCH_LOCALS; - - /* We don't hold the lock at this point, so want to make sure that - * there won't be a buffer wrap. - * - * It might be better to talk about explicit places where - * scheduling is allowed, rather than assume that it is whenever a - * batchbuffer fills up. - */ - intel_batchbuffer_require_space(intel->batch, get_state_size(state), 0); - - /* Do this here as we may have flushed the batchbuffer above, - * causing more state to be dirty! - */ - dirty = get_dirty(state); - - if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr, "%s dirty: %x\n", __FUNCTION__, dirty); - - if (dirty & I915_UPLOAD_INVARIENT) { - if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr, "I915_UPLOAD_INVARIENT:\n"); - i915_emit_invarient_state(intel); - } - - if (dirty & I915_UPLOAD_CTX) { - if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr, "I915_UPLOAD_CTX:\n"); - - emit(intel, state->Ctx, sizeof(state->Ctx)); - } - - if (dirty & I915_UPLOAD_BUFFERS) { - if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr, "I915_UPLOAD_BUFFERS:\n"); - BEGIN_BATCH(I915_DEST_SETUP_SIZE + 2, 0); - if (state->draw_region) { - OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR0]); - OUT_BATCH(state->Buffer[I915_DESTREG_CBUFADDR1]); - OUT_RELOC(state->draw_region->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - state->draw_region->draw_offset); - } - if (state->depth_region) { - OUT_BATCH(state->Buffer[I915_DESTREG_DBUFADDR0]); - OUT_BATCH(state->Buffer[I915_DESTREG_DBUFADDR1]); - OUT_RELOC(state->depth_region->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - state->depth_region->draw_offset); - } - - OUT_BATCH(state->Buffer[I915_DESTREG_DV0]); - OUT_BATCH(state->Buffer[I915_DESTREG_DV1]); - OUT_BATCH(state->Buffer[I915_DESTREG_SENABLE]); - OUT_BATCH(state->Buffer[I915_DESTREG_SR0]); - OUT_BATCH(state->Buffer[I915_DESTREG_SR1]); - OUT_BATCH(state->Buffer[I915_DESTREG_SR2]); - ADVANCE_BATCH(); - } - - if (dirty & I915_UPLOAD_STIPPLE) { - if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr, "I915_UPLOAD_STIPPLE:\n"); - emit(intel, state->Stipple, sizeof(state->Stipple)); - } - - if (dirty & I915_UPLOAD_FOG) { - if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr, "I915_UPLOAD_FOG:\n"); - emit(intel, state->Fog, sizeof(state->Fog)); - } - - /* Combine all the dirty texture state into a single command to - * avoid lockups on I915 hardware. - */ - if (dirty & I915_UPLOAD_TEX_ALL) { - int nr = 0; - - for (i = 0; i < I915_TEX_UNITS; i++) - if (dirty & I915_UPLOAD_TEX(i)) - nr++; - - BEGIN_BATCH(2 + nr * 3, 0); - OUT_BATCH(_3DSTATE_MAP_STATE | (3 * nr)); - OUT_BATCH((dirty & I915_UPLOAD_TEX_ALL) >> I915_UPLOAD_TEX_0_SHIFT); - for (i = 0; i < I915_TEX_UNITS; i++) - if (dirty & I915_UPLOAD_TEX(i)) { - - if (state->tex_buffer[i]) { - OUT_RELOC(state->tex_buffer[i], - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, - state->tex_offset[i]); - } - else if (state == &i915->meta) { - assert(i == 0); - OUT_BATCH(0); - } - else { - OUT_BATCH(state->tex_offset[i]); - } - - OUT_BATCH(state->Tex[i][I915_TEXREG_MS3]); - OUT_BATCH(state->Tex[i][I915_TEXREG_MS4]); - } - ADVANCE_BATCH(); - - BEGIN_BATCH(2 + nr * 3, 0); - OUT_BATCH(_3DSTATE_SAMPLER_STATE | (3 * nr)); - OUT_BATCH((dirty & I915_UPLOAD_TEX_ALL) >> I915_UPLOAD_TEX_0_SHIFT); - for (i = 0; i < I915_TEX_UNITS; i++) - if (dirty & I915_UPLOAD_TEX(i)) { - OUT_BATCH(state->Tex[i][I915_TEXREG_SS2]); - OUT_BATCH(state->Tex[i][I915_TEXREG_SS3]); - OUT_BATCH(state->Tex[i][I915_TEXREG_SS4]); - } - ADVANCE_BATCH(); - } - - if (dirty & I915_UPLOAD_CONSTANTS) { - if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr, "I915_UPLOAD_CONSTANTS:\n"); - emit(intel, state->Constant, state->ConstantSize * sizeof(GLuint)); - } - - if (dirty & I915_UPLOAD_PROGRAM) { - if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr, "I915_UPLOAD_PROGRAM:\n"); - - assert((state->Program[0] & 0x1ff) + 2 == state->ProgramSize); - - emit(intel, state->Program, state->ProgramSize * sizeof(GLuint)); - if (INTEL_DEBUG & DEBUG_STATE) - i915_disassemble_program(state->Program, state->ProgramSize); - } - - state->emitted |= dirty; -} - -static void -i915_destroy_context(struct intel_context *intel) -{ - _tnl_free_vertices(&intel->ctx); -} - - -/** - * Set the drawing regions for the color and depth/stencil buffers. - * This involves setting the pitch, cpp and buffer ID/location. - * Also set pixel format for color and Z rendering - * Used for setting both regular and meta state. - */ -void -i915_state_draw_region(struct intel_context *intel, - struct i915_hw_state *state, - struct intel_region *color_region, - struct intel_region *depth_region) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - GLuint value; - - ASSERT(state == &i915->state || state == &i915->meta); - - if (state->draw_region != color_region) { - intel_region_release(&state->draw_region); - intel_region_reference(&state->draw_region, color_region); - } - if (state->depth_region != depth_region) { - intel_region_release(&state->depth_region); - intel_region_reference(&state->depth_region, depth_region); - } - - /* - * Set stride/cpp values - */ - if (color_region) { -// fprintf(stderr, "color pitch %d\n", color_region->pitch); - state->Buffer[I915_DESTREG_CBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - state->Buffer[I915_DESTREG_CBUFADDR1] = - (BUF_3D_ID_COLOR_BACK | - BUF_3D_PITCH(color_region->pitch * color_region->cpp) | - BUF_3D_USE_FENCE); - } - - if (depth_region) { -// fprintf(stderr, "depth pitch %d\n", depth_region->pitch); - state->Buffer[I915_DESTREG_DBUFADDR0] = _3DSTATE_BUF_INFO_CMD; - state->Buffer[I915_DESTREG_DBUFADDR1] = - (BUF_3D_ID_DEPTH | - BUF_3D_PITCH(depth_region->pitch * depth_region->cpp) | - BUF_3D_USE_FENCE); - } - - /* - * Compute/set I915_DESTREG_DV1 value - */ - value = (DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - LOD_PRECLAMP_OGL | TEX_DEFAULT_COLOR_OGL); - if (color_region && color_region->cpp == 4) { - value |= DV_PF_8888; - } - else { - value |= (DITHER_FULL_ALWAYS | DV_PF_565); - } - if (depth_region && depth_region->cpp == 4) { - value |= DEPTH_FRMT_24_FIXED_8_OTHER; - } - else { - value |= DEPTH_FRMT_16_FIXED; - } - state->Buffer[I915_DESTREG_DV1] = value; - - I915_STATECHANGE(i915, I915_UPLOAD_BUFFERS); -} - - -static void -i915_set_draw_region(struct intel_context *intel, - struct intel_region *color_region, - struct intel_region *depth_region) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - i915_state_draw_region(intel, &i915->state, color_region, depth_region); -} - - - -static void -i915_lost_hardware(struct intel_context *intel) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - i915->state.emitted = 0; -} - -static GLuint -i915_flush_cmd(void) -{ - return MI_FLUSH | FLUSH_MAP_CACHE; -} - -static void -i915_assert_not_dirty( struct intel_context *intel ) -{ - struct i915_context *i915 = i915_context(&intel->ctx); - struct i915_hw_state *state = i915->current; - GLuint dirty = get_dirty(state); - assert(!dirty); -} - - -void -i915InitVtbl(struct i915_context *i915) -{ - i915->intel.vtbl.check_vertex_size = i915_check_vertex_size; - i915->intel.vtbl.destroy = i915_destroy_context; - i915->intel.vtbl.emit_state = i915_emit_state; - i915->intel.vtbl.lost_hardware = i915_lost_hardware; - i915->intel.vtbl.reduced_primitive_state = i915_reduced_primitive_state; - i915->intel.vtbl.render_start = i915_render_start; - i915->intel.vtbl.render_prevalidate = i915_render_prevalidate; - i915->intel.vtbl.set_draw_region = i915_set_draw_region; - i915->intel.vtbl.update_texture_state = i915UpdateTextureState; - i915->intel.vtbl.flush_cmd = i915_flush_cmd; - i915->intel.vtbl.assert_not_dirty = i915_assert_not_dirty; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c deleted file mode 100644 index c92b83bcb3..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c +++ /dev/null @@ -1,340 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_batchbuffer.h" -#include "intel_ioctl.h" - -/* Relocations in kernel space: - * - pass dma buffer seperately - * - memory manager knows how to patch - * - pass list of dependent buffers - * - pass relocation list - * - * Either: - * - get back an offset for buffer to fire - * - memory manager knows how to fire buffer - * - * Really want the buffer to be AGP and pinned. - * - */ - -/* Cliprect fence: The highest fence protecting a dma buffer - * containing explicit cliprect information. Like the old drawable - * lock but irq-driven. X server must wait for this fence to expire - * before changing cliprects [and then doing sw rendering?]. For - * other dma buffers, the scheduler will grab current cliprect info - * and mix into buffer. X server must hold the lock while changing - * cliprects??? Make per-drawable. Need cliprects in shared memory - * -- beats storing them with every cmd buffer in the queue. - * - * ==> X server must wait for this fence to expire before touching the - * framebuffer with new cliprects. - * - * ==> Cliprect-dependent buffers associated with a - * cliprect-timestamp. All of the buffers associated with a timestamp - * must go to hardware before any buffer with a newer timestamp. - * - * ==> Dma should be queued per-drawable for correct X/GL - * synchronization. Or can fences be used for this? - * - * Applies to: Blit operations, metaops, X server operations -- X - * server automatically waits on its own dma to complete before - * modifying cliprects ??? - */ - -static void -intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) -{ - int i; - fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); - for (i = 0; i < count / 4; i += 4) - fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", - offset + i * 4, ptr[i], ptr[i + 1], ptr[i + 2], ptr[i + 3]); - fprintf(stderr, "END BATCH\n\n\n"); -} - -void -intel_batchbuffer_reset(struct intel_batchbuffer *batch) -{ - - int i; - - /* - * Get a new, free batchbuffer. - */ - - batch->size = batch->intel->intelScreen->maxBatchSize; - driBOData(batch->buffer, batch->size, NULL, 0); - - driBOResetList(&batch->list); - - /* - * Unreference buffers previously on the relocation list. - */ - - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - driBOUnReference(r->buf); - } - - batch->list_count = 0; - batch->nr_relocs = 0; - batch->flags = 0; - - /* - * We don't refcount the batchbuffer itself since we can't destroy it - * while it's on the list. - */ - - - driBOAddListItem(&batch->list, batch->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE); - - - batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); - batch->ptr = batch->map; -} - -/*====================================================================== - * Public functions - */ -struct intel_batchbuffer * -intel_batchbuffer_alloc(struct intel_context *intel) -{ - struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); - - batch->intel = intel; - - driGenBuffers(intel->intelScreen->batchPool, "batchbuffer", 1, - &batch->buffer, 4096, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); - batch->last_fence = NULL; - driBOCreateList(20, &batch->list); - intel_batchbuffer_reset(batch); - return batch; -} - -void -intel_batchbuffer_free(struct intel_batchbuffer *batch) -{ - if (batch->last_fence) { - driFenceFinish(batch->last_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); - driFenceUnReference(batch->last_fence); - batch->last_fence = NULL; - } - if (batch->map) { - driBOUnmap(batch->buffer); - batch->map = NULL; - } - driBOUnReference(batch->buffer); - batch->buffer = NULL; - free(batch); -} - -/* TODO: Push this whole function into bufmgr. - */ -static void -do_flush_locked(struct intel_batchbuffer *batch, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) -{ - GLuint *ptr; - GLuint i; - struct intel_context *intel = batch->intel; - unsigned fenceFlags; - struct _DriFenceObject *fo; - - driBOValidateList(batch->intel->driFd, &batch->list); - - /* Apply the relocations. This nasty map indicates to me that the - * whole task should be done internally by the memory manager, and - * that dma buffers probably need to be pinned within agp space. - */ - ptr = (GLuint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, - DRM_BO_HINT_ALLOW_UNFENCED_MAP); - - - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - - ptr[r->offset / 4] = driBOOffset(r->buf) + r->delta; - } - - if (INTEL_DEBUG & DEBUG_BATCH) - intel_dump_batchbuffer(0, ptr, used); - - driBOUnmap(batch->buffer); - batch->map = NULL; - - /* Throw away non-effective packets. Won't work once we have - * hardware contexts which would preserve statechanges beyond a - * single buffer. - */ - - if (!(intel->numClipRects == 0 && !ignore_cliprects)) { - intel_batch_ioctl(batch->intel, - driBOOffset(batch->buffer), - used, ignore_cliprects, allow_unlock); - } - - - /* - * Kernel fencing. The flags tells the kernel that we've - * programmed an MI_FLUSH. - */ - - fenceFlags = DRM_I915_FENCE_FLAG_FLUSHED; - fo = driFenceBuffers(batch->intel->driFd, - "Batch fence", fenceFlags); - - /* - * User space fencing. - */ - - driBOFence(batch->buffer, fo); - - if (driFenceType(fo) == DRM_FENCE_TYPE_EXE) { - - /* - * Oops. We only validated a batch buffer. This means we - * didn't do any proper rendering. Discard this fence object. - */ - - driFenceUnReference(fo); - } else { - driFenceUnReference(batch->last_fence); - batch->last_fence = fo; - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - driBOFence(r->buf, fo); - } - } - - if (intel->numClipRects == 0 && !ignore_cliprects) { - if (allow_unlock) { - UNLOCK_HARDWARE(intel); - sched_yield(); - LOCK_HARDWARE(intel); - } - intel->vtbl.lost_hardware(intel); - } -} - - -struct _DriFenceObject * -intel_batchbuffer_flush(struct intel_batchbuffer *batch) -{ - struct intel_context *intel = batch->intel; - GLuint used = batch->ptr - batch->map; - GLboolean was_locked = intel->locked; - - if (used == 0) - return batch->last_fence; - - /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a - * performance drain that we would like to avoid. - */ - if (used & 4) { - ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->ptr)[1] = 0; - ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; - used += 8; - } - - driBOUnmap(batch->buffer); - batch->ptr = NULL; - batch->map = NULL; - - /* TODO: Just pass the relocation list and dma buffer up to the - * kernel. - */ - if (!was_locked) - LOCK_HARDWARE(intel); - - do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), - GL_FALSE); - - if (!was_locked) - UNLOCK_HARDWARE(intel); - - /* Reset the buffer: - */ - intel_batchbuffer_reset(batch); - return batch->last_fence; -} - -void -intel_batchbuffer_finish(struct intel_batchbuffer *batch) -{ - struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); - driFenceReference(fence); - driFenceFinish(fence, 3, GL_FALSE); - driFenceUnReference(fence); -} - - -/* This is the only way buffers get added to the validate list. - */ -GLboolean -intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, - struct _DriBufferObject *buffer, - GLuint flags, GLuint mask, GLuint delta) -{ - assert(batch->nr_relocs < MAX_RELOCS); - - driBOAddListItem(&batch->list, buffer, flags, mask); - - { - struct buffer_reloc *r = &batch->reloc[batch->nr_relocs++]; - driBOReference(buffer); - r->buf = buffer; - r->offset = batch->ptr - batch->map; - r->delta = delta; - } - - batch->ptr += 4; - return GL_TRUE; -} - - - -void -intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags) -{ - assert((bytes & 3) == 0); - intel_batchbuffer_require_space(batch, bytes, flags); - __memcpy(batch->ptr, data, bytes); - batch->ptr += bytes; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h deleted file mode 100644 index 212f130101..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h +++ /dev/null @@ -1,124 +0,0 @@ -#ifndef INTEL_BATCHBUFFER_H -#define INTEL_BATCHBUFFER_H - -#include "mtypes.h" -#include "dri_bufmgr.h" - -struct intel_context; - -#define BATCH_SZ 16384 -#define BATCH_RESERVED 16 - -#define MAX_RELOCS 4096 - -#define INTEL_BATCH_NO_CLIPRECTS 0x1 -#define INTEL_BATCH_CLIPRECTS 0x2 - -struct buffer_reloc -{ - struct _DriBufferObject *buf; - GLuint offset; - GLuint delta; /* not needed? */ -}; - -struct intel_batchbuffer -{ - struct bufmgr *bm; - struct intel_context *intel; - - struct _DriBufferObject *buffer; - struct _DriFenceObject *last_fence; - GLuint flags; - - drmBOList list; - GLuint list_count; - GLubyte *map; - GLubyte *ptr; - - struct buffer_reloc reloc[MAX_RELOCS]; - GLuint nr_relocs; - GLuint size; -}; - -struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context - *intel); - -void intel_batchbuffer_free(struct intel_batchbuffer *batch); - - -void intel_batchbuffer_finish(struct intel_batchbuffer *batch); - -struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer - *batch); - -void intel_batchbuffer_reset(struct intel_batchbuffer *batch); - - -/* Unlike bmBufferData, this currently requires the buffer be mapped. - * Consider it a convenience function wrapping multple - * intel_buffer_dword() calls. - */ -void intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags); - -void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, - GLuint bytes); - -GLboolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, - struct _DriBufferObject *buffer, - GLuint flags, - GLuint mask, GLuint offset); - -/* Inline functions - might actually be better off with these - * non-inlined. Certainly better off switching all command packets to - * be passed as structs rather than dwords, but that's a little bit of - * work... - */ -static INLINE GLuint -intel_batchbuffer_space(struct intel_batchbuffer *batch) -{ - return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); -} - - -static INLINE void -intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) -{ - assert(batch->map); - assert(intel_batchbuffer_space(batch) >= 4); - *(GLuint *) (batch->ptr) = dword; - batch->ptr += 4; -} - -static INLINE void -intel_batchbuffer_require_space(struct intel_batchbuffer *batch, - GLuint sz, GLuint flags) -{ - assert(sz < batch->size - 8); - if (intel_batchbuffer_space(batch) < sz || - (batch->flags != 0 && flags != 0 && batch->flags != flags)) - intel_batchbuffer_flush(batch); - - batch->flags |= flags; -} - -/* Here are the crusty old macros, to be removed: - */ -#define BATCH_LOCALS - -#define BEGIN_BATCH(n, flags) do { \ - assert(!intel->prim.flush); \ - intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ -} while (0) - -#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) - -#define OUT_RELOC(buf,flags,mask,delta) do { \ - assert((delta) >= 0); \ - intel_batchbuffer_emit_reloc(intel->batch, buf, flags, mask, delta); \ -} while (0) - -#define ADVANCE_BATCH() do { } while(0) - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_batchpool.c b/src/mesa/drivers/dri/i915tex/intel_batchpool.c deleted file mode 100644 index 2503b8a62a..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_batchpool.c +++ /dev/null @@ -1,418 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#include -#include -#include -#include "imports.h" -#include "glthread.h" -#include "dri_bufpool.h" -#include "dri_bufmgr.h" -#include "intel_screen.h" - -typedef struct -{ - drmMMListHead head; - struct _BPool *parent; - struct _DriFenceObject *fence; - unsigned long start; - int unfenced; - int mapped; -} BBuf; - -typedef struct _BPool -{ - _glthread_Mutex mutex; - unsigned long bufSize; - unsigned poolSize; - unsigned numFree; - unsigned numTot; - unsigned numDelayed; - unsigned checkDelayed; - drmMMListHead free; - drmMMListHead delayed; - drmMMListHead head; - drmBO kernelBO; - void *virtual; - BBuf *bufs; -} BPool; - - -static BPool * -createBPool(int fd, unsigned long bufSize, unsigned numBufs, unsigned flags, - unsigned checkDelayed) -{ - BPool *p = (BPool *) malloc(sizeof(*p)); - BBuf *buf; - int i; - - if (!p) - return NULL; - - p->bufs = (BBuf *) malloc(numBufs * sizeof(*p->bufs)); - if (!p->bufs) { - free(p); - return NULL; - } - - DRMINITLISTHEAD(&p->free); - DRMINITLISTHEAD(&p->head); - DRMINITLISTHEAD(&p->delayed); - - p->numTot = numBufs; - p->numFree = numBufs; - p->bufSize = bufSize; - p->numDelayed = 0; - p->checkDelayed = checkDelayed; - - _glthread_INIT_MUTEX(p->mutex); - - if (drmBOCreate(fd, 0, numBufs * bufSize, 0, NULL, drm_bo_type_dc, - flags, DRM_BO_HINT_DONT_FENCE, &p->kernelBO)) { - free(p->bufs); - free(p); - return NULL; - } - if (drmBOMap(fd, &p->kernelBO, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, - &p->virtual)) { - drmBODestroy(fd, &p->kernelBO); - free(p->bufs); - free(p); - return NULL; - } - - /* - * We unmap the buffer so that we can validate it later. Note that this is - * just a synchronizing operation. The buffer will have a virtual mapping - * until it is destroyed. - */ - - drmBOUnmap(fd, &p->kernelBO); - - buf = p->bufs; - for (i = 0; i < numBufs; ++i) { - buf->parent = p; - buf->fence = NULL; - buf->start = i * bufSize; - buf->mapped = 0; - buf->unfenced = 0; - DRMLISTADDTAIL(&buf->head, &p->free); - buf++; - } - - return p; -} - - -static void -pool_checkFree(BPool * p, int wait) -{ - drmMMListHead *list, *prev; - BBuf *buf; - int signaled = 0; - int i; - - list = p->delayed.next; - - if (p->numDelayed > 3) { - for (i = 0; i < p->numDelayed; i += 3) { - list = list->next; - } - } - - prev = list->prev; - for (; list != &p->delayed; list = prev, prev = list->prev) { - - buf = DRMLISTENTRY(BBuf, list, head); - - if (!signaled) { - if (wait) { - driFenceFinish(buf->fence, DRM_FENCE_TYPE_EXE, 1); - signaled = 1; - } - else { - signaled = driFenceSignaled(buf->fence, DRM_FENCE_TYPE_EXE); - } - } - - if (!signaled) - break; - - driFenceUnReference(buf->fence); - buf->fence = NULL; - DRMLISTDEL(list); - p->numDelayed--; - DRMLISTADD(list, &p->free); - p->numFree++; - } -} - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, unsigned flags, unsigned hint, - unsigned alignment) -{ - BPool *p = (BPool *) pool->data; - - drmMMListHead *item; - - if (alignment && (alignment != 4096)) - return NULL; - - _glthread_LOCK_MUTEX(p->mutex); - - if (p->numFree == 0) - pool_checkFree(p, GL_TRUE); - - if (p->numFree == 0) { - fprintf(stderr, "Out of fixed size buffer objects\n"); - BM_CKFATAL(-ENOMEM); - } - - item = p->free.next; - - if (item == &p->free) { - fprintf(stderr, "Fixed size buffer pool corruption\n"); - } - - DRMLISTDEL(item); - --p->numFree; - - _glthread_UNLOCK_MUTEX(p->mutex); - return (void *) DRMLISTENTRY(BBuf, item, head); -} - - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - - if (buf->fence) { - DRMLISTADDTAIL(&buf->head, &p->delayed); - p->numDelayed++; - } - else { - buf->unfenced = 0; - DRMLISTADD(&buf->head, &p->free); - p->numFree++; - } - - if ((p->numDelayed % p->checkDelayed) == 0) - pool_checkFree(p, 0); - - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, void **virtual) -{ - - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - - /* - * Currently Mesa doesn't have any condition variables to resolve this - * cleanly in a multithreading environment. - * We bail out instead. - */ - - if (buf->mapped) { - fprintf(stderr, "Trying to map already mapped buffer object\n"); - BM_CKFATAL(-EINVAL); - } - -#if 0 - if (buf->unfenced && !(hint & DRM_BO_HINT_ALLOW_UNFENCED_MAP)) { - fprintf(stderr, "Trying to map an unfenced buffer object 0x%08x" - " 0x%08x %d\n", hint, flags, buf->start); - BM_CKFATAL(-EINVAL); - } - -#endif - - if (buf->fence) { - _glthread_UNLOCK_MUTEX(p->mutex); - return -EBUSY; - } - - buf->mapped = GL_TRUE; - *virtual = (unsigned char *) p->virtual + buf->start; - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, int lazy) -{ - BBuf *buf = (BBuf *) private; - driFenceFinish(buf->fence, 0, lazy); - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - - buf->mapped = 0; - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - return p->kernelBO.offset + buf->start; -} - -static unsigned -pool_flags(struct _DriBufferPool *pool, void *private) -{ - BPool *p = (BPool *) pool->data; - - return p->kernelBO.flags; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - BPool *p = (BPool *) pool->data; - - return p->bufSize; -} - - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - if (buf->fence) { - driFenceUnReference(buf->fence); - } - buf->fence = fence; - buf->unfenced = 0; - driFenceReference(buf->fence); - _glthread_UNLOCK_MUTEX(p->mutex); - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - return &p->kernelBO; -} - -static int -pool_validate(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - _glthread_LOCK_MUTEX(p->mutex); - buf->unfenced = GL_TRUE; - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - BPool *p = (BPool *) pool->data; - - /* - * Wait on outstanding fences. - */ - - _glthread_LOCK_MUTEX(p->mutex); - while ((p->numFree < p->numTot) && p->numDelayed) { - _glthread_UNLOCK_MUTEX(p->mutex); - sched_yield(); - pool_checkFree(p, GL_TRUE); - _glthread_LOCK_MUTEX(p->mutex); - } - - drmBODestroy(pool->fd, &p->kernelBO); - free(p->bufs); - _glthread_UNLOCK_MUTEX(p->mutex); - free(p); - free(pool); -} - - -struct _DriBufferPool * -driBatchPoolInit(int fd, unsigned flags, - unsigned long bufSize, - unsigned numBufs, unsigned checkDelayed) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - if (!pool) - return NULL; - - pool->data = createBPool(fd, bufSize, numBufs, flags, checkDelayed); - if (!pool->data) - return NULL; - - pool->fd = fd; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = &pool_validate; - pool->waitIdle = &pool_waitIdle; - pool->setstatic = NULL; - pool->takeDown = &pool_takedown; - return pool; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.c b/src/mesa/drivers/dri/i915tex/intel_blit.c deleted file mode 100644 index 28441f4b83..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_blit.c +++ /dev/null @@ -1,527 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include -#include - -#include "mtypes.h" -#include "context.h" -#include "enums.h" - -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "intel_buffers.h" -#include "intel_context.h" -#include "intel_fbo.h" -#include "intel_reg.h" -#include "intel_regions.h" -#include "vblank.h" - -#define FILE_DEBUG_FLAG DEBUG_BLIT - -/** - * Copy the back color buffer to the front color buffer. - * Used for SwapBuffers(). - */ -void -intelCopyBuffer(__DRIdrawablePrivate * dPriv, - const drm_clip_rect_t * rect) -{ - - struct intel_context *intel; - const intelScreenPrivate *intelScreen; - - DBG("%s\n", __FUNCTION__); - - assert(dPriv); - - intel = intelScreenContext(dPriv->driScreenPriv->private); - if (!intel) - return; - - intelScreen = intel->intelScreen; - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - current context (which is what intelScreenContext should return) might - not get a contended lock and thus cliprects not updated (tests/manywin) */ - if ((struct intel_context *)dPriv->driContextPriv->driverPrivate != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - - if (dPriv && dPriv->numClipRects) { - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - const struct intel_region *frontRegion - = intelScreen->front_region; - const struct intel_region *backRegion - = intel_fb->Base._ColorDrawBufferMask[0] == BUFFER_BIT_FRONT_LEFT ? - intel_get_rb_region(&intel_fb->Base, BUFFER_FRONT_LEFT) : - intel_get_rb_region(&intel_fb->Base, BUFFER_BACK_LEFT); - const int backWidth = intel_fb->Base.Width; - const int backHeight = intel_fb->Base.Height; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = frontRegion->pitch; - const int srcpitch = backRegion->pitch; - const int cpp = frontRegion->cpp; - int BR13, CMD; - int i; - - ASSERT(intel_fb); - ASSERT(intel_fb->Base.Name == 0); /* Not a user-created FBO */ - ASSERT(frontRegion); - ASSERT(backRegion); - ASSERT(frontRegion->cpp == backRegion->cpp); - - DBG("front pitch %d back pitch %d\n", - frontRegion->pitch, backRegion->pitch); - - if (cpp == 2) { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - } - else { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - } - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->width || pbox->y2 > intelScreen->height) - continue; - - box = *pbox; - - if (rect) { - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > backWidth) - box.x2 = backWidth + box.x1; - if (box.y2 - box.y1 > backHeight) - box.y2 = backHeight + box.y1; - - DBG("box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((box.y1 << 16) | box.x1); - OUT_BATCH((box.y2 << 16) | box.x2); - - OUT_RELOC(frontRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((sbox.y1 << 16) | sbox.x1); - OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(backRegion->buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - ADVANCE_BATCH(); - } - - if (intel->first_swap_fence) - driFenceUnReference(intel->first_swap_fence); - intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); - driFenceReference(intel->first_swap_fence); - } - - UNLOCK_HARDWARE(intel); - - /* XXX this is bogus. The context here may not even be bound to this drawable! */ - if (intel->lastStamp != dPriv->lastStamp) { - GET_CURRENT_CONTEXT(currctx); - struct intel_context *intelcurrent = intel_context(currctx); - if (intelcurrent == intel && intelcurrent->driDrawable == dPriv) { - intelWindowMoved(intel); - intel->lastStamp = dPriv->lastStamp; - } - } - -} - - - - -void -intelEmitFillBlit(struct intel_context *intel, - GLuint cpp, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort x, GLshort y, GLshort w, GLshort h, GLuint color) -{ - GLuint BR13, CMD; - BATCH_LOCALS; - - dst_pitch *= cpp; - - switch (cpp) { - case 1: - case 2: - case 3: - BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); - CMD = XY_COLOR_BLT_CMD; - break; - case 4: - BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); - CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); - break; - default: - return; - } - - DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); - - - BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((y << 16) | x); - OUT_BATCH(((y + h) << 16) | (x + w)); - OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); - OUT_BATCH(color); - ADVANCE_BATCH(); -} - - -static GLuint translate_raster_op(GLenum logicop) -{ - switch(logicop) { - case GL_CLEAR: return 0x00; - case GL_AND: return 0x88; - case GL_AND_REVERSE: return 0x44; - case GL_COPY: return 0xCC; - case GL_AND_INVERTED: return 0x22; - case GL_NOOP: return 0xAA; - case GL_XOR: return 0x66; - case GL_OR: return 0xEE; - case GL_NOR: return 0x11; - case GL_EQUIV: return 0x99; - case GL_INVERT: return 0x55; - case GL_OR_REVERSE: return 0xDD; - case GL_COPY_INVERTED: return 0x33; - case GL_OR_INVERTED: return 0xBB; - case GL_NAND: return 0x77; - case GL_SET: return 0xFF; - default: return 0; - } -} - - -/* Copy BitBlt - */ -void -intelEmitCopyBlit(struct intel_context *intel, - GLuint cpp, - GLshort src_pitch, - struct _DriBufferObject *src_buffer, - GLuint src_offset, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort src_x, GLshort src_y, - GLshort dst_x, GLshort dst_y, - GLshort w, GLshort h, - GLenum logic_op) -{ - GLuint CMD, BR13; - int dst_y2 = dst_y + h; - int dst_x2 = dst_x + w; - BATCH_LOCALS; - - - DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, - src_buffer, src_pitch, src_offset, src_x, src_y, - dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); - - src_pitch *= cpp; - dst_pitch *= cpp; - - switch (cpp) { - case 1: - case 2: - case 3: - BR13 = (((GLint) dst_pitch) & 0xffff) | - (translate_raster_op(logic_op) << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - break; - case 4: - BR13 = - (((GLint) dst_pitch) & 0xffff) | - (translate_raster_op(logic_op) << 16) | (1 << 24) | (1 << 25); - CMD = - (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - break; - default: - return; - } - - if (dst_y2 < dst_y || dst_x2 < dst_x) { - return; - } - - /* Initial y values don't seem to work with negative pitches. If - * we adjust the offsets manually (below), it seems to work fine. - * - * On the other hand, if we always adjust, the hardware doesn't - * know which blit directions to use, so overlapping copypixels get - * the wrong result. - */ - if (dst_pitch > 0 && src_pitch > 0) { - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((dst_y << 16) | dst_x); - OUT_BATCH((dst_y2 << 16) | dst_x2); - OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, dst_offset); - OUT_BATCH((src_y << 16) | src_x); - OUT_BATCH(((GLint) src_pitch & 0xffff)); - OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, src_offset); - ADVANCE_BATCH(); - } - else { - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((0 << 16) | dst_x); - OUT_BATCH((h << 16) | dst_x2); - OUT_RELOC(dst_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - dst_offset + dst_y * dst_pitch); - OUT_BATCH((0 << 16) | src_x); - OUT_BATCH(((GLint) src_pitch & 0xffff)); - OUT_RELOC(src_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, - src_offset + src_y * src_pitch); - ADVANCE_BATCH(); - } -} - - -/** - * Use blitting to clear the renderbuffers named by 'flags'. - * Note: we can't use the ctx->DrawBuffer->_ColorDrawBufferMask field - * since that might include software renderbuffers or renderbuffers - * which we're clearing with triangles. - * \param mask bitmask of BUFFER_BIT_* values indicating buffers to clear - */ -void -intelClearWithBlit(GLcontext * ctx, GLbitfield mask) -{ - struct intel_context *intel = intel_context(ctx); - struct gl_framebuffer *fb = ctx->DrawBuffer; - GLuint clear_depth; - GLbitfield skipBuffers = 0; - BATCH_LOCALS; - - DBG("%s %x\n", __FUNCTION__, mask); - - /* - * Compute values for clearing the buffers. - */ - clear_depth = 0; - if (mask & BUFFER_BIT_DEPTH) { - clear_depth = (GLuint) (fb->_DepthMax * ctx->Depth.Clear); - } - if (mask & BUFFER_BIT_STENCIL) { - clear_depth |= (ctx->Stencil.Clear & 0xff) << 24; - } - - /* If clearing both depth and stencil, skip BUFFER_BIT_STENCIL in - * the loop below. - */ - if ((mask & BUFFER_BIT_DEPTH) && (mask & BUFFER_BIT_STENCIL)) { - skipBuffers = BUFFER_BIT_STENCIL; - } - - /* XXX Move this flush/lock into the following conditional? */ - intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); - - if (intel->numClipRects) { - GLint cx, cy, cw, ch; - drm_clip_rect_t b; - - /* Get clear bounds after locking */ - cx = fb->_Xmin; - cy = fb->_Ymin; - cw = fb->_Xmax - cx; - ch = fb->_Ymax - cy; - - if (fb->Name == 0) { - /* clearing a window */ - /* flip top to bottom */ - b.x1 = cx; - b.y1 = fb->Height - cy - ch; - b.x2 = b.x1 + cw; - b.y2 = b.y1 + ch; - } - else { - /* clearing FBO */ - b.x1 = cx; - b.y1 = cy; - b.x2 = b.x1 + cw; - b.y2 = b.y1 + ch; - /* no change to mask */ - } - - { - GLuint buf; - GLuint clearMask = mask; /* use copy, since we modify it below */ - GLboolean all = (cw == fb->Width && ch == fb->Height); - - DBG("clear %d,%d..%d,%d, mask %x\n", - b.x1, b.y1, b.x2, b.y2, mask); - - /* Loop over all renderbuffers */ - for (buf = 0; buf < BUFFER_COUNT && clearMask; buf++) { - const GLbitfield bufBit = 1 << buf; - if ((clearMask & bufBit) && !(bufBit & skipBuffers)) { - /* OK, clear this renderbuffer */ - struct intel_region *irb_region = - intel_get_rb_region(fb, buf); - struct _DriBufferObject *write_buffer = - intel_region_buffer(intel->intelScreen, irb_region, - all ? INTEL_WRITE_FULL : - INTEL_WRITE_PART); - - GLuint clearVal; - GLint pitch, cpp; - GLuint BR13, CMD; - - ASSERT(irb_region); - - pitch = irb_region->pitch; - cpp = irb_region->cpp; - - DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, - irb_region->buffer, (pitch * cpp), - irb_region->draw_offset, - b.x1, b.y1, b.x2 - b.x1, b.y2 - b.y1); - - - /* Setup the blit command */ - if (cpp == 4) { - BR13 = (0xF0 << 16) | (pitch * cpp) | (1 << 24) | (1 << 25); - if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) { - CMD = XY_COLOR_BLT_CMD; - if (clearMask & BUFFER_BIT_DEPTH) - CMD |= XY_COLOR_BLT_WRITE_RGB; - if (clearMask & BUFFER_BIT_STENCIL) - CMD |= XY_COLOR_BLT_WRITE_ALPHA; - } - else { - /* clearing RGBA */ - CMD = (XY_COLOR_BLT_CMD | - XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); - } - } - else { - ASSERT(cpp == 2 || cpp == 0); - BR13 = (0xF0 << 16) | (pitch * cpp) | (1 << 24); - CMD = XY_COLOR_BLT_CMD; - } - - if (buf == BUFFER_DEPTH || buf == BUFFER_STENCIL) { - clearVal = clear_depth; - } - else { - clearVal = (cpp == 4) - ? intel->ClearColor8888 : intel->ClearColor565; - } - /* - _mesa_debug(ctx, "hardware blit clear buf %d rb id %d\n", - buf, irb->Base.Name); - */ - intel_wait_flips(intel, INTEL_BATCH_NO_CLIPRECTS); - - BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((b.y1 << 16) | b.x1); - OUT_BATCH((b.y2 << 16) | b.x2); - OUT_RELOC(write_buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, - irb_region->draw_offset); - OUT_BATCH(clearVal); - ADVANCE_BATCH(); - clearMask &= ~bufBit; /* turn off bit, for faster loop exit */ - } - } - } - intel_batchbuffer_flush(intel->batch); - } - - UNLOCK_HARDWARE(intel); -} diff --git a/src/mesa/drivers/dri/i915tex/intel_blit.h b/src/mesa/drivers/dri/i915tex/intel_blit.h deleted file mode 100644 index 77686444fa..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_blit.h +++ /dev/null @@ -1,62 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BLIT_H -#define INTEL_BLIT_H - -#include "intel_context.h" -#include "intel_ioctl.h" -#include "dri_bufmgr.h" - -extern void intelCopyBuffer(__DRIdrawablePrivate * dpriv, - const drm_clip_rect_t * rect); - -extern void intelClearWithBlit(GLcontext * ctx, GLbitfield mask); - -extern void intelEmitCopyBlit(struct intel_context *intel, - GLuint cpp, - GLshort src_pitch, - struct _DriBufferObject *src_buffer, - GLuint src_offset, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort srcx, GLshort srcy, - GLshort dstx, GLshort dsty, - GLshort w, GLshort h, - GLenum logicop ); - -extern void intelEmitFillBlit(struct intel_context *intel, - GLuint cpp, - GLshort dst_pitch, - struct _DriBufferObject *dst_buffer, - GLuint dst_offset, - GLshort x, GLshort y, - GLshort w, GLshort h, GLuint color); - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_buffer_objects.c b/src/mesa/drivers/dri/i915tex/intel_buffer_objects.c deleted file mode 100644 index 91c45ad95b..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_buffer_objects.c +++ /dev/null @@ -1,257 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "imports.h" -#include "mtypes.h" -#include "bufferobj.h" - -#include "intel_context.h" -#include "intel_buffer_objects.h" -#include "intel_regions.h" -#include "dri_bufmgr.h" - -/** - * There is some duplication between mesa's bufferobjects and our - * bufmgr buffers. Both have an integer handle and a hashtable to - * lookup an opaque structure. It would be nice if the handles and - * internal structure where somehow shared. - */ -static struct gl_buffer_object * -intel_bufferobj_alloc(GLcontext * ctx, GLuint name, GLenum target) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_buffer_object *obj = CALLOC_STRUCT(intel_buffer_object); - - _mesa_initialize_buffer_object(&obj->Base, name, target); - - driGenBuffers(intel->intelScreen->regionPool, - "bufferobj", 1, &obj->buffer, 64, - DRM_BO_FLAG_MEM_LOCAL | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0); - - return &obj->Base; -} - - -/* Break the COW tie to the region. The region gets to keep the data. - */ -void -intel_bufferobj_release_region(struct intel_context *intel, - struct intel_buffer_object *intel_obj) -{ - assert(intel_obj->region->buffer == intel_obj->buffer); - intel_obj->region->pbo = NULL; - intel_obj->region = NULL; - driBOUnReference(intel_obj->buffer); - intel_obj->buffer = NULL; - - /* This leads to a large number of buffer deletion/creation events. - * Currently the drm doesn't like that: - */ - driGenBuffers(intel->intelScreen->regionPool, - "buffer object", 1, &intel_obj->buffer, 64, 0, 0); - LOCK_HARDWARE(intel); - driBOData(intel_obj->buffer, intel_obj->Base.Size, NULL, 0); - UNLOCK_HARDWARE(intel); -} - -/* Break the COW tie to the region. Both the pbo and the region end - * up with a copy of the data. - */ -void -intel_bufferobj_cow(struct intel_context *intel, - struct intel_buffer_object *intel_obj) -{ - assert(intel_obj->region); - intel_region_cow(intel->intelScreen, intel_obj->region); -} - - -/** - * Deallocate/free a vertex/pixel buffer object. - * Called via glDeleteBuffersARB(). - */ -static void -intel_bufferobj_free(GLcontext * ctx, struct gl_buffer_object *obj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - assert(intel_obj); - - if (intel_obj->region) { - intel_bufferobj_release_region(intel, intel_obj); - } - else if (intel_obj->buffer) { - driDeleteBuffers(1, &intel_obj->buffer); - } - - _mesa_free(intel_obj); -} - - - -/** - * Allocate space for and store data in a buffer object. Any data that was - * previously stored in the buffer object is lost. If data is NULL, - * memory will be allocated, but no copy will occur. - * Called via glBufferDataARB(). - */ -static void -intel_bufferobj_data(GLcontext * ctx, - GLenum target, - GLsizeiptrARB size, - const GLvoid * data, - GLenum usage, struct gl_buffer_object *obj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - intel_obj->Base.Size = size; - intel_obj->Base.Usage = usage; - - if (intel_obj->region) - intel_bufferobj_release_region(intel, intel_obj); - - LOCK_HARDWARE(intel); - driBOData(intel_obj->buffer, size, data, 0); - UNLOCK_HARDWARE(intel); -} - - -/** - * Replace data in a subrange of buffer object. If the data range - * specified by size + offset extends beyond the end of the buffer or - * if data is NULL, no copy is performed. - * Called via glBufferSubDataARB(). - */ -static void -intel_bufferobj_subdata(GLcontext * ctx, - GLenum target, - GLintptrARB offset, - GLsizeiptrARB size, - const GLvoid * data, struct gl_buffer_object *obj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - assert(intel_obj); - - if (intel_obj->region) - intel_bufferobj_cow(intel, intel_obj); - - driBOSubData(intel_obj->buffer, offset, size, data); -} - - -/** - * Called via glGetBufferSubDataARB(). - */ -static void -intel_bufferobj_get_subdata(GLcontext * ctx, - GLenum target, - GLintptrARB offset, - GLsizeiptrARB size, - GLvoid * data, struct gl_buffer_object *obj) -{ - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - assert(intel_obj); - driBOGetSubData(intel_obj->buffer, offset, size, data); -} - - - -/** - * Called via glMapBufferARB(). - */ -static void * -intel_bufferobj_map(GLcontext * ctx, - GLenum target, - GLenum access, struct gl_buffer_object *obj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - /* XXX: Translate access to flags arg below: - */ - assert(intel_obj); - - if (intel_obj->region) - intel_bufferobj_cow(intel, intel_obj); - - obj->Pointer = driBOMap(intel_obj->buffer, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - return obj->Pointer; -} - - -/** - * Called via glMapBufferARB(). - */ -static GLboolean -intel_bufferobj_unmap(GLcontext * ctx, - GLenum target, struct gl_buffer_object *obj) -{ - struct intel_buffer_object *intel_obj = intel_buffer_object(obj); - - assert(intel_obj); - assert(obj->Pointer); - driBOUnmap(intel_obj->buffer); - obj->Pointer = NULL; - return GL_TRUE; -} - -struct _DriBufferObject * -intel_bufferobj_buffer(struct intel_context *intel, - struct intel_buffer_object *intel_obj, GLuint flag) -{ - if (intel_obj->region) { - if (flag == INTEL_WRITE_PART) - intel_bufferobj_cow(intel, intel_obj); - else if (flag == INTEL_WRITE_FULL) - intel_bufferobj_release_region(intel, intel_obj); - } - - return intel_obj->buffer; -} - -void -intel_bufferobj_init(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - - ctx->Driver.NewBufferObject = intel_bufferobj_alloc; - ctx->Driver.DeleteBuffer = intel_bufferobj_free; - ctx->Driver.BufferData = intel_bufferobj_data; - ctx->Driver.BufferSubData = intel_bufferobj_subdata; - ctx->Driver.GetBufferSubData = intel_bufferobj_get_subdata; - ctx->Driver.MapBuffer = intel_bufferobj_map; - ctx->Driver.UnmapBuffer = intel_bufferobj_unmap; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_buffer_objects.h b/src/mesa/drivers/dri/i915tex/intel_buffer_objects.h deleted file mode 100644 index afe9b2f7cf..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_buffer_objects.h +++ /dev/null @@ -1,86 +0,0 @@ - /************************************************************************** - * - * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BUFFEROBJ_H -#define INTEL_BUFFEROBJ_H - -#include "mtypes.h" - -struct intel_context; -struct intel_region; -struct gl_buffer_object; - - -/** - * Intel vertex/pixel buffer object, derived from Mesa's gl_buffer_object. - */ -struct intel_buffer_object -{ - struct gl_buffer_object Base; - struct _DriBufferObject *buffer; /* the low-level buffer manager's buffer handle */ - - struct intel_region *region; /* Is there a zero-copy texture - associated with this (pixel) - buffer object? */ -}; - - -/* Get the bm buffer associated with a GL bufferobject: - */ -struct _DriBufferObject *intel_bufferobj_buffer(struct intel_context *intel, - struct intel_buffer_object - *obj, GLuint flag); - -/* Hook the bufferobject implementation into mesa: - */ -void intel_bufferobj_init(struct intel_context *intel); - - - -/* Are the obj->Name tests necessary? Unfortunately yes, mesa - * allocates a couple of gl_buffer_object structs statically, and - * the Name == 0 test is the only way to identify them and avoid - * casting them erroneously to our structs. - */ -static INLINE struct intel_buffer_object * -intel_buffer_object(struct gl_buffer_object *obj) -{ - if (obj->Name) - return (struct intel_buffer_object *) obj; - else - return NULL; -} - -/* Helpers for zerocopy image uploads. See also intel_regions.h: - */ -void intel_bufferobj_cow(struct intel_context *intel, - struct intel_buffer_object *intel_obj); -void intel_bufferobj_release_region(struct intel_context *intel, - struct intel_buffer_object *intel_obj); - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.c b/src/mesa/drivers/dri/i915tex/intel_buffers.c deleted file mode 100644 index 8944145afa..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.c +++ /dev/null @@ -1,890 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_blit.h" -#include "intel_buffers.h" -#include "intel_depthstencil.h" -#include "intel_fbo.h" -#include "intel_tris.h" -#include "intel_regions.h" -#include "intel_batchbuffer.h" -#include "intel_reg.h" -#include "context.h" -#include "utils.h" -#include "drirenderbuffer.h" -#include "framebuffer.h" -#include "swrast/swrast.h" -#include "vblank.h" - - -/* This block can be removed when libdrm >= 2.3.1 is required */ - -#ifndef DRM_VBLANK_FLIP - -#define DRM_VBLANK_FLIP 0x8000000 - -typedef struct drm_i915_flip { - int pipes; -} drm_i915_flip_t; - -#undef DRM_IOCTL_I915_FLIP -#define DRM_IOCTL_I915_FLIP DRM_IOW(DRM_COMMAND_BASE + DRM_I915_FLIP, \ - drm_i915_flip_t) - -#endif - - -/** - * XXX move this into a new dri/common/cliprects.c file. - */ -GLboolean -intel_intersect_cliprects(drm_clip_rect_t * dst, - const drm_clip_rect_t * a, - const drm_clip_rect_t * b) -{ - GLint bx = b->x1; - GLint by = b->y1; - GLint bw = b->x2 - bx; - GLint bh = b->y2 - by; - - if (bx < a->x1) - bw -= a->x1 - bx, bx = a->x1; - if (by < a->y1) - bh -= a->y1 - by, by = a->y1; - if (bx + bw > a->x2) - bw = a->x2 - bx; - if (by + bh > a->y2) - bh = a->y2 - by; - if (bw <= 0) - return GL_FALSE; - if (bh <= 0) - return GL_FALSE; - - dst->x1 = bx; - dst->y1 = by; - dst->x2 = bx + bw; - dst->y2 = by + bh; - - return GL_TRUE; -} - -/** - * Return pointer to current color drawing region, or NULL. - */ -struct intel_region * -intel_drawbuf_region(struct intel_context *intel) -{ - struct intel_renderbuffer *irbColor = - intel_renderbuffer(intel->ctx.DrawBuffer->_ColorDrawBuffers[0][0]); - if (irbColor) - return irbColor->region; - else - return NULL; -} - -/** - * Return pointer to current color reading region, or NULL. - */ -struct intel_region * -intel_readbuf_region(struct intel_context *intel) -{ - struct intel_renderbuffer *irb - = intel_renderbuffer(intel->ctx.ReadBuffer->_ColorReadBuffer); - if (irb) - return irb->region; - else - return NULL; -} - - - -/** - * Update the following fields for rendering: - * intel->numClipRects - * intel->pClipRects - */ -static void -intelSetRenderbufferClipRects(struct intel_context *intel) -{ - /* zero-sized buffers might be legal? */ - assert(intel->ctx.DrawBuffer->Width > 0); - assert(intel->ctx.DrawBuffer->Height > 0); - intel->fboRect.x1 = 0; - intel->fboRect.y1 = 0; - intel->fboRect.x2 = intel->ctx.DrawBuffer->Width; - intel->fboRect.y2 = intel->ctx.DrawBuffer->Height; - intel->numClipRects = 1; - intel->pClipRects = &intel->fboRect; -} - - -/** - * This will be called whenever the currently bound window is moved/resized. - * XXX: actually, it seems to NOT be called when the window is only moved (BP). - */ -void -intelWindowMoved(struct intel_context *intel) -{ - GLcontext *ctx = &intel->ctx; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - - if (!intel->ctx.DrawBuffer) { - /* when would this happen? -BP */ - assert(0); - intel->numClipRects = 0; - } - - /* Update Mesa's notion of window size */ - driUpdateFramebufferSize(ctx, dPriv); - intel_fb->Base.Initialized = GL_TRUE; /* XXX remove someday */ - - { - drmI830Sarea *sarea = intel->sarea; - drm_clip_rect_t drw_rect = { .x1 = dPriv->x, .x2 = dPriv->x + dPriv->w, - .y1 = dPriv->y, .y2 = dPriv->y + dPriv->h }; - drm_clip_rect_t pipeA_rect = { .x1 = sarea->pipeA_x, .y1 = sarea->pipeA_y, - .x2 = sarea->pipeA_x + sarea->pipeA_w, - .y2 = sarea->pipeA_y + sarea->pipeA_h }; - drm_clip_rect_t pipeB_rect = { .x1 = sarea->pipeB_x, .y1 = sarea->pipeB_y, - .x2 = sarea->pipeB_x + sarea->pipeB_w, - .y2 = sarea->pipeB_y + sarea->pipeB_h }; - GLint areaA = driIntersectArea( drw_rect, pipeA_rect ); - GLint areaB = driIntersectArea( drw_rect, pipeB_rect ); - GLuint flags = intel_fb->vblank_flags; - GLboolean pf_active; - GLint pf_pipes; - - /* Update page flipping info - */ - pf_pipes = 0; - - if (areaA > 0) - pf_pipes |= 1; - - if (areaB > 0) - pf_pipes |= 2; - - intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3; - - intel_fb->pf_num_pages = 2 /*intel->intelScreen->third.handle ? 3 : 2*/; - - pf_active = pf_pipes && (pf_pipes & intel->sarea->pf_active) == pf_pipes; - - if (INTEL_DEBUG & DEBUG_LOCK) - if (pf_active != intel_fb->pf_active) - _mesa_printf("%s - Page flipping %sactive\n", __progname, - pf_active ? "" : "in"); - - if (pf_active) { - /* Sync pages between pipes if we're flipping on both at the same time */ - if (pf_pipes == 0x3 && pf_pipes != intel_fb->pf_pipes && - (intel->sarea->pf_current_page & 0x3) != - (((intel->sarea->pf_current_page) >> 2) & 0x3)) { - drm_i915_flip_t flip; - - if (intel_fb->pf_current_page == - (intel->sarea->pf_current_page & 0x3)) { - /* XXX: This is ugly, but emitting two flips 'in a row' can cause - * lockups for unknown reasons. - */ - intel->sarea->pf_current_page = - intel->sarea->pf_current_page & 0x3; - intel->sarea->pf_current_page |= - ((intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % - intel_fb->pf_num_pages) << 2; - - flip.pipes = 0x2; - } else { - intel->sarea->pf_current_page = - intel->sarea->pf_current_page & (0x3 << 2); - intel->sarea->pf_current_page |= - (intel_fb->pf_current_page + intel_fb->pf_num_pages - 1) % - intel_fb->pf_num_pages; - - flip.pipes = 0x1; - } - - drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); - } - - intel_fb->pf_pipes = pf_pipes; - } - - intel_fb->pf_active = pf_active; - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); - - /* Update vblank info - */ - if (areaB > areaA || (areaA == areaB && areaB > 0)) { - flags = intel_fb->vblank_flags | VBLANK_FLAG_SECONDARY; - } else { - flags = intel_fb->vblank_flags & ~VBLANK_FLAG_SECONDARY; - } - - if (flags != intel_fb->vblank_flags && intel_fb->vblank_flags && - !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ)) { - drmVBlank vbl; - int i; - - vbl.request.type = DRM_VBLANK_ABSOLUTE; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - vbl.request.type |= DRM_VBLANK_SECONDARY; - } - - for (i = 0; i < intel_fb->pf_num_pages; i++) { - if (!intel_fb->color_rb[i] || - (intel_fb->vbl_waited - intel_fb->color_rb[i]->vbl_pending) <= - (1<<23)) - continue; - - vbl.request.sequence = intel_fb->color_rb[i]->vbl_pending; - drmWaitVBlank(intel->driFd, &vbl); - } - - intel_fb->vblank_flags = flags; - driGetCurrentVBlank(dPriv, intel_fb->vblank_flags, &intel_fb->vbl_seq); - intel_fb->vbl_waited = intel_fb->vbl_seq; - - for (i = 0; i < intel_fb->pf_num_pages; i++) { - if (intel_fb->color_rb[i]) - intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_waited; - } - } - } - - /* Update hardware scissor */ - ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, - ctx->Scissor.Width, ctx->Scissor.Height); - - /* Re-calculate viewport related state */ - ctx->Driver.DepthRange( ctx, ctx->Viewport.Near, ctx->Viewport.Far ); -} - - - -/* A true meta version of this would be very simple and additionally - * machine independent. Maybe we'll get there one day. - */ -static void -intelClearWithTris(struct intel_context *intel, GLbitfield mask) -{ - GLcontext *ctx = &intel->ctx; - struct gl_framebuffer *fb = ctx->DrawBuffer; - drm_clip_rect_t clear; - - if (INTEL_DEBUG & DEBUG_BLIT) - _mesa_printf("%s 0x%x\n", __FUNCTION__, mask); - - LOCK_HARDWARE(intel); - - if (intel->numClipRects) { - GLuint buf; - - intel->vtbl.install_meta_state(intel); - - clear.x1 = fb->_Xmin; - clear.y1 = fb->_Ymin; - clear.x2 = fb->_Xmax; - clear.y2 = fb->_Ymax; - - /* Back and stencil cliprects are the same. Try and do both - * buffers at once: - */ - if (mask & - (BUFFER_BIT_BACK_LEFT | BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH)) { - struct intel_region *backRegion = - intel_get_rb_region(fb, BUFFER_BACK_LEFT); - struct intel_region *depthRegion = - intel_get_rb_region(fb, BUFFER_DEPTH); - const GLuint clearColor = (backRegion && backRegion->cpp == 4) - ? intel->ClearColor8888 : intel->ClearColor565; - - intel->vtbl.meta_draw_region(intel, backRegion, depthRegion); - - if (mask & BUFFER_BIT_BACK_LEFT) - intel->vtbl.meta_color_mask(intel, GL_TRUE); - else - intel->vtbl.meta_color_mask(intel, GL_FALSE); - - if (mask & BUFFER_BIT_STENCIL) - intel->vtbl.meta_stencil_replace(intel, - intel->ctx.Stencil.WriteMask[0], - intel->ctx.Stencil.Clear); - else - intel->vtbl.meta_no_stencil_write(intel); - - if (mask & BUFFER_BIT_DEPTH) - intel->vtbl.meta_depth_replace(intel); - else - intel->vtbl.meta_no_depth_write(intel); - - /* XXX: Using INTEL_BATCH_NO_CLIPRECTS here is dangerous as the - * drawing origin may not be correctly emitted. - */ - intel_meta_draw_quad(intel, clear.x1, clear.x2, clear.y1, clear.y2, - intel->ctx.Depth.Clear, clearColor, 0, 0, 0, 0); /* texcoords */ - - mask &= - ~(BUFFER_BIT_BACK_LEFT | BUFFER_BIT_STENCIL | BUFFER_BIT_DEPTH); - } - - /* clear the remaining (color) renderbuffers */ - for (buf = 0; buf < BUFFER_COUNT && mask; buf++) { - const GLuint bufBit = 1 << buf; - if (mask & bufBit) { - struct intel_renderbuffer *irbColor = - intel_renderbuffer(fb->Attachment[buf].Renderbuffer); - GLuint color = (irbColor->region->cpp == 4) - ? intel->ClearColor8888 : intel->ClearColor565; - - ASSERT(irbColor); - - intel->vtbl.meta_no_depth_write(intel); - intel->vtbl.meta_no_stencil_write(intel); - intel->vtbl.meta_color_mask(intel, GL_TRUE); - intel->vtbl.meta_draw_region(intel, irbColor->region, NULL); - - /* XXX: Using INTEL_BATCH_NO_CLIPRECTS here is dangerous as the - * drawing origin may not be correctly emitted. - */ - intel_meta_draw_quad(intel, clear.x1, clear.x2, clear.y1, clear.y2, 0, /* depth clear val */ - color, 0, 0, 0, 0); /* texcoords */ - - mask &= ~bufBit; - } - } - - intel->vtbl.leave_meta_state(intel); - intel_batchbuffer_flush(intel->batch); - } - UNLOCK_HARDWARE(intel); -} - - -/** - * Called by ctx->Driver.Clear. - */ -static void -intelClear(GLcontext *ctx, GLbitfield mask) -{ - struct intel_context *intel = intel_context(ctx); - const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask); - GLbitfield tri_mask = 0; - GLbitfield blit_mask = 0; - GLbitfield swrast_mask = 0; - struct gl_framebuffer *fb = ctx->DrawBuffer; - GLuint i; - - if (0) - fprintf(stderr, "%s\n", __FUNCTION__); - - /* HW color buffers (front, back, aux, generic FBO, etc) */ - if (colorMask == ~0) { - /* clear all R,G,B,A */ - /* XXX FBO: need to check if colorbuffers are software RBOs! */ - blit_mask |= (mask & BUFFER_BITS_COLOR); - } - else { - /* glColorMask in effect */ - tri_mask |= (mask & BUFFER_BITS_COLOR); - } - - /* HW stencil */ - if (mask & BUFFER_BIT_STENCIL) { - const struct intel_region *stencilRegion - = intel_get_rb_region(fb, BUFFER_STENCIL); - if (stencilRegion) { - /* have hw stencil */ - if ((ctx->Stencil.WriteMask[0] & 0xff) != 0xff) { - /* not clearing all stencil bits, so use triangle clearing */ - tri_mask |= BUFFER_BIT_STENCIL; - } - else { - /* clearing all stencil bits, use blitting */ - blit_mask |= BUFFER_BIT_STENCIL; - } - } - } - - /* HW depth */ - if (mask & BUFFER_BIT_DEPTH) { - /* clear depth with whatever method is used for stencil (see above) */ - if (tri_mask & BUFFER_BIT_STENCIL) - tri_mask |= BUFFER_BIT_DEPTH; - else - blit_mask |= BUFFER_BIT_DEPTH; - } - - /* SW fallback clearing */ - swrast_mask = mask & ~tri_mask & ~blit_mask; - - for (i = 0; i < BUFFER_COUNT; i++) { - GLuint bufBit = 1 << i; - if ((blit_mask | tri_mask) & bufBit) { - if (!fb->Attachment[i].Renderbuffer->ClassID) { - blit_mask &= ~bufBit; - tri_mask &= ~bufBit; - swrast_mask |= bufBit; - } - } - } - - - intelFlush(ctx); /* XXX intelClearWithBlit also does this */ - - if (blit_mask) - intelClearWithBlit(ctx, blit_mask); - - if (tri_mask) - intelClearWithTris(intel, tri_mask); - - if (swrast_mask) - _swrast_Clear(ctx, swrast_mask); -} - - -/* Emit wait for pending flips */ -void -intel_wait_flips(struct intel_context *intel, GLuint batch_flags) -{ - struct intel_framebuffer *intel_fb = - (struct intel_framebuffer *) intel->ctx.DrawBuffer; - struct intel_renderbuffer *intel_rb = - intel_get_renderbuffer(&intel_fb->Base, - intel_fb->Base._ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : - BUFFER_BACK_LEFT); - - if (intel_fb->Base.Name == 0 && intel_rb->pf_pending == intel_fb->pf_seq) { - GLint pf_pipes = intel_fb->pf_pipes; - BATCH_LOCALS; - - /* Wait for pending flips to take effect */ - BEGIN_BATCH(2, batch_flags); - OUT_BATCH(pf_pipes & 0x1 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_A_FLIP) - : 0); - OUT_BATCH(pf_pipes & 0x2 ? (MI_WAIT_FOR_EVENT | MI_WAIT_FOR_PLANE_B_FLIP) - : 0); - ADVANCE_BATCH(); - - intel_rb->pf_pending--; - } -} - -#if 0 -/* Flip the front & back buffers - */ -static GLboolean -intelPageFlip(const __DRIdrawablePrivate * dPriv) -{ - struct intel_context *intel; - int ret; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s\n", __FUNCTION__); - - assert(dPriv); - assert(dPriv->driContextPriv); - assert(dPriv->driContextPriv->driverPrivate); - - intel = (struct intel_context *) dPriv->driContextPriv->driverPrivate; - - if (intel->intelScreen->drmMinor < 9) - return GL_FALSE; - - intelFlush(&intel->ctx); - - ret = 0; - - LOCK_HARDWARE(intel); - - if (dPriv->numClipRects && intel_fb->pf_active) { - drm_i915_flip_t flip; - - flip.pipes = intel_fb->pf_pipes; - - ret = drmCommandWrite(intel->driFd, DRM_I915_FLIP, &flip, sizeof(flip)); - } - - UNLOCK_HARDWARE(intel); - - if (ret || !intel_fb->pf_active) - return GL_FALSE; - - if (!dPriv->numClipRects) { - usleep(10000); /* throttle invisible client 10ms */ - } - - intel_fb->pf_current_page = (intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3; - - if (dPriv->numClipRects != 0) { - intel_get_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT)->pf_pending = - intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->pf_pending = - ++intel_fb->pf_seq; - } - - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, &intel_fb->Base); - - if (INTEL_DEBUG & DEBUG_IOCTL) - fprintf(stderr, "%s: success\n", __FUNCTION__); - - return GL_TRUE; -} -#endif - - -static GLboolean -intelScheduleSwap(const __DRIdrawablePrivate * dPriv, GLboolean *missed_target) -{ - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - unsigned int interval = driGetVBlankInterval(dPriv, intel_fb->vblank_flags); - struct intel_context *intel = - intelScreenContext(dPriv->driScreenPriv->private); - const intelScreenPrivate *intelScreen = intel->intelScreen; - unsigned int target; - drm_i915_vblank_swap_t swap; - GLboolean ret; - - if (!intel_fb->vblank_flags || - (intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) || - intelScreen->drmMinor < (intel_fb->pf_active ? 9 : 6)) - return GL_FALSE; - - swap.seqtype = DRM_VBLANK_ABSOLUTE; - - if (intel_fb->vblank_flags & VBLANK_FLAG_SYNC) { - swap.seqtype |= DRM_VBLANK_NEXTONMISS; - } else if (interval == 0) { - return GL_FALSE; - } - - swap.drawable = dPriv->hHWDrawable; - target = swap.sequence = intel_fb->vbl_seq + interval; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - swap.seqtype |= DRM_VBLANK_SECONDARY; - } - - LOCK_HARDWARE(intel); - - intel_batchbuffer_flush(intel->batch); - - if ( intel_fb->pf_active ) { - swap.seqtype |= DRM_VBLANK_FLIP; - - intel_fb->pf_current_page = (((intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3) + 1) % - intel_fb->pf_num_pages; - } - - if (!drmCommandWriteRead(intel->driFd, DRM_I915_VBLANK_SWAP, &swap, - sizeof(swap))) { - intel_fb->vbl_seq = swap.sequence; - swap.sequence -= target; - *missed_target = swap.sequence > 0 && swap.sequence <= (1 << 23); - - intel_get_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT)->vbl_pending = - intel_get_renderbuffer(&intel_fb->Base, - BUFFER_FRONT_LEFT)->vbl_pending = - intel_fb->vbl_seq; - - if (swap.seqtype & DRM_VBLANK_FLIP) { - intel_flip_renderbuffers(intel_fb); - intel_draw_buffer(&intel->ctx, intel->ctx.DrawBuffer); - } - - ret = GL_TRUE; - } else { - if (swap.seqtype & DRM_VBLANK_FLIP) { - intel_fb->pf_current_page = ((intel->sarea->pf_current_page >> - (intel_fb->pf_pipes & 0x2)) & 0x3) % - intel_fb->pf_num_pages; - } - - ret = GL_FALSE; - } - - UNLOCK_HARDWARE(intel); - - return ret; -} - -void -intelSwapBuffers(__DRIdrawablePrivate * dPriv) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel; - - if (ctx == NULL) - return; - - intel = intel_context(ctx); - - if (ctx->Visual.doubleBufferMode) { - GLboolean missed_target; - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - int64_t ust; - - _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - - if (!intelScheduleSwap(dPriv, &missed_target)) { - driWaitForVBlank(dPriv, &intel_fb->vbl_seq, intel_fb->vblank_flags, - &missed_target); - - intelCopyBuffer(dPriv, NULL); - } - - intel_fb->swap_count++; - (*dri_interface->getUST) (&ust); - if (missed_target) { - intel_fb->swap_missed_count++; - intel_fb->swap_missed_ust = ust - intel_fb->swap_ust; - } - - intel_fb->swap_ust = ust; - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); - } -} - -void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) -{ - if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) { - struct intel_context *intel = - (struct intel_context *) dPriv->driContextPriv->driverPrivate; - GLcontext *ctx = &intel->ctx; - - if (ctx->Visual.doubleBufferMode) { - drm_clip_rect_t rect; - /* fixup cliprect (driDrawable may have changed?) later */ - rect.x1 = x; - rect.y1 = y; - rect.x2 = w; - rect.y2 = h; - _mesa_notifySwapBuffers(ctx); /* flush pending rendering comands */ - intelCopyBuffer(dPriv, &rect); - } - } - else { - /* XXX this shouldn't be an error but we can't handle it for now */ - fprintf(stderr, "%s: drawable has no context!\n", __FUNCTION__); - } -} - - -/** - * Update the hardware state for drawing into a window or framebuffer object. - * - * Called by glDrawBuffer, glBindFramebufferEXT, MakeCurrent, and other - * places within the driver. - * - * Basically, this needs to be called any time the current framebuffer - * changes, the renderbuffers change, or we need to draw into different - * color buffers. - */ -void -intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_region *colorRegion, *depthRegion = NULL; - struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL; - - if (!fb) { - /* this can happen during the initial context initialization */ - return; - } - - /* Do this here, not core Mesa, since this function is called from - * many places within the driver. - */ - if (ctx->NewState & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) { - /* this updates the DrawBuffer->_NumColorDrawBuffers fields, etc */ - _mesa_update_framebuffer(ctx); - /* this updates the DrawBuffer's Width/Height if it's a FBO */ - _mesa_update_draw_buffer_bounds(ctx); - } - - if (fb->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { - /* this may occur when we're called by glBindFrameBuffer() during - * the process of someone setting up renderbuffers, etc. - */ - /*_mesa_debug(ctx, "DrawBuffer: incomplete user FBO\n");*/ - return; - } - - if (fb->Name) - intel_validate_paired_depth_stencil(ctx, fb); - - /* - * How many color buffers are we drawing into? - */ - if (fb->_NumColorDrawBuffers[0] != 1) { - /* writing to 0 or 2 or 4 color buffers */ - /*_mesa_debug(ctx, "Software rendering\n");*/ - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); - } - else { - /* draw to exactly one color buffer */ - /*_mesa_debug(ctx, "Hardware rendering\n");*/ - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); - } - - /* - * Get the intel_renderbuffer for the colorbuffer we're drawing into. - * And set up cliprects. - */ - { - struct intel_renderbuffer *irb; - intelSetRenderbufferClipRects(intel); - irb = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); - colorRegion = (irb && irb->region) ? irb->region : NULL; - } - - /* Update culling direction which changes depending on the - * orientation of the buffer: - */ - if (ctx->Driver.FrontFace) - ctx->Driver.FrontFace(ctx, ctx->Polygon.FrontFace); - else - ctx->NewState |= _NEW_POLYGON; - - if (!colorRegion) { - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE); - } - else { - FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE); - } - - /*** - *** Get depth buffer region and check if we need a software fallback. - *** Note that the depth buffer is usually a DEPTH_STENCIL buffer. - ***/ - if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped) { - irbDepth = intel_renderbuffer(fb->_DepthBuffer->Wrapped); - if (irbDepth && irbDepth->region) { - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); - depthRegion = irbDepth->region; - } - else { - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_TRUE); - depthRegion = NULL; - } - } - else { - /* not using depth buffer */ - FALLBACK(intel, INTEL_FALLBACK_DEPTH_BUFFER, GL_FALSE); - depthRegion = NULL; - } - - /*** - *** Stencil buffer - *** This can only be hardware accelerated if we're using a - *** combined DEPTH_STENCIL buffer (for now anyway). - ***/ - if (fb->_StencilBuffer && fb->_StencilBuffer->Wrapped) { - irbStencil = intel_renderbuffer(fb->_StencilBuffer->Wrapped); - if (irbStencil && irbStencil->region) { - ASSERT(irbStencil->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); - /* need to re-compute stencil hw state */ - ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); - if (!depthRegion) - depthRegion = irbStencil->region; - } - else { - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_TRUE); - } - } - else { - /* XXX FBO: instead of FALSE, pass ctx->Stencil.Enabled ??? */ - FALLBACK(intel, INTEL_FALLBACK_STENCIL_BUFFER, GL_FALSE); - /* need to re-compute stencil hw state */ - ctx->Driver.Enable(ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled); - } - - - /** - ** Release old regions, reference new regions - **/ - - intel->vtbl.set_draw_region(intel, colorRegion, depthRegion); - - /* update viewport since it depends on window size */ - ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y, - ctx->Viewport.Width, ctx->Viewport.Height); - - /* Update hardware scissor */ - ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y, - ctx->Scissor.Width, ctx->Scissor.Height); -} - - -static void -intelDrawBuffer(GLcontext * ctx, GLenum mode) -{ - intel_draw_buffer(ctx, ctx->DrawBuffer); -} - - -static void -intelReadBuffer(GLcontext * ctx, GLenum mode) -{ - if (ctx->ReadBuffer == ctx->DrawBuffer) { - /* This will update FBO completeness status. - * A framebuffer will be incomplete if the GL_READ_BUFFER setting - * refers to a missing renderbuffer. Calling glReadBuffer can set - * that straight and can make the drawing buffer complete. - */ - intel_draw_buffer(ctx, ctx->DrawBuffer); - } - /* Generally, functions which read pixels (glReadPixels, glCopyPixels, etc) - * reference ctx->ReadBuffer and do appropriate state checks. - */ -} - - -void -intelInitBufferFuncs(struct dd_function_table *functions) -{ - functions->Clear = intelClear; - functions->DrawBuffer = intelDrawBuffer; - functions->ReadBuffer = intelReadBuffer; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_buffers.h b/src/mesa/drivers/dri/i915tex/intel_buffers.h deleted file mode 100644 index 13d1a15ffb..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_buffers.h +++ /dev/null @@ -1,55 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BUFFERS_H -#define INTEL_BUFFERS_H - - -struct intel_context; -struct intel_framebuffer; - - -extern GLboolean -intel_intersect_cliprects(drm_clip_rect_t * dest, - const drm_clip_rect_t * a, - const drm_clip_rect_t * b); - -extern struct intel_region *intel_readbuf_region(struct intel_context *intel); - -extern struct intel_region *intel_drawbuf_region(struct intel_context *intel); - -extern void intel_wait_flips(struct intel_context *intel, GLuint batch_flags); - -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void intelWindowMoved(struct intel_context *intel); - -extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb); - -extern void intelInitBufferFuncs(struct dd_function_table *functions); - -#endif /* INTEL_BUFFERS_H */ diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c deleted file mode 100644 index 5ab919ecb6..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_context.c +++ /dev/null @@ -1,782 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "glheader.h" -#include "context.h" -#include "matrix.h" -#include "simple_list.h" -#include "extensions.h" -#include "framebuffer.h" -#include "imports.h" -#include "points.h" - -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#include "tnl/tnl.h" - -#include "tnl/t_pipeline.h" -#include "tnl/t_vertex.h" - -#include "drivers/common/driverfuncs.h" - -#include "intel_screen.h" - -#include "i830_dri.h" - -#include "intel_buffers.h" -#include "intel_tex.h" -#include "intel_span.h" -#include "intel_tris.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "intel_pixel.h" -#include "intel_regions.h" -#include "intel_buffer_objects.h" -#include "intel_fbo.h" - -#include "pipe/softpipe/sp_context.h" -#include "state_tracker/st_public.h" - - -#include "drirenderbuffer.h" -#include "vblank.h" -#include "utils.h" -#include "xmlpool.h" /* for symbolic values of enum-type options */ -#ifndef INTEL_DEBUG -int INTEL_DEBUG = (0); -#endif - -#define need_GL_ARB_multisample -#define need_GL_ARB_point_parameters -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_buffer_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_window_pos -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_equation_separate -#define need_GL_EXT_blend_func_separate -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_cull_vertex -#define need_GL_EXT_fog_coord -#define need_GL_EXT_framebuffer_object -#define need_GL_EXT_multi_draw_arrays -#define need_GL_EXT_secondary_color -#define need_GL_NV_vertex_program -#include "extension_helper.h" - - -#define DRIVER_DATE "20061102" - -_glthread_Mutex lockMutex; -static GLboolean lockMutexInit = GL_FALSE; - - -static const GLubyte * -intelGetString(GLcontext * ctx, GLenum name) -{ - const char *chipset; - static char buffer[128]; - - switch (name) { - case GL_VENDOR: - return (GLubyte *) "Tungsten Graphics, Inc"; - break; - - case GL_RENDERER: - switch (intel_context(ctx)->intelScreen->deviceID) { - case PCI_CHIP_845_G: - chipset = "Intel(R) 845G"; - break; - case PCI_CHIP_I830_M: - chipset = "Intel(R) 830M"; - break; - case PCI_CHIP_I855_GM: - chipset = "Intel(R) 852GM/855GM"; - break; - case PCI_CHIP_I865_G: - chipset = "Intel(R) 865G"; - break; - case PCI_CHIP_I915_G: - chipset = "Intel(R) 915G"; - break; - case PCI_CHIP_I915_GM: - chipset = "Intel(R) 915GM"; - break; - case PCI_CHIP_I945_G: - chipset = "Intel(R) 945G"; - break; - case PCI_CHIP_I945_GM: - chipset = "Intel(R) 945GM"; - break; - case PCI_CHIP_I945_GME: - chipset = "Intel(R) 945GME"; - break; - case PCI_CHIP_G33_G: - chipset = "Intel(R) G33"; - break; - case PCI_CHIP_Q35_G: - chipset = "Intel(R) Q35"; - break; - case PCI_CHIP_Q33_G: - chipset = "Intel(R) Q33"; - break; - default: - chipset = "Unknown Intel Chipset"; - break; - } - - (void) driGetRendererString(buffer, chipset, DRIVER_DATE, 0); - return (GLubyte *) buffer; - - default: - return NULL; - } -} - - -/** - * Extension strings exported by the intel driver. - * - * \note - * It appears that ARB_texture_env_crossbar has "disappeared" compared to the - * old i830-specific driver. - */ -const struct dri_extension card_extensions[] = { - {"GL_ARB_multisample", GL_ARB_multisample_functions}, - {"GL_ARB_multitexture", NULL}, - {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, - {"GL_ARB_texture_border_clamp", NULL}, - {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, - {"GL_ARB_texture_cube_map", NULL}, - {"GL_ARB_texture_env_add", NULL}, - {"GL_ARB_texture_env_combine", NULL}, - {"GL_ARB_texture_env_dot3", NULL}, - {"GL_ARB_texture_mirrored_repeat", NULL}, - {"GL_ARB_texture_rectangle", NULL}, - {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, - {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, - {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, - {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", - GL_EXT_blend_equation_separate_functions}, - {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, - {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_subtract", NULL}, - {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, - {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, -#if 1 /* XXX FBO temporary? */ - {"GL_EXT_packed_depth_stencil", NULL}, -#endif - {"GL_EXT_pixel_buffer_object", NULL}, - {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, - {"GL_EXT_stencil_wrap", NULL}, - {"GL_EXT_texture_edge_clamp", NULL}, - {"GL_EXT_texture_env_combine", NULL}, - {"GL_EXT_texture_env_dot3", NULL}, - {"GL_EXT_texture_filter_anisotropic", NULL}, - {"GL_EXT_texture_lod_bias", NULL}, - {"GL_3DFX_texture_compression_FXT1", NULL}, - {"GL_APPLE_client_storage", NULL}, - {"GL_MESA_pack_invert", NULL}, - {"GL_MESA_ycbcr_texture", NULL}, - {"GL_NV_blend_square", NULL}, - {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, - {"GL_NV_vertex_program1_1", NULL}, -/* { "GL_SGIS_generate_mipmap", NULL }, */ - {NULL, NULL} -}; - -extern const struct tnl_pipeline_stage _intel_render_stage; - -static const struct tnl_pipeline_stage *intel_pipeline[] = { - &_tnl_vertex_transform_stage, - &_tnl_vertex_cull_stage, - &_tnl_normal_transform_stage, - &_tnl_lighting_stage, - &_tnl_fog_coordinate_stage, - &_tnl_texgen_stage, - &_tnl_texture_transform_stage, - &_tnl_point_attenuation_stage, - &_tnl_vertex_program_stage, -#if 1 - &_intel_render_stage, /* ADD: unclipped rastersetup-to-dma */ -#endif - &_tnl_render_stage, - 0, -}; - - -static const struct dri_debug_control debug_control[] = { - {"tex", DEBUG_TEXTURE}, - {"state", DEBUG_STATE}, - {"ioctl", DEBUG_IOCTL}, - {"blit", DEBUG_BLIT}, - {"mip", DEBUG_MIPTREE}, - {"fall", DEBUG_FALLBACKS}, - {"verb", DEBUG_VERBOSE}, - {"bat", DEBUG_BATCH}, - {"pix", DEBUG_PIXEL}, - {"buf", DEBUG_BUFMGR}, - {"reg", DEBUG_REGION}, - {"fbo", DEBUG_FBO}, - {"lock", DEBUG_LOCK}, - {NULL, 0} -}; - - -static void -intelInvalidateState(GLcontext * ctx, GLuint new_state) -{ - _swrast_InvalidateState(ctx, new_state); - _swsetup_InvalidateState(ctx, new_state); - _vbo_InvalidateState(ctx, new_state); - _tnl_InvalidateState(ctx, new_state); - _tnl_invalidate_vertex_state(ctx, new_state); - - st_invalidate_state( ctx, new_state ); - - intel_context(ctx)->NewGLState |= new_state; -} - - -void -intelFlush(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - - if (intel->Fallback) - _swrast_flush(ctx); - - INTEL_FIREVERTICES(intel); - - if (intel->batch->map != intel->batch->ptr) - intel_batchbuffer_flush(intel->batch); - - /* XXX: Need to do an MI_FLUSH here. - */ -} - - -/** - * Check if we need to rotate/warp the front color buffer to the - * rotated screen. We generally need to do this when we get a glFlush - * or glFinish after drawing to the front color buffer. - * If no rotation, just copy the private fake front buffer to the real one. - */ -static void -intelCheckFrontUpdate(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - /* rely on _ColorDrawBufferMask being kept up to date by mesa - even for window-fbos. */ - /* not sure. Might need that for all masks including - BUFFER_BIT_FRONT_LEFT maybe? */ - if (intel->ctx.DrawBuffer->_ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT) { - __DRIdrawablePrivate *dPriv = intel->driDrawable; - intelCopyBuffer(dPriv, NULL); - } -} - - -/** - * Called via glFlush. - */ -static void -intelglFlush(GLcontext * ctx) -{ - intelFlush(ctx); - intelCheckFrontUpdate(ctx); -} - -void -intelFinish(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - intelFlush(ctx); - if (intel->batch->last_fence) { - driFenceFinish(intel->batch->last_fence, - 0, GL_FALSE); - driFenceUnReference(intel->batch->last_fence); - intel->batch->last_fence = NULL; - } - intelCheckFrontUpdate(ctx); -} - - -void -intelInitDriverFunctions(struct dd_function_table *functions) -{ - _mesa_init_driver_functions(functions); - - functions->Flush = intelglFlush; - functions->Finish = intelFinish; - functions->GetString = intelGetString; - functions->UpdateState = intelInvalidateState; - functions->CopyColorTable = _swrast_CopyColorTable; - functions->CopyColorSubTable = _swrast_CopyColorSubTable; - functions->CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D; - functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D; - - intelInitTextureFuncs(functions); - intelInitPixelFuncs(functions); - intelInitStateFuncs(functions); - intelInitBufferFuncs(functions); -} - - -GLboolean -intelInitContext(struct intel_context *intel, - const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate, - struct dd_function_table *functions) -{ - GLcontext *ctx = &intel->ctx; - GLcontext *shareCtx = (GLcontext *) sharedContextPrivate; - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - drmI830Sarea *saPriv = (drmI830Sarea *) - (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset); - int fthrottle_mode; - GLboolean havePools; - - DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - havePools = intelCreatePools(intelScreen); - DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - - if (!havePools) - return GL_FALSE; - - if (!_mesa_initialize_context(&intel->ctx, - mesaVis, shareCtx, - functions, (void *) intel)) - return GL_FALSE; - - driContextPriv->driverPrivate = intel; - intel->intelScreen = intelScreen; - intel->driScreen = sPriv; - intel->sarea = saPriv; - - if (!lockMutexInit) { - lockMutexInit = GL_TRUE; - _glthread_INIT_MUTEX(lockMutex); - } - - driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, - intel->driScreen->myNum, "i915"); - - ctx->Const.MaxTextureMaxAnisotropy = 2.0; - - /* This doesn't yet catch all non-conformant rendering, but it's a - * start. - */ - if (getenv("INTEL_STRICT_CONFORMANCE")) { - intel->strict_conformance = 1; - } - - ctx->Const.MinLineWidth = 1.0; - ctx->Const.MinLineWidthAA = 1.0; - ctx->Const.MaxLineWidth = 3.0; - ctx->Const.MaxLineWidthAA = 3.0; - ctx->Const.LineWidthGranularity = 1.0; - - ctx->Const.MinPointSize = 1.0; - ctx->Const.MinPointSizeAA = 1.0; - ctx->Const.MaxPointSize = 255.0; - ctx->Const.MaxPointSizeAA = 3.0; - ctx->Const.PointSizeGranularity = 1.0; - - /* reinitialize the context point state. - * It depend on constants in __GLcontextRec::Const - */ - _mesa_init_point(ctx); - - ctx->Const.MaxColorAttachments = 4; /* XXX FBO: review this */ - - /* Initialize the software rasterizer and helper modules. */ - _swrast_CreateContext(ctx); - _vbo_CreateContext(ctx); - _tnl_CreateContext(ctx); - _swsetup_CreateContext(ctx); - - /* Install the customized pipeline: */ - _tnl_destroy_pipeline(ctx); - _tnl_install_pipeline(ctx, intel_pipeline); - - /* Configure swrast to match hardware characteristics: */ - _swrast_allow_pixel_fog(ctx, GL_FALSE); - _swrast_allow_vertex_fog(ctx, GL_TRUE); - - /* Dri stuff */ - intel->hHWContext = driContextPriv->hHWContext; - intel->driFd = sPriv->fd; - intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; - - intel->hw_stipple = 1; - - /* XXX FBO: this doesn't seem to be used anywhere */ - switch (mesaVis->depthBits) { - case 0: /* what to do in this case? */ - case 16: - intel->polygon_offset_scale = 1.0 / 0xffff; - break; - case 24: - intel->polygon_offset_scale = 2.0 / 0xffffff; /* req'd to pass glean */ - break; - default: - assert(0); - break; - } - - /* Initialize swrast, tnl driver tables: */ - intelInitSpanFuncs(ctx); - intelInitTriFuncs(ctx); - - - intel->RenderIndex = ~0; - - fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->iw.irq_seq = -1; - intel->irqsEmitted = 0; - - intel->do_irqs = (intel->intelScreen->irq_active && - fthrottle_mode == DRI_CONF_FTHROTTLE_IRQS); - - intel->do_usleeps = (fthrottle_mode == DRI_CONF_FTHROTTLE_USLEEPS); - - _math_matrix_ctr(&intel->ViewportMatrix); - - /* Disable imaging extension until convolution is working in - * teximage paths: - */ - driInitExtensions(ctx, card_extensions, -/* GL_TRUE, */ - GL_FALSE); - - - intel->batch = intel_batchbuffer_alloc(intel); - intel->last_swap_fence = NULL; - intel->first_swap_fence = NULL; - - intel_bufferobj_init(intel); - intel_fbo_init(intel); - - if (intel->ctx.Mesa_DXTn) { - _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); - _mesa_enable_extension(ctx, "GL_S3_s3tc"); - } - else if (driQueryOptionb(&intel->optionCache, "force_s3tc_enable")) { - _mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc"); - } - - intel->prim.primitive = ~0; - - -#if DO_DEBUG - INTEL_DEBUG = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); -#endif - - if (getenv("INTEL_NO_RAST")) { - fprintf(stderr, "disabling 3D rasterization\n"); - FALLBACK(intel, INTEL_FALLBACK_USER, 1); - } - - - st_create_context( &intel->ctx, - softpipe_create() ); - - - return GL_TRUE; -} - -void -intelDestroyContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = - (struct intel_context *) driContextPriv->driverPrivate; - - assert(intel); /* should never be null */ - if (intel) { - GLboolean release_texture_heaps; - - INTEL_FIREVERTICES(intel); - - intel->vtbl.destroy(intel); - - release_texture_heaps = (intel->ctx.Shared->RefCount == 1); - _swsetup_DestroyContext(&intel->ctx); - _tnl_DestroyContext(&intel->ctx); - _vbo_DestroyContext(&intel->ctx); - - _swrast_DestroyContext(&intel->ctx); - intel->Fallback = 0; /* don't call _swrast_Flush later */ - - intel_batchbuffer_free(intel->batch); - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - if (intel->first_swap_fence) { - driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->first_swap_fence); - intel->first_swap_fence = NULL; - } - - - if (release_texture_heaps) { - /* This share group is about to go away, free our private - * texture object data. - */ - if (INTEL_DEBUG & DEBUG_TEXTURE) - fprintf(stderr, "do something to free texture heaps\n"); - } - - /* free the Mesa context */ - _mesa_free_context_data(&intel->ctx); - } -} - -GLboolean -intelUnbindContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = (struct intel_context *) driContextPriv->driverPrivate; - /* XXX UnbindContext is called AFTER the new context is made current. - Hopefully shouldn't be a problem ? */ - FLUSH_VERTICES((&intel->ctx), 0); - intelFlush(&intel->ctx); - return GL_TRUE; -} - -GLboolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv) -{ - -#if 0 - if (driDrawPriv) { - fprintf(stderr, "x %d, y %d, width %d, height %d\n", - driDrawPriv->x, driDrawPriv->y, driDrawPriv->w, driDrawPriv->h); - } -#endif - - if (driContextPriv) { - struct intel_context *intel = - (struct intel_context *) driContextPriv->driverPrivate; - struct intel_framebuffer *intel_fb = - (struct intel_framebuffer *) driDrawPriv->driverPrivate; - GLframebuffer *readFb = (GLframebuffer *) driReadPriv->driverPrivate; - - /* this is a hack so we have a valid context when the region allocation - is done. Need a per-screen context? */ - intel->intelScreen->dummyctxptr = intel; - - /* update GLframebuffer size to match window if needed */ - driUpdateFramebufferSize(&intel->ctx, driDrawPriv); - - if (driReadPriv != driDrawPriv) { - driUpdateFramebufferSize(&intel->ctx, driReadPriv); - } - - _mesa_make_current(&intel->ctx, &intel_fb->Base, readFb); - - /* The drawbuffer won't always be updated by _mesa_make_current: - */ - if (intel->ctx.DrawBuffer == &intel_fb->Base) { - - if (intel->driDrawable != driDrawPriv) { - if (driDrawPriv->pdraw->swap_interval == (unsigned)-1) { - int i; - - intel_fb->vblank_flags = (intel->intelScreen->irq_active != 0) - ? driGetDefaultVBlankFlags(&intel->optionCache) - : VBLANK_FLAG_NO_IRQ; - - (*dri_interface->getUST) (&intel_fb->swap_ust); - driDrawableInitVBlank(driDrawPriv, intel_fb->vblank_flags, - &intel_fb->vbl_seq); - intel_fb->vbl_waited = intel_fb->vbl_seq; - - for (i = 0; i < 2; i++) { - if (intel_fb->color_rb[i]) - intel_fb->color_rb[i]->vbl_pending = intel_fb->vbl_seq; - } - } - } - } - - if ((intel->driDrawable != driDrawPriv) || - (intel->lastStamp != driDrawPriv->lastStamp)) { - intel->driDrawable = driDrawPriv; - intelWindowMoved(intel); - intel->lastStamp = driDrawPriv->lastStamp; - } - - } - else { - _mesa_make_current(NULL, NULL, NULL); - } - - return GL_TRUE; -} - -static void -intelContendedLock(struct intel_context *intel, GLuint flags) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - __DRIscreenPrivate *sPriv = intel->driScreen; - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - drmI830Sarea *sarea = intel->sarea; - - drmGetLock(intel->driFd, intel->hHWContext, flags); - - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("%s - got contended lock\n", __progname); - - /* If the window moved, may need to set a new cliprect now. - * - * NOTE: This releases and regains the hw lock, so all state - * checking must be done *after* this call: - */ - if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - - if (sarea->width != intelScreen->width || - sarea->height != intelScreen->height) { - - intelUpdateScreenRotation(sPriv, sarea); - } - -#if 0 - if (sarea->width != intel->width || - sarea->height != intel->height || - sarea->rotation != intel->current_rotation) { - int numClipRects = intel->numClipRects; - - /* - * FIXME: Really only need to do this when drawing to a - * common back- or front buffer. - */ - - /* - * This will essentially drop the outstanding batchbuffer on the floor. - */ - intel->numClipRects = 0; - - if (intel->Fallback) - _swrast_flush(&intel->ctx); - - INTEL_FIREVERTICES(intel); - - if (intel->batch->map != intel->batch->ptr) - intel_batchbuffer_flush(intel->batch); - - intel->numClipRects = numClipRects; - - /* force window update */ - intel->lastStamp = 0; - - intel->width = sarea->width; - intel->height = sarea->height; - intel->current_rotation = sarea->rotation; - } -#endif -} - - -/* Lock the hardware and validate our state. - */ -void LOCK_HARDWARE( struct intel_context *intel ) -{ - char __ret=0; - struct intel_framebuffer *intel_fb = NULL; - struct intel_renderbuffer *intel_rb = NULL; - _glthread_LOCK_MUTEX(lockMutex); - assert(!intel->locked); - - if (intel->driDrawable) { - intel_fb = intel->driDrawable->driverPrivate; - - if (intel_fb) - intel_rb = - intel_get_renderbuffer(&intel_fb->Base, - intel_fb->Base._ColorDrawBufferMask[0] == - BUFFER_BIT_FRONT_LEFT ? BUFFER_FRONT_LEFT : - BUFFER_BACK_LEFT); - } - - if (intel_rb && intel_fb->vblank_flags && - !(intel_fb->vblank_flags & VBLANK_FLAG_NO_IRQ) && - (intel_fb->vbl_waited - intel_rb->vbl_pending) > (1<<23)) { - drmVBlank vbl; - - vbl.request.type = DRM_VBLANK_ABSOLUTE; - - if ( intel_fb->vblank_flags & VBLANK_FLAG_SECONDARY ) { - vbl.request.type |= DRM_VBLANK_SECONDARY; - } - - vbl.request.sequence = intel_rb->vbl_pending; - drmWaitVBlank(intel->driFd, &vbl); - intel_fb->vbl_waited = vbl.reply.sequence; - } - - DRM_CAS(intel->driHwLock, intel->hHWContext, - (DRM_LOCK_HELD|intel->hHWContext), __ret); - - if (__ret) - intelContendedLock( intel, 0 ); - - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("%s - locked\n", __progname); - - intel->locked = 1; -} - - - /* Unlock the hardware using the global current context - */ -void UNLOCK_HARDWARE( struct intel_context *intel ) -{ - intel->locked = 0; - - DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); - - _glthread_UNLOCK_MUTEX(lockMutex); - - if (INTEL_DEBUG & DEBUG_LOCK) - _mesa_printf("%s - unlocked\n", __progname); -} - diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h deleted file mode 100644 index ee5be45533..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_context.h +++ /dev/null @@ -1,487 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTELCONTEXT_INC -#define INTELCONTEXT_INC - - - -#include "mtypes.h" -#include "drm.h" -#include "mm.h" -#include "texmem.h" - -#include "intel_screen.h" -#include "i915_drm.h" -#include "i830_common.h" -#include "tnl/t_vertex.h" - -#define TAG(x) intel##x -#include "tnl_dd/t_dd_vertex.h" -#undef TAG - -#define DV_PF_555 (1<<8) -#define DV_PF_565 (2<<8) -#define DV_PF_8888 (3<<8) - -struct intel_region; -struct intel_context; -struct _DriBufferObject; - -typedef void (*intel_tri_func) (struct intel_context *, intelVertex *, - intelVertex *, intelVertex *); -typedef void (*intel_line_func) (struct intel_context *, intelVertex *, - intelVertex *); -typedef void (*intel_point_func) (struct intel_context *, intelVertex *); - -#define INTEL_FALLBACK_DRAW_BUFFER 0x1 -#define INTEL_FALLBACK_READ_BUFFER 0x2 -#define INTEL_FALLBACK_DEPTH_BUFFER 0x4 -#define INTEL_FALLBACK_STENCIL_BUFFER 0x8 -#define INTEL_FALLBACK_USER 0x10 -#define INTEL_FALLBACK_RENDERMODE 0x20 - -extern void intelFallback(struct intel_context *intel, GLuint bit, - GLboolean mode); -#define FALLBACK( intel, bit, mode ) intelFallback( intel, bit, mode ) - - -#define INTEL_WRITE_PART 0x1 -#define INTEL_WRITE_FULL 0x2 -#define INTEL_READ 0x4 - -struct intel_texture_object -{ - struct gl_texture_object base; /* The "parent" object */ - - /* The mipmap tree must include at least these levels once - * validated: - */ - GLuint firstLevel; - GLuint lastLevel; - - /* Offset for firstLevel image: - */ - GLuint textureOffset; - - /* On validation any active images held in main memory or in other - * regions will be copied to this region and the old storage freed. - */ - struct intel_mipmap_tree *mt; - - GLboolean imageOverride; - GLint depthOverride; - GLuint pitchOverride; -}; - - - -struct intel_texture_image -{ - struct gl_texture_image base; - - /* These aren't stored in gl_texture_image - */ - GLuint level; - GLuint face; - - /* If intelImage->mt != NULL, image data is stored here. - * Else if intelImage->base.Data != NULL, image is stored there. - * Else there is no image data. - */ - struct intel_mipmap_tree *mt; -}; - - -#define INTEL_MAX_FIXUP 64 - -struct intel_context -{ - GLcontext ctx; /* the parent class */ - - struct - { - void (*destroy) (struct intel_context * intel); - void (*emit_state) (struct intel_context * intel); - void (*lost_hardware) (struct intel_context * intel); - void (*update_texture_state) (struct intel_context * intel); - - void (*render_start) (struct intel_context * intel); - void (*render_prevalidate) (struct intel_context * intel); - void (*set_draw_region) (struct intel_context * intel, - struct intel_region * draw_region, - struct intel_region * depth_region); - - GLuint(*flush_cmd) (void); - - void (*reduced_primitive_state) (struct intel_context * intel, - GLenum rprim); - - GLboolean(*check_vertex_size) (struct intel_context * intel, - GLuint expected); - - - /* Metaops: - */ - void (*install_meta_state) (struct intel_context * intel); - void (*leave_meta_state) (struct intel_context * intel); - - void (*meta_draw_region) (struct intel_context * intel, - struct intel_region * draw_region, - struct intel_region * depth_region); - - void (*meta_color_mask) (struct intel_context * intel, GLboolean); - - void (*meta_stencil_replace) (struct intel_context * intel, - GLuint mask, GLuint clear); - - void (*meta_depth_replace) (struct intel_context * intel); - - void (*meta_texture_blend_replace) (struct intel_context * intel); - - void (*meta_no_stencil_write) (struct intel_context * intel); - void (*meta_no_depth_write) (struct intel_context * intel); - void (*meta_no_texture) (struct intel_context * intel); - - void (*meta_import_pixel_state) (struct intel_context * intel); - - GLboolean(*meta_tex_rect_source) (struct intel_context * intel, - struct _DriBufferObject * buffer, - GLuint offset, - GLuint pitch, - GLuint height, - GLenum format, GLenum type); - void (*rotate_window) (struct intel_context * intel, - __DRIdrawablePrivate * dPriv, GLuint srcBuf); - - void (*assert_not_dirty) (struct intel_context *intel); - - } vtbl; - - GLint refcount; - GLuint Fallback; - GLuint NewGLState; - - struct _DriFenceObject *last_swap_fence; - struct _DriFenceObject *first_swap_fence; - - struct intel_batchbuffer *batch; - - struct - { - GLuint id; - GLuint primitive; - GLubyte *start_ptr; - void (*flush) (struct intel_context *); - } prim; - - GLboolean locked; - char *prevLockFile; - int prevLockLine; - - GLuint ClearColor565; - GLuint ClearColor8888; - - /* Offsets of fields within the current vertex: - */ - GLuint coloroffset; - GLuint specoffset; - GLuint wpos_offset; - GLuint wpos_size; - - struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX]; - GLuint vertex_attr_count; - - GLfloat polygon_offset_scale; /* dependent on depth_scale, bpp */ - - GLboolean hw_stipple; - GLboolean strict_conformance; - - /* AGP memory buffer manager: - */ - struct bufmgr *bm; - - - /* State for intelvb.c and inteltris.c. - */ - GLuint RenderIndex; - GLmatrix ViewportMatrix; - GLenum render_primitive; - GLenum reduced_primitive; - GLuint vertex_size; - GLubyte *verts; /* points to tnl->clipspace.vertex_buf */ - - - /* Fallback rasterization functions - */ - intel_point_func draw_point; - intel_line_func draw_line; - intel_tri_func draw_tri; - - /* These refer to the current drawing buffer: - */ - GLuint numClipRects; /**< cliprects for drawing */ - drm_clip_rect_t *pClipRects; - drm_clip_rect_t fboRect; /**< cliprect for rendering */ - - GLuint do_usleeps; - int do_irqs; - GLuint irqsEmitted; - drm_i915_irq_wait_t iw; - - drm_context_t hHWContext; - drmLock *driHwLock; - int driFd; - - __DRIdrawablePrivate *driDrawable; - __DRIscreenPrivate *driScreen; - intelScreenPrivate *intelScreen; - drmI830Sarea *sarea; - - GLuint lastStamp; - - /** - * Configuration cache - */ - driOptionCache optionCache; -}; - -/* These are functions now: - */ -void LOCK_HARDWARE( struct intel_context *intel ); -void UNLOCK_HARDWARE( struct intel_context *intel ); - -extern char *__progname; - - -#define SUBPIXEL_X 0.125 -#define SUBPIXEL_Y 0.125 - -#define INTEL_FIREVERTICES(intel) \ -do { \ - if ((intel)->prim.flush) \ - (intel)->prim.flush(intel); \ -} while (0) - -/* ================================================================ - * Color packing: - */ - -#define INTEL_PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define INTEL_PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define INTEL_PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define INTEL_PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - -/* ================================================================ - * From linux kernel i386 header files, copes with odd sizes better - * than COPY_DWORDS would: - * XXX Put this in src/mesa/main/imports.h ??? - */ -#if defined(i386) || defined(__i386__) -static INLINE void * -__memcpy(void *to, const void *from, size_t n) -{ - int d0, d1, d2; - __asm__ __volatile__("rep ; movsl\n\t" - "testb $2,%b4\n\t" - "je 1f\n\t" - "movsw\n" - "1:\ttestb $1,%b4\n\t" - "je 2f\n\t" - "movsb\n" "2:":"=&c"(d0), "=&D"(d1), "=&S"(d2) - :"0"(n / 4), "q"(n), "1"((long) to), "2"((long) from) - :"memory"); - return (to); -} -#else -#define __memcpy(a,b,c) memcpy(a,b,c) -#endif - - - -/* ================================================================ - * Debugging: - */ -#define DO_DEBUG 0 -#if DO_DEBUG -extern int INTEL_DEBUG; -#else -#define INTEL_DEBUG 0 -#endif - -#define DEBUG_TEXTURE 0x1 -#define DEBUG_STATE 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_BLIT 0x8 -#define DEBUG_MIPTREE 0x10 -#define DEBUG_FALLBACKS 0x20 -#define DEBUG_VERBOSE 0x40 -#define DEBUG_BATCH 0x80 -#define DEBUG_PIXEL 0x100 -#define DEBUG_BUFMGR 0x200 -#define DEBUG_REGION 0x400 -#define DEBUG_FBO 0x800 -#define DEBUG_LOCK 0x1000 - -#define DBG(...) do { if (INTEL_DEBUG & FILE_DEBUG_FLAG) _mesa_printf(__VA_ARGS__); } while(0) - - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GME 0x27AE -#define PCI_CHIP_G33_G 0x29C2 -#define PCI_CHIP_Q35_G 0x29B2 -#define PCI_CHIP_Q33_G 0x29D2 - - -/* ================================================================ - * intel_context.c: - */ - -extern GLboolean intelInitContext(struct intel_context *intel, - const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate, - struct dd_function_table *functions); - -extern void intelGetLock(struct intel_context *intel, GLuint flags); - -extern void intelFinish(GLcontext * ctx); -extern void intelFlush(GLcontext * ctx); - -extern void intelInitDriverFunctions(struct dd_function_table *functions); - - -/* ================================================================ - * intel_state.c: - */ -extern void intelInitStateFuncs(struct dd_function_table *functions); - -#define COMPAREFUNC_ALWAYS 0 -#define COMPAREFUNC_NEVER 0x1 -#define COMPAREFUNC_LESS 0x2 -#define COMPAREFUNC_EQUAL 0x3 -#define COMPAREFUNC_LEQUAL 0x4 -#define COMPAREFUNC_GREATER 0x5 -#define COMPAREFUNC_NOTEQUAL 0x6 -#define COMPAREFUNC_GEQUAL 0x7 - -#define STENCILOP_KEEP 0 -#define STENCILOP_ZERO 0x1 -#define STENCILOP_REPLACE 0x2 -#define STENCILOP_INCRSAT 0x3 -#define STENCILOP_DECRSAT 0x4 -#define STENCILOP_INCR 0x5 -#define STENCILOP_DECR 0x6 -#define STENCILOP_INVERT 0x7 - -#define LOGICOP_CLEAR 0 -#define LOGICOP_NOR 0x1 -#define LOGICOP_AND_INV 0x2 -#define LOGICOP_COPY_INV 0x3 -#define LOGICOP_AND_RVRSE 0x4 -#define LOGICOP_INV 0x5 -#define LOGICOP_XOR 0x6 -#define LOGICOP_NAND 0x7 -#define LOGICOP_AND 0x8 -#define LOGICOP_EQUIV 0x9 -#define LOGICOP_NOOP 0xa -#define LOGICOP_OR_INV 0xb -#define LOGICOP_COPY 0xc -#define LOGICOP_OR_RVRSE 0xd -#define LOGICOP_OR 0xe -#define LOGICOP_SET 0xf - -#define BLENDFACT_ZERO 0x01 -#define BLENDFACT_ONE 0x02 -#define BLENDFACT_SRC_COLR 0x03 -#define BLENDFACT_INV_SRC_COLR 0x04 -#define BLENDFACT_SRC_ALPHA 0x05 -#define BLENDFACT_INV_SRC_ALPHA 0x06 -#define BLENDFACT_DST_ALPHA 0x07 -#define BLENDFACT_INV_DST_ALPHA 0x08 -#define BLENDFACT_DST_COLR 0x09 -#define BLENDFACT_INV_DST_COLR 0x0a -#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b -#define BLENDFACT_CONST_COLOR 0x0c -#define BLENDFACT_INV_CONST_COLOR 0x0d -#define BLENDFACT_CONST_ALPHA 0x0e -#define BLENDFACT_INV_CONST_ALPHA 0x0f -#define BLENDFACT_MASK 0x0f - -#define MI_BATCH_BUFFER_END (0xA<<23) - - -extern int intel_translate_compare_func(GLenum func); -extern int intel_translate_stencil_op(GLenum op); -extern int intel_translate_blend_factor(GLenum factor); -extern int intel_translate_logic_op(GLenum opcode); - - -/*====================================================================== - * Inline conversion functions. - * These are better-typed than the macros used previously: - */ -static INLINE struct intel_context * -intel_context(GLcontext * ctx) -{ - return (struct intel_context *) ctx; -} - -static INLINE struct intel_texture_object * -intel_texture_object(struct gl_texture_object *obj) -{ - return (struct intel_texture_object *) obj; -} - -static INLINE struct intel_texture_image * -intel_texture_image(struct gl_texture_image *img) -{ - return (struct intel_texture_image *) img; -} - -extern struct intel_renderbuffer *intel_renderbuffer(struct gl_renderbuffer - *rb); - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_depthstencil.c b/src/mesa/drivers/dri/i915tex/intel_depthstencil.c deleted file mode 100644 index d269a85a3c..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_depthstencil.c +++ /dev/null @@ -1,282 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "imports.h" -#include "context.h" -#include "depthstencil.h" -#include "fbobject.h" -#include "framebuffer.h" -#include "hash.h" -#include "mtypes.h" -#include "renderbuffer.h" - -#include "intel_context.h" -#include "intel_fbo.h" -#include "intel_depthstencil.h" -#include "intel_regions.h" - - -/** - * The GL_EXT_framebuffer_object allows the user to create their own - * framebuffer objects consisting of color renderbuffers (0 or more), - * depth renderbuffers (0 or 1) and stencil renderbuffers (0 or 1). - * - * The spec considers depth and stencil renderbuffers to be totally independent - * buffers. In reality, most graphics hardware today uses a combined - * depth+stencil buffer (one 32-bit pixel = 24 bits of Z + 8 bits of stencil). - * - * This causes difficulty because the user may create some number of depth - * renderbuffers and some number of stencil renderbuffers and bind them - * together in framebuffers in any combination. - * - * This code manages all that. - * - * 1. Depth renderbuffers are always allocated in hardware as 32bpp - * GL_DEPTH24_STENCIL8 buffers. - * - * 2. Stencil renderbuffers are initially allocated in software as 8bpp - * GL_STENCIL_INDEX8 buffers. - * - * 3. Depth and Stencil renderbuffers use the PairedStencil and PairedDepth - * fields (respectively) to indicate if the buffer's currently paired - * with another stencil or depth buffer (respectively). - * - * 4. When a depth and stencil buffer are initially both attached to the - * current framebuffer, we merge the stencil buffer values into the - * depth buffer (really a depth+stencil buffer). The then hardware uses - * the combined buffer. - * - * 5. Whenever a depth or stencil buffer is reallocated (with - * glRenderbufferStorage) we undo the pairing and copy the stencil values - * from the combined depth/stencil buffer back to the stencil-only buffer. - * - * 6. We also undo the pairing when we find a change in buffer bindings. - * - * 7. If a framebuffer is only using a depth renderbuffer (no stencil), we - * just use the combined depth/stencil buffer and ignore the stencil values. - * - * 8. If a framebuffer is only using a stencil renderbuffer (no depth) we have - * to promote the 8bpp software stencil buffer to a 32bpp hardware - * depth+stencil buffer. - * - */ - - - -static void -map_regions(GLcontext * ctx, - struct intel_renderbuffer *depthRb, - struct intel_renderbuffer *stencilRb) -{ - struct intel_context *intel = intel_context(ctx); - if (depthRb && depthRb->region) { - intel_region_map(intel->intelScreen, depthRb->region); - depthRb->pfMap = depthRb->region->map; - depthRb->pfPitch = depthRb->region->pitch; - } - if (stencilRb && stencilRb->region) { - intel_region_map(intel->intelScreen, stencilRb->region); - stencilRb->pfMap = stencilRb->region->map; - stencilRb->pfPitch = stencilRb->region->pitch; - } -} - -static void -unmap_regions(GLcontext * ctx, - struct intel_renderbuffer *depthRb, - struct intel_renderbuffer *stencilRb) -{ - struct intel_context *intel = intel_context(ctx); - if (depthRb && depthRb->region) { - intel_region_unmap(intel->intelScreen, depthRb->region); - depthRb->pfMap = NULL; - depthRb->pfPitch = 0; - } - if (stencilRb && stencilRb->region) { - intel_region_unmap(intel->intelScreen, stencilRb->region); - stencilRb->pfMap = NULL; - stencilRb->pfPitch = 0; - } -} - - - -/** - * Undo the pairing/interleaving between depth and stencil buffers. - * irb should be a depth/stencil or stencil renderbuffer. - */ -void -intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb) -{ - if (irb->PairedStencil) { - /* irb is a depth/stencil buffer */ - struct gl_renderbuffer *stencilRb; - struct intel_renderbuffer *stencilIrb; - - ASSERT(irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - - stencilRb = _mesa_lookup_renderbuffer(ctx, irb->PairedStencil); - stencilIrb = intel_renderbuffer(stencilRb); - if (stencilIrb) { - /* need to extract stencil values from the depth buffer */ - ASSERT(stencilIrb->PairedDepth == irb->Base.Name); - map_regions(ctx, irb, stencilIrb); - _mesa_extract_stencil(ctx, &irb->Base, &stencilIrb->Base); - unmap_regions(ctx, irb, stencilIrb); - stencilIrb->PairedDepth = 0; - } - irb->PairedStencil = 0; - } - else if (irb->PairedDepth) { - /* irb is a stencil buffer */ - struct gl_renderbuffer *depthRb; - struct intel_renderbuffer *depthIrb; - - ASSERT(irb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || - irb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - - depthRb = _mesa_lookup_renderbuffer(ctx, irb->PairedDepth); - depthIrb = intel_renderbuffer(depthRb); - if (depthIrb) { - /* need to extract stencil values from the depth buffer */ - ASSERT(depthIrb->PairedStencil == irb->Base.Name); - map_regions(ctx, depthIrb, irb); - _mesa_extract_stencil(ctx, &depthIrb->Base, &irb->Base); - unmap_regions(ctx, depthIrb, irb); - depthIrb->PairedStencil = 0; - } - irb->PairedDepth = 0; - } - else { - _mesa_problem(ctx, "Problem in undo_depth_stencil_pairing"); - } - - ASSERT(irb->PairedStencil == 0); - ASSERT(irb->PairedDepth == 0); -} - - -/** - * Examine the depth and stencil renderbuffers which are attached to the - * framebuffer. If both depth and stencil are attached, make sure that the - * renderbuffers are 'paired' (combined). If only depth or only stencil is - * attached, undo any previous pairing. - * - * Must be called if NewState & _NEW_BUFFER (when renderbuffer attachments - * change, for example). - */ -void -intel_validate_paired_depth_stencil(GLcontext * ctx, - struct gl_framebuffer *fb) -{ - struct intel_renderbuffer *depthRb, *stencilRb; - - depthRb = intel_get_renderbuffer(fb, BUFFER_DEPTH); - stencilRb = intel_get_renderbuffer(fb, BUFFER_STENCIL); - - if (depthRb && stencilRb) { - if (depthRb == stencilRb) { - /* Using a user-created combined depth/stencil buffer. - * Nothing to do. - */ - ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_STENCIL_EXT); - ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - } - else { - /* Separate depth/stencil buffers, need to interleave now */ - ASSERT(depthRb->Base._BaseFormat == GL_DEPTH_COMPONENT); - ASSERT(stencilRb->Base._BaseFormat == GL_STENCIL_INDEX); - /* may need to interleave depth/stencil now */ - if (depthRb->PairedStencil == stencilRb->Base.Name) { - /* OK, the depth and stencil buffers are already interleaved */ - ASSERT(stencilRb->PairedDepth == depthRb->Base.Name); - } - else { - /* need to setup new pairing/interleaving */ - if (depthRb->PairedStencil) { - intel_unpair_depth_stencil(ctx, depthRb); - } - if (stencilRb->PairedDepth) { - intel_unpair_depth_stencil(ctx, stencilRb); - } - - ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - ASSERT(stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT || - stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - - /* establish new pairing: interleave stencil into depth buffer */ - map_regions(ctx, depthRb, stencilRb); - _mesa_insert_stencil(ctx, &depthRb->Base, &stencilRb->Base); - unmap_regions(ctx, depthRb, stencilRb); - depthRb->PairedStencil = stencilRb->Base.Name; - stencilRb->PairedDepth = depthRb->Base.Name; - } - - } - } - else if (depthRb) { - /* Depth buffer but no stencil buffer. - * We'll use a GL_DEPTH24_STENCIL8 buffer and ignore the stencil bits. - */ - /* can't assert this until storage is allocated: - ASSERT(depthRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - */ - /* intel_undo any previous pairing */ - if (depthRb->PairedStencil) { - intel_unpair_depth_stencil(ctx, depthRb); - } - } - else if (stencilRb) { - /* Stencil buffer but no depth buffer. - * Since h/w doesn't typically support just 8bpp stencil w/out Z, - * we'll use a GL_DEPTH24_STENCIL8 buffer and ignore the depth bits. - */ - /* undo any previous pairing */ - if (stencilRb->PairedDepth) { - intel_unpair_depth_stencil(ctx, stencilRb); - } - if (stencilRb->Base._ActualFormat == GL_STENCIL_INDEX8_EXT) { - /* promote buffer to GL_DEPTH24_STENCIL8 for hw rendering */ - _mesa_promote_stencil(ctx, &stencilRb->Base); - ASSERT(stencilRb->Base._ActualFormat == GL_DEPTH24_STENCIL8_EXT); - } - } - - /* Finally, update the fb->_DepthBuffer and fb->_StencilBuffer fields */ - _mesa_update_depth_buffer(ctx, fb, BUFFER_DEPTH); - if (depthRb && depthRb->PairedStencil) - _mesa_update_stencil_buffer(ctx, fb, BUFFER_DEPTH); - else - _mesa_update_stencil_buffer(ctx, fb, BUFFER_STENCIL); - - - /* The hardware should use fb->Attachment[BUFFER_DEPTH].Renderbuffer - * first, if present, then fb->Attachment[BUFFER_STENCIL].Renderbuffer - * if present. - */ -} diff --git a/src/mesa/drivers/dri/i915tex/intel_depthstencil.h b/src/mesa/drivers/dri/i915tex/intel_depthstencil.h deleted file mode 100644 index 2d3fc48b3a..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_depthstencil.h +++ /dev/null @@ -1,14 +0,0 @@ - -#ifndef INTEL_DEPTH_STENCIL_H -#define INTEL_DEPTH_STENCIL_H - - -extern void -intel_unpair_depth_stencil(GLcontext * ctx, struct intel_renderbuffer *irb); - -extern void -intel_validate_paired_depth_stencil(GLcontext * ctx, - struct gl_framebuffer *fb); - - -#endif /* INTEL_DEPTH_STENCIL_H */ diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.c b/src/mesa/drivers/dri/i915tex/intel_fbo.c deleted file mode 100644 index 928139b5d9..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.c +++ /dev/null @@ -1,621 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "imports.h" -#include "mtypes.h" -#include "fbobject.h" -#include "framebuffer.h" -#include "renderbuffer.h" -#include "context.h" -#include "texformat.h" -#include "texrender.h" - -#include "intel_context.h" -#include "intel_buffers.h" -#include "intel_depthstencil.h" -#include "intel_fbo.h" -#include "intel_mipmap_tree.h" -#include "intel_regions.h" -#include "intel_span.h" - - -#define FILE_DEBUG_FLAG DEBUG_FBO - -#define INTEL_RB_CLASS 0x12345678 - - -/* XXX FBO: move this to intel_context.h (inlined) */ -/** - * Return a gl_renderbuffer ptr casted to intel_renderbuffer. - * NULL will be returned if the rb isn't really an intel_renderbuffer. - * This is determiend by checking the ClassID. - */ -struct intel_renderbuffer * -intel_renderbuffer(struct gl_renderbuffer *rb) -{ - struct intel_renderbuffer *irb = (struct intel_renderbuffer *) rb; - if (irb && irb->Base.ClassID == INTEL_RB_CLASS) { - /*_mesa_warning(NULL, "Returning non-intel Rb\n");*/ - return irb; - } - else - return NULL; -} - - -struct intel_renderbuffer * -intel_get_renderbuffer(struct gl_framebuffer *fb, GLuint attIndex) -{ - return intel_renderbuffer(fb->Attachment[attIndex].Renderbuffer); -} - - -void -intel_flip_renderbuffers(struct intel_framebuffer *intel_fb) -{ - int current_page = intel_fb->pf_current_page; - int next_page = (current_page + 1) % intel_fb->pf_num_pages; - struct gl_renderbuffer *tmp_rb; - - /* Exchange renderbuffers if necessary but make sure their reference counts - * are preserved. - */ - if (intel_fb->color_rb[current_page] && - intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer != - &intel_fb->color_rb[current_page]->Base) { - tmp_rb = NULL; - _mesa_reference_renderbuffer(&tmp_rb, - intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer); - tmp_rb = &intel_fb->color_rb[current_page]->Base; - _mesa_reference_renderbuffer( - &intel_fb->Base.Attachment[BUFFER_FRONT_LEFT].Renderbuffer, tmp_rb); - _mesa_reference_renderbuffer(&tmp_rb, NULL); - } - - if (intel_fb->color_rb[next_page] && - intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer != - &intel_fb->color_rb[next_page]->Base) { - tmp_rb = NULL; - _mesa_reference_renderbuffer(&tmp_rb, - intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer); - tmp_rb = &intel_fb->color_rb[next_page]->Base; - _mesa_reference_renderbuffer( - &intel_fb->Base.Attachment[BUFFER_BACK_LEFT].Renderbuffer, tmp_rb); - _mesa_reference_renderbuffer(&tmp_rb, NULL); - } -} - - -struct intel_region * -intel_get_rb_region(struct gl_framebuffer *fb, GLuint attIndex) -{ - struct intel_renderbuffer *irb = intel_get_renderbuffer(fb, attIndex); - - if (irb) - return irb->region; - else - return NULL; -} - - - -/** - * Create a new framebuffer object. - */ -static struct gl_framebuffer * -intel_new_framebuffer(GLcontext * ctx, GLuint name) -{ - /* Only drawable state in intel_framebuffer at this time, just use Mesa's - * class - */ - return _mesa_new_framebuffer(ctx, name); -} - - -static void -intel_delete_renderbuffer(struct gl_renderbuffer *rb) -{ - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); - struct intel_renderbuffer *irb = intel_renderbuffer(rb); - - ASSERT(irb); - - DBG("freeing renderbuffer\n"); - - if (irb->PairedStencil || irb->PairedDepth) { - intel_unpair_depth_stencil(ctx, irb); - } - - if (intel && irb->region) { - intel_region_release(&irb->region); - } - - _mesa_free(irb); -} - - - -/** - * Return a pointer to a specific pixel in a renderbuffer. - */ -static void * -intel_get_pointer(GLcontext * ctx, struct gl_renderbuffer *rb, - GLint x, GLint y) -{ - /* By returning NULL we force all software rendering to go through - * the span routines. - */ - return NULL; -} - - - -/** - * Called via glRenderbufferStorageEXT() to set the format and allocate - * storage for a user-created (or priv buffer) renderbuffer. - */ -static GLboolean -intel_alloc_renderbuffer_storage(GLcontext * ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, - GLuint width, GLuint height) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_renderbuffer *irb = intel_renderbuffer(rb); - GLboolean softwareBuffer = GL_FALSE; - int cpp; - - switch (internalFormat) { - case GL_R3_G3_B2: - case GL_RGB4: - case GL_RGB5: - rb->_ActualFormat = GL_RGB5; - rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 5; - rb->GreenBits = 6; - rb->BlueBits = 5; - cpp = 2; - break; - case GL_RGB: - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - case GL_RGBA: - case GL_RGBA2: - case GL_RGBA4: - case GL_RGB5_A1: - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - rb->_ActualFormat = GL_RGBA8; - rb->DataType = GL_UNSIGNED_BYTE; - rb->RedBits = 8; - rb->GreenBits = 8; - rb->BlueBits = 8; - rb->AlphaBits = 8; - cpp = 4; - break; - case GL_STENCIL_INDEX: - case GL_STENCIL_INDEX1_EXT: - case GL_STENCIL_INDEX4_EXT: - case GL_STENCIL_INDEX8_EXT: - case GL_STENCIL_INDEX16_EXT: - /* alloc a depth+stencil buffer */ - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->StencilBits = 8; - cpp = 4; - break; - case GL_DEPTH_COMPONENT16: - rb->_ActualFormat = GL_DEPTH_COMPONENT16; - rb->DataType = GL_UNSIGNED_SHORT; - rb->DepthBits = 16; - cpp = 2; - break; - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - cpp = 4; - break; - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - rb->_ActualFormat = GL_DEPTH24_STENCIL8_EXT; - rb->DataType = GL_UNSIGNED_INT_24_8_EXT; - rb->DepthBits = 24; - rb->StencilBits = 8; - cpp = 4; - break; - default: - _mesa_problem(ctx, - "Unexpected format (%x) in intel_alloc_renderbuffer_storage", internalFormat); - return GL_FALSE; - } - - intelFlush(ctx); - - /* free old region */ - if (irb->region) { - intel_region_release(&irb->region); - } - - /* allocate new memory region/renderbuffer */ - if (softwareBuffer) { - return _mesa_soft_renderbuffer_storage(ctx, rb, internalFormat, - width, height); - } - else { - /* Choose a pitch to match hardware requirements: - */ - GLuint pitch = ((cpp * width + 63) & ~63) / cpp; - - /* alloc hardware renderbuffer */ - DBG("Allocating %d x %d Intel RBO (pitch %d)\n", width, - height, pitch); - - irb->region = intel_region_alloc(intel->intelScreen, cpp, pitch, height); - if (!irb->region) - return GL_FALSE; /* out of memory? */ - - ASSERT(irb->region->buffer); - - rb->Width = width; - rb->Height = height; - -#if 1 - /* update the surface's size too */ - rb->surface->width = width; - rb->surface->height = height; -#endif - - /* This sets the Get/PutRow/Value functions */ - intel_set_span_functions(&irb->Base); - - return GL_TRUE; - } -} - - -static void -intel_resize_buffers(GLcontext *ctx, struct gl_framebuffer *fb, - GLuint width, GLuint height) -{ - struct intel_framebuffer *intel_fb = (struct intel_framebuffer*)fb; - int i; - - _mesa_resize_framebuffer(ctx, fb, width, height); - - fb->Initialized = GL_TRUE; /* XXX remove someday */ - - if (fb->Name != 0) { - return; - } - - /* Make sure all window system renderbuffers are up to date */ - for (i = 0; i < 3; i++) { - struct gl_renderbuffer *rb = &intel_fb->color_rb[i]->Base; - - /* only resize if size is changing */ - if (rb && (rb->Width != width || rb->Height != height)) { - rb->AllocStorage(ctx, rb, rb->InternalFormat, width, height); - } - } -} - -static GLboolean -intel_nop_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, - GLenum internalFormat, GLuint width, GLuint height) -{ - _mesa_problem(ctx, "intel_op_alloc_storage should never be called."); - return GL_FALSE; -} - - - -struct intel_renderbuffer * -intel_new_renderbuffer_fb(GLuint intFormat) -{ - struct intel_renderbuffer *irb; - - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(NULL, GL_OUT_OF_MEMORY, "creating renderbuffer"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, 0); - irb->Base.ClassID = INTEL_RB_CLASS; - irb->Base.InternalFormat = intFormat; - - switch (intFormat) { - case GL_RGB5: - case GL_RGBA8: - irb->Base._BaseFormat = GL_RGBA; - break; - case GL_DEPTH_COMPONENT16: - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - break; - case GL_DEPTH24_STENCIL8_EXT: - irb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; - break; - default: - _mesa_problem(NULL, - "Unexpected intFormat in intel_create_renderbuffer"); - return NULL; - } - - /* intel-specific methods */ - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; - irb->Base.GetPointer = intel_get_pointer; - /* span routines set in alloc_storage function */ - - irb->Base.surface = intel_new_surface(intFormat); - irb->Base.surface->rb = irb; - - return irb; -} - -/** - * Create a new renderbuffer object. - * Typically called via glBindRenderbufferEXT(). - */ -static struct gl_renderbuffer * -intel_new_renderbuffer(GLcontext * ctx, GLuint name) -{ - /*struct intel_context *intel = intel_context(ctx); */ - struct intel_renderbuffer *irb; - - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "creating renderbuffer"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, name); - irb->Base.ClassID = INTEL_RB_CLASS; - - /* intel-specific methods */ - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_alloc_renderbuffer_storage; - irb->Base.GetPointer = intel_get_pointer; - /* span routines set in alloc_storage function */ - - irb->Base.surface = intel_new_surface(0 /*unknown format*/); - irb->Base.surface->rb = irb; - - return &irb->Base; -} - - -/** - * Called via glBindFramebufferEXT(). - */ -static void -intel_bind_framebuffer(GLcontext * ctx, GLenum target, - struct gl_framebuffer *fb, struct gl_framebuffer *fbread) -{ - if (target == GL_FRAMEBUFFER_EXT || target == GL_DRAW_FRAMEBUFFER_EXT) { - intel_draw_buffer(ctx, fb); - /* Integer depth range depends on depth buffer bits */ - ctx->Driver.DepthRange(ctx, ctx->Viewport.Near, ctx->Viewport.Far); - } - else { - /* don't need to do anything if target == GL_READ_FRAMEBUFFER_EXT */ - } -} - - -/** - * Called via glFramebufferRenderbufferEXT(). - */ -static void -intel_framebuffer_renderbuffer(GLcontext * ctx, - struct gl_framebuffer *fb, - GLenum attachment, struct gl_renderbuffer *rb) -{ - DBG("Intel FramebufferRenderbuffer %u %u\n", fb->Name, rb ? rb->Name : 0); - - intelFlush(ctx); - - _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb); - intel_draw_buffer(ctx, fb); -} - - -/** - * When glFramebufferTexture[123]D is called this function sets up the - * gl_renderbuffer wrapper around the texture image. - * This will have the region info needed for hardware rendering. - */ -static struct intel_renderbuffer * -intel_wrap_texture(GLcontext * ctx, struct gl_texture_image *texImage) -{ - const GLuint name = ~0; /* not significant, but distinct for debugging */ - struct intel_renderbuffer *irb; - - /* make an intel_renderbuffer to wrap the texture image */ - irb = CALLOC_STRUCT(intel_renderbuffer); - if (!irb) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glFramebufferTexture"); - return NULL; - } - - _mesa_init_renderbuffer(&irb->Base, name); - irb->Base.ClassID = INTEL_RB_CLASS; - - if (texImage->TexFormat == &_mesa_texformat_argb8888) { - irb->Base._ActualFormat = GL_RGBA8; - irb->Base._BaseFormat = GL_RGBA; - DBG("Render to RGBA8 texture OK\n"); - } - else if (texImage->TexFormat == &_mesa_texformat_rgb565) { - irb->Base._ActualFormat = GL_RGB5; - irb->Base._BaseFormat = GL_RGB; - DBG("Render to RGB5 texture OK\n"); - } - else if (texImage->TexFormat == &_mesa_texformat_z16) { - irb->Base._ActualFormat = GL_DEPTH_COMPONENT16; - irb->Base._BaseFormat = GL_DEPTH_COMPONENT; - DBG("Render to DEPTH16 texture OK\n"); - } - else { - DBG("Render to texture BAD FORMAT %d\n", - texImage->TexFormat->MesaFormat); - _mesa_free(irb); - return NULL; - } - - irb->Base.InternalFormat = irb->Base._ActualFormat; - irb->Base.Width = texImage->Width; - irb->Base.Height = texImage->Height; - irb->Base.DataType = GL_UNSIGNED_BYTE; /* FBO XXX fix */ - irb->Base.RedBits = texImage->TexFormat->RedBits; - irb->Base.GreenBits = texImage->TexFormat->GreenBits; - irb->Base.BlueBits = texImage->TexFormat->BlueBits; - irb->Base.AlphaBits = texImage->TexFormat->AlphaBits; - irb->Base.DepthBits = texImage->TexFormat->DepthBits; - - irb->Base.Delete = intel_delete_renderbuffer; - irb->Base.AllocStorage = intel_nop_alloc_storage; - intel_set_span_functions(&irb->Base); - - irb->RenderToTexture = GL_TRUE; - - return irb; -} - - -/** - * Called by glFramebufferTexture[123]DEXT() (and other places) to - * prepare for rendering into texture memory. This might be called - * many times to choose different texture levels, cube faces, etc - * before intel_finish_render_texture() is ever called. - */ -static void -intel_render_texture(GLcontext * ctx, - struct gl_framebuffer *fb, - struct gl_renderbuffer_attachment *att) -{ - struct gl_texture_image *newImage - = att->Texture->Image[att->CubeMapFace][att->TextureLevel]; - struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); - struct intel_texture_image *intel_image; - GLuint imageOffset; - - (void) fb; - - ASSERT(newImage); - - if (!irb) { - irb = intel_wrap_texture(ctx, newImage); - if (irb) { - /* bind the wrapper to the attachment point */ - _mesa_reference_renderbuffer(&att->Renderbuffer, &irb->Base); - } - else { - /* fallback to software rendering */ - _mesa_render_texture(ctx, fb, att); - return; - } - } - - DBG("Begin render texture tid %x tex=%u w=%d h=%d refcount=%d\n", - _glthread_GetID(), - att->Texture->Name, newImage->Width, newImage->Height, - irb->Base.RefCount); - - /* point the renderbufer's region to the texture image region */ - intel_image = intel_texture_image(newImage); - if (irb->region != intel_image->mt->region) { - if (irb->region) - intel_region_release(&irb->region); - intel_region_reference(&irb->region, intel_image->mt->region); - } - - /* compute offset of the particular 2D image within the texture region */ - imageOffset = intel_miptree_image_offset(intel_image->mt, - att->CubeMapFace, - att->TextureLevel); - - if (att->Texture->Target == GL_TEXTURE_3D) { - const GLuint *offsets = intel_miptree_depth_offsets(intel_image->mt, - att->TextureLevel); - imageOffset += offsets[att->Zoffset]; - } - - /* store that offset in the region */ - intel_image->mt->region->draw_offset = imageOffset; - - /* update drawing region, etc */ - intel_draw_buffer(ctx, fb); -} - - -/** - * Called by Mesa when rendering to a texture is done. - */ -static void -intel_finish_render_texture(GLcontext * ctx, - struct gl_renderbuffer_attachment *att) -{ - struct intel_renderbuffer *irb = intel_renderbuffer(att->Renderbuffer); - - DBG("End render texture (tid %x) tex %u\n", _glthread_GetID(), att->Texture->Name); - - if (irb) { - /* just release the region */ - intel_region_release(&irb->region); - } - else if (att->Renderbuffer) { - /* software fallback */ - _mesa_finish_render_texture(ctx, att); - /* XXX FBO: Need to unmap the buffer (or in intelSpanRenderStart???) */ - } -} - - -/** - * Do one-time context initializations related to GL_EXT_framebuffer_object. - * Hook in device driver functions. - */ -void -intel_fbo_init(struct intel_context *intel) -{ - intel->ctx.Driver.NewFramebuffer = intel_new_framebuffer; - intel->ctx.Driver.NewRenderbuffer = intel_new_renderbuffer; - intel->ctx.Driver.BindFramebuffer = intel_bind_framebuffer; - intel->ctx.Driver.FramebufferRenderbuffer = intel_framebuffer_renderbuffer; - intel->ctx.Driver.RenderTexture = intel_render_texture; - intel->ctx.Driver.FinishRenderTexture = intel_finish_render_texture; - intel->ctx.Driver.ResizeBuffers = intel_resize_buffers; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_fbo.h b/src/mesa/drivers/dri/i915tex/intel_fbo.h deleted file mode 100644 index 7dc3967533..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_fbo.h +++ /dev/null @@ -1,127 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_FBO_H -#define INTEL_FBO_H - - -#include "pipe/p_state.h" -#include "pipe/softpipe/sp_surface.h" - - -struct intel_context; -struct intel_region; - - -/** - * Intel framebuffer, derived from gl_framebuffer. - */ -struct intel_framebuffer -{ - struct gl_framebuffer Base; - - struct intel_renderbuffer *color_rb[3]; - - /* Drawable page flipping state */ - GLboolean pf_active; - GLuint pf_seq; - GLint pf_pipes; - GLint pf_current_page; - GLint pf_num_pages; - - /* VBI - */ - GLuint vbl_seq; - GLuint vblank_flags; - GLuint vbl_waited; - - int64_t swap_ust; - int64_t swap_missed_ust; - - GLuint swap_count; - GLuint swap_missed_count; -}; - - -/** - * Intel renderbuffer, derived from gl_renderbuffer. - * Note: The PairedDepth and PairedStencil fields use renderbuffer IDs, - * not pointers because in some circumstances a deleted renderbuffer could - * result in a dangling pointer here. - */ -struct intel_renderbuffer -{ - struct gl_renderbuffer Base; - struct intel_region *region; - void *pfMap; /* possibly paged flipped map pointer */ - GLuint pfPitch; /* possibly paged flipped pitch */ - GLboolean RenderToTexture; /* RTT? */ - - GLuint PairedDepth; /**< only used if this is a depth renderbuffer */ - GLuint PairedStencil; /**< only used if this is a stencil renderbuffer */ - - GLuint pf_pending; /**< sequence number of pending flip */ - - GLuint vbl_pending; /**< vblank sequence number of pending flip */ - - struct intel_surface *surface; -}; - -#if 0 -extern struct intel_renderbuffer *intel_create_renderbuffer(GLenum intFormat, - GLsizei width, - GLsizei height, - int offset, - int pitch, - int cpp, - void *map); -#endif - -extern struct intel_renderbuffer *intel_new_renderbuffer_fb(GLuint intFormat); - -extern void intel_fbo_init(struct intel_context *intel); - - -/* XXX make inline or macro */ -extern struct intel_renderbuffer *intel_get_renderbuffer(struct gl_framebuffer - *fb, - GLuint attIndex); - -extern void intel_flip_renderbuffers(struct intel_framebuffer *intel_fb); - - -/* XXX make inline or macro */ -extern struct intel_region *intel_get_rb_region(struct gl_framebuffer *fb, - GLuint attIndex); - - - -extern struct pipe_surface * -intel_new_surface(GLuint intFormat); - - -#endif /* INTEL_FBO_H */ diff --git a/src/mesa/drivers/dri/i915tex/intel_ioctl.c b/src/mesa/drivers/dri/i915tex/intel_ioctl.c deleted file mode 100644 index 54721c78ab..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_ioctl.c +++ /dev/null @@ -1,137 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include -#include -#include -#include - -#include "mtypes.h" -#include "context.h" -#include "swrast/swrast.h" - -#include "intel_context.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "intel_regions.h" -#include "drm.h" - -#define FILE_DEBUG_FLAG DEBUG_IOCTL - -int -intelEmitIrqLocked(struct intel_context *intel) -{ - drmI830IrqEmit ie; - int ret, seq; - - assert(((*(int *) intel->driHwLock) & ~DRM_LOCK_CONT) == - (DRM_LOCK_HELD | intel->hHWContext)); - - ie.irq_seq = &seq; - - ret = drmCommandWriteRead(intel->driFd, DRM_I830_IRQ_EMIT, - &ie, sizeof(ie)); - if (ret) { - fprintf(stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret); - exit(1); - } - - DBG("%s --> %d\n", __FUNCTION__, seq); - - return seq; -} - -void -intelWaitIrq(struct intel_context *intel, int seq) -{ - int ret; - - DBG("%s %d\n", __FUNCTION__, seq); - - intel->iw.irq_seq = seq; - - do { - ret = - drmCommandWrite(intel->driFd, DRM_I830_IRQ_WAIT, &intel->iw, - sizeof(intel->iw)); - } while (ret == -EAGAIN || ret == -EINTR); - - if (ret) { - fprintf(stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret); - exit(1); - } -} - - -void -intel_batch_ioctl(struct intel_context *intel, - GLuint start_offset, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) -{ - drmI830BatchBuffer batch; - - assert(intel->locked); - assert(used); - - DBG("%s used %d offset %x..%x ignore_cliprects %d\n", - __FUNCTION__, - used, start_offset, start_offset + used, ignore_cliprects); - - /* Throw away non-effective packets. Won't work once we have - * hardware contexts which would preserve statechanges beyond a - * single buffer. - */ - - - - batch.start = start_offset; - batch.used = used; - batch.cliprects = intel->pClipRects; - batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; - batch.DR1 = 0; - batch.DR4 = 0; /* still need this ? */ - - DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n", - __FUNCTION__, - batch.start, - batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); - - if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, - sizeof(batch))) { - fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n", -errno); - UNLOCK_HARDWARE(intel); - exit(1); - } - - /* FIXME: use hardware contexts to avoid 'losing' hardware after - * each buffer flush. - */ - intel->vtbl.lost_hardware(intel); -} diff --git a/src/mesa/drivers/dri/i915tex/intel_ioctl.h b/src/mesa/drivers/dri/i915tex/intel_ioctl.h deleted file mode 100644 index e8d07de893..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_ioctl.h +++ /dev/null @@ -1,40 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_IOCTL_H -#define INTEL_IOCTL_H - -#include "intel_context.h" - -void intelWaitIrq(struct intel_context *intel, int seq); -int intelEmitIrqLocked(struct intel_context *intel); - -void intel_batch_ioctl(struct intel_context *intel, - GLuint start_offset, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock); -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c b/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c deleted file mode 100644 index ab3ec71f9e..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.c +++ /dev/null @@ -1,357 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_context.h" -#include "intel_mipmap_tree.h" -#include "intel_regions.h" -#include "enums.h" - -#define FILE_DEBUG_FLAG DEBUG_MIPTREE - -static GLenum -target_to_target(GLenum target) -{ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - return GL_TEXTURE_CUBE_MAP_ARB; - default: - return target; - } -} - -struct intel_mipmap_tree * -intel_miptree_create(struct intel_context *intel, - GLenum target, - GLenum internal_format, - GLuint first_level, - GLuint last_level, - GLuint width0, - GLuint height0, - GLuint depth0, GLuint cpp, GLuint compress_byte) -{ - GLboolean ok; - struct intel_mipmap_tree *mt = calloc(sizeof(*mt), 1); - - DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), - _mesa_lookup_enum_by_nr(internal_format), first_level, last_level); - - mt->target = target_to_target(target); - mt->internal_format = internal_format; - mt->first_level = first_level; - mt->last_level = last_level; - mt->width0 = width0; - mt->height0 = height0; - mt->depth0 = depth0; - mt->cpp = compress_byte ? compress_byte : cpp; - mt->compressed = compress_byte ? 1 : 0; - mt->refcount = 1; - - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - ok = i945_miptree_layout(mt); - break; - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - case PCI_CHIP_I830_M: - case PCI_CHIP_I855_GM: - case PCI_CHIP_I865_G: - default: - /* All the i830 chips and the i915 use this layout: - */ - ok = i915_miptree_layout(mt); - break; - } - - if (ok) - mt->region = intel_region_alloc(intel->intelScreen, - mt->cpp, mt->pitch, mt->total_height); - - if (!mt->region) { - free(mt); - return NULL; - } - - return mt; -} - - -void -intel_miptree_reference(struct intel_mipmap_tree **dst, - struct intel_mipmap_tree *src) -{ - src->refcount++; - *dst = src; - DBG("%s %p refcount now %d\n", __FUNCTION__, src, src->refcount); -} - -void -intel_miptree_release(struct intel_context *intel, - struct intel_mipmap_tree **mt) -{ - if (!*mt) - return; - - DBG("%s %p refcount will be %d\n", __FUNCTION__, *mt, (*mt)->refcount - 1); - if (--(*mt)->refcount <= 0) { - GLuint i; - - DBG("%s deleting %p\n", __FUNCTION__, *mt); - - intel_region_release(&((*mt)->region)); - - for (i = 0; i < MAX_TEXTURE_LEVELS; i++) - if ((*mt)->level[i].image_offset) - free((*mt)->level[i].image_offset); - - free(*mt); - } - *mt = NULL; -} - - - - -/* Can the image be pulled into a unified mipmap tree. This mirrors - * the completeness test in a lot of ways. - * - * Not sure whether I want to pass gl_texture_image here. - */ -GLboolean -intel_miptree_match_image(struct intel_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level) -{ - /* Images with borders are never pulled into mipmap trees. - */ - if (image->Border) - return GL_FALSE; - - if (image->InternalFormat != mt->internal_format || - image->IsCompressed != mt->compressed) - return GL_FALSE; - - /* Test image dimensions against the base level image adjusted for - * minification. This will also catch images not present in the - * tree, changed targets, etc. - */ - if (image->Width != mt->level[level].width || - image->Height != mt->level[level].height || - image->Depth != mt->level[level].depth) - return GL_FALSE; - - return GL_TRUE; -} - - -void -intel_miptree_set_level_info(struct intel_mipmap_tree *mt, - GLuint level, - GLuint nr_images, - GLuint x, GLuint y, GLuint w, GLuint h, GLuint d) -{ - - mt->level[level].width = w; - mt->level[level].height = h; - mt->level[level].depth = d; - mt->level[level].level_offset = (x + y * mt->pitch) * mt->cpp; - mt->level[level].nr_images = nr_images; - - DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - level, w, h, d, x, y, mt->level[level].level_offset); - - /* Not sure when this would happen, but anyway: - */ - if (mt->level[level].image_offset) { - free(mt->level[level].image_offset); - mt->level[level].image_offset = NULL; - } - - assert(nr_images); - - mt->level[level].image_offset = malloc(nr_images * sizeof(GLuint)); - mt->level[level].image_offset[0] = 0; -} - - - -void -intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, - GLuint level, GLuint img, GLuint x, GLuint y) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < mt->level[level].nr_images); - - mt->level[level].image_offset[img] = (x + y * mt->pitch); - - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, mt->level[level].image_offset[img]); -} - - -/* Although we use the image_offset[] array to store relative offsets - * to cube faces, Mesa doesn't know anything about this and expects - * each cube face to be treated as a separate image. - * - * These functions present that view to mesa: - */ -const GLuint * -intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, GLuint level) -{ - static const GLuint zero = 0; - - if (mt->target != GL_TEXTURE_3D || mt->level[level].nr_images == 1) - return &zero; - else - return mt->level[level].image_offset; -} - - -GLuint -intel_miptree_image_offset(struct intel_mipmap_tree * mt, - GLuint face, GLuint level) -{ - if (mt->target == GL_TEXTURE_CUBE_MAP_ARB) - return (mt->level[level].level_offset + - mt->level[level].image_offset[face] * mt->cpp); - else - return mt->level[level].level_offset; -} - - - -/** - * Map a teximage in a mipmap tree. - * \param row_stride returns row stride in bytes - * \param image_offsets returns array [texture.depth] of image offsets - * (in bytes) for 3D textures. - * \return address of mapping - */ -GLubyte * -intel_miptree_image_map(struct intel_context * intel, - struct intel_mipmap_tree * mt, - GLuint face, - GLuint level, - GLuint * row_stride, GLuint image_offsets[]) -{ - DBG("%s \n", __FUNCTION__); - - if (row_stride) - *row_stride = mt->pitch * mt->cpp; - - if (image_offsets) - memcpy(image_offsets, mt->level[level].image_offset, - mt->level[level].depth * sizeof(GLuint)); - - return (intel_region_map(intel->intelScreen, mt->region) + - intel_miptree_image_offset(mt, face, level)); -} - -void -intel_miptree_image_unmap(struct intel_context *intel, - struct intel_mipmap_tree *mt) -{ - DBG("%s\n", __FUNCTION__); - intel_region_unmap(intel->intelScreen, mt->region); -} - - - -/* Upload data for a particular image. - */ -void -intel_miptree_image_data(struct intel_context *intel, - struct intel_mipmap_tree *dst, - GLuint face, - GLuint level, - void *src, - GLuint src_row_pitch, GLuint src_image_pitch) -{ - GLuint depth = dst->level[level].depth; - GLuint dst_offset = intel_miptree_image_offset(dst, face, level); - const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); - GLuint i; - GLuint height = 0; - - DBG("%s\n", __FUNCTION__); - for (i = 0; i < depth; i++) { - height = dst->level[level].height; - if(dst->compressed) - height /= 4; - intel_region_data(intel->intelScreen, dst->region, - dst_offset + dst_depth_offset[i], /* dst_offset */ - 0, 0, /* dstx, dsty */ - src, - src_row_pitch, - 0, 0, /* source x, y */ - dst->level[level].width, height); /* width, height */ - - src += src_image_pitch * dst->cpp; - } -} - -/* Copy mipmap image between trees - */ -void -intel_miptree_image_copy(struct intel_context *intel, - struct intel_mipmap_tree *dst, - GLuint face, GLuint level, - struct intel_mipmap_tree *src) -{ - GLuint width = src->level[level].width; - GLuint height = src->level[level].height; - GLuint depth = src->level[level].depth; - GLuint dst_offset = intel_miptree_image_offset(dst, face, level); - GLuint src_offset = intel_miptree_image_offset(src, face, level); - const GLuint *dst_depth_offset = intel_miptree_depth_offsets(dst, level); - const GLuint *src_depth_offset = intel_miptree_depth_offsets(src, level); - GLuint i; - - if (dst->compressed) - height /= 4; - for (i = 0; i < depth; i++) { - intel_region_copy(intel->intelScreen, - dst->region, dst_offset + dst_depth_offset[i], - 0, - 0, - src->region, src_offset + src_depth_offset[i], - 0, 0, width, height); - } - -} diff --git a/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h b/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h deleted file mode 100644 index 675a438c20..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_mipmap_tree.h +++ /dev/null @@ -1,198 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_MIPMAP_TREE_H -#define INTEL_MIPMAP_TREE_H - -#include "intel_regions.h" - -/* A layer on top of the intel_regions code which adds: - * - * - Code to size and layout a region to hold a set of mipmaps. - * - Query to determine if a new image fits in an existing tree. - * - More refcounting - * - maybe able to remove refcounting from intel_region? - * - ? - * - * The fixed mipmap layout of intel hardware where one offset - * specifies the position of all images in a mipmap hierachy - * complicates the implementation of GL texture image commands, - * compared to hardware where each image is specified with an - * independent offset. - * - * In an ideal world, each texture object would be associated with a - * single bufmgr buffer or 2d intel_region, and all the images within - * the texture object would slot into the tree as they arrive. The - * reality can be a little messier, as images can arrive from the user - * with sizes that don't fit in the existing tree, or in an order - * where the tree layout cannot be guessed immediately. - * - * This structure encodes an idealized mipmap tree. The GL image - * commands build these where possible, otherwise store the images in - * temporary system buffers. - */ - - -/** - * Describes the location of each texture image within a texture region. - */ -struct intel_mipmap_level -{ - GLuint level_offset; - GLuint width; - GLuint height; - GLuint depth; - GLuint nr_images; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - GLuint *image_offset; -}; - -struct intel_mipmap_tree -{ - /* Effectively the key: - */ - GLenum target; - GLenum internal_format; - - GLuint first_level; - GLuint last_level; - - GLuint width0, height0, depth0; /**< Level zero image dimensions */ - GLuint cpp; - GLboolean compressed; - - /* Derived from the above: - */ - GLuint pitch; - GLuint depth_pitch; /* per-image on i945? */ - GLuint total_height; - - /* Includes image offset tables: - */ - struct intel_mipmap_level level[MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct intel_region *region; - - /* These are also refcounted: - */ - GLuint refcount; -}; - - - -struct intel_mipmap_tree *intel_miptree_create(struct intel_context *intel, - GLenum target, - GLenum internal_format, - GLuint first_level, - GLuint last_level, - GLuint width0, - GLuint height0, - GLuint depth0, - GLuint cpp, - GLuint compress_byte); - -void intel_miptree_reference(struct intel_mipmap_tree **dst, - struct intel_mipmap_tree *src); - -void intel_miptree_release(struct intel_context *intel, - struct intel_mipmap_tree **mt); - -/* Check if an image fits an existing mipmap tree layout - */ -GLboolean intel_miptree_match_image(struct intel_mipmap_tree *mt, - struct gl_texture_image *image, - GLuint face, GLuint level); - -/* Return a pointer to an image within a tree. Return image stride as - * well. - */ -GLubyte *intel_miptree_image_map(struct intel_context *intel, - struct intel_mipmap_tree *mt, - GLuint face, - GLuint level, - GLuint * row_stride, GLuint image_offsets[]); - -void intel_miptree_image_unmap(struct intel_context *intel, - struct intel_mipmap_tree *mt); - - -/* Return the linear offset of an image relative to the start of the - * tree: - */ -GLuint intel_miptree_image_offset(struct intel_mipmap_tree *mt, - GLuint face, GLuint level); - -/* Return pointers to each 2d slice within an image. Indexed by depth - * value. - */ -const GLuint *intel_miptree_depth_offsets(struct intel_mipmap_tree *mt, - GLuint level); - - -void intel_miptree_set_level_info(struct intel_mipmap_tree *mt, - GLuint level, - GLuint nr_images, - GLuint x, GLuint y, - GLuint w, GLuint h, GLuint d); - -void intel_miptree_set_image_offset(struct intel_mipmap_tree *mt, - GLuint level, - GLuint img, GLuint x, GLuint y); - - -/* Upload an image into a tree - */ -void intel_miptree_image_data(struct intel_context *intel, - struct intel_mipmap_tree *dst, - GLuint face, - GLuint level, - void *src, - GLuint src_row_pitch, GLuint src_image_pitch); - -/* Copy an image between two trees - */ -void intel_miptree_image_copy(struct intel_context *intel, - struct intel_mipmap_tree *dst, - GLuint face, GLuint level, - struct intel_mipmap_tree *src); - -/* i915_mipmap_tree.c: - */ -GLboolean i915_miptree_layout(struct intel_mipmap_tree *mt); -GLboolean i945_miptree_layout(struct intel_mipmap_tree *mt); - - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel.c b/src/mesa/drivers/dri/i915tex/intel_pixel.c deleted file mode 100644 index 8e61cc8184..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_pixel.c +++ /dev/null @@ -1,119 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portionsalloc - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "enums.h" -#include "state.h" -#include "swrast/swrast.h" - -#include "intel_context.h" -#include "intel_pixel.h" -#include "intel_regions.h" - - -/** - * Check if any fragment operations are in effect which might effect - * glDraw/CopyPixels. - */ -GLboolean -intel_check_blit_fragment_ops(GLcontext * ctx) -{ - if (ctx->NewState) - _mesa_update_state(ctx); - - /* XXX Note: Scissor done with blitter: - */ - return !(ctx->_ImageTransferState || - ctx->Color.AlphaEnabled || - ctx->Depth.Test || - ctx->Fog.Enabled || - ctx->Stencil.Enabled || - !ctx->Color.ColorMask[0] || - !ctx->Color.ColorMask[1] || - !ctx->Color.ColorMask[2] || - !ctx->Color.ColorMask[3] || - ctx->Texture._EnabledUnits || - ctx->FragmentProgram._Enabled || - ctx->Color.BlendEnabled); -} - - -GLboolean -intel_check_meta_tex_fragment_ops(GLcontext * ctx) -{ - if (ctx->NewState) - _mesa_update_state(ctx); - - /* Some of _ImageTransferState (scale, bias) could be done with - * fragment programs on i915. - */ - return !(ctx->_ImageTransferState || ctx->Fog.Enabled || /* not done yet */ - ctx->Texture._EnabledUnits || ctx->FragmentProgram._Enabled); -} - -/* The intel_region struct doesn't really do enough to capture the - * format of the pixels in the region. For now this code assumes that - * the region is a display surface and hence is either ARGB8888 or - * RGB565. - * XXX FBO: If we'd pass in the intel_renderbuffer instead of region, we'd - * know the buffer's pixel format. - * - * \param format as given to glDraw/ReadPixels - * \param type as given to glDraw/ReadPixels - */ -GLboolean -intel_check_blit_format(struct intel_region * region, - GLenum format, GLenum type) -{ - if (region->cpp == 4 && - (type == GL_UNSIGNED_INT_8_8_8_8_REV || - type == GL_UNSIGNED_BYTE) && format == GL_BGRA) { - return GL_TRUE; - } - - if (region->cpp == 2 && - type == GL_UNSIGNED_SHORT_5_6_5_REV && format == GL_BGR) { - return GL_TRUE; - } - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s: bad format for blit (cpp %d, type %s format %s)\n", - __FUNCTION__, region->cpp, - _mesa_lookup_enum_by_nr(type), _mesa_lookup_enum_by_nr(format)); - - return GL_FALSE; -} - - -void -intelInitPixelFuncs(struct dd_function_table *functions) -{ - functions->Accum = _swrast_Accum; - functions->Bitmap = _swrast_Bitmap; - functions->CopyPixels = intelCopyPixels; - functions->ReadPixels = intelReadPixels; - functions->DrawPixels = intelDrawPixels; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel.h b/src/mesa/drivers/dri/i915tex/intel_pixel.h deleted file mode 100644 index a6fcf90ce0..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_pixel.h +++ /dev/null @@ -1,63 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_PIXEL_H -#define INTEL_PIXEL_H - -#include "mtypes.h" - -void intelInitPixelFuncs(struct dd_function_table *functions); - -GLboolean intel_check_blit_fragment_ops(GLcontext * ctx); - -GLboolean intel_check_meta_tex_fragment_ops(GLcontext * ctx); - -GLboolean intel_check_blit_format(struct intel_region *region, - GLenum format, GLenum type); - - -void intelReadPixels(GLcontext * ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - GLvoid * pixels); - -void intelDrawPixels(GLcontext * ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - GLenum format, - GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid * pixels); - -void intelCopyPixels(GLcontext * ctx, - GLint srcx, GLint srcy, - GLsizei width, GLsizei height, - GLint destx, GLint desty, GLenum type); - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c b/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c deleted file mode 100644 index c205c6a715..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_bitmap.c +++ /dev/null @@ -1,351 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portionsalloc - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "enums.h" -#include "image.h" -#include "colormac.h" -#include "mtypes.h" -#include "macros.h" -#include "bufferobj.h" -#include "swrast/swrast.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "intel_regions.h" -#include "intel_buffer_objects.h" - - - -#define FILE_DEBUG_FLAG DEBUG_PIXEL - - -/* Unlike the other intel_pixel_* functions, the expectation here is - * that the incoming data is not in a PBO. With the XY_TEXT blit - * method, there's no benefit haveing it in a PBO, but we could - * implement a path based on XY_MONO_SRC_COPY_BLIT which might benefit - * PBO bitmaps. I think they are probably pretty rare though - I - * wonder if Xgl uses them? - */ -static const GLubyte *map_pbo( GLcontext *ctx, - GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, - const GLubyte *bitmap ) -{ - GLubyte *buf; - - if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, - GL_COLOR_INDEX, GL_BITMAP, - (GLvoid *) bitmap)) { - _mesa_error(ctx, GL_INVALID_OPERATION,"glBitmap(invalid PBO access)"); - return NULL; - } - - buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, - GL_READ_ONLY_ARB, - unpack->BufferObj); - if (!buf) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glBitmap(PBO is mapped)"); - return NULL; - } - - return ADD_POINTERS(buf, bitmap); -} - -static GLboolean test_bit( const GLubyte *src, - GLuint bit ) -{ - return (src[bit/8] & (1<<(bit % 8))) ? 1 : 0; -} - -static void set_bit( GLubyte *dest, - GLuint bit ) -{ - dest[bit/8] |= 1 << (bit % 8); -} - -static int align(int x, int align) -{ - return (x + align - 1) & ~(align - 1); -} - -/* Extract a rectangle's worth of data from the bitmap. Called - * per-cliprect. - */ -static GLuint get_bitmap_rect(GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, - const GLubyte *bitmap, - GLuint x, GLuint y, - GLuint w, GLuint h, - GLubyte *dest, - GLuint row_align, - GLboolean invert) -{ - GLuint src_offset = (x + unpack->SkipPixels) & 0x7; - GLuint mask = unpack->LsbFirst ? 0 : 7; - GLuint bit = 0; - GLint row, col; - GLint first, last; - GLint incr; - GLuint count = 0; - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s %d,%d %dx%d bitmap %dx%d skip %d src_offset %d mask %d\n", - __FUNCTION__, x,y,w,h,width,height,unpack->SkipPixels, src_offset, mask); - - if (invert) { - first = h-1; - last = 0; - incr = -1; - } - else { - first = 0; - last = h-1; - incr = 1; - } - - /* Require that dest be pre-zero'd. - */ - for (row = first; row != (last+incr); row += incr) { - const GLubyte *rowsrc = _mesa_image_address2d(unpack, bitmap, - width, height, - GL_COLOR_INDEX, GL_BITMAP, - y + row, x); - - for (col = 0; col < w; col++, bit++) { - if (test_bit(rowsrc, (col + src_offset) ^ mask)) { - set_bit(dest, bit ^ 7); - count++; - } - } - - if (row_align) - bit = (bit + row_align - 1) & ~(row_align - 1); - } - - return count; -} - - - - -/* - * Render a bitmap. - */ -static GLboolean -do_blit_bitmap( GLcontext *ctx, - GLint dstx, GLint dsty, - GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, - const GLubyte *bitmap ) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_region *dst = intel_drawbuf_region(intel); - - union { - GLuint ui; - GLubyte ub[4]; - } color; - - - if (unpack->BufferObj->Name) { - bitmap = map_pbo(ctx, width, height, unpack, bitmap); - if (bitmap == NULL) - return GL_TRUE; /* even though this is an error, we're done */ - } - - UNCLAMPED_FLOAT_TO_CHAN(color.ub[0], ctx->Current.RasterColor[2]); - UNCLAMPED_FLOAT_TO_CHAN(color.ub[1], ctx->Current.RasterColor[1]); - UNCLAMPED_FLOAT_TO_CHAN(color.ub[2], ctx->Current.RasterColor[0]); - UNCLAMPED_FLOAT_TO_CHAN(color.ub[3], ctx->Current.RasterColor[3]); - - /* Does zoom apply to bitmaps? - */ - if (!intel_check_blit_fragment_ops(ctx) || - ctx->Pixel.ZoomX != 1.0F || - ctx->Pixel.ZoomY != 1.0F) - return GL_FALSE; - - LOCK_HARDWARE(intel); - - { - drm_clip_rect_t box; - drm_clip_rect_t dest_rect; - GLint srcx = 0, srcy = 0; - GLint orig_screen_x1, orig_screen_y2; - GLuint i; - - box.x1 = 0; - box.y1 = 0; - box.x2 = ctx->DrawBuffer->Width; - box.y2 = ctx->DrawBuffer->Height; - orig_screen_x1 = dstx; - orig_screen_y2 = box->y2 - dsty; - - /* Do scissoring in GL coordinates: - */ -x if (ctx->Scissor.Enabled) - { - GLint x = ctx->Scissor.X; - GLint y = ctx->Scissor.Y; - GLuint w = ctx->Scissor.Width; - GLuint h = ctx->Scissor.Height; - - if (!_mesa_clip_to_region(x, y, x+w-1, y+h-1, &dstx, &dsty, &width, &height)) - goto out; - } - - /* Convert from GL to hardware coordinates: - */ - dsty = box->y2 - dsty - height; - dstx = dstx; - - dest_rect.x1 = dstx; - dest_rect.y1 = dsty; - dest_rect.x2 = dstx + width; - dest_rect.y2 = dsty + height; - - for (i = 0; i < 1; i++) { - drm_clip_rect_t rect; - int box_w, box_h; - GLint px, py; - GLuint stipple[32]; - - if (!intel_intersect_cliprects(&rect, &dest_rect, &box[0])) - continue; - - /* Now go back to GL coordinates to figure out what subset of - * the bitmap we are uploading for this cliprect: - */ - box_w = rect.x2 - rect.x1; - box_h = rect.y2 - rect.y1; - srcx = rect.x1 - orig_screen_x1; - srcy = orig_screen_y2 - rect.y2; - - -#define DY 32 -#define DX 32 - - /* Then, finally, chop it all into chunks that can be - * digested by hardware: - */ - for (py = 0; py < box_h; py += DY) { - for (px = 0; px < box_w; px += DX) { - int h = MIN2(DY, box_h - py); - int w = MIN2(DX, box_w - px); - GLuint sz = align(align(w,8) * h, 64)/8; - - assert(sz <= sizeof(stipple)); - memset(stipple, 0, sz); - - /* May need to adjust this when padding has been introduced in - * sz above: - */ - if (get_bitmap_rect(width, height, unpack, - bitmap, - srcx + px, srcy + py, w, h, - (GLubyte *)stipple, - 8, - GL_TRUE) == 0) - continue; - - /* - */ - intelEmitImmediateColorExpandBlit( intel, - dst->cpp, - (GLubyte *)stipple, - sz, - color.ui, - dst->pitch, - dst->buffer, - 0, - dst->tiled, - rect.x1 + px, - rect.y2 - (py + h), - w, h); - } - } - } - intel->need_flush = GL_TRUE; - out: - intel_batchbuffer_flush(intel->batch); - } - UNLOCK_HARDWARE(intel); - - - if (unpack->BufferObj->Name) { - /* done with PBO so unmap it now */ - ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, - unpack->BufferObj); - } - - return GL_TRUE; -} - - - - - -/* There are a large number of possible ways to implement bitmap on - * this hardware, most of them have some sort of drawback. Here are a - * few that spring to mind: - * - * Blit: - * - XY_MONO_SRC_BLT_CMD - * - use XY_SETUP_CLIP_BLT for cliprect clipping. - * - XY_TEXT_BLT - * - XY_TEXT_IMMEDIATE_BLT - * - blit per cliprect, subject to maximum immediate data size. - * - XY_COLOR_BLT - * - per pixel or run of pixels - * - XY_PIXEL_BLT - * - good for sparse bitmaps - * - * 3D engine: - * - Point per pixel - * - Translate bitmap to an alpha texture and render as a quad - * - Chop bitmap up into 32x32 squares and render w/polygon stipple. - */ -void -intelBitmap(GLcontext * ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, - const GLubyte * pixels) -{ - if (do_blit_bitmap(ctx, x, y, width, height, - unpack, pixels)) - return; - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s: fallback to swrast\n", __FUNCTION__); - - _swrast_Bitmap(ctx, x, y, width, height, unpack, pixels); -} diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c b/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c deleted file mode 100644 index 2ecacc3c09..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_copy.c +++ /dev/null @@ -1,358 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "enums.h" -#include "image.h" -#include "state.h" -#include "mtypes.h" -#include "macros.h" -#include "swrast/swrast.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "intel_buffers.h" -#include "intel_blit.h" -#include "intel_regions.h" -#include "intel_tris.h" -#include "intel_pixel.h" -#include "intel_fbo.h" - -#define FILE_DEBUG_FLAG DEBUG_PIXEL - -static struct intel_region * -copypix_src_region(struct intel_context *intel, GLenum type) -{ - struct intel_renderbuffer *irb; - - switch (type) { - case GL_COLOR: - return intel_readbuf_region(intel); - case GL_DEPTH: - /* Don't think this is really possible execpt at 16bpp, when we have no stencil. - */ - irb = intel_renderbuffer(intel->ctx.ReadBuffer->_DepthBuffer->Wrapped); - if (irb && irb->region && irb->region->cpp == 2) - return irb->region; - break; - case GL_STENCIL: - /* Don't think this is really possible. - */ - break; - case GL_DEPTH_STENCIL_EXT: - /* Does it matter whether it is stencil/depth or depth/stencil? - */ - irb = intel_renderbuffer(intel->ctx.ReadBuffer->_DepthBuffer->Wrapped); - if (irb && irb->region) - return irb->region; - break; - default: - break; - } - - return NULL; -} - - -/* Doesn't work for overlapping regions. Could do a double copy or - * just fallback. - */ -static GLboolean -do_texture_copypixels(GLcontext * ctx, - GLint srcx, GLint srcy, - GLsizei width, GLsizei height, - GLint dstx, GLint dsty, GLenum type) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_region *dst = intel_drawbuf_region(intel); - struct intel_region *src = copypix_src_region(intel, type); - struct intel_region *depthreg = NULL; - GLenum src_format; - GLenum src_type; - - DBG("%s %d,%d %dx%d --> %d,%d\n", __FUNCTION__, - srcx, srcy, width, height, dstx, dsty); - - if (!src || !dst || type != GL_COLOR) - return GL_FALSE; - - /* Can't handle overlapping regions. Don't have sufficient control - * over rasterization to pull it off in-place. Punt on these for - * now. - * - * XXX: do a copy to a temporary. - */ - if (src->buffer == dst->buffer) { - drm_clip_rect_t srcbox; - drm_clip_rect_t dstbox; - drm_clip_rect_t tmp; - - srcbox.x1 = srcx; - srcbox.y1 = srcy; - srcbox.x2 = srcx + width; - srcbox.y2 = srcy + height; - - dstbox.x1 = dstx; - dstbox.y1 = dsty; - dstbox.x2 = dstx + width * ctx->Pixel.ZoomX; - dstbox.y2 = dsty + height * ctx->Pixel.ZoomY; - - DBG("src %d,%d %d,%d\n", srcbox.x1, srcbox.y1, srcbox.x2, srcbox.y2); - DBG("dst %d,%d %d,%d (%dx%d) (%f,%f)\n", dstbox.x1, dstbox.y1, dstbox.x2, dstbox.y2, - width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY); - - if (intel_intersect_cliprects(&tmp, &srcbox, &dstbox)) { - DBG("%s: regions overlap\n", __FUNCTION__); - return GL_FALSE; - } - } - - intelFlush(&intel->ctx); - - intel->vtbl.install_meta_state(intel); - - /* Is this true? Also will need to turn depth testing on according - * to state: - */ - intel->vtbl.meta_no_stencil_write(intel); - intel->vtbl.meta_no_depth_write(intel); - - /* Set the 3d engine to draw into the destination region: - */ - if (ctx->DrawBuffer->_DepthBuffer && - ctx->DrawBuffer->_DepthBuffer->Wrapped) - depthreg = (intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped))->region; - - intel->vtbl.meta_draw_region(intel, dst, depthreg); - - intel->vtbl.meta_import_pixel_state(intel); - - if (src->cpp == 2) { - src_format = GL_RGB; - src_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - src_format = GL_BGRA; - src_type = GL_UNSIGNED_BYTE; - } - - /* Set the frontbuffer up as a large rectangular texture. - */ - if (!intel->vtbl.meta_tex_rect_source(intel, src->buffer, 0, - src->pitch, - src->height, src_format, src_type)) { - intel->vtbl.leave_meta_state(intel); - return GL_FALSE; - } - - - intel->vtbl.meta_texture_blend_replace(intel); - - LOCK_HARDWARE(intel); - - { - int dstbufHeight = ctx->DrawBuffer->Height; - /* convert from gl to hardware coords */ - srcy = ctx->ReadBuffer->Height - srcy - height; - - /* Clip against the source region. This is the only source - * clipping we do. XXX: Just set the texcord wrap mode to clamp - * or similar. - * - */ - if (0) { - GLint orig_x = srcx; - GLint orig_y = srcy; - - if (!_mesa_clip_to_region(0, 0, ctx->ReadBuffer->Width - 1, - ctx->ReadBuffer->Height - 1, - &srcx, &srcy, &width, &height)) - goto out; - - dstx += srcx - orig_x; - dsty += (srcy - orig_y) * ctx->Pixel.ZoomY; - } - - /* Just use the regular cliprect mechanism... Does this need to - * even hold the lock??? - */ - intel_meta_draw_quad(intel, - dstx, - dstx + width * ctx->Pixel.ZoomX, - dstbufHeight - (dsty + height * ctx->Pixel.ZoomY), - dstbufHeight - (dsty), 0, /* XXX: what z value? */ - 0x00ff00ff, - srcx, srcx + width, srcy, srcy + height); - - out: - intel->vtbl.leave_meta_state(intel); - intel_batchbuffer_flush(intel->batch); - } - UNLOCK_HARDWARE(intel); - - DBG("%s: success\n", __FUNCTION__); - return GL_TRUE; -} - - - - - -/** - * CopyPixels with the blitter. Don't support zooming, pixel transfer, etc. - */ -static GLboolean -do_blit_copypixels(GLcontext * ctx, - GLint srcx, GLint srcy, - GLsizei width, GLsizei height, - GLint dstx, GLint dsty, GLenum type) -{ - struct intel_context *intel = intel_context(ctx); - struct gl_framebuffer *fb = ctx->DrawBuffer; - struct gl_framebuffer *fbread = ctx->ReadBuffer; - struct intel_renderbuffer *irbread; - struct intel_renderbuffer *irbdraw; - struct intel_region *dst; - struct intel_region *src; - - /* Copypixels can be more than a straight copy. Ensure all the - * extra operations are disabled: - */ - if (!intel_check_blit_fragment_ops(ctx) || - ctx->Pixel.ZoomX != 1.0F || ctx->Pixel.ZoomY != 1.0F) - return GL_FALSE; - - intelFlush(&intel->ctx); - - if (type == GL_COLOR) { - irbread = intel_renderbuffer(fbread->_ColorReadBuffer); - irbdraw = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); - if (!irbread || !irbread->region || !irbdraw || !irbdraw->region) - return GL_FALSE; - } - else if (type == GL_DEPTH) { - /* Don't think this is really possible execpt at 16bpp, when we have no stencil. - */ - irbread = intel_renderbuffer(fbread->_DepthBuffer->Wrapped); - irbdraw = intel_renderbuffer(fb->_DepthBuffer->Wrapped); - if (!irbread || !irbread->region || !irbdraw || !irbdraw->region - || !(irbread->region->cpp == 2)) - return GL_FALSE; - } - else if (type == GL_DEPTH_STENCIL_EXT) { - /* Does it matter whether it is stencil/depth or depth/stencil? - */ - irbread = intel_renderbuffer(fbread->_DepthBuffer->Wrapped); - irbdraw = intel_renderbuffer(fb->_DepthBuffer->Wrapped); - if (!irbread || !irbread->region || !irbdraw || !irbdraw->region) - return GL_FALSE; - } - else /* GL_STENCIL */ { - /* Don't think this is really possible. - */ - return GL_FALSE; - } - - src = irbread->region; - dst = irbdraw->region; - - { - GLint dx = dstx - srcx; - GLint dy = dsty - srcy; - - /* Clip against dest, including scissor, in GL coordinates: - */ - - if (!_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, - fb->_Xmax - 1, fb->_Ymax - 1, - &dstx, &dsty, &width, &height)) - goto out; - - srcx = dstx - dx; - srcy = dsty - dy; - - /* Convert from GL to hardware coordinates: - */ - dsty = fb->Height - dsty - height; - srcy = fbread->Height - srcy - height; - - /* Clip against the source region: - */ - dx = srcx - dstx; - dy = srcy - dsty; - - if (!_mesa_clip_to_region(0, 0, irbread->Base.Width - 1, - irbread->Base.Height - 1, - &srcx, &srcy, &width, &height)) - goto out; - - dstx = srcx - dx; - dsty = srcy - dy; - - - { - - intelEmitCopyBlit(intel, dst->cpp, - src->pitch, src->buffer, 0, - dst->pitch, dst->buffer, 0, - srcx, srcy, - dstx, dsty, - width, height, - ctx->Color.ColorLogicOpEnabled ? - ctx->Color.LogicOp : GL_COPY); - } - - out: - intel_batchbuffer_flush(intel->batch); - } - - DBG("%s: success\n", __FUNCTION__); - return GL_TRUE; -} - - -void -intelCopyPixels(GLcontext * ctx, - GLint srcx, GLint srcy, - GLsizei width, GLsizei height, - GLint destx, GLint desty, GLenum type) -{ - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (do_blit_copypixels(ctx, srcx, srcy, width, height, destx, desty, type)) - return; - - if (do_texture_copypixels(ctx, srcx, srcy, width, height, destx, desty, type)) - return; - - DBG("fallback to _swrast_CopyPixels\n"); - - _swrast_CopyPixels(ctx, srcx, srcy, width, height, destx, desty, type); -} diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c b/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c deleted file mode 100644 index 4139117206..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_draw.c +++ /dev/null @@ -1,431 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portionsalloc - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "enums.h" -#include "image.h" -#include "mtypes.h" -#include "macros.h" -#include "bufferobj.h" -#include "swrast/swrast.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "intel_buffers.h" -#include "intel_regions.h" -#include "intel_pixel.h" -#include "intel_buffer_objects.h" -#include "intel_tris.h" -#include "intel_fbo.h" - - -static GLboolean -do_texture_drawpixels(GLcontext * ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid * pixels) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_region *dst = intel_drawbuf_region(intel); - struct intel_buffer_object *src = intel_buffer_object(unpack->BufferObj); - struct intel_region *depthreg = NULL; - GLuint rowLength = unpack->RowLength ? unpack->RowLength : width; - GLuint src_offset; - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - intelFlush(&intel->ctx); - intel->vtbl.render_start(intel); - intel->vtbl.emit_state(intel); - - if (!dst) - return GL_FALSE; - - if (src) { - if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels"); - return GL_TRUE; - } - } - else { - /* PBO only for now: - */ -/* _mesa_printf("%s - not PBO\n", __FUNCTION__); */ - return GL_FALSE; - } - - /* There are a couple of things we can't do yet, one of which is - * set the correct state for pixel operations when GL texturing is - * enabled. That's a pretty rare state and probably not worth the - * effort. A completely device-independent version of this may do - * more. - * - * Similarly, we make no attempt to merge metaops processing with - * an enabled fragment program, though it would certainly be - * possible. - */ - if (!intel_check_meta_tex_fragment_ops(ctx)) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - bad GL fragment state for metaops texture\n", - __FUNCTION__); - return GL_FALSE; - } - - intel->vtbl.install_meta_state(intel); - - - /* Is this true? Also will need to turn depth testing on according - * to state: - */ - intel->vtbl.meta_no_stencil_write(intel); - intel->vtbl.meta_no_depth_write(intel); - - /* Set the 3d engine to draw into the destination region: - */ - if (ctx->DrawBuffer->_DepthBuffer && - ctx->DrawBuffer->_DepthBuffer->Wrapped) - depthreg = (intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped))->region; - - intel->vtbl.meta_draw_region(intel, dst, depthreg); - - intel->vtbl.meta_import_pixel_state(intel); - - src_offset = (GLuint) _mesa_image_address(2, unpack, pixels, width, height, - format, type, 0, 0, 0); - - - /* Setup the pbo up as a rectangular texture, if possible. - * - * TODO: This is almost always possible if the i915 fragment - * program is adjusted to correctly swizzle the sampled colors. - * The major exception is any 24bit texture, like RGB888, for which - * there is no hardware support. - */ - if (!intel->vtbl.meta_tex_rect_source(intel, src->buffer, src_offset, - rowLength, height, format, type)) { - intel->vtbl.leave_meta_state(intel); - return GL_FALSE; - } - - intel->vtbl.meta_texture_blend_replace(intel); - - - LOCK_HARDWARE(intel); - - { - int bufHeight = ctx->DrawBuffer->Height; - - GLint srcx, srcy; - GLint dstx, dsty; - - dstx = x; - dsty = bufHeight - (y + height); - - srcx = 0; /* skiprows/pixels already done */ - srcy = 0; - - if (0) { - const GLint orig_x = dstx; - const GLint orig_y = dsty; - - if (!_mesa_clip_to_region(0, 0, dst->pitch, dst->height, - &dstx, &dsty, &width, &height)) - goto out; - - srcx += dstx - orig_x; - srcy += dsty - orig_y; - } - - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("draw %d,%d %dx%d\n", dstx, dsty, width, height); - - /* Must use the regular cliprect mechanism in order to get the - * drawing origin set correctly. Otherwise scissor state is in - * incorrect coordinate space. Does this even need to hold the - * lock??? - */ - intel_meta_draw_quad(intel, - dstx, dstx + width * ctx->Pixel.ZoomX, - bufHeight - (y + height * ctx->Pixel.ZoomY), - bufHeight - (y), - -ctx->Current.RasterPos[2] * .5, - 0x00ff00ff, - srcx, srcx + width, srcy + height, srcy); - out: - intel->vtbl.leave_meta_state(intel); - intel_batchbuffer_flush(intel->batch); - } - UNLOCK_HARDWARE(intel); - return GL_TRUE; -} - - - - - -/* Pros: - * - no waiting for idle before updating framebuffer. - * - * Cons: - * - if upload is by memcpy, this may actually be slower than fallback path. - * - uploads the whole image even if destination is clipped - * - * Need to benchmark. - * - * Given the questions about performance, implement for pbo's only. - * This path is definitely a win if the pbo is already in agp. If it - * turns out otherwise, we can add the code necessary to upload client - * data to agp space before performing the blit. (Though it may turn - * out to be better/simpler just to use the texture engine). - */ -static GLboolean -do_blit_drawpixels(GLcontext * ctx, - GLint dstx, GLint dsty, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid * pixels) -{ - struct intel_context *intel = intel_context(ctx); - struct gl_framebuffer *fb = ctx->DrawBuffer; - struct intel_renderbuffer *irbdraw; - struct intel_region *dest; - struct intel_buffer_object *src = intel_buffer_object(unpack->BufferObj); - GLuint src_offset; - GLuint rowLength; - struct _DriFenceObject *fence = NULL; - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s\n", __FUNCTION__); - - if (type == GL_COLOR) { - irbdraw = intel_renderbuffer(fb->_ColorDrawBuffers[0][0]); - if (!irbdraw || !irbdraw->region) - return GL_FALSE; - } - else if (type == GL_DEPTH) { - /* Don't think this is really possible execpt at 16bpp, when we have no stencil. - */ - irbdraw = intel_renderbuffer(fb->_DepthBuffer->Wrapped); - if (!irbdraw || !irbdraw->region || !(irbdraw->region->cpp == 2)) - return GL_FALSE; - } - else if (type == GL_DEPTH_STENCIL_EXT) { - /* Does it matter whether it is stencil/depth or depth/stencil? - */ - irbdraw = intel_renderbuffer(fb->_DepthBuffer->Wrapped); - if (!irbdraw || !irbdraw->region) - return GL_FALSE; - } - else /* GL_STENCIL */ { - /* Don't think this is really possible. - */ - return GL_FALSE; - } - - dest = irbdraw->region; - - if (!dest) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - no dest\n", __FUNCTION__); - return GL_FALSE; - } - - if (src) { - /* This validation should be done by core mesa: - */ - if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels"); - return GL_TRUE; - } - } - else { - /* PBO only for now: - */ - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - not PBO\n", __FUNCTION__); - return GL_FALSE; - } - - if (!intel_check_blit_format(dest, format, type)) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - bad format for blit\n", __FUNCTION__); - return GL_FALSE; - } - - if (!intel_check_blit_fragment_ops(ctx)) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - bad GL fragment state for blitter\n", - __FUNCTION__); - return GL_FALSE; - } - - if (ctx->Pixel.ZoomX != 1.0F) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - bad PixelZoomX for blit\n", __FUNCTION__); - return GL_FALSE; - } - - - if (unpack->RowLength > 0) - rowLength = unpack->RowLength; - else - rowLength = width; - - if (ctx->Pixel.ZoomY == -1.0F) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - bad PixelZoomY for blit\n", __FUNCTION__); - return GL_FALSE; /* later */ - dsty -= height; - } - else if (ctx->Pixel.ZoomY == 1.0F) { - rowLength = -rowLength; - } - else { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - bad PixelZoomY for blit\n", __FUNCTION__); - return GL_FALSE; - } - - src_offset = (GLuint) _mesa_image_address(2, unpack, pixels, width, height, - format, type, 0, 0, 0); - - /* don't need a lock as we have no cliprects ? */ - intelFlush(&intel->ctx); - - { - GLuint srcx = 0; - GLuint srcy = 0; - GLint dx = dstx; - GLint dy = dsty; - GLuint height_orig = height; - - /* Do scissoring and clipping in GL coordinates, no need to clip against - * pbo src region (note fbo fields include scissor already): - */ - height_orig = height; - if (!_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, - fb->_Xmax - 1, fb->_Ymax - 1, - &dstx, &dsty, &width, &height)) - goto out; - - srcx = dstx - dx; - srcy = dsty - dy; - - struct _DriBufferObject *src_buffer = - intel_bufferobj_buffer(intel, src, INTEL_READ); - - /* Convert from GL to hardware coordinates: - */ - dsty = fb->Height - dsty - height; - srcy = height_orig - srcy - height; - - { - intelEmitCopyBlit(intel, - dest->cpp, - rowLength, - src_buffer, src_offset, - dest->pitch, - dest->buffer, 0, - srcx, srcy, - dstx, dsty, width, height, - ctx->Color.ColorLogicOpEnabled ? - ctx->Color.LogicOp : GL_COPY); - } - fence = intel_batchbuffer_flush(intel->batch); - driFenceReference(fence); - } - - if (fence) { - driFenceFinish(fence, DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, GL_FALSE); - driFenceUnReference(fence); - } - - out: - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - DONE\n", __FUNCTION__); - - return GL_TRUE; -} - - - -void -intelDrawPixels(GLcontext * ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - GLenum format, - GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid * pixels) -{ - if (do_blit_drawpixels(ctx, x, y, width, height, format, type, - unpack, pixels)) - return; - - if (do_texture_drawpixels(ctx, x, y, width, height, format, type, - unpack, pixels)) - return; - - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s: fallback to swrast\n", __FUNCTION__); - - if (ctx->FragmentProgram._Current == ctx->FragmentProgram._TexEnvProgram) { - /* - * We don't want the i915 texenv program to be applied to DrawPixels. - * This is really just a performance optimization (mesa will other- - * wise happily run the fragment program on each pixel in the image). - */ - struct gl_fragment_program *fpSave = ctx->FragmentProgram._Current; - /* can't just set current frag prog to 0 here as on buffer resize - we'll get new state checks which will segfault (actually don't get them - with current priv buffers). Remains a hack. */ - ctx->FragmentProgram._Current = NULL; - ctx->FragmentProgram._UseTexEnvProgram = GL_FALSE; - ctx->FragmentProgram._Active = GL_FALSE; - _swrast_DrawPixels( ctx, x, y, width, height, format, type, - unpack, pixels ); - ctx->FragmentProgram._Current = fpSave; - ctx->FragmentProgram._UseTexEnvProgram = GL_TRUE; - ctx->FragmentProgram._Active = GL_TRUE; - } - else { - _swrast_DrawPixels( ctx, x, y, width, height, format, type, - unpack, pixels ); - } -} diff --git a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c b/src/mesa/drivers/dri/i915tex/intel_pixel_read.c deleted file mode 100644 index 76a6959625..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_pixel_read.c +++ /dev/null @@ -1,327 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "enums.h" -#include "mtypes.h" -#include "macros.h" -#include "image.h" -#include "bufferobj.h" -#include "swrast/swrast.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "intel_buffers.h" -#include "intel_regions.h" -#include "intel_pixel.h" -#include "intel_buffer_objects.h" -#include "intel_fbo.h" -#include "intel_tris.h" - -/* For many applications, the new ability to pull the source buffers - * back out of the GTT and then do the packing/conversion operations - * in software will be as much of an improvement as trying to get the - * blitter and/or texture engine to do the work. - * - * This step is gated on private backbuffers. - * - * Obviously the frontbuffer can't be pulled back, so that is either - * an argument for blit/texture readpixels, or for blitting to a - * temporary and then pulling that back. - * - * XXX this is not true for fake frontbuffers... - - * When the destination is a pbo, however, it's not clear if it is - * ever going to be pulled to main memory (though the access param - * will be a good hint). So it sounds like we do want to be able to - * choose between blit/texture implementation on the gpu and pullback - * and cpu-based copying. - * - * Unless you can magically turn client memory into a PBO for the - * duration of this call, there will be a cpu-based copying step in - * any case. - */ - - -static GLboolean -do_texture_readpixels(GLcontext * ctx, - GLint srcx, GLint srcy, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - struct intel_region *dest_region) -{ -#if 0 - struct intel_context *intel = intel_context(ctx); - struct intel_region *depthreg = NULL; - struct intel_region *src = intel_readbuf_region(intel); -// struct intel_region *src = copypix_src_region(intel, type); - GLenum src_format; - GLenum src_type; - - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - if (!src || type != GL_COLOR) - return GL_FALSE; - - intelFlush(&intel->ctx); - - intel->vtbl.install_meta_state(intel); - - /* Is this true? Also will need to turn depth testing on according - * to state: - */ - intel->vtbl.meta_no_stencil_write(intel); - intel->vtbl.meta_no_depth_write(intel); - - /* Set the 3d engine to draw into the destination region: - */ - if (ctx->DrawBuffer->_DepthBuffer && - ctx->DrawBuffer->_DepthBuffer->Wrapped) - depthreg = (intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped))->region; - - intel->vtbl.meta_draw_region(intel, dest_region, depthreg); - - intel->vtbl.meta_import_pixel_state(intel); - - if (src->cpp == 2) { - src_format = GL_RGB; - src_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - src_format = GL_BGRA; - src_type = GL_UNSIGNED_BYTE; - } - - /* Set the frontbuffer up as a large rectangular texture. - */ - if (!intel->vtbl.meta_tex_rect_source(intel, src->buffer, 0, - src->pitch, - src->height, src_format, src_type)) { - intel->vtbl.leave_meta_state(intel); - return GL_FALSE; - } - - - intel->vtbl.meta_texture_blend_replace(intel); - - LOCK_HARDWARE(intel); - - { - int dstbufHeight = height * ctx->Pixel.ZoomY; /* ? */ - /* convert from gl to hardware coords */ - srcy = ctx->ReadBuffer->Height - srcy - height; - - /* Clip against the source region. This is the only source - * clipping we do. XXX: Just set the texcord wrap mode to clamp - * or similar. - * - */ - if (0) { - if (!_mesa_clip_to_region(0, 0, ctx->ReadBuffer->Width - 1, - ctx->ReadBuffer->Height - 1, - &srcx, &srcy, &width, &height)) - goto out; - - } - - /* Just use the regular cliprect mechanism... Does this need to - * even hold the lock??? - */ - intel_meta_draw_quad(intel, - 0, - width * ctx->Pixel.ZoomX, - dstbufHeight - (height * ctx->Pixel.ZoomY), - dstbufHeight, 0, /* XXX: what z value? */ - 0x00ff00ff, - srcx, srcx + width, srcy, srcy + height); - - out: - intel->vtbl.leave_meta_state(intel); - intel_batchbuffer_flush(intel->batch); - } - UNLOCK_HARDWARE(intel); - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s: success\n", __FUNCTION__); - return GL_TRUE; -#endif - - return GL_FALSE; -} - - - - -static GLboolean -do_blit_readpixels(GLcontext * ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, GLvoid * pixels) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_renderbuffer *irbread; - struct intel_region *src; - struct intel_buffer_object *dst = intel_buffer_object(pack->BufferObj); - GLuint dst_offset; - GLuint rowLength; - struct _DriFenceObject *fence = NULL; - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s\n", __FUNCTION__); - - irbread = intel_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); - if (!irbread || !irbread->region) - return GL_FALSE; - src = irbread->region; - - if (dst) { - /* XXX This validation should be done by core mesa: - */ - if (!_mesa_validate_pbo_access(2, pack, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels"); - return GL_TRUE; - } - } - else { - /* PBO only for now: - */ - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - not PBO\n", __FUNCTION__); - return GL_FALSE; - } - - - if (ctx->_ImageTransferState || - !intel_check_blit_format(src, format, type)) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - bad format for blit\n", __FUNCTION__); - return GL_FALSE; - } - - if (pack->RowLength > 0) - rowLength = pack->RowLength; - else - rowLength = width; - - if (((rowLength * src->cpp) % pack->Alignment) || - pack->SwapBytes || pack->LsbFirst) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s: bad packing params\n", __FUNCTION__); - return GL_FALSE; - } - - if (pack->Invert) { - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s: MESA_PACK_INVERT not done yet\n", __FUNCTION__); - return GL_FALSE; - } - else { - rowLength = -rowLength; - } - - /* XXX 64-bit cast? */ - dst_offset = (GLuint) _mesa_image_address(2, pack, pixels, width, height, - format, type, 0, 0, 0); - - intelFlush(&intel->ctx); - - { - GLint srcx = x; - GLint srcy = irbread->Base.Height - (y + height); - GLint srcwidth = width; - GLint srcheight = height; - - GLboolean all = (width * height * src->cpp == dst->Base.Size && - x == 0 && dst_offset == 0); - - struct _DriBufferObject *dst_buffer = - intel_bufferobj_buffer(intel, dst, all ? INTEL_WRITE_FULL : - INTEL_WRITE_PART); - - /* clip to src region */ - if (_mesa_clip_to_region(0, 0, irbread->Base.Width - 1, - irbread->Base.Height - 1, - &srcx, &srcy, &srcwidth, &srcheight)); - - { - intelEmitCopyBlit(intel, - src->cpp, - src->pitch, src->buffer, 0, - rowLength, - dst_buffer, dst_offset, - srcx, srcy, - 0, 0, - width, height, - GL_COPY); - } - - fence = intel_batchbuffer_flush(intel->batch); - driFenceReference(fence); - - } - - if (fence) { - driFenceFinish(fence, DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, - GL_FALSE); - driFenceUnReference(fence); - } - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s - DONE\n", __FUNCTION__); - - return GL_TRUE; -} - -void -intelReadPixels(GLcontext * ctx, - GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, GLvoid * pixels) -{ - if (INTEL_DEBUG & DEBUG_PIXEL) - fprintf(stderr, "%s\n", __FUNCTION__); - - intelFlush(ctx); - - if (do_blit_readpixels - (ctx, x, y, width, height, format, type, pack, pixels)) - return; - - if (do_texture_readpixels - (ctx, x, y, width, height, format, type, pack, pixels)) - return; - - if (INTEL_DEBUG & DEBUG_PIXEL) - _mesa_printf("%s: fallback to swrast\n", __FUNCTION__); - - _swrast_ReadPixels(ctx, x, y, width, height, format, type, pack, pixels); -} diff --git a/src/mesa/drivers/dri/i915tex/intel_reg.h b/src/mesa/drivers/dri/i915tex/intel_reg.h deleted file mode 100644 index 7828ba6ad3..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_reg.h +++ /dev/null @@ -1,88 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef _INTEL_REG_H_ -#define _INTEL_REG_H_ - - - -#define CMD_3D (0x3<<29) - - -#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) -#define PRIM_INDIRECT (1<<23) -#define PRIM_INLINE (0<<23) -#define PRIM_INDIRECT_SEQUENTIAL (0<<17) -#define PRIM_INDIRECT_ELTS (1<<17) - -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_MASK (0x1f<<18) - -#define I915PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define I915PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define I915PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define I915PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - - -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 - -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) - -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) - -#define MI_WAIT_FOR_EVENT ((0x3<<23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_regions.c b/src/mesa/drivers/dri/i915tex/intel_regions.c deleted file mode 100644 index 7d19bd07d3..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_regions.c +++ /dev/null @@ -1,480 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Provide additional functionality on top of bufmgr buffers: - * - 2d semantics and blit operations - * - refcounting of buffers for multiple images in a buffer. - * - refcounting of buffer mappings. - * - some logic for moving the buffers to the best memory pools for - * given operations. - * - * Most of this is to make it easier to implement the fixed-layout - * mipmap tree required by intel hardware in the face of GL's - * programming interface where each image can be specifed in random - * order and it isn't clear what layout the tree should have until the - * last moment. - */ - -#include "intel_context.h" -#include "intel_regions.h" -#include "intel_blit.h" -#include "intel_buffer_objects.h" -#include "dri_bufmgr.h" -#include "intel_batchbuffer.h" - -#define FILE_DEBUG_FLAG DEBUG_REGION - -void -intel_region_idle(intelScreenPrivate *intelScreen, struct intel_region *region) -{ - DBG("%s\n", __FUNCTION__); - if (region && region->buffer) - driBOWaitIdle(region->buffer, GL_FALSE); -} - -/* XXX: Thread safety? - */ -GLubyte * -intel_region_map(intelScreenPrivate *intelScreen, struct intel_region *region) -{ - DBG("%s\n", __FUNCTION__); - if (!region->map_refcount++) { - if (region->pbo) - intel_region_cow(intelScreen, region); - - region->map = driBOMap(region->buffer, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - } - - return region->map; -} - -void -intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *region) -{ - DBG("%s\n", __FUNCTION__); - if (!--region->map_refcount) { - driBOUnmap(region->buffer); - region->map = NULL; - } -} - -#undef TEST_CACHED_TEXTURES - -struct intel_region * -intel_region_alloc(intelScreenPrivate *intelScreen, - GLuint cpp, GLuint pitch, GLuint height) -{ - struct intel_region *region = calloc(sizeof(*region), 1); - struct intel_context *intel = intelScreenContext(intelScreen); - - DBG("%s\n", __FUNCTION__); - - region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ - region->refcount = 1; - - driGenBuffers(intelScreen->regionPool, - "region", 1, ®ion->buffer, 64, -#ifdef TEST_CACHED_TEXTURES - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_BIND_CACHED | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, -#else - 0, -#endif - 0); - LOCK_HARDWARE(intel); - driBOData(region->buffer, pitch * cpp * height, NULL, 0); - UNLOCK_HARDWARE(intel); - return region; -} - -void -intel_region_reference(struct intel_region **dst, struct intel_region *src) -{ - assert(*dst == NULL); - if (src) { - src->refcount++; - *dst = src; - } -} - -void -intel_region_release(struct intel_region **region) -{ - if (!*region) - return; - - DBG("%s %d\n", __FUNCTION__, (*region)->refcount - 1); - - ASSERT((*region)->refcount > 0); - (*region)->refcount--; - - if ((*region)->refcount == 0) { - assert((*region)->map_refcount == 0); - - if ((*region)->pbo) - (*region)->pbo->region = NULL; - (*region)->pbo = NULL; - driBOUnReference((*region)->buffer); - free(*region); - } - *region = NULL; -} - - -struct intel_region * -intel_region_create_static(intelScreenPrivate *intelScreen, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height) -{ - struct intel_region *region = calloc(sizeof(*region), 1); - DBG("%s\n", __FUNCTION__); - - region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ - region->refcount = 1; - - /* - * We use a "shared" buffer type to indicate buffers created and - * shared by others. - */ - - driGenBuffers(intelScreen->staticPool, "static region", 1, - ®ion->buffer, 64, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - driBOSetStatic(region->buffer, offset, pitch * cpp * height, virtual, 0); - - return region; -} - - - -void -intel_region_update_static(intelScreenPrivate *intelScreen, - struct intel_region *region, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height) -{ - DBG("%s\n", __FUNCTION__); - - region->cpp = cpp; - region->pitch = pitch; - region->height = height; /* needed? */ - - /* - * We use a "shared" buffer type to indicate buffers created and - * shared by others. - */ - - driDeleteBuffers(1, ®ion->buffer); - driGenBuffers(intelScreen->staticPool, "static region", 1, - ®ion->buffer, 64, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - driBOSetStatic(region->buffer, offset, pitch * cpp * height, virtual, 0); - -} - - - -/* - * XXX Move this into core Mesa? - */ -static void -_mesa_copy_rect(GLubyte * dst, - GLuint cpp, - GLuint dst_pitch, - GLuint dst_x, - GLuint dst_y, - GLuint width, - GLuint height, - const GLubyte * src, - GLuint src_pitch, GLuint src_x, GLuint src_y) -{ - GLuint i; - - dst_pitch *= cpp; - src_pitch *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * dst_pitch; - width *= cpp; - - if (width == dst_pitch && width == src_pitch) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; - } - } -} - - -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -void -intel_region_data(intelScreenPrivate *intelScreen, - struct intel_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - const void *src, GLuint src_pitch, - GLuint srcx, GLuint srcy, GLuint width, GLuint height) -{ - struct intel_context *intel = intelScreenContext(intelScreen); - - DBG("%s\n", __FUNCTION__); - - if (intel == NULL) - return; - - if (dst->pbo) { - if (dstx == 0 && - dsty == 0 && width == dst->pitch && height == dst->height) - intel_region_release_pbo(intelScreen, dst); - else - intel_region_cow(intelScreen, dst); - } - - - LOCK_HARDWARE(intel); - - _mesa_copy_rect(intel_region_map(intelScreen, dst) + dst_offset, - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - intel_region_unmap(intelScreen, dst); - - UNLOCK_HARDWARE(intel); - -} - -/* Copy rectangular sub-regions. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -void -intel_region_copy(intelScreenPrivate *intelScreen, - struct intel_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - struct intel_region *src, - GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height) -{ - struct intel_context *intel = intelScreenContext(intelScreen); - - DBG("%s\n", __FUNCTION__); - - if (intel == NULL) - return; - - if (dst->pbo) { - if (dstx == 0 && - dsty == 0 && width == dst->pitch && height == dst->height) - intel_region_release_pbo(intelScreen, dst); - else - intel_region_cow(intelScreen, dst); - } - - assert(src->cpp == dst->cpp); - - intelEmitCopyBlit(intel, - dst->cpp, - src->pitch, src->buffer, src_offset, - dst->pitch, dst->buffer, dst_offset, - srcx, srcy, dstx, dsty, width, height, - GL_COPY); -} - -/* Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -void -intel_region_fill(intelScreenPrivate *intelScreen, - struct intel_region *dst, - GLuint dst_offset, - GLuint dstx, GLuint dsty, - GLuint width, GLuint height, GLuint color) -{ - struct intel_context *intel = intelScreenContext(intelScreen); - - DBG("%s\n", __FUNCTION__); - - if (intel == NULL) - return; - - if (dst->pbo) { - if (dstx == 0 && - dsty == 0 && width == dst->pitch && height == dst->height) - intel_region_release_pbo(intelScreen, dst); - else - intel_region_cow(intelScreen, dst); - } - - intelEmitFillBlit(intel, - dst->cpp, - dst->pitch, dst->buffer, dst_offset, - dstx, dsty, width, height, color); -} - -/* Attach to a pbo, discarding our data. Effectively zero-copy upload - * the pbo's data. - */ -void -intel_region_attach_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region, - struct intel_buffer_object *pbo) -{ - if (region->pbo == pbo) - return; - - /* If there is already a pbo attached, break the cow tie now. - * Don't call intel_region_release_pbo() as that would - * unnecessarily allocate a new buffer we would have to immediately - * discard. - */ - if (region->pbo) { - region->pbo->region = NULL; - region->pbo = NULL; - } - - if (region->buffer) { - driDeleteBuffers(1, ®ion->buffer); - region->buffer = NULL; - } - - region->pbo = pbo; - region->pbo->region = region; - region->buffer = driBOReference(pbo->buffer); -} - - -/* Break the COW tie to the pbo. The pbo gets to keep the data. - */ -void -intel_region_release_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region) -{ - struct intel_context *intel = intelScreenContext(intelScreen); - - assert(region->buffer == region->pbo->buffer); - region->pbo->region = NULL; - region->pbo = NULL; - driBOUnReference(region->buffer); - region->buffer = NULL; - - driGenBuffers(intelScreen->regionPool, - "region", 1, ®ion->buffer, 64, 0, 0); - - LOCK_HARDWARE(intel); - driBOData(region->buffer, - region->cpp * region->pitch * region->height, NULL, 0); - UNLOCK_HARDWARE(intel); -} - -/* Break the COW tie to the pbo. Both the pbo and the region end up - * with a copy of the data. - */ -void -intel_region_cow(intelScreenPrivate *intelScreen, struct intel_region *region) -{ - struct intel_context *intel = intelScreenContext(intelScreen); - struct intel_buffer_object *pbo = region->pbo; - - if (intel == NULL) - return; - - intel_region_release_pbo(intelScreen, region); - - assert(region->cpp * region->pitch * region->height == pbo->Base.Size); - - DBG("%s (%d bytes)\n", __FUNCTION__, pbo->Base.Size); - - /* Now blit from the texture buffer to the new buffer: - */ - - intel_batchbuffer_flush(intel->batch); - - if (!intel->locked) { - LOCK_HARDWARE(intel); - intelEmitCopyBlit(intel, - region->cpp, - region->pitch, - region->buffer, 0, - region->pitch, - pbo->buffer, 0, - 0, 0, 0, 0, - region->pitch, region->height, - GL_COPY); - - intel_batchbuffer_flush(intel->batch); - UNLOCK_HARDWARE(intel); - } - else { - intelEmitCopyBlit(intel, - region->cpp, - region->pitch, - region->buffer, 0, - region->pitch, - pbo->buffer, 0, - 0, 0, 0, 0, - region->pitch, region->height, - GL_COPY); - - intel_batchbuffer_flush(intel->batch); - } -} - -struct _DriBufferObject * -intel_region_buffer(intelScreenPrivate *intelScreen, - struct intel_region *region, GLuint flag) -{ - if (region->pbo) { - if (flag == INTEL_WRITE_PART) - intel_region_cow(intelScreen, region); - else if (flag == INTEL_WRITE_FULL) - intel_region_release_pbo(intelScreen, region); - } - - return region->buffer; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_regions.h b/src/mesa/drivers/dri/i915tex/intel_regions.h deleted file mode 100644 index d938c107a4..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_regions.h +++ /dev/null @@ -1,141 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_REGIONS_H -#define INTEL_REGIONS_H - -#include "mtypes.h" -#include "intel_screen.h" - -struct intel_context; -struct intel_buffer_object; - -/** - * A layer on top of the bufmgr buffers that adds a few useful things: - * - * - Refcounting for local buffer references. - * - Refcounting for buffer maps - * - Buffer dimensions - pitch and height. - * - Blitter commands for copying 2D regions between buffers. (really???) - */ -struct intel_region -{ - struct _DriBufferObject *buffer; /**< buffer manager's buffer ID */ - GLuint refcount; /**< Reference count for region */ - GLuint cpp; /**< bytes per pixel */ - GLuint pitch; /**< in pixels */ - GLuint height; /**< in pixels */ - GLubyte *map; /**< only non-NULL when region is actually mapped */ - GLuint map_refcount; /**< Reference count for mapping */ - - GLuint draw_offset; /**< Offset of drawing address within the region */ - - struct intel_buffer_object *pbo; /* zero-copy uploads */ -}; - - -/* Allocate a refcounted region. Pointers to regions should only be - * copied by calling intel_reference_region(). - */ -struct intel_region *intel_region_alloc(intelScreenPrivate *intelScreen, - GLuint cpp, - GLuint pitch, GLuint height); - -void intel_region_reference(struct intel_region **dst, - struct intel_region *src); - -void intel_region_release(struct intel_region **ib); - -extern struct intel_region -*intel_region_create_static(intelScreenPrivate *intelScreen, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, - GLuint pitch, GLuint height); -extern void -intel_region_update_static(intelScreenPrivate *intelScreen, - struct intel_region *region, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height); - - -void intel_region_idle(intelScreenPrivate *intelScreen, - struct intel_region *ib); - -/* Map/unmap regions. This is refcounted also: - */ -GLubyte *intel_region_map(intelScreenPrivate *intelScreen, - struct intel_region *ib); - -void intel_region_unmap(intelScreenPrivate *intelScreen, struct intel_region *ib); - - -/* Upload data to a rectangular sub-region - */ -void intel_region_data(intelScreenPrivate *intelScreen, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - const void *src, GLuint src_stride, - GLuint srcx, GLuint srcy, GLuint width, GLuint height); - -/* Copy rectangular sub-regions - */ -void intel_region_copy(intelScreenPrivate *intelScreen, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - struct intel_region *src, - GLuint src_offset, - GLuint srcx, GLuint srcy, GLuint width, GLuint height); - -/* Fill a rectangular sub-region - */ -void intel_region_fill(intelScreenPrivate *intelScreen, - struct intel_region *dest, - GLuint dest_offset, - GLuint destx, GLuint desty, - GLuint width, GLuint height, GLuint color); - -/* Helpers for zerocopy uploads, particularly texture image uploads: - */ -void intel_region_attach_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region, - struct intel_buffer_object *pbo); -void intel_region_release_pbo(intelScreenPrivate *intelScreen, - struct intel_region *region); -void intel_region_cow(intelScreenPrivate *intelScreen, - struct intel_region *region); - -struct _DriBufferObject *intel_region_buffer(intelScreenPrivate *intelScreen, - struct intel_region *region, - GLuint flag); - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_render.c b/src/mesa/drivers/dri/i915tex/intel_render.c deleted file mode 100644 index dddbe9906e..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_render.c +++ /dev/null @@ -1,247 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* - * Render unclipped vertex buffers by emitting vertices directly to - * dma buffers. Use strip/fan hardware acceleration where possible. - * - */ -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" -#include "enums.h" - -#include "tnl/t_context.h" -#include "tnl/t_vertex.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_tris.h" -#include "intel_batchbuffer.h" -#include "intel_reg.h" - -/* - * Render unclipped vertex buffers by emitting vertices directly to - * dma buffers. Use strip/fan hardware primitives where possible. - * Try to simulate missing primitives with indexed vertices. - */ -#define HAVE_POINTS 0 /* Has it, but can't use because subpixel has to - * be adjusted for points on the INTEL/I845G - */ -#define HAVE_LINES 1 -#define HAVE_LINE_STRIPS 1 -#define HAVE_TRIANGLES 1 -#define HAVE_TRI_STRIPS 1 -#define HAVE_TRI_STRIP_1 0 /* has it, template can't use it yet */ -#define HAVE_TRI_FANS 1 -#define HAVE_POLYGONS 1 -#define HAVE_QUADS 0 -#define HAVE_QUAD_STRIPS 0 - -#define HAVE_ELTS 0 - -static GLuint hw_prim[GL_POLYGON + 1] = { - 0, - PRIM3D_LINELIST, - PRIM3D_LINESTRIP, - PRIM3D_LINESTRIP, - PRIM3D_TRILIST, - PRIM3D_TRISTRIP, - PRIM3D_TRIFAN, - 0, - 0, - PRIM3D_POLY -}; - -static const GLenum reduced_prim[GL_POLYGON + 1] = { - GL_POINTS, - GL_LINES, - GL_LINES, - GL_LINES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES -}; - -static const int scale_prim[GL_POLYGON + 1] = { - 0, /* fallback case */ - 1, - 2, - 2, - 1, - 3, - 3, - 0, /* fallback case */ - 0, /* fallback case */ - 3 -}; - - -static void -intelDmaPrimitive(struct intel_context *intel, GLenum prim) -{ - if (0) - fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim)); - INTEL_FIREVERTICES(intel); - intel->vtbl.reduced_primitive_state(intel, reduced_prim[prim]); - intelStartInlinePrimitive(intel, hw_prim[prim], INTEL_BATCH_CLIPRECTS); -} - - -#define LOCAL_VARS struct intel_context *intel = intel_context(ctx) -#define INIT( prim ) \ -do { \ - intelDmaPrimitive( intel, prim ); \ -} while (0) - -#define FLUSH() INTEL_FIREVERTICES(intel) - -#define GET_SUBSEQUENT_VB_MAX_VERTS() \ - ((intel->batch->size - 1500) / (intel->vertex_size*4)) -#define GET_CURRENT_VB_MAX_VERTS() GET_SUBSEQUENT_VB_MAX_VERTS() - -#define ALLOC_VERTS( nr ) NULL - -// intelExtendInlinePrimitive( intel, (nr) * intel->vertex_size ) - -#define EMIT_VERTS( ctx, j, nr, buf ) \ - _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf ) - -#define TAG(x) intel_##x -#include "tnl_dd/t_dd_dmatmp.h" - - -/**********************************************************************/ -/* Render pipeline stage */ -/**********************************************************************/ - -/* Heuristic to choose between the two render paths: - */ -static GLboolean -choose_render(struct intel_context *intel, struct vertex_buffer *VB) -{ - int vertsz = intel->vertex_size; - int cost_render = 0; - int cost_fallback = 0; - int nr_prims = 0; - int nr_rprims = 0; - int nr_rverts = 0; - int rprim = intel->reduced_primitive; - int i = 0; - - for (i = 0; i < VB->PrimitiveCount; i++) { - GLuint prim = VB->Primitive[i].mode; - GLuint length = VB->Primitive[i].count; - - if (!length) - continue; - - nr_prims++; - nr_rverts += length * scale_prim[prim & PRIM_MODE_MASK]; - - if (reduced_prim[prim & PRIM_MODE_MASK] != rprim) { - nr_rprims++; - rprim = reduced_prim[prim & PRIM_MODE_MASK]; - } - } - - /* One point for each generated primitive: - */ - cost_render = nr_prims; - cost_fallback = nr_rprims; - - /* One point for every 1024 dwords (4k) of dma: - */ - cost_render += (vertsz * i) / 1024; - cost_fallback += (vertsz * nr_rverts) / 1024; - - if (0) - fprintf(stderr, "cost render: %d fallback: %d\n", - cost_render, cost_fallback); - - if (cost_render > cost_fallback) - return GL_FALSE; - - return GL_TRUE; -} - - -static GLboolean -intel_run_render(GLcontext * ctx, struct tnl_pipeline_stage *stage) -{ - struct intel_context *intel = intel_context(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *VB = &tnl->vb; - GLuint i; - - intel->vtbl.render_prevalidate( intel ); - - return GL_TRUE; - - /* Don't handle clipping or indexed vertices. - */ - if (intel->RenderIndex != 0 || - !intel_validate_render(ctx, VB) || !choose_render(intel, VB)) { - return GL_TRUE; - } - - tnl->clipspace.new_inputs |= VERT_BIT_POS; - - tnl->Driver.Render.Start(ctx); - - for (i = 0; i < VB->PrimitiveCount; i++) { - GLuint prim = VB->Primitive[i].mode; - GLuint start = VB->Primitive[i].start; - GLuint length = VB->Primitive[i].count; - - if (!length) - continue; - - intel_render_tab_verts[prim & PRIM_MODE_MASK] (ctx, start, - start + length, prim); - } - - tnl->Driver.Render.Finish(ctx); - - INTEL_FIREVERTICES(intel); - - return GL_FALSE; /* finished the pipe */ -} - -const struct tnl_pipeline_stage _intel_render_stage = { - "intel render", - NULL, - NULL, - NULL, - NULL, - intel_run_render /* run */ -}; diff --git a/src/mesa/drivers/dri/i915tex/intel_rotate.c b/src/mesa/drivers/dri/i915tex/intel_rotate.c deleted file mode 100644 index 12d98c4ad2..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_rotate.c +++ /dev/null @@ -1,237 +0,0 @@ - -/** - * Routines for simple 2D->2D transformations for rotated, flipped screens. - * - * XXX This code is not intel-specific. Move it into a common/utility - * someday. - */ - -#include "intel_rotate.h" - -#define MIN2(A, B) ( ((A) < (B)) ? (A) : (B) ) - -#define ABS(A) ( ((A) < 0) ? -(A) : (A) ) - - -void -matrix23Set(struct matrix23 *m, - int m00, int m01, int m02, int m10, int m11, int m12) -{ - m->m00 = m00; - m->m01 = m01; - m->m02 = m02; - m->m10 = m10; - m->m11 = m11; - m->m12 = m12; -} - - -/* - * Transform (x,y) coordinate by the given matrix. - */ -void -matrix23TransformCoordf(const struct matrix23 *m, float *x, float *y) -{ - const float x0 = *x; - const float y0 = *y; - - *x = m->m00 * x0 + m->m01 * y0 + m->m02; - *y = m->m10 * x0 + m->m11 * y0 + m->m12; -} - - -void -matrix23TransformCoordi(const struct matrix23 *m, int *x, int *y) -{ - const int x0 = *x; - const int y0 = *y; - - *x = m->m00 * x0 + m->m01 * y0 + m->m02; - *y = m->m10 * x0 + m->m11 * y0 + m->m12; -} - - -/* - * Transform a width and height by the given matrix. - * XXX this could be optimized quite a bit. - */ -void -matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist) -{ - int x0 = 0, y0 = 0; - int x1 = *xDist, y1 = 0; - int x2 = 0, y2 = *yDist; - matrix23TransformCoordi(m, &x0, &y0); - matrix23TransformCoordi(m, &x1, &y1); - matrix23TransformCoordi(m, &x2, &y2); - - *xDist = (x1 - x0) + (x2 - x0); - *yDist = (y1 - y0) + (y2 - y0); - - if (*xDist < 0) - *xDist = -*xDist; - if (*yDist < 0) - *yDist = -*yDist; -} - - -/** - * Transform the rect defined by (x, y, w, h) by m. - */ -void -matrix23TransformRect(const struct matrix23 *m, int *x, int *y, int *w, - int *h) -{ - int x0 = *x, y0 = *y; - int x1 = *x + *w, y1 = *y; - int x2 = *x + *w, y2 = *y + *h; - int x3 = *x, y3 = *y + *h; - matrix23TransformCoordi(m, &x0, &y0); - matrix23TransformCoordi(m, &x1, &y1); - matrix23TransformCoordi(m, &x2, &y2); - matrix23TransformCoordi(m, &x3, &y3); - *w = ABS(x1 - x0) + ABS(x2 - x1); - /**w = ABS(*w);*/ - *h = ABS(y1 - y0) + ABS(y2 - y1); - /**h = ABS(*h);*/ - *x = MIN2(x0, x1); - *x = MIN2(*x, x2); - *y = MIN2(y0, y1); - *y = MIN2(*y, y2); -} - - -/* - * Make rotation matrix for width X height screen. - */ -void -matrix23Rotate(struct matrix23 *m, int width, int height, int angle) -{ - switch (angle) { - case 0: - matrix23Set(m, 1, 0, 0, 0, 1, 0); - break; - case 90: - matrix23Set(m, 0, 1, 0, -1, 0, width); - break; - case 180: - matrix23Set(m, -1, 0, width, 0, -1, height); - break; - case 270: - matrix23Set(m, 0, -1, height, 1, 0, 0); - break; - default: - /*abort() */ ; - } -} - - -/* - * Make flip/reflection matrix for width X height screen. - */ -void -matrix23Flip(struct matrix23 *m, int width, int height, int xflip, int yflip) -{ - if (xflip) { - m->m00 = -1; - m->m01 = 0; - m->m02 = width - 1; - } - else { - m->m00 = 1; - m->m01 = 0; - m->m02 = 0; - } - if (yflip) { - m->m10 = 0; - m->m11 = -1; - m->m12 = height - 1; - } - else { - m->m10 = 0; - m->m11 = 1; - m->m12 = 0; - } -} - - -/* - * result = a * b - */ -void -matrix23Multiply(struct matrix23 *result, - const struct matrix23 *a, const struct matrix23 *b) -{ - result->m00 = a->m00 * b->m00 + a->m01 * b->m10; - result->m01 = a->m00 * b->m01 + a->m01 * b->m11; - result->m02 = a->m00 * b->m02 + a->m01 * b->m12 + a->m02; - - result->m10 = a->m10 * b->m00 + a->m11 * b->m10; - result->m11 = a->m10 * b->m01 + a->m11 * b->m11; - result->m12 = a->m10 * b->m02 + a->m11 * b->m12 + a->m12; -} - - -#if 000 - -#include - -int -main(int argc, char *argv[]) -{ - int width = 500, height = 400; - int rot; - int fx = 0, fy = 0; /* flip x and/or y ? */ - int coords[4][2]; - - /* four corner coords to test with */ - coords[0][0] = 0; - coords[0][1] = 0; - coords[1][0] = width - 1; - coords[1][1] = 0; - coords[2][0] = width - 1; - coords[2][1] = height - 1; - coords[3][0] = 0; - coords[3][1] = height - 1; - - - for (rot = 0; rot < 360; rot += 90) { - struct matrix23 rotate, flip, m; - int i; - - printf("Rot %d, xFlip %d, yFlip %d:\n", rot, fx, fy); - - /* make transformation matrix 'm' */ - matrix23Rotate(&rotate, width, height, rot); - matrix23Flip(&flip, width, height, fx, fy); - matrix23Multiply(&m, &rotate, &flip); - - /* xform four coords */ - for (i = 0; i < 4; i++) { - int x = coords[i][0]; - int y = coords[i][1]; - matrix23TransformCoordi(&m, &x, &y); - printf(" %d, %d -> %d %d\n", coords[i][0], coords[i][1], x, y); - } - - /* xform width, height */ - { - int x = width; - int y = height; - matrix23TransformDistance(&m, &x, &y); - printf(" %d x %d -> %d x %d\n", width, height, x, y); - } - - /* xform rect */ - { - int x = 50, y = 10, w = 200, h = 100; - matrix23TransformRect(&m, &x, &y, &w, &h); - printf(" %d,%d %d x %d -> %d, %d %d x %d\n", 50, 10, 200, 100, - x, y, w, h); - } - - } - - return 0; -} -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_rotate.h b/src/mesa/drivers/dri/i915tex/intel_rotate.h deleted file mode 100644 index 9c8802ca47..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_rotate.h +++ /dev/null @@ -1,39 +0,0 @@ -#ifndef INTEL_ROTATE_H -#define INTEL_ROTATE_H 1 - -struct matrix23 -{ - int m00, m01, m02; - int m10, m11, m12; -}; - - - -extern void -matrix23Set(struct matrix23 *m, - int m00, int m01, int m02, int m10, int m11, int m12); - -extern void matrix23TransformCoordi(const struct matrix23 *m, int *x, int *y); - -extern void -matrix23TransformCoordf(const struct matrix23 *m, float *x, float *y); - -extern void -matrix23TransformDistance(const struct matrix23 *m, int *xDist, int *yDist); - -extern void -matrix23TransformRect(const struct matrix23 *m, - int *x, int *y, int *w, int *h); - -extern void -matrix23Rotate(struct matrix23 *m, int width, int height, int angle); - -extern void -matrix23Flip(struct matrix23 *m, int width, int height, int xflip, int yflip); - -extern void -matrix23Multiply(struct matrix23 *result, - const struct matrix23 *a, const struct matrix23 *b); - - -#endif /* INTEL_ROTATE_H */ diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c deleted file mode 100644 index 3c0ae854e4..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_screen.c +++ /dev/null @@ -1,755 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "context.h" -#include "framebuffer.h" -#include "matrix.h" -#include "renderbuffer.h" -#include "simple_list.h" -#include "utils.h" -#include "vblank.h" -#include "xmlpool.h" - - -#include "intel_screen.h" - -#include "intel_buffers.h" -#include "intel_tex.h" -#include "intel_span.h" -#include "intel_tris.h" -#include "intel_ioctl.h" -#include "intel_fbo.h" - -#include "i830_dri.h" -#include "dri_bufpool.h" -#include "intel_regions.h" -#include "intel_batchbuffer.h" - -PUBLIC const char __driConfigOptions[] = - DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE - DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) - DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) - DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY - DRI_CONF_FORCE_S3TC_ENABLE(false) - DRI_CONF_ALLOW_LARGE_TEXTURES(1) - DRI_CONF_SECTION_END DRI_CONF_END; - const GLuint __driNConfigOptions = 4; - -#ifdef USE_NEW_INTERFACE - static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; -#endif /*USE_NEW_INTERFACE */ - - extern const struct dri_extension card_extensions[]; - -/** - * Map all the memory regions described by the screen. - * \return GL_TRUE if success, GL_FALSE if error. - */ -GLboolean -intelMapScreenRegions(__DRIscreenPrivate * sPriv) -{ - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - - if (intelScreen->front.handle) { - if (drmMap(sPriv->fd, - intelScreen->front.handle, - intelScreen->front.size, - (drmAddress *) & intelScreen->front.map) != 0) { - _mesa_problem(NULL, "drmMap(frontbuffer) failed!"); - return GL_FALSE; - } - } - else { - _mesa_warning(NULL, "no front buffer handle in intelMapScreenRegions!"); - } - - if (0) - printf("Mappings: front: %p\n", intelScreen->front.map); - return GL_TRUE; -} - - -static struct intel_region * -intel_recreate_static(intelScreenPrivate *intelScreen, - struct intel_region *region, - GLuint mem_type, - GLuint offset, - void *virtual, - GLuint cpp, GLuint pitch, GLuint height) -{ - if (region) { - intel_region_update_static(intelScreen, region, mem_type, offset, - virtual, cpp, pitch, height); - } else { - region = intel_region_create_static(intelScreen, mem_type, offset, - virtual, cpp, pitch, height); - } - return region; -} - - -/* Create intel_region structs to describe the static front,back,depth - * buffers created by the xserver. - * Only used for real front buffer now. - * - * Note that these don't allocate video memory, just describe - * allocations already made by the X server. - */ -static void -intel_recreate_static_regions(intelScreenPrivate *intelScreen) -{ -/* this is the real front buffer which is only used for blitting to */ - intelScreen->front_region = - intel_recreate_static(intelScreen, - intelScreen->front_region, - DRM_BO_FLAG_MEM_TT, - intelScreen->front.offset, - intelScreen->front.map, - intelScreen->cpp, - intelScreen->front.pitch / intelScreen->cpp, - intelScreen->height); - -} - -/** - * Use the information in the sarea to update the screen parameters - * related to screen rotation. Needs to be called locked. - */ -void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) -{ - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - - intelUnmapScreenRegions(intelScreen); - intelUpdateScreenFromSAREA(intelScreen, sarea); - if (!intelMapScreenRegions(sPriv)) { - fprintf(stderr, "ERROR Remapping screen regions!!!\n"); - } - intel_recreate_static_regions(intelScreen); -} - - -void -intelUnmapScreenRegions(intelScreenPrivate * intelScreen) -{ -#define REALLY_UNMAP 1 - if (intelScreen->front.map) { -#if REALLY_UNMAP - if (drmUnmap(intelScreen->front.map, intelScreen->front.size) != 0) - printf("drmUnmap front failed!\n"); -#endif - intelScreen->front.map = NULL; - } -} - - -static void -intelPrintDRIInfo(intelScreenPrivate * intelScreen, - __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) -{ - fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", - intelScreen->front.size, intelScreen->front.offset, - intelScreen->front.pitch); - fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); -} - - -static void -intelPrintSAREA(const drmI830Sarea * sarea) -{ - fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, - sarea->height); - fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); - fprintf(stderr, - "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->front_offset, sarea->front_size, - (unsigned) sarea->front_handle); - fprintf(stderr, - "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->back_offset, sarea->back_size, - (unsigned) sarea->back_handle); - fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->depth_offset, sarea->depth_size, - (unsigned) sarea->depth_handle); - fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); - fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); - fprintf(stderr, - "SAREA: rotated offset: 0x%08x size: 0x%x\n", - sarea->rotated_offset, sarea->rotated_size); - fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); -} - - -/** - * A number of the screen parameters are obtained/computed from - * information in the SAREA. This function updates those parameters. - */ -void -intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, - drmI830Sarea * sarea) -{ - intelScreen->width = sarea->width; - intelScreen->height = sarea->height; - - intelScreen->front.offset = sarea->front_offset; - intelScreen->front.pitch = sarea->pitch * intelScreen->cpp; - intelScreen->front.handle = sarea->front_handle; - intelScreen->front.size = sarea->front_size; - - if (0) - intelPrintSAREA(sarea); -} - -GLboolean -intelCreatePools(intelScreenPrivate *intelScreen) -{ - unsigned batchPoolSize = 1024*1024; - __DRIscreenPrivate * sPriv = intelScreen->driScrnPriv; - - if (intelScreen->havePools) - return GL_TRUE; - - batchPoolSize /= intelScreen->maxBatchSize; - intelScreen->regionPool = driDRMPoolInit(sPriv->fd); - - if (!intelScreen->regionPool) - return GL_FALSE; - - intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); - - if (!intelScreen->staticPool) - return GL_FALSE; - - intelScreen->texPool = intelScreen->regionPool; - - intelScreen->batchPool = driBatchPoolInit(sPriv->fd, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT | - DRM_BO_FLAG_MEM_LOCAL, - intelScreen->maxBatchSize, - batchPoolSize, 5); - if (!intelScreen->batchPool) { - fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); - return GL_FALSE; - } - - intel_recreate_static_regions(intelScreen); - intelScreen->havePools = GL_TRUE; - - return GL_TRUE; -} - - -static GLboolean -intelInitDriver(__DRIscreenPrivate * sPriv) -{ - intelScreenPrivate *intelScreen; - I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; - drmI830Sarea *sarea; - - PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = - (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> - getProcAddress("glxEnableExtension")); - void *const psc = sPriv->psc->screenConfigs; - - if (sPriv->devPrivSize != sizeof(I830DRIRec)) { - fprintf(stderr, - "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); - return GL_FALSE; - } - - /* Allocate the private area */ - intelScreen = (intelScreenPrivate *) CALLOC(sizeof(intelScreenPrivate)); - if (!intelScreen) { - fprintf(stderr, "\nERROR! Allocating private area failed\n"); - return GL_FALSE; - } - /* parse information in __driConfigOptions */ - driParseOptionInfo(&intelScreen->optionCache, - __driConfigOptions, __driNConfigOptions); - - intelScreen->driScrnPriv = sPriv; - sPriv->private = (void *) intelScreen; - intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset; - sarea = (drmI830Sarea *) - (((GLubyte *) sPriv->pSAREA) + intelScreen->sarea_priv_offset); - - intelScreen->maxBatchSize = BATCH_SZ; - intelScreen->deviceID = gDRIPriv->deviceID; - if (intelScreen->deviceID == PCI_CHIP_I865_G) - intelScreen->maxBatchSize = 4096; - - intelScreen->mem = gDRIPriv->mem; - intelScreen->cpp = gDRIPriv->cpp; - - switch (gDRIPriv->bitsPerPixel) { - case 16: - break; - case 32: - break; - default: - exit(1); - break; - } - - intelUpdateScreenFromSAREA(intelScreen, sarea); - - if (!intelMapScreenRegions(sPriv)) { - fprintf(stderr, "\nERROR! mapping regions\n"); - _mesa_free(intelScreen); - sPriv->private = NULL; - return GL_FALSE; - } - - - intelScreen->sarea_priv_offset = gDRIPriv->sarea_priv_offset; - - if (0) - intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); - - intelScreen->drmMinor = sPriv->drmMinor; - - /* Determine if IRQs are active? */ - { - int ret; - drmI830GetParam gp; - - gp.param = I830_PARAM_IRQ_ACTIVE; - gp.value = &intelScreen->irq_active; - - ret = drmCommandWriteRead(sPriv->fd, DRM_I830_GETPARAM, - &gp, sizeof(gp)); - if (ret) { - fprintf(stderr, "drmI830GetParam: %d\n", ret); - return GL_FALSE; - } - } - - /* Determine if batchbuffers are allowed */ - { - int ret; - drmI830GetParam gp; - - gp.param = I830_PARAM_ALLOW_BATCHBUFFER; - gp.value = &intelScreen->allow_batchbuffer; - - ret = drmCommandWriteRead(sPriv->fd, DRM_I830_GETPARAM, - &gp, sizeof(gp)); - if (ret) { - fprintf(stderr, "drmI830GetParam: (%d) %d\n", gp.param, ret); - return GL_FALSE; - } - } - - if (glx_enable_extension != NULL) { - (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); - (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); - (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); - } - - return GL_TRUE; -} - - -static void -intelDestroyScreen(__DRIscreenPrivate * sPriv) -{ - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - - intelUnmapScreenRegions(intelScreen); - - if (intelScreen->havePools) { - driPoolTakeDown(intelScreen->regionPool); - driPoolTakeDown(intelScreen->staticPool); - driPoolTakeDown(intelScreen->batchPool); - } - FREE(intelScreen); - sPriv->private = NULL; -} - - -/** - * This is called when we need to set up GL rendering to a new X window. - */ -static GLboolean -intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, - __DRIdrawablePrivate * driDrawPriv, - const __GLcontextModes * mesaVis, GLboolean isPixmap) -{ -#if 0 - intelScreenPrivate *screen = (intelScreenPrivate *) driScrnPriv->private; -#endif - - if (isPixmap) { - return GL_FALSE; /* not implemented */ - } - else { - GLboolean swStencil = (mesaVis->stencilBits > 0 && - mesaVis->depthBits != 24); - GLenum rgbFormat = (mesaVis->redBits == 5 ? GL_RGB5 : GL_RGBA8); - - struct intel_framebuffer *intel_fb = CALLOC_STRUCT(intel_framebuffer); - - if (!intel_fb) - return GL_FALSE; - - _mesa_initialize_framebuffer(&intel_fb->Base, mesaVis); - - { - /* fake frontbuffer */ - /* XXX allocation should only happen in the unusual case - it's actually needed */ - intel_fb->color_rb[0] - = intel_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_FRONT_LEFT, - &intel_fb->color_rb[0]->Base); - } - - if (mesaVis->doubleBufferMode) { - intel_fb->color_rb[1] - = intel_new_renderbuffer_fb(rgbFormat); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_BACK_LEFT, - &intel_fb->color_rb[1]->Base); - -#if 0 - if (screen->third.handle) { - struct gl_renderbuffer *tmp_rb = NULL; - - intel_fb->color_rb[2] - = intel_new_renderbuffer_fb(rgbFormat); - _mesa_reference_renderbuffer(&tmp_rb, &intel_fb->color_rb[2]->Base); - } -#endif - } - - if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) { - /* combined depth/stencil buffer */ - struct intel_renderbuffer *depthStencilRb - = intel_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT); - /* note: bind RB to two attachment points */ - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, - &depthStencilRb->Base); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_STENCIL, - &depthStencilRb->Base); - } - else if (mesaVis->depthBits == 16) { - /* just 16-bit depth buffer, no hw stencil */ - struct intel_renderbuffer *depthRb - = intel_new_renderbuffer_fb(GL_DEPTH_COMPONENT16); - _mesa_add_renderbuffer(&intel_fb->Base, BUFFER_DEPTH, &depthRb->Base); - } - - - /* now add any/all software-based renderbuffers we may need */ - _mesa_add_soft_renderbuffers(&intel_fb->Base, - GL_FALSE, /* never sw color */ - GL_FALSE, /* never sw depth */ - swStencil, mesaVis->accumRedBits > 0, - GL_FALSE, /* never sw alpha */ - GL_FALSE /* never sw aux */ ); - driDrawPriv->driverPrivate = (void *) intel_fb; - - return GL_TRUE; - } -} - -static void -intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) -{ - _mesa_unreference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate))); -} - - -/** - * Get information about previous buffer swaps. - */ -static int -intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) -{ - struct intel_framebuffer *intel_fb; - - if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) - || (sInfo == NULL)) { - return -1; - } - - intel_fb = dPriv->driverPrivate; - sInfo->swap_count = intel_fb->swap_count; - sInfo->swap_ust = intel_fb->swap_ust; - sInfo->swap_missed_count = intel_fb->swap_missed_count; - - sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0) - ? driCalculateSwapUsage(dPriv, 0, intel_fb->swap_missed_ust) - : 0.0; - - return 0; -} - - -/* There are probably better ways to do this, such as an - * init-designated function to register chipids and createcontext - * functions. - */ -extern GLboolean i830CreateContext(const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - -extern GLboolean i915CreateContext(const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - - - - -static GLboolean -intelCreateContext(const __GLcontextModes * mesaVis, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate) -{ - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - intelScreenPrivate *intelScreen = (intelScreenPrivate *) sPriv->private; - - switch (intelScreen->deviceID) { - /* Don't deal with i830 until texture work complete: - */ - case PCI_CHIP_845_G: - case PCI_CHIP_I830_M: - case PCI_CHIP_I855_GM: - case PCI_CHIP_I865_G: - return i830CreateContext(mesaVis, driContextPriv, sharedContextPrivate); - - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q35_G: - case PCI_CHIP_Q33_G: - return i915CreateContext(mesaVis, driContextPriv, sharedContextPrivate); - - default: - fprintf(stderr, "Unrecognized deviceID %x\n", intelScreen->deviceID); - return GL_FALSE; - } -} - - -static const struct __DriverAPIRec intelAPI = { - .InitDriver = intelInitDriver, - .DestroyScreen = intelDestroyScreen, - .CreateContext = intelCreateContext, - .DestroyContext = intelDestroyContext, - .CreateBuffer = intelCreateBuffer, - .DestroyBuffer = intelDestroyBuffer, - .SwapBuffers = intelSwapBuffers, - .MakeCurrent = intelMakeCurrent, - .UnbindContext = intelUnbindContext, - .GetSwapInfo = intelGetSwapInfo, - .GetMSC = driGetMSC32, - .WaitForMSC = driWaitForMSC32, - .WaitForSBC = NULL, - .SwapBuffersMSC = NULL, - .CopySubBuffer = intelCopySubBuffer, - .setTexOffset = intelSetTexOffset, -}; - - -static __GLcontextModes * -intelFillInModes(unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, GLboolean have_back_buffer) -{ - __GLcontextModes *modes; - __GLcontextModes *m; - unsigned num_modes; - unsigned depth_buffer_factor; - unsigned back_buffer_factor; - GLenum fb_format; - GLenum fb_type; - - /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't - * support pageflipping at all. - */ - static const GLenum back_buffer_modes[] = { - GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML - }; - - u_int8_t depth_bits_array[3]; - u_int8_t stencil_bits_array[3]; - - - depth_bits_array[0] = 0; - depth_bits_array[1] = depth_bits; - depth_bits_array[2] = depth_bits; - - /* Just like with the accumulation buffer, always provide some modes - * with a stencil buffer. It will be a sw fallback, but some apps won't - * care about that. - */ - stencil_bits_array[0] = 0; - stencil_bits_array[1] = 0; - if (depth_bits == 24) - stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; - - stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; - - depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; - back_buffer_factor = (have_back_buffer) ? 3 : 1; - - num_modes = depth_buffer_factor * back_buffer_factor * 4; - - if (pixel_bits == 16) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; - } - - modes = - (*dri_interface->createContextModes) (num_modes, - sizeof(__GLcontextModes)); - m = modes; - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_TRUE_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_DIRECT_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - - /* Mark the visual as slow if there are "fake" stencil bits. - */ - for (m = modes; m != NULL; m = m->next) { - if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { - m->visualRating = GLX_SLOW_CONFIG; - } - } - - return modes; -} - - -/** - * This is the bootstrap function for the driver. libGL supplies all of the - * requisite information about the system, and the driver initializes itself. - * This routine also fills in the linked list pointed to by \c driver_modes - * with the \c __GLcontextModes that the driver can support for windows or - * pbuffers. - * - * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on - * failure. - */ -PUBLIC void * -__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, - __DRIscreen * psc, - const __GLcontextModes * modes, - const __DRIversion * ddx_version, - const __DRIversion * dri_version, - const __DRIversion * drm_version, - const __DRIframebuffer * frame_buffer, - drmAddress pSAREA, int fd, - int internal_api_version, - const __DRIinterfaceMethods * interface, - __GLcontextModes ** driver_modes) -{ - __DRIscreenPrivate *psp; - static const __DRIversion ddx_expected = { 1, 7, 0 }; - static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 7, 0 }; - - dri_interface = interface; - - if (!driCheckDriDdxDrmVersions2("i915", - dri_version, &dri_expected, - ddx_version, &ddx_expected, - drm_version, &drm_expected)) { - return NULL; - } - - psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, - ddx_version, dri_version, drm_version, - frame_buffer, pSAREA, fd, - internal_api_version, &intelAPI); - - if (psp != NULL) { - I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; - *driver_modes = intelFillInModes(dri_priv->cpp * 8, - (dri_priv->cpp == 2) ? 16 : 24, - (dri_priv->cpp == 2) ? 0 : 8, 1); - - /* Calling driInitExtensions here, with a NULL context pointer, does not actually - * enable the extensions. It just makes sure that all the dispatch offsets for all - * the extensions that *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is called, but we can't - * enable the extensions until we have a context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions(NULL, card_extensions, GL_FALSE); - } - - return (void *) psp; -} - -struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen) -{ - /* - * This should probably change to have the screen allocate a dummy - * context at screen creation. For now just use the current context. - */ - - GET_CURRENT_CONTEXT(ctx); - if (ctx == NULL) { -/* _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; */ - /* need a context for the first time makecurrent is called (for hw lock - when allocating priv buffers) */ - if (intelScreen->dummyctxptr == NULL) { - _mesa_problem(NULL, "No current context in intelScreenContext\n"); - return NULL; - } - return intelScreen->dummyctxptr; - } - return intel_context(ctx); - -} - diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.h b/src/mesa/drivers/dri/i915tex/intel_screen.h deleted file mode 100644 index 8e4e0b9c12..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_screen.h +++ /dev/null @@ -1,123 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef _INTEL_INIT_H_ -#define _INTEL_INIT_H_ - -#include -#include "dri_util.h" -#include "intel_rotate.h" -#include "i830_common.h" -#include "xmlconfig.h" -#include "dri_bufpool.h" - -/* XXX: change name or eliminate to avoid conflict with "struct - * intel_region"!!! - */ -typedef struct -{ - drm_handle_t handle; - drmSize size; /* region size in bytes */ - char *map; /* memory map */ - int offset; /* from start of video mem, in bytes */ - int pitch; /* row stride, in bytes */ -} intelRegion; - -typedef struct -{ - intelRegion front; - - struct intel_region *front_region; - - int deviceID; - int width; - int height; - int mem; /* unused */ - - int cpp; /* for front and back buffers */ - - __DRIscreenPrivate *driScrnPriv; - unsigned int sarea_priv_offset; - - int drmMinor; - - int irq_active; - int allow_batchbuffer; - - struct matrix23 rotMatrix; - - - /** - * Configuration cache with default values for all contexts - */ - driOptionCache optionCache; - struct _DriBufferPool *batchPool; - struct _DriBufferPool *texPool; - struct _DriBufferPool *regionPool; - struct _DriBufferPool *staticPool; - unsigned int maxBatchSize; - GLboolean havePools; - struct intel_context *dummyctxptr; -} intelScreenPrivate; - - - -extern GLboolean intelMapScreenRegions(__DRIscreenPrivate * sPriv); - -extern void intelUnmapScreenRegions(intelScreenPrivate * intelScreen); - -extern void -intelUpdateScreenFromSAREA(intelScreenPrivate * intelScreen, - drmI830Sarea * sarea); - -extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); - -extern GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); - -extern GLboolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); - -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h); - -extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, - unsigned long bufSize, - unsigned numBufs, - unsigned checkDelayed); - -extern struct intel_context *intelScreenContext(intelScreenPrivate *intelScreen); - -extern void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); -extern GLboolean -intelCreatePools(intelScreenPrivate *intelScreen); - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_span.c b/src/mesa/drivers/dri/i915tex/intel_span.c deleted file mode 100644 index 5a978d9ce2..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_span.c +++ /dev/null @@ -1,401 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "macros.h" -#include "mtypes.h" -#include "colormac.h" - -#include "intel_fbo.h" -#include "intel_screen.h" -#include "intel_span.h" -#include "intel_regions.h" -#include "intel_ioctl.h" -#include "intel_tex.h" - -#include "swrast/swrast.h" - -/* - break intelWriteRGBASpan_ARGB8888 -*/ - -#undef DBG -#define DBG 0 - -#define LOCAL_VARS \ - struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ - const GLint yScale = irb->RenderToTexture ? 1 : -1; \ - const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ - GLubyte *buf = (GLubyte *) irb->pfMap; \ - GLuint p; \ - assert(irb->pfMap);\ - (void) p; - -/* There is just a single cliploop! - */ -#define HW_CLIPLOOP() \ - do { \ - int minx = 0; \ - int miny = 0; \ - int maxx = irb->Base.Width - 1; \ - int maxy = irb->Base.Height - 1; - -#define HW_ENDCLIPLOOP() \ - } while (0) - - -#define Y_FLIP(_y) ((_y) * yScale + yBias) - -#define HW_LOCK() - -#define HW_UNLOCK() - -/* 16 bit, RGB565 color spanline and pixel functions - */ -#define SPANTMP_PIXEL_FMT GL_RGB -#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5 - -#define TAG(x) intel##x##_RGB565 -#define TAG2(x,y) intel##x##_RGB565##y -#define GET_PTR(X,Y) (buf + ((Y) * irb->pfPitch + (X)) * 2) -#include "spantmp2.h" - -/* 32 bit, ARGB8888 color spanline and pixel functions - */ -#define SPANTMP_PIXEL_FMT GL_BGRA -#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV - -#define TAG(x) intel##x##_ARGB8888 -#define TAG2(x,y) intel##x##_ARGB8888##y -#define GET_PTR(X,Y) (buf + ((Y) * irb->pfPitch + (X)) * 4) -#include "spantmp2.h" - - -#define LOCAL_DEPTH_VARS \ - struct intel_renderbuffer *irb = intel_renderbuffer(rb); \ - const GLuint pitch = irb->pfPitch/***XXX region->pitch*/; /* in pixels */ \ - const GLint yScale = irb->RenderToTexture ? 1 : -1; \ - const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \ - char *buf = (char *) irb->pfMap/*XXX use region->map*/ ; - -#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS - -/** - ** 16-bit depthbuffer functions. - **/ -#define WRITE_DEPTH( _x, _y, d ) \ - ((GLushort *)buf)[(_x) + (_y) * pitch] = d; - -#define READ_DEPTH( d, _x, _y ) \ - d = ((GLushort *)buf)[(_x) + (_y) * pitch]; - - -#define TAG(x) intel##x##_z16 -#include "depthtmp.h" - - -/** - ** 24/8-bit interleaved depth/stencil functions - ** Note: we're actually reading back combined depth+stencil values. - ** The wrappers in main/depthstencil.c are used to extract the depth - ** and stencil values. - **/ -/* Change ZZZS -> SZZZ */ -#define WRITE_DEPTH( _x, _y, d ) { \ - GLuint tmp = ((d) >> 8) | ((d) << 24); \ - ((GLuint *)buf)[(_x) + (_y) * pitch] = tmp; \ -} - -/* Change SZZZ -> ZZZS */ -#define READ_DEPTH( d, _x, _y ) { \ - GLuint tmp = ((GLuint *)buf)[(_x) + (_y) * pitch]; \ - d = (tmp << 8) | (tmp >> 24); \ -} - -#define TAG(x) intel##x##_z24_s8 -#include "depthtmp.h" - - -/** - ** 8-bit stencil function (XXX FBO: This is obsolete) - **/ -#define WRITE_STENCIL( _x, _y, d ) { \ - GLuint tmp = ((GLuint *)buf)[(_x) + (_y) * pitch]; \ - tmp &= 0xffffff; \ - tmp |= ((d) << 24); \ - ((GLuint *) buf)[(_x) + (_y) * pitch] = tmp; \ -} - -#define READ_STENCIL( d, _x, _y ) \ - d = ((GLuint *)buf)[(_x) + (_y) * pitch] >> 24; - -#define TAG(x) intel##x##_z24_s8 -#include "stenciltmp.h" - - - -/** - * Map or unmap all the renderbuffers which we may need during - * software rendering. - * XXX in the future, we could probably convey extra information to - * reduce the number of mappings needed. I.e. if doing a glReadPixels - * from the depth buffer, we really only need one mapping. - * - * XXX Rewrite this function someday. - * We can probably just loop over all the renderbuffer attachments, - * map/unmap all of them, and not worry about the _ColorDrawBuffers - * _ColorReadBuffer, _DepthBuffer or _StencilBuffer fields. - */ -static void -intel_map_unmap_buffers(struct intel_context *intel, GLboolean map) -{ - GLcontext *ctx = &intel->ctx; - GLuint i, j; - struct intel_renderbuffer *irb; - - /* color draw buffers */ - for (i = 0; i < ctx->Const.MaxDrawBuffers; i++) { - for (j = 0; j < ctx->DrawBuffer->_NumColorDrawBuffers[i]; j++) { - struct gl_renderbuffer *rb = - ctx->DrawBuffer->_ColorDrawBuffers[i][j]; - irb = intel_renderbuffer(rb); - if (irb) { - /* this is a user-created intel_renderbuffer */ - if (irb->region) { - if (map) - intel_region_map(intel->intelScreen, irb->region); - else - intel_region_unmap(intel->intelScreen, irb->region); - } - irb->pfMap = irb->region->map; - irb->pfPitch = irb->region->pitch; - } - } - } - - /* check for render to textures */ - for (i = 0; i < BUFFER_COUNT; i++) { - struct gl_renderbuffer_attachment *att = - ctx->DrawBuffer->Attachment + i; - struct gl_texture_object *tex = att->Texture; - if (tex) { - /* render to texture */ - ASSERT(att->Renderbuffer); - if (map) { - struct gl_texture_image *texImg; - texImg = tex->Image[att->CubeMapFace][att->TextureLevel]; - intel_tex_map_images(intel, intel_texture_object(tex)); - } - else { - intel_tex_unmap_images(intel, intel_texture_object(tex)); - } - } - } - - /* color read buffers */ - irb = intel_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); - if (irb && irb->region) { - if (map) - intel_region_map(intel->intelScreen, irb->region); - else - intel_region_unmap(intel->intelScreen, irb->region); - irb->pfMap = irb->region->map; - irb->pfPitch = irb->region->pitch; - } - - /* Account for front/back color page flipping. - * The span routines use the pfMap and pfPitch fields which will - * swap the front/back region map/pitch if we're page flipped. - * Do this after mapping, above, so the map field is valid. - */ -#if 0 - if (map && ctx->DrawBuffer->Name == 0) { - struct intel_renderbuffer *irbFront - = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_FRONT_LEFT); - struct intel_renderbuffer *irbBack - = intel_get_renderbuffer(ctx->DrawBuffer, BUFFER_BACK_LEFT); - if (irbBack) { - /* double buffered */ - if (intel->sarea->pf_current_page == 0) { - irbFront->pfMap = irbFront->region->map; - irbFront->pfPitch = irbFront->region->pitch; - irbBack->pfMap = irbBack->region->map; - irbBack->pfPitch = irbBack->region->pitch; - } - else { - irbFront->pfMap = irbBack->region->map; - irbFront->pfPitch = irbBack->region->pitch; - irbBack->pfMap = irbFront->region->map; - irbBack->pfPitch = irbFront->region->pitch; - } - } - } -#endif - - /* depth buffer (Note wrapper!) */ - if (ctx->DrawBuffer->_DepthBuffer) { - irb = intel_renderbuffer(ctx->DrawBuffer->_DepthBuffer->Wrapped); - if (irb && irb->region) { - if (map) { - intel_region_map(intel->intelScreen, irb->region); - irb->pfMap = irb->region->map; - irb->pfPitch = irb->region->pitch; - } - else { - intel_region_unmap(intel->intelScreen, irb->region); - irb->pfMap = NULL; - irb->pfPitch = 0; - } - } - } - - /* stencil buffer (Note wrapper!) */ - if (ctx->DrawBuffer->_StencilBuffer) { - irb = intel_renderbuffer(ctx->DrawBuffer->_StencilBuffer->Wrapped); - if (irb && irb->region) { - if (map) { - intel_region_map(intel->intelScreen, irb->region); - irb->pfMap = irb->region->map; - irb->pfPitch = irb->region->pitch; - } - else { - intel_region_unmap(intel->intelScreen, irb->region); - irb->pfMap = NULL; - irb->pfPitch = 0; - } - } - } -} - - - -/** - * Prepare for softare rendering. Map current read/draw framebuffers' - * renderbuffes and all currently bound texture objects. - * - * Old note: Moved locking out to get reasonable span performance. - */ -void -intelSpanRenderStart(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - GLuint i; - - intelFinish(&intel->ctx); - LOCK_HARDWARE(intel); - -#if 0 - /* Just map the framebuffer and all textures. Bufmgr code will - * take care of waiting on the necessary fences: - */ - intel_region_map(intel->intelScreen, intel->front_region); - intel_region_map(intel->intelScreen, intel->back_region); - intel_region_map(intel->intelScreen, intel->intelScreen->depth_region); -#endif - - for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { - struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; - intel_tex_map_images(intel, intel_texture_object(texObj)); - } - } - - intel_map_unmap_buffers(intel, GL_TRUE); -} - -/** - * Called when done softare rendering. Unmap the buffers we mapped in - * the above function. - */ -void -intelSpanRenderFinish(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - GLuint i; - - _swrast_flush(ctx); - - /* Now unmap the framebuffer: - */ -#if 0 - intel_region_unmap(intel, intel->front_region); - intel_region_unmap(intel, intel->back_region); - intel_region_unmap(intel, intel->intelScreen->depth_region); -#endif - - for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) { - if (ctx->Texture.Unit[i]._ReallyEnabled) { - struct gl_texture_object *texObj = ctx->Texture.Unit[i]._Current; - intel_tex_unmap_images(intel, intel_texture_object(texObj)); - } - } - - intel_map_unmap_buffers(intel, GL_FALSE); - - UNLOCK_HARDWARE(intel); -} - - -void -intelInitSpanFuncs(GLcontext * ctx) -{ - struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx); - swdd->SpanRenderStart = intelSpanRenderStart; - swdd->SpanRenderFinish = intelSpanRenderFinish; -} - - -/** - * Plug in appropriate span read/write functions for the given renderbuffer. - * These are used for the software fallbacks. - */ -void -intel_set_span_functions(struct gl_renderbuffer *rb) -{ - if (rb->_ActualFormat == GL_RGB5) { - /* 565 RGB */ - intelInitPointers_RGB565(rb); - } - else if (rb->_ActualFormat == GL_RGBA8) { - /* 8888 RGBA */ - intelInitPointers_ARGB8888(rb); - } - else if (rb->_ActualFormat == GL_DEPTH_COMPONENT16) { - intelInitDepthPointers_z16(rb); - } - else if (rb->_ActualFormat == GL_DEPTH_COMPONENT24 || /* XXX FBO remove */ - rb->_ActualFormat == GL_DEPTH24_STENCIL8_EXT) { - intelInitDepthPointers_z24_s8(rb); - } - else if (rb->_ActualFormat == GL_STENCIL_INDEX8_EXT) { /* XXX FBO remove */ - intelInitStencilPointers_z24_s8(rb); - } - else { - _mesa_problem(NULL, - "Unexpected _ActualFormat in intelSetSpanFunctions"); - } -} diff --git a/src/mesa/drivers/dri/i915tex/intel_span.h b/src/mesa/drivers/dri/i915tex/intel_span.h deleted file mode 100644 index 5201f6d6c6..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_span.h +++ /dev/null @@ -1,38 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef _INTEL_SPAN_H -#define _INTEL_SPAN_H - -extern void intelInitSpanFuncs(GLcontext * ctx); - -extern void intelSpanRenderFinish(GLcontext * ctx); -extern void intelSpanRenderStart(GLcontext * ctx); - -extern void intel_set_span_functions(struct gl_renderbuffer *rb); - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_state.c b/src/mesa/drivers/dri/i915tex/intel_state.c deleted file mode 100644 index 5c5f2c6de5..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_state.c +++ /dev/null @@ -1,261 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "enums.h" -#include "colormac.h" -#include "dd.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_fbo.h" -#include "intel_regions.h" -#include "swrast/swrast.h" - -int -intel_translate_compare_func(GLenum func) -{ - switch (func) { - case GL_NEVER: - return COMPAREFUNC_NEVER; - case GL_LESS: - return COMPAREFUNC_LESS; - case GL_LEQUAL: - return COMPAREFUNC_LEQUAL; - case GL_GREATER: - return COMPAREFUNC_GREATER; - case GL_GEQUAL: - return COMPAREFUNC_GEQUAL; - case GL_NOTEQUAL: - return COMPAREFUNC_NOTEQUAL; - case GL_EQUAL: - return COMPAREFUNC_EQUAL; - case GL_ALWAYS: - return COMPAREFUNC_ALWAYS; - } - - fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); - return COMPAREFUNC_ALWAYS; -} - -int -intel_translate_stencil_op(GLenum op) -{ - switch (op) { - case GL_KEEP: - return STENCILOP_KEEP; - case GL_ZERO: - return STENCILOP_ZERO; - case GL_REPLACE: - return STENCILOP_REPLACE; - case GL_INCR: - return STENCILOP_INCRSAT; - case GL_DECR: - return STENCILOP_DECRSAT; - case GL_INCR_WRAP: - return STENCILOP_INCR; - case GL_DECR_WRAP: - return STENCILOP_DECR; - case GL_INVERT: - return STENCILOP_INVERT; - default: - return STENCILOP_ZERO; - } -} - -int -intel_translate_blend_factor(GLenum factor) -{ - switch (factor) { - case GL_ZERO: - return BLENDFACT_ZERO; - case GL_SRC_ALPHA: - return BLENDFACT_SRC_ALPHA; - case GL_ONE: - return BLENDFACT_ONE; - case GL_SRC_COLOR: - return BLENDFACT_SRC_COLR; - case GL_ONE_MINUS_SRC_COLOR: - return BLENDFACT_INV_SRC_COLR; - case GL_DST_COLOR: - return BLENDFACT_DST_COLR; - case GL_ONE_MINUS_DST_COLOR: - return BLENDFACT_INV_DST_COLR; - case GL_ONE_MINUS_SRC_ALPHA: - return BLENDFACT_INV_SRC_ALPHA; - case GL_DST_ALPHA: - return BLENDFACT_DST_ALPHA; - case GL_ONE_MINUS_DST_ALPHA: - return BLENDFACT_INV_DST_ALPHA; - case GL_SRC_ALPHA_SATURATE: - return BLENDFACT_SRC_ALPHA_SATURATE; - case GL_CONSTANT_COLOR: - return BLENDFACT_CONST_COLOR; - case GL_ONE_MINUS_CONSTANT_COLOR: - return BLENDFACT_INV_CONST_COLOR; - case GL_CONSTANT_ALPHA: - return BLENDFACT_CONST_ALPHA; - case GL_ONE_MINUS_CONSTANT_ALPHA: - return BLENDFACT_INV_CONST_ALPHA; - } - - fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, factor); - return BLENDFACT_ZERO; -} - -int -intel_translate_logic_op(GLenum opcode) -{ - switch (opcode) { - case GL_CLEAR: - return LOGICOP_CLEAR; - case GL_AND: - return LOGICOP_AND; - case GL_AND_REVERSE: - return LOGICOP_AND_RVRSE; - case GL_COPY: - return LOGICOP_COPY; - case GL_COPY_INVERTED: - return LOGICOP_COPY_INV; - case GL_AND_INVERTED: - return LOGICOP_AND_INV; - case GL_NOOP: - return LOGICOP_NOOP; - case GL_XOR: - return LOGICOP_XOR; - case GL_OR: - return LOGICOP_OR; - case GL_OR_INVERTED: - return LOGICOP_OR_INV; - case GL_NOR: - return LOGICOP_NOR; - case GL_EQUIV: - return LOGICOP_EQUIV; - case GL_INVERT: - return LOGICOP_INV; - case GL_OR_REVERSE: - return LOGICOP_OR_RVRSE; - case GL_NAND: - return LOGICOP_NAND; - case GL_SET: - return LOGICOP_SET; - default: - return LOGICOP_SET; - } -} - - -static void -intelClearColor(GLcontext * ctx, const GLfloat color[4]) -{ - struct intel_context *intel = intel_context(ctx); - GLubyte clear[4]; - - CLAMPED_FLOAT_TO_UBYTE(clear[0], color[0]); - CLAMPED_FLOAT_TO_UBYTE(clear[1], color[1]); - CLAMPED_FLOAT_TO_UBYTE(clear[2], color[2]); - CLAMPED_FLOAT_TO_UBYTE(clear[3], color[3]); - - /* compute both 32 and 16-bit clear values */ - intel->ClearColor8888 = INTEL_PACKCOLOR8888(clear[0], clear[1], - clear[2], clear[3]); - intel->ClearColor565 = INTEL_PACKCOLOR565(clear[0], clear[1], clear[2]); -} - - -/** - * Update the viewport transformation matrix. Depends on: - * - viewport pos/size - * - depthrange - * - window size or FBO size - */ -static void -intelCalcViewport(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - const GLfloat *v = ctx->Viewport._WindowMap.m; - const GLfloat depthScale = 1.0F / ctx->DrawBuffer->_DepthMaxF; - GLfloat *m = intel->ViewportMatrix.m; - GLfloat yScale, yBias; - - struct intel_renderbuffer *irb - = intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); - if (irb && !irb->RenderToTexture) { - /* y=0=top */ - yScale = -1.0; - yBias = irb->Base.Height; - } - else { - /* y=0=bottom */ - yScale = 1.0; - yBias = 0.0; - } - - m[MAT_SX] = v[MAT_SX]; - m[MAT_TX] = v[MAT_TX] + SUBPIXEL_X; - - m[MAT_SY] = v[MAT_SY] * yScale; - m[MAT_TY] = v[MAT_TY] * yScale + yBias + SUBPIXEL_Y; - - m[MAT_SZ] = v[MAT_SZ] * depthScale; - m[MAT_TZ] = v[MAT_TZ] * depthScale; -} - -static void -intelViewport(GLcontext * ctx, - GLint x, GLint y, GLsizei width, GLsizei height) -{ - intelCalcViewport(ctx); -} - -static void -intelDepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval) -{ - intelCalcViewport(ctx); -} - -/* Fallback to swrast for select and feedback. - */ -static void -intelRenderMode(GLcontext * ctx, GLenum mode) -{ - struct intel_context *intel = intel_context(ctx); - FALLBACK(intel, INTEL_FALLBACK_RENDERMODE, (mode != GL_RENDER)); -} - - -void -intelInitStateFuncs(struct dd_function_table *functions) -{ - functions->RenderMode = intelRenderMode; - functions->Viewport = intelViewport; - functions->DepthRange = intelDepthRange; - functions->ClearColor = intelClearColor; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_structs.h b/src/mesa/drivers/dri/i915tex/intel_structs.h deleted file mode 100644 index 522e3bd92c..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_structs.h +++ /dev/null @@ -1,132 +0,0 @@ -#ifndef INTEL_STRUCTS_H -#define INTEL_STRUCTS_H - -struct br0 { - GLuint length:8; - GLuint pad0:3; - GLuint dst_tiled:1; - GLuint pad1:8; - GLuint write_rgb:1; - GLuint write_alpha:1; - GLuint opcode:7; - GLuint client:3; -}; - - -struct br13 { - GLint dest_pitch:16; - GLuint rop:8; - GLuint color_depth:2; - GLuint pad1:3; - GLuint mono_source_transparency:1; - GLuint clipping_enable:1; - GLuint pad0:1; -}; - - - -/* This is an attempt to move some of the 2D interaction in this - * driver to using structs for packets rather than a bunch of #defines - * and dwords. - */ -struct xy_color_blit { - struct br0 br0; - struct br13 br13; - - struct { - GLuint dest_x1:16; - GLuint dest_y1:16; - } dw2; - - struct { - GLuint dest_x2:16; - GLuint dest_y2:16; - } dw3; - - GLuint dest_base_addr; - GLuint color; -}; - -struct xy_src_copy_blit { - struct br0 br0; - struct br13 br13; - - struct { - GLuint dest_x1:16; - GLuint dest_y1:16; - } dw2; - - struct { - GLuint dest_x2:16; - GLuint dest_y2:16; - } dw3; - - GLuint dest_base_addr; - - struct { - GLuint src_x1:16; - GLuint src_y1:16; - } dw5; - - struct { - GLint src_pitch:16; - GLuint pad:16; - } dw6; - - GLuint src_base_addr; -}; - -struct xy_setup_blit { - struct br0 br0; - struct br13 br13; - - struct { - GLuint clip_x1:16; - GLuint clip_y1:16; - } dw2; - - struct { - GLuint clip_x2:16; - GLuint clip_y2:16; - } dw3; - - GLuint dest_base_addr; - GLuint background_color; - GLuint foreground_color; - GLuint pattern_base_addr; -}; - - -struct xy_text_immediate_blit { - struct { - GLuint length:8; - GLuint pad2:3; - GLuint dst_tiled:1; - GLuint pad1:4; - GLuint byte_packed:1; - GLuint pad0:5; - GLuint opcode:7; - GLuint client:3; - } dw0; - - struct { - GLuint dest_x1:16; - GLuint dest_y1:16; - } dw1; - - struct { - GLuint dest_x2:16; - GLuint dest_y2:16; - } dw2; - - /* Src bitmap data follows as inline dwords. - */ -}; - - -#define CLIENT_2D 0x2 -#define OPCODE_XY_SETUP_BLT 0x1 -#define OPCODE_XY_COLOR_BLT 0x50 -#define OPCODE_XY_TEXT_IMMEDIATE_BLT 0x31 - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_surface.c b/src/mesa/drivers/dri/i915tex/intel_surface.c deleted file mode 100644 index 3e86835de2..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_surface.c +++ /dev/null @@ -1,250 +0,0 @@ -#include "glheader.h" -#include "context.h" -#include "framebuffer.h" -#include "renderbuffer.h" -#include "utils.h" -#include "main/macros.h" - - -#include "intel_screen.h" - -#include "intel_context.h" -#include "intel_buffers.h" -#include "intel_regions.h" -#include "intel_span.h" -#include "intel_fbo.h" - -#include "pipe/p_state.h" -#include "pipe/p_defines.h" -#include "pipe/softpipe/sp_surface.h" - - -/* - * XXX a lof of this is a temporary kludge - */ - -/** - * Note: the arithmetic/addressing in these functions is a little - * tricky since we need to invert the Y axis. - */ - - -static void -read_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) -{ - const GLint bytesPerRow = sps->surface.stride * sps->surface.cpp; - const GLint invY = sps->surface.height - y - 1; - const GLubyte *src = sps->surface.ptr + invY * bytesPerRow + x * sps->surface.cpp; - GLfloat *dst = (GLfloat *) rrrr; - GLubyte temp[16]; - GLuint j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - - memcpy(temp + 8, src, 8); - memcpy(temp + 0, src + bytesPerRow, 8); - - for (j = 0; j < 4; j++) { - dst[0 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 2]); /*R*/ - dst[1 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 1]); /*G*/ - dst[2 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 0]); /*B*/ - dst[3 * 4 + j] = UBYTE_TO_FLOAT(temp[j * 4 + 3]); /*A*/ - } -} - - -static void -write_quad_f_swz(struct softpipe_surface *sps, GLint x, GLint y, - GLfloat (*rrrr)[QUAD_SIZE]) -{ - const GLfloat *src = (const GLfloat *) rrrr; - const GLint bytesPerRow = sps->surface.stride * sps->surface.cpp; - const GLint invY = sps->surface.height - y - 1; - GLubyte *dst = sps->surface.ptr + invY * bytesPerRow + x * sps->surface.cpp; - GLubyte temp[16]; - GLuint j; - - assert(sps->surface.format == PIPE_FORMAT_U_A8_R8_G8_B8); - - for (j = 0; j < 4; j++) { - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 2], src[0 * 4 + j]); /*R*/ - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 1], src[1 * 4 + j]); /*G*/ - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 0], src[2 * 4 + j]); /*B*/ - UNCLAMPED_FLOAT_TO_UBYTE(temp[j * 4 + 3], src[3 * 4 + j]); /*A*/ - } - - memcpy(dst, temp + 8, 8); - memcpy(dst + bytesPerRow, temp + 0, 8); -} - - - -static void -read_quad_z24(struct softpipe_surface *sps, - GLint x, GLint y, GLuint zzzz[QUAD_SIZE]) -{ - static const GLuint mask = 0xffffff; - const GLint invY = sps->surface.height - y - 1; - const GLuint *src - = (GLuint *) (sps->surface.ptr - + (invY * sps->surface.stride + x) * sps->surface.cpp); - - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); - - /* extract lower three bytes */ - zzzz[0] = src[0] & mask; - zzzz[1] = src[1] & mask; - src -= sps->surface.stride; - zzzz[2] = src[0] & mask; - zzzz[3] = src[1] & mask; -} - -static void -write_quad_z24(struct softpipe_surface *sps, - GLint x, GLint y, const GLuint zzzz[QUAD_SIZE]) -{ - static const GLuint mask = 0xff000000; - const GLint invY = sps->surface.height - y - 1; - GLuint *dst - = (GLuint *) (sps->surface.ptr - + (invY * sps->surface.stride + x) * sps->surface.cpp); - - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); - - /* write lower three bytes */ - dst[0] = (dst[0] & mask) | zzzz[0]; - dst[1] = (dst[1] & mask) | zzzz[1]; - dst -= sps->surface.stride; - dst[0] = (dst[0] & mask) | zzzz[2]; - dst[1] = (dst[1] & mask) | zzzz[3]; -} - - -static void -read_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, GLubyte ssss[QUAD_SIZE]) -{ - const GLint invY = sps->surface.height - y - 1; - const GLuint *src = (const GLuint *) (sps->surface.ptr - + (invY * sps->surface.stride + x) * sps->surface.cpp); - - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); - - /* extract high byte */ - ssss[0] = src[0] >> 24; - ssss[1] = src[1] >> 24; - src -= sps->surface.stride; - ssss[2] = src[0] >> 24; - ssss[3] = src[1] >> 24; -} - -static void -write_quad_stencil(struct softpipe_surface *sps, - GLint x, GLint y, const GLubyte ssss[QUAD_SIZE]) -{ - static const GLuint mask = 0x00ffffff; - const GLint invY = sps->surface.height - y - 1; - GLuint *dst = (GLuint *) (sps->surface.ptr - + (invY * sps->surface.stride + x) * sps->surface.cpp); - - assert(sps->surface.format == PIPE_FORMAT_Z24_S8); - - /* write high byte */ - dst[0] = (dst[0] & mask) | (ssss[0] << 24); - dst[1] = (dst[1] & mask) | (ssss[1] << 24); - dst -= sps->surface.stride; - dst[0] = (dst[0] & mask) | (ssss[2] << 24); - dst[1] = (dst[1] & mask) | (ssss[3] << 24); -} - - -static void * -map_surface_buffer(struct pipe_buffer *pb, GLuint access_mode) -{ - struct softpipe_surface *sps = (struct softpipe_surface *) pb; - struct intel_renderbuffer *irb = (struct intel_renderbuffer *) sps->surface.rb; - assert(access_mode == PIPE_MAP_READ_WRITE); - - /*LOCK_HARDWARE(intel);*/ - - if (irb->region) { - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); -#if 0 - intelFinish(&intel->ctx); /* XXX need this? */ -#endif - intel_region_map(intel->intelScreen, irb->region); - } - pb->ptr = irb->region->map; - - sps->surface.stride = irb->region->pitch; - sps->surface.cpp = irb->region->cpp; - sps->surface.ptr = irb->region->map; - - return pb->ptr; -} - - -static void -unmap_surface_buffer(struct pipe_buffer *pb) -{ - struct softpipe_surface *sps = (struct softpipe_surface *) pb; - struct intel_renderbuffer *irb = (struct intel_renderbuffer *) sps->surface.rb; - - if (irb->region) { - GET_CURRENT_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); - intel_region_unmap(intel->intelScreen, irb->region); - } - pb->ptr = NULL; - - sps->surface.stride = 0; - sps->surface.cpp = 0; - sps->surface.ptr = NULL; - - /*UNLOCK_HARDWARE(intel);*/ -} - - -struct pipe_surface * -intel_new_surface(GLuint intFormat) -{ - struct softpipe_surface *sps = CALLOC_STRUCT(softpipe_surface); - if (!sps) - return NULL; - - sps->surface.width = 0; /* set in intel_alloc_renderbuffer_storage() */ - sps->surface.height = 0; - - if (intFormat == GL_RGBA8) { - sps->surface.format = PIPE_FORMAT_U_A8_R8_G8_B8; - sps->read_quad_f_swz = read_quad_f_swz; - sps->write_quad_f_swz = write_quad_f_swz; - } - else if (intFormat == GL_RGB5) { - sps->surface.format = PIPE_FORMAT_U_R5_G6_B5; - - } - else if (intFormat == GL_DEPTH_COMPONENT16) { - sps->surface.format = PIPE_FORMAT_U_Z16; - - } - else if (intFormat == GL_DEPTH24_STENCIL8_EXT) { - sps->surface.format = PIPE_FORMAT_Z24_S8; - sps->read_quad_z = read_quad_z24; - sps->write_quad_z = write_quad_z24; - sps->read_quad_stencil = read_quad_stencil; - sps->write_quad_stencil = write_quad_stencil; - } - else { - /* TBD / unknown */ - - } - - sps->surface.buffer.map = map_surface_buffer; - sps->surface.buffer.unmap = unmap_surface_buffer; - - return &sps->surface; -} - diff --git a/src/mesa/drivers/dri/i915tex/intel_tex.c b/src/mesa/drivers/dri/i915tex/intel_tex.c deleted file mode 100644 index b08dee43bc..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tex.c +++ /dev/null @@ -1,192 +0,0 @@ -#include "texobj.h" -#include "intel_context.h" -#include "intel_mipmap_tree.h" -#include "intel_tex.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -static GLboolean -intelIsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) -{ -#if 0 - struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); - - return - intelObj->mt && - intelObj->mt->region && - intel_is_region_resident(intel, intelObj->mt->region); -#endif - return 1; -} - - - -static struct gl_texture_image * -intelNewTextureImage(GLcontext * ctx) -{ - DBG("%s\n", __FUNCTION__); - (void) ctx; - return (struct gl_texture_image *) CALLOC_STRUCT(intel_texture_image); -} - - -static struct gl_texture_object * -intelNewTextureObject(GLcontext * ctx, GLuint name, GLenum target) -{ - struct intel_texture_object *obj = CALLOC_STRUCT(intel_texture_object); - - DBG("%s\n", __FUNCTION__); - _mesa_initialize_texture_object(&obj->base, name, target); - - return &obj->base; -} - -static void -intelDeleteTextureObject(GLcontext *ctx, - struct gl_texture_object *texObj) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); - - if (intelObj->mt) - intel_miptree_release(intel, &intelObj->mt); - - _mesa_delete_texture_object(ctx, texObj); -} - - -static void -intelFreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_texture_image *intelImage = intel_texture_image(texImage); - - DBG("%s\n", __FUNCTION__); - - if (intelImage->mt) { - intel_miptree_release(intel, &intelImage->mt); - } - - if (texImage->Data) { - free(texImage->Data); - texImage->Data = NULL; - } -} - - -/* The system memcpy (at least on ubuntu 5.10) has problems copying - * to agp (writecombined) memory from a source which isn't 64-byte - * aligned - there is a 4x performance falloff. - * - * The x86 __memcpy is immune to this but is slightly slower - * (10%-ish) than the system memcpy. - * - * The sse_memcpy seems to have a slight cliff at 64/32 bytes, but - * isn't much faster than x86_memcpy for agp copies. - * - * TODO: switch dynamically. - */ -static void * -do_memcpy(void *dest, const void *src, size_t n) -{ - if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) { - return __memcpy(dest, src, n); - } - else - return memcpy(dest, src, n); -} - - -#if DO_DEBUG - -#ifndef __x86_64__ -static unsigned -fastrdtsc(void) -{ - unsigned eax; - __asm__ volatile ("\t" - "pushl %%ebx\n\t" - "cpuid\n\t" ".byte 0x0f, 0x31\n\t" - "popl %%ebx\n":"=a" (eax) - :"0"(0) - :"ecx", "edx", "cc"); - - return eax; -} -#else -static unsigned -fastrdtsc(void) -{ - unsigned eax; - __asm__ volatile ("\t" "cpuid\n\t" ".byte 0x0f, 0x31\n\t":"=a" (eax) - :"0"(0) - :"ecx", "edx", "ebx", "cc"); - - return eax; -} -#endif - -static unsigned -time_diff(unsigned t, unsigned t2) -{ - return ((t < t2) ? t2 - t : 0xFFFFFFFFU - (t - t2 - 1)); -} - - -static void * -timed_memcpy(void *dest, const void *src, size_t n) -{ - void *ret; - unsigned t1, t2; - double rate; - - if ((((unsigned) src) & 63) || (((unsigned) dest) & 63)) - _mesa_printf("Warning - non-aligned texture copy!\n"); - - t1 = fastrdtsc(); - ret = do_memcpy(dest, src, n); - t2 = fastrdtsc(); - - rate = time_diff(t1, t2); - rate /= (double) n; - _mesa_printf("timed_memcpy: %u %u --> %f clocks/byte\n", t1, t2, rate); - return ret; -} -#endif /* DO_DEBUG */ - - -void -intelInitTextureFuncs(struct dd_function_table *functions) -{ - functions->ChooseTextureFormat = intelChooseTextureFormat; - functions->TexImage1D = intelTexImage1D; - functions->TexImage2D = intelTexImage2D; - functions->TexImage3D = intelTexImage3D; - functions->TexSubImage1D = intelTexSubImage1D; - functions->TexSubImage2D = intelTexSubImage2D; - functions->TexSubImage3D = intelTexSubImage3D; - functions->CopyTexImage1D = intelCopyTexImage1D; - functions->CopyTexImage2D = intelCopyTexImage2D; - functions->CopyTexSubImage1D = intelCopyTexSubImage1D; - functions->CopyTexSubImage2D = intelCopyTexSubImage2D; - functions->GetTexImage = intelGetTexImage; - - /* compressed texture functions */ - functions->CompressedTexImage2D = intelCompressedTexImage2D; - functions->GetCompressedTexImage = intelGetCompressedTexImage; - - functions->NewTextureObject = intelNewTextureObject; - functions->NewTextureImage = intelNewTextureImage; - functions->DeleteTexture = intelDeleteTextureObject; - functions->FreeTexImageData = intelFreeTextureImageData; - functions->UpdateTexturePalette = 0; - functions->IsTextureResident = intelIsTextureResident; - -#if DO_DEBUG - if (INTEL_DEBUG & DEBUG_BUFMGR) - functions->TextureMemCpy = timed_memcpy; - else -#endif - functions->TextureMemCpy = do_memcpy; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_tex.h b/src/mesa/drivers/dri/i915tex/intel_tex.h deleted file mode 100644 index b77d7a1d8a..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tex.h +++ /dev/null @@ -1,151 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTELTEX_INC -#define INTELTEX_INC - -#include "mtypes.h" -#include "intel_context.h" -#include "texmem.h" - - -void intelInitTextureFuncs(struct dd_function_table *functions); - -const struct gl_texture_format *intelChooseTextureFormat(GLcontext * ctx, - GLint internalFormat, - GLenum format, - GLenum type); - - -void intelTexImage3D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage3D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexImage2D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage2D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexImage1D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelTexSubImage1D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLint border); - -void intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint border); - -void intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint x, GLint y, GLsizei width); - -void intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLint x, GLint y, GLsizei width, GLsizei height); - -void intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage); - -void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ); - -void intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, - GLvoid *pixels, - const struct gl_texture_object *texObj, - const struct gl_texture_image *texImage); - -void intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, - unsigned long long offset, GLint depth, GLuint pitch); - -GLuint intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit); - -void intel_tex_map_images(struct intel_context *intel, - struct intel_texture_object *intelObj); - -void intel_tex_unmap_images(struct intel_context *intel, - struct intel_texture_object *intelObj); - -int intel_compressed_num_bytes(GLuint mesaFormat); - -#endif diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_copy.c b/src/mesa/drivers/dri/i915tex/intel_tex_copy.c deleted file mode 100644 index 8d7f8f9d91..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tex_copy.c +++ /dev/null @@ -1,293 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "mtypes.h" -#include "enums.h" -#include "image.h" -#include "teximage.h" -#include "swrast/swrast.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_buffers.h" -#include "intel_mipmap_tree.h" -#include "intel_regions.h" -#include "intel_fbo.h" -#include "intel_tex.h" -#include "intel_blit.h" -#include "intel_pixel.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -/** - * Get the intel_region which is the source for any glCopyTex[Sub]Image call. - * - * Do the best we can using the blitter. A future project is to use - * the texture engine and fragment programs for these copies. - */ -static const struct intel_region * -get_teximage_source(struct intel_context *intel, GLenum internalFormat) -{ - struct intel_renderbuffer *irb; - - DBG("%s %s\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(internalFormat)); - - switch (internalFormat) { - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16_ARB: - irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); - if (irb && irb->region && irb->region->cpp == 2) - return irb->region; - return NULL; - case GL_DEPTH24_STENCIL8_EXT: - case GL_DEPTH_STENCIL_EXT: - irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH); - if (irb && irb->region && irb->region->cpp == 4) - return irb->region; - return NULL; - case GL_RGBA: - case GL_RGBA8: - return intel_readbuf_region(intel); - case GL_RGB: - if (intel->intelScreen->cpp == 2) - return intel_readbuf_region(intel); - return NULL; - default: - return NULL; - } -} - - -static GLboolean -do_copy_texsubimage(struct intel_context *intel, - struct intel_texture_image *intelImage, - GLenum internalFormat, - GLint dstx, GLint dsty, - GLint x, GLint y, GLsizei width, GLsizei height) -{ - GLcontext *ctx = &intel->ctx; - const struct intel_region *src = - get_teximage_source(intel, internalFormat); - - if (!intelImage->mt || !src) { - DBG("%s fail %p %p\n", __FUNCTION__, intelImage->mt, src); - return GL_FALSE; - } - - intelFlush(ctx); - /* XXX still need the lock ? */ - LOCK_HARDWARE(intel); - { - GLuint image_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); - const GLint orig_x = x; - const GLint orig_y = y; - const struct gl_framebuffer *fb = ctx->DrawBuffer; - - if (_mesa_clip_to_region(fb->_Xmin, fb->_Ymin, fb->_Xmax, fb->_Ymax, - &x, &y, &width, &height)) { - /* Update dst for clipped src. Need to also clip the source rect. - */ - dstx += x - orig_x; - dsty += y - orig_y; - - if (!(ctx->ReadBuffer->Name == 0)) { - /* XXX this looks bogus ? */ - /* FBO: invert Y */ - y = ctx->ReadBuffer->Height - y - 1; - } - - /* A bit of fiddling to get the blitter to work with -ve - * pitches. But we get a nice inverted blit this way, so it's - * worth it: - */ - intelEmitCopyBlit(intel, - intelImage->mt->cpp, - -src->pitch, - src->buffer, - src->height * src->pitch * src->cpp, - intelImage->mt->pitch, - intelImage->mt->region->buffer, - image_offset, - x, y + height, dstx, dsty, width, height, - GL_COPY); /* ? */ - - intel_batchbuffer_flush(intel->batch); - } - } - - - UNLOCK_HARDWARE(intel); - -#if 0 - /* GL_SGIS_generate_mipmap -- this can be accelerated now. - * XXX Add a ctx->Driver.GenerateMipmaps() function? - */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); - } -#endif - - return GL_TRUE; -} - - - - - -void -intelCopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLint border) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - - if (border) - goto fail; - - /* Setup or redefine the texture object, mipmap tree and texture - * image. Don't populate yet. - */ - ctx->Driver.TexImage1D(ctx, target, level, internalFormat, - width, border, - GL_RGBA, CHAN_TYPE, NULL, - &ctx->DefaultPacking, texObj, texImage); - - if (!do_copy_texsubimage(intel_context(ctx), - intel_texture_image(texImage), - internalFormat, 0, 0, x, y, width, 1)) - goto fail; - - return; - - fail: - _swrast_copy_teximage1d(ctx, target, level, internalFormat, x, y, - width, border); -} - -void -intelCopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, - GLenum internalFormat, - GLint x, GLint y, GLsizei width, GLsizei height, - GLint border) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - - if (border) - goto fail; - - /* Setup or redefine the texture object, mipmap tree and texture - * image. Don't populate yet. - */ - ctx->Driver.TexImage2D(ctx, target, level, internalFormat, - width, height, border, - GL_RGBA, CHAN_TYPE, NULL, - &ctx->DefaultPacking, texObj, texImage); - - - if (!do_copy_texsubimage(intel_context(ctx), - intel_texture_image(texImage), - internalFormat, 0, 0, x, y, width, height)) - goto fail; - - return; - - fail: - _swrast_copy_teximage2d(ctx, target, level, internalFormat, x, y, - width, height, border); -} - - -void -intelCopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint x, GLint y, GLsizei width) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - GLenum internalFormat = texImage->InternalFormat; - - /* XXX need to check as in above function? */ - - /* Need to check texture is compatible with source format. - */ - - if (!do_copy_texsubimage(intel_context(ctx), - intel_texture_image(texImage), - internalFormat, xoffset, 0, x, y, width, 1)) { - _swrast_copy_texsubimage1d(ctx, target, level, xoffset, x, y, width); - } -} - - - -void -intelCopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLint x, GLint y, GLsizei width, GLsizei height) -{ - struct gl_texture_unit *texUnit = - &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - struct gl_texture_object *texObj = - _mesa_select_tex_object(ctx, texUnit, target); - struct gl_texture_image *texImage = - _mesa_select_tex_image(ctx, texObj, target, level); - GLenum internalFormat = texImage->InternalFormat; - - - /* Need to check texture is compatible with source format. - */ - - if (!do_copy_texsubimage(intel_context(ctx), - intel_texture_image(texImage), - internalFormat, - xoffset, yoffset, x, y, width, height)) { - - DBG("%s - fallback to swrast\n", __FUNCTION__); - - _swrast_copy_texsubimage2d(ctx, target, level, - xoffset, yoffset, x, y, width, height); - } -} diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_format.c b/src/mesa/drivers/dri/i915tex/intel_tex_format.c deleted file mode 100644 index 6e058dff69..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tex_format.c +++ /dev/null @@ -1,172 +0,0 @@ -#include "intel_context.h" -#include "intel_tex.h" -#include "texformat.h" -#include "enums.h" - -/* It works out that this function is fine for all the supported - * hardware. However, there is still a need to map the formats onto - * hardware descriptors. - */ -/* Note that the i915 can actually support many more formats than - * these if we take the step of simply swizzling the colors - * immediately after sampling... - */ -const struct gl_texture_format * -intelChooseTextureFormat(GLcontext * ctx, GLint internalFormat, - GLenum format, GLenum type) -{ - struct intel_context *intel = intel_context(ctx); - const GLboolean do32bpt = (intel->intelScreen->cpp == 4); - - switch (internalFormat) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - if (format == GL_BGRA) { - if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - return &_mesa_texformat_argb8888; - } - else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - return &_mesa_texformat_argb4444; - } - else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - return &_mesa_texformat_argb1555; - } - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return &_mesa_texformat_rgb565; - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case GL_RGBA4: - case GL_RGBA2: - return &_mesa_texformat_argb4444; - - case GL_RGB5_A1: - return &_mesa_texformat_argb1555; - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: - return &_mesa_texformat_argb8888; - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: - return &_mesa_texformat_rgb565; - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: - return &_mesa_texformat_a8; - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: - return &_mesa_texformat_l8; - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: - return &_mesa_texformat_i8; - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; - - case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; - - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - return &_mesa_texformat_z16; - - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_z24_s8; - - default: - fprintf(stderr, "unexpected texture format %s in %s\n", - _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); - return NULL; - } - - return NULL; /* never get here */ -} - -int intel_compressed_num_bytes(GLuint mesaFormat) -{ - int bytes = 0; - switch(mesaFormat) { - - case MESA_FORMAT_RGB_FXT1: - case MESA_FORMAT_RGBA_FXT1: - case MESA_FORMAT_RGB_DXT1: - case MESA_FORMAT_RGBA_DXT1: - bytes = 2; - break; - - case MESA_FORMAT_RGBA_DXT3: - case MESA_FORMAT_RGBA_DXT5: - bytes = 4; - default: - break; - } - - return bytes; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_image.c b/src/mesa/drivers/dri/i915tex/intel_tex_image.c deleted file mode 100644 index f790b1e6f7..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tex_image.c +++ /dev/null @@ -1,691 +0,0 @@ - -#include -#include - -#include "glheader.h" -#include "macros.h" -#include "mtypes.h" -#include "enums.h" -#include "colortab.h" -#include "convolve.h" -#include "context.h" -#include "simple_list.h" -#include "texcompress.h" -#include "texformat.h" -#include "texobj.h" -#include "texstore.h" - -#include "intel_context.h" -#include "intel_mipmap_tree.h" -#include "intel_buffer_objects.h" -#include "intel_batchbuffer.h" -#include "intel_tex.h" -#include "intel_ioctl.h" -#include "intel_blit.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -/* Functions to store texture images. Where possible, mipmap_tree's - * will be created or further instantiated with image data, otherwise - * images will be stored in malloc'd memory. A validation step is - * required to pull those images into a mipmap tree, or otherwise - * decide a fallback is required. - */ - - -static int -logbase2(int n) -{ - GLint i = 1; - GLint log2 = 0; - - while (n > i) { - i *= 2; - log2++; - } - - return log2; -} - - -/* Otherwise, store it in memory if (Border != 0) or (any dimension == - * 1). - * - * Otherwise, if max_level >= level >= min_level, create tree with - * space for textures from min_level down to max_level. - * - * Otherwise, create tree with space for textures from (level - * 0)..(1x1). Consider pruning this tree at a validation if the - * saving is worth it. - */ -static void -guess_and_alloc_mipmap_tree(struct intel_context *intel, - struct intel_texture_object *intelObj, - struct intel_texture_image *intelImage) -{ - GLuint firstLevel; - GLuint lastLevel; - GLuint width = intelImage->base.Width; - GLuint height = intelImage->base.Height; - GLuint depth = intelImage->base.Depth; - GLuint l2width, l2height, l2depth; - GLuint i, comp_byte = 0; - - DBG("%s\n", __FUNCTION__); - - if (intelImage->base.Border) - return; - - if (intelImage->level > intelObj->base.BaseLevel && - (intelImage->base.Width == 1 || - (intelObj->base.Target != GL_TEXTURE_1D && - intelImage->base.Height == 1) || - (intelObj->base.Target == GL_TEXTURE_3D && - intelImage->base.Depth == 1))) - return; - - /* If this image disrespects BaseLevel, allocate from level zero. - * Usually BaseLevel == 0, so it's unlikely to happen. - */ - if (intelImage->level < intelObj->base.BaseLevel) - firstLevel = 0; - else - firstLevel = intelObj->base.BaseLevel; - - - /* Figure out image dimensions at start level. - */ - for (i = intelImage->level; i > firstLevel; i--) { - width <<= 1; - if (height != 1) - height <<= 1; - if (depth != 1) - depth <<= 1; - } - - /* Guess a reasonable value for lastLevel. This is probably going - * to be wrong fairly often and might mean that we have to look at - * resizable buffers, or require that buffers implement lazy - * pagetable arrangements. - */ - if ((intelObj->base.MinFilter == GL_NEAREST || - intelObj->base.MinFilter == GL_LINEAR) && - intelImage->level == firstLevel) { - lastLevel = firstLevel; - } - else { - l2width = logbase2(width); - l2height = logbase2(height); - l2depth = logbase2(depth); - lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); - } - - assert(!intelObj->mt); - if (intelImage->base.IsCompressed) - comp_byte = intel_compressed_num_bytes(intelImage->base.TexFormat->MesaFormat); - intelObj->mt = intel_miptree_create(intel, - intelObj->base.Target, - intelImage->base.InternalFormat, - firstLevel, - lastLevel, - width, - height, - depth, - intelImage->base.TexFormat->TexelBytes, - comp_byte); - - DBG("%s - success\n", __FUNCTION__); -} - - - - -static GLuint -target_to_face(GLenum target) -{ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - return ((GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X); - default: - return 0; - } -} - -/* There are actually quite a few combinations this will work for, - * more than what I've listed here. - */ -static GLboolean -check_pbo_format(GLint internalFormat, - GLenum format, GLenum type, - const struct gl_texture_format *mesa_format) -{ - switch (internalFormat) { - case 4: - case GL_RGBA: - return (format == GL_BGRA && - (type == GL_UNSIGNED_BYTE || - type == GL_UNSIGNED_INT_8_8_8_8_REV) && - mesa_format == &_mesa_texformat_argb8888); - case 3: - case GL_RGB: - return (format == GL_RGB && - type == GL_UNSIGNED_SHORT_5_6_5 && - mesa_format == &_mesa_texformat_rgb565); - case GL_YCBCR_MESA: - return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); - default: - return GL_FALSE; - } -} - - -/* XXX: Do this for TexSubImage also: - */ -static GLboolean -try_pbo_upload(struct intel_context *intel, - struct intel_texture_image *intelImage, - const struct gl_pixelstore_attrib *unpack, - GLint internalFormat, - GLint width, GLint height, - GLenum format, GLenum type, const void *pixels) -{ - struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); - GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; - - if (!pbo || - intel->ctx._ImageTransferState || - unpack->SkipPixels || unpack->SkipRows) { - _mesa_printf("%s: failure 1\n", __FUNCTION__); - return GL_FALSE; - } - - src_offset = (GLuint) pixels; - - if (unpack->RowLength > 0) - src_stride = unpack->RowLength; - else - src_stride = width; - - dst_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); - - dst_stride = intelImage->mt->pitch; - - intelFlush(&intel->ctx); - LOCK_HARDWARE(intel); - { - struct _DriBufferObject *src_buffer = - intel_bufferobj_buffer(intel, pbo, INTEL_READ); - struct _DriBufferObject *dst_buffer = - intel_region_buffer(intel->intelScreen, intelImage->mt->region, - INTEL_WRITE_FULL); - - - intelEmitCopyBlit(intel, - intelImage->mt->cpp, - src_stride, src_buffer, src_offset, - dst_stride, dst_buffer, dst_offset, - 0, 0, 0, 0, width, height, - GL_COPY); - - intel_batchbuffer_flush(intel->batch); - } - UNLOCK_HARDWARE(intel); - - return GL_TRUE; -} - - - -static GLboolean -try_pbo_zcopy(struct intel_context *intel, - struct intel_texture_image *intelImage, - const struct gl_pixelstore_attrib *unpack, - GLint internalFormat, - GLint width, GLint height, - GLenum format, GLenum type, const void *pixels) -{ - struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); - GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; - - if (!pbo || - intel->ctx._ImageTransferState || - unpack->SkipPixels || unpack->SkipRows) { - _mesa_printf("%s: failure 1\n", __FUNCTION__); - return GL_FALSE; - } - - src_offset = (GLuint) pixels; - - if (unpack->RowLength > 0) - src_stride = unpack->RowLength; - else - src_stride = width; - - dst_offset = intel_miptree_image_offset(intelImage->mt, - intelImage->face, - intelImage->level); - - dst_stride = intelImage->mt->pitch; - - if (src_stride != dst_stride || dst_offset != 0 || src_offset != 0) { - _mesa_printf("%s: failure 2\n", __FUNCTION__); - return GL_FALSE; - } - - intel_region_attach_pbo(intel->intelScreen, intelImage->mt->region, pbo); - - return GL_TRUE; -} - - - - - - -static void -intelTexImage(GLcontext * ctx, - GLint dims, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage, GLsizei imageSize, int compressed) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_texture_object *intelObj = intel_texture_object(texObj); - struct intel_texture_image *intelImage = intel_texture_image(texImage); - GLint postConvWidth = width; - GLint postConvHeight = height; - GLint texelBytes, sizeInBytes; - GLuint dstRowStride; - - - DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); - - intelFlush(ctx); - - intelImage->face = target_to_face(target); - intelImage->level = level; - - if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { - _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth, - &postConvHeight); - } - - /* choose the texture format */ - texImage->TexFormat = intelChooseTextureFormat(ctx, internalFormat, - format, type); - - _mesa_set_fetch_functions(texImage, dims); - - if (texImage->TexFormat->TexelBytes == 0) { - /* must be a compressed format */ - texelBytes = 0; - texImage->IsCompressed = GL_TRUE; - texImage->CompressedSize = - ctx->Driver.CompressedTextureSize(ctx, texImage->Width, - texImage->Height, texImage->Depth, - texImage->TexFormat->MesaFormat); - } else { - texelBytes = texImage->TexFormat->TexelBytes; - - /* Minimum pitch of 32 bytes */ - if (postConvWidth * texelBytes < 32) { - postConvWidth = 32 / texelBytes; - texImage->RowStride = postConvWidth; - } - - assert(texImage->RowStride == postConvWidth); - } - - /* Release the reference to a potentially orphaned buffer. - * Release any old malloced memory. - */ - if (intelImage->mt) { - intel_miptree_release(intel, &intelImage->mt); - assert(!texImage->Data); - } - else if (texImage->Data) { - _mesa_align_free(texImage->Data); - } - - /* If this is the only texture image in the tree, could call - * bmBufferData with NULL data to free the old block and avoid - * waiting on any outstanding fences. - */ - if (intelObj->mt && - intelObj->mt->first_level == level && - intelObj->mt->last_level == level && - intelObj->mt->target != GL_TEXTURE_CUBE_MAP_ARB && - !intel_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { - - DBG("release it\n"); - intel_miptree_release(intel, &intelObj->mt); - assert(!intelObj->mt); - } - - if (!intelObj->mt) { - guess_and_alloc_mipmap_tree(intel, intelObj, intelImage); - if (!intelObj->mt) { - DBG("guess_and_alloc_mipmap_tree: failed\n"); - } - } - - assert(!intelImage->mt); - - if (intelObj->mt && - intel_miptree_match_image(intelObj->mt, &intelImage->base, - intelImage->face, intelImage->level)) { - - intel_miptree_reference(&intelImage->mt, intelObj->mt); - assert(intelImage->mt); - } - - if (!intelImage->mt) - DBG("XXX: Image did not fit into tree - storing in local memory!\n"); - - /* PBO fastpaths: - */ - if (dims <= 2 && - intelImage->mt && - intel_buffer_object(unpack->BufferObj) && - check_pbo_format(internalFormat, format, - type, intelImage->base.TexFormat)) { - - DBG("trying pbo upload\n"); - - /* Attempt to texture directly from PBO data (zero copy upload). - * - * Currently disable as it can lead to worse as well as better - * performance (in particular when intel_region_cow() is - * required). - */ - if (intelObj->mt == intelImage->mt && - intelObj->mt->first_level == level && - intelObj->mt->last_level == level) { - - if (try_pbo_zcopy(intel, intelImage, unpack, - internalFormat, - width, height, format, type, pixels)) { - - DBG("pbo zcopy upload succeeded\n"); - return; - } - } - - - /* Otherwise, attempt to use the blitter for PBO image uploads. - */ - if (try_pbo_upload(intel, intelImage, unpack, - internalFormat, - width, height, format, type, pixels)) { - DBG("pbo upload succeeded\n"); - return; - } - - DBG("pbo upload failed\n"); - } - - - - /* intelCopyTexImage calls this function with pixels == NULL, with - * the expectation that the mipmap tree will be set up but nothing - * more will be done. This is where those calls return: - */ - if (compressed) { - pixels = _mesa_validate_pbo_compressed_teximage(ctx, imageSize, pixels, - unpack, - "glCompressedTexImage"); - } else { - pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, - format, type, - pixels, unpack, "glTexImage"); - } - if (!pixels) - return; - - - if (intelImage->mt) - intel_region_idle(intel->intelScreen, intelImage->mt->region); - - LOCK_HARDWARE(intel); - - if (intelImage->mt) { - texImage->Data = intel_miptree_image_map(intel, - intelImage->mt, - intelImage->face, - intelImage->level, - &dstRowStride, - intelImage->base.ImageOffsets); - } - else { - /* Allocate regular memory and store the image there temporarily. */ - if (texImage->IsCompressed) { - sizeInBytes = texImage->CompressedSize; - dstRowStride = - _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width); - assert(dims != 3); - } - else { - dstRowStride = postConvWidth * texelBytes; - sizeInBytes = depth * dstRowStride * postConvHeight; - } - - texImage->Data = malloc(sizeInBytes); - } - - DBG("Upload image %dx%dx%d row_len %x " - "pitch %x\n", - width, height, depth, width * texelBytes, dstRowStride); - - /* Copy data. Would like to know when it's ok for us to eg. use - * the blitter to copy. Or, use the hardware to do the format - * conversion and copy: - */ - if (compressed) { - memcpy(texImage->Data, pixels, imageSize); - } else if (!texImage->TexFormat->StoreImage(ctx, dims, - texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, 0, 0, 0, /* dstX/Y/Zoffset */ - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, unpack)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); - } - - _mesa_unmap_teximage_pbo(ctx, unpack); - - if (intelImage->mt) { - intel_miptree_image_unmap(intel, intelImage->mt); - texImage->Data = NULL; - } - - UNLOCK_HARDWARE(intel); - -#if 0 - /* GL_SGIS_generate_mipmap -- this can be accelerated now. - */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); - } -#endif -} - -void -intelTexImage3D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexImage(ctx, 3, target, level, - internalFormat, width, height, depth, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); -} - - -void -intelTexImage2D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexImage(ctx, 2, target, level, - internalFormat, width, height, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); -} - -void -intelTexImage1D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexImage(ctx, 1, target, level, - internalFormat, width, 1, 1, border, - format, type, pixels, unpack, texObj, texImage, 0, 0); -} - -void intelCompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) -{ - intelTexImage(ctx, 2, target, level, - internalFormat, width, height, 1, border, - 0, 0, data, &ctx->Unpack, texObj, texImage, imageSize, 1); -} - -/** - * Need to map texture image into memory before copying image data, - * then unmap it. - */ -static void -intel_get_tex_image(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage, int compressed) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_texture_image *intelImage = intel_texture_image(texImage); - - /* Map */ - if (intelImage->mt) { - /* Image is stored in hardware format in a buffer managed by the - * kernel. Need to explicitly map and unmap it. - */ - intelImage->base.Data = - intel_miptree_image_map(intel, - intelImage->mt, - intelImage->face, - intelImage->level, - &intelImage->base.RowStride, - intelImage->base.ImageOffsets); - intelImage->base.RowStride /= intelImage->mt->cpp; - } - else { - /* Otherwise, the image should actually be stored in - * intelImage->base.Data. This is pretty confusing for - * everybody, I'd much prefer to separate the two functions of - * texImage->Data - storage for texture images in main memory - * and access (ie mappings) of images. In other words, we'd - * create a new texImage->Map field and leave Data simply for - * storage. - */ - assert(intelImage->base.Data); - } - - - if (compressed) { - _mesa_get_compressed_teximage(ctx, target, level, pixels, - texObj, texImage); - } else { - _mesa_get_teximage(ctx, target, level, format, type, pixels, - texObj, texImage); - } - - - /* Unmap */ - if (intelImage->mt) { - intel_miptree_image_unmap(intel, intelImage->mt); - intelImage->base.Data = NULL; - } -} - -void -intelGetTexImage(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intel_get_tex_image(ctx, target, level, format, type, pixels, - texObj, texImage, 0); - - -} - -void -intelGetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, - GLvoid *pixels, - const struct gl_texture_object *texObj, - const struct gl_texture_image *texImage) -{ - intel_get_tex_image(ctx, target, level, 0, 0, pixels, - texObj, texImage, 1); - -} - -void -intelSetTexOffset(__DRIcontext *pDRICtx, GLint texname, - unsigned long long offset, GLint depth, GLuint pitch) -{ - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct intel_texture_object *intelObj = intel_texture_object(tObj); - - if (!intelObj) - return; - - if (intelObj->mt) - intel_miptree_release(intel, &intelObj->mt); - - intelObj->imageOverride = GL_TRUE; - intelObj->depthOverride = depth; - intelObj->pitchOverride = pitch; - - if (offset) - intelObj->textureOffset = offset; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_layout.c b/src/mesa/drivers/dri/i915tex/intel_tex_layout.c deleted file mode 120000 index fe61b44194..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tex_layout.c +++ /dev/null @@ -1 +0,0 @@ -../intel/intel_tex_layout.c \ No newline at end of file diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_subimage.c b/src/mesa/drivers/dri/i915tex/intel_tex_subimage.c deleted file mode 100644 index 3935787806..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tex_subimage.c +++ /dev/null @@ -1,182 +0,0 @@ - -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "mtypes.h" -#include "texobj.h" -#include "texstore.h" -#include "enums.h" - -#include "intel_context.h" -#include "intel_tex.h" -#include "intel_mipmap_tree.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -static void -intelTexSubimage(GLcontext * ctx, - GLint dims, - GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLint width, GLint height, GLint depth, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - struct intel_context *intel = intel_context(ctx); - struct intel_texture_image *intelImage = intel_texture_image(texImage); - GLuint dstRowStride; - - DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(target), - level, xoffset, yoffset, width, height); - - intelFlush(ctx); - - pixels = - _mesa_validate_pbo_teximage(ctx, dims, width, height, depth, format, - type, pixels, packing, "glTexSubImage2D"); - if (!pixels) - return; - - if (intelImage->mt) - intel_region_idle(intel->intelScreen, intelImage->mt->region); - - LOCK_HARDWARE(intel); - - /* Map buffer if necessary. Need to lock to prevent other contexts - * from uploading the buffer under us. - */ - if (intelImage->mt) - texImage->Data = intel_miptree_image_map(intel, - intelImage->mt, - intelImage->face, - intelImage->level, - &dstRowStride, - texImage->ImageOffsets); - - assert(dstRowStride); - - if (!texImage->TexFormat->StoreImage(ctx, dims, texImage->_BaseFormat, - texImage->TexFormat, - texImage->Data, - xoffset, yoffset, zoffset, - dstRowStride, - texImage->ImageOffsets, - width, height, depth, - format, type, pixels, packing)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "intelTexSubImage"); - } - -#if 0 - /* GL_SGIS_generate_mipmap */ - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); - } -#endif - - _mesa_unmap_teximage_pbo(ctx, packing); - - if (intelImage->mt) { - intel_miptree_image_unmap(intel, intelImage->mt); - texImage->Data = NULL; - } - - UNLOCK_HARDWARE(intel); -} - - - - - -void -intelTexSubImage3D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - - intelTexSubimage(ctx, 3, - target, level, - xoffset, yoffset, zoffset, - width, height, depth, - format, type, pixels, packing, texObj, texImage); - -} - - - -void -intelTexSubImage2D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - - intelTexSubimage(ctx, 2, - target, level, - xoffset, yoffset, 0, - width, height, 1, - format, type, pixels, packing, texObj, texImage); - -} - - -void -intelTexSubImage1D(GLcontext * ctx, - GLenum target, - GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLenum type, - const GLvoid * pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) -{ - intelTexSubimage(ctx, 1, - target, level, - xoffset, 0, 0, - width, 1, 1, - format, type, pixels, packing, texObj, texImage); - -} diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_validate.c b/src/mesa/drivers/dri/i915tex/intel_tex_validate.c deleted file mode 100644 index af18c26d55..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tex_validate.c +++ /dev/null @@ -1,272 +0,0 @@ -#include "mtypes.h" -#include "macros.h" - -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_mipmap_tree.h" -#include "intel_tex.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -/** - * Compute which mipmap levels that really need to be sent to the hardware. - * This depends on the base image size, GL_TEXTURE_MIN_LOD, - * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL. - */ -static void -intel_calculate_first_last_level(struct intel_texture_object *intelObj) -{ - struct gl_texture_object *tObj = &intelObj->base; - const struct gl_texture_image *const baseImage = - tObj->Image[0][tObj->BaseLevel]; - - /* These must be signed values. MinLod and MaxLod can be negative numbers, - * and having firstLevel and lastLevel as signed prevents the need for - * extra sign checks. - */ - int firstLevel; - int lastLevel; - - /* Yes, this looks overly complicated, but it's all needed. - */ - switch (tObj->Target) { - case GL_TEXTURE_1D: - case GL_TEXTURE_2D: - case GL_TEXTURE_3D: - case GL_TEXTURE_CUBE_MAP: - if (tObj->MinFilter == GL_NEAREST || tObj->MinFilter == GL_LINEAR) { - /* GL_NEAREST and GL_LINEAR only care about GL_TEXTURE_BASE_LEVEL. - */ - firstLevel = lastLevel = tObj->BaseLevel; - } - else { - firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); - firstLevel = MAX2(firstLevel, tObj->BaseLevel); - lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); - lastLevel = MAX2(lastLevel, tObj->BaseLevel); - lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); - lastLevel = MIN2(lastLevel, tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ - } - break; - case GL_TEXTURE_RECTANGLE_NV: - case GL_TEXTURE_4D_SGIS: - firstLevel = lastLevel = 0; - break; - default: - return; - } - - /* save these values */ - intelObj->firstLevel = firstLevel; - intelObj->lastLevel = lastLevel; -} - -static void -copy_image_data_to_tree(struct intel_context *intel, - struct intel_texture_object *intelObj, - struct intel_texture_image *intelImage) -{ - if (intelImage->mt) { - /* Copy potentially with the blitter: - */ - intel_miptree_image_copy(intel, - intelObj->mt, - intelImage->face, - intelImage->level, intelImage->mt); - - intel_miptree_release(intel, &intelImage->mt); - } - else { - assert(intelImage->base.Data != NULL); - - /* More straightforward upload. - */ - intel_miptree_image_data(intel, - intelObj->mt, - intelImage->face, - intelImage->level, - intelImage->base.Data, - intelImage->base.RowStride, - intelImage->base.RowStride * - intelImage->base.Height); - _mesa_align_free(intelImage->base.Data); - intelImage->base.Data = NULL; - } - - intel_miptree_reference(&intelImage->mt, intelObj->mt); -} - - -/* - */ -GLuint -intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit) -{ - struct gl_texture_object *tObj = intel->ctx.Texture.Unit[unit]._Current; - struct intel_texture_object *intelObj = intel_texture_object(tObj); - int comp_byte = 0; - int cpp; - - GLuint face, i; - GLuint nr_faces = 0; - struct intel_texture_image *firstImage; - - GLboolean need_flush = GL_FALSE; - - /* We know/require this is true by now: - */ - assert(intelObj->base._Complete); - - /* What levels must the tree include at a minimum? - */ - intel_calculate_first_last_level(intelObj); - firstImage = - intel_texture_image(intelObj->base.Image[0][intelObj->firstLevel]); - - /* Fallback case: - */ - if (firstImage->base.Border) { - if (intelObj->mt) { - intel_miptree_release(intel, &intelObj->mt); - } - return GL_FALSE; - } - - - /* If both firstImage and intelObj have a tree which can contain - * all active images, favour firstImage. Note that because of the - * completeness requirement, we know that the image dimensions - * will match. - */ - if (firstImage->mt && - firstImage->mt != intelObj->mt && - firstImage->mt->first_level <= intelObj->firstLevel && - firstImage->mt->last_level >= intelObj->lastLevel) { - - if (intelObj->mt) - intel_miptree_release(intel, &intelObj->mt); - - intel_miptree_reference(&intelObj->mt, firstImage->mt); - } - - if (firstImage->base.IsCompressed) { - comp_byte = intel_compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); - cpp = comp_byte; - } - else cpp = firstImage->base.TexFormat->TexelBytes; - - /* Check tree can hold all active levels. Check tree matches - * target, imageFormat, etc. - * - * XXX: For some layouts (eg i945?), the test might have to be - * first_level == firstLevel, as the tree isn't valid except at the - * original start level. Hope to get around this by - * programming minLod, maxLod, baseLevel into the hardware and - * leaving the tree alone. - */ - if (intelObj->mt && - (intelObj->mt->target != intelObj->base.Target || - intelObj->mt->internal_format != firstImage->base.InternalFormat || - intelObj->mt->first_level != intelObj->firstLevel || - intelObj->mt->last_level != intelObj->lastLevel || - intelObj->mt->width0 != firstImage->base.Width || - intelObj->mt->height0 != firstImage->base.Height || - intelObj->mt->depth0 != firstImage->base.Depth || - intelObj->mt->cpp != cpp || - intelObj->mt->compressed != firstImage->base.IsCompressed)) { - intel_miptree_release(intel, &intelObj->mt); - } - - - /* May need to create a new tree: - */ - if (!intelObj->mt) { - intelObj->mt = intel_miptree_create(intel, - intelObj->base.Target, - firstImage->base.InternalFormat, - intelObj->firstLevel, - intelObj->lastLevel, - firstImage->base.Width, - firstImage->base.Height, - firstImage->base.Depth, - cpp, - comp_byte); - } - - /* Pull in any images not in the object's tree: - */ - nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct intel_texture_image *intelImage = - intel_texture_image(intelObj->base.Image[face][i]); - - /* Need to import images in main memory or held in other trees. - */ - if (intelObj->mt != intelImage->mt) { - copy_image_data_to_tree(intel, intelObj, intelImage); - need_flush = GL_TRUE; - } - } - } - - if (need_flush) - intel_batchbuffer_flush(intel->batch); - - return GL_TRUE; -} - - - -void -intel_tex_map_images(struct intel_context *intel, - struct intel_texture_object *intelObj) -{ - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - GLuint face, i; - - DBG("%s\n", __FUNCTION__); - - for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct intel_texture_image *intelImage = - intel_texture_image(intelObj->base.Image[face][i]); - - if (intelImage->mt) { - intelImage->base.Data = - intel_miptree_image_map(intel, - intelImage->mt, - intelImage->face, - intelImage->level, - &intelImage->base.RowStride, - intelImage->base.ImageOffsets); - /* convert stride to texels, not bytes */ - intelImage->base.RowStride /= intelImage->mt->cpp; -/* intelImage->base.ImageStride /= intelImage->mt->cpp; */ - } - } - } -} - - - -void -intel_tex_unmap_images(struct intel_context *intel, - struct intel_texture_object *intelObj) -{ - GLuint nr_faces = (intelObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; - GLuint face, i; - - for (face = 0; face < nr_faces; face++) { - for (i = intelObj->firstLevel; i <= intelObj->lastLevel; i++) { - struct intel_texture_image *intelImage = - intel_texture_image(intelObj->base.Image[face][i]); - - if (intelImage->mt) { - intel_miptree_image_unmap(intel, intelImage->mt); - intelImage->base.Data = NULL; - } - } - } -} diff --git a/src/mesa/drivers/dri/i915tex/intel_tris.c b/src/mesa/drivers/dri/i915tex/intel_tris.c deleted file mode 100644 index 48dae9d720..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tris.c +++ /dev/null @@ -1,1164 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "enums.h" -#include "texobj.h" -#include "state.h" -#include "dd.h" - -#include "swrast/swrast.h" -#include "swrast_setup/swrast_setup.h" -#include "tnl/t_context.h" -#include "tnl/t_pipeline.h" -#include "tnl/t_vertex.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_tris.h" -#include "intel_batchbuffer.h" -#include "intel_buffers.h" -#include "intel_reg.h" -#include "intel_span.h" -#include "intel_tex.h" - -static void intelRenderPrimitive(GLcontext * ctx, GLenum prim); -static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim, - GLuint hwprim); - -/* - */ -static void -intel_flush_inline_primitive(struct intel_context *intel) -{ - GLuint used = intel->batch->ptr - intel->prim.start_ptr; - - assert(intel->prim.primitive != ~0); - -/* _mesa_printf("/\n"); */ - - if (used < 8) - goto do_discard; - - *(int *) intel->prim.start_ptr = (_3DPRIMITIVE | - intel->prim.primitive | (used / 4 - 2)); - - goto finished; - - do_discard: - intel->batch->ptr -= used; - - finished: - intel->prim.primitive = ~0; - intel->prim.start_ptr = 0; - intel->prim.flush = 0; -} - - -/* Emit a primitive referencing vertices in a vertex buffer. - */ -void -intelStartInlinePrimitive(struct intel_context *intel, - GLuint prim, GLuint batch_flags) -{ - BATCH_LOCALS; - - intel->vtbl.emit_state(intel); - - /* Need to make sure at the very least that we don't wrap - * batchbuffers in BEGIN_BATCH below, otherwise the primitive will - * be emitted to a batchbuffer missing the required full-state - * preamble. - */ - if (intel_batchbuffer_space(intel->batch) < 100) { - intel_batchbuffer_flush(intel->batch); - intel->vtbl.emit_state(intel); - } - -/* _mesa_printf("%s *", __progname); */ - - intel_wait_flips(intel, batch_flags); - - /* Emit a slot which will be filled with the inline primitive - * command later. - */ - BEGIN_BATCH(2, batch_flags); - OUT_BATCH(0); - - intel->prim.start_ptr = intel->batch->ptr; - intel->prim.primitive = prim; - intel->prim.flush = intel_flush_inline_primitive; - - OUT_BATCH(0); - ADVANCE_BATCH(); - -/* _mesa_printf(">"); */ -} - - -void -intelWrapInlinePrimitive(struct intel_context *intel) -{ - GLuint prim = intel->prim.primitive; - GLuint batchflags = intel->batch->flags; - - intel_flush_inline_primitive(intel); - intel_batchbuffer_flush(intel->batch); - intelStartInlinePrimitive(intel, prim, batchflags); /* ??? */ -} - -GLuint * -intelExtendInlinePrimitive(struct intel_context *intel, GLuint dwords) -{ - GLuint sz = dwords * sizeof(GLuint); - GLuint *ptr; - - assert(intel->prim.flush == intel_flush_inline_primitive); - - if (intel_batchbuffer_space(intel->batch) < sz) - intelWrapInlinePrimitive(intel); - -/* _mesa_printf("."); */ - - intel->vtbl.assert_not_dirty(intel); - - ptr = (GLuint *) intel->batch->ptr; - intel->batch->ptr += sz; - - return ptr; -} - - - -/*********************************************************************** - * Emit primitives as inline vertices * - ***********************************************************************/ - -#ifdef __i386__ -#define COPY_DWORDS( j, vb, vertsize, v ) \ -do { \ - int __tmp; \ - __asm__ __volatile__( "rep ; movsl" \ - : "=%c" (j), "=D" (vb), "=S" (__tmp) \ - : "0" (vertsize), \ - "D" ((long)vb), \ - "S" ((long)v) ); \ -} while (0) -#else -#define COPY_DWORDS( j, vb, vertsize, v ) \ -do { \ - for ( j = 0 ; j < vertsize ; j++ ) { \ - vb[j] = ((GLuint *)v)[j]; \ - } \ - vb += vertsize; \ -} while (0) -#endif - -static void -intel_draw_quad(struct intel_context *intel, - intelVertexPtr v0, - intelVertexPtr v1, intelVertexPtr v2, intelVertexPtr v3) -{ -#ifndef NO_REND - GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive(intel, 6 * vertsize); - int j; - - COPY_DWORDS(j, vb, vertsize, v0); - COPY_DWORDS(j, vb, vertsize, v1); - - /* If smooth shading, draw like a trifan which gives better - * rasterization. Otherwise draw as two triangles with provoking - * vertex in third position as required for flat shading. - */ - if (intel->ctx.Light.ShadeModel == GL_FLAT) { - COPY_DWORDS(j, vb, vertsize, v3); - COPY_DWORDS(j, vb, vertsize, v1); - } - else { - COPY_DWORDS(j, vb, vertsize, v2); - COPY_DWORDS(j, vb, vertsize, v0); - } - - COPY_DWORDS(j, vb, vertsize, v2); - COPY_DWORDS(j, vb, vertsize, v3); -#endif -} - -static void -intel_draw_triangle(struct intel_context *intel, - intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2) -{ -#ifndef NO_REND - GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive(intel, 3 * vertsize); - int j; - - COPY_DWORDS(j, vb, vertsize, v0); - COPY_DWORDS(j, vb, vertsize, v1); - COPY_DWORDS(j, vb, vertsize, v2); -#endif -} - - -static void -intel_draw_line(struct intel_context *intel, - intelVertexPtr v0, intelVertexPtr v1) -{ -#ifndef NO_REND - GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive(intel, 2 * vertsize); - int j; - - COPY_DWORDS(j, vb, vertsize, v0); - COPY_DWORDS(j, vb, vertsize, v1); -#endif -} - - -static void -intel_draw_point(struct intel_context *intel, intelVertexPtr v0) -{ -#ifndef NO_REND - GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive(intel, vertsize); - int j; - - /* Adjust for sub pixel position -- still required for conform. */ - *(float *) &vb[0] = v0->v.x - 0.125; - *(float *) &vb[1] = v0->v.y - 0.125; - for (j = 2; j < vertsize; j++) - vb[j] = v0->ui[j]; -#endif -} - - - -/*********************************************************************** - * Fixup for ARB_point_parameters * - ***********************************************************************/ - -/* Currently not working - VERT_ATTRIB_POINTSIZE isn't correctly - * represented in the fragment program InputsRead field. - */ -static void -intel_atten_point(struct intel_context *intel, intelVertexPtr v0) -{ - GLcontext *ctx = &intel->ctx; - GLfloat psz[4], col[4], restore_psz, restore_alpha; - - _tnl_get_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz); - _tnl_get_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col); - - restore_psz = psz[0]; - restore_alpha = col[3]; - - if (psz[0] >= ctx->Point.Threshold) { - psz[0] = MIN2(psz[0], ctx->Point.MaxSize); - } - else { - GLfloat dsize = psz[0] / ctx->Point.Threshold; - psz[0] = MAX2(ctx->Point.Threshold, ctx->Point.MinSize); - col[3] *= dsize * dsize; - } - - if (psz[0] < 1.0) - psz[0] = 1.0; - - if (restore_psz != psz[0] || restore_alpha != col[3]) { - _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz); - _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col); - - intel_draw_point(intel, v0); - - psz[0] = restore_psz; - col[3] = restore_alpha; - - _tnl_set_attr(ctx, v0, _TNL_ATTRIB_POINTSIZE, psz); - _tnl_set_attr(ctx, v0, _TNL_ATTRIB_COLOR0, col); - } - else - intel_draw_point(intel, v0); -} - - - - - -/*********************************************************************** - * Fixup for I915 WPOS texture coordinate * - ***********************************************************************/ - - - -static void -intel_wpos_triangle(struct intel_context *intel, - intelVertexPtr v0, intelVertexPtr v1, intelVertexPtr v2) -{ - GLuint offset = intel->wpos_offset; - GLuint size = intel->wpos_size; - - __memcpy(((char *) v0) + offset, v0, size); - __memcpy(((char *) v1) + offset, v1, size); - __memcpy(((char *) v2) + offset, v2, size); - - intel_draw_triangle(intel, v0, v1, v2); -} - - -static void -intel_wpos_line(struct intel_context *intel, - intelVertexPtr v0, intelVertexPtr v1) -{ - GLuint offset = intel->wpos_offset; - GLuint size = intel->wpos_size; - - __memcpy(((char *) v0) + offset, v0, size); - __memcpy(((char *) v1) + offset, v1, size); - - intel_draw_line(intel, v0, v1); -} - - -static void -intel_wpos_point(struct intel_context *intel, intelVertexPtr v0) -{ - GLuint offset = intel->wpos_offset; - GLuint size = intel->wpos_size; - - __memcpy(((char *) v0) + offset, v0, size); - - intel_draw_point(intel, v0); -} - - - - - - -/*********************************************************************** - * Macros for t_dd_tritmp.h to draw basic primitives * - ***********************************************************************/ - -#define TRI( a, b, c ) \ -do { \ - if (DO_FALLBACK) \ - intel->draw_tri( intel, a, b, c ); \ - else \ - intel_draw_triangle( intel, a, b, c ); \ -} while (0) - -#define QUAD( a, b, c, d ) \ -do { \ - if (DO_FALLBACK) { \ - intel->draw_tri( intel, a, b, d ); \ - intel->draw_tri( intel, b, c, d ); \ - } else \ - intel_draw_quad( intel, a, b, c, d ); \ -} while (0) - -#define LINE( v0, v1 ) \ -do { \ - if (DO_FALLBACK) \ - intel->draw_line( intel, v0, v1 ); \ - else \ - intel_draw_line( intel, v0, v1 ); \ -} while (0) - -#define POINT( v0 ) \ -do { \ - if (DO_FALLBACK) \ - intel->draw_point( intel, v0 ); \ - else \ - intel_draw_point( intel, v0 ); \ -} while (0) - - -/*********************************************************************** - * Build render functions from dd templates * - ***********************************************************************/ - -#define INTEL_OFFSET_BIT 0x01 -#define INTEL_TWOSIDE_BIT 0x02 -#define INTEL_UNFILLED_BIT 0x04 -#define INTEL_FALLBACK_BIT 0x08 -#define INTEL_MAX_TRIFUNC 0x10 - - -static struct -{ - tnl_points_func points; - tnl_line_func line; - tnl_triangle_func triangle; - tnl_quad_func quad; -} rast_tab[INTEL_MAX_TRIFUNC]; - - -#define DO_FALLBACK (IND & INTEL_FALLBACK_BIT) -#define DO_OFFSET (IND & INTEL_OFFSET_BIT) -#define DO_UNFILLED (IND & INTEL_UNFILLED_BIT) -#define DO_TWOSIDE (IND & INTEL_TWOSIDE_BIT) -#define DO_FLAT 0 -#define DO_TRI 1 -#define DO_QUAD 1 -#define DO_LINE 1 -#define DO_POINTS 1 -#define DO_FULL_QUAD 1 - -#define HAVE_RGBA 1 -#define HAVE_SPEC 1 -#define HAVE_BACK_COLORS 0 -#define HAVE_HW_FLATSHADE 1 -#define VERTEX intelVertex -#define TAB rast_tab - -/* Only used to pull back colors into vertices (ie, we know color is - * floating point). - */ -#define INTEL_COLOR( dst, src ) \ -do { \ - UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \ - UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \ - UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \ - UNCLAMPED_FLOAT_TO_UBYTE((dst)[3], (src)[3]); \ -} while (0) - -#define INTEL_SPEC( dst, src ) \ -do { \ - UNCLAMPED_FLOAT_TO_UBYTE((dst)[0], (src)[2]); \ - UNCLAMPED_FLOAT_TO_UBYTE((dst)[1], (src)[1]); \ - UNCLAMPED_FLOAT_TO_UBYTE((dst)[2], (src)[0]); \ -} while (0) - - -#define DEPTH_SCALE intel->polygon_offset_scale -#define UNFILLED_TRI unfilled_tri -#define UNFILLED_QUAD unfilled_quad -#define VERT_X(_v) _v->v.x -#define VERT_Y(_v) _v->v.y -#define VERT_Z(_v) _v->v.z -#define AREA_IS_CCW( a ) (a > 0) -#define GET_VERTEX(e) (intel->verts + (e * intel->vertex_size * sizeof(GLuint))) - -#define VERT_SET_RGBA( v, c ) if (coloroffset) INTEL_COLOR( v->ub4[coloroffset], c ) -#define VERT_COPY_RGBA( v0, v1 ) if (coloroffset) v0->ui[coloroffset] = v1->ui[coloroffset] -#define VERT_SAVE_RGBA( idx ) if (coloroffset) color[idx] = v[idx]->ui[coloroffset] -#define VERT_RESTORE_RGBA( idx ) if (coloroffset) v[idx]->ui[coloroffset] = color[idx] - -#define VERT_SET_SPEC( v, c ) if (specoffset) INTEL_SPEC( v->ub4[specoffset], c ) -#define VERT_COPY_SPEC( v0, v1 ) if (specoffset) COPY_3V(v0->ub4[specoffset], v1->ub4[specoffset]) -#define VERT_SAVE_SPEC( idx ) if (specoffset) spec[idx] = v[idx]->ui[specoffset] -#define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx] - -#define LOCAL_VARS(n) \ - struct intel_context *intel = intel_context(ctx); \ - GLuint color[n], spec[n]; \ - GLuint coloroffset = intel->coloroffset; \ - GLboolean specoffset = intel->specoffset; \ - (void) color; (void) spec; (void) coloroffset; (void) specoffset; - - -/*********************************************************************** - * Helpers for rendering unfilled primitives * - ***********************************************************************/ - -static const GLuint hw_prim[GL_POLYGON + 1] = { - PRIM3D_POINTLIST, - PRIM3D_LINELIST, - PRIM3D_LINELIST, - PRIM3D_LINELIST, - PRIM3D_TRILIST, - PRIM3D_TRILIST, - PRIM3D_TRILIST, - PRIM3D_TRILIST, - PRIM3D_TRILIST, - PRIM3D_TRILIST -}; - -#define RASTERIZE(x) intelRasterPrimitive( ctx, x, hw_prim[x] ) -#define RENDER_PRIMITIVE intel->render_primitive -#define TAG(x) x -#define IND INTEL_FALLBACK_BIT -#include "tnl_dd/t_dd_unfilled.h" -#undef IND - -/*********************************************************************** - * Generate GL render functions * - ***********************************************************************/ - -#define IND (0) -#define TAG(x) x -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_OFFSET_BIT) -#define TAG(x) x##_offset -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_TWOSIDE_BIT) -#define TAG(x) x##_twoside -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT) -#define TAG(x) x##_twoside_offset -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_UNFILLED_BIT) -#define TAG(x) x##_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT) -#define TAG(x) x##_offset_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT) -#define TAG(x) x##_twoside_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT) -#define TAG(x) x##_twoside_offset_unfilled -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_FALLBACK_BIT) -#define TAG(x) x##_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT) -#define TAG(x) x##_offset_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_TWOSIDE_BIT|INTEL_FALLBACK_BIT) -#define TAG(x) x##_twoside_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_FALLBACK_BIT) -#define TAG(x) x##_twoside_offset_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT) -#define TAG(x) x##_unfilled_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT) -#define TAG(x) x##_offset_unfilled_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_TWOSIDE_BIT|INTEL_UNFILLED_BIT|INTEL_FALLBACK_BIT) -#define TAG(x) x##_twoside_unfilled_fallback -#include "tnl_dd/t_dd_tritmp.h" - -#define IND (INTEL_TWOSIDE_BIT|INTEL_OFFSET_BIT|INTEL_UNFILLED_BIT| \ - INTEL_FALLBACK_BIT) -#define TAG(x) x##_twoside_offset_unfilled_fallback -#include "tnl_dd/t_dd_tritmp.h" - - -static void -init_rast_tab(void) -{ - init(); - init_offset(); - init_twoside(); - init_twoside_offset(); - init_unfilled(); - init_offset_unfilled(); - init_twoside_unfilled(); - init_twoside_offset_unfilled(); - init_fallback(); - init_offset_fallback(); - init_twoside_fallback(); - init_twoside_offset_fallback(); - init_unfilled_fallback(); - init_offset_unfilled_fallback(); - init_twoside_unfilled_fallback(); - init_twoside_offset_unfilled_fallback(); -} - - -/*********************************************************************** - * Rasterization fallback helpers * - ***********************************************************************/ - - -/* This code is hit only when a mix of accelerated and unaccelerated - * primitives are being drawn, and only for the unaccelerated - * primitives. - */ -static void -intel_fallback_tri(struct intel_context *intel, - intelVertex * v0, intelVertex * v1, intelVertex * v2) -{ - GLcontext *ctx = &intel->ctx; - SWvertex v[3]; - - if (0) - fprintf(stderr, "\n%s\n", __FUNCTION__); - - INTEL_FIREVERTICES(intel); - - _swsetup_Translate(ctx, v0, &v[0]); - _swsetup_Translate(ctx, v1, &v[1]); - _swsetup_Translate(ctx, v2, &v[2]); - intelSpanRenderStart(ctx); - _swrast_Triangle(ctx, &v[0], &v[1], &v[2]); - intelSpanRenderFinish(ctx); -} - - -static void -intel_fallback_line(struct intel_context *intel, - intelVertex * v0, intelVertex * v1) -{ - GLcontext *ctx = &intel->ctx; - SWvertex v[2]; - - if (0) - fprintf(stderr, "\n%s\n", __FUNCTION__); - - INTEL_FIREVERTICES(intel); - - _swsetup_Translate(ctx, v0, &v[0]); - _swsetup_Translate(ctx, v1, &v[1]); - intelSpanRenderStart(ctx); - _swrast_Line(ctx, &v[0], &v[1]); - intelSpanRenderFinish(ctx); -} - -static void -intel_fallback_point(struct intel_context *intel, - intelVertex * v0) -{ - GLcontext *ctx = &intel->ctx; - SWvertex v[1]; - - if (0) - fprintf(stderr, "\n%s\n", __FUNCTION__); - - INTEL_FIREVERTICES(intel); - - _swsetup_Translate(ctx, v0, &v[0]); - intelSpanRenderStart(ctx); - _swrast_Point(ctx, &v[0]); - intelSpanRenderFinish(ctx); -} - - -/**********************************************************************/ -/* Render unclipped begin/end objects */ -/**********************************************************************/ - -#define IND 0 -#define V(x) (intelVertex *)(vertptr + ((x)*vertsize*sizeof(GLuint))) -#define RENDER_POINTS( start, count ) \ - for ( ; start < count ; start++) POINT( V(ELT(start)) ); -#define RENDER_LINE( v0, v1 ) LINE( V(v0), V(v1) ) -#define RENDER_TRI( v0, v1, v2 ) TRI( V(v0), V(v1), V(v2) ) -#define RENDER_QUAD( v0, v1, v2, v3 ) QUAD( V(v0), V(v1), V(v2), V(v3) ) -#define INIT(x) intelRenderPrimitive( ctx, x ) -#undef LOCAL_VARS -#define LOCAL_VARS \ - struct intel_context *intel = intel_context(ctx); \ - GLubyte *vertptr = (GLubyte *)intel->verts; \ - const GLuint vertsize = intel->vertex_size; \ - const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts; \ - (void) elt; -#define RESET_STIPPLE -#define RESET_OCCLUSION -#define PRESERVE_VB_DEFS -#define ELT(x) x -#define TAG(x) intel_##x##_verts -#include "tnl/t_vb_rendertmp.h" -#undef ELT -#undef TAG -#define TAG(x) intel_##x##_elts -#define ELT(x) elt[x] -#include "tnl/t_vb_rendertmp.h" - -/**********************************************************************/ -/* Render clipped primitives */ -/**********************************************************************/ - - - -static void -intelRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n) -{ - struct intel_context *intel = intel_context(ctx); - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb; - GLuint prim = intel->render_primitive; - - /* Render the new vertices as an unclipped polygon. - */ - { - GLuint *tmp = VB->Elts; - VB->Elts = (GLuint *) elts; - tnl->Driver.Render.PrimTabElts[GL_POLYGON] (ctx, 0, n, - PRIM_BEGIN | PRIM_END); - VB->Elts = tmp; - } - - /* Restore the render primitive - */ - if (prim != GL_POLYGON) - tnl->Driver.Render.PrimitiveNotify(ctx, prim); -} - -static void -intelRenderClippedLine(GLcontext * ctx, GLuint ii, GLuint jj) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - - tnl->Driver.Render.Line(ctx, ii, jj); -} - -static void -intelFastRenderClippedPoly(GLcontext * ctx, const GLuint * elts, GLuint n) -{ -#ifndef NO_REND - struct intel_context *intel = intel_context(ctx); - const GLuint vertsize = intel->vertex_size; - GLuint *vb = intelExtendInlinePrimitive(intel, (n - 2) * 3 * vertsize); - GLubyte *vertptr = (GLubyte *) intel->verts; - const GLuint *start = (const GLuint *) V(elts[0]); - int i, j; - - for (i = 2; i < n; i++) { - COPY_DWORDS(j, vb, vertsize, V(elts[i - 1])); - COPY_DWORDS(j, vb, vertsize, V(elts[i])); - COPY_DWORDS(j, vb, vertsize, start); - } -#endif -} - -/**********************************************************************/ -/* Choose render functions */ -/**********************************************************************/ - - - - -#define ANY_FALLBACK_FLAGS (DD_LINE_STIPPLE | DD_TRI_STIPPLE | DD_POINT_ATTEN | DD_POINT_SMOOTH | DD_TRI_SMOOTH) -#define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE | DD_TRI_OFFSET | DD_TRI_UNFILLED) - -void -intelChooseRenderState(GLcontext * ctx) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - struct intel_context *intel = intel_context(ctx); - GLuint flags = ctx->_TriangleCaps; - const struct gl_fragment_program *fprog = ctx->FragmentProgram._Current; - GLboolean have_wpos = (fprog && (fprog->Base.InputsRead & FRAG_BIT_WPOS)); - GLuint index = 0; - - if (INTEL_DEBUG & DEBUG_STATE) - fprintf(stderr, "\n%s\n", __FUNCTION__); - - if ((flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) || have_wpos) { - - if (flags & ANY_RASTER_FLAGS) { - if (flags & DD_TRI_LIGHT_TWOSIDE) - index |= INTEL_TWOSIDE_BIT; - if (flags & DD_TRI_OFFSET) - index |= INTEL_OFFSET_BIT; - if (flags & DD_TRI_UNFILLED) - index |= INTEL_UNFILLED_BIT; - } - - if (have_wpos) { - intel->draw_point = intel_wpos_point; - intel->draw_line = intel_wpos_line; - intel->draw_tri = intel_wpos_triangle; - - /* Make sure these get called: - */ - index |= INTEL_FALLBACK_BIT; - } - else { - intel->draw_point = intel_draw_point; - intel->draw_line = intel_draw_line; - intel->draw_tri = intel_draw_triangle; - } - - /* Hook in fallbacks for specific primitives. - */ - if (flags & ANY_FALLBACK_FLAGS) { - if (flags & DD_LINE_STIPPLE) - intel->draw_line = intel_fallback_line; - - if ((flags & DD_TRI_STIPPLE) && !intel->hw_stipple) - intel->draw_tri = intel_fallback_tri; - - if (flags & DD_TRI_SMOOTH) { - if (intel->strict_conformance) - intel->draw_tri = intel_fallback_tri; - } - - if (flags & DD_POINT_ATTEN) { - if (0) - intel->draw_point = intel_atten_point; - else - intel->draw_point = intel_fallback_point; - } - - if (flags & DD_POINT_SMOOTH) { - if (intel->strict_conformance) - intel->draw_point = intel_fallback_point; - } - - index |= INTEL_FALLBACK_BIT; - } - } - - if (intel->RenderIndex != index) { - intel->RenderIndex = index; - - tnl->Driver.Render.Points = rast_tab[index].points; - tnl->Driver.Render.Line = rast_tab[index].line; - tnl->Driver.Render.Triangle = rast_tab[index].triangle; - tnl->Driver.Render.Quad = rast_tab[index].quad; - - if (index == 0) { - tnl->Driver.Render.PrimTabVerts = intel_render_tab_verts; - tnl->Driver.Render.PrimTabElts = intel_render_tab_elts; - tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */ - tnl->Driver.Render.ClippedPolygon = intelFastRenderClippedPoly; - } - else { - tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts; - tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts; - tnl->Driver.Render.ClippedLine = intelRenderClippedLine; - tnl->Driver.Render.ClippedPolygon = intelRenderClippedPoly; - } - } -} - -static const GLenum reduced_prim[GL_POLYGON + 1] = { - GL_POINTS, - GL_LINES, - GL_LINES, - GL_LINES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES, - GL_TRIANGLES -}; - - -/**********************************************************************/ -/* High level hooks for t_vb_render.c */ -/**********************************************************************/ - - - - -static void -intelRunPipeline(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - - _mesa_lock_context_textures(ctx); - - if (ctx->NewState) - _mesa_update_state_locked(ctx); - - if (intel->NewGLState) { - if (intel->NewGLState & _NEW_TEXTURE) { - intel->vtbl.update_texture_state(intel); - } - - if (!intel->Fallback) { - if (intel->NewGLState & _INTEL_NEW_RENDERSTATE) - intelChooseRenderState(ctx); - } - - intel->NewGLState = 0; - } - - _tnl_run_pipeline(ctx); - - _mesa_unlock_context_textures(ctx); -} - -static void -intelRenderStart(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - - intel->vtbl.render_start(intel_context(ctx)); - intel->vtbl.emit_state(intel); -} - -static void -intelRenderFinish(GLcontext * ctx) -{ - struct intel_context *intel = intel_context(ctx); - - if (intel->RenderIndex & INTEL_FALLBACK_BIT) - _swrast_flush(ctx); - - INTEL_FIREVERTICES(intel); -} - - - - - /* System to flush dma and emit state changes based on the rasterized - * primitive. - */ -static void -intelRasterPrimitive(GLcontext * ctx, GLenum rprim, GLuint hwprim) -{ - struct intel_context *intel = intel_context(ctx); - - if (0) - fprintf(stderr, "%s %s %x\n", __FUNCTION__, - _mesa_lookup_enum_by_nr(rprim), hwprim); - - intel->vtbl.reduced_primitive_state(intel, rprim); - - /* Start a new primitive. Arrange to have it flushed later on. - */ - if (hwprim != intel->prim.primitive) { - INTEL_FIREVERTICES(intel); - - intelStartInlinePrimitive(intel, hwprim, INTEL_BATCH_CLIPRECTS); - } -} - - - /* - */ -static void -intelRenderPrimitive(GLcontext * ctx, GLenum prim) -{ - struct intel_context *intel = intel_context(ctx); - - if (0) - fprintf(stderr, "%s %s\n", __FUNCTION__, _mesa_lookup_enum_by_nr(prim)); - - /* Let some clipping routines know which primitive they're dealing - * with. - */ - intel->render_primitive = prim; - - /* Shortcircuit this when called from t_dd_rendertmp.h for unfilled - * triangles. The rasterized primitive will always be reset by - * lower level functions in that case, potentially pingponging the - * state: - */ - if (reduced_prim[prim] == GL_TRIANGLES && - (ctx->_TriangleCaps & DD_TRI_UNFILLED)) - return; - - /* Set some primitive-dependent state and Start? a new primitive. - */ - intelRasterPrimitive(ctx, reduced_prim[prim], hw_prim[prim]); -} - - - /**********************************************************************/ - /* Transition to/from hardware rasterization. */ - /**********************************************************************/ - -static char *fallbackStrings[] = { - [0] = "Draw buffer", - [1] = "Read buffer", - [2] = "Depth buffer", - [3] = "Stencil buffer", - [4] = "User disable", - [5] = "Render mode", - - [12] = "Texture", - [13] = "Color mask", - [14] = "Stencil", - [15] = "Stipple", - [16] = "Program", - [17] = "Logic op", - [18] = "Smooth polygon", - [19] = "Smooth point", -}; - - -static char * -getFallbackString(GLuint bit) -{ - int i = 0; - while (bit > 1) { - i++; - bit >>= 1; - } - return fallbackStrings[i]; -} - - - -void -intelFallback(struct intel_context *intel, GLuint bit, GLboolean mode) -{ - GLcontext *ctx = &intel->ctx; - TNLcontext *tnl = TNL_CONTEXT(ctx); - GLuint oldfallback = intel->Fallback; - - if (mode) { - intel->Fallback |= bit; - if (oldfallback == 0) { - intelFlush(ctx); - if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "ENTER FALLBACK %x: %s\n", - bit, getFallbackString(bit)); - _swsetup_Wakeup(ctx); - intel->RenderIndex = ~0; - } - } - else { - intel->Fallback &= ~bit; - if (oldfallback == bit) { - _swrast_flush(ctx); - if (INTEL_DEBUG & DEBUG_FALLBACKS) - fprintf(stderr, "LEAVE FALLBACK %s\n", getFallbackString(bit)); - tnl->Driver.Render.Start = intelRenderStart; - tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive; - tnl->Driver.Render.Finish = intelRenderFinish; - tnl->Driver.Render.BuildVertices = _tnl_build_vertices; - tnl->Driver.Render.CopyPV = _tnl_copy_pv; - tnl->Driver.Render.Interp = _tnl_interp; - - _tnl_invalidate_vertex_state(ctx, ~0); - _tnl_invalidate_vertices(ctx, ~0); - _tnl_install_attrs(ctx, - intel->vertex_attrs, - intel->vertex_attr_count, - intel->ViewportMatrix.m, 0); - - intel->NewGLState |= _INTEL_NEW_RENDERSTATE; - } - } -} - -union fi -{ - GLfloat f; - GLint i; -}; - - -/**********************************************************************/ -/* Used only with the metaops callbacks. */ -/**********************************************************************/ -void -intel_meta_draw_poly(struct intel_context *intel, - GLuint n, - GLfloat xy[][2], - GLfloat z, GLuint color, GLfloat tex[][2]) -{ -#ifndef NO_REND - union fi *vb; - GLint i; - - /* All 3d primitives should be emitted with INTEL_BATCH_CLIPRECTS, - * otherwise the drawing origin (DR4) might not be set correctly. - */ - intelStartInlinePrimitive(intel, PRIM3D_TRIFAN, INTEL_BATCH_CLIPRECTS); - vb = (union fi *) intelExtendInlinePrimitive(intel, n * 6); - - for (i = 0; i < n; i++) { - vb[0].f = xy[i][0]; - vb[1].f = xy[i][1]; - vb[2].f = z; - vb[3].i = color; - vb[4].f = tex[i][0]; - vb[5].f = tex[i][1]; - vb += 6; - } - - INTEL_FIREVERTICES(intel); -#endif -} - -void -intel_meta_draw_quad(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, - GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1) -{ - GLfloat xy[4][2]; - GLfloat tex[4][2]; - - xy[0][0] = x0; - xy[0][1] = y0; - xy[1][0] = x1; - xy[1][1] = y0; - xy[2][0] = x1; - xy[2][1] = y1; - xy[3][0] = x0; - xy[3][1] = y1; - - tex[0][0] = s0; - tex[0][1] = t0; - tex[1][0] = s1; - tex[1][1] = t0; - tex[2][0] = s1; - tex[2][1] = t1; - tex[3][0] = s0; - tex[3][1] = t1; - - intel_meta_draw_poly(intel, 4, xy, z, color, tex); -} - - - -/**********************************************************************/ -/* Initialization. */ -/**********************************************************************/ - - -void -intelInitTriFuncs(GLcontext * ctx) -{ - TNLcontext *tnl = TNL_CONTEXT(ctx); - static int firsttime = 1; - - if (firsttime) { - init_rast_tab(); - firsttime = 0; - } - - tnl->Driver.RunPipeline = intelRunPipeline; - tnl->Driver.Render.Start = intelRenderStart; - tnl->Driver.Render.Finish = intelRenderFinish; - tnl->Driver.Render.PrimitiveNotify = intelRenderPrimitive; - tnl->Driver.Render.ResetLineStipple = _swrast_ResetLineStipple; - tnl->Driver.Render.BuildVertices = _tnl_build_vertices; - tnl->Driver.Render.CopyPV = _tnl_copy_pv; - tnl->Driver.Render.Interp = _tnl_interp; -} diff --git a/src/mesa/drivers/dri/i915tex/intel_tris.h b/src/mesa/drivers/dri/i915tex/intel_tris.h deleted file mode 100644 index b7bae8cd3b..0000000000 --- a/src/mesa/drivers/dri/i915tex/intel_tris.h +++ /dev/null @@ -1,69 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTELTRIS_INC -#define INTELTRIS_INC - -#include "mtypes.h" - - - -#define _INTEL_NEW_RENDERSTATE (_DD_NEW_LINE_STIPPLE | \ - _DD_NEW_TRI_UNFILLED | \ - _DD_NEW_TRI_LIGHT_TWOSIDE | \ - _DD_NEW_TRI_OFFSET | \ - _DD_NEW_TRI_STIPPLE | \ - _NEW_PROGRAM | \ - _NEW_POLYGONSTIPPLE) - -extern void intelInitTriFuncs(GLcontext * ctx); - -extern void intelChooseRenderState(GLcontext * ctx); - -extern void intelStartInlinePrimitive(struct intel_context *intel, - GLuint prim, GLuint flags); -extern void intelWrapInlinePrimitive(struct intel_context *intel); - -GLuint *intelExtendInlinePrimitive(struct intel_context *intel, - GLuint dwords); - - -void intel_meta_draw_quad(struct intel_context *intel, - GLfloat x0, GLfloat x1, - GLfloat y0, GLfloat y1, - GLfloat z, - GLuint color, - GLfloat s0, GLfloat s1, GLfloat t0, GLfloat t1); - -void intel_meta_draw_poly(struct intel_context *intel, - GLuint n, - GLfloat xy[][2], - GLfloat z, GLuint color, GLfloat tex[][2]); - - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/server/i830_common.h b/src/mesa/drivers/dri/i915tex/server/i830_common.h deleted file mode 100644 index d4d58886ce..0000000000 --- a/src/mesa/drivers/dri/i915tex/server/i830_common.h +++ /dev/null @@ -1,226 +0,0 @@ -/************************************************************************** - -Copyright 2001 VA Linux Systems Inc., Fremont, California. -Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas. - -All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining a -copy of this software and associated documentation files (the "Software"), -to deal in the Software without restriction, including without limitation -on the rights to use, copy, modify, merge, publish, distribute, sub -license, and/or sell copies of the Software, and to permit persons to whom -the Software is furnished to do so, subject to the following conditions: - -The above copyright notice and this permission notice (including the next -paragraph) shall be included in all copies or substantial portions of the -Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL -ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, -DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR -OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE -USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ - -#ifndef _I830_COMMON_H_ -#define _I830_COMMON_H_ - - -#define I830_NR_TEX_REGIONS 255 /* maximum due to use of chars for next/prev */ -#define I830_LOG_MIN_TEX_REGION_SIZE 14 - - -/* Driver specific DRM command indices - * NOTE: these are not OS specific, but they are driver specific - */ -#define DRM_I830_INIT 0x00 -#define DRM_I830_FLUSH 0x01 -#define DRM_I830_FLIP 0x02 -#define DRM_I830_BATCHBUFFER 0x03 -#define DRM_I830_IRQ_EMIT 0x04 -#define DRM_I830_IRQ_WAIT 0x05 -#define DRM_I830_GETPARAM 0x06 -#define DRM_I830_SETPARAM 0x07 -#define DRM_I830_ALLOC 0x08 -#define DRM_I830_FREE 0x09 -#define DRM_I830_INIT_HEAP 0x0a -#define DRM_I830_CMDBUFFER 0x0b -#define DRM_I830_DESTROY_HEAP 0x0c -#define DRM_I830_SET_VBLANK_PIPE 0x0d -#define DRM_I830_GET_VBLANK_PIPE 0x0e - -typedef struct { - enum { - I830_INIT_DMA = 0x01, - I830_CLEANUP_DMA = 0x02, - I830_RESUME_DMA = 0x03 - } func; - unsigned int mmio_offset; - int sarea_priv_offset; - unsigned int ring_start; - unsigned int ring_end; - unsigned int ring_size; - unsigned int front_offset; - unsigned int back_offset; - unsigned int depth_offset; - unsigned int w; - unsigned int h; - unsigned int pitch; - unsigned int pitch_bits; - unsigned int back_pitch; - unsigned int depth_pitch; - unsigned int cpp; - unsigned int chipset; -} drmI830Init; - -typedef struct { - drmTextureRegion texList[I830_NR_TEX_REGIONS+1]; - int last_upload; /* last time texture was uploaded */ - int last_enqueue; /* last time a buffer was enqueued */ - int last_dispatch; /* age of the most recently dispatched buffer */ - int ctxOwner; /* last context to upload state */ - int texAge; - int pf_enabled; /* is pageflipping allowed? */ - int pf_active; - int pf_current_page; /* which buffer is being displayed? */ - int perf_boxes; /* performance boxes to be displayed */ - int width, height; /* screen size in pixels */ - - drm_handle_t front_handle; - int front_offset; - int front_size; - - drm_handle_t back_handle; - int back_offset; - int back_size; - - drm_handle_t depth_handle; - int depth_offset; - int depth_size; - - drm_handle_t tex_handle; - int tex_offset; - int tex_size; - int log_tex_granularity; - int pitch; - int rotation; /* 0, 90, 180 or 270 */ - int rotated_offset; - int rotated_size; - int rotated_pitch; - int virtualX, virtualY; - - unsigned int front_tiled; - unsigned int back_tiled; - unsigned int depth_tiled; - unsigned int rotated_tiled; - unsigned int rotated2_tiled; - - int pipeA_x; - int pipeA_y; - int pipeA_w; - int pipeA_h; - int pipeB_x; - int pipeB_y; - int pipeB_w; - int pipeB_h; - - /* Triple buffering */ - drm_handle_t third_handle; - int third_offset; - int third_size; - unsigned int third_tiled; -} drmI830Sarea; - -/* Flags for perf_boxes - */ -#define I830_BOX_RING_EMPTY 0x1 /* populated by kernel */ -#define I830_BOX_FLIP 0x2 /* populated by kernel */ -#define I830_BOX_WAIT 0x4 /* populated by kernel & client */ -#define I830_BOX_TEXTURE_LOAD 0x8 /* populated by kernel */ -#define I830_BOX_LOST_CONTEXT 0x10 /* populated by client */ - - -typedef struct { - int start; /* agp offset */ - int used; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830BatchBuffer; - -typedef struct { - char *buf; /* agp offset */ - int sz; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830CmdBuffer; - -typedef struct { - int *irq_seq; -} drmI830IrqEmit; - -typedef struct { - int irq_seq; -} drmI830IrqWait; - -typedef struct { - int param; - int *value; -} drmI830GetParam; - -#define I830_PARAM_IRQ_ACTIVE 1 -#define I830_PARAM_ALLOW_BATCHBUFFER 2 - -typedef struct { - int param; - int value; -} drmI830SetParam; - -#define I830_SETPARAM_USE_MI_BATCHBUFFER_START 1 -#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY 2 -#define I830_SETPARAM_ALLOW_BATCHBUFFER 3 - - -/* A memory manager for regions of shared memory: - */ -#define I830_MEM_REGION_AGP 1 - -typedef struct { - int region; - int alignment; - int size; - int *region_offset; /* offset from start of fb or agp */ -} drmI830MemAlloc; - -typedef struct { - int region; - int region_offset; -} drmI830MemFree; - -typedef struct { - int region; - int size; - int start; -} drmI830MemInitHeap; - -typedef struct { - int region; -} drmI830MemDestroyHeap; - -#define DRM_I830_VBLANK_PIPE_A 1 -#define DRM_I830_VBLANK_PIPE_B 2 - -typedef struct { - int pipe; -} drmI830VBlankPipe; - -#endif /* _I830_DRM_H_ */ diff --git a/src/mesa/drivers/dri/i915tex/server/i830_dri.h b/src/mesa/drivers/dri/i915tex/server/i830_dri.h deleted file mode 100644 index c2a3af8cbf..0000000000 --- a/src/mesa/drivers/dri/i915tex/server/i830_dri.h +++ /dev/null @@ -1,63 +0,0 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ - -#ifndef _I830_DRI_H -#define _I830_DRI_H - -#include "xf86drm.h" -#include "i830_common.h" - -#define I830_MAX_DRAWABLES 256 - -#define I830_MAJOR_VERSION 1 -#define I830_MINOR_VERSION 7 -#define I830_PATCHLEVEL 2 - -#define I830_REG_SIZE 0x80000 - -typedef struct _I830DRIRec { - drm_handle_t regs; - drmSize regsSize; - - drmSize unused1; /* backbufferSize */ - drm_handle_t unused2; /* backbuffer */ - - drmSize unused3; /* depthbufferSize */ - drm_handle_t unused4; /* depthbuffer */ - - drmSize unused5; /* rotatedSize */ - drm_handle_t unused6; /* rotatedbuffer */ - - drm_handle_t unused7; /* textures */ - int unused8; /* textureSize */ - - drm_handle_t unused9; /* agp_buffers */ - drmSize unused10; /* agp_buf_size */ - - int deviceID; - int width; - int height; - int mem; - int cpp; - int bitsPerPixel; - - int unused11[8]; /* was front/back/depth/rotated offset/pitch */ - - int unused12; /* logTextureGranularity */ - int unused13; /* textureOffset */ - - int irq; - int sarea_priv_offset; -} I830DRIRec, *I830DRIPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830ConfigPrivRec, *I830ConfigPrivPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830DRIContextRec, *I830DRIContextPtr; - - -#endif diff --git a/src/mesa/drivers/dri/i915tex/server/intel.h b/src/mesa/drivers/dri/i915tex/server/intel.h deleted file mode 100644 index 6ea72499c1..0000000000 --- a/src/mesa/drivers/dri/i915tex/server/intel.h +++ /dev/null @@ -1,331 +0,0 @@ -#ifndef _INTEL_H_ -#define _INTEL_H_ - -#include "xf86drm.h" /* drm_handle_t, etc */ - -/* Intel */ -#ifndef PCI_CHIP_I810 -#define PCI_CHIP_I810 0x7121 -#define PCI_CHIP_I810_DC100 0x7123 -#define PCI_CHIP_I810_E 0x7125 -#define PCI_CHIP_I815 0x1132 -#define PCI_CHIP_I810_BRIDGE 0x7120 -#define PCI_CHIP_I810_DC100_BRIDGE 0x7122 -#define PCI_CHIP_I810_E_BRIDGE 0x7124 -#define PCI_CHIP_I815_BRIDGE 0x1130 -#endif - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 - -#ifndef PCI_CHIP_I855_GM -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I855_GM_BRIDGE 0x3580 -#endif - -#ifndef PCI_CHIP_I865_G -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I865_G_BRIDGE 0x2570 -#endif - -#ifndef PCI_CHIP_I915_G -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I915_GM -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I915_GM_BRIDGE 0x2590 -#endif - -#ifndef PCI_CHIP_E7221_G -#define PCI_CHIP_E7221_G 0x258A -/* Same as I915_G_BRIDGE */ -#define PCI_CHIP_E7221_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I945_G -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_G_BRIDGE 0x2770 -#endif - -#ifndef PCI_CHIP_I945_GM -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GM_BRIDGE 0x27A0 -#endif - -#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 || \ - pI810->Chipset == PCI_CHIP_I810_DC100 || \ - pI810->Chipset == PCI_CHIP_I810_E) -#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815) -#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M) -#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G) -#define IS_I85X(pI810) (pI810->Chipset == PCI_CHIP_I855_GM) -#define IS_I852(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME)) -#define IS_I855(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME)) -#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G) - -#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G) -#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM) -#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G) -#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM) -#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810)) - -#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810)) - -#define I830_GMCH_CTRL 0x52 - -#define I830_GMCH_MEM_MASK 0x1 -#define I830_GMCH_MEM_64M 0x1 -#define I830_GMCH_MEM_128M 0 - -#define I830_GMCH_GMS_MASK 0x70 -#define I830_GMCH_GMS_DISABLED 0x00 -#define I830_GMCH_GMS_LOCAL 0x10 -#define I830_GMCH_GMS_STOLEN_512 0x20 -#define I830_GMCH_GMS_STOLEN_1024 0x30 -#define I830_GMCH_GMS_STOLEN_8192 0x40 - -#define I855_GMCH_GMS_MASK (0x7 << 4) -#define I855_GMCH_GMS_DISABLED 0x00 -#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) -#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) -#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) -#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) -#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) -#define I915G_GMCH_GMS_STOLEN_48M (0x6 << 4) -#define I915G_GMCH_GMS_STOLEN_64M (0x7 << 4) - -typedef unsigned char Bool; -#define TRUE 1 -#define FALSE 0 - -#define PIPE_NONE 0<<0 -#define PIPE_CRT 1<<0 -#define PIPE_TV 1<<1 -#define PIPE_DFP 1<<2 -#define PIPE_LFP 1<<3 -#define PIPE_CRT2 1<<4 -#define PIPE_TV2 1<<5 -#define PIPE_DFP2 1<<6 -#define PIPE_LFP2 1<<7 - -typedef struct _I830MemPool *I830MemPoolPtr; -typedef struct _I830MemRange *I830MemRangePtr; -typedef struct _I830MemRange { - long Start; - long End; - long Size; - unsigned long Physical; - unsigned long Offset; /* Offset of AGP-allocated portion */ - unsigned long Alignment; - drm_handle_t Key; - unsigned long Pitch; // add pitch - I830MemPoolPtr Pool; -} I830MemRange; - -typedef struct _I830MemPool { - I830MemRange Total; - I830MemRange Free; - I830MemRange Fixed; - I830MemRange Allocated; -} I830MemPool; - -typedef struct { - int tail_mask; - I830MemRange mem; - unsigned char *virtual_start; - int head; - int tail; - int space; -} I830RingBuffer; - -typedef struct _I830Rec { - unsigned char *MMIOBase; - unsigned char *FbBase; - int cpp; - uint32_t aper_size; - unsigned int bios_version; - - /* These are set in PreInit and never changed. */ - long FbMapSize; - long TotalVideoRam; - I830MemRange StolenMemory; /* pre-allocated memory */ - long BIOSMemorySize; /* min stolen pool size */ - int BIOSMemSizeLoc; - - /* These change according to what has been allocated. */ - long FreeMemory; - I830MemRange MemoryAperture; - I830MemPool StolenPool; - long allocatedMemory; - - /* Regions allocated either from the above pools, or from agpgart. */ - /* for single and dual head configurations */ - I830MemRange FrontBuffer; - I830MemRange FrontBuffer2; - I830MemRange Scratch; - I830MemRange Scratch2; - - I830RingBuffer *LpRing; - - I830MemRange BackBuffer; - I830MemRange DepthBuffer; - I830MemRange TexMem; - int TexGranularity; - I830MemRange ContextMem; - int drmMinor; - Bool have3DWindows; - - Bool NeedRingBufferLow; - Bool allowPageFlip; - Bool disableTiling; - - int Chipset; - unsigned long LinearAddr; - unsigned long MMIOAddr; - - drmSize registerSize; /**< \brief MMIO register map size */ - drm_handle_t registerHandle; /**< \brief MMIO register map handle */ - // IOADDRESS ioBase; - int irq; /**< \brief IRQ number */ - int GttBound; - - drm_handle_t ring_map; - unsigned int Fence[8]; - -} I830Rec; - -/* - * 12288 is set as the maximum, chosen because it is enough for - * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare. - */ -#define I830_MAXIMUM_VBIOS_MEM 12288 -#define I830_DEFAULT_VIDEOMEM_2D (MB(32) / 1024) -#define I830_DEFAULT_VIDEOMEM_3D (MB(64) / 1024) - -/* Flags for memory allocation function */ -#define FROM_ANYWHERE 0x00000000 -#define FROM_POOL_ONLY 0x00000001 -#define FROM_NEW_ONLY 0x00000002 -#define FROM_MASK 0x0000000f - -#define ALLOCATE_AT_TOP 0x00000010 -#define ALLOCATE_AT_BOTTOM 0x00000020 -#define FORCE_GAPS 0x00000040 - -#define NEED_PHYSICAL_ADDR 0x00000100 -#define ALIGN_BOTH_ENDS 0x00000200 -#define FORCE_LOW 0x00000400 - -#define ALLOC_NO_TILING 0x00001000 -#define ALLOC_INITIAL 0x00002000 - -#define ALLOCATE_DRY_RUN 0x80000000 - -/* Chipset registers for VIDEO BIOS memory RW access */ -#define _855_DRAM_RW_CONTROL 0x58 -#define _845_DRAM_RW_CONTROL 0x90 -#define DRAM_WRITE 0x33330000 - -#define KB(x) ((x) * 1024) -#define MB(x) ((x) * KB(1024)) - -#define GTT_PAGE_SIZE KB(4) -#define ROUND_TO(x, y) (((x) + (y) - 1) / (y) * (y)) -#define ROUND_DOWN_TO(x, y) ((x) / (y) * (y)) -#define ROUND_TO_PAGE(x) ROUND_TO((x), GTT_PAGE_SIZE) -#define ROUND_TO_MB(x) ROUND_TO((x), MB(1)) -#define PRIMARY_RINGBUFFER_SIZE KB(128) - - -/* Ring buffer registers, p277, overview p19 - */ -#define LP_RING 0x2030 -#define HP_RING 0x2040 - -#define RING_TAIL 0x00 -#define TAIL_ADDR 0x000FFFF8 -#define I830_TAIL_MASK 0x001FFFF8 - -#define RING_HEAD 0x04 -#define HEAD_WRAP_COUNT 0xFFE00000 -#define HEAD_WRAP_ONE 0x00200000 -#define HEAD_ADDR 0x001FFFFC -#define I830_HEAD_MASK 0x001FFFFC - -#define RING_START 0x08 -#define START_ADDR 0x03FFFFF8 -#define I830_RING_START_MASK 0xFFFFF000 - -#define RING_LEN 0x0C -#define RING_NR_PAGES 0x001FF000 -#define I830_RING_NR_PAGES 0x001FF000 -#define RING_REPORT_MASK 0x00000006 -#define RING_REPORT_64K 0x00000002 -#define RING_REPORT_128K 0x00000004 -#define RING_NO_REPORT 0x00000000 -#define RING_VALID_MASK 0x00000001 -#define RING_VALID 0x00000001 -#define RING_INVALID 0x00000000 - - -/* Fence/Tiling ranges [0..7] - */ -#define FENCE 0x2000 -#define FENCE_NR 8 - -#define I915G_FENCE_START_MASK 0x0ff00000 - -#define I830_FENCE_START_MASK 0x07f80000 - -#define FENCE_START_MASK 0x03F80000 -#define FENCE_X_MAJOR 0x00000000 -#define FENCE_Y_MAJOR 0x00001000 -#define FENCE_SIZE_MASK 0x00000700 -#define FENCE_SIZE_512K 0x00000000 -#define FENCE_SIZE_1M 0x00000100 -#define FENCE_SIZE_2M 0x00000200 -#define FENCE_SIZE_4M 0x00000300 -#define FENCE_SIZE_8M 0x00000400 -#define FENCE_SIZE_16M 0x00000500 -#define FENCE_SIZE_32M 0x00000600 -#define FENCE_SIZE_64M 0x00000700 -#define I915G_FENCE_SIZE_1M 0x00000000 -#define I915G_FENCE_SIZE_2M 0x00000100 -#define I915G_FENCE_SIZE_4M 0x00000200 -#define I915G_FENCE_SIZE_8M 0x00000300 -#define I915G_FENCE_SIZE_16M 0x00000400 -#define I915G_FENCE_SIZE_32M 0x00000500 -#define I915G_FENCE_SIZE_64M 0x00000600 -#define I915G_FENCE_SIZE_128M 0x00000700 -#define FENCE_PITCH_1 0x00000000 -#define FENCE_PITCH_2 0x00000010 -#define FENCE_PITCH_4 0x00000020 -#define FENCE_PITCH_8 0x00000030 -#define FENCE_PITCH_16 0x00000040 -#define FENCE_PITCH_32 0x00000050 -#define FENCE_PITCH_64 0x00000060 -#define FENCE_VALID 0x00000001 - -#include - -# define MMIO_IN8(base, offset) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) -# define MMIO_IN32(base, offset) \ - read_MMIO_LE32(base, offset) -# define MMIO_OUT8(base, offset, val) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val) -# define MMIO_OUT32(base, offset, val) \ - *(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val) - - - /* Memory mapped register access macros */ -#define INREG8(addr) MMIO_IN8(MMIO, addr) -#define INREG(addr) MMIO_IN32(MMIO, addr) -#define OUTREG8(addr, val) MMIO_OUT8(MMIO, addr, val) -#define OUTREG(addr, val) MMIO_OUT32(MMIO, addr, val) - -#define DSPABASE 0x70184 - -#endif diff --git a/src/mesa/drivers/dri/i915tex/server/intel_dri.c b/src/mesa/drivers/dri/i915tex/server/intel_dri.c deleted file mode 100644 index e49c4214ad..0000000000 --- a/src/mesa/drivers/dri/i915tex/server/intel_dri.c +++ /dev/null @@ -1,1306 +0,0 @@ -/** - * \file server/intel_dri.c - * \brief File to perform the device-specific initialization tasks typically - * done in the X server. - * - * Here they are converted to run in the client (or perhaps a standalone - * process), and to work with the frame buffer device rather than the X - * server infrastructure. - * - * Copyright (C) 2006 Dave Airlie (airlied@linux.ie) - - Permission is hereby granted, free of charge, to any person obtaining a - copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sub license, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial portions - of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR - ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. -*/ - -#include -#include -#include -#include -#include - -#include "driver.h" -#include "drm.h" - -#include "intel.h" -#include "i830_dri.h" - -#include "memops.h" -#include "pciaccess.h" - -static size_t drm_page_size; -static int nextTile = 0; -#define xf86DrvMsg(...) do {} while(0) - -static const int pitches[] = { - 128 * 8, - 128 * 16, - 128 * 32, - 128 * 64, - 0 -}; - -static Bool I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea); - -static unsigned long -GetBestTileAlignment(unsigned long size) -{ - unsigned long i; - - for (i = KB(512); i < size; i <<= 1) - ; - - if (i > MB(64)) - i = MB(64); - - return i; -} - -static void SetFenceRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - unsigned char *MMIO = ctx->MMIOAddress; - - for (i = 0; i < 8; i++) { - OUTREG(FENCE + i * 4, pI830->Fence[i]); - // if (I810_DEBUG & DEBUG_VERBOSE_VGA) - fprintf(stderr,"Fence Register : %x\n", pI830->Fence[i]); - } -} - -/* Tiled memory is good... really, really good... - * - * Need to make it less likely that we miss out on this - probably - * need to move the frontbuffer away from the 'guarenteed' alignment - * of the first memory segment, or perhaps allocate a discontigous - * framebuffer to get more alignment 'sweet spots'. - */ -static void -SetFence(const DRIDriverContext *ctx, I830Rec *pI830, - int nr, unsigned int start, unsigned int pitch, - unsigned int size) -{ - unsigned int val; - unsigned int fence_mask = 0; - unsigned int fence_pitch; - - if (nr < 0 || nr > 7) { - fprintf(stderr, - "SetFence: fence %d out of range\n",nr); - return; - } - - pI830->Fence[nr] = 0; - - if (IS_I9XX(pI830)) - fence_mask = ~I915G_FENCE_START_MASK; - else - fence_mask = ~I830_FENCE_START_MASK; - - if (start & fence_mask) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not %s aligned\n", - nr, start, (IS_I9XX(pI830)) ? "1MB" : "512k"); - return; - } - - if (start % size) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not size (%dk) aligned\n", - nr, start, size / 1024); - return; - } - - if (pitch & 127) { - fprintf(stderr, - "SetFence: %d: pitch (%d) not a multiple of 128 bytes\n", - nr, pitch); - return; - } - - val = (start | FENCE_X_MAJOR | FENCE_VALID); - - if (IS_I9XX(pI830)) { - switch (size) { - case MB(1): - val |= I915G_FENCE_SIZE_1M; - break; - case MB(2): - val |= I915G_FENCE_SIZE_2M; - break; - case MB(4): - val |= I915G_FENCE_SIZE_4M; - break; - case MB(8): - val |= I915G_FENCE_SIZE_8M; - break; - case MB(16): - val |= I915G_FENCE_SIZE_16M; - break; - case MB(32): - val |= I915G_FENCE_SIZE_32M; - break; - case MB(64): - val |= I915G_FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } else { - switch (size) { - case KB(512): - val |= FENCE_SIZE_512K; - break; - case MB(1): - val |= FENCE_SIZE_1M; - break; - case MB(2): - val |= FENCE_SIZE_2M; - break; - case MB(4): - val |= FENCE_SIZE_4M; - break; - case MB(8): - val |= FENCE_SIZE_8M; - break; - case MB(16): - val |= FENCE_SIZE_16M; - break; - case MB(32): - val |= FENCE_SIZE_32M; - break; - case MB(64): - val |= FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } - - if (IS_I9XX(pI830)) - fence_pitch = pitch / 512; - else - fence_pitch = pitch / 128; - - switch (fence_pitch) { - case 1: - val |= FENCE_PITCH_1; - break; - case 2: - val |= FENCE_PITCH_2; - break; - case 4: - val |= FENCE_PITCH_4; - break; - case 8: - val |= FENCE_PITCH_8; - break; - case 16: - val |= FENCE_PITCH_16; - break; - case 32: - val |= FENCE_PITCH_32; - break; - case 64: - val |= FENCE_PITCH_64; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal pitch (%d)\n", nr, pitch); - return; - } - - pI830->Fence[nr] = val; -} - -static Bool -MakeTiles(const DRIDriverContext *ctx, I830Rec *pI830, I830MemRange *pMem) -{ - int pitch, ntiles, i; - - pitch = pMem->Pitch * ctx->cpp; - /* - * Simply try to break the region up into at most four pieces of size - * equal to the alignment. - */ - ntiles = ROUND_TO(pMem->Size, pMem->Alignment) / pMem->Alignment; - if (ntiles >= 4) { - return FALSE; - } - - for (i = 0; i < ntiles; i++, nextTile++) { - SetFence(ctx, pI830, nextTile, pMem->Start + i * pMem->Alignment, - pitch, pMem->Alignment); - } - return TRUE; -} - -static void I830SetupMemoryTiling(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - - /* Clear out */ - for (i = 0; i < 8; i++) - pI830->Fence[i] = 0; - - nextTile = 0; - - if (pI830->BackBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->BackBuffer))) { - fprintf(stderr, - "Activating tiled memory for the back buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the back buffer.\n"); - pI830->allowPageFlip = FALSE; - } - } - - if (pI830->DepthBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->DepthBuffer))) { - fprintf(stderr, - "Activating tiled memory for the depth buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the depth buffer.\n"); - } - } - - return; -} - -static int I830DetectMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - struct pci_device host_bridge, ig_dev; - uint32_t gmch_ctrl; - int memsize = 0; - int range; - uint32_t aper_size; - uint32_t membase2 = 0; - - memset(&host_bridge, 0, sizeof(host_bridge)); - memset(&ig_dev, 0, sizeof(ig_dev)); - - ig_dev.dev = 2; - - pci_device_cfg_read_u32(&host_bridge, &gmch_ctrl, I830_GMCH_CTRL); - - if (IS_I830(pI830) || IS_845G(pI830)) { - if ((gmch_ctrl & I830_GMCH_MEM_MASK) == I830_GMCH_MEM_128M) { - aper_size = 0x80000000; - } else { - aper_size = 0x40000000; - } - } else { - if (IS_I9XX(pI830)) { - int ret; - ret = pci_device_cfg_read_u32(&ig_dev, &membase2, 0x18); - if (membase2 & 0x08000000) - aper_size = 0x8000000; - else - aper_size = 0x10000000; - - fprintf(stderr,"aper size is %08X %08x %d\n", aper_size, membase2, ret); - } else - aper_size = 0x8000000; - } - - pI830->aper_size = aper_size; - - - /* We need to reduce the stolen size, by the GTT and the popup. - * The GTT varying according the the FbMapSize and the popup is 4KB */ - range = (ctx->shared.fbSize / (1024*1024)) + 4; - - if (IS_I85X(pI830) || IS_I865G(pI830) || IS_I9XX(pI830)) { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I855_GMCH_GMS_STOLEN_1M: - memsize = MB(1) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_4M: - memsize = MB(4) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_8M: - memsize = MB(8) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_16M: - memsize = MB(16) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_32M: - memsize = MB(32) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_48M: - if (IS_I9XX(pI830)) - memsize = MB(48) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_64M: - if (IS_I9XX(pI830)) - memsize = MB(64) - KB(range); - break; - } - } else { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I830_GMCH_GMS_STOLEN_512: - memsize = KB(512) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_1024: - memsize = MB(1) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_8192: - memsize = MB(8) - KB(range); - break; - case I830_GMCH_GMS_LOCAL: - memsize = 0; - xf86DrvMsg(pScrn->scrnIndex, X_WARNING, - "Local memory found, but won't be used.\n"); - break; - } - } - if (memsize > 0) { - fprintf(stderr, - "detected %d kB stolen memory.\n", memsize / 1024); - } else { - fprintf(stderr, - "no video memory detected.\n"); - } - return memsize; -} - -static int AgpInit(const DRIDriverContext *ctx, I830Rec *info) -{ - unsigned long mode = 0x4; - - if (drmAgpAcquire(ctx->drmFD) < 0) { - fprintf(stderr, "[gart] AGP not available\n"); - return 0; - } - - if (drmAgpEnable(ctx->drmFD, mode) < 0) { - fprintf(stderr, "[gart] AGP not enabled\n"); - drmAgpRelease(ctx->drmFD); - return 0; - } - else - fprintf(stderr, "[gart] AGP enabled at %dx\n", ctx->agpmode); - - return 1; -} - -/* - * Allocate memory from the given pool. Grow the pool if needed and if - * possible. - */ -static unsigned long -AllocFromPool(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, - long size, unsigned long alignment, int flags) -{ - long needed, start, end; - - if (!result || !pool || !size) - return 0; - - /* Calculate how much space is needed. */ - if (alignment <= GTT_PAGE_SIZE) - needed = size; - else { - start = ROUND_TO(pool->Free.Start, alignment); - end = ROUND_TO(start + size, alignment); - needed = end - pool->Free.Start; - } - if (needed > pool->Free.Size) { - return 0; - } - - result->Start = ROUND_TO(pool->Free.Start, alignment); - pool->Free.Start += needed; - result->End = pool->Free.Start; - - pool->Free.Size = pool->Free.End - pool->Free.Start; - result->Size = result->End - result->Start; - result->Pool = pool; - result->Alignment = alignment; - return needed; -} - -static unsigned long AllocFromAGP(const DRIDriverContext *ctx, I830Rec *pI830, long size, unsigned long alignment, I830MemRange *result) -{ - unsigned long start, end; - unsigned long newApStart, newApEnd; - int ret; - if (!result || !size) - return 0; - - if (!alignment) - alignment = 4; - - start = ROUND_TO(pI830->MemoryAperture.Start, alignment); - end = ROUND_TO(start + size, alignment); - newApStart = end; - newApEnd = pI830->MemoryAperture.End; - - ret=drmAgpAlloc(ctx->drmFD, size, 0, &(result->Physical), (drm_handle_t *)&(result->Key)); - - if (ret) - { - fprintf(stderr,"drmAgpAlloc failed %d\n", ret); - return 0; - } - pI830->allocatedMemory += size; - pI830->MemoryAperture.Start = newApStart; - pI830->MemoryAperture.End = newApEnd; - pI830->MemoryAperture.Size = newApEnd - newApStart; - // pI830->FreeMemory -= size; - result->Start = start; - result->End = start + size; - result->Size = size; - result->Offset = start; - result->Alignment = alignment; - result->Pool = NULL; - - return size; -} - -unsigned long -I830AllocVidMem(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, long size, - unsigned long alignment, int flags) -{ - unsigned long ret; - - if (!result) - return 0; - - /* Make sure these are initialised. */ - result->Size = 0; - result->Key = -1; - - if (!size) { - return 0; - } - - if (pool->Free.Size < size) { - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - else { - ret = AllocFromPool(ctx, pI830, result, pool, size, alignment, flags); - if (ret == 0) - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - return ret; -} - -static Bool BindAgpRange(const DRIDriverContext *ctx, I830MemRange *mem) -{ - if (!mem) - return FALSE; - - if (mem->Key == -1) - return TRUE; - - return !drmAgpBind(ctx->drmFD, mem->Key, mem->Offset); -} - -/* simple memory allocation routines needed */ -/* put ring buffer in low memory */ -/* need to allocate front, back, depth buffers aligned correctly, - allocate ring buffer, -*/ - -/* */ -static Bool -I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned long size, ret; - unsigned long lines, lineSize, align; - - /* allocate ring buffer */ - memset(pI830->LpRing, 0, sizeof(I830RingBuffer)); - pI830->LpRing->mem.Key = -1; - - size = PRIMARY_RINGBUFFER_SIZE; - - ret = I830AllocVidMem(ctx, pI830, &pI830->LpRing->mem, &pI830->StolenPool, size, 0x1000, 0); - - if (ret != size) - { - fprintf(stderr,"unable to allocate ring buffer %ld\n", ret); - return FALSE; - } - - pI830->LpRing->tail_mask = pI830->LpRing->mem.Size - 1; - - - /* allocate front buffer */ - memset(&(pI830->FrontBuffer), 0, sizeof(pI830->FrontBuffer)); - pI830->FrontBuffer.Key = -1; - pI830->FrontBuffer.Pitch = ctx->shared.virtualWidth; - - align = KB(512); - - lineSize = ctx->shared.virtualWidth * ctx->cpp; - lines = (ctx->shared.virtualHeight + 15) / 16 * 16; - size = lineSize * lines; - size = ROUND_TO_PAGE(size); - - align = GetBestTileAlignment(size); - - ret = I830AllocVidMem(ctx, pI830, &pI830->FrontBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate front buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->BackBuffer), 0, sizeof(pI830->BackBuffer)); - pI830->BackBuffer.Key = -1; - pI830->BackBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->BackBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate back buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->DepthBuffer), 0, sizeof(pI830->DepthBuffer)); - pI830->DepthBuffer.Key = -1; - pI830->DepthBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->DepthBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate depth buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->ContextMem), 0, sizeof(pI830->ContextMem)); - pI830->ContextMem.Key = -1; - size = KB(32); - - ret = I830AllocVidMem(ctx, pI830, &pI830->ContextMem, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate context buffer %ld\n", ret); - return FALSE; - } - -#if 0 - memset(&(pI830->TexMem), 0, sizeof(pI830->TexMem)); - pI830->TexMem.Key = -1; - - size = 32768 * 1024; - ret = AllocFromAGP(ctx, pI830, size, align, &pI830->TexMem); - if (ret < size) - { - fprintf(stderr,"unable to allocate texture memory %ld\n", ret); - return FALSE; - } -#endif - - return TRUE; -} - -static Bool -I830BindMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - if (!BindAgpRange(ctx, &pI830->LpRing->mem)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->FrontBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->BackBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->DepthBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->ContextMem)) - return FALSE; -#if 0 - if (!BindAgpRange(ctx, &pI830->TexMem)) - return FALSE; -#endif - return TRUE; -} - -static void SetupDRIMM(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned long aperEnd = ROUND_DOWN_TO(pI830->aper_size, GTT_PAGE_SIZE) / GTT_PAGE_SIZE; - unsigned long aperStart = ROUND_TO(pI830->aper_size - KB(32768), GTT_PAGE_SIZE) / GTT_PAGE_SIZE; - - fprintf(stderr, "aper size is %08X\n", ctx->shared.fbSize); - if (drmMMInit(ctx->drmFD, aperStart, aperEnd - aperStart, DRM_BO_MEM_TT)) { - fprintf(stderr, - "DRM MM Initialization Failed\n"); - } else { - fprintf(stderr, - "DRM MM Initialized at offset 0x%lx length %d page\n", aperStart, aperEnd-aperStart); - } - -} - -static Bool -I830CleanupDma(const DRIDriverContext *ctx) -{ - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_CLEANUP_DMA; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, "I830 Dma Cleanup Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830InitDma(const DRIDriverContext *ctx, I830Rec *pI830) -{ - I830RingBuffer *ring = pI830->LpRing; - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_INIT_DMA; - - info.ring_start = ring->mem.Start + pI830->LinearAddr; - info.ring_end = ring->mem.End + pI830->LinearAddr; - info.ring_size = ring->mem.Size; - - info.mmio_offset = (unsigned int)ctx->MMIOStart; - - info.sarea_priv_offset = sizeof(drm_sarea_t); - - info.front_offset = pI830->FrontBuffer.Start; - info.back_offset = pI830->BackBuffer.Start; - info.depth_offset = pI830->DepthBuffer.Start; - info.w = ctx->shared.virtualWidth; - info.h = ctx->shared.virtualHeight; - info.pitch = ctx->shared.virtualWidth; - info.back_pitch = pI830->BackBuffer.Pitch; - info.depth_pitch = pI830->DepthBuffer.Pitch; - info.cpp = ctx->cpp; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, - "I830 Dma Initialization Failed\n"); - return FALSE; - } - - return TRUE; -} - -static int I830CheckDRMVersion( const DRIDriverContext *ctx, - I830Rec *pI830 ) -{ - drmVersionPtr version; - - version = drmGetVersion(ctx->drmFD); - - if (version) { - int req_minor, req_patch; - - req_minor = 4; - req_patch = 0; - - if (version->version_major != 1 || - version->version_minor < req_minor || - (version->version_minor == req_minor && - version->version_patchlevel < req_patch)) { - /* Incompatible drm version */ - fprintf(stderr, - "[dri] I830DRIScreenInit failed because of a version " - "mismatch.\n" - "[dri] i915.o kernel module version is %d.%d.%d " - "but version 1.%d.%d or newer is needed.\n" - "[dri] Disabling DRI.\n", - version->version_major, - version->version_minor, - version->version_patchlevel, - req_minor, - req_patch); - drmFreeVersion(version); - return 0; - } - - pI830->drmMinor = version->version_minor; - drmFreeVersion(version); - } - return 1; -} - -static void -I830SetRingRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned int itemp; - unsigned char *MMIO = ctx->MMIOAddress; - - OUTREG(LP_RING + RING_LEN, 0); - OUTREG(LP_RING + RING_TAIL, 0); - OUTREG(LP_RING + RING_HEAD, 0); - - if ((long)(pI830->LpRing->mem.Start & I830_RING_START_MASK) != - pI830->LpRing->mem.Start) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer start (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Start, I830_RING_START_MASK); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = pI830->LpRing->mem.Start & I830_RING_START_MASK; - OUTREG(LP_RING + RING_START, itemp); - - if (((pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES) != - pI830->LpRing->mem.Size - 4096) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer size - 4096 (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Size - 4096, - I830_RING_NR_PAGES); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = (pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES; - itemp |= (RING_NO_REPORT | RING_VALID); - OUTREG(LP_RING + RING_LEN, itemp); - - pI830->LpRing->head = INREG(LP_RING + RING_HEAD) & I830_HEAD_MASK; - pI830->LpRing->tail = INREG(LP_RING + RING_TAIL); - pI830->LpRing->space = pI830->LpRing->head - (pI830->LpRing->tail + 8); - if (pI830->LpRing->space < 0) - pI830->LpRing->space += pI830->LpRing->mem.Size; - - SetFenceRegs(ctx, pI830); - - /* RESET THE DISPLAY PIPE TO POINT TO THE FRONTBUFFER - hacky - hacky hacky */ - OUTREG(DSPABASE, pI830->FrontBuffer.Start + pI830->LinearAddr); - -} - -static Bool -I830SetParam(const DRIDriverContext *ctx, int param, int value) -{ - drmI830SetParam sp; - - memset(&sp, 0, sizeof(sp)); - sp.param = param; - sp.value = value; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_SETPARAM, &sp, sizeof(sp))) { - fprintf(stderr, "I830 SetParam Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - fprintf(stderr, - "[drm] Mapping front buffer\n"); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->front_offset + pI830->LinearAddr), - sarea->front_size, - DRM_FRAME_BUFFER, /*DRM_AGP,*/ - 0, - &sarea->front_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(front_handle) failed. Disabling DRI\n"); - return FALSE; - } - ctx->shared.hFrameBuffer = sarea->front_handle; - ctx->shared.fbSize = sarea->front_size; - fprintf(stderr, "[drm] Front Buffer = 0x%08x\n", - sarea->front_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->back_offset), - sarea->back_size, DRM_AGP, 0, - &sarea->back_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(back_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Back Buffer = 0x%08x\n", - sarea->back_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->depth_offset, - sarea->depth_size, DRM_AGP, 0, - &sarea->depth_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(depth_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Depth Buffer = 0x%08x\n", - sarea->depth_handle); - -#if 0 - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->tex_offset, - sarea->tex_size, DRM_AGP, 0, - &sarea->tex_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(tex_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] textures = 0x%08x\n", - sarea->tex_handle); -#endif - return TRUE; -} - - -static void -I830DRIUnmapScreenRegions(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ -#if 1 - if (sarea->front_handle) { - drmRmMap(ctx->drmFD, sarea->front_handle); - sarea->front_handle = 0; - } -#endif - if (sarea->back_handle) { - drmRmMap(ctx->drmFD, sarea->back_handle); - sarea->back_handle = 0; - } - if (sarea->depth_handle) { - drmRmMap(ctx->drmFD, sarea->depth_handle); - sarea->depth_handle = 0; - } - if (sarea->tex_handle) { - drmRmMap(ctx->drmFD, sarea->tex_handle); - sarea->tex_handle = 0; - } -} - -static Bool -I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - if (drmAddMap(ctx->drmFD, - (drm_handle_t)pI830->LpRing->mem.Start, - pI830->LpRing->mem.Size, DRM_AGP, 0, - &pI830->ring_map) < 0) { - fprintf(stderr, - "[drm] drmAddMap(ring_map) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] ring buffer = 0x%08x\n", - pI830->ring_map); - - if (I830InitDma(ctx, pI830) == FALSE) { - return FALSE; - } - - /* init to zero to be safe */ - - I830DRIMapScreenRegions(ctx, pI830, sarea); - SetupDRIMM(ctx, pI830); - - if (ctx->pciDevice != PCI_CHIP_845_G && - ctx->pciDevice != PCI_CHIP_I830_M) { - I830SetParam(ctx, I830_SETPARAM_USE_MI_BATCHBUFFER_START, 1 ); - } - - /* Okay now initialize the dma engine */ - { - pI830->irq = drmGetInterruptFromBusID(ctx->drmFD, - ctx->pciBus, - ctx->pciDevice, - ctx->pciFunc); - - if (drmCtlInstHandler(ctx->drmFD, pI830->irq)) { - fprintf(stderr, - "[drm] failure adding irq handler\n"); - pI830->irq = 0; - return FALSE; - } - else - fprintf(stderr, - "[drm] dma control initialized, using IRQ %d\n", - pI830->irq); - } - - fprintf(stderr, "[dri] visual configs initialized\n"); - - return TRUE; -} - -static Bool -I830ClearScreen(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - /* need to drmMap front and back buffers and zero them */ - drmAddress map_addr; - int ret; - - ret = drmMap(ctx->drmFD, - sarea->front_handle, - sarea->front_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map front buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->front_size); - drmUnmap(map_addr, sarea->front_size); - - - ret = drmMap(ctx->drmFD, - sarea->back_handle, - sarea->back_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map back buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->back_size); - drmUnmap(map_addr, sarea->back_size); - - return TRUE; -} - -static Bool -I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) - -{ - I830DRIPtr pI830DRI; - drmI830Sarea *pSAREAPriv; - int err; - - drm_page_size = getpagesize(); - - pI830->registerSize = ctx->MMIOSize; - /* This is a hack for now. We have to have more than a 4k page here - * because of the size of the state. However, the state should be - * in a per-context mapping. This will be added in the Mesa 3.5 port - * of the I830 driver. - */ - ctx->shared.SAREASize = SAREA_MAX; - - /* Note that drmOpen will try to load the kernel module, if needed. */ - ctx->drmFD = drmOpen("i915", NULL ); - if (ctx->drmFD < 0) { - fprintf(stderr, "[drm] drmOpen failed\n"); - return 0; - } - - if ((err = drmSetBusid(ctx->drmFD, ctx->pciBusID)) < 0) { - fprintf(stderr, "[drm] drmSetBusid failed (%d, %s), %s\n", - ctx->drmFD, ctx->pciBusID, strerror(-err)); - return 0; - } - - if (drmAddMap( ctx->drmFD, - 0, - ctx->shared.SAREASize, - DRM_SHM, - DRM_CONTAINS_LOCK, - &ctx->shared.hSAREA) < 0) - { - fprintf(stderr, "[drm] drmAddMap failed\n"); - return 0; - } - - fprintf(stderr, "[drm] added %d byte SAREA at 0x%08x\n", - ctx->shared.SAREASize, ctx->shared.hSAREA); - - if (drmMap( ctx->drmFD, - ctx->shared.hSAREA, - ctx->shared.SAREASize, - (drmAddressPtr)(&ctx->pSAREA)) < 0) - { - fprintf(stderr, "[drm] drmMap failed\n"); - return 0; - - } - - memset(ctx->pSAREA, 0, ctx->shared.SAREASize); - fprintf(stderr, "[drm] mapped SAREA 0x%08x to %p, size %d\n", - ctx->shared.hSAREA, ctx->pSAREA, ctx->shared.SAREASize); - - - if (drmAddMap(ctx->drmFD, - ctx->MMIOStart, - ctx->MMIOSize, - DRM_REGISTERS, - DRM_READ_ONLY, - &pI830->registerHandle) < 0) { - fprintf(stderr, "[drm] drmAddMap mmio failed\n"); - return 0; - } - fprintf(stderr, - "[drm] register handle = 0x%08x\n", pI830->registerHandle); - - - if (!I830CheckDRMVersion(ctx, pI830)) { - return FALSE; - } - - /* Create a 'server' context so we can grab the lock for - * initialization ioctls. - */ - if ((err = drmCreateContext(ctx->drmFD, &ctx->serverContext)) != 0) { - fprintf(stderr, "%s: drmCreateContext failed %d\n", __FUNCTION__, err); - return 0; - } - - DRM_LOCK(ctx->drmFD, ctx->pSAREA, ctx->serverContext, 0); - - /* Initialize the SAREA private data structure */ - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - memset(pSAREAPriv, 0, sizeof(*pSAREAPriv)); - - pI830->StolenMemory.Size = I830DetectMemory(ctx, pI830); - pI830->StolenMemory.Start = 0; - pI830->StolenMemory.End = pI830->StolenMemory.Size; - - pI830->MemoryAperture.Start = pI830->StolenMemory.End; - pI830->MemoryAperture.End = KB(40000); - pI830->MemoryAperture.Size = pI830->MemoryAperture.End - pI830->MemoryAperture.Start; - - pI830->StolenPool.Fixed = pI830->StolenMemory; - pI830->StolenPool.Total = pI830->StolenMemory; - pI830->StolenPool.Free = pI830->StolenPool.Total; - pI830->FreeMemory = pI830->StolenPool.Total.Size; - - if (!AgpInit(ctx, pI830)) - return FALSE; - - if (I830AllocateMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - if (I830BindMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - pSAREAPriv->rotated_offset = -1; - pSAREAPriv->rotated_size = 0; - pSAREAPriv->rotated_pitch = ctx->shared.virtualWidth; - - pSAREAPriv->front_offset = pI830->FrontBuffer.Start; - pSAREAPriv->front_size = pI830->FrontBuffer.Size; - pSAREAPriv->width = ctx->shared.virtualWidth; - pSAREAPriv->height = ctx->shared.virtualHeight; - pSAREAPriv->pitch = ctx->shared.virtualWidth; - pSAREAPriv->virtualX = ctx->shared.virtualWidth; - pSAREAPriv->virtualY = ctx->shared.virtualHeight; - pSAREAPriv->back_offset = pI830->BackBuffer.Start; - pSAREAPriv->back_size = pI830->BackBuffer.Size; - pSAREAPriv->depth_offset = pI830->DepthBuffer.Start; - pSAREAPriv->depth_size = pI830->DepthBuffer.Size; -#if 0 - pSAREAPriv->tex_offset = pI830->TexMem.Start; - pSAREAPriv->tex_size = pI830->TexMem.Size; -#endif - pSAREAPriv->log_tex_granularity = pI830->TexGranularity; - - ctx->driverClientMsg = malloc(sizeof(I830DRIRec)); - ctx->driverClientMsgSize = sizeof(I830DRIRec); - pI830DRI = (I830DRIPtr)ctx->driverClientMsg; - pI830DRI->deviceID = pI830->Chipset; - pI830DRI->regsSize = I830_REG_SIZE; - pI830DRI->width = ctx->shared.virtualWidth; - pI830DRI->height = ctx->shared.virtualHeight; - pI830DRI->mem = ctx->shared.fbSize; - pI830DRI->cpp = ctx->cpp; - - pI830DRI->bitsPerPixel = ctx->bpp; - pI830DRI->sarea_priv_offset = sizeof(drm_sarea_t); - - err = I830DRIDoMappings(ctx, pI830, pSAREAPriv); - if (err == FALSE) - return FALSE; - - I830SetupMemoryTiling(ctx, pI830); - - /* Quick hack to clear the front & back buffers. Could also use - * the clear ioctl to do this, but would need to setup hw state - * first. - */ - I830ClearScreen(ctx, pI830, pSAREAPriv); - - I830SetRingRegs(ctx, pI830); - - return TRUE; -} - - -/** - * \brief Validate the fbdev mode. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Saves some registers and returns 1. - * - * \sa radeonValidateMode(). - */ -static int i830ValidateMode( const DRIDriverContext *ctx ) -{ - return 1; -} - -/** - * \brief Examine mode returned by fbdev. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Restores registers that fbdev has clobbered and returns 1. - * - * \sa i810ValidateMode(). - */ -static int i830PostValidateMode( const DRIDriverContext *ctx ) -{ - I830Rec *pI830 = ctx->driverPrivate; - - I830SetRingRegs(ctx, pI830); - return 1; -} - - -/** - * \brief Initialize the framebuffer device mode - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Fills in \p info with some default values and some information from \p ctx - * and then calls I810ScreenInit() for the screen initialization. - * - * Before exiting clears the framebuffer memory accessing it directly. - */ -static int i830InitFBDev( DRIDriverContext *ctx ) -{ - I830Rec *pI830 = calloc(1, sizeof(I830Rec)); - int i; - - { - int dummy = ctx->shared.virtualWidth; - - switch (ctx->bpp / 8) { - case 1: dummy = (ctx->shared.virtualWidth + 127) & ~127; break; - case 2: dummy = (ctx->shared.virtualWidth + 31) & ~31; break; - case 3: - case 4: dummy = (ctx->shared.virtualWidth + 15) & ~15; break; - } - - ctx->shared.virtualWidth = dummy; - ctx->shared.Width = ctx->shared.virtualWidth; - } - - - for (i = 0; pitches[i] != 0; i++) { - if (pitches[i] >= ctx->shared.virtualWidth) { - ctx->shared.virtualWidth = pitches[i]; - break; - } - } - - ctx->driverPrivate = (void *)pI830; - - pI830->LpRing = calloc(1, sizeof(I830RingBuffer)); - pI830->Chipset = ctx->chipset; - pI830->LinearAddr = ctx->FBStart; - - if (!I830ScreenInit( ctx, pI830 )) - return 0; - - - return 1; -} - - -/** - * \brief The screen is being closed, so clean up any state and free any - * resources used by the DRI. - * - * \param ctx display handle. - * - * Unmaps the SAREA, closes the DRM device file descriptor and frees the driver - * private data. - */ -static void i830HaltFBDev( DRIDriverContext *ctx ) -{ - drmI830Sarea *pSAREAPriv; - I830Rec *pI830 = ctx->driverPrivate; - - if (pI830->irq) { - drmCtlUninstHandler(ctx->drmFD); - pI830->irq = 0; } - - I830CleanupDma(ctx); - - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - - I830DRIUnmapScreenRegions(ctx, pI830, pSAREAPriv); - drmUnmap( ctx->pSAREA, ctx->shared.SAREASize ); - drmClose(ctx->drmFD); - - if (ctx->driverPrivate) { - free(ctx->driverPrivate); - ctx->driverPrivate = 0; - } -} - - -extern void i810NotifyFocus( int ); - -/** - * \brief Exported driver interface for Mini GLX. - * - * \sa DRIDriverRec. - */ -const struct DRIDriverRec __driDriver = { - i830ValidateMode, - i830PostValidateMode, - i830InitFBDev, - i830HaltFBDev, - NULL,//I830EngineShutdown, - NULL, //I830EngineRestore, -#ifndef _EMBEDDED - 0, -#else - i810NotifyFocus, -#endif -}; -- cgit v1.2.3 From eaa82d33bffda1fb9a787b92b5f1a2d15c99f045 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 12 Dec 2007 16:04:58 +0000 Subject: Add src/mesa to include path. --- src/glx/x11/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/glx/x11/Makefile b/src/glx/x11/Makefile index c7b5874faa..5f74fcff06 100644 --- a/src/glx/x11/Makefile +++ b/src/glx/x11/Makefile @@ -44,6 +44,7 @@ OBJECTS = $(SOURCES:.c=.o) $(MESA_GLAPI_OBJECTS) INCLUDES = -I. \ -I$(TOP)/include \ -I$(TOP)/include/GL/internal \ + -I$(TOP)/src/mesa \ -I$(TOP)/src/mesa/main \ -I$(TOP)/src/mesa/glapi \ $(LIBDRM_CFLAGS) \ -- cgit v1.2.3 From a308a3e3f8f3eb6ad459c79c36a988c1e33c3f56 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 12 Dec 2007 12:26:00 -0500 Subject: i965: compile vertex shaders and emit hardware instructions impossible to test because fragment shaders haven't been done yet and it crahses in the fragment shading code. but given that i'm infallible (and highly dillusional) i'm sure it's correct. --- src/mesa/pipe/i965simple/brw_vs_emit.c | 478 +++++++++++++++++++-------------- 1 file changed, 275 insertions(+), 203 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index 4121eae632..59459d4200 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -732,11 +732,11 @@ static struct brw_reg get_arg( struct brw_vs_compile *c, static struct brw_reg get_dst( struct brw_vs_compile *c, - struct tgsi_dst_register dst ) + const struct tgsi_dst_register *dst ) { - struct brw_reg reg = get_reg(c, dst.File, dst.Index); + struct brw_reg reg = get_reg(c, dst->File, dst->Index); - reg.dw1.bits.writemask = dst.WriteMask; + reg.dw1.bits.writemask = dst->WriteMask; return reg; } @@ -950,14 +950,14 @@ post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst ) { struct tgsi_parse_context parse; const struct tgsi_token *tokens = c->vp->program.tokens; - struct brw_instruction *brw_inst1, *brw_inst2; - const struct tgsi_full_instruction *inst1, *inst2; - int offset; tgsi_parse_init(&parse, tokens); while (!tgsi_parse_end_of_tokens(&parse)) { tgsi_parse_token(&parse); if (parse.FullToken.Token.Type == TGSI_TOKEN_TYPE_INSTRUCTION) { #if 0 + struct brw_instruction *brw_inst1, *brw_inst2; + const struct tgsi_full_instruction *inst1, *inst2; + int offset; inst1 = &parse.FullToken.FullInstruction; brw_inst1 = inst1->Data; switch (inst1->Opcode) { @@ -982,29 +982,277 @@ post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst ) tgsi_parse_free(&parse); } +static void process_declaration(const struct tgsi_full_declaration *decl, + struct brw_prog_info *info) +{ + switch(decl->Declaration.File) { + case TGSI_FILE_CONSTANT: { + } + break; + case TGSI_FILE_INPUT: { + } + break; + case TGSI_FILE_OUTPUT: { + if (decl->Declaration.Semantic) { + int idx = 0; + if (decl->Declaration.Declare == TGSI_DECLARE_MASK) { + printf("DECLARATION MASK = %d\n", + decl->u.DeclarationMask.Mask); + assert(0); + } else { //range + idx = decl->u.DeclarationRange.First; + } + switch (decl->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: { + info->pos_idx = idx; + } + break; + case TGSI_SEMANTIC_COLOR: + break; + case TGSI_SEMANTIC_BCOLOR: + break; + case TGSI_SEMANTIC_FOG: + break; + case TGSI_SEMANTIC_PSIZE: { + info->writes_psize = TRUE; + info->psize_idx = idx; + } + break; + case TGSI_SEMANTIC_GENERIC: + break; + } + } + } + break; + case TGSI_FILE_TEMPORARY: { + info->num_temps++; + } + break; + case TGSI_FILE_SAMPLER: { + } + break; + case TGSI_FILE_ADDRESS: { + info->num_addrs++; + } + break; + case TGSI_FILE_IMMEDIATE: { + } + break; + case TGSI_FILE_NULL: { + } + break; + } +} + +static void process_instruction(struct brw_vs_compile *c, + struct tgsi_full_instruction *inst, + struct brw_prog_info *info) +{ + struct brw_reg args[3], dst; + struct brw_compile *p = &c->func; + struct brw_indirect stack_index = brw_indirect(0, 0); + unsigned i; + unsigned index; + unsigned file; + /*FIXME: might not be the only one*/ + const struct tgsi_dst_register *dst_reg = &inst->FullDstRegisters[0].DstRegister; + /* + struct brw_instruction *if_inst[MAX_IFSN]; + unsigned insn, if_insn = 0; + */ + + /* Get argument regs. SWZ is special and does this itself. */ + if (inst->Instruction.Opcode != TGSI_OPCODE_SWZ) + for (i = 0; i < 3; i++) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + index = src->SrcRegister.Index; + file = src->SrcRegister.File; + if (file == TGSI_FILE_OUTPUT&&c->output_regs[index].used_in_src) + args[i] = c->output_regs[index].reg; + else + args[i] = get_arg(c, &src->SrcRegister); + } + + /* Get dest regs. Note that it is possible for a reg to be both + * dst and arg, given the static allocation of registers. So + * care needs to be taken emitting multi-operation instructions. + */ + index = dst_reg->Index; + file = dst_reg->File; + if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src) + dst = c->output_regs[index].reg; + else + dst = get_dst(c, dst_reg); + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + brw_MOV(p, dst, brw_abs(args[0])); + break; + case TGSI_OPCODE_ADD: + brw_ADD(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DP3: + brw_DP3(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DP4: + brw_DP4(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DPH: + brw_DPH(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DST: + unalias2(c, dst, args[0], args[1], emit_dst_noalias); + break; + case TGSI_OPCODE_EXP: + unalias1(c, dst, args[0], emit_exp_noalias); + break; + case TGSI_OPCODE_EX2: + emit_math1(c, BRW_MATH_FUNCTION_EXP, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_ARL: + emit_arl(c, dst, args[0]); + break; + case TGSI_OPCODE_FLR: + brw_RNDD(p, dst, args[0]); + break; + case TGSI_OPCODE_FRC: + brw_FRC(p, dst, args[0]); + break; + case TGSI_OPCODE_LOG: + unalias1(c, dst, args[0], emit_log_noalias); + break; + case TGSI_OPCODE_LG2: + emit_math1(c, BRW_MATH_FUNCTION_LOG, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_LIT: + unalias1(c, dst, args[0], emit_lit_noalias); + break; + case TGSI_OPCODE_MAD: + brw_MOV(p, brw_acc_reg(), args[2]); + brw_MAC(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MAX: + emit_max(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MIN: + emit_min(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MOV: +#if 0 + case TGSI_OPCODE_SWZ: + /* The args[0] value can't be used here as it won't have + * correctly encoded the full swizzle: + */ + emit_swz(c, dst, inst->SrcReg[0] ); +#endif + brw_MOV(p, dst, args[0]); + break; + case TGSI_OPCODE_MUL: + brw_MUL(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_POW: + emit_math2(c, BRW_MATH_FUNCTION_POW, dst, args[0], args[1], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_RCP: + emit_math1(c, BRW_MATH_FUNCTION_INV, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_RSQ: + emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + + case TGSI_OPCODE_SEQ: + emit_seq(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SNE: + emit_sne(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SGE: + emit_sge(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SGT: + emit_sgt(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SLT: + emit_slt(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SLE: + emit_sle(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SUB: + brw_ADD(p, dst, args[0], negate(args[1])); + break; + case TGSI_OPCODE_XPD: + emit_xpd(p, dst, args[0], args[1]); + break; +#if 0 + case TGSI_OPCODE_IF: + assert(if_insn < MAX_IFSN); + if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8); + break; + case TGSI_OPCODE_ELSE: + if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]); + break; + case TGSI_OPCODE_ENDIF: + assert(if_insn > 0); + brw_ENDIF(p, if_inst[--if_insn]); + break; + case TGSI_OPCODE_BRA: + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + brw_set_predicate_control_flag_value(p, 0xff); + break; + case TGSI_OPCODE_CAL: + brw_set_access_mode(p, BRW_ALIGN_1); + brw_ADD(p, deref_1uw(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(4)); + inst->Data = &p->store[p->nr_insn]; + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + break; +#endif + case TGSI_OPCODE_RET: + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(-4)); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, brw_ip_reg(), deref_1uw(stack_index, 0)); + brw_set_access_mode(p, BRW_ALIGN_16); + case TGSI_OPCODE_END: + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + break; + case TGSI_OPCODE_BGNSUB: + case TGSI_OPCODE_ENDSUB: + break; + default: + printf("Unsupport opcode %d in vertex shader\n", inst->Instruction.Opcode); + break; + } + + if (dst_reg->File == TGSI_FILE_OUTPUT + && dst_reg->Index != info->pos_idx + && c->output_regs[dst_reg->Index].used_in_src) + brw_MOV(p, get_dst(c, dst_reg), dst); + + release_tmps(c); +} + /* Emit the fragment program instructions here. */ void brw_vs_emit(struct brw_vs_compile *c) { #define MAX_IFSN 32 struct brw_compile *p = &c->func; - unsigned insn, if_insn = 0; struct brw_instruction *end_inst; - struct brw_instruction *if_inst[MAX_IFSN]; - struct brw_indirect stack_index = brw_indirect(0, 0); struct tgsi_parse_context parse; + struct brw_indirect stack_index = brw_indirect(0, 0); const struct tgsi_token *tokens = c->vp->program.tokens; struct brw_prog_info prog_info; unsigned allocated_registers = 0; memset(&prog_info, 0, sizeof(struct brw_prog_info)); - unsigned index; - unsigned file; - brw_set_compression_control(p, BRW_COMPRESSION_NONE); brw_set_access_mode(p, BRW_ALIGN_16); -#if 0 tgsi_parse_init(&parse, tokens); /* Message registers can't be read, so copy the output into GRF register if they are used in source registers */ @@ -1014,11 +1262,13 @@ void brw_vs_emit(struct brw_vs_compile *c) switch (parse.FullToken.Token.Type) { case TGSI_TOKEN_TYPE_INSTRUCTION: { const struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction; - struct prog_src_register *src = &inst->SrcReg[i]; - unsigned index = src->Index; - unsigned file = src->File; - if (file == TGSI_FILE_OUTPUT) - c->output_regs[index].used_in_src = TRUE; + for (i = 0; i < 3; ++i) { + const struct tgsi_src_register *src = &inst->FullSrcRegisters[i].SrcRegister; + unsigned index = src->Index; + unsigned file = src->File; + if (file == TGSI_FILE_OUTPUT) + c->output_regs[index].used_in_src = TRUE; + } } break; default: @@ -1026,7 +1276,7 @@ void brw_vs_emit(struct brw_vs_compile *c) break; } } -#endif + tgsi_parse_free(&parse); tgsi_parse_init(&parse, tokens); @@ -1036,205 +1286,27 @@ void brw_vs_emit(struct brw_vs_compile *c) switch (parse.FullToken.Token.Type) { case TGSI_TOKEN_TYPE_DECLARATION: { struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; - /* FIXME: fill in brw_prog_info based on declarations here */ + process_declaration(decl, &prog_info); } break; case TGSI_TOKEN_TYPE_IMMEDIATE: { - } break; case TGSI_TOKEN_TYPE_INSTRUCTION: { + struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction; if (!allocated_registers) { - /* first instruction after declerations. - * know that we know used vars allocate - * registers */ + /* first instruction (declerations finished). + * now that we know what vars are being used allocate + * registers for them.*/ brw_vs_alloc_regs(c, &prog_info); brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); allocated_registers = 1; } + process_instruction(c, inst, &prog_info); } break; } } -#if 0 - struct prog_instruction *inst = &c->vp->program.Base.Instructions[insn]; - struct brw_reg args[3], dst; - unsigned i; - - /* Get argument regs. SWZ is special and does this itself. - */ - inst->Data = &p->store[p->nr_insn]; - if (inst->Opcode != TGSI_OPCODE_SWZ) - for (i = 0; i < 3; i++) { - struct prog_src_register *src = &inst->SrcReg[i]; - index = src->Index; - file = src->File; - if (file == TGSI_FILE_OUTPUT&&c->output_regs[index].used_in_src) - args[i] = c->output_regs[index].reg; - else - args[i] = get_arg(c, src); - } - - /* Get dest regs. Note that it is possible for a reg to be both - * dst and arg, given the static allocation of registers. So - * care needs to be taken emitting multi-operation instructions. - */ - index = inst->DstReg.Index; - file = inst->DstReg.File; - if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src) - dst = c->output_regs[index].reg; - else - dst = get_dst(c, inst->DstReg); - - switch (inst->Opcode) { - case TGSI_OPCODE_ABS: - brw_MOV(p, dst, brw_abs(args[0])); - break; - case TGSI_OPCODE_ADD: - brw_ADD(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_DP3: - brw_DP3(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_DP4: - brw_DP4(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_DPH: - brw_DPH(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_DST: - unalias2(c, dst, args[0], args[1], emit_dst_noalias); - break; - case TGSI_OPCODE_EXP: - unalias1(c, dst, args[0], emit_exp_noalias); - break; - case TGSI_OPCODE_EX2: - emit_math1(c, BRW_MATH_FUNCTION_EXP, dst, args[0], BRW_MATH_PRECISION_FULL); - break; - case TGSI_OPCODE_ARL: - emit_arl(c, dst, args[0]); - break; - case TGSI_OPCODE_FLR: - brw_RNDD(p, dst, args[0]); - break; - case TGSI_OPCODE_FRC: - brw_FRC(p, dst, args[0]); - break; - case TGSI_OPCODE_LOG: - unalias1(c, dst, args[0], emit_log_noalias); - break; - case TGSI_OPCODE_LG2: - emit_math1(c, BRW_MATH_FUNCTION_LOG, dst, args[0], BRW_MATH_PRECISION_FULL); - break; - case TGSI_OPCODE_LIT: - unalias1(c, dst, args[0], emit_lit_noalias); - break; - case TGSI_OPCODE_MAD: - brw_MOV(p, brw_acc_reg(), args[2]); - brw_MAC(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_MAX: - emit_max(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_MIN: - emit_min(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_MOV: - brw_MOV(p, dst, args[0]); - break; - case TGSI_OPCODE_MUL: - brw_MUL(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_POW: - emit_math2(c, BRW_MATH_FUNCTION_POW, dst, args[0], args[1], BRW_MATH_PRECISION_FULL); - break; - case TGSI_OPCODE_RCP: - emit_math1(c, BRW_MATH_FUNCTION_INV, dst, args[0], BRW_MATH_PRECISION_FULL); - break; - case TGSI_OPCODE_RSQ: - emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, args[0], BRW_MATH_PRECISION_FULL); - break; - - case TGSI_OPCODE_SEQ: - emit_seq(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SNE: - emit_sne(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SGE: - emit_sge(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SGT: - emit_sgt(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SLT: - emit_slt(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SLE: - emit_sle(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SUB: - brw_ADD(p, dst, args[0], negate(args[1])); - break; - case TGSI_OPCODE_SWZ: - /* The args[0] value can't be used here as it won't have - * correctly encoded the full swizzle: - */ - emit_swz(c, dst, inst->SrcReg[0] ); - break; - case TGSI_OPCODE_XPD: - emit_xpd(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_IF: - assert(if_insn < MAX_IFSN); - if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8); - break; - case TGSI_OPCODE_ELSE: - if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]); - break; - case TGSI_OPCODE_ENDIF: - assert(if_insn > 0); - brw_ENDIF(p, if_inst[--if_insn]); - break; - case TGSI_OPCODE_BRA: - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); - brw_set_predicate_control_flag_value(p, 0xff); - break; - case TGSI_OPCODE_CAL: - brw_set_access_mode(p, BRW_ALIGN_1); - brw_ADD(p, deref_1uw(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); - brw_set_access_mode(p, BRW_ALIGN_16); - brw_ADD(p, get_addr_reg(stack_index), - get_addr_reg(stack_index), brw_imm_d(4)); - inst->Data = &p->store[p->nr_insn]; - brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); - break; - case TGSI_OPCODE_RET: - brw_ADD(p, get_addr_reg(stack_index), - get_addr_reg(stack_index), brw_imm_d(-4)); - brw_set_access_mode(p, BRW_ALIGN_1); - brw_MOV(p, brw_ip_reg(), deref_1uw(stack_index, 0)); - brw_set_access_mode(p, BRW_ALIGN_16); - case TGSI_OPCODE_END: - brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); - break; - case TGSI_OPCODE_BGNSUB: - case TGSI_OPCODE_ENDSUB: - break; - default: - printf("Unsupport opcode %d in vertex shader\n", inst->Opcode); - break; - } - - if (inst->DstReg.File == TGSI_FILE_OUTPUT - && inst->DstReg.Index != VERT_RESULT_HPOS - && c->output_regs[inst->DstReg.Index].used_in_src) - brw_MOV(p, get_dst(c, inst->DstReg), dst); - - release_tmps(c); - } -#endif end_inst = &p->store[p->nr_insn]; emit_vertex_write(c, &prog_info); -- cgit v1.2.3 From 77baacbab156cd3859fe50e05c2ff55fe8e27926 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 12:58:04 -0700 Subject: fix some mistakes in the _PIPE_FORMAT_x #defines --- src/mesa/pipe/p_format.h | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index 95a50c18af..b1772b352f 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -150,8 +150,8 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) * Shorthand macro for common format swizzles. */ #define _PIPE_FORMAT_R000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_RG00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_A ) -#define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_RG00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) #define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) #define _PIPE_FORMAT_BGRA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_A ) @@ -366,7 +366,7 @@ static INLINE char *pf_sprint_name( char *str, uint format ) return str; } -static INLINE uint pf_get_component_bits( uint format, uint comp ) +static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) { uint size; @@ -388,7 +388,7 @@ static INLINE uint pf_get_component_bits( uint format, uint comp ) return size << (pf_exp8(format) * 3); } -static INLINE uint pf_get_bits( uint format ) +static INLINE uint pf_get_bits( enum pipe_format format ) { if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { return @@ -408,7 +408,7 @@ static INLINE uint pf_get_bits( uint format ) } } -static INLINE uint pf_get_size( uint format ) { +static INLINE uint pf_get_size( enum pipe_format format ) { assert(pf_get_bits(format) % 8 == 0); return pf_get_bits(format) / 8; } -- cgit v1.2.3 From d7aa44636988941d5d09ba943104033dd1912c80 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 12:58:59 -0700 Subject: Try PIPE_FORMAT_R5G6B5_UNORM for GL_RGB5 request. --- src/mesa/state_tracker/st_format.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 98cc2084f0..287a0054b9 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -413,6 +413,8 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_R3_G3_B2: if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) return PIPE_FORMAT_A1R5G5B5_UNORM; + if (pipe->is_format_supported( pipe, PIPE_FORMAT_R5G6B5_UNORM, surfType )) + return PIPE_FORMAT_R5G6B5_UNORM; return default_rgba_format( pipe, surfType ); case GL_ALPHA: -- cgit v1.2.3 From b22287d0972ab37c822457a31609adfc4fb98573 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 12:59:19 -0700 Subject: update format param type --- src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c index f9ed6253f1..cec3437831 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c @@ -31,9 +31,10 @@ #include "intel_context.h" #include "intel_winsys.h" -#include "pipe/softpipe/sp_winsys.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "pipe/p_format.h" +#include "pipe/softpipe/sp_winsys.h" struct intel_softpipe_winsys { @@ -45,7 +46,8 @@ struct intel_softpipe_winsys { * Return list of surface formats supported by this driver. */ static boolean -intel_is_format_supported(struct softpipe_winsys *sws, uint format) +intel_is_format_supported(struct softpipe_winsys *sws, + enum pipe_format format) { switch(format) { case PIPE_FORMAT_A8R8G8B8_UNORM: -- cgit v1.2.3 From 829099210ba442e7035659e8db3c2e059e059c24 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 12:59:45 -0700 Subject: add PIPE_FORMAT_R5G6B5_UNORM to list of supported screen surface types --- src/mesa/pipe/i915simple/i915_context.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index a08cf5087e..0e3d6d64e4 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -69,6 +69,7 @@ i915_is_format_supported( struct pipe_context *pipe, }; static const enum pipe_format screen_surface_supported[] = { PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, PIPE_FORMAT_NONE /* list terminator */ }; const enum pipe_format *list; -- cgit v1.2.3 From 69292a0067ef77cde607183f1719bae61ef0f3a5 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 13:08:00 -0700 Subject: remove old assertion --- src/mesa/pipe/i915simple/i915_surface.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index f93a75b0f0..3db4093d22 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -240,7 +240,6 @@ i915_get_tex_surface(struct pipe_context *pipe, ps = pipe->winsys->surface_alloc(pipe->winsys); if (ps) { - assert(ps->format); assert(ps->refcount); pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); ps->format = pt->format; -- cgit v1.2.3 From 20e6a9b86d715c38f64828d6a306c048571499a6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 12 Dec 2007 17:22:51 +0000 Subject: 965: delete dead files --- src/mesa/pipe/i965simple/Makefile | 1 - src/mesa/pipe/i965simple/brw_metaops.c | 538 --------------------------------- src/mesa/pipe/i965simple/brw_program.c | 130 -------- 3 files changed, 669 deletions(-) delete mode 100644 src/mesa/pipe/i965simple/brw_metaops.c delete mode 100644 src/mesa/pipe/i965simple/brw_program.c (limited to 'src') diff --git a/src/mesa/pipe/i965simple/Makefile b/src/mesa/pipe/i965simple/Makefile index 460bc52c35..2b7fae0a7a 100644 --- a/src/mesa/pipe/i965simple/Makefile +++ b/src/mesa/pipe/i965simple/Makefile @@ -28,7 +28,6 @@ DRIVER_SOURCES = \ brw_gs_emit.c \ brw_gs_state.c \ brw_misc_state.c \ - brw_program.c \ brw_regions.c \ brw_sf.c \ brw_sf_emit.c \ diff --git a/src/mesa/pipe/i965simple/brw_metaops.c b/src/mesa/pipe/i965simple/brw_metaops.c deleted file mode 100644 index 09f5d28961..0000000000 --- a/src/mesa/pipe/i965simple/brw_metaops.c +++ /dev/null @@ -1,538 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - * frame buffer texture by Gary Wong - */ - - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_draw.h" - -#if 0 - -#define INIT(brw, STRUCT, ATTRIB) \ -do { \ - brw->attribs.ATTRIB = &ctx->ATTRIB; \ -} while (0) - -#define DUP(brw, STRUCT, ATTRIB) \ -do { \ - brw->metaops.attribs.ATTRIB = MALLOC_STRUCT(STRUCT); \ - memcpy(brw->metaops.attribs.ATTRIB, \ - brw->attribs.ATTRIB, \ - sizeof(struct STRUCT)); \ -} while (0) - - -#define INSTALL(brw, ATTRIB, STATE) \ -do { \ - brw->attribs.ATTRIB = brw->metaops.attribs.ATTRIB; \ - brw->state.dirty.mesa |= STATE; \ -} while (0) - -#define RESTORE(brw, ATTRIB, STATE) \ -do { \ - brw->attribs.ATTRIB = &brw->intel.ctx.ATTRIB; \ - brw->state.dirty.mesa |= STATE; \ -} while (0) - -static void init_attribs( struct brw_context *brw ) -{ - DUP(brw, gl_colorbuffer_attrib, Color); - DUP(brw, gl_depthbuffer_attrib, Depth); - DUP(brw, gl_fog_attrib, Fog); - DUP(brw, gl_hint_attrib, Hint); - DUP(brw, gl_light_attrib, Light); - DUP(brw, gl_line_attrib, Line); - DUP(brw, gl_point_attrib, Point); - DUP(brw, gl_polygon_attrib, Polygon); - DUP(brw, gl_scissor_attrib, Scissor); - DUP(brw, gl_stencil_attrib, Stencil); - DUP(brw, gl_texture_attrib, Texture); - DUP(brw, gl_transform_attrib, Transform); - DUP(brw, gl_viewport_attrib, Viewport); - DUP(brw, gl_vertex_program_state, VertexProgram); - DUP(brw, gl_fragment_program_state, FragmentProgram); -} - -static void install_attribs( struct brw_context *brw ) -{ - INSTALL(brw, Color, _NEW_COLOR); - INSTALL(brw, Depth, _NEW_DEPTH); - INSTALL(brw, Fog, _NEW_FOG); - INSTALL(brw, Hint, _NEW_HINT); - INSTALL(brw, Light, _NEW_LIGHT); - INSTALL(brw, Line, _NEW_LINE); - INSTALL(brw, Point, _NEW_POINT); - INSTALL(brw, Polygon, _NEW_POLYGON); - INSTALL(brw, Scissor, _NEW_SCISSOR); - INSTALL(brw, Stencil, _NEW_STENCIL); - INSTALL(brw, Texture, _NEW_TEXTURE); - INSTALL(brw, Transform, _NEW_TRANSFORM); - INSTALL(brw, Viewport, _NEW_VIEWPORT); - INSTALL(brw, VertexProgram, _NEW_PROGRAM); - INSTALL(brw, FragmentProgram, _NEW_PROGRAM); -} - -static void restore_attribs( struct brw_context *brw ) -{ - RESTORE(brw, Color, _NEW_COLOR); - RESTORE(brw, Depth, _NEW_DEPTH); - RESTORE(brw, Fog, _NEW_FOG); - RESTORE(brw, Hint, _NEW_HINT); - RESTORE(brw, Light, _NEW_LIGHT); - RESTORE(brw, Line, _NEW_LINE); - RESTORE(brw, Point, _NEW_POINT); - RESTORE(brw, Polygon, _NEW_POLYGON); - RESTORE(brw, Scissor, _NEW_SCISSOR); - RESTORE(brw, Stencil, _NEW_STENCIL); - RESTORE(brw, Texture, _NEW_TEXTURE); - RESTORE(brw, Transform, _NEW_TRANSFORM); - RESTORE(brw, Viewport, _NEW_VIEWPORT); - RESTORE(brw, VertexProgram, _NEW_PROGRAM); - RESTORE(brw, FragmentProgram, _NEW_PROGRAM); -} - - -static const char *vp_prog = - "!!ARBvp1.0\n" - "MOV result.color, vertex.color;\n" - "MOV result.position, vertex.position;\n" - "END\n"; - -static const char *fp_prog = - "!!ARBfp1.0\n" - "MOV result.color, fragment.color;\n" - "END\n"; - -static const char *fp_tex_prog = - "!!ARBfp1.0\n" - "TEMP a;\n" - "ADD a, fragment.position, program.local[0];\n" - "MUL a, a, program.local[1];\n" - "TEX result.color, a, texture[0], 2D;\n" - "MOV result.depth.z, fragment.position;\n" - "END\n"; - -/* Derived values of importance: - * - * FragmentProgram->_Current - * VertexProgram->_Enabled - * brw->vertex_program - * DrawBuffer->_ColorDrawBufferMask[0] - * - * - * More if drawpixels-through-texture is added. - */ -static void init_metaops_state( struct brw_context *brw ) -{ - GLcontext *ctx = &brw->intel.ctx; - - brw->metaops.vbo = ctx->Driver.NewBufferObject(ctx, 1, GL_ARRAY_BUFFER_ARB); - - ctx->Driver.BufferData(ctx, - GL_ARRAY_BUFFER_ARB, - 4096, - NULL, - GL_DYNAMIC_DRAW_ARB, - brw->metaops.vbo); - - brw->metaops.fp = (struct gl_fragment_program *) - ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 1 ); - - brw->metaops.fp_tex = (struct gl_fragment_program *) - ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 1 ); - - brw->metaops.vp = (struct gl_vertex_program *) - ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 1 ); - - _mesa_parse_arb_fragment_program(ctx, GL_FRAGMENT_PROGRAM_ARB, - fp_prog, strlen(fp_prog), - brw->metaops.fp); - - _mesa_parse_arb_fragment_program(ctx, GL_FRAGMENT_PROGRAM_ARB, - fp_tex_prog, strlen(fp_tex_prog), - brw->metaops.fp_tex); - - _mesa_parse_arb_vertex_program(ctx, GL_VERTEX_PROGRAM_ARB, - vp_prog, strlen(vp_prog), - brw->metaops.vp); - - brw->metaops.attribs.VertexProgram->Current = brw->metaops.vp; - brw->metaops.attribs.VertexProgram->_Enabled = TRUE; - - brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp; -} - -static void meta_flat_shade( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Raster->flatshade = TRUE; - brw->state.dirty.mesa |= _NEW_LIGHT; -} - - -static void meta_no_stencil_write( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Stencil->Enabled = FALSE; - brw->metaops.attribs.Stencil->WriteMask[0] = FALSE; - brw->state.dirty.mesa |= _NEW_STENCIL; -} - -static void meta_no_depth_write( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Depth->Test = FALSE; - brw->metaops.attribs.Depth->Mask = FALSE; - brw->state.dirty.mesa |= _NEW_DEPTH; -} - - -static void meta_depth_replace( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - /* ctx->Driver.Enable( ctx, GL_DEPTH_TEST, TRUE ) - * ctx->Driver.DepthMask( ctx, TRUE ) - */ - brw->metaops.attribs.Depth->Test = TRUE; - brw->metaops.attribs.Depth->Mask = TRUE; - brw->state.dirty.mesa |= _NEW_DEPTH; - - /* ctx->Driver.DepthFunc( ctx, GL_ALWAYS ) - */ - brw->metaops.attribs.Depth->Func = GL_ALWAYS; - - brw->state.dirty.mesa |= _NEW_DEPTH; -} - - -static void meta_stencil_replace( struct intel_context *intel, - unsigned s_mask, - unsigned s_clear) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Stencil->Enabled = TRUE; - brw->metaops.attribs.Stencil->WriteMask[0] = s_mask; - brw->metaops.attribs.Stencil->ValueMask[0] = 0xff; - brw->metaops.attribs.Stencil->Ref[0] = s_clear; - brw->metaops.attribs.Stencil->Function[0] = GL_ALWAYS; - brw->metaops.attribs.Stencil->FailFunc[0] = GL_REPLACE; - brw->metaops.attribs.Stencil->ZPassFunc[0] = GL_REPLACE; - brw->metaops.attribs.Stencil->ZFailFunc[0] = GL_REPLACE; - brw->state.dirty.mesa |= _NEW_STENCIL; -} - - -static void meta_color_mask( struct intel_context *intel, boolean state ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - if (state) - COPY_4V(brw->metaops.attribs.Color->ColorMask, - brw->intel.ctx.Color.ColorMask); - else - ASSIGN_4V(brw->metaops.attribs.Color->ColorMask, 0, 0, 0, 0); - - brw->state.dirty.mesa |= _NEW_COLOR; -} - -static void meta_no_texture( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp; - - brw->metaops.attribs.Texture->CurrentUnit = 0; - brw->metaops.attribs.Texture->_EnabledUnits = 0; - brw->metaops.attribs.Texture->_EnabledCoordUnits = 0; - brw->metaops.attribs.Texture->Unit[ 0 ].Enabled = 0; - brw->metaops.attribs.Texture->Unit[ 0 ]._ReallyEnabled = 0; - - brw->state.dirty.mesa |= _NEW_TEXTURE | _NEW_PROGRAM; -} - -static void meta_texture_blend_replace(struct intel_context *intel) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw->metaops.attribs.Texture->CurrentUnit = 0; - brw->metaops.attribs.Texture->_EnabledUnits = 1; - brw->metaops.attribs.Texture->_EnabledCoordUnits = 1; - brw->metaops.attribs.Texture->Unit[ 0 ].Enabled = TEXTURE_2D_BIT; - brw->metaops.attribs.Texture->Unit[ 0 ]._ReallyEnabled = TEXTURE_2D_BIT; - brw->metaops.attribs.Texture->Unit[ 0 ].Current2D = - intel->frame_buffer_texobj; - brw->metaops.attribs.Texture->Unit[ 0 ]._Current = - intel->frame_buffer_texobj; - - brw->state.dirty.mesa |= _NEW_TEXTURE | _NEW_PROGRAM; -} - -static void meta_import_pixel_state(struct intel_context *intel) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - RESTORE(brw, Color, _NEW_COLOR); - RESTORE(brw, Depth, _NEW_DEPTH); - RESTORE(brw, Fog, _NEW_FOG); - RESTORE(brw, Scissor, _NEW_SCISSOR); - RESTORE(brw, Stencil, _NEW_STENCIL); - RESTORE(brw, Texture, _NEW_TEXTURE); - RESTORE(brw, FragmentProgram, _NEW_PROGRAM); -} - -static void meta_frame_buffer_texture( struct intel_context *intel, - int xoff, int yoff ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - struct intel_region *region = intel_drawbuf_region( intel ); - - INSTALL(brw, FragmentProgram, _NEW_PROGRAM); - - brw->metaops.attribs.FragmentProgram->_Current = brw->metaops.fp_tex; - /* This is unfortunate, but seems to be necessary, since later on we - will end up calling _mesa_load_state_parameters to lookup the - local params (below), and that will want to look in ctx.FragmentProgram - instead of brw->attribs.FragmentProgram. */ - intel->ctx.FragmentProgram.Current = brw->metaops.fp_tex; - - brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 0 ] = xoff; - brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 1 ] = yoff; - brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 2 ] = 0.0; - brw->metaops.fp_tex->Base.LocalParams[ 0 ][ 3 ] = 0.0; - brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 0 ] = - 1.0 / region->pitch; - brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 1 ] = - -1.0 / region->height; - brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 2 ] = 0.0; - brw->metaops.fp_tex->Base.LocalParams[ 1 ][ 3 ] = 1.0; - - brw->state.dirty.mesa |= _NEW_PROGRAM; -} - - -static void meta_draw_region( struct intel_context *intel, - struct intel_region *draw_region, - struct intel_region *depth_region ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - if (!brw->metaops.saved_draw_region) { - brw->metaops.saved_draw_region = brw->state.draw_region; - brw->metaops.saved_depth_region = brw->state.depth_region; - } - - brw->state.draw_region = draw_region; - brw->state.depth_region = depth_region; - - brw->state.dirty.mesa |= _NEW_BUFFERS; -} - - -static void meta_draw_quad(struct intel_context *intel, - float x0, float x1, - float y0, float y1, - float z, - ubyte red, ubyte green, - ubyte blue, ubyte alpha, - float s0, float s1, - float t0, float t1) -{ - GLcontext *ctx = &intel->ctx; - struct brw_context *brw = brw_context(&intel->ctx); - struct gl_client_array pos_array; - struct gl_client_array color_array; - struct gl_client_array *attribs[PIPE_ATTRIB_MAX]; - struct _mesa_prim prim[1]; - float pos[4][3]; - ubyte color[4]; - - ctx->Driver.BufferData(ctx, - GL_ARRAY_BUFFER_ARB, - sizeof(pos) + sizeof(color), - NULL, - GL_DYNAMIC_DRAW_ARB, - brw->metaops.vbo); - - pos[0][0] = x0; - pos[0][1] = y0; - pos[0][2] = z; - - pos[1][0] = x1; - pos[1][1] = y0; - pos[1][2] = z; - - pos[2][0] = x1; - pos[2][1] = y1; - pos[2][2] = z; - - pos[3][0] = x0; - pos[3][1] = y1; - pos[3][2] = z; - - - ctx->Driver.BufferSubData(ctx, - GL_ARRAY_BUFFER_ARB, - 0, - sizeof(pos), - pos, - brw->metaops.vbo); - - color[0] = red; - color[1] = green; - color[2] = blue; - color[3] = alpha; - - ctx->Driver.BufferSubData(ctx, - GL_ARRAY_BUFFER_ARB, - sizeof(pos), - sizeof(color), - color, - brw->metaops.vbo); - - /* Ignoring texture coords. - */ - - memset(attribs, 0, PIPE_ATTRIB_MAX * sizeof(*attribs)); - - attribs[VERT_ATTRIB_POS] = &pos_array; - attribs[VERT_ATTRIB_POS]->Ptr = 0; - attribs[VERT_ATTRIB_POS]->Type = GL_FLOAT; - attribs[VERT_ATTRIB_POS]->Enabled = 1; - attribs[VERT_ATTRIB_POS]->Size = 3; - attribs[VERT_ATTRIB_POS]->StrideB = 3 * sizeof(float); - attribs[VERT_ATTRIB_POS]->Stride = 3 * sizeof(float); - attribs[VERT_ATTRIB_POS]->_MaxElement = 4; - attribs[VERT_ATTRIB_POS]->Normalized = 0; - attribs[VERT_ATTRIB_POS]->BufferObj = brw->metaops.vbo; - - attribs[VERT_ATTRIB_COLOR0] = &color_array; - attribs[VERT_ATTRIB_COLOR0]->Ptr = (const ubyte *)sizeof(pos); - attribs[VERT_ATTRIB_COLOR0]->Type = GL_UNSIGNED_BYTE; - attribs[VERT_ATTRIB_COLOR0]->Enabled = 1; - attribs[VERT_ATTRIB_COLOR0]->Size = 4; - attribs[VERT_ATTRIB_COLOR0]->StrideB = 0; - attribs[VERT_ATTRIB_COLOR0]->Stride = 0; - attribs[VERT_ATTRIB_COLOR0]->_MaxElement = 1; - attribs[VERT_ATTRIB_COLOR0]->Normalized = 1; - attribs[VERT_ATTRIB_COLOR0]->BufferObj = brw->metaops.vbo; - - /* Just ignoring texture coordinates for now. - */ - - memset(prim, 0, sizeof(*prim)); - - prim[0].mode = GL_TRIANGLE_FAN; - prim[0].begin = 1; - prim[0].end = 1; - prim[0].weak = 0; - prim[0].pad = 0; - prim[0].start = 0; - prim[0].count = 4; - - brw_draw_prims(&brw->intel.ctx, - (const struct gl_client_array **)attribs, - prim, 1, - NULL, - 0, - 3 ); -} - - -static void install_meta_state( struct intel_context *intel ) -{ - GLcontext *ctx = &intel->ctx; - struct brw_context *brw = brw_context(ctx); - - if (!brw->metaops.vbo) { - init_metaops_state(brw); - } - - install_attribs(brw); - - meta_no_texture(&brw->intel); - meta_flat_shade(&brw->intel); - brw->metaops.restore_draw_mask = ctx->DrawBuffer->_ColorDrawBufferMask[0]; - brw->metaops.restore_fp = ctx->FragmentProgram.Current; - - /* This works without adjusting refcounts. Fix later? - */ - brw->metaops.saved_draw_region = brw->state.draw_region; - brw->metaops.saved_depth_region = brw->state.depth_region; - brw->metaops.active = 1; - - brw->state.dirty.brw |= BRW_NEW_METAOPS; -} - -static void leave_meta_state( struct intel_context *intel ) -{ - GLcontext *ctx = &intel->ctx; - struct brw_context *brw = brw_context(ctx); - - restore_attribs(brw); - - ctx->DrawBuffer->_ColorDrawBufferMask[0] = brw->metaops.restore_draw_mask; - ctx->FragmentProgram.Current = brw->metaops.restore_fp; - - brw->state.draw_region = brw->metaops.saved_draw_region; - brw->state.depth_region = brw->metaops.saved_depth_region; - brw->metaops.saved_draw_region = NULL; - brw->metaops.saved_depth_region = NULL; - brw->metaops.active = 0; - - brw->state.dirty.mesa |= _NEW_BUFFERS; - brw->state.dirty.brw |= BRW_NEW_METAOPS; -} - - - -void brw_init_metaops( struct brw_context *brw ) -{ - init_attribs(brw); -} - -void brw_destroy_metaops( struct brw_context *brw ) -{ - GLcontext *ctx = &brw->intel.ctx; - - if (brw->metaops.vbo) - ctx->Driver.DeleteBuffer( ctx, brw->metaops.vbo ); - -/* ctx->Driver.DeleteProgram( ctx, brw->metaops.fp ); */ -/* ctx->Driver.DeleteProgram( ctx, brw->metaops.fp_tex ); */ -/* ctx->Driver.DeleteProgram( ctx, brw->metaops.vp ); */ -} -#endif diff --git a/src/mesa/pipe/i965simple/brw_program.c b/src/mesa/pipe/i965simple/brw_program.c deleted file mode 100644 index c4640a2b06..0000000000 --- a/src/mesa/pipe/i965simple/brw_program.c +++ /dev/null @@ -1,130 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_context.h" -#include "brw_util.h" - -#if 0 -static void brwBindProgram( GLcontext *ctx, - int target, - struct gl_program *prog ) -{ - struct brw_context *brw = brw_context(ctx); - - switch (target) { - case GL_VERTEX_PROGRAM_ARB: - brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM; - break; - case GL_FRAGMENT_PROGRAM_ARB: - brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM; - break; - } -} - -static struct gl_program *brwNewProgram( GLcontext *ctx, - int target, - unsigned id ) -{ - struct brw_context *brw = brw_context(ctx); - - switch (target) { - case GL_VERTEX_PROGRAM_ARB: { - struct brw_vertex_program *prog = CALLOC_STRUCT(brw_vertex_program); - if (prog) { - prog->id = brw->program_id++; - - return _mesa_init_vertex_program( ctx, &prog->program, - target, id ); - } - else - return NULL; - } - - case GL_FRAGMENT_PROGRAM_ARB: { - struct brw_fragment_program *prog = CALLOC_STRUCT(brw_fragment_program); - if (prog) { - prog->id = brw->program_id++; - - return _mesa_init_fragment_program( ctx, &prog->program, - target, id ); - } - else - return NULL; - } - - default: - return _mesa_new_program(ctx, target, id); - } -} - -static void brwDeleteProgram( GLcontext *ctx, - struct gl_program *prog ) -{ - - _mesa_delete_program( ctx, prog ); -} - - -static boolean brwIsProgramNative( GLcontext *ctx, - int target, - struct gl_program *prog ) -{ - return TRUE; -} - -static void brwProgramStringNotify( GLcontext *ctx, - int target, - struct gl_program *prog ) -{ - if (target == GL_FRAGMENT_PROGRAM_ARB) { - struct brw_context *brw = brw_context(ctx); - struct brw_fragment_program *p = (struct brw_fragment_program *)prog; - struct brw_fragment_program *fp = (struct brw_fragment_program *)brw->fragment_program; - if (p == fp) - brw->state.dirty.brw |= BRW_NEW_FRAGMENT_PROGRAM; - p->id = brw->program_id++; - p->param_state = p->program.Base.Parameters->StateFlags; - } - else if (target == GL_VERTEX_PROGRAM_ARB) { - struct brw_context *brw = brw_context(ctx); - struct brw_vertex_program *p = (struct brw_vertex_program *)prog; - struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->vertex_program; - if (p == vp) - brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM; - p->id = brw->program_id++; - p->param_state = p->program.Base.Parameters->StateFlags; - - /* Also tell tnl about it: - */ - _tnl_program_string(ctx, target, prog); - } -} -#endif -- cgit v1.2.3 From 5978289df1ea406c684b7193e5f4694eac6c00fb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 12 Dec 2007 17:24:10 +0000 Subject: 965: delete dead files --- src/mesa/pipe/i965simple/Makefile | 1 - src/mesa/pipe/i965simple/brw_regions.c | 39 ---------------------------------- src/mesa/pipe/i965simple/brw_regions.h | 5 ----- 3 files changed, 45 deletions(-) delete mode 100644 src/mesa/pipe/i965simple/brw_regions.c delete mode 100644 src/mesa/pipe/i965simple/brw_regions.h (limited to 'src') diff --git a/src/mesa/pipe/i965simple/Makefile b/src/mesa/pipe/i965simple/Makefile index 2b7fae0a7a..526e020726 100644 --- a/src/mesa/pipe/i965simple/Makefile +++ b/src/mesa/pipe/i965simple/Makefile @@ -28,7 +28,6 @@ DRIVER_SOURCES = \ brw_gs_emit.c \ brw_gs_state.c \ brw_misc_state.c \ - brw_regions.c \ brw_sf.c \ brw_sf_emit.c \ brw_sf_state.c \ diff --git a/src/mesa/pipe/i965simple/brw_regions.c b/src/mesa/pipe/i965simple/brw_regions.c deleted file mode 100644 index 32ad988279..0000000000 --- a/src/mesa/pipe/i965simple/brw_regions.c +++ /dev/null @@ -1,39 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Provide additional functionality on top of bufmgr buffers: - * - 2d semantics and blit operations (XXX: remove/simplify blits??) - * - refcounting of buffers for multiple images in a buffer. - * - refcounting of buffer mappings. - */ - -#include "pipe/p_defines.h" -#include "pipe/p_winsys.h" -#include "brw_context.h" -#include "brw_blit.h" - - diff --git a/src/mesa/pipe/i965simple/brw_regions.h b/src/mesa/pipe/i965simple/brw_regions.h deleted file mode 100644 index 1eebd1d5ee..0000000000 --- a/src/mesa/pipe/i965simple/brw_regions.h +++ /dev/null @@ -1,5 +0,0 @@ -#ifndef BRW_REGIONS_H -#define BRW_REGIONS_H - - -#endif -- cgit v1.2.3 From 216de6ab5bd87dece8e68055018e5b91ff549089 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 12 Dec 2007 18:44:03 +0000 Subject: 965: delete dead files --- src/mesa/pipe/i965simple/Makefile | 1 - src/mesa/pipe/i965simple/brw_tex.c | 65 -------------------------------------- 2 files changed, 66 deletions(-) delete mode 100644 src/mesa/pipe/i965simple/brw_tex.c (limited to 'src') diff --git a/src/mesa/pipe/i965simple/Makefile b/src/mesa/pipe/i965simple/Makefile index 526e020726..dbab152da1 100644 --- a/src/mesa/pipe/i965simple/Makefile +++ b/src/mesa/pipe/i965simple/Makefile @@ -36,7 +36,6 @@ DRIVER_SOURCES = \ brw_state_cache.c \ brw_state_pool.c \ brw_state_upload.c \ - brw_tex.c \ brw_tex_layout.c \ brw_urb.c \ brw_util.c \ diff --git a/src/mesa/pipe/i965simple/brw_tex.c b/src/mesa/pipe/i965simple/brw_tex.c deleted file mode 100644 index e9ff67c3f4..0000000000 --- a/src/mesa/pipe/i965simple/brw_tex.c +++ /dev/null @@ -1,65 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_defines.h" - -#if 0 -void brw_FrameBufferTexInit( struct brw_context *brw ) -{ - struct intel_context *intel = &brw->intel; - GLcontext *ctx = &intel->ctx; - struct intel_region *region = intel->intelScreen->front_region; - struct gl_texture_object *obj; - struct gl_texture_image *img; - - intel->frame_buffer_texobj = obj = - ctx->Driver.NewTextureObject( ctx, (unsigned) -1, GL_TEXTURE_2D ); - - obj->MinFilter = GL_NEAREST; - obj->MagFilter = GL_NEAREST; - - img = ctx->Driver.NewTextureImage( ctx ); - - _mesa_init_teximage_fields( ctx, GL_TEXTURE_2D, img, - region->pitch, region->height, 1, 0, - region->cpp == 4 ? GL_RGBA : GL_RGB ); - - _mesa_set_tex_image( obj, GL_TEXTURE_2D, 0, img ); -} - -void brw_FrameBufferTexDestroy( struct brw_context *brw ) -{ - brw->intel.ctx.Driver.DeleteTexture( &brw->intel.ctx, - brw->intel.frame_buffer_texobj ); -} -#endif -- cgit v1.2.3 From ea7c533d40f78a1f62c7b7129a2ad6b33f810da4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 12 Dec 2007 19:53:41 +0000 Subject: 965: sketch changes to brw_wm_surface_state.c --- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index c04fd57494..844cfc54a9 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -147,7 +147,7 @@ void brw_update_texture_surface( GLcontext *ctx, /* surf.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */ /* Updated in emit_reloc */ - surf.ss1.base_addr = intelObj->mt->region->buffer->offset; + surf.ss1.base_addr = brw_buffer_offset( intelObj->mt->region->buffer ); surf.ss2.mip_count = intelObj->lastLevel - intelObj->firstLevel; surf.ss2.width = firstImage->Width - 1; @@ -198,6 +198,8 @@ static void upload_wm_surfaces(struct brw_context *brw ) surf.ss0.surface_type = BRW_SURFACE_2D; + surf.ss1.base_addr = brw_buffer_offset( region->buffer ); + surf.ss2.width = region->pitch - 1; /* XXX: not really! */ surf.ss2.height = region->height - 1; surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; @@ -218,6 +220,9 @@ static void upload_wm_surfaces(struct brw_context *brw ) surf.ss0.writedisable_blue = !brw->attribs.Color->ColorMask[2]; surf.ss0.writedisable_alpha = !brw->attribs.Color->ColorMask[3]; + + + brw->wm.bind.surf_ss_offset[0] = brw_cache_data( &brw->cache[BRW_SS_SURFACE], &surf ); brw->wm.nr_surfaces = 1; @@ -251,6 +256,11 @@ static void upload_wm_surfaces(struct brw_context *brw ) &brw->wm.bind ); } + +/* KW: Will find a different way to acheive this, see for example the + * state caches with relocs in the i915 swz driver. + */ +#if 0 static void emit_reloc_wm_surfaces(struct brw_context *brw) { int unit; @@ -281,6 +291,9 @@ static void emit_reloc_wm_surfaces(struct brw_context *brw) } } } +#endif + + const struct brw_tracked_state brw_wm_surfaces = { .dirty = { @@ -289,6 +302,5 @@ const struct brw_tracked_state brw_wm_surfaces = { .cache = 0 }, .update = upload_wm_surfaces, - .emit_reloc = emit_reloc_wm_surfaces, }; #endif -- cgit v1.2.3 From 06036bc8c4d31d36f2e143f5d7f2db866f44b342 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 12 Dec 2007 20:07:35 +0000 Subject: 965: remove complex compiler for simple fragment programs. Keep the simple compiler for complex fragment programs... --- src/mesa/pipe/i965simple/Makefile | 6 - src/mesa/pipe/i965simple/brw_wm_debug.c | 172 ----- src/mesa/pipe/i965simple/brw_wm_emit.c | 1289 ------------------------------- src/mesa/pipe/i965simple/brw_wm_pass0.c | 466 ----------- src/mesa/pipe/i965simple/brw_wm_pass1.c | 277 ------- src/mesa/pipe/i965simple/brw_wm_pass2.c | 335 -------- 6 files changed, 2545 deletions(-) delete mode 100644 src/mesa/pipe/i965simple/brw_wm_debug.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm_emit.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm_pass0.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm_pass1.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm_pass2.c (limited to 'src') diff --git a/src/mesa/pipe/i965simple/Makefile b/src/mesa/pipe/i965simple/Makefile index dbab152da1..eda5afaea5 100644 --- a/src/mesa/pipe/i965simple/Makefile +++ b/src/mesa/pipe/i965simple/Makefile @@ -45,14 +45,8 @@ DRIVER_SOURCES = \ brw_vs_state.c \ brw_vtbl.c \ brw_wm.c \ - brw_wm_debug.c \ - brw_wm_emit.c \ - brw_wm_fp.c \ brw_wm_iz.c \ brw_wm_glsl.c \ - brw_wm_pass0.c \ - brw_wm_pass1.c \ - brw_wm_pass2.c \ brw_wm_sampler_state.c \ brw_wm_state.c \ brw_wm_surface_state.c diff --git a/src/mesa/pipe/i965simple/brw_wm_debug.c b/src/mesa/pipe/i965simple/brw_wm_debug.c deleted file mode 100644 index b31a608901..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_debug.c +++ /dev/null @@ -1,172 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_wm.h" - -#if 0 -void brw_wm_print_value( struct brw_wm_compile *c, - struct brw_wm_value *value ) -{ - assert(value); - if (c->state >= PASS2_DONE) - brw_print_reg(value->hw_reg); - else if( value == &c->undef_value ) - _mesa_printf("undef"); - else if( value - c->vreg >= 0 && - value - c->vreg < BRW_WM_MAX_VREG) - _mesa_printf("r%d", value - c->vreg); - else if (value - c->creg >= 0 && - value - c->creg < BRW_WM_MAX_PARAM) - _mesa_printf("c%d", value - c->creg); - else if (value - c->payload.input_interp >= 0 && - value - c->payload.input_interp < FRAG_ATTRIB_MAX) - _mesa_printf("i%d", value - c->payload.input_interp); - else if (value - c->payload.depth >= 0 && - value - c->payload.depth < FRAG_ATTRIB_MAX) - _mesa_printf("d%d", value - c->payload.depth); - else - _mesa_printf("?"); -} - -void brw_wm_print_ref( struct brw_wm_compile *c, - struct brw_wm_ref *ref ) -{ - struct brw_reg hw_reg = ref->hw_reg; - - if (ref->unspill_reg) - _mesa_printf("UNSPILL(%x)/", ref->value->spill_slot); - - if (c->state >= PASS2_DONE) - brw_print_reg(ref->hw_reg); - else { - _mesa_printf("%s", hw_reg.negate ? "-" : ""); - _mesa_printf("%s", hw_reg.abs ? "abs/" : ""); - brw_wm_print_value(c, ref->value); - if ((hw_reg.nr&1) || hw_reg.subnr) { - _mesa_printf("->%d.%d", (hw_reg.nr&1), hw_reg.subnr); - } - } -} - -void brw_wm_print_insn( struct brw_wm_compile *c, - struct brw_wm_instruction *inst ) -{ - unsigned i, arg; - unsigned nr_args = brw_wm_nr_args(inst->opcode); - - _mesa_printf("["); - for (i = 0; i < 4; i++) { - if (inst->dst[i]) { - brw_wm_print_value(c, inst->dst[i]); - if (inst->dst[i]->spill_slot) - _mesa_printf("/SPILL(%x)",inst->dst[i]->spill_slot); - } - else - _mesa_printf("#"); - if (i < 3) - _mesa_printf(","); - } - _mesa_printf("]"); - - if (inst->writemask != WRITEMASK_XYZW) - _mesa_printf(".%s%s%s%s", - GET_BIT(inst->writemask, 0) ? "x" : "", - GET_BIT(inst->writemask, 1) ? "y" : "", - GET_BIT(inst->writemask, 2) ? "z" : "", - GET_BIT(inst->writemask, 3) ? "w" : ""); - - switch (inst->opcode) { - case WM_PIXELXY: - _mesa_printf(" = PIXELXY"); - break; - case WM_DELTAXY: - _mesa_printf(" = DELTAXY"); - break; - case WM_PIXELW: - _mesa_printf(" = PIXELW"); - break; - case WM_WPOSXY: - _mesa_printf(" = WPOSXY"); - break; - case WM_PINTERP: - _mesa_printf(" = PINTERP"); - break; - case WM_LINTERP: - _mesa_printf(" = LINTERP"); - break; - case WM_CINTERP: - _mesa_printf(" = CINTERP"); - break; - case WM_FB_WRITE: - _mesa_printf(" = FB_WRITE"); - break; - default: - _mesa_printf(" = %s", _mesa_opcode_string(inst->opcode)); - break; - } - - if (inst->saturate) - _mesa_printf("_SAT"); - - for (arg = 0; arg < nr_args; arg++) { - - _mesa_printf(" ["); - - for (i = 0; i < 4; i++) { - if (inst->src[arg][i]) { - brw_wm_print_ref(c, inst->src[arg][i]); - } - else - _mesa_printf("%%"); - - if (i < 3) - _mesa_printf(","); - else - _mesa_printf("]"); - } - } - _mesa_printf("\n"); -} - -void brw_wm_print_program( struct brw_wm_compile *c, - const char *stage ) -{ - unsigned insn; - - _mesa_printf("\n\n\n%s:\n", stage); - for (insn = 0; insn < c->nr_insns; insn++) - brw_wm_print_insn(c, &c->instruction[insn]); - _mesa_printf("\n\n\n"); -} - -#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_emit.c b/src/mesa/pipe/i965simple/brw_wm_emit.c deleted file mode 100644 index 75bc4cd419..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_emit.c +++ /dev/null @@ -1,1289 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_wm.h" - -#define SATURATE (1<<5) - -/* Not quite sure how correct this is - need to understand horiz - * vs. vertical strides a little better. - */ -static __inline struct brw_reg sechalf( struct brw_reg reg ) -{ - if (reg.vstride) - reg.nr++; - return reg; -} - -/* Payload R0: - * - * R0.0 -- pixel mask, one bit for each of 4 pixels in 4 tiles, - * corresponding to each of the 16 execution channels. - * R0.1..8 -- ? - * R1.0 -- triangle vertex 0.X - * R1.1 -- triangle vertex 0.Y - * R1.2 -- tile 0 x,y coords (2 packed uwords) - * R1.3 -- tile 1 x,y coords (2 packed uwords) - * R1.4 -- tile 2 x,y coords (2 packed uwords) - * R1.5 -- tile 3 x,y coords (2 packed uwords) - * R1.6 -- ? - * R1.7 -- ? - * R1.8 -- ? - */ - - -static void emit_pixel_xy(struct brw_compile *p, - const struct brw_reg *dst, - unsigned mask, - const struct brw_reg *arg0) -{ - struct brw_reg r1 = brw_vec1_grf(1, 0); - struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW); - - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - - /* Calculate pixel centers by adding 1 or 0 to each of the - * micro-tile coordinates passed in r1. - */ - if (mask & TGSI_WRITEMASK_X) { - brw_ADD(p, - vec16(retype(dst[0], BRW_REGISTER_TYPE_UW)), - stride(suboffset(r1_uw, 4), 2, 4, 0), - brw_imm_v(0x10101010)); - } - - if (mask & TGSI_WRITEMASK_Y) { - brw_ADD(p, - vec16(retype(dst[1], BRW_REGISTER_TYPE_UW)), - stride(suboffset(r1_uw,5), 2, 4, 0), - brw_imm_v(0x11001100)); - } - - brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); -} - - - -static void emit_delta_xy(struct brw_compile *p, - const struct brw_reg *dst, - unsigned mask, - const struct brw_reg *arg0, - const struct brw_reg *arg1) -{ - struct brw_reg r1 = brw_vec1_grf(1, 0); - - /* Calc delta X,Y by subtracting origin in r1 from the pixel - * centers. - */ - if (mask & TGSI_WRITEMASK_X) { - brw_ADD(p, - dst[0], - retype(arg0[0], BRW_REGISTER_TYPE_UW), - negate(r1)); - } - - if (mask & TGSI_WRITEMASK_Y) { - brw_ADD(p, - dst[1], - retype(arg0[1], BRW_REGISTER_TYPE_UW), - negate(suboffset(r1,1))); - - } -} - -static void emit_wpos_xy(struct brw_compile *p, - const struct brw_reg *dst, - unsigned mask, - const struct brw_reg *arg0) -{ - /* Calc delta X,Y by subtracting origin in r1 from the pixel - * centers. - */ - if (mask & TGSI_WRITEMASK_X) { - brw_MOV(p, - dst[0], - retype(arg0[0], BRW_REGISTER_TYPE_UW)); - } - - if (mask & TGSI_WRITEMASK_Y) { - /* TODO -- window_height - Y */ - brw_MOV(p, - dst[1], - negate(retype(arg0[1], BRW_REGISTER_TYPE_UW))); - - } -} - - -static void emit_pixel_w( struct brw_compile *p, - const struct brw_reg *dst, - unsigned mask, - const struct brw_reg *arg0, - const struct brw_reg *deltas) -{ - /* Don't need this if all you are doing is interpolating color, for - * instance. - */ - if (mask & TGSI_WRITEMASK_W) { - struct brw_reg interp3 = brw_vec1_grf(arg0[0].nr+1, 4); - - /* Calc 1/w - just linterp wpos[3] optimized by putting the - * result straight into a message reg. - */ - brw_LINE(p, brw_null_reg(), interp3, deltas[0]); - brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), deltas[1]); - - /* Calc w */ - brw_math_16( p, dst[3], - BRW_MATH_FUNCTION_INV, - BRW_MATH_SATURATE_NONE, - 2, brw_null_reg(), - BRW_MATH_PRECISION_FULL); - } -} - - - -static void emit_linterp( struct brw_compile *p, - const struct brw_reg *dst, - unsigned mask, - const struct brw_reg *arg0, - const struct brw_reg *deltas ) -{ - struct brw_reg interp[4]; - unsigned nr = arg0[0].nr; - unsigned i; - - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - - for(i = 0; i < 4; i++ ) { - if (mask & (1<func; - unsigned msgLength, responseLength; - boolean shadow = (c->key.shadowtex_mask & (1<tex_unit)) ? 1 : 0; - unsigned i, nr; - unsigned emit; - - /* How many input regs are there? - */ - switch (inst->tex_idx) { - case TGSI_TEXTURE_1D: - emit = TGSI_WRITEMASK_X; - nr = 1; - break; - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: - emit = TGSI_WRITEMASK_XY; - nr = 2; - break; - default: - emit = TGSI_WRITEMASK_XYZ; - nr = 3; - break; - } - - if (shadow) { - nr = 4; - emit |= TGSI_WRITEMASK_W; - } - - msgLength = 1; - - for (i = 0; i < nr; i++) { - static const unsigned swz[4] = {0,1,2,2}; - if (emit & (1<payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), - inst->tex_unit + 1, /* surface */ - inst->tex_unit, /* sampler */ - inst->writemask, - (shadow ? - BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE : - BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE), - responseLength, - msgLength, - 0); - - if (shadow) - brw_MOV(p, dst[3], brw_imm_f(1.0)); -} - - -static void emit_txb( struct brw_wm_compile *c, - const struct brw_wm_instruction *inst, - struct brw_reg *dst, - unsigned dst_flags, - struct brw_reg *arg ) -{ - struct brw_compile *p = &c->func; - unsigned msgLength; - - /* Shadow ignored for txb. - */ - switch (inst->tex_idx) { - case TGSI_TEXTURE_1D: - brw_MOV(p, brw_message_reg(2), arg[0]); - brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); - brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); - break; - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: - brw_MOV(p, brw_message_reg(2), arg[0]); - brw_MOV(p, brw_message_reg(4), arg[1]); - brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); - break; - default: - brw_MOV(p, brw_message_reg(2), arg[0]); - brw_MOV(p, brw_message_reg(4), arg[1]); - brw_MOV(p, brw_message_reg(6), arg[2]); - break; - } - - brw_MOV(p, brw_message_reg(8), arg[3]); - msgLength = 9; - - - brw_SAMPLE(p, - retype(vec16(dst[0]), BRW_REGISTER_TYPE_UW), - 1, - retype(c->payload.depth[0].hw_reg, BRW_REGISTER_TYPE_UW), - inst->tex_unit + 1, /* surface */ - inst->tex_unit, /* sampler */ - inst->writemask, - BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS, - 8, /* responseLength */ - msgLength, - 0); - -} - - -static void emit_lit( struct brw_compile *p, - const struct brw_reg *dst, - unsigned mask, - const struct brw_reg *arg0 ) -{ - assert((mask & TGSI_WRITEMASK_XW) == 0); - - if (mask & TGSI_WRITEMASK_Y) { - brw_set_saturate(p, (mask & SATURATE) ? 1 : 0); - brw_MOV(p, dst[1], arg0[0]); - brw_set_saturate(p, 0); - } - - if (mask & TGSI_WRITEMASK_Z) { - emit_math2(p, BRW_MATH_FUNCTION_POW, - &dst[2], - TGSI_WRITEMASK_X | (mask & SATURATE), - &arg0[1], - &arg0[3]); - } - - /* Ordinarily you'd use an iff statement to skip or shortcircuit - * some of the POW calculations above, but 16-wide iff statements - * seem to lock c1 hardware, so this is a nasty workaround: - */ - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_LE, arg0[0], brw_imm_f(0)); - { - if (mask & TGSI_WRITEMASK_Y) - brw_MOV(p, dst[1], brw_imm_f(0)); - - if (mask & TGSI_WRITEMASK_Z) - brw_MOV(p, dst[2], brw_imm_f(0)); - } - brw_set_predicate_control(p, BRW_PREDICATE_NONE); -} - - -/* Kill pixel - set execution mask to zero for those pixels which - * fail. - */ -static void emit_kil( struct brw_wm_compile *c, - struct brw_reg *arg0) -{ - struct brw_compile *p = &c->func; - struct brw_reg r0uw = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); - unsigned i; - - - /* XXX - usually won't need 4 compares! - */ - for (i = 0; i < 4; i++) { - brw_push_insn_state(p); - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_GE, arg0[i], brw_imm_f(0)); - brw_set_predicate_control_flag_value(p, 0xff); - brw_AND(p, r0uw, brw_flag_reg(), r0uw); - brw_pop_insn_state(p); - } -} - -static void fire_fb_write( struct brw_wm_compile *c, - unsigned base_reg, - unsigned nr ) -{ - struct brw_compile *p = &c->func; - - /* Pass through control information: - */ -/* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ - { - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - brw_MOV(p, - brw_message_reg(base_reg + 1), - brw_vec8_grf(1, 0)); - brw_pop_insn_state(p); - } - - /* Send framebuffer write message: */ -/* send (16) null.0<1>:uw m0 r0.0<8;8,1>:uw 0x85a04000:ud { Align1 EOT } */ - brw_fb_WRITE(p, - retype(vec16(brw_null_reg()), BRW_REGISTER_TYPE_UW), - base_reg, - retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), - 0, /* render surface always 0 */ - nr, - 0, - 1); -} - -static void emit_aa( struct brw_wm_compile *c, - struct brw_reg *arg1, - unsigned reg ) -{ - struct brw_compile *p = &c->func; - unsigned comp = c->key.aa_dest_stencil_reg / 2; - unsigned off = c->key.aa_dest_stencil_reg % 2; - struct brw_reg aa = offset(arg1[comp], off); - - brw_push_insn_state(p); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); /* ?? */ - brw_MOV(p, brw_message_reg(reg), aa); - brw_pop_insn_state(p); -} - - -/* Post-fragment-program processing. Send the results to the - * framebuffer. - */ -static void emit_fb_write( struct brw_wm_compile *c, - struct brw_reg *arg0, - struct brw_reg *arg1, - struct brw_reg *arg2) -{ - struct brw_compile *p = &c->func; - unsigned nr = 2; - unsigned channel; - - /* Reserve a space for AA - may not be needed: - */ - if (c->key.aa_dest_stencil_reg) - nr += 1; - - /* I don't really understand how this achieves the color interleave - * (ie RGBARGBA) in the result: [Do the saturation here] - */ - { - brw_push_insn_state(p); - - for (channel = 0; channel < 4; channel++) { - /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ - /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ - - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - brw_MOV(p, - brw_message_reg(nr + channel), - arg0[channel]); - - brw_set_compression_control(p, BRW_COMPRESSION_2NDHALF); - brw_MOV(p, - brw_message_reg(nr + channel + 4), - sechalf(arg0[channel])); - } - - /* skip over the regs populated above: - */ - nr += 8; - - brw_pop_insn_state(p); - } - - if (c->key.source_depth_to_render_target) - { - if (c->key.computes_depth) - brw_MOV(p, brw_message_reg(nr), arg2[2]); - else - brw_MOV(p, brw_message_reg(nr), arg1[1]); /* ? */ - - nr += 2; - } - - if (c->key.dest_depth_reg) - { - unsigned comp = c->key.dest_depth_reg / 2; - unsigned off = c->key.dest_depth_reg % 2; - - if (off != 0) { - brw_push_insn_state(p); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - brw_MOV(p, brw_message_reg(nr), arg1[comp]); - /* 2nd half? */ - brw_MOV(p, brw_message_reg(nr+1), offset(arg1[comp],1)); - brw_pop_insn_state(p); - } - else { - brw_MOV(p, brw_message_reg(nr), arg1[comp]); - } - nr += 2; - } - - - if (!c->key.runtime_check_aads_emit) { - if (c->key.aa_dest_stencil_reg) - emit_aa(c, arg1, 2); - - fire_fb_write(c, 0, nr); - } - else { - struct brw_reg v1_null_ud = vec1(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD)); - struct brw_reg ip = brw_ip_reg(); - struct brw_instruction *jmp; - - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); - brw_AND(p, - v1_null_ud, - get_element_ud(brw_vec8_grf(1,0), 6), - brw_imm_ud(1<<26)); - - jmp = brw_JMPI(p, ip, ip, brw_imm_w(0)); - { - emit_aa(c, arg1, 2); - fire_fb_write(c, 0, nr); - /* note - thread killed in subroutine */ - } - brw_land_fwd_jump(p, jmp); - - /* ELSE: Shuffle up one register to fill in the hole left for AA: - */ - fire_fb_write(c, 1, nr-1); - } -} - - - - -/* Post-fragment-program processing. Send the results to the - * framebuffer. - */ -static void emit_spill( struct brw_wm_compile *c, - struct brw_reg reg, - unsigned slot ) -{ - struct brw_compile *p = &c->func; - - /* - mov (16) m2.0<1>:ud r2.0<8;8,1>:ud { Align1 Compr } - */ - brw_MOV(p, brw_message_reg(2), reg); - - /* - mov (1) r0.2<1>:d 0x00000080:d { Align1 NoMask } - send (16) null.0<1>:uw m1 r0.0<8;8,1>:uw 0x053003ff:ud { Align1 } - */ - brw_dp_WRITE_16(p, - retype(vec16(brw_vec8_grf(0, 0)), BRW_REGISTER_TYPE_UW), - 1, - slot); -} - -static void emit_unspill( struct brw_wm_compile *c, - struct brw_reg reg, - unsigned slot ) -{ - struct brw_compile *p = &c->func; - - /* Slot 0 is the undef value. - */ - if (slot == 0) { - brw_MOV(p, reg, brw_imm_f(0)); - return; - } - - /* - mov (1) r0.2<1>:d 0x000000c0:d { Align1 NoMask } - send (16) r110.0<1>:uw m1 r0.0<8;8,1>:uw 0x041243ff:ud { Align1 } - */ - - brw_dp_READ_16(p, - retype(vec16(reg), BRW_REGISTER_TYPE_UW), - 1, - slot); -} - - - -/** - * Retrieve upto 4 GEN4 register pairs for the given wm reg: - */ -static void get_argument_regs( struct brw_wm_compile *c, - struct brw_wm_ref *arg[], - struct brw_reg *regs ) -{ - unsigned i; - - for (i = 0; i < 4; i++) { - if (arg[i]) { - - if (arg[i]->unspill_reg) - emit_unspill(c, - brw_vec8_grf(arg[i]->unspill_reg, 0), - arg[i]->value->spill_slot); - - regs[i] = arg[i]->hw_reg; - } - else { - regs[i] = brw_null_reg(); - } - } -} - -static void spill_values( struct brw_wm_compile *c, - struct brw_wm_value *values, - unsigned nr ) -{ - unsigned i; - - for (i = 0; i < nr; i++) - if (values[i].spill_slot) - emit_spill(c, values[i].hw_reg, values[i].spill_slot); -} - - - -/* Emit the fragment program instructions here. - */ -void brw_wm_emit( struct brw_wm_compile *c ) -{ - struct brw_compile *p = &c->func; - unsigned insn; - - brw_set_compression_control(p, BRW_COMPRESSION_COMPRESSED); - - /* Check if any of the payload regs need to be spilled: - */ - spill_values(c, c->payload.depth, 4); - spill_values(c, c->creg, c->nr_creg); - spill_values(c, c->payload.input_interp, PIPE_ATTRIB_MAX); - - - for (insn = 0; insn < c->nr_insns; insn++) { - - struct brw_wm_instruction *inst = &c->instruction[insn]; - struct brw_reg args[3][4], dst[4]; - unsigned i, dst_flags; - - /* Get argument regs: - */ - for (i = 0; i < 3; i++) - get_argument_regs(c, inst->src[i], args[i]); - - /* Get dest regs: - */ - for (i = 0; i < 4; i++) - if (inst->dst[i]) - dst[i] = inst->dst[i]->hw_reg; - else - dst[i] = brw_null_reg(); - - /* Flags - */ - dst_flags = inst->writemask; - if (inst->saturate) - dst_flags |= SATURATE; - - switch (inst->opcode) { - /* Generated instructions for calculating triangle interpolants: - */ - case WM_PIXELXY: - emit_pixel_xy(p, dst, dst_flags, args[0]); - break; - - case WM_DELTAXY: - emit_delta_xy(p, dst, dst_flags, args[0], args[1]); - break; - - case WM_WPOSXY: - emit_wpos_xy(p, dst, dst_flags, args[0]); - break; - - case WM_PIXELW: - emit_pixel_w(p, dst, dst_flags, args[0], args[1]); - break; - - case WM_LINTERP: - emit_linterp(p, dst, dst_flags, args[0], args[1]); - break; - - case WM_PINTERP: - emit_pinterp(p, dst, dst_flags, args[0], args[1], args[2]); - break; - - case WM_CINTERP: - emit_cinterp(p, dst, dst_flags, args[0]); - break; - - case WM_FB_WRITE: - emit_fb_write(c, args[0], args[1], args[2]); - break; - - /* Straightforward arithmetic: - */ - case TGSI_OPCODE_ADD: - emit_alu2(p, brw_ADD, dst, dst_flags, args[0], args[1]); - break; - - case TGSI_OPCODE_FRC: - emit_alu1(p, brw_FRC, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_FLR: - emit_alu1(p, brw_RNDD, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_DP3: /* */ - emit_dp3(p, dst, dst_flags, args[0], args[1]); - break; - - case TGSI_OPCODE_DP4: - emit_dp4(p, dst, dst_flags, args[0], args[1]); - break; - - case TGSI_OPCODE_DPH: - emit_dph(p, dst, dst_flags, args[0], args[1]); - break; - - case TGSI_OPCODE_LRP: /* */ - emit_lrp(p, dst, dst_flags, args[0], args[1], args[2]); - break; - - case TGSI_OPCODE_MAD: - emit_mad(p, dst, dst_flags, args[0], args[1], args[2]); - break; - - case TGSI_OPCODE_MOV: - emit_alu1(p, brw_MOV, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_MUL: - emit_alu2(p, brw_MUL, dst, dst_flags, args[0], args[1]); - break; - - case TGSI_OPCODE_XPD: - emit_xpd(p, dst, dst_flags, args[0], args[1]); - break; - - /* Higher math functions: - */ - case TGSI_OPCODE_RCP: - emit_math1(p, BRW_MATH_FUNCTION_INV, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_RSQ: - emit_math1(p, BRW_MATH_FUNCTION_RSQ, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_SIN: - emit_math1(p, BRW_MATH_FUNCTION_SIN, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_COS: - emit_math1(p, BRW_MATH_FUNCTION_COS, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_EX2: - emit_math1(p, BRW_MATH_FUNCTION_EXP, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_LG2: - emit_math1(p, BRW_MATH_FUNCTION_LOG, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_SCS: - /* There is an scs math function, but it would need some - * fixup for 16-element execution. - */ - if (dst_flags & TGSI_WRITEMASK_X) - emit_math1(p, BRW_MATH_FUNCTION_COS, dst, (dst_flags&SATURATE)|TGSI_WRITEMASK_X, - args[0]); - if (dst_flags & TGSI_WRITEMASK_Y) - emit_math1(p, BRW_MATH_FUNCTION_SIN, dst+1, (dst_flags&SATURATE)|TGSI_WRITEMASK_X, - args[0]); - break; - - case TGSI_OPCODE_POW: - emit_math2(p, BRW_MATH_FUNCTION_POW, dst, dst_flags, args[0], args[1]); - break; - - /* Comparisons: - */ - case TGSI_OPCODE_CMP: - emit_cmp(p, dst, dst_flags, args[0], args[1], args[2]); - break; - - case TGSI_OPCODE_MAX: - emit_max(p, dst, dst_flags, args[0], args[1]); - break; - - case TGSI_OPCODE_MIN: - emit_min(p, dst, dst_flags, args[0], args[1]); - break; - - case TGSI_OPCODE_SLT: - emit_slt(p, dst, dst_flags, args[0], args[1]); - break; - - case TGSI_OPCODE_SLE: - emit_sle(p, dst, dst_flags, args[0], args[1]); - break; - case TGSI_OPCODE_SGT: - emit_sgt(p, dst, dst_flags, args[0], args[1]); - break; - case TGSI_OPCODE_SGE: - emit_sge(p, dst, dst_flags, args[0], args[1]); - break; - case TGSI_OPCODE_SEQ: - emit_seq(p, dst, dst_flags, args[0], args[1]); - break; - case TGSI_OPCODE_SNE: - emit_sne(p, dst, dst_flags, args[0], args[1]); - break; - - case TGSI_OPCODE_LIT: - emit_lit(p, dst, dst_flags, args[0]); - break; - - /* Texturing operations: - */ - case TGSI_OPCODE_TEX: - emit_tex(c, inst, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_TXB: - emit_txb(c, inst, dst, dst_flags, args[0]); - break; - - case TGSI_OPCODE_KIL: - emit_kil(c, args[0]); - break; - - default: - printf("unsupport opcode %d in fragment program\n", - inst->opcode); - } - - for (i = 0; i < 4; i++) - if (inst->dst[i] && inst->dst[i]->spill_slot) - emit_spill(c, - inst->dst[i]->hw_reg, - inst->dst[i]->spill_slot); - } -} diff --git a/src/mesa/pipe/i965simple/brw_wm_pass0.c b/src/mesa/pipe/i965simple/brw_wm_pass0.c deleted file mode 100644 index 2ec3ac00c1..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_pass0.c +++ /dev/null @@ -1,466 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_wm.h" - - -#if 0 -/*********************************************************************** - */ - -static struct brw_wm_ref *get_ref( struct brw_wm_compile *c ) -{ - assert(c->nr_refs < BRW_WM_MAX_REF); - return &c->refs[c->nr_refs++]; -} - -static struct brw_wm_value *get_value( struct brw_wm_compile *c) -{ - assert(c->nr_refs < BRW_WM_MAX_VREG); - return &c->vreg[c->nr_vreg++]; -} - -static struct brw_wm_instruction *get_instruction( struct brw_wm_compile *c ) -{ - assert(c->nr_insns < BRW_WM_MAX_INSN); - return &c->instruction[c->nr_insns++]; -} - -/*********************************************************************** - */ - -static void pass0_init_undef( struct brw_wm_compile *c) -{ - struct brw_wm_ref *ref = &c->undef_ref; - ref->value = &c->undef_value; - ref->hw_reg = brw_vec8_grf(0, 0); - ref->insn = 0; - ref->prevuse = NULL; -} - -static void pass0_set_fpreg_value( struct brw_wm_compile *c, - unsigned file, - unsigned idx, - unsigned component, - struct brw_wm_value *value ) -{ - struct brw_wm_ref *ref = get_ref(c); - ref->value = value; - ref->hw_reg = brw_vec8_grf(0, 0); - ref->insn = 0; - ref->prevuse = NULL; - c->pass0_fp_reg[file][idx][component] = ref; -} - -static void pass0_set_fpreg_ref( struct brw_wm_compile *c, - unsigned file, - unsigned idx, - unsigned component, - const struct brw_wm_ref *src_ref ) -{ - c->pass0_fp_reg[file][idx][component] = src_ref; -} - -static const struct brw_wm_ref *get_param_ref( struct brw_wm_compile *c, - const float *param_ptr ) -{ - unsigned i = c->prog_data.nr_params++; - - if (i >= BRW_WM_MAX_PARAM) { - _mesa_printf("%s: out of params\n", __FUNCTION__); - c->prog_data.error = 1; - return NULL; - } - else { - struct brw_wm_ref *ref = get_ref(c); - - c->prog_data.param[i] = param_ptr; - c->nr_creg = (i+16)/16; - - /* Push the offsets into hw_reg. These will be added to the - * real register numbers once one is allocated in pass2. - */ - ref->hw_reg = brw_vec1_grf((i&8)?1:0, i%8); - ref->value = &c->creg[i/16]; - ref->insn = 0; - ref->prevuse = NULL; - - return ref; - } -} - - -static const struct brw_wm_ref *get_const_ref( struct brw_wm_compile *c, - const float *constval ) -{ - unsigned i; - - /* Search for an existing const value matching the request: - */ - for (i = 0; i < c->nr_constrefs; i++) { - if (c->constref[i].constval == *constval) - return c->constref[i].ref; - } - - /* Else try to add a new one: - */ - if (c->nr_constrefs < BRW_WM_MAX_CONST) { - unsigned i = c->nr_constrefs++; - - /* A constant is a special type of parameter: - */ - c->constref[i].constval = *constval; - c->constref[i].ref = get_param_ref(c, constval); - - return c->constref[i].ref; - } - else { - _mesa_printf("%s: out of constrefs\n", __FUNCTION__); - c->prog_data.error = 1; - return NULL; - } -} - - -/* Lookup our internal registers - */ -static const struct brw_wm_ref *pass0_get_reg( struct brw_wm_compile *c, - unsigned file, - unsigned idx, - unsigned component ) -{ - const struct brw_wm_ref *ref = c->pass0_fp_reg[file][idx][component]; - - if (!ref) { - switch (file) { - case PROGRAM_INPUT: - case PROGRAM_PAYLOAD: - case PROGRAM_TEMPORARY: - case PROGRAM_OUTPUT: - case PROGRAM_VARYING: - break; - - case PROGRAM_LOCAL_PARAM: - ref = get_param_ref(c, &c->fp->program.Base.LocalParams[idx][component]); - break; - - case PROGRAM_ENV_PARAM: - ref = get_param_ref(c, &c->env_param[idx][component]); - break; - - case PROGRAM_STATE_VAR: - case PROGRAM_UNIFORM: - case PROGRAM_CONSTANT: - case PROGRAM_NAMED_PARAM: { - struct gl_program_parameter_list *plist = c->fp->program.Base.Parameters; - - /* There's something really hokey about parameters parsed in - * arb programs - they all end up in here, whether they be - * state values, paramters or constants. This duplicates the - * structure above & also seems to subvert the limits set for - * each type of constant/param. - */ - switch (plist->Parameters[idx].Type) { - case PROGRAM_NAMED_PARAM: - case PROGRAM_CONSTANT: - /* These are invarient: - */ - ref = get_const_ref(c, &plist->ParameterValues[idx][component]); - break; - - case PROGRAM_STATE_VAR: - case PROGRAM_UNIFORM: - /* These may change from run to run: - */ - ref = get_param_ref(c, &plist->ParameterValues[idx][component] ); - break; - - default: - assert(0); - break; - } - break; - } - - default: - assert(0); - break; - } - - c->pass0_fp_reg[file][idx][component] = ref; - } - - if (!ref) - ref = &c->undef_ref; - - return ref; -} - - - - -/*********************************************************************** - * Straight translation to internal instruction format - */ - -static void pass0_set_dst( struct brw_wm_compile *c, - struct brw_wm_instruction *out, - const struct prog_instruction *inst, - unsigned writemask ) -{ - const struct prog_dst_register *dst = &inst->DstReg; - unsigned i; - - for (i = 0; i < 4; i++) { - if (writemask & (1<dst[i] = get_value(c); - - pass0_set_fpreg_value(c, dst->File, dst->Index, i, out->dst[i]); - } - } - - out->writemask = writemask; -} - - -static void pass0_set_dst_scalar( struct brw_wm_compile *c, - struct brw_wm_instruction *out, - const struct prog_instruction *inst, - unsigned writemask ) -{ - if (writemask) { - const struct prog_dst_register *dst = &inst->DstReg; - unsigned i; - - /* Compute only the first (X) value: - */ - out->writemask = WRITEMASK_X; - out->dst[0] = get_value(c); - - /* Update our tracking register file for all the components in - * writemask: - */ - for (i = 0; i < 4; i++) { - if (writemask & (1<File, dst->Index, i, out->dst[0]); - } - } - } - else - out->writemask = 0; -} - - - -static const struct brw_wm_ref *get_fp_src_reg_ref( struct brw_wm_compile *c, - struct prog_src_register src, - unsigned i ) -{ - unsigned component = GET_SWZ(src.Swizzle,i); - const struct brw_wm_ref *src_ref; - static const float const_zero = 0.0; - static const float const_one = 1.0; - - - if (component == SWIZZLE_ZERO) - src_ref = get_const_ref(c, &const_zero); - else if (component == SWIZZLE_ONE) - src_ref = get_const_ref(c, &const_one); - else - src_ref = pass0_get_reg(c, src.File, src.Index, component); - - return src_ref; -} - - -static struct brw_wm_ref *get_new_ref( struct brw_wm_compile *c, - struct prog_src_register src, - unsigned i, - struct brw_wm_instruction *insn) -{ - const struct brw_wm_ref *ref = get_fp_src_reg_ref(c, src, i); - struct brw_wm_ref *newref = get_ref(c); - - newref->value = ref->value; - newref->hw_reg = ref->hw_reg; - - if (insn) { - newref->insn = insn - c->instruction; - newref->prevuse = newref->value->lastuse; - newref->value->lastuse = newref; - } - - if (src.NegateBase & (1<hw_reg.negate ^= 1; - - if (src.Abs) { - newref->hw_reg.negate = 0; - newref->hw_reg.abs = 1; - } - - return newref; -} - - - -static struct brw_wm_instruction *translate_insn( struct brw_wm_compile *c, - const struct prog_instruction *inst ) -{ - struct brw_wm_instruction *out = get_instruction(c); - unsigned writemask = inst->DstReg.WriteMask; - unsigned nr_args = brw_wm_nr_args(inst->Opcode); - unsigned i, j; - - /* Copy some data out of the instruction - */ - out->opcode = inst->Opcode; - out->saturate = (inst->SaturateMode != SATURATE_OFF); - out->tex_unit = inst->TexSrcUnit; - out->tex_idx = inst->TexSrcTarget; - - /* Args: - */ - for (i = 0; i < nr_args; i++) { - for (j = 0; j < 4; j++) { - out->src[i][j] = get_new_ref(c, inst->SrcReg[i], j, out); - } - } - - /* Dst: - */ - if (brw_wm_is_scalar_result(out->opcode)) - pass0_set_dst_scalar(c, out, inst, writemask); - else - pass0_set_dst(c, out, inst, writemask); - - return out; -} - - - -/*********************************************************************** - * Optimize moves and swizzles away: - */ -static void pass0_precalc_mov( struct brw_wm_compile *c, - const struct prog_instruction *inst ) -{ - const struct prog_dst_register *dst = &inst->DstReg; - unsigned writemask = inst->DstReg.WriteMask; - unsigned i; - - /* Get the effect of a MOV by manipulating our register table: - */ - for (i = 0; i < 4; i++) { - if (writemask & (1<File, dst->Index, i, - get_new_ref(c, inst->SrcReg[0], i, NULL)); - } - } -} - - -/* Initialize payload "registers". - */ -static void pass0_init_payload( struct brw_wm_compile *c ) -{ - unsigned i; - - for (i = 0; i < 4; i++) { - unsigned j = i >= c->key.nr_depth_regs ? 0 : i; - pass0_set_fpreg_value( c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, i, - &c->payload.depth[j] ); - } - -#if 0 - /* This seems to be an alternative to the INTERP_WPOS stuff I do - * elsewhere: - */ - if (c->key.source_depth_reg) - pass0_set_fpreg_value(c, PROGRAM_INPUT, FRAG_ATTRIB_WPOS, 2, - &c->payload.depth[c->key.source_depth_reg/2]); -#endif - - for (i = 0; i < FRAG_ATTRIB_MAX; i++) - pass0_set_fpreg_value( c, PROGRAM_PAYLOAD, i, 0, - &c->payload.input_interp[i] ); -} - -/*********************************************************************** - * PASS 0 - * - * Work forwards to give each calculated value a unique number. Where - * an instruction produces duplicate values (eg DP3), all are given - * the same number. - * - * Translate away swizzling and eliminate non-saturating moves. - */ -void brw_wm_pass0( struct brw_wm_compile *c ) -{ - unsigned insn; - - c->nr_vreg = 0; - c->nr_insns = 0; - - pass0_init_undef(c); - pass0_init_payload(c); - - for (insn = 0; insn < c->nr_fp_insns; insn++) { - const struct prog_instruction *inst = &c->prog_instructions[insn]; - - - /* Optimize away moves, otherwise emit translated instruction: - */ - switch (inst->Opcode) { - case OPCODE_MOV: - case OPCODE_SWZ: - if (!inst->SaturateMode) { - pass0_precalc_mov(c, inst); - } - else { - translate_insn(c, inst); - } - break; - - - default: - translate_insn(c, inst); - break; - } - } - - if (BRW_DEBUG & DEBUG_WM) { - brw_wm_print_program(c, "pass0"); - } -} - -#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_pass1.c b/src/mesa/pipe/i965simple/brw_wm_pass1.c deleted file mode 100644 index e6736bd6df..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_pass1.c +++ /dev/null @@ -1,277 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_wm.h" - -#if 0 -static unsigned get_tracked_mask(struct brw_wm_compile *c, - struct brw_wm_instruction *inst) -{ - unsigned i; - for (i = 0; i < 4; i++) { - if (inst->writemask & (1<dst[i]->contributes_to_output) { - inst->writemask &= ~(1<dst[i] = 0; - } - } - } - - return inst->writemask; -} - -/* Remove a reference from a value's usage chain. - */ -static void unlink_ref(struct brw_wm_ref *ref) -{ - struct brw_wm_value *value = ref->value; - - if (ref == value->lastuse) { - value->lastuse = ref->prevuse; - } else { - struct brw_wm_ref *i = value->lastuse; - while (i->prevuse != ref) i = i->prevuse; - i->prevuse = ref->prevuse; - } -} - -static void track_arg(struct brw_wm_compile *c, - struct brw_wm_instruction *inst, - unsigned arg, - unsigned readmask) -{ - unsigned i; - - for (i = 0; i < 4; i++) { - struct brw_wm_ref *ref = inst->src[arg][i]; - if (ref) { - if (readmask & (1<value->contributes_to_output = 1; - else { - unlink_ref(ref); - inst->src[arg][i] = NULL; - } - } - } -} - -static unsigned get_texcoord_mask( unsigned tex_idx ) -{ - switch (tex_idx) { - case TEXTURE_1D_INDEX: return WRITEMASK_X; - case TEXTURE_2D_INDEX: return WRITEMASK_XY; - case TEXTURE_3D_INDEX: return WRITEMASK_XYZ; - case TEXTURE_CUBE_INDEX: return WRITEMASK_XYZ; - case TEXTURE_RECT_INDEX: return WRITEMASK_XY; - default: return 0; - } -} - -/* Step two: Basically this is dead code elimination. - * - * Iterate backwards over instructions, noting which values - * contribute to the final result. Adjust writemasks to only - * calculate these values. - */ -void brw_wm_pass1( struct brw_wm_compile *c ) -{ - int insn; - - for (insn = c->nr_insns-1; insn >= 0; insn--) { - struct brw_wm_instruction *inst = &c->instruction[insn]; - unsigned writemask; - unsigned read0, read1, read2; - - if (inst->opcode == OPCODE_KIL) { - track_arg(c, inst, 0, WRITEMASK_XYZW); /* All args contribute to final */ - continue; - } - - if (inst->opcode == WM_FB_WRITE) { - track_arg(c, inst, 0, WRITEMASK_XYZW); - track_arg(c, inst, 1, WRITEMASK_XYZW); - if (c->key.source_depth_to_render_target && - c->key.computes_depth) - track_arg(c, inst, 2, WRITEMASK_Z); - else - track_arg(c, inst, 2, 0); - continue; - } - - /* Lookup all the registers which were written by this - * instruction and get a mask of those that contribute to the output: - */ - writemask = get_tracked_mask(c, inst); - if (!writemask) { - unsigned arg; - for (arg = 0; arg < 3; arg++) - track_arg(c, inst, arg, 0); - continue; - } - - read0 = 0; - read1 = 0; - read2 = 0; - - /* Mark all inputs which contribute to the marked outputs: - */ - switch (inst->opcode) { - case OPCODE_ABS: - case OPCODE_FLR: - case OPCODE_FRC: - case OPCODE_MOV: - read0 = writemask; - break; - - case OPCODE_SUB: - case OPCODE_SLT: - case OPCODE_SLE: - case OPCODE_SGE: - case OPCODE_SGT: - case OPCODE_SEQ: - case OPCODE_SNE: - case OPCODE_ADD: - case OPCODE_MAX: - case OPCODE_MIN: - case OPCODE_MUL: - read0 = writemask; - read1 = writemask; - break; - - case OPCODE_MAD: - case OPCODE_CMP: - case OPCODE_LRP: - read0 = writemask; - read1 = writemask; - read2 = writemask; - break; - - case OPCODE_XPD: - if (writemask & WRITEMASK_X) read0 |= WRITEMASK_YZ; - if (writemask & WRITEMASK_Y) read0 |= WRITEMASK_XZ; - if (writemask & WRITEMASK_Z) read0 |= WRITEMASK_XY; - read1 = read0; - break; - - case OPCODE_COS: - case OPCODE_EX2: - case OPCODE_LG2: - case OPCODE_RCP: - case OPCODE_RSQ: - case OPCODE_SIN: - case OPCODE_SCS: - case WM_CINTERP: - case WM_PIXELXY: - read0 = WRITEMASK_X; - break; - - case OPCODE_POW: - read0 = WRITEMASK_X; - read1 = WRITEMASK_X; - break; - - case OPCODE_TEX: - read0 = get_texcoord_mask(inst->tex_idx); - - if (c->key.shadowtex_mask & (1<tex_unit)) - read0 |= WRITEMASK_Z; - break; - - case OPCODE_TXB: - /* Shadow ignored for txb. - */ - read0 = get_texcoord_mask(inst->tex_idx) | WRITEMASK_W; - break; - - case WM_WPOSXY: - read0 = writemask & WRITEMASK_XY; - break; - - case WM_DELTAXY: - read0 = writemask & WRITEMASK_XY; - read1 = WRITEMASK_X; - break; - - case WM_PIXELW: - read0 = WRITEMASK_X; - read1 = WRITEMASK_XY; - break; - - case WM_LINTERP: - read0 = WRITEMASK_X; - read1 = WRITEMASK_XY; - break; - - case WM_PINTERP: - read0 = WRITEMASK_X; /* interpolant */ - read1 = WRITEMASK_XY; /* deltas */ - read2 = WRITEMASK_W; /* pixel w */ - break; - - case OPCODE_DP3: - read0 = WRITEMASK_XYZ; - read1 = WRITEMASK_XYZ; - break; - - case OPCODE_DPH: - read0 = WRITEMASK_XYZ; - read1 = WRITEMASK_XYZW; - break; - - case OPCODE_DP4: - read0 = WRITEMASK_XYZW; - read1 = WRITEMASK_XYZW; - break; - - case OPCODE_LIT: - read0 = WRITEMASK_XYW; - break; - - case OPCODE_SWZ: - case OPCODE_DST: - case OPCODE_TXP: - default: - break; - } - - track_arg(c, inst, 0, read0); - track_arg(c, inst, 1, read1); - track_arg(c, inst, 2, read2); - } - - if (BRW_DEBUG & DEBUG_WM) { - brw_wm_print_program(c, "pass1"); - } -} - -#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_pass2.c b/src/mesa/pipe/i965simple/brw_wm_pass2.c deleted file mode 100644 index b1b404706f..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_pass2.c +++ /dev/null @@ -1,335 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_wm.h" - -#if 0 -/* Use these to force spilling so that that functionality can be - * tested with known-good examples rather than having to construct new - * tests. - */ -#define TEST_PAYLOAD_SPILLS 0 -#define TEST_DST_SPILLS 0 - -static void spill_value(struct brw_wm_compile *c, - struct brw_wm_value *value); - -static void prealloc_reg(struct brw_wm_compile *c, - struct brw_wm_value *value, - unsigned reg) -{ - if (value->lastuse) { - /* Set nextuse to zero, it will be corrected by - * update_register_usage(). - */ - c->pass2_grf[reg].value = value; - c->pass2_grf[reg].nextuse = 0; - - value->resident = &c->pass2_grf[reg]; - value->hw_reg = brw_vec8_grf(reg*2, 0); - - if (TEST_PAYLOAD_SPILLS) - spill_value(c, value); - } -} - - -/* Initialize all the register values. Do the initial setup - * calculations for interpolants. - */ -static void init_registers( struct brw_wm_compile *c ) -{ - unsigned inputs = FRAG_BIT_WPOS | c->fp_interp_emitted; - unsigned nr_interp_regs = 0; - unsigned i = 0; - unsigned j; - - for (j = 0; j < c->grf_limit; j++) - c->pass2_grf[j].nextuse = BRW_WM_MAX_INSN; - - for (j = 0; j < c->key.nr_depth_regs; j++) - prealloc_reg(c, &c->payload.depth[j], i++); - - for (j = 0; j < c->nr_creg; j++) - prealloc_reg(c, &c->creg[j], i++); - - for (j = 0; j < FRAG_ATTRIB_MAX; j++) - if (inputs & (1<payload.input_interp[j], i++); - } - - assert(nr_interp_regs >= 1); - - c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; - c->prog_data.urb_read_length = nr_interp_regs * 2; - c->prog_data.curb_read_length = c->nr_creg * 2; - - c->max_wm_grf = i * 2; -} - - -/* Update the nextuse value for each register in our file. - */ -static void update_register_usage(struct brw_wm_compile *c, - unsigned thisinsn) -{ - unsigned i; - - for (i = 1; i < c->grf_limit; i++) { - struct brw_wm_grf *grf = &c->pass2_grf[i]; - - /* Only search those which can change: - */ - if (grf->nextuse < thisinsn) { - struct brw_wm_ref *ref = grf->value->lastuse; - - /* Has last use of value been passed? - */ - if (ref->insn < thisinsn) { - grf->value->resident = 0; - grf->value = 0; - grf->nextuse = BRW_WM_MAX_INSN; - } - else { - /* Else loop through chain to update: - */ - while (ref->prevuse && ref->prevuse->insn >= thisinsn) - ref = ref->prevuse; - - grf->nextuse = ref->insn; - } - } - } -} - - -static void spill_value(struct brw_wm_compile *c, - struct brw_wm_value *value) -{ - /* Allocate a spill slot. Note that allocations start from 0x40 - - * the first slot is reserved to mean "undef" in brw_wm_emit.c - */ - if (!value->spill_slot) { - c->last_scratch += 0x40; - value->spill_slot = c->last_scratch; - } - - /* The spill will be done in brw_wm_emit.c immediately after the - * value is calculated, so we can just take this reg without any - * further work. - */ - value->resident->value = NULL; - value->resident->nextuse = BRW_WM_MAX_INSN; - value->resident = NULL; -} - - - -/* Search for contiguous region with the most distant nearest - * member. Free regs count as very distant. - * - * TODO: implement spill-to-reg so that we can rearrange discontigous - * free regs and then spill the oldest non-free regs in sequence. - * This would mean inserting instructions in this pass. - */ -static unsigned search_contiguous_regs(struct brw_wm_compile *c, - unsigned nr, - unsigned thisinsn) -{ - struct brw_wm_grf *grf = c->pass2_grf; - unsigned furthest = 0; - unsigned reg = 0; - unsigned i, j; - - /* Start search at 1: r0 is special and can't be used or spilled. - */ - for (i = 1; i < c->grf_limit && furthest < BRW_WM_MAX_INSN; i++) { - unsigned group_nextuse = BRW_WM_MAX_INSN; - - for (j = 0; j < nr; j++) { - if (grf[i+j].nextuse < group_nextuse) - group_nextuse = grf[i+j].nextuse; - } - - if (group_nextuse > furthest) { - furthest = group_nextuse; - reg = i; - } - } - - assert(furthest != thisinsn); - - /* Any non-empty regs will need to be spilled: - */ - for (j = 0; j < nr; j++) - if (grf[reg+j].value) - spill_value(c, grf[reg+j].value); - - return reg; -} - - -static void alloc_contiguous_dest(struct brw_wm_compile *c, - struct brw_wm_value *dst[], - unsigned nr, - unsigned thisinsn) -{ - unsigned reg = search_contiguous_regs(c, nr, thisinsn); - unsigned i; - - for (i = 0; i < nr; i++) { - if (!dst[i]) { - /* Need to grab a dummy value in TEX case. Don't introduce - * it into the tracking scheme. - */ - dst[i] = &c->vreg[c->nr_vreg++]; - } - else { - assert(!dst[i]->resident); - assert(c->pass2_grf[reg+i].nextuse != thisinsn); - - c->pass2_grf[reg+i].value = dst[i]; - c->pass2_grf[reg+i].nextuse = thisinsn; - - dst[i]->resident = &c->pass2_grf[reg+i]; - } - - dst[i]->hw_reg = brw_vec8_grf((reg+i)*2, 0); - } - - if ((reg+nr)*2 > c->max_wm_grf) - c->max_wm_grf = (reg+nr) * 2; -} - - -static void load_args(struct brw_wm_compile *c, - struct brw_wm_instruction *inst) -{ - unsigned thisinsn = inst - c->instruction; - unsigned i,j; - - for (i = 0; i < 3; i++) { - for (j = 0; j < 4; j++) { - struct brw_wm_ref *ref = inst->src[i][j]; - - if (ref) { - if (!ref->value->resident) { - /* Need to bring the value in from scratch space. The code for - * this will be done in brw_wm_emit.c, here we just do the - * register allocation and mark the ref as requiring a fill. - */ - unsigned reg = search_contiguous_regs(c, 1, thisinsn); - - c->pass2_grf[reg].value = ref->value; - c->pass2_grf[reg].nextuse = thisinsn; - - ref->value->resident = &c->pass2_grf[reg]; - - /* Note that a fill is required: - */ - ref->unspill_reg = reg*2; - } - - /* Adjust the hw_reg to point at the value's current location: - */ - assert(ref->value == ref->value->resident->value); - ref->hw_reg.nr += (ref->value->resident - c->pass2_grf) * 2; - } - } - } -} - - - -/* Step 3: Work forwards once again. Perform register allocations, - * taking into account instructions like TEX which require contiguous - * result registers. Where necessary spill registers to scratch space - * and reload later. - */ -void brw_wm_pass2( struct brw_wm_compile *c ) -{ - unsigned insn; - unsigned i; - - init_registers(c); - - for (insn = 0; insn < c->nr_insns; insn++) { - struct brw_wm_instruction *inst = &c->instruction[insn]; - - /* Update registers' nextuse values: - */ - update_register_usage(c, insn); - - /* May need to unspill some args. - */ - load_args(c, inst); - - /* Allocate registers to hold results: - */ - switch (inst->opcode) { - case OPCODE_TEX: - case OPCODE_TXB: - case OPCODE_TXP: - alloc_contiguous_dest(c, inst->dst, 4, insn); - break; - - default: - for (i = 0; i < 4; i++) { - if (inst->writemask & (1<dst[i]); - alloc_contiguous_dest(c, &inst->dst[i], 1, insn); - } - } - break; - } - - if (TEST_DST_SPILLS && inst->opcode != WM_PIXELXY) - for (i = 0; i < 4; i++) - if (inst->dst[i]) - spill_value(c, inst->dst[i]); - - } - - if (BRW_DEBUG & DEBUG_WM) { - brw_wm_print_program(c, "pass2"); - } - - c->state = PASS2_DONE; - - if (BRW_DEBUG & DEBUG_WM) { - brw_wm_print_program(c, "pass2/done"); - } -} - -#endif -- cgit v1.2.3 From a51d0e419a285c5445061a38fdaf3aca02ad2c3c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 14:45:22 -0700 Subject: move some code to avoid deref of NULL ptr --- src/mesa/pipe/softpipe/sp_tile_cache.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 25c6dd4d17..d637c7291f 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -147,10 +147,10 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, pipe_surface_reference(&tc->surface, ps); - if (!ps->map) - pipe_surface_map(ps); - if (ps) { + if (!ps->map) + pipe_surface_map(ps); + tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || ps->format == PIPE_FORMAT_Z16_UNORM || ps->format == PIPE_FORMAT_Z32_UNORM || -- cgit v1.2.3 From 20eae595faa20dba8a59d8a4bfd01aa6b458cecd Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 14:55:57 -0700 Subject: Re-org of st_create_framebuffer() and renderbuffer format selection. st_create_framebuffer() now takes pipe_formats for the color, depth, stencil buffers. This avoids a round-about chain of calls to pipe->is_format_supported() for window renderbuffers (their format never changes). Renderbuffer format selection code in st_format.c is simpler now too. --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 25 ++++- src/mesa/pipe/softpipe/sp_surface.c | 5 + src/mesa/pipe/xlib/xm_api.c | 120 ++++++++++++++--------- src/mesa/state_tracker/st_cb_fbo.c | 59 +++++++---- src/mesa/state_tracker/st_cb_fbo.h | 5 +- src/mesa/state_tracker/st_format.c | 58 ++--------- src/mesa/state_tracker/st_format.h | 5 +- src/mesa/state_tracker/st_framebuffer.c | 31 +++--- src/mesa/state_tracker/st_public.h | 4 + 9 files changed, 173 insertions(+), 139 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index 1b520f7135..bce6c5699a 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -274,11 +274,34 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, return GL_FALSE; /* not implemented */ } else { + enum pipe_format colorFormat, depthFormat, stencilFormat; struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); + if (!intelfb) return GL_FALSE; - intelfb->stfb = st_create_framebuffer(visual, GL_TRUE, (void*) intelfb); + if (visual->redBits == 5) + colorFormat = PIPE_FORMAT_R5G6B5_UNORM; + else + colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; + + if (visual->depthBits == 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (visual->depthBits == 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_NONE; + + if (visual->stencilBits == 8) + stencilFormat = PIPE_FORMAT_S8Z24_UNORM; + else + stencilFormat = PIPE_FORMAT_NONE; + + intelfb->stfb = st_create_framebuffer(visual, GL_TRUE, + colorFormat, + depthFormat, + stencilFormat, + (void*) intelfb); if (!intelfb->stfb) { free(intelfb); return GL_FALSE; diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 1dc494d6ff..c81e0f9c18 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -783,6 +783,11 @@ softpipe_put_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_A1R5G5B5_UNORM: /*a1r5g5b5_put_tile(ps, x, y, w, h, p);*/ break; + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_R8G8B8A8_UNORM: + /* XXX need these */ + fprintf(stderr, "unsup pipe format in softpipe_put_tile_rgba()\n"); + break; case PIPE_FORMAT_U_L8: /*l8_put_tile(ps, x, y, w, h, p);*/ break; diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 4021af187c..142074bc65 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -251,6 +251,52 @@ xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b, } +/** + * Choose the pixel format for the given visual. + * This will tell the gallium driver how to pack pixel data into + * drawing surfaces. + */ +static GLuint +choose_pixel_format(XMesaVisual v) +{ + if ( GET_REDMASK(v) == 0x0000ff + && GET_GREENMASK(v) == 0x00ff00 + && GET_BLUEMASK(v) == 0xff0000 + && v->BitsPerPixel == 32) { + if (CHECK_BYTE_ORDER(v)) { + /* no byteswapping needed */ + return 0 /* PIXEL_FORMAT_U_A8_B8_G8_R8 */; + } + else { + return PIPE_FORMAT_R8G8B8A8_UNORM; + } + } + else if ( GET_REDMASK(v) == 0xff0000 + && GET_GREENMASK(v) == 0x00ff00 + && GET_BLUEMASK(v) == 0x0000ff + && v->BitsPerPixel == 32) { + if (CHECK_BYTE_ORDER(v)) { + /* no byteswapping needed */ + return PIPE_FORMAT_A8R8G8B8_UNORM; + } + else { + return PIPE_FORMAT_B8G8R8A8_UNORM; + } + } + else if ( GET_REDMASK(v) == 0xf800 + && GET_GREENMASK(v) == 0x07e0 + && GET_BLUEMASK(v) == 0x001f + && CHECK_BYTE_ORDER(v) + && v->BitsPerPixel == 16) { + /* 5-6-5 RGB */ + return PIPE_FORMAT_R5G6B5_UNORM; + } + + assert(0); + return 0; +} + + /**********************************************************************/ /***** Linked list of XMesaBuffers *****/ @@ -276,6 +322,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, { XMesaBuffer b; GLframebuffer *fb; + enum pipe_format colorFormat, depthFormat, stencilFormat; ASSERT(type == WINDOW || type == PIXMAP || type == PBUFFER); @@ -289,10 +336,35 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, b->type = type; b->cmap = cmap; + /* determine PIPE_FORMATs for buffers */ + colorFormat = choose_pixel_format(vis); + + if (vis->mesa_visual.depthBits == 0) + depthFormat = PIPE_FORMAT_NONE; + else if (vis->mesa_visual.depthBits <= 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (vis->mesa_visual.depthBits <= 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_Z32_UNORM; + + if (vis->mesa_visual.stencilBits == 8) { + if (depthFormat == PIPE_FORMAT_S8Z24_UNORM) + stencilFormat = depthFormat; + else + stencilFormat = PIPE_FORMAT_S8_UNORM; + } + else { + stencilFormat = PIPE_FORMAT_NONE; + } + + /* * Create framebuffer, but we'll plug in our own renderbuffers below. */ - b->stfb = st_create_framebuffer(&vis->mesa_visual, GL_TRUE, (void *) b); + b->stfb = st_create_framebuffer(&vis->mesa_visual, GL_TRUE, + colorFormat, depthFormat, stencilFormat, + (void *) b); fb = &b->stfb->Base; /* @@ -387,52 +459,6 @@ xmesa_free_buffer(XMesaBuffer buffer) /**********************************************************************/ -/** - * Choose the pixel format for the given visual. - * This will tell the gallium driver how to pack pixel data into - * drawing surfaces. - */ -static GLuint -choose_pixel_format(XMesaVisual v) -{ - if ( GET_REDMASK(v) == 0x0000ff - && GET_GREENMASK(v) == 0x00ff00 - && GET_BLUEMASK(v) == 0xff0000 - && v->BitsPerPixel == 32) { - if (CHECK_BYTE_ORDER(v)) { - /* no byteswapping needed */ - return 0 /* PIXEL_FORMAT_U_A8_B8_G8_R8 */; - } - else { - return PIPE_FORMAT_R8G8B8A8_UNORM; - } - } - else if ( GET_REDMASK(v) == 0xff0000 - && GET_GREENMASK(v) == 0x00ff00 - && GET_BLUEMASK(v) == 0x0000ff - && v->BitsPerPixel == 32) { - if (CHECK_BYTE_ORDER(v)) { - /* no byteswapping needed */ - return PIPE_FORMAT_A8R8G8B8_UNORM; - } - else { - return PIPE_FORMAT_B8G8R8A8_UNORM; - } - } - else if ( GET_REDMASK(v) == 0xf800 - && GET_GREENMASK(v) == 0x07e0 - && GET_BLUEMASK(v) == 0x001f - && CHECK_BYTE_ORDER(v) - && v->BitsPerPixel == 16) { - /* 5-6-5 RGB */ - return PIPE_FORMAT_R5G6B5_UNORM; - } - - assert(0); - return 0; -} - - /** * When a context is bound for the first time, we can finally finish * initializing the context's visual and buffer information. diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 047de412e3..254740ff20 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -87,9 +87,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); - uint type = strb->screenSurface ? PIPE_SCREEN_SURFACE : PIPE_SURFACE; - const enum pipe_format pipeFormat - = st_choose_pipe_format(pipe, internalFormat, GL_NONE, GL_NONE, type); + enum pipe_format pipeFormat; GLbitfield flags = 0x0; /* XXX needed? */ if (!strb->surface) { @@ -106,6 +104,18 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipe->winsys->buffer_reference(pipe->winsys, &strb->surface->buffer, NULL); + /* Determine surface format here */ + if (strb->format != PIPE_FORMAT_NONE) { + assert(strb->format != 0); + /* we'll hit this for front/back color bufs */ + pipeFormat = strb->format; + } + else { + pipeFormat = st_choose_renderbuffer_format(pipe, internalFormat); + } + + init_renderbuffer_bits(strb, pipeFormat); + pipe->winsys->surface_alloc_storage(pipe->winsys, strb->surface, width, @@ -185,6 +195,7 @@ st_new_renderbuffer(GLcontext *ctx, GLuint name) strb->Base.Delete = st_renderbuffer_delete; strb->Base.AllocStorage = st_renderbuffer_alloc_storage; strb->Base.GetPointer = null_get_pointer; + strb->format = PIPE_FORMAT_NONE; return &strb->Base; } return NULL; @@ -193,12 +204,10 @@ st_new_renderbuffer(GLcontext *ctx, GLuint name) /** * Allocate a renderbuffer for a an on-screen window (not a user-created - * renderbuffer). The window system code determines the internal format. - * \param screenSurface indicates if the renderbuffer is a front/back color - * buffer that'll be displayed/copied to the screen + * renderbuffer). The window system code determines the format. */ struct gl_renderbuffer * -st_new_renderbuffer_fb(GLenum intFormat, GLboolean screenSurface) +st_new_renderbuffer_fb(enum pipe_format format) { struct st_renderbuffer *strb; @@ -210,28 +219,37 @@ st_new_renderbuffer_fb(GLenum intFormat, GLboolean screenSurface) _mesa_init_renderbuffer(&strb->Base, 0); strb->Base.ClassID = 0x4242; /* just a unique value */ - strb->Base.InternalFormat = intFormat; - strb->screenSurface = screenSurface; - - switch (intFormat) { - case GL_RGB5: - case GL_RGBA8: - case GL_RGBA16: + strb->format = format; + + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_B8G8R8A8_UNORM: + case PIPE_FORMAT_A1R5G5B5_UNORM: + case PIPE_FORMAT_A4R4G4B4_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + strb->Base.InternalFormat = GL_RGBA; strb->Base._BaseFormat = GL_RGBA; break; - case GL_DEPTH_COMPONENT16: - case GL_DEPTH_COMPONENT32: + case PIPE_FORMAT_Z16_UNORM: + strb->Base.InternalFormat = GL_DEPTH_COMPONENT16; strb->Base._BaseFormat = GL_DEPTH_COMPONENT; break; - case GL_DEPTH24_STENCIL8_EXT: + case PIPE_FORMAT_Z32_UNORM: + strb->Base.InternalFormat = GL_DEPTH_COMPONENT32; + strb->Base._BaseFormat = GL_DEPTH_COMPONENT; + break; + case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: + strb->Base.InternalFormat = GL_DEPTH24_STENCIL8_EXT; strb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; break; - case GL_STENCIL_INDEX8_EXT: - strb->Base._BaseFormat = GL_STENCIL_INDEX; + case PIPE_FORMAT_R16G16B16A16_SNORM: + strb->Base.InternalFormat = GL_RGBA16; + strb->Base._BaseFormat = GL_RGBA; break; default: _mesa_problem(NULL, - "Unexpected intFormat in st_new_renderbuffer"); + "Unexpected format in st_new_renderbuffer_fb"); return NULL; } @@ -248,6 +266,7 @@ st_new_renderbuffer_fb(GLenum intFormat, GLboolean screenSurface) + /** * Called via ctx->Driver.BindFramebufferEXT(). */ diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index bd85bfc549..21e531d1d0 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -39,7 +39,7 @@ struct st_renderbuffer { struct gl_renderbuffer Base; struct pipe_surface *surface; - GLboolean screenSurface; /**< A front/back colorbuffer? */ + enum pipe_format format; /** preferred format, or PIPE_FORMAT_NONE */ }; @@ -51,8 +51,7 @@ st_renderbuffer(struct gl_renderbuffer *rb) extern struct gl_renderbuffer * -st_new_renderbuffer_fb(GLenum intFormat, GLboolean screen_surface); - +st_new_renderbuffer_fb(enum pipe_format format); extern void st_init_fbo_functions(struct dd_function_table *functions); diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 287a0054b9..2a23445ca2 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -283,9 +283,9 @@ static GLuint default_rgba_format(struct pipe_context *pipe, uint type) { static const enum pipe_format colorFormats[] = { - PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_B8G8R8A8_UNORM, + PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_R5G6B5_UNORM }; uint i; @@ -334,56 +334,22 @@ default_depth_format(struct pipe_context *pipe, uint type) /** - * Choose the PIPE_FORMAT_ to use for storing a texture image based - * on the user's internalFormat, format and type parameters. - * We query the pipe device for a list of formats which it supports - * and choose from them. - * If we find a device that needs a more intricate selection mechanism, - * this function _could_ get pushed down into the pipe device. - * - * Note: also used for glRenderbufferStorageEXT() - * - * Note: format and type may be GL_NONE (see renderbuffers) + * Choose the PIPE_FORMAT_ to use for user-created renderbuffers. * * \return PIPE_FORMAT_NONE if error/problem. */ -GLuint -st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, - GLenum format, GLenum type, uint surfType) +enum pipe_format +st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) { - assert(surfType == PIPE_TEXTURE || - surfType == PIPE_SURFACE || - surfType == PIPE_SCREEN_SURFACE); + uint surfType = PIPE_SURFACE; switch (internalFormat) { case 4: case GL_RGBA: case GL_COMPRESSED_RGBA: - if (format == GL_BGRA) { - if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A8R8G8B8_UNORM, surfType )) - return PIPE_FORMAT_A8R8G8B8_UNORM; - } - else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A4R4G4B4_UNORM, surfType )) - return PIPE_FORMAT_A4R4G4B4_UNORM; - } - else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) - return PIPE_FORMAT_A1R5G5B5_UNORM; - } - } - return default_rgba_format( pipe, surfType ); - case 3: case GL_RGB: case GL_COMPRESSED_RGB: - if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_R5G6B5_UNORM, surfType )) - return PIPE_FORMAT_R5G6B5_UNORM; - } - return default_rgba_format( pipe, surfType ); - case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: @@ -434,7 +400,7 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8, surfType )) + if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_L8, surfType )) return PIPE_FORMAT_U_A8; return default_rgba_format( pipe, surfType ); @@ -461,18 +427,10 @@ st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, return PIPE_FORMAT_U_I8; return default_rgba_format( pipe, surfType ); +#if 0 + /* not supported for renderbuffers */ case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_YCBCR, surfType )) - return PIPE_FORMAT_YCBCR; - } - else { - if (pipe->is_format_supported( pipe, PIPE_FORMAT_YCBCR_REV, surfType )) - return PIPE_FORMAT_YCBCR_REV; - } return PIPE_FORMAT_NONE; - -#if 0 case GL_COMPRESSED_RGB_FXT1_3DFX: return &_mesa_texformat_rgb_fxt1; case GL_COMPRESSED_RGBA_FXT1_3DFX: diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index ebff7c5b91..c9a11de504 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -63,9 +63,8 @@ extern enum pipe_format st_mesa_format_to_pipe_format(GLuint mesaFormat); -extern GLuint -st_choose_pipe_format(struct pipe_context *pipe, GLint internalFormat, - GLenum format, GLenum type, uint surfType); +extern enum pipe_format +st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat); extern const struct gl_texture_format * diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 454306b874..b81a894ef1 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -35,15 +35,16 @@ #include "st_cb_fbo.h" -struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, - boolean createRenderbuffers, - void *private) +struct st_framebuffer * +st_create_framebuffer( const __GLcontextModes *visual, + boolean createRenderbuffers, /* XXX remove? */ + enum pipe_format colorFormat, + enum pipe_format depthFormat, + enum pipe_format stencilFormat, + void *private) { - struct st_framebuffer *stfb - = CALLOC_STRUCT(st_framebuffer); + struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer); if (stfb) { - GLenum rgbFormat = (visual->redBits == 5 ? GL_RGB5 : GL_RGBA8); - _mesa_initialize_framebuffer(&stfb->Base, visual); if (createRenderbuffers) { @@ -52,20 +53,20 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, /* XXX allocation should only happen in the unusual case it's actually needed */ struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(rgbFormat, GL_TRUE); + = st_new_renderbuffer_fb(colorFormat); _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb); } if (visual->doubleBufferMode) { struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(rgbFormat, GL_TRUE); + = st_new_renderbuffer_fb(colorFormat); _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); } if (visual->depthBits == 24 && visual->stencilBits == 8) { /* combined depth/stencil buffer */ struct gl_renderbuffer *depthStencilRb - = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT, GL_FALSE); + = st_new_renderbuffer_fb(depthFormat); /* note: bind RB to two attachment points */ _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb); @@ -76,26 +77,26 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, if (visual->depthBits == 32) { /* 32-bit depth buffer */ struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT32, GL_FALSE); + = st_new_renderbuffer_fb(depthFormat); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } else if (visual->depthBits == 24) { /* 24-bit depth buffer, ignore stencil bits */ struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(GL_DEPTH24_STENCIL8_EXT, GL_FALSE); + = st_new_renderbuffer_fb(depthFormat); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } else if (visual->depthBits > 0) { /* 16-bit depth buffer */ struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(GL_DEPTH_COMPONENT16, GL_FALSE); + = st_new_renderbuffer_fb(depthFormat); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } if (visual->stencilBits > 0) { /* 8-bit stencil */ struct gl_renderbuffer *stencilRb - = st_new_renderbuffer_fb(GL_STENCIL_INDEX8_EXT, GL_FALSE); + = st_new_renderbuffer_fb(stencilFormat); _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb); } } @@ -103,7 +104,7 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, if (visual->accumRedBits > 0) { /* 16-bit/channel accum */ struct gl_renderbuffer *accumRb - = st_new_renderbuffer_fb(GL_RGBA16, GL_FALSE); + = st_new_renderbuffer_fb(PIPE_FORMAT_R16G16B16A16_SNORM); _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb); } } diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 307635936b..558b20f1e8 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -29,6 +29,7 @@ #define ST_PUBLIC_H #include "pipe/p_compiler.h" +#include "pipe/p_format.h" #define ST_SURFACE_FRONT_LEFT 0 @@ -55,6 +56,9 @@ void st_copy_context_state(struct st_context *dst, struct st_context *src, struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, boolean createRenderbuffers, + enum pipe_format colorFormat, + enum pipe_format depthFormat, + enum pipe_format stencilFormat, void *privateData); void st_resize_framebuffer( struct st_framebuffer *stfb, -- cgit v1.2.3 From 708251558c4eb1719a0f73d83c233c7a492f2297 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 15:24:01 -0700 Subject: PIPE_SCREEN_SURFACE no longer needed --- src/mesa/pipe/cell/ppu/cell_context.c | 2 -- src/mesa/pipe/i915simple/i915_context.c | 12 ++---------- src/mesa/pipe/i965simple/brw_context.c | 4 +++- src/mesa/pipe/p_defines.h | 1 - src/mesa/pipe/softpipe/sp_context.c | 4 +--- 5 files changed, 6 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 966f355bb3..eff33da969 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -60,8 +60,6 @@ cell_is_format_supported( struct pipe_context *pipe, case PIPE_SURFACE: /* cell supports all (off-screen) surface formats, XXX for now */ return TRUE; - case PIPE_SCREEN_SURFACE: - return format == cell->winsys->preferredFormat; default: assert(0); return FALSE; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 0e3d6d64e4..1c6b8cbb05 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -41,7 +41,7 @@ /** * Query format support for creating a texture, drawing surface, etc. * \param format the format to test - * \param type one of PIPE_TEXTURE, PIPE_SURFACE, PIPE_SCREEN_SURFACE + * \param type one of PIPE_TEXTURE, PIPE_SURFACE */ static boolean i915_is_format_supported( struct pipe_context *pipe, @@ -64,12 +64,7 @@ i915_is_format_supported( struct pipe_context *pipe, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_R5G6B5_UNORM, PIPE_FORMAT_S8Z24_UNORM, - PIPE_FORMAT_R16G16B16A16_SNORM, - PIPE_FORMAT_NONE /* list terminator */ - }; - static const enum pipe_format screen_surface_supported[] = { - PIPE_FORMAT_A8R8G8B8_UNORM, - PIPE_FORMAT_R5G6B5_UNORM, + /*PIPE_FORMAT_R16G16B16A16_SNORM,*/ PIPE_FORMAT_NONE /* list terminator */ }; const enum pipe_format *list; @@ -82,9 +77,6 @@ i915_is_format_supported( struct pipe_context *pipe, case PIPE_SURFACE: list = surface_supported; break; - case PIPE_SCREEN_SURFACE: - list = screen_surface_supported; - break; default: assert(0); } diff --git a/src/mesa/pipe/i965simple/brw_context.c b/src/mesa/pipe/i965simple/brw_context.c index 13bed97ff2..e69ba6938e 100644 --- a/src/mesa/pipe/i965simple/brw_context.c +++ b/src/mesa/pipe/i965simple/brw_context.c @@ -134,7 +134,7 @@ brw_get_paramf(struct pipe_context *pipe, int param) static boolean brw_is_format_supported( struct pipe_context *pipe, - uint format ) + enum pipe_format format, uint type ) { #if 0 /* XXX: This is broken -- rewrite if still needed. */ @@ -190,6 +190,8 @@ brw_is_format_supported( struct pipe_context *pipe, case PIPE_FORMAT_R5G6B5_UNORM: case PIPE_FORMAT_S8Z24_UNORM: return TRUE; + default: + return FALSE; }; return FALSE; #endif diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index d3afef95b4..a67ed60547 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -165,7 +165,6 @@ */ #define PIPE_TEXTURE 1 #define PIPE_SURFACE 2 /**< user-created surfaces */ -#define PIPE_SCREEN_SURFACE 3 /**< On-screen front/back colorbuffer */ /** diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 107c8f8597..dcf0444f6e 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -49,7 +49,7 @@ /** * Query format support for creating a texture, drawing surface, etc. * \param format the format to test - * \param type one of PIPE_TEXTURE, PIPE_SURFACE, PIPE_SCREEN_SURFACE + * \param type one of PIPE_TEXTURE, PIPE_SURFACE */ static boolean softpipe_is_format_supported( struct pipe_context *pipe, @@ -64,8 +64,6 @@ softpipe_is_format_supported( struct pipe_context *pipe, case PIPE_SURFACE: /* softpipe supports all (off-screen) surface formats */ return TRUE; - case PIPE_SCREEN_SURFACE: - return softpipe->winsys->is_format_supported( softpipe->winsys, format ); default: assert(0); return FALSE; -- cgit v1.2.3 From a4ca2591f67a5d088fb7acc56d3d6d98aede753a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 16:00:11 -0700 Subject: Move float/rgba tile get/put functions into new file. This should be moved into a gallium util lib (location tbd) so it can also be used by the state tracker. Then, remove the pipe->get/put_tile_rgba() functions. --- src/mesa/pipe/i915simple/i915_surface.c | 7 + src/mesa/pipe/softpipe/Makefile | 1 + src/mesa/pipe/softpipe/sp_rgba_tile.c | 727 ++++++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_rgba_tile.h | 48 +++ src/mesa/pipe/softpipe/sp_surface.c | 667 +---------------------------- 5 files changed, 784 insertions(+), 666 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_rgba_tile.c create mode 100644 src/mesa/pipe/softpipe/sp_rgba_tile.h (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 3db4093d22..b61ad17561 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -34,6 +34,8 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" +#include "pipe/softpipe/sp_rgba_tile.h" /* XXX TEMPORARY */ + #define CLIP_TILE \ do { \ @@ -420,8 +422,13 @@ i915_init_surface_functions(struct i915_context *i915) i915->pipe.get_tex_surface = i915_get_tex_surface; i915->pipe.get_tile = i915_get_tile; i915->pipe.put_tile = i915_put_tile; +#if 0 i915->pipe.get_tile_rgba = i915_get_tile_rgba; i915->pipe.put_tile_rgba = i915_put_tile_rgba; +#else + i915->pipe.get_tile_rgba = softpipe_get_tile_rgba; + i915->pipe.put_tile_rgba = softpipe_put_tile_rgba; +#endif i915->pipe.surface_data = i915_surface_data; i915->pipe.surface_copy = i915_surface_copy; i915->pipe.surface_fill = i915_surface_fill; diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 31438a882e..7608ff126a 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -25,6 +25,7 @@ DRIVER_SOURCES = \ sp_quad_output.c \ sp_quad_stencil.c \ sp_quad_stipple.c \ + sp_rgba_tile.c \ sp_state_blend.c \ sp_state_clip.c \ sp_state_derived.c \ diff --git a/src/mesa/pipe/softpipe/sp_rgba_tile.c b/src/mesa/pipe/softpipe/sp_rgba_tile.c new file mode 100644 index 0000000000..71332746fb --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_rgba_tile.c @@ -0,0 +1,727 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * RGBA/float tile get/put functions. + * XXX eventually this should go into a utility library usable by + * the state tracker too. + * Then, remove pipe->get/put_tile_rgba() + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "sp_surface.h" +#include "sp_rgba_tile.h" + + +/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */ +#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) + +#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ + us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) + + + +#define CLIP_TILE \ + do { \ + if (x >= ps->width) \ + return; \ + if (y >= ps->height) \ + return; \ + if (x + w > ps->width) \ + w = ps->width - x; \ + if (y + h > ps->height) \ + h = ps->height - y; \ + } while(0) + + +/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ + +static void +a8r8g8b8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const unsigned *src + = ((const unsigned *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const unsigned pixel = src[j]; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +static void +a8r8g8b8_put_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + unsigned *dst + = ((unsigned *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + dst[j] = (a << 24) | (r << 16) | (g << 8) | b; + pRow += 4; + } + dst += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ + +static void +b8g8r8a8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const unsigned *src + = ((const unsigned *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const unsigned pixel = src[j]; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +static void +b8g8r8a8_put_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + unsigned *dst + = ((unsigned *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + dst[j] = (b << 24) | (g << 16) | (r << 8) | a; + pRow += 4; + } + dst += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ + +static void +a1r5g5b5_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + + assert(ps->format == PIPE_FORMAT_A1R5G5B5_UNORM); + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const ushort pixel = src[j]; + p[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); + p[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); + p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + p[3] = ((pixel >> 15) ) * 1.0f; + p += 4; + } + src += ps->pitch; + } +} + + +/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ + +static void +a4r4g4b4_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + + assert(ps->format == PIPE_FORMAT_A4R4G4B4_UNORM); + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const ushort pixel = src[j]; + p[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); + p[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); + p[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); + p[3] = ((pixel >> 12) ) * (1.0f / 15.0f); + p += 4; + } + src += ps->pitch; + } +} + + +/*** PIPE_FORMAT_R5G6B5_UNORM ***/ + +static void +r5g6b5_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + + assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const ushort pixel = src[j]; + p[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); + p[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); + p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + p[3] = 1.0f; + p += 4; + } + src += ps->pitch; + } +} + + +/*** PIPE_FORMAT_Z16_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z16_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + const float scale = 1.0f / 65535.0f; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_Z16_UNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = src[j] * scale; + } + src += ps->pitch; + p += 4 * w0; + } +} + + + + +/*** PIPE_FORMAT_U_L8 ***/ + +static void +l8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ubyte *src + = ((const ubyte *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_L8); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(src[j]); + pRow[3] = 1.0; + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_U_A8 ***/ + +static void +a8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ubyte *src + = ((const ubyte *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_A8); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[0] = + pRow[1] = + pRow[2] = 0.0; + pRow[3] = UBYTE_TO_FLOAT(src[j]); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ + +static void +r16g16b16a16_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const short *src + = ((const short *) (ps->map)) + + (y * ps->pitch + x) * 4; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + const short *pixel = src; + for (j = 0; j < w; j++) { + pRow[0] = SHORT_TO_FLOAT(pixel[0]); + pRow[1] = SHORT_TO_FLOAT(pixel[1]); + pRow[2] = SHORT_TO_FLOAT(pixel[2]); + pRow[3] = SHORT_TO_FLOAT(pixel[3]); + pRow += 4; + pixel += 4; + } + src += ps->pitch * 4; + p += w0 * 4; + } +} + + +static void +r16g16b16a16_put_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + short *dst + = ((short *) (ps->map)) + + (y * ps->pitch + x) * 4; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + short r, g, b, a; + UNCLAMPED_FLOAT_TO_SHORT(r, pRow[0]); + UNCLAMPED_FLOAT_TO_SHORT(g, pRow[1]); + UNCLAMPED_FLOAT_TO_SHORT(b, pRow[2]); + UNCLAMPED_FLOAT_TO_SHORT(a, pRow[3]); + dst[j*4+0] = r; + dst[j*4+1] = g; + dst[j*4+2] = b; + dst[j*4+3] = a; + pRow += 4; + } + dst += ps->pitch * 4; + p += w0 * 4; + } +} + + + +/*** PIPE_FORMAT_U_I8 ***/ + +static void +i8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ubyte *src + = ((const ubyte *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_I8); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = UBYTE_TO_FLOAT(src[j]); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_U_A8_L8 ***/ + +static void +a8_l8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_A8_L8); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const ushort p = src[j]; + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(p & 0xff); + pRow[3] = UBYTE_TO_FLOAT(p >> 8); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + + + +/*** PIPE_FORMAT_Z32_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z32_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const uint *src + = ((const uint *) (ps->map)) + + y * ps->pitch + x; + const double scale = 1.0 / (double) 0xffffffff; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_Z16_UNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = (float) (scale * src[j]); + } + src += ps->pitch; + p += 4 * w0; + } +} + + +/*** PIPE_FORMAT_S8Z24_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +s8z24_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const uint *src + = ((const uint *) (ps->map)) + + y * ps->pitch + x; + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_S8Z24_UNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = (float) (scale * (src[j] & 0xffffff)); + } + src += ps->pitch; + p += 4 * w0; + } +} + + +/*** PIPE_FORMAT_Z24S8_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +z24s8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const uint *src + = ((const uint *) (ps->map)) + + y * ps->pitch + x; + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_Z24S8_UNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = (float) (scale * (src[j] >> 8)); + } + src += ps->pitch; + p += 4 * w0; + } +} + + +void +softpipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p) +{ + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + a1r5g5b5_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_A4R4G4B4_UNORM: + a4r4g4b4_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g6b5_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_L8: + l8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_A8: + a8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_I8: + i8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_A8_L8: + a8_l8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_Z16_UNORM: + z16_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_Z32_UNORM: + z32_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_S8Z24_UNORM: + s8z24_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_Z24S8_UNORM: + z24s8_get_tile_rgba(ps, x, y, w, h, p); + break; + default: + assert(0); + } +} + + +void +softpipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p) +{ + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_put_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_put_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + /*a1r5g5b5_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_R8G8B8A8_UNORM: + /* XXX need these */ + fprintf(stderr, "unsup pipe format in softpipe_put_tile_rgba()\n"); + break; + case PIPE_FORMAT_U_L8: + /*l8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_A8: + /*a8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_I8: + /*i8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_A8_L8: + /*a8_l8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_put_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_Z16_UNORM: + /*z16_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_Z32_UNORM: + /*z32_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_S8Z24_UNORM: + /*s8z24_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_Z24S8_UNORM: + /*z24s8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + default: + assert(0); + } +} diff --git a/src/mesa/pipe/softpipe/sp_rgba_tile.h b/src/mesa/pipe/softpipe/sp_rgba_tile.h new file mode 100644 index 0000000000..ffa1b65de1 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_rgba_tile.h @@ -0,0 +1,48 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_RGBA_TILE_H +#define SP_RGBA_TILE_H + +#include "pipe/p_compiler.h" + +struct pipe_context; +struct pipe_surface; + +extern void +softpipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p); + +extern void +softpipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p); + +#endif diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index c81e0f9c18..2a6909007f 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -32,18 +32,7 @@ #include "sp_context.h" #include "sp_surface.h" #include "sp_texture.h" - - -/** - * Softpipe surface functions. - */ - - -/** Convert GLshort in [-32768,32767] to GLfloat in [-1.0,1.0] */ -#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) - -#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ - us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) +#include "sp_rgba_tile.h" @@ -60,549 +49,6 @@ } while(0) -/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ - -static void -a8r8g8b8_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const unsigned *src - = ((const unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -static void -a8r8g8b8_put_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) -{ - unsigned *dst - = ((unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - dst[j] = (a << 24) | (r << 16) | (g << 8) | b; - pRow += 4; - } - dst += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ - -static void -b8g8r8a8_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const unsigned *src - = ((const unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -static void -b8g8r8a8_put_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) -{ - unsigned *dst - = ((unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - dst[j] = (b << 24) | (g << 16) | (r << 8) | a; - pRow += 4; - } - dst += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ - -static void -a1r5g5b5_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - - assert(ps->format == PIPE_FORMAT_A1R5G5B5_UNORM); - - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - const ushort pixel = src[j]; - p[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); - p[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); - p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - p[3] = ((pixel >> 15) ) * 1.0f; - p += 4; - } - src += ps->pitch; - } -} - - -/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ - -static void -a4r4g4b4_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - - assert(ps->format == PIPE_FORMAT_A4R4G4B4_UNORM); - - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - const ushort pixel = src[j]; - p[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); - p[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); - p[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); - p[3] = ((pixel >> 12) ) * (1.0f / 15.0f); - p += 4; - } - src += ps->pitch; - } -} - - -/*** PIPE_FORMAT_R5G6B5_UNORM ***/ - -static void -r5g6b5_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - - assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); - - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - const ushort pixel = src[j]; - p[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); - p[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); - p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - p[3] = 1.0f; - p += 4; - } - src += ps->pitch; - } -} - - -/*** PIPE_FORMAT_Z16_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z16_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - const float scale = 1.0f / 65535.0f; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_Z16_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = src[j] * scale; - } - src += ps->pitch; - p += 4 * w0; - } -} - - - - -/*** PIPE_FORMAT_U_L8 ***/ - -static void -l8_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const ubyte *src - = ((const ubyte *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_L8); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(src[j]); - pRow[3] = 1.0; - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_U_A8 ***/ - -static void -a8_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const ubyte *src - = ((const ubyte *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_A8); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[0] = - pRow[1] = - pRow[2] = 0.0; - pRow[3] = UBYTE_TO_FLOAT(src[j]); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ - -static void -r16g16b16a16_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const short *src - = ((const short *) (ps->map)) - + (y * ps->pitch + x) * 4; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - const short *pixel = src; - for (j = 0; j < w; j++) { - pRow[0] = SHORT_TO_FLOAT(pixel[0]); - pRow[1] = SHORT_TO_FLOAT(pixel[1]); - pRow[2] = SHORT_TO_FLOAT(pixel[2]); - pRow[3] = SHORT_TO_FLOAT(pixel[3]); - pRow += 4; - pixel += 4; - } - src += ps->pitch * 4; - p += w0 * 4; - } -} - - -static void -r16g16b16a16_put_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) -{ - short *dst - = ((short *) (ps->map)) - + (y * ps->pitch + x) * 4; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++) { - short r, g, b, a; - UNCLAMPED_FLOAT_TO_SHORT(r, pRow[0]); - UNCLAMPED_FLOAT_TO_SHORT(g, pRow[1]); - UNCLAMPED_FLOAT_TO_SHORT(b, pRow[2]); - UNCLAMPED_FLOAT_TO_SHORT(a, pRow[3]); - dst[j*4+0] = r; - dst[j*4+1] = g; - dst[j*4+2] = b; - dst[j*4+3] = a; - pRow += 4; - } - dst += ps->pitch * 4; - p += w0 * 4; - } -} - - - -/*** PIPE_FORMAT_U_I8 ***/ - -static void -i8_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const ubyte *src - = ((const ubyte *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_I8); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = UBYTE_TO_FLOAT(src[j]); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_U_A8_L8 ***/ - -static void -a8_l8_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_A8_L8); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const ushort p = src[j]; - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(p & 0xff); - pRow[3] = UBYTE_TO_FLOAT(p >> 8); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - - - -/*** PIPE_FORMAT_Z32_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z32_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const uint *src - = ((const uint *) (ps->map)) - + y * ps->pitch + x; - const double scale = 1.0 / (double) 0xffffffff; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_Z16_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (float) (scale * src[j]); - } - src += ps->pitch; - p += 4 * w0; - } -} - - -/*** PIPE_FORMAT_S8Z24_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -s8z24_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const uint *src - = ((const uint *) (ps->map)) - + y * ps->pitch + x; - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_S8Z24_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (float) (scale * (src[j] & 0xffffff)); - } - src += ps->pitch; - p += 4 * w0; - } -} - - -/*** PIPE_FORMAT_Z24S8_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -z24s8_get_tile(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, float *p) -{ - const uint *src - = ((const uint *) (ps->map)) - + y * ps->pitch + x; - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_Z24S8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (float) (scale * (src[j] >> 8)); - } - src += ps->pitch; - p += 4 * w0; - } -} - - /** * Called via pipe->get_tex_surface() * XXX is this in the right place? @@ -710,117 +156,6 @@ softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, } -/* XXX TEMPORARY */ -void -softpipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p) -{ - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - a1r5g5b5_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_A4R4G4B4_UNORM: - a4r4g4b4_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_R5G6B5_UNORM: - r5g6b5_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_U_L8: - l8_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_U_A8: - a8_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_U_I8: - i8_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_U_A8_L8: - a8_l8_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_Z16_UNORM: - z16_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_Z32_UNORM: - z32_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_S8Z24_UNORM: - s8z24_get_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_Z24S8_UNORM: - z24s8_get_tile(ps, x, y, w, h, p); - break; - default: - assert(0); - } -} - - -/* XXX TEMPORARY */ -void -softpipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p) -{ - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_put_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_put_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - /*a1r5g5b5_put_tile(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_R8G8B8A8_UNORM: - /* XXX need these */ - fprintf(stderr, "unsup pipe format in softpipe_put_tile_rgba()\n"); - break; - case PIPE_FORMAT_U_L8: - /*l8_put_tile(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_U_A8: - /*a8_put_tile(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_U_I8: - /*i8_put_tile(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_U_A8_L8: - /*a8_l8_put_tile(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_put_tile(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_Z16_UNORM: - /*z16_put_tile(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_Z32_UNORM: - /*z32_put_tile(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_S8Z24_UNORM: - /*s8z24_put_tile(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_Z24S8_UNORM: - /*z24s8_put_tile(ps, x, y, w, h, p);*/ - break; - default: - assert(0); - } -} - - /** * Copy 2D rect from one place to another. * Position and sizes are in pixels. -- cgit v1.2.3 From 034476fc78b626915e0c597b08df437ea59e22c8 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 16:04:52 -0700 Subject: remove #include --- src/mesa/pipe/softpipe/sp_rgba_tile.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_rgba_tile.c b/src/mesa/pipe/softpipe/sp_rgba_tile.c index 71332746fb..9d8d6a79e0 100644 --- a/src/mesa/pipe/softpipe/sp_rgba_tile.c +++ b/src/mesa/pipe/softpipe/sp_rgba_tile.c @@ -36,7 +36,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "sp_surface.h" + #include "sp_rgba_tile.h" -- cgit v1.2.3 From 4416514aa82e59eae559b59e49846931fdf85f56 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 16:05:38 -0700 Subject: Remove get/put_tile_rgba() functions. Temporarily, use the functions from sp_rgba_tile.c --- src/mesa/pipe/i915simple/i915_surface.c | 103 +------------------------------- src/mesa/pipe/i965simple/brw_surface.c | 73 ++-------------------- 2 files changed, 5 insertions(+), 171 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index b61ad17561..d4c02c5359 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -50,103 +50,6 @@ } while(0) -/** - * Note: this is exactly like a8r8g8b8_get_tile() in sp_surface.c - * Share it someday. - */ -/** XXX this will go away eventually */ -static void -i915_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *p) -{ - const unsigned *src - = ((const unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - CLIP_TILE; - - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } - break; - case PIPE_FORMAT_S8Z24_UNORM: - { - const float scale = 1.0f / (float) 0xffffff; - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (pixel & 0xffffff) * scale; - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } - } - break; - default: - assert(0); - } -} - - -/** XXX this will go away eventually */ -static void -i915_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *p) -{ - unsigned *dst - = ((unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); - - CLIP_TILE; - - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - dst[j] = (a << 24) | (r << 16) | (g << 8) | b; - pRow += 4; - } - dst += ps->pitch; - p += w0 * 4; - } - break; - default: - assert(0); - } -} - - /* * XXX note: same as code in sp_surface.c */ @@ -422,13 +325,9 @@ i915_init_surface_functions(struct i915_context *i915) i915->pipe.get_tex_surface = i915_get_tex_surface; i915->pipe.get_tile = i915_get_tile; i915->pipe.put_tile = i915_put_tile; -#if 0 - i915->pipe.get_tile_rgba = i915_get_tile_rgba; - i915->pipe.put_tile_rgba = i915_put_tile_rgba; -#else i915->pipe.get_tile_rgba = softpipe_get_tile_rgba; i915->pipe.put_tile_rgba = softpipe_put_tile_rgba; -#endif + i915->pipe.surface_data = i915_surface_data; i915->pipe.surface_copy = i915_surface_copy; i915->pipe.surface_fill = i915_surface_fill; diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c index b5a0318047..b0d2fec8e0 100644 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -33,6 +33,8 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" +#include "pipe/softpipe/sp_rgba_tile.h" /* XXX TEMPORARY */ + #define CLIP_TILE \ do { \ @@ -47,73 +49,6 @@ } while(0) -/** - * Note: this is exactly like a8r8g8b8_get_tile() in sp_surface.c - * Share it someday. - */ -static void -brw_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *p) -{ - const unsigned *src - = ((const unsigned *) (ps->map + ps->offset)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - CLIP_TILE; - - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } - break; - case PIPE_FORMAT_S8Z24_UNORM: - { - const float scale = 1.0f / (float) 0xffffff; - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (pixel & 0xffffff) * scale; - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } - } - break; - default: - assert(0); - } -} - - -static void -brw_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *p) -{ - /* TODO */ - assert(0); -} - - /* * XXX note: same as code in sp_surface.c */ @@ -388,8 +323,8 @@ brw_init_surface_functions(struct brw_context *brw) brw->pipe.get_tex_surface = brw_get_tex_surface; brw->pipe.get_tile = brw_get_tile; brw->pipe.put_tile = brw_put_tile; - brw->pipe.get_tile_rgba = brw_get_tile_rgba; - brw->pipe.put_tile_rgba = brw_put_tile_rgba; + brw->pipe.get_tile_rgba = softpipe_get_tile_rgba; + brw->pipe.put_tile_rgba = softpipe_put_tile_rgba; brw->pipe.surface_data = brw_surface_data; brw->pipe.surface_copy = brw_surface_copy; -- cgit v1.2.3 From 417c2953339bc281f0ac486777afbbd00a385294 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Dec 2007 16:30:49 -0700 Subject: added r5g5b5_put_tile_rgba() --- src/mesa/pipe/softpipe/sp_rgba_tile.c | 35 +++++++++++++++++++++++++++++++++-- 1 file changed, 33 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_rgba_tile.c b/src/mesa/pipe/softpipe/sp_rgba_tile.c index 9d8d6a79e0..c42635a1ad 100644 --- a/src/mesa/pipe/softpipe/sp_rgba_tile.c +++ b/src/mesa/pipe/softpipe/sp_rgba_tile.c @@ -275,6 +275,37 @@ r5g6b5_get_tile_rgba(struct pipe_surface *ps, } +static void +r5g5b5_put_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + ushort *dst + = ((ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); + + CLIP_TILE; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); + uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); + uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); + dst[j] = (r << 11) | (g << 5) | (b); + pRow += 4; + } + dst += ps->pitch; + p += w0 * 4; + } +} + + + /*** PIPE_FORMAT_Z16_UNORM ***/ /** @@ -690,9 +721,9 @@ softpipe_put_tile_rgba(struct pipe_context *pipe, /*a1r5g5b5_put_tile_rgba(ps, x, y, w, h, p);*/ break; case PIPE_FORMAT_R5G6B5_UNORM: + r5g5b5_put_tile_rgba(ps, x, y, w, h, p); + break; case PIPE_FORMAT_R8G8B8A8_UNORM: - /* XXX need these */ - fprintf(stderr, "unsup pipe format in softpipe_put_tile_rgba()\n"); break; case PIPE_FORMAT_U_L8: /*l8_put_tile_rgba(ps, x, y, w, h, p);*/ -- cgit v1.2.3 From da89104c576b5b7a0c581a018a8b1c2d770e13c2 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 13 Dec 2007 10:32:36 -0500 Subject: Actually allocate the output registers. In tgsi the output's will (hopefully) consecutive so this check (besides not being currently implemented) is not necessary. --- src/mesa/pipe/i965simple/brw_vs_emit.c | 38 +++++++++++++++------------------- 1 file changed, 17 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index 59459d4200..530e17a736 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -103,28 +103,26 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c, c->first_output = reg; mrf = 4; for (i = 0; i < c->vp->program.num_outputs; i++) { - if (c->prog_data.outputs_written & (1<nr_outputs++; + c->nr_outputs++; #if 0 - if (i == VERT_RESULT_HPOS) { - c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); - reg++; - } - else if (i == VERT_RESULT_PSIZ) { - c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); - reg++; - mrf++; /* just a placeholder? XXX fix later stages & remove this */ - } - else { - c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf); - mrf++; - } -#else - /* for now stuff everything in grf */ + if (i == VERT_RESULT_HPOS) { c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); reg++; -#endif } + else if (i == VERT_RESULT_PSIZ) { + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); + reg++; + mrf++; /* just a placeholder? XXX fix later stages & remove this */ + } + else { + c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf); + mrf++; + } +#else + /* for now stuff everything in grf */ + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); + reg++; +#endif } /* Allocate program temporaries: @@ -627,11 +625,9 @@ static struct brw_reg get_reg( struct brw_vs_compile *c, case TGSI_FILE_TEMPORARY: case TGSI_FILE_INPUT: case TGSI_FILE_OUTPUT: + case TGSI_FILE_CONSTANT: assert(c->regs[file][index].nr != 0); return c->regs[file][index]; - case TGSI_FILE_CONSTANT: - assert(c->regs[TGSI_FILE_CONSTANT][index].nr != 0); - return c->regs[TGSI_FILE_CONSTANT][index]; case TGSI_FILE_ADDRESS: assert(index == 0); return c->regs[file][index]; -- cgit v1.2.3 From 4844b9fb9d24b119935d540dd345707dd81d9f4a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 10:54:31 +0000 Subject: 965: calculate nr enabled vbos on the fly --- src/mesa/pipe/i965simple/brw_draw_upload.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_draw_upload.c b/src/mesa/pipe/i965simple/brw_draw_upload.c index 186a6274fa..b2168ab78b 100644 --- a/src/mesa/pipe/i965simple/brw_draw_upload.c +++ b/src/mesa/pipe/i965simple/brw_draw_upload.c @@ -207,18 +207,21 @@ static unsigned get_index_type(int type) boolean brw_upload_vertex_buffers( struct brw_context *brw ) { struct brw_array_state vbp; + unsigned nr_enabled = 0; unsigned i; - int nr_enabled = brw->vb.last_vb + 1; memset(&vbp, 0, sizeof(vbp)); /* This is a hardware limit: */ - if (nr_enabled >= BRW_VEP_MAX) - return FALSE; - for (i = 0; i < nr_enabled; i++) + for (i = 0; i < BRW_VEP_MAX; i++) { + if (brw->vb.vbo_array[i] == NULL) { + nr_enabled = i; + break; + } + vbp.vb[i].vb0.bits.pitch = brw->vb.vbo_array[i].pitch; vbp.vb[i].vb0.bits.pad = 0; vbp.vb[i].vb0.bits.access_type = BRW_VERTEXBUFFER_ACCESS_VERTEXDATA; -- cgit v1.2.3 From 1f1506ec94f3d44091de88d8bcd7ab9c20a6e7b2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 10:55:16 +0000 Subject: 965: use state atom array directly --- src/mesa/pipe/i965simple/brw_state_upload.c | 37 ++++------------------------- 1 file changed, 5 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_state_upload.c b/src/mesa/pipe/i965simple/brw_state_upload.c index 1ca7484958..10f1d4812a 100644 --- a/src/mesa/pipe/i965simple/brw_state_upload.c +++ b/src/mesa/pipe/i965simple/brw_state_upload.c @@ -102,34 +102,13 @@ void brw_init_state( struct brw_context *brw ) brw_init_pools(brw); brw_init_caches(brw); - brw->state.atoms = MALLOC(sizeof(atoms)); - brw->state.nr_atoms = sizeof(atoms)/sizeof(*atoms); - memcpy(brw->state.atoms, atoms, sizeof(atoms)); - - /* Patch in a pointer to the dynamic state atom: - */ - for (i = 0; i < brw->state.nr_atoms; i++) - if (brw->state.atoms[i] == NULL) - brw->state.atoms[i] = &brw->curbe.tracked_state; - - memcpy(&brw->curbe.tracked_state, - &brw_constant_buffer, - sizeof(brw_constant_buffer)); - brw->state.dirty.brw = ~0; brw->emit_state_always = 0; - - } void brw_destroy_state( struct brw_context *brw ) { - if (brw->state.atoms) { - FREE(brw->state.atoms); - brw->state.atoms = NULL; - } - brw_destroy_caches(brw); brw_destroy_batch_cache(brw); brw_destroy_pools(brw); @@ -189,21 +168,17 @@ void brw_validate_state( struct brw_context *brw ) memset(&examined, 0, sizeof(examined)); prev = *state; - for (i = 0; i < brw->state.nr_atoms; i++) { - const struct brw_tracked_state *atom = brw->state.atoms[i]; + for (i = 0; i < Elements(atoms); i++) { + const struct brw_tracked_state *atom = atoms[i]; struct brw_state_flags generated; assert(atom->dirty.brw || atom->dirty.cache); assert(atom->update); - if (check_state(state, &atom->dirty) || atom->always_update) { + if (check_state(state, &atom->dirty)) { atom->update( brw ); - -/* emit_foo(brw); */ } - if (atom->emit_reloc != NULL) - atom->emit_reloc(brw); accumulate_state(&examined, &atom->dirty); @@ -218,12 +193,10 @@ void brw_validate_state( struct brw_context *brw ) } else { for (i = 0; i < Elements(atoms); i++) { - const struct brw_tracked_state *atom = brw->state.atoms[i]; + const struct brw_tracked_state *atom = atoms[i]; - if (check_state(state, &atom->dirty) || atom->always_update) + if (check_state(state, &atom->dirty)) atom->update( brw ); - if (atom->emit_reloc != NULL) - atom->emit_reloc(brw); } } -- cgit v1.2.3 From 54f1ce3b273d81c19aea5cc060cfc9d088b003ef Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 10:55:47 +0000 Subject: 965: remove reference to metaops --- src/mesa/pipe/i965simple/brw_vs.c | 10 ---------- 1 file changed, 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_vs.c b/src/mesa/pipe/i965simple/brw_vs.c index 33c6624214..738c6346d5 100644 --- a/src/mesa/pipe/i965simple/brw_vs.c +++ b/src/mesa/pipe/i965simple/brw_vs.c @@ -97,13 +97,6 @@ static void brw_upload_vs_prog( struct brw_context *brw ) key.copy_edgeflag = (brw->attribs.Raster->fill_cw != PIPE_POLYGON_MODE_FILL || brw->attribs.Raster->fill_ccw != PIPE_POLYGON_MODE_FILL); -#if 0 - /* BRW_NEW_METAOPS - */ - if (brw->metaops.active) - key.know_w_is_one = 1; -#endif - /* Make an early check for the key. */ if (brw_search_cache(&brw->cache[BRW_VS_PROG], @@ -120,9 +113,6 @@ static void brw_upload_vs_prog( struct brw_context *brw ) */ const struct brw_tracked_state brw_vs_prog = { .dirty = { -#if 0 - .brw = BRW_NEW_VS | BRW_NEW_METAOPS, -#endif .brw = BRW_NEW_VS, .cache = 0 }, -- cgit v1.2.3 From 7abe2bfefadfdf6dbce44395ab0d206652f5bff4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 10:56:09 +0000 Subject: 965: remove dead code --- src/mesa/pipe/i965simple/brw_wm.c | 143 +++----------------------------------- 1 file changed, 9 insertions(+), 134 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm.c b/src/mesa/pipe/i965simple/brw_wm.c index 65271f22fd..f4e3166e1c 100644 --- a/src/mesa/pipe/i965simple/brw_wm.c +++ b/src/mesa/pipe/i965simple/brw_wm.c @@ -35,86 +35,6 @@ #include "brw_wm.h" #include "brw_state.h" -unsigned brw_wm_nr_args( unsigned opcode ) -{ - switch (opcode) { - - case WM_PIXELXY: - case TGSI_OPCODE_ABS: - case TGSI_OPCODE_FLR: - case TGSI_OPCODE_FRC: - case TGSI_OPCODE_MOV: - case TGSI_OPCODE_COS: - case TGSI_OPCODE_EX2: - case TGSI_OPCODE_LG2: - case TGSI_OPCODE_RCP: - case TGSI_OPCODE_RSQ: - case TGSI_OPCODE_SIN: - case TGSI_OPCODE_SCS: - case TGSI_OPCODE_TEX: - case TGSI_OPCODE_TXB: - case TGSI_OPCODE_TXD: - case TGSI_OPCODE_KIL: - case TGSI_OPCODE_LIT: - case WM_CINTERP: - case WM_WPOSXY: - return 1; - - case TGSI_OPCODE_POW: - case TGSI_OPCODE_SUB: - case TGSI_OPCODE_SGE: - case TGSI_OPCODE_SGT: - case TGSI_OPCODE_SLE: - case TGSI_OPCODE_SLT: - case TGSI_OPCODE_SEQ: - case TGSI_OPCODE_SNE: - case TGSI_OPCODE_ADD: - case TGSI_OPCODE_MAX: - case TGSI_OPCODE_MIN: - case TGSI_OPCODE_MUL: - case TGSI_OPCODE_XPD: - case TGSI_OPCODE_DP3: - case TGSI_OPCODE_DP4: - case TGSI_OPCODE_DPH: - case TGSI_OPCODE_DST: - case WM_LINTERP: - case WM_DELTAXY: - case WM_PIXELW: - return 2; - - case WM_FB_WRITE: - case WM_PINTERP: - case TGSI_OPCODE_MAD: - case TGSI_OPCODE_CMP: - case TGSI_OPCODE_LRP: - return 3; - - default: - return 0; - } -} - - -unsigned brw_wm_is_scalar_result( unsigned opcode ) -{ - switch (opcode) { - case TGSI_OPCODE_COS: - case TGSI_OPCODE_EX2: - case TGSI_OPCODE_LG2: - case TGSI_OPCODE_POW: - case TGSI_OPCODE_RCP: - case TGSI_OPCODE_RSQ: - case TGSI_OPCODE_SIN: - case TGSI_OPCODE_DP3: - case TGSI_OPCODE_DP4: - case TGSI_OPCODE_DPH: - case TGSI_OPCODE_DST: - return 1; - - default: - return 0; - } -} static void do_wm_prog( struct brw_context *brw, @@ -135,51 +55,12 @@ static void do_wm_prog( struct brw_context *brw, memcpy(&c->key, key, sizeof(*key)); c->fp = fp; + fprintf(stderr, "XXXXXXXX FP\n"); -#if 0 - c->env_param = brw->intel.ctx.FragmentProgram.Parameters; - - if (brw_wm_is_glsl(&c->fp->program)) { - brw_wm_glsl_emit(c); - } else - { - /* Augment fragment program. Add instructions for pre- and - * post-fragment-program tasks such as interpolation and fogging. - */ - brw_wm_pass_fp(c); - - /* Translate to intermediate representation. Build register usage - * chains. - */ - brw_wm_pass0(c); - - /* Dead code removal. - */ - brw_wm_pass1(c); - - /* Register allocation. - */ - c->grf_limit = BRW_WM_MAX_GRF/2; - - /* This is where we start emitting gen4 code: - */ - brw_init_compile(&c->func); - - brw_wm_pass2(c); - - c->prog_data.total_grf = c->max_wm_grf; - if (c->last_scratch) { - c->prog_data.total_scratch = - c->last_scratch + 0x40; - } else { - c->prog_data.total_scratch = 0; - } - - /* Emit GEN4 code. - */ - brw_wm_emit(c); - } + + brw_wm_glsl_emit(c); + /* get the program */ program = brw_get_program(&c->func, &program_size); @@ -193,7 +74,6 @@ static void do_wm_prog( struct brw_context *brw, program_size, &c->prog_data, &brw->wm.prog_data ); -#endif } @@ -274,14 +154,10 @@ static void brw_wm_populate_key( struct brw_context *brw, #if 0 - /* BRW_NEW_WM_INPUT_DIMENSIONS */ - key->projtex_mask = brw->wm.input_size_masks[4-1] >> (FRAG_ATTRIB_TEX0 - FRAG_ATTRIB_WPOS); -#endif - - /* _NEW_LIGHT */ - key->flat_shade = (brw->attribs.Raster->flatshade); - - /* _NEW_TEXTURE */ + /* BRW_NEW_SAMPLER + * + * Not doing any of this at the moment: + */ for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { const struct pipe_sampler_state *unit = brw->attribs.Samplers[i]; @@ -291,12 +167,11 @@ static void brw_wm_populate_key( struct brw_context *brw, unit->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { key->shadowtex_mask |= 1<Image[0][t->BaseLevel]->InternalFormat == GL_YCBCR_MESA) key->yuvtex_mask |= 1< Date: Thu, 13 Dec 2007 10:57:02 +0000 Subject: 965: remove dead code --- src/mesa/pipe/i965simple/brw_wm_fp.c | 1007 ---------------------------------- 1 file changed, 1007 deletions(-) delete mode 100644 src/mesa/pipe/i965simple/brw_wm_fp.c (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_fp.c b/src/mesa/pipe/i965simple/brw_wm_fp.c deleted file mode 100644 index 20e90bc612..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_fp.c +++ /dev/null @@ -1,1007 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_wm.h" -#include "brw_util.h" - - -#define FIRST_INTERNAL_TEMP MAX_NV_FRAGMENT_PROGRAM_TEMPS - -#define X 0 -#define Y 1 -#define Z 2 -#define W 3 - - -static const char *wm_opcode_strings[] = { - "PIXELXY", - "DELTAXY", - "PIXELW", - "LINTERP", - "PINTERP", - "CINTERP", - "WPOSXY", - "FB_WRITE" -}; - -#if 0 -static const char *wm_file_strings[] = { - "PAYLOAD" -}; -#endif - - -/*********************************************************************** - * Source regs - */ -#if 0 -static struct prog_src_register src_reg(unsigned file, unsigned idx) -{ - struct prog_src_register reg; - reg.File = file; - reg.Index = idx; - reg.Swizzle = SWIZZLE_NOOP; - reg.RelAddr = 0; - reg.NegateBase = 0; - reg.Abs = 0; - reg.NegateAbs = 0; - return reg; -} - -static struct prog_src_register src_reg_from_dst(struct prog_dst_register dst) -{ - return src_reg(dst.File, dst.Index); -} - -static struct prog_src_register src_undef( void ) -{ - return src_reg(PROGRAM_UNDEFINED, 0); -} - -static boolean src_is_undef(struct prog_src_register src) -{ - return src.File == PROGRAM_UNDEFINED; -} - -static struct prog_src_register src_swizzle( struct prog_src_register reg, int x, int y, int z, int w ) -{ - reg.Swizzle = MAKE_SWIZZLE4(x,y,z,w); - return reg; -} - -static struct prog_src_register src_swizzle1( struct prog_src_register reg, int x ) -{ - return src_swizzle(reg, x, x, x, x); -} - - -/*********************************************************************** - * Dest regs - */ - -static struct prog_dst_register dst_reg(unsigned file, unsigned idx) -{ - struct prog_dst_register reg; - reg.File = file; - reg.Index = idx; - reg.WriteMask = WRITEMASK_XYZW; - reg.CondMask = 0; - reg.CondSwizzle = 0; - reg.pad = 0; - reg.CondSrc = 0; - return reg; -} - -static struct prog_dst_register dst_mask( struct prog_dst_register reg, int mask ) -{ - reg.WriteMask &= mask; - return reg; -} - -static struct prog_dst_register dst_undef( void ) -{ - return dst_reg(PROGRAM_UNDEFINED, 0); -} - - - -static struct prog_dst_register get_temp( struct brw_wm_compile *c ) -{ - int bit = ffs( ~c->fp_temp ); - - if (!bit) { - _mesa_printf("%s: out of temporaries\n", __FILE__); - exit(1); - } - - c->fp_temp |= 1<<(bit-1); - return dst_reg(PROGRAM_TEMPORARY, FIRST_INTERNAL_TEMP+(bit-1)); -} - - -static void release_temp( struct brw_wm_compile *c, struct prog_dst_register temp ) -{ - c->fp_temp &= ~1<<(temp.Index + 1 - FIRST_INTERNAL_TEMP); -} - - -/*********************************************************************** - * Instructions - */ - -static const struct tgsi_token *get_fp_inst(struct brw_wm_compile *c) -{ - return &c->prog_instructions->tokens[c->nr_fp_insns++]; -} - -static struct prog_instruction *emit_insn(struct brw_wm_compile *c, - const struct prog_instruction *inst0) -{ - struct prog_instruction *inst = get_fp_inst(c); - *inst = *inst0; - inst->Data = (void *)inst0; - return inst; -} - -static struct prog_instruction * emit_op(struct brw_wm_compile *c, - unsigned op, - struct prog_dst_register dest, - unsigned saturate, - unsigned tex_src_unit, - unsigned tex_src_target, - struct prog_src_register src0, - struct prog_src_register src1, - struct prog_src_register src2 ) -{ - struct prog_instruction *inst = get_fp_inst(c); - - memset(inst, 0, sizeof(*inst)); - - inst->Opcode = op; - inst->DstReg = dest; - inst->SaturateMode = saturate; - inst->TexSrcUnit = tex_src_unit; - inst->TexSrcTarget = tex_src_target; - inst->SrcReg[0] = src0; - inst->SrcReg[1] = src1; - inst->SrcReg[2] = src2; - return inst; -} - - - - -/*********************************************************************** - * Special instructions for interpolation and other tasks - */ - -static struct prog_src_register get_pixel_xy( struct brw_wm_compile *c ) -{ - if (src_is_undef(c->pixel_xy)) { - struct prog_dst_register pixel_xy = get_temp(c); - struct prog_src_register payload_r0_depth = src_reg(PROGRAM_PAYLOAD, PAYLOAD_DEPTH); - - - /* Emit the out calculations, and hold onto the results. Use - * two instructions as a temporary is required. - */ - /* pixel_xy.xy = PIXELXY payload[0]; - */ - emit_op(c, - WM_PIXELXY, - dst_mask(pixel_xy, WRITEMASK_XY), - 0, 0, 0, - payload_r0_depth, - src_undef(), - src_undef()); - - c->pixel_xy = src_reg_from_dst(pixel_xy); - } - - return c->pixel_xy; -} - -static struct prog_src_register get_delta_xy( struct brw_wm_compile *c ) -{ - if (src_is_undef(c->delta_xy)) { - struct prog_dst_register delta_xy = get_temp(c); - struct prog_src_register pixel_xy = get_pixel_xy(c); - struct prog_src_register payload_r0_depth = src_reg(PROGRAM_PAYLOAD, PAYLOAD_DEPTH); - - /* deltas.xy = DELTAXY pixel_xy, payload[0] - */ - emit_op(c, - WM_DELTAXY, - dst_mask(delta_xy, WRITEMASK_XY), - 0, 0, 0, - pixel_xy, - payload_r0_depth, - src_undef()); - - c->delta_xy = src_reg_from_dst(delta_xy); - } - - return c->delta_xy; -} - -static struct prog_src_register get_pixel_w( struct brw_wm_compile *c ) -{ - if (src_is_undef(c->pixel_w)) { - struct prog_dst_register pixel_w = get_temp(c); - struct prog_src_register deltas = get_delta_xy(c); - struct prog_src_register interp_wpos = src_reg(PROGRAM_PAYLOAD, FRAG_ATTRIB_WPOS); - - - /* deltas.xyw = DELTAS2 deltas.xy, payload.interp_wpos.x - */ - emit_op(c, - WM_PIXELW, - dst_mask(pixel_w, WRITEMASK_W), - 0, 0, 0, - interp_wpos, - deltas, - src_undef()); - - - c->pixel_w = src_reg_from_dst(pixel_w); - } - - return c->pixel_w; -} - -static void emit_interp( struct brw_wm_compile *c, - unsigned idx ) -{ - struct prog_dst_register dst = dst_reg(PROGRAM_INPUT, idx); - struct prog_src_register interp = src_reg(PROGRAM_PAYLOAD, idx); - struct prog_src_register deltas = get_delta_xy(c); - struct prog_src_register arg2; - unsigned opcode; - - /* Need to use PINTERP on attributes which have been - * multiplied by 1/W in the SF program, and LINTERP on those - * which have not: - */ - switch (idx) { - case FRAG_ATTRIB_WPOS: - opcode = WM_LINTERP; - arg2 = src_undef(); - - /* Have to treat wpos.xy specially: - */ - emit_op(c, - WM_WPOSXY, - dst_mask(dst, WRITEMASK_XY), - 0, 0, 0, - get_pixel_xy(c), - src_undef(), - src_undef()); - - dst = dst_mask(dst, WRITEMASK_ZW); - - /* PROGRAM_INPUT.attr.xyzw = INTERP payload.interp[attr].x, deltas.xyw - */ - emit_op(c, - WM_LINTERP, - dst, - 0, 0, 0, - interp, - deltas, - arg2); - break; - case FRAG_ATTRIB_COL0: - case FRAG_ATTRIB_COL1: - if (c->key.flat_shade) { - emit_op(c, - WM_CINTERP, - dst, - 0, 0, 0, - interp, - src_undef(), - src_undef()); - } - else { - emit_op(c, - WM_LINTERP, - dst, - 0, 0, 0, - interp, - deltas, - src_undef()); - } - break; - default: - emit_op(c, - WM_PINTERP, - dst, - 0, 0, 0, - interp, - deltas, - get_pixel_w(c)); - break; - } - - c->fp_interp_emitted |= 1<SrcReg[0].Index; - struct prog_src_register interp = src_reg(PROGRAM_PAYLOAD, idx); - - c->fp_deriv_emitted |= 1<DstReg, - 0, 0, 0, - interp, - get_pixel_w(c), - src_undef()); -} - -static void emit_ddy( struct brw_wm_compile *c, - const struct prog_instruction *inst ) -{ - unsigned idx = inst->SrcReg[0].Index; - struct prog_src_register interp = src_reg(PROGRAM_PAYLOAD, idx); - - c->fp_deriv_emitted |= 1<DstReg, - 0, 0, 0, - interp, - get_pixel_w(c), - src_undef()); -} - -/*********************************************************************** - * Hacks to extend the program parameter and constant lists. - */ - -/* Add the fog parameters to the parameter list of the original - * program, rather than creating a new list. Doesn't really do any - * harm and it's not as if the parameter handling isn't a big hack - * anyway. - */ -static struct prog_src_register search_or_add_param5(struct brw_wm_compile *c, - int s0, - int s1, - int s2, - int s3, - int s4) -{ - struct gl_program_parameter_list *paramList = c->fp->program.Base.Parameters; - gl_state_index tokens[STATE_LENGTH]; - unsigned idx; - tokens[0] = s0; - tokens[1] = s1; - tokens[2] = s2; - tokens[3] = s3; - tokens[4] = s4; - - for (idx = 0; idx < paramList->NumParameters; idx++) { - if (paramList->Parameters[idx].Type == PROGRAM_STATE_VAR && - memcmp(paramList->Parameters[idx].StateIndexes, tokens, sizeof(tokens)) == 0) - return src_reg(PROGRAM_STATE_VAR, idx); - } - - idx = _mesa_add_state_reference( paramList, tokens ); - - /* Recalculate state dependency: - */ - c->fp->param_state = paramList->StateFlags; - - return src_reg(PROGRAM_STATE_VAR, idx); -} - - -static struct prog_src_register search_or_add_const4f( struct brw_wm_compile *c, - float s0, - float s1, - float s2, - float s3) -{ - struct gl_program_parameter_list *paramList = c->fp->program.Base.Parameters; - float values[4]; - unsigned idx; - unsigned swizzle; - - values[0] = s0; - values[1] = s1; - values[2] = s2; - values[3] = s3; - - /* Have to search, otherwise multiple compilations will each grow - * the parameter list. - */ - for (idx = 0; idx < paramList->NumParameters; idx++) { - if (paramList->Parameters[idx].Type == PROGRAM_CONSTANT && - memcmp(paramList->ParameterValues[idx], values, sizeof(values)) == 0) - - /* XXX: this mimics the mesa bug which puts all constants and - * parameters into the "PROGRAM_STATE_VAR" category: - */ - return src_reg(PROGRAM_STATE_VAR, idx); - } - - idx = _mesa_add_unnamed_constant( paramList, values, 4, &swizzle ); - /* XXX what about swizzle? */ - return src_reg(PROGRAM_STATE_VAR, idx); -} - - - -/*********************************************************************** - * Expand various instructions here to simpler forms. - */ -static void precalc_dst( struct brw_wm_compile *c, - const struct prog_instruction *inst ) -{ - struct prog_src_register src0 = inst->SrcReg[0]; - struct prog_src_register src1 = inst->SrcReg[1]; - struct prog_dst_register dst = inst->DstReg; - - if (dst.WriteMask & WRITEMASK_Y) { - /* dst.y = mul src0.y, src1.y - */ - emit_op(c, - OPCODE_MUL, - dst_mask(dst, WRITEMASK_Y), - inst->SaturateMode, 0, 0, - src0, - src1, - src_undef()); - } - - - if (dst.WriteMask & WRITEMASK_XZ) { - unsigned z = GET_SWZ(src0.Swizzle, Z); - - /* dst.xz = swz src0.1zzz - */ - emit_op(c, - OPCODE_SWZ, - dst_mask(dst, WRITEMASK_XZ), - inst->SaturateMode, 0, 0, - src_swizzle(src0, SWIZZLE_ONE, z, z, z), - src_undef(), - src_undef()); - } - if (dst.WriteMask & WRITEMASK_W) { - /* dst.w = mov src1.w - */ - emit_op(c, - OPCODE_MOV, - dst_mask(dst, WRITEMASK_W), - inst->SaturateMode, 0, 0, - src1, - src_undef(), - src_undef()); - } -} - - -static void precalc_lit( struct brw_wm_compile *c, - const struct prog_instruction *inst ) -{ - struct prog_src_register src0 = inst->SrcReg[0]; - struct prog_dst_register dst = inst->DstReg; - - if (dst.WriteMask & WRITEMASK_XW) { - /* dst.xw = swz src0.1111 - */ - emit_op(c, - OPCODE_SWZ, - dst_mask(dst, WRITEMASK_XW), - 0, 0, 0, - src_swizzle1(src0, SWIZZLE_ONE), - src_undef(), - src_undef()); - } - - - if (dst.WriteMask & WRITEMASK_YZ) { - emit_op(c, - OPCODE_LIT, - dst_mask(dst, WRITEMASK_YZ), - inst->SaturateMode, 0, 0, - src0, - src_undef(), - src_undef()); - } -} - -static void precalc_tex( struct brw_wm_compile *c, - const struct prog_instruction *inst ) -{ - struct prog_src_register coord; - struct prog_dst_register tmpcoord; - - if (inst->TexSrcTarget == TEXTURE_CUBE_INDEX) { - struct prog_instruction *out; - struct prog_dst_register tmp0 = get_temp(c); - struct prog_src_register tmp0src = src_reg_from_dst(tmp0); - struct prog_dst_register tmp1 = get_temp(c); - struct prog_src_register tmp1src = src_reg_from_dst(tmp1); - struct prog_src_register src0 = inst->SrcReg[0]; - - tmpcoord = get_temp(c); - coord = src_reg_from_dst(tmpcoord); - - out = emit_op(c, OPCODE_MOV, - tmpcoord, - 0, 0, 0, - src0, - src_undef(), - src_undef()); - out->SrcReg[0].NegateBase = 0; - out->SrcReg[0].Abs = 1; - - emit_op(c, OPCODE_MAX, - tmp0, - 0, 0, 0, - src_swizzle1(coord, X), - src_swizzle1(coord, Y), - src_undef()); - - emit_op(c, OPCODE_MAX, - tmp1, - 0, 0, 0, - tmp0src, - src_swizzle1(coord, Z), - src_undef()); - - emit_op(c, OPCODE_RCP, - tmp0, - 0, 0, 0, - tmp1src, - src_undef(), - src_undef()); - - emit_op(c, OPCODE_MUL, - tmpcoord, - 0, 0, 0, - src0, - tmp0src, - src_undef()); - - release_temp(c, tmp0); - release_temp(c, tmp1); - } else if (inst->TexSrcTarget == TEXTURE_RECT_INDEX) { - struct prog_src_register scale = - search_or_add_param5( c, - STATE_INTERNAL, - STATE_TEXRECT_SCALE, - inst->TexSrcUnit, - 0,0 ); - - tmpcoord = get_temp(c); - - /* coord.xy = MUL inst->SrcReg[0], { 1/width, 1/height } - */ - emit_op(c, - OPCODE_MUL, - tmpcoord, - 0, 0, 0, - inst->SrcReg[0], - scale, - src_undef()); - - coord = src_reg_from_dst(tmpcoord); - } - else { - coord = inst->SrcReg[0]; - } - - /* Need to emit YUV texture conversions by hand. Probably need to - * do this here - the alternative is in brw_wm_emit.c, but the - * conversion requires allocating a temporary variable which we - * don't have the facility to do that late in the compilation. - */ - if (!(c->key.yuvtex_mask & (1<TexSrcUnit))) { - emit_op(c, - OPCODE_TEX, - inst->DstReg, - inst->SaturateMode, - inst->TexSrcUnit, - inst->TexSrcTarget, - coord, - src_undef(), - src_undef()); - } - else { - /* - CONST C0 = { -.5, -.0625, -.5, 1.164 } - CONST C1 = { 1.596, -0.813, 2.018, -.391 } - UYV = TEX ... - UYV.xyz = ADD UYV, C0 - UYV.y = MUL UYV.y, C0.w - RGB.xyz = MAD UYV.xxz, C1, UYV.y - RGB.y = MAD UYV.z, C1.w, RGB.y - */ - struct prog_dst_register dst = inst->DstReg; - struct prog_src_register src0 = inst->SrcReg[0]; - struct prog_dst_register tmp = get_temp(c); - struct prog_src_register tmpsrc = src_reg_from_dst(tmp); - struct prog_src_register C0 = search_or_add_const4f( c, -.5, -.0625, -.5, 1.164 ); - struct prog_src_register C1 = search_or_add_const4f( c, 1.596, -0.813, 2.018, -.391 ); - - /* tmp = TEX ... - */ - emit_op(c, - OPCODE_TEX, - tmp, - inst->SaturateMode, - inst->TexSrcUnit, - inst->TexSrcTarget, - src0, - src_undef(), - src_undef()); - - /* tmp.xyz = ADD TMP, C0 - */ - emit_op(c, - OPCODE_ADD, - dst_mask(tmp, WRITEMASK_XYZ), - 0, 0, 0, - tmpsrc, - C0, - src_undef()); - - /* YUV.y = MUL YUV.y, C0.w - */ - emit_op(c, - OPCODE_MUL, - dst_mask(tmp, WRITEMASK_Y), - 0, 0, 0, - tmpsrc, - src_swizzle1(C0, W), - src_undef()); - - /* RGB.xyz = MAD YUV.xxz, C1, YUV.y - */ - emit_op(c, - OPCODE_MAD, - dst_mask(dst, WRITEMASK_XYZ), - 0, 0, 0, - src_swizzle(tmpsrc, X,X,Z,Z), - C1, - src_swizzle1(tmpsrc, Y)); - - /* RGB.y = MAD YUV.z, C1.w, RGB.y - */ - emit_op(c, - OPCODE_MAD, - dst_mask(dst, WRITEMASK_Y), - 0, 0, 0, - src_swizzle1(tmpsrc, Z), - src_swizzle1(C1, W), - src_swizzle1(src_reg_from_dst(dst), Y)); - - release_temp(c, tmp); - } - - if (inst->TexSrcTarget == GL_TEXTURE_RECTANGLE_NV) - release_temp(c, tmpcoord); -} - - -static boolean projtex( struct brw_wm_compile *c, - const struct prog_instruction *inst ) -{ - struct prog_src_register src = inst->SrcReg[0]; - - /* Only try to detect the simplest cases. Could detect (later) - * cases where we are trying to emit code like RCP {1.0}, MUL x, - * {1.0}, and so on. - * - * More complex cases than this typically only arise from - * user-provided fragment programs anyway: - */ - if (inst->TexSrcTarget == TEXTURE_CUBE_INDEX) - return 0; /* ut2004 gun rendering !?! */ - else if (src.File == PROGRAM_INPUT && - GET_SWZ(src.Swizzle, W) == W && - (c->key.projtex_mask & (1<<(src.Index + FRAG_ATTRIB_WPOS - FRAG_ATTRIB_TEX0))) == 0) - return 0; - else - return 1; -} - - -static void precalc_txp( struct brw_wm_compile *c, - const struct prog_instruction *inst ) -{ - struct prog_src_register src0 = inst->SrcReg[0]; - - if (projtex(c, inst)) { - struct prog_dst_register tmp = get_temp(c); - struct prog_instruction tmp_inst; - - /* tmp0.w = RCP inst.arg[0][3] - */ - emit_op(c, - OPCODE_RCP, - dst_mask(tmp, WRITEMASK_W), - 0, 0, 0, - src_swizzle1(src0, GET_SWZ(src0.Swizzle, W)), - src_undef(), - src_undef()); - - /* tmp0.xyz = MUL inst.arg[0], tmp0.wwww - */ - emit_op(c, - OPCODE_MUL, - dst_mask(tmp, WRITEMASK_XYZ), - 0, 0, 0, - src0, - src_swizzle1(src_reg_from_dst(tmp), W), - src_undef()); - - /* dst = precalc(TEX tmp0) - */ - tmp_inst = *inst; - tmp_inst.SrcReg[0] = src_reg_from_dst(tmp); - precalc_tex(c, &tmp_inst); - - release_temp(c, tmp); - } - else - { - /* dst = precalc(TEX src0) - */ - precalc_tex(c, inst); - } -} - - - - - -/*********************************************************************** - * Add instructions to perform fog blending - */ - -static void fog_blend( struct brw_wm_compile *c, - struct prog_src_register fog_factor ) -{ - struct prog_dst_register outcolor = dst_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLR); - struct prog_src_register fogcolor = search_or_add_param5( c, STATE_FOG_COLOR, 0,0,0,0 ); - - /* color.xyz = LRP fog_factor.xxxx, output_color, fog_color */ - - emit_op(c, - OPCODE_LRP, - dst_mask(outcolor, WRITEMASK_XYZ), - 0, 0, 0, - fog_factor, - src_reg_from_dst(outcolor), - fogcolor); -} - - - -/* This one is simple - just take the interpolated fog coordinate and - * use it as the fog blend factor. - */ -static void fog_interpolated( struct brw_wm_compile *c ) -{ - struct prog_src_register fogc = src_reg(PROGRAM_INPUT, FRAG_ATTRIB_FOGC); - - if (!(c->fp_interp_emitted & (1<fp->program.FogOption) - return; - - if (1) - fog_interpolated( c ); - else { - /* TODO: per-pixel fog */ - assert(0); - } -} - -static void emit_fb_write( struct brw_wm_compile *c ) -{ - struct prog_src_register outcolor = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_COLR); - struct prog_src_register payload_r0_depth = src_reg(PROGRAM_PAYLOAD, PAYLOAD_DEPTH); - struct prog_src_register outdepth = src_reg(PROGRAM_OUTPUT, FRAG_RESULT_DEPR); - - emit_op(c, - WM_FB_WRITE, - dst_mask(dst_undef(),0), - 0, 0, 0, - outcolor, - payload_r0_depth, - outdepth); -} - - - - -/*********************************************************************** - * Emit INTERP instructions ahead of first use of each attrib. - */ - -static void validate_src_regs( struct brw_wm_compile *c, - const struct prog_instruction *inst ) -{ - unsigned nr_args = brw_wm_nr_args( inst->Opcode ); - unsigned i; - - for (i = 0; i < nr_args; i++) { - if (inst->SrcReg[i].File == PROGRAM_INPUT) { - unsigned idx = inst->SrcReg[i].Index; - if (!(c->fp_interp_emitted & (1<Opcode < MAX_OPCODE) - _mesa_print_instruction(insn); - else if (insn->Opcode < MAX_WM_OPCODE) { - unsigned idx = insn->Opcode - MAX_OPCODE; - - _mesa_print_alu_instruction(insn, - wm_opcode_strings[idx], - 3); - } - else - _mesa_printf("UNKNOWN\n"); - - } -} -void brw_wm_pass_fp( struct brw_wm_compile *c ) -{ - struct brw_fragment_program *fp = c->fp; - unsigned insn; - if (BRW_DEBUG & DEBUG_WM) { - _mesa_printf("\n\n\npre-fp:\n"); - _mesa_print_program(&fp->program.Base); - _mesa_printf("\n"); - } - - c->pixel_xy = src_undef(); - c->delta_xy = src_undef(); - c->pixel_w = src_undef(); - c->nr_fp_insns = 0; - - /* Emit preamble instructions: - */ - - - for (insn = 0; insn < fp->program.Base.NumInstructions; insn++) { - const struct prog_instruction *inst = &fp->program.Base.Instructions[insn]; - struct prog_instruction *out; - - /* Check for INPUT values, emit INTERP instructions where - * necessary: - */ - validate_src_regs(c, inst); - - - switch (inst->Opcode) { - case OPCODE_SWZ: - out = emit_insn(c, inst); - out->Opcode = OPCODE_MOV; - break; - - case OPCODE_ABS: - out = emit_insn(c, inst); - out->Opcode = OPCODE_MOV; - out->SrcReg[0].NegateBase = 0; - out->SrcReg[0].Abs = 1; - break; - - case OPCODE_SUB: - out = emit_insn(c, inst); - out->Opcode = OPCODE_ADD; - out->SrcReg[1].NegateBase ^= 0xf; - break; - - case OPCODE_SCS: - out = emit_insn(c, inst); - /* This should probably be done in the parser. - */ - out->DstReg.WriteMask &= WRITEMASK_XY; - break; - - case OPCODE_DST: - precalc_dst(c, inst); - break; - - case OPCODE_LIT: - precalc_lit(c, inst); - break; - - case OPCODE_TXP: - precalc_txp(c, inst); - break; - - case OPCODE_XPD: - out = emit_insn(c, inst); - /* This should probably be done in the parser. - */ - out->DstReg.WriteMask &= WRITEMASK_XYZ; - break; - - case OPCODE_KIL: - out = emit_insn(c, inst); - /* This should probably be done in the parser. - */ - out->DstReg.WriteMask = 0; - break; - case OPCODE_DDX: - emit_ddx(c, inst); - break; - case OPCODE_DDY: - emit_ddy(c, inst); - break; - case OPCODE_END: - emit_fog(c); - emit_fb_write(c); - break; - case OPCODE_PRINT: - break; - - default: - emit_insn(c, inst); - break; - } - } - - if (BRW_DEBUG & DEBUG_WM) { - _mesa_printf("\n\n\npass_fp:\n"); - print_insns( c->prog_instructions, c->nr_fp_insns ); - _mesa_printf("\n"); - } -} -#endif -- cgit v1.2.3 From 0c2e87d565ced28594156e71ecd387227746e2e3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 11:12:44 +0000 Subject: 965: remove dead code --- src/mesa/pipe/i965simple/Makefile | 1 - src/mesa/pipe/i965simple/brw_vs_constval.c | 223 ----------------------------- 2 files changed, 224 deletions(-) delete mode 100644 src/mesa/pipe/i965simple/brw_vs_constval.c (limited to 'src') diff --git a/src/mesa/pipe/i965simple/Makefile b/src/mesa/pipe/i965simple/Makefile index eda5afaea5..b0085df65f 100644 --- a/src/mesa/pipe/i965simple/Makefile +++ b/src/mesa/pipe/i965simple/Makefile @@ -40,7 +40,6 @@ DRIVER_SOURCES = \ brw_urb.c \ brw_util.c \ brw_vs.c \ - brw_vs_constval.c \ brw_vs_emit.c \ brw_vs_state.c \ brw_vtbl.c \ diff --git a/src/mesa/pipe/i965simple/brw_vs_constval.c b/src/mesa/pipe/i965simple/brw_vs_constval.c deleted file mode 100644 index de43e72c1d..0000000000 --- a/src/mesa/pipe/i965simple/brw_vs_constval.c +++ /dev/null @@ -1,223 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_vs.h" - -#if 0 -/* Component is active if it may diverge from [0,0,0,1]. Undef values - * are promoted to [0,0,0,1] for the purposes of this analysis. - */ -struct tracker { - boolean twoside; - ubyte active[PROGRAM_OUTPUT+1][128]; - unsigned size_masks[4]; -}; - - -static void set_active_component( struct tracker *t, - unsigned file, - unsigned index, - ubyte active ) -{ - switch (file) { - case PROGRAM_TEMPORARY: - case PROGRAM_INPUT: - case PROGRAM_OUTPUT: - t->active[file][index] |= active; - - default: - break; - } -} - -static void set_active( struct tracker *t, - struct prog_dst_register dst, - unsigned active ) -{ - set_active_component( t, dst.File, dst.Index, active & dst.WriteMask ); -} - - -static ubyte get_active_component( struct tracker *t, - unsigned file, - unsigned index, - unsigned component, - ubyte swz ) -{ - switch (swz) { - case SWIZZLE_ZERO: - return component < 3 ? 0 : (1<active[file][index] & (1<active[PROGRAM_OUTPUT][idx]; - if (active & (1<<3)) return 4; - if (active & (1<<2)) return 3; - if (active & (1<<1)) return 2; - if (active & (1<<0)) return 1; - return 0; -} - -/* Note the potential copying that occurs in the setup program: - */ -static void calc_sizes( struct tracker *t ) -{ - unsigned i; - - if (t->twoside) { - t->active[PROGRAM_OUTPUT][VERT_RESULT_COL0] |= - t->active[PROGRAM_OUTPUT][VERT_RESULT_BFC0]; - - t->active[PROGRAM_OUTPUT][VERT_RESULT_COL1] |= - t->active[PROGRAM_OUTPUT][VERT_RESULT_BFC1]; - } - - for (i = 0; i < FRAG_ATTRIB_MAX; i++) { - switch (get_output_size(t, i)) { - case 4: t->size_masks[4-1] |= 1<size_masks[3-1] |= 1<size_masks[2-1] |= 1<size_masks[1-1] |= 1<vb.info.sizes[attr/16]; - unsigned sizes_bits = (sizes_dword>>((attr%16)*2)) & 0x3; - return sizes_bits + 1; -/* return brw->vb.inputs[attr].glarray->Size; */ -} - -/* Calculate sizes of vertex program outputs. Size is the largest - * component index which might vary from [0,0,0,1] - */ -static void calc_wm_input_sizes( struct brw_context *brw ) -{ - /* BRW_NEW_VERTEX_PROGRAM */ - struct brw_vertex_program *vp = - (struct brw_vertex_program *)brw->vertex_program; - /* BRW_NEW_INPUT_DIMENSIONS */ - struct tracker t; - unsigned insn; - unsigned i; - - memset(&t, 0, sizeof(t)); - - /* _NEW_LIGHT */ - if (brw->attribs.Light->Model.TwoSide) - t.twoside = 1; - - for (i = 0; i < PIPE_ATTRIB_MAX; i++) - if (vp->program.Base.InputsRead & (1<program.Base.NumInstructions; insn++) { - struct prog_instruction *inst = &vp->program.Base.Instructions[insn]; - - switch (inst->Opcode) { - case OPCODE_ARL: - break; - - case OPCODE_MOV: - set_active(&t, inst->DstReg, get_active(&t, inst->SrcReg[0])); - break; - - default: - set_active(&t, inst->DstReg, 0xf); - break; - } - } - - calc_sizes(&t); - - if (memcmp(brw->wm.input_size_masks, t.size_masks, sizeof(t.size_masks)) != 0) { - memcpy(brw->wm.input_size_masks, t.size_masks, sizeof(t.size_masks)); - brw->state.dirty.brw |= BRW_NEW_WM_INPUT_DIMENSIONS; - } -} - -const struct brw_tracked_state brw_wm_input_sizes = { - .dirty = { - .mesa = _NEW_LIGHT, - .brw = BRW_NEW_VERTEX_PROGRAM | BRW_NEW_INPUT_DIMENSIONS, - .cache = 0 - }, - .update = calc_wm_input_sizes -}; -#endif -- cgit v1.2.3 From 559750298a40ac845262365aa0d3d69e6dca0ced Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 14:52:31 +0000 Subject: 965: remove dead code --- src/mesa/pipe/i965simple/Makefile | 1 - src/mesa/pipe/i965simple/brw_vtbl.c | 149 ------------------------------------ 2 files changed, 150 deletions(-) delete mode 100644 src/mesa/pipe/i965simple/brw_vtbl.c (limited to 'src') diff --git a/src/mesa/pipe/i965simple/Makefile b/src/mesa/pipe/i965simple/Makefile index b0085df65f..21f40f72a0 100644 --- a/src/mesa/pipe/i965simple/Makefile +++ b/src/mesa/pipe/i965simple/Makefile @@ -42,7 +42,6 @@ DRIVER_SOURCES = \ brw_vs.c \ brw_vs_emit.c \ brw_vs_state.c \ - brw_vtbl.c \ brw_wm.c \ brw_wm_iz.c \ brw_wm_glsl.c \ diff --git a/src/mesa/pipe/i965simple/brw_vtbl.c b/src/mesa/pipe/i965simple/brw_vtbl.c deleted file mode 100644 index 6dc3bd838b..0000000000 --- a/src/mesa/pipe/i965simple/brw_vtbl.c +++ /dev/null @@ -1,149 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_state.h" - -#include "brw_draw.h" -#include "brw_state.h" -#include "brw_vs.h" -#include - -#if 0 -/* called from intelDestroyContext() - */ -static void brw_destroy_context( struct intel_context *intel ) -{ - GLcontext *ctx = &intel->ctx; - struct brw_context *brw = brw_context(&intel->ctx); - - brw_destroy_metaops(brw); - brw_destroy_state(brw); - brw_draw_destroy( brw ); - - brw_ProgramCacheDestroy( ctx ); - brw_FrameBufferTexDestroy( brw ); -} - -/* called from intelDrawBuffer() - */ -static void brw_set_draw_region( struct intel_context *intel, - struct intel_region *draw_region, - struct intel_region *depth_region) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - intel_region_release(&brw->state.draw_region); - intel_region_release(&brw->state.depth_region); - intel_region_reference(&brw->state.draw_region, draw_region); - intel_region_reference(&brw->state.depth_region, depth_region); -} - - -/* called from intelFlushBatchLocked - */ -static void brw_lost_hardware( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - /* Note that we effectively lose the context after this. - * - * Setting this flag provokes a state buffer wrap and also flushes - * the hardware caches. - */ - brw->state.dirty.brw |= BRW_NEW_CONTEXT; - - /* Which means there shouldn't be any commands already queued: - */ - assert(intel->batch->ptr == intel->batch->map); - - brw->state.dirty.mesa |= ~0; - brw->state.dirty.brw |= ~0; - brw->state.dirty.cache |= ~0; -} - -static void brw_note_fence( struct intel_context *intel, - unsigned fence ) -{ - brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_FENCE; -} - -static void brw_note_unlock( struct intel_context *intel ) -{ - struct brw_context *brw = brw_context(&intel->ctx); - - brw_pool_check_wrap(brw, &brw->pool[BRW_GS_POOL]); - brw_pool_check_wrap(brw, &brw->pool[BRW_SS_POOL]); - - brw_context(&intel->ctx)->state.dirty.brw |= BRW_NEW_LOCK; -} - - -void brw_do_flush( struct brw_context *brw, - unsigned flags ) -{ - struct brw_mi_flush flush; - memset(&flush, 0, sizeof(flush)); - flush.opcode = CMD_MI_FLUSH; - flush.flags = flags; - BRW_BATCH_STRUCT(brw, &flush); -} - - -static void brw_emit_flush( struct intel_context *intel, - unsigned unused ) -{ - brw_do_flush(brw_context(&intel->ctx), - BRW_FLUSH_STATE_CACHE|BRW_FLUSH_READ_CACHE); -} - - -/* called from intelWaitForIdle() and intelFlush() - * - * For now, just flush everything. Could be smarter later. - */ -static unsigned brw_flush_cmd( void ) -{ - struct brw_mi_flush flush; - flush.opcode = CMD_MI_FLUSH; - flush.pad = 0; - flush.flags = BRW_FLUSH_READ_CACHE | BRW_FLUSH_STATE_CACHE; - return *(unsigned *)&flush; -} - -static void brw_invalidate_state( struct intel_context *intel, unsigned new_state ) -{ - /* nothing */ -} -#endif -- cgit v1.2.3 From 8cd608d5ba538cdebea7e9bb51c1458ce7eb5137 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 15:39:34 +0000 Subject: 965: missing code from last commit --- src/mesa/pipe/i965simple/brw_context.h | 2 +- src/mesa/pipe/i965simple/brw_draw.c | 1 - src/mesa/pipe/i965simple/brw_draw_upload.c | 12 ++++++------ src/mesa/pipe/i965simple/brw_misc_state.c | 4 +--- src/mesa/pipe/i965simple/brw_state.c | 9 +++------ 5 files changed, 11 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h index 53f66cd6a9..20528f00b3 100644 --- a/src/mesa/pipe/i965simple/brw_context.h +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -489,7 +489,7 @@ struct brw_context /* Arrays with buffer objects to copy non-bufferobj arrays into * for upload: */ - struct pipe_vertex_buffer vbo_array[PIPE_ATTRIB_MAX]; + struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX]; struct brw_vertex_element inputs[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/i965simple/brw_draw.c b/src/mesa/pipe/i965simple/brw_draw.c index f443f41c6f..01c8ddb227 100644 --- a/src/mesa/pipe/i965simple/brw_draw.c +++ b/src/mesa/pipe/i965simple/brw_draw.c @@ -151,7 +151,6 @@ static boolean brw_try_draw_elements( struct pipe_context *pipe, unsigned count ) { struct brw_context *brw = brw_context(pipe); - boolean retval = FALSE; /* Set the first primitive ahead of validate_state: */ diff --git a/src/mesa/pipe/i965simple/brw_draw_upload.c b/src/mesa/pipe/i965simple/brw_draw_upload.c index b2168ab78b..79144837e8 100644 --- a/src/mesa/pipe/i965simple/brw_draw_upload.c +++ b/src/mesa/pipe/i965simple/brw_draw_upload.c @@ -217,18 +217,18 @@ boolean brw_upload_vertex_buffers( struct brw_context *brw ) for (i = 0; i < BRW_VEP_MAX; i++) { - if (brw->vb.vbo_array[i] == NULL) { + if (brw->vb.vbo_array[i]->buffer == NULL) { nr_enabled = i; break; } - vbp.vb[i].vb0.bits.pitch = brw->vb.vbo_array[i].pitch; + vbp.vb[i].vb0.bits.pitch = brw->vb.vbo_array[i]->pitch; vbp.vb[i].vb0.bits.pad = 0; vbp.vb[i].vb0.bits.access_type = BRW_VERTEXBUFFER_ACCESS_VERTEXDATA; vbp.vb[i].vb0.bits.vb_index = i; - vbp.vb[i].offset = brw->vb.vbo_array[i].buffer_offset; - vbp.vb[i].buffer = brw->vb.vbo_array[i].buffer; - vbp.vb[i].max_index = brw->vb.vbo_array[i].max_index; + vbp.vb[i].offset = brw->vb.vbo_array[i]->buffer_offset; + vbp.vb[i].buffer = brw->vb.vbo_array[i]->buffer; + vbp.vb[i].max_index = brw->vb.vbo_array[i]->max_index; } @@ -263,7 +263,7 @@ boolean brw_upload_vertex_elements( struct brw_context *brw ) for (i = 0; i < nr_enabled; i++) { struct brw_vertex_element *input = &brw->vb.inputs[i]; - switch (brw->vb.vbo_array[input->vep.ve0.vertex_buffer_index].pitch) { + switch (brw->vb.vbo_array[input->vep.ve0.vertex_buffer_index]->pitch) { case 0: input->vep.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_0; case 1: input->vep.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; case 2: input->vep.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; diff --git a/src/mesa/pipe/i965simple/brw_misc_state.c b/src/mesa/pipe/i965simple/brw_misc_state.c index 0750502334..2ba4d3a928 100644 --- a/src/mesa/pipe/i965simple/brw_misc_state.c +++ b/src/mesa/pipe/i965simple/brw_misc_state.c @@ -202,9 +202,7 @@ static void upload_depthbuffer(struct brw_context *brw) OUT_BATCH(((depth_surface->pitch * depth_surface->cpp) - 1) | (format << 18) | (BRW_TILEWALK_YMAJOR << 26) | -#if 0 - (depth_surface->region->tiled << 27) | -#endif +// (depth_surface->region->tiled << 27) | (BRW_SURFACE_2D << 29)); OUT_RELOC(depth_surface->buffer, PIPE_BUFFER_FLAG_READ | PIPE_BUFFER_FLAG_WRITE, 0); diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c index ff4ae7999b..f1eba146c4 100644 --- a/src/mesa/pipe/i965simple/brw_state.c +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -273,14 +273,11 @@ static void brw_set_viewport_state( struct pipe_context *pipe, static void brw_set_vertex_buffer( struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer ) + unsigned index, + const struct pipe_vertex_buffer *buffer ) { struct brw_context *brw = brw_context(pipe); - brw->vb.vbo_array[index] = *buffer; - if (index > brw->vb.last_vb) - brw->vb.last_vb = index; - assert(brw->vb.last_vb < BRW_VEP_MAX); + brw->vb.vbo_array[index] = buffer; } static void brw_set_vertex_element(struct pipe_context *pipe, -- cgit v1.2.3 From 81c9058ea90615cd0c819da6e51c3a539a540a7e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Dec 2007 09:05:15 -0700 Subject: Add QuadX,Y fields to tgsi_exec_machine to pass quad's position to tgsi interpreter. The tgsi linear/perspective attribute code now uses these values rather than input[0]. Need to update SSE path to take quad x,y as function params. Then, we can remove additional code. --- src/mesa/pipe/softpipe/sp_quad_fs.c | 11 ++++++++++- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 6 ++---- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 1 + 3 files changed, 13 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 251b47341a..da590b256a 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -100,6 +100,7 @@ shade_quad( machine->InterpCoefs = quad->coef; +#if 1 /* XXX only do this if the fp really reads fragment.position */ machine->Inputs[0].xyzw[0].f[0] = fx; machine->Inputs[0].xyzw[0].f[1] = fx + 1.0f; machine->Inputs[0].xyzw[0].f[2] = fx; @@ -113,6 +114,10 @@ shade_quad( machine->Inputs[0].xyzw[1].f[1] = fy; machine->Inputs[0].xyzw[1].f[2] = fy + 1.0f; machine->Inputs[0].xyzw[1].f[3] = fy + 1.0f; +#endif + + machine->QuadX = quad->x0; + machine->QuadY = quad->y0; /* run shader */ #if defined(__i386__) || defined(__386__) @@ -123,7 +128,11 @@ shade_quad( machine->Outputs, machine->Consts, machine->Temps, - machine->InterpCoefs ); + machine->InterpCoefs +#if 0 + ,quad->x0, quad->y0 +#endif + ); quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]); } else diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 8636271a34..03997f9099 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1352,8 +1352,7 @@ linear_interpolation( unsigned attrib, unsigned chan ) { - const float x = mach->Inputs[0].xyzw[0].f[0]; - const float y = mach->Inputs[0].xyzw[1].f[0]; + const float x = mach->QuadX, y = mach->QuadY; const float dadx = mach->InterpCoefs[attrib].dadx[chan]; const float dady = mach->InterpCoefs[attrib].dady[chan]; const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; @@ -1369,8 +1368,7 @@ perspective_interpolation( unsigned attrib, unsigned chan ) { - const float x = mach->Inputs[0].xyzw[0].f[0]; - const float y = mach->Inputs[0].xyzw[1].f[0]; + const float x = mach->QuadX, y = mach->QuadY; const float dadx = mach->InterpCoefs[attrib].dadx[chan]; const float dady = mach->InterpCoefs[attrib].dady[chan]; const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index e7952a08e3..8d166bb5f4 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -170,6 +170,7 @@ struct tgsi_exec_machine struct tgsi_exec_vector *Inputs; struct tgsi_exec_vector *Outputs; const struct tgsi_token *Tokens; + float QuadX, QuadY; /**< for frag progs only */ unsigned Processor; /* GEOMETRY processor only. */ -- cgit v1.2.3 From a6ae4ea854c62d7d96060ba7044b5e2b73f761f5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 16:35:27 +0000 Subject: 965: magic callback to get buffer offset --- src/mesa/pipe/i965simple/brw_winsys.h | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_winsys.h b/src/mesa/pipe/i965simple/brw_winsys.h index cc0a210a9f..49a12a1c27 100644 --- a/src/mesa/pipe/i965simple/brw_winsys.h +++ b/src/mesa/pipe/i965simple/brw_winsys.h @@ -95,6 +95,8 @@ enum brw_cache_id { BRW_MAX_CACHE }; +#define BRW_CONSTANT_BUFFER BRW_MAX_CACHE + /** * Additional winsys interface for i965simple. * @@ -164,6 +166,13 @@ struct brw_winsys { unsigned data_type); + /* A cheat so we don't have to think about relocations in a couple + * of places yet: + */ + unsigned (*get_buffer_offset)( struct brw_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned flags ); + }; #define BRW_BUFFER_ACCESS_WRITE 0x1 -- cgit v1.2.3 From c4d71c719c0648909ae98bd7cbfadc590b064b13 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 16:39:49 +0000 Subject: 965: implement magic buffer offset callback --- src/mesa/pipe/xlib/xm_winsys_aub.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index ee3c2d6181..0348c2ad40 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -493,6 +493,13 @@ static void aub_i965_batch_reloc( struct brw_winsys *sws, iws->data[iws->nr++] = aub_bo(buf)->offset + delta; } +static unsigned aub_i965_get_buffer_offset( struct brw_winsys *sws, + struct pipe_buffer_handle *buf, + unsigned access_flags ) +{ + return aub_bo(buf)->offset; +} + static void aub_i965_batch_flush( struct brw_winsys *sws, @@ -605,6 +612,7 @@ xmesa_create_i965simple( struct pipe_winsys *winsys ) iws->winsys.batch_reloc = aub_i965_batch_reloc; iws->winsys.batch_flush = aub_i965_batch_flush; iws->winsys.buffer_subdata_typed = aub_i965_buffer_subdata_typed; + iws->winsys.get_buffer_offset = aub_i965_get_buffer_offset; iws->pipe_winsys = winsys; -- cgit v1.2.3 From c605a55e9f771a2f0e85d69ff60059f7ea95320f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 13 Dec 2007 11:42:30 -0500 Subject: i965: make the wm_surface compile --- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 90 ++++++++++++------------- 1 file changed, 45 insertions(+), 45 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index 844cfc54a9..2901e28adb 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -33,8 +33,6 @@ #include "brw_state.h" #include "brw_defines.h" - -#if 0 static unsigned translate_tex_target( int target ) { switch (target) { @@ -75,13 +73,13 @@ static unsigned translate_tex_format( unsigned mesa_format ) assert(0); /* not supported for sampling */ return BRW_SURFACEFORMAT_R8G8B8_UNORM; - case PIPE_FORMAT_U_A8_R8_G8_B8: + case PIPE_FORMAT_B8G8R8A8_UNORM: return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; - case PIPE_FORMAT_RGBA8888_REV: + case PIPE_FORMAT_R8G8B8A8_UNORM: return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; - case PIPE_FORMAT_U_R5_G6_B5: + case PIPE_FORMAT_R5G6B5_UNORM: return BRW_SURFACEFORMAT_B5G6R5_UNORM; case PIPE_FORMAT_A1R5G5B5_UNORM: @@ -95,14 +93,15 @@ static unsigned translate_tex_format( unsigned mesa_format ) case PIPE_FORMAT_YCBCR: return BRW_SURFACEFORMAT_YCRCB_SWAPUVY; - +#if 0 case PIPE_FORMAT_RGB_FXT1: case PIPE_FORMAT_RGBA_FXT1: return BRW_SURFACEFORMAT_FXT1; +#endif case PIPE_FORMAT_Z16_UNORM: return BRW_SURFACEFORMAT_I16_UNORM; - +#if 0 case PIPE_FORMAT_RGB_DXT1: return BRW_SURFACEFORMAT_DXT1_RGB; @@ -119,6 +118,7 @@ static unsigned translate_tex_format( unsigned mesa_format ) return BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB; case PIPE_FORMAT_SRGB_DXT1: return BRW_SURFACEFORMAT_BC1_UNORM_SRGB; +#endif default: assert(0); @@ -126,41 +126,46 @@ static unsigned translate_tex_format( unsigned mesa_format ) } } +static unsigned brw_buffer_offset(struct brw_context *brw, + struct pipe_buffer_handle *buffer) +{ + return brw->winsys->get_buffer_offset(brw->winsys, + buffer, + 0); +} + static -void brw_update_texture_surface( GLcontext *ctx, +void brw_update_texture_surface( struct brw_context *brw, unsigned unit ) { - struct brw_context *brw = brw_context(ctx); - struct gl_texture_object *tObj = brw->attribs.Texture->Unit[unit]._Current; - struct intel_texture_object *intelObj = intel_texture_object(tObj); - struct gl_texture_image *firstImage = tObj->Image[0][intelObj->firstLevel]; + const struct brw_texture *tObj = brw->attribs.Texture[unit]; struct brw_surface_state surf; memset(&surf, 0, sizeof(surf)); surf.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW; - surf.ss0.surface_type = translate_tex_target(tObj->Target); - surf.ss0.surface_format = translate_tex_format(firstImage->TexFormat->MesaFormat); + surf.ss0.surface_type = translate_tex_target(tObj->base.target); + surf.ss0.surface_format = translate_tex_format(tObj->base.format); /* This is ok for all textures with channel width 8bit or less: */ /* surf.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */ /* Updated in emit_reloc */ - surf.ss1.base_addr = brw_buffer_offset( intelObj->mt->region->buffer ); + surf.ss1.base_addr = brw_buffer_offset( brw, tObj->buffer ); - surf.ss2.mip_count = intelObj->lastLevel - intelObj->firstLevel; - surf.ss2.width = firstImage->Width - 1; - surf.ss2.height = firstImage->Height - 1; + surf.ss2.mip_count = tObj->base.last_level - tObj->base.first_level; + surf.ss2.width = tObj->base.width[0]; + surf.ss2.height = tObj->base.height[0]; surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; - surf.ss3.tiled_surface = intelObj->mt->region->tiled; /* always zero */ - surf.ss3.pitch = (intelObj->mt->pitch * intelObj->mt->cpp) - 1; - surf.ss3.depth = firstImage->Depth - 1; + surf.ss3.tiled_surface = 0; /* always zero */ + surf.ss3.pitch = tObj->pitch; + surf.ss3.depth = tObj->base.depth[0]; surf.ss4.min_lod = 0; - if (tObj->Target == GL_TEXTURE_CUBE_MAP) { + if (tObj->base.target == PIPE_TEXTURE_CUBE) { surf.ss0.cube_pos_x = 1; surf.ss0.cube_pos_y = 1; surf.ss0.cube_pos_z = 1; @@ -180,13 +185,11 @@ void brw_update_texture_surface( GLcontext *ctx, static void upload_wm_surfaces(struct brw_context *brw ) { - GLcontext *ctx = &brw->intel.ctx; - struct intel_context *intel = &brw->intel; unsigned i; { struct brw_surface_state surf; - struct intel_region *region = brw->state.draw_region; + struct pipe_surface *region = brw->attribs.FrameBuffer.cbufs[0];/*fixme*/ memset(&surf, 0, sizeof(surf)); @@ -198,27 +201,27 @@ static void upload_wm_surfaces(struct brw_context *brw ) surf.ss0.surface_type = BRW_SURFACE_2D; - surf.ss1.base_addr = brw_buffer_offset( region->buffer ); + surf.ss1.base_addr = brw_buffer_offset( brw, region->buffer ); - surf.ss2.width = region->pitch - 1; /* XXX: not really! */ - surf.ss2.height = region->height - 1; + surf.ss2.width = region->width; + surf.ss2.height = region->height; surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; - surf.ss3.tiled_surface = region->tiled; - surf.ss3.pitch = (region->pitch * region->cpp) - 1; + surf.ss3.tiled_surface = 0; + surf.ss3.pitch = region->pitch; } else { surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; surf.ss0.surface_type = BRW_SURFACE_NULL; } /* _NEW_COLOR */ - surf.ss0.color_blend = (!brw->attribs.Color->_LogicOpEnabled && - brw->attribs.Color->BlendEnabled); + surf.ss0.color_blend = (!brw->attribs.Blend->logicop_enable && + brw->attribs.Blend->blend_enable); - surf.ss0.writedisable_red = !brw->attribs.Color->ColorMask[0]; - surf.ss0.writedisable_green = !brw->attribs.Color->ColorMask[1]; - surf.ss0.writedisable_blue = !brw->attribs.Color->ColorMask[2]; - surf.ss0.writedisable_alpha = !brw->attribs.Color->ColorMask[3]; + surf.ss0.writedisable_red = !brw->attribs.BlendColor.color[0]; + surf.ss0.writedisable_green = !brw->attribs.BlendColor.color[1]; + surf.ss0.writedisable_blue = !brw->attribs.BlendColor.color[2]; + surf.ss0.writedisable_alpha = !brw->attribs.BlendColor.color[3]; @@ -230,23 +233,24 @@ static void upload_wm_surfaces(struct brw_context *brw ) for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { - struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i]; + const struct brw_texture *texUnit = brw->attribs.Texture[i]; /* _NEW_TEXTURE, BRW_NEW_TEXDATA */ - if (texUnit->_ReallyEnabled && - intel_finalize_mipmap_tree(intel, i)) { + if (texUnit->base.refcount/*(texUnit->refcount > 0) == really used */) { - brw_update_texture_surface(ctx, i); + brw_update_texture_surface(brw, i); brw->wm.nr_surfaces = i+2; } - else if( texUnit->_ReallyEnabled && +#if 0 + else if( texUnit->refcount && texUnit->_Current == intel->frame_buffer_texobj ) { brw->wm.bind.surf_ss_offset[i+1] = brw->wm.bind.surf_ss_offset[0]; brw->wm.nr_surfaces = i+2; } +#endif else { brw->wm.bind.surf_ss_offset[i+1] = 0; } @@ -293,14 +297,10 @@ static void emit_reloc_wm_surfaces(struct brw_context *brw) } #endif - - const struct brw_tracked_state brw_wm_surfaces = { .dirty = { - .mesa = _NEW_COLOR | _NEW_TEXTURE | _NEW_BUFFERS, .brw = BRW_NEW_CONTEXT, .cache = 0 }, .update = upload_wm_surfaces, }; -#endif -- cgit v1.2.3 From 568fcf64c774f5f8e9f65bb86c121f5d550b1632 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 20:38:56 +0000 Subject: 965: get fragment shader compiler compiling Don't think that it will run though. --- src/mesa/pipe/i915simple/i915_texture.c | 4 - src/mesa/pipe/i965simple/Makefile | 2 + src/mesa/pipe/i965simple/brw_batch.h | 3 +- src/mesa/pipe/i965simple/brw_cc.c | 2 +- src/mesa/pipe/i965simple/brw_clip_state.c | 4 +- src/mesa/pipe/i965simple/brw_context.c | 1 - src/mesa/pipe/i965simple/brw_context.h | 99 +- src/mesa/pipe/i965simple/brw_curbe.c | 79 +- src/mesa/pipe/i965simple/brw_eu.h | 11 + src/mesa/pipe/i965simple/brw_eu_emit.c | 18 +- src/mesa/pipe/i965simple/brw_gs_state.c | 3 +- src/mesa/pipe/i965simple/brw_misc_state.c | 6 +- src/mesa/pipe/i965simple/brw_sf.c | 213 ++- src/mesa/pipe/i965simple/brw_sf.h | 17 +- src/mesa/pipe/i965simple/brw_sf_emit.c | 340 +--- src/mesa/pipe/i965simple/brw_sf_state.c | 124 +- src/mesa/pipe/i965simple/brw_shader_info.c | 49 + src/mesa/pipe/i965simple/brw_state.c | 7 + src/mesa/pipe/i965simple/brw_state.h | 7 + src/mesa/pipe/i965simple/brw_state_cache.c | 5 +- src/mesa/pipe/i965simple/brw_state_pool.c | 9 +- src/mesa/pipe/i965simple/brw_state_upload.c | 4 +- src/mesa/pipe/i965simple/brw_tex_layout.c | 26 +- src/mesa/pipe/i965simple/brw_vs_state.c | 2 +- src/mesa/pipe/i965simple/brw_wm.c | 28 +- src/mesa/pipe/i965simple/brw_wm.h | 184 +- src/mesa/pipe/i965simple/brw_wm_decl.c | 377 ++++ src/mesa/pipe/i965simple/brw_wm_glsl.c | 2141 ++++++++++------------- src/mesa/pipe/i965simple/brw_wm_sampler_state.c | 9 +- src/mesa/pipe/i965simple/brw_wm_state.c | 50 +- src/mesa/pipe/p_util.h | 5 + src/mesa/pipe/softpipe/sp_texture.c | 4 - 32 files changed, 1795 insertions(+), 2038 deletions(-) create mode 100644 src/mesa/pipe/i965simple/brw_shader_info.c create mode 100644 src/mesa/pipe/i965simple/brw_wm_decl.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c index fefd105adf..44f72e63cc 100644 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -47,10 +47,6 @@ static unsigned minify( unsigned d ) return MAX2(1, d>>1); } -static int align(int value, int alignment) -{ - return (value + alignment - 1) & ~(alignment - 1); -} static void diff --git a/src/mesa/pipe/i965simple/Makefile b/src/mesa/pipe/i965simple/Makefile index 21f40f72a0..48c00ab50b 100644 --- a/src/mesa/pipe/i965simple/Makefile +++ b/src/mesa/pipe/i965simple/Makefile @@ -31,6 +31,7 @@ DRIVER_SOURCES = \ brw_sf.c \ brw_sf_emit.c \ brw_sf_state.c \ + brw_shader_info.c \ brw_state.c \ brw_state_batch.c \ brw_state_cache.c \ @@ -44,6 +45,7 @@ DRIVER_SOURCES = \ brw_vs_state.c \ brw_wm.c \ brw_wm_iz.c \ + brw_wm_decl.c \ brw_wm_glsl.c \ brw_wm_sampler_state.c \ brw_wm_state.c \ diff --git a/src/mesa/pipe/i965simple/brw_batch.h b/src/mesa/pipe/i965simple/brw_batch.h index 7c778f360b..bef69ac871 100644 --- a/src/mesa/pipe/i965simple/brw_batch.h +++ b/src/mesa/pipe/i965simple/brw_batch.h @@ -36,7 +36,7 @@ #define INTEL_BATCH_CLIPRECTS 0x2 #define BEGIN_BATCH( dwords, relocs ) \ - (brw->batch_start = brw->winsys->batch_start(brw->winsys, dwords, relocs)) + brw->winsys->batch_start(brw->winsys, dwords, relocs) #define OUT_BATCH( dword ) \ brw->winsys->batch_dword(brw->winsys, dword) @@ -50,7 +50,6 @@ */ #define FLUSH_BATCH(fence) do { \ brw->winsys->batch_flush(brw->winsys, fence); \ - brw->batch_start = NULL; \ brw->hardware_dirty = ~0; \ } while (0) diff --git a/src/mesa/pipe/i965simple/brw_cc.c b/src/mesa/pipe/i965simple/brw_cc.c index fc7fdba53f..6cc1505311 100644 --- a/src/mesa/pipe/i965simple/brw_cc.c +++ b/src/mesa/pipe/i965simple/brw_cc.c @@ -142,7 +142,7 @@ static void upload_cc_vp( struct brw_context *brw ) const struct brw_tracked_state brw_cc_vp = { .dirty = { - .brw = BRW_NEW_CONTEXT, + .brw = BRW_NEW_SCENE, .cache = 0 }, .update = upload_cc_vp diff --git a/src/mesa/pipe/i965simple/brw_clip_state.c b/src/mesa/pipe/i965simple/brw_clip_state.c index 51a4666a0b..ea5c05a279 100644 --- a/src/mesa/pipe/i965simple/brw_clip_state.c +++ b/src/mesa/pipe/i965simple/brw_clip_state.c @@ -32,7 +32,7 @@ #include "brw_context.h" #include "brw_state.h" #include "brw_defines.h" - +#include "pipe/p_util.h" static void upload_clip_unit( struct brw_context *brw ) @@ -43,7 +43,7 @@ static void upload_clip_unit( struct brw_context *brw ) /* CACHE_NEW_CLIP_PROG */ clip.thread0.grf_reg_count = - ALIGN(brw->clip.prog_data->total_grf, 16) / 16 - 1; + align(brw->clip.prog_data->total_grf, 16) / 16 - 1; clip.thread0.kernel_start_pointer = brw->clip.prog_gs_offset >> 6; clip.thread3.urb_entry_read_length = brw->clip.prog_data->urb_read_length; clip.thread3.const_urb_entry_read_length = brw->clip.prog_data->curb_read_length; diff --git a/src/mesa/pipe/i965simple/brw_context.c b/src/mesa/pipe/i965simple/brw_context.c index e69ba6938e..5e58701e91 100644 --- a/src/mesa/pipe/i965simple/brw_context.c +++ b/src/mesa/pipe/i965simple/brw_context.c @@ -237,7 +237,6 @@ struct pipe_context *brw_create(struct pipe_winsys *pipe_winsys, brw->pci_id = pci_id; brw->dirty = ~0; brw->hardware_dirty = ~0; - brw->batch_start = NULL; memset(&brw->wm.bind, ~0, sizeof(brw->wm.bind)); diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h index 20528f00b3..318c6a7049 100644 --- a/src/mesa/pipe/i965simple/brw_context.h +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -119,7 +119,6 @@ * Handles blending and (presumably) depth and stencil testing. */ -#define BRW_FALLBACK_TEXTURE 0x1 #define BRW_MAX_CURBE (32*16) struct brw_context; @@ -147,16 +146,13 @@ struct brw_winsys; /* Raised for other internal events: */ #define BRW_NEW_URB_FENCE 0x10000 -#define BRW_NEW_INPUT_DIMENSIONS 0x20000 +#define BRW_NEW_PSP 0x20000 #define BRW_NEW_CURBE_OFFSETS 0x40000 #define BRW_NEW_REDUCED_PRIMITIVE 0x80000 #define BRW_NEW_PRIMITIVE 0x100000 -#define BRW_NEW_CONTEXT 0x200000 -#define BRW_NEW_WM_INPUT_DIMENSIONS 0x400000 -#define BRW_NEW_INPUT_VARYING 0x800000 -#define BRW_NEW_PSP 0x1000000 +#define BRW_NEW_SCENE 0x200000 +#define BRW_NEW_SF_LINKAGE 0x400000 -#define ALIGN(value, alignment) ((value + alignment - 1) & ~(alignment - 1)) extern int BRW_DEBUG; #define DEBUG_TEXTURE 0x1 @@ -198,23 +194,47 @@ struct brw_state_flags { unsigned brw; }; + +struct brw_shader_info { + int nr_regs[8]; /* TGSI_FILE_* */ +}; + + + struct brw_vertex_program { struct pipe_shader_state program; - unsigned id; - unsigned param_state; /* flags indicating state tracked by params */ + struct brw_shader_info info; + int id; }; struct brw_fragment_program { struct pipe_shader_state program; - unsigned id; - unsigned param_state; /* flags indicating state tracked by params */ + struct brw_shader_info info; + + boolean UsesDepth; boolean UsesKill; boolean ComputesDepth; + int id; }; + + +struct pipe_setup_linkage { + struct { + unsigned vp_output:5; + unsigned interp_mode:4; + unsigned bf_vp_output:5; + } fp_input[PIPE_MAX_SHADER_INPUTS]; + + unsigned fp_input_count:5; + unsigned max_vp_output:5; +}; + + + struct brw_texture { struct pipe_texture base; @@ -248,6 +268,12 @@ struct brw_texture { * corresponding to a different brw_wm_prog_key struct, with different * compiled programs: */ +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ + struct brw_wm_prog_data { unsigned curb_read_length; unsigned urb_read_length; @@ -256,13 +282,14 @@ struct brw_wm_prog_data { unsigned total_grf; unsigned total_scratch; - unsigned nr_params; - boolean error; - - /* Pointer to tracked values (only valid once - * _mesa_load_state_parameters has been called at runtime). + /* Internally generated constants for the CURBE. These are loaded + * ahead of the data from the constant buffer. */ - const float *param[BRW_MAX_CURBE]; + const float internal_const[8]; + unsigned nr_internal_consts; + unsigned max_const; + + boolean error; }; struct brw_sf_prog_data { @@ -298,19 +325,14 @@ struct brw_vs_prog_data { unsigned inputs_read; + unsigned max_const; + /* Used for calculating urb partitions: */ unsigned urb_entry_size; }; -/* Size == 0 if output either not written, or always [0,0,0,1] - */ -struct brw_vs_ouput_sizes { - ubyte output_size[PIPE_MAX_SHADER_OUTPUTS]; -}; - - #define BRW_MAX_TEX_UNIT 8 #define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 @@ -374,8 +396,6 @@ struct brw_cache { struct brw_tracked_state { struct brw_state_flags dirty; void (*update)( struct brw_context *brw ); - void (*emit_reloc)( struct brw_context *brw ); - boolean always_update; }; @@ -455,8 +475,6 @@ struct brw_context struct { struct brw_state_flags dirty; - struct brw_tracked_state **atoms; - unsigned nr_atoms; } state; @@ -496,27 +514,16 @@ struct brw_context #define BRW_NR_UPLOAD_BUFS 17 #define BRW_UPLOAD_INIT_SIZE (128*1024) - struct { - struct pipe_buffer_handle *vbo[BRW_NR_UPLOAD_BUFS]; - unsigned buf; - unsigned offset; - unsigned size; - unsigned wrap; - } upload; - /* Summary of size and varying of active arrays, so we can check * for changes to this state: */ struct brw_vertex_info info; - int last_vb; } vb; - unsigned *batch_start; unsigned hardware_dirty; unsigned dirty; unsigned pci_id; - /* BRW_NEW_URB_ALLOCATIONS: */ struct { @@ -557,11 +564,6 @@ struct brw_context unsigned vs_size; unsigned total_size; - /* Dynamic tracker which changes to reflect the state referenced - * by active fp and vp program parameters: - */ - struct brw_tracked_state tracked_state; - unsigned gs_offset; float *last_buf; @@ -595,6 +597,8 @@ struct brw_context struct { struct brw_sf_prog_data *prog_data; + struct pipe_setup_linkage linkage; + unsigned prog_gs_offset; unsigned vp_gs_offset; unsigned state_gs_offset; @@ -602,11 +606,8 @@ struct brw_context struct { struct brw_wm_prog_data *prog_data; - struct brw_wm_compile *compile_data; - /* Input sizes, calculated from active vertex program: - */ - unsigned input_size_masks[4]; +// struct brw_wm_compiler *compile_data; /** @@ -667,8 +668,6 @@ void brw_destroy_state(struct brw_context *brw); * brw_tex.c */ void brwUpdateTextureState( struct brw_context *brw ); -void brw_FrameBufferTexInit( struct brw_context *brw ); -void brw_FrameBufferTexDestroy( struct brw_context *brw ); /* brw_urb.c diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index 0894e82d56..b943a7af98 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -35,6 +35,9 @@ #include "brw_defines.h" #include "brw_state.h" #include "brw_util.h" +#include "brw_wm.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" #define FILE_DEBUG_FLAG DEBUG_FALLBACKS @@ -43,11 +46,10 @@ static void calculate_curbe_offsets( struct brw_context *brw ) { /* CACHE_NEW_WM_PROG */ - unsigned nr_fp_regs = (brw->wm.prog_data->nr_params + 15) / 16; + unsigned nr_fp_regs = align(brw->wm.prog_data->max_const, 16); /* BRW_NEW_VERTEX_PROGRAM */ - struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->attribs.VertexProgram; - unsigned nr_vp_regs = (vp->program.num_inputs * 4 + 15) / 16; + unsigned nr_vp_regs = align(brw->vs.prog_data->max_const, 16); unsigned nr_clip_regs = 0; unsigned total_regs; @@ -55,7 +57,7 @@ static void calculate_curbe_offsets( struct brw_context *brw ) /* BRW_NEW_CLIP ? */ if (brw->attribs.Transform->ClipPlanesEnabled) { unsigned nr_planes = 6 + brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled); - nr_clip_regs = (nr_planes * 4 + 15) / 16; + nr_clip_regs = align(nr_planes * 4, 16); } #endif @@ -172,28 +174,18 @@ static float fixed_plane[6][4] = { { 1, 0, 0, 1 } }; -#if 0 /* Upload a new set of constants. Too much variability to go into the * cache mechanism, but maybe would benefit from a comparison against * the current uploaded set of constants. */ static void upload_constant_buffer(struct brw_context *brw) { - GLcontext *ctx = &brw->intel.ctx; - struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->vertex_program; - struct brw_fragment_program *fp = (struct brw_fragment_program *)brw->fragment_program; struct brw_mem_pool *pool = &brw->pool[BRW_GS_POOL]; unsigned sz = brw->curbe.total_size; unsigned bufsz = sz * 16 * sizeof(float); float *buf; unsigned i; - /* Update our own dependency flags. This works because this - * function will also be called whenever fp or vp changes. - */ - brw->curbe.tracked_state.dirty.mesa = (_NEW_TRANSFORM|_NEW_PROJECTION); - brw->curbe.tracked_state.dirty.mesa |= vp->param_state; - brw->curbe.tracked_state.dirty.mesa |= fp->param_state; if (sz == 0) { struct brw_constant_buffer cb; @@ -220,10 +212,16 @@ static void upload_constant_buffer(struct brw_context *brw) if (brw->curbe.wm_size) { unsigned offset = brw->curbe.wm_start * 16; - _mesa_load_state_parameters(ctx, fp->program.Base.Parameters); + /* First the constant buffer constants: + */ + + /* Then any internally generated constants: + */ + for (i = 0; i < brw->wm.prog_data->nr_internal_consts; i++) + buf[offset + i] = brw->wm.prog_data->internal_const[i]; - for (i = 0; i < brw->wm.prog_data->nr_params; i++) - buf[offset + i] = brw->wm.prog_data->param[i][0]; + assert(brw->wm.prog_data->max_const == + brw->wm.prog_data->nr_internal_consts); } @@ -243,34 +241,26 @@ static void upload_constant_buffer(struct brw_context *brw) buf[offset + i * 4 + 3] = fixed_plane[i][3]; } - /* Clip planes: _NEW_TRANSFORM plus _NEW_PROJECTION to get to - * clip-space: + /* Clip planes: BRW_NEW_CLIP: */ - assert(MAX_CLIP_PLANES == 6); - for (j = 0; j < MAX_CLIP_PLANES; j++) { - if (brw->attribs.Transform->ClipPlanesEnabled & (1<attribs.Transform->_ClipUserPlane[j][0]; - buf[offset + i * 4 + 1] = brw->attribs.Transform->_ClipUserPlane[j][1]; - buf[offset + i * 4 + 2] = brw->attribs.Transform->_ClipUserPlane[j][2]; - buf[offset + i * 4 + 3] = brw->attribs.Transform->_ClipUserPlane[j][3]; - i++; - } + for (j = 0; j < brw->attribs.Clip.nr; j++) { + buf[offset + i * 4 + 0] = brw->attribs.Clip.ucp[j][0]; + buf[offset + i * 4 + 1] = brw->attribs.Clip.ucp[j][1]; + buf[offset + i * 4 + 2] = brw->attribs.Clip.ucp[j][2]; + buf[offset + i * 4 + 3] = brw->attribs.Clip.ucp[j][3]; + i++; } } if (brw->curbe.vs_size) { - unsigned offset = brw->curbe.vs_start * 16; - unsigned nr = vp->program.Base.Parameters->NumParameters; +// unsigned offset = brw->curbe.vs_start * 16; +// unsigned nr = vp->max_const; - _mesa_load_state_parameters(ctx, vp->program.Base.Parameters); + /* map the vertex constant buffer and copy to curbe: */ - for (i = 0; i < nr; i++) { - buf[offset + i * 4 + 0] = vp->program.Base.Parameters->ParameterValues[i][0]; - buf[offset + i * 4 + 1] = vp->program.Base.Parameters->ParameterValues[i][1]; - buf[offset + i * 4 + 2] = vp->program.Base.Parameters->ParameterValues[i][2]; - buf[offset + i * 4 + 3] = vp->program.Base.Parameters->ParameterValues[i][3]; - } +// assert(nr == 0); + assert(0); } if (0) { @@ -309,7 +299,12 @@ static void upload_constant_buffer(struct brw_context *brw) /* Copy data to the buffer: */ - dri_bo_subdata(pool->buffer, brw->curbe.gs_offset, bufsz, buf); + brw->winsys->buffer_subdata_typed(brw->winsys, + pool->buffer, + brw->curbe.gs_offset, + bufsz, + buf, + BRW_CONSTANT_BUFFER ); } /* TODO: only emit the constant_buffer packet when necessary, ie: @@ -341,9 +336,7 @@ static void upload_constant_buffer(struct brw_context *brw) * flushes as necessary when doublebuffering of CURBEs isn't * possible. */ -/* intel_batchbuffer_align(brw->intel.batch, 64, sizeof(cb)); */ BRW_BATCH_STRUCT(brw, &cb); -/* intel_batchbuffer_align(brw->intel.batch, 64, 0); */ } } @@ -355,9 +348,8 @@ static void upload_constant_buffer(struct brw_context *brw) */ const struct brw_tracked_state brw_constant_buffer = { .dirty = { - .mesa = (_NEW_TRANSFORM|_NEW_PROJECTION), /* plus fp and vp flags */ - .brw = (BRW_NEW_FRAGMENT_PROGRAM | - BRW_NEW_VERTEX_PROGRAM | + .brw = (BRW_NEW_CLIP | + BRW_NEW_CONSTANTS | BRW_NEW_URB_FENCE | /* Implicit - hardware requires this, not used above */ BRW_NEW_PSP | /* Implicit - hardware requires this, not used above */ BRW_NEW_CURBE_OFFSETS), @@ -366,4 +358,3 @@ const struct brw_tracked_state brw_constant_buffer = { .update = upload_constant_buffer }; -#endif diff --git a/src/mesa/pipe/i965simple/brw_eu.h b/src/mesa/pipe/i965simple/brw_eu.h index 111edb1506..23151ae9ed 100644 --- a/src/mesa/pipe/i965simple/brw_eu.h +++ b/src/mesa/pipe/i965simple/brw_eu.h @@ -694,6 +694,17 @@ void brw_init_compile( struct brw_compile *p ); const unsigned *brw_get_program( struct brw_compile *p, unsigned *sz ); +struct brw_instruction *brw_alu1( struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src ); + +struct brw_instruction *brw_alu2(struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1 ); + /* Helpers for regular instructions: */ #define ALU1(OP) \ diff --git a/src/mesa/pipe/i965simple/brw_eu_emit.c b/src/mesa/pipe/i965simple/brw_eu_emit.c index bda63e8b9a..2423536dd1 100644 --- a/src/mesa/pipe/i965simple/brw_eu_emit.c +++ b/src/mesa/pipe/i965simple/brw_eu_emit.c @@ -363,10 +363,10 @@ static struct brw_instruction *next_insn( struct brw_compile *p, } -static struct brw_instruction *brw_alu1( struct brw_compile *p, - unsigned opcode, - struct brw_reg dest, - struct brw_reg src ) +struct brw_instruction *brw_alu1( struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src ) { struct brw_instruction *insn = next_insn(p, opcode); brw_set_dest(insn, dest); @@ -374,11 +374,11 @@ static struct brw_instruction *brw_alu1( struct brw_compile *p, return insn; } -static struct brw_instruction *brw_alu2(struct brw_compile *p, - unsigned opcode, - struct brw_reg dest, - struct brw_reg src0, - struct brw_reg src1 ) +struct brw_instruction *brw_alu2(struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1 ) { struct brw_instruction *insn = next_insn(p, opcode); brw_set_dest(insn, dest); diff --git a/src/mesa/pipe/i965simple/brw_gs_state.c b/src/mesa/pipe/i965simple/brw_gs_state.c index 8e62eb4bd7..3932e9e939 100644 --- a/src/mesa/pipe/i965simple/brw_gs_state.c +++ b/src/mesa/pipe/i965simple/brw_gs_state.c @@ -34,6 +34,7 @@ #include "brw_context.h" #include "brw_state.h" #include "brw_defines.h" +#include "pipe/p_util.h" @@ -46,7 +47,7 @@ static void upload_gs_unit( struct brw_context *brw ) /* CACHE_NEW_GS_PROG */ if (brw->gs.prog_active) { gs.thread0.grf_reg_count = - ALIGN(brw->gs.prog_data->total_grf, 16) / 16 - 1; + align(brw->gs.prog_data->total_grf, 16) / 16 - 1; gs.thread0.kernel_start_pointer = brw->gs.prog_gs_offset >> 6; gs.thread3.urb_entry_read_length = brw->gs.prog_data->urb_read_length; } diff --git a/src/mesa/pipe/i965simple/brw_misc_state.c b/src/mesa/pipe/i965simple/brw_misc_state.c index 2ba4d3a928..e600e9d8de 100644 --- a/src/mesa/pipe/i965simple/brw_misc_state.c +++ b/src/mesa/pipe/i965simple/brw_misc_state.c @@ -315,7 +315,7 @@ static void upload_pipe_control(struct brw_context *brw) const struct brw_tracked_state brw_pipe_control = { .dirty = { - .brw = BRW_NEW_CONTEXT, + .brw = BRW_NEW_SCENE, .cache = 0 }, .update = upload_pipe_control @@ -380,7 +380,7 @@ static void upload_invarient_state( struct brw_context *brw ) const struct brw_tracked_state brw_invarient_state = { .dirty = { - .brw = BRW_NEW_CONTEXT, + .brw = BRW_NEW_SCENE, .cache = 0 }, .update = upload_invarient_state @@ -416,7 +416,7 @@ static void upload_state_base_address( struct brw_context *brw ) const struct brw_tracked_state brw_state_base_address = { .dirty = { - .brw = BRW_NEW_CONTEXT, + .brw = BRW_NEW_SCENE, .cache = 0 }, .update = upload_state_base_address diff --git a/src/mesa/pipe/i965simple/brw_sf.c b/src/mesa/pipe/i965simple/brw_sf.c index f009ff37d9..e7c02beda5 100644 --- a/src/mesa/pipe/i965simple/brw_sf.c +++ b/src/mesa/pipe/i965simple/brw_sf.c @@ -36,9 +36,8 @@ #include "brw_util.h" #include "brw_sf.h" #include "brw_state.h" +#include "tgsi/util/tgsi_parse.h" -#if 0 -#define DO_SETUP_BITS ((1<<(FRAG_ATTRIB_MAX)) - 1) static void compile_sf_prog( struct brw_context *brw, struct brw_sf_prog_key *key ) @@ -46,7 +45,6 @@ static void compile_sf_prog( struct brw_context *brw, struct brw_sf_compile c; const unsigned *program; unsigned program_size; - unsigned i, idx; memset(&c, 0, sizeof(c)); @@ -55,27 +53,17 @@ static void compile_sf_prog( struct brw_context *brw, brw_init_compile(&c.func); c.key = *key; - c.nr_attrs = brw_count_bits(c.key.attrs); + + + c.nr_attrs = c.key.vp_output_count; c.nr_attr_regs = (c.nr_attrs+1)/2; - c.nr_setup_attrs = brw_count_bits(c.key.attrs & DO_SETUP_BITS); + + c.nr_setup_attrs = c.key.fp_input_count; c.nr_setup_regs = (c.nr_setup_attrs+1)/2; c.prog_data.urb_read_length = c.nr_attr_regs; c.prog_data.urb_entry_size = c.nr_setup_regs * 2; - /* Construct map from attribute number to position in the vertex. - */ - for (i = idx = 0; i < VERT_RESULT_MAX; i++) - if (c.key.attrs & (1<= VERT_RESULT_TEX0 && i <= VERT_RESULT_TEX7) { - c.point_attrs[i].CoordReplace = - brw->attribs.Point->CoordReplace[i - VERT_RESULT_TEX0]; - } else - c.point_attrs[i].CoordReplace = FALSE; - idx++; - } /* Which primitive? Or all three? */ @@ -90,21 +78,17 @@ static void compile_sf_prog( struct brw_context *brw, break; case SF_POINTS: c.nr_verts = 1; - if (key->do_point_sprite) - brw_emit_point_sprite_setup( &c ); - else - brw_emit_point_setup( &c ); + brw_emit_point_setup( &c ); break; + case SF_UNFILLED_TRIS: - c.nr_verts = 3; - brw_emit_anyprim_setup( &c ); - break; default: assert(0); return; } + /* get the program */ program = brw_get_program(&c.func, &program_size); @@ -142,20 +126,15 @@ static void upload_sf_prog( struct brw_context *brw ) /* Populate the key, noting state dependencies: */ /* CACHE_NEW_VS_PROG */ - key.attrs = brw->vs.prog_data->outputs_written; + key.vp_output_count = brw->vs.prog_data->outputs_written; /* BRW_NEW_REDUCED_PRIMITIVE */ switch (brw->reduced_primitive) { case PIPE_PRIM_TRIANGLES: - /* NOTE: We just use the edgeflag attribute as an indicator that - * unfilled triangles are active. We don't actually do the - * edgeflag testing here, it is already done in the clip - * program. - */ - if (key.attrs & (1<attribs.Point->PointSprite; - key.SpriteOrigin = brw->attribs.Point->SpriteOrigin; - /* BRW_NEW_RASTER */ - key.do_flat_shading = (brw->attribs.Raster->flatshade); - key.do_twoside_color = (brw->attribs.Light->Enabled && brw->attribs.Light->Model.TwoSide); - /* _NEW_POLYGON */ - if (key.do_twoside_color) - key.frontface_ccw = (brw->attribs.Polygon->FrontFace == GL_CCW); +// key.do_point_sprite = brw->attribs.Point->PointSprite; +// key.SpriteOrigin = brw->attribs.Point->SpriteOrigin; + +// key.do_flat_shading = (brw->attribs.Raster->flatshade); +// key.do_twoside_color = (brw->attribs.Light->Enabled && brw->attribs.Light->Model.TwoSide); + +// if (key.do_twoside_color) +// key.frontface_ccw = (brw->attribs.Polygon->FrontFace == GL_CCW); if (!search_cache(brw, &key)) @@ -184,11 +162,150 @@ static void upload_sf_prog( struct brw_context *brw ) const struct brw_tracked_state brw_sf_prog = { .dirty = { - .brw = (BRW_NEW_RASTER | - BRW_NEW_REDUCED_PRIMITIVE), - .cache = CACHE_NEW_VS_PROG + .brw = (BRW_NEW_RASTERIZER | + BRW_NEW_REDUCED_PRIMITIVE | + BRW_NEW_VS | + BRW_NEW_FS), + .cache = 0, }, .update = upload_sf_prog }; -#endif + +/* Build a struct like the one we'd like the state tracker to pass to + * us. + */ +static void update_sf_linkage( struct brw_context *brw ) +{ + const struct brw_vertex_program *vs = brw->attribs.VertexProgram; + const struct brw_fragment_program *fs = brw->attribs.FragmentProgram; + struct pipe_setup_linkage state; + struct tgsi_parse_context parse; + + int i, j; + int nr_vp_outputs = 0; + int done = 0; + + struct { + unsigned semantic:8; + unsigned semantic_index:16; + } fp_semantic[32], vp_semantic[32]; + + memset(&state, 0, sizeof(state)); + + state.fp_input_count = 0; + + + + + /* First scan fp inputs + */ + tgsi_parse_init( &parse, fs->program.tokens ); + while( !done && + !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) + { + int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + + for (i = first; i < last; i++) { + state.fp_input[i].vp_output = ~0; + state.fp_input[i].bf_vp_output = ~0; + state.fp_input[i].interp_mode = + parse.FullToken.FullDeclaration.Interpolation.Interpolate; + + fp_semantic[i].semantic = + parse.FullToken.FullDeclaration.Semantic.SemanticName; + fp_semantic[i].semantic_index = + parse.FullToken.FullDeclaration.Semantic.SemanticIndex; + + } + + assert(last > state.fp_input_count); + state.fp_input_count = last; + } + break; + default: + done = 1; + break; + } + } + + + assert(state.fp_input_count == fs->program.num_inputs); + + + /* Then scan vp outputs + */ + done = 0; + tgsi_parse_init( &parse, vs->program.tokens ); + while( !done && + !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) + { + int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + + for (i = first; i < last; i++) { + vp_semantic[i].semantic = + parse.FullToken.FullDeclaration.Semantic.SemanticName; + vp_semantic[i].semantic_index = + parse.FullToken.FullDeclaration.Semantic.SemanticIndex; + } + + assert(last > nr_vp_outputs); + nr_vp_outputs = last; + } + break; + default: + done = 1; + break; + } + } + + + /* Now match based on semantic information. + */ + for (i = 0; i< state.fp_input_count; i++) { + for (j = 0; j < nr_vp_outputs; j++) { + if (fp_semantic[i].semantic == vp_semantic[j].semantic && + fp_semantic[i].semantic_index == vp_semantic[j].semantic_index) { + state.fp_input[i].vp_output = j; + } + } + if (fp_semantic[i].semantic == TGSI_SEMANTIC_COLOR) { + for (j = 0; j < nr_vp_outputs; j++) { + if (TGSI_SEMANTIC_BCOLOR == vp_semantic[j].semantic && + fp_semantic[i].semantic_index == vp_semantic[j].semantic_index) { + state.fp_input[i].bf_vp_output = j; + } + } + } + } + + if (memcmp(&brw->sf.linkage, &state, sizeof(state)) != 0) { + brw->sf.linkage = state; + brw->state.dirty.brw |= BRW_NEW_SF_LINKAGE; + } +} + + +const struct brw_tracked_state brw_sf_linkage = { + .dirty = { + .brw = (BRW_NEW_VS | + BRW_NEW_FS), + .cache = 0, + }, + .update = update_sf_linkage +}; + diff --git a/src/mesa/pipe/i965simple/brw_sf.h b/src/mesa/pipe/i965simple/brw_sf.h index d04388325d..b7ada47560 100644 --- a/src/mesa/pipe/i965simple/brw_sf.h +++ b/src/mesa/pipe/i965simple/brw_sf.h @@ -42,15 +42,26 @@ #define SF_TRIANGLES 2 #define SF_UNFILLED_TRIS 3 + + struct brw_sf_prog_key { - unsigned attrs:32; + unsigned vp_output_count:5; + unsigned fp_input_count:5; + unsigned primitive:2; unsigned do_twoside_color:1; unsigned do_flat_shading:1; unsigned frontface_ccw:1; unsigned do_point_sprite:1; - unsigned pad:10; - int SpriteOrigin; + + /* Interpolation masks; + */ + unsigned linear_mask; + unsigned persp_mask; + unsigned const_mask; + + +// int SpriteOrigin; }; struct brw_sf_point_tex { diff --git a/src/mesa/pipe/i965simple/brw_sf_emit.c b/src/mesa/pipe/i965simple/brw_sf_emit.c index 93f23171f2..834b5efdfe 100644 --- a/src/mesa/pipe/i965simple/brw_sf_emit.c +++ b/src/mesa/pipe/i965simple/brw_sf_emit.c @@ -36,171 +36,6 @@ #include "brw_util.h" #include "brw_sf.h" -#if 0 -static struct brw_reg get_vert_attr(struct brw_sf_compile *c, - struct brw_reg vert, - unsigned attr) -{ - unsigned off = c->attr_to_idx[attr] / 2; - unsigned sub = c->attr_to_idx[attr] % 2; - - return brw_vec4_grf(vert.nr + off, sub * 4); -} - -static boolean have_attr(struct brw_sf_compile *c, - unsigned attr) -{ - return (c->key.attrs & (1<func; - unsigned i; - - for (i = 0; i < 2; i++) { - if (have_attr(c, VERT_RESULT_COL0+i) && - have_attr(c, VERT_RESULT_BFC0+i)) - brw_MOV(p, - get_vert_attr(c, vert, VERT_RESULT_COL0+i), - get_vert_attr(c, vert, VERT_RESULT_BFC0+i)); - } -} - - -static void do_twoside_color( struct brw_sf_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *if_insn; - unsigned backface_conditional = c->key.frontface_ccw ? BRW_CONDITIONAL_G : BRW_CONDITIONAL_L; - - /* Already done in clip program: - */ - if (c->key.primitive == SF_UNFILLED_TRIS) - return; - - /* XXX: What happens if BFC isn't present? This could only happen - * for user-supplied vertex programs, as t_vp_build.c always does - * the right thing. - */ - if (!(have_attr(c, VERT_RESULT_COL0) && have_attr(c, VERT_RESULT_BFC0)) && - !(have_attr(c, VERT_RESULT_COL1) && have_attr(c, VERT_RESULT_BFC1))) - return; - - /* Need to use BRW_EXECUTE_4 and also do an 4-wide compare in order - * to get all channels active inside the IF. In the clipping code - * we run with NoMask, so it's not an option and we can use - * BRW_EXECUTE_1 for all comparisions. - */ - brw_push_insn_state(p); - brw_CMP(p, vec4(brw_null_reg()), backface_conditional, c->det, brw_imm_f(0)); - if_insn = brw_IF(p, BRW_EXECUTE_4); - { - switch (c->nr_verts) { - case 3: copy_bfc(c, c->vert[2]); - case 2: copy_bfc(c, c->vert[1]); - case 1: copy_bfc(c, c->vert[0]); - } - } - brw_ENDIF(p, if_insn); - brw_pop_insn_state(p); -} - - - -/*********************************************************************** - * Flat shading - */ - -#define VERT_RESULT_COLOR_BITS ((1<func; - unsigned i; - - for (i = VERT_RESULT_COL0; i <= VERT_RESULT_COL1; i++) { - if (have_attr(c,i)) - brw_MOV(p, - get_vert_attr(c, dst, i), - get_vert_attr(c, src, i)); - } -} - - - -/* Need to use a computed jump to copy flatshaded attributes as the - * vertices are ordered according to y-coordinate before reaching this - * point, so the PV could be anywhere. - */ -static void do_flatshade_triangle( struct brw_sf_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_reg ip = brw_ip_reg(); - unsigned nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS); - if (!nr) - return; - - /* Already done in clip program: - */ - if (c->key.primitive == SF_UNFILLED_TRIS) - return; - - brw_push_insn_state(p); - - brw_MUL(p, c->pv, c->pv, brw_imm_ud(nr*2+1)); - brw_JMPI(p, ip, ip, c->pv); - - copy_colors(c, c->vert[1], c->vert[0]); - copy_colors(c, c->vert[2], c->vert[0]); - brw_JMPI(p, ip, ip, brw_imm_ud(nr*4+1)); - - copy_colors(c, c->vert[0], c->vert[1]); - copy_colors(c, c->vert[2], c->vert[1]); - brw_JMPI(p, ip, ip, brw_imm_ud(nr*2)); - - copy_colors(c, c->vert[0], c->vert[2]); - copy_colors(c, c->vert[1], c->vert[2]); - - brw_pop_insn_state(p); -} - - -static void do_flatshade_line( struct brw_sf_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_reg ip = brw_ip_reg(); - unsigned nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS); - - if (!nr) - return; - - /* Already done in clip program: - */ - if (c->key.primitive == SF_UNFILLED_TRIS) - return; - - brw_push_insn_state(p); - - brw_MUL(p, c->pv, c->pv, brw_imm_ud(nr+1)); - brw_JMPI(p, ip, ip, c->pv); - copy_colors(c, c->vert[1], c->vert[0]); - - brw_JMPI(p, ip, ip, brw_imm_ud(nr)); - copy_colors(c, c->vert[0], c->vert[1]); - - brw_pop_insn_state(p); -} - /*********************************************************************** @@ -277,9 +112,6 @@ static void copy_z_inv_w( struct brw_sf_compile *c ) static void invert_det( struct brw_sf_compile *c) { - /* Looks like we invert all 8 elements just to get 1/det in - * position 2 !?! - */ brw_math(&c->func, c->inv_det, BRW_MATH_FUNCTION_INV, @@ -302,22 +134,16 @@ static boolean calculate_masks( struct brw_sf_compile *c, ushort *pc_linear) { boolean is_last_attr = (reg == c->nr_setup_regs - 1); - unsigned persp_mask = c->key.attrs & ~NON_PERPECTIVE_ATTRS; - unsigned linear_mask; - if (c->key.do_flat_shading) - linear_mask = c->key.attrs & ~(FRAG_BIT_COL0|FRAG_BIT_COL1); - else - linear_mask = c->key.attrs; *pc_persp = 0; *pc_linear = 0; *pc = 0xf; - if (persp_mask & (1 << c->idx_to_attr[reg*2])) - *pc_persp = 0xf; +// if (persp_mask & (1 << c->idx_to_attr[reg*2])) +// *pc_persp = 0xf; - if (linear_mask & (1 << c->idx_to_attr[reg*2])) +// if (linear_mask & (1 << c->idx_to_attr[reg*2])) *pc_linear = 0xf; /* Maybe only processs one attribute on the final round: @@ -325,10 +151,10 @@ static boolean calculate_masks( struct brw_sf_compile *c, if (reg*2+1 < c->nr_setup_attrs) { *pc |= 0xf0; - if (persp_mask & (1 << c->idx_to_attr[reg*2+1])) - *pc_persp |= 0xf0; +// if (persp_mask & (1 << c->idx_to_attr[reg*2+1])) +// *pc_persp |= 0xf0; - if (linear_mask & (1 << c->idx_to_attr[reg*2+1])) +// if (linear_mask & (1 << c->idx_to_attr[reg*2+1])) *pc_linear |= 0xf0; } @@ -347,12 +173,6 @@ void brw_emit_tri_setup( struct brw_sf_compile *c ) invert_det(c); copy_z_inv_w(c); - if (c->key.do_twoside_color) - do_twoside_color(c); - - if (c->key.do_flat_shading) - do_flatshade_triangle(c); - for (i = 0; i < c->nr_setup_regs; i++) { @@ -433,9 +253,6 @@ void brw_emit_line_setup( struct brw_sf_compile *c ) invert_det(c); copy_z_inv_w(c); - if (c->key.do_flat_shading) - do_flatshade_line(c); - for (i = 0; i < c->nr_setup_regs; i++) { /* Pair of incoming attributes: @@ -491,86 +308,6 @@ void brw_emit_line_setup( struct brw_sf_compile *c ) } } -void brw_emit_point_sprite_setup( struct brw_sf_compile *c ) -{ - struct brw_compile *p = &c->func; - unsigned i; - - c->nr_verts = 1; - alloc_regs(c); - copy_z_inv_w(c); - for (i = 0; i < c->nr_setup_regs; i++) - { - struct brw_sf_point_tex *tex = &c->point_attrs[c->idx_to_attr[2*i]]; - struct brw_reg a0 = offset(c->vert[0], i); - ushort pc, pc_persp, pc_linear; - boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); - - if (pc_persp) - { - if (!tex->CoordReplace) { - brw_set_predicate_control_flag_value(p, pc_persp); - brw_MUL(p, a0, a0, c->inv_w[0]); - } - } - - if (tex->CoordReplace) { - /* Caculate 1.0/PointWidth */ - brw_math(&c->func, - c->tmp, - BRW_MATH_FUNCTION_INV, - BRW_MATH_SATURATE_NONE, - 0, - c->dx0, - BRW_MATH_DATA_SCALAR, - BRW_MATH_PRECISION_FULL); - - if (c->key.SpriteOrigin == GL_UPPER_LEFT) { - brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]); - brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0)); - brw_MUL(p, c->m2Cy, c->tmp, negate(c->inv_w[0])); - brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0)); - } else { - brw_MUL(p, c->m1Cx, c->tmp, c->inv_w[0]); - brw_MOV(p, vec1(suboffset(c->m1Cx, 1)), brw_imm_f(0.0)); - brw_MUL(p, c->m2Cy, c->tmp, c->inv_w[0]); - brw_MOV(p, vec1(suboffset(c->m2Cy, 0)), brw_imm_f(0.0)); - } - } else { - brw_MOV(p, c->m1Cx, brw_imm_ud(0)); - brw_MOV(p, c->m2Cy, brw_imm_ud(0)); - } - - { - brw_set_predicate_control_flag_value(p, pc); - if (tex->CoordReplace) { - if (c->key.SpriteOrigin == GL_UPPER_LEFT) { - brw_MUL(p, c->m3C0, c->inv_w[0], brw_imm_f(1.0)); - brw_MOV(p, vec1(suboffset(c->m3C0, 0)), brw_imm_f(0.0)); - } - else - brw_MOV(p, c->m3C0, brw_imm_f(0.0)); - } else { - brw_MOV(p, c->m3C0, a0); /* constant value */ - } - - /* Copy m0..m3 to URB. - */ - brw_urb_WRITE(p, - brw_null_reg(), - 0, - brw_vec8_grf(0, 0), - 0, /* allocate */ - 1, /* used */ - 4, /* msg len */ - 0, /* response len */ - last, /* eot */ - last, /* writes complete */ - i*4, /* urb destination offset */ - BRW_URB_SWIZZLE_TRANSPOSE); - } - } -} /* Points setup - several simplifications as all attributes are * constant across the face of the point (point sprites excluded!) @@ -629,68 +366,3 @@ void brw_emit_point_setup( struct brw_sf_compile *c ) } } } - -void brw_emit_anyprim_setup( struct brw_sf_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_reg ip = brw_ip_reg(); - struct brw_reg payload_prim = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0); - struct brw_reg payload_attr = get_element_ud(brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, 1, 0), 0); - struct brw_reg primmask; - struct brw_instruction *jmp; - struct brw_reg v1_null_ud = vec1(retype(brw_null_reg(), BRW_REGISTER_TYPE_UD)); - - alloc_regs(c); - - primmask = retype(get_element(c->tmp, 0), BRW_REGISTER_TYPE_UD); - - brw_MOV(p, primmask, brw_imm_ud(1)); - brw_SHL(p, primmask, primmask, payload_prim); - - brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); - brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_TRILIST) | - (1<<_3DPRIM_TRISTRIP) | - (1<<_3DPRIM_TRIFAN) | - (1<<_3DPRIM_TRISTRIP_REVERSE) | - (1<<_3DPRIM_POLYGON) | - (1<<_3DPRIM_RECTLIST) | - (1<<_3DPRIM_TRIFAN_NOSTIPPLE))); - jmp = brw_JMPI(p, ip, ip, brw_imm_w(0)); - { - brw_push_insn_state(p); - brw_emit_tri_setup( c ); - brw_pop_insn_state(p); - /* note - thread killed in subroutine */ - } - brw_land_fwd_jump(p, jmp); - - brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); - brw_AND(p, v1_null_ud, primmask, brw_imm_ud((1<<_3DPRIM_LINELIST) | - (1<<_3DPRIM_LINESTRIP) | - (1<<_3DPRIM_LINELOOP) | - (1<<_3DPRIM_LINESTRIP_CONT) | - (1<<_3DPRIM_LINESTRIP_BF) | - (1<<_3DPRIM_LINESTRIP_CONT_BF))); - jmp = brw_JMPI(p, ip, ip, brw_imm_w(0)); - { - brw_push_insn_state(p); - brw_emit_line_setup( c ); - brw_pop_insn_state(p); - /* note - thread killed in subroutine */ - } - brw_land_fwd_jump(p, jmp); - - brw_set_conditionalmod(p, BRW_CONDITIONAL_Z); - brw_AND(p, v1_null_ud, payload_attr, brw_imm_ud(1<intel.ctx; struct brw_sf_viewport sfv; - struct intel_renderbuffer *irb = - intel_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); - float y_scale, y_bias; - int x, y, w, h, x1, x2, y1, y2; - int draw_h = ctx->DrawBuffer->Height; memset(&sfv, 0, sizeof(sfv)); - if (ctx->DrawBuffer->Name) { - /* User-created FBO */ - if (irb && !irb->RenderToTexture) { - y_scale = -1.0; - y_bias = draw_h; - } else { - y_scale = 1.0; - y_bias = 0; - } - } else { - if (brw->intel.driDrawable != NULL) { - y_scale = -1.0; - y_bias = draw_h; - } else { - y_scale = 1.0; - y_bias = 0; - } - } - /* _NEW_VIEWPORT, BRW_NEW_METAOPS */ - - if (!brw->metaops.active) { - const float *v = brw->intel.ctx.Viewport._WindowMap.m; - - sfv.viewport.m00 = v[MAT_SX]; - sfv.viewport.m11 = v[MAT_SY] * y_scale; - sfv.viewport.m22 = v[MAT_SZ] * brw->intel.depth_scale; - sfv.viewport.m30 = v[MAT_TX]; - sfv.viewport.m31 = v[MAT_TY] * y_scale + y_bias; - sfv.viewport.m32 = v[MAT_TZ] * brw->intel.depth_scale; - } else { - sfv.viewport.m00 = 1; - sfv.viewport.m11 = - 1; - sfv.viewport.m22 = 1; - sfv.viewport.m30 = 0; - sfv.viewport.m31 = brw->intel.driDrawable->h; - sfv.viewport.m32 = 0; - } + /* BRW_NEW_VIEWPORT */ + { + const float *scale = brw->attribs.Viewport.scale; + const float *trans = brw->attribs.Viewport.translate; - /* _NEW_SCISSOR */ - x = brw->attribs.Scissor->X; - y = brw->attribs.Scissor->Y; - w = brw->attribs.Scissor->Width; - h = brw->attribs.Scissor->Height; - - if (ctx->DrawBuffer->Name == 0) { - x1 = x; - y1 = draw_h - (y + h); - x2 = x + w - 1; - y2 = y1 + h - 1; - } else { - /* FBO has non-inverted coords. */ - x1 = x; - y1 = y; - x2 = x + w - 1; - y2 = y + h - 1; + sfv.viewport.m00 = scale[0]; + sfv.viewport.m11 = scale[1]; + sfv.viewport.m22 = scale[2]; + sfv.viewport.m30 = trans[0]; + sfv.viewport.m31 = trans[1]; + sfv.viewport.m32 = trans[2]; } - sfv.scissor.xmin = CLAMP(x1, 0, ctx->DrawBuffer->Width - 1); - sfv.scissor.xmax = CLAMP(y1, 0, ctx->DrawBuffer->Height - 1); - sfv.scissor.ymin = CLAMP(x2, 0, ctx->DrawBuffer->Width - 1); - sfv.scissor.ymax = CLAMP(y2, 0, ctx->DrawBuffer->Height - 1); + /* _NEW_SCISSOR */ + sfv.scissor.xmin = brw->attribs.Scissor.minx; + sfv.scissor.xmax = brw->attribs.Scissor.maxx; + sfv.scissor.ymin = brw->attribs.Scissor.miny; + sfv.scissor.ymax = brw->attribs.Scissor.maxy; brw->sf.vp_gs_offset = brw_cache_data( &brw->cache[BRW_SF_VP], &sfv ); } const struct brw_tracked_state brw_sf_vp = { .dirty = { - .mesa = (_NEW_VIEWPORT | - _NEW_SCISSOR), - .brw = BRW_NEW_METAOPS, + .brw = (BRW_NEW_SCISSOR | + BRW_NEW_VIEWPORT), .cache = 0 }, .update = upload_sf_vp @@ -130,7 +80,7 @@ static void upload_sf_unit( struct brw_context *brw ) memset(&sf, 0, sizeof(sf)); /* CACHE_NEW_SF_PROG */ - sf.thread0.grf_reg_count = ALIGN(brw->sf.prog_data->total_grf, 16) / 16 - 1; + sf.thread0.grf_reg_count = align(brw->sf.prog_data->total_grf, 16) / 16 - 1; sf.thread0.kernel_start_pointer = brw->sf.prog_gs_offset >> 6; sf.thread3.urb_entry_read_length = brw->sf.prog_data->urb_read_length; @@ -151,19 +101,19 @@ static void upload_sf_unit( struct brw_context *brw ) /* CACHE_NEW_SF_VP */ sf.sf5.sf_viewport_state_offset = brw->sf.vp_gs_offset >> 5; - sf.sf5.viewport_transform = 1; - /* _NEW_SCISSOR */ - if (brw->attribs.Scissor->Enabled) + /* BRW_NEW_RASTER */ + if (brw->attribs.Raster->scissor) sf.sf6.scissor = 1; - /* _NEW_POLYGON */ +#if 0 if (brw->attribs.Polygon->FrontFace == GL_CCW) sf.sf5.front_winding = BRW_FRONTWINDING_CCW; else sf.sf5.front_winding = BRW_FRONTWINDING_CW; + if (brw->attribs.Polygon->CullFlag) { switch (brw->attribs.Polygon->CullFaceMode) { case GL_FRONT: @@ -182,25 +132,24 @@ static void upload_sf_unit( struct brw_context *brw ) } else sf.sf6.cull_mode = BRW_CULLMODE_NONE; +#else + sf.sf5.front_winding = BRW_FRONTWINDING_CW; + sf.sf6.cull_mode = BRW_CULLMODE_NONE; +#endif - - /* _NEW_LINE */ - /* XXX use ctx->Const.Min/MaxLineWidth here */ - sf.sf6.line_width = CLAMP(brw->attribs.Line->Width, 1.0, 5.0) * (1<<1); + sf.sf6.line_width = CLAMP(brw->attribs.Raster->line_width, 1.0, 5.0) * (1<<1); sf.sf6.line_endcap_aa_region_width = 1; - if (brw->attribs.Line->SmoothFlag) + if (brw->attribs.Raster->line_smooth) sf.sf6.aa_enable = 1; else if (sf.sf6.line_width <= 0x2) sf.sf6.line_width = 0; - /* _NEW_POINT */ sf.sf6.point_rast_rule = 1; /* opengl conventions */ - /* XXX clamp max depends on AA vs. non-AA */ - sf.sf7.sprite_point = brw->attribs.Point->PointSprite; - sf.sf7.point_size = CLAMP(brw->attribs.Point->Size, 1.0, 255.0) * (1<<3); - sf.sf7.use_point_size_state = !brw->attribs.Point->_Attenuated; + sf.sf7.sprite_point = brw->attribs.Raster->point_sprite; + sf.sf7.point_size = CLAMP(brw->attribs.Raster->line_width, 1.0, 255.0) * (1<<3); + sf.sf7.use_point_size_state = brw->attribs.Raster->point_size_per_vertex; /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons: */ @@ -220,12 +169,8 @@ static void upload_sf_unit( struct brw_context *brw ) const struct brw_tracked_state brw_sf_unit = { .dirty = { - .mesa = (_NEW_POLYGON | - _NEW_LINE | - _NEW_POINT | - _NEW_SCISSOR), - .brw = (BRW_NEW_URB_FENCE | - BRW_NEW_METAOPS), + .brw = (BRW_NEW_RASTERIZER | + BRW_NEW_URB_FENCE), .cache = (CACHE_NEW_SF_VP | CACHE_NEW_SF_PROG) }, @@ -233,4 +178,3 @@ const struct brw_tracked_state brw_sf_unit = { }; -#endif diff --git a/src/mesa/pipe/i965simple/brw_shader_info.c b/src/mesa/pipe/i965simple/brw_shader_info.c new file mode 100644 index 0000000000..431b45466a --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_shader_info.c @@ -0,0 +1,49 @@ + +#include "brw_context.h" +#include "brw_state.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" + + + + +void brw_shader_info(const struct tgsi_token *tokens, + struct brw_shader_info *info ) +{ + struct tgsi_parse_context parse; + int done = 0; + + tgsi_parse_init( &parse, tokens ); + + while( !done && + !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; + unsigned last = decl->u.DeclarationRange.Last; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + // Broken by crazy wpos init: + //assert( info->nr_regs[decl->Declaration.File] <= last); + + info->nr_regs[decl->Declaration.File] = MAX2(info->nr_regs[decl->Declaration.File], + last+1); + break; + } + case TGSI_TOKEN_TYPE_IMMEDIATE: + case TGSI_TOKEN_TYPE_INSTRUCTION: + default: + done = 1; + break; + } + } + + tgsi_parse_free (&parse); + +} diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c index f1eba146c4..032a4e0285 100644 --- a/src/mesa/pipe/i965simple/brw_state.c +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -198,6 +198,10 @@ static void * brw_create_fs_state(struct pipe_context *pipe, /* XXX: Do I have to duplicate the tokens as well?? */ brw_fp->program = *shader; + brw_fp->id = brw_context(pipe)->program_id++; + + brw_shader_info(shader->tokens, + &brw_fp->info); return (void *)brw_fp; } @@ -228,6 +232,9 @@ static void *brw_create_vs_state(struct pipe_context *pipe, /* XXX: Do I have to duplicate the tokens as well?? */ brw_vp->program = *shader; + brw_vp->id = brw_context(pipe)->program_id++; + brw_shader_info(shader->tokens, + &brw_vp->info); tgsi_dump(shader->tokens, 0); diff --git a/src/mesa/pipe/i965simple/brw_state.h b/src/mesa/pipe/i965simple/brw_state.h index 4dabfe8082..d09711f6f0 100644 --- a/src/mesa/pipe/i965simple/brw_state.h +++ b/src/mesa/pipe/i965simple/brw_state.h @@ -154,4 +154,11 @@ void brw_upload_clip_prog(struct brw_context *brw); void brw_upload_blend_constant_color(struct brw_context *brw); void brw_upload_wm_samplers(struct brw_context *brw); +/* brw_shader_info.c + */ + +void brw_shader_info(const struct tgsi_token *tokens, + struct brw_shader_info *info ); + + #endif diff --git a/src/mesa/pipe/i965simple/brw_state_cache.c b/src/mesa/pipe/i965simple/brw_state_cache.c index 13e262d2e5..c5738733f4 100644 --- a/src/mesa/pipe/i965simple/brw_state_cache.c +++ b/src/mesa/pipe/i965simple/brw_state_cache.c @@ -178,8 +178,9 @@ unsigned brw_upload_cache( struct brw_cache *cache, if (BRW_DEBUG & DEBUG_STATE) printf("upload %s: %d bytes to pool buffer %p offset %x\n", - cache->name, data_size, - cache->pool->buffer, + cache->name, + data_size, + (void*)cache->pool->buffer, offset); /* Copy data to the buffer: diff --git a/src/mesa/pipe/i965simple/brw_state_pool.c b/src/mesa/pipe/i965simple/brw_state_pool.c index a490049024..78268ed8f2 100644 --- a/src/mesa/pipe/i965simple/brw_state_pool.c +++ b/src/mesa/pipe/i965simple/brw_state_pool.c @@ -43,17 +43,18 @@ */ #include "pipe/p_winsys.h" +#include "pipe/p_util.h" #include "brw_context.h" #include "brw_state.h" boolean brw_pool_alloc( struct brw_mem_pool *pool, unsigned size, - unsigned align, + unsigned alignment, unsigned *offset_return) { - unsigned fixup = ALIGN(pool->offset, align) - pool->offset; + unsigned fixup = align(pool->offset, alignment) - pool->offset; - size = ALIGN(size, 4); + size = align(size, 4); if (pool->offset + fixup + size >= pool->size) { printf("%s failed\n", __FUNCTION__); @@ -114,7 +115,7 @@ void brw_pool_check_wrap( struct brw_context *brw, struct brw_mem_pool *pool ) { if (pool->offset > (pool->size * 3) / 4) { - brw->state.dirty.brw |= BRW_NEW_CONTEXT; + brw->state.dirty.brw |= BRW_NEW_SCENE; } } diff --git a/src/mesa/pipe/i965simple/brw_state_upload.c b/src/mesa/pipe/i965simple/brw_state_upload.c index 10f1d4812a..1fb480172d 100644 --- a/src/mesa/pipe/i965simple/brw_state_upload.c +++ b/src/mesa/pipe/i965simple/brw_state_upload.c @@ -97,8 +97,6 @@ const struct brw_tracked_state *atoms[] = void brw_init_state( struct brw_context *brw ) { - unsigned i; - brw_init_pools(brw); brw_init_caches(brw); @@ -156,7 +154,7 @@ void brw_validate_state( struct brw_context *brw ) state->brw == 0) return; - if (brw->state.dirty.brw & BRW_NEW_CONTEXT) + if (brw->state.dirty.brw & BRW_NEW_SCENE) brw_clear_batch_cache_flush(brw); if (BRW_DEBUG) { diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.c b/src/mesa/pipe/i965simple/brw_tex_layout.c index b9514be0c2..7d6e2851b1 100644 --- a/src/mesa/pipe/i965simple/brw_tex_layout.c +++ b/src/mesa/pipe/i965simple/brw_tex_layout.c @@ -149,10 +149,10 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) unsigned mip1_width; if (pt->compressed) { - mip1_width = ALIGN(minify(pt->width[0]), align_w) - + ALIGN(minify(minify(pt->width[0])), align_w); + mip1_width = align(minify(pt->width[0]), align_w) + + align(minify(minify(pt->width[0])), align_w); } else { - mip1_width = ALIGN(minify(pt->width[0]), align_w) + mip1_width = align(minify(pt->width[0]), align_w) + minify(minify(pt->width[0])); } @@ -164,7 +164,7 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) /* Pitch must be a whole number of dwords, even though we * express it in texels. */ - tex->pitch = ALIGN(tex->pitch * pt->cpp, 4) / pt->cpp; + tex->pitch = align(tex->pitch * pt->cpp, 4) / pt->cpp; tex->total_height = 0; for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { @@ -176,7 +176,7 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) if (pt->compressed) img_height = MAX2(1, height/4); else - img_height = ALIGN(height, align_h); + img_height = align(height, align_h); /* Because the images are packed better, the final offset @@ -187,7 +187,7 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) /* Layout_below: step right after second mipmap. */ if (level == pt->first_level + 1) { - x += ALIGN(width, align_w); + x += align(width, align_w); } else { y += img_height; @@ -221,13 +221,13 @@ static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture #if 0 if (pt->compressed) { align_w = intel_compressed_alignment(pt->internal_format); - pt->pitch = ALIGN(width, align_w); + pt->pitch = align(width, align_w); pack_y_pitch = (height + 3) / 4; } else #endif { - tex->pitch = ALIGN(pt->width[0] * pt->cpp, 4) / pt->cpp; - pack_y_pitch = ALIGN(pt->height[0], align_h); + tex->pitch = align(pt->width[0] * pt->cpp, 4) / pt->cpp; + pack_y_pitch = align(pt->height[0], align_h); } pack_x_pitch = tex->pitch; @@ -262,8 +262,8 @@ static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture if (pt->compressed) { pack_y_pitch = (height + 3) / 4; - if (pack_x_pitch > ALIGN(width, align_w)) { - pack_x_pitch = ALIGN(width, align_w); + if (pack_x_pitch > align(width, align_w)) { + pack_x_pitch = align(width, align_w); pack_x_nr <<= 1; } } else { @@ -275,7 +275,7 @@ static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture if (pack_y_pitch > 2) { pack_y_pitch >>= 1; - pack_y_pitch = ALIGN(pack_y_pitch, align_h); + pack_y_pitch = align(pack_y_pitch, align_h); } } @@ -305,8 +305,6 @@ brw_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) sizeof(struct brw_texture)); if (tex) { - struct brw_context *brw = brw_context(pipe); - memset(&tex->base + 1, 0, sizeof(struct brw_texture) - sizeof(struct pipe_texture)); diff --git a/src/mesa/pipe/i965simple/brw_vs_state.c b/src/mesa/pipe/i965simple/brw_vs_state.c index 7d6fb383b9..c73469929c 100644 --- a/src/mesa/pipe/i965simple/brw_vs_state.c +++ b/src/mesa/pipe/i965simple/brw_vs_state.c @@ -44,7 +44,7 @@ static void upload_vs_unit( struct brw_context *brw ) /* CACHE_NEW_VS_PROG */ vs.thread0.kernel_start_pointer = brw->vs.prog_gs_offset >> 6; - vs.thread0.grf_reg_count = ALIGN(brw->vs.prog_data->total_grf, 16) / 16 - 1; + vs.thread0.grf_reg_count = align(brw->vs.prog_data->total_grf, 16) / 16 - 1; vs.thread3.urb_entry_read_length = brw->vs.prog_data->urb_read_length; vs.thread3.const_urb_entry_read_length = brw->vs.prog_data->curb_read_length; vs.thread3.dispatch_grf_start_reg = 1; diff --git a/src/mesa/pipe/i965simple/brw_wm.c b/src/mesa/pipe/i965simple/brw_wm.c index f4e3166e1c..f0a38d384b 100644 --- a/src/mesa/pipe/i965simple/brw_wm.c +++ b/src/mesa/pipe/i965simple/brw_wm.c @@ -33,7 +33,9 @@ #include "brw_context.h" #include "brw_util.h" #include "brw_wm.h" +#include "brw_eu.h" #include "brw_state.h" +#include "pipe/p_util.h" @@ -41,24 +43,22 @@ static void do_wm_prog( struct brw_context *brw, struct brw_fragment_program *fp, struct brw_wm_prog_key *key) { - struct brw_wm_compile *c; + struct brw_wm_compile *c = CALLOC_STRUCT(brw_wm_compile); const unsigned *program; unsigned program_size; - c = brw->wm.compile_data; - if (c == NULL) { - brw->wm.compile_data = calloc(1, sizeof(*brw->wm.compile_data)); - c = brw->wm.compile_data; - } else { - memset(c, 0, sizeof(*brw->wm.compile_data)); - } - memcpy(&c->key, key, sizeof(*key)); - + c->key = *key; c->fp = fp; + + c->delta_xy[0] = brw_null_reg(); + c->delta_xy[1] = brw_null_reg(); + c->pixel_xy[0] = brw_null_reg(); + c->pixel_xy[1] = brw_null_reg(); + c->pixel_w = brw_null_reg(); + fprintf(stderr, "XXXXXXXX FP\n"); - brw_wm_glsl_emit(c); /* get the program @@ -74,6 +74,8 @@ static void do_wm_prog( struct brw_context *brw, program_size, &c->prog_data, &brw->wm.prog_data ); + + FREE(c); } @@ -86,8 +88,7 @@ static void brw_wm_populate_key( struct brw_context *brw, (struct brw_fragment_program *)brw->attribs.FragmentProgram; unsigned lookup = 0; unsigned line_aa; - unsigned i; - + memset(key, 0, sizeof(*key)); /* Build the index for table lookup @@ -204,7 +205,6 @@ static void brw_upload_wm_prog( struct brw_context *brw ) const struct brw_tracked_state brw_wm_prog = { .dirty = { .brw = (BRW_NEW_FS | - BRW_NEW_WM_INPUT_DIMENSIONS | BRW_NEW_REDUCED_PRIMITIVE), .cache = 0 }, diff --git a/src/mesa/pipe/i965simple/brw_wm.h b/src/mesa/pipe/i965simple/brw_wm.h index a394e25da3..a1ac0f504a 100644 --- a/src/mesa/pipe/i965simple/brw_wm.h +++ b/src/mesa/pipe/i965simple/brw_wm.h @@ -60,86 +60,19 @@ struct brw_wm_prog_key { unsigned aa_dest_stencil_reg:3; unsigned dest_depth_reg:3; unsigned nr_depth_regs:3; - unsigned projtex_mask:8; unsigned shadowtex_mask:8; unsigned computes_depth:1; /* could be derived from program string */ unsigned source_depth_to_render_target:1; - unsigned flat_shade:1; unsigned runtime_check_aads_emit:1; - - unsigned yuvtex_mask:8; - unsigned pad1:24; - - unsigned program_string_id:32; -}; - - -/* A bit of a glossary: - * - * brw_wm_value: A computed value or program input. Values are - * constant, they are created once and are never modified. When a - * fragment program register is written or overwritten, new values are - * created fresh, preserving the rule that values are constant. - * - * brw_wm_ref: A reference to a value. Wherever a value used is by an - * instruction or as a program output, that is tracked with an - * instance of this struct. All references to a value occur after it - * is created. After the last reference, a value is dead and can be - * discarded. - * - * brw_wm_grf: Represents a physical hardware register. May be either - * empty or hold a value. Register allocation is the process of - * assigning values to grf registers. This occurs in pass2 and the - * brw_wm_grf struct is not used before that. - * - * Fragment program registers: These are time-varying constructs that - * are hard to reason about and which we translate away in pass0. A - * single fragment program register element (eg. temp[0].x) will be - * translated to one or more brw_wm_value structs, one for each time - * that temp[0].x is written to during the program. - */ - - -/* Used in pass2 to track register allocation. - */ -struct brw_wm_grf { - struct brw_wm_value *value; - unsigned nextuse; -}; + unsigned yuvtex_mask:8; -struct brw_wm_value { - struct brw_reg hw_reg; /* emitted to this reg, may not always be there */ - struct brw_wm_ref *lastuse; - struct brw_wm_grf *resident; - unsigned contributes_to_output:1; - unsigned spill_slot:16; /* if non-zero, spill immediately after calculation */ + unsigned program_string_id; }; -struct brw_wm_ref { - struct brw_reg hw_reg; /* nr filled in in pass2, everything else, pass0 */ - struct brw_wm_value *value; - struct brw_wm_ref *prevuse; - unsigned unspill_reg:7; /* unspill to reg */ - unsigned emitted:1; - unsigned insn:24; -}; -struct brw_wm_constref { - const struct brw_wm_ref *ref; - float constval; -}; -struct brw_wm_instruction { - struct brw_wm_value *dst[4]; - struct brw_wm_ref *src[3][4]; - unsigned opcode:8; - unsigned saturate:1; - unsigned writemask:4; - unsigned tex_unit:4; /* texture unit for TEX, TXD, TXP instructions */ - unsigned tex_idx:3; /* TEXTURE_1D,2D,3D,CUBE,RECT_INDEX source target */ -}; #define PROGRAM_INTERNAL_PARAM #define MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS 1024 /* 72 for GL_ARB_f_p */ @@ -151,124 +84,59 @@ struct brw_wm_instruction { #define BRW_WM_MAX_CONST 256 #define BRW_WM_MAX_KILLS MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS - - -/* New opcodes to track internal operations required for WM unit. - * These are added early so that the registers used can be tracked, - * freed and reused like those of other instructions. - */ -#define WM_PIXELXY (TGSI_OPCODE_LAST) -#define WM_DELTAXY (TGSI_OPCODE_LAST + 1) -#define WM_PIXELW (TGSI_OPCODE_LAST + 2) -#define WM_LINTERP (TGSI_OPCODE_LAST + 3) -#define WM_PINTERP (TGSI_OPCODE_LAST + 4) -#define WM_CINTERP (TGSI_OPCODE_LAST + 5) -#define WM_WPOSXY (TGSI_OPCODE_LAST + 6) -#define WM_FB_WRITE (TGSI_OPCODE_LAST + 7) -#define MAX_WM_OPCODE (TGSI_OPCODE_LAST + 8) - #define PAYLOAD_DEPTH (PIPE_ATTRIB_MAX) +#define MAX_IFSN 32 +#define MAX_LOOP_DEPTH 32 + struct brw_wm_compile { struct brw_compile func; struct brw_wm_prog_key key; - struct brw_wm_prog_data prog_data; + struct brw_wm_prog_data prog_data; /* result */ struct brw_fragment_program *fp; - float (*env_param)[4]; - - enum { - START, - PASS2_DONE - } state; - - /* Initial pass - translate fp instructions to fp instructions, - * simplifying and adding instructions for interpolation and - * framebuffer writes. - */ - const struct pipe_shader_state *prog_instructions; - unsigned nr_fp_insns; - unsigned fp_temp; - unsigned fp_interp_emitted; - unsigned fp_deriv_emitted; - - struct tgsi_src_register pixel_xy; - struct tgsi_src_register delta_xy; - struct tgsi_src_register pixel_w; - - - struct brw_wm_value vreg[BRW_WM_MAX_VREG]; - unsigned nr_vreg; - - struct brw_wm_value creg[BRW_WM_MAX_PARAM]; - unsigned nr_creg; + unsigned grf_limit; + unsigned max_wm_grf; - struct { - struct brw_wm_value depth[4]; /* includes r0/r1 */ - struct brw_wm_value input_interp[PIPE_ATTRIB_MAX]; - } payload; + struct brw_reg pixel_xy[2]; + struct brw_reg delta_xy[2]; + struct brw_reg pixel_w; - const struct brw_wm_ref *pass0_fp_reg[16][256][4]; - struct brw_wm_ref undef_ref; - struct brw_wm_value undef_value; + struct brw_reg wm_regs[8][32][4]; - struct brw_wm_ref refs[BRW_WM_MAX_REF]; - unsigned nr_refs; + struct brw_reg payload_depth[4]; + struct brw_reg payload_coef[16]; - struct brw_wm_instruction instruction[BRW_WM_MAX_INSN]; - unsigned nr_insns; + struct brw_reg emit_mask_reg; - struct brw_wm_constref constref[BRW_WM_MAX_CONST]; - unsigned nr_constrefs; + struct brw_instruction *if_inst[MAX_IFSN]; + int if_insn; - struct brw_wm_grf pass2_grf[BRW_WM_MAX_GRF/2]; + struct brw_instruction *loop_inst[MAX_LOOP_DEPTH]; + int loop_insn; - unsigned grf_limit; - unsigned max_wm_grf; - unsigned last_scratch; + struct brw_instruction *inst0; + struct brw_instruction *inst1; - struct { - boolean inited; - struct brw_reg reg; - } wm_regs[16][256][4]; struct brw_reg stack; - struct brw_reg emit_mask_reg; + struct brw_indirect stack_index; + unsigned reg_index; + + unsigned tmp_start; unsigned tmp_index; }; -unsigned brw_wm_nr_args( unsigned opcode ); -unsigned brw_wm_is_scalar_result( unsigned opcode ); - -void brw_wm_pass_fp( struct brw_wm_compile *c ); -void brw_wm_pass0( struct brw_wm_compile *c ); -void brw_wm_pass1( struct brw_wm_compile *c ); -void brw_wm_pass2( struct brw_wm_compile *c ); -void brw_wm_emit( struct brw_wm_compile *c ); - -void brw_wm_print_value( struct brw_wm_compile *c, - struct brw_wm_value *value ); - -void brw_wm_print_ref( struct brw_wm_compile *c, - struct brw_wm_ref *ref ); - -void brw_wm_print_insn( struct brw_wm_compile *c, - struct brw_wm_instruction *inst ); - -void brw_wm_print_program( struct brw_wm_compile *c, - const char *stage ); void brw_wm_lookup_iz( unsigned line_aa, unsigned lookup, struct brw_wm_prog_key *key ); -#if 0 -boolean brw_wm_is_glsl(struct gl_fragment_program *fp); void brw_wm_glsl_emit(struct brw_wm_compile *c); -#endif +void brw_wm_emit_decls(struct brw_wm_compile *c); #endif diff --git a/src/mesa/pipe/i965simple/brw_wm_decl.c b/src/mesa/pipe/i965simple/brw_wm_decl.c new file mode 100644 index 0000000000..392f17fad6 --- /dev/null +++ b/src/mesa/pipe/i965simple/brw_wm_decl.c @@ -0,0 +1,377 @@ + +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_wm.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" + +static struct brw_reg alloc_tmp(struct brw_wm_compile *c) +{ + c->tmp_index++; + c->reg_index = MAX2(c->reg_index, c->tmp_index); + return brw_vec8_grf(c->tmp_start + c->tmp_index, 0); +} + +static void release_tmps(struct brw_wm_compile *c) +{ + c->tmp_index = 0; +} + + + +static int is_null( struct brw_reg reg ) +{ + return (reg.file == BRW_ARCHITECTURE_REGISTER_FILE && + reg.nr == BRW_ARF_NULL); +} + +static void emit_pixel_xy( struct brw_wm_compile *c ) +{ + if (is_null(c->pixel_xy[0])) { + + struct brw_compile *p = &c->func; + struct brw_reg r1_uw = retype(brw_vec1_grf(1, 0), BRW_REGISTER_TYPE_UW); + + c->pixel_xy[0] = vec8(retype(alloc_tmp(c), BRW_REGISTER_TYPE_UW)); + c->pixel_xy[1] = vec8(retype(alloc_tmp(c), BRW_REGISTER_TYPE_UW)); + + /* Calculate pixel centers by adding 1 or 0 to each of the + * micro-tile coordinates passed in r1. + */ + brw_ADD(p, + c->pixel_xy[0], + stride(suboffset(r1_uw, 4), 2, 4, 0), + brw_imm_v(0x10101010)); + + brw_ADD(p, + c->pixel_xy[1], + stride(suboffset(r1_uw, 5), 2, 4, 0), + brw_imm_v(0x11001100)); + } +} + + + + + + +static void emit_delta_xy( struct brw_wm_compile *c ) +{ + if (is_null(c->delta_xy[0])) { + struct brw_compile *p = &c->func; + struct brw_reg r1 = brw_vec1_grf(1, 0); + + emit_pixel_xy(c); + + c->delta_xy[0] = alloc_tmp(c); + c->delta_xy[1] = alloc_tmp(c); + + /* Calc delta X,Y by subtracting origin in r1 from the pixel + * centers. + */ + brw_ADD(p, + c->delta_xy[0], + retype(c->pixel_xy[0], BRW_REGISTER_TYPE_UW), + negate(r1)); + + brw_ADD(p, + c->delta_xy[1], + retype(c->pixel_xy[1], BRW_REGISTER_TYPE_UW), + negate(suboffset(r1,1))); + } +} + + + +#if 0 +static void emit_pixel_w( struct brw_wm_compile *c ) +{ + if (is_null(c->pixel_w)) { + struct brw_compile *p = &c->func; + + struct brw_reg interp_wpos = c->coef_wpos; + + c->pixel_w = alloc_tmp(c); + + emit_delta_xy(c); + + /* Calc 1/w - just linterp wpos[3] optimized by putting the + * result straight into a message reg. + */ + struct brw_reg interp3 = brw_vec1_grf(interp_wpos.nr+1, 4); + brw_LINE(p, brw_null_reg(), interp3, c->delta_xy[0]); + brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), c->delta_xy[1]); + + /* Calc w */ + brw_math_16( p, + c->pixel_w, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 2, + brw_null_reg(), + BRW_MATH_PRECISION_FULL); + } +} +#endif + + +static void emit_cinterp(struct brw_wm_compile *c, + int idx, + int mask ) +{ + struct brw_compile *p = &c->func; + struct brw_reg interp[4]; + struct brw_reg coef = c->payload_coef[idx]; + int i; + + interp[0] = brw_vec1_grf(coef.nr, 0); + interp[1] = brw_vec1_grf(coef.nr, 4); + interp[2] = brw_vec1_grf(coef.nr+1, 0); + interp[3] = brw_vec1_grf(coef.nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<wm_regs[TGSI_FILE_INPUT][idx][i]; + brw_MOV(p, dst, suboffset(interp[i],3)); + } + } +} + +static void emit_linterp(struct brw_wm_compile *c, + int idx, + int mask ) +{ + struct brw_compile *p = &c->func; + struct brw_reg interp[4]; + struct brw_reg coef = c->payload_coef[idx]; + int i; + + emit_delta_xy(c); + + interp[0] = brw_vec1_grf(coef.nr, 0); + interp[1] = brw_vec1_grf(coef.nr, 4); + interp[2] = brw_vec1_grf(coef.nr+1, 0); + interp[3] = brw_vec1_grf(coef.nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<wm_regs[TGSI_FILE_INPUT][idx][i]; + brw_LINE(p, brw_null_reg(), interp[i], c->delta_xy[0]); + brw_MAC(p, dst, suboffset(interp[i],1), c->delta_xy[1]); + } + } +} + +#if 0 +static void emit_pinterp(struct brw_wm_compile *c, + int idx, + int mask ) +{ + struct brw_compile *p = &c->func; + struct brw_reg interp[4]; + struct brw_reg coef = c->payload_coef[idx]; + int i; + + get_delta_xy(c); + get_pixel_w(c); + + interp[0] = brw_vec1_grf(coef.nr, 0); + interp[1] = brw_vec1_grf(coef.nr, 4); + interp[2] = brw_vec1_grf(coef.nr+1, 0); + interp[3] = brw_vec1_grf(coef.nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<delta_xy[0]); + brw_MAC(p, dst, suboffset(interp[i],1), c->delta_xy[1]); + brw_MUL(p, dst, dst, c->pixel_w); + } + } +} +#endif + + + +#if 0 +static void emit_wpos( ) +{ + struct prog_dst_register dst = dst_reg(PROGRAM_INPUT, idx); + struct tgsi_full_src_register interp = src_reg(PROGRAM_PAYLOAD, idx); + struct tgsi_full_src_register deltas = get_delta_xy(c); + struct tgsi_full_src_register arg2; + unsigned opcode; + + opcode = WM_LINTERP; + arg2 = src_undef(); + + /* Have to treat wpos.xy specially: + */ + emit_op(c, + WM_WPOSXY, + dst_mask(dst, WRITEMASK_XY), + 0, 0, 0, + get_pixel_xy(c), + src_undef(), + src_undef()); + + dst = dst_mask(dst, WRITEMASK_ZW); + + /* PROGRAM_INPUT.attr.xyzw = INTERP payload.interp[attr].x, deltas.xyw + */ + emit_op(c, + WM_LINTERP, + dst, + 0, 0, 0, + interp, + deltas, + arg2); +} +#endif + + + + +/* Perform register allocation: + * + * -- r0??? + * -- passthrough depth regs (and stencil/aa??) + * -- curbe ?? + * -- inputs (coefficients) + * + * Use a totally static register allocation. This will perform poorly + * but is an easy way to get started (again). + */ +static void prealloc_reg(struct brw_wm_compile *c) +{ + int i, j; + int nr_curbe_regs = 0; + + /* R0, then some depth related regs: + */ + for (i = 0; i < c->key.nr_depth_regs; i++) { + c->payload_depth[i] = brw_vec8_grf(i*2, 0); + c->reg_index += 2; + } + + + /* Then a copy of our part of the CURBE entry: + */ + { + int nr_constants = c->fp->info.nr_regs[TGSI_FILE_CONSTANT]; + int index = 0; + + c->prog_data.max_const = 4*nr_constants; + for (i = 0; i < nr_constants; i++) { + for (j = 0; j < 4; j++, index++) + c->wm_regs[TGSI_FILE_CONSTANT][i][j] = brw_vec1_grf(c->reg_index + index/8, + index%8); + } + + nr_curbe_regs = 2*((4*nr_constants+15)/16); + c->reg_index += nr_curbe_regs; + } + + /* Next we receive the plane coefficients for parameter + * interpolation: + */ + for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_INPUT]; i++) { + c->payload_coef[i] = brw_vec8_grf(c->reg_index, 0); + c->reg_index += 2; + } + + c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; + c->prog_data.urb_read_length = c->fp->program.num_inputs * 2; + c->prog_data.curb_read_length = nr_curbe_regs; + + /* That's the end of the payload, now we can start allocating registers. + */ + c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); + c->reg_index++; + + c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); + c->reg_index += 2; + + /* Now allocate room for the interpolated inputs and staging + * registers for the outputs: + */ + for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_INPUT]; i++) + for (j = 0; j < 4; j++) + c->wm_regs[TGSI_FILE_INPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); + + for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_OUTPUT]; i++) + for (j = 0; j < 4; j++) + c->wm_regs[TGSI_FILE_OUTPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); + + /* Beyond this we should only need registers for internal temporaries: + */ + c->tmp_start = c->reg_index; +} + + + + + +/* Need to interpolate fragment program inputs in as a preamble to the + * shader. A more sophisticated compiler would do this on demand, but + * we'll do it up front: + */ +void brw_wm_emit_decls(struct brw_wm_compile *c) +{ + struct tgsi_parse_context parse; + int done = 0; + + prealloc_reg(c); + + tgsi_parse_init( &parse, c->fp->program.tokens ); + + while( !done && + !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; + unsigned first = decl->u.DeclarationRange.First; + unsigned last = decl->u.DeclarationRange.Last; + unsigned mask = decl->Declaration.UsageMask; /* ? */ + unsigned i; + + if (decl->Declaration.File != TGSI_FILE_INPUT) + break; + + assert(decl->Declaration.Interpolate); + + for( i = first; i <= last; i++ ) { + switch (decl->Interpolation.Interpolate) { + case TGSI_INTERPOLATE_CONSTANT: + emit_cinterp(c, i, mask); + break; + + case TGSI_INTERPOLATE_LINEAR: + emit_linterp(c, i, mask); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + //emit_pinterp(c, i, mask); + emit_linterp(c, i, mask); + break; + } + } + break; + } + case TGSI_TOKEN_TYPE_IMMEDIATE: + case TGSI_TOKEN_TYPE_INSTRUCTION: + default: + done = 1; + break; + } + } + + tgsi_parse_free (&parse); + + release_tmps(c); +} diff --git a/src/mesa/pipe/i965simple/brw_wm_glsl.c b/src/mesa/pipe/i965simple/brw_wm_glsl.c index 90e73a605a..d6dfaed826 100644 --- a/src/mesa/pipe/i965simple/brw_wm_glsl.c +++ b/src/mesa/pipe/i965simple/brw_wm_glsl.c @@ -2,753 +2,437 @@ #include "brw_context.h" #include "brw_eu.h" #include "brw_wm.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" -#if 0 - -/* Only guess, need a flag in gl_fragment_program later */ -boolean brw_wm_is_glsl(struct gl_fragment_program *fp) -{ - int i; - for (i = 0; i < fp->Base.NumInstructions; i++) { - struct prog_instruction *inst = &fp->Base.Instructions[i]; - switch (inst->Opcode) { - case OPCODE_IF: - case OPCODE_INT: - case OPCODE_ENDIF: - case OPCODE_CAL: - case OPCODE_BRK: - case OPCODE_RET: - case OPCODE_DDX: - case OPCODE_DDY: - case OPCODE_BGNLOOP: - return TRUE; - default: - break; - } - } - return FALSE; -} -static void set_reg(struct brw_wm_compile *c, int file, int index, - int component, struct brw_reg reg) -{ - c->wm_regs[file][index][component].reg = reg; - c->wm_regs[file][index][component].inited = TRUE; -} -static int get_scalar_dst_index(struct prog_instruction *inst) +static int get_scalar_dst_index(struct tgsi_full_instruction *inst) { - int i; - for (i = 0; i < 4; i++) - if (inst->DstReg.WriteMask & (1<FullDstRegisters[0].DstRegister; + int i; + for (i = 0; i < 4; i++) + if (dst.WriteMask & (1<tmp_index--, 0); - return reg; + c->tmp_index++; + c->reg_index = MAX2(c->reg_index, c->tmp_index); + return brw_vec8_grf(c->tmp_start + c->tmp_index, 0); } static void release_tmps(struct brw_wm_compile *c) { - c->tmp_index = 127; + c->tmp_index = 0; } + static struct brw_reg -get_reg(struct brw_wm_compile *c, int file, int index, int component, int nr, unsigned neg, unsigned abs) +get_reg(struct brw_wm_compile *c, int file, int index, int component ) { - struct brw_reg reg; - switch (file) { - case PROGRAM_STATE_VAR: - case PROGRAM_CONSTANT: - case PROGRAM_UNIFORM: - file = PROGRAM_STATE_VAR; - break; - case PROGRAM_UNDEFINED: - return brw_null_reg(); - default: - break; - } - - if(c->wm_regs[file][index][component].inited) - reg = c->wm_regs[file][index][component].reg; - else - reg = brw_vec8_grf(c->reg_index, 0); - - if(!c->wm_regs[file][index][component].inited) { - set_reg(c, file, index, component, reg); - c->reg_index++; - } - - if (neg & (1<< component)) { - reg = negate(reg); - } - if (abs) - reg = brw_abs(reg); - return reg; + switch (file) { + case TGSI_FILE_NULL: + return brw_null_reg(); + + case TGSI_FILE_SAMPLER: + /* Should never get here: + */ + assert (0); + return brw_null_reg(); + + case TGSI_FILE_IMMEDIATE: + /* These need a different path: + */ + assert(0); + return brw_null_reg(); + + + case TGSI_FILE_CONSTANT: + case TGSI_FILE_INPUT: + case TGSI_FILE_OUTPUT: + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_ADDRESS: + return c->wm_regs[file][index][component]; + + default: + assert(0); + return brw_null_reg(); + } } -static void prealloc_reg(struct brw_wm_compile *c) -{ - int i, j; - struct brw_reg reg; - int nr_interp_regs = 0; - unsigned inputs = FRAG_BIT_WPOS | c->fp_interp_emitted | c->fp_deriv_emitted; - - for (i = 0; i < 4; i++) { - reg = (i < c->key.nr_depth_regs) - ? brw_vec8_grf(i*2, 0) : brw_vec8_grf(0, 0); - set_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, i, reg); - } - c->reg_index += 2*c->key.nr_depth_regs; - { - int nr_params = c->fp->program.Base.Parameters->NumParameters; - struct gl_program_parameter_list *plist = - c->fp->program.Base.Parameters; - int index = 0; - c->prog_data.nr_params = 4*nr_params; - for (i = 0; i < nr_params; i++) { - for (j = 0; j < 4; j++, index++) { - reg = brw_vec1_grf(c->reg_index + index/8, - index%8); - c->prog_data.param[index] = - &plist->ParameterValues[i][j]; - set_reg(c, PROGRAM_STATE_VAR, i, j, reg); - } - } - c->nr_creg = 2*((4*nr_params+15)/16); - c->reg_index += c->nr_creg; - } - for (i = 0; i < FRAG_ATTRIB_MAX; i++) { - if (inputs & (1<reg_index, 0); - for (j = 0; j < 4; j++) - set_reg(c, PROGRAM_PAYLOAD, i, j, reg); - c->reg_index += 2; - - } - } - c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; - c->prog_data.urb_read_length = nr_interp_regs * 2; - c->prog_data.curb_read_length = c->nr_creg; - c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); - c->reg_index++; - c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); - c->reg_index += 2; -} static struct brw_reg get_dst_reg(struct brw_wm_compile *c, - struct prog_instruction *inst, int component, int nr) -{ - return get_reg(c, inst->DstReg.File, inst->DstReg.Index, component, nr, - 0, 0); -} - -static struct brw_reg get_src_reg(struct brw_wm_compile *c, - struct prog_src_register *src, int index, int nr) + struct tgsi_full_instruction *inst, + int component) { - int component = GET_SWZ(src->Swizzle, index); - return get_reg(c, src->File, src->Index, component, nr, - src->NegateBase, src->Abs); + return get_reg(c, + inst->FullDstRegisters[0].DstRegister.File, + inst->FullDstRegisters[0].DstRegister.Index, + component); } -static void emit_abs( struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - int i; - struct brw_compile *p = &c->func; - brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); - for (i = 0; i < 4; i++) { - if (inst->DstReg.WriteMask & (1<SrcReg[0], i, 1); - brw_MOV(p, dst, brw_abs(src)); - } - } - brw_set_saturate(p, 0); -} - -static void emit_int( struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - int i; - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); - for (i = 0; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - brw_RNDD(p, dst, src); - } - } - brw_set_saturate(p, 0); -} - -static void emit_mov( struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - int i; - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); - for (i = 0; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - brw_MOV(p, dst, src); - } - } - brw_set_saturate(p, 0); -} - -static void emit_pixel_xy(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_reg r1 = brw_vec1_grf(1, 0); - struct brw_reg r1_uw = retype(r1, BRW_REGISTER_TYPE_UW); - - struct brw_reg dst0, dst1; - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - - dst0 = get_dst_reg(c, inst, 0, 1); - dst1 = get_dst_reg(c, inst, 1, 1); - /* Calculate pixel centers by adding 1 or 0 to each of the - * micro-tile coordinates passed in r1. - */ - if (mask & WRITEMASK_X) { - brw_ADD(p, - vec8(retype(dst0, BRW_REGISTER_TYPE_UW)), - stride(suboffset(r1_uw, 4), 2, 4, 0), - brw_imm_v(0x10101010)); - } - - if (mask & WRITEMASK_Y) { - brw_ADD(p, - vec8(retype(dst1, BRW_REGISTER_TYPE_UW)), - stride(suboffset(r1_uw, 5), 2, 4, 0), - brw_imm_v(0x11001100)); - } - -} - -static void emit_delta_xy(struct brw_wm_compile *c, - struct prog_instruction *inst) +static int get_swz( struct tgsi_src_register src, int index ) { - struct brw_reg r1 = brw_vec1_grf(1, 0); - struct brw_reg dst0, dst1, src0, src1; - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - - dst0 = get_dst_reg(c, inst, 0, 1); - dst1 = get_dst_reg(c, inst, 1, 1); - src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); - src1 = get_src_reg(c, &inst->SrcReg[0], 1, 1); - /* Calc delta X,Y by subtracting origin in r1 from the pixel - * centers. - */ - if (mask & WRITEMASK_X) { - brw_ADD(p, - dst0, - retype(src0, BRW_REGISTER_TYPE_UW), - negate(r1)); - } - - if (mask & WRITEMASK_Y) { - brw_ADD(p, - dst1, - retype(src1, BRW_REGISTER_TYPE_UW), - negate(suboffset(r1,1))); - - } - + switch (index & 3) { + case 0: return src.SwizzleX; + case 1: return src.SwizzleY; + case 2: return src.SwizzleZ; + case 3: return src.SwizzleW; + default: return 0; + } } - -static void fire_fb_write( struct brw_wm_compile *c, - unsigned base_reg, - unsigned nr ) +static int get_ext_swz( struct tgsi_src_register_ext_swz src, int index ) { - struct brw_compile *p = &c->func; - - /* Pass through control information: - */ - /* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ - { - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ - brw_MOV(p, - brw_message_reg(base_reg + 1), - brw_vec8_grf(1, 0)); - brw_pop_insn_state(p); - } - /* Send framebuffer write message: */ - brw_fb_WRITE(p, - retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW), - base_reg, - retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), - 0, /* render surface always 0 */ - nr, - 0, - 1); + switch (index & 3) { + case 0: return src.ExtSwizzleX; + case 1: return src.ExtSwizzleY; + case 2: return src.ExtSwizzleZ; + case 3: return src.ExtSwizzleW; + default: return 0; + } } -static void emit_fb_write(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - int nr = 2; - int channel; - struct brw_reg src0;//, src1, src2, dst; - - /* Reserve a space for AA - may not be needed: - */ - if (c->key.aa_dest_stencil_reg) - nr += 1; - { - brw_push_insn_state(p); - for (channel = 0; channel < 4; channel++) { - src0 = get_src_reg(c, &inst->SrcReg[0], channel, 1); - /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ - /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ - brw_MOV(p, brw_message_reg(nr + channel), src0); - } - /* skip over the regs populated above: */ - nr += 8; - brw_pop_insn_state(p); - } - fire_fb_write(c, 0, nr); -} - -static void emit_pixel_w( struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - if (mask & WRITEMASK_W) { - struct brw_reg dst, src0, delta0, delta1; - struct brw_reg interp3; - - dst = get_dst_reg(c, inst, 3, 1); - src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); - delta0 = get_src_reg(c, &inst->SrcReg[1], 0, 1); - delta1 = get_src_reg(c, &inst->SrcReg[1], 1, 1); - - interp3 = brw_vec1_grf(src0.nr+1, 4); - /* Calc 1/w - just linterp wpos[3] optimized by putting the - * result straight into a message reg. - */ - brw_LINE(p, brw_null_reg(), interp3, delta0); - brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), delta1); - - /* Calc w */ - brw_math_16( p, dst, - BRW_MATH_FUNCTION_INV, - BRW_MATH_SATURATE_NONE, - 2, brw_null_reg(), - BRW_MATH_PRECISION_FULL); - } -} - -static void emit_linterp(struct brw_wm_compile *c, - struct prog_instruction *inst) +static struct brw_reg get_src_reg(struct brw_wm_compile *c, + struct tgsi_full_src_register *src, + int index) { - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - struct brw_reg interp[4]; - struct brw_reg dst, delta0, delta1; - struct brw_reg src0; - - src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); - delta0 = get_src_reg(c, &inst->SrcReg[1], 0, 1); - delta1 = get_src_reg(c, &inst->SrcReg[1], 1, 1); - unsigned nr = src0.nr; - int i; - - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - - for(i = 0; i < 4; i++ ) { - if (mask & (1<SrcRegister.Negate) + neg = 1; + + component = get_swz(src->SrcRegister, component); + + /* Yes, there are multiple negates: + */ + switch (component & 3) { + case 0: neg ^= src->SrcRegisterExtSwz.NegateX; break; + case 1: neg ^= src->SrcRegisterExtSwz.NegateY; break; + case 2: neg ^= src->SrcRegisterExtSwz.NegateZ; break; + case 3: neg ^= src->SrcRegisterExtSwz.NegateW; break; + } + + /* And multiple swizzles, fun isn't it: + */ + component = get_ext_swz(src->SrcRegisterExtSwz, component); + + /* Can't handle this, don't know if we need to: + */ + assert(src->SrcRegisterExtSwz.ExtDivide == 0); + + /* Not handling indirect lookups yet: + */ + assert(src->SrcRegister.Indirect == 0); + + /* Don't know what dimension means: + */ + assert(src->SrcRegister.Dimension == 0); + + /* Will never handle any of this stuff: + */ + assert(src->SrcRegisterExtMod.Complement == 0); + assert(src->SrcRegisterExtMod.Bias == 0); + assert(src->SrcRegisterExtMod.Scale2X == 0); + + if (src->SrcRegisterExtMod.Absolute) + abs = 1; + + /* Another negate! This is a post-absolute negate, which we + * can't do. Need to clean the crap out of tgsi somehow. + */ + assert(src->SrcRegisterExtMod.Negate == 0); + + switch( component ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + reg = get_reg(c, + src->SrcRegister.File, + src->SrcRegister.Index, + component ); + + if (neg) + reg = negate(reg); + + if (abs) + reg = brw_abs(reg); + + break; + + /* XXX: this won't really work in the general case, but we know + * that the extended swizzle is only allowed in the SWZ + * instruction (right??), in which case using an immediate + * directly will work. + */ + case TGSI_EXTSWIZZLE_ZERO: + reg = brw_imm_f(0); + break; + + case TGSI_EXTSWIZZLE_ONE: + if (neg && !abs) + reg = brw_imm_f(-1.0); + else + reg = brw_imm_f(1.0); + break; + + default: + assert(0); + break; + } + + + return reg; } -static void emit_cinterp(struct brw_wm_compile *c, - struct prog_instruction *inst) +static void emit_abs( struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) { - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - - struct brw_reg interp[4]; - struct brw_reg dst, src0; - - src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); - unsigned nr = src0.nr; - int i; - - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - - for(i = 0; i < 4; i++ ) { - if (mask & (1<FullDstRegisters[0].DstRegister.WriteMask; + + int i; + struct brw_compile *p = &c->func; + brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + brw_MOV(p, dst, brw_abs(src)); /* NOTE */ + } + } + brw_set_saturate(p, 0); } -static void emit_pinterp(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - - struct brw_reg interp[4]; - struct brw_reg dst, delta0, delta1; - struct brw_reg src0, w; - - src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); - delta0 = get_src_reg(c, &inst->SrcReg[1], 0, 1); - delta1 = get_src_reg(c, &inst->SrcReg[1], 1, 1); - w = get_src_reg(c, &inst->SrcReg[2], 3, 1); - unsigned nr = src0.nr; - int i; - - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - - for(i = 0; i < 4; i++ ) { - if (mask & (1<func; - unsigned mask = inst->DstReg.WriteMask; - for (i = 0; i < 4; i++) { - unsigned i2 = (i+2)%3; - unsigned i1 = (i+1)%3; - if (mask & (1<SrcReg[0], i2, 1)); - src1 = get_src_reg(c, &inst->SrcReg[1], i1, 1); - brw_MUL(p, brw_null_reg(), src0, src1); - src0 = get_src_reg(c, &inst->SrcReg[0], i1, 1); - src1 = get_src_reg(c, &inst->SrcReg[1], i2, 1); - brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); - brw_MAC(p, dst, src0, src1); - brw_set_saturate(p, 0); - } - } - brw_set_saturate(p, 0); + int i; + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + for (i = 0; i < 4; i++) { + unsigned i2 = (i+2)%3; + unsigned i1 = (i+1)%3; + if (mask & (1<FullSrcRegisters[0], i2)); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i1); + brw_MUL(p, brw_null_reg(), src0, src1); + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i1); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i2); + brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); + brw_MAC(p, dst, src0, src1); + brw_set_saturate(p, 0); + } + } + brw_set_saturate(p, 0); } static void emit_dp3(struct brw_wm_compile *c, - struct prog_instruction *inst) + struct tgsi_full_instruction *inst) { - struct brw_reg src0[3], src1[3], dst; - int i; - struct brw_compile *p = &c->func; - for (i = 0; i < 3; i++) { - src0[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); - src1[i] = get_src_reg(c, &inst->SrcReg[1], i, 1); - } - - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); - brw_MUL(p, brw_null_reg(), src0[0], src1[0]); - brw_MAC(p, brw_null_reg(), src0[1], src1[1]); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_MAC(p, dst, src0[2], src1[2]); - brw_set_saturate(p, 0); + struct brw_reg src0[3], src1[3], dst; + int i; + struct brw_compile *p = &c->func; + for (i = 0; i < 3; i++) { + src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); + src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i); + } + + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MAC(p, dst, src0[2], src1[2]); + brw_set_saturate(p, 0); } static void emit_dp4(struct brw_wm_compile *c, - struct prog_instruction *inst) + struct tgsi_full_instruction *inst) { - struct brw_reg src0[4], src1[4], dst; - int i; - struct brw_compile *p = &c->func; - for (i = 0; i < 4; i++) { - src0[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); - src1[i] = get_src_reg(c, &inst->SrcReg[1], i, 1); - } - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); - brw_MUL(p, brw_null_reg(), src0[0], src1[0]); - brw_MAC(p, brw_null_reg(), src0[1], src1[1]); - brw_MAC(p, brw_null_reg(), src0[2], src1[2]); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_MAC(p, dst, src0[3], src1[3]); - brw_set_saturate(p, 0); + struct brw_reg src0[4], src1[4], dst; + int i; + struct brw_compile *p = &c->func; + for (i = 0; i < 4; i++) { + src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); + src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i); + } + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_MAC(p, brw_null_reg(), src0[2], src1[2]); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MAC(p, dst, src0[3], src1[3]); + brw_set_saturate(p, 0); } static void emit_dph(struct brw_wm_compile *c, - struct prog_instruction *inst) + struct tgsi_full_instruction *inst) { - struct brw_reg src0[4], src1[4], dst; - int i; - struct brw_compile *p = &c->func; - for (i = 0; i < 4; i++) { - src0[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); - src1[i] = get_src_reg(c, &inst->SrcReg[1], i, 1); - } - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); - brw_MUL(p, brw_null_reg(), src0[0], src1[0]); - brw_MAC(p, brw_null_reg(), src0[1], src1[1]); - brw_MAC(p, dst, src0[2], src1[2]); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_ADD(p, dst, src0[3], src1[3]); - brw_set_saturate(p, 0); + struct brw_reg src0[4], src1[4], dst; + int i; + struct brw_compile *p = &c->func; + for (i = 0; i < 4; i++) { + src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); + src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i); + } + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_MAC(p, dst, src0[2], src1[2]); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_ADD(p, dst, src0[3], src1[3]); + brw_set_saturate(p, 0); } static void emit_math1(struct brw_wm_compile *c, - struct prog_instruction *inst, unsigned func) + struct tgsi_full_instruction *inst, unsigned func) { - struct brw_compile *p = &c->func; - struct brw_reg src0, dst; + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; - src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); - brw_MOV(p, brw_message_reg(2), src0); - brw_math(p, + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + brw_MOV(p, brw_message_reg(2), src0); + brw_math(p, dst, func, - (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + ((inst->Instruction.Saturate != TGSI_SAT_NONE) + ? BRW_MATH_SATURATE_SATURATE + : BRW_MATH_SATURATE_NONE), 2, brw_null_reg(), BRW_MATH_DATA_VECTOR, BRW_MATH_PRECISION_FULL); } -static void emit_rcp(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_INV); -} -static void emit_rsq(struct brw_wm_compile *c, - struct prog_instruction *inst) +static void emit_alu2(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst, + unsigned opcode) { - emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ); + struct brw_compile *p = &c->func; + struct brw_reg src0, src1, dst; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + int i; + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + brw_alu2(p, opcode, dst, src0, src1); + } + } + brw_set_saturate(p, 0); } -static void emit_sin(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_SIN); -} -static void emit_cos(struct brw_wm_compile *c, - struct prog_instruction *inst) +static void emit_alu1(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst, + unsigned opcode) { - emit_math1(c, inst, BRW_MATH_FUNCTION_COS); + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + int i; + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + brw_alu1(p, opcode, dst, src0); + } + } + if (inst->Instruction.Saturate != TGSI_SAT_NONE) + brw_set_saturate(p, 0); } -static void emit_ex2(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_EXP); -} - -static void emit_lg2(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - emit_math1(c, inst, BRW_MATH_FUNCTION_LOG); -} - -static void emit_add(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, src1, dst; - unsigned mask = inst->DstReg.WriteMask; - int i; - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); - brw_ADD(p, dst, src0, src1); - } - } - brw_set_saturate(p, 0); -} - -static void emit_sub(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, src1, dst; - unsigned mask = inst->DstReg.WriteMask; - int i; - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); - brw_ADD(p, dst, src0, negate(src1)); - } - } - brw_set_saturate(p, 0); -} - -static void emit_mul(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, src1, dst; - unsigned mask = inst->DstReg.WriteMask; - int i; - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); - brw_MUL(p, dst, src0, src1); - } - } - brw_set_saturate(p, 0); -} - -static void emit_frc(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, dst; - unsigned mask = inst->DstReg.WriteMask; - int i; - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - brw_FRC(p, dst, src0); - } - } - if (inst->SaturateMode != SATURATE_OFF) - brw_set_saturate(p, 0); -} - -static void emit_flr(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, dst; - unsigned mask = inst->DstReg.WriteMask; - int i; - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - brw_RNDD(p, dst, src0); - } - } - brw_set_saturate(p, 0); -} static void emit_max(struct brw_wm_compile *c, - struct prog_instruction *inst) + struct tgsi_full_instruction *inst) { - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - struct brw_reg src0, src1, dst; - int i; - brw_push_insn_state(p); - for (i = 0; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_MOV(p, dst, src0); - brw_set_saturate(p, 0); - - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src0, src1); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_MOV(p, dst, src1); - brw_set_saturate(p, 0); - brw_set_predicate_control_flag_value(p, 0xff); - } - } - brw_pop_insn_state(p); + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg src0, src1, dst; + int i; + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MOV(p, dst, src0); + brw_set_saturate(p, 0); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src0, src1); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, src1); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + } + } + brw_pop_insn_state(p); } static void emit_min(struct brw_wm_compile *c, - struct prog_instruction *inst) + struct tgsi_full_instruction *inst) { - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - struct brw_reg src0, src1, dst; - int i; - brw_push_insn_state(p); - for (i = 0; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_MOV(p, dst, src0); - brw_set_saturate(p, 0); - - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src1, src0); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_MOV(p, dst, src1); - brw_set_saturate(p, 0); - brw_set_predicate_control_flag_value(p, 0xff); - } - } - brw_pop_insn_state(p); + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg src0, src1, dst; + int i; + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MOV(p, dst, src0); + brw_set_saturate(p, 0); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src1, src0); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, src1); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + } + } + brw_pop_insn_state(p); } static void emit_pow(struct brw_wm_compile *c, - struct prog_instruction *inst) + struct tgsi_full_instruction *inst) { - struct brw_compile *p = &c->func; - struct brw_reg dst, src0, src1; - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst), 1); - src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); - src1 = get_src_reg(c, &inst->SrcReg[1], 0, 1); + struct brw_compile *p = &c->func; + struct brw_reg dst, src0, src1; + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], 0); - brw_MOV(p, brw_message_reg(2), src0); - brw_MOV(p, brw_message_reg(3), src1); + brw_MOV(p, brw_message_reg(2), src0); + brw_MOV(p, brw_message_reg(3), src1); - brw_math(p, + brw_math(p, dst, BRW_MATH_FUNCTION_POW, - (inst->SaturateMode != SATURATE_OFF) ? BRW_MATH_SATURATE_SATURATE : BRW_MATH_SATURATE_NONE, + (inst->Instruction.Saturate != TGSI_SAT_NONE + ? BRW_MATH_SATURATE_SATURATE + : BRW_MATH_SATURATE_NONE), 2, brw_null_reg(), BRW_MATH_DATA_VECTOR, @@ -756,601 +440,636 @@ static void emit_pow(struct brw_wm_compile *c, } static void emit_lrp(struct brw_wm_compile *c, - struct prog_instruction *inst) + struct tgsi_full_instruction *inst) { - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - struct brw_reg dst, tmp1, tmp2, src0, src1, src2; - int i; - for (i = 0; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - - src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); - - if (src1.nr == dst.nr) { - tmp1 = alloc_tmp(c); - brw_MOV(p, tmp1, src1); - } else - tmp1 = src1; - - src2 = get_src_reg(c, &inst->SrcReg[2], i, 1); - if (src2.nr == dst.nr) { - tmp2 = alloc_tmp(c); - brw_MOV(p, tmp2, src2); - } else - tmp2 = src2; - - brw_ADD(p, dst, negate(src0), brw_imm_f(1.0)); - brw_MUL(p, brw_null_reg(), dst, tmp2); - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_MAC(p, dst, src0, tmp1); - brw_set_saturate(p, 0); - } - release_tmps(c); - } + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg dst, tmp1, tmp2, src0, src1, src2; + int i; + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + + if (src1.nr == dst.nr) { + tmp1 = alloc_tmp(c); + brw_MOV(p, tmp1, src1); + } else + tmp1 = src1; + + src2 = get_src_reg(c, &inst->FullSrcRegisters[2], i); + if (src2.nr == dst.nr) { + tmp2 = alloc_tmp(c); + brw_MOV(p, tmp2, src2); + } else + tmp2 = src2; + + brw_ADD(p, dst, negate(src0), brw_imm_f(1.0)); + brw_MUL(p, brw_null_reg(), dst, tmp2); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MAC(p, dst, src0, tmp1); + brw_set_saturate(p, 0); + } + release_tmps(c); + } } static void emit_kil(struct brw_wm_compile *c) { - struct brw_compile *p = &c->func; - struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); - brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK - brw_AND(p, depth, c->emit_mask_reg, depth); - brw_pop_insn_state(p); + struct brw_compile *p = &c->func; + struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK + brw_AND(p, depth, c->emit_mask_reg, depth); + brw_pop_insn_state(p); } static void emit_mad(struct brw_wm_compile *c, - struct prog_instruction *inst) + struct tgsi_full_instruction *inst) { - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - struct brw_reg dst, src0, src1, src2; - int i; - - for (i = 0; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); - src2 = get_src_reg(c, &inst->SrcReg[2], i, 1); - brw_MUL(p, dst, src0, src1); - - brw_set_saturate(p, (inst->SaturateMode != SATURATE_OFF) ? 1 : 0); - brw_ADD(p, dst, dst, src2); - brw_set_saturate(p, 0); - } - } + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg dst, src0, src1, src2; + int i; + + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + src2 = get_src_reg(c, &inst->FullSrcRegisters[2], i); + brw_MUL(p, dst, src0, src1); + + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_ADD(p, dst, dst, src2); + brw_set_saturate(p, 0); + } + } } static void emit_sop(struct brw_wm_compile *c, - struct prog_instruction *inst, unsigned cond) + struct tgsi_full_instruction *inst, unsigned cond) { - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - struct brw_reg dst, src0, src1; - int i; - - brw_push_insn_state(p); - for (i = 0; i < 4; i++) { - if (mask & (1<SrcReg[0], i, 1); - src1 = get_src_reg(c, &inst->SrcReg[1], i, 1); - brw_CMP(p, brw_null_reg(), cond, src0, src1); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_MOV(p, dst, brw_imm_f(0.0)); - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_MOV(p, dst, brw_imm_f(1.0)); - } - } - brw_pop_insn_state(p); + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg dst, src0, src1; + int i; + + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + brw_CMP(p, brw_null_reg(), cond, src0, src1); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_MOV(p, dst, brw_imm_f(0.0)); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, brw_imm_f(1.0)); + } + } + brw_pop_insn_state(p); } -static void emit_slt(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - emit_sop(c, inst, BRW_CONDITIONAL_L); -} -static void emit_sle(struct brw_wm_compile *c, - struct prog_instruction *inst) +static void emit_ddx(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) { - emit_sop(c, inst, BRW_CONDITIONAL_LE); + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg interp[4]; + struct brw_reg dst; + struct brw_reg src0, w; + unsigned nr, i; + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); + w = get_src_reg(c, &inst->FullSrcRegisters[1], 3); + nr = src0.nr; + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg interp[4]; + struct brw_reg dst; + struct brw_reg src0, w; + unsigned nr, i; + + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); + nr = src0.nr; + w = get_src_reg(c, &inst->FullSrcRegisters[1], 3); + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + struct brw_reg payload_reg = c->payload_depth[0]; + struct brw_reg dst[4], src[4]; + unsigned i; + for (i = 0; i < 4; i++) + dst[i] = get_dst_reg(c, inst, i); + for (i = 0; i < 4; i++) + src[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); -static void emit_seq(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - emit_sop(c, inst, BRW_CONDITIONAL_EQ); -} +#if 0 + switch (inst->TexSrcTarget) { + case TEXTURE_1D_INDEX: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), brw_imm_f(0)); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); + break; + case TEXTURE_2D_INDEX: + case TEXTURE_RECT_INDEX: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); + break; + default: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), src[2]); + break; + } +#else + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); +#endif -static void emit_sne(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - emit_sop(c, inst, BRW_CONDITIONAL_NEQ); + brw_MOV(p, brw_message_reg(5), src[3]); + brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); + brw_SAMPLE(p, + retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), + 1, + retype(payload_reg, BRW_REGISTER_TYPE_UW), + inst->TexSrcUnit + 1, /* surface */ + inst->TexSrcUnit, /* sampler */ + inst->FullDstRegisters[0].DstRegister.WriteMask, + BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS, + 4, + 4, + 0); +#endif } -static void emit_ddx(struct brw_wm_compile *c, - struct prog_instruction *inst) +static void emit_tex(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) { - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - struct brw_reg interp[4]; - struct brw_reg dst; - struct brw_reg src0, w; - unsigned nr, i; - src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); - w = get_src_reg(c, &inst->SrcReg[1], 3, 1); - nr = src0.nr; - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); - for(i = 0; i < 4; i++ ) { - if (mask & (1<func; + struct brw_reg payload_reg = c->payload_depth[0]; + struct brw_reg dst[4], src[4]; + unsigned msg_len; + unsigned i, nr; + unsigned emit; + boolean shadow = (c->key.shadowtex_mask & (1<TexSrcUnit)) ? 1 : 0; + + for (i = 0; i < 4; i++) + dst[i] = get_dst_reg(c, inst, i); + for (i = 0; i < 4; i++) + src[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); -static void emit_ddy(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - struct brw_reg interp[4]; - struct brw_reg dst; - struct brw_reg src0, w; - unsigned nr, i; - - src0 = get_src_reg(c, &inst->SrcReg[0], 0, 1); - nr = src0.nr; - w = get_src_reg(c, &inst->SrcReg[1], 3, 1); - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - brw_set_saturate(p, inst->SaturateMode != SATURATE_OFF); - for(i = 0; i < 4; i++ ) { - if (mask & (1<TexSrcTarget) { + case TEXTURE_1D_INDEX: + emit = WRITEMASK_X; + nr = 1; + break; + case TEXTURE_2D_INDEX: + case TEXTURE_RECT_INDEX: + emit = WRITEMASK_XY; + nr = 2; + break; + default: + emit = WRITEMASK_XYZ; + nr = 3; + break; + } +#else + emit = WRITEMASK_XY; + nr = 2; +#endif -static void emit_wpos_xy(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->DstReg.WriteMask; - struct brw_reg src0[2], dst[2]; - - dst[0] = get_dst_reg(c, inst, 0, 1); - dst[1] = get_dst_reg(c, inst, 1, 1); - - src0[0] = get_src_reg(c, &inst->SrcReg[0], 0, 1); - src0[1] = get_src_reg(c, &inst->SrcReg[0], 1, 1); - - /* Calc delta X,Y by subtracting origin in r1 from the pixel - * centers. - */ - if (mask & WRITEMASK_X) { - brw_MOV(p, - dst[0], - retype(src0[0], BRW_REGISTER_TYPE_UW)); - } - - if (mask & WRITEMASK_Y) { - /* TODO -- window_height - Y */ - brw_MOV(p, - dst[1], - retype(src0[1], BRW_REGISTER_TYPE_UW)); - - } + msg_len = 1; + + for (i = 0; i < nr; i++) { + static const unsigned swz[4] = {0,1,2,2}; + if (emit & (1<TexSrcUnit + 1, /* surface */ + inst->TexSrcUnit, /* sampler */ + inst->FullDstRegisters[0].DstRegister.WriteMask, + BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE, + 4, + shadow ? 6 : 4, + 0); + + if (shadow) + brw_MOV(p, dst[3], brw_imm_f(1.0)); +#endif } -/* TODO - BIAS on SIMD8 not workind yet... - */ -static void emit_txb(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg dst[4], src[4], payload_reg; - unsigned i; - payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0); - for (i = 0; i < 4; i++) - dst[i] = get_dst_reg(c, inst, i, 1); - for (i = 0; i < 4; i++) - src[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); - - switch (inst->TexSrcTarget) { - case TEXTURE_1D_INDEX: - brw_MOV(p, brw_message_reg(2), src[0]); - brw_MOV(p, brw_message_reg(3), brw_imm_f(0)); - brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); - break; - case TEXTURE_2D_INDEX: - case TEXTURE_RECT_INDEX: - brw_MOV(p, brw_message_reg(2), src[0]); - brw_MOV(p, brw_message_reg(3), src[1]); - brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); - break; - default: - brw_MOV(p, brw_message_reg(2), src[0]); - brw_MOV(p, brw_message_reg(3), src[1]); - brw_MOV(p, brw_message_reg(4), src[2]); - break; - } - brw_MOV(p, brw_message_reg(5), src[3]); - brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); - brw_SAMPLE(p, - retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), - 1, - retype(payload_reg, BRW_REGISTER_TYPE_UW), - inst->TexSrcUnit + 1, /* surface */ - inst->TexSrcUnit, /* sampler */ - inst->DstReg.WriteMask, - BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS, - 4, - 4, - 0); -} -static void emit_tex(struct brw_wm_compile *c, - struct prog_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg dst[4], src[4], payload_reg; - unsigned msg_len; - unsigned i, nr; - unsigned emit; - boolean shadow = (c->key.shadowtex_mask & (1<TexSrcUnit)) ? 1 : 0; - payload_reg = get_reg(c, PROGRAM_PAYLOAD, PAYLOAD_DEPTH, 0, 1, 0, 0); - for (i = 0; i < 4; i++) - dst[i] = get_dst_reg(c, inst, i, 1); - for (i = 0; i < 4; i++) - src[i] = get_src_reg(c, &inst->SrcReg[0], i, 1); - switch (inst->TexSrcTarget) { - case TEXTURE_1D_INDEX: - emit = WRITEMASK_X; - nr = 1; - break; - case TEXTURE_2D_INDEX: - case TEXTURE_RECT_INDEX: - emit = WRITEMASK_XY; - nr = 2; - break; - default: - emit = WRITEMASK_XYZ; - nr = 3; - break; - } - msg_len = 1; - - for (i = 0; i < nr; i++) { - static const unsigned swz[4] = {0,1,2,2}; - if (emit & (1<TexSrcUnit + 1, /* surface */ - inst->TexSrcUnit, /* sampler */ - inst->DstReg.WriteMask, - BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE, - 4, - shadow ? 6 : 4, - 0); - - if (shadow) - brw_MOV(p, dst[3], brw_imm_f(1.0)); -} -static void post_wm_emit( struct brw_wm_compile *c ) + +static void emit_fb_write(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) { - unsigned nr_insns = c->fp->program.Base.NumInstructions; - unsigned insn, target_insn; - struct prog_instruction *inst1, *inst2; - struct brw_instruction *brw_inst1, *brw_inst2; - int offset; - for (insn = 0; insn < nr_insns; insn++) { - inst1 = &c->fp->program.Base.Instructions[insn]; - brw_inst1 = inst1->Data; - switch (inst1->Opcode) { - case OPCODE_CAL: - target_insn = inst1->BranchTarget; - inst2 = &c->fp->program.Base.Instructions[target_insn]; - brw_inst2 = inst2->Data; - offset = brw_inst2 - brw_inst1; - brw_set_src1(brw_inst1, brw_imm_d(offset*16)); - break; - default: - break; - } - } + struct brw_compile *p = &c->func; + int nr = 2; + int channel; + int base_reg = 0; + + // src0 = output color + // src1 = payload_depth[0] + // src2 = output depth + // dst = ??? + + + + /* Reserve a space for AA - may not be needed: + */ + if (c->key.aa_dest_stencil_reg) + nr += 1; + + { + brw_push_insn_state(p); + for (channel = 0; channel < 4; channel++) { + struct brw_reg src0 = c->wm_regs[TGSI_FILE_OUTPUT][0][channel]; + + /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ + /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ + brw_MOV(p, brw_message_reg(nr + channel), src0); + } + /* skip over the regs populated above: */ + nr += 8; + brw_pop_insn_state(p); + } + + + /* Pass through control information: + */ + /* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ + { + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ + brw_MOV(p, + brw_message_reg(base_reg + 1), + brw_vec8_grf(1, 0)); + brw_pop_insn_state(p); + } + + /* Send framebuffer write message: */ + brw_fb_WRITE(p, + retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW), + base_reg, + retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), + 0, /* render surface always 0 */ + nr, + 0, + 1); + } -static void brw_wm_emit_glsl(struct brw_wm_compile *c) +static void brw_wm_emit_instruction( struct brw_wm_compile *c, + struct tgsi_full_instruction *inst ) { -#define MAX_IFSN 32 -#define MAX_LOOP_DEPTH 32 - struct brw_instruction *if_inst[MAX_IFSN], *loop_inst[MAX_LOOP_DEPTH]; - struct brw_instruction *inst0, *inst1; - int i, if_insn = 0, loop_insn = 0; - struct brw_compile *p = &c->func; - struct brw_indirect stack_index = brw_indirect(0, 0); - - brw_init_compile(&c->func); - c->reg_index = 0; - prealloc_reg(c); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); - - for (i = 0; i < c->nr_fp_insns; i++) { - struct prog_instruction *inst = &c->prog_instructions[i]; - struct prog_instruction *orig_inst; - - if ((orig_inst = inst->Data) != 0) - orig_inst->Data = current_insn(p); - - if (inst->CondUpdate) - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - else - brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); - - switch (inst->Opcode) { - case WM_PIXELXY: - emit_pixel_xy(c, inst); - break; - case WM_DELTAXY: - emit_delta_xy(c, inst); - break; - case WM_PIXELW: - emit_pixel_w(c, inst); - break; - case WM_LINTERP: - emit_linterp(c, inst); - break; - case WM_PINTERP: - emit_pinterp(c, inst); - break; - case WM_CINTERP: - emit_cinterp(c, inst); - break; - case WM_WPOSXY: - emit_wpos_xy(c, inst); - break; - case WM_FB_WRITE: - emit_fb_write(c, inst); - break; - case OPCODE_ABS: - emit_abs(c, inst); - break; - case OPCODE_ADD: - emit_add(c, inst); - break; - case OPCODE_SUB: - emit_sub(c, inst); - break; - case OPCODE_FRC: - emit_frc(c, inst); - break; - case OPCODE_FLR: - emit_flr(c, inst); - break; - case OPCODE_LRP: - emit_lrp(c, inst); - break; - case OPCODE_INT: - emit_int(c, inst); - break; - case OPCODE_MOV: - emit_mov(c, inst); - break; - case OPCODE_DP3: - emit_dp3(c, inst); - break; - case OPCODE_DP4: - emit_dp4(c, inst); - break; - case OPCODE_XPD: - emit_xpd(c, inst); - break; - case OPCODE_DPH: - emit_dph(c, inst); - break; - case OPCODE_RCP: - emit_rcp(c, inst); - break; - case OPCODE_RSQ: - emit_rsq(c, inst); - break; - case OPCODE_SIN: - emit_sin(c, inst); - break; - case OPCODE_COS: - emit_cos(c, inst); - break; - case OPCODE_EX2: - emit_ex2(c, inst); - break; - case OPCODE_LG2: - emit_lg2(c, inst); - break; - case OPCODE_MAX: - emit_max(c, inst); - break; - case OPCODE_MIN: - emit_min(c, inst); - break; - case OPCODE_DDX: - emit_ddx(c, inst); - break; - case OPCODE_DDY: - emit_ddy(c, inst); - break; - case OPCODE_SLT: - emit_slt(c, inst); - break; - case OPCODE_SLE: - emit_sle(c, inst); - break; - case OPCODE_SGT: - emit_sgt(c, inst); - break; - case OPCODE_SGE: - emit_sge(c, inst); - break; - case OPCODE_SEQ: - emit_seq(c, inst); - break; - case OPCODE_SNE: - emit_sne(c, inst); - break; - case OPCODE_MUL: - emit_mul(c, inst); - break; - case OPCODE_POW: - emit_pow(c, inst); - break; - case OPCODE_MAD: - emit_mad(c, inst); - break; - case OPCODE_TEX: - emit_tex(c, inst); - break; - case OPCODE_TXB: - emit_txb(c, inst); - break; - case OPCODE_KIL_NV: - emit_kil(c); - break; - case OPCODE_IF: - assert(if_insn < MAX_IFSN); - if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8); - break; - case OPCODE_ELSE: - if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]); - break; - case OPCODE_ENDIF: - assert(if_insn > 0); - brw_ENDIF(p, if_inst[--if_insn]); - break; - case OPCODE_BGNSUB: - case OPCODE_ENDSUB: - break; - case OPCODE_CAL: - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); - brw_set_access_mode(p, BRW_ALIGN_1); - brw_ADD(p, deref_1ud(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); - brw_set_access_mode(p, BRW_ALIGN_16); - brw_ADD(p, get_addr_reg(stack_index), - get_addr_reg(stack_index), brw_imm_d(4)); - orig_inst = inst->Data; - orig_inst->Data = &p->store[p->nr_insn]; - brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); - brw_pop_insn_state(p); - break; - - case OPCODE_RET: - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); - brw_ADD(p, get_addr_reg(stack_index), - get_addr_reg(stack_index), brw_imm_d(-4)); - brw_set_access_mode(p, BRW_ALIGN_1); - brw_MOV(p, brw_ip_reg(), deref_1ud(stack_index, 0)); - brw_set_access_mode(p, BRW_ALIGN_16); - brw_pop_insn_state(p); - - break; - case OPCODE_BGNLOOP: - loop_inst[loop_insn++] = brw_DO(p, BRW_EXECUTE_8); - break; - case OPCODE_BRK: - brw_BREAK(p); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - break; - case OPCODE_CONT: - brw_CONT(p); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - break; - case OPCODE_ENDLOOP: - loop_insn--; - inst0 = inst1 = brw_WHILE(p, loop_inst[loop_insn]); - /* patch all the BREAK instructions from - last BEGINLOOP */ - while (inst0 > loop_inst[loop_insn]) { - inst0--; - if (inst0->header.opcode == BRW_OPCODE_BREAK) { - inst0->bits3.if_else.jump_count = inst1 - inst0 + 1; - inst0->bits3.if_else.pop_count = 0; - } else if (inst0->header.opcode == BRW_OPCODE_CONTINUE) { - inst0->bits3.if_else.jump_count = inst1 - inst0; - inst0->bits3.if_else.pop_count = 0; - } - } - break; - default: - _mesa_printf("unsupported IR in fragment shader %d\n", - inst->Opcode); - } - if (inst->CondUpdate) - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - else - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - } - post_wm_emit(c); - for (i = 0; i < c->fp->program.Base.NumInstructions; i++) - c->fp->program.Base.Instructions[i].Data = NULL; + struct brw_compile *p = &c->func; + +#if 0 + if (inst->CondUpdate) + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + else + brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); +#else + brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); +#endif + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + emit_abs(c, inst); + break; + case TGSI_OPCODE_ADD: + emit_alu2(c, inst, BRW_OPCODE_ADD); + break; + case TGSI_OPCODE_SUB: + assert(0); +// emit_alu2(c, inst, BRW_OPCODE_SUB); + break; + case TGSI_OPCODE_FRC: + emit_alu1(c, inst, BRW_OPCODE_FRC); + break; + case TGSI_OPCODE_FLR: + assert(0); +// emit_alu1(c, inst, BRW_OPCODE_FLR); + break; + case TGSI_OPCODE_LRP: + emit_lrp(c, inst); + break; + case TGSI_OPCODE_INT: + emit_alu1(c, inst, BRW_OPCODE_RNDD); + break; + case TGSI_OPCODE_MOV: + emit_alu1(c, inst, BRW_OPCODE_MOV); + break; + case TGSI_OPCODE_DP3: + emit_dp3(c, inst); + break; + case TGSI_OPCODE_DP4: + emit_dp4(c, inst); + break; + case TGSI_OPCODE_XPD: + emit_xpd(c, inst); + break; + case TGSI_OPCODE_DPH: + emit_dph(c, inst); + break; + case TGSI_OPCODE_RCP: + emit_math1(c, inst, BRW_MATH_FUNCTION_INV); + break; + case TGSI_OPCODE_RSQ: + emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ); + break; + case TGSI_OPCODE_SIN: + emit_math1(c, inst, BRW_MATH_FUNCTION_SIN); + break; + case TGSI_OPCODE_COS: + emit_math1(c, inst, BRW_MATH_FUNCTION_COS); + break; + case TGSI_OPCODE_EX2: + emit_math1(c, inst, BRW_MATH_FUNCTION_EXP); + break; + case TGSI_OPCODE_LG2: + emit_math1(c, inst, BRW_MATH_FUNCTION_LOG); + break; + case TGSI_OPCODE_MAX: + emit_max(c, inst); + break; + case TGSI_OPCODE_MIN: + emit_min(c, inst); + break; + case TGSI_OPCODE_DDX: + emit_ddx(c, inst); + break; + case TGSI_OPCODE_DDY: + emit_ddy(c, inst); + break; + case TGSI_OPCODE_SLT: + emit_sop(c, inst, BRW_CONDITIONAL_L); + break; + case TGSI_OPCODE_SLE: + emit_sop(c, inst, BRW_CONDITIONAL_LE); + break; + case TGSI_OPCODE_SGT: + emit_sop(c, inst, BRW_CONDITIONAL_G); + break; + case TGSI_OPCODE_SGE: + emit_sop(c, inst, BRW_CONDITIONAL_GE); + break; + case TGSI_OPCODE_SEQ: + emit_sop(c, inst, BRW_CONDITIONAL_EQ); + break; + case TGSI_OPCODE_SNE: + emit_sop(c, inst, BRW_CONDITIONAL_NEQ); + break; + case TGSI_OPCODE_MUL: + emit_alu2(c, inst, BRW_OPCODE_MUL); + break; + case TGSI_OPCODE_POW: + emit_pow(c, inst); + break; + case TGSI_OPCODE_MAD: + emit_mad(c, inst); + break; + case TGSI_OPCODE_TEX: + emit_tex(c, inst); + break; + case TGSI_OPCODE_TXB: + emit_txb(c, inst); + break; + case TGSI_OPCODE_TEXKILL: + emit_kil(c); + break; + case TGSI_OPCODE_IF: + assert(c->if_insn < MAX_IFSN); + c->if_inst[c->if_insn++] = brw_IF(p, BRW_EXECUTE_8); + break; + case TGSI_OPCODE_ELSE: + c->if_inst[c->if_insn-1] = brw_ELSE(p, c->if_inst[c->if_insn-1]); + break; + case TGSI_OPCODE_ENDIF: + assert(c->if_insn > 0); + brw_ENDIF(p, c->if_inst[--c->if_insn]); + break; + case TGSI_OPCODE_BGNSUB: + case TGSI_OPCODE_ENDSUB: + break; + case TGSI_OPCODE_CAL: + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_ADD(p, deref_1ud(c->stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_ADD(p, + get_addr_reg(c->stack_index), + get_addr_reg(c->stack_index), brw_imm_d(4)); +// orig_inst = inst->Data; +// orig_inst->Data = &p->store[p->nr_insn]; + assert(0); + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + brw_pop_insn_state(p); + break; + + case TGSI_OPCODE_RET: + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_ADD(p, + get_addr_reg(c->stack_index), + get_addr_reg(c->stack_index), brw_imm_d(-4)); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, brw_ip_reg(), deref_1ud(c->stack_index, 0)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_pop_insn_state(p); + + break; + case TGSI_OPCODE_LOOP: + c->loop_inst[c->loop_insn++] = brw_DO(p, BRW_EXECUTE_8); + break; + case TGSI_OPCODE_BRK: + brw_BREAK(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case TGSI_OPCODE_CONT: + brw_CONT(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case TGSI_OPCODE_ENDLOOP: + c->loop_insn--; + c->inst0 = c->inst1 = brw_WHILE(p, c->loop_inst[c->loop_insn]); + /* patch all the BREAK instructions from + last BEGINLOOP */ + while (c->inst0 > c->loop_inst[c->loop_insn]) { + c->inst0--; + if (c->inst0->header.opcode == BRW_OPCODE_BREAK) { + c->inst0->bits3.if_else.jump_count = c->inst1 - c->inst0 + 1; + c->inst0->bits3.if_else.pop_count = 0; + } else if (c->inst0->header.opcode == BRW_OPCODE_CONTINUE) { + c->inst0->bits3.if_else.jump_count = c->inst1 - c->inst0; + c->inst0->bits3.if_else.pop_count = 0; + } + } + break; + case TGSI_OPCODE_END: + emit_fb_write(c, inst); + break; + + default: + _mesa_printf("unsupported IR in fragment shader %d\n", + inst->Instruction.Opcode); + } +#if 0 + if (inst->CondUpdate) + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + else + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +#endif } + + + + + void brw_wm_glsl_emit(struct brw_wm_compile *c) { - brw_wm_pass_fp(c); - c->tmp_index = 127; - brw_wm_emit_glsl(c); - c->prog_data.total_grf = c->reg_index; - c->prog_data.total_scratch = 0; -} + struct tgsi_parse_context parse; + struct brw_compile *p = &c->func; + + brw_init_compile(&c->func); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + c->reg_index = 0; + c->if_insn = 0; + c->loop_insn = 0; + c->stack_index = brw_indirect(0,0); + + /* Do static register allocation and parameter interpolation: + */ + brw_wm_emit_decls( c ); + + /* Emit the actual program. All done with very direct translation, + * hopefully we can improve on this shortly... + */ + brw_MOV(p, get_addr_reg(c->stack_index), brw_address(c->stack)); + + tgsi_parse_init( &parse, c->fp->program.tokens ); + + while( !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + /* already done */ + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* not handled yet */ + assert(0); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + brw_wm_emit_instruction(c, &parse.FullToken.FullInstruction); + break; + + default: + assert( 0 ); + } + } + + tgsi_parse_free (&parse); + + /* Fix up call targets: + */ +#if 0 + { + unsigned nr_insns = c->fp->program.Base.NumInstructions; + unsigned insn, target_insn; + struct tgsi_full_instruction *inst1, *inst2; + struct brw_instruction *brw_inst1, *brw_inst2; + int offset; + for (insn = 0; insn < nr_insns; insn++) { + inst1 = &c->fp->program.Base.Instructions[insn]; + brw_inst1 = inst1->Data; + switch (inst1->Opcode) { + case TGSI_OPCODE_CAL: + target_insn = inst1->BranchTarget; + inst2 = &c->fp->program.Base.Instructions[target_insn]; + brw_inst2 = inst2->Data; + offset = brw_inst2 - brw_inst1; + brw_set_src1(brw_inst1, brw_imm_d(offset*16)); + break; + default: + break; + } + } + } #endif + + c->prog_data.total_grf = c->reg_index; + c->prog_data.total_scratch = 0; +} diff --git a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c b/src/mesa/pipe/i965simple/brw_wm_sampler_state.c index 4ea0dd7db0..fbeea8c809 100644 --- a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_sampler_state.c @@ -229,12 +229,12 @@ static void brw_update_sampler_state( const struct pipe_sampler_state *pipe_samp * complicates various things. However, this is still too confusing - * FIXME: simplify all the different new texture state flags. */ -void brw_upload_wm_samplers(struct brw_context *brw) +static void upload_wm_samplers(struct brw_context *brw) { unsigned unit; unsigned sampler_count = 0; - /* _NEW_TEXTURE */ + /* BRW_NEW_SAMPLER */ for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { if (brw->attribs.Samplers[unit]) { /* FIXME: correctly detect enabled ones */ const struct pipe_sampler_state *sampler = brw->attribs.Samplers[unit]; @@ -262,14 +262,11 @@ void brw_upload_wm_samplers(struct brw_context *brw) sizeof(struct brw_sampler_state) * brw->wm.sampler_count); } -#if 0 const struct brw_tracked_state brw_wm_samplers = { .dirty = { - .mesa = _NEW_TEXTURE, - .brw = 0, + .brw = BRW_NEW_SAMPLER, .cache = 0 }, .update = upload_wm_samplers }; -#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_state.c b/src/mesa/pipe/i965simple/brw_wm_state.c index b45fb2f56b..52d2c85423 100644 --- a/src/mesa/pipe/i965simple/brw_wm_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_state.c @@ -34,15 +34,13 @@ #include "brw_state.h" #include "brw_defines.h" #include "brw_wm.h" +#include "pipe/p_util.h" /*********************************************************************** * WM unit - fragment programs and rasterization */ - -#if 0 static void upload_wm_unit(struct brw_context *brw ) { - struct intel_context *intel = &brw->intel; struct brw_wm_unit_state wm; unsigned max_threads; unsigned per_thread; @@ -56,7 +54,7 @@ static void upload_wm_unit(struct brw_context *brw ) memset(&wm, 0, sizeof(wm)); /* CACHE_NEW_WM_PROG */ - wm.thread0.grf_reg_count = ALIGN(brw->wm.prog_data->total_grf, 16) / 16 - 1; + wm.thread0.grf_reg_count = align(brw->wm.prog_data->total_grf, 16) / 16 - 1; wm.thread0.kernel_start_pointer = brw->wm.prog_gs_offset >> 6; wm.thread3.dispatch_grf_start_reg = brw->wm.prog_data->first_curbe_grf; wm.thread3.urb_entry_read_length = brw->wm.prog_data->urb_read_length; @@ -64,9 +62,10 @@ static void upload_wm_unit(struct brw_context *brw ) wm.wm5.max_threads = max_threads; - per_thread = ALIGN(brw->wm.prog_data->total_scratch, 1024); + per_thread = align(brw->wm.prog_data->total_scratch, 1024); assert(per_thread <= 12 * 1024); +#if 0 if (brw->wm.prog_data->total_scratch) { unsigned total = per_thread * (max_threads + 1); @@ -95,6 +94,7 @@ static void upload_wm_unit(struct brw_context *brw ) * so just fail for now if we hit that path. */ assert(brw->wm.prog_data->total_scratch == 0); +#endif /* CACHE_NEW_SURFACE */ wm.thread1.binding_table_entry_count = brw->wm.nr_surfaces; @@ -112,23 +112,20 @@ static void upload_wm_unit(struct brw_context *brw ) /* BRW_NEW_FRAGMENT_PROGRAM */ { - const struct gl_fragment_program *fp = brw->fragment_program; + const struct brw_fragment_program *fp = brw->attribs.FragmentProgram; - if (fp->Base.InputsRead & (1<UsesDepth) wm.wm5.program_uses_depth = 1; /* as far as we can tell */ - if (fp->Base.OutputsWritten & (1<ComputesDepth) wm.wm5.program_computes_depth = 1; - /* _NEW_COLOR */ + /* BRW_NEW_ALPHA_TEST */ if (fp->UsesKill || - brw->attribs.Color->AlphaEnabled) + brw->attribs.AlphaTest->enabled) wm.wm5.program_uses_killpixel = 1; - if (brw_wm_is_glsl(fp)) - wm.wm5.enable_8_pix = 1; - else - wm.wm5.enable_16_pix = 1; + wm.wm5.enable_8_pix = 1; } wm.wm5.thread_dispatch_enable = 1; /* AKA: color_write */ @@ -138,11 +135,11 @@ static void upload_wm_unit(struct brw_context *brw ) wm.wm5.line_aa_region_width = 0; wm.wm5.line_endcap_aa_region_width = 1; - /* _NEW_POLYGONSTIPPLE */ - if (brw->attribs.Polygon->StippleFlag) + /* BRW_NEW_RASTERIZER */ + if (brw->attribs.Raster->poly_stipple_enable) wm.wm5.polygon_stipple = 1; - /* _NEW_POLYGON */ +#if 0 if (brw->attribs.Polygon->OffsetFill) { wm.wm5.depth_offset = 1; /* Something wierd going on with legacy_global_depth_bias, @@ -156,13 +153,13 @@ static void upload_wm_unit(struct brw_context *brw ) */ wm.global_depth_offset_scale = brw->attribs.Polygon->OffsetFactor; } +#endif - /* _NEW_LINE */ - if (brw->attribs.Line->StippleFlag) { + if (brw->attribs.Raster->line_stipple_enable) { wm.wm5.line_stipple = 1; } - if (BRW_DEBUG & DEBUG_STATS || intel->stats_wm) + if (BRW_DEBUG & DEBUG_STATS) wm.wm4.stats_enable = 1; brw->wm.state_gs_offset = brw_cache_data( &brw->cache[BRW_WM_UNIT], &wm ); @@ -183,14 +180,10 @@ static void upload_wm_unit(struct brw_context *brw ) const struct brw_tracked_state brw_wm_unit = { .dirty = { - .mesa = (_NEW_POLYGON | - _NEW_POLYGONSTIPPLE | - _NEW_LINE | - _NEW_COLOR), - - .brw = (BRW_NEW_FRAGMENT_PROGRAM | - BRW_NEW_CURBE_OFFSETS | - BRW_NEW_LOCK), + .brw = (BRW_NEW_RASTERIZER | + BRW_NEW_ALPHA_TEST | + BRW_NEW_FS | + BRW_NEW_CURBE_OFFSETS), .cache = (CACHE_NEW_SURFACE | CACHE_NEW_WM_PROG | @@ -199,4 +192,3 @@ const struct brw_tracked_state brw_wm_unit = { .update = upload_wm_unit }; -#endif diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index e6d284d932..46edcf3075 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -360,6 +360,11 @@ static INLINE float LOG2(float val) #define CEILF(x) ((float) ceil(x)) #endif +static INLINE int align(int value, int alignment) +{ + return (value + alignment - 1) & ~(alignment - 1); +} + /* Convenient... */ extern void _mesa_printf(const char *str, ...); diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index 2dd1add6f7..44512e4281 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -52,10 +52,6 @@ static unsigned minify( unsigned d ) return MAX2(1, d>>1); } -static int align(int value, int alignment) -{ - return (value + alignment - 1) & ~(alignment - 1); -} static void -- cgit v1.2.3 From 7a2d3ac5bcad87d139f97eaa9ca2aa90d174240a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Dec 2007 21:00:10 +0000 Subject: 965: get brw_wm_surface_state compiling again --- src/mesa/pipe/i965simple/brw_state.c | 3 +++ src/mesa/pipe/i965simple/brw_wm_surface_state.c | 11 ++++++++--- 2 files changed, 11 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c index 032a4e0285..26450ae597 100644 --- a/src/mesa/pipe/i965simple/brw_state.c +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -203,6 +203,9 @@ static void * brw_create_fs_state(struct pipe_context *pipe, brw_shader_info(shader->tokens, &brw_fp->info); + tgsi_dump(shader->tokens, 0); + + return (void *)brw_fp; } diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index 2901e28adb..6e68c4c660 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -189,6 +189,9 @@ static void upload_wm_surfaces(struct brw_context *brw ) { struct brw_surface_state surf; + + /* BRW_NEW_FRAMEBUFFER + */ struct pipe_surface *region = brw->attribs.FrameBuffer.cbufs[0];/*fixme*/ memset(&surf, 0, sizeof(surf)); @@ -213,7 +216,7 @@ static void upload_wm_surfaces(struct brw_context *brw ) surf.ss0.surface_type = BRW_SURFACE_NULL; } - /* _NEW_COLOR */ + /* BRW_NEW_BLEND */ surf.ss0.color_blend = (!brw->attribs.Blend->logicop_enable && brw->attribs.Blend->blend_enable); @@ -235,7 +238,7 @@ static void upload_wm_surfaces(struct brw_context *brw ) for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { const struct brw_texture *texUnit = brw->attribs.Texture[i]; - /* _NEW_TEXTURE, BRW_NEW_TEXDATA + /* BRW_NEW_TEXTURE */ if (texUnit->base.refcount/*(texUnit->refcount > 0) == really used */) { @@ -299,7 +302,9 @@ static void emit_reloc_wm_surfaces(struct brw_context *brw) const struct brw_tracked_state brw_wm_surfaces = { .dirty = { - .brw = BRW_NEW_CONTEXT, + .brw = (BRW_NEW_FRAMEBUFFER | + BRW_NEW_BLEND | + BRW_NEW_TEXTURE), .cache = 0 }, .update = upload_wm_surfaces, -- cgit v1.2.3 From e2d7107cc2427a0e61949ead50cf0ac2c2bc6f11 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Dec 2007 18:26:10 -0700 Subject: remove unused var --- src/mesa/pipe/softpipe/sp_context.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index dcf0444f6e..b62e691e87 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -55,8 +55,6 @@ static boolean softpipe_is_format_supported( struct pipe_context *pipe, enum pipe_format format, uint type ) { - struct softpipe_context *softpipe = softpipe_context( pipe ); - switch (type) { case PIPE_TEXTURE: /* softpipe supports all texture formats */ -- cgit v1.2.3 From a2b4d4a8db77443b834daf811802a909768fa926 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Dec 2007 18:26:48 -0700 Subject: add missing code for PIPE_FORMAT_S8_UNORM renderbuffer --- src/mesa/state_tracker/st_cb_fbo.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 254740ff20..c40f75417f 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -243,6 +243,10 @@ st_new_renderbuffer_fb(enum pipe_format format) strb->Base.InternalFormat = GL_DEPTH24_STENCIL8_EXT; strb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; break; + case PIPE_FORMAT_S8_UNORM: + strb->Base.InternalFormat = GL_STENCIL_INDEX8_EXT; + strb->Base._BaseFormat = GL_STENCIL_INDEX; + break; case PIPE_FORMAT_R16G16B16A16_SNORM: strb->Base.InternalFormat = GL_RGBA16; strb->Base._BaseFormat = GL_RGBA; -- cgit v1.2.3 From 23e36c2dfb1f9501a6a1023afc1d0c151f2e99c3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Dec 2007 10:46:29 -0700 Subject: update comment for TGSI_TOKEN_TYPE_IMMEDIATE case --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 1cd554250c..d517b88acc 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -928,8 +928,9 @@ i915_translate_instructions(struct i915_fp_compile *p, break; case TGSI_TOKEN_TYPE_IMMEDIATE: - /* XXX no-op? */ - assert(0); + /* This is a no-op. We'll get immediates from the usual constant/ + * uniform buffer. + */ break; case TGSI_TOKEN_TYPE_INSTRUCTION: -- cgit v1.2.3 From e785f190f0d49f0367f7468c22b77962d0f14ea0 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Dec 2007 11:00:46 -0700 Subject: Don't always declare frag shader INPUT[0] as fragment position. We were doing this for the sake of softpipe and the tgsi intergrepter since we always need the fragment position and W-coordinate information in order to compute fragment interpolants. But that's not appropriate for hardware drivers. The tgsi interpreter now get x,y,w information from a separate tgsi_exec_vector variable setup by softpipe. The new pipe_shader_state->input_map[] defines how vert shader outputs map to frag shader inputs. It may go away though, since one can also examine the semantic label on frag shader input[0] to figure things out. --- src/mesa/pipe/p_state.h | 1 + src/mesa/pipe/softpipe/sp_context.h | 2 - src/mesa/pipe/softpipe/sp_headers.h | 1 + src/mesa/pipe/softpipe/sp_prim_setup.c | 339 +++++++++++++++++++----------- src/mesa/pipe/softpipe/sp_quad_earlyz.c | 6 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 68 +++--- src/mesa/pipe/softpipe/sp_state_derived.c | 12 +- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 27 +-- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 2 +- src/mesa/state_tracker/st_atom_shader.c | 25 ++- src/mesa/state_tracker/st_mesa_to_tgsi.c | 46 ++-- src/mesa/state_tracker/st_program.c | 25 ++- 12 files changed, 324 insertions(+), 230 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 4e42838f1d..43b710ff3b 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -139,6 +139,7 @@ struct pipe_shader_state { const struct tgsi_token *tokens; ubyte num_inputs; ubyte num_outputs; + ubyte input_map[PIPE_MAX_SHADER_INPUTS]; /* XXX this may be temporary */ ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 2c038de5f7..8fd44933f2 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -110,8 +110,6 @@ struct softpipe_context { struct vertex_info vertex_info; unsigned attr_mask; unsigned nr_frag_attrs; /**< number of active fragment attribs */ - boolean need_z; /**< produce quad/fragment Z values? */ - boolean need_w; /**< produce quad/fragment W values? */ int psize_slot; #if 0 diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h index b9f2b2205a..0ae31d8796 100644 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ b/src/mesa/pipe/softpipe/sp_headers.h @@ -73,6 +73,7 @@ struct quad_header { float coverage[QUAD_SIZE]; /** fragment coverage for antialiasing */ const struct tgsi_interp_coef *coef; + const struct tgsi_interp_coef *posCoef; unsigned nr_attrs; }; diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index fc96f92af1..8d8dceadc5 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -36,10 +36,12 @@ #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" +#include "sp_state.h" #include "sp_prim_setup.h" #include "pipe/draw/draw_private.h" #include "pipe/draw/draw_vertex.h" #include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" #define DEBUG_VERTS 0 @@ -80,8 +82,11 @@ struct setup_stage { float oneoverarea; struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; + struct tgsi_interp_coef posCoef; /* For Z, W */ struct quad_header quad; + uint firstFpInput; /** Semantic type of first frag input */ + struct { int left[2]; /**< [0] = row0, [1] = row1 */ int right[2]; @@ -365,18 +370,17 @@ static boolean setup_sort_vertices( struct setup_stage *setup, * \param i which component of the slot (0..3) */ static void const_coeff( struct setup_stage *setup, - unsigned slot, - unsigned i ) + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) { - assert(slot < PIPE_MAX_SHADER_INPUTS); assert(i <= 3); - setup->coef[slot].dadx[i] = 0; - setup->coef[slot].dady[i] = 0; + coef->dadx[i] = 0; + coef->dady[i] = 0; /* need provoking vertex info! */ - setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; + coef->a0[i] = setup->vprovoke->data[vertSlot][i]; } @@ -385,19 +389,20 @@ static void const_coeff( struct setup_stage *setup, * for a triangle. */ static void tri_linear_coeff( struct setup_stage *setup, - unsigned slot, - unsigned i) + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) { - float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; - float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + float botda = setup->vmid->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; + float majda = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; float a = setup->ebot.dy * majda - botda * setup->emaj.dy; float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - - assert(slot < PIPE_MAX_SHADER_INPUTS); + float dadx = a * setup->oneoverarea; + float dady = b * setup->oneoverarea; + assert(i <= 3); - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; + coef->dadx[i] = dadx; + coef->dady[i] = dady; /* calculate a0 as the value which would be sampled for the * fragment at (0,0), taking into account that we want to sample at @@ -411,9 +416,9 @@ static void tri_linear_coeff( struct setup_stage *setup, * to define a0 as the sample at a pixel center somewhere near vmin * instead - i'll switch to this later. */ - setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); + coef->a0[i] = (setup->vmin->data[vertSlot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); /* _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", @@ -434,39 +439,68 @@ static void tri_linear_coeff( struct setup_stage *setup, * divide the interpolated value by the interpolated W at that fragment. */ static void tri_persp_coeff( struct setup_stage *setup, - unsigned slot, - unsigned i ) + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) { - /* premultiply by 1/w: + /* premultiply by 1/w (v->data[0][3] is always W): */ - float mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; - float mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; - float maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; - + float mina = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; + float mida = setup->vmid->data[vertSlot][i] * setup->vmid->data[0][3]; + float maxa = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; float botda = mida - mina; float majda = maxa - mina; float a = setup->ebot.dy * majda - botda * setup->emaj.dy; float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float dadx = a * setup->oneoverarea; + float dady = b * setup->oneoverarea; /* - printf("tri persp %d,%d: %f %f %f\n", slot, i, - setup->vmin->data[slot][i], - setup->vmid->data[slot][i], - setup->vmax->data[slot][i] + printf("tri persp %d,%d: %f %f %f\n", vertSlot, i, + setup->vmin->data[vertSlot][i], + setup->vmid->data[vertSlot][i], + setup->vmax->data[vertSlot][i] ); */ - - assert(slot < PIPE_MAX_SHADER_INPUTS); assert(i <= 3); - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; - setup->coef[slot].a0[i] = (mina - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (mina - + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); +} + + +/** + * Special coefficient setup for gl_FragCoord. + * X and Y are trivial, though Y has to be inverted for OpenGL. + * Z and W are copied from posCoef which should have already been computed. + * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask. + */ +static void +setup_fragcoord_coeff(struct setup_stage *setup) +{ + const int winHeight = setup->softpipe->framebuffer.cbufs[0]->height; + /*X*/ + setup->coef[0].a0[0] = 0; + setup->coef[0].dadx[0] = 1.0; + setup->coef[0].dady[0] = 0.0; + /*Y*/ + setup->coef[0].a0[1] = winHeight - 1; + setup->coef[0].dadx[1] = 0.0; + setup->coef[0].dady[1] = -1.0; + /*Z*/ + setup->coef[0].a0[2] = setup->posCoef.a0[2]; + setup->coef[0].dadx[2] = setup->posCoef.dadx[2]; + setup->coef[0].dady[2] = setup->posCoef.dady[2]; + /*w*/ + setup->coef[0].a0[3] = setup->posCoef.a0[3]; + setup->coef[0].dadx[3] = setup->posCoef.dadx[3]; + setup->coef[0].dady[3] = setup->posCoef.dady[3]; } + /** * Compute the setup->coef[] array dadx, dady, a0 values. * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. @@ -474,36 +508,67 @@ static void tri_persp_coeff( struct setup_stage *setup, static void setup_tri_coefficients( struct setup_stage *setup ) { const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; - unsigned slot, j; +#define USE_INPUT_MAP 0 +#if USE_INPUT_MAP + const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; +#endif + uint fragSlot; /* z and w are done by linear interpolation: */ - tri_linear_coeff(setup, 0, 2); - tri_linear_coeff(setup, 0, 3); + tri_linear_coeff(setup, &setup->posCoef, 0, 2); + tri_linear_coeff(setup, &setup->posCoef, 0, 3); /* setup interpolation for all the remaining attributes: */ - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - switch (interp[slot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - tri_linear_coeff(setup, slot, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - tri_persp_coeff(setup, slot, j); - break; - - default: - /* invalid interp mode */ - assert(0); + for (fragSlot = 0; fragSlot < setup->quad.nr_attrs; fragSlot++) { + /* which vertex output maps to this fragment input: */ +#if !USE_INPUT_MAP + uint vertSlot; + if (setup->firstFpInput == TGSI_SEMANTIC_POSITION) { + if (fragSlot == 0) { + setup_fragcoord_coeff(setup); + continue; + } + vertSlot = fragSlot; + } + else { + vertSlot = fragSlot + 1; } + +#else + uint vertSlot = fs->input_map[fragSlot]; + + if (vertSlot == 0) { + /* special case: shader is reading gl_FragCoord */ + /* XXX with a new INTERP_POSITION token, we could just add a + * new case to the switch below. + */ + setup_fragcoord_coeff(setup); + } + else { +#endif + uint j; + switch (interp[vertSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + default: + /* invalid interp mode */ + assert(0); + } +#if USE_INPUT_MAP + } +#endif } } @@ -660,17 +725,18 @@ static void setup_tri( struct draw_stage *stage, * for a line. */ static void -line_linear_coeff(struct setup_stage *setup, unsigned slot, unsigned i) +line_linear_coeff(struct setup_stage *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) { - const float da = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + const float da = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; const float dadx = da * setup->emaj.dx * setup->oneoverarea; const float dady = da * setup->emaj.dy * setup->oneoverarea; - setup->coef[slot].dadx[i] = dadx; - setup->coef[slot].dady[i] = dady; - setup->coef[slot].a0[i] - = (setup->vmin->data[slot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (setup->vmin->data[vertSlot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); } @@ -679,21 +745,21 @@ line_linear_coeff(struct setup_stage *setup, unsigned slot, unsigned i) * for a line. */ static void -line_persp_coeff(struct setup_stage *setup, unsigned slot, unsigned i) +line_persp_coeff(struct setup_stage *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) { /* XXX double-check/verify this arithmetic */ - const float a0 = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; - const float a1 = setup->vmax->data[slot][i] * setup->vmin->data[0][3]; + const float a0 = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; + const float a1 = setup->vmax->data[vertSlot][i] * setup->vmin->data[0][3]; const float da = a1 - a0; const float dadx = da * setup->emaj.dx * setup->oneoverarea; const float dady = da * setup->emaj.dy * setup->oneoverarea; - setup->coef[slot].dadx[i] = dadx; - setup->coef[slot].dady[i] = dady; - setup->coef[slot].a0[i] - = (setup->vmin->data[slot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); - + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (setup->vmin->data[vertSlot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); } @@ -705,7 +771,8 @@ static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; - unsigned slot, j; + const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; + unsigned fragSlot; /* use setup->vmin, vmax to point to vertices */ setup->vprovoke = prim->v[1]; @@ -720,31 +787,39 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) /* z and w are done by linear interpolation: */ - line_linear_coeff(setup, 0, 2); - line_linear_coeff(setup, 0, 3); + line_linear_coeff(setup, &setup->posCoef, 0, 2); + line_linear_coeff(setup, &setup->posCoef, 0, 3); /* setup interpolation for all the remaining attributes: */ - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - switch (interp[slot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - line_linear_coeff(setup, slot, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - line_persp_coeff(setup, slot, j); - break; - - default: - /* invalid interp mode */ - assert(0); + for (fragSlot = 0; fragSlot < setup->quad.nr_attrs; fragSlot++) { + /* which vertex output maps to this fragment input: */ + uint vertSlot = fs->input_map[fragSlot]; + + if (vertSlot == 0) { + /* special case: shader is reading gl_FragCoord */ + setup_fragcoord_coeff(setup); + } + else { + uint j; + switch (interp[vertSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + + default: + /* invalid interp mode */ + assert(0); + } } } } @@ -910,14 +985,15 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) static void -point_persp_coeff(struct setup_stage *setup, const struct vertex_header *vert, - uint slot, uint i) +point_persp_coeff(struct setup_stage *setup, + const struct vertex_header *vert, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) { - assert(slot < PIPE_MAX_SHADER_INPUTS); assert(i <= 3); - setup->coef[slot].dadx[i] = 0.0F; - setup->coef[slot].dady[i] = 0.0F; - setup->coef[slot].a0[i] = vert->data[slot][i] * vert->data[0][3]; + coef->dadx[i] = 0.0F; + coef->dady[i] = 0.0F; + coef->a0[i] = vert->data[vertSlot][i] * vert->data[0][3]; } @@ -930,6 +1006,7 @@ static void setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); + const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; const struct vertex_header *v0 = prim->v[0]; const int sizeAttr = setup->softpipe->psize_slot; @@ -940,7 +1017,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth; const float x = v0->data[0][0]; /* Note: data[0] is always position */ const float y = v0->data[0][1]; - unsigned slot, j; + uint fragSlot; /* For points, all interpolants are constant-valued. * However, for point sprites, we'll need to setup texcoords appropriately. @@ -959,22 +1036,36 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) * probably should be ruled out on that basis. */ setup->vprovoke = prim->v[0]; - const_coeff(setup, 0, 2); - const_coeff(setup, 0, 3); - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - switch (interp[slot]) { - case INTERP_CONSTANT: - /* fall-through */ - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - point_persp_coeff(setup, v0, slot, j); - break; - default: - assert(0); + + /* setup Z, W */ + const_coeff(setup, &setup->posCoef, 0, 2); + const_coeff(setup, &setup->posCoef, 0, 3); + + for (fragSlot = 0; fragSlot < setup->quad.nr_attrs; fragSlot++) { + /* which vertex output maps to this fragment input: */ + uint vertSlot = fs->input_map[fragSlot]; + + if (vertSlot == 0) { + /* special case: shader is reading gl_FragCoord */ + setup_fragcoord_coeff(setup); + } + else { + uint j; + switch (interp[vertSlot]) { + case INTERP_CONSTANT: + /* fall-through */ + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + point_persp_coeff(setup, setup->vprovoke, + &setup->coef[fragSlot], vertSlot, j); + break; + default: + assert(0); + } } } @@ -1108,9 +1199,12 @@ static void setup_begin( struct draw_stage *stage ) { struct setup_stage *setup = setup_stage(stage); struct softpipe_context *sp = setup->softpipe; + const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; + setup->firstFpInput = fs->input_semantic_name[0]; + sp->quad.first->begin(sp->quad.first); } @@ -1151,6 +1245,7 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) setup->stage.destroy = render_destroy; setup->quad.coef = setup->coef; + setup->quad.posCoef = &setup->posCoef; return &setup->stage; } diff --git a/src/mesa/pipe/softpipe/sp_quad_earlyz.c b/src/mesa/pipe/softpipe/sp_quad_earlyz.c index 3abd1f1fb9..22ea99049f 100644 --- a/src/mesa/pipe/softpipe/sp_quad_earlyz.c +++ b/src/mesa/pipe/softpipe/sp_quad_earlyz.c @@ -47,9 +47,9 @@ earlyz_quad( { const float fx = (float) quad->x0; const float fy = (float) quad->y0; - const float dzdx = quad->coef[0].dadx[2]; - const float dzdy = quad->coef[0].dady[2]; - const float z0 = quad->coef[0].a0[2] + dzdx * fx + dzdy * fy; + const float dzdx = quad->posCoef->dadx[2]; + const float dzdy = quad->posCoef->dady[2]; + const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy; quad->outputs.depth[0] = z0; quad->outputs.depth[1] = z0 + dzdx; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index da590b256a..6e7e7eb074 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -74,15 +74,49 @@ quad_shade_stage(struct quad_stage *qs) } +/** + * Compute quad X,Y,Z,W for the four fragments in a quad. + * Note that we only need to "compute" X and Y for the upper-left fragment. + * We could do less work if we're not depth testing, or there's no + * perspective-corrected attributes, but that's seldom. + */ +static void +setup_pos_vector(const struct tgsi_interp_coef *coef, + float x, float y, + struct tgsi_exec_vector *quadpos) +{ + uint chan; + /* do X */ + quadpos->xyzw[0].f[0] = x; + /* do Y */ + quadpos->xyzw[1].f[0] = y; + /* do Z and W for all fragments in the quad */ + for (chan = 2; chan < 4; chan++) { + const float dadx = coef->dadx[chan]; + const float dady = coef->dady[chan]; + const float a0 = coef->a0[chan] + dadx * x + dady * y; + quadpos->xyzw[chan].f[0] = a0; + quadpos->xyzw[chan].f[1] = a0 + dadx; + quadpos->xyzw[chan].f[2] = a0 + dady; + quadpos->xyzw[chan].f[3] = a0 + dadx + dady; + } +} + + typedef void (XSTDCALL *codegen_function)( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, float (*constant)[4], struct tgsi_exec_vector *temporary, - const struct tgsi_interp_coef *coef ); + const struct tgsi_interp_coef *coef +#if 0 + ,const struct tgsi_exec_vector *quadPos +#endif + ); + -/* This should be done by the fragment shader execution unit (code - * generated from the decl instructions). Do it here for now. +/** + * Execute fragment shader for the four fragments in the quad. */ static void shade_quad( @@ -91,33 +125,15 @@ shade_quad( { struct quad_shade_stage *qss = quad_shade_stage( qs ); struct softpipe_context *softpipe = qs->softpipe; - const float fx = (float) quad->x0; - const float fy = (float) quad->y0; struct tgsi_exec_machine *machine = &qss->machine; - /* Consts does not require 16 byte alignment. */ + /* Consts do not require 16 byte alignment. */ machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; machine->InterpCoefs = quad->coef; -#if 1 /* XXX only do this if the fp really reads fragment.position */ - machine->Inputs[0].xyzw[0].f[0] = fx; - machine->Inputs[0].xyzw[0].f[1] = fx + 1.0f; - machine->Inputs[0].xyzw[0].f[2] = fx; - machine->Inputs[0].xyzw[0].f[3] = fx + 1.0f; - - /* XXX for OpenGL we need to invert the Y pos here (y=0=top). - * but that'll mess up linear/perspective interpolation of other - * attributes... - */ - machine->Inputs[0].xyzw[1].f[0] = fy; - machine->Inputs[0].xyzw[1].f[1] = fy; - machine->Inputs[0].xyzw[1].f[2] = fy + 1.0f; - machine->Inputs[0].xyzw[1].f[3] = fy + 1.0f; -#endif - - machine->QuadX = quad->x0; - machine->QuadY = quad->y0; + /* Compute X, Y, Z, W vals for this quad */ + setup_pos_vector(quad->posCoef, quad->x0, quad->y0, &machine->QuadPos); /* run shader */ #if defined(__i386__) || defined(__386__) @@ -130,9 +146,9 @@ shade_quad( machine->Temps, machine->InterpCoefs #if 0 - ,quad->x0, quad->y0 + ,machine->QuadPos #endif - ); + ); quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]); } else diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index c4f1a0a01a..736ac1c33b 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -51,18 +51,11 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) memset(vinfo, 0, sizeof(*vinfo)); - if (softpipe->depth_stencil->depth.enabled) - softpipe->need_z = TRUE; - else - softpipe->need_z = FALSE; - softpipe->need_w = FALSE; if (fs->input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { /* Need Z if depth test is enabled or the fragment program uses the * fragment position (XYZW). */ - softpipe->need_z = TRUE; - softpipe->need_w = TRUE; } softpipe->psize_slot = -1; @@ -121,7 +114,6 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); - softpipe->need_w = TRUE; break; default: @@ -129,7 +121,11 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) } } +#if 00 softpipe->nr_frag_attrs = vinfo->num_attribs; +#else + softpipe->nr_frag_attrs = fs->num_inputs; +#endif /* We want these after all other attribs since they won't get passed * to the fragment shader. All prior vertex output attribs should match diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 03997f9099..1f43f3643e 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1352,7 +1352,8 @@ linear_interpolation( unsigned attrib, unsigned chan ) { - const float x = mach->QuadX, y = mach->QuadY; + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; const float dadx = mach->InterpCoefs[attrib].dadx[chan]; const float dady = mach->InterpCoefs[attrib].dady[chan]; const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; @@ -1368,14 +1369,17 @@ perspective_interpolation( unsigned attrib, unsigned chan ) { - const float x = mach->QuadX, y = mach->QuadY; + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; const float dadx = mach->InterpCoefs[attrib].dadx[chan]; const float dady = mach->InterpCoefs[attrib].dady[chan]; const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; - mach->Inputs[attrib].xyzw[chan].f[0] = a0 / mach->Inputs[0].xyzw[3].f[0]; - mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / mach->Inputs[0].xyzw[3].f[1]; - mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / mach->Inputs[0].xyzw[3].f[2]; - mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / mach->Inputs[0].xyzw[3].f[3]; + const float *w = mach->QuadPos.xyzw[3].f; + /* divide by W here */ + mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; + mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; + mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; + mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; } @@ -1400,17 +1404,6 @@ exec_declaration( last = decl->u.DeclarationRange.Last; mask = decl->Declaration.UsageMask; - /* Do not touch WPOS.xy */ - if( first == 0 ) { - mask &= ~TGSI_WRITEMASK_XY; - if( mask == TGSI_WRITEMASK_NONE ) { - first++; - if( first > last ) { - return; - } - } - } - switch( decl->Interpolation.Interpolate ) { case TGSI_INTERPOLATE_CONSTANT: interp = constant_interpolation; diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 8d166bb5f4..db92e282df 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -170,7 +170,6 @@ struct tgsi_exec_machine struct tgsi_exec_vector *Inputs; struct tgsi_exec_vector *Outputs; const struct tgsi_token *Tokens; - float QuadX, QuadY; /**< for frag progs only */ unsigned Processor; /* GEOMETRY processor only. */ @@ -178,6 +177,7 @@ struct tgsi_exec_machine /* FRAGMENT processor only. */ const struct tgsi_interp_coef *InterpCoefs; + struct tgsi_exec_vector QuadPos; /* Conditional execution masks */ uint CondMask; /**< For IF/ELSE/ENDIF */ diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 4ec10badad..33372b0f39 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -151,8 +151,7 @@ find_translated_vp(struct st_context *st, { static const GLuint UNUSED = ~0; struct translated_vertex_program *xvp; - const GLbitfield fragInputsRead - = stfp->Base.Base.InputsRead | FRAG_BIT_WPOS; + const GLbitfield fragInputsRead = stfp->Base.Base.InputsRead; /* * Translate fragment program if needed. @@ -206,6 +205,7 @@ find_translated_vp(struct st_context *st, if (xvp->serialNo != stvp->serialNo) { GLuint outAttr, dummySlot; const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten; + GLuint numVpOuts = 0; /* Compute mapping of vertex program outputs to slots, which depends * on the fragment program's input->slot mapping. @@ -214,11 +214,24 @@ find_translated_vp(struct st_context *st, /* set default: */ xvp->output_to_slot[outAttr] = UNUSED; - if (outputsWritten & (1 << outAttr)) { + if (outAttr == VERT_RESULT_HPOS) { + /* always put xformed position into slot zero */ + xvp->output_to_slot[VERT_RESULT_HPOS] = 0; + numVpOuts++; + } + else if (outputsWritten & (1 << outAttr)) { /* see if the frag prog wants this vert output */ - GLint fpIn = vp_out_to_fp_in(outAttr); - if (fpIn >= 0) { - xvp->output_to_slot[outAttr] = stfp->input_to_slot[fpIn]; + GLint fpInAttrib = vp_out_to_fp_in(outAttr); + if (fpInAttrib >= 0) { + GLuint fpInSlot = stfp->input_to_slot[fpInAttrib]; + GLuint vpOutSlot = stfp->fs->state.input_map[fpInSlot]; + xvp->output_to_slot[outAttr] = vpOutSlot; + numVpOuts++; + } + else if (outAttr == VERT_RESULT_BFC0 || + outAttr == VERT_RESULT_BFC1) { + /* backface colors go into last slots */ + xvp->output_to_slot[outAttr] = numVpOuts++; } } } diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index b392edf16d..27dab5b9c0 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -675,44 +675,22 @@ tgsi_translate_mesa_program( if (procType == TGSI_PROCESSOR_FRAGMENT) { for (i = 0; i < numInputs; i++) { struct tgsi_full_declaration fulldecl; - switch (inputSemanticName[i]) { - case TGSI_SEMANTIC_POSITION: - /* Fragment XY pos */ - fulldecl = make_input_decl(i, - GL_TRUE, TGSI_INTERPOLATE_CONSTANT, - TGSI_WRITEMASK_XY, - GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); - ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - /* Fragment ZW pos */ - fulldecl = make_input_decl(i, - GL_TRUE, TGSI_INTERPOLATE_LINEAR, - TGSI_WRITEMASK_ZW, - GL_TRUE, TGSI_SEMANTIC_POSITION, 0 ); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - default: - fulldecl = make_input_decl(i, - GL_TRUE, interpMode[i], - TGSI_WRITEMASK_XYZW, - GL_TRUE, inputSemanticName[i], - inputSemanticIndex[i]); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); - break; - } + fulldecl = make_input_decl(i, + GL_TRUE, interpMode[i], + TGSI_WRITEMASK_XYZW, + GL_TRUE, inputSemanticName[i], + inputSemanticIndex[i]); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); } } else { /* vertex prog */ + /* XXX: this could probaby be merged with the clause above. + * the only difference is the semantic tags. + */ for (i = 0; i < numInputs; i++) { struct tgsi_full_declaration fulldecl; fulldecl = make_input_decl(i, diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index e64bf14d56..fe22233c93 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -47,7 +47,7 @@ #include "st_mesa_to_tgsi.h" -#define TGSI_DEBUG 0 +#define TGSI_DEBUG 01 /** @@ -283,16 +283,17 @@ st_translate_fragment_program(struct st_context *st, const struct cso_fragment_shader *cso; GLuint interpMode[16]; /* XXX size? */ GLuint attr; - GLbitfield inputsRead = stfp->Base.Base.InputsRead; - - /* For software rendering, we always need the fragment input position - * in order to calculate interpolated values. - * For i915, we always want to emit the semantic info for position. - */ - inputsRead |= FRAG_BIT_WPOS; + const GLbitfield inputsRead = stfp->Base.Base.InputsRead; + GLuint vslot = 0; memset(&fs, 0, sizeof(fs)); + /* which vertex output goes to the first fragment input: */ + if (inputsRead & FRAG_BIT_WPOS) + vslot = 0; + else + vslot = 1; + /* * Convert Mesa program inputs to TGSI input register semantics. */ @@ -300,15 +301,17 @@ st_translate_fragment_program(struct st_context *st, if (inputsRead & (1 << attr)) { const GLuint slot = fs.num_inputs; - fs.num_inputs++; - defaultInputMapping[attr] = slot; + fs.input_map[slot] = vslot++; + + fs.num_inputs++; + switch (attr) { case FRAG_ATTRIB_WPOS: fs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; fs.input_semantic_index[slot] = 0; - interpMode[slot] = TGSI_INTERPOLATE_CONSTANT; + interpMode[slot] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_COL0: fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; -- cgit v1.2.3 From 1e3b07f363e6bf512ab1a5c620656985aece40fd Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Dec 2007 11:16:49 -0700 Subject: set program->SamplersUsed bit when using a texture instruction --- src/mesa/main/texenvprogram.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index efb3b35f6a..1e46d8c375 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -936,10 +936,12 @@ static void load_texture( struct texenv_fragment_program *p, GLuint unit ) /* TODO: Use D0_MASK_XY where possible. */ - if (p->state->unit[unit].enabled) + if (p->state->unit[unit].enabled) { p->src_texture[unit] = emit_texld( p, OPCODE_TXP, tmp, WRITEMASK_XYZW, unit, dim, texcoord ); + p->program->Base.SamplersUsed |= (1 << unit); + } else p->src_texture[unit] = get_zero(p); } -- cgit v1.2.3 From 814d4ff83a515fe161cfb86c8b1492e2218051fe Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Dec 2007 11:17:48 -0700 Subject: emit tgsi declarations for texture samplers --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 27dab5b9c0..bccabd8004 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -585,6 +585,20 @@ make_temp_decl( } +static struct tgsi_full_declaration +make_sampler_decl(GLuint index) +{ + struct tgsi_full_declaration decl; + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_SAMPLER; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.u.DeclarationRange.First = index; + decl.u.DeclarationRange.Last = index; + return decl; +} + + + /** * Find the temporaries which are used in the given program. */ @@ -788,6 +802,19 @@ tgsi_translate_mesa_program( } } + /* texture samplers */ + for (i = 0; i < 8; i++) { + if (program->SamplersUsed & (1 << i)) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_sampler_decl( i ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + } + + for( i = 0; i < program->NumInstructions; i++ ) { compile_instruction( &program->Instructions[i], -- cgit v1.2.3 From fce4612f8a29ee1798c9326a431a139d856c7a04 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Dec 2007 11:42:28 -0700 Subject: set SamplerUnit[] entry in load_texture() just to be safe --- src/mesa/main/texenvprogram.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index 1e46d8c375..d866d10017 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -941,6 +941,10 @@ static void load_texture( struct texenv_fragment_program *p, GLuint unit ) tmp, WRITEMASK_XYZW, unit, dim, texcoord ); p->program->Base.SamplersUsed |= (1 << unit); + /* This identity mapping should already be in place + * (see _mesa_init_program_struct()) but let's be safe. + */ + p->program->Base.SamplerUnits[unit] = unit; } else p->src_texture[unit] = get_zero(p); -- cgit v1.2.3 From f3b3ea9742e6511fa46332c2c6d2433f96cc5c10 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Dec 2007 11:43:15 -0700 Subject: update_samplers() didn't respect the sampler->texunit mapping. This fixes the glsl/texdemo1.c program. --- src/mesa/state_tracker/st_atom_sampler.c | 25 ++++++++++++++++--------- 1 file changed, 16 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 052b6dd144..6241e70b55 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -35,6 +35,7 @@ #include "st_context.h" #include "st_cache.h" #include "st_atom.h" +#include "st_program.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -116,17 +117,23 @@ gl_filter_to_img_filter(GLenum filter) static void update_samplers(struct st_context *st) { - GLuint u; + const struct st_fragment_program *fs = st->fp; + GLuint su; - for (u = 0; u < st->ctx->Const.MaxTextureImageUnits; u++) { - const struct gl_texture_object *texobj - = st->ctx->Texture.Unit[u]._Current; + /* loop over sampler units (aka tex image units) */ + for (su = 0; su < st->ctx->Const.MaxTextureImageUnits; su++) { struct pipe_sampler_state sampler; const struct cso_sampler *cso; memset(&sampler, 0, sizeof(sampler)); - if (texobj) { + if (fs->Base.Base.SamplersUsed & (1 << su)) { + GLuint texUnit = fs->Base.Base.SamplerUnits[su]; + const struct gl_texture_object *texobj + = st->ctx->Texture.Unit[texUnit]._Current; + + assert(texobj); + sampler.wrap_s = gl_wrap_to_sp(texobj->WrapS); sampler.wrap_t = gl_wrap_to_sp(texobj->WrapT); sampler.wrap_r = gl_wrap_to_sp(texobj->WrapR); @@ -138,7 +145,7 @@ update_samplers(struct st_context *st) if (texobj->Target != GL_TEXTURE_RECTANGLE_ARB) sampler.normalized_coords = 1; - sampler.lod_bias = st->ctx->Texture.Unit[u].LodBias; + sampler.lod_bias = st->ctx->Texture.Unit[su].LodBias; #if 1 sampler.min_lod = texobj->MinLod; sampler.max_lod = texobj->MaxLod; @@ -166,10 +173,10 @@ update_samplers(struct st_context *st) cso = st_cached_sampler_state(st, &sampler); - if (cso != st->state.sampler[u]) { + if (cso != st->state.sampler[su]) { /* state has changed */ - st->state.sampler[u] = cso; - st->pipe->bind_sampler_state(st->pipe, u, cso->data); + st->state.sampler[su] = cso; + st->pipe->bind_sampler_state(st->pipe, su, cso->data); } } } -- cgit v1.2.3 From 017f862de1f857bca29f09794539aaf411014f13 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Dec 2007 12:25:25 -0700 Subject: Added origin_lower_left field to pipe_rasterizer_state This controls whether the window origin is considered to be the lower-left or upper-left corner. This effects computation of gl_FragCoord and the application of polygon stipple. --- src/mesa/pipe/p_state.h | 1 + src/mesa/pipe/softpipe/sp_prim_setup.c | 14 +++++++++++--- src/mesa/pipe/softpipe/sp_quad_stipple.c | 16 ++++++++++++---- src/mesa/state_tracker/st_atom_rasterizer.c | 2 ++ 4 files changed, 26 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 43b710ff3b..af65d365bf 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -94,6 +94,7 @@ struct pipe_rasterizer_state unsigned line_stipple_factor:8; /**< [1..256] actually */ unsigned line_stipple_pattern:16; unsigned bypass_clipping:1; + unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ float line_width; float point_size; /**< used when no per-vertex size */ diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 8d8dceadc5..2ccf5e2624 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -480,15 +480,23 @@ static void tri_persp_coeff( struct setup_stage *setup, static void setup_fragcoord_coeff(struct setup_stage *setup) { - const int winHeight = setup->softpipe->framebuffer.cbufs[0]->height; /*X*/ setup->coef[0].a0[0] = 0; setup->coef[0].dadx[0] = 1.0; setup->coef[0].dady[0] = 0.0; /*Y*/ - setup->coef[0].a0[1] = winHeight - 1; + if (setup->softpipe->rasterizer->origin_lower_left) { + /* y=0=bottom */ + const int winHeight = setup->softpipe->framebuffer.cbufs[0]->height; + setup->coef[0].a0[1] = winHeight - 1; + setup->coef[0].dady[1] = -1.0; + } + else { + /* y=0=top */ + setup->coef[0].a0[1] = 0.0; + setup->coef[0].dady[1] = 1.0; + } setup->coef[0].dadx[1] = 0.0; - setup->coef[0].dady[1] = -1.0; /*Z*/ setup->coef[0].a0[2] = setup->posCoef.a0[2]; setup->coef[0].dadx[2] = setup->posCoef.dadx[2]; diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index 04d95989c4..0c42963dfe 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -22,10 +22,18 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) if (quad->prim == PRIM_TRI) { struct softpipe_context *softpipe = qs->softpipe; /* need to invert Y to index into OpenGL's stipple pattern */ - const int y0 = softpipe->framebuffer.cbufs[0]->height - 1 - quad->y0; - const int y1 = y0 - 1; - const unsigned stipple0 = softpipe->poly_stipple.stipple[y0 % 32]; - const unsigned stipple1 = softpipe->poly_stipple.stipple[y1 % 32]; + int y0, y1; + uint stipple0, stipple1; + if (softpipe->rasterizer->origin_lower_left) { + y0 = softpipe->framebuffer.cbufs[0]->height - 1 - quad->y0; + y1 = y0 - 1; + } + else { + y0 = quad->y0; + y1 = y0 + 1; + } + stipple0 = softpipe->poly_stipple.stipple[y0 % 32]; + stipple1 = softpipe->poly_stipple.stipple[y1 % 32]; #if 1 const int col0 = quad->x0 % 32; diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 2a7128dd27..5c6b89d78c 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -77,6 +77,8 @@ static void update_raster_state( struct st_context *st ) uint i; memset(&raster, 0, sizeof(raster)); + + raster.origin_lower_left = 1; /* Always true for OpenGL */ /* _NEW_POLYGON, _NEW_BUFFERS */ -- cgit v1.2.3 From de1201a0ba05068ccdc731f0a79395ce58b9121a Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 15 Dec 2007 09:49:20 -0700 Subject: don't allocate scratch tile in sp_tile_cache_flush_clear() to avoid stack overflow --- src/mesa/pipe/softpipe/sp_tile_cache.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index d637c7291f..a7be3812fe 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -58,6 +58,8 @@ struct softpipe_tile_cache struct pipe_surface *tex_surf; int tex_face, tex_level, tex_z; + + struct softpipe_cached_tile tile; /**< scratch tile for clears */ }; @@ -279,11 +281,10 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, const uint w = tc->surface->width; const uint h = tc->surface->height; uint x, y; - struct softpipe_cached_tile tile; uint numCleared = 0; - /* clear one tile to the clear value */ - clear_tile(&tile, ps->format, tc->clear_val); + /* clear the scratch tile to the clear value */ + clear_tile(&tc->tile, ps->format, tc->clear_val); /* push the tile to all positions marked as clear */ for (y = 0; y < h; y += TILE_SIZE) { @@ -291,7 +292,7 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, if (is_clear_flag_set(tc->clear_flags, x, y)) { pipe->put_tile(pipe, ps, x, y, TILE_SIZE, TILE_SIZE, - tile.data.color32, 0/*STRIDE*/); + tc->tile.data.color32, 0/*STRIDE*/); /* do this? */ clear_clear_flag(tc->clear_flags, x, y); -- cgit v1.2.3 From bccc4c5c0116987041dec2327ddb43eff1bcfadf Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 15 Dec 2007 16:42:22 -0700 Subject: Re-usable tile get/put functions --- src/mesa/pipe/util/p_tile.c | 834 ++++++++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/util/p_tile.h | 81 +++++ 2 files changed, 915 insertions(+) create mode 100644 src/mesa/pipe/util/p_tile.c create mode 100644 src/mesa/pipe/util/p_tile.h (limited to 'src') diff --git a/src/mesa/pipe/util/p_tile.c b/src/mesa/pipe/util/p_tile.c new file mode 100644 index 0000000000..8bc3c22c83 --- /dev/null +++ b/src/mesa/pipe/util/p_tile.c @@ -0,0 +1,834 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * RGBA/float tile get/put functions. + * Usable both by drivers and state trackers. + * Surfaces should already be in a mapped state. + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" + +#include "p_tile.h" + + + +/** + * Move raw block of pixels from surface to user memory. + * This should be usable by any hw driver that has mappable surfaces. + */ +void +pipe_get_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride) +{ + const uint cpp = ps->cpp; + const ubyte *pSrc; + const uint src_stride = ps->pitch * cpp; + ubyte *pDest; + uint i; + + assert(ps->map); + + if (dst_stride == 0) { + dst_stride = w * cpp; + } + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + pSrc = ps->map + (y * ps->pitch + x) * cpp; + pDest = (ubyte *) p; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w * cpp); + pDest += dst_stride; + pSrc += src_stride; + } +} + + +/** + * Move raw block of pixels from user memory to surface. + * This should be usable by any hw driver that has mappable surfaces. + */ +void +pipe_put_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride) +{ + const uint cpp = ps->cpp; + const ubyte *pSrc; + const uint dst_stride = ps->pitch * cpp; + ubyte *pDest; + uint i; + + assert(ps->map); + + if (src_stride == 0) { + src_stride = w * cpp; + } + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + pSrc = (const ubyte *) p; + pDest = ps->map + (y * ps->pitch + x) * cpp; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w * cpp); + pDest += dst_stride; + pSrc += src_stride; + } +} + + + + +/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */ +#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) + +#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ + us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) + + + +/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ + +static void +a8r8g8b8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const unsigned *src + = ((const unsigned *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const unsigned pixel = src[j]; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +static void +a8r8g8b8_put_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + unsigned *dst + = ((unsigned *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + dst[j] = (a << 24) | (r << 16) | (g << 8) | b; + pRow += 4; + } + dst += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ + +static void +b8g8r8a8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const unsigned *src + = ((const unsigned *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const unsigned pixel = src[j]; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +static void +b8g8r8a8_put_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + unsigned *dst + = ((unsigned *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + dst[j] = (b << 24) | (g << 16) | (r << 8) | a; + pRow += 4; + } + dst += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ + +static void +a1r5g5b5_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + + assert(ps->format == PIPE_FORMAT_A1R5G5B5_UNORM); + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const ushort pixel = src[j]; + p[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); + p[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); + p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + p[3] = ((pixel >> 15) ) * 1.0f; + p += 4; + } + src += ps->pitch; + } +} + + +/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ + +static void +a4r4g4b4_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + + assert(ps->format == PIPE_FORMAT_A4R4G4B4_UNORM); + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const ushort pixel = src[j]; + p[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); + p[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); + p[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); + p[3] = ((pixel >> 12) ) * (1.0f / 15.0f); + p += 4; + } + src += ps->pitch; + } +} + + +/*** PIPE_FORMAT_R5G6B5_UNORM ***/ + +static void +r5g6b5_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + + assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); + + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + const ushort pixel = src[j]; + p[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); + p[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); + p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + p[3] = 1.0f; + p += 4; + } + src += ps->pitch; + } +} + + +static void +r5g5b5_put_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + ushort *dst + = ((ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); + uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); + uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); + dst[j] = (r << 11) | (g << 5) | (b); + pRow += 4; + } + dst += ps->pitch; + p += w0 * 4; + } +} + + + +/*** PIPE_FORMAT_Z16_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z16_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + const float scale = 1.0f / 65535.0f; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_Z16_UNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = src[j] * scale; + } + src += ps->pitch; + p += 4 * w0; + } +} + + + + +/*** PIPE_FORMAT_U_L8 ***/ + +static void +l8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ubyte *src + = ((const ubyte *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_L8); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(src[j]); + pRow[3] = 1.0; + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_U_A8 ***/ + +static void +a8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ubyte *src + = ((const ubyte *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_A8); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[0] = + pRow[1] = + pRow[2] = 0.0; + pRow[3] = UBYTE_TO_FLOAT(src[j]); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ + +static void +r16g16b16a16_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const short *src + = ((const short *) (ps->map)) + + (y * ps->pitch + x) * 4; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + const short *pixel = src; + for (j = 0; j < w; j++) { + pRow[0] = SHORT_TO_FLOAT(pixel[0]); + pRow[1] = SHORT_TO_FLOAT(pixel[1]); + pRow[2] = SHORT_TO_FLOAT(pixel[2]); + pRow[3] = SHORT_TO_FLOAT(pixel[3]); + pRow += 4; + pixel += 4; + } + src += ps->pitch * 4; + p += w0 * 4; + } +} + + +static void +r16g16b16a16_put_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + const float *p) +{ + short *dst + = ((short *) (ps->map)) + + (y * ps->pitch + x) * 4; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++) { + short r, g, b, a; + UNCLAMPED_FLOAT_TO_SHORT(r, pRow[0]); + UNCLAMPED_FLOAT_TO_SHORT(g, pRow[1]); + UNCLAMPED_FLOAT_TO_SHORT(b, pRow[2]); + UNCLAMPED_FLOAT_TO_SHORT(a, pRow[3]); + dst[j*4+0] = r; + dst[j*4+1] = g; + dst[j*4+2] = b; + dst[j*4+3] = a; + pRow += 4; + } + dst += ps->pitch * 4; + p += w0 * 4; + } +} + + + +/*** PIPE_FORMAT_U_I8 ***/ + +static void +i8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ubyte *src + = ((const ubyte *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_I8); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = UBYTE_TO_FLOAT(src[j]); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + +/*** PIPE_FORMAT_U_A8_L8 ***/ + +static void +a8_l8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const ushort *src + = ((const ushort *) (ps->map)) + + y * ps->pitch + x; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_U_A8_L8); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + const ushort p = src[j]; + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(p & 0xff); + pRow[3] = UBYTE_TO_FLOAT(p >> 8); + pRow += 4; + } + src += ps->pitch; + p += w0 * 4; + } +} + + + + +/*** PIPE_FORMAT_Z32_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z32_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const uint *src + = ((const uint *) (ps->map)) + + y * ps->pitch + x; + const double scale = 1.0 / (double) 0xffffffff; + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_Z16_UNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = (float) (scale * src[j]); + } + src += ps->pitch; + p += 4 * w0; + } +} + + +/*** PIPE_FORMAT_S8Z24_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +s8z24_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const uint *src + = ((const uint *) (ps->map)) + + y * ps->pitch + x; + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_S8Z24_UNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = (float) (scale * (src[j] & 0xffffff)); + } + src += ps->pitch; + p += 4 * w0; + } +} + + +/*** PIPE_FORMAT_Z24S8_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +z24s8_get_tile_rgba(struct pipe_surface *ps, + unsigned x, unsigned y, unsigned w, unsigned h, + float *p) +{ + const uint *src + = ((const uint *) (ps->map)) + + y * ps->pitch + x; + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + unsigned w0 = w; + + assert(ps->format == PIPE_FORMAT_Z24S8_UNORM); + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++) { + pRow[j * 4 + 0] = + pRow[j * 4 + 1] = + pRow[j * 4 + 2] = + pRow[j * 4 + 3] = (float) (scale * (src[j] >> 8)); + } + src += ps->pitch; + p += 4 * w0; + } +} + + +void +pipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p) +{ + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + a1r5g5b5_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_A4R4G4B4_UNORM: + a4r4g4b4_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g6b5_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_L8: + l8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_A8: + a8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_I8: + i8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_U_A8_L8: + a8_l8_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_Z16_UNORM: + z16_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_Z32_UNORM: + z32_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_S8Z24_UNORM: + s8z24_get_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_Z24S8_UNORM: + z24s8_get_tile_rgba(ps, x, y, w, h, p); + break; + default: + assert(0); + } +} + + +void +pipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p) +{ + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_put_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_put_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + /*a1r5g5b5_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g5b5_put_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + break; + case PIPE_FORMAT_U_L8: + /*l8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_A8: + /*a8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_I8: + /*i8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_U_A8_L8: + /*a8_l8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_put_tile_rgba(ps, x, y, w, h, p); + break; + case PIPE_FORMAT_Z16_UNORM: + /*z16_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_Z32_UNORM: + /*z32_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_S8Z24_UNORM: + /*s8z24_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + case PIPE_FORMAT_Z24S8_UNORM: + /*z24s8_put_tile_rgba(ps, x, y, w, h, p);*/ + break; + default: + assert(0); + } +} diff --git a/src/mesa/pipe/util/p_tile.h b/src/mesa/pipe/util/p_tile.h new file mode 100644 index 0000000000..cd8124bf11 --- /dev/null +++ b/src/mesa/pipe/util/p_tile.h @@ -0,0 +1,81 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_TILE_H +#define P_TILE_H + +#include "pipe/p_compiler.h" + +struct pipe_context; +struct pipe_surface; + + +/** + * Clip tile against surface dims. + * \return TRUE if tile is totally clipped, FALSE otherwise + */ +static INLINE boolean +pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps) +{ + if (x >= ps->width) + return TRUE; + if (y >= ps->height) + return TRUE; + if (x + *w > ps->width) + *w = ps->width - x; + if (y + *h > ps->height) + *h = ps->height - y; + return FALSE; +} + + +extern void +pipe_get_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride); + +extern void +pipe_put_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride); + + +extern void +pipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p); + +extern void +pipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p); + +#endif -- cgit v1.2.3 From 263e8f057290961c9dd603ffa69f25a7a253301a Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 15 Dec 2007 16:44:29 -0700 Subject: Use tile functions in new p_tile.[ch] This removes quite a bit of code duplicated in the drivers. --- src/mesa/pipe/i915simple/i915_surface.c | 92 ++----------------------------- src/mesa/pipe/i965simple/brw_surface.c | 92 ++----------------------------- src/mesa/pipe/softpipe/Makefile | 1 - src/mesa/pipe/softpipe/sp_surface.c | 95 ++++----------------------------- src/mesa/sources | 4 ++ 5 files changed, 23 insertions(+), 261 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index d4c02c5359..34064612a7 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -33,89 +33,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" - -#include "pipe/softpipe/sp_rgba_tile.h" /* XXX TEMPORARY */ - - -#define CLIP_TILE \ - do { \ - if (x >= ps->width) \ - return; \ - if (y >= ps->height) \ - return; \ - if (x + w > ps->width) \ - w = ps->width - x; \ - if (y + h > ps->height) \ - h = ps->height -y; \ - } while(0) - - -/* - * XXX note: same as code in sp_surface.c - */ -static void -i915_get_tile(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride) -{ - const uint cpp = ps->cpp; - const uint w0 = w; - const ubyte *pSrc; - ubyte *pDest; - uint i; - - assert(ps->map); - - CLIP_TILE; - - if (dst_stride == 0) { - dst_stride = w0 * cpp; - } - - pSrc = ps->map + (y * ps->pitch + x) * cpp; - pDest = (ubyte *) p; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w0 * cpp); - pDest += dst_stride; - pSrc += ps->pitch * cpp; - } -} - - -/* - * XXX note: same as code in sp_surface.c - */ -static void -i915_put_tile(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride) -{ - const uint cpp = ps->cpp; - const uint w0 = w; - const ubyte *pSrc; - ubyte *pDest; - uint i; - - assert(ps->map); - - CLIP_TILE; - - if (src_stride == 0) { - src_stride = w0 * cpp; - } - - pSrc = (const ubyte *) p; - pDest = ps->map + (y * ps->pitch + x) * cpp; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w0 * cpp); - pDest += ps->pitch * cpp; - pSrc += src_stride; - } -} +#include "pipe/util/p_tile.h" /* @@ -323,10 +241,10 @@ void i915_init_surface_functions(struct i915_context *i915) { i915->pipe.get_tex_surface = i915_get_tex_surface; - i915->pipe.get_tile = i915_get_tile; - i915->pipe.put_tile = i915_put_tile; - i915->pipe.get_tile_rgba = softpipe_get_tile_rgba; - i915->pipe.put_tile_rgba = softpipe_put_tile_rgba; + i915->pipe.get_tile = pipe_get_tile_raw; + i915->pipe.put_tile = pipe_put_tile_raw; + i915->pipe.get_tile_rgba = pipe_get_tile_rgba; + i915->pipe.put_tile_rgba = pipe_put_tile_rgba; i915->pipe.surface_data = i915_surface_data; i915->pipe.surface_copy = i915_surface_copy; diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c index b0d2fec8e0..96f17e434f 100644 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -32,89 +32,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" - -#include "pipe/softpipe/sp_rgba_tile.h" /* XXX TEMPORARY */ - - -#define CLIP_TILE \ - do { \ - if (x >= ps->width) \ - return; \ - if (y >= ps->height) \ - return; \ - if (x + w > ps->width) \ - w = ps->width - x; \ - if (y + h > ps->height) \ - h = ps->height -y; \ - } while(0) - - -/* - * XXX note: same as code in sp_surface.c - */ -static void -brw_get_tile(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride) -{ - const uint cpp = ps->cpp; - const uint w0 = w; - const ubyte *pSrc; - ubyte *pDest; - uint i; - - assert(ps->map); - - CLIP_TILE; - - if (dst_stride == 0) { - dst_stride = w0 * cpp; - } - - pSrc = ps->map + ps->offset + (y * ps->pitch + x) * cpp; - pDest = (ubyte *) p; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w0 * cpp); - pDest += dst_stride; - pSrc += ps->pitch * cpp; - } -} - - -/* - * XXX note: same as code in sp_surface.c - */ -static void -brw_put_tile(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride) -{ - const uint cpp = ps->cpp; - const uint w0 = w; - const ubyte *pSrc; - ubyte *pDest; - uint i; - - assert(ps->map); - - CLIP_TILE; - - if (src_stride == 0) { - src_stride = w0 * cpp; - } - - pSrc = (const ubyte *) p; - pDest = ps->map + ps->offset + (y * ps->pitch + x) * cpp; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w0 * cpp); - pDest += ps->pitch * cpp; - pSrc += src_stride; - } -} +#include "pipe/util/p_tile.h" /* @@ -321,10 +239,10 @@ void brw_init_surface_functions(struct brw_context *brw) { brw->pipe.get_tex_surface = brw_get_tex_surface; - brw->pipe.get_tile = brw_get_tile; - brw->pipe.put_tile = brw_put_tile; - brw->pipe.get_tile_rgba = softpipe_get_tile_rgba; - brw->pipe.put_tile_rgba = softpipe_put_tile_rgba; + brw->pipe.get_tile = pipe_get_tile_raw; + brw->pipe.put_tile = pipe_put_tile_raw; + brw->pipe.get_tile_rgba = pipe_get_tile_rgba; + brw->pipe.put_tile_rgba = pipe_put_tile_rgba; brw->pipe.surface_data = brw_surface_data; brw->pipe.surface_copy = brw_surface_copy; diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 7608ff126a..31438a882e 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -25,7 +25,6 @@ DRIVER_SOURCES = \ sp_quad_output.c \ sp_quad_stencil.c \ sp_quad_stipple.c \ - sp_rgba_tile.c \ sp_state_blend.c \ sp_state_clip.c \ sp_state_derived.c \ diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 2a6909007f..5259fbfd20 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -29,24 +29,10 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" +#include "pipe/util/p_tile.h" #include "sp_context.h" #include "sp_surface.h" #include "sp_texture.h" -#include "sp_rgba_tile.h" - - - -#define CLIP_TILE \ - do { \ - if (x >= ps->width) \ - return; \ - if (y >= ps->height) \ - return; \ - if (x + w > ps->width) \ - w = ps->width - x; \ - if (y + h > ps->height) \ - h = ps->height - y; \ - } while(0) /** @@ -90,72 +76,6 @@ softpipe_get_tex_surface(struct pipe_context *pipe, } -/** - * Move raw block of pixels from surface to user memory. - */ -void -softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride) -{ - const uint cpp = ps->cpp; - const ubyte *pSrc; - const uint src_stride = ps->pitch * cpp; - ubyte *pDest; - uint i; - - assert(ps->map); - - if (dst_stride == 0) { - dst_stride = w * cpp; - } - - CLIP_TILE; - - pSrc = ps->map + (y * ps->pitch + x) * cpp; - pDest = (ubyte *) p; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w * cpp); - pDest += dst_stride; - pSrc += src_stride; - } -} - - -/** - * Move raw block of pixels from user memory to surface. - */ -void -softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride) -{ - const uint cpp = ps->cpp; - const ubyte *pSrc; - const uint dst_stride = ps->pitch * cpp; - ubyte *pDest; - uint i; - - assert(ps->map); - - if (src_stride == 0) { - src_stride = w * cpp; - } - - CLIP_TILE; - - pSrc = (const ubyte *) p; - pDest = ps->map + (y * ps->pitch + x) * cpp; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w * cpp); - pDest += dst_stride; - pSrc += src_stride; - } -} - - /** * Copy 2D rect from one place to another. * Position and sizes are in pixels. @@ -308,6 +228,10 @@ sp_surface_fill(struct pipe_context *pipe, ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff); + val0 = (val0 << 8) | val0; + val1 = (val1 << 8) | val1; + val2 = (val2 << 8) | val2; + val3 = (val3 << 8) | val3; for (i = 0; i < height; i++) { for (j = 0; j < width; j++) { row[j*4+0] = val0; @@ -331,11 +255,10 @@ sp_surface_fill(struct pipe_context *pipe, void sp_init_surface_functions(struct softpipe_context *sp) { - sp->pipe.get_tile = softpipe_get_tile; - sp->pipe.put_tile = softpipe_put_tile; - - sp->pipe.get_tile_rgba = softpipe_get_tile_rgba; - sp->pipe.put_tile_rgba = softpipe_put_tile_rgba; + sp->pipe.get_tile = pipe_get_tile_raw; + sp->pipe.put_tile = pipe_put_tile_raw; + sp->pipe.get_tile_rgba = pipe_get_tile_rgba; + sp->pipe.put_tile_rgba = pipe_put_tile_rgba; sp->pipe.surface_data = sp_surface_data; sp->pipe.surface_copy = sp_surface_copy; diff --git a/src/mesa/sources b/src/mesa/sources index f889fb30c1..5d29d20aed 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -196,6 +196,9 @@ STATECACHE_SOURCES = \ pipe/cso_cache/cso_hash.c \ pipe/cso_cache/cso_cache.c +PIPEUTIL_SOURCES = \ + pipe/util/p_tile.c + STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ state_tracker/st_atom_alphatest.c \ @@ -383,6 +386,7 @@ SOLO_SOURCES = \ $(DRAW_SOURCES) \ $(TGSIEXEC_SOURCES) \ $(TGSIUTIL_SOURCES) \ + $(PIPEUTIL_SOURCES) \ $(STATECACHE_SOURCES) \ $(STATETRACKER_SOURCES) \ $(TNL_SOURCES) \ -- cgit v1.2.3 From 47b5138d2d6533ac1cd818713fd0678ec3c7aa1b Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 15 Dec 2007 16:51:57 -0700 Subject: rename some 'mesa' functions --- src/mesa/pipe/i915simple/i915_surface.c | 34 ++++++++++++++------------------ src/mesa/pipe/i965simple/brw_surface.c | 35 +++++++++++++++------------------ 2 files changed, 31 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 34064612a7..79e74e1143 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -76,21 +76,18 @@ i915_get_tex_surface(struct pipe_context *pipe, } -/* - * XXX Move this into core Mesa? - */ static void -_mesa_copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte * src, - unsigned src_pitch, - unsigned src_x, - unsigned src_y) +copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte *src, + unsigned src_pitch, + unsigned src_x, + unsigned src_y) { unsigned i; @@ -128,10 +125,9 @@ i915_surface_data(struct pipe_context *pipe, const void *src, unsigned src_pitch, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - _mesa_copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); + copy_rect(pipe_surface_map(dst), + dst->cpp, dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); pipe_surface_unmap(dst); } @@ -151,7 +147,7 @@ i915_surface_copy(struct pipe_context *pipe, assert( dst->cpp == src->cpp ); if (0) { - _mesa_copy_rect(pipe_surface_map(dst), + copy_rect(pipe_surface_map(dst), dst->cpp, dst->pitch, dstx, dsty, diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c index 96f17e434f..d0e7229d5c 100644 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -75,21 +75,19 @@ brw_get_tex_surface(struct pipe_context *pipe, return ps; } -/* - * XXX Move this into core Mesa? - */ + static void -_mesa_copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte * src, - unsigned src_pitch, - unsigned src_x, - unsigned src_y) +copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte *src, + unsigned src_pitch, + unsigned src_x, + unsigned src_y) { unsigned i; @@ -126,10 +124,9 @@ brw_surface_data(struct pipe_context *pipe, const void *src, unsigned src_pitch, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - _mesa_copy_rect(pipe_surface_map(dst) + dst->offset, - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); + copy_rect(pipe_surface_map(dst) + dst->offset, + dst->cpp, dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); pipe_surface_unmap(dst); } @@ -149,7 +146,7 @@ brw_surface_copy(struct pipe_context *pipe, assert(dst->cpp == src->cpp); if (0) { - _mesa_copy_rect(pipe_surface_map(dst) + dst->offset, + copy_rect(pipe_surface_map(dst) + dst->offset, dst->cpp, dst->pitch, dstx, dsty, -- cgit v1.2.3 From 41b1aa51096e844c0b06f950b1891dc7e5256db7 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 16 Dec 2007 10:00:51 -0700 Subject: switch on cpp instead of format --- src/mesa/pipe/softpipe/sp_tile_cache.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index a7be3812fe..ba1f71b01e 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -236,13 +236,11 @@ clear_tile(struct softpipe_cached_tile *tile, { uint i, j; - switch (format) { - case PIPE_FORMAT_U_S8: - /* 8 bpp */ + switch (pf_get_size(format)) { + case 1: memset(tile->data.any, 0, TILE_SIZE * TILE_SIZE); break; - case PIPE_FORMAT_Z16_UNORM: - /* 16 bpp */ + case 2: if (clear_value == 0) { memset(tile->data.any, 0, 2 * TILE_SIZE * TILE_SIZE); } @@ -254,8 +252,7 @@ clear_tile(struct softpipe_cached_tile *tile, } } break; - default: - /* 32 bpp */ + case 4: if (clear_value == 0) { memset(tile->data.any, 0, 4 * TILE_SIZE * TILE_SIZE); } @@ -266,6 +263,9 @@ clear_tile(struct softpipe_cached_tile *tile, } } } + break; + default: + assert(0); } } -- cgit v1.2.3 From ae858511d00b99070f95e1c22b447d76db9b76c0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 14 Dec 2007 20:13:44 +0000 Subject: 965: put brw_constant_buffer atom directly in update list --- src/mesa/pipe/i965simple/brw_state_upload.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_state_upload.c b/src/mesa/pipe/i965simple/brw_state_upload.c index 1fb480172d..2502e54929 100644 --- a/src/mesa/pipe/i965simple/brw_state_upload.c +++ b/src/mesa/pipe/i965simple/brw_state_upload.c @@ -90,8 +90,7 @@ const struct brw_tracked_state *atoms[] = &brw_psp_urb_cbs, - - NULL, /* brw_constant_buffer */ + &brw_constant_buffer }; -- cgit v1.2.3 From b52e4477f81b8ab7e0db6a73a8f1358b88cf3ea0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 14 Dec 2007 20:14:05 +0000 Subject: 965: fix ExtDivide check --- src/mesa/pipe/i965simple/brw_wm_glsl.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_glsl.c b/src/mesa/pipe/i965simple/brw_wm_glsl.c index d6dfaed826..37d5b216b2 100644 --- a/src/mesa/pipe/i965simple/brw_wm_glsl.c +++ b/src/mesa/pipe/i965simple/brw_wm_glsl.c @@ -126,7 +126,7 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c, /* Can't handle this, don't know if we need to: */ - assert(src->SrcRegisterExtSwz.ExtDivide == 0); + assert(src->SrcRegisterExtSwz.ExtDivide == TGSI_EXTSWIZZLE_ONE); /* Not handling indirect lookups yet: */ -- cgit v1.2.3 From 95adfb9fec808fff1e72c3feb407be13a96397f9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 14 Dec 2007 20:14:30 +0000 Subject: 965: fix/hack check for NULL texunit in state upload --- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index 6e68c4c660..db8f670970 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -237,6 +237,9 @@ static void upload_wm_surfaces(struct brw_context *brw ) for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { const struct brw_texture *texUnit = brw->attribs.Texture[i]; + if (texUnit == NULL) + continue; + /* BRW_NEW_TEXTURE */ -- cgit v1.2.3 From 3be91c5a493a219dacf12b4cc279e37f26420cbd Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 14 Dec 2007 20:14:58 +0000 Subject: 965: fix for non-indexed draw path --- src/mesa/pipe/i965simple/brw_draw.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_draw.c b/src/mesa/pipe/i965simple/brw_draw.c index 01c8ddb227..498bf6b7ee 100644 --- a/src/mesa/pipe/i965simple/brw_draw.c +++ b/src/mesa/pipe/i965simple/brw_draw.c @@ -158,7 +158,8 @@ static boolean brw_try_draw_elements( struct pipe_context *pipe, /* Upload index, vertex data: */ - if (!brw_upload_indices( brw, index_buffer, index_size, start, count )) + if (index_buffer && + !brw_upload_indices( brw, index_buffer, index_size, start, count )) return FALSE; if (!brw_upload_vertex_elements( brw )) @@ -169,7 +170,9 @@ static boolean brw_try_draw_elements( struct pipe_context *pipe, if (brw->state.dirty.brw) brw_validate_state( brw ); - if (brw_emit_prim(brw, TRUE, start, count)) + if (brw_emit_prim(brw, + index_buffer != NULL, + start, count)) return FALSE; return TRUE; -- cgit v1.2.3 From 556e247cee905f84d639b4a292e891c24b36bea1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 14 Dec 2007 20:15:22 +0000 Subject: 965: remove dead state atom decls --- src/mesa/pipe/i965simple/brw_curbe.c | 8 -------- src/mesa/pipe/i965simple/brw_urb.c | 12 ------------ 2 files changed, 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index b943a7af98..66a04b9c38 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -156,14 +156,6 @@ void brw_upload_constant_buffer_state(struct brw_context *brw) BRW_CACHED_BATCH_STRUCT(brw, &cbs); } -const struct brw_tracked_state brw_constant_buffer_state = { - .dirty = { - .brw = BRW_NEW_URB_FENCE, - .cache = 0 - }, - .update = brw_upload_constant_buffer_state -}; - static float fixed_plane[6][4] = { { 0, 0, -1, 1 }, diff --git a/src/mesa/pipe/i965simple/brw_urb.c b/src/mesa/pipe/i965simple/brw_urb.c index 64255b67c4..8cc86e26a3 100644 --- a/src/mesa/pipe/i965simple/brw_urb.c +++ b/src/mesa/pipe/i965simple/brw_urb.c @@ -183,15 +183,3 @@ void brw_upload_urb_fence(struct brw_context *brw) BRW_BATCH_STRUCT(brw, &uf); } - - -#if 0 -const struct brw_tracked_state brw_urb_fence = { - .dirty = { - .mesa = 0, - .brw = BRW_NEW_URB_FENCE | BRW_NEW_PSP, - .cache = 0 - }, - .update = brw_upload_urb_fence -}; -#endif -- cgit v1.2.3 From bfe79babf99e6b9435195178d1ea64687c60d161 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 17 Dec 2007 16:14:29 +0000 Subject: gallium: incorporate alpha state into depth_stencil state object. --- src/mesa/pipe/cell/ppu/cell_context.c | 4 -- src/mesa/pipe/cell/ppu/cell_context.h | 3 +- src/mesa/pipe/cell/ppu/cell_state.h | 9 +-- src/mesa/pipe/cell/ppu/cell_state_blend.c | 40 ++--------- src/mesa/pipe/cso_cache/cso_cache.c | 13 +--- src/mesa/pipe/cso_cache/cso_cache.h | 14 ++-- src/mesa/pipe/failover/fo_context.h | 1 - src/mesa/pipe/failover/fo_state.c | 60 +++------------- src/mesa/pipe/failover/fo_state_emit.c | 8 +-- src/mesa/pipe/i915simple/i915_context.h | 4 -- src/mesa/pipe/i915simple/i915_state.c | 86 ++++++++--------------- src/mesa/pipe/i915simple/i915_state_immediate.c | 6 +- src/mesa/pipe/i965simple/brw_cc.c | 49 ++++++------- src/mesa/pipe/i965simple/brw_context.h | 3 +- src/mesa/pipe/i965simple/brw_state.c | 42 ++--------- src/mesa/pipe/i965simple/brw_wm.c | 13 ++-- src/mesa/pipe/i965simple/brw_wm_state.c | 2 +- src/mesa/pipe/p_context.h | 13 ++-- src/mesa/pipe/p_state.h | 36 +++++----- src/mesa/pipe/softpipe/sp_context.c | 10 +-- src/mesa/pipe/softpipe/sp_context.h | 9 ++- src/mesa/pipe/softpipe/sp_quad.c | 8 +-- src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 4 +- src/mesa/pipe/softpipe/sp_quad_stencil.c | 25 ++----- src/mesa/pipe/softpipe/sp_state.h | 9 +-- src/mesa/pipe/softpipe/sp_state_blend.c | 38 ++-------- src/mesa/pipe/softpipe/sp_state_derived.c | 8 +-- src/mesa/sources | 1 - src/mesa/state_tracker/st_atom.c | 3 +- src/mesa/state_tracker/st_atom.h | 3 +- src/mesa/state_tracker/st_atom_alphatest.c | 92 ------------------------- src/mesa/state_tracker/st_atom_depth.c | 55 ++++++++------- src/mesa/state_tracker/st_cache.c | 40 +++-------- src/mesa/state_tracker/st_cache.h | 9 +-- src/mesa/state_tracker/st_cb_clear.c | 35 ++++------ src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_context.h | 2 +- 37 files changed, 206 insertions(+), 553 deletions(-) delete mode 100644 src/mesa/state_tracker/st_atom_alphatest.c (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index eff33da969..b448a8aa8c 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -176,10 +176,6 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) /* state setters */ - cell->pipe.create_alpha_test_state = cell_create_alpha_test_state; - cell->pipe.bind_alpha_test_state = cell_bind_alpha_test_state; - cell->pipe.delete_alpha_test_state = cell_delete_alpha_test_state; - cell->pipe.create_blend_state = cell_create_blend_state; cell->pipe.bind_blend_state = cell_bind_blend_state; cell->pipe.delete_blend_state = cell_delete_blend_state; diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 96f000eef4..f8d6cc5d32 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -40,10 +40,9 @@ struct cell_context struct cell_winsys *winsys; - const struct pipe_alpha_test_state *alpha_test; const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct pipe_depth_stencil_state *depth_stencil; + const struct pipe_depth_stencil_alpha_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; struct pipe_blend_color blend_color; diff --git a/src/mesa/pipe/cell/ppu/cell_state.h b/src/mesa/pipe/cell/ppu/cell_state.h index 4bad45950b..b01814202d 100644 --- a/src/mesa/pipe/cell/ppu/cell_state.h +++ b/src/mesa/pipe/cell/ppu/cell_state.h @@ -27,13 +27,6 @@ cell_set_framebuffer_state( struct pipe_context *, const struct pipe_framebuffer_state * ); -extern void * -cell_create_alpha_test_state(struct pipe_context *, - const struct pipe_alpha_test_state *); -extern void -cell_bind_alpha_test_state(struct pipe_context *, void *); -extern void -cell_delete_alpha_test_state(struct pipe_context *, void *); extern void * cell_create_blend_state(struct pipe_context *, const struct pipe_blend_state *); @@ -57,7 +50,7 @@ cell_delete_sampler_state(struct pipe_context *, void *); extern void * cell_create_depth_stencil_state(struct pipe_context *, - const struct pipe_depth_stencil_state *); + const struct pipe_depth_stencil_alpha_state *); extern void cell_bind_depth_stencil_state(struct pipe_context *, void *); diff --git a/src/mesa/pipe/cell/ppu/cell_state_blend.c b/src/mesa/pipe/cell/ppu/cell_state_blend.c index e807463d90..efcb9e38a4 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_blend.c +++ b/src/mesa/pipe/cell/ppu/cell_state_blend.c @@ -69,44 +69,14 @@ void cell_set_blend_color( struct pipe_context *pipe, } -/** XXX move someday? Or consolidate all these simple state setters - * into one file. - */ - -void * -cell_create_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *alpha) -{ - struct pipe_alpha_test_state *state = MALLOC( sizeof(struct pipe_alpha_test_state) ); - memcpy(state, alpha, sizeof(struct pipe_alpha_test_state)); - return state; -} - -void -cell_bind_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - struct cell_context *cell = cell_context(pipe); - - cell->alpha_test = (const struct pipe_alpha_test_state *)alpha; - - cell->dirty |= CELL_NEW_ALPHA_TEST; -} - -void -cell_delete_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - FREE( alpha ); -} void * cell_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *depth_stencil) + const struct pipe_depth_stencil_alpha_state *depth_stencil) { - struct pipe_depth_stencil_state *state = - MALLOC( sizeof(struct pipe_depth_stencil_state) ); - memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_state)); + struct pipe_depth_stencil_alpha_state *state = + MALLOC( sizeof(struct pipe_depth_stencil_alpha_state) ); + memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_alpha_state)); return state; } @@ -116,7 +86,7 @@ cell_bind_depth_stencil_state(struct pipe_context *pipe, { struct cell_context *cell = cell_context(pipe); - cell->depth_stencil = (const struct pipe_depth_stencil_state *)depth_stencil; + cell->depth_stencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; cell->dirty |= CELL_NEW_DEPTH_STENCIL; } diff --git a/src/mesa/pipe/cso_cache/cso_cache.c b/src/mesa/pipe/cso_cache/cso_cache.c index 0bba5914dc..9e77e0774d 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.c +++ b/src/mesa/pipe/cso_cache/cso_cache.c @@ -78,7 +78,7 @@ static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_ case CSO_SAMPLER: hash = sc->sampler_hash; break; - case CSO_DEPTH_STENCIL: + case CSO_DEPTH_STENCIL_ALPHA: hash = sc->depth_stencil_hash; break; case CSO_RASTERIZER: @@ -90,9 +90,6 @@ static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_ case CSO_VERTEX_SHADER: hash = sc->vs_hash; break; - case CSO_ALPHA_TEST: - hash = sc->alpha_hash; - break; } return hash; @@ -105,16 +102,14 @@ static int _cso_size_for_type(enum cso_cache_type type) return sizeof(struct pipe_blend_state); case CSO_SAMPLER: return sizeof(struct pipe_sampler_state); - case CSO_DEPTH_STENCIL: - return sizeof(struct pipe_depth_stencil_state); + case CSO_DEPTH_STENCIL_ALPHA: + return sizeof(struct pipe_depth_stencil_alpha_state); case CSO_RASTERIZER: return sizeof(struct pipe_rasterizer_state); case CSO_FRAGMENT_SHADER: return sizeof(struct pipe_shader_state); case CSO_VERTEX_SHADER: return sizeof(struct pipe_shader_state); - case CSO_ALPHA_TEST: - return sizeof(struct pipe_alpha_test_state); } return 0; } @@ -169,7 +164,6 @@ struct cso_cache *cso_cache_create(void) sc->rasterizer_hash = cso_hash_create(); sc->fs_hash = cso_hash_create(); sc->vs_hash = cso_hash_create(); - sc->alpha_hash = cso_hash_create(); return sc; } @@ -183,6 +177,5 @@ void cso_cache_delete(struct cso_cache *sc) cso_hash_delete(sc->rasterizer_hash); cso_hash_delete(sc->fs_hash); cso_hash_delete(sc->vs_hash); - cso_hash_delete(sc->alpha_hash); free(sc); } diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h index cd36dd51e9..116e2eaa2c 100644 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ b/src/mesa/pipe/cso_cache/cso_cache.h @@ -40,7 +40,6 @@ struct cso_hash; struct cso_cache { - struct cso_hash *alpha_hash; struct cso_hash *blend_hash; struct cso_hash *depth_stencil_hash; struct cso_hash *fs_hash; @@ -54,8 +53,8 @@ struct cso_blend { void *data; }; -struct cso_depth_stencil { - struct pipe_depth_stencil_state state; +struct cso_depth_stencil_alpha { + struct pipe_depth_stencil_alpha_state state; void *data; }; @@ -79,19 +78,14 @@ struct cso_sampler { void *data; }; -struct cso_alpha_test { - struct pipe_alpha_test_state state; - void *data; -}; enum cso_cache_type { CSO_BLEND, CSO_SAMPLER, - CSO_DEPTH_STENCIL, + CSO_DEPTH_STENCIL_ALPHA, CSO_RASTERIZER, CSO_FRAGMENT_SHADER, - CSO_VERTEX_SHADER, - CSO_ALPHA_TEST + CSO_VERTEX_SHADER }; unsigned cso_construct_key(void *item, int item_size); diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h index f5eaa0b5fa..1dc87291c9 100644 --- a/src/mesa/pipe/failover/fo_context.h +++ b/src/mesa/pipe/failover/fo_context.h @@ -70,7 +70,6 @@ struct failover_context { /* The most recent drawing state as set by the driver: */ - const struct fo_state *alpha_test; const struct fo_state *blend; const struct fo_state *sampler[PIPE_MAX_SAMPLERS]; const struct fo_state *depth_stencil; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index 6b4f1517ac..fa700b9674 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -45,45 +45,6 @@ * lower overheads. */ -static void * -failover_create_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *templ) -{ - struct fo_state *state = malloc(sizeof(struct fo_state)); - struct failover_context *failover = failover_context(pipe); - - state->sw_state = failover->sw->create_alpha_test_state(pipe, templ); - state->hw_state = failover->hw->create_alpha_test_state(pipe, templ); - - return state; -} - -static void -failover_bind_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - struct failover_context *failover = failover_context(pipe); - struct fo_state *state = (struct fo_state *)alpha; - - failover->alpha_test = state; - failover->dirty |= FO_NEW_ALPHA_TEST; - failover->hw->bind_alpha_test_state(failover->hw, - state->hw_state); -} - -static void -failover_delete_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - struct fo_state *state = (struct fo_state*)alpha; - struct failover_context *failover = failover_context(pipe); - - failover->sw->delete_alpha_test_state(pipe, state->sw_state); - failover->hw->delete_alpha_test_state(pipe, state->hw_state); - state->sw_state = 0; - state->hw_state = 0; - free(state); -} static void * @@ -149,13 +110,13 @@ failover_set_clip_state( struct pipe_context *pipe, static void * failover_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *templ) + const struct pipe_depth_stencil_alpha_state *templ) { struct fo_state *state = malloc(sizeof(struct fo_state)); struct failover_context *failover = failover_context(pipe); - state->sw_state = failover->sw->create_depth_stencil_state(pipe, templ); - state->hw_state = failover->hw->create_depth_stencil_state(pipe, templ); + state->sw_state = failover->sw->create_depth_stencil_alpha_state(pipe, templ); + state->hw_state = failover->hw->create_depth_stencil_alpha_state(pipe, templ); return state; } @@ -168,7 +129,7 @@ failover_bind_depth_stencil_state(struct pipe_context *pipe, struct fo_state *state = (struct fo_state *)depth_stencil; failover->depth_stencil = state; failover->dirty |= FO_NEW_DEPTH_STENCIL; - failover->hw->bind_depth_stencil_state(failover->hw, state->hw_state); + failover->hw->bind_depth_stencil_alpha_state(failover->hw, state->hw_state); } static void @@ -178,8 +139,8 @@ failover_delete_depth_stencil_state(struct pipe_context *pipe, struct fo_state *state = (struct fo_state*)ds; struct failover_context *failover = failover_context(pipe); - failover->sw->delete_depth_stencil_state(pipe, state->sw_state); - failover->hw->delete_depth_stencil_state(pipe, state->hw_state); + failover->sw->delete_depth_stencil_alpha_state(pipe, state->sw_state); + failover->hw->delete_depth_stencil_alpha_state(pipe, state->hw_state); state->sw_state = 0; state->hw_state = 0; free(state); @@ -434,18 +395,15 @@ failover_set_vertex_element(struct pipe_context *pipe, void failover_init_state_functions( struct failover_context *failover ) { - failover->pipe.create_alpha_test_state = failover_create_alpha_test_state; - failover->pipe.bind_alpha_test_state = failover_bind_alpha_test_state; - failover->pipe.delete_alpha_test_state = failover_delete_alpha_test_state; failover->pipe.create_blend_state = failover_create_blend_state; failover->pipe.bind_blend_state = failover_bind_blend_state; failover->pipe.delete_blend_state = failover_delete_blend_state; failover->pipe.create_sampler_state = failover_create_sampler_state; failover->pipe.bind_sampler_state = failover_bind_sampler_state; failover->pipe.delete_sampler_state = failover_delete_sampler_state; - failover->pipe.create_depth_stencil_state = failover_create_depth_stencil_state; - failover->pipe.bind_depth_stencil_state = failover_bind_depth_stencil_state; - failover->pipe.delete_depth_stencil_state = failover_delete_depth_stencil_state; + failover->pipe.create_depth_stencil_alpha_state = failover_create_depth_stencil_state; + failover->pipe.bind_depth_stencil_alpha_state = failover_bind_depth_stencil_state; + failover->pipe.delete_depth_stencil_alpha_state = failover_delete_depth_stencil_state; failover->pipe.create_rasterizer_state = failover_create_rasterizer_state; failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state; failover->pipe.delete_rasterizer_state = failover_delete_rasterizer_state; diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c index c99ecd4f8d..c663dd4947 100644 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ b/src/mesa/pipe/failover/fo_state_emit.c @@ -55,10 +55,6 @@ failover_state_emit( struct failover_context *failover ) { unsigned i; - if (failover->dirty & FO_NEW_ALPHA_TEST) - failover->sw->bind_alpha_test_state( failover->sw, - failover->alpha_test->sw_state ); - if (failover->dirty & FO_NEW_BLEND) failover->sw->bind_blend_state( failover->sw, failover->blend->sw_state ); @@ -70,8 +66,8 @@ failover_state_emit( struct failover_context *failover ) failover->sw->set_clip_state( failover->sw, &failover->clip ); if (failover->dirty & FO_NEW_DEPTH_STENCIL) - failover->sw->bind_depth_stencil_state( failover->sw, - failover->depth_stencil->sw_state ); + failover->sw->bind_depth_stencil_alpha_state( failover->sw, + failover->depth_stencil->sw_state ); if (failover->dirty & FO_NEW_FRAMEBUFFER) failover->sw->set_framebuffer_state( failover->sw, &failover->framebuffer ); diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 80df7f0fba..2f1f036993 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -146,9 +146,6 @@ struct i915_sampler_state { const struct pipe_sampler_state *templ; }; -struct i915_alpha_test_state { - unsigned LIS6; -}; struct i915_texture { struct pipe_texture base; @@ -186,7 +183,6 @@ struct i915_context /* The most recent drawing state as set by the driver: */ - const struct i915_alpha_test_state *alpha_test; const struct i915_blend_state *blend; const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct i915_depth_stencil_state *depth_stencil; diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 2a9a587a37..f8332aab37 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -284,13 +284,13 @@ static void i915_delete_sampler_state(struct pipe_context *pipe, static void * i915_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *depth_stencil) + const struct pipe_depth_stencil_alpha_state *depth_stencil) { struct i915_depth_stencil_state *cso = CALLOC_STRUCT( i915_depth_stencil_state ); { - int testmask = depth_stencil->stencil.value_mask[0] & 0xff; - int writemask = depth_stencil->stencil.write_mask[0] & 0xff; + int testmask = depth_stencil->stencil[0].value_mask & 0xff; + int writemask = depth_stencil->stencil[0].write_mask & 0xff; cso->stencil_modes4 |= (_3DSTATE_MODES_4_CMD | ENABLE_STENCIL_TEST_MASK | @@ -299,12 +299,12 @@ i915_create_depth_stencil_state(struct pipe_context *pipe, STENCIL_WRITE_MASK(writemask)); } - if (depth_stencil->stencil.front_enabled) { - int test = i915_translate_compare_func(depth_stencil->stencil.front_func); - int fop = i915_translate_stencil_op(depth_stencil->stencil.front_fail_op); - int dfop = i915_translate_stencil_op(depth_stencil->stencil.front_zfail_op); - int dpop = i915_translate_stencil_op(depth_stencil->stencil.front_zpass_op); - int ref = depth_stencil->stencil.ref_value[0] & 0xff; + if (depth_stencil->stencil[0].enabled) { + int test = i915_translate_compare_func(depth_stencil->stencil[0].func); + int fop = i915_translate_stencil_op(depth_stencil->stencil[0].fail_op); + int dfop = i915_translate_stencil_op(depth_stencil->stencil[0].zfail_op); + int dpop = i915_translate_stencil_op(depth_stencil->stencil[0].zpass_op); + int ref = depth_stencil->stencil[0].ref_value & 0xff; cso->stencil_LIS5 |= (S5_STENCIL_TEST_ENABLE | S5_STENCIL_WRITE_ENABLE | @@ -315,14 +315,14 @@ i915_create_depth_stencil_state(struct pipe_context *pipe, (dpop << S5_STENCIL_PASS_Z_PASS_SHIFT)); } - if (depth_stencil->stencil.back_enabled) { - int test = i915_translate_compare_func(depth_stencil->stencil.back_func); - int fop = i915_translate_stencil_op(depth_stencil->stencil.back_fail_op); - int dfop = i915_translate_stencil_op(depth_stencil->stencil.back_zfail_op); - int dpop = i915_translate_stencil_op(depth_stencil->stencil.back_zpass_op); - int ref = depth_stencil->stencil.ref_value[1] & 0xff; - int tmask = depth_stencil->stencil.value_mask[1] & 0xff; - int wmask = depth_stencil->stencil.write_mask[1] & 0xff; + if (depth_stencil->stencil[1].enabled) { + int test = i915_translate_compare_func(depth_stencil->stencil[1].func); + int fop = i915_translate_stencil_op(depth_stencil->stencil[1].fail_op); + int dfop = i915_translate_stencil_op(depth_stencil->stencil[1].zfail_op); + int dpop = i915_translate_stencil_op(depth_stencil->stencil[1].zpass_op); + int ref = depth_stencil->stencil[1].ref_value & 0xff; + int tmask = depth_stencil->stencil[1].value_mask & 0xff; + int wmask = depth_stencil->stencil[1].write_mask & 0xff; cso->bfo[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | BFO_ENABLE_STENCIL_FUNCS | @@ -363,6 +363,15 @@ i915_create_depth_stencil_state(struct pipe_context *pipe, cso->depth_LIS6 |= S6_DEPTH_WRITE_ENABLE; } + if (depth_stencil->alpha.enabled) { + int test = i915_translate_compare_func(depth_stencil->alpha.func); + ubyte refByte = float_to_ubyte(depth_stencil->alpha.ref); + + cso->depth_LIS6 |= (S6_ALPHA_TEST_ENABLE | + (test << S6_ALPHA_TEST_FUNC_SHIFT) | + (((unsigned) refByte) << S6_ALPHA_REF_SHIFT)); + } + return cso; } @@ -383,39 +392,6 @@ static void i915_delete_depth_stencil_state(struct pipe_context *pipe, } -static void * -i915_create_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *alpha_test) -{ - struct i915_alpha_test_state *cso = CALLOC_STRUCT( i915_alpha_test_state ); - - if (alpha_test->enabled) { - int test = i915_translate_compare_func(alpha_test->func); - ubyte refByte = float_to_ubyte(alpha_test->ref); - - cso->LIS6 |= (S6_ALPHA_TEST_ENABLE | - (test << S6_ALPHA_TEST_FUNC_SHIFT) | - (((unsigned) refByte) << S6_ALPHA_REF_SHIFT)); - } - return cso; -} - -static void i915_bind_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->alpha_test = (const struct i915_alpha_test_state*)alpha; - - i915->dirty |= I915_NEW_ALPHA_TEST; -} - -static void i915_delete_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - FREE(alpha); -} - static void i915_set_scissor_state( struct pipe_context *pipe, const struct pipe_scissor_state *scissor ) { @@ -674,10 +650,6 @@ static void i915_set_vertex_element( struct pipe_context *pipe, void i915_init_state_functions( struct i915_context *i915 ) { - i915->pipe.create_alpha_test_state = i915_create_alpha_test_state; - i915->pipe.bind_alpha_test_state = i915_bind_alpha_test_state; - i915->pipe.delete_alpha_test_state = i915_delete_alpha_test_state; - i915->pipe.create_blend_state = i915_create_blend_state; i915->pipe.bind_blend_state = i915_bind_blend_state; i915->pipe.delete_blend_state = i915_delete_blend_state; @@ -686,9 +658,9 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.bind_sampler_state = i915_bind_sampler_state; i915->pipe.delete_sampler_state = i915_delete_sampler_state; - i915->pipe.create_depth_stencil_state = i915_create_depth_stencil_state; - i915->pipe.bind_depth_stencil_state = i915_bind_depth_stencil_state; - i915->pipe.delete_depth_stencil_state = i915_delete_depth_stencil_state; + i915->pipe.create_depth_stencil_alpha_state = i915_create_depth_stencil_state; + i915->pipe.bind_depth_stencil_alpha_state = i915_bind_depth_stencil_state; + i915->pipe.delete_depth_stencil_alpha_state = i915_delete_depth_stencil_state; i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index da2402c018..752d25f233 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -159,10 +159,6 @@ static void upload_S6( struct i915_context *i915 ) unsigned LIS6 = (S6_COLOR_WRITE_ENABLE | (2 << S6_TRISTRIP_PV_SHIFT)); - /* I915_NEW_ALPHA_TEST - */ - LIS6 |= i915->alpha_test->LIS6; - /* I915_NEW_BLEND */ LIS6 |= i915->blend->LIS6; @@ -178,7 +174,7 @@ static void upload_S6( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_S6 = { - I915_NEW_ALPHA_TEST | I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, + I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, upload_S6 }; diff --git a/src/mesa/pipe/i965simple/brw_cc.c b/src/mesa/pipe/i965simple/brw_cc.c index 6cc1505311..dcee731895 100644 --- a/src/mesa/pipe/i965simple/brw_cc.c +++ b/src/mesa/pipe/i965simple/brw_cc.c @@ -156,38 +156,37 @@ static void upload_cc_unit( struct brw_context *brw ) memset(&cc, 0, sizeof(cc)); /* BRW_NEW_DEPTH_STENCIL */ - if (brw->attribs.DepthStencil->stencil.front_enabled) { - cc.cc0.stencil_enable = brw->attribs.DepthStencil->stencil.front_enabled; - cc.cc0.stencil_func = brw_translate_compare_func(brw->attribs.DepthStencil->stencil.front_func); - cc.cc0.stencil_fail_op = brw_translate_stencil_op(brw->attribs.DepthStencil->stencil.front_fail_op); + if (brw->attribs.DepthStencil->stencil[0].enabled) { + cc.cc0.stencil_enable = brw->attribs.DepthStencil->stencil[0].enabled; + cc.cc0.stencil_func = brw_translate_compare_func(brw->attribs.DepthStencil->stencil[0].func); + cc.cc0.stencil_fail_op = brw_translate_stencil_op(brw->attribs.DepthStencil->stencil[0].fail_op); cc.cc0.stencil_pass_depth_fail_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil.front_zfail_op); + brw->attribs.DepthStencil->stencil[0].zfail_op); cc.cc0.stencil_pass_depth_pass_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil.front_zpass_op); - cc.cc1.stencil_ref = brw->attribs.DepthStencil->stencil.ref_value[0]; - cc.cc1.stencil_write_mask = brw->attribs.DepthStencil->stencil.write_mask[0]; - cc.cc1.stencil_test_mask = brw->attribs.DepthStencil->stencil.value_mask[0]; + brw->attribs.DepthStencil->stencil[0].zpass_op); + cc.cc1.stencil_ref = brw->attribs.DepthStencil->stencil[0].ref_value; + cc.cc1.stencil_write_mask = brw->attribs.DepthStencil->stencil[0].write_mask; + cc.cc1.stencil_test_mask = brw->attribs.DepthStencil->stencil[0].value_mask; - if (brw->attribs.DepthStencil->stencil.back_enabled) { - cc.cc0.bf_stencil_enable = brw->attribs.DepthStencil->stencil.back_enabled; + if (brw->attribs.DepthStencil->stencil[1].enabled) { + cc.cc0.bf_stencil_enable = brw->attribs.DepthStencil->stencil[1].enabled; cc.cc0.bf_stencil_func = brw_translate_compare_func( - brw->attribs.DepthStencil->stencil.back_func); + brw->attribs.DepthStencil->stencil[1].func); cc.cc0.bf_stencil_fail_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil.back_fail_op); + brw->attribs.DepthStencil->stencil[1].fail_op); cc.cc0.bf_stencil_pass_depth_fail_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil.back_zfail_op); + brw->attribs.DepthStencil->stencil[1].zfail_op); cc.cc0.bf_stencil_pass_depth_pass_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil.back_zpass_op); - cc.cc1.bf_stencil_ref = brw->attribs.DepthStencil->stencil.ref_value[1]; - cc.cc2.bf_stencil_write_mask = brw->attribs.DepthStencil->stencil.write_mask[1]; - cc.cc2.bf_stencil_test_mask = brw->attribs.DepthStencil->stencil.value_mask[1]; + brw->attribs.DepthStencil->stencil[1].zpass_op); + cc.cc1.bf_stencil_ref = brw->attribs.DepthStencil->stencil[1].ref_value; + cc.cc2.bf_stencil_write_mask = brw->attribs.DepthStencil->stencil[1].write_mask; + cc.cc2.bf_stencil_test_mask = brw->attribs.DepthStencil->stencil[1].value_mask; } /* Not really sure about this: */ - if (brw->attribs.DepthStencil->stencil.write_mask[0] || - (brw->attribs.DepthStencil->stencil.back_enabled && - brw->attribs.DepthStencil->stencil.write_mask[1])) + if (brw->attribs.DepthStencil->stencil[0].write_mask || + brw->attribs.DepthStencil->stencil[1].write_mask) cc.cc0.stencil_write_enable = 1; } @@ -228,11 +227,13 @@ static void upload_cc_unit( struct brw_context *brw ) /* BRW_NEW_ALPHATEST */ - if (brw->attribs.AlphaTest->enabled) { + if (brw->attribs.DepthStencil->alpha.enabled) { cc.cc3.alpha_test = 1; - cc.cc3.alpha_test_func = brw_translate_compare_func(brw->attribs.AlphaTest->func); + cc.cc3.alpha_test_func = + brw_translate_compare_func(brw->attribs.DepthStencil->alpha.func); - UNCLAMPED_FLOAT_TO_UBYTE(cc.cc7.alpha_ref.ub[0], brw->attribs.AlphaTest->ref); + UNCLAMPED_FLOAT_TO_UBYTE(cc.cc7.alpha_ref.ub[0], + brw->attribs.DepthStencil->alpha.ref); cc.cc3.alpha_test_format = BRW_ALPHATEST_FORMAT_UNORM8; } diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h index 318c6a7049..11146570be 100644 --- a/src/mesa/pipe/i965simple/brw_context.h +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -479,9 +479,8 @@ struct brw_context struct { - const struct pipe_alpha_test_state *AlphaTest; const struct pipe_blend_state *Blend; - const struct pipe_depth_stencil_state *DepthStencil; + const struct pipe_depth_stencil_alpha_state *DepthStencil; const struct pipe_poly_stipple *PolygonStipple; const struct pipe_rasterizer_state *Raster; const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c index 26450ae597..e7f5a27a38 100644 --- a/src/mesa/pipe/i965simple/brw_state.c +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -116,9 +116,9 @@ static void brw_delete_sampler_state(struct pipe_context *pipe, static void * brw_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *depth_stencil) + const struct pipe_depth_stencil_alpha_state *depth_stencil) { - DUP( pipe_depth_stencil_state, depth_stencil ); + DUP( pipe_depth_stencil_alpha_state, depth_stencil ); } static void brw_bind_depth_stencil_state(struct pipe_context *pipe, @@ -126,7 +126,7 @@ static void brw_bind_depth_stencil_state(struct pipe_context *pipe, { struct brw_context *brw = brw_context(pipe); - brw->attribs.DepthStencil = (const struct pipe_depth_stencil_state *)depth_stencil; + brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; } @@ -137,32 +137,6 @@ static void brw_delete_depth_stencil_state(struct pipe_context *pipe, free(depth_stencil); } -/************************************************************************ - * Alpha test - */ -static void * -brw_create_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *alpha_test) -{ - DUP(pipe_alpha_test_state, alpha_test); -} - -static void brw_bind_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.AlphaTest = (const struct pipe_alpha_test_state*)alpha; - - brw->state.dirty.brw |= BRW_NEW_ALPHA_TEST; -} - -static void brw_delete_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - free(alpha); -} - /************************************************************************ * Scissor */ @@ -415,10 +389,6 @@ static void brw_delete_rasterizer_state(struct pipe_context *pipe, void brw_init_state_functions( struct brw_context *brw ) { - brw->pipe.create_alpha_test_state = brw_create_alpha_test_state; - brw->pipe.bind_alpha_test_state = brw_bind_alpha_test_state; - brw->pipe.delete_alpha_test_state = brw_delete_alpha_test_state; - brw->pipe.create_blend_state = brw_create_blend_state; brw->pipe.bind_blend_state = brw_bind_blend_state; brw->pipe.delete_blend_state = brw_delete_blend_state; @@ -427,9 +397,9 @@ brw_init_state_functions( struct brw_context *brw ) brw->pipe.bind_sampler_state = brw_bind_sampler_state; brw->pipe.delete_sampler_state = brw_delete_sampler_state; - brw->pipe.create_depth_stencil_state = brw_create_depth_stencil_state; - brw->pipe.bind_depth_stencil_state = brw_bind_depth_stencil_state; - brw->pipe.delete_depth_stencil_state = brw_delete_depth_stencil_state; + brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; + brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; + brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; diff --git a/src/mesa/pipe/i965simple/brw_wm.c b/src/mesa/pipe/i965simple/brw_wm.c index f0a38d384b..0ee0fbed51 100644 --- a/src/mesa/pipe/i965simple/brw_wm.c +++ b/src/mesa/pipe/i965simple/brw_wm.c @@ -93,15 +93,14 @@ static void brw_wm_populate_key( struct brw_context *brw, /* Build the index for table lookup */ - /* _NEW_COLOR */ + /* BRW_NEW_DEPTH_STENCIL */ if (fp->UsesKill || - brw->attribs.AlphaTest->enabled) + brw->attribs.DepthStencil->alpha.enabled) lookup |= IZ_PS_KILL_ALPHATEST_BIT; if (fp->ComputesDepth) lookup |= IZ_PS_COMPUTES_DEPTH_BIT; - /* _NEW_DEPTH */ if (brw->attribs.DepthStencil->depth.enabled) lookup |= IZ_DEPTH_TEST_ENABLE_BIT; @@ -109,13 +108,11 @@ static void brw_wm_populate_key( struct brw_context *brw, brw->attribs.DepthStencil->depth.writemask) /* ?? */ lookup |= IZ_DEPTH_WRITE_ENABLE_BIT; - /* _NEW_STENCIL */ - if (brw->attribs.DepthStencil->stencil.front_enabled) { + if (brw->attribs.DepthStencil->stencil[0].enabled) { lookup |= IZ_STENCIL_TEST_ENABLE_BIT; - if (brw->attribs.DepthStencil->stencil.write_mask[0] || - (brw->attribs.DepthStencil->stencil.back_enabled && - brw->attribs.DepthStencil->stencil.write_mask[1])) + if (brw->attribs.DepthStencil->stencil[0].write_mask || + brw->attribs.DepthStencil->stencil[1].write_mask) lookup |= IZ_STENCIL_WRITE_ENABLE_BIT; } diff --git a/src/mesa/pipe/i965simple/brw_wm_state.c b/src/mesa/pipe/i965simple/brw_wm_state.c index 52d2c85423..5ccd488842 100644 --- a/src/mesa/pipe/i965simple/brw_wm_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_state.c @@ -122,7 +122,7 @@ static void upload_wm_unit(struct brw_context *brw ) /* BRW_NEW_ALPHA_TEST */ if (fp->UsesKill || - brw->attribs.AlphaTest->enabled) + brw->attribs.DepthStencil->alpha.enabled) wm.wm5.program_uses_killpixel = 1; wm.wm5.enable_8_pix = 1; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 92ca7dd8e3..1afb38a868 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -102,11 +102,6 @@ struct pipe_context { /* * State functions */ - void * (*create_alpha_test_state)(struct pipe_context *, - const struct pipe_alpha_test_state *); - void (*bind_alpha_test_state)(struct pipe_context *, void *); - void (*delete_alpha_test_state)(struct pipe_context *, void *); - void * (*create_blend_state)(struct pipe_context *, const struct pipe_blend_state *); void (*bind_blend_state)(struct pipe_context *, void *); @@ -122,10 +117,10 @@ struct pipe_context { void (*bind_rasterizer_state)(struct pipe_context *, void *); void (*delete_rasterizer_state)(struct pipe_context *, void *); - void * (*create_depth_stencil_state)(struct pipe_context *, - const struct pipe_depth_stencil_state *); - void (*bind_depth_stencil_state)(struct pipe_context *, void *); - void (*delete_depth_stencil_state)(struct pipe_context *, void *); + void * (*create_depth_stencil_alpha_state)(struct pipe_context *, + const struct pipe_depth_stencil_alpha_state *); + void (*bind_depth_stencil_alpha_state)(struct pipe_context *, void *); + void (*delete_depth_stencil_alpha_state)(struct pipe_context *, void *); void * (*create_fs_state)(struct pipe_context *, const struct pipe_shader_state *); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index af65d365bf..b7793c6d31 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -147,7 +147,7 @@ struct pipe_shader_state { ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; }; -struct pipe_depth_stencil_state +struct pipe_depth_stencil_alpha_state { struct { unsigned enabled:1; /**< depth test enabled? */ @@ -156,28 +156,24 @@ struct pipe_depth_stencil_state unsigned occlusion_count:1; /**< XXX move this elsewhere? */ } depth; struct { - unsigned front_enabled:1; - unsigned front_func:3; /**< PIPE_FUNC_x */ - unsigned front_fail_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned front_zpass_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned front_zfail_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned back_enabled:1; - unsigned back_func:3; /**< PIPE_FUNC_x */ - unsigned back_fail_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned back_zpass_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned back_zfail_op:3; /**< PIPE_STENCIL_OP_x */ - ubyte ref_value[2]; /**< [0] = front, [1] = back */ - ubyte value_mask[2]; - ubyte write_mask[2]; - } stencil; -}; + unsigned enabled:1; + unsigned func:3; /**< PIPE_FUNC_x */ + unsigned fail_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned zpass_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned zfail_op:3; /**< PIPE_STENCIL_OP_x */ + ubyte ref_value; + ubyte value_mask; + ubyte write_mask; + } stencil[2]; /**< [0] = front, [1] = back */ -struct pipe_alpha_test_state { - unsigned enabled:1; - unsigned func:3; /**< PIPE_FUNC_x */ - float ref; /**< reference value */ + struct { + unsigned enabled:1; + unsigned func:3; /**< PIPE_FUNC_x */ + float ref; /**< reference value */ + } alpha; }; + struct pipe_blend_state { unsigned blend_enable:1; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index b62e691e87..b6995b8a6c 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -240,10 +240,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.get_paramf = softpipe_get_paramf; /* state setters */ - softpipe->pipe.create_alpha_test_state = softpipe_create_alpha_test_state; - softpipe->pipe.bind_alpha_test_state = softpipe_bind_alpha_test_state; - softpipe->pipe.delete_alpha_test_state = softpipe_delete_alpha_test_state; - softpipe->pipe.create_blend_state = softpipe_create_blend_state; softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; @@ -252,9 +248,9 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.bind_sampler_state = softpipe_bind_sampler_state; softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; - softpipe->pipe.create_depth_stencil_state = softpipe_create_depth_stencil_state; - softpipe->pipe.bind_depth_stencil_state = softpipe_bind_depth_stencil_state; - softpipe->pipe.delete_depth_stencil_state = softpipe_delete_depth_stencil_state; + softpipe->pipe.create_depth_stencil_alpha_state = softpipe_create_depth_stencil_state; + softpipe->pipe.bind_depth_stencil_alpha_state = softpipe_bind_depth_stencil_state; + softpipe->pipe.delete_depth_stencil_alpha_state = softpipe_delete_depth_stencil_state; softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 8fd44933f2..8f14dd11d1 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -53,13 +53,13 @@ struct softpipe_tile_cache; #define SP_NEW_SCISSOR 0x20 #define SP_NEW_STIPPLE 0x40 #define SP_NEW_FRAMEBUFFER 0x80 -#define SP_NEW_ALPHA_TEST 0x100 -#define SP_NEW_DEPTH_STENCIL 0x200 +#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100 +#define SP_NEW_CONSTANTS 0x200 #define SP_NEW_SAMPLER 0x400 #define SP_NEW_TEXTURE 0x800 #define SP_NEW_VERTEX 0x1000 #define SP_NEW_VS 0x2000 -#define SP_NEW_CONSTANTS 0x4000 +#define SP_NEW_QUERY 0x4000 struct sp_vertex_shader_state { struct pipe_shader_state *state; @@ -73,10 +73,9 @@ struct softpipe_context { /* The most recent drawing state as set by the driver: */ - const struct pipe_alpha_test_state *alpha_test; const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct pipe_depth_stencil_state *depth_stencil; + const struct pipe_depth_stencil_alpha_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct sp_fragment_shader_state *fs; const struct sp_vertex_shader_state *vs; diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index 6330465a8b..a10c9c3e02 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -43,8 +43,8 @@ static void sp_build_depth_stencil( struct softpipe_context *sp ) { - if (sp->depth_stencil->stencil.front_enabled || - sp->depth_stencil->stencil.back_enabled) { + if (sp->depth_stencil->stencil[0].enabled || + sp->depth_stencil->stencil[1].enabled) { sp_push_quad_first( sp, sp->quad.stencil_test ); } else if (sp->depth_stencil->depth.enabled && @@ -59,7 +59,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) boolean early_depth_test = sp->depth_stencil->depth.enabled && sp->framebuffer.zbuf && - !sp->alpha_test->enabled && + !sp->depth_stencil->alpha.enabled && sp->fs->shader.output_semantic_name[0] != TGSI_SEMANTIC_POSITION; /* build up the pipeline in reverse order... */ @@ -98,7 +98,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp_build_depth_stencil( sp ); } - if (sp->alpha_test->enabled) { + if (sp->depth_stencil->alpha.enabled) { sp_push_quad_first( sp, sp->quad.alpha_test ); } diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c index d056abe98d..4ffeac35e1 100644 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c @@ -14,11 +14,11 @@ static void alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - const float ref = softpipe->alpha_test->ref; + const float ref = softpipe->depth_stencil->alpha.ref; unsigned passMask = 0x0, j; const float *aaaa = quad->outputs.color[3]; - switch (softpipe->alpha_test->func) { + switch (softpipe->depth_stencil->alpha.func) { case PIPE_FUNC_NEVER: quad->mask = 0x0; break; diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 3f3eca078b..a688a06c74 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -211,24 +211,13 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) /* choose front or back face function, operator, etc */ /* XXX we could do these initializations once per primitive */ - if (softpipe->depth_stencil->stencil.back_enabled && quad->facing) { - func = softpipe->depth_stencil->stencil.back_func; - failOp = softpipe->depth_stencil->stencil.back_fail_op; - zFailOp = softpipe->depth_stencil->stencil.back_zfail_op; - zPassOp = softpipe->depth_stencil->stencil.back_zpass_op; - ref = softpipe->depth_stencil->stencil.ref_value[1]; - wrtMask = softpipe->depth_stencil->stencil.write_mask[1]; - valMask = softpipe->depth_stencil->stencil.value_mask[1]; - } - else { - func = softpipe->depth_stencil->stencil.front_func; - failOp = softpipe->depth_stencil->stencil.front_fail_op; - zFailOp = softpipe->depth_stencil->stencil.front_zfail_op; - zPassOp = softpipe->depth_stencil->stencil.front_zpass_op; - ref = softpipe->depth_stencil->stencil.ref_value[0]; - wrtMask = softpipe->depth_stencil->stencil.write_mask[0]; - valMask = softpipe->depth_stencil->stencil.value_mask[0]; - } + func = softpipe->depth_stencil->stencil[quad->facing].func; + failOp = softpipe->depth_stencil->stencil[quad->facing].fail_op; + zFailOp = softpipe->depth_stencil->stencil[quad->facing].zfail_op; + zPassOp = softpipe->depth_stencil->stencil[quad->facing].zpass_op; + ref = softpipe->depth_stencil->stencil[quad->facing].ref_value; + wrtMask = softpipe->depth_stencil->stencil[quad->facing].write_mask; + valMask = softpipe->depth_stencil->stencil[quad->facing].value_mask; assert(ps); /* shouldn't get here if there's no stencil buffer */ diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index a3bd078a71..76b79b5280 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -52,13 +52,6 @@ struct sp_fragment_shader_state { #endif }; -void * -softpipe_create_alpha_test_state(struct pipe_context *, - const struct pipe_alpha_test_state *); -void -softpipe_bind_alpha_test_state(struct pipe_context *, void *); -void -softpipe_delete_alpha_test_state(struct pipe_context *, void *); void * softpipe_create_blend_state(struct pipe_context *, @@ -76,7 +69,7 @@ void softpipe_delete_sampler_state(struct pipe_context *, void *); void * softpipe_create_depth_stencil_state(struct pipe_context *, - const struct pipe_depth_stencil_state *); + const struct pipe_depth_stencil_alpha_state *); void softpipe_bind_depth_stencil_state(struct pipe_context *, void *); void softpipe_delete_depth_stencil_state(struct pipe_context *, void *); diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 5ceec2513f..160ca5cbc0 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -73,40 +73,14 @@ void softpipe_set_blend_color( struct pipe_context *pipe, * into one file. */ -void * -softpipe_create_alpha_test_state(struct pipe_context *pipe, - const struct pipe_alpha_test_state *alpha) -{ - struct pipe_alpha_test_state *state = MALLOC( sizeof(struct pipe_alpha_test_state) ); - memcpy(state, alpha, sizeof(struct pipe_alpha_test_state)); - return state; -} - -void -softpipe_bind_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->alpha_test = (const struct pipe_alpha_test_state *)alpha; - - softpipe->dirty |= SP_NEW_ALPHA_TEST; -} - -void -softpipe_delete_alpha_test_state(struct pipe_context *pipe, - void *alpha) -{ - FREE( alpha ); -} void * softpipe_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_state *depth_stencil) + const struct pipe_depth_stencil_alpha_state *depth_stencil) { - struct pipe_depth_stencil_state *state = - MALLOC( sizeof(struct pipe_depth_stencil_state) ); - memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_state)); + struct pipe_depth_stencil_alpha_state *state = + MALLOC( sizeof(struct pipe_depth_stencil_alpha_state) ); + memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_alpha_state)); return state; } @@ -116,9 +90,9 @@ softpipe_bind_depth_stencil_state(struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->depth_stencil = (const struct pipe_depth_stencil_state *)depth_stencil; + softpipe->depth_stencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; - softpipe->dirty |= SP_NEW_DEPTH_STENCIL; + softpipe->dirty |= SP_NEW_DEPTH_STENCIL_ALPHA; } void diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 736ac1c33b..94072a2d30 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -201,16 +201,16 @@ void softpipe_update_derived( struct softpipe_context *softpipe ) calculate_vertex_layout( softpipe ); if (softpipe->dirty & (SP_NEW_SCISSOR | - SP_NEW_DEPTH_STENCIL | + SP_NEW_DEPTH_STENCIL_ALPHA | SP_NEW_FRAMEBUFFER)) compute_cliprect(softpipe); if (softpipe->dirty & (SP_NEW_BLEND | - SP_NEW_DEPTH_STENCIL | - SP_NEW_ALPHA_TEST | + SP_NEW_DEPTH_STENCIL_ALPHA | SP_NEW_FRAMEBUFFER | SP_NEW_RASTERIZER | - SP_NEW_FS)) + SP_NEW_FS | + SP_NEW_QUERY)) sp_build_quad_pipeline(softpipe); softpipe->dirty = 0; diff --git a/src/mesa/sources b/src/mesa/sources index 5d29d20aed..56ea6dbce2 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -201,7 +201,6 @@ PIPEUTIL_SOURCES = \ STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ - state_tracker/st_atom_alphatest.c \ state_tracker/st_atom_blend.c \ state_tracker/st_atom_clip.c \ state_tracker/st_atom_constbuf.c \ diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index bde81edd8c..0e22a2fa6e 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -45,7 +45,7 @@ static const struct st_tracked_state *atoms[] = { &st_update_framebuffer, - &st_update_depth_stencil, + &st_update_depth_stencil_alpha, &st_update_clip, &st_update_shader, @@ -59,7 +59,6 @@ static const struct st_tracked_state *atoms[] = &st_update_texture, &st_update_vs_constants, &st_update_fs_constants, - &st_update_alpha_test, &st_update_pixel_transfer }; diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 0114f42ba5..3a63e2dec0 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -46,7 +46,7 @@ void st_validate_state( struct st_context *st ); const struct st_tracked_state st_update_framebuffer; const struct st_tracked_state st_update_clip; -const struct st_tracked_state st_update_depth_stencil; +const struct st_tracked_state st_update_depth_stencil_alpha; const struct st_tracked_state st_update_shader; const struct st_tracked_state st_update_rasterizer; const struct st_tracked_state st_update_polygon_stipple; @@ -57,7 +57,6 @@ const struct st_tracked_state st_update_sampler; const struct st_tracked_state st_update_texture; const struct st_tracked_state st_update_fs_constants; const struct st_tracked_state st_update_vs_constants; -const struct st_tracked_state st_update_alpha_test; const struct st_tracked_state st_update_pixel_transfer; diff --git a/src/mesa/state_tracker/st_atom_alphatest.c b/src/mesa/state_tracker/st_atom_alphatest.c deleted file mode 100644 index 873520ab02..0000000000 --- a/src/mesa/state_tracker/st_atom_alphatest.c +++ /dev/null @@ -1,92 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#include "st_context.h" -#include "st_cache.h" -#include "st_atom.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" - - -/** - * Convert GLenum stencil func tokens to pipe tokens. - */ -static GLuint -gl_alpha_func_to_sp(GLenum func) -{ - /* Same values, just biased */ - assert(PIPE_FUNC_NEVER == GL_NEVER - GL_NEVER); - assert(PIPE_FUNC_LESS == GL_LESS - GL_NEVER); - assert(PIPE_FUNC_EQUAL == GL_EQUAL - GL_NEVER); - assert(PIPE_FUNC_LEQUAL == GL_LEQUAL - GL_NEVER); - assert(PIPE_FUNC_GREATER == GL_GREATER - GL_NEVER); - assert(PIPE_FUNC_NOTEQUAL == GL_NOTEQUAL - GL_NEVER); - assert(PIPE_FUNC_GEQUAL == GL_GEQUAL - GL_NEVER); - assert(PIPE_FUNC_ALWAYS == GL_ALWAYS - GL_NEVER); - assert(func >= GL_NEVER); - assert(func <= GL_ALWAYS); - return func - GL_NEVER; -} - - -static void -update_alpha_test( struct st_context *st ) -{ - struct pipe_alpha_test_state alpha; - const struct cso_alpha_test *cso; - - memset(&alpha, 0, sizeof(alpha)); - - if (st->ctx->Color.AlphaEnabled) { - alpha.enabled = 1; - alpha.func = gl_alpha_func_to_sp(st->ctx->Color.AlphaFunc); - alpha.ref = st->ctx->Color.AlphaRef; - } - cso = st_cached_alpha_test_state(st, &alpha); - if (st->state.alpha_test != cso) { - /* state has changed */ - st->state.alpha_test = cso; - st->pipe->bind_alpha_test_state(st->pipe, cso->data); /* bind new state */ - } -} - - -const struct st_tracked_state st_update_alpha_test = { - .name = "st_update_alpha_test", - .dirty = { - .mesa = (_NEW_COLOR), - .st = 0, - }, - .update = update_alpha_test -}; diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index e785434cec..7aecdbfbcc 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -91,10 +91,10 @@ gl_stencil_op_to_pipe(GLenum func) } static void -update_depth_stencil(struct st_context *st) +update_depth_stencil_alpha(struct st_context *st) { - struct pipe_depth_stencil_state depth_stencil; - const struct cso_depth_stencil *cso; + struct pipe_depth_stencil_alpha_state depth_stencil; + const struct cso_depth_stencil_alpha *cso; memset(&depth_stencil, 0, sizeof(depth_stencil)); @@ -107,40 +107,47 @@ update_depth_stencil(struct st_context *st) depth_stencil.depth.occlusion_count = 1; if (st->ctx->Stencil.Enabled) { - depth_stencil.stencil.front_enabled = 1; - depth_stencil.stencil.front_func = st_compare_func_to_pipe(st->ctx->Stencil.Function[0]); - depth_stencil.stencil.front_fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[0]); - depth_stencil.stencil.front_zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[0]); - depth_stencil.stencil.front_zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[0]); - depth_stencil.stencil.ref_value[0] = st->ctx->Stencil.Ref[0] & 0xff; - depth_stencil.stencil.value_mask[0] = st->ctx->Stencil.ValueMask[0] & 0xff; - depth_stencil.stencil.write_mask[0] = st->ctx->Stencil.WriteMask[0] & 0xff; + depth_stencil.stencil[0].enabled = 1; + depth_stencil.stencil[0].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[0]); + depth_stencil.stencil[0].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[0]); + depth_stencil.stencil[0].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[0]); + depth_stencil.stencil[0].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[0]); + depth_stencil.stencil[0].ref_value = st->ctx->Stencil.Ref[0] & 0xff; + depth_stencil.stencil[0].value_mask = st->ctx->Stencil.ValueMask[0] & 0xff; + depth_stencil.stencil[0].write_mask = st->ctx->Stencil.WriteMask[0] & 0xff; + if (st->ctx->Stencil.TestTwoSide) { - depth_stencil.stencil.back_enabled = 1; - depth_stencil.stencil.back_func = st_compare_func_to_pipe(st->ctx->Stencil.Function[1]); - depth_stencil.stencil.back_fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[1]); - depth_stencil.stencil.back_zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[1]); - depth_stencil.stencil.back_zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[1]); - depth_stencil.stencil.ref_value[1] = st->ctx->Stencil.Ref[1] & 0xff; - depth_stencil.stencil.value_mask[1] = st->ctx->Stencil.ValueMask[1] & 0xff; - depth_stencil.stencil.write_mask[1] = st->ctx->Stencil.WriteMask[1] & 0xff; + depth_stencil.stencil[1].enabled = 1; + depth_stencil.stencil[1].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[1]); + depth_stencil.stencil[1].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[1]); + depth_stencil.stencil[1].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[1]); + depth_stencil.stencil[1].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[1]); + depth_stencil.stencil[1].ref_value = st->ctx->Stencil.Ref[1] & 0xff; + depth_stencil.stencil[1].value_mask = st->ctx->Stencil.ValueMask[1] & 0xff; + depth_stencil.stencil[1].write_mask = st->ctx->Stencil.WriteMask[1] & 0xff; } } - cso = st_cached_depth_stencil_state(st, &depth_stencil); + if (st->ctx->Color.AlphaEnabled) { + depth_stencil.alpha.enabled = 1; + depth_stencil.alpha.func = st_compare_func_to_pipe(st->ctx->Color.AlphaFunc); + depth_stencil.alpha.ref = st->ctx->Color.AlphaRef; + } + + cso = st_cached_depth_stencil_alpha_state(st, &depth_stencil); if (st->state.depth_stencil != cso) { /* state has changed */ st->state.depth_stencil = cso; - st->pipe->bind_depth_stencil_state(st->pipe, cso->data); /* bind new state */ + st->pipe->bind_depth_stencil_alpha_state(st->pipe, cso->data); /* bind new state */ } } -const struct st_tracked_state st_update_depth_stencil = { +const struct st_tracked_state st_update_depth_stencil_alpha = { .name = "st_update_depth_stencil", .dirty = { - .mesa = (_NEW_DEPTH|_NEW_STENCIL), + .mesa = (_NEW_DEPTH|_NEW_STENCIL|_NEW_COLOR), .st = 0, }, - .update = update_depth_stencil + .update = update_depth_stencil_alpha }; diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index c0f712ba1d..e0965b217a 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -87,24 +87,25 @@ st_cached_sampler_state(struct st_context *st, return (struct cso_sampler*)(cso_hash_iter_data(iter)); } -const struct cso_depth_stencil * -st_cached_depth_stencil_state(struct st_context *st, - const struct pipe_depth_stencil_state *templ) +const struct cso_depth_stencil_alpha * +st_cached_depth_stencil_alpha_state(struct st_context *st, + const struct pipe_depth_stencil_alpha_state *templ) { unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_depth_stencil_state)); + sizeof(struct pipe_depth_stencil_alpha_state)); struct cso_hash_iter iter = cso_find_state_template(st->cache, - hash_key, CSO_DEPTH_STENCIL, + hash_key, + CSO_DEPTH_STENCIL_ALPHA, (void*)templ); if (cso_hash_iter_is_null(iter)) { - struct cso_depth_stencil *cso = malloc(sizeof(struct cso_depth_stencil)); - memcpy(&cso->state, templ, sizeof(struct pipe_depth_stencil_state)); - cso->data = st->pipe->create_depth_stencil_state(st->pipe, &cso->state); + struct cso_depth_stencil_alpha *cso = malloc(sizeof(struct cso_depth_stencil_alpha)); + memcpy(&cso->state, templ, sizeof(struct pipe_depth_stencil_alpha_state)); + cso->data = st->pipe->create_depth_stencil_alpha_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; - iter = cso_insert_state(st->cache, hash_key, CSO_DEPTH_STENCIL, cso); + iter = cso_insert_state(st->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso); } - return (struct cso_depth_stencil*)(cso_hash_iter_data(iter)); + return (struct cso_depth_stencil_alpha*)(cso_hash_iter_data(iter)); } const struct cso_rasterizer* st_cached_rasterizer_state( @@ -167,22 +168,3 @@ st_cached_vs_state(struct st_context *st, return (struct cso_vertex_shader*)(cso_hash_iter_data(iter)); } -const struct cso_alpha_test * -st_cached_alpha_test_state(struct st_context *st, - const struct pipe_alpha_test_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_alpha_test_state)); - struct cso_hash_iter iter = cso_find_state_template(st->cache, - hash_key, CSO_ALPHA_TEST, - (void*)templ); - if (cso_hash_iter_is_null(iter)) { - struct cso_alpha_test *cso = malloc(sizeof(struct cso_alpha_test)); - memcpy(&cso->state, templ, sizeof(struct pipe_alpha_test_state)); - cso->data = st->pipe->create_alpha_test_state(st->pipe, &cso->state); - if (!cso->data) - cso->data = &cso->state; - iter = cso_insert_state(st->cache, hash_key, CSO_ALPHA_TEST, cso); - } - return ((struct cso_alpha_test *)cso_hash_iter_data(iter)); -} diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index 422f668c56..e0c176b0ff 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -39,9 +39,6 @@ struct pipe_blend_state; struct pipe_sampler_state; struct st_context; -const struct cso_alpha_test * -st_cached_alpha_test_state(struct st_context *st, - const struct pipe_alpha_test_state *alpha); const struct cso_blend * st_cached_blend_state(struct st_context *st, @@ -51,9 +48,9 @@ const struct cso_sampler * st_cached_sampler_state(struct st_context *st, const struct pipe_sampler_state *sampler); -const struct cso_depth_stencil * -st_cached_depth_stencil_state(struct st_context *st, - const struct pipe_depth_stencil_state *depth_stencil); +const struct cso_depth_stencil_alpha * +st_cached_depth_stencil_alpha_state(struct st_context *st, + const struct pipe_depth_stencil_alpha_state *depth_stencil); const struct cso_rasterizer * st_cached_rasterizer_state(struct st_context *st, diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index b4b2429a2a..40319f4b4b 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -272,14 +272,6 @@ clear_with_quad(GLcontext *ctx, const GLfloat x1 = ctx->DrawBuffer->_Xmax; const GLfloat y1 = ctx->DrawBuffer->_Ymax; - /* alpha state: disabled */ - { - struct pipe_alpha_test_state alpha_test; - const struct cso_alpha_test *cso; - memset(&alpha_test, 0, sizeof(alpha_test)); - cso = st_cached_alpha_test_state(st, &alpha_test); - pipe->bind_alpha_test_state(pipe, cso->data); - } /* blend state: RGBA masking */ { @@ -304,8 +296,8 @@ clear_with_quad(GLcontext *ctx, /* depth_stencil state: always pass/set to ref value */ { - struct pipe_depth_stencil_state depth_stencil; - const struct cso_depth_stencil *cso; + struct pipe_depth_stencil_alpha_state depth_stencil; + const struct cso_depth_stencil_alpha *cso; memset(&depth_stencil, 0, sizeof(depth_stencil)); if (depth) { depth_stencil.depth.enabled = 1; @@ -314,17 +306,17 @@ clear_with_quad(GLcontext *ctx, } if (stencil) { - depth_stencil.stencil.front_enabled = 1; - depth_stencil.stencil.front_func = PIPE_FUNC_ALWAYS; - depth_stencil.stencil.front_fail_op = PIPE_STENCIL_OP_REPLACE; - depth_stencil.stencil.front_zpass_op = PIPE_STENCIL_OP_REPLACE; - depth_stencil.stencil.front_zfail_op = PIPE_STENCIL_OP_REPLACE; - depth_stencil.stencil.ref_value[0] = ctx->Stencil.Clear; - depth_stencil.stencil.value_mask[0] = 0xff; - depth_stencil.stencil.write_mask[0] = ctx->Stencil.WriteMask[0] & 0xff; + depth_stencil.stencil[0].enabled = 1; + depth_stencil.stencil[0].func = PIPE_FUNC_ALWAYS; + depth_stencil.stencil[0].fail_op = PIPE_STENCIL_OP_REPLACE; + depth_stencil.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE; + depth_stencil.stencil[0].zfail_op = PIPE_STENCIL_OP_REPLACE; + depth_stencil.stencil[0].ref_value = ctx->Stencil.Clear; + depth_stencil.stencil[0].value_mask = 0xff; + depth_stencil.stencil[0].write_mask = ctx->Stencil.WriteMask[0] & 0xff; } - cso = st_cached_depth_stencil_state(st, &depth_stencil); - pipe->bind_depth_stencil_state(pipe, cso->data); + cso = st_cached_depth_stencil_alpha_state(st, &depth_stencil); + pipe->bind_depth_stencil_alpha_state(pipe, cso->data); } /* rasterizer state: nothing */ @@ -381,9 +373,8 @@ clear_with_quad(GLcontext *ctx, draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); /* Restore pipe state */ - pipe->bind_alpha_test_state(pipe, st->state.alpha_test->data); pipe->bind_blend_state(pipe, st->state.blend->data); - pipe->bind_depth_stencil_state(pipe, st->state.depth_stencil->data); + pipe->bind_depth_stencil_alpha_state(pipe, st->state.depth_stencil->data); pipe->bind_fs_state(pipe, st->state.fs->data); pipe->bind_vs_state(pipe, st->state.vs->data); pipe->bind_rasterizer_state(pipe, st->state.rasterizer->data); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 0bc48b7039..e70a5b49e1 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -789,7 +789,7 @@ compatible_formats(GLenum format, GLenum type, enum pipe_format pipeFormat) static GLboolean any_fragment_ops(const struct st_context *st) { - if (st->state.alpha_test->state.enabled || + if (st->state.depth_stencil->state.alpha.enabled || st->state.blend->state.blend_enable || st->state.blend->state.logicop_enable || st->state.depth_stencil->state.depth.enabled) diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 87646b3c71..c3919d474c 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -97,7 +97,7 @@ struct st_context const struct cso_alpha_test *alpha_test; const struct cso_blend *blend; const struct cso_sampler *sampler[PIPE_MAX_SAMPLERS]; - const struct cso_depth_stencil *depth_stencil; + const struct cso_depth_stencil_alpha *depth_stencil; const struct cso_rasterizer *rasterizer; const struct cso_fragment_shader *fs; const struct cso_vertex_shader *vs; -- cgit v1.2.3 From 531efbab75ea7d05e1af4640814f19cb5594374e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 17 Dec 2007 14:05:46 -0500 Subject: i965: return false when the function fails, not succeeds --- src/mesa/pipe/i965simple/brw_draw.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_draw.c b/src/mesa/pipe/i965simple/brw_draw.c index 498bf6b7ee..25861a4373 100644 --- a/src/mesa/pipe/i965simple/brw_draw.c +++ b/src/mesa/pipe/i965simple/brw_draw.c @@ -170,9 +170,8 @@ static boolean brw_try_draw_elements( struct pipe_context *pipe, if (brw->state.dirty.brw) brw_validate_state( brw ); - if (brw_emit_prim(brw, - index_buffer != NULL, - start, count)) + if (!brw_emit_prim(brw, index_buffer != NULL, + start, count)) return FALSE; return TRUE; -- cgit v1.2.3 From aeb53622814d340323c766dd33e1dd9a25f33e7a Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 17 Dec 2007 14:36:31 -0500 Subject: i965: index the destination offset with regards to the current index --- src/mesa/pipe/i965simple/brw_state.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c index e7f5a27a38..2008853654 100644 --- a/src/mesa/pipe/i965simple/brw_state.c +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -289,7 +289,7 @@ static void brw_set_vertex_element(struct pipe_context *pipe, el.vep.ve0.valid = 1; el.vep.ve0.vertex_buffer_index = element->vertex_buffer_index; - el.vep.ve1.dst_offset = element->dst_offset; + el.vep.ve1.dst_offset = index * 4; el.vep.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; el.vep.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; el.vep.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; -- cgit v1.2.3 From 8978627ad0e236796758d4912822b3428a649987 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Dec 2007 13:59:55 -0700 Subject: Emit declarations for constants (references to the const buffer) --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 58 +++++++++++++++++++++++++++++++- 1 file changed, 57 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index bccabd8004..c8b76978f1 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -597,6 +597,19 @@ make_sampler_decl(GLuint index) return decl; } +/** Reference into a constant buffer */ +static struct tgsi_full_declaration +make_constant_decl(GLuint first, GLuint last) +{ + struct tgsi_full_declaration decl; + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_CONSTANT; + decl.Declaration.Declare = TGSI_DECLARE_RANGE; + decl.u.DeclarationRange.First = first; + decl.u.DeclarationRange.Last = last; + return decl; +} + /** @@ -788,7 +801,7 @@ tgsi_translate_mesa_program( /* immediates/literals */ for (i = 0; program->Parameters && i < program->Parameters->NumParameters; - i++) { + i++) { if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { struct tgsi_full_immediate fullimm = make_immediate(program->Parameters->ParameterValues[i], @@ -802,6 +815,49 @@ tgsi_translate_mesa_program( } } + /* constant buffer refs */ + { + GLint start = -1, end = -1; + + for (i = 0; + program->Parameters && i < program->Parameters->NumParameters; + i++) { + GLboolean emit = (i == program->Parameters->NumParameters - 1); + + switch (program->Parameters->Parameters[i].Type) { + case PROGRAM_ENV_PARAM: + case PROGRAM_STATE_VAR: + case PROGRAM_NAMED_PARAM: + case PROGRAM_UNIFORM: + if (start == -1) { + /* begin a sequence */ + start = i; + end = i; + } + else { + /* continue sequence */ + end = i; + } + break; + default: + if (start != -1) { + /* end of sequence */ + emit = GL_TRUE; + } + } + + if (emit && start >= 0) { + struct tgsi_full_declaration fulldecl; + fulldecl = make_constant_decl( start, end ); + ti += tgsi_build_full_declaration(&fulldecl, + &tokens[ti], + header, + maxTokens - ti); + start = end = -1; + } + } + } + /* texture samplers */ for (i = 0; i < 8; i++) { if (program->SamplersUsed & (1 << i)) { -- cgit v1.2.3 From 04a5a8b3bcba64c506ee5646a3a8b737e124a66a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Dec 2007 14:05:53 -0700 Subject: updates for depth/stencil/alpha state --- src/mesa/pipe/cell/ppu/cell_context.c | 6 +++--- src/mesa/pipe/cell/ppu/cell_state.h | 6 +++--- src/mesa/pipe/cell/ppu/cell_state_blend.c | 10 +++++----- 3 files changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index b448a8aa8c..281fc6ec31 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -184,9 +184,9 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->pipe.bind_sampler_state = cell_bind_sampler_state; cell->pipe.delete_sampler_state = cell_delete_sampler_state; - cell->pipe.create_depth_stencil_state = cell_create_depth_stencil_state; - cell->pipe.bind_depth_stencil_state = cell_bind_depth_stencil_state; - cell->pipe.delete_depth_stencil_state = cell_delete_depth_stencil_state; + cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state; + cell->pipe.bind_depth_stencil_alpha_state = cell_bind_depth_stencil_alpha_state; + cell->pipe.delete_depth_stencil_alpha_state = cell_delete_depth_stencil_alpha_state; cell->pipe.create_rasterizer_state = cell_create_rasterizer_state; cell->pipe.bind_rasterizer_state = cell_bind_rasterizer_state; diff --git a/src/mesa/pipe/cell/ppu/cell_state.h b/src/mesa/pipe/cell/ppu/cell_state.h index b01814202d..63dbac2f02 100644 --- a/src/mesa/pipe/cell/ppu/cell_state.h +++ b/src/mesa/pipe/cell/ppu/cell_state.h @@ -49,14 +49,14 @@ cell_delete_sampler_state(struct pipe_context *, void *); extern void * -cell_create_depth_stencil_state(struct pipe_context *, +cell_create_depth_stencil_alpha_state(struct pipe_context *, const struct pipe_depth_stencil_alpha_state *); extern void -cell_bind_depth_stencil_state(struct pipe_context *, void *); +cell_bind_depth_stencil_alpha_state(struct pipe_context *, void *); extern void -cell_delete_depth_stencil_state(struct pipe_context *, void *); +cell_delete_depth_stencil_alpha_state(struct pipe_context *, void *); void *cell_create_fs_state(struct pipe_context *, diff --git a/src/mesa/pipe/cell/ppu/cell_state_blend.c b/src/mesa/pipe/cell/ppu/cell_state_blend.c index efcb9e38a4..da3fcfd3a5 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_blend.c +++ b/src/mesa/pipe/cell/ppu/cell_state_blend.c @@ -71,8 +71,8 @@ void cell_set_blend_color( struct pipe_context *pipe, void * -cell_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) +cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil) { struct pipe_depth_stencil_alpha_state *state = MALLOC( sizeof(struct pipe_depth_stencil_alpha_state) ); @@ -81,8 +81,8 @@ cell_create_depth_stencil_state(struct pipe_context *pipe, } void -cell_bind_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) +cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, + void *depth_stencil) { struct cell_context *cell = cell_context(pipe); @@ -92,7 +92,7 @@ cell_bind_depth_stencil_state(struct pipe_context *pipe, } void -cell_delete_depth_stencil_state(struct pipe_context *pipe, void *depth) +cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth) { FREE( depth ); } -- cgit v1.2.3 From e2a669aed4bc6f9c94b6b664bf667777078e02c0 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Dec 2007 20:41:20 -0700 Subject: obsolete --- src/mesa/pipe/softpipe/sp_rgba_tile.c | 758 ---------------------------------- src/mesa/pipe/softpipe/sp_rgba_tile.h | 48 --- 2 files changed, 806 deletions(-) delete mode 100644 src/mesa/pipe/softpipe/sp_rgba_tile.c delete mode 100644 src/mesa/pipe/softpipe/sp_rgba_tile.h (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_rgba_tile.c b/src/mesa/pipe/softpipe/sp_rgba_tile.c deleted file mode 100644 index c42635a1ad..0000000000 --- a/src/mesa/pipe/softpipe/sp_rgba_tile.c +++ /dev/null @@ -1,758 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * RGBA/float tile get/put functions. - * XXX eventually this should go into a utility library usable by - * the state tracker too. - * Then, remove pipe->get/put_tile_rgba() - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" - -#include "sp_rgba_tile.h" - - -/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */ -#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) - -#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ - us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) - - - -#define CLIP_TILE \ - do { \ - if (x >= ps->width) \ - return; \ - if (y >= ps->height) \ - return; \ - if (x + w > ps->width) \ - w = ps->width - x; \ - if (y + h > ps->height) \ - h = ps->height - y; \ - } while(0) - - -/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ - -static void -a8r8g8b8_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const unsigned *src - = ((const unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -static void -a8r8g8b8_put_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) -{ - unsigned *dst - = ((unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - dst[j] = (a << 24) | (r << 16) | (g << 8) | b; - pRow += 4; - } - dst += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ - -static void -b8g8r8a8_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const unsigned *src - = ((const unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -static void -b8g8r8a8_put_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) -{ - unsigned *dst - = ((unsigned *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - dst[j] = (b << 24) | (g << 16) | (r << 8) | a; - pRow += 4; - } - dst += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ - -static void -a1r5g5b5_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - - assert(ps->format == PIPE_FORMAT_A1R5G5B5_UNORM); - - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - const ushort pixel = src[j]; - p[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); - p[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); - p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - p[3] = ((pixel >> 15) ) * 1.0f; - p += 4; - } - src += ps->pitch; - } -} - - -/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ - -static void -a4r4g4b4_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - - assert(ps->format == PIPE_FORMAT_A4R4G4B4_UNORM); - - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - const ushort pixel = src[j]; - p[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); - p[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); - p[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); - p[3] = ((pixel >> 12) ) * (1.0f / 15.0f); - p += 4; - } - src += ps->pitch; - } -} - - -/*** PIPE_FORMAT_R5G6B5_UNORM ***/ - -static void -r5g6b5_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - - assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); - - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - const ushort pixel = src[j]; - p[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); - p[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); - p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - p[3] = 1.0f; - p += 4; - } - src += ps->pitch; - } -} - - -static void -r5g5b5_put_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) -{ - ushort *dst - = ((ushort *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++) { - uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); - uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); - uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); - dst[j] = (r << 11) | (g << 5) | (b); - pRow += 4; - } - dst += ps->pitch; - p += w0 * 4; - } -} - - - -/*** PIPE_FORMAT_Z16_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z16_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - const float scale = 1.0f / 65535.0f; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_Z16_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = src[j] * scale; - } - src += ps->pitch; - p += 4 * w0; - } -} - - - - -/*** PIPE_FORMAT_U_L8 ***/ - -static void -l8_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const ubyte *src - = ((const ubyte *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_L8); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(src[j]); - pRow[3] = 1.0; - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_U_A8 ***/ - -static void -a8_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const ubyte *src - = ((const ubyte *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_A8); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[0] = - pRow[1] = - pRow[2] = 0.0; - pRow[3] = UBYTE_TO_FLOAT(src[j]); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ - -static void -r16g16b16a16_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const short *src - = ((const short *) (ps->map)) - + (y * ps->pitch + x) * 4; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - const short *pixel = src; - for (j = 0; j < w; j++) { - pRow[0] = SHORT_TO_FLOAT(pixel[0]); - pRow[1] = SHORT_TO_FLOAT(pixel[1]); - pRow[2] = SHORT_TO_FLOAT(pixel[2]); - pRow[3] = SHORT_TO_FLOAT(pixel[3]); - pRow += 4; - pixel += 4; - } - src += ps->pitch * 4; - p += w0 * 4; - } -} - - -static void -r16g16b16a16_put_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) -{ - short *dst - = ((short *) (ps->map)) - + (y * ps->pitch + x) * 4; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++) { - short r, g, b, a; - UNCLAMPED_FLOAT_TO_SHORT(r, pRow[0]); - UNCLAMPED_FLOAT_TO_SHORT(g, pRow[1]); - UNCLAMPED_FLOAT_TO_SHORT(b, pRow[2]); - UNCLAMPED_FLOAT_TO_SHORT(a, pRow[3]); - dst[j*4+0] = r; - dst[j*4+1] = g; - dst[j*4+2] = b; - dst[j*4+3] = a; - pRow += 4; - } - dst += ps->pitch * 4; - p += w0 * 4; - } -} - - - -/*** PIPE_FORMAT_U_I8 ***/ - -static void -i8_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const ubyte *src - = ((const ubyte *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_I8); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = UBYTE_TO_FLOAT(src[j]); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - -/*** PIPE_FORMAT_U_A8_L8 ***/ - -static void -a8_l8_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const ushort *src - = ((const ushort *) (ps->map)) - + y * ps->pitch + x; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_A8_L8); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - const ushort p = src[j]; - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(p & 0xff); - pRow[3] = UBYTE_TO_FLOAT(p >> 8); - pRow += 4; - } - src += ps->pitch; - p += w0 * 4; - } -} - - - - -/*** PIPE_FORMAT_Z32_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z32_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const uint *src - = ((const uint *) (ps->map)) - + y * ps->pitch + x; - const double scale = 1.0 / (double) 0xffffffff; - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_Z16_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (float) (scale * src[j]); - } - src += ps->pitch; - p += 4 * w0; - } -} - - -/*** PIPE_FORMAT_S8Z24_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -s8z24_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const uint *src - = ((const uint *) (ps->map)) - + y * ps->pitch + x; - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_S8Z24_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (float) (scale * (src[j] & 0xffffff)); - } - src += ps->pitch; - p += 4 * w0; - } -} - - -/*** PIPE_FORMAT_Z24S8_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -z24s8_get_tile_rgba(struct pipe_surface *ps, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) -{ - const uint *src - = ((const uint *) (ps->map)) - + y * ps->pitch + x; - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_Z24S8_UNORM); - - CLIP_TILE; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (float) (scale * (src[j] >> 8)); - } - src += ps->pitch; - p += 4 * w0; - } -} - - -void -softpipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p) -{ - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - a1r5g5b5_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_A4R4G4B4_UNORM: - a4r4g4b4_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_R5G6B5_UNORM: - r5g6b5_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_U_L8: - l8_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_U_A8: - a8_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_U_I8: - i8_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_U_A8_L8: - a8_l8_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_Z16_UNORM: - z16_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_Z32_UNORM: - z32_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_S8Z24_UNORM: - s8z24_get_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_Z24S8_UNORM: - z24s8_get_tile_rgba(ps, x, y, w, h, p); - break; - default: - assert(0); - } -} - - -void -softpipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p) -{ - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_put_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_put_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - /*a1r5g5b5_put_tile_rgba(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_R5G6B5_UNORM: - r5g5b5_put_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_R8G8B8A8_UNORM: - break; - case PIPE_FORMAT_U_L8: - /*l8_put_tile_rgba(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_U_A8: - /*a8_put_tile_rgba(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_U_I8: - /*i8_put_tile_rgba(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_U_A8_L8: - /*a8_l8_put_tile_rgba(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_put_tile_rgba(ps, x, y, w, h, p); - break; - case PIPE_FORMAT_Z16_UNORM: - /*z16_put_tile_rgba(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_Z32_UNORM: - /*z32_put_tile_rgba(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_S8Z24_UNORM: - /*s8z24_put_tile_rgba(ps, x, y, w, h, p);*/ - break; - case PIPE_FORMAT_Z24S8_UNORM: - /*z24s8_put_tile_rgba(ps, x, y, w, h, p);*/ - break; - default: - assert(0); - } -} diff --git a/src/mesa/pipe/softpipe/sp_rgba_tile.h b/src/mesa/pipe/softpipe/sp_rgba_tile.h deleted file mode 100644 index ffa1b65de1..0000000000 --- a/src/mesa/pipe/softpipe/sp_rgba_tile.h +++ /dev/null @@ -1,48 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SP_RGBA_TILE_H -#define SP_RGBA_TILE_H - -#include "pipe/p_compiler.h" - -struct pipe_context; -struct pipe_surface; - -extern void -softpipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p); - -extern void -softpipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p); - -#endif -- cgit v1.2.3 From 7cef9237ae663f107dce82a688e8e0a9ce8193bc Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 18 Dec 2007 09:59:54 -0500 Subject: i965: don't treat swz differently and upload vertex buffers --- src/mesa/pipe/i965simple/brw_draw.c | 5 ++++- src/mesa/pipe/i965simple/brw_draw_upload.c | 2 +- src/mesa/pipe/i965simple/brw_vs_emit.c | 31 ++++++++++++++++++------------ 3 files changed, 24 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_draw.c b/src/mesa/pipe/i965simple/brw_draw.c index 25861a4373..acfb524a30 100644 --- a/src/mesa/pipe/i965simple/brw_draw.c +++ b/src/mesa/pipe/i965simple/brw_draw.c @@ -158,10 +158,13 @@ static boolean brw_try_draw_elements( struct pipe_context *pipe, /* Upload index, vertex data: */ - if (index_buffer && + if (index_buffer && !brw_upload_indices( brw, index_buffer, index_size, start, count )) return FALSE; + if (!brw_upload_vertex_buffers(brw)) + return FALSE; + if (!brw_upload_vertex_elements( brw )) return FALSE; diff --git a/src/mesa/pipe/i965simple/brw_draw_upload.c b/src/mesa/pipe/i965simple/brw_draw_upload.c index 79144837e8..19626ca633 100644 --- a/src/mesa/pipe/i965simple/brw_draw_upload.c +++ b/src/mesa/pipe/i965simple/brw_draw_upload.c @@ -217,7 +217,7 @@ boolean brw_upload_vertex_buffers( struct brw_context *brw ) for (i = 0; i < BRW_VEP_MAX; i++) { - if (brw->vb.vbo_array[i]->buffer == NULL) { + if (brw->vb.vbo_array[i] == NULL) { nr_enabled = i; break; } diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index 530e17a736..f4d61eade0 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -995,7 +995,7 @@ static void process_declaration(const struct tgsi_full_declaration *decl, printf("DECLARATION MASK = %d\n", decl->u.DeclarationMask.Mask); assert(0); - } else { //range + } else { /*range*/ idx = decl->u.DeclarationRange.First; } switch (decl->Semantic.SemanticName) { @@ -1057,17 +1057,15 @@ static void process_instruction(struct brw_vs_compile *c, unsigned insn, if_insn = 0; */ - /* Get argument regs. SWZ is special and does this itself. */ - if (inst->Instruction.Opcode != TGSI_OPCODE_SWZ) - for (i = 0; i < 3; i++) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - index = src->SrcRegister.Index; - file = src->SrcRegister.File; - if (file == TGSI_FILE_OUTPUT&&c->output_regs[index].used_in_src) - args[i] = c->output_regs[index].reg; - else - args[i] = get_arg(c, &src->SrcRegister); - } + for (i = 0; i < 3; i++) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + index = src->SrcRegister.Index; + file = src->SrcRegister.File; + if (file == TGSI_FILE_OUTPUT&&c->output_regs[index].used_in_src) + args[i] = c->output_regs[index].reg; + else + args[i] = get_arg(c, &src->SrcRegister); + } /* Get dest regs. Note that it is possible for a reg to be both * dst and arg, given the static allocation of registers. So @@ -1208,11 +1206,16 @@ static void process_instruction(struct brw_vs_compile *c, break; #endif case TGSI_OPCODE_RET: +#if 0 brw_ADD(p, get_addr_reg(stack_index), get_addr_reg(stack_index), brw_imm_d(-4)); brw_set_access_mode(p, BRW_ALIGN_1); brw_MOV(p, brw_ip_reg(), deref_1uw(stack_index, 0)); brw_set_access_mode(p, BRW_ALIGN_16); +#else + /*brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));*/ +#endif + break; case TGSI_OPCODE_END: brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); break; @@ -1295,7 +1298,11 @@ void brw_vs_emit(struct brw_vs_compile *c) * now that we know what vars are being used allocate * registers for them.*/ brw_vs_alloc_regs(c, &prog_info); + + brw_set_access_mode(p, BRW_ALIGN_1); brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); + brw_set_access_mode(p, BRW_ALIGN_16); + allocated_registers = 1; } process_instruction(c, inst, &prog_info); -- cgit v1.2.3 From 9d4ab42f4be3a26f702729cc79ef67f8afc2eca5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 18 Dec 2007 16:56:22 +0000 Subject: vbo: unmap and remap immediate vbo before/after each draw. Also use BufferData(NULL) to get fresh storage and avoid synchronous operation where we would have to flush and wait for the fence after each draw because of the map. This will chew through a whole load of buffer space on small draws, so it isn't a proper solution. Need to support a no-fence or append mapping mode to do this right, or use user buffers. --- src/mesa/vbo/vbo_exec_draw.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c index 68aba1df66..23a3658386 100644 --- a/src/mesa/vbo/vbo_exec_draw.c +++ b/src/mesa/vbo/vbo_exec_draw.c @@ -221,8 +221,18 @@ void vbo_exec_vtx_flush( struct vbo_exec_context *exec ) if (exec->vtx.copied.nr != exec->vtx.vert_count) { GLcontext *ctx = exec->ctx; + GLenum target = GL_ARRAY_BUFFER_ARB; + GLenum access = GL_READ_WRITE_ARB; + GLenum usage = GL_STREAM_DRAW_ARB; + GLsizei size = VBO_VERT_BUFFER_SIZE * sizeof(GLfloat); + + /* Before the unmap (why?) + */ vbo_exec_bind_arrays( ctx ); + ctx->Driver.UnmapBuffer(ctx, target, exec->vtx.bufferobj); + exec->vtx.buffer_map = NULL; + vbo_context(ctx)->draw_prims( ctx, exec->vtx.inputs, exec->vtx.prim, @@ -230,6 +240,12 @@ void vbo_exec_vtx_flush( struct vbo_exec_context *exec ) NULL, 0, exec->vtx.vert_count - 1); + + /* Get new data: + */ + ctx->Driver.BufferData(ctx, target, size, NULL, usage, exec->vtx.bufferobj); + exec->vtx.buffer_map + = ctx->Driver.MapBuffer(ctx, target, access, exec->vtx.bufferobj); } } -- cgit v1.2.3 From 208b2ad8ab51c472886388fdd872e3a86e2c1c5c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 18 Dec 2007 16:57:17 +0000 Subject: gallium: give userbuffers some storage in the aub buffer pool --- src/mesa/pipe/xlib/xm_winsys_aub.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index 0348c2ad40..ef3d975afb 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -51,7 +51,6 @@ struct aub_buffer { unsigned refcount; unsigned map_count; boolean dump_on_unmap; - boolean userbuffer; }; @@ -276,9 +275,15 @@ static struct pipe_buffer_handle * aub_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) { struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer); - sbo->size = bytes; - sbo->userbuffer = 1; - sbo->data = ptr; + + /* Lets hope this is meant for upload, not as a result! + */ + aub_buffer_data( winsys, + pipe_bo(sbo), + bytes, + ptr, + 0 ); + return pipe_bo(sbo); } -- cgit v1.2.3 From c31416971e4eac148f8e82d6c4392bd6f9cbc05d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 18 Dec 2007 17:25:33 +0000 Subject: gallium: rationalize vertex_element state packet Remove dst_offset (not used) Add nr_components, which could be calculated from format, but would be too much effort. Update i965 driver to cope. --- src/mesa/pipe/i965simple/brw_context.h | 10 +------ src/mesa/pipe/i965simple/brw_draw_upload.c | 14 ++------- src/mesa/pipe/i965simple/brw_state.c | 46 ++++++++++++++---------------- src/mesa/pipe/p_state.h | 4 +-- src/mesa/state_tracker/st_draw.c | 6 ++-- 5 files changed, 29 insertions(+), 51 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h index 11146570be..fc2cb055e9 100644 --- a/src/mesa/pipe/i965simple/brw_context.h +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -443,14 +443,6 @@ struct brw_cached_batch_item { */ #define ATTRIB_BIT_DWORDS ((PIPE_ATTRIB_MAX+31)/32) -struct brw_vertex_element { - struct brw_vertex_element_state vep; - - unsigned index; - unsigned element_size; - unsigned count; - unsigned vbo_rebase_offset; -}; @@ -508,7 +500,7 @@ struct brw_context */ struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX]; - struct brw_vertex_element inputs[PIPE_ATTRIB_MAX]; + struct brw_vertex_element_state inputs[PIPE_ATTRIB_MAX]; #define BRW_NR_UPLOAD_BUFS 17 #define BRW_UPLOAD_INIT_SIZE (128*1024) diff --git a/src/mesa/pipe/i965simple/brw_draw_upload.c b/src/mesa/pipe/i965simple/brw_draw_upload.c index 19626ca633..88d6c9d111 100644 --- a/src/mesa/pipe/i965simple/brw_draw_upload.c +++ b/src/mesa/pipe/i965simple/brw_draw_upload.c @@ -260,18 +260,8 @@ boolean brw_upload_vertex_elements( struct brw_context *brw ) memset(&vep, 0, sizeof(vep)); - for (i = 0; i < nr_enabled; i++) { - struct brw_vertex_element *input = &brw->vb.inputs[i]; - - switch (brw->vb.vbo_array[input->vep.ve0.vertex_buffer_index]->pitch) { - case 0: input->vep.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_0; - case 1: input->vep.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; - case 2: input->vep.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; - case 3: input->vep.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; - break; - } - vep.ve[i] = input->vep; - } + for (i = 0; i < nr_enabled; i++) + vep.ve[i] = brw->vb.inputs[i]; vep.header.length = (1 + nr_enabled * sizeof(vep.ve[0])/4) - 2; diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c index 2008853654..daf14ff4ff 100644 --- a/src/mesa/pipe/i965simple/brw_state.c +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -272,31 +272,27 @@ static void brw_set_vertex_element(struct pipe_context *pipe, struct brw_context *brw = brw_context(pipe); assert(index < PIPE_ATTRIB_MAX); - struct brw_vertex_element el; - memset(&el, 0, sizeof(struct brw_vertex_element)); - - /* do we need those anymore?*/ - el.index = index; -#if 0 - /*FIXME*/ - el.element_size = 0; - el.count = 0; - el.vbo_rebase_offset = 0; -#endif - - el.vep.ve0.src_offset = element->src_offset; - el.vep.ve0.src_format = brw_translate_surface_format(element->src_format); - el.vep.ve0.valid = 1; - el.vep.ve0.vertex_buffer_index = element->vertex_buffer_index; - - el.vep.ve1.dst_offset = index * 4; - el.vep.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; - el.vep.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; - el.vep.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; - el.vep.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; - /*can we count of brw->vb.vbo_array[element->vertex_buffer_index] - * being initialized ok to actually compute vbcomponent's - * correctly? */ + struct brw_vertex_element_state el; + memset(&el, 0, sizeof(el)); + + el.ve0.src_offset = element->src_offset; + el.ve0.src_format = brw_translate_surface_format(element->src_format); + el.ve0.valid = 1; + el.ve0.vertex_buffer_index = element->vertex_buffer_index; + + el.ve1.dst_offset = index * 4; + + el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; + + switch (element->nr_components) { + case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; + case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; + case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; + break; + } brw->vb.inputs[index] = el; } diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index b7793c6d31..16d50fdb82 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -307,8 +307,8 @@ struct pipe_vertex_element * this attribute live in? */ unsigned vertex_buffer_index:5; - - unsigned dst_offset:8; + unsigned nr_components:3; + enum pipe_format src_format; /**< PIPE_FORMAT_* */ }; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 274ae86a3e..a3e061d604 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -289,7 +289,7 @@ st_draw_vbo(GLcontext *ctx, vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ vbuffer[attr].max_index = 0; /* need this? */ velement.vertex_buffer_index = attr; - velement.dst_offset = 0; /* need this? */ + velement.nr_components = arrays[mesaAttr]->Size; velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, arrays[mesaAttr]->Size, arrays[mesaAttr]->Normalized); @@ -415,7 +415,7 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, velement.src_offset = i * 4 * sizeof(GLfloat); velement.vertex_buffer_index = 0; velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - velement.dst_offset = 0; + velement.nr_components = 4; pipe->set_vertex_element(pipe, i, &velement); } @@ -547,7 +547,7 @@ st_feedback_draw_vbo(GLcontext *ctx, vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ vbuffer[attr].max_index = 0; /* need this? */ velement.vertex_buffer_index = attr; - velement.dst_offset = 0; /* need this? */ + velement.nr_components = arrays[mesaAttr]->Size; velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, arrays[mesaAttr]->Size, arrays[mesaAttr]->Normalized); -- cgit v1.2.3 From fb4eb8c91b7e76bf66d92da91c1e6d994b6e7e3e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Dec 2007 15:59:54 -0700 Subject: fix some semantic info mix-ups in calculate_vertex_layout() --- src/mesa/pipe/softpipe/sp_state_derived.c | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 94072a2d30..a5e766781f 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -81,21 +81,21 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) break; case TGSI_SEMANTIC_COLOR: - if (fs->input_semantic_index[i] == 0) { + if (vs->output_semantic_index[i] == 0) { front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); } else { - assert(fs->input_semantic_index[i] == 1); + assert(vs->output_semantic_index[i] == 1); front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); } break; case TGSI_SEMANTIC_BCOLOR: - if (fs->input_semantic_index[i] == 0) { + if (vs->output_semantic_index[i] == 0) { emitBack0 = TRUE; } else { - assert(fs->input_semantic_index[i] == 1); + assert(vs->output_semantic_index[i] == 1); emitBack1 = TRUE; } break; @@ -121,11 +121,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) } } -#if 00 - softpipe->nr_frag_attrs = vinfo->num_attribs; -#else softpipe->nr_frag_attrs = fs->num_inputs; -#endif /* We want these after all other attribs since they won't get passed * to the fragment shader. All prior vertex output attribs should match -- cgit v1.2.3 From 8c20747834c2ea7006f127e974560534ab279da2 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Dec 2007 16:00:31 -0700 Subject: setup the frontface register (fog.y, ATM) --- src/mesa/pipe/softpipe/sp_prim_setup.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 2ccf5e2624..de45529bf9 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -516,7 +516,7 @@ setup_fragcoord_coeff(struct setup_stage *setup) static void setup_tri_coefficients( struct setup_stage *setup ) { const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; -#define USE_INPUT_MAP 0 +#define USE_INPUT_MAP 01 #if USE_INPUT_MAP const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; #endif @@ -554,6 +554,12 @@ static void setup_tri_coefficients( struct setup_stage *setup ) */ setup_fragcoord_coeff(setup); } + else if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } else { #endif uint j; -- cgit v1.2.3 From 52da6b559a47eca2c1a8ec1b713e188f38e1d16a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Dec 2007 16:00:58 -0700 Subject: fix bug on GL_VERTEX_PROGRAM_TWO_SIDE path --- src/mesa/state_tracker/st_atom_rasterizer.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 5c6b89d78c..35fd506458 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -109,7 +109,7 @@ static void update_raster_state( struct st_context *st ) * GL_LIGHT_MODEL_TWO_SIDE is set) or from vertex programs (when * GL_VERTEX_PROGRAM_TWO_SIDE is set). Note the logic here. */ - if (ctx->VertexProgram._Enabled) { + if (ctx->VertexProgram._Current) { raster.light_twoside = ctx->VertexProgram.TwoSideEnabled; } else if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { -- cgit v1.2.3 From d0a63de37888966591735a190d69b0333d31bef5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Dec 2007 16:01:25 -0700 Subject: turn off TGSI_DEBUG --- src/mesa/state_tracker/st_program.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index fe22233c93..6802bb3b06 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -47,7 +47,7 @@ #include "st_mesa_to_tgsi.h" -#define TGSI_DEBUG 01 +#define TGSI_DEBUG 0 /** -- cgit v1.2.3 From 4e1a4458662e1766e9de7ce0e23de0766dcb1b9e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Dec 2007 04:58:29 -0500 Subject: i965: be smarter about register allocation --- src/mesa/pipe/i965simple/brw_vs_emit.c | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index f4d61eade0..174f56ff34 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -74,25 +74,23 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c, /* Vertex program parameters from curbe: */ +#if 0 nr_params = c->vp->program.num_inputs; /*FIXME: i think this is wrong... */ for (i = 0; i < nr_params; i++) { - c->regs[TGSI_FILE_INPUT][i] = stride( brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1); + c->regs[TGSI_FILE_INPUT][i] = stride(brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1); } reg += (nr_params+1)/2; - +#endif c->prog_data.curb_read_length = reg - 1; /* Allocate input regs: */ - c->nr_inputs = 0; - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { - if (c->prog_data.inputs_read & (1<nr_inputs++; + c->nr_inputs = c->vp->program.num_inputs; + for (i = 0; i < c->nr_inputs; i++) { c->regs[TGSI_FILE_INPUT][i] = brw_vec8_grf(reg, 0); reg++; - } } @@ -1061,7 +1059,7 @@ static void process_instruction(struct brw_vs_compile *c, struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; index = src->SrcRegister.Index; file = src->SrcRegister.File; - if (file == TGSI_FILE_OUTPUT&&c->output_regs[index].used_in_src) + if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src) args[i] = c->output_regs[index].reg; else args[i] = get_arg(c, &src->SrcRegister); -- cgit v1.2.3 From 93d90c6b570298ea96c5952af1acb31b0a89362c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Dec 2007 05:28:28 -0500 Subject: i965: stuff the outputs into mrf registers when possible --- src/mesa/pipe/i965simple/brw_vs_emit.c | 11 ++++++++--- 1 file changed, 8 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index 174f56ff34..d8483f213a 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -117,9 +117,14 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c, mrf++; } #else - /* for now stuff everything in grf */ - c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); - reg++; + /*treat pos differently for now */ + if (i == info->pos_idx) { + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); + reg++; + } else { + c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf); + mrf++; + } #endif } -- cgit v1.2.3 From 9791c16b317d1536ba5c2d67398d17b80da2f015 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 10:20:14 +0000 Subject: 965: populate fp_input_count in setup key --- src/mesa/pipe/i965simple/brw_sf.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_sf.c b/src/mesa/pipe/i965simple/brw_sf.c index e7c02beda5..7b6fd3fff6 100644 --- a/src/mesa/pipe/i965simple/brw_sf.c +++ b/src/mesa/pipe/i965simple/brw_sf.c @@ -128,6 +128,10 @@ static void upload_sf_prog( struct brw_context *brw ) /* CACHE_NEW_VS_PROG */ key.vp_output_count = brw->vs.prog_data->outputs_written; + /* BRW_NEW_FS */ + key.fp_input_count = brw->attribs.FragmentProgram->info.nr_regs[TGSI_FILE_INPUT]; + + /* BRW_NEW_REDUCED_PRIMITIVE */ switch (brw->reduced_primitive) { case PIPE_PRIM_TRIANGLES: -- cgit v1.2.3 From 4fd2a2ac6f6593487d59fac2d8c0365179509ae5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 10:49:16 +0000 Subject: 965: fix colormask state --- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index db8f670970..0f851dd1ed 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -221,10 +221,10 @@ static void upload_wm_surfaces(struct brw_context *brw ) brw->attribs.Blend->blend_enable); - surf.ss0.writedisable_red = !brw->attribs.BlendColor.color[0]; - surf.ss0.writedisable_green = !brw->attribs.BlendColor.color[1]; - surf.ss0.writedisable_blue = !brw->attribs.BlendColor.color[2]; - surf.ss0.writedisable_alpha = !brw->attribs.BlendColor.color[3]; + surf.ss0.writedisable_red = !(brw->attribs.Blend->colormask & PIPE_MASK_R); + surf.ss0.writedisable_green = !(brw->attribs.Blend->colormask & PIPE_MASK_G); + surf.ss0.writedisable_blue = !(brw->attribs.Blend->colormask & PIPE_MASK_B); + surf.ss0.writedisable_alpha = !(brw->attribs.Blend->colormask & PIPE_MASK_A); -- cgit v1.2.3 From 94c2ab2895f220d5d5156db71197446b3b89f52b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 10:55:49 +0000 Subject: 965: fix vertex pointsize state, match default cull mode --- src/mesa/pipe/i965simple/brw_sf_state.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_sf_state.c b/src/mesa/pipe/i965simple/brw_sf_state.c index 0de6e7240e..2dde277991 100644 --- a/src/mesa/pipe/i965simple/brw_sf_state.c +++ b/src/mesa/pipe/i965simple/brw_sf_state.c @@ -133,7 +133,7 @@ static void upload_sf_unit( struct brw_context *brw ) else sf.sf6.cull_mode = BRW_CULLMODE_NONE; #else - sf.sf5.front_winding = BRW_FRONTWINDING_CW; + sf.sf5.front_winding = BRW_FRONTWINDING_CCW; sf.sf6.cull_mode = BRW_CULLMODE_NONE; #endif @@ -149,7 +149,7 @@ static void upload_sf_unit( struct brw_context *brw ) sf.sf7.sprite_point = brw->attribs.Raster->point_sprite; sf.sf7.point_size = CLAMP(brw->attribs.Raster->line_width, 1.0, 255.0) * (1<<3); - sf.sf7.use_point_size_state = brw->attribs.Raster->point_size_per_vertex; + sf.sf7.use_point_size_state = !brw->attribs.Raster->point_size_per_vertex; /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons: */ -- cgit v1.2.3 From e03982475795f19a674b79940c135584b13fd79a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 11:10:06 +0000 Subject: 965: disable clipping for now --- src/mesa/pipe/i965simple/brw_clip.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_clip.c b/src/mesa/pipe/i965simple/brw_clip.c index fc86e0a446..268124cc53 100644 --- a/src/mesa/pipe/i965simple/brw_clip.c +++ b/src/mesa/pipe/i965simple/brw_clip.c @@ -154,6 +154,8 @@ static void upload_clip_prog(struct brw_context *brw) key.do_flat_shading = (brw->attribs.Raster->flatshade); /* BRW_NEW_CLIP */ key.nr_userclip = brw->attribs.Clip.nr; /* XXX */ + +#if 0 key.clip_mode = BRW_CLIPMODE_NORMAL; if (key.primitive == PIPE_PRIM_TRIANGLES) { @@ -185,6 +187,9 @@ static void upload_clip_prog(struct brw_context *brw) } } } +#else + key.clip_mode = BRW_CLIPMODE_ACCEPT_ALL; +#endif if (!search_cache(brw, &key)) compile_clip_prog( brw, &key ); -- cgit v1.2.3 From aba4ee24cf1de8e1ec9a7f02c2d97c3156d87eef Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 11:10:30 +0000 Subject: 965: disable clipping more, and restore drawing rect packet --- src/mesa/pipe/i965simple/brw_misc_state.c | 29 ++++++++++++++++++++++++++++- 1 file changed, 28 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_misc_state.c b/src/mesa/pipe/i965simple/brw_misc_state.c index e600e9d8de..70b1c43f2c 100644 --- a/src/mesa/pipe/i965simple/brw_misc_state.c +++ b/src/mesa/pipe/i965simple/brw_misc_state.c @@ -57,6 +57,33 @@ void brw_upload_blend_constant_color(struct brw_context *brw) BRW_CACHED_BATCH_STRUCT(brw, &bcc); } +static void upload_drawing_rect(struct brw_context *brw) +{ + struct brw_drawrect bdr; + + memset(&bdr, 0, sizeof(bdr)); + bdr.header.opcode = CMD_DRAW_RECT; + bdr.header.length = sizeof(bdr)/4 - 2; + bdr.xmin = 0; + bdr.ymin = 0; + bdr.xmax = brw->attribs.FrameBuffer.cbufs[0]->width; + bdr.ymax = brw->attribs.FrameBuffer.cbufs[0]->height; + bdr.xorg = 0; + bdr.yorg = 0; + + /* Can't use BRW_CACHED_BATCH_STRUCT because this is also emitted + * uncached in brw_draw.c: + */ + BRW_BATCH_STRUCT(brw, &bdr); +} + +const struct brw_tracked_state brw_drawing_rect = { + .dirty = { + .brw = BRW_NEW_SCENE, + .cache = 0 + }, + .update = upload_drawing_rect +}; /** * Upload the binding table pointers, which point each stage's array of surface @@ -117,7 +144,7 @@ static void upload_pipelined_state_pointers(struct brw_context *brw ) psp.gs.enable = 1; } - { + if (0) { psp.clp.offset = brw->clip.state_gs_offset >> 5; psp.clp.enable = 1; } -- cgit v1.2.3 From 38de4c5715209171bc168454f7a6dc6c43b341b8 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Dec 2007 06:00:30 -0500 Subject: i965: actually detect whether a sampler is enabled/disabled --- src/mesa/pipe/i965simple/brw_wm_sampler_state.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c b/src/mesa/pipe/i965simple/brw_wm_sampler_state.c index fbeea8c809..cfb430eb09 100644 --- a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_sampler_state.c @@ -236,7 +236,8 @@ static void upload_wm_samplers(struct brw_context *brw) /* BRW_NEW_SAMPLER */ for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { - if (brw->attribs.Samplers[unit]) { /* FIXME: correctly detect enabled ones */ + /* determine unit enable/disable by looking for a bound texture */ + if (brw->attribs.Texture[unit]) { const struct pipe_sampler_state *sampler = brw->attribs.Samplers[unit]; unsigned sdc_gs_offset = upload_default_color(brw, sampler->border_color); -- cgit v1.2.3 From 1f37b4d87e798690ed080e5da356366906ff55cd Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 11:32:56 +0000 Subject: 965: respect surface width when dumping bitmap --- src/mesa/pipe/xlib/brw_aub.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/brw_aub.c b/src/mesa/pipe/xlib/brw_aub.c index 1b7fc3c20e..541d50c6e4 100644 --- a/src/mesa/pipe/xlib/brw_aub.c +++ b/src/mesa/pipe/xlib/brw_aub.c @@ -331,7 +331,7 @@ void brw_aub_dump_bmp( struct brw_aubfile *aubfile, db.format = format; db.bpp = surface->cpp * 8; db.pitch = surface->pitch; - db.xsize = surface->pitch; + db.xsize = surface->width; db.ysize = surface->height; db.addr = gtt_offset; db.unknown = /* surface->tiled ? 0x4 : */ 0x0; -- cgit v1.2.3 From 2f53713beb9aa425d8248bae6a3c388a7de189de Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 11:35:21 +0000 Subject: 965: fix off-by-one in surface dimensions --- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index 0f851dd1ed..1eb11cbb91 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -155,13 +155,13 @@ void brw_update_texture_surface( struct brw_context *brw, surf.ss1.base_addr = brw_buffer_offset( brw, tObj->buffer ); surf.ss2.mip_count = tObj->base.last_level - tObj->base.first_level; - surf.ss2.width = tObj->base.width[0]; - surf.ss2.height = tObj->base.height[0]; + surf.ss2.width = tObj->base.width[0] - 1; + surf.ss2.height = tObj->base.height[0] - 1; surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; surf.ss3.tiled_surface = 0; /* always zero */ - surf.ss3.pitch = tObj->pitch; - surf.ss3.depth = tObj->base.depth[0]; + surf.ss3.pitch = tObj->pitch - 1; + surf.ss3.depth = tObj->base.depth[0] - 1; surf.ss4.min_lod = 0; @@ -192,25 +192,25 @@ static void upload_wm_surfaces(struct brw_context *brw ) /* BRW_NEW_FRAMEBUFFER */ - struct pipe_surface *region = brw->attribs.FrameBuffer.cbufs[0];/*fixme*/ + struct pipe_surface *pipe_surface = brw->attribs.FrameBuffer.cbufs[0];/*fixme*/ memset(&surf, 0, sizeof(surf)); - if (region != NULL) { - if (region->cpp == 4) + if (pipe_surface != NULL) { + if (pipe_surface->cpp == 4) surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; else surf.ss0.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM; surf.ss0.surface_type = BRW_SURFACE_2D; - surf.ss1.base_addr = brw_buffer_offset( brw, region->buffer ); + surf.ss1.base_addr = brw_buffer_offset( brw, pipe_surface->buffer ); - surf.ss2.width = region->width; - surf.ss2.height = region->height; + surf.ss2.width = pipe_surface->width - 1; + surf.ss2.height = pipe_surface->height - 1; surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; surf.ss3.tiled_surface = 0; - surf.ss3.pitch = region->pitch; + surf.ss3.pitch = pipe_surface->pitch - 1; } else { surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; surf.ss0.surface_type = BRW_SURFACE_NULL; -- cgit v1.2.3 From 05ec043d1ca01f0be94e8b8ceaeb730475b308f9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 11:36:03 +0000 Subject: 965: fix off-by-one in scissor rect --- src/mesa/pipe/i965simple/brw_sf_state.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_sf_state.c b/src/mesa/pipe/i965simple/brw_sf_state.c index 2dde277991..9acd3ea61b 100644 --- a/src/mesa/pipe/i965simple/brw_sf_state.c +++ b/src/mesa/pipe/i965simple/brw_sf_state.c @@ -58,9 +58,9 @@ static void upload_sf_vp(struct brw_context *brw) /* _NEW_SCISSOR */ sfv.scissor.xmin = brw->attribs.Scissor.minx; - sfv.scissor.xmax = brw->attribs.Scissor.maxx; + sfv.scissor.xmax = brw->attribs.Scissor.maxx - 1; sfv.scissor.ymin = brw->attribs.Scissor.miny; - sfv.scissor.ymax = brw->attribs.Scissor.maxy; + sfv.scissor.ymax = brw->attribs.Scissor.maxy - 1; brw->sf.vp_gs_offset = brw_cache_data( &brw->cache[BRW_SF_VP], &sfv ); } -- cgit v1.2.3 From a0ce71b51dd7aac4e219a023bae6fc9771d64c03 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 11:49:56 +0000 Subject: 965: pitch is in bytes not pixels --- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index 1eb11cbb91..69b6202fc8 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -210,7 +210,7 @@ static void upload_wm_surfaces(struct brw_context *brw ) surf.ss2.height = pipe_surface->height - 1; surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; surf.ss3.tiled_surface = 0; - surf.ss3.pitch = pipe_surface->pitch - 1; + surf.ss3.pitch = (pipe_surface->pitch * pipe_surface->cpp) - 1; } else { surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; surf.ss0.surface_type = BRW_SURFACE_NULL; -- cgit v1.2.3 From bc1fc7d6739476748c902a79dcf1676b0ddc17fd Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Dec 2007 07:08:13 -0500 Subject: i965: emit fb write on RET for now (until we get END back) --- src/mesa/pipe/i965simple/brw_wm_glsl.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_glsl.c b/src/mesa/pipe/i965simple/brw_wm_glsl.c index 37d5b216b2..f4b5c13c06 100644 --- a/src/mesa/pipe/i965simple/brw_wm_glsl.c +++ b/src/mesa/pipe/i965simple/brw_wm_glsl.c @@ -935,6 +935,7 @@ static void brw_wm_emit_instruction( struct brw_wm_compile *c, break; case TGSI_OPCODE_RET: +#if 0 brw_push_insn_state(p); brw_set_mask_control(p, BRW_MASK_DISABLE); brw_ADD(p, @@ -944,6 +945,9 @@ static void brw_wm_emit_instruction( struct brw_wm_compile *c, brw_MOV(p, brw_ip_reg(), deref_1ud(c->stack_index, 0)); brw_set_access_mode(p, BRW_ALIGN_16); brw_pop_insn_state(p); +#else + emit_fb_write(c, inst); +#endif break; case TGSI_OPCODE_LOOP: -- cgit v1.2.3 From f367c57802f3800f695acae975fdc30ece6a63b3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 12:23:27 +0000 Subject: 965: fill unused surface pointers with zero --- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 18 ++++-------------- 1 file changed, 4 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index 69b6202fc8..5c7dee5790 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -235,28 +235,18 @@ static void upload_wm_surfaces(struct brw_context *brw ) } + /* BRW_NEW_TEXTURE + */ for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { const struct brw_texture *texUnit = brw->attribs.Texture[i]; - if (texUnit == NULL) - continue; - - /* BRW_NEW_TEXTURE - */ - if (texUnit->base.refcount/*(texUnit->refcount > 0) == really used */) { + if (texUnit && + texUnit->base.refcount/*(texUnit->refcount > 0) == really used */) { brw_update_texture_surface(brw, i); brw->wm.nr_surfaces = i+2; } -#if 0 - else if( texUnit->refcount && - texUnit->_Current == intel->frame_buffer_texobj ) - { - brw->wm.bind.surf_ss_offset[i+1] = brw->wm.bind.surf_ss_offset[0]; - brw->wm.nr_surfaces = i+2; - } -#endif else { brw->wm.bind.surf_ss_offset[i+1] = 0; } -- cgit v1.2.3 From f13a200c8367c58fc962bdad9bcb22263d385886 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Dec 2007 07:49:42 -0500 Subject: actually set the max_index. useful in the driver --- src/mesa/state_tracker/st_draw.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index a3e061d604..b3d463043d 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -287,7 +287,7 @@ st_draw_vbo(GLcontext *ctx, /* common-case setup */ vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ - vbuffer[attr].max_index = 0; /* need this? */ + vbuffer[attr].max_index = max_index; velement.vertex_buffer_index = attr; velement.nr_components = arrays[mesaAttr]->Size; velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, @@ -545,7 +545,7 @@ st_feedback_draw_vbo(GLcontext *ctx, /* common-case setup */ vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ - vbuffer[attr].max_index = 0; /* need this? */ + vbuffer[attr].max_index = max_index; velement.vertex_buffer_index = attr; velement.nr_components = arrays[mesaAttr]->Size; velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, -- cgit v1.2.3 From 2e3dfe97ee514a9ef8bd0a862360854d9be392d3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 13:11:56 +0000 Subject: 965: allocate buffer space to hold batch commands --- src/mesa/pipe/xlib/xm_winsys_aub.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index ef3d975afb..dd7c725507 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -226,10 +226,16 @@ void xmesa_commands_aub(struct pipe_winsys *winsys, unsigned nr_dwords) { struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + unsigned size = nr_dwords * 4; + + assert(iws->used + size < iws->size); + brw_aub_gtt_cmds( iws->aubfile, - 0, /* ?? */ + AUB_BUF_START + iws->used, cmds, nr_dwords * sizeof(int) ); + + iws->used += size; } -- cgit v1.2.3 From 65f67baa425af71fc52639d22c1683da3b386a1b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 13:12:34 +0000 Subject: 965: make sure stipple state gets uploaded --- src/mesa/pipe/i965simple/brw_context.h | 2 - src/mesa/pipe/i965simple/brw_misc_state.c | 86 ++++++++++++++++++++--------- src/mesa/pipe/i965simple/brw_state.h | 5 -- src/mesa/pipe/i965simple/brw_state_upload.c | 7 ++- 4 files changed, 65 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h index fc2cb055e9..59e3d24a1d 100644 --- a/src/mesa/pipe/i965simple/brw_context.h +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -666,8 +666,6 @@ void brwUpdateTextureState( struct brw_context *brw ); void brw_upload_urb_fence(struct brw_context *brw); void brw_upload_constant_buffer_state(struct brw_context *brw); -void brw_upload_polygon_stipple(struct brw_context *brw); -void brw_upload_line_stipple(struct brw_context *brw); void brw_init_surface_functions(struct brw_context *brw); void brw_init_state_functions(struct brw_context *brw); diff --git a/src/mesa/pipe/i965simple/brw_misc_state.c b/src/mesa/pipe/i965simple/brw_misc_state.c index 70b1c43f2c..13b3b1671d 100644 --- a/src/mesa/pipe/i965simple/brw_misc_state.c +++ b/src/mesa/pipe/i965simple/brw_misc_state.c @@ -42,7 +42,7 @@ * Blend color */ -void brw_upload_blend_constant_color(struct brw_context *brw) +static void upload_blend_constant_color(struct brw_context *brw) { struct brw_blend_constant_color bcc; @@ -57,6 +57,19 @@ void brw_upload_blend_constant_color(struct brw_context *brw) BRW_CACHED_BATCH_STRUCT(brw, &bcc); } + +const struct brw_tracked_state brw_blend_constant_color = { + .dirty = { + .brw = BRW_NEW_BLEND, + .cache = 0 + }, + .update = upload_blend_constant_color +}; + + +/*********************************************************************** + * Drawing rectangle + */ static void upload_drawing_rect(struct brw_context *brw) { struct brw_drawrect bdr; @@ -243,7 +256,7 @@ static void upload_depthbuffer(struct brw_context *brw) const struct brw_tracked_state brw_depthbuffer = { .dirty = { - .brw = 0, + .brw = BRW_NEW_SCENE, .cache = 0 }, .update = upload_depthbuffer, @@ -251,50 +264,43 @@ const struct brw_tracked_state brw_depthbuffer = { -/*********************************************************************** - * Polygon stipple offset packet - */ - -static void upload_polygon_stipple_offset(struct brw_context *brw) -{ - struct brw_polygon_stipple_offset bpso; - - memset(&bpso, 0, sizeof(bpso)); - bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET; - bpso.header.length = sizeof(bpso)/4-2; - - bpso.bits0.x_offset = 0; - bpso.bits0.y_offset = 0; - - BRW_CACHED_BATCH_STRUCT(brw, &bpso); -} /*********************************************************************** * Polygon stipple packet */ -void brw_upload_polygon_stipple(struct brw_context *brw) +static void upload_polygon_stipple(struct brw_context *brw) { struct brw_polygon_stipple bps; unsigned i; - /*update the offset at the same time it will always be 0*/ - upload_polygon_stipple_offset(brw); memset(&bps, 0, sizeof(bps)); bps.header.opcode = CMD_POLY_STIPPLE_PATTERN; bps.header.length = sizeof(bps)/4-2; - for (i = 0; i < 32; i++) - bps.stipple[i] = brw->attribs.PolygonStipple->stipple[31 - i]; /* invert */ + /* XXX: state tracker should send *all* state down initially! + */ + if (brw->attribs.PolygonStipple) + for (i = 0; i < 32; i++) + bps.stipple[i] = brw->attribs.PolygonStipple->stipple[31 - i]; /* invert */ BRW_CACHED_BATCH_STRUCT(brw, &bps); } +const struct brw_tracked_state brw_polygon_stipple = { + .dirty = { + .brw = BRW_NEW_STIPPLE, + .cache = 0 + }, + .update = upload_polygon_stipple +}; + + /*********************************************************************** * Line stipple packet */ -void brw_upload_line_stipple(struct brw_context *brw) +static void upload_line_stipple(struct brw_context *brw) { struct brw_line_stipple bls; float tmp; @@ -316,6 +322,14 @@ void brw_upload_line_stipple(struct brw_context *brw) BRW_CACHED_BATCH_STRUCT(brw, &bls); } +const struct brw_tracked_state brw_line_stipple = { + .dirty = { + .brw = BRW_NEW_STIPPLE, + .cache = 0 + }, + .update = upload_line_stipple +}; + /*********************************************************************** * Misc constant state packets @@ -355,6 +369,15 @@ const struct brw_tracked_state brw_pipe_control = { static void upload_invarient_state( struct brw_context *brw ) { + { + struct brw_mi_flush flush; + + memset(&flush, 0, sizeof(flush)); + flush.opcode = CMD_MI_FLUSH; + flush.flags = BRW_FLUSH_STATE_CACHE | BRW_FLUSH_READ_CACHE; + BRW_BATCH_STRUCT(brw, &flush); + } + { /* 0x61040000 Pipeline Select */ /* PipelineSelect : 0 */ @@ -403,6 +426,19 @@ static void upload_invarient_state( struct brw_context *brw ) BRW_BATCH_STRUCT(brw, &vfs); } + + + { + struct brw_polygon_stipple_offset bpso; + + memset(&bpso, 0, sizeof(bpso)); + bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET; + bpso.header.length = sizeof(bpso)/4-2; + bpso.bits0.x_offset = 0; + bpso.bits0.y_offset = 0; + + BRW_BATCH_STRUCT(brw, &bpso); + } } const struct brw_tracked_state brw_invarient_state = { diff --git a/src/mesa/pipe/i965simple/brw_state.h b/src/mesa/pipe/i965simple/brw_state.h index d09711f6f0..62055e5506 100644 --- a/src/mesa/pipe/i965simple/brw_state.h +++ b/src/mesa/pipe/i965simple/brw_state.h @@ -65,7 +65,6 @@ const struct brw_tracked_state brw_urb_fence; const struct brw_tracked_state brw_vertex_state; const struct brw_tracked_state brw_vs_prog; const struct brw_tracked_state brw_vs_unit; -const struct brw_tracked_state brw_wm_input_sizes; const struct brw_tracked_state brw_wm_prog; const struct brw_tracked_state brw_wm_samplers; const struct brw_tracked_state brw_wm_surfaces; @@ -149,10 +148,6 @@ void brw_clear_all_caches( struct brw_context *brw ); void brw_invalidate_pools( struct brw_context *brw ); void brw_clear_batch_cache_flush( struct brw_context *brw ); -void brw_upload_cc_unit(struct brw_context *brw); -void brw_upload_clip_prog(struct brw_context *brw); -void brw_upload_blend_constant_color(struct brw_context *brw); -void brw_upload_wm_samplers(struct brw_context *brw); /* brw_shader_info.c */ diff --git a/src/mesa/pipe/i965simple/brw_state_upload.c b/src/mesa/pipe/i965simple/brw_state_upload.c index 2502e54929..e727601e1e 100644 --- a/src/mesa/pipe/i965simple/brw_state_upload.c +++ b/src/mesa/pipe/i965simple/brw_state_upload.c @@ -43,7 +43,6 @@ */ const struct brw_tracked_state *atoms[] = { - &brw_wm_input_sizes, &brw_vs_prog, &brw_gs_prog, &brw_clip_prog, @@ -84,8 +83,6 @@ const struct brw_tracked_state *atoms[] = &brw_depthbuffer, &brw_polygon_stipple, - &brw_polygon_stipple_offset, - &brw_line_stipple, &brw_psp_urb_cbs, @@ -192,6 +189,10 @@ void brw_validate_state( struct brw_context *brw ) for (i = 0; i < Elements(atoms); i++) { const struct brw_tracked_state *atom = atoms[i]; + assert(atom->dirty.brw || + atom->dirty.cache); + assert(atom->update); + if (check_state(state, &atom->dirty)) atom->update( brw ); } -- cgit v1.2.3 From ae280f1ce90a7d95bf761efdfebc91ae5b379079 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 14:20:23 +0000 Subject: 965: align buffer allocations to 4k --- src/mesa/pipe/xlib/xm_winsys_aub.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index dd7c725507..60ba233f35 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -35,9 +35,9 @@ #include "glxheader.h" #include "xmesaP.h" -#include "main/macros.h" #include "pipe/p_winsys.h" +#include "pipe/p_util.h" #include "pipe/i965simple/brw_winsys.h" #include "brw_aub.h" #include "xm_winsys_aub.h" @@ -159,7 +159,7 @@ static int aub_buffer_data(struct pipe_winsys *winsys, assert(iws->used + size < iws->size); sbo->data = iws->pool + iws->used; sbo->offset = AUB_BUF_START + iws->used; - iws->used += size; + iws->used += align(size, 4096); } sbo->size = size; @@ -235,7 +235,7 @@ void xmesa_commands_aub(struct pipe_winsys *winsys, cmds, nr_dwords * sizeof(int) ); - iws->used += size; + iws->used += align(size, 4096); } -- cgit v1.2.3 From 86ddaa9d0e3c80e0fd7b7c99b66e46e0bcbdb04d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 15:16:25 +0000 Subject: 965: hardwire correct behaviour for vp-tri setup (for now...) --- src/mesa/pipe/i965simple/brw_sf_emit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_sf_emit.c b/src/mesa/pipe/i965simple/brw_sf_emit.c index 834b5efdfe..0fa61f14b6 100644 --- a/src/mesa/pipe/i965simple/brw_sf_emit.c +++ b/src/mesa/pipe/i965simple/brw_sf_emit.c @@ -148,7 +148,7 @@ static boolean calculate_masks( struct brw_sf_compile *c, /* Maybe only processs one attribute on the final round: */ - if (reg*2+1 < c->nr_setup_attrs) { + if (1 || reg*2+1 < c->nr_setup_attrs) { *pc |= 0xf0; // if (persp_mask & (1 << c->idx_to_attr[reg*2+1])) -- cgit v1.2.3 From dc461d8c8764e2bd8303c1955f244bacab4467fa Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 15:48:53 +0000 Subject: 965: count grf allocation correctly for wm prog, first aub triangle --- src/mesa/pipe/i965simple/brw_wm_decl.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_decl.c b/src/mesa/pipe/i965simple/brw_wm_decl.c index 392f17fad6..5b1abb9d48 100644 --- a/src/mesa/pipe/i965simple/brw_wm_decl.c +++ b/src/mesa/pipe/i965simple/brw_wm_decl.c @@ -9,7 +9,7 @@ static struct brw_reg alloc_tmp(struct brw_wm_compile *c) { c->tmp_index++; - c->reg_index = MAX2(c->reg_index, c->tmp_index); + c->reg_index = MAX2(c->reg_index, c->tmp_start + c->tmp_index); return brw_vec8_grf(c->tmp_start + c->tmp_index, 0); } -- cgit v1.2.3 From a65f39f918f3a28c3c4e38036099d4ce97fcfac7 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 19 Dec 2007 13:22:27 -0500 Subject: consts --- src/mesa/pipe/i965simple/brw_curbe.c | 16 ++++++++++------ src/mesa/pipe/i965simple/brw_vs_emit.c | 17 ++++++++++++----- src/mesa/pipe/xlib/xm_winsys_aub.c | 2 +- 3 files changed, 23 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index 66a04b9c38..1bdd870bd6 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -37,6 +37,7 @@ #include "brw_util.h" #include "brw_wm.h" #include "pipe/p_state.h" +#include "pipe/p_winsys.h" #include "pipe/p_util.h" #define FILE_DEBUG_FLAG DEBUG_FALLBACKS @@ -246,13 +247,16 @@ static void upload_constant_buffer(struct brw_context *brw) if (brw->curbe.vs_size) { -// unsigned offset = brw->curbe.vs_start * 16; -// unsigned nr = vp->max_const; - + unsigned offset = brw->curbe.vs_start * 16; + /*unsigned nr = vp->max_const;*/ + const struct pipe_constant_buffer *cbuffer = brw->attribs.Constants[0]; + struct pipe_winsys *ws = brw->pipe.winsys; /* map the vertex constant buffer and copy to curbe: */ - -// assert(nr == 0); - assert(0); + ws->buffer_map(ws, cbuffer->buffer, 0); + ws->buffer_get_subdata(ws, cbuffer->buffer, + 0, cbuffer->size, + &buf[offset]); + ws->buffer_unmap(ws, cbuffer->buffer); } if (0) { diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index d8483f213a..d70df158f9 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -38,6 +38,7 @@ struct brw_prog_info { unsigned num_temps; unsigned num_addrs; + unsigned num_consts; unsigned writes_psize; @@ -74,13 +75,11 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c, /* Vertex program parameters from curbe: */ -#if 0 - nr_params = c->vp->program.num_inputs; /*FIXME: i think this is wrong... */ + nr_params = c->prog_data.max_const; for (i = 0; i < nr_params; i++) { - c->regs[TGSI_FILE_INPUT][i] = stride(brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1); + c->regs[TGSI_FILE_CONSTANT][i] = stride(brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1); } reg += (nr_params+1)/2; -#endif c->prog_data.curb_read_length = reg - 1; @@ -986,6 +985,13 @@ static void process_declaration(const struct tgsi_full_declaration *decl, { switch(decl->Declaration.File) { case TGSI_FILE_CONSTANT: { + if (decl->Declaration.Declare == TGSI_DECLARE_MASK) { + printf("DECLARATION MASK = %d\n", + decl->u.DeclarationMask.Mask); + assert(0); + } else { /*range*/ + info->num_consts += decl->u.DeclarationRange.Last - decl->u.DeclarationRange.First + 1; + } } break; case TGSI_FILE_INPUT: { @@ -1300,12 +1306,12 @@ void brw_vs_emit(struct brw_vs_compile *c) /* first instruction (declerations finished). * now that we know what vars are being used allocate * registers for them.*/ + c->prog_data.max_const = prog_info.num_consts; brw_vs_alloc_regs(c, &prog_info); brw_set_access_mode(p, BRW_ALIGN_1); brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); brw_set_access_mode(p, BRW_ALIGN_16); - allocated_registers = 1; } process_instruction(c, inst, &prog_info); @@ -1318,4 +1324,5 @@ void brw_vs_emit(struct brw_vs_compile *c) emit_vertex_write(c, &prog_info); post_vs_emit(c, end_inst); tgsi_parse_free(&parse); + } diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index 60ba233f35..2d276480cb 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -259,7 +259,7 @@ static int aub_buffer_get_subdata(struct pipe_winsys *winsys, void *data) { struct aub_buffer *sbo = aub_bo(buf); - assert(sbo->size > offset + size); + assert(sbo->size >= offset + size); memcpy(data, sbo->data + offset, size); return 0; } -- cgit v1.2.3 From 5ff69cfd68970ef67ecf3fc10f9abd31356e793f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 19:04:35 +0000 Subject: 965: handle BRW_CONSTANT_BUFFER data type --- src/mesa/pipe/xlib/xm_winsys_aub.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index 2d276480cb..bedfcab40f 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -596,6 +596,14 @@ static void aub_i965_buffer_subdata_typed(struct brw_winsys *winsys, aub_type = DW_SURFACE_STATE; aub_sub_type = DWSS_BINDING_TABLE_STATE; break; + case BRW_CONSTANT_BUFFER: + aub_type = DW_CONSTANT_URB_ENTRY; + aub_sub_type = 0; + break; + + default: + assert(0); + break; } xmesa_buffer_subdata_aub( iws->pipe_winsys, -- cgit v1.2.3 From d24a9b26fc44b586a19d46e731e61a5d133868d9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Dec 2007 19:05:19 +0000 Subject: 965: dump curbe contents to stderr --- src/mesa/pipe/i965simple/brw_curbe.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index 1bdd870bd6..411263f565 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -254,12 +254,13 @@ static void upload_constant_buffer(struct brw_context *brw) /* map the vertex constant buffer and copy to curbe: */ ws->buffer_map(ws, cbuffer->buffer, 0); ws->buffer_get_subdata(ws, cbuffer->buffer, - 0, cbuffer->size, + 0, + cbuffer->size, &buf[offset]); ws->buffer_unmap(ws, cbuffer->buffer); } - if (0) { + if (1) { for (i = 0; i < sz*16; i+=4) _mesa_printf("curbe %d.%d: %f %f %f %f\n", i/8, i&4, buf[i+0], buf[i+1], buf[i+2], buf[i+3]); -- cgit v1.2.3 From c61b32777b5b3b138c8bf5fac7e7e70ec084c8c1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Dec 2007 08:49:59 -0700 Subject: new assertions --- src/mesa/state_tracker/st_program.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 6802bb3b06..1852228b29 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -167,6 +167,7 @@ st_translate_vertex_program(struct st_context *st, switch (attr) { case VERT_RESULT_HPOS: + assert(slot == 0); vs.output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; vs.output_semantic_index[slot] = 0; break; @@ -218,6 +219,8 @@ st_translate_vertex_program(struct st_context *st, } } + assert(vs.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); + if (outputMapping) { /* find max output slot referenced to compute vs.num_outputs */ -- cgit v1.2.3 From dbf6eced87c16eae0834d38d86a60b4f643ee3b7 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Dec 2007 08:50:35 -0700 Subject: move st_make_current() before buffer size check so renderbuffer alloc storage works --- src/mesa/pipe/xlib/xm_api.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 142074bc65..186a712b52 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1067,12 +1067,12 @@ GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, */ _glapi_check_multithread(); + st_make_current(c->st, drawBuffer->stfb, readBuffer->stfb); + xmesa_check_and_update_buffer_size(c, drawBuffer); if (readBuffer != drawBuffer) xmesa_check_and_update_buffer_size(c, readBuffer); - st_make_current(c->st, drawBuffer->stfb, readBuffer->stfb); - /* Solution to Stephane Rehel's problem with glXReleaseBuffersMESA(): */ drawBuffer->wasCurrent = GL_TRUE; } -- cgit v1.2.3 From cedf6892e7cc29aab404b142012eec0b0603a414 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Dec 2007 08:50:52 -0700 Subject: special-case PSIZE too --- src/mesa/state_tracker/st_atom_shader.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 33372b0f39..2a182c7d9c 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -228,12 +228,17 @@ find_translated_vp(struct st_context *st, xvp->output_to_slot[outAttr] = vpOutSlot; numVpOuts++; } - else if (outAttr == VERT_RESULT_BFC0 || + else if (outAttr == VERT_RESULT_PSIZ || + outAttr == VERT_RESULT_BFC0 || outAttr == VERT_RESULT_BFC1) { /* backface colors go into last slots */ xvp->output_to_slot[outAttr] = numVpOuts++; } } + /* + printf("output_to_slot[%d] = %d\n", outAttr, + xvp->output_to_slot[outAttr]); + */ } /* Unneeded vertex program outputs will go to this slot. -- cgit v1.2.3 From 127ab73b380d9c26974d19062e3e16758005bd80 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Dec 2007 08:51:17 -0700 Subject: remove obsolete TXP, add some sanity checks --- src/mesa/pipe/tgsi/util/tgsi_dump.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.c b/src/mesa/pipe/tgsi/util/tgsi_dump.c index 3f4d930dec..cdbc0dbc9c 100644 --- a/src/mesa/pipe/tgsi/util/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/util/tgsi_dump.c @@ -453,7 +453,6 @@ static const char *TGSI_OPCODES[] = "OPCODE_CALLNZ", "OPCODE_IFC", "OPCODE_BREAKC", - "OPCODE_TXP", "OPCODE_KIL", "OPCODE_END" }; @@ -592,7 +591,6 @@ static const char *TGSI_OPCODES_SHORT[] = "CALLNZ", "IFC", "BREAKC", - "TXP", "KIL", "END" }; @@ -1518,6 +1516,15 @@ dump_gen( tgsi_parse_free( &parse ); } + +static void +sanity_checks(void) +{ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); + assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "END") == 0); +} + + void tgsi_dump( const struct tgsi_token *tokens, @@ -1525,6 +1532,8 @@ tgsi_dump( { struct file_dump dump; + sanity_checks(); + dump.base.write = _file_dump_write; #if 0 { -- cgit v1.2.3 From 1575763a6f57d1f13c707b709f188b0617c8955a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Dec 2007 11:50:50 -0700 Subject: convert Mesa OPCODE_END to TGSI_OPCODE_END, not TGSI_OPCOD_RET --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index c8b76978f1..470cef8e85 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -499,7 +499,7 @@ compile_instruction( fulldst->DstRegister.WriteMask &= TGSI_WRITEMASK_XYZ; break; case OPCODE_END: - fullinst->Instruction.Opcode = TGSI_OPCODE_RET; + fullinst->Instruction.Opcode = TGSI_OPCODE_END; break; default: assert( 0 ); -- cgit v1.2.3 From c664302c3e34a29b4bbb02fd3789dd3f7d92849c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Dec 2007 13:45:00 -0700 Subject: Fix problem with initial viewport/scissor size. If an app never called glViewport, the viewport size was always 0 by 0 pixels. Now pass initial size to st_create_framebuffer() and initialize the viewport and scissor bounds in st_make_current(). This could also be fixed by ensuring the gl_framebuffers passed to _mesa_make_current() were initialized to the right size. But that involves allocating the renderbuffers/pipe_surfaces earlier and that runs into some other issues ATM. Also remove obsolete createRenderbuffers param to st_create_framebuffer(). --- src/mesa/drivers/dri/intel_winsys/intel_screen.c | 4 +- src/mesa/pipe/xlib/xm_api.c | 8 +- src/mesa/state_tracker/st_context.c | 14 +++ src/mesa/state_tracker/st_context.h | 1 + src/mesa/state_tracker/st_framebuffer.c | 109 +++++++++++------------ src/mesa/state_tracker/st_public.h | 2 +- 6 files changed, 79 insertions(+), 59 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c index bce6c5699a..9e31c013a9 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_screen.c @@ -297,10 +297,12 @@ intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, else stencilFormat = PIPE_FORMAT_NONE; - intelfb->stfb = st_create_framebuffer(visual, GL_TRUE, + intelfb->stfb = st_create_framebuffer(visual, colorFormat, depthFormat, stencilFormat, + driDrawPriv->w, + driDrawPriv->h, (void*) intelfb); if (!intelfb->stfb) { free(intelfb); diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 186a712b52..ebf4c21eaf 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -207,7 +207,7 @@ static GLboolean window_exists( XMesaDisplay *dpy, Window win ) } static Status -get_drawable_size( XMesaDisplay *dpy, Drawable d, GLuint *width, GLuint *height ) +get_drawable_size( XMesaDisplay *dpy, Drawable d, uint *width, uint *height ) { Window root; Status stat; @@ -323,6 +323,7 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, XMesaBuffer b; GLframebuffer *fb; enum pipe_format colorFormat, depthFormat, stencilFormat; + uint width, height; ASSERT(type == WINDOW || type == PIXMAP || type == PBUFFER); @@ -359,11 +360,14 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, } + get_drawable_size(vis->display, d, &width, &height); + /* * Create framebuffer, but we'll plug in our own renderbuffers below. */ - b->stfb = st_create_framebuffer(&vis->mesa_visual, GL_TRUE, + b->stfb = st_create_framebuffer(&vis->mesa_visual, colorFormat, depthFormat, stencilFormat, + width, height, (void *) b); fb = &b->stfb->Base; diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 524e06fb00..1d26da474e 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -28,6 +28,8 @@ #include "main/imports.h" #include "main/context.h" #include "main/extensions.h" +#include "main/matrix.h" +#include "main/buffers.h" #include "vbo/vbo.h" #include "shader/shader_api.h" #include "st_public.h" @@ -163,7 +165,19 @@ void st_make_current(struct st_context *st, struct st_framebuffer *read) { if (st) { + GLboolean firstTime = st->ctx->FirstTimeCurrent; _mesa_make_current(st->ctx, &draw->Base, &read->Base); + /* Need to initialize viewport here since draw->Base->Width/Height + * will still be zero at this point. + * This could be improved, but would require rather extensive work + * elsewhere (allocate rb surface storage sooner) + */ + if (firstTime) { + GLuint w = draw->InitWidth, h = draw->InitHeight; + _mesa_set_viewport(st->ctx, 0, 0, w, h); + _mesa_set_scissor(st->ctx, 0, 0, w, h); + + } } else { _mesa_make_current(NULL, NULL, NULL); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index c3919d474c..0f40f3ceee 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -191,6 +191,7 @@ struct st_framebuffer { GLframebuffer Base; void *Private; + GLuint InitWidth, InitHeight; }; diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index b81a894ef1..5e0943f75c 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -37,80 +37,79 @@ struct st_framebuffer * st_create_framebuffer( const __GLcontextModes *visual, - boolean createRenderbuffers, /* XXX remove? */ enum pipe_format colorFormat, enum pipe_format depthFormat, enum pipe_format stencilFormat, + uint width, uint height, void *private) { struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer); if (stfb) { _mesa_initialize_framebuffer(&stfb->Base, visual); - if (createRenderbuffers) { - { - /* fake frontbuffer */ - /* XXX allocation should only happen in the unusual case - it's actually needed */ - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(colorFormat); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb); - } + { + /* fake frontbuffer */ + /* XXX allocation should only happen in the unusual case + it's actually needed */ + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(colorFormat); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb); + } - if (visual->doubleBufferMode) { - struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(colorFormat); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); - } + if (visual->doubleBufferMode) { + struct gl_renderbuffer *rb + = st_new_renderbuffer_fb(colorFormat); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); + } + + if (visual->depthBits == 24 && visual->stencilBits == 8) { + /* combined depth/stencil buffer */ + struct gl_renderbuffer *depthStencilRb + = st_new_renderbuffer_fb(depthFormat); + /* note: bind RB to two attachment points */ + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb); + } + else { + /* separate depth and/or stencil */ - if (visual->depthBits == 24 && visual->stencilBits == 8) { - /* combined depth/stencil buffer */ - struct gl_renderbuffer *depthStencilRb + if (visual->depthBits == 32) { + /* 32-bit depth buffer */ + struct gl_renderbuffer *depthRb = st_new_renderbuffer_fb(depthFormat); - /* note: bind RB to two attachment points */ - _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } - else { - /* separate depth and/or stencil */ - - if (visual->depthBits == 32) { - /* 32-bit depth buffer */ - struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(depthFormat); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); - } - else if (visual->depthBits == 24) { - /* 24-bit depth buffer, ignore stencil bits */ - struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(depthFormat); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); - } - else if (visual->depthBits > 0) { - /* 16-bit depth buffer */ - struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(depthFormat); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); - } - - if (visual->stencilBits > 0) { - /* 8-bit stencil */ - struct gl_renderbuffer *stencilRb - = st_new_renderbuffer_fb(stencilFormat); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb); - } + else if (visual->depthBits == 24) { + /* 24-bit depth buffer, ignore stencil bits */ + struct gl_renderbuffer *depthRb + = st_new_renderbuffer_fb(depthFormat); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); + } + else if (visual->depthBits > 0) { + /* 16-bit depth buffer */ + struct gl_renderbuffer *depthRb + = st_new_renderbuffer_fb(depthFormat); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } - if (visual->accumRedBits > 0) { - /* 16-bit/channel accum */ - struct gl_renderbuffer *accumRb - = st_new_renderbuffer_fb(PIPE_FORMAT_R16G16B16A16_SNORM); - _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb); + if (visual->stencilBits > 0) { + /* 8-bit stencil */ + struct gl_renderbuffer *stencilRb + = st_new_renderbuffer_fb(stencilFormat); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb); } } - stfb->Base.Initialized = GL_TRUE; + if (visual->accumRedBits > 0) { + /* 16-bit/channel accum */ + struct gl_renderbuffer *accumRb + = st_new_renderbuffer_fb(PIPE_FORMAT_R16G16B16A16_SNORM); + _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb); + } + stfb->Base.Initialized = GL_TRUE; + stfb->InitWidth = width; + stfb->InitHeight = height; stfb->Private = private; } return stfb; diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 558b20f1e8..ed5ef1f159 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -55,10 +55,10 @@ void st_copy_context_state(struct st_context *dst, struct st_context *src, uint mask); struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, - boolean createRenderbuffers, enum pipe_format colorFormat, enum pipe_format depthFormat, enum pipe_format stencilFormat, + uint width, uint height, void *privateData); void st_resize_framebuffer( struct st_framebuffer *stfb, -- cgit v1.2.3 From 4a1776a763f2c3b3d1d5bdb5720cafad3ac932c6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Dec 2007 13:53:28 -0700 Subject: temporarily defeat an assertion --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 1f43f3643e..e469281d22 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -2415,6 +2415,7 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) mach->FuncMask = 0xf; mach->ExecMask = 0xf; + mach->CondStackTop = 0; /* temporarily subvert this assertion */ assert(mach->CondStackTop == 0); assert(mach->LoopStackTop == 0); assert(mach->ContStackTop == 0); -- cgit v1.2.3 From e9207430cefd44beb1a16cfb5879a7ace475e2f5 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Dec 2007 14:06:22 -0700 Subject: Add some prototype code for converting RET to END for main(). Disabled for now. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 470cef8e85..01dc70b66e 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -180,7 +180,8 @@ compile_instruction( const GLuint outputMapping[], const GLuint immediateMapping[], GLuint preamble_size, - GLuint processor ) + GLuint processor, + GLboolean *insideSubroutine) { GLuint i; struct tgsi_full_dst_register *fulldst; @@ -283,6 +284,7 @@ compile_instruction( break; case OPCODE_BGNSUB: fullinst->Instruction.Opcode = TGSI_OPCODE_BGNSUB; + *insideSubroutine = GL_TRUE; break; case OPCODE_BRA: fullinst->Instruction.Opcode = TGSI_OPCODE_BRA; @@ -334,6 +336,7 @@ compile_instruction( break; case OPCODE_ENDSUB: fullinst->Instruction.Opcode = TGSI_OPCODE_ENDSUB; + *insideSubroutine = GL_FALSE; break; case OPCODE_EX2: fullinst->Instruction.Opcode = TGSI_OPCODE_EX2; @@ -412,7 +415,16 @@ compile_instruction( fullinst->Instruction.Opcode = TGSI_OPCODE_RCP; break; case OPCODE_RET: - fullinst->Instruction.Opcode = TGSI_OPCODE_RET; + /* If RET is used inside main (not a real subroutine) we may want + * to execute END instead of RET. TBD... + */ + if (1 /* *insideSubroutine */) { + fullinst->Instruction.Opcode = TGSI_OPCODE_RET; + } + else { + /* inside main() pseudo-function */ + fullinst->Instruction.Opcode = TGSI_OPCODE_END; + } break; case OPCODE_RSQ: fullinst->Instruction.Opcode = TGSI_OPCODE_RSQ; @@ -682,6 +694,7 @@ tgsi_translate_mesa_program( GLuint preamble_size = 0; GLuint immediates[1000]; GLuint numImmediates = 0; + GLboolean insideSubroutine = GL_FALSE; assert(procType == TGSI_PROCESSOR_FRAGMENT || procType == TGSI_PROCESSOR_VERTEX); @@ -879,7 +892,8 @@ tgsi_translate_mesa_program( outputMapping, immediates, preamble_size, - procType ); + procType, + &insideSubroutine); ti += tgsi_build_full_instruction( &fullinst, -- cgit v1.2.3 From ebf78c0dcccbe2c458b945e014f2bd53ab137e91 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 20 Dec 2007 07:05:52 -0500 Subject: 965: fix the constant buffers --- src/mesa/pipe/i965simple/brw_curbe.c | 4 ++-- src/mesa/pipe/xlib/xm_winsys_aub.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index 411263f565..ee20e00362 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -175,7 +175,7 @@ static void upload_constant_buffer(struct brw_context *brw) { struct brw_mem_pool *pool = &brw->pool[BRW_GS_POOL]; unsigned sz = brw->curbe.total_size; - unsigned bufsz = sz * 16 * sizeof(float); + unsigned bufsz = sz * sizeof(float); float *buf; unsigned i; @@ -261,7 +261,7 @@ static void upload_constant_buffer(struct brw_context *brw) } if (1) { - for (i = 0; i < sz*16; i+=4) + for (i = 0; i < sz; i+=4) _mesa_printf("curbe %d.%d: %f %f %f %f\n", i/8, i&4, buf[i+0], buf[i+1], buf[i+2], buf[i+3]); diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index bedfcab40f..b207638390 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -165,7 +165,7 @@ static int aub_buffer_data(struct pipe_winsys *winsys, sbo->size = size; if (data != NULL) { - memcpy(iws->pool, data, size); + memcpy(sbo->data, data, size); brw_aub_gtt_data( iws->aubfile, sbo->offset, -- cgit v1.2.3 From 21c67b70d4bcffad8f0f07c0423c18e59f4259ea Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 20 Dec 2007 13:19:56 +0000 Subject: gallium: translate ARB fp/vp immediates consistently to tgsi immediates --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 24 ++++++++++++++++++++---- 1 file changed, 20 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 01dc70b66e..325aa20173 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -47,7 +47,9 @@ */ static GLuint map_register_file( - enum register_file file ) + enum register_file file, + GLuint index, + const GLuint immediateMapping[] ) { switch( file ) { case PROGRAM_UNDEFINED: @@ -56,10 +58,20 @@ map_register_file( return TGSI_FILE_TEMPORARY; //case PROGRAM_LOCAL_PARAM: //case PROGRAM_ENV_PARAM: + + /* Because of the longstanding problem with mesa arb shaders + * where constants, immediates and state variables are all + * bundled together as PROGRAM_STATE_VAR, we can't tell from the + * mesa register file whether this is a CONSTANT or an + * IMMEDIATE, hence we need all the other information. + */ case PROGRAM_STATE_VAR: case PROGRAM_NAMED_PARAM: case PROGRAM_UNIFORM: - return TGSI_FILE_CONSTANT; + if (immediateMapping[index] != ~0) + return TGSI_FILE_IMMEDIATE; + else + return TGSI_FILE_CONSTANT; case PROGRAM_CONSTANT: return TGSI_FILE_IMMEDIATE; case PROGRAM_INPUT: @@ -194,7 +206,7 @@ compile_instruction( fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode ); fulldst = &fullinst->FullDstRegisters[0]; - fulldst->DstRegister.File = map_register_file( inst->DstReg.File ); + fulldst->DstRegister.File = map_register_file( inst->DstReg.File, 0, NULL ); fulldst->DstRegister.Index = map_register_file_index( fulldst->DstRegister.File, inst->DstReg.Index, @@ -208,7 +220,9 @@ compile_instruction( GLuint j; fullsrc = &fullinst->FullSrcRegisters[i]; - fullsrc->SrcRegister.File = map_register_file( inst->SrcReg[i].File ); + fullsrc->SrcRegister.File = map_register_file( inst->SrcReg[i].File, + inst->SrcReg[i].Index, + immediateMapping ); fullsrc->SrcRegister.Index = map_register_file_index( fullsrc->SrcRegister.File, inst->SrcReg[i].Index, @@ -813,6 +827,8 @@ tgsi_translate_mesa_program( } /* immediates/literals */ + memset(immediates, ~0, sizeof(immediates)); + for (i = 0; program->Parameters && i < program->Parameters->NumParameters; i++) { if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { -- cgit v1.2.3 From 9e41d547db6669ff669f1d60cb35df9edf306370 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 20 Dec 2007 13:47:11 +0000 Subject: 965: respect pipe flush flags Now we emit way too many flushes instead of none at all. --- src/mesa/pipe/i965simple/brw_batch.h | 2 ++ src/mesa/pipe/i965simple/brw_curbe.c | 1 + src/mesa/pipe/i965simple/brw_flush.c | 20 ++++++++------------ src/mesa/pipe/i965simple/brw_state.h | 1 - src/mesa/pipe/i965simple/brw_urb.c | 3 ++- 5 files changed, 13 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_batch.h b/src/mesa/pipe/i965simple/brw_batch.h index bef69ac871..8605d7c108 100644 --- a/src/mesa/pipe/i965simple/brw_batch.h +++ b/src/mesa/pipe/i965simple/brw_batch.h @@ -53,4 +53,6 @@ brw->hardware_dirty = ~0; \ } while (0) +#define BRW_BATCH_STRUCT(brw, s) brw_batchbuffer_data( brw->winsys, (s), sizeof(*(s))) + #endif diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index ee20e00362..785affa2d8 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -34,6 +34,7 @@ #include "brw_context.h" #include "brw_defines.h" #include "brw_state.h" +#include "brw_batch.h" #include "brw_util.h" #include "brw_wm.h" #include "pipe/p_state.h" diff --git a/src/mesa/pipe/i965simple/brw_flush.c b/src/mesa/pipe/i965simple/brw_flush.c index 1f11c5f164..5216c680cf 100644 --- a/src/mesa/pipe/i965simple/brw_flush.c +++ b/src/mesa/pipe/i965simple/brw_flush.c @@ -49,22 +49,18 @@ static void brw_flush( struct pipe_context *pipe, * caches? */ if (flags & (PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE)) { - unsigned flush = CMD_MI_FLUSH; - -#if 0 + struct brw_mi_flush flush; + + memset(&flush, 0, sizeof(flush)); + flush.opcode = CMD_MI_FLUSH; + if (!(flags & PIPE_FLUSH_RENDER_CACHE)) - flush |= INHIBIT_FLUSH_RENDER_CACHE; + flush.flags |= BRW_INHIBIT_FLUSH_RENDER_CACHE; if (flags & PIPE_FLUSH_TEXTURE_CACHE) - flush |= FLUSH_MAP_CACHE; -#endif + flush.flags |= BRW_FLUSH_READ_CACHE; - if (!BEGIN_BATCH(1, 0)) { - FLUSH_BATCH( &fence ); - assert(BEGIN_BATCH(1, 0)); - } - OUT_BATCH( flush ); - ADVANCE_BATCH(); + BRW_BATCH_STRUCT(brw, &flush); } /* If there are no flags, just flush pending commands to hardware: diff --git a/src/mesa/pipe/i965simple/brw_state.h b/src/mesa/pipe/i965simple/brw_state.h index 62055e5506..8b4d7aabf0 100644 --- a/src/mesa/pipe/i965simple/brw_state.h +++ b/src/mesa/pipe/i965simple/brw_state.h @@ -115,7 +115,6 @@ static inline struct pipe_buffer_handle *brw_cache_buffer(struct brw_context *br /*********************************************************************** * brw_state_batch.c */ -#define BRW_BATCH_STRUCT(brw, s) brw_batchbuffer_data( brw->winsys, (s), sizeof(*(s))) #define BRW_CACHED_BATCH_STRUCT(brw, s) brw_cached_batch_struct( brw, (s), sizeof(*(s)) ) boolean brw_cached_batch_struct( struct brw_context *brw, diff --git a/src/mesa/pipe/i965simple/brw_urb.c b/src/mesa/pipe/i965simple/brw_urb.c index 8cc86e26a3..b284526aa6 100644 --- a/src/mesa/pipe/i965simple/brw_urb.c +++ b/src/mesa/pipe/i965simple/brw_urb.c @@ -31,7 +31,8 @@ #include "brw_context.h" -#include "brw_state.h" +//#include "brw_state.h" +#include "brw_batch.h" #include "brw_defines.h" #define VS 0 -- cgit v1.2.3 From a85535b7cb0886f23e5686e37d1fa54394cdece4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 20 Dec 2007 13:47:46 +0000 Subject: gallium: make state tracker explictly ask for rendercache flushes --- src/mesa/state_tracker/st_cb_accum.c | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- src/mesa/state_tracker/st_cb_fbo.c | 2 +- src/mesa/state_tracker/st_cb_flush.c | 4 ++-- src/mesa/state_tracker/st_cb_readpixels.c | 2 +- src/mesa/state_tracker/st_framebuffer.c | 3 ++- 6 files changed, 9 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index ea0b1187fc..cf2e9db51c 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -321,7 +321,7 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value) const GLint height = ctx->DrawBuffer->_Ymax - ypos; /* make sure color bufs aren't cached */ - pipe->flush(pipe, 0); + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); switch (op) { case GL_ADD: diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index e70a5b49e1..2db12c653b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -849,7 +849,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, GLint skipPixels; ubyte *stmap; - pipe->flush(pipe, 0); + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); /* map the stencil buffer */ stmap = pipe_surface_map(ps); @@ -1208,7 +1208,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, uint format; /* make sure rendering has completed */ - pipe->flush(pipe, 0x0); + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); st_validate_state(st); diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index c40f75417f..bc0f26ffb0 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -365,7 +365,7 @@ st_finish_render_texture(GLcontext *ctx, assert(strb); - ctx->st->pipe->flush(ctx->st->pipe, 0x0); + ctx->st->pipe->flush(ctx->st->pipe, PIPE_FLUSH_RENDER_CACHE); /* printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface); diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 95149a3200..9808b1f8f6 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -83,7 +83,7 @@ void st_flush( struct st_context *st, uint pipeFlushFlags ) */ static void st_Flush(GLcontext *ctx) { - st_flush(ctx->st, 0x0); + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE); } @@ -92,7 +92,7 @@ static void st_Flush(GLcontext *ctx) */ static void st_Finish(GLcontext *ctx) { - st_flush(ctx->st, PIPE_FLUSH_WAIT); + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_WAIT); } diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 96829fcfa0..b0c9275c6f 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -155,7 +155,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } /* make sure rendering has completed */ - pipe->flush(pipe, 0x0); + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); if (pack->BufferObj && pack->BufferObj->Name) { /* reading into a PBO */ diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 5e0943f75c..7ddc74e355 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -33,6 +33,7 @@ #include "st_public.h" #include "st_context.h" #include "st_cb_fbo.h" +#include "pipe/p_defines.h" struct st_framebuffer * @@ -170,7 +171,7 @@ st_notify_swapbuffers(struct st_framebuffer *stfb) GET_CURRENT_CONTEXT(ctx); if (ctx && ctx->DrawBuffer == &stfb->Base) { - st_flush(ctx->st, 0x0); + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE); } } -- cgit v1.2.3 From 4fa7afabc966a3d37324f2f9b03e1cc466db7773 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 20 Dec 2007 12:54:23 -0500 Subject: i965: very crude and hacky way of handling immediates --- src/mesa/pipe/i965simple/brw_context.h | 3 +++ src/mesa/pipe/i965simple/brw_curbe.c | 24 +++++++++++++++++------- src/mesa/pipe/i965simple/brw_vs_emit.c | 15 +++++++++++++-- 3 files changed, 33 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h index 59e3d24a1d..139f13ad4e 100644 --- a/src/mesa/pipe/i965simple/brw_context.h +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -327,6 +327,9 @@ struct brw_vs_prog_data { unsigned max_const; + float imm_buf[PIPE_MAX_CONSTANT][4]; + unsigned num_imm; + /* Used for calculating urb partitions: */ unsigned urb_entry_size; diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index 785affa2d8..603116c863 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -252,13 +252,23 @@ static void upload_constant_buffer(struct brw_context *brw) /*unsigned nr = vp->max_const;*/ const struct pipe_constant_buffer *cbuffer = brw->attribs.Constants[0]; struct pipe_winsys *ws = brw->pipe.winsys; - /* map the vertex constant buffer and copy to curbe: */ - ws->buffer_map(ws, cbuffer->buffer, 0); - ws->buffer_get_subdata(ws, cbuffer->buffer, - 0, - cbuffer->size, - &buf[offset]); - ws->buffer_unmap(ws, cbuffer->buffer); + if (cbuffer->size) { + /* map the vertex constant buffer and copy to curbe: */ + ws->buffer_map(ws, cbuffer->buffer, 0); + ws->buffer_get_subdata(ws, cbuffer->buffer, + 0, + cbuffer->size, + &buf[offset]); + ws->buffer_unmap(ws, cbuffer->buffer); + offset += cbuffer->size; + } + /*immediates*/ +#if 0 + if (brw->vs.prog_data->num_imm) { + memcpy(&buf[offset], brw->vs.prog_data->imm_buf, + brw->vs.prog_data->num_imm * 4 * sizeof(float)); + } +#endif } if (1) { diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index d70df158f9..f3507f60f7 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -627,9 +627,12 @@ static struct brw_reg get_reg( struct brw_vs_compile *c, case TGSI_FILE_TEMPORARY: case TGSI_FILE_INPUT: case TGSI_FILE_OUTPUT: - case TGSI_FILE_CONSTANT: assert(c->regs[file][index].nr != 0); return c->regs[file][index]; + case TGSI_FILE_CONSTANT: + case TGSI_FILE_IMMEDIATE: + assert(c->regs[TGSI_FILE_CONSTANT][index].nr != 0); + return c->regs[TGSI_FILE_CONSTANT][index]; case TGSI_FILE_ADDRESS: assert(index == 0); return c->regs[file][index]; @@ -1298,6 +1301,14 @@ void brw_vs_emit(struct brw_vs_compile *c) } break; case TGSI_TOKEN_TYPE_IMMEDIATE: { + int i; + struct tgsi_full_immediate *imm = &parse.FullToken.FullImmediate; + /*assert(imm->Immediate.Size == 4);*/ + c->prog_data.imm_buf[c->prog_data.num_imm][0] = imm->u.ImmediateFloat32[0].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][1] = imm->u.ImmediateFloat32[1].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][2] = imm->u.ImmediateFloat32[2].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][3] = imm->u.ImmediateFloat32[3].Float; + c->prog_data.num_imm++; } break; case TGSI_TOKEN_TYPE_INSTRUCTION: { @@ -1306,7 +1317,7 @@ void brw_vs_emit(struct brw_vs_compile *c) /* first instruction (declerations finished). * now that we know what vars are being used allocate * registers for them.*/ - c->prog_data.max_const = prog_info.num_consts; + c->prog_data.max_const = prog_info.num_consts + c->prog_data.num_imm; brw_vs_alloc_regs(c, &prog_info); brw_set_access_mode(p, BRW_ALIGN_1); -- cgit v1.2.3 From f9e0e2b3efbc45a290f0a84a1beb63cd8c4f8428 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 24 Dec 2007 07:57:34 -0500 Subject: i965: a little better way of handling immediates --- src/mesa/pipe/i965simple/brw_context.h | 1 + src/mesa/pipe/i965simple/brw_curbe.c | 9 ++++++--- src/mesa/pipe/i965simple/brw_vs_emit.c | 3 +++ 3 files changed, 10 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h index 139f13ad4e..c610a435e0 100644 --- a/src/mesa/pipe/i965simple/brw_context.h +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -329,6 +329,7 @@ struct brw_vs_prog_data { float imm_buf[PIPE_MAX_CONSTANT][4]; unsigned num_imm; + unsigned num_consts; /* Used for calculating urb partitions: */ diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index 603116c863..4d79a7abe2 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -252,9 +252,14 @@ static void upload_constant_buffer(struct brw_context *brw) /*unsigned nr = vp->max_const;*/ const struct pipe_constant_buffer *cbuffer = brw->attribs.Constants[0]; struct pipe_winsys *ws = brw->pipe.winsys; - if (cbuffer->size) { + /* FIXME: buffer size is num_consts + num_immediates */ + if (brw->vs.prog_data->num_consts) { /* map the vertex constant buffer and copy to curbe: */ ws->buffer_map(ws, cbuffer->buffer, 0); + /* FIXME: this is wrong. the cbuffer->size currently + * represents size of consts + immediates. so if we'll + * have both we'll copy over the end of the buffer + * with the subsequent memcpy */ ws->buffer_get_subdata(ws, cbuffer->buffer, 0, cbuffer->size, @@ -263,12 +268,10 @@ static void upload_constant_buffer(struct brw_context *brw) offset += cbuffer->size; } /*immediates*/ -#if 0 if (brw->vs.prog_data->num_imm) { memcpy(&buf[offset], brw->vs.prog_data->imm_buf, brw->vs.prog_data->num_imm * 4 * sizeof(float)); } -#endif } if (1) { diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index f3507f60f7..0cc0a437b0 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -630,6 +630,8 @@ static struct brw_reg get_reg( struct brw_vs_compile *c, assert(c->regs[file][index].nr != 0); return c->regs[file][index]; case TGSI_FILE_CONSTANT: + assert(c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_consts].nr != 0); + return c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_consts]; case TGSI_FILE_IMMEDIATE: assert(c->regs[TGSI_FILE_CONSTANT][index].nr != 0); return c->regs[TGSI_FILE_CONSTANT][index]; @@ -1317,6 +1319,7 @@ void brw_vs_emit(struct brw_vs_compile *c) /* first instruction (declerations finished). * now that we know what vars are being used allocate * registers for them.*/ + c->prog_data.num_consts = prog_info.num_consts; c->prog_data.max_const = prog_info.num_consts + c->prog_data.num_imm; brw_vs_alloc_regs(c, &prog_info); -- cgit v1.2.3 From 552907d8a497d42f6693ca0f9324f003cfe3a66d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Dec 2007 17:37:30 -0700 Subject: free Default1D/2DArray objects --- src/mesa/main/context.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 563a89e686..a9f9bd9da4 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -554,8 +554,13 @@ alloc_shared_state( GLcontext *ctx ) (*ctx->Driver.DeleteTexture)(ctx, ss->DefaultCubeMap); if (ss->DefaultRect) (*ctx->Driver.DeleteTexture)(ctx, ss->DefaultRect); - if (ss) - _mesa_free(ss); + if (ss->Default1DArray) + (*ctx->Driver.DeleteTexture)(ctx, ss->Default1DArray); + if (ss->Default2DArray) + (*ctx->Driver.DeleteTexture)(ctx, ss->Default2DArray); + + _mesa_free(ss); + return GL_FALSE; } @@ -678,6 +683,9 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) ctx->Driver.DeleteTexture(ctx, ss->Default3D); ctx->Driver.DeleteTexture(ctx, ss->DefaultCubeMap); ctx->Driver.DeleteTexture(ctx, ss->DefaultRect); + ctx->Driver.DeleteTexture(ctx, ss->Default1DArray); + ctx->Driver.DeleteTexture(ctx, ss->Default2DArray); + /* all other textures */ _mesa_HashDeleteAll(ss->TexObjects, delete_texture_cb, ctx); _mesa_DeleteHashTable(ss->TexObjects); -- cgit v1.2.3 From eba2e044a04da99fea2214ba10ca5981f596702e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Dec 2007 17:37:59 -0700 Subject: added tgsi_exec_machine_free_data() --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 16 ++++++++++++++++ src/mesa/pipe/tgsi/exec/tgsi_exec.h | 5 +++++ 2 files changed, 21 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index e469281d22..49affc12f3 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -275,6 +275,22 @@ tgsi_exec_machine_init( } +void +tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach) +{ + if (mach->Instructions) { + FREE(mach->Instructions); + mach->Instructions = NULL; + mach->NumInstructions = 0; + } + if (mach->Declarations) { + FREE(mach->Declarations); + mach->Declarations = NULL; + mach->NumDeclarations = 0; + } +} + + static void micro_abs( union tgsi_exec_channel *dst, diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index db92e282df..1fb66ee960 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -227,6 +227,11 @@ uint tgsi_exec_machine_run( struct tgsi_exec_machine *mach ); + +void +tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach); + + #if defined __cplusplus } /* extern "C" */ #endif -- cgit v1.2.3 From f6136f993a2f06e1d4719b3e9a76187222e9a927 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Dec 2007 17:38:35 -0700 Subject: free tgsi machine state --- src/mesa/pipe/draw/draw_context.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 179f7ed0d6..a9ff54404f 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -102,6 +102,7 @@ void draw_destroy( struct draw_context *draw ) draw->pipeline.validate->destroy( draw->pipeline.validate ); if (draw->pipeline.rasterize) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); + tgsi_exec_machine_free_data(&draw->machine); FREE( draw->vcache.vertex[0] ); /* Frees all the vertices. */ FREE( draw ); } -- cgit v1.2.3 From 57a5421d658df0ef330fc2c9c34ff9fa0923867e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Dec 2007 17:39:01 -0700 Subject: free tgsi machine state --- src/mesa/pipe/softpipe/sp_quad_fs.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 6e7e7eb074..921dfbaccb 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -335,6 +335,7 @@ static void shade_destroy(struct quad_stage *qs) { struct quad_shade_stage *qss = (struct quad_shade_stage *) qs; + tgsi_exec_machine_free_data(&qss->machine); FREE( qss->inputs ); FREE( qss->outputs ); FREE( qs ); -- cgit v1.2.3 From aec6009ece010f334f5777d727620cc042746e31 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Dec 2007 17:39:21 -0700 Subject: fix a memleak --- src/mesa/shader/prog_cache.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/shader/prog_cache.c b/src/mesa/shader/prog_cache.c index d97e27dd6b..dd0241ef24 100644 --- a/src/mesa/shader/prog_cache.c +++ b/src/mesa/shader/prog_cache.c @@ -143,6 +143,7 @@ _mesa_delete_program_cache(GLcontext *ctx, struct gl_program_cache *cache) { clear_cache(ctx, cache); _mesa_free(cache->items); + _mesa_free(cache); } -- cgit v1.2.3 From 171a9674c4caf32303a8d09ce48db18cde1d3e7e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Dec 2007 17:40:02 -0700 Subject: free surface caches in softpipe_destroy() --- src/mesa/pipe/softpipe/sp_context.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index b6995b8a6c..52672a9b29 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -127,6 +127,7 @@ softpipe_unmap_surfaces(struct softpipe_context *sp) static void softpipe_destroy( struct pipe_context *pipe ) { struct softpipe_context *softpipe = softpipe_context( pipe ); + uint i; draw_destroy( softpipe->draw ); @@ -143,6 +144,14 @@ static void softpipe_destroy( struct pipe_context *pipe ) softpipe->quad.colormask->destroy( softpipe->quad.colormask ); softpipe->quad.output->destroy( softpipe->quad.output ); + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + sp_destroy_tile_cache(softpipe->cbuf_cache[i]); + sp_destroy_tile_cache(softpipe->zbuf_cache); + sp_destroy_tile_cache(softpipe->sbuf_cache_sep); + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + sp_destroy_tile_cache(softpipe->tex_cache[i]); + FREE( softpipe ); } -- cgit v1.2.3 From 573b4414b926bb86e9a1e8f3ffad64426aa4bda4 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Dec 2007 06:56:42 -0700 Subject: fix mem leak (free key) --- src/mesa/main/ffvertex_prog.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 8fcb9e5b14..4a9a0cd975 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -1529,6 +1529,8 @@ _mesa_get_fixed_func_vertex_program(GLcontext *ctx) prog = (struct gl_vertex_program *) ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); + if (!prog) + return NULL; create_new_program( key, prog, ctx->Const.VertexProgram.MaxTemps ); @@ -1541,10 +1543,8 @@ _mesa_get_fixed_func_vertex_program(GLcontext *ctx) _mesa_program_cache_insert(ctx, ctx->VertexProgram.Cache, key, sizeof(*key), &prog->Base); } - else { - /* use cached program */ - _mesa_free(key); - } + + _mesa_free(key); return prog; } -- cgit v1.2.3 From f44e2f14d24b6d2730e7ff85e96695867ee96cbe Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Dec 2007 06:56:58 -0700 Subject: fix mem leaks --- src/mesa/pipe/xlib/xm_api.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index ebf4c21eaf..168eba0784 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -447,6 +447,10 @@ xmesa_free_buffer(XMesaBuffer buffer) /* Unreference. If count = zero we'll really delete the buffer */ _mesa_unreference_framebuffer(&fb); + XFreeGC(b->xm_visual->display, b->gc); + + free(buffer); + return; } /* continue search */ -- cgit v1.2.3 From 1631a9513d32ddcc5467225ad25d0e7a154af895 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Dec 2007 06:57:24 -0700 Subject: free program caches --- src/mesa/shader/program.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index f13aa18e95..fc53b5737c 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -101,6 +101,7 @@ _mesa_free_program_data(GLcontext *ctx) if (ctx->VertexProgram.Current->Base.RefCount <= 0) ctx->Driver.DeleteProgram(ctx, &(ctx->VertexProgram.Current->Base)); } + _mesa_delete_program_cache(ctx, ctx->VertexProgram.Cache); #endif #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program if (ctx->FragmentProgram.Current) { @@ -108,6 +109,7 @@ _mesa_free_program_data(GLcontext *ctx) if (ctx->FragmentProgram.Current->Base.RefCount <= 0) ctx->Driver.DeleteProgram(ctx, &(ctx->FragmentProgram.Current->Base)); } + _mesa_delete_program_cache(ctx, ctx->FragmentProgram.Cache); #endif /* XXX probably move this stuff */ #if FEATURE_ATI_fragment_shader -- cgit v1.2.3 From 6128c938650f3937ba54dd7225d258d753e16ade Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Dec 2007 07:16:12 -0700 Subject: unref const buffers during context destroy --- src/mesa/state_tracker/st_context.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 1d26da474e..668ac139f7 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -51,6 +51,7 @@ #include "st_extensions.h" #include "st_program.h" #include "pipe/p_context.h" +#include "pipe/p_winsys.h" #include "pipe/draw/draw_context.h" #include "pipe/cso_cache/cso_cache.h" @@ -136,6 +137,9 @@ struct st_context *st_create_context(struct pipe_context *pipe, static void st_destroy_context_priv( struct st_context *st ) { + struct pipe_winsys *ws = st->pipe->winsys; + uint i; + draw_destroy(st->draw); st_destroy_atoms( st ); st_destroy_draw( st ); @@ -146,6 +150,12 @@ static void st_destroy_context_priv( struct st_context *st ) _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache); + for (i = 0; i < Elements(st->state.constants); i++) { + if (st->state.constants[i].buffer) { + ws->buffer_reference(ws, &st->state.constants[i].buffer, NULL); + } + } + st->pipe->destroy( st->pipe ); free( st ); } -- cgit v1.2.3 From 4fabee9d4d217bb18411b6c481416775aead029e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 26 Dec 2007 07:16:41 -0700 Subject: unref const buffers during context destroy --- src/mesa/pipe/softpipe/sp_context.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 52672a9b29..c7af63cc2d 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -127,6 +127,7 @@ softpipe_unmap_surfaces(struct softpipe_context *sp) static void softpipe_destroy( struct pipe_context *pipe ) { struct softpipe_context *softpipe = softpipe_context( pipe ); + struct pipe_winsys *ws = pipe->winsys; uint i; draw_destroy( softpipe->draw ); @@ -152,6 +153,12 @@ static void softpipe_destroy( struct pipe_context *pipe ) for (i = 0; i < PIPE_MAX_SAMPLERS; i++) sp_destroy_tile_cache(softpipe->tex_cache[i]); + for (i = 0; i < Elements(softpipe->constants); i++) { + if (softpipe->constants[i].buffer) { + ws->buffer_reference(ws, &softpipe->constants[i].buffer, NULL); + } + } + FREE( softpipe ); } -- cgit v1.2.3 From 6883e02a5cab528623e11ad7717f1052ab2e031d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Dec 2007 19:17:41 -0700 Subject: fix a mem leak, document another existing leak --- src/mesa/pipe/cso_cache/cso_hash.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cso_cache/cso_hash.c b/src/mesa/pipe/cso_cache/cso_hash.c index fe033c9e91..0338cb3b47 100644 --- a/src/mesa/pipe/cso_cache/cso_hash.c +++ b/src/mesa/pipe/cso_cache/cso_hash.c @@ -101,8 +101,15 @@ static void *cso_data_allocate_node(struct cso_hash_data *hash) return malloc(hash->nodeSize); } -static void cso_data_free_node(void *node) +static void cso_data_free_node(struct cso_node *node) { + /* XXX still a leak here. + * Need to cast value ptr to original cso type, then free the + * driver-specific data hanging off of it. For example: + struct cso_sampler *csamp = (struct cso_sampler *) node->value; + free(csamp->data); + */ + free(node->value); free(node); } -- cgit v1.2.3 From 6656864135411d379c06e071c5a5e73b4e4baea5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Dec 2007 19:18:12 -0700 Subject: fix leak, fix refcount error --- src/mesa/pipe/softpipe/sp_tile_cache.c | 17 +++++++++++------ 1 file changed, 11 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index ba1f71b01e..1dbcc5aadd 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -126,9 +126,17 @@ void sp_destroy_tile_cache(struct softpipe_tile_cache *tc) { uint pos; + for (pos = 0; pos < NUM_ENTRIES; pos++) { - assert(tc->entries[pos].x < 0); + //assert(tc->entries[pos].x < 0); + } + if (tc->surface) { + pipe_surface_reference(&tc->surface, NULL); + } + if (tc->tex_surf) { + pipe_surface_reference(&tc->tex_surf, NULL); } + FREE( tc ); } @@ -466,15 +474,12 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, tc->tex_level != level || tc->tex_z != z) { /* get new surface (view into texture) */ - struct pipe_surface *ps; if (tc->tex_surf && tc->tex_surf->map) pipe_surface_unmap(tc->tex_surf); - ps = pipe->get_tex_surface(pipe, tc->texture, face, level, z); - pipe_surface_reference(&tc->tex_surf, ps); - - pipe_surface_map(ps); + tc->tex_surf = pipe->get_tex_surface(pipe, tc->texture, face, level, z); + pipe_surface_map(tc->tex_surf); tc->tex_face = face; tc->tex_level = level; -- cgit v1.2.3 From c37d2c9325a39287e29c22ab0e6838e7e35c7175 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Dec 2007 14:18:18 -0700 Subject: fix single-sided stencil test bug --- src/mesa/pipe/softpipe/sp_quad_stencil.c | 20 +++++++++++++------- 1 file changed, 13 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index a688a06c74..33740883d3 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -208,16 +208,22 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) struct softpipe_cached_tile *tile = sp_get_cached_tile(softpipe, softpipe->sbuf_cache, quad->x0, quad->y0); uint j; + uint face = quad->facing; + + if (!softpipe->depth_stencil->stencil[1].enabled) { + /* single-sided stencil test, use front (face=0) state */ + face = 0; + } /* choose front or back face function, operator, etc */ /* XXX we could do these initializations once per primitive */ - func = softpipe->depth_stencil->stencil[quad->facing].func; - failOp = softpipe->depth_stencil->stencil[quad->facing].fail_op; - zFailOp = softpipe->depth_stencil->stencil[quad->facing].zfail_op; - zPassOp = softpipe->depth_stencil->stencil[quad->facing].zpass_op; - ref = softpipe->depth_stencil->stencil[quad->facing].ref_value; - wrtMask = softpipe->depth_stencil->stencil[quad->facing].write_mask; - valMask = softpipe->depth_stencil->stencil[quad->facing].value_mask; + func = softpipe->depth_stencil->stencil[face].func; + failOp = softpipe->depth_stencil->stencil[face].fail_op; + zFailOp = softpipe->depth_stencil->stencil[face].zfail_op; + zPassOp = softpipe->depth_stencil->stencil[face].zpass_op; + ref = softpipe->depth_stencil->stencil[face].ref_value; + wrtMask = softpipe->depth_stencil->stencil[face].write_mask; + valMask = softpipe->depth_stencil->stencil[face].value_mask; assert(ps); /* shouldn't get here if there's no stencil buffer */ -- cgit v1.2.3 From 678db26b9d2db0c76e7676b5499d8d55a5619a94 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Dec 2007 14:25:09 -0700 Subject: unref buffer after drawing, fixes rastpos mem leak --- src/mesa/state_tracker/st_draw.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index b3d463043d..96db9e3c03 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -616,6 +616,7 @@ st_feedback_draw_vbo(GLcontext *ctx, if (draw->vertex_buffer[i].buffer) { pipe->winsys->buffer_unmap(pipe->winsys, draw->vertex_buffer[i].buffer); + winsys->buffer_reference(winsys, &draw->vertex_buffer[i].buffer, NULL); draw_set_mapped_vertex_buffer(draw, i, NULL); } } -- cgit v1.2.3 From 520a455f55f6c6c529dc6b725a5225793c0b4b01 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Dec 2007 14:52:12 -0700 Subject: fix vbo display list memleak upon context destruction --- src/mesa/vbo/vbo_context.c | 1 + src/mesa/vbo/vbo_save.c | 15 +++++++++++++++ 2 files changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c index 60d0b76ab6..235cee2429 100644 --- a/src/mesa/vbo/vbo_context.c +++ b/src/mesa/vbo/vbo_context.c @@ -246,6 +246,7 @@ void _vbo_DestroyContext( GLcontext *ctx ) } vbo_exec_destroy(ctx); + vbo_save_destroy(ctx); FREE(vbo_context(ctx)); ctx->swtnl_im = NULL; } diff --git a/src/mesa/vbo/vbo_save.c b/src/mesa/vbo/vbo_save.c index 87248e10f3..8dd87141c0 100644 --- a/src/mesa/vbo/vbo_save.c +++ b/src/mesa/vbo/vbo_save.c @@ -69,6 +69,21 @@ void vbo_save_init( GLcontext *ctx ) void vbo_save_destroy( GLcontext *ctx ) { + struct vbo_context *vbo = vbo_context(ctx); + struct vbo_save_context *save = &vbo->save; + if (save->prim_store) { + if ( --save->prim_store->refcount == 0 ) { + FREE( save->prim_store ); + save->prim_store = NULL; + } + if ( --save->vertex_store->refcount == 0 ) { + if (save->vertex_store->bufferobj) + ctx->Driver.DeleteBuffer( ctx, save->vertex_store->bufferobj ); + + FREE( save->vertex_store ); + save->vertex_store = NULL; + } + } } -- cgit v1.2.3 From 292bbd4a7250b96c4edadc2da5ebb7fc72b6159f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Jan 2008 15:09:16 -0700 Subject: move SP_NEW_ #defines into sp_state.h --- src/mesa/pipe/softpipe/sp_context.h | 16 ---------------- src/mesa/pipe/softpipe/sp_state.h | 21 +++++++++++++++++++++ 2 files changed, 21 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 8f14dd11d1..0b25e4c40b 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -45,22 +45,6 @@ struct draw_stage; struct softpipe_tile_cache; -#define SP_NEW_VIEWPORT 0x1 -#define SP_NEW_RASTERIZER 0x2 -#define SP_NEW_FS 0x4 -#define SP_NEW_BLEND 0x8 -#define SP_NEW_CLIP 0x10 -#define SP_NEW_SCISSOR 0x20 -#define SP_NEW_STIPPLE 0x40 -#define SP_NEW_FRAMEBUFFER 0x80 -#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100 -#define SP_NEW_CONSTANTS 0x200 -#define SP_NEW_SAMPLER 0x400 -#define SP_NEW_TEXTURE 0x800 -#define SP_NEW_VERTEX 0x1000 -#define SP_NEW_VS 0x2000 -#define SP_NEW_QUERY 0x4000 - struct sp_vertex_shader_state { struct pipe_shader_state *state; void *draw_data; diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index 76b79b5280..ed16228ea2 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -35,10 +35,31 @@ #include "x86/rtasm/x86sse.h" + +#define SP_NEW_VIEWPORT 0x1 +#define SP_NEW_RASTERIZER 0x2 +#define SP_NEW_FS 0x4 +#define SP_NEW_BLEND 0x8 +#define SP_NEW_CLIP 0x10 +#define SP_NEW_SCISSOR 0x20 +#define SP_NEW_STIPPLE 0x40 +#define SP_NEW_FRAMEBUFFER 0x80 +#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100 +#define SP_NEW_CONSTANTS 0x200 +#define SP_NEW_SAMPLER 0x400 +#define SP_NEW_TEXTURE 0x800 +#define SP_NEW_VERTEX 0x1000 +#define SP_NEW_VS 0x2000 +#define SP_NEW_QUERY 0x4000 + + + #ifdef MESA_LLVM struct gallivm_prog; #endif + + /** * Softpipe fs state is derived from pipe_shader_state. */ -- cgit v1.2.3 From 52659e3c238d961de1f25bed9254747f2f931547 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Jan 2008 15:17:30 -0700 Subject: Clean-up, re-org some vertex/fragment shader state code. --- src/mesa/pipe/softpipe/sp_context.h | 7 ++----- src/mesa/pipe/softpipe/sp_state.h | 12 ++++++++--- src/mesa/pipe/softpipe/sp_state_derived.c | 2 +- src/mesa/pipe/softpipe/sp_state_fs.c | 35 +++++++++++++------------------ 4 files changed, 27 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 0b25e4c40b..394baf0109 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -43,13 +43,10 @@ struct softpipe_winsys; struct draw_context; struct draw_stage; struct softpipe_tile_cache; +struct sp_fragment_shader_state; +struct sp_vertex_shader_state; -struct sp_vertex_shader_state { - struct pipe_shader_state *state; - void *draw_data; -}; - struct softpipe_context { struct pipe_context pipe; /**< base class */ struct softpipe_winsys *winsys; /**< window system interface */ diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index ed16228ea2..c1f5555a86 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -60,9 +60,7 @@ struct gallivm_prog; -/** - * Softpipe fs state is derived from pipe_shader_state. - */ +/** Subclass of pipe_shader_state */ struct sp_fragment_shader_state { struct pipe_shader_state shader; #if defined(__i386__) || defined(__386__) @@ -74,6 +72,14 @@ struct sp_fragment_shader_state { }; +/** Subclass of pipe_shader_state */ +struct sp_vertex_shader_state { + struct pipe_shader_state shader; + void *draw_data; +}; + + + void * softpipe_create_blend_state(struct pipe_context *, const struct pipe_blend_state *); diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index a5e766781f..630ae3163f 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -40,7 +40,7 @@ */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { - const struct pipe_shader_state *vs = softpipe->vs->state; + const struct pipe_shader_state *vs = &softpipe->vs->shader; const struct pipe_shader_state *fs = &softpipe->fs->shader; const enum interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index a360b4f02b..f72a91485f 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -42,18 +42,20 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { struct softpipe_context *softpipe = softpipe_context(pipe); + struct sp_fragment_shader_state *state; /* Decide whether we'll be codegenerating this shader and if so do * that now. */ - struct sp_fragment_shader_state *state = MALLOC( sizeof(struct sp_fragment_shader_state) ); + state = CALLOC_STRUCT(sp_fragment_shader_state); + if (!state) + return NULL; + state->shader = *templ; - if( softpipe->dump_fs ) { - tgsi_dump( - state->shader.tokens, - 0 ); + if (softpipe->dump_fs) { + tgsi_dump(state->shader.tokens, 0); } #if defined(__i386__) || defined(__386__) @@ -75,6 +77,7 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, return state; } + void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -84,6 +87,7 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) softpipe->dirty |= SP_NEW_FS; } + void softpipe_delete_fs_state(struct pipe_context *pipe, void *shader) { @@ -103,22 +107,16 @@ void * softpipe_create_vs_state(struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_vertex_shader_state *state; - state = MALLOC( sizeof(struct sp_vertex_shader_state) ); + state = CALLOC_STRUCT(sp_vertex_shader_state); if (state == NULL ) { return NULL; } - state->state = MALLOC( sizeof(struct pipe_shader_state) ); - if (state->state == NULL) { - FREE( state ); - return NULL; - } - memcpy( state->state, templ, sizeof(struct pipe_shader_state) ); + state->shader = *templ; state->draw_data = draw_create_vertex_shader(softpipe->draw, - state->state); + &state->shader); if (state->draw_data == NULL) { - FREE( state->state ); FREE( state ); return NULL; } @@ -126,6 +124,7 @@ void * softpipe_create_vs_state(struct pipe_context *pipe, return state; } + void softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -137,8 +136,8 @@ void softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) softpipe->dirty |= SP_NEW_VS; } -void softpipe_delete_vs_state(struct pipe_context *pipe, - void *vs) + +void softpipe_delete_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -146,12 +145,10 @@ void softpipe_delete_vs_state(struct pipe_context *pipe, (struct sp_vertex_shader_state *)vs; draw_delete_vertex_shader(softpipe->draw, state->draw_data); - FREE( state->state ); FREE( state ); } - void softpipe_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, const struct pipe_constant_buffer *buf) @@ -170,5 +167,3 @@ void softpipe_set_constant_buffer(struct pipe_context *pipe, softpipe->dirty |= SP_NEW_CONSTANTS; } - - -- cgit v1.2.3 From 57a711f7275993e75aa00918065b876776618a17 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Jan 2008 14:01:14 -0700 Subject: hack/fix pack_color() for correct ps3 format --- src/mesa/pipe/cell/spu/tri.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 90452f14ba..cd648db360 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -42,6 +42,7 @@ #include "pipe/p_compiler.h" +#include "pipe/p_format.h" #include "pipe/p_util.h" #include "main.h" #include "tri.h" @@ -235,8 +236,16 @@ pack_color(const float color[4]) uint g = (uint) (color[1] * 255.0); uint b = (uint) (color[2] * 255.0); uint a = (uint) (color[3] * 255.0); - uint icolor = (b << 24) | (g << 16) | (r << 8) | a; - return icolor; + const enum pipe_format format = PIPE_FORMAT_A8R8G8B8_UNORM; /* XXX temp */ + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + return (a << 24) | (r << 16) | (g << 8) | b; + case PIPE_FORMAT_B8G8R8A8_UNORM: + return (b << 24) | (g << 16) | (r << 8) | a; + default: + assert(0); + return 0; + } } -- cgit v1.2.3 From da92ac01e80e8a83233b1d4a881503bfc2806a1a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Jan 2008 15:21:14 -0700 Subject: Plug in more infrastructure for actual rendering. Track vertex/fragment shader state. Plug in pipe->draw_arrays(), pipe->draw_elements(). Plug render stage (a stub) into end of 'draw' pipeline. Specify a hard-coded vertex format for now. --- src/mesa/pipe/cell/ppu/Makefile | 3 + src/mesa/pipe/cell/ppu/cell_context.c | 9 +- src/mesa/pipe/cell/ppu/cell_context.h | 27 +++- src/mesa/pipe/cell/ppu/cell_draw_arrays.c | 169 +++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_draw_arrays.h | 42 ++++++ src/mesa/pipe/cell/ppu/cell_render.c | 129 ++++++++++++++++ src/mesa/pipe/cell/ppu/cell_render.h | 36 +++++ src/mesa/pipe/cell/ppu/cell_state.h | 2 + src/mesa/pipe/cell/ppu/cell_state_derived.c | 222 ++++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_state_fs.c | 103 +++++-------- 10 files changed, 675 insertions(+), 67 deletions(-) create mode 100644 src/mesa/pipe/cell/ppu/cell_draw_arrays.c create mode 100644 src/mesa/pipe/cell/ppu/cell_draw_arrays.h create mode 100644 src/mesa/pipe/cell/ppu/cell_render.c create mode 100644 src/mesa/pipe/cell/ppu/cell_render.h create mode 100644 src/mesa/pipe/cell/ppu/cell_state_derived.c (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index f597784d65..2758e1b77f 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -17,9 +17,12 @@ SPU_CODE_MODULE = ../spu/g3d_spu.a SOURCES = \ cell_context.c \ + cell_draw_arrays.c \ cell_flush.c \ + cell_render.c \ cell_state_blend.c \ cell_state_clip.c \ + cell_state_derived.c \ cell_state_fs.c \ cell_state_rasterizer.c \ cell_state_sampler.c \ diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 281fc6ec31..82b69ac33e 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -40,7 +40,9 @@ #include "pipe/cell/common.h" #include "pipe/draw/draw_context.h" #include "cell_context.h" +#include "cell_draw_arrays.h" #include "cell_flush.h" +#include "cell_render.h" #include "cell_state.h" #include "cell_surface.h" #include "cell_spu.h" @@ -212,12 +214,9 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; cell->pipe.set_vertex_element = cell_set_vertex_element; -#if 0 - cell->pipe.set_feedback_buffer = cell_set_feedback_buffer; cell->pipe.draw_arrays = cell_draw_arrays; cell->pipe.draw_elements = cell_draw_elements; -#endif cell->pipe.clear = cell_clear_surface; cell->pipe.flush = cell_flush; @@ -235,6 +234,10 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->draw = draw_create(); + cell->render_stage = cell_draw_render_stage(cell); + draw_set_rasterize_stage(cell->draw, cell->render_stage); + + /* * SPU stuff */ diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index f8d6cc5d32..5f6f987d8f 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -31,9 +31,25 @@ #include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/draw/draw_vertex.h" #include "cell_winsys.h" +struct cell_vertex_shader_state +{ + struct pipe_shader_state shader; + void *draw_data; +}; + + +struct cell_fragment_shader_state +{ + struct pipe_shader_state shader; + void *data; +}; + + struct cell_context { struct pipe_context pipe; @@ -44,6 +60,8 @@ struct cell_context const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_alpha_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; + const struct cell_vertex_shader_state *vs; + const struct cell_fragment_shader_state *fs; struct pipe_blend_color blend_color; struct pipe_clip_state clip; @@ -55,13 +73,20 @@ struct cell_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + + uint dirty; /** The primitive drawing context */ struct draw_context *draw; - struct draw_stage *setup; + struct draw_stage *render_stage; struct draw_stage *vbuf; + struct vertex_info vertex_info; + + /** Mapped constant buffers */ + void *mapped_constants[PIPE_SHADER_TYPES]; + uint num_spus; diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c new file mode 100644 index 0000000000..c014f5b29d --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c @@ -0,0 +1,169 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +#include "cell_context.h" +#include "cell_draw_arrays.h" +#include "cell_state.h" + +#include "pipe/draw/draw_context.h" + + + +static void +cell_map_constant_buffers(struct cell_context *sp) +{ + struct pipe_winsys *ws = sp->pipe.winsys; + uint i; + for (i = 0; i < 2; i++) { + if (sp->constants[i].size) + sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, + PIPE_BUFFER_FLAG_READ); + } + + draw_set_mapped_constant_buffer(sp->draw, + sp->mapped_constants[PIPE_SHADER_VERTEX]); +} + +static void +cell_unmap_constant_buffers(struct cell_context *sp) +{ + struct pipe_winsys *ws = sp->pipe.winsys; + uint i; + for (i = 0; i < 2; i++) { + if (sp->constants[i].size) + ws->buffer_unmap(ws, sp->constants[i].buffer); + sp->mapped_constants[i] = NULL; + } +} + + +boolean +cell_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count) +{ + return cell_draw_elements(pipe, NULL, 0, mode, start, count); +} + + + +/** + * Draw vertex arrays, with optional indexing. + * Basically, map the vertex buffers (and drawing surfaces), then hand off + * the drawing to the 'draw' module. + * + * XXX should the element buffer be specified/bound with a separate function? + */ +boolean +cell_draw_elements(struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + struct cell_context *sp = cell_context(pipe); + struct draw_context *draw = sp->draw; + unsigned i; + + /* first, check that the primitive is not malformed. It is the + * state tracker's responsibility to do send only correctly formed + * primitives down. It currently isn't doing that though... + */ +#if 1 + count = draw_trim_prim( mode, count ); +#else + if (!draw_validate_prim( mode, count )) + assert(0); +#endif + + if (sp->dirty) + cell_update_derived( sp ); + +#if 0 + cell_map_surfaces(sp); +#endif + cell_map_constant_buffers(sp); + + /* + * Map vertex buffers + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (sp->vertex_buffer[i].buffer) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + sp->vertex_buffer[i].buffer, + PIPE_BUFFER_FLAG_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); + } + } + /* Map index buffer, if present */ + if (indexBuffer) { + void *mapped_indexes + = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, + PIPE_BUFFER_FLAG_READ); + draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + } + else { + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + + /* draw! */ + draw_arrays(draw, mode, start, count); + + /* always flush for now */ + draw_flush(draw); + + /* + * unmap vertex/index buffers + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (sp->vertex_buffer[i].buffer) { + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); + } + } + if (indexBuffer) { + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + + /* Note: leave drawing surfaces mapped */ + cell_unmap_constant_buffers(sp); + + return TRUE; +} diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.h b/src/mesa/pipe/cell/ppu/cell_draw_arrays.h new file mode 100644 index 0000000000..bd5b703f3b --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_draw_arrays.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_DRAW_ARRAYS_H +#define CELL_DRAW_ARRAYS_H + + +boolean cell_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count); + +boolean cell_draw_elements(struct pipe_context *pipe, + struct pipe_buffer_handle *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); + + + +#endif /* CELL_DRAW_ARRAYS_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_render.c b/src/mesa/pipe/cell/ppu/cell_render.c new file mode 100644 index 0000000000..c7f0cf6be6 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_render.c @@ -0,0 +1,129 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Last stage of 'draw' pipeline: send tris to SPUs. + * \author Brian Paul + */ + +#include "cell_context.h" +#include "cell_render.h" +#include "pipe/p_util.h" +#include "pipe/draw/draw_private.h" + + +struct render_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct cell_context *cell; +}; + + +static INLINE struct render_stage * +render_stage(struct draw_stage *stage) +{ + return (struct render_stage *) stage; +} + + +static void render_begin( struct draw_stage *stage ) +{ +#if 0 + struct render_stage *render = render_stage(stage); + struct cell_context *sp = render->cell; + const struct pipe_shader_state *fs = &render->cell->fs->shader; + render->quad.nr_attrs = render->cell->nr_frag_attrs; + + render->firstFpInput = fs->input_semantic_name[0]; + + sp->quad.first->begin(sp->quad.first); +#endif +} + + +static void render_end( struct draw_stage *stage ) +{ +} + + +static void reset_stipple_counter( struct draw_stage *stage ) +{ + struct render_stage *render = render_stage(stage); + /*render->cell->line_stipple_counter = 0;*/ +} + + +static void +render_point(struct draw_stage *stage, struct prim_header *prim) +{ +} + + +static void +render_line(struct draw_stage *stage, struct prim_header *prim) +{ +} + + +static void +render_tri(struct draw_stage *stage, struct prim_header *prim) +{ + printf("Cell render tri\n"); +} + + +static void render_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + +/** + * Create a new draw/render stage. This will be plugged into the + * draw module as the last pipeline stage. + */ +struct draw_stage *cell_draw_render_stage( struct cell_context *cell ) +{ + struct render_stage *render = CALLOC_STRUCT(render_stage); + + render->cell = cell; + render->stage.draw = cell->draw; + render->stage.begin = render_begin; + render->stage.point = render_point; + render->stage.line = render_line; + render->stage.tri = render_tri; + render->stage.end = render_end; + render->stage.reset_stipple_counter = reset_stipple_counter; + render->stage.destroy = render_destroy; + + /* + render->quad.coef = render->coef; + render->quad.posCoef = &render->posCoef; + */ + + return &render->stage; +} diff --git a/src/mesa/pipe/cell/ppu/cell_render.h b/src/mesa/pipe/cell/ppu/cell_render.h new file mode 100644 index 0000000000..d66e1bdb79 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_render.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_RENDER_H +#define CELL_RENDER_H + +struct cell_context; +struct draw_stage; + +extern struct draw_stage *cell_draw_render_stage( struct cell_context *cell ); + +#endif /* CELL_RENDER_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_state.h b/src/mesa/pipe/cell/ppu/cell_state.h index 63dbac2f02..d2091f8edf 100644 --- a/src/mesa/pipe/cell/ppu/cell_state.h +++ b/src/mesa/pipe/cell/ppu/cell_state.h @@ -104,4 +104,6 @@ void cell_set_viewport_state( struct pipe_context *, const struct pipe_viewport_state * ); +void cell_update_derived( struct cell_context *softpipe ); + #endif diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c new file mode 100644 index 0000000000..ee7c0204b8 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -0,0 +1,222 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_vertex.h" +#include "cell_context.h" +#include "cell_state.h" + + +/** + * Determine which post-transform / pre-rasterization vertex attributes + * we need. + * Derived from: fs, setup states. + */ +static void calculate_vertex_layout( struct cell_context *cell ) +{ +#if 0 + const struct pipe_shader_state *vs = cell->vs->state; + const struct pipe_shader_state *fs = &cell->fs->shader; + const enum interp_mode colorInterp + = cell->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + struct vertex_info *vinfo = &cell->vertex_info; + boolean emitBack0 = FALSE, emitBack1 = FALSE, emitPsize = FALSE; + uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; + uint i; +#endif + const enum interp_mode colorInterp + = cell->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + struct vertex_info *vinfo = &cell->vertex_info; + uint front0; + + memset(vinfo, 0, sizeof(*vinfo)); + +#if 0 + if (fs->input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { + /* Need Z if depth test is enabled or the fragment program uses the + * fragment position (XYZW). + */ + } + + cell->psize_slot = -1; +#endif + + /* always emit vertex pos */ + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR); + +#if 1 + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); +#endif + +#if 0 + /* + * XXX I think we need to reconcile the vertex shader outputs with + * the fragment shader inputs here to make sure the slots line up. + * Might just be getting lucky so far. + * Or maybe do that in the state tracker? + */ + + for (i = 0; i < vs->num_outputs; i++) { + switch (vs->output_semantic_name[i]) { + + case TGSI_SEMANTIC_POSITION: + /* vertex programs always emit position, but might not be + * needed for fragment progs. + */ + /* no-op */ + break; + + case TGSI_SEMANTIC_COLOR: + if (vs->output_semantic_index[i] == 0) { + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + } + else { + assert(vs->output_semantic_index[i] == 1); + front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + } + break; + + case TGSI_SEMANTIC_BCOLOR: + if (vs->output_semantic_index[i] == 0) { + emitBack0 = TRUE; + } + else { + assert(vs->output_semantic_index[i] == 1); + emitBack1 = TRUE; + } + break; + + case TGSI_SEMANTIC_FOG: + draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE); + break; + + case TGSI_SEMANTIC_PSIZE: + /* XXX only emit if drawing points or front/back polygon mode + * is point mode + */ + emitPsize = TRUE; + break; + + case TGSI_SEMANTIC_GENERIC: + /* this includes texcoords and varying vars */ + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); + break; + + default: + assert(0); + } + } + + cell->nr_frag_attrs = fs->num_inputs; + + /* We want these after all other attribs since they won't get passed + * to the fragment shader. All prior vertex output attribs should match + * up 1:1 with the fragment shader inputs. + */ + if (emitBack0) { + back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + } + if (emitBack1) { + back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + } + if (emitPsize) { + cell->psize_slot + = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT); + } + + /* If the attributes have changed, tell the draw module about + * the new vertex layout. + */ + /* XXX we also need to do this when the shading mode (interp modes) change: */ +#endif + + if (1/*vinfo->attr_mask != cell->attr_mask*/) { + /*cell->attr_mask = vinfo->attr_mask;*/ + + draw_set_vertex_info( cell->draw, vinfo); + +#if 0 + draw_set_twoside_attributes(cell->draw, + front0, back0, front1, back1); +#endif + } +} + + +#if 0 +/** + * Recompute cliprect from scissor bounds, scissor enable and surface size. + */ +static void +compute_cliprect(struct cell_context *sp) +{ + unsigned surfWidth, surfHeight; + + if (sp->framebuffer.num_cbufs > 0) { + surfWidth = sp->framebuffer.cbufs[0]->width; + surfHeight = sp->framebuffer.cbufs[0]->height; + } + else { + /* no surface? */ + surfWidth = sp->scissor.maxx; + surfHeight = sp->scissor.maxy; + } + + if (sp->rasterizer->scissor) { + /* clip to scissor rect */ + sp->cliprect.minx = MAX2(sp->scissor.minx, 0); + sp->cliprect.miny = MAX2(sp->scissor.miny, 0); + sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth); + sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight); + } + else { + /* clip to surface bounds */ + sp->cliprect.minx = 0; + sp->cliprect.miny = 0; + sp->cliprect.maxx = surfWidth; + sp->cliprect.maxy = surfHeight; + } +} +#endif + + +void cell_update_derived( struct cell_context *cell ) +{ + if (cell->dirty & (CELL_NEW_RASTERIZER | CELL_NEW_FS)) + calculate_vertex_layout( cell ); + +#if 0 + if (cell->dirty & (CELL_NEW_SCISSOR | + CELL_NEW_DEPTH_STENCIL_ALPHA | + CELL_NEW_FRAMEBUFFER)) + compute_cliprect(cell); +#endif + + cell->dirty = 0; +} diff --git a/src/mesa/pipe/cell/ppu/cell_state_fs.c b/src/mesa/pipe/cell/ppu/cell_state_fs.c index 910f210cdc..5a935f2de1 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_fs.c +++ b/src/mesa/pipe/cell/ppu/cell_state_fs.c @@ -28,8 +28,8 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "pipe/p_winsys.h" -#if 0 #include "pipe/draw/draw_context.h" +#if 0 #include "pipe/p_shader_tokens.h" #include "pipe/llvm/gallivm.h" #include "pipe/tgsi/util/tgsi_dump.h" @@ -40,25 +40,22 @@ #include "cell_state.h" -void * cell_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) +void * +cell_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { struct cell_context *cell = cell_context(pipe); + struct cell_fragment_shader_state *state; - return malloc(5); /* XXX temp */ - -#if 0 - /* Decide whether we'll be codegenerating this shader and if so do - * that now. - */ + state = CALLOC_STRUCT(cell_fragment_shader_state); + if (!state) + return NULL; - struct sp_fragment_shader_state *state = MALLOC( sizeof(struct sp_fragment_shader_state) ); state->shader = *templ; - if( cell->dump_fs ) { - tgsi_dump( - state->shader.tokens, - 0 ); +#if 0 + if (cell->dump_fs) { + tgsi_dump(state->shader.tokens, 0); } #if defined(__i386__) || defined(__386__) @@ -76,32 +73,32 @@ void * cell_create_fs_state(struct pipe_context *pipe, else gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); #endif - return state; #endif + + return state; } -void cell_bind_fs_state(struct pipe_context *pipe, void *fs) + +void +cell_bind_fs_state(struct pipe_context *pipe, void *fs) { struct cell_context *cell = cell_context(pipe); -#if 0 - cell->fs = (struct sp_fragment_shader_state *) fs; - cell->dirty |= SP_NEW_FS; -#endif + cell->fs = (struct cell_fragment_shader_state *) fs; + + cell->dirty |= CELL_NEW_FS; } -void cell_delete_fs_state(struct pipe_context *pipe, - void *shader) + +void +cell_delete_fs_state(struct pipe_context *pipe, void *fs) { -#if 0 - struct sp_fragment_shader_state *state = shader; + struct cell_context *cell = cell_context(pipe); -#if defined(__i386__) || defined(__386__) - x86_release_func( &state->sse2_program ); -#endif + struct cell_fragment_shader_state *state = + (struct cell_fragment_shader_state *) fs; FREE( state ); -#endif } @@ -109,73 +106,56 @@ void * cell_create_vs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { - return malloc(5); /* XXX */ -#if 0 struct cell_context *cell = cell_context(pipe); - struct sp_vertex_shader_state *state; + struct cell_vertex_shader_state *state; - state = MALLOC( sizeof(struct sp_vertex_shader_state) ); - if (state == NULL ) { + state = CALLOC_STRUCT(cell_vertex_shader_state); + if (!state) return NULL; - } - state->state = MALLOC( sizeof(struct pipe_shader_state) ); - if (state->state == NULL) { - FREE( state ); - return NULL; - } - memcpy( state->state, templ, sizeof(struct pipe_shader_state) ); + state->shader = *templ; - state->draw_data = draw_create_vertex_shader(cell->draw, - state->state); + state->draw_data = draw_create_vertex_shader(cell->draw, &state->shader); if (state->draw_data == NULL) { - FREE( state->state ); FREE( state ); return NULL; } return state; -#endif } - void cell_bind_vs_state(struct pipe_context *pipe, void *vs) { -#if 0 struct cell_context *cell = cell_context(pipe); - cell->vs = (const struct sp_vertex_shader_state *)vs; + cell->vs = (const struct cell_vertex_shader_state *) vs; draw_bind_vertex_shader(cell->draw, cell->vs->draw_data); - cell->dirty |= SP_NEW_VS; -#endif + cell->dirty |= CELL_NEW_VS; } + void cell_delete_vs_state(struct pipe_context *pipe, void *vs) { -#if 0 struct cell_context *cell = cell_context(pipe); - struct sp_vertex_shader_state *state = - (struct sp_vertex_shader_state *)vs; + struct cell_vertex_shader_state *state = + (struct cell_vertex_shader_state *) vs; draw_delete_vertex_shader(cell->draw, state->draw_data); - FREE( state->state ); FREE( state ); -#endif } - -void cell_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) +void +cell_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) { -#if 0 struct cell_context *cell = cell_context(pipe); struct pipe_winsys *ws = pipe->winsys; @@ -188,8 +168,5 @@ void cell_set_constant_buffer(struct pipe_context *pipe, buf->buffer); cell->constants[shader].size = buf->size; - cell->dirty |= SP_NEW_CONSTANTS; -#endif + cell->dirty |= CELL_NEW_CONSTANTS; } - - -- cgit v1.2.3 From de9f8e8b717aa4b4ab94af73be5aa70088cd6b81 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 2 Jan 2008 18:53:33 -0700 Subject: Cell: basic triangle rendering works. The cell "render_stage" (last in the "draw" pipeline) emits vertices into a buffer which is pulled by the SPUs in response to a "RENDER" command. This is pretty much temporary/scaffold code for now. --- src/mesa/pipe/cell/common.h | 21 ++++++++ src/mesa/pipe/cell/ppu/cell_context.c | 5 +- src/mesa/pipe/cell/ppu/cell_context.h | 3 +- src/mesa/pipe/cell/ppu/cell_flush.c | 3 ++ src/mesa/pipe/cell/ppu/cell_render.c | 73 +++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_render.h | 3 ++ src/mesa/pipe/cell/ppu/cell_surface.c | 8 ++- src/mesa/pipe/cell/spu/main.c | 95 ++++++++++++++++++++++++++++++++--- src/mesa/pipe/cell/spu/main.h | 13 +++-- src/mesa/pipe/cell/spu/tri.c | 21 ++++++-- src/mesa/pipe/cell/spu/tri.h | 4 ++ src/mesa/pipe/xlib/xm_winsys.c | 9 +++- 12 files changed, 240 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index f7f1e2eb41..0868e8d90f 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -51,6 +51,7 @@ #define CELL_CMD_CLEAR_TILES 3 #define CELL_CMD_TRIANGLE 4 #define CELL_CMD_FINISH 5 +#define CELL_CMD_RENDER 6 /** @@ -80,12 +81,22 @@ struct cell_command_triangle } ALIGN16_ATTRIB; +struct cell_command_render +{ + uint prim_type; + uint num_verts; + float xmin, ymin, xmax, ymax; + void *vertex_data; +} ALIGN16_ATTRIB; + + /** XXX unions don't seem to work */ struct cell_command { struct cell_command_framebuffer fb; struct cell_command_clear_tiles clear; struct cell_command_triangle tri; + struct cell_command_render render; } ALIGN16_ATTRIB; @@ -98,6 +109,16 @@ struct cell_init_info } ALIGN16_ATTRIB; +/** Temporary */ +#define CELL_MAX_VERTS 48 +#define CELL_MAX_ATTRIBS 2 +struct cell_prim_buffer +{ + float vertex[CELL_MAX_VERTS][CELL_MAX_ATTRIBS][4] ALIGN16_ATTRIB; + float xmin, ymin, xmax, ymax; + uint num_verts; +} ALIGN16_ATTRIB; + #endif /* CELL_COMMON_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 82b69ac33e..fb89837a7c 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -161,10 +161,13 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) { struct cell_context *cell; - cell = CALLOC_STRUCT(cell_context); + /* some fields need to be 16-byte aligned, so align the whole object */ + cell = (struct cell_context*) align_malloc(sizeof(struct cell_context), 16); if (!cell) return NULL; + memset(cell, 0, sizeof(*cell)); + cell->winsys = cws; cell->pipe.winsys = winsys; cell->pipe.destroy = cell_destroy_context; diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 5f6f987d8f..be985820d1 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -34,6 +34,7 @@ #include "pipe/p_defines.h" #include "pipe/draw/draw_vertex.h" #include "cell_winsys.h" +#include "pipe/cell/common.h" struct cell_vertex_shader_state @@ -90,7 +91,7 @@ struct cell_context uint num_spus; - + struct cell_prim_buffer prim_buffer; }; diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c index b1ff0e51cd..47003bef18 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -29,6 +29,7 @@ #include "cell_context.h" #include "cell_flush.h" #include "cell_spu.h" +#include "cell_render.h" void @@ -39,6 +40,8 @@ cell_flush(struct pipe_context *pipe, unsigned flags) printf("%s\n", __FUNCTION__); + cell_flush_prim_buffer(cell); + /* Send CMD_FINISH to all SPUs */ for (i = 0; i < cell->num_spus; i++) { send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FINISH); diff --git a/src/mesa/pipe/cell/ppu/cell_render.c b/src/mesa/pipe/cell/ppu/cell_render.c index c7f0cf6be6..672406aa23 100644 --- a/src/mesa/pipe/cell/ppu/cell_render.c +++ b/src/mesa/pipe/cell/ppu/cell_render.c @@ -32,6 +32,7 @@ #include "cell_context.h" #include "cell_render.h" +#include "cell_spu.h" #include "pipe/p_util.h" #include "pipe/draw/draw_private.h" @@ -89,13 +90,85 @@ render_line(struct draw_stage *stage, struct prim_header *prim) } +/** Write a vertex into the prim buffer */ +static void +save_vertex(struct cell_prim_buffer *buf, uint pos, + const struct vertex_header *vert) +{ + uint attr, j; + + for (attr = 0; attr < 2; attr++) { + for (j = 0; j < 4; j++) { + buf->vertex[pos][attr][j] = vert->data[attr][j]; + } + } + + /* update bounding box */ + if (vert->data[0][0] < buf->xmin) + buf->xmin = vert->data[0][0]; + if (vert->data[0][0] > buf->xmax) + buf->xmax = vert->data[0][0]; + if (vert->data[0][1] < buf->ymin) + buf->ymin = vert->data[0][1]; + if (vert->data[0][1] > buf->ymax) + buf->ymax = vert->data[0][1]; +} + + static void render_tri(struct draw_stage *stage, struct prim_header *prim) { + struct render_stage *rs = render_stage(stage); + struct cell_context *cell = rs->cell; + struct cell_prim_buffer *buf = &cell->prim_buffer; + uint i; + printf("Cell render tri\n"); + + if (buf->num_verts + 3 > CELL_MAX_VERTS) { + cell_flush_prim_buffer(cell); + } + + i = buf->num_verts; + assert(i+2 <= CELL_MAX_VERTS); + save_vertex(buf, i+0, prim->v[0]); + save_vertex(buf, i+1, prim->v[1]); + save_vertex(buf, i+2, prim->v[2]); + buf->num_verts += 3; } +/** + * Send the a RENDER command to all SPUs to have them render the prims + * in the current prim_buffer. + */ +void +cell_flush_prim_buffer(struct cell_context *cell) +{ + uint i; + + if (cell->prim_buffer.num_verts == 0) + return; + + printf("*** Flushing prim buffer\n"); + for (i = 0; i < cell->num_spus; i++) { + struct cell_command_render *render = &cell_global.command[i].render; + render->prim_type = PIPE_PRIM_TRIANGLES; + render->num_verts = cell->prim_buffer.num_verts; + render->xmin = cell->prim_buffer.xmin; + render->ymin = cell->prim_buffer.ymin; + render->xmax = cell->prim_buffer.xmax; + render->ymax = cell->prim_buffer.ymax; + render->vertex_data = &cell->prim_buffer.vertex; + ASSERT_ALIGN16(render->vertex_data); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER); + } + + cell->prim_buffer.num_verts = 0; +} + + + static void render_destroy( struct draw_stage *stage ) { FREE( stage ); diff --git a/src/mesa/pipe/cell/ppu/cell_render.h b/src/mesa/pipe/cell/ppu/cell_render.h index d66e1bdb79..826dcbafeb 100644 --- a/src/mesa/pipe/cell/ppu/cell_render.h +++ b/src/mesa/pipe/cell/ppu/cell_render.h @@ -31,6 +31,9 @@ struct cell_context; struct draw_stage; +extern void +cell_flush_prim_buffer(struct cell_context *cell); + extern struct draw_stage *cell_draw_render_stage( struct cell_context *cell ); #endif /* CELL_RENDER_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 185eeb26e8..c487d0439a 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -51,8 +51,14 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, if (!ps->map) pipe_surface_map(ps); + if (pf_get_size(ps->format) != 4) { + printf("Cell: Skipping non 32bpp clear_surface\n"); + return; + } + for (i = 0; i < cell->num_spus; i++) { struct cell_command_framebuffer *fb = &cell_global.command[i].fb; + printf("%s %u start = 0x%x\n", __FUNCTION__, i, ps->map); fb->start = ps->map; fb->width = ps->width; fb->height = ps->height; @@ -66,7 +72,7 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_TILES); } -#if 1 +#if 0 /* XXX Draw a test triangle over the cleared surface */ for (i = 0; i < cell->num_spus; i++) { /* Same triangle data for all SPUs */ diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index cc5eddb0f5..183397a5ff 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -37,6 +37,7 @@ #include "main.h" #include "tri.h" #include "pipe/cell/common.h" +#include "pipe/p_defines.h" /* helpful headers: @@ -48,6 +49,8 @@ volatile struct cell_init_info init; struct framebuffer fb; +uint tile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; + int DefaultTag; @@ -62,12 +65,12 @@ wait_on_mask(unsigned tag) void -get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile) +get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, + int tag) { uint offset = ty * fb->width_tiles + tx; uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; ubyte *src = (ubyte *) fb->start + offset * bytesPerTile; - int tag = DefaultTag; assert(tx < fb->width_tiles); assert(ty < fb->height_tiles); @@ -85,12 +88,12 @@ get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile) } void -put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile) +put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, + int tag) { uint offset = ty * fb->width_tiles + tx; uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; ubyte *dst = (ubyte *) fb->start + offset * bytesPerTile; - int tag = DefaultTag; assert(tx < fb->width_tiles); assert(ty < fb->height_tiles); @@ -100,7 +103,7 @@ put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile) tile, (unsigned int) dst, bytesPerTile); */ mfc_put((void *) tile, /* src in local memory */ - (unsigned int) dst, /* dst in main mory */ + (unsigned int) dst, /* dst in main memory */ bytesPerTile, tag, 0, /* tid */ @@ -119,15 +122,19 @@ clear_tiles(const struct cell_command_clear_tiles *clear) for (i = 0; i < TILE_SIZE * TILE_SIZE; i++) tile[i] = clear->value; + /* printf("SPU: %s num=%d w=%d h=%d\n", __FUNCTION__, num_tiles, fb.width_tiles, fb.height_tiles); + */ + for (i = init.id; i < num_tiles; i += init.num_spus) { uint tx = i % fb.width_tiles; uint ty = i / fb.width_tiles; - put_tile(&fb, tx, ty, tile); + put_tile(&fb, tx, ty, tile, DefaultTag); /* XXX we don't want this here, but it fixes bad tile results */ wait_on_mask(1 << DefaultTag); } + } @@ -155,6 +162,76 @@ triangle(const struct cell_command_triangle *tri) +static void +render(const struct cell_command_render *render) +{ + const uint num_tiles = fb.width_tiles * fb.height_tiles; + struct cell_prim_buffer prim_buffer ALIGN16_ATTRIB; + int tag = DefaultTag; + uint i, j; + + /* + printf("SPU %u: RENDER buffer dst=%p src=%p size=%d\n", + init.id, + &prim_buffer, render->vertex_data, (int)sizeof(prim_buffer)); + */ + + ASSERT_ALIGN16(render->vertex_data); + ASSERT_ALIGN16(&prim_buffer); + + /* get vertex data from main memory */ + mfc_get(&prim_buffer, /* dest */ + (unsigned int) render->vertex_data, /* src */ + sizeof(prim_buffer), /* bytes */ + tag, + 0, /* tid */ + 0 /* rid */); + wait_on_mask( 1 << tag ); /* XXX temporary */ + + /* loop over tiles */ + for (i = init.id; i < num_tiles; i += init.num_spus) { + uint tx = i % fb.width_tiles; + uint ty = i / fb.width_tiles; + + get_tile(&fb, tx, ty, (uint *) tile, DefaultTag); + wait_on_mask(1 << DefaultTag); /* XXX temporary */ + + assert(render->prim_type == PIPE_PRIM_TRIANGLES); + + /* loop over tris */ + for (j = 0; j < render->num_verts; j += 3) { + struct prim_header prim; + + /* + printf(" %u: Triangle %g,%g %g,%g %g,%g\n", + init.id, + prim_buffer.vertex[j*3+0][0][0], + prim_buffer.vertex[j*3+0][0][1], + prim_buffer.vertex[j*3+1][0][0], + prim_buffer.vertex[j*3+1][0][1], + prim_buffer.vertex[j*3+2][0][0], + prim_buffer.vertex[j*3+2][0][1]); + */ + + /* pos */ + COPY_4V(prim.v[0].data[0], prim_buffer.vertex[j+0][0]); + COPY_4V(prim.v[1].data[0], prim_buffer.vertex[j+1][0]); + COPY_4V(prim.v[2].data[0], prim_buffer.vertex[j+2][0]); + + /* color */ + COPY_4V(prim.v[0].data[1], prim_buffer.vertex[j+0][1]); + COPY_4V(prim.v[1].data[1], prim_buffer.vertex[j+1][1]); + COPY_4V(prim.v[2].data[1], prim_buffer.vertex[j+2][1]); + + draw_triangle(&prim, tx, ty); + } + + put_tile(&fb, tx, ty, (uint *) tile, DefaultTag); + wait_on_mask(1 << DefaultTag); /* XXX temp */ + } +} + + /** * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. */ @@ -215,6 +292,12 @@ main_loop(void) printf("SPU %u: TRIANGLE\n", init.id); triangle(&cmd.tri); break; + case CELL_CMD_RENDER: + printf("SPU %u: RENDER %u verts, prim %u\n", + init.id, cmd.render.num_verts, cmd.render.prim_type); + render(&cmd.render); + break; + case CELL_CMD_FINISH: printf("SPU %u: FINISH\n", init.id); /* wait for all outstanding DMAs to finish */ diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index 8c2796387f..47ce4be4b4 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -37,13 +37,16 @@ extern volatile struct cell_init_info init; struct framebuffer { - void *start; - uint width, height; + void *start; /**< addr of surface in main memory */ + uint width, height; /**< size in pixels */ uint width_tiles, height_tiles; /**< width and height in tiles */ }; +/* XXX Collect these globals in a struct: */ + extern struct framebuffer fb; +extern uint tile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; extern int DefaultTag; @@ -52,10 +55,12 @@ void wait_on_mask(unsigned tag); void -get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile); +get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, + int tag); void -put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile); +put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, + int tag); #endif /* MAIN_H */ diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index cd648db360..ce759a5647 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -74,8 +74,6 @@ static int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; -static uint tile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; - #endif @@ -879,11 +877,26 @@ draw_triangle(struct prim_header *tri, uint tx, uint ty) cliprect_maxx = (tx + 1) * TILE_SIZE; cliprect_maxy = (ty + 1) * TILE_SIZE; - get_tile(&fb, tx, ty, (uint *) tile); + get_tile(&fb, tx, ty, (uint *) tile, DefaultTag); wait_on_mask(1 << DefaultTag); setup_tri(tri); - put_tile(&fb, tx, ty, (uint *) tile); + put_tile(&fb, tx, ty, (uint *) tile, DefaultTag); wait_on_mask(1 << DefaultTag); } + + +void +tri_draw(struct prim_header *tri, uint tx, uint ty) +{ + /* set clipping bounds to tile bounds */ + cliprect_minx = tx * TILE_SIZE; + cliprect_miny = ty * TILE_SIZE; + cliprect_maxx = (tx + 1) * TILE_SIZE; + cliprect_maxy = (ty + 1) * TILE_SIZE; + + setup_tri(tri); +} + + diff --git a/src/mesa/pipe/cell/spu/tri.h b/src/mesa/pipe/cell/spu/tri.h index dc66ad0f47..576030579f 100644 --- a/src/mesa/pipe/cell/spu/tri.h +++ b/src/mesa/pipe/cell/spu/tri.h @@ -49,4 +49,8 @@ extern void draw_triangle(struct prim_header *tri, uint tx, uint ty); +extern void +tri_draw(struct prim_header *tri, uint tx, uint ty); + + #endif /* TRI_H */ diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 42c43387af..10dc09b13c 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -45,7 +45,10 @@ #ifdef GALLIUM_CELL #include "pipe/cell/ppu/cell_context.h" #include "pipe/cell/ppu/cell_winsys.h" +#else +#define TILE_SIZE 32 /* avoid compilation errors */ #endif + #include "xm_winsys_aub.h" @@ -214,7 +217,6 @@ xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) { XImage *ximage = b->tempImage; struct xm_buffer *xm_buf = xm_bo(surf->buffer); - const int TILE_SIZE = 32; const uint tilesPerRow = (surf->width + TILE_SIZE - 1) / TILE_SIZE; uint x, y; @@ -234,6 +236,7 @@ xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) int tx = x / TILE_SIZE; int ty = y / TILE_SIZE; int offset = ty * tilesPerRow + tx; + offset *= 4 * TILE_SIZE * TILE_SIZE; ximage->data = (char *) xm_buf->data + offset; @@ -364,6 +367,10 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys, surf->cpp = pf_get_size(format); surf->pitch = round_up(width, alignment / surf->cpp); +#ifdef GALLIUM_CELL /* XXX a bit of a hack */ + height = round_up(height, TILE_SIZE); +#endif + assert(!surf->buffer); surf->buffer = winsys->buffer_create(winsys, alignment, 0, 0); if(!surf->buffer) -- cgit v1.2.3 From d55c4ec9d2c0a8cd9ba75985962297381e6c0364 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 2 Jan 2008 18:58:44 -0700 Subject: remove previous triangle test code --- src/mesa/pipe/cell/common.h | 9 --------- src/mesa/pipe/cell/ppu/cell_surface.c | 33 --------------------------------- src/mesa/pipe/cell/spu/main.c | 30 +----------------------------- src/mesa/pipe/cell/spu/tri.c | 23 ++++------------------- src/mesa/pipe/cell/spu/tri.h | 4 ---- 5 files changed, 5 insertions(+), 94 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 0868e8d90f..8202bbbe76 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -49,7 +49,6 @@ #define CELL_CMD_EXIT 1 #define CELL_CMD_FRAMEBUFFER 2 #define CELL_CMD_CLEAR_TILES 3 -#define CELL_CMD_TRIANGLE 4 #define CELL_CMD_FINISH 5 #define CELL_CMD_RENDER 6 @@ -74,13 +73,6 @@ struct cell_command_clear_tiles } ALIGN16_ATTRIB; -struct cell_command_triangle -{ - float vert[3][4]; - float color[3][4]; -} ALIGN16_ATTRIB; - - struct cell_command_render { uint prim_type; @@ -95,7 +87,6 @@ struct cell_command { struct cell_command_framebuffer fb; struct cell_command_clear_tiles clear; - struct cell_command_triangle tri; struct cell_command_render render; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index c487d0439a..8b3ed4a23a 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -71,37 +71,4 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, cell_global.command[i].clear.value = clearValue | (i << 21); send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_TILES); } - -#if 0 - /* XXX Draw a test triangle over the cleared surface */ - for (i = 0; i < cell->num_spus; i++) { - /* Same triangle data for all SPUs */ - struct cell_command_triangle *tri = &cell_global.command[i].tri; - tri->vert[0][0] = 20.0; - tri->vert[0][1] = ps->height - 20; - - tri->vert[1][0] = ps->width - 20.0; - tri->vert[1][1] = ps->height - 20; - - tri->vert[2][0] = ps->width / 2; - tri->vert[2][1] = 20.0; - - tri->color[0][0] = 1.0; - tri->color[0][1] = 0.0; - tri->color[0][2] = 0.0; - tri->color[0][3] = 0.0; - - tri->color[1][0] = 0.0; - tri->color[1][1] = 1.0; - tri->color[1][2] = 0.0; - tri->color[1][3] = 0.0; - - tri->color[2][0] = 0.0; - tri->color[2][1] = 0.0; - tri->color[2][2] = 1.0; - tri->color[2][3] = 0.0; - - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_TRIANGLE); - } -#endif } diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 183397a5ff..94b9eda5c3 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -138,30 +138,6 @@ clear_tiles(const struct cell_command_clear_tiles *clear) } -static void -triangle(const struct cell_command_triangle *tri) -{ - uint num_tiles = fb.width_tiles * fb.height_tiles; - struct prim_header prim; - uint i; - - COPY_4V(prim.v[0].data[0], tri->vert[0]); - COPY_4V(prim.v[1].data[0], tri->vert[1]); - COPY_4V(prim.v[2].data[0], tri->vert[2]); - - COPY_4V(prim.v[0].data[1], tri->color[0]); - COPY_4V(prim.v[1].data[1], tri->color[1]); - COPY_4V(prim.v[2].data[1], tri->color[2]); - - for (i = init.id; i < num_tiles; i += init.num_spus) { - uint tx = i % fb.width_tiles; - uint ty = i / fb.width_tiles; - draw_triangle(&prim, tx, ty); - } -} - - - static void render(const struct cell_command_render *render) { @@ -223,7 +199,7 @@ render(const struct cell_command_render *render) COPY_4V(prim.v[1].data[1], prim_buffer.vertex[j+1][1]); COPY_4V(prim.v[2].data[1], prim_buffer.vertex[j+2][1]); - draw_triangle(&prim, tx, ty); + tri_draw(&prim, tx, ty); } put_tile(&fb, tx, ty, (uint *) tile, DefaultTag); @@ -288,10 +264,6 @@ main_loop(void) printf("SPU %u: CLEAR to 0x%08x\n", init.id, cmd.clear.value); clear_tiles(&cmd.clear); break; - case CELL_CMD_TRIANGLE: - printf("SPU %u: TRIANGLE\n", init.id); - triangle(&cmd.tri); - break; case CELL_CMD_RENDER: printf("SPU %u: RENDER %u verts, prim %u\n", init.id, cmd.render.num_verts, cmd.render.prim_type); diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index ce759a5647..4e198e638c 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -868,25 +868,10 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) #endif -void -draw_triangle(struct prim_header *tri, uint tx, uint ty) -{ - /* set clipping bounds to tile bounds */ - cliprect_minx = tx * TILE_SIZE; - cliprect_miny = ty * TILE_SIZE; - cliprect_maxx = (tx + 1) * TILE_SIZE; - cliprect_maxy = (ty + 1) * TILE_SIZE; - - get_tile(&fb, tx, ty, (uint *) tile, DefaultTag); - wait_on_mask(1 << DefaultTag); - - setup_tri(tri); - - put_tile(&fb, tx, ty, (uint *) tile, DefaultTag); - wait_on_mask(1 << DefaultTag); -} - - +/** + * Draw triangle into tile at (tx, ty) (tile coords) + * The tile data should have already been fetched. + */ void tri_draw(struct prim_header *tri, uint tx, uint ty) { diff --git a/src/mesa/pipe/cell/spu/tri.h b/src/mesa/pipe/cell/spu/tri.h index 576030579f..01612a2579 100644 --- a/src/mesa/pipe/cell/spu/tri.h +++ b/src/mesa/pipe/cell/spu/tri.h @@ -45,10 +45,6 @@ struct prim_header { }; -extern void -draw_triangle(struct prim_header *tri, uint tx, uint ty); - - extern void tri_draw(struct prim_header *tri, uint tx, uint ty); -- cgit v1.2.3 From 54090bd841b302c6d48e7f130dbe07c8fd5a0a96 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 2 Jan 2008 19:05:34 -0700 Subject: only fetch as much vertex data as needed --- src/mesa/pipe/cell/common.h | 2 +- src/mesa/pipe/cell/ppu/cell_render.c | 1 + src/mesa/pipe/cell/spu/main.c | 11 +++++++---- 3 files changed, 9 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 8202bbbe76..ee20bd150b 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -76,7 +76,7 @@ struct cell_command_clear_tiles struct cell_command_render { uint prim_type; - uint num_verts; + uint num_verts, num_attribs; float xmin, ymin, xmax, ymax; void *vertex_data; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_render.c b/src/mesa/pipe/cell/ppu/cell_render.c index 672406aa23..b97f4ff536 100644 --- a/src/mesa/pipe/cell/ppu/cell_render.c +++ b/src/mesa/pipe/cell/ppu/cell_render.c @@ -155,6 +155,7 @@ cell_flush_prim_buffer(struct cell_context *cell) struct cell_command_render *render = &cell_global.command[i].render; render->prim_type = PIPE_PRIM_TRIANGLES; render->num_verts = cell->prim_buffer.num_verts; + render->num_attribs = CELL_MAX_ATTRIBS; render->xmin = cell->prim_buffer.xmin; render->ymin = cell->prim_buffer.ymin; render->xmax = cell->prim_buffer.xmax; diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 94b9eda5c3..5e29d4faaa 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -144,7 +144,7 @@ render(const struct cell_command_render *render) const uint num_tiles = fb.width_tiles * fb.height_tiles; struct cell_prim_buffer prim_buffer ALIGN16_ATTRIB; int tag = DefaultTag; - uint i, j; + uint i, j, vertex_bytes; /* printf("SPU %u: RENDER buffer dst=%p src=%p size=%d\n", @@ -155,10 +155,13 @@ render(const struct cell_command_render *render) ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(&prim_buffer); + /* how much vertex data */ + vertex_bytes = render->num_verts * render->num_attribs * 4 * sizeof(float); + /* get vertex data from main memory */ mfc_get(&prim_buffer, /* dest */ (unsigned int) render->vertex_data, /* src */ - sizeof(prim_buffer), /* bytes */ + vertex_bytes, /* size */ tag, 0, /* tid */ 0 /* rid */); @@ -166,8 +169,8 @@ render(const struct cell_command_render *render) /* loop over tiles */ for (i = init.id; i < num_tiles; i += init.num_spus) { - uint tx = i % fb.width_tiles; - uint ty = i / fb.width_tiles; + const uint tx = i % fb.width_tiles; + const uint ty = i / fb.width_tiles; get_tile(&fb, tx, ty, (uint *) tile, DefaultTag); wait_on_mask(1 << DefaultTag); /* XXX temporary */ -- cgit v1.2.3 From f6b7e2d3bf9af34704e9624246614c1583b655da Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 2 Jan 2008 19:31:36 -0700 Subject: make use of prim bounds box info --- src/mesa/pipe/cell/ppu/cell_context.c | 5 +++++ src/mesa/pipe/cell/ppu/cell_render.c | 5 +++++ src/mesa/pipe/cell/spu/main.c | 40 ++++++++++++++++++++++++++++++++--- 3 files changed, 47 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index fb89837a7c..4fcf804d82 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -241,6 +241,11 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) draw_set_rasterize_stage(cell->draw, cell->render_stage); + cell->prim_buffer.xmin = 1e100; + cell->prim_buffer.ymin = 1e100; + cell->prim_buffer.xmax = -1e100; + cell->prim_buffer.ymax = -1e100; + /* * SPU stuff */ diff --git a/src/mesa/pipe/cell/ppu/cell_render.c b/src/mesa/pipe/cell/ppu/cell_render.c index b97f4ff536..79aa37ea02 100644 --- a/src/mesa/pipe/cell/ppu/cell_render.c +++ b/src/mesa/pipe/cell/ppu/cell_render.c @@ -166,6 +166,11 @@ cell_flush_prim_buffer(struct cell_context *cell) } cell->prim_buffer.num_verts = 0; + + cell->prim_buffer.xmin = 1e100; + cell->prim_buffer.ymin = 1e100; + cell->prim_buffer.xmax = -1e100; + cell->prim_buffer.ymax = -1e100; } diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 5e29d4faaa..94f13029ff 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -87,6 +87,7 @@ get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, 0 /* rid */); } + void put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, int tag) @@ -138,6 +139,33 @@ clear_tiles(const struct cell_command_clear_tiles *clear) } +/** + * Given a rendering command's bounding box (in pixels) compute the + * location of the corresponding screen tile bounding box. + */ +static INLINE void +tile_bounding_box(const struct cell_command_render *render, + uint *txmin, uint *tymin, + uint *box_num_tiles, uint *box_width_tiles) +{ + uint txmax, tymax, box_height_tiles; + + *txmin = (uint) render->xmin / TILE_SIZE; + *tymin = (uint) render->ymin / TILE_SIZE; + txmax = (uint) render->xmax / TILE_SIZE; + tymax = (uint) render->ymax / TILE_SIZE; + *box_width_tiles = txmax - *txmin + 1; + box_height_tiles = tymax - *tymin + 1; + *box_num_tiles = *box_width_tiles * box_height_tiles; + /* + printf("Render bounds: %g, %g ... %g, %g\n", + render->xmin, render->ymin, render->xmax, render->ymax); + printf("Render tiles: %u, %u .. %u, %u\n", *txmin, *tymin, txmax, tymax); + */ +} + + + static void render(const struct cell_command_render *render) { @@ -167,10 +195,16 @@ render(const struct cell_command_render *render) 0 /* rid */); wait_on_mask( 1 << tag ); /* XXX temporary */ + /* find tiles which intersect the prim bounding box */ + uint txmin, tymin, box_width_tiles, box_num_tiles; + tile_bounding_box(render, &txmin, &tymin, + &box_num_tiles, &box_width_tiles); + + /* loop over tiles */ - for (i = init.id; i < num_tiles; i += init.num_spus) { - const uint tx = i % fb.width_tiles; - const uint ty = i / fb.width_tiles; + for (i = init.id; i < box_num_tiles; i += init.num_spus) { + const uint tx = txmin + i % box_width_tiles; + const uint ty = tymin + i / box_width_tiles; get_tile(&fb, tx, ty, (uint *) tile, DefaultTag); wait_on_mask(1 << DefaultTag); /* XXX temporary */ -- cgit v1.2.3 From 7aee3eff18decd6ce46f63a44c9a531414565265 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 07:49:13 -0700 Subject: better debug code --- src/mesa/pipe/cell/ppu/cell_surface.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 8b3ed4a23a..5fa37dd2cd 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -67,8 +67,21 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, } for (i = 0; i < cell->num_spus; i++) { - /* XXX clear color varies per-SPU for debugging */ - cell_global.command[i].clear.value = clearValue | (i << 21); +#if 1 + uint clr = clearValue; + /* XXX debug: clear color varied per-SPU to visualize tiles */ + if ((clr & 0xff) == 0) + clr |= 64 + i * 8; + if ((clr & 0xff00) == 0) + clr |= (64 + i * 8) << 8; + if ((clr & 0xff0000) == 0) + clr |= (64 + i * 8) << 16; + if ((clr & 0xff000000) == 0) + clr |= (64 + i * 8) << 24; + cell_global.command[i].clear.value = clr; +#else + cell_global.command[i].clear.value = clearValue; +#endif send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_TILES); } } -- cgit v1.2.3 From 5cd96f7684adc7ba52a216fb676c7eb88ba3f66a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 07:49:59 -0700 Subject: pass surface format in cell_command_framebuffer struct --- src/mesa/pipe/cell/common.h | 3 ++- src/mesa/pipe/cell/spu/main.c | 6 ++++-- src/mesa/pipe/cell/spu/main.h | 1 + src/mesa/pipe/cell/spu/tri.c | 3 +-- 4 files changed, 8 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index ee20bd150b..983ccd0775 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -35,6 +35,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_util.h" +#include "pipe/p_format.h" /** for sanity checking */ @@ -60,7 +61,7 @@ struct cell_command_framebuffer { void *start; int width, height; - unsigned format; + enum pipe_format format; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 94f13029ff..02b23866a5 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -285,10 +285,12 @@ main_loop(void) exitFlag = 1; break; case CELL_CMD_FRAMEBUFFER: - printf("SPU %u: FRAMEBUFFER: %d x %d at %p\n", init.id, + printf("SPU %u: FRAMEBUFFER: %d x %d at %p, format 0x%x\n", init.id, cmd.fb.width, cmd.fb.height, - cmd.fb.start); + cmd.fb.start, + cmd.fb.format); + fb.format = cmd.fb.format; fb.width = cmd.fb.width; fb.height = cmd.fb.height; fb.width_tiles = (fb.width + TILE_SIZE - 1) / TILE_SIZE; diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index 47ce4be4b4..3240e01615 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -38,6 +38,7 @@ extern volatile struct cell_init_info init; struct framebuffer { void *start; /**< addr of surface in main memory */ + enum pipe_format format; uint width, height; /**< size in pixels */ uint width_tiles, height_tiles; /**< width and height in tiles */ }; diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 4e198e638c..3707ebea45 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -234,8 +234,7 @@ pack_color(const float color[4]) uint g = (uint) (color[1] * 255.0); uint b = (uint) (color[2] * 255.0); uint a = (uint) (color[3] * 255.0); - const enum pipe_format format = PIPE_FORMAT_A8R8G8B8_UNORM; /* XXX temp */ - switch (format) { + switch (fb.format) { case PIPE_FORMAT_A8R8G8B8_UNORM: return (a << 24) | (r << 16) | (g << 8) | b; case PIPE_FORMAT_B8G8R8A8_UNORM: -- cgit v1.2.3 From bb55835fc07e4d29e94c7db1e3854a5bd37c8aeb Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 08:58:01 -0700 Subject: insert a temporary flush to fix missing triangles artifact --- src/mesa/pipe/cell/ppu/cell_render.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_render.c b/src/mesa/pipe/cell/ppu/cell_render.c index 79aa37ea02..f5a91776a5 100644 --- a/src/mesa/pipe/cell/ppu/cell_render.c +++ b/src/mesa/pipe/cell/ppu/cell_render.c @@ -171,6 +171,11 @@ cell_flush_prim_buffer(struct cell_context *cell) cell->prim_buffer.ymin = 1e100; cell->prim_buffer.xmax = -1e100; cell->prim_buffer.ymax = -1e100; + + /* XXX temporary, need to double-buffer the prim buffer until we get + * a real command buffer/list system. + */ + cell_flush(&cell->pipe, 0x0); } -- cgit v1.2.3 From 9b598df95ebe99d9aaf2043ce8786847978de4aa Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 08:58:51 -0700 Subject: asst changes in bbox code, dma tags, etc --- src/mesa/pipe/cell/spu/main.c | 32 +++++++++++++++++++++++--------- 1 file changed, 23 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 02b23866a5..a131bd5198 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -119,6 +119,7 @@ clear_tiles(const struct cell_command_clear_tiles *clear) uint num_tiles = fb.width_tiles * fb.height_tiles; uint i; uint tile[TILE_SIZE * TILE_SIZE] ALIGN16_ATTRIB; + int tag = init.id; for (i = 0; i < TILE_SIZE * TILE_SIZE; i++) tile[i] = clear->value; @@ -131,9 +132,9 @@ clear_tiles(const struct cell_command_clear_tiles *clear) for (i = init.id; i < num_tiles; i += init.num_spus) { uint tx = i % fb.width_tiles; uint ty = i / fb.width_tiles; - put_tile(&fb, tx, ty, tile, DefaultTag); + put_tile(&fb, tx, ty, tile, tag); /* XXX we don't want this here, but it fixes bad tile results */ - wait_on_mask(1 << DefaultTag); + wait_on_mask(1 << tag); } } @@ -148,6 +149,15 @@ tile_bounding_box(const struct cell_command_render *render, uint *txmin, uint *tymin, uint *box_num_tiles, uint *box_width_tiles) { +#if 0 + /* Debug: full-window bounding box */ + uint txmax = fb.width_tiles - 1; + uint tymax = fb.height_tiles - 1; + *txmin = 0; + *tymin = 0; + *box_num_tiles = fb.width_tiles * fb.height_tiles; + *box_width_tiles = fb.width_tiles; +#else uint txmax, tymax, box_height_tiles; *txmin = (uint) render->xmin / TILE_SIZE; @@ -157,11 +167,12 @@ tile_bounding_box(const struct cell_command_render *render, *box_width_tiles = txmax - *txmin + 1; box_height_tiles = tymax - *tymin + 1; *box_num_tiles = *box_width_tiles * box_height_tiles; - /* +#endif +#if 0 printf("Render bounds: %g, %g ... %g, %g\n", render->xmin, render->ymin, render->xmax, render->ymax); printf("Render tiles: %u, %u .. %u, %u\n", *txmin, *tymin, txmax, tymax); - */ +#endif } @@ -171,7 +182,7 @@ render(const struct cell_command_render *render) { const uint num_tiles = fb.width_tiles * fb.height_tiles; struct cell_prim_buffer prim_buffer ALIGN16_ATTRIB; - int tag = DefaultTag; + int tag = init.id /**DefaultTag**/; uint i, j, vertex_bytes; /* @@ -206,8 +217,11 @@ render(const struct cell_command_render *render) const uint tx = txmin + i % box_width_tiles; const uint ty = tymin + i / box_width_tiles; - get_tile(&fb, tx, ty, (uint *) tile, DefaultTag); - wait_on_mask(1 << DefaultTag); /* XXX temporary */ + assert(tx < fb.width_tiles); + assert(ty < fb.height_tiles); + + get_tile(&fb, tx, ty, (uint *) tile, tag); + wait_on_mask(1 << tag); /* XXX temporary */ assert(render->prim_type == PIPE_PRIM_TRIANGLES); @@ -239,8 +253,8 @@ render(const struct cell_command_render *render) tri_draw(&prim, tx, ty); } - put_tile(&fb, tx, ty, (uint *) tile, DefaultTag); - wait_on_mask(1 << DefaultTag); /* XXX temp */ + put_tile(&fb, tx, ty, (uint *) tile, tag); + wait_on_mask(1 << tag); /* XXX temp */ } } -- cgit v1.2.3 From 4ff6367295bc266cf1e3390570c9aee50fe716a0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 09:40:02 -0700 Subject: Cell: improve surface state code to replace some temporary code. --- src/mesa/pipe/cell/common.h | 4 +- src/mesa/pipe/cell/ppu/cell_spu.c | 15 +++--- src/mesa/pipe/cell/ppu/cell_state_surface.c | 77 +++++++++++++++++++++++++++-- src/mesa/pipe/cell/ppu/cell_surface.c | 7 +-- src/mesa/pipe/cell/spu/main.c | 17 ++++--- src/mesa/pipe/cell/spu/main.h | 6 ++- src/mesa/pipe/cell/spu/tri.c | 2 +- 7 files changed, 103 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 983ccd0775..dff0b1f8ec 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -59,9 +59,9 @@ */ struct cell_command_framebuffer { - void *start; int width, height; - enum pipe_format format; + void *color_start, *depth_start; + enum pipe_format color_format, depth_format; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 55a0d5038b..15b89682fa 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -158,16 +158,19 @@ void test_spus(struct cell_context *cell) { uint i; - struct pipe_surface *surf = cell->framebuffer.cbufs[0]; + struct pipe_surface *csurf = cell->framebuffer.cbufs[0]; + struct pipe_surface *zsurf = cell->framebuffer.zbuf; printf("PPU: sleep(2)\n\n\n"); sleep(2); for (i = 0; i < cell->num_spus; i++) { - cell_global.command[i].fb.start = surf->map; - cell_global.command[i].fb.width = surf->width; - cell_global.command[i].fb.height = surf->height; - cell_global.command[i].fb.format = PIPE_FORMAT_A8R8G8B8_UNORM; + cell_global.command[i].fb.color_start = csurf->map; + cell_global.command[i].fb.depth_start = zsurf ? zsurf->map : NULL; + cell_global.command[i].fb.width = csurf->width; + cell_global.command[i].fb.height = csurf->height; + cell_global.command[i].fb.color_format = PIPE_FORMAT_A8R8G8B8_UNORM; + cell_global.command[i].fb.depth_format = PIPE_FORMAT_Z32_UNORM; send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FRAMEBUFFER); } @@ -179,7 +182,7 @@ test_spus(struct cell_context *cell) finish_all(cell->num_spus); { - uint *b = (uint*) surf->map; + uint *b = (uint*) csurf->map; printf("PPU: Clear results: 0x%x 0x%x 0x%x 0x%x\n", b[0], b[1000], b[2000], b[3000]); } diff --git a/src/mesa/pipe/cell/ppu/cell_state_surface.c b/src/mesa/pipe/cell/ppu/cell_state_surface.c index f7330caf5e..cd8e5def31 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_state_surface.c @@ -1,7 +1,36 @@ - - +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_inlines.h" #include "cell_context.h" #include "cell_state.h" +#include "cell_spu.h" + void cell_set_framebuffer_state(struct pipe_context *pipe, @@ -9,9 +38,49 @@ cell_set_framebuffer_state(struct pipe_context *pipe, { struct cell_context *cell = cell_context(pipe); - cell->framebuffer = *fb; + if (memcmp(&cell->framebuffer, fb, sizeof(*fb))) { + struct pipe_surface *csurf = fb->cbufs[0]; + struct pipe_surface *zsurf = fb->zbuf; + uint i; + + /* change in fb state */ + + /* unmap old surfaces */ + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + if (cell->framebuffer.cbufs[i] && + cell->framebuffer.cbufs[i]->map) { + pipe_surface_unmap(cell->framebuffer.cbufs[i]); + } + } - cell->dirty |= CELL_NEW_FRAMEBUFFER; + if (cell->framebuffer.zbuf && + cell->framebuffer.zbuf->map) { + pipe_surface_unmap(cell->framebuffer.zbuf); + } + + /* update my state */ + cell->framebuffer = *fb; + + /* map new surfaces */ + if (csurf && !csurf->map) + pipe_surface_map(csurf); + + if (zsurf && !zsurf->map) + pipe_surface_map(zsurf); + + for (i = 0; i < cell->num_spus; i++) { + struct cell_command_framebuffer *fb = &cell_global.command[i].fb; + fb->color_start = csurf->map; + fb->color_format = csurf->format; + fb->depth_start = zsurf ? zsurf->map : NULL; + fb->depth_format = zsurf ? zsurf->format : PIPE_FORMAT_NONE; + fb->width = csurf->width; + fb->height = csurf->height; + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FRAMEBUFFER); + } + + cell->dirty |= CELL_NEW_FRAMEBUFFER; + } #if 0 struct pipe_surface *ps; diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 5fa37dd2cd..62e0febc0c 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -55,16 +55,17 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, printf("Cell: Skipping non 32bpp clear_surface\n"); return; } - +#if 0 for (i = 0; i < cell->num_spus; i++) { struct cell_command_framebuffer *fb = &cell_global.command[i].fb; printf("%s %u start = 0x%x\n", __FUNCTION__, i, ps->map); - fb->start = ps->map; + fb->color_start = ps->map; fb->width = ps->width; fb->height = ps->height; - fb->format = ps->format; + fb->color_format = ps->format; send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FRAMEBUFFER); } +#endif for (i = 0; i < cell->num_spus; i++) { #if 1 diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index a131bd5198..6292962f44 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -70,7 +70,7 @@ get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, { uint offset = ty * fb->width_tiles + tx; uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; - ubyte *src = (ubyte *) fb->start + offset * bytesPerTile; + ubyte *src = (ubyte *) fb->color_start + offset * bytesPerTile; assert(tx < fb->width_tiles); assert(ty < fb->height_tiles); @@ -94,7 +94,7 @@ put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, { uint offset = ty * fb->width_tiles + tx; uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; - ubyte *dst = (ubyte *) fb->start + offset * bytesPerTile; + ubyte *dst = (ubyte *) fb->color_start + offset * bytesPerTile; assert(tx < fb->width_tiles); assert(ty < fb->height_tiles); @@ -180,7 +180,6 @@ tile_bounding_box(const struct cell_command_render *render, static void render(const struct cell_command_render *render) { - const uint num_tiles = fb.width_tiles * fb.height_tiles; struct cell_prim_buffer prim_buffer ALIGN16_ATTRIB; int tag = init.id /**DefaultTag**/; uint i, j, vertex_bytes; @@ -302,16 +301,20 @@ main_loop(void) printf("SPU %u: FRAMEBUFFER: %d x %d at %p, format 0x%x\n", init.id, cmd.fb.width, cmd.fb.height, - cmd.fb.start, - cmd.fb.format); - fb.format = cmd.fb.format; + cmd.fb.color_start, + cmd.fb.color_format); + fb.color_start = cmd.fb.color_start; + fb.depth_start = cmd.fb.depth_start; + fb.color_format = cmd.fb.color_format; + fb.depth_format = cmd.fb.depth_format; fb.width = cmd.fb.width; fb.height = cmd.fb.height; fb.width_tiles = (fb.width + TILE_SIZE - 1) / TILE_SIZE; fb.height_tiles = (fb.height + TILE_SIZE - 1) / TILE_SIZE; + /* printf("SPU %u: %u x %u tiles\n", init.id, fb.width_tiles, fb.height_tiles); - fb.start = cmd.fb.start; + */ break; case CELL_CMD_CLEAR_TILES: printf("SPU %u: CLEAR to 0x%08x\n", init.id, cmd.clear.value); diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index 3240e01615..60a565bfa7 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -37,8 +37,10 @@ extern volatile struct cell_init_info init; struct framebuffer { - void *start; /**< addr of surface in main memory */ - enum pipe_format format; + void *color_start; /**< addr of color surface in main memory */ + void *depth_start; /**< addr of depth surface in main memory */ + enum pipe_format color_format; + enum pipe_format depth_format; uint width, height; /**< size in pixels */ uint width_tiles, height_tiles; /**< width and height in tiles */ }; diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 3707ebea45..f58cc4b024 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -234,7 +234,7 @@ pack_color(const float color[4]) uint g = (uint) (color[1] * 255.0); uint b = (uint) (color[2] * 255.0); uint a = (uint) (color[3] * 255.0); - switch (fb.format) { + switch (fb.color_format) { case PIPE_FORMAT_A8R8G8B8_UNORM: return (a << 24) | (r << 16) | (g << 8) | b; case PIPE_FORMAT_B8G8R8A8_UNORM: -- cgit v1.2.3 From 3ffef8de825f843ed504a8177fd08af9196be696 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 09:40:32 -0700 Subject: disable bbox code until glitches are fixed --- src/mesa/pipe/cell/spu/main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 6292962f44..bdb814e5d8 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -149,7 +149,7 @@ tile_bounding_box(const struct cell_command_render *render, uint *txmin, uint *tymin, uint *box_num_tiles, uint *box_width_tiles) { -#if 0 +#if 1 /* Debug: full-window bounding box */ uint txmax = fb.width_tiles - 1; uint tymax = fb.height_tiles - 1; -- cgit v1.2.3 From 5e00ae3fea51afd5de14d559f693645837b3903b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 09:56:48 -0700 Subject: Cell: initial work for getting/putting Z tiles --- src/mesa/pipe/cell/spu/main.c | 47 ++++++++++++++++++++++++++++--------------- src/mesa/pipe/cell/spu/main.h | 7 ++++--- src/mesa/pipe/cell/spu/tri.c | 8 ++++---- 3 files changed, 39 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index bdb814e5d8..9580281971 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -49,7 +49,8 @@ volatile struct cell_init_info init; struct framebuffer fb; -uint tile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +uint ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; int DefaultTag; @@ -66,11 +67,13 @@ wait_on_mask(unsigned tag) void get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, - int tag) + int tag, int zBuf) { - uint offset = ty * fb->width_tiles + tx; - uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; - ubyte *src = (ubyte *) fb->color_start + offset * bytesPerTile; + const uint offset = ty * fb->width_tiles + tx; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; + const ubyte *src = zBuf ? fb->depth_start : fb->color_start; + + src += offset * bytesPerTile; assert(tx < fb->width_tiles); assert(ty < fb->height_tiles); @@ -90,11 +93,13 @@ get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, void put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, - int tag) + int tag, int zBuf) { - uint offset = ty * fb->width_tiles + tx; - uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; - ubyte *dst = (ubyte *) fb->color_start + offset * bytesPerTile; + const uint offset = ty * fb->width_tiles + tx; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; + ubyte *dst = zBuf ? fb->depth_start : fb->color_start; + + dst += offset * bytesPerTile; assert(tx < fb->width_tiles); assert(ty < fb->height_tiles); @@ -117,12 +122,12 @@ static void clear_tiles(const struct cell_command_clear_tiles *clear) { uint num_tiles = fb.width_tiles * fb.height_tiles; - uint i; - uint tile[TILE_SIZE * TILE_SIZE] ALIGN16_ATTRIB; + uint i, j; int tag = init.id; - for (i = 0; i < TILE_SIZE * TILE_SIZE; i++) - tile[i] = clear->value; + for (i = 0; i < TILE_SIZE; i++) + for (j = 0; j < TILE_SIZE; j++) + ctile[i][j] = clear->value; /* printf("SPU: %s num=%d w=%d h=%d\n", @@ -132,7 +137,7 @@ clear_tiles(const struct cell_command_clear_tiles *clear) for (i = init.id; i < num_tiles; i += init.num_spus) { uint tx = i % fb.width_tiles; uint ty = i / fb.width_tiles; - put_tile(&fb, tx, ty, tile, tag); + put_tile(&fb, tx, ty, (uint *) ctile, tag, 0); /* XXX we don't want this here, but it fixes bad tile results */ wait_on_mask(1 << tag); } @@ -219,9 +224,14 @@ render(const struct cell_command_render *render) assert(tx < fb.width_tiles); assert(ty < fb.height_tiles); - get_tile(&fb, tx, ty, (uint *) tile, tag); + get_tile(&fb, tx, ty, (uint *) ctile, tag, 0); wait_on_mask(1 << tag); /* XXX temporary */ + if (fb.depth_format == PIPE_FORMAT_Z32_UNORM) { + get_tile(&fb, tx, ty, (uint *) ztile, tag+1, 1); + wait_on_mask(1 << (tag+1)); /* XXX temporary */ + } + assert(render->prim_type == PIPE_PRIM_TRIANGLES); /* loop over tris */ @@ -252,8 +262,13 @@ render(const struct cell_command_render *render) tri_draw(&prim, tx, ty); } - put_tile(&fb, tx, ty, (uint *) tile, tag); + put_tile(&fb, tx, ty, (uint *) ctile, tag, 0); wait_on_mask(1 << tag); /* XXX temp */ + + if (fb.depth_format == PIPE_FORMAT_Z32_UNORM) { + put_tile(&fb, tx, ty, (uint *) ztile, tag+1, 1); + wait_on_mask(1 << (tag+1)); /* XXX temporary */ + } } } diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index 60a565bfa7..c22679c5d9 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -49,7 +49,8 @@ struct framebuffer { extern struct framebuffer fb; -extern uint tile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +extern uint ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; extern int DefaultTag; @@ -59,11 +60,11 @@ wait_on_mask(unsigned tag); void get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, - int tag); + int tag, int zBuf); void put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, - int tag); + int tag, int zBuf); #endif /* MAIN_H */ diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index f58cc4b024..36599297b9 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -267,13 +267,13 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) eval_coeff(setup, 1, (float) x, (float) y, colors); if (mask & MASK_TOP_LEFT) - tile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); + ctile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); if (mask & MASK_TOP_RIGHT) - tile[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); + ctile[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); if (mask & MASK_BOTTOM_LEFT) - tile[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); + ctile[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); if (mask & MASK_BOTTOM_RIGHT) - tile[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); + ctile[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); #endif } -- cgit v1.2.3 From aa7f2333675f3e005f3eb6a40ac55d2fb55ea36e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 15:03:52 -0700 Subject: replace void * with struct draw_vertex_shader opaque type --- src/mesa/pipe/draw/draw_context.h | 10 ++++++---- src/mesa/pipe/draw/draw_vertex_shader.c | 16 ++++++---------- src/mesa/pipe/i915simple/i915_state.c | 6 +++--- src/mesa/pipe/softpipe/sp_state.h | 2 +- 4 files changed, 16 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 6dc6e4ce82..60be3e194d 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -45,6 +45,7 @@ struct vertex_buffer; struct vertex_info; struct draw_context; struct draw_stage; +struct draw_vertex_shader; /** @@ -89,12 +90,13 @@ void draw_set_rasterize_stage( struct draw_context *draw, struct draw_stage *stage ); -void * draw_create_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader); +struct draw_vertex_shader * +draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader); void draw_bind_vertex_shader(struct draw_context *draw, - void *vcso); + struct draw_vertex_shader *dvs); void draw_delete_vertex_shader(struct draw_context *draw, - void *vcso); + struct draw_vertex_shader *dvs); boolean draw_use_sse(struct draw_context *draw); diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index d34d923018..1fa9af8cec 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -227,7 +227,7 @@ void draw_vertex_shader_queue_flush( struct draw_context *draw ) } -void * +struct draw_vertex_shader * draw_create_vertex_shader(struct draw_context *draw, const struct pipe_shader_state *shader) { @@ -263,10 +263,10 @@ draw_create_vertex_shader(struct draw_context *draw, } void draw_bind_vertex_shader(struct draw_context *draw, - void *vcso) + struct draw_vertex_shader *dvs) { draw_flush(draw); - draw->vertex_shader = (struct draw_vertex_shader*)(vcso); + draw->vertex_shader = dvs; /* specify the fragment program to interpret/execute */ tgsi_exec_machine_init(&draw->machine, @@ -276,15 +276,11 @@ void draw_bind_vertex_shader(struct draw_context *draw, } void draw_delete_vertex_shader(struct draw_context *draw, - void *vcso) + struct draw_vertex_shader *dvs) { - struct draw_vertex_shader *vs; - - vs = (struct draw_vertex_shader *) vcso; - #if defined(__i386__) || defined(__386__) - x86_release_func( (struct x86_function *) &vs->sse2_program ); + x86_release_func( (struct x86_function *) &dvs->sse2_program ); #endif - FREE( vs ); + FREE( dvs ); } diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index f8332aab37..1190e05699 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -438,12 +438,12 @@ i915_create_vs_state(struct pipe_context *pipe, return draw_create_vertex_shader(i915->draw, templ); } -static void i915_bind_vs_state(struct pipe_context *pipe, void *vs) +static void i915_bind_vs_state(struct pipe_context *pipe, void *shader) { struct i915_context *i915 = i915_context(pipe); /* just pass-through to draw module */ - draw_bind_vertex_shader(i915->draw, vs); + draw_bind_vertex_shader(i915->draw, (struct draw_vertex_shader *) shader); } static void i915_delete_vs_state(struct pipe_context *pipe, void *shader) @@ -451,7 +451,7 @@ static void i915_delete_vs_state(struct pipe_context *pipe, void *shader) struct i915_context *i915 = i915_context(pipe); /* just pass-through to draw module */ - draw_delete_vertex_shader(i915->draw, shader); + draw_delete_vertex_shader(i915->draw, (struct draw_vertex_shader *) shader); } static void i915_set_constant_buffer(struct pipe_context *pipe, diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index c1f5555a86..bac7b0876f 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -75,7 +75,7 @@ struct sp_fragment_shader_state { /** Subclass of pipe_shader_state */ struct sp_vertex_shader_state { struct pipe_shader_state shader; - void *draw_data; + struct draw_vertex_shader *draw_data; }; -- cgit v1.2.3 From 2b40838972bb84a0dff8f8a3c933b0d2b8384f10 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 15:05:51 -0700 Subject: rename vars, clean-up formatting --- src/mesa/pipe/softpipe/sp_state_fs.c | 33 ++++++++++++++++++++------------- 1 file changed, 20 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index f72a91485f..830c873b25 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -38,8 +38,9 @@ #include "pipe/tgsi/exec/tgsi_sse2.h" -void * softpipe_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) +void * +softpipe_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_fragment_shader_state *state; @@ -78,7 +79,8 @@ void * softpipe_create_fs_state(struct pipe_context *pipe, } -void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) +void +softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -88,10 +90,10 @@ void softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) } -void softpipe_delete_fs_state(struct pipe_context *pipe, - void *shader) +void +softpipe_delete_fs_state(struct pipe_context *pipe, void *fs) { - struct sp_fragment_shader_state *state = shader; + struct sp_fragment_shader_state *state = fs; #if defined(__i386__) || defined(__386__) x86_release_func( &state->sse2_program ); @@ -101,8 +103,9 @@ void softpipe_delete_fs_state(struct pipe_context *pipe, } -void * softpipe_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) +void * +softpipe_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_vertex_shader_state *state; @@ -125,7 +128,8 @@ void * softpipe_create_vs_state(struct pipe_context *pipe, } -void softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) +void +softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -137,7 +141,8 @@ void softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) } -void softpipe_delete_vs_state(struct pipe_context *pipe, void *vs) +void +softpipe_delete_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); @@ -149,9 +154,11 @@ void softpipe_delete_vs_state(struct pipe_context *pipe, void *vs) } -void softpipe_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) + +void +softpipe_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) { struct softpipe_context *softpipe = softpipe_context(pipe); struct pipe_winsys *ws = pipe->winsys; -- cgit v1.2.3 From 4e62fbbfc5bcf368c3f11d4384008e64aad06ec8 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Jan 2008 15:09:30 -0700 Subject: clean-ups, silence warnings --- src/mesa/pipe/draw/draw_vertex_shader.c | 26 ++++++++++++++++++-------- 1 file changed, 18 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 1fa9af8cec..240149118f 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -117,7 +117,11 @@ run_vertex_program(struct draw_context *draw, #if defined(__i386__) || defined(__386__) if (draw->use_sse) { /* SSE */ - codegen_function func = (codegen_function) x86_get_func( &draw->vertex_shader->sse2_program ); + /* cast away const */ + struct draw_vertex_shader *shader + = (struct draw_vertex_shader *)draw->vertex_shader; + codegen_function func + = (codegen_function) x86_get_func( &shader->sse2_program ); func( machine->Inputs, machine->Outputs, @@ -193,7 +197,8 @@ run_vertex_program(struct draw_context *draw, * Run the vertex shader on all vertices in the vertex queue. * Called by the draw module when the vertx cache needs to be flushed. */ -void draw_vertex_shader_queue_flush( struct draw_context *draw ) +void +draw_vertex_shader_queue_flush(struct draw_context *draw) { unsigned i, j; @@ -246,7 +251,7 @@ draw_create_vertex_shader(struct draw_context *draw, struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; x86_init_func( &vs->sse2_program ); - tgsi_emit_sse2( sh->tokens, &vs->sse2_program ); + tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, &vs->sse2_program ); } #endif #ifdef MESA_LLVM @@ -255,15 +260,18 @@ draw_create_vertex_shader(struct draw_context *draw, if (!draw->engine) { draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); } - else + else { gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); + } #endif return vs; } -void draw_bind_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs) + +void +draw_bind_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs) { draw_flush(draw); draw->vertex_shader = dvs; @@ -275,8 +283,10 @@ void draw_bind_vertex_shader(struct draw_context *draw, NULL /*samplers*/ ); } -void draw_delete_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs) + +void +draw_delete_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs) { #if defined(__i386__) || defined(__386__) x86_release_func( (struct x86_function *) &dvs->sse2_program ); -- cgit v1.2.3 From 25c9728644becd6342d025bdf355f311d00d5cb5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 3 Jan 2008 16:44:56 +0000 Subject: 965: scan fs inputs to work out interpolation in setup program --- src/mesa/pipe/i965simple/brw_sf.c | 102 +++++++++++++++++++++------------ src/mesa/pipe/i965simple/brw_sf_emit.c | 30 +++++++--- 2 files changed, 87 insertions(+), 45 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_sf.c b/src/mesa/pipe/i965simple/brw_sf.c index 7b6fd3fff6..c04234ecf4 100644 --- a/src/mesa/pipe/i965simple/brw_sf.c +++ b/src/mesa/pipe/i965simple/brw_sf.c @@ -119,7 +119,11 @@ static boolean search_cache( struct brw_context *brw, */ static void upload_sf_prog( struct brw_context *brw ) { + const struct brw_fragment_program *fs = brw->attribs.FragmentProgram; struct brw_sf_prog_key key; + struct tgsi_parse_context parse; + int i, done = 0; + memset(&key, 0, sizeof(key)); @@ -149,6 +153,63 @@ static void upload_sf_prog( struct brw_context *brw ) } + + /* Scan fp inputs to figure out what interpolation modes are + * required for each incoming vp output. There is an assumption + * that the state tracker makes sure there is a 1:1 linkage between + * these sets of attributes (XXX: position??) + */ + tgsi_parse_init( &parse, fs->program.tokens ); + while( !done && + !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) + { + int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + int interp_mode = parse.FullToken.FullDeclaration.Interpolation.Interpolate; + //int semantic = parse.FullToken.FullDeclaration.Semantic.SemanticName; + //int semantic_index = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; + + _mesa_printf("fs input %d..%d interp mode %d\n", first, last, interp_mode); + + switch (interp_mode) { + case TGSI_INTERPOLATE_CONSTANT: + for (i = first; i <= last; i++) + key.const_mask |= (1 << i); + break; + case TGSI_INTERPOLATE_LINEAR: + for (i = first; i <= last; i++) + key.linear_mask |= (1 << i); + break; + case TGSI_INTERPOLATE_PERSPECTIVE: + for (i = first; i <= last; i++) + key.persp_mask |= (1 << i); + break; + default: + break; + } + + /* Also need stuff for flat shading, twosided color. + */ + + } + break; + default: + done = 1; + break; + } + } + + _mesa_printf("key.persp_mask: %x\n", key.persp_mask); + _mesa_printf("key.linear_mask: %x\n", key.linear_mask); + _mesa_printf("key.const_mask: %x\n", key.const_mask); + + // key.do_point_sprite = brw->attribs.Point->PointSprite; // key.SpriteOrigin = brw->attribs.Point->SpriteOrigin; @@ -176,6 +237,8 @@ const struct brw_tracked_state brw_sf_prog = { }; + +#if 0 /* Build a struct like the one we'd like the state tracker to pass to * us. */ @@ -202,43 +265,6 @@ static void update_sf_linkage( struct brw_context *brw ) - /* First scan fp inputs - */ - tgsi_parse_init( &parse, fs->program.tokens ); - while( !done && - !tgsi_parse_end_of_tokens( &parse ) ) - { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) - { - int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; - - for (i = first; i < last; i++) { - state.fp_input[i].vp_output = ~0; - state.fp_input[i].bf_vp_output = ~0; - state.fp_input[i].interp_mode = - parse.FullToken.FullDeclaration.Interpolation.Interpolate; - - fp_semantic[i].semantic = - parse.FullToken.FullDeclaration.Semantic.SemanticName; - fp_semantic[i].semantic_index = - parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - - } - - assert(last > state.fp_input_count); - state.fp_input_count = last; - } - break; - default: - done = 1; - break; - } - } assert(state.fp_input_count == fs->program.num_inputs); @@ -313,3 +339,5 @@ const struct brw_tracked_state brw_sf_linkage = { .update = update_sf_linkage }; + +#endif diff --git a/src/mesa/pipe/i965simple/brw_sf_emit.c b/src/mesa/pipe/i965simple/brw_sf_emit.c index 0fa61f14b6..bc3878a93a 100644 --- a/src/mesa/pipe/i965simple/brw_sf_emit.c +++ b/src/mesa/pipe/i965simple/brw_sf_emit.c @@ -134,30 +134,39 @@ static boolean calculate_masks( struct brw_sf_compile *c, ushort *pc_linear) { boolean is_last_attr = (reg == c->nr_setup_regs - 1); + unsigned persp_mask = c->key.persp_mask; + unsigned linear_mask = c->key.linear_mask; + _mesa_printf("persp_mask: %x\n", persp_mask); + _mesa_printf("linear_mask: %x\n", linear_mask); *pc_persp = 0; *pc_linear = 0; *pc = 0xf; -// if (persp_mask & (1 << c->idx_to_attr[reg*2])) -// *pc_persp = 0xf; + if (persp_mask & (1 << (reg*2))) + *pc_persp = 0xf; -// if (linear_mask & (1 << c->idx_to_attr[reg*2])) + if (linear_mask & (1 << (reg*2))) *pc_linear = 0xf; /* Maybe only processs one attribute on the final round: */ - if (1 || reg*2+1 < c->nr_setup_attrs) { + if (reg*2+1 < c->nr_setup_attrs) { *pc |= 0xf0; -// if (persp_mask & (1 << c->idx_to_attr[reg*2+1])) -// *pc_persp |= 0xf0; + if (persp_mask & (1 << (reg*2+1))) + *pc_persp |= 0xf0; -// if (linear_mask & (1 << c->idx_to_attr[reg*2+1])) + if (linear_mask & (1 << (reg*2+1))) *pc_linear |= 0xf0; } + _mesa_printf("pc: %x\n", *pc); + _mesa_printf("pc_persp: %x\n", *pc_persp); + _mesa_printf("pc_linear: %x\n", *pc_linear); + + return is_last_attr; } @@ -168,6 +177,8 @@ void brw_emit_tri_setup( struct brw_sf_compile *c ) struct brw_compile *p = &c->func; unsigned i; + _mesa_printf("%s START ==============\n", __FUNCTION__); + c->nr_verts = 3; alloc_regs(c); invert_det(c); @@ -181,7 +192,7 @@ void brw_emit_tri_setup( struct brw_sf_compile *c ) struct brw_reg a0 = offset(c->vert[0], i); struct brw_reg a1 = offset(c->vert[1], i); struct brw_reg a2 = offset(c->vert[2], i); - ushort pc, pc_persp, pc_linear; + ushort pc = 0, pc_persp = 0, pc_linear = 0; boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); if (pc_persp) @@ -238,6 +249,9 @@ void brw_emit_tri_setup( struct brw_sf_compile *c ) BRW_URB_SWIZZLE_TRANSPOSE); /* XXX: Swizzle control "SF to windower" */ } } + + _mesa_printf("%s DONE ==============\n", __FUNCTION__); + } -- cgit v1.2.3 From 65426b144e9c3f6c8c1e6d0c0dce0ae955f7d1c3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 3 Jan 2008 17:21:22 +0000 Subject: 965: always perform SF parameter setup for position Match behaviour of DRI driver. Fix fragment shader to find the other parameters one slot further on. Will need more work to cope with FP's that actually reference position. --- src/mesa/pipe/i965simple/brw_sf.c | 10 +++++++++- src/mesa/pipe/i965simple/brw_wm_decl.c | 8 +++++++- 2 files changed, 16 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_sf.c b/src/mesa/pipe/i965simple/brw_sf.c index c04234ecf4..362196a0d1 100644 --- a/src/mesa/pipe/i965simple/brw_sf.c +++ b/src/mesa/pipe/i965simple/brw_sf.c @@ -58,7 +58,7 @@ static void compile_sf_prog( struct brw_context *brw, c.nr_attrs = c.key.vp_output_count; c.nr_attr_regs = (c.nr_attrs+1)/2; - c.nr_setup_attrs = c.key.fp_input_count; + c.nr_setup_attrs = c.key.fp_input_count + 1; /* +1 for position */ c.nr_setup_regs = (c.nr_setup_attrs+1)/2; c.prog_data.urb_read_length = c.nr_attr_regs; @@ -205,6 +205,14 @@ static void upload_sf_prog( struct brw_context *brw ) } } + /* Hack: Adjust for position. Optimize away when not required (ie + * for perspective interpolation). + */ + key.persp_mask <<= 1; + key.linear_mask <<= 1; + key.linear_mask |= 1; + key.const_mask <<= 1; + _mesa_printf("key.persp_mask: %x\n", key.persp_mask); _mesa_printf("key.linear_mask: %x\n", key.linear_mask); _mesa_printf("key.const_mask: %x\n", key.const_mask); diff --git a/src/mesa/pipe/i965simple/brw_wm_decl.c b/src/mesa/pipe/i965simple/brw_wm_decl.c index 5b1abb9d48..b45a333a2e 100644 --- a/src/mesa/pipe/i965simple/brw_wm_decl.c +++ b/src/mesa/pipe/i965simple/brw_wm_decl.c @@ -273,6 +273,12 @@ static void prealloc_reg(struct brw_wm_compile *c) c->reg_index += nr_curbe_regs; } + /* Adjust for parameter coefficients for position, which are + * currently always provided. + */ +// c->position_coef[i] = brw_vec8_grf(c->reg_index, 0); + c->reg_index += 2; + /* Next we receive the plane coefficients for parameter * interpolation: */ @@ -282,7 +288,7 @@ static void prealloc_reg(struct brw_wm_compile *c) } c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; - c->prog_data.urb_read_length = c->fp->program.num_inputs * 2; + c->prog_data.urb_read_length = (c->fp->program.num_inputs + 1) * 2; c->prog_data.curb_read_length = nr_curbe_regs; /* That's the end of the payload, now we can start allocating registers. -- cgit v1.2.3 From f54012650eebd940a6a7006ebcc0b9160c0368cf Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 3 Jan 2008 19:46:16 +0000 Subject: 965: use correct offset for constants vs immediates --- src/mesa/pipe/i965simple/brw_vs_emit.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index 0cc0a437b0..5f212bd055 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -630,8 +630,8 @@ static struct brw_reg get_reg( struct brw_vs_compile *c, assert(c->regs[file][index].nr != 0); return c->regs[file][index]; case TGSI_FILE_CONSTANT: - assert(c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_consts].nr != 0); - return c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_consts]; + assert(c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_imm].nr != 0); + return c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_imm]; case TGSI_FILE_IMMEDIATE: assert(c->regs[TGSI_FILE_CONSTANT][index].nr != 0); return c->regs[TGSI_FILE_CONSTANT][index]; -- cgit v1.2.3 From b1e67c5d840a2c07c4ccf73bf1d3200b8531efa3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 3 Jan 2008 19:46:59 +0000 Subject: 965: fix various refcount issues --- src/mesa/pipe/xlib/xm_winsys_aub.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index b207638390..2be8f8793d 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -99,6 +99,8 @@ static void *aub_buffer_map(struct pipe_winsys *winsys, { struct aub_buffer *sbo = aub_bo(buf); + assert(sbo->data); + if (flags & PIPE_BUFFER_FLAG_WRITE) sbo->dump_on_unmap = 1; @@ -116,6 +118,9 @@ static void aub_buffer_unmap(struct pipe_winsys *winsys, if (sbo->map_count == 0 && sbo->dump_on_unmap) { + + sbo->dump_on_unmap = 0; + brw_aub_gtt_data( iws->aubfile, sbo->offset, sbo->data, @@ -132,6 +137,7 @@ aub_buffer_reference(struct pipe_winsys *winsys, struct pipe_buffer_handle *buf) { if (*ptr) { + assert(aub_bo(*ptr)->refcount != 0); if (--(aub_bo(*ptr)->refcount) == 0) free(*ptr); *ptr = NULL; @@ -273,7 +279,9 @@ aub_buffer_create(struct pipe_winsys *winsys, unsigned flags, unsigned hint) { - return pipe_bo(CALLOC_STRUCT(aub_buffer)); + struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer); + sbo->refcount = 1; + return pipe_bo(sbo); } @@ -282,6 +290,8 @@ aub_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) { struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer); + sbo->refcount = 1; + /* Lets hope this is meant for upload, not as a result! */ aub_buffer_data( winsys, -- cgit v1.2.3 From 3385f3d250df1ad6899c0ac59a8f6b845cc5d504 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 4 Jan 2008 08:32:20 -0500 Subject: llvm: update llvm sources the latest svn --- src/mesa/pipe/llvm/gallivm_builtins.cpp | 342 +++++++----- src/mesa/pipe/llvm/instructions.cpp | 239 ++++---- src/mesa/pipe/llvm/llvm_base_shader.cpp | 960 +++++++++++++++++--------------- 3 files changed, 833 insertions(+), 708 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm_builtins.cpp b/src/mesa/pipe/llvm/gallivm_builtins.cpp index 48693ca2ed..1796f0a177 100644 --- a/src/mesa/pipe/llvm/gallivm_builtins.cpp +++ b/src/mesa/pipe/llvm/gallivm_builtins.cpp @@ -8,144 +8,170 @@ mod->setModuleIdentifier("shader"); // Type Definitions ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 25); -PointerType* PointerTy_1 = PointerType::get(ArrayTy_0); +PointerType* PointerTy_1 = PointerType::get(ArrayTy_0, 0); std::vectorFuncTy_2_args; FuncTy_2_args.push_back(Type::FloatTy); FuncTy_2_args.push_back(Type::FloatTy); -ParamAttrsList *FuncTy_2_PAL = 0; FunctionType* FuncTy_2 = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/FuncTy_2_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_2_PAL); + /*isVarArg=*/false); -PointerType* PointerTy_3 = PointerType::get(FuncTy_2); +PointerType* PointerTy_3 = PointerType::get(FuncTy_2, 0); VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); std::vectorFuncTy_5_args; FuncTy_5_args.push_back(VectorTy_4); -ParamAttrsList *FuncTy_5_PAL = 0; FunctionType* FuncTy_5 = FunctionType::get( /*Result=*/VectorTy_4, /*Params=*/FuncTy_5_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_5_PAL); + /*isVarArg=*/false); std::vectorFuncTy_6_args; FuncTy_6_args.push_back(VectorTy_4); FuncTy_6_args.push_back(VectorTy_4); FuncTy_6_args.push_back(VectorTy_4); -ParamAttrsList *FuncTy_6_PAL = 0; FunctionType* FuncTy_6 = FunctionType::get( /*Result=*/VectorTy_4, /*Params=*/FuncTy_6_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_6_PAL); + /*isVarArg=*/false); VectorType* VectorTy_7 = VectorType::get(IntegerType::get(32), 4); std::vectorFuncTy_9_args; -ParamAttrsList *FuncTy_9_PAL = 0; FunctionType* FuncTy_9 = FunctionType::get( /*Result=*/IntegerType::get(32), /*Params=*/FuncTy_9_args, - /*isVarArg=*/true, - /*ParamAttrs=*/FuncTy_9_PAL); + /*isVarArg=*/true); -PointerType* PointerTy_8 = PointerType::get(FuncTy_9); +PointerType* PointerTy_8 = PointerType::get(FuncTy_9, 0); -PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8)); +PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8), 0); std::vectorFuncTy_12_args; FuncTy_12_args.push_back(Type::FloatTy); -ParamAttrsList *FuncTy_12_PAL = 0; FunctionType* FuncTy_12 = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/FuncTy_12_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_12_PAL); + /*isVarArg=*/false); -PointerType* PointerTy_11 = PointerType::get(FuncTy_12); +PointerType* PointerTy_11 = PointerType::get(FuncTy_12, 0); std::vectorFuncTy_13_args; FuncTy_13_args.push_back(VectorTy_4); -ParamAttrsList *FuncTy_13_PAL = 0; FunctionType* FuncTy_13 = FunctionType::get( /*Result=*/IntegerType::get(32), /*Params=*/FuncTy_13_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_13_PAL); + /*isVarArg=*/false); // Function Declarations Function* func_approx = new Function( /*Type=*/FuncTy_2, - /*Linkage=*/GlobalValue::ExternalLinkage, + /*Linkage=*/GlobalValue::WeakLinkage, /*Name=*/"approx", mod); func_approx->setCallingConv(CallingConv::C); +const ParamAttrsList *func_approx_PAL = 0; +func_approx->setParamAttrs(func_approx_PAL); Function* func_powf = new Function( /*Type=*/FuncTy_2, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"powf", mod); // (external, no body) func_powf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_powf_PAL = 0; +func_powf->setParamAttrs(func_powf_PAL); Function* func_lit = new Function( /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::ExternalLinkage, + /*Linkage=*/GlobalValue::WeakLinkage, /*Name=*/"lit", mod); func_lit->setCallingConv(CallingConv::C); +const ParamAttrsList *func_lit_PAL = 0; +func_lit->setParamAttrs(func_lit_PAL); Function* func_cmp = new Function( /*Type=*/FuncTy_6, - /*Linkage=*/GlobalValue::ExternalLinkage, + /*Linkage=*/GlobalValue::WeakLinkage, /*Name=*/"cmp", mod); func_cmp->setCallingConv(CallingConv::C); +const ParamAttrsList *func_cmp_PAL = 0; +{ + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_cmp_PAL = ParamAttrsList::get(Attrs); + +} +func_cmp->setParamAttrs(func_cmp_PAL); Function* func_vcos = new Function( /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::ExternalLinkage, + /*Linkage=*/GlobalValue::WeakLinkage, /*Name=*/"vcos", mod); func_vcos->setCallingConv(CallingConv::C); +const ParamAttrsList *func_vcos_PAL = 0; +func_vcos->setParamAttrs(func_vcos_PAL); Function* func_printf = new Function( /*Type=*/FuncTy_9, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"printf", mod); // (external, no body) func_printf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_printf_PAL = 0; +func_printf->setParamAttrs(func_printf_PAL); Function* func_cosf = new Function( /*Type=*/FuncTy_12, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"cosf", mod); // (external, no body) func_cosf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_cosf_PAL = 0; +func_cosf->setParamAttrs(func_cosf_PAL); Function* func_scs = new Function( /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::ExternalLinkage, + /*Linkage=*/GlobalValue::WeakLinkage, /*Name=*/"scs", mod); func_scs->setCallingConv(CallingConv::C); +const ParamAttrsList *func_scs_PAL = 0; +func_scs->setParamAttrs(func_scs_PAL); Function* func_sinf = new Function( /*Type=*/FuncTy_12, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"sinf", mod); // (external, no body) func_sinf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_sinf_PAL = 0; +func_sinf->setParamAttrs(func_sinf_PAL); Function* func_vsin = new Function( /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::ExternalLinkage, + /*Linkage=*/GlobalValue::WeakLinkage, /*Name=*/"vsin", mod); func_vsin->setCallingConv(CallingConv::C); +const ParamAttrsList *func_vsin_PAL = 0; +func_vsin->setParamAttrs(func_vsin_PAL); Function* func_kilp = new Function( /*Type=*/FuncTy_13, - /*Linkage=*/GlobalValue::ExternalLinkage, + /*Linkage=*/GlobalValue::WeakLinkage, /*Name=*/"kilp", mod); func_kilp->setCallingConv(CallingConv::C); +const ParamAttrsList *func_kilp_PAL = 0; +{ + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_kilp_PAL = ParamAttrsList::get(Attrs); + +} +func_kilp->setParamAttrs(func_kilp_PAL); // Global Variable Declarations @@ -250,7 +276,9 @@ gvar_array__str1->setInitializer(const_array_15); float_call_params.push_back(float_b_addr_1); CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); float_call->setCallingConv(CallingConv::C); - float_call->setTailCall(true); + float_call->setTailCall(true);const ParamAttrsList *float_call_PAL = 0; + float_call->setParamAttrs(float_call_PAL); + new ReturnInst(float_call, label_entry); } @@ -266,28 +294,24 @@ gvar_array__str1->setInitializer(const_array_15); BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); // Block entry (label_entry_38) - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp7", label_entry_38); - FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp7, const_float_18, "cmp", label_entry_38); + ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp6", label_entry_38); + FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp6, const_float_18, "cmp", label_entry_38); new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_39, label_entry_38); // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp12 = new InsertElementInst(const_packed_20, float_tmp7, const_int32_23, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp14", label_ifthen); - ExtractElementInst* float_tmp16 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp16", label_ifthen); - FCmpInst* int1_cmp_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp16, const_float_16, "cmp.i", label_ifthen); - SelectInst* float_b_addr_0_i = new SelectInst(int1_cmp_i, const_float_16, float_tmp16, "b.addr.0.i", label_ifthen); - FCmpInst* int1_cmp3_i = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0_i, const_float_17, "cmp3.i", label_ifthen); - SelectInst* float_b_addr_1_i = new SelectInst(int1_cmp3_i, const_float_17, float_b_addr_0_i, "b.addr.1.i", label_ifthen); - FCmpInst* int1_cmp7_i = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp14, const_float_18, "cmp7.i", label_ifthen); - SelectInst* float_a_addr_0_i = new SelectInst(int1_cmp7_i, const_float_18, float_tmp14, "a.addr.0.i", label_ifthen); - std::vector float_call_i_params; - float_call_i_params.push_back(float_a_addr_0_i); - float_call_i_params.push_back(float_b_addr_1_i); - CallInst* float_call_i = new CallInst(func_powf, float_call_i_params.begin(), float_call_i_params.end(), "call.i", label_ifthen); - float_call_i->setCallingConv(CallingConv::C); - float_call_i->setTailCall(true); - InsertElementInst* packed_tmp18 = new InsertElementInst(packed_tmp12, float_call_i, const_int32_25, "tmp18", label_ifthen); - new ReturnInst(packed_tmp18, label_ifthen); + InsertElementInst* packed_tmp10 = new InsertElementInst(const_packed_20, float_tmp6, const_int32_23, "tmp10", label_ifthen); + ExtractElementInst* float_tmp12 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp14", label_ifthen); + std::vector float_call_41_params; + float_call_41_params.push_back(float_tmp12); + float_call_41_params.push_back(float_tmp14); + CallInst* float_call_41 = new CallInst(func_approx, float_call_41_params.begin(), float_call_41_params.end(), "call", label_ifthen); + float_call_41->setCallingConv(CallingConv::C); + float_call_41->setTailCall(true);const ParamAttrsList *float_call_41_PAL = 0; + float_call_41->setParamAttrs(float_call_41_PAL); + + InsertElementInst* packed_tmp16 = new InsertElementInst(packed_tmp10, float_call_41, const_int32_25, "tmp16", label_ifthen); + new ReturnInst(packed_tmp16, label_ifthen); // Block UnifiedReturnBlock (label_UnifiedReturnBlock) new ReturnInst(const_packed_26, label_UnifiedReturnBlock); @@ -304,7 +328,7 @@ gvar_array__str1->setInitializer(const_array_15); Value* packed_tmp2 = args++; packed_tmp2->setName("tmp2"); - BasicBlock* label_entry_43 = new BasicBlock("entry",func_cmp,0); + BasicBlock* label_entry_44 = new BasicBlock("entry",func_cmp,0); BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); @@ -312,15 +336,15 @@ gvar_array__str1->setInitializer(const_array_15); BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); - // Block entry (label_entry_43) - ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_43); - CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_43); - FCmpInst* int1_cmp_44 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_43); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_43); - CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_43); - FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_43); - SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_44, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_43); - new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_43); + // Block entry (label_entry_44) + ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_44); + CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_44); + FCmpInst* int1_cmp_45 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_44); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_44); + CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_44); + FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_44); + SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_45, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_44); + new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_44); // Block cond.?14 (label_cond__14) ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_28, "tmp233", label_cond__14); @@ -386,125 +410,155 @@ gvar_array__str1->setInitializer(const_array_15); Value* packed_val = args++; packed_val->setName("val"); - BasicBlock* label_entry_52 = new BasicBlock("entry",func_vcos,0); - - // Block entry (label_entry_52) - ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_52); - CastInst* double_conv_53 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_52); - ExtractElementInst* float_tmp3_54 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_52); - CastInst* double_conv4 = new FPExtInst(float_tmp3_54, Type::DoubleTy, "conv4", label_entry_52); - ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_52); - CastInst* double_conv7 = new FPExtInst(float_tmp6, Type::DoubleTy, "conv7", label_entry_52); - ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_52); - CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_52); + BasicBlock* label_entry_53 = new BasicBlock("entry",func_vcos,0); + + // Block entry (label_entry_53) + ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_53); + CastInst* double_conv_54 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_53); + ExtractElementInst* float_tmp3_55 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_53); + CastInst* double_conv4 = new FPExtInst(float_tmp3_55, Type::DoubleTy, "conv4", label_entry_53); + ExtractElementInst* float_tmp6_56 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_53); + CastInst* double_conv7 = new FPExtInst(float_tmp6_56, Type::DoubleTy, "conv7", label_entry_53); + ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_53); + CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_53); std::vector int32_call_params; int32_call_params.push_back(const_ptr_34); - int32_call_params.push_back(double_conv_53); + int32_call_params.push_back(double_conv_54); int32_call_params.push_back(double_conv4); int32_call_params.push_back(double_conv7); int32_call_params.push_back(double_conv10); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_52); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_53); int32_call->setCallingConv(CallingConv::C); - int32_call->setTailCall(true); - CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_52); + int32_call->setTailCall(true);const ParamAttrsList *int32_call_PAL = 0; + int32_call->setParamAttrs(int32_call_PAL); + + CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_53); float_call13->setCallingConv(CallingConv::C); - float_call13->setTailCall(true); - InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_52); - InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call13, const_int32_23, "tmp20", label_entry_52); - InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call13, const_int32_25, "tmp25", label_entry_52); - InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call13, const_int32_24, "tmp30", label_entry_52); - CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_52); + float_call13->setTailCall(true);const ParamAttrsList *float_call13_PAL = 0; + float_call13->setParamAttrs(float_call13_PAL); + + InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_53); + CallInst* float_call18 = new CallInst(func_cosf, float_tmp1, "call18", label_entry_53); + float_call18->setCallingConv(CallingConv::C); + float_call18->setTailCall(true);const ParamAttrsList *float_call18_PAL = 0; + float_call18->setParamAttrs(float_call18_PAL); + + InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call18, const_int32_23, "tmp20", label_entry_53); + CallInst* float_call23 = new CallInst(func_cosf, float_tmp1, "call23", label_entry_53); + float_call23->setCallingConv(CallingConv::C); + float_call23->setTailCall(true);const ParamAttrsList *float_call23_PAL = 0; + float_call23->setParamAttrs(float_call23_PAL); + + InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call23, const_int32_25, "tmp25", label_entry_53); + CallInst* float_call28 = new CallInst(func_cosf, float_tmp1, "call28", label_entry_53); + float_call28->setCallingConv(CallingConv::C); + float_call28->setTailCall(true);const ParamAttrsList *float_call28_PAL = 0; + float_call28->setParamAttrs(float_call28_PAL); + + InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call28, const_int32_24, "tmp30", label_entry_53); + CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_53); + CastInst* double_conv36 = new FPExtInst(float_call18, Type::DoubleTy, "conv36", label_entry_53); + CastInst* double_conv39 = new FPExtInst(float_call23, Type::DoubleTy, "conv39", label_entry_53); + CastInst* double_conv42 = new FPExtInst(float_call28, Type::DoubleTy, "conv42", label_entry_53); std::vector int32_call43_params; int32_call43_params.push_back(const_ptr_36); int32_call43_params.push_back(double_conv33); - int32_call43_params.push_back(double_conv33); - int32_call43_params.push_back(double_conv33); - int32_call43_params.push_back(double_conv33); - CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_52); + int32_call43_params.push_back(double_conv36); + int32_call43_params.push_back(double_conv39); + int32_call43_params.push_back(double_conv42); + CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_53); int32_call43->setCallingConv(CallingConv::C); - int32_call43->setTailCall(true); - new ReturnInst(packed_tmp30, label_entry_52); + int32_call43->setTailCall(true);const ParamAttrsList *int32_call43_PAL = 0; + int32_call43->setParamAttrs(int32_call43_PAL); + + new ReturnInst(packed_tmp30, label_entry_53); } // Function: scs (func_scs) { Function::arg_iterator args = func_scs->arg_begin(); - Value* packed_val_56 = args++; - packed_val_56->setName("val"); - - BasicBlock* label_entry_57 = new BasicBlock("entry",func_scs,0); - - // Block entry (label_entry_57) - ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_56, const_int32_19, "tmp2", label_entry_57); - CallInst* float_call_58 = new CallInst(func_cosf, float_tmp2, "call", label_entry_57); - float_call_58->setCallingConv(CallingConv::C); - float_call_58->setTailCall(true); - InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_58, const_int32_19, "tmp5", label_entry_57); - CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_57); + Value* packed_val_58 = args++; + packed_val_58->setName("val"); + + BasicBlock* label_entry_59 = new BasicBlock("entry",func_scs,0); + + // Block entry (label_entry_59) + ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_58, const_int32_19, "tmp2", label_entry_59); + CallInst* float_call_60 = new CallInst(func_cosf, float_tmp2, "call", label_entry_59); + float_call_60->setCallingConv(CallingConv::C); + float_call_60->setTailCall(true);const ParamAttrsList *float_call_60_PAL = 0; + float_call_60->setParamAttrs(float_call_60_PAL); + + InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_60, const_int32_19, "tmp5", label_entry_59); + CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_59); float_call7->setCallingConv(CallingConv::C); - float_call7->setTailCall(true); - InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_57); - new ReturnInst(packed_tmp9, label_entry_57); + float_call7->setTailCall(true);const ParamAttrsList *float_call7_PAL = 0; + float_call7->setParamAttrs(float_call7_PAL); + + InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_59); + new ReturnInst(packed_tmp9, label_entry_59); } // Function: vsin (func_vsin) { Function::arg_iterator args = func_vsin->arg_begin(); - Value* packed_val_60 = args++; - packed_val_60->setName("val"); - - BasicBlock* label_entry_61 = new BasicBlock("entry",func_vsin,0); - - // Block entry (label_entry_61) - ExtractElementInst* float_tmp2_62 = new ExtractElementInst(packed_val_60, const_int32_19, "tmp2", label_entry_61); - CallInst* float_call_63 = new CallInst(func_sinf, float_tmp2_62, "call", label_entry_61); - float_call_63->setCallingConv(CallingConv::C); - float_call_63->setTailCall(true); - InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_63, const_int32_19, "tmp6", label_entry_61); - InsertElementInst* packed_tmp9_64 = new InsertElementInst(packed_tmp6, float_call_63, const_int32_23, "tmp9", label_entry_61); - InsertElementInst* packed_tmp12_65 = new InsertElementInst(packed_tmp9_64, float_call_63, const_int32_25, "tmp12", label_entry_61); - InsertElementInst* packed_tmp15_66 = new InsertElementInst(packed_tmp12_65, float_call_63, const_int32_24, "tmp15", label_entry_61); - new ReturnInst(packed_tmp15_66, label_entry_61); + Value* packed_val_62 = args++; + packed_val_62->setName("val"); + + BasicBlock* label_entry_63 = new BasicBlock("entry",func_vsin,0); + + // Block entry (label_entry_63) + ExtractElementInst* float_tmp2_64 = new ExtractElementInst(packed_val_62, const_int32_19, "tmp2", label_entry_63); + CallInst* float_call_65 = new CallInst(func_sinf, float_tmp2_64, "call", label_entry_63); + float_call_65->setCallingConv(CallingConv::C); + float_call_65->setTailCall(true);const ParamAttrsList *float_call_65_PAL = 0; + float_call_65->setParamAttrs(float_call_65_PAL); + + InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_65, const_int32_19, "tmp6", label_entry_63); + InsertElementInst* packed_tmp9_66 = new InsertElementInst(packed_tmp6, float_call_65, const_int32_23, "tmp9", label_entry_63); + InsertElementInst* packed_tmp12 = new InsertElementInst(packed_tmp9_66, float_call_65, const_int32_25, "tmp12", label_entry_63); + InsertElementInst* packed_tmp15_67 = new InsertElementInst(packed_tmp12, float_call_65, const_int32_24, "tmp15", label_entry_63); + new ReturnInst(packed_tmp15_67, label_entry_63); } // Function: kilp (func_kilp) { Function::arg_iterator args = func_kilp->arg_begin(); - Value* packed_val_68 = args++; - packed_val_68->setName("val"); + Value* packed_val_69 = args++; + packed_val_69->setName("val"); - BasicBlock* label_entry_69 = new BasicBlock("entry",func_kilp,0); + BasicBlock* label_entry_70 = new BasicBlock("entry",func_kilp,0); BasicBlock* label_lor_rhs = new BasicBlock("lor_rhs",func_kilp,0); - BasicBlock* label_lor_rhs6 = new BasicBlock("lor_rhs6",func_kilp,0); - BasicBlock* label_lor_rhs13 = new BasicBlock("lor_rhs13",func_kilp,0); - BasicBlock* label_UnifiedReturnBlock_70 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); + BasicBlock* label_lor_rhs5 = new BasicBlock("lor_rhs5",func_kilp,0); + BasicBlock* label_lor_rhs11 = new BasicBlock("lor_rhs11",func_kilp,0); + BasicBlock* label_UnifiedReturnBlock_71 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); - // Block entry (label_entry_69) - ExtractElementInst* float_tmp1_71 = new ExtractElementInst(packed_val_68, const_int32_19, "tmp1", label_entry_69); - FCmpInst* int1_cmp_72 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_71, const_float_18, "cmp", label_entry_69); - new BranchInst(label_UnifiedReturnBlock_70, label_lor_rhs, int1_cmp_72, label_entry_69); + // Block entry (label_entry_70) + ExtractElementInst* float_tmp1_72 = new ExtractElementInst(packed_val_69, const_int32_19, "tmp1", label_entry_70); + FCmpInst* int1_cmp_73 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_72, const_float_18, "cmp", label_entry_70); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs, int1_cmp_73, label_entry_70); // Block lor_rhs (label_lor_rhs) - ExtractElementInst* float_tmp3_74 = new ExtractElementInst(packed_val_68, const_int32_23, "tmp3", label_lor_rhs); - FCmpInst* int1_cmp5 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_74, const_float_18, "cmp5", label_lor_rhs); - new BranchInst(label_UnifiedReturnBlock_70, label_lor_rhs6, int1_cmp5, label_lor_rhs); - - // Block lor_rhs6 (label_lor_rhs6) - ExtractElementInst* float_tmp8 = new ExtractElementInst(packed_val_68, const_int32_25, "tmp8", label_lor_rhs6); - FCmpInst* int1_cmp10 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp8, const_float_18, "cmp10", label_lor_rhs6); - new BranchInst(label_UnifiedReturnBlock_70, label_lor_rhs13, int1_cmp10, label_lor_rhs6); - - // Block lor_rhs13 (label_lor_rhs13) - ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_val_68, const_int32_24, "tmp15", label_lor_rhs13); - FCmpInst* int1_cmp17 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp15, const_float_18, "cmp17", label_lor_rhs13); - CastInst* int32_retval = new ZExtInst(int1_cmp17, IntegerType::get(32), "retval", label_lor_rhs13); - new ReturnInst(int32_retval, label_lor_rhs13); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock_70) - new ReturnInst(const_int32_23, label_UnifiedReturnBlock_70); + ExtractElementInst* float_tmp3_75 = new ExtractElementInst(packed_val_69, const_int32_23, "tmp3", label_lor_rhs); + FCmpInst* int1_cmp4 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_75, const_float_18, "cmp4", label_lor_rhs); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs5, int1_cmp4, label_lor_rhs); + + // Block lor_rhs5 (label_lor_rhs5) + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_val_69, const_int32_25, "tmp7", label_lor_rhs5); + FCmpInst* int1_cmp8 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp7, const_float_18, "cmp8", label_lor_rhs5); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs11, int1_cmp8, label_lor_rhs5); + + // Block lor_rhs11 (label_lor_rhs11) + ExtractElementInst* float_tmp13 = new ExtractElementInst(packed_val_69, const_int32_24, "tmp13", label_lor_rhs11); + FCmpInst* int1_cmp14 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp13, const_float_18, "cmp14", label_lor_rhs11); + CastInst* int32_retval = new ZExtInst(int1_cmp14, IntegerType::get(32), "retval", label_lor_rhs11); + new ReturnInst(int32_retval, label_lor_rhs11); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock_71) + new ReturnInst(const_int32_23, label_UnifiedReturnBlock_71); } diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp index c8d1992587..55d39fa5f1 100644 --- a/src/mesa/pipe/llvm/instructions.cpp +++ b/src/mesa/pipe/llvm/instructions.cpp @@ -41,6 +41,7 @@ #include #include #include +#include #include #include @@ -103,13 +104,13 @@ const char * Instructions::name(const char *prefix) llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2) { Value *mulRes = mul(in1, in2); - ExtractElementInst *x = m_builder.CreateExtractElement(mulRes, + Value *x = m_builder.CreateExtractElement(mulRes, m_storage->constantInt(0), name("extractx")); - ExtractElementInst *y = m_builder.CreateExtractElement(mulRes, + Value *y = m_builder.CreateExtractElement(mulRes, m_storage->constantInt(1), name("extracty")); - ExtractElementInst *z = m_builder.CreateExtractElement(mulRes, + Value *z = m_builder.CreateExtractElement(mulRes, m_storage->constantInt(2), name("extractz")); Value *xy = m_builder.CreateAdd(x, y,name("xy")); @@ -127,13 +128,13 @@ llvm::Value *Instructions::callFSqrt(llvm::Value *val) FunctionType* fsqrtType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/fsqrtArgs, - /*isVarArg=*/false, - /*ParamAttrs=*/fsqrtPal); + /*isVarArg=*/false); m_llvmFSqrt = new Function( /*Type=*/fsqrtType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"llvm.sqrt.f32", m_mod); m_llvmFSqrt->setCallingConv(CallingConv::C); + m_llvmFSqrt->setParamAttrs(fsqrtPal); } CallInst *call = m_builder.CreateCall(m_llvmFSqrt, val, name("sqrt")); @@ -144,9 +145,9 @@ llvm::Value *Instructions::callFSqrt(llvm::Value *val) llvm::Value * Instructions::rsq(llvm::Value *in1) { - ExtractElementInst *x = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("extractx")); + Value *x = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("extractx")); Value *abs = callFAbs(x); Value *sqrt = callFSqrt(abs); @@ -161,9 +162,9 @@ llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, llvm::Value *z, llvm::Value *w) { Constant *const_vec = Constant::getNullValue(m_floatVecType); - InsertElementInst *res = m_builder.CreateInsertElement(const_vec, x, - m_storage->constantInt(0), - name("vecx")); + Value *res = m_builder.CreateInsertElement(const_vec, x, + m_storage->constantInt(0), + name("vecx")); res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), name("vecxy")); res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), @@ -184,13 +185,13 @@ llvm::Value *Instructions::callFAbs(llvm::Value *val) FunctionType* fabsType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/fabsArgs, - /*isVarArg=*/false, - /*ParamAttrs=*/fabsPal); + /*isVarArg=*/false); m_llvmFAbs = new Function( /*Type=*/fabsType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"fabs", m_mod); m_llvmFAbs->setCallingConv(CallingConv::C); + m_llvmFAbs->setParamAttrs(fabsPal); } CallInst *call = m_builder.CreateCall(m_llvmFAbs, val, name("fabs")); @@ -227,13 +228,13 @@ llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) FunctionType* powType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/powArgs, - /*isVarArg=*/false, - /*ParamAttrs=*/powPal); + /*isVarArg=*/false); m_llvmPow = new Function( /*Type=*/powType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"llvm.pow.f32", m_mod); m_llvmPow->setCallingConv(CallingConv::C); + m_llvmPow->setParamAttrs(powPal); } std::vector params; params.push_back(val1); @@ -247,21 +248,21 @@ llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2) { - ExtractElementInst *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - ExtractElementInst *x2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(0), - name("x2")); + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); llvm::Value *val = callPow(x1, x2); return vectorFromVals(val, val, val, val); } llvm::Value * Instructions::rcp(llvm::Value *in1) { - ExtractElementInst *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); Value *res = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, APFloat(1.f)), x1, name("rcp")); @@ -290,18 +291,18 @@ llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) { - ExtractElementInst *y1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(1), - name("y1")); - ExtractElementInst *z = m_builder.CreateExtractElement(in1, - m_storage->constantInt(2), - name("z")); - ExtractElementInst *y2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(1), - name("y2")); - ExtractElementInst *w = m_builder.CreateExtractElement(in2, - m_storage->constantInt(3), - name("w")); + Value *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + Value *z = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z")); + Value *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + Value *w = m_builder.CreateExtractElement(in2, + m_storage->constantInt(3), + name("w")); Value *ry = m_builder.CreateMul(y1, y2, name("tyuy")); return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), ry, z, w); @@ -326,13 +327,13 @@ llvm::Value * Instructions::callFloor(llvm::Value *val) FunctionType* floorType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/floorArgs, - /*isVarArg=*/false, - /*ParamAttrs=*/floorPal); + /*isVarArg=*/false); m_llvmFloor = new Function( /*Type=*/floorType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"floorf", m_mod); m_llvmFloor->setCallingConv(CallingConv::C); + m_llvmFloor->setParamAttrs(floorPal); } CallInst *call = m_builder.CreateCall(m_llvmFloor, val, name("floorf")); @@ -369,13 +370,13 @@ llvm::Value * Instructions::callFLog(llvm::Value *val) FunctionType* flogType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/flogArgs, - /*isVarArg=*/false, - /*ParamAttrs=*/flogPal); + /*isVarArg=*/false); m_llvmFlog = new Function( /*Type=*/flogType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"logf", m_mod); m_llvmFlog->setCallingConv(CallingConv::C); + m_llvmFlog->setParamAttrs(flogPal); } CallInst *call = m_builder.CreateCall(m_llvmFlog, val, name("logf")); @@ -399,20 +400,20 @@ llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) std::vector vec2 = extractVector(in2); Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); - SelectInst *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], - name("selx")); + Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); - SelectInst *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], - name("sely")); + Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); - SelectInst *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], - name("selz")); + Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); - SelectInst *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], - name("selw")); + Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); return vectorFromVals(selx, sely, selz, selw); } @@ -423,24 +424,24 @@ llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) std::vector vec2 = extractVector(in2); Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], - name("xcmp")); - SelectInst *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], - name("selx")); + name("xcmp")); + Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], - name("ycmp")); - SelectInst *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], - name("sely")); + name("ycmp")); + Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], - name("zcmp")); - SelectInst *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], - name("selz")); + name("zcmp")); + Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], - name("wcmp")); - SelectInst *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], - name("selw")); + name("wcmp")); + Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); return vectorFromVals(selx, sely, selz, selw); } @@ -474,17 +475,17 @@ void Instructions::printVector(llvm::Value *val) func_printf = declarePrintf(); assert(func_printf); std::vector vec = extractVector(val); - CastInst *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx")); - CastInst *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy")); - CastInst *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz")); - CastInst *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw")); + Value *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx")); + Value *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy")); + Value *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz")); + Value *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw")); std::vector params; params.push_back(m_fmtPtr); params.push_back(dx); params.push_back(dy); params.push_back(dz); params.push_back(dw); - CallInst* call = m_builder.CreateCall(func_printf, params.begin(), params.end(), + CallInst *call = m_builder.CreateCall(func_printf, params.begin(), params.end(), name("printf")); call->setCallingConv(CallingConv::C); call->setTailCall(true); @@ -497,13 +498,13 @@ llvm::Function * Instructions::declarePrintf() FunctionType* funcTy = FunctionType::get( /*Result=*/IntegerType::get(32), /*Params=*/args, - /*isVarArg=*/true, - /*ParamAttrs=*/params); + /*isVarArg=*/true); Function* func_printf = new Function( /*Type=*/funcTy, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"printf", m_mod); func_printf->setCallingConv(CallingConv::C); + func_printf->setParamAttrs(params); return func_printf; } @@ -516,16 +517,16 @@ llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) std::vector vec1 = extractVector(in1); std::vector vec2 = extractVector(in2); Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], name("xcmp")); - SelectInst *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], name("ycmp")); - SelectInst *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], name("zcmp")); - SelectInst *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], name("wcmp")); - SelectInst *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); return vectorFromVals(x, y, z, w); } @@ -538,16 +539,16 @@ llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) std::vector vec2 = extractVector(in2); Value *xcmp = m_builder.CreateFCmpOGE(vec1[0], vec2[0], name("xcmp")); - SelectInst *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); Value *ycmp = m_builder.CreateFCmpOGE(vec1[1], vec2[1], name("ycmp")); - SelectInst *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); Value *zcmp = m_builder.CreateFCmpOGE(vec1[2], vec2[2], name("zcmp")); - SelectInst *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); Value *wcmp = m_builder.CreateFCmpOGE(vec1[3], vec2[3], name("wcmp")); - SelectInst *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); return vectorFromVals(x, y, z, w); } @@ -562,41 +563,41 @@ llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) std::vector vec2 = extractVector(in2); Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); - SelectInst *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); - SelectInst *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); - SelectInst *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); - SelectInst *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); return vectorFromVals(x, y, z, w); } llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) { - ExtractElementInst *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - ExtractElementInst *y1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(1), - name("y1")); - ExtractElementInst *z1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(2), - name("z1")); - - ExtractElementInst *x2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(0), - name("x2")); - ExtractElementInst *y2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(1), - name("y2")); - ExtractElementInst *z2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(2), - name("z2")); + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + Value *z1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z1")); + + Value *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); + Value *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + Value *z2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(2), + name("z2")); Value *y1z2 = mul(y1, z2); Value *z1y2 = mul(z1, y2); @@ -631,8 +632,8 @@ void Instructions::ifop(llvm::Value *in) Constant *float0 = Constant::getNullValue(Type::FloatTy); - ExtractElementInst *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0), - name("extractx")); + Value *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0), + name("extractx")); Value *xcmp = m_builder.CreateFCmpUNE(x, float0, name("xcmp")); m_builder.CreateCondBr(xcmp, ifthen, ifend); //m_builder.SetInsertPoint(yblock); @@ -708,22 +709,22 @@ void Instructions::brk() llvm::Value * Instructions::trunc(llvm::Value *in) { std::vector vec = extractVector(in); - CastInst *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32), - name("ftoix")); - CastInst *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32), - name("ftoiy")); - CastInst *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32), - name("ftoiz")); - CastInst *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32), - name("ftoiw")); - CastInst *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy, - name("fx")); - CastInst *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy, - name("fy")); - CastInst *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy, - name("fz")); - CastInst *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy, - name("fw")); + Value *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32), + name("ftoix")); + Value *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32), + name("ftoiy")); + Value *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32), + name("ftoiz")); + Value *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32), + name("ftoiw")); + Value *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy, + name("fx")); + Value *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy, + name("fy")); + Value *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy, + name("fz")); + Value *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy, + name("fw")); return vectorFromVals(fx, fy, fz, fw); } @@ -743,7 +744,7 @@ void Instructions::cal(int label, llvm::Value *input) llvm::Function * Instructions::declareFunc(int label) { - PointerType *vecPtr = PointerType::get(m_floatVecType); + PointerType *vecPtr = PointerType::getUnqual(m_floatVecType); std::vector args; args.push_back(vecPtr); args.push_back(vecPtr); @@ -753,14 +754,14 @@ llvm::Function * Instructions::declareFunc(int label) FunctionType *funcType = FunctionType::get( /*Result=*/Type::VoidTy, /*Params=*/args, - /*isVarArg=*/false, - /*ParamAttrs=*/params); + /*isVarArg=*/false); std::string name = createFuncName(label); Function *func = new Function( /*Type=*/funcType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/name.c_str(), m_mod); func->setCallingConv(CallingConv::C); + func->setParamAttrs(params); return func; } diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index 85fa389d79..951afc390e 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -4,14 +4,14 @@ Module* createBaseShader() { // Module Construction Module* mod = new Module("Shader"); - mod->setDataLayout(""); - mod->setTargetTriple("i686-apple-darwin9"); + mod->setDataLayout("e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"); + mod->setTargetTriple("i686-pc-linux-gnu"); // Type Definitions std::vectorStructTy_struct_ShaderInput_fields; VectorType* VectorTy_1 = VectorType::get(Type::FloatTy, 4); - PointerType* PointerTy_0 = PointerType::get(VectorTy_1); + PointerType* PointerTy_0 = PointerType::get(VectorTy_1, 0); StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); @@ -21,9 +21,6 @@ Module* createBaseShader() { StructType* StructTy_struct_ShaderInput = StructType::get(StructTy_struct_ShaderInput_fields, /*isPacked=*/false); mod->addTypeName("struct.ShaderInput", StructTy_struct_ShaderInput); - OpaqueType* OpaqueTy_struct_pipe_mipmap_tree = OpaqueType::get(); - mod->addTypeName("struct.pipe_mipmap_tree", OpaqueTy_struct_pipe_mipmap_tree); - OpaqueType* OpaqueTy_struct_pipe_sampler_state = OpaqueType::get(); mod->addTypeName("struct.pipe_sampler_state", OpaqueTy_struct_pipe_sampler_state); @@ -31,199 +28,217 @@ Module* createBaseShader() { mod->addTypeName("struct.softpipe_tile_cache", OpaqueTy_struct_softpipe_tile_cache); std::vectorStructTy_struct_tgsi_sampler_fields; - PointerType* PointerTy_2 = PointerType::get(OpaqueTy_struct_pipe_sampler_state); + PointerType* PointerTy_2 = PointerType::get(OpaqueTy_struct_pipe_sampler_state, 0); StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_2); - PointerType* PointerTy_3 = PointerType::get(OpaqueTy_struct_pipe_mipmap_tree); - - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_3); - std::vectorFuncTy_5_args; + std::vectorFuncTy_4_args; PATypeHolder StructTy_struct_tgsi_sampler_fwd = OpaqueType::get(); - PointerType* PointerTy_6 = PointerType::get(StructTy_struct_tgsi_sampler_fwd); + PointerType* PointerTy_5 = PointerType::get(StructTy_struct_tgsi_sampler_fwd, 0); - FuncTy_5_args.push_back(PointerTy_6); - PointerType* PointerTy_7 = PointerType::get(Type::FloatTy); + FuncTy_4_args.push_back(PointerTy_5); + PointerType* PointerTy_6 = PointerType::get(Type::FloatTy, 0); - FuncTy_5_args.push_back(PointerTy_7); - FuncTy_5_args.push_back(PointerTy_7); - FuncTy_5_args.push_back(PointerTy_7); - FuncTy_5_args.push_back(Type::FloatTy); - ArrayType* ArrayTy_9 = ArrayType::get(Type::FloatTy, 4); + FuncTy_4_args.push_back(PointerTy_6); + FuncTy_4_args.push_back(PointerTy_6); + FuncTy_4_args.push_back(PointerTy_6); + FuncTy_4_args.push_back(Type::FloatTy); + ArrayType* ArrayTy_8 = ArrayType::get(Type::FloatTy, 4); - PointerType* PointerTy_8 = PointerType::get(ArrayTy_9); + PointerType* PointerTy_7 = PointerType::get(ArrayTy_8, 0); - FuncTy_5_args.push_back(PointerTy_8); - ParamAttrsList *FuncTy_5_PAL = 0; - FunctionType* FuncTy_5 = FunctionType::get( + FuncTy_4_args.push_back(PointerTy_7); + FunctionType* FuncTy_4 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_5_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_5_PAL); + /*Params=*/FuncTy_4_args, + /*isVarArg=*/false); - PointerType* PointerTy_4 = PointerType::get(FuncTy_5); + PointerType* PointerTy_3 = PointerType::get(FuncTy_4, 0); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_4); - PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8)); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_3); + PointerType* PointerTy_9 = PointerType::get(IntegerType::get(8), 0); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_10); - PointerType* PointerTy_11 = PointerType::get(OpaqueTy_struct_softpipe_tile_cache); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_9); + PointerType* PointerTy_10 = PointerType::get(OpaqueTy_struct_softpipe_tile_cache, 0); - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_11); + StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_10); StructType* StructTy_struct_tgsi_sampler = StructType::get(StructTy_struct_tgsi_sampler_fields, /*isPacked=*/false); mod->addTypeName("struct.tgsi_sampler", StructTy_struct_tgsi_sampler); cast(StructTy_struct_tgsi_sampler_fwd.get())->refineAbstractTypeTo(StructTy_struct_tgsi_sampler); StructTy_struct_tgsi_sampler = cast(StructTy_struct_tgsi_sampler_fwd.get()); - std::vectorFuncTy_12_args; - ArrayType* ArrayTy_14 = ArrayType::get(VectorTy_1, 16); + std::vectorFuncTy_11_args; + ArrayType* ArrayTy_13 = ArrayType::get(VectorTy_1, 16); - PointerType* PointerTy_13 = PointerType::get(ArrayTy_14); + PointerType* PointerTy_12 = PointerType::get(ArrayTy_13, 0); - FuncTy_12_args.push_back(PointerTy_13); - ArrayType* ArrayTy_16 = ArrayType::get(ArrayTy_9, 16); + FuncTy_11_args.push_back(PointerTy_12); + ArrayType* ArrayTy_15 = ArrayType::get(ArrayTy_8, 16); - PointerType* PointerTy_15 = PointerType::get(ArrayTy_16); + PointerType* PointerTy_14 = PointerType::get(ArrayTy_15, 0); - FuncTy_12_args.push_back(PointerTy_15); - FuncTy_12_args.push_back(IntegerType::get(32)); - FuncTy_12_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_12_PAL = 0; - FunctionType* FuncTy_12 = FunctionType::get( + FuncTy_11_args.push_back(PointerTy_14); + FuncTy_11_args.push_back(IntegerType::get(32)); + FuncTy_11_args.push_back(IntegerType::get(32)); + FunctionType* FuncTy_11 = FunctionType::get( /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_12_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_12_PAL); + /*Params=*/FuncTy_11_args, + /*isVarArg=*/false); + + std::vectorFuncTy_16_args; + FuncTy_16_args.push_back(PointerTy_0); + FuncTy_16_args.push_back(PointerTy_7); + FuncTy_16_args.push_back(IntegerType::get(32)); + FunctionType* FuncTy_16 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_16_args, + /*isVarArg=*/false); std::vectorFuncTy_17_args; + FuncTy_17_args.push_back(PointerTy_7); FuncTy_17_args.push_back(PointerTy_0); - FuncTy_17_args.push_back(PointerTy_8); FuncTy_17_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_17_PAL = 0; FunctionType* FuncTy_17 = FunctionType::get( /*Result=*/Type::VoidTy, /*Params=*/FuncTy_17_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_17_PAL); + /*isVarArg=*/false); std::vectorFuncTy_18_args; - FuncTy_18_args.push_back(PointerTy_8); - FuncTy_18_args.push_back(PointerTy_0); + FuncTy_18_args.push_back(PointerTy_12); + FuncTy_18_args.push_back(PointerTy_12); + FuncTy_18_args.push_back(PointerTy_7); + FuncTy_18_args.push_back(IntegerType::get(32)); + FuncTy_18_args.push_back(IntegerType::get(32)); + FuncTy_18_args.push_back(IntegerType::get(32)); FuncTy_18_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_18_PAL = 0; FunctionType* FuncTy_18 = FunctionType::get( /*Result=*/Type::VoidTy, /*Params=*/FuncTy_18_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_18_PAL); - - std::vectorFuncTy_19_args; - FuncTy_19_args.push_back(PointerTy_13); - FuncTy_19_args.push_back(PointerTy_13); - FuncTy_19_args.push_back(PointerTy_8); - FuncTy_19_args.push_back(IntegerType::get(32)); - FuncTy_19_args.push_back(IntegerType::get(32)); - FuncTy_19_args.push_back(IntegerType::get(32)); - FuncTy_19_args.push_back(IntegerType::get(32)); - ParamAttrsList *FuncTy_19_PAL = 0; - FunctionType* FuncTy_19 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_19_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_19_PAL); + /*isVarArg=*/false); - ArrayType* ArrayTy_21 = ArrayType::get(VectorTy_1, 32); + ArrayType* ArrayTy_20 = ArrayType::get(VectorTy_1, 32); - PointerType* PointerTy_20 = PointerType::get(ArrayTy_21); + PointerType* PointerTy_19 = PointerType::get(ArrayTy_20, 0); - ArrayType* ArrayTy_23 = ArrayType::get(VectorTy_1, 128); + ArrayType* ArrayTy_22 = ArrayType::get(VectorTy_1, 128); - PointerType* PointerTy_22 = PointerType::get(ArrayTy_23); + PointerType* PointerTy_21 = PointerType::get(ArrayTy_22, 0); - PointerType* PointerTy_24 = PointerType::get(StructTy_struct_ShaderInput); + PointerType* PointerTy_23 = PointerType::get(StructTy_struct_ShaderInput, 0); - PointerType* PointerTy_25 = PointerType::get(PointerTy_0); + PointerType* PointerTy_24 = PointerType::get(PointerTy_0, 0); + + std::vectorFuncTy_26_args; + FuncTy_26_args.push_back(PointerTy_23); + FunctionType* FuncTy_26 = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/FuncTy_26_args, + /*isVarArg=*/false); + + PointerType* PointerTy_25 = PointerType::get(FuncTy_26, 0); std::vectorFuncTy_27_args; - FuncTy_27_args.push_back(PointerTy_24); - ParamAttrsList *FuncTy_27_PAL = 0; + FuncTy_27_args.push_back(Type::FloatTy); + FuncTy_27_args.push_back(Type::FloatTy); + FuncTy_27_args.push_back(PointerTy_12); + FuncTy_27_args.push_back(PointerTy_12); + FuncTy_27_args.push_back(IntegerType::get(32)); + FuncTy_27_args.push_back(PointerTy_7); + FuncTy_27_args.push_back(IntegerType::get(32)); + FuncTy_27_args.push_back(PointerTy_5); FunctionType* FuncTy_27 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_27_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_27_PAL); - - PointerType* PointerTy_26 = PointerType::get(FuncTy_27); - - std::vectorFuncTy_28_args; - FuncTy_28_args.push_back(Type::FloatTy); - FuncTy_28_args.push_back(Type::FloatTy); - FuncTy_28_args.push_back(PointerTy_13); - FuncTy_28_args.push_back(PointerTy_13); - FuncTy_28_args.push_back(IntegerType::get(32)); - FuncTy_28_args.push_back(PointerTy_8); - FuncTy_28_args.push_back(IntegerType::get(32)); - FuncTy_28_args.push_back(PointerTy_6); - PointerType* PointerTy_29 = PointerType::get(IntegerType::get(32)); - - FuncTy_28_args.push_back(PointerTy_29); - ParamAttrsList *FuncTy_28_PAL = 0; - FunctionType* FuncTy_28 = FunctionType::get( /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_28_args, - /*isVarArg=*/false, - /*ParamAttrs=*/FuncTy_28_PAL); + /*Params=*/FuncTy_27_args, + /*isVarArg=*/false); + + PointerType* PointerTy_28 = PointerType::get(IntegerType::get(32), 0); // Function Declarations Function* func_from_array = new Function( - /*Type=*/FuncTy_12, + /*Type=*/FuncTy_11, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_array", mod); func_from_array->setCallingConv(CallingConv::C); + const ParamAttrsList *func_from_array_PAL = 0; + { + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_from_array_PAL = ParamAttrsList::get(Attrs); + + } + func_from_array->setParamAttrs(func_from_array_PAL); Function* func_from_consts = new Function( - /*Type=*/FuncTy_17, + /*Type=*/FuncTy_16, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"from_consts", mod); func_from_consts->setCallingConv(CallingConv::C); + const ParamAttrsList *func_from_consts_PAL = 0; + { + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_from_consts_PAL = ParamAttrsList::get(Attrs); + + } + func_from_consts->setParamAttrs(func_from_consts_PAL); Function* func_to_array = new Function( - /*Type=*/FuncTy_18, + /*Type=*/FuncTy_17, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"to_array", mod); func_to_array->setCallingConv(CallingConv::C); + const ParamAttrsList *func_to_array_PAL = 0; + { + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_to_array_PAL = ParamAttrsList::get(Attrs); + + } + func_to_array->setParamAttrs(func_to_array_PAL); Function* func_run_vertex_shader = new Function( - /*Type=*/FuncTy_19, + /*Type=*/FuncTy_18, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_vertex_shader", mod); func_run_vertex_shader->setCallingConv(CallingConv::C); + ParamAttrsList *func_run_vertex_shader_PAL = 0; + func_run_vertex_shader->setParamAttrs(func_run_vertex_shader_PAL); Function* func_execute_shader = new Function( - /*Type=*/FuncTy_27, + /*Type=*/FuncTy_26, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"execute_shader", mod); // (external, no body) func_execute_shader->setCallingConv(CallingConv::C); + ParamAttrsList *func_execute_shader_PAL = 0; + func_execute_shader->setParamAttrs(func_execute_shader_PAL); Function* func_run_fragment_shader = new Function( - /*Type=*/FuncTy_28, + /*Type=*/FuncTy_27, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_fragment_shader", mod); func_run_fragment_shader->setCallingConv(CallingConv::C); + ParamAttrsList *func_run_fragment_shader_PAL = 0; + func_run_fragment_shader->setParamAttrs(func_run_fragment_shader_PAL); // Global Variable Declarations // Constant Definitions - Constant* const_int32_30 = Constant::getNullValue(IntegerType::get(32)); - UndefValue* const_packed_31 = UndefValue::get(VectorTy_1); - ConstantInt* const_int32_32 = ConstantInt::get(APInt(32, "1", 10)); + Constant* const_int32_29 = Constant::getNullValue(IntegerType::get(32)); + ConstantInt* const_int32_30 = ConstantInt::get(APInt(32, "-1", 10)); + ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "1", 10)); + UndefValue* const_packed_32 = UndefValue::get(VectorTy_1); ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "2", 10)); ConstantInt* const_int32_34 = ConstantInt::get(APInt(32, "3", 10)); ConstantInt* const_int32_35 = ConstantInt::get(APInt(32, "4", 10)); - ConstantInt* const_int32_36 = ConstantInt::get(APInt(32, "-1", 10)); // Global Variable Definitions @@ -242,53 +257,68 @@ Module* createBaseShader() { int32_num_attribs->setName("num_attribs"); BasicBlock* label_entry = new BasicBlock("entry",func_from_array,0); + BasicBlock* label_forcond2_preheader_split = new BasicBlock("forcond2.preheader.split",func_from_array,0); + BasicBlock* label_forcond2 = new BasicBlock("forcond2",func_from_array,0); BasicBlock* label_forbody6 = new BasicBlock("forbody6",func_from_array,0); BasicBlock* label_forinc57 = new BasicBlock("forinc57",func_from_array,0); BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); // Block entry (label_entry) - ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_30, "cmp", label_entry); - ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_30, "cmp5", label_entry); + ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_29, "cmp", label_entry); + ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_29, "cmp5", label_entry); BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); - new BranchInst(label_forbody6, label_afterfor60, int1_bothcond, label_entry); + new BranchInst(label_forcond2_preheader_split, label_afterfor60, int1_bothcond, label_entry); - // Block forbody6 (label_forbody6) + // Block forcond2.preheader.split (label_forcond2_preheader_split) + BinaryOperator* int32_tmp21 = BinaryOperator::create(Instruction::Add, int32_count, const_int32_30, "tmp21", label_forcond2_preheader_split); + ICmpInst* int1_tmp22 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp21, const_int32_29, "tmp22", label_forcond2_preheader_split); + SelectInst* int32_tmp25 = new SelectInst(int1_tmp22, const_int32_31, int32_count, "tmp25", label_forcond2_preheader_split); + new BranchInst(label_forcond2, label_forcond2_preheader_split); + + // Block forcond2 (label_forcond2) Argument* fwdref_38 = new Argument(IntegerType::get(32)); - Argument* fwdref_39 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody6); - int32_i_0_reg2mem_0->reserveOperandSpace(3); - int32_i_0_reg2mem_0->addIncoming(const_int32_30, label_entry); + PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forcond2); + int32_i_0_reg2mem_0->reserveOperandSpace(2); + int32_i_0_reg2mem_0->addIncoming(const_int32_29, label_forcond2_preheader_split); int32_i_0_reg2mem_0->addIncoming(fwdref_38, label_forinc57); - int32_i_0_reg2mem_0->addIncoming(fwdref_39, label_forbody6); - Argument* fwdref_40 = new Argument(IntegerType::get(32)); + Argument* fwdref_39 = new Argument(VectorTy_1); + PHINode* packed_vec_1_reg2mem_0 = new PHINode(VectorTy_1, "vec.1.reg2mem.0", label_forcond2); + packed_vec_1_reg2mem_0->reserveOperandSpace(2); + packed_vec_1_reg2mem_0->addIncoming(const_packed_32, label_forcond2_preheader_split); + packed_vec_1_reg2mem_0->addIncoming(fwdref_39, label_forinc57); + + BinaryOperator* int32_tmp = BinaryOperator::create(Instruction::Add, int32_num_attribs, const_int32_30, "tmp", label_forcond2); + ICmpInst* int1_tmp18 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp, const_int32_29, "tmp18", label_forcond2); + SelectInst* int32_tmp19 = new SelectInst(int1_tmp18, const_int32_31, int32_num_attribs, "tmp19", label_forcond2); + new BranchInst(label_forbody6, label_forcond2); + + // Block forbody6 (label_forbody6) + Argument* fwdref_41 = new Argument(IntegerType::get(32)); PHINode* int32_j_0_reg2mem_0 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0", label_forbody6); - int32_j_0_reg2mem_0->reserveOperandSpace(3); - int32_j_0_reg2mem_0->addIncoming(fwdref_40, label_forbody6); - int32_j_0_reg2mem_0->addIncoming(const_int32_30, label_forinc57); - int32_j_0_reg2mem_0->addIncoming(const_int32_30, label_entry); + int32_j_0_reg2mem_0->reserveOperandSpace(2); + int32_j_0_reg2mem_0->addIncoming(const_int32_29, label_forcond2); + int32_j_0_reg2mem_0->addIncoming(fwdref_41, label_forbody6); - Argument* fwdref_41 = new Argument(VectorTy_1); PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody6); - packed_vec_0_reg2mem_0->reserveOperandSpace(3); - packed_vec_0_reg2mem_0->addIncoming(fwdref_41, label_forbody6); - packed_vec_0_reg2mem_0->addIncoming(const_packed_31, label_entry); - packed_vec_0_reg2mem_0->addIncoming(fwdref_41, label_forinc57); + packed_vec_0_reg2mem_0->reserveOperandSpace(2); + packed_vec_0_reg2mem_0->addIncoming(packed_vec_1_reg2mem_0, label_forcond2); + packed_vec_0_reg2mem_0->addIncoming(fwdref_39, label_forbody6); std::vector ptr_arraydecay11_indices; ptr_arraydecay11_indices.push_back(int32_i_0_reg2mem_0); ptr_arraydecay11_indices.push_back(int32_j_0_reg2mem_0); - ptr_arraydecay11_indices.push_back(const_int32_30); + ptr_arraydecay11_indices.push_back(const_int32_29); Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); - InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_30, "tmp15", label_forbody6); + InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_29, "tmp15", label_forbody6); std::vector ptr_arrayidx23_indices; ptr_arrayidx23_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx23_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx23_indices.push_back(const_int32_32); + ptr_arrayidx23_indices.push_back(const_int32_31); Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_32, "tmp26", label_forbody6); + InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_31, "tmp26", label_forbody6); std::vector ptr_arrayidx34_indices; ptr_arrayidx34_indices.push_back(int32_i_0_reg2mem_0); ptr_arrayidx34_indices.push_back(int32_j_0_reg2mem_0); @@ -308,23 +338,22 @@ Module* createBaseShader() { ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0); Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); StoreInst* void_42 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); - BinaryOperator* int32_inc = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_32, "inc", label_forbody6); - ICmpInst* int1_cmp59 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc, int32_num_attribs, "cmp59", label_forbody6); - new BranchInst(label_forbody6, label_forinc57, int1_cmp59, label_forbody6); + BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_31, "indvar.next", label_forbody6); + ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_tmp19, "exitcond", label_forbody6); + new BranchInst(label_forinc57, label_forbody6, int1_exitcond, label_forbody6); // Block forinc57 (label_forinc57) - BinaryOperator* int32_inc59 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_32, "inc59", label_forinc57); - ICmpInst* int1_cmp17 = new ICmpInst(ICmpInst::ICMP_SLT, int32_inc59, int32_count, "cmp17", label_forinc57); - new BranchInst(label_forbody6, label_afterfor60, int1_cmp17, label_forinc57); + BinaryOperator* int32_indvar_next20 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_31, "indvar.next20", label_forinc57); + ICmpInst* int1_exitcond26 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next20, int32_tmp25, "exitcond26", label_forinc57); + new BranchInst(label_afterfor60, label_forcond2, int1_exitcond26, label_forinc57); // Block afterfor60 (label_afterfor60) new ReturnInst(label_afterfor60); // Resolve Forward References - fwdref_39->replaceAllUsesWith(int32_i_0_reg2mem_0); delete fwdref_39; - fwdref_41->replaceAllUsesWith(packed_tmp48); delete fwdref_41; - fwdref_40->replaceAllUsesWith(int32_inc); delete fwdref_40; - fwdref_38->replaceAllUsesWith(int32_inc59); delete fwdref_38; + fwdref_39->replaceAllUsesWith(packed_tmp48); delete fwdref_39; + fwdref_41->replaceAllUsesWith(int32_indvar_next); delete fwdref_41; + fwdref_38->replaceAllUsesWith(int32_indvar_next20); delete fwdref_38; } @@ -339,62 +368,69 @@ Module* createBaseShader() { int32_count_48->setName("count"); BasicBlock* label_entry_49 = new BasicBlock("entry",func_from_consts,0); + BasicBlock* label_forbody_preheader = new BasicBlock("forbody.preheader",func_from_consts,0); BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); // Block entry (label_entry_49) - ICmpInst* int1_cmp_50 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_48, const_int32_30, "cmp", label_entry_49); - new BranchInst(label_forbody, label_afterfor, int1_cmp_50, label_entry_49); + ICmpInst* int1_cmp_50 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_48, const_int32_29, "cmp", label_entry_49); + new BranchInst(label_forbody_preheader, label_afterfor, int1_cmp_50, label_entry_49); + + // Block forbody.preheader (label_forbody_preheader) + BinaryOperator* int32_tmp_52 = BinaryOperator::create(Instruction::Add, int32_count_48, const_int32_30, "tmp", label_forbody_preheader); + ICmpInst* int1_tmp9 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_52, const_int32_29, "tmp9", label_forbody_preheader); + SelectInst* int32_tmp10 = new SelectInst(int1_tmp9, const_int32_31, int32_count_48, "tmp10", label_forbody_preheader); + new BranchInst(label_forbody, label_forbody_preheader); // Block forbody (label_forbody) - Argument* fwdref_53 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_52 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); - int32_i_0_reg2mem_0_52->reserveOperandSpace(2); - int32_i_0_reg2mem_0_52->addIncoming(const_int32_30, label_entry_49); - int32_i_0_reg2mem_0_52->addIncoming(fwdref_53, label_forbody); - - Argument* fwdref_55 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_54 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody); - packed_vec_0_reg2mem_0_54->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_54->addIncoming(const_packed_31, label_entry_49); - packed_vec_0_reg2mem_0_54->addIncoming(fwdref_55, label_forbody); + Argument* fwdref_55 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_54 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); + int32_i_0_reg2mem_0_54->reserveOperandSpace(2); + int32_i_0_reg2mem_0_54->addIncoming(const_int32_29, label_forbody_preheader); + int32_i_0_reg2mem_0_54->addIncoming(fwdref_55, label_forbody); + + Argument* fwdref_57 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_56 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody); + packed_vec_0_reg2mem_0_56->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_56->addIncoming(const_packed_32, label_forbody_preheader); + packed_vec_0_reg2mem_0_56->addIncoming(fwdref_57, label_forbody); std::vector ptr_arraydecay_indices; - ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_52); - ptr_arraydecay_indices.push_back(const_int32_30); + ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_54); + ptr_arraydecay_indices.push_back(const_int32_29); Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_47, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); - InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_54, float_tmp5, const_int32_30, "tmp7", label_forbody); + InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_56, float_tmp5, const_int32_29, "tmp7", label_forbody); std::vector ptr_arrayidx12_indices; - ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_52); - ptr_arrayidx12_indices.push_back(const_int32_32); + ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_54); + ptr_arrayidx12_indices.push_back(const_int32_31); Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); - LoadInst* float_tmp13_56 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); - InsertElementInst* packed_tmp15_57 = new InsertElementInst(packed_tmp7, float_tmp13_56, const_int32_32, "tmp15", label_forbody); + LoadInst* float_tmp13_58 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); + InsertElementInst* packed_tmp15_59 = new InsertElementInst(packed_tmp7, float_tmp13_58, const_int32_31, "tmp15", label_forbody); std::vector ptr_arrayidx20_indices; - ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_52); + ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_54); ptr_arrayidx20_indices.push_back(const_int32_33); Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); - InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_57, float_tmp21, const_int32_33, "tmp23", label_forbody); + InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_59, float_tmp21, const_int32_33, "tmp23", label_forbody); std::vector ptr_arrayidx28_indices; - ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_52); + ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_54); ptr_arrayidx28_indices.push_back(const_int32_34); Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_34, "tmp31", label_forbody); - GetElementPtrInst* ptr_arrayidx34_58 = new GetElementPtrInst(ptr_res_46, int32_i_0_reg2mem_0_52, "arrayidx34", label_forbody); - StoreInst* void_59 = new StoreInst(packed_tmp31, ptr_arrayidx34_58, false, label_forbody); - BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_52, const_int32_32, "indvar.next", label_forbody); - ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_count_48, "exitcond", label_forbody); - new BranchInst(label_afterfor, label_forbody, int1_exitcond, label_forbody); + GetElementPtrInst* ptr_arrayidx34_60 = new GetElementPtrInst(ptr_res_46, int32_i_0_reg2mem_0_54, "arrayidx34", label_forbody); + StoreInst* void_61 = new StoreInst(packed_tmp31, ptr_arrayidx34_60, false, label_forbody); + BinaryOperator* int32_indvar_next_62 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_54, const_int32_31, "indvar.next", label_forbody); + ICmpInst* int1_exitcond_63 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_62, int32_tmp10, "exitcond", label_forbody); + new BranchInst(label_afterfor, label_forbody, int1_exitcond_63, label_forbody); // Block afterfor (label_afterfor) new ReturnInst(label_afterfor); // Resolve Forward References - fwdref_55->replaceAllUsesWith(packed_tmp31); delete fwdref_55; - fwdref_53->replaceAllUsesWith(int32_indvar_next); delete fwdref_53; + fwdref_57->replaceAllUsesWith(packed_tmp31); delete fwdref_57; + fwdref_55->replaceAllUsesWith(int32_indvar_next_62); delete fwdref_55; } @@ -405,59 +441,66 @@ Module* createBaseShader() { ptr_dests->setName("dests"); Value* ptr_in = args++; ptr_in->setName("in"); - Value* int32_num_attribs_62 = args++; - int32_num_attribs_62->setName("num_attribs"); - - BasicBlock* label_entry_63 = new BasicBlock("entry",func_to_array,0); - BasicBlock* label_forbody_64 = new BasicBlock("forbody",func_to_array,0); - BasicBlock* label_afterfor_65 = new BasicBlock("afterfor",func_to_array,0); - - // Block entry (label_entry_63) - ICmpInst* int1_cmp_66 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_62, const_int32_30, "cmp", label_entry_63); - new BranchInst(label_forbody_64, label_afterfor_65, int1_cmp_66, label_entry_63); - - // Block forbody (label_forbody_64) - Argument* fwdref_69 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_68 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_64); - int32_i_0_reg2mem_0_68->reserveOperandSpace(2); - int32_i_0_reg2mem_0_68->addIncoming(const_int32_30, label_entry_63); - int32_i_0_reg2mem_0_68->addIncoming(fwdref_69, label_forbody_64); - - std::vector ptr_arraydecay_70_indices; - ptr_arraydecay_70_indices.push_back(int32_i_0_reg2mem_0_68); - ptr_arraydecay_70_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay_70 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_70_indices.begin(), ptr_arraydecay_70_indices.end(), "arraydecay", label_forbody_64); - GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_68, "arrayidx6", label_forbody_64); - LoadInst* packed_tmp7_71 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_64); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_71, const_int32_30, "tmp11", label_forbody_64); - StoreInst* void_72 = new StoreInst(float_tmp11, ptr_arraydecay_70, false, label_forbody_64); + Value* int32_num_attribs_66 = args++; + int32_num_attribs_66->setName("num_attribs"); + + BasicBlock* label_entry_67 = new BasicBlock("entry",func_to_array,0); + BasicBlock* label_forbody_preheader_68 = new BasicBlock("forbody.preheader",func_to_array,0); + BasicBlock* label_forbody_69 = new BasicBlock("forbody",func_to_array,0); + BasicBlock* label_afterfor_70 = new BasicBlock("afterfor",func_to_array,0); + + // Block entry (label_entry_67) + ICmpInst* int1_cmp_71 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_66, const_int32_29, "cmp", label_entry_67); + new BranchInst(label_forbody_preheader_68, label_afterfor_70, int1_cmp_71, label_entry_67); + + // Block forbody.preheader (label_forbody_preheader_68) + BinaryOperator* int32_tmp_73 = BinaryOperator::create(Instruction::Add, int32_num_attribs_66, const_int32_30, "tmp", label_forbody_preheader_68); + ICmpInst* int1_tmp9_74 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_73, const_int32_29, "tmp9", label_forbody_preheader_68); + SelectInst* int32_tmp10_75 = new SelectInst(int1_tmp9_74, const_int32_31, int32_num_attribs_66, "tmp10", label_forbody_preheader_68); + new BranchInst(label_forbody_69, label_forbody_preheader_68); + + // Block forbody (label_forbody_69) + Argument* fwdref_78 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_77 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_69); + int32_i_0_reg2mem_0_77->reserveOperandSpace(2); + int32_i_0_reg2mem_0_77->addIncoming(const_int32_29, label_forbody_preheader_68); + int32_i_0_reg2mem_0_77->addIncoming(fwdref_78, label_forbody_69); + + std::vector ptr_arraydecay_79_indices; + ptr_arraydecay_79_indices.push_back(int32_i_0_reg2mem_0_77); + ptr_arraydecay_79_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay_79 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_79_indices.begin(), ptr_arraydecay_79_indices.end(), "arraydecay", label_forbody_69); + GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_77, "arrayidx6", label_forbody_69); + LoadInst* packed_tmp7_80 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_69); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_80, const_int32_29, "tmp11", label_forbody_69); + StoreInst* void_81 = new StoreInst(float_tmp11, ptr_arraydecay_79, false, label_forbody_69); std::vector ptr_arrayidx13_indices; - ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_68); - ptr_arrayidx13_indices.push_back(const_int32_32); - Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_64); - ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_71, const_int32_32, "tmp15", label_forbody_64); - StoreInst* void_73 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_64); + ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_77); + ptr_arrayidx13_indices.push_back(const_int32_31); + Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_69); + ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_80, const_int32_31, "tmp15", label_forbody_69); + StoreInst* void_82 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_69); std::vector ptr_arrayidx17_indices; - ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_68); + ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_77); ptr_arrayidx17_indices.push_back(const_int32_33); - Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_64); - ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_71, const_int32_33, "tmp19", label_forbody_64); - StoreInst* void_74 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_64); + Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_69); + ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_80, const_int32_33, "tmp19", label_forbody_69); + StoreInst* void_83 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_69); std::vector ptr_arrayidx21_indices; - ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_68); + ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_77); ptr_arrayidx21_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_64); - ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_71, const_int32_34, "tmp23", label_forbody_64); - StoreInst* void_75 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_64); - BinaryOperator* int32_indvar_next_76 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_68, const_int32_32, "indvar.next", label_forbody_64); - ICmpInst* int1_exitcond_77 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_76, int32_num_attribs_62, "exitcond", label_forbody_64); - new BranchInst(label_afterfor_65, label_forbody_64, int1_exitcond_77, label_forbody_64); + Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_69); + ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_80, const_int32_34, "tmp23", label_forbody_69); + StoreInst* void_84 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_69); + BinaryOperator* int32_indvar_next_85 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_77, const_int32_31, "indvar.next", label_forbody_69); + ICmpInst* int1_exitcond_86 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_85, int32_tmp10_75, "exitcond", label_forbody_69); + new BranchInst(label_afterfor_70, label_forbody_69, int1_exitcond_86, label_forbody_69); - // Block afterfor (label_afterfor_65) - new ReturnInst(label_afterfor_65); + // Block afterfor (label_afterfor_70) + new ReturnInst(label_afterfor_70); // Resolve Forward References - fwdref_69->replaceAllUsesWith(int32_indvar_next_76); delete fwdref_69; + fwdref_78->replaceAllUsesWith(int32_indvar_next_85); delete fwdref_78; } @@ -474,50 +517,57 @@ Module* createBaseShader() { int32_num_vertices->setName("num_vertices"); Value* int32_num_inputs = args++; int32_num_inputs->setName("num_inputs"); - Value* int32_num_attribs_80 = args++; - int32_num_attribs_80->setName("num_attribs"); + Value* int32_num_attribs_89 = args++; + int32_num_attribs_89->setName("num_attribs"); Value* int32_num_consts = args++; int32_num_consts->setName("num_consts"); - BasicBlock* label_entry_81 = new BasicBlock("entry",func_run_vertex_shader,0); + BasicBlock* label_entry_90 = new BasicBlock("entry",func_run_vertex_shader,0); + BasicBlock* label_forbody_preheader_i = new BasicBlock("forbody.preheader.i",func_run_vertex_shader,0); BasicBlock* label_forbody_i = new BasicBlock("forbody.i",func_run_vertex_shader,0); BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_vertex_shader,0); - BasicBlock* label_forbody_preheader = new BasicBlock("forbody.preheader",func_run_vertex_shader,0); - BasicBlock* label_forbody_82 = new BasicBlock("forbody",func_run_vertex_shader,0); - BasicBlock* label_afterfor_83 = new BasicBlock("afterfor",func_run_vertex_shader,0); - - // Block entry (label_entry_81) - AllocaInst* ptr_consts = new AllocaInst(ArrayTy_21, "consts", label_entry_81); - AllocaInst* ptr_temps = new AllocaInst(ArrayTy_23, "temps", label_entry_81); - AllocaInst* ptr_args = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_81); - ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_30, "cmp.i", label_entry_81); - new BranchInst(label_forbody_i, label_from_consts_exit, int1_cmp_i, label_entry_81); + BasicBlock* label_forbody_preheader_91 = new BasicBlock("forbody.preheader",func_run_vertex_shader,0); + BasicBlock* label_forbody_92 = new BasicBlock("forbody",func_run_vertex_shader,0); + BasicBlock* label_afterfor_93 = new BasicBlock("afterfor",func_run_vertex_shader,0); + + // Block entry (label_entry_90) + AllocaInst* ptr_consts = new AllocaInst(ArrayTy_20, "consts", label_entry_90); + AllocaInst* ptr_temps = new AllocaInst(ArrayTy_22, "temps", label_entry_90); + AllocaInst* ptr_args = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_90); + ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_29, "cmp.i", label_entry_90); + new BranchInst(label_forbody_preheader_i, label_from_consts_exit, int1_cmp_i, label_entry_90); + + // Block forbody.preheader.i (label_forbody_preheader_i) + BinaryOperator* int32_tmp_i = BinaryOperator::create(Instruction::Add, int32_num_consts, const_int32_30, "tmp.i", label_forbody_preheader_i); + ICmpInst* int1_tmp9_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_i, const_int32_29, "tmp9.i", label_forbody_preheader_i); + SelectInst* int32_tmp10_i = new SelectInst(int1_tmp9_i, const_int32_31, int32_num_consts, "tmp10.i", label_forbody_preheader_i); + new BranchInst(label_forbody_i, label_forbody_preheader_i); // Block forbody.i (label_forbody_i) - Argument* fwdref_85 = new Argument(IntegerType::get(32)); + Argument* fwdref_96 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i); int32_i_0_reg2mem_0_i->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i->addIncoming(const_int32_30, label_entry_81); - int32_i_0_reg2mem_0_i->addIncoming(fwdref_85, label_forbody_i); + int32_i_0_reg2mem_0_i->addIncoming(const_int32_29, label_forbody_preheader_i); + int32_i_0_reg2mem_0_i->addIncoming(fwdref_96, label_forbody_i); - Argument* fwdref_86 = new Argument(VectorTy_1); + Argument* fwdref_97 = new Argument(VectorTy_1); PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i); packed_vec_0_reg2mem_0_i->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i->addIncoming(const_packed_31, label_entry_81); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_86, label_forbody_i); + packed_vec_0_reg2mem_0_i->addIncoming(const_packed_32, label_forbody_preheader_i); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_97, label_forbody_i); std::vector ptr_arraydecay_i_indices; ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arraydecay_i_indices.push_back(const_int32_30); + ptr_arraydecay_i_indices.push_back(const_int32_29); Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_i); LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i, "tmp5.i", false, label_forbody_i); - InsertElementInst* packed_tmp7_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp5_i, const_int32_30, "tmp7.i", label_forbody_i); + InsertElementInst* packed_tmp7_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp5_i, const_int32_29, "tmp7.i", label_forbody_i); std::vector ptr_arrayidx12_i_indices; ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arrayidx12_i_indices.push_back(const_int32_32); + ptr_arrayidx12_i_indices.push_back(const_int32_31); Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i); LoadInst* float_tmp13_i = new LoadInst(ptr_arrayidx12_i, "tmp13.i", false, label_forbody_i); - InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_tmp7_i, float_tmp13_i, const_int32_32, "tmp15.i", label_forbody_i); + InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_tmp7_i, float_tmp13_i, const_int32_31, "tmp15.i", label_forbody_i); std::vector ptr_arrayidx20_i_indices; ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i); ptr_arrayidx20_i_indices.push_back(const_int32_33); @@ -531,78 +581,83 @@ Module* createBaseShader() { LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i); InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i, float_tmp29_i, const_int32_34, "tmp31.i", label_forbody_i); std::vector ptr_arrayidx34_i_indices; - ptr_arrayidx34_i_indices.push_back(const_int32_30); + ptr_arrayidx34_i_indices.push_back(const_int32_29); ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i); Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody_i); - StoreInst* void_87 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i, false, label_forbody_i); - BinaryOperator* int32_indvar_next6 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i, const_int32_32, "indvar.next6", label_forbody_i); - ICmpInst* int1_exitcond7 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next6, int32_num_consts, "exitcond7", label_forbody_i); - new BranchInst(label_from_consts_exit, label_forbody_i, int1_exitcond7, label_forbody_i); + StoreInst* void_98 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i, false, label_forbody_i); + BinaryOperator* int32_indvar_next8 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i, const_int32_31, "indvar.next8", label_forbody_i); + ICmpInst* int1_exitcond9 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next8, int32_tmp10_i, "exitcond9", label_forbody_i); + new BranchInst(label_from_consts_exit, label_forbody_i, int1_exitcond9, label_forbody_i); // Block from_consts.exit (label_from_consts_exit) std::vector ptr_tmp2_indices; - ptr_tmp2_indices.push_back(const_int32_30); + ptr_tmp2_indices.push_back(const_int32_29); ptr_tmp2_indices.push_back(const_int32_34); Instruction* ptr_tmp2 = new GetElementPtrInst(ptr_args, ptr_tmp2_indices.begin(), ptr_tmp2_indices.end(), "tmp2", label_from_consts_exit); std::vector ptr_arraydecay3_indices; - ptr_arraydecay3_indices.push_back(const_int32_30); - ptr_arraydecay3_indices.push_back(const_int32_30); + ptr_arraydecay3_indices.push_back(const_int32_29); + ptr_arraydecay3_indices.push_back(const_int32_29); Instruction* ptr_arraydecay3 = new GetElementPtrInst(ptr_consts, ptr_arraydecay3_indices.begin(), ptr_arraydecay3_indices.end(), "arraydecay3", label_from_consts_exit); - StoreInst* void_89 = new StoreInst(ptr_arraydecay3, ptr_tmp2, false, label_from_consts_exit); + StoreInst* void_100 = new StoreInst(ptr_arraydecay3, ptr_tmp2, false, label_from_consts_exit); std::vector ptr_tmp4_indices; - ptr_tmp4_indices.push_back(const_int32_30); + ptr_tmp4_indices.push_back(const_int32_29); ptr_tmp4_indices.push_back(const_int32_33); Instruction* ptr_tmp4 = new GetElementPtrInst(ptr_args, ptr_tmp4_indices.begin(), ptr_tmp4_indices.end(), "tmp4", label_from_consts_exit); std::vector ptr_arraydecay5_indices; - ptr_arraydecay5_indices.push_back(const_int32_30); - ptr_arraydecay5_indices.push_back(const_int32_30); + ptr_arraydecay5_indices.push_back(const_int32_29); + ptr_arraydecay5_indices.push_back(const_int32_29); Instruction* ptr_arraydecay5 = new GetElementPtrInst(ptr_temps, ptr_arraydecay5_indices.begin(), ptr_arraydecay5_indices.end(), "arraydecay5", label_from_consts_exit); - StoreInst* void_90 = new StoreInst(ptr_arraydecay5, ptr_tmp4, false, label_from_consts_exit); - ICmpInst* int1_cmp_91 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_30, "cmp", label_from_consts_exit); - new BranchInst(label_forbody_preheader, label_afterfor_83, int1_cmp_91, label_from_consts_exit); + StoreInst* void_101 = new StoreInst(ptr_arraydecay5, ptr_tmp4, false, label_from_consts_exit); + ICmpInst* int1_cmp_102 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_29, "cmp", label_from_consts_exit); + new BranchInst(label_forbody_preheader_91, label_afterfor_93, int1_cmp_102, label_from_consts_exit); - // Block forbody.preheader (label_forbody_preheader) + // Block forbody.preheader (label_forbody_preheader_91) std::vector ptr_tmp8_indices; - ptr_tmp8_indices.push_back(const_int32_30); - ptr_tmp8_indices.push_back(const_int32_30); - Instruction* ptr_tmp8 = new GetElementPtrInst(ptr_args, ptr_tmp8_indices.begin(), ptr_tmp8_indices.end(), "tmp8", label_forbody_preheader); + ptr_tmp8_indices.push_back(const_int32_29); + ptr_tmp8_indices.push_back(const_int32_29); + Instruction* ptr_tmp8 = new GetElementPtrInst(ptr_args, ptr_tmp8_indices.begin(), ptr_tmp8_indices.end(), "tmp8", label_forbody_preheader_91); std::vector ptr_tmp12_indices; - ptr_tmp12_indices.push_back(const_int32_30); - ptr_tmp12_indices.push_back(const_int32_32); - Instruction* ptr_tmp12 = new GetElementPtrInst(ptr_args, ptr_tmp12_indices.begin(), ptr_tmp12_indices.end(), "tmp12", label_forbody_preheader); - new BranchInst(label_forbody_82, label_forbody_preheader); - - // Block forbody (label_forbody_82) - Argument* fwdref_95 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_94 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_82); - int32_i_0_reg2mem_0_94->reserveOperandSpace(2); - int32_i_0_reg2mem_0_94->addIncoming(const_int32_30, label_forbody_preheader); - int32_i_0_reg2mem_0_94->addIncoming(fwdref_95, label_forbody_82); - - std::vector ptr_arraydecay11_96_indices; - ptr_arraydecay11_96_indices.push_back(int32_i_0_reg2mem_0_94); - ptr_arraydecay11_96_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay11_96 = new GetElementPtrInst(ptr_results, ptr_arraydecay11_96_indices.begin(), ptr_arraydecay11_96_indices.end(), "arraydecay11", label_forbody_82); - StoreInst* void_97 = new StoreInst(ptr_arraydecay11_96, ptr_tmp8, false, label_forbody_82); + ptr_tmp12_indices.push_back(const_int32_29); + ptr_tmp12_indices.push_back(const_int32_31); + Instruction* ptr_tmp12 = new GetElementPtrInst(ptr_args, ptr_tmp12_indices.begin(), ptr_tmp12_indices.end(), "tmp12", label_forbody_preheader_91); + BinaryOperator* int32_tmp_104 = BinaryOperator::create(Instruction::Add, int32_num_vertices, const_int32_30, "tmp", label_forbody_preheader_91); + ICmpInst* int1_tmp6 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_104, const_int32_29, "tmp6", label_forbody_preheader_91); + SelectInst* int32_tmp7 = new SelectInst(int1_tmp6, const_int32_31, int32_num_vertices, "tmp7", label_forbody_preheader_91); + new BranchInst(label_forbody_92, label_forbody_preheader_91); + + // Block forbody (label_forbody_92) + Argument* fwdref_107 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_106 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_92); + int32_i_0_reg2mem_0_106->reserveOperandSpace(2); + int32_i_0_reg2mem_0_106->addIncoming(const_int32_29, label_forbody_preheader_91); + int32_i_0_reg2mem_0_106->addIncoming(fwdref_107, label_forbody_92); + + std::vector ptr_arraydecay11_108_indices; + ptr_arraydecay11_108_indices.push_back(int32_i_0_reg2mem_0_106); + ptr_arraydecay11_108_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay11_108 = new GetElementPtrInst(ptr_results, ptr_arraydecay11_108_indices.begin(), ptr_arraydecay11_108_indices.end(), "arraydecay11", label_forbody_92); + StoreInst* void_109 = new StoreInst(ptr_arraydecay11_108, ptr_tmp8, false, label_forbody_92); std::vector ptr_arraydecay16_indices; - ptr_arraydecay16_indices.push_back(int32_i_0_reg2mem_0_94); - ptr_arraydecay16_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay16 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay16_indices.begin(), ptr_arraydecay16_indices.end(), "arraydecay16", label_forbody_82); - StoreInst* void_98 = new StoreInst(ptr_arraydecay16, ptr_tmp12, false, label_forbody_82); - CallInst* void_99 = new CallInst(func_execute_shader, ptr_args, "", label_forbody_82); - void_99->setCallingConv(CallingConv::C); - void_99->setTailCall(false); - BinaryOperator* int32_indvar_next_100 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_94, const_int32_32, "indvar.next", label_forbody_82); - ICmpInst* int1_exitcond_101 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_100, int32_num_vertices, "exitcond", label_forbody_82); - new BranchInst(label_afterfor_83, label_forbody_82, int1_exitcond_101, label_forbody_82); - - // Block afterfor (label_afterfor_83) - new ReturnInst(label_afterfor_83); + ptr_arraydecay16_indices.push_back(int32_i_0_reg2mem_0_106); + ptr_arraydecay16_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay16 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay16_indices.begin(), ptr_arraydecay16_indices.end(), "arraydecay16", label_forbody_92); + StoreInst* void_110 = new StoreInst(ptr_arraydecay16, ptr_tmp12, false, label_forbody_92); + CallInst* void_111 = new CallInst(func_execute_shader, ptr_args, "", label_forbody_92); + void_111->setCallingConv(CallingConv::C); + void_111->setTailCall(false);ParamAttrsList *void_111_PAL = 0; + void_111->setParamAttrs(void_111_PAL); + + BinaryOperator* int32_indvar_next_112 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_106, const_int32_31, "indvar.next", label_forbody_92); + ICmpInst* int1_exitcond_113 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_112, int32_tmp7, "exitcond", label_forbody_92); + new BranchInst(label_afterfor_93, label_forbody_92, int1_exitcond_113, label_forbody_92); + + // Block afterfor (label_afterfor_93) + new ReturnInst(label_afterfor_93); // Resolve Forward References - fwdref_86->replaceAllUsesWith(packed_tmp31_i); delete fwdref_86; - fwdref_85->replaceAllUsesWith(int32_indvar_next6); delete fwdref_85; - fwdref_95->replaceAllUsesWith(int32_indvar_next_100); delete fwdref_95; + fwdref_107->replaceAllUsesWith(int32_indvar_next_112); delete fwdref_107; + fwdref_97->replaceAllUsesWith(packed_tmp31_i); delete fwdref_97; + fwdref_96->replaceAllUsesWith(int32_indvar_next8); delete fwdref_96; } @@ -613,180 +668,195 @@ Module* createBaseShader() { float_x->setName("x"); Value* float_y = args++; float_y->setName("y"); - Value* ptr_results_104 = args++; - ptr_results_104->setName("results"); - Value* ptr_inputs_105 = args++; - ptr_inputs_105->setName("inputs"); - Value* int32_num_inputs_106 = args++; - int32_num_inputs_106->setName("num_inputs"); - Value* ptr_aconsts_107 = args++; - ptr_aconsts_107->setName("aconsts"); - Value* int32_num_consts_108 = args++; - int32_num_consts_108->setName("num_consts"); + Value* ptr_results_116 = args++; + ptr_results_116->setName("results"); + Value* ptr_inputs_117 = args++; + ptr_inputs_117->setName("inputs"); + Value* int32_num_inputs_118 = args++; + int32_num_inputs_118->setName("num_inputs"); + Value* ptr_aconsts_119 = args++; + ptr_aconsts_119->setName("aconsts"); + Value* int32_num_consts_120 = args++; + int32_num_consts_120->setName("num_consts"); Value* ptr_samplers = args++; ptr_samplers->setName("samplers"); - BasicBlock* label_entry_109 = new BasicBlock("entry",func_run_fragment_shader,0); - BasicBlock* label_forbody_i_110 = new BasicBlock("forbody.i",func_run_fragment_shader,0); - BasicBlock* label_from_consts_exit_111 = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); + BasicBlock* label_entry_121 = new BasicBlock("entry",func_run_fragment_shader,0); + BasicBlock* label_forbody_preheader_i_122 = new BasicBlock("forbody.preheader.i",func_run_fragment_shader,0); + BasicBlock* label_forbody_i_123 = new BasicBlock("forbody.i",func_run_fragment_shader,0); + BasicBlock* label_from_consts_exit_124 = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); - // Block entry (label_entry_109) - AllocaInst* ptr_consts_112 = new AllocaInst(ArrayTy_21, "consts", label_entry_109); - AllocaInst* ptr_temps_113 = new AllocaInst(ArrayTy_23, "temps", label_entry_109); - AllocaInst* ptr_args_114 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_109); + // Block entry (label_entry_121) + AllocaInst* ptr_consts_125 = new AllocaInst(ArrayTy_20, "consts", label_entry_121); + AllocaInst* ptr_temps_126 = new AllocaInst(ArrayTy_22, "temps", label_entry_121); + AllocaInst* ptr_args_127 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_121); std::vector ptr_tmp_indices; - ptr_tmp_indices.push_back(const_int32_30); + ptr_tmp_indices.push_back(const_int32_29); ptr_tmp_indices.push_back(const_int32_35); - Instruction* ptr_tmp = new GetElementPtrInst(ptr_args_114, ptr_tmp_indices.begin(), ptr_tmp_indices.end(), "tmp", label_entry_109); - StoreInst* void_115 = new StoreInst(const_int32_30, ptr_tmp, false, label_entry_109); - ICmpInst* int1_cmp_i_116 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_108, const_int32_30, "cmp.i", label_entry_109); - new BranchInst(label_forbody_i_110, label_from_consts_exit_111, int1_cmp_i_116, label_entry_109); - - // Block forbody.i (label_forbody_i_110) - Argument* fwdref_119 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_i_118 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i_110); - int32_i_0_reg2mem_0_i_118->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i_118->addIncoming(const_int32_30, label_entry_109); - int32_i_0_reg2mem_0_i_118->addIncoming(fwdref_119, label_forbody_i_110); - - Argument* fwdref_121 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i_120 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i_110); - packed_vec_0_reg2mem_0_i_120->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_120->addIncoming(const_packed_31, label_entry_109); - packed_vec_0_reg2mem_0_i_120->addIncoming(fwdref_121, label_forbody_i_110); - - std::vector ptr_arraydecay_i_122_indices; - ptr_arraydecay_i_122_indices.push_back(int32_i_0_reg2mem_0_i_118); - ptr_arraydecay_i_122_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay_i_122 = new GetElementPtrInst(ptr_aconsts_107, ptr_arraydecay_i_122_indices.begin(), ptr_arraydecay_i_122_indices.end(), "arraydecay.i", label_forbody_i_110); - LoadInst* float_tmp5_i_123 = new LoadInst(ptr_arraydecay_i_122, "tmp5.i", false, label_forbody_i_110); - InsertElementInst* packed_tmp7_i_124 = new InsertElementInst(packed_vec_0_reg2mem_0_i_120, float_tmp5_i_123, const_int32_30, "tmp7.i", label_forbody_i_110); - std::vector ptr_arrayidx12_i_125_indices; - ptr_arrayidx12_i_125_indices.push_back(int32_i_0_reg2mem_0_i_118); - ptr_arrayidx12_i_125_indices.push_back(const_int32_32); - Instruction* ptr_arrayidx12_i_125 = new GetElementPtrInst(ptr_aconsts_107, ptr_arrayidx12_i_125_indices.begin(), ptr_arrayidx12_i_125_indices.end(), "arrayidx12.i", label_forbody_i_110); - LoadInst* float_tmp13_i_126 = new LoadInst(ptr_arrayidx12_i_125, "tmp13.i", false, label_forbody_i_110); - InsertElementInst* packed_tmp15_i_127 = new InsertElementInst(packed_tmp7_i_124, float_tmp13_i_126, const_int32_32, "tmp15.i", label_forbody_i_110); - std::vector ptr_arrayidx20_i_128_indices; - ptr_arrayidx20_i_128_indices.push_back(int32_i_0_reg2mem_0_i_118); - ptr_arrayidx20_i_128_indices.push_back(const_int32_33); - Instruction* ptr_arrayidx20_i_128 = new GetElementPtrInst(ptr_aconsts_107, ptr_arrayidx20_i_128_indices.begin(), ptr_arrayidx20_i_128_indices.end(), "arrayidx20.i", label_forbody_i_110); - LoadInst* float_tmp21_i_129 = new LoadInst(ptr_arrayidx20_i_128, "tmp21.i", false, label_forbody_i_110); - InsertElementInst* packed_tmp23_i_130 = new InsertElementInst(packed_tmp15_i_127, float_tmp21_i_129, const_int32_33, "tmp23.i", label_forbody_i_110); - std::vector ptr_arrayidx28_i_131_indices; - ptr_arrayidx28_i_131_indices.push_back(int32_i_0_reg2mem_0_i_118); - ptr_arrayidx28_i_131_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx28_i_131 = new GetElementPtrInst(ptr_aconsts_107, ptr_arrayidx28_i_131_indices.begin(), ptr_arrayidx28_i_131_indices.end(), "arrayidx28.i", label_forbody_i_110); - LoadInst* float_tmp29_i_132 = new LoadInst(ptr_arrayidx28_i_131, "tmp29.i", false, label_forbody_i_110); - InsertElementInst* packed_tmp31_i_133 = new InsertElementInst(packed_tmp23_i_130, float_tmp29_i_132, const_int32_34, "tmp31.i", label_forbody_i_110); - std::vector ptr_arrayidx34_i_134_indices; - ptr_arrayidx34_i_134_indices.push_back(const_int32_30); - ptr_arrayidx34_i_134_indices.push_back(int32_i_0_reg2mem_0_i_118); - Instruction* ptr_arrayidx34_i_134 = new GetElementPtrInst(ptr_consts_112, ptr_arrayidx34_i_134_indices.begin(), ptr_arrayidx34_i_134_indices.end(), "arrayidx34.i", label_forbody_i_110); - StoreInst* void_135 = new StoreInst(packed_tmp31_i_133, ptr_arrayidx34_i_134, false, label_forbody_i_110); - BinaryOperator* int32_indvar_next7 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_118, const_int32_32, "indvar.next7", label_forbody_i_110); - ICmpInst* int1_exitcond8 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next7, int32_num_consts_108, "exitcond8", label_forbody_i_110); - new BranchInst(label_from_consts_exit_111, label_forbody_i_110, int1_exitcond8, label_forbody_i_110); - - // Block from_consts.exit (label_from_consts_exit_111) + Instruction* ptr_tmp = new GetElementPtrInst(ptr_args_127, ptr_tmp_indices.begin(), ptr_tmp_indices.end(), "tmp", label_entry_121); + StoreInst* void_128 = new StoreInst(const_int32_29, ptr_tmp, false, label_entry_121); + ICmpInst* int1_cmp_i_129 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_120, const_int32_29, "cmp.i", label_entry_121); + new BranchInst(label_forbody_preheader_i_122, label_from_consts_exit_124, int1_cmp_i_129, label_entry_121); + + // Block forbody.preheader.i (label_forbody_preheader_i_122) + BinaryOperator* int32_tmp_i_131 = BinaryOperator::create(Instruction::Add, int32_num_consts_120, const_int32_30, "tmp.i", label_forbody_preheader_i_122); + ICmpInst* int1_tmp9_i_132 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_i_131, const_int32_29, "tmp9.i", label_forbody_preheader_i_122); + SelectInst* int32_tmp10_i_133 = new SelectInst(int1_tmp9_i_132, const_int32_31, int32_num_consts_120, "tmp10.i", label_forbody_preheader_i_122); + new BranchInst(label_forbody_i_123, label_forbody_preheader_i_122); + + // Block forbody.i (label_forbody_i_123) + Argument* fwdref_136 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_i_135 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i_123); + int32_i_0_reg2mem_0_i_135->reserveOperandSpace(2); + int32_i_0_reg2mem_0_i_135->addIncoming(const_int32_29, label_forbody_preheader_i_122); + int32_i_0_reg2mem_0_i_135->addIncoming(fwdref_136, label_forbody_i_123); + + Argument* fwdref_138 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i_137 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i_123); + packed_vec_0_reg2mem_0_i_137->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_137->addIncoming(const_packed_32, label_forbody_preheader_i_122); + packed_vec_0_reg2mem_0_i_137->addIncoming(fwdref_138, label_forbody_i_123); + + std::vector ptr_arraydecay_i_139_indices; + ptr_arraydecay_i_139_indices.push_back(int32_i_0_reg2mem_0_i_135); + ptr_arraydecay_i_139_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay_i_139 = new GetElementPtrInst(ptr_aconsts_119, ptr_arraydecay_i_139_indices.begin(), ptr_arraydecay_i_139_indices.end(), "arraydecay.i", label_forbody_i_123); + LoadInst* float_tmp5_i_140 = new LoadInst(ptr_arraydecay_i_139, "tmp5.i", false, label_forbody_i_123); + InsertElementInst* packed_tmp7_i_141 = new InsertElementInst(packed_vec_0_reg2mem_0_i_137, float_tmp5_i_140, const_int32_29, "tmp7.i", label_forbody_i_123); + std::vector ptr_arrayidx12_i_142_indices; + ptr_arrayidx12_i_142_indices.push_back(int32_i_0_reg2mem_0_i_135); + ptr_arrayidx12_i_142_indices.push_back(const_int32_31); + Instruction* ptr_arrayidx12_i_142 = new GetElementPtrInst(ptr_aconsts_119, ptr_arrayidx12_i_142_indices.begin(), ptr_arrayidx12_i_142_indices.end(), "arrayidx12.i", label_forbody_i_123); + LoadInst* float_tmp13_i_143 = new LoadInst(ptr_arrayidx12_i_142, "tmp13.i", false, label_forbody_i_123); + InsertElementInst* packed_tmp15_i_144 = new InsertElementInst(packed_tmp7_i_141, float_tmp13_i_143, const_int32_31, "tmp15.i", label_forbody_i_123); + std::vector ptr_arrayidx20_i_145_indices; + ptr_arrayidx20_i_145_indices.push_back(int32_i_0_reg2mem_0_i_135); + ptr_arrayidx20_i_145_indices.push_back(const_int32_33); + Instruction* ptr_arrayidx20_i_145 = new GetElementPtrInst(ptr_aconsts_119, ptr_arrayidx20_i_145_indices.begin(), ptr_arrayidx20_i_145_indices.end(), "arrayidx20.i", label_forbody_i_123); + LoadInst* float_tmp21_i_146 = new LoadInst(ptr_arrayidx20_i_145, "tmp21.i", false, label_forbody_i_123); + InsertElementInst* packed_tmp23_i_147 = new InsertElementInst(packed_tmp15_i_144, float_tmp21_i_146, const_int32_33, "tmp23.i", label_forbody_i_123); + std::vector ptr_arrayidx28_i_148_indices; + ptr_arrayidx28_i_148_indices.push_back(int32_i_0_reg2mem_0_i_135); + ptr_arrayidx28_i_148_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx28_i_148 = new GetElementPtrInst(ptr_aconsts_119, ptr_arrayidx28_i_148_indices.begin(), ptr_arrayidx28_i_148_indices.end(), "arrayidx28.i", label_forbody_i_123); + LoadInst* float_tmp29_i_149 = new LoadInst(ptr_arrayidx28_i_148, "tmp29.i", false, label_forbody_i_123); + InsertElementInst* packed_tmp31_i_150 = new InsertElementInst(packed_tmp23_i_147, float_tmp29_i_149, const_int32_34, "tmp31.i", label_forbody_i_123); + std::vector ptr_arrayidx34_i_151_indices; + ptr_arrayidx34_i_151_indices.push_back(const_int32_29); + ptr_arrayidx34_i_151_indices.push_back(int32_i_0_reg2mem_0_i_135); + Instruction* ptr_arrayidx34_i_151 = new GetElementPtrInst(ptr_consts_125, ptr_arrayidx34_i_151_indices.begin(), ptr_arrayidx34_i_151_indices.end(), "arrayidx34.i", label_forbody_i_123); + StoreInst* void_152 = new StoreInst(packed_tmp31_i_150, ptr_arrayidx34_i_151, false, label_forbody_i_123); + BinaryOperator* int32_indvar_next7 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_135, const_int32_31, "indvar.next7", label_forbody_i_123); + ICmpInst* int1_exitcond8 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next7, int32_tmp10_i_133, "exitcond8", label_forbody_i_123); + new BranchInst(label_from_consts_exit_124, label_forbody_i_123, int1_exitcond8, label_forbody_i_123); + + // Block from_consts.exit (label_from_consts_exit_124) std::vector ptr_tmp3_indices; - ptr_tmp3_indices.push_back(const_int32_30); + ptr_tmp3_indices.push_back(const_int32_29); ptr_tmp3_indices.push_back(const_int32_34); - Instruction* ptr_tmp3 = new GetElementPtrInst(ptr_args_114, ptr_tmp3_indices.begin(), ptr_tmp3_indices.end(), "tmp3", label_from_consts_exit_111); + Instruction* ptr_tmp3 = new GetElementPtrInst(ptr_args_127, ptr_tmp3_indices.begin(), ptr_tmp3_indices.end(), "tmp3", label_from_consts_exit_124); std::vector ptr_arraydecay4_indices; - ptr_arraydecay4_indices.push_back(const_int32_30); - ptr_arraydecay4_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay4 = new GetElementPtrInst(ptr_consts_112, ptr_arraydecay4_indices.begin(), ptr_arraydecay4_indices.end(), "arraydecay4", label_from_consts_exit_111); - StoreInst* void_137 = new StoreInst(ptr_arraydecay4, ptr_tmp3, false, label_from_consts_exit_111); + ptr_arraydecay4_indices.push_back(const_int32_29); + ptr_arraydecay4_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay4 = new GetElementPtrInst(ptr_consts_125, ptr_arraydecay4_indices.begin(), ptr_arraydecay4_indices.end(), "arraydecay4", label_from_consts_exit_124); + StoreInst* void_154 = new StoreInst(ptr_arraydecay4, ptr_tmp3, false, label_from_consts_exit_124); std::vector ptr_tmp5_indices; - ptr_tmp5_indices.push_back(const_int32_30); + ptr_tmp5_indices.push_back(const_int32_29); ptr_tmp5_indices.push_back(const_int32_33); - Instruction* ptr_tmp5 = new GetElementPtrInst(ptr_args_114, ptr_tmp5_indices.begin(), ptr_tmp5_indices.end(), "tmp5", label_from_consts_exit_111); + Instruction* ptr_tmp5 = new GetElementPtrInst(ptr_args_127, ptr_tmp5_indices.begin(), ptr_tmp5_indices.end(), "tmp5", label_from_consts_exit_124); std::vector ptr_arraydecay6_indices; - ptr_arraydecay6_indices.push_back(const_int32_30); - ptr_arraydecay6_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay6 = new GetElementPtrInst(ptr_temps_113, ptr_arraydecay6_indices.begin(), ptr_arraydecay6_indices.end(), "arraydecay6", label_from_consts_exit_111); - StoreInst* void_138 = new StoreInst(ptr_arraydecay6, ptr_tmp5, false, label_from_consts_exit_111); - std::vector ptr_tmp8_139_indices; - ptr_tmp8_139_indices.push_back(const_int32_30); - ptr_tmp8_139_indices.push_back(const_int32_32); - Instruction* ptr_tmp8_139 = new GetElementPtrInst(ptr_args_114, ptr_tmp8_139_indices.begin(), ptr_tmp8_139_indices.end(), "tmp8", label_from_consts_exit_111); - std::vector ptr_tmp12_140_indices; - ptr_tmp12_140_indices.push_back(const_int32_30); - ptr_tmp12_140_indices.push_back(const_int32_30); - Instruction* ptr_tmp12_140 = new GetElementPtrInst(ptr_args_114, ptr_tmp12_140_indices.begin(), ptr_tmp12_140_indices.end(), "tmp12", label_from_consts_exit_111); - std::vector ptr_arraydecay11_141_indices; - ptr_arraydecay11_141_indices.push_back(const_int32_30); - ptr_arraydecay11_141_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay11_141 = new GetElementPtrInst(ptr_inputs_105, ptr_arraydecay11_141_indices.begin(), ptr_arraydecay11_141_indices.end(), "arraydecay11", label_from_consts_exit_111); - StoreInst* void_142 = new StoreInst(ptr_arraydecay11_141, ptr_tmp8_139, false, label_from_consts_exit_111); - std::vector ptr_arraydecay16_143_indices; - ptr_arraydecay16_143_indices.push_back(const_int32_30); - ptr_arraydecay16_143_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay16_143 = new GetElementPtrInst(ptr_results_104, ptr_arraydecay16_143_indices.begin(), ptr_arraydecay16_143_indices.end(), "arraydecay16", label_from_consts_exit_111); - StoreInst* void_144 = new StoreInst(ptr_arraydecay16_143, ptr_tmp12_140, false, label_from_consts_exit_111); - StoreInst* void_145 = new StoreInst(const_int32_30, ptr_tmp, false, label_from_consts_exit_111); - CallInst* void_146 = new CallInst(func_execute_shader, ptr_args_114, "", label_from_consts_exit_111); - void_146->setCallingConv(CallingConv::C); - void_146->setTailCall(false); - LoadInst* int32_tmp23 = new LoadInst(ptr_tmp, "tmp23", false, label_from_consts_exit_111); + ptr_arraydecay6_indices.push_back(const_int32_29); + ptr_arraydecay6_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay6 = new GetElementPtrInst(ptr_temps_126, ptr_arraydecay6_indices.begin(), ptr_arraydecay6_indices.end(), "arraydecay6", label_from_consts_exit_124); + StoreInst* void_155 = new StoreInst(ptr_arraydecay6, ptr_tmp5, false, label_from_consts_exit_124); + std::vector ptr_tmp8_156_indices; + ptr_tmp8_156_indices.push_back(const_int32_29); + ptr_tmp8_156_indices.push_back(const_int32_31); + Instruction* ptr_tmp8_156 = new GetElementPtrInst(ptr_args_127, ptr_tmp8_156_indices.begin(), ptr_tmp8_156_indices.end(), "tmp8", label_from_consts_exit_124); + std::vector ptr_tmp12_157_indices; + ptr_tmp12_157_indices.push_back(const_int32_29); + ptr_tmp12_157_indices.push_back(const_int32_29); + Instruction* ptr_tmp12_157 = new GetElementPtrInst(ptr_args_127, ptr_tmp12_157_indices.begin(), ptr_tmp12_157_indices.end(), "tmp12", label_from_consts_exit_124); + std::vector ptr_arraydecay11_158_indices; + ptr_arraydecay11_158_indices.push_back(const_int32_29); + ptr_arraydecay11_158_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay11_158 = new GetElementPtrInst(ptr_inputs_117, ptr_arraydecay11_158_indices.begin(), ptr_arraydecay11_158_indices.end(), "arraydecay11", label_from_consts_exit_124); + StoreInst* void_159 = new StoreInst(ptr_arraydecay11_158, ptr_tmp8_156, false, label_from_consts_exit_124); + std::vector ptr_arraydecay16_160_indices; + ptr_arraydecay16_160_indices.push_back(const_int32_29); + ptr_arraydecay16_160_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay16_160 = new GetElementPtrInst(ptr_results_116, ptr_arraydecay16_160_indices.begin(), ptr_arraydecay16_160_indices.end(), "arraydecay16", label_from_consts_exit_124); + StoreInst* void_161 = new StoreInst(ptr_arraydecay16_160, ptr_tmp12_157, false, label_from_consts_exit_124); + StoreInst* void_162 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); + CallInst* void_163 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); + void_163->setCallingConv(CallingConv::C); + void_163->setTailCall(false);ParamAttrsList *void_163_PAL = 0; + void_163->setParamAttrs(void_163_PAL); + + LoadInst* int32_tmp23 = new LoadInst(ptr_tmp, "tmp23", false, label_from_consts_exit_124); std::vector ptr_arraydecay11_1_indices; - ptr_arraydecay11_1_indices.push_back(const_int32_32); - ptr_arraydecay11_1_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay11_1 = new GetElementPtrInst(ptr_inputs_105, ptr_arraydecay11_1_indices.begin(), ptr_arraydecay11_1_indices.end(), "arraydecay11.1", label_from_consts_exit_111); - StoreInst* void_147 = new StoreInst(ptr_arraydecay11_1, ptr_tmp8_139, false, label_from_consts_exit_111); + ptr_arraydecay11_1_indices.push_back(const_int32_31); + ptr_arraydecay11_1_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay11_1 = new GetElementPtrInst(ptr_inputs_117, ptr_arraydecay11_1_indices.begin(), ptr_arraydecay11_1_indices.end(), "arraydecay11.1", label_from_consts_exit_124); + StoreInst* void_164 = new StoreInst(ptr_arraydecay11_1, ptr_tmp8_156, false, label_from_consts_exit_124); std::vector ptr_arraydecay16_1_indices; - ptr_arraydecay16_1_indices.push_back(const_int32_32); - ptr_arraydecay16_1_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay16_1 = new GetElementPtrInst(ptr_results_104, ptr_arraydecay16_1_indices.begin(), ptr_arraydecay16_1_indices.end(), "arraydecay16.1", label_from_consts_exit_111); - StoreInst* void_148 = new StoreInst(ptr_arraydecay16_1, ptr_tmp12_140, false, label_from_consts_exit_111); - StoreInst* void_149 = new StoreInst(const_int32_30, ptr_tmp, false, label_from_consts_exit_111); - CallInst* void_150 = new CallInst(func_execute_shader, ptr_args_114, "", label_from_consts_exit_111); - void_150->setCallingConv(CallingConv::C); - void_150->setTailCall(false); - LoadInst* int32_tmp23_1 = new LoadInst(ptr_tmp, "tmp23.1", false, label_from_consts_exit_111); - BinaryOperator* int32_shl_1 = BinaryOperator::create(Instruction::Shl, int32_tmp23_1, const_int32_32, "shl.1", label_from_consts_exit_111); - BinaryOperator* int32_or_1 = BinaryOperator::create(Instruction::Or, int32_shl_1, int32_tmp23, "or.1", label_from_consts_exit_111); + ptr_arraydecay16_1_indices.push_back(const_int32_31); + ptr_arraydecay16_1_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay16_1 = new GetElementPtrInst(ptr_results_116, ptr_arraydecay16_1_indices.begin(), ptr_arraydecay16_1_indices.end(), "arraydecay16.1", label_from_consts_exit_124); + StoreInst* void_165 = new StoreInst(ptr_arraydecay16_1, ptr_tmp12_157, false, label_from_consts_exit_124); + StoreInst* void_166 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); + CallInst* void_167 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); + void_167->setCallingConv(CallingConv::C); + void_167->setTailCall(false);ParamAttrsList *void_167_PAL = 0; + void_167->setParamAttrs(void_167_PAL); + + LoadInst* int32_tmp23_1 = new LoadInst(ptr_tmp, "tmp23.1", false, label_from_consts_exit_124); + BinaryOperator* int32_shl_1 = BinaryOperator::create(Instruction::Shl, int32_tmp23_1, const_int32_31, "shl.1", label_from_consts_exit_124); + BinaryOperator* int32_or_1 = BinaryOperator::create(Instruction::Or, int32_shl_1, int32_tmp23, "or.1", label_from_consts_exit_124); std::vector ptr_arraydecay11_2_indices; ptr_arraydecay11_2_indices.push_back(const_int32_33); - ptr_arraydecay11_2_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay11_2 = new GetElementPtrInst(ptr_inputs_105, ptr_arraydecay11_2_indices.begin(), ptr_arraydecay11_2_indices.end(), "arraydecay11.2", label_from_consts_exit_111); - StoreInst* void_151 = new StoreInst(ptr_arraydecay11_2, ptr_tmp8_139, false, label_from_consts_exit_111); + ptr_arraydecay11_2_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay11_2 = new GetElementPtrInst(ptr_inputs_117, ptr_arraydecay11_2_indices.begin(), ptr_arraydecay11_2_indices.end(), "arraydecay11.2", label_from_consts_exit_124); + StoreInst* void_168 = new StoreInst(ptr_arraydecay11_2, ptr_tmp8_156, false, label_from_consts_exit_124); std::vector ptr_arraydecay16_2_indices; ptr_arraydecay16_2_indices.push_back(const_int32_33); - ptr_arraydecay16_2_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay16_2 = new GetElementPtrInst(ptr_results_104, ptr_arraydecay16_2_indices.begin(), ptr_arraydecay16_2_indices.end(), "arraydecay16.2", label_from_consts_exit_111); - StoreInst* void_152 = new StoreInst(ptr_arraydecay16_2, ptr_tmp12_140, false, label_from_consts_exit_111); - StoreInst* void_153 = new StoreInst(const_int32_30, ptr_tmp, false, label_from_consts_exit_111); - CallInst* void_154 = new CallInst(func_execute_shader, ptr_args_114, "", label_from_consts_exit_111); - void_154->setCallingConv(CallingConv::C); - void_154->setTailCall(false); - LoadInst* int32_tmp23_2 = new LoadInst(ptr_tmp, "tmp23.2", false, label_from_consts_exit_111); - BinaryOperator* int32_shl_2 = BinaryOperator::create(Instruction::Shl, int32_tmp23_2, const_int32_33, "shl.2", label_from_consts_exit_111); - BinaryOperator* int32_or_2 = BinaryOperator::create(Instruction::Or, int32_shl_2, int32_or_1, "or.2", label_from_consts_exit_111); + ptr_arraydecay16_2_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay16_2 = new GetElementPtrInst(ptr_results_116, ptr_arraydecay16_2_indices.begin(), ptr_arraydecay16_2_indices.end(), "arraydecay16.2", label_from_consts_exit_124); + StoreInst* void_169 = new StoreInst(ptr_arraydecay16_2, ptr_tmp12_157, false, label_from_consts_exit_124); + StoreInst* void_170 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); + CallInst* void_171 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); + void_171->setCallingConv(CallingConv::C); + void_171->setTailCall(false);ParamAttrsList *void_171_PAL = 0; + void_171->setParamAttrs(void_171_PAL); + + LoadInst* int32_tmp23_2 = new LoadInst(ptr_tmp, "tmp23.2", false, label_from_consts_exit_124); + BinaryOperator* int32_shl_2 = BinaryOperator::create(Instruction::Shl, int32_tmp23_2, const_int32_33, "shl.2", label_from_consts_exit_124); + BinaryOperator* int32_or_2 = BinaryOperator::create(Instruction::Or, int32_shl_2, int32_or_1, "or.2", label_from_consts_exit_124); std::vector ptr_arraydecay11_3_indices; ptr_arraydecay11_3_indices.push_back(const_int32_34); - ptr_arraydecay11_3_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay11_3 = new GetElementPtrInst(ptr_inputs_105, ptr_arraydecay11_3_indices.begin(), ptr_arraydecay11_3_indices.end(), "arraydecay11.3", label_from_consts_exit_111); - StoreInst* void_155 = new StoreInst(ptr_arraydecay11_3, ptr_tmp8_139, false, label_from_consts_exit_111); + ptr_arraydecay11_3_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay11_3 = new GetElementPtrInst(ptr_inputs_117, ptr_arraydecay11_3_indices.begin(), ptr_arraydecay11_3_indices.end(), "arraydecay11.3", label_from_consts_exit_124); + StoreInst* void_172 = new StoreInst(ptr_arraydecay11_3, ptr_tmp8_156, false, label_from_consts_exit_124); std::vector ptr_arraydecay16_3_indices; ptr_arraydecay16_3_indices.push_back(const_int32_34); - ptr_arraydecay16_3_indices.push_back(const_int32_30); - Instruction* ptr_arraydecay16_3 = new GetElementPtrInst(ptr_results_104, ptr_arraydecay16_3_indices.begin(), ptr_arraydecay16_3_indices.end(), "arraydecay16.3", label_from_consts_exit_111); - StoreInst* void_156 = new StoreInst(ptr_arraydecay16_3, ptr_tmp12_140, false, label_from_consts_exit_111); - StoreInst* void_157 = new StoreInst(const_int32_30, ptr_tmp, false, label_from_consts_exit_111); - CallInst* void_158 = new CallInst(func_execute_shader, ptr_args_114, "", label_from_consts_exit_111); - void_158->setCallingConv(CallingConv::C); - void_158->setTailCall(false); - LoadInst* int32_tmp23_3 = new LoadInst(ptr_tmp, "tmp23.3", false, label_from_consts_exit_111); - BinaryOperator* int32_shl_3 = BinaryOperator::create(Instruction::Shl, int32_tmp23_3, const_int32_34, "shl.3", label_from_consts_exit_111); - BinaryOperator* int32_or_3 = BinaryOperator::create(Instruction::Or, int32_shl_3, int32_or_2, "or.3", label_from_consts_exit_111); - BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_or_3, const_int32_36, "neg", label_from_consts_exit_111); - new ReturnInst(int32_neg, label_from_consts_exit_111); + ptr_arraydecay16_3_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay16_3 = new GetElementPtrInst(ptr_results_116, ptr_arraydecay16_3_indices.begin(), ptr_arraydecay16_3_indices.end(), "arraydecay16.3", label_from_consts_exit_124); + StoreInst* void_173 = new StoreInst(ptr_arraydecay16_3, ptr_tmp12_157, false, label_from_consts_exit_124); + StoreInst* void_174 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); + CallInst* void_175 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); + void_175->setCallingConv(CallingConv::C); + void_175->setTailCall(false);ParamAttrsList *void_175_PAL = 0; + void_175->setParamAttrs(void_175_PAL); + + LoadInst* int32_tmp23_3 = new LoadInst(ptr_tmp, "tmp23.3", false, label_from_consts_exit_124); + BinaryOperator* int32_shl_3 = BinaryOperator::create(Instruction::Shl, int32_tmp23_3, const_int32_34, "shl.3", label_from_consts_exit_124); + BinaryOperator* int32_or_3 = BinaryOperator::create(Instruction::Or, int32_shl_3, int32_or_2, "or.3", label_from_consts_exit_124); + BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_or_3, const_int32_30, "neg", label_from_consts_exit_124); + new ReturnInst(int32_neg, label_from_consts_exit_124); // Resolve Forward References - fwdref_121->replaceAllUsesWith(packed_tmp31_i_133); delete fwdref_121; - fwdref_119->replaceAllUsesWith(int32_indvar_next7); delete fwdref_119; + fwdref_138->replaceAllUsesWith(packed_tmp31_i_150); delete fwdref_138; + fwdref_136->replaceAllUsesWith(int32_indvar_next7); delete fwdref_136; } -- cgit v1.2.3 From 9789c301b476b1127f847144fd5c8313e6ba8be8 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 4 Jan 2008 09:27:42 -0500 Subject: llvm: we need custom rules so had to redo the build a little bit also don't use sse when llvm is enabled --- src/mesa/Makefile | 11 +++-- src/mesa/pipe/Makefile | 6 ++- src/mesa/pipe/Makefile.template | 10 +++-- src/mesa/pipe/llvm/Makefile | 80 +++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/llvm_base_shader.cpp | 16 +++---- src/mesa/pipe/softpipe/sp_state_fs.c | 12 +++-- src/mesa/sources | 11 +---- 7 files changed, 113 insertions(+), 33 deletions(-) create mode 100644 src/mesa/pipe/llvm/Makefile (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index dd8e2c3762..681c566140 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -20,14 +20,17 @@ CELL_LIB = $(TOP)/src/mesa/pipe/cell/ppu/libcell.a CELL_LIB_SPU = $(TOP)/src/mesa/pipe/cell/spu/g3d_spu.a endif +ifeq ($(CONFIG_NAME), linux-llvm) +LLVM_LIB = $(TOP)/src/mesa/pipe/llvm/libgallivm.a +endif .SUFFIXES : .cpp .c.o: - $(CC) -c $(LLVM_CFLAGS) $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ .cpp.o: - $(CXX) -c $(LLVM_CXXFLAGS) $(INCLUDE_DIRS) $(CXXFLAGS) $< -o $@ + $(CXX) -c $(INCLUDE_DIRS) $(CXXFLAGS) $< -o $@ .S.o: $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ @@ -120,12 +123,12 @@ stand-alone: depend subdirs $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) $(TOP)/$(LIB_DIR)/$ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(LLVM_LIB) @ $(TOP)/bin/mklib -o $(GL_LIB) \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ - $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) + $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(LLVM_LIB) $(GL_LIB_DEPS) # Make the OSMesa library $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECTS) diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile index c0345a9cb5..da044036ff 100644 --- a/src/mesa/pipe/Makefile +++ b/src/mesa/pipe/Makefile @@ -6,7 +6,11 @@ ifeq ($(CONFIG_NAME), linux-cell) CELL_DIR = cell endif -SUBDIRS = softpipe i915simple i965simple failover pipebuffer $(CELL_DIR) +ifeq ($(CONFIG_NAME), linux-llvm) +LLVM_DIR = llvm +endif + +SUBDIRS = softpipe llvm i915simple i965simple failover pipebuffer $(CELL_DIR) $(LLVM_DIR) default: subdirs diff --git a/src/mesa/pipe/Makefile.template b/src/mesa/pipe/Makefile.template index 8c2f84b328..3cd07660b6 100644 --- a/src/mesa/pipe/Makefile.template +++ b/src/mesa/pipe/Makefile.template @@ -8,6 +8,7 @@ COMMON_SOURCES = OBJECTS = $(C_SOURCES:.c=.o) \ + $(CPP_SOURCES:.cpp=.o) \ $(ASM_SOURCES:.S=.o) @@ -22,7 +23,10 @@ INCLUDES = \ ##### RULES ##### .c.o: - $(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ .S.o: $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ @@ -37,10 +41,10 @@ $(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/mesa/pipe/Makefile.template $(TOP)/bin/mklib -o $@ -static $(OBJECTS) -depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) +depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) rm -f depend touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ $(ASM_SOURCES) 2> /dev/null diff --git a/src/mesa/pipe/llvm/Makefile b/src/mesa/pipe/llvm/Makefile new file mode 100644 index 0000000000..b1463f67cf --- /dev/null +++ b/src/mesa/pipe/llvm/Makefile @@ -0,0 +1,80 @@ +# -*-makefile-*- +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = gallivm + + +GALLIVM_SOURCES = \ + gallivm.cpp \ + instructions.cpp \ + storage.cpp + +INC_SOURCES = gallivm_builtins.cpp llvm_base_shader.cpp + +CPP_SOURCES = \ + $(GALLIVM_SOURCES) + +C_SOURCES = +ASM_SOURCES = + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(CPP_SOURCES:.cpp=.o) \ + $(ASM_SOURCES:.S=.o) + +### Include directories +INCLUDES = \ + -I. \ + -I$(TOP)/src/mesa/pipe \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDES) $(LLVM_CXXFLAGS) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +##### TARGETS ##### + +default:: depend symlinks $(LIBNAME) + + +$(LIBNAME): $(OBJECTS) Makefile + $(TOP)/bin/mklib -o $@ -static $(OBJECTS) + + +depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ + $(ASM_SOURCES) $(INC_SOURCES) 2> /dev/null + + +gallivm_builtins.cpp: llvm_builtins.c + clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins + +llvm_base_shader.cpp: llvm_entry.c + clang --emit-llvm $< |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -o=$@ -funcname=createBaseShader + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean: + -rm -f *.o */*.o *~ *.so *~ server/*.o + -rm -f depend depend.bak + -rm -f gallivm_builtins.cpp llvm_base_shader.cpp + +symlinks: + + +include depend diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index 951afc390e..b574b550ae 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -209,7 +209,7 @@ Module* createBaseShader() { /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_vertex_shader", mod); func_run_vertex_shader->setCallingConv(CallingConv::C); - ParamAttrsList *func_run_vertex_shader_PAL = 0; + const ParamAttrsList *func_run_vertex_shader_PAL = 0; func_run_vertex_shader->setParamAttrs(func_run_vertex_shader_PAL); Function* func_execute_shader = new Function( @@ -217,7 +217,7 @@ Module* createBaseShader() { /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"execute_shader", mod); // (external, no body) func_execute_shader->setCallingConv(CallingConv::C); - ParamAttrsList *func_execute_shader_PAL = 0; + const ParamAttrsList *func_execute_shader_PAL = 0; func_execute_shader->setParamAttrs(func_execute_shader_PAL); Function* func_run_fragment_shader = new Function( @@ -225,7 +225,7 @@ Module* createBaseShader() { /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"run_fragment_shader", mod); func_run_fragment_shader->setCallingConv(CallingConv::C); - ParamAttrsList *func_run_fragment_shader_PAL = 0; + const ParamAttrsList *func_run_fragment_shader_PAL = 0; func_run_fragment_shader->setParamAttrs(func_run_fragment_shader_PAL); // Global Variable Declarations @@ -644,7 +644,7 @@ Module* createBaseShader() { StoreInst* void_110 = new StoreInst(ptr_arraydecay16, ptr_tmp12, false, label_forbody_92); CallInst* void_111 = new CallInst(func_execute_shader, ptr_args, "", label_forbody_92); void_111->setCallingConv(CallingConv::C); - void_111->setTailCall(false);ParamAttrsList *void_111_PAL = 0; + void_111->setTailCall(false);const ParamAttrsList *void_111_PAL = 0; void_111->setParamAttrs(void_111_PAL); BinaryOperator* int32_indvar_next_112 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_106, const_int32_31, "indvar.next", label_forbody_92); @@ -790,7 +790,7 @@ Module* createBaseShader() { StoreInst* void_162 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); CallInst* void_163 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); void_163->setCallingConv(CallingConv::C); - void_163->setTailCall(false);ParamAttrsList *void_163_PAL = 0; + void_163->setTailCall(false);const ParamAttrsList *void_163_PAL = 0; void_163->setParamAttrs(void_163_PAL); LoadInst* int32_tmp23 = new LoadInst(ptr_tmp, "tmp23", false, label_from_consts_exit_124); @@ -807,7 +807,7 @@ Module* createBaseShader() { StoreInst* void_166 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); CallInst* void_167 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); void_167->setCallingConv(CallingConv::C); - void_167->setTailCall(false);ParamAttrsList *void_167_PAL = 0; + void_167->setTailCall(false);const ParamAttrsList *void_167_PAL = 0; void_167->setParamAttrs(void_167_PAL); LoadInst* int32_tmp23_1 = new LoadInst(ptr_tmp, "tmp23.1", false, label_from_consts_exit_124); @@ -826,7 +826,7 @@ Module* createBaseShader() { StoreInst* void_170 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); CallInst* void_171 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); void_171->setCallingConv(CallingConv::C); - void_171->setTailCall(false);ParamAttrsList *void_171_PAL = 0; + void_171->setTailCall(false);const ParamAttrsList *void_171_PAL = 0; void_171->setParamAttrs(void_171_PAL); LoadInst* int32_tmp23_2 = new LoadInst(ptr_tmp, "tmp23.2", false, label_from_consts_exit_124); @@ -845,7 +845,7 @@ Module* createBaseShader() { StoreInst* void_174 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); CallInst* void_175 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); void_175->setCallingConv(CallingConv::C); - void_175->setTailCall(false);ParamAttrsList *void_175_PAL = 0; + void_175->setTailCall(false);const ParamAttrsList *void_175_PAL = 0; void_175->setParamAttrs(void_175_PAL); LoadInst* int32_tmp23_3 = new LoadInst(ptr_tmp, "tmp23.3", false, label_from_consts_exit_124); diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 830c873b25..945c93411f 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -59,13 +59,6 @@ softpipe_create_fs_state(struct pipe_context *pipe, tgsi_dump(state->shader.tokens, 0); } -#if defined(__i386__) || defined(__386__) - if (softpipe->use_sse) { - x86_init_func( &state->sse2_program ); - tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); - } -#endif - #ifdef MESA_LLVM state->llvm_prog = gallivm_from_tgsi(state->shader.tokens, GALLIVM_FS); if (!gallivm_global_cpu_engine()) { @@ -73,6 +66,11 @@ softpipe_create_fs_state(struct pipe_context *pipe, } else gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); +#elif defined(__i386__) || defined(__386__) + if (softpipe->use_sse) { + x86_init_func( &state->sse2_program ); + tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); + } #endif return state; diff --git a/src/mesa/sources b/src/mesa/sources index 56ea6dbce2..90ce9ce370 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -186,12 +186,6 @@ TGSIUTIL_SOURCES = \ pipe/tgsi/util/tgsi_parse.c \ pipe/tgsi/util/tgsi_util.c - -LLVMTGSI_SOURCES = \ - pipe/llvm/gallivm.cpp \ - pipe/llvm/storage.cpp \ - pipe/llvm/instructions.cpp - STATECACHE_SOURCES = \ pipe/cso_cache/cso_hash.c \ pipe/cso_cache/cso_cache.c @@ -370,7 +364,6 @@ FBDEV_DRIVER_SOURCES = \ ALL_SOURCES = \ $(GLAPI_SOURCES) \ $(SOLO_SOURCES) \ - $(LLVMTGSI_SOURCES) \ $(ASM_SOURCES) \ $(COMMON_DRIVER_SOURCES)\ $(X11_DRIVER_SOURCES) \ @@ -397,15 +390,13 @@ SOLO_SOURCES = \ CORE_SOURCES = \ $(GLAPI_SOURCES) \ - $(SOLO_SOURCES) \ - $(LLVMTGSI_SOURCES) + $(SOLO_SOURCES) ### Object files SOLO_OBJECTS = \ $(SOLO_SOURCES:.c=.o) \ - $(LLVMTGSI_SOURCES:.cpp=.o) \ $(ASM_SOURCES:.S=.o) GLAPI_OBJECTS = \ -- cgit v1.2.3 From 6f012904318311207a20bbf586f1a9f9f8b7fc20 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 4 Jan 2008 10:01:32 -0500 Subject: llvm: if llvm is enabled don't even try to use sse for vs --- src/mesa/pipe/draw/draw_vertex_shader.c | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 240149118f..c2e038453e 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -245,15 +245,6 @@ draw_create_vertex_shader(struct draw_context *draw, vs->state = shader; -#if defined(__i386__) || defined(__386__) - if (draw->use_sse) { - /* cast-away const */ - struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; - - x86_init_func( &vs->sse2_program ); - tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, &vs->sse2_program ); - } -#endif #ifdef MESA_LLVM vs->llvm_prog = gallivm_from_tgsi(shader->tokens, GALLIVM_VS); draw->engine = gallivm_global_cpu_engine(); @@ -263,6 +254,14 @@ draw_create_vertex_shader(struct draw_context *draw, else { gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); } +#elif defined(__i386__) || defined(__386__) + if (draw->use_sse) { + /* cast-away const */ + struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; + + x86_init_func( &vs->sse2_program ); + tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, &vs->sse2_program ); + } #endif return vs; -- cgit v1.2.3 From 1c5f27a18b775b3784fcd265d60e0affa0b31581 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 4 Jan 2008 17:06:55 +0100 Subject: gallium: Make texture target an enum for better debuggability. Also make enum pipe_format used in a couple more places. --- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 6 +++--- src/mesa/pipe/p_defines.h | 10 ++++++---- src/mesa/pipe/p_state.h | 5 +++-- src/mesa/state_tracker/st_cb_texture.c | 2 +- src/mesa/state_tracker/st_texture.c | 4 ++-- src/mesa/state_tracker/st_texture.h | 4 ++-- 6 files changed, 17 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index 5c7dee5790..fc40e0438c 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -33,7 +33,7 @@ #include "brw_state.h" #include "brw_defines.h" -static unsigned translate_tex_target( int target ) +static unsigned translate_tex_target( enum pipe_texture_target target ) { switch (target) { case PIPE_TEXTURE_1D: @@ -54,9 +54,9 @@ static unsigned translate_tex_target( int target ) } } -static unsigned translate_tex_format( unsigned mesa_format ) +static unsigned translate_tex_format( enum pipe_format pipe_format ) { - switch( mesa_format ) { + switch( pipe_format ) { case PIPE_FORMAT_U_L8: return BRW_SURFACEFORMAT_L8_UNORM; diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index a67ed60547..c6d9c02bd9 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -115,10 +115,12 @@ #define PIPE_STENCIL_OP_INVERT 7 /** Texture types */ -#define PIPE_TEXTURE_1D 0 -#define PIPE_TEXTURE_2D 1 -#define PIPE_TEXTURE_3D 2 -#define PIPE_TEXTURE_CUBE 3 +enum pipe_texture_target { + PIPE_TEXTURE_1D = 0, + PIPE_TEXTURE_2D = 1, + PIPE_TEXTURE_3D = 2, + PIPE_TEXTURE_CUBE = 3 +}; #define PIPE_TEX_FACE_POS_X 0 #define PIPE_TEX_FACE_NEG_X 1 diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 16d50fdb82..76e633e930 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -39,6 +39,7 @@ #define PIPE_STATE_H #include "p_compiler.h" +#include "p_defines.h" #include "p_format.h" /** @@ -262,8 +263,8 @@ struct pipe_texture { /* Effectively the key: */ - unsigned target; /**< PIPE_TEXTURE_x */ - enum pipe_format format; /**< PIPE_FORMAT_x */ + enum pipe_texture_target target; /**< PIPE_TEXTURE_x */ + enum pipe_format format; /**< PIPE_FORMAT_x */ unsigned first_level; unsigned last_level; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index e813bdb47a..773fc0012e 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -97,7 +97,7 @@ st_get_texobj_texture(struct gl_texture_object *texObj) } -static unsigned +static enum pipe_texture_target gl_target_to_pipe(GLenum target) { switch (target) { diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index c7d28eeca2..a2bdf846ca 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -61,8 +61,8 @@ target_to_target(GLenum target) struct pipe_texture * st_texture_create(struct st_context *st, - unsigned target, - unsigned format, + enum pipe_texture_target target, + enum pipe_format format, GLuint first_level, GLuint last_level, GLuint width0, diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 7524c219e0..d8b1bcad9d 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -37,8 +37,8 @@ struct pipe_texture; extern struct pipe_texture * st_texture_create(struct st_context *st, - unsigned target, - unsigned format, + enum pipe_texture_target target, + enum pipe_format format, GLuint first_level, GLuint last_level, GLuint width0, -- cgit v1.2.3 From 934468296c33a04125e42e5d0f8747f65405a393 Mon Sep 17 00:00:00 2001 From: Jerome Glisse Date: Thu, 10 Jan 2008 01:11:53 +0100 Subject: softpipe: map only once in softpipe_map_surfaces softpipe_map_surfaces get call several time but softpipe_unmap_surfaces get call only once. So to make sure stuff are properly unmap when softpipe_unmap_surfaces get call we map surfaces only one time in softpipe_map_surfaces. --- src/mesa/pipe/softpipe/sp_context.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index c7af63cc2d..0759092305 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -80,16 +80,16 @@ softpipe_map_surfaces(struct softpipe_context *sp) for (i = 0; i < sp->framebuffer.num_cbufs; i++) { ps = sp->framebuffer.cbufs[i]; - if (ps->buffer) + if (ps->buffer && !ps->map) pipe_surface_map(ps); } ps = sp->framebuffer.zbuf; - if (ps && ps->buffer) + if (ps && ps->buffer && !ps->map) pipe_surface_map(ps); ps = sp->framebuffer.sbuf; - if (ps && ps->buffer) + if (ps && ps->buffer && !ps->map) pipe_surface_map(ps); } -- cgit v1.2.3 From abee68a7220e5ee16216caf22841ad934fb37334 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 9 Jan 2008 14:10:59 -0700 Subject: Cell: implemement basic Z testing Also, improve some surface clearing code --- src/mesa/pipe/cell/common.h | 17 +++++----- src/mesa/pipe/cell/ppu/cell_spu.c | 3 +- src/mesa/pipe/cell/ppu/cell_surface.c | 47 +++++++++++++------------- src/mesa/pipe/cell/spu/main.c | 44 ++++++++++++++++--------- src/mesa/pipe/cell/spu/main.h | 2 +- src/mesa/pipe/cell/spu/tri.c | 62 +++++++++++++++++++++++++++++++++-- 6 files changed, 124 insertions(+), 51 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index dff0b1f8ec..4c770f5c32 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -47,11 +47,11 @@ #define TILE_SIZE 32 -#define CELL_CMD_EXIT 1 -#define CELL_CMD_FRAMEBUFFER 2 -#define CELL_CMD_CLEAR_TILES 3 -#define CELL_CMD_FINISH 5 -#define CELL_CMD_RENDER 6 +#define CELL_CMD_EXIT 1 +#define CELL_CMD_FRAMEBUFFER 2 +#define CELL_CMD_CLEAR_SURFACE 3 +#define CELL_CMD_FINISH 4 +#define CELL_CMD_RENDER 5 /** @@ -66,10 +66,11 @@ struct cell_command_framebuffer /** - * Clear framebuffer tiles to given value/color. + * Clear framebuffer to the given value/color. */ -struct cell_command_clear_tiles +struct cell_command_clear_surface { + uint surface; /**< Temporary: 0=color, 1=Z */ uint value; } ALIGN16_ATTRIB; @@ -87,7 +88,7 @@ struct cell_command_render struct cell_command { struct cell_command_framebuffer fb; - struct cell_command_clear_tiles clear; + struct cell_command_clear_surface clear; struct cell_command_render render; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 15b89682fa..a7dbf24dd8 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -176,7 +176,8 @@ test_spus(struct cell_context *cell) for (i = 0; i < cell->num_spus; i++) { cell_global.command[i].clear.value = 0xff880044; /* XXX */ - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_TILES); + cell_global.command[i].clear.surface = 0; + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_SURFACE); } finish_all(cell->num_spus); diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 62e0febc0c..1e1548c8b6 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -47,42 +47,43 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, { struct cell_context *cell = cell_context(pipe); uint i; + uint surfIndex; if (!ps->map) pipe_surface_map(ps); - if (pf_get_size(ps->format) != 4) { - printf("Cell: Skipping non 32bpp clear_surface\n"); - return; + if (ps == cell->framebuffer.zbuf) { + surfIndex = 1; } -#if 0 - for (i = 0; i < cell->num_spus; i++) { - struct cell_command_framebuffer *fb = &cell_global.command[i].fb; - printf("%s %u start = 0x%x\n", __FUNCTION__, i, ps->map); - fb->color_start = ps->map; - fb->width = ps->width; - fb->height = ps->height; - fb->color_format = ps->format; - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FRAMEBUFFER); + else { + surfIndex = 0; } -#endif + + printf("Clear surf %u\n", surfIndex); for (i = 0; i < cell->num_spus; i++) { #if 1 uint clr = clearValue; - /* XXX debug: clear color varied per-SPU to visualize tiles */ - if ((clr & 0xff) == 0) - clr |= 64 + i * 8; - if ((clr & 0xff00) == 0) - clr |= (64 + i * 8) << 8; - if ((clr & 0xff0000) == 0) - clr |= (64 + i * 8) << 16; - if ((clr & 0xff000000) == 0) - clr |= (64 + i * 8) << 24; + if (surfIndex == 0) { + /* XXX debug: clear color varied per-SPU to visualize tiles */ + if ((clr & 0xff) == 0) + clr |= 64 + i * 8; + if ((clr & 0xff00) == 0) + clr |= (64 + i * 8) << 8; + if ((clr & 0xff0000) == 0) + clr |= (64 + i * 8) << 16; + if ((clr & 0xff000000) == 0) + clr |= (64 + i * 8) << 24; + } cell_global.command[i].clear.value = clr; #else cell_global.command[i].clear.value = clearValue; #endif - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_TILES); + cell_global.command[i].clear.surface = surfIndex; + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_SURFACE); } + + /* XXX temporary */ + cell_flush(&cell->pipe, 0x0); + } diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 9580281971..1552452ab7 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -50,7 +50,7 @@ volatile struct cell_init_info init; struct framebuffer fb; uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -uint ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; int DefaultTag; @@ -70,7 +70,7 @@ get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, int tag, int zBuf) { const uint offset = ty * fb->width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); const ubyte *src = zBuf ? fb->depth_start : fb->color_start; src += offset * bytesPerTile; @@ -96,7 +96,7 @@ put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, int tag, int zBuf) { const uint offset = ty * fb->width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * 4; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); ubyte *dst = zBuf ? fb->depth_start : fb->color_start; dst += offset * bytesPerTile; @@ -119,15 +119,22 @@ put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, static void -clear_tiles(const struct cell_command_clear_tiles *clear) +clear_surface(const struct cell_command_clear_surface *clear) { uint num_tiles = fb.width_tiles * fb.height_tiles; uint i, j; int tag = init.id; - for (i = 0; i < TILE_SIZE; i++) - for (j = 0; j < TILE_SIZE; j++) - ctile[i][j] = clear->value; + if (clear->surface == 0) { + for (i = 0; i < TILE_SIZE; i++) + for (j = 0; j < TILE_SIZE; j++) + ctile[i][j] = clear->value; + } + else { + for (i = 0; i < TILE_SIZE; i++) + for (j = 0; j < TILE_SIZE; j++) + ztile[i][j] = clear->value; + } /* printf("SPU: %s num=%d w=%d h=%d\n", @@ -137,7 +144,10 @@ clear_tiles(const struct cell_command_clear_tiles *clear) for (i = init.id; i < num_tiles; i += init.num_spus) { uint tx = i % fb.width_tiles; uint ty = i / fb.width_tiles; - put_tile(&fb, tx, ty, (uint *) ctile, tag, 0); + if (clear->surface == 0) + put_tile(&fb, tx, ty, (uint *) ctile, tag, 0); + else + put_tile(&fb, tx, ty, (uint *) ztile, tag, 1); /* XXX we don't want this here, but it fixes bad tile results */ wait_on_mask(1 << tag); } @@ -227,7 +237,7 @@ render(const struct cell_command_render *render) get_tile(&fb, tx, ty, (uint *) ctile, tag, 0); wait_on_mask(1 << tag); /* XXX temporary */ - if (fb.depth_format == PIPE_FORMAT_Z32_UNORM) { + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { get_tile(&fb, tx, ty, (uint *) ztile, tag+1, 1); wait_on_mask(1 << (tag+1)); /* XXX temporary */ } @@ -265,7 +275,7 @@ render(const struct cell_command_render *render) put_tile(&fb, tx, ty, (uint *) ctile, tag, 0); wait_on_mask(1 << tag); /* XXX temp */ - if (fb.depth_format == PIPE_FORMAT_Z32_UNORM) { + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { put_tile(&fb, tx, ty, (uint *) ztile, tag+1, 1); wait_on_mask(1 << (tag+1)); /* XXX temporary */ } @@ -313,11 +323,14 @@ main_loop(void) exitFlag = 1; break; case CELL_CMD_FRAMEBUFFER: - printf("SPU %u: FRAMEBUFFER: %d x %d at %p, format 0x%x\n", init.id, + printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", + init.id, cmd.fb.width, cmd.fb.height, cmd.fb.color_start, - cmd.fb.color_format); + cmd.fb.color_format, + cmd.fb.depth_format); + printf("Z16 = 0x%x\n", PIPE_FORMAT_Z16_UNORM); fb.color_start = cmd.fb.color_start; fb.depth_start = cmd.fb.depth_start; fb.color_format = cmd.fb.color_format; @@ -331,9 +344,10 @@ main_loop(void) init.id, fb.width_tiles, fb.height_tiles); */ break; - case CELL_CMD_CLEAR_TILES: - printf("SPU %u: CLEAR to 0x%08x\n", init.id, cmd.clear.value); - clear_tiles(&cmd.clear); + case CELL_CMD_CLEAR_SURFACE: + printf("SPU %u: CLEAR SURF %u to 0x%08x\n", init.id, + cmd.clear.surface, cmd.clear.value); + clear_surface(&cmd.clear); break; case CELL_CMD_RENDER: printf("SPU %u: RENDER %u verts, prim %u\n", diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index c22679c5d9..5e0516d0c1 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -50,7 +50,7 @@ struct framebuffer { extern struct framebuffer fb; extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -extern uint ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; extern int DefaultTag; diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 36599297b9..58fca2e34a 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -227,6 +227,22 @@ eval_coeff( struct setup_stage *setup, uint slot, } +static INLINE void +eval_z( struct setup_stage *setup, + float x, float y, float result[4]) +{ + uint slot = 0; + uint i = 2; + const float *dadx = setup->coef[slot].dadx; + const float *dady = setup->coef[slot].dady; + + result[QUAD_TOP_LEFT] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; + result[QUAD_TOP_RIGHT] = result[0] + dadx[i]; + result[QUAD_BOTTOM_LEFT] = result[0] + dady[i]; + result[QUAD_BOTTOM_RIGHT] = result[0] + dadx[i] + dady[i]; +} + + static INLINE uint pack_color(const float color[4]) { @@ -263,9 +279,48 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) int ix = x - cliprect_minx; int iy = y - cliprect_miny; float colors[4][4]; + uint z; eval_coeff(setup, 1, (float) x, (float) y, colors); + + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + float zvals[4]; + eval_z(setup, (float) x, (float) y, zvals); + + if (mask & MASK_TOP_LEFT) { + z = (uint) (zvals[0] * 65535.0); + if (z < ztile[iy][ix]) + ztile[iy][ix] = z; + else + mask &= ~MASK_TOP_LEFT; + } + + if (mask & MASK_TOP_RIGHT) { + z = (uint) (zvals[1] * 65535.0); + if (z < ztile[iy][ix+1]) + ztile[iy][ix+1] = z; + else + mask &= ~MASK_TOP_RIGHT; + } + + if (mask & MASK_BOTTOM_LEFT) { + z = (uint) (zvals[2] * 65535.0); + if (z < ztile[iy+1][ix]) + ztile[iy+1][ix] = z; + else + mask &= ~MASK_BOTTOM_LEFT; + } + + if (mask & MASK_BOTTOM_RIGHT) { + z = (uint) (zvals[3] * 65535.0); + if (z < ztile[iy+1][ix+1]) + ztile[iy+1][ix+1] = z; + else + mask &= ~MASK_BOTTOM_RIGHT; + } + } + if (mask & MASK_TOP_LEFT) ctile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); if (mask & MASK_TOP_RIGHT) @@ -512,10 +567,10 @@ static void const_coeff( struct setup_stage *setup, * for a triangle. */ static void tri_linear_coeff( struct setup_stage *setup, - unsigned slot ) + uint slot, uint firstComp, uint lastComp ) { uint i; - for (i = 0; i < 4; i++) { + for (i = firstComp; i < lastComp; i++) { float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; float a = setup->ebot.dy * majda - botda * setup->emaj.dy; @@ -637,7 +692,8 @@ static void setup_tri_coefficients( struct setup_stage *setup ) } } #else - tri_linear_coeff(setup, 1); /* slot 1 = color */ + tri_linear_coeff(setup, 0, 2, 3); /* slot 0, z */ + tri_linear_coeff(setup, 1, 0, 4); /* slot 1, color */ #endif } -- cgit v1.2.3 From d48c6e7b21839791ff53308187cc2aeed9fb618c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 9 Jan 2008 14:14:24 -0700 Subject: Cell: remove some debug printfs, predicate others with Debug boolean --- src/mesa/pipe/cell/ppu/cell_flush.c | 2 -- src/mesa/pipe/cell/ppu/cell_render.c | 3 --- src/mesa/pipe/cell/ppu/cell_surface.c | 2 -- src/mesa/pipe/cell/spu/main.c | 49 +++++++++++++++++++++-------------- 4 files changed, 30 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c index 47003bef18..977c50da32 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -38,8 +38,6 @@ cell_flush(struct pipe_context *pipe, unsigned flags) struct cell_context *cell = cell_context(pipe); uint i; - printf("%s\n", __FUNCTION__); - cell_flush_prim_buffer(cell); /* Send CMD_FINISH to all SPUs */ diff --git a/src/mesa/pipe/cell/ppu/cell_render.c b/src/mesa/pipe/cell/ppu/cell_render.c index f5a91776a5..ecdd47e28b 100644 --- a/src/mesa/pipe/cell/ppu/cell_render.c +++ b/src/mesa/pipe/cell/ppu/cell_render.c @@ -123,8 +123,6 @@ render_tri(struct draw_stage *stage, struct prim_header *prim) struct cell_prim_buffer *buf = &cell->prim_buffer; uint i; - printf("Cell render tri\n"); - if (buf->num_verts + 3 > CELL_MAX_VERTS) { cell_flush_prim_buffer(cell); } @@ -150,7 +148,6 @@ cell_flush_prim_buffer(struct cell_context *cell) if (cell->prim_buffer.num_verts == 0) return; - printf("*** Flushing prim buffer\n"); for (i = 0; i < cell->num_spus; i++) { struct cell_command_render *render = &cell_global.command[i].render; render->prim_type = PIPE_PRIM_TRIANGLES; diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 1e1548c8b6..03dd41583f 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -59,8 +59,6 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, surfIndex = 0; } - printf("Clear surf %u\n", surfIndex); - for (i = 0; i < cell->num_spus; i++) { #if 1 uint clr = clearValue; diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 1552452ab7..8e7f8be1eb 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -45,6 +45,8 @@ helpful headers: /opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h */ +static boolean Debug = TRUE; + volatile struct cell_init_info init; struct framebuffer fb; @@ -292,7 +294,8 @@ main_loop(void) struct cell_command cmd; int exitFlag = 0; - printf("SPU %u: Enter main loop\n", init.id); + if (Debug) + printf("SPU %u: Enter main loop\n", init.id); assert((sizeof(struct cell_command) & 0xf) == 0); ASSERT_ALIGN16(&cmd); @@ -301,12 +304,14 @@ main_loop(void) unsigned opcode; int tag = 0; - printf("SPU %u: Wait for cmd...\n", init.id); + if (Debug) + printf("SPU %u: Wait for cmd...\n", init.id); /* read/wait from mailbox */ opcode = (unsigned int) spu_read_in_mbox(); - printf("SPU %u: got cmd %u\n", init.id, opcode); + if (Debug) + printf("SPU %u: got cmd %u\n", init.id, opcode); /* command payload */ mfc_get(&cmd, /* dest */ @@ -319,18 +324,19 @@ main_loop(void) switch (opcode) { case CELL_CMD_EXIT: - printf("SPU %u: EXIT\n", init.id); + if (Debug) + printf("SPU %u: EXIT\n", init.id); exitFlag = 1; break; case CELL_CMD_FRAMEBUFFER: - printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", - init.id, - cmd.fb.width, - cmd.fb.height, - cmd.fb.color_start, - cmd.fb.color_format, - cmd.fb.depth_format); - printf("Z16 = 0x%x\n", PIPE_FORMAT_Z16_UNORM); + if (Debug) + printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", + init.id, + cmd.fb.width, + cmd.fb.height, + cmd.fb.color_start, + cmd.fb.color_format, + cmd.fb.depth_format); fb.color_start = cmd.fb.color_start; fb.depth_start = cmd.fb.depth_start; fb.color_format = cmd.fb.color_format; @@ -345,18 +351,21 @@ main_loop(void) */ break; case CELL_CMD_CLEAR_SURFACE: - printf("SPU %u: CLEAR SURF %u to 0x%08x\n", init.id, - cmd.clear.surface, cmd.clear.value); + if (Debug) + printf("SPU %u: CLEAR SURF %u to 0x%08x\n", init.id, + cmd.clear.surface, cmd.clear.value); clear_surface(&cmd.clear); break; case CELL_CMD_RENDER: - printf("SPU %u: RENDER %u verts, prim %u\n", - init.id, cmd.render.num_verts, cmd.render.prim_type); + if (Debug) + printf("SPU %u: RENDER %u verts, prim %u\n", + init.id, cmd.render.num_verts, cmd.render.prim_type); render(&cmd.render); break; case CELL_CMD_FINISH: - printf("SPU %u: FINISH\n", init.id); + if (Debug) + printf("SPU %u: FINISH\n", init.id); /* wait for all outstanding DMAs to finish */ mfc_write_tag_mask(~0); mfc_read_tag_status_all(); @@ -369,7 +378,8 @@ main_loop(void) } - printf("SPU %u: Exit main loop\n", init.id); + if (Debug) + printf("SPU %u: Exit main loop\n", init.id); } @@ -388,7 +398,8 @@ main(unsigned long speid, unsigned long argp) DefaultTag = 1; - printf("SPU: main() speid=%lu\n", speid); + if (Debug) + printf("SPU: main() speid=%lu\n", speid); mfc_get(&init, /* dest */ (unsigned int) argp, /* src */ -- cgit v1.2.3 From c7a22c39792d6ff430af84502d578824081dbd3f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 9 Jan 2008 14:44:19 -0700 Subject: move cliprect bounds, do trivial rejection triangle clipping --- src/mesa/pipe/cell/spu/tri.c | 72 ++++++++++++++++++++++---------------------- 1 file changed, 36 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 58fca2e34a..5d4790fe6c 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -72,8 +72,6 @@ #define PIPE_MAX_SHADER_INPUTS 8 /* XXX temp */ -static int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; - #endif @@ -124,6 +122,8 @@ struct setup_stage { float oneoverarea; + int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; + #if 0 struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; #else @@ -276,8 +276,8 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) sp->quad.first->run(sp->quad.first, &setup->quad); #else /* Cell: "write" quad fragments to the tile by setting prim color */ - int ix = x - cliprect_minx; - int iy = y - cliprect_miny; + int ix = x - setup->cliprect_minx; + int iy = y - setup->cliprect_miny; float colors[4][4]; uint z; @@ -497,6 +497,20 @@ static boolean setup_sort_vertices( struct setup_stage *setup, } } + /* Check if triangle is completely outside the tile bounds */ + if (setup->vmin->data[0][1] > setup->cliprect_maxy) + return FALSE; + if (setup->vmax->data[0][1] < setup->cliprect_miny) + return FALSE; + if (setup->vmin->data[0][0] < setup->cliprect_minx && + setup->vmid->data[0][0] < setup->cliprect_minx && + setup->vmax->data[0][0] < setup->cliprect_minx) + return FALSE; + if (setup->vmin->data[0][0] > setup->cliprect_maxx && + setup->vmid->data[0][0] > setup->cliprect_maxx && + setup->vmax->data[0][0] > setup->cliprect_maxx) + return FALSE; + setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; @@ -733,18 +747,10 @@ static void subtriangle( struct setup_stage *setup, struct edge *eright, unsigned lines ) { -#if 0 - const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; -#else - const int minx = cliprect_minx; - const int maxx = cliprect_maxx; - const int miny = cliprect_miny; - const int maxy = cliprect_maxy; -#endif + const int minx = setup->cliprect_minx; + const int maxx = setup->cliprect_maxx; + const int miny = setup->cliprect_miny; + const int maxy = setup->cliprect_maxy; int y, start_y, finish_y; int sy = (int)eleft->sy; @@ -810,25 +816,19 @@ static void subtriangle( struct setup_stage *setup, /** * Do setup for triangle rasterization, then render the triangle. */ -static void setup_tri( -#if 0 - struct draw_stage *stage, -#endif - struct prim_header *prim ) +static void +setup_tri(struct setup_stage *setup, struct prim_header *prim) { -#if 0 - struct setup_stage *setup = setup_stage( stage ); -#else - struct setup_stage ss; - struct setup_stage *setup = &ss; - ss.color = prim->color; -#endif + setup->color = prim->color; /* XXX temporary */ /* _mesa_printf("%s\n", __FUNCTION__ ); */ - setup_sort_vertices( setup, prim ); + if (!setup_sort_vertices( setup, prim )) { + return; /* totally clipped */ + } + setup_tri_coefficients( setup ); setup_tri_edges( setup ); @@ -930,13 +930,13 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) void tri_draw(struct prim_header *tri, uint tx, uint ty) { + struct setup_stage setup; + /* set clipping bounds to tile bounds */ - cliprect_minx = tx * TILE_SIZE; - cliprect_miny = ty * TILE_SIZE; - cliprect_maxx = (tx + 1) * TILE_SIZE; - cliprect_maxy = (ty + 1) * TILE_SIZE; + setup.cliprect_minx = tx * TILE_SIZE; + setup.cliprect_miny = ty * TILE_SIZE; + setup.cliprect_maxx = (tx + 1) * TILE_SIZE; + setup.cliprect_maxy = (ty + 1) * TILE_SIZE; - setup_tri(tri); + setup_tri(&setup, tri); } - - -- cgit v1.2.3 From 18b0f345212a5950bddebe00edc3ec5f54781456 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 9 Jan 2008 16:12:25 -0700 Subject: Cell: start using DMA tags in a more sensible way, move waits() later when possible. --- src/mesa/pipe/cell/spu/main.c | 52 +++++++++++++++++++++++++++---------------- src/mesa/pipe/cell/spu/main.h | 12 ++++++++++ src/mesa/pipe/cell/spu/tri.c | 7 +++++- 3 files changed, 51 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 8e7f8be1eb..7b63e85ae2 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -45,7 +45,7 @@ helpful headers: /opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h */ -static boolean Debug = TRUE; +static boolean Debug = FALSE; volatile struct cell_init_info init; @@ -59,9 +59,10 @@ int DefaultTag; void -wait_on_mask(unsigned tag) +wait_on_mask(unsigned tagMask) { - mfc_write_tag_mask( tag ); + mfc_write_tag_mask( tagMask ); + /* wait for completion of _any_ DMAs specified by tagMask */ mfc_read_tag_status_any(); } @@ -125,7 +126,6 @@ clear_surface(const struct cell_command_clear_surface *clear) { uint num_tiles = fb.width_tiles * fb.height_tiles; uint i, j; - int tag = init.id; if (clear->surface == 0) { for (i = 0; i < TILE_SIZE; i++) @@ -147,13 +147,15 @@ clear_surface(const struct cell_command_clear_surface *clear) uint tx = i % fb.width_tiles; uint ty = i / fb.width_tiles; if (clear->surface == 0) - put_tile(&fb, tx, ty, (uint *) ctile, tag, 0); + put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); else - put_tile(&fb, tx, ty, (uint *) ztile, tag, 1); + put_tile(&fb, tx, ty, (uint *) ztile, TAG_SURFACE_CLEAR, 1); /* XXX we don't want this here, but it fixes bad tile results */ - wait_on_mask(1 << tag); } +#if 0 + wait_on_mask(1 << TAG_SURFACE_CLEAR); +#endif } @@ -198,7 +200,6 @@ static void render(const struct cell_command_render *render) { struct cell_prim_buffer prim_buffer ALIGN16_ATTRIB; - int tag = init.id /**DefaultTag**/; uint i, j, vertex_bytes; /* @@ -217,16 +218,18 @@ render(const struct cell_command_render *render) mfc_get(&prim_buffer, /* dest */ (unsigned int) render->vertex_data, /* src */ vertex_bytes, /* size */ - tag, + TAG_VERTEX_BUFFER, 0, /* tid */ 0 /* rid */); - wait_on_mask( 1 << tag ); /* XXX temporary */ + wait_on_mask(1 << TAG_VERTEX_BUFFER); /* find tiles which intersect the prim bounding box */ uint txmin, tymin, box_width_tiles, box_num_tiles; tile_bounding_box(render, &txmin, &tymin, &box_num_tiles, &box_width_tiles); + /* make sure any pending clears have completed */ + wait_on_mask(1 << TAG_SURFACE_CLEAR); /* loop over tiles */ for (i = init.id; i < box_num_tiles; i += init.num_spus) { @@ -236,12 +239,12 @@ render(const struct cell_command_render *render) assert(tx < fb.width_tiles); assert(ty < fb.height_tiles); - get_tile(&fb, tx, ty, (uint *) ctile, tag, 0); - wait_on_mask(1 << tag); /* XXX temporary */ - + /* Start fetching color/z tiles. We'll wait for completion when + * we need read/write to them later in triangle rasterization. + */ + get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - get_tile(&fb, tx, ty, (uint *) ztile, tag+1, 1); - wait_on_mask(1 << (tag+1)); /* XXX temporary */ + get_tile(&fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); } assert(render->prim_type == PIPE_PRIM_TRIANGLES); @@ -274,12 +277,23 @@ render(const struct cell_command_render *render) tri_draw(&prim, tx, ty); } - put_tile(&fb, tx, ty, (uint *) ctile, tag, 0); - wait_on_mask(1 << tag); /* XXX temp */ + /* in case nothing was drawn, wait now for completion */ + /* XXX temporary */ + wait_on_mask(1 << TAG_READ_TILE_COLOR); + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + wait_on_mask(1 << TAG_READ_TILE_Z); /* XXX temporary */ + } + + /* XXX IF we wrote anything into the tile... */ + + put_tile(&fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + put_tile(&fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); + } + wait_on_mask(1 << TAG_WRITE_TILE_COLOR); /* XXX temp */ if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - put_tile(&fb, tx, ty, (uint *) ztile, tag+1, 1); - wait_on_mask(1 << (tag+1)); /* XXX temporary */ + wait_on_mask(1 << TAG_WRITE_TILE_Z); /* XXX temporary */ } } } diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index 5e0516d0c1..656d28ea0e 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -55,6 +55,18 @@ extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; extern int DefaultTag; +/* DMA TAGS */ + +#define TAG_SURFACE_CLEAR 10 +#define TAG_VERTEX_BUFFER 11 +#define TAG_READ_TILE_COLOR 12 +#define TAG_READ_TILE_Z 13 +#define TAG_WRITE_TILE_COLOR 14 +#define TAG_WRITE_TILE_Z 15 + + + + void wait_on_mask(unsigned tag); diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 5d4790fe6c..b7dfd6ab8c 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -283,11 +283,13 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) eval_coeff(setup, 1, (float) x, (float) y, colors); - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { float zvals[4]; eval_z(setup, (float) x, (float) y, zvals); + wait_on_mask(1 << TAG_READ_TILE_Z); /* XXX temporary */ + + if (mask & MASK_TOP_LEFT) { z = (uint) (zvals[0] * 65535.0); if (z < ztile[iy][ix]) @@ -321,6 +323,9 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) } } + if (mask) + wait_on_mask(1 << TAG_READ_TILE_COLOR); + if (mask & MASK_TOP_LEFT) ctile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); if (mask & MASK_TOP_RIGHT) -- cgit v1.2.3 From d23869a88a1e9e41c9ebbd5f918ede16a8ee838f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 9 Jan 2008 17:49:43 -0700 Subject: Cell: initial implementation of tile status optimizations Tiles are marked as CLEAR, DEFINED or DIRTY to avoid making unnecessary get_tile() and put_tile() calls. --- src/mesa/pipe/cell/spu/main.c | 128 ++++++++++++++++++++++++++++++++++++------ src/mesa/pipe/cell/spu/main.h | 22 +++++++- src/mesa/pipe/cell/spu/tri.c | 47 +++++++++++----- 3 files changed, 164 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 7b63e85ae2..5d47ca85d3 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -54,7 +54,9 @@ struct framebuffer fb; uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -int DefaultTag; +ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; +ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + @@ -120,13 +122,91 @@ put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, } +void +clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value) +{ + uint i, j; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile[i][j] = value; + } + } +} + +void +clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value) +{ + uint i, j; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile[i][j] = value; + } + } +} + + +/** + * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled + * tiles back to the main framebuffer. + */ +static void +really_clear_tiles(uint surfaceIndex) +{ + const uint num_tiles = fb.width_tiles * fb.height_tiles; + uint i, j; + + if (surfaceIndex == 0) { + for (i = 0; i < TILE_SIZE; i++) + for (j = 0; j < TILE_SIZE; j++) + ctile[i][j] = fb.color_clear_value; + + for (i = init.id; i < num_tiles; i += init.num_spus) { + uint tx = i % fb.width_tiles; + uint ty = i / fb.width_tiles; + if (tile_status[ty][tx] == TILE_STATUS_CLEAR) { + put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); + } + } + } + else { + for (i = 0; i < TILE_SIZE; i++) + for (j = 0; j < TILE_SIZE; j++) + ztile[i][j] = fb.depth_clear_value; + + for (i = init.id; i < num_tiles; i += init.num_spus) { + uint tx = i % fb.width_tiles; + uint ty = i / fb.width_tiles; + if (tile_status_z[ty][tx] == TILE_STATUS_CLEAR) + put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 1); + } + } + +#if 0 + wait_on_mask(1 << TAG_SURFACE_CLEAR); +#endif +} + static void clear_surface(const struct cell_command_clear_surface *clear) { - uint num_tiles = fb.width_tiles * fb.height_tiles; + const uint num_tiles = fb.width_tiles * fb.height_tiles; uint i, j; +#define CLEAR_OPT 1 +#if CLEAR_OPT + /* set all tile's status to CLEAR */ + if (clear->surface == 0) { + memset(tile_status, TILE_STATUS_CLEAR, sizeof(tile_status)); + fb.color_clear_value = clear->value; + } + else { + memset(tile_status_z, TILE_STATUS_CLEAR, sizeof(tile_status_z)); + fb.depth_clear_value = clear->value; + } + return; +#endif + if (clear->surface == 0) { for (i = 0; i < TILE_SIZE; i++) for (j = 0; j < TILE_SIZE; j++) @@ -195,7 +275,6 @@ tile_bounding_box(const struct cell_command_render *render, } - static void render(const struct cell_command_render *render) { @@ -242,9 +321,14 @@ render(const struct cell_command_render *render) /* Start fetching color/z tiles. We'll wait for completion when * we need read/write to them later in triangle rasterization. */ - get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - get_tile(&fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); + if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(&fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); + } + } + + if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); } assert(render->prim_type == PIPE_PRIM_TRIANGLES); @@ -277,23 +361,22 @@ render(const struct cell_command_render *render) tri_draw(&prim, tx, ty); } - /* in case nothing was drawn, wait now for completion */ - /* XXX temporary */ - wait_on_mask(1 << TAG_READ_TILE_COLOR); - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - wait_on_mask(1 << TAG_READ_TILE_Z); /* XXX temporary */ + /* write color/z tiles back to main framebuffer, if dirtied */ + if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(&fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); + tile_status[ty][tx] = TILE_STATUS_DEFINED; } - - /* XXX IF we wrote anything into the tile... */ - - put_tile(&fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - put_tile(&fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); + if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(&fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); + tile_status_z[ty][tx] = TILE_STATUS_DEFINED; + } } - wait_on_mask(1 << TAG_WRITE_TILE_COLOR); /* XXX temp */ + /* XXX move these... */ + wait_on_mask(1 << TAG_WRITE_TILE_COLOR); if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - wait_on_mask(1 << TAG_WRITE_TILE_Z); /* XXX temporary */ + wait_on_mask(1 << TAG_WRITE_TILE_Z); } } } @@ -380,6 +463,7 @@ main_loop(void) case CELL_CMD_FINISH: if (Debug) printf("SPU %u: FINISH\n", init.id); + really_clear_tiles(0); /* wait for all outstanding DMAs to finish */ mfc_write_tag_mask(~0); mfc_read_tag_status_all(); @@ -398,6 +482,14 @@ main_loop(void) +static void +one_time_init(void) +{ + memset(tile_status, TILE_STATUS_DEFINED, sizeof(tile_status)); + memset(tile_status_z, TILE_STATUS_DEFINED, sizeof(tile_status_z)); +} + + /** * SPE entrypoint. * Note: example programs declare params as 'unsigned long long' but @@ -410,7 +502,7 @@ main(unsigned long speid, unsigned long argp) (void) speid; - DefaultTag = 1; + one_time_init(); if (Debug) printf("SPU: main() speid=%lu\n", speid); diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index 656d28ea0e..ee4248ead6 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -34,6 +34,10 @@ #include "pipe/cell/common.h" +#define MAX_WIDTH 1024 +#define MAX_HEIGHT 1024 + + extern volatile struct cell_init_info init; struct framebuffer { @@ -43,6 +47,9 @@ struct framebuffer { enum pipe_format depth_format; uint width, height; /**< size in pixels */ uint width_tiles, height_tiles; /**< width and height in tiles */ + + uint color_clear_value; + uint depth_clear_value; }; /* XXX Collect these globals in a struct: */ @@ -52,8 +59,6 @@ extern struct framebuffer fb; extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -extern int DefaultTag; - /* DMA TAGS */ @@ -66,6 +71,13 @@ extern int DefaultTag; +#define TILE_STATUS_CLEAR 1 +#define TILE_STATUS_DEFINED 2 /**< defined pixel data */ +#define TILE_STATUS_DIRTY 3 /**< modified, but not put back yet */ + +extern ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; +extern ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + void wait_on_mask(unsigned tag); @@ -78,5 +90,11 @@ void put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, int tag, int zBuf); +void +clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value); + +void +clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value); + #endif /* MAIN_H */ diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index b7dfd6ab8c..78cc7a591f 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -122,6 +122,8 @@ struct setup_stage { float oneoverarea; + uint tx, ty; + int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; #if 0 @@ -287,8 +289,15 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) float zvals[4]; eval_z(setup, (float) x, (float) y, zvals); - wait_on_mask(1 << TAG_READ_TILE_Z); /* XXX temporary */ - + if (tile_status_z[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { + /* now, _really_ clear the tile */ + clear_tile_z(ztile, fb.depth_clear_value); + } + else { + /* make sure we've got the tile from main mem */ + wait_on_mask(1 << TAG_READ_TILE_Z); + } + tile_status_z[setup->ty][setup->tx] = TILE_STATUS_DIRTY; if (mask & MASK_TOP_LEFT) { z = (uint) (zvals[0] * 65535.0); @@ -323,17 +332,26 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) } } - if (mask) - wait_on_mask(1 << TAG_READ_TILE_COLOR); - - if (mask & MASK_TOP_LEFT) - ctile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); - if (mask & MASK_TOP_RIGHT) - ctile[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); - if (mask & MASK_BOTTOM_LEFT) - ctile[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); - if (mask & MASK_BOTTOM_RIGHT) - ctile[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); + if (mask) { + if (tile_status[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { + /* now, _really_ clear the tile */ + clear_tile(ctile, fb.color_clear_value); + } + else { + /* make sure we've got the tile from main mem */ + wait_on_mask(1 << TAG_READ_TILE_COLOR); + } + tile_status[setup->ty][setup->tx] = TILE_STATUS_DIRTY; + + if (mask & MASK_TOP_LEFT) + ctile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); + if (mask & MASK_TOP_RIGHT) + ctile[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); + if (mask & MASK_BOTTOM_LEFT) + ctile[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); + if (mask & MASK_BOTTOM_RIGHT) + ctile[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); + } #endif } @@ -937,6 +955,9 @@ tri_draw(struct prim_header *tri, uint tx, uint ty) { struct setup_stage setup; + setup.tx = tx; + setup.ty = ty; + /* set clipping bounds to tile bounds */ setup.cliprect_minx = tx * TILE_SIZE; setup.cliprect_miny = ty * TILE_SIZE; -- cgit v1.2.3 From 51ea675745f4212c0bd859a940350faf466df102 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 10 Jan 2008 10:03:17 +0100 Subject: Add glapi/ path for inclusion of glapioffsets.h. Not sure why this is only needed now... --- src/mesa/x86/glapi_x86.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S index 1d04779d14..8b7204cdb7 100644 --- a/src/mesa/x86/glapi_x86.S +++ b/src/mesa/x86/glapi_x86.S @@ -27,7 +27,7 @@ */ #include "assyntax.h" -#include "glapioffsets.h" +#include "glapi/glapioffsets.h" #if defined(STDCALL_API) # if defined(USE_MGL_NAMESPACE) -- cgit v1.2.3 From ede7b00b59b37f078de0663918c0c84d572c27e8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 10 Jan 2008 10:03:47 +0100 Subject: softpipe: Simplify texture memory layout. --- src/mesa/pipe/softpipe/sp_surface.c | 27 ++- src/mesa/pipe/softpipe/sp_texture.c | 330 ++---------------------------------- src/mesa/pipe/softpipe/sp_texture.h | 19 +-- 3 files changed, 30 insertions(+), 346 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 5259fbfd20..ece30e36ec 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -46,20 +46,6 @@ softpipe_get_tex_surface(struct pipe_context *pipe, { struct softpipe_texture *spt = softpipe_texture(pt); struct pipe_surface *ps; - unsigned offset; /* in bytes */ - - offset = spt->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE) { - offset += spt->image_offset[level][face] * pt->cpp; - } - else if (pt->target == PIPE_TEXTURE_3D) { - offset += spt->image_offset[level][zslice] * pt->cpp; - } - else { - assert(face == 0); - assert(zslice == 0); - } ps = pipe->winsys->surface_alloc(pipe->winsys); if (ps) { @@ -69,8 +55,17 @@ softpipe_get_tex_surface(struct pipe_context *pipe, ps->cpp = pt->cpp; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->pitch = spt->pitch; - ps->offset = offset; + ps->pitch = ps->width; + ps->offset = spt->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { + ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * + (pt->compressed ? ps->height/4 : ps->height) * + ps->width * ps->cpp; + } else { + assert(face == 0); + assert(zslice == 0); + } } return ps; } diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index 44512e4281..e5e6bfe01b 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -41,10 +41,7 @@ #include "sp_texture.h" -/* At the moment, just make softpipe use the same layout for its - * textures as the i945. Softpipe needs some sort of texture layout, - * this one was handy. May be worthwhile to simplify this code a - * little. +/* Simple, maximally packed layout. */ static unsigned minify( unsigned d ) @@ -53,319 +50,35 @@ static unsigned minify( unsigned d ) } - -static void -sp_miptree_set_level_info(struct softpipe_texture *spt, - unsigned level, - unsigned nr_images, - unsigned x, unsigned y, unsigned w, unsigned h, - unsigned d) -{ - struct pipe_texture *pt = &spt->base; - - assert(level < PIPE_MAX_TEXTURE_LEVELS); - - pt->width[level] = w; - pt->height[level] = h; - pt->depth[level] = d; - - spt->nr_images[level] = nr_images; - spt->level_offset[level] = (x + y * spt->pitch) * pt->cpp; - - /* - DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - level, w, h, d, x, y, spt->level_offset[level]); - */ - - /* Not sure when this would happen, but anyway: - */ - if (spt->image_offset[level]) { - FREE( spt->image_offset[level] ); - spt->image_offset[level] = NULL; - } - - assert(nr_images); - assert(!spt->image_offset[level]); - - spt->image_offset[level] = (unsigned *) MALLOC( nr_images * sizeof(unsigned) ); - spt->image_offset[level][0] = 0; -} - - -static void -sp_miptree_set_image_offset(struct softpipe_texture *spt, - unsigned level, unsigned img, unsigned x, unsigned y) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < spt->nr_images[level]); - - spt->image_offset[level][img] = (x + y * spt->pitch); - - /* - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, spt->image_offset[level][img]); - */ -} - - static void -sp_miptree_layout_2d( struct softpipe_texture *spt ) +softpipe_texture_layout(struct softpipe_texture * spt) { struct pipe_texture *pt = &spt->base; - int align_h = 2, align_w = 4; unsigned level; - unsigned x = 0; - unsigned y = 0; unsigned width = pt->width[0]; unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; - spt->pitch = pt->width[0]; - /* XXX FIX THIS: - * we use alignment=64 bytes in sp_region_alloc(). If we change - * that, change this too. - */ - if (spt->pitch < 16) - spt->pitch = 16; - - /* May need to adjust pitch to accomodate the placement of - * the 2nd mipmap. This occurs when the alignment - * constraints of mipmap placement push the right edge of the - * 2nd mipmap out past the width of its parent. - */ - if (pt->first_level != pt->last_level) { - unsigned mip1_width = align(minify(pt->width[0]), align_w) - + minify(minify(pt->width[0])); - - if (mip1_width > pt->width[0]) - spt->pitch = mip1_width; - } - - /* Pitch must be a whole number of dwords, even though we - * express it in texels. - */ - spt->pitch = align(spt->pitch * pt->cpp, 4) / pt->cpp; - spt->total_height = 0; + spt->buffer_size = 0; for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { - unsigned img_height; - - sp_miptree_set_level_info(spt, level, 1, x, y, width, height, 1); - - if (pt->compressed) - img_height = MAX2(1, height/4); - else - img_height = align(height, align_h); + pt->width[level] = width; + pt->height[level] = height; + pt->depth[level] = depth; + spt->level_offset[level] = spt->buffer_size; - /* Because the images are packed better, the final offset - * might not be the maximal one: - */ - spt->total_height = MAX2(spt->total_height, y + img_height); - - /* Layout_below: step right after second mipmap. - */ - if (level == pt->first_level + 1) { - x += align(width, align_w); - } - else { - y += img_height; - } + spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * + ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * + width * pt->cpp; width = minify(width); height = minify(height); + depth = minify(depth); } } -static const int initial_offsets[6][2] = { - {0, 0}, - {0, 2}, - {1, 0}, - {1, 2}, - {1, 1}, - {1, 3} -}; - -static const int step_offsets[6][2] = { - {0, 2}, - {0, 2}, - {-1, 2}, - {-1, 2}, - {-1, 1}, - {-1, 1} -}; - - - -static boolean -softpipe_mipmap_tree_layout(struct pipe_context *pipe, struct softpipe_texture * spt) -{ - struct pipe_texture *pt = &spt->base; - unsigned level; - - switch (pt->target) { - case PIPE_TEXTURE_CUBE:{ - const unsigned dim = pt->width[0]; - unsigned face; - unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* Depending on the size of the largest images, pitch can be - * determined either by the old-style packing of cubemap faces, - * or the final row of 4x4, 2x2 and 1x1 faces below this. - */ - if (dim > 32) - spt->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; - else - spt->pitch = 14 * 8; - - spt->total_height = dim * 4 + 4; - - /* Set all the levels to effectively occupy the whole rectangular region. - */ - for (level = pt->first_level; level <= pt->last_level; level++) { - sp_miptree_set_level_info(spt, level, 6, - 0, 0, - lvlWidth, lvlHeight, 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - - for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * dim; - unsigned y = initial_offsets[face][1] * dim; - unsigned d = dim; - - if (dim == 4 && face >= 4) { - y = spt->total_height - 4; - x = (face - 4) * 8; - } - else if (dim < 4 && (face > 0 || pt->first_level > 0)) { - y = spt->total_height - 4; - x = face * 8; - } - - for (level = pt->first_level; level <= pt->last_level; level++) { - sp_miptree_set_image_offset(spt, level, face, x, y); - - d >>= 1; - - switch (d) { - case 4: - switch (face) { - case PIPE_TEX_FACE_POS_X: - case PIPE_TEX_FACE_NEG_X: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - case PIPE_TEX_FACE_POS_Y: - case PIPE_TEX_FACE_NEG_Y: - y += 12; - x -= 8; - break; - case PIPE_TEX_FACE_POS_Z: - case PIPE_TEX_FACE_NEG_Z: - y = spt->total_height - 4; - x = (face - 4) * 8; - break; - } - - case 2: - y = spt->total_height - 4; - x = 16 + face * 8; - break; - - case 1: - x += 48; - break; - - default: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - } - } - } - break; - } - case PIPE_TEXTURE_3D:{ - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - unsigned pack_x_pitch, pack_x_nr; - unsigned pack_y_pitch; - unsigned level; - - spt->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; - spt->total_height = 0; - - pack_y_pitch = MAX2(pt->height[0], 2); - pack_x_pitch = spt->pitch; - pack_x_nr = 1; - - for (level = pt->first_level; level <= pt->last_level; level++) { - unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; - int x = 0; - int y = 0; - unsigned q, j; - - sp_miptree_set_level_info(spt, level, nr_images, - 0, spt->total_height, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - sp_miptree_set_image_offset(spt, level, q, x, y); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - spt->total_height += y; - - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= spt->pitch); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - } - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - break; - } - - case PIPE_TEXTURE_1D: - case PIPE_TEXTURE_2D: -// case PIPE_TEXTURE_RECTANGLE: - sp_miptree_layout_2d(spt); - break; - default: - assert(0); - break; - } - - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - spt->pitch, - spt->total_height, pt->cpp, spt->pitch * spt->total_height * pt->cpp); - */ - - return TRUE; -} - void softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) { @@ -376,15 +89,13 @@ softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) memset(&spt->base + 1, 0, sizeof(struct softpipe_texture) - sizeof(struct pipe_texture)); - if (softpipe_mipmap_tree_layout(pipe, spt)) { - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); + softpipe_texture_layout(spt); - if (spt->buffer) { - pipe->winsys->buffer_data(pipe->winsys, spt->buffer, - spt->pitch * spt->base.cpp * - spt->total_height, NULL, - PIPE_BUFFER_USAGE_PIXEL); - } + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); + + if (spt->buffer) { + pipe->winsys->buffer_data(pipe->winsys, spt->buffer, spt->buffer_size, + NULL, PIPE_BUFFER_USAGE_PIXEL); } if (!spt->buffer) { @@ -408,7 +119,6 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) */ if (--(*pt)->refcount <= 0) { struct softpipe_texture *spt = softpipe_texture(*pt); - uint i; /* DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); @@ -416,10 +126,6 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) pipe->winsys->buffer_reference(pipe->winsys, &spt->buffer, NULL); - for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) - if (spt->image_offset[i]) - free(spt->image_offset[i]); - free(spt); } *pt = NULL; diff --git a/src/mesa/pipe/softpipe/sp_texture.h b/src/mesa/pipe/softpipe/sp_texture.h index 732064d986..e1a5db2791 100644 --- a/src/mesa/pipe/softpipe/sp_texture.h +++ b/src/mesa/pipe/softpipe/sp_texture.h @@ -10,29 +10,12 @@ struct softpipe_texture { struct pipe_texture base; - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; /* The data is held here: */ struct pipe_buffer_handle *buffer; + unsigned long buffer_size; }; -- cgit v1.2.3 From fd8b5ede950eb809165c783db2844130fb4c072d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 10:22:01 -0700 Subject: clean-up comments, code --- src/mesa/pipe/softpipe/sp_context.c | 2 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 16 +++--- src/mesa/pipe/softpipe/sp_prim_setup.h | 99 ++++++++++++++++++---------------- 3 files changed, 63 insertions(+), 54 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 0759092305..4f22539629 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -345,7 +345,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, if (GETENV( "SP_VBUF" ) != NULL) { softpipe->vbuf = sp_draw_vbuf_stage(softpipe->draw, &softpipe->pipe, - sp_vbuf_setup_draw); + sp_vbuf_render); draw_set_rasterize_stage(softpipe->draw, softpipe->vbuf); } diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index de45529bf9..722cadc5c0 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -1286,18 +1286,20 @@ static void calc_det( struct prim_header *header ) -/* Test harness - feed vertex buffer back into prim pipeline. +/** + * Render buffer of points/lines/triangles. + * Called by vbuf code when the vertex or index buffer is filled. * * The big issue at this point is that reset_stipple doesn't make it * through the interface. Probably need to split primitives at reset * stipple, perhaps using the ~0 index marker. */ -void sp_vbuf_setup_draw( struct pipe_context *pipe, - unsigned primitive, - const ushort *elements, - unsigned nr_elements, - const void *vertex_buffer, - unsigned nr_vertices ) +void sp_vbuf_render( struct pipe_context *pipe, + unsigned primitive, + const ushort *elements, + unsigned nr_elements, + const void *vertex_buffer, + unsigned nr_vertices ) { struct softpipe_context *softpipe = softpipe_context( pipe ); struct setup_stage *setup = setup_stage( softpipe->setup ); diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h index 598394f73e..c16dc634b0 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.h +++ b/src/mesa/pipe/softpipe/sp_prim_setup.h @@ -1,33 +1,49 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5 - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ #ifndef SP_PRIM_SETUP_H #define SP_PRIM_SETUP_H -/* Vertices are just an array of floats, with all the attributes +/** + * vbuf is a special stage to gather the stream of triangles, lines, points + * together and reconstruct vertex buffers for hardware upload. + * + * First attempt, work in progress. + * + * TODO: + * - separate out vertex buffer building and primitive emit, ie >1 draw per vb. + * - tell vbuf stage how to build hw vertices directly + * - pass vbuf stage a buffer pointer for direct emit to agp/vram. + * + * + * + * Vertices are just an array of floats, with all the attributes * packed. We currently assume a layout like: * * attr[0][0..3] - window position @@ -37,23 +53,11 @@ * all the enabled attributes run contiguously. */ + struct draw_stage; struct softpipe_context; -extern struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ); - - -/* A special stage to gather the stream of triangles, lines, points - * together and reconstruct vertex buffers for hardware upload. - * - * First attempt, work in progress. - * - * TODO: - * - separate out vertex buffer building and primitive emit, ie >1 draw per vb. - * - tell vbuf stage how to build hw vertices directly - * - pass vbuf stage a buffer pointer for direct emit to agp/vram. - */ typedef void (*vbuf_draw_func)( struct pipe_context *pipe, unsigned prim, const ushort *elements, @@ -62,20 +66,23 @@ typedef void (*vbuf_draw_func)( struct pipe_context *pipe, unsigned nr_vertices ); -extern struct draw_stage *sp_draw_vbuf_stage( struct draw_context *draw_context, - struct pipe_context *pipe, - vbuf_draw_func draw ); +extern struct draw_stage * +sp_draw_render_stage( struct softpipe_context *softpipe ); +extern struct draw_stage * +sp_draw_vbuf_stage( struct draw_context *draw_context, + struct pipe_context *pipe, + vbuf_draw_func draw ); -/* Test harness - */ -void sp_vbuf_setup_draw( struct pipe_context *pipe, - unsigned prim, - const ushort *elements, - unsigned nr_elements, - const void *vertex_buffer, - unsigned nr_vertices ); + +extern void +sp_vbuf_render( struct pipe_context *pipe, + unsigned prim, + const ushort *elements, + unsigned nr_elements, + const void *vertex_buffer, + unsigned nr_vertices ); -- cgit v1.2.3 From 3d9c84c1b88d30115682763d2762a3c2bf4e5ff3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 10:38:28 -0700 Subject: comment about vertex data in emit_vertex() --- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index 2cfdeb5809..e9bd4dd5e4 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -120,6 +120,12 @@ static void emit_vertex( struct vbuf_stage *vbuf, vertex->vertex_id = vbuf->nr_vertices++; //vbuf->emit_vertex( vbuf->vertex_ptr, vertex ); + + /* Note: for softpipe, the vertex includes the vertex header info + * such as clip flags and clip coords. In the future when vbuf is + * always used, we could just copy the vertex attributes/data here. + * The sp_prim_setup.c code doesn't use any of the vertex header info. + */ memcpy(vbuf->vertex_ptr, vertex, vbuf->vertex_size); vbuf->vertex_ptr += vbuf->vertex_size; -- cgit v1.2.3 From e6b33b6f35ab2f7b240ac902cc748d8e1a2fd4ef Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 08:32:27 -0700 Subject: bump CELL_MAX_VERTS to 240 --- src/mesa/pipe/cell/common.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 4c770f5c32..d7de0beece 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -103,7 +103,7 @@ struct cell_init_info /** Temporary */ -#define CELL_MAX_VERTS 48 +#define CELL_MAX_VERTS 240 #define CELL_MAX_ATTRIBS 2 struct cell_prim_buffer { -- cgit v1.2.3 From d07b86dedfd87ed7c301fa81242314c891031693 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 09:13:10 -0700 Subject: Cell: clean-up cell_spu_exit() code --- src/mesa/pipe/cell/ppu/cell_context.c | 2 -- src/mesa/pipe/cell/ppu/cell_spu.c | 25 ++++++++----------------- src/mesa/pipe/cell/ppu/cell_spu.h | 3 --- 3 files changed, 8 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 4fcf804d82..57178c835f 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -148,7 +148,6 @@ cell_destroy_context( struct pipe_context *pipe ) struct cell_context *cell = cell_context(pipe); cell_spu_exit(cell); - wait_spus(cell->num_spus); free(cell); } @@ -255,7 +254,6 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) #if 0 test_spus(cell); - wait_spus(); #endif return &cell->pipe; diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index a7dbf24dd8..44ad3f22b3 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -194,32 +194,23 @@ test_spus(struct cell_context *cell) } -/** - * Wait for all SPUs to exit/return. - */ -void -wait_spus(uint num_spus) -{ - uint i; - void *value; - - for (i = 0; i < num_spus; i++) { - pthread_join(cell_global.spe_threads[i], &value); - } -} - - /** * Tell all the SPUs to stop/exit. */ void cell_spu_exit(struct cell_context *cell) { - unsigned i; + uint i; for (i = 0; i < cell->num_spus; i++) { send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_EXIT); } - wait_spus(cell->num_spus); + /* wait for threads to exit */ + for (i = 0; i < cell->num_spus; i++) { + void *value; + pthread_join(cell_global.spe_threads[i], &value); + cell_global.spe_threads[i] = 0; + cell_global.spe_contexts[i] = 0; + } } diff --git a/src/mesa/pipe/cell/ppu/cell_spu.h b/src/mesa/pipe/cell/ppu/cell_spu.h index 612cb45c59..d11001af62 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.h +++ b/src/mesa/pipe/cell/ppu/cell_spu.h @@ -82,9 +82,6 @@ void test_spus(struct cell_context *cell); -void -wait_spus(uint num_spus); - void cell_spu_exit(struct cell_context *cell); -- cgit v1.2.3 From cd89fe2d390d2e8cdf7039c8405fd34d28419b5e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 09:22:22 -0700 Subject: Cell: s/free/align_free/ --- src/mesa/pipe/cell/ppu/cell_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 57178c835f..fee72eabc2 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -149,7 +149,7 @@ cell_destroy_context( struct pipe_context *pipe ) cell_spu_exit(cell); - free(cell); + align_free(cell); } -- cgit v1.2.3 From 2da5afbd3ffd50409fc729e166fe5133c7a7e7a8 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 15:13:14 -0700 Subject: Cell: call draw_compute_vertex_size() --- src/mesa/pipe/cell/ppu/cell_state_derived.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c index ee7c0204b8..ec3a8f3a81 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -159,7 +159,8 @@ static void calculate_vertex_layout( struct cell_context *cell ) if (1/*vinfo->attr_mask != cell->attr_mask*/) { /*cell->attr_mask = vinfo->attr_mask;*/ - draw_set_vertex_info( cell->draw, vinfo); + draw_compute_vertex_size(vinfo); + draw_set_vertex_info(cell->draw, vinfo); #if 0 draw_set_twoside_attributes(cell->draw, -- cgit v1.2.3 From 763e30748e99064fafeb0d9b34de8d732732689c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 15:14:58 -0700 Subject: Cell: initial implementation of vbuf code. The draw module's vbuf stage builds buffers of post-transformed vertices and issues draw-elements calls to render them. We'll pass the vertex and index buffers to the SPUs... --- src/mesa/pipe/cell/ppu/Makefile | 1 + src/mesa/pipe/cell/ppu/cell_context.c | 8 +- src/mesa/pipe/cell/ppu/cell_context.h | 6 ++ src/mesa/pipe/cell/ppu/cell_vbuf.c | 136 ++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_vbuf.h | 38 ++++++++++ 5 files changed, 188 insertions(+), 1 deletion(-) create mode 100644 src/mesa/pipe/cell/ppu/cell_vbuf.c create mode 100644 src/mesa/pipe/cell/ppu/cell_vbuf.h (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index 2758e1b77f..9bf89702d5 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -30,6 +30,7 @@ SOURCES = \ cell_state_vertex.c \ cell_spu.c \ cell_surface.c \ + cell_vbuf.c \ cell_winsys.c diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index fee72eabc2..091e593e1f 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -46,6 +46,7 @@ #include "cell_state.h" #include "cell_surface.h" #include "cell_spu.h" +#include "cell_vbuf.h" @@ -236,9 +237,14 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->draw = draw_create(); +#define VBUF 0 +#if VBUF + cell_init_vbuf(cell); + draw_set_rasterize_stage(cell->draw, cell->vbuf); +#else cell->render_stage = cell_draw_render_stage(cell); draw_set_rasterize_stage(cell->draw, cell->render_stage); - +#endif cell->prim_buffer.xmin = 1e100; cell->prim_buffer.ymin = 1e100; diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index be985820d1..4ff39a53c6 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -33,10 +33,13 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/draw/draw_vertex.h" +#include "pipe/draw/draw_vbuf.h" #include "cell_winsys.h" #include "pipe/cell/common.h" +struct cell_vbuf_render; + struct cell_vertex_shader_state { struct pipe_shader_state shader; @@ -81,6 +84,9 @@ struct cell_context /** The primitive drawing context */ struct draw_context *draw; struct draw_stage *render_stage; + + /** For post-transformed vertex buffering: */ + struct cell_vbuf_render *vbuf_render; struct draw_stage *vbuf; struct vertex_info vertex_info; diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c new file mode 100644 index 0000000000..1a43d237d9 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -0,0 +1,136 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Authors + * Brian Paul + */ + + +#include "cell_context.h" +#include "cell_vbuf.h" +#include "pipe/draw/draw_vbuf.h" + + +/** + * Subclass of vbuf_render because we need a cell_context pointer in + * a few places. + */ +struct cell_vbuf_render +{ + struct vbuf_render base; + struct cell_context *cell; + uint prim; +}; + + +/** cast wrapper */ +static struct cell_vbuf_render * +cell_vbuf_render(struct vbuf_render *vbr) +{ + return (struct cell_vbuf_render *) vbr; +} + + + +static const struct vertex_info * +cell_vbuf_get_vertex_info(struct vbuf_render *vbr) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + return &cvbr->cell->vertex_info; +} + + +static void * +cell_vbuf_allocate_vertices(struct vbuf_render *vbr, + ushort vertex_size, ushort nr_vertices) +{ + printf("Alloc verts %u * %u\n", vertex_size, nr_vertices); + return align_malloc(vertex_size * nr_vertices, 16); +} + + +static void +cell_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + cvbr->prim = prim; + printf("cell_set_prim %u\n", prim); +} + + +static void +cell_vbuf_draw(struct vbuf_render *vbr, + const ushort *indices, unsigned nr_indices) +{ + printf("cell_vbuf_draw nr_indices = %u\n", nr_indices); +} + + +static void +cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, + unsigned vertex_size, unsigned vertices_used) +{ + printf("Free verts %u * %u\n", vertex_size, vertices_used); + align_free(vertices); +} + + +static void +cell_vbuf_destroy(struct vbuf_render *vbr) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + cvbr->cell->vbuf_render = NULL; + FREE(cvbr); +} + + +/** + * Initialize the post-transform vertex buffer information for the given + * context. + */ +void +cell_init_vbuf(struct cell_context *cell) +{ + assert(cell->draw); + + cell->vbuf_render = CALLOC_STRUCT(cell_vbuf_render); + + cell->vbuf_render->base.max_indices = 100; + cell->vbuf_render->base.max_vertex_buffer_bytes = 16 * 1024; + + cell->vbuf_render->base.get_vertex_info = cell_vbuf_get_vertex_info; + cell->vbuf_render->base.allocate_vertices = cell_vbuf_allocate_vertices; + cell->vbuf_render->base.set_primitive = cell_vbuf_set_primitive; + cell->vbuf_render->base.draw = cell_vbuf_draw; + cell->vbuf_render->base.release_vertices = cell_vbuf_release_vertices; + cell->vbuf_render->base.destroy = cell_vbuf_destroy; + + cell->vbuf_render->cell = cell; + + cell->vbuf = draw_vbuf_stage(cell->draw, &cell->vbuf_render->base); +} diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.h b/src/mesa/pipe/cell/ppu/cell_vbuf.h new file mode 100644 index 0000000000..d265cbf770 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_VBUF_H +#define CELL_VBUF_H + + +struct cell_context; + +extern void +cell_init_vbuf(struct cell_context *cell); + + +#endif /* CELL_VBUF_H */ -- cgit v1.2.3 From cac8892ddb24ddb92b6f367689712925ee6c2d86 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 17:01:11 -0700 Subject: Additional parameters to vbuf_render->draw() Pass complete information about vertex/index buffer location, size, etc. --- src/mesa/pipe/draw/draw_vbuf.c | 14 ++++++++++---- src/mesa/pipe/draw/draw_vbuf.h | 6 +++++- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 6 +++++- 3 files changed, 20 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index d010aaba07..3f6606de86 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -291,8 +291,14 @@ vbuf_flush_indices( struct draw_stage *stage ) assert(0); } - vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); - + vbuf->render->draw( vbuf->render, + vbuf->prim, + vbuf->indices, + vbuf->nr_indices, + vbuf->vertices, + vbuf->nr_vertices, + vbuf->vertex_size ); + vbuf->nr_indices = 0; } @@ -373,7 +379,7 @@ static void vbuf_destroy( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); - FREE( vbuf->indices ); + align_free( vbuf->indices ); FREE( stage ); } @@ -399,7 +405,7 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, assert(render->max_indices < UNDEFINED_VERTEX_ID); vbuf->max_indices = render->max_indices; - vbuf->indices = MALLOC( vbuf->max_indices ); + vbuf->indices = align_malloc( vbuf->max_indices, 16 ); vbuf->vertices = NULL; vbuf->vertex_ptr = vbuf->vertices; diff --git a/src/mesa/pipe/draw/draw_vbuf.h b/src/mesa/pipe/draw/draw_vbuf.h index 43aa740f64..be4c4ab77b 100644 --- a/src/mesa/pipe/draw/draw_vbuf.h +++ b/src/mesa/pipe/draw/draw_vbuf.h @@ -82,8 +82,12 @@ struct vbuf_render { * DrawElements, note indices are ushort: */ void (*draw)( struct vbuf_render *, + uint prim, const ushort *indices, - unsigned nr_indices ); + uint nr_indices, + const void *vertices, + uint nr_vertices, + uint vertex_bytes); /** * Called when vbuf is done with this set of vertices: diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 571ad40595..bdcc027ed7 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -136,8 +136,12 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render, static void i915_vbuf_render_draw( struct vbuf_render *render, + uint prim, const ushort *indices, - unsigned nr_indices ) + uint nr_indices, + const void *vertices, + uint nr_vertices, + uint vertex_bytes) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; -- cgit v1.2.3 From 152f1d84d4d942d9e912b116c5fc67ba96ed9859 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 17:01:52 -0700 Subject: Cell: checkpoint: draw_vbuf code in place and works, but not enabled by default yet. --- src/mesa/pipe/cell/common.h | 15 ++++ src/mesa/pipe/cell/ppu/cell_vbuf.c | 59 +++++++++++++-- src/mesa/pipe/cell/spu/main.c | 146 +++++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/main.h | 1 + 4 files changed, 213 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index d7de0beece..dc2db299c4 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -52,6 +52,7 @@ #define CELL_CMD_CLEAR_SURFACE 3 #define CELL_CMD_FINISH 4 #define CELL_CMD_RENDER 5 +#define CELL_CMD_RENDER_VBUF 6 /** @@ -84,12 +85,26 @@ struct cell_command_render } ALIGN16_ATTRIB; +#define CELL_MAX_VBUF_SIZE (16 * 1024) +#define CELL_MAX_VBUF_INDEXES 1024 +struct cell_command_render_vbuf +{ + uint prim_type; + uint num_verts, num_attribs; + uint num_indexes; + const void *vertex_data; + const ushort *index_data; + float xmin, ymin, xmax, ymax; +} ALIGN16_ATTRIB; + + /** XXX unions don't seem to work */ struct cell_command { struct cell_command_framebuffer fb; struct cell_command_clear_surface clear; struct cell_command_render render; + struct cell_command_render_vbuf render_vbuf; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 1a43d237d9..cf468204bb 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -32,6 +32,7 @@ #include "cell_context.h" +#include "cell_spu.h" #include "cell_vbuf.h" #include "pipe/draw/draw_vbuf.h" @@ -69,7 +70,7 @@ static void * cell_vbuf_allocate_vertices(struct vbuf_render *vbr, ushort vertex_size, ushort nr_vertices) { - printf("Alloc verts %u * %u\n", vertex_size, nr_vertices); + /*printf("Alloc verts %u * %u\n", vertex_size, nr_vertices);*/ return align_malloc(vertex_size * nr_vertices, 16); } @@ -79,23 +80,63 @@ cell_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) { struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); cvbr->prim = prim; - printf("cell_set_prim %u\n", prim); + /*printf("cell_set_prim %u\n", prim);*/ } +#if 0 static void cell_vbuf_draw(struct vbuf_render *vbr, - const ushort *indices, unsigned nr_indices) + uint prim, + const ushort *indices, + uint nr_indices, + const void *vertices, + uint nr_vertices, + uint vertex_size) { - printf("cell_vbuf_draw nr_indices = %u\n", nr_indices); + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + struct cell_context *cell = cvbr->cell; + uint i; + + /*printf("cell_vbuf_draw nr_indices = %u\n", nr_indices);*/ + + if (prim != PIPE_PRIM_TRIANGLES) + return; /* only render tris for now */ + + for (i = 0; i < cell->num_spus; i++) { + struct cell_command_render_vbuf *render = &cell_global.command[i].render_vbuf; + render->prim_type = prim; + render->num_verts = nr_vertices; + render->num_attribs = CELL_MAX_ATTRIBS; /* XXX fix */ + render->vertex_data = vertices; + render->index_data = indices; + render->num_indexes = nr_indices; +#if 0 + render->xmin = cell->prim_buffer.xmin; + render->ymin = cell->prim_buffer.ymin; + render->xmax = cell->prim_buffer.xmax; + render->ymax = cell->prim_buffer.ymax; +#else + render->xmin = -100; + render->ymin = -100; + render->xmax = 100; + render->ymax = 100; +#endif + + ASSERT_ALIGN16(render->vertex_data); + ASSERT_ALIGN16(render->index_data); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER_VBUF); + } + + cell_flush(&cell->pipe, 0x0); } - +#endif static void cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, unsigned vertex_size, unsigned vertices_used) { - printf("Free verts %u * %u\n", vertex_size, vertices_used); + /*printf("Free verts %u * %u\n", vertex_size, vertices_used);*/ align_free(vertices); } @@ -120,13 +161,15 @@ cell_init_vbuf(struct cell_context *cell) cell->vbuf_render = CALLOC_STRUCT(cell_vbuf_render); - cell->vbuf_render->base.max_indices = 100; - cell->vbuf_render->base.max_vertex_buffer_bytes = 16 * 1024; + cell->vbuf_render->base.max_indices = CELL_MAX_VBUF_INDEXES; + cell->vbuf_render->base.max_vertex_buffer_bytes = CELL_MAX_VBUF_SIZE; cell->vbuf_render->base.get_vertex_info = cell_vbuf_get_vertex_info; cell->vbuf_render->base.allocate_vertices = cell_vbuf_allocate_vertices; cell->vbuf_render->base.set_primitive = cell_vbuf_set_primitive; +#if 0 cell->vbuf_render->base.draw = cell_vbuf_draw; +#endif cell->vbuf_render->base.release_vertices = cell_vbuf_release_vertices; cell->vbuf_render->base.destroy = cell_vbuf_destroy; diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 5d47ca85d3..6dbd93f530 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -382,6 +382,143 @@ render(const struct cell_command_render *render) } +static void +render_vbuf(const struct cell_command_render_vbuf *render) +{ + /* we'll DMA into these buffers */ + ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; + ushort indexes[CELL_MAX_VBUF_INDEXES] ALIGN16_ATTRIB; + uint i, j, vertex_bytes, index_bytes; + + ASSERT_ALIGN16(render->vertex_data); + ASSERT_ALIGN16(render->index_data); + + /* how much vertex data */ + vertex_bytes = render->num_verts * render->num_attribs * 4 * sizeof(float); + index_bytes = render->num_indexes * sizeof(ushort); + if (index_bytes < 8) + index_bytes = 8; + else + index_bytes = (index_bytes + 15) & ~0xf; /* multiple of 16 */ + + /* + printf("VBUF: indices at %p, vertices at %p vertex_bytes %u ind_bytes %u\n", + render->index_data, render->vertex_data, vertex_bytes, index_bytes); + */ + + /* get vertex data from main memory */ + mfc_get(vertex_data, /* dest */ + (unsigned int) render->vertex_data, /* src */ + vertex_bytes, /* size */ + TAG_VERTEX_BUFFER, + 0, /* tid */ + 0 /* rid */); + + /* get index data from main memory */ + mfc_get(indexes, /* dest */ + (unsigned int) render->index_data, /* src */ + index_bytes, + TAG_INDEX_BUFFER, + 0, /* tid */ + 0 /* rid */); + + wait_on_mask(1 << TAG_VERTEX_BUFFER); + wait_on_mask(1 << TAG_INDEX_BUFFER); + + /* find tiles which intersect the prim bounding box */ + uint txmin, tymin, box_width_tiles, box_num_tiles; +#if 0 + tile_bounding_box(render, &txmin, &tymin, + &box_num_tiles, &box_width_tiles); +#else + txmin = 0; + tymin = 0; + box_num_tiles = fb.width_tiles * fb.height_tiles; + box_width_tiles = fb.width_tiles; +#endif + + /* make sure any pending clears have completed */ + wait_on_mask(1 << TAG_SURFACE_CLEAR); + + /* loop over tiles */ + for (i = init.id; i < box_num_tiles; i += init.num_spus) { + const uint tx = txmin + i % box_width_tiles; + const uint ty = tymin + i / box_width_tiles; + + assert(tx < fb.width_tiles); + assert(ty < fb.height_tiles); + + /* Start fetching color/z tiles. We'll wait for completion when + * we need read/write to them later in triangle rasterization. + */ + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(&fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); + } + } + + if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); + } + + assert(render->prim_type == PIPE_PRIM_TRIANGLES); + + /* loop over tris */ + for (j = 0; j < render->num_indexes; j += 3) { + struct prim_header prim; + const float *vbuf = (const float *) vertex_data; + const float *v0, *v1, *v2; + + v0 = vbuf + indexes[j] * render->num_attribs * 4; + v1 = vbuf + indexes[j+1] * render->num_attribs * 4; + v2 = vbuf + indexes[j+2] * render->num_attribs * 4; + + /* + printf(" %u: Triangle %g,%g %g,%g %g,%g\n", + init.id, + prim_buffer.vertex[j*3+0][0][0], + prim_buffer.vertex[j*3+0][0][1], + prim_buffer.vertex[j*3+1][0][0], + prim_buffer.vertex[j*3+1][0][1], + prim_buffer.vertex[j*3+2][0][0], + prim_buffer.vertex[j*3+2][0][1]); + */ + + /* pos */ + COPY_4V(prim.v[0].data[0], v0); + COPY_4V(prim.v[1].data[0], v1); + COPY_4V(prim.v[2].data[0], v2); + + /* color */ + COPY_4V(prim.v[0].data[1], v0 + 4); + COPY_4V(prim.v[1].data[1], v1 + 4); + COPY_4V(prim.v[2].data[1], v2 + 4); + + tri_draw(&prim, tx, ty); + } + + /* write color/z tiles back to main framebuffer, if dirtied */ + if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(&fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); + tile_status[ty][tx] = TILE_STATUS_DEFINED; + } + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(&fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); + tile_status_z[ty][tx] = TILE_STATUS_DEFINED; + } + } + + /* XXX move these... */ + wait_on_mask(1 << TAG_WRITE_TILE_COLOR); + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + wait_on_mask(1 << TAG_WRITE_TILE_Z); + } + } +} + + + /** * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. */ @@ -459,6 +596,15 @@ main_loop(void) init.id, cmd.render.num_verts, cmd.render.prim_type); render(&cmd.render); break; + case CELL_CMD_RENDER_VBUF: + if (Debug) + printf("SPU %u: RENDER_VBUF prim %u, indices: %u, nr_vert: %u\n", + init.id, + cmd.render_vbuf.prim_type, + cmd.render_vbuf.num_verts, + cmd.render_vbuf.num_indexes); + render_vbuf(&cmd.render_vbuf); + break; case CELL_CMD_FINISH: if (Debug) diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index ee4248ead6..fce113b77d 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -64,6 +64,7 @@ extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; #define TAG_SURFACE_CLEAR 10 #define TAG_VERTEX_BUFFER 11 +#define TAG_INDEX_BUFFER 16 #define TAG_READ_TILE_COLOR 12 #define TAG_READ_TILE_Z 13 #define TAG_WRITE_TILE_COLOR 14 -- cgit v1.2.3 From b3247225423213c156ce4f428d1d246758a96d50 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 17:03:21 -0700 Subject: Cell: enable vbuf path by default --- src/mesa/pipe/cell/ppu/cell_context.c | 2 +- src/mesa/pipe/cell/ppu/cell_vbuf.c | 5 +---- 2 files changed, 2 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 091e593e1f..52c97c939b 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -237,7 +237,7 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->draw = draw_create(); -#define VBUF 0 +#define VBUF 1 #if VBUF cell_init_vbuf(cell); draw_set_rasterize_stage(cell->draw, cell->vbuf); diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index cf468204bb..059a098914 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -84,7 +84,6 @@ cell_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) } -#if 0 static void cell_vbuf_draw(struct vbuf_render *vbr, uint prim, @@ -130,7 +129,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, cell_flush(&cell->pipe, 0x0); } -#endif + static void cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, @@ -167,9 +166,7 @@ cell_init_vbuf(struct cell_context *cell) cell->vbuf_render->base.get_vertex_info = cell_vbuf_get_vertex_info; cell->vbuf_render->base.allocate_vertices = cell_vbuf_allocate_vertices; cell->vbuf_render->base.set_primitive = cell_vbuf_set_primitive; -#if 0 cell->vbuf_render->base.draw = cell_vbuf_draw; -#endif cell->vbuf_render->base.release_vertices = cell_vbuf_release_vertices; cell->vbuf_render->base.destroy = cell_vbuf_destroy; -- cgit v1.2.3 From 6c11485405700865895b7c5f14e08bc5bede2a35 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 17:14:06 -0700 Subject: Cell: use new ASSERT macro instead of standard assert The later doesn't seem to work properly in SPU code. --- src/mesa/pipe/cell/spu/main.c | 23 +++++++++++------------ src/mesa/pipe/cell/spu/main.h | 10 ++++++++++ src/mesa/pipe/cell/spu/tri.c | 6 +++--- 3 files changed, 24 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 6dbd93f530..5ad15bc639 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -30,7 +30,6 @@ #include -#include #include #include @@ -80,8 +79,8 @@ get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, src += offset * bytesPerTile; - assert(tx < fb->width_tiles); - assert(ty < fb->height_tiles); + ASSERT(tx < fb->width_tiles); + ASSERT(ty < fb->height_tiles); ASSERT_ALIGN16(tile); /* printf("get_tile: dest: %p src: 0x%x size: %d\n", @@ -106,8 +105,8 @@ put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, dst += offset * bytesPerTile; - assert(tx < fb->width_tiles); - assert(ty < fb->height_tiles); + ASSERT(tx < fb->width_tiles); + ASSERT(ty < fb->height_tiles); ASSERT_ALIGN16(tile); /* printf("put_tile: src: %p dst: 0x%x size: %d\n", @@ -315,8 +314,8 @@ render(const struct cell_command_render *render) const uint tx = txmin + i % box_width_tiles; const uint ty = tymin + i / box_width_tiles; - assert(tx < fb.width_tiles); - assert(ty < fb.height_tiles); + ASSERT(tx < fb.width_tiles); + ASSERT(ty < fb.height_tiles); /* Start fetching color/z tiles. We'll wait for completion when * we need read/write to them later in triangle rasterization. @@ -331,7 +330,7 @@ render(const struct cell_command_render *render) get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); } - assert(render->prim_type == PIPE_PRIM_TRIANGLES); + ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); /* loop over tris */ for (j = 0; j < render->num_verts; j += 3) { @@ -445,8 +444,8 @@ render_vbuf(const struct cell_command_render_vbuf *render) const uint tx = txmin + i % box_width_tiles; const uint ty = tymin + i / box_width_tiles; - assert(tx < fb.width_tiles); - assert(ty < fb.height_tiles); + ASSERT(tx < fb.width_tiles); + ASSERT(ty < fb.height_tiles); /* Start fetching color/z tiles. We'll wait for completion when * we need read/write to them later in triangle rasterization. @@ -461,7 +460,7 @@ render_vbuf(const struct cell_command_render_vbuf *render) get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); } - assert(render->prim_type == PIPE_PRIM_TRIANGLES); + ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); /* loop over tris */ for (j = 0; j < render->num_indexes; j += 3) { @@ -531,7 +530,7 @@ main_loop(void) if (Debug) printf("SPU %u: Enter main loop\n", init.id); - assert((sizeof(struct cell_command) & 0xf) == 0); + ASSERT((sizeof(struct cell_command) & 0xf) == 0); ASSERT_ALIGN16(&cmd); while (!exitFlag) { diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index fce113b77d..c539385a07 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -98,4 +98,14 @@ void clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value); + +/** The standard assert macro doesn't seem to work on SPUs */ +#define ASSERT(x) \ + if (!(x)) { \ + fprintf(stderr, "SPU %d: %s:%d: %s(): assertion %s failed.\n", \ + init.id, __FILE__, __LINE__, __FUNCTION__, #x); \ + exit(1); \ + } + + #endif /* MAIN_H */ diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 78cc7a591f..b40f35bc69 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -258,7 +258,7 @@ pack_color(const float color[4]) case PIPE_FORMAT_B8G8R8A8_UNORM: return (b << 24) | (g << 16) | (r << 8) | a; default: - assert(0); + ASSERT(0); return 0; } } @@ -613,7 +613,7 @@ static void tri_linear_coeff( struct setup_stage *setup, float a = setup->ebot.dy * majda - botda * setup->emaj.dy; float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - assert(slot < PIPE_MAX_SHADER_INPUTS); + ASSERT(slot < PIPE_MAX_SHADER_INPUTS); setup->coef[slot].dadx[i] = a * setup->oneoverarea; setup->coef[slot].dady[i] = b * setup->oneoverarea; @@ -777,7 +777,7 @@ static void subtriangle( struct setup_stage *setup, int y, start_y, finish_y; int sy = (int)eleft->sy; - assert((int)eleft->sy == (int) eright->sy); + ASSERT((int)eleft->sy == (int) eright->sy); /* clip top/bottom */ start_y = sy; -- cgit v1.2.3 From 02f6f9f8d47fc36c8edf4661c4e78c9c1a1941fc Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 17:30:51 -0700 Subject: Cell: move tile-related code into new tile.[ch] files. --- src/mesa/pipe/cell/spu/Makefile | 4 ++ src/mesa/pipe/cell/spu/main.c | 83 +---------------------------- src/mesa/pipe/cell/spu/main.h | 39 ++------------ src/mesa/pipe/cell/spu/tile.c | 115 ++++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/tile.h | 69 ++++++++++++++++++++++++ src/mesa/pipe/cell/spu/tri.c | 1 + 6 files changed, 194 insertions(+), 117 deletions(-) create mode 100644 src/mesa/pipe/cell/spu/tile.c create mode 100644 src/mesa/pipe/cell/spu/tile.h (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index 8e606dd1a2..bf97a49e78 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -17,6 +17,7 @@ PROG_SPU_EMBED_O = $(PROG)_spu-embed.o SOURCES = \ main.c \ + tile.c \ tri.c SPU_OBJECTS = $(SOURCES:.c=.o) \ @@ -40,6 +41,9 @@ $(PROG_SPU): $(SPU_OBJECTS) main.o: main.c $(SPU_CC) $(SPU_CFLAGS) -c main.c +tile.o: tile.c + $(SPU_CC) $(SPU_CFLAGS) -c tile.c + tri.o: tri.c $(SPU_CC) $(SPU_CFLAGS) -c tri.c diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 5ad15bc639..84134b893a 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -35,6 +35,7 @@ #include "main.h" #include "tri.h" +#include "tile.h" #include "pipe/cell/common.h" #include "pipe/p_defines.h" @@ -50,13 +51,6 @@ volatile struct cell_init_info init; struct framebuffer fb; -uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; - -ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; -ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - - void @@ -69,81 +63,6 @@ wait_on_mask(unsigned tagMask) -void -get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, - int tag, int zBuf) -{ - const uint offset = ty * fb->width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); - const ubyte *src = zBuf ? fb->depth_start : fb->color_start; - - src += offset * bytesPerTile; - - ASSERT(tx < fb->width_tiles); - ASSERT(ty < fb->height_tiles); - ASSERT_ALIGN16(tile); - /* - printf("get_tile: dest: %p src: 0x%x size: %d\n", - tile, (unsigned int) src, bytesPerTile); - */ - mfc_get(tile, /* dest in local memory */ - (unsigned int) src, /* src in main memory */ - bytesPerTile, - tag, - 0, /* tid */ - 0 /* rid */); -} - - -void -put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, - int tag, int zBuf) -{ - const uint offset = ty * fb->width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); - ubyte *dst = zBuf ? fb->depth_start : fb->color_start; - - dst += offset * bytesPerTile; - - ASSERT(tx < fb->width_tiles); - ASSERT(ty < fb->height_tiles); - ASSERT_ALIGN16(tile); - /* - printf("put_tile: src: %p dst: 0x%x size: %d\n", - tile, (unsigned int) dst, bytesPerTile); - */ - mfc_put((void *) tile, /* src in local memory */ - (unsigned int) dst, /* dst in main memory */ - bytesPerTile, - tag, - 0, /* tid */ - 0 /* rid */); -} - - -void -clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value) -{ - uint i, j; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile[i][j] = value; - } - } -} - -void -clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value) -{ - uint i, j; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile[i][j] = value; - } - } -} - - /** * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled * tiles back to the main framebuffer. diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index c539385a07..68280eb67b 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -34,10 +34,6 @@ #include "pipe/cell/common.h" -#define MAX_WIDTH 1024 -#define MAX_HEIGHT 1024 - - extern volatile struct cell_init_info init; struct framebuffer { @@ -56,9 +52,6 @@ struct framebuffer { extern struct framebuffer fb; -extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; - /* DMA TAGS */ @@ -71,34 +64,6 @@ extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; #define TAG_WRITE_TILE_Z 15 - -#define TILE_STATUS_CLEAR 1 -#define TILE_STATUS_DEFINED 2 /**< defined pixel data */ -#define TILE_STATUS_DIRTY 3 /**< modified, but not put back yet */ - -extern ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; -extern ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - - -void -wait_on_mask(unsigned tag); - -void -get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, - int tag, int zBuf); - -void -put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, - int tag, int zBuf); - -void -clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value); - -void -clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value); - - - /** The standard assert macro doesn't seem to work on SPUs */ #define ASSERT(x) \ if (!(x)) { \ @@ -108,4 +73,8 @@ clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value); } +void +wait_on_mask(unsigned tag); + + #endif /* MAIN_H */ diff --git a/src/mesa/pipe/cell/spu/tile.c b/src/mesa/pipe/cell/spu/tile.c new file mode 100644 index 0000000000..ac89d0fab0 --- /dev/null +++ b/src/mesa/pipe/cell/spu/tile.c @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + + +#include "tile.h" + + + +uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; + +ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; +ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + + + +void +get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, + int tag, int zBuf) +{ + const uint offset = ty * fb->width_tiles + tx; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); + const ubyte *src = zBuf ? fb->depth_start : fb->color_start; + + src += offset * bytesPerTile; + + ASSERT(tx < fb->width_tiles); + ASSERT(ty < fb->height_tiles); + ASSERT_ALIGN16(tile); + /* + printf("get_tile: dest: %p src: 0x%x size: %d\n", + tile, (unsigned int) src, bytesPerTile); + */ + mfc_get(tile, /* dest in local memory */ + (unsigned int) src, /* src in main memory */ + bytesPerTile, + tag, + 0, /* tid */ + 0 /* rid */); +} + + +void +put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, + int tag, int zBuf) +{ + const uint offset = ty * fb->width_tiles + tx; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); + ubyte *dst = zBuf ? fb->depth_start : fb->color_start; + + dst += offset * bytesPerTile; + + ASSERT(tx < fb->width_tiles); + ASSERT(ty < fb->height_tiles); + ASSERT_ALIGN16(tile); + /* + printf("put_tile: src: %p dst: 0x%x size: %d\n", + tile, (unsigned int) dst, bytesPerTile); + */ + mfc_put((void *) tile, /* src in local memory */ + (unsigned int) dst, /* dst in main memory */ + bytesPerTile, + tag, + 0, /* tid */ + 0 /* rid */); +} + + +void +clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value) +{ + uint i, j; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile[i][j] = value; + } + } +} + +void +clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value) +{ + uint i, j; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile[i][j] = value; + } + } +} + diff --git a/src/mesa/pipe/cell/spu/tile.h b/src/mesa/pipe/cell/spu/tile.h new file mode 100644 index 0000000000..832cf29e30 --- /dev/null +++ b/src/mesa/pipe/cell/spu/tile.h @@ -0,0 +1,69 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TILE_H +#define TILE_H + + +#include +#include +#include "main.h" +#include "pipe/cell/common.h" + + +#define MAX_WIDTH 1024 +#define MAX_HEIGHT 1024 + + +extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; + + +#define TILE_STATUS_CLEAR 1 +#define TILE_STATUS_DEFINED 2 /**< defined pixel data */ +#define TILE_STATUS_DIRTY 3 /**< modified, but not put back yet */ + +extern ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; +extern ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + + +void +get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, + int tag, int zBuf); + +void +put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, + int tag, int zBuf); + +void +clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value); + +void +clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value); + + +#endif /* TILE_H */ diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index b40f35bc69..f902bf2126 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -45,6 +45,7 @@ #include "pipe/p_format.h" #include "pipe/p_util.h" #include "main.h" +#include "tile.h" #include "tri.h" /* -- cgit v1.2.3 From 44f4b9b9ea81974a8e7de444280e471ca05e9261 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 21:22:03 -0700 Subject: Cell: avoid copying vertex data --- src/mesa/pipe/cell/spu/main.c | 49 ++++++++----------------------------------- src/mesa/pipe/cell/spu/tri.c | 6 ------ src/mesa/pipe/cell/spu/tri.h | 2 +- 3 files changed, 10 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 84134b893a..27e1169a7f 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -174,6 +174,9 @@ tile_bounding_box(const struct cell_command_render *render, *tymin = 0; *box_num_tiles = fb.width_tiles * fb.height_tiles; *box_width_tiles = fb.width_tiles; + (void) render; + (void) txmax; + (void) tymax; #else uint txmax, tymax, box_height_tiles; @@ -255,26 +258,9 @@ render(const struct cell_command_render *render) for (j = 0; j < render->num_verts; j += 3) { struct prim_header prim; - /* - printf(" %u: Triangle %g,%g %g,%g %g,%g\n", - init.id, - prim_buffer.vertex[j*3+0][0][0], - prim_buffer.vertex[j*3+0][0][1], - prim_buffer.vertex[j*3+1][0][0], - prim_buffer.vertex[j*3+1][0][1], - prim_buffer.vertex[j*3+2][0][0], - prim_buffer.vertex[j*3+2][0][1]); - */ - - /* pos */ - COPY_4V(prim.v[0].data[0], prim_buffer.vertex[j+0][0]); - COPY_4V(prim.v[1].data[0], prim_buffer.vertex[j+1][0]); - COPY_4V(prim.v[2].data[0], prim_buffer.vertex[j+2][0]); - - /* color */ - COPY_4V(prim.v[0].data[1], prim_buffer.vertex[j+0][1]); - COPY_4V(prim.v[1].data[1], prim_buffer.vertex[j+1][1]); - COPY_4V(prim.v[2].data[1], prim_buffer.vertex[j+2][1]); + prim.v[0] = (struct vertex_header *) prim_buffer.vertex[j+0]; + prim.v[1] = (struct vertex_header *) prim_buffer.vertex[j+1]; + prim.v[2] = (struct vertex_header *) prim_buffer.vertex[j+2]; tri_draw(&prim, tx, ty); } @@ -391,26 +377,9 @@ render_vbuf(const struct cell_command_render_vbuf *render) v1 = vbuf + indexes[j+1] * render->num_attribs * 4; v2 = vbuf + indexes[j+2] * render->num_attribs * 4; - /* - printf(" %u: Triangle %g,%g %g,%g %g,%g\n", - init.id, - prim_buffer.vertex[j*3+0][0][0], - prim_buffer.vertex[j*3+0][0][1], - prim_buffer.vertex[j*3+1][0][0], - prim_buffer.vertex[j*3+1][0][1], - prim_buffer.vertex[j*3+2][0][0], - prim_buffer.vertex[j*3+2][0][1]); - */ - - /* pos */ - COPY_4V(prim.v[0].data[0], v0); - COPY_4V(prim.v[1].data[0], v1); - COPY_4V(prim.v[2].data[0], v2); - - /* color */ - COPY_4V(prim.v[0].data[1], v0 + 4); - COPY_4V(prim.v[1].data[1], v1 + 4); - COPY_4V(prim.v[2].data[1], v2 + 4); + prim.v[0] = (struct vertex_header *) v0; + prim.v[1] = (struct vertex_header *) v1; + prim.v[2] = (struct vertex_header *) v2; tri_draw(&prim, tx, ty); } diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index f902bf2126..77f7052cd5 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -455,15 +455,9 @@ static void print_vertex(const struct setup_stage *setup, static boolean setup_sort_vertices( struct setup_stage *setup, const struct prim_header *prim ) { -#if 0 const struct vertex_header *v0 = prim->v[0]; const struct vertex_header *v1 = prim->v[1]; const struct vertex_header *v2 = prim->v[2]; -#else - const struct vertex_header *v0 = &prim->v[0]; - const struct vertex_header *v1 = &prim->v[1]; - const struct vertex_header *v2 = &prim->v[2]; -#endif #if DEBUG_VERTS fprintf(stderr, "Triangle:\n"); diff --git a/src/mesa/pipe/cell/spu/tri.h b/src/mesa/pipe/cell/spu/tri.h index 01612a2579..bcb42852b2 100644 --- a/src/mesa/pipe/cell/spu/tri.h +++ b/src/mesa/pipe/cell/spu/tri.h @@ -40,7 +40,7 @@ struct vertex_header { struct prim_header { - struct vertex_header v[3]; + struct vertex_header *v[3]; uint color; }; -- cgit v1.2.3 From c56b20971bfe554f2b9ba74c40b350f950bb31ff Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 21:35:13 -0700 Subject: Cell: make vertex_header and prim_header types private to tri.c --- src/mesa/pipe/cell/spu/main.c | 31 +++++++++++++------------------ src/mesa/pipe/cell/spu/tri.c | 26 ++++++++++++++++++++++++-- src/mesa/pipe/cell/spu/tri.h | 17 +---------------- 3 files changed, 38 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 27e1169a7f..7c42013279 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -256,13 +256,11 @@ render(const struct cell_command_render *render) /* loop over tris */ for (j = 0; j < render->num_verts; j += 3) { - struct prim_header prim; + const float *v0 = (const float *) prim_buffer.vertex[j+0]; + const float *v1 = (const float *) prim_buffer.vertex[j+1]; + const float *v2 = (const float *) prim_buffer.vertex[j+2]; - prim.v[0] = (struct vertex_header *) prim_buffer.vertex[j+0]; - prim.v[1] = (struct vertex_header *) prim_buffer.vertex[j+1]; - prim.v[2] = (struct vertex_header *) prim_buffer.vertex[j+2]; - - tri_draw(&prim, tx, ty); + tri_draw(v0, v1, v2, tx, ty); } /* write color/z tiles back to main framebuffer, if dirtied */ @@ -292,13 +290,16 @@ render_vbuf(const struct cell_command_render_vbuf *render) /* we'll DMA into these buffers */ ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; ushort indexes[CELL_MAX_VBUF_INDEXES] ALIGN16_ATTRIB; - uint i, j, vertex_bytes, index_bytes; + + uint i, j, vertex_size, vertex_bytes, index_bytes; ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(render->index_data); + vertex_size = render->num_attribs * 4 * sizeof(float); + /* how much vertex data */ - vertex_bytes = render->num_verts * render->num_attribs * 4 * sizeof(float); + vertex_bytes = render->num_verts * vertex_size; index_bytes = render->num_indexes * sizeof(ushort); if (index_bytes < 8) index_bytes = 8; @@ -369,19 +370,13 @@ render_vbuf(const struct cell_command_render_vbuf *render) /* loop over tris */ for (j = 0; j < render->num_indexes; j += 3) { - struct prim_header prim; - const float *vbuf = (const float *) vertex_data; const float *v0, *v1, *v2; - v0 = vbuf + indexes[j] * render->num_attribs * 4; - v1 = vbuf + indexes[j+1] * render->num_attribs * 4; - v2 = vbuf + indexes[j+2] * render->num_attribs * 4; - - prim.v[0] = (struct vertex_header *) v0; - prim.v[1] = (struct vertex_header *) v1; - prim.v[2] = (struct vertex_header *) v2; + v0 = (const float *) (vertex_data + indexes[j+0] * vertex_size); + v1 = (const float *) (vertex_data + indexes[j+1] * vertex_size); + v2 = (const float *) (vertex_data + indexes[j+2] * vertex_size); - tri_draw(&prim, tx, ty); + tri_draw(v0, v1, v2, tx, ty); } /* write color/z tiles back to main framebuffer, if dirtied */ diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 77f7052cd5..9eaf1df90b 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -54,6 +54,23 @@ */ +/** + * Simplified types taken from other parts of Gallium + */ + +struct vertex_header { + float data[2][4]; /* pos and color */ +}; + + +struct prim_header { + struct vertex_header *v[3]; + uint color; +}; + + + + #if 1 /* XXX fix this */ @@ -946,10 +963,15 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) * The tile data should have already been fetched. */ void -tri_draw(struct prim_header *tri, uint tx, uint ty) +tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) { + struct prim_header tri; struct setup_stage setup; + tri.v[0] = (struct vertex_header *) v0; + tri.v[1] = (struct vertex_header *) v1; + tri.v[2] = (struct vertex_header *) v2; + setup.tx = tx; setup.ty = ty; @@ -959,5 +981,5 @@ tri_draw(struct prim_header *tri, uint tx, uint ty) setup.cliprect_maxx = (tx + 1) * TILE_SIZE; setup.cliprect_maxy = (ty + 1) * TILE_SIZE; - setup_tri(&setup, tri); + setup_tri(&setup, &tri); } diff --git a/src/mesa/pipe/cell/spu/tri.h b/src/mesa/pipe/cell/spu/tri.h index bcb42852b2..f10c4077d3 100644 --- a/src/mesa/pipe/cell/spu/tri.h +++ b/src/mesa/pipe/cell/spu/tri.h @@ -30,23 +30,8 @@ #define TRI_H -/** - * Simplified types taken from other parts of Gallium - */ - -struct vertex_header { - float data[2][4]; /* pos and color */ -}; - - -struct prim_header { - struct vertex_header *v[3]; - uint color; -}; - - extern void -tri_draw(struct prim_header *tri, uint tx, uint ty); +tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty); #endif /* TRI_H */ -- cgit v1.2.3 From 6059ecaabefc273f39353825dde568d1a2e2ba19 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 21:37:55 -0700 Subject: Cell: remove unused code --- src/mesa/pipe/cell/spu/tri.c | 83 -------------------------------------------- 1 file changed, 83 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 9eaf1df90b..52e48e3dd2 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -29,18 +29,6 @@ * Triangle rendering within a tile. */ - -#if 0 -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_quad.h" -#include "sp_prim_setup.h" -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vertex.h" -#include "pipe/p_util.h" -#endif - - #include "pipe/p_compiler.h" #include "pipe/p_format.h" #include "pipe/p_util.h" @@ -48,10 +36,6 @@ #include "tile.h" #include "tri.h" -/* -#include -#include -*/ /** @@ -62,7 +46,6 @@ struct vertex_header { float data[2][4]; /* pos and color */ }; - struct prim_header { struct vertex_header *v[3]; uint color; @@ -119,11 +102,6 @@ struct interp_coef * Also used for line drawing (taking some liberties). */ struct setup_stage { -#if 0 - struct draw_stage stage; /**< This must be first (base class) */ - - struct softpipe_context *softpipe; -#endif /* Vertices are just an array of floats making up each attribute in * turn. Currently fixed at 4 floats, but should change in time. @@ -897,67 +875,6 @@ setup_tri(struct setup_stage *setup, struct prim_header *prim) - -#if 0 -static void setup_begin( struct draw_stage *stage ) -{ - struct setup_stage *setup = setup_stage(stage); - struct softpipe_context *sp = setup->softpipe; - - setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; - - sp->quad.first->begin(sp->quad.first); -} -#endif - -#if 0 -static void setup_end( struct draw_stage *stage ) -{ -} -#endif - - -#if 0 -static void reset_stipple_counter( struct draw_stage *stage ) -{ - struct setup_stage *setup = setup_stage(stage); - setup->softpipe->line_stipple_counter = 0; -} -#endif - -#if 0 -static void render_destroy( struct draw_stage *stage ) -{ - FREE( stage ); -} -#endif - - -#if 0 -/** - * Create a new primitive setup/render stage. - */ -struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) -{ - struct setup_stage *setup = CALLOC_STRUCT(setup_stage); - - setup->softpipe = softpipe; - setup->stage.draw = softpipe->draw; - setup->stage.begin = setup_begin; - setup->stage.point = setup_point; - setup->stage.line = setup_line; - setup->stage.tri = setup_tri; - setup->stage.end = setup_end; - setup->stage.reset_stipple_counter = reset_stipple_counter; - setup->stage.destroy = render_destroy; - - setup->quad.coef = setup->coef; - - return &setup->stage; -} -#endif - - /** * Draw triangle into tile at (tx, ty) (tile coords) * The tile data should have already been fetched. -- cgit v1.2.3 From ea190f4b41d8959aa3edfd46503a3077201cdefd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 21:50:31 -0700 Subject: Cell: remove unused color field --- src/mesa/pipe/cell/spu/tri.c | 10 ---------- 1 file changed, 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c index 52e48e3dd2..a4fcb71d20 100644 --- a/src/mesa/pipe/cell/spu/tri.c +++ b/src/mesa/pipe/cell/spu/tri.c @@ -48,7 +48,6 @@ struct vertex_header { struct prim_header { struct vertex_header *v[3]; - uint color; }; @@ -131,9 +130,6 @@ struct setup_stage { #if 0 struct quad_header quad; #endif -#if 1 - uint color; -#endif struct { int left[2]; /**< [0] = row0, [1] = row1 */ @@ -832,12 +828,6 @@ static void subtriangle( struct setup_stage *setup, static void setup_tri(struct setup_stage *setup, struct prim_header *prim) { - setup->color = prim->color; /* XXX temporary */ - - /* - _mesa_printf("%s\n", __FUNCTION__ ); - */ - if (!setup_sort_vertices( setup, prim )) { return; /* totally clipped */ } -- cgit v1.2.3 From 71caa922e0bf9f6378bd02374402eaea2990b493 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 21:50:55 -0700 Subject: Cell: compute bounding box in cell_vbuf_draw() --- src/mesa/pipe/cell/ppu/cell_vbuf.c | 31 ++++++++++++++++++++----------- 1 file changed, 20 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 059a098914..62d453cb7f 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -95,8 +95,24 @@ cell_vbuf_draw(struct vbuf_render *vbr, { struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); struct cell_context *cell = cvbr->cell; + float xmin, ymin, xmax, ymax; uint i; + /* compute x/y bounding box */ + xmin = ymin = 1e50; + xmax = ymax = -1e50; + for (i = 0; i < nr_vertices; i++) { + const float *v = (float *) ((ubyte *) vertices + i * vertex_size); + if (v[0] < xmin) + xmin = v[0]; + if (v[0] > xmax) + xmax = v[0]; + if (v[1] < ymin) + ymin = v[1]; + if (v[1] > ymax) + ymax = v[1]; + } + /*printf("cell_vbuf_draw nr_indices = %u\n", nr_indices);*/ if (prim != PIPE_PRIM_TRIANGLES) @@ -110,17 +126,10 @@ cell_vbuf_draw(struct vbuf_render *vbr, render->vertex_data = vertices; render->index_data = indices; render->num_indexes = nr_indices; -#if 0 - render->xmin = cell->prim_buffer.xmin; - render->ymin = cell->prim_buffer.ymin; - render->xmax = cell->prim_buffer.xmax; - render->ymax = cell->prim_buffer.ymax; -#else - render->xmin = -100; - render->ymin = -100; - render->xmax = 100; - render->ymax = 100; -#endif + render->xmin = xmin; + render->ymin = ymin; + render->xmax = xmax; + render->ymax = ymax; ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(render->index_data); -- cgit v1.2.3 From aee5f471ce30f0511769c02a24160e9bb5047792 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 21:53:38 -0700 Subject: Cell: remove unneeded #includes --- src/mesa/pipe/cell/spu/main.h | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index 68280eb67b..9351aa340b 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -29,8 +29,6 @@ #define MAIN_H -#include -#include #include "pipe/cell/common.h" -- cgit v1.2.3 From 299dffce4fafa2ed03a6e1f7ca7a5357c147477e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 10 Jan 2008 21:59:15 -0700 Subject: Cell: wait_on_mask_all() --- src/mesa/pipe/cell/spu/main.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 7c42013279..548ef127ba 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -62,6 +62,15 @@ wait_on_mask(unsigned tagMask) } +static void +wait_on_mask_all(unsigned tagMask) +{ + mfc_write_tag_mask( tagMask ); + /* wait for completion of _any_ DMAs specified by tagMask */ + mfc_read_tag_status_all(); +} + + /** * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled @@ -327,8 +336,8 @@ render_vbuf(const struct cell_command_render_vbuf *render) 0, /* tid */ 0 /* rid */); - wait_on_mask(1 << TAG_VERTEX_BUFFER); - wait_on_mask(1 << TAG_INDEX_BUFFER); + wait_on_mask_all((1 << TAG_VERTEX_BUFFER) | + (1 << TAG_INDEX_BUFFER)); /* find tiles which intersect the prim bounding box */ uint txmin, tymin, box_width_tiles, box_num_tiles; -- cgit v1.2.3 From 1217d5cca3503103cc40221ea8287960236e3734 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 08:14:42 -0700 Subject: Cell: Remove the pre-vbuf rendering code --- src/mesa/pipe/cell/common.h | 13 +---- src/mesa/pipe/cell/ppu/Makefile | 1 - src/mesa/pipe/cell/ppu/cell_context.c | 8 +-- src/mesa/pipe/cell/ppu/cell_vbuf.c | 4 +- src/mesa/pipe/cell/spu/main.c | 104 ++-------------------------------- 5 files changed, 9 insertions(+), 121 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index dc2db299c4..60abe2d9bc 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -52,7 +52,6 @@ #define CELL_CMD_CLEAR_SURFACE 3 #define CELL_CMD_FINISH 4 #define CELL_CMD_RENDER 5 -#define CELL_CMD_RENDER_VBUF 6 /** @@ -76,18 +75,9 @@ struct cell_command_clear_surface } ALIGN16_ATTRIB; -struct cell_command_render -{ - uint prim_type; - uint num_verts, num_attribs; - float xmin, ymin, xmax, ymax; - void *vertex_data; -} ALIGN16_ATTRIB; - - #define CELL_MAX_VBUF_SIZE (16 * 1024) #define CELL_MAX_VBUF_INDEXES 1024 -struct cell_command_render_vbuf +struct cell_command_render { uint prim_type; uint num_verts, num_attribs; @@ -104,7 +94,6 @@ struct cell_command struct cell_command_framebuffer fb; struct cell_command_clear_surface clear; struct cell_command_render render; - struct cell_command_render_vbuf render_vbuf; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index 9bf89702d5..44f14c0211 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -19,7 +19,6 @@ SOURCES = \ cell_context.c \ cell_draw_arrays.c \ cell_flush.c \ - cell_render.c \ cell_state_blend.c \ cell_state_clip.c \ cell_state_derived.c \ diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 52c97c939b..defce2869a 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -237,14 +237,8 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->draw = draw_create(); -#define VBUF 1 -#if VBUF cell_init_vbuf(cell); draw_set_rasterize_stage(cell->draw, cell->vbuf); -#else - cell->render_stage = cell_draw_render_stage(cell); - draw_set_rasterize_stage(cell->draw, cell->render_stage); -#endif cell->prim_buffer.xmin = 1e100; cell->prim_buffer.ymin = 1e100; @@ -254,7 +248,7 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) /* * SPU stuff */ - cell->num_spus = 6; /* XXX >6 seems to fail */ + cell->num_spus = 6; cell_start_spus(cell->num_spus); diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 62d453cb7f..1d9a57ce44 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -119,7 +119,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, return; /* only render tris for now */ for (i = 0; i < cell->num_spus; i++) { - struct cell_command_render_vbuf *render = &cell_global.command[i].render_vbuf; + struct cell_command_render *render = &cell_global.command[i].render; render->prim_type = prim; render->num_verts = nr_vertices; render->num_attribs = CELL_MAX_ATTRIBS; /* XXX fix */ @@ -133,7 +133,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(render->index_data); - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER_VBUF); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER); } cell_flush(&cell->pipe, 0x0); diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 548ef127ba..20f15029e1 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -207,94 +207,6 @@ tile_bounding_box(const struct cell_command_render *render, static void render(const struct cell_command_render *render) -{ - struct cell_prim_buffer prim_buffer ALIGN16_ATTRIB; - uint i, j, vertex_bytes; - - /* - printf("SPU %u: RENDER buffer dst=%p src=%p size=%d\n", - init.id, - &prim_buffer, render->vertex_data, (int)sizeof(prim_buffer)); - */ - - ASSERT_ALIGN16(render->vertex_data); - ASSERT_ALIGN16(&prim_buffer); - - /* how much vertex data */ - vertex_bytes = render->num_verts * render->num_attribs * 4 * sizeof(float); - - /* get vertex data from main memory */ - mfc_get(&prim_buffer, /* dest */ - (unsigned int) render->vertex_data, /* src */ - vertex_bytes, /* size */ - TAG_VERTEX_BUFFER, - 0, /* tid */ - 0 /* rid */); - wait_on_mask(1 << TAG_VERTEX_BUFFER); - - /* find tiles which intersect the prim bounding box */ - uint txmin, tymin, box_width_tiles, box_num_tiles; - tile_bounding_box(render, &txmin, &tymin, - &box_num_tiles, &box_width_tiles); - - /* make sure any pending clears have completed */ - wait_on_mask(1 << TAG_SURFACE_CLEAR); - - /* loop over tiles */ - for (i = init.id; i < box_num_tiles; i += init.num_spus) { - const uint tx = txmin + i % box_width_tiles; - const uint ty = tymin + i / box_width_tiles; - - ASSERT(tx < fb.width_tiles); - ASSERT(ty < fb.height_tiles); - - /* Start fetching color/z tiles. We'll wait for completion when - * we need read/write to them later in triangle rasterization. - */ - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(&fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); - } - } - - if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); - } - - ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); - - /* loop over tris */ - for (j = 0; j < render->num_verts; j += 3) { - const float *v0 = (const float *) prim_buffer.vertex[j+0]; - const float *v1 = (const float *) prim_buffer.vertex[j+1]; - const float *v2 = (const float *) prim_buffer.vertex[j+2]; - - tri_draw(v0, v1, v2, tx, ty); - } - - /* write color/z tiles back to main framebuffer, if dirtied */ - if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(&fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); - tile_status[ty][tx] = TILE_STATUS_DEFINED; - } - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(&fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); - tile_status_z[ty][tx] = TILE_STATUS_DEFINED; - } - } - - /* XXX move these... */ - wait_on_mask(1 << TAG_WRITE_TILE_COLOR); - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - wait_on_mask(1 << TAG_WRITE_TILE_Z); - } - } -} - - -static void -render_vbuf(const struct cell_command_render_vbuf *render) { /* we'll DMA into these buffers */ ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; @@ -483,18 +395,12 @@ main_loop(void) break; case CELL_CMD_RENDER: if (Debug) - printf("SPU %u: RENDER %u verts, prim %u\n", - init.id, cmd.render.num_verts, cmd.render.prim_type); - render(&cmd.render); - break; - case CELL_CMD_RENDER_VBUF: - if (Debug) - printf("SPU %u: RENDER_VBUF prim %u, indices: %u, nr_vert: %u\n", + printf("SPU %u: RENDER prim %u, indices: %u, nr_vert: %u\n", init.id, - cmd.render_vbuf.prim_type, - cmd.render_vbuf.num_verts, - cmd.render_vbuf.num_indexes); - render_vbuf(&cmd.render_vbuf); + cmd.render.prim_type, + cmd.render.num_verts, + cmd.render.num_indexes); + render(&cmd.render); break; case CELL_CMD_FINISH: -- cgit v1.2.3 From 45230b4e082473f7e1f0c0a18dd0e2fc9e5508e1 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 10:10:45 -0700 Subject: Fix malloc size buf, silence signed/unsigned comparision warning --- src/mesa/pipe/draw/draw_vbuf.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 3f6606de86..62aa6de68f 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -275,7 +275,7 @@ vbuf_flush_indices( struct draw_stage *stage ) if(!vbuf->nr_indices) return; - assert(vbuf->vertex_ptr - vbuf->vertices == + assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); switch(vbuf->prim) { @@ -405,7 +405,8 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, assert(render->max_indices < UNDEFINED_VERTEX_ID); vbuf->max_indices = render->max_indices; - vbuf->indices = align_malloc( vbuf->max_indices, 16 ); + vbuf->indices + = align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); vbuf->vertices = NULL; vbuf->vertex_ptr = vbuf->vertices; -- cgit v1.2.3 From d4c806302e96a0fd97a4613bc29f4c9863ae89fe Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 11:27:09 -0700 Subject: s/int/uint/ to silence warning --- src/mesa/pipe/draw/draw_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index a9ff54404f..d134b05717 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -69,7 +69,7 @@ struct draw_context *draw_create( void ) /* Statically allocate maximum sized vertices for the cache - could be cleverer... */ { - int i; + uint i; char *tmp = MALLOC( Elements(draw->vcache.vertex) * MAX_VERTEX_SIZE ); for (i = 0; i < Elements(draw->vcache.vertex); i++) -- cgit v1.2.3 From f313a1ece297e2a3bea5b611082f8cc0012c64b8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 11:27:24 -0700 Subject: Cell: reformattting --- src/mesa/pipe/cell/ppu/cell_draw_arrays.c | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c index c014f5b29d..537cec8785 100644 --- a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c +++ b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c @@ -121,18 +121,17 @@ cell_draw_elements(struct pipe_context *pipe, */ for (i = 0; i < PIPE_ATTRIB_MAX; i++) { if (sp->vertex_buffer[i].buffer) { - void *buf - = pipe->winsys->buffer_map(pipe->winsys, - sp->vertex_buffer[i].buffer, - PIPE_BUFFER_FLAG_READ); + void *buf = pipe->winsys->buffer_map(pipe->winsys, + sp->vertex_buffer[i].buffer, + PIPE_BUFFER_FLAG_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } } /* Map index buffer, if present */ if (indexBuffer) { - void *mapped_indexes - = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_FLAG_READ); + void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, + indexBuffer, + PIPE_BUFFER_FLAG_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { @@ -161,7 +160,6 @@ cell_draw_elements(struct pipe_context *pipe, draw_set_mapped_element_buffer(draw, 0, NULL); } - /* Note: leave drawing surfaces mapped */ cell_unmap_constant_buffers(sp); -- cgit v1.2.3 From 21b282e9347d06272e5dfc795aa7c0b861c240e1 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 11:30:02 -0700 Subject: vbuf_flush_vertices() instead of vbuf_flush_indices() in vbuf_end() --- src/mesa/pipe/draw/draw_vbuf.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 62aa6de68f..4f59b1b25d 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -360,8 +360,14 @@ vbuf_begin( struct draw_stage *stage ) static void vbuf_end( struct draw_stage *stage ) { +#if 0 /* XXX: Overkill */ vbuf_flush_indices( stage ); +#else + /* By flushing vertices we avoid having the vertex buffer grow and grow */ + struct vbuf_stage *vbuf = vbuf_stage(stage); + vbuf_flush_vertices( stage, vbuf->vertex_size ); +#endif stage->point = vbuf_first_point; stage->line = vbuf_first_line; -- cgit v1.2.3 From 6b5d674f915b67f3b51c331e3a6ba02e9f82473f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 11:31:03 -0700 Subject: Cell: debug code, comments --- src/mesa/pipe/cell/ppu/cell_vbuf.c | 17 +++++++++++++++-- 1 file changed, 15 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 1d9a57ce44..63c3b39ec7 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -98,6 +98,20 @@ cell_vbuf_draw(struct vbuf_render *vbr, float xmin, ymin, xmax, ymax; uint i; +#if 0 + printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u\n", + nr_indices, nr_vertices); + printf(" "); + for (i = 0; i < nr_indices; i += 3) { + printf("%u %u %u, ", indices[i+0], indices[i+1], indices[i+2]); + } + printf("\n"); +#elif 0 + printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u indexes = [%u %u %u ...]\n", + nr_indices, nr_vertices, + indices[0], indices[1], indices[2]); +#endif + /* compute x/y bounding box */ xmin = ymin = 1e50; xmax = ymax = -1e50; @@ -113,8 +127,6 @@ cell_vbuf_draw(struct vbuf_render *vbr, ymax = v[1]; } - /*printf("cell_vbuf_draw nr_indices = %u\n", nr_indices);*/ - if (prim != PIPE_PRIM_TRIANGLES) return; /* only render tris for now */ @@ -136,6 +148,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER); } + /* XXX this is temporary */ cell_flush(&cell->pipe, 0x0); } -- cgit v1.2.3 From 7db94ba03173f345a47fd727e8d866a87a484415 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 11:35:03 -0700 Subject: Cell: remove obsolete cell_prim_buffer struct and code --- src/mesa/pipe/cell/common.h | 14 ++------------ src/mesa/pipe/cell/ppu/cell_context.c | 5 ----- src/mesa/pipe/cell/ppu/cell_context.h | 2 -- 3 files changed, 2 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 60abe2d9bc..c00cee6a02 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -54,6 +54,7 @@ #define CELL_CMD_RENDER 5 + /** * Tell SPUs about the framebuffer size, location */ @@ -77,6 +78,7 @@ struct cell_command_clear_surface #define CELL_MAX_VBUF_SIZE (16 * 1024) #define CELL_MAX_VBUF_INDEXES 1024 +#define CELL_MAX_ATTRIBS 2 /* temporary! */ struct cell_command_render { uint prim_type; @@ -106,16 +108,4 @@ struct cell_init_info } ALIGN16_ATTRIB; -/** Temporary */ -#define CELL_MAX_VERTS 240 -#define CELL_MAX_ATTRIBS 2 -struct cell_prim_buffer -{ - float vertex[CELL_MAX_VERTS][CELL_MAX_ATTRIBS][4] ALIGN16_ATTRIB; - float xmin, ymin, xmax, ymax; - uint num_verts; -} ALIGN16_ATTRIB; - - - #endif /* CELL_COMMON_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index defce2869a..000217b259 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -240,11 +240,6 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell_init_vbuf(cell); draw_set_rasterize_stage(cell->draw, cell->vbuf); - cell->prim_buffer.xmin = 1e100; - cell->prim_buffer.ymin = 1e100; - cell->prim_buffer.xmax = -1e100; - cell->prim_buffer.ymax = -1e100; - /* * SPU stuff */ diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 4ff39a53c6..d7754d5d61 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -96,8 +96,6 @@ struct cell_context uint num_spus; - - struct cell_prim_buffer prim_buffer; }; -- cgit v1.2.3 From 08c2571fb48d41731c81cc402acabf709523c831 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 12:04:55 -0700 Subject: Cell: initial work for a hw-like batch buffer system --- src/mesa/pipe/cell/common.h | 22 +++++++++ src/mesa/pipe/cell/ppu/Makefile | 1 + src/mesa/pipe/cell/ppu/cell_batch.c | 85 +++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_batch.h | 40 +++++++++++++++++ src/mesa/pipe/cell/ppu/cell_context.c | 7 ++- src/mesa/pipe/cell/ppu/cell_context.h | 5 +++ src/mesa/pipe/cell/ppu/cell_spu.c | 13 +++--- src/mesa/pipe/cell/ppu/cell_spu.h | 10 ++--- src/mesa/pipe/cell/spu/main.c | 19 +++++++- 9 files changed, 190 insertions(+), 12 deletions(-) create mode 100644 src/mesa/pipe/cell/ppu/cell_batch.c create mode 100644 src/mesa/pipe/cell/ppu/cell_batch.h (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index c00cee6a02..450ca1d722 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -47,13 +47,23 @@ #define TILE_SIZE 32 +/** + * The low byte of a mailbox word contains the command opcode. + * Remaining higher bytes are command specific. + */ +#define CELL_CMD_OPCODE_MASK 0xf + #define CELL_CMD_EXIT 1 #define CELL_CMD_FRAMEBUFFER 2 #define CELL_CMD_CLEAR_SURFACE 3 #define CELL_CMD_FINISH 4 #define CELL_CMD_RENDER 5 +#define CELL_CMD_BATCH 6 +#define CELL_NUM_BATCH_BUFFERS 2 +#define CELL_BATCH_BUFFER_SIZE 1024 /**< 16KB would be the max */ + /** * Tell SPUs about the framebuffer size, location @@ -90,12 +100,23 @@ struct cell_command_render } ALIGN16_ATTRIB; +/** + * Execute a command/batch buffer. + */ +struct cell_command_batch +{ + ushort buffer; /**< which buffer [0, CELL_NUM_CMD_BUFFFERS-1] */ + ushort length; /**< in bytes */ +} ALIGN16_ATTRIB; + + /** XXX unions don't seem to work */ struct cell_command { struct cell_command_framebuffer fb; struct cell_command_clear_surface clear; struct cell_command_render render; + struct cell_command_batch batch; } ALIGN16_ATTRIB; @@ -105,6 +126,7 @@ struct cell_init_info unsigned id; unsigned num_spus; struct cell_command *cmd; + ubyte *batch_buffers[CELL_NUM_BATCH_BUFFERS]; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index 44f14c0211..f7e3dd09f4 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -16,6 +16,7 @@ SPU_CODE_MODULE = ../spu/g3d_spu.a SOURCES = \ + cell_batch.c \ cell_context.c \ cell_draw_arrays.c \ cell_flush.c \ diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c new file mode 100644 index 0000000000..a71573cbf7 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -0,0 +1,85 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "cell_context.h" +#include "cell_batch.h" +#include "cell_spu.h" + + +void +cell_batch_flush(struct cell_context *cell) +{ + const uint batch = cell->cur_batch; + const uint size = cell->batch_buffer_size[batch]; + uint i, cmd_word; + + if (size == 0) + return; + + assert(batch < CELL_NUM_BATCH_BUFFERS); + + printf("cell_batch_dispatch: buf %u, size %u\n", batch, size); + + cmd_word = CELL_CMD_BATCH | (batch << 8) | (size << 16); + + for (i = 0; i < cell->num_spus; i++) { + send_mbox_message(cell_global.spe_contexts[i], cmd_word); + } + + /* XXX wait on DMA xfer of prev buffer to complete */ + + cell->cur_batch = (batch + 1) % CELL_NUM_BATCH_BUFFERS; + + cell->batch_buffer_size[cell->cur_batch] = 0; /* empty */ +} + + +/** + * \param cmd command to append + * \param length command size in bytes + */ +void +cell_batch_append(struct cell_context *cell, const void *cmd, uint length) +{ + uint size; + + assert(cell->cur_batch >= 0); + + size = cell->batch_buffer_size[cell->cur_batch]; + + if (size + length > CELL_BATCH_BUFFER_SIZE) { + cell_batch_flush(cell); + size = 0; + } + + assert(size + length <= CELL_BATCH_BUFFER_SIZE); + + memcpy(cell->batch_buffer[cell->cur_batch] + size, cmd, length); + + cell->batch_buffer_size[cell->cur_batch] = size + length; +} diff --git a/src/mesa/pipe/cell/ppu/cell_batch.h b/src/mesa/pipe/cell/ppu/cell_batch.h new file mode 100644 index 0000000000..02d7edf8b9 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_batch.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_BATCH_H +#define CELL_BATCH_H + + +extern void +cell_batch_flush(struct cell_context *cell); + +extern void +cell_batch_append(struct cell_context *cell, const void *cmd, uint length); + + +#endif /* CELL_BATCH_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 000217b259..6ba3b0d413 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -160,6 +160,7 @@ struct pipe_context * cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) { struct cell_context *cell; + uint i; /* some fields need to be 16-byte aligned, so align the whole object */ cell = (struct cell_context*) align_malloc(sizeof(struct cell_context), 16); @@ -245,7 +246,11 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) */ cell->num_spus = 6; - cell_start_spus(cell->num_spus); + cell_start_spus(cell); + + for (i = 0; i < CELL_NUM_BATCH_BUFFERS; i++) { + cell->batch_buffer_size[i] = 0; + } #if 0 test_spus(cell); diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index d7754d5d61..94d8f08da8 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -96,6 +96,11 @@ struct cell_context uint num_spus; + + ubyte batch_buffer_size[CELL_NUM_BATCH_BUFFERS]; + ubyte batch_buffer[CELL_NUM_BATCH_BUFFERS][CELL_BATCH_BUFFER_SIZE] ALIGN16_ATTRIB; + int cur_batch; /**< which batch buffer is being filled */ + }; diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 44ad3f22b3..88a620e669 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -95,11 +95,11 @@ static void *cell_thread_function(void *arg) * Create the SPU threads */ void -cell_start_spus(uint num_spus) +cell_start_spus(struct cell_context *cell) { - uint i; + uint i, j; - assert(num_spus <= MAX_SPUS); + assert(cell->num_spus <= MAX_SPUS); ASSERT_ALIGN16(&cell_global.command[0]); ASSERT_ALIGN16(&cell_global.command[1]); @@ -107,10 +107,13 @@ cell_start_spus(uint num_spus) ASSERT_ALIGN16(&cell_global.inits[0]); ASSERT_ALIGN16(&cell_global.inits[1]); - for (i = 0; i < num_spus; i++) { + for (i = 0; i < cell->num_spus; i++) { cell_global.inits[i].id = i; - cell_global.inits[i].num_spus = num_spus; + cell_global.inits[i].num_spus = cell->num_spus; cell_global.inits[i].cmd = &cell_global.command[i]; + for (j = 0; j < CELL_NUM_BATCH_BUFFERS; j++) { + cell_global.inits[i].batch_buffers[j] = cell->batch_buffer[j]; + } cell_global.spe_contexts[i] = spe_context_create(0, NULL); if (!cell_global.spe_contexts[i]) { diff --git a/src/mesa/pipe/cell/ppu/cell_spu.h b/src/mesa/pipe/cell/ppu/cell_spu.h index d11001af62..b4bfbced80 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.h +++ b/src/mesa/pipe/cell/ppu/cell_spu.h @@ -71,18 +71,18 @@ extern uint wait_mbox_message(spe_context_ptr_t ctx); -void -cell_start_spus(uint num_spus); +extern void +cell_start_spus(struct cell_context *cell); -void +extern void finish_all(uint num_spus); -void +extern void test_spus(struct cell_context *cell); -void +extern void cell_spu_exit(struct cell_context *cell); diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 20f15029e1..04bb087cf9 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -322,6 +322,13 @@ render(const struct cell_command_render *render) +static void +batch(const struct cell_command_batch *batch) +{ +} + + + /** * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. */ @@ -359,7 +366,7 @@ main_loop(void) 0 /* rid */); wait_on_mask( 1 << tag ); - switch (opcode) { + switch (opcode & CELL_CMD_OPCODE_MASK) { case CELL_CMD_EXIT: if (Debug) printf("SPU %u: EXIT\n", init.id); @@ -403,6 +410,16 @@ main_loop(void) render(&cmd.render); break; + case CELL_CMD_BATCH: + /* execute a batch buffer */ + if (Debug) + printf("SPU %u: BATCH buffer %u, len %u\n", + init.id, + cmd.batch.buffer, + cmd.batch.length); + batch(&cmd.batch); + break; + case CELL_CMD_FINISH: if (Debug) printf("SPU %u: FINISH\n", init.id); -- cgit v1.2.3 From 8dd678208e4dcd90e07dc271d11d73d87465e0fd Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 16:08:53 -0700 Subject: Cell: basic batch buffer working --- src/mesa/pipe/cell/common.h | 14 +-- src/mesa/pipe/cell/ppu/cell_batch.c | 27 ++++- src/mesa/pipe/cell/ppu/cell_batch.h | 6 + src/mesa/pipe/cell/ppu/cell_flush.c | 19 ++- src/mesa/pipe/cell/ppu/cell_state_surface.c | 20 +++- src/mesa/pipe/cell/ppu/cell_surface.c | 12 ++ src/mesa/pipe/cell/ppu/cell_vbuf.c | 27 ++++- src/mesa/pipe/cell/spu/main.c | 175 +++++++++++++++++++--------- src/mesa/pipe/cell/spu/main.h | 4 +- 9 files changed, 231 insertions(+), 73 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 450ca1d722..3d6f1c4ba0 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -70,6 +70,7 @@ */ struct cell_command_framebuffer { + uint opcode; int width, height; void *color_start, *depth_start; enum pipe_format color_format, depth_format; @@ -81,6 +82,7 @@ struct cell_command_framebuffer */ struct cell_command_clear_surface { + uint opcode; uint surface; /**< Temporary: 0=color, 1=Z */ uint value; } ALIGN16_ATTRIB; @@ -91,6 +93,7 @@ struct cell_command_clear_surface #define CELL_MAX_ATTRIBS 2 /* temporary! */ struct cell_command_render { + uint opcode; uint prim_type; uint num_verts, num_attribs; uint num_indexes; @@ -100,23 +103,12 @@ struct cell_command_render } ALIGN16_ATTRIB; -/** - * Execute a command/batch buffer. - */ -struct cell_command_batch -{ - ushort buffer; /**< which buffer [0, CELL_NUM_CMD_BUFFFERS-1] */ - ushort length; /**< in bytes */ -} ALIGN16_ATTRIB; - - /** XXX unions don't seem to work */ struct cell_command { struct cell_command_framebuffer fb; struct cell_command_clear_surface clear; struct cell_command_render render; - struct cell_command_batch batch; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index a71573cbf7..7eb1b50ddb 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -43,7 +43,7 @@ cell_batch_flush(struct cell_context *cell) assert(batch < CELL_NUM_BATCH_BUFFERS); - printf("cell_batch_dispatch: buf %u, size %u\n", batch, size); + /*printf("cell_batch_dispatch: buf %u, size %u\n", batch, size);*/ cmd_word = CELL_CMD_BATCH | (batch << 8) | (size << 16); @@ -83,3 +83,28 @@ cell_batch_append(struct cell_context *cell, const void *cmd, uint length) cell->batch_buffer_size[cell->cur_batch] = size + length; } + + +void * +cell_batch_alloc(struct cell_context *cell, uint bytes) +{ + void *pos; + uint size; + + assert(cell->cur_batch >= 0); + + size = cell->batch_buffer_size[cell->cur_batch]; + + if (size + bytes > CELL_BATCH_BUFFER_SIZE) { + cell_batch_flush(cell); + size = 0; + } + + assert(size + bytes <= CELL_BATCH_BUFFER_SIZE); + + pos = (void *) (cell->batch_buffer[cell->cur_batch] + size); + + cell->batch_buffer_size[cell->cur_batch] = size + bytes; + + return pos; +} diff --git a/src/mesa/pipe/cell/ppu/cell_batch.h b/src/mesa/pipe/cell/ppu/cell_batch.h index 02d7edf8b9..7a5c6392d7 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.h +++ b/src/mesa/pipe/cell/ppu/cell_batch.h @@ -30,11 +30,17 @@ #define CELL_BATCH_H +struct cell_context; + + extern void cell_batch_flush(struct cell_context *cell); extern void cell_batch_append(struct cell_context *cell, const void *cmd, uint length); +extern void * +cell_batch_alloc(struct cell_context *cell, uint bytes); + #endif /* CELL_BATCH_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c index 977c50da32..33cbe2a085 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -38,16 +38,25 @@ cell_flush(struct pipe_context *pipe, unsigned flags) struct cell_context *cell = cell_context(pipe); uint i; - cell_flush_prim_buffer(cell); + if (flags & PIPE_FLUSH_WAIT) { + uint *cmd = (uint *) cell_batch_alloc(cell, sizeof(uint)); + *cmd = CELL_CMD_FINISH; + } + + cell_batch_flush(cell); +#if 0 /* Send CMD_FINISH to all SPUs */ for (i = 0; i < cell->num_spus; i++) { send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FINISH); } +#endif - /* Wait for ack */ - for (i = 0; i < cell->num_spus; i++) { - uint k = wait_mbox_message(cell_global.spe_contexts[i]); - assert(k == CELL_CMD_FINISH); + if (flags & PIPE_FLUSH_WAIT) { + /* Wait for ack */ + for (i = 0; i < cell->num_spus; i++) { + uint k = wait_mbox_message(cell_global.spe_contexts[i]); + assert(k == CELL_CMD_FINISH); + } } } diff --git a/src/mesa/pipe/cell/ppu/cell_state_surface.c b/src/mesa/pipe/cell/ppu/cell_state_surface.c index cd8e5def31..ff8fdbd334 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_state_surface.c @@ -27,6 +27,7 @@ #include "pipe/p_inlines.h" +#include "cell_batch.h" #include "cell_context.h" #include "cell_state.h" #include "cell_spu.h" @@ -68,8 +69,10 @@ cell_set_framebuffer_state(struct pipe_context *pipe, if (zsurf && !zsurf->map) pipe_surface_map(zsurf); +#if 0 for (i = 0; i < cell->num_spus; i++) { struct cell_command_framebuffer *fb = &cell_global.command[i].fb; + fb->opcode = CELL_CMD_FRAMEBUFFER; fb->color_start = csurf->map; fb->color_format = csurf->format; fb->depth_start = zsurf ? zsurf->map : NULL; @@ -78,7 +81,22 @@ cell_set_framebuffer_state(struct pipe_context *pipe, fb->height = csurf->height; send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FRAMEBUFFER); } - +#endif +#if 1 + { + struct cell_command_framebuffer *fb + = cell_batch_alloc(cell, sizeof(*fb)); + fb->opcode = CELL_CMD_FRAMEBUFFER; + fb->color_start = csurf->map; + fb->color_format = csurf->format; + fb->depth_start = zsurf ? zsurf->map : NULL; + fb->depth_format = zsurf ? zsurf->format : PIPE_FORMAT_NONE; + fb->width = csurf->width; + fb->height = csurf->height; + /*cell_batch_flush(cell);*/ + /*cell_flush(&cell->pipe, 0x0);*/ + } +#endif cell->dirty |= CELL_NEW_FRAMEBUFFER; } diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 03dd41583f..ea1c2bd644 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -37,6 +37,7 @@ #include "pipe/p_util.h" #include "pipe/cell/common.h" #include "cell_context.h" +#include "cell_batch.h" #include "cell_surface.h" #include "cell_spu.h" @@ -59,6 +60,7 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, surfIndex = 0; } +#if 0 for (i = 0; i < cell->num_spus; i++) { #if 1 uint clr = clearValue; @@ -80,6 +82,16 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, cell_global.command[i].clear.surface = surfIndex; send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_SURFACE); } +#else + { + struct cell_command_clear_surface *clr + = (struct cell_command_clear_surface *) + cell_batch_alloc(cell, sizeof(*clr)); + clr->opcode = CELL_CMD_CLEAR_SURFACE; + clr->surface = surfIndex; + clr->value = clearValue; + } +#endif /* XXX temporary */ cell_flush(&cell->pipe, 0x0); diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 63c3b39ec7..d27d718dae 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -130,8 +130,10 @@ cell_vbuf_draw(struct vbuf_render *vbr, if (prim != PIPE_PRIM_TRIANGLES) return; /* only render tris for now */ +#if 0 for (i = 0; i < cell->num_spus; i++) { struct cell_command_render *render = &cell_global.command[i].render; + render->opcode = CELL_CMD_RENDER; render->prim_type = prim; render->num_verts = nr_vertices; render->num_attribs = CELL_MAX_ATTRIBS; /* XXX fix */ @@ -147,9 +149,32 @@ cell_vbuf_draw(struct vbuf_render *vbr, ASSERT_ALIGN16(render->index_data); send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER); } +#else + { + struct cell_command_render *render + = (struct cell_command_render *) + cell_batch_alloc(cell, sizeof(*render)); + render->opcode = CELL_CMD_RENDER; + render->prim_type = prim; + render->num_verts = nr_vertices; + render->num_attribs = CELL_MAX_ATTRIBS; /* XXX fix */ + render->vertex_data = vertices; + render->index_data = indices; + render->num_indexes = nr_indices; + render->xmin = xmin; + render->ymin = ymin; + render->xmax = xmax; + render->ymax = ymax; + + ASSERT_ALIGN16(render->vertex_data); + ASSERT_ALIGN16(render->index_data); + } +#endif +#if 01 /* XXX this is temporary */ - cell_flush(&cell->pipe, 0x0); + cell_flush(&cell->pipe, PIPE_FLUSH_WAIT); +#endif } diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c index 04bb087cf9..0b1e3d46da 100644 --- a/src/mesa/pipe/cell/spu/main.c +++ b/src/mesa/pipe/cell/spu/main.c @@ -115,11 +115,15 @@ really_clear_tiles(uint surfaceIndex) static void -clear_surface(const struct cell_command_clear_surface *clear) +cmd_clear_surface(const struct cell_command_clear_surface *clear) { const uint num_tiles = fb.width_tiles * fb.height_tiles; uint i, j; + if (Debug) + printf("SPU %u: CLEAR SURF %u to 0x%08x\n", init.id, + clear->surface, clear->value); + #define CLEAR_OPT 1 #if CLEAR_OPT /* set all tile's status to CLEAR */ @@ -206,7 +210,7 @@ tile_bounding_box(const struct cell_command_render *render, static void -render(const struct cell_command_render *render) +cmd_render(const struct cell_command_render *render) { /* we'll DMA into these buffers */ ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; @@ -214,6 +218,14 @@ render(const struct cell_command_render *render) uint i, j, vertex_size, vertex_bytes, index_bytes; + if (Debug) + printf("SPU %u: RENDER prim %u, indices: %u, nr_vert: %u\n", + init.id, + render->prim_type, + render->num_verts, + render->num_indexes); + + ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(render->index_data); @@ -321,13 +333,115 @@ render(const struct cell_command_render *render) } +static void +cmd_framebuffer(const struct cell_command_framebuffer *cmd) +{ + if (Debug) + printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", + init.id, + cmd->width, + cmd->height, + cmd->color_start, + cmd->color_format, + cmd->depth_format); + + fb.color_start = cmd->color_start; + fb.depth_start = cmd->depth_start; + fb.color_format = cmd->color_format; + fb.depth_format = cmd->depth_format; + fb.width = cmd->width; + fb.height = cmd->height; + fb.width_tiles = (fb.width + TILE_SIZE - 1) / TILE_SIZE; + fb.height_tiles = (fb.height + TILE_SIZE - 1) / TILE_SIZE; +} + static void -batch(const struct cell_command_batch *batch) +cmd_finish(void) { + if (Debug) + printf("SPU %u: FINISH\n", init.id); + really_clear_tiles(0); + /* wait for all outstanding DMAs to finish */ + mfc_write_tag_mask(~0); + mfc_read_tag_status_all(); + /* send mbox message to PPU */ + spu_write_out_mbox(CELL_CMD_FINISH); } +/** + * Execute a batch of commands + * The opcode param encodes the location of the buffer and its size. + */ +static void +cmd_batch(uint opcode) +{ + const uint buf = (opcode >> 8) & 0xff; + uint size = (opcode >> 16); + uint buffer[CELL_BATCH_BUFFER_SIZE / 4] ALIGN16_ATTRIB; + const uint usize = size / sizeof(uint); + uint pos; + + if (Debug) + printf("SPU %u: BATCH buffer %u, len %u, from %p\n", + init.id, buf, size, init.batch_buffers[buf]); + + ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH); + + ASSERT_ALIGN16(init.batch_buffers[buf]); + + size = (size + 0xf) & ~0xf; + + mfc_get(buffer, /* dest */ + (unsigned int) init.batch_buffers[buf], /* src */ + size, + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); + + for (pos = 0; pos < usize; /* no incr */) { + switch (buffer[pos]) { + case CELL_CMD_FRAMEBUFFER: + { + struct cell_command_framebuffer *fb + = (struct cell_command_framebuffer *) &buffer[pos]; + cmd_framebuffer(fb); + pos += sizeof(*fb) / 4; + } + break; + case CELL_CMD_CLEAR_SURFACE: + { + struct cell_command_clear_surface *clr + = (struct cell_command_clear_surface *) &buffer[pos]; + cmd_clear_surface(clr); + pos += sizeof(*clr) / 4; + } + break; + case CELL_CMD_RENDER: + { + struct cell_command_render *render + = (struct cell_command_render *) &buffer[pos]; + cmd_render(render); + pos += sizeof(*render) / 4; + } + break; + case CELL_CMD_FINISH: + cmd_finish(); + pos += 1; + break; + default: + printf("SPU %u: bad opcode: 0x%x\n", init.id, buffer[pos]); + ASSERT(0); + break; + } + } + + if (Debug) + printf("SPU %u: BATCH complete\n", init.id); +} + /** * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. @@ -355,7 +469,7 @@ main_loop(void) opcode = (unsigned int) spu_read_in_mbox(); if (Debug) - printf("SPU %u: got cmd %u\n", init.id, opcode); + printf("SPU %u: got cmd 0x%x\n", init.id, opcode); /* command payload */ mfc_get(&cmd, /* dest */ @@ -373,62 +487,19 @@ main_loop(void) exitFlag = 1; break; case CELL_CMD_FRAMEBUFFER: - if (Debug) - printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", - init.id, - cmd.fb.width, - cmd.fb.height, - cmd.fb.color_start, - cmd.fb.color_format, - cmd.fb.depth_format); - fb.color_start = cmd.fb.color_start; - fb.depth_start = cmd.fb.depth_start; - fb.color_format = cmd.fb.color_format; - fb.depth_format = cmd.fb.depth_format; - fb.width = cmd.fb.width; - fb.height = cmd.fb.height; - fb.width_tiles = (fb.width + TILE_SIZE - 1) / TILE_SIZE; - fb.height_tiles = (fb.height + TILE_SIZE - 1) / TILE_SIZE; - /* - printf("SPU %u: %u x %u tiles\n", - init.id, fb.width_tiles, fb.height_tiles); - */ + cmd_framebuffer(&cmd.fb); break; case CELL_CMD_CLEAR_SURFACE: - if (Debug) - printf("SPU %u: CLEAR SURF %u to 0x%08x\n", init.id, - cmd.clear.surface, cmd.clear.value); - clear_surface(&cmd.clear); + cmd_clear_surface(&cmd.clear); break; case CELL_CMD_RENDER: - if (Debug) - printf("SPU %u: RENDER prim %u, indices: %u, nr_vert: %u\n", - init.id, - cmd.render.prim_type, - cmd.render.num_verts, - cmd.render.num_indexes); - render(&cmd.render); + cmd_render(&cmd.render); break; - case CELL_CMD_BATCH: - /* execute a batch buffer */ - if (Debug) - printf("SPU %u: BATCH buffer %u, len %u\n", - init.id, - cmd.batch.buffer, - cmd.batch.length); - batch(&cmd.batch); + cmd_batch(opcode); break; - case CELL_CMD_FINISH: - if (Debug) - printf("SPU %u: FINISH\n", init.id); - really_clear_tiles(0); - /* wait for all outstanding DMAs to finish */ - mfc_write_tag_mask(~0); - mfc_read_tag_status_all(); - /* send mbox message to PPU */ - spu_write_out_mbox(CELL_CMD_FINISH); + cmd_finish(); break; default: printf("Bad opcode!\n"); diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h index 9351aa340b..cc70fdd7a6 100644 --- a/src/mesa/pipe/cell/spu/main.h +++ b/src/mesa/pipe/cell/spu/main.h @@ -55,12 +55,12 @@ extern struct framebuffer fb; #define TAG_SURFACE_CLEAR 10 #define TAG_VERTEX_BUFFER 11 -#define TAG_INDEX_BUFFER 16 #define TAG_READ_TILE_COLOR 12 #define TAG_READ_TILE_Z 13 #define TAG_WRITE_TILE_COLOR 14 #define TAG_WRITE_TILE_Z 15 - +#define TAG_INDEX_BUFFER 16 +#define TAG_BATCH_BUFFER 17 /** The standard assert macro doesn't seem to work on SPUs */ #define ASSERT(x) \ -- cgit v1.2.3 From 50eb29ed9492a34db4ba53f1f28a2868b808955a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 16:31:58 -0700 Subject: whitespace, comment changes --- src/mesa/pipe/p_state.h | 30 ++++++++++++++---------------- 1 file changed, 14 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 76e633e930..f9c5bfb6c6 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -56,18 +56,13 @@ #define PIPE_MAX_SHADER_OUTPUTS 16 -/* fwd decl */ +/* fwd decls */ struct pipe_surface; +struct pipe_winsys; /* opaque type */ struct pipe_buffer_handle; -struct pipe_winsys; - - -/*** - *** State objects - ***/ /** @@ -115,6 +110,7 @@ struct pipe_viewport_state { float translate[4]; }; + struct pipe_scissor_state { unsigned minx:16; unsigned miny:16; @@ -122,6 +118,7 @@ struct pipe_scissor_state { unsigned maxy:16; }; + struct pipe_clip_state { float ucp[PIPE_MAX_CLIP_PLANES][4]; unsigned nr; @@ -148,13 +145,14 @@ struct pipe_shader_state { ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; }; + struct pipe_depth_stencil_alpha_state { struct { - unsigned enabled:1; /**< depth test enabled? */ - unsigned writemask:1; /**< allow depth buffer writes? */ - unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ - unsigned occlusion_count:1; /**< XXX move this elsewhere? */ + unsigned enabled:1; /**< depth test enabled? */ + unsigned writemask:1; /**< allow depth buffer writes? */ + unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ + unsigned occlusion_count:1; /**< do occlusion counting? */ } depth; struct { unsigned enabled:1; @@ -165,12 +163,11 @@ struct pipe_depth_stencil_alpha_state ubyte ref_value; ubyte value_mask; ubyte write_mask; - } stencil[2]; /**< [0] = front, [1] = back */ - + } stencil[2]; /**< [0] = front, [1] = back */ struct { unsigned enabled:1; - unsigned func:3; /**< PIPE_FUNC_x */ - float ref; /**< reference value */ + unsigned func:3; /**< PIPE_FUNC_x */ + float ref; /**< reference value */ } alpha; }; @@ -193,10 +190,12 @@ struct pipe_blend_state { unsigned dither:1; }; + struct pipe_blend_color { float color[4]; }; + struct pipe_framebuffer_state { /** multiple colorbuffers for multiple render targets */ @@ -314,5 +313,4 @@ struct pipe_vertex_element }; - #endif -- cgit v1.2.3 From 1c22b5955953973c2c7988bef998f336493e11bc Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 11 Jan 2008 16:32:20 -0700 Subject: s/int/uint/ to silence warnings --- src/mesa/pipe/i965simple/brw_tex_layout.c | 2 +- src/mesa/pipe/i965simple/brw_winsys.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.c b/src/mesa/pipe/i965simple/brw_tex_layout.c index 7d6e2851b1..2b2bf16f1b 100644 --- a/src/mesa/pipe/i965simple/brw_tex_layout.c +++ b/src/mesa/pipe/i965simple/brw_tex_layout.c @@ -237,7 +237,7 @@ static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; int x = 0; int y = 0; - int q, j; + uint q, j; intel_miptree_set_level_info(tex, level, nr_images, 0, tex->total_height, diff --git a/src/mesa/pipe/i965simple/brw_winsys.h b/src/mesa/pipe/i965simple/brw_winsys.h index 49a12a1c27..253599896c 100644 --- a/src/mesa/pipe/i965simple/brw_winsys.h +++ b/src/mesa/pipe/i965simple/brw_winsys.h @@ -190,7 +190,7 @@ static inline boolean brw_batchbuffer_data(struct brw_winsys *winsys, unsigned bytes) { static const unsigned incr = sizeof(unsigned); - int i; + uint i; const unsigned *udata = (const unsigned*)(data); unsigned size = bytes/incr; for (i = 0; i < size; ++i) { -- cgit v1.2.3 From 07276d676c757d3be2e3090d6c67fbbb2f9768eb Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 12 Jan 2008 10:06:27 -0700 Subject: Cell: sketch out needed rasterizer state --- src/mesa/pipe/cell/ppu/cell_state_rasterizer.c | 42 ++++++++++++++++++++++++++ 1 file changed, 42 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c b/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c index 11e7de7309..6b1675af26 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c +++ b/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c @@ -32,6 +32,48 @@ #include "cell_state.h" + +struct spu_rasterizer_state +{ + unsigned flatshade:1; +#if 0 + unsigned light_twoside:1; + unsigned front_winding:2; /**< PIPE_WINDING_x */ + unsigned cull_mode:2; /**< PIPE_WINDING_x */ + unsigned fill_cw:2; /**< PIPE_POLYGON_MODE_x */ + unsigned fill_ccw:2; /**< PIPE_POLYGON_MODE_x */ + unsigned offset_cw:1; + unsigned offset_ccw:1; +#endif + unsigned scissor:1; + unsigned poly_smooth:1; + unsigned poly_stipple_enable:1; + unsigned point_smooth:1; +#if 0 + unsigned point_sprite:1; + unsigned point_size_per_vertex:1; /**< size computed in vertex shader */ +#endif + unsigned multisample:1; /* XXX maybe more ms state in future */ + unsigned line_smooth:1; + unsigned line_stipple_enable:1; + unsigned line_stipple_factor:8; /**< [1..256] actually */ + unsigned line_stipple_pattern:16; +#if 0 + unsigned bypass_clipping:1; +#endif + unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ + + float line_width; + float point_size; /**< used when no per-vertex size */ +#if 0 + float offset_units; + float offset_scale; + ubyte sprite_coord_mode[PIPE_MAX_SHADER_OUTPUTS]; /**< PIPE_SPRITE_COORD_ */ +#endif +}; + + + void * cell_create_rasterizer_state(struct pipe_context *pipe, const struct pipe_rasterizer_state *setup) -- cgit v1.2.3 From a9a84674726a58c183fe8983321208a32cdf940e Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 12 Jan 2008 10:33:24 -0700 Subject: Cell: prefix SPU sources with spu_ --- src/mesa/pipe/cell/spu/Makefile | 19 +- src/mesa/pipe/cell/spu/spu_main.c | 553 +++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_main.h | 78 ++++ src/mesa/pipe/cell/spu/spu_tile.c | 115 +++++ src/mesa/pipe/cell/spu/spu_tile.h | 69 +++ src/mesa/pipe/cell/spu/spu_tri.c | 892 ++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_tri.h | 37 ++ 7 files changed, 1751 insertions(+), 12 deletions(-) create mode 100644 src/mesa/pipe/cell/spu/spu_main.c create mode 100644 src/mesa/pipe/cell/spu/spu_main.h create mode 100644 src/mesa/pipe/cell/spu/spu_tile.c create mode 100644 src/mesa/pipe/cell/spu/spu_tile.h create mode 100644 src/mesa/pipe/cell/spu/spu_tri.c create mode 100644 src/mesa/pipe/cell/spu/spu_tri.h (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index bf97a49e78..b92a756cd5 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -16,15 +16,19 @@ PROG_SPU_EMBED_O = $(PROG)_spu-embed.o SOURCES = \ - main.c \ - tile.c \ - tri.c + spu_main.c \ + spu_tile.c \ + spu_tri.c SPU_OBJECTS = $(SOURCES:.c=.o) \ INCLUDE_DIRS = -I$(TOP)/src/mesa +.c.o: + $(SPU_CC) $(SPU_CFLAGS) -c $< + + # The .a file will be linked into the main/PPU executable default: $(PROG_SPU_A) @@ -38,15 +42,6 @@ $(PROG_SPU): $(SPU_OBJECTS) $(SPU_CC) -o $(PROG_SPU) $(SPU_OBJECTS) $(SPU_LFLAGS) -main.o: main.c - $(SPU_CC) $(SPU_CFLAGS) -c main.c - -tile.o: tile.c - $(SPU_CC) $(SPU_CFLAGS) -c tile.c - -tri.o: tri.c - $(SPU_CC) $(SPU_CFLAGS) -c tri.c - clean: rm -f *~ *.o *.a *.d $(PROG_SPU) diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c new file mode 100644 index 0000000000..df708e65d1 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -0,0 +1,553 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/* main() for Cell SPU code */ + + +#include +#include +#include + +#include "spu_main.h" +#include "spu_tri.h" +#include "spu_tile.h" +#include "pipe/cell/common.h" +#include "pipe/p_defines.h" + +/* +helpful headers: +/usr/lib/gcc/spu/4.1.1/include/spu_mfcio.h +/opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h +*/ + +static boolean Debug = FALSE; + +volatile struct cell_init_info init; + +struct framebuffer fb; + + + +void +wait_on_mask(unsigned tagMask) +{ + mfc_write_tag_mask( tagMask ); + /* wait for completion of _any_ DMAs specified by tagMask */ + mfc_read_tag_status_any(); +} + + +static void +wait_on_mask_all(unsigned tagMask) +{ + mfc_write_tag_mask( tagMask ); + /* wait for completion of _any_ DMAs specified by tagMask */ + mfc_read_tag_status_all(); +} + + + +/** + * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled + * tiles back to the main framebuffer. + */ +static void +really_clear_tiles(uint surfaceIndex) +{ + const uint num_tiles = fb.width_tiles * fb.height_tiles; + uint i, j; + + if (surfaceIndex == 0) { + for (i = 0; i < TILE_SIZE; i++) + for (j = 0; j < TILE_SIZE; j++) + ctile[i][j] = fb.color_clear_value; + + for (i = init.id; i < num_tiles; i += init.num_spus) { + uint tx = i % fb.width_tiles; + uint ty = i / fb.width_tiles; + if (tile_status[ty][tx] == TILE_STATUS_CLEAR) { + put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); + } + } + } + else { + for (i = 0; i < TILE_SIZE; i++) + for (j = 0; j < TILE_SIZE; j++) + ztile[i][j] = fb.depth_clear_value; + + for (i = init.id; i < num_tiles; i += init.num_spus) { + uint tx = i % fb.width_tiles; + uint ty = i / fb.width_tiles; + if (tile_status_z[ty][tx] == TILE_STATUS_CLEAR) + put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 1); + } + } + +#if 0 + wait_on_mask(1 << TAG_SURFACE_CLEAR); +#endif +} + + +static void +cmd_clear_surface(const struct cell_command_clear_surface *clear) +{ + const uint num_tiles = fb.width_tiles * fb.height_tiles; + uint i, j; + + if (Debug) + printf("SPU %u: CLEAR SURF %u to 0x%08x\n", init.id, + clear->surface, clear->value); + +#define CLEAR_OPT 1 +#if CLEAR_OPT + /* set all tile's status to CLEAR */ + if (clear->surface == 0) { + memset(tile_status, TILE_STATUS_CLEAR, sizeof(tile_status)); + fb.color_clear_value = clear->value; + } + else { + memset(tile_status_z, TILE_STATUS_CLEAR, sizeof(tile_status_z)); + fb.depth_clear_value = clear->value; + } + return; +#endif + + if (clear->surface == 0) { + for (i = 0; i < TILE_SIZE; i++) + for (j = 0; j < TILE_SIZE; j++) + ctile[i][j] = clear->value; + } + else { + for (i = 0; i < TILE_SIZE; i++) + for (j = 0; j < TILE_SIZE; j++) + ztile[i][j] = clear->value; + } + + /* + printf("SPU: %s num=%d w=%d h=%d\n", + __FUNCTION__, num_tiles, fb.width_tiles, fb.height_tiles); + */ + + for (i = init.id; i < num_tiles; i += init.num_spus) { + uint tx = i % fb.width_tiles; + uint ty = i / fb.width_tiles; + if (clear->surface == 0) + put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); + else + put_tile(&fb, tx, ty, (uint *) ztile, TAG_SURFACE_CLEAR, 1); + /* XXX we don't want this here, but it fixes bad tile results */ + } + +#if 0 + wait_on_mask(1 << TAG_SURFACE_CLEAR); +#endif +} + + +/** + * Given a rendering command's bounding box (in pixels) compute the + * location of the corresponding screen tile bounding box. + */ +static INLINE void +tile_bounding_box(const struct cell_command_render *render, + uint *txmin, uint *tymin, + uint *box_num_tiles, uint *box_width_tiles) +{ +#if 1 + /* Debug: full-window bounding box */ + uint txmax = fb.width_tiles - 1; + uint tymax = fb.height_tiles - 1; + *txmin = 0; + *tymin = 0; + *box_num_tiles = fb.width_tiles * fb.height_tiles; + *box_width_tiles = fb.width_tiles; + (void) render; + (void) txmax; + (void) tymax; +#else + uint txmax, tymax, box_height_tiles; + + *txmin = (uint) render->xmin / TILE_SIZE; + *tymin = (uint) render->ymin / TILE_SIZE; + txmax = (uint) render->xmax / TILE_SIZE; + tymax = (uint) render->ymax / TILE_SIZE; + *box_width_tiles = txmax - *txmin + 1; + box_height_tiles = tymax - *tymin + 1; + *box_num_tiles = *box_width_tiles * box_height_tiles; +#endif +#if 0 + printf("Render bounds: %g, %g ... %g, %g\n", + render->xmin, render->ymin, render->xmax, render->ymax); + printf("Render tiles: %u, %u .. %u, %u\n", *txmin, *tymin, txmax, tymax); +#endif +} + + +static void +cmd_render(const struct cell_command_render *render) +{ + /* we'll DMA into these buffers */ + ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; + ushort indexes[CELL_MAX_VBUF_INDEXES] ALIGN16_ATTRIB; + + uint i, j, vertex_size, vertex_bytes, index_bytes; + + if (Debug) + printf("SPU %u: RENDER prim %u, indices: %u, nr_vert: %u\n", + init.id, + render->prim_type, + render->num_verts, + render->num_indexes); + + + ASSERT_ALIGN16(render->vertex_data); + ASSERT_ALIGN16(render->index_data); + + vertex_size = render->num_attribs * 4 * sizeof(float); + + /* how much vertex data */ + vertex_bytes = render->num_verts * vertex_size; + index_bytes = render->num_indexes * sizeof(ushort); + if (index_bytes < 8) + index_bytes = 8; + else + index_bytes = (index_bytes + 15) & ~0xf; /* multiple of 16 */ + + /* + printf("VBUF: indices at %p, vertices at %p vertex_bytes %u ind_bytes %u\n", + render->index_data, render->vertex_data, vertex_bytes, index_bytes); + */ + + /* get vertex data from main memory */ + mfc_get(vertex_data, /* dest */ + (unsigned int) render->vertex_data, /* src */ + vertex_bytes, /* size */ + TAG_VERTEX_BUFFER, + 0, /* tid */ + 0 /* rid */); + + /* get index data from main memory */ + mfc_get(indexes, /* dest */ + (unsigned int) render->index_data, /* src */ + index_bytes, + TAG_INDEX_BUFFER, + 0, /* tid */ + 0 /* rid */); + + wait_on_mask_all((1 << TAG_VERTEX_BUFFER) | + (1 << TAG_INDEX_BUFFER)); + + /* find tiles which intersect the prim bounding box */ + uint txmin, tymin, box_width_tiles, box_num_tiles; +#if 0 + tile_bounding_box(render, &txmin, &tymin, + &box_num_tiles, &box_width_tiles); +#else + txmin = 0; + tymin = 0; + box_num_tiles = fb.width_tiles * fb.height_tiles; + box_width_tiles = fb.width_tiles; +#endif + + /* make sure any pending clears have completed */ + wait_on_mask(1 << TAG_SURFACE_CLEAR); + + /* loop over tiles */ + for (i = init.id; i < box_num_tiles; i += init.num_spus) { + const uint tx = txmin + i % box_width_tiles; + const uint ty = tymin + i / box_width_tiles; + + ASSERT(tx < fb.width_tiles); + ASSERT(ty < fb.height_tiles); + + /* Start fetching color/z tiles. We'll wait for completion when + * we need read/write to them later in triangle rasterization. + */ + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(&fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); + } + } + + if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); + } + + ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); + + /* loop over tris */ + for (j = 0; j < render->num_indexes; j += 3) { + const float *v0, *v1, *v2; + + v0 = (const float *) (vertex_data + indexes[j+0] * vertex_size); + v1 = (const float *) (vertex_data + indexes[j+1] * vertex_size); + v2 = (const float *) (vertex_data + indexes[j+2] * vertex_size); + + tri_draw(v0, v1, v2, tx, ty); + } + + /* write color/z tiles back to main framebuffer, if dirtied */ + if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(&fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); + tile_status[ty][tx] = TILE_STATUS_DEFINED; + } + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(&fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); + tile_status_z[ty][tx] = TILE_STATUS_DEFINED; + } + } + + /* XXX move these... */ + wait_on_mask(1 << TAG_WRITE_TILE_COLOR); + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + wait_on_mask(1 << TAG_WRITE_TILE_Z); + } + } +} + + +static void +cmd_framebuffer(const struct cell_command_framebuffer *cmd) +{ + if (Debug) + printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", + init.id, + cmd->width, + cmd->height, + cmd->color_start, + cmd->color_format, + cmd->depth_format); + + fb.color_start = cmd->color_start; + fb.depth_start = cmd->depth_start; + fb.color_format = cmd->color_format; + fb.depth_format = cmd->depth_format; + fb.width = cmd->width; + fb.height = cmd->height; + fb.width_tiles = (fb.width + TILE_SIZE - 1) / TILE_SIZE; + fb.height_tiles = (fb.height + TILE_SIZE - 1) / TILE_SIZE; +} + + +static void +cmd_finish(void) +{ + if (Debug) + printf("SPU %u: FINISH\n", init.id); + really_clear_tiles(0); + /* wait for all outstanding DMAs to finish */ + mfc_write_tag_mask(~0); + mfc_read_tag_status_all(); + /* send mbox message to PPU */ + spu_write_out_mbox(CELL_CMD_FINISH); +} + + +/** + * Execute a batch of commands + * The opcode param encodes the location of the buffer and its size. + */ +static void +cmd_batch(uint opcode) +{ + const uint buf = (opcode >> 8) & 0xff; + uint size = (opcode >> 16); + uint buffer[CELL_BATCH_BUFFER_SIZE / 4] ALIGN16_ATTRIB; + const uint usize = size / sizeof(uint); + uint pos; + + if (Debug) + printf("SPU %u: BATCH buffer %u, len %u, from %p\n", + init.id, buf, size, init.batch_buffers[buf]); + + ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH); + + ASSERT_ALIGN16(init.batch_buffers[buf]); + + size = (size + 0xf) & ~0xf; + + mfc_get(buffer, /* dest */ + (unsigned int) init.batch_buffers[buf], /* src */ + size, + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); + + for (pos = 0; pos < usize; /* no incr */) { + switch (buffer[pos]) { + case CELL_CMD_FRAMEBUFFER: + { + struct cell_command_framebuffer *fb + = (struct cell_command_framebuffer *) &buffer[pos]; + cmd_framebuffer(fb); + pos += sizeof(*fb) / 4; + } + break; + case CELL_CMD_CLEAR_SURFACE: + { + struct cell_command_clear_surface *clr + = (struct cell_command_clear_surface *) &buffer[pos]; + cmd_clear_surface(clr); + pos += sizeof(*clr) / 4; + } + break; + case CELL_CMD_RENDER: + { + struct cell_command_render *render + = (struct cell_command_render *) &buffer[pos]; + cmd_render(render); + pos += sizeof(*render) / 4; + } + break; + case CELL_CMD_FINISH: + cmd_finish(); + pos += 1; + break; + default: + printf("SPU %u: bad opcode: 0x%x\n", init.id, buffer[pos]); + ASSERT(0); + break; + } + } + + if (Debug) + printf("SPU %u: BATCH complete\n", init.id); +} + + +/** + * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. + */ +static void +main_loop(void) +{ + struct cell_command cmd; + int exitFlag = 0; + + if (Debug) + printf("SPU %u: Enter main loop\n", init.id); + + ASSERT((sizeof(struct cell_command) & 0xf) == 0); + ASSERT_ALIGN16(&cmd); + + while (!exitFlag) { + unsigned opcode; + int tag = 0; + + if (Debug) + printf("SPU %u: Wait for cmd...\n", init.id); + + /* read/wait from mailbox */ + opcode = (unsigned int) spu_read_in_mbox(); + + if (Debug) + printf("SPU %u: got cmd 0x%x\n", init.id, opcode); + + /* command payload */ + mfc_get(&cmd, /* dest */ + (unsigned int) init.cmd, /* src */ + sizeof(struct cell_command), /* bytes */ + tag, + 0, /* tid */ + 0 /* rid */); + wait_on_mask( 1 << tag ); + + switch (opcode & CELL_CMD_OPCODE_MASK) { + case CELL_CMD_EXIT: + if (Debug) + printf("SPU %u: EXIT\n", init.id); + exitFlag = 1; + break; + case CELL_CMD_FRAMEBUFFER: + cmd_framebuffer(&cmd.fb); + break; + case CELL_CMD_CLEAR_SURFACE: + cmd_clear_surface(&cmd.clear); + break; + case CELL_CMD_RENDER: + cmd_render(&cmd.render); + break; + case CELL_CMD_BATCH: + cmd_batch(opcode); + break; + case CELL_CMD_FINISH: + cmd_finish(); + break; + default: + printf("Bad opcode!\n"); + } + + } + + if (Debug) + printf("SPU %u: Exit main loop\n", init.id); +} + + + +static void +one_time_init(void) +{ + memset(tile_status, TILE_STATUS_DEFINED, sizeof(tile_status)); + memset(tile_status_z, TILE_STATUS_DEFINED, sizeof(tile_status_z)); +} + + +/** + * SPE entrypoint. + * Note: example programs declare params as 'unsigned long long' but + * that doesn't work. + */ +int +main(unsigned long speid, unsigned long argp) +{ + int tag = 0; + + (void) speid; + + one_time_init(); + + if (Debug) + printf("SPU: main() speid=%lu\n", speid); + + mfc_get(&init, /* dest */ + (unsigned int) argp, /* src */ + sizeof(struct cell_init_info), /* bytes */ + tag, + 0, /* tid */ + 0 /* rid */); + wait_on_mask( 1 << tag ); + + + main_loop(); + + return 0; +} diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h new file mode 100644 index 0000000000..9ef8cf0709 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -0,0 +1,78 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SPU_MAIN_H +#define SPU_MAIN_H + + +#include "pipe/cell/common.h" + + +extern volatile struct cell_init_info init; + +struct framebuffer { + void *color_start; /**< addr of color surface in main memory */ + void *depth_start; /**< addr of depth surface in main memory */ + enum pipe_format color_format; + enum pipe_format depth_format; + uint width, height; /**< size in pixels */ + uint width_tiles, height_tiles; /**< width and height in tiles */ + + uint color_clear_value; + uint depth_clear_value; +}; + +/* XXX Collect these globals in a struct: */ + +extern struct framebuffer fb; + + +/* DMA TAGS */ + +#define TAG_SURFACE_CLEAR 10 +#define TAG_VERTEX_BUFFER 11 +#define TAG_READ_TILE_COLOR 12 +#define TAG_READ_TILE_Z 13 +#define TAG_WRITE_TILE_COLOR 14 +#define TAG_WRITE_TILE_Z 15 +#define TAG_INDEX_BUFFER 16 +#define TAG_BATCH_BUFFER 17 + +/** The standard assert macro doesn't seem to work on SPUs */ +#define ASSERT(x) \ + if (!(x)) { \ + fprintf(stderr, "SPU %d: %s:%d: %s(): assertion %s failed.\n", \ + init.id, __FILE__, __LINE__, __FUNCTION__, #x); \ + exit(1); \ + } + + +void +wait_on_mask(unsigned tag); + + +#endif /* SPU_MAIN_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c new file mode 100644 index 0000000000..8fd9ab2905 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_tile.c @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + + +#include "spu_tile.h" + + + +uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; + +ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; +ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + + + +void +get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, + int tag, int zBuf) +{ + const uint offset = ty * fb->width_tiles + tx; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); + const ubyte *src = zBuf ? fb->depth_start : fb->color_start; + + src += offset * bytesPerTile; + + ASSERT(tx < fb->width_tiles); + ASSERT(ty < fb->height_tiles); + ASSERT_ALIGN16(tile); + /* + printf("get_tile: dest: %p src: 0x%x size: %d\n", + tile, (unsigned int) src, bytesPerTile); + */ + mfc_get(tile, /* dest in local memory */ + (unsigned int) src, /* src in main memory */ + bytesPerTile, + tag, + 0, /* tid */ + 0 /* rid */); +} + + +void +put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, + int tag, int zBuf) +{ + const uint offset = ty * fb->width_tiles + tx; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); + ubyte *dst = zBuf ? fb->depth_start : fb->color_start; + + dst += offset * bytesPerTile; + + ASSERT(tx < fb->width_tiles); + ASSERT(ty < fb->height_tiles); + ASSERT_ALIGN16(tile); + /* + printf("put_tile: src: %p dst: 0x%x size: %d\n", + tile, (unsigned int) dst, bytesPerTile); + */ + mfc_put((void *) tile, /* src in local memory */ + (unsigned int) dst, /* dst in main memory */ + bytesPerTile, + tag, + 0, /* tid */ + 0 /* rid */); +} + + +void +clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value) +{ + uint i, j; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile[i][j] = value; + } + } +} + +void +clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value) +{ + uint i, j; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile[i][j] = value; + } + } +} + diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h new file mode 100644 index 0000000000..2ad469b584 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -0,0 +1,69 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SPU_TILE_H +#define SPU_TILE_H + + +#include +#include +#include "spu_main.h" +#include "pipe/cell/common.h" + + +#define MAX_WIDTH 1024 +#define MAX_HEIGHT 1024 + + +extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; + + +#define TILE_STATUS_CLEAR 1 +#define TILE_STATUS_DEFINED 2 /**< defined pixel data */ +#define TILE_STATUS_DIRTY 3 /**< modified, but not put back yet */ + +extern ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; +extern ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + + +void +get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, + int tag, int zBuf); + +void +put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, + int tag, int zBuf); + +void +clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value); + +void +clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value); + + +#endif /* SPU_TILE_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c new file mode 100644 index 0000000000..af24e42435 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -0,0 +1,892 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Triangle rendering within a tile. + */ + +#include "pipe/p_compiler.h" +#include "pipe/p_format.h" +#include "pipe/p_util.h" +#include "spu_main.h" +#include "spu_tile.h" +#include "spu_tri.h" + + + +/** + * Simplified types taken from other parts of Gallium + */ + +struct vertex_header { + float data[2][4]; /* pos and color */ +}; + +struct prim_header { + struct vertex_header *v[3]; +}; + + + + +#if 1 + +/* XXX fix this */ +#undef CEILF +#define CEILF(X) ((float) (int) ((X) + 0.99999)) + + +#define QUAD_TOP_LEFT 0 +#define QUAD_TOP_RIGHT 1 +#define QUAD_BOTTOM_LEFT 2 +#define QUAD_BOTTOM_RIGHT 3 +#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) +#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) +#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) +#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) +#define MASK_ALL 0xf + +#define PIPE_MAX_SHADER_INPUTS 8 /* XXX temp */ + +#endif + + +#define DEBUG_VERTS 0 + +/** + * Triangle edge info + */ +struct edge { + float dx; /**< X(v1) - X(v0), used only during setup */ + float dy; /**< Y(v1) - Y(v0), used only during setup */ + float dxdy; /**< dx/dy */ + float sx, sy; /**< first sample point coord */ + int lines; /**< number of lines on this edge */ +}; + + +struct interp_coef +{ + float a0[4]; + float dadx[4]; + float dady[4]; +}; + +/** + * Triangle setup info (derived from draw_stage). + * Also used for line drawing (taking some liberties). + */ +struct setup_stage { + + /* Vertices are just an array of floats making up each attribute in + * turn. Currently fixed at 4 floats, but should change in time. + * Codegen will help cope with this. + */ + const struct vertex_header *vmax; + const struct vertex_header *vmid; + const struct vertex_header *vmin; + const struct vertex_header *vprovoke; + + struct edge ebot; + struct edge etop; + struct edge emaj; + + float oneoverarea; + + uint tx, ty; + + int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; + +#if 0 + struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; +#else + struct interp_coef coef[PIPE_MAX_SHADER_INPUTS]; +#endif + +#if 0 + struct quad_header quad; +#endif + + struct { + int left[2]; /**< [0] = row0, [1] = row1 */ + int right[2]; + int y; + unsigned y_flags; + unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ + } span; +}; + + +#if 0 +/** + * Basically a cast wrapper. + */ +static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) +{ + return (struct setup_stage *)stage; +} +#endif + +#if 0 +/** + * Clip setup->quad against the scissor/surface bounds. + */ +static INLINE void +quad_clip(struct setup_stage *setup) +{ + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + + if (setup->quad.x0 >= maxx || + setup->quad.y0 >= maxy || + setup->quad.x0 + 1 < minx || + setup->quad.y0 + 1 < miny) { + /* totally clipped */ + setup->quad.mask = 0x0; + return; + } + if (setup->quad.x0 < minx) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + if (setup->quad.y0 < miny) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + if (setup->quad.x0 == maxx - 1) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + if (setup->quad.y0 == maxy - 1) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); +} +#endif + +#if 0 +/** + * Emit a quad (pass to next stage) with clipping. + */ +static INLINE void +clip_emit_quad(struct setup_stage *setup) +{ + quad_clip(setup); + if (setup->quad.mask) { + struct softpipe_context *sp = setup->softpipe; + sp->quad.first->run(sp->quad.first, &setup->quad); + } +} +#endif + +/** + * Evaluate attribute coefficients (plane equations) to compute + * attribute values for the four fragments in a quad. + * Eg: four colors will be compute. + */ +static INLINE void +eval_coeff( struct setup_stage *setup, uint slot, + float x, float y, float result[4][4]) +{ + uint i; + const float *dadx = setup->coef[slot].dadx; + const float *dady = setup->coef[slot].dady; + + /* loop over XYZW comps */ + for (i = 0; i < 4; i++) { + result[QUAD_TOP_LEFT][i] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; + result[QUAD_TOP_RIGHT][i] = result[0][i] + dadx[i]; + result[QUAD_BOTTOM_LEFT][i] = result[0][i] + dady[i]; + result[QUAD_BOTTOM_RIGHT][i] = result[0][i] + dadx[i] + dady[i]; + } +} + + +static INLINE void +eval_z( struct setup_stage *setup, + float x, float y, float result[4]) +{ + uint slot = 0; + uint i = 2; + const float *dadx = setup->coef[slot].dadx; + const float *dady = setup->coef[slot].dady; + + result[QUAD_TOP_LEFT] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; + result[QUAD_TOP_RIGHT] = result[0] + dadx[i]; + result[QUAD_BOTTOM_LEFT] = result[0] + dady[i]; + result[QUAD_BOTTOM_RIGHT] = result[0] + dadx[i] + dady[i]; +} + + +static INLINE uint +pack_color(const float color[4]) +{ + uint r = (uint) (color[0] * 255.0); + uint g = (uint) (color[1] * 255.0); + uint b = (uint) (color[2] * 255.0); + uint a = (uint) (color[3] * 255.0); + switch (fb.color_format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + return (a << 24) | (r << 16) | (g << 8) | b; + case PIPE_FORMAT_B8G8R8A8_UNORM: + return (b << 24) | (g << 16) | (r << 8) | a; + default: + ASSERT(0); + return 0; + } +} + + +/** + * Emit a quad (pass to next stage). No clipping is done. + */ +static INLINE void +emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) +{ +#if 0 + struct softpipe_context *sp = setup->softpipe; + setup->quad.x0 = x; + setup->quad.y0 = y; + setup->quad.mask = mask; + sp->quad.first->run(sp->quad.first, &setup->quad); +#else + /* Cell: "write" quad fragments to the tile by setting prim color */ + int ix = x - setup->cliprect_minx; + int iy = y - setup->cliprect_miny; + float colors[4][4]; + uint z; + + eval_coeff(setup, 1, (float) x, (float) y, colors); + + if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + float zvals[4]; + eval_z(setup, (float) x, (float) y, zvals); + + if (tile_status_z[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { + /* now, _really_ clear the tile */ + clear_tile_z(ztile, fb.depth_clear_value); + } + else { + /* make sure we've got the tile from main mem */ + wait_on_mask(1 << TAG_READ_TILE_Z); + } + tile_status_z[setup->ty][setup->tx] = TILE_STATUS_DIRTY; + + if (mask & MASK_TOP_LEFT) { + z = (uint) (zvals[0] * 65535.0); + if (z < ztile[iy][ix]) + ztile[iy][ix] = z; + else + mask &= ~MASK_TOP_LEFT; + } + + if (mask & MASK_TOP_RIGHT) { + z = (uint) (zvals[1] * 65535.0); + if (z < ztile[iy][ix+1]) + ztile[iy][ix+1] = z; + else + mask &= ~MASK_TOP_RIGHT; + } + + if (mask & MASK_BOTTOM_LEFT) { + z = (uint) (zvals[2] * 65535.0); + if (z < ztile[iy+1][ix]) + ztile[iy+1][ix] = z; + else + mask &= ~MASK_BOTTOM_LEFT; + } + + if (mask & MASK_BOTTOM_RIGHT) { + z = (uint) (zvals[3] * 65535.0); + if (z < ztile[iy+1][ix+1]) + ztile[iy+1][ix+1] = z; + else + mask &= ~MASK_BOTTOM_RIGHT; + } + } + + if (mask) { + if (tile_status[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { + /* now, _really_ clear the tile */ + clear_tile(ctile, fb.color_clear_value); + } + else { + /* make sure we've got the tile from main mem */ + wait_on_mask(1 << TAG_READ_TILE_COLOR); + } + tile_status[setup->ty][setup->tx] = TILE_STATUS_DIRTY; + + if (mask & MASK_TOP_LEFT) + ctile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); + if (mask & MASK_TOP_RIGHT) + ctile[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); + if (mask & MASK_BOTTOM_LEFT) + ctile[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); + if (mask & MASK_BOTTOM_RIGHT) + ctile[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); + } +#endif +} + + +/** + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. + */ +static INLINE int block( int x ) +{ + return x & ~1; +} + + +/** + * Compute mask which indicates which pixels in the 2x2 quad are actually inside + * the triangle's bounds. + * + * this is pretty nasty... may need to rework flush_spans again to + * fix it, if possible. + */ +static unsigned calculate_mask( struct setup_stage *setup, int x ) +{ + unsigned mask = 0x0; + + if (x >= setup->span.left[0] && x < setup->span.right[0]) + mask |= MASK_TOP_LEFT; + + if (x >= setup->span.left[1] && x < setup->span.right[1]) + mask |= MASK_BOTTOM_LEFT; + + if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) + mask |= MASK_TOP_RIGHT; + + if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) + mask |= MASK_BOTTOM_RIGHT; + + return mask; +} + + +/** + * Render a horizontal span of quads + */ +static void flush_spans( struct setup_stage *setup ) +{ + int minleft, maxright; + int x; + + switch (setup->span.y_flags) { + case 0x3: + /* both odd and even lines written (both quad rows) */ + minleft = MIN2(setup->span.left[0], setup->span.left[1]); + maxright = MAX2(setup->span.right[0], setup->span.right[1]); + break; + + case 0x1: + /* only even line written (quad top row) */ + minleft = setup->span.left[0]; + maxright = setup->span.right[0]; + break; + + case 0x2: + /* only odd line written (quad bottom row) */ + minleft = setup->span.left[1]; + maxright = setup->span.right[1]; + break; + + default: + return; + } + + /* XXX this loop could be moved into the above switch cases and + * calculate_mask() could be simplified a bit... + */ + for (x = block(minleft); x <= block(maxright); x += 2) { + emit_quad( setup, x, setup->span.y, + calculate_mask( setup, x ) ); + } + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +} + +#if DEBUG_VERTS +static void print_vertex(const struct setup_stage *setup, + const struct vertex_header *v) +{ + int i; + fprintf(stderr, "Vertex: (%p)\n", v); + for (i = 0; i < setup->quad.nr_attrs; i++) { + fprintf(stderr, " %d: %f %f %f %f\n", i, + v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); + } +} +#endif + +static boolean setup_sort_vertices( struct setup_stage *setup, + const struct prim_header *prim ) +{ + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + const struct vertex_header *v2 = prim->v[2]; + +#if DEBUG_VERTS + fprintf(stderr, "Triangle:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); + print_vertex(setup, v2); +#endif + + setup->vprovoke = v2; + + /* determine bottom to top order of vertices */ + { + float y0 = v0->data[0][1]; + float y1 = v1->data[0][1]; + float y2 = v2->data[0][1]; + if (y0 <= y1) { + if (y1 <= y2) { + /* y0<=y1<=y2 */ + setup->vmin = v0; + setup->vmid = v1; + setup->vmax = v2; + } + else if (y2 <= y0) { + /* y2<=y0<=y1 */ + setup->vmin = v2; + setup->vmid = v0; + setup->vmax = v1; + } + else { + /* y0<=y2<=y1 */ + setup->vmin = v0; + setup->vmid = v2; + setup->vmax = v1; + } + } + else { + if (y0 <= y2) { + /* y1<=y0<=y2 */ + setup->vmin = v1; + setup->vmid = v0; + setup->vmax = v2; + } + else if (y2 <= y1) { + /* y2<=y1<=y0 */ + setup->vmin = v2; + setup->vmid = v1; + setup->vmax = v0; + } + else { + /* y1<=y2<=y0 */ + setup->vmin = v1; + setup->vmid = v2; + setup->vmax = v0; + } + } + } + + /* Check if triangle is completely outside the tile bounds */ + if (setup->vmin->data[0][1] > setup->cliprect_maxy) + return FALSE; + if (setup->vmax->data[0][1] < setup->cliprect_miny) + return FALSE; + if (setup->vmin->data[0][0] < setup->cliprect_minx && + setup->vmid->data[0][0] < setup->cliprect_minx && + setup->vmax->data[0][0] < setup->cliprect_minx) + return FALSE; + if (setup->vmin->data[0][0] > setup->cliprect_maxx && + setup->vmid->data[0][0] > setup->cliprect_maxx && + setup->vmax->data[0][0] > setup->cliprect_maxx) + return FALSE; + + setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; + setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; + setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; + + /* + * Compute triangle's area. Use 1/area to compute partial + * derivatives of attributes later. + * + * The area will be the same as prim->det, but the sign may be + * different depending on how the vertices get sorted above. + * + * To determine whether the primitive is front or back facing we + * use the prim->det value because its sign is correct. + */ + { + const float area = (setup->emaj.dx * setup->ebot.dy - + setup->ebot.dx * setup->emaj.dy); + + setup->oneoverarea = 1.0f / area; + /* + _mesa_printf("%s one-over-area %f area %f det %f\n", + __FUNCTION__, setup->oneoverarea, area, prim->det ); + */ + } + +#if 0 + /* We need to know if this is a front or back-facing triangle for: + * - the GLSL gl_FrontFacing fragment attribute (bool) + * - two-sided stencil test + */ + setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); +#endif + + return TRUE; +} + + +#if 0 +/** + * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + * The value value comes from vertex->data[slot][i]. + * The result will be put into setup->coef[slot].a0[i]. + * \param slot which attribute slot + * \param i which component of the slot (0..3) + */ +static void const_coeff( struct setup_stage *setup, + unsigned slot, + unsigned i ) +{ + assert(slot < PIPE_MAX_SHADER_INPUTS); + assert(i <= 3); + + setup->coef[slot].dadx[i] = 0; + setup->coef[slot].dady[i] = 0; + + /* need provoking vertex info! + */ + setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; +} +#endif + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a triangle. + */ +static void tri_linear_coeff( struct setup_stage *setup, + uint slot, uint firstComp, uint lastComp ) +{ + uint i; + for (i = firstComp; i < lastComp; i++) { + float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; + float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + ASSERT(slot < PIPE_MAX_SHADER_INPUTS); + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); + } + + /* + _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", + slot, "xyzw"[i], + setup->coef[slot].a0[i], + setup->coef[slot].dadx[i], + setup->coef[slot].dady[i]); + */ +} + + +#if 0 +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a triangle. + * We basically multiply the vertex value by 1/w before computing + * the plane coefficients (a0, dadx, dady). + * Later, when we compute the value at a particular fragment position we'll + * divide the interpolated value by the interpolated W at that fragment. + */ +static void tri_persp_coeff( struct setup_stage *setup, + unsigned slot, + unsigned i ) +{ + /* premultiply by 1/w: + */ + float mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; + float mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; + float maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; + + float botda = mida - mina; + float majda = maxa - mina; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + + /* + printf("tri persp %d,%d: %f %f %f\n", slot, i, + setup->vmin->data[slot][i], + setup->vmid->data[slot][i], + setup->vmax->data[slot][i] + ); + */ + + assert(slot < PIPE_MAX_SHADER_INPUTS); + assert(i <= 3); + + setup->coef[slot].dadx[i] = a * setup->oneoverarea; + setup->coef[slot].dady[i] = b * setup->oneoverarea; + setup->coef[slot].a0[i] = (mina - + (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + + setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); +} +#endif + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. + */ +static void setup_tri_coefficients( struct setup_stage *setup ) +{ +#if 0 + const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + unsigned slot, j; + + /* z and w are done by linear interpolation: + */ + tri_linear_coeff(setup, 0, 2); + tri_linear_coeff(setup, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (slot = 1; slot < setup->quad.nr_attrs; slot++) { + switch (interp[slot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, slot, j); + break; + + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + tri_linear_coeff(setup, slot, j); + break; + + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + tri_persp_coeff(setup, slot, j); + break; + + default: + /* invalid interp mode */ + assert(0); + } + } +#else + tri_linear_coeff(setup, 0, 2, 3); /* slot 0, z */ + tri_linear_coeff(setup, 1, 0, 4); /* slot 1, color */ +#endif +} + + +static void setup_tri_edges( struct setup_stage *setup ) +{ + float vmin_x = setup->vmin->data[0][0] + 0.5f; + float vmid_x = setup->vmid->data[0][0] + 0.5f; + + float vmin_y = setup->vmin->data[0][1] - 0.5f; + float vmid_y = setup->vmid->data[0][1] - 0.5f; + float vmax_y = setup->vmax->data[0][1] - 0.5f; + + setup->emaj.sy = CEILF(vmin_y); + setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); + setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; + setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; + + setup->etop.sy = CEILF(vmid_y); + setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); + setup->etop.dxdy = setup->etop.dx / setup->etop.dy; + setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; + + setup->ebot.sy = CEILF(vmin_y); + setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); + setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; + setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; +} + + +/** + * Render the upper or lower half of a triangle. + * Scissoring/cliprect is applied here too. + */ +static void subtriangle( struct setup_stage *setup, + struct edge *eleft, + struct edge *eright, + unsigned lines ) +{ + const int minx = setup->cliprect_minx; + const int maxx = setup->cliprect_maxx; + const int miny = setup->cliprect_miny; + const int maxy = setup->cliprect_maxy; + int y, start_y, finish_y; + int sy = (int)eleft->sy; + + ASSERT((int)eleft->sy == (int) eright->sy); + + /* clip top/bottom */ + start_y = sy; + finish_y = sy + lines; + + if (start_y < miny) + start_y = miny; + + if (finish_y > maxy) + finish_y = maxy; + + start_y -= sy; + finish_y -= sy; + + /* + _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + */ + + for (y = start_y; y < finish_y; y++) { + + /* avoid accumulating adds as floats don't have the precision to + * accurately iterate large triangle edges that way. luckily we + * can just multiply these days. + * + * this is all drowned out by the attribute interpolation anyway. + */ + int left = (int)(eleft->sx + y * eleft->dxdy); + int right = (int)(eright->sx + y * eright->dxdy); + + /* clip left/right */ + if (left < minx) + left = minx; + if (right > maxx) + right = maxx; + + if (left < right) { + int _y = sy + y; + if (block(_y) != setup->span.y) { + flush_spans(setup); + setup->span.y = block(_y); + } + + setup->span.left[_y&1] = left; + setup->span.right[_y&1] = right; + setup->span.y_flags |= 1<<(_y&1); + } + } + + + /* save the values so that emaj can be restarted: + */ + eleft->sx += lines * eleft->dxdy; + eright->sx += lines * eright->dxdy; + eleft->sy += lines; + eright->sy += lines; +} + + +/** + * Do setup for triangle rasterization, then render the triangle. + */ +static void +setup_tri(struct setup_stage *setup, struct prim_header *prim) +{ + if (!setup_sort_vertices( setup, prim )) { + return; /* totally clipped */ + } + + setup_tri_coefficients( setup ); + setup_tri_edges( setup ); + +#if 0 + setup->quad.prim = PRIM_TRI; +#endif + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; + /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ + + /* init_constant_attribs( setup ); */ + + if (setup->oneoverarea < 0.0) { + /* emaj on left: + */ + subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); + subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); + } + else { + /* emaj on right: + */ + subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); + subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); + } + + flush_spans( setup ); +} + + + +/** + * Draw triangle into tile at (tx, ty) (tile coords) + * The tile data should have already been fetched. + */ +void +tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) +{ + struct prim_header tri; + struct setup_stage setup; + + tri.v[0] = (struct vertex_header *) v0; + tri.v[1] = (struct vertex_header *) v1; + tri.v[2] = (struct vertex_header *) v2; + + setup.tx = tx; + setup.ty = ty; + + /* set clipping bounds to tile bounds */ + setup.cliprect_minx = tx * TILE_SIZE; + setup.cliprect_miny = ty * TILE_SIZE; + setup.cliprect_maxx = (tx + 1) * TILE_SIZE; + setup.cliprect_maxy = (ty + 1) * TILE_SIZE; + + setup_tri(&setup, &tri); +} diff --git a/src/mesa/pipe/cell/spu/spu_tri.h b/src/mesa/pipe/cell/spu/spu_tri.h new file mode 100644 index 0000000000..86c42b6339 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_tri.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SPU_TRI_H +#define SPU_TRI_H + + +extern void +tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty); + + +#endif /* SPU_TRI_H */ -- cgit v1.2.3 From 2e469775b3f52adf496802f286e46ac4a31cdba1 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 12 Jan 2008 10:33:50 -0700 Subject: Cell: prefix SPU files with spu_ --- src/mesa/pipe/cell/spu/main.c | 553 -------------------------- src/mesa/pipe/cell/spu/main.h | 78 ---- src/mesa/pipe/cell/spu/tile.c | 115 ------ src/mesa/pipe/cell/spu/tile.h | 69 ---- src/mesa/pipe/cell/spu/tri.c | 892 ------------------------------------------ src/mesa/pipe/cell/spu/tri.h | 37 -- 6 files changed, 1744 deletions(-) delete mode 100644 src/mesa/pipe/cell/spu/main.c delete mode 100644 src/mesa/pipe/cell/spu/main.h delete mode 100644 src/mesa/pipe/cell/spu/tile.c delete mode 100644 src/mesa/pipe/cell/spu/tile.h delete mode 100644 src/mesa/pipe/cell/spu/tri.c delete mode 100644 src/mesa/pipe/cell/spu/tri.h (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/main.c b/src/mesa/pipe/cell/spu/main.c deleted file mode 100644 index 0b1e3d46da..0000000000 --- a/src/mesa/pipe/cell/spu/main.c +++ /dev/null @@ -1,553 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/* main() for Cell SPU code */ - - -#include -#include -#include - -#include "main.h" -#include "tri.h" -#include "tile.h" -#include "pipe/cell/common.h" -#include "pipe/p_defines.h" - -/* -helpful headers: -/usr/lib/gcc/spu/4.1.1/include/spu_mfcio.h -/opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h -*/ - -static boolean Debug = FALSE; - -volatile struct cell_init_info init; - -struct framebuffer fb; - - - -void -wait_on_mask(unsigned tagMask) -{ - mfc_write_tag_mask( tagMask ); - /* wait for completion of _any_ DMAs specified by tagMask */ - mfc_read_tag_status_any(); -} - - -static void -wait_on_mask_all(unsigned tagMask) -{ - mfc_write_tag_mask( tagMask ); - /* wait for completion of _any_ DMAs specified by tagMask */ - mfc_read_tag_status_all(); -} - - - -/** - * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled - * tiles back to the main framebuffer. - */ -static void -really_clear_tiles(uint surfaceIndex) -{ - const uint num_tiles = fb.width_tiles * fb.height_tiles; - uint i, j; - - if (surfaceIndex == 0) { - for (i = 0; i < TILE_SIZE; i++) - for (j = 0; j < TILE_SIZE; j++) - ctile[i][j] = fb.color_clear_value; - - for (i = init.id; i < num_tiles; i += init.num_spus) { - uint tx = i % fb.width_tiles; - uint ty = i / fb.width_tiles; - if (tile_status[ty][tx] == TILE_STATUS_CLEAR) { - put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); - } - } - } - else { - for (i = 0; i < TILE_SIZE; i++) - for (j = 0; j < TILE_SIZE; j++) - ztile[i][j] = fb.depth_clear_value; - - for (i = init.id; i < num_tiles; i += init.num_spus) { - uint tx = i % fb.width_tiles; - uint ty = i / fb.width_tiles; - if (tile_status_z[ty][tx] == TILE_STATUS_CLEAR) - put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 1); - } - } - -#if 0 - wait_on_mask(1 << TAG_SURFACE_CLEAR); -#endif -} - - -static void -cmd_clear_surface(const struct cell_command_clear_surface *clear) -{ - const uint num_tiles = fb.width_tiles * fb.height_tiles; - uint i, j; - - if (Debug) - printf("SPU %u: CLEAR SURF %u to 0x%08x\n", init.id, - clear->surface, clear->value); - -#define CLEAR_OPT 1 -#if CLEAR_OPT - /* set all tile's status to CLEAR */ - if (clear->surface == 0) { - memset(tile_status, TILE_STATUS_CLEAR, sizeof(tile_status)); - fb.color_clear_value = clear->value; - } - else { - memset(tile_status_z, TILE_STATUS_CLEAR, sizeof(tile_status_z)); - fb.depth_clear_value = clear->value; - } - return; -#endif - - if (clear->surface == 0) { - for (i = 0; i < TILE_SIZE; i++) - for (j = 0; j < TILE_SIZE; j++) - ctile[i][j] = clear->value; - } - else { - for (i = 0; i < TILE_SIZE; i++) - for (j = 0; j < TILE_SIZE; j++) - ztile[i][j] = clear->value; - } - - /* - printf("SPU: %s num=%d w=%d h=%d\n", - __FUNCTION__, num_tiles, fb.width_tiles, fb.height_tiles); - */ - - for (i = init.id; i < num_tiles; i += init.num_spus) { - uint tx = i % fb.width_tiles; - uint ty = i / fb.width_tiles; - if (clear->surface == 0) - put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); - else - put_tile(&fb, tx, ty, (uint *) ztile, TAG_SURFACE_CLEAR, 1); - /* XXX we don't want this here, but it fixes bad tile results */ - } - -#if 0 - wait_on_mask(1 << TAG_SURFACE_CLEAR); -#endif -} - - -/** - * Given a rendering command's bounding box (in pixels) compute the - * location of the corresponding screen tile bounding box. - */ -static INLINE void -tile_bounding_box(const struct cell_command_render *render, - uint *txmin, uint *tymin, - uint *box_num_tiles, uint *box_width_tiles) -{ -#if 1 - /* Debug: full-window bounding box */ - uint txmax = fb.width_tiles - 1; - uint tymax = fb.height_tiles - 1; - *txmin = 0; - *tymin = 0; - *box_num_tiles = fb.width_tiles * fb.height_tiles; - *box_width_tiles = fb.width_tiles; - (void) render; - (void) txmax; - (void) tymax; -#else - uint txmax, tymax, box_height_tiles; - - *txmin = (uint) render->xmin / TILE_SIZE; - *tymin = (uint) render->ymin / TILE_SIZE; - txmax = (uint) render->xmax / TILE_SIZE; - tymax = (uint) render->ymax / TILE_SIZE; - *box_width_tiles = txmax - *txmin + 1; - box_height_tiles = tymax - *tymin + 1; - *box_num_tiles = *box_width_tiles * box_height_tiles; -#endif -#if 0 - printf("Render bounds: %g, %g ... %g, %g\n", - render->xmin, render->ymin, render->xmax, render->ymax); - printf("Render tiles: %u, %u .. %u, %u\n", *txmin, *tymin, txmax, tymax); -#endif -} - - -static void -cmd_render(const struct cell_command_render *render) -{ - /* we'll DMA into these buffers */ - ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; - ushort indexes[CELL_MAX_VBUF_INDEXES] ALIGN16_ATTRIB; - - uint i, j, vertex_size, vertex_bytes, index_bytes; - - if (Debug) - printf("SPU %u: RENDER prim %u, indices: %u, nr_vert: %u\n", - init.id, - render->prim_type, - render->num_verts, - render->num_indexes); - - - ASSERT_ALIGN16(render->vertex_data); - ASSERT_ALIGN16(render->index_data); - - vertex_size = render->num_attribs * 4 * sizeof(float); - - /* how much vertex data */ - vertex_bytes = render->num_verts * vertex_size; - index_bytes = render->num_indexes * sizeof(ushort); - if (index_bytes < 8) - index_bytes = 8; - else - index_bytes = (index_bytes + 15) & ~0xf; /* multiple of 16 */ - - /* - printf("VBUF: indices at %p, vertices at %p vertex_bytes %u ind_bytes %u\n", - render->index_data, render->vertex_data, vertex_bytes, index_bytes); - */ - - /* get vertex data from main memory */ - mfc_get(vertex_data, /* dest */ - (unsigned int) render->vertex_data, /* src */ - vertex_bytes, /* size */ - TAG_VERTEX_BUFFER, - 0, /* tid */ - 0 /* rid */); - - /* get index data from main memory */ - mfc_get(indexes, /* dest */ - (unsigned int) render->index_data, /* src */ - index_bytes, - TAG_INDEX_BUFFER, - 0, /* tid */ - 0 /* rid */); - - wait_on_mask_all((1 << TAG_VERTEX_BUFFER) | - (1 << TAG_INDEX_BUFFER)); - - /* find tiles which intersect the prim bounding box */ - uint txmin, tymin, box_width_tiles, box_num_tiles; -#if 0 - tile_bounding_box(render, &txmin, &tymin, - &box_num_tiles, &box_width_tiles); -#else - txmin = 0; - tymin = 0; - box_num_tiles = fb.width_tiles * fb.height_tiles; - box_width_tiles = fb.width_tiles; -#endif - - /* make sure any pending clears have completed */ - wait_on_mask(1 << TAG_SURFACE_CLEAR); - - /* loop over tiles */ - for (i = init.id; i < box_num_tiles; i += init.num_spus) { - const uint tx = txmin + i % box_width_tiles; - const uint ty = tymin + i / box_width_tiles; - - ASSERT(tx < fb.width_tiles); - ASSERT(ty < fb.height_tiles); - - /* Start fetching color/z tiles. We'll wait for completion when - * we need read/write to them later in triangle rasterization. - */ - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(&fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); - } - } - - if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); - } - - ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); - - /* loop over tris */ - for (j = 0; j < render->num_indexes; j += 3) { - const float *v0, *v1, *v2; - - v0 = (const float *) (vertex_data + indexes[j+0] * vertex_size); - v1 = (const float *) (vertex_data + indexes[j+1] * vertex_size); - v2 = (const float *) (vertex_data + indexes[j+2] * vertex_size); - - tri_draw(v0, v1, v2, tx, ty); - } - - /* write color/z tiles back to main framebuffer, if dirtied */ - if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(&fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); - tile_status[ty][tx] = TILE_STATUS_DEFINED; - } - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(&fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); - tile_status_z[ty][tx] = TILE_STATUS_DEFINED; - } - } - - /* XXX move these... */ - wait_on_mask(1 << TAG_WRITE_TILE_COLOR); - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - wait_on_mask(1 << TAG_WRITE_TILE_Z); - } - } -} - - -static void -cmd_framebuffer(const struct cell_command_framebuffer *cmd) -{ - if (Debug) - printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", - init.id, - cmd->width, - cmd->height, - cmd->color_start, - cmd->color_format, - cmd->depth_format); - - fb.color_start = cmd->color_start; - fb.depth_start = cmd->depth_start; - fb.color_format = cmd->color_format; - fb.depth_format = cmd->depth_format; - fb.width = cmd->width; - fb.height = cmd->height; - fb.width_tiles = (fb.width + TILE_SIZE - 1) / TILE_SIZE; - fb.height_tiles = (fb.height + TILE_SIZE - 1) / TILE_SIZE; -} - - -static void -cmd_finish(void) -{ - if (Debug) - printf("SPU %u: FINISH\n", init.id); - really_clear_tiles(0); - /* wait for all outstanding DMAs to finish */ - mfc_write_tag_mask(~0); - mfc_read_tag_status_all(); - /* send mbox message to PPU */ - spu_write_out_mbox(CELL_CMD_FINISH); -} - - -/** - * Execute a batch of commands - * The opcode param encodes the location of the buffer and its size. - */ -static void -cmd_batch(uint opcode) -{ - const uint buf = (opcode >> 8) & 0xff; - uint size = (opcode >> 16); - uint buffer[CELL_BATCH_BUFFER_SIZE / 4] ALIGN16_ATTRIB; - const uint usize = size / sizeof(uint); - uint pos; - - if (Debug) - printf("SPU %u: BATCH buffer %u, len %u, from %p\n", - init.id, buf, size, init.batch_buffers[buf]); - - ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH); - - ASSERT_ALIGN16(init.batch_buffers[buf]); - - size = (size + 0xf) & ~0xf; - - mfc_get(buffer, /* dest */ - (unsigned int) init.batch_buffers[buf], /* src */ - size, - TAG_BATCH_BUFFER, - 0, /* tid */ - 0 /* rid */); - wait_on_mask(1 << TAG_BATCH_BUFFER); - - for (pos = 0; pos < usize; /* no incr */) { - switch (buffer[pos]) { - case CELL_CMD_FRAMEBUFFER: - { - struct cell_command_framebuffer *fb - = (struct cell_command_framebuffer *) &buffer[pos]; - cmd_framebuffer(fb); - pos += sizeof(*fb) / 4; - } - break; - case CELL_CMD_CLEAR_SURFACE: - { - struct cell_command_clear_surface *clr - = (struct cell_command_clear_surface *) &buffer[pos]; - cmd_clear_surface(clr); - pos += sizeof(*clr) / 4; - } - break; - case CELL_CMD_RENDER: - { - struct cell_command_render *render - = (struct cell_command_render *) &buffer[pos]; - cmd_render(render); - pos += sizeof(*render) / 4; - } - break; - case CELL_CMD_FINISH: - cmd_finish(); - pos += 1; - break; - default: - printf("SPU %u: bad opcode: 0x%x\n", init.id, buffer[pos]); - ASSERT(0); - break; - } - } - - if (Debug) - printf("SPU %u: BATCH complete\n", init.id); -} - - -/** - * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. - */ -static void -main_loop(void) -{ - struct cell_command cmd; - int exitFlag = 0; - - if (Debug) - printf("SPU %u: Enter main loop\n", init.id); - - ASSERT((sizeof(struct cell_command) & 0xf) == 0); - ASSERT_ALIGN16(&cmd); - - while (!exitFlag) { - unsigned opcode; - int tag = 0; - - if (Debug) - printf("SPU %u: Wait for cmd...\n", init.id); - - /* read/wait from mailbox */ - opcode = (unsigned int) spu_read_in_mbox(); - - if (Debug) - printf("SPU %u: got cmd 0x%x\n", init.id, opcode); - - /* command payload */ - mfc_get(&cmd, /* dest */ - (unsigned int) init.cmd, /* src */ - sizeof(struct cell_command), /* bytes */ - tag, - 0, /* tid */ - 0 /* rid */); - wait_on_mask( 1 << tag ); - - switch (opcode & CELL_CMD_OPCODE_MASK) { - case CELL_CMD_EXIT: - if (Debug) - printf("SPU %u: EXIT\n", init.id); - exitFlag = 1; - break; - case CELL_CMD_FRAMEBUFFER: - cmd_framebuffer(&cmd.fb); - break; - case CELL_CMD_CLEAR_SURFACE: - cmd_clear_surface(&cmd.clear); - break; - case CELL_CMD_RENDER: - cmd_render(&cmd.render); - break; - case CELL_CMD_BATCH: - cmd_batch(opcode); - break; - case CELL_CMD_FINISH: - cmd_finish(); - break; - default: - printf("Bad opcode!\n"); - } - - } - - if (Debug) - printf("SPU %u: Exit main loop\n", init.id); -} - - - -static void -one_time_init(void) -{ - memset(tile_status, TILE_STATUS_DEFINED, sizeof(tile_status)); - memset(tile_status_z, TILE_STATUS_DEFINED, sizeof(tile_status_z)); -} - - -/** - * SPE entrypoint. - * Note: example programs declare params as 'unsigned long long' but - * that doesn't work. - */ -int -main(unsigned long speid, unsigned long argp) -{ - int tag = 0; - - (void) speid; - - one_time_init(); - - if (Debug) - printf("SPU: main() speid=%lu\n", speid); - - mfc_get(&init, /* dest */ - (unsigned int) argp, /* src */ - sizeof(struct cell_init_info), /* bytes */ - tag, - 0, /* tid */ - 0 /* rid */); - wait_on_mask( 1 << tag ); - - - main_loop(); - - return 0; -} diff --git a/src/mesa/pipe/cell/spu/main.h b/src/mesa/pipe/cell/spu/main.h deleted file mode 100644 index cc70fdd7a6..0000000000 --- a/src/mesa/pipe/cell/spu/main.h +++ /dev/null @@ -1,78 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef MAIN_H -#define MAIN_H - - -#include "pipe/cell/common.h" - - -extern volatile struct cell_init_info init; - -struct framebuffer { - void *color_start; /**< addr of color surface in main memory */ - void *depth_start; /**< addr of depth surface in main memory */ - enum pipe_format color_format; - enum pipe_format depth_format; - uint width, height; /**< size in pixels */ - uint width_tiles, height_tiles; /**< width and height in tiles */ - - uint color_clear_value; - uint depth_clear_value; -}; - -/* XXX Collect these globals in a struct: */ - -extern struct framebuffer fb; - - -/* DMA TAGS */ - -#define TAG_SURFACE_CLEAR 10 -#define TAG_VERTEX_BUFFER 11 -#define TAG_READ_TILE_COLOR 12 -#define TAG_READ_TILE_Z 13 -#define TAG_WRITE_TILE_COLOR 14 -#define TAG_WRITE_TILE_Z 15 -#define TAG_INDEX_BUFFER 16 -#define TAG_BATCH_BUFFER 17 - -/** The standard assert macro doesn't seem to work on SPUs */ -#define ASSERT(x) \ - if (!(x)) { \ - fprintf(stderr, "SPU %d: %s:%d: %s(): assertion %s failed.\n", \ - init.id, __FILE__, __LINE__, __FUNCTION__, #x); \ - exit(1); \ - } - - -void -wait_on_mask(unsigned tag); - - -#endif /* MAIN_H */ diff --git a/src/mesa/pipe/cell/spu/tile.c b/src/mesa/pipe/cell/spu/tile.c deleted file mode 100644 index ac89d0fab0..0000000000 --- a/src/mesa/pipe/cell/spu/tile.c +++ /dev/null @@ -1,115 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - - -#include "tile.h" - - - -uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; - -ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; -ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - - - -void -get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, - int tag, int zBuf) -{ - const uint offset = ty * fb->width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); - const ubyte *src = zBuf ? fb->depth_start : fb->color_start; - - src += offset * bytesPerTile; - - ASSERT(tx < fb->width_tiles); - ASSERT(ty < fb->height_tiles); - ASSERT_ALIGN16(tile); - /* - printf("get_tile: dest: %p src: 0x%x size: %d\n", - tile, (unsigned int) src, bytesPerTile); - */ - mfc_get(tile, /* dest in local memory */ - (unsigned int) src, /* src in main memory */ - bytesPerTile, - tag, - 0, /* tid */ - 0 /* rid */); -} - - -void -put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, - int tag, int zBuf) -{ - const uint offset = ty * fb->width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); - ubyte *dst = zBuf ? fb->depth_start : fb->color_start; - - dst += offset * bytesPerTile; - - ASSERT(tx < fb->width_tiles); - ASSERT(ty < fb->height_tiles); - ASSERT_ALIGN16(tile); - /* - printf("put_tile: src: %p dst: 0x%x size: %d\n", - tile, (unsigned int) dst, bytesPerTile); - */ - mfc_put((void *) tile, /* src in local memory */ - (unsigned int) dst, /* dst in main memory */ - bytesPerTile, - tag, - 0, /* tid */ - 0 /* rid */); -} - - -void -clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value) -{ - uint i, j; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile[i][j] = value; - } - } -} - -void -clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value) -{ - uint i, j; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile[i][j] = value; - } - } -} - diff --git a/src/mesa/pipe/cell/spu/tile.h b/src/mesa/pipe/cell/spu/tile.h deleted file mode 100644 index 832cf29e30..0000000000 --- a/src/mesa/pipe/cell/spu/tile.h +++ /dev/null @@ -1,69 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TILE_H -#define TILE_H - - -#include -#include -#include "main.h" -#include "pipe/cell/common.h" - - -#define MAX_WIDTH 1024 -#define MAX_HEIGHT 1024 - - -extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; - - -#define TILE_STATUS_CLEAR 1 -#define TILE_STATUS_DEFINED 2 /**< defined pixel data */ -#define TILE_STATUS_DIRTY 3 /**< modified, but not put back yet */ - -extern ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; -extern ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - - -void -get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, - int tag, int zBuf); - -void -put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, - int tag, int zBuf); - -void -clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value); - -void -clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value); - - -#endif /* TILE_H */ diff --git a/src/mesa/pipe/cell/spu/tri.c b/src/mesa/pipe/cell/spu/tri.c deleted file mode 100644 index a4fcb71d20..0000000000 --- a/src/mesa/pipe/cell/spu/tri.c +++ /dev/null @@ -1,892 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Triangle rendering within a tile. - */ - -#include "pipe/p_compiler.h" -#include "pipe/p_format.h" -#include "pipe/p_util.h" -#include "main.h" -#include "tile.h" -#include "tri.h" - - - -/** - * Simplified types taken from other parts of Gallium - */ - -struct vertex_header { - float data[2][4]; /* pos and color */ -}; - -struct prim_header { - struct vertex_header *v[3]; -}; - - - - -#if 1 - -/* XXX fix this */ -#undef CEILF -#define CEILF(X) ((float) (int) ((X) + 0.99999)) - - -#define QUAD_TOP_LEFT 0 -#define QUAD_TOP_RIGHT 1 -#define QUAD_BOTTOM_LEFT 2 -#define QUAD_BOTTOM_RIGHT 3 -#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) -#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) -#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) -#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) -#define MASK_ALL 0xf - -#define PIPE_MAX_SHADER_INPUTS 8 /* XXX temp */ - -#endif - - -#define DEBUG_VERTS 0 - -/** - * Triangle edge info - */ -struct edge { - float dx; /**< X(v1) - X(v0), used only during setup */ - float dy; /**< Y(v1) - Y(v0), used only during setup */ - float dxdy; /**< dx/dy */ - float sx, sy; /**< first sample point coord */ - int lines; /**< number of lines on this edge */ -}; - - -struct interp_coef -{ - float a0[4]; - float dadx[4]; - float dady[4]; -}; - -/** - * Triangle setup info (derived from draw_stage). - * Also used for line drawing (taking some liberties). - */ -struct setup_stage { - - /* Vertices are just an array of floats making up each attribute in - * turn. Currently fixed at 4 floats, but should change in time. - * Codegen will help cope with this. - */ - const struct vertex_header *vmax; - const struct vertex_header *vmid; - const struct vertex_header *vmin; - const struct vertex_header *vprovoke; - - struct edge ebot; - struct edge etop; - struct edge emaj; - - float oneoverarea; - - uint tx, ty; - - int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; - -#if 0 - struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; -#else - struct interp_coef coef[PIPE_MAX_SHADER_INPUTS]; -#endif - -#if 0 - struct quad_header quad; -#endif - - struct { - int left[2]; /**< [0] = row0, [1] = row1 */ - int right[2]; - int y; - unsigned y_flags; - unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ - } span; -}; - - -#if 0 -/** - * Basically a cast wrapper. - */ -static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) -{ - return (struct setup_stage *)stage; -} -#endif - -#if 0 -/** - * Clip setup->quad against the scissor/surface bounds. - */ -static INLINE void -quad_clip(struct setup_stage *setup) -{ - const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; - - if (setup->quad.x0 >= maxx || - setup->quad.y0 >= maxy || - setup->quad.x0 + 1 < minx || - setup->quad.y0 + 1 < miny) { - /* totally clipped */ - setup->quad.mask = 0x0; - return; - } - if (setup->quad.x0 < minx) - setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup->quad.y0 < miny) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - if (setup->quad.x0 == maxx - 1) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup->quad.y0 == maxy - 1) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); -} -#endif - -#if 0 -/** - * Emit a quad (pass to next stage) with clipping. - */ -static INLINE void -clip_emit_quad(struct setup_stage *setup) -{ - quad_clip(setup); - if (setup->quad.mask) { - struct softpipe_context *sp = setup->softpipe; - sp->quad.first->run(sp->quad.first, &setup->quad); - } -} -#endif - -/** - * Evaluate attribute coefficients (plane equations) to compute - * attribute values for the four fragments in a quad. - * Eg: four colors will be compute. - */ -static INLINE void -eval_coeff( struct setup_stage *setup, uint slot, - float x, float y, float result[4][4]) -{ - uint i; - const float *dadx = setup->coef[slot].dadx; - const float *dady = setup->coef[slot].dady; - - /* loop over XYZW comps */ - for (i = 0; i < 4; i++) { - result[QUAD_TOP_LEFT][i] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; - result[QUAD_TOP_RIGHT][i] = result[0][i] + dadx[i]; - result[QUAD_BOTTOM_LEFT][i] = result[0][i] + dady[i]; - result[QUAD_BOTTOM_RIGHT][i] = result[0][i] + dadx[i] + dady[i]; - } -} - - -static INLINE void -eval_z( struct setup_stage *setup, - float x, float y, float result[4]) -{ - uint slot = 0; - uint i = 2; - const float *dadx = setup->coef[slot].dadx; - const float *dady = setup->coef[slot].dady; - - result[QUAD_TOP_LEFT] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; - result[QUAD_TOP_RIGHT] = result[0] + dadx[i]; - result[QUAD_BOTTOM_LEFT] = result[0] + dady[i]; - result[QUAD_BOTTOM_RIGHT] = result[0] + dadx[i] + dady[i]; -} - - -static INLINE uint -pack_color(const float color[4]) -{ - uint r = (uint) (color[0] * 255.0); - uint g = (uint) (color[1] * 255.0); - uint b = (uint) (color[2] * 255.0); - uint a = (uint) (color[3] * 255.0); - switch (fb.color_format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - return (a << 24) | (r << 16) | (g << 8) | b; - case PIPE_FORMAT_B8G8R8A8_UNORM: - return (b << 24) | (g << 16) | (r << 8) | a; - default: - ASSERT(0); - return 0; - } -} - - -/** - * Emit a quad (pass to next stage). No clipping is done. - */ -static INLINE void -emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) -{ -#if 0 - struct softpipe_context *sp = setup->softpipe; - setup->quad.x0 = x; - setup->quad.y0 = y; - setup->quad.mask = mask; - sp->quad.first->run(sp->quad.first, &setup->quad); -#else - /* Cell: "write" quad fragments to the tile by setting prim color */ - int ix = x - setup->cliprect_minx; - int iy = y - setup->cliprect_miny; - float colors[4][4]; - uint z; - - eval_coeff(setup, 1, (float) x, (float) y, colors); - - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - float zvals[4]; - eval_z(setup, (float) x, (float) y, zvals); - - if (tile_status_z[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { - /* now, _really_ clear the tile */ - clear_tile_z(ztile, fb.depth_clear_value); - } - else { - /* make sure we've got the tile from main mem */ - wait_on_mask(1 << TAG_READ_TILE_Z); - } - tile_status_z[setup->ty][setup->tx] = TILE_STATUS_DIRTY; - - if (mask & MASK_TOP_LEFT) { - z = (uint) (zvals[0] * 65535.0); - if (z < ztile[iy][ix]) - ztile[iy][ix] = z; - else - mask &= ~MASK_TOP_LEFT; - } - - if (mask & MASK_TOP_RIGHT) { - z = (uint) (zvals[1] * 65535.0); - if (z < ztile[iy][ix+1]) - ztile[iy][ix+1] = z; - else - mask &= ~MASK_TOP_RIGHT; - } - - if (mask & MASK_BOTTOM_LEFT) { - z = (uint) (zvals[2] * 65535.0); - if (z < ztile[iy+1][ix]) - ztile[iy+1][ix] = z; - else - mask &= ~MASK_BOTTOM_LEFT; - } - - if (mask & MASK_BOTTOM_RIGHT) { - z = (uint) (zvals[3] * 65535.0); - if (z < ztile[iy+1][ix+1]) - ztile[iy+1][ix+1] = z; - else - mask &= ~MASK_BOTTOM_RIGHT; - } - } - - if (mask) { - if (tile_status[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { - /* now, _really_ clear the tile */ - clear_tile(ctile, fb.color_clear_value); - } - else { - /* make sure we've got the tile from main mem */ - wait_on_mask(1 << TAG_READ_TILE_COLOR); - } - tile_status[setup->ty][setup->tx] = TILE_STATUS_DIRTY; - - if (mask & MASK_TOP_LEFT) - ctile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); - if (mask & MASK_TOP_RIGHT) - ctile[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); - if (mask & MASK_BOTTOM_LEFT) - ctile[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); - if (mask & MASK_BOTTOM_RIGHT) - ctile[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); - } -#endif -} - - -/** - * Given an X or Y coordinate, return the block/quad coordinate that it - * belongs to. - */ -static INLINE int block( int x ) -{ - return x & ~1; -} - - -/** - * Compute mask which indicates which pixels in the 2x2 quad are actually inside - * the triangle's bounds. - * - * this is pretty nasty... may need to rework flush_spans again to - * fix it, if possible. - */ -static unsigned calculate_mask( struct setup_stage *setup, int x ) -{ - unsigned mask = 0x0; - - if (x >= setup->span.left[0] && x < setup->span.right[0]) - mask |= MASK_TOP_LEFT; - - if (x >= setup->span.left[1] && x < setup->span.right[1]) - mask |= MASK_BOTTOM_LEFT; - - if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) - mask |= MASK_TOP_RIGHT; - - if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) - mask |= MASK_BOTTOM_RIGHT; - - return mask; -} - - -/** - * Render a horizontal span of quads - */ -static void flush_spans( struct setup_stage *setup ) -{ - int minleft, maxright; - int x; - - switch (setup->span.y_flags) { - case 0x3: - /* both odd and even lines written (both quad rows) */ - minleft = MIN2(setup->span.left[0], setup->span.left[1]); - maxright = MAX2(setup->span.right[0], setup->span.right[1]); - break; - - case 0x1: - /* only even line written (quad top row) */ - minleft = setup->span.left[0]; - maxright = setup->span.right[0]; - break; - - case 0x2: - /* only odd line written (quad bottom row) */ - minleft = setup->span.left[1]; - maxright = setup->span.right[1]; - break; - - default: - return; - } - - /* XXX this loop could be moved into the above switch cases and - * calculate_mask() could be simplified a bit... - */ - for (x = block(minleft); x <= block(maxright); x += 2) { - emit_quad( setup, x, setup->span.y, - calculate_mask( setup, x ) ); - } - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; -} - -#if DEBUG_VERTS -static void print_vertex(const struct setup_stage *setup, - const struct vertex_header *v) -{ - int i; - fprintf(stderr, "Vertex: (%p)\n", v); - for (i = 0; i < setup->quad.nr_attrs; i++) { - fprintf(stderr, " %d: %f %f %f %f\n", i, - v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); - } -} -#endif - -static boolean setup_sort_vertices( struct setup_stage *setup, - const struct prim_header *prim ) -{ - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - const struct vertex_header *v2 = prim->v[2]; - -#if DEBUG_VERTS - fprintf(stderr, "Triangle:\n"); - print_vertex(setup, v0); - print_vertex(setup, v1); - print_vertex(setup, v2); -#endif - - setup->vprovoke = v2; - - /* determine bottom to top order of vertices */ - { - float y0 = v0->data[0][1]; - float y1 = v1->data[0][1]; - float y2 = v2->data[0][1]; - if (y0 <= y1) { - if (y1 <= y2) { - /* y0<=y1<=y2 */ - setup->vmin = v0; - setup->vmid = v1; - setup->vmax = v2; - } - else if (y2 <= y0) { - /* y2<=y0<=y1 */ - setup->vmin = v2; - setup->vmid = v0; - setup->vmax = v1; - } - else { - /* y0<=y2<=y1 */ - setup->vmin = v0; - setup->vmid = v2; - setup->vmax = v1; - } - } - else { - if (y0 <= y2) { - /* y1<=y0<=y2 */ - setup->vmin = v1; - setup->vmid = v0; - setup->vmax = v2; - } - else if (y2 <= y1) { - /* y2<=y1<=y0 */ - setup->vmin = v2; - setup->vmid = v1; - setup->vmax = v0; - } - else { - /* y1<=y2<=y0 */ - setup->vmin = v1; - setup->vmid = v2; - setup->vmax = v0; - } - } - } - - /* Check if triangle is completely outside the tile bounds */ - if (setup->vmin->data[0][1] > setup->cliprect_maxy) - return FALSE; - if (setup->vmax->data[0][1] < setup->cliprect_miny) - return FALSE; - if (setup->vmin->data[0][0] < setup->cliprect_minx && - setup->vmid->data[0][0] < setup->cliprect_minx && - setup->vmax->data[0][0] < setup->cliprect_minx) - return FALSE; - if (setup->vmin->data[0][0] > setup->cliprect_maxx && - setup->vmid->data[0][0] > setup->cliprect_maxx && - setup->vmax->data[0][0] > setup->cliprect_maxx) - return FALSE; - - setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; - setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; - setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; - - /* - * Compute triangle's area. Use 1/area to compute partial - * derivatives of attributes later. - * - * The area will be the same as prim->det, but the sign may be - * different depending on how the vertices get sorted above. - * - * To determine whether the primitive is front or back facing we - * use the prim->det value because its sign is correct. - */ - { - const float area = (setup->emaj.dx * setup->ebot.dy - - setup->ebot.dx * setup->emaj.dy); - - setup->oneoverarea = 1.0f / area; - /* - _mesa_printf("%s one-over-area %f area %f det %f\n", - __FUNCTION__, setup->oneoverarea, area, prim->det ); - */ - } - -#if 0 - /* We need to know if this is a front or back-facing triangle for: - * - the GLSL gl_FrontFacing fragment attribute (bool) - * - two-sided stencil test - */ - setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); -#endif - - return TRUE; -} - - -#if 0 -/** - * Compute a0 for a constant-valued coefficient (GL_FLAT shading). - * The value value comes from vertex->data[slot][i]. - * The result will be put into setup->coef[slot].a0[i]. - * \param slot which attribute slot - * \param i which component of the slot (0..3) - */ -static void const_coeff( struct setup_stage *setup, - unsigned slot, - unsigned i ) -{ - assert(slot < PIPE_MAX_SHADER_INPUTS); - assert(i <= 3); - - setup->coef[slot].dadx[i] = 0; - setup->coef[slot].dady[i] = 0; - - /* need provoking vertex info! - */ - setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; -} -#endif - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a triangle. - */ -static void tri_linear_coeff( struct setup_stage *setup, - uint slot, uint firstComp, uint lastComp ) -{ - uint i; - for (i = firstComp; i < lastComp; i++) { - float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; - float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - - ASSERT(slot < PIPE_MAX_SHADER_INPUTS); - - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; - - /* calculate a0 as the value which would be sampled for the - * fragment at (0,0), taking into account that we want to sample at - * pixel centers, in other words (0.5, 0.5). - * - * this is neat but unfortunately not a good way to do things for - * triangles with very large values of dadx or dady as it will - * result in the subtraction and re-addition from a0 of a very - * large number, which means we'll end up loosing a lot of the - * fractional bits and precision from a0. the way to fix this is - * to define a0 as the sample at a pixel center somewhere near vmin - * instead - i'll switch to this later. - */ - setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); - } - - /* - _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", - slot, "xyzw"[i], - setup->coef[slot].a0[i], - setup->coef[slot].dadx[i], - setup->coef[slot].dady[i]); - */ -} - - -#if 0 -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a triangle. - * We basically multiply the vertex value by 1/w before computing - * the plane coefficients (a0, dadx, dady). - * Later, when we compute the value at a particular fragment position we'll - * divide the interpolated value by the interpolated W at that fragment. - */ -static void tri_persp_coeff( struct setup_stage *setup, - unsigned slot, - unsigned i ) -{ - /* premultiply by 1/w: - */ - float mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; - float mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; - float maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; - - float botda = mida - mina; - float majda = maxa - mina; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - - /* - printf("tri persp %d,%d: %f %f %f\n", slot, i, - setup->vmin->data[slot][i], - setup->vmid->data[slot][i], - setup->vmax->data[slot][i] - ); - */ - - assert(slot < PIPE_MAX_SHADER_INPUTS); - assert(i <= 3); - - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; - setup->coef[slot].a0[i] = (mina - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); -} -#endif - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. - */ -static void setup_tri_coefficients( struct setup_stage *setup ) -{ -#if 0 - const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; - unsigned slot, j; - - /* z and w are done by linear interpolation: - */ - tri_linear_coeff(setup, 0, 2); - tri_linear_coeff(setup, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - switch (interp[slot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - break; - - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - tri_linear_coeff(setup, slot, j); - break; - - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - tri_persp_coeff(setup, slot, j); - break; - - default: - /* invalid interp mode */ - assert(0); - } - } -#else - tri_linear_coeff(setup, 0, 2, 3); /* slot 0, z */ - tri_linear_coeff(setup, 1, 0, 4); /* slot 1, color */ -#endif -} - - -static void setup_tri_edges( struct setup_stage *setup ) -{ - float vmin_x = setup->vmin->data[0][0] + 0.5f; - float vmid_x = setup->vmid->data[0][0] + 0.5f; - - float vmin_y = setup->vmin->data[0][1] - 0.5f; - float vmid_y = setup->vmid->data[0][1] - 0.5f; - float vmax_y = setup->vmax->data[0][1] - 0.5f; - - setup->emaj.sy = CEILF(vmin_y); - setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); - setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; - setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; - - setup->etop.sy = CEILF(vmid_y); - setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); - setup->etop.dxdy = setup->etop.dx / setup->etop.dy; - setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; - - setup->ebot.sy = CEILF(vmin_y); - setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); - setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; - setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; -} - - -/** - * Render the upper or lower half of a triangle. - * Scissoring/cliprect is applied here too. - */ -static void subtriangle( struct setup_stage *setup, - struct edge *eleft, - struct edge *eright, - unsigned lines ) -{ - const int minx = setup->cliprect_minx; - const int maxx = setup->cliprect_maxx; - const int miny = setup->cliprect_miny; - const int maxy = setup->cliprect_maxy; - int y, start_y, finish_y; - int sy = (int)eleft->sy; - - ASSERT((int)eleft->sy == (int) eright->sy); - - /* clip top/bottom */ - start_y = sy; - finish_y = sy + lines; - - if (start_y < miny) - start_y = miny; - - if (finish_y > maxy) - finish_y = maxy; - - start_y -= sy; - finish_y -= sy; - - /* - _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); - */ - - for (y = start_y; y < finish_y; y++) { - - /* avoid accumulating adds as floats don't have the precision to - * accurately iterate large triangle edges that way. luckily we - * can just multiply these days. - * - * this is all drowned out by the attribute interpolation anyway. - */ - int left = (int)(eleft->sx + y * eleft->dxdy); - int right = (int)(eright->sx + y * eright->dxdy); - - /* clip left/right */ - if (left < minx) - left = minx; - if (right > maxx) - right = maxx; - - if (left < right) { - int _y = sy + y; - if (block(_y) != setup->span.y) { - flush_spans(setup); - setup->span.y = block(_y); - } - - setup->span.left[_y&1] = left; - setup->span.right[_y&1] = right; - setup->span.y_flags |= 1<<(_y&1); - } - } - - - /* save the values so that emaj can be restarted: - */ - eleft->sx += lines * eleft->dxdy; - eright->sx += lines * eright->dxdy; - eleft->sy += lines; - eright->sy += lines; -} - - -/** - * Do setup for triangle rasterization, then render the triangle. - */ -static void -setup_tri(struct setup_stage *setup, struct prim_header *prim) -{ - if (!setup_sort_vertices( setup, prim )) { - return; /* totally clipped */ - } - - setup_tri_coefficients( setup ); - setup_tri_edges( setup ); - -#if 0 - setup->quad.prim = PRIM_TRI; -#endif - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; - /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ - - /* init_constant_attribs( setup ); */ - - if (setup->oneoverarea < 0.0) { - /* emaj on left: - */ - subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); - subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); - } - else { - /* emaj on right: - */ - subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); - subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); - } - - flush_spans( setup ); -} - - - -/** - * Draw triangle into tile at (tx, ty) (tile coords) - * The tile data should have already been fetched. - */ -void -tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) -{ - struct prim_header tri; - struct setup_stage setup; - - tri.v[0] = (struct vertex_header *) v0; - tri.v[1] = (struct vertex_header *) v1; - tri.v[2] = (struct vertex_header *) v2; - - setup.tx = tx; - setup.ty = ty; - - /* set clipping bounds to tile bounds */ - setup.cliprect_minx = tx * TILE_SIZE; - setup.cliprect_miny = ty * TILE_SIZE; - setup.cliprect_maxx = (tx + 1) * TILE_SIZE; - setup.cliprect_maxy = (ty + 1) * TILE_SIZE; - - setup_tri(&setup, &tri); -} diff --git a/src/mesa/pipe/cell/spu/tri.h b/src/mesa/pipe/cell/spu/tri.h deleted file mode 100644 index f10c4077d3..0000000000 --- a/src/mesa/pipe/cell/spu/tri.h +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef TRI_H -#define TRI_H - - -extern void -tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty); - - -#endif /* TRI_H */ -- cgit v1.2.3 From 68f5a6f74335e252e6a04dd6ae9ef7ae1482be97 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 12 Jan 2008 11:41:23 -0700 Subject: Cell: whitespace/cleanup --- src/mesa/pipe/cell/ppu/cell_state_blend.c | 34 ++++++++++++++++---------- src/mesa/pipe/cell/ppu/cell_state_rasterizer.c | 9 +++---- 2 files changed, 25 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_blend.c b/src/mesa/pipe/cell/ppu/cell_state_blend.c index da3fcfd3a5..4e49655820 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_blend.c +++ b/src/mesa/pipe/cell/ppu/cell_state_blend.c @@ -32,17 +32,19 @@ #include "cell_context.h" #include "cell_state.h" + void * cell_create_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend) + const struct pipe_blend_state *blend) { - struct pipe_blend_state *state = MALLOC( sizeof(struct pipe_blend_state) ); + struct pipe_blend_state *state = MALLOC(sizeof(struct pipe_blend_state)); memcpy(state, blend, sizeof(struct pipe_blend_state)); return state; } -void cell_bind_blend_state( struct pipe_context *pipe, - void *blend ) + +void +cell_bind_blend_state(struct pipe_context *pipe, void *blend) { struct cell_context *cell = cell_context(pipe); @@ -51,15 +53,17 @@ void cell_bind_blend_state( struct pipe_context *pipe, cell->dirty |= CELL_NEW_BLEND; } -void cell_delete_blend_state(struct pipe_context *pipe, - void *blend) + +void +cell_delete_blend_state(struct pipe_context *pipe, void *blend) { - FREE( blend ); + FREE(blend); } -void cell_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ) +void +cell_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *blend_color) { struct cell_context *cell = cell_context(pipe); @@ -70,29 +74,33 @@ void cell_set_blend_color( struct pipe_context *pipe, + void * cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) + const struct pipe_depth_stencil_alpha_state *depth_stencil) { struct pipe_depth_stencil_alpha_state *state = - MALLOC( sizeof(struct pipe_depth_stencil_alpha_state) ); + MALLOC(sizeof(struct pipe_depth_stencil_alpha_state)); memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_alpha_state)); return state; } + void cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth_stencil) { struct cell_context *cell = cell_context(pipe); - cell->depth_stencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; + cell->depth_stencil + = (const struct pipe_depth_stencil_alpha_state *) depth_stencil; cell->dirty |= CELL_NEW_DEPTH_STENCIL; } + void cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth) { - FREE( depth ); + FREE(depth); } diff --git a/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c b/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c index 6b1675af26..d8128ece54 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c +++ b/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c @@ -78,8 +78,8 @@ void * cell_create_rasterizer_state(struct pipe_context *pipe, const struct pipe_rasterizer_state *setup) { - struct pipe_rasterizer_state *state = - MALLOC( sizeof(struct pipe_rasterizer_state) ); + struct pipe_rasterizer_state *state + = MALLOC(sizeof(struct pipe_rasterizer_state)); memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); return state; } @@ -98,10 +98,9 @@ cell_bind_rasterizer_state(struct pipe_context *pipe, void *setup) cell->dirty |= CELL_NEW_RASTERIZER; } + void cell_delete_rasterizer_state(struct pipe_context *pipe, void *rasterizer) { - FREE( rasterizer ); + FREE(rasterizer); } - - -- cgit v1.2.3 From dae719a68173bddedbb531c030cd4a12bcb0435b Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 12 Jan 2008 12:39:26 -0700 Subject: Cell: first state object (depth/stencil/alpha) emitted to SPUs --- src/mesa/pipe/cell/common.h | 1 + src/mesa/pipe/cell/ppu/Makefile | 1 + src/mesa/pipe/cell/ppu/cell_batch.c | 1 + src/mesa/pipe/cell/ppu/cell_state_blend.c | 1 + src/mesa/pipe/cell/ppu/cell_state_derived.c | 2 ++ src/mesa/pipe/cell/ppu/cell_state_emit.c | 45 +++++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_state_emit.h | 36 +++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_main.c | 20 ++++++++++++- 8 files changed, 106 insertions(+), 1 deletion(-) create mode 100644 src/mesa/pipe/cell/ppu/cell_state_emit.c create mode 100644 src/mesa/pipe/cell/ppu/cell_state_emit.h (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 3d6f1c4ba0..74b3d3cbee 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -59,6 +59,7 @@ #define CELL_CMD_FINISH 4 #define CELL_CMD_RENDER 5 #define CELL_CMD_BATCH 6 +#define CELL_CMD_STATE_DEPTH_STENCIL 7 #define CELL_NUM_BATCH_BUFFERS 2 diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index f7e3dd09f4..6a1bd5982f 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -23,6 +23,7 @@ SOURCES = \ cell_state_blend.c \ cell_state_clip.c \ cell_state_derived.c \ + cell_state_emit.c \ cell_state_fs.c \ cell_state_rasterizer.c \ cell_state_sampler.c \ diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index 7eb1b50ddb..45d49d0465 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -68,6 +68,7 @@ cell_batch_append(struct cell_context *cell, const void *cmd, uint length) { uint size; + assert(length % 4 == 0); assert(cell->cur_batch >= 0); size = cell->batch_buffer_size[cell->cur_batch]; diff --git a/src/mesa/pipe/cell/ppu/cell_state_blend.c b/src/mesa/pipe/cell/ppu/cell_state_blend.c index 4e49655820..34ae0128ea 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_blend.c +++ b/src/mesa/pipe/cell/ppu/cell_state_blend.c @@ -33,6 +33,7 @@ #include "cell_state.h" + void * cell_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c index ec3a8f3a81..dc2879b915 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -219,5 +219,7 @@ void cell_update_derived( struct cell_context *cell ) compute_cliprect(cell); #endif + cell_emit_state(cell); + cell->dirty = 0; } diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c new file mode 100644 index 0000000000..e1a1458f39 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.c @@ -0,0 +1,45 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "cell_context.h" +#include "cell_state.h" +#include "cell_state_emit.h" +#include "cell_batch.h" + + + +void +cell_emit_state(struct cell_context *cell) +{ + if (cell->dirty & CELL_NEW_DEPTH_STENCIL) { + uint cmd = CELL_CMD_STATE_DEPTH_STENCIL; + cell_batch_append(cell, &cmd, 4); + cell_batch_append(cell, cell->depth_stencil, + sizeof(struct pipe_depth_stencil_alpha_state)); + } +} diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.h b/src/mesa/pipe/cell/ppu/cell_state_emit.h new file mode 100644 index 0000000000..59f8affe8d --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_STATE_EMIT_H +#define CELL_STATE_EMIT_H + + +extern void +cell_emit_state(struct cell_context *cell); + + +#endif /* CELL_STATE_EMIT_H */ diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index df708e65d1..f9d7302c36 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -38,6 +38,8 @@ #include "spu_tile.h" #include "pipe/cell/common.h" #include "pipe/p_defines.h" +#include "pipe/p_state.h" + /* helpful headers: @@ -45,7 +47,7 @@ helpful headers: /opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h */ -static boolean Debug = FALSE; +static boolean Debug = TRUE; volatile struct cell_init_info init; @@ -356,6 +358,17 @@ cmd_framebuffer(const struct cell_command_framebuffer *cmd) } +static void +cmd_state_depth_stencil(const struct pipe_depth_stencil_alpha_state *state) +{ + if (Debug) + printf("SPU %u: DEPTH_STENCIL: ztest %d\n", + init.id, + state->depth.enabled); + /* XXX copy/save the state */ +} + + static void cmd_finish(void) { @@ -431,6 +444,11 @@ cmd_batch(uint opcode) cmd_finish(); pos += 1; break; + case CELL_CMD_STATE_DEPTH_STENCIL: + cmd_state_depth_stencil((struct pipe_depth_stencil_alpha_state *) + &buffer[pos+1]); + pos += (1 + sizeof(struct pipe_depth_stencil_alpha_state) / 4); + break; default: printf("SPU %u: bad opcode: 0x%x\n", init.id, buffer[pos]); ASSERT(0); -- cgit v1.2.3 From d53e1c255aad83ee6c183f6e144d309327898669 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 12 Jan 2008 12:53:49 -0700 Subject: Cell: collect vars in a spu_global struct --- src/mesa/pipe/cell/spu/spu_main.c | 128 ++++++++++++++++++-------------------- src/mesa/pipe/cell/spu/spu_main.h | 27 ++++++-- src/mesa/pipe/cell/spu/spu_tri.c | 18 ++---- 3 files changed, 88 insertions(+), 85 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index f9d7302c36..7446cc3eff 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -38,7 +38,6 @@ #include "spu_tile.h" #include "pipe/cell/common.h" #include "pipe/p_defines.h" -#include "pipe/p_state.h" /* @@ -49,10 +48,7 @@ helpful headers: static boolean Debug = TRUE; -volatile struct cell_init_info init; - -struct framebuffer fb; - +struct spu_global spu; void @@ -81,32 +77,32 @@ wait_on_mask_all(unsigned tagMask) static void really_clear_tiles(uint surfaceIndex) { - const uint num_tiles = fb.width_tiles * fb.height_tiles; + const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; uint i, j; if (surfaceIndex == 0) { for (i = 0; i < TILE_SIZE; i++) for (j = 0; j < TILE_SIZE; j++) - ctile[i][j] = fb.color_clear_value; + ctile[i][j] = spu.fb.color_clear_value; - for (i = init.id; i < num_tiles; i += init.num_spus) { - uint tx = i % fb.width_tiles; - uint ty = i / fb.width_tiles; + for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { + uint tx = i % spu.fb.width_tiles; + uint ty = i / spu.fb.width_tiles; if (tile_status[ty][tx] == TILE_STATUS_CLEAR) { - put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); + put_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); } } } else { for (i = 0; i < TILE_SIZE; i++) for (j = 0; j < TILE_SIZE; j++) - ztile[i][j] = fb.depth_clear_value; + ztile[i][j] = spu.fb.depth_clear_value; - for (i = init.id; i < num_tiles; i += init.num_spus) { - uint tx = i % fb.width_tiles; - uint ty = i / fb.width_tiles; + for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { + uint tx = i % spu.fb.width_tiles; + uint ty = i / spu.fb.width_tiles; if (tile_status_z[ty][tx] == TILE_STATUS_CLEAR) - put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 1); + put_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 1); } } @@ -119,11 +115,11 @@ really_clear_tiles(uint surfaceIndex) static void cmd_clear_surface(const struct cell_command_clear_surface *clear) { - const uint num_tiles = fb.width_tiles * fb.height_tiles; + const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; uint i, j; if (Debug) - printf("SPU %u: CLEAR SURF %u to 0x%08x\n", init.id, + printf("SPU %u: CLEAR SURF %u to 0x%08x\n", spu.init.id, clear->surface, clear->value); #define CLEAR_OPT 1 @@ -131,11 +127,11 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) /* set all tile's status to CLEAR */ if (clear->surface == 0) { memset(tile_status, TILE_STATUS_CLEAR, sizeof(tile_status)); - fb.color_clear_value = clear->value; + spu.fb.color_clear_value = clear->value; } else { memset(tile_status_z, TILE_STATUS_CLEAR, sizeof(tile_status_z)); - fb.depth_clear_value = clear->value; + spu.fb.depth_clear_value = clear->value; } return; #endif @@ -153,16 +149,16 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) /* printf("SPU: %s num=%d w=%d h=%d\n", - __FUNCTION__, num_tiles, fb.width_tiles, fb.height_tiles); + __FUNCTION__, num_tiles, spu.fb.width_tiles, spu.fb.height_tiles); */ - for (i = init.id; i < num_tiles; i += init.num_spus) { - uint tx = i % fb.width_tiles; - uint ty = i / fb.width_tiles; + for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { + uint tx = i % spu.fb.width_tiles; + uint ty = i / spu.fb.width_tiles; if (clear->surface == 0) - put_tile(&fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); + put_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); else - put_tile(&fb, tx, ty, (uint *) ztile, TAG_SURFACE_CLEAR, 1); + put_tile(&spu.fb, tx, ty, (uint *) ztile, TAG_SURFACE_CLEAR, 1); /* XXX we don't want this here, but it fixes bad tile results */ } @@ -183,12 +179,12 @@ tile_bounding_box(const struct cell_command_render *render, { #if 1 /* Debug: full-window bounding box */ - uint txmax = fb.width_tiles - 1; - uint tymax = fb.height_tiles - 1; + uint txmax = spu.fb.width_tiles - 1; + uint tymax = spu.fb.height_tiles - 1; *txmin = 0; *tymin = 0; - *box_num_tiles = fb.width_tiles * fb.height_tiles; - *box_width_tiles = fb.width_tiles; + *box_num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; + *box_width_tiles = spu.fb.width_tiles; (void) render; (void) txmax; (void) tymax; @@ -222,7 +218,7 @@ cmd_render(const struct cell_command_render *render) if (Debug) printf("SPU %u: RENDER prim %u, indices: %u, nr_vert: %u\n", - init.id, + spu.init.id, render->prim_type, render->num_verts, render->num_indexes); @@ -273,32 +269,32 @@ cmd_render(const struct cell_command_render *render) #else txmin = 0; tymin = 0; - box_num_tiles = fb.width_tiles * fb.height_tiles; - box_width_tiles = fb.width_tiles; + box_num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; + box_width_tiles = spu.fb.width_tiles; #endif /* make sure any pending clears have completed */ wait_on_mask(1 << TAG_SURFACE_CLEAR); /* loop over tiles */ - for (i = init.id; i < box_num_tiles; i += init.num_spus) { + for (i = spu.init.id; i < box_num_tiles; i += spu.init.num_spus) { const uint tx = txmin + i % box_width_tiles; const uint ty = tymin + i / box_width_tiles; - ASSERT(tx < fb.width_tiles); - ASSERT(ty < fb.height_tiles); + ASSERT(tx < spu.fb.width_tiles); + ASSERT(ty < spu.fb.height_tiles); /* Start fetching color/z tiles. We'll wait for completion when * we need read/write to them later in triangle rasterization. */ - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(&fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); + get_tile(&spu.fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); } } if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(&fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); + get_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); } ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); @@ -316,19 +312,19 @@ cmd_render(const struct cell_command_render *render) /* write color/z tiles back to main framebuffer, if dirtied */ if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(&fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); + put_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); tile_status[ty][tx] = TILE_STATUS_DEFINED; } - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(&fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); + put_tile(&spu.fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); tile_status_z[ty][tx] = TILE_STATUS_DEFINED; } } /* XXX move these... */ wait_on_mask(1 << TAG_WRITE_TILE_COLOR); - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { wait_on_mask(1 << TAG_WRITE_TILE_Z); } } @@ -340,21 +336,21 @@ cmd_framebuffer(const struct cell_command_framebuffer *cmd) { if (Debug) printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", - init.id, + spu.init.id, cmd->width, cmd->height, cmd->color_start, cmd->color_format, cmd->depth_format); - fb.color_start = cmd->color_start; - fb.depth_start = cmd->depth_start; - fb.color_format = cmd->color_format; - fb.depth_format = cmd->depth_format; - fb.width = cmd->width; - fb.height = cmd->height; - fb.width_tiles = (fb.width + TILE_SIZE - 1) / TILE_SIZE; - fb.height_tiles = (fb.height + TILE_SIZE - 1) / TILE_SIZE; + spu.fb.color_start = cmd->color_start; + spu.fb.depth_start = cmd->depth_start; + spu.fb.color_format = cmd->color_format; + spu.fb.depth_format = cmd->depth_format; + spu.fb.width = cmd->width; + spu.fb.height = cmd->height; + spu.fb.width_tiles = (spu.fb.width + TILE_SIZE - 1) / TILE_SIZE; + spu.fb.height_tiles = (spu.fb.height + TILE_SIZE - 1) / TILE_SIZE; } @@ -363,7 +359,7 @@ cmd_state_depth_stencil(const struct pipe_depth_stencil_alpha_state *state) { if (Debug) printf("SPU %u: DEPTH_STENCIL: ztest %d\n", - init.id, + spu.init.id, state->depth.enabled); /* XXX copy/save the state */ } @@ -373,7 +369,7 @@ static void cmd_finish(void) { if (Debug) - printf("SPU %u: FINISH\n", init.id); + printf("SPU %u: FINISH\n", spu.init.id); really_clear_tiles(0); /* wait for all outstanding DMAs to finish */ mfc_write_tag_mask(~0); @@ -398,16 +394,16 @@ cmd_batch(uint opcode) if (Debug) printf("SPU %u: BATCH buffer %u, len %u, from %p\n", - init.id, buf, size, init.batch_buffers[buf]); + spu.init.id, buf, size, spu.init.batch_buffers[buf]); ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH); - ASSERT_ALIGN16(init.batch_buffers[buf]); + ASSERT_ALIGN16(spu.init.batch_buffers[buf]); size = (size + 0xf) & ~0xf; mfc_get(buffer, /* dest */ - (unsigned int) init.batch_buffers[buf], /* src */ + (unsigned int) spu.init.batch_buffers[buf], /* src */ size, TAG_BATCH_BUFFER, 0, /* tid */ @@ -450,14 +446,14 @@ cmd_batch(uint opcode) pos += (1 + sizeof(struct pipe_depth_stencil_alpha_state) / 4); break; default: - printf("SPU %u: bad opcode: 0x%x\n", init.id, buffer[pos]); + printf("SPU %u: bad opcode: 0x%x\n", spu.init.id, buffer[pos]); ASSERT(0); break; } } if (Debug) - printf("SPU %u: BATCH complete\n", init.id); + printf("SPU %u: BATCH complete\n", spu.init.id); } @@ -471,7 +467,7 @@ main_loop(void) int exitFlag = 0; if (Debug) - printf("SPU %u: Enter main loop\n", init.id); + printf("SPU %u: Enter main loop\n", spu.init.id); ASSERT((sizeof(struct cell_command) & 0xf) == 0); ASSERT_ALIGN16(&cmd); @@ -481,17 +477,17 @@ main_loop(void) int tag = 0; if (Debug) - printf("SPU %u: Wait for cmd...\n", init.id); + printf("SPU %u: Wait for cmd...\n", spu.init.id); /* read/wait from mailbox */ opcode = (unsigned int) spu_read_in_mbox(); if (Debug) - printf("SPU %u: got cmd 0x%x\n", init.id, opcode); + printf("SPU %u: got cmd 0x%x\n", spu.init.id, opcode); /* command payload */ mfc_get(&cmd, /* dest */ - (unsigned int) init.cmd, /* src */ + (unsigned int) spu.init.cmd, /* src */ sizeof(struct cell_command), /* bytes */ tag, 0, /* tid */ @@ -501,7 +497,7 @@ main_loop(void) switch (opcode & CELL_CMD_OPCODE_MASK) { case CELL_CMD_EXIT: if (Debug) - printf("SPU %u: EXIT\n", init.id); + printf("SPU %u: EXIT\n", spu.init.id); exitFlag = 1; break; case CELL_CMD_FRAMEBUFFER: @@ -526,7 +522,7 @@ main_loop(void) } if (Debug) - printf("SPU %u: Exit main loop\n", init.id); + printf("SPU %u: Exit main loop\n", spu.init.id); } @@ -556,7 +552,7 @@ main(unsigned long speid, unsigned long argp) if (Debug) printf("SPU: main() speid=%lu\n", speid); - mfc_get(&init, /* dest */ + mfc_get(&spu.init, /* dest */ (unsigned int) argp, /* src */ sizeof(struct cell_init_info), /* bytes */ tag, diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 9ef8cf0709..ea43224d02 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -30,10 +30,9 @@ #include "pipe/cell/common.h" +#include "pipe/p_state.h" -extern volatile struct cell_init_info init; - struct framebuffer { void *color_start; /**< addr of color surface in main memory */ void *depth_start; /**< addr of depth surface in main memory */ @@ -44,11 +43,27 @@ struct framebuffer { uint color_clear_value; uint depth_clear_value; -}; +} ALIGN16_ATTRIB; + + +/** + * All SPU global/context state will be in singleton object of this type: + */ +struct spu_global +{ + struct cell_init_info init; + + struct framebuffer fb; + struct pipe_depth_stencil_alpha_state depth_stencil; + struct pipe_blend_state blend; + /* XXX more state to come */ + +} ALIGN16_ATTRIB; + + +extern struct spu_global spu; -/* XXX Collect these globals in a struct: */ -extern struct framebuffer fb; /* DMA TAGS */ @@ -66,7 +81,7 @@ extern struct framebuffer fb; #define ASSERT(x) \ if (!(x)) { \ fprintf(stderr, "SPU %d: %s:%d: %s(): assertion %s failed.\n", \ - init.id, __FILE__, __LINE__, __FUNCTION__, #x); \ + spu.init.id, __FILE__, __LINE__, __FUNCTION__, #x); \ exit(1); \ } diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index af24e42435..b42b321c32 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -41,7 +41,6 @@ /** * Simplified types taken from other parts of Gallium */ - struct vertex_header { float data[2][4]; /* pos and color */ }; @@ -51,10 +50,6 @@ struct prim_header { }; - - -#if 1 - /* XXX fix this */ #undef CEILF #define CEILF(X) ((float) (int) ((X) + 0.99999)) @@ -70,10 +65,6 @@ struct prim_header { #define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) #define MASK_ALL 0xf -#define PIPE_MAX_SHADER_INPUTS 8 /* XXX temp */ - -#endif - #define DEBUG_VERTS 0 @@ -96,6 +87,7 @@ struct interp_coef float dady[4]; }; + /** * Triangle setup info (derived from draw_stage). * Also used for line drawing (taking some liberties). @@ -244,7 +236,7 @@ pack_color(const float color[4]) uint g = (uint) (color[1] * 255.0); uint b = (uint) (color[2] * 255.0); uint a = (uint) (color[3] * 255.0); - switch (fb.color_format) { + switch (spu.fb.color_format) { case PIPE_FORMAT_A8R8G8B8_UNORM: return (a << 24) | (r << 16) | (g << 8) | b; case PIPE_FORMAT_B8G8R8A8_UNORM: @@ -277,13 +269,13 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) eval_coeff(setup, 1, (float) x, (float) y, colors); - if (fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { float zvals[4]; eval_z(setup, (float) x, (float) y, zvals); if (tile_status_z[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ - clear_tile_z(ztile, fb.depth_clear_value); + clear_tile_z(ztile, spu.fb.depth_clear_value); } else { /* make sure we've got the tile from main mem */ @@ -327,7 +319,7 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) if (mask) { if (tile_status[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ - clear_tile(ctile, fb.color_clear_value); + clear_tile(ctile, spu.fb.color_clear_value); } else { /* make sure we've got the tile from main mem */ -- cgit v1.2.3 From 8b1bfd1d27989c9a880ce269d56ea1dfd88a5811 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 12 Jan 2008 16:57:55 -0700 Subject: Cell: disable cell_emit_state() for now --- src/mesa/pipe/cell/ppu/cell_state_derived.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c index dc2879b915..c654990325 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -219,7 +219,7 @@ void cell_update_derived( struct cell_context *cell ) compute_cliprect(cell); #endif - cell_emit_state(cell); + //cell_emit_state(cell); cell->dirty = 0; } -- cgit v1.2.3 From a511200e5f0c384e68258879bab76563d8e01f01 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 12 Jan 2008 16:58:50 -0700 Subject: Cell: remove fb parameter to get/put_tile() --- src/mesa/pipe/cell/spu/spu_main.c | 30 +++++++++++++++++------------- src/mesa/pipe/cell/spu/spu_main.h | 4 ++-- src/mesa/pipe/cell/spu/spu_tile.c | 22 ++++++++++------------ src/mesa/pipe/cell/spu/spu_tile.h | 6 ++---- 4 files changed, 31 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 7446cc3eff..0aee2aa6f9 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -83,13 +83,13 @@ really_clear_tiles(uint surfaceIndex) if (surfaceIndex == 0) { for (i = 0; i < TILE_SIZE; i++) for (j = 0; j < TILE_SIZE; j++) - ctile[i][j] = spu.fb.color_clear_value; + ctile[i][j] = spu.fb.color_clear_value; /*0xff00ff;*/ for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { uint tx = i % spu.fb.width_tiles; uint ty = i / spu.fb.width_tiles; if (tile_status[ty][tx] == TILE_STATUS_CLEAR) { - put_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); + put_tile(tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); } } } @@ -102,11 +102,11 @@ really_clear_tiles(uint surfaceIndex) uint tx = i % spu.fb.width_tiles; uint ty = i / spu.fb.width_tiles; if (tile_status_z[ty][tx] == TILE_STATUS_CLEAR) - put_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 1); + put_tile(tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 1); } } -#if 0 +#if 01 wait_on_mask(1 << TAG_SURFACE_CLEAR); #endif } @@ -156,9 +156,9 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) uint tx = i % spu.fb.width_tiles; uint ty = i / spu.fb.width_tiles; if (clear->surface == 0) - put_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); + put_tile(tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); else - put_tile(&spu.fb, tx, ty, (uint *) ztile, TAG_SURFACE_CLEAR, 1); + put_tile(tx, ty, (uint *) ztile, TAG_SURFACE_CLEAR, 1); /* XXX we don't want this here, but it fixes bad tile results */ } @@ -216,13 +216,15 @@ cmd_render(const struct cell_command_render *render) uint i, j, vertex_size, vertex_bytes, index_bytes; - if (Debug) + if (Debug) { printf("SPU %u: RENDER prim %u, indices: %u, nr_vert: %u\n", spu.init.id, render->prim_type, render->num_verts, render->num_indexes); - + printf(" bound: %g, %g .. %g, %g\n", + render->xmin, render->ymin, render->xmax, render->ymax); + } ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(render->index_data); @@ -289,12 +291,12 @@ cmd_render(const struct cell_command_render *render) */ if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(&spu.fb, tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); + get_tile(tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); } } if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); + get_tile(tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); } ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); @@ -312,12 +314,12 @@ cmd_render(const struct cell_command_render *render) /* write color/z tiles back to main framebuffer, if dirtied */ if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(&spu.fb, tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); + put_tile(tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); tile_status[ty][tx] = TILE_STATUS_DEFINED; } if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(&spu.fb, tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); + put_tile(tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); tile_status_z[ty][tx] = TILE_STATUS_DEFINED; } } @@ -361,7 +363,9 @@ cmd_state_depth_stencil(const struct pipe_depth_stencil_alpha_state *state) printf("SPU %u: DEPTH_STENCIL: ztest %d\n", spu.init.id, state->depth.enabled); - /* XXX copy/save the state */ + /* + memcpy(&spu.depth_stencil, state, sizeof(*state)); + */ } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index ea43224d02..75fb5b388b 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -33,7 +33,7 @@ #include "pipe/p_state.h" -struct framebuffer { +struct spu_framebuffer { void *color_start; /**< addr of color surface in main memory */ void *depth_start; /**< addr of depth surface in main memory */ enum pipe_format color_format; @@ -53,7 +53,7 @@ struct spu_global { struct cell_init_info init; - struct framebuffer fb; + struct spu_framebuffer fb; struct pipe_depth_stencil_alpha_state depth_stencil; struct pipe_blend_state blend; /* XXX more state to come */ diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c index 8fd9ab2905..99d779007e 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.c +++ b/src/mesa/pipe/cell/spu/spu_tile.c @@ -40,17 +40,16 @@ ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; void -get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, - int tag, int zBuf) +get_tile(uint tx, uint ty, uint *tile, int tag, int zBuf) { - const uint offset = ty * fb->width_tiles + tx; + const uint offset = ty * spu.fb.width_tiles + tx; const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); - const ubyte *src = zBuf ? fb->depth_start : fb->color_start; + const ubyte *src = zBuf ? spu.fb.depth_start : spu.fb.color_start; src += offset * bytesPerTile; - ASSERT(tx < fb->width_tiles); - ASSERT(ty < fb->height_tiles); + ASSERT(tx < spu.fb.width_tiles); + ASSERT(ty < spu.fb.height_tiles); ASSERT_ALIGN16(tile); /* printf("get_tile: dest: %p src: 0x%x size: %d\n", @@ -66,17 +65,16 @@ get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, void -put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, - int tag, int zBuf) +put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf) { - const uint offset = ty * fb->width_tiles + tx; + const uint offset = ty * spu.fb.width_tiles + tx; const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); - ubyte *dst = zBuf ? fb->depth_start : fb->color_start; + ubyte *dst = zBuf ? spu.fb.depth_start : spu.fb.color_start; dst += offset * bytesPerTile; - ASSERT(tx < fb->width_tiles); - ASSERT(ty < fb->height_tiles); + ASSERT(tx < spu.fb.width_tiles); + ASSERT(ty < spu.fb.height_tiles); ASSERT_ALIGN16(tile); /* printf("put_tile: src: %p dst: 0x%x size: %d\n", diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h index 2ad469b584..637923764b 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -52,12 +52,10 @@ extern ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_AT void -get_tile(const struct framebuffer *fb, uint tx, uint ty, uint *tile, - int tag, int zBuf); +get_tile(uint tx, uint ty, uint *tile, int tag, int zBuf); void -put_tile(const struct framebuffer *fb, uint tx, uint ty, const uint *tile, - int tag, int zBuf); +put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf); void clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value); -- cgit v1.2.3 From c76efb96b405e43e3261d1dc9e8812fdb2cfbac8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 10 Jan 2008 17:44:04 +0100 Subject: Remove mapping fields from struct pipe_surface. It's now the responsibility of surface users to keep track of their mappings. --- src/mesa/pipe/i915simple/i915_surface.c | 17 ++-- src/mesa/pipe/i965simple/brw_surface.c | 17 ++-- src/mesa/pipe/p_context.h | 4 +- src/mesa/pipe/p_inlines.h | 26 ++---- src/mesa/pipe/p_state.h | 2 - src/mesa/pipe/softpipe/sp_context.c | 26 ++---- src/mesa/pipe/softpipe/sp_state_surface.c | 42 ++------- src/mesa/pipe/softpipe/sp_surface.c | 16 ++-- src/mesa/pipe/softpipe/sp_tile_cache.c | 45 ++++++++-- src/mesa/pipe/softpipe/sp_tile_cache.h | 6 ++ src/mesa/pipe/util/p_tile.c | 128 +++++++++++++++++----------- src/mesa/pipe/xlib/xm_api.c | 21 ++--- src/mesa/state_tracker/st_cb_accum.c | 137 ++++-------------------------- src/mesa/state_tracker/st_cb_drawpixels.c | 6 -- src/mesa/state_tracker/st_cb_fbo.c | 3 - src/mesa/state_tracker/st_cb_readpixels.c | 4 - src/mesa/state_tracker/st_cb_texture.c | 13 +-- 17 files changed, 191 insertions(+), 322 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 79e74e1143..4e1b0929ee 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -171,10 +171,10 @@ i915_surface_copy(struct pipe_context *pipe, /* Fill a rectangular sub-region. Need better logic about when to * push buffers into AGP - will currently do so whenever possible. */ -static ubyte * -get_pointer(struct pipe_surface *dst, unsigned x, unsigned y) +static void * +get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) { - return dst->map + (y * dst->pitch + x) * dst->cpp; + return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; } @@ -186,12 +186,11 @@ i915_surface_fill(struct pipe_context *pipe, { if (0) { unsigned i, j; - - (void)pipe_surface_map(dst); + void *dst_map = pipe_surface_map(dst); switch (dst->cpp) { case 1: { - ubyte *row = get_pointer(dst, dstx, dsty); + ubyte *row = get_pointer(dst, dst_map, dstx, dsty); for (i = 0; i < height; i++) { memset(row, value, width); row += dst->pitch; @@ -199,7 +198,7 @@ i915_surface_fill(struct pipe_context *pipe, } break; case 2: { - ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + ushort *row = get_pointer(dst, dst_map, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = (ushort) value; @@ -208,7 +207,7 @@ i915_surface_fill(struct pipe_context *pipe, } break; case 4: { - unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); + unsigned *row = get_pointer(dst, dst_map, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = value; @@ -220,6 +219,8 @@ i915_surface_fill(struct pipe_context *pipe, assert(0); break; } + + pipe_surface_unmap( dst ); } else { i915_fill_blit( i915_context(pipe), diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c index d0e7229d5c..850423bdd5 100644 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -171,10 +171,10 @@ brw_surface_copy(struct pipe_context *pipe, /* Fill a rectangular sub-region. Need better logic about when to * push buffers into AGP - will currently do so whenever possible. */ -static ubyte * -get_pointer(struct pipe_surface *dst, unsigned x, unsigned y) +static void * +get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) { - return dst->map + (y * dst->pitch + x) * dst->cpp; + return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; } @@ -186,12 +186,11 @@ brw_surface_fill(struct pipe_context *pipe, { if (0) { unsigned i, j; - - (void)pipe_surface_map(dst); + void *dst_map = pipe_surface_map(dst); switch (dst->cpp) { case 1: { - ubyte *row = get_pointer(dst, dstx, dsty); + ubyte *row = get_pointer(dst, dst_map, dstx, dsty); for (i = 0; i < height; i++) { memset(row, value, width); row += dst->pitch; @@ -199,7 +198,7 @@ brw_surface_fill(struct pipe_context *pipe, } break; case 2: { - ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + ushort *row = get_pointer(dst, dst_map, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = (ushort) value; @@ -208,7 +207,7 @@ brw_surface_fill(struct pipe_context *pipe, } break; case 4: { - unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); + unsigned *row = get_pointer(dst, dst_map, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = value; @@ -220,6 +219,8 @@ brw_surface_fill(struct pipe_context *pipe, assert(0); break; } + + pipe_surface_unmap( dst ); } else { brw_fill_blit(brw_context(pipe), diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 1afb38a868..1a1f4f4e78 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -194,9 +194,9 @@ struct pipe_context { const void *p, int src_stride); /* XXX temporary here, move these to softpipe */ void (*get_tile_rgba)(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *p); + uint x, uint y, uint w, uint h, float *p); void (*put_tile_rgba)(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *p); + uint x, uint y, uint w, uint h, const float *p); /* diff --git a/src/mesa/pipe/p_inlines.h b/src/mesa/pipe/p_inlines.h index e7303c45c1..6976d087f9 100644 --- a/src/mesa/pipe/p_inlines.h +++ b/src/mesa/pipe/p_inlines.h @@ -33,35 +33,21 @@ #include "p_winsys.h" -static INLINE ubyte * +static INLINE void * pipe_surface_map(struct pipe_surface *surface) { - if (!surface->map_refcount++) { - surface->map - = (ubyte *) surface->winsys->buffer_map( surface->winsys, - surface->buffer, - PIPE_BUFFER_FLAG_WRITE | - PIPE_BUFFER_FLAG_READ ) - + surface->offset; - } - - return surface->map; + return (char *)surface->winsys->buffer_map( surface->winsys, surface->buffer, + PIPE_BUFFER_FLAG_WRITE | + PIPE_BUFFER_FLAG_READ ) + + surface->offset; } static INLINE void pipe_surface_unmap(struct pipe_surface *surface) { - if (surface->map_refcount > 0) { - assert(surface->map); - if (!--surface->map_refcount) { - surface->winsys->buffer_unmap( surface->winsys, - surface->buffer ); - surface->map = NULL; - } - } + surface->winsys->buffer_unmap( surface->winsys, surface->buffer ); } - /** * Set 'ptr' to point to 'surf' and update reference counting. * The old thing pointed to, if any, will be unreferenced first. diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index f9c5bfb6c6..ccd2a5f9e2 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -242,8 +242,6 @@ struct pipe_sampler_state struct pipe_surface { struct pipe_buffer_handle *buffer; /**< driver private buffer handle */ - ubyte *map; /**< only non-NULL when surface is actually mapped */ - unsigned map_refcount; /**< Reference count for mapping */ enum pipe_format format; /**< PIPE_FORMAT_x */ unsigned cpp; /**< bytes per pixel */ unsigned width, height; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 4f22539629..68c18e2d05 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -75,22 +75,15 @@ softpipe_is_format_supported( struct pipe_context *pipe, void softpipe_map_surfaces(struct softpipe_context *sp) { - struct pipe_surface *ps; unsigned i; for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - ps = sp->framebuffer.cbufs[i]; - if (ps->buffer && !ps->map) - pipe_surface_map(ps); + sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); } - ps = sp->framebuffer.zbuf; - if (ps && ps->buffer && !ps->map) - pipe_surface_map(ps); + sp_tile_cache_map_surfaces(sp->zbuf_cache); - ps = sp->framebuffer.sbuf; - if (ps && ps->buffer && !ps->map) - pipe_surface_map(ps); + sp_tile_cache_map_surfaces(sp->sbuf_cache); } @@ -100,7 +93,6 @@ softpipe_map_surfaces(struct softpipe_context *sp) void softpipe_unmap_surfaces(struct softpipe_context *sp) { - struct pipe_surface *ps; uint i; for (i = 0; i < sp->framebuffer.num_cbufs; i++) @@ -109,18 +101,12 @@ softpipe_unmap_surfaces(struct softpipe_context *sp) sp_flush_tile_cache(sp, sp->sbuf_cache); for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - ps = sp->framebuffer.cbufs[i]; - if (ps->map) - pipe_surface_unmap(ps); + sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); } - ps = sp->framebuffer.zbuf; - if (ps && ps->map) - pipe_surface_unmap(ps); + sp_tile_cache_unmap_surfaces(sp->zbuf_cache); - ps = sp->framebuffer.sbuf; - if (ps && ps->map) - pipe_surface_unmap(ps); + sp_tile_cache_unmap_surfaces(sp->sbuf_cache); } diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index ee72aaf4c5..4a9a28cc4d 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -38,7 +38,7 @@ /** * XXX this might get moved someday * Set the framebuffer surface info: color buffers, zbuffer, stencil buffer. - * Here, we map the surfaces and update the tile cache to point to the new + * Here, we flush the old surfaces and update the tile cache to point to the new * surfaces. */ void @@ -46,7 +46,6 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, const struct pipe_framebuffer_state *fb) { struct softpipe_context *sp = softpipe_context(pipe); - struct pipe_surface *ps; uint i; for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { @@ -54,19 +53,12 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) { /* flush old */ sp_flush_tile_cache(sp, sp->cbuf_cache[i]); - /* unmap old */ - ps = sp->framebuffer.cbufs[i]; - if (ps && ps->map) - pipe_surface_unmap(ps); - /* map new */ - ps = fb->cbufs[i]; - if (ps) - pipe_surface_map(ps); + /* assign new */ sp->framebuffer.cbufs[i] = fb->cbufs[i]; /* update cache */ - sp_tile_cache_set_surface(sp->cbuf_cache[i], ps); + sp_tile_cache_set_surface(sp->cbuf_cache[i], fb->cbufs[i]); } } @@ -76,23 +68,12 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, if (sp->framebuffer.zbuf != fb->zbuf) { /* flush old */ sp_flush_tile_cache(sp, sp->zbuf_cache); - /* unmap old */ - ps = sp->framebuffer.zbuf; - if (ps && ps->map) - pipe_surface_unmap(ps); - if (sp->framebuffer.sbuf == sp->framebuffer.zbuf) { - /* combined z/stencil */ - sp->framebuffer.sbuf = NULL; - } - /* map new */ - ps = fb->zbuf; - if (ps) - pipe_surface_map(ps); + /* assign new */ sp->framebuffer.zbuf = fb->zbuf; /* update cache */ - sp_tile_cache_set_surface(sp->zbuf_cache, ps); + sp_tile_cache_set_surface(sp->zbuf_cache, fb->zbuf); } /* XXX combined depth/stencil here */ @@ -101,14 +82,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, if (sp->framebuffer.sbuf != fb->sbuf) { /* flush old */ sp_flush_tile_cache(sp, sp->sbuf_cache_sep); - /* unmap old */ - ps = sp->framebuffer.sbuf; - if (ps && ps->map) - pipe_surface_unmap(ps); - /* map new */ - ps = fb->sbuf; - if (ps && fb->sbuf != fb->zbuf) - pipe_surface_map(ps); + /* assign new */ sp->framebuffer.sbuf = fb->sbuf; @@ -116,12 +90,12 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, if (fb->sbuf != fb->zbuf) { /* separate stencil buf */ sp->sbuf_cache = sp->sbuf_cache_sep; - sp_tile_cache_set_surface(sp->sbuf_cache, ps); + sp_tile_cache_set_surface(sp->sbuf_cache, fb->sbuf); } else { /* combined depth/stencil */ sp->sbuf_cache = sp->zbuf_cache; - sp_tile_cache_set_surface(sp->sbuf_cache, ps); + sp_tile_cache_set_surface(sp->sbuf_cache, fb->sbuf); } } diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index ece30e36ec..b44ba3e957 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -158,10 +158,10 @@ sp_surface_copy(struct pipe_context *pipe, } -static ubyte * -get_pointer(struct pipe_surface *dst, unsigned x, unsigned y) +static void * +get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) { - return dst->map + (y * dst->pitch + x) * dst->cpp; + return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; } @@ -179,16 +179,16 @@ sp_surface_fill(struct pipe_context *pipe, unsigned width, unsigned height, unsigned value) { unsigned i, j; + void *dst_map = pipe_surface_map(dst); assert(dst->pitch > 0); assert(width <= dst->pitch); - (void)pipe_surface_map(dst); switch (dst->cpp) { case 1: { - ubyte *row = get_pointer(dst, dstx, dsty); + ubyte *row = get_pointer(dst, dst_map, dstx, dsty); for (i = 0; i < height; i++) { memset(row, value, width); row += dst->pitch; @@ -197,7 +197,7 @@ sp_surface_fill(struct pipe_context *pipe, break; case 2: { - ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + ushort *row = get_pointer(dst, dst_map, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = (ushort) value; @@ -207,7 +207,7 @@ sp_surface_fill(struct pipe_context *pipe, break; case 4: { - unsigned *row = (unsigned *) get_pointer(dst, dstx, dsty); + unsigned *row = get_pointer(dst, dst_map, dstx, dsty); for (i = 0; i < height; i++) { for (j = 0; j < width; j++) row[j] = value; @@ -218,7 +218,7 @@ sp_surface_fill(struct pipe_context *pipe, case 8: { /* expand the 4-byte clear value to an 8-byte value */ - ushort *row = (ushort *) get_pointer(dst, dstx, dsty); + ushort *row = (ushort *) get_pointer(dst, dst_map, dstx, dsty); ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 1dbcc5aadd..d6f60807e6 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -49,6 +49,7 @@ struct softpipe_tile_cache { struct pipe_surface *surface; /**< the surface we're caching */ + void *surface_map; struct pipe_texture *texture; /**< if caching a texture */ struct softpipe_cached_tile entries[NUM_ENTRIES]; uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; @@ -57,6 +58,7 @@ struct softpipe_tile_cache boolean depth_stencil; /** Is the surface a depth/stencil format? */ struct pipe_surface *tex_surf; + void *tex_surf_map; int tex_face, tex_level, tex_z; struct softpipe_cached_tile tile; /**< scratch tile for clears */ @@ -150,7 +152,7 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, { assert(!tc->texture); - if (tc->surface && tc->surface->map) { + if (tc->surface_map) { /*assert(tc->surface != ps);*/ pipe_surface_unmap(tc->surface); } @@ -158,8 +160,8 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, pipe_surface_reference(&tc->surface, ps); if (ps) { - if (!ps->map) - pipe_surface_map(ps); + if (tc->surface_map) + tc->surface_map = pipe_surface_map(ps); tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || ps->format == PIPE_FORMAT_Z16_UNORM || @@ -179,6 +181,32 @@ sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) } +void +sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc) +{ + if (tc->surface && !tc->surface_map) + tc->surface_map = pipe_surface_map(tc->surface); + + if (tc->tex_surf && !tc->tex_surf_map) + tc->tex_surf_map = pipe_surface_map(tc->tex_surf); +} + + +void +sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc) +{ + if (tc->surface_map) { + pipe_surface_unmap(tc->surface); + tc->surface_map = NULL; + } + + if (tc->tex_surf_map) { + pipe_surface_unmap(tc->tex_surf); + tc->tex_surf_map = NULL; + } +} + + /** * Specify the texture to cache. */ @@ -192,8 +220,10 @@ sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, tc->texture = texture; - if (tc->tex_surf && tc->tex_surf->map) + if (tc->tex_surf_map) { pipe_surface_unmap(tc->tex_surf); + tc->tex_surf_map = NULL; + } pipe_surface_reference(&tc->tex_surf, NULL); /* mark as entries as invalid/empty */ @@ -330,9 +360,6 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, if (!ps || !ps->buffer) return; - if (!ps->map) - pipe_surface_map(ps); - for (pos = 0; pos < NUM_ENTRIES; pos++) { struct softpipe_cached_tile *tile = tc->entries + pos; if (tile->x >= 0) { @@ -475,11 +502,11 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, tc->tex_z != z) { /* get new surface (view into texture) */ - if (tc->tex_surf && tc->tex_surf->map) + if (tc->tex_surf_map) pipe_surface_unmap(tc->tex_surf); tc->tex_surf = pipe->get_tex_surface(pipe, tc->texture, face, level, z); - pipe_surface_map(tc->tex_surf); + tc->tex_surf_map = pipe_surface_map(tc->tex_surf); tc->tex_face = face; tc->tex_level = level; diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h index 91fb2795a2..7fd1081286 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ b/src/mesa/pipe/softpipe/sp_tile_cache.h @@ -73,6 +73,12 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, extern struct pipe_surface * sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); +extern void +sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc); + +extern void +sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc); + extern void sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, struct pipe_texture *texture); diff --git a/src/mesa/pipe/util/p_tile.c b/src/mesa/pipe/util/p_tile.c index 8bc3c22c83..f2e19f19dc 100644 --- a/src/mesa/pipe/util/p_tile.c +++ b/src/mesa/pipe/util/p_tile.c @@ -56,8 +56,6 @@ pipe_get_tile_raw(struct pipe_context *pipe, ubyte *pDest; uint i; - assert(ps->map); - if (dst_stride == 0) { dst_stride = w * cpp; } @@ -65,7 +63,7 @@ pipe_get_tile_raw(struct pipe_context *pipe, if (pipe_clip_tile(x, y, &w, &h, ps)) return; - pSrc = ps->map + (y * ps->pitch + x) * cpp; + pSrc = (const ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; pDest = (ubyte *) p; for (i = 0; i < h; i++) { @@ -73,6 +71,8 @@ pipe_get_tile_raw(struct pipe_context *pipe, pDest += dst_stride; pSrc += src_stride; } + + pipe_surface_unmap(ps); } @@ -92,8 +92,6 @@ pipe_put_tile_raw(struct pipe_context *pipe, ubyte *pDest; uint i; - assert(ps->map); - if (src_stride == 0) { src_stride = w * cpp; } @@ -102,13 +100,15 @@ pipe_put_tile_raw(struct pipe_context *pipe, return; pSrc = (const ubyte *) p; - pDest = ps->map + (y * ps->pitch + x) * cpp; + pDest = (ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w * cpp); pDest += dst_stride; pSrc += src_stride; } + + pipe_surface_unmap(ps); } @@ -126,11 +126,12 @@ pipe_put_tile_raw(struct pipe_context *pipe, static void a8r8g8b8_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const unsigned *src - = ((const unsigned *) (ps->map)) + = ((const unsigned *) map) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -158,11 +159,12 @@ a8r8g8b8_get_tile_rgba(struct pipe_surface *ps, static void a8r8g8b8_put_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, const float *p) { unsigned *dst - = ((unsigned *) (ps->map)) + = ((unsigned *) map) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -193,11 +195,12 @@ a8r8g8b8_put_tile_rgba(struct pipe_surface *ps, static void b8g8r8a8_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const unsigned *src - = ((const unsigned *) (ps->map)) + = ((const unsigned *) map) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -225,11 +228,12 @@ b8g8r8a8_get_tile_rgba(struct pipe_surface *ps, static void b8g8r8a8_put_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, const float *p) { unsigned *dst - = ((unsigned *) (ps->map)) + = ((unsigned *) map) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -260,11 +264,12 @@ b8g8r8a8_put_tile_rgba(struct pipe_surface *ps, static void a1r5g5b5_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ushort *src - = ((const ushort *) (ps->map)) + = ((const ushort *) map) + y * ps->pitch + x; unsigned i, j; @@ -288,11 +293,12 @@ a1r5g5b5_get_tile_rgba(struct pipe_surface *ps, static void a4r4g4b4_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ushort *src - = ((const ushort *) (ps->map)) + = ((const ushort *) map) + y * ps->pitch + x; unsigned i, j; @@ -316,11 +322,12 @@ a4r4g4b4_get_tile_rgba(struct pipe_surface *ps, static void r5g6b5_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ushort *src - = ((const ushort *) (ps->map)) + = ((const ushort *) map) + y * ps->pitch + x; unsigned i, j; @@ -342,11 +349,12 @@ r5g6b5_get_tile_rgba(struct pipe_surface *ps, static void r5g5b5_put_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, const float *p) { ushort *dst - = ((ushort *) (ps->map)) + = ((ushort *) map) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -379,11 +387,12 @@ r5g5b5_put_tile_rgba(struct pipe_surface *ps, */ static void z16_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ushort *src - = ((const ushort *) (ps->map)) + = ((const ushort *) map) + y * ps->pitch + x; const float scale = 1.0f / 65535.0f; unsigned i, j; @@ -414,11 +423,12 @@ z16_get_tile_rgba(struct pipe_surface *ps, static void l8_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ubyte *src - = ((const ubyte *) (ps->map)) + = ((const ubyte *) map) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -447,11 +457,12 @@ l8_get_tile_rgba(struct pipe_surface *ps, static void a8_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ubyte *src - = ((const ubyte *) (ps->map)) + = ((const ubyte *) map) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -480,11 +491,12 @@ a8_get_tile_rgba(struct pipe_surface *ps, static void r16g16b16a16_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const short *src - = ((const short *) (ps->map)) + = ((const short *) map) + (y * ps->pitch + x) * 4; unsigned i, j; unsigned w0 = w; @@ -513,11 +525,12 @@ r16g16b16a16_get_tile_rgba(struct pipe_surface *ps, static void r16g16b16a16_put_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, const float *p) { short *dst - = ((short *) (ps->map)) + = ((short *) map) + (y * ps->pitch + x) * 4; unsigned i, j; unsigned w0 = w; @@ -552,11 +565,12 @@ r16g16b16a16_put_tile_rgba(struct pipe_surface *ps, static void i8_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ubyte *src - = ((const ubyte *) (ps->map)) + = ((const ubyte *) map) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -585,11 +599,12 @@ i8_get_tile_rgba(struct pipe_surface *ps, static void a8_l8_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const ushort *src - = ((const ushort *) (ps->map)) + = ((const ushort *) map) + y * ps->pitch + x; unsigned i, j; unsigned w0 = w; @@ -624,11 +639,12 @@ a8_l8_get_tile_rgba(struct pipe_surface *ps, */ static void z32_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const uint *src - = ((const uint *) (ps->map)) + = ((const uint *) map) + y * ps->pitch + x; const double scale = 1.0 / (double) 0xffffffff; unsigned i, j; @@ -660,11 +676,12 @@ z32_get_tile_rgba(struct pipe_surface *ps, */ static void s8z24_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const uint *src - = ((const uint *) (ps->map)) + = ((const uint *) map) + y * ps->pitch + x; const double scale = 1.0 / ((1 << 24) - 1); unsigned i, j; @@ -696,11 +713,12 @@ s8z24_get_tile_rgba(struct pipe_surface *ps, */ static void z24s8_get_tile_rgba(struct pipe_surface *ps, + void *map, unsigned x, unsigned y, unsigned w, unsigned h, float *p) { const uint *src - = ((const uint *) (ps->map)) + = ((const uint *) map) + y * ps->pitch + x; const double scale = 1.0 / ((1 << 24) - 1); unsigned i, j; @@ -731,52 +749,56 @@ pipe_get_tile_rgba(struct pipe_context *pipe, uint x, uint y, uint w, uint h, float *p) { + void *map = pipe_surface_map(ps); + switch (ps->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_get_tile_rgba(ps, x, y, w, h, p); + a8r8g8b8_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_get_tile_rgba(ps, x, y, w, h, p); + b8g8r8a8_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_A1R5G5B5_UNORM: - a1r5g5b5_get_tile_rgba(ps, x, y, w, h, p); + a1r5g5b5_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_A4R4G4B4_UNORM: - a4r4g4b4_get_tile_rgba(ps, x, y, w, h, p); + a4r4g4b4_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_R5G6B5_UNORM: - r5g6b5_get_tile_rgba(ps, x, y, w, h, p); + r5g6b5_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_U_L8: - l8_get_tile_rgba(ps, x, y, w, h, p); + l8_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_U_A8: - a8_get_tile_rgba(ps, x, y, w, h, p); + a8_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_U_I8: - i8_get_tile_rgba(ps, x, y, w, h, p); + i8_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_U_A8_L8: - a8_l8_get_tile_rgba(ps, x, y, w, h, p); + a8_l8_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_get_tile_rgba(ps, x, y, w, h, p); + r16g16b16a16_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_Z16_UNORM: - z16_get_tile_rgba(ps, x, y, w, h, p); + z16_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_Z32_UNORM: - z32_get_tile_rgba(ps, x, y, w, h, p); + z32_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_S8Z24_UNORM: - s8z24_get_tile_rgba(ps, x, y, w, h, p); + s8z24_get_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_Z24S8_UNORM: - z24s8_get_tile_rgba(ps, x, y, w, h, p); + z24s8_get_tile_rgba(ps, map, x, y, w, h, p); break; default: assert(0); } + + pipe_surface_unmap(ps); } @@ -786,49 +808,53 @@ pipe_put_tile_rgba(struct pipe_context *pipe, uint x, uint y, uint w, uint h, const float *p) { + void *map = pipe_surface_map(ps); + switch (ps->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_put_tile_rgba(ps, x, y, w, h, p); + a8r8g8b8_put_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_put_tile_rgba(ps, x, y, w, h, p); + b8g8r8a8_put_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_A1R5G5B5_UNORM: - /*a1r5g5b5_put_tile_rgba(ps, x, y, w, h, p);*/ + /*a1r5g5b5_put_tile_rgba(ps, map, x, y, w, h, p);*/ break; case PIPE_FORMAT_R5G6B5_UNORM: - r5g5b5_put_tile_rgba(ps, x, y, w, h, p); + r5g5b5_put_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_R8G8B8A8_UNORM: break; case PIPE_FORMAT_U_L8: - /*l8_put_tile_rgba(ps, x, y, w, h, p);*/ + /*l8_put_tile_rgba(ps, map, x, y, w, h, p);*/ break; case PIPE_FORMAT_U_A8: - /*a8_put_tile_rgba(ps, x, y, w, h, p);*/ + /*a8_put_tile_rgba(ps, map, x, y, w, h, p);*/ break; case PIPE_FORMAT_U_I8: - /*i8_put_tile_rgba(ps, x, y, w, h, p);*/ + /*i8_put_tile_rgba(ps, map, x, y, w, h, p);*/ break; case PIPE_FORMAT_U_A8_L8: - /*a8_l8_put_tile_rgba(ps, x, y, w, h, p);*/ + /*a8_l8_put_tile_rgba(ps, map, x, y, w, h, p);*/ break; case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_put_tile_rgba(ps, x, y, w, h, p); + r16g16b16a16_put_tile_rgba(ps, map, x, y, w, h, p); break; case PIPE_FORMAT_Z16_UNORM: - /*z16_put_tile_rgba(ps, x, y, w, h, p);*/ + /*z16_put_tile_rgba(ps, map, x, y, w, h, p);*/ break; case PIPE_FORMAT_Z32_UNORM: - /*z32_put_tile_rgba(ps, x, y, w, h, p);*/ + /*z32_put_tile_rgba(ps, map, x, y, w, h, p);*/ break; case PIPE_FORMAT_S8Z24_UNORM: - /*s8z24_put_tile_rgba(ps, x, y, w, h, p);*/ + /*s8z24_put_tile_rgba(ps, map, x, y, w, h, p);*/ break; case PIPE_FORMAT_Z24S8_UNORM: - /*z24s8_put_tile_rgba(ps, x, y, w, h, p);*/ + /*z24s8_put_tile_rgba(ps, map, x, y, w, h, p);*/ break; default: assert(0); } + + pipe_surface_unmap(ps); } diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 168eba0784..03985eab5a 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1247,22 +1247,11 @@ void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height ) GLboolean XMesaGetDepthBuffer( XMesaBuffer b, GLint *width, GLint *height, GLint *bytesPerValue, void **buffer ) { - struct pipe_surface *surf - = st_get_framebuffer_surface(b->stfb, ST_SURFACE_DEPTH); - if (surf) { - *width = surf->width; - *height = surf->pitch; - *bytesPerValue = surf->cpp; - *buffer = surf->map; - return GL_TRUE; - } - else { - *width = 0; - *height = 0; - *bytesPerValue = 0; - *buffer = 0; - return GL_FALSE; - } + *width = 0; + *height = 0; + *bytesPerValue = 0; + *buffer = 0; + return GL_FALSE; } diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index cf2e9db51c..73cd7db18d 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -59,6 +59,7 @@ void st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { + struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *acc_strb = st_renderbuffer(rb); struct pipe_surface *acc_ps = acc_strb->surface; const GLint xpos = ctx->DrawBuffer->_Xmin; @@ -69,102 +70,17 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) const GLfloat g = ctx->Accum.ClearColor[1]; const GLfloat b = ctx->Accum.ClearColor[2]; const GLfloat a = ctx->Accum.ClearColor[3]; + GLfloat *accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + int i; - (void) pipe_surface_map(acc_ps); - - switch (acc_ps->format) { - case PIPE_FORMAT_R16G16B16A16_SNORM: - { - const short sr = (short) (32767 * r); - const short sg = (short) (32767 * g); - const short sb = (short) (32767 * b); - const short sa = (short) (32767 * a); - short *acc = ((short *) acc_ps->map) - + (ypos * acc_ps->pitch + xpos) * 4; - int i, j; - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) { - acc[j*4+0] = sr; - acc[j*4+1] = sg; - acc[j*4+2] = sb; - acc[j*4+3] = sa; - } - acc += acc_ps->pitch * 4; - } - } - break; - default: - assert(0); + for (i = 0; i < width * height; i++) { + accBuf[i*4+0] = r; + accBuf[i*4+1] = g; + accBuf[i*4+2] = b; + accBuf[i*4+3] = a; } - pipe_surface_unmap(acc_ps); -} - - -/** Get block of values from accum buffer, converting to float */ -static void -get_accum_tile(struct pipe_context *pipe, - struct pipe_surface *acc_surf, - int xpos, int ypos, int width, int height, - float *buf) -{ - switch (acc_surf->format) { - case PIPE_FORMAT_R16G16B16A16_SNORM: - { - const short *acc = ((const short *) acc_surf->map) - + (ypos * acc_surf->pitch + xpos) * 4; - int i, j; - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) { - buf[j*4+0] = SHORT_TO_FLOAT(acc[j*4+0]); - buf[j*4+1] = SHORT_TO_FLOAT(acc[j*4+1]); - buf[j*4+2] = SHORT_TO_FLOAT(acc[j*4+2]); - buf[j*4+3] = SHORT_TO_FLOAT(acc[j*4+3]); - } - acc += acc_surf->pitch * 4; - buf += width * 4; - } - } - break; - default: - assert(0); - } -} - - -/** Put block of values into accum buffer, converting from float */ -static void -put_accum_tile(struct pipe_context *pipe, - struct pipe_surface *acc_surf, - int xpos, int ypos, int width, int height, - const float *buf) -{ - switch (acc_surf->format) { - case PIPE_FORMAT_R16G16B16A16_SNORM: - { - short *acc = ((short *) acc_surf->map) - + (ypos * acc_surf->pitch + xpos) * 4; - int i, j; - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) { - short r, g, b, a; - UNCLAMPED_FLOAT_TO_SHORT(r, buf[j*4+0]); - UNCLAMPED_FLOAT_TO_SHORT(g, buf[j*4+1]); - UNCLAMPED_FLOAT_TO_SHORT(b, buf[j*4+2]); - UNCLAMPED_FLOAT_TO_SHORT(a, buf[j*4+3]); - acc[j*4+0] = r; - acc[j*4+1] = g; - acc[j*4+2] = b; - acc[j*4+3] = a; - } - acc += acc_surf->pitch * 4; - buf += width * 4; - } - } - break; - default: - assert(0); - } + pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); } @@ -179,19 +95,15 @@ accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - (void) pipe_surface_map(acc_ps); - - get_accum_tile(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] * scale + bias; } - put_accum_tile(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(accBuf); - - pipe_surface_unmap(acc_ps); } @@ -201,30 +113,23 @@ accum_accum(struct pipe_context *pipe, GLfloat value, struct pipe_surface *acc_ps, struct pipe_surface *color_ps) { - ubyte *colorMap, *accMap; GLfloat *colorBuf, *accBuf; GLint i; colorBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - colorMap = pipe_surface_map(color_ps); - accMap = pipe_surface_map(acc_ps); - pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, colorBuf); - get_accum_tile(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] + colorBuf[i] * value; } - put_accum_tile(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(colorBuf); free(accBuf); - - pipe_surface_unmap(color_ps); - pipe_surface_unmap(acc_ps); } @@ -239,21 +144,15 @@ accum_load(struct pipe_context *pipe, GLfloat value, buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - (void) pipe_surface_map(color_ps); - (void) pipe_surface_map(acc_ps); - pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, buf); for (i = 0; i < 4 * width * height; i++) { buf[i] = buf[i] * value; } - put_accum_tile(pipe, acc_ps, xpos, ypos, width, height, buf); + pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, buf); free(buf); - - pipe_surface_unmap(color_ps); - pipe_surface_unmap(acc_ps); } @@ -270,10 +169,7 @@ accum_return(GLcontext *ctx, GLfloat value, abuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - (void) pipe_surface_map(color_ps); - (void) pipe_surface_map(acc_ps); - - get_accum_tile(pipe, acc_ps, xpos, ypos, width, height, abuf); + pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) { cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); @@ -297,9 +193,6 @@ accum_return(GLcontext *ctx, GLfloat value, free(abuf); if (cbuf) free(cbuf); - - pipe_surface_unmap(color_ps); - pipe_surface_unmap(acc_ps); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 2db12c653b..5ea2f08b8a 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1261,15 +1261,9 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, /* alternate path using get/put_tile() */ GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - (void) pipe_surface_map(psRead); - (void) pipe_surface_map(psTex); - pipe->get_tile_rgba(pipe, psRead, srcx, srcy, width, height, buf); pipe->put_tile_rgba(pipe, psTex, 0, 0, width, height, buf); - pipe_surface_unmap(psRead); - pipe_surface_unmap(psTex); - free(buf); } diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index bc0f26ffb0..2d6b3fc749 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -97,9 +97,6 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, return GL_FALSE; } - /* loop here since mapping is refcounted */ - while (strb->surface->map) - pipe_surface_unmap(strb->surface); if (strb->surface->buffer) pipe->winsys->buffer_reference(pipe->winsys, &strb->surface->buffer, NULL); diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index b0c9275c6f..585fe04dee 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -179,8 +179,6 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (!strb) return; - pipe_surface_map(strb->surface); - if (format == GL_RGBA && type == GL_FLOAT) { /* write tile(row) directly into user's buffer */ df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width, @@ -229,8 +227,6 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } } } - - pipe_surface_unmap(strb->surface); } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 773fc0012e..19274570a1 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1063,15 +1063,13 @@ fallback_copy_texsubimage(GLcontext *ctx, dest_surf = pipe->get_tex_surface(pipe, pt, face, level, destZ); - (void) pipe_surface_map(dest_surf); - (void) pipe_surface_map(src_surf); - /* buffer for one row */ data = (GLfloat *) malloc(width * 4 * sizeof(GLfloat)); /* do copy row by row */ for (row = 0; row < height; row++) { - pipe->get_tile_rgba(pipe, src_surf, srcX, srcY + row, width, 1, data); + pipe->get_tile_rgba(pipe, src_surf, srcX, srcY + row, width, 1, + data); /* XXX we're ignoring convolution for now */ if (ctx->_ImageTransferState) { @@ -1080,14 +1078,11 @@ fallback_copy_texsubimage(GLcontext *ctx, width, (GLfloat (*)[4])data); } - pipe->put_tile_rgba(pipe, dest_surf, destX, destY, width, 1, data); + pipe->put_tile_rgba(pipe, dest_surf, destX, destY, width, 1, + data); destY += yStep; } - - (void) pipe_surface_unmap(dest_surf); - (void) pipe_surface_unmap(src_surf); - free(data); } -- cgit v1.2.3 From 2014e0bacbd2661bf98d084120a109b1c0bf0df2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Mon, 14 Jan 2008 16:17:01 +0100 Subject: Remove pipe->get/put_tile_rgba. pipe_get/put_tile_rgba() now use pipe->get/put_tile internally. Also simplify the _get/put_tile_rgba() helper functions and clean up some inconsitencies in them. --- src/mesa/pipe/i915simple/i915_surface.c | 2 - src/mesa/pipe/i965simple/brw_surface.c | 2 - src/mesa/pipe/p_context.h | 5 - src/mesa/pipe/softpipe/sp_surface.c | 2 - src/mesa/pipe/softpipe/sp_tile_cache.c | 25 +- src/mesa/pipe/util/p_tile.c | 585 +++++++++++------------------- src/mesa/state_tracker/st_cb_accum.c | 23 +- src/mesa/state_tracker/st_cb_drawpixels.c | 5 +- src/mesa/state_tracker/st_cb_readpixels.c | 3 +- src/mesa/state_tracker/st_cb_texture.c | 7 +- 10 files changed, 245 insertions(+), 414 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 4e1b0929ee..e3c3cdd2e4 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -240,8 +240,6 @@ i915_init_surface_functions(struct i915_context *i915) i915->pipe.get_tex_surface = i915_get_tex_surface; i915->pipe.get_tile = pipe_get_tile_raw; i915->pipe.put_tile = pipe_put_tile_raw; - i915->pipe.get_tile_rgba = pipe_get_tile_rgba; - i915->pipe.put_tile_rgba = pipe_put_tile_rgba; i915->pipe.surface_data = i915_surface_data; i915->pipe.surface_copy = i915_surface_copy; diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c index 850423bdd5..4eacbdf82b 100644 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -239,8 +239,6 @@ brw_init_surface_functions(struct brw_context *brw) brw->pipe.get_tex_surface = brw_get_tex_surface; brw->pipe.get_tile = pipe_get_tile_raw; brw->pipe.put_tile = pipe_put_tile_raw; - brw->pipe.get_tile_rgba = pipe_get_tile_rgba; - brw->pipe.put_tile_rgba = pipe_put_tile_rgba; brw->pipe.surface_data = brw_surface_data; brw->pipe.surface_copy = brw_surface_copy; diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 1a1f4f4e78..25b5dc35e4 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -192,11 +192,6 @@ struct pipe_context { struct pipe_surface *ps, uint x, uint y, uint w, uint h, const void *p, int src_stride); - /* XXX temporary here, move these to softpipe */ - void (*get_tile_rgba)(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, float *p); - void (*put_tile_rgba)(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, const float *p); /* diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index b44ba3e957..6c080d5b5c 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -252,8 +252,6 @@ sp_init_surface_functions(struct softpipe_context *sp) { sp->pipe.get_tile = pipe_get_tile_raw; sp->pipe.put_tile = pipe_put_tile_raw; - sp->pipe.get_tile_rgba = pipe_get_tile_rgba; - sp->pipe.put_tile_rgba = pipe_put_tile_rgba; sp->pipe.surface_data = sp_surface_data; sp->pipe.surface_copy = sp_surface_copy; diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index d6f60807e6..6515ce668c 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -34,6 +34,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "pipe/util/p_tile.h" #include "sp_context.h" #include "sp_surface.h" #include "sp_tile_cache.h" @@ -369,9 +370,9 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, tile->data.depth32, 0/*STRIDE*/); } else { - pipe->put_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + pipe_put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); } tile->x = tile->y = -1; /* mark as empty */ inuse++; @@ -418,9 +419,9 @@ sp_get_cached_tile(struct softpipe_context *softpipe, tile->data.depth32, 0/*STRIDE*/); } else { - pipe->put_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + pipe_put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); } } @@ -445,9 +446,9 @@ sp_get_cached_tile(struct softpipe_context *softpipe, tile->data.depth32, 0/*STRIDE*/); } else { - pipe->get_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + pipe_get_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); } } } @@ -514,9 +515,9 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, } /* get tile from the surface (view into texture) */ - pipe->get_tile_rgba(pipe, tc->tex_surf, - tile_x, tile_y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + pipe_get_tile_rgba(pipe, tc->tex_surf, + tile_x, tile_y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); tile->x = tile_x; tile->y = tile_y; tile->z = z; diff --git a/src/mesa/pipe/util/p_tile.c b/src/mesa/pipe/util/p_tile.c index f2e19f19dc..85a863db8a 100644 --- a/src/mesa/pipe/util/p_tile.c +++ b/src/mesa/pipe/util/p_tile.c @@ -125,68 +125,46 @@ pipe_put_tile_raw(struct pipe_context *pipe, /*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ static void -a8r8g8b8_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +a8r8g8b8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const unsigned *src - = ((const unsigned *) map) - + y * ps->pitch + x; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow += 4; } - src += ps->pitch; - p += w0 * 4; + p += dst_stride; } } static void -a8r8g8b8_put_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) +a8r8g8b8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) { - unsigned *dst - = ((unsigned *) map) - + y * ps->pitch + x; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_A8R8G8B8_UNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { const float *pRow = p; - for (j = 0; j < w; j++) { + for (j = 0; j < w; j++, pRow += 4) { unsigned r, g, b, a; UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - dst[j] = (a << 24) | (r << 16) | (g << 8) | b; - pRow += 4; + *dst++ = (a << 24) | (r << 16) | (g << 8) | b; } - dst += ps->pitch; - p += w0 * 4; + p += src_stride; } } @@ -194,68 +172,46 @@ a8r8g8b8_put_tile_rgba(struct pipe_surface *ps, /*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ static void -b8g8r8a8_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +b8g8r8a8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const unsigned *src - = ((const unsigned *) map) - + y * ps->pitch + x; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { - const unsigned pixel = src[j]; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow += 4; } - src += ps->pitch; - p += w0 * 4; + p += dst_stride; } } static void -b8g8r8a8_put_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) +b8g8r8a8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) { - unsigned *dst - = ((unsigned *) map) - + y * ps->pitch + x; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_B8G8R8A8_UNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { const float *pRow = p; - for (j = 0; j < w; j++) { + for (j = 0; j < w; j++, pRow += 4) { unsigned r, g, b, a; UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - dst[j] = (b << 24) | (g << 16) | (r << 8) | a; - pRow += 4; + *dst++ = (b << 24) | (g << 16) | (r << 8) | a; } - dst += ps->pitch; - p += w0 * 4; + p += src_stride; } } @@ -263,28 +219,23 @@ b8g8r8a8_put_tile_rgba(struct pipe_surface *ps, /*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ static void -a1r5g5b5_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +a1r5g5b5_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const ushort *src - = ((const ushort *) map) - + y * ps->pitch + x; unsigned i, j; - assert(ps->format == PIPE_FORMAT_A1R5G5B5_UNORM); - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - const ushort pixel = src[j]; - p[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); - p[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); - p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - p[3] = ((pixel >> 15) ) * 1.0f; - p += 4; + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); + pRow[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); + pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + pRow[3] = ((pixel >> 15) ) * 1.0f; } - src += ps->pitch; + p += dst_stride; } } @@ -292,28 +243,23 @@ a1r5g5b5_get_tile_rgba(struct pipe_surface *ps, /*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ static void -a4r4g4b4_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +a4r4g4b4_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const ushort *src - = ((const ushort *) map) - + y * ps->pitch + x; unsigned i, j; - assert(ps->format == PIPE_FORMAT_A4R4G4B4_UNORM); - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - const ushort pixel = src[j]; - p[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); - p[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); - p[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); - p[3] = ((pixel >> 12) ) * (1.0f / 15.0f); - p += 4; + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); + pRow[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); + pRow[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); + pRow[3] = ((pixel >> 12) ) * (1.0f / 15.0f); } - src += ps->pitch; + p += dst_stride; } } @@ -321,60 +267,44 @@ a4r4g4b4_get_tile_rgba(struct pipe_surface *ps, /*** PIPE_FORMAT_R5G6B5_UNORM ***/ static void -r5g6b5_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +r5g6b5_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const ushort *src - = ((const ushort *) map) - + y * ps->pitch + x; unsigned i, j; - assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - const ushort pixel = src[j]; - p[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); - p[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); - p[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - p[3] = 1.0f; - p += 4; + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); + pRow[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); + pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + pRow[3] = 1.0f; } - src += ps->pitch; + p += dst_stride; } } static void -r5g5b5_put_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) +r5g5b5_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) { - ushort *dst - = ((ushort *) map) - + y * ps->pitch + x; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_R5G6B5_UNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { const float *pRow = p; - for (j = 0; j < w; j++) { + for (j = 0; j < w; j++, pRow += 4) { uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); - dst[j] = (r << 11) | (g << 5) | (b); - pRow += 4; + *dst++ = (r << 11) | (g << 5) | (b); } - dst += ps->pitch; - p += w0 * 4; + p += src_stride; } } @@ -386,33 +316,23 @@ r5g5b5_put_tile_rgba(struct pipe_surface *ps, * Return each Z value as four floats in [0,1]. */ static void -z16_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +z16_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const ushort *src - = ((const ushort *) map) - + y * ps->pitch + x; const float scale = 1.0f / 65535.0f; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_Z16_UNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = src[j] * scale; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = *src++ * scale; } - src += ps->pitch; - p += 4 * w0; + p += dst_stride; } } @@ -422,33 +342,22 @@ z16_get_tile_rgba(struct pipe_surface *ps, /*** PIPE_FORMAT_U_L8 ***/ static void -l8_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +l8_get_tile_rgba(ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const ubyte *src - = ((const ubyte *) map) - + y * ps->pitch + x; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_L8); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { + for (j = 0; j < w; j++, src++, pRow += 4) { pRow[0] = pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(src[j]); + pRow[2] = UBYTE_TO_FLOAT(*src); pRow[3] = 1.0; - pRow += 4; } - src += ps->pitch; - p += w0 * 4; + p += dst_stride; } } @@ -456,33 +365,22 @@ l8_get_tile_rgba(struct pipe_surface *ps, /*** PIPE_FORMAT_U_A8 ***/ static void -a8_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +a8_get_tile_rgba(ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const ubyte *src - = ((const ubyte *) map) - + y * ps->pitch + x; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_A8); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { + for (j = 0; j < w; j++, src++, pRow += 4) { pRow[0] = pRow[1] = pRow[2] = 0.0; - pRow[3] = UBYTE_TO_FLOAT(src[j]); - pRow += 4; + pRow[3] = UBYTE_TO_FLOAT(*src); } - src += ps->pitch; - p += w0 * 4; + p += dst_stride; } } @@ -490,72 +388,43 @@ a8_get_tile_rgba(struct pipe_surface *ps, /*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ static void -r16g16b16a16_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +r16g16b16a16_get_tile_rgba(short *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const short *src - = ((const short *) map) - + (y * ps->pitch + x) * 4; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - const short *pixel = src; - for (j = 0; j < w; j++) { - pRow[0] = SHORT_TO_FLOAT(pixel[0]); - pRow[1] = SHORT_TO_FLOAT(pixel[1]); - pRow[2] = SHORT_TO_FLOAT(pixel[2]); - pRow[3] = SHORT_TO_FLOAT(pixel[3]); - pRow += 4; - pixel += 4; + for (j = 0; j < w; j++, src += 4, pRow += 4) { + pRow[0] = SHORT_TO_FLOAT(src[0]); + pRow[1] = SHORT_TO_FLOAT(src[1]); + pRow[2] = SHORT_TO_FLOAT(src[2]); + pRow[3] = SHORT_TO_FLOAT(src[3]); } - src += ps->pitch * 4; - p += w0 * 4; + p += dst_stride; } } static void -r16g16b16a16_put_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - const float *p) +r16g16b16a16_put_tile_rgba(short *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) { - short *dst - = ((short *) map) - + (y * ps->pitch + x) * 4; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_R16G16B16A16_SNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { const float *pRow = p; - for (j = 0; j < w; j++) { - short r, g, b, a; - UNCLAMPED_FLOAT_TO_SHORT(r, pRow[0]); - UNCLAMPED_FLOAT_TO_SHORT(g, pRow[1]); - UNCLAMPED_FLOAT_TO_SHORT(b, pRow[2]); - UNCLAMPED_FLOAT_TO_SHORT(a, pRow[3]); - dst[j*4+0] = r; - dst[j*4+1] = g; - dst[j*4+2] = b; - dst[j*4+3] = a; - pRow += 4; + for (j = 0; j < w; j++, dst += 4, pRow += 4) { + UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); + UNCLAMPED_FLOAT_TO_SHORT(dst[1], pRow[1]); + UNCLAMPED_FLOAT_TO_SHORT(dst[2], pRow[2]); + UNCLAMPED_FLOAT_TO_SHORT(dst[3], pRow[3]); } - dst += ps->pitch * 4; - p += w0 * 4; + p += src_stride; } } @@ -564,33 +433,22 @@ r16g16b16a16_put_tile_rgba(struct pipe_surface *ps, /*** PIPE_FORMAT_U_I8 ***/ static void -i8_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +i8_get_tile_rgba(ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const ubyte *src - = ((const ubyte *) map) - + y * ps->pitch + x; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_I8); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { + for (j = 0; j < w; j++, src++, pRow += 4) { pRow[0] = pRow[1] = pRow[2] = - pRow[3] = UBYTE_TO_FLOAT(src[j]); - pRow += 4; + pRow[3] = UBYTE_TO_FLOAT(*src); } - src += ps->pitch; - p += w0 * 4; + p += dst_stride; } } @@ -598,34 +456,23 @@ i8_get_tile_rgba(struct pipe_surface *ps, /*** PIPE_FORMAT_U_A8_L8 ***/ static void -a8_l8_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +a8_l8_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const ushort *src - = ((const ushort *) map) - + y * ps->pitch + x; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_U_A8_L8); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { - const ushort p = src[j]; + for (j = 0; j < w; j++, pRow += 4) { + ushort p = *src++; pRow[0] = pRow[1] = pRow[2] = UBYTE_TO_FLOAT(p & 0xff); pRow[3] = UBYTE_TO_FLOAT(p >> 8); - pRow += 4; } - src += ps->pitch; - p += w0 * 4; + p += dst_stride; } } @@ -638,33 +485,23 @@ a8_l8_get_tile_rgba(struct pipe_surface *ps, * Return each Z value as four floats in [0,1]. */ static void -z32_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +z32_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const uint *src - = ((const uint *) map) - + y * ps->pitch + x; const double scale = 1.0 / (double) 0xffffffff; unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_Z16_UNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (float) (scale * src[j]); + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (*src++ * scale); } - src += ps->pitch; - p += 4 * w0; + p += dst_stride; } } @@ -675,33 +512,23 @@ z32_get_tile_rgba(struct pipe_surface *ps, * Return Z component as four float in [0,1]. Stencil part ignored. */ static void -s8z24_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +s8z24_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const uint *src - = ((const uint *) map) - + y * ps->pitch + x; const double scale = 1.0 / ((1 << 24) - 1); unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_S8Z24_UNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (float) (scale * (src[j] & 0xffffff)); + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (scale * (*src++ & 0xffffff)); } - src += ps->pitch; - p += 4 * w0; + p += dst_stride; } } @@ -712,33 +539,23 @@ s8z24_get_tile_rgba(struct pipe_surface *ps, * Return Z component as four float in [0,1]. Stencil part ignored. */ static void -z24s8_get_tile_rgba(struct pipe_surface *ps, - void *map, - unsigned x, unsigned y, unsigned w, unsigned h, - float *p) +z24s8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { - const uint *src - = ((const uint *) map) - + y * ps->pitch + x; const double scale = 1.0 / ((1 << 24) - 1); unsigned i, j; - unsigned w0 = w; - - assert(ps->format == PIPE_FORMAT_Z24S8_UNORM); - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; for (i = 0; i < h; i++) { float *pRow = p; - for (j = 0; j < w; j++) { - pRow[j * 4 + 0] = - pRow[j * 4 + 1] = - pRow[j * 4 + 2] = - pRow[j * 4 + 3] = (float) (scale * (src[j] >> 8)); + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (scale * (*src++ >> 8)); } - src += ps->pitch; - p += 4 * w0; + p += dst_stride; } } @@ -749,56 +566,67 @@ pipe_get_tile_rgba(struct pipe_context *pipe, uint x, uint y, uint w, uint h, float *p) { - void *map = pipe_surface_map(ps); + unsigned dst_stride = w * 4; + void *packed; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + packed = MALLOC(h * w * ps->cpp); + + if (!packed) + return; + + pipe->get_tile(pipe, ps, x, y, w, h, packed, w * ps->cpp); switch (ps->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_get_tile_rgba(ps, map, x, y, w, h, p); + a8r8g8b8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_get_tile_rgba(ps, map, x, y, w, h, p); + b8g8r8a8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_A1R5G5B5_UNORM: - a1r5g5b5_get_tile_rgba(ps, map, x, y, w, h, p); + a1r5g5b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_A4R4G4B4_UNORM: - a4r4g4b4_get_tile_rgba(ps, map, x, y, w, h, p); + a4r4g4b4_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_R5G6B5_UNORM: - r5g6b5_get_tile_rgba(ps, map, x, y, w, h, p); + r5g6b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_U_L8: - l8_get_tile_rgba(ps, map, x, y, w, h, p); + l8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_U_A8: - a8_get_tile_rgba(ps, map, x, y, w, h, p); + a8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_U_I8: - i8_get_tile_rgba(ps, map, x, y, w, h, p); + i8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_U_A8_L8: - a8_l8_get_tile_rgba(ps, map, x, y, w, h, p); + a8_l8_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_get_tile_rgba(ps, map, x, y, w, h, p); + r16g16b16a16_get_tile_rgba((short *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_Z16_UNORM: - z16_get_tile_rgba(ps, map, x, y, w, h, p); + z16_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_Z32_UNORM: - z32_get_tile_rgba(ps, map, x, y, w, h, p); + z32_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_S8Z24_UNORM: - s8z24_get_tile_rgba(ps, map, x, y, w, h, p); + s8z24_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_Z24S8_UNORM: - z24s8_get_tile_rgba(ps, map, x, y, w, h, p); + z24s8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; default: assert(0); } - pipe_surface_unmap(ps); + FREE(packed); } @@ -808,53 +636,64 @@ pipe_put_tile_rgba(struct pipe_context *pipe, uint x, uint y, uint w, uint h, const float *p) { - void *map = pipe_surface_map(ps); + unsigned src_stride = w * 4; + void *packed; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + packed = MALLOC(h * w * ps->cpp); + + if (!packed) + return; switch (ps->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_put_tile_rgba(ps, map, x, y, w, h, p); + a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_put_tile_rgba(ps, map, x, y, w, h, p); + b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_A1R5G5B5_UNORM: - /*a1r5g5b5_put_tile_rgba(ps, map, x, y, w, h, p);*/ + /*a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_R5G6B5_UNORM: - r5g5b5_put_tile_rgba(ps, map, x, y, w, h, p); + r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_R8G8B8A8_UNORM: break; case PIPE_FORMAT_U_L8: - /*l8_put_tile_rgba(ps, map, x, y, w, h, p);*/ + /*l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_U_A8: - /*a8_put_tile_rgba(ps, map, x, y, w, h, p);*/ + /*a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_U_I8: - /*i8_put_tile_rgba(ps, map, x, y, w, h, p);*/ + /*i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_U_A8_L8: - /*a8_l8_put_tile_rgba(ps, map, x, y, w, h, p);*/ + /*a8_l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_put_tile_rgba(ps, map, x, y, w, h, p); + r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_Z16_UNORM: - /*z16_put_tile_rgba(ps, map, x, y, w, h, p);*/ + /*z16_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_Z32_UNORM: - /*z32_put_tile_rgba(ps, map, x, y, w, h, p);*/ + /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_S8Z24_UNORM: - /*s8z24_put_tile_rgba(ps, map, x, y, w, h, p);*/ + /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_Z24S8_UNORM: - /*z24s8_put_tile_rgba(ps, map, x, y, w, h, p);*/ + /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ break; default: assert(0); } - pipe_surface_unmap(ps); + pipe->put_tile(pipe, ps, x, y, w, h, packed, w * ps->cpp); + + FREE(packed); } diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 73cd7db18d..3a3bf9016d 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -43,6 +43,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "pipe/util/p_tile.h" #define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ @@ -80,7 +81,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) accBuf[i*4+3] = a; } - pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); } @@ -95,13 +96,13 @@ accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] * scale + bias; } - pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(accBuf); } @@ -119,14 +120,14 @@ accum_accum(struct pipe_context *pipe, GLfloat value, colorBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, colorBuf); - pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, colorBuf); + pipe_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] + colorBuf[i] * value; } - pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(colorBuf); free(accBuf); @@ -144,13 +145,13 @@ accum_load(struct pipe_context *pipe, GLfloat value, buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, buf); + pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, buf); for (i = 0; i < 4 * width * height; i++) { buf[i] = buf[i] * value; } - pipe->put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, buf); + pipe_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, buf); free(buf); } @@ -169,11 +170,11 @@ accum_return(GLcontext *ctx, GLfloat value, abuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe->get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); + pipe_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) { cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe->get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, cbuf); + pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, cbuf); } for (i = 0; i < width * height; i++) { @@ -188,7 +189,7 @@ accum_return(GLcontext *ctx, GLfloat value, } } - pipe->put_tile_rgba(pipe, color_ps, xpos, ypos, width, height, abuf); + pipe_put_tile_rgba(pipe, color_ps, xpos, ypos, width, height, abuf); free(abuf); if (cbuf) diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 5ea2f08b8a..be5434f91b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -56,6 +56,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" +#include "pipe/util/p_tile.h" #include "shader/prog_instruction.h" @@ -1261,8 +1262,8 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, /* alternate path using get/put_tile() */ GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe->get_tile_rgba(pipe, psRead, srcx, srcy, width, height, buf); - pipe->put_tile_rgba(pipe, psTex, 0, 0, width, height, buf); + pipe_get_tile_rgba(pipe, psRead, srcx, srcy, width, height, buf); + pipe_put_tile_rgba(pipe, psTex, 0, 0, width, height, buf); free(buf); } diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 585fe04dee..a1bbb3a831 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -40,6 +40,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "pipe/util/p_tile.h" #include "st_context.h" #include "st_cb_readpixels.h" #include "st_cb_fbo.h" @@ -210,7 +211,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, /* Do a row at a time to flip image data vertically */ for (i = 0; i < height; i++) { - pipe->get_tile_rgba(pipe, strb->surface, x, y, width, 1, df); + pipe_get_tile_rgba(pipe, strb->surface, x, y, width, 1, df); y += yStep; df += dfStride; if (!dfStride) { diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 19274570a1..ba0950e295 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -45,6 +45,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "pipe/util/p_tile.h" #define DBG if (0) printf @@ -1068,8 +1069,7 @@ fallback_copy_texsubimage(GLcontext *ctx, /* do copy row by row */ for (row = 0; row < height; row++) { - pipe->get_tile_rgba(pipe, src_surf, srcX, srcY + row, width, 1, - data); + pipe_get_tile_rgba(pipe, src_surf, srcX, srcY + row, width, 1, data); /* XXX we're ignoring convolution for now */ if (ctx->_ImageTransferState) { @@ -1078,8 +1078,7 @@ fallback_copy_texsubimage(GLcontext *ctx, width, (GLfloat (*)[4])data); } - pipe->put_tile_rgba(pipe, dest_surf, destX, destY, width, 1, - data); + pipe_put_tile_rgba(pipe, dest_surf, destX, destY, width, 1, data); destY += yStep; } -- cgit v1.2.3 From eff70f67c558c90c5419b43729a6dc8a9a8c47ec Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Jan 2008 11:11:08 -0700 Subject: Cell: update surface map code to match recent changes in pipe_surface struct --- src/mesa/pipe/cell/ppu/cell_context.h | 2 ++ src/mesa/pipe/cell/ppu/cell_spu.c | 6 +++--- src/mesa/pipe/cell/ppu/cell_state_surface.c | 21 +++++++++++---------- src/mesa/pipe/cell/ppu/cell_surface.c | 4 ++-- 4 files changed, 18 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 94d8f08da8..43e32a8abb 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -78,6 +78,8 @@ struct cell_context struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS]; + ubyte *zbuf_map; uint dirty; diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 88a620e669..77ddd9ccbf 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -168,8 +168,8 @@ test_spus(struct cell_context *cell) sleep(2); for (i = 0; i < cell->num_spus; i++) { - cell_global.command[i].fb.color_start = csurf->map; - cell_global.command[i].fb.depth_start = zsurf ? zsurf->map : NULL; + cell_global.command[i].fb.color_start = cell->cbuf_map[0]; + cell_global.command[i].fb.depth_start = cell->zbuf_map; cell_global.command[i].fb.width = csurf->width; cell_global.command[i].fb.height = csurf->height; cell_global.command[i].fb.color_format = PIPE_FORMAT_A8R8G8B8_UNORM; @@ -186,7 +186,7 @@ test_spus(struct cell_context *cell) finish_all(cell->num_spus); { - uint *b = (uint*) csurf->map; + uint *b = (uint*) cell->cbuf_map[0]; printf("PPU: Clear results: 0x%x 0x%x 0x%x 0x%x\n", b[0], b[1000], b[2000], b[3000]); } diff --git a/src/mesa/pipe/cell/ppu/cell_state_surface.c b/src/mesa/pipe/cell/ppu/cell_state_surface.c index ff8fdbd334..9bf3c339d2 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_state_surface.c @@ -39,6 +39,7 @@ cell_set_framebuffer_state(struct pipe_context *pipe, { struct cell_context *cell = cell_context(pipe); + /* XXX revisit this memcmp! */ if (memcmp(&cell->framebuffer, fb, sizeof(*fb))) { struct pipe_surface *csurf = fb->cbufs[0]; struct pipe_surface *zsurf = fb->zbuf; @@ -48,26 +49,26 @@ cell_set_framebuffer_state(struct pipe_context *pipe, /* unmap old surfaces */ for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { - if (cell->framebuffer.cbufs[i] && - cell->framebuffer.cbufs[i]->map) { + if (cell->framebuffer.cbufs[i] && cell->cbuf_map[i]) { pipe_surface_unmap(cell->framebuffer.cbufs[i]); + cell->cbuf_map[i] = NULL; } } - if (cell->framebuffer.zbuf && - cell->framebuffer.zbuf->map) { + if (cell->framebuffer.zbuf && cell->zbuf_map) { pipe_surface_unmap(cell->framebuffer.zbuf); + cell->zbuf_map = NULL; } /* update my state */ cell->framebuffer = *fb; /* map new surfaces */ - if (csurf && !csurf->map) - pipe_surface_map(csurf); + if (csurf) + cell->cbuf_map[0] = pipe_surface_map(csurf); - if (zsurf && !zsurf->map) - pipe_surface_map(zsurf); + if (zsurf) + cell->zbuf_map = pipe_surface_map(zsurf); #if 0 for (i = 0; i < cell->num_spus; i++) { @@ -87,9 +88,9 @@ cell_set_framebuffer_state(struct pipe_context *pipe, struct cell_command_framebuffer *fb = cell_batch_alloc(cell, sizeof(*fb)); fb->opcode = CELL_CMD_FRAMEBUFFER; - fb->color_start = csurf->map; + fb->color_start = cell->cbuf_map[0]; fb->color_format = csurf->format; - fb->depth_start = zsurf ? zsurf->map : NULL; + fb->depth_start = cell->zbuf_map; fb->depth_format = zsurf ? zsurf->format : PIPE_FORMAT_NONE; fb->width = csurf->width; fb->height = csurf->height; diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index ea1c2bd644..53c5325ddd 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -50,8 +50,8 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, uint i; uint surfIndex; - if (!ps->map) - pipe_surface_map(ps); + if (!cell->cbuf_map[0]) + cell->cbuf_map[0] = pipe_surface_map(ps); if (ps == cell->framebuffer.zbuf) { surfIndex = 1; -- cgit v1.2.3 From c28b112ce37022aa6e00ac4557ad6fe5a57ae578 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Jan 2008 16:37:13 -0700 Subject: clamp colors during float->int conversion --- src/mesa/pipe/cell/spu/spu_tri.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index b42b321c32..ddd5e662d2 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -236,6 +236,10 @@ pack_color(const float color[4]) uint g = (uint) (color[1] * 255.0); uint b = (uint) (color[2] * 255.0); uint a = (uint) (color[3] * 255.0); + r = MIN2(r, 255); + g = MIN2(g, 255); + b = MIN2(b, 255); + a = MIN2(a, 255); switch (spu.fb.color_format) { case PIPE_FORMAT_A8R8G8B8_UNORM: return (a << 24) | (r << 16) | (g << 8) | b; -- cgit v1.2.3 From f20cb1d81b120c8de9b00e2fb523e2367b25e875 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Jan 2008 16:39:26 -0700 Subject: Cell: after sending a batch, wait for a DMA completion signal. This fixes sporadic rendering glitches. Using a mailbox msg for now, until spe_mfcio_tag_status_read() or similar is found to work. --- src/mesa/pipe/cell/common.h | 2 ++ src/mesa/pipe/cell/ppu/cell_batch.c | 17 ++++++++++++++--- src/mesa/pipe/cell/spu/spu_main.c | 23 +++++++++++++++++++++-- src/mesa/pipe/cell/spu/spu_tile.c | 4 +++- 4 files changed, 40 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 74b3d3cbee..29d1cbedda 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -65,6 +65,8 @@ #define CELL_NUM_BATCH_BUFFERS 2 #define CELL_BATCH_BUFFER_SIZE 1024 /**< 16KB would be the max */ +#define CELL_BATCH_FINISHED 0x1234 /**< mbox message */ + /** * Tell SPUs about the framebuffer size, location diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index 45d49d0465..45f62ac3ad 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -43,16 +43,27 @@ cell_batch_flush(struct cell_context *cell) assert(batch < CELL_NUM_BATCH_BUFFERS); - /*printf("cell_batch_dispatch: buf %u, size %u\n", batch, size);*/ - + /* + printf("cell_batch_dispatch: buf %u at %p, size %u\n", + batch, &cell->batch_buffer[batch][0], size); + */ + cmd_word = CELL_CMD_BATCH | (batch << 8) | (size << 16); for (i = 0; i < cell->num_spus; i++) { send_mbox_message(cell_global.spe_contexts[i], cmd_word); } - /* XXX wait on DMA xfer of prev buffer to complete */ + /* XXX wait for the DMX xfer to finish. + * Using mailboxes here is temporary. + * Ideally, we want to use a PPE-side DMA status check function... + */ + for (i = 0; i < cell->num_spus; i++) { + uint k = wait_mbox_message(cell_global.spe_contexts[i]); + assert(k == CELL_BATCH_FINISHED); + } + /* next buffer */ cell->cur_batch = (batch + 1) % CELL_NUM_BATCH_BUFFERS; cell->batch_buffer_size[cell->cur_batch] = 0; /* empty */ diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 0aee2aa6f9..2727b03756 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -46,7 +46,7 @@ helpful headers: /opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h */ -static boolean Debug = TRUE; +static boolean Debug = FALSE; struct spu_global spu; @@ -106,7 +106,7 @@ really_clear_tiles(uint surfaceIndex) } } -#if 01 +#if 0 wait_on_mask(1 << TAG_SURFACE_CLEAR); #endif } @@ -165,6 +165,9 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) #if 0 wait_on_mask(1 << TAG_SURFACE_CLEAR); #endif + + if (Debug) + printf("SPU %u: CLEAR SURF done\n", spu.init.id); } @@ -222,8 +225,10 @@ cmd_render(const struct cell_command_render *render) render->prim_type, render->num_verts, render->num_indexes); + /* printf(" bound: %g, %g .. %g, %g\n", render->xmin, render->ymin, render->xmax, render->ymax); + */ } ASSERT_ALIGN16(render->vertex_data); @@ -244,6 +249,7 @@ cmd_render(const struct cell_command_render *render) render->index_data, render->vertex_data, vertex_bytes, index_bytes); */ + ASSERT(vertex_bytes % 16 == 0); /* get vertex data from main memory */ mfc_get(vertex_data, /* dest */ (unsigned int) render->vertex_data, /* src */ @@ -252,6 +258,8 @@ cmd_render(const struct cell_command_render *render) 0, /* tid */ 0 /* rid */); + ASSERT(index_bytes % 16 == 0); + /* get index data from main memory */ mfc_get(indexes, /* dest */ (unsigned int) render->index_data, /* src */ @@ -330,6 +338,10 @@ cmd_render(const struct cell_command_render *render) wait_on_mask(1 << TAG_WRITE_TILE_Z); } } + + if (Debug) + printf("SPU %u: RENDER done\n", + spu.init.id); } @@ -406,6 +418,9 @@ cmd_batch(uint opcode) size = (size + 0xf) & ~0xf; + ASSERT(size % 16 == 0); + ASSERT((unsigned int) spu.init.batch_buffers[buf] % 16 == 0); + mfc_get(buffer, /* dest */ (unsigned int) spu.init.batch_buffers[buf], /* src */ size, @@ -414,6 +429,10 @@ cmd_batch(uint opcode) 0 /* rid */); wait_on_mask(1 << TAG_BATCH_BUFFER); + /* send mbox message to indicate DMA completed */ + /* XXX temporary */ + spu_write_out_mbox(CELL_BATCH_FINISHED); + for (pos = 0; pos < usize; /* no incr */) { switch (buffer[pos]) { case CELL_CMD_FRAMEBUFFER: diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c index 99d779007e..1505cb322b 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.c +++ b/src/mesa/pipe/cell/spu/spu_tile.c @@ -77,9 +77,11 @@ put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf) ASSERT(ty < spu.fb.height_tiles); ASSERT_ALIGN16(tile); /* - printf("put_tile: src: %p dst: 0x%x size: %d\n", + printf("SPU %u: put_tile: src: %p dst: 0x%x size: %d\n", + spu.init.id, tile, (unsigned int) dst, bytesPerTile); */ + mfc_put((void *) tile, /* src in local memory */ (unsigned int) dst, /* dst in main memory */ bytesPerTile, -- cgit v1.2.3 From b4e4fafb4157d416077e985c03204ed5bbe0f2e1 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Jan 2008 19:04:45 -0700 Subject: disable debug printf --- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index e9bd4dd5e4..055cb19f9a 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -227,7 +227,9 @@ static void vbuf_flush_elements( struct draw_stage *stage ) struct vbuf_stage *vbuf = vbuf_stage( stage ); if (vbuf->nr_elements) { + /* fprintf(stderr, "%s (%d elts)\n", __FUNCTION__, vbuf->nr_elements); + */ /* Draw now or add to list of primitives??? */ -- cgit v1.2.3 From ac95fee4fffee77bb7bd798d094ed2e3a7c4019b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Jan 2008 19:12:46 -0700 Subject: Fix problems with vertex shaders and the private draw module. The CSO returned by pipe->create_vs_state() can't be passed to the private draw module. That was causing glRasterPos to blow up. Add a 'draw_shader' field to st_vertex_program for use with the private draw module. Change st_context->state.vs type from cso_vertex_shader to st_vertex_program. --- src/mesa/state_tracker/st_atom_shader.c | 23 +++++++++++++---------- src/mesa/state_tracker/st_cb_clear.c | 6 +++--- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- src/mesa/state_tracker/st_cb_program.c | 11 +++++++---- src/mesa/state_tracker/st_context.h | 2 +- src/mesa/state_tracker/st_debug.c | 4 ++-- src/mesa/state_tracker/st_draw.c | 12 ++++++++---- src/mesa/state_tracker/st_program.c | 6 ++---- src/mesa/state_tracker/st_program.h | 7 +++++-- 9 files changed, 43 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 2a182c7d9c..1ed9333556 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -43,6 +43,8 @@ #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" +#include "pipe/cso_cache/cso_cache.h" + #include "st_context.h" #include "st_cache.h" #include "st_atom.h" @@ -71,8 +73,8 @@ struct translated_vertex_program /** The program in TGSI format */ struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; - /** Pointer to the translated, cached vertex shader */ - const struct cso_vertex_shader *vs; + /** Pointer to the translated vertex program */ + struct st_vertex_program *vp; struct translated_vertex_program *next; /**< next in linked list */ }; @@ -257,12 +259,13 @@ find_translated_vp(struct st_context *st, assert(stvp->Base.Base.NumInstructions > 1); - xvp->vs = st_translate_vertex_program(st, stvp, - xvp->output_to_slot, - xvp->tokens, - ST_MAX_SHADER_TOKENS); - assert(xvp->vs); - stvp->vs = NULL; /* don't want to use this */ + st_translate_vertex_program(st, stvp, + xvp->output_to_slot, + xvp->tokens, + ST_MAX_SHADER_TOKENS); + + assert(stvp->cso); + xvp->vp = stvp; /* translated VP is up to date now */ xvp->serialNo = stvp->serialNo; @@ -291,8 +294,8 @@ update_linkage( struct st_context *st ) xvp = find_translated_vp(st, stvp, stfp); st->vp = stvp; - st->state.vs = xvp->vs; - st->pipe->bind_vs_state(st->pipe, st->state.vs->data); + st->state.vs = xvp->vp; + st->pipe->bind_vs_state(st->pipe, st->state.vs->cso->data); st->fp = stfp; st->state.fs = stfp->fs; diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 40319f4b4b..758d4a4086 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -208,7 +208,7 @@ make_vertex_shader(struct st_context *st) stvp = (struct st_vertex_program *) p; st_translate_vertex_program(st, stvp, NULL, stvp->tokens, ST_MAX_SHADER_TOKENS); - assert(stvp->vs); + assert(stvp->cso); return stvp; } @@ -350,7 +350,7 @@ clear_with_quad(GLcontext *ctx, if (!stvp) { stvp = make_vertex_shader(st); } - pipe->bind_vs_state(pipe, stvp->vs->data); + pipe->bind_vs_state(pipe, stvp->cso->data); } /* viewport state: viewport matching window dims */ @@ -376,7 +376,7 @@ clear_with_quad(GLcontext *ctx, pipe->bind_blend_state(pipe, st->state.blend->data); pipe->bind_depth_stencil_alpha_state(pipe, st->state.depth_stencil->data); pipe->bind_fs_state(pipe, st->state.fs->data); - pipe->bind_vs_state(pipe, st->state.vs->data); + pipe->bind_vs_state(pipe, st->state.vs->cso->data); pipe->bind_rasterizer_state(pipe, st->state.rasterizer->data); pipe->set_viewport_state(pipe, &st->state.viewport); /* OR: diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index be5434f91b..f4d6b9362c 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -662,7 +662,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->bind_fs_state(pipe, stfp->fs->data); /* vertex shader state: position + texcoord pass-through */ - pipe->bind_vs_state(pipe, stvp->vs->data); + pipe->bind_vs_state(pipe, stvp->cso->data); /* texture sampling state: */ { @@ -719,7 +719,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* restore GL state */ pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); pipe->bind_fs_state(pipe, ctx->st->state.fs->data); - pipe->bind_vs_state(pipe, ctx->st->state.vs->data); + pipe->bind_vs_state(pipe, ctx->st->state.vs->cso->data); pipe->set_sampler_texture(pipe, unit, ctx->st->state.sampler_texture[unit]); pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]->data); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index aed6b1ee97..63a954264e 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -39,6 +39,8 @@ #include "shader/programopt.h" #include "shader/shader_api.h" +#include "pipe/cso_cache/cso_cache.h" + #include "st_context.h" #include "st_program.h" #include "st_atom_shader.h" @@ -181,10 +183,11 @@ static void st_program_string_notify( GLcontext *ctx, stvp->serialNo++; - if (stvp->vs) { - /* free the TGSI code */ - // cso_delete(stfp->vs); - stvp->vs = NULL; + if (stvp->cso) { + /* free the CSO data */ + st->pipe->delete_vs_state(st->pipe, stvp->cso->data); + FREE((void *) stvp->cso); + stvp->cso = NULL; } stvp->param_state = stvp->Base.Base.Parameters->StateFlags; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 0f40f3ceee..5ae21c93f9 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -100,7 +100,7 @@ struct st_context const struct cso_depth_stencil_alpha *depth_stencil; const struct cso_rasterizer *rasterizer; const struct cso_fragment_shader *fs; - const struct cso_vertex_shader *vs; + struct st_vertex_program *vs; struct pipe_blend_color blend_color; struct pipe_clip_state clip; diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index cffd66751d..57450e52bf 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -53,11 +53,11 @@ st_print_current(void) int i; printf("Vertex Transform Inputs:\n"); - for (i = 0; i < st->state.vs->state.num_inputs; i++) { + for (i = 0; i < st->state.vs->cso->state.num_inputs; i++) { printf(" Slot %d: VERT_ATTRIB_%d\n", i, st->vp->index_to_input[i]); } - tgsi_dump( st->state.vs->state.tokens, 0 ); + tgsi_dump( st->state.vs->cso->state.tokens, 0 ); if (st->vp->Base.Base.Parameters) _mesa_print_parameter_list(st->vp->Base.Base.Parameters); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 96db9e3c03..94b3a9531a 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -241,7 +241,7 @@ st_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = &ctx->st->state.vs->state; + vs = &ctx->st->state.vs->cso->state; /* loop over TGSI shader inputs to determine vertex buffer * and attribute info @@ -447,7 +447,7 @@ set_feedback_vertex_format(GLcontext *ctx) else { /* GL_FEEDBACK, or glRasterPos */ /* emit all attribs (pos, color, texcoord) as GLfloat[4] */ - vinfo.num_attribs = st->state.vs->state.num_outputs; + vinfo.num_attribs = st->state.vs->cso->state.num_outputs; for (i = 0; i < vinfo.num_attribs; i++) { vinfo.format[i] = FORMAT_4F; vinfo.interp_mode[i] = INTERP_LINEAR; @@ -491,7 +491,11 @@ st_feedback_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = &ctx->st->state.vs->state; + vs = &ctx->st->state.vs->cso->state; + + if (!st->state.vs->draw_shader) { + st->state.vs->draw_shader = draw_create_vertex_shader(draw, vs); + } /* * Set up the draw module's state. @@ -503,7 +507,7 @@ st_feedback_draw_vbo(GLcontext *ctx, draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); draw_set_rasterizer_state(draw, &st->state.rasterizer->state); - draw_bind_vertex_shader(draw, st->state.vs->data); + draw_bind_vertex_shader(draw, st->state.vs->draw_shader); set_feedback_vertex_format(ctx); /* loop over TGSI shader inputs to determine vertex buffer diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 1852228b29..1f1e6500e0 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -57,7 +57,7 @@ * \param tokensOut destination for TGSI tokens * \return pointer to cached pipe_shader object. */ -const struct cso_vertex_shader * +void st_translate_vertex_program(struct st_context *st, struct st_vertex_program *stvp, const GLuint outputMapping[], @@ -256,12 +256,10 @@ st_translate_vertex_program(struct st_context *st, vs.tokens = tokensOut; cso = st_cached_vs_state(st, &vs); - stvp->vs = cso; + stvp->cso = cso; if (TGSI_DEBUG) tgsi_dump( tokensOut, 0 ); - - return cso; } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 049f9f659f..de02c3185f 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -90,7 +90,10 @@ struct st_vertex_program struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; /** Pointer to the corresponding cached shader */ - const struct cso_vertex_shader *vs; + const struct cso_vertex_shader *cso; + + /** For using our private draw module (glRasterPos) */ + struct draw_vertex_shader *draw_shader; GLuint param_state; }; @@ -122,7 +125,7 @@ st_translate_fragment_program(struct st_context *st, GLuint maxTokens); -extern const struct cso_vertex_shader * +extern void st_translate_vertex_program(struct st_context *st, struct st_vertex_program *vp, const GLuint vert_output_to_slot[], -- cgit v1.2.3 From d280206c7f74c6c0fc22798b5945db3bf369364e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Jan 2008 19:18:35 -0700 Subject: free stvp->draw_shader --- src/mesa/state_tracker/st_cb_program.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 63a954264e..1f34fc86ad 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -190,11 +190,15 @@ static void st_program_string_notify( GLcontext *ctx, stvp->cso = NULL; } + if (stvp->draw_shader) { + draw_delete_vertex_shader(st->draw, stvp->draw_shader); + stvp->draw_shader = NULL; + } + stvp->param_state = stvp->Base.Base.Parameters->StateFlags; if (st->vp == stvp) st->dirty.st |= ST_NEW_VERTEX_PROGRAM; - } } -- cgit v1.2.3 From dd5a8d234b95c6f85d0a6ecb18db8aadefb71dbd Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 15 Jan 2008 14:18:17 -0700 Subject: Add surface status/clear_value fields, PIPE_SURFACE_STATUS_x tokens --- src/mesa/pipe/p_defines.h | 8 ++++++++ src/mesa/pipe/p_state.h | 2 ++ 2 files changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index c6d9c02bd9..50bea691e7 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -169,6 +169,14 @@ enum pipe_texture_target { #define PIPE_SURFACE 2 /**< user-created surfaces */ +/** + * Surface status + */ +#define PIPE_SURFACE_STATUS_UNDEFINED 0 +#define PIPE_SURFACE_STATUS_DEFINED 1 +#define PIPE_SURFACE_STATUS_CLEAR 2 + + /** * Buffer access flags */ diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index ccd2a5f9e2..46328d2a8f 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -243,6 +243,8 @@ struct pipe_surface { struct pipe_buffer_handle *buffer; /**< driver private buffer handle */ enum pipe_format format; /**< PIPE_FORMAT_x */ + unsigned status; /**< PIPE_SURFACE_STATUS_x */ + unsigned clear_value; /**< may be temporary */ unsigned cpp; /**< bytes per pixel */ unsigned width, height; unsigned pitch; /**< in pixels */ -- cgit v1.2.3 From 14a1e5908dc204f033eaff88e6b3f5cbc9793c0c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 15 Jan 2008 18:22:23 -0700 Subject: include pipe/draw/draw_context.h --- src/mesa/state_tracker/st_cb_program.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 1f34fc86ad..f1f33fb0dd 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -40,6 +40,7 @@ #include "shader/shader_api.h" #include "pipe/cso_cache/cso_cache.h" +#include "pipe/draw/draw_context.h" #include "st_context.h" #include "st_program.h" -- cgit v1.2.3 From 0cec2e18d5a11244f1b2c989b3f17161a95ee416 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 15 Jan 2008 18:22:35 -0700 Subject: remove redundant llvm subdir --- src/mesa/pipe/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile index da044036ff..d880d090c1 100644 --- a/src/mesa/pipe/Makefile +++ b/src/mesa/pipe/Makefile @@ -10,7 +10,7 @@ ifeq ($(CONFIG_NAME), linux-llvm) LLVM_DIR = llvm endif -SUBDIRS = softpipe llvm i915simple i965simple failover pipebuffer $(CELL_DIR) $(LLVM_DIR) +SUBDIRS = softpipe i915simple i965simple failover pipebuffer $(CELL_DIR) $(LLVM_DIR) default: subdirs -- cgit v1.2.3 From 5a185ca09ae9d0774c99a187463d975a40ce9770 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 15 Jan 2008 18:26:21 -0700 Subject: replace _mesa_printf() with fprintf() --- src/mesa/pipe/i965simple/brw_sf.c | 8 ++++---- src/mesa/pipe/i965simple/brw_sf_emit.c | 14 +++++++------- 2 files changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_sf.c b/src/mesa/pipe/i965simple/brw_sf.c index 362196a0d1..b89b2e4087 100644 --- a/src/mesa/pipe/i965simple/brw_sf.c +++ b/src/mesa/pipe/i965simple/brw_sf.c @@ -175,7 +175,7 @@ static void upload_sf_prog( struct brw_context *brw ) //int semantic = parse.FullToken.FullDeclaration.Semantic.SemanticName; //int semantic_index = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - _mesa_printf("fs input %d..%d interp mode %d\n", first, last, interp_mode); + fprintf(stderr, "fs input %d..%d interp mode %d\n", first, last, interp_mode); switch (interp_mode) { case TGSI_INTERPOLATE_CONSTANT: @@ -213,9 +213,9 @@ static void upload_sf_prog( struct brw_context *brw ) key.linear_mask |= 1; key.const_mask <<= 1; - _mesa_printf("key.persp_mask: %x\n", key.persp_mask); - _mesa_printf("key.linear_mask: %x\n", key.linear_mask); - _mesa_printf("key.const_mask: %x\n", key.const_mask); + fprintf(stderr, "key.persp_mask: %x\n", key.persp_mask); + fprintf(stderr, "key.linear_mask: %x\n", key.linear_mask); + fprintf(stderr, "key.const_mask: %x\n", key.const_mask); // key.do_point_sprite = brw->attribs.Point->PointSprite; diff --git a/src/mesa/pipe/i965simple/brw_sf_emit.c b/src/mesa/pipe/i965simple/brw_sf_emit.c index bc3878a93a..6ff5254ff7 100644 --- a/src/mesa/pipe/i965simple/brw_sf_emit.c +++ b/src/mesa/pipe/i965simple/brw_sf_emit.c @@ -137,8 +137,8 @@ static boolean calculate_masks( struct brw_sf_compile *c, unsigned persp_mask = c->key.persp_mask; unsigned linear_mask = c->key.linear_mask; - _mesa_printf("persp_mask: %x\n", persp_mask); - _mesa_printf("linear_mask: %x\n", linear_mask); + fprintf(stderr, "persp_mask: %x\n", persp_mask); + fprintf(stderr, "linear_mask: %x\n", linear_mask); *pc_persp = 0; *pc_linear = 0; @@ -162,9 +162,9 @@ static boolean calculate_masks( struct brw_sf_compile *c, *pc_linear |= 0xf0; } - _mesa_printf("pc: %x\n", *pc); - _mesa_printf("pc_persp: %x\n", *pc_persp); - _mesa_printf("pc_linear: %x\n", *pc_linear); + fprintf(stderr, "pc: %x\n", *pc); + fprintf(stderr, "pc_persp: %x\n", *pc_persp); + fprintf(stderr, "pc_linear: %x\n", *pc_linear); return is_last_attr; @@ -177,7 +177,7 @@ void brw_emit_tri_setup( struct brw_sf_compile *c ) struct brw_compile *p = &c->func; unsigned i; - _mesa_printf("%s START ==============\n", __FUNCTION__); + fprintf(stderr, "%s START ==============\n", __FUNCTION__); c->nr_verts = 3; alloc_regs(c); @@ -250,7 +250,7 @@ void brw_emit_tri_setup( struct brw_sf_compile *c ) } } - _mesa_printf("%s DONE ==============\n", __FUNCTION__); + fprintf(stderr, "%s DONE ==============\n", __FUNCTION__); } -- cgit v1.2.3 From 62ef6376bef7dcc7f23c676519683e99cd9f7717 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 15 Jan 2008 18:27:14 -0700 Subject: Don't include stdint.h or inttypes. Use the uint64 typedef instead of uint64_t. --- src/mesa/pipe/p_context.h | 4 ++-- src/mesa/pipe/p_util.h | 1 - src/mesa/pipe/softpipe/sp_context.h | 2 +- src/mesa/pipe/softpipe/sp_query.c | 6 +++--- 4 files changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 25b5dc35e4..7a18d48865 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -29,7 +29,7 @@ #define PIPE_CONTEXT_H #include "p_state.h" -#include + struct pipe_state_cache; @@ -97,7 +97,7 @@ struct pipe_context { boolean (*get_query_result)(struct pipe_context *pipe, struct pipe_query *q, boolean wait, - uint64_t *result); + uint64 *result); /* * State functions diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 46edcf3075..1bf606bf4b 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -30,7 +30,6 @@ #include "p_compiler.h" #include -#include #ifdef WIN32 diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 394baf0109..3537f86a61 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -76,7 +76,7 @@ struct softpipe_context { /* Counter for occlusion queries. Note this supports overlapping * queries. */ - uint64_t occlusion_count; + uint64 occlusion_count; /* * Mapped vertex buffers diff --git a/src/mesa/pipe/softpipe/sp_query.c b/src/mesa/pipe/softpipe/sp_query.c index bf753dad98..6a8a43aeda 100644 --- a/src/mesa/pipe/softpipe/sp_query.c +++ b/src/mesa/pipe/softpipe/sp_query.c @@ -37,8 +37,8 @@ #include "sp_query.h" struct softpipe_query { - uint64_t start; - uint64_t end; + uint64 start; + uint64 end; }; @@ -87,7 +87,7 @@ static boolean softpipe_get_query_result(struct pipe_context *pipe, struct pipe_query *q, boolean wait, - uint64_t *result ) + uint64 *result ) { struct softpipe_query *sq = softpipe_query(q); *result = sq->end - sq->start; -- cgit v1.2.3 From 587e2becc237bc1c900a1c0ba114a1a0192690ff Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 15 Jan 2008 18:32:51 -0700 Subject: typedef uintptr_t for non-HAVE_POSIX_MEMALIGN build --- src/mesa/pipe/p_util.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 1bf606bf4b..69ec62e307 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -123,6 +123,7 @@ align_malloc(size_t bytes, uint alignment) (void) posix_memalign(& mem, alignment, bytes); return mem; #else + typedef unsigned long int uintptr_t; uintptr_t ptr, buf; assert( alignment > 0 ); -- cgit v1.2.3 From dd235ff1db6a2839607931cc07f9ba6e602ac3d4 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 08:30:00 -0700 Subject: Fix a two-sided lighting bug (fixes samples/wave.c) --- src/mesa/state_tracker/st_atom_rasterizer.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 35fd506458..beae36bca0 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -110,7 +110,14 @@ static void update_raster_state( struct st_context *st ) * GL_VERTEX_PROGRAM_TWO_SIDE is set). Note the logic here. */ if (ctx->VertexProgram._Current) { - raster.light_twoside = ctx->VertexProgram.TwoSideEnabled; + if (ctx->VertexProgram._Enabled) { + /* user-defined program */ + raster.light_twoside = ctx->VertexProgram.TwoSideEnabled; + } + else { + /* TNL-generated program */ + raster.light_twoside = ctx->Light.Enabled && ctx->Light.Model.TwoSide; + } } else if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { raster.light_twoside = 1; -- cgit v1.2.3 From 85cfe029b958c1954aefe0279eb2388bb272ec5c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 10:56:23 -0700 Subject: init front/back color attrib indexes to 0, not -1 --- src/mesa/pipe/draw/draw_context.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index d134b05717..39d324f97f 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -76,10 +76,10 @@ struct draw_context *draw_create( void ) draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * MAX_VERTEX_SIZE); } - draw->attrib_front0 = -1; - draw->attrib_back0 = -1; - draw->attrib_front1 = -1; - draw->attrib_back1 = -1; + draw->attrib_front0 = 0; + draw->attrib_back0 = 0; + draw->attrib_front1 = 0; + draw->attrib_back1 = 0; draw->prim = ~0; /* != any of PIPE_PRIM_x */ -- cgit v1.2.3 From 1b5f20c665d1174a52f5d1ec96b23a343ce0ac09 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 10:56:38 -0700 Subject: remove useless assertions --- src/mesa/pipe/draw/draw_twoside.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 00b4ee45cd..c7e268f11e 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -79,11 +79,9 @@ static struct vertex_header *copy_bfc( struct twoside_stage *twoside, const struct draw_context *draw = twoside->stage.draw; if (draw->attrib_front0 && draw->attrib_back0) { - assert(draw->attrib_back0); copy_attrib(draw->attrib_front0, draw->attrib_back0, tmp); } if (draw->attrib_front1 && draw->attrib_back1) { - assert(draw->attrib_back1); copy_attrib(draw->attrib_front1, draw->attrib_back1, tmp); } -- cgit v1.2.3 From 0e6b78c7eb7ebee9074489664416be489a27c742 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 10:57:10 -0700 Subject: s/DBG/DRAW_DBG --- src/mesa/pipe/draw/draw_vertex_fetch.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 7d983ebd29..3ca17f8829 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -37,7 +37,7 @@ #include "draw_vertex.h" -#define DBG 0 +#define DRAW_DBG 0 /** @@ -107,7 +107,7 @@ void draw_vertex_fetch( struct draw_context *draw, for (j = 0; j < count; j++) { uint attr; -#if DBG +#if DRAW_DBG printf("fetch vertex %u: \n", j); #endif @@ -124,7 +124,7 @@ void draw_vertex_fetch( struct draw_context *draw, fetch_attrib4(src, draw->vertex_element[attr].src_format, p); -#if DBG +#if DRAW_DBG printf(" %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]); #endif -- cgit v1.2.3 From d11fd189ff84ec24df0fb988e5c3e1a9260e038c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 10:57:53 -0700 Subject: vbuf updates/fixes (KeithW) Plus, update i915 and Cell drivers for latest vbuf->draw() params. --- src/mesa/pipe/cell/ppu/cell_vbuf.c | 34 ++++++++++++----- src/mesa/pipe/draw/draw_vbuf.c | 63 ++++++++++++++++++------------- src/mesa/pipe/draw/draw_vbuf.h | 6 +-- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 6 +-- 4 files changed, 63 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index d27d718dae..a420de041a 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -46,6 +46,8 @@ struct cell_vbuf_render struct vbuf_render base; struct cell_context *cell; uint prim; + uint vertex_size; + void *vertex_buffer; }; @@ -70,8 +72,12 @@ static void * cell_vbuf_allocate_vertices(struct vbuf_render *vbr, ushort vertex_size, ushort nr_vertices) { + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); /*printf("Alloc verts %u * %u\n", vertex_size, nr_vertices);*/ - return align_malloc(vertex_size * nr_vertices, 16); + assert(!cvbr->vertex_buffer); + cvbr->vertex_buffer = align_malloc(vertex_size * nr_vertices, 16); + cvbr->vertex_size = vertex_size; + return cvbr->vertex_buffer; } @@ -86,17 +92,22 @@ cell_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) static void cell_vbuf_draw(struct vbuf_render *vbr, - uint prim, const ushort *indices, - uint nr_indices, - const void *vertices, - uint nr_vertices, - uint vertex_size) + uint nr_indices) { struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); struct cell_context *cell = cvbr->cell; float xmin, ymin, xmax, ymax; uint i; + uint nr_vertices = 0; + const void *vertices = cvbr->vertex_buffer; + const uint vertex_size = cvbr->vertex_size; + + for (i = 0; i < nr_indices; i++) { + if (indices[i] > nr_vertices) + nr_vertices = indices[i]; + } + nr_vertices++; #if 0 printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u\n", @@ -127,14 +138,14 @@ cell_vbuf_draw(struct vbuf_render *vbr, ymax = v[1]; } - if (prim != PIPE_PRIM_TRIANGLES) + if (cvbr->prim != PIPE_PRIM_TRIANGLES) return; /* only render tris for now */ #if 0 for (i = 0; i < cell->num_spus; i++) { struct cell_command_render *render = &cell_global.command[i].render; render->opcode = CELL_CMD_RENDER; - render->prim_type = prim; + render->prim_type = cvbr->prim; render->num_verts = nr_vertices; render->num_attribs = CELL_MAX_ATTRIBS; /* XXX fix */ render->vertex_data = vertices; @@ -155,7 +166,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, = (struct cell_command_render *) cell_batch_alloc(cell, sizeof(*render)); render->opcode = CELL_CMD_RENDER; - render->prim_type = prim; + render->prim_type = cvbr->prim; render->num_verts = nr_vertices; render->num_attribs = CELL_MAX_ATTRIBS; /* XXX fix */ render->vertex_data = vertices; @@ -182,8 +193,13 @@ static void cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, unsigned vertex_size, unsigned vertices_used) { + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + /*printf("Free verts %u * %u\n", vertex_size, vertices_used);*/ align_free(vertices); + + assert(vertices == cvbr->vertex_buffer); + cvbr->vertex_buffer = NULL; } diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 4f59b1b25d..6cda122c3a 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -83,7 +83,8 @@ vbuf_stage( struct draw_stage *stage ) static void vbuf_flush_indices( struct draw_stage *stage ); -static void vbuf_flush_vertices( struct draw_stage *stage, +static void vbuf_flush_vertices( struct draw_stage *stage ); +static void vbuf_alloc_vertices( struct draw_stage *stage, unsigned new_vertex_size ); @@ -98,8 +99,10 @@ overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) static INLINE void check_space( struct vbuf_stage *vbuf, unsigned nr ) { - if (vbuf->nr_vertices + nr > vbuf->max_vertices ) - vbuf_flush_vertices(&vbuf->stage, vbuf->vertex_size ); + if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { + vbuf_flush_vertices(&vbuf->stage); + vbuf_alloc_vertices(&vbuf->stage, vbuf->vertex_size); + } if (vbuf->nr_indices + nr > vbuf->max_indices ) vbuf_flush_indices(&vbuf->stage); @@ -232,10 +235,12 @@ vbuf_first_tri( struct draw_stage *stage, struct vbuf_stage *vbuf = vbuf_stage( stage ); vbuf_flush_indices( stage ); + stage->tri = vbuf_tri; - stage->tri( stage, prim ); vbuf->prim = PIPE_PRIM_TRIANGLES; vbuf->render->set_primitive(vbuf->render, PIPE_PRIM_TRIANGLES); + + stage->tri( stage, prim ); } @@ -247,9 +252,10 @@ vbuf_first_line( struct draw_stage *stage, vbuf_flush_indices( stage ); stage->line = vbuf_line; - stage->line( stage, prim ); vbuf->prim = PIPE_PRIM_LINES; vbuf->render->set_primitive(vbuf->render, PIPE_PRIM_LINES); + + stage->line( stage, prim ); } @@ -260,10 +266,12 @@ vbuf_first_point( struct draw_stage *stage, struct vbuf_stage *vbuf = vbuf_stage( stage ); vbuf_flush_indices( stage ); + stage->point = vbuf_point; - stage->point( stage, prim ); vbuf->prim = PIPE_PRIM_POINTS; vbuf->render->set_primitive(vbuf->render, PIPE_PRIM_POINTS); + + stage->point( stage, prim ); } @@ -291,15 +299,13 @@ vbuf_flush_indices( struct draw_stage *stage ) assert(0); } - vbuf->render->draw( vbuf->render, - vbuf->prim, - vbuf->indices, - vbuf->nr_indices, - vbuf->vertices, - vbuf->nr_vertices, - vbuf->vertex_size ); - + vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); + vbuf->nr_indices = 0; + + stage->point = vbuf_first_point; + stage->line = vbuf_first_line; + stage->tri = vbuf_first_tri; } @@ -311,8 +317,7 @@ vbuf_flush_indices( struct draw_stage *stage ) * we flush. */ static void -vbuf_flush_vertices( struct draw_stage *stage, - unsigned new_vertex_size ) +vbuf_flush_vertices( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); @@ -332,15 +337,24 @@ vbuf_flush_vertices( struct draw_stage *stage, vbuf->vertex_ptr = vbuf->vertices = NULL; } +} + +static void +vbuf_alloc_vertices( struct draw_stage *stage, + unsigned new_vertex_size ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + assert(!vbuf->nr_indices); + assert(!vbuf->vertices); /* Allocate a new vertex buffer */ vbuf->vertex_size = new_vertex_size; vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; vbuf->vertices = vbuf->render->allocate_vertices(vbuf->render, - vbuf->vertex_size, - vbuf->max_vertices) ; + (ushort) vbuf->vertex_size, + (ushort) vbuf->max_vertices); vbuf->vertex_ptr = vbuf->vertices; } @@ -352,22 +366,17 @@ vbuf_begin( struct draw_stage *stage ) const struct vertex_info *vinfo = vbuf->render->get_vertex_info(vbuf->render); unsigned vertex_size = vinfo->size * sizeof(float); - if(vbuf->vertex_size != vertex_size) - vbuf_flush_vertices(&vbuf->stage, vertex_size); + /* XXX: Overkill */ + vbuf_alloc_vertices(&vbuf->stage, vertex_size); } static void vbuf_end( struct draw_stage *stage ) { -#if 0 +// vbuf_flush_indices( stage ); /* XXX: Overkill */ - vbuf_flush_indices( stage ); -#else - /* By flushing vertices we avoid having the vertex buffer grow and grow */ - struct vbuf_stage *vbuf = vbuf_stage(stage); - vbuf_flush_vertices( stage, vbuf->vertex_size ); -#endif + vbuf_flush_vertices( stage ); stage->point = vbuf_first_point; stage->line = vbuf_first_line; diff --git a/src/mesa/pipe/draw/draw_vbuf.h b/src/mesa/pipe/draw/draw_vbuf.h index be4c4ab77b..cfd2b9820c 100644 --- a/src/mesa/pipe/draw/draw_vbuf.h +++ b/src/mesa/pipe/draw/draw_vbuf.h @@ -82,12 +82,8 @@ struct vbuf_render { * DrawElements, note indices are ushort: */ void (*draw)( struct vbuf_render *, - uint prim, const ushort *indices, - uint nr_indices, - const void *vertices, - uint nr_vertices, - uint vertex_bytes); + uint nr_indices ); /** * Called when vbuf is done with this set of vertices: diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index bdcc027ed7..d85a6d3c0f 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -136,12 +136,8 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render, static void i915_vbuf_render_draw( struct vbuf_render *render, - uint prim, const ushort *indices, - uint nr_indices, - const void *vertices, - uint nr_vertices, - uint vertex_bytes) + uint nr_indices) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; -- cgit v1.2.3 From 27cff4402e9582fed8152e88a2c36e488bf76198 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 13:54:32 -0700 Subject: copy UsesKill state in _mesa_combine_programs() --- src/mesa/shader/program.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index fc53b5737c..3869e31714 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -497,6 +497,13 @@ _mesa_combine_programs(GLcontext *ctx, newProg->NumInstructions = newLength; if (newProg->Target == GL_FRAGMENT_PROGRAM_ARB) { + struct gl_fragment_program *fprogA, *fprogB, *newFprog; + fprogA = (struct gl_fragment_program *) progA; + fprogB = (struct gl_fragment_program *) progB; + newFprog = (struct gl_fragment_program *) newProg; + + newFprog->UsesKill = fprogA->UsesKill || fprogB->UsesKill; + /* connect color outputs/inputs */ if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) && (progB->InputsRead & (1 << FRAG_ATTRIB_COL0))) { -- cgit v1.2.3 From 13bfa87b09294607e73d3993878956d5b344bbb3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 13:55:08 -0700 Subject: tweaks to quad drawing, set UsesKill in bitmap shader --- src/mesa/state_tracker/st_cb_drawpixels.c | 27 +++++++++++++++------------ 1 file changed, 15 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index f4d6b9362c..eaa88d3c6c 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -158,6 +158,7 @@ make_bitmap_fragment_program(GLcontext *ctx) p->OutputsWritten = 0x0; stfp = (struct st_fragment_program *) p; + stfp->Base.UsesKill = GL_TRUE; st_translate_fragment_program(ctx->st, stfp, NULL, stfp->tokens, ST_MAX_SHADER_TOKENS); @@ -536,30 +537,31 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, { GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ GLuint i; + GLfloat sLeft = 0.0, sRight = 1.0; GLfloat tTop = invertTex, tBot = 1.0 - tTop; /* upper-left */ verts[0][0][0] = x0; /* attr[0].x */ - verts[0][0][1] = y0; /* attr[0].x */ - verts[0][1][0] = 0.0; /* attr[1].s */ + verts[0][0][1] = y0; /* attr[0].y */ + verts[0][1][0] = sLeft; /* attr[1].s */ verts[0][1][1] = tTop; /* attr[1].t */ /* upper-right */ verts[1][0][0] = x1; verts[1][0][1] = y0; - verts[1][1][0] = 1.0; + verts[1][1][0] = sRight; verts[1][1][1] = tTop; /* lower-right */ verts[2][0][0] = x1; verts[2][0][1] = y1; - verts[2][1][0] = 1.0; + verts[2][1][0] = sRight; verts[2][1][1] = tBot; /* lower-left */ verts[3][0][0] = x0; verts[3][0][1] = y1; - verts[3][1][0] = 0.0; + verts[3][1][0] = sLeft; verts[3][1][1] = tBot; /* same for all verts: */ @@ -581,30 +583,31 @@ draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, { GLfloat verts[4][3][4]; /* four verts, three attribs, XYZW */ GLuint i; + GLfloat sLeft = 0.0, sRight = 1.0; GLfloat tTop = invertTex, tBot = 1.0 - tTop; /* upper-left */ verts[0][0][0] = x0; /* attr[0].x */ verts[0][0][1] = y0; /* attr[0].y */ - verts[0][2][0] = 0.0; /* attr[2].s */ + verts[0][2][0] = sLeft; /* attr[2].s */ verts[0][2][1] = tTop; /* attr[2].t */ /* upper-right */ verts[1][0][0] = x1; verts[1][0][1] = y0; - verts[1][2][0] = 1.0; + verts[1][2][0] = sRight; verts[1][2][1] = tTop; /* lower-right */ verts[2][0][0] = x1; verts[2][0][1] = y1; - verts[2][2][0] = 1.0; + verts[2][2][0] = sRight; verts[2][2][1] = tBot; /* lower-left */ verts[3][0][0] = x0; verts[3][0][1] = y1; - verts[3][2][0] = 0.0; + verts[3][2][0] = sLeft; verts[3][2][1] = tBot; /* same for all verts: */ @@ -669,9 +672,9 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, struct pipe_sampler_state sampler; const struct cso_sampler *cso; memset(&sampler, 0, sizeof(sampler)); - sampler.wrap_s = PIPE_TEX_WRAP_REPEAT; - sampler.wrap_t = PIPE_TEX_WRAP_REPEAT; - sampler.wrap_r = PIPE_TEX_WRAP_REPEAT; + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP; sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; -- cgit v1.2.3 From ecc9cf43af658413f91818fe90fdb0c2eaad9a46 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 16:29:36 -0700 Subject: Fix incorrect [0,1] clamping in store_dest(). --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 49affc12f3..dcc39362a9 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1135,10 +1135,8 @@ store_dest( case TGSI_SAT_ZERO_ONE: /* XXX need to obey ExecMask here */ - micro_lt( dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], - &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], chan ); - micro_lt( dst, chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], - chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); + micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); break; case TGSI_SAT_MINUS_PLUS_ONE: @@ -1646,6 +1644,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); + /* XXX use micro_min()?? */ micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); STORE(&r[0], 0, chan_index); @@ -1657,6 +1656,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); + /* XXX use micro_max()?? */ micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); STORE(&r[0], 0, chan_index ); -- cgit v1.2.3 From bdc38254618d5f984e206b5ddb484c3cbc55c3bb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 16:31:37 -0700 Subject: Fix broken fog. Note that fogcoord and front-facing attribs are in the same register, in the X and Y channels, respectively. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 722cadc5c0..7847027cae 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -554,12 +554,6 @@ static void setup_tri_coefficients( struct setup_stage *setup ) */ setup_fragcoord_coeff(setup); } - else if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } else { #endif uint j; @@ -578,8 +572,18 @@ static void setup_tri_coefficients( struct setup_stage *setup ) break; default: /* invalid interp mode */ - assert(0); + /* assert(0); re-enable this and run demos/fogcoord.c ... */ + ; + } + + if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; } + + #if USE_INPUT_MAP } #endif -- cgit v1.2.3 From 846969917e164bc85b56c8d61d9549114b1321eb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 16:43:36 -0700 Subject: clean-up swizzle fields in fog code, fix NegateBase --- src/mesa/shader/programopt.c | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index fc5b0497fe..9eeb71db1b 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -192,13 +192,13 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog) inst->DstReg.WriteMask = WRITEMASK_X; inst->SrcReg[0].File = PROGRAM_INPUT; inst->SrcReg[0].Index = FRAG_ATTRIB_FOGC; - inst->SrcReg[0].Swizzle = SWIZZLE_X; + inst->SrcReg[0].Swizzle = SWIZZLE_XXXX; inst->SrcReg[1].File = PROGRAM_STATE_VAR; inst->SrcReg[1].Index = fogPRefOpt; - inst->SrcReg[1].Swizzle = SWIZZLE_X; + inst->SrcReg[1].Swizzle = SWIZZLE_XXXX; inst->SrcReg[2].File = PROGRAM_STATE_VAR; inst->SrcReg[2].Index = fogPRefOpt; - inst->SrcReg[2].Swizzle = SWIZZLE_Y; + inst->SrcReg[2].Swizzle = SWIZZLE_YYYY; inst->SaturateMode = SATURATE_ZERO_ONE; inst++; } @@ -214,10 +214,10 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog) inst->SrcReg[0].File = PROGRAM_STATE_VAR; inst->SrcReg[0].Index = fogPRefOpt; inst->SrcReg[0].Swizzle - = (fprog->FogOption == GL_EXP) ? SWIZZLE_Z : SWIZZLE_W; + = (fprog->FogOption == GL_EXP) ? SWIZZLE_ZZZZ : SWIZZLE_WWWW; inst->SrcReg[1].File = PROGRAM_INPUT; inst->SrcReg[1].Index = FRAG_ATTRIB_FOGC; - inst->SrcReg[1].Swizzle = SWIZZLE_X; + inst->SrcReg[1].Swizzle = SWIZZLE_XXXX; inst++; if (fprog->FogOption == GL_EXP2) { /* MUL fogFactorTemp.x, fogFactorTemp.x, fogFactorTemp.x; */ @@ -227,10 +227,10 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog) inst->DstReg.WriteMask = WRITEMASK_X; inst->SrcReg[0].File = PROGRAM_TEMPORARY; inst->SrcReg[0].Index = fogFactorTemp; - inst->SrcReg[0].Swizzle = SWIZZLE_X; + inst->SrcReg[0].Swizzle = SWIZZLE_XXXX; inst->SrcReg[1].File = PROGRAM_TEMPORARY; inst->SrcReg[1].Index = fogFactorTemp; - inst->SrcReg[1].Swizzle = SWIZZLE_X; + inst->SrcReg[1].Swizzle = SWIZZLE_XXXX; inst++; } /* EX2_SAT fogFactorTemp.x, -fogFactorTemp.x; */ @@ -240,8 +240,8 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog) inst->DstReg.WriteMask = WRITEMASK_X; inst->SrcReg[0].File = PROGRAM_TEMPORARY; inst->SrcReg[0].Index = fogFactorTemp; - inst->SrcReg[0].NegateBase = GL_TRUE; - inst->SrcReg[0].Swizzle = SWIZZLE_X; + inst->SrcReg[0].NegateBase = NEGATE_XYZW; + inst->SrcReg[0].Swizzle = SWIZZLE_XXXX; inst->SaturateMode = SATURATE_ZERO_ONE; inst++; } @@ -252,8 +252,7 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog) inst->DstReg.WriteMask = WRITEMASK_XYZ; inst->SrcReg[0].File = PROGRAM_TEMPORARY; inst->SrcReg[0].Index = fogFactorTemp; - inst->SrcReg[0].Swizzle - = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X); + inst->SrcReg[0].Swizzle = SWIZZLE_XXXX; inst->SrcReg[1].File = PROGRAM_TEMPORARY; inst->SrcReg[1].Index = colorTemp; inst->SrcReg[1].Swizzle = SWIZZLE_NOOP; -- cgit v1.2.3 From 5ea37f591a06073611b8276a87cc7c745e8b3fcb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 16:43:50 -0700 Subject: use NEGATE_X/Y/Z/W tokens --- src/mesa/shader/prog_print.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index 969b9757cc..1145f0c7a4 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -304,7 +304,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) if (!extended) s[i++] = '.'; - if (negateBase & 0x1) + if (negateBase & NEGATE_X) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 0)]; @@ -312,7 +312,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) s[i++] = ','; } - if (negateBase & 0x2) + if (negateBase & NEGATE_Y) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 1)]; @@ -320,7 +320,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) s[i++] = ','; } - if (negateBase & 0x4) + if (negateBase & NEGATE_Z) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 2)]; @@ -328,7 +328,7 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) s[i++] = ','; } - if (negateBase & 0x8) + if (negateBase & NEGATE_W) s[i++] = '-'; s[i++] = swz[GET_SWZ(swizzle, 3)]; -- cgit v1.2.3 From 83417a8b37211e290cfdb06530533cd7215500c4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 14 Dec 2007 15:50:06 +0000 Subject: gallium: whitespace changes --- src/mesa/pipe/draw/draw_context.h | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 60be3e194d..91e11e6930 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -100,15 +100,13 @@ void draw_delete_vertex_shader(struct draw_context *draw, boolean draw_use_sse(struct draw_context *draw); -void -draw_set_vertex_buffer(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_buffer *buffer); - -void -draw_set_vertex_element(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_element *element); +void draw_set_vertex_buffer(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_buffer *buffer); + +void draw_set_vertex_element(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_element *element); void draw_set_mapped_element_buffer( struct draw_context *draw, unsigned eltSize, void *elements ); -- cgit v1.2.3 From 66719dc63a169d5bec5e56fa5ad46853121a8dc0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 14 Dec 2007 15:51:07 +0000 Subject: gallium: hardwire reset_temps functionality --- src/mesa/pipe/draw/draw_context.c | 20 +++++--------------- src/mesa/pipe/draw/draw_linestipple.c | 1 - src/mesa/pipe/draw/draw_private.h | 7 ------- 3 files changed, 5 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 39d324f97f..fd43d690f6 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -238,17 +238,6 @@ void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) } } -/** - * Reset the vertex ids for the stage's temp verts. - */ -void draw_reset_tmps( struct draw_stage *stage ) -{ - unsigned i; - - if (stage->tmp) - for (i = 0; i < stage->nr_tmps; i++) - stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; -} void draw_free_tmps( struct draw_stage *stage ) { @@ -271,10 +260,11 @@ void draw_reset_vertex_ids(struct draw_context *draw) struct draw_stage *stage = draw->pipeline.first; while (stage) { - if (stage->reset_tmps) - stage->reset_tmps(stage); - else - draw_reset_tmps(stage); + unsigned i; + + for (i = 0; i < stage->nr_tmps; i++) + stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; + stage = stage->next; } diff --git a/src/mesa/pipe/draw/draw_linestipple.c b/src/mesa/pipe/draw/draw_linestipple.c index 5f0db99b23..d7fe1071f0 100644 --- a/src/mesa/pipe/draw/draw_linestipple.c +++ b/src/mesa/pipe/draw/draw_linestipple.c @@ -265,7 +265,6 @@ struct clip_pipe_stage *clip_pipe_stipple( struct clip_pipeline *pipe ) stipple->stage.point = clip_passthrough_point; stipple->stage.line = stipple_line; stipple->stage.tri = clip_passthrough_tri; - stipple->stage.reset_tmps = clip_pipe_reset_tmps; stipple->stage.end = stipple_end; stipple->stage.destroy = stipple_destroy; diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 5f89ac121e..d8832449ea 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -116,13 +116,6 @@ struct draw_stage void (*end)( struct draw_stage * ); - /** - * Reset temporary vertex ids in this stage - * - * draw_free_tmps will be called instead if null. - */ - void (*reset_tmps)( struct draw_stage * ); - void (*reset_stipple_counter)( struct draw_stage * ); void (*destroy)( struct draw_stage * ); -- cgit v1.2.3 From 7771519862ade53bbd29dd25987c51e3c444e364 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 17 Dec 2007 11:59:02 +0000 Subject: gallium: fix warning --- src/mesa/pipe/i915simple/i915_debug.h | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h index 13197f3595..0bcd094233 100644 --- a/src/mesa/pipe/i915simple/i915_debug.h +++ b/src/mesa/pipe/i915simple/i915_debug.h @@ -69,11 +69,13 @@ void i915_print_ureg(const char *msg, unsigned ureg); #define DEBUG_SURFACE 0x2000 #define DEBUG_WINSYS 0x4000 +#include "pipe/p_compiler.h" + #if defined(DEBUG) && defined(FILE_DEBUG_FLAG) #include "pipe/p_winsys.h" -static void +static INLINE void I915_DBG( struct i915_context *i915, const char *fmt, @@ -92,7 +94,7 @@ I915_DBG( #else -static void +static INLINE void I915_DBG( struct i915_context *i915, const char *fmt, -- cgit v1.2.3 From af51d188c3e894c2ac3d261831320f178d1bd6fe Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 17 Dec 2007 12:03:51 +0000 Subject: gallium: fix warning --- src/mesa/pipe/softpipe/sp_winsys.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h index cbf64ebb85..d6b379f58c 100644 --- a/src/mesa/pipe/softpipe/sp_winsys.h +++ b/src/mesa/pipe/softpipe/sp_winsys.h @@ -37,6 +37,7 @@ #include "pipe/p_compiler.h" /* for boolean */ +enum pipe_format; struct softpipe_winsys { /** test if the given format is supported for front/back color bufs */ -- cgit v1.2.3 From 39b8303d4af97235b4daeba5a9ee0e12640e0be5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 18 Dec 2007 18:59:49 +0000 Subject: gallium: Detect if a null buffer is passed. --- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c index e69f88a323..dfb80b1dcf 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -230,6 +230,9 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list, { struct fenced_buffer *buf; + if(!buffer) + return NULL; + buf = (struct fenced_buffer *)calloc(1, sizeof(struct fenced_buffer)); if(!buf) return NULL; -- cgit v1.2.3 From 3eaf882a662f719845ead1017e28a121cbce86fb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 18 Dec 2007 22:01:35 +0000 Subject: gallium: Alternative mm_bufmgr_create_from_buffer constructor for mm pool managers. --- src/mesa/pipe/pipebuffer/pb_bufmgr.h | 11 ++++++++++- src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 34 ++++++++++++++++++++++++++------- 2 files changed, 37 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr.h b/src/mesa/pipe/pipebuffer/pb_bufmgr.h index 0e6c3a8d37..12e36323a8 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr.h +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr.h @@ -65,7 +65,7 @@ struct buffer_manager /* XXX: we will likely need more allocation flags */ struct pipe_buffer * (*create_buffer)( struct buffer_manager *mgr, - size_t size); + size_t size ); void (*destroy)( struct buffer_manager *mgr ); @@ -96,6 +96,15 @@ struct buffer_manager * mm_bufmgr_create(struct buffer_manager *provider, size_t size, size_t align2); +/** + * Same as mm_bufmgr_create. + * + * Buffer will be release when the manager is destroyed. + */ +struct buffer_manager * +mm_bufmgr_create_from_buffer(struct pipe_buffer *buffer, + size_t size, size_t align2); + /** * Fenced buffer manager. diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index 8e6dcf1116..e184655f1a 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -218,11 +218,14 @@ mm_bufmgr_destroy(struct buffer_manager *mgr) struct buffer_manager * -mm_bufmgr_create(struct buffer_manager *provider, - size_t size, size_t align2) +mm_bufmgr_create_from_buffer(struct pipe_buffer *buffer, + size_t size, size_t align2) { struct mm_buffer_manager *mm; + if(!buffer) + return NULL; + mm = (struct mm_buffer_manager *)calloc(1, sizeof(*mm)); if (!mm) return NULL; @@ -237,9 +240,7 @@ mm_bufmgr_create(struct buffer_manager *provider, _glthread_INIT_MUTEX(mm->mutex); - mm->buffer = provider->create_buffer(provider, size); - if (!mm->buffer) - goto failure; + mm->buffer = buffer; mm->map = buffer_map(mm->buffer, PIPE_BUFFER_FLAG_READ | PIPE_BUFFER_FLAG_WRITE ); @@ -257,9 +258,28 @@ if(mm->heap) mmDestroy(mm->heap); if(mm->map) buffer_unmap(mm->buffer); - if(mm->buffer) - buffer_release(mm->buffer); if(mm) free(mm); return NULL; } + + +struct buffer_manager * +mm_bufmgr_create(struct buffer_manager *provider, + size_t size, size_t align2) +{ + struct pipe_buffer *buffer; + struct buffer_manager *mgr; + + buffer = provider->create_buffer(provider, size); + if (!buffer) + return NULL; + + mgr = mm_bufmgr_create_from_buffer(buffer, size, align2); + if (!mgr) { + buffer_release(buffer); + return NULL; + } + + return mgr; +} -- cgit v1.2.3 From bf3bdb3b42632e7621e1cbae0d24e60b05ce16e0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 18 Dec 2007 22:19:40 +0000 Subject: gallium: Fix cut-n-paste error. --- src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index e184655f1a..e9bc34b4a4 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -230,8 +230,6 @@ mm_bufmgr_create_from_buffer(struct pipe_buffer *buffer, if (!mm) return NULL; - assert(provider); - assert(provider->create_buffer); mm->base.create_buffer = mm_bufmgr_create_buffer; mm->base.destroy = mm_bufmgr_destroy; @@ -271,6 +269,8 @@ mm_bufmgr_create(struct buffer_manager *provider, struct pipe_buffer *buffer; struct buffer_manager *mgr; + assert(provider); + assert(provider->create_buffer); buffer = provider->create_buffer(provider, size); if (!buffer) return NULL; -- cgit v1.2.3 From 310e7ca44b33558a9aa22f43024d1bfb63642128 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 16 Jan 2008 17:21:05 -0700 Subject: temporarily don't emit TGSI immediates, use the constant buffer This allows a few more programs to run and not crash. Need to get constant buffer uploads working... --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 325aa20173..936ba9ed85 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -41,6 +41,7 @@ #define TGSI_DEBUG 0 +#define EMIT_IMMEDIATES 0 /* * Map mesa register file to TGSI register file. @@ -68,12 +69,17 @@ map_register_file( case PROGRAM_STATE_VAR: case PROGRAM_NAMED_PARAM: case PROGRAM_UNIFORM: - if (immediateMapping[index] != ~0) + if (immediateMapping[index] != ~0) { return TGSI_FILE_IMMEDIATE; + } else return TGSI_FILE_CONSTANT; case PROGRAM_CONSTANT: +#if EMIT_IMMEDIATES return TGSI_FILE_IMMEDIATE; +#else + return TGSI_FILE_CONSTANT; +#endif case PROGRAM_INPUT: return TGSI_FILE_INPUT; case PROGRAM_OUTPUT: @@ -828,7 +834,7 @@ tgsi_translate_mesa_program( /* immediates/literals */ memset(immediates, ~0, sizeof(immediates)); - +#if EMIT_IMMEDIATES for (i = 0; program->Parameters && i < program->Parameters->NumParameters; i++) { if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { @@ -843,6 +849,7 @@ tgsi_translate_mesa_program( numImmediates++; } } +#endif /* constant buffer refs */ { -- cgit v1.2.3 From c04f4174923cda7beed1945f8ab77fd5cea1b50c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 17 Jan 2008 11:00:51 +0900 Subject: gallium: Handle TGSI_OPCODE_END on sse2. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 8ee43f59b5..f8660e7ad1 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -1967,6 +1967,7 @@ emit_instruction( break; case TGSI_OPCODE_RET: + case TGSI_OPCODE_END: #ifdef WIN32 emit_retw( func, 16 ); #else -- cgit v1.2.3 From b016f0adba8278f3744d3aaa207a1b586d51756d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 17 Jan 2008 13:37:17 +0900 Subject: gallium: Pass PIPE_BUFFER_USAGE_PIXEL to buffer_data. --- src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index c7b519d95b..340fa2fb85 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -227,7 +227,7 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, surf->buffer, surf->pitch * surf->cpp * height, NULL, - 0); + PIPE_BUFFER_USAGE_PIXEL); if(ret) { winsys->buffer_reference(winsys, &surf->buffer, NULL); return ret; -- cgit v1.2.3 From 271f9dac79a9247de9a57f4d248e404bf1652a13 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 17 Jan 2008 13:39:14 +0900 Subject: Back-port miscellaneous fixes from internal branch (mostly portability fixes). These are changes that are in our internal branch, but somehow were skipped so far. It was done using visual comparison of the branches -- it is likely that changes are being carried on the wrong way --- src/mesa/pipe/Makefile.template | 9 +++-- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 2 +- src/mesa/pipe/i915simple/i915_surface.c | 1 + src/mesa/pipe/i915simple/i915_texture.c | 12 +++--- src/mesa/pipe/p_compiler.h | 19 +++++++++ src/mesa/pipe/p_format.h | 2 + src/mesa/pipe/p_shader_tokens.h | 3 +- src/mesa/pipe/p_util.h | 64 +++++++++++++++++++------------ src/mesa/pipe/p_winsys.h | 5 ++- src/mesa/pipe/softpipe/sp_prim_setup.c | 2 +- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 +- src/mesa/pipe/softpipe/sp_quad_stipple.c | 2 + src/mesa/pipe/softpipe/sp_surface.c | 1 + src/mesa/pipe/softpipe/sp_texture.c | 2 +- src/mesa/pipe/softpipe/sp_tile_cache.c | 10 ++--- src/mesa/pipe/tgsi/util/tgsi_build.c | 6 +-- src/mesa/state_tracker/st_cb_fbo.c | 2 + src/mesa/state_tracker/st_public.h | 3 ++ 18 files changed, 98 insertions(+), 49 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/Makefile.template b/src/mesa/pipe/Makefile.template index 3cd07660b6..8e84f8eb2d 100644 --- a/src/mesa/pipe/Makefile.template +++ b/src/mesa/pipe/Makefile.template @@ -17,7 +17,8 @@ INCLUDES = \ -I. \ -I$(TOP)/src/mesa/pipe \ -I$(TOP)/src/mesa \ - -I$(TOP)/include + -I$(TOP)/include \ + $(DRIVER_INCLUDES) ##### RULES ##### @@ -34,11 +35,11 @@ INCLUDES = \ ##### TARGETS ##### -default:: depend symlinks $(LIBNAME) +default: depend symlinks $(LIBNAME) $(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/mesa/pipe/Makefile.template - $(TOP)/bin/mklib -o $@ -static $(OBJECTS) + $(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS) depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) @@ -54,7 +55,7 @@ tags: # Remove .o and backup files -clean: +clean:: -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) -rm -f depend depend.bak diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index d85a6d3c0f..edc62e25e5 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -202,7 +202,7 @@ static void i915_vbuf_render_destroy( struct vbuf_render *render ) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); - free(i915_render); + FREE(i915_render); } diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index e3c3cdd2e4..72a65e0fb4 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -64,6 +64,7 @@ i915_get_tex_surface(struct pipe_context *pipe, ps = pipe->winsys->surface_alloc(pipe->winsys); if (ps) { assert(ps->refcount); + assert(ps->winsys); pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); ps->format = pt->format; ps->cpp = pt->cpp; diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c index 44f72e63cc..6b0a4a96f3 100644 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -123,7 +123,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) * 2nd mipmap out past the width of its parent. */ if (pt->first_level != pt->last_level) { - unsigned mip1_width = align(minify(pt->width[0]), align_w) + unsigned mip1_width = align_int(minify(pt->width[0]), align_w) + minify(minify(pt->width[0])); if (mip1_width > pt->width[0]) @@ -133,7 +133,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) /* Pitch must be a whole number of dwords, even though we * express it in texels. */ - tex->pitch = align(tex->pitch * pt->cpp, 4) / pt->cpp; + tex->pitch = align_int(tex->pitch * pt->cpp, 4) / pt->cpp; tex->total_height = 0; for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { @@ -144,7 +144,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) if (pt->compressed) img_height = MAX2(1, height/4); else - img_height = align(height, align_h); + img_height = align_int(height, align_h); /* Because the images are packed better, the final offset @@ -155,7 +155,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) /* Layout_below: step right after second mipmap. */ if (level == pt->first_level + 1) { - x += align(width, align_w); + x += align_int(width, align_w); } else { y += img_height; @@ -531,9 +531,9 @@ i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) if (tex->image_offset[i]) - free(tex->image_offset[i]); + FREE(tex->image_offset[i]); - free(tex); + FREE(tex); } *pt = NULL; } diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h index 4f2c9ef88a..ab9609deab 100644 --- a/src/mesa/pipe/p_compiler.h +++ b/src/mesa/pipe/p_compiler.h @@ -50,6 +50,25 @@ typedef unsigned short ushort; typedef unsigned long long uint64; +#if defined(__MSC__) + +typedef unsigned short uint16_t; +typedef long int32_t; +typedef unsigned long uint32_t; +typedef long long int64_t; +typedef unsigned long long uint64_t; + +#if defined(_WIN64) +typedef unsigned __int64 uintptr_t; +#else +typedef unsigned int uintptr_t; +#endif + +#else +#include +#endif + + #define TRUE 1 #define FALSE 0 diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index b1772b352f..9f60cdbb04 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -97,6 +97,8 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) return (f >> shift) & mask; } +/* XXX: The bit layout needs to be revised, can't currently encode 10-bit components. */ + #define pf_swizzle_x(f) pf_get(f, 2, 0x7) /**< PIPE_FORMAT_COMP_ */ #define pf_swizzle_y(f) pf_get(f, 5, 0x7) /**< PIPE_FORMAT_COMP_ */ #define pf_swizzle_z(f) pf_get(f, 8, 0x7) /**< PIPE_FORMAT_COMP_ */ diff --git a/src/mesa/pipe/p_shader_tokens.h b/src/mesa/pipe/p_shader_tokens.h index e5922b439f..e9d1d66bda 100644 --- a/src/mesa/pipe/p_shader_tokens.h +++ b/src/mesa/pipe/p_shader_tokens.h @@ -109,7 +109,8 @@ struct tgsi_declaration_interpolation #define TGSI_SEMANTIC_FOG 3 #define TGSI_SEMANTIC_PSIZE 4 #define TGSI_SEMANTIC_GENERIC 5 -#define TGSI_SEMANTIC_COUNT 6 /**< number of semantic values */ +#define TGSI_SEMANTIC_NORMAL 6 +#define TGSI_SEMANTIC_COUNT 7 /**< number of semantic values */ struct tgsi_declaration_semantic { diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 69ec62e307..573fef9b83 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -111,6 +111,32 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) +/** + * Return a pointer aligned to next multiple of N bytes. + */ +static INLINE void * +align_pointer( void *unaligned, uint alignment ) +{ + if (sizeof(void *) == 64) { + union { + void *p; + uint64 u; + } pu; + pu.p = unaligned; + pu.u = (pu.u + alignment - 1) & ~(uint64) (alignment - 1); + return pu.p; + } + else { + /* 32-bit pointers */ + union { + void *p; + uint u; + } pu; + pu.p = unaligned; + pu.u = (pu.u + alignment - 1) & ~(alignment - 1); + return pu.p; + } +} /** * Return memory on given byte alignment @@ -123,19 +149,18 @@ align_malloc(size_t bytes, uint alignment) (void) posix_memalign(& mem, alignment, bytes); return mem; #else - typedef unsigned long int uintptr_t; - uintptr_t ptr, buf; + char *ptr, *buf; assert( alignment > 0 ); - ptr = (uintptr_t) MALLOC(bytes + alignment + sizeof(void *)); + ptr = (char *) MALLOC(bytes + alignment + sizeof(void *)); if (!ptr) return NULL; - buf = (ptr + alignment + sizeof(void *)) & ~(uintptr_t)(alignment - 1); - *(uintptr_t *)(buf - sizeof(void *)) = ptr; + buf = (char *) align_pointer( ptr + sizeof(void *), alignment ); + *(char **)(buf - sizeof(void *)) = ptr; - return (void *) buf; + return buf; #endif /* defined(HAVE_POSIX_MEMALIGN) */ } @@ -169,28 +194,17 @@ align_free(void *ptr) static INLINE void * align16( void *unaligned ) { - if (sizeof(void *) == 64) { - union { - void *p; - uint64 u; - } pu; - pu.p = unaligned; - pu.u = (pu.u + 15) & ~15; - return pu.p; - } - else { - /* 32-bit pointers */ - union { - void *p; - uint u; - } pu; - pu.p = unaligned; - pu.u = (pu.u + 15) & ~15; - return pu.p; - } + return align_pointer( unaligned, 16 ); } +static INLINE int align_int(int x, int align) +{ + return (x + align - 1) & ~(align - 1); +} + + + #if defined(__MSC__) && defined(__WIN32__) static INLINE unsigned ffs( unsigned u ) { diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index aa9362ec0b..75c6dc7e85 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -79,7 +79,10 @@ struct pipe_winsys /** allocate a new surface (no context dependency) */ struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws); - /** allocate storage for a pipe_surface */ + /** + * Allocate storage for a pipe_surface. + * Returns 0 if succeeds. + */ int (*surface_alloc_storage)(struct pipe_winsys *ws, struct pipe_surface *surf, unsigned width, unsigned height, diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 7847027cae..89f8df945c 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -488,7 +488,7 @@ setup_fragcoord_coeff(struct setup_stage *setup) if (setup->softpipe->rasterizer->origin_lower_left) { /* y=0=bottom */ const int winHeight = setup->softpipe->framebuffer.cbufs[0]->height; - setup->coef[0].a0[1] = winHeight - 1; + setup->coef[0].a0[1] = (float) (winHeight - 1); setup->coef[0].dady[1] = -1.0; } else { diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 921dfbaccb..0001c76a80 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -133,7 +133,7 @@ shade_quad( machine->InterpCoefs = quad->coef; /* Compute X, Y, Z, W vals for this quad */ - setup_pos_vector(quad->posCoef, quad->x0, quad->y0, &machine->QuadPos); + setup_pos_vector(quad->posCoef, (float) quad->x0, (float) quad->y0, &machine->QuadPos); /* run shader */ #if defined(__i386__) || defined(__386__) diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c index 0c42963dfe..8660432259 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ b/src/mesa/pipe/softpipe/sp_quad_stipple.c @@ -36,6 +36,7 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) stipple1 = softpipe->poly_stipple.stipple[y1 % 32]; #if 1 + { const int col0 = quad->x0 % 32; if ((stipple0 & (bit31 >> col0)) == 0) quad->mask &= ~MASK_TOP_LEFT; @@ -48,6 +49,7 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) if ((stipple1 & (bit30 >> col0)) == 0) quad->mask &= ~MASK_BOTTOM_RIGHT; + } #else /* We'd like to use this code, but we'd need to redefine * MASK_TOP_LEFT to be (1 << 1) and MASK_TOP_RIGHT to be (1 << 0), diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 6c080d5b5c..a580fb3e82 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -50,6 +50,7 @@ softpipe_get_tex_surface(struct pipe_context *pipe, ps = pipe->winsys->surface_alloc(pipe->winsys); if (ps) { assert(ps->refcount); + assert(ps->winsys); pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); ps->format = pt->format; ps->cpp = pt->cpp; diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index e5e6bfe01b..532bcfcc51 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -126,7 +126,7 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) pipe->winsys->buffer_reference(pipe->winsys, &spt->buffer, NULL); - free(spt); + FREE(spt); } *pt = NULL; } diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 6515ce668c..1974f77459 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -286,7 +286,7 @@ clear_tile(struct softpipe_cached_tile *tile, else { for (i = 0; i < TILE_SIZE; i++) { for (j = 0; j < TILE_SIZE; j++) { - tile->data.depth16[i][j] = clear_value; + tile->data.depth16[i][j] = (ushort) clear_value; } } } @@ -564,10 +564,10 @@ sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue) r = g = b = a = 0; } - tc->clear_color[0] = r / 255.0; - tc->clear_color[1] = g / 255.0; - tc->clear_color[2] = b / 255.0; - tc->clear_color[3] = a / 255.0; + tc->clear_color[0] = r / 255.0f; + tc->clear_color[1] = g / 255.0f; + tc->clear_color[2] = b / 255.0f; + tc->clear_color[3] = a / 255.0f; #if TILE_CLEAR_OPTIMIZATION /* set flags to indicate all the tiles are cleared */ diff --git a/src/mesa/pipe/tgsi/util/tgsi_build.c b/src/mesa/pipe/tgsi/util/tgsi_build.c index 19b2aad921..67f7d2c2c2 100644 --- a/src/mesa/pipe/tgsi/util/tgsi_build.c +++ b/src/mesa/pipe/tgsi/util/tgsi_build.c @@ -365,7 +365,7 @@ tgsi_build_immediate( immediate = tgsi_default_immediate(); - header_bodysize_grow( header ); + header_bodysize_grow( header ); return immediate; } @@ -415,7 +415,7 @@ tgsi_build_full_immediate( struct tgsi_header *header, unsigned maxsize ) { - unsigned size = 0, i; + unsigned size = 0, i; struct tgsi_immediate *immediate; if( maxsize <= size ) @@ -433,7 +433,7 @@ tgsi_build_full_immediate( if32 = (struct tgsi_immediate_float32 *) &tokens[size]; size++; - *if32 = tgsi_build_immediate_float32( + *if32 = tgsi_build_immediate_float32( full_imm->u.ImmediateFloat32[i].Float, immediate, header ); diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 2d6b3fc749..f02cb3d133 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -93,6 +93,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, if (!strb->surface) { strb->surface = pipe->winsys->surface_alloc(pipe->winsys); assert(strb->surface); + assert(strb->surface->refcount); + assert(strb->surface->winsys); if (!strb->surface) return GL_FALSE; } diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index ed5ef1f159..78a8fde82b 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -28,6 +28,9 @@ #ifndef ST_PUBLIC_H #define ST_PUBLIC_H +#include "GL/gl.h" +#include "GL/internal/glcore.h" /* for __GLcontextModes */ + #include "pipe/p_compiler.h" #include "pipe/p_format.h" -- cgit v1.2.3 From 25e2b8d669e01aac551276af7f34d8708d8cb9d0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 17 Jan 2008 17:20:46 +0900 Subject: Revert "temporarily don't emit TGSI immediates, use the constant buffer" This reverts commit 310e7ca44b33558a9aa22f43024d1bfb63642128. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 936ba9ed85..325aa20173 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -41,7 +41,6 @@ #define TGSI_DEBUG 0 -#define EMIT_IMMEDIATES 0 /* * Map mesa register file to TGSI register file. @@ -69,17 +68,12 @@ map_register_file( case PROGRAM_STATE_VAR: case PROGRAM_NAMED_PARAM: case PROGRAM_UNIFORM: - if (immediateMapping[index] != ~0) { + if (immediateMapping[index] != ~0) return TGSI_FILE_IMMEDIATE; - } else return TGSI_FILE_CONSTANT; case PROGRAM_CONSTANT: -#if EMIT_IMMEDIATES return TGSI_FILE_IMMEDIATE; -#else - return TGSI_FILE_CONSTANT; -#endif case PROGRAM_INPUT: return TGSI_FILE_INPUT; case PROGRAM_OUTPUT: @@ -834,7 +828,7 @@ tgsi_translate_mesa_program( /* immediates/literals */ memset(immediates, ~0, sizeof(immediates)); -#if EMIT_IMMEDIATES + for (i = 0; program->Parameters && i < program->Parameters->NumParameters; i++) { if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { @@ -849,7 +843,6 @@ tgsi_translate_mesa_program( numImmediates++; } } -#endif /* constant buffer refs */ { -- cgit v1.2.3 From e44bdcf9789caf9971c3c94605fbff7ce66af7ba Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 18 Jan 2008 19:08:30 +0100 Subject: Unify copy_rect helpers. Some of the copies were buggy. --- src/mesa/pipe/i915simple/i915_surface.c | 57 +++++-------------------- src/mesa/pipe/i965simple/brw_surface.c | 56 +++++-------------------- src/mesa/pipe/p_util.h | 9 ++++ src/mesa/pipe/softpipe/sp_surface.c | 63 ++++++---------------------- src/mesa/pipe/util/p_util.c | 73 +++++++++++++++++++++++++++++++++ src/mesa/sources | 3 +- 6 files changed, 118 insertions(+), 143 deletions(-) create mode 100644 src/mesa/pipe/util/p_util.c (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 72a65e0fb4..e625f5a68c 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -77,41 +77,6 @@ i915_get_tex_surface(struct pipe_context *pipe, } -static void -copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte *src, - unsigned src_pitch, - unsigned src_x, - unsigned src_y) -{ - unsigned i; - - dst_pitch *= cpp; - src_pitch *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * dst_pitch; - width *= cpp; - - if (width == dst_pitch && width == src_pitch) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; - } - } -} - - /* Upload data to a rectangular sub-region. Lots of choices how to do this: * * - memcpy by span to current destination @@ -126,9 +91,9 @@ i915_surface_data(struct pipe_context *pipe, const void *src, unsigned src_pitch, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - copy_rect(pipe_surface_map(dst), - dst->cpp, dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); pipe_surface_unmap(dst); } @@ -148,14 +113,14 @@ i915_surface_copy(struct pipe_context *pipe, assert( dst->cpp == src->cpp ); if (0) { - copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - pipe_surface_map(src), - src->pitch, - srcx, srcy); + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe_surface_map(src), + src->pitch, + srcx, srcy); pipe_surface_unmap(src); pipe_surface_unmap(dst); diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c index 4eacbdf82b..76b8c73d5c 100644 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -76,40 +76,6 @@ brw_get_tex_surface(struct pipe_context *pipe, } -static void -copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte *src, - unsigned src_pitch, - unsigned src_x, - unsigned src_y) -{ - unsigned i; - - dst_pitch *= cpp; - src_pitch *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * dst_pitch; - width *= cpp; - - if (width == dst_pitch && width == src_pitch) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; - } - } -} - /* Upload data to a rectangular sub-region. Lots of choices how to do this: * * - memcpy by span to current destination @@ -124,9 +90,9 @@ brw_surface_data(struct pipe_context *pipe, const void *src, unsigned src_pitch, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - copy_rect(pipe_surface_map(dst) + dst->offset, - dst->cpp, dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); + pipe_copy_rect(pipe_surface_map(dst) + dst->offset, + dst->cpp, dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); pipe_surface_unmap(dst); } @@ -146,14 +112,14 @@ brw_surface_copy(struct pipe_context *pipe, assert(dst->cpp == src->cpp); if (0) { - copy_rect(pipe_surface_map(dst) + dst->offset, - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - pipe_surface_map(src) + src->offset, - src->pitch, - srcx, srcy); + pipe_copy_rect(pipe_surface_map(dst) + dst->offset, + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe_surface_map(src) + src->offset, + src->pitch, + srcx, srcy); pipe_surface_unmap(src); pipe_surface_unmap(dst); diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 573fef9b83..a2bc330424 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -383,4 +383,13 @@ static INLINE int align(int value, int alignment) */ extern void _mesa_printf(const char *str, ...); + +/* util/p_util.c + */ +extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, + unsigned dst_x, unsigned dst_y, unsigned width, + unsigned height, const ubyte * src, + unsigned src_pitch, unsigned src_x, unsigned src_y); + + #endif diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index a580fb3e82..d5119654ed 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -72,45 +72,6 @@ softpipe_get_tex_surface(struct pipe_context *pipe, } -/** - * Copy 2D rect from one place to another. - * Position and sizes are in pixels. - */ -static void -copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte * src, - unsigned src_pitch, - unsigned src_x, - unsigned src_y) -{ - unsigned i; - - dst_pitch *= cpp; - src_pitch *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * src_pitch; - width *= cpp; - - if (width == dst_pitch && width == src_pitch) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; - } - } -} - - /* Upload data to a rectangular sub-region. Lots of choices how to do this: * * - memcpy by span to current destination @@ -125,10 +86,10 @@ sp_surface_data(struct pipe_context *pipe, const void *src, unsigned src_pitch, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); pipe_surface_unmap(dst); } @@ -145,14 +106,14 @@ sp_surface_copy(struct pipe_context *pipe, { assert( dst->cpp == src->cpp ); - copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - pipe_surface_map(src), - src->pitch, - srcx, srcy); + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe_surface_map(src), + src->pitch, + srcx, srcy); pipe_surface_unmap(src); pipe_surface_unmap(dst); diff --git a/src/mesa/pipe/util/p_util.c b/src/mesa/pipe/util/p_util.c new file mode 100644 index 0000000000..c4882b77d2 --- /dev/null +++ b/src/mesa/pipe/util/p_util.c @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Miscellaneous utility functions. + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" + + +/** + * Copy 2D rect from one place to another. + * Position and sizes are in pixels. + */ +void +pipe_copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + unsigned src_pitch, + unsigned src_x, + unsigned src_y) +{ + unsigned i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * src_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } +} diff --git a/src/mesa/sources b/src/mesa/sources index 90ce9ce370..d8760ba134 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -191,7 +191,8 @@ STATECACHE_SOURCES = \ pipe/cso_cache/cso_cache.c PIPEUTIL_SOURCES = \ - pipe/util/p_tile.c + pipe/util/p_tile.c \ + pipe/util/p_util.c STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ -- cgit v1.2.3 From fdc8636bdc65deb0d95a62a51c8d9bca05bc6bb8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 18 Jan 2008 12:45:27 -0700 Subject: use PROGRAM_CONSTANT instead of PROGRAM_STATE_VAR when generating immediates/literals --- src/mesa/main/ffvertex_prog.c | 2 +- src/mesa/main/texenvprogram.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 4a9a0cd975..04ece4cda8 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -436,7 +436,7 @@ static struct ureg register_const4f( struct tnl_program *p, idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4, &swizzle ); ASSERT(swizzle == SWIZZLE_NOOP); - return make_ureg(PROGRAM_STATE_VAR, idx); + return make_ureg(PROGRAM_CONSTANT, idx); } #define register_const1f(p, s0) register_const4f(p, s0, 0, 0, 1) diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index d866d10017..644b1f39c7 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -582,7 +582,7 @@ static struct ureg register_const4f( struct texenv_fragment_program *p, idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4, &swizzle ); ASSERT(swizzle == SWIZZLE_NOOP); - return make_ureg(PROGRAM_STATE_VAR, idx); + return make_ureg(PROGRAM_CONSTANT, idx); } #define register_scalar_const(p, s0) register_const4f(p, s0, s0, s0, s0) -- cgit v1.2.3 From 7cbfe8c51dfa8ef1f4671e817759edc7a79ce17a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 18 Jan 2008 12:45:55 -0700 Subject: set param type to PROGRAM_CONSTANT when parsing immediates --- src/mesa/shader/arbprogparse.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 9a5290d920..3f3c4c91aa 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -1825,12 +1825,14 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst, break; case PARAM_CONSTANT: + /* parsing something like {1.0, 2.0, 3.0, 4.0} */ parse_constant (inst, const_values, Program, use); idx = _mesa_add_named_constant(Program->Base.Parameters, (char *) param_var->name, const_values, 4); if (param_var->param_binding_begin == ~0U) param_var->param_binding_begin = idx; + param_var->param_binding_type = PROGRAM_CONSTANT; param_var->param_binding_length++; Program->Base.NumParameters++; break; -- cgit v1.2.3 From 06b019d16bc20d772a8aed2a68d1c5d37a402a81 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 18 Jan 2008 12:47:20 -0700 Subject: s/varible/variable/ --- src/mesa/shader/arbprogparse.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 3f3c4c91aa..d662e0b8b5 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -1695,7 +1695,7 @@ parse_attrib (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head if (found) { error_msg = (char *) _mesa_malloc (_mesa_strlen ((char *) attrib_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Varible Declaration: %s", + _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", attrib_var->name); program_error(ctx, Program->Position, error_msg); _mesa_free (error_msg); @@ -1880,7 +1880,7 @@ parse_param (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, if (found) { char *error_msg = (char *) _mesa_malloc (_mesa_strlen ((char *) param_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Varible Declaration: %s", + _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", param_var->name); program_error (ctx, Program->Position, error_msg); _mesa_free (error_msg); @@ -1980,7 +1980,7 @@ parse_temp (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, if (found) { char *error_msg = (char *) _mesa_malloc (_mesa_strlen ((char *) temp_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Varible Declaration: %s", + _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", temp_var->name); program_error(ctx, Program->Position, error_msg); _mesa_free (error_msg); @@ -2026,7 +2026,7 @@ parse_output (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head if (found) { char *error_msg = (char *) _mesa_malloc (_mesa_strlen ((char *) output_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Varible Declaration: %s", + _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", output_var->name); program_error (ctx, Program->Position, error_msg); _mesa_free (error_msg); @@ -2057,7 +2057,7 @@ parse_alias (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, if (found) { char *error_msg = (char *) _mesa_malloc (_mesa_strlen ((char *) temp_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Varible Declaration: %s", + _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", temp_var->name); program_error(ctx, Program->Position, error_msg); _mesa_free (error_msg); @@ -2100,7 +2100,7 @@ parse_address (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_hea if (found) { char *error_msg = (char *) _mesa_malloc (_mesa_strlen ((char *) temp_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Varible Declaration: %s", + _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", temp_var->name); program_error (ctx, Program->Position, error_msg); _mesa_free (error_msg); -- cgit v1.2.3 From a1f4a5e802ad62c88fca6834b9de1c83672230a6 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 19 Jan 2008 12:04:06 -0700 Subject: Cell: improve "finished copying batch buffer" signalling. When the SPU is done copying a batch buffer to local store, use an mfc_put() to write a "done" message back to the buffer status array in main memory. We were previously using a mailbox message for synchronization. --- src/mesa/pipe/cell/common.h | 7 ++++-- src/mesa/pipe/cell/ppu/cell_batch.c | 47 ++++++++++++++++++++++++----------- src/mesa/pipe/cell/ppu/cell_context.c | 23 ++++++++++++++--- src/mesa/pipe/cell/ppu/cell_context.h | 7 ++++++ src/mesa/pipe/cell/ppu/cell_spu.c | 1 + src/mesa/pipe/cell/spu/spu_main.c | 38 +++++++++++++++++++++++++--- src/mesa/pipe/cell/spu/spu_main.h | 2 ++ 7 files changed, 103 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 29d1cbedda..f05070d25a 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -62,10 +62,12 @@ #define CELL_CMD_STATE_DEPTH_STENCIL 7 -#define CELL_NUM_BATCH_BUFFERS 2 +#define CELL_NUM_BATCH_BUFFERS 3 #define CELL_BATCH_BUFFER_SIZE 1024 /**< 16KB would be the max */ -#define CELL_BATCH_FINISHED 0x1234 /**< mbox message */ +#define CELL_BUFFER_STATUS_FREE 10 +#define CELL_BUFFER_STATUS_USED 20 + /** @@ -122,6 +124,7 @@ struct cell_init_info unsigned num_spus; struct cell_command *cmd; ubyte *batch_buffers[CELL_NUM_BATCH_BUFFERS]; + uint *buffer_status; /**< points at cell_context->buffer_status */ } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index 45f62ac3ad..ab4553f16c 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -34,9 +34,9 @@ void cell_batch_flush(struct cell_context *cell) { - const uint batch = cell->cur_batch; + uint batch = cell->cur_batch; const uint size = cell->batch_buffer_size[batch]; - uint i, cmd_word; + uint spu, cmd_word; if (size == 0) return; @@ -48,25 +48,44 @@ cell_batch_flush(struct cell_context *cell) batch, &cell->batch_buffer[batch][0], size); */ + /* + * Build "BATCH" command and sent to all SPUs. + */ cmd_word = CELL_CMD_BATCH | (batch << 8) | (size << 16); - for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(cell_global.spe_contexts[i], cmd_word); + for (spu = 0; spu < cell->num_spus; spu++) { + assert(cell->buffer_status[spu][batch][0] == CELL_BUFFER_STATUS_USED); + send_mbox_message(cell_global.spe_contexts[spu], cmd_word); } - /* XXX wait for the DMX xfer to finish. - * Using mailboxes here is temporary. - * Ideally, we want to use a PPE-side DMA status check function... + /* When the SPUs are done copying the buffer into their locals stores + * they'll write a BUFFER_STATUS_FREE message into the buffer_status[] + * array indicating that the PPU can re-use the buffer. */ - for (i = 0; i < cell->num_spus; i++) { - uint k = wait_mbox_message(cell_global.spe_contexts[i]); - assert(k == CELL_BATCH_FINISHED); - } - /* next buffer */ - cell->cur_batch = (batch + 1) % CELL_NUM_BATCH_BUFFERS; - cell->batch_buffer_size[cell->cur_batch] = 0; /* empty */ + /* Find a buffer that's marked as free by all SPUs */ + while (1) { + uint num_free = 0; + + batch = (batch + 1) % CELL_NUM_BATCH_BUFFERS; + + for (spu = 0; spu < cell->num_spus; spu++) { + if (cell->buffer_status[spu][batch][0] == CELL_BUFFER_STATUS_FREE) + num_free++; + } + + if (num_free == cell->num_spus) { + /* found a free buffer, now mark status as used */ + for (spu = 0; spu < cell->num_spus; spu++) { + cell->buffer_status[spu][batch][0] = CELL_BUFFER_STATUS_USED; + } + break; + } + } + + cell->batch_buffer_size[batch] = 0; /* empty */ + cell->cur_batch = batch; } diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 6ba3b0d413..897c187d65 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -160,7 +160,7 @@ struct pipe_context * cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) { struct cell_context *cell; - uint i; + uint spu, buf; /* some fields need to be 16-byte aligned, so align the whole object */ cell = (struct cell_context*) align_malloc(sizeof(struct cell_context), 16); @@ -248,13 +248,30 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell_start_spus(cell); - for (i = 0; i < CELL_NUM_BATCH_BUFFERS; i++) { - cell->batch_buffer_size[i] = 0; + for (buf = 0; buf < CELL_NUM_BATCH_BUFFERS; buf++) { + cell->batch_buffer_size[buf] = 0; + + /* init batch buffer status values, + * mark 0th buffer as used, rest as free. + */ + for (spu = 0; spu < cell->num_spus; spu++) { + if (buf == 0) + cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED; + else + cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_FREE; + } } + #if 0 test_spus(cell); #endif return &cell->pipe; } + + +#if 0 +/** [4] to ensure 16-byte alignment for each status word */ +uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BATCH_BUFFERS][4] ALIGN16_ATTRIB; +#endif diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 43e32a8abb..08e448f14f 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -38,6 +38,9 @@ #include "pipe/cell/common.h" +#define CELL_MAX_SPUS 6 + + struct cell_vbuf_render; struct cell_vertex_shader_state @@ -103,10 +106,14 @@ struct cell_context ubyte batch_buffer[CELL_NUM_BATCH_BUFFERS][CELL_BATCH_BUFFER_SIZE] ALIGN16_ATTRIB; int cur_batch; /**< which batch buffer is being filled */ + /** [4] to ensure 16-byte alignment for each status word */ + uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BATCH_BUFFERS][4] ALIGN16_ATTRIB; + }; + static INLINE struct cell_context * cell_context(struct pipe_context *pipe) { diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 77ddd9ccbf..a21ab9ff3a 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -114,6 +114,7 @@ cell_start_spus(struct cell_context *cell) for (j = 0; j < CELL_NUM_BATCH_BUFFERS; j++) { cell_global.inits[i].batch_buffers[j] = cell->batch_buffer[j]; } + cell_global.inits[i].buffer_status = &cell->buffer_status[0][0][0]; cell_global.spe_contexts[i] = spe_context_create(0, NULL); if (!cell_global.spe_contexts[i]) { diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 2727b03756..bff098f06e 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -395,6 +395,38 @@ cmd_finish(void) } +/** + * Tell the PPU that this SPU has finished copying a batch buffer to + * local store and that it may be reused by the PPU. + * This is done by writting a 16-byte batch-buffer-status block back into + * main memory (in cell_contex->buffer_status[]). + */ +static void +release_batch_buffer(uint buffer) +{ + /* Evidently, using less than a 16-byte status doesn't work reliably */ + static const uint status[4] ALIGN16_ATTRIB + = {CELL_BUFFER_STATUS_FREE, 0, 0, 0}; + + const uint index = 4 * (spu.init.id * CELL_NUM_BATCH_BUFFERS + buffer); + uint *dst = spu.init.buffer_status + index; + + ASSERT(buffer < CELL_NUM_BATCH_BUFFERS); + + /* + printf("SPU %u: Set batch status buf=%u, index %u, at %p to FREE\n", + spu.init.id, buffer, index, dst); + */ + + mfc_put((void *) &status, /* src in local memory */ + (unsigned int) dst, /* dst in main memory */ + sizeof(status), /* size */ + TAG_MISC, /* tag is unimportant */ + 0, /* tid */ + 0 /* rid */); +} + + /** * Execute a batch of commands * The opcode param encodes the location of the buffer and its size. @@ -429,9 +461,9 @@ cmd_batch(uint opcode) 0 /* rid */); wait_on_mask(1 << TAG_BATCH_BUFFER); - /* send mbox message to indicate DMA completed */ - /* XXX temporary */ - spu_write_out_mbox(CELL_BATCH_FINISHED); + /* Tell PPU we're done copying the buffer to local store */ + release_batch_buffer(buf); + for (pos = 0; pos < usize; /* no incr */) { switch (buffer[pos]) { diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 75fb5b388b..f39f375d24 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -76,6 +76,8 @@ extern struct spu_global spu; #define TAG_WRITE_TILE_Z 15 #define TAG_INDEX_BUFFER 16 #define TAG_BATCH_BUFFER 17 +#define TAG_MISC 18 + /** The standard assert macro doesn't seem to work on SPUs */ #define ASSERT(x) \ -- cgit v1.2.3 From 45b5d3b1fc996e0f460ea163e4f4b3e750e8914f Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 20 Jan 2008 14:58:29 -0700 Subject: enable call to cell_emit_state() --- src/mesa/pipe/cell/ppu/cell_state_derived.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c index c654990325..dc2879b915 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -219,7 +219,7 @@ void cell_update_derived( struct cell_context *cell ) compute_cliprect(cell); #endif - //cell_emit_state(cell); + cell_emit_state(cell); cell->dirty = 0; } -- cgit v1.2.3 From 661be1ae7b1cd5837d8e7224a5ebe1b8d428137e Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 20 Jan 2008 15:00:18 -0700 Subject: Cell: use depth/stencil state to enable ztest Move z-test code into do_depth_test(). Add ZSIZE symbol to support 2 or 4-byte Z values. --- src/mesa/pipe/cell/spu/spu_main.c | 22 ++++--- src/mesa/pipe/cell/spu/spu_main.h | 2 + src/mesa/pipe/cell/spu/spu_tile.c | 17 ++++-- src/mesa/pipe/cell/spu/spu_tile.h | 12 +++- src/mesa/pipe/cell/spu/spu_tri.c | 122 ++++++++++++++++++++++---------------- 5 files changed, 113 insertions(+), 62 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index bff098f06e..30c39aab68 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -239,8 +239,8 @@ cmd_render(const struct cell_command_render *render) /* how much vertex data */ vertex_bytes = render->num_verts * vertex_size; index_bytes = render->num_indexes * sizeof(ushort); - if (index_bytes < 8) - index_bytes = 8; + if (index_bytes < 16) + index_bytes = 16; else index_bytes = (index_bytes + 15) & ~0xf; /* multiple of 16 */ @@ -297,7 +297,7 @@ cmd_render(const struct cell_command_render *render) /* Start fetching color/z tiles. We'll wait for completion when * we need read/write to them later in triangle rasterization. */ - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (spu.depth_stencil.depth.enabled) { if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { get_tile(tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); } @@ -325,7 +325,7 @@ cmd_render(const struct cell_command_render *render) put_tile(tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); tile_status[ty][tx] = TILE_STATUS_DEFINED; } - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (spu.depth_stencil.depth.enabled) { if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { put_tile(tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); tile_status_z[ty][tx] = TILE_STATUS_DEFINED; @@ -334,7 +334,7 @@ cmd_render(const struct cell_command_render *render) /* XXX move these... */ wait_on_mask(1 << TAG_WRITE_TILE_COLOR); - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (spu.depth_stencil.depth.enabled) { wait_on_mask(1 << TAG_WRITE_TILE_Z); } } @@ -365,6 +365,14 @@ cmd_framebuffer(const struct cell_command_framebuffer *cmd) spu.fb.height = cmd->height; spu.fb.width_tiles = (spu.fb.width + TILE_SIZE - 1) / TILE_SIZE; spu.fb.height_tiles = (spu.fb.height + TILE_SIZE - 1) / TILE_SIZE; + + if (cmd->depth_format == PIPE_FORMAT_Z16_UNORM) { + ASSERT(ZSIZE == 2); + } + else if (cmd->depth_format == PIPE_FORMAT_Z32_UNORM) { + ASSERT(ZSIZE == 4); + } + } @@ -375,9 +383,9 @@ cmd_state_depth_stencil(const struct pipe_depth_stencil_alpha_state *state) printf("SPU %u: DEPTH_STENCIL: ztest %d\n", spu.init.id, state->depth.enabled); - /* + memcpy(&spu.depth_stencil, state, sizeof(*state)); - */ + } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index f39f375d24..cd2afbe9bf 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -32,6 +32,8 @@ #include "pipe/cell/common.h" #include "pipe/p_state.h" +/** XXX temp bytes/z value */ +#define ZSIZE 2 struct spu_framebuffer { void *color_start; /**< addr of color surface in main memory */ diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c index 1505cb322b..13fc5e345f 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.c +++ b/src/mesa/pipe/cell/spu/spu_tile.c @@ -32,7 +32,11 @@ uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +#if ZSIZE == 2 ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +#else +uint ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +#endif ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; @@ -43,7 +47,7 @@ void get_tile(uint tx, uint ty, uint *tile, int tag, int zBuf) { const uint offset = ty * spu.fb.width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? ZSIZE : 4); const ubyte *src = zBuf ? spu.fb.depth_start : spu.fb.color_start; src += offset * bytesPerTile; @@ -68,7 +72,7 @@ void put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf) { const uint offset = ty * spu.fb.width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? 2 : 4); + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? ZSIZE : 4); ubyte *dst = zBuf ? spu.fb.depth_start : spu.fb.color_start; dst += offset * bytesPerTile; @@ -81,7 +85,6 @@ put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf) spu.init.id, tile, (unsigned int) dst, bytesPerTile); */ - mfc_put((void *) tile, /* src in local memory */ (unsigned int) dst, /* dst in main memory */ bytesPerTile, @@ -103,7 +106,13 @@ clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value) } void -clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value) +clear_tile_z( +#if ZSIZE == 2 + ushort tile[TILE_SIZE][TILE_SIZE], +#else + uint tile[TILE_SIZE][TILE_SIZE], +#endif + uint value) { uint i, j; for (i = 0; i < TILE_SIZE; i++) { diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h index 637923764b..7c288a1f91 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -40,7 +40,11 @@ extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +#if ZSIZE == 2 extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +#else +extern uint ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +#endif #define TILE_STATUS_CLEAR 1 @@ -61,7 +65,13 @@ void clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value); void -clear_tile_z(ushort tile[TILE_SIZE][TILE_SIZE], uint value); +clear_tile_z( +#if ZSIZE == 2 + ushort tile[TILE_SIZE][TILE_SIZE], +#else + uint tile[TILE_SIZE][TILE_SIZE], +#endif + uint value); #endif /* SPU_TILE_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index ddd5e662d2..0b8533092e 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -217,8 +217,8 @@ static INLINE void eval_z( struct setup_stage *setup, float x, float y, float result[4]) { - uint slot = 0; - uint i = 2; + const uint slot = 0; + const uint i = 2; const float *dadx = setup->coef[slot].dadx; const float *dady = setup->coef[slot].dady; @@ -252,6 +252,72 @@ pack_color(const float color[4]) } +static uint +do_depth_test(struct setup_stage *setup, int x, int y, unsigned mask) +{ + int ix = x - setup->cliprect_minx; + int iy = y - setup->cliprect_miny; + float zvals[4]; + float zscale = 65535.0; + + if (ZSIZE == 2) { + ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM); + } + else { + ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); + } + ASSERT(sizeof(ztile[0][0]) == ZSIZE); + + + eval_z(setup, (float) x, (float) y, zvals); + + if (tile_status_z[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { + /* now, _really_ clear the tile */ + clear_tile_z(ztile, spu.fb.depth_clear_value); + } + else { + /* make sure we've got the tile from main mem */ + wait_on_mask(1 << TAG_READ_TILE_Z); + } + tile_status_z[setup->ty][setup->tx] = TILE_STATUS_DIRTY; + + + if (mask & MASK_TOP_LEFT) { + uint z = (uint) (zvals[0] * zscale); + if (z < ztile[iy][ix]) + ztile[iy][ix] = z; + else + mask &= ~MASK_TOP_LEFT; + } + + if (mask & MASK_TOP_RIGHT) { + uint z = (uint) (zvals[1] * zscale); + if (z < ztile[iy][ix+1]) + ztile[iy][ix+1] = z; + else + mask &= ~MASK_TOP_RIGHT; + } + + if (mask & MASK_BOTTOM_LEFT) { + uint z = (uint) (zvals[2] * zscale); + if (z < ztile[iy+1][ix]) + ztile[iy+1][ix] = z; + else + mask &= ~MASK_BOTTOM_LEFT; + } + + if (mask & MASK_BOTTOM_RIGHT) { + uint z = (uint) (zvals[3] * zscale); + if (z < ztile[iy+1][ix+1]) + ztile[iy+1][ix+1] = z; + else + mask &= ~MASK_BOTTOM_RIGHT; + } + + return mask; +} + + /** * Emit a quad (pass to next stage). No clipping is done. */ @@ -266,58 +332,14 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) sp->quad.first->run(sp->quad.first, &setup->quad); #else /* Cell: "write" quad fragments to the tile by setting prim color */ - int ix = x - setup->cliprect_minx; - int iy = y - setup->cliprect_miny; + const int ix = x - setup->cliprect_minx; + const int iy = y - setup->cliprect_miny; float colors[4][4]; - uint z; eval_coeff(setup, 1, (float) x, (float) y, colors); - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - float zvals[4]; - eval_z(setup, (float) x, (float) y, zvals); - - if (tile_status_z[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { - /* now, _really_ clear the tile */ - clear_tile_z(ztile, spu.fb.depth_clear_value); - } - else { - /* make sure we've got the tile from main mem */ - wait_on_mask(1 << TAG_READ_TILE_Z); - } - tile_status_z[setup->ty][setup->tx] = TILE_STATUS_DIRTY; - - if (mask & MASK_TOP_LEFT) { - z = (uint) (zvals[0] * 65535.0); - if (z < ztile[iy][ix]) - ztile[iy][ix] = z; - else - mask &= ~MASK_TOP_LEFT; - } - - if (mask & MASK_TOP_RIGHT) { - z = (uint) (zvals[1] * 65535.0); - if (z < ztile[iy][ix+1]) - ztile[iy][ix+1] = z; - else - mask &= ~MASK_TOP_RIGHT; - } - - if (mask & MASK_BOTTOM_LEFT) { - z = (uint) (zvals[2] * 65535.0); - if (z < ztile[iy+1][ix]) - ztile[iy+1][ix] = z; - else - mask &= ~MASK_BOTTOM_LEFT; - } - - if (mask & MASK_BOTTOM_RIGHT) { - z = (uint) (zvals[3] * 65535.0); - if (z < ztile[iy+1][ix+1]) - ztile[iy+1][ix+1] = z; - else - mask &= ~MASK_BOTTOM_RIGHT; - } + if (spu.depth_stencil.depth.enabled) { + mask &= do_depth_test(setup, x, y, mask); } if (mask) { -- cgit v1.2.3 From f0be276c2e84716856ae87b4b0f0411700ed5be4 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 20 Jan 2008 17:27:05 -0700 Subject: Cell: clean-up/re-org tile code Also, support 16 or 32-bit Z buffer at runtime. --- src/mesa/pipe/cell/spu/spu_main.c | 41 +++++++-------- src/mesa/pipe/cell/spu/spu_main.h | 24 +++++++-- src/mesa/pipe/cell/spu/spu_tile.c | 39 ++------------ src/mesa/pipe/cell/spu/spu_tile.h | 44 ++++++++++------ src/mesa/pipe/cell/spu/spu_tri.c | 104 ++++++++++++++++++++++++-------------- 5 files changed, 136 insertions(+), 116 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 30c39aab68..8e9352deff 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -78,12 +78,10 @@ static void really_clear_tiles(uint surfaceIndex) { const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; - uint i, j; + uint i; if (surfaceIndex == 0) { - for (i = 0; i < TILE_SIZE; i++) - for (j = 0; j < TILE_SIZE; j++) - ctile[i][j] = spu.fb.color_clear_value; /*0xff00ff;*/ + clear_c_tile(ctile); for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { uint tx = i % spu.fb.width_tiles; @@ -94,9 +92,7 @@ really_clear_tiles(uint surfaceIndex) } } else { - for (i = 0; i < TILE_SIZE; i++) - for (j = 0; j < TILE_SIZE; j++) - ztile[i][j] = spu.fb.depth_clear_value; + clear_z_tile(&ztile); for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { uint tx = i % spu.fb.width_tiles; @@ -116,7 +112,7 @@ static void cmd_clear_surface(const struct cell_command_clear_surface *clear) { const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; - uint i, j; + uint i; if (Debug) printf("SPU %u: CLEAR SURF %u to 0x%08x\n", spu.init.id, @@ -137,14 +133,12 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) #endif if (clear->surface == 0) { - for (i = 0; i < TILE_SIZE; i++) - for (j = 0; j < TILE_SIZE; j++) - ctile[i][j] = clear->value; + spu.fb.color_clear_value = clear->value; + clear_c_tile(ctile); } else { - for (i = 0; i < TILE_SIZE; i++) - for (j = 0; j < TILE_SIZE; j++) - ztile[i][j] = clear->value; + spu.fb.depth_clear_value = clear->value; + clear_z_tile(&ztile); } /* @@ -158,7 +152,7 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) if (clear->surface == 0) put_tile(tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); else - put_tile(tx, ty, (uint *) ztile, TAG_SURFACE_CLEAR, 1); + put_tile(tx, ty, (uint *) ztile.t32, TAG_SURFACE_CLEAR, 1); /* XXX we don't want this here, but it fixes bad tile results */ } @@ -299,7 +293,7 @@ cmd_render(const struct cell_command_render *render) */ if (spu.depth_stencil.depth.enabled) { if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(tx, ty, (uint *) ztile, TAG_READ_TILE_Z, 1); + get_tile(tx, ty, (uint *) ztile.t32, TAG_READ_TILE_Z, 1); } } @@ -327,7 +321,7 @@ cmd_render(const struct cell_command_render *render) } if (spu.depth_stencil.depth.enabled) { if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(tx, ty, (uint *) ztile, TAG_WRITE_TILE_Z, 1); + put_tile(tx, ty, (uint *) ztile.t32, TAG_WRITE_TILE_Z, 1); tile_status_z[ty][tx] = TILE_STATUS_DEFINED; } } @@ -366,13 +360,12 @@ cmd_framebuffer(const struct cell_command_framebuffer *cmd) spu.fb.width_tiles = (spu.fb.width + TILE_SIZE - 1) / TILE_SIZE; spu.fb.height_tiles = (spu.fb.height + TILE_SIZE - 1) / TILE_SIZE; - if (cmd->depth_format == PIPE_FORMAT_Z16_UNORM) { - ASSERT(ZSIZE == 2); - } - else if (cmd->depth_format == PIPE_FORMAT_Z32_UNORM) { - ASSERT(ZSIZE == 4); - } - + if (spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM) + spu.fb.zsize = 4; + else if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) + spu.fb.zsize = 2; + else + spu.fb.zsize = 0; } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index cd2afbe9bf..2aa7015ae3 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -32,8 +32,6 @@ #include "pipe/cell/common.h" #include "pipe/p_state.h" -/** XXX temp bytes/z value */ -#define ZSIZE 2 struct spu_framebuffer { void *color_start; /**< addr of color surface in main memory */ @@ -45,6 +43,8 @@ struct spu_framebuffer { uint color_clear_value; uint depth_clear_value; + + uint zsize; /**< 0, 2 or 4 bytes per Z */ } ALIGN16_ATTRIB; @@ -90,8 +90,26 @@ extern struct spu_global spu; } -void +extern void wait_on_mask(unsigned tag); +static INLINE void +memset16(ushort *d, ushort value, uint count) +{ + uint i; + for (i = 0; i < count; i++) + d[i] = value; +} + + +static INLINE void +memset32(uint *d, uint value, uint count) +{ + uint i; + for (i = 0; i < count; i++) + d[i] = value; +} + + #endif /* SPU_MAIN_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c index 13fc5e345f..9895360f5f 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.c +++ b/src/mesa/pipe/cell/spu/spu_tile.c @@ -32,11 +32,7 @@ uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -#if ZSIZE == 2 -ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -#else -uint ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -#endif +tile_t ztile ALIGN16_ATTRIB; ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; @@ -47,7 +43,7 @@ void get_tile(uint tx, uint ty, uint *tile, int tag, int zBuf) { const uint offset = ty * spu.fb.width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? ZSIZE : 4); + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4); const ubyte *src = zBuf ? spu.fb.depth_start : spu.fb.color_start; src += offset * bytesPerTile; @@ -72,7 +68,7 @@ void put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf) { const uint offset = ty * spu.fb.width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? ZSIZE : 4); + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4); ubyte *dst = zBuf ? spu.fb.depth_start : spu.fb.color_start; dst += offset * bytesPerTile; @@ -93,32 +89,3 @@ put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf) 0 /* rid */); } - -void -clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value) -{ - uint i, j; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile[i][j] = value; - } - } -} - -void -clear_tile_z( -#if ZSIZE == 2 - ushort tile[TILE_SIZE][TILE_SIZE], -#else - uint tile[TILE_SIZE][TILE_SIZE], -#endif - uint value) -{ - uint i, j; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile[i][j] = value; - } - } -} - diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h index 7c288a1f91..4c8db581f3 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -39,12 +39,14 @@ #define MAX_HEIGHT 1024 +typedef union { + ushort t16[TILE_SIZE][TILE_SIZE]; + uint t32[TILE_SIZE][TILE_SIZE]; +} tile_t; + + extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -#if ZSIZE == 2 -extern ushort ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -#else -extern uint ztile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; -#endif +extern tile_t ztile ALIGN16_ATTRIB; #define TILE_STATUS_CLEAR 1 @@ -61,17 +63,29 @@ get_tile(uint tx, uint ty, uint *tile, int tag, int zBuf); void put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf); -void -clear_tile(uint tile[TILE_SIZE][TILE_SIZE], uint value); -void -clear_tile_z( -#if ZSIZE == 2 - ushort tile[TILE_SIZE][TILE_SIZE], -#else - uint tile[TILE_SIZE][TILE_SIZE], -#endif - uint value); + +static INLINE void +clear_c_tile(uint tile[TILE_SIZE][TILE_SIZE]) +{ + memset32((uint*) tile, spu.fb.color_clear_value, TILE_SIZE * TILE_SIZE); +} + + +static INLINE void +clear_z_tile(tile_t *ztile) +{ + if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + memset16((ushort*) ztile->t16, + spu.fb.depth_clear_value, + TILE_SIZE * TILE_SIZE); + } + else { + memset32((uint*) ztile->t32, + spu.fb.depth_clear_value, + TILE_SIZE * TILE_SIZE); + } +} #endif /* SPU_TILE_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 0b8533092e..6de07614fb 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -258,22 +258,12 @@ do_depth_test(struct setup_stage *setup, int x, int y, unsigned mask) int ix = x - setup->cliprect_minx; int iy = y - setup->cliprect_miny; float zvals[4]; - float zscale = 65535.0; - - if (ZSIZE == 2) { - ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM); - } - else { - ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); - } - ASSERT(sizeof(ztile[0][0]) == ZSIZE); - eval_z(setup, (float) x, (float) y, zvals); if (tile_status_z[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ - clear_tile_z(ztile, spu.fb.depth_clear_value); + clear_z_tile(&ztile); } else { /* make sure we've got the tile from main mem */ @@ -282,36 +272,74 @@ do_depth_test(struct setup_stage *setup, int x, int y, unsigned mask) tile_status_z[setup->ty][setup->tx] = TILE_STATUS_DIRTY; - if (mask & MASK_TOP_LEFT) { - uint z = (uint) (zvals[0] * zscale); - if (z < ztile[iy][ix]) - ztile[iy][ix] = z; - else - mask &= ~MASK_TOP_LEFT; - } + if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + const float zscale = 65535.0; + if (mask & MASK_TOP_LEFT) { + uint z = (uint) (zvals[0] * zscale); + if (z < ztile.t16[iy][ix]) + ztile.t16[iy][ix] = z; + else + mask &= ~MASK_TOP_LEFT; + } - if (mask & MASK_TOP_RIGHT) { - uint z = (uint) (zvals[1] * zscale); - if (z < ztile[iy][ix+1]) - ztile[iy][ix+1] = z; - else - mask &= ~MASK_TOP_RIGHT; - } + if (mask & MASK_TOP_RIGHT) { + uint z = (uint) (zvals[1] * zscale); + if (z < ztile.t16[iy][ix+1]) + ztile.t16[iy][ix+1] = z; + else + mask &= ~MASK_TOP_RIGHT; + } - if (mask & MASK_BOTTOM_LEFT) { - uint z = (uint) (zvals[2] * zscale); - if (z < ztile[iy+1][ix]) - ztile[iy+1][ix] = z; - else - mask &= ~MASK_BOTTOM_LEFT; + if (mask & MASK_BOTTOM_LEFT) { + uint z = (uint) (zvals[2] * zscale); + if (z < ztile.t16[iy+1][ix]) + ztile.t16[iy+1][ix] = z; + else + mask &= ~MASK_BOTTOM_LEFT; + } + + if (mask & MASK_BOTTOM_RIGHT) { + uint z = (uint) (zvals[3] * zscale); + if (z < ztile.t16[iy+1][ix+1]) + ztile.t16[iy+1][ix+1] = z; + else + mask &= ~MASK_BOTTOM_RIGHT; + } } + else { + const float zscale = (float) 0xffffffff; + ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); + if (mask & MASK_TOP_LEFT) { + uint z = (uint) (zvals[0] * zscale); + if (z < ztile.t32[iy][ix]) + ztile.t32[iy][ix] = z; + else + mask &= ~MASK_TOP_LEFT; + } - if (mask & MASK_BOTTOM_RIGHT) { - uint z = (uint) (zvals[3] * zscale); - if (z < ztile[iy+1][ix+1]) - ztile[iy+1][ix+1] = z; - else - mask &= ~MASK_BOTTOM_RIGHT; + if (mask & MASK_TOP_RIGHT) { + uint z = (uint) (zvals[1] * zscale); + if (z < ztile.t32[iy][ix+1]) + ztile.t32[iy][ix+1] = z; + else + mask &= ~MASK_TOP_RIGHT; + } + + if (mask & MASK_BOTTOM_LEFT) { + uint z = (uint) (zvals[2] * zscale); + if (z < ztile.t32[iy+1][ix]) + ztile.t32[iy+1][ix] = z; + else + mask &= ~MASK_BOTTOM_LEFT; + } + + if (mask & MASK_BOTTOM_RIGHT) { + uint z = (uint) (zvals[3] * zscale); + if (z < ztile.t32[iy+1][ix+1]) + ztile.t32[iy+1][ix+1] = z; + else + mask &= ~MASK_BOTTOM_RIGHT; + } } return mask; @@ -345,7 +373,7 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) if (mask) { if (tile_status[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ - clear_tile(ctile, spu.fb.color_clear_value); + clear_c_tile(ctile); } else { /* make sure we've got the tile from main mem */ -- cgit v1.2.3 From f4b89be70111793a6b5eb511e1c92be72bb6b3d9 Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 20 Jan 2008 17:39:07 -0700 Subject: Cell: use tile_t for color tile --- src/mesa/pipe/cell/spu/spu_main.c | 20 ++++++++++---------- src/mesa/pipe/cell/spu/spu_tile.c | 10 +++++----- src/mesa/pipe/cell/spu/spu_tile.h | 12 +++++++----- src/mesa/pipe/cell/spu/spu_tri.c | 10 +++++----- 4 files changed, 27 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 8e9352deff..2b32c26854 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -81,13 +81,13 @@ really_clear_tiles(uint surfaceIndex) uint i; if (surfaceIndex == 0) { - clear_c_tile(ctile); + clear_c_tile(&ctile); for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { uint tx = i % spu.fb.width_tiles; uint ty = i / spu.fb.width_tiles; if (tile_status[ty][tx] == TILE_STATUS_CLEAR) { - put_tile(tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); + put_tile(tx, ty, &ctile, TAG_SURFACE_CLEAR, 0); } } } @@ -98,7 +98,7 @@ really_clear_tiles(uint surfaceIndex) uint tx = i % spu.fb.width_tiles; uint ty = i / spu.fb.width_tiles; if (tile_status_z[ty][tx] == TILE_STATUS_CLEAR) - put_tile(tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 1); + put_tile(tx, ty, &ctile, TAG_SURFACE_CLEAR, 1); } } @@ -134,7 +134,7 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) if (clear->surface == 0) { spu.fb.color_clear_value = clear->value; - clear_c_tile(ctile); + clear_c_tile(&ctile); } else { spu.fb.depth_clear_value = clear->value; @@ -150,9 +150,9 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) uint tx = i % spu.fb.width_tiles; uint ty = i / spu.fb.width_tiles; if (clear->surface == 0) - put_tile(tx, ty, (uint *) ctile, TAG_SURFACE_CLEAR, 0); + put_tile(tx, ty, &ctile, TAG_SURFACE_CLEAR, 0); else - put_tile(tx, ty, (uint *) ztile.t32, TAG_SURFACE_CLEAR, 1); + put_tile(tx, ty, &ztile, TAG_SURFACE_CLEAR, 1); /* XXX we don't want this here, but it fixes bad tile results */ } @@ -293,12 +293,12 @@ cmd_render(const struct cell_command_render *render) */ if (spu.depth_stencil.depth.enabled) { if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(tx, ty, (uint *) ztile.t32, TAG_READ_TILE_Z, 1); + get_tile(tx, ty, &ztile, TAG_READ_TILE_Z, 1); } } if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(tx, ty, (uint *) ctile, TAG_READ_TILE_COLOR, 0); + get_tile(tx, ty, &ctile, TAG_READ_TILE_COLOR, 0); } ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); @@ -316,12 +316,12 @@ cmd_render(const struct cell_command_render *render) /* write color/z tiles back to main framebuffer, if dirtied */ if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(tx, ty, (uint *) ctile, TAG_WRITE_TILE_COLOR, 0); + put_tile(tx, ty, &ctile, TAG_WRITE_TILE_COLOR, 0); tile_status[ty][tx] = TILE_STATUS_DEFINED; } if (spu.depth_stencil.depth.enabled) { if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(tx, ty, (uint *) ztile.t32, TAG_WRITE_TILE_Z, 1); + put_tile(tx, ty, &ztile, TAG_WRITE_TILE_Z, 1); tile_status_z[ty][tx] = TILE_STATUS_DEFINED; } } diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c index 9895360f5f..ca1352f9f8 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.c +++ b/src/mesa/pipe/cell/spu/spu_tile.c @@ -31,7 +31,7 @@ -uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +tile_t ctile ALIGN16_ATTRIB; tile_t ztile ALIGN16_ATTRIB; ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; @@ -40,7 +40,7 @@ ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; void -get_tile(uint tx, uint ty, uint *tile, int tag, int zBuf) +get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf) { const uint offset = ty * spu.fb.width_tiles + tx; const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4); @@ -55,7 +55,7 @@ get_tile(uint tx, uint ty, uint *tile, int tag, int zBuf) printf("get_tile: dest: %p src: 0x%x size: %d\n", tile, (unsigned int) src, bytesPerTile); */ - mfc_get(tile, /* dest in local memory */ + mfc_get(tile->t32, /* dest in local memory */ (unsigned int) src, /* src in main memory */ bytesPerTile, tag, @@ -65,7 +65,7 @@ get_tile(uint tx, uint ty, uint *tile, int tag, int zBuf) void -put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf) +put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf) { const uint offset = ty * spu.fb.width_tiles + tx; const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4); @@ -81,7 +81,7 @@ put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf) spu.init.id, tile, (unsigned int) dst, bytesPerTile); */ - mfc_put((void *) tile, /* src in local memory */ + mfc_put((void *) tile->t32, /* src in local memory */ (unsigned int) dst, /* dst in main memory */ bytesPerTile, tag, diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h index 4c8db581f3..f83dc009c2 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -45,7 +45,7 @@ typedef union { } tile_t; -extern uint ctile[TILE_SIZE][TILE_SIZE] ALIGN16_ATTRIB; +extern tile_t ctile ALIGN16_ATTRIB; extern tile_t ztile ALIGN16_ATTRIB; @@ -58,17 +58,19 @@ extern ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_AT void -get_tile(uint tx, uint ty, uint *tile, int tag, int zBuf); +get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf); void -put_tile(uint tx, uint ty, const uint *tile, int tag, int zBuf); +put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf); static INLINE void -clear_c_tile(uint tile[TILE_SIZE][TILE_SIZE]) +clear_c_tile(tile_t *ctile) { - memset32((uint*) tile, spu.fb.color_clear_value, TILE_SIZE * TILE_SIZE); + memset32((uint*) ctile->t32, + spu.fb.color_clear_value, + TILE_SIZE * TILE_SIZE); } diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 6de07614fb..1d73c5171c 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -373,7 +373,7 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) if (mask) { if (tile_status[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ - clear_c_tile(ctile); + clear_c_tile(&ctile); } else { /* make sure we've got the tile from main mem */ @@ -382,13 +382,13 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) tile_status[setup->ty][setup->tx] = TILE_STATUS_DIRTY; if (mask & MASK_TOP_LEFT) - ctile[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); + ctile.t32[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); if (mask & MASK_TOP_RIGHT) - ctile[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); + ctile.t32[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); if (mask & MASK_BOTTOM_LEFT) - ctile[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); + ctile.t32[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); if (mask & MASK_BOTTOM_RIGHT) - ctile[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); + ctile.t32[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); } #endif } -- cgit v1.2.3 From 382b86e90f69fa0493fae3c7e5c9cd482984af8f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 11:16:22 -0700 Subject: gallium: add a src_index[] array to draw's vertex_info for mapping post-xform vertex attribs to hw vertex attribs --- src/mesa/pipe/cell/ppu/cell_state_derived.c | 19 +++++++++--------- src/mesa/pipe/draw/draw_vbuf.c | 29 ++++++++++++++------------- src/mesa/pipe/draw/draw_vertex.h | 7 ++++++- src/mesa/pipe/i915simple/i915_state_derived.c | 15 +++++++------- src/mesa/pipe/softpipe/sp_state_derived.c | 17 ++++++++-------- 5 files changed, 48 insertions(+), 39 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c index dc2879b915..1e31c11ecd 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -54,6 +54,7 @@ static void calculate_vertex_layout( struct cell_context *cell ) = cell->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &cell->vertex_info; uint front0; + uint src = 0; memset(vinfo, 0, sizeof(*vinfo)); @@ -68,10 +69,10 @@ static void calculate_vertex_layout( struct cell_context *cell ) #endif /* always emit vertex pos */ - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR, src++); #if 1 - front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); #endif #if 0 @@ -94,11 +95,11 @@ static void calculate_vertex_layout( struct cell_context *cell ) case TGSI_SEMANTIC_COLOR: if (vs->output_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); } else { assert(vs->output_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); } break; @@ -113,7 +114,7 @@ static void calculate_vertex_layout( struct cell_context *cell ) break; case TGSI_SEMANTIC_FOG: - draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE, src++); break; case TGSI_SEMANTIC_PSIZE: @@ -125,7 +126,7 @@ static void calculate_vertex_layout( struct cell_context *cell ) case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src++); break; default: @@ -140,14 +141,14 @@ static void calculate_vertex_layout( struct cell_context *cell ) * up 1:1 with the fragment shader inputs. */ if (emitBack0) { - back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); } if (emitBack1) { - back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); } if (emitPsize) { cell->psize_slot - = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT); + = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT, src++); } /* If the attributes have changed, tell the draw module about diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 6cda122c3a..7683e3381c 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -139,37 +139,38 @@ emit_vertex( struct vbuf_stage *vbuf, vertex->vertex_id = vbuf->nr_vertices++; for (i = 0; i < vinfo->num_attribs; i++) { + uint j = vinfo->src_index[i]; switch (vinfo->format[i]) { case FORMAT_OMIT: /* no-op */ break; case FORMAT_1F: - *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); count++; break; case FORMAT_2F: - *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); count += 2; break; case FORMAT_3F: - *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][2]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); count += 3; break; case FORMAT_4F: - *vbuf->vertex_ptr++ = fui(vertex->data[i][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][2]); - *vbuf->vertex_ptr++ = fui(vertex->data[i][3]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][3]); count += 4; break; case FORMAT_4UB: - *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[i][2] ), - float_to_ubyte( vertex->data[i][1] ), - float_to_ubyte( vertex->data[i][0] ), - float_to_ubyte( vertex->data[i][3] )); + *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), + float_to_ubyte( vertex->data[j][1] ), + float_to_ubyte( vertex->data[j][0] ), + float_to_ubyte( vertex->data[j][3] )); count += 1; break; default: diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 8bb328affa..ab0425e106 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -70,6 +70,7 @@ struct vertex_info uint hwfmt[4]; /**< hardware format info for this format */ enum interp_mode interp_mode[PIPE_MAX_SHADER_OUTPUTS]; enum attrib_format format[PIPE_MAX_SHADER_OUTPUTS]; /**< FORMAT_x */ + uint src_index[PIPE_MAX_SHADER_OUTPUTS]; uint size; /**< total vertex size in dwords */ }; @@ -77,16 +78,20 @@ struct vertex_info /** * Add another attribute to the given vertex_info object. + * \param src_index indicates which post-transformed vertex attrib slot + * corresponds to this attribute. * \return slot in which the attribute was added */ static INLINE uint draw_emit_vertex_attr(struct vertex_info *vinfo, - enum attrib_format format, enum interp_mode interp) + enum attrib_format format, enum interp_mode interp, + uint src_index) { const uint n = vinfo->num_attribs; assert(n < PIPE_MAX_SHADER_OUTPUTS); vinfo->format[n] = format; vinfo->interp_mode[n] = interp; + vinfo->src_index[n] = src_index; vinfo->num_attribs++; return n; } diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index be73769cf2..466c704d87 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -50,12 +50,13 @@ static void calculate_vertex_layout( struct i915_context *i915 ) boolean needW = 0; uint i; boolean texCoords[8]; + uint src = 0; memset(texCoords, 0, sizeof(texCoords)); memset(&vinfo, 0, sizeof(vinfo)); /* pos */ - draw_emit_vertex_attr(&vinfo, FORMAT_3F, INTERP_LINEAR); + draw_emit_vertex_attr(&vinfo, FORMAT_3F, INTERP_LINEAR, src++); /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ for (i = 0; i < fs->num_inputs; i++) { @@ -64,12 +65,12 @@ static void calculate_vertex_layout( struct i915_context *i915 ) break; case TGSI_SEMANTIC_COLOR: if (fs->input_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(&vinfo, FORMAT_4UB, colorInterp); + front0 = draw_emit_vertex_attr(&vinfo, FORMAT_4UB, colorInterp, src++); vinfo.hwfmt[0] |= S4_VFMT_COLOR; } else { assert(fs->input_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(&vinfo, FORMAT_4UB, colorInterp); + front1 = draw_emit_vertex_attr(&vinfo, FORMAT_4UB, colorInterp, src++); vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG; } break; @@ -79,7 +80,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) const uint unit = fs->input_semantic_index[i]; uint hwtc; texCoords[unit] = TRUE; - draw_emit_vertex_attr(&vinfo, FORMAT_4F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(&vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src++); hwtc = TEXCOORDFMT_4D; needW = TRUE; vinfo.hwfmt[1] |= hwtc << (unit * 4); @@ -87,7 +88,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) break; case TGSI_SEMANTIC_FOG: fprintf(stderr, "i915 fogcoord not implemented yet\n"); - draw_emit_vertex_attr(&vinfo, FORMAT_1F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(&vinfo, FORMAT_1F, INTERP_PERSPECTIVE, src++); break; default: assert(0); @@ -119,10 +120,10 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ if (i915->rasterizer->light_twoside) { if (front0) { - back0 = draw_emit_vertex_attr(&vinfo, FORMAT_OMIT, colorInterp); + back0 = draw_emit_vertex_attr(&vinfo, FORMAT_OMIT, colorInterp, src++); } if (back0) { - back1 = draw_emit_vertex_attr(&vinfo, FORMAT_OMIT, colorInterp); + back1 = draw_emit_vertex_attr(&vinfo, FORMAT_OMIT, colorInterp, src++); } } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 630ae3163f..0f1410e5de 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -48,6 +48,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) boolean emitBack0 = FALSE, emitBack1 = FALSE, emitPsize = FALSE; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; uint i; + int src = 0; memset(vinfo, 0, sizeof(*vinfo)); @@ -61,7 +62,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) softpipe->psize_slot = -1; /* always emit vertex pos */ - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR, src++); /* * XXX I think we need to reconcile the vertex shader outputs with @@ -82,11 +83,11 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) case TGSI_SEMANTIC_COLOR: if (vs->output_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); } else { assert(vs->output_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); } break; @@ -101,7 +102,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) break; case TGSI_SEMANTIC_FOG: - draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE, src++); break; case TGSI_SEMANTIC_PSIZE: @@ -113,7 +114,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src++); break; default: @@ -128,14 +129,14 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) * up 1:1 with the fragment shader inputs. */ if (emitBack0) { - back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); } if (emitBack1) { - back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp); + back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); } if (emitPsize) { softpipe->psize_slot - = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT); + = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT, src++); } /* If the attributes have changed, tell the draw module about -- cgit v1.2.3 From 7f21d04a518ac24348934961646501d6ed029466 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 11:37:58 -0700 Subject: fix broken point rendering in sp_vbuf_render() --- src/mesa/pipe/softpipe/sp_prim_setup.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 89f8df945c..444c80232f 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -1341,8 +1341,8 @@ void sp_vbuf_render( struct pipe_context *pipe, break; case PIPE_PRIM_POINTS: - for (i = 0; i < nr_elements; i += 2) { - prim.v[i] = (struct vertex_header *)((char *)vertex_buffer + + for (i = 0; i < nr_elements; i++) { + prim.v[0] = (struct vertex_header *)((char *)vertex_buffer + elements[i] * vertex_size); setup->stage.point( &setup->stage, &prim ); } -- cgit v1.2.3 From 329a8479b69a800b5fc6485767fb676c3eae26db Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 14:08:20 -0700 Subject: gallium: new 'draw' stage for line stipple Stippled lines are converted into a series of shorter line segments --- src/mesa/pipe/draw/draw_context.c | 2 + src/mesa/pipe/draw/draw_private.h | 2 + src/mesa/pipe/draw/draw_stipple.c | 242 +++++++++++++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_validate.c | 5 + src/mesa/sources | 1 + 5 files changed, 252 insertions(+) create mode 100644 src/mesa/pipe/draw/draw_stipple.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index fd43d690f6..e08873c5e1 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -49,6 +49,7 @@ struct draw_context *draw_create( void ) /* create pipeline stages */ draw->pipeline.wide = draw_wide_stage( draw ); + draw->pipeline.stipple = draw_stipple_stage( draw ); draw->pipeline.unfilled = draw_unfilled_stage( draw ); draw->pipeline.twoside = draw_twoside_stage( draw ); draw->pipeline.offset = draw_offset_stage( draw ); @@ -93,6 +94,7 @@ struct draw_context *draw_create( void ) void draw_destroy( struct draw_context *draw ) { draw->pipeline.wide->destroy( draw->pipeline.wide ); + draw->pipeline.stipple->destroy( draw->pipeline.stipple ); draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); draw->pipeline.twoside->destroy( draw->pipeline.twoside ); draw->pipeline.offset->destroy( draw->pipeline.offset ); diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index d8832449ea..20503fe158 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -158,6 +158,7 @@ struct draw_context struct draw_stage *twoside; struct draw_stage *offset; struct draw_stage *unfilled; + struct draw_stage *stipple; struct draw_stage *wide; struct draw_stage *rasterize; } pipeline; @@ -248,6 +249,7 @@ extern struct draw_stage *draw_offset_stage( struct draw_context *context ); extern struct draw_stage *draw_clip_stage( struct draw_context *context ); extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); extern struct draw_stage *draw_cull_stage( struct draw_context *context ); +extern struct draw_stage *draw_stipple_stage( struct draw_context *context ); extern struct draw_stage *draw_wide_stage( struct draw_context *context ); extern struct draw_stage *draw_validate_stage( struct draw_context *context ); diff --git a/src/mesa/pipe/draw/draw_stipple.c b/src/mesa/pipe/draw/draw_stipple.c new file mode 100644 index 0000000000..6d5226b1b8 --- /dev/null +++ b/src/mesa/pipe/draw/draw_stipple.c @@ -0,0 +1,242 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +/* Implement line stipple by cutting lines up into smaller lines. + * There are hundreds of ways to implement line stipple, this is one + * choice that should work in all situations, requires no state + * manipulations, but with a penalty in terms of large amounts of + * generated geometry. + */ + + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +/** Subclass of draw_stage */ +struct stipple_stage { + struct draw_stage stage; + float counter; + uint pattern; + uint factor; +}; + + +static INLINE struct stipple_stage * +stipple_stage(struct draw_stage *stage) +{ + return (struct stipple_stage *) stage; +} + + +/** + * Compute interpolated vertex attributes for 'dst' at position 't' + * between 'v0' and 'v1'. + */ +static void +screen_interp( struct draw_context *draw, + struct vertex_header *dst, + float t, + const struct vertex_header *v0, + const struct vertex_header *v1 ) +{ + uint attr; + for (attr = 0; attr < draw->vertex_info.num_attribs; attr++) { + switch (draw->vertex_info.interp_mode[attr]) { + case INTERP_NONE: + case INTERP_CONSTANT: + COPY_4FV(dst->data[attr], v0->data[attr]); + break; + case INTERP_PERSPECTIVE: + /* Fall-through */ + /* XXX special-case perspective? */ + case INTERP_LINEAR: + { + const float *val0 = v0->data[attr]; + const float *val1 = v1->data[attr]; + float *newv = dst->data[attr]; + uint i; + for (i = 0; i < 4; i++) { + newv[i] = val0[i] + t * (val1[i] - val0[i]); + } + } + break; + default: + abort(); + } + } +} + + +static void +emit_segment(struct draw_stage *stage, struct prim_header *header, + float t0, float t1) +{ + struct vertex_header *v0new = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1new = dup_vert(stage, header->v[1], 1); + struct prim_header newprim = *header; + + if (t0 > 0.0) { + screen_interp( stage->draw, v0new, t0, header->v[0], header->v[1] ); + newprim.v[0] = v0new; + } + + if (t1 < 1.0) { + screen_interp( stage->draw, v1new, t1, header->v[0], header->v[1] ); + newprim.v[1] = v1new; + } + + stage->next->line( stage->next, &newprim ); +} + + +static INLINE unsigned +stipple_test(int counter, ushort pattern, int factor) +{ + int b = (counter / factor) & 0xf; + return (1 << b) & pattern; +} + + +static void +stipple_line(struct draw_stage *stage, struct prim_header *header) +{ + struct stipple_stage *stipple = stipple_stage(stage); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const float *pos0 = v0->data[0]; + const float *pos1 = v1->data[0]; + float start = 0; + int state = 0; + + float x0 = pos0[0]; + float x1 = pos1[0]; + float y0 = pos0[1]; + float y1 = pos1[1]; + + float dx = x0 > x1 ? x0 - x1 : x1 - x0; + float dy = y0 > y1 ? y0 - y1 : y1 - y0; + + float length = MAX2(dx, dy); + int i; + + /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table. + */ + for (i = 0; i < length; i++) { + int result = stipple_test( stipple->counter+i, + stipple->pattern, stipple->factor ); + if (result != state) { + /* changing from "off" to "on" or vice versa */ + if (state) { + if (start != i) { + /* finishing an "on" segment */ + emit_segment( stage, header, start / length, i / length ); + } + } + else { + /* starting an "on" segment */ + start = i; + } + state = result; + } + } + + if (state && start < length) + emit_segment( stage, header, start / length, 1.0 ); + + stipple->counter += length; +} + + +static void +reset_stipple_counter(struct draw_stage *stage) +{ + struct stipple_stage *stipple = stipple_stage(stage); + stipple->counter = 0; + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +stipple_begin(struct draw_stage *stage) +{ + struct stipple_stage *stipple = stipple_stage(stage); + struct draw_context *draw = stage->draw; + + stipple->pattern = draw->rasterizer->line_stipple_pattern; + stipple->factor = draw->rasterizer->line_stipple_factor + 1; + + stage->next->begin( stage->next ); +} + + +static void +stipple_end(struct draw_stage *stage) +{ + stage->next->end( stage->next ); +} + + +static void +passthrough_point(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->point( stage->next, header ); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Create line stippler stage + */ +struct draw_stage *draw_stipple_stage( struct draw_context *draw ) +{ + struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); + + draw_alloc_tmps( &stipple->stage, 2 ); + + stipple->stage.draw = draw; + stipple->stage.next = NULL; + stipple->stage.begin = stipple_begin; + stipple->stage.point = passthrough_point; + stipple->stage.line = stipple_line; + stipple->stage.tri = passthrough_tri; + stipple->stage.reset_stipple_counter = reset_stipple_counter; + stipple->stage.end = stipple_end; + + return &stipple->stage; +} diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index 58cf340281..df06e0426c 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -58,6 +58,11 @@ static void validate_begin( struct draw_stage *stage ) next = draw->pipeline.wide; } + if (draw->rasterizer->line_stipple_enable) { + draw->pipeline.stipple->next = next; + next = draw->pipeline.stipple; + } + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { draw->pipeline.unfilled->next = next; diff --git a/src/mesa/sources b/src/mesa/sources index d8760ba134..97ef7e1936 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -165,6 +165,7 @@ DRAW_SOURCES = \ pipe/draw/draw_flatshade.c \ pipe/draw/draw_offset.c \ pipe/draw/draw_prim.c \ + pipe/draw/draw_stipple.c \ pipe/draw/draw_twoside.c \ pipe/draw/draw_unfilled.c \ pipe/draw/draw_validate.c \ -- cgit v1.2.3 From ea3a2b440207f319438fe76d9a451e78f0e575d4 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 14:10:16 -0700 Subject: gallium: remove per-fragment line stipple code --- src/mesa/pipe/softpipe/sp_prim_setup.c | 40 +++++----------------------------- 1 file changed, 5 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 444c80232f..cdc385e4db 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -769,7 +769,7 @@ line_persp_coeff(struct setup_stage *setup, { /* XXX double-check/verify this arithmetic */ const float a0 = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; - const float a1 = setup->vmax->data[vertSlot][i] * setup->vmin->data[0][3]; + const float a1 = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; const float da = a1 - a0; const float dadx = da * setup->emaj.dx * setup->oneoverarea; const float dady = da * setup->emaj.dy * setup->oneoverarea; @@ -872,21 +872,10 @@ plot(struct setup_stage *setup, int x, int y) } -/** - * Determine whether or not to emit a line fragment by checking - * line stipple pattern. - */ -static INLINE unsigned -stipple_test(int counter, ushort pattern, int factor) -{ - int b = (counter / factor) & 0xf; - return (1 << b) & pattern; -} - - /** * Do setup for line rasterization, then render the line. - * XXX single-pixel width, no stipple, etc + * Single-pixel width, no stipple, etc. We rely on the 'draw' module + * to handle stippling and wide lines. */ static void setup_line(struct draw_stage *stage, struct prim_header *prim) @@ -894,8 +883,6 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) const struct vertex_header *v0 = prim->v[0]; const struct vertex_header *v1 = prim->v[1]; struct setup_stage *setup = setup_stage( stage ); - struct softpipe_context *sp = setup->softpipe; - int x0 = (int) v0->data[0][0]; int x1 = (int) v1->data[0][0]; int y0 = (int) v0->data[0][1]; @@ -947,12 +934,7 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) const int errorDec = error - dx; for (i = 0; i < dx; i++) { - if (!sp->rasterizer->line_stipple_enable || - stipple_test(sp->line_stipple_counter, - (ushort) sp->rasterizer->line_stipple_pattern, - sp->rasterizer->line_stipple_factor + 1)) { - plot(setup, x0, y0); - } + plot(setup, x0, y0); x0 += xstep; if (error < 0) { @@ -962,8 +944,6 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) error += errorDec; y0 += ystep; } - - sp->line_stipple_counter++; } } else { @@ -974,15 +954,9 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) const int errorDec = error - dy; for (i = 0; i < dy; i++) { - if (!sp->rasterizer->line_stipple_enable || - stipple_test(sp->line_stipple_counter, - (ushort) sp->rasterizer->line_stipple_pattern, - sp->rasterizer->line_stipple_factor + 1)) { - plot(setup, x0, y0); - } + plot(setup, x0, y0); y0 += ystep; - if (error < 0) { error += errorInc; } @@ -990,8 +964,6 @@ setup_line(struct draw_stage *stage, struct prim_header *prim) error += errorDec; x0 += xstep; } - - sp->line_stipple_counter++; } } @@ -1234,8 +1206,6 @@ static void setup_end( struct draw_stage *stage ) static void reset_stipple_counter( struct draw_stage *stage ) { - struct setup_stage *setup = setup_stage(stage); - setup->softpipe->line_stipple_counter = 0; } -- cgit v1.2.3 From be92796eb9b0ef35ad84c59ad183863ed2dbe002 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 15:29:54 -0700 Subject: gallium: prototype draw_linestipple.c replaced by draw_stipple.c --- src/mesa/pipe/draw/draw_linestipple.c | 272 ---------------------------------- 1 file changed, 272 deletions(-) delete mode 100644 src/mesa/pipe/draw/draw_linestipple.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_linestipple.c b/src/mesa/pipe/draw/draw_linestipple.c deleted file mode 100644 index d7fe1071f0..0000000000 --- a/src/mesa/pipe/draw/draw_linestipple.c +++ /dev/null @@ -1,272 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -/* Implement line stipple by cutting lines up into smaller lines. - * There are hundreds of ways to implement line stipple, this is one - * choice that should work in all situations, requires no state - * manipulations, but with a penalty in terms of large amounts of - * generated geometry. - */ - -#include "imports.h" -#include "macros.h" - -#define CLIP_PRIVATE -#include "clip/clip_context.h" - -#define CLIP_PIPE_PRIVATE -#include "clip/clip_pipe.h" - - - -struct stipple_stage { - struct clip_pipe_stage stage; - - GLuint hw_data_offset; - - GLfloat counter; - GLuint pattern; - GLuint factor; -}; - - - - -static INLINE struct stipple_stage *stipple_stage( struct clip_pipe_stage *stage ) -{ - return (struct stipple_stage *)stage; -} - - - - -static void interp_attr( const struct vf_attr *a, - GLubyte *vdst, - GLfloat t, - const GLubyte *vin, - const GLubyte *vout ) -{ - GLuint offset = a->vertoffset; - GLfloat fin[4], fout[4], fdst[4]; - - a->extract( a, fin, vin + offset ); - a->extract( a, fout, vout + offset ); - - fdst[0] = LINTERP( t, fout[0], fin[0] ); - fdst[1] = LINTERP( t, fout[1], fin[1] ); - fdst[2] = LINTERP( t, fout[2], fin[2] ); - fdst[3] = LINTERP( t, fout[3], fin[3] ); - - a->insert[4-1]( a, vdst + offset, fdst ); -} - - - - -/* Weird screen-space interpolation?? Otherwise do something special - * with pos.w or fix vertices to always have clip coords available. - */ -static void screen_interp( struct vertex_fetch *vf, - struct vertex_header *dst, - GLfloat t, - const struct vertex_header *out, - const struct vertex_header *in ) -{ - GLubyte *vdst = (GLubyte *)dst; - const GLubyte *vin = (const GLubyte *)in; - const GLubyte *vout = (const GLubyte *)out; - - const struct vf_attr *a = vf->attr; - const GLuint attr_count = vf->attr_count; - GLuint j; - - /* Vertex header. - */ - { - assert(a[0].attrib == VF_ATTRIB_VERTEX_HEADER); - dst->clipmask = 0; - dst->edgeflag = 0; - dst->pad = 0; - dst->index = 0xffff; - } - - - /* Other attributes - */ - for (j = 1; j < attr_count; j++) { - interp_attr(&a[j], vdst, t, vin, vout); - } -} - - - -/* Clip a line against the viewport and user clip planes. - */ -static void draw_line_segment( struct clip_pipe_stage *stage, - struct prim_header *header, - GLfloat t0, - GLfloat t1 ) -{ - struct vertex_fetch *vf = stage->pipe->draw->vb.vf; - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - - struct prim_header newprim = *header; - header = &newprim; - - _mesa_printf("%s %f %f\n", __FUNCTION__, t0, t1); - - if (t0 > 0.0) { - screen_interp( vf, stage->tmp[0], t0, v0, v1 ); - header->v[0] = stage->tmp[0]; - } - - if (t1 < 1.0) { - screen_interp( vf, stage->tmp[1], t1, v0, v1 ); - header->v[1] = stage->tmp[1]; - } - - stage->next->line( stage->next, header ); -} - - - -/* XXX: don't really want to iterate like this. - */ -static INLINE unsigned -stipple_test(int counter, ushort pattern, int factor) -{ - int b = (counter / factor) & 0xf; - return (1 << b) & pattern; -} - - - -/* XXX: Need to have precalculated flatshading already. - */ -static void stipple_line( struct clip_pipe_stage *stage, - struct prim_header *header ) -{ - struct stipple_stage *stipple = stipple_stage(stage); - GLuint hw_data_offset = stipple->hw_data_offset; - const GLfloat *pos0 = (GLfloat *)&(header->v[0]->data[hw_data_offset]); - const GLfloat *pos1 = (GLfloat *)&(header->v[1]->data[hw_data_offset]); - GLfloat start = 0; - int state = 0; - - GLfloat x0 = (GLfloat)pos0[0]; - GLfloat x1 = (GLfloat)pos1[0]; - GLfloat y0 = (GLfloat)pos0[1]; - GLfloat y1 = (GLfloat)pos1[1]; - - GLfloat dx = x0 > x1 ? x0 - x1 : x1 - x0; - GLfloat dy = y0 > y1 ? y0 - y1 : y1 - y0; - - GLfloat length = MAX2(dx, dy); - GLint i; - - if (header->reset_line_stipple) - stipple->counter = 0; - - /* XXX: iterating like this is lame - */ - for (i = 0; i < length; i++) { - int result = !!stipple_test( stipple->counter+i, stipple->pattern, stipple->factor ); -// _mesa_printf("%d %f %d\n", i, length, result); - if (result != state) { - if (state) { - if (start != i) - draw_line_segment( stage, header, start / length, i / length ); - } - else { - start = i; - } - state = result; - } - } - - if (state && start < length) - draw_line_segment( stage, header, start / length, 1.0 ); - - stipple->counter += length; -} - - - -static void stipple_begin( struct clip_pipe_stage *stage ) -{ - struct stipple_stage *stipple = stipple_stage(stage); - struct clip_context *draw = stage->pipe->draw; - - if (stage->pipe->draw->vb_state.clipped_prims) - stipple->hw_data_offset = 16; - else - stipple->hw_data_offset = 0; - - stipple->stage.tri = clip_passthrough_tri; - stipple->stage.point = clip_passthrough_point; - - stipple->stage.line = stipple_line; - stipple->factor = draw->state.line_stipple_factor + 1; - stipple->pattern = draw->state.line_stipple_pattern; - - stage->next->begin( stage->next ); -} - - -static void stipple_end( struct clip_pipe_stage *stage ) -{ - stage->next->end( stage->next ); -} - - -static void stipple_destroy( struct clip_pipe_stage *stage ) -{ - FREE( stage ); -} - - -struct clip_pipe_stage *clip_pipe_stipple( struct clip_pipeline *pipe ) -{ - struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); - - clip_pipe_alloc_tmps( &stipple->stage, 4 ); - - stipple->stage.pipe = pipe; - stipple->stage.next = NULL; - stipple->stage.begin = stipple_begin; - stipple->stage.point = clip_passthrough_point; - stipple->stage.line = stipple_line; - stipple->stage.tri = clip_passthrough_tri; - stipple->stage.end = stipple_end; - stipple->stage.destroy = stipple_destroy; - - return &stipple->stage; -} -- cgit v1.2.3 From 5a57ffafd623a04b86aea19c000ff4a64c47fd43 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 19:25:10 -0700 Subject: Cell: added asmfiles rule --- src/mesa/pipe/cell/spu/Makefile | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index b92a756cd5..417ae1b072 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -22,12 +22,17 @@ SOURCES = \ SPU_OBJECTS = $(SOURCES:.c=.o) \ +SPU_ASM_OUT = $(SOURCES:.c=.s) \ + INCLUDE_DIRS = -I$(TOP)/src/mesa .c.o: $(SPU_CC) $(SPU_CFLAGS) -c $< +.c.s: + $(SPU_CC) $(SPU_CFLAGS) -S $< + # The .a file will be linked into the main/PPU executable default: $(PROG_SPU_A) @@ -43,8 +48,11 @@ $(PROG_SPU): $(SPU_OBJECTS) +asmfiles: $(SPU_ASM_OUT) + + clean: - rm -f *~ *.o *.a *.d $(PROG_SPU) + rm -f *~ *.o *.a *.d *.s $(PROG_SPU) -- cgit v1.2.3 From e2612f4aa057ef20cc6eaf37337239bfc580d328 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:14:22 -0700 Subject: gallium: move softpipe_get_tex_surface() into sp_texture.c --- src/mesa/pipe/softpipe/sp_surface.c | 38 ------------------------------------- src/mesa/pipe/softpipe/sp_texture.c | 36 +++++++++++++++++++++++++++++++++++ 2 files changed, 36 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index d5119654ed..e115705507 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -32,44 +32,6 @@ #include "pipe/util/p_tile.h" #include "sp_context.h" #include "sp_surface.h" -#include "sp_texture.h" - - -/** - * Called via pipe->get_tex_surface() - * XXX is this in the right place? - */ -struct pipe_surface * -softpipe_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - struct softpipe_texture *spt = softpipe_texture(pt); - struct pipe_surface *ps; - - ps = pipe->winsys->surface_alloc(pipe->winsys); - if (ps) { - assert(ps->refcount); - assert(ps->winsys); - pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); - ps->format = pt->format; - ps->cpp = pt->cpp; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->pitch = ps->width; - ps->offset = spt->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { - ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - (pt->compressed ? ps->height/4 : ps->height) * - ps->width * ps->cpp; - } else { - assert(face == 0); - assert(zslice == 0); - } - } - return ps; -} /* Upload data to a rectangular sub-region. Lots of choices how to do this: diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index 532bcfcc51..d43a6996e9 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -130,3 +130,39 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } *pt = NULL; } + + +/** + * Called via pipe->get_tex_surface() + */ +struct pipe_surface * +softpipe_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct softpipe_texture *spt = softpipe_texture(pt); + struct pipe_surface *ps; + + ps = pipe->winsys->surface_alloc(pipe->winsys); + if (ps) { + assert(ps->refcount); + assert(ps->winsys); + pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = ps->width; + ps->offset = spt->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { + ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * + (pt->compressed ? ps->height/4 : ps->height) * + ps->width * ps->cpp; + } else { + assert(face == 0); + assert(zslice == 0); + } + } + return ps; +} -- cgit v1.2.3 From c14773b06be1229ada36be8c1a2b3a2dad846285 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:16:13 -0700 Subject: gallium: add license comment --- src/mesa/pipe/softpipe/sp_texture.h | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_texture.h b/src/mesa/pipe/softpipe/sp_texture.h index e1a5db2791..d142bdddac 100644 --- a/src/mesa/pipe/softpipe/sp_texture.h +++ b/src/mesa/pipe/softpipe/sp_texture.h @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #ifndef SP_TEXTURE_H #define SP_TEXTURE_H -- cgit v1.2.3 From 4211d2d4618ace01c70b86f69d8d0d297f83d74d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:23:48 -0700 Subject: gallium: move softpipe_get_tex_surface() prototype --- src/mesa/pipe/softpipe/sp_surface.h | 7 ------- src/mesa/pipe/softpipe/sp_texture.h | 7 +++++-- 2 files changed, 5 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index b652e7598e..f06a325dae 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -36,13 +36,6 @@ struct pipe_context; struct softpipe_context; -struct softpipe_tile_cache; - - -extern struct pipe_surface * -softpipe_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice); extern void diff --git a/src/mesa/pipe/softpipe/sp_texture.h b/src/mesa/pipe/softpipe/sp_texture.h index d142bdddac..0494bf365b 100644 --- a/src/mesa/pipe/softpipe/sp_texture.h +++ b/src/mesa/pipe/softpipe/sp_texture.h @@ -61,7 +61,10 @@ softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); extern void softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); - -#endif /* SP_TEXTURE */ +extern struct pipe_surface * +softpipe_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice); +#endif /* SP_TEXTURE */ -- cgit v1.2.3 From fd9c03fc428b325103051d4543e464def71d0b53 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:25:44 -0700 Subject: Cell: add cell_set_sampler_texture() --- src/mesa/pipe/cell/ppu/cell_state.h | 5 +++++ src/mesa/pipe/cell/ppu/cell_state_sampler.c | 14 ++++++++++++++ 2 files changed, 19 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state.h b/src/mesa/pipe/cell/ppu/cell_state.h index d2091f8edf..fbca7c9574 100644 --- a/src/mesa/pipe/cell/ppu/cell_state.h +++ b/src/mesa/pipe/cell/ppu/cell_state.h @@ -86,6 +86,11 @@ void cell_set_constant_buffer(struct pipe_context *pipe, void cell_set_polygon_stipple( struct pipe_context *, const struct pipe_poly_stipple * ); +void +cell_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, + struct pipe_texture *texture); + void cell_set_scissor_state( struct pipe_context *, const struct pipe_scissor_state * ); diff --git a/src/mesa/pipe/cell/ppu/cell_state_sampler.c b/src/mesa/pipe/cell/ppu/cell_state_sampler.c index 1e7d4f08b8..ae1eeb4620 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_sampler.c +++ b/src/mesa/pipe/cell/ppu/cell_state_sampler.c @@ -66,3 +66,17 @@ cell_delete_sampler_state(struct pipe_context *pipe, { FREE( sampler ); } + + + +void +cell_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, + struct pipe_texture *texture) +{ + struct cell_context *cell = cell_context(pipe); + + cell->texture[sampler] = texture; + + cell->dirty |= CELL_NEW_TEXTURE; +} -- cgit v1.2.3 From 1811965091341fb2c5c94ac0e3bb3c2ca4de161e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:26:40 -0700 Subject: Cell: Add initial texture functions --- src/mesa/pipe/cell/ppu/Makefile | 1 + src/mesa/pipe/cell/ppu/cell_context.c | 12 ++- src/mesa/pipe/cell/ppu/cell_context.h | 2 +- src/mesa/pipe/cell/ppu/cell_texture.c | 168 ++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_texture.h | 73 +++++++++++++++ 5 files changed, 251 insertions(+), 5 deletions(-) create mode 100644 src/mesa/pipe/cell/ppu/cell_texture.c create mode 100644 src/mesa/pipe/cell/ppu/cell_texture.h (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index 6a1bd5982f..0df07004a4 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -31,6 +31,7 @@ SOURCES = \ cell_state_vertex.c \ cell_spu.c \ cell_surface.c \ + cell_texture.c \ cell_vbuf.c \ cell_winsys.c diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 897c187d65..fc1ca28f3a 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -46,6 +46,7 @@ #include "cell_state.h" #include "cell_surface.h" #include "cell_spu.h" +#include "cell_texture.h" #include "cell_vbuf.h" @@ -225,14 +226,17 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->pipe.clear = cell_clear_surface; cell->pipe.flush = cell_flush; + /* textures */ + cell->pipe.texture_create = cell_texture_create; + cell->pipe.texture_release = cell_texture_release; + cell->pipe.get_tex_surface = cell_get_tex_surface; + + cell->pipe.set_sampler_texture = cell_set_sampler_texture; + #if 0 cell->pipe.begin_query = cell_begin_query; cell->pipe.end_query = cell_end_query; cell->pipe.wait_query = cell_wait_query; - - /* textures */ - cell->pipe.mipmap_tree_layout = cell_mipmap_tree_layout; - cell->pipe.get_tex_surface = cell_get_tex_surface; #endif diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 08e448f14f..1937812e2d 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -76,7 +76,7 @@ struct cell_context struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct softpipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; diff --git a/src/mesa/pipe/cell/ppu/cell_texture.c b/src/mesa/pipe/cell/ppu/cell_texture.c new file mode 100644 index 0000000000..8f342d8aad --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_texture.c @@ -0,0 +1,168 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "cell_context.h" +#include "cell_state.h" +#include "cell_texture.h" + + +/* Simple, maximally packed layout. + */ + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + + +static void +cell_texture_layout(struct cell_texture * spt) +{ + struct pipe_texture *pt = &spt->base; + unsigned level; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + + spt->buffer_size = 0; + + for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + pt->width[level] = width; + pt->height[level] = height; + pt->depth[level] = depth; + + spt->level_offset[level] = spt->buffer_size; + + spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * + ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * + width * pt->cpp; + + width = minify(width); + height = minify(height); + depth = minify(depth); + } +} + + +void +cell_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) +{ + struct cell_texture *spt = REALLOC(*pt, sizeof(struct pipe_texture), + sizeof(struct cell_texture)); + + if (spt) { + memset(&spt->base + 1, 0, + sizeof(struct cell_texture) - sizeof(struct pipe_texture)); + + cell_texture_layout(spt); + + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); + + if (spt->buffer) { + pipe->winsys->buffer_data(pipe->winsys, spt->buffer, spt->buffer_size, + NULL, PIPE_BUFFER_USAGE_PIXEL); + } + + if (!spt->buffer) { + FREE(spt); + spt = NULL; + } + } + + *pt = &spt->base; +} + +void +cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct cell_texture *spt = cell_texture(*pt); + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); + */ + + pipe->winsys->buffer_reference(pipe->winsys, &spt->buffer, NULL); + + FREE(spt); + } + *pt = NULL; +} + + +/** + * Called via pipe->get_tex_surface() + */ +struct pipe_surface * +cell_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct cell_texture *spt = cell_texture(pt); + struct pipe_surface *ps; + + ps = pipe->winsys->surface_alloc(pipe->winsys); + if (ps) { + assert(ps->refcount); + assert(ps->winsys); + pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = ps->width; + ps->offset = spt->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { + ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * + (pt->compressed ? ps->height/4 : ps->height) * + ps->width * ps->cpp; + } else { + assert(face == 0); + assert(zslice == 0); + } + } + return ps; +} diff --git a/src/mesa/pipe/cell/ppu/cell_texture.h b/src/mesa/pipe/cell/ppu/cell_texture.h new file mode 100644 index 0000000000..97d8bd1230 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_texture.h @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_TEXTURE_H +#define CELL_TEXTURE_H + + +struct pipe_context; +struct pipe_texture; + + +/** + * Subclass of pipe_texture + */ +struct cell_texture +{ + struct pipe_texture base; + + unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer_handle *buffer; + unsigned long buffer_size; +}; + + +/** cast wrapper */ +static INLINE struct cell_texture * +cell_texture(struct pipe_texture *pt) +{ + return (struct cell_texture *) pt; +} + + + +extern void +cell_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); + +extern void +cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); + +extern struct pipe_surface * +cell_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice); + + +#endif /* CELL_TEXTURE */ -- cgit v1.2.3 From c27c3529845b535826f56c16fb0e9a258e5d5841 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:31:16 -0700 Subject: Cell: rename cell_surface.[ch] cell_clear.[ch] --- src/mesa/pipe/cell/ppu/Makefile | 2 +- src/mesa/pipe/cell/ppu/cell_clear.c | 99 +++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_clear.h | 43 +++++++++++++++ src/mesa/pipe/cell/ppu/cell_context.c | 2 +- 4 files changed, 144 insertions(+), 2 deletions(-) create mode 100644 src/mesa/pipe/cell/ppu/cell_clear.c create mode 100644 src/mesa/pipe/cell/ppu/cell_clear.h (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index 0df07004a4..33ce582143 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -17,6 +17,7 @@ SPU_CODE_MODULE = ../spu/g3d_spu.a SOURCES = \ cell_batch.c \ + cell_clear.c \ cell_context.c \ cell_draw_arrays.c \ cell_flush.c \ @@ -30,7 +31,6 @@ SOURCES = \ cell_state_surface.c \ cell_state_vertex.c \ cell_spu.c \ - cell_surface.c \ cell_texture.c \ cell_vbuf.c \ cell_winsys.c diff --git a/src/mesa/pipe/cell/ppu/cell_clear.c b/src/mesa/pipe/cell/ppu/cell_clear.c new file mode 100644 index 0000000000..48c1c8e2c8 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_clear.c @@ -0,0 +1,99 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Authors + * Brian Paul + */ + +#include +#include +#include +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/cell/common.h" +#include "cell_clear.h" +#include "cell_context.h" +#include "cell_batch.h" +#include "cell_spu.h" + + +void +cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + struct cell_context *cell = cell_context(pipe); + uint i; + uint surfIndex; + + if (!cell->cbuf_map[0]) + cell->cbuf_map[0] = pipe_surface_map(ps); + + if (ps == cell->framebuffer.zbuf) { + surfIndex = 1; + } + else { + surfIndex = 0; + } + +#if 0 + for (i = 0; i < cell->num_spus; i++) { +#if 1 + uint clr = clearValue; + if (surfIndex == 0) { + /* XXX debug: clear color varied per-SPU to visualize tiles */ + if ((clr & 0xff) == 0) + clr |= 64 + i * 8; + if ((clr & 0xff00) == 0) + clr |= (64 + i * 8) << 8; + if ((clr & 0xff0000) == 0) + clr |= (64 + i * 8) << 16; + if ((clr & 0xff000000) == 0) + clr |= (64 + i * 8) << 24; + } + cell_global.command[i].clear.value = clr; +#else + cell_global.command[i].clear.value = clearValue; +#endif + cell_global.command[i].clear.surface = surfIndex; + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_SURFACE); + } +#else + { + struct cell_command_clear_surface *clr + = (struct cell_command_clear_surface *) + cell_batch_alloc(cell, sizeof(*clr)); + clr->opcode = CELL_CMD_CLEAR_SURFACE; + clr->surface = surfIndex; + clr->value = clearValue; + } +#endif + + /* XXX temporary */ + cell_flush(&cell->pipe, 0x0); + +} diff --git a/src/mesa/pipe/cell/ppu/cell_clear.h b/src/mesa/pipe/cell/ppu/cell_clear.h new file mode 100644 index 0000000000..4c69c4c89f --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_clear.h @@ -0,0 +1,43 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_SURFACE_H +#define CELL_SURFACE_H + + +struct pipe_context; +struct pipe_surface; + + +extern void +cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + + + +#endif /* CELL_SURFACE_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index fc1ca28f3a..be9336259d 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -44,7 +44,7 @@ #include "cell_flush.h" #include "cell_render.h" #include "cell_state.h" -#include "cell_surface.h" +#include "cell_clear.h" #include "cell_spu.h" #include "cell_texture.h" #include "cell_vbuf.h" -- cgit v1.2.3 From 8538da0d7a3a4fc0cffa700d5fa220895688e44e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:31:36 -0700 Subject: Cell: renamed cell_surface.[ch] to cell_clear.[ch] --- src/mesa/pipe/cell/ppu/cell_surface.c | 99 ----------------------------------- src/mesa/pipe/cell/ppu/cell_surface.h | 43 --------------- 2 files changed, 142 deletions(-) delete mode 100644 src/mesa/pipe/cell/ppu/cell_surface.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_surface.h (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c deleted file mode 100644 index 53c5325ddd..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ /dev/null @@ -1,99 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Authors - * Brian Paul - */ - -#include -#include -#include -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "pipe/cell/common.h" -#include "cell_context.h" -#include "cell_batch.h" -#include "cell_surface.h" -#include "cell_spu.h" - - -void -cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue) -{ - struct cell_context *cell = cell_context(pipe); - uint i; - uint surfIndex; - - if (!cell->cbuf_map[0]) - cell->cbuf_map[0] = pipe_surface_map(ps); - - if (ps == cell->framebuffer.zbuf) { - surfIndex = 1; - } - else { - surfIndex = 0; - } - -#if 0 - for (i = 0; i < cell->num_spus; i++) { -#if 1 - uint clr = clearValue; - if (surfIndex == 0) { - /* XXX debug: clear color varied per-SPU to visualize tiles */ - if ((clr & 0xff) == 0) - clr |= 64 + i * 8; - if ((clr & 0xff00) == 0) - clr |= (64 + i * 8) << 8; - if ((clr & 0xff0000) == 0) - clr |= (64 + i * 8) << 16; - if ((clr & 0xff000000) == 0) - clr |= (64 + i * 8) << 24; - } - cell_global.command[i].clear.value = clr; -#else - cell_global.command[i].clear.value = clearValue; -#endif - cell_global.command[i].clear.surface = surfIndex; - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_SURFACE); - } -#else - { - struct cell_command_clear_surface *clr - = (struct cell_command_clear_surface *) - cell_batch_alloc(cell, sizeof(*clr)); - clr->opcode = CELL_CMD_CLEAR_SURFACE; - clr->surface = surfIndex; - clr->value = clearValue; - } -#endif - - /* XXX temporary */ - cell_flush(&cell->pipe, 0x0); - -} diff --git a/src/mesa/pipe/cell/ppu/cell_surface.h b/src/mesa/pipe/cell/ppu/cell_surface.h deleted file mode 100644 index 4c69c4c89f..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_surface.h +++ /dev/null @@ -1,43 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef CELL_SURFACE_H -#define CELL_SURFACE_H - - -struct pipe_context; -struct pipe_surface; - - -extern void -cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue); - - - -#endif /* CELL_SURFACE_H */ -- cgit v1.2.3 From b91de8a6e7d74d38698b6dad9f34040e484e69af Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:50:39 -0700 Subject: gallium: include sp_headers.h --- src/mesa/pipe/softpipe/sp_tex_sample.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 9e48ed0cd2..5e215c433a 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -33,6 +33,7 @@ */ #include "sp_context.h" +#include "sp_headers.h" #include "sp_surface.h" #include "sp_tex_sample.h" #include "sp_tile_cache.h" -- cgit v1.2.3 From a1799ecd82c5a04123d1f06ef7a2e18c6b719f96 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:51:03 -0700 Subject: gallium: remove unneeded protos, #includes, etc --- src/mesa/pipe/softpipe/sp_surface.h | 25 ------------------------- 1 file changed, 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h index f06a325dae..22de3ba43f 100644 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ b/src/mesa/pipe/softpipe/sp_surface.h @@ -31,35 +31,10 @@ #ifndef SP_SURFACE_H #define SP_SURFACE_H -#include "sp_headers.h" -#include "pipe/p_state.h" -struct pipe_context; struct softpipe_context; -extern void -softpipe_get_tile(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, void *p, int dst_stride); - -extern void -softpipe_put_tile(struct pipe_context *pipe, struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride); - -extern void -softpipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p); - -extern void -softpipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p); - - extern void sp_init_surface_functions(struct softpipe_context *sp); -- cgit v1.2.3 From 13189412d93cd7ce5863aaf21b8c91d21bd535c0 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:53:23 -0700 Subject: Cell: Added cell_surface.[ch] files --- src/mesa/pipe/cell/ppu/Makefile | 1 + src/mesa/pipe/cell/ppu/cell_context.c | 4 +- src/mesa/pipe/cell/ppu/cell_surface.c | 182 ++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_surface.h | 42 ++++++++ 4 files changed, 228 insertions(+), 1 deletion(-) create mode 100644 src/mesa/pipe/cell/ppu/cell_surface.c create mode 100644 src/mesa/pipe/cell/ppu/cell_surface.h (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index 33ce582143..e7f2562da7 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -31,6 +31,7 @@ SOURCES = \ cell_state_surface.c \ cell_state_vertex.c \ cell_spu.c \ + cell_surface.c \ cell_texture.c \ cell_vbuf.c \ cell_winsys.c diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index be9336259d..1403a9b669 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -39,12 +39,13 @@ #include "pipe/p_winsys.h" #include "pipe/cell/common.h" #include "pipe/draw/draw_context.h" +#include "cell_clear.h" #include "cell_context.h" #include "cell_draw_arrays.h" #include "cell_flush.h" #include "cell_render.h" #include "cell_state.h" -#include "cell_clear.h" +#include "cell_surface.h" #include "cell_spu.h" #include "cell_texture.h" #include "cell_vbuf.h" @@ -239,6 +240,7 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->pipe.wait_query = cell_wait_query; #endif + cell_init_surface_funcs(cell); cell->draw = draw_create(); diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c new file mode 100644 index 0000000000..c8796dab43 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -0,0 +1,182 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "pipe/util/p_tile.h" +#include "cell_context.h" +#include "cell_surface.h" + + +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +static void +cell_surface_data(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + const void *src, unsigned src_pitch, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + pipe_surface_unmap(dst); +} + + +static void +cell_surface_copy(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + assert( dst->cpp == src->cpp ); + + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe_surface_map(src), + src->pitch, + srcx, srcy); + + pipe_surface_unmap(src); + pipe_surface_unmap(dst); +} + + +static void * +get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) +{ + return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; +} + + +#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8)) + + +/** + * Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +static void +cell_surface_fill(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) +{ + unsigned i, j; + void *dst_map = pipe_surface_map(dst); + + assert(dst->pitch > 0); + assert(width <= dst->pitch); + + switch (dst->cpp) { + case 1: + { + ubyte *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: + { + ushort *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (ushort) value; + row += dst->pitch; + } + } + break; + case 4: + { + unsigned *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + case 8: + { + /* expand the 4-byte clear value to an 8-byte value */ + ushort *row = (ushort *) get_pointer(dst, dst_map, dstx, dsty); + ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); + ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); + ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); + ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff); + val0 = (val0 << 8) | val0; + val1 = (val1 << 8) | val1; + val2 = (val2 << 8) | val2; + val3 = (val3 << 8) | val3; + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) { + row[j*4+0] = val0; + row[j*4+1] = val1; + row[j*4+2] = val2; + row[j*4+3] = val3; + } + row += dst->pitch * 4; + } + } + break; + default: + assert(0); + break; + } + + pipe_surface_unmap( dst ); +} + + +void +cell_init_surface_functions(struct cell_context *cell) +{ + cell->pipe.get_tile = pipe_get_tile_raw; + cell->pipe.put_tile = pipe_put_tile_raw; + + cell->pipe.surface_data = cell_surface_data; + cell->pipe.surface_copy = cell_surface_copy; + cell->pipe.surface_fill = cell_surface_fill; +} diff --git a/src/mesa/pipe/cell/ppu/cell_surface.h b/src/mesa/pipe/cell/ppu/cell_surface.h new file mode 100644 index 0000000000..9e58f32944 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_surface.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef CELL_SURFACE_H +#define CELL_SURFACE_H + + +struct cell_context; + + +extern void +cell_init_surface_functions(struct cell_context *cell); + + +#endif /* SP_SURFACE_H */ -- cgit v1.2.3 From 0c085717cbb63f00223dc1df3d514f1ef2aaeda1 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 20:54:38 -0700 Subject: Cell: s/cell_init_surface_funcs/cell_init_surface_functions/ --- src/mesa/pipe/cell/ppu/cell_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 1403a9b669..3db1dbdb46 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -240,7 +240,7 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->pipe.wait_query = cell_wait_query; #endif - cell_init_surface_funcs(cell); + cell_init_surface_functions(cell); cell->draw = draw_create(); -- cgit v1.2.3 From cfeda4165b2845d2253d3082aa771a2346bac6a0 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 21:15:43 -0700 Subject: Cell: initial support for sampler state --- src/mesa/pipe/cell/common.h | 1 + src/mesa/pipe/cell/ppu/cell_state_emit.c | 7 +++++++ src/mesa/pipe/cell/spu/spu_main.c | 15 +++++++++++++++ src/mesa/pipe/cell/spu/spu_main.h | 1 + 4 files changed, 24 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index f05070d25a..4c998722a2 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -60,6 +60,7 @@ #define CELL_CMD_RENDER 5 #define CELL_CMD_BATCH 6 #define CELL_CMD_STATE_DEPTH_STENCIL 7 +#define CELL_CMD_STATE_SAMPLER 8 #define CELL_NUM_BATCH_BUFFERS 3 diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c index e1a1458f39..e7d14d0d25 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.c +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.c @@ -42,4 +42,11 @@ cell_emit_state(struct cell_context *cell) cell_batch_append(cell, cell->depth_stencil, sizeof(struct pipe_depth_stencil_alpha_state)); } + + if (cell->dirty & CELL_NEW_SAMPLER) { + uint cmd = CELL_CMD_STATE_SAMPLER; + cell_batch_append(cell, &cmd, 4); + cell_batch_append(cell, cell->sampler[0], + sizeof(struct pipe_sampler_state)); + } } diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 2b32c26854..2e5cb76b4a 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -378,7 +378,17 @@ cmd_state_depth_stencil(const struct pipe_depth_stencil_alpha_state *state) state->depth.enabled); memcpy(&spu.depth_stencil, state, sizeof(*state)); +} + + +static void +cmd_state_sampler(const struct pipe_sampler_state *state) +{ + if (Debug) + printf("SPU %u: SAMPLER\n", + spu.init.id); + memcpy(&spu.sampler[0], state, sizeof(*state)); } @@ -501,6 +511,11 @@ cmd_batch(uint opcode) &buffer[pos+1]); pos += (1 + sizeof(struct pipe_depth_stencil_alpha_state) / 4); break; + case CELL_CMD_STATE_SAMPLER: + cmd_state_sampler((struct pipe_sampler_state *) + &buffer[pos+1]); + pos += (1 + sizeof(struct pipe_sampler_state) / 4); + break; default: printf("SPU %u: bad opcode: 0x%x\n", spu.init.id, buffer[pos]); ASSERT(0); diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 2aa7015ae3..3ef73c9473 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -58,6 +58,7 @@ struct spu_global struct spu_framebuffer fb; struct pipe_depth_stencil_alpha_state depth_stencil; struct pipe_blend_state blend; + struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; /* XXX more state to come */ } ALIGN16_ATTRIB; -- cgit v1.2.3 From 7049ff53f640aeccc9523a103468183ffda996fd Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jan 2008 21:19:24 -0700 Subject: gallium: silence warnings --- src/mesa/state_tracker/st_cb_feedback.c | 4 ++-- src/mesa/state_tracker/st_cb_rasterpos.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index ea775b9452..43543df1a8 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -103,13 +103,13 @@ feedback_vertex(GLcontext *ctx, const struct draw_context *draw, */ slot = st->vertex_result_to_slot[VERT_RESULT_COL0]; - if (slot != ~0) + if (slot != ~0U) color = v->data[slot]; else color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; slot = st->vertex_result_to_slot[VERT_RESULT_TEX0]; - if (slot != ~0) + if (slot != ~0U) texcoord = v->data[slot]; else texcoord = ctx->Current.Attrib[VERT_ATTRIB_TEX0]; diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 852cff6490..f1c2d2d7c3 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -117,7 +117,7 @@ update_attrib(GLcontext *ctx, const GLuint *outputMapping, { const GLfloat *src; const GLuint k = outputMapping[result]; - if (k != ~0) + if (k != ~0U) src = vert->data[k]; else src = ctx->Current.Attrib[defaultAttrib]; -- cgit v1.2.3 From f1fb69a6e52260193ec16a9820a66e3e4bb03edd Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 22 Jan 2008 10:16:30 -0700 Subject: gallium: Added FORMAT_1F_PSIZE to insert constant point size into vertices Also, added draw_convert_wide_points/lines() so a driver can tell the draw module whether to convert wide points/lines into triangles, or just pass them through. --- src/mesa/pipe/draw/draw_context.c | 25 ++++++++++++++++++++++++- src/mesa/pipe/draw/draw_context.h | 4 ++++ src/mesa/pipe/draw/draw_private.h | 3 +++ src/mesa/pipe/draw/draw_validate.c | 4 ++-- src/mesa/pipe/draw/draw_vbuf.c | 11 ++++++++--- src/mesa/pipe/draw/draw_vertex.c | 2 ++ src/mesa/pipe/draw/draw_vertex.h | 5 +++-- 7 files changed, 46 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index e08873c5e1..5b9ea55630 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -71,7 +71,7 @@ struct draw_context *draw_create( void ) */ { uint i; - char *tmp = MALLOC( Elements(draw->vcache.vertex) * MAX_VERTEX_SIZE ); + char *tmp = (char*) MALLOC( Elements(draw->vcache.vertex) * MAX_VERTEX_SIZE ); for (i = 0; i < Elements(draw->vcache.vertex); i++) draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * MAX_VERTEX_SIZE); @@ -82,6 +82,9 @@ struct draw_context *draw_create( void ) draw->attrib_front1 = 0; draw->attrib_back1 = 0; + draw->convert_wide_points = TRUE; + draw->convert_wide_lines = TRUE; + draw->prim = ~0; /* != any of PIPE_PRIM_x */ draw_vertex_cache_invalidate( draw ); @@ -217,6 +220,26 @@ draw_set_mapped_constant_buffer(struct draw_context *draw, } +/** + * Tells the draw module whether to convert wide points (size != 1) + * into triangles. + */ +void +draw_convert_wide_points(struct draw_context *draw, boolean enable) +{ + draw->convert_wide_points = enable; +} + + +/** + * Tells the draw module whether to convert wide lines (width != 1) + * into triangles. + */ +void +draw_convert_wide_lines(struct draw_context *draw, boolean enable) +{ + draw->convert_wide_lines = enable; +} diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index 91e11e6930..cfde26ceb7 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -89,6 +89,10 @@ void draw_set_rasterizer_state( struct draw_context *draw, void draw_set_rasterize_stage( struct draw_context *draw, struct draw_stage *stage ); +void draw_convert_wide_points(struct draw_context *draw, boolean enable); + +void draw_convert_wide_lines(struct draw_context *draw, boolean enable); + struct draw_vertex_shader * draw_create_vertex_shader(struct draw_context *draw, diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 20503fe158..a264fabfb4 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -195,6 +195,9 @@ struct draw_context uint attrib_front0, attrib_back0; uint attrib_front1, attrib_back1; + boolean convert_wide_points; /**< convert wide points to tris? */ + boolean convert_wide_lines; /**< convert side lines to tris? */ + boolean drawing; /**< do we presently have something queued for drawing? */ unsigned prim; /**< current prim type: PIPE_PRIM_x */ unsigned reduced_prim; diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index df06e0426c..3b1f5179a9 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -51,8 +51,8 @@ static void validate_begin( struct draw_stage *stage ) * shorter pipelines for lines & points. */ - if (draw->rasterizer->line_width != 1.0 || - draw->rasterizer->point_size != 1.0 || + if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines) || + (draw->rasterizer->point_size != 1.0 && draw->convert_wide_points) || draw->rasterizer->point_sprite) { draw->pipeline.wide->next = next; next = draw->pipeline.wide; diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 7683e3381c..0e07ee7690 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -148,6 +148,10 @@ emit_vertex( struct vbuf_stage *vbuf, *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); count++; break; + case FORMAT_1F_PSIZE: + *vbuf->vertex_ptr++ = fui(vbuf->stage.draw->rasterizer->point_size); + count++; + break; case FORMAT_2F: *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); @@ -353,7 +357,7 @@ vbuf_alloc_vertices( struct draw_stage *stage, /* Allocate a new vertex buffer */ vbuf->vertex_size = new_vertex_size; vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; - vbuf->vertices = vbuf->render->allocate_vertices(vbuf->render, + vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, (ushort) vbuf->vertex_size, (ushort) vbuf->max_vertices); vbuf->vertex_ptr = vbuf->vertices; @@ -388,6 +392,7 @@ vbuf_end( struct draw_stage *stage ) static void vbuf_reset_stipple_counter( struct draw_stage *stage ) { + (void) stage; } @@ -421,8 +426,8 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, assert(render->max_indices < UNDEFINED_VERTEX_ID); vbuf->max_indices = render->max_indices; - vbuf->indices - = align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); + vbuf->indices = (ushort *) + align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); vbuf->vertices = NULL; vbuf->vertex_ptr = vbuf->vertices; diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index a1926d951a..6191fcedbf 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -65,6 +65,8 @@ draw_compute_vertex_size(struct vertex_info *vinfo) break; case FORMAT_4UB: /* fall-through */ + case FORMAT_1F_PSIZE: + /* fall-through */ case FORMAT_1F: vinfo->size += 1; break; diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index ab0425e106..e4f85bc49f 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -41,12 +41,13 @@ struct draw_context; * Vertex attribute format */ enum attrib_format { - FORMAT_OMIT, + FORMAT_OMIT, /**< don't emit the attribute */ FORMAT_1F, + FORMAT_1F_PSIZE, /**< insert constant point size */ FORMAT_2F, FORMAT_3F, FORMAT_4F, - FORMAT_4UB + FORMAT_4UB /**< XXX may need variations for RGBA vs BGRA, etc */ }; -- cgit v1.2.3 From 52264ec53d8d3b20ed7575e2ce09856cb15ab461 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 22 Jan 2008 16:35:50 -0700 Subject: added stipple_destroy() --- src/mesa/pipe/draw/draw_stipple.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_stipple.c b/src/mesa/pipe/draw/draw_stipple.c index 6d5226b1b8..3e0d5689e1 100644 --- a/src/mesa/pipe/draw/draw_stipple.c +++ b/src/mesa/pipe/draw/draw_stipple.c @@ -152,8 +152,8 @@ stipple_line(struct draw_stage *stage, struct prim_header *header) /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table. */ for (i = 0; i < length; i++) { - int result = stipple_test( stipple->counter+i, - stipple->pattern, stipple->factor ); + int result = stipple_test( (int) stipple->counter+i, + (ushort) stipple->pattern, stipple->factor ); if (result != state) { /* changing from "off" to "on" or vice versa */ if (state) { @@ -164,7 +164,7 @@ stipple_line(struct draw_stage *stage, struct prim_header *header) } else { /* starting an "on" segment */ - start = i; + start = (float) i; } state = result; } @@ -220,6 +220,13 @@ passthrough_tri(struct draw_stage *stage, struct prim_header *header) } +static void +stipple_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + /** * Create line stippler stage */ @@ -237,6 +244,7 @@ struct draw_stage *draw_stipple_stage( struct draw_context *draw ) stipple->stage.tri = passthrough_tri; stipple->stage.reset_stipple_counter = reset_stipple_counter; stipple->stage.end = stipple_end; + stipple->stage.destroy = stipple_destroy; return &stipple->stage; } -- cgit v1.2.3 From 59e8f3e5e91f27c5d12b810375247fb1ddea68fe Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 22 Jan 2008 16:37:19 -0700 Subject: gallium: Don't allocate a vertex buffer until after setting the prim type. The vertex size may depend on whether we're drawing points, lines or triangles. --- src/mesa/pipe/draw/draw_vbuf.c | 64 +++++++++++++++++++++++++----------------- 1 file changed, 39 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 0e07ee7690..a4b6247e18 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -110,10 +110,11 @@ check_space( struct vbuf_stage *vbuf, unsigned nr ) /** - * Extract the needed fields from vertex_header and emit i915 dwords. + * Extract the needed fields from post-transformed vertex and emit + * a hardware(driver) vertex. * Recall that the vertices are constructed by the 'draw' module and * have a couple of slots at the beginning (1-dword header, 4-dword - * clip pos) that we ignore here. + * clip pos) that we ignore here. We only use the vertex->data[] fields. */ static INLINE void emit_vertex( struct vbuf_stage *vbuf, @@ -233,18 +234,43 @@ vbuf_point( struct draw_stage *stage, } +/** + * Set the prim type for subsequent vertices. + * This may result in a new vertex size. The existing vbuffer (if any) + * will be flushed if needed and a new one allocated. + */ +static void +vbuf_set_prim( struct draw_stage *stage, uint newprim ) +{ + struct vbuf_stage *vbuf = vbuf_stage(stage); + const struct vertex_info *vinfo; + unsigned vertex_size; + + assert(newprim == PIPE_PRIM_POINTS || + newprim == PIPE_PRIM_LINES || + newprim == PIPE_PRIM_TRIANGLES); + + vbuf->prim = newprim; + vbuf->render->set_primitive(vbuf->render, newprim); + + vinfo = vbuf->render->get_vertex_info(vbuf->render); + vertex_size = vinfo->size * sizeof(float); + + if (vertex_size != vbuf->vertex_size) + vbuf_flush_vertices(stage); + + if (!vbuf->vertices) + vbuf_alloc_vertices(stage, vertex_size); +} + + static void vbuf_first_tri( struct draw_stage *stage, struct prim_header *prim ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - vbuf_flush_indices( stage ); - stage->tri = vbuf_tri; - vbuf->prim = PIPE_PRIM_TRIANGLES; - vbuf->render->set_primitive(vbuf->render, PIPE_PRIM_TRIANGLES); - + vbuf_set_prim(stage, PIPE_PRIM_TRIANGLES); stage->tri( stage, prim ); } @@ -253,13 +279,9 @@ static void vbuf_first_line( struct draw_stage *stage, struct prim_header *prim ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - vbuf_flush_indices( stage ); stage->line = vbuf_line; - vbuf->prim = PIPE_PRIM_LINES; - vbuf->render->set_primitive(vbuf->render, PIPE_PRIM_LINES); - + vbuf_set_prim(stage, PIPE_PRIM_LINES); stage->line( stage, prim ); } @@ -268,14 +290,9 @@ static void vbuf_first_point( struct draw_stage *stage, struct prim_header *prim ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - vbuf_flush_indices( stage ); - stage->point = vbuf_point; - vbuf->prim = PIPE_PRIM_POINTS; - vbuf->render->set_primitive(vbuf->render, PIPE_PRIM_POINTS); - + vbuf_set_prim(stage, PIPE_PRIM_POINTS); stage->point( stage, prim ); } @@ -367,12 +384,7 @@ vbuf_alloc_vertices( struct draw_stage *stage, static void vbuf_begin( struct draw_stage *stage ) { - struct vbuf_stage *vbuf = vbuf_stage(stage); - const struct vertex_info *vinfo = vbuf->render->get_vertex_info(vbuf->render); - unsigned vertex_size = vinfo->size * sizeof(float); - - /* XXX: Overkill */ - vbuf_alloc_vertices(&vbuf->stage, vertex_size); + /* no-op, vbuffer allocated by first point/line/tri */ } @@ -431,6 +443,8 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, vbuf->vertices = NULL; vbuf->vertex_ptr = vbuf->vertices; + + vbuf->prim = ~0; return &vbuf->stage; } -- cgit v1.2.3 From 286380020b146d600ac86d519ddfbf765a5965b2 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 22 Jan 2008 20:52:48 -0700 Subject: gallium: use temp reg for storing color to avoid reading 'result.color' --- src/mesa/state_tracker/st_atom_pixeltransfer.c | 65 +++++++++++++++----------- 1 file changed, 37 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index 8cf340e685..1e3effd549 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -109,6 +109,7 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) struct gl_program_parameter_list *params; struct gl_fragment_program *fp; GLuint ic = 0; + const GLuint colorTemp = 0; fp = (struct gl_fragment_program *) ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); @@ -117,11 +118,11 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) params = _mesa_new_parameter_list(); - /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ + /* TEX colorTemp, fragment.texcoord[0], texture[0], 2D; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_TEX; - inst[ic].DstReg.File = PROGRAM_OUTPUT; - inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = colorTemp; inst[ic].SrcReg[0].File = PROGRAM_INPUT; inst[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; inst[ic].TexSrcUnit = 0; @@ -130,7 +131,7 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) fp->Base.InputsRead = (1 << FRAG_ATTRIB_TEX0); fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR); - /* MAD result.color, result.color, scale, bias; */ + /* MAD colorTemp, colorTemp, scale, bias; */ if (key->scaleAndBias) { static const gl_state_index scale_state[STATE_LENGTH] = { STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 }; @@ -153,10 +154,10 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_MAD; - inst[ic].DstReg.File = PROGRAM_OUTPUT; - inst[ic].DstReg.Index = FRAG_RESULT_COLR; - inst[ic].SrcReg[0].File = PROGRAM_OUTPUT; - inst[ic].SrcReg[0].Index = FRAG_RESULT_COLR; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = colorTemp; + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = colorTemp; inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = scale_p; inst[ic].SrcReg[2].File = PROGRAM_STATE_VAR; @@ -178,67 +179,75 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) GLint row1_p = _mesa_add_state_reference(params, row1_state); GLint row2_p = _mesa_add_state_reference(params, row2_state); GLint row3_p = _mesa_add_state_reference(params, row3_state); + const GLuint temp = 1; - /* MOV temp0, result.color; */ + /* MOV temp, colorTemp; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_MOV; inst[ic].DstReg.File = PROGRAM_TEMPORARY; - inst[ic].DstReg.Index = 0; - inst[ic].SrcReg[0].File = PROGRAM_OUTPUT; - inst[ic].SrcReg[0].Index = FRAG_RESULT_COLR; + inst[ic].DstReg.Index = temp; + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = colorTemp; ic++; /* XXX reimplement in terms of MUL/MAD (see t_vp_build.c) */ - /* DP4 result.color.x, tmp0, matrow0; */ + /* DP4 colorTemp.x, temp, matrow0; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; - inst[ic].DstReg.File = PROGRAM_OUTPUT; - inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = colorTemp; inst[ic].DstReg.WriteMask = WRITEMASK_X; inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - inst[ic].SrcReg[0].Index = 0; + inst[ic].SrcReg[0].Index = temp; inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = row0_p; ic++; - /* DP4 result.color.y, tmp0, matrow1; */ + /* DP4 colorTemp.y, temp, matrow1; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; - inst[ic].DstReg.File = PROGRAM_OUTPUT; - inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = colorTemp; inst[ic].DstReg.WriteMask = WRITEMASK_Y; inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - inst[ic].SrcReg[0].Index = 0; + inst[ic].SrcReg[0].Index = temp; inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = row1_p; ic++; - /* DP4 result.color.z, tmp0, matrow2; */ + /* DP4 colorTemp.z, temp, matrow2; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; - inst[ic].DstReg.File = PROGRAM_OUTPUT; - inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = colorTemp; inst[ic].DstReg.WriteMask = WRITEMASK_Z; inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - inst[ic].SrcReg[0].Index = 0; + inst[ic].SrcReg[0].Index = temp; inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = row2_p; ic++; - /* DP4 result.color.w, tmp0, matrow3; */ + /* DP4 colorTemp.w, temp, matrow3; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; - inst[ic].DstReg.File = PROGRAM_OUTPUT; - inst[ic].DstReg.Index = FRAG_RESULT_COLR; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index =colorTemp; inst[ic].DstReg.WriteMask = WRITEMASK_W; inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - inst[ic].SrcReg[0].Index = 0; + inst[ic].SrcReg[0].Index = temp; inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = row3_p; ic++; } + /* Modify last instruction's dst reg to write to result.color */ + { + struct prog_instruction *last = &inst[ic - 1]; + last->DstReg.File = PROGRAM_OUTPUT; + last->DstReg.Index = FRAG_RESULT_COLR; + } + /* END; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_END; -- cgit v1.2.3 From 46a787ad979e16b827caad5a123b41f8f59327f1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 08:24:08 -0700 Subject: gallium: multiply vertex point size by 0.5 --- src/mesa/pipe/draw/draw_wide_prims.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index a56c9b8893..f71a38efba 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -217,7 +217,7 @@ static void wide_point( struct draw_stage *stage, /* point size is either per-vertex or fixed size */ if (wide->psize_slot >= 0) { - half_size = header->v[0]->data[wide->psize_slot][0]; + half_size = 0.5 * header->v[0]->data[wide->psize_slot][0]; } else { half_size = wide->half_point_size; -- cgit v1.2.3 From 2444f6c6a4c2f54c3198882c2256419628ca49d1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 08:24:30 -0700 Subject: gallium: fix computation of raster.point_size_per_vertex flag --- src/mesa/state_tracker/st_atom_rasterizer.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index beae36bca0..435d604af7 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -74,6 +74,7 @@ static void update_raster_state( struct st_context *st ) GLcontext *ctx = st->ctx; struct pipe_rasterizer_state raster; const struct cso_rasterizer *cso; + const struct gl_vertex_program *vertProg = ctx->VertexProgram._Current; uint i; memset(&raster, 0, sizeof(raster)); @@ -210,7 +211,19 @@ static void update_raster_state( struct st_context *st ) raster.sprite_coord_mode[i] = PIPE_SPRITE_COORD_NONE; } } - raster.point_size_per_vertex = ctx->VertexProgram.PointSizeEnabled; + if (vertProg) { + if (vertProg->Base.Id == 0) { + if (vertProg->Base.OutputsWritten & (1 << VERT_RESULT_PSIZ)) { + /* generated program which emits point size */ + raster.point_size_per_vertex = TRUE; + } + } + else if (ctx->VertexProgram.PointSizeEnabled) { + /* user-defined program and GL_VERTEX_PROGRAM_POINT_SIZE set */ + raster.point_size_per_vertex = ctx->VertexProgram.PointSizeEnabled; + } + } + /* _NEW_LINE */ -- cgit v1.2.3 From b3f081999f8eb904370df7492e39184268c77c13 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 09:27:39 -0700 Subject: gallium: s/INTERP_NONE/INTERP_LINEAR/ to fix assertion with progs/samples/select.c --- src/mesa/state_tracker/st_draw.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 94b3a9531a..8674b3edd1 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -442,7 +442,7 @@ set_feedback_vertex_format(GLcontext *ctx) assert(ctx->RenderMode == GL_SELECT); vinfo.num_attribs = 1; vinfo.format[0] = FORMAT_4F; - vinfo.interp_mode[0] = INTERP_NONE; + vinfo.interp_mode[0] = INTERP_LINEAR; } else { /* GL_FEEDBACK, or glRasterPos */ -- cgit v1.2.3 From cd3643698eafa0869a8317b002e5b066de0172e7 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 12:48:41 -0700 Subject: gallium: overhaul usage of vertex_info in draw module. Remove all dependencies on vertex_info, except for draw_vbuf. Drawing stages now strictly operate on post-transformed vertices and don't know anything about hw vertices. Use vertex program output info for two-side/flat/etc stages. Temporarily disable vbuf module in softpipe driver. --- src/mesa/pipe/draw/draw_clip.c | 4 +- src/mesa/pipe/draw/draw_context.c | 6 --- src/mesa/pipe/draw/draw_context.h | 1 - src/mesa/pipe/draw/draw_flatshade.c | 73 +++++++++++++++++---------- src/mesa/pipe/draw/draw_private.h | 14 ++--- src/mesa/pipe/draw/draw_stipple.c | 30 +++-------- src/mesa/pipe/draw/draw_twoside.c | 35 +++++++++++-- src/mesa/pipe/draw/draw_vertex.c | 70 ------------------------- src/mesa/pipe/draw/draw_vertex.h | 14 +---- src/mesa/pipe/draw/draw_vertex_cache.c | 1 - src/mesa/pipe/draw/draw_vertex_fetch.c | 1 - src/mesa/pipe/draw/draw_vertex_shader.c | 5 +- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 5 +- src/mesa/pipe/i915simple/i915_state_derived.c | 6 +-- src/mesa/pipe/softpipe/sp_context.c | 2 +- src/mesa/pipe/softpipe/sp_context.h | 2 + src/mesa/pipe/softpipe/sp_prim_setup.c | 2 +- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 5 +- src/mesa/pipe/softpipe/sp_state_derived.c | 7 ++- src/mesa/state_tracker/st_draw.c | 4 ++ 20 files changed, 117 insertions(+), 170 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index c50376f11f..e0eb656364 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -93,7 +93,7 @@ static void interp( const struct clipper *clip, const struct vertex_header *out, const struct vertex_header *in ) { - const unsigned nr_attrs = clip->stage.draw->vertex_info.num_attribs; + const unsigned nr_attrs = clip->stage.draw->num_vs_outputs; unsigned j; /* Vertex header. @@ -349,7 +349,7 @@ do_clip_line( struct draw_stage *stage, static void clip_begin( struct draw_stage *stage ) { /* should always have position, at least */ - assert(stage->draw->vertex_info.num_attribs >= 1); + assert(stage->draw->num_vs_outputs > 0); stage->next->begin( stage->next ); } diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 5b9ea55630..ff23288fa8 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -77,11 +77,6 @@ struct draw_context *draw_create( void ) draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * MAX_VERTEX_SIZE); } - draw->attrib_front0 = 0; - draw->attrib_back0 = 0; - draw->attrib_front1 = 0; - draw->attrib_back1 = 0; - draw->convert_wide_points = TRUE; draw->convert_wide_lines = TRUE; @@ -242,7 +237,6 @@ draw_convert_wide_lines(struct draw_context *draw, boolean enable) } - /** * Allocate space for temporary post-transform vertices, such as for clipping. */ diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h index cfde26ceb7..ddeb184497 100644 --- a/src/mesa/pipe/draw/draw_context.h +++ b/src/mesa/pipe/draw/draw_context.h @@ -122,7 +122,6 @@ void draw_set_mapped_constant_buffer(struct draw_context *draw, const void *buffer); - /*********************************************************************** * draw_prim.c */ diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index d7551e7948..46b0f7970f 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -29,42 +29,59 @@ */ #include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" #include "draw_private.h" -static void flatshade_begin( struct draw_stage *stage ) +/** subclass of draw_stage */ +struct flat_stage { - stage->next->begin( stage->next ); -} + struct draw_stage stage; + + uint num_color_attribs; + uint color_attribs[4]; /* front/back primary/secondary colors */ +}; +static INLINE struct flat_stage * +flat_stage(struct draw_stage *stage) +{ + return (struct flat_stage *) stage; +} + -static INLINE void copy_attr( unsigned attr, - struct vertex_header *dst, - const struct vertex_header *src ) +static void flatshade_begin( struct draw_stage *stage ) { - if (attr) { - memcpy( dst->data[attr], - src->data[attr], - sizeof(src->data[0]) ); + struct flat_stage *flat = flat_stage(stage); + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + /* Find which vertex shader outputs are colors, make a list */ + flat->num_color_attribs = 0; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + flat->color_attribs[flat->num_color_attribs++] = i; + } } + + stage->next->begin( stage->next ); } +/** Copy all the color attributes from 'src' vertex to 'dst' vertex */ static INLINE void copy_colors( struct draw_stage *stage, struct vertex_header *dst, const struct vertex_header *src ) { - const uint num_attribs = stage->draw->vertex_info.num_attribs; - const enum interp_mode *interp = stage->draw->vertex_info.interp_mode; + const struct flat_stage *flat = flat_stage(stage); uint i; /* Look for constant/flat attribs and duplicate from src to dst vertex */ /* skip attrib[0] which is vert pos */ - for (i = 1; i < num_attribs; i++) { - if (interp[i] == INTERP_CONSTANT) { - copy_attr( i, dst, src ); - } + for (i = 0; i < flat->num_color_attribs; i++) { + const uint attr = flat->color_attribs[i]; + memcpy(dst->data[attr], src->data[attr], sizeof(src->data[0])); } } @@ -139,21 +156,21 @@ static void flatshade_destroy( struct draw_stage *stage ) */ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) { - struct draw_stage *flatshade = CALLOC_STRUCT(draw_stage); + struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage); - draw_alloc_tmps( flatshade, 2 ); + draw_alloc_tmps( &flatshade->stage, 2 ); - flatshade->draw = draw; - flatshade->next = NULL; - flatshade->begin = flatshade_begin; - flatshade->point = flatshade_point; - flatshade->line = flatshade_line; - flatshade->tri = flatshade_tri; - flatshade->end = flatshade_end; - flatshade->reset_stipple_counter = flatshade_reset_stipple_counter; - flatshade->destroy = flatshade_destroy; + flatshade->stage.draw = draw; + flatshade->stage.next = NULL; + flatshade->stage.begin = flatshade_begin; + flatshade->stage.point = flatshade_point; + flatshade->stage.line = flatshade_line; + flatshade->stage.tri = flatshade_tri; + flatshade->stage.end = flatshade_end; + flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; + flatshade->stage.destroy = flatshade_destroy; - return flatshade; + return &flatshade->stage; } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index a264fabfb4..685ec4a644 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -44,8 +44,6 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "draw_vertex.h" - #include "x86/rtasm/x86sse.h" #include "pipe/tgsi/exec/tgsi_exec.h" @@ -170,6 +168,8 @@ struct draw_context struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; const struct draw_vertex_shader *vertex_shader; + uint num_vs_outputs; /**< convenience, from vertex_shader */ + /* user-space vertex data, buffers */ struct { /** vertex element/index buffer (ex: glDrawElements) */ @@ -189,12 +189,6 @@ struct draw_context float plane[12][4]; unsigned nr_planes; - /** Describes the layout of post-transformation vertices */ - struct vertex_info vertex_info; - /** Two-sided attributes: */ - uint attrib_front0, attrib_back0; - uint attrib_front1, attrib_back1; - boolean convert_wide_points; /**< convert wide points to tris? */ boolean convert_wide_lines; /**< convert side lines to tris? */ @@ -309,7 +303,9 @@ dup_vert( struct draw_stage *stage, unsigned idx ) { struct vertex_header *tmp = stage->tmp[idx]; - memcpy(tmp, vert, stage->draw->vertex_info.size * sizeof(float) ); + const uint vsize = sizeof(struct vertex_header) + + stage->draw->num_vs_outputs * 4 * sizeof(float); + memcpy(tmp, vert, vsize); tmp->vertex_id = UNDEFINED_VERTEX_ID; return tmp; } diff --git a/src/mesa/pipe/draw/draw_stipple.c b/src/mesa/pipe/draw/draw_stipple.c index 3e0d5689e1..2a47eb7be2 100644 --- a/src/mesa/pipe/draw/draw_stipple.c +++ b/src/mesa/pipe/draw/draw_stipple.c @@ -61,6 +61,7 @@ stipple_stage(struct draw_stage *stage) /** * Compute interpolated vertex attributes for 'dst' at position 't' * between 'v0' and 'v1'. + * XXX using linear interpolation for all attribs at this time. */ static void screen_interp( struct draw_context *draw, @@ -70,28 +71,13 @@ screen_interp( struct draw_context *draw, const struct vertex_header *v1 ) { uint attr; - for (attr = 0; attr < draw->vertex_info.num_attribs; attr++) { - switch (draw->vertex_info.interp_mode[attr]) { - case INTERP_NONE: - case INTERP_CONSTANT: - COPY_4FV(dst->data[attr], v0->data[attr]); - break; - case INTERP_PERSPECTIVE: - /* Fall-through */ - /* XXX special-case perspective? */ - case INTERP_LINEAR: - { - const float *val0 = v0->data[attr]; - const float *val1 = v1->data[attr]; - float *newv = dst->data[attr]; - uint i; - for (i = 0; i < 4; i++) { - newv[i] = val0[i] + t * (val1[i] - val0[i]); - } - } - break; - default: - abort(); + for (attr = 0; attr < draw->num_vs_outputs; attr++) { + const float *val0 = v0->data[attr]; + const float *val1 = v1->data[attr]; + float *newv = dst->data[attr]; + uint i; + for (i = 0; i < 4; i++) { + newv[i] = val0[i] + t * (val1[i] - val0[i]); } } } diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index c7e268f11e..da8e7bd90d 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -30,12 +30,15 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" #include "draw_private.h" struct twoside_stage { struct draw_stage stage; float sign; /**< +1 or -1 */ + uint attrib_front0, attrib_back0; + uint attrib_front1, attrib_back1; }; @@ -48,6 +51,29 @@ static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage ) static void twoside_begin( struct draw_stage *stage ) { struct twoside_stage *twoside = twoside_stage(stage); + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + twoside->attrib_front0 = 0; + twoside->attrib_front1 = 0; + twoside->attrib_back0 = 0; + twoside->attrib_back1 = 0; + + /* Find which vertex shader outputs are front/back colors */ + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { + if (vs->output_semantic_index[i] == 0) + twoside->attrib_front0 = i; + else + twoside->attrib_front1 = i; + } + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + if (vs->output_semantic_index[i] == 0) + twoside->attrib_back0 = i; + else + twoside->attrib_back1 = i; + } + } /* * We'll multiply the primitive's determinant by this sign to determine @@ -76,13 +102,12 @@ static struct vertex_header *copy_bfc( struct twoside_stage *twoside, unsigned idx ) { struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); - const struct draw_context *draw = twoside->stage.draw; - if (draw->attrib_front0 && draw->attrib_back0) { - copy_attrib(draw->attrib_front0, draw->attrib_back0, tmp); + if (twoside->attrib_front0 && twoside->attrib_back0) { + copy_attrib(twoside->attrib_front0, twoside->attrib_back0, tmp); } - if (draw->attrib_front1 && draw->attrib_back1) { - copy_attrib(draw->attrib_front1, draw->attrib_back1, tmp); + if (twoside->attrib_front1 && twoside->attrib_back1) { + copy_attrib(twoside->attrib_front1, twoside->attrib_back1, tmp); } return tmp; diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index 6191fcedbf..1ab004be10 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -38,17 +38,6 @@ #include "pipe/draw/draw_vertex.h" -static INLINE void -emit_vertex_attr(struct vertex_info *vinfo, - enum attrib_format format, enum interp_mode interp) -{ - const uint n = vinfo->num_attribs; - vinfo->interp_mode[n] = interp; - vinfo->format[n] = format; - vinfo->num_attribs++; -} - - /** * Compute the size of a vertex, in dwords/floats, to update the * vinfo->size field. @@ -86,62 +75,3 @@ draw_compute_vertex_size(struct vertex_info *vinfo) assert(vinfo->size * 4 <= MAX_VERTEX_SIZE); } - - -/** - * Tell the drawing module about the contents of post-transformation vertices. - * Note that the vertex attribute format info isn't used by 'draw'; all - * attributes are handled as float[4]. But when the driver emits vertices - * it'll use that info. - * We _do_ care about the number of attributes and their interpolation modes. - */ -void -draw_set_vertex_info( struct draw_context *draw, - const struct vertex_info *info) -{ - assert(info->interp_mode[0] == INTERP_LINEAR); /* should be vert pos */ - assert(info->num_attribs <= PIPE_MAX_SHADER_OUTPUTS); - - memcpy(&draw->vertex_info, info, sizeof(*info)); - - /* Need to know vertex size (in words) for vertex copying elsewhere. - * Four words per attribute, plus vertex header (uint) and clip - * position (float[4]). - */ - draw->vertex_info.size = draw->vertex_info.num_attribs * 4 + 5; -} - - -/** - * This function is used to tell the draw module about attributes - * (like colors) that need to be selected based on front/back face - * orientation. - * - * The logic is: - * if (polygon is back-facing) { - * vertex->attrib[front0] = vertex->attrib[back0]; - * vertex->attrib[front1] = vertex->attrib[back1]; - * } - * - * \param front0 first attrib to replace if the polygon is back-facing - * \param back0 first attrib to copy if the polygon is back-facing - * \param front1 second attrib to replace if the polygon is back-facing - * \param back1 second attrib to copy if the polygon is back-facing - * - * Pass -1 to disable two-sided attributes. - */ -void -draw_set_twoside_attributes(struct draw_context *draw, - uint front0, uint back0, - uint front1, uint back1) -{ - /* XXX we could alternately pass an array of front/back attribs if there's - * ever need for more than two. One could imagine a shader extension - * that allows arbitrary attributes to be selected based on polygon - * orientation... - */ - draw->attrib_front0 = front0; - draw->attrib_back0 = back0; - draw->attrib_front1 = front1; - draw->attrib_back1 = back1; -} diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index e4f85bc49f..28abd8ab91 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -34,9 +34,6 @@ #define DRAW_VERTEX_H -struct draw_context; - - /** * Vertex attribute format */ @@ -63,7 +60,7 @@ enum interp_mode { /** - * Information about post-transformed vertex layout. + * Information about hardware/rasterization vertex layout. */ struct vertex_info { @@ -71,7 +68,7 @@ struct vertex_info uint hwfmt[4]; /**< hardware format info for this format */ enum interp_mode interp_mode[PIPE_MAX_SHADER_OUTPUTS]; enum attrib_format format[PIPE_MAX_SHADER_OUTPUTS]; /**< FORMAT_x */ - uint src_index[PIPE_MAX_SHADER_OUTPUTS]; + uint src_index[PIPE_MAX_SHADER_OUTPUTS]; /**< map to post-xform attribs */ uint size; /**< total vertex size in dwords */ }; @@ -98,13 +95,6 @@ draw_emit_vertex_attr(struct vertex_info *vinfo, } -extern void draw_set_vertex_info( struct draw_context *draw, - const struct vertex_info *info); - -extern void draw_set_twoside_attributes(struct draw_context *draw, - uint front0, uint back0, - uint front1, uint back1); - extern void draw_compute_vertex_size(struct vertex_info *vinfo); diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index 29993f14d2..97a40b876e 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -33,7 +33,6 @@ #include "pipe/p_util.h" #include "draw_private.h" #include "draw_context.h" -#include "draw_vertex.h" void draw_vertex_cache_invalidate( struct draw_context *draw ) diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 3ca17f8829..4d64d3d4f2 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -34,7 +34,6 @@ #include "pipe/p_shader_tokens.h" #include "draw_private.h" #include "draw_context.h" -#include "draw_vertex.h" #define DRAW_DBG 0 diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index c2e038453e..5ca659dbf5 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -38,7 +38,6 @@ #endif #include "draw_private.h" #include "draw_context.h" -#include "draw_vertex.h" #include "x86/rtasm/x86sse.h" #include "pipe/llvm/gallivm.h" @@ -176,7 +175,7 @@ run_vertex_program(struct draw_context *draw, /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. */ - for (slot = 1; slot < draw->vertex_info.num_attribs; slot++) { + for (slot = 1; slot < draw->num_vs_outputs; slot++) { vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; @@ -275,6 +274,8 @@ draw_bind_vertex_shader(struct draw_context *draw, draw_flush(draw); draw->vertex_shader = dvs; + draw->num_vs_outputs = dvs->state->num_outputs; + /* specify the fragment program to interpret/execute */ tgsi_exec_machine_init(&draw->machine, draw->vertex_shader->state->tokens, diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index acd61163fa..08be1b83f8 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -33,7 +33,6 @@ #include "pipe/p_util.h" #include "draw_private.h" #include "draw_context.h" -#include "draw_vertex.h" #ifdef MESA_LLVM @@ -132,7 +131,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) gallivm_prog_exec(prog, inputs, outputs, consts, draw->vs.queue_nr, draw->vertex_shader->state->num_inputs, - draw->vertex_info.num_attribs - 2); + draw->vertex_shader->state->num_outputs); /* store machine results */ @@ -173,7 +172,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. */ - for (slot = 1; slot < draw->vertex_info.num_attribs; slot++) { + for (slot = 1; slot < draw->vs_num_outputs; slot++) { vOut->data[slot][0] = dests[slot][0]; vOut->data[slot][1] = dests[slot][1]; vOut->data[slot][2] = dests[slot][2]; diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 466c704d87..1f17de9cef 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -133,10 +133,10 @@ static void calculate_vertex_layout( struct i915_context *i915 ) /* If the attributes have changed, tell the draw module about the new * vertex layout. We'll also update the hardware vertex format info. */ - draw_set_vertex_info( i915->draw, &vinfo); + /*draw_set_vertex_info( i915->draw, &vinfo);*/ - draw_set_twoside_attributes(i915->draw, - front0, back0, front1, back1); + /*draw_set_twoside_attributes(i915->draw, + front0, back0, front1, back1);*/ /* Need to set this flag so that the LIS2/4 registers get set. * It also means the i915_update_immediate() function must be called diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 68c18e2d05..6dccedbdcb 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -328,7 +328,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, assert(softpipe->draw); softpipe->setup = sp_draw_render_stage(softpipe); - if (GETENV( "SP_VBUF" ) != NULL) { + if (0 && GETENV( "SP_VBUF" ) != NULL) { softpipe->vbuf = sp_draw_vbuf_stage(softpipe->draw, &softpipe->pipe, sp_vbuf_render); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 3537f86a61..4ce0410075 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -40,6 +40,7 @@ struct softpipe_winsys; +struct softpipe_vbuf_render; struct draw_context; struct draw_stage; struct softpipe_tile_cache; @@ -125,6 +126,7 @@ struct softpipe_context { struct draw_context *draw; struct draw_stage *setup; struct draw_stage *vbuf; + struct softpipe_vbuf_render *vbuf_render; uint current_cbuf; /**< current color buffer being written to */ diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index cdc385e4db..533a2fbd4d 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -1278,7 +1278,7 @@ void sp_vbuf_render( struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context( pipe ); struct setup_stage *setup = setup_stage( softpipe->setup ); struct prim_header prim; - unsigned vertex_size = setup->stage.draw->vertex_info.size * sizeof(float); + unsigned vertex_size = softpipe->vertex_info.size * sizeof(float); unsigned i, j; prim.det = 0; diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index 055cb19f9a..1d58e375a9 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -263,8 +263,11 @@ static void vbuf_flush_elements( struct draw_stage *stage ) static void vbuf_begin( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage(stage); + struct softpipe_context *softpipe = softpipe_context(vbuf->pipe); - vbuf->vertex_size = vbuf->draw_context->vertex_info.size * sizeof(float); + //vbuf->vertex_size = vbuf->draw_context->vertex_info.size * sizeof(float); + + vbuf->vertex_size = softpipe->vertex_info.size * sizeof(float); } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 0f1410e5de..64e86a0d86 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -122,6 +122,8 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) } } + draw_compute_vertex_size(vinfo); + softpipe->nr_frag_attrs = fs->num_inputs; /* We want these after all other attribs since they won't get passed @@ -146,10 +148,11 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) if (1/*vinfo->attr_mask != softpipe->attr_mask*/) { /*softpipe->attr_mask = vinfo->attr_mask;*/ - draw_set_vertex_info( softpipe->draw, vinfo); + /*draw_set_vertex_info( softpipe->draw, vinfo);*/ - draw_set_twoside_attributes(softpipe->draw, + /*draw_set_twoside_attributes(softpipe->draw, front0, back0, front1, back1); + */ } } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 8674b3edd1..71acf6e862 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -434,10 +434,13 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, static void set_feedback_vertex_format(GLcontext *ctx) { +#if 0 struct st_context *st = ctx->st; struct vertex_info vinfo; GLuint i; + memset(&vinfo, 0, sizeof(vinfo)); + if (ctx->RenderMode == GL_SELECT) { assert(ctx->RenderMode == GL_SELECT); vinfo.num_attribs = 1; @@ -455,6 +458,7 @@ set_feedback_vertex_format(GLcontext *ctx) } draw_set_vertex_info(st->draw, &vinfo); +#endif } -- cgit v1.2.3 From 2d37e78e636e5e1e7d5d00230e50a00f7a71e868 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 13:17:50 -0700 Subject: gallium: restore/rewrite vbuf code for softpipe Now based on the draw_vbuf code, instead being a custom one-off. Disabled by default, enable with SP_VBUF env var. --- src/mesa/pipe/draw/draw_vbuf.c | 5 + src/mesa/pipe/draw/draw_vertex.c | 3 + src/mesa/pipe/draw/draw_vertex.h | 1 + src/mesa/pipe/i915simple/i915_state_derived.c | 8 - src/mesa/pipe/softpipe/sp_context.c | 9 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 84 ------ src/mesa/pipe/softpipe/sp_prim_setup.h | 10 - src/mesa/pipe/softpipe/sp_prim_vbuf.c | 392 ++++++++++---------------- src/mesa/pipe/softpipe/sp_prim_vbuf.h | 38 +++ src/mesa/pipe/softpipe/sp_state_derived.c | 69 +---- 10 files changed, 206 insertions(+), 413 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_prim_vbuf.h (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index a4b6247e18..c179b5e9dd 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -145,6 +145,11 @@ emit_vertex( struct vbuf_stage *vbuf, case FORMAT_OMIT: /* no-op */ break; + case FORMAT_HEADER: + memcpy(vbuf->vertex_ptr, vertex, sizeof(*vertex)); + vbuf->vertex_ptr += sizeof(*vertex) / 4; + count += sizeof(*vertex) / 4; + break; case FORMAT_1F: *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); count++; diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index 1ab004be10..37d8b04ef9 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -52,6 +52,9 @@ draw_compute_vertex_size(struct vertex_info *vinfo) switch (vinfo->format[i]) { case FORMAT_OMIT: break; + case FORMAT_HEADER: + vinfo->size += sizeof(struct vertex_header) / 4; + break; case FORMAT_4UB: /* fall-through */ case FORMAT_1F_PSIZE: diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 28abd8ab91..e76a17c198 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -39,6 +39,7 @@ */ enum attrib_format { FORMAT_OMIT, /**< don't emit the attribute */ + FORMAT_HEADER, /**< The 5-byte vertex header */ FORMAT_1F, FORMAT_1F_PSIZE, /**< insert constant point size */ FORMAT_2F, diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 1f17de9cef..03a46edc55 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -130,14 +130,6 @@ static void calculate_vertex_layout( struct i915_context *i915 ) draw_compute_vertex_size(&vinfo); if (memcmp(&i915->current.vertex_info, &vinfo, sizeof(vinfo))) { - /* If the attributes have changed, tell the draw module about the new - * vertex layout. We'll also update the hardware vertex format info. - */ - /*draw_set_vertex_info( i915->draw, &vinfo);*/ - - /*draw_set_twoside_attributes(i915->draw, - front0, back0, front1, back1);*/ - /* Need to set this flag so that the LIS2/4 registers get set. * It also means the i915_update_immediate() function must be called * after this one, in i915_update_derived(). diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 6dccedbdcb..81436c6d79 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -37,6 +37,7 @@ #include "sp_context.h" #include "sp_flush.h" #include "sp_prim_setup.h" +#include "sp_prim_vbuf.h" #include "sp_state.h" #include "sp_surface.h" #include "sp_tile_cache.h" @@ -328,12 +329,8 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, assert(softpipe->draw); softpipe->setup = sp_draw_render_stage(softpipe); - if (0 && GETENV( "SP_VBUF" ) != NULL) { - softpipe->vbuf = sp_draw_vbuf_stage(softpipe->draw, - &softpipe->pipe, - sp_vbuf_render); - - draw_set_rasterize_stage(softpipe->draw, softpipe->vbuf); + if (GETENV( "SP_VBUF" ) != NULL) { + sp_init_vbuf(softpipe); } else { draw_set_rasterize_stage(softpipe->draw, softpipe->setup); diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 533a2fbd4d..9b2f10bbbb 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -1237,87 +1237,3 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) return &setup->stage; } - - -/* Recalculate det. This is only used in the test harness below: - */ -static void calc_det( struct prim_header *header ) -{ - /* Window coords: */ - const float *v0 = header->v[0]->data[0]; - const float *v1 = header->v[1]->data[0]; - const float *v2 = header->v[2]->data[0]; - - /* edge vectors e = v0 - v2, f = v1 - v2 */ - const float ex = v0[0] - v2[0]; - const float ey = v0[1] - v2[1]; - const float fx = v1[0] - v2[0]; - const float fy = v1[1] - v2[1]; - - /* det = cross(e,f).z */ - header->det = ex * fy - ey * fx; -} - - - -/** - * Render buffer of points/lines/triangles. - * Called by vbuf code when the vertex or index buffer is filled. - * - * The big issue at this point is that reset_stipple doesn't make it - * through the interface. Probably need to split primitives at reset - * stipple, perhaps using the ~0 index marker. - */ -void sp_vbuf_render( struct pipe_context *pipe, - unsigned primitive, - const ushort *elements, - unsigned nr_elements, - const void *vertex_buffer, - unsigned nr_vertices ) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - struct setup_stage *setup = setup_stage( softpipe->setup ); - struct prim_header prim; - unsigned vertex_size = softpipe->vertex_info.size * sizeof(float); - unsigned i, j; - - prim.det = 0; - prim.reset_line_stipple = 0; - prim.edgeflags = 0; - prim.pad = 0; - - setup->stage.begin( &setup->stage ); - - switch (primitive) { - case PIPE_PRIM_TRIANGLES: - for (i = 0; i < nr_elements; i += 3) { - for (j = 0; j < 3; j++) - prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + - elements[i+j] * vertex_size); - - calc_det(&prim); - setup->stage.tri( &setup->stage, &prim ); - } - break; - - case PIPE_PRIM_LINES: - for (i = 0; i < nr_elements; i += 2) { - for (j = 0; j < 2; j++) - prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + - elements[i+j] * vertex_size); - - setup->stage.line( &setup->stage, &prim ); - } - break; - - case PIPE_PRIM_POINTS: - for (i = 0; i < nr_elements; i++) { - prim.v[0] = (struct vertex_header *)((char *)vertex_buffer + - elements[i] * vertex_size); - setup->stage.point( &setup->stage, &prim ); - } - break; - } - - setup->stage.end( &setup->stage ); -} diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h index c16dc634b0..f3e8a79dd9 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.h +++ b/src/mesa/pipe/softpipe/sp_prim_setup.h @@ -76,14 +76,4 @@ sp_draw_vbuf_stage( struct draw_context *draw_context, vbuf_draw_func draw ); -extern void -sp_vbuf_render( struct pipe_context *pipe, - unsigned prim, - const ushort *elements, - unsigned nr_elements, - const void *vertex_buffer, - unsigned nr_vertices ); - - - #endif /* SP_PRIM_SETUP_H */ diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index 1d58e375a9..b76405f179 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -2,7 +2,7 @@ * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -26,302 +26,198 @@ **************************************************************************/ /** - * Build post-transformation, post-clipping vertex buffers and element - * lists by hooking into the end of the primitive pipeline and - * manipulating the vertex_id field in the vertex headers. - * - * Keith Whitwell + * Post-transform vertex buffering. This is an optional part of the + * softpipe rendering pipeline. + * Probably not desired in general, but useful for testing/debuggin. + * Enabled/Disabled with SP_VBUF env var. + * + * Authors + * Brian Paul */ #include "sp_context.h" -#include "sp_headers.h" -#include "sp_quad.h" -#include "sp_prim_setup.h" +#include "sp_prim_vbuf.h" +#include "pipe/draw/draw_context.h" #include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vertex.h" -#include "pipe/p_util.h" - -static void vbuf_flush_elements( struct draw_stage *stage ); +#include "pipe/draw/draw_vbuf.h" -#define VBUF_SIZE (64*1024) -#define IBUF_SIZE (16*1024) +#define SP_MAX_VBUF_INDEXES 1024 +#define SP_MAX_VBUF_SIZE 4096 /** - * Vertex buffer emit stage. + * Subclass of vbuf_render. */ -struct vbuf_stage { - struct draw_stage stage; /**< This must be first (base class) */ - - struct draw_context *draw_context; - struct pipe_context *pipe; - vbuf_draw_func draw; - - /* Vertices are passed in as an array of floats making up each - * attribute in turn. Will eventually convert to hardware format - * in this stage. - */ - char *vertex_map; - char *vertex_ptr; - unsigned vertex_size; - unsigned nr_vertices; - - unsigned max_vertices; - - ushort *element_map; - unsigned nr_elements; - - unsigned prim; - -}; - -/** - * Basically a cast wrapper. - */ -static INLINE struct vbuf_stage *vbuf_stage( struct draw_stage *stage ) -{ - return (struct vbuf_stage *)stage; -} - - - - -static boolean overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) +struct softpipe_vbuf_render { - unsigned long used = (unsigned long) ((char *) ptr - (char *) map); - return (used + bytes) > bufsz; -} - - -static boolean check_space( struct vbuf_stage *vbuf ) -{ - if (overflow( vbuf->vertex_map, - vbuf->vertex_ptr, - 4 * vbuf->vertex_size, - VBUF_SIZE )) - return FALSE; - - - if (vbuf->nr_elements + 4 > IBUF_SIZE / sizeof(ushort) ) - return FALSE; - - return TRUE; -} + struct vbuf_render base; + struct softpipe_context *softpipe; + uint prim; + uint vertex_size; + void *vertex_buffer; +}; -static void emit_vertex( struct vbuf_stage *vbuf, - struct vertex_header *vertex ) +/** cast wrapper */ +static struct softpipe_vbuf_render * +softpipe_vbuf_render(struct vbuf_render *vbr) { -// fprintf(stderr, "emit vertex %d to %p\n", -// vbuf->nr_vertices, vbuf->vertex_ptr); - - vertex->vertex_id = vbuf->nr_vertices++; - - //vbuf->emit_vertex( vbuf->vertex_ptr, vertex ); - - /* Note: for softpipe, the vertex includes the vertex header info - * such as clip flags and clip coords. In the future when vbuf is - * always used, we could just copy the vertex attributes/data here. - * The sp_prim_setup.c code doesn't use any of the vertex header info. - */ - memcpy(vbuf->vertex_ptr, vertex, vbuf->vertex_size); - - vbuf->vertex_ptr += vbuf->vertex_size; + return (struct softpipe_vbuf_render *) vbr; } -/** - * - */ -static void vbuf_tri( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - if (!check_space( vbuf )) - vbuf_flush_elements( stage ); - - for (i = 0; i < 3; i++) { - if (prim->v[i]->vertex_id == UNDEFINED_VERTEX_ID) - emit_vertex( vbuf, prim->v[i] ); - - vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; - } -} - - -static void vbuf_line(struct draw_stage *stage, - struct prim_header *prim) +static const struct vertex_info * +sp_vbuf_get_vertex_info(struct vbuf_render *vbr) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - if (!check_space( vbuf )) - vbuf_flush_elements( stage ); - - for (i = 0; i < 2; i++) { - if (prim->v[i]->vertex_id == UNDEFINED_VERTEX_ID) - emit_vertex( vbuf, prim->v[i] ); - - vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[i]->vertex_id; - } + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + /* XXX check for state changes? */ + return &cvbr->softpipe->vertex_info; } -static void vbuf_point(struct draw_stage *stage, - struct prim_header *prim) +static void * +sp_vbuf_allocate_vertices(struct vbuf_render *vbr, + ushort vertex_size, ushort nr_vertices) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - if (!check_space( vbuf )) - vbuf_flush_elements( stage ); - - if (prim->v[0]->vertex_id == UNDEFINED_VERTEX_ID) - emit_vertex( vbuf, prim->v[0] ); - - vbuf->element_map[vbuf->nr_elements++] = (ushort) prim->v[0]->vertex_id; + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + assert(!cvbr->vertex_buffer); + cvbr->vertex_buffer = align_malloc(vertex_size * nr_vertices, 16); + cvbr->vertex_size = vertex_size; + return cvbr->vertex_buffer; } -static void vbuf_first_tri( struct draw_stage *stage, - struct prim_header *prim ) +static void +sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, + unsigned vertex_size, unsigned vertices_used) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_elements( stage ); - stage->tri = vbuf_tri; - stage->tri( stage, prim ); - vbuf->prim = PIPE_PRIM_TRIANGLES; + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + align_free(vertices); + assert(vertices == cvbr->vertex_buffer); + cvbr->vertex_buffer = NULL; } -static void vbuf_first_line( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_elements( stage ); - stage->line = vbuf_line; - stage->line( stage, prim ); - vbuf->prim = PIPE_PRIM_LINES; -} -static void vbuf_first_point( struct draw_stage *stage, - struct prim_header *prim ) +static void +sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_elements( stage ); - stage->point = vbuf_point; - stage->point( stage, prim ); - vbuf->prim = PIPE_PRIM_POINTS; + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + cvbr->prim = prim; } - -static void vbuf_flush_elements( struct draw_stage *stage ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - if (vbuf->nr_elements) { - /* - fprintf(stderr, "%s (%d elts)\n", __FUNCTION__, vbuf->nr_elements); - */ - - /* Draw now or add to list of primitives??? - */ - vbuf->draw( vbuf->pipe, - vbuf->prim, - vbuf->element_map, - vbuf->nr_elements, - vbuf->vertex_map, - (unsigned) (vbuf->vertex_ptr - vbuf->vertex_map) / vbuf->vertex_size ); - - vbuf->nr_elements = 0; - - vbuf->vertex_ptr = vbuf->vertex_map; - vbuf->nr_vertices = 0; - - /* Reset vertex ids? Actually, want to not do that unless our - * vertex buffer is full. Would like separate - * flush-on-index-full and flush-on-vb-full, but may raise - * issues uploading vertices if the hardware wants to flush when - * we flush. - */ - draw_reset_vertex_ids( vbuf->draw_context ); +/** + * Recalculate prim's determinant. + * XXX is this needed? + */ +static void +calc_det(struct prim_header *header) +{ + /* Window coords: */ + const float *v0 = header->v[0]->data[0]; + const float *v1 = header->v[1]->data[0]; + const float *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0] - v2[0]; + const float ey = v0[1] - v2[1]; + const float fx = v1[0] - v2[0]; + const float fy = v1[1] - v2[1]; + + /* det = cross(e,f).z */ + header->det = ex * fy - ey * fx; +} + + +static void +sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + struct softpipe_context *softpipe = cvbr->softpipe; + struct draw_stage *setup = softpipe->setup; + struct prim_header prim; + unsigned vertex_size = softpipe->vertex_info.size * sizeof(float); + unsigned i, j; + void *vertex_buffer = cvbr->vertex_buffer; + + prim.det = 0; + prim.reset_line_stipple = 0; + prim.edgeflags = 0; + prim.pad = 0; + + setup->begin( setup ); + + switch (cvbr->prim) { + case PIPE_PRIM_TRIANGLES: + for (i = 0; i < nr_indices; i += 3) { + for (j = 0; j < 3; j++) + prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + + indices[i+j] * vertex_size); + + calc_det(&prim); + setup->tri( setup, &prim ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i < nr_indices; i += 2) { + for (j = 0; j < 2; j++) + prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + + indices[i+j] * vertex_size); + + setup->line( setup, &prim ); + } + break; + + case PIPE_PRIM_POINTS: + for (i = 0; i < nr_indices; i++) { + prim.v[0] = (struct vertex_header *)((char *)vertex_buffer + + indices[i] * vertex_size); + setup->point( setup, &prim ); + } + break; } - stage->tri = vbuf_first_tri; - stage->line = vbuf_first_line; - stage->point = vbuf_first_point; + setup->end( setup ); } -static void vbuf_begin( struct draw_stage *stage ) +static void +sp_vbuf_destroy(struct vbuf_render *vbr) { - struct vbuf_stage *vbuf = vbuf_stage(stage); - struct softpipe_context *softpipe = softpipe_context(vbuf->pipe); - - //vbuf->vertex_size = vbuf->draw_context->vertex_info.size * sizeof(float); - - vbuf->vertex_size = softpipe->vertex_info.size * sizeof(float); + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + cvbr->softpipe->vbuf_render = NULL; + FREE(cvbr); } -static void vbuf_end( struct draw_stage *stage ) -{ - /* Overkill. - */ - vbuf_flush_elements( stage ); -} - - -static void reset_stipple_counter( struct draw_stage *stage ) +/** + * Initialize the post-transform vertex buffer information for the given + * context. + */ +void +sp_init_vbuf(struct softpipe_context *sp) { - /* XXX: This doesn't work. - */ -} + assert(sp->draw); + sp->vbuf_render = CALLOC_STRUCT(softpipe_vbuf_render); -static void vbuf_destroy( struct draw_stage *stage ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); + sp->vbuf_render->base.max_indices = SP_MAX_VBUF_INDEXES; + sp->vbuf_render->base.max_vertex_buffer_bytes = SP_MAX_VBUF_SIZE; - FREE( vbuf->element_map ); - FREE( vbuf->vertex_map ); - FREE( stage ); -} + sp->vbuf_render->base.get_vertex_info = sp_vbuf_get_vertex_info; + sp->vbuf_render->base.allocate_vertices = sp_vbuf_allocate_vertices; + sp->vbuf_render->base.set_primitive = sp_vbuf_set_primitive; + sp->vbuf_render->base.draw = sp_vbuf_draw; + sp->vbuf_render->base.release_vertices = sp_vbuf_release_vertices; + sp->vbuf_render->base.destroy = sp_vbuf_destroy; + sp->vbuf_render->softpipe = sp; -/** - * Create a new primitive vbuf/render stage. - */ -struct draw_stage *sp_draw_vbuf_stage( struct draw_context *draw_context, - struct pipe_context *pipe, - vbuf_draw_func draw ) -{ - struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); - - vbuf->stage.begin = vbuf_begin; - vbuf->stage.point = vbuf_first_point; - vbuf->stage.line = vbuf_first_line; - vbuf->stage.tri = vbuf_first_tri; - vbuf->stage.end = vbuf_end; - vbuf->stage.reset_stipple_counter = reset_stipple_counter; - vbuf->stage.destroy = vbuf_destroy; - - vbuf->pipe = pipe; - vbuf->draw = draw; - vbuf->draw_context = draw_context; - - vbuf->element_map = MALLOC( IBUF_SIZE ); - vbuf->vertex_map = MALLOC( VBUF_SIZE ); - - vbuf->vertex_ptr = vbuf->vertex_map; - + sp->vbuf = draw_vbuf_stage(sp->draw, &sp->vbuf_render->base); - return &vbuf->stage; + draw_set_rasterize_stage(sp->draw, sp->vbuf); } diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.h b/src/mesa/pipe/softpipe/sp_prim_vbuf.h new file mode 100644 index 0000000000..1de9cc2a89 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_VBUF_H +#define SP_VBUF_H + + +struct softpipe_context; + +extern void +sp_init_vbuf(struct softpipe_context *softpipe); + + +#endif /* SP_VBUF_H */ diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 64e86a0d86..74d34c50a8 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -45,24 +45,16 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) const enum interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; - boolean emitBack0 = FALSE, emitBack1 = FALSE, emitPsize = FALSE; - uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; uint i; - int src = 0; memset(vinfo, 0, sizeof(*vinfo)); - - if (fs->input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { - /* Need Z if depth test is enabled or the fragment program uses the - * fragment position (XYZW). - */ - } - softpipe->psize_slot = -1; - /* always emit vertex pos */ - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR, src++); + if (softpipe->vbuf) { + /* softpipe's setup/rasterizer stage expects vertex to have a header */ + draw_emit_vertex_attr(vinfo, FORMAT_HEADER, INTERP_LINEAR, 0); + } /* * XXX I think we need to reconcile the vertex shader outputs with @@ -75,46 +67,38 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) switch (vs->output_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: - /* vertex programs always emit position, but might not be - * needed for fragment progs. - */ - /* no-op */ + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR, i); break; case TGSI_SEMANTIC_COLOR: if (vs->output_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); + draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, i); } else { assert(vs->output_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); + draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, i); } break; case TGSI_SEMANTIC_BCOLOR: - if (vs->output_semantic_index[i] == 0) { - emitBack0 = TRUE; - } - else { - assert(vs->output_semantic_index[i] == 1); - emitBack1 = TRUE; - } + /* no-op */ break; case TGSI_SEMANTIC_FOG: - draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE, src++); + draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE, i); break; case TGSI_SEMANTIC_PSIZE: /* XXX only emit if drawing points or front/back polygon mode * is point mode */ - emitPsize = TRUE; + softpipe->psize_slot + = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT, i); break; case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src++); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, i); break; default: @@ -125,35 +109,6 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) draw_compute_vertex_size(vinfo); softpipe->nr_frag_attrs = fs->num_inputs; - - /* We want these after all other attribs since they won't get passed - * to the fragment shader. All prior vertex output attribs should match - * up 1:1 with the fragment shader inputs. - */ - if (emitBack0) { - back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); - } - if (emitBack1) { - back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); - } - if (emitPsize) { - softpipe->psize_slot - = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT, src++); - } - - /* If the attributes have changed, tell the draw module about - * the new vertex layout. - */ - /* XXX we also need to do this when the shading mode (interp modes) change: */ - if (1/*vinfo->attr_mask != softpipe->attr_mask*/) { - /*softpipe->attr_mask = vinfo->attr_mask;*/ - - /*draw_set_vertex_info( softpipe->draw, vinfo);*/ - - /*draw_set_twoside_attributes(softpipe->draw, - front0, back0, front1, back1); - */ - } } -- cgit v1.2.3 From 336e2cb9ec52e5cf8b2fac3142d5a67e3dcd55c4 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 14:23:49 -0700 Subject: gallium: rework compute_vertex_layout() to intelligently map vs outputs to fs inputs Some follow-on simplification in prim setup is possible... --- src/mesa/pipe/softpipe/sp_prim_setup.c | 13 +++++- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 1 + src/mesa/pipe/softpipe/sp_state_derived.c | 73 ++++++++++++++++++++----------- 3 files changed, 60 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 9b2f10bbbb..4b6b7a0fd1 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -515,13 +515,22 @@ setup_fragcoord_coeff(struct setup_stage *setup) */ static void setup_tri_coefficients( struct setup_stage *setup ) { - const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + const enum interp_mode *interp; #define USE_INPUT_MAP 01 #if USE_INPUT_MAP const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; #endif uint fragSlot; + if (setup->softpipe->vertex_info.format[0] == FORMAT_HEADER) { + /* skip header, pos slots */ + interp = setup->softpipe->vertex_info.interp_mode + 2; + } + else { + /* skip pos slot */ + interp = setup->softpipe->vertex_info.interp_mode + 1; + } + /* z and w are done by linear interpolation: */ tri_linear_coeff(setup, &setup->posCoef, 0, 2); @@ -557,7 +566,7 @@ static void setup_tri_coefficients( struct setup_stage *setup ) else { #endif uint j; - switch (interp[vertSlot]) { + switch (interp[fragSlot]) { case INTERP_CONSTANT: for (j = 0; j < NUM_CHANNELS; j++) const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index b76405f179..0ab8a55d7d 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -74,6 +74,7 @@ sp_vbuf_get_vertex_info(struct vbuf_render *vbr) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); /* XXX check for state changes? */ + assert(!cvbr->softpipe->dirty ); return &cvbr->softpipe->vertex_info; } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 74d34c50a8..d399868f24 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -33,6 +33,22 @@ #include "sp_state.h" +static int +find_vs_output(const struct pipe_shader_state *vs, + uint semantic_name, + uint semantic_index) +{ + uint i; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == semantic_name && + vs->output_semantic_index[i] == semantic_index) + return i; + } + return -1; +} + + + /** * Determine which post-transform / pre-rasterization vertex attributes * we need. @@ -46,6 +62,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; uint i; + int src; memset(vinfo, 0, sizeof(*vinfo)); @@ -56,6 +73,11 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) draw_emit_vertex_attr(vinfo, FORMAT_HEADER, INTERP_LINEAR, 0); } + /* always emit pos for softpipe rasterization */ + src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR, src); + /* * XXX I think we need to reconcile the vertex shader outputs with * the fragment shader inputs here to make sure the slots line up. @@ -63,42 +85,36 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) * Or maybe do that in the state tracker? */ - for (i = 0; i < vs->num_outputs; i++) { - switch (vs->output_semantic_name[i]) { + for (i = 0; i < fs->num_inputs; i++) { + switch (fs->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR, i); + /* handled above */ break; case TGSI_SEMANTIC_COLOR: - if (vs->output_semantic_index[i] == 0) { - draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, i); - } - else { - assert(vs->output_semantic_index[i] == 1); - draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, i); - } - break; - - case TGSI_SEMANTIC_BCOLOR: - /* no-op */ + src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, + fs->input_semantic_index[i]); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src); break; case TGSI_SEMANTIC_FOG: - draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE, i); - break; - - case TGSI_SEMANTIC_PSIZE: - /* XXX only emit if drawing points or front/back polygon mode - * is point mode - */ - softpipe->psize_slot - = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT, i); + src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); +#if 1 + if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */ + src = 0; +#endif + assert(src >= 0); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src); break; case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, i); + src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, + fs->input_semantic_index[i]); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src); break; default: @@ -106,6 +122,13 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) } } + src = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); + if (src >= 0) { + softpipe->psize_slot = src; + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_CONSTANT, src); + } + + draw_compute_vertex_size(vinfo); softpipe->nr_frag_attrs = fs->num_inputs; @@ -152,7 +175,7 @@ compute_cliprect(struct softpipe_context *sp) */ void softpipe_update_derived( struct softpipe_context *softpipe ) { - if (softpipe->dirty & (SP_NEW_RASTERIZER | SP_NEW_FS)) + if (softpipe->dirty & (SP_NEW_RASTERIZER | SP_NEW_FS | SP_NEW_VS)) calculate_vertex_layout( softpipe ); if (softpipe->dirty & (SP_NEW_SCISSOR | -- cgit v1.2.3 From 1cf3c77e086d3f8b01921aae9b7309aa62b3d15e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 15:39:34 -0700 Subject: gallium: in sp_tile_cache_clear() mark all cache slots as empty/free. Fixes progs/redbook/stencil.c --- src/mesa/pipe/softpipe/sp_tile_cache.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 1974f77459..bc7e8d01e4 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -538,6 +538,7 @@ void sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue) { uint r, g, b, a; + uint pos; tc->clear_val = clearValue; @@ -576,4 +577,9 @@ sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue) /* disable the optimization */ memset(tc->clear_flags, 0, sizeof(tc->clear_flags)); #endif + + for (pos = 0; pos < NUM_ENTRIES; pos++) { + struct softpipe_cached_tile *tile = tc->entries + pos; + tile->x = tile->y = -1; + } } -- cgit v1.2.3 From 8de2331e432e2ea6f978acb6c80666da99c6c4a1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 15:45:52 -0700 Subject: gallium: remove support for separate depth/stencil buffers. Always combined now. --- src/mesa/pipe/i915simple/i915_state_emit.c | 15 +++++---------- src/mesa/pipe/i965simple/brw_misc_state.c | 2 +- src/mesa/pipe/p_state.h | 3 +-- src/mesa/pipe/softpipe/sp_clear.c | 11 +++-------- src/mesa/pipe/softpipe/sp_context.c | 19 +++++-------------- src/mesa/pipe/softpipe/sp_context.h | 6 +----- src/mesa/pipe/softpipe/sp_flush.c | 7 ++----- src/mesa/pipe/softpipe/sp_quad.c | 4 ++-- src/mesa/pipe/softpipe/sp_quad_depth_test.c | 4 ++-- src/mesa/pipe/softpipe/sp_quad_stencil.c | 4 ++-- src/mesa/pipe/softpipe/sp_state_surface.c | 10 ++++++---- src/mesa/pipe/xlib/fakeglx.c | 8 +++++--- src/mesa/state_tracker/st_atom_framebuffer.c | 15 ++++++++------- src/mesa/state_tracker/st_cb_drawpixels.c | 6 +++--- 14 files changed, 46 insertions(+), 68 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 09bf1fa2d6..8598eacc9c 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -74,14 +74,9 @@ framebuffer_size(const struct pipe_framebuffer_state *fb, *height = fb->cbufs[0]->height; return TRUE; } - else if (fb->zbuf) { - *width = fb->zbuf->width; - *height = fb->zbuf->height; - return TRUE; - } - else if (fb->sbuf) { - *width = fb->sbuf->width; - *height = fb->sbuf->height; + else if (fb->zsbuf) { + *width = fb->zsbuf->width; + *height = fb->zsbuf->height; return TRUE; } else { @@ -209,7 +204,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) if (i915->hardware_dirty & I915_HW_STATIC) { struct pipe_surface *cbuf_surface = i915->framebuffer.cbufs[0]; - struct pipe_surface *depth_surface = i915->framebuffer.zbuf; + struct pipe_surface *depth_surface = i915->framebuffer.zsbuf; if (cbuf_surface) { unsigned pitch = (cbuf_surface->pitch * cbuf_surface->cpp); @@ -251,7 +246,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) cformat = translate_format(cformat); if (depth_surface) - zformat = translate_depth_format( i915->framebuffer.zbuf->format ); + zformat = translate_depth_format( i915->framebuffer.zsbuf->format ); OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */ diff --git a/src/mesa/pipe/i965simple/brw_misc_state.c b/src/mesa/pipe/i965simple/brw_misc_state.c index 13b3b1671d..3b7f36dd1d 100644 --- a/src/mesa/pipe/i965simple/brw_misc_state.c +++ b/src/mesa/pipe/i965simple/brw_misc_state.c @@ -211,7 +211,7 @@ const struct brw_tracked_state brw_psp_urb_cbs = { */ static void upload_depthbuffer(struct brw_context *brw) { - struct pipe_surface *depth_surface = brw->attribs.FrameBuffer.zbuf; + struct pipe_surface *depth_surface = brw->attribs.FrameBuffer.zsbuf; BEGIN_BATCH(5, INTEL_BATCH_NO_CLIPRECTS); OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (5 - 2)); diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 46328d2a8f..146a479cf3 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -202,8 +202,7 @@ struct pipe_framebuffer_state unsigned num_cbufs; struct pipe_surface *cbufs[PIPE_MAX_COLOR_BUFS]; - struct pipe_surface *zbuf; /**< Z buffer */ - struct pipe_surface *sbuf; /**< Stencil buffer */ + struct pipe_surface *zsbuf; /**< Z/stencil buffer */ }; diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 5b3857145d..571f64b38d 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -53,24 +53,19 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, softpipe_update_derived(softpipe); /* not needed?? */ #endif -#if TILE_CLEAR_OPTIMIZATION - if (ps == sp_tile_cache_get_surface(softpipe->zbuf_cache)) { - sp_tile_cache_clear(softpipe->zbuf_cache, clearValue); + if (ps == sp_tile_cache_get_surface(softpipe->zsbuf_cache)) { + sp_tile_cache_clear(softpipe->zsbuf_cache, clearValue); return; } for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[i])) { sp_tile_cache_clear(softpipe->cbuf_cache[i], clearValue); - return; } } -#endif +#if !TILE_CLEAR_OPTIMIZATION /* non-cached surface */ pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); - -#if 0 - sp_clear_tile_cache(ps, clearValue); #endif } diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 81436c6d79..8dadd9aa74 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -82,9 +82,7 @@ softpipe_map_surfaces(struct softpipe_context *sp) sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); } - sp_tile_cache_map_surfaces(sp->zbuf_cache); - - sp_tile_cache_map_surfaces(sp->sbuf_cache); + sp_tile_cache_map_surfaces(sp->zsbuf_cache); } @@ -98,16 +96,12 @@ softpipe_unmap_surfaces(struct softpipe_context *sp) for (i = 0; i < sp->framebuffer.num_cbufs; i++) sp_flush_tile_cache(sp, sp->cbuf_cache[i]); - sp_flush_tile_cache(sp, sp->zbuf_cache); - sp_flush_tile_cache(sp, sp->sbuf_cache); + sp_flush_tile_cache(sp, sp->zsbuf_cache); for (i = 0; i < sp->framebuffer.num_cbufs; i++) { sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); } - - sp_tile_cache_unmap_surfaces(sp->zbuf_cache); - - sp_tile_cache_unmap_surfaces(sp->sbuf_cache); + sp_tile_cache_unmap_surfaces(sp->zsbuf_cache); } @@ -134,8 +128,7 @@ static void softpipe_destroy( struct pipe_context *pipe ) for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) sp_destroy_tile_cache(softpipe->cbuf_cache[i]); - sp_destroy_tile_cache(softpipe->zbuf_cache); - sp_destroy_tile_cache(softpipe->sbuf_cache_sep); + sp_destroy_tile_cache(softpipe->zsbuf_cache); for (i = 0; i < PIPE_MAX_SAMPLERS; i++) sp_destroy_tile_cache(softpipe->tex_cache[i]); @@ -298,9 +291,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, */ for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) softpipe->cbuf_cache[i] = sp_create_tile_cache(); - softpipe->zbuf_cache = sp_create_tile_cache(); - softpipe->sbuf_cache_sep = sp_create_tile_cache(); - softpipe->sbuf_cache = softpipe->sbuf_cache_sep; /* initial value */ + softpipe->zsbuf_cache = sp_create_tile_cache(); for (i = 0; i < PIPE_MAX_SAMPLERS; i++) softpipe->tex_cache[i] = sp_create_tile_cache(); diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index 4ce0410075..daf7dd5acc 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -131,11 +131,7 @@ struct softpipe_context { uint current_cbuf; /**< current color buffer being written to */ struct softpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS]; - struct softpipe_tile_cache *zbuf_cache; - /** Stencil buffer cache, for stencil separate from Z */ - struct softpipe_tile_cache *sbuf_cache_sep; - /** This either points to zbuf_cache or sbuf_cache_sep */ - struct softpipe_tile_cache *sbuf_cache; + struct softpipe_tile_cache *zsbuf_cache; struct softpipe_tile_cache *tex_cache[PIPE_MAX_SAMPLERS]; diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index 47b11803ce..ec6bb4a0dc 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -59,11 +59,8 @@ softpipe_flush( struct pipe_context *pipe, if (softpipe->cbuf_cache[i]) sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]); - if (softpipe->zbuf_cache) - sp_flush_tile_cache(softpipe, softpipe->zbuf_cache); - - if (softpipe->sbuf_cache) - sp_flush_tile_cache(softpipe, softpipe->sbuf_cache); + if (softpipe->zsbuf_cache) + sp_flush_tile_cache(softpipe, softpipe->zsbuf_cache); /* Need this call for hardware buffers before swapbuffers. * diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c index a10c9c3e02..6bd468a51c 100644 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ b/src/mesa/pipe/softpipe/sp_quad.c @@ -48,7 +48,7 @@ sp_build_depth_stencil( sp_push_quad_first( sp, sp->quad.stencil_test ); } else if (sp->depth_stencil->depth.enabled && - sp->framebuffer.zbuf) { + sp->framebuffer.zsbuf) { sp_push_quad_first( sp, sp->quad.depth_test ); } } @@ -58,7 +58,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) { boolean early_depth_test = sp->depth_stencil->depth.enabled && - sp->framebuffer.zbuf && + sp->framebuffer.zsbuf && !sp->depth_stencil->alpha.enabled && sp->fs->shader.output_semantic_name[0] != TGSI_SEMANTIC_POSITION; diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c index 1b8a2960af..a9a0754f27 100644 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ b/src/mesa/pipe/softpipe/sp_quad_depth_test.c @@ -53,14 +53,14 @@ void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct pipe_surface *ps = softpipe->framebuffer.zbuf; + struct pipe_surface *ps = softpipe->framebuffer.zsbuf; const enum pipe_format format = ps->format; unsigned bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ unsigned qzzzz[QUAD_SIZE]; /**< Z values from the quad */ unsigned zmask = 0; unsigned j; struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, softpipe->zbuf_cache, quad->x0, quad->y0); + = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->x0, quad->y0); assert(ps); /* shouldn't get here if there's no zbuffer */ diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c index 33740883d3..92a0da0083 100644 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ b/src/mesa/pipe/softpipe/sp_quad_stencil.c @@ -201,12 +201,12 @@ static void stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - struct pipe_surface *ps = softpipe->framebuffer.sbuf; + struct pipe_surface *ps = softpipe->framebuffer.zsbuf; unsigned func, zFailOp, zPassOp, failOp; ubyte ref, wrtMask, valMask; ubyte stencilVals[QUAD_SIZE]; struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, softpipe->sbuf_cache, quad->x0, quad->y0); + = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->x0, quad->y0); uint j; uint face = quad->facing; diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c index 4a9a28cc4d..e2c6893e9f 100644 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ b/src/mesa/pipe/softpipe/sp_state_surface.c @@ -65,17 +65,18 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, sp->framebuffer.num_cbufs = fb->num_cbufs; /* zbuf changing? */ - if (sp->framebuffer.zbuf != fb->zbuf) { + if (sp->framebuffer.zsbuf != fb->zsbuf) { /* flush old */ - sp_flush_tile_cache(sp, sp->zbuf_cache); + sp_flush_tile_cache(sp, sp->zsbuf_cache); /* assign new */ - sp->framebuffer.zbuf = fb->zbuf; + sp->framebuffer.zsbuf = fb->zsbuf; /* update cache */ - sp_tile_cache_set_surface(sp->zbuf_cache, fb->zbuf); + sp_tile_cache_set_surface(sp->zsbuf_cache, fb->zsbuf); } +#if 0 /* XXX combined depth/stencil here */ /* sbuf changing? */ @@ -98,6 +99,7 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, sp_tile_cache_set_surface(sp->sbuf_cache, fb->sbuf); } } +#endif sp->dirty |= SP_NEW_FRAMEBUFFER; } diff --git a/src/mesa/pipe/xlib/fakeglx.c b/src/mesa/pipe/xlib/fakeglx.c index 6965ee0794..902a755075 100644 --- a/src/mesa/pipe/xlib/fakeglx.c +++ b/src/mesa/pipe/xlib/fakeglx.c @@ -419,7 +419,7 @@ static XMesaVisual create_glx_visual( Display *dpy, XVisualInfo *visinfo ) { int vislevel; - GLint zBits = default_depth_bits(); + GLint zBits = 24; /*default_depth_bits();*/ GLint accBits = default_accum_bits(); GLboolean alphaFlag = default_alpha_bits() > 0; @@ -1289,7 +1289,7 @@ choose_visual( Display *dpy, int screen, const int *list, GLboolean fbConfig ) double_flag = GL_TRUE; if (vis->depth > 8) rgb_flag = GL_TRUE; - depth_size = default_depth_bits(); + depth_size = 24; /*default_depth_bits();*/ stencil_size = STENCIL_BITS; /* XXX accum??? */ } @@ -1336,7 +1336,9 @@ choose_visual( Display *dpy, int screen, const int *list, GLboolean fbConfig ) * largest depth buffer size, which is 32bits/value. Instead, we * return 16 to maintain performance with earlier versions of Mesa. */ - if (depth_size > 24) + if (stencil_size > 0) + depth_size = 24; /* if Z and stencil, always use 24+8 format */ + else if (depth_size > 24) depth_size = 32; else if (depth_size > 16) depth_size = 24; diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index aec51f5eed..3e58d49f1f 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -65,14 +65,15 @@ update_framebuffer_state( struct st_context *st ) if (strb) { strb = st_renderbuffer(strb->Base.Wrapped); assert(strb->surface); - framebuffer.zbuf = strb->surface; + framebuffer.zsbuf = strb->surface; } - - strb = st_renderbuffer(fb->Attachment[BUFFER_STENCIL].Renderbuffer); - if (strb) { - strb = st_renderbuffer(strb->Base.Wrapped); - assert(strb->surface); - framebuffer.sbuf = strb->surface; + else { + strb = st_renderbuffer(fb->Attachment[BUFFER_STENCIL].Renderbuffer); + if (strb) { + strb = st_renderbuffer(strb->Base.Wrapped); + assert(strb->surface); + framebuffer.zsbuf = strb->surface; + } } /* XXX: The memcmp is insufficient for eliminating redundant state changes, diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index eaa88d3c6c..6b44cba2e4 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -848,7 +848,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, { struct st_context *st = ctx->st; struct pipe_context *pipe = st->pipe; - struct pipe_surface *ps = st->state.framebuffer.sbuf; + struct pipe_surface *ps = st->state.framebuffer.zsbuf; const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0; GLint skipPixels; ubyte *stmap; @@ -940,13 +940,13 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, st_validate_state(st); if (format == GL_DEPTH_COMPONENT) { - ps = st->state.framebuffer.zbuf; + ps = st->state.framebuffer.zsbuf; stfp = make_fragment_shader_z(ctx->st); stvp = make_vertex_shader(ctx->st, GL_TRUE); color = ctx->Current.RasterColor; } else if (format == GL_STENCIL_INDEX) { - ps = st->state.framebuffer.sbuf; + ps = st->state.framebuffer.zsbuf; /* XXX special case - can't use texture map */ color = NULL; } -- cgit v1.2.3 From f19b2d72214fe4354d1e123ebe892255a11525c3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 16:17:49 -0700 Subject: Cell: asst. updates to sync up with today's earlier gallium changes --- src/mesa/pipe/cell/ppu/cell_clear.c | 2 +- src/mesa/pipe/cell/ppu/cell_context.h | 2 +- src/mesa/pipe/cell/ppu/cell_spu.c | 4 ++-- src/mesa/pipe/cell/ppu/cell_state_derived.c | 12 +----------- src/mesa/pipe/cell/ppu/cell_state_surface.c | 12 ++++++------ 5 files changed, 11 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_clear.c b/src/mesa/pipe/cell/ppu/cell_clear.c index 48c1c8e2c8..9f7cfdcb46 100644 --- a/src/mesa/pipe/cell/ppu/cell_clear.c +++ b/src/mesa/pipe/cell/ppu/cell_clear.c @@ -53,7 +53,7 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, if (!cell->cbuf_map[0]) cell->cbuf_map[0] = pipe_surface_map(ps); - if (ps == cell->framebuffer.zbuf) { + if (ps == cell->framebuffer.zsbuf) { surfIndex = 1; } else { diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 1937812e2d..b6c59ff596 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -82,7 +82,7 @@ struct cell_context struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS]; - ubyte *zbuf_map; + ubyte *zsbuf_map; uint dirty; diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index a21ab9ff3a..5a52af1f47 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -163,14 +163,14 @@ test_spus(struct cell_context *cell) { uint i; struct pipe_surface *csurf = cell->framebuffer.cbufs[0]; - struct pipe_surface *zsurf = cell->framebuffer.zbuf; + struct pipe_surface *zsurf = cell->framebuffer.zsbuf; printf("PPU: sleep(2)\n\n\n"); sleep(2); for (i = 0; i < cell->num_spus; i++) { cell_global.command[i].fb.color_start = cell->cbuf_map[0]; - cell_global.command[i].fb.depth_start = cell->zbuf_map; + cell_global.command[i].fb.depth_start = cell->zsbuf_map; cell_global.command[i].fb.width = csurf->width; cell_global.command[i].fb.height = csurf->height; cell_global.command[i].fb.color_format = PIPE_FORMAT_A8R8G8B8_UNORM; diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c index 1e31c11ecd..f907ce97e6 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -157,17 +157,7 @@ static void calculate_vertex_layout( struct cell_context *cell ) /* XXX we also need to do this when the shading mode (interp modes) change: */ #endif - if (1/*vinfo->attr_mask != cell->attr_mask*/) { - /*cell->attr_mask = vinfo->attr_mask;*/ - - draw_compute_vertex_size(vinfo); - draw_set_vertex_info(cell->draw, vinfo); - -#if 0 - draw_set_twoside_attributes(cell->draw, - front0, back0, front1, back1); -#endif - } + draw_compute_vertex_size(vinfo); } diff --git a/src/mesa/pipe/cell/ppu/cell_state_surface.c b/src/mesa/pipe/cell/ppu/cell_state_surface.c index 9bf3c339d2..1f7faca080 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_state_surface.c @@ -42,7 +42,7 @@ cell_set_framebuffer_state(struct pipe_context *pipe, /* XXX revisit this memcmp! */ if (memcmp(&cell->framebuffer, fb, sizeof(*fb))) { struct pipe_surface *csurf = fb->cbufs[0]; - struct pipe_surface *zsurf = fb->zbuf; + struct pipe_surface *zsurf = fb->zsbuf; uint i; /* change in fb state */ @@ -55,9 +55,9 @@ cell_set_framebuffer_state(struct pipe_context *pipe, } } - if (cell->framebuffer.zbuf && cell->zbuf_map) { - pipe_surface_unmap(cell->framebuffer.zbuf); - cell->zbuf_map = NULL; + if (cell->framebuffer.zsbuf && cell->zsbuf_map) { + pipe_surface_unmap(cell->framebuffer.zsbuf); + cell->zsbuf_map = NULL; } /* update my state */ @@ -68,7 +68,7 @@ cell_set_framebuffer_state(struct pipe_context *pipe, cell->cbuf_map[0] = pipe_surface_map(csurf); if (zsurf) - cell->zbuf_map = pipe_surface_map(zsurf); + cell->zsbuf_map = pipe_surface_map(zsurf); #if 0 for (i = 0; i < cell->num_spus; i++) { @@ -90,7 +90,7 @@ cell_set_framebuffer_state(struct pipe_context *pipe, fb->opcode = CELL_CMD_FRAMEBUFFER; fb->color_start = cell->cbuf_map[0]; fb->color_format = csurf->format; - fb->depth_start = cell->zbuf_map; + fb->depth_start = cell->zsbuf_map; fb->depth_format = zsurf ? zsurf->format : PIPE_FORMAT_NONE; fb->width = csurf->width; fb->height = csurf->height; -- cgit v1.2.3 From f5d2c1eef943bcd1c4c42dcb2b2634cdf6e243e0 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 16:19:40 -0700 Subject: Cell: remove old test code --- src/mesa/pipe/cell/ppu/cell_context.c | 11 ------ src/mesa/pipe/cell/ppu/cell_spu.c | 65 ----------------------------------- src/mesa/pipe/cell/ppu/cell_spu.h | 7 ---- 3 files changed, 83 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 3db1dbdb46..98e13d8dde 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -268,16 +268,5 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) } } - -#if 0 - test_spus(cell); -#endif - return &cell->pipe; } - - -#if 0 -/** [4] to ensure 16-byte alignment for each status word */ -uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BATCH_BUFFERS][4] ALIGN16_ATTRIB; -#endif diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 5a52af1f47..4627bc8d1f 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -133,71 +133,6 @@ cell_start_spus(struct cell_context *cell) } -/** wait for all SPUs to finish working */ -/** XXX temporary */ -void -finish_all(uint num_spus) -{ - uint i; - - for (i = 0; i < num_spus; i++) { - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FINISH); - } - for (i = 0; i < num_spus; i++) { - /* wait for mbox message */ - unsigned k; - - while (spe_out_mbox_read(cell_global.spe_contexts[i], &k, 1) < 1) - ; - - assert(k == CELL_CMD_FINISH); - } -} - - -/** - ** Send test commands (XXX temporary) - **/ -void -test_spus(struct cell_context *cell) -{ - uint i; - struct pipe_surface *csurf = cell->framebuffer.cbufs[0]; - struct pipe_surface *zsurf = cell->framebuffer.zsbuf; - - printf("PPU: sleep(2)\n\n\n"); - sleep(2); - - for (i = 0; i < cell->num_spus; i++) { - cell_global.command[i].fb.color_start = cell->cbuf_map[0]; - cell_global.command[i].fb.depth_start = cell->zsbuf_map; - cell_global.command[i].fb.width = csurf->width; - cell_global.command[i].fb.height = csurf->height; - cell_global.command[i].fb.color_format = PIPE_FORMAT_A8R8G8B8_UNORM; - cell_global.command[i].fb.depth_format = PIPE_FORMAT_Z32_UNORM; - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FRAMEBUFFER); - } - - for (i = 0; i < cell->num_spus; i++) { - cell_global.command[i].clear.value = 0xff880044; /* XXX */ - cell_global.command[i].clear.surface = 0; - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_SURFACE); - } - - finish_all(cell->num_spus); - - { - uint *b = (uint*) cell->cbuf_map[0]; - printf("PPU: Clear results: 0x%x 0x%x 0x%x 0x%x\n", - b[0], b[1000], b[2000], b[3000]); - } - - for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_EXIT); - } -} - - /** * Tell all the SPUs to stop/exit. */ diff --git a/src/mesa/pipe/cell/ppu/cell_spu.h b/src/mesa/pipe/cell/ppu/cell_spu.h index b4bfbced80..19eff94f96 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.h +++ b/src/mesa/pipe/cell/ppu/cell_spu.h @@ -75,13 +75,6 @@ extern void cell_start_spus(struct cell_context *cell); -extern void -finish_all(uint num_spus); - -extern void -test_spus(struct cell_context *cell); - - extern void cell_spu_exit(struct cell_context *cell); -- cgit v1.2.3 From b3492a2052a8877739ddf430922ed48ae30d84da Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 3 Jan 2008 15:08:53 -0800 Subject: Use HOST_CC for programs that will run on the build host --- src/mesa/x86/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/x86/Makefile b/src/mesa/x86/Makefile index 3c6a6b11c0..09481dc509 100644 --- a/src/mesa/x86/Makefile +++ b/src/mesa/x86/Makefile @@ -21,7 +21,7 @@ clean: gen_matypes: gen_matypes.c - $(CC) $(INCLUDE_DIRS) $(CFLAGS) gen_matypes.c -o gen_matypes + $(HOST_CC) $(INCLUDE_DIRS) $(HOST_CFLAGS) gen_matypes.c -o gen_matypes # need some special rules here, unfortunately matypes.h: ../main/mtypes.h ../tnl/t_context.h gen_matypes -- cgit v1.2.3 From dc1abc393206c3eb00e7cca60ce3499c993bc1f6 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 3 Jan 2008 15:11:02 -0800 Subject: Specify linker instead of letting mklib use gcc This is the final bit required to get Cell builds on x86 working. Without this, mklib tries to link using the x86 gcc instead of ppu32-gcc. --- src/mesa/Makefile | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 681c566140..b16d74bf49 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -125,6 +125,7 @@ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(LLVM_LIB) @ $(TOP)/bin/mklib -o $(GL_LIB) \ + -linker $(CC) \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ @@ -134,12 +135,14 @@ $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECTS) @ if [ "${DRIVER_DIRS}" = "osmesa" ] ; then \ $(TOP)/bin/mklib -o $(OSMESA_LIB) \ + -linker $(CC) \ -major $(MESA_MAJOR) \ -minor $(MESA_MINOR) -patch $(MESA_TINY) \ -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ $(OSMESA_LIB_DEPS) $(OSMESA16_OBJECTS) ; \ else \ $(TOP)/bin/mklib -o $(OSMESA_LIB) \ + -linker $(CC) \ -major $(MESA_MAJOR) \ -minor $(MESA_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ -- cgit v1.2.3 From 6000dcc973cdc1bd61621b432aed27d21059ae5d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 23:08:59 -0700 Subject: gallium: improved vertex layout and interpolant setup in softpipe --- src/mesa/pipe/draw/draw_vertex.h | 2 + src/mesa/pipe/softpipe/sp_context.h | 4 +- src/mesa/pipe/softpipe/sp_prim_setup.c | 217 +++++++++++++----------------- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 4 +- src/mesa/pipe/softpipe/sp_state_derived.c | 55 ++++---- 5 files changed, 123 insertions(+), 159 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index e76a17c198..7e9d8abcce 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -36,6 +36,7 @@ /** * Vertex attribute format + * XXX rename to "EMIT" */ enum attrib_format { FORMAT_OMIT, /**< don't emit the attribute */ @@ -54,6 +55,7 @@ enum attrib_format { */ enum interp_mode { INTERP_NONE, /**< never interpolate vertex header info */ + INTERP_POS, /**< special case for frag position */ INTERP_CONSTANT, INTERP_LINEAR, INTERP_PERSPECTIVE diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index daf7dd5acc..aff8c2cc5d 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -89,8 +89,8 @@ struct softpipe_context { /** Vertex format */ struct vertex_info vertex_info; - unsigned attr_mask; - unsigned nr_frag_attrs; /**< number of active fragment attribs */ + struct vertex_info vertex_info_vbuf; + int psize_slot; #if 0 diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 4b6b7a0fd1..478ecff2fb 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -85,8 +85,6 @@ struct setup_stage { struct tgsi_interp_coef posCoef; /* For Z, W */ struct quad_header quad; - uint firstFpInput; /** Semantic type of first frag input */ - struct { int left[2]; /**< [0] = row0, [1] = row1 */ int right[2]; @@ -515,22 +513,10 @@ setup_fragcoord_coeff(struct setup_stage *setup) */ static void setup_tri_coefficients( struct setup_stage *setup ) { - const enum interp_mode *interp; -#define USE_INPUT_MAP 01 -#if USE_INPUT_MAP - const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; -#endif + const struct softpipe_context *softpipe = setup->softpipe; + const struct pipe_shader_state *fs = &softpipe->fs->shader; uint fragSlot; - if (setup->softpipe->vertex_info.format[0] == FORMAT_HEADER) { - /* skip header, pos slots */ - interp = setup->softpipe->vertex_info.interp_mode + 2; - } - else { - /* skip pos slot */ - interp = setup->softpipe->vertex_info.interp_mode + 1; - } - /* z and w are done by linear interpolation: */ tri_linear_coeff(setup, &setup->posCoef, 0, 2); @@ -538,64 +524,37 @@ static void setup_tri_coefficients( struct setup_stage *setup ) /* setup interpolation for all the remaining attributes: */ - for (fragSlot = 0; fragSlot < setup->quad.nr_attrs; fragSlot++) { - /* which vertex output maps to this fragment input: */ -#if !USE_INPUT_MAP - uint vertSlot; - if (setup->firstFpInput == TGSI_SEMANTIC_POSITION) { - if (fragSlot == 0) { - setup_fragcoord_coeff(setup); - continue; - } - vertSlot = fragSlot; - } - else { - vertSlot = fragSlot + 1; - } - -#else - uint vertSlot = fs->input_map[fragSlot]; - - if (vertSlot == 0) { - /* special case: shader is reading gl_FragCoord */ - /* XXX with a new INTERP_POSITION token, we could just add a - * new case to the switch below. - */ + for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { + const uint vertSlot = softpipe->vertex_info.src_index[fragSlot]; + uint j; + + switch (softpipe->vertex_info.interp_mode[fragSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + assert(fragSlot == 0); setup_fragcoord_coeff(setup); + break; + default: + assert(0); } - else { -#endif - uint j; - switch (interp[fragSlot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - default: - /* invalid interp mode */ - /* assert(0); re-enable this and run demos/fogcoord.c ... */ - ; - } - if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - - -#if USE_INPUT_MAP + if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; } -#endif } } @@ -797,9 +756,9 @@ line_persp_coeff(struct setup_stage *setup, static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { - const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + const struct softpipe_context *softpipe = setup->softpipe; const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; - unsigned fragSlot; + uint fragSlot; /* use setup->vmin, vmax to point to vertices */ setup->vprovoke = prim->v[1]; @@ -819,34 +778,37 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) /* setup interpolation for all the remaining attributes: */ - for (fragSlot = 0; fragSlot < setup->quad.nr_attrs; fragSlot++) { - /* which vertex output maps to this fragment input: */ - uint vertSlot = fs->input_map[fragSlot]; - - if (vertSlot == 0) { - /* special case: shader is reading gl_FragCoord */ + for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { + const uint vertSlot = softpipe->vertex_info.src_index[fragSlot]; + uint j; + + switch (softpipe->vertex_info.interp_mode[fragSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + assert(fragSlot == 0); + assert(0); /* XXX fix this: */ setup_fragcoord_coeff(setup); + break; + default: + assert(0); } - else { - uint j; - switch (interp[vertSlot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - - default: - /* invalid interp mode */ - assert(0); - } + + if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; } } } @@ -1005,8 +967,8 @@ static void setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; - const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; + struct softpipe_context *softpipe = setup->softpipe; + const struct pipe_shader_state *fs = &softpipe->fs->shader; const struct vertex_header *v0 = prim->v[0]; const int sizeAttr = setup->softpipe->psize_slot; const float size @@ -1040,31 +1002,36 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const_coeff(setup, &setup->posCoef, 0, 2); const_coeff(setup, &setup->posCoef, 0, 3); - for (fragSlot = 0; fragSlot < setup->quad.nr_attrs; fragSlot++) { - /* which vertex output maps to this fragment input: */ - uint vertSlot = fs->input_map[fragSlot]; - - if (vertSlot == 0) { - /* special case: shader is reading gl_FragCoord */ + for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { + const uint vertSlot = softpipe->vertex_info.src_index[fragSlot]; + uint j; + + switch (softpipe->vertex_info.interp_mode[fragSlot]) { + case INTERP_CONSTANT: + /* fall-through */ + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + point_persp_coeff(setup, setup->vprovoke, + &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + assert(fragSlot == 0); + assert(0); /* XXX fix this: */ setup_fragcoord_coeff(setup); + break; + default: + assert(0); } - else { - uint j; - switch (interp[vertSlot]) { - case INTERP_CONSTANT: - /* fall-through */ - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - point_persp_coeff(setup, setup->vprovoke, - &setup->coef[fragSlot], vertSlot, j); - break; - default: - assert(0); - } + + if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; } } @@ -1200,9 +1167,7 @@ static void setup_begin( struct draw_stage *stage ) struct softpipe_context *sp = setup->softpipe; const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; - setup->quad.nr_attrs = setup->softpipe->nr_frag_attrs; - - setup->firstFpInput = fs->input_semantic_name[0]; + setup->quad.nr_attrs = fs->num_inputs; sp->quad.first->begin(sp->quad.first); } diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index 0ab8a55d7d..dfabae8302 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -75,7 +75,7 @@ sp_vbuf_get_vertex_info(struct vbuf_render *vbr) struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); /* XXX check for state changes? */ assert(!cvbr->softpipe->dirty ); - return &cvbr->softpipe->vertex_info; + return &cvbr->softpipe->vertex_info_vbuf; } @@ -140,7 +140,7 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) struct softpipe_context *softpipe = cvbr->softpipe; struct draw_stage *setup = softpipe->setup; struct prim_header prim; - unsigned vertex_size = softpipe->vertex_info.size * sizeof(float); + unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); unsigned i, j; void *vertex_buffer = cvbr->vertex_buffer; diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index d399868f24..4be3641d67 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -50,9 +50,9 @@ find_vs_output(const struct pipe_shader_state *vs, /** - * Determine which post-transform / pre-rasterization vertex attributes - * we need. - * Derived from: fs, setup states. + * Determine how to map vertex program outputs to fragment program inputs. + * Basically, this will be used when computing the triangle interpolation + * coefficients from the post-transform vertex attributes. */ static void calculate_vertex_layout( struct softpipe_context *softpipe ) { @@ -61,35 +61,33 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) const enum interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; + struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; uint i; int src; - memset(vinfo, 0, sizeof(*vinfo)); - - softpipe->psize_slot = -1; - if (softpipe->vbuf) { - /* softpipe's setup/rasterizer stage expects vertex to have a header */ - draw_emit_vertex_attr(vinfo, FORMAT_HEADER, INTERP_LINEAR, 0); + /* if using the post-transform vertex buffer, tell draw_vbuf to + * simply emit the whole post-xform vertex as-is: + */ + vinfo_vbuf->num_attribs = 0; + draw_emit_vertex_attr(vinfo_vbuf, FORMAT_HEADER, INTERP_NONE, 0); + for (i = 0; i < vs->num_outputs; i++) { + draw_emit_vertex_attr(vinfo_vbuf, FORMAT_4F, INTERP_NONE, i); + } + draw_compute_vertex_size(vinfo_vbuf); } - /* always emit pos for softpipe rasterization */ - src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); - assert(src >= 0); - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR, src); - /* - * XXX I think we need to reconcile the vertex shader outputs with - * the fragment shader inputs here to make sure the slots line up. - * Might just be getting lucky so far. - * Or maybe do that in the state tracker? + * Loop over fragment shader inputs, searching for the matching output + * from the vertex shader. */ - + vinfo->num_attribs = 0; for (i = 0; i < fs->num_inputs; i++) { switch (fs->input_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - /* handled above */ + src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_POS, src); break; case TGSI_SEMANTIC_COLOR: @@ -122,16 +120,13 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) } } - src = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); - if (src >= 0) { - softpipe->psize_slot = src; - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_CONSTANT, src); + softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); + if (softpipe->psize_slot >= 0) { + draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_CONSTANT, + softpipe->psize_slot); } - draw_compute_vertex_size(vinfo); - - softpipe->nr_frag_attrs = fs->num_inputs; } @@ -175,7 +170,9 @@ compute_cliprect(struct softpipe_context *sp) */ void softpipe_update_derived( struct softpipe_context *softpipe ) { - if (softpipe->dirty & (SP_NEW_RASTERIZER | SP_NEW_FS | SP_NEW_VS)) + if (softpipe->dirty & (SP_NEW_RASTERIZER | + SP_NEW_FS | + SP_NEW_VS)) calculate_vertex_layout( softpipe ); if (softpipe->dirty & (SP_NEW_SCISSOR | -- cgit v1.2.3 From 2549d79ae5342a2b97c06e3b0db9da6eb7821d2e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 23:27:31 -0700 Subject: gallium: rename FORMAT_x to EMIT_x and use EMIT_ALL to emit whole vertex as-is in vbuf code --- src/mesa/pipe/draw/draw_vbuf.c | 27 ++++++++++++++------------- src/mesa/pipe/draw/draw_vertex.c | 21 ++++++++++----------- src/mesa/pipe/draw/draw_vertex.h | 27 +++++++++++++-------------- src/mesa/pipe/i915simple/i915_prim_emit.c | 14 +++++++------- src/mesa/pipe/i915simple/i915_state_derived.c | 18 +++++++++--------- src/mesa/pipe/softpipe/sp_state_derived.c | 22 ++++++++++------------ 6 files changed, 63 insertions(+), 66 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index c179b5e9dd..82051d2e65 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -141,42 +141,43 @@ emit_vertex( struct vbuf_stage *vbuf, for (i = 0; i < vinfo->num_attribs; i++) { uint j = vinfo->src_index[i]; - switch (vinfo->format[i]) { - case FORMAT_OMIT: + switch (vinfo->emit[i]) { + case EMIT_OMIT: /* no-op */ break; - case FORMAT_HEADER: - memcpy(vbuf->vertex_ptr, vertex, sizeof(*vertex)); - vbuf->vertex_ptr += sizeof(*vertex) / 4; - count += sizeof(*vertex) / 4; - break; - case FORMAT_1F: + case EMIT_ALL: + /* just copy the whole vertex as-is to the vbuf */ + assert(i == 0); + memcpy(vbuf->vertex_ptr, vertex, vinfo->size * 4); + vbuf->vertex_ptr += vinfo->size; + return; + case EMIT_1F: *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); count++; break; - case FORMAT_1F_PSIZE: + case EMIT_1F_PSIZE: *vbuf->vertex_ptr++ = fui(vbuf->stage.draw->rasterizer->point_size); count++; break; - case FORMAT_2F: + case EMIT_2F: *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); count += 2; break; - case FORMAT_3F: + case EMIT_3F: *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); count += 3; break; - case FORMAT_4F: + case EMIT_4F: *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); *vbuf->vertex_ptr++ = fui(vertex->data[j][3]); count += 4; break; - case FORMAT_4UB: + case EMIT_4UB: *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), float_to_ubyte( vertex->data[j][1] ), float_to_ubyte( vertex->data[j][0] ), diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c index 37d8b04ef9..2d6592150f 100644 --- a/src/mesa/pipe/draw/draw_vertex.c +++ b/src/mesa/pipe/draw/draw_vertex.c @@ -49,28 +49,27 @@ draw_compute_vertex_size(struct vertex_info *vinfo) vinfo->size = 0; for (i = 0; i < vinfo->num_attribs; i++) { - switch (vinfo->format[i]) { - case FORMAT_OMIT: + switch (vinfo->emit[i]) { + case EMIT_OMIT: break; - case FORMAT_HEADER: - vinfo->size += sizeof(struct vertex_header) / 4; - break; - case FORMAT_4UB: + case EMIT_4UB: /* fall-through */ - case FORMAT_1F_PSIZE: + case EMIT_1F_PSIZE: /* fall-through */ - case FORMAT_1F: + case EMIT_1F: vinfo->size += 1; break; - case FORMAT_2F: + case EMIT_2F: vinfo->size += 2; break; - case FORMAT_3F: + case EMIT_3F: vinfo->size += 3; break; - case FORMAT_4F: + case EMIT_4F: vinfo->size += 4; break; + case EMIT_ALL: + /* fall-through */ default: assert(0); } diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index 7e9d8abcce..f20b10f8d7 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -35,18 +35,17 @@ /** - * Vertex attribute format - * XXX rename to "EMIT" + * Vertex attribute emit modes */ -enum attrib_format { - FORMAT_OMIT, /**< don't emit the attribute */ - FORMAT_HEADER, /**< The 5-byte vertex header */ - FORMAT_1F, - FORMAT_1F_PSIZE, /**< insert constant point size */ - FORMAT_2F, - FORMAT_3F, - FORMAT_4F, - FORMAT_4UB /**< XXX may need variations for RGBA vs BGRA, etc */ +enum attrib_emit { + EMIT_OMIT, /**< don't emit the attribute */ + EMIT_ALL, /**< emit whole post-xform vertex, w/ header */ + EMIT_1F, + EMIT_1F_PSIZE, /**< insert constant point size */ + EMIT_2F, + EMIT_3F, + EMIT_4F, + EMIT_4UB /**< XXX may need variations for RGBA vs BGRA, etc */ }; @@ -70,7 +69,7 @@ struct vertex_info uint num_attribs; uint hwfmt[4]; /**< hardware format info for this format */ enum interp_mode interp_mode[PIPE_MAX_SHADER_OUTPUTS]; - enum attrib_format format[PIPE_MAX_SHADER_OUTPUTS]; /**< FORMAT_x */ + enum attrib_emit emit[PIPE_MAX_SHADER_OUTPUTS]; /**< EMIT_x */ uint src_index[PIPE_MAX_SHADER_OUTPUTS]; /**< map to post-xform attribs */ uint size; /**< total vertex size in dwords */ }; @@ -85,12 +84,12 @@ struct vertex_info */ static INLINE uint draw_emit_vertex_attr(struct vertex_info *vinfo, - enum attrib_format format, enum interp_mode interp, + enum attrib_emit emit, enum interp_mode interp, uint src_index) { const uint n = vinfo->num_attribs; assert(n < PIPE_MAX_SHADER_OUTPUTS); - vinfo->format[n] = format; + vinfo->emit[n] = emit; vinfo->interp_mode[n] = interp; vinfo->src_index[n] = src_index; vinfo->num_attribs++; diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index c50a591589..f74671b39c 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -73,33 +73,33 @@ emit_hw_vertex( struct i915_context *i915, uint count = 0; /* for debug/sanity */ for (i = 0; i < vinfo->num_attribs; i++) { - switch (vinfo->format[i]) { - case FORMAT_OMIT: + switch (vinfo->emit[i]) { + case EMIT_OMIT: /* no-op */ break; - case FORMAT_1F: + case EMIT_1F: OUT_BATCH( fui(vertex->data[i][0]) ); count++; break; - case FORMAT_2F: + case EMIT_2F: OUT_BATCH( fui(vertex->data[i][0]) ); OUT_BATCH( fui(vertex->data[i][1]) ); count += 2; break; - case FORMAT_3F: + case EMIT_3F: OUT_BATCH( fui(vertex->data[i][0]) ); OUT_BATCH( fui(vertex->data[i][1]) ); OUT_BATCH( fui(vertex->data[i][2]) ); count += 3; break; - case FORMAT_4F: + case EMIT_4F: OUT_BATCH( fui(vertex->data[i][0]) ); OUT_BATCH( fui(vertex->data[i][1]) ); OUT_BATCH( fui(vertex->data[i][2]) ); OUT_BATCH( fui(vertex->data[i][3]) ); count += 4; break; - case FORMAT_4UB: + case EMIT_4UB: OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[i][2] ), float_to_ubyte( vertex->data[i][1] ), float_to_ubyte( vertex->data[i][0] ), diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 03a46edc55..62741e30f8 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -56,7 +56,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) memset(&vinfo, 0, sizeof(vinfo)); /* pos */ - draw_emit_vertex_attr(&vinfo, FORMAT_3F, INTERP_LINEAR, src++); + draw_emit_vertex_attr(&vinfo, EMIT_3F, INTERP_LINEAR, src++); /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ for (i = 0; i < fs->num_inputs; i++) { @@ -65,12 +65,12 @@ static void calculate_vertex_layout( struct i915_context *i915 ) break; case TGSI_SEMANTIC_COLOR: if (fs->input_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(&vinfo, FORMAT_4UB, colorInterp, src++); + front0 = draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src++); vinfo.hwfmt[0] |= S4_VFMT_COLOR; } else { assert(fs->input_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(&vinfo, FORMAT_4UB, colorInterp, src++); + front1 = draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src++); vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG; } break; @@ -80,7 +80,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) const uint unit = fs->input_semantic_index[i]; uint hwtc; texCoords[unit] = TRUE; - draw_emit_vertex_attr(&vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src++); + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); hwtc = TEXCOORDFMT_4D; needW = TRUE; vinfo.hwfmt[1] |= hwtc << (unit * 4); @@ -88,7 +88,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) break; case TGSI_SEMANTIC_FOG: fprintf(stderr, "i915 fogcoord not implemented yet\n"); - draw_emit_vertex_attr(&vinfo, FORMAT_1F, INTERP_PERSPECTIVE, src++); + draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_PERSPECTIVE, src++); break; default: assert(0); @@ -107,11 +107,11 @@ static void calculate_vertex_layout( struct i915_context *i915 ) /* go back and fill in the vertex position info now that we have needW */ if (needW) { vinfo.hwfmt[0] |= S4_VFMT_XYZW; - vinfo.format[0] = FORMAT_4F; + vinfo.emit[0] = EMIT_4F; } else { vinfo.hwfmt[0] |= S4_VFMT_XYZ; - vinfo.format[0] = FORMAT_3F; + vinfo.emit[0] = EMIT_3F; } /* Additional attributes required for setup: Just twosided @@ -120,10 +120,10 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ if (i915->rasterizer->light_twoside) { if (front0) { - back0 = draw_emit_vertex_attr(&vinfo, FORMAT_OMIT, colorInterp, src++); + back0 = draw_emit_vertex_attr(&vinfo, EMIT_OMIT, colorInterp, src++); } if (back0) { - back1 = draw_emit_vertex_attr(&vinfo, FORMAT_OMIT, colorInterp, src++); + back1 = draw_emit_vertex_attr(&vinfo, EMIT_OMIT, colorInterp, src++); } } diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 4be3641d67..39c3e1afe1 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -29,6 +29,7 @@ #include "pipe/p_shader_tokens.h" #include "pipe/draw/draw_context.h" #include "pipe/draw/draw_vertex.h" +#include "pipe/draw/draw_private.h" #include "sp_context.h" #include "sp_state.h" @@ -61,20 +62,16 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) const enum interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; - struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; uint i; - int src; if (softpipe->vbuf) { /* if using the post-transform vertex buffer, tell draw_vbuf to * simply emit the whole post-xform vertex as-is: */ + struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; vinfo_vbuf->num_attribs = 0; - draw_emit_vertex_attr(vinfo_vbuf, FORMAT_HEADER, INTERP_NONE, 0); - for (i = 0; i < vs->num_outputs; i++) { - draw_emit_vertex_attr(vinfo_vbuf, FORMAT_4F, INTERP_NONE, i); - } - draw_compute_vertex_size(vinfo_vbuf); + draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); + vinfo_vbuf->size = 4 * vs->num_outputs + sizeof(struct vertex_header)/4; } /* @@ -83,18 +80,19 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) */ vinfo->num_attribs = 0; for (i = 0; i < fs->num_inputs; i++) { + int src; switch (fs->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); assert(src >= 0); - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_POS, src); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); break; case TGSI_SEMANTIC_COLOR: src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, fs->input_semantic_index[i]); assert(src >= 0); - draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src); + draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); break; case TGSI_SEMANTIC_FOG: @@ -104,7 +102,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) src = 0; #endif assert(src >= 0); - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; case TGSI_SEMANTIC_GENERIC: @@ -112,7 +110,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, fs->input_semantic_index[i]); assert(src >= 0); - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; default: @@ -122,7 +120,7 @@ static void calculate_vertex_layout( struct softpipe_context *softpipe ) softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); if (softpipe->psize_slot >= 0) { - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_CONSTANT, + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, softpipe->psize_slot); } -- cgit v1.2.3 From 11349b6a5f648bbad8765bf744a8e42b0f5e5f50 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 23 Jan 2008 23:33:35 -0700 Subject: Cell: s/FORMAT/EMIT/ --- src/mesa/pipe/cell/ppu/cell_state_derived.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c index f907ce97e6..f236224bce 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -69,10 +69,10 @@ static void calculate_vertex_layout( struct cell_context *cell ) #endif /* always emit vertex pos */ - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_LINEAR, src++); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_LINEAR, src++); #if 1 - front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); + front0 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); #endif #if 0 @@ -95,11 +95,11 @@ static void calculate_vertex_layout( struct cell_context *cell ) case TGSI_SEMANTIC_COLOR: if (vs->output_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); + front0 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); } else { assert(vs->output_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); + front1 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); } break; @@ -114,7 +114,7 @@ static void calculate_vertex_layout( struct cell_context *cell ) break; case TGSI_SEMANTIC_FOG: - draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_PERSPECTIVE, src++); + draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src++); break; case TGSI_SEMANTIC_PSIZE: @@ -126,7 +126,7 @@ static void calculate_vertex_layout( struct cell_context *cell ) case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ - draw_emit_vertex_attr(vinfo, FORMAT_4F, INTERP_PERSPECTIVE, src++); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); break; default: @@ -141,14 +141,14 @@ static void calculate_vertex_layout( struct cell_context *cell ) * up 1:1 with the fragment shader inputs. */ if (emitBack0) { - back0 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); + back0 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); } if (emitBack1) { - back1 = draw_emit_vertex_attr(vinfo, FORMAT_4F, colorInterp, src++); + back1 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); } if (emitPsize) { cell->psize_slot - = draw_emit_vertex_attr(vinfo, FORMAT_1F, INTERP_CONSTANT, src++); + = draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_CONSTANT, src++); } /* If the attributes have changed, tell the draw module about -- cgit v1.2.3 From bac991d6452a12a9c6cb85293a17448540761ca7 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 24 Jan 2008 07:49:36 -0500 Subject: change the name of the shaders to reflect what they do. remove stale code and do some general cleanups. --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 2 +- src/mesa/pipe/llvm/Makefile | 2 +- src/mesa/pipe/llvm/gallivm.cpp | 11 +++- src/mesa/pipe/llvm/gallivm.h | 4 +- src/mesa/pipe/llvm/llvm_entry.c | 89 ---------------------------- 5 files changed, 12 insertions(+), 96 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 08be1b83f8..4228c4f388 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -172,7 +172,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. */ - for (slot = 1; slot < draw->vs_num_outputs; slot++) { + for (slot = 1; slot < draw->num_vs_outputs; slot++) { vOut->data[slot][0] = dests[slot][0]; vOut->data[slot][1] = dests[slot][1]; vOut->data[slot][2] = dests[slot][2]; diff --git a/src/mesa/pipe/llvm/Makefile b/src/mesa/pipe/llvm/Makefile index b1463f67cf..f655fb8340 100644 --- a/src/mesa/pipe/llvm/Makefile +++ b/src/mesa/pipe/llvm/Makefile @@ -61,7 +61,7 @@ gallivm_builtins.cpp: llvm_builtins.c clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins llvm_base_shader.cpp: llvm_entry.c - clang --emit-llvm $< |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -o=$@ -funcname=createBaseShader + clang --emit-llvm $< |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -o=$@ -f -funcname=createBaseShader # Emacs tags tags: diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index 49bbf753c4..afa1446890 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -733,7 +733,11 @@ tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) unsigned instno = 0; Function* shader = mod->getFunction("execute_shader"); std::ostringstream stream; - stream << "execute_shader"; + if (prog->type == GALLIVM_VS) { + stream << "vs_shader"; + } else { + stream << "fs_shader"; + } stream << prog->id; std::string func_name = stream.str(); shader->setName(func_name.c_str()); @@ -799,6 +803,7 @@ gallivm_from_tgsi(const struct tgsi_token *tokens, enum gallivm_shader_type type struct gallivm_prog *gallivm = (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); gallivm->id = GLOBAL_ID; + gallivm->type = type; tgsi_dump(tokens, 0); llvm::Module *mod = tgsi_to_llvm(gallivm, tokens); @@ -812,7 +817,6 @@ gallivm_from_tgsi(const struct tgsi_token *tokens, enum gallivm_shader_type type passes.run(*mod); gallivm->module = mod; - gallivm->type = type; gallivm_prog_dump(gallivm, 0); @@ -966,7 +970,8 @@ void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) const llvm::Function &func = (*itr); std::string name = func.getName(); const llvm::Function *found = 0; - if (name.find("execute_shader") != std::string::npos || + if (name.find("vs_shader") != std::string::npos || + name.find("fs_shader") != std::string::npos || name.find("function") != std::string::npos) found = &func; if (found) { diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h index fd9a11e5b6..4695de3127 100644 --- a/src/mesa/pipe/llvm/gallivm.h +++ b/src/mesa/pipe/llvm/gallivm.h @@ -30,8 +30,8 @@ * Zack Rusin zack@tungstengraphics.com */ -#ifndef LLVMTGSI_H -#define LLVMTGSI_H +#ifndef GALLIVM_H +#define GALLIVM_H #if defined __cplusplus extern "C" { diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index 909bef340a..c3b34584e1 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -34,95 +34,6 @@ /* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm-dis */ typedef __attribute__(( ocu_vector_type(4) )) float float4; -#if 0 -//clang doesn't suppoer "struct->member" notation yet -struct vertex_header { - unsigned clipmask:12; - unsigned edgeflag:1; - unsigned pad:3; - unsigned vertex_id:16; - - float clip[4]; - - float data[][4]; -}; - -inline float -dot4(float4 a, float4 b) -{ - float4 c = a*b; - return c.x + c.y + c.z + c.w; -} - -inline unsigned -compute_clipmask(float4 clip, float4 (*plane), unsigned nr) -{ - unsigned mask = 0; - unsigned i; - - for (i = 0; i < nr; i++) { - if (dot4(clip, plane[i]) < 0) - mask |= (1<clip; - x = clip[0] = res0.x; - y = clip[1] = res0.y; - z = clip[2] = res0.z; - w = clip[3] = res0.w; - vOut->clipmask = compute_clipmask(res0, planes, nr_planes); - vOut->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - res0.x = x; res0.y = y; res0.z = z; res0.w = 1; - - /* Viewport mapping */ - res0 = res0 * scale + trans; - vOut->data[0][0] = res0.x; - vOut->data[0][1] = res0.y; - vOut->data[0][2] = res0.z; - vOut->data[0][3] = w; - - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - * Skip 0 since we just did it above. - * Subtract two because of the VERTEX_HEADER, CLIP_POS attribs. - */ - for (slot = 1; slot < num_attribs - 2; slot++) { - float4 vec = results[slot]; - vOut->data[slot][0] = vec.x; - vOut->data[slot][1] = vec.y; - vOut->data[slot][2] = vec.z; - vOut->data[slot][3] = vec.w; - - printf("output %d: %f %f %f %f\n", slot, - vOut->data[slot][0], - vOut->data[slot][1], - vOut->data[slot][2], - vOut->data[slot][3]); - } -} -#endif - void from_array(float4 (*res)[16], float (*ainputs)[16][4], int count, int num_attribs) { -- cgit v1.2.3 From 6e2c3b3a494f7dd44e7718ee1c006848cd835100 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 24 Jan 2008 08:29:19 -0500 Subject: disable usage of llvm in fragment shaders it produces wrong results because it hasn't been adjusted to some new changes and it will just be in the way while changing llvm code to a different vector layout --- src/mesa/pipe/softpipe/sp_quad_fs.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 0001c76a80..c9cc8afa0c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -356,7 +356,12 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) qss->stage.softpipe = softpipe; qss->stage.begin = shade_begin; #ifdef MESA_LLVM + /* disable until ported to accept + * x/y and soa layout qss->stage.run = shade_quad_llvm; + */ + softpipe->use_sse = FALSE; + qss->stage.run = shade_quad; #else qss->stage.run = shade_quad; #endif -- cgit v1.2.3 From 47b3afc53f25b1374afe38664c90404bce42ec22 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 Jan 2008 09:18:15 -0700 Subject: Cell: use Cell driver by default if linux-cell is built To disable the Cell driver and use softpipe instead, export GALLIUM_NOCELL=1 --- src/mesa/pipe/xlib/xm_winsys.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 10dc09b13c..432431aca1 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -405,7 +405,7 @@ xm_surface_alloc(struct pipe_winsys *ws) xms->surface.winsys = ws; #ifdef GALLIUM_CELL - if (getenv("GALLIUM_CELL")) { + if (!getenv("GALLIUM_NOCELL")) { xms->tileSize = 32; /** probably temporary */ } #endif @@ -511,7 +511,7 @@ xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) struct pipe_context *pipe; #ifdef GALLIUM_CELL - if (getenv("GALLIUM_CELL")) { + if (!getenv("GALLIUM_NOCELL")) { struct cell_winsys *cws = cell_get_winsys(pixelformat); pipe = cell_create_context(pws, cws); if (pipe) -- cgit v1.2.3 From 6c29be9f41470b22c87ef571e10b5a9824800608 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 Jan 2008 09:59:39 -0700 Subject: i915: add case for TGSI_FILE_IMMEDIATE to avoid error, but still needs proper handling... --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index d517b88acc..0185512aeb 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -200,6 +200,9 @@ src_vector(struct i915_fp_compile *p, } break; + case TGSI_FILE_IMMEDIATE: + /* XXX unfinished - need to append immediates onto const buffer */ + /* fall-through */ case TGSI_FILE_CONSTANT: src = UREG(REG_TYPE_CONST, index); break; @@ -928,9 +931,7 @@ i915_translate_instructions(struct i915_fp_compile *p, break; case TGSI_TOKEN_TYPE_IMMEDIATE: - /* This is a no-op. We'll get immediates from the usual constant/ - * uniform buffer. - */ + /* XXX append the immediate to the const buffer... */ break; case TGSI_TOKEN_TYPE_INSTRUCTION: -- cgit v1.2.3 From c753e7adde9a7a2c8ff772fe8e2a42084c5966e0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 Jan 2008 16:05:33 -0700 Subject: gallium: added rastpos_destroy() --- src/mesa/state_tracker/st_cb_rasterpos.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index f1c2d2d7c3..dc6456c775 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -104,6 +104,12 @@ rastpos_line( struct draw_stage *stage, struct prim_header *prim ) assert(0); } +static void +rastpos_destroy(struct draw_stage *stage) +{ + free(stage); +} + /** * Update a raster pos attribute from the vertex result if it's present, @@ -183,6 +189,7 @@ new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw) rs->stage.tri = rastpos_tri; rs->stage.end = rastpos_end; rs->stage.reset_stipple_counter = rastpos_reset_stipple_counter; + rs->stage.destroy = rastpos_destroy; rs->ctx = ctx; for (i = 0; i < VERT_ATTRIB_MAX; i++) { -- cgit v1.2.3 From 95ff06d610b976dbd88f78a0d209916c463dda91 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 Jan 2008 16:39:34 -0700 Subject: Cell: fix asst. warnings --- src/mesa/pipe/cell/ppu/cell_batch.h | 2 ++ src/mesa/pipe/cell/ppu/cell_clear.c | 3 ++- src/mesa/pipe/cell/ppu/cell_clear.h | 6 +++--- src/mesa/pipe/cell/ppu/cell_context.c | 2 +- src/mesa/pipe/cell/ppu/cell_flush.c | 1 + 5 files changed, 9 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_batch.h b/src/mesa/pipe/cell/ppu/cell_batch.h index 7a5c6392d7..47e3287626 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.h +++ b/src/mesa/pipe/cell/ppu/cell_batch.h @@ -29,6 +29,8 @@ #ifndef CELL_BATCH_H #define CELL_BATCH_H +#include "pipe/p_compiler.h" + struct cell_context; diff --git a/src/mesa/pipe/cell/ppu/cell_clear.c b/src/mesa/pipe/cell/ppu/cell_clear.c index 9f7cfdcb46..e01640b994 100644 --- a/src/mesa/pipe/cell/ppu/cell_clear.c +++ b/src/mesa/pipe/cell/ppu/cell_clear.c @@ -39,6 +39,7 @@ #include "cell_clear.h" #include "cell_context.h" #include "cell_batch.h" +#include "cell_flush.h" #include "cell_spu.h" @@ -47,7 +48,7 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct cell_context *cell = cell_context(pipe); - uint i; + /*uint i;*/ uint surfIndex; if (!cell->cbuf_map[0]) diff --git a/src/mesa/pipe/cell/ppu/cell_clear.h b/src/mesa/pipe/cell/ppu/cell_clear.h index 4c69c4c89f..ff47d43f4c 100644 --- a/src/mesa/pipe/cell/ppu/cell_clear.h +++ b/src/mesa/pipe/cell/ppu/cell_clear.h @@ -26,8 +26,8 @@ **************************************************************************/ -#ifndef CELL_SURFACE_H -#define CELL_SURFACE_H +#ifndef CELL_CLEAR_H +#define CELL_CLEAR_H struct pipe_context; @@ -40,4 +40,4 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, -#endif /* CELL_SURFACE_H */ +#endif /* CELL_CLEAR_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 98e13d8dde..8cb0c48f40 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -56,7 +56,7 @@ static boolean cell_is_format_supported( struct pipe_context *pipe, enum pipe_format format, uint type ) { - struct cell_context *cell = cell_context( pipe ); + /*struct cell_context *cell = cell_context( pipe );*/ switch (type) { case PIPE_TEXTURE: diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c index 33cbe2a085..d54eeb8524 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -27,6 +27,7 @@ #include "cell_context.h" +#include "cell_batch.h" #include "cell_flush.h" #include "cell_spu.h" #include "cell_render.h" -- cgit v1.2.3 From 0673c571c79d00d05328bf41e9c6a52bfb0f7324 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 Jan 2008 16:41:03 -0700 Subject: Cell: rewrite compute_vertex_layout() --- src/mesa/pipe/cell/ppu/cell_state_derived.c | 149 +++++++++++----------------- 1 file changed, 60 insertions(+), 89 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c index f236224bce..f3b3ae8544 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -31,102 +31,93 @@ #include "pipe/draw/draw_vertex.h" #include "cell_context.h" #include "cell_state.h" +#include "cell_state_emit.h" + + +static int +find_vs_output(const struct pipe_shader_state *vs, + uint semantic_name, + uint semantic_index) +{ + uint i; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == semantic_name && + vs->output_semantic_index[i] == semantic_index) + return i; + } + return -1; +} /** - * Determine which post-transform / pre-rasterization vertex attributes - * we need. - * Derived from: fs, setup states. + * Determine how to map vertex program outputs to fragment program inputs. + * Basically, this will be used when computing the triangle interpolation + * coefficients from the post-transform vertex attributes. */ -static void calculate_vertex_layout( struct cell_context *cell ) +static void +calculate_vertex_layout( struct cell_context *cell ) { -#if 0 - const struct pipe_shader_state *vs = cell->vs->state; + const struct pipe_shader_state *vs = &cell->vs->shader; const struct pipe_shader_state *fs = &cell->fs->shader; const enum interp_mode colorInterp = cell->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &cell->vertex_info; - boolean emitBack0 = FALSE, emitBack1 = FALSE, emitPsize = FALSE; - uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; uint i; -#endif - const enum interp_mode colorInterp - = cell->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; - struct vertex_info *vinfo = &cell->vertex_info; - uint front0; - uint src = 0; - - memset(vinfo, 0, sizeof(*vinfo)); + int src; #if 0 - if (fs->input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { - /* Need Z if depth test is enabled or the fragment program uses the - * fragment position (XYZW). + if (cell->vbuf) { + /* if using the post-transform vertex buffer, tell draw_vbuf to + * simply emit the whole post-xform vertex as-is: */ + struct vertex_info *vinfo_vbuf = &cell->vertex_info_vbuf; + vinfo_vbuf->num_attribs = 0; + draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); + vinfo_vbuf->size = 4 * vs->num_outputs + sizeof(struct vertex_header)/4; } - - cell->psize_slot = -1; #endif - /* always emit vertex pos */ - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_LINEAR, src++); + /* reset vinfo */ + vinfo->num_attribs = 0; -#if 1 - front0 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); -#endif + /* we always want to emit vertex pos */ + src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); -#if 0 /* - * XXX I think we need to reconcile the vertex shader outputs with - * the fragment shader inputs here to make sure the slots line up. - * Might just be getting lucky so far. - * Or maybe do that in the state tracker? + * Loop over fragment shader inputs, searching for the matching output + * from the vertex shader. */ - - for (i = 0; i < vs->num_outputs; i++) { - switch (vs->output_semantic_name[i]) { - + for (i = 0; i < fs->num_inputs; i++) { + switch (fs->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: - /* vertex programs always emit position, but might not be - * needed for fragment progs. - */ - /* no-op */ + /* already done above */ break; case TGSI_SEMANTIC_COLOR: - if (vs->output_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); - } - else { - assert(vs->output_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); - } - break; - - case TGSI_SEMANTIC_BCOLOR: - if (vs->output_semantic_index[i] == 0) { - emitBack0 = TRUE; - } - else { - assert(vs->output_semantic_index[i] == 1); - emitBack1 = TRUE; - } + src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, + fs->input_semantic_index[i]); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); break; case TGSI_SEMANTIC_FOG: - draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src++); - break; - - case TGSI_SEMANTIC_PSIZE: - /* XXX only emit if drawing points or front/back polygon mode - * is point mode - */ - emitPsize = TRUE; + src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); +#if 1 + if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */ + src = 0; +#endif + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src); break; case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); + src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, + fs->input_semantic_index[i]); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; default: @@ -134,29 +125,6 @@ static void calculate_vertex_layout( struct cell_context *cell ) } } - cell->nr_frag_attrs = fs->num_inputs; - - /* We want these after all other attribs since they won't get passed - * to the fragment shader. All prior vertex output attribs should match - * up 1:1 with the fragment shader inputs. - */ - if (emitBack0) { - back0 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); - } - if (emitBack1) { - back1 = draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src++); - } - if (emitPsize) { - cell->psize_slot - = draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_CONSTANT, src++); - } - - /* If the attributes have changed, tell the draw module about - * the new vertex layout. - */ - /* XXX we also need to do this when the shading mode (interp modes) change: */ -#endif - draw_compute_vertex_size(vinfo); } @@ -198,9 +166,12 @@ compute_cliprect(struct cell_context *sp) #endif + void cell_update_derived( struct cell_context *cell ) { - if (cell->dirty & (CELL_NEW_RASTERIZER | CELL_NEW_FS)) + if (cell->dirty & (CELL_NEW_RASTERIZER | + CELL_NEW_FS | + CELL_NEW_VS)) calculate_vertex_layout( cell ); #if 0 -- cgit v1.2.3 From fa82d863684a89ff8c56c3b0196488c15024f743 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 Jan 2008 16:47:23 -0700 Subject: Cell: send actual vertex size in the render command --- src/mesa/pipe/cell/common.h | 6 ++++-- src/mesa/pipe/cell/ppu/cell_render.c | 2 +- src/mesa/pipe/cell/ppu/cell_vbuf.c | 26 ++++---------------------- src/mesa/pipe/cell/spu/spu_main.c | 28 +++++++++++++--------------- 4 files changed, 22 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 4c998722a2..cd2b61e6e9 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -96,12 +96,14 @@ struct cell_command_clear_surface #define CELL_MAX_VBUF_SIZE (16 * 1024) #define CELL_MAX_VBUF_INDEXES 1024 -#define CELL_MAX_ATTRIBS 2 /* temporary! */ + + struct cell_command_render { uint opcode; uint prim_type; - uint num_verts, num_attribs; + uint num_verts; + uint vertex_size; /**< bytes per vertex */ uint num_indexes; const void *vertex_data; const ushort *index_data; diff --git a/src/mesa/pipe/cell/ppu/cell_render.c b/src/mesa/pipe/cell/ppu/cell_render.c index ecdd47e28b..4ab277a4b2 100644 --- a/src/mesa/pipe/cell/ppu/cell_render.c +++ b/src/mesa/pipe/cell/ppu/cell_render.c @@ -152,7 +152,7 @@ cell_flush_prim_buffer(struct cell_context *cell) struct cell_command_render *render = &cell_global.command[i].render; render->prim_type = PIPE_PRIM_TRIANGLES; render->num_verts = cell->prim_buffer.num_verts; - render->num_attribs = CELL_MAX_ATTRIBS; + render->vertex_size = cell->vertex_info->size * 4; render->xmin = cell->prim_buffer.xmin; render->ymin = cell->prim_buffer.ymin; render->xmax = cell->prim_buffer.xmax; diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index a420de041a..711d03b49f 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -31,7 +31,9 @@ */ +#include "cell_batch.h" #include "cell_context.h" +#include "cell_flush.h" #include "cell_spu.h" #include "cell_vbuf.h" #include "pipe/draw/draw_vbuf.h" @@ -141,26 +143,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, if (cvbr->prim != PIPE_PRIM_TRIANGLES) return; /* only render tris for now */ -#if 0 - for (i = 0; i < cell->num_spus; i++) { - struct cell_command_render *render = &cell_global.command[i].render; - render->opcode = CELL_CMD_RENDER; - render->prim_type = cvbr->prim; - render->num_verts = nr_vertices; - render->num_attribs = CELL_MAX_ATTRIBS; /* XXX fix */ - render->vertex_data = vertices; - render->index_data = indices; - render->num_indexes = nr_indices; - render->xmin = xmin; - render->ymin = ymin; - render->xmax = xmax; - render->ymax = ymax; - - ASSERT_ALIGN16(render->vertex_data); - ASSERT_ALIGN16(render->index_data); - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER); - } -#else + /* build/insert batch RENDER command */ { struct cell_command_render *render = (struct cell_command_render *) @@ -168,7 +151,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, render->opcode = CELL_CMD_RENDER; render->prim_type = cvbr->prim; render->num_verts = nr_vertices; - render->num_attribs = CELL_MAX_ATTRIBS; /* XXX fix */ + render->vertex_size = 4 * cell->vertex_info.size; render->vertex_data = vertices; render->index_data = indices; render->num_indexes = nr_indices; @@ -180,7 +163,6 @@ cell_vbuf_draw(struct vbuf_render *vbr, ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(render->index_data); } -#endif #if 01 /* XXX this is temporary */ diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 2e5cb76b4a..03dbef051b 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -210,8 +210,8 @@ cmd_render(const struct cell_command_render *render) /* we'll DMA into these buffers */ ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; ushort indexes[CELL_MAX_VBUF_INDEXES] ALIGN16_ATTRIB; - - uint i, j, vertex_size, vertex_bytes, index_bytes; + uint i, j, total_vertex_bytes, total_index_bytes; + const uint vertex_size = render->vertex_size; /* in bytes */ if (Debug) { printf("SPU %u: RENDER prim %u, indices: %u, nr_vert: %u\n", @@ -228,36 +228,34 @@ cmd_render(const struct cell_command_render *render) ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(render->index_data); - vertex_size = render->num_attribs * 4 * sizeof(float); - /* how much vertex data */ - vertex_bytes = render->num_verts * vertex_size; - index_bytes = render->num_indexes * sizeof(ushort); - if (index_bytes < 16) - index_bytes = 16; + total_vertex_bytes = render->num_verts * vertex_size; + total_index_bytes = render->num_indexes * sizeof(ushort); + if (total_index_bytes < 16) + total_index_bytes = 16; else - index_bytes = (index_bytes + 15) & ~0xf; /* multiple of 16 */ + total_index_bytes = (total_index_bytes + 15) & ~0xf; /* multiple of 16 */ /* - printf("VBUF: indices at %p, vertices at %p vertex_bytes %u ind_bytes %u\n", - render->index_data, render->vertex_data, vertex_bytes, index_bytes); + printf("VBUF: indices at %p, vertices at %p total_vertex_bytes %u ind_bytes %u\n", + render->index_data, render->vertex_data, total_vertex_bytes, total_index_bytes); */ - ASSERT(vertex_bytes % 16 == 0); + ASSERT(total_vertex_bytes % 16 == 0); /* get vertex data from main memory */ mfc_get(vertex_data, /* dest */ (unsigned int) render->vertex_data, /* src */ - vertex_bytes, /* size */ + total_vertex_bytes, /* size */ TAG_VERTEX_BUFFER, 0, /* tid */ 0 /* rid */); - ASSERT(index_bytes % 16 == 0); + ASSERT(total_index_bytes % 16 == 0); /* get index data from main memory */ mfc_get(indexes, /* dest */ (unsigned int) render->index_data, /* src */ - index_bytes, + total_index_bytes, TAG_INDEX_BUFFER, 0, /* tid */ 0 /* rid */); -- cgit v1.2.3 From 547e52f4a04f39d5a96bf5b1ebe97c83b9219dab Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 Jan 2008 18:20:07 -0700 Subject: Cell: insert dummy field into cell_command_render struct to work around apparent compiler bug Without this dummy field, it seems the SPU code in cmd_render() gets a bogus value when dereferencing render->num_indices, sometimes. This showed up as missing tries in gears.c in the first frame rendered. Using spu-gcc version 4.1.1, Cell SDK 2.1 --- src/mesa/pipe/cell/common.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index cd2b61e6e9..b9a00ccbb5 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -104,6 +104,7 @@ struct cell_command_render uint prim_type; uint num_verts; uint vertex_size; /**< bytes per vertex */ + uint dummy; /* XXX this dummy field works around a compiler bug */ uint num_indexes; const void *vertex_data; const ushort *index_data; -- cgit v1.2.3 From 48355538a65fe9c0be234c61080edd70f6a86736 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 Jan 2008 15:23:27 -0700 Subject: gallium: added rastpos_destroy() --- src/mesa/state_tracker/st_cb_rasterpos.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index dc6456c775..9e20e94dc7 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -188,6 +188,7 @@ new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw) rs->stage.line = rastpos_line; rs->stage.tri = rastpos_tri; rs->stage.end = rastpos_end; + rs->stage.destroy = rastpos_destroy; rs->stage.reset_stipple_counter = rastpos_reset_stipple_counter; rs->stage.destroy = rastpos_destroy; rs->ctx = ctx; -- cgit v1.2.3 From a148cc51fbe56266199057cfde2abb2b59eb8713 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 08:15:04 -0700 Subject: gallium: optimizations to flatshade, two-side prim stages --- src/mesa/pipe/draw/draw_flatshade.c | 26 +++++++++++++++++++------- src/mesa/pipe/draw/draw_twoside.c | 31 ++++++++++++++++--------------- 2 files changed, 35 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 46b0f7970f..8444c53310 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -70,18 +70,31 @@ static void flatshade_begin( struct draw_stage *stage ) /** Copy all the color attributes from 'src' vertex to 'dst' vertex */ -static INLINE void copy_colors( struct draw_stage *stage, - struct vertex_header *dst, +static INLINE void copy_colors( struct draw_stage *stage, + struct vertex_header *dst, const struct vertex_header *src ) { const struct flat_stage *flat = flat_stage(stage); uint i; + for (i = 0; i < flat->num_color_attribs; i++) { + const uint attr = flat->color_attribs[i]; + COPY_4FV(dst->data[attr], src->data[attr]); + } +} - /* Look for constant/flat attribs and duplicate from src to dst vertex */ - /* skip attrib[0] which is vert pos */ + +/** Copy all the color attributes from src vertex to dst0 & dst1 vertices */ +static INLINE void copy_colors2( struct draw_stage *stage, + struct vertex_header *dst0, + struct vertex_header *dst1, + const struct vertex_header *src ) +{ + const struct flat_stage *flat = flat_stage(stage); + uint i; for (i = 0; i < flat->num_color_attribs; i++) { const uint attr = flat->color_attribs[i]; - memcpy(dst->data[attr], src->data[attr], sizeof(src->data[0])); + COPY_4FV(dst0->data[attr], src->data[attr]); + COPY_4FV(dst1->data[attr], src->data[attr]); } } @@ -101,8 +114,7 @@ static void flatshade_tri( struct draw_stage *stage, tmp.v[1] = dup_vert(stage, header->v[1], 1); tmp.v[2] = header->v[2]; - copy_colors(stage, tmp.v[0], tmp.v[2]); - copy_colors(stage, tmp.v[1], tmp.v[2]); + copy_colors2(stage, tmp.v[0], tmp.v[1], tmp.v[2]); stage->next->tri( stage->next, &tmp ); } diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index da8e7bd90d..75c51ec6a9 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -75,6 +75,12 @@ static void twoside_begin( struct draw_stage *stage ) } } + if (!twoside->attrib_back0) + twoside->attrib_front0 = 0; + + if (!twoside->attrib_back1) + twoside->attrib_front1 = 0; + /* * We'll multiply the primitive's determinant by this sign to determine * if the triangle is back-facing (negative). @@ -86,28 +92,23 @@ static void twoside_begin( struct draw_stage *stage ) } -static INLINE void copy_attrib( unsigned attr_dst, - unsigned attr_src, - struct vertex_header *v ) -{ - COPY_4FV(v->data[attr_dst], v->data[attr_src]); -} - - /** * Copy back color(s) to front color(s). */ -static struct vertex_header *copy_bfc( struct twoside_stage *twoside, - const struct vertex_header *v, - unsigned idx ) +static INLINE struct vertex_header * +copy_bfc( struct twoside_stage *twoside, + const struct vertex_header *v, + unsigned idx ) { struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); - if (twoside->attrib_front0 && twoside->attrib_back0) { - copy_attrib(twoside->attrib_front0, twoside->attrib_back0, tmp); + if (twoside->attrib_back0) { + COPY_4FV(tmp->data[twoside->attrib_front0], + tmp->data[twoside->attrib_back0]); } - if (twoside->attrib_front1 && twoside->attrib_back1) { - copy_attrib(twoside->attrib_front1, twoside->attrib_back1, tmp); + if (twoside->attrib_back1) { + COPY_4FV(tmp->data[twoside->attrib_front1], + tmp->data[twoside->attrib_back1]); } return tmp; -- cgit v1.2.3 From 78bff07f73a546363b737b33958311efa7b3ce53 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 24 Jan 2008 18:26:37 -0700 Subject: Cell: new ROUNDUP16() macro --- src/mesa/pipe/cell/common.h | 3 +++ src/mesa/pipe/cell/spu/spu_main.c | 7 +++---- 2 files changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index b9a00ccbb5..5c437daac6 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -43,6 +43,9 @@ assert((((unsigned long) (ptr)) & 0xf) == 0); +/** round up value to next multiple of 16 */ +#define ROUNDUP16(k) (((k) + 0xf) & ~0xf) + #define TILE_SIZE 32 diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 03dbef051b..1d0de7b1f9 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -234,7 +234,7 @@ cmd_render(const struct cell_command_render *render) if (total_index_bytes < 16) total_index_bytes = 16; else - total_index_bytes = (total_index_bytes + 15) & ~0xf; /* multiple of 16 */ + total_index_bytes = ROUNDUP16(total_index_bytes); /* printf("VBUF: indices at %p, vertices at %p total_vertex_bytes %u ind_bytes %u\n", @@ -457,10 +457,9 @@ cmd_batch(uint opcode) ASSERT_ALIGN16(spu.init.batch_buffers[buf]); - size = (size + 0xf) & ~0xf; + size = ROUNDUP16(size); - ASSERT(size % 16 == 0); - ASSERT((unsigned int) spu.init.batch_buffers[buf] % 16 == 0); + ASSERT_ALIGN16(spu.init.batch_buffers[buf]); mfc_get(buffer, /* dest */ (unsigned int) spu.init.batch_buffers[buf], /* src */ -- cgit v1.2.3 From 4c4f691069411c77b33a6bfca64766f861cbc823 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 08:17:04 -0700 Subject: Cell: batch_buffer_size should be uint, not ubyte --- src/mesa/pipe/cell/ppu/cell_context.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index b6c59ff596..3bd88bfd5b 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -102,7 +102,7 @@ struct cell_context uint num_spus; - ubyte batch_buffer_size[CELL_NUM_BATCH_BUFFERS]; + uint batch_buffer_size[CELL_NUM_BATCH_BUFFERS]; ubyte batch_buffer[CELL_NUM_BATCH_BUFFERS][CELL_BATCH_BUFFER_SIZE] ALIGN16_ATTRIB; int cur_batch; /**< which batch buffer is being filled */ -- cgit v1.2.3 From 235da629dceb23bfddea1eadfcf771d2794d6119 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 08:17:52 -0700 Subject: Cell: propogate vertex info to SPUs, use it for attrib interpolation --- src/mesa/pipe/cell/common.h | 7 +++-- src/mesa/pipe/cell/ppu/cell_state.h | 1 + src/mesa/pipe/cell/ppu/cell_state_derived.c | 5 ++++ src/mesa/pipe/cell/ppu/cell_state_emit.c | 6 ++++ src/mesa/pipe/cell/spu/spu_main.c | 19 ++++++++++-- src/mesa/pipe/cell/spu/spu_main.h | 4 +++ src/mesa/pipe/cell/spu/spu_tri.c | 46 ++++++++++++----------------- 7 files changed, 55 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 5c437daac6..af7f27bc83 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -64,6 +64,7 @@ #define CELL_CMD_BATCH 6 #define CELL_CMD_STATE_DEPTH_STENCIL 7 #define CELL_CMD_STATE_SAMPLER 8 +#define CELL_CMD_STATE_VERTEX_INFO 9 #define CELL_NUM_BATCH_BUFFERS 3 @@ -103,11 +104,11 @@ struct cell_command_clear_surface struct cell_command_render { - uint opcode; - uint prim_type; + uint opcode; /**< CELL_CMD_RENDER */ + uint prim_type; /**< PIPE_PRIM_x */ uint num_verts; uint vertex_size; /**< bytes per vertex */ - uint dummy; /* XXX this dummy field works around a compiler bug */ + uint dummy; /* XXX this dummy field works around a compiler bug */ uint num_indexes; const void *vertex_data; const ushort *index_data; diff --git a/src/mesa/pipe/cell/ppu/cell_state.h b/src/mesa/pipe/cell/ppu/cell_state.h index fbca7c9574..3a71ba14fa 100644 --- a/src/mesa/pipe/cell/ppu/cell_state.h +++ b/src/mesa/pipe/cell/ppu/cell_state.h @@ -19,6 +19,7 @@ #define CELL_NEW_VERTEX 0x1000 #define CELL_NEW_VS 0x2000 #define CELL_NEW_CONSTANTS 0x4000 +#define CELL_NEW_VERTEX_INFO 0x8000 diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c index f3b3ae8544..56daf5dfde 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ b/src/mesa/pipe/cell/ppu/cell_state_derived.c @@ -30,6 +30,7 @@ #include "pipe/draw/draw_context.h" #include "pipe/draw/draw_vertex.h" #include "cell_context.h" +#include "cell_batch.h" #include "cell_state.h" #include "cell_state_emit.h" @@ -85,6 +86,7 @@ calculate_vertex_layout( struct cell_context *cell ) assert(src >= 0); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); + /* * Loop over fragment shader inputs, searching for the matching output * from the vertex shader. @@ -126,6 +128,9 @@ calculate_vertex_layout( struct cell_context *cell ) } draw_compute_vertex_size(vinfo); + + /* XXX only signal this if format really changes */ + cell->dirty |= CELL_NEW_VERTEX_INFO; } diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c index e7d14d0d25..e5b7c92514 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.c +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.c @@ -49,4 +49,10 @@ cell_emit_state(struct cell_context *cell) cell_batch_append(cell, cell->sampler[0], sizeof(struct pipe_sampler_state)); } + + if (cell->dirty & CELL_NEW_VERTEX_INFO) { + uint cmd = CELL_CMD_STATE_VERTEX_INFO; + cell_batch_append(cell, &cmd, 4); + cell_batch_append(cell, &cell->vertex_info, sizeof(struct vertex_info)); + } } diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 1d0de7b1f9..880f8de550 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -390,6 +390,17 @@ cmd_state_sampler(const struct pipe_sampler_state *state) } +static void +cmd_state_vertex_info(const struct vertex_info *vinfo) +{ + if (Debug) + printf("SPU %u: VERTEX_INFO num_attribs=%u\n", spu.init.id, + vinfo->num_attribs); + memcpy(&spu.vertex_info, vinfo, sizeof(*vinfo)); +} + + + static void cmd_finish(void) { @@ -472,7 +483,6 @@ cmd_batch(uint opcode) /* Tell PPU we're done copying the buffer to local store */ release_batch_buffer(buf); - for (pos = 0; pos < usize; /* no incr */) { switch (buffer[pos]) { case CELL_CMD_FRAMEBUFFER: @@ -509,10 +519,13 @@ cmd_batch(uint opcode) pos += (1 + sizeof(struct pipe_depth_stencil_alpha_state) / 4); break; case CELL_CMD_STATE_SAMPLER: - cmd_state_sampler((struct pipe_sampler_state *) - &buffer[pos+1]); + cmd_state_sampler((struct pipe_sampler_state *) &buffer[pos+1]); pos += (1 + sizeof(struct pipe_sampler_state) / 4); break; + case CELL_CMD_STATE_VERTEX_INFO: + cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]); + pos += (1 + sizeof(struct vertex_info) / 4); + break; default: printf("SPU %u: bad opcode: 0x%x\n", spu.init.id, buffer[pos]); ASSERT(0); diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 3ef73c9473..e4359bf60d 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -30,6 +30,7 @@ #include "pipe/cell/common.h" +#include "pipe/draw/draw_vertex.h" #include "pipe/p_state.h" @@ -59,6 +60,9 @@ struct spu_global struct pipe_depth_stencil_alpha_state depth_stencil; struct pipe_blend_state blend; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; + + struct vertex_info vertex_info; + /* XXX more state to come */ } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 1d73c5171c..3d0d106c10 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -42,7 +42,7 @@ * Simplified types taken from other parts of Gallium */ struct vertex_header { - float data[2][4]; /* pos and color */ + float data[0][4]; }; struct prim_header { @@ -727,40 +727,32 @@ static void tri_persp_coeff( struct setup_stage *setup, */ static void setup_tri_coefficients( struct setup_stage *setup ) { -#if 0 - const enum interp_mode *interp = setup->softpipe->vertex_info.interp_mode; - unsigned slot, j; - - /* z and w are done by linear interpolation: - */ - tri_linear_coeff(setup, 0, 2); - tri_linear_coeff(setup, 0, 3); +#if 1 + uint i; - /* setup interpolation for all the remaining attributes: - */ - for (slot = 1; slot < setup->quad.nr_attrs; slot++) { - switch (interp[slot]) { + for (i = 0; i < spu.vertex_info.num_attribs; i++) { + switch (spu.vertex_info.interp_mode[i]) { + case INTERP_NONE: + break; + case INTERP_POS: + tri_linear_coeff(setup, i, 2, 3); /* slot 0, z */ + /* XXX interp W if PERSPECTIVE... */ + break; case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, slot, j); - break; - + /* fall-through */ case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - tri_linear_coeff(setup, slot, j); - break; - + tri_linear_coeff(setup, i, 0, 4); /* slot 1, color */ + break; case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - tri_persp_coeff(setup, slot, j); - break; - + break; default: - /* invalid interp mode */ - assert(0); + ASSERT(0); } } #else + ASSERT(spu.vertex_info.interp_mode[0] == INTERP_POS); + ASSERT(spu.vertex_info.interp_mode[1] == INTERP_LINEAR || + spu.vertex_info.interp_mode[1] == INTERP_CONSTANT); tri_linear_coeff(setup, 0, 2, 3); /* slot 0, z */ tri_linear_coeff(setup, 1, 0, 4); /* slot 1, color */ #endif -- cgit v1.2.3 From c344bda34dcbc1731573edc94de0c20549557adf Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 08:19:45 -0700 Subject: gallium: include p_state.h, replace PIPE_MAX_SHADER_OUTPUTS with PIPE_MAX_SHADER_INPUTS --- src/mesa/pipe/draw/draw_vertex.h | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index f20b10f8d7..dfc637b19b 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -34,6 +34,9 @@ #define DRAW_VERTEX_H +#include "pipe/p_state.h" + + /** * Vertex attribute emit modes */ @@ -68,9 +71,9 @@ struct vertex_info { uint num_attribs; uint hwfmt[4]; /**< hardware format info for this format */ - enum interp_mode interp_mode[PIPE_MAX_SHADER_OUTPUTS]; - enum attrib_emit emit[PIPE_MAX_SHADER_OUTPUTS]; /**< EMIT_x */ - uint src_index[PIPE_MAX_SHADER_OUTPUTS]; /**< map to post-xform attribs */ + enum interp_mode interp_mode[PIPE_MAX_SHADER_INPUTS]; + enum attrib_emit emit[PIPE_MAX_SHADER_INPUTS]; /**< EMIT_x */ + uint src_index[PIPE_MAX_SHADER_INPUTS]; /**< map to post-xform attribs */ uint size; /**< total vertex size in dwords */ }; @@ -88,7 +91,7 @@ draw_emit_vertex_attr(struct vertex_info *vinfo, uint src_index) { const uint n = vinfo->num_attribs; - assert(n < PIPE_MAX_SHADER_OUTPUTS); + assert(n < PIPE_MAX_SHADER_INPUTS); vinfo->emit[n] = emit; vinfo->interp_mode[n] = interp; vinfo->src_index[n] = src_index; -- cgit v1.2.3 From 59d66bf9df3a6b964f177a8aff286cac63476add Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 10 Jan 2008 14:57:07 +0000 Subject: 965: cope better with range decls for temps, immediates, etc --- src/mesa/pipe/i965simple/brw_vs_emit.c | 34 ++++++++++++---------------------- 1 file changed, 12 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index 5f212bd055..b32c233dd2 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -988,33 +988,24 @@ post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst ) static void process_declaration(const struct tgsi_full_declaration *decl, struct brw_prog_info *info) { + int first = decl->u.DeclarationRange.First; + int last = decl->u.DeclarationRange.Last; + + assert (decl->Declaration.Declare != TGSI_DECLARE_MASK); + switch(decl->Declaration.File) { - case TGSI_FILE_CONSTANT: { - if (decl->Declaration.Declare == TGSI_DECLARE_MASK) { - printf("DECLARATION MASK = %d\n", - decl->u.DeclarationMask.Mask); - assert(0); - } else { /*range*/ - info->num_consts += decl->u.DeclarationRange.Last - decl->u.DeclarationRange.First + 1; - } - } + case TGSI_FILE_CONSTANT: + info->num_consts += last - first + 1; break; case TGSI_FILE_INPUT: { } break; case TGSI_FILE_OUTPUT: { + assert(last == first); /* for now */ if (decl->Declaration.Semantic) { - int idx = 0; - if (decl->Declaration.Declare == TGSI_DECLARE_MASK) { - printf("DECLARATION MASK = %d\n", - decl->u.DeclarationMask.Mask); - assert(0); - } else { /*range*/ - idx = decl->u.DeclarationRange.First; - } switch (decl->Semantic.SemanticName) { case TGSI_SEMANTIC_POSITION: { - info->pos_idx = idx; + info->pos_idx = first; } break; case TGSI_SEMANTIC_COLOR: @@ -1025,7 +1016,7 @@ static void process_declaration(const struct tgsi_full_declaration *decl, break; case TGSI_SEMANTIC_PSIZE: { info->writes_psize = TRUE; - info->psize_idx = idx; + info->psize_idx = first; } break; case TGSI_SEMANTIC_GENERIC: @@ -1035,14 +1026,14 @@ static void process_declaration(const struct tgsi_full_declaration *decl, } break; case TGSI_FILE_TEMPORARY: { - info->num_temps++; + info->num_temps += (last - first) + 1; } break; case TGSI_FILE_SAMPLER: { } break; case TGSI_FILE_ADDRESS: { - info->num_addrs++; + info->num_addrs += (last - first) + 1; } break; case TGSI_FILE_IMMEDIATE: { @@ -1303,7 +1294,6 @@ void brw_vs_emit(struct brw_vs_compile *c) } break; case TGSI_TOKEN_TYPE_IMMEDIATE: { - int i; struct tgsi_full_immediate *imm = &parse.FullToken.FullImmediate; /*assert(imm->Immediate.Size == 4);*/ c->prog_data.imm_buf[c->prog_data.num_imm][0] = imm->u.ImmediateFloat32[0].Float; -- cgit v1.2.3 From 7a207682aafc05c62cbc5851cc6c98c43aa3d9bd Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 10 Jan 2008 14:57:55 +0000 Subject: 965: add asserts to catch batch overrun --- src/mesa/pipe/i965simple/brw_batch.h | 3 ++- src/mesa/pipe/i965simple/brw_winsys.h | 4 ++++ src/mesa/pipe/xlib/xm_winsys_aub.c | 37 +++++++++++++++++++++++------------ 3 files changed, 31 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i965simple/brw_batch.h b/src/mesa/pipe/i965simple/brw_batch.h index 8605d7c108..5f5932a488 100644 --- a/src/mesa/pipe/i965simple/brw_batch.h +++ b/src/mesa/pipe/i965simple/brw_batch.h @@ -44,7 +44,8 @@ #define OUT_RELOC( buf, flags, delta ) \ brw->winsys->batch_reloc(brw->winsys, buf, flags, delta) -#define ADVANCE_BATCH() +#define ADVANCE_BATCH() \ + brw->winsys->batch_end( brw->winsys ) /* XXX: this is bogus - need proper handling for out-of-memory in batchbuffer. */ diff --git a/src/mesa/pipe/i965simple/brw_winsys.h b/src/mesa/pipe/i965simple/brw_winsys.h index 253599896c..b60f63aa5b 100644 --- a/src/mesa/pipe/i965simple/brw_winsys.h +++ b/src/mesa/pipe/i965simple/brw_winsys.h @@ -193,9 +193,13 @@ static inline boolean brw_batchbuffer_data(struct brw_winsys *winsys, uint i; const unsigned *udata = (const unsigned*)(data); unsigned size = bytes/incr; + + winsys->batch_start(winsys, size, 0); for (i = 0; i < size; ++i) { winsys->batch_dword(winsys, udata[i]); } + winsys->batch_end(winsys); + return (i == size); } #endif diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index 2be8f8793d..980c0be540 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -466,9 +466,10 @@ struct aub_brw_winsys { struct pipe_winsys *pipe_winsys; - unsigned data[IWS_BATCHBUFFER_SIZE]; - unsigned nr; - unsigned size; + unsigned batch_data[IWS_BATCHBUFFER_SIZE]; + unsigned batch_nr; + unsigned batch_size; + unsigned batch_alloc; }; @@ -490,9 +491,10 @@ static unsigned *aub_i965_batch_start( struct brw_winsys *sws, { struct aub_brw_winsys *iws = aub_brw_winsys(sws); - if (iws->size < iws->nr + dwords) + if (iws->batch_size < iws->batch_nr + dwords) return NULL; + iws->batch_alloc = iws->batch_nr + dwords; return (void *)1; /* not a valid pointer! */ } @@ -501,7 +503,8 @@ static void aub_i965_batch_dword( struct brw_winsys *sws, { struct aub_brw_winsys *iws = aub_brw_winsys(sws); - iws->data[iws->nr++] = dword; + assert(iws->batch_nr < iws->batch_alloc); + iws->batch_data[iws->batch_nr++] = dword; } static void aub_i965_batch_reloc( struct brw_winsys *sws, @@ -511,7 +514,8 @@ static void aub_i965_batch_reloc( struct brw_winsys *sws, { struct aub_brw_winsys *iws = aub_brw_winsys(sws); - iws->data[iws->nr++] = aub_bo(buf)->offset + delta; + assert(iws->batch_nr < iws->batch_alloc); + iws->batch_data[iws->batch_nr++] = aub_bo(buf)->offset + delta; } static unsigned aub_i965_get_buffer_offset( struct brw_winsys *sws, @@ -521,19 +525,27 @@ static unsigned aub_i965_get_buffer_offset( struct brw_winsys *sws, return aub_bo(buf)->offset; } +static void aub_i965_batch_end( struct brw_winsys *sws ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + assert(iws->batch_nr <= iws->batch_alloc); + iws->batch_alloc = 0; +} static void aub_i965_batch_flush( struct brw_winsys *sws, struct pipe_fence_handle **fence ) { struct aub_brw_winsys *iws = aub_brw_winsys(sws); - assert(iws->nr <= iws->size); + assert(iws->batch_nr <= iws->batch_size); - if (iws->nr) + if (iws->batch_nr) { xmesa_commands_aub( iws->pipe_winsys, - iws->data, - iws->nr ); - iws->nr = 0; + iws->batch_data, + iws->batch_nr ); + } + + iws->batch_nr = 0; } @@ -639,13 +651,14 @@ xmesa_create_i965simple( struct pipe_winsys *winsys ) iws->winsys.batch_start = aub_i965_batch_start; iws->winsys.batch_dword = aub_i965_batch_dword; iws->winsys.batch_reloc = aub_i965_batch_reloc; + iws->winsys.batch_end = aub_i965_batch_end; iws->winsys.batch_flush = aub_i965_batch_flush; iws->winsys.buffer_subdata_typed = aub_i965_buffer_subdata_typed; iws->winsys.get_buffer_offset = aub_i965_get_buffer_offset; iws->pipe_winsys = winsys; - iws->size = IWS_BATCHBUFFER_SIZE; + iws->batch_size = IWS_BATCHBUFFER_SIZE; /* Create the i965simple context: */ -- cgit v1.2.3 From 756d52ec12c41ee90ee9598dc9028cc134806bd2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 25 Jan 2008 17:01:01 +0100 Subject: gallium: Simplify winsys buffer interface. The properties of a buffer represented by struct pipe_buffer_handle are now basically constant over its lifetime. The state tracker gets to deal with any more complex buffer semantics it may need to provide. --- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 82 +++++++------- src/mesa/pipe/i915simple/i915_context.c | 4 +- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 8 +- src/mesa/pipe/i915simple/i915_state.c | 3 +- src/mesa/pipe/i915simple/i915_texture.c | 14 +-- src/mesa/pipe/i965simple/brw_curbe.c | 7 +- src/mesa/pipe/i965simple/brw_draw_upload.c | 6 +- src/mesa/pipe/i965simple/brw_misc_state.c | 6 +- src/mesa/pipe/i965simple/brw_state_pool.c | 10 +- src/mesa/pipe/i965simple/brw_tex_layout.c | 14 +-- src/mesa/pipe/p_defines.h | 26 ++--- src/mesa/pipe/p_inlines.h | 4 +- src/mesa/pipe/p_winsys.h | 59 +++------- src/mesa/pipe/pipebuffer/pb_buffer_handle.c | 52 +-------- src/mesa/pipe/pipebuffer/pb_buffer_handle.h | 4 +- src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 3 +- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 4 +- src/mesa/pipe/softpipe/sp_draw_arrays.c | 6 +- src/mesa/pipe/softpipe/sp_texture.c | 9 +- src/mesa/pipe/xlib/xm_winsys.c | 69 ++---------- src/mesa/pipe/xlib/xm_winsys_aub.c | 119 +++++---------------- src/mesa/state_tracker/st_atom_constbuf.c | 16 ++- src/mesa/state_tracker/st_cb_bufferobjects.c | 106 ++++++++++-------- src/mesa/state_tracker/st_cb_bufferobjects.h | 1 + src/mesa/state_tracker/st_context.h | 7 -- src/mesa/state_tracker/st_draw.c | 45 ++------ 26 files changed, 228 insertions(+), 456 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 340fa2fb85..43ed0602a4 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -70,10 +70,10 @@ static void *intel_buffer_map(struct pipe_winsys *winsys, { unsigned drm_flags = 0; - if (flags & PIPE_BUFFER_FLAG_WRITE) + if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) drm_flags |= DRM_BO_FLAG_WRITE; - if (flags & PIPE_BUFFER_FLAG_READ) + if (flags & PIPE_BUFFER_USAGE_CPU_READ) drm_flags |= DRM_BO_FLAG_READ; return driBOMap( dri_bo(buf), drm_flags, 0 ); @@ -103,50 +103,48 @@ intel_buffer_reference(struct pipe_winsys *winsys, } -/* Grabs the hardware lock! - */ -static int intel_buffer_data(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, - unsigned size, const void *data, - unsigned usage ) -{ - driBOData( dri_bo(buf), size, data, 0 ); - return 0; -} - -static int intel_buffer_subdata(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) -{ - driBOSubData( dri_bo(buf), offset, size, data ); - return 0; -} - -static int intel_buffer_get_subdata(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) -{ - driBOGetSubData( dri_bo(buf), offset, size, data ); - return 0; -} - /* Pipe has no concept of pools. We choose the tex/region pool * for all buffers. + * Grabs the hardware lock! */ static struct pipe_buffer_handle * intel_buffer_create(struct pipe_winsys *winsys, unsigned alignment, - unsigned flags, - unsigned hint ) + unsigned usage, + unsigned size ) { struct _DriBufferObject *buffer; struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + unsigned flags = 0; + + if (usage & (PIPE_BUFFER_USAGE_VERTEX /*| IWS_BUFFER_USAGE_LOCAL*/)) { + flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; + } else { + flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; + } + + if (usage & PIPE_BUFFER_USAGE_GPU_READ) + flags |= DRM_BO_FLAG_READ; + + if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) + flags |= DRM_BO_FLAG_WRITE; + + /* drm complains if we don't set any read/write flags. + */ + if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) + flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + +#if 0 + if (flags & IWS_BUFFER_USAGE_EXE) + flags |= DRM_BO_FLAG_EXE; + + if (usage & IWS_BUFFER_USAGE_CACHED) + flags |= DRM_BO_FLAG_CACHED; +#endif + driGenBuffers( iws->regionPool, - "pipe buffer", 1, &buffer, alignment, flags, hint ); + "pipe buffer", 1, &buffer, alignment, flags, 0 ); + driBOData( buffer, size, NULL, 0 ); return pipe_bo(buffer); } @@ -219,15 +217,12 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, surf->pitch = round_up(width, alignment / surf->cpp); assert(!surf->buffer); - surf->buffer = winsys->buffer_create(winsys, alignment, 0, 0); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->cpp * height); if(!surf->buffer) return -1; - ret = winsys->buffer_data(winsys, - surf->buffer, - surf->pitch * surf->cpp * height, - NULL, - PIPE_BUFFER_USAGE_PIXEL); if(ret) { winsys->buffer_reference(winsys, &surf->buffer, NULL); return ret; @@ -285,9 +280,6 @@ intel_create_pipe_winsys( int fd ) iws->winsys.buffer_map = intel_buffer_map; iws->winsys.buffer_unmap = intel_buffer_unmap; iws->winsys.buffer_reference = intel_buffer_reference; - iws->winsys.buffer_data = intel_buffer_data; - iws->winsys.buffer_subdata = intel_buffer_subdata; - iws->winsys.buffer_get_subdata = intel_buffer_get_subdata; iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; iws->winsys.printf = intel_printf; iws->winsys.get_name = intel_get_name; diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index 1c6b8cbb05..c2b96d2554 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -185,7 +185,7 @@ i915_draw_elements( struct pipe_context *pipe, void *buf = pipe->winsys->buffer_map(pipe->winsys, i915->vertex_buffer[i].buffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } } @@ -193,7 +193,7 @@ i915_draw_elements( struct pipe_context *pipe, if (indexBuffer) { void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index edc62e25e5..0887f9a1c6 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -99,16 +99,14 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, /* FIXME: handle failure */ assert(!i915->vbo); - i915->vbo = winsys->buffer_create(winsys, 64, 0, 0); - winsys->buffer_data( winsys, i915->vbo, - size, NULL, - I915_BUFFER_USAGE_LIT_VERTEX ); + i915->vbo = winsys->buffer_create(winsys, 64, I915_BUFFER_USAGE_LIT_VERTEX, + size); i915->dirty |= I915_NEW_VBO; return winsys->buffer_map(winsys, i915->vbo, - PIPE_BUFFER_FLAG_WRITE ); + PIPE_BUFFER_USAGE_CPU_WRITE); } diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 1190e05699..950ea52d60 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -476,7 +476,8 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, { void *mapped; if (buf->size && - (mapped = ws->buffer_map(ws, buf->buffer, PIPE_BUFFER_FLAG_READ))) { + (mapped = ws->buffer_map(ws, buf->buffer, + PIPE_BUFFER_USAGE_CPU_READ))) { memcpy(i915->current.constants[shader], mapped, buf->size); ws->buffer_unmap(ws, buf->buffer); i915->current.num_user_constants[shader] diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c index 6b0a4a96f3..bf80e18233 100644 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -490,15 +490,11 @@ i915_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) sizeof(struct i915_texture) - sizeof(struct pipe_texture)); if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : - i915_miptree_layout(pipe, tex)) { - tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, 0, 0); - - if (tex->buffer) - pipe->winsys->buffer_data(pipe->winsys, tex->buffer, - tex->pitch * tex->base.cpp * - tex->total_height, NULL, - PIPE_BUFFER_USAGE_PIXEL); - } + i915_miptree_layout(pipe, tex)) + tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, + PIPE_BUFFER_USAGE_PIXEL, + tex->pitch * tex->base.cpp * + tex->total_height); if (!tex->buffer) { FREE(tex); diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index 4d79a7abe2..2733eb4e75 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -255,15 +255,12 @@ static void upload_constant_buffer(struct brw_context *brw) /* FIXME: buffer size is num_consts + num_immediates */ if (brw->vs.prog_data->num_consts) { /* map the vertex constant buffer and copy to curbe: */ - ws->buffer_map(ws, cbuffer->buffer, 0); + void *data = ws->buffer_map(ws, cbuffer->buffer, 0); /* FIXME: this is wrong. the cbuffer->size currently * represents size of consts + immediates. so if we'll * have both we'll copy over the end of the buffer * with the subsequent memcpy */ - ws->buffer_get_subdata(ws, cbuffer->buffer, - 0, - cbuffer->size, - &buf[offset]); + memcpy(&buf[offset], data, cbuffer->size); ws->buffer_unmap(ws, cbuffer->buffer); offset += cbuffer->size; } diff --git a/src/mesa/pipe/i965simple/brw_draw_upload.c b/src/mesa/pipe/i965simple/brw_draw_upload.c index 88d6c9d111..43e53914e9 100644 --- a/src/mesa/pipe/i965simple/brw_draw_upload.c +++ b/src/mesa/pipe/i965simple/brw_draw_upload.c @@ -240,7 +240,7 @@ boolean brw_upload_vertex_buffers( struct brw_context *brw ) for (i = 0; i < nr_enabled; i++) { OUT_BATCH( vbp.vb[i].vb0.dword ); - OUT_RELOC( vbp.vb[i].buffer, PIPE_BUFFER_FLAG_READ, + OUT_RELOC( vbp.vb[i].buffer, PIPE_BUFFER_USAGE_GPU_READ, vbp.vb[i].offset); OUT_BATCH( vbp.vb[i].max_index ); OUT_BATCH( vbp.vb[i].instance_data_step_rate ); @@ -290,8 +290,8 @@ boolean brw_upload_indices( struct brw_context *brw, BEGIN_BATCH(4, 0); OUT_BATCH( ib.header.dword ); - OUT_RELOC( index_buffer, PIPE_BUFFER_FLAG_READ, start); - OUT_RELOC( index_buffer, PIPE_BUFFER_FLAG_READ, start + count); + OUT_RELOC( index_buffer, PIPE_BUFFER_USAGE_GPU_READ, start); + OUT_RELOC( index_buffer, PIPE_BUFFER_USAGE_GPU_READ, start + count); OUT_BATCH( 0 ); ADVANCE_BATCH(); } diff --git a/src/mesa/pipe/i965simple/brw_misc_state.c b/src/mesa/pipe/i965simple/brw_misc_state.c index 3b7f36dd1d..925049ecc1 100644 --- a/src/mesa/pipe/i965simple/brw_misc_state.c +++ b/src/mesa/pipe/i965simple/brw_misc_state.c @@ -245,7 +245,7 @@ static void upload_depthbuffer(struct brw_context *brw) // (depth_surface->region->tiled << 27) | (BRW_SURFACE_2D << 29)); OUT_RELOC(depth_surface->buffer, - PIPE_BUFFER_FLAG_READ | PIPE_BUFFER_FLAG_WRITE, 0); + PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE, 0); OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) | ((depth_surface->pitch - 1) << 6) | ((depth_surface->height - 1) << 19)); @@ -465,10 +465,10 @@ static void upload_state_base_address( struct brw_context *brw ) BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2)); OUT_RELOC(brw->pool[BRW_GS_POOL].buffer, - PIPE_BUFFER_FLAG_READ, + PIPE_BUFFER_USAGE_GPU_READ, 1); /* General state base address */ OUT_RELOC(brw->pool[BRW_SS_POOL].buffer, - PIPE_BUFFER_FLAG_READ, + PIPE_BUFFER_USAGE_GPU_READ, 1); /* Surface state base address */ OUT_BATCH(1); /* Indirect object base address */ OUT_BATCH(1); /* General state upper bound */ diff --git a/src/mesa/pipe/i965simple/brw_state_pool.c b/src/mesa/pipe/i965simple/brw_state_pool.c index 78268ed8f2..2f930be837 100644 --- a/src/mesa/pipe/i965simple/brw_state_pool.c +++ b/src/mesa/pipe/i965simple/brw_state_pool.c @@ -91,13 +91,9 @@ static void brw_init_pool( struct brw_context *brw, pool->brw = brw; pool->buffer = brw->pipe.winsys->buffer_create(brw->pipe.winsys, - 4096, 0, 0); - - brw->pipe.winsys->buffer_data(brw->pipe.winsys, - pool->buffer, - size, - NULL, - 0 /* DRM_BO_FLAG_MEM_TT */); + 4096, + 0 /* DRM_BO_FLAG_MEM_TT */, + size); } static void brw_destroy_pool( struct brw_context *brw, diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.c b/src/mesa/pipe/i965simple/brw_tex_layout.c index 2b2bf16f1b..eadacbf09e 100644 --- a/src/mesa/pipe/i965simple/brw_tex_layout.c +++ b/src/mesa/pipe/i965simple/brw_tex_layout.c @@ -308,15 +308,11 @@ brw_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) memset(&tex->base + 1, 0, sizeof(struct brw_texture) - sizeof(struct pipe_texture)); - if (brw_miptree_layout(pipe, tex)) { - tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, 0, 0); - - if (tex->buffer) - pipe->winsys->buffer_data(pipe->winsys, tex->buffer, - tex->pitch * tex->base.cpp * - tex->total_height, NULL, - PIPE_BUFFER_USAGE_PIXEL); - } + if (brw_miptree_layout(pipe, tex)) + tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, + PIPE_BUFFER_USAGE_PIXEL, + tex->pitch * tex->base.cpp * + tex->total_height); if (!tex->buffer) { FREE(tex); diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 50bea691e7..90eebd5e6c 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -177,26 +177,20 @@ enum pipe_texture_target { #define PIPE_SURFACE_STATUS_CLEAR 2 -/** - * Buffer access flags - */ -#define PIPE_BUFFER_FLAG_READ 0x1 -#define PIPE_BUFFER_FLAG_WRITE 0x2 -#define PIPE_BUFFER_FLAG_MEM_LOCAL 0x4 -#define PIPE_BUFFER_FLAG_CACHED 0x8 -#define PIPE_BUFFER_FLAG_CUSTOM (1<<16) - - - /** * Buffer usage flags */ -#define PIPE_BUFFER_USAGE_PIXEL (1 << 0) -#define PIPE_BUFFER_USAGE_VERTEX (1 << 1) -#define PIPE_BUFFER_USAGE_INDEX (1 << 2) -#define PIPE_BUFFER_USAGE_CONSTANT (1 << 3) +#define PIPE_BUFFER_USAGE_CPU_READ (1 << 0) +#define PIPE_BUFFER_USAGE_CPU_WRITE (1 << 1) +#define PIPE_BUFFER_USAGE_GPU_READ (1 << 2) +#define PIPE_BUFFER_USAGE_GPU_WRITE (1 << 3) +#define PIPE_BUFFER_USAGE_PIXEL (1 << 4) +#define PIPE_BUFFER_USAGE_VERTEX (1 << 5) +#define PIPE_BUFFER_USAGE_INDEX (1 << 6) +#define PIPE_BUFFER_USAGE_CONSTANT (1 << 7) /** Pipe driver custam usage flags should be greater or equal to this value */ -#define PIPE_BUFFER_USAGE_CUSTOM (1 << 16) +#define PIPE_BUFFER_USAGE_CUSTOM (1 << 16) + /** * Flush types: diff --git a/src/mesa/pipe/p_inlines.h b/src/mesa/pipe/p_inlines.h index 6976d087f9..8ee0820f92 100644 --- a/src/mesa/pipe/p_inlines.h +++ b/src/mesa/pipe/p_inlines.h @@ -37,8 +37,8 @@ static INLINE void * pipe_surface_map(struct pipe_surface *surface) { return (char *)surface->winsys->buffer_map( surface->winsys, surface->buffer, - PIPE_BUFFER_FLAG_WRITE | - PIPE_BUFFER_FLAG_READ ) + PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_CPU_READ ) + surface->offset; } diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 75c6dc7e85..9742f59a4d 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -1,4 +1,4 @@ -/************************************************************************** + /************************************************************************** * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. @@ -93,17 +93,20 @@ struct pipe_winsys /** - * The buffer manager is modeled after the dri_bufmgr interface, which - * in turn is modeled after the ARB_vertex_buffer_object extension, - * but this is the subset that gallium cares about. Remember that - * gallium gets to choose the interface it needs, and the window - * systems must then implement that interface (rather than the + * Buffer management. Buffer attributes are mostly fixed over its lifetime. + * + * Remember that gallium gets to choose the interface it needs, and the + * window systems must then implement that interface (rather than the * other way around...). + * + * usage is a bitmask of PIPE_BUFFER_USAGE_PIXEL/VERTEX/INDEX/CONSTANT. This + * usage argument is only an optimization hint, not a guarantee, therefore + * proper behavior must be observed in all circumstances. */ struct pipe_buffer_handle *(*buffer_create)( struct pipe_winsys *sws, unsigned alignment, - unsigned flags, - unsigned hint ); + unsigned usage, + unsigned size ); /** Create a buffer that wraps user-space data */ struct pipe_buffer_handle *(*user_buffer_create)(struct pipe_winsys *sws, @@ -116,7 +119,7 @@ struct pipe_winsys */ void *(*buffer_map)( struct pipe_winsys *sws, struct pipe_buffer_handle *buf, - unsigned flags ); + unsigned usage ); void (*buffer_unmap)( struct pipe_winsys *sws, struct pipe_buffer_handle *buf ); @@ -126,44 +129,8 @@ struct pipe_winsys struct pipe_buffer_handle **ptr, struct pipe_buffer_handle *buf ); - /** - * Create the data store of a buffer and optionally initialize it. - * - * usage is a bitmask of PIPE_BUFFER_USAGE_PIXEL/VERTEX/INDEX/CONSTANT. This - * usage argument is only an optimization hint, not a guarantee, therefore - * proper behavior must be observed in all circumstances. - * - * Returns zero on success. - */ - int (*buffer_data)(struct pipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned size, const void *data, - unsigned usage); - /** - * Modify some or all of the data contained in a buffer's data store. - * - * Returns zero on success. - */ - int (*buffer_subdata)(struct pipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data); - - /** - * Query some or all of the data contained in a buffer's data store. - * - * Returns zero on success. - */ - int (*buffer_get_subdata)(struct pipe_winsys *sws, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data); - - - /** Set ptr = buf, with reference counting */ + /** Set ptr = fence, with reference counting */ void (*fence_reference)( struct pipe_winsys *sws, struct pipe_fence_handle **ptr, struct pipe_fence_handle *fence ); diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_handle.c b/src/mesa/pipe/pipebuffer/pb_buffer_handle.c index 5a5eaee7ac..9aeff976f3 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_handle.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_handle.c @@ -46,8 +46,8 @@ static struct pipe_buffer_handle * buffer_handle_create(struct pipe_winsys *winsys, unsigned alignment, - unsigned flags, - unsigned hint) + unsigned usage, + unsigned size) { struct pipe_buffer_handle *handle; @@ -57,8 +57,8 @@ buffer_handle_create(struct pipe_winsys *winsys, handle->refcount = 1; handle->alignment = alignment; - handle->flags = flags; - handle->hint = hint; + handle->usage = usage; + handle->size = size; handle->buf = &null_buffer; @@ -73,7 +73,7 @@ buffer_handle_create_user(struct pipe_winsys *winsys, struct pipe_buffer_handle *handle; struct pipe_buffer *buf; - handle = buffer_handle_create(winsys, 1, 0, 0); + handle = buffer_handle_create(winsys, 1, 0, size); if(!handle) return NULL; @@ -129,46 +129,6 @@ buffer_handle_reference(struct pipe_winsys *winsys, } -static int -buffer_handle_subdata(struct pipe_winsys *winsys, - struct pipe_buffer_handle *handle, - unsigned long offset, - unsigned long size, - const void *data) -{ - void *map; - assert(handle); - assert(data); - map = buffer_handle_map(winsys, handle, PIPE_BUFFER_FLAG_WRITE); - if(map) { - memcpy((char *)map + offset, data, size); - buffer_handle_unmap(winsys, handle); - return 0; - } - return -1; -} - - -static int -buffer_handle_get_subdata(struct pipe_winsys *winsys, - struct pipe_buffer_handle *handle, - unsigned long offset, - unsigned long size, - void *data) -{ - void *map; - assert(handle); - assert(data); - map = buffer_handle_map(winsys, handle, PIPE_BUFFER_FLAG_READ); - if(map) { - memcpy(data, (char *)map + offset, size); - buffer_handle_unmap(winsys, handle); - return 0; - } - return -1; -} - - void buffer_handle_init_winsys(struct pipe_winsys *winsys) { @@ -177,6 +137,4 @@ buffer_handle_init_winsys(struct pipe_winsys *winsys) winsys->buffer_map = buffer_handle_map; winsys->buffer_unmap = buffer_handle_unmap; winsys->buffer_reference = buffer_handle_reference; - winsys->buffer_subdata = buffer_handle_subdata; - winsys->buffer_get_subdata = buffer_handle_get_subdata; } diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_handle.h b/src/mesa/pipe/pipebuffer/pb_buffer_handle.h index 076eec2fdc..34133c9fba 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_handle.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer_handle.h @@ -55,8 +55,8 @@ struct pipe_buffer_handle /** Allocation characteristics */ unsigned alignment; - unsigned flags; - unsigned hint; + unsigned usage; + unsigned size; /** * The actual buffer. diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index e9bc34b4a4..d174f24e32 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -241,7 +241,8 @@ mm_bufmgr_create_from_buffer(struct pipe_buffer *buffer, mm->buffer = buffer; mm->map = buffer_map(mm->buffer, - PIPE_BUFFER_FLAG_READ | PIPE_BUFFER_FLAG_WRITE ); + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); if(!mm->map) goto failure; diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index ee6fa62500..e6a8c78668 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -246,8 +246,8 @@ pool_bufmgr_create(struct buffer_manager *provider, goto failure; pool->map = buffer_map(pool->buffer, - PIPE_BUFFER_FLAG_READ | - PIPE_BUFFER_FLAG_WRITE ); + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); if(!pool->map) goto failure; diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index b7626f8a5f..8ee9177e5b 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -50,7 +50,7 @@ softpipe_map_constant_buffers(struct softpipe_context *sp) for (i = 0; i < 2; i++) { if (sp->constants[i].size) sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); } draw_set_mapped_constant_buffer(sp->draw, @@ -122,7 +122,7 @@ softpipe_draw_elements(struct pipe_context *pipe, void *buf = pipe->winsys->buffer_map(pipe->winsys, sp->vertex_buffer[i].buffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } } @@ -130,7 +130,7 @@ softpipe_draw_elements(struct pipe_context *pipe, if (indexBuffer) { void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index d43a6996e9..23e9c1b376 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -91,12 +91,9 @@ softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) softpipe_texture_layout(spt); - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); - - if (spt->buffer) { - pipe->winsys->buffer_data(pipe->winsys, spt->buffer, spt->buffer_size, - NULL, PIPE_BUFFER_USAGE_PIXEL); - } + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); if (!spt->buffer) { FREE(spt); diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index 432431aca1..cb043ef394 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -165,48 +165,6 @@ xm_buffer_reference(struct pipe_winsys *pws, } } -static int -xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, - unsigned size, const void *data, unsigned usage ) -{ - struct xm_buffer *xm_buf = xm_bo(buf); - assert(!xm_buf->userBuffer); - if (xm_buf->size != size) { - if (xm_buf->data) - align_free(xm_buf->data); - /* align to 16-byte multiple for Cell */ - xm_buf->data = align_malloc(size, 16); - xm_buf->size = size; - } - if (data) - memcpy(xm_buf->data, data, size); - return 0; -} - -static int -xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, - unsigned long offset, unsigned long size, const void *data) -{ - struct xm_buffer *xm_buf = xm_bo(buf); - GLubyte *b = (GLubyte *) xm_buf->data; - assert(!xm_buf->userBuffer); - assert(b); - memcpy(b + offset, data, size); - return 0; -} - -static int -xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, - unsigned long offset, unsigned long size, void *data) -{ - const struct xm_buffer *xm_buf = xm_bo(buf); - const GLubyte *b = (GLubyte *) xm_buf->data; - assert(!xm_buf->userBuffer); - assert(b); - memcpy(data, b + offset, size); - return 0; -} - /** * Display a surface that's in a tiled configuration. That is, all the @@ -317,11 +275,16 @@ xm_get_name(struct pipe_winsys *pws) static struct pipe_buffer_handle * xm_buffer_create(struct pipe_winsys *pws, unsigned alignment, - unsigned flags, - unsigned hints) + unsigned usage, + unsigned size) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); buffer->refcount = 1; + + /* align to 16-byte multiple for Cell */ + buffer->data = align_malloc(size, max(alignment, 16)); + buffer->size = size; + return pipe_bo(buffer); } @@ -359,7 +322,6 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys, unsigned flags) { const unsigned alignment = 64; - int ret; surf->width = width; surf->height = height; @@ -372,19 +334,11 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys, #endif assert(!surf->buffer); - surf->buffer = winsys->buffer_create(winsys, alignment, 0, 0); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->cpp * height); if(!surf->buffer) return -1; - - ret = winsys->buffer_data(winsys, - surf->buffer, - surf->pitch * surf->cpp * height, - NULL, - 0); - if(ret) { - winsys->buffer_reference(winsys, &surf->buffer, NULL); - return ret; - } return 0; } @@ -454,9 +408,6 @@ xmesa_get_pipe_winsys_aub(void) ws->buffer_map = xm_buffer_map; ws->buffer_unmap = xm_buffer_unmap; ws->buffer_reference = xm_buffer_reference; - ws->buffer_data = xm_buffer_data; - ws->buffer_subdata = xm_buffer_subdata; - ws->buffer_get_subdata = xm_buffer_get_subdata; ws->surface_alloc = xm_surface_alloc; ws->surface_alloc_storage = xm_surface_alloc_storage; diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index 980c0be540..28dd07bc6e 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -101,7 +101,7 @@ static void *aub_buffer_map(struct pipe_winsys *winsys, assert(sbo->data); - if (flags & PIPE_BUFFER_FLAG_WRITE) + if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) sbo->dump_on_unmap = 1; sbo->map_count++; @@ -150,61 +150,6 @@ aub_buffer_reference(struct pipe_winsys *winsys, } -static int aub_buffer_data(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, - unsigned size, const void *data, - unsigned usage ) -{ - struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); - struct aub_buffer *sbo = aub_bo(buf); - - /* Could reuse buffers that are not referenced in current - * batchbuffer. Can't do that atm, so always reallocate: - */ - if (1 || sbo->size < size) { - assert(iws->used + size < iws->size); - sbo->data = iws->pool + iws->used; - sbo->offset = AUB_BUF_START + iws->used; - iws->used += align(size, 4096); - } - - sbo->size = size; - - if (data != NULL) { - memcpy(sbo->data, data, size); - - brw_aub_gtt_data( iws->aubfile, - sbo->offset, - sbo->data, - sbo->size, - 0, - 0 ); - } - return 0; -} - -static int aub_buffer_subdata(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - const void *data) -{ - struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); - struct aub_buffer *sbo = aub_bo(buf); - - assert(sbo->size > offset + size); - memcpy(sbo->data + offset, data, size); - - brw_aub_gtt_data( iws->aubfile, - sbo->offset + offset, - sbo->data + offset, - size, - 0, - 0 ); - return 0; -} - - void xmesa_buffer_subdata_aub(struct pipe_winsys *winsys, struct pipe_buffer_handle *buf, unsigned long offset, @@ -258,29 +203,30 @@ void xmesa_display_aub( /* struct pipe_winsys *winsys, */ -static int aub_buffer_get_subdata(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, - unsigned long offset, - unsigned long size, - void *data) -{ - struct aub_buffer *sbo = aub_bo(buf); - assert(sbo->size >= offset + size); - memcpy(data, sbo->data + offset, size); - return 0; -} - /* Pipe has no concept of pools. We choose the tex/region pool * for all buffers. */ static struct pipe_buffer_handle * aub_buffer_create(struct pipe_winsys *winsys, - unsigned alignment, - unsigned flags, - unsigned hint) + unsigned alignment, + unsigned usage, + unsigned size) { + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer); + sbo->refcount = 1; + + /* Could reuse buffers that are not referenced in current + * batchbuffer. Can't do that atm, so always reallocate: + */ + assert(iws->used + size < iws->size); + sbo->data = iws->pool + iws->used; + sbo->offset = AUB_BUF_START + iws->used; + iws->used += align(size, 4096); + + sbo->size = size; + return pipe_bo(sbo); } @@ -288,17 +234,14 @@ aub_buffer_create(struct pipe_winsys *winsys, static struct pipe_buffer_handle * aub_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) { - struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer); - - sbo->refcount = 1; + struct aub_buffer *sbo; /* Lets hope this is meant for upload, not as a result! */ - aub_buffer_data( winsys, - pipe_bo(sbo), - bytes, - ptr, - 0 ); + sbo = aub_bo(aub_buffer_create( winsys, 0, 0, 0 )); + + sbo->data = ptr; + sbo->size = bytes; return pipe_bo(sbo); } @@ -345,7 +288,6 @@ aub_i915_surface_alloc_storage(struct pipe_winsys *winsys, unsigned flags) { const unsigned alignment = 64; - int ret; surf->width = width; surf->height = height; @@ -354,20 +296,12 @@ aub_i915_surface_alloc_storage(struct pipe_winsys *winsys, surf->pitch = round_up(width, alignment / surf->cpp); assert(!surf->buffer); - surf->buffer = winsys->buffer_create(winsys, alignment, 0, 0); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->cpp * height); if(!surf->buffer) return -1; - ret = winsys->buffer_data(winsys, - surf->buffer, - surf->pitch * surf->cpp * height, - NULL, - 0); - if(ret) { - winsys->buffer_reference(winsys, &surf->buffer, NULL); - return ret; - } - return 0; } @@ -418,9 +352,6 @@ xmesa_create_pipe_winsys_aub( void ) iws->winsys.buffer_map = aub_buffer_map; iws->winsys.buffer_unmap = aub_buffer_unmap; iws->winsys.buffer_reference = aub_buffer_reference; - iws->winsys.buffer_data = aub_buffer_data; - iws->winsys.buffer_subdata = aub_buffer_subdata; - iws->winsys.buffer_get_subdata = aub_buffer_get_subdata; iws->winsys.flush_frontbuffer = aub_flush_frontbuffer; iws->winsys.printf = aub_printf; iws->winsys.get_name = aub_get_name; diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index 57f5ec68d2..c9d63136b5 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -69,8 +69,13 @@ void st_upload_constants( struct st_context *st, _mesa_load_state_parameters(st->ctx, params); - if (!cbuf->buffer) - cbuf->buffer = ws->buffer_create(ws, 1, 0, 0); + if (cbuf->buffer && cbuf->size != paramBytes) + ws->buffer_reference( ws, &cbuf->buffer, NULL ); + + if (!cbuf->buffer) { + cbuf->buffer = ws->buffer_create(ws, 1, PIPE_BUFFER_USAGE_CONSTANT, + paramBytes); + } if (0) { @@ -80,8 +85,11 @@ void st_upload_constants( struct st_context *st, } /* load Mesa constants into the constant buffer */ - ws->buffer_data(ws, cbuf->buffer, paramBytes, params->ParameterValues, - PIPE_BUFFER_USAGE_CONSTANT); + if (cbuf->buffer) { + memcpy(ws->buffer_map(ws, cbuf->buffer, PIPE_BUFFER_USAGE_CPU_WRITE), + params->ParameterValues, paramBytes); + ws->buffer_unmap(ws, cbuf->buffer); + } cbuf->size = paramBytes; diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index 872248cdb5..60bd1d5f0d 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -55,7 +55,6 @@ static struct gl_buffer_object * st_bufferobj_alloc(GLcontext *ctx, GLuint name, GLenum target) { - struct st_context *st = st_context(ctx); struct st_buffer_object *st_obj = CALLOC_STRUCT(st_buffer_object); if (!st_obj) @@ -63,8 +62,6 @@ st_bufferobj_alloc(GLcontext *ctx, GLuint name, GLenum target) _mesa_initialize_buffer_object(&st_obj->Base, name, target); - st_obj->buffer = st->pipe->winsys->buffer_create( st->pipe->winsys, 32, 0, 0 ); - return &st_obj->Base; } @@ -88,6 +85,57 @@ st_bufferobj_free(GLcontext *ctx, struct gl_buffer_object *obj) +/** + * Replace data in a subrange of buffer object. If the data range + * specified by size + offset extends beyond the end of the buffer or + * if data is NULL, no copy is performed. + * Called via glBufferSubDataARB(). + */ +static void +st_bufferobj_subdata(GLcontext *ctx, + GLenum target, + GLintptrARB offset, + GLsizeiptrARB size, + const GLvoid * data, struct gl_buffer_object *obj) +{ + struct pipe_context *pipe = st_context(ctx)->pipe; + struct st_buffer_object *st_obj = st_buffer_object(obj); + char *map; + + if (offset >= st_obj->size || size > (st_obj->size - offset)) + return; + + map = pipe->winsys->buffer_map(pipe->winsys, st_obj->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE); + memcpy(map + offset, data, size); + pipe->winsys->buffer_unmap(pipe->winsys, st_obj->buffer); +} + + +/** + * Called via glGetBufferSubDataARB(). + */ +static void +st_bufferobj_get_subdata(GLcontext *ctx, + GLenum target, + GLintptrARB offset, + GLsizeiptrARB size, + GLvoid * data, struct gl_buffer_object *obj) +{ + struct pipe_context *pipe = st_context(ctx)->pipe; + struct st_buffer_object *st_obj = st_buffer_object(obj); + char *map; + + if (offset >= st_obj->size || size > (st_obj->size - offset)) + return; + + map = pipe->winsys->buffer_map(pipe->winsys, st_obj->buffer, + PIPE_BUFFER_USAGE_CPU_READ); + memcpy(data, map + offset, size); + pipe->winsys->buffer_unmap(pipe->winsys, st_obj->buffer); +} + + /** * Allocate space for and store data in a buffer object. Any data that was * previously stored in the buffer object is lost. If data is NULL, @@ -102,7 +150,8 @@ st_bufferobj_data(GLcontext *ctx, GLenum usage, struct gl_buffer_object *obj) { - struct pipe_context *pipe = st_context(ctx)->pipe; + struct st_context *st = st_context(ctx); + struct pipe_context *pipe = st->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); unsigned buffer_usage; @@ -124,46 +173,15 @@ st_bufferobj_data(GLcontext *ctx, buffer_usage = 0; } - pipe->winsys->buffer_data( pipe->winsys, st_obj->buffer, - size, data, - buffer_usage ); -} + pipe->winsys->buffer_reference( pipe->winsys, &st_obj->buffer, NULL ); + st_obj->buffer = pipe->winsys->buffer_create( pipe->winsys, 32, buffer_usage, + size ); -/** - * Replace data in a subrange of buffer object. If the data range - * specified by size + offset extends beyond the end of the buffer or - * if data is NULL, no copy is performed. - * Called via glBufferSubDataARB(). - */ -static void -st_bufferobj_subdata(GLcontext *ctx, - GLenum target, - GLintptrARB offset, - GLsizeiptrARB size, - const GLvoid * data, struct gl_buffer_object *obj) -{ - struct pipe_context *pipe = st_context(ctx)->pipe; - struct st_buffer_object *st_obj = st_buffer_object(obj); - - pipe->winsys->buffer_subdata(pipe->winsys, st_obj->buffer, offset, size, data); -} - - -/** - * Called via glGetBufferSubDataARB(). - */ -static void -st_bufferobj_get_subdata(GLcontext *ctx, - GLenum target, - GLintptrARB offset, - GLsizeiptrARB size, - GLvoid * data, struct gl_buffer_object *obj) -{ - struct pipe_context *pipe = st_context(ctx)->pipe; - struct st_buffer_object *st_obj = st_buffer_object(obj); + st_obj->size = size; - pipe->winsys->buffer_get_subdata(pipe->winsys, st_obj->buffer, offset, size, data); + if (data) + st_bufferobj_subdata(ctx, target, 0, size, data, obj); } @@ -180,15 +198,15 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access, switch (access) { case GL_WRITE_ONLY: - flags = PIPE_BUFFER_FLAG_WRITE; + flags = PIPE_BUFFER_USAGE_CPU_WRITE; break; case GL_READ_ONLY: - flags = PIPE_BUFFER_FLAG_READ; + flags = PIPE_BUFFER_USAGE_CPU_READ; break; case GL_READ_WRITE: /* fall-through */ default: - flags = PIPE_BUFFER_FLAG_READ | PIPE_BUFFER_FLAG_WRITE; + flags = PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE; break; } diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.h b/src/mesa/state_tracker/st_cb_bufferobjects.h index 2090a743e0..15003ae15d 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.h +++ b/src/mesa/state_tracker/st_cb_bufferobjects.h @@ -40,6 +40,7 @@ struct st_buffer_object { struct gl_buffer_object Base; struct pipe_buffer_handle *buffer; + GLsizeiptrARB size; }; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 5ae21c93f9..2b6f8743f3 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -164,13 +164,6 @@ struct st_context struct st_fragment_program *combined_prog; } bitmap; - /** - * Buffer object which stores the ctx->Current.Attrib[] values. - * Used for vertex array drawing when we we need an attribute for - * which there's no enabled array. - */ - struct pipe_buffer_handle *default_attrib_buffer; - struct cso_cache *cache; }; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 71acf6e862..371d4e7966 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -190,29 +190,6 @@ pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized) } -/** - * The default attribute buffer is basically a copy of the - * ctx->Current.Attrib[] array. It's used when the vertex program - * references an attribute for which we don't have a VBO/array. - */ -static void -create_default_attribs_buffer(struct st_context *st) -{ - struct pipe_context *pipe = st->pipe; - /* XXX don't hardcode magic 32 here */ - st->default_attrib_buffer = pipe->winsys->buffer_create( pipe->winsys, 32, 0, 0 ); -} - - -static void -destroy_default_attribs_buffer(struct st_context *st) -{ - struct pipe_context *pipe = st->pipe; - pipe->winsys->buffer_reference(pipe->winsys, - &st->default_attrib_buffer, NULL); -} - - /** * This function gets plugged into the VBO module and is called when * we have something to render. @@ -399,10 +376,14 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, } /* XXX create one-time */ - vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); - pipe->winsys->buffer_data(pipe->winsys, vbuf, - vertex_bytes, verts, - PIPE_BUFFER_USAGE_VERTEX); + vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_VERTEX, vertex_bytes); + assert(vbuf); + + memcpy(pipe->winsys->buffer_map(pipe->winsys, vbuf, + PIPE_BUFFER_USAGE_CPU_WRITE), + verts, vertex_bytes); + pipe->winsys->buffer_unmap(pipe->winsys, vbuf); /* tell pipe about the vertex buffer */ vbuffer.buffer = vbuf; @@ -568,7 +549,7 @@ st_feedback_draw_vbo(GLcontext *ctx, /* map the attrib buffer */ map = pipe->winsys->buffer_map(pipe->winsys, vbuffer[attr].buffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, attr, map); } @@ -592,7 +573,7 @@ st_feedback_draw_vbo(GLcontext *ctx, map = pipe->winsys->buffer_map(pipe->winsys, index_buffer_handle, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(draw, indexSize, map); } else { @@ -604,7 +585,7 @@ st_feedback_draw_vbo(GLcontext *ctx, /* map constant buffers */ mapped_constants = winsys->buffer_map(winsys, st->state.constants[PIPE_SHADER_VERTEX].buffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_constant_buffer(st->draw, mapped_constants); @@ -640,16 +621,12 @@ void st_init_draw( struct st_context *st ) { GLcontext *ctx = st->ctx; - /* actually, not used here, but elsewhere */ - create_default_attribs_buffer(st); - vbo_set_draw_func(ctx, st_draw_vbo); } void st_destroy_draw( struct st_context *st ) { - destroy_default_attribs_buffer(st); } -- cgit v1.2.3 From 1e0d30a515e4cac891b6c590f12a33e0e8a8e295 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 25 Jan 2008 20:53:31 +0000 Subject: gallium: rename pipe_buffer_handle to pipe_buffer, rework pipebuffer/ code Provide an actual definition of the pipe_buffer struct, containing the parameters used to create the buffer, and its refcount. Shift refcounting buffers out of the winsys interface, similar to surfaces & textures. Rework pipebuffer/ to reflect the fact these changes, and also Michel's reworking of the buffer interface. --- src/mesa/drivers/dri/intel_winsys/intel_winsys.h | 23 ++-- .../drivers/dri/intel_winsys/intel_winsys_i915.c | 2 +- .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 53 ++++---- src/mesa/drivers/x11/xm_winsys.c | 24 ++-- src/mesa/pipe/cell/ppu/cell_draw_arrays.c | 2 +- src/mesa/pipe/cell/ppu/cell_draw_arrays.h | 2 +- src/mesa/pipe/cell/ppu/cell_texture.h | 2 +- src/mesa/pipe/failover/fo_context.c | 2 +- src/mesa/pipe/i915simple/i915_blit.c | 6 +- src/mesa/pipe/i915simple/i915_blit.h | 6 +- src/mesa/pipe/i915simple/i915_context.c | 2 +- src/mesa/pipe/i915simple/i915_context.h | 4 +- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 3 +- src/mesa/pipe/i915simple/i915_state_emit.c | 2 +- src/mesa/pipe/i915simple/i915_surface.c | 2 +- src/mesa/pipe/i915simple/i915_texture.c | 2 +- src/mesa/pipe/i915simple/i915_winsys.h | 4 +- src/mesa/pipe/i965simple/brw_blit.c | 6 +- src/mesa/pipe/i965simple/brw_blit.h | 8 +- src/mesa/pipe/i965simple/brw_context.h | 6 +- src/mesa/pipe/i965simple/brw_draw.c | 4 +- src/mesa/pipe/i965simple/brw_draw.h | 2 +- src/mesa/pipe/i965simple/brw_draw_upload.c | 4 +- src/mesa/pipe/i965simple/brw_state.h | 2 +- src/mesa/pipe/i965simple/brw_state_pool.c | 7 +- src/mesa/pipe/i965simple/brw_surface.c | 2 +- src/mesa/pipe/i965simple/brw_tex_layout.c | 3 +- src/mesa/pipe/i965simple/brw_winsys.h | 8 +- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 2 +- src/mesa/pipe/p_context.h | 2 +- src/mesa/pipe/p_inlines.h | 43 +++++-- src/mesa/pipe/p_state.h | 23 +++- src/mesa/pipe/p_winsys.h | 17 +-- src/mesa/pipe/pipebuffer/Makefile | 2 - src/mesa/pipe/pipebuffer/pb_buffer.c | 49 ++++++-- src/mesa/pipe/pipebuffer/pb_buffer.h | 111 ++++++++-------- src/mesa/pipe/pipebuffer/pb_buffer_client.c | 60 ++++----- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 102 ++++++--------- src/mesa/pipe/pipebuffer/pb_buffer_fenced.h | 10 +- src/mesa/pipe/pipebuffer/pb_buffer_handle.c | 140 --------------------- src/mesa/pipe/pipebuffer/pb_buffer_handle.h | 120 ------------------ src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 41 +++--- src/mesa/pipe/pipebuffer/pb_buffer_null.c | 98 --------------- src/mesa/pipe/pipebuffer/pb_bufmgr.h | 24 ++-- src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c | 36 +++--- src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 88 ++++++------- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 75 +++++------ src/mesa/pipe/softpipe/sp_context.c | 2 +- src/mesa/pipe/softpipe/sp_draw_arrays.c | 2 +- src/mesa/pipe/softpipe/sp_state.h | 2 +- src/mesa/pipe/softpipe/sp_state_fs.c | 7 +- src/mesa/pipe/softpipe/sp_texture.c | 4 +- src/mesa/pipe/softpipe/sp_texture.h | 2 +- src/mesa/pipe/xlib/xm_winsys.c | 80 +++++------- src/mesa/pipe/xlib/xm_winsys_aub.c | 44 +++---- src/mesa/pipe/xlib/xm_winsys_aub.h | 4 +- src/mesa/state_tracker/st_atom_constbuf.c | 3 +- src/mesa/state_tracker/st_cb_bufferobjects.c | 5 +- src/mesa/state_tracker/st_cb_bufferobjects.h | 4 +- src/mesa/state_tracker/st_cb_fbo.c | 4 +- src/mesa/state_tracker/st_context.c | 3 +- src/mesa/state_tracker/st_draw.c | 21 ++-- 62 files changed, 529 insertions(+), 894 deletions(-) delete mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_handle.c delete mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_handle.h delete mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_null.c (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h index 89e63e0a79..ffc40782be 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h @@ -28,10 +28,12 @@ #ifndef INTEL_WINSYS_H #define INTEL_WINSYS_H +#include "pipe/p_state.h" + struct intel_context; struct pipe_context; struct pipe_winsys; -struct pipe_buffer_handle; +struct pipe_buffer; struct _DriBufferObject; struct pipe_winsys * @@ -49,20 +51,21 @@ intel_create_i915simple( struct intel_context *intel, struct pipe_winsys *winsys ); +struct intel_buffer { + struct pipe_buffer base; + struct _DriBufferObject *driBO; +}; -/* Turn the pipe opaque buffer pointer into a dri_bufmgr opaque - * buffer pointer... - */ -static INLINE struct _DriBufferObject * -dri_bo( struct pipe_buffer_handle *bo ) +static INLINE struct intel_buffer * +intel_buffer( struct pipe_buffer *buf ) { - return (struct _DriBufferObject *)bo; + return (struct intel_buffer *)buf; } -static INLINE struct pipe_buffer_handle * -pipe_bo( struct _DriBufferObject *bo ) +static INLINE struct _DriBufferObject * +dri_bo( struct pipe_buffer *buf ) { - return (struct pipe_buffer_handle *)bo; + return intel_buffer(buf)->driBO; } diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c index 8e0eea4392..1ba6a9e1b2 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c @@ -85,7 +85,7 @@ static void intel_i915_batch_dword( struct i915_winsys *sws, } static void intel_i915_batch_reloc( struct i915_winsys *sws, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned access_flags, unsigned delta ) { diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 43ed0602a4..910c0d2cc5 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -43,6 +43,7 @@ #include "pipe/p_defines.h" #include "pipe/p_state.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" @@ -65,7 +66,7 @@ intel_pipe_winsys( struct pipe_winsys *winsys ) /* Most callbacks map direcly onto dri_bufmgr operations: */ static void *intel_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned flags ) { unsigned drm_flags = 0; @@ -80,26 +81,17 @@ static void *intel_buffer_map(struct pipe_winsys *winsys, } static void intel_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf) + struct pipe_buffer *buf) { driBOUnmap( dri_bo(buf) ); } static void -intel_buffer_reference(struct pipe_winsys *winsys, - struct pipe_buffer_handle **ptr, - struct pipe_buffer_handle *buf) +intel_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) { - if (*ptr) { - driBOUnReference( dri_bo(*ptr) ); - *ptr = NULL; - } - - if (buf) { - driBOReference( dri_bo(buf) ); - *ptr = buf; - } + driBOUnReference( dri_bo(buf) ); } @@ -107,16 +99,21 @@ intel_buffer_reference(struct pipe_winsys *winsys, * for all buffers. * Grabs the hardware lock! */ -static struct pipe_buffer_handle * +static struct pipe_buffer * intel_buffer_create(struct pipe_winsys *winsys, unsigned alignment, unsigned usage, unsigned size ) { - struct _DriBufferObject *buffer; + struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); unsigned flags = 0; + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + if (usage & (PIPE_BUFFER_USAGE_VERTEX /*| IWS_BUFFER_USAGE_LOCAL*/)) { flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; } else { @@ -143,20 +140,24 @@ intel_buffer_create(struct pipe_winsys *winsys, #endif driGenBuffers( iws->regionPool, - "pipe buffer", 1, &buffer, alignment, flags, 0 ); - driBOData( buffer, size, NULL, 0 ); - return pipe_bo(buffer); + "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); + + driBOData( buffer->driBO, size, NULL, 0 ); + + return &buffer->base; } -static struct pipe_buffer_handle * +static struct pipe_buffer * intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) { - struct _DriBufferObject *buffer; + struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + driGenUserBuffer( iws->regionPool, - "pipe user buffer", &buffer, ptr, bytes); - return pipe_bo(buffer); + "pipe user buffer", &buffer->driBO, ptr, bytes); + + return &buffer->base; } @@ -224,7 +225,7 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, return -1; if(ret) { - winsys->buffer_reference(winsys, &surf->buffer, NULL); + pipe_buffer_reference(winsys, &surf->buffer, NULL); return ret; } @@ -239,7 +240,7 @@ intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) surf->refcount--; if (surf->refcount == 0) { if (surf->buffer) - winsys->buffer_reference(winsys, &surf->buffer, NULL); + pipe_buffer_reference(winsys, &surf->buffer, NULL); free(surf); } *s = NULL; @@ -279,7 +280,7 @@ intel_create_pipe_winsys( int fd ) iws->winsys.user_buffer_create = intel_user_buffer_create; iws->winsys.buffer_map = intel_buffer_map; iws->winsys.buffer_unmap = intel_buffer_unmap; - iws->winsys.buffer_reference = intel_buffer_reference; + iws->winsys.buffer_destroy = intel_buffer_destroy; iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; iws->winsys.printf = intel_printf; iws->winsys.get_name = intel_get_name; diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index dafbe96a1e..a690df2772 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -71,15 +71,15 @@ struct xm_buffer * buffer pointer... */ static inline struct xm_buffer * -xm_bo( struct pipe_buffer_handle *bo ) +xm_bo( struct pipe_buffer *bo ) { return (struct xm_buffer *) bo; } -static inline struct pipe_buffer_handle * +static inline struct pipe_buffer * pipe_bo( struct xm_buffer *bo ) { - return (struct pipe_buffer_handle *) bo; + return (struct pipe_buffer *) bo; } /* Turn a softpipe winsys into an xm/softpipe winsys: @@ -94,7 +94,7 @@ xm_winsys(struct softpipe_winsys *sws) /* Most callbacks map direcly onto dri_bufmgr operations: */ static void * -xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, +xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buf, unsigned flags) { struct xm_buffer *xm_buf = xm_bo(buf); @@ -103,7 +103,7 @@ xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, } static void -xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) +xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) { struct xm_buffer *xm_buf = xm_bo(buf); xm_buf->mapped = NULL; @@ -111,8 +111,8 @@ xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) static void xm_buffer_reference(struct pipe_winsys *pws, - struct pipe_buffer_handle **ptr, - struct pipe_buffer_handle *buf) + struct pipe_buffer **ptr, + struct pipe_buffer *buf) { if (*ptr) { struct xm_buffer *oldBuf = xm_bo(*ptr); @@ -139,7 +139,7 @@ xm_buffer_reference(struct pipe_winsys *pws, } static void -xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, +xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer *buf, unsigned size, const void *data, unsigned usage) { struct xm_buffer *xm_buf = xm_bo(buf); @@ -155,7 +155,7 @@ xm_buffer_data(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, } static void -xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, +xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer *buf, unsigned long offset, unsigned long size, const void *data) { struct xm_buffer *xm_buf = xm_bo(buf); @@ -166,7 +166,7 @@ xm_buffer_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, } static void -xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, +xm_buffer_get_subdata(struct pipe_winsys *pws, struct pipe_buffer *buf, unsigned long offset, unsigned long size, void *data) { const struct xm_buffer *xm_buf = xm_bo(buf); @@ -209,7 +209,7 @@ xm_get_name(struct pipe_winsys *pws) } -static struct pipe_buffer_handle * +static struct pipe_buffer * xm_buffer_create(struct pipe_winsys *pws, unsigned alignment, unsigned flags, @@ -224,7 +224,7 @@ xm_buffer_create(struct pipe_winsys *pws, /** * Create buffer which wraps user-space data. */ -static struct pipe_buffer_handle * +static struct pipe_buffer * xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c index 537cec8785..8286da712c 100644 --- a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c +++ b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c @@ -89,7 +89,7 @@ cell_draw_arrays(struct pipe_context *pipe, unsigned mode, */ boolean cell_draw_elements(struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, + struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count) { diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.h b/src/mesa/pipe/cell/ppu/cell_draw_arrays.h index bd5b703f3b..d5df4aa05f 100644 --- a/src/mesa/pipe/cell/ppu/cell_draw_arrays.h +++ b/src/mesa/pipe/cell/ppu/cell_draw_arrays.h @@ -33,7 +33,7 @@ boolean cell_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count); boolean cell_draw_elements(struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, + struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count); diff --git a/src/mesa/pipe/cell/ppu/cell_texture.h b/src/mesa/pipe/cell/ppu/cell_texture.h index 97d8bd1230..ef5808c086 100644 --- a/src/mesa/pipe/cell/ppu/cell_texture.h +++ b/src/mesa/pipe/cell/ppu/cell_texture.h @@ -44,7 +44,7 @@ struct cell_texture /* The data is held here: */ - struct pipe_buffer_handle *buffer; + struct pipe_buffer *buffer; unsigned long buffer_size; }; diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index a1291f583b..3799c4dfb0 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -46,7 +46,7 @@ static void failover_destroy( struct pipe_context *pipe ) static boolean failover_draw_elements( struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, + struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count) { diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/mesa/pipe/i915simple/i915_blit.c index 6e95313a0d..d49876f970 100644 --- a/src/mesa/pipe/i915simple/i915_blit.c +++ b/src/mesa/pipe/i915simple/i915_blit.c @@ -38,7 +38,7 @@ void i915_fill_blit(struct i915_context *i915, unsigned cpp, short dst_pitch, - struct pipe_buffer_handle *dst_buffer, + struct pipe_buffer *dst_buffer, unsigned dst_offset, short x, short y, short w, short h, @@ -87,10 +87,10 @@ void i915_copy_blit( struct i915_context *i915, unsigned cpp, short src_pitch, - struct pipe_buffer_handle *src_buffer, + struct pipe_buffer *src_buffer, unsigned src_offset, short dst_pitch, - struct pipe_buffer_handle *dst_buffer, + struct pipe_buffer *dst_buffer, unsigned dst_offset, short src_x, short src_y, short dst_x, short dst_y, diff --git a/src/mesa/pipe/i915simple/i915_blit.h b/src/mesa/pipe/i915simple/i915_blit.h index 7ea4e979ec..d7a66be10a 100644 --- a/src/mesa/pipe/i915simple/i915_blit.h +++ b/src/mesa/pipe/i915simple/i915_blit.h @@ -33,10 +33,10 @@ extern void i915_copy_blit(struct i915_context *i915, unsigned cpp, short src_pitch, - struct pipe_buffer_handle *src_buffer, + struct pipe_buffer *src_buffer, unsigned src_offset, short dst_pitch, - struct pipe_buffer_handle *dst_buffer, + struct pipe_buffer *dst_buffer, unsigned dst_offset, short srcx, short srcy, short dstx, short dsty, @@ -45,7 +45,7 @@ extern void i915_copy_blit(struct i915_context *i915, extern void i915_fill_blit(struct i915_context *i915, unsigned cpp, short dst_pitch, - struct pipe_buffer_handle *dst_buffer, + struct pipe_buffer *dst_buffer, unsigned dst_offset, short x, short y, short w, short h, unsigned color); diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c index c2b96d2554..497623a700 100644 --- a/src/mesa/pipe/i915simple/i915_context.c +++ b/src/mesa/pipe/i915simple/i915_context.c @@ -166,7 +166,7 @@ static void i915_destroy( struct pipe_context *pipe ) static boolean i915_draw_elements( struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, + struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned prim, unsigned start, unsigned count) { diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h index 2f1f036993..b4ea63c3e7 100644 --- a/src/mesa/pipe/i915simple/i915_context.h +++ b/src/mesa/pipe/i915simple/i915_context.h @@ -172,7 +172,7 @@ struct i915_texture { /* The data is held here: */ - struct pipe_buffer_handle *buffer; + struct pipe_buffer *buffer; }; struct i915_context @@ -204,7 +204,7 @@ struct i915_context unsigned *batch_start; /** Vertex buffer */ - struct pipe_buffer_handle *vbo; + struct pipe_buffer *vbo; struct i915_state current; unsigned hardware_dirty; diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 0887f9a1c6..39154b2488 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -42,6 +42,7 @@ #include "pipe/draw/draw_vbuf.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "i915_context.h" @@ -192,7 +193,7 @@ i915_vbuf_render_release_vertices( struct vbuf_render *render, assert(i915->vbo); winsys->buffer_unmap(winsys, i915->vbo); - winsys->buffer_reference(winsys, &i915->vbo, NULL); + pipe_buffer_reference(winsys, &i915->vbo, NULL); } diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 8598eacc9c..657f523893 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -278,7 +278,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(enabled); for (unit = 0; unit < I915_TEX_UNITS; unit++) { if (enabled & (1 << unit)) { - struct pipe_buffer_handle *buf = + struct pipe_buffer *buf = i915->texture[unit]->buffer; uint offset = 0; assert(buf); diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index e625f5a68c..8ef02b99cd 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -65,7 +65,7 @@ i915_get_tex_surface(struct pipe_context *pipe, if (ps) { assert(ps->refcount); assert(ps->winsys); - pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c index bf80e18233..61944fe7d9 100644 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -523,7 +523,7 @@ i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); */ - pipe->winsys->buffer_reference(pipe->winsys, &tex->buffer, NULL); + pipe_buffer_reference(pipe->winsys, &tex->buffer, NULL); for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) if (tex->image_offset[i]) diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h index 2c0f335d34..fe49710852 100644 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ b/src/mesa/pipe/i915simple/i915_winsys.h @@ -50,7 +50,7 @@ * etc. */ -struct pipe_buffer_handle; +struct pipe_buffer; struct pipe_winsys; @@ -93,7 +93,7 @@ struct i915_winsys { * I915_BUFFER_ACCESS_READ macros. */ void (*batch_reloc)( struct i915_winsys *sws, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned access_flags, unsigned delta ); diff --git a/src/mesa/pipe/i965simple/brw_blit.c b/src/mesa/pipe/i965simple/brw_blit.c index 4692129e40..bbd366294f 100644 --- a/src/mesa/pipe/i965simple/brw_blit.c +++ b/src/mesa/pipe/i965simple/brw_blit.c @@ -42,7 +42,7 @@ void brw_fill_blit(struct brw_context *brw, unsigned cpp, short dst_pitch, - struct pipe_buffer_handle *dst_buffer, + struct pipe_buffer *dst_buffer, unsigned dst_offset, boolean dst_tiled, short x, short y, @@ -113,11 +113,11 @@ static unsigned translate_raster_op(unsigned logicop) void brw_copy_blit(struct brw_context *brw, unsigned cpp, short src_pitch, - struct pipe_buffer_handle *src_buffer, + struct pipe_buffer *src_buffer, unsigned src_offset, boolean src_tiled, short dst_pitch, - struct pipe_buffer_handle *dst_buffer, + struct pipe_buffer *dst_buffer, unsigned dst_offset, boolean dst_tiled, short src_x, short src_y, diff --git a/src/mesa/pipe/i965simple/brw_blit.h b/src/mesa/pipe/i965simple/brw_blit.h index 371a135375..7f17a70173 100644 --- a/src/mesa/pipe/i965simple/brw_blit.h +++ b/src/mesa/pipe/i965simple/brw_blit.h @@ -3,13 +3,13 @@ #include "pipe/p_compiler.h" -struct pipe_buffer_handle; +struct pipe_buffer; struct brw_context; void brw_fill_blit(struct brw_context *intel, unsigned cpp, short dst_pitch, - struct pipe_buffer_handle *dst_buffer, + struct pipe_buffer *dst_buffer, unsigned dst_offset, boolean dst_tiled, short x, short y, @@ -18,11 +18,11 @@ void brw_fill_blit(struct brw_context *intel, void brw_copy_blit(struct brw_context *intel, unsigned cpp, short src_pitch, - struct pipe_buffer_handle *src_buffer, + struct pipe_buffer *src_buffer, unsigned src_offset, boolean src_tiled, short dst_pitch, - struct pipe_buffer_handle *dst_buffer, + struct pipe_buffer *dst_buffer, unsigned dst_offset, boolean dst_tiled, short src_x, short src_y, diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h index c610a435e0..65664d853d 100644 --- a/src/mesa/pipe/i965simple/brw_context.h +++ b/src/mesa/pipe/i965simple/brw_context.h @@ -260,7 +260,7 @@ struct brw_texture { /* The data is held here: */ - struct pipe_buffer_handle *buffer; + struct pipe_buffer *buffer; }; /* Data about a particular attempt to compile a program. Note that @@ -350,7 +350,7 @@ struct brw_surface_binding_table { struct brw_cache; struct brw_mem_pool { - struct pipe_buffer_handle *buffer; + struct pipe_buffer *buffer; unsigned size; unsigned offset; /* offset of first free byte */ @@ -615,7 +615,7 @@ struct brw_context unsigned nr_surfaces; unsigned max_threads; - struct pipe_buffer_handle *scratch_buffer; + struct pipe_buffer *scratch_buffer; unsigned scratch_buffer_size; unsigned sampler_count; diff --git a/src/mesa/pipe/i965simple/brw_draw.c b/src/mesa/pipe/i965simple/brw_draw.c index acfb524a30..7598e3dc8a 100644 --- a/src/mesa/pipe/i965simple/brw_draw.c +++ b/src/mesa/pipe/i965simple/brw_draw.c @@ -144,7 +144,7 @@ static boolean brw_emit_prim( struct brw_context *brw, * fallback conditions. */ static boolean brw_try_draw_elements( struct pipe_context *pipe, - struct pipe_buffer_handle *index_buffer, + struct pipe_buffer *index_buffer, unsigned index_size, unsigned mode, unsigned start, @@ -183,7 +183,7 @@ static boolean brw_try_draw_elements( struct pipe_context *pipe, static boolean brw_draw_elements( struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, + struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, diff --git a/src/mesa/pipe/i965simple/brw_draw.h b/src/mesa/pipe/i965simple/brw_draw.h index 053f2efb9d..62fe0d5d0e 100644 --- a/src/mesa/pipe/i965simple/brw_draw.h +++ b/src/mesa/pipe/i965simple/brw_draw.h @@ -42,7 +42,7 @@ boolean brw_upload_vertices( struct brw_context *brw, unsigned max_index ); boolean brw_upload_indices(struct brw_context *brw, - const struct pipe_buffer_handle *index_buffer, + const struct pipe_buffer *index_buffer, int ib_size, int start, int count); boolean brw_upload_vertex_buffers( struct brw_context *brw ); diff --git a/src/mesa/pipe/i965simple/brw_draw_upload.c b/src/mesa/pipe/i965simple/brw_draw_upload.c index 43e53914e9..aa85d93866 100644 --- a/src/mesa/pipe/i965simple/brw_draw_upload.c +++ b/src/mesa/pipe/i965simple/brw_draw_upload.c @@ -47,7 +47,7 @@ struct brw_array_state { unsigned dword; } vb0; - struct pipe_buffer_handle *buffer; + struct pipe_buffer *buffer; unsigned offset; unsigned max_index; @@ -272,7 +272,7 @@ boolean brw_upload_vertex_elements( struct brw_context *brw ) } boolean brw_upload_indices( struct brw_context *brw, - const struct pipe_buffer_handle *index_buffer, + const struct pipe_buffer *index_buffer, int ib_size, int start, int count) { /* Emit the indexbuffer packet: diff --git a/src/mesa/pipe/i965simple/brw_state.h b/src/mesa/pipe/i965simple/brw_state.h index 8b4d7aabf0..258e9a556e 100644 --- a/src/mesa/pipe/i965simple/brw_state.h +++ b/src/mesa/pipe/i965simple/brw_state.h @@ -106,7 +106,7 @@ boolean brw_search_cache( struct brw_cache *cache, void brw_init_caches( struct brw_context *brw ); void brw_destroy_caches( struct brw_context *brw ); -static inline struct pipe_buffer_handle *brw_cache_buffer(struct brw_context *brw, +static inline struct pipe_buffer *brw_cache_buffer(struct brw_context *brw, enum brw_cache_id id) { return brw->cache[id].pool->buffer; diff --git a/src/mesa/pipe/i965simple/brw_state_pool.c b/src/mesa/pipe/i965simple/brw_state_pool.c index 2f930be837..7c67f0ee25 100644 --- a/src/mesa/pipe/i965simple/brw_state_pool.c +++ b/src/mesa/pipe/i965simple/brw_state_pool.c @@ -44,6 +44,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "brw_context.h" #include "brw_state.h" @@ -101,9 +102,9 @@ static void brw_destroy_pool( struct brw_context *brw, { struct brw_mem_pool *pool = &brw->pool[pool_id]; - pool->brw->pipe.winsys->buffer_reference( pool->brw->pipe.winsys, - &pool->buffer, - NULL ); + pipe_buffer_reference( pool->brw->pipe.winsys, + &pool->buffer, + NULL ); } diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c index 76b8c73d5c..252aec7d29 100644 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -64,7 +64,7 @@ brw_get_tex_surface(struct pipe_context *pipe, if (ps) { assert(ps->format); assert(ps->refcount); - pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.c b/src/mesa/pipe/i965simple/brw_tex_layout.c index eadacbf09e..b8b6b579e2 100644 --- a/src/mesa/pipe/i965simple/brw_tex_layout.c +++ b/src/mesa/pipe/i965simple/brw_tex_layout.c @@ -39,6 +39,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "brw_context.h" @@ -341,7 +342,7 @@ brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); */ - pipe->winsys->buffer_reference(pipe->winsys, &tex->buffer, NULL); + pipe_buffer_reference(pipe->winsys, &tex->buffer, NULL); for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) if (tex->image_offset[i]) diff --git a/src/mesa/pipe/i965simple/brw_winsys.h b/src/mesa/pipe/i965simple/brw_winsys.h index b60f63aa5b..3523a58614 100644 --- a/src/mesa/pipe/i965simple/brw_winsys.h +++ b/src/mesa/pipe/i965simple/brw_winsys.h @@ -50,7 +50,7 @@ * etc. */ -struct pipe_buffer_handle; +struct pipe_buffer; struct pipe_fence_handle; struct pipe_winsys; @@ -136,7 +136,7 @@ struct brw_winsys { * I915_BUFFER_ACCESS_READ macros. */ void (*batch_reloc)(struct brw_winsys *sws, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned access_flags, unsigned delta); @@ -159,7 +159,7 @@ struct brw_winsys { * simulator: */ void (*buffer_subdata_typed)(struct brw_winsys *sws, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned long offset, unsigned long size, const void *data, @@ -170,7 +170,7 @@ struct brw_winsys { * of places yet: */ unsigned (*get_buffer_offset)( struct brw_winsys *sws, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned flags ); }; diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index fc40e0438c..cbb4f2efd3 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -127,7 +127,7 @@ static unsigned translate_tex_format( enum pipe_format pipe_format ) } static unsigned brw_buffer_offset(struct brw_context *brw, - struct pipe_buffer_handle *buffer) + struct pipe_buffer *buffer) { return brw->winsys->get_buffer_offset(brw->winsys, buffer, diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 7a18d48865..37464c88a1 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -74,7 +74,7 @@ struct pipe_context { unsigned mode, unsigned start, unsigned count); boolean (*draw_elements)( struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, + struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count); diff --git a/src/mesa/pipe/p_inlines.h b/src/mesa/pipe/p_inlines.h index 8ee0820f92..ebf6ed86bc 100644 --- a/src/mesa/pipe/p_inlines.h +++ b/src/mesa/pipe/p_inlines.h @@ -56,20 +56,38 @@ pipe_surface_unmap(struct pipe_surface *surface) static INLINE void pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) { - assert(ptr); - if (*ptr) { + /* bump the refcount first */ + if (surf) + surf->refcount++; + + if (*ptr /* && --(*ptr)->refcount == 0 */) { struct pipe_winsys *winsys = (*ptr)->winsys; winsys->surface_release(winsys, ptr); assert(!*ptr); } - if (surf) { - /* reference the new thing */ - surf->refcount++; - *ptr = surf; - } + + *ptr = surf; +} + + +/* XXX: thread safety issues! + */ +static INLINE void +pipe_buffer_reference(struct pipe_winsys *winsys, + struct pipe_buffer **ptr, + struct pipe_buffer *buf) +{ + if (buf) + buf->refcount++; + + if (*ptr && --(*ptr)->refcount == 0) + winsys->buffer_destroy( winsys, *ptr ); + + *ptr = buf; } + /** * \sa pipe_surface_reference */ @@ -78,15 +96,16 @@ pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, struct pipe_texture *pt) { assert(ptr); + + if (pt) + pt->refcount++; + if (*ptr) { pipe->texture_release(pipe, ptr); assert(!*ptr); } - if (pt) { - /* reference the new thing */ - pt->refcount++; - *ptr = pt; - } + + *ptr = pt; } diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 146a479cf3..83ca43f678 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -60,8 +60,21 @@ struct pipe_surface; struct pipe_winsys; -/* opaque type */ -struct pipe_buffer_handle; + + +/** + * The driver will certainly subclass this to include actual memory + * management information. + */ +struct pipe_buffer { + unsigned alignment; + unsigned usage; + unsigned size; + + /** Reference count */ + unsigned refcount; +}; + @@ -129,7 +142,7 @@ struct pipe_clip_state { * Constants for vertex/fragment shaders */ struct pipe_constant_buffer { - struct pipe_buffer_handle *buffer; + struct pipe_buffer *buffer; unsigned size; /** in bytes */ }; @@ -240,7 +253,7 @@ struct pipe_sampler_state */ struct pipe_surface { - struct pipe_buffer_handle *buffer; /**< driver private buffer handle */ + struct pipe_buffer *buffer; /**< driver private buffer handle */ enum pipe_format format; /**< PIPE_FORMAT_x */ unsigned status; /**< PIPE_SURFACE_STATUS_x */ unsigned clear_value; /**< may be temporary */ @@ -290,7 +303,7 @@ struct pipe_vertex_buffer unsigned pitch:11; /**< stride to same attrib in next vertex, in bytes */ unsigned max_index; /**< number of vertices in this buffer */ unsigned buffer_offset; /**< offset to start of data in buffer, in bytes */ - struct pipe_buffer_handle *buffer; /**< the actual buffer */ + struct pipe_buffer *buffer; /**< the actual buffer */ }; diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 9742f59a4d..95e3684008 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -39,9 +39,6 @@ */ -/** Opaque type for a buffer */ -struct pipe_buffer_handle; - /** Opaque type */ struct pipe_fence_handle; @@ -103,13 +100,13 @@ struct pipe_winsys * usage argument is only an optimization hint, not a guarantee, therefore * proper behavior must be observed in all circumstances. */ - struct pipe_buffer_handle *(*buffer_create)( struct pipe_winsys *sws, + struct pipe_buffer *(*buffer_create)( struct pipe_winsys *sws, unsigned alignment, unsigned usage, unsigned size ); /** Create a buffer that wraps user-space data */ - struct pipe_buffer_handle *(*user_buffer_create)(struct pipe_winsys *sws, + struct pipe_buffer *(*user_buffer_create)(struct pipe_winsys *sws, void *ptr, unsigned bytes); @@ -118,16 +115,14 @@ struct pipe_winsys * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. */ void *(*buffer_map)( struct pipe_winsys *sws, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned usage ); void (*buffer_unmap)( struct pipe_winsys *sws, - struct pipe_buffer_handle *buf ); + struct pipe_buffer *buf ); - /** Set ptr = buf, with reference counting */ - void (*buffer_reference)( struct pipe_winsys *sws, - struct pipe_buffer_handle **ptr, - struct pipe_buffer_handle *buf ); + void (*buffer_destroy)( struct pipe_winsys *sws, + struct pipe_buffer *buf ); /** Set ptr = fence, with reference counting */ diff --git a/src/mesa/pipe/pipebuffer/Makefile b/src/mesa/pipe/pipebuffer/Makefile index 061d8a060f..0d23e8b8f1 100644 --- a/src/mesa/pipe/pipebuffer/Makefile +++ b/src/mesa/pipe/pipebuffer/Makefile @@ -8,10 +8,8 @@ LIBNAME = pipebuffer DRIVER_SOURCES = \ pb_buffer.c \ pb_buffer_client.c \ - pb_buffer_handle.c \ pb_buffer_fenced.c \ pb_buffer_malloc.c \ - pb_buffer_null.c \ pb_bufmgr_fenced.c \ pb_bufmgr_mm.c \ pb_bufmgr_pool.c diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.c b/src/mesa/pipe/pipebuffer/pb_buffer.c index 99c960b697..90ab9044ff 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer.c @@ -34,19 +34,44 @@ #include "pb_buffer.h" +#include "pipe/p_winsys.h" -void -buffer_reference(struct pipe_buffer **dst, - struct pipe_buffer *src) + +static void * +pb_winsys_map(struct pipe_winsys *winsys, + struct pipe_buffer *ws_buf, + unsigned flags) +{ + struct pb_buffer *buf = pb_buffer(ws_buf); + + return buf->vtbl->map(buf, flags); +} + +static void +pb_winsys_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *ws_buf) +{ + struct pb_buffer *buf = pb_buffer(ws_buf); + + buf->vtbl->unmap(buf); +} + +static void +pb_winsys_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *ws_buf) +{ + struct pb_buffer *buf = pb_buffer(ws_buf); + + buf->vtbl->destroy(buf); +} + + + +void +pb_init_winsys(struct pipe_winsys *winsys) { - if(*dst != src) { - if (src) - src->vtbl->reference(src); - - if (*dst) - (*dst)->vtbl->release(*dst); - - *dst = src; - } + winsys->buffer_map = pb_winsys_map; + winsys->buffer_unmap = pb_winsys_unmap; + winsys->buffer_destroy = pb_winsys_destroy; } diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.h b/src/mesa/pipe/pipebuffer/pb_buffer.h index 0523531395..f909bded0d 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer.h @@ -47,53 +47,43 @@ #include #include +#include "pipe/p_state.h" -struct pipe_buffer_vtbl; - +struct pb_vtbl; /** - * Base class for all pipe buffers. + * Base class for all pb_* buffers. */ -struct pipe_buffer +struct pb_buffer { + struct pipe_buffer base; + /** * Pointer to the virtual function table. * * Avoid accessing this table directly. Use the inline functions below * instead to avoid mistakes. */ - const struct pipe_buffer_vtbl *vtbl; + const struct pb_vtbl *vtbl; }; - /** * Virtual function table for the buffer storage operations. * * Note that creation is not done through this table. */ -struct pipe_buffer_vtbl +struct pb_vtbl { - /** - * Add a reference to the buffer. - * - * This method can be a no-op for buffers that don't need reference - * counting. - */ - void (*reference)( struct pipe_buffer *buf ); - - /** - * Release a reference to this buffer and destroy it. - */ - void (*release)( struct pipe_buffer *buf ); + void (*destroy)( struct pb_buffer *buf ); /** * Map the entire data store of a buffer object into the client's address. * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. */ - void *(*map)( struct pipe_buffer *buf, + void *(*map)( struct pb_buffer *buf, unsigned flags ); - void (*unmap)( struct pipe_buffer *buf ); + void (*unmap)( struct pb_buffer *buf ); /** * Get the base buffer and the offset. @@ -106,29 +96,17 @@ struct pipe_buffer_vtbl * * Note that this will increase the reference count of the base buffer. */ - void (*get_base_buffer)( struct pipe_buffer *buf, - struct pipe_buffer **base_buf, + void (*get_base_buffer)( struct pb_buffer *buf, + struct pb_buffer **base_buf, unsigned *offset ); }; -/** *dst = src with reference counting */ -void -buffer_reference(struct pipe_buffer **dst, - struct pipe_buffer *src); - - -static inline void -buffer_release(struct pipe_buffer *buf) -{ - assert(buf); - buf->vtbl->release(buf); -} - - +/* Accessor functions for pb->vtbl: + */ static inline void * -buffer_map(struct pipe_buffer *buf, - unsigned flags) +pb_map(struct pb_buffer *buf, + unsigned flags) { assert(buf); return buf->vtbl->map(buf, flags); @@ -136,7 +114,7 @@ buffer_map(struct pipe_buffer *buf, static inline void -buffer_unmap(struct pipe_buffer *buf) +pb_unmap(struct pb_buffer *buf) { assert(buf); buf->vtbl->unmap(buf); @@ -144,32 +122,63 @@ buffer_unmap(struct pipe_buffer *buf) static inline void -buffer_get_base_buffer( struct pipe_buffer *buf, - struct pipe_buffer **base_buf, - unsigned *offset ) +pb_get_base_buffer( struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset ) { buf->vtbl->get_base_buffer(buf, base_buf, offset); } +static inline void +pb_destroy(struct pb_buffer *buf) +{ + assert(buf); + buf->vtbl->destroy(buf); +} + -/** Placeholder for empty buffers. */ -extern struct pipe_buffer null_buffer; /** - * Client buffers (also designated as user buffers) are just for convenience - * of the state tracker, so that it can masquerade its own data as a buffer. + * User buffers are special buffers that initially reference memory + * held by the user but which may if necessary copy that memory into + * device memory behind the scenes, for submission to hardware. + * + * These are particularly useful when the referenced data is never + * submitted to hardware at all, in the particular case of software + * vertex processing. */ -struct pipe_buffer * -client_buffer_create(void *data); +struct pb_buffer * +pb_user_buffer_create(void *data, unsigned bytes); /** * Malloc-based buffer to store data that can't be used by the graphics * hardware. */ -struct pipe_buffer * -malloc_buffer_create(unsigned size); +struct pb_buffer * +pb_malloc_buffer_create( unsigned alignment, + unsigned usage, + unsigned size ); + + + +static inline struct pipe_buffer * +pb_pipe_buffer( struct pb_buffer *pbuf ) +{ + return &pbuf->base; +} + +static inline struct pb_buffer * +pb_buffer( struct pipe_buffer *buf ) +{ + /* Could add a magic cookie check on debug builds. + */ + return (struct pb_buffer *)buf; +} + +void +pb_init_winsys(struct pipe_winsys *winsys); #endif /*PB_BUFFER_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_client.c b/src/mesa/pipe/pipebuffer/pb_buffer_client.c index bb7f5a9a94..6bf4745301 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_client.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_client.c @@ -35,36 +35,30 @@ #include "pb_buffer.h" +#include "pipe/p_util.h" -struct client_buffer +struct pb_user_buffer { - struct pipe_buffer base; + struct pb_buffer base; void *data; }; -extern const struct pipe_buffer_vtbl client_buffer_vtbl; +extern const struct pb_vtbl pb_user_buffer_vtbl; -static inline struct client_buffer * -client_buffer(struct pipe_buffer *buf) +static INLINE struct pb_user_buffer * +pb_user_buffer(struct pb_buffer *buf) { assert(buf); - assert(buf->vtbl == &client_buffer_vtbl); - return (struct client_buffer *)buf; + assert(buf->vtbl == &pb_user_buffer_vtbl); + return (struct pb_user_buffer *)buf; } static void -client_buffer_reference(struct pipe_buffer *buf) -{ - /* No-op */ -} - - -static void -client_buffer_release(struct pipe_buffer *buf) +pb_user_buffer_destroy(struct pb_buffer *buf) { assert(buf); free(buf); @@ -72,23 +66,23 @@ client_buffer_release(struct pipe_buffer *buf) static void * -client_buffer_map(struct pipe_buffer *buf, +pb_user_buffer_map(struct pb_buffer *buf, unsigned flags) { - return client_buffer(buf)->data; + return pb_user_buffer(buf)->data; } static void -client_buffer_unmap(struct pipe_buffer *buf) +pb_user_buffer_unmap(struct pb_buffer *buf) { /* No-op */ } static void -client_buffer_get_base_buffer(struct pipe_buffer *buf, - struct pipe_buffer **base_buf, +pb_user_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, unsigned *offset) { *base_buf = buf; @@ -96,27 +90,25 @@ client_buffer_get_base_buffer(struct pipe_buffer *buf, } -const struct pipe_buffer_vtbl -client_buffer_vtbl = { - client_buffer_reference, - client_buffer_release, - client_buffer_map, - client_buffer_unmap, - client_buffer_get_base_buffer +const struct pb_vtbl +pb_user_buffer_vtbl = { + pb_user_buffer_destroy, + pb_user_buffer_map, + pb_user_buffer_unmap, + pb_user_buffer_get_base_buffer }; -struct pipe_buffer * -client_buffer_create(void *data) +struct pb_buffer * +pb_user_buffer_create(void *data, unsigned bytes) { - struct client_buffer *buf; - - buf = (struct client_buffer *)malloc(sizeof(struct client_buffer)); + struct pb_user_buffer *buf = CALLOC_STRUCT(pb_user_buffer); + if(!buf) return NULL; - buf->base.vtbl = &client_buffer_vtbl; - + buf->base.vtbl = &pb_user_buffer_vtbl; + buf->base.base.size = bytes; buf->data = data; return &buf->base; diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c index dfb80b1dcf..625120d714 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -72,9 +72,9 @@ struct fenced_buffer_list */ struct fenced_buffer { - struct pipe_buffer base; + struct pb_buffer base; - struct pipe_buffer *buffer; + struct pb_buffer *buffer; unsigned refcount; struct pipe_fence_handle *fence; @@ -84,8 +84,8 @@ struct fenced_buffer }; -static inline struct fenced_buffer * -fenced_buffer(struct pipe_buffer *buf) +static INLINE struct fenced_buffer * +fenced_buffer(struct pb_buffer *buf) { assert(buf); assert(buf->vtbl == &fenced_buffer_vtbl); @@ -93,12 +93,6 @@ fenced_buffer(struct pipe_buffer *buf) } -static void -_fenced_buffer_destroy(struct fenced_buffer *fenced_buf) -{ - buffer_release(fenced_buf->buffer); - free(fenced_buf); -} static void @@ -143,97 +137,81 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, LIST_DEL(list); fenced_list->numDelayed--; - - _fenced_buffer_destroy(fenced_buf); - } -} - -static void -fenced_buffer_reference(struct pipe_buffer *buf) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - struct fenced_buffer_list *fenced_list = fenced_buf->list; - - _glthread_LOCK_MUTEX(fenced_list->mutex); - fenced_buf->refcount++; - _glthread_UNLOCK_MUTEX(fenced_list->mutex); + /* Do the delayed destroy: + */ + pb_destroy(fenced_buf->buffer); + free(fenced_buf); + } } static void -fenced_buffer_release(struct pipe_buffer *buf) +fenced_buffer_destroy(struct pb_buffer *buf) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); struct fenced_buffer_list *fenced_list = fenced_buf->list; - _glthread_LOCK_MUTEX(fenced_list->mutex); - - fenced_buf->refcount--; - if(!fenced_buf->refcount) { - if (fenced_buf->fence) { - LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); - fenced_list->numDelayed++; - } - else { - _fenced_buffer_destroy(fenced_buf); - } - - if ((fenced_list->numDelayed % fenced_list->checkDelayed) == 0) - _fenced_buffer_list_check_free(fenced_list, 0); + if (fenced_buf->fence) { + LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); + fenced_list->numDelayed++; + } + else { + pb_destroy(fenced_buf->buffer); + free(fenced_buf); } - _glthread_UNLOCK_MUTEX(fenced_list->mutex); + if ((fenced_list->numDelayed % fenced_list->checkDelayed) == 0) + _fenced_buffer_list_check_free(fenced_list, 0); } static void * -fenced_buffer_map(struct pipe_buffer *buf, +fenced_buffer_map(struct pb_buffer *buf, unsigned flags) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); - return buffer_map(fenced_buf->buffer, flags); + return pb_map(fenced_buf->buffer, flags); } static void -fenced_buffer_unmap(struct pipe_buffer *buf) +fenced_buffer_unmap(struct pb_buffer *buf) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); - buffer_unmap(fenced_buf->buffer); + pb_unmap(fenced_buf->buffer); } static void -fenced_buffer_get_base_buffer(struct pipe_buffer *buf, - struct pipe_buffer **base_buf, +fenced_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, unsigned *offset) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); - buffer_get_base_buffer(fenced_buf->buffer, base_buf, offset); + pb_get_base_buffer(fenced_buf->buffer, base_buf, offset); } -const struct pipe_buffer_vtbl +const struct pb_vtbl fenced_buffer_vtbl = { - fenced_buffer_reference, - fenced_buffer_release, + fenced_buffer_destroy, fenced_buffer_map, fenced_buffer_unmap, fenced_buffer_get_base_buffer }; -struct pipe_buffer * +struct pb_buffer * fenced_buffer_create(struct fenced_buffer_list *fenced_list, - struct pipe_buffer *buffer) + struct pb_buffer *buffer) { struct fenced_buffer *buf; if(!buffer) return NULL; - buf = (struct fenced_buffer *)calloc(1, sizeof(struct fenced_buffer)); + buf = CALLOC_STRUCT(fenced_buffer); if(!buf) return NULL; @@ -247,20 +225,16 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list, void -buffer_fence(struct pipe_buffer *buf, +buffer_fence(struct pb_buffer *buf, struct pipe_fence_handle *fence) { - if(buf->vtbl == &fenced_buffer_vtbl) { - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - struct fenced_buffer_list *fenced_list = fenced_buf->list; - struct pipe_winsys *winsys = fenced_list->winsys; + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + struct pipe_winsys *winsys = fenced_list->winsys; - _glthread_LOCK_MUTEX(fenced_list->mutex); - winsys->fence_reference(winsys, &fenced_buf->fence, fence); - _glthread_UNLOCK_MUTEX(fenced_list->mutex); - } - else - assert(0); + _glthread_LOCK_MUTEX(fenced_list->mutex); + winsys->fence_reference(winsys, &fenced_buf->fence, fence); + _glthread_UNLOCK_MUTEX(fenced_list->mutex); } diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h index 07e42a67f8..09082a5390 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h @@ -70,7 +70,7 @@ struct fenced_buffer_list; * * NOTE: Made public for debugging purposes. */ -extern const struct pipe_buffer_vtbl fenced_buffer_vtbl; +extern const struct pb_vtbl fenced_buffer_vtbl; /** @@ -98,19 +98,19 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list); * * NOTE: this will not increase the buffer reference count. */ -struct pipe_buffer * +struct pb_buffer * fenced_buffer_create(struct fenced_buffer_list *fenced, - struct pipe_buffer *buffer); + struct pb_buffer *buffer); /** * Set a buffer's fence. * - * NOTE: Although it takes a generic pipe buffer argument, it will fail + * NOTE: Although it takes a generic pb_buffer argument, it will fail * on everything but buffers returned by fenced_buffer_create. */ void -buffer_fence(struct pipe_buffer *buf, +buffer_fence(struct pb_buffer *buf, struct pipe_fence_handle *fence); diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_handle.c b/src/mesa/pipe/pipebuffer/pb_buffer_handle.c deleted file mode 100644 index 9aeff976f3..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_buffer_handle.c +++ /dev/null @@ -1,140 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Drop-in implementation of the winsys driver functions for buffer handles. - * - * \author José Fonseca - */ - - -#include -#include - -#include "pipe/p_winsys.h" -#include "pipe/p_defines.h" - -#include "pb_buffer.h" -#include "pb_buffer_handle.h" - - -static struct pipe_buffer_handle * -buffer_handle_create(struct pipe_winsys *winsys, - unsigned alignment, - unsigned usage, - unsigned size) -{ - struct pipe_buffer_handle *handle; - - handle = (struct pipe_buffer_handle *)malloc(sizeof(struct pipe_buffer_handle)); - if(!handle) - return NULL; - - handle->refcount = 1; - handle->alignment = alignment; - handle->usage = usage; - handle->size = size; - - handle->buf = &null_buffer; - - return handle; -} - - -static struct pipe_buffer_handle * -buffer_handle_create_user(struct pipe_winsys *winsys, - void *data, unsigned size) -{ - struct pipe_buffer_handle *handle; - struct pipe_buffer *buf; - - handle = buffer_handle_create(winsys, 1, 0, size); - if(!handle) - return NULL; - - buf = client_buffer_create(data); - if(!buf) { - free(handle); - return NULL; - } - - buffer_handle_data(handle, buf); - - return handle; -} - - -static void * -buffer_handle_map(struct pipe_winsys *winsys, - struct pipe_buffer_handle *handle, - unsigned flags) -{ - return buffer_map(handle->buf, flags); -} - - -static void -buffer_handle_unmap(struct pipe_winsys *winsys, - struct pipe_buffer_handle *handle) -{ - buffer_unmap(handle->buf); -} - - -static void -buffer_handle_reference(struct pipe_winsys *winsys, - struct pipe_buffer_handle **dst, - struct pipe_buffer_handle *src) -{ - /* XXX: should this be thread safe? */ - - if (src) { - src->refcount++; - } - - if (*dst) { - (*dst)->refcount--; - if ((*dst)->refcount == 0) { - buffer_release((*dst)->buf); - free(*dst); - } - } - - *dst = src; -} - - -void -buffer_handle_init_winsys(struct pipe_winsys *winsys) -{ - winsys->buffer_create = buffer_handle_create; - winsys->user_buffer_create = buffer_handle_create_user; - winsys->buffer_map = buffer_handle_map; - winsys->buffer_unmap = buffer_handle_unmap; - winsys->buffer_reference = buffer_handle_reference; -} diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_handle.h b/src/mesa/pipe/pipebuffer/pb_buffer_handle.h deleted file mode 100644 index 34133c9fba..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_buffer_handle.h +++ /dev/null @@ -1,120 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Buffer handle interface. - * - * \author José Fonseca - */ - -#ifndef PB_BUFFER_HANDLE_H_ -#define PB_BUFFER_HANDLE_H_ - - -#include - -#include "pb_buffer.h" - - -/** - * Buffer handle. - * - * The buffer handle and the buffer data storage are separate entities. This - * is modelled after ARB_vertex_buffer_object, which is the interface that - * Gallium requires. See p_winsys.h for more information. - */ -struct pipe_buffer_handle -{ - /** Reference count */ - unsigned refcount; - - /** Allocation characteristics */ - unsigned alignment; - unsigned usage; - unsigned size; - - /** - * The actual buffer. - * - * It should never be NULL. Use null_buffer instead. - */ - struct pipe_buffer *buf; -}; - - -/** - * Set buffer storage. - * - * NOTE: this will not increase the buffer reference count. - */ -static inline void -buffer_handle_data(struct pipe_buffer_handle *handle, - struct pipe_buffer *buf) -{ - assert(handle); - assert(handle->buf); - buffer_release(handle->buf); - assert(buf); - handle->buf = buf; -} - - -static inline void -buffer_handle_clear(struct pipe_buffer_handle *handle) -{ - buffer_handle_data(handle, &null_buffer); -} - - -static inline int -buffer_handle_has_data(struct pipe_buffer_handle *handle) { - assert(handle); - assert(handle->buf); - return handle->buf != &null_buffer; -} - - -/** - * Fill in the pipe_winsys' buffer-related callbacks. - * - * Specifically, the fullfilled functions are: - * - buffer_create - * - user_buffer_create - * - buffer_map - * - buffer_unmap - * - buffer_reference - * - buffer_subdata - * - buffer_get_subdata - * - * NOTE: buffer_data is left untouched. - */ -void -buffer_handle_init_winsys(struct pipe_winsys *winsys); - - -#endif /*PB_BUFFER_HANDLE_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c index 65ad51e1e7..e02eb1eebd 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c @@ -42,15 +42,15 @@ struct malloc_buffer { - struct pipe_buffer base; + struct pb_buffer base; void *data; }; -extern const struct pipe_buffer_vtbl malloc_buffer_vtbl; +extern const struct pb_vtbl malloc_buffer_vtbl; -static inline struct malloc_buffer * -malloc_buffer(struct pipe_buffer *buf) +static INLINE struct malloc_buffer * +malloc_buffer(struct pb_buffer *buf) { assert(buf); assert(buf->vtbl == &malloc_buffer_vtbl); @@ -59,14 +59,7 @@ malloc_buffer(struct pipe_buffer *buf) static void -malloc_buffer_reference(struct pipe_buffer *buf) -{ - /* no-op */ -} - - -static void -malloc_buffer_release(struct pipe_buffer *buf) +malloc_buffer_destroy(struct pb_buffer *buf) { free(malloc_buffer(buf)->data); free(buf); @@ -74,7 +67,7 @@ malloc_buffer_release(struct pipe_buffer *buf) static void * -malloc_buffer_map(struct pipe_buffer *buf, +malloc_buffer_map(struct pb_buffer *buf, unsigned flags) { return malloc_buffer(buf)->data; @@ -82,15 +75,15 @@ malloc_buffer_map(struct pipe_buffer *buf, static void -malloc_buffer_unmap(struct pipe_buffer *buf) +malloc_buffer_unmap(struct pb_buffer *buf) { /* No-op */ } static void -malloc_buffer_get_base_buffer(struct pipe_buffer *buf, - struct pipe_buffer **base_buf, +malloc_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, unsigned *offset) { *base_buf = buf; @@ -98,18 +91,19 @@ malloc_buffer_get_base_buffer(struct pipe_buffer *buf, } -const struct pipe_buffer_vtbl +const struct pb_vtbl malloc_buffer_vtbl = { - malloc_buffer_reference, - malloc_buffer_release, + malloc_buffer_destroy, malloc_buffer_map, malloc_buffer_unmap, malloc_buffer_get_base_buffer }; -struct pipe_buffer * -malloc_buffer_create(unsigned size) +struct pb_buffer * +pb_malloc_buffer_create( unsigned alignment, + unsigned usage, + unsigned size ) { struct malloc_buffer *buf; @@ -121,7 +115,10 @@ malloc_buffer_create(unsigned size) return NULL; buf->base.vtbl = &malloc_buffer_vtbl; - + buf->base.base.alignment = alignment; + buf->base.base.usage = usage; + buf->base.base.size = size; + buf->data = malloc(size); if(!buf->data) { free(buf); diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_null.c b/src/mesa/pipe/pipebuffer/pb_buffer_null.c deleted file mode 100644 index a356c6b2d5..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_buffer_null.c +++ /dev/null @@ -1,98 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Null buffer implementation. - * - * We have a special null buffer object so that we can safely call buffer - * operations without having to check whether the buffer pointer is null or not. - * - * \author José Fonseca - */ - - -#include "pipe/p_winsys.h" -#include "pipe/p_defines.h" - -#include "pb_buffer.h" - - -static void -null_buffer_reference(struct pipe_buffer *buf) -{ - /* No-op */ -} - - -static void -null_buffer_release(struct pipe_buffer *buf) -{ - /* No-op */ -} - - -static void * -null_buffer_map(struct pipe_buffer *buf, - unsigned flags) -{ - assert(0); - return NULL; -} - - -static void -null_buffer_unmap(struct pipe_buffer *buf) -{ - assert(0); -} - - -static void -null_buffer_get_base_buffer(struct pipe_buffer *buf, - struct pipe_buffer **base_buf, - unsigned *offset) -{ - *base_buf = buf; - *offset = 0; -} - - -const struct pipe_buffer_vtbl -pipe_buffer_vtbl = { - null_buffer_reference, - null_buffer_release, - null_buffer_map, - null_buffer_unmap, - null_buffer_get_base_buffer -}; - - -struct pipe_buffer -null_buffer = { - &pipe_buffer_vtbl -}; diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr.h b/src/mesa/pipe/pipebuffer/pb_bufmgr.h index 12e36323a8..13d4ea7545 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr.h +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr.h @@ -60,15 +60,15 @@ struct pipe_winsys; /** * Abstract base class for all buffer managers. */ -struct buffer_manager +struct pb_manager { /* XXX: we will likely need more allocation flags */ - struct pipe_buffer * - (*create_buffer)( struct buffer_manager *mgr, + struct pb_buffer * + (*create_buffer)( struct pb_manager *mgr, size_t size ); void - (*destroy)( struct buffer_manager *mgr ); + (*destroy)( struct pb_manager *mgr ); }; @@ -80,8 +80,8 @@ struct buffer_manager * * It is meant to manage the allocation of batch buffer pools. */ -struct buffer_manager * -pool_bufmgr_create(struct buffer_manager *provider, +struct pb_manager * +pool_bufmgr_create(struct pb_manager *provider, size_t n, size_t size); @@ -92,8 +92,8 @@ pool_bufmgr_create(struct buffer_manager *provider, * with the size of the heap, and then using the old mm memory manager to manage * that heap. */ -struct buffer_manager * -mm_bufmgr_create(struct buffer_manager *provider, +struct pb_manager * +mm_bufmgr_create(struct pb_manager *provider, size_t size, size_t align2); /** @@ -101,8 +101,8 @@ mm_bufmgr_create(struct buffer_manager *provider, * * Buffer will be release when the manager is destroyed. */ -struct buffer_manager * -mm_bufmgr_create_from_buffer(struct pipe_buffer *buffer, +struct pb_manager * +mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, size_t size, size_t align2); @@ -115,8 +115,8 @@ mm_bufmgr_create_from_buffer(struct pipe_buffer *buffer, * NOTE: the buffer manager that provides the buffers will be destroyed * at the same time. */ -struct buffer_manager * -fenced_bufmgr_create(struct buffer_manager *provider, +struct pb_manager * +fenced_bufmgr_create(struct pb_manager *provider, struct pipe_winsys *winsys); diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c index defd8e4df7..8cdbbed272 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c @@ -42,30 +42,30 @@ #include "pb_bufmgr.h" -struct fenced_buffer_manager +struct fenced_pb_manager { - struct buffer_manager base; + struct pb_manager base; - struct buffer_manager *provider; + struct pb_manager *provider; struct fenced_buffer_list *fenced_list; }; -static inline struct fenced_buffer_manager * -fenced_buffer_manager(struct buffer_manager *mgr) +static inline struct fenced_pb_manager * +fenced_pb_manager(struct pb_manager *mgr) { assert(mgr); - return (struct fenced_buffer_manager *)mgr; + return (struct fenced_pb_manager *)mgr; } -static struct pipe_buffer * -fenced_bufmgr_create_buffer(struct buffer_manager *mgr, size_t size) +static struct pb_buffer * +fenced_bufmgr_create_buffer(struct pb_manager *mgr, size_t size) { - struct fenced_buffer_manager *fenced_mgr = fenced_buffer_manager(mgr); - struct pipe_buffer *buf; - struct pipe_buffer *fenced_buf; + struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); + struct pb_buffer *buf; + struct pb_buffer *fenced_buf; /* check for free buffers before allocating new ones */ fenced_buffer_list_check_free(fenced_mgr->fenced_list, 0); @@ -84,7 +84,7 @@ fenced_bufmgr_create_buffer(struct buffer_manager *mgr, size_t size) fenced_buf = fenced_buffer_create(fenced_mgr->fenced_list, buf); if(!fenced_buf) { - buffer_release(buf); + pb_destroy(buf); } return fenced_buf; @@ -92,9 +92,9 @@ fenced_bufmgr_create_buffer(struct buffer_manager *mgr, size_t size) static void -fenced_bufmgr_destroy(struct buffer_manager *mgr) +fenced_bufmgr_destroy(struct pb_manager *mgr) { - struct fenced_buffer_manager *fenced_mgr = fenced_buffer_manager(mgr); + struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); fenced_buffer_list_destroy(fenced_mgr->fenced_list); @@ -104,13 +104,13 @@ fenced_bufmgr_destroy(struct buffer_manager *mgr) } -struct buffer_manager * -fenced_bufmgr_create(struct buffer_manager *provider, +struct pb_manager * +fenced_bufmgr_create(struct pb_manager *provider, struct pipe_winsys *winsys) { - struct fenced_buffer_manager *fenced_mgr; + struct fenced_pb_manager *fenced_mgr; - fenced_mgr = (struct fenced_buffer_manager *)calloc(1, sizeof(*fenced_mgr)); + fenced_mgr = (struct fenced_pb_manager *)calloc(1, sizeof(*fenced_mgr)); if (!fenced_mgr) return NULL; diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index d174f24e32..f456b571d9 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -53,9 +53,9 @@ #define SUPER(__derived) (&(__derived)->base) -struct mm_buffer_manager +struct mm_pb_manager { - struct buffer_manager base; + struct pb_manager base; _glthread_Mutex mutex; @@ -64,31 +64,31 @@ struct mm_buffer_manager size_t align2; - struct pipe_buffer *buffer; + struct pb_buffer *buffer; void *map; }; -static inline struct mm_buffer_manager * -mm_buffer_manager(struct buffer_manager *mgr) +static inline struct mm_pb_manager * +mm_pb_manager(struct pb_manager *mgr) { assert(mgr); - return (struct mm_buffer_manager *)mgr; + return (struct mm_pb_manager *)mgr; } struct mm_buffer { - struct pipe_buffer base; + struct pb_buffer base; - struct mm_buffer_manager *mgr; + struct mm_pb_manager *mgr; struct mem_block *block; }; static inline struct mm_buffer * -mm_buffer(struct pipe_buffer *buf) +mm_buffer(struct pb_buffer *buf) { assert(buf); return (struct mm_buffer *)buf; @@ -96,17 +96,10 @@ mm_buffer(struct pipe_buffer *buf) static void -mm_buffer_reference(struct pipe_buffer *buf) -{ - /* No-op */ -} - - -static void -mm_buffer_release(struct pipe_buffer *buf) +mm_buffer_destroy(struct pb_buffer *buf) { struct mm_buffer *mm_buf = mm_buffer(buf); - struct mm_buffer_manager *mm = mm_buf->mgr; + struct mm_pb_manager *mm = mm_buf->mgr; _glthread_LOCK_MUTEX(mm->mutex); mmFreeMem(mm_buf->block); @@ -116,50 +109,49 @@ mm_buffer_release(struct pipe_buffer *buf) static void * -mm_buffer_map(struct pipe_buffer *buf, +mm_buffer_map(struct pb_buffer *buf, unsigned flags) { struct mm_buffer *mm_buf = mm_buffer(buf); - struct mm_buffer_manager *mm = mm_buf->mgr; + struct mm_pb_manager *mm = mm_buf->mgr; return (unsigned char *) mm->map + mm_buf->block->ofs; } static void -mm_buffer_unmap(struct pipe_buffer *buf) +mm_buffer_unmap(struct pb_buffer *buf) { /* No-op */ } static void -mm_buffer_get_base_buffer(struct pipe_buffer *buf, - struct pipe_buffer **base_buf, +mm_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, unsigned *offset) { struct mm_buffer *mm_buf = mm_buffer(buf); - struct mm_buffer_manager *mm = mm_buf->mgr; - buffer_get_base_buffer(mm->buffer, base_buf, offset); + struct mm_pb_manager *mm = mm_buf->mgr; + pb_get_base_buffer(mm->buffer, base_buf, offset); *offset += mm_buf->block->ofs; } -static const struct pipe_buffer_vtbl +static const struct pb_vtbl mm_buffer_vtbl = { - mm_buffer_reference, - mm_buffer_release, + mm_buffer_destroy, mm_buffer_map, mm_buffer_unmap, mm_buffer_get_base_buffer }; -static struct pipe_buffer * -mm_bufmgr_create_buffer(struct buffer_manager *mgr, +static struct pb_buffer * +mm_bufmgr_create_buffer(struct pb_manager *mgr, size_t size) { - struct mm_buffer_manager *mm = mm_buffer_manager(mgr); + struct mm_pb_manager *mm = mm_pb_manager(mgr); struct mm_buffer *mm_buf; _glthread_LOCK_MUTEX(mm->mutex); @@ -200,16 +192,16 @@ mm_bufmgr_create_buffer(struct buffer_manager *mgr, static void -mm_bufmgr_destroy(struct buffer_manager *mgr) +mm_bufmgr_destroy(struct pb_manager *mgr) { - struct mm_buffer_manager *mm = mm_buffer_manager(mgr); + struct mm_pb_manager *mm = mm_pb_manager(mgr); _glthread_LOCK_MUTEX(mm->mutex); mmDestroy(mm->heap); - buffer_unmap(mm->buffer); - buffer_release(mm->buffer); + pb_unmap(mm->buffer); + pb_destroy(mm->buffer); _glthread_UNLOCK_MUTEX(mm->mutex); @@ -217,16 +209,16 @@ mm_bufmgr_destroy(struct buffer_manager *mgr) } -struct buffer_manager * -mm_bufmgr_create_from_buffer(struct pipe_buffer *buffer, +struct pb_manager * +mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, size_t size, size_t align2) { - struct mm_buffer_manager *mm; + struct mm_pb_manager *mm; if(!buffer) return NULL; - mm = (struct mm_buffer_manager *)calloc(1, sizeof(*mm)); + mm = (struct mm_pb_manager *)calloc(1, sizeof(*mm)); if (!mm) return NULL; @@ -240,9 +232,9 @@ mm_bufmgr_create_from_buffer(struct pipe_buffer *buffer, mm->buffer = buffer; - mm->map = buffer_map(mm->buffer, - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE); + mm->map = pb_map(mm->buffer, + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); if(!mm->map) goto failure; @@ -256,19 +248,19 @@ failure: if(mm->heap) mmDestroy(mm->heap); if(mm->map) - buffer_unmap(mm->buffer); + pb_unmap(mm->buffer); if(mm) free(mm); return NULL; } -struct buffer_manager * -mm_bufmgr_create(struct buffer_manager *provider, +struct pb_manager * +mm_bufmgr_create(struct pb_manager *provider, size_t size, size_t align2) { - struct pipe_buffer *buffer; - struct buffer_manager *mgr; + struct pb_buffer *buffer; + struct pb_manager *mgr; assert(provider); assert(provider->create_buffer); @@ -278,7 +270,7 @@ mm_bufmgr_create(struct buffer_manager *provider, mgr = mm_bufmgr_create_from_buffer(buffer, size, align2); if (!mgr) { - buffer_release(buffer); + pb_destroy(buffer); return NULL; } diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index e6a8c78668..7e8494f615 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -55,9 +55,9 @@ #define SUPER(__derived) (&(__derived)->base) -struct pool_buffer_manager +struct pool_pb_manager { - struct buffer_manager base; + struct pb_manager base; _glthread_Mutex mutex; @@ -68,26 +68,26 @@ struct pool_buffer_manager struct list_head free; - struct pipe_buffer *buffer; + struct pb_buffer *buffer; void *map; struct pool_buffer *bufs; }; -static inline struct pool_buffer_manager * -pool_buffer_manager(struct buffer_manager *mgr) +static inline struct pool_pb_manager * +pool_pb_manager(struct pb_manager *mgr) { assert(mgr); - return (struct pool_buffer_manager *)mgr; + return (struct pool_pb_manager *)mgr; } struct pool_buffer { - struct pipe_buffer base; + struct pb_buffer base; - struct pool_buffer_manager *mgr; + struct pool_pb_manager *mgr; struct list_head head; @@ -96,25 +96,19 @@ struct pool_buffer static inline struct pool_buffer * -pool_buffer(struct pipe_buffer *buf) +pool_buffer(struct pb_buffer *buf) { assert(buf); return (struct pool_buffer *)buf; } -static void -pool_buffer_reference(struct pipe_buffer *buf) -{ - /* No-op */ -} - static void -pool_buffer_release(struct pipe_buffer *buf) +pool_buffer_destroy(struct pb_buffer *buf) { struct pool_buffer *pool_buf = pool_buffer(buf); - struct pool_buffer_manager *pool = pool_buf->mgr; + struct pool_pb_manager *pool = pool_buf->mgr; _glthread_LOCK_MUTEX(pool->mutex); LIST_ADD(&pool_buf->head, &pool->free); @@ -124,10 +118,10 @@ pool_buffer_release(struct pipe_buffer *buf) static void * -pool_buffer_map(struct pipe_buffer *buf, unsigned flags) +pool_buffer_map(struct pb_buffer *buf, unsigned flags) { struct pool_buffer *pool_buf = pool_buffer(buf); - struct pool_buffer_manager *pool = pool_buf->mgr; + struct pool_pb_manager *pool = pool_buf->mgr; void *map; _glthread_LOCK_MUTEX(pool->mutex); @@ -138,38 +132,37 @@ pool_buffer_map(struct pipe_buffer *buf, unsigned flags) static void -pool_buffer_unmap(struct pipe_buffer *buf) +pool_buffer_unmap(struct pb_buffer *buf) { /* No-op */ } static void -pool_buffer_get_base_buffer(struct pipe_buffer *buf, - struct pipe_buffer **base_buf, +pool_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, unsigned *offset) { struct pool_buffer *pool_buf = pool_buffer(buf); - struct pool_buffer_manager *pool = pool_buf->mgr; - buffer_get_base_buffer(pool->buffer, base_buf, offset); + struct pool_pb_manager *pool = pool_buf->mgr; + pb_get_base_buffer(pool->buffer, base_buf, offset); *offset += pool_buf->start; } -static const struct pipe_buffer_vtbl +static const struct pb_vtbl pool_buffer_vtbl = { - pool_buffer_reference, - pool_buffer_release, + pool_buffer_destroy, pool_buffer_map, pool_buffer_unmap, pool_buffer_get_base_buffer }; -static struct pipe_buffer * -pool_bufmgr_create_buffer(struct buffer_manager *mgr, size_t size) +static struct pb_buffer * +pool_bufmgr_create_buffer(struct pb_manager *mgr, size_t size) { - struct pool_buffer_manager *pool = pool_buffer_manager(mgr); + struct pool_pb_manager *pool = pool_pb_manager(mgr); struct pool_buffer *pool_buf; struct list_head *item; @@ -201,15 +194,15 @@ pool_bufmgr_create_buffer(struct buffer_manager *mgr, size_t size) static void -pool_bufmgr_destroy(struct buffer_manager *mgr) +pool_bufmgr_destroy(struct pb_manager *mgr) { - struct pool_buffer_manager *pool = pool_buffer_manager(mgr); + struct pool_pb_manager *pool = pool_pb_manager(mgr); _glthread_LOCK_MUTEX(pool->mutex); free(pool->bufs); - buffer_unmap(pool->buffer); - buffer_release(pool->buffer); + pb_unmap(pool->buffer); + pb_destroy(pool->buffer); _glthread_UNLOCK_MUTEX(pool->mutex); @@ -217,16 +210,16 @@ pool_bufmgr_destroy(struct buffer_manager *mgr) } -struct buffer_manager * -pool_bufmgr_create(struct buffer_manager *provider, +struct pb_manager * +pool_bufmgr_create(struct pb_manager *provider, size_t numBufs, size_t bufSize) { - struct pool_buffer_manager *pool; + struct pool_pb_manager *pool; struct pool_buffer *pool_buf; int i; - pool = (struct pool_buffer_manager *)calloc(1, sizeof(*pool)); + pool = (struct pool_pb_manager *)calloc(1, sizeof(*pool)); if (!pool) return NULL; @@ -245,7 +238,7 @@ pool_bufmgr_create(struct buffer_manager *provider, if (!pool->buffer) goto failure; - pool->map = buffer_map(pool->buffer, + pool->map = pb_map(pool->buffer, PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE); if(!pool->map) @@ -270,9 +263,9 @@ failure: if(pool->bufs) free(pool->bufs); if(pool->map) - buffer_unmap(pool->buffer); + pb_unmap(pool->buffer); if(pool->buffer) - buffer_release(pool->buffer); + pb_destroy(pool->buffer); if(pool) free(pool); return NULL; diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c index 8dadd9aa74..cea6b90104 100644 --- a/src/mesa/pipe/softpipe/sp_context.c +++ b/src/mesa/pipe/softpipe/sp_context.c @@ -135,7 +135,7 @@ static void softpipe_destroy( struct pipe_context *pipe ) for (i = 0; i < Elements(softpipe->constants); i++) { if (softpipe->constants[i].buffer) { - ws->buffer_reference(ws, &softpipe->constants[i].buffer, NULL); + pipe_buffer_reference(ws, &softpipe->constants[i].buffer, NULL); } } diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 8ee9177e5b..423c91d4b8 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -88,7 +88,7 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, */ boolean softpipe_draw_elements(struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, + struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count) { diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index bac7b0876f..af955c1e17 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -157,7 +157,7 @@ boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count); boolean softpipe_draw_elements(struct pipe_context *pipe, - struct pipe_buffer_handle *indexBuffer, + struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count); diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 945c93411f..1430be7869 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -30,6 +30,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "pipe/draw/draw_context.h" #include "pipe/p_shader_tokens.h" @@ -165,9 +166,9 @@ softpipe_set_constant_buffer(struct pipe_context *pipe, assert(index == 0); /* note: reference counting */ - ws->buffer_reference(ws, - &softpipe->constants[shader].buffer, - buf->buffer); + pipe_buffer_reference(ws, + &softpipe->constants[shader].buffer, + buf->buffer); softpipe->constants[shader].size = buf->size; softpipe->dirty |= SP_NEW_CONSTANTS; diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index 23e9c1b376..172234843d 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -121,7 +121,7 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); */ - pipe->winsys->buffer_reference(pipe->winsys, &spt->buffer, NULL); + pipe_buffer_reference(pipe->winsys, &spt->buffer, NULL); FREE(spt); } @@ -144,7 +144,7 @@ softpipe_get_tex_surface(struct pipe_context *pipe, if (ps) { assert(ps->refcount); assert(ps->winsys); - pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); + pipe_buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; diff --git a/src/mesa/pipe/softpipe/sp_texture.h b/src/mesa/pipe/softpipe/sp_texture.h index 0494bf365b..c6cf370351 100644 --- a/src/mesa/pipe/softpipe/sp_texture.h +++ b/src/mesa/pipe/softpipe/sp_texture.h @@ -41,7 +41,7 @@ struct softpipe_texture /* The data is held here: */ - struct pipe_buffer_handle *buffer; + struct pipe_buffer *buffer; unsigned long buffer_size; }; diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c index cb043ef394..c3cd22eea3 100644 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ b/src/mesa/pipe/xlib/xm_winsys.c @@ -40,6 +40,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/softpipe/sp_winsys.h" #ifdef GALLIUM_CELL @@ -57,9 +58,8 @@ */ struct xm_buffer { + struct pipe_buffer base; boolean userBuffer; /** Is this a user-space buffer? */ - int refcount; - unsigned size; void *data; void *mapped; }; @@ -106,63 +106,44 @@ xmesa_softpipe_winsys(struct softpipe_winsys *spws) * buffer pointer... */ static INLINE struct xm_buffer * -xm_bo( struct pipe_buffer_handle *bo ) +xm_buffer( struct pipe_buffer *buf ) { - return (struct xm_buffer *) bo; + return (struct xm_buffer *)buf; } -static INLINE struct pipe_buffer_handle * -pipe_bo( struct xm_buffer *bo ) -{ - return (struct pipe_buffer_handle *) bo; -} /* Most callbacks map direcly onto dri_bufmgr operations: */ static void * -xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer_handle *buf, +xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buf, unsigned flags) { - struct xm_buffer *xm_buf = xm_bo(buf); + struct xm_buffer *xm_buf = xm_buffer(buf); xm_buf->mapped = xm_buf->data; return xm_buf->mapped; } static void -xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer_handle *buf) +xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) { - struct xm_buffer *xm_buf = xm_bo(buf); + struct xm_buffer *xm_buf = xm_buffer(buf); xm_buf->mapped = NULL; } static void -xm_buffer_reference(struct pipe_winsys *pws, - struct pipe_buffer_handle **ptr, - struct pipe_buffer_handle *buf) +xm_buffer_destroy(struct pipe_winsys *pws, + struct pipe_buffer *buf) { - if (*ptr) { - struct xm_buffer *oldBuf = xm_bo(*ptr); - oldBuf->refcount--; - assert(oldBuf->refcount >= 0); - if (oldBuf->refcount == 0) { - if (oldBuf->data) { - if (!oldBuf->userBuffer) - align_free(oldBuf->data); - oldBuf->data = NULL; - } - free(oldBuf); - } - *ptr = NULL; - } + struct xm_buffer *oldBuf = xm_buffer(buf); - assert(!(*ptr)); - - if (buf) { - struct xm_buffer *newBuf = xm_bo(buf); - newBuf->refcount++; - *ptr = buf; + if (oldBuf->data) { + if (!oldBuf->userBuffer) + align_free(oldBuf->data); + oldBuf->data = NULL; } + + free(oldBuf); } @@ -174,7 +155,7 @@ static void xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) { XImage *ximage = b->tempImage; - struct xm_buffer *xm_buf = xm_bo(surf->buffer); + struct xm_buffer *xm_buf = xm_buffer(surf->buffer); const uint tilesPerRow = (surf->width + TILE_SIZE - 1) / TILE_SIZE; uint x, y; @@ -214,7 +195,7 @@ void xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) { XImage *ximage = b->tempImage; - struct xm_buffer *xm_buf = xm_bo(surf->buffer); + struct xm_buffer *xm_buf = xm_buffer(surf->buffer); const struct xmesa_surface *xm_surf = xmesa_surface((struct pipe_surface *) surf); @@ -272,35 +253,38 @@ xm_get_name(struct pipe_winsys *pws) } -static struct pipe_buffer_handle * +static struct pipe_buffer * xm_buffer_create(struct pipe_winsys *pws, unsigned alignment, unsigned usage, unsigned size) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); - buffer->refcount = 1; + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; /* align to 16-byte multiple for Cell */ buffer->data = align_malloc(size, max(alignment, 16)); - buffer->size = size; - return pipe_bo(buffer); + return &buffer->base; } /** * Create buffer which wraps user-space data. */ -static struct pipe_buffer_handle * +static struct pipe_buffer * xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); + buffer->base.refcount = 1; + buffer->base.size = bytes; buffer->userBuffer = TRUE; - buffer->refcount = 1; buffer->data = ptr; - buffer->size = bytes; - return pipe_bo(buffer); + + return &buffer->base; } @@ -376,7 +360,7 @@ xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) surf->refcount--; if (surf->refcount == 0) { if (surf->buffer) - winsys->buffer_reference(winsys, &surf->buffer, NULL); + pipe_buffer_reference(winsys, &surf->buffer, NULL); free(surf); } *s = NULL; @@ -407,7 +391,7 @@ xmesa_get_pipe_winsys_aub(void) ws->user_buffer_create = xm_user_buffer_create; ws->buffer_map = xm_buffer_map; ws->buffer_unmap = xm_buffer_unmap; - ws->buffer_reference = xm_buffer_reference; + ws->buffer_destroy = xm_buffer_destroy; ws->surface_alloc = xm_surface_alloc; ws->surface_alloc_storage = xm_surface_alloc_storage; diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c index 28dd07bc6e..bf41570257 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ b/src/mesa/pipe/xlib/xm_winsys_aub.c @@ -38,6 +38,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/i965simple/brw_winsys.h" #include "brw_aub.h" #include "xm_winsys_aub.h" @@ -79,22 +80,22 @@ aub_pipe_winsys( struct pipe_winsys *winsys ) static INLINE struct aub_buffer * -aub_bo( struct pipe_buffer_handle *bo ) +aub_bo( struct pipe_buffer *bo ) { return (struct aub_buffer *)bo; } -static INLINE struct pipe_buffer_handle * +static INLINE struct pipe_buffer * pipe_bo( struct aub_buffer *bo ) { - return (struct pipe_buffer_handle *)bo; + return (struct pipe_buffer *)bo; } static void *aub_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned flags ) { struct aub_buffer *sbo = aub_bo(buf); @@ -109,7 +110,7 @@ static void *aub_buffer_map(struct pipe_winsys *winsys, } static void aub_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf) + struct pipe_buffer *buf) { struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); struct aub_buffer *sbo = aub_bo(buf); @@ -132,26 +133,15 @@ static void aub_buffer_unmap(struct pipe_winsys *winsys, static void -aub_buffer_reference(struct pipe_winsys *winsys, - struct pipe_buffer_handle **ptr, - struct pipe_buffer_handle *buf) +aub_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) { - if (*ptr) { - assert(aub_bo(*ptr)->refcount != 0); - if (--(aub_bo(*ptr)->refcount) == 0) - free(*ptr); - *ptr = NULL; - } - - if (buf) { - aub_bo(buf)->refcount++; - *ptr = buf; - } + free(buf); } void xmesa_buffer_subdata_aub(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned long offset, unsigned long size, const void *data, @@ -206,7 +196,7 @@ void xmesa_display_aub( /* struct pipe_winsys *winsys, */ /* Pipe has no concept of pools. We choose the tex/region pool * for all buffers. */ -static struct pipe_buffer_handle * +static struct pipe_buffer * aub_buffer_create(struct pipe_winsys *winsys, unsigned alignment, unsigned usage, @@ -231,7 +221,7 @@ aub_buffer_create(struct pipe_winsys *winsys, } -static struct pipe_buffer_handle * +static struct pipe_buffer * aub_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) { struct aub_buffer *sbo; @@ -312,7 +302,7 @@ aub_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) surf->refcount--; if (surf->refcount == 0) { if (surf->buffer) - winsys->buffer_reference(winsys, &surf->buffer, NULL); + pipe_buffer_reference(winsys, &surf->buffer, NULL); free(surf); } *s = NULL; @@ -351,7 +341,7 @@ xmesa_create_pipe_winsys_aub( void ) iws->winsys.user_buffer_create = aub_user_buffer_create; iws->winsys.buffer_map = aub_buffer_map; iws->winsys.buffer_unmap = aub_buffer_unmap; - iws->winsys.buffer_reference = aub_buffer_reference; + iws->winsys.buffer_destroy = aub_buffer_destroy; iws->winsys.flush_frontbuffer = aub_flush_frontbuffer; iws->winsys.printf = aub_printf; iws->winsys.get_name = aub_get_name; @@ -439,7 +429,7 @@ static void aub_i965_batch_dword( struct brw_winsys *sws, } static void aub_i965_batch_reloc( struct brw_winsys *sws, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned access_flags, unsigned delta ) { @@ -450,7 +440,7 @@ static void aub_i965_batch_reloc( struct brw_winsys *sws, } static unsigned aub_i965_get_buffer_offset( struct brw_winsys *sws, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned access_flags ) { return aub_bo(buf)->offset; @@ -482,7 +472,7 @@ static void aub_i965_batch_flush( struct brw_winsys *sws, static void aub_i965_buffer_subdata_typed(struct brw_winsys *winsys, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned long offset, unsigned long size, const void *data, diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.h b/src/mesa/pipe/xlib/xm_winsys_aub.h index c0fe449107..7bee199116 100644 --- a/src/mesa/pipe/xlib/xm_winsys_aub.h +++ b/src/mesa/pipe/xlib/xm_winsys_aub.h @@ -30,7 +30,7 @@ struct pipe_context; struct pipe_winsys; -struct pipe_buffer_handle; +struct pipe_buffer; struct pipe_surface; struct pipe_winsys * @@ -47,7 +47,7 @@ xmesa_create_i965simple( struct pipe_winsys *winsys ); void xmesa_buffer_subdata_aub(struct pipe_winsys *winsys, - struct pipe_buffer_handle *buf, + struct pipe_buffer *buf, unsigned long offset, unsigned long size, const void *data, diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index c9d63136b5..21416da2e0 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -37,6 +37,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" #include "st_context.h" #include "st_atom.h" @@ -70,7 +71,7 @@ void st_upload_constants( struct st_context *st, _mesa_load_state_parameters(st->ctx, params); if (cbuf->buffer && cbuf->size != paramBytes) - ws->buffer_reference( ws, &cbuf->buffer, NULL ); + pipe_buffer_reference( ws, &cbuf->buffer, NULL ); if (!cbuf->buffer) { cbuf->buffer = ws->buffer_create(ws, 1, PIPE_BUFFER_USAGE_CONSTANT, diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index 60bd1d5f0d..fa1254ff7c 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -36,6 +36,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" @@ -78,7 +79,7 @@ st_bufferobj_free(GLcontext *ctx, struct gl_buffer_object *obj) struct st_buffer_object *st_obj = st_buffer_object(obj); if (st_obj->buffer) - pipe->winsys->buffer_reference(pipe->winsys, &st_obj->buffer, NULL); + pipe_buffer_reference(pipe->winsys, &st_obj->buffer, NULL); free(st_obj); } @@ -173,7 +174,7 @@ st_bufferobj_data(GLcontext *ctx, buffer_usage = 0; } - pipe->winsys->buffer_reference( pipe->winsys, &st_obj->buffer, NULL ); + pipe_buffer_reference( pipe->winsys, &st_obj->buffer, NULL ); st_obj->buffer = pipe->winsys->buffer_create( pipe->winsys, 32, buffer_usage, size ); diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.h b/src/mesa/state_tracker/st_cb_bufferobjects.h index 15003ae15d..dcbb5a5233 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.h +++ b/src/mesa/state_tracker/st_cb_bufferobjects.h @@ -30,7 +30,7 @@ struct st_context; struct gl_buffer_object; -struct pipe_buffer_handle; +struct pipe_buffer; /** * State_tracker vertex/pixel buffer object, derived from Mesa's @@ -39,7 +39,7 @@ struct pipe_buffer_handle; struct st_buffer_object { struct gl_buffer_object Base; - struct pipe_buffer_handle *buffer; + struct pipe_buffer *buffer; GLsizeiptrARB size; }; diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index f02cb3d133..4341623267 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -100,8 +100,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } if (strb->surface->buffer) - pipe->winsys->buffer_reference(pipe->winsys, &strb->surface->buffer, - NULL); + pipe_buffer_reference(pipe->winsys, &strb->surface->buffer, + NULL); /* Determine surface format here */ if (strb->format != PIPE_FORMAT_NONE) { diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 668ac139f7..9c206c057a 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -52,6 +52,7 @@ #include "st_program.h" #include "pipe/p_context.h" #include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" #include "pipe/draw/draw_context.h" #include "pipe/cso_cache/cso_cache.h" @@ -152,7 +153,7 @@ static void st_destroy_context_priv( struct st_context *st ) for (i = 0; i < Elements(st->state.constants); i++) { if (st->state.constants[i].buffer) { - ws->buffer_reference(ws, &st->state.constants[i].buffer, NULL); + pipe_buffer_reference(ws, &st->state.constants[i].buffer, NULL); } } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 371d4e7966..8ef50ee768 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -45,6 +45,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" +#include "pipe/p_inlines.h" #include "pipe/draw/draw_private.h" #include "pipe/draw/draw_context.h" @@ -237,7 +238,7 @@ st_draw_vbo(GLcontext *ctx, assert(stobj->buffer); vbuffer[attr].buffer = NULL; - winsys->buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); + pipe_buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); vbuffer[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ velement.src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; assert(velement.src_offset <= 2048); /* 11-bit field */ @@ -282,7 +283,7 @@ st_draw_vbo(GLcontext *ctx, if (ib) { /* indexed primitive */ struct gl_buffer_object *bufobj = ib->obj; - struct pipe_buffer_handle *indexBuf = NULL; + struct pipe_buffer *indexBuf = NULL; unsigned indexSize, indexOffset, i; switch (ib->type) { @@ -303,7 +304,7 @@ st_draw_vbo(GLcontext *ctx, if (bufobj && bufobj->Name) { /* elements/indexes are in a real VBO */ struct st_buffer_object *stobj = st_buffer_object(bufobj); - winsys->buffer_reference(winsys, &indexBuf, stobj->buffer); + pipe_buffer_reference(winsys, &indexBuf, stobj->buffer); indexOffset = (unsigned) ib->ptr / indexSize; } else { @@ -321,7 +322,7 @@ st_draw_vbo(GLcontext *ctx, prims[i].start + indexOffset, prims[i].count); } - winsys->buffer_reference(winsys, &indexBuf, NULL); + pipe_buffer_reference(winsys, &indexBuf, NULL); } else { /* non-indexed */ @@ -333,7 +334,7 @@ st_draw_vbo(GLcontext *ctx, /* unreference buffers (frees wrapped user-space buffer objects) */ for (attr = 0; attr < vs->num_inputs; attr++) { - winsys->buffer_reference(winsys, &vbuffer[attr].buffer, NULL); + pipe_buffer_reference(winsys, &vbuffer[attr].buffer, NULL); assert(!vbuffer[attr].buffer); pipe->set_vertex_buffer(pipe, attr, &vbuffer[attr]); } @@ -358,7 +359,7 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, const float height = ctx->DrawBuffer->Height; const unsigned vertex_bytes = numVertex * numAttribs * 4 * sizeof(float); struct pipe_context *pipe = ctx->st->pipe; - struct pipe_buffer_handle *vbuf; + struct pipe_buffer *vbuf; struct pipe_vertex_buffer vbuffer; struct pipe_vertex_element velement; unsigned i; @@ -404,7 +405,7 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, pipe->draw_arrays(pipe, prim, 0, numVertex); /* XXX: do one-time */ - pipe->winsys->buffer_reference(pipe->winsys, &vbuf, NULL); + pipe_buffer_reference(pipe->winsys, &vbuf, NULL); } @@ -465,7 +466,7 @@ st_feedback_draw_vbo(GLcontext *ctx, struct pipe_winsys *winsys = pipe->winsys; const struct st_vertex_program *vp; const struct pipe_shader_state *vs; - struct pipe_buffer_handle *index_buffer_handle = 0; + struct pipe_buffer *index_buffer_handle = 0; struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; GLuint attr, i; ubyte *mapped_constants; @@ -513,7 +514,7 @@ st_feedback_draw_vbo(GLcontext *ctx, assert(stobj->buffer); vbuffer[attr].buffer = NULL; - winsys->buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); + pipe_buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); vbuffer[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ velement.src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; } @@ -605,7 +606,7 @@ st_feedback_draw_vbo(GLcontext *ctx, if (draw->vertex_buffer[i].buffer) { pipe->winsys->buffer_unmap(pipe->winsys, draw->vertex_buffer[i].buffer); - winsys->buffer_reference(winsys, &draw->vertex_buffer[i].buffer, NULL); + pipe_buffer_reference(winsys, &draw->vertex_buffer[i].buffer, NULL); draw_set_mapped_vertex_buffer(draw, i, NULL); } } -- cgit v1.2.3 From bd8bf60b6f2402895e7159a4df644f8a4a307cf5 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 25 Jan 2008 14:46:43 -0800 Subject: Work around parameter differences for main in different SDK versions --- src/mesa/pipe/cell/spu/spu_main.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 880f8de550..2b88c54576 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -616,13 +616,21 @@ one_time_init(void) } +/* In some versions of the SDK the SPE main takes 'unsigned long' as a + * parameter. In others it takes 'unsigned long long'. Use a define to + * select between the two. + */ +#ifdef SPU_MAIN_PARAM_LONG_LONG +typedef unsigned long long main_param_t +#else +typedef unsigned long main_param_t +#endif + /** * SPE entrypoint. - * Note: example programs declare params as 'unsigned long long' but - * that doesn't work. */ int -main(unsigned long speid, unsigned long argp) +main(main_param_t speid, main_param_t argp) { int tag = 0; -- cgit v1.2.3 From 0bfd085e2866fbbd40209dcee23f0e6240583fe8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 15:59:27 -0700 Subject: gallium: replace prim pipeline begin/end() functions with flush() This is basically half of Keith's draw/flush patch. The stage->point/line/tri() functions are now self-validating, the validator functions are installed by the flush() function. There were excessive calls to validate_pipeline(), however. This was caused by draw_prim_queue_flush() keeping a local 'first' variable that always pointed to the validate functions. Replaced 'first' with 'draw->pipeline.first'. Performance in gears is up just slightly with this patch. --- src/mesa/pipe/cell/ppu/cell_draw_arrays.c | 9 +-- src/mesa/pipe/draw/draw_clip.c | 17 +----- src/mesa/pipe/draw/draw_cull.c | 31 +++++----- src/mesa/pipe/draw/draw_flatshade.c | 67 +++++++++++++-------- src/mesa/pipe/draw/draw_offset.c | 37 +++++++----- src/mesa/pipe/draw/draw_prim.c | 23 +++---- src/mesa/pipe/draw/draw_private.h | 5 +- src/mesa/pipe/draw/draw_stipple.c | 16 ++--- src/mesa/pipe/draw/draw_twoside.c | 99 ++++++++++++++++--------------- src/mesa/pipe/draw/draw_unfilled.c | 33 ++++++----- src/mesa/pipe/draw/draw_validate.c | 54 ++++++++++++++--- src/mesa/pipe/draw/draw_vbuf.c | 5 +- src/mesa/pipe/draw/draw_wide_prims.c | 51 ++++++++++------ src/mesa/pipe/i915simple/i915_prim_emit.c | 11 +--- src/mesa/pipe/softpipe/sp_draw_arrays.c | 9 +-- src/mesa/pipe/softpipe/sp_prim_setup.c | 43 ++++++++++++-- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 4 +- src/mesa/state_tracker/st_cb_feedback.c | 24 ++------ src/mesa/state_tracker/st_cb_rasterpos.c | 11 +--- 19 files changed, 309 insertions(+), 240 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c index 8286da712c..9c41b4a523 100644 --- a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c +++ b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c @@ -143,21 +143,18 @@ cell_draw_elements(struct pipe_context *pipe, /* draw! */ draw_arrays(draw, mode, start, count); - /* always flush for now */ - draw_flush(draw); - /* - * unmap vertex/index buffers + * unmap vertex/index buffers - will cause draw module to flush */ for (i = 0; i < PIPE_ATTRIB_MAX; i++) { if (sp->vertex_buffer[i].buffer) { - pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); draw_set_mapped_vertex_buffer(draw, i, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); } } if (indexBuffer) { - pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); draw_set_mapped_element_buffer(draw, 0, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); } /* Note: leave drawing surfaces mapped */ diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index e0eb656364..2d410e3244 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -346,15 +346,6 @@ do_clip_line( struct draw_stage *stage, } -static void clip_begin( struct draw_stage *stage ) -{ - /* should always have position, at least */ - assert(stage->draw->num_vs_outputs > 0); - - stage->next->begin( stage->next ); -} - - static void clip_point( struct draw_stage *stage, struct prim_header *header ) @@ -402,10 +393,9 @@ clip_tri( struct draw_stage *stage, } } - -static void clip_end( struct draw_stage *stage ) +static void clip_flush( struct draw_stage *stage, unsigned flags ) { - stage->next->end( stage->next ); + stage->next->flush( stage->next, flags ); } @@ -433,11 +423,10 @@ struct draw_stage *draw_clip_stage( struct draw_context *draw ) draw_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); clipper->stage.draw = draw; - clipper->stage.begin = clip_begin; clipper->stage.point = clip_point; clipper->stage.line = clip_line; clipper->stage.tri = clip_tri; - clipper->stage.end = clip_end; + clipper->stage.flush = clip_flush; clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; clipper->stage.destroy = clip_destroy; diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index 9bd53f45f2..05c274e4dc 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -50,14 +50,6 @@ static INLINE struct cull_stage *cull_stage( struct draw_stage *stage ) } -static void cull_begin( struct draw_stage *stage ) -{ - struct cull_stage *cull = cull_stage(stage); - - cull->winding = stage->draw->rasterizer->cull_mode; - - stage->next->begin( stage->next ); -} static void cull_tri( struct draw_stage *stage, @@ -90,6 +82,18 @@ static void cull_tri( struct draw_stage *stage, } } +static void cull_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct cull_stage *cull = cull_stage(stage); + + cull->winding = stage->draw->rasterizer->cull_mode; + + stage->tri = cull_tri; + stage->tri( stage, header ); +} + + static void cull_line( struct draw_stage *stage, struct prim_header *header ) @@ -105,12 +109,12 @@ static void cull_point( struct draw_stage *stage, } -static void cull_end( struct draw_stage *stage ) +static void cull_flush( struct draw_stage *stage, unsigned flags ) { - stage->next->end( stage->next ); + stage->tri = cull_first_tri; + stage->next->flush( stage->next, flags ); } - static void cull_reset_stipple_counter( struct draw_stage *stage ) { stage->next->reset_stipple_counter( stage->next ); @@ -135,11 +139,10 @@ struct draw_stage *draw_cull_stage( struct draw_context *draw ) cull->stage.draw = draw; cull->stage.next = NULL; - cull->stage.begin = cull_begin; cull->stage.point = cull_point; cull->stage.line = cull_line; - cull->stage.tri = cull_tri; - cull->stage.end = cull_end; + cull->stage.tri = cull_first_tri; + cull->stage.flush = cull_flush; cull->stage.reset_stipple_counter = cull_reset_stipple_counter; cull->stage.destroy = cull_destroy; diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 8444c53310..1419f287d2 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -50,25 +50,6 @@ flat_stage(struct draw_stage *stage) } -static void flatshade_begin( struct draw_stage *stage ) -{ - struct flat_stage *flat = flat_stage(stage); - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; - uint i; - - /* Find which vertex shader outputs are colors, make a list */ - flat->num_color_attribs = 0; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || - vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - flat->color_attribs[flat->num_color_attribs++] = i; - } - } - - stage->next->begin( stage->next ); -} - - /** Copy all the color attributes from 'src' vertex to 'dst' vertex */ static INLINE void copy_colors( struct draw_stage *stage, struct vertex_header *dst, @@ -144,9 +125,46 @@ static void flatshade_point( struct draw_stage *stage, } -static void flatshade_end( struct draw_stage *stage ) +static void flatshade_init_state( struct draw_stage *stage ) +{ + struct flat_stage *flat = flat_stage(stage); + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + /* Find which vertex shader outputs are colors, make a list */ + flat->num_color_attribs = 0; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + flat->color_attribs[flat->num_color_attribs++] = i; + } + } + + stage->line = flatshade_line; + stage->tri = flatshade_tri; +} + +static void flatshade_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + flatshade_init_state( stage ); + stage->tri( stage, header ); +} + +static void flatshade_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + flatshade_init_state( stage ); + stage->line( stage, header ); +} + + +static void flatshade_flush( struct draw_stage *stage, + unsigned flags ) { - stage->next->end( stage->next ); + stage->tri = flatshade_first_tri; + stage->line = flatshade_first_line; + stage->next->flush( stage->next, flags ); } @@ -174,11 +192,10 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) flatshade->stage.draw = draw; flatshade->stage.next = NULL; - flatshade->stage.begin = flatshade_begin; flatshade->stage.point = flatshade_point; - flatshade->stage.line = flatshade_line; - flatshade->stage.tri = flatshade_tri; - flatshade->stage.end = flatshade_end; + flatshade->stage.line = flatshade_first_line; + flatshade->stage.tri = flatshade_first_tri; + flatshade->stage.flush = flatshade_flush; flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; flatshade->stage.destroy = flatshade_destroy; diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index f8a01db3dd..a2990ee8a8 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -52,16 +52,7 @@ static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) } -static void offset_begin( struct draw_stage *stage ) -{ - struct offset_stage *offset = offset_stage(stage); - float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ - offset->units = stage->draw->rasterizer->offset_units * mrd; - offset->scale = stage->draw->rasterizer->offset_scale; - - stage->next->begin( stage->next ); -} /** @@ -124,24 +115,39 @@ static void offset_tri( struct draw_stage *stage, } +static void offset_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ + + offset->units = stage->draw->rasterizer->offset_units * mrd; + offset->scale = stage->draw->rasterizer->offset_scale; + + stage->tri = offset_tri; + stage->tri( stage, header ); +} + static void offset_line( struct draw_stage *stage, - struct prim_header *header ) + struct prim_header *header ) { stage->next->line( stage->next, header ); } static void offset_point( struct draw_stage *stage, - struct prim_header *header ) + struct prim_header *header ) { stage->next->point( stage->next, header ); } -static void offset_end( struct draw_stage *stage ) +static void offset_flush( struct draw_stage *stage, + unsigned flags ) { - stage->next->end( stage->next ); + stage->tri = offset_first_tri; + stage->next->flush( stage->next, flags ); } @@ -169,11 +175,10 @@ struct draw_stage *draw_offset_stage( struct draw_context *draw ) offset->stage.draw = draw; offset->stage.next = NULL; - offset->stage.begin = offset_begin; offset->stage.point = offset_point; offset->stage.line = offset_line; - offset->stage.tri = offset_tri; - offset->stage.end = offset_end; + offset->stage.tri = offset_first_tri; + offset->stage.flush = offset_flush; offset->stage.reset_stipple_counter = offset_reset_stipple_counter; offset->stage.destroy = offset_destroy; diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index f8fc23b510..5703f5f0b0 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -57,7 +57,7 @@ static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { static void draw_prim_queue_flush( struct draw_context *draw ) { - struct draw_stage *first = draw->pipeline.first; + // struct draw_stage *first = draw->pipeline.first; unsigned i; if (0) @@ -69,27 +69,31 @@ static void draw_prim_queue_flush( struct draw_context *draw ) if (draw->vs.queue_nr) draw_vertex_shader_queue_flush(draw); + /* NOTE: we cannot save draw->pipeline->first in a local var because + * draw->pipeline->first is often changed by the first call to tri(), + * line(), etc. + */ switch (draw->reduced_prim) { case RP_TRI: for (i = 0; i < draw->pq.queue_nr; i++) { if (draw->pq.queue[i].reset_line_stipple) - first->reset_stipple_counter( first ); + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - first->tri( first, &draw->pq.queue[i] ); + draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); } break; case RP_LINE: for (i = 0; i < draw->pq.queue_nr; i++) { if (draw->pq.queue[i].reset_line_stipple) - first->reset_stipple_counter( first ); + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - first->line( first, &draw->pq.queue[i] ); + draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); } break; case RP_POINT: - first->reset_stipple_counter( first ); + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); for (i = 0; i < draw->pq.queue_nr; i++) - first->point( first, &draw->pq.queue[i] ); + draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); break; } @@ -119,7 +123,7 @@ void draw_do_flush( struct draw_context *draw, if ((flush & DRAW_FLUSH_DRAW) && draw->drawing) { - draw->pipeline.first->end( draw->pipeline.first ); + draw->pipeline.first->flush( draw->pipeline.first, ~0 ); draw->drawing = FALSE; draw->prim = ~0; draw->pipeline.first = draw->pipeline.validate; @@ -415,9 +419,6 @@ draw_arrays(struct draw_context *draw, unsigned prim, { if (!draw->drawing) { draw->drawing = TRUE; - - /* tell drawing pipeline we're beginning drawing */ - draw->pipeline.first->begin( draw->pipeline.first ); } if (draw->prim != prim) { diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 685ec4a644..e393fa5fe2 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -101,8 +101,6 @@ struct draw_stage struct vertex_header **tmp; /**< temp vert storage, such as for clipping */ unsigned nr_tmps; - void (*begin)( struct draw_stage * ); - void (*point)( struct draw_stage *, struct prim_header * ); @@ -112,7 +110,8 @@ struct draw_stage void (*tri)( struct draw_stage *, struct prim_header * ); - void (*end)( struct draw_stage * ); + void (*flush)( struct draw_stage *, + unsigned flags ); void (*reset_stipple_counter)( struct draw_stage * ); diff --git a/src/mesa/pipe/draw/draw_stipple.c b/src/mesa/pipe/draw/draw_stipple.c index 2a47eb7be2..9029101916 100644 --- a/src/mesa/pipe/draw/draw_stipple.c +++ b/src/mesa/pipe/draw/draw_stipple.c @@ -173,7 +173,8 @@ reset_stipple_counter(struct draw_stage *stage) static void -stipple_begin(struct draw_stage *stage) +stipple_first_line(struct draw_stage *stage, + struct prim_header *header) { struct stipple_stage *stipple = stipple_stage(stage); struct draw_context *draw = stage->draw; @@ -181,14 +182,16 @@ stipple_begin(struct draw_stage *stage) stipple->pattern = draw->rasterizer->line_stipple_pattern; stipple->factor = draw->rasterizer->line_stipple_factor + 1; - stage->next->begin( stage->next ); + stage->line = stipple_line; + stage->line( stage, header ); } static void -stipple_end(struct draw_stage *stage) +stipple_flush(struct draw_stage *stage, unsigned flags) { - stage->next->end( stage->next ); + stage->line = stipple_first_line; + stage->next->flush( stage->next, flags ); } @@ -224,12 +227,11 @@ struct draw_stage *draw_stipple_stage( struct draw_context *draw ) stipple->stage.draw = draw; stipple->stage.next = NULL; - stipple->stage.begin = stipple_begin; stipple->stage.point = passthrough_point; - stipple->stage.line = stipple_line; + stipple->stage.line = stipple_first_line; stipple->stage.tri = passthrough_tri; stipple->stage.reset_stipple_counter = reset_stipple_counter; - stipple->stage.end = stipple_end; + stipple->stage.flush = stipple_flush; stipple->stage.destroy = stipple_destroy; return &stipple->stage; diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index 75c51ec6a9..ad2aaf10bb 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -48,48 +48,6 @@ static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage ) } -static void twoside_begin( struct draw_stage *stage ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; - uint i; - - twoside->attrib_front0 = 0; - twoside->attrib_front1 = 0; - twoside->attrib_back0 = 0; - twoside->attrib_back1 = 0; - - /* Find which vertex shader outputs are front/back colors */ - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { - if (vs->output_semantic_index[i] == 0) - twoside->attrib_front0 = i; - else - twoside->attrib_front1 = i; - } - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - if (vs->output_semantic_index[i] == 0) - twoside->attrib_back0 = i; - else - twoside->attrib_back1 = i; - } - } - - if (!twoside->attrib_back0) - twoside->attrib_front0 = 0; - - if (!twoside->attrib_back1) - twoside->attrib_front1 = 0; - - /* - * We'll multiply the primitive's determinant by this sign to determine - * if the triangle is back-facing (negative). - * sign = -1 for CCW, +1 for CW - */ - twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; - - stage->next->begin( stage->next ); -} /** @@ -157,10 +115,56 @@ static void twoside_point( struct draw_stage *stage, } -static void twoside_end( struct draw_stage *stage ) +static void twoside_first_tri( struct draw_stage *stage, + struct prim_header *header ) { - /* pass-through */ - stage->next->end( stage->next ); + struct twoside_stage *twoside = twoside_stage(stage); + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + twoside->attrib_front0 = 0; + twoside->attrib_front1 = 0; + twoside->attrib_back0 = 0; + twoside->attrib_back1 = 0; + + /* Find which vertex shader outputs are front/back colors */ + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { + if (vs->output_semantic_index[i] == 0) + twoside->attrib_front0 = i; + else + twoside->attrib_front1 = i; + } + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + if (vs->output_semantic_index[i] == 0) + twoside->attrib_back0 = i; + else + twoside->attrib_back1 = i; + } + } + + if (!twoside->attrib_back0) + twoside->attrib_front0 = 0; + + if (!twoside->attrib_back1) + twoside->attrib_front1 = 0; + + /* + * We'll multiply the primitive's determinant by this sign to determine + * if the triangle is back-facing (negative). + * sign = -1 for CCW, +1 for CW + */ + twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; + + stage->tri = twoside_tri; + stage->tri( stage, header ); +} + + +static void twoside_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->tri = twoside_first_tri; + stage->next->flush( stage->next, flags ); } @@ -188,11 +192,10 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) twoside->stage.draw = draw; twoside->stage.next = NULL; - twoside->stage.begin = twoside_begin; twoside->stage.point = twoside_point; twoside->stage.line = twoside_line; - twoside->stage.tri = twoside_tri; - twoside->stage.end = twoside_end; + twoside->stage.tri = twoside_first_tri; + twoside->stage.flush = twoside_flush; twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; twoside->stage.destroy = twoside_destroy; diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index 786826b33c..364bda8b79 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -55,15 +55,6 @@ static INLINE struct unfilled_stage *unfilled_stage( struct draw_stage *stage ) } -static void unfilled_begin( struct draw_stage *stage ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - - unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */ - unfilled->mode[1] = stage->draw->rasterizer->fill_cw; /* back */ - - stage->next->begin( stage->next ); -} static void point( struct draw_stage *stage, struct vertex_header *v0 ) @@ -142,6 +133,20 @@ static void unfilled_tri( struct draw_stage *stage, } } + +static void unfilled_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + + unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */ + unfilled->mode[1] = stage->draw->rasterizer->fill_cw; /* back */ + + stage->tri = unfilled_tri; + stage->tri( stage, header ); +} + + static void unfilled_line( struct draw_stage *stage, struct prim_header *header ) { @@ -156,9 +161,10 @@ static void unfilled_point( struct draw_stage *stage, } -static void unfilled_end( struct draw_stage *stage ) +static void unfilled_flush( struct draw_stage *stage, + unsigned flags ) { - stage->next->end( stage->next ); + stage->next->flush( stage->next, flags ); } @@ -187,11 +193,10 @@ struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) unfilled->stage.draw = draw; unfilled->stage.next = NULL; unfilled->stage.tmp = NULL; - unfilled->stage.begin = unfilled_begin; unfilled->stage.point = unfilled_point; unfilled->stage.line = unfilled_line; - unfilled->stage.tri = unfilled_tri; - unfilled->stage.end = unfilled_end; + unfilled->stage.tri = unfilled_first_tri; + unfilled->stage.flush = unfilled_flush; unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; unfilled->stage.destroy = unfilled_destroy; diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index 3b1f5179a9..a626fb1fba 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -39,7 +39,7 @@ /** * Rebuild the rendering pipeline. */ -static void validate_begin( struct draw_stage *stage ) +static struct draw_stage *validate_pipeline( struct draw_stage *stage ) { struct draw_context *draw = stage->draw; struct draw_stage *next = draw->pipeline.rasterize; @@ -106,7 +106,45 @@ static void validate_begin( struct draw_stage *stage ) } draw->pipeline.first = next; - draw->pipeline.first->begin( draw->pipeline.first ); + //BP draw->pipeline.first->begin( draw->pipeline.first ); + return next; +} + + +static void validate_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->tri( pipeline, header ); +} + +static void validate_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->line( pipeline, header ); +} + +static void validate_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->point( pipeline, header ); +} + +static void validate_reset_stipple_counter( struct draw_stage *stage ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->reset_stipple_counter( pipeline ); +} + +static void validate_flush( struct draw_stage *stage, + unsigned flags ) +{ + /* May need to pass a backend flush on to the rasterize stage. + */ + if (stage->next) + stage->next->flush( stage->next, flags ); } @@ -124,13 +162,13 @@ struct draw_stage *draw_validate_stage( struct draw_context *draw ) struct draw_stage *stage = CALLOC_STRUCT(draw_stage); stage->draw = draw; + stage->next = NULL; - stage->begin = validate_begin; - stage->point = NULL; - stage->line = NULL; - stage->tri = NULL; - stage->end = NULL; - stage->reset_stipple_counter = NULL; + stage->point = validate_point; + stage->line = validate_line; + stage->tri = validate_tri; + stage->flush = validate_flush; + stage->reset_stipple_counter = validate_reset_stipple_counter; stage->destroy = validate_destroy; return stage; diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 82051d2e65..d827f51d56 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -395,7 +395,7 @@ vbuf_begin( struct draw_stage *stage ) static void -vbuf_end( struct draw_stage *stage ) +vbuf_flush( struct draw_stage *stage, unsigned flags ) { // vbuf_flush_indices( stage ); /* XXX: Overkill */ @@ -432,11 +432,10 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); vbuf->stage.draw = draw; - vbuf->stage.begin = vbuf_begin; vbuf->stage.point = vbuf_first_point; vbuf->stage.line = vbuf_first_line; vbuf->stage.tri = vbuf_first_tri; - vbuf->stage.end = vbuf_end; + vbuf->stage.flush = vbuf_flush; vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter; vbuf->stage.destroy = vbuf_destroy; diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index f71a38efba..26f5d10dc0 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -262,26 +262,19 @@ static void wide_point( struct draw_stage *stage, } -static void wide_begin( struct draw_stage *stage ) +static void wide_first_point( struct draw_stage *stage, + struct prim_header *header ) { struct wide_stage *wide = wide_stage(stage); struct draw_context *draw = stage->draw; wide->half_point_size = 0.5f * draw->rasterizer->point_size; - wide->half_line_width = 0.5f * draw->rasterizer->line_width; - - if (draw->rasterizer->line_width != 1.0) { - wide->stage.line = wide_line; - } - else { - wide->stage.line = passthrough_line; - } if (draw->rasterizer->point_size != 1.0) { - wide->stage.point = wide_point; + stage->point = wide_point; } else { - wide->stage.point = passthrough_point; + stage->point = passthrough_point; } if (draw->rasterizer->point_sprite) { @@ -299,6 +292,7 @@ static void wide_begin( struct draw_stage *stage ) } wide->psize_slot = -1; + if (draw->rasterizer->point_size_per_vertex) { /* find PSIZ vertex output */ const struct draw_vertex_shader *vs = draw->vertex_shader; @@ -311,13 +305,35 @@ static void wide_begin( struct draw_stage *stage ) } } - stage->next->begin( stage->next ); + stage->point( stage, header ); +} + + + +static void wide_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct wide_stage *wide = wide_stage(stage); + struct draw_context *draw = stage->draw; + + wide->half_line_width = 0.5f * draw->rasterizer->line_width; + + if (draw->rasterizer->line_width != 1.0) { + wide->stage.line = wide_line; + } + else { + wide->stage.line = passthrough_line; + } + + stage->line( stage, header ); } -static void wide_end( struct draw_stage *stage ) +static void wide_flush( struct draw_stage *stage, unsigned flags ) { - stage->next->end( stage->next ); + stage->line = wide_first_line; + stage->point = wide_first_point; + stage->next->flush( stage->next, flags ); } @@ -342,11 +358,10 @@ struct draw_stage *draw_wide_stage( struct draw_context *draw ) wide->stage.draw = draw; wide->stage.next = NULL; - wide->stage.begin = wide_begin; - wide->stage.point = wide_point; - wide->stage.line = wide_line; + wide->stage.point = wide_first_point; + wide->stage.line = wide_first_line; wide->stage.tri = passthrough_tri; - wide->stage.end = wide_end; + wide->stage.flush = wide_flush; wide->stage.reset_stipple_counter = draw_reset_stipple_counter; wide->stage.destroy = wide_destroy; diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c index f74671b39c..c4a706c37d 100644 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ b/src/mesa/pipe/i915simple/i915_prim_emit.c @@ -180,13 +180,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) } - -static void setup_begin( struct draw_stage *stage ) -{ -} - - -static void setup_end( struct draw_stage *stage ) +static void setup_flush( struct draw_stage *stage, unsigned flags ) { } @@ -210,11 +204,10 @@ struct draw_stage *i915_draw_render_stage( struct i915_context *i915 ) setup->i915 = i915; setup->stage.draw = i915->draw; - setup->stage.begin = setup_begin; setup->stage.point = setup_point; setup->stage.line = setup_line; setup->stage.tri = setup_tri; - setup->stage.end = setup_end; + setup->stage.flush = setup_flush; setup->stage.reset_stipple_counter = reset_stipple_counter; setup->stage.destroy = render_destroy; diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c index 423c91d4b8..71a303a8b5 100644 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ b/src/mesa/pipe/softpipe/sp_draw_arrays.c @@ -142,21 +142,18 @@ softpipe_draw_elements(struct pipe_context *pipe, /* draw! */ draw_arrays(draw, mode, start, count); - /* always flush for now */ - draw_flush(draw); - /* - * unmap vertex/index buffers + * unmap vertex/index buffers - will cause draw module to flush */ for (i = 0; i < PIPE_ATTRIB_MAX; i++) { if (sp->vertex_buffer[i].buffer) { - pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); draw_set_mapped_vertex_buffer(draw, i, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); } } if (indexBuffer) { - pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); draw_set_mapped_element_buffer(draw, 0, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); } diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 478ecff2fb..088d8a4c07 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -1170,11 +1170,43 @@ static void setup_begin( struct draw_stage *stage ) setup->quad.nr_attrs = fs->num_inputs; sp->quad.first->begin(sp->quad.first); + + stage->point = setup_point; + stage->line = setup_line; + stage->tri = setup_tri; } -static void setup_end( struct draw_stage *stage ) +static void setup_first_point( struct draw_stage *stage, + struct prim_header *header ) +{ + setup_begin(stage); + stage->point( stage, header ); +} + +static void setup_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + setup_begin(stage); + stage->line( stage, header ); +} + + +static void setup_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + setup_begin(stage); + stage->tri( stage, header ); +} + + + +static void setup_flush( struct draw_stage *stage, + unsigned flags ) { + stage->point = setup_first_point; + stage->line = setup_first_line; + stage->tri = setup_first_tri; } @@ -1198,11 +1230,10 @@ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) setup->softpipe = softpipe; setup->stage.draw = softpipe->draw; - setup->stage.begin = setup_begin; - setup->stage.point = setup_point; - setup->stage.line = setup_line; - setup->stage.tri = setup_tri; - setup->stage.end = setup_end; + setup->stage.point = setup_first_point; + setup->stage.line = setup_first_line; + setup->stage.tri = setup_first_tri; + setup->stage.flush = setup_flush; setup->stage.reset_stipple_counter = reset_stipple_counter; setup->stage.destroy = render_destroy; diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index dfabae8302..c9089e7eb2 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -149,8 +149,6 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) prim.edgeflags = 0; prim.pad = 0; - setup->begin( setup ); - switch (cvbr->prim) { case PIPE_PRIM_TRIANGLES: for (i = 0; i < nr_indices; i += 3) { @@ -182,7 +180,7 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) break; } - setup->end( setup ); + setup->flush( setup, 0 ); } diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 43543df1a8..31744151f1 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -159,14 +159,7 @@ feedback_point( struct draw_stage *stage, struct prim_header *prim ) static void -feedback_end( struct draw_stage *stage ) -{ - /* no-op */ -} - - -static void -feedback_begin( struct draw_stage *stage ) +feedback_flush( struct draw_stage *stage, unsigned flags ) { /* no-op */ } @@ -190,11 +183,10 @@ draw_glfeedback_stage(GLcontext *ctx, struct draw_context *draw) fs->stage.draw = draw; fs->stage.next = NULL; - fs->stage.begin = feedback_begin; fs->stage.point = feedback_point; fs->stage.line = feedback_line; fs->stage.tri = feedback_tri; - fs->stage.end = feedback_end; + fs->stage.flush = feedback_flush; fs->stage.reset_stipple_counter = feedback_reset_stipple_counter; fs->ctx = ctx; @@ -234,14 +226,7 @@ select_point( struct draw_stage *stage, struct prim_header *prim ) static void -select_begin( struct draw_stage *stage ) -{ - /* no-op */ -} - - -static void -select_end( struct draw_stage *stage ) +select_flush( struct draw_stage *stage, unsigned flags ) { /* no-op */ } @@ -264,11 +249,10 @@ draw_glselect_stage(GLcontext *ctx, struct draw_context *draw) fs->stage.draw = draw; fs->stage.next = NULL; - fs->stage.begin = select_begin; fs->stage.point = select_point; fs->stage.line = select_line; fs->stage.tri = select_tri; - fs->stage.end = select_end; + fs->stage.flush = select_flush; fs->stage.reset_stipple_counter = select_reset_stipple_counter; fs->ctx = ctx; diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 9e20e94dc7..7e347c4893 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -73,13 +73,7 @@ rastpos_stage( struct draw_stage *stage ) } static void -rastpos_begin( struct draw_stage *stage ) -{ - /* no-op */ -} - -static void -rastpos_end( struct draw_stage *stage ) +rastpos_flush( struct draw_stage *stage, unsigned flags ) { /* no-op */ } @@ -183,11 +177,10 @@ new_draw_rastpos_stage(GLcontext *ctx, struct draw_context *draw) rs->stage.draw = draw; rs->stage.next = NULL; - rs->stage.begin = rastpos_begin; rs->stage.point = rastpos_point; rs->stage.line = rastpos_line; rs->stage.tri = rastpos_tri; - rs->stage.end = rastpos_end; + rs->stage.flush = rastpos_flush; rs->stage.destroy = rastpos_destroy; rs->stage.reset_stipple_counter = rastpos_reset_stipple_counter; rs->stage.destroy = rastpos_destroy; -- cgit v1.2.3 From 1603a33fb276d7e78a2e872dfa05aa0093d1329a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 17:21:05 -0700 Subject: gallium: better flush logic in draw module This is the other half of Keith's draw/flush patch. There are now 5 flush flags to control what's flushed (post-xform vertex cache, prim cache, vbuf, etc). The gears slow-down in this part of the patch was due to the cull stage not getting invoked. It was unconditional before, but is now gated by 'need_det'. But it also needs to be gated by draw->rasterizer->cull_mode. Gears uses back-face culling. --- src/mesa/pipe/draw/draw_context.c | 29 ++++++----- src/mesa/pipe/draw/draw_prim.c | 85 +++++++++++++-------------------- src/mesa/pipe/draw/draw_private.h | 16 +++---- src/mesa/pipe/draw/draw_validate.c | 32 ++++++++----- src/mesa/pipe/draw/draw_vbuf.c | 17 +++---- src/mesa/pipe/draw/draw_vertex_cache.c | 9 ++-- src/mesa/pipe/draw/draw_vertex_shader.c | 4 +- 7 files changed, 89 insertions(+), 103 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index ff23288fa8..e8ca1f035b 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -80,7 +80,7 @@ struct draw_context *draw_create( void ) draw->convert_wide_points = TRUE; draw->convert_wide_lines = TRUE; - draw->prim = ~0; /* != any of PIPE_PRIM_x */ + draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ draw_vertex_cache_invalidate( draw ); draw_set_mapped_element_buffer( draw, 0, NULL ); @@ -111,8 +111,7 @@ void draw_destroy( struct draw_context *draw ) void draw_flush( struct draw_context *draw ) { - if (draw->drawing) - draw_do_flush( draw, DRAW_FLUSH_DRAW ); + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); } @@ -124,7 +123,8 @@ void draw_flush( struct draw_context *draw ) void draw_set_rasterizer_state( struct draw_context *draw, const struct pipe_rasterizer_state *raster ) { - draw_flush( draw ); + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->rasterizer = raster; } @@ -137,7 +137,8 @@ void draw_set_rasterizer_state( struct draw_context *draw, void draw_set_rasterize_stage( struct draw_context *draw, struct draw_stage *stage ) { - draw_flush( draw ); + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->pipeline.rasterize = stage; } @@ -148,7 +149,7 @@ void draw_set_rasterize_stage( struct draw_context *draw, void draw_set_clip_state( struct draw_context *draw, const struct pipe_clip_state *clip ) { - draw_flush( draw ); + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); assert(clip->nr <= PIPE_MAX_CLIP_PLANES); memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); @@ -162,7 +163,7 @@ void draw_set_clip_state( struct draw_context *draw, void draw_set_viewport_state( struct draw_context *draw, const struct pipe_viewport_state *viewport ) { - draw_flush( draw ); + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); draw->viewport = *viewport; /* struct copy */ } @@ -173,8 +174,7 @@ draw_set_vertex_buffer(struct draw_context *draw, unsigned attr, const struct pipe_vertex_buffer *buffer) { - draw_flush( draw ); - + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); assert(attr < PIPE_ATTRIB_MAX); draw->vertex_buffer[attr] = *buffer; } @@ -185,8 +185,7 @@ draw_set_vertex_element(struct draw_context *draw, unsigned attr, const struct pipe_vertex_element *element) { - draw_flush( draw ); - + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); assert(attr < PIPE_ATTRIB_MAX); draw->vertex_element[attr] = *element; } @@ -199,8 +198,7 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw, unsigned attr, const void *buffer) { - draw_flush( draw ); - + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); draw->user.vbuffer[attr] = buffer; } @@ -209,8 +207,7 @@ void draw_set_mapped_constant_buffer(struct draw_context *draw, const void *buffer) { - draw_flush( draw ); - + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); draw->user.constants = buffer; } @@ -222,6 +219,7 @@ draw_set_mapped_constant_buffer(struct draw_context *draw, void draw_convert_wide_points(struct draw_context *draw, boolean enable) { + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); draw->convert_wide_points = enable; } @@ -233,6 +231,7 @@ draw_convert_wide_points(struct draw_context *draw, boolean enable) void draw_convert_wide_lines(struct draw_context *draw, boolean enable) { + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); draw->convert_wide_lines = enable; } diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 5703f5f0b0..243381aec0 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -57,17 +57,14 @@ static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { static void draw_prim_queue_flush( struct draw_context *draw ) { - // struct draw_stage *first = draw->pipeline.first; unsigned i; if (0) fprintf(stdout,"Flushing with %d prims, %d verts\n", draw->pq.queue_nr, draw->vs.queue_nr); - /* Make sure all vertices are available/shaded: - */ - if (draw->vs.queue_nr) - draw_vertex_shader_queue_flush(draw); + if (draw->pq.queue_nr == 0) + return; /* NOTE: we cannot save draw->pipeline->first in a local var because * draw->pipeline->first is often changed by the first call to tri(), @@ -102,33 +99,32 @@ static void draw_prim_queue_flush( struct draw_context *draw ) } -void draw_do_flush( struct draw_context *draw, - unsigned flush ) + +void draw_do_flush( struct draw_context *draw, unsigned flags ) { - if ((flush & (DRAW_FLUSH_PRIM_QUEUE | - DRAW_FLUSH_VERTEX_CACHE_INVALIDATE | - DRAW_FLUSH_DRAW)) && - draw->pq.queue_nr) - { - draw_prim_queue_flush(draw); - } + if (0) + fprintf(stdout,"Flushing with %d verts, %d prims\n", + draw->vs.queue_nr, + draw->pq.queue_nr ); - if ((flush & (DRAW_FLUSH_VERTEX_CACHE_INVALIDATE | - DRAW_FLUSH_DRAW)) && - draw->drawing) - { - draw_vertex_cache_invalidate(draw); - } - if ((flush & DRAW_FLUSH_DRAW) && - draw->drawing) - { - draw->pipeline.first->flush( draw->pipeline.first, ~0 ); - draw->drawing = FALSE; - draw->prim = ~0; - draw->pipeline.first = draw->pipeline.validate; - } + if (flags >= DRAW_FLUSH_SHADER_QUEUE) { + draw_vertex_shader_queue_flush(draw); + + if (flags >= DRAW_FLUSH_PRIM_QUEUE) { + draw_prim_queue_flush(draw); + if (flags >= DRAW_FLUSH_VERTEX_CACHE) { + draw_vertex_cache_invalidate(draw); + + if (flags >= DRAW_FLUSH_STATE_CHANGE) { + draw->pipeline.first->flush( draw->pipeline.first, flags ); + draw->pipeline.first = draw->pipeline.validate; + draw->reduced_prim = ~0; + } + } + } + } } @@ -143,7 +139,7 @@ static struct prim_header *get_queued_prim( struct draw_context *draw, { if (!draw_vertex_cache_check_space( draw, nr_verts )) { // fprintf(stderr, "v"); - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE_INVALIDATE ); + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE ); } else if (draw->pq.queue_nr == PRIM_QUEUE_LENGTH) { // fprintf(stderr, "p"); @@ -251,13 +247,14 @@ static void do_quad( struct draw_context *draw, * Main entrypoint to draw some number of points/lines/triangles */ static void -draw_prim( struct draw_context *draw, unsigned start, unsigned count ) +draw_prim( struct draw_context *draw, + unsigned prim, unsigned start, unsigned count ) { unsigned i; // _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - switch (draw->prim) { + switch (prim) { case PIPE_PRIM_POINTS: for (i = 0; i < count; i ++) { do_point( draw, @@ -389,21 +386,6 @@ draw_prim( struct draw_context *draw, unsigned start, unsigned count ) } -static void -draw_set_prim( struct draw_context *draw, unsigned prim ) -{ - assert(prim >= PIPE_PRIM_POINTS); - assert(prim <= PIPE_PRIM_POLYGON); - - if (reduced_prim[prim] != draw->reduced_prim) { - draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); - draw->reduced_prim = reduced_prim[prim]; - } - - draw->prim = prim; -} - - /** @@ -417,16 +399,13 @@ void draw_arrays(struct draw_context *draw, unsigned prim, unsigned start, unsigned count) { - if (!draw->drawing) { - draw->drawing = TRUE; - } - - if (draw->prim != prim) { - draw_set_prim( draw, prim ); + if (reduced_prim[prim] != draw->reduced_prim) { + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->reduced_prim = reduced_prim[prim]; } /* drawing done here: */ - draw_prim(draw, start, count); + draw_prim(draw, prim, start, count); } diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index e393fa5fe2..1c2e88264f 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -111,7 +111,7 @@ struct draw_stage struct prim_header * ); void (*flush)( struct draw_stage *, - unsigned flags ); + unsigned flags ); void (*reset_stipple_counter)( struct draw_stage * ); @@ -191,8 +191,6 @@ struct draw_context boolean convert_wide_points; /**< convert wide points to tris? */ boolean convert_wide_lines; /**< convert side lines to tris? */ - boolean drawing; /**< do we presently have something queued for drawing? */ - unsigned prim; /**< current prim type: PIPE_PRIM_x */ unsigned reduced_prim; /** TGSI program interpreter runtime state */ @@ -278,14 +276,14 @@ extern void draw_vertex_fetch( struct draw_context *draw, unsigned count ); -#define DRAW_FLUSH_PRIM_QUEUE 0x1 -#define DRAW_FLUSH_VERTEX_CACHE_INVALIDATE 0x2 -#define DRAW_FLUSH_DRAW 0x4 +#define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ +#define DRAW_FLUSH_PRIM_QUEUE 0x2 +#define DRAW_FLUSH_VERTEX_CACHE 0x4 +#define DRAW_FLUSH_STATE_CHANGE 0x8 +#define DRAW_FLUSH_BACKEND 0x10 -void draw_do_flush( struct draw_context *draw, - unsigned flags ); - +void draw_do_flush( struct draw_context *draw, unsigned flags ); diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index a626fb1fba..86d5a5f814 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -43,6 +43,13 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) { struct draw_context *draw = stage->draw; struct draw_stage *next = draw->pipeline.rasterize; + int need_det = 0; + int precalc_flat = 0; + + /* Set the validate's next stage to the rasterize stage, so that it + * can be found later if needed for flushing. + */ + stage->next = next; /* * NOTE: we build up the pipeline in end-to-start order. @@ -61,29 +68,38 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) if (draw->rasterizer->line_stipple_enable) { draw->pipeline.stipple->next = next; next = draw->pipeline.stipple; + precalc_flat = 1; /* only needed for lines really */ } if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { draw->pipeline.unfilled->next = next; next = draw->pipeline.unfilled; + precalc_flat = 1; /* only needed for triangles really */ + need_det = 1; } if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) { draw->pipeline.offset->next = next; next = draw->pipeline.offset; + need_det = 1; } if (draw->rasterizer->light_twoside) { draw->pipeline.twoside->next = next; next = draw->pipeline.twoside; + need_det = 1; } /* Always run the cull stage as we calculate determinant there - * also. Fix this.. + * also. + * + * This can actually be a win as culling out the triangles can lead + * to less work emitting vertices, smaller vertex buffers, etc. + * It's difficult to say whether this will be true in general. */ - { + if (need_det || draw->rasterizer->cull_mode) { draw->pipeline.cull->next = next; next = draw->pipeline.cull; } @@ -94,23 +110,18 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) { draw->pipeline.clip->next = next; next = draw->pipeline.clip; + precalc_flat = 1; /* XXX: FIX ME! Only needed for clipped prims */ } - /* Do software flatshading prior to clipping. XXX: should only do - * this for clipped primitives, ie it is a part of the clip - * routine. - */ - if (draw->rasterizer->flatshade) { + if (draw->rasterizer->flatshade && precalc_flat) { draw->pipeline.flatshade->next = next; next = draw->pipeline.flatshade; } - + draw->pipeline.first = next; - //BP draw->pipeline.first->begin( draw->pipeline.first ); return next; } - static void validate_tri( struct draw_stage *stage, struct prim_header *header ) { @@ -162,7 +173,6 @@ struct draw_stage *draw_validate_stage( struct draw_context *draw ) struct draw_stage *stage = CALLOC_STRUCT(draw_stage); stage->draw = draw; - stage->next = NULL; stage->point = validate_point; stage->line = validate_line; diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index d827f51d56..cd0b4fbbb9 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -387,29 +387,26 @@ vbuf_alloc_vertices( struct draw_stage *stage, } -static void -vbuf_begin( struct draw_stage *stage ) -{ - /* no-op, vbuffer allocated by first point/line/tri */ -} - static void vbuf_flush( struct draw_stage *stage, unsigned flags ) { -// vbuf_flush_indices( stage ); - /* XXX: Overkill */ - vbuf_flush_vertices( stage ); - + vbuf_flush_indices( stage ); + stage->point = vbuf_first_point; stage->line = vbuf_first_line; stage->tri = vbuf_first_tri; + + if (flags & DRAW_FLUSH_BACKEND) + vbuf_flush_vertices( stage ); } static void vbuf_reset_stipple_counter( struct draw_stage *stage ) { + /* XXX: Need to do something here for hardware with linestipple. + */ (void) stage; } diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index 97a40b876e..b4b4906d70 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -42,10 +42,13 @@ void draw_vertex_cache_invalidate( struct draw_context *draw ) assert(draw->pq.queue_nr == 0); assert(draw->vs.queue_nr == 0); assert(draw->vcache.referenced == 0); - + /* XXX memset() here */ +#if 0 for (i = 0; i < Elements( draw->vcache.idx ); i++) draw->vcache.idx[i] = ~0; - +#else + memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); +#endif // fprintf(stderr, "x\n"); } @@ -148,7 +151,7 @@ void draw_vertex_cache_unreference( struct draw_context *draw ) int draw_vertex_cache_check_space( struct draw_context *draw, - unsigned nr_verts ) + unsigned nr_verts ) { if (draw->vcache.overflow + nr_verts < VCACHE_OVERFLOW) { /* The vs queue is sized so that this can never happen: diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 5ca659dbf5..d19b60198d 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -271,9 +271,9 @@ void draw_bind_vertex_shader(struct draw_context *draw, struct draw_vertex_shader *dvs) { - draw_flush(draw); - draw->vertex_shader = dvs; + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->vertex_shader = dvs; draw->num_vs_outputs = dvs->state->num_outputs; /* specify the fragment program to interpret/execute */ -- cgit v1.2.3 From 42d38ac02fff63406cd59914a3eb1e5ad27134d0 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 17:22:02 -0700 Subject: gallium: replace loop w/ memset in draw_vertex_cache_invalidate() --- src/mesa/pipe/draw/draw_vertex_cache.c | 10 +--------- 1 file changed, 1 insertion(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c index b4b4906d70..44427999cc 100644 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ b/src/mesa/pipe/draw/draw_vertex_cache.c @@ -37,19 +37,11 @@ void draw_vertex_cache_invalidate( struct draw_context *draw ) { - unsigned i; - assert(draw->pq.queue_nr == 0); assert(draw->vs.queue_nr == 0); assert(draw->vcache.referenced == 0); - /* XXX memset() here */ -#if 0 - for (i = 0; i < Elements( draw->vcache.idx ); i++) - draw->vcache.idx[i] = ~0; -#else + memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); -#endif -// fprintf(stderr, "x\n"); } -- cgit v1.2.3 From 4d534124e77109a0405bf68dc3fa1a370f5ea014 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 09:01:36 -0700 Subject: Cell: emit framebuffer state like other state Plus misc clean-up. Window resize sometimes works now. --- src/mesa/pipe/cell/common.h | 18 ++--- src/mesa/pipe/cell/ppu/cell_state_emit.c | 14 ++++ src/mesa/pipe/cell/ppu/cell_state_surface.c | 118 +--------------------------- src/mesa/pipe/cell/spu/spu_main.c | 13 +-- 4 files changed, 32 insertions(+), 131 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index af7f27bc83..f9edd0cc26 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -56,15 +56,15 @@ */ #define CELL_CMD_OPCODE_MASK 0xf -#define CELL_CMD_EXIT 1 -#define CELL_CMD_FRAMEBUFFER 2 -#define CELL_CMD_CLEAR_SURFACE 3 -#define CELL_CMD_FINISH 4 -#define CELL_CMD_RENDER 5 -#define CELL_CMD_BATCH 6 -#define CELL_CMD_STATE_DEPTH_STENCIL 7 -#define CELL_CMD_STATE_SAMPLER 8 -#define CELL_CMD_STATE_VERTEX_INFO 9 +#define CELL_CMD_EXIT 1 +#define CELL_CMD_CLEAR_SURFACE 2 +#define CELL_CMD_FINISH 3 +#define CELL_CMD_RENDER 4 +#define CELL_CMD_BATCH 5 +#define CELL_CMD_STATE_FRAMEBUFFER 10 +#define CELL_CMD_STATE_DEPTH_STENCIL 11 +#define CELL_CMD_STATE_SAMPLER 12 +#define CELL_CMD_STATE_VERTEX_INFO 13 #define CELL_NUM_BATCH_BUFFERS 3 diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c index e5b7c92514..dbca900c35 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.c +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.c @@ -36,6 +36,20 @@ void cell_emit_state(struct cell_context *cell) { + if (cell->dirty & CELL_NEW_FRAMEBUFFER) { + struct pipe_surface *cbuf = cell->framebuffer.cbufs[0]; + struct pipe_surface *zbuf = cell->framebuffer.zsbuf; + struct cell_command_framebuffer *fb + = cell_batch_alloc(cell, sizeof(*fb)); + fb->opcode = CELL_CMD_STATE_FRAMEBUFFER; + fb->color_start = cell->cbuf_map[0]; + fb->color_format = cbuf->format; + fb->depth_start = cell->zsbuf_map; + fb->depth_format = zbuf ? zbuf->format : PIPE_FORMAT_NONE; + fb->width = cell->framebuffer.cbufs[0]->width; + fb->height = cell->framebuffer.cbufs[0]->height; + } + if (cell->dirty & CELL_NEW_DEPTH_STENCIL) { uint cmd = CELL_CMD_STATE_DEPTH_STENCIL; cell_batch_append(cell, &cmd, 4); diff --git a/src/mesa/pipe/cell/ppu/cell_state_surface.c b/src/mesa/pipe/cell/ppu/cell_state_surface.c index 1f7faca080..287610b76b 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_state_surface.c @@ -27,10 +27,8 @@ #include "pipe/p_inlines.h" -#include "cell_batch.h" #include "cell_context.h" #include "cell_state.h" -#include "cell_spu.h" void @@ -39,14 +37,11 @@ cell_set_framebuffer_state(struct pipe_context *pipe, { struct cell_context *cell = cell_context(pipe); - /* XXX revisit this memcmp! */ - if (memcmp(&cell->framebuffer, fb, sizeof(*fb))) { + if (1 /*memcmp(&cell->framebuffer, fb, sizeof(*fb))*/) { struct pipe_surface *csurf = fb->cbufs[0]; struct pipe_surface *zsurf = fb->zsbuf; uint i; - /* change in fb state */ - /* unmap old surfaces */ for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { if (cell->framebuffer.cbufs[i] && cell->cbuf_map[i]) { @@ -70,118 +65,7 @@ cell_set_framebuffer_state(struct pipe_context *pipe, if (zsurf) cell->zsbuf_map = pipe_surface_map(zsurf); -#if 0 - for (i = 0; i < cell->num_spus; i++) { - struct cell_command_framebuffer *fb = &cell_global.command[i].fb; - fb->opcode = CELL_CMD_FRAMEBUFFER; - fb->color_start = csurf->map; - fb->color_format = csurf->format; - fb->depth_start = zsurf ? zsurf->map : NULL; - fb->depth_format = zsurf ? zsurf->format : PIPE_FORMAT_NONE; - fb->width = csurf->width; - fb->height = csurf->height; - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FRAMEBUFFER); - } -#endif -#if 1 - { - struct cell_command_framebuffer *fb - = cell_batch_alloc(cell, sizeof(*fb)); - fb->opcode = CELL_CMD_FRAMEBUFFER; - fb->color_start = cell->cbuf_map[0]; - fb->color_format = csurf->format; - fb->depth_start = cell->zsbuf_map; - fb->depth_format = zsurf ? zsurf->format : PIPE_FORMAT_NONE; - fb->width = csurf->width; - fb->height = csurf->height; - /*cell_batch_flush(cell);*/ - /*cell_flush(&cell->pipe, 0x0);*/ - } -#endif cell->dirty |= CELL_NEW_FRAMEBUFFER; } - -#if 0 - struct pipe_surface *ps; - uint i; - - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { - /* check if changing cbuf */ - if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) { - /* flush old */ - sp_flush_tile_cache(sp, sp->cbuf_cache[i]); - /* unmap old */ - ps = sp->framebuffer.cbufs[i]; - if (ps && ps->map) - pipe_surface_unmap(ps); - /* map new */ - ps = fb->cbufs[i]; - if (ps) - pipe_surface_map(ps); - /* assign new */ - sp->framebuffer.cbufs[i] = fb->cbufs[i]; - - /* update cache */ - sp_tile_cache_set_surface(sp, sp->cbuf_cache[i], ps); - } - } - - sp->framebuffer.num_cbufs = fb->num_cbufs; - - /* zbuf changing? */ - if (sp->framebuffer.zbuf != fb->zbuf) { - /* flush old */ - sp_flush_tile_cache(sp, sp->zbuf_cache); - /* unmap old */ - ps = sp->framebuffer.zbuf; - if (ps && ps->map) - pipe_surface_unmap(ps); - if (sp->framebuffer.sbuf == sp->framebuffer.zbuf) { - /* combined z/stencil */ - sp->framebuffer.sbuf = NULL; - } - /* map new */ - ps = fb->zbuf; - if (ps) - pipe_surface_map(ps); - /* assign new */ - sp->framebuffer.zbuf = fb->zbuf; - - /* update cache */ - sp_tile_cache_set_surface(sp, sp->zbuf_cache, ps); - } - - /* XXX combined depth/stencil here */ - - /* sbuf changing? */ - if (sp->framebuffer.sbuf != fb->sbuf) { - /* flush old */ - sp_flush_tile_cache(sp, sp->sbuf_cache_sep); - /* unmap old */ - ps = sp->framebuffer.sbuf; - if (ps && ps->map) - pipe_surface_unmap(ps); - /* map new */ - ps = fb->sbuf; - if (ps && fb->sbuf != fb->zbuf) - pipe_surface_map(ps); - /* assign new */ - sp->framebuffer.sbuf = fb->sbuf; - - /* update cache */ - if (fb->sbuf != fb->zbuf) { - /* separate stencil buf */ - sp->sbuf_cache = sp->sbuf_cache_sep; - sp_tile_cache_set_surface(sp, sp->sbuf_cache, ps); - } - else { - /* combined depth/stencil */ - sp->sbuf_cache = sp->zbuf_cache; - sp_tile_cache_set_surface(sp, sp->sbuf_cache, ps); - } - } - - sp->dirty |= SP_NEW_FRAMEBUFFER; -#endif } diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 2b88c54576..f9e15538b8 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -338,7 +338,7 @@ cmd_render(const struct cell_command_render *render) static void -cmd_framebuffer(const struct cell_command_framebuffer *cmd) +cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) { if (Debug) printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", @@ -349,6 +349,9 @@ cmd_framebuffer(const struct cell_command_framebuffer *cmd) cmd->color_format, cmd->depth_format); + ASSERT_ALIGN16(cmd->color_start); + ASSERT_ALIGN16(cmd->depth_start); + spu.fb.color_start = cmd->color_start; spu.fb.depth_start = cmd->depth_start; spu.fb.color_format = cmd->color_format; @@ -485,11 +488,11 @@ cmd_batch(uint opcode) for (pos = 0; pos < usize; /* no incr */) { switch (buffer[pos]) { - case CELL_CMD_FRAMEBUFFER: + case CELL_CMD_STATE_FRAMEBUFFER: { struct cell_command_framebuffer *fb = (struct cell_command_framebuffer *) &buffer[pos]; - cmd_framebuffer(fb); + cmd_state_framebuffer(fb); pos += sizeof(*fb) / 4; } break; @@ -581,8 +584,8 @@ main_loop(void) printf("SPU %u: EXIT\n", spu.init.id); exitFlag = 1; break; - case CELL_CMD_FRAMEBUFFER: - cmd_framebuffer(&cmd.fb); + case CELL_CMD_STATE_FRAMEBUFFER: + cmd_state_framebuffer(&cmd.fb); break; case CELL_CMD_CLEAR_SURFACE: cmd_clear_surface(&cmd.clear); -- cgit v1.2.3 From b5618330c64d8bdb260c258e0b47c8047ce41112 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 17:34:04 -0700 Subject: Cell: typedefs need semicolons, Ian :) --- src/mesa/pipe/cell/spu/spu_main.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index f9e15538b8..ef605a5197 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -624,9 +624,9 @@ one_time_init(void) * select between the two. */ #ifdef SPU_MAIN_PARAM_LONG_LONG -typedef unsigned long long main_param_t +typedef unsigned long long main_param_t; #else -typedef unsigned long main_param_t +typedef unsigned long main_param_t; #endif /** -- cgit v1.2.3 From 0b958e3b7c34ffdd63fb1cd1b4c9084640ed9a78 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Jan 2008 17:34:22 -0700 Subject: Cell: updates to match pipe_buffer changes --- src/mesa/pipe/cell/ppu/cell_draw_arrays.c | 6 +++--- src/mesa/pipe/cell/ppu/cell_state_fs.c | 3 ++- src/mesa/pipe/cell/ppu/cell_texture.c | 13 +++++-------- 3 files changed, 10 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c index 9c41b4a523..717cd8370f 100644 --- a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c +++ b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c @@ -51,7 +51,7 @@ cell_map_constant_buffers(struct cell_context *sp) for (i = 0; i < 2; i++) { if (sp->constants[i].size) sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); } draw_set_mapped_constant_buffer(sp->draw, @@ -123,7 +123,7 @@ cell_draw_elements(struct pipe_context *pipe, if (sp->vertex_buffer[i].buffer) { void *buf = pipe->winsys->buffer_map(pipe->winsys, sp->vertex_buffer[i].buffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } } @@ -131,7 +131,7 @@ cell_draw_elements(struct pipe_context *pipe, if (indexBuffer) { void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_FLAG_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { diff --git a/src/mesa/pipe/cell/ppu/cell_state_fs.c b/src/mesa/pipe/cell/ppu/cell_state_fs.c index 5a935f2de1..81c2ac14dd 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_fs.c +++ b/src/mesa/pipe/cell/ppu/cell_state_fs.c @@ -27,6 +27,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "pipe/draw/draw_context.h" #if 0 @@ -163,7 +164,7 @@ cell_set_constant_buffer(struct pipe_context *pipe, assert(index == 0); /* note: reference counting */ - ws->buffer_reference(ws, + pipe_buffer_reference(ws, &cell->constants[shader].buffer, buf->buffer); cell->constants[shader].size = buf->size; diff --git a/src/mesa/pipe/cell/ppu/cell_texture.c b/src/mesa/pipe/cell/ppu/cell_texture.c index 8f342d8aad..0a8190d983 100644 --- a/src/mesa/pipe/cell/ppu/cell_texture.c +++ b/src/mesa/pipe/cell/ppu/cell_texture.c @@ -91,12 +91,9 @@ cell_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) cell_texture_layout(spt); - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, 0, 0); - - if (spt->buffer) { - pipe->winsys->buffer_data(pipe->winsys, spt->buffer, spt->buffer_size, - NULL, PIPE_BUFFER_USAGE_PIXEL); - } + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); if (!spt->buffer) { FREE(spt); @@ -124,7 +121,7 @@ cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); */ - pipe->winsys->buffer_reference(pipe->winsys, &spt->buffer, NULL); + pipe_buffer_reference(pipe->winsys, &spt->buffer, NULL); FREE(spt); } @@ -147,7 +144,7 @@ cell_get_tex_surface(struct pipe_context *pipe, if (ps) { assert(ps->refcount); assert(ps->winsys); - pipe->winsys->buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); + pipe_buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; -- cgit v1.2.3 From a97b3f64b36aa8d5905a7ed8b7c9f00324661060 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Jan 2008 17:30:21 +0000 Subject: gallium: remove makefile whitespace --- src/mesa/pipe/pipebuffer/Makefile | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/Makefile b/src/mesa/pipe/pipebuffer/Makefile index 0d23e8b8f1..ea8c3440c3 100644 --- a/src/mesa/pipe/pipebuffer/Makefile +++ b/src/mesa/pipe/pipebuffer/Makefile @@ -4,7 +4,6 @@ include $(TOP)/configs/current LIBNAME = pipebuffer - DRIVER_SOURCES = \ pb_buffer.c \ pb_buffer_client.c \ -- cgit v1.2.3 From 596a92ee7590cd3819aad0139cf779d28e57874d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Jan 2008 20:22:49 +0000 Subject: gallium: specialize glFlush vs other flush semantics --- src/mesa/state_tracker/st_cb_flush.c | 44 ++++++++++++++++++++++++--------- src/mesa/state_tracker/st_framebuffer.c | 7 +++++- 2 files changed, 38 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 9808b1f8f6..dbec993f1b 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -45,21 +45,32 @@ void st_flush( struct st_context *st, uint pipeFlushFlags ) { - GLframebuffer *fb = st->ctx->DrawBuffer; - FLUSH_VERTICES(st->ctx, 0); - /* If there has been no rendering to the frontbuffer, consider - * short-circuiting this, or perhaps pass an "optional" flag down - * to the driver so that it can make the decision. - */ st->pipe->flush( st->pipe, pipeFlushFlags ); +} + + +static void st_gl_flush( struct st_context *st, uint pipeFlushFlags ) +{ + GLframebuffer *fb = st->ctx->DrawBuffer; + + FLUSH_VERTICES(st->ctx, 0); if (!fb) return; /* XXX: temporary hack. This flag should only be set if we do any * rendering to the front buffer. + * + * Further more, the scissor rectangle could be tracked to + * construct a dirty region of the front buffer, to avoid + * situations where it must be copied repeatedly. + * + * In the extreme case, some kind of timer could be set up to allow + * coalescing of multiple flushes to the frontbuffer, which can be + * quite a performance drain if there are a sufficient number of + * them. */ st->flags.frontbuffer_dirty = (fb->_ColorDrawBufferMask[0] & BUFFER_BIT_FRONT_LEFT); @@ -69,6 +80,15 @@ void st_flush( struct st_context *st, uint pipeFlushFlags ) = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); struct pipe_surface *front_surf = strb->surface; + /* If we aren't rendering to the frontbuffer, this is a noop. + * This should be uncontroversial for glFlush, though people may + * feel more strongly about glFinish. + * + * Additionally, need to make sure that the frontbuffer_dirty + * flag really gets set on frontbuffer rendering. + */ + st->pipe->flush( st->pipe, pipeFlushFlags ); + /* Hook for copying "fake" frontbuffer if necessary: */ st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf, @@ -81,23 +101,23 @@ void st_flush( struct st_context *st, uint pipeFlushFlags ) /** * Called via ctx->Driver.Flush() */ -static void st_Flush(GLcontext *ctx) +static void st_glFlush(GLcontext *ctx) { - st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE); + st_gl_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE); } /** * Called via ctx->Driver.Finish() */ -static void st_Finish(GLcontext *ctx) +static void st_glFinish(GLcontext *ctx) { - st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_WAIT); + st_gl_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_WAIT); } void st_init_flush_functions(struct dd_function_table *functions) { - functions->Flush = st_Flush; - functions->Finish = st_Finish; + functions->Flush = st_glFlush; + functions->Finish = st_glFinish; } diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 7ddc74e355..42e8e24896 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -34,6 +34,7 @@ #include "st_context.h" #include "st_cb_fbo.h" #include "pipe/p_defines.h" +#include "pipe/p_context.h" struct st_framebuffer * @@ -171,7 +172,11 @@ st_notify_swapbuffers(struct st_framebuffer *stfb) GET_CURRENT_CONTEXT(ctx); if (ctx && ctx->DrawBuffer == &stfb->Base) { - st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE); + st_flush( ctx->st, + PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_SWAPBUFFERS); + } +} + } } -- cgit v1.2.3 From 6d2d41294ce7de78d56051261bb4fc345517bf9b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Jan 2008 20:23:32 +0000 Subject: gallium: add notify_swapbuffers_complete, use it to set surfaces to undefined state --- src/mesa/state_tracker/st_framebuffer.c | 20 ++++++++++++++++++++ src/mesa/state_tracker/st_public.h | 1 + 2 files changed, 21 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 42e8e24896..bca3fa5c38 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -177,6 +177,26 @@ st_notify_swapbuffers(struct st_framebuffer *stfb) } } + +/** + * Quick hack - allows the winsys to inform the driver that surface + * states are now undefined after a glXSwapBuffers or similar. + */ +void +st_notify_swapbuffers_complete(struct st_framebuffer *stfb) +{ + GET_CURRENT_CONTEXT(ctx); + + if (ctx && ctx->DrawBuffer == &stfb->Base) { + struct st_renderbuffer *strb; + int i; + + for (i = 0; i < BUFFER_COUNT; i++) { + if (stfb->Base.Attachment[i].Renderbuffer) { + strb = st_renderbuffer(stfb->Base.Attachment[i].Renderbuffer); + strb->surface->status = PIPE_SURFACE_STATUS_UNDEFINED; + } + } } } diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 78a8fde82b..3c397b126a 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -81,5 +81,6 @@ void st_make_current(struct st_context *st, void st_flush( struct st_context *st, uint pipeFlushFlags ); void st_notify_swapbuffers(struct st_framebuffer *stfb); +void st_notify_swapbuffers_complete(struct st_framebuffer *stfb); #endif -- cgit v1.2.3 From 1e9e4341e08a45c93e7caa90c2cca844469f4629 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Jan 2008 20:25:19 +0000 Subject: gallium: add PIPE_FLUSH_SWAPBUFFERS flag --- src/mesa/pipe/p_defines.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 90eebd5e6c..85adf2d61d 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -198,6 +198,7 @@ enum pipe_texture_target { #define PIPE_FLUSH_RENDER_CACHE 0x1 #define PIPE_FLUSH_TEXTURE_CACHE 0x2 #define PIPE_FLUSH_WAIT 0x4 +#define PIPE_FLUSH_SWAPBUFFERS 0x8 /** -- cgit v1.2.3 From f3aa4de034b0d791ce2e38e8aeb3b3abdb4e3b50 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Jan 2008 14:38:36 +0000 Subject: gallium: minor cleanups to pipe interface - Remove put/get tile, just have users call put_tile_raw, etc directly. - Remove surface_data call, just map it locally. --- src/mesa/pipe/failover/fo_context.c | 1 - src/mesa/pipe/i915simple/i915_surface.c | 24 ---------------------- src/mesa/pipe/i965simple/brw_surface.c | 4 ---- src/mesa/pipe/p_context.h | 36 +++++++++------------------------ src/mesa/pipe/softpipe/sp_surface.c | 25 ----------------------- src/mesa/pipe/softpipe/sp_tile_cache.c | 8 ++++---- src/mesa/pipe/util/p_tile.c | 4 ++-- src/mesa/state_tracker/st_texture.c | 35 ++++++++++++++++++++++++++------ 8 files changed, 44 insertions(+), 93 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index 3799c4dfb0..cf6c9fed50 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -140,7 +140,6 @@ struct pipe_context *failover_create( struct pipe_context *hw, #endif failover->pipe.get_tex_surface = hw->get_tex_surface; - failover->pipe.surface_data = hw->surface_data; failover->pipe.surface_copy = hw->surface_copy; failover->pipe.surface_fill = hw->surface_fill; failover->pipe.texture_create = hw->texture_create; diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 8ef02b99cd..1bdaba773f 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -77,27 +77,6 @@ i915_get_tex_surface(struct pipe_context *pipe, } -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -static void -i915_surface_data(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - const void *src, unsigned src_pitch, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - pipe_copy_rect(pipe_surface_map(dst), - dst->cpp, dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - pipe_surface_unmap(dst); -} - /* Assumes all values are within bounds -- no checking at this level - * do it higher up if required. @@ -204,10 +183,7 @@ void i915_init_surface_functions(struct i915_context *i915) { i915->pipe.get_tex_surface = i915_get_tex_surface; - i915->pipe.get_tile = pipe_get_tile_raw; - i915->pipe.put_tile = pipe_put_tile_raw; - i915->pipe.surface_data = i915_surface_data; i915->pipe.surface_copy = i915_surface_copy; i915->pipe.surface_fill = i915_surface_fill; } diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c index 252aec7d29..eb7835836e 100644 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -203,10 +203,6 @@ void brw_init_surface_functions(struct brw_context *brw) { brw->pipe.get_tex_surface = brw_get_tex_surface; - brw->pipe.get_tile = pipe_get_tile_raw; - brw->pipe.put_tile = pipe_put_tile_raw; - - brw->pipe.surface_data = brw_surface_data; brw->pipe.surface_copy = brw_surface_copy; brw->pipe.surface_fill = brw_surface_fill; } diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 37464c88a1..0dda06c53b 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -78,9 +78,6 @@ struct pipe_context { unsigned indexSize, unsigned mode, unsigned start, unsigned count); - /** Clear a surface to given value (no scissor; clear whole surface) */ - void (*clear)(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue); /** * Query objects @@ -176,33 +173,9 @@ struct pipe_context { const struct pipe_vertex_element * ); - /** Get a surface which is a "view" into a texture */ - struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, - struct pipe_texture *texture, - unsigned face, unsigned level, - unsigned zslice); - - /** Get a block of raw pixel data from a surface */ - void (*get_tile)(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride); - /** Put a block of raw pixel data into a surface */ - void (*put_tile)(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride); - - /* * Surface functions */ - void (*surface_data)(struct pipe_context *pipe, - struct pipe_surface *dest, - unsigned destx, unsigned desty, - const void *src, unsigned src_stride, - unsigned srcx, unsigned srcy, - unsigned width, unsigned height); void (*surface_copy)(struct pipe_context *pipe, struct pipe_surface *dest, @@ -218,6 +191,10 @@ struct pipe_context { unsigned width, unsigned height, unsigned value); + void (*clear)(struct pipe_context *pipe, + struct pipe_surface *ps, + unsigned clearValue); + /* * Texture functions @@ -228,6 +205,11 @@ struct pipe_context { void (*texture_release)(struct pipe_context *pipe, struct pipe_texture **pt); + /** Get a surface which is a "view" into a texture */ + struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level, + unsigned zslice); /* Flush rendering: */ diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index e115705507..5978ee48bd 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -34,27 +34,6 @@ #include "sp_surface.h" -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -static void -sp_surface_data(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - const void *src, unsigned src_pitch, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - pipe_copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - pipe_surface_unmap(dst); -} /* Assumes all values are within bounds -- no checking at this level - * do it higher up if required. @@ -174,10 +153,6 @@ sp_surface_fill(struct pipe_context *pipe, void sp_init_surface_functions(struct softpipe_context *sp) { - sp->pipe.get_tile = pipe_get_tile_raw; - sp->pipe.put_tile = pipe_put_tile_raw; - - sp->pipe.surface_data = sp_surface_data; sp->pipe.surface_copy = sp_surface_copy; sp->pipe.surface_fill = sp_surface_fill; } diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index bc7e8d01e4..451e157abf 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -329,7 +329,7 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, for (y = 0; y < h; y += TILE_SIZE) { for (x = 0; x < w; x += TILE_SIZE) { if (is_clear_flag_set(tc->clear_flags, x, y)) { - pipe->put_tile(pipe, ps, + pipe_put_tile_raw(pipe, ps, x, y, TILE_SIZE, TILE_SIZE, tc->tile.data.color32, 0/*STRIDE*/); @@ -365,7 +365,7 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_cached_tile *tile = tc->entries + pos; if (tile->x >= 0) { if (tc->depth_stencil) { - pipe->put_tile(pipe, ps, + pipe_put_tile_raw(pipe, ps, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } @@ -414,7 +414,7 @@ sp_get_cached_tile(struct softpipe_context *softpipe, if (tile->x != -1) { /* put dirty tile back in framebuffer */ if (tc->depth_stencil) { - pipe->put_tile(pipe, ps, + pipe_put_tile_raw(pipe, ps, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } @@ -441,7 +441,7 @@ sp_get_cached_tile(struct softpipe_context *softpipe, else { /* get new tile data from surface */ if (tc->depth_stencil) { - pipe->get_tile(pipe, ps, + pipe_put_tile_raw(pipe, ps, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } diff --git a/src/mesa/pipe/util/p_tile.c b/src/mesa/pipe/util/p_tile.c index 85a863db8a..3f795a3898 100644 --- a/src/mesa/pipe/util/p_tile.c +++ b/src/mesa/pipe/util/p_tile.c @@ -577,7 +577,7 @@ pipe_get_tile_rgba(struct pipe_context *pipe, if (!packed) return; - pipe->get_tile(pipe, ps, x, y, w, h, packed, w * ps->cpp); + pipe_get_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); switch (ps->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: @@ -693,7 +693,7 @@ pipe_put_tile_rgba(struct pipe_context *pipe, assert(0); } - pipe->put_tile(pipe, ps, x, y, w, h, packed, w * ps->cpp); + pipe_put_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); FREE(packed); } diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index a2bdf846ca..15cc458be8 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -201,6 +201,29 @@ st_texture_image_unmap(struct st_texture_image *stImage) +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +static void +st_surface_data(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + const void *src, unsigned src_pitch, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + pipe_surface_unmap(dst); +} + + /* Upload data for a particular image. */ void @@ -225,12 +248,12 @@ st_texture_image_data(struct pipe_context *pipe, dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); - pipe->surface_data(pipe, dst_surface, - 0, 0, /* dstx, dsty */ - srcUB, - src_row_pitch, - 0, 0, /* source x, y */ - dst->width[level], height); /* width, height */ + st_surface_data(pipe, dst_surface, + 0, 0, /* dstx, dsty */ + srcUB, + src_row_pitch, + 0, 0, /* source x, y */ + dst->width[level], height); /* width, height */ pipe_surface_reference(&dst_surface, NULL); -- cgit v1.2.3 From 2bca9f6457d8f766ef12891657fc2d10ea72ef54 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 24 Jan 2008 19:37:37 +0900 Subject: gallium: Complete integer types. --- src/mesa/pipe/p_compiler.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h index ab9609deab..e939d9cd9b 100644 --- a/src/mesa/pipe/p_compiler.h +++ b/src/mesa/pipe/p_compiler.h @@ -52,6 +52,9 @@ typedef unsigned long long uint64; #if defined(__MSC__) +typedef char int8_t; +typedef unsigned char uint8_t; +typedef short int16_t; typedef unsigned short uint16_t; typedef long int32_t; typedef unsigned long uint32_t; -- cgit v1.2.3 From c3e4e9260d7527bb0369650b861cba43834f7106 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 24 Jan 2008 19:40:14 +0900 Subject: gallium: Fix MSVC float/double conversion warning. --- src/mesa/pipe/draw/draw_wide_prims.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 26f5d10dc0..9759e7e2e8 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -217,7 +217,7 @@ static void wide_point( struct draw_stage *stage, /* point size is either per-vertex or fixed size */ if (wide->psize_slot >= 0) { - half_size = 0.5 * header->v[0]->data[wide->psize_slot][0]; + half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; } else { half_size = wide->half_point_size; -- cgit v1.2.3 From 027983f5850afea753381be454122166c6d56777 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 24 Jan 2008 11:19:06 +0000 Subject: gallium: restructure vertex fetch code slightly --- src/mesa/pipe/draw/draw_private.h | 17 +++ src/mesa/pipe/draw/draw_vertex_fetch.c | 193 +++++++++++++++++++++----------- src/mesa/pipe/draw/draw_vertex_shader.c | 4 + 3 files changed, 149 insertions(+), 65 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 1c2e88264f..1e59f5bd8d 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -137,6 +137,13 @@ struct draw_vertex_shader { #endif }; + +/* Internal function for vertex fetch. + */ +typedef void (*fetch_func)(const void *ptr, float *attrib); + + + /** * Private context for the drawing module. */ @@ -196,6 +203,15 @@ struct draw_context /** TGSI program interpreter runtime state */ struct tgsi_exec_machine machine; + /* Vertex fetch internal state + */ + struct { + const ubyte *src_ptr[PIPE_ATTRIB_MAX]; + unsigned pitch[PIPE_ATTRIB_MAX]; + fetch_func fetch[PIPE_ATTRIB_MAX]; + unsigned nr_attrs; + } vertex_fetch; + /* Post-tnl vertex cache: */ struct { @@ -270,6 +286,7 @@ extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw ); struct tgsi_exec_machine; +extern void draw_update_vertex_fetch( struct draw_context *draw ); extern void draw_vertex_fetch( struct draw_context *draw, struct tgsi_exec_machine *machine, const unsigned *elts, diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 4d64d3d4f2..23e8187899 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -42,53 +42,108 @@ /** * Fetch a float[4] vertex attribute from memory, doing format/type * conversion as needed. - * XXX this might be a temporary thing. + * + * This is probably needed/dupliocated elsewhere, eg format + * conversion, texture sampling etc. */ -static void -fetch_attrib4(const void *ptr, enum pipe_format format, float attrib[4]) +#define FETCH_ATTRIB( NAME, SZ, CVT ) \ +static void \ +fetch_##NAME(const void *ptr, float *attrib) \ +{ \ + static const float defaults[4] = { 0,0,0,1 }; \ + int i; \ + \ + for (i = 0; i < SZ; i++) { \ + attrib[i] = CVT; \ + } \ + \ + for (; i < 4; i++) { \ + attrib[i] = defaults[i]; \ + } \ +} + +#define CVT_32_FLOAT ((float *) ptr)[i] +#define CVT_32_SSCALED (float) ((int *) ptr)[i] +#define CVT_8_UNORM (float) ((unsigned char *) ptr)[i] / 255.0f + +FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) +FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) +FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) +FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) + + + +static fetch_func get_fetch_func( unsigned format ) { - /* defaults */ - attrib[1] = 0.0; - attrib[2] = 0.0; - attrib[3] = 1.0; switch (format) { case PIPE_FORMAT_R32G32B32A32_FLOAT: - attrib[3] = ((float *) ptr)[3]; - /* fall-through */ + return fetch_R32G32B32A32_FLOAT; case PIPE_FORMAT_R32G32B32_FLOAT: - attrib[2] = ((float *) ptr)[2]; - /* fall-through */ + return fetch_R32G32B32_FLOAT; case PIPE_FORMAT_R32G32_FLOAT: - attrib[1] = ((float *) ptr)[1]; - /* fall-through */ + return fetch_R32G32_FLOAT; case PIPE_FORMAT_R32_FLOAT: - attrib[0] = ((float *) ptr)[0]; - break; - + return fetch_R32_FLOAT; case PIPE_FORMAT_R32G32B32A32_SSCALED: - attrib[3] = (float) ((int *) ptr)[3]; - /* fall-through */ + return fetch_R32G32B32A32_SSCALED; case PIPE_FORMAT_R32G32B32_SSCALED: - attrib[2] = (float) ((int *) ptr)[2]; - /* fall-through */ + return fetch_R32G32B32_SSCALED; case PIPE_FORMAT_R32G32_SSCALED: - attrib[1] = (float) ((int *) ptr)[1]; - /* fall-through */ + return fetch_R32G32_SSCALED; case PIPE_FORMAT_R32_SSCALED: - attrib[0] = (float) ((int *) ptr)[0]; - break; - + return fetch_R32_SSCALED; case PIPE_FORMAT_A8R8G8B8_UNORM: + return fetch_A8R8G8B8_UNORM; case PIPE_FORMAT_R8G8B8A8_UNORM: - attrib[0] = (float) ((unsigned char *) ptr)[2] / 255.0f; - attrib[1] = (float) ((unsigned char *) ptr)[1] / 255.0f; - attrib[2] = (float) ((unsigned char *) ptr)[0] / 255.0f; - attrib[3] = (float) ((unsigned char *) ptr)[3] / 255.0f; - break; - + return fetch_R8G8B8A8_UNORM; default: + /* Lots of missing cases! */ assert(0); + return NULL; + } +} + + +static void +transpose_4x4( float *out, const float *in ) +{ + /* This can be achieved in 12 sse instructions, plus the final + * stores I guess. This is probably a bit more than that - maybe + * 32 or so? + */ + out[0] = in[0]; out[1] = in[4]; out[2] = in[8]; out[3] = in[12]; + out[4] = in[1]; out[5] = in[5]; out[6] = in[9]; out[7] = in[13]; + out[8] = in[2]; out[9] = in[6]; out[10] = in[10]; out[11] = in[14]; + out[12] = in[3]; out[13] = in[7]; out[14] = in[11]; out[15] = in[15]; +} + + + +void draw_update_vertex_fetch( struct draw_context *draw ) +{ + //unsigned nr_attrs = draw->vertex_element_count; + unsigned nr_attrs = draw->vertex_shader->state->num_inputs; + unsigned i; + + for (i = 0; i < nr_attrs; i++) { + unsigned buf = draw->vertex_element[i].vertex_buffer_index; + unsigned format = draw->vertex_element[i].src_format; + + draw->vertex_fetch.src_ptr[i] = (const ubyte *) (draw->user.vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[i].src_offset ); + + draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; + draw->vertex_fetch.fetch[i] = get_fetch_func( format ); } + + draw->vertex_fetch.nr_attrs = nr_attrs; } @@ -100,40 +155,48 @@ void draw_vertex_fetch( struct draw_context *draw, const unsigned *elts, unsigned count ) { - unsigned j; - - /* loop over vertices */ - for (j = 0; j < count; j++) { - uint attr; - -#if DRAW_DBG - printf("fetch vertex %u: \n", j); -#endif - - /* loop over vertex attributes (vertex shader inputs) */ - for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { - - unsigned buf = draw->vertex_element[attr].vertex_buffer_index; - const void *src - = (const void *) ((const ubyte *) draw->user.vbuffer[buf] - + draw->vertex_buffer[buf].buffer_offset - + draw->vertex_element[attr].src_offset - + elts[j] * draw->vertex_buffer[buf].pitch); - float p[4]; - - fetch_attrib4(src, draw->vertex_element[attr].src_format, p); - -#if DRAW_DBG - printf(" %u: %f %f %f %f\n", attr, p[0], p[1], p[2], p[3]); -#endif - - /* Transform to AoS xxxx/yyyy/zzzz/wwww representation: - */ - machine->Inputs[attr].xyzw[0].f[j] = p[0]; /*X*/ - machine->Inputs[attr].xyzw[1].f[j] = p[1]; /*Y*/ - machine->Inputs[attr].xyzw[2].f[j] = p[2]; /*Z*/ - machine->Inputs[attr].xyzw[3].f[j] = p[3]; /*W*/ - } + unsigned nr_attrs = draw->vertex_fetch.nr_attrs; + unsigned attr; + + assert(count <= 4); + +// _mesa_printf("%s %d\n", __FUNCTION__, count); + + /* loop over vertex attributes (vertex shader inputs) + */ + for (attr = 0; attr < nr_attrs; attr++) { + + const unsigned pitch = draw->vertex_fetch.pitch[attr]; + const ubyte *src = draw->vertex_fetch.src_ptr[attr]; + const fetch_func fetch = draw->vertex_fetch.fetch[attr]; + unsigned i; + float p[4][4]; + + + /* Fetch four attributes for four vertices. + * + * Could fetch directly into AOS format, but this is meant to be + * a prototype for an sse implementation, which would have + * difficulties doing that. + */ + for (i = 0; i < count; i++) + fetch( src + elts[i] * pitch, p[i] ); + + /* Be nice and zero out any missing vertices: + */ + for ( ; i < 4; i++) + p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; + + /* Transpose/swizzle into sse-friendly format. Currently + * assuming that all vertex shader inputs are float[4], but this + * isn't true -- if the vertex shader only wants tex0.xy, we + * could optimize for that. + * + * To do so fully without codegen would probably require an + * excessive number of fetch functions, but we could at least + * minimize the transpose step: + */ + transpose_4x4( (float *)&machine->Inputs[attr].xyzw[0].f[0], (float *)p ); } } diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index d19b60198d..3041974b9a 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -201,6 +201,10 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) { unsigned i, j; + /* XXX: do this on statechange: + */ + draw_update_vertex_fetch( draw ); + // fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); #ifdef MESA_LLVM if (draw->vertex_shader->llvm_prog) { -- cgit v1.2.3 From 337ebdd8f987a76579498d84bdea76a3ee819fcf Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 24 Jan 2008 16:41:29 +0100 Subject: gallium: Fix build on Windows. --- src/mesa/pipe/draw/draw_vertex_fetch.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 23e8187899..1a9e38158d 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -135,9 +135,9 @@ void draw_update_vertex_fetch( struct draw_context *draw ) unsigned buf = draw->vertex_element[i].vertex_buffer_index; unsigned format = draw->vertex_element[i].src_format; - draw->vertex_fetch.src_ptr[i] = (const ubyte *) (draw->user.vbuffer[buf] + + draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + draw->vertex_buffer[buf].buffer_offset + - draw->vertex_element[i].src_offset ); + draw->vertex_element[i].src_offset; draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; draw->vertex_fetch.fetch[i] = get_fetch_func( format ); -- cgit v1.2.3 From 9cbff50cb8d8c727bb0da1c60813fffe66f390e7 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 24 Jan 2008 18:15:30 +0100 Subject: gallium: Fix build on Windows. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 088d8a4c07..b17801d13d 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -551,7 +551,7 @@ static void setup_tri_coefficients( struct setup_stage *setup ) if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; setup->coef[fragSlot].dadx[1] = 0.0; setup->coef[fragSlot].dady[1] = 0.0; } @@ -806,7 +806,7 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; setup->coef[fragSlot].dadx[1] = 0.0; setup->coef[fragSlot].dady[1] = 0.0; } @@ -1029,7 +1029,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1 - setup->quad.facing; + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; setup->coef[fragSlot].dadx[1] = 0.0; setup->coef[fragSlot].dady[1] = 0.0; } -- cgit v1.2.3 From 7411a84c2bd6f4dd4f027c6c6a647822cbb4c061 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 25 Jan 2008 19:33:58 +0900 Subject: gallium: Define MALLOC_STRUCT. --- src/mesa/pipe/p_util.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index a2bc330424..059528787d 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -108,6 +108,8 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #endif /* WIN32 */ +#define MALLOC_STRUCT(T) (struct T *) MALLOC(sizeof(struct T)) + #define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) -- cgit v1.2.3 From 6c3d12e374939d36a23af7c9c502de18ce02f423 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jan 2008 19:36:23 +0100 Subject: gallium: Fix build for WinXP. --- src/mesa/pipe/p_thread.h | 54 +++++++++++++++++++++++++++++ src/mesa/pipe/pipebuffer/pb_buffer.h | 13 +++---- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 15 +++++--- src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c | 2 +- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 16 +++++---- 5 files changed, 82 insertions(+), 18 deletions(-) create mode 100644 src/mesa/pipe/p_thread.h (limited to 'src') diff --git a/src/mesa/pipe/p_thread.h b/src/mesa/pipe/p_thread.h new file mode 100644 index 0000000000..cd432c547c --- /dev/null +++ b/src/mesa/pipe/p_thread.h @@ -0,0 +1,54 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_THREAD_H +#define P_THREAD_H + +#include "p_compiler.h" + +/* + * XXX: We should come up with a system-independent thread definitions. + * XXX: Patching glthread defs for now. + */ + +#ifndef __MSC__ + +#include "glapi/glthread.h" + +#else /* __MSC__ */ + +typedef int _glthread_Mutex; + +#define _glthread_INIT_MUTEX( M ) ((void) (M)) +#define _glthread_LOCK_MUTEX( M ) ((void) (M)) +#define _glthread_UNLOCK_MUTEX( M ) ((void) (M)) + +#define sched_yield() ((void) 0) + +#endif /* __MSC__ */ + +#endif /* P_THREAD_H */ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.h b/src/mesa/pipe/pipebuffer/pb_buffer.h index f909bded0d..b8cb731338 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer.h @@ -46,6 +46,7 @@ #include #include +#include "p_compiler.h" #include "pipe/p_state.h" @@ -104,7 +105,7 @@ struct pb_vtbl /* Accessor functions for pb->vtbl: */ -static inline void * +static INLINE void * pb_map(struct pb_buffer *buf, unsigned flags) { @@ -113,7 +114,7 @@ pb_map(struct pb_buffer *buf, } -static inline void +static INLINE void pb_unmap(struct pb_buffer *buf) { assert(buf); @@ -121,7 +122,7 @@ pb_unmap(struct pb_buffer *buf) } -static inline void +static INLINE void pb_get_base_buffer( struct pb_buffer *buf, struct pb_buffer **base_buf, unsigned *offset ) @@ -129,7 +130,7 @@ pb_get_base_buffer( struct pb_buffer *buf, buf->vtbl->get_base_buffer(buf, base_buf, offset); } -static inline void +static INLINE void pb_destroy(struct pb_buffer *buf) { assert(buf); @@ -163,13 +164,13 @@ pb_malloc_buffer_create( unsigned alignment, -static inline struct pipe_buffer * +static INLINE struct pipe_buffer * pb_pipe_buffer( struct pb_buffer *pbuf ) { return &pbuf->base; } -static inline struct pb_buffer * +static INLINE struct pb_buffer * pb_buffer( struct pipe_buffer *buf ) { /* Could add a magic cookie check on debug builds. diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c index 625120d714..ad2da71c4a 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -36,17 +36,21 @@ #include #include -#include -#include "main/imports.h" -#include "glapi/glthread.h" #include "linked_list.h" +#include "p_compiler.h" #include "p_winsys.h" +#include "p_thread.h" #include "pb_buffer.h" #include "pb_buffer_fenced.h" +#ifndef __MSC__ +#include +#include "main/imports.h" +#endif + /** * Convenience macro (type safe). @@ -103,11 +107,12 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, struct fenced_buffer *fenced_buf; struct list_head *list, *prev; int signaled = -1; - int i; list = fenced_list->delayed.next; if (fenced_list->numDelayed > 3) { + unsigned i; + for (i = 0; i < fenced_list->numDelayed; i += 3) { list = list->next; } @@ -281,7 +286,7 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) while (fenced_list->numDelayed) { _glthread_UNLOCK_MUTEX(fenced_list->mutex); sched_yield(); - _fenced_buffer_list_check_free(fenced_list, GL_TRUE); + _fenced_buffer_list_check_free(fenced_list, 1); _glthread_LOCK_MUTEX(fenced_list->mutex); } diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c index 8cdbbed272..28996b9a17 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c @@ -52,7 +52,7 @@ struct fenced_pb_manager }; -static inline struct fenced_pb_manager * +static INLINE struct fenced_pb_manager * fenced_pb_manager(struct pb_manager *mgr) { assert(mgr); diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index 7e8494f615..1e213074f1 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -37,17 +37,21 @@ #include #include -#include -#include "main/imports.h" -#include "glapi/glthread.h" #include "linked_list.h" +#include "p_compiler.h" +#include "p_thread.h" #include "p_defines.h" #include "pb_buffer.h" #include "pb_bufmgr.h" +#ifndef __MSC__ +#include +#include "main/imports.h" +#endif + /** * Convenience macro (type safe). @@ -75,7 +79,7 @@ struct pool_pb_manager }; -static inline struct pool_pb_manager * +static INLINE struct pool_pb_manager * pool_pb_manager(struct pb_manager *mgr) { assert(mgr); @@ -95,7 +99,7 @@ struct pool_buffer }; -static inline struct pool_buffer * +static INLINE struct pool_buffer * pool_buffer(struct pb_buffer *buf) { assert(buf); @@ -217,7 +221,7 @@ pool_bufmgr_create(struct pb_manager *provider, { struct pool_pb_manager *pool; struct pool_buffer *pool_buf; - int i; + size_t i; pool = (struct pool_pb_manager *)calloc(1, sizeof(*pool)); if (!pool) -- cgit v1.2.3 From 1d6c98651f73b2f110d26c906c6626d5b750d38b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 22 Jan 2008 00:13:50 +0900 Subject: gallium: Use directory name in include. --- src/mesa/pipe/pipebuffer/pb_buffer.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.h b/src/mesa/pipe/pipebuffer/pb_buffer.h index b8cb731338..570986314a 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer.h @@ -46,7 +46,8 @@ #include #include -#include "p_compiler.h" + +#include "pipe/p_compiler.h" #include "pipe/p_state.h" -- cgit v1.2.3 From b8dc2956df0f534924fc45287cccc04499e70959 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jan 2008 21:33:35 +0100 Subject: gallium: Fix build on WinXP. --- src/mesa/pipe/pipebuffer/pb_buffer_client.c | 3 ++- src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 6 +++--- 2 files changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_client.c b/src/mesa/pipe/pipebuffer/pb_buffer_client.c index 6bf4745301..477ecfe32c 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_client.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_client.c @@ -34,6 +34,7 @@ */ +#include "pipe/p_util.h" #include "pb_buffer.h" #include "pipe/p_util.h" @@ -61,7 +62,7 @@ static void pb_user_buffer_destroy(struct pb_buffer *buf) { assert(buf); - free(buf); + FREE(buf); } diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c index e02eb1eebd..42aae7acfa 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c @@ -62,7 +62,7 @@ static void malloc_buffer_destroy(struct pb_buffer *buf) { free(malloc_buffer(buf)->data); - free(buf); + FREE(buf); } @@ -110,7 +110,7 @@ pb_malloc_buffer_create( unsigned alignment, /* TODO: accept an alignment parameter */ /* TODO: do a single allocation */ - buf = (struct malloc_buffer *)malloc(sizeof(struct malloc_buffer)); + buf = (struct malloc_buffer *)MALLOC(sizeof(struct malloc_buffer)); if(!buf) return NULL; @@ -121,7 +121,7 @@ pb_malloc_buffer_create( unsigned alignment, buf->data = malloc(size); if(!buf->data) { - free(buf); + FREE(buf); return NULL; } -- cgit v1.2.3 From 0fe428f1dfafe6bfb8896e0a08cbe12d569497d7 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 23 Jan 2008 15:47:10 +0100 Subject: d3d: Fix build on Windows. --- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 5 +++-- src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c | 8 +++++--- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 13 +++++++------ 3 files changed, 15 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c index ad2da71c4a..5b33a42e19 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -42,6 +42,7 @@ #include "p_compiler.h" #include "p_winsys.h" #include "p_thread.h" +#include "p_util.h" #include "pb_buffer.h" #include "pb_buffer_fenced.h" @@ -248,7 +249,7 @@ fenced_buffer_list_create(struct pipe_winsys *winsys) { struct fenced_buffer_list *fenced_list; - fenced_list = (struct fenced_buffer_list *)calloc(1, sizeof(*fenced_list)); + fenced_list = (struct fenced_buffer_list *)CALLOC(1, sizeof(*fenced_list)); if (!fenced_list) return NULL; @@ -292,7 +293,7 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) _glthread_UNLOCK_MUTEX(fenced_list->mutex); - free(fenced_list); + FREE(fenced_list); } diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c index 28996b9a17..734d5bdceb 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c @@ -37,6 +37,8 @@ #include #include +#include "p_util.h" + #include "pb_buffer.h" #include "pb_buffer_fenced.h" #include "pb_bufmgr.h" @@ -100,7 +102,7 @@ fenced_bufmgr_destroy(struct pb_manager *mgr) fenced_mgr->provider->destroy(fenced_mgr->provider); - free(fenced_mgr); + FREE(fenced_mgr); } @@ -110,7 +112,7 @@ fenced_bufmgr_create(struct pb_manager *provider, { struct fenced_pb_manager *fenced_mgr; - fenced_mgr = (struct fenced_pb_manager *)calloc(1, sizeof(*fenced_mgr)); + fenced_mgr = (struct fenced_pb_manager *)CALLOC(1, sizeof(*fenced_mgr)); if (!fenced_mgr) return NULL; @@ -120,7 +122,7 @@ fenced_bufmgr_create(struct pb_manager *provider, fenced_mgr->provider = provider; fenced_mgr->fenced_list = fenced_buffer_list_create(winsys); if(!fenced_mgr->fenced_list) { - free(fenced_mgr); + FREE(fenced_mgr); return NULL; } diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index 1e213074f1..9d7dbd5eb9 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -43,6 +43,7 @@ #include "p_compiler.h" #include "p_thread.h" #include "p_defines.h" +#include "p_util.h" #include "pb_buffer.h" #include "pb_bufmgr.h" @@ -203,14 +204,14 @@ pool_bufmgr_destroy(struct pb_manager *mgr) struct pool_pb_manager *pool = pool_pb_manager(mgr); _glthread_LOCK_MUTEX(pool->mutex); - free(pool->bufs); + FREE(pool->bufs); pb_unmap(pool->buffer); pb_destroy(pool->buffer); _glthread_UNLOCK_MUTEX(pool->mutex); - free(mgr); + FREE(mgr); } @@ -223,7 +224,7 @@ pool_bufmgr_create(struct pb_manager *provider, struct pool_buffer *pool_buf; size_t i; - pool = (struct pool_pb_manager *)calloc(1, sizeof(*pool)); + pool = (struct pool_pb_manager *)CALLOC(1, sizeof(*pool)); if (!pool) return NULL; @@ -248,7 +249,7 @@ pool_bufmgr_create(struct pb_manager *provider, if(!pool->map) goto failure; - pool->bufs = (struct pool_buffer *) malloc(numBufs * sizeof(*pool->bufs)); + pool->bufs = (struct pool_buffer *) MALLOC(numBufs * sizeof(*pool->bufs)); if (!pool->bufs) goto failure; @@ -265,12 +266,12 @@ pool_bufmgr_create(struct pb_manager *provider, failure: if(pool->bufs) - free(pool->bufs); + FREE(pool->bufs); if(pool->map) pb_unmap(pool->buffer); if(pool->buffer) pb_destroy(pool->buffer); if(pool) - free(pool); + FREE(pool); return NULL; } -- cgit v1.2.3 From 9387784fbf22495898abe3dc41cfcd42397792cd Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 23 Jan 2008 15:47:10 +0100 Subject: gallium: Fix build on Windows. --- src/mesa/pipe/pipebuffer/pb_buffer_client.c | 1 - src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 5 +++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_client.c b/src/mesa/pipe/pipebuffer/pb_buffer_client.c index 477ecfe32c..c316aabd32 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_client.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_client.c @@ -36,7 +36,6 @@ #include "pipe/p_util.h" #include "pb_buffer.h" -#include "pipe/p_util.h" struct pb_user_buffer diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c index 42aae7acfa..b80ee8dda3 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c @@ -37,6 +37,7 @@ #include #include +#include "pipe/p_util.h" #include "pb_buffer.h" @@ -61,7 +62,7 @@ malloc_buffer(struct pb_buffer *buf) static void malloc_buffer_destroy(struct pb_buffer *buf) { - free(malloc_buffer(buf)->data); + FREE(malloc_buffer(buf)->data); FREE(buf); } @@ -119,7 +120,7 @@ pb_malloc_buffer_create( unsigned alignment, buf->base.base.usage = usage; buf->base.base.size = size; - buf->data = malloc(size); + buf->data = MALLOC(size); if(!buf->data) { FREE(buf); return NULL; -- cgit v1.2.3 From 8ee1452e184e030eb7cc865a7828c8a5ea08d6f7 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 24 Jan 2008 11:09:07 +0000 Subject: gallium: fix linux build --- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 1 - src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 1 - 2 files changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c index 5b33a42e19..b2edc321ef 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -49,7 +49,6 @@ #ifndef __MSC__ #include -#include "main/imports.h" #endif diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index 9d7dbd5eb9..e462462610 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -50,7 +50,6 @@ #ifndef __MSC__ #include -#include "main/imports.h" #endif -- cgit v1.2.3 From 0afe449bfc568c4bc6fa69afa5de6edae93b27a9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 25 Jan 2008 18:37:41 +0900 Subject: gallium: Remove unneeded headers. --- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index e462462610..33b2f628dd 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -48,10 +48,6 @@ #include "pb_buffer.h" #include "pb_bufmgr.h" -#ifndef __MSC__ -#include -#endif - /** * Convenience macro (type safe). -- cgit v1.2.3 From ca01ed45e8dee7970c6eefb528b7576ce6717459 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 25 Jan 2008 18:38:25 +0900 Subject: gallium: Make mm buffer manager more portable. --- src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 320 ++++++++++++++++++++++++++++++-- 1 file changed, 309 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index f456b571d9..038406e953 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -1,6 +1,7 @@ /************************************************************************** * * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 1999 Wittawat Yamwong * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -34,15 +35,12 @@ #include -#include -#include -#include "main/imports.h" -#include "glapi/glthread.h" -#include "main/mm.h" #include "linked_list.h" #include "p_defines.h" +#include "p_thread.h" +#include "p_util.h" #include "pb_buffer.h" #include "pb_bufmgr.h" @@ -53,6 +51,306 @@ #define SUPER(__derived) (&(__derived)->base) +struct mem_block +{ + struct mem_block *next, *prev; + struct mem_block *next_free, *prev_free; + struct mem_block *heap; + int ofs, size; + unsigned int free:1; + unsigned int reserved:1; +}; + + +#ifdef DEBUG +/** + * For debugging purposes. + */ +static void +mmDumpMemInfo(const struct mem_block *heap) +{ + fprintf(stderr, "Memory heap %p:\n", (void *)heap); + if (heap == 0) { + fprintf(stderr, " heap == 0\n"); + } else { + const struct mem_block *p; + + for(p = heap->next; p != heap; p = p->next) { + fprintf(stderr, " Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + fprintf(stderr, "\nFree list:\n"); + + for(p = heap->next_free; p != heap; p = p->next_free) { + fprintf(stderr, " FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + } + fprintf(stderr, "End of memory blocks\n"); +} +#endif + + +/** + * input: total size in bytes + * return: a heap pointer if OK, NULL if error + */ +static struct mem_block * +mmInit(int ofs, int size) +{ + struct mem_block *heap, *block; + + if (size <= 0) + return NULL; + + heap = CALLOC_STRUCT(mem_block); + if (!heap) + return NULL; + + block = CALLOC_STRUCT(mem_block); + if (!block) { + FREE(heap); + return NULL; + } + + heap->next = block; + heap->prev = block; + heap->next_free = block; + heap->prev_free = block; + + block->heap = heap; + block->next = heap; + block->prev = heap; + block->next_free = heap; + block->prev_free = heap; + + block->ofs = ofs; + block->size = size; + block->free = 1; + + return heap; +} + + +static struct mem_block * +SliceBlock(struct mem_block *p, + int startofs, int size, + int reserved, int alignment) +{ + struct mem_block *newblock; + + /* break left [p, newblock, p->next], then p = newblock */ + if (startofs > p->ofs) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs; + newblock->size = p->size - (startofs - p->ofs); + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size -= newblock->size; + p = newblock; + } + + /* break right, also [p, newblock, p->next] */ + if (size < p->size) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs + size; + newblock->size = p->size - size; + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size = size; + } + + /* p = middle block */ + p->free = 0; + + /* Remove p from the free list: + */ + p->next_free->prev_free = p->prev_free; + p->prev_free->next_free = p->next_free; + + p->next_free = 0; + p->prev_free = 0; + + p->reserved = reserved; + return p; +} + + +/** + * Allocate 'size' bytes with 2^align2 bytes alignment, + * restrict the search to free memory after 'startSearch' + * depth and back buffers should be in different 4mb banks + * to get better page hits if possible + * input: size = size of block + * align2 = 2^align2 bytes alignment + * startSearch = linear offset from start of heap to begin search + * return: pointer to the allocated block, 0 if error + */ +static struct mem_block * +mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) +{ + struct mem_block *p; + const int mask = (1 << align2)-1; + int startofs = 0; + int endofs; + + if (!heap || align2 < 0 || size <= 0) + return NULL; + + for (p = heap->next_free; p != heap; p = p->next_free) { + assert(p->free); + + startofs = (p->ofs + mask) & ~mask; + if ( startofs < startSearch ) { + startofs = startSearch; + } + endofs = startofs+size; + if (endofs <= (p->ofs+p->size)) + break; + } + + if (p == heap) + return NULL; + + assert(p->free); + p = SliceBlock(p,startofs,size,0,mask+1); + + return p; +} + + +#if 0 +/** + * Free block starts at offset + * input: pointer to a heap, start offset + * return: pointer to a block + */ +static struct mem_block * +mmFindBlock(struct mem_block *heap, int start) +{ + struct mem_block *p; + + for (p = heap->next; p != heap; p = p->next) { + if (p->ofs == start) + return p; + } + + return NULL; +} +#endif + + +static INLINE int +Join2Blocks(struct mem_block *p) +{ + /* XXX there should be some assertions here */ + + /* NOTE: heap->free == 0 */ + + if (p->free && p->next->free) { + struct mem_block *q = p->next; + + assert(p->ofs + p->size == q->ofs); + p->size += q->size; + + p->next = q->next; + q->next->prev = p; + + q->next_free->prev_free = q->prev_free; + q->prev_free->next_free = q->next_free; + + FREE(q); + return 1; + } + return 0; +} + + +/** + * Free block starts at offset + * input: pointer to a block + * return: 0 if OK, -1 if error + */ +static int +mmFreeMem(struct mem_block *b) +{ + if (!b) + return 0; + + if (b->free) { + fprintf(stderr, "block already free\n"); + return -1; + } + if (b->reserved) { + fprintf(stderr, "block is reserved\n"); + return -1; + } + + b->free = 1; + b->next_free = b->heap->next_free; + b->prev_free = b->heap; + b->next_free->prev_free = b; + b->prev_free->next_free = b; + + Join2Blocks(b); + if (b->prev != b->heap) + Join2Blocks(b->prev); + + return 0; +} + + +/** + * destroy MM + */ +static void +mmDestroy(struct mem_block *heap) +{ + struct mem_block *p; + + if (!heap) + return; + + for (p = heap->next; p != heap; ) { + struct mem_block *next = p->next; + FREE(p); + p = next; + } + + FREE(heap); +} + + struct mm_pb_manager { struct pb_manager base; @@ -103,7 +401,7 @@ mm_buffer_destroy(struct pb_buffer *buf) _glthread_LOCK_MUTEX(mm->mutex); mmFreeMem(mm_buf->block); - free(buf); + FREE(buf); _glthread_UNLOCK_MUTEX(mm->mutex); } @@ -156,7 +454,7 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, _glthread_LOCK_MUTEX(mm->mutex); - mm_buf = (struct mm_buffer *)malloc(sizeof(*mm_buf)); + mm_buf = CALLOC_STRUCT(mm_buffer); if (!mm_buf) { _glthread_UNLOCK_MUTEX(mm->mutex); return NULL; @@ -176,7 +474,7 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); if(!mm_buf->block) { assert(0); - free(mm_buf); + FREE(mm_buf); _glthread_UNLOCK_MUTEX(mm->mutex); return NULL; } @@ -205,7 +503,7 @@ mm_bufmgr_destroy(struct pb_manager *mgr) _glthread_UNLOCK_MUTEX(mm->mutex); - free(mgr); + FREE(mgr); } @@ -218,7 +516,7 @@ mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, if(!buffer) return NULL; - mm = (struct mm_pb_manager *)calloc(1, sizeof(*mm)); + mm = CALLOC_STRUCT(mm_pb_manager); if (!mm) return NULL; @@ -250,7 +548,7 @@ if(mm->heap) if(mm->map) pb_unmap(mm->buffer); if(mm) - free(mm); + FREE(mm); return NULL; } -- cgit v1.2.3 From 65df0241465b2dae4979d71cad17b83cfd1fda11 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 17 Jan 2008 17:06:16 +0900 Subject: gallium: Aggregate all buffer allocation info into a single struct. Obey requested alignment. --- src/mesa/pipe/pipebuffer/pb_buffer.h | 18 ++++++++++++++---- src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 13 ++++++------- src/mesa/pipe/pipebuffer/pb_bufmgr.h | 7 +++++-- src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c | 8 +++++--- src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 15 +++++++++++++-- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 12 +++++++++--- 6 files changed, 52 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.h b/src/mesa/pipe/pipebuffer/pb_buffer.h index 570986314a..2f570ef9de 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer.h @@ -53,6 +53,18 @@ struct pb_vtbl; +/** + * Buffer description. + * + * Used when allocating the buffer. + */ +struct pb_desc +{ + unsigned alignment; + unsigned usage; +}; + + /** * Base class for all pb_* buffers. */ @@ -159,10 +171,8 @@ pb_user_buffer_create(void *data, unsigned bytes); * hardware. */ struct pb_buffer * -pb_malloc_buffer_create( unsigned alignment, - unsigned usage, - unsigned size ); - +pb_malloc_buffer_create(size_t size, + const struct pb_desc *desc); static INLINE struct pipe_buffer * diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c index b80ee8dda3..f0ff1d347e 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c @@ -62,7 +62,7 @@ malloc_buffer(struct pb_buffer *buf) static void malloc_buffer_destroy(struct pb_buffer *buf) { - FREE(malloc_buffer(buf)->data); + align_free(malloc_buffer(buf)->data); FREE(buf); } @@ -102,9 +102,8 @@ malloc_buffer_vtbl = { struct pb_buffer * -pb_malloc_buffer_create( unsigned alignment, - unsigned usage, - unsigned size ) +pb_malloc_buffer_create(size_t size, + const struct pb_desc *desc) { struct malloc_buffer *buf; @@ -116,11 +115,11 @@ pb_malloc_buffer_create( unsigned alignment, return NULL; buf->base.vtbl = &malloc_buffer_vtbl; - buf->base.base.alignment = alignment; - buf->base.base.usage = usage; + buf->base.base.alignment = desc->alignment; + buf->base.base.usage = desc->usage; buf->base.base.size = size; - buf->data = MALLOC(size); + buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment); if(!buf->data) { FREE(buf); return NULL; diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr.h b/src/mesa/pipe/pipebuffer/pb_bufmgr.h index 13d4ea7545..1ddf784c97 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr.h +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr.h @@ -53,6 +53,7 @@ #include +struct pb_desc; struct pipe_buffer; struct pipe_winsys; @@ -65,7 +66,8 @@ struct pb_manager /* XXX: we will likely need more allocation flags */ struct pb_buffer * (*create_buffer)( struct pb_manager *mgr, - size_t size ); + size_t size, + const struct pb_desc *desc); void (*destroy)( struct pb_manager *mgr ); @@ -82,7 +84,8 @@ struct pb_manager */ struct pb_manager * pool_bufmgr_create(struct pb_manager *provider, - size_t n, size_t size); + size_t n, size_t size, + const struct pb_desc *desc); /** diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c index 734d5bdceb..6a81cbdae0 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c @@ -63,7 +63,9 @@ fenced_pb_manager(struct pb_manager *mgr) static struct pb_buffer * -fenced_bufmgr_create_buffer(struct pb_manager *mgr, size_t size) +fenced_bufmgr_create_buffer(struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc) { struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); struct pb_buffer *buf; @@ -72,12 +74,12 @@ fenced_bufmgr_create_buffer(struct pb_manager *mgr, size_t size) /* check for free buffers before allocating new ones */ fenced_buffer_list_check_free(fenced_mgr->fenced_list, 0); - buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size); + buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); if(!buf) { /* try harder to get a buffer */ fenced_buffer_list_check_free(fenced_mgr->fenced_list, 1); - buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size); + buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); if(!buf) { /* give up */ return NULL; diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index 038406e953..2a62702c36 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -447,11 +447,17 @@ mm_buffer_vtbl = { static struct pb_buffer * mm_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size) + size_t size, + const struct pb_desc *desc) { struct mm_pb_manager *mm = mm_pb_manager(mgr); struct mm_buffer *mm_buf; + /* We don't handle alignments larger then the one initially setup */ + assert(desc->alignment % (1 << mm->align2) == 0); + if(desc->alignment % (1 << mm->align2)) + return NULL; + _glthread_LOCK_MUTEX(mm->mutex); mm_buf = CALLOC_STRUCT(mm_buffer); @@ -559,10 +565,15 @@ mm_bufmgr_create(struct pb_manager *provider, { struct pb_buffer *buffer; struct pb_manager *mgr; + struct pb_desc desc; assert(provider); assert(provider->create_buffer); - buffer = provider->create_buffer(provider, size); + + memset(&desc, 0, sizeof(desc)); + desc.alignment = 1 << align2; + + buffer = provider->create_buffer(provider, size, &desc); if (!buffer) return NULL; diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index 33b2f628dd..61c06ec824 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -62,6 +62,7 @@ struct pool_pb_manager _glthread_Mutex mutex; size_t bufSize; + size_t bufAlign; size_t numFree; size_t numTot; @@ -160,13 +161,16 @@ pool_buffer_vtbl = { static struct pb_buffer * -pool_bufmgr_create_buffer(struct pb_manager *mgr, size_t size) +pool_bufmgr_create_buffer(struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc) { struct pool_pb_manager *pool = pool_pb_manager(mgr); struct pool_buffer *pool_buf; struct list_head *item; assert(size == pool->bufSize); + assert(desc->alignment % pool->bufAlign == 0); _glthread_LOCK_MUTEX(pool->mutex); @@ -213,7 +217,8 @@ pool_bufmgr_destroy(struct pb_manager *mgr) struct pb_manager * pool_bufmgr_create(struct pb_manager *provider, size_t numBufs, - size_t bufSize) + size_t bufSize, + const struct pb_desc *desc) { struct pool_pb_manager *pool; struct pool_buffer *pool_buf; @@ -231,10 +236,11 @@ pool_bufmgr_create(struct pb_manager *provider, pool->numTot = numBufs; pool->numFree = numBufs; pool->bufSize = bufSize; + pool->bufAlign = desc->alignment; _glthread_INIT_MUTEX(pool->mutex); - pool->buffer = provider->create_buffer(provider, numBufs*bufSize); + pool->buffer = provider->create_buffer(provider, numBufs*bufSize, desc); if (!pool->buffer) goto failure; -- cgit v1.2.3 From 7f4bec264e8180609846379da6f63555fc2eafef Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 14:11:50 -0700 Subject: Cell: move ASSERT macro into common.h --- src/mesa/pipe/cell/common.h | 11 +++++++++++ src/mesa/pipe/cell/spu/spu_main.h | 9 --------- 2 files changed, 11 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index f9edd0cc26..e925299119 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -38,6 +38,17 @@ #include "pipe/p_format.h" +/** The standard assert macro doesn't seem to work reliably */ +#define ASSERT(x) \ + if (!(x)) { \ + ubyte *p = NULL; \ + fprintf(stderr, "%s:%d: %s(): assertion %s failed.\n", \ + __FILE__, __LINE__, __FUNCTION__, #x); \ + *p = 0; \ + exit(1); \ + } + + /** for sanity checking */ #define ASSERT_ALIGN16(ptr) \ assert((((unsigned long) (ptr)) & 0xf) == 0); diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index e4359bf60d..5bc5d9fa99 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -86,15 +86,6 @@ extern struct spu_global spu; #define TAG_MISC 18 -/** The standard assert macro doesn't seem to work on SPUs */ -#define ASSERT(x) \ - if (!(x)) { \ - fprintf(stderr, "SPU %d: %s:%d: %s(): assertion %s failed.\n", \ - spu.init.id, __FILE__, __LINE__, __FUNCTION__, #x); \ - exit(1); \ - } - - extern void wait_on_mask(unsigned tag); -- cgit v1.2.3 From d4bd6ca8162be0684e4bff06a17b9dea1a02d217 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 14:12:33 -0700 Subject: Cell: add check to catch recursive batch flushing --- src/mesa/pipe/cell/ppu/cell_batch.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index ab4553f16c..5a25f1b266 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -34,13 +34,18 @@ void cell_batch_flush(struct cell_context *cell) { + static boolean flushing = FALSE; uint batch = cell->cur_batch; const uint size = cell->batch_buffer_size[batch]; uint spu, cmd_word; + assert(!flushing); + if (size == 0) return; + flushing = TRUE; + assert(batch < CELL_NUM_BATCH_BUFFERS); /* @@ -86,6 +91,8 @@ cell_batch_flush(struct cell_context *cell) cell->batch_buffer_size[batch] = 0; /* empty */ cell->cur_batch = batch; + + flushing = FALSE; } -- cgit v1.2.3 From 856db7dee69083733a1d6e02a33bec4189387f60 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 14:14:56 -0700 Subject: Cell: fix a recursive flushing bug --- src/mesa/pipe/cell/ppu/cell_flush.c | 18 ++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_flush.h | 3 +++ src/mesa/pipe/cell/ppu/cell_vbuf.c | 2 +- 3 files changed, 22 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c index d54eeb8524..b98bb566b1 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -31,14 +31,30 @@ #include "cell_flush.h" #include "cell_spu.h" #include "cell_render.h" +#include "pipe/draw/draw_context.h" void cell_flush(struct pipe_context *pipe, unsigned flags) { struct cell_context *cell = cell_context(pipe); + + draw_flush( cell->draw ); + cell_flush_int(pipe, flags); +} + + +/** internal flush */ +void +cell_flush_int(struct pipe_context *pipe, unsigned flags) +{ + static boolean flushing = FALSE; /* recursion catcher */ + struct cell_context *cell = cell_context(pipe); uint i; + ASSERT(!flushing); + flushing = TRUE; + if (flags & PIPE_FLUSH_WAIT) { uint *cmd = (uint *) cell_batch_alloc(cell, sizeof(uint)); *cmd = CELL_CMD_FINISH; @@ -60,4 +76,6 @@ cell_flush(struct pipe_context *pipe, unsigned flags) assert(k == CELL_CMD_FINISH); } } + + flushing = FALSE; } diff --git a/src/mesa/pipe/cell/ppu/cell_flush.h b/src/mesa/pipe/cell/ppu/cell_flush.h index cf1a104f97..eda351b1cb 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.h +++ b/src/mesa/pipe/cell/ppu/cell_flush.h @@ -32,4 +32,7 @@ extern void cell_flush(struct pipe_context *pipe, unsigned flags); +extern void +cell_flush_int(struct pipe_context *pipe, unsigned flags); + #endif diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 711d03b49f..00ff990eab 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -166,7 +166,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, #if 01 /* XXX this is temporary */ - cell_flush(&cell->pipe, PIPE_FLUSH_WAIT); + cell_flush_int(&cell->pipe, PIPE_FLUSH_WAIT); #endif } -- cgit v1.2.3 From 9ef61b98087603617702043c07633282148883e3 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 16:04:47 -0700 Subject: gallium: fix a few segfaults/assertions that can happen during context init --- src/mesa/pipe/draw/draw_vertex_fetch.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 1a9e38158d..fb64723a19 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -79,7 +79,7 @@ FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) -static fetch_func get_fetch_func( unsigned format ) +static fetch_func get_fetch_func( enum pipe_format format ) { switch (format) { case PIPE_FORMAT_R32G32B32A32_FLOAT: @@ -102,6 +102,8 @@ static fetch_func get_fetch_func( unsigned format ) return fetch_A8R8G8B8_UNORM; case PIPE_FORMAT_R8G8B8A8_UNORM: return fetch_R8G8B8A8_UNORM; + case 0: + return NULL; default: /* Lots of missing cases! */ assert(0); @@ -127,13 +129,17 @@ transpose_4x4( float *out, const float *in ) void draw_update_vertex_fetch( struct draw_context *draw ) { - //unsigned nr_attrs = draw->vertex_element_count; - unsigned nr_attrs = draw->vertex_shader->state->num_inputs; - unsigned i; + unsigned nr_attrs, i; + + /* this may happend during context init */ + if (!draw->vertex_shader) + return; + + nr_attrs = draw->vertex_shader->state->num_inputs; for (i = 0; i < nr_attrs; i++) { unsigned buf = draw->vertex_element[i].vertex_buffer_index; - unsigned format = draw->vertex_element[i].src_format; + enum pipe_format format = draw->vertex_element[i].src_format; draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + draw->vertex_buffer[buf].buffer_offset + -- cgit v1.2.3 From ffeae750314b918839c118896a645d4f4f15d230 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 16:05:35 -0700 Subject: gallium: s/GLuint/unsigned/ to work with gallium when THREADS not defined --- src/mesa/glapi/glthread.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/glapi/glthread.h b/src/mesa/glapi/glthread.h index a61086d0dc..afb04f7bf2 100644 --- a/src/mesa/glapi/glthread.h +++ b/src/mesa/glapi/glthread.h @@ -259,11 +259,11 @@ typedef benaphore _glthread_Mutex; * THREADS not defined */ -typedef GLuint _glthread_TSD; +typedef unsigned _glthread_TSD; -typedef GLuint _glthread_Thread; +typedef unsigned _glthread_Thread; -typedef GLuint _glthread_Mutex; +typedef unsigned _glthread_Mutex; #define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 -- cgit v1.2.3 From b717de3238a028a3fdfbaf13eb02dbde262f03e7 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 16:06:26 -0700 Subject: Cell: remove some old tile code --- src/mesa/pipe/cell/ppu/cell_surface.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index c8796dab43..6b7b918128 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -173,10 +173,6 @@ cell_surface_fill(struct pipe_context *pipe, void cell_init_surface_functions(struct cell_context *cell) { - cell->pipe.get_tile = pipe_get_tile_raw; - cell->pipe.put_tile = pipe_put_tile_raw; - - cell->pipe.surface_data = cell_surface_data; cell->pipe.surface_copy = cell_surface_copy; cell->pipe.surface_fill = cell_surface_fill; } -- cgit v1.2.3 From 651e8e9a73b4f0c3424a78b978f710d098f47ae2 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 16:46:52 -0700 Subject: gallium: disable unnecessary point/line/tri re-validation in vbuf_flush_indices() --- src/mesa/pipe/draw/draw_vbuf.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index cd0b4fbbb9..aea5b4aeee 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -331,9 +331,12 @@ vbuf_flush_indices( struct draw_stage *stage ) vbuf->nr_indices = 0; + /* don't need to reset point/line/tri functions */ +#if 0 stage->point = vbuf_first_point; stage->line = vbuf_first_line; stage->tri = vbuf_first_tri; +#endif } -- cgit v1.2.3 From 8dc597290813f58e2f2b7ddcb4ad762ea379c1f7 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 18:27:37 -0700 Subject: Cell: added cell_batch_free_space() --- src/mesa/pipe/cell/ppu/cell_batch.c | 11 +++++++++++ src/mesa/pipe/cell/ppu/cell_batch.h | 3 +++ 2 files changed, 14 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index 5a25f1b266..c894ef8608 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -96,6 +96,15 @@ cell_batch_flush(struct cell_context *cell) } +uint +cell_batch_free_space(const struct cell_context *cell) +{ + uint free = CELL_BATCH_BUFFER_SIZE + - cell->batch_buffer_size[cell->cur_batch]; + return free; +} + + /** * \param cmd command to append * \param length command size in bytes @@ -129,6 +138,8 @@ cell_batch_alloc(struct cell_context *cell, uint bytes) void *pos; uint size; + ASSERT(bytes % 4 == 0); + assert(cell->cur_batch >= 0); size = cell->batch_buffer_size[cell->cur_batch]; diff --git a/src/mesa/pipe/cell/ppu/cell_batch.h b/src/mesa/pipe/cell/ppu/cell_batch.h index 47e3287626..c4ba7feb3d 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.h +++ b/src/mesa/pipe/cell/ppu/cell_batch.h @@ -38,6 +38,9 @@ struct cell_context; extern void cell_batch_flush(struct cell_context *cell); +extern uint +cell_batch_free_space(const struct cell_context *cell); + extern void cell_batch_append(struct cell_context *cell, const void *cmd, uint length); -- cgit v1.2.3 From 1bab5bd24e09b6e11cd99f989bb00c587119aed2 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 18:30:44 -0700 Subject: Cell: added support for inlined indexes If there's room in the batch buffer after the rendering command to accomodate the indexes, put them there rather than in a separate buffer. --- src/mesa/pipe/cell/common.h | 1 + src/mesa/pipe/cell/ppu/cell_vbuf.c | 31 +++++++-- src/mesa/pipe/cell/spu/spu_main.c | 125 +++++++++++++++++++++++++------------ 3 files changed, 114 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index e925299119..e955bb9ec5 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -124,6 +124,7 @@ struct cell_command_render const void *vertex_data; const ushort *index_data; float xmin, ymin, xmax, ymax; + boolean inline_indexes; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 00ff990eab..59a4a2b6e9 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -39,6 +39,10 @@ #include "pipe/draw/draw_vbuf.h" +/** Allow render indexes to be inlined after RENDER command */ +#define ALLOW_INLINING 1 + + /** * Subclass of vbuf_render because we need a cell_context pointer in * a few places. @@ -123,6 +127,8 @@ cell_vbuf_draw(struct vbuf_render *vbr, printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u indexes = [%u %u %u ...]\n", nr_indices, nr_vertices, indices[0], indices[1], indices[2]); + printf("ind space = %u, space = %u\n", + nr_indices * 2, cell_batch_free_space(cell)); #endif /* compute x/y bounding box */ @@ -145,23 +151,40 @@ cell_vbuf_draw(struct vbuf_render *vbr, /* build/insert batch RENDER command */ { + const uint index_bytes = (nr_indices * 2 + 3) & ~0x3; + struct cell_command_render *render = (struct cell_command_render *) cell_batch_alloc(cell, sizeof(*render)); render->opcode = CELL_CMD_RENDER; render->prim_type = cvbr->prim; + render->num_verts = nr_vertices; render->vertex_size = 4 * cell->vertex_info.size; render->vertex_data = vertices; - render->index_data = indices; + ASSERT_ALIGN16(render->vertex_data); + render->num_indexes = nr_indices; + + if (ALLOW_INLINING && + index_bytes <= cell_batch_free_space(cell)) { + /* indices inlined, right after render cmd */ + void *dst = cell_batch_alloc(cell, index_bytes); + memcpy(dst, indices, nr_indices * 2); + render->inline_indexes = TRUE; + render->index_data = NULL; + } + else { + /* indices in separate buffer */ + render->inline_indexes = FALSE; + render->index_data = indices; + ASSERT_ALIGN16(render->index_data); + } + render->xmin = xmin; render->ymin = ymin; render->xmax = xmax; render->ymax = ymax; - - ASSERT_ALIGN16(render->vertex_data); - ASSERT_ALIGN16(render->index_data); } #if 01 diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index ef605a5197..92be0b4a4a 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -204,66 +204,104 @@ tile_bounding_box(const struct cell_command_render *render, } +/** + * Render primitives + * \param pos_incr returns value indicating how may words to skip after + * this command in the batch buffer + */ static void -cmd_render(const struct cell_command_render *render) +cmd_render(const struct cell_command_render *render, uint *pos_incr) { /* we'll DMA into these buffers */ ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; - ushort indexes[CELL_MAX_VBUF_INDEXES] ALIGN16_ATTRIB; - uint i, j, total_vertex_bytes, total_index_bytes; + ushort index_data[CELL_MAX_VBUF_INDEXES] ALIGN16_ATTRIB; const uint vertex_size = render->vertex_size; /* in bytes */ + const ushort *indexes; + uint i, j; + if (Debug) { - printf("SPU %u: RENDER prim %u, indices: %u, nr_vert: %u\n", + printf("SPU %u: RENDER prim %u, num_vert=%u num_ind=%u inlined=%u\n", spu.init.id, render->prim_type, render->num_verts, - render->num_indexes); + render->num_indexes, + render->inline_indexes); + /* printf(" bound: %g, %g .. %g, %g\n", render->xmin, render->ymin, render->xmax, render->ymax); */ + printf("SPU %u: indices at %p vertices at %p\n", + spu.init.id, + render->index_data, render->vertex_data); } ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(render->index_data); - /* how much vertex data */ - total_vertex_bytes = render->num_verts * vertex_size; - total_index_bytes = render->num_indexes * sizeof(ushort); - if (total_index_bytes < 16) - total_index_bytes = 16; - else - total_index_bytes = ROUNDUP16(total_index_bytes); - /* - printf("VBUF: indices at %p, vertices at %p total_vertex_bytes %u ind_bytes %u\n", - render->index_data, render->vertex_data, total_vertex_bytes, total_index_bytes); - */ + /** + ** Get vertices + **/ + { + const uint total_vertex_bytes = render->num_verts * vertex_size; - ASSERT(total_vertex_bytes % 16 == 0); - /* get vertex data from main memory */ - mfc_get(vertex_data, /* dest */ - (unsigned int) render->vertex_data, /* src */ - total_vertex_bytes, /* size */ - TAG_VERTEX_BUFFER, - 0, /* tid */ - 0 /* rid */); + ASSERT(total_vertex_bytes % 16 == 0); + + /* get vertex data from main memory */ + mfc_get(vertex_data, /* dest */ + (unsigned int) render->vertex_data, /* src */ + total_vertex_bytes, /* size */ + TAG_VERTEX_BUFFER, + 0, /* tid */ + 0 /* rid */); + } - ASSERT(total_index_bytes % 16 == 0); - /* get index data from main memory */ - mfc_get(indexes, /* dest */ - (unsigned int) render->index_data, /* src */ - total_index_bytes, - TAG_INDEX_BUFFER, - 0, /* tid */ - 0 /* rid */); + /** + ** Get indexes + **/ + if (render->inline_indexes) { + /* indexes are right after the render command in the batch buffer */ + ASSERT(sizeof(*render) % 4 == 0); + indexes = (ushort *) (render + 1); - wait_on_mask_all((1 << TAG_VERTEX_BUFFER) | - (1 << TAG_INDEX_BUFFER)); + *pos_incr = (render->num_indexes * 2 + 3) / 4; - /* find tiles which intersect the prim bounding box */ + /* wait for vertex data */ + wait_on_mask_all(1 << TAG_VERTEX_BUFFER); + } + else { + /* indexes are in separate buffer */ + uint total_index_bytes; + + *pos_incr = 0; + + total_index_bytes = render->num_indexes * sizeof(ushort); + if (total_index_bytes < 16) + total_index_bytes = 16; + else + total_index_bytes = ROUNDUP16(total_index_bytes); + + indexes = index_data; + + /* get index data from main memory */ + mfc_get(index_data, /* dest */ + (unsigned int) render->index_data, /* src */ + total_index_bytes, + TAG_INDEX_BUFFER, + 0, /* tid */ + 0 /* rid */); + + wait_on_mask_all((1 << TAG_VERTEX_BUFFER) | + (1 << TAG_INDEX_BUFFER)); + } + + + /** + ** find tiles which intersect the prim bounding box + **/ uint txmin, tymin, box_width_tiles, box_num_tiles; #if 0 tile_bounding_box(render, &txmin, &tymin, @@ -278,7 +316,10 @@ cmd_render(const struct cell_command_render *render) /* make sure any pending clears have completed */ wait_on_mask(1 << TAG_SURFACE_CLEAR); - /* loop over tiles */ + + /** + ** loop over tiles, rendering tris + **/ for (i = spu.init.id; i < box_num_tiles; i += spu.init.num_spus) { const uint tx = txmin + i % box_width_tiles; const uint ty = tymin + i / box_width_tiles; @@ -300,6 +341,7 @@ cmd_render(const struct cell_command_render *render) } ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); + ASSERT(render->num_indexes % 3 == 0); /* loop over tris */ for (j = 0; j < render->num_indexes; j += 3) { @@ -508,8 +550,9 @@ cmd_batch(uint opcode) { struct cell_command_render *render = (struct cell_command_render *) &buffer[pos]; - cmd_render(render); - pos += sizeof(*render) / 4; + uint pos_incr; + cmd_render(render, &pos_incr); + pos += sizeof(*render) / 4 + pos_incr; } break; case CELL_CMD_FINISH: @@ -591,7 +634,11 @@ main_loop(void) cmd_clear_surface(&cmd.clear); break; case CELL_CMD_RENDER: - cmd_render(&cmd.render); + { + uint pos_incr; + cmd_render(&cmd.render, &pos_incr); + assert(pos_incr == 0); + } break; case CELL_CMD_BATCH: cmd_batch(opcode); -- cgit v1.2.3 From 419a5cca34d931e61587eaeb8d32a44b415c43ad Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 19:31:20 -0700 Subject: Cell: added ROUNUP4 --- src/mesa/pipe/cell/common.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index e955bb9ec5..bdde166630 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -54,6 +54,9 @@ assert((((unsigned long) (ptr)) & 0xf) == 0); +/** round up value to next multiple of 4 */ +#define ROUNDUP4(k) (((k) + 0x3) & ~0x3) + /** round up value to next multiple of 16 */ #define ROUNDUP16(k) (((k) + 0xf) & ~0xf) -- cgit v1.2.3 From 87a8a339d7c8973168ffb5e5506f7ec4b3a524ba Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 26 Jan 2008 19:38:16 -0700 Subject: Cell: added support for inlined vertex buffers. Small prims are now self-contained in batch buffers when space allows. --- src/mesa/pipe/cell/common.h | 1 + src/mesa/pipe/cell/ppu/cell_vbuf.c | 40 ++++++++++++++++------- src/mesa/pipe/cell/spu/spu_main.c | 66 ++++++++++++++++++++++---------------- 3 files changed, 68 insertions(+), 39 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index bdde166630..0b63ed39be 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -128,6 +128,7 @@ struct cell_command_render const ushort *index_data; float xmin, ymin, xmax, ymax; boolean inline_indexes; + boolean inline_verts; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 59a4a2b6e9..ee572b3a51 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -39,8 +39,9 @@ #include "pipe/draw/draw_vbuf.h" -/** Allow render indexes to be inlined after RENDER command */ -#define ALLOW_INLINING 1 +/** Allow prim indexes, verts to be inlined after RENDER command */ +#define ALLOW_INLINE_INDEXES 1 +#define ALLOW_INLINE_VERTS 1 /** @@ -127,8 +128,10 @@ cell_vbuf_draw(struct vbuf_render *vbr, printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u indexes = [%u %u %u ...]\n", nr_indices, nr_vertices, indices[0], indices[1], indices[2]); - printf("ind space = %u, space = %u\n", - nr_indices * 2, cell_batch_free_space(cell)); + printf("ind space = %u, vert space = %u, space = %u\n", + nr_indices * 2, + nr_vertices * 4 * cell->vertex_info.size, + cell_batch_free_space(cell)); #endif /* compute x/y bounding box */ @@ -151,7 +154,8 @@ cell_vbuf_draw(struct vbuf_render *vbr, /* build/insert batch RENDER command */ { - const uint index_bytes = (nr_indices * 2 + 3) & ~0x3; + const uint index_bytes = ROUNDUP4(nr_indices * 2); + const uint vertex_bytes = nr_vertices * 4 * cell->vertex_info.size; struct cell_command_render *render = (struct cell_command_render *) @@ -159,14 +163,8 @@ cell_vbuf_draw(struct vbuf_render *vbr, render->opcode = CELL_CMD_RENDER; render->prim_type = cvbr->prim; - render->num_verts = nr_vertices; - render->vertex_size = 4 * cell->vertex_info.size; - render->vertex_data = vertices; - ASSERT_ALIGN16(render->vertex_data); - render->num_indexes = nr_indices; - - if (ALLOW_INLINING && + if (ALLOW_INLINE_INDEXES && index_bytes <= cell_batch_free_space(cell)) { /* indices inlined, right after render cmd */ void *dst = cell_batch_alloc(cell, index_bytes); @@ -181,6 +179,24 @@ cell_vbuf_draw(struct vbuf_render *vbr, ASSERT_ALIGN16(render->index_data); } + render->vertex_size = 4 * cell->vertex_info.size; + render->num_verts = nr_vertices; + if (ALLOW_INLINE_VERTS && + render->inline_indexes && + vertex_bytes <= cell_batch_free_space(cell)) { + /* vertex data inlined, after indices */ + void *dst = cell_batch_alloc(cell, vertex_bytes); + memcpy(dst, vertices, vertex_bytes); + render->inline_verts = TRUE; + render->vertex_data = NULL; + } + else { + render->inline_verts = FALSE; + render->vertex_data = vertices; + ASSERT_ALIGN16(render->vertex_data); + } + + render->xmin = xmin; render->ymin = ymin; render->xmax = xmax; diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 92be0b4a4a..0c83900a18 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -216,16 +216,21 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; ushort index_data[CELL_MAX_VBUF_INDEXES] ALIGN16_ATTRIB; const uint vertex_size = render->vertex_size; /* in bytes */ + const uint total_vertex_bytes = render->num_verts * vertex_size; + const ubyte *vertices; const ushort *indexes; + uint mask; uint i, j; if (Debug) { - printf("SPU %u: RENDER prim %u, num_vert=%u num_ind=%u inlined=%u\n", + printf("SPU %u: RENDER prim %u, num_vert=%u num_ind=%u " + "inline_vert=%u inline_ind=%u\n", spu.init.id, render->prim_type, render->num_verts, render->num_indexes, + render->inline_verts, render->inline_indexes); /* @@ -237,43 +242,28 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) render->index_data, render->vertex_data); } + ASSERT(sizeof(*render) % 4 == 0); ASSERT_ALIGN16(render->vertex_data); ASSERT_ALIGN16(render->index_data); /** - ** Get vertices + ** Get vertex, index buffers if not inlined **/ - { - const uint total_vertex_bytes = render->num_verts * vertex_size; - + if (!render->inline_verts) { ASSERT(total_vertex_bytes % 16 == 0); - /* get vertex data from main memory */ mfc_get(vertex_data, /* dest */ (unsigned int) render->vertex_data, /* src */ total_vertex_bytes, /* size */ TAG_VERTEX_BUFFER, 0, /* tid */ 0 /* rid */); - } - - /** - ** Get indexes - **/ - if (render->inline_indexes) { - /* indexes are right after the render command in the batch buffer */ - ASSERT(sizeof(*render) % 4 == 0); - indexes = (ushort *) (render + 1); - - *pos_incr = (render->num_indexes * 2 + 3) / 4; - - /* wait for vertex data */ - wait_on_mask_all(1 << TAG_VERTEX_BUFFER); + vertices = vertex_data; } - else { - /* indexes are in separate buffer */ + + if (!render->inline_indexes) { uint total_index_bytes; *pos_incr = 0; @@ -293,12 +283,34 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) TAG_INDEX_BUFFER, 0, /* tid */ 0 /* rid */); + } + + + /** + ** Get pointers to inlined indexes, verts, if present + **/ + if (render->inline_indexes) { + /* indexes are right after the render command in the batch buffer */ + indexes = (ushort *) (render + 1); + *pos_incr = (render->num_indexes * 2 + 3) / 4; - wait_on_mask_all((1 << TAG_VERTEX_BUFFER) | - (1 << TAG_INDEX_BUFFER)); + if (render->inline_verts) { + /* vertices are after indexes, if inlined */ + vertices = (const ubyte *) (render + 1) + *pos_incr * 4; + *pos_incr = *pos_incr + total_vertex_bytes / 4; + } } + /* wait for vertex and/or index buffers if not inlined */ + mask = 0x0; + if (!render->inline_verts) + mask |= (1 << TAG_VERTEX_BUFFER); + if (!render->inline_indexes) + mask |= (1 << TAG_INDEX_BUFFER); + wait_on_mask_all(mask); + + /** ** find tiles which intersect the prim bounding box **/ @@ -347,9 +359,9 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) for (j = 0; j < render->num_indexes; j += 3) { const float *v0, *v1, *v2; - v0 = (const float *) (vertex_data + indexes[j+0] * vertex_size); - v1 = (const float *) (vertex_data + indexes[j+1] * vertex_size); - v2 = (const float *) (vertex_data + indexes[j+2] * vertex_size); + v0 = (const float *) (vertices + indexes[j+0] * vertex_size); + v1 = (const float *) (vertices + indexes[j+1] * vertex_size); + v2 = (const float *) (vertices + indexes[j+2] * vertex_size); tri_draw(v0, v1, v2, tx, ty); } -- cgit v1.2.3 From e6c8278c04518b8b8b0960a9e21b48b6816fcc20 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 27 Jan 2008 19:20:48 +0900 Subject: Fullfill pipe_winsys->user_buffer_create. Aggregate winsys functions. --- src/mesa/pipe/pipebuffer/Makefile | 5 +- src/mesa/pipe/pipebuffer/pb_buffer.c | 77 ------------- src/mesa/pipe/pipebuffer/pb_buffer.h | 62 +++++----- src/mesa/pipe/pipebuffer/pb_buffer_client.c | 115 ------------------- src/mesa/pipe/pipebuffer/pb_winsys.c | 170 ++++++++++++++++++++++++++++ 5 files changed, 207 insertions(+), 222 deletions(-) delete mode 100644 src/mesa/pipe/pipebuffer/pb_buffer.c delete mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_client.c create mode 100644 src/mesa/pipe/pipebuffer/pb_winsys.c (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/Makefile b/src/mesa/pipe/pipebuffer/Makefile index ea8c3440c3..75764a9a18 100644 --- a/src/mesa/pipe/pipebuffer/Makefile +++ b/src/mesa/pipe/pipebuffer/Makefile @@ -5,13 +5,12 @@ include $(TOP)/configs/current LIBNAME = pipebuffer DRIVER_SOURCES = \ - pb_buffer.c \ - pb_buffer_client.c \ pb_buffer_fenced.c \ pb_buffer_malloc.c \ pb_bufmgr_fenced.c \ pb_bufmgr_mm.c \ - pb_bufmgr_pool.c + pb_bufmgr_pool.c \ + pb_winsys.c C_SOURCES = \ $(DRIVER_SOURCES) diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.c b/src/mesa/pipe/pipebuffer/pb_buffer.c deleted file mode 100644 index 90ab9044ff..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_buffer.c +++ /dev/null @@ -1,77 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Buffer implementation. - * - * \author José Fonseca - */ - - -#include "pb_buffer.h" -#include "pipe/p_winsys.h" - - - -static void * -pb_winsys_map(struct pipe_winsys *winsys, - struct pipe_buffer *ws_buf, - unsigned flags) -{ - struct pb_buffer *buf = pb_buffer(ws_buf); - - return buf->vtbl->map(buf, flags); -} - -static void -pb_winsys_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *ws_buf) -{ - struct pb_buffer *buf = pb_buffer(ws_buf); - - buf->vtbl->unmap(buf); -} - -static void -pb_winsys_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *ws_buf) -{ - struct pb_buffer *buf = pb_buffer(ws_buf); - - buf->vtbl->destroy(buf); -} - - - -void -pb_init_winsys(struct pipe_winsys *winsys) -{ - winsys->buffer_map = pb_winsys_map; - winsys->buffer_unmap = pb_winsys_unmap; - winsys->buffer_destroy = pb_winsys_destroy; -} diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.h b/src/mesa/pipe/pipebuffer/pb_buffer.h index 2f570ef9de..17551b3b50 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer.h @@ -48,8 +48,9 @@ #include #include "pipe/p_compiler.h" - #include "pipe/p_state.h" +#include "pipe/p_inlines.h" + struct pb_vtbl; @@ -81,6 +82,7 @@ struct pb_buffer const struct pb_vtbl *vtbl; }; + /** * Virtual function table for the buffer storage operations. * @@ -116,6 +118,24 @@ struct pb_vtbl }; +static INLINE struct pipe_buffer * +pb_pipe_buffer( struct pb_buffer *pbuf ) +{ + assert(pbuf); + return &pbuf->base; +} + + +static INLINE struct pb_buffer * +pb_buffer( struct pipe_buffer *buf ) +{ + assert(buf); + /* Could add a magic cookie check on debug builds. + */ + return (struct pb_buffer *)buf; +} + + /* Accessor functions for pb->vtbl: */ static INLINE void * @@ -143,6 +163,7 @@ pb_get_base_buffer( struct pb_buffer *buf, buf->vtbl->get_base_buffer(buf, base_buf, offset); } + static INLINE void pb_destroy(struct pb_buffer *buf) { @@ -151,19 +172,20 @@ pb_destroy(struct pb_buffer *buf) } +/* XXX: thread safety issues! + */ +static INLINE void +pb_reference(struct pb_buffer **dst, + struct pb_buffer *src) +{ + if (src) + src->base.refcount++; + if (*dst && --(*dst)->base.refcount == 0) + pb_destroy( *dst ); -/** - * User buffers are special buffers that initially reference memory - * held by the user but which may if necessary copy that memory into - * device memory behind the scenes, for submission to hardware. - * - * These are particularly useful when the referenced data is never - * submitted to hardware at all, in the particular case of software - * vertex processing. - */ -struct pb_buffer * -pb_user_buffer_create(void *data, unsigned bytes); + *dst = src; +} /** @@ -175,22 +197,8 @@ pb_malloc_buffer_create(size_t size, const struct pb_desc *desc); -static INLINE struct pipe_buffer * -pb_pipe_buffer( struct pb_buffer *pbuf ) -{ - return &pbuf->base; -} - -static INLINE struct pb_buffer * -pb_buffer( struct pipe_buffer *buf ) -{ - /* Could add a magic cookie check on debug builds. - */ - return (struct pb_buffer *)buf; -} - - void pb_init_winsys(struct pipe_winsys *winsys); + #endif /*PB_BUFFER_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_client.c b/src/mesa/pipe/pipebuffer/pb_buffer_client.c deleted file mode 100644 index c316aabd32..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_buffer_client.c +++ /dev/null @@ -1,115 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Implementation of client buffer (also designated as "user buffers"), which - * are just state-tracker owned data masqueraded as buffers. - * - * \author José Fonseca - */ - - -#include "pipe/p_util.h" -#include "pb_buffer.h" - - -struct pb_user_buffer -{ - struct pb_buffer base; - void *data; -}; - - -extern const struct pb_vtbl pb_user_buffer_vtbl; - - -static INLINE struct pb_user_buffer * -pb_user_buffer(struct pb_buffer *buf) -{ - assert(buf); - assert(buf->vtbl == &pb_user_buffer_vtbl); - return (struct pb_user_buffer *)buf; -} - - -static void -pb_user_buffer_destroy(struct pb_buffer *buf) -{ - assert(buf); - FREE(buf); -} - - -static void * -pb_user_buffer_map(struct pb_buffer *buf, - unsigned flags) -{ - return pb_user_buffer(buf)->data; -} - - -static void -pb_user_buffer_unmap(struct pb_buffer *buf) -{ - /* No-op */ -} - - -static void -pb_user_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - *base_buf = buf; - *offset = 0; -} - - -const struct pb_vtbl -pb_user_buffer_vtbl = { - pb_user_buffer_destroy, - pb_user_buffer_map, - pb_user_buffer_unmap, - pb_user_buffer_get_base_buffer -}; - - -struct pb_buffer * -pb_user_buffer_create(void *data, unsigned bytes) -{ - struct pb_user_buffer *buf = CALLOC_STRUCT(pb_user_buffer); - - if(!buf) - return NULL; - - buf->base.vtbl = &pb_user_buffer_vtbl; - buf->base.base.size = bytes; - buf->data = data; - - return &buf->base; -} diff --git a/src/mesa/pipe/pipebuffer/pb_winsys.c b/src/mesa/pipe/pipebuffer/pb_winsys.c new file mode 100644 index 0000000000..978944091f --- /dev/null +++ b/src/mesa/pipe/pipebuffer/pb_winsys.c @@ -0,0 +1,170 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of client buffer (also designated as "user buffers"), which + * are just state-tracker owned data masqueraded as buffers. + * + * \author José Fonseca + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" + +#include "pb_buffer.h" + + +/** + * User buffers are special buffers that initially reference memory + * held by the user but which may if necessary copy that memory into + * device memory behind the scenes, for submission to hardware. + * + * These are particularly useful when the referenced data is never + * submitted to hardware at all, in the particular case of software + * vertex processing. + */ +struct pb_user_buffer +{ + struct pb_buffer base; + void *data; +}; + + +extern const struct pb_vtbl pb_user_buffer_vtbl; + + +static INLINE struct pb_user_buffer * +pb_user_buffer(struct pb_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &pb_user_buffer_vtbl); + return (struct pb_user_buffer *)buf; +} + + +static void +pb_user_buffer_destroy(struct pb_buffer *buf) +{ + assert(buf); + FREE(buf); +} + + +static void * +pb_user_buffer_map(struct pb_buffer *buf, + unsigned flags) +{ + return pb_user_buffer(buf)->data; +} + + +static void +pb_user_buffer_unmap(struct pb_buffer *buf) +{ + /* No-op */ +} + + +static void +pb_user_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + *base_buf = buf; + *offset = 0; +} + + +const struct pb_vtbl +pb_user_buffer_vtbl = { + pb_user_buffer_destroy, + pb_user_buffer_map, + pb_user_buffer_unmap, + pb_user_buffer_get_base_buffer +}; + + +static struct pipe_buffer * +pb_winsys_user_buffer_create(struct pipe_winsys *winsys, + void *data, + unsigned bytes) +{ + struct pb_user_buffer *buf = CALLOC_STRUCT(pb_user_buffer); + + if(!buf) + return NULL; + + buf->base.base.refcount = 1; + buf->base.base.size = bytes; + buf->base.base.alignment = 0; + buf->base.base.usage = 0; + + buf->base.vtbl = &pb_user_buffer_vtbl; + buf->data = data; + + return &buf->base.base; +} + + +static void * +pb_winsys_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags) +{ + (void)winsys; + return pb_map(pb_buffer(buf), flags); +} + + +static void +pb_winsys_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + (void)winsys; + pb_unmap(pb_buffer(buf)); +} + + +static void +pb_winsys_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + (void)winsys; + pb_destroy(pb_buffer(buf)); +} + + +void +pb_init_winsys(struct pipe_winsys *winsys) +{ + winsys->user_buffer_create = pb_winsys_user_buffer_create; + winsys->buffer_map = pb_winsys_buffer_map; + winsys->buffer_unmap = pb_winsys_buffer_unmap; + winsys->buffer_destroy = pb_winsys_buffer_destroy; +} -- cgit v1.2.3 From dac124081d4bbc9d7527661e4a96aa78077b9f52 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 27 Jan 2008 19:22:25 +0900 Subject: Do refcounting trhoughout all buffer objects, since it is now a base requirement. --- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 11 +++++++---- src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 3 ++- src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c | 1 + src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 9 +++++++-- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 16 ++++++++++++++-- 5 files changed, 31 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c index b2edc321ef..349647fe6e 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -80,7 +80,6 @@ struct fenced_buffer struct pb_buffer *buffer; - unsigned refcount; struct pipe_fence_handle *fence; struct list_head head; @@ -145,7 +144,7 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, /* Do the delayed destroy: */ - pb_destroy(fenced_buf->buffer); + pb_reference(&fenced_buf->buffer, NULL); free(fenced_buf); } } @@ -162,7 +161,7 @@ fenced_buffer_destroy(struct pb_buffer *buf) fenced_list->numDelayed++; } else { - pb_destroy(fenced_buf->buffer); + pb_reference(&fenced_buf->buffer, NULL); free(fenced_buf); } @@ -220,9 +219,13 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list, if(!buf) return NULL; + buf->base.base.refcount = 1; + buf->base.base.alignment = buffer->base.alignment; + buf->base.base.usage = buffer->base.usage; + buf->base.base.size = buffer->base.size; + buf->base.vtbl = &fenced_buffer_vtbl; buf->buffer = buffer; - buf->refcount = 1; buf->list = fenced_list; return &buf->base; diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c index f0ff1d347e..fc83a00f36 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c @@ -114,10 +114,11 @@ pb_malloc_buffer_create(size_t size, if(!buf) return NULL; - buf->base.vtbl = &malloc_buffer_vtbl; + buf->base.base.refcount = 1; buf->base.base.alignment = desc->alignment; buf->base.base.usage = desc->usage; buf->base.base.size = size; + buf->base.vtbl = &malloc_buffer_vtbl; buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment); if(!buf->data) { diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c index 6a81cbdae0..3b341c64c2 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c @@ -88,6 +88,7 @@ fenced_bufmgr_create_buffer(struct pb_manager *mgr, fenced_buf = fenced_buffer_create(fenced_mgr->fenced_list, buf); if(!fenced_buf) { + assert(buf->base.refcount == 1); pb_destroy(buf); } diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index 2a62702c36..2694f57bca 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -466,6 +466,11 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, return NULL; } + mm_buf->base.base.refcount = 1; + mm_buf->base.base.alignment = desc->alignment; + mm_buf->base.base.usage = desc->usage; + mm_buf->base.base.size = size; + mm_buf->base.vtbl = &mm_buffer_vtbl; mm_buf->mgr = mm; @@ -505,7 +510,7 @@ mm_bufmgr_destroy(struct pb_manager *mgr) mmDestroy(mm->heap); pb_unmap(mm->buffer); - pb_destroy(mm->buffer); + pb_reference(&mm->buffer, NULL); _glthread_UNLOCK_MUTEX(mm->mutex); @@ -579,7 +584,7 @@ mm_bufmgr_create(struct pb_manager *provider, mgr = mm_bufmgr_create_from_buffer(buffer, size, align2); if (!mgr) { - pb_destroy(buffer); + pb_reference(&buffer, NULL); return NULL; } diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index 61c06ec824..7c29954112 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -111,6 +111,8 @@ pool_buffer_destroy(struct pb_buffer *buf) struct pool_buffer *pool_buf = pool_buffer(buf); struct pool_pb_manager *pool = pool_buf->mgr; + assert(pool_buf->base.base.refcount == 0); + _glthread_LOCK_MUTEX(pool->mutex); LIST_ADD(&pool_buf->head, &pool->free); pool->numFree++; @@ -192,7 +194,13 @@ pool_bufmgr_create_buffer(struct pb_manager *mgr, --pool->numFree; _glthread_UNLOCK_MUTEX(pool->mutex); + pool_buf = LIST_ENTRY(struct pool_buffer, item, head); + assert(pool_buf->base.base.refcount == 0); + pool_buf->base.base.refcount = 1; + pool_buf->base.base.alignment = desc->alignment; + pool_buf->base.base.usage = desc->usage; + return SUPER(pool_buf); } @@ -206,7 +214,7 @@ pool_bufmgr_destroy(struct pb_manager *mgr) FREE(pool->bufs); pb_unmap(pool->buffer); - pb_destroy(pool->buffer); + pb_reference(&pool->buffer, NULL); _glthread_UNLOCK_MUTEX(pool->mutex); @@ -256,6 +264,10 @@ pool_bufmgr_create(struct pb_manager *provider, pool_buf = pool->bufs; for (i = 0; i < numBufs; ++i) { + pool_buf->base.base.refcount = 0; + pool_buf->base.base.alignment = 0; + pool_buf->base.base.usage = 0; + pool_buf->base.base.size = bufSize; pool_buf->base.vtbl = &pool_buffer_vtbl; pool_buf->mgr = pool; pool_buf->start = i * bufSize; @@ -271,7 +283,7 @@ failure: if(pool->map) pb_unmap(pool->buffer); if(pool->buffer) - pb_destroy(pool->buffer); + pb_reference(&pool->buffer, NULL); if(pool) FREE(pool); return NULL; -- cgit v1.2.3 From 3b93c74a8d6e36039b79ddf38c11e27aa0bd3b9b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 28 Jan 2008 14:51:14 +0900 Subject: Cache the vinfo in vbuf_stage. --- src/mesa/pipe/draw/draw_vbuf.c | 22 +++++++++++++--------- 1 file changed, 13 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index aea5b4aeee..60ad93133c 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -50,6 +50,8 @@ struct vbuf_stage { struct vbuf_render *render; + const struct vertex_info *vinfo; + /** Vertex size in bytes */ unsigned vertex_size; @@ -84,8 +86,7 @@ vbuf_stage( struct draw_stage *stage ) static void vbuf_flush_indices( struct draw_stage *stage ); static void vbuf_flush_vertices( struct draw_stage *stage ); -static void vbuf_alloc_vertices( struct draw_stage *stage, - unsigned new_vertex_size ); +static void vbuf_alloc_vertices( struct draw_stage *stage ); static INLINE boolean @@ -101,7 +102,7 @@ check_space( struct vbuf_stage *vbuf, unsigned nr ) { if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { vbuf_flush_vertices(&vbuf->stage); - vbuf_alloc_vertices(&vbuf->stage, vbuf->vertex_size); + vbuf_alloc_vertices(&vbuf->stage); } if (vbuf->nr_indices + nr > vbuf->max_indices ) @@ -120,10 +121,12 @@ static INLINE void emit_vertex( struct vbuf_stage *vbuf, struct vertex_header *vertex ) { - const struct vertex_info *vinfo = vbuf->render->get_vertex_info(vbuf->render); + const struct vertex_info *vinfo = vbuf->vinfo; uint i; uint count = 0; /* for debug/sanity */ + + assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); // fprintf(stderr, "emit vertex %d to %p\n", // vbuf->nr_vertices, vbuf->vertex_ptr); @@ -265,8 +268,11 @@ vbuf_set_prim( struct draw_stage *stage, uint newprim ) if (vertex_size != vbuf->vertex_size) vbuf_flush_vertices(stage); + vbuf->vinfo = vinfo; + vbuf->vertex_size = vertex_size; + if (!vbuf->vertices) - vbuf_alloc_vertices(stage, vertex_size); + vbuf_alloc_vertices(stage); } @@ -364,7 +370,7 @@ vbuf_flush_vertices( struct draw_stage *stage ) vbuf->vertices, vbuf->vertex_size, vbuf->nr_vertices); - vbuf->nr_vertices = 0; + vbuf->max_vertices = vbuf->nr_vertices = 0; vbuf->vertex_ptr = vbuf->vertices = NULL; } @@ -372,8 +378,7 @@ vbuf_flush_vertices( struct draw_stage *stage ) static void -vbuf_alloc_vertices( struct draw_stage *stage, - unsigned new_vertex_size ) +vbuf_alloc_vertices( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); @@ -381,7 +386,6 @@ vbuf_alloc_vertices( struct draw_stage *stage, assert(!vbuf->vertices); /* Allocate a new vertex buffer */ - vbuf->vertex_size = new_vertex_size; vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, (ushort) vbuf->vertex_size, -- cgit v1.2.3 From c3f10aef386e0af90f8735d8b9598959c17a590f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 28 Jan 2008 15:00:08 +0900 Subject: Simplify prototypes of draw_vbuf's internal functions. --- src/mesa/pipe/draw/draw_vbuf.c | 57 +++++++++++++++++++++--------------------- 1 file changed, 29 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 60ad93133c..1e260c6156 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -84,9 +84,9 @@ vbuf_stage( struct draw_stage *stage ) } -static void vbuf_flush_indices( struct draw_stage *stage ); -static void vbuf_flush_vertices( struct draw_stage *stage ); -static void vbuf_alloc_vertices( struct draw_stage *stage ); +static void vbuf_flush_indices( struct vbuf_stage *vbuf ); +static void vbuf_flush_vertices( struct vbuf_stage *vbuf ); +static void vbuf_alloc_vertices( struct vbuf_stage *vbuf ); static INLINE boolean @@ -101,12 +101,12 @@ static INLINE void check_space( struct vbuf_stage *vbuf, unsigned nr ) { if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { - vbuf_flush_vertices(&vbuf->stage); - vbuf_alloc_vertices(&vbuf->stage); + vbuf_flush_vertices(vbuf); + vbuf_alloc_vertices(vbuf); } if (vbuf->nr_indices + nr > vbuf->max_indices ) - vbuf_flush_indices(&vbuf->stage); + vbuf_flush_indices(vbuf); } @@ -249,9 +249,8 @@ vbuf_point( struct draw_stage *stage, * will be flushed if needed and a new one allocated. */ static void -vbuf_set_prim( struct draw_stage *stage, uint newprim ) +vbuf_set_prim( struct vbuf_stage *vbuf, uint newprim ) { - struct vbuf_stage *vbuf = vbuf_stage(stage); const struct vertex_info *vinfo; unsigned vertex_size; @@ -266,13 +265,13 @@ vbuf_set_prim( struct draw_stage *stage, uint newprim ) vertex_size = vinfo->size * sizeof(float); if (vertex_size != vbuf->vertex_size) - vbuf_flush_vertices(stage); + vbuf_flush_vertices(vbuf); vbuf->vinfo = vinfo; vbuf->vertex_size = vertex_size; if (!vbuf->vertices) - vbuf_alloc_vertices(stage); + vbuf_alloc_vertices(vbuf); } @@ -280,9 +279,11 @@ static void vbuf_first_tri( struct draw_stage *stage, struct prim_header *prim ) { - vbuf_flush_indices( stage ); + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); stage->tri = vbuf_tri; - vbuf_set_prim(stage, PIPE_PRIM_TRIANGLES); + vbuf_set_prim(vbuf, PIPE_PRIM_TRIANGLES); stage->tri( stage, prim ); } @@ -291,9 +292,11 @@ static void vbuf_first_line( struct draw_stage *stage, struct prim_header *prim ) { - vbuf_flush_indices( stage ); + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); stage->line = vbuf_line; - vbuf_set_prim(stage, PIPE_PRIM_LINES); + vbuf_set_prim(vbuf, PIPE_PRIM_LINES); stage->line( stage, prim ); } @@ -302,18 +305,18 @@ static void vbuf_first_point( struct draw_stage *stage, struct prim_header *prim ) { - vbuf_flush_indices( stage ); + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); stage->point = vbuf_point; - vbuf_set_prim(stage, PIPE_PRIM_POINTS); + vbuf_set_prim(vbuf, PIPE_PRIM_POINTS); stage->point( stage, prim ); } static void -vbuf_flush_indices( struct draw_stage *stage ) +vbuf_flush_indices( struct vbuf_stage *vbuf ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - if(!vbuf->nr_indices) return; @@ -354,12 +357,10 @@ vbuf_flush_indices( struct draw_stage *stage ) * we flush. */ static void -vbuf_flush_vertices( struct draw_stage *stage ) +vbuf_flush_vertices( struct vbuf_stage *vbuf ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - if(vbuf->vertices) { - vbuf_flush_indices(stage); + vbuf_flush_indices(vbuf); /* Reset temporary vertices ids */ if(vbuf->nr_vertices) @@ -378,10 +379,8 @@ vbuf_flush_vertices( struct draw_stage *stage ) static void -vbuf_alloc_vertices( struct draw_stage *stage ) +vbuf_alloc_vertices( struct vbuf_stage *vbuf ) { - struct vbuf_stage *vbuf = vbuf_stage( stage ); - assert(!vbuf->nr_indices); assert(!vbuf->vertices); @@ -398,14 +397,16 @@ vbuf_alloc_vertices( struct draw_stage *stage ) static void vbuf_flush( struct draw_stage *stage, unsigned flags ) { - vbuf_flush_indices( stage ); + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); stage->point = vbuf_first_point; stage->line = vbuf_first_line; stage->tri = vbuf_first_tri; if (flags & DRAW_FLUSH_BACKEND) - vbuf_flush_vertices( stage ); + vbuf_flush_vertices( vbuf ); } -- cgit v1.2.3 From 19caf4e4f0eae82ff5f36e5bc99463b6677467e6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 28 Jan 2008 15:48:51 +0900 Subject: Clone vf module. --- src/mesa/pipe/draw/draw_vf.c | 374 +++++++++++++ src/mesa/pipe/draw/draw_vf.h | 249 +++++++++ src/mesa/pipe/draw/draw_vf_generic.c | 983 +++++++++++++++++++++++++++++++++++ src/mesa/pipe/draw/draw_vf_sse.c | 664 +++++++++++++++++++++++ src/mesa/sources | 3 + 5 files changed, 2273 insertions(+) create mode 100644 src/mesa/pipe/draw/draw_vf.c create mode 100644 src/mesa/pipe/draw/draw_vf.h create mode 100644 src/mesa/pipe/draw/draw_vf_generic.c create mode 100644 src/mesa/pipe/draw/draw_vf_sse.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c new file mode 100644 index 0000000000..f758460b5f --- /dev/null +++ b/src/mesa/pipe/draw/draw_vf.c @@ -0,0 +1,374 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + +#include "glheader.h" +#include "context.h" +#include "colormac.h" + +#include "pipe/p_compiler.h" + +#include "draw_vf.h" + +#define DBG 0 + + + +static boolean match_fastpath( struct draw_vertex_fetch *vf, + const struct draw_vf_fastpath *fp) +{ + unsigned j; + + if (vf->attr_count != fp->attr_count) + return FALSE; + + for (j = 0; j < vf->attr_count; j++) + if (vf->attr[j].format != fp->attr[j].format || + vf->attr[j].inputsize != fp->attr[j].size || + vf->attr[j].vertoffset != fp->attr[j].offset) + return FALSE; + + if (fp->match_strides) { + if (vf->vertex_stride != fp->vertex_stride) + return FALSE; + + for (j = 0; j < vf->attr_count; j++) + if (vf->attr[j].inputstride != fp->attr[j].stride) + return FALSE; + } + + return TRUE; +} + +static boolean search_fastpath_emit( struct draw_vertex_fetch *vf ) +{ + struct draw_vf_fastpath *fp = vf->fastpath; + + for ( ; fp ; fp = fp->next) { + if (match_fastpath(vf, fp)) { + vf->emit = fp->func; + return TRUE; + } + } + + return FALSE; +} + +void draw_vf_register_fastpath( struct draw_vertex_fetch *vf, + boolean match_strides ) +{ + struct draw_vf_fastpath *fastpath = CALLOC_STRUCT(draw_vf_fastpath); + unsigned i; + + fastpath->vertex_stride = vf->vertex_stride; + fastpath->attr_count = vf->attr_count; + fastpath->match_strides = match_strides; + fastpath->func = vf->emit; + fastpath->attr = (struct draw_vf_attr_type *) + _mesa_malloc(vf->attr_count * sizeof(fastpath->attr[0])); + + for (i = 0; i < vf->attr_count; i++) { + fastpath->attr[i].format = vf->attr[i].format; + fastpath->attr[i].stride = vf->attr[i].inputstride; + fastpath->attr[i].size = vf->attr[i].inputsize; + fastpath->attr[i].offset = vf->attr[i].vertoffset; + } + + fastpath->next = vf->fastpath; + vf->fastpath = fastpath; +} + + + + +/*********************************************************************** + * Build codegen functions or return generic ones: + */ +static void choose_emit_func( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *dest) +{ + vf->emit = NULL; + + /* Does this match an existing (hardwired, codegen or known-bad) + * fastpath? + */ + if (search_fastpath_emit(vf)) { + /* Use this result. If it is null, then it is already known + * that the current state will fail for codegen and there is no + * point trying again. + */ + } + else if (vf->codegen_emit) { + vf->codegen_emit( vf ); + } + + if (!vf->emit) { + draw_vf_generate_hardwired_emit(vf); + } + + /* Otherwise use the generic version: + */ + if (!vf->emit) + vf->emit = draw_vf_generic_emit; + + vf->emit( vf, count, dest ); +} + + + + + +/*********************************************************************** + * Public entrypoints, mostly dispatch to the above: + */ + + + +unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, + const struct draw_vf_attr_map *map, + unsigned nr, + unsigned vertex_stride ) +{ + unsigned offset = 0; + unsigned i, j; + + assert(nr < DRAW_VF_ATTRIB_MAX); + + memset(vf->lookup, 0, sizeof(vf->lookup)); + + for (j = 0, i = 0; i < nr; i++) { + const unsigned format = map[i].format; + if (format == EMIT_PAD) { + if (DBG) + _mesa_printf("%d: pad %d, offset %d\n", i, + map[i].offset, offset); + + offset += map[i].offset; + + } + else { + assert(vf->lookup[map[i].attrib] == 0); + vf->lookup[map[i].attrib] = &vf->attr[j]; + + vf->attr[j].attrib = map[i].attrib; + vf->attr[j].format = format; + vf->attr[j].insert = draw_vf_format_info[format].insert; + vf->attr[j].extract = draw_vf_format_info[format].extract; + vf->attr[j].vertattrsize = draw_vf_format_info[format].attrsize; + vf->attr[j].vertoffset = offset; + + if (DBG) + _mesa_printf("%d: %s, offset %d\n", i, + draw_vf_format_info[format].name, + vf->attr[j].vertoffset); + + offset += draw_vf_format_info[format].attrsize; + j++; + } + } + + vf->attr_count = j; + vf->vertex_stride = vertex_stride ? vertex_stride : offset; + vf->emit = choose_emit_func; + + assert(vf->vertex_stride >= offset); + return vf->vertex_stride; +} + + + +void draw_vf_set_vp_matrix( struct draw_vertex_fetch *vf, + const float *viewport ) +{ + assert(vf->allow_viewport_emits); + + /* scale */ + vf->vp[0] = viewport[MAT_SX]; + vf->vp[1] = viewport[MAT_SY]; + vf->vp[2] = viewport[MAT_SZ]; + vf->vp[3] = 1.0; + + /* translate */ + vf->vp[4] = viewport[MAT_TX]; + vf->vp[5] = viewport[MAT_TY]; + vf->vp[6] = viewport[MAT_TZ]; + vf->vp[7] = 0.0; +} + +void draw_vf_set_vp_scale_translate( struct draw_vertex_fetch *vf, + const float *scale, + const float *translate ) +{ + assert(vf->allow_viewport_emits); + + vf->vp[0] = scale[0]; + vf->vp[1] = scale[1]; + vf->vp[2] = scale[2]; + vf->vp[3] = scale[3]; + + vf->vp[4] = translate[0]; + vf->vp[5] = translate[1]; + vf->vp[6] = translate[2]; + vf->vp[7] = translate[3]; +} + + +/* Set attribute pointers, adjusted for start position: + */ +void draw_vf_set_sources( struct draw_vertex_fetch *vf, + GLvector4f * const sources[], + unsigned start ) +{ + struct draw_vf_attr *a = vf->attr; + unsigned j; + + for (j = 0; j < vf->attr_count; j++) { + const GLvector4f *vptr = sources[a[j].attrib]; + + if ((a[j].inputstride != vptr->stride) || + (a[j].inputsize != vptr->size)) + vf->emit = choose_emit_func; + + a[j].inputstride = vptr->stride; + a[j].inputsize = vptr->size; + a[j].do_insert = a[j].insert[vptr->size - 1]; + a[j].inputptr = ((uint8_t *)vptr->data) + start * vptr->stride; + } +} + + + +/* Emit count VB vertices to dest. + */ +void draw_vf_emit_vertices( struct draw_vertex_fetch *vf, + unsigned count, + void *dest ) +{ + vf->emit( vf, count, (uint8_t*) dest ); +} + + +/* Extract a named attribute from a hardware vertex. Will have to + * reverse any viewport transformation, swizzling or other conversions + * which may have been applied. + * + * This is mainly required for on-the-fly vertex translations to + * swrast format. + */ +void draw_vf_get_attr( struct draw_vertex_fetch *vf, + const void *vertex, + GLenum attr, + const float *dflt, + float *dest ) +{ + const struct draw_vf_attr *a = vf->attr; + const unsigned attr_count = vf->attr_count; + unsigned j; + + for (j = 0; j < attr_count; j++) { + if (a[j].attrib == attr) { + a[j].extract( &a[j], dest, (uint8_t *)vertex + a[j].vertoffset ); + return; + } + } + + /* Else return the value from ctx->Current. + */ + _mesa_memcpy( dest, dflt, 4*sizeof(float)); +} + + + + +struct draw_vertex_fetch *draw_vf_create( boolean allow_viewport_emits ) +{ + struct draw_vertex_fetch *vf = CALLOC_STRUCT(draw_vertex_fetch); + unsigned i; + + for (i = 0; i < DRAW_VF_ATTRIB_MAX; i++) + vf->attr[i].vf = vf; + + vf->allow_viewport_emits = allow_viewport_emits; + + switch(CHAN_TYPE) { + case GL_UNSIGNED_BYTE: + vf->chan_scale[0] = 255.0; + vf->chan_scale[1] = 255.0; + vf->chan_scale[2] = 255.0; + vf->chan_scale[3] = 255.0; + break; + case GL_UNSIGNED_SHORT: + vf->chan_scale[0] = 65535.0; + vf->chan_scale[1] = 65535.0; + vf->chan_scale[2] = 65535.0; + vf->chan_scale[3] = 65535.0; + break; + default: + vf->chan_scale[0] = 1.0; + vf->chan_scale[1] = 1.0; + vf->chan_scale[2] = 1.0; + vf->chan_scale[3] = 1.0; + break; + } + + vf->identity[0] = 0.0; + vf->identity[1] = 0.0; + vf->identity[2] = 0.0; + vf->identity[3] = 1.0; + + vf->codegen_emit = NULL; + +#ifdef USE_SSE_ASM + if (!_mesa_getenv("MESA_NO_CODEGEN")) + vf->codegen_emit = draw_vf_generate_sse_emit; +#endif + + return vf; +} + + +void draw_vf_destroy( struct draw_vertex_fetch *vf ) +{ + struct draw_vf_fastpath *fp, *tmp; + + for (fp = vf->fastpath ; fp ; fp = tmp) { + tmp = fp->next; + FREE(fp->attr); + + /* KW: At the moment, fp->func is constrained to be allocated by + * _mesa_exec_alloc(), as the hardwired fastpaths in + * t_vertex_generic.c are handled specially. It would be nice + * to unify them, but this probably won't change until this + * module gets another overhaul. + */ + _mesa_exec_free((void *) fp->func); + FREE(fp); + } + + vf->fastpath = NULL; + FREE(vf); +} diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h new file mode 100644 index 0000000000..279570aad5 --- /dev/null +++ b/src/mesa/pipe/draw/draw_vf.h @@ -0,0 +1,249 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + +#ifndef DRAW_VF_H +#define DRAW_VF_H + + +#include "pipe/p_compiler.h" +#include "math/m_vector.h" + + +enum { + DRAW_VF_ATTRIB_POS = 0, + DRAW_VF_ATTRIB_WEIGHT = 1, + DRAW_VF_ATTRIB_NORMAL = 2, + DRAW_VF_ATTRIB_COLOR0 = 3, + DRAW_VF_ATTRIB_COLOR1 = 4, + DRAW_VF_ATTRIB_FOG = 5, + DRAW_VF_ATTRIB_COLOR_INDEX = 6, + DRAW_VF_ATTRIB_EDGEFLAG = 7, + DRAW_VF_ATTRIB_TEX0 = 8, + DRAW_VF_ATTRIB_TEX1 = 9, + DRAW_VF_ATTRIB_TEX2 = 10, + DRAW_VF_ATTRIB_TEX3 = 11, + DRAW_VF_ATTRIB_TEX4 = 12, + DRAW_VF_ATTRIB_TEX5 = 13, + DRAW_VF_ATTRIB_TEX6 = 14, + DRAW_VF_ATTRIB_TEX7 = 15, + DRAW_VF_ATTRIB_VAR0 = 16, + DRAW_VF_ATTRIB_VAR1 = 17, + DRAW_VF_ATTRIB_VAR2 = 18, + DRAW_VF_ATTRIB_VAR3 = 19, + DRAW_VF_ATTRIB_VAR4 = 20, + DRAW_VF_ATTRIB_VAR5 = 21, + DRAW_VF_ATTRIB_VAR6 = 22, + DRAW_VF_ATTRIB_VAR7 = 23, + DRAW_VF_ATTRIB_POINTSIZE = 24, + DRAW_VF_ATTRIB_BFC0 = 25, + DRAW_VF_ATTRIB_BFC1 = 26, + DRAW_VF_ATTRIB_CLIP_POS = 27, + DRAW_VF_ATTRIB_VERTEX_HEADER = 28, + DRAW_VF_ATTRIB_MAX = 29 +}; + +enum draw_vf_attr_format { + EMIT_1F, + EMIT_2F, + EMIT_3F, + EMIT_4F, + EMIT_2F_VIEWPORT, /**< do viewport transform and emit */ + EMIT_3F_VIEWPORT, /**< do viewport transform and emit */ + EMIT_4F_VIEWPORT, /**< do viewport transform and emit */ + EMIT_3F_XYW, /**< for projective texture */ + EMIT_1UB_1F, /**< for fog coordinate */ + EMIT_3UB_3F_RGB, /**< for specular color */ + EMIT_3UB_3F_BGR, /**< for specular color */ + EMIT_4UB_4F_RGBA, /**< for color */ + EMIT_4UB_4F_BGRA, /**< for color */ + EMIT_4UB_4F_ARGB, /**< for color */ + EMIT_4UB_4F_ABGR, /**< for color */ + EMIT_4CHAN_4F_RGBA, /**< for swrast color */ + EMIT_PAD, /**< leave a hole of 'offset' bytes */ + EMIT_MAX +}; + +struct draw_vf_attr_map { + unsigned attrib; + enum draw_vf_attr_format format; + unsigned offset; +}; + +struct draw_vertex_fetch; + + +void +draw_vf_set_vp_matrix( struct draw_vertex_fetch *vf, + const float *viewport ); + +void +draw_vf_set_vp_scale_translate( struct draw_vertex_fetch *vf, + const float *scale, + const float *translate ); + +unsigned +draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, + const struct draw_vf_attr_map *map, + unsigned nr, + unsigned vertex_stride ); + +void +draw_vf_set_sources( struct draw_vertex_fetch *vf, + GLvector4f * const attrib[], + unsigned start ); + +void +draw_vf_emit_vertices( struct draw_vertex_fetch *vf, + unsigned count, + void *dest ); + +void +draw_vf_get_attr( struct draw_vertex_fetch *vf, + const void *vertex, + GLenum attr, + const float *dflt, + float *dest ); + +struct draw_vertex_fetch * +draw_vf_create( boolean allow_viewport_emits ); + +void +draw_vf_destroy( struct draw_vertex_fetch *vf ); + + + +/*********************************************************************** + * Internal functions and structs: + */ + +struct draw_vf_attr; + +typedef void (*draw_vf_extract_func)( const struct draw_vf_attr *a, + float *out, + const uint8_t *v ); + +typedef void (*draw_vf_insert_func)( const struct draw_vf_attr *a, + uint8_t *v, + const float *in ); + +typedef void (*draw_vf_emit_func)( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *dest ); + + + +/** + * Describes how to convert/move a vertex attribute from a vertex + * array to a vertex structure. + */ +struct draw_vf_attr +{ + struct draw_vertex_fetch *vf; + + unsigned format; + unsigned inputsize; + unsigned inputstride; + unsigned vertoffset; /**< position of the attrib in the vertex struct */ + + unsigned attrib; /**< which vertex attrib (0=position, etc) */ + unsigned vertattrsize; /**< size of the attribute in bytes */ + + uint8_t *inputptr; + const draw_vf_insert_func *insert; + draw_vf_insert_func do_insert; + draw_vf_extract_func extract; +}; + +struct draw_vertex_fetch +{ + struct draw_vf_attr attr[DRAW_VF_ATTRIB_MAX]; + unsigned attr_count; + unsigned vertex_stride; + + struct draw_vf_attr *lookup[DRAW_VF_ATTRIB_MAX]; + + draw_vf_emit_func emit; + + /* Parameters and constants for codegen: + */ + boolean allow_viewport_emits; + float vp[8]; + float chan_scale[4]; + float identity[4]; + + struct draw_vf_fastpath *fastpath; + + void (*codegen_emit)( struct draw_vertex_fetch *vf ); +}; + + +struct draw_vf_attr_type { + unsigned format; + unsigned size; + unsigned stride; + unsigned offset; +}; + +struct draw_vf_fastpath { + unsigned vertex_stride; + unsigned attr_count; + boolean match_strides; + + struct draw_vf_attr_type *attr; + + draw_vf_emit_func func; + struct draw_vf_fastpath *next; +}; + + +void +draw_vf_register_fastpath( struct draw_vertex_fetch *vtx, + boolean match_strides ); + +void +draw_vf_generic_emit( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *v ); + +void +draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ); + +void +draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ); + + +struct draw_vf_format_info { + const char *name; + draw_vf_extract_func extract; + draw_vf_insert_func insert[4]; + const unsigned attrsize; +}; + +const struct draw_vf_format_info draw_vf_format_info[EMIT_MAX]; + + +#endif diff --git a/src/mesa/pipe/draw/draw_vf_generic.c b/src/mesa/pipe/draw/draw_vf_generic.c new file mode 100644 index 0000000000..19e6c587e5 --- /dev/null +++ b/src/mesa/pipe/draw/draw_vf_generic.c @@ -0,0 +1,983 @@ + +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + +#include "glheader.h" +#include "context.h" +#include "colormac.h" +#include "simple_list.h" + +#include "pipe/p_compiler.h" + +#include "draw_vf.h" + + +/* + * These functions take the NDC coordinates pointed to by 'in', apply the + * NDC->Viewport mapping and store the results at 'v'. + */ + +static INLINE void insert_4f_viewport_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + float *out = (float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = scale[2] * in[2] + trans[2]; + out[3] = in[3]; +} + +static INLINE void insert_4f_viewport_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + float *out = (float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = scale[2] * in[2] + trans[2]; + out[3] = 1; +} + +static INLINE void insert_4f_viewport_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + float *out = (float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = trans[2]; + out[3] = 1; +} + +static INLINE void insert_4f_viewport_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + float *out = (float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = trans[1]; + out[2] = trans[2]; + out[3] = 1; +} + +static INLINE void insert_3f_viewport_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + float *out = (float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = scale[2] * in[2] + trans[2]; +} + +static INLINE void insert_3f_viewport_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + float *out = (float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; + out[2] = scale[2] * in[2] + trans[2]; +} + +static INLINE void insert_3f_viewport_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + float *out = (float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = trans[1]; + out[2] = trans[2]; +} + +static INLINE void insert_2f_viewport_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + float *out = (float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = scale[1] * in[1] + trans[1]; +} + +static INLINE void insert_2f_viewport_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + float *out = (float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = scale[0] * in[0] + trans[0]; + out[1] = trans[1]; +} + + +/* + * These functions do the same as above, except for the viewport mapping. + */ + +static INLINE void insert_4f_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = in[3]; +} + +static INLINE void insert_4f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = 1; +} + +static INLINE void insert_4f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; + out[3] = 1; +} + +static INLINE void insert_4f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; + out[3] = 1; +} + +static INLINE void insert_3f_xyw_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[3]; +} + +static INLINE void insert_3f_xyw_err( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + (void) a; (void) v; (void) in; + _mesa_exit(1); +} + +static INLINE void insert_3f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; +} + +static INLINE void insert_3f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; +} + +static INLINE void insert_3f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; +} + + +static INLINE void insert_2f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; +} + +static INLINE void insert_2f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; +} + +static INLINE void insert_1f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; +} + +static INLINE void insert_null( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + (void) a; (void) v; (void) in; +} + +static INLINE void insert_4chan_4f_rgba_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + GLchan *c = (GLchan *)v; + (void) a; + UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); + UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); + UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); + UNCLAMPED_FLOAT_TO_CHAN(c[3], in[3]); +} + +static INLINE void insert_4chan_4f_rgba_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + GLchan *c = (GLchan *)v; + (void) a; + UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); + UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); + UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); + c[3] = CHAN_MAX; +} + +static INLINE void insert_4chan_4f_rgba_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + GLchan *c = (GLchan *)v; + (void) a; + UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); + UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); + c[2] = 0; + c[3] = CHAN_MAX; +} + +static INLINE void insert_4chan_4f_rgba_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + GLchan *c = (GLchan *)v; + (void) a; + UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); + c[1] = 0; + c[2] = 0; + c[3] = CHAN_MAX; +} + +static INLINE void insert_4ub_4f_rgba_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); +} + +static INLINE void insert_4ub_4f_rgba_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_rgba_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[2] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_rgba_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + v[1] = 0; + v[2] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); +} + +static INLINE void insert_4ub_4f_bgra_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[0] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + v[1] = 0; + v[0] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); +} + +static INLINE void insert_4ub_4f_argb_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + v[3] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + v[2] = 0x00; + v[3] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); +} + +static INLINE void insert_4ub_4f_abgr_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + v[1] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + v[2] = 0x00; + v[1] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_3ub_3f_rgb_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); +} + +static INLINE void insert_3ub_3f_rgb_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[2] = 0; +} + +static INLINE void insert_3ub_3f_rgb_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + v[1] = 0; + v[2] = 0; +} + +static INLINE void insert_3ub_3f_bgr_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); +} + +static INLINE void insert_3ub_3f_bgr_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[0] = 0; +} + +static INLINE void insert_3ub_3f_bgr_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + v[1] = 0; + v[0] = 0; +} + + +static INLINE void insert_1ub_1f_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); +} + + +/*********************************************************************** + * Functions to perform the reverse operations to the above, for + * swrast translation and clip-interpolation. + * + * Currently always extracts a full 4 floats. + */ + +static void extract_4f_viewport( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + const float *in = (const float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + /* Although included for completeness, the position coordinate is + * usually handled differently during clipping. + */ + out[0] = (in[0] - trans[0]) / scale[0]; + out[1] = (in[1] - trans[1]) / scale[1]; + out[2] = (in[2] - trans[2]) / scale[2]; + out[3] = in[3]; +} + +static void extract_3f_viewport( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + const float *in = (const float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = (in[0] - trans[0]) / scale[0]; + out[1] = (in[1] - trans[1]) / scale[1]; + out[2] = (in[2] - trans[2]) / scale[2]; + out[3] = 1; +} + + +static void extract_2f_viewport( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + const float *in = (const float *)v; + const float *scale = a->vf->vp; + const float *trans = a->vf->vp + 4; + + out[0] = (in[0] - trans[0]) / scale[0]; + out[1] = (in[1] - trans[1]) / scale[1]; + out[2] = 0; + out[3] = 1; +} + + +static void extract_4f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) +{ + const float *in = (const float *)v; + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = in[3]; +} + +static void extract_3f_xyw( const struct draw_vf_attr *a, float *out, const uint8_t *v ) +{ + const float *in = (const float *)v; + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; + out[3] = in[2]; +} + + +static void extract_3f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) +{ + const float *in = (const float *)v; + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = 1; +} + + +static void extract_2f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) +{ + const float *in = (const float *)v; + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; + out[3] = 1; +} + +static void extract_1f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) +{ + const float *in = (const float *)v; + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; + out[3] = 1; +} + +static void extract_4chan_4f_rgba( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + GLchan *c = (GLchan *)v; + (void) a; + + out[0] = CHAN_TO_FLOAT(c[0]); + out[1] = CHAN_TO_FLOAT(c[1]); + out[2] = CHAN_TO_FLOAT(c[2]); + out[3] = CHAN_TO_FLOAT(c[3]); +} + +static void extract_4ub_4f_rgba( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + (void) a; + out[0] = UBYTE_TO_FLOAT(v[0]); + out[1] = UBYTE_TO_FLOAT(v[1]); + out[2] = UBYTE_TO_FLOAT(v[2]); + out[3] = UBYTE_TO_FLOAT(v[3]); +} + +static void extract_4ub_4f_bgra( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + (void) a; + out[2] = UBYTE_TO_FLOAT(v[0]); + out[1] = UBYTE_TO_FLOAT(v[1]); + out[0] = UBYTE_TO_FLOAT(v[2]); + out[3] = UBYTE_TO_FLOAT(v[3]); +} + +static void extract_4ub_4f_argb( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + (void) a; + out[3] = UBYTE_TO_FLOAT(v[0]); + out[0] = UBYTE_TO_FLOAT(v[1]); + out[1] = UBYTE_TO_FLOAT(v[2]); + out[2] = UBYTE_TO_FLOAT(v[3]); +} + +static void extract_4ub_4f_abgr( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + (void) a; + out[3] = UBYTE_TO_FLOAT(v[0]); + out[2] = UBYTE_TO_FLOAT(v[1]); + out[1] = UBYTE_TO_FLOAT(v[2]); + out[0] = UBYTE_TO_FLOAT(v[3]); +} + +static void extract_3ub_3f_rgb( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + (void) a; + out[0] = UBYTE_TO_FLOAT(v[0]); + out[1] = UBYTE_TO_FLOAT(v[1]); + out[2] = UBYTE_TO_FLOAT(v[2]); + out[3] = 1; +} + +static void extract_3ub_3f_bgr( const struct draw_vf_attr *a, float *out, + const uint8_t *v ) +{ + (void) a; + out[2] = UBYTE_TO_FLOAT(v[0]); + out[1] = UBYTE_TO_FLOAT(v[1]); + out[0] = UBYTE_TO_FLOAT(v[2]); + out[3] = 1; +} + +static void extract_1ub_1f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) +{ + (void) a; + out[0] = UBYTE_TO_FLOAT(v[0]); + out[1] = 0; + out[2] = 0; + out[3] = 1; +} + + +const struct draw_vf_format_info draw_vf_format_info[EMIT_MAX] = +{ + { "1f", + extract_1f, + { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, + sizeof(float) }, + + { "2f", + extract_2f, + { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, + 2 * sizeof(float) }, + + { "3f", + extract_3f, + { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, + 3 * sizeof(float) }, + + { "4f", + extract_4f, + { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, + 4 * sizeof(float) }, + + { "2f_viewport", + extract_2f_viewport, + { insert_2f_viewport_1, insert_2f_viewport_2, insert_2f_viewport_2, + insert_2f_viewport_2 }, + 2 * sizeof(float) }, + + { "3f_viewport", + extract_3f_viewport, + { insert_3f_viewport_1, insert_3f_viewport_2, insert_3f_viewport_3, + insert_3f_viewport_3 }, + 3 * sizeof(float) }, + + { "4f_viewport", + extract_4f_viewport, + { insert_4f_viewport_1, insert_4f_viewport_2, insert_4f_viewport_3, + insert_4f_viewport_4 }, + 4 * sizeof(float) }, + + { "3f_xyw", + extract_3f_xyw, + { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, + insert_3f_xyw_4 }, + 3 * sizeof(float) }, + + { "1ub_1f", + extract_1ub_1f, + { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 }, + sizeof(uint8_t) }, + + { "3ub_3f_rgb", + extract_3ub_3f_rgb, + { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3, + insert_3ub_3f_rgb_3 }, + 3 * sizeof(uint8_t) }, + + { "3ub_3f_bgr", + extract_3ub_3f_bgr, + { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3, + insert_3ub_3f_bgr_3 }, + 3 * sizeof(uint8_t) }, + + { "4ub_4f_rgba", + extract_4ub_4f_rgba, + { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, + insert_4ub_4f_rgba_4 }, + 4 * sizeof(uint8_t) }, + + { "4ub_4f_bgra", + extract_4ub_4f_bgra, + { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3, + insert_4ub_4f_bgra_4 }, + 4 * sizeof(uint8_t) }, + + { "4ub_4f_argb", + extract_4ub_4f_argb, + { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3, + insert_4ub_4f_argb_4 }, + 4 * sizeof(uint8_t) }, + + { "4ub_4f_abgr", + extract_4ub_4f_abgr, + { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3, + insert_4ub_4f_abgr_4 }, + 4 * sizeof(uint8_t) }, + + { "4chan_4f_rgba", + extract_4chan_4f_rgba, + { insert_4chan_4f_rgba_1, insert_4chan_4f_rgba_2, insert_4chan_4f_rgba_3, + insert_4chan_4f_rgba_4 }, + 4 * sizeof(GLchan) }, + + { "pad", + NULL, + { NULL, NULL, NULL, NULL }, + 0 } + +}; + + + + +/*********************************************************************** + * Hardwired fastpaths for emitting whole vertices or groups of + * vertices + */ +#define EMIT5(NR, F0, F1, F2, F3, F4, NAME) \ +static void NAME( struct draw_vertex_fetch *vf, \ + unsigned count, \ + uint8_t *v ) \ +{ \ + struct draw_vf_attr *a = vf->attr; \ + unsigned i; \ + \ + for (i = 0 ; i < count ; i++, v += vf->vertex_stride) { \ + if (NR > 0) { \ + F0( &a[0], v + a[0].vertoffset, (float *)a[0].inputptr ); \ + a[0].inputptr += a[0].inputstride; \ + } \ + \ + if (NR > 1) { \ + F1( &a[1], v + a[1].vertoffset, (float *)a[1].inputptr ); \ + a[1].inputptr += a[1].inputstride; \ + } \ + \ + if (NR > 2) { \ + F2( &a[2], v + a[2].vertoffset, (float *)a[2].inputptr ); \ + a[2].inputptr += a[2].inputstride; \ + } \ + \ + if (NR > 3) { \ + F3( &a[3], v + a[3].vertoffset, (float *)a[3].inputptr ); \ + a[3].inputptr += a[3].inputstride; \ + } \ + \ + if (NR > 4) { \ + F4( &a[4], v + a[4].vertoffset, (float *)a[4].inputptr ); \ + a[4].inputptr += a[4].inputstride; \ + } \ + } \ +} + + +#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \ + insert_null, insert_null, NAME) + +#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \ + insert_null, NAME) + +#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \ + insert_null, NAME) + + +EMIT2(insert_3f_viewport_3, insert_4ub_4f_rgba_4, emit_viewport3_rgba4) +EMIT2(insert_3f_viewport_3, insert_4ub_4f_bgra_4, emit_viewport3_bgra4) +EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4) + +EMIT3(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_viewport4_rgba4_st2) +EMIT3(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2, emit_viewport4_bgra4_st2) +EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2) + +EMIT4(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_viewport4_rgba4_st2_st2) +EMIT4(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2, insert_2f_2, emit_viewport4_bgra4_st2_st2) +EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2) + + +/* Use the codegen paths to select one of a number of hardwired + * fastpaths. + */ +void draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ) +{ + draw_vf_emit_func func = NULL; + + /* Does it fit a hardwired fastpath? Help! this is growing out of + * control! + */ + switch (vf->attr_count) { + case 2: + if (vf->attr[0].do_insert == insert_3f_viewport_3) { + if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4) + func = emit_viewport3_bgra4; + else if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) + func = emit_viewport3_rgba4; + } + else if (vf->attr[0].do_insert == insert_3f_3 && + vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + func = emit_xyz3_rgba4; + } + break; + case 3: + if (vf->attr[2].do_insert == insert_2f_2) { + if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + if (vf->attr[0].do_insert == insert_4f_viewport_4) + func = emit_viewport4_rgba4_st2; + else if (vf->attr[0].do_insert == insert_4f_4) + func = emit_xyzw4_rgba4_st2; + } + else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 && + vf->attr[0].do_insert == insert_4f_viewport_4) + func = emit_viewport4_bgra4_st2; + } + break; + case 4: + if (vf->attr[2].do_insert == insert_2f_2 && + vf->attr[3].do_insert == insert_2f_2) { + if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + if (vf->attr[0].do_insert == insert_4f_viewport_4) + func = emit_viewport4_rgba4_st2_st2; + else if (vf->attr[0].do_insert == insert_4f_4) + func = emit_xyzw4_rgba4_st2_st2; + } + else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 && + vf->attr[0].do_insert == insert_4f_viewport_4) + func = emit_viewport4_bgra4_st2_st2; + } + break; + } + + vf->emit = func; +} + +/*********************************************************************** + * Generic (non-codegen) functions for whole vertices or groups of + * vertices + */ + +void draw_vf_generic_emit( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *v ) +{ + struct draw_vf_attr *a = vf->attr; + const unsigned attr_count = vf->attr_count; + const unsigned stride = vf->vertex_stride; + unsigned i, j; + + for (i = 0 ; i < count ; i++, v += stride) { + for (j = 0; j < attr_count; j++) { + float *in = (float *)a[j].inputptr; + a[j].inputptr += a[j].inputstride; + a[j].do_insert( &a[j], v + a[j].vertoffset, in ); + } + } +} + + diff --git a/src/mesa/pipe/draw/draw_vf_sse.c b/src/mesa/pipe/draw/draw_vf_sse.c new file mode 100644 index 0000000000..2cf3a45ff9 --- /dev/null +++ b/src/mesa/pipe/draw/draw_vf_sse.c @@ -0,0 +1,664 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + +#include "glheader.h" +#include "colormac.h" +#include "simple_list.h" +#include "enums.h" + +#include "pipe/p_compiler.h" + +#include "draw_vf.h" + +#if defined(USE_SSE_ASM) + +#include "x86/rtasm/x86sse.h" +#include "x86/common_x86_asm.h" + + +#define X 0 +#define Y 1 +#define Z 2 +#define W 3 + + +struct x86_program { + struct x86_function func; + + struct draw_vertex_fetch *vf; + boolean inputs_safe; + boolean outputs_safe; + boolean have_sse2; + + struct x86_reg identity; + struct x86_reg chan0; +}; + + +static struct x86_reg get_identity( struct x86_program *p ) +{ + return p->identity; +} + +static void emit_load4f_4( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movups(&p->func, dest, arg0); +} + +static void emit_load4f_3( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Have to jump through some hoops: + * + * c 0 0 0 + * c 0 0 1 + * 0 0 c 1 + * a b c 1 + */ + sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); + sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); + sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) ); + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load4f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Initialize from identity, then pull in low two words: + */ + sse_movups(&p->func, dest, get_identity(p)); + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load4f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Pull in low word, then swizzle in identity */ + sse_movss(&p->func, dest, arg0); + sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); +} + + + +static void emit_load3f_3( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Over-reads by 1 dword - potential SEGV if input is a vertex + * array. + */ + if (p->inputs_safe) { + sse_movups(&p->func, dest, arg0); + } + else { + /* c 0 0 0 + * c c c c + * a b c c + */ + sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); + sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X)); + sse_movlps(&p->func, dest, arg0); + } +} + +static void emit_load3f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_2(p, dest, arg0); +} + +static void emit_load3f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_1(p, dest, arg0); +} + +static void emit_load2f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load2f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_1(p, dest, arg0); +} + +static void emit_load1f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movss(&p->func, dest, arg0); +} + +static void (*load[4][4])( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) = { + { emit_load1f_1, + emit_load1f_1, + emit_load1f_1, + emit_load1f_1 }, + + { emit_load2f_1, + emit_load2f_2, + emit_load2f_2, + emit_load2f_2 }, + + { emit_load3f_1, + emit_load3f_2, + emit_load3f_3, + emit_load3f_3 }, + + { emit_load4f_1, + emit_load4f_2, + emit_load4f_3, + emit_load4f_4 } +}; + +static void emit_load( struct x86_program *p, + struct x86_reg dest, + unsigned sz, + struct x86_reg src, + unsigned src_sz) +{ + load[sz-1][src_sz-1](p, dest, src); +} + +static void emit_store4f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movups(&p->func, dest, arg0); +} + +static void emit_store3f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + if (p->outputs_safe) { + /* Emit the extra dword anyway. This may hurt writecombining, + * may cause other problems. + */ + sse_movups(&p->func, dest, arg0); + } + else { + /* Alternate strategy - emit two, shuffle, emit one. + */ + sse_movlps(&p->func, dest, arg0); + sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */ + sse_movss(&p->func, x86_make_disp(dest,8), arg0); + } +} + +static void emit_store2f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movlps(&p->func, dest, arg0); +} + +static void emit_store1f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movss(&p->func, dest, arg0); +} + + +static void (*store[4])( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) = +{ + emit_store1f, + emit_store2f, + emit_store3f, + emit_store4f +}; + +static void emit_store( struct x86_program *p, + struct x86_reg dest, + unsigned sz, + struct x86_reg temp ) + +{ + store[sz-1](p, dest, temp); +} + +static void emit_pack_store_4ub( struct x86_program *p, + struct x86_reg dest, + struct x86_reg temp ) +{ + /* Scale by 255.0 + */ + sse_mulps(&p->func, temp, p->chan0); + + if (p->have_sse2) { + sse2_cvtps2dq(&p->func, temp, temp); + sse2_packssdw(&p->func, temp, temp); + sse2_packuswb(&p->func, temp, temp); + sse_movss(&p->func, dest, temp); + } + else { + struct x86_reg mmx0 = x86_make_reg(file_MMX, 0); + struct x86_reg mmx1 = x86_make_reg(file_MMX, 1); + sse_cvtps2pi(&p->func, mmx0, temp); + sse_movhlps(&p->func, temp, temp); + sse_cvtps2pi(&p->func, mmx1, temp); + mmx_packssdw(&p->func, mmx0, mmx1); + mmx_packuswb(&p->func, mmx0, mmx0); + mmx_movd(&p->func, dest, mmx0); + } +} + +static int get_offset( const void *a, const void *b ) +{ + return (const char *)b - (const char *)a; +} + +/* Not much happens here. Eventually use this function to try and + * avoid saving/reloading the source pointers each vertex (if some of + * them can fit in registers). + */ +static void get_src_ptr( struct x86_program *p, + struct x86_reg srcREG, + struct x86_reg vfREG, + struct draw_vf_attr *a ) +{ + struct draw_vertex_fetch *vf = p->vf; + struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); + + /* Load current a[j].inputptr + */ + x86_mov(&p->func, srcREG, ptr_to_src); +} + +static void update_src_ptr( struct x86_program *p, + struct x86_reg srcREG, + struct x86_reg vfREG, + struct draw_vf_attr *a ) +{ + if (a->inputstride) { + struct draw_vertex_fetch *vf = p->vf; + struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); + + /* add a[j].inputstride (hardcoded value - could just as easily + * pull the stride value from memory each time). + */ + x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride)); + + /* save new value of a[j].inputptr + */ + x86_mov(&p->func, ptr_to_src, srcREG); + } +} + + +/* Lots of hardcoding + * + * EAX -- pointer to current output vertex + * ECX -- pointer to current attribute + * + */ +static boolean build_vertex_emit( struct x86_program *p ) +{ + struct draw_vertex_fetch *vf = p->vf; + unsigned j = 0; + + struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX); + struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX); + struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); + struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI); + struct x86_reg temp = x86_make_reg(file_XMM, 0); + struct x86_reg vp0 = x86_make_reg(file_XMM, 1); + struct x86_reg vp1 = x86_make_reg(file_XMM, 2); + uint8_t *fixup, *label; + + /* Push a few regs? + */ + x86_push(&p->func, countEBP); + x86_push(&p->func, vfESI); + + + /* Get vertex count, compare to zero + */ + x86_xor(&p->func, srcECX, srcECX); + x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2)); + x86_cmp(&p->func, countEBP, srcECX); + fixup = x86_jcc_forward(&p->func, cc_E); + + /* Initialize destination register. + */ + x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3)); + + /* Move argument 1 (vf) into a reg: + */ + x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1)); + + + /* Possibly load vp0, vp1 for viewport calcs: + */ + if (vf->allow_viewport_emits) { + sse_movups(&p->func, vp0, x86_make_disp(vfESI, get_offset(vf, &vf->vp[0]))); + sse_movups(&p->func, vp1, x86_make_disp(vfESI, get_offset(vf, &vf->vp[4]))); + } + + /* always load, needed or not: + */ + sse_movups(&p->func, p->chan0, x86_make_disp(vfESI, get_offset(vf, &vf->chan_scale[0]))); + sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0]))); + + /* Note address for loop jump */ + label = x86_get_label(&p->func); + + /* Emit code for each of the attributes. Currently routes + * everything through SSE registers, even when it might be more + * efficient to stick with regular old x86. No optimization or + * other tricks - enough new ground to cover here just getting + * things working. + */ + while (j < vf->attr_count) { + struct draw_vf_attr *a = &vf->attr[j]; + struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset); + + /* Now, load an XMM reg from src, perhaps transform, then save. + * Could be shortcircuited in specific cases: + */ + switch (a->format) { + case EMIT_1F: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 1, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_2F: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 2, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_3F: + /* Potentially the worst case - hardcode 2+1 copying: + */ + if (0) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + } + else { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 2, temp); + if (a->inputsize > 2) { + emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1); + emit_store(p, x86_make_disp(dest,8), 1, temp); + } + else { + sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p)); + } + update_src_ptr(p, srcECX, vfESI, a); + } + break; + case EMIT_4F: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 4, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_2F_VIEWPORT: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + sse_mulps(&p->func, temp, vp0); + sse_addps(&p->func, temp, vp1); + emit_store(p, dest, 2, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_3F_VIEWPORT: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + sse_mulps(&p->func, temp, vp0); + sse_addps(&p->func, temp, vp1); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4F_VIEWPORT: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_mulps(&p->func, temp, vp0); + sse_addps(&p->func, temp, vp1); + emit_store(p, dest, 4, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_3F_XYW: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z)); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + + case EMIT_1UB_1F: + /* Test for PAD3 + 1UB: + */ + if (j > 0 && + a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3) + { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X)); + emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */ + update_src_ptr(p, srcECX, vfESI, a); + } + else { + _mesa_printf("Can't emit 1ub %x %x %d\n", a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); + return FALSE; + } + break; + case EMIT_3UB_3F_RGB: + case EMIT_3UB_3F_BGR: + /* Test for 3UB + PAD1: + */ + if (j == vf->attr_count - 1 || + a[1].vertoffset >= a->vertoffset + 4) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + if (a->format == EMIT_3UB_3F_BGR) + sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + } + /* Test for 3UB + 1UB: + */ + else if (j < vf->attr_count - 1 && + a[1].format == EMIT_1UB_1F && + a[1].vertoffset == a->vertoffset + 3) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + update_src_ptr(p, srcECX, vfESI, a); + + /* Make room for incoming value: + */ + sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); + + get_src_ptr(p, srcECX, vfESI, &a[1]); + emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize); + update_src_ptr(p, srcECX, vfESI, &a[1]); + + /* Rearrange and possibly do BGR conversion: + */ + if (a->format == EMIT_3UB_3F_BGR) + sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); + else + sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X)); + + emit_pack_store_4ub(p, dest, temp); + j++; /* NOTE: two attrs consumed */ + } + else { + _mesa_printf("Can't emit 3ub\n"); + } + return FALSE; /* add this later */ + break; + + case EMIT_4UB_4F_RGBA: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4UB_4F_BGRA: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4UB_4F_ARGB: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4UB_4F_ABGR: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case EMIT_4CHAN_4F_RGBA: + switch (CHAN_TYPE) { + case GL_UNSIGNED_BYTE: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case GL_FLOAT: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 4, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case GL_UNSIGNED_SHORT: + default: + _mesa_printf("unknown CHAN_TYPE %s\n", _mesa_lookup_enum_by_nr(CHAN_TYPE)); + return FALSE; + } + break; + default: + _mesa_printf("unknown a[%d].format %d\n", j, a->format); + return FALSE; /* catch any new opcodes */ + } + + /* Increment j by at least 1 - may have been incremented above also: + */ + j++; + } + + /* Next vertex: + */ + x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vf->vertex_stride)); + + /* decr count, loop if not zero + */ + x86_dec(&p->func, countEBP); + x86_test(&p->func, countEBP, countEBP); + x86_jcc(&p->func, cc_NZ, label); + + /* Exit mmx state? + */ + if (p->func.need_emms) + mmx_emms(&p->func); + + /* Land forward jump here: + */ + x86_fixup_fwd_jump(&p->func, fixup); + + /* Pop regs and return + */ + x86_pop(&p->func, x86_get_base_reg(vfESI)); + x86_pop(&p->func, countEBP); + x86_ret(&p->func); + + vf->emit = (draw_vf_emit_func)x86_get_func(&p->func); + return TRUE; +} + + + +void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) +{ + struct x86_program p; + + if (!cpu_has_xmm) { + vf->codegen_emit = NULL; + return; + } + + _mesa_memset(&p, 0, sizeof(p)); + + p.vf = vf; + p.inputs_safe = 0; /* for now */ + p.outputs_safe = 1; /* for now */ + p.have_sse2 = cpu_has_xmm2; + p.identity = x86_make_reg(file_XMM, 6); + p.chan0 = x86_make_reg(file_XMM, 7); + + x86_init_func(&p.func); + + if (build_vertex_emit(&p)) { + draw_vf_register_fastpath( vf, TRUE ); + } + else { + /* Note the failure so that we don't keep trying to codegen an + * impossible state: + */ + draw_vf_register_fastpath( vf, FALSE ); + x86_release_func(&p.func); + } +} + +#else + +void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) +{ + /* Dummy version for when USE_SSE_ASM not defined */ +} + +#endif diff --git a/src/mesa/sources b/src/mesa/sources index 97ef7e1936..e31d8cc466 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -175,6 +175,9 @@ DRAW_SOURCES = \ pipe/draw/draw_vertex_fetch.c \ pipe/draw/draw_vertex_shader.c \ pipe/draw/draw_vertex_shader_llvm.c \ + pipe/draw/draw_vf.c \ + pipe/draw/draw_vf_generic.c \ + pipe/draw/draw_vf_sse.c \ pipe/draw/draw_wide_prims.c TGSIEXEC_SOURCES = \ -- cgit v1.2.3 From 09059259bed779360158664625e41a67f7496a74 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 28 Jan 2008 18:46:21 +0900 Subject: First stab at hooking draw_vbuf & vf. Emit disabled for now. Tested with softpipe. Only one vertex at a time for now (slow). --- src/mesa/pipe/draw/draw_vbuf.c | 183 ++++++++++++++++++++++++++++++++++- src/mesa/pipe/draw/draw_vf.c | 18 +++- src/mesa/pipe/draw/draw_vf.h | 46 +++++---- src/mesa/pipe/draw/draw_vf_generic.c | 2 +- src/mesa/pipe/draw/draw_vf_sse.c | 38 ++++---- 5 files changed, 241 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 1e260c6156..a3d0b5bca3 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -35,12 +35,15 @@ #include +#include -#include "pipe/draw/draw_vbuf.h" -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vertex.h" #include "pipe/p_util.h" +#include "draw_vbuf.h" +#include "draw_private.h" +#include "draw_vertex.h" +#include "draw_vf.h" + /** * Vertex buffer emit stage. @@ -55,6 +58,8 @@ struct vbuf_stage { /** Vertex size in bytes */ unsigned vertex_size; + struct draw_vertex_fetch *vf; + /* FIXME: we have no guarantee that 'unsigned' is 32bit */ /** Vertices in hardware format */ @@ -121,6 +126,7 @@ static INLINE void emit_vertex( struct vbuf_stage *vbuf, struct vertex_header *vertex ) { +#if 0 const struct vertex_info *vinfo = vbuf->vinfo; uint i; @@ -151,9 +157,11 @@ emit_vertex( struct vbuf_stage *vbuf, case EMIT_ALL: /* just copy the whole vertex as-is to the vbuf */ assert(i == 0); + assert(j == 0); memcpy(vbuf->vertex_ptr, vertex, vinfo->size * 4); vbuf->vertex_ptr += vinfo->size; - return; + count += vinfo->size; + break; case EMIT_1F: *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); count++; @@ -192,6 +200,156 @@ emit_vertex( struct vbuf_stage *vbuf, } } assert(count == vinfo->size); +#else + if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { + if(vertex->vertex_id < vbuf->nr_vertices) + return; + else + fprintf(stderr, "Bad vertex id 0x%04x (>= 0x%04x)\n", + vertex->vertex_id, vbuf->nr_vertices); + return; + } + + vertex->vertex_id = vbuf->nr_vertices++; + + draw_vf_set_data(vbuf->vf, vertex->data); + draw_vf_emit_vertices(vbuf->vf, 1, vbuf->vertex_ptr); + + vbuf->vertex_ptr += vbuf->vertex_size/4; +#endif +} + + +static void +vbuf_set_vf_attributes(struct vbuf_stage *vbuf ) +{ + const struct vertex_info *vinfo = vbuf->vinfo; + struct draw_vf_attr_map attrs[PIPE_MAX_SHADER_INPUTS]; + uint i; + uint count = 0; /* for debug/sanity */ + unsigned nr_attrs = 0; + +// fprintf(stderr, "emit vertex %d to %p\n", +// vbuf->nr_vertices, vbuf->vertex_ptr); + +#if 0 + if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { + if(vertex->vertex_id < vbuf->nr_vertices) + return; + else + fprintf(stderr, "Bad vertex id 0x%04x (>= 0x%04x)\n", + vertex->vertex_id, vbuf->nr_vertices); + return; + } +#endif + + for (i = 0; i < vinfo->num_attribs; i++) { + uint j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + /* no-op */ + break; + case EMIT_ALL: { + /* just copy the whole vertex as-is to the vbuf */ + unsigned k, s = vinfo->size; + assert(i == 0); + assert(j == 0); + /* copy the vertex header */ + /* XXX: we actually don't copy the header, just pad it */ + attrs[nr_attrs].attrib = 0; + attrs[nr_attrs].format = DRAW_EMIT_PAD; + attrs[nr_attrs].offset = offsetof(struct vertex_header, data); + s -= offsetof(struct vertex_header, data)/4; + count += offsetof(struct vertex_header, data)/4; + nr_attrs++; + /* copy the vertex data */ + for(k = 0; k < (s & ~0x3); k += 4) { + attrs[nr_attrs].attrib = k/4; + attrs[nr_attrs].format = DRAW_EMIT_4F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 4; + } + /* tail */ + /* XXX: actually, this shouldn't be needed */ + attrs[nr_attrs].attrib = k/4; + attrs[nr_attrs].offset = 0; + switch(s & 0x3) { + case 0: + break; + case 1: + attrs[nr_attrs].format = DRAW_EMIT_1F; + nr_attrs++; + count += 1; + break; + case 2: + attrs[nr_attrs].format = DRAW_EMIT_2F; + nr_attrs++; + count += 2; + break; + case 3: + attrs[nr_attrs].format = DRAW_EMIT_3F; + nr_attrs++; + count += 3; + break; + } + break; + } + case EMIT_1F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_1F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count++; + break; + case EMIT_1F_PSIZE: + /* FIXME */ + assert(0); + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_PAD; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count++; + break; + case EMIT_2F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_2F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 2; + break; + case EMIT_3F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_3F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 3; + break; + case EMIT_4F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_4F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 4; + break; + case EMIT_4UB: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_4UB_4F_BGRA; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 1; + break; + default: + assert(0); + } + } + + assert(count == vinfo->size); + + draw_vf_set_vertex_attributes(vbuf->vf, + attrs, + nr_attrs, + vbuf->vertex_size); } @@ -269,6 +427,7 @@ vbuf_set_prim( struct vbuf_stage *vbuf, uint newprim ) vbuf->vinfo = vinfo; vbuf->vertex_size = vertex_size; + vbuf_set_vf_attributes(vbuf); if (!vbuf->vertices) vbuf_alloc_vertices(vbuf); @@ -423,7 +582,12 @@ static void vbuf_destroy( struct draw_stage *stage ) { struct vbuf_stage *vbuf = vbuf_stage( stage ); - align_free( vbuf->indices ); + if(vbuf->indices) + align_free( vbuf->indices ); + + if(vbuf->vf) + draw_vf_destroy( vbuf->vf ); + FREE( stage ); } @@ -436,6 +600,9 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, { struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); + if(!vbuf) + return NULL; + vbuf->stage.draw = draw; vbuf->stage.point = vbuf_first_point; vbuf->stage.line = vbuf_first_line; @@ -450,11 +617,17 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, vbuf->max_indices = render->max_indices; vbuf->indices = (ushort *) align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); + if(!vbuf->indices) + vbuf_destroy(&vbuf->stage); vbuf->vertices = NULL; vbuf->vertex_ptr = vbuf->vertices; vbuf->prim = ~0; + vbuf->vf = draw_vf_create(FALSE); + if(!vbuf->vf) + vbuf_destroy(&vbuf->stage); + return &vbuf->stage; } diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index f758460b5f..675974c6bc 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -162,7 +162,7 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, for (j = 0, i = 0; i < nr; i++) { const unsigned format = map[i].format; - if (format == EMIT_PAD) { + if (format == DRAW_EMIT_PAD) { if (DBG) _mesa_printf("%d: pad %d, offset %d\n", i, map[i].offset, offset); @@ -261,6 +261,22 @@ void draw_vf_set_sources( struct draw_vertex_fetch *vf, } +/* Set attribute pointers, adjusted for start position: + */ +void draw_vf_set_data( struct draw_vertex_fetch *vf, + float data[][4]) +{ + struct draw_vf_attr *a = vf->attr; + unsigned j; + + for (j = 0; j < vf->attr_count; j++) { + a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ + a[j].inputsize = 4; + a[j].do_insert = a[j].insert[4 - 1]; + a[j].inputptr = (uint8_t *)&data[a[j].attrib][0]; + } +} + /* Emit count VB vertices to dest. */ diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h index 279570aad5..7619c0ee27 100644 --- a/src/mesa/pipe/draw/draw_vf.h +++ b/src/mesa/pipe/draw/draw_vf.h @@ -29,9 +29,11 @@ #define DRAW_VF_H -#include "pipe/p_compiler.h" #include "math/m_vector.h" +#include "pipe/p_compiler.h" +#include "draw_vertex.h" + enum { DRAW_VF_ATTRIB_POS = 0, @@ -67,24 +69,24 @@ enum { }; enum draw_vf_attr_format { - EMIT_1F, - EMIT_2F, - EMIT_3F, - EMIT_4F, - EMIT_2F_VIEWPORT, /**< do viewport transform and emit */ - EMIT_3F_VIEWPORT, /**< do viewport transform and emit */ - EMIT_4F_VIEWPORT, /**< do viewport transform and emit */ - EMIT_3F_XYW, /**< for projective texture */ - EMIT_1UB_1F, /**< for fog coordinate */ - EMIT_3UB_3F_RGB, /**< for specular color */ - EMIT_3UB_3F_BGR, /**< for specular color */ - EMIT_4UB_4F_RGBA, /**< for color */ - EMIT_4UB_4F_BGRA, /**< for color */ - EMIT_4UB_4F_ARGB, /**< for color */ - EMIT_4UB_4F_ABGR, /**< for color */ - EMIT_4CHAN_4F_RGBA, /**< for swrast color */ - EMIT_PAD, /**< leave a hole of 'offset' bytes */ - EMIT_MAX + DRAW_EMIT_1F, + DRAW_EMIT_2F, + DRAW_EMIT_3F, + DRAW_EMIT_4F, + DRAW_EMIT_2F_VIEWPORT, /**< do viewport transform and emit */ + DRAW_EMIT_3F_VIEWPORT, /**< do viewport transform and emit */ + DRAW_EMIT_4F_VIEWPORT, /**< do viewport transform and emit */ + DRAW_EMIT_3F_XYW, /**< for projective texture */ + DRAW_EMIT_1UB_1F, /**< for fog coordinate */ + DRAW_EMIT_3UB_3F_RGB, /**< for specular color */ + DRAW_EMIT_3UB_3F_BGR, /**< for specular color */ + DRAW_EMIT_4UB_4F_RGBA, /**< for color */ + DRAW_EMIT_4UB_4F_BGRA, /**< for color */ + DRAW_EMIT_4UB_4F_ARGB, /**< for color */ + DRAW_EMIT_4UB_4F_ABGR, /**< for color */ + DRAW_EMIT_4CHAN_4F_RGBA, /**< for swrast color */ + DRAW_EMIT_PAD, /**< leave a hole of 'offset' bytes */ + DRAW_EMIT_MAX }; struct draw_vf_attr_map { @@ -116,6 +118,10 @@ draw_vf_set_sources( struct draw_vertex_fetch *vf, GLvector4f * const attrib[], unsigned start ); +void +draw_vf_set_data( struct draw_vertex_fetch *vf, + float data[][4]); + void draw_vf_emit_vertices( struct draw_vertex_fetch *vf, unsigned count, @@ -243,7 +249,7 @@ struct draw_vf_format_info { const unsigned attrsize; }; -const struct draw_vf_format_info draw_vf_format_info[EMIT_MAX]; +const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX]; #endif diff --git a/src/mesa/pipe/draw/draw_vf_generic.c b/src/mesa/pipe/draw/draw_vf_generic.c index 19e6c587e5..42effc0c65 100644 --- a/src/mesa/pipe/draw/draw_vf_generic.c +++ b/src/mesa/pipe/draw/draw_vf_generic.c @@ -735,7 +735,7 @@ static void extract_1ub_1f( const struct draw_vf_attr *a, float *out, const uint } -const struct draw_vf_format_info draw_vf_format_info[EMIT_MAX] = +const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX] = { { "1f", extract_1f, diff --git a/src/mesa/pipe/draw/draw_vf_sse.c b/src/mesa/pipe/draw/draw_vf_sse.c index 2cf3a45ff9..a7019a47e6 100644 --- a/src/mesa/pipe/draw/draw_vf_sse.c +++ b/src/mesa/pipe/draw/draw_vf_sse.c @@ -398,19 +398,19 @@ static boolean build_vertex_emit( struct x86_program *p ) * Could be shortcircuited in specific cases: */ switch (a->format) { - case EMIT_1F: + case DRAW_EMIT_1F: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); emit_store(p, dest, 1, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_2F: + case DRAW_EMIT_2F: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); emit_store(p, dest, 2, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_3F: + case DRAW_EMIT_3F: /* Potentially the worst case - hardcode 2+1 copying: */ if (0) { @@ -433,13 +433,13 @@ static boolean build_vertex_emit( struct x86_program *p ) update_src_ptr(p, srcECX, vfESI, a); } break; - case EMIT_4F: + case DRAW_EMIT_4F: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); emit_store(p, dest, 4, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_2F_VIEWPORT: + case DRAW_EMIT_2F_VIEWPORT: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); sse_mulps(&p->func, temp, vp0); @@ -447,7 +447,7 @@ static boolean build_vertex_emit( struct x86_program *p ) emit_store(p, dest, 2, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_3F_VIEWPORT: + case DRAW_EMIT_3F_VIEWPORT: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); sse_mulps(&p->func, temp, vp0); @@ -455,7 +455,7 @@ static boolean build_vertex_emit( struct x86_program *p ) emit_store(p, dest, 3, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_4F_VIEWPORT: + case DRAW_EMIT_4F_VIEWPORT: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); sse_mulps(&p->func, temp, vp0); @@ -463,7 +463,7 @@ static boolean build_vertex_emit( struct x86_program *p ) emit_store(p, dest, 4, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_3F_XYW: + case DRAW_EMIT_3F_XYW: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z)); @@ -471,7 +471,7 @@ static boolean build_vertex_emit( struct x86_program *p ) update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_1UB_1F: + case DRAW_EMIT_1UB_1F: /* Test for PAD3 + 1UB: */ if (j > 0 && @@ -488,15 +488,15 @@ static boolean build_vertex_emit( struct x86_program *p ) return FALSE; } break; - case EMIT_3UB_3F_RGB: - case EMIT_3UB_3F_BGR: + case DRAW_EMIT_3UB_3F_RGB: + case DRAW_EMIT_3UB_3F_BGR: /* Test for 3UB + PAD1: */ if (j == vf->attr_count - 1 || a[1].vertoffset >= a->vertoffset + 4) { get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - if (a->format == EMIT_3UB_3F_BGR) + if (a->format == DRAW_EMIT_3UB_3F_BGR) sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); emit_pack_store_4ub(p, dest, temp); update_src_ptr(p, srcECX, vfESI, a); @@ -504,7 +504,7 @@ static boolean build_vertex_emit( struct x86_program *p ) /* Test for 3UB + 1UB: */ else if (j < vf->attr_count - 1 && - a[1].format == EMIT_1UB_1F && + a[1].format == DRAW_EMIT_1UB_1F && a[1].vertoffset == a->vertoffset + 3) { get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); @@ -520,7 +520,7 @@ static boolean build_vertex_emit( struct x86_program *p ) /* Rearrange and possibly do BGR conversion: */ - if (a->format == EMIT_3UB_3F_BGR) + if (a->format == DRAW_EMIT_3UB_3F_BGR) sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); else sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X)); @@ -534,34 +534,34 @@ static boolean build_vertex_emit( struct x86_program *p ) return FALSE; /* add this later */ break; - case EMIT_4UB_4F_RGBA: + case DRAW_EMIT_4UB_4F_RGBA: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); emit_pack_store_4ub(p, dest, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_4UB_4F_BGRA: + case DRAW_EMIT_4UB_4F_BGRA: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); emit_pack_store_4ub(p, dest, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_4UB_4F_ARGB: + case DRAW_EMIT_4UB_4F_ARGB: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); emit_pack_store_4ub(p, dest, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_4UB_4F_ABGR: + case DRAW_EMIT_4UB_4F_ABGR: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); emit_pack_store_4ub(p, dest, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case EMIT_4CHAN_4F_RGBA: + case DRAW_EMIT_4CHAN_4F_RGBA: switch (CHAN_TYPE) { case GL_UNSIGNED_BYTE: get_src_ptr(p, srcECX, vfESI, a); -- cgit v1.2.3 From 5abc8d9e23b1b8cde9c4183b73bfced3d4f01c87 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Jan 2008 10:14:13 +0000 Subject: gallium: remove dead code from draw_vf* --- src/mesa/pipe/draw/Makefile | 2 + src/mesa/pipe/draw/draw_vbuf.c | 2 +- src/mesa/pipe/draw/draw_vf.c | 90 +------- src/mesa/pipe/draw/draw_vf.h | 18 +- src/mesa/pipe/draw/draw_vf_generic.c | 420 +---------------------------------- src/mesa/pipe/draw/draw_vf_sse.c | 51 ----- 6 files changed, 9 insertions(+), 574 deletions(-) create mode 100644 src/mesa/pipe/draw/Makefile (limited to 'src') diff --git a/src/mesa/pipe/draw/Makefile b/src/mesa/pipe/draw/Makefile new file mode 100644 index 0000000000..451911a354 --- /dev/null +++ b/src/mesa/pipe/draw/Makefile @@ -0,0 +1,2 @@ +default: + cd .. ; make diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index a3d0b5bca3..8ca225c65a 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -625,7 +625,7 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, vbuf->prim = ~0; - vbuf->vf = draw_vf_create(FALSE); + vbuf->vf = draw_vf_create(); if(!vbuf->vf) vbuf_destroy(&vbuf->stage); diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index 675974c6bc..deedfc7bc7 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -177,7 +177,6 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, vf->attr[j].attrib = map[i].attrib; vf->attr[j].format = format; vf->attr[j].insert = draw_vf_format_info[format].insert; - vf->attr[j].extract = draw_vf_format_info[format].extract; vf->attr[j].vertattrsize = draw_vf_format_info[format].attrsize; vf->attr[j].vertoffset = offset; @@ -201,41 +200,6 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, -void draw_vf_set_vp_matrix( struct draw_vertex_fetch *vf, - const float *viewport ) -{ - assert(vf->allow_viewport_emits); - - /* scale */ - vf->vp[0] = viewport[MAT_SX]; - vf->vp[1] = viewport[MAT_SY]; - vf->vp[2] = viewport[MAT_SZ]; - vf->vp[3] = 1.0; - - /* translate */ - vf->vp[4] = viewport[MAT_TX]; - vf->vp[5] = viewport[MAT_TY]; - vf->vp[6] = viewport[MAT_TZ]; - vf->vp[7] = 0.0; -} - -void draw_vf_set_vp_scale_translate( struct draw_vertex_fetch *vf, - const float *scale, - const float *translate ) -{ - assert(vf->allow_viewport_emits); - - vf->vp[0] = scale[0]; - vf->vp[1] = scale[1]; - vf->vp[2] = scale[2]; - vf->vp[3] = scale[3]; - - vf->vp[4] = translate[0]; - vf->vp[5] = translate[1]; - vf->vp[6] = translate[2]; - vf->vp[7] = translate[3]; -} - /* Set attribute pointers, adjusted for start position: */ @@ -288,39 +252,10 @@ void draw_vf_emit_vertices( struct draw_vertex_fetch *vf, } -/* Extract a named attribute from a hardware vertex. Will have to - * reverse any viewport transformation, swizzling or other conversions - * which may have been applied. - * - * This is mainly required for on-the-fly vertex translations to - * swrast format. - */ -void draw_vf_get_attr( struct draw_vertex_fetch *vf, - const void *vertex, - GLenum attr, - const float *dflt, - float *dest ) -{ - const struct draw_vf_attr *a = vf->attr; - const unsigned attr_count = vf->attr_count; - unsigned j; - - for (j = 0; j < attr_count; j++) { - if (a[j].attrib == attr) { - a[j].extract( &a[j], dest, (uint8_t *)vertex + a[j].vertoffset ); - return; - } - } - - /* Else return the value from ctx->Current. - */ - _mesa_memcpy( dest, dflt, 4*sizeof(float)); -} - -struct draw_vertex_fetch *draw_vf_create( boolean allow_viewport_emits ) +struct draw_vertex_fetch *draw_vf_create( void ) { struct draw_vertex_fetch *vf = CALLOC_STRUCT(draw_vertex_fetch); unsigned i; @@ -328,29 +263,6 @@ struct draw_vertex_fetch *draw_vf_create( boolean allow_viewport_emits ) for (i = 0; i < DRAW_VF_ATTRIB_MAX; i++) vf->attr[i].vf = vf; - vf->allow_viewport_emits = allow_viewport_emits; - - switch(CHAN_TYPE) { - case GL_UNSIGNED_BYTE: - vf->chan_scale[0] = 255.0; - vf->chan_scale[1] = 255.0; - vf->chan_scale[2] = 255.0; - vf->chan_scale[3] = 255.0; - break; - case GL_UNSIGNED_SHORT: - vf->chan_scale[0] = 65535.0; - vf->chan_scale[1] = 65535.0; - vf->chan_scale[2] = 65535.0; - vf->chan_scale[3] = 65535.0; - break; - default: - vf->chan_scale[0] = 1.0; - vf->chan_scale[1] = 1.0; - vf->chan_scale[2] = 1.0; - vf->chan_scale[3] = 1.0; - break; - } - vf->identity[0] = 0.0; vf->identity[1] = 0.0; vf->identity[2] = 0.0; diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h index 7619c0ee27..c6a8fe0d53 100644 --- a/src/mesa/pipe/draw/draw_vf.h +++ b/src/mesa/pipe/draw/draw_vf.h @@ -73,9 +73,6 @@ enum draw_vf_attr_format { DRAW_EMIT_2F, DRAW_EMIT_3F, DRAW_EMIT_4F, - DRAW_EMIT_2F_VIEWPORT, /**< do viewport transform and emit */ - DRAW_EMIT_3F_VIEWPORT, /**< do viewport transform and emit */ - DRAW_EMIT_4F_VIEWPORT, /**< do viewport transform and emit */ DRAW_EMIT_3F_XYW, /**< for projective texture */ DRAW_EMIT_1UB_1F, /**< for fog coordinate */ DRAW_EMIT_3UB_3F_RGB, /**< for specular color */ @@ -84,7 +81,6 @@ enum draw_vf_attr_format { DRAW_EMIT_4UB_4F_BGRA, /**< for color */ DRAW_EMIT_4UB_4F_ARGB, /**< for color */ DRAW_EMIT_4UB_4F_ABGR, /**< for color */ - DRAW_EMIT_4CHAN_4F_RGBA, /**< for swrast color */ DRAW_EMIT_PAD, /**< leave a hole of 'offset' bytes */ DRAW_EMIT_MAX }; @@ -98,14 +94,6 @@ struct draw_vf_attr_map { struct draw_vertex_fetch; -void -draw_vf_set_vp_matrix( struct draw_vertex_fetch *vf, - const float *viewport ); - -void -draw_vf_set_vp_scale_translate( struct draw_vertex_fetch *vf, - const float *scale, - const float *translate ); unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, @@ -135,7 +123,7 @@ draw_vf_get_attr( struct draw_vertex_fetch *vf, float *dest ); struct draw_vertex_fetch * -draw_vf_create( boolean allow_viewport_emits ); +draw_vf_create( void ); void draw_vf_destroy( struct draw_vertex_fetch *vf ); @@ -196,9 +184,6 @@ struct draw_vertex_fetch /* Parameters and constants for codegen: */ - boolean allow_viewport_emits; - float vp[8]; - float chan_scale[4]; float identity[4]; struct draw_vf_fastpath *fastpath; @@ -244,7 +229,6 @@ draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ); struct draw_vf_format_info { const char *name; - draw_vf_extract_func extract; draw_vf_insert_func insert[4]; const unsigned attrsize; }; diff --git a/src/mesa/pipe/draw/draw_vf_generic.c b/src/mesa/pipe/draw/draw_vf_generic.c index 42effc0c65..343428d26c 100644 --- a/src/mesa/pipe/draw/draw_vf_generic.c +++ b/src/mesa/pipe/draw/draw_vf_generic.c @@ -36,125 +36,6 @@ #include "draw_vf.h" -/* - * These functions take the NDC coordinates pointed to by 'in', apply the - * NDC->Viewport mapping and store the results at 'v'. - */ - -static INLINE void insert_4f_viewport_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - float *out = (float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = scale[0] * in[0] + trans[0]; - out[1] = scale[1] * in[1] + trans[1]; - out[2] = scale[2] * in[2] + trans[2]; - out[3] = in[3]; -} - -static INLINE void insert_4f_viewport_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - float *out = (float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = scale[0] * in[0] + trans[0]; - out[1] = scale[1] * in[1] + trans[1]; - out[2] = scale[2] * in[2] + trans[2]; - out[3] = 1; -} - -static INLINE void insert_4f_viewport_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - float *out = (float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = scale[0] * in[0] + trans[0]; - out[1] = scale[1] * in[1] + trans[1]; - out[2] = trans[2]; - out[3] = 1; -} - -static INLINE void insert_4f_viewport_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - float *out = (float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = scale[0] * in[0] + trans[0]; - out[1] = trans[1]; - out[2] = trans[2]; - out[3] = 1; -} - -static INLINE void insert_3f_viewport_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - float *out = (float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = scale[0] * in[0] + trans[0]; - out[1] = scale[1] * in[1] + trans[1]; - out[2] = scale[2] * in[2] + trans[2]; -} - -static INLINE void insert_3f_viewport_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - float *out = (float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = scale[0] * in[0] + trans[0]; - out[1] = scale[1] * in[1] + trans[1]; - out[2] = scale[2] * in[2] + trans[2]; -} - -static INLINE void insert_3f_viewport_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - float *out = (float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = scale[0] * in[0] + trans[0]; - out[1] = trans[1]; - out[2] = trans[2]; -} - -static INLINE void insert_2f_viewport_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - float *out = (float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = scale[0] * in[0] + trans[0]; - out[1] = scale[1] * in[1] + trans[1]; -} - -static INLINE void insert_2f_viewport_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - float *out = (float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = scale[0] * in[0] + trans[0]; - out[1] = trans[1]; -} - - -/* - * These functions do the same as above, except for the viewport mapping. - */ static INLINE void insert_4f_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) { @@ -278,50 +159,6 @@ static INLINE void insert_null( const struct draw_vf_attr *a, uint8_t *v, const (void) a; (void) v; (void) in; } -static INLINE void insert_4chan_4f_rgba_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - GLchan *c = (GLchan *)v; - (void) a; - UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); - UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); - UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); - UNCLAMPED_FLOAT_TO_CHAN(c[3], in[3]); -} - -static INLINE void insert_4chan_4f_rgba_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - GLchan *c = (GLchan *)v; - (void) a; - UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); - UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); - UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); - c[3] = CHAN_MAX; -} - -static INLINE void insert_4chan_4f_rgba_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - GLchan *c = (GLchan *)v; - (void) a; - UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); - UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); - c[2] = 0; - c[3] = CHAN_MAX; -} - -static INLINE void insert_4chan_4f_rgba_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - GLchan *c = (GLchan *)v; - (void) a; - UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); - c[1] = 0; - c[2] = 0; - c[3] = CHAN_MAX; -} - static INLINE void insert_4ub_4f_rgba_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) { @@ -545,291 +382,64 @@ static INLINE void insert_1ub_1f_1( const struct draw_vf_attr *a, uint8_t *v, } -/*********************************************************************** - * Functions to perform the reverse operations to the above, for - * swrast translation and clip-interpolation. - * - * Currently always extracts a full 4 floats. - */ - -static void extract_4f_viewport( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - const float *in = (const float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - /* Although included for completeness, the position coordinate is - * usually handled differently during clipping. - */ - out[0] = (in[0] - trans[0]) / scale[0]; - out[1] = (in[1] - trans[1]) / scale[1]; - out[2] = (in[2] - trans[2]) / scale[2]; - out[3] = in[3]; -} - -static void extract_3f_viewport( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - const float *in = (const float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = (in[0] - trans[0]) / scale[0]; - out[1] = (in[1] - trans[1]) / scale[1]; - out[2] = (in[2] - trans[2]) / scale[2]; - out[3] = 1; -} - - -static void extract_2f_viewport( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - const float *in = (const float *)v; - const float *scale = a->vf->vp; - const float *trans = a->vf->vp + 4; - - out[0] = (in[0] - trans[0]) / scale[0]; - out[1] = (in[1] - trans[1]) / scale[1]; - out[2] = 0; - out[3] = 1; -} - - -static void extract_4f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) -{ - const float *in = (const float *)v; - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; - out[3] = in[3]; -} - -static void extract_3f_xyw( const struct draw_vf_attr *a, float *out, const uint8_t *v ) -{ - const float *in = (const float *)v; - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = 0; - out[3] = in[2]; -} - - -static void extract_3f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) -{ - const float *in = (const float *)v; - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; - out[3] = 1; -} - - -static void extract_2f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) -{ - const float *in = (const float *)v; - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = 0; - out[3] = 1; -} - -static void extract_1f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) -{ - const float *in = (const float *)v; - (void) a; - - out[0] = in[0]; - out[1] = 0; - out[2] = 0; - out[3] = 1; -} - -static void extract_4chan_4f_rgba( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - GLchan *c = (GLchan *)v; - (void) a; - - out[0] = CHAN_TO_FLOAT(c[0]); - out[1] = CHAN_TO_FLOAT(c[1]); - out[2] = CHAN_TO_FLOAT(c[2]); - out[3] = CHAN_TO_FLOAT(c[3]); -} - -static void extract_4ub_4f_rgba( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - (void) a; - out[0] = UBYTE_TO_FLOAT(v[0]); - out[1] = UBYTE_TO_FLOAT(v[1]); - out[2] = UBYTE_TO_FLOAT(v[2]); - out[3] = UBYTE_TO_FLOAT(v[3]); -} - -static void extract_4ub_4f_bgra( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - (void) a; - out[2] = UBYTE_TO_FLOAT(v[0]); - out[1] = UBYTE_TO_FLOAT(v[1]); - out[0] = UBYTE_TO_FLOAT(v[2]); - out[3] = UBYTE_TO_FLOAT(v[3]); -} - -static void extract_4ub_4f_argb( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - (void) a; - out[3] = UBYTE_TO_FLOAT(v[0]); - out[0] = UBYTE_TO_FLOAT(v[1]); - out[1] = UBYTE_TO_FLOAT(v[2]); - out[2] = UBYTE_TO_FLOAT(v[3]); -} - -static void extract_4ub_4f_abgr( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - (void) a; - out[3] = UBYTE_TO_FLOAT(v[0]); - out[2] = UBYTE_TO_FLOAT(v[1]); - out[1] = UBYTE_TO_FLOAT(v[2]); - out[0] = UBYTE_TO_FLOAT(v[3]); -} - -static void extract_3ub_3f_rgb( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - (void) a; - out[0] = UBYTE_TO_FLOAT(v[0]); - out[1] = UBYTE_TO_FLOAT(v[1]); - out[2] = UBYTE_TO_FLOAT(v[2]); - out[3] = 1; -} - -static void extract_3ub_3f_bgr( const struct draw_vf_attr *a, float *out, - const uint8_t *v ) -{ - (void) a; - out[2] = UBYTE_TO_FLOAT(v[0]); - out[1] = UBYTE_TO_FLOAT(v[1]); - out[0] = UBYTE_TO_FLOAT(v[2]); - out[3] = 1; -} - -static void extract_1ub_1f( const struct draw_vf_attr *a, float *out, const uint8_t *v ) -{ - (void) a; - out[0] = UBYTE_TO_FLOAT(v[0]); - out[1] = 0; - out[2] = 0; - out[3] = 1; -} - - const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX] = { { "1f", - extract_1f, { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, sizeof(float) }, { "2f", - extract_2f, { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, 2 * sizeof(float) }, { "3f", - extract_3f, { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, 3 * sizeof(float) }, { "4f", - extract_4f, { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, 4 * sizeof(float) }, - { "2f_viewport", - extract_2f_viewport, - { insert_2f_viewport_1, insert_2f_viewport_2, insert_2f_viewport_2, - insert_2f_viewport_2 }, - 2 * sizeof(float) }, - - { "3f_viewport", - extract_3f_viewport, - { insert_3f_viewport_1, insert_3f_viewport_2, insert_3f_viewport_3, - insert_3f_viewport_3 }, - 3 * sizeof(float) }, - - { "4f_viewport", - extract_4f_viewport, - { insert_4f_viewport_1, insert_4f_viewport_2, insert_4f_viewport_3, - insert_4f_viewport_4 }, - 4 * sizeof(float) }, - { "3f_xyw", - extract_3f_xyw, { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_4 }, 3 * sizeof(float) }, { "1ub_1f", - extract_1ub_1f, { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 }, sizeof(uint8_t) }, { "3ub_3f_rgb", - extract_3ub_3f_rgb, { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3, insert_3ub_3f_rgb_3 }, 3 * sizeof(uint8_t) }, { "3ub_3f_bgr", - extract_3ub_3f_bgr, { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3, insert_3ub_3f_bgr_3 }, 3 * sizeof(uint8_t) }, { "4ub_4f_rgba", - extract_4ub_4f_rgba, { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, insert_4ub_4f_rgba_4 }, 4 * sizeof(uint8_t) }, { "4ub_4f_bgra", - extract_4ub_4f_bgra, { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3, insert_4ub_4f_bgra_4 }, 4 * sizeof(uint8_t) }, { "4ub_4f_argb", - extract_4ub_4f_argb, { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3, insert_4ub_4f_argb_4 }, 4 * sizeof(uint8_t) }, { "4ub_4f_abgr", - extract_4ub_4f_abgr, { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3, insert_4ub_4f_abgr_4 }, 4 * sizeof(uint8_t) }, - { "4chan_4f_rgba", - extract_4chan_4f_rgba, - { insert_4chan_4f_rgba_1, insert_4chan_4f_rgba_2, insert_4chan_4f_rgba_3, - insert_4chan_4f_rgba_4 }, - 4 * sizeof(GLchan) }, - { "pad", - NULL, { NULL, NULL, NULL, NULL }, 0 } @@ -889,16 +499,10 @@ static void NAME( struct draw_vertex_fetch *vf, \ insert_null, NAME) -EMIT2(insert_3f_viewport_3, insert_4ub_4f_rgba_4, emit_viewport3_rgba4) -EMIT2(insert_3f_viewport_3, insert_4ub_4f_bgra_4, emit_viewport3_bgra4) EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4) -EMIT3(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_viewport4_rgba4_st2) -EMIT3(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2, emit_viewport4_bgra4_st2) EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2) -EMIT4(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_viewport4_rgba4_st2_st2) -EMIT4(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2, insert_2f_2, emit_viewport4_bgra4_st2_st2) EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2) @@ -914,42 +518,26 @@ void draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ) */ switch (vf->attr_count) { case 2: - if (vf->attr[0].do_insert == insert_3f_viewport_3) { - if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4) - func = emit_viewport3_bgra4; - else if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) - func = emit_viewport3_rgba4; - } - else if (vf->attr[0].do_insert == insert_3f_3 && - vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + if (vf->attr[0].do_insert == insert_3f_3 && + vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { func = emit_xyz3_rgba4; } break; case 3: if (vf->attr[2].do_insert == insert_2f_2) { if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - if (vf->attr[0].do_insert == insert_4f_viewport_4) - func = emit_viewport4_rgba4_st2; - else if (vf->attr[0].do_insert == insert_4f_4) + if (vf->attr[0].do_insert == insert_4f_4) func = emit_xyzw4_rgba4_st2; } - else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 && - vf->attr[0].do_insert == insert_4f_viewport_4) - func = emit_viewport4_bgra4_st2; } break; case 4: if (vf->attr[2].do_insert == insert_2f_2 && vf->attr[3].do_insert == insert_2f_2) { if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - if (vf->attr[0].do_insert == insert_4f_viewport_4) - func = emit_viewport4_rgba4_st2_st2; - else if (vf->attr[0].do_insert == insert_4f_4) + if (vf->attr[0].do_insert == insert_4f_4) func = emit_xyzw4_rgba4_st2_st2; } - else if (vf->attr[1].do_insert == insert_4ub_4f_bgra_4 && - vf->attr[0].do_insert == insert_4f_viewport_4) - func = emit_viewport4_bgra4_st2_st2; } break; } diff --git a/src/mesa/pipe/draw/draw_vf_sse.c b/src/mesa/pipe/draw/draw_vf_sse.c index a7019a47e6..b238b542e7 100644 --- a/src/mesa/pipe/draw/draw_vf_sse.c +++ b/src/mesa/pipe/draw/draw_vf_sse.c @@ -369,13 +369,6 @@ static boolean build_vertex_emit( struct x86_program *p ) x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1)); - /* Possibly load vp0, vp1 for viewport calcs: - */ - if (vf->allow_viewport_emits) { - sse_movups(&p->func, vp0, x86_make_disp(vfESI, get_offset(vf, &vf->vp[0]))); - sse_movups(&p->func, vp1, x86_make_disp(vfESI, get_offset(vf, &vf->vp[4]))); - } - /* always load, needed or not: */ sse_movups(&p->func, p->chan0, x86_make_disp(vfESI, get_offset(vf, &vf->chan_scale[0]))); @@ -439,30 +432,6 @@ static boolean build_vertex_emit( struct x86_program *p ) emit_store(p, dest, 4, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case DRAW_EMIT_2F_VIEWPORT: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); - sse_mulps(&p->func, temp, vp0); - sse_addps(&p->func, temp, vp1); - emit_store(p, dest, 2, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_3F_VIEWPORT: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - sse_mulps(&p->func, temp, vp0); - sse_addps(&p->func, temp, vp1); - emit_store(p, dest, 3, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4F_VIEWPORT: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_mulps(&p->func, temp, vp0); - sse_addps(&p->func, temp, vp1); - emit_store(p, dest, 4, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; case DRAW_EMIT_3F_XYW: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); @@ -561,26 +530,6 @@ static boolean build_vertex_emit( struct x86_program *p ) emit_pack_store_4ub(p, dest, temp); update_src_ptr(p, srcECX, vfESI, a); break; - case DRAW_EMIT_4CHAN_4F_RGBA: - switch (CHAN_TYPE) { - case GL_UNSIGNED_BYTE: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case GL_FLOAT: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 4, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case GL_UNSIGNED_SHORT: - default: - _mesa_printf("unknown CHAN_TYPE %s\n", _mesa_lookup_enum_by_nr(CHAN_TYPE)); - return FALSE; - } - break; default: _mesa_printf("unknown a[%d].format %d\n", j, a->format); return FALSE; /* catch any new opcodes */ -- cgit v1.2.3 From 4141ebdf59cddbb412b388c7f38f50e5e80c49d2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Jan 2008 10:46:52 +0000 Subject: gallium: fill in missing formats for vertex_fetch --- src/mesa/pipe/draw/draw_vertex_fetch.c | 220 ++++++++++++++++++++++++++++++--- 1 file changed, 203 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index fb64723a19..0789dc8e8c 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -62,50 +62,236 @@ fetch_##NAME(const void *ptr, float *attrib) \ } \ } +#define CVT_64_FLOAT ((double *) ptr)[i] #define CVT_32_FLOAT ((float *) ptr)[i] + +#define CVT_8_USCALED (float) ((unsigned char *) ptr)[i] +#define CVT_16_USCALED (float) ((unsigned short *) ptr)[i] +#define CVT_32_USCALED (float) ((unsigned int *) ptr)[i] + +#define CVT_8_SSCALED (float) ((char *) ptr)[i] +#define CVT_16_SSCALED (float) ((short *) ptr)[i] #define CVT_32_SSCALED (float) ((int *) ptr)[i] + #define CVT_8_UNORM (float) ((unsigned char *) ptr)[i] / 255.0f +#define CVT_16_UNORM (float) ((unsigned short *) ptr)[i] / 65535.0f +#define CVT_32_UNORM (float) ((unsigned int *) ptr)[i] / 4294967295.0f + +#define CVT_8_SNORM (float) ((char *) ptr)[i] / 127.0f +#define CVT_16_SNORM (float) ((short *) ptr)[i] / 32767.0f +#define CVT_32_SNORM (float) ((int *) ptr)[i] / 2147483647.0f + +FETCH_ATTRIB( R64G64B64A64_FLOAT, 4, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) +FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) + +FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) +FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) + FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) + +FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) +FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) + +FETCH_ATTRIB( R32G32B32A32_SNORM, 4, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32B32_SNORM, 3, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32_SNORM, 2, CVT_32_SNORM ) +FETCH_ATTRIB( R32_SNORM, 1, CVT_32_SNORM ) + +FETCH_ATTRIB( R16G16B16A16_USCALED, 4, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16B16_USCALED, 3, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16_USCALED, 2, CVT_16_USCALED ) +FETCH_ATTRIB( R16_USCALED, 1, CVT_16_USCALED ) + +FETCH_ATTRIB( R16G16B16A16_SSCALED, 4, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16B16_SSCALED, 3, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16_SSCALED, 2, CVT_16_SSCALED ) +FETCH_ATTRIB( R16_SSCALED, 1, CVT_16_SSCALED ) + +FETCH_ATTRIB( R16G16B16A16_UNORM, 4, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16B16_UNORM, 3, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16_UNORM, 2, CVT_16_UNORM ) +FETCH_ATTRIB( R16_UNORM, 1, CVT_16_UNORM ) + +FETCH_ATTRIB( R16G16B16A16_SNORM, 4, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16B16_SNORM, 3, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16_SNORM, 2, CVT_16_SNORM ) +FETCH_ATTRIB( R16_SNORM, 1, CVT_16_SNORM ) + +FETCH_ATTRIB( R8G8B8A8_USCALED, 4, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8B8_USCALED, 3, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8_USCALED, 2, CVT_8_USCALED ) +FETCH_ATTRIB( R8_USCALED, 1, CVT_8_USCALED ) + +FETCH_ATTRIB( R8G8B8A8_SSCALED, 4, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8B8_SSCALED, 3, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8_SSCALED, 2, CVT_8_SSCALED ) +FETCH_ATTRIB( R8_SSCALED, 1, CVT_8_SSCALED ) + +FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8B8_UNORM, 3, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8_UNORM, 2, CVT_8_UNORM ) +FETCH_ATTRIB( R8_UNORM, 1, CVT_8_UNORM ) + +FETCH_ATTRIB( R8G8B8A8_SNORM, 4, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8B8_SNORM, 3, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8_SNORM, 2, CVT_8_SNORM ) +FETCH_ATTRIB( R8_SNORM, 1, CVT_8_SNORM ) + FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) +//FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) static fetch_func get_fetch_func( enum pipe_format format ) { switch (format) { - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return fetch_R32G32B32A32_FLOAT; - case PIPE_FORMAT_R32G32B32_FLOAT: - return fetch_R32G32B32_FLOAT; - case PIPE_FORMAT_R32G32_FLOAT: - return fetch_R32G32_FLOAT; + case PIPE_FORMAT_R64_FLOAT: + return fetch_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return fetch_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return fetch_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return fetch_R64G64B64A64_FLOAT; + case PIPE_FORMAT_R32_FLOAT: return fetch_R32_FLOAT; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - return fetch_R32G32B32A32_SSCALED; - case PIPE_FORMAT_R32G32B32_SSCALED: - return fetch_R32G32B32_SSCALED; - case PIPE_FORMAT_R32G32_SSCALED: - return fetch_R32G32_SSCALED; + case PIPE_FORMAT_R32G32_FLOAT: + return fetch_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return fetch_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return fetch_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return fetch_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return fetch_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return fetch_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return fetch_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return fetch_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return fetch_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return fetch_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return fetch_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return fetch_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return fetch_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return fetch_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return fetch_R32G32B32A32_SNORM; + case PIPE_FORMAT_R32_SSCALED: return fetch_R32_SSCALED; - case PIPE_FORMAT_A8R8G8B8_UNORM: - return fetch_A8R8G8B8_UNORM; + case PIPE_FORMAT_R32G32_SSCALED: + return fetch_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return fetch_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return fetch_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return fetch_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return fetch_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return fetch_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return fetch_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return fetch_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return fetch_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return fetch_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return fetch_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return fetch_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return fetch_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return fetch_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return fetch_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return fetch_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return fetch_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return fetch_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return fetch_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return fetch_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return fetch_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return fetch_R8G8B8_UNORM; case PIPE_FORMAT_R8G8B8A8_UNORM: return fetch_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return fetch_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return fetch_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return fetch_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return fetch_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return fetch_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return fetch_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return fetch_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return fetch_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return fetch_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return fetch_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return fetch_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return fetch_R8G8B8A8_SSCALED; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + return fetch_A8R8G8B8_UNORM; + case 0: - return NULL; + return NULL; /* not sure why this is needed */ + default: - /* Lots of missing cases! */ assert(0); return NULL; } -- cgit v1.2.3 From a46181044fd5573895180ee5f1a016c4c1e4a653 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Jan 2008 10:48:22 +0000 Subject: gallium: explictly cast double to float in vertex fetch --- src/mesa/pipe/draw/draw_vertex_fetch.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 0789dc8e8c..af3983b7f0 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -62,7 +62,7 @@ fetch_##NAME(const void *ptr, float *attrib) \ } \ } -#define CVT_64_FLOAT ((double *) ptr)[i] +#define CVT_64_FLOAT (float) ((double *) ptr)[i] #define CVT_32_FLOAT ((float *) ptr)[i] #define CVT_8_USCALED (float) ((unsigned char *) ptr)[i] -- cgit v1.2.3 From 99f297651198c3424aab202595064d6f5596b2fc Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Jan 2008 11:12:11 +0000 Subject: gallium: only call vertex/prim queue flush when there is something to flush --- src/mesa/pipe/draw/draw_prim.c | 9 +++++---- src/mesa/pipe/draw/draw_vertex_fetch.c | 10 ++++++++++ src/mesa/pipe/draw/draw_vertex_shader.c | 2 ++ 3 files changed, 17 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 243381aec0..2a612a1673 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -63,8 +63,7 @@ static void draw_prim_queue_flush( struct draw_context *draw ) fprintf(stdout,"Flushing with %d prims, %d verts\n", draw->pq.queue_nr, draw->vs.queue_nr); - if (draw->pq.queue_nr == 0) - return; + assert (draw->pq.queue_nr != 0); /* NOTE: we cannot save draw->pipeline->first in a local var because * draw->pipeline->first is often changed by the first call to tri(), @@ -109,10 +108,12 @@ void draw_do_flush( struct draw_context *draw, unsigned flags ) if (flags >= DRAW_FLUSH_SHADER_QUEUE) { - draw_vertex_shader_queue_flush(draw); + if (draw->vs.queue_nr) + draw_vertex_shader_queue_flush(draw); if (flags >= DRAW_FLUSH_PRIM_QUEUE) { - draw_prim_queue_flush(draw); + if (draw->pq.queue_nr) + draw_prim_queue_flush(draw); if (flags >= DRAW_FLUSH_VERTEX_CACHE) { draw_vertex_cache_invalidate(draw); diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index af3983b7f0..143acdd3b4 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -158,6 +158,14 @@ FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) static fetch_func get_fetch_func( enum pipe_format format ) { +#if 0 + { + char tmp[80]; + pf_sprint_name(tmp, format); + _mesa_printf("%s: %s\n", __FUNCTION__, tmp); + } +#endif + switch (format) { case PIPE_FORMAT_R64_FLOAT: return fetch_R64_FLOAT; @@ -317,6 +325,8 @@ void draw_update_vertex_fetch( struct draw_context *draw ) { unsigned nr_attrs, i; +// _mesa_printf("%s\n", __FUNCTION__); + /* this may happend during context init */ if (!draw->vertex_shader) return; diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 3041974b9a..289c35c7ae 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -201,6 +201,8 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) { unsigned i, j; + assert(draw->vs.queue_nr != 0); + /* XXX: do this on statechange: */ draw_update_vertex_fetch( draw ); -- cgit v1.2.3 From 88858e046888d0bcb763537adc74a64e564678df Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Jan 2008 12:40:29 +0000 Subject: gallium: add a couple of hardwired vertex fetch functions --- src/mesa/pipe/draw/draw_private.h | 9 +- src/mesa/pipe/draw/draw_vertex_fetch.c | 150 +++++++++++++++++++++++++++----- src/mesa/pipe/draw/draw_vertex_shader.c | 12 ++- 3 files changed, 141 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 1e59f5bd8d..21de400676 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -141,6 +141,10 @@ struct draw_vertex_shader { /* Internal function for vertex fetch. */ typedef void (*fetch_func)(const void *ptr, float *attrib); +typedef void (*full_fetch_func)( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ); @@ -210,6 +214,7 @@ struct draw_context unsigned pitch[PIPE_ATTRIB_MAX]; fetch_func fetch[PIPE_ATTRIB_MAX]; unsigned nr_attrs; + full_fetch_func fetch_func; } vertex_fetch; /* Post-tnl vertex cache: @@ -287,10 +292,6 @@ extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw ); struct tgsi_exec_machine; extern void draw_update_vertex_fetch( struct draw_context *draw ); -extern void draw_vertex_fetch( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ); #define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 143acdd3b4..afdf1971d2 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -320,42 +320,101 @@ transpose_4x4( float *out, const float *in ) } - -void draw_update_vertex_fetch( struct draw_context *draw ) + +static void fetch_xyz_rgb( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) { - unsigned nr_attrs, i; + assert(count <= 4); // _mesa_printf("%s\n", __FUNCTION__); - - /* this may happend during context init */ - if (!draw->vertex_shader) - return; - nr_attrs = draw->vertex_shader->state->num_inputs; + /* loop over vertex attributes (vertex shader inputs) + */ - for (i = 0; i < nr_attrs; i++) { - unsigned buf = draw->vertex_element[i].vertex_buffer_index; - enum pipe_format format = draw->vertex_element[i].src_format; + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + + for (i = 0; i < 4; i++) { + { + const float *in = (const float *)(src[0] + elts[i] * pitch[0]); + float *out = &machine->Inputs[0].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[1] + elts[i] * pitch[1]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + } +} - draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + - draw->vertex_buffer[buf].buffer_offset + - draw->vertex_element[i].src_offset; - draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; - draw->vertex_fetch.fetch[i] = get_fetch_func( format ); - } - draw->vertex_fetch.nr_attrs = nr_attrs; + +static void fetch_xyz_rgb_st( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + assert(count <= 4); + + /* loop over vertex attributes (vertex shader inputs) + */ + + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + + for (i = 0; i < 4; i++) { + { + const float *in = (const float *)(src[0] + elts[i] * pitch[0]); + float *out = &machine->Inputs[0].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[1] + elts[i] * pitch[1]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[2] + elts[i] * pitch[2]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = 0.0f; + out[12] = 1.0f; + } + } } + + /** * Fetch vertex attributes for 'count' vertices. */ -void draw_vertex_fetch( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) +static void generic_vertex_fetch( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) { unsigned nr_attrs = draw->vertex_fetch.nr_attrs; unsigned attr; @@ -402,3 +461,50 @@ void draw_vertex_fetch( struct draw_context *draw, } } + + +void draw_update_vertex_fetch( struct draw_context *draw ) +{ + unsigned nr_attrs, i; + +// _mesa_printf("%s\n", __FUNCTION__); + + /* this may happend during context init */ + if (!draw->vertex_shader) + return; + + nr_attrs = draw->vertex_shader->state->num_inputs; + + for (i = 0; i < nr_attrs; i++) { + unsigned buf = draw->vertex_element[i].vertex_buffer_index; + enum pipe_format format = draw->vertex_element[i].src_format; + + draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[i].src_offset; + + draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; + draw->vertex_fetch.fetch[i] = get_fetch_func( format ); + } + + draw->vertex_fetch.nr_attrs = nr_attrs; + + draw->vertex_fetch.fetch_func = generic_vertex_fetch; + + switch (nr_attrs) { + case 2: + if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT) + draw->vertex_fetch.fetch_func = fetch_xyz_rgb; + break; + case 3: + if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32_FLOAT) + draw->vertex_fetch.fetch_func = fetch_xyz_rgb_st; + break; + default: + break; + } + +} diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 289c35c7ae..0806e23d6c 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -110,7 +110,7 @@ run_vertex_program(struct draw_context *draw, machine->Inputs = ALIGN16_ASSIGN(inputs); machine->Outputs = ALIGN16_ASSIGN(outputs); - draw_vertex_fetch( draw, machine, elts, count ); + draw->vertex_fetch.fetch_func( draw, machine, elts, count ); /* run shader */ #if defined(__i386__) || defined(__386__) @@ -219,14 +219,18 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) for (i = 0; i < draw->vs.queue_nr; i += 4) { struct vertex_header *dests[4]; unsigned elts[4]; - int n; + int n = MIN2(4, draw->vs.queue_nr - i); - for (j = 0; j < 4; j++) { + for (j = 0; j < n; j++) { elts[j] = draw->vs.queue[i + j].elt; dests[j] = draw->vs.queue[i + j].dest; } - n = MIN2(4, draw->vs.queue_nr - i); + for ( ; j < 4; j++) { + elts[j] = elts[0]; + dests[j] = dests[0]; + } + assert(n > 0); assert(n <= 4); -- cgit v1.2.3 From 82d9063708539d53c7670b2ab732bed24230b94d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Jan 2008 15:50:02 +0000 Subject: gallium: handle flatshading explicitly in clipper stage We can do a better job in the clip stage than just relying on the brute-force approach of copying colors to all incoming vertices applied in the flatshade stage. At very least, it is only necessary to do this in the clipper when a primitive is actually being clipped. --- src/mesa/pipe/draw/draw_clip.c | 136 +++++++++++++++++++++++++------------ src/mesa/pipe/draw/draw_validate.c | 10 +-- 2 files changed, 99 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 2d410e3244..da20028904 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -33,6 +33,8 @@ #include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + #include "draw_context.h" #include "draw_private.h" @@ -54,6 +56,12 @@ struct clipper { struct draw_stage stage; /**< base class */ + /* Basically duplicate some of the flatshading logic here: + */ + boolean flat; + uint num_color_attribs; + uint color_attribs[4]; /* front/back primary/secondary colors */ + float (*plane)[4]; }; @@ -82,6 +90,17 @@ static void interp_attr( float *fdst, fdst[3] = LINTERP( t, fout[3], fin[3] ); } +static void copy_colors( struct draw_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + const struct clipper *clipper = clipper_stage(stage); + uint i; + for (i = 0; i < clipper->num_color_attribs; i++) { + const uint attr = clipper->color_attribs[i]; + COPY_4FV(dst->data[attr], src->data[attr]); + } +} @@ -134,27 +153,11 @@ static void interp( const struct clipper *clip, } } -#if 0 -static INLINE void do_tri( struct draw_stage *next, - struct prim_header *header ) -{ - unsigned i; - for (i = 0; i < 3; i++) { - float *ndc = header->v[i]->data[0]; - _mesa_printf("ndc %f %f %f\n", ndc[0], ndc[1], ndc[2]); - assert(ndc[0] >= -1 && ndc[0] <= 641); - assert(ndc[1] >= 30 && ndc[1] <= 481); - } - _mesa_printf("\n"); - next->tri(next, header); -} -#endif - static void emit_poly( struct draw_stage *stage, struct vertex_header **inlist, unsigned n, - const struct prim_header *origPrim) + const struct prim_header *origPrim) { struct prim_header header; unsigned i; @@ -163,16 +166,16 @@ static void emit_poly( struct draw_stage *stage, header.det = origPrim->det; for (i = 2; i < n; i++) { - header.v[0] = inlist[0]; - header.v[1] = inlist[i-1]; - header.v[2] = inlist[i]; + header.v[0] = inlist[i-1]; + header.v[1] = inlist[i]; + header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ { - unsigned tmp0 = header.v[0]->edgeflag; + unsigned tmp1 = header.v[1]->edgeflag; unsigned tmp2 = header.v[2]->edgeflag; - if (i != 2) header.v[0]->edgeflag = 0; - if (i != n-1) header.v[2]->edgeflag = 0; + if (i != n-1) header.v[1]->edgeflag = 0; + if (i != 2) header.v[2]->edgeflag = 0; header.edgeflags = ((header.v[0]->edgeflag << 0) | (header.v[1]->edgeflag << 1) | @@ -180,27 +183,13 @@ static void emit_poly( struct draw_stage *stage, stage->next->tri( stage->next, &header ); - header.v[0]->edgeflag = tmp0; + header.v[1]->edgeflag = tmp1; header.v[2]->edgeflag = tmp2; } } } -#if 0 -static void emit_poly( struct draw_stage *stage ) -{ - unsigned i; - - for (i = 2; i < n; i++) { - header->v[0] = inlist[0]; - header->v[1] = inlist[i-1]; - header->v[2] = inlist[i]; - - stage->next->tri( stage->next, header ); - } -} -#endif /* Clip a triangle against the viewport and user clip planes. @@ -281,6 +270,18 @@ do_clip_tri( struct draw_stage *stage, } } + /* If flat-shading, copy color to new provoking vertex. + */ + if (clipper->flat && inlist[0] != header->v[2]) { + if (1) { + inlist[0] = dup_vert(stage, inlist[0], tmpnr++); + } + + copy_colors(stage, inlist[0], header->v[2]); + } + + + /* Emit the polygon as triangles to the setup stage: */ if (n >= 3) @@ -328,6 +329,10 @@ do_clip_line( struct draw_stage *stage, if (v0->clipmask) { interp( clipper, stage->tmp[0], t0, v0, v1 ); + + if (clipper->flat) + copy_colors(stage, stage->tmp[0], v0); + newprim.v[0] = stage->tmp[0]; } else { @@ -393,8 +398,55 @@ clip_tri( struct draw_stage *stage, } } -static void clip_flush( struct draw_stage *stage, unsigned flags ) +/* Update state. Could further delay this until we hit the first + * primitive that really requires clipping. + */ +static void +clip_init_state( struct draw_stage *stage ) +{ + struct clipper *clipper = clipper_stage( stage ); + + clipper->flat = stage->draw->rasterizer->flatshade; + + if (clipper->flat) { + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + clipper->num_color_attribs = 0; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + clipper->color_attribs[clipper->num_color_attribs++] = i; + } + } + } + + stage->tri = clip_tri; + stage->line = clip_line; +} + + + +static void clip_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + clip_init_state( stage ); + stage->tri( stage, header ); +} + +static void clip_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + clip_init_state( stage ); + stage->line( stage, header ); +} + + +static void clip_flush( struct draw_stage *stage, + unsigned flags ) { + stage->tri = clip_first_tri; + stage->line = clip_first_line; stage->next->flush( stage->next, flags ); } @@ -420,12 +472,12 @@ struct draw_stage *draw_clip_stage( struct draw_context *draw ) { struct clipper *clipper = CALLOC_STRUCT(clipper); - draw_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES ); + draw_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); clipper->stage.draw = draw; clipper->stage.point = clip_point; - clipper->stage.line = clip_line; - clipper->stage.tri = clip_tri; + clipper->stage.line = clip_first_line; + clipper->stage.tri = clip_first_tri; clipper->stage.flush = clip_flush; clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; clipper->stage.destroy = clip_destroy; diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c index 86d5a5f814..4375ebabbc 100644 --- a/src/mesa/pipe/draw/draw_validate.c +++ b/src/mesa/pipe/draw/draw_validate.c @@ -78,6 +78,11 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) precalc_flat = 1; /* only needed for triangles really */ need_det = 1; } + + if (draw->rasterizer->flatshade && precalc_flat) { + draw->pipeline.flatshade->next = next; + next = draw->pipeline.flatshade; + } if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) { @@ -110,13 +115,8 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) { draw->pipeline.clip->next = next; next = draw->pipeline.clip; - precalc_flat = 1; /* XXX: FIX ME! Only needed for clipped prims */ } - if (draw->rasterizer->flatshade && precalc_flat) { - draw->pipeline.flatshade->next = next; - next = draw->pipeline.flatshade; - } draw->pipeline.first = next; return next; -- cgit v1.2.3 From 16ed55c6412d2bdc5bff78e99114490223fb4afe Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 09:23:29 -0700 Subject: gallium: check if surface has defined status in check_clear_depth_with_quad() This was part of Keith's patch from Friday. --- src/mesa/state_tracker/st_cb_clear.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 758d4a4086..0cd469c156 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -408,7 +408,9 @@ check_clear_depth_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) const struct st_renderbuffer *strb = st_renderbuffer(rb); const GLboolean isDS = is_depth_stencil_format(strb->surface->format); return ctx->Scissor.Enabled - || (isDS && ctx->DrawBuffer->Visual.stencilBits > 0); + || (isDS && + strb->surface->status == PIPE_SURFACE_STATUS_DEFINED && + ctx->DrawBuffer->Visual.stencilBits > 0); } -- cgit v1.2.3 From 2194675196260c0a5d44242d698b85c86f84074b Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 27 Jan 2008 12:01:47 -0700 Subject: Cell: generalize the batch buffer code for vertex buffers... --- src/mesa/pipe/cell/common.h | 8 ++-- src/mesa/pipe/cell/ppu/cell_batch.c | 84 ++++++++++++++++++++--------------- src/mesa/pipe/cell/ppu/cell_batch.h | 3 ++ src/mesa/pipe/cell/ppu/cell_context.c | 5 ++- src/mesa/pipe/cell/ppu/cell_context.h | 10 +++-- src/mesa/pipe/cell/ppu/cell_spu.c | 4 +- src/mesa/pipe/cell/spu/spu_main.c | 22 ++++----- 7 files changed, 79 insertions(+), 57 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 0b63ed39be..ce9c381907 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -81,8 +81,8 @@ #define CELL_CMD_STATE_VERTEX_INFO 13 -#define CELL_NUM_BATCH_BUFFERS 3 -#define CELL_BATCH_BUFFER_SIZE 1024 /**< 16KB would be the max */ +#define CELL_NUM_BUFFERS 4 +#define CELL_BUFFER_SIZE (4*1024) /**< 16KB would be the max */ #define CELL_BUFFER_STATUS_FREE 10 #define CELL_BUFFER_STATUS_USED 20 @@ -147,7 +147,9 @@ struct cell_init_info unsigned id; unsigned num_spus; struct cell_command *cmd; - ubyte *batch_buffers[CELL_NUM_BATCH_BUFFERS]; + + /** Buffers for command batches, vertex/index data */ + ubyte *buffers[CELL_NUM_BUFFERS]; uint *buffer_status; /**< points at cell_context->buffer_status */ } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index c894ef8608..178caa74e1 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -31,12 +31,46 @@ #include "cell_spu.h" + +uint +cell_get_empty_buffer(struct cell_context *cell) +{ + uint buf = 0; + + /* Find a buffer that's marked as free by all SPUs */ + while (1) { + uint spu, num_free = 0; + + for (spu = 0; spu < cell->num_spus; spu++) { + if (cell->buffer_status[spu][buf][0] == CELL_BUFFER_STATUS_FREE) { + num_free++; + + if (num_free == cell->num_spus) { + /* found a free buffer, now mark status as used */ + for (spu = 0; spu < cell->num_spus; spu++) { + cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED; + } + return buf; + } + } + else { + break; + } + } + + /* try next buf */ + buf = (buf + 1) % CELL_NUM_BUFFERS; + } +} + + + void cell_batch_flush(struct cell_context *cell) { static boolean flushing = FALSE; uint batch = cell->cur_batch; - const uint size = cell->batch_buffer_size[batch]; + const uint size = cell->buffer_size[batch]; uint spu, cmd_word; assert(!flushing); @@ -46,7 +80,7 @@ cell_batch_flush(struct cell_context *cell) flushing = TRUE; - assert(batch < CELL_NUM_BATCH_BUFFERS); + assert(batch < CELL_NUM_BUFFERS); /* printf("cell_batch_dispatch: buf %u at %p, size %u\n", @@ -68,28 +102,9 @@ cell_batch_flush(struct cell_context *cell) * array indicating that the PPU can re-use the buffer. */ + batch = cell_get_empty_buffer(cell); - /* Find a buffer that's marked as free by all SPUs */ - while (1) { - uint num_free = 0; - - batch = (batch + 1) % CELL_NUM_BATCH_BUFFERS; - - for (spu = 0; spu < cell->num_spus; spu++) { - if (cell->buffer_status[spu][batch][0] == CELL_BUFFER_STATUS_FREE) - num_free++; - } - - if (num_free == cell->num_spus) { - /* found a free buffer, now mark status as used */ - for (spu = 0; spu < cell->num_spus; spu++) { - cell->buffer_status[spu][batch][0] = CELL_BUFFER_STATUS_USED; - } - break; - } - } - - cell->batch_buffer_size[batch] = 0; /* empty */ + cell->buffer_size[batch] = 0; /* empty */ cell->cur_batch = batch; flushing = FALSE; @@ -99,8 +114,7 @@ cell_batch_flush(struct cell_context *cell) uint cell_batch_free_space(const struct cell_context *cell) { - uint free = CELL_BATCH_BUFFER_SIZE - - cell->batch_buffer_size[cell->cur_batch]; + uint free = CELL_BUFFER_SIZE - cell->buffer_size[cell->cur_batch]; return free; } @@ -117,18 +131,18 @@ cell_batch_append(struct cell_context *cell, const void *cmd, uint length) assert(length % 4 == 0); assert(cell->cur_batch >= 0); - size = cell->batch_buffer_size[cell->cur_batch]; + size = cell->buffer_size[cell->cur_batch]; - if (size + length > CELL_BATCH_BUFFER_SIZE) { + if (size + length > CELL_BUFFER_SIZE) { cell_batch_flush(cell); size = 0; } - assert(size + length <= CELL_BATCH_BUFFER_SIZE); + assert(size + length <= CELL_BUFFER_SIZE); - memcpy(cell->batch_buffer[cell->cur_batch] + size, cmd, length); + memcpy(cell->buffer[cell->cur_batch] + size, cmd, length); - cell->batch_buffer_size[cell->cur_batch] = size + length; + cell->buffer_size[cell->cur_batch] = size + length; } @@ -142,18 +156,18 @@ cell_batch_alloc(struct cell_context *cell, uint bytes) assert(cell->cur_batch >= 0); - size = cell->batch_buffer_size[cell->cur_batch]; + size = cell->buffer_size[cell->cur_batch]; - if (size + bytes > CELL_BATCH_BUFFER_SIZE) { + if (size + bytes > CELL_BUFFER_SIZE) { cell_batch_flush(cell); size = 0; } - assert(size + bytes <= CELL_BATCH_BUFFER_SIZE); + assert(size + bytes <= CELL_BUFFER_SIZE); - pos = (void *) (cell->batch_buffer[cell->cur_batch] + size); + pos = (void *) (cell->buffer[cell->cur_batch] + size); - cell->batch_buffer_size[cell->cur_batch] = size + bytes; + cell->buffer_size[cell->cur_batch] = size + bytes; return pos; } diff --git a/src/mesa/pipe/cell/ppu/cell_batch.h b/src/mesa/pipe/cell/ppu/cell_batch.h index c4ba7feb3d..b4c96f465a 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.h +++ b/src/mesa/pipe/cell/ppu/cell_batch.h @@ -35,6 +35,9 @@ struct cell_context; +extern uint +cell_get_empty_buffer(struct cell_context *cell); + extern void cell_batch_flush(struct cell_context *cell); diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 8cb0c48f40..e8020a49bc 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -254,8 +254,9 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell_start_spus(cell); - for (buf = 0; buf < CELL_NUM_BATCH_BUFFERS; buf++) { - cell->batch_buffer_size[buf] = 0; + /* init command, vertex/index buffer info */ + for (buf = 0; buf < CELL_NUM_BUFFERS; buf++) { + cell->buffer_size[buf] = 0; /* init batch buffer status values, * mark 0th buffer as used, rest as free. diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 3bd88bfd5b..de65fb5e9a 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -102,12 +102,14 @@ struct cell_context uint num_spus; - uint batch_buffer_size[CELL_NUM_BATCH_BUFFERS]; - ubyte batch_buffer[CELL_NUM_BATCH_BUFFERS][CELL_BATCH_BUFFER_SIZE] ALIGN16_ATTRIB; - int cur_batch; /**< which batch buffer is being filled */ + /** Buffers for command batches, vertex/index data */ + uint buffer_size[CELL_NUM_BUFFERS]; + ubyte buffer[CELL_NUM_BUFFERS][CELL_BUFFER_SIZE] ALIGN16_ATTRIB; + + int cur_batch; /**< which buffer is being filled w/ commands */ /** [4] to ensure 16-byte alignment for each status word */ - uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BATCH_BUFFERS][4] ALIGN16_ATTRIB; + uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BUFFERS][4] ALIGN16_ATTRIB; }; diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c index 4627bc8d1f..7c83a47e57 100644 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ b/src/mesa/pipe/cell/ppu/cell_spu.c @@ -111,8 +111,8 @@ cell_start_spus(struct cell_context *cell) cell_global.inits[i].id = i; cell_global.inits[i].num_spus = cell->num_spus; cell_global.inits[i].cmd = &cell_global.command[i]; - for (j = 0; j < CELL_NUM_BATCH_BUFFERS; j++) { - cell_global.inits[i].batch_buffers[j] = cell->batch_buffer[j]; + for (j = 0; j < CELL_NUM_BUFFERS; j++) { + cell_global.inits[i].buffers[j] = cell->buffer[j]; } cell_global.inits[i].buffer_status = &cell->buffer_status[0][0][0]; diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 0c83900a18..2097683b82 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -473,22 +473,22 @@ cmd_finish(void) /** - * Tell the PPU that this SPU has finished copying a batch buffer to + * Tell the PPU that this SPU has finished copying a buffer to * local store and that it may be reused by the PPU. * This is done by writting a 16-byte batch-buffer-status block back into - * main memory (in cell_contex->buffer_status[]). + * main memory (in cell_context->buffer_status[]). */ static void -release_batch_buffer(uint buffer) +release_buffer(uint buffer) { /* Evidently, using less than a 16-byte status doesn't work reliably */ static const uint status[4] ALIGN16_ATTRIB = {CELL_BUFFER_STATUS_FREE, 0, 0, 0}; - const uint index = 4 * (spu.init.id * CELL_NUM_BATCH_BUFFERS + buffer); + const uint index = 4 * (spu.init.id * CELL_NUM_BUFFERS + buffer); uint *dst = spu.init.buffer_status + index; - ASSERT(buffer < CELL_NUM_BATCH_BUFFERS); + ASSERT(buffer < CELL_NUM_BUFFERS); /* printf("SPU %u: Set batch status buf=%u, index %u, at %p to FREE\n", @@ -513,24 +513,24 @@ cmd_batch(uint opcode) { const uint buf = (opcode >> 8) & 0xff; uint size = (opcode >> 16); - uint buffer[CELL_BATCH_BUFFER_SIZE / 4] ALIGN16_ATTRIB; + uint buffer[CELL_BUFFER_SIZE / 4] ALIGN16_ATTRIB; const uint usize = size / sizeof(uint); uint pos; if (Debug) printf("SPU %u: BATCH buffer %u, len %u, from %p\n", - spu.init.id, buf, size, spu.init.batch_buffers[buf]); + spu.init.id, buf, size, spu.init.buffers[buf]); ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH); - ASSERT_ALIGN16(spu.init.batch_buffers[buf]); + ASSERT_ALIGN16(spu.init.buffers[buf]); size = ROUNDUP16(size); - ASSERT_ALIGN16(spu.init.batch_buffers[buf]); + ASSERT_ALIGN16(spu.init.buffers[buf]); mfc_get(buffer, /* dest */ - (unsigned int) spu.init.batch_buffers[buf], /* src */ + (unsigned int) spu.init.buffers[buf], /* src */ size, TAG_BATCH_BUFFER, 0, /* tid */ @@ -538,7 +538,7 @@ cmd_batch(uint opcode) wait_on_mask(1 << TAG_BATCH_BUFFER); /* Tell PPU we're done copying the buffer to local store */ - release_batch_buffer(buf); + release_buffer(buf); for (pos = 0; pos < usize; /* no incr */) { switch (buffer[pos]) { -- cgit v1.2.3 From aaea9a121bc739db87e539214c23f76d4cd5bf49 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 09:57:13 -0700 Subject: Cell: additional debug code, misc clean-up --- src/mesa/pipe/cell/ppu/cell_batch.c | 52 +++++++++++++++++++++++++++++-------- src/mesa/pipe/cell/ppu/cell_batch.h | 2 +- 2 files changed, 42 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index 178caa74e1..2d032fc902 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -35,7 +35,7 @@ uint cell_get_empty_buffer(struct cell_context *cell) { - uint buf = 0; + uint buf = 0, tries = 0; /* Find a buffer that's marked as free by all SPUs */ while (1) { @@ -50,6 +50,9 @@ cell_get_empty_buffer(struct cell_context *cell) for (spu = 0; spu < cell->num_spus; spu++) { cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED; } + /* + printf("PPU: ALLOC BUFFER %u\n", buf); + */ return buf; } } @@ -60,11 +63,17 @@ cell_get_empty_buffer(struct cell_context *cell) /* try next buf */ buf = (buf + 1) % CELL_NUM_BUFFERS; + + tries++; + if (tries == 100) { + /* + printf("PPU WAITING for buffer...\n"); + */ + } } } - void cell_batch_flush(struct cell_context *cell) { @@ -120,29 +129,39 @@ cell_batch_free_space(const struct cell_context *cell) /** - * \param cmd command to append - * \param length command size in bytes + * Append data to current batch. */ void -cell_batch_append(struct cell_context *cell, const void *cmd, uint length) +cell_batch_append(struct cell_context *cell, const void *data, uint bytes) { uint size; - assert(length % 4 == 0); - assert(cell->cur_batch >= 0); + ASSERT(bytes % 4 == 0); + ASSERT(bytes <= CELL_BUFFER_SIZE); + ASSERT(cell->cur_batch >= 0); + +#ifdef ASSERT + { + uint spu; + for (spu = 0; spu < cell->num_spus; spu++) { + ASSERT(cell->buffer_status[spu][cell->cur_batch][0] + == CELL_BUFFER_STATUS_USED); + } + } +#endif size = cell->buffer_size[cell->cur_batch]; - if (size + length > CELL_BUFFER_SIZE) { + if (size + bytes > CELL_BUFFER_SIZE) { cell_batch_flush(cell); size = 0; } - assert(size + length <= CELL_BUFFER_SIZE); + assert(size + bytes <= CELL_BUFFER_SIZE); - memcpy(cell->buffer[cell->cur_batch] + size, cmd, length); + memcpy(cell->buffer[cell->cur_batch] + size, data, bytes); - cell->buffer_size[cell->cur_batch] = size + length; + cell->buffer_size[cell->cur_batch] = size + bytes; } @@ -153,9 +172,20 @@ cell_batch_alloc(struct cell_context *cell, uint bytes) uint size; ASSERT(bytes % 4 == 0); + ASSERT(bytes <= CELL_BUFFER_SIZE); assert(cell->cur_batch >= 0); +#ifdef ASSERT + { + uint spu; + for (spu = 0; spu < cell->num_spus; spu++) { + ASSERT(cell->buffer_status[spu][cell->cur_batch][0] + == CELL_BUFFER_STATUS_USED); + } + } +#endif + size = cell->buffer_size[cell->cur_batch]; if (size + bytes > CELL_BUFFER_SIZE) { diff --git a/src/mesa/pipe/cell/ppu/cell_batch.h b/src/mesa/pipe/cell/ppu/cell_batch.h index b4c96f465a..f4f37314a4 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.h +++ b/src/mesa/pipe/cell/ppu/cell_batch.h @@ -45,7 +45,7 @@ extern uint cell_batch_free_space(const struct cell_context *cell); extern void -cell_batch_append(struct cell_context *cell, const void *cmd, uint length); +cell_batch_append(struct cell_context *cell, const void *data, uint bytes); extern void * cell_batch_alloc(struct cell_context *cell, uint bytes); -- cgit v1.2.3 From 200dcb4760960f0d9c74a7053de63337e93dd85b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 09:57:51 -0700 Subject: Cell: If flushing for swapbuffers, wait for frame completion --- src/mesa/pipe/cell/ppu/cell_flush.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c index b98bb566b1..cf4e676645 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -39,6 +39,9 @@ cell_flush(struct pipe_context *pipe, unsigned flags) { struct cell_context *cell = cell_context(pipe); + if (flags & PIPE_FLUSH_SWAPBUFFERS) + flags |= PIPE_FLUSH_WAIT; + draw_flush( cell->draw ); cell_flush_int(pipe, flags); } -- cgit v1.2.3 From 7024019d4e6e2a1618e910a127bea8c3b7661a54 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 10:00:27 -0700 Subject: Cell: checkpoint commit: always inline prim indexes into batch buffer Also, explicit release-vertex-buffer command. Lots of debug/stale code still in place... --- src/mesa/pipe/cell/common.h | 12 ++++ src/mesa/pipe/cell/ppu/cell_vbuf.c | 113 ++++++++++++++++++++++++++----------- src/mesa/pipe/cell/spu/spu_main.c | 110 +++++++++++++++++++++++++----------- src/mesa/pipe/cell/spu/spu_main.h | 2 + 4 files changed, 171 insertions(+), 66 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index ce9c381907..31637ed1cc 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -75,6 +75,7 @@ #define CELL_CMD_FINISH 3 #define CELL_CMD_RENDER 4 #define CELL_CMD_BATCH 5 +#define CELL_CMD_RELEASE_VERTS 6 #define CELL_CMD_STATE_FRAMEBUFFER 10 #define CELL_CMD_STATE_DEPTH_STENCIL 11 #define CELL_CMD_STATE_SAMPLER 12 @@ -124,7 +125,11 @@ struct cell_command_render uint vertex_size; /**< bytes per vertex */ uint dummy; /* XXX this dummy field works around a compiler bug */ uint num_indexes; +#if 0 const void *vertex_data; +#else + uint vertex_buf; /**< which cell->buffer[] contains the vertex data */ +#endif const ushort *index_data; float xmin, ymin, xmax, ymax; boolean inline_indexes; @@ -132,6 +137,13 @@ struct cell_command_render } ALIGN16_ATTRIB; +struct cell_command_release_verts +{ + int opcode; /**< CELL_CMD_RELEASE_VERTS */ + uint vertex_buf; /**< in [0, CELL_NUM_BUFFERS-1] */ +}; + + /** XXX unions don't seem to work */ struct cell_command { diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index ee572b3a51..6e12e16fe0 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -40,8 +40,8 @@ /** Allow prim indexes, verts to be inlined after RENDER command */ -#define ALLOW_INLINE_INDEXES 1 -#define ALLOW_INLINE_VERTS 1 +#define ALLOW_INLINE_INDEXES 01 +#define ALLOW_INLINE_VERTS 0 /** @@ -55,6 +55,9 @@ struct cell_vbuf_render uint prim; uint vertex_size; void *vertex_buffer; +#if 1 + uint vertex_buf; +#endif }; @@ -81,13 +84,52 @@ cell_vbuf_allocate_vertices(struct vbuf_render *vbr, { struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); /*printf("Alloc verts %u * %u\n", vertex_size, nr_vertices);*/ +#if 0 assert(!cvbr->vertex_buffer); cvbr->vertex_buffer = align_malloc(vertex_size * nr_vertices, 16); +#else + assert(cvbr->vertex_buf == ~0); + cvbr->vertex_buf = cell_get_empty_buffer(cvbr->cell); + cvbr->vertex_buffer = cvbr->cell->buffer[cvbr->vertex_buf]; + printf("%s vertex_buf = %u\n", __FUNCTION__, cvbr->vertex_buf); +#endif cvbr->vertex_size = vertex_size; return cvbr->vertex_buffer; } +static void +cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, + unsigned vertex_size, unsigned vertices_used) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + struct cell_context *cell = cvbr->cell; + + /*printf("Free verts %u * %u\n", vertex_size, vertices_used);*/ +#if 0 + align_free(vertices); +#else + printf("%s vertex_buf = %u count = %u\n", + __FUNCTION__, cvbr->vertex_buf, vertices_used); + + { + struct cell_command_release_verts *release + = (struct cell_command_release_verts *) + cell_batch_alloc(cell, sizeof(struct cell_command_release_verts)); + release->opcode = CELL_CMD_RELEASE_VERTS; + release->vertex_buf = cvbr->vertex_buf; + } + + cvbr->vertex_buf = ~0; + cell_flush_int(&cell->pipe, 0x0);/*NEW*/ +#endif + + assert(vertices == cvbr->vertex_buffer); + cvbr->vertex_buffer = NULL; +} + + + static void cell_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) { @@ -124,7 +166,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, printf("%u %u %u, ", indices[i+0], indices[i+1], indices[i+2]); } printf("\n"); -#elif 0 +#elif 01 printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u indexes = [%u %u %u ...]\n", nr_indices, nr_vertices, indices[0], indices[1], indices[2]); @@ -157,28 +199,26 @@ cell_vbuf_draw(struct vbuf_render *vbr, const uint index_bytes = ROUNDUP4(nr_indices * 2); const uint vertex_bytes = nr_vertices * 4 * cell->vertex_info.size; + const uint batch_size = sizeof(struct cell_command_render) + + index_bytes; + struct cell_command_render *render = (struct cell_command_render *) - cell_batch_alloc(cell, sizeof(*render)); + cell_batch_alloc(cell, batch_size); + render->opcode = CELL_CMD_RENDER; render->prim_type = cvbr->prim; render->num_indexes = nr_indices; - if (ALLOW_INLINE_INDEXES && - index_bytes <= cell_batch_free_space(cell)) { - /* indices inlined, right after render cmd */ - void *dst = cell_batch_alloc(cell, index_bytes); - memcpy(dst, indices, nr_indices * 2); - render->inline_indexes = TRUE; - render->index_data = NULL; - } - else { - /* indices in separate buffer */ - render->inline_indexes = FALSE; - render->index_data = indices; - ASSERT_ALIGN16(render->index_data); - } + /* append indices after render command */ + memcpy(render + 1, indices, nr_indices * 2); + render->inline_indexes = TRUE; + render->index_data = NULL; + + /* if there's room, append vertices after the indices, else leave + * vertices in the original/separate buffer. + */ render->vertex_size = 4 * cell->vertex_info.size; render->num_verts = nr_vertices; if (ALLOW_INLINE_VERTS && @@ -188,12 +228,21 @@ cell_vbuf_draw(struct vbuf_render *vbr, void *dst = cell_batch_alloc(cell, vertex_bytes); memcpy(dst, vertices, vertex_bytes); render->inline_verts = TRUE; +#if 0 render->vertex_data = NULL; +#else + render->vertex_buf = ~0; +#endif } else { render->inline_verts = FALSE; +#if 0 render->vertex_data = vertices; ASSERT_ALIGN16(render->vertex_data); +#else + ASSERT(cvbr->vertex_buf >= 0); + render->vertex_buf = cvbr->vertex_buf; +#endif } @@ -203,27 +252,13 @@ cell_vbuf_draw(struct vbuf_render *vbr, render->ymax = ymax; } -#if 01 +#if 0 /* XXX this is temporary */ cell_flush_int(&cell->pipe, PIPE_FLUSH_WAIT); #endif } -static void -cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, - unsigned vertex_size, unsigned vertices_used) -{ - struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); - - /*printf("Free verts %u * %u\n", vertex_size, vertices_used);*/ - align_free(vertices); - - assert(vertices == cvbr->vertex_buffer); - cvbr->vertex_buffer = NULL; -} - - static void cell_vbuf_destroy(struct vbuf_render *vbr) { @@ -244,8 +279,17 @@ cell_init_vbuf(struct cell_context *cell) cell->vbuf_render = CALLOC_STRUCT(cell_vbuf_render); +#if 0 cell->vbuf_render->base.max_indices = CELL_MAX_VBUF_INDEXES; cell->vbuf_render->base.max_vertex_buffer_bytes = CELL_MAX_VBUF_SIZE; +#else + cell->vbuf_render->base.max_indices + = (CELL_BUFFER_SIZE + - sizeof(struct cell_command_render) + - sizeof(struct cell_command_release_verts)) + / sizeof(ushort); + cell->vbuf_render->base.max_vertex_buffer_bytes = CELL_BUFFER_SIZE; +#endif cell->vbuf_render->base.get_vertex_info = cell_vbuf_get_vertex_info; cell->vbuf_render->base.allocate_vertices = cell_vbuf_allocate_vertices; @@ -255,6 +299,9 @@ cell_init_vbuf(struct cell_context *cell) cell->vbuf_render->base.destroy = cell_vbuf_destroy; cell->vbuf_render->cell = cell; +#if 1 + cell->vbuf_render->vertex_buf = ~0; +#endif cell->vbuf = draw_vbuf_stage(cell->draw, &cell->vbuf_render->base); } diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 2097683b82..eb979718f8 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -69,6 +69,32 @@ wait_on_mask_all(unsigned tagMask) } +/** + * Tell the PPU that this SPU has finished copying a buffer to + * local store and that it may be reused by the PPU. + * This is done by writting a 16-byte batch-buffer-status block back into + * main memory (in cell_context->buffer_status[]). + */ +static void +release_buffer(uint buffer) +{ + /* Evidently, using less than a 16-byte status doesn't work reliably */ + static const uint status[4] ALIGN16_ATTRIB + = {CELL_BUFFER_STATUS_FREE, 0, 0, 0}; + + const uint index = 4 * (spu.init.id * CELL_NUM_BUFFERS + buffer); + uint *dst = spu.init.buffer_status + index; + + ASSERT(buffer < CELL_NUM_BUFFERS); + + mfc_put((void *) &status, /* src in local memory */ + (unsigned int) dst, /* dst in main memory */ + sizeof(status), /* size */ + TAG_MISC, /* tag is unimportant */ + 0, /* tid */ + 0 /* rid */); +} + /** * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled @@ -237,13 +263,18 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) printf(" bound: %g, %g .. %g, %g\n", render->xmin, render->ymin, render->xmax, render->ymax); */ + /* printf("SPU %u: indices at %p vertices at %p\n", spu.init.id, render->index_data, render->vertex_data); + */ } ASSERT(sizeof(*render) % 4 == 0); +#if 0 ASSERT_ALIGN16(render->vertex_data); +#else +#endif ASSERT_ALIGN16(render->index_data); @@ -251,10 +282,18 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) ** Get vertex, index buffers if not inlined **/ if (!render->inline_verts) { + void *src; ASSERT(total_vertex_bytes % 16 == 0); +#if 0 + src = render->vertex_data; +#else + spu.cur_vertex_buf = render->vertex_buf; + src = spu.init.buffers[render->vertex_buf]; +#endif + mfc_get(vertex_data, /* dest */ - (unsigned int) render->vertex_data, /* src */ + (unsigned int) src, total_vertex_bytes, /* size */ TAG_VERTEX_BUFFER, 0, /* tid */ @@ -298,6 +337,7 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) /* vertices are after indexes, if inlined */ vertices = (const ubyte *) (render + 1) + *pos_incr * 4; *pos_incr = *pos_incr + total_vertex_bytes / 4; + spu.cur_vertex_buf = ~0; } } @@ -310,6 +350,12 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) mask |= (1 << TAG_INDEX_BUFFER); wait_on_mask_all(mask); +#if 0 + if (!render->inline_verts) { + printf("SPU %u: release vbuf %u\n", spu.init.id, render->vertex_buf); + release_buffer(render->vertex_buf); + } +#endif /** ** find tiles which intersect the prim bounding box @@ -359,6 +405,14 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) for (j = 0; j < render->num_indexes; j += 3) { const float *v0, *v1, *v2; + if (indexes[j] == 0xffff) { + printf("index[%u] = 0xffff\n", j); + } + + ASSERT(indexes[j] != 0xffff); + ASSERT(indexes[j+1] != 0xffff); + ASSERT(indexes[j+2] != 0xffff); + v0 = (const float *) (vertices + indexes[j+0] * vertex_size); v1 = (const float *) (vertices + indexes[j+1] * vertex_size); v2 = (const float *) (vertices + indexes[j+2] * vertex_size); @@ -391,6 +445,17 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) } +static void +cmd_release_verts(const struct cell_command_release_verts *release) +{ + if (Debug) + printf("SPU %u: RELEASE VERTS %u\n", + spu.init.id, spu.cur_vertex_buf); + ASSERT(spu.cur_vertex_buf == release->vertex_buf); + release_buffer(release->vertex_buf); +} + + static void cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) { @@ -472,38 +537,6 @@ cmd_finish(void) } -/** - * Tell the PPU that this SPU has finished copying a buffer to - * local store and that it may be reused by the PPU. - * This is done by writting a 16-byte batch-buffer-status block back into - * main memory (in cell_context->buffer_status[]). - */ -static void -release_buffer(uint buffer) -{ - /* Evidently, using less than a 16-byte status doesn't work reliably */ - static const uint status[4] ALIGN16_ATTRIB - = {CELL_BUFFER_STATUS_FREE, 0, 0, 0}; - - const uint index = 4 * (spu.init.id * CELL_NUM_BUFFERS + buffer); - uint *dst = spu.init.buffer_status + index; - - ASSERT(buffer < CELL_NUM_BUFFERS); - - /* - printf("SPU %u: Set batch status buf=%u, index %u, at %p to FREE\n", - spu.init.id, buffer, index, dst); - */ - - mfc_put((void *) &status, /* src in local memory */ - (unsigned int) dst, /* dst in main memory */ - sizeof(status), /* size */ - TAG_MISC, /* tag is unimportant */ - 0, /* tid */ - 0 /* rid */); -} - - /** * Execute a batch of commands * The opcode param encodes the location of the buffer and its size. @@ -538,6 +571,8 @@ cmd_batch(uint opcode) wait_on_mask(1 << TAG_BATCH_BUFFER); /* Tell PPU we're done copying the buffer to local store */ + if (Debug) + printf("SPU %u: release batch buf %u\n", spu.init.id, buf); release_buffer(buf); for (pos = 0; pos < usize; /* no incr */) { @@ -567,6 +602,15 @@ cmd_batch(uint opcode) pos += sizeof(*render) / 4 + pos_incr; } break; + case CELL_CMD_RELEASE_VERTS: + { + struct cell_command_release_verts *release + = (struct cell_command_release_verts *) &buffer[pos]; + cmd_release_verts(release); + ASSERT(sizeof(*release) == 8); + pos += sizeof(*release) / 4; + } + break; case CELL_CMD_FINISH: cmd_finish(); pos += 1; diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 5bc5d9fa99..68c7263b7f 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -65,6 +65,8 @@ struct spu_global /* XXX more state to come */ + uint cur_vertex_buf; + } ALIGN16_ATTRIB; -- cgit v1.2.3 From 5b5ec94663d566b4840975c4ef4740abb138bb12 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 10:41:27 -0700 Subject: Cell: clean-up of render path Finally removed a number of unneeded flush commands. Vertex buffers are allocated from the general buffer pool, freed by SPUs when done. Still an occasional failed assertion (invalid batch buffer command)... --- src/mesa/pipe/cell/common.h | 12 +--- src/mesa/pipe/cell/ppu/cell_vbuf.c | 60 ++++++-------------- src/mesa/pipe/cell/spu/spu_main.c | 112 +++++++------------------------------ src/mesa/pipe/cell/spu/spu_main.h | 2 - 4 files changed, 38 insertions(+), 148 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 31637ed1cc..d6e1dd4f7d 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -68,7 +68,7 @@ * The low byte of a mailbox word contains the command opcode. * Remaining higher bytes are command specific. */ -#define CELL_CMD_OPCODE_MASK 0xf +#define CELL_CMD_OPCODE_MASK 0xff #define CELL_CMD_EXIT 1 #define CELL_CMD_CLEAR_SURFACE 2 @@ -113,10 +113,6 @@ struct cell_command_clear_surface } ALIGN16_ATTRIB; -#define CELL_MAX_VBUF_SIZE (16 * 1024) -#define CELL_MAX_VBUF_INDEXES 1024 - - struct cell_command_render { uint opcode; /**< CELL_CMD_RENDER */ @@ -125,14 +121,8 @@ struct cell_command_render uint vertex_size; /**< bytes per vertex */ uint dummy; /* XXX this dummy field works around a compiler bug */ uint num_indexes; -#if 0 - const void *vertex_data; -#else uint vertex_buf; /**< which cell->buffer[] contains the vertex data */ -#endif - const ushort *index_data; float xmin, ymin, xmax, ymax; - boolean inline_indexes; boolean inline_verts; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 6e12e16fe0..b2a25d767b 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -39,9 +39,8 @@ #include "pipe/draw/draw_vbuf.h" -/** Allow prim indexes, verts to be inlined after RENDER command */ -#define ALLOW_INLINE_INDEXES 01 -#define ALLOW_INLINE_VERTS 0 +/** Allow vertex data to be inlined after RENDER command */ +#define ALLOW_INLINE_VERTS 1 /** @@ -52,12 +51,10 @@ struct cell_vbuf_render { struct vbuf_render base; struct cell_context *cell; - uint prim; - uint vertex_size; - void *vertex_buffer; -#if 1 - uint vertex_buf; -#endif + uint prim; /**< PIPE_PRIM_x */ + uint vertex_size; /**< in bytes */ + void *vertex_buffer; /**< just for debug, really */ + uint vertex_buf; /**< in [0, CELL_NUM_BUFFERS-1] */ }; @@ -84,15 +81,10 @@ cell_vbuf_allocate_vertices(struct vbuf_render *vbr, { struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); /*printf("Alloc verts %u * %u\n", vertex_size, nr_vertices);*/ -#if 0 - assert(!cvbr->vertex_buffer); - cvbr->vertex_buffer = align_malloc(vertex_size * nr_vertices, 16); -#else + assert(cvbr->vertex_buf == ~0); cvbr->vertex_buf = cell_get_empty_buffer(cvbr->cell); cvbr->vertex_buffer = cvbr->cell->buffer[cvbr->vertex_buf]; - printf("%s vertex_buf = %u\n", __FUNCTION__, cvbr->vertex_buf); -#endif cvbr->vertex_size = vertex_size; return cvbr->vertex_buffer; } @@ -105,14 +97,13 @@ cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); struct cell_context *cell = cvbr->cell; - /*printf("Free verts %u * %u\n", vertex_size, vertices_used);*/ -#if 0 - align_free(vertices); -#else + /* printf("%s vertex_buf = %u count = %u\n", __FUNCTION__, cvbr->vertex_buf, vertices_used); + */ - { + /* Tell SPUs they can release the vert buf */ + if (cvbr->vertex_buf != ~0U) { struct cell_command_release_verts *release = (struct cell_command_release_verts *) cell_batch_alloc(cell, sizeof(struct cell_command_release_verts)); @@ -121,8 +112,7 @@ cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, } cvbr->vertex_buf = ~0; - cell_flush_int(&cell->pipe, 0x0);/*NEW*/ -#endif + cell_flush_int(&cell->pipe, 0x0); assert(vertices == cvbr->vertex_buffer); cvbr->vertex_buffer = NULL; @@ -166,7 +156,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, printf("%u %u %u, ", indices[i+0], indices[i+1], indices[i+2]); } printf("\n"); -#elif 01 +#elif 0 printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u indexes = [%u %u %u ...]\n", nr_indices, nr_vertices, indices[0], indices[1], indices[2]); @@ -213,8 +203,6 @@ cell_vbuf_draw(struct vbuf_render *vbr, /* append indices after render command */ memcpy(render + 1, indices, nr_indices * 2); - render->inline_indexes = TRUE; - render->index_data = NULL; /* if there's room, append vertices after the indices, else leave * vertices in the original/separate buffer. @@ -222,30 +210,20 @@ cell_vbuf_draw(struct vbuf_render *vbr, render->vertex_size = 4 * cell->vertex_info.size; render->num_verts = nr_vertices; if (ALLOW_INLINE_VERTS && - render->inline_indexes && vertex_bytes <= cell_batch_free_space(cell)) { /* vertex data inlined, after indices */ void *dst = cell_batch_alloc(cell, vertex_bytes); memcpy(dst, vertices, vertex_bytes); render->inline_verts = TRUE; -#if 0 - render->vertex_data = NULL; -#else render->vertex_buf = ~0; -#endif } else { + /* vertex data in separate buffer */ render->inline_verts = FALSE; -#if 0 - render->vertex_data = vertices; - ASSERT_ALIGN16(render->vertex_data); -#else ASSERT(cvbr->vertex_buf >= 0); render->vertex_buf = cvbr->vertex_buf; -#endif } - render->xmin = xmin; render->ymin = ymin; render->xmax = xmax; @@ -253,7 +231,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, } #if 0 - /* XXX this is temporary */ + /* helpful for debug */ cell_flush_int(&cell->pipe, PIPE_FLUSH_WAIT); #endif } @@ -279,17 +257,15 @@ cell_init_vbuf(struct cell_context *cell) cell->vbuf_render = CALLOC_STRUCT(cell_vbuf_render); -#if 0 - cell->vbuf_render->base.max_indices = CELL_MAX_VBUF_INDEXES; - cell->vbuf_render->base.max_vertex_buffer_bytes = CELL_MAX_VBUF_SIZE; -#else + /* The max number of indexes is what can fix into a batch buffer, + * minus the render and release-verts commands. + */ cell->vbuf_render->base.max_indices = (CELL_BUFFER_SIZE - sizeof(struct cell_command_render) - sizeof(struct cell_command_release_verts)) / sizeof(ushort); cell->vbuf_render->base.max_vertex_buffer_bytes = CELL_BUFFER_SIZE; -#endif cell->vbuf_render->base.get_vertex_info = cell_vbuf_get_vertex_info; cell->vbuf_render->base.allocate_vertices = cell_vbuf_allocate_vertices; diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index eb979718f8..5b50ec6953 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -239,59 +239,45 @@ static void cmd_render(const struct cell_command_render *render, uint *pos_incr) { /* we'll DMA into these buffers */ - ubyte vertex_data[CELL_MAX_VBUF_SIZE] ALIGN16_ATTRIB; - ushort index_data[CELL_MAX_VBUF_INDEXES] ALIGN16_ATTRIB; + ubyte vertex_data[CELL_BUFFER_SIZE] ALIGN16_ATTRIB; const uint vertex_size = render->vertex_size; /* in bytes */ const uint total_vertex_bytes = render->num_verts * vertex_size; const ubyte *vertices; const ushort *indexes; - uint mask; uint i, j; if (Debug) { printf("SPU %u: RENDER prim %u, num_vert=%u num_ind=%u " - "inline_vert=%u inline_ind=%u\n", + "inline_vert=%u\n", spu.init.id, render->prim_type, render->num_verts, render->num_indexes, - render->inline_verts, - render->inline_indexes); + render->inline_verts); /* printf(" bound: %g, %g .. %g, %g\n", render->xmin, render->ymin, render->xmax, render->ymax); */ - /* - printf("SPU %u: indices at %p vertices at %p\n", - spu.init.id, - render->index_data, render->vertex_data); - */ } ASSERT(sizeof(*render) % 4 == 0); -#if 0 - ASSERT_ALIGN16(render->vertex_data); -#else -#endif - ASSERT_ALIGN16(render->index_data); + ASSERT(total_vertex_bytes % 16 == 0); + /* indexes are right after the render command in the batch buffer */ + indexes = (const ushort *) (render + 1); + *pos_incr = (render->num_indexes * 2 + 3) / 4; - /** - ** Get vertex, index buffers if not inlined - **/ - if (!render->inline_verts) { - void *src; - ASSERT(total_vertex_bytes % 16 == 0); - -#if 0 - src = render->vertex_data; -#else - spu.cur_vertex_buf = render->vertex_buf; - src = spu.init.buffers[render->vertex_buf]; -#endif + if (render->inline_verts) { + /* Vertices are right after indexes in batch buffer */ + vertices = (const ubyte *) (render + 1) + *pos_incr * 4; + *pos_incr = *pos_incr + total_vertex_bytes / 4; + } + else { + /* Begin DMA fetch of vertex buffer */ + void *src = spu.init.buffers[render->vertex_buf]; mfc_get(vertex_data, /* dest */ (unsigned int) src, total_vertex_bytes, /* size */ @@ -300,63 +286,11 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) 0 /* rid */); vertices = vertex_data; - } - - if (!render->inline_indexes) { - uint total_index_bytes; - - *pos_incr = 0; - - total_index_bytes = render->num_indexes * sizeof(ushort); - if (total_index_bytes < 16) - total_index_bytes = 16; - else - total_index_bytes = ROUNDUP16(total_index_bytes); - indexes = index_data; - - /* get index data from main memory */ - mfc_get(index_data, /* dest */ - (unsigned int) render->index_data, /* src */ - total_index_bytes, - TAG_INDEX_BUFFER, - 0, /* tid */ - 0 /* rid */); - } - - - /** - ** Get pointers to inlined indexes, verts, if present - **/ - if (render->inline_indexes) { - /* indexes are right after the render command in the batch buffer */ - indexes = (ushort *) (render + 1); - *pos_incr = (render->num_indexes * 2 + 3) / 4; - - if (render->inline_verts) { - /* vertices are after indexes, if inlined */ - vertices = (const ubyte *) (render + 1) + *pos_incr * 4; - *pos_incr = *pos_incr + total_vertex_bytes / 4; - spu.cur_vertex_buf = ~0; - } + wait_on_mask(1 << TAG_VERTEX_BUFFER); } - /* wait for vertex and/or index buffers if not inlined */ - mask = 0x0; - if (!render->inline_verts) - mask |= (1 << TAG_VERTEX_BUFFER); - if (!render->inline_indexes) - mask |= (1 << TAG_INDEX_BUFFER); - wait_on_mask_all(mask); - -#if 0 - if (!render->inline_verts) { - printf("SPU %u: release vbuf %u\n", spu.init.id, render->vertex_buf); - release_buffer(render->vertex_buf); - } -#endif - /** ** find tiles which intersect the prim bounding box **/ @@ -372,7 +306,7 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) #endif /* make sure any pending clears have completed */ - wait_on_mask(1 << TAG_SURFACE_CLEAR); + wait_on_mask(1 << TAG_SURFACE_CLEAR); /* XXX temporary */ /** @@ -405,14 +339,6 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) for (j = 0; j < render->num_indexes; j += 3) { const float *v0, *v1, *v2; - if (indexes[j] == 0xffff) { - printf("index[%u] = 0xffff\n", j); - } - - ASSERT(indexes[j] != 0xffff); - ASSERT(indexes[j+1] != 0xffff); - ASSERT(indexes[j+2] != 0xffff); - v0 = (const float *) (vertices + indexes[j+0] * vertex_size); v1 = (const float *) (vertices + indexes[j+1] * vertex_size); v2 = (const float *) (vertices + indexes[j+2] * vertex_size); @@ -450,8 +376,8 @@ cmd_release_verts(const struct cell_command_release_verts *release) { if (Debug) printf("SPU %u: RELEASE VERTS %u\n", - spu.init.id, spu.cur_vertex_buf); - ASSERT(spu.cur_vertex_buf == release->vertex_buf); + spu.init.id, release->vertex_buf); + ASSERT(release->vertex_buf != ~0U); release_buffer(release->vertex_buf); } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 68c7263b7f..5bc5d9fa99 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -65,8 +65,6 @@ struct spu_global /* XXX more state to come */ - uint cur_vertex_buf; - } ALIGN16_ATTRIB; -- cgit v1.2.3 From 3f8a8eada693c9501b3e52d47986e46028c172b0 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 10:41:51 -0700 Subject: Cell: remove unneeded flush(), dead code --- src/mesa/pipe/cell/ppu/cell_clear.c | 30 +----------------------------- 1 file changed, 1 insertion(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_clear.c b/src/mesa/pipe/cell/ppu/cell_clear.c index e01640b994..e61bfd9b0f 100644 --- a/src/mesa/pipe/cell/ppu/cell_clear.c +++ b/src/mesa/pipe/cell/ppu/cell_clear.c @@ -48,7 +48,6 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { struct cell_context *cell = cell_context(pipe); - /*uint i;*/ uint surfIndex; if (!cell->cbuf_map[0]) @@ -61,29 +60,7 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, surfIndex = 0; } -#if 0 - for (i = 0; i < cell->num_spus; i++) { -#if 1 - uint clr = clearValue; - if (surfIndex == 0) { - /* XXX debug: clear color varied per-SPU to visualize tiles */ - if ((clr & 0xff) == 0) - clr |= 64 + i * 8; - if ((clr & 0xff00) == 0) - clr |= (64 + i * 8) << 8; - if ((clr & 0xff0000) == 0) - clr |= (64 + i * 8) << 16; - if ((clr & 0xff000000) == 0) - clr |= (64 + i * 8) << 24; - } - cell_global.command[i].clear.value = clr; -#else - cell_global.command[i].clear.value = clearValue; -#endif - cell_global.command[i].clear.surface = surfIndex; - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_CLEAR_SURFACE); - } -#else + { struct cell_command_clear_surface *clr = (struct cell_command_clear_surface *) @@ -92,9 +69,4 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, clr->surface = surfIndex; clr->value = clearValue; } -#endif - - /* XXX temporary */ - cell_flush(&cell->pipe, 0x0); - } -- cgit v1.2.3 From a8590e097e965c01ede7df47ff752b0e7adabace Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 11:20:47 -0700 Subject: Cell: make sure state commands aren't split across batches --- src/mesa/pipe/cell/ppu/cell_state_emit.c | 29 ++++++++++++++++++----------- 1 file changed, 18 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c index dbca900c35..6776ec88c7 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.c +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.c @@ -33,6 +33,17 @@ +static void +emit_state_cmd(struct cell_context *cell, uint cmd, + const void *state, uint state_size) +{ + uint *dst = (uint *) cell_batch_alloc(cell, sizeof(uint) + state_size); + *dst = cmd; + memcpy(dst + 1, state, state_size); +} + + + void cell_emit_state(struct cell_context *cell) { @@ -51,22 +62,18 @@ cell_emit_state(struct cell_context *cell) } if (cell->dirty & CELL_NEW_DEPTH_STENCIL) { - uint cmd = CELL_CMD_STATE_DEPTH_STENCIL; - cell_batch_append(cell, &cmd, 4); - cell_batch_append(cell, cell->depth_stencil, - sizeof(struct pipe_depth_stencil_alpha_state)); + emit_state_cmd(cell, CELL_CMD_STATE_DEPTH_STENCIL, + cell->depth_stencil, + sizeof(struct pipe_depth_stencil_alpha_state)); } if (cell->dirty & CELL_NEW_SAMPLER) { - uint cmd = CELL_CMD_STATE_SAMPLER; - cell_batch_append(cell, &cmd, 4); - cell_batch_append(cell, cell->sampler[0], - sizeof(struct pipe_sampler_state)); + emit_state_cmd(cell, CELL_CMD_STATE_SAMPLER, + cell->sampler[0], sizeof(struct pipe_sampler_state)); } if (cell->dirty & CELL_NEW_VERTEX_INFO) { - uint cmd = CELL_CMD_STATE_VERTEX_INFO; - cell_batch_append(cell, &cmd, 4); - cell_batch_append(cell, &cell->vertex_info, sizeof(struct vertex_info)); + emit_state_cmd(cell, CELL_CMD_STATE_VERTEX_INFO, + &cell->vertex_info, sizeof(struct vertex_info)); } } -- cgit v1.2.3 From 3d1b0f4c57edaf5707e4952617dcd6c57dfbdc65 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 11:21:25 -0700 Subject: Cell: additional assertions --- src/mesa/pipe/cell/spu/spu_main.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 5b50ec6953..62f6a357ba 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -441,9 +441,12 @@ cmd_state_sampler(const struct pipe_sampler_state *state) static void cmd_state_vertex_info(const struct vertex_info *vinfo) { - if (Debug) + if (Debug) { printf("SPU %u: VERTEX_INFO num_attribs=%u\n", spu.init.id, vinfo->num_attribs); + } + ASSERT(vinfo->num_attribs >= 1); + ASSERT(vinfo->num_attribs <= 8); memcpy(&spu.vertex_info, vinfo, sizeof(*vinfo)); } -- cgit v1.2.3 From c50ba44095ceef6395727769663ed46c63a1a514 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Jan 2008 20:57:16 +0000 Subject: gallium: fix typos in hardwired fetch path --- src/mesa/pipe/draw/draw_vertex_fetch.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index afdf1971d2..89e4c256a7 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -396,7 +396,7 @@ static void fetch_xyz_rgb_st( struct draw_context *draw, { const float *in = (const float *)(src[2] + elts[i] * pitch[2]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; + float *out = &machine->Inputs[2].xyzw[0].f[i]; out[0] = in[0]; out[4] = in[1]; out[8] = 0.0f; @@ -500,7 +500,7 @@ void draw_update_vertex_fetch( struct draw_context *draw ) case 3: if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32_FLOAT) + draw->vertex_element[2].src_format == PIPE_FORMAT_R32G32_FLOAT) draw->vertex_fetch.fetch_func = fetch_xyz_rgb_st; break; default: -- cgit v1.2.3 From 1e2d6b1b82aaa8bc57535e56c5e6eac9387e22e6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Jan 2008 20:57:58 +0000 Subject: gallium: remove dead vars, code --- src/mesa/pipe/draw/draw_vf_sse.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vf_sse.c b/src/mesa/pipe/draw/draw_vf_sse.c index b238b542e7..066d6c0b7b 100644 --- a/src/mesa/pipe/draw/draw_vf_sse.c +++ b/src/mesa/pipe/draw/draw_vf_sse.c @@ -343,8 +343,6 @@ static boolean build_vertex_emit( struct x86_program *p ) struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI); struct x86_reg temp = x86_make_reg(file_XMM, 0); - struct x86_reg vp0 = x86_make_reg(file_XMM, 1); - struct x86_reg vp1 = x86_make_reg(file_XMM, 2); uint8_t *fixup, *label; /* Push a few regs? @@ -371,7 +369,6 @@ static boolean build_vertex_emit( struct x86_program *p ) /* always load, needed or not: */ - sse_movups(&p->func, p->chan0, x86_make_disp(vfESI, get_offset(vf, &vf->chan_scale[0]))); sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0]))); /* Note address for loop jump */ -- cgit v1.2.3 From f3d0882c0218612a91a6feac91d23b34f6447d8e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 29 Jan 2008 09:34:09 +0900 Subject: gallium: Remove direct dependencies to mesa internals. _mesa_exec_free is still being called. More invasive refactoring is necessary to clean it out. --- src/mesa/pipe/draw/draw_vf.c | 20 ++++++++------ src/mesa/pipe/draw/draw_vf.h | 52 +++++------------------------------- src/mesa/pipe/draw/draw_vf_generic.c | 9 ++++--- src/mesa/pipe/draw/draw_vf_sse.c | 14 +++++----- 4 files changed, 31 insertions(+), 64 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index deedfc7bc7..d36f6293b1 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -25,17 +25,20 @@ * Keith Whitwell */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" #include "pipe/p_compiler.h" +#include "pipe/p_util.h" #include "draw_vf.h" + #define DBG 0 +/* TODO: remove this */ +extern void +_mesa_exec_free( void *addr ); + static boolean match_fastpath( struct draw_vertex_fetch *vf, const struct draw_vf_fastpath *fp) @@ -88,7 +91,7 @@ void draw_vf_register_fastpath( struct draw_vertex_fetch *vf, fastpath->match_strides = match_strides; fastpath->func = vf->emit; fastpath->attr = (struct draw_vf_attr_type *) - _mesa_malloc(vf->attr_count * sizeof(fastpath->attr[0])); + MALLOC(vf->attr_count * sizeof(fastpath->attr[0])); for (i = 0; i < vf->attr_count; i++) { fastpath->attr[i].format = vf->attr[i].format; @@ -156,7 +159,7 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, unsigned offset = 0; unsigned i, j; - assert(nr < DRAW_VF_ATTRIB_MAX); + assert(nr < PIPE_ATTRIB_MAX); memset(vf->lookup, 0, sizeof(vf->lookup)); @@ -200,7 +203,7 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, - +#if 0 /* Set attribute pointers, adjusted for start position: */ void draw_vf_set_sources( struct draw_vertex_fetch *vf, @@ -223,6 +226,7 @@ void draw_vf_set_sources( struct draw_vertex_fetch *vf, a[j].inputptr = ((uint8_t *)vptr->data) + start * vptr->stride; } } +#endif /* Set attribute pointers, adjusted for start position: @@ -260,7 +264,7 @@ struct draw_vertex_fetch *draw_vf_create( void ) struct draw_vertex_fetch *vf = CALLOC_STRUCT(draw_vertex_fetch); unsigned i; - for (i = 0; i < DRAW_VF_ATTRIB_MAX; i++) + for (i = 0; i < PIPE_ATTRIB_MAX; i++) vf->attr[i].vf = vf; vf->identity[0] = 0.0; @@ -271,7 +275,7 @@ struct draw_vertex_fetch *draw_vf_create( void ) vf->codegen_emit = NULL; #ifdef USE_SSE_ASM - if (!_mesa_getenv("MESA_NO_CODEGEN")) + if (!GETENV("MESA_NO_CODEGEN")) vf->codegen_emit = draw_vf_generate_sse_emit; #endif diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h index c6a8fe0d53..7d90f35b0f 100644 --- a/src/mesa/pipe/draw/draw_vf.h +++ b/src/mesa/pipe/draw/draw_vf.h @@ -29,44 +29,11 @@ #define DRAW_VF_H -#include "math/m_vector.h" - #include "pipe/p_compiler.h" -#include "draw_vertex.h" +#include "pipe/p_state.h" +#include "draw_vertex.h" -enum { - DRAW_VF_ATTRIB_POS = 0, - DRAW_VF_ATTRIB_WEIGHT = 1, - DRAW_VF_ATTRIB_NORMAL = 2, - DRAW_VF_ATTRIB_COLOR0 = 3, - DRAW_VF_ATTRIB_COLOR1 = 4, - DRAW_VF_ATTRIB_FOG = 5, - DRAW_VF_ATTRIB_COLOR_INDEX = 6, - DRAW_VF_ATTRIB_EDGEFLAG = 7, - DRAW_VF_ATTRIB_TEX0 = 8, - DRAW_VF_ATTRIB_TEX1 = 9, - DRAW_VF_ATTRIB_TEX2 = 10, - DRAW_VF_ATTRIB_TEX3 = 11, - DRAW_VF_ATTRIB_TEX4 = 12, - DRAW_VF_ATTRIB_TEX5 = 13, - DRAW_VF_ATTRIB_TEX6 = 14, - DRAW_VF_ATTRIB_TEX7 = 15, - DRAW_VF_ATTRIB_VAR0 = 16, - DRAW_VF_ATTRIB_VAR1 = 17, - DRAW_VF_ATTRIB_VAR2 = 18, - DRAW_VF_ATTRIB_VAR3 = 19, - DRAW_VF_ATTRIB_VAR4 = 20, - DRAW_VF_ATTRIB_VAR5 = 21, - DRAW_VF_ATTRIB_VAR6 = 22, - DRAW_VF_ATTRIB_VAR7 = 23, - DRAW_VF_ATTRIB_POINTSIZE = 24, - DRAW_VF_ATTRIB_BFC0 = 25, - DRAW_VF_ATTRIB_BFC1 = 26, - DRAW_VF_ATTRIB_CLIP_POS = 27, - DRAW_VF_ATTRIB_VERTEX_HEADER = 28, - DRAW_VF_ATTRIB_MAX = 29 -}; enum draw_vf_attr_format { DRAW_EMIT_1F, @@ -101,10 +68,12 @@ draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, unsigned nr, unsigned vertex_stride ); +#if 0 void draw_vf_set_sources( struct draw_vertex_fetch *vf, GLvector4f * const attrib[], - unsigned start ); + unsigned start ); +#endif void draw_vf_set_data( struct draw_vertex_fetch *vf, @@ -115,13 +84,6 @@ draw_vf_emit_vertices( struct draw_vertex_fetch *vf, unsigned count, void *dest ); -void -draw_vf_get_attr( struct draw_vertex_fetch *vf, - const void *vertex, - GLenum attr, - const float *dflt, - float *dest ); - struct draw_vertex_fetch * draw_vf_create( void ); @@ -174,11 +136,11 @@ struct draw_vf_attr struct draw_vertex_fetch { - struct draw_vf_attr attr[DRAW_VF_ATTRIB_MAX]; + struct draw_vf_attr attr[PIPE_ATTRIB_MAX]; unsigned attr_count; unsigned vertex_stride; - struct draw_vf_attr *lookup[DRAW_VF_ATTRIB_MAX]; + struct draw_vf_attr *lookup[PIPE_ATTRIB_MAX]; draw_vf_emit_func emit; diff --git a/src/mesa/pipe/draw/draw_vf_generic.c b/src/mesa/pipe/draw/draw_vf_generic.c index 343428d26c..a16eb456b7 100644 --- a/src/mesa/pipe/draw/draw_vf_generic.c +++ b/src/mesa/pipe/draw/draw_vf_generic.c @@ -26,12 +26,13 @@ * Keith Whitwell */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" + +#include + #include "simple_list.h" #include "pipe/p_compiler.h" +#include "pipe/p_util.h" #include "draw_vf.h" @@ -94,7 +95,7 @@ static INLINE void insert_3f_xyw_4( const struct draw_vf_attr *a, uint8_t *v, co static INLINE void insert_3f_xyw_err( const struct draw_vf_attr *a, uint8_t *v, const float *in ) { (void) a; (void) v; (void) in; - _mesa_exit(1); + assert(0); } static INLINE void insert_3f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) diff --git a/src/mesa/pipe/draw/draw_vf_sse.c b/src/mesa/pipe/draw/draw_vf_sse.c index 066d6c0b7b..4036ded1d8 100644 --- a/src/mesa/pipe/draw/draw_vf_sse.c +++ b/src/mesa/pipe/draw/draw_vf_sse.c @@ -25,15 +25,14 @@ * Keith Whitwell */ -#include "glheader.h" -#include "colormac.h" + #include "simple_list.h" -#include "enums.h" #include "pipe/p_compiler.h" #include "draw_vf.h" + #if defined(USE_SSE_ASM) #include "x86/rtasm/x86sse.h" @@ -450,7 +449,8 @@ static boolean build_vertex_emit( struct x86_program *p ) update_src_ptr(p, srcECX, vfESI, a); } else { - _mesa_printf("Can't emit 1ub %x %x %d\n", a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); + fprintf(stderr, "Can't emit 1ub %x %x %d\n", + a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); return FALSE; } break; @@ -495,7 +495,7 @@ static boolean build_vertex_emit( struct x86_program *p ) j++; /* NOTE: two attrs consumed */ } else { - _mesa_printf("Can't emit 3ub\n"); + fprintf(stderr, "Can't emit 3ub\n"); } return FALSE; /* add this later */ break; @@ -528,7 +528,7 @@ static boolean build_vertex_emit( struct x86_program *p ) update_src_ptr(p, srcECX, vfESI, a); break; default: - _mesa_printf("unknown a[%d].format %d\n", j, a->format); + fprintf(stderr, "unknown a[%d].format %d\n", j, a->format); return FALSE; /* catch any new opcodes */ } @@ -577,7 +577,7 @@ void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) return; } - _mesa_memset(&p, 0, sizeof(p)); + memset(&p, 0, sizeof(p)); p.vf = vf; p.inputs_safe = 0; /* for now */ -- cgit v1.2.3 From 93d727eea75812ecb21706804033a33d2e761eb8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 29 Jan 2008 09:41:21 +0900 Subject: gallium: Use CALLOC for pb_buffer to ensure that all fields of pipe_buffer are initiallized. --- src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 3 +-- src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 2 ++ src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 2 +- 3 files changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c index fc83a00f36..2151f1d691 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c @@ -107,10 +107,9 @@ pb_malloc_buffer_create(size_t size, { struct malloc_buffer *buf; - /* TODO: accept an alignment parameter */ /* TODO: do a single allocation */ - buf = (struct malloc_buffer *)MALLOC(sizeof(struct malloc_buffer)); + buf = CALLOC_STRUCT(malloc_buffer); if(!buf) return NULL; diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index 2694f57bca..a2657dac59 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -399,6 +399,8 @@ mm_buffer_destroy(struct pb_buffer *buf) struct mm_buffer *mm_buf = mm_buffer(buf); struct mm_pb_manager *mm = mm_buf->mgr; + assert(buf->base.refcount == 0); + _glthread_LOCK_MUTEX(mm->mutex); mmFreeMem(mm_buf->block); FREE(buf); diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index 7c29954112..f80c7e34c0 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -258,7 +258,7 @@ pool_bufmgr_create(struct pb_manager *provider, if(!pool->map) goto failure; - pool->bufs = (struct pool_buffer *) MALLOC(numBufs * sizeof(*pool->bufs)); + pool->bufs = (struct pool_buffer *)CALLOC(numBufs, sizeof(*pool->bufs)); if (!pool->bufs) goto failure; -- cgit v1.2.3 From 1cc0b0dda7eaf6bdf891d6915b36e7b2ff41133c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 29 Jan 2008 09:42:03 +0900 Subject: gallium: Use GALLIUM_ prefix for env vars. --- src/mesa/pipe/draw/draw_vf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index d36f6293b1..4fc2312ad1 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -275,7 +275,7 @@ struct draw_vertex_fetch *draw_vf_create( void ) vf->codegen_emit = NULL; #ifdef USE_SSE_ASM - if (!GETENV("MESA_NO_CODEGEN")) + if (!GETENV("GALLIUM_NO_CODEGEN")) vf->codegen_emit = draw_vf_generate_sse_emit; #endif -- cgit v1.2.3 From 043fc00a60377f8cd1878e0d0e5157dfb4567289 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 12:46:05 -0700 Subject: Cell: re-enable bounding boxes The geometry bounding box is used to restrict rasterization to just those tiles that are relevant. Note another dummy field had to be added to the cell_command_render struct. Apparently, every 4th word in a struct is susceptible to corruption in some circumstances. Might be a compiler bug. --- src/mesa/pipe/cell/common.h | 2 +- src/mesa/pipe/cell/ppu/cell_vbuf.c | 4 ++++ src/mesa/pipe/cell/spu/spu_main.c | 30 +++++++++++++++++++----------- 3 files changed, 24 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index d6e1dd4f7d..5e32b209e6 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -122,7 +122,7 @@ struct cell_command_render uint dummy; /* XXX this dummy field works around a compiler bug */ uint num_indexes; uint vertex_buf; /**< which cell->buffer[] contains the vertex data */ - float xmin, ymin, xmax, ymax; + float xmin, dummy2, ymin, xmax, ymax; /* XXX another dummy field */ boolean inline_verts; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index b2a25d767b..9f737287ad 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -180,6 +180,10 @@ cell_vbuf_draw(struct vbuf_render *vbr, if (v[1] > ymax) ymax = v[1]; } +#if 0 + printf("PPU Bounds %g, %g .. %g, %g\n", xmin, ymin, xmax, ymax); + fflush(stdout); +#endif if (cvbr->prim != PIPE_PRIM_TRIANGLES) return; /* only render tris for now */ diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 62f6a357ba..c2b05ed5a2 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -200,7 +200,7 @@ tile_bounding_box(const struct cell_command_render *render, uint *txmin, uint *tymin, uint *box_num_tiles, uint *box_width_tiles) { -#if 1 +#if 0 /* Debug: full-window bounding box */ uint txmax = spu.fb.width_tiles - 1; uint tymax = spu.fb.height_tiles - 1; @@ -223,13 +223,24 @@ tile_bounding_box(const struct cell_command_render *render, *box_num_tiles = *box_width_tiles * box_height_tiles; #endif #if 0 - printf("Render bounds: %g, %g ... %g, %g\n", + printf("SPU %u: bounds: %g, %g ... %g, %g\n", spu.init.id, render->xmin, render->ymin, render->xmax, render->ymax); - printf("Render tiles: %u, %u .. %u, %u\n", *txmin, *tymin, txmax, tymax); + printf("SPU %u: tiles: %u, %u .. %u, %u\n", + spu.init.id, *txmin, *tymin, txmax, tymax); + ASSERT(render->xmin <= render->xmax); + ASSERT(render->ymin <= render->ymax); #endif } +/** Check if the tile at (tx,ty) belongs to this SPU */ +static INLINE boolean +my_tile(uint tx, uint ty) +{ + return (spu.fb.width_tiles * ty + tx) % spu.init.num_spus == spu.init.id; +} + + /** * Render primitives * \param pos_incr returns value indicating how may words to skip after @@ -295,15 +306,9 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) ** find tiles which intersect the prim bounding box **/ uint txmin, tymin, box_width_tiles, box_num_tiles; -#if 0 tile_bounding_box(render, &txmin, &tymin, &box_num_tiles, &box_width_tiles); -#else - txmin = 0; - tymin = 0; - box_num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; - box_width_tiles = spu.fb.width_tiles; -#endif + /* make sure any pending clears have completed */ wait_on_mask(1 << TAG_SURFACE_CLEAR); /* XXX temporary */ @@ -312,13 +317,16 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) /** ** loop over tiles, rendering tris **/ - for (i = spu.init.id; i < box_num_tiles; i += spu.init.num_spus) { + for (i = 0; i < box_num_tiles; i++) { const uint tx = txmin + i % box_width_tiles; const uint ty = tymin + i / box_width_tiles; ASSERT(tx < spu.fb.width_tiles); ASSERT(ty < spu.fb.height_tiles); + if (!my_tile(tx, ty)) + continue; + /* Start fetching color/z tiles. We'll wait for completion when * we need read/write to them later in triangle rasterization. */ -- cgit v1.2.3 From 41899c70a72cd6206acec6c4c41953fea17d4ecf Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 13:02:11 -0700 Subject: Cell: emit state in cell_clear_surface() if dirty. Without this a program that does nothing but glClear() doesn't work. We need the framebuffer state. --- src/mesa/pipe/cell/ppu/cell_clear.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_clear.c b/src/mesa/pipe/cell/ppu/cell_clear.c index e61bfd9b0f..07b908eec5 100644 --- a/src/mesa/pipe/cell/ppu/cell_clear.c +++ b/src/mesa/pipe/cell/ppu/cell_clear.c @@ -50,6 +50,10 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, struct cell_context *cell = cell_context(pipe); uint surfIndex; + if (cell->dirty) + cell_update_derived(cell); + + if (!cell->cbuf_map[0]) cell->cbuf_map[0] = pipe_surface_map(ps); -- cgit v1.2.3 From c2372cc7481bf3985a6a3126952ab9d5dab4bf77 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 17:22:12 -0700 Subject: Cell: initial texture cache/sampling code --- src/mesa/pipe/cell/spu/spu_texture.c | 139 +++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_texture.h | 43 +++++++++++ 2 files changed, 182 insertions(+) create mode 100644 src/mesa/pipe/cell/spu/spu_texture.c create mode 100644 src/mesa/pipe/cell/spu/spu_texture.h (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c new file mode 100644 index 0000000000..6d566a5006 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -0,0 +1,139 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_compiler.h" +#include "spu_main.h" +#include "spu_texture.h" +#include "spu_tile.h" + + +/** + * Number of texture tiles to cache. + * Note that this will probably be the largest consumer of SPU local store/ + * memory for this driver! + */ +#define CACHE_SIZE 16 + +static tile_t tex_tiles[CACHE_SIZE] ALIGN16_ATTRIB; + +static int tex_tile_x[CACHE_SIZE], tex_tile_y[CACHE_SIZE]; + + + +/** + * Mark all tex cache entries as invalid. + */ +void +invalidate_tex_cache(void) +{ + /* XXX memset? */ + uint i; + for (i = 0; i < CACHE_SIZE; i++) + tex_tile_x[i] = tex_tile_y[i] = -1; +} + + +/** + * Return the cache pos/index which corresponds to texel (i,j) + */ +static INLINE uint +cache_pos(uint i, uint j) +{ + uint tx = i / TILE_SIZE; + uint ty = j / TILE_SIZE; + uint pos = (tx + ty * 4) % CACHE_SIZE; + return pos; +} + + +/** + * Make sure the tile for texel (i,j) is present, return its position/index + * in the cache. + */ +static uint +get_tex_tile(uint i, uint j) +{ + const int tx = i / TILE_SIZE; + const int ty = j / TILE_SIZE; + const uint pos = cache_pos(i, j); + + if (tex_tile_x[pos] != tx || tex_tile_y[pos] != ty) { + /* texture cache miss, fetch tile from main memory */ + const uint tiles_per_row = spu.texture.width / TILE_SIZE; + const uint bytes_per_tile = sizeof(tile_t); + const void *src = (const ubyte *) spu.texture.start + + (ty * tiles_per_row + tx) * bytes_per_tile; + + printf("SPU %u: tex cache miss at %d, %d pos=%u old=%d,%d\n", + spu.init.id, tx, ty, pos, + tex_tile_x[pos], tex_tile_y[pos]); +#if 0 + printf("SPU %u: get tex tile from %p to %p\n", + spu.init.id, src, tex_tiles[pos].t32); +#endif + + ASSERT_ALIGN16(tex_tiles[pos].t32); + ASSERT_ALIGN16(src); + + mfc_get(tex_tiles[pos].t32, /* dest */ + (unsigned int) src, + bytes_per_tile, /* size */ + TAG_TEXTURE_TILE, + 0, /* tid */ + 0 /* rid */); + + wait_on_mask(1 << TAG_TEXTURE_TILE); + + tex_tile_x[pos] = tx; + tex_tile_y[pos] = ty; + } + else { +#if 0 + printf("SPU %u: tex cache HIT at %d, %d\n", + spu.init.id, tx, ty); +#endif + } + + return pos; +} + + +/** + * Get texture sample at texcoord. + * XXX this is extremely primitive for now. + */ +uint +sample_texture(const float *texcoord) +{ + /* wrap/repeat */ + uint i = (uint) (texcoord[0] * spu.texture.width) % spu.texture.width; + uint j = (uint) (texcoord[1] * spu.texture.height) % spu.texture.height; + uint pos = get_tex_tile(i, j); + uint texel = tex_tiles[pos].t32[j % TILE_SIZE][i % TILE_SIZE]; + return texel; +} diff --git a/src/mesa/pipe/cell/spu/spu_texture.h b/src/mesa/pipe/cell/spu/spu_texture.h new file mode 100644 index 0000000000..b75b7ac44f --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_texture.h @@ -0,0 +1,43 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SPU_TEXTURE_H +#define SPU_TEXTURE_H + + +#include "pipe/p_compiler.h" + + +extern void +invalidate_tex_cache(void); + + +extern uint +sample_texture(const float *texcoord); + + +#endif /* SPU_TEXTURE_H */ -- cgit v1.2.3 From 425f270fcbfdbfce98adaf9da4b8eb7360f34447 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 17:23:44 -0700 Subject: Cell: basic texture mapping Texture images are tiled in PPU code. SPUs use a texture cache for getting texels from textures. This is very rough code, but demos/texcyl.c works. --- src/mesa/pipe/cell/common.h | 10 +++- src/mesa/pipe/cell/ppu/cell_context.h | 5 +- src/mesa/pipe/cell/ppu/cell_state_emit.c | 12 +++- src/mesa/pipe/cell/ppu/cell_state_sampler.c | 10 +++- src/mesa/pipe/cell/ppu/cell_texture.c | 87 +++++++++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_texture.h | 6 ++ src/mesa/pipe/cell/spu/Makefile | 1 + src/mesa/pipe/cell/spu/spu_main.c | 17 ++++++ src/mesa/pipe/cell/spu/spu_main.h | 3 + src/mesa/pipe/cell/spu/spu_tri.c | 60 ++++++++++++-------- 10 files changed, 183 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 5e32b209e6..f0d48ff403 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -79,7 +79,8 @@ #define CELL_CMD_STATE_FRAMEBUFFER 10 #define CELL_CMD_STATE_DEPTH_STENCIL 11 #define CELL_CMD_STATE_SAMPLER 12 -#define CELL_CMD_STATE_VERTEX_INFO 13 +#define CELL_CMD_STATE_TEXTURE 13 +#define CELL_CMD_STATE_VERTEX_INFO 14 #define CELL_NUM_BUFFERS 4 @@ -134,6 +135,13 @@ struct cell_command_release_verts }; +struct cell_command_texture +{ + void *start; /**< Address in main memory */ + uint width, height; +}; + + /** XXX unions don't seem to work */ struct cell_command { diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index de65fb5e9a..7d234f3e45 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -76,7 +76,7 @@ struct cell_context struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct cell_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; @@ -84,6 +84,9 @@ struct cell_context ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS]; ubyte *zsbuf_map; + struct pipe_surface *tex_surf; + uint *tex_map; + uint dirty; /** The primitive drawing context */ diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c index 6776ec88c7..391ff454ac 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.c +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.c @@ -30,7 +30,7 @@ #include "cell_state.h" #include "cell_state_emit.h" #include "cell_batch.h" - +#include "cell_texture.h" static void @@ -72,6 +72,16 @@ cell_emit_state(struct cell_context *cell) cell->sampler[0], sizeof(struct pipe_sampler_state)); } + if (cell->dirty & CELL_NEW_TEXTURE) { + struct cell_command_texture texture; + texture.start = cell->texture[0]->tiled_data; + texture.width = cell->texture[0]->base.width[0]; + texture.height = cell->texture[0]->base.height[0]; + + emit_state_cmd(cell, CELL_CMD_STATE_TEXTURE, + &texture, sizeof(struct cell_command_texture)); + } + if (cell->dirty & CELL_NEW_VERTEX_INFO) { emit_state_cmd(cell, CELL_CMD_STATE_VERTEX_INFO, &cell->vertex_info, sizeof(struct vertex_info)); diff --git a/src/mesa/pipe/cell/ppu/cell_state_sampler.c b/src/mesa/pipe/cell/ppu/cell_state_sampler.c index ae1eeb4620..317f7603bb 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_sampler.c +++ b/src/mesa/pipe/cell/ppu/cell_state_sampler.c @@ -30,12 +30,10 @@ */ #include "pipe/p_util.h" +#include "pipe/draw/draw_context.h" #include "cell_context.h" #include "cell_state.h" -#if 0 #include "cell_texture.h" -#include "cell_tile_cache.h" -#endif void * @@ -53,6 +51,8 @@ cell_bind_sampler_state(struct pipe_context *pipe, { struct cell_context *cell = cell_context(pipe); + draw_flush(cell->draw); + assert(unit < PIPE_MAX_SAMPLERS); cell->sampler[unit] = (struct pipe_sampler_state *)sampler; @@ -76,7 +76,11 @@ cell_set_sampler_texture(struct pipe_context *pipe, { struct cell_context *cell = cell_context(pipe); + draw_flush(cell->draw); + cell->texture[sampler] = texture; + cell_update_texture_mapping(cell); + cell->dirty |= CELL_NEW_TEXTURE; } diff --git a/src/mesa/pipe/cell/ppu/cell_texture.c b/src/mesa/pipe/cell/ppu/cell_texture.c index 0a8190d983..acbe4c79f0 100644 --- a/src/mesa/pipe/cell/ppu/cell_texture.c +++ b/src/mesa/pipe/cell/ppu/cell_texture.c @@ -163,3 +163,90 @@ cell_get_tex_surface(struct pipe_context *pipe, } return ps; } + + + +static void +tile_copy_data(uint w, uint h, uint tile_size, uint *dst, const uint *src) +{ + const uint tile_size2 = tile_size * tile_size; + const uint h_t = h / tile_size, w_t = w / tile_size; + + uint it, jt; /* tile counters */ + uint i, j; /* intra-tile counters */ + + for (it = 0; it < h_t; it++) { + for (jt = 0; jt < w_t; jt++) { + /* fill in tile (i, j) */ + uint *tdst = dst + (it * w_t + jt) * tile_size2; + for (i = 0; i < tile_size; i++) { + for (j = 0; j < tile_size; j++) { + const uint srci = it * tile_size + i; + const uint srcj = jt * tile_size + j; + *tdst++ = src[srci * h + srcj]; + } + } + } + } +} + + + +/** + * Convert linear texture image data to tiled format for SPU usage. + */ +static void +cell_tile_texture(struct cell_context *cell, + struct cell_texture *texture) +{ + uint face = 0, level = 0, zslice = 0; + struct pipe_surface *surf; + const uint w = texture->base.width[0], h = texture->base.height[0]; + const uint *src; + + /* temporary restrictions: */ + assert(w >= TILE_SIZE); + assert(h >= TILE_SIZE); + assert(w % TILE_SIZE == 0); + assert(h % TILE_SIZE == 0); + + surf = cell_get_tex_surface(&cell->pipe, &texture->base, face, level, zslice); + ASSERT(surf); + + src = (const uint *) pipe_surface_map(surf); + + if (texture->tiled_data) { + align_free(texture->tiled_data); + } + texture->tiled_data = align_malloc(w * h * 4, 16); + + tile_copy_data(w, h, TILE_SIZE, texture->tiled_data, src); + + pipe_surface_unmap(surf); + + pipe_surface_reference(&surf, NULL); +} + + + +void +cell_update_texture_mapping(struct cell_context *cell) +{ + uint face = 0, level = 0, zslice = 0; + + cell_tile_texture(cell, cell->texture[0]); +#if 0 + if (cell->tex_surf && cell->tex_map) { + pipe_surface_unmap(cell->tex_surf); + cell->tex_map = NULL; + } + + /* XXX free old surface */ + + cell->tex_surf = cell_get_tex_surface(&cell->pipe, + &cell->texture[0]->base, + face, level, zslice); + + cell->tex_map = pipe_surface_map(cell->tex_surf); +#endif +} diff --git a/src/mesa/pipe/cell/ppu/cell_texture.h b/src/mesa/pipe/cell/ppu/cell_texture.h index ef5808c086..bd434c8776 100644 --- a/src/mesa/pipe/cell/ppu/cell_texture.h +++ b/src/mesa/pipe/cell/ppu/cell_texture.h @@ -46,6 +46,8 @@ struct cell_texture */ struct pipe_buffer *buffer; unsigned long buffer_size; + + void *tiled_data; /* XXX this may be temporary */ /*ALIGN16*/ }; @@ -70,4 +72,8 @@ cell_get_tex_surface(struct pipe_context *pipe, unsigned face, unsigned level, unsigned zslice); +extern void +cell_update_texture_mapping(struct cell_context *cell); + + #endif /* CELL_TEXTURE */ diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index 417ae1b072..011fdcefe3 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -17,6 +17,7 @@ PROG_SPU_EMBED_O = $(PROG)_spu-embed.o SOURCES = \ spu_main.c \ + spu_texture.c \ spu_tile.c \ spu_tri.c diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index c2b05ed5a2..5a5b17dd89 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -34,6 +34,7 @@ #include #include "spu_main.h" +#include "spu_texture.h" #include "spu_tri.h" #include "spu_tile.h" #include "pipe/cell/common.h" @@ -446,6 +447,17 @@ cmd_state_sampler(const struct pipe_sampler_state *state) } +static void +cmd_state_texture(const struct cell_command_texture *texture) +{ + if (Debug) + printf("SPU %u: TEXTURE at %p size %u x %u\n", + spu.init.id, texture->start, texture->width, texture->height); + + memcpy(&spu.texture, texture, sizeof(*texture)); +} + + static void cmd_state_vertex_info(const struct vertex_info *vinfo) { @@ -561,6 +573,10 @@ cmd_batch(uint opcode) cmd_state_sampler((struct pipe_sampler_state *) &buffer[pos+1]); pos += (1 + sizeof(struct pipe_sampler_state) / 4); break; + case CELL_CMD_STATE_TEXTURE: + cmd_state_texture((struct cell_command_texture *) &buffer[pos+1]); + pos += (1 + sizeof(struct cell_command_texture) / 4); + break; case CELL_CMD_STATE_VERTEX_INFO: cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]); pos += (1 + sizeof(struct vertex_info) / 4); @@ -656,6 +672,7 @@ one_time_init(void) { memset(tile_status, TILE_STATUS_DEFINED, sizeof(tile_status)); memset(tile_status_z, TILE_STATUS_DEFINED, sizeof(tile_status_z)); + invalidate_tex_cache(); } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 5bc5d9fa99..480c54ebd0 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -60,6 +60,7 @@ struct spu_global struct pipe_depth_stencil_alpha_state depth_stencil; struct pipe_blend_state blend; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; + struct cell_command_texture texture; struct vertex_info vertex_info; @@ -84,6 +85,8 @@ extern struct spu_global spu; #define TAG_INDEX_BUFFER 16 #define TAG_BATCH_BUFFER 17 #define TAG_MISC 18 +#define TAG_TEXTURE_TILE 19 + extern void diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 3d0d106c10..aad28f1036 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -33,6 +33,7 @@ #include "pipe/p_format.h" #include "pipe/p_util.h" #include "spu_main.h" +#include "spu_texture.h" #include "spu_tile.h" #include "spu_tri.h" @@ -362,9 +363,24 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) /* Cell: "write" quad fragments to the tile by setting prim color */ const int ix = x - setup->cliprect_minx; const int iy = y - setup->cliprect_miny; - float colors[4][4]; - - eval_coeff(setup, 1, (float) x, (float) y, colors); + uint colors[4]; /* indexed by QUAD_x */ + + if (spu.texture.start) { + float texcoords[4][4]; + uint i; + eval_coeff(setup, 2, (float) x, (float) y, texcoords); + for (i = 0; i < 4; i++) { + colors[i] = sample_texture(texcoords[i]); + } + } + else { + float fcolors[4][4]; + eval_coeff(setup, 1, (float) x, (float) y, fcolors); + colors[QUAD_TOP_LEFT] = pack_color(fcolors[QUAD_TOP_LEFT]); + colors[QUAD_TOP_RIGHT] = pack_color(fcolors[QUAD_TOP_RIGHT]); + colors[QUAD_BOTTOM_LEFT] = pack_color(fcolors[QUAD_BOTTOM_LEFT]); + colors[QUAD_BOTTOM_RIGHT] = pack_color(fcolors[QUAD_BOTTOM_RIGHT]); + } if (spu.depth_stencil.depth.enabled) { mask &= do_depth_test(setup, x, y, mask); @@ -382,13 +398,13 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) tile_status[setup->ty][setup->tx] = TILE_STATUS_DIRTY; if (mask & MASK_TOP_LEFT) - ctile.t32[iy][ix] = pack_color(colors[QUAD_TOP_LEFT]); + ctile.t32[iy][ix] = colors[QUAD_TOP_LEFT]; if (mask & MASK_TOP_RIGHT) - ctile.t32[iy][ix+1] = pack_color(colors[QUAD_TOP_RIGHT]); + ctile.t32[iy][ix+1] = colors[QUAD_TOP_RIGHT]; if (mask & MASK_BOTTOM_LEFT) - ctile.t32[iy+1][ix] = pack_color(colors[QUAD_BOTTOM_LEFT]); + ctile.t32[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; if (mask & MASK_BOTTOM_RIGHT) - ctile.t32[iy+1][ix+1] = pack_color(colors[QUAD_BOTTOM_RIGHT]); + ctile.t32[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; } #endif } @@ -606,7 +622,6 @@ static boolean setup_sort_vertices( struct setup_stage *setup, } -#if 0 /** * Compute a0 for a constant-valued coefficient (GL_FLAT shading). * The value value comes from vertex->data[slot][i]. @@ -614,21 +629,20 @@ static boolean setup_sort_vertices( struct setup_stage *setup, * \param slot which attribute slot * \param i which component of the slot (0..3) */ -static void const_coeff( struct setup_stage *setup, - unsigned slot, - unsigned i ) +static void const_coeff(struct setup_stage *setup, uint slot) { - assert(slot < PIPE_MAX_SHADER_INPUTS); - assert(i <= 3); + uint i; + ASSERT(slot < PIPE_MAX_SHADER_INPUTS); - setup->coef[slot].dadx[i] = 0; - setup->coef[slot].dady[i] = 0; + for (i = 0; i < 4; i++) { + setup->coef[slot].dadx[i] = 0; + setup->coef[slot].dady[i] = 0; - /* need provoking vertex info! - */ - setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; + /* need provoking vertex info! + */ + setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; + } } -#endif /** @@ -735,15 +749,17 @@ static void setup_tri_coefficients( struct setup_stage *setup ) case INTERP_NONE: break; case INTERP_POS: - tri_linear_coeff(setup, i, 2, 3); /* slot 0, z */ + tri_linear_coeff(setup, i, 2, 3); /* XXX interp W if PERSPECTIVE... */ break; case INTERP_CONSTANT: - /* fall-through */ + const_coeff(setup, i); + break; case INTERP_LINEAR: - tri_linear_coeff(setup, i, 0, 4); /* slot 1, color */ + tri_linear_coeff(setup, i, 0, 4); break; case INTERP_PERSPECTIVE: + tri_linear_coeff(setup, i, 0, 4); /* XXX temporary */ break; default: ASSERT(0); -- cgit v1.2.3 From 25105276b38451439516928d188e07f2eb3e250e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 17:32:23 -0700 Subject: Cell: minor optimization for flat shading --- src/mesa/pipe/cell/spu/spu_tri.c | 37 ++++++++++++++++++++++++++++--------- 1 file changed, 28 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index aad28f1036..19a231d9c4 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -200,16 +200,35 @@ static INLINE void eval_coeff( struct setup_stage *setup, uint slot, float x, float y, float result[4][4]) { - uint i; - const float *dadx = setup->coef[slot].dadx; - const float *dady = setup->coef[slot].dady; + switch (spu.vertex_info.interp_mode[slot]) { + case INTERP_CONSTANT: + { + uint i; + for (i = 0; i < 4; i++) { + result[QUAD_TOP_LEFT][i] = + result[QUAD_TOP_RIGHT][i] = + result[QUAD_BOTTOM_LEFT][i] = + result[QUAD_BOTTOM_RIGHT][i] = setup->coef[slot].a0[i]; + } + } + break; - /* loop over XYZW comps */ - for (i = 0; i < 4; i++) { - result[QUAD_TOP_LEFT][i] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; - result[QUAD_TOP_RIGHT][i] = result[0][i] + dadx[i]; - result[QUAD_BOTTOM_LEFT][i] = result[0][i] + dady[i]; - result[QUAD_BOTTOM_RIGHT][i] = result[0][i] + dadx[i] + dady[i]; + case INTERP_LINEAR: + /* fall-through, for now */ + default: + { + uint i; + const float *dadx = setup->coef[slot].dadx; + const float *dady = setup->coef[slot].dady; + + /* loop over XYZW comps */ + for (i = 0; i < 4; i++) { + result[QUAD_TOP_LEFT][i] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; + result[QUAD_TOP_RIGHT][i] = result[0][i] + dadx[i]; + result[QUAD_BOTTOM_LEFT][i] = result[0][i] + dady[i]; + result[QUAD_BOTTOM_RIGHT][i] = result[0][i] + dadx[i] + dady[i]; + } + } } } -- cgit v1.2.3 From e2406b47883d74933e74507af65695c8c7d7861a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 18:03:45 -0700 Subject: Cell: compute min index referenced in draw command, use it to reduce size of vertex data payload --- src/mesa/pipe/cell/common.h | 2 ++ src/mesa/pipe/cell/ppu/cell_vbuf.c | 13 +++++++++++-- src/mesa/pipe/cell/spu/spu_main.c | 20 ++++++++++++++++---- 3 files changed, 29 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index f0d48ff403..90aa46a534 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -124,6 +124,8 @@ struct cell_command_render uint num_indexes; uint vertex_buf; /**< which cell->buffer[] contains the vertex data */ float xmin, dummy2, ymin, xmax, ymax; /* XXX another dummy field */ + uint dummy3; + uint min_index; boolean inline_verts; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 9f737287ad..e63b34cf52 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -138,16 +138,23 @@ cell_vbuf_draw(struct vbuf_render *vbr, struct cell_context *cell = cvbr->cell; float xmin, ymin, xmax, ymax; uint i; - uint nr_vertices = 0; + uint nr_vertices = 0, min_index = ~0; const void *vertices = cvbr->vertex_buffer; const uint vertex_size = cvbr->vertex_size; for (i = 0; i < nr_indices; i++) { if (indices[i] > nr_vertices) nr_vertices = indices[i]; + if (indices[i] < min_index) + min_index = indices[i]; } nr_vertices++; +#if 0 + /*if (min_index > 0)*/ + printf("%s min_index = %u\n", __FUNCTION__, min_index); +#endif + #if 0 printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u\n", nr_indices, nr_vertices); @@ -169,7 +176,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, /* compute x/y bounding box */ xmin = ymin = 1e50; xmax = ymax = -1e50; - for (i = 0; i < nr_vertices; i++) { + for (i = min_index; i < nr_vertices; i++) { const float *v = (float *) ((ubyte *) vertices + i * vertex_size); if (v[0] < xmin) xmin = v[0]; @@ -204,6 +211,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, render->prim_type = cvbr->prim; render->num_indexes = nr_indices; + render->min_index = min_index; /* append indices after render command */ memcpy(render + 1, indices, nr_indices * 2); @@ -214,6 +222,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, render->vertex_size = 4 * cell->vertex_info.size; render->num_verts = nr_vertices; if (ALLOW_INLINE_VERTS && + min_index == 0 && vertex_bytes <= cell_batch_free_space(cell)) { /* vertex data inlined, after indices */ void *dst = cell_batch_alloc(cell, vertex_bytes); diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 5a5b17dd89..3c9efb4741 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -253,7 +253,7 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) /* we'll DMA into these buffers */ ubyte vertex_data[CELL_BUFFER_SIZE] ALIGN16_ATTRIB; const uint vertex_size = render->vertex_size; /* in bytes */ - const uint total_vertex_bytes = render->num_verts * vertex_size; + /*const*/ uint total_vertex_bytes = render->num_verts * vertex_size; const ubyte *vertices; const ushort *indexes; uint i, j; @@ -289,9 +289,21 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) } else { /* Begin DMA fetch of vertex buffer */ - void *src = spu.init.buffers[render->vertex_buf]; - mfc_get(vertex_data, /* dest */ - (unsigned int) src, + ubyte *src = spu.init.buffers[render->vertex_buf]; + ubyte *dest = vertex_data; + + /* skip vertex data we won't use */ +#if 01 + src += render->min_index * vertex_size; + dest += render->min_index * vertex_size; + total_vertex_bytes -= render->min_index * vertex_size; +#endif + ASSERT(total_vertex_bytes % 16 == 0); + ASSERT_ALIGN16(dest); + ASSERT_ALIGN16(src); + + mfc_get(dest, /* in vertex_data[] array */ + (unsigned int) src, /* src in main memory */ total_vertex_bytes, /* size */ TAG_VERTEX_BUFFER, 0, /* tid */ -- cgit v1.2.3 From 4bede9219be1f93844c5897216c6674b46a23a88 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 18:09:16 -0700 Subject: Cell: add a few null texture tests --- src/mesa/pipe/cell/ppu/cell_state_emit.c | 13 ++++++++++--- src/mesa/pipe/cell/ppu/cell_texture.c | 3 ++- 2 files changed, 12 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c index 391ff454ac..702184416b 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.c +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.c @@ -74,9 +74,16 @@ cell_emit_state(struct cell_context *cell) if (cell->dirty & CELL_NEW_TEXTURE) { struct cell_command_texture texture; - texture.start = cell->texture[0]->tiled_data; - texture.width = cell->texture[0]->base.width[0]; - texture.height = cell->texture[0]->base.height[0]; + if (cell->texture[0]) { + texture.start = cell->texture[0]->tiled_data; + texture.width = cell->texture[0]->base.width[0]; + texture.height = cell->texture[0]->base.height[0]; + } + else { + texture.start = NULL; + texture.width = 0; + texture.height = 0; + } emit_state_cmd(cell, CELL_CMD_STATE_TEXTURE, &texture, sizeof(struct cell_command_texture)); diff --git a/src/mesa/pipe/cell/ppu/cell_texture.c b/src/mesa/pipe/cell/ppu/cell_texture.c index acbe4c79f0..2cf6022939 100644 --- a/src/mesa/pipe/cell/ppu/cell_texture.c +++ b/src/mesa/pipe/cell/ppu/cell_texture.c @@ -234,7 +234,8 @@ cell_update_texture_mapping(struct cell_context *cell) { uint face = 0, level = 0, zslice = 0; - cell_tile_texture(cell, cell->texture[0]); + if (cell->texture[0]) + cell_tile_texture(cell, cell->texture[0]); #if 0 if (cell->tex_surf && cell->tex_map) { pipe_surface_unmap(cell->tex_surf); -- cgit v1.2.3 From 64935c875128d2d1254b6b39ced72b9848d477fe Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 28 Jan 2008 18:17:30 -0700 Subject: Cell: move cmd_render() into new spu_render.c file --- src/mesa/pipe/cell/spu/Makefile | 1 + src/mesa/pipe/cell/spu/spu_main.c | 206 +------------------------------ src/mesa/pipe/cell/spu/spu_main.h | 1 + src/mesa/pipe/cell/spu/spu_render.c | 240 ++++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_render.h | 38 ++++++ 5 files changed, 283 insertions(+), 203 deletions(-) create mode 100644 src/mesa/pipe/cell/spu/spu_render.c create mode 100644 src/mesa/pipe/cell/spu/spu_render.h (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index 011fdcefe3..d5b30e1f27 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -17,6 +17,7 @@ PROG_SPU_EMBED_O = $(PROG)_spu-embed.o SOURCES = \ spu_main.c \ + spu_render.c \ spu_texture.c \ spu_tile.c \ spu_tri.c diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 3c9efb4741..6e02f2c964 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -34,8 +34,8 @@ #include #include "spu_main.h" +#include "spu_render.h" #include "spu_texture.h" -#include "spu_tri.h" #include "spu_tile.h" #include "pipe/cell/common.h" #include "pipe/p_defines.h" @@ -47,7 +47,7 @@ helpful headers: /opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h */ -static boolean Debug = FALSE; +boolean Debug = FALSE; struct spu_global spu; @@ -61,7 +61,7 @@ wait_on_mask(unsigned tagMask) } -static void +static INLINE void wait_on_mask_all(unsigned tagMask) { mfc_write_tag_mask( tagMask ); @@ -192,206 +192,6 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) } -/** - * Given a rendering command's bounding box (in pixels) compute the - * location of the corresponding screen tile bounding box. - */ -static INLINE void -tile_bounding_box(const struct cell_command_render *render, - uint *txmin, uint *tymin, - uint *box_num_tiles, uint *box_width_tiles) -{ -#if 0 - /* Debug: full-window bounding box */ - uint txmax = spu.fb.width_tiles - 1; - uint tymax = spu.fb.height_tiles - 1; - *txmin = 0; - *tymin = 0; - *box_num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; - *box_width_tiles = spu.fb.width_tiles; - (void) render; - (void) txmax; - (void) tymax; -#else - uint txmax, tymax, box_height_tiles; - - *txmin = (uint) render->xmin / TILE_SIZE; - *tymin = (uint) render->ymin / TILE_SIZE; - txmax = (uint) render->xmax / TILE_SIZE; - tymax = (uint) render->ymax / TILE_SIZE; - *box_width_tiles = txmax - *txmin + 1; - box_height_tiles = tymax - *tymin + 1; - *box_num_tiles = *box_width_tiles * box_height_tiles; -#endif -#if 0 - printf("SPU %u: bounds: %g, %g ... %g, %g\n", spu.init.id, - render->xmin, render->ymin, render->xmax, render->ymax); - printf("SPU %u: tiles: %u, %u .. %u, %u\n", - spu.init.id, *txmin, *tymin, txmax, tymax); - ASSERT(render->xmin <= render->xmax); - ASSERT(render->ymin <= render->ymax); -#endif -} - - -/** Check if the tile at (tx,ty) belongs to this SPU */ -static INLINE boolean -my_tile(uint tx, uint ty) -{ - return (spu.fb.width_tiles * ty + tx) % spu.init.num_spus == spu.init.id; -} - - -/** - * Render primitives - * \param pos_incr returns value indicating how may words to skip after - * this command in the batch buffer - */ -static void -cmd_render(const struct cell_command_render *render, uint *pos_incr) -{ - /* we'll DMA into these buffers */ - ubyte vertex_data[CELL_BUFFER_SIZE] ALIGN16_ATTRIB; - const uint vertex_size = render->vertex_size; /* in bytes */ - /*const*/ uint total_vertex_bytes = render->num_verts * vertex_size; - const ubyte *vertices; - const ushort *indexes; - uint i, j; - - - if (Debug) { - printf("SPU %u: RENDER prim %u, num_vert=%u num_ind=%u " - "inline_vert=%u\n", - spu.init.id, - render->prim_type, - render->num_verts, - render->num_indexes, - render->inline_verts); - - /* - printf(" bound: %g, %g .. %g, %g\n", - render->xmin, render->ymin, render->xmax, render->ymax); - */ - } - - ASSERT(sizeof(*render) % 4 == 0); - ASSERT(total_vertex_bytes % 16 == 0); - - /* indexes are right after the render command in the batch buffer */ - indexes = (const ushort *) (render + 1); - *pos_incr = (render->num_indexes * 2 + 3) / 4; - - - if (render->inline_verts) { - /* Vertices are right after indexes in batch buffer */ - vertices = (const ubyte *) (render + 1) + *pos_incr * 4; - *pos_incr = *pos_incr + total_vertex_bytes / 4; - } - else { - /* Begin DMA fetch of vertex buffer */ - ubyte *src = spu.init.buffers[render->vertex_buf]; - ubyte *dest = vertex_data; - - /* skip vertex data we won't use */ -#if 01 - src += render->min_index * vertex_size; - dest += render->min_index * vertex_size; - total_vertex_bytes -= render->min_index * vertex_size; -#endif - ASSERT(total_vertex_bytes % 16 == 0); - ASSERT_ALIGN16(dest); - ASSERT_ALIGN16(src); - - mfc_get(dest, /* in vertex_data[] array */ - (unsigned int) src, /* src in main memory */ - total_vertex_bytes, /* size */ - TAG_VERTEX_BUFFER, - 0, /* tid */ - 0 /* rid */); - - vertices = vertex_data; - - wait_on_mask(1 << TAG_VERTEX_BUFFER); - } - - - /** - ** find tiles which intersect the prim bounding box - **/ - uint txmin, tymin, box_width_tiles, box_num_tiles; - tile_bounding_box(render, &txmin, &tymin, - &box_num_tiles, &box_width_tiles); - - - /* make sure any pending clears have completed */ - wait_on_mask(1 << TAG_SURFACE_CLEAR); /* XXX temporary */ - - - /** - ** loop over tiles, rendering tris - **/ - for (i = 0; i < box_num_tiles; i++) { - const uint tx = txmin + i % box_width_tiles; - const uint ty = tymin + i / box_width_tiles; - - ASSERT(tx < spu.fb.width_tiles); - ASSERT(ty < spu.fb.height_tiles); - - if (!my_tile(tx, ty)) - continue; - - /* Start fetching color/z tiles. We'll wait for completion when - * we need read/write to them later in triangle rasterization. - */ - if (spu.depth_stencil.depth.enabled) { - if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(tx, ty, &ztile, TAG_READ_TILE_Z, 1); - } - } - - if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(tx, ty, &ctile, TAG_READ_TILE_COLOR, 0); - } - - ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); - ASSERT(render->num_indexes % 3 == 0); - - /* loop over tris */ - for (j = 0; j < render->num_indexes; j += 3) { - const float *v0, *v1, *v2; - - v0 = (const float *) (vertices + indexes[j+0] * vertex_size); - v1 = (const float *) (vertices + indexes[j+1] * vertex_size); - v2 = (const float *) (vertices + indexes[j+2] * vertex_size); - - tri_draw(v0, v1, v2, tx, ty); - } - - /* write color/z tiles back to main framebuffer, if dirtied */ - if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(tx, ty, &ctile, TAG_WRITE_TILE_COLOR, 0); - tile_status[ty][tx] = TILE_STATUS_DEFINED; - } - if (spu.depth_stencil.depth.enabled) { - if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(tx, ty, &ztile, TAG_WRITE_TILE_Z, 1); - tile_status_z[ty][tx] = TILE_STATUS_DEFINED; - } - } - - /* XXX move these... */ - wait_on_mask(1 << TAG_WRITE_TILE_COLOR); - if (spu.depth_stencil.depth.enabled) { - wait_on_mask(1 << TAG_WRITE_TILE_Z); - } - } - - if (Debug) - printf("SPU %u: RENDER done\n", - spu.init.id); -} - - static void cmd_release_verts(const struct cell_command_release_verts *release) { diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 480c54ebd0..009e046ba5 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -70,6 +70,7 @@ struct spu_global extern struct spu_global spu; +extern boolean Debug; diff --git a/src/mesa/pipe/cell/spu/spu_render.c b/src/mesa/pipe/cell/spu/spu_render.c new file mode 100644 index 0000000000..21a286a23d --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_render.c @@ -0,0 +1,240 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include +#include +#include + +#include "spu_main.h" +#include "spu_render.h" +#include "spu_tri.h" +#include "spu_tile.h" +#include "pipe/cell/common.h" + + + +/** + * Given a rendering command's bounding box (in pixels) compute the + * location of the corresponding screen tile bounding box. + */ +static INLINE void +tile_bounding_box(const struct cell_command_render *render, + uint *txmin, uint *tymin, + uint *box_num_tiles, uint *box_width_tiles) +{ +#if 0 + /* Debug: full-window bounding box */ + uint txmax = spu.fb.width_tiles - 1; + uint tymax = spu.fb.height_tiles - 1; + *txmin = 0; + *tymin = 0; + *box_num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; + *box_width_tiles = spu.fb.width_tiles; + (void) render; + (void) txmax; + (void) tymax; +#else + uint txmax, tymax, box_height_tiles; + + *txmin = (uint) render->xmin / TILE_SIZE; + *tymin = (uint) render->ymin / TILE_SIZE; + txmax = (uint) render->xmax / TILE_SIZE; + tymax = (uint) render->ymax / TILE_SIZE; + *box_width_tiles = txmax - *txmin + 1; + box_height_tiles = tymax - *tymin + 1; + *box_num_tiles = *box_width_tiles * box_height_tiles; +#endif +#if 0 + printf("SPU %u: bounds: %g, %g ... %g, %g\n", spu.init.id, + render->xmin, render->ymin, render->xmax, render->ymax); + printf("SPU %u: tiles: %u, %u .. %u, %u\n", + spu.init.id, *txmin, *tymin, txmax, tymax); + ASSERT(render->xmin <= render->xmax); + ASSERT(render->ymin <= render->ymax); +#endif +} + + +/** Check if the tile at (tx,ty) belongs to this SPU */ +static INLINE boolean +my_tile(uint tx, uint ty) +{ + return (spu.fb.width_tiles * ty + tx) % spu.init.num_spus == spu.init.id; +} + + +/** + * Render primitives + * \param pos_incr returns value indicating how may words to skip after + * this command in the batch buffer + */ +void +cmd_render(const struct cell_command_render *render, uint *pos_incr) +{ + /* we'll DMA into these buffers */ + ubyte vertex_data[CELL_BUFFER_SIZE] ALIGN16_ATTRIB; + const uint vertex_size = render->vertex_size; /* in bytes */ + /*const*/ uint total_vertex_bytes = render->num_verts * vertex_size; + const ubyte *vertices; + const ushort *indexes; + uint i, j; + + + if (Debug) { + printf("SPU %u: RENDER prim %u, num_vert=%u num_ind=%u " + "inline_vert=%u\n", + spu.init.id, + render->prim_type, + render->num_verts, + render->num_indexes, + render->inline_verts); + + /* + printf(" bound: %g, %g .. %g, %g\n", + render->xmin, render->ymin, render->xmax, render->ymax); + */ + } + + ASSERT(sizeof(*render) % 4 == 0); + ASSERT(total_vertex_bytes % 16 == 0); + + /* indexes are right after the render command in the batch buffer */ + indexes = (const ushort *) (render + 1); + *pos_incr = (render->num_indexes * 2 + 3) / 4; + + + if (render->inline_verts) { + /* Vertices are right after indexes in batch buffer */ + vertices = (const ubyte *) (render + 1) + *pos_incr * 4; + *pos_incr = *pos_incr + total_vertex_bytes / 4; + } + else { + /* Begin DMA fetch of vertex buffer */ + ubyte *src = spu.init.buffers[render->vertex_buf]; + ubyte *dest = vertex_data; + + /* skip vertex data we won't use */ +#if 01 + src += render->min_index * vertex_size; + dest += render->min_index * vertex_size; + total_vertex_bytes -= render->min_index * vertex_size; +#endif + ASSERT(total_vertex_bytes % 16 == 0); + ASSERT_ALIGN16(dest); + ASSERT_ALIGN16(src); + + mfc_get(dest, /* in vertex_data[] array */ + (unsigned int) src, /* src in main memory */ + total_vertex_bytes, /* size */ + TAG_VERTEX_BUFFER, + 0, /* tid */ + 0 /* rid */); + + vertices = vertex_data; + + wait_on_mask(1 << TAG_VERTEX_BUFFER); + } + + + /** + ** find tiles which intersect the prim bounding box + **/ + uint txmin, tymin, box_width_tiles, box_num_tiles; + tile_bounding_box(render, &txmin, &tymin, + &box_num_tiles, &box_width_tiles); + + + /* make sure any pending clears have completed */ + wait_on_mask(1 << TAG_SURFACE_CLEAR); /* XXX temporary */ + + + /** + ** loop over tiles, rendering tris + **/ + for (i = 0; i < box_num_tiles; i++) { + const uint tx = txmin + i % box_width_tiles; + const uint ty = tymin + i / box_width_tiles; + + ASSERT(tx < spu.fb.width_tiles); + ASSERT(ty < spu.fb.height_tiles); + + if (!my_tile(tx, ty)) + continue; + + /* Start fetching color/z tiles. We'll wait for completion when + * we need read/write to them later in triangle rasterization. + */ + if (spu.depth_stencil.depth.enabled) { + if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(tx, ty, &ztile, TAG_READ_TILE_Z, 1); + } + } + + if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(tx, ty, &ctile, TAG_READ_TILE_COLOR, 0); + } + + ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); + ASSERT(render->num_indexes % 3 == 0); + + /* loop over tris */ + for (j = 0; j < render->num_indexes; j += 3) { + const float *v0, *v1, *v2; + + v0 = (const float *) (vertices + indexes[j+0] * vertex_size); + v1 = (const float *) (vertices + indexes[j+1] * vertex_size); + v2 = (const float *) (vertices + indexes[j+2] * vertex_size); + + tri_draw(v0, v1, v2, tx, ty); + } + + /* write color/z tiles back to main framebuffer, if dirtied */ + if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(tx, ty, &ctile, TAG_WRITE_TILE_COLOR, 0); + tile_status[ty][tx] = TILE_STATUS_DEFINED; + } + if (spu.depth_stencil.depth.enabled) { + if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(tx, ty, &ztile, TAG_WRITE_TILE_Z, 1); + tile_status_z[ty][tx] = TILE_STATUS_DEFINED; + } + } + + /* XXX move these... */ + wait_on_mask(1 << TAG_WRITE_TILE_COLOR); + if (spu.depth_stencil.depth.enabled) { + wait_on_mask(1 << TAG_WRITE_TILE_Z); + } + } + + if (Debug) + printf("SPU %u: RENDER done\n", + spu.init.id); +} + + diff --git a/src/mesa/pipe/cell/spu/spu_render.h b/src/mesa/pipe/cell/spu/spu_render.h new file mode 100644 index 0000000000..fbcdc5ec31 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_render.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SPU_RENDER_H +#define SPU_RENDER_H + +#include "pipe/cell/common.h" + +extern void +cmd_render(const struct cell_command_render *render, uint *pos_incr); + +#endif /* SPU_RENDER_H */ + -- cgit v1.2.3 From 3d3f7cf06e4ac25aeb604703cb8113db9fd2f8eb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 29 Jan 2008 11:21:29 +0900 Subject: gallium: Add extern keyword to global. --- src/mesa/pipe/draw/draw_vf.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h index 7d90f35b0f..c0fa063c52 100644 --- a/src/mesa/pipe/draw/draw_vf.h +++ b/src/mesa/pipe/draw/draw_vf.h @@ -195,7 +195,8 @@ struct draw_vf_format_info { const unsigned attrsize; }; -const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX]; +extern const struct draw_vf_format_info +draw_vf_format_info[DRAW_EMIT_MAX]; #endif -- cgit v1.2.3 From 6f75de06ff2ea899b43b94236dbfbfaee12ba88c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 29 Jan 2008 18:01:17 +0900 Subject: gallium: Allow draw_vf usage to be controlled at runtime. --- src/mesa/pipe/draw/draw_vbuf.c | 149 ++++++++++++++++++++--------------------- 1 file changed, 71 insertions(+), 78 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 8ca225c65a..2309ed9f12 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -127,15 +127,9 @@ emit_vertex( struct vbuf_stage *vbuf, struct vertex_header *vertex ) { #if 0 - const struct vertex_info *vinfo = vbuf->vinfo; - - uint i; - uint count = 0; /* for debug/sanity */ - - assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); - -// fprintf(stderr, "emit vertex %d to %p\n", -// vbuf->nr_vertices, vbuf->vertex_ptr); + fprintf(stderr, "emit vertex %d to %p\n", + vbuf->nr_vertices, vbuf->vertex_ptr); +#endif if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { if(vertex->vertex_id < vbuf->nr_vertices) @@ -148,75 +142,72 @@ emit_vertex( struct vbuf_stage *vbuf, vertex->vertex_id = vbuf->nr_vertices++; - for (i = 0; i < vinfo->num_attribs; i++) { - uint j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; - case EMIT_ALL: - /* just copy the whole vertex as-is to the vbuf */ - assert(i == 0); - assert(j == 0); - memcpy(vbuf->vertex_ptr, vertex, vinfo->size * 4); - vbuf->vertex_ptr += vinfo->size; - count += vinfo->size; - break; - case EMIT_1F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - count++; - break; - case EMIT_1F_PSIZE: - *vbuf->vertex_ptr++ = fui(vbuf->stage.draw->rasterizer->point_size); - count++; - break; - case EMIT_2F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - count += 2; - break; - case EMIT_3F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); - count += 3; - break; - case EMIT_4F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][3]); - count += 4; - break; - case EMIT_4UB: - *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), - float_to_ubyte( vertex->data[j][1] ), - float_to_ubyte( vertex->data[j][0] ), - float_to_ubyte( vertex->data[j][3] )); - count += 1; - break; - default: - assert(0); + if(!vbuf->vf) { + const struct vertex_info *vinfo = vbuf->vinfo; + uint i; + uint count = 0; /* for debug/sanity */ + + assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); + + for (i = 0; i < vinfo->num_attribs; i++) { + uint j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + /* no-op */ + break; + case EMIT_ALL: + /* just copy the whole vertex as-is to the vbuf */ + assert(i == 0); + assert(j == 0); + memcpy(vbuf->vertex_ptr, vertex, vinfo->size * 4); + vbuf->vertex_ptr += vinfo->size; + count += vinfo->size; + break; + case EMIT_1F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + count++; + break; + case EMIT_1F_PSIZE: + *vbuf->vertex_ptr++ = fui(vbuf->stage.draw->rasterizer->point_size); + count++; + break; + case EMIT_2F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + count += 2; + break; + case EMIT_3F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); + count += 3; + break; + case EMIT_4F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][3]); + count += 4; + break; + case EMIT_4UB: + *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), + float_to_ubyte( vertex->data[j][1] ), + float_to_ubyte( vertex->data[j][0] ), + float_to_ubyte( vertex->data[j][3] )); + count += 1; + break; + default: + assert(0); + } } + assert(count == vinfo->size); } - assert(count == vinfo->size); -#else - if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { - if(vertex->vertex_id < vbuf->nr_vertices) - return; - else - fprintf(stderr, "Bad vertex id 0x%04x (>= 0x%04x)\n", - vertex->vertex_id, vbuf->nr_vertices); - return; + else { + draw_vf_set_data(vbuf->vf, vertex->data); + draw_vf_emit_vertices(vbuf->vf, 1, vbuf->vertex_ptr); + + vbuf->vertex_ptr += vbuf->vertex_size/4; } - - vertex->vertex_id = vbuf->nr_vertices++; - - draw_vf_set_data(vbuf->vf, vertex->data); - draw_vf_emit_vertices(vbuf->vf, 1, vbuf->vertex_ptr); - - vbuf->vertex_ptr += vbuf->vertex_size/4; -#endif } @@ -229,6 +220,9 @@ vbuf_set_vf_attributes(struct vbuf_stage *vbuf ) uint count = 0; /* for debug/sanity */ unsigned nr_attrs = 0; + if(!vbuf->vf) + return; + // fprintf(stderr, "emit vertex %d to %p\n", // vbuf->nr_vertices, vbuf->vertex_ptr); @@ -625,9 +619,8 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, vbuf->prim = ~0; - vbuf->vf = draw_vf_create(); - if(!vbuf->vf) - vbuf_destroy(&vbuf->stage); + if(!GETENV("GALLIUM_NOVF")) + vbuf->vf = draw_vf_create(); return &vbuf->stage; } -- cgit v1.2.3 From 5022344c656c0e004222a0a77c98838e8ae0a1ac Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 29 Jan 2008 20:46:48 +0900 Subject: gallium: Emit constants. --- src/mesa/pipe/draw/draw_vf.c | 10 +++++++-- src/mesa/pipe/draw/draw_vf.h | 22 ++++++++++++++++++- src/mesa/pipe/draw/draw_vf_generic.c | 42 +++++++++++++++++++++++++----------- src/mesa/pipe/draw/draw_vf_sse.c | 4 ++++ 4 files changed, 62 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index 4fc2312ad1..958d31933b 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -182,6 +182,9 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, vf->attr[j].insert = draw_vf_format_info[format].insert; vf->attr[j].vertattrsize = draw_vf_format_info[format].attrsize; vf->attr[j].vertoffset = offset; + vf->attr[j].isconst = draw_vf_format_info[format].isconst; + if(vf->attr[j].isconst) + memcpy(vf->attr[j].data, &map[i].data, vf->attr[j].vertattrsize); if (DBG) _mesa_printf("%d: %s, offset %d\n", i, @@ -240,8 +243,11 @@ void draw_vf_set_data( struct draw_vertex_fetch *vf, for (j = 0; j < vf->attr_count; j++) { a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ a[j].inputsize = 4; - a[j].do_insert = a[j].insert[4 - 1]; - a[j].inputptr = (uint8_t *)&data[a[j].attrib][0]; + a[j].do_insert = a[j].insert[4 - 1]; + if(a[j].isconst) + a[j].inputptr = a[j].data; + else + a[j].inputptr = (uint8_t *)&data[a[j].attrib][0]; } } diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h index c0fa063c52..911ea07bdf 100644 --- a/src/mesa/pipe/draw/draw_vf.h +++ b/src/mesa/pipe/draw/draw_vf.h @@ -48,14 +48,30 @@ enum draw_vf_attr_format { DRAW_EMIT_4UB_4F_BGRA, /**< for color */ DRAW_EMIT_4UB_4F_ARGB, /**< for color */ DRAW_EMIT_4UB_4F_ABGR, /**< for color */ + DRAW_EMIT_1F_CONST, + DRAW_EMIT_2F_CONST, + DRAW_EMIT_3F_CONST, + DRAW_EMIT_4F_CONST, DRAW_EMIT_PAD, /**< leave a hole of 'offset' bytes */ DRAW_EMIT_MAX }; -struct draw_vf_attr_map { +struct draw_vf_attr_map +{ + /** Input attribute number */ unsigned attrib; + enum draw_vf_attr_format format; + unsigned offset; + + /** + * Constant data for DRAW_EMIT_*_CONST + */ + union { + uint8_t ub[4]; + float f[4]; + } data; }; struct draw_vertex_fetch; @@ -124,6 +140,9 @@ struct draw_vf_attr unsigned inputsize; unsigned inputstride; unsigned vertoffset; /**< position of the attrib in the vertex struct */ + + boolean isconst; /**< read from const data below */ + uint8_t data[16]; unsigned attrib; /**< which vertex attrib (0=position, etc) */ unsigned vertattrsize; /**< size of the attribute in bytes */ @@ -193,6 +212,7 @@ struct draw_vf_format_info { const char *name; draw_vf_insert_func insert[4]; const unsigned attrsize; + const boolean isconst; }; extern const struct draw_vf_format_info diff --git a/src/mesa/pipe/draw/draw_vf_generic.c b/src/mesa/pipe/draw/draw_vf_generic.c index a16eb456b7..0caa798396 100644 --- a/src/mesa/pipe/draw/draw_vf_generic.c +++ b/src/mesa/pipe/draw/draw_vf_generic.c @@ -387,62 +387,78 @@ const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX] = { { "1f", { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, - sizeof(float) }, + sizeof(float), FALSE }, { "2f", { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, - 2 * sizeof(float) }, + 2 * sizeof(float), FALSE }, { "3f", { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, - 3 * sizeof(float) }, + 3 * sizeof(float), FALSE }, { "4f", { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, - 4 * sizeof(float) }, + 4 * sizeof(float), FALSE }, { "3f_xyw", { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_4 }, - 3 * sizeof(float) }, + 3 * sizeof(float), FALSE }, { "1ub_1f", { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 }, - sizeof(uint8_t) }, + sizeof(uint8_t), FALSE }, { "3ub_3f_rgb", { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3, insert_3ub_3f_rgb_3 }, - 3 * sizeof(uint8_t) }, + 3 * sizeof(uint8_t), FALSE }, { "3ub_3f_bgr", { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3, insert_3ub_3f_bgr_3 }, - 3 * sizeof(uint8_t) }, + 3 * sizeof(uint8_t), FALSE }, { "4ub_4f_rgba", { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, insert_4ub_4f_rgba_4 }, - 4 * sizeof(uint8_t) }, + 4 * sizeof(uint8_t), FALSE }, { "4ub_4f_bgra", { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3, insert_4ub_4f_bgra_4 }, - 4 * sizeof(uint8_t) }, + 4 * sizeof(uint8_t), FALSE }, { "4ub_4f_argb", { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3, insert_4ub_4f_argb_4 }, - 4 * sizeof(uint8_t) }, + 4 * sizeof(uint8_t), FALSE }, { "4ub_4f_abgr", { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3, insert_4ub_4f_abgr_4 }, - 4 * sizeof(uint8_t) }, + 4 * sizeof(uint8_t), FALSE }, + + { "1f_const", + { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, + sizeof(float), TRUE }, + + { "2f_const", + { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, + 2 * sizeof(float), TRUE }, + + { "3f_const", + { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, + 3 * sizeof(float), TRUE }, + + { "4f_const", + { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, + 4 * sizeof(float), TRUE }, { "pad", { NULL, NULL, NULL, NULL }, - 0 } + 0, FALSE }, }; diff --git a/src/mesa/pipe/draw/draw_vf_sse.c b/src/mesa/pipe/draw/draw_vf_sse.c index 4036ded1d8..1389e6cfb9 100644 --- a/src/mesa/pipe/draw/draw_vf_sse.c +++ b/src/mesa/pipe/draw/draw_vf_sse.c @@ -388,18 +388,21 @@ static boolean build_vertex_emit( struct x86_program *p ) */ switch (a->format) { case DRAW_EMIT_1F: + case DRAW_EMIT_1F_CONST: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); emit_store(p, dest, 1, temp); update_src_ptr(p, srcECX, vfESI, a); break; case DRAW_EMIT_2F: + case DRAW_EMIT_2F_CONST: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); emit_store(p, dest, 2, temp); update_src_ptr(p, srcECX, vfESI, a); break; case DRAW_EMIT_3F: + case DRAW_EMIT_3F_CONST: /* Potentially the worst case - hardcode 2+1 copying: */ if (0) { @@ -423,6 +426,7 @@ static boolean build_vertex_emit( struct x86_program *p ) } break; case DRAW_EMIT_4F: + case DRAW_EMIT_4F_CONST: get_src_ptr(p, srcECX, vfESI, a); emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); emit_store(p, dest, 4, temp); -- cgit v1.2.3 From f94425b316b57ad19ce067a449b20ebee50064f9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 29 Jan 2008 20:47:30 +0900 Subject: gallium: Emit point size as a constant. --- src/mesa/pipe/draw/draw_vbuf.c | 85 ++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 81 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 2309ed9f12..92a8b9fbcf 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -115,6 +115,70 @@ check_space( struct vbuf_stage *vbuf, unsigned nr ) } +#if 0 +static INLINE void +dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) +{ + assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); + unsigned i, j, k; + + for (i = 0; i < vinfo->num_attribs; i++) { + j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + fprintf(stderr, "EMIT_OMIT:"); + break; + case EMIT_ALL: + assert(i == 0); + assert(j == 0); + fprintf(stderr, "EMIT_ALL:\t"); + for(k = 0; k < vinfo->size*4; ++k) + fprintf(stderr, "%02x ", *data++); + break; + case EMIT_1F: + fprintf(stderr, "EMIT_1F:\t"); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_1F_PSIZE: + fprintf(stderr, "EMIT_1F_PSIZE:\t"); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_2F: + fprintf(stderr, "EMIT_2F:\t"); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_3F: + fprintf(stderr, "EMIT_3F:\t"); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + data += sizeof(float); + break; + case EMIT_4F: + fprintf(stderr, "EMIT_4F:\t"); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_4UB: + fprintf(stderr, "EMIT_4UB:\t"); + fprintf(stderr, "%u ", *data++); + fprintf(stderr, "%u ", *data++); + fprintf(stderr, "%u ", *data++); + fprintf(stderr, "%u ", *data++); + break; + default: + assert(0); + } + fprintf(stderr, "\n"); + } + fprintf(stderr, "\n"); +} +#endif + + /** * Extract the needed fields from post-transformed vertex and emit * a hardware(driver) vertex. @@ -190,7 +254,7 @@ emit_vertex( struct vbuf_stage *vbuf, count += 4; break; case EMIT_4UB: - *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), + *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), float_to_ubyte( vertex->data[j][1] ), float_to_ubyte( vertex->data[j][0] ), float_to_ubyte( vertex->data[j][3] )); @@ -201,6 +265,20 @@ emit_vertex( struct vbuf_stage *vbuf, } } assert(count == vinfo->size); +#if 0 + { + static float data[256]; + draw_vf_set_data(vbuf->vf, vertex->data); + draw_vf_emit_vertices(vbuf->vf, 1, data); + if(memcmp((uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size, data, vbuf->vertex_size)) { + fprintf(stderr, "With VF:\n"); + dump_emitted_vertex(vbuf->vinfo, (uint8_t *)data); + fprintf(stderr, "Without VF:\n"); + dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size); + assert(0); + } + } +#endif } else { draw_vf_set_data(vbuf->vf, vertex->data); @@ -297,11 +375,10 @@ vbuf_set_vf_attributes(struct vbuf_stage *vbuf ) count++; break; case EMIT_1F_PSIZE: - /* FIXME */ - assert(0); attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_PAD; + attrs[nr_attrs].format = DRAW_EMIT_1F_CONST; attrs[nr_attrs].offset = 0; + attrs[nr_attrs].data.f[0] = vbuf->stage.draw->rasterizer->point_size; nr_attrs++; count++; break; -- cgit v1.2.3 From c185c55aec8c6d0e47a2d7b84acf7d063acfce61 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 29 Jan 2008 12:37:07 +0000 Subject: gallium: don't rely on assert(0) for error handling - may be disabled --- src/mesa/state_tracker/st_draw.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 8ef50ee768..c9b8e78485 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -298,6 +298,7 @@ st_draw_vbo(GLcontext *ctx, break; default: assert(0); + return; } /* get/create the index buffer object */ @@ -570,6 +571,7 @@ st_feedback_draw_vbo(GLcontext *ctx, break; default: assert(0); + return; } map = pipe->winsys->buffer_map(pipe->winsys, -- cgit v1.2.3 From c81848210e744650724a63fbf5a5795fa4e019c5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 29 Jan 2008 12:37:47 +0000 Subject: gallium: streamline various unfilled & stippled paths --- src/mesa/pipe/draw/draw_prim.c | 158 +++++++++++++++++++++++++++++------------ 1 file changed, 113 insertions(+), 45 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 2a612a1673..41b3fddcc1 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -69,28 +69,46 @@ static void draw_prim_queue_flush( struct draw_context *draw ) * draw->pipeline->first is often changed by the first call to tri(), * line(), etc. */ - switch (draw->reduced_prim) { - case RP_TRI: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - - draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); + if (draw->rasterizer->line_stipple_enable) { + switch (draw->reduced_prim) { + case RP_TRI: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); + + draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); + } + break; + case RP_LINE: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); + + draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); + } + break; + case RP_POINT: + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); + break; } - break; - case RP_LINE: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - - draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); + } + else { + switch (draw->reduced_prim) { + case RP_TRI: + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); + break; + case RP_LINE: + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); + break; + case RP_POINT: + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); + break; } - break; - case RP_POINT: - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); - break; } draw->pq.queue_nr = 0; @@ -231,7 +249,7 @@ static void do_ef_triangle( struct draw_context *draw, } -static void do_quad( struct draw_context *draw, +static void do_ef_quad( struct draw_context *draw, unsigned v0, unsigned v1, unsigned v2, @@ -243,6 +261,16 @@ static void do_quad( struct draw_context *draw, do_ef_triangle( draw, 0, omitEdge3, v1, v2, v3 ); } +static void do_quad( struct draw_context *draw, + unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3 ) +{ + do_triangle( draw, v0, v1, v3 ); + do_triangle( draw, v1, v2, v3 ); +} + /** * Main entrypoint to draw some number of points/lines/triangles @@ -252,6 +280,8 @@ draw_prim( struct draw_context *draw, unsigned prim, unsigned start, unsigned count ) { unsigned i; + boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); // _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); @@ -289,24 +319,32 @@ draw_prim( struct draw_context *draw, break; case PIPE_PRIM_LINE_STRIP: - if (count >= 2) { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, - start + i - 1, - start + i ); - } + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, + start + i - 1, + start + i ); } break; case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - do_ef_triangle( draw, - 1, - ~0, + if (unfilled) { + for (i = 0; i+2 < count; i += 3) { + do_ef_triangle( draw, + 1, + ~0, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + else { + for (i = 0; i+2 < count; i += 3) { + do_triangle( draw, start + i + 0, start + i + 1, start + i + 2 ); + } } break; @@ -340,27 +378,49 @@ draw_prim( struct draw_context *draw, case PIPE_PRIM_QUADS: - for (i = 0; i+3 < count; i += 4) { - do_quad( draw, - start + i + 0, - start + i + 1, - start + i + 2, - start + i + 3); + if (unfilled) { + for (i = 0; i+3 < count; i += 4) { + do_ef_quad( draw, + start + i + 0, + start + i + 1, + start + i + 2, + start + i + 3); + } + } + else { + for (i = 0; i+3 < count; i += 4) { + do_quad( draw, + start + i + 0, + start + i + 1, + start + i + 2, + start + i + 3); + } } break; case PIPE_PRIM_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - do_quad( draw, - start + i + 2, - start + i + 0, - start + i + 1, - start + i + 3); + if (unfilled) { + for (i = 0; i+3 < count; i += 2) { + do_ef_quad( draw, + start + i + 2, + start + i + 0, + start + i + 1, + start + i + 3); + } + } + else { + for (i = 0; i+3 < count; i += 2) { + do_quad( draw, + start + i + 2, + start + i + 0, + start + i + 1, + start + i + 3); + } } break; case PIPE_PRIM_POLYGON: - if (count >= 3) { + if (unfilled) { unsigned ef_mask = (1<<2) | (1<<0); for (i = 0; i+2 < count; i++) { @@ -378,6 +438,14 @@ draw_prim( struct draw_context *draw, ef_mask &= ~(1<<2); } } + else { + for (i = 0; i+2 < count; i++) { + do_triangle( draw, + start + i + 1, + start + i + 2, + start + 0); + } + } break; default: -- cgit v1.2.3 From 1aaed3856878a39beb5aab0402d2adb8b277e812 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 29 Jan 2008 15:17:56 +0000 Subject: gallium: weaken assert slightly --- src/mesa/pipe/draw/draw_vf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index 958d31933b..06b84b93cc 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -174,7 +174,7 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, } else { - assert(vf->lookup[map[i].attrib] == 0); + assert(vf->lookup[map[i].attrib] == 0 || format == DRAW_EMIT_1F_CONST); vf->lookup[map[i].attrib] = &vf->attr[j]; vf->attr[j].attrib = map[i].attrib; -- cgit v1.2.3 From 7f2713a29ff46a608de0feac2f56f034dbc738cb Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 29 Jan 2008 11:22:57 -0700 Subject: Cell: use _pack_rgba8() from pack_rgba8.h to do float[4]->uint color conversion texcyl.c is twice as fast now in non-texture mode --- src/mesa/pipe/cell/spu/spu_tri.c | 15 ++++----------- 1 file changed, 4 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 19a231d9c4..7c6a54134f 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -29,6 +29,8 @@ * Triangle rendering within a tile. */ +#include + #include "pipe/p_compiler.h" #include "pipe/p_format.h" #include "pipe/p_util.h" @@ -38,7 +40,6 @@ #include "spu_tri.h" - /** * Simplified types taken from other parts of Gallium */ @@ -252,19 +253,11 @@ eval_z( struct setup_stage *setup, static INLINE uint pack_color(const float color[4]) { - uint r = (uint) (color[0] * 255.0); - uint g = (uint) (color[1] * 255.0); - uint b = (uint) (color[2] * 255.0); - uint a = (uint) (color[3] * 255.0); - r = MIN2(r, 255); - g = MIN2(g, 255); - b = MIN2(b, 255); - a = MIN2(a, 255); switch (spu.fb.color_format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - return (a << 24) | (r << 16) | (g << 8) | b; + return _pack_rgba8(color[3], color[0], color[1], color[2]); case PIPE_FORMAT_B8G8R8A8_UNORM: - return (b << 24) | (g << 16) | (r << 8) | a; + return _pack_rgba8(color[2], color[1], color[0], color[3]); default: ASSERT(0); return 0; -- cgit v1.2.3 From 17ef840af40c9228ee0f4f7453bc00e318d9e6c4 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 29 Jan 2008 16:41:10 +0100 Subject: gallium: Fix build on WinXP. --- src/mesa/pipe/draw/draw_clip.c | 2 +- src/mesa/pipe/draw/draw_vertex_fetch.c | 16 ++++++++-------- src/mesa/pipe/draw/draw_vertex_shader.c | 4 ++-- src/mesa/pipe/draw/draw_vf.c | 10 ++++++---- src/mesa/pipe/draw/draw_vf_generic.c | 2 -- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 4 ++-- 6 files changed, 19 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index da20028904..61130c5600 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -406,7 +406,7 @@ clip_init_state( struct draw_stage *stage ) { struct clipper *clipper = clipper_stage( stage ); - clipper->flat = stage->draw->rasterizer->flatshade; + clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; if (clipper->flat) { const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index 89e4c256a7..b23f487e74 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -326,6 +326,10 @@ static void fetch_xyz_rgb( struct draw_context *draw, const unsigned *elts, unsigned count ) { + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + assert(count <= 4); // _mesa_printf("%s\n", __FUNCTION__); @@ -333,10 +337,6 @@ static void fetch_xyz_rgb( struct draw_context *draw, /* loop over vertex attributes (vertex shader inputs) */ - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - for (i = 0; i < 4; i++) { { const float *in = (const float *)(src[0] + elts[i] * pitch[0]); @@ -366,15 +366,15 @@ static void fetch_xyz_rgb_st( struct draw_context *draw, const unsigned *elts, unsigned count ) { + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + assert(count <= 4); /* loop over vertex attributes (vertex shader inputs) */ - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - for (i = 0; i < 4; i++) { { const float *in = (const float *)(src[0] + elts[i] * pitch[0]); diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 0806e23d6c..b851da845f 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -199,7 +199,7 @@ run_vertex_program(struct draw_context *draw, void draw_vertex_shader_queue_flush(struct draw_context *draw) { - unsigned i, j; + unsigned i; assert(draw->vs.queue_nr != 0); @@ -219,7 +219,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) for (i = 0; i < draw->vs.queue_nr; i += 4) { struct vertex_header *dests[4]; unsigned elts[4]; - int n = MIN2(4, draw->vs.queue_nr - i); + int j, n = MIN2(4, draw->vs.queue_nr - i); for (j = 0; j < n; j++) { elts[j] = draw->vs.queue[i + j].elt; diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index 06b84b93cc..0debea1f12 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -32,7 +32,7 @@ #include "draw_vf.h" -#define DBG 0 +#define DRAW_VF_DBG 0 /* TODO: remove this */ @@ -166,9 +166,10 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, for (j = 0, i = 0; i < nr; i++) { const unsigned format = map[i].format; if (format == DRAW_EMIT_PAD) { - if (DBG) +#if (DRAW_VF_DBG) _mesa_printf("%d: pad %d, offset %d\n", i, map[i].offset, offset); +#endif offset += map[i].offset; @@ -186,10 +187,11 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, if(vf->attr[j].isconst) memcpy(vf->attr[j].data, &map[i].data, vf->attr[j].vertattrsize); - if (DBG) +#if (DRAW_VF_DBG) _mesa_printf("%d: %s, offset %d\n", i, draw_vf_format_info[format].name, vf->attr[j].vertoffset); +#endif offset += draw_vf_format_info[format].attrsize; j++; @@ -303,7 +305,7 @@ void draw_vf_destroy( struct draw_vertex_fetch *vf ) * to unify them, but this probably won't change until this * module gets another overhaul. */ - _mesa_exec_free((void *) fp->func); + //_mesa_exec_free((void *) fp->func); FREE(fp); } diff --git a/src/mesa/pipe/draw/draw_vf_generic.c b/src/mesa/pipe/draw/draw_vf_generic.c index 0caa798396..7f5f56ef9c 100644 --- a/src/mesa/pipe/draw/draw_vf_generic.c +++ b/src/mesa/pipe/draw/draw_vf_generic.c @@ -29,8 +29,6 @@ #include -#include "simple_list.h" - #include "pipe/p_compiler.h" #include "pipe/p_util.h" diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c index 349647fe6e..4cf4222db9 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -145,7 +145,7 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, /* Do the delayed destroy: */ pb_reference(&fenced_buf->buffer, NULL); - free(fenced_buf); + FREE(fenced_buf); } } @@ -162,7 +162,7 @@ fenced_buffer_destroy(struct pb_buffer *buf) } else { pb_reference(&fenced_buf->buffer, NULL); - free(fenced_buf); + FREE(fenced_buf); } if ((fenced_list->numDelayed % fenced_list->checkDelayed) == 0) -- cgit v1.2.3 From ed0327980a73947cab0ae619cdcfa7455259bff2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 30 Jan 2008 15:24:56 +0900 Subject: gallium: Remove draw_vertex_fetch::lookup. It is not being used, and would be dangerous to use given the possibility of constants. --- src/mesa/pipe/draw/draw_vf.c | 5 ----- src/mesa/pipe/draw/draw_vf.h | 2 -- 2 files changed, 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index 0debea1f12..64d9ed02a9 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -161,8 +161,6 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, assert(nr < PIPE_ATTRIB_MAX); - memset(vf->lookup, 0, sizeof(vf->lookup)); - for (j = 0, i = 0; i < nr; i++) { const unsigned format = map[i].format; if (format == DRAW_EMIT_PAD) { @@ -175,9 +173,6 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, } else { - assert(vf->lookup[map[i].attrib] == 0 || format == DRAW_EMIT_1F_CONST); - vf->lookup[map[i].attrib] = &vf->attr[j]; - vf->attr[j].attrib = map[i].attrib; vf->attr[j].format = format; vf->attr[j].insert = draw_vf_format_info[format].insert; diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h index 911ea07bdf..09cf4d3a6a 100644 --- a/src/mesa/pipe/draw/draw_vf.h +++ b/src/mesa/pipe/draw/draw_vf.h @@ -159,8 +159,6 @@ struct draw_vertex_fetch unsigned attr_count; unsigned vertex_stride; - struct draw_vf_attr *lookup[PIPE_ATTRIB_MAX]; - draw_vf_emit_func emit; /* Parameters and constants for codegen: -- cgit v1.2.3 From eb0e0d38eae02df17e2c11503dc047718c1244ad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 30 Jan 2008 16:46:41 +0900 Subject: gallium: Teach draw_vf about draw vertices. This reduces the emit overhead, which is significant since we're emiting one vertex at a time. --- src/mesa/pipe/draw/draw_vbuf.c | 147 ++--------------------------------- src/mesa/pipe/draw/draw_vf.c | 171 ++++++++++++++++++++++++++++++++++------- src/mesa/pipe/draw/draw_vf.h | 17 ++-- 3 files changed, 161 insertions(+), 174 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index 92a8b9fbcf..ac03001d8f 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -35,7 +35,6 @@ #include -#include #include "pipe/p_util.h" @@ -268,8 +267,7 @@ emit_vertex( struct vbuf_stage *vbuf, #if 0 { static float data[256]; - draw_vf_set_data(vbuf->vf, vertex->data); - draw_vf_emit_vertices(vbuf->vf, 1, data); + draw_vf_emit_vertex(vbuf->vf, vertex, data); if(memcmp((uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size, data, vbuf->vertex_size)) { fprintf(stderr, "With VF:\n"); dump_emitted_vertex(vbuf->vinfo, (uint8_t *)data); @@ -281,149 +279,13 @@ emit_vertex( struct vbuf_stage *vbuf, #endif } else { - draw_vf_set_data(vbuf->vf, vertex->data); - draw_vf_emit_vertices(vbuf->vf, 1, vbuf->vertex_ptr); + draw_vf_emit_vertex(vbuf->vf, vertex, vbuf->vertex_ptr); vbuf->vertex_ptr += vbuf->vertex_size/4; } } -static void -vbuf_set_vf_attributes(struct vbuf_stage *vbuf ) -{ - const struct vertex_info *vinfo = vbuf->vinfo; - struct draw_vf_attr_map attrs[PIPE_MAX_SHADER_INPUTS]; - uint i; - uint count = 0; /* for debug/sanity */ - unsigned nr_attrs = 0; - - if(!vbuf->vf) - return; - -// fprintf(stderr, "emit vertex %d to %p\n", -// vbuf->nr_vertices, vbuf->vertex_ptr); - -#if 0 - if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { - if(vertex->vertex_id < vbuf->nr_vertices) - return; - else - fprintf(stderr, "Bad vertex id 0x%04x (>= 0x%04x)\n", - vertex->vertex_id, vbuf->nr_vertices); - return; - } -#endif - - for (i = 0; i < vinfo->num_attribs; i++) { - uint j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; - case EMIT_ALL: { - /* just copy the whole vertex as-is to the vbuf */ - unsigned k, s = vinfo->size; - assert(i == 0); - assert(j == 0); - /* copy the vertex header */ - /* XXX: we actually don't copy the header, just pad it */ - attrs[nr_attrs].attrib = 0; - attrs[nr_attrs].format = DRAW_EMIT_PAD; - attrs[nr_attrs].offset = offsetof(struct vertex_header, data); - s -= offsetof(struct vertex_header, data)/4; - count += offsetof(struct vertex_header, data)/4; - nr_attrs++; - /* copy the vertex data */ - for(k = 0; k < (s & ~0x3); k += 4) { - attrs[nr_attrs].attrib = k/4; - attrs[nr_attrs].format = DRAW_EMIT_4F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 4; - } - /* tail */ - /* XXX: actually, this shouldn't be needed */ - attrs[nr_attrs].attrib = k/4; - attrs[nr_attrs].offset = 0; - switch(s & 0x3) { - case 0: - break; - case 1: - attrs[nr_attrs].format = DRAW_EMIT_1F; - nr_attrs++; - count += 1; - break; - case 2: - attrs[nr_attrs].format = DRAW_EMIT_2F; - nr_attrs++; - count += 2; - break; - case 3: - attrs[nr_attrs].format = DRAW_EMIT_3F; - nr_attrs++; - count += 3; - break; - } - break; - } - case EMIT_1F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_1F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count++; - break; - case EMIT_1F_PSIZE: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_1F_CONST; - attrs[nr_attrs].offset = 0; - attrs[nr_attrs].data.f[0] = vbuf->stage.draw->rasterizer->point_size; - nr_attrs++; - count++; - break; - case EMIT_2F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_2F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 2; - break; - case EMIT_3F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_3F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 3; - break; - case EMIT_4F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_4F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 4; - break; - case EMIT_4UB: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_4UB_4F_BGRA; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 1; - break; - default: - assert(0); - } - } - - assert(count == vinfo->size); - - draw_vf_set_vertex_attributes(vbuf->vf, - attrs, - nr_attrs, - vbuf->vertex_size); -} - - static void vbuf_tri( struct draw_stage *stage, struct prim_header *prim ) @@ -498,7 +360,10 @@ vbuf_set_prim( struct vbuf_stage *vbuf, uint newprim ) vbuf->vinfo = vinfo; vbuf->vertex_size = vertex_size; - vbuf_set_vf_attributes(vbuf); + if(vbuf->vf) + draw_vf_set_vertex_info(vbuf->vf, + vbuf->vinfo, + vbuf->stage.draw->rasterizer->point_size); if (!vbuf->vertices) vbuf_alloc_vertices(vbuf); diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index 64d9ed02a9..0da8e59ad6 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -26,6 +26,8 @@ */ +#include + #include "pipe/p_compiler.h" #include "pipe/p_util.h" @@ -151,10 +153,11 @@ static void choose_emit_func( struct draw_vertex_fetch *vf, -unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, - const struct draw_vf_attr_map *map, - unsigned nr, - unsigned vertex_stride ) +static unsigned +draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, + const struct draw_vf_attr_map *map, + unsigned nr, + unsigned vertex_stride ) { unsigned offset = 0; unsigned i, j; @@ -202,6 +205,133 @@ unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, } +void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, + const struct vertex_info *vinfo, + float point_size ) +{ + unsigned i, j, k; + struct draw_vf_attr *a = vf->attr; + struct draw_vf_attr_map attrs[PIPE_MAX_SHADER_INPUTS]; + unsigned count = 0; /* for debug/sanity */ + unsigned nr_attrs = 0; + + for (i = 0; i < vinfo->num_attribs; i++) { + j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + /* no-op */ + break; + case EMIT_ALL: { + /* just copy the whole vertex as-is to the vbuf */ + unsigned s = vinfo->size; + assert(i == 0); + assert(j == 0); + /* copy the vertex header */ + /* XXX: we actually don't copy the header, just pad it */ + attrs[nr_attrs].attrib = 0; + attrs[nr_attrs].format = DRAW_EMIT_PAD; + attrs[nr_attrs].offset = offsetof(struct vertex_header, data); + s -= offsetof(struct vertex_header, data)/4; + count += offsetof(struct vertex_header, data)/4; + nr_attrs++; + /* copy the vertex data */ + for(k = 0; k < (s & ~0x3); k += 4) { + attrs[nr_attrs].attrib = k/4; + attrs[nr_attrs].format = DRAW_EMIT_4F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 4; + } + /* tail */ + /* XXX: actually, this shouldn't be needed */ + attrs[nr_attrs].attrib = k/4; + attrs[nr_attrs].offset = 0; + switch(s & 0x3) { + case 0: + break; + case 1: + attrs[nr_attrs].format = DRAW_EMIT_1F; + nr_attrs++; + count += 1; + break; + case 2: + attrs[nr_attrs].format = DRAW_EMIT_2F; + nr_attrs++; + count += 2; + break; + case 3: + attrs[nr_attrs].format = DRAW_EMIT_3F; + nr_attrs++; + count += 3; + break; + } + break; + } + case EMIT_1F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_1F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count++; + break; + case EMIT_1F_PSIZE: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_1F_CONST; + attrs[nr_attrs].offset = 0; + attrs[nr_attrs].data.f[0] = point_size; + nr_attrs++; + count++; + break; + case EMIT_2F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_2F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 2; + break; + case EMIT_3F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_3F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 3; + break; + case EMIT_4F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_4F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 4; + break; + case EMIT_4UB: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_4UB_4F_BGRA; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 1; + break; + default: + assert(0); + } + } + + assert(count == vinfo->size); + + draw_vf_set_vertex_attributes(vf, + attrs, + nr_attrs, + vinfo->size * sizeof(float) ); + + for (j = 0; j < vf->attr_count; j++) { + a[j].inputsize = 4; + a[j].do_insert = a[j].insert[4 - 1]; + if(a[j].isconst) { + a[j].inputptr = a[j].data; + a[j].inputstride = 0; + } + } +} + #if 0 /* Set attribute pointers, adjusted for start position: @@ -229,38 +359,27 @@ void draw_vf_set_sources( struct draw_vertex_fetch *vf, #endif -/* Set attribute pointers, adjusted for start position: +/** + * Emit a vertex to dest. */ -void draw_vf_set_data( struct draw_vertex_fetch *vf, - float data[][4]) +void draw_vf_emit_vertex( struct draw_vertex_fetch *vf, + struct vertex_header *vertex, + void *dest ) { struct draw_vf_attr *a = vf->attr; unsigned j; for (j = 0; j < vf->attr_count; j++) { - a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ - a[j].inputsize = 4; - a[j].do_insert = a[j].insert[4 - 1]; - if(a[j].isconst) - a[j].inputptr = a[j].data; - else - a[j].inputptr = (uint8_t *)&data[a[j].attrib][0]; + if(!a[j].isconst) { + a[j].inputptr = (uint8_t *)&vertex->data[a[j].attrib][0]; + a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ + } } + + vf->emit( vf, 1, (uint8_t*) dest ); } -/* Emit count VB vertices to dest. - */ -void draw_vf_emit_vertices( struct draw_vertex_fetch *vf, - unsigned count, - void *dest ) -{ - vf->emit( vf, count, (uint8_t*) dest ); -} - - - - struct draw_vertex_fetch *draw_vf_create( void ) { diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h index 09cf4d3a6a..e694b98675 100644 --- a/src/mesa/pipe/draw/draw_vf.h +++ b/src/mesa/pipe/draw/draw_vf.h @@ -33,6 +33,7 @@ #include "pipe/p_state.h" #include "draw_vertex.h" +#include "draw_private.h" // for vertex_header enum draw_vf_attr_format { @@ -78,11 +79,17 @@ struct draw_vertex_fetch; +#if 0 unsigned draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, const struct draw_vf_attr_map *map, unsigned nr, unsigned vertex_stride ); +#endif + +void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, + const struct vertex_info *vinfo, + float point_size ); #if 0 void @@ -92,13 +99,9 @@ draw_vf_set_sources( struct draw_vertex_fetch *vf, #endif void -draw_vf_set_data( struct draw_vertex_fetch *vf, - float data[][4]); - -void -draw_vf_emit_vertices( struct draw_vertex_fetch *vf, - unsigned count, - void *dest ); +draw_vf_emit_vertex( struct draw_vertex_fetch *vf, + struct vertex_header *vertex, + void *dest ); struct draw_vertex_fetch * draw_vf_create( void ); -- cgit v1.2.3 From da6eac242d9b79ad77389b6ab579804bc0261005 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 Jan 2008 11:49:26 -0700 Subject: Cell: move CELL_MAX_SPUS --- src/mesa/pipe/cell/common.h | 2 ++ src/mesa/pipe/cell/ppu/cell_context.h | 3 --- 2 files changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 90aa46a534..d5e86863d4 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -61,6 +61,8 @@ #define ROUNDUP16(k) (((k) + 0xf) & ~0xf) +#define CELL_MAX_SPUS 6 + #define TILE_SIZE 32 diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 7d234f3e45..65b89518ad 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -38,9 +38,6 @@ #include "pipe/cell/common.h" -#define CELL_MAX_SPUS 6 - - struct cell_vbuf_render; struct cell_vertex_shader_state -- cgit v1.2.3 From 41bdf4cf4c924e4c04c62dc144584cf7ead3cf44 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 Jan 2008 11:49:51 -0700 Subject: Cell: make wait_on_mask() static/inlined --- src/mesa/pipe/cell/spu/spu_main.c | 19 ------------------- src/mesa/pipe/cell/spu/spu_main.h | 23 +++++++++++++++++++++-- 2 files changed, 21 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 6e02f2c964..6886f283be 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -31,7 +31,6 @@ #include #include -#include #include "spu_main.h" #include "spu_render.h" @@ -52,24 +51,6 @@ boolean Debug = FALSE; struct spu_global spu; -void -wait_on_mask(unsigned tagMask) -{ - mfc_write_tag_mask( tagMask ); - /* wait for completion of _any_ DMAs specified by tagMask */ - mfc_read_tag_status_any(); -} - - -static INLINE void -wait_on_mask_all(unsigned tagMask) -{ - mfc_write_tag_mask( tagMask ); - /* wait for completion of _any_ DMAs specified by tagMask */ - mfc_read_tag_status_all(); -} - - /** * Tell the PPU that this SPU has finished copying a buffer to * local store and that it may be reused by the PPU. diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 009e046ba5..8908bf8bc0 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -29,6 +29,8 @@ #define SPU_MAIN_H +#include + #include "pipe/cell/common.h" #include "pipe/draw/draw_vertex.h" #include "pipe/p_state.h" @@ -90,8 +92,25 @@ extern boolean Debug; -extern void -wait_on_mask(unsigned tag); +static INLINE void +wait_on_mask(unsigned tagMask) +{ + mfc_write_tag_mask( tagMask ); + /* wait for completion of _any_ DMAs specified by tagMask */ + mfc_read_tag_status_any(); +} + + +static INLINE void +wait_on_mask_all(unsigned tagMask) +{ + mfc_write_tag_mask( tagMask ); + /* wait for completion of _any_ DMAs specified by tagMask */ + mfc_read_tag_status_all(); +} + + + static INLINE void -- cgit v1.2.3 From 0d3f60ec64965a07ef26b551436f0d768154e4d3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 Jan 2008 11:56:14 -0700 Subject: Cell: check tile status before wait_on_mask() --- src/mesa/pipe/cell/spu/spu_tri.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 7c6a54134f..01a47a4851 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -278,7 +278,7 @@ do_depth_test(struct setup_stage *setup, int x, int y, unsigned mask) /* now, _really_ clear the tile */ clear_z_tile(&ztile); } - else { + else if (tile_status_z[setup->ty][setup->tx] != TILE_STATUS_DIRTY) { /* make sure we've got the tile from main mem */ wait_on_mask(1 << TAG_READ_TILE_Z); } @@ -403,7 +403,7 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) /* now, _really_ clear the tile */ clear_c_tile(&ctile); } - else { + else if (tile_status[setup->ty][setup->tx] != TILE_STATUS_DIRTY) { /* make sure we've got the tile from main mem */ wait_on_mask(1 << TAG_READ_TILE_COLOR); } -- cgit v1.2.3 From dcf41a0eed71a67060b4efa9ab4befc86eafc177 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 Jan 2008 11:56:41 -0700 Subject: Cell: minor code refactoring, movement --- src/mesa/pipe/cell/spu/spu_render.c | 85 ++++++++++++++++++++++++------------- 1 file changed, 55 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_render.c b/src/mesa/pipe/cell/spu/spu_render.c index 21a286a23d..f506095116 100644 --- a/src/mesa/pipe/cell/spu/spu_render.c +++ b/src/mesa/pipe/cell/spu/spu_render.c @@ -88,6 +88,55 @@ my_tile(uint tx, uint ty) } +/** + * Start fetching non-clear color/Z tiles from main memory + */ +static INLINE void +get_cz_tiles(uint tx, uint ty) +{ + if (spu.depth_stencil.depth.enabled) { + if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(tx, ty, &ztile, TAG_READ_TILE_Z, 1); + } + } + + if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { + get_tile(tx, ty, &ctile, TAG_READ_TILE_COLOR, 0); + } +} + + +/** + * Start putting dirty color/Z tiles back to main memory + */ +static INLINE void +put_cz_tiles(uint tx, uint ty) +{ + if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(tx, ty, &ztile, TAG_WRITE_TILE_Z, 1); + tile_status_z[ty][tx] = TILE_STATUS_DEFINED; + } + + if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { + put_tile(tx, ty, &ctile, TAG_WRITE_TILE_COLOR, 0); + tile_status[ty][tx] = TILE_STATUS_DEFINED; + } +} + + +/** + * Wait for 'put' of color/z tiles to complete. + */ +static INLINE void +wait_put_cz_tiles(void) +{ + wait_on_mask(1 << TAG_WRITE_TILE_COLOR); + if (spu.depth_stencil.depth.enabled) { + wait_on_mask(1 << TAG_WRITE_TILE_Z); + } +} + + /** * Render primitives * \param pos_incr returns value indicating how may words to skip after @@ -122,6 +171,9 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) ASSERT(sizeof(*render) % 4 == 0); ASSERT(total_vertex_bytes % 16 == 0); + ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); + ASSERT(render->num_indexes % 3 == 0); + /* indexes are right after the render command in the batch buffer */ indexes = (const ushort *) (render + 1); @@ -186,21 +238,7 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) if (!my_tile(tx, ty)) continue; - /* Start fetching color/z tiles. We'll wait for completion when - * we need read/write to them later in triangle rasterization. - */ - if (spu.depth_stencil.depth.enabled) { - if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(tx, ty, &ztile, TAG_READ_TILE_Z, 1); - } - } - - if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { - get_tile(tx, ty, &ctile, TAG_READ_TILE_COLOR, 0); - } - - ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); - ASSERT(render->num_indexes % 3 == 0); + get_cz_tiles(tx, ty); /* loop over tris */ for (j = 0; j < render->num_indexes; j += 3) { @@ -214,22 +252,9 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) } /* write color/z tiles back to main framebuffer, if dirtied */ - if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(tx, ty, &ctile, TAG_WRITE_TILE_COLOR, 0); - tile_status[ty][tx] = TILE_STATUS_DEFINED; - } - if (spu.depth_stencil.depth.enabled) { - if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { - put_tile(tx, ty, &ztile, TAG_WRITE_TILE_Z, 1); - tile_status_z[ty][tx] = TILE_STATUS_DEFINED; - } - } + put_cz_tiles(tx, ty); - /* XXX move these... */ - wait_on_mask(1 << TAG_WRITE_TILE_COLOR); - if (spu.depth_stencil.depth.enabled) { - wait_on_mask(1 << TAG_WRITE_TILE_Z); - } + wait_put_cz_tiles(); /* XXX seems unnecessary... */ } if (Debug) -- cgit v1.2.3 From 022bf6dfa1ef1c18f0439553e39e473b678848e2 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 Jan 2008 12:08:23 -0700 Subject: Cell: make 'setup' a regular var instead of passing around a pointer everywhere We'll never have more than one of these objects. Avoiding pointer deref improves performance a bit. --- src/mesa/pipe/cell/spu/spu_tri.c | 419 +++++++++++++++++++-------------------- 1 file changed, 209 insertions(+), 210 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 01a47a4851..5bb2cb12e3 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -135,6 +135,12 @@ struct setup_stage { }; + +static struct setup_stage setup; + + + + #if 0 /** * Basically a cast wrapper. @@ -147,33 +153,33 @@ static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) #if 0 /** - * Clip setup->quad against the scissor/surface bounds. + * Clip setup.quad against the scissor/surface bounds. */ static INLINE void quad_clip(struct setup_stage *setup) { - const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const struct pipe_scissor_state *cliprect = &setup.softpipe->cliprect; const int minx = (int) cliprect->minx; const int maxx = (int) cliprect->maxx; const int miny = (int) cliprect->miny; const int maxy = (int) cliprect->maxy; - if (setup->quad.x0 >= maxx || - setup->quad.y0 >= maxy || - setup->quad.x0 + 1 < minx || - setup->quad.y0 + 1 < miny) { + if (setup.quad.x0 >= maxx || + setup.quad.y0 >= maxy || + setup.quad.x0 + 1 < minx || + setup.quad.y0 + 1 < miny) { /* totally clipped */ - setup->quad.mask = 0x0; + setup.quad.mask = 0x0; return; } - if (setup->quad.x0 < minx) - setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup->quad.y0 < miny) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - if (setup->quad.x0 == maxx - 1) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup->quad.y0 == maxy - 1) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + if (setup.quad.x0 < minx) + setup.quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + if (setup.quad.y0 < miny) + setup.quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + if (setup.quad.x0 == maxx - 1) + setup.quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + if (setup.quad.y0 == maxy - 1) + setup.quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); } #endif @@ -185,9 +191,9 @@ static INLINE void clip_emit_quad(struct setup_stage *setup) { quad_clip(setup); - if (setup->quad.mask) { - struct softpipe_context *sp = setup->softpipe; - sp->quad.first->run(sp->quad.first, &setup->quad); + if (setup.quad.mask) { + struct softpipe_context *sp = setup.softpipe; + sp->quad.first->run(sp->quad.first, &setup.quad); } } #endif @@ -198,8 +204,7 @@ clip_emit_quad(struct setup_stage *setup) * Eg: four colors will be compute. */ static INLINE void -eval_coeff( struct setup_stage *setup, uint slot, - float x, float y, float result[4][4]) +eval_coeff(uint slot, float x, float y, float result[4][4]) { switch (spu.vertex_info.interp_mode[slot]) { case INTERP_CONSTANT: @@ -209,7 +214,7 @@ eval_coeff( struct setup_stage *setup, uint slot, result[QUAD_TOP_LEFT][i] = result[QUAD_TOP_RIGHT][i] = result[QUAD_BOTTOM_LEFT][i] = - result[QUAD_BOTTOM_RIGHT][i] = setup->coef[slot].a0[i]; + result[QUAD_BOTTOM_RIGHT][i] = setup.coef[slot].a0[i]; } } break; @@ -219,12 +224,12 @@ eval_coeff( struct setup_stage *setup, uint slot, default: { uint i; - const float *dadx = setup->coef[slot].dadx; - const float *dady = setup->coef[slot].dady; + const float *dadx = setup.coef[slot].dadx; + const float *dady = setup.coef[slot].dady; /* loop over XYZW comps */ for (i = 0; i < 4; i++) { - result[QUAD_TOP_LEFT][i] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; + result[QUAD_TOP_LEFT][i] = setup.coef[slot].a0[i] + x * dadx[i] + y * dady[i]; result[QUAD_TOP_RIGHT][i] = result[0][i] + dadx[i]; result[QUAD_BOTTOM_LEFT][i] = result[0][i] + dady[i]; result[QUAD_BOTTOM_RIGHT][i] = result[0][i] + dadx[i] + dady[i]; @@ -235,15 +240,14 @@ eval_coeff( struct setup_stage *setup, uint slot, static INLINE void -eval_z( struct setup_stage *setup, - float x, float y, float result[4]) +eval_z(float x, float y, float result[4]) { const uint slot = 0; const uint i = 2; - const float *dadx = setup->coef[slot].dadx; - const float *dady = setup->coef[slot].dady; + const float *dadx = setup.coef[slot].dadx; + const float *dady = setup.coef[slot].dady; - result[QUAD_TOP_LEFT] = setup->coef[slot].a0[i] + x * dadx[i] + y * dady[i]; + result[QUAD_TOP_LEFT] = setup.coef[slot].a0[i] + x * dadx[i] + y * dady[i]; result[QUAD_TOP_RIGHT] = result[0] + dadx[i]; result[QUAD_BOTTOM_LEFT] = result[0] + dady[i]; result[QUAD_BOTTOM_RIGHT] = result[0] + dadx[i] + dady[i]; @@ -266,23 +270,23 @@ pack_color(const float color[4]) static uint -do_depth_test(struct setup_stage *setup, int x, int y, unsigned mask) +do_depth_test(int x, int y, unsigned mask) { - int ix = x - setup->cliprect_minx; - int iy = y - setup->cliprect_miny; + int ix = x - setup.cliprect_minx; + int iy = y - setup.cliprect_miny; float zvals[4]; - eval_z(setup, (float) x, (float) y, zvals); + eval_z((float) x, (float) y, zvals); - if (tile_status_z[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { + if (tile_status_z[setup.ty][setup.tx] == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ clear_z_tile(&ztile); } - else if (tile_status_z[setup->ty][setup->tx] != TILE_STATUS_DIRTY) { + else if (tile_status_z[setup.ty][setup.tx] != TILE_STATUS_DIRTY) { /* make sure we've got the tile from main mem */ wait_on_mask(1 << TAG_READ_TILE_Z); } - tile_status_z[setup->ty][setup->tx] = TILE_STATUS_DIRTY; + tile_status_z[setup.ty][setup.tx] = TILE_STATUS_DIRTY; if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { @@ -363,31 +367,31 @@ do_depth_test(struct setup_stage *setup, int x, int y, unsigned mask) * Emit a quad (pass to next stage). No clipping is done. */ static INLINE void -emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) +emit_quad( int x, int y, unsigned mask ) { #if 0 - struct softpipe_context *sp = setup->softpipe; - setup->quad.x0 = x; - setup->quad.y0 = y; - setup->quad.mask = mask; - sp->quad.first->run(sp->quad.first, &setup->quad); + struct softpipe_context *sp = setup.softpipe; + setup.quad.x0 = x; + setup.quad.y0 = y; + setup.quad.mask = mask; + sp->quad.first->run(sp->quad.first, &setup.quad); #else /* Cell: "write" quad fragments to the tile by setting prim color */ - const int ix = x - setup->cliprect_minx; - const int iy = y - setup->cliprect_miny; + const int ix = x - setup.cliprect_minx; + const int iy = y - setup.cliprect_miny; uint colors[4]; /* indexed by QUAD_x */ if (spu.texture.start) { float texcoords[4][4]; uint i; - eval_coeff(setup, 2, (float) x, (float) y, texcoords); + eval_coeff(2, (float) x, (float) y, texcoords); for (i = 0; i < 4; i++) { colors[i] = sample_texture(texcoords[i]); } } else { float fcolors[4][4]; - eval_coeff(setup, 1, (float) x, (float) y, fcolors); + eval_coeff(1, (float) x, (float) y, fcolors); colors[QUAD_TOP_LEFT] = pack_color(fcolors[QUAD_TOP_LEFT]); colors[QUAD_TOP_RIGHT] = pack_color(fcolors[QUAD_TOP_RIGHT]); colors[QUAD_BOTTOM_LEFT] = pack_color(fcolors[QUAD_BOTTOM_LEFT]); @@ -395,19 +399,19 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) } if (spu.depth_stencil.depth.enabled) { - mask &= do_depth_test(setup, x, y, mask); + mask &= do_depth_test(x, y, mask); } if (mask) { - if (tile_status[setup->ty][setup->tx] == TILE_STATUS_CLEAR) { + if (tile_status[setup.ty][setup.tx] == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ clear_c_tile(&ctile); } - else if (tile_status[setup->ty][setup->tx] != TILE_STATUS_DIRTY) { + else if (tile_status[setup.ty][setup.tx] != TILE_STATUS_DIRTY) { /* make sure we've got the tile from main mem */ wait_on_mask(1 << TAG_READ_TILE_COLOR); } - tile_status[setup->ty][setup->tx] = TILE_STATUS_DIRTY; + tile_status[setup.ty][setup.tx] = TILE_STATUS_DIRTY; if (mask & MASK_TOP_LEFT) ctile.t32[iy][ix] = colors[QUAD_TOP_LEFT]; @@ -439,20 +443,20 @@ static INLINE int block( int x ) * this is pretty nasty... may need to rework flush_spans again to * fix it, if possible. */ -static unsigned calculate_mask( struct setup_stage *setup, int x ) +static unsigned calculate_mask( int x ) { unsigned mask = 0x0; - if (x >= setup->span.left[0] && x < setup->span.right[0]) + if (x >= setup.span.left[0] && x < setup.span.right[0]) mask |= MASK_TOP_LEFT; - if (x >= setup->span.left[1] && x < setup->span.right[1]) + if (x >= setup.span.left[1] && x < setup.span.right[1]) mask |= MASK_BOTTOM_LEFT; - if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) + if (x+1 >= setup.span.left[0] && x+1 < setup.span.right[0]) mask |= MASK_TOP_RIGHT; - if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) + if (x+1 >= setup.span.left[1] && x+1 < setup.span.right[1]) mask |= MASK_BOTTOM_RIGHT; return mask; @@ -462,28 +466,28 @@ static unsigned calculate_mask( struct setup_stage *setup, int x ) /** * Render a horizontal span of quads */ -static void flush_spans( struct setup_stage *setup ) +static void flush_spans( void ) { int minleft, maxright; int x; - switch (setup->span.y_flags) { + switch (setup.span.y_flags) { case 0x3: /* both odd and even lines written (both quad rows) */ - minleft = MIN2(setup->span.left[0], setup->span.left[1]); - maxright = MAX2(setup->span.right[0], setup->span.right[1]); + minleft = MIN2(setup.span.left[0], setup.span.left[1]); + maxright = MAX2(setup.span.right[0], setup.span.right[1]); break; case 0x1: /* only even line written (quad top row) */ - minleft = setup->span.left[0]; - maxright = setup->span.right[0]; + minleft = setup.span.left[0]; + maxright = setup.span.right[0]; break; case 0x2: /* only odd line written (quad bottom row) */ - minleft = setup->span.left[1]; - maxright = setup->span.right[1]; + minleft = setup.span.left[1]; + maxright = setup.span.right[1]; break; default: @@ -494,31 +498,29 @@ static void flush_spans( struct setup_stage *setup ) * calculate_mask() could be simplified a bit... */ for (x = block(minleft); x <= block(maxright); x += 2) { - emit_quad( setup, x, setup->span.y, - calculate_mask( setup, x ) ); + emit_quad( x, setup.span.y, + calculate_mask( x ) ); } - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; + setup.span.y = 0; + setup.span.y_flags = 0; + setup.span.right[0] = 0; + setup.span.right[1] = 0; } #if DEBUG_VERTS -static void print_vertex(const struct setup_stage *setup, - const struct vertex_header *v) +static void print_vertex(const struct vertex_header *v) { int i; fprintf(stderr, "Vertex: (%p)\n", v); - for (i = 0; i < setup->quad.nr_attrs; i++) { + for (i = 0; i < setup.quad.nr_attrs; i++) { fprintf(stderr, " %d: %f %f %f %f\n", i, v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); } } #endif -static boolean setup_sort_vertices( struct setup_stage *setup, - const struct prim_header *prim ) +static boolean setup_sort_vertices(const struct prim_header *prim ) { const struct vertex_header *v0 = prim->v[0]; const struct vertex_header *v1 = prim->v[1]; @@ -526,12 +528,12 @@ static boolean setup_sort_vertices( struct setup_stage *setup, #if DEBUG_VERTS fprintf(stderr, "Triangle:\n"); - print_vertex(setup, v0); - print_vertex(setup, v1); - print_vertex(setup, v2); + print_vertex(v0); + print_vertex(v1); + print_vertex(v2); #endif - setup->vprovoke = v2; + setup.vprovoke = v2; /* determine bottom to top order of vertices */ { @@ -541,65 +543,65 @@ static boolean setup_sort_vertices( struct setup_stage *setup, if (y0 <= y1) { if (y1 <= y2) { /* y0<=y1<=y2 */ - setup->vmin = v0; - setup->vmid = v1; - setup->vmax = v2; + setup.vmin = v0; + setup.vmid = v1; + setup.vmax = v2; } else if (y2 <= y0) { /* y2<=y0<=y1 */ - setup->vmin = v2; - setup->vmid = v0; - setup->vmax = v1; + setup.vmin = v2; + setup.vmid = v0; + setup.vmax = v1; } else { /* y0<=y2<=y1 */ - setup->vmin = v0; - setup->vmid = v2; - setup->vmax = v1; + setup.vmin = v0; + setup.vmid = v2; + setup.vmax = v1; } } else { if (y0 <= y2) { /* y1<=y0<=y2 */ - setup->vmin = v1; - setup->vmid = v0; - setup->vmax = v2; + setup.vmin = v1; + setup.vmid = v0; + setup.vmax = v2; } else if (y2 <= y1) { /* y2<=y1<=y0 */ - setup->vmin = v2; - setup->vmid = v1; - setup->vmax = v0; + setup.vmin = v2; + setup.vmid = v1; + setup.vmax = v0; } else { /* y1<=y2<=y0 */ - setup->vmin = v1; - setup->vmid = v2; - setup->vmax = v0; + setup.vmin = v1; + setup.vmid = v2; + setup.vmax = v0; } } } /* Check if triangle is completely outside the tile bounds */ - if (setup->vmin->data[0][1] > setup->cliprect_maxy) + if (setup.vmin->data[0][1] > setup.cliprect_maxy) return FALSE; - if (setup->vmax->data[0][1] < setup->cliprect_miny) + if (setup.vmax->data[0][1] < setup.cliprect_miny) return FALSE; - if (setup->vmin->data[0][0] < setup->cliprect_minx && - setup->vmid->data[0][0] < setup->cliprect_minx && - setup->vmax->data[0][0] < setup->cliprect_minx) + if (setup.vmin->data[0][0] < setup.cliprect_minx && + setup.vmid->data[0][0] < setup.cliprect_minx && + setup.vmax->data[0][0] < setup.cliprect_minx) return FALSE; - if (setup->vmin->data[0][0] > setup->cliprect_maxx && - setup->vmid->data[0][0] > setup->cliprect_maxx && - setup->vmax->data[0][0] > setup->cliprect_maxx) + if (setup.vmin->data[0][0] > setup.cliprect_maxx && + setup.vmid->data[0][0] > setup.cliprect_maxx && + setup.vmax->data[0][0] > setup.cliprect_maxx) return FALSE; - setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; - setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; - setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; + setup.ebot.dx = setup.vmid->data[0][0] - setup.vmin->data[0][0]; + setup.ebot.dy = setup.vmid->data[0][1] - setup.vmin->data[0][1]; + setup.emaj.dx = setup.vmax->data[0][0] - setup.vmin->data[0][0]; + setup.emaj.dy = setup.vmax->data[0][1] - setup.vmin->data[0][1]; + setup.etop.dx = setup.vmax->data[0][0] - setup.vmid->data[0][0]; + setup.etop.dy = setup.vmax->data[0][1] - setup.vmid->data[0][1]; /* * Compute triangle's area. Use 1/area to compute partial @@ -612,13 +614,13 @@ static boolean setup_sort_vertices( struct setup_stage *setup, * use the prim->det value because its sign is correct. */ { - const float area = (setup->emaj.dx * setup->ebot.dy - - setup->ebot.dx * setup->emaj.dy); + const float area = (setup.emaj.dx * setup.ebot.dy - + setup.ebot.dx * setup.emaj.dy); - setup->oneoverarea = 1.0f / area; + setup.oneoverarea = 1.0f / area; /* _mesa_printf("%s one-over-area %f area %f det %f\n", - __FUNCTION__, setup->oneoverarea, area, prim->det ); + __FUNCTION__, setup.oneoverarea, area, prim->det ); */ } @@ -627,7 +629,7 @@ static boolean setup_sort_vertices( struct setup_stage *setup, * - the GLSL gl_FrontFacing fragment attribute (bool) * - two-sided stencil test */ - setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); + setup.quad.facing = (prim->det > 0.0) ^ (setup.softpipe->rasterizer->front_winding == PIPE_WINDING_CW); #endif return TRUE; @@ -637,22 +639,22 @@ static boolean setup_sort_vertices( struct setup_stage *setup, /** * Compute a0 for a constant-valued coefficient (GL_FLAT shading). * The value value comes from vertex->data[slot][i]. - * The result will be put into setup->coef[slot].a0[i]. + * The result will be put into setup.coef[slot].a0[i]. * \param slot which attribute slot * \param i which component of the slot (0..3) */ -static void const_coeff(struct setup_stage *setup, uint slot) +static void const_coeff(uint slot) { uint i; ASSERT(slot < PIPE_MAX_SHADER_INPUTS); for (i = 0; i < 4; i++) { - setup->coef[slot].dadx[i] = 0; - setup->coef[slot].dady[i] = 0; + setup.coef[slot].dadx[i] = 0; + setup.coef[slot].dady[i] = 0; /* need provoking vertex info! */ - setup->coef[slot].a0[i] = setup->vprovoke->data[slot][i]; + setup.coef[slot].a0[i] = setup.vprovoke->data[slot][i]; } } @@ -661,20 +663,19 @@ static void const_coeff(struct setup_stage *setup, uint slot) * Compute a0, dadx and dady for a linearly interpolated coefficient, * for a triangle. */ -static void tri_linear_coeff( struct setup_stage *setup, - uint slot, uint firstComp, uint lastComp ) +static void tri_linear_coeff( uint slot, uint firstComp, uint lastComp ) { uint i; for (i = firstComp; i < lastComp; i++) { - float botda = setup->vmid->data[slot][i] - setup->vmin->data[slot][i]; - float majda = setup->vmax->data[slot][i] - setup->vmin->data[slot][i]; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float botda = setup.vmid->data[slot][i] - setup.vmin->data[slot][i]; + float majda = setup.vmax->data[slot][i] - setup.vmin->data[slot][i]; + float a = setup.ebot.dy * majda - botda * setup.emaj.dy; + float b = setup.emaj.dx * botda - majda * setup.ebot.dx; ASSERT(slot < PIPE_MAX_SHADER_INPUTS); - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; + setup.coef[slot].dadx[i] = a * setup.oneoverarea; + setup.coef[slot].dady[i] = b * setup.oneoverarea; /* calculate a0 as the value which would be sampled for the * fragment at (0,0), taking into account that we want to sample at @@ -688,17 +689,17 @@ static void tri_linear_coeff( struct setup_stage *setup, * to define a0 as the sample at a pixel center somewhere near vmin * instead - i'll switch to this later. */ - setup->coef[slot].a0[i] = (setup->vmin->data[slot][i] - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); + setup.coef[slot].a0[i] = (setup.vmin->data[slot][i] - + (setup.coef[slot].dadx[i] * (setup.vmin->data[0][0] - 0.5f) + + setup.coef[slot].dady[i] * (setup.vmin->data[0][1] - 0.5f))); } /* _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", slot, "xyzw"[i], - setup->coef[slot].a0[i], - setup->coef[slot].dadx[i], - setup->coef[slot].dady[i]); + setup.coef[slot].a0[i], + setup.coef[slot].dadx[i], + setup.coef[slot].dady[i]); */ } @@ -712,46 +713,45 @@ static void tri_linear_coeff( struct setup_stage *setup, * Later, when we compute the value at a particular fragment position we'll * divide the interpolated value by the interpolated W at that fragment. */ -static void tri_persp_coeff( struct setup_stage *setup, - unsigned slot, +static void tri_persp_coeff( unsigned slot, unsigned i ) { /* premultiply by 1/w: */ - float mina = setup->vmin->data[slot][i] * setup->vmin->data[0][3]; - float mida = setup->vmid->data[slot][i] * setup->vmid->data[0][3]; - float maxa = setup->vmax->data[slot][i] * setup->vmax->data[0][3]; + float mina = setup.vmin->data[slot][i] * setup.vmin->data[0][3]; + float mida = setup.vmid->data[slot][i] * setup.vmid->data[0][3]; + float maxa = setup.vmax->data[slot][i] * setup.vmax->data[0][3]; float botda = mida - mina; float majda = maxa - mina; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float a = setup.ebot.dy * majda - botda * setup.emaj.dy; + float b = setup.emaj.dx * botda - majda * setup.ebot.dx; /* printf("tri persp %d,%d: %f %f %f\n", slot, i, - setup->vmin->data[slot][i], - setup->vmid->data[slot][i], - setup->vmax->data[slot][i] + setup.vmin->data[slot][i], + setup.vmid->data[slot][i], + setup.vmax->data[slot][i] ); */ assert(slot < PIPE_MAX_SHADER_INPUTS); assert(i <= 3); - setup->coef[slot].dadx[i] = a * setup->oneoverarea; - setup->coef[slot].dady[i] = b * setup->oneoverarea; - setup->coef[slot].a0[i] = (mina - - (setup->coef[slot].dadx[i] * (setup->vmin->data[0][0] - 0.5f) + - setup->coef[slot].dady[i] * (setup->vmin->data[0][1] - 0.5f))); + setup.coef[slot].dadx[i] = a * setup.oneoverarea; + setup.coef[slot].dady[i] = b * setup.oneoverarea; + setup.coef[slot].a0[i] = (mina - + (setup.coef[slot].dadx[i] * (setup.vmin->data[0][0] - 0.5f) + + setup.coef[slot].dady[i] * (setup.vmin->data[0][1] - 0.5f))); } #endif /** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. + * Compute the setup.coef[] array dadx, dady, a0 values. + * Must be called after setup.vmin,vmid,vmax,vprovoke are initialized. */ -static void setup_tri_coefficients( struct setup_stage *setup ) +static void setup_tri_coefficients(void) { #if 1 uint i; @@ -761,17 +761,17 @@ static void setup_tri_coefficients( struct setup_stage *setup ) case INTERP_NONE: break; case INTERP_POS: - tri_linear_coeff(setup, i, 2, 3); + tri_linear_coeff(i, 2, 3); /* XXX interp W if PERSPECTIVE... */ break; case INTERP_CONSTANT: - const_coeff(setup, i); + const_coeff(i); break; case INTERP_LINEAR: - tri_linear_coeff(setup, i, 0, 4); + tri_linear_coeff(i, 0, 4); break; case INTERP_PERSPECTIVE: - tri_linear_coeff(setup, i, 0, 4); /* XXX temporary */ + tri_linear_coeff(i, 0, 4); /* XXX temporary */ break; default: ASSERT(0); @@ -781,35 +781,35 @@ static void setup_tri_coefficients( struct setup_stage *setup ) ASSERT(spu.vertex_info.interp_mode[0] == INTERP_POS); ASSERT(spu.vertex_info.interp_mode[1] == INTERP_LINEAR || spu.vertex_info.interp_mode[1] == INTERP_CONSTANT); - tri_linear_coeff(setup, 0, 2, 3); /* slot 0, z */ - tri_linear_coeff(setup, 1, 0, 4); /* slot 1, color */ + tri_linear_coeff(0, 2, 3); /* slot 0, z */ + tri_linear_coeff(1, 0, 4); /* slot 1, color */ #endif } -static void setup_tri_edges( struct setup_stage *setup ) +static void setup_tri_edges(void) { - float vmin_x = setup->vmin->data[0][0] + 0.5f; - float vmid_x = setup->vmid->data[0][0] + 0.5f; - - float vmin_y = setup->vmin->data[0][1] - 0.5f; - float vmid_y = setup->vmid->data[0][1] - 0.5f; - float vmax_y = setup->vmax->data[0][1] - 0.5f; - - setup->emaj.sy = CEILF(vmin_y); - setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); - setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; - setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; - - setup->etop.sy = CEILF(vmid_y); - setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); - setup->etop.dxdy = setup->etop.dx / setup->etop.dy; - setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; - - setup->ebot.sy = CEILF(vmin_y); - setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); - setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; - setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; + float vmin_x = setup.vmin->data[0][0] + 0.5f; + float vmid_x = setup.vmid->data[0][0] + 0.5f; + + float vmin_y = setup.vmin->data[0][1] - 0.5f; + float vmid_y = setup.vmid->data[0][1] - 0.5f; + float vmax_y = setup.vmax->data[0][1] - 0.5f; + + setup.emaj.sy = CEILF(vmin_y); + setup.emaj.lines = (int) CEILF(vmax_y - setup.emaj.sy); + setup.emaj.dxdy = setup.emaj.dx / setup.emaj.dy; + setup.emaj.sx = vmin_x + (setup.emaj.sy - vmin_y) * setup.emaj.dxdy; + + setup.etop.sy = CEILF(vmid_y); + setup.etop.lines = (int) CEILF(vmax_y - setup.etop.sy); + setup.etop.dxdy = setup.etop.dx / setup.etop.dy; + setup.etop.sx = vmid_x + (setup.etop.sy - vmid_y) * setup.etop.dxdy; + + setup.ebot.sy = CEILF(vmin_y); + setup.ebot.lines = (int) CEILF(vmid_y - setup.ebot.sy); + setup.ebot.dxdy = setup.ebot.dx / setup.ebot.dy; + setup.ebot.sx = vmin_x + (setup.ebot.sy - vmin_y) * setup.ebot.dxdy; } @@ -817,15 +817,14 @@ static void setup_tri_edges( struct setup_stage *setup ) * Render the upper or lower half of a triangle. * Scissoring/cliprect is applied here too. */ -static void subtriangle( struct setup_stage *setup, - struct edge *eleft, +static void subtriangle( struct edge *eleft, struct edge *eright, unsigned lines ) { - const int minx = setup->cliprect_minx; - const int maxx = setup->cliprect_maxx; - const int miny = setup->cliprect_miny; - const int maxy = setup->cliprect_maxy; + const int minx = setup.cliprect_minx; + const int maxx = setup.cliprect_maxx; + const int miny = setup.cliprect_miny; + const int maxy = setup.cliprect_maxy; int y, start_y, finish_y; int sy = (int)eleft->sy; @@ -867,14 +866,14 @@ static void subtriangle( struct setup_stage *setup, if (left < right) { int _y = sy + y; - if (block(_y) != setup->span.y) { - flush_spans(setup); - setup->span.y = block(_y); + if (block(_y) != setup.span.y) { + flush_spans(); + setup.span.y = block(_y); } - setup->span.left[_y&1] = left; - setup->span.right[_y&1] = right; - setup->span.y_flags |= 1<<(_y&1); + setup.span.left[_y&1] = left; + setup.span.right[_y&1] = right; + setup.span.y_flags |= 1<<(_y&1); } } @@ -892,41 +891,41 @@ static void subtriangle( struct setup_stage *setup, * Do setup for triangle rasterization, then render the triangle. */ static void -setup_tri(struct setup_stage *setup, struct prim_header *prim) +setup_tri(struct prim_header *prim) { - if (!setup_sort_vertices( setup, prim )) { + if (!setup_sort_vertices( prim )) { return; /* totally clipped */ } - setup_tri_coefficients( setup ); - setup_tri_edges( setup ); + setup_tri_coefficients(); + setup_tri_edges(); #if 0 - setup->quad.prim = PRIM_TRI; + setup.quad.prim = PRIM_TRI; #endif - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; - /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ + setup.span.y = 0; + setup.span.y_flags = 0; + setup.span.right[0] = 0; + setup.span.right[1] = 0; + /* setup.span.z_mode = tri_z_mode( setup.ctx ); */ /* init_constant_attribs( setup ); */ - if (setup->oneoverarea < 0.0) { + if (setup.oneoverarea < 0.0) { /* emaj on left: */ - subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); - subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); + subtriangle( &setup.emaj, &setup.ebot, setup.ebot.lines ); + subtriangle( &setup.emaj, &setup.etop, setup.etop.lines ); } else { /* emaj on right: */ - subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); - subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); + subtriangle( &setup.ebot, &setup.emaj, setup.ebot.lines ); + subtriangle( &setup.etop, &setup.emaj, setup.etop.lines ); } - flush_spans( setup ); + flush_spans(); } @@ -939,7 +938,7 @@ void tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) { struct prim_header tri; - struct setup_stage setup; + /*struct setup_stage setup;*/ tri.v[0] = (struct vertex_header *) v0; tri.v[1] = (struct vertex_header *) v1; @@ -954,5 +953,5 @@ tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) setup.cliprect_maxx = (tx + 1) * TILE_SIZE; setup.cliprect_maxy = (ty + 1) * TILE_SIZE; - setup_tri(&setup, &tri); + setup_tri(&tri); } -- cgit v1.2.3 From 7b149449df3a7de62f79eb96d5b722cc9d3b5912 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 Jan 2008 12:13:04 -0700 Subject: Cell: fold setup_tri() into tri_draw() --- src/mesa/pipe/cell/spu/spu_tri.c | 62 ++++++++++++---------------------------- 1 file changed, 19 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 5bb2cb12e3..1c615a6e6a 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -47,9 +47,6 @@ struct vertex_header { float data[0][4]; }; -struct prim_header { - struct vertex_header *v[3]; -}; /* XXX fix this */ @@ -520,11 +517,10 @@ static void print_vertex(const struct vertex_header *v) } #endif -static boolean setup_sort_vertices(const struct prim_header *prim ) +static boolean setup_sort_vertices(const struct vertex_header *v0, + const struct vertex_header *v1, + const struct vertex_header *v2) { - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - const struct vertex_header *v2 = prim->v[2]; #if DEBUG_VERTS fprintf(stderr, "Triangle:\n"); @@ -888,22 +884,30 @@ static void subtriangle( struct edge *eleft, /** - * Do setup for triangle rasterization, then render the triangle. + * Draw triangle into tile at (tx, ty) (tile coords) + * The tile data should have already been fetched. */ -static void -setup_tri(struct prim_header *prim) +void +tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) { - if (!setup_sort_vertices( prim )) { + setup.tx = tx; + setup.ty = ty; + + /* set clipping bounds to tile bounds */ + setup.cliprect_minx = tx * TILE_SIZE; + setup.cliprect_miny = ty * TILE_SIZE; + setup.cliprect_maxx = (tx + 1) * TILE_SIZE; + setup.cliprect_maxy = (ty + 1) * TILE_SIZE; + + if (!setup_sort_vertices((struct vertex_header *) v0, + (struct vertex_header *) v1, + (struct vertex_header *) v2)) { return; /* totally clipped */ } setup_tri_coefficients(); setup_tri_edges(); -#if 0 - setup.quad.prim = PRIM_TRI; -#endif - setup.span.y = 0; setup.span.y_flags = 0; setup.span.right[0] = 0; @@ -927,31 +931,3 @@ setup_tri(struct prim_header *prim) flush_spans(); } - - - -/** - * Draw triangle into tile at (tx, ty) (tile coords) - * The tile data should have already been fetched. - */ -void -tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) -{ - struct prim_header tri; - /*struct setup_stage setup;*/ - - tri.v[0] = (struct vertex_header *) v0; - tri.v[1] = (struct vertex_header *) v1; - tri.v[2] = (struct vertex_header *) v2; - - setup.tx = tx; - setup.ty = ty; - - /* set clipping bounds to tile bounds */ - setup.cliprect_minx = tx * TILE_SIZE; - setup.cliprect_miny = ty * TILE_SIZE; - setup.cliprect_maxx = (tx + 1) * TILE_SIZE; - setup.cliprect_maxy = (ty + 1) * TILE_SIZE; - - setup_tri(&tri); -} -- cgit v1.2.3 From 24f0e54c1b9ff43dcb75758c8e0faba355c0617c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 Jan 2008 15:26:51 -0700 Subject: Cell: start to SIMD-ize triangle attribute interpolation Using the spu_add(), etc intrinsics. About a 15% speed-up with some tests. --- src/mesa/pipe/cell/spu/spu_main.h | 7 ++ src/mesa/pipe/cell/spu/spu_texture.c | 6 +- src/mesa/pipe/cell/spu/spu_texture.h | 2 +- src/mesa/pipe/cell/spu/spu_tri.c | 126 +++++++++++++++++++---------------- 4 files changed, 79 insertions(+), 62 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 8908bf8bc0..73f9ed29d6 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -36,6 +36,13 @@ #include "pipe/p_state.h" +typedef union +{ + vector float v; + float f[4]; +} float4; + + struct spu_framebuffer { void *color_start; /**< addr of color surface in main memory */ void *depth_start; /**< addr of depth surface in main memory */ diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index 6d566a5006..7a1ca097c0 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -128,11 +128,11 @@ get_tex_tile(uint i, uint j) * XXX this is extremely primitive for now. */ uint -sample_texture(const float *texcoord) +sample_texture(float4 texcoord) { /* wrap/repeat */ - uint i = (uint) (texcoord[0] * spu.texture.width) % spu.texture.width; - uint j = (uint) (texcoord[1] * spu.texture.height) % spu.texture.height; + uint i = (uint) (texcoord.f[0] * spu.texture.width) % spu.texture.width; + uint j = (uint) (texcoord.f[1] * spu.texture.height) % spu.texture.height; uint pos = get_tex_tile(i, j); uint texel = tex_tiles[pos].t32[j % TILE_SIZE][i % TILE_SIZE]; return texel; diff --git a/src/mesa/pipe/cell/spu/spu_texture.h b/src/mesa/pipe/cell/spu/spu_texture.h index b75b7ac44f..938a42b549 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.h +++ b/src/mesa/pipe/cell/spu/spu_texture.h @@ -37,7 +37,7 @@ invalidate_tex_cache(void); extern uint -sample_texture(const float *texcoord); +sample_texture(float4 texcoord); #endif /* SPU_TEXTURE_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 1c615a6e6a..4fc6d90895 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -81,9 +81,9 @@ struct edge { struct interp_coef { - float a0[4]; - float dadx[4]; - float dady[4]; + float4 a0; + float4 dadx; + float4 dady; }; @@ -201,36 +201,31 @@ clip_emit_quad(struct setup_stage *setup) * Eg: four colors will be compute. */ static INLINE void -eval_coeff(uint slot, float x, float y, float result[4][4]) +eval_coeff(uint slot, float x, float y, float4 result[4]) { switch (spu.vertex_info.interp_mode[slot]) { case INTERP_CONSTANT: - { - uint i; - for (i = 0; i < 4; i++) { - result[QUAD_TOP_LEFT][i] = - result[QUAD_TOP_RIGHT][i] = - result[QUAD_BOTTOM_LEFT][i] = - result[QUAD_BOTTOM_RIGHT][i] = setup.coef[slot].a0[i]; - } - } + result[QUAD_TOP_LEFT] = + result[QUAD_TOP_RIGHT] = + result[QUAD_BOTTOM_LEFT] = + result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0; break; case INTERP_LINEAR: /* fall-through, for now */ default: { - uint i; - const float *dadx = setup.coef[slot].dadx; - const float *dady = setup.coef[slot].dady; - - /* loop over XYZW comps */ - for (i = 0; i < 4; i++) { - result[QUAD_TOP_LEFT][i] = setup.coef[slot].a0[i] + x * dadx[i] + y * dady[i]; - result[QUAD_TOP_RIGHT][i] = result[0][i] + dadx[i]; - result[QUAD_BOTTOM_LEFT][i] = result[0][i] + dady[i]; - result[QUAD_BOTTOM_RIGHT][i] = result[0][i] + dadx[i] + dady[i]; - } + register vector float dadx = setup.coef[slot].dadx.v; + register vector float dady = setup.coef[slot].dady.v; + register vector float topLeft + = spu_add(setup.coef[slot].a0.v, + spu_add(spu_mul(spu_splats(x), dadx), + spu_mul(spu_splats(y), dady))); + + result[QUAD_TOP_LEFT].v = topLeft; + result[QUAD_TOP_RIGHT].v = spu_add(topLeft, dadx); + result[QUAD_BOTTOM_LEFT].v = spu_add(topLeft, dady); + result[QUAD_BOTTOM_RIGHT].v = spu_add(spu_add(topLeft, dadx), dady); } } } @@ -240,28 +235,46 @@ static INLINE void eval_z(float x, float y, float result[4]) { const uint slot = 0; - const uint i = 2; - const float *dadx = setup.coef[slot].dadx; - const float *dady = setup.coef[slot].dady; - - result[QUAD_TOP_LEFT] = setup.coef[slot].a0[i] + x * dadx[i] + y * dady[i]; - result[QUAD_TOP_RIGHT] = result[0] + dadx[i]; - result[QUAD_BOTTOM_LEFT] = result[0] + dady[i]; - result[QUAD_BOTTOM_RIGHT] = result[0] + dadx[i] + dady[i]; + const float dzdx = setup.coef[slot].dadx.f[2]; + const float dzdy = setup.coef[slot].dady.f[2]; + const float topLeft = setup.coef[slot].a0.f[2] + x * dzdx + y * dzdy; +#if 1 + result[QUAD_TOP_LEFT] = topLeft; + result[QUAD_TOP_RIGHT] = topLeft + dzdx; + result[QUAD_BOTTOM_LEFT] = topLeft + dzdy; + result[QUAD_BOTTOM_RIGHT] = topLeft + dzdx + dzdy; +#else + /* XXX vectorize */ + const vector float topLeftv = spu_splats(topLeft); + const vector float derivs + = (vector float) { 0.0, dzdx, dzdy, dzdx + dzdy }; + vector float *res = (vector float *) result; + *res = spu_add(topLeftv, derivs); +#endif } -static INLINE uint -pack_color(const float color[4]) +static INLINE void +pack_colors(uint uicolors[4], const float4 fcolors[4]) { + /* XXX grab the code for _pack_rgba8() and use the shuffle + * command to do the swizzling seen here. + */ switch (spu.fb.color_format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - return _pack_rgba8(color[3], color[0], color[1], color[2]); + uicolors[0] = _pack_rgba8(fcolors[0].f[3], fcolors[0].f[0], fcolors[0].f[1], fcolors[0].f[2]); + uicolors[1] = _pack_rgba8(fcolors[1].f[3], fcolors[1].f[0], fcolors[1].f[1], fcolors[1].f[2]); + uicolors[2] = _pack_rgba8(fcolors[2].f[3], fcolors[2].f[0], fcolors[2].f[1], fcolors[2].f[2]); + uicolors[3] = _pack_rgba8(fcolors[3].f[3], fcolors[0].f[0], fcolors[3].f[1], fcolors[3].f[2]); + break; case PIPE_FORMAT_B8G8R8A8_UNORM: - return _pack_rgba8(color[2], color[1], color[0], color[3]); + uicolors[0] = _pack_rgba8(fcolors[0].f[2], fcolors[0].f[1], fcolors[0].f[0], fcolors[0].f[3]); + uicolors[1] = _pack_rgba8(fcolors[1].f[2], fcolors[1].f[1], fcolors[1].f[0], fcolors[1].f[3]); + uicolors[2] = _pack_rgba8(fcolors[2].f[2], fcolors[2].f[1], fcolors[2].f[0], fcolors[2].f[3]); + uicolors[3] = _pack_rgba8(fcolors[3].f[2], fcolors[3].f[1], fcolors[3].f[0], fcolors[3].f[3]); + break; default: ASSERT(0); - return 0; } } @@ -379,7 +392,7 @@ emit_quad( int x, int y, unsigned mask ) uint colors[4]; /* indexed by QUAD_x */ if (spu.texture.start) { - float texcoords[4][4]; + float4 texcoords[4]; uint i; eval_coeff(2, (float) x, (float) y, texcoords); for (i = 0; i < 4; i++) { @@ -387,12 +400,9 @@ emit_quad( int x, int y, unsigned mask ) } } else { - float fcolors[4][4]; + float4 fcolors[4]; eval_coeff(1, (float) x, (float) y, fcolors); - colors[QUAD_TOP_LEFT] = pack_color(fcolors[QUAD_TOP_LEFT]); - colors[QUAD_TOP_RIGHT] = pack_color(fcolors[QUAD_TOP_RIGHT]); - colors[QUAD_BOTTOM_LEFT] = pack_color(fcolors[QUAD_BOTTOM_LEFT]); - colors[QUAD_BOTTOM_RIGHT] = pack_color(fcolors[QUAD_BOTTOM_RIGHT]); + pack_colors(colors, fcolors); } if (spu.depth_stencil.depth.enabled) { @@ -645,12 +655,12 @@ static void const_coeff(uint slot) ASSERT(slot < PIPE_MAX_SHADER_INPUTS); for (i = 0; i < 4; i++) { - setup.coef[slot].dadx[i] = 0; - setup.coef[slot].dady[i] = 0; + setup.coef[slot].dadx.f[i] = 0; + setup.coef[slot].dady.f[i] = 0; /* need provoking vertex info! */ - setup.coef[slot].a0[i] = setup.vprovoke->data[slot][i]; + setup.coef[slot].a0.f[i] = setup.vprovoke->data[slot][i]; } } @@ -670,8 +680,8 @@ static void tri_linear_coeff( uint slot, uint firstComp, uint lastComp ) ASSERT(slot < PIPE_MAX_SHADER_INPUTS); - setup.coef[slot].dadx[i] = a * setup.oneoverarea; - setup.coef[slot].dady[i] = b * setup.oneoverarea; + setup.coef[slot].dadx.f[i] = a * setup.oneoverarea; + setup.coef[slot].dady.f[i] = b * setup.oneoverarea; /* calculate a0 as the value which would be sampled for the * fragment at (0,0), taking into account that we want to sample at @@ -685,17 +695,17 @@ static void tri_linear_coeff( uint slot, uint firstComp, uint lastComp ) * to define a0 as the sample at a pixel center somewhere near vmin * instead - i'll switch to this later. */ - setup.coef[slot].a0[i] = (setup.vmin->data[slot][i] - - (setup.coef[slot].dadx[i] * (setup.vmin->data[0][0] - 0.5f) + - setup.coef[slot].dady[i] * (setup.vmin->data[0][1] - 0.5f))); + setup.coef[slot].a0.f[i] = (setup.vmin->data[slot][i] - + (setup.coef[slot].dadx.f[i] * (setup.vmin->data[0][0] - 0.5f) + + setup.coef[slot].dady.f[i] * (setup.vmin->data[0][1] - 0.5f))); } /* _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", slot, "xyzw"[i], setup.coef[slot].a0[i], - setup.coef[slot].dadx[i], - setup.coef[slot].dady[i]); + setup.coef[slot].dadx.f[i], + setup.coef[slot].dady.f[i]); */ } @@ -734,11 +744,11 @@ static void tri_persp_coeff( unsigned slot, assert(slot < PIPE_MAX_SHADER_INPUTS); assert(i <= 3); - setup.coef[slot].dadx[i] = a * setup.oneoverarea; - setup.coef[slot].dady[i] = b * setup.oneoverarea; - setup.coef[slot].a0[i] = (mina - - (setup.coef[slot].dadx[i] * (setup.vmin->data[0][0] - 0.5f) + - setup.coef[slot].dady[i] * (setup.vmin->data[0][1] - 0.5f))); + setup.coef[slot].dadx.f[i] = a * setup.oneoverarea; + setup.coef[slot].dady.f[i] = b * setup.oneoverarea; + setup.coef[slot].a0.f[i] = (mina - + (setup.coef[slot].dadx.f[i] * (setup.vmin->data[0][0] - 0.5f) + + setup.coef[slot].dady.f[i] * (setup.vmin->data[0][1] - 0.5f))); } #endif -- cgit v1.2.3 From 8fb73a59939ac9ec1e41abf89a4a8c8dde09b8df Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 30 Jan 2008 20:40:26 -0700 Subject: Cell: prototype SIMD code for z testing --- src/mesa/pipe/cell/spu/spu_tile.h | 10 +++ src/mesa/pipe/cell/spu/spu_tri.c | 147 +++++++++++++++++++++++++++++--------- 2 files changed, 123 insertions(+), 34 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h index f83dc009c2..18d1b3c117 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -42,6 +42,7 @@ typedef union { ushort t16[TILE_SIZE][TILE_SIZE]; uint t32[TILE_SIZE][TILE_SIZE]; + float4 f4[TILE_SIZE/2][TILE_SIZE/2]; } tile_t; @@ -83,9 +84,18 @@ clear_z_tile(tile_t *ztile) TILE_SIZE * TILE_SIZE); } else { + ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); +#if SIMD_Z + union fi z; + z.f = 1.0; + memset32((uint*) ztile->t32, + z.i,/*spu.fb.depth_clear_value,*/ + TILE_SIZE * TILE_SIZE); +#else memset32((uint*) ztile->t32, spu.fb.depth_clear_value, TILE_SIZE * TILE_SIZE); +#endif } } diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 4fc6d90895..e436e153ec 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -40,6 +40,19 @@ #include "spu_tri.h" +/* + * If SIMD_Z=1 the Z buffer is floating point and we use vector instructions + * to do Z testing/updating. + */ +#define SIMD_Z 0 + +#if SIMD_Z +typedef vector unsigned int mask_t; +#else +typedef uint mask_t; +#endif + + /** * Simplified types taken from other parts of Gallium */ @@ -231,26 +244,16 @@ eval_coeff(uint slot, float x, float y, float4 result[4]) } -static INLINE void -eval_z(float x, float y, float result[4]) +static INLINE vector float +eval_z(float x, float y) { const uint slot = 0; const float dzdx = setup.coef[slot].dadx.f[2]; const float dzdy = setup.coef[slot].dady.f[2]; const float topLeft = setup.coef[slot].a0.f[2] + x * dzdx + y * dzdy; -#if 1 - result[QUAD_TOP_LEFT] = topLeft; - result[QUAD_TOP_RIGHT] = topLeft + dzdx; - result[QUAD_BOTTOM_LEFT] = topLeft + dzdy; - result[QUAD_BOTTOM_RIGHT] = topLeft + dzdx + dzdy; -#else - /* XXX vectorize */ const vector float topLeftv = spu_splats(topLeft); - const vector float derivs - = (vector float) { 0.0, dzdx, dzdy, dzdx + dzdy }; - vector float *res = (vector float *) result; - *res = spu_add(topLeftv, derivs); -#endif + const vector float derivs = (vector float) { 0.0, dzdx, dzdy, dzdx + dzdy }; + return spu_add(topLeftv, derivs); } @@ -279,14 +282,22 @@ pack_colors(uint uicolors[4], const float4 fcolors[4]) } -static uint -do_depth_test(int x, int y, unsigned mask) + +static unsigned int +do_depth_test(int x, int y, unsigned int mask) { + static const float4 zscale16 + = {.f={65535.0, 65535.0, 65535.0, 65535.0}}; + static const float4 zscale32 + = {.f={(float)0xffffffff, + (float)0xffffffff, + (float)0xffffffff, + (float)0xffffffff}}; int ix = x - setup.cliprect_minx; int iy = y - setup.cliprect_miny; - float zvals[4]; + float4 zvals; - eval_z((float) x, (float) y, zvals); + zvals.v = eval_z((float) x, (float) y); if (tile_status_z[setup.ty][setup.tx] == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ @@ -300,9 +311,9 @@ do_depth_test(int x, int y, unsigned mask) if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - const float zscale = 65535.0; + zvals.v = spu_mul(zvals.v, zscale16.v); if (mask & MASK_TOP_LEFT) { - uint z = (uint) (zvals[0] * zscale); + uint z = (uint) zvals.f[0]; if (z < ztile.t16[iy][ix]) ztile.t16[iy][ix] = z; else @@ -310,7 +321,7 @@ do_depth_test(int x, int y, unsigned mask) } if (mask & MASK_TOP_RIGHT) { - uint z = (uint) (zvals[1] * zscale); + uint z = (uint) zvals.f[1]; if (z < ztile.t16[iy][ix+1]) ztile.t16[iy][ix+1] = z; else @@ -318,7 +329,7 @@ do_depth_test(int x, int y, unsigned mask) } if (mask & MASK_BOTTOM_LEFT) { - uint z = (uint) (zvals[2] * zscale); + uint z = (uint) zvals.f[2]; if (z < ztile.t16[iy+1][ix]) ztile.t16[iy+1][ix] = z; else @@ -326,7 +337,7 @@ do_depth_test(int x, int y, unsigned mask) } if (mask & MASK_BOTTOM_RIGHT) { - uint z = (uint) (zvals[3] * zscale); + uint z = (uint) zvals.f[3]; if (z < ztile.t16[iy+1][ix+1]) ztile.t16[iy+1][ix+1] = z; else @@ -334,10 +345,10 @@ do_depth_test(int x, int y, unsigned mask) } } else { - const float zscale = (float) 0xffffffff; + zvals.v = spu_mul(zvals.v, zscale32.v); ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); if (mask & MASK_TOP_LEFT) { - uint z = (uint) (zvals[0] * zscale); + uint z = (uint) zvals.f[0]; if (z < ztile.t32[iy][ix]) ztile.t32[iy][ix] = z; else @@ -345,7 +356,7 @@ do_depth_test(int x, int y, unsigned mask) } if (mask & MASK_TOP_RIGHT) { - uint z = (uint) (zvals[1] * zscale); + uint z = (uint) zvals.f[1]; if (z < ztile.t32[iy][ix+1]) ztile.t32[iy][ix+1] = z; else @@ -353,7 +364,7 @@ do_depth_test(int x, int y, unsigned mask) } if (mask & MASK_BOTTOM_LEFT) { - uint z = (uint) (zvals[2] * zscale); + uint z = (uint) zvals.f[2]; if (z < ztile.t32[iy+1][ix]) ztile.t32[iy+1][ix] = z; else @@ -361,7 +372,7 @@ do_depth_test(int x, int y, unsigned mask) } if (mask & MASK_BOTTOM_RIGHT) { - uint z = (uint) (zvals[3] * zscale); + uint z = (uint) zvals.f[3]; if (z < ztile.t32[iy+1][ix+1]) ztile.t32[iy+1][ix+1] = z; else @@ -373,11 +384,45 @@ do_depth_test(int x, int y, unsigned mask) } + + +static vector unsigned int +do_depth_test_simd(int x, int y, vector unsigned int quadmask) +{ + int ix = (x - setup.cliprect_minx) / 2; + int iy = (y - setup.cliprect_miny) / 2; + float4 zvals; + + vector unsigned int zmask; + + zvals.v = eval_z((float) x, (float) y); + + if (tile_status_z[setup.ty][setup.tx] == TILE_STATUS_CLEAR) { + /* now, _really_ clear the tile */ + clear_z_tile(&ztile); + } + else if (tile_status_z[setup.ty][setup.tx] != TILE_STATUS_DIRTY) { + /* make sure we've got the tile from main mem */ + wait_on_mask(1 << TAG_READ_TILE_Z); + } + tile_status_z[setup.ty][setup.tx] = TILE_STATUS_DIRTY; + + /* XXX fetch Z value sooner to hide latency here */ + zmask = spu_cmpgt(ztile.f4[ix][iy].v, zvals.v); + zmask = spu_and(zmask, quadmask); + + ztile.f4[ix][iy].v = spu_sel(ztile.f4[ix][iy].v, zvals.v, zmask); + //ztile.f4[ix][iy].v = spu_sel(zvals.v, ztile.f4[ix][iy].v, mask4); + + return zmask; +} + + /** * Emit a quad (pass to next stage). No clipping is done. */ static INLINE void -emit_quad( int x, int y, unsigned mask ) +emit_quad( int x, int y, mask_t mask ) { #if 0 struct softpipe_context *sp = setup.softpipe; @@ -406,10 +451,17 @@ emit_quad( int x, int y, unsigned mask ) } if (spu.depth_stencil.depth.enabled) { - mask &= do_depth_test(x, y, mask); +#if SIMD_Z + mask = do_depth_test_simd(x, y, mask); +#else + mask = do_depth_test(x, y, mask); +#endif } - if (mask) { +#if !SIMD_Z + if (mask) +#endif + { if (tile_status[setup.ty][setup.tx] == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ clear_c_tile(&ctile); @@ -420,6 +472,21 @@ emit_quad( int x, int y, unsigned mask ) } tile_status[setup.ty][setup.tx] = TILE_STATUS_DIRTY; +#if SIMD_Z + if (spu_extract(mask, 0)) + ctile.t32[iy][ix] = colors[QUAD_TOP_LEFT]; + if (spu_extract(mask, 1)) + ctile.t32[iy][ix+1] = colors[QUAD_TOP_RIGHT]; + if (spu_extract(mask, 2)) + ctile.t32[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; + if (spu_extract(mask, 3)) + ctile.t32[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; +#elif 0 + /* SIMD_Z with swizzled color buffer (someday) */ + vector float icolors = *((vector float *) &colors); + ctile.f4[iy/2][ix/2].v = spu_sel(ctile.f4[iy/2][ix/2].v, icolors, mask); + +#else if (mask & MASK_TOP_LEFT) ctile.t32[iy][ix] = colors[QUAD_TOP_LEFT]; if (mask & MASK_TOP_RIGHT) @@ -428,7 +495,9 @@ emit_quad( int x, int y, unsigned mask ) ctile.t32[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; if (mask & MASK_BOTTOM_RIGHT) ctile.t32[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; +#endif } + #endif } @@ -450,8 +519,18 @@ static INLINE int block( int x ) * this is pretty nasty... may need to rework flush_spans again to * fix it, if possible. */ -static unsigned calculate_mask( int x ) +static mask_t calculate_mask( int x ) { +#if SIMD_Z + uint m0, m1, m2, m3; + + m0 = (x >= setup.span.left[0] && x < setup.span.right[0]) * ~0; + m1 = (x+1 >= setup.span.left[0] && x+1 < setup.span.right[0]) * ~0; + m2 = (x >= setup.span.left[1] && x < setup.span.right[1]) * ~0; + m3 = (x+1 >= setup.span.left[1] && x+1 < setup.span.right[1]) * ~0; + + return (vector unsigned int) {m0, m1, m2, m3}; +#else unsigned mask = 0x0; if (x >= setup.span.left[0] && x < setup.span.right[0]) @@ -467,6 +546,7 @@ static unsigned calculate_mask( int x ) mask |= MASK_BOTTOM_RIGHT; return mask; +#endif } @@ -505,8 +585,7 @@ static void flush_spans( void ) * calculate_mask() could be simplified a bit... */ for (x = block(minleft); x <= block(maxright); x += 2) { - emit_quad( x, setup.span.y, - calculate_mask( x ) ); + emit_quad( x, setup.span.y, calculate_mask( x ) ); } setup.span.y = 0; -- cgit v1.2.3 From 524bba17a75cee597f588da9c19f25d758aa237b Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 29 Jan 2008 10:37:18 -0800 Subject: Initial pass at vertex shader on SPU using TGSI VM All of the code is wired in on the SPU side, but it is not called from the PPU yet. Instruction / declaration fetch still needs to be implemented in spu_exec.c. --- src/mesa/pipe/cell/common.h | 38 + src/mesa/pipe/cell/spu/Makefile | 6 +- src/mesa/pipe/cell/spu/spu_exec.c | 2355 ++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_exec.h | 171 ++ src/mesa/pipe/cell/spu/spu_main.c | 28 + src/mesa/pipe/cell/spu/spu_util.c | 165 ++ src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 493 ++++++ src/mesa/pipe/cell/spu/spu_vertex_shader.c | 224 +++ src/mesa/pipe/cell/spu/spu_vertex_shader.h | 61 + 9 files changed, 3540 insertions(+), 1 deletion(-) create mode 100644 src/mesa/pipe/cell/spu/spu_exec.c create mode 100644 src/mesa/pipe/cell/spu/spu_exec.h create mode 100644 src/mesa/pipe/cell/spu/spu_util.c create mode 100644 src/mesa/pipe/cell/spu/spu_vertex_fetch.c create mode 100644 src/mesa/pipe/cell/spu/spu_vertex_shader.c create mode 100644 src/mesa/pipe/cell/spu/spu_vertex_shader.h (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index d5e86863d4..80a1425ec7 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -83,6 +83,9 @@ #define CELL_CMD_STATE_SAMPLER 12 #define CELL_CMD_STATE_TEXTURE 13 #define CELL_CMD_STATE_VERTEX_INFO 14 +#define CELL_CMD_STATE_VIEWPORT 15 +#define CELL_CMD_STATE_VS_ARRAY_INFO 16 +#define CELL_CMD_VS_EXECUTE 17 #define CELL_NUM_BUFFERS 4 @@ -116,6 +119,41 @@ struct cell_command_clear_surface } ALIGN16_ATTRIB; +/** + * Array info used by the vertex shader's vertex puller. + */ +struct cell_array_info +{ + void *base; /**< Base address of the 0th element. */ + uint attr; /**< Attribute that this state if for. */ + uint pitch; /**< Byte pitch from one entry to the next. */ + enum pipe_format format; /**< Pipe format of each entry. */ +} ALIGN16_ATTRIB; + + +struct cell_shader_info +{ + unsigned processor; + unsigned num_outputs; + + void *declarations; + unsigned num_declarations; + void *instructions; + unsigned num_instructions; + void *uniforms; +} ALIGN16_ATTRIB; + + +struct cell_command_vs +{ + struct cell_shader_info shader; + void *elts; + unsigned num_elts; + unsigned bytes_per_elt; + void *vOut; +} ALIGN16_ATTRIB; + + struct cell_command_render { uint opcode; /**< CELL_CMD_RENDER */ diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index d5b30e1f27..2d031bfbc6 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -20,7 +20,11 @@ SOURCES = \ spu_render.c \ spu_texture.c \ spu_tile.c \ - spu_tri.c + spu_tri.c \ + spu_exec.c \ + spu_util.c \ + spu_vertex_fetch.c \ + spu_vertex_shader.c SPU_OBJECTS = $(SOURCES:.c=.o) \ diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c new file mode 100644 index 0000000000..6888e97caf --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_exec.c @@ -0,0 +1,2355 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI interpretor/executor. + * + * Flow control information: + * + * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) + * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special + * care since a condition may be true for some quad components but false + * for other components. + * + * We basically execute all statements (even if they're in the part of + * an IF/ELSE clause that's "not taken") and use a special mask to + * control writing to destination registers. This is the ExecMask. + * See store_dest(). + * + * The ExecMask is computed from three other masks (CondMask, LoopMask and + * ContMask) which are controlled by the flow control instructions (namely: + * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). + * + * + * Authors: + * Michal Krol + * Brian Paul + */ + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/util/tgsi_util.h" +#include "spu_exec.h" + +#define TILE_TOP_LEFT 0 +#define TILE_TOP_RIGHT 1 +#define TILE_BOTTOM_LEFT 2 +#define TILE_BOTTOM_RIGHT 3 + +/* + * Shorthand locations of various utility registers (_I = Index, _C = Channel) + */ +#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I +#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C +#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I +#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C +#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I +#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C +#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I +#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C +#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I +#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C +#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I +#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C +#define TEMP_128_I TGSI_EXEC_TEMP_128_I +#define TEMP_128_C TGSI_EXEC_TEMP_128_C +#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I +#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C +#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I +#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C +#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I +#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C +#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I +#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +#define FOR_EACH_CHANNEL(CHAN)\ + for (CHAN = 0; CHAN < 4; CHAN++) + +#define IS_CHANNEL_ENABLED(INST, CHAN)\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IS_CHANNEL_ENABLED2(INST, CHAN)\ + ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) + +#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED2( INST, CHAN )) + + +/** The execution mask depends on the conditional mask and the loop mask */ +#define UPDATE_EXEC_MASK(MACH) \ + MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask + + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + + + +/** + * Initialize machine state by expanding tokens to full instructions, + * allocating temporary storage, setting up constants, etc. + * After this, we can call spu_exec_machine_run() many times. + */ +void +spu_exec_machine_init(struct spu_exec_machine *mach, + uint numSamplers, + struct spu_sampler *samplers, + unsigned processor) +{ + uint i; + + mach->Samplers = samplers; + mach->Processor = processor; + mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; + + /* Setup constants. */ + for( i = 0; i < 4; i++ ) { + mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; + mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; + mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; + mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; + mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; + mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; + mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; + mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; + } +} + + +static void +micro_abs( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->f[0] = (float) fabs( (double) src->f[0] ); + dst->f[1] = (float) fabs( (double) src->f[1] ); + dst->f[2] = (float) fabs( (double) src->f[2] ); + dst->f[3] = (float) fabs( (double) src->f[3] ); +} + +static void +micro_add( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] + src1->f[0]; + dst->f[1] = src0->f[1] + src1->f[1]; + dst->f[2] = src0->f[2] + src1->f[2]; + dst->f[3] = src0->f[3] + src1->f[3]; +} + +static void +micro_iadd( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] + src1->i[0]; + dst->i[1] = src0->i[1] + src1->i[1]; + dst->i[2] = src0->i[2] + src1->i[2]; + dst->i[3] = src0->i[3] + src1->i[3]; +} + +static void +micro_and( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] & src1->u[0]; + dst->u[1] = src0->u[1] & src1->u[1]; + dst->u[2] = src0->u[2] & src1->u[2]; + dst->u[3] = src0->u[3] & src1->u[3]; +} + +static void +micro_ceil( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ +#if 0 + dst->f[0] = (float) ceil( (double) src->f[0] ); + dst->f[1] = (float) ceil( (double) src->f[1] ); + dst->f[2] = (float) ceil( (double) src->f[2] ); + dst->f[3] = (float) ceil( (double) src->f[3] ); +#endif +} + +static void +micro_cos( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ +#if 0 + dst->f[0] = (float) cos( (double) src->f[0] ); + dst->f[1] = (float) cos( (double) src->f[1] ); + dst->f[2] = (float) cos( (double) src->f[2] ); + dst->f[3] = (float) cos( (double) src->f[3] ); +#endif +} + +static void +micro_ddx( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_ddy( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_div( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] / src1->f[0]; + dst->f[1] = src0->f[1] / src1->f[1]; + dst->f[2] = src0->f[2] / src1->f[2]; + dst->f[3] = src0->f[3] / src1->f[3]; +} + +static void +micro_udiv( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] / src1->u[0]; + dst->u[1] = src0->u[1] / src1->u[1]; + dst->u[2] = src0->u[2] / src1->u[2]; + dst->u[3] = src0->u[3] / src1->u[3]; +} + +static void +micro_eq( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1, + const union spu_exec_channel *src2, + const union spu_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ieq( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1, + const union spu_exec_channel *src2, + const union spu_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_exp2( + union spu_exec_channel *dst, + const union spu_exec_channel *src) +{ +#if 0 + dst->f[0] = (float) pow( 2.0, (double) src->f[0] ); + dst->f[1] = (float) pow( 2.0, (double) src->f[1] ); + dst->f[2] = (float) pow( 2.0, (double) src->f[2] ); + dst->f[3] = (float) pow( 2.0, (double) src->f[3] ); +#endif +} + +static void +micro_f2it( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->i[0] = (int) src->f[0]; + dst->i[1] = (int) src->f[1]; + dst->i[2] = (int) src->f[2]; + dst->i[3] = (int) src->f[3]; +} + +static void +micro_f2ut( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->u[0] = (uint) src->f[0]; + dst->u[1] = (uint) src->f[1]; + dst->u[2] = (uint) src->f[2]; + dst->u[3] = (uint) src->f[3]; +} + +static void +micro_flr( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ +#if 0 + dst->f[0] = (float) floor( (double) src->f[0] ); + dst->f[1] = (float) floor( (double) src->f[1] ); + dst->f[2] = (float) floor( (double) src->f[2] ); + dst->f[3] = (float) floor( (double) src->f[3] ); +#endif +} + +static void +micro_frc( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ +#if 0 + dst->f[0] = src->f[0] - (float) floor( (double) src->f[0] ); + dst->f[1] = src->f[1] - (float) floor( (double) src->f[1] ); + dst->f[2] = src->f[2] - (float) floor( (double) src->f[2] ); + dst->f[3] = src->f[3] - (float) floor( (double) src->f[3] ); +#endif +} + +static void +micro_ge( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1, + const union spu_exec_channel *src2, + const union spu_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_i2f( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->f[0] = (float) src->i[0]; + dst->f[1] = (float) src->i[1]; + dst->f[2] = (float) src->i[2]; + dst->f[3] = (float) src->i[3]; +} + +static void +micro_lg2( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ +#if 0 + dst->f[0] = (float) log( (double) src->f[0] ) * 1.442695f; + dst->f[1] = (float) log( (double) src->f[1] ) * 1.442695f; + dst->f[2] = (float) log( (double) src->f[2] ) * 1.442695f; + dst->f[3] = (float) log( (double) src->f[3] ) * 1.442695f; +#endif +} + +static void +micro_lt( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1, + const union spu_exec_channel *src2, + const union spu_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ilt( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1, + const union spu_exec_channel *src2, + const union spu_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_ult( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1, + const union spu_exec_channel *src2, + const union spu_exec_channel *src3 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; +} + +static void +micro_max( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imax( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umax( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_min( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imin( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umin( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_umod( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] % src1->u[0]; + dst->u[1] = src0->u[1] % src1->u[1]; + dst->u[2] = src0->u[2] % src1->u[2]; + dst->u[3] = src0->u[3] % src1->u[3]; +} + +static void +micro_mul( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] * src1->f[0]; + dst->f[1] = src0->f[1] * src1->f[1]; + dst->f[2] = src0->f[2] * src1->f[2]; + dst->f[3] = src0->f[3] * src1->f[3]; +} + +static void +micro_imul( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] * src1->i[0]; + dst->i[1] = src0->i[1] * src1->i[1]; + dst->i[2] = src0->i[2] * src1->i[2]; + dst->i[3] = src0->i[3] * src1->i[3]; +} + +static void +micro_imul64( + union spu_exec_channel *dst0, + union spu_exec_channel *dst1, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst1->i[0] = src0->i[0] * src1->i[0]; + dst1->i[1] = src0->i[1] * src1->i[1]; + dst1->i[2] = src0->i[2] * src1->i[2]; + dst1->i[3] = src0->i[3] * src1->i[3]; + dst0->i[0] = 0; + dst0->i[1] = 0; + dst0->i[2] = 0; + dst0->i[3] = 0; +} + +static void +micro_umul64( + union spu_exec_channel *dst0, + union spu_exec_channel *dst1, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst1->u[0] = src0->u[0] * src1->u[0]; + dst1->u[1] = src0->u[1] * src1->u[1]; + dst1->u[2] = src0->u[2] * src1->u[2]; + dst1->u[3] = src0->u[3] * src1->u[3]; + dst0->u[0] = 0; + dst0->u[1] = 0; + dst0->u[2] = 0; + dst0->u[3] = 0; +} + +static void +micro_movc( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1, + const union spu_exec_channel *src2 ) +{ + dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; + dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; + dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; + dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; +} + +static void +micro_neg( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->f[0] = -src->f[0]; + dst->f[1] = -src->f[1]; + dst->f[2] = -src->f[2]; + dst->f[3] = -src->f[3]; +} + +static void +micro_ineg( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->i[0] = -src->i[0]; + dst->i[1] = -src->i[1]; + dst->i[2] = -src->i[2]; + dst->i[3] = -src->i[3]; +} + +static void +micro_not( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->u[0] = ~src->u[0]; + dst->u[1] = ~src->u[1]; + dst->u[2] = ~src->u[2]; + dst->u[3] = ~src->u[3]; +} + +static void +micro_or( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] | src1->u[0]; + dst->u[1] = src0->u[1] | src1->u[1]; + dst->u[2] = src0->u[2] | src1->u[2]; + dst->u[3] = src0->u[3] | src1->u[3]; +} + +static void +micro_pow( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ +#if 0 + dst->f[0] = (float) pow( (double) src0->f[0], (double) src1->f[0] ); + dst->f[1] = (float) pow( (double) src0->f[1], (double) src1->f[1] ); + dst->f[2] = (float) pow( (double) src0->f[2], (double) src1->f[2] ); + dst->f[3] = (float) pow( (double) src0->f[3], (double) src1->f[3] ); +#endif +} + +static void +micro_rnd( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ +#if 0 + dst->f[0] = (float) floor( (double) (src->f[0] + 0.5f) ); + dst->f[1] = (float) floor( (double) (src->f[1] + 0.5f) ); + dst->f[2] = (float) floor( (double) (src->f[2] + 0.5f) ); + dst->f[3] = (float) floor( (double) (src->f[3] + 0.5f) ); +#endif +} + +static void +micro_shl( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] << src1->i[0]; + dst->i[1] = src0->i[1] << src1->i[1]; + dst->i[2] = src0->i[2] << src1->i[2]; + dst->i[3] = src0->i[3] << src1->i[3]; +} + +static void +micro_ishr( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] >> src1->i[0]; + dst->i[1] = src0->i[1] >> src1->i[1]; + dst->i[2] = src0->i[2] >> src1->i[2]; + dst->i[3] = src0->i[3] >> src1->i[3]; +} + +static void +micro_trunc( + union spu_exec_channel *dst, + const union spu_exec_channel *src0 ) +{ + dst->f[0] = (float) (int) src0->f[0]; + dst->f[1] = (float) (int) src0->f[1]; + dst->f[2] = (float) (int) src0->f[2]; + dst->f[3] = (float) (int) src0->f[3]; +} + +static void +micro_ushr( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] >> src1->u[0]; + dst->u[1] = src0->u[1] >> src1->u[1]; + dst->u[2] = src0->u[2] >> src1->u[2]; + dst->u[3] = src0->u[3] >> src1->u[3]; +} + +static void +micro_sin( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ +#if 0 + dst->f[0] = (float) sin( (double) src->f[0] ); + dst->f[1] = (float) sin( (double) src->f[1] ); + dst->f[2] = (float) sin( (double) src->f[2] ); + dst->f[3] = (float) sin( (double) src->f[3] ); +#endif +} + +static void +micro_sqrt( union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ +#if 0 + dst->f[0] = (float) sqrt( (double) src->f[0] ); + dst->f[1] = (float) sqrt( (double) src->f[1] ); + dst->f[2] = (float) sqrt( (double) src->f[2] ); + dst->f[3] = (float) sqrt( (double) src->f[3] ); +#endif +} + +static void +micro_sub( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] - src1->f[0]; + dst->f[1] = src0->f[1] - src1->f[1]; + dst->f[2] = src0->f[2] - src1->f[2]; + dst->f[3] = src0->f[3] - src1->f[3]; +} + +static void +micro_u2f( + union spu_exec_channel *dst, + const union spu_exec_channel *src ) +{ + dst->f[0] = (float) src->u[0]; + dst->f[1] = (float) src->u[1]; + dst->f[2] = (float) src->u[2]; + dst->f[3] = (float) src->u[3]; +} + +static void +micro_xor( + union spu_exec_channel *dst, + const union spu_exec_channel *src0, + const union spu_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] ^ src1->u[0]; + dst->u[1] = src0->u[1] ^ src1->u[1]; + dst->u[2] = src0->u[2] ^ src1->u[2]; + dst->u[3] = src0->u[3] ^ src1->u[3]; +} + +static void +fetch_src_file_channel( + const struct spu_exec_machine *mach, + const uint file, + const uint swizzle, + const union spu_exec_channel *index, + union spu_exec_channel *chan ) +{ + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( file ) { + case TGSI_FILE_CONSTANT: + chan->f[0] = mach->Consts[index->i[0]][swizzle]; + chan->f[1] = mach->Consts[index->i[1]][swizzle]; + chan->f[2] = mach->Consts[index->i[2]][swizzle]; + chan->f[3] = mach->Consts[index->i[3]][swizzle]; + break; + + case TGSI_FILE_INPUT: + chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_TEMPORARY: + chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_IMMEDIATE: + assert( index->i[0] < (int) mach->ImmLimit ); + assert( index->i[1] < (int) mach->ImmLimit ); + assert( index->i[2] < (int) mach->ImmLimit ); + assert( index->i[3] < (int) mach->ImmLimit ); + + chan->f[0] = mach->Imms[index->i[0]][swizzle]; + chan->f[1] = mach->Imms[index->i[1]][swizzle]; + chan->f[2] = mach->Imms[index->i[2]][swizzle]; + chan->f[3] = mach->Imms[index->i[3]][swizzle]; + break; + + case TGSI_FILE_ADDRESS: + chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_OUTPUT: + /* vertex/fragment output vars can be read too */ + chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; + break; + + case TGSI_EXTSWIZZLE_ONE: + *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; + break; + + default: + assert( 0 ); + } +} + +static void +fetch_source( + const struct spu_exec_machine *mach, + union spu_exec_channel *chan, + const struct tgsi_full_src_register *reg, + const uint chan_index ) +{ + union spu_exec_channel index; + uint swizzle; + + index.i[0] = + index.i[1] = + index.i[2] = + index.i[3] = reg->SrcRegister.Index; + + if (reg->SrcRegister.Indirect) { + union spu_exec_channel index2; + union spu_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle(®->SrcRegisterInd, + CHAN_X); + fetch_src_file_channel( + mach, + reg->SrcRegisterInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + + if( reg->SrcRegister.Dimension ) { + switch( reg->SrcRegister.File ) { + case TGSI_FILE_INPUT: + index.i[0] *= 17; + index.i[1] *= 17; + index.i[2] *= 17; + index.i[3] *= 17; + break; + case TGSI_FILE_CONSTANT: + index.i[0] *= 4096; + index.i[1] *= 4096; + index.i[2] *= 4096; + index.i[3] *= 4096; + break; + default: + assert( 0 ); + } + + index.i[0] += reg->SrcRegisterDim.Index; + index.i[1] += reg->SrcRegisterDim.Index; + index.i[2] += reg->SrcRegisterDim.Index; + index.i[3] += reg->SrcRegisterDim.Index; + + if (reg->SrcRegisterDim.Indirect) { + union spu_exec_channel index2; + union spu_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterDimInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterDimInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + } + + swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + fetch_src_file_channel( + mach, + reg->SrcRegister.File, + swizzle, + &index, + chan ); + + switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { + case TGSI_UTIL_SIGN_CLEAR: + micro_abs( chan, chan ); + break; + + case TGSI_UTIL_SIGN_SET: + micro_abs( chan, chan ); + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } + + if (reg->SrcRegisterExtMod.Complement) { + micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan ); + } +} + +static void +store_dest( + struct spu_exec_machine *mach, + const union spu_exec_channel *chan, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + uint chan_index ) +{ + union spu_exec_channel *dst; + + switch( reg->DstRegister.File ) { + case TGSI_FILE_NULL: + return; + + case TGSI_FILE_OUTPUT: + dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + + reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_TEMPORARY: + dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_ADDRESS: + dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; + break; + + default: + assert( 0 ); + return; + } + + switch (inst->Instruction.Saturate) + { + case TGSI_SAT_NONE: + if (mach->ExecMask & 0x1) + dst->i[0] = chan->i[0]; + if (mach->ExecMask & 0x2) + dst->i[1] = chan->i[1]; + if (mach->ExecMask & 0x4) + dst->i[2] = chan->i[2]; + if (mach->ExecMask & 0x8) + dst->i[3] = chan->i[3]; + break; + + case TGSI_SAT_ZERO_ONE: + /* XXX need to obey ExecMask here */ + micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); + micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + + default: + assert( 0 ); + } +} + +#define FETCH(VAL,INDEX,CHAN)\ + fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) + +#define STORE(VAL,INDEX,CHAN)\ + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) + + +/** + * Execute ARB-style KIL which is predicated by a src register. + * Kill fragment if any of the four values is less than zero. + */ +static void +exec_kilp(struct spu_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + uint uniquemask; + uint chan_index; + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + union spu_exec_channel r[1]; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + uint swizzle; + uint i; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle ( + &inst->FullSrcRegisters[0], + chan_index); + + /* check if the component has not been already tested */ + if (uniquemask & (1 << swizzle)) + continue; + uniquemask |= 1 << swizzle; + + FETCH(&r[0], 0, chan_index); + for (i = 0; i < 4; i++) + if (r[0].f[i] < 0.0f) + kilmask |= 1 << i; + } + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; +} + + +/* + * Fetch a texel using STR texture coordinates. + */ +static void +fetch_texel( struct spu_sampler *sampler, + const union spu_exec_channel *s, + const union spu_exec_channel *t, + const union spu_exec_channel *p, + float lodbias, /* XXX should be float[4] */ + union spu_exec_channel *r, + union spu_exec_channel *g, + union spu_exec_channel *b, + union spu_exec_channel *a ) +{ + uint j; + float rgba[NUM_CHANNELS][QUAD_SIZE]; + + sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); + + for (j = 0; j < 4; j++) { + r->f[j] = rgba[0][j]; + g->f[j] = rgba[1][j]; + b->f[j] = rgba[2][j]; + a->f[j] = rgba[3][j]; + } +} + + +static void +exec_tex(struct spu_exec_machine *mach, + const struct tgsi_full_instruction *inst, + boolean biasLod) +{ + const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + union spu_exec_channel r[8]; + uint chan_index; + float lodBias; + + /* printf("Sampler %u unit %u\n", sampler, unit); */ + + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: + + FETCH(&r[0], 0, CHAN_X); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[1], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[1] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[1], 0, CHAN_W); + lodBias = r[2].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ + &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ + break; + + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, /* inputs */ + &r[0], &r[1], &r[2], &r[3]); /* outputs */ + break; + + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, + &r[0], &r[1], &r[2], &r[3]); + break; + + default: + assert (0); + } + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } +} + + + +static void +constant_interpolation( + struct spu_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; + } +} + +static void +linear_interpolation( + struct spu_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + mach->Inputs[attrib].xyzw[chan].f[0] = a0; + mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; + mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; + mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; +} + +static void +perspective_interpolation( + struct spu_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + const float *w = mach->QuadPos.xyzw[3].f; + /* divide by W here */ + mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; + mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; + mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; + mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; +} + + +typedef void (* interpolation_func)( + struct spu_exec_machine *mach, + unsigned attrib, + unsigned chan ); + +static void +exec_declaration(struct spu_exec_machine *mach, + const struct tgsi_full_declaration *decl) +{ + if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + interpolation_func interp; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + switch( decl->Interpolation.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + interp = constant_interpolation; + break; + + case TGSI_INTERPOLATE_LINEAR: + interp = linear_interpolation; + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + interp = perspective_interpolation; + break; + + default: + assert( 0 ); + } + + if( mask == TGSI_WRITEMASK_XYZW ) { + unsigned i, j; + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + interp( mach, i, j ); + } + } + } + else { + unsigned i, j; + + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + interp( mach, i, j ); + } + } + } + } + } + } +} + +static void +exec_instruction( + struct spu_exec_machine *mach, + const struct tgsi_full_instruction *inst, + int *pc ) +{ + uint chan_index; + union spu_exec_channel r[8]; + + (*pc)++; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_f2it( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_X ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[1], 0, CHAN_Y ); + micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + + FETCH( &r[2], 0, CHAN_W ); + micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); + micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); + micro_pow( &r[1], &r[1], &r[2] ); + micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Z ); + } + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( &r[0], 0, CHAN_X ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( &r[0], 0, CHAN_X ); + micro_sqrt( &r[0], &r[0] ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + assert (0); + break; + + case TGSI_OPCODE_LOG: + assert (0); + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_mul( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Z ); + FETCH( &r[2], 1, CHAN_Z ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_W); + FETCH(&r[2], 1, CHAN_W); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + FETCH( &r[0], 0, CHAN_Y ); + FETCH( &r[1], 1, CHAN_Y); + micro_mul( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_Z ); + STORE( &r[0], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + FETCH( &r[0], 1, CHAN_W ); + STORE( &r[0], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + /* XXX use micro_min()?? */ + micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + /* XXX use micro_max()?? */ + micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); + + STORE(&r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_mul( &r[0], &r[0], &r[1] ); + FETCH( &r[1], 2, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_sub( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_sub( &r[1], &r[1], &r[2] ); + micro_mul( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_CND: + assert (0); + break; + + case TGSI_OPCODE_CND0: + assert (0); + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert (0); + break; + + case TGSI_OPCODE_INDEX: + assert (0); + break; + + case TGSI_OPCODE_NEGATE: + assert (0); + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_frc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + assert (0); + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_flr( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_rnd( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH(&r[0], 0, CHAN_X); + + micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( &r[0], 0, CHAN_X ); + micro_lg2( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_pow( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + FETCH(&r[0], 0, CHAN_Y); + FETCH(&r[1], 1, CHAN_Z); + + micro_mul( &r[2], &r[0], &r[1] ); + + FETCH(&r[3], 0, CHAN_Z); + FETCH(&r[4], 1, CHAN_Y); + + micro_mul( &r[5], &r[3], &r[4] ); + micro_sub( &r[2], &r[2], &r[5] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &r[2], 0, CHAN_X ); + } + + FETCH(&r[2], 1, CHAN_X); + + micro_mul( &r[3], &r[3], &r[2] ); + + FETCH(&r[5], 0, CHAN_X); + + micro_mul( &r[1], &r[1], &r[5] ); + micro_sub( &r[3], &r[3], &r[1] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + STORE( &r[3], 0, CHAN_Y ); + } + + micro_mul( &r[5], &r[5], &r[4] ); + micro_mul( &r[0], &r[0], &r[2] ); + micro_sub( &r[5], &r[5], &r[0] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( &r[5], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + assert (0); + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + + micro_abs( &r[0], &r[0] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_RCC: + assert (0); + break; + + case TGSI_OPCODE_DPH: + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 1, CHAN_W); + + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH(&r[0], 0, CHAN_X); + + micro_cos( &r[0], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddx( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDY: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddy( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_KILP: + exec_kilp (mach, inst); + break; + + case TGSI_OPCODE_KIL: + /* for enabled ExecMask bits, set the killed bit */ + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; + break; + + case TGSI_OPCODE_PK2H: + assert (0); + break; + + case TGSI_OPCODE_PK2US: + assert (0); + break; + + case TGSI_OPCODE_PK4B: + assert (0); + break; + + case TGSI_OPCODE_PK4UB: + assert (0); + break; + + case TGSI_OPCODE_RFL: + assert (0); + break; + + case TGSI_OPCODE_SEQ: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_eq( &r[0], &r[0], &r[1], + &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], + &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SFL: + assert (0); + break; + + case TGSI_OPCODE_SGT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SIN: + FETCH( &r[0], 0, CHAN_X ); + micro_sin( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SNE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_STR: + assert (0); + break; + + case TGSI_OPCODE_TEX: + /* simple texture lookup */ + /* src[0] = texcoord */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, FALSE); + break; + + case TGSI_OPCODE_TXB: + /* Texture lookup with lod bias */ + /* src[0] = texcoord (src[0].w = load bias) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE); + break; + + case TGSI_OPCODE_TXD: + /* Texture lookup with explict partial derivatives */ + /* src[0] = texcoord */ + /* src[1] = d[strq]/dx */ + /* src[2] = d[strq]/dy */ + /* src[3] = sampler unit */ + assert (0); + break; + + case TGSI_OPCODE_TXL: + /* Texture lookup with explit LOD */ + /* src[0] = texcoord (src[0].w = load bias) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE); + break; + + case TGSI_OPCODE_UP2H: + assert (0); + break; + + case TGSI_OPCODE_UP2US: + assert (0); + break; + + case TGSI_OPCODE_UP4B: + assert (0); + break; + + case TGSI_OPCODE_UP4UB: + assert (0); + break; + + case TGSI_OPCODE_X2D: + assert (0); + break; + + case TGSI_OPCODE_ARA: + assert (0); + break; + + case TGSI_OPCODE_ARR: + assert (0); + break; + + case TGSI_OPCODE_BRA: + assert (0); + break; + + case TGSI_OPCODE_CAL: + /* skip the call if no execution channels are enabled */ + if (mach->ExecMask) { + /* do the call */ + + /* push the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + + assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); + mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; + + /* note that PC was already incremented above */ + mach->CallStack[mach->CallStackTop++] = *pc; + *pc = inst->InstructionExtLabel.Label; + } + break; + + case TGSI_OPCODE_RET: + mach->FuncMask &= ~mach->ExecMask; + UPDATE_EXEC_MASK(mach); + + if (mach->ExecMask == 0x0) { + /* really return now (otherwise, keep executing */ + + if (mach->CallStackTop == 0) { + /* returning from main() */ + *pc = -1; + return; + } + *pc = mach->CallStack[--mach->CallStackTop]; + + /* pop the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->FuncStackTop > 0); + mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; + + UPDATE_EXEC_MASK(mach); + } + break; + + case TGSI_OPCODE_SSG: + assert (0); + break; + + case TGSI_OPCODE_CMP: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_SCS: + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( &r[0], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + micro_cos( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + micro_sin( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_Y ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_NRM: + assert (0); + break; + + case TGSI_OPCODE_DIV: + assert( 0 ); + break; + + case TGSI_OPCODE_DP2: + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_IF: + /* push CondMask */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + FETCH( &r[0], 0, CHAN_X ); + /* update CondMask */ + if( ! r[0].u[0] ) { + mach->CondMask &= ~0x1; + } + if( ! r[0].u[1] ) { + mach->CondMask &= ~0x2; + } + if( ! r[0].u[2] ) { + mach->CondMask &= ~0x4; + } + if( ! r[0].u[3] ) { + mach->CondMask &= ~0x8; + } + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ELSE */ + break; + + case TGSI_OPCODE_ELSE: + /* invert CondMask wrt previous mask */ + { + uint prevMask; + assert(mach->CondStackTop > 0); + prevMask = mach->CondStack[mach->CondStackTop - 1]; + mach->CondMask = ~mach->CondMask & prevMask; + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ENDIF */ + } + break; + + case TGSI_OPCODE_ENDIF: + /* pop CondMask */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_END: + /* halt execution */ + *pc = -1; + break; + + case TGSI_OPCODE_REP: + assert (0); + break; + + case TGSI_OPCODE_ENDREP: + assert (0); + break; + + case TGSI_OPCODE_PUSHA: + assert (0); + break; + + case TGSI_OPCODE_POPA: + assert (0); + break; + + case TGSI_OPCODE_CEIL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ceil( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_I2F: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_i2f( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_NOT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_not( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TRUNC: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_trunc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_shl( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ishr( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_AND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_and( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_OR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_or( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOD: + assert (0); + break; + + case TGSI_OPCODE_XOR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_xor( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SAD: + assert (0); + break; + + case TGSI_OPCODE_TXF: + assert (0); + break; + + case TGSI_OPCODE_TXQ: + assert (0); + break; + + case TGSI_OPCODE_EMIT: + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; + break; + + case TGSI_OPCODE_ENDPRIM: + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; + break; + + case TGSI_OPCODE_LOOP: + /* fall-through (for now) */ + case TGSI_OPCODE_BGNLOOP2: + /* push LoopMask and ContMasks */ + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + break; + + case TGSI_OPCODE_ENDLOOP: + /* fall-through (for now at least) */ + case TGSI_OPCODE_ENDLOOP2: + /* Restore ContMask, but don't pop */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; + if (mach->LoopMask) { + /* repeat loop: jump to instruction just past BGNLOOP */ + *pc = inst->InstructionExtLabel.Label + 1; + } + else { + /* exit loop: pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + /* pop ContMask */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + } + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BRK: + /* turn off loop channels for each enabled exec channel */ + mach->LoopMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_CONT: + /* turn off cont channels for each enabled exec channel */ + mach->ContMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BGNSUB: + /* no-op */ + break; + + case TGSI_OPCODE_ENDSUB: + /* no-op */ + break; + + case TGSI_OPCODE_NOISE1: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE2: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE3: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE4: + assert( 0 ); + break; + + case TGSI_OPCODE_NOP: + break; + + default: + assert( 0 ); + } +} + + +/** + * Run TGSI interpreter. + * \return bitmask of "alive" quad components + */ +uint +spu_exec_machine_run( struct spu_exec_machine *mach ) +{ + uint i; + int pc = 0; + + mach->CondMask = 0xf; + mach->LoopMask = 0xf; + mach->ContMask = 0xf; + mach->FuncMask = 0xf; + mach->ExecMask = 0xf; + + mach->CondStackTop = 0; /* temporarily subvert this assertion */ + assert(mach->CondStackTop == 0); + assert(mach->LoopStackTop == 0); + assert(mach->ContStackTop == 0); + assert(mach->CallStackTop == 0); + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; + + if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; + mach->Primitives[0] = 0; + } + + + /* execute declarations (interpolants) */ + for (i = 0; i < mach->NumDeclarations; i++) { + exec_declaration( mach, mach->Declarations+i ); + } + + /* execute instructions, until pc is set to -1 */ + while (pc != -1) { + assert(pc < mach->NumInstructions); + exec_instruction( mach, mach->Instructions + pc, &pc ); + } + +#if 0 + /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ + if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { + /* + * Scale back depth component. + */ + for (i = 0; i < 4; i++) + mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; + } +#endif + + return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; +} + + diff --git a/src/mesa/pipe/cell/spu/spu_exec.h b/src/mesa/pipe/cell/spu/spu_exec.h new file mode 100644 index 0000000000..89e422ba48 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_exec.h @@ -0,0 +1,171 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#if !defined SPU_EXEC_H +#define SPU_EXEC_H + +#include "pipe/p_compiler.h" +#include "pipe/tgsi/exec/tgsi_exec.h" + +#if defined __cplusplus +extern "C" { +#endif + +/** + * Registers may be treated as float, signed int or unsigned int. + */ +union spu_exec_channel +{ + float f[QUAD_SIZE]; + int i[QUAD_SIZE]; + unsigned u[QUAD_SIZE]; +}; + +/** + * A vector[RGBA] of channels[4 pixels] + */ +struct spu_exec_vector +{ + union spu_exec_channel xyzw[NUM_CHANNELS]; +}; + +/** + * For fragment programs, information for computing fragment input + * values from plane equation of the triangle/line. + */ +struct spu_interp_coef +{ + float a0[NUM_CHANNELS]; /* in an xyzw layout */ + float dadx[NUM_CHANNELS]; + float dady[NUM_CHANNELS]; +}; + + +struct softpipe_tile_cache; /**< Opaque to TGSI */ + +/** + * Information for sampling textures, which must be implemented + * by code outside the TGSI executor. + */ +struct spu_sampler +{ + const struct pipe_sampler_state *state; + struct pipe_texture *texture; + /** Get samples for four fragments in a quad */ + void (*get_samples)(struct spu_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); + void *pipe; /*XXX temporary*/ + struct softpipe_tile_cache *cache; +}; + + +/** + * Run-time virtual machine state for executing TGSI shader. + */ +struct spu_exec_machine +{ + /* + * 32 program temporaries + * 4 internal temporaries + * 1 address + */ + struct spu_exec_vector Temps[TGSI_EXEC_NUM_TEMPS + + TGSI_EXEC_NUM_ADDRS + 1] + ALIGN16_ATTRIB; + + struct spu_exec_vector *Addrs; + + struct spu_sampler *Samplers; + + float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; + unsigned ImmLimit; + float (*Consts)[4]; + struct spu_exec_vector *Inputs; + struct spu_exec_vector *Outputs; + unsigned Processor; + + /* GEOMETRY processor only. */ + unsigned *Primitives; + + /* FRAGMENT processor only. */ + const struct spu_interp_coef *InterpCoefs; + struct spu_exec_vector QuadPos; + + /* Conditional execution masks */ + uint CondMask; /**< For IF/ELSE/ENDIF */ + uint LoopMask; /**< For BGNLOOP/ENDLOOP */ + uint ContMask; /**< For loop CONT statements */ + uint FuncMask; /**< For function calls */ + uint ExecMask; /**< = CondMask & LoopMask */ + + /** Condition mask stack (for nested conditionals) */ + uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; + int CondStackTop; + + /** Loop mask stack (for nested loops) */ + uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopStackTop; + + /** Loop continue mask stack (see comments in tgsi_exec.c) */ + uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int ContStackTop; + + /** Function execution mask stack (for executing subroutine code) */ + uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; + int FuncStackTop; + + /** Function call stack for saving/restoring the program counter */ + uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; + int CallStackTop; + + struct tgsi_full_instruction *Instructions; + uint NumInstructions; + + struct tgsi_full_declaration *Declarations; + uint NumDeclarations; +}; + + +extern void +spu_exec_machine_init(struct spu_exec_machine *mach, + uint numSamplers, + struct spu_sampler *samplers, + unsigned processor); + +extern uint +spu_exec_machine_run( struct spu_exec_machine *mach ); + + +#if defined __cplusplus +} /* extern "C" */ +#endif + +#endif /* SPU_EXEC_H */ diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 6886f283be..9daa3ec735 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -36,6 +36,7 @@ #include "spu_render.h" #include "spu_texture.h" #include "spu_tile.h" +#include "spu_vertex_shader.h" #include "pipe/cell/common.h" #include "pipe/p_defines.h" @@ -50,6 +51,7 @@ boolean Debug = FALSE; struct spu_global spu; +struct spu_vs_context draw; /** * Tell the PPU that this SPU has finished copying a buffer to @@ -264,6 +266,18 @@ cmd_state_vertex_info(const struct vertex_info *vinfo) } +static void +cmd_state_vs_array_info(const struct cell_array_info *vs_info) +{ + const unsigned attr = vs_info->attr; + + ASSERT(attr < PIPE_ATTRIB_MAX); + draw.vertex_fetch.src_ptr[attr] = vs_info->base; + draw.vertex_fetch.pitch[attr] = vs_info->pitch; + draw.vertex_fetch.format[attr] = vs_info->format; + draw.vertex_fetch.dirty = 1; +} + static void cmd_finish(void) @@ -374,6 +388,20 @@ cmd_batch(uint opcode) cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]); pos += (1 + sizeof(struct vertex_info) / 4); break; + case CELL_CMD_STATE_VIEWPORT: + (void) memcpy(& draw.viewport, &buffer[pos+1], + sizeof(struct pipe_viewport_state)); + pos += (1 + sizeof(struct pipe_viewport_state) / 4); + break; + case CELL_CMD_STATE_VS_ARRAY_INFO: + cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); + pos += (1 + sizeof(struct cell_array_info) / 4); + break; + case CELL_CMD_VS_EXECUTE: + spu_execute_vertex_shader(&draw, + (struct cell_command_vs *) &buffer[pos+1]); + pos += (1 + sizeof(struct cell_command_vs) / 4); + break; default: printf("SPU %u: bad opcode: 0x%x\n", spu.init.id, buffer[pos]); ASSERT(0); diff --git a/src/mesa/pipe/cell/spu/spu_util.c b/src/mesa/pipe/cell/spu/spu_util.c new file mode 100644 index 0000000000..ac373240c1 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_util.c @@ -0,0 +1,165 @@ +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +//#include "tgsi_build.h" +#include "pipe/tgsi/util/tgsi_util.h" + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->SwizzleX; + case 1: + return reg->SwizzleY; + case 2: + return reg->SwizzleZ; + case 3: + return reg->SwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->ExtSwizzleX; + case 1: + return reg->ExtSwizzleY; + case 2: + return reg->ExtSwizzleZ; + case 3: + return reg->ExtSwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned swizzle; + + /* + * First, calculate the extended swizzle for a given channel. This will give + * us either a channel index into the simple swizzle or a constant 1 or 0. + */ + swizzle = tgsi_util_get_src_register_extswizzle( + ®->SrcRegisterExtSwz, + component ); + + assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); + assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); + assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); + assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); + + /* + * Second, calculate the simple swizzle for the unswizzled channel index. + * Leave the constants intact, they are not affected by the simple swizzle. + */ + if( swizzle <= TGSI_SWIZZLE_W ) { + swizzle = tgsi_util_get_src_register_swizzle( + ®->SrcRegister, + component ); + } + + return swizzle; +} + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->NegateX; + case 1: + return reg->NegateY; + case 2: + return reg->NegateZ; + case 3: + return reg->NegateW; + default: + assert( 0 ); + } + return 0; +} + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ) +{ + switch( component ) { + case 0: + reg->NegateX = negate; + break; + case 1: + reg->NegateY = negate; + break; + case 2: + reg->NegateZ = negate; + break; + case 3: + reg->NegateW = negate; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned sign_mode; + + if( reg->SrcRegisterExtMod.Absolute ) { + /* Consider only the post-abs negation. */ + + if( reg->SrcRegisterExtMod.Negate ) { + sign_mode = TGSI_UTIL_SIGN_SET; + } + else { + sign_mode = TGSI_UTIL_SIGN_CLEAR; + } + } + else { + /* Accumulate the three negations. */ + + unsigned negate; + + negate = reg->SrcRegister.Negate; + if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { + negate = !negate; + } + if( reg->SrcRegisterExtMod.Negate ) { + negate = !negate; + } + + if( negate ) { + sign_mode = TGSI_UTIL_SIGN_TOGGLE; + } + else { + sign_mode = TGSI_UTIL_SIGN_KEEP; + } + } + + return sign_mode; +} diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c new file mode 100644 index 0000000000..b8f8c52eed --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -0,0 +1,493 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_state.h" +#include "pipe/p_shader_tokens.h" +#include "spu_exec.h" +#include "spu_vertex_shader.h" + + +#define DRAW_DBG 0 + + +/** + * Fetch a float[4] vertex attribute from memory, doing format/type + * conversion as needed. + * + * This is probably needed/dupliocated elsewhere, eg format + * conversion, texture sampling etc. + */ +#define FETCH_ATTRIB( NAME, SZ, CVT ) \ +static void \ +fetch_##NAME(const void *ptr, float *attrib) \ +{ \ + static const float defaults[4] = { 0,0,0,1 }; \ + int i; \ + \ + for (i = 0; i < SZ; i++) { \ + attrib[i] = CVT; \ + } \ + \ + for (; i < 4; i++) { \ + attrib[i] = defaults[i]; \ + } \ +} + +#define CVT_64_FLOAT (float) ((double *) ptr)[i] +#define CVT_32_FLOAT ((float *) ptr)[i] + +#define CVT_8_USCALED (float) ((unsigned char *) ptr)[i] +#define CVT_16_USCALED (float) ((unsigned short *) ptr)[i] +#define CVT_32_USCALED (float) ((unsigned int *) ptr)[i] + +#define CVT_8_SSCALED (float) ((char *) ptr)[i] +#define CVT_16_SSCALED (float) ((short *) ptr)[i] +#define CVT_32_SSCALED (float) ((int *) ptr)[i] + +#define CVT_8_UNORM (float) ((unsigned char *) ptr)[i] / 255.0f +#define CVT_16_UNORM (float) ((unsigned short *) ptr)[i] / 65535.0f +#define CVT_32_UNORM (float) ((unsigned int *) ptr)[i] / 4294967295.0f + +#define CVT_8_SNORM (float) ((char *) ptr)[i] / 127.0f +#define CVT_16_SNORM (float) ((short *) ptr)[i] / 32767.0f +#define CVT_32_SNORM (float) ((int *) ptr)[i] / 2147483647.0f + +FETCH_ATTRIB( R64G64B64A64_FLOAT, 4, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) +FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) + +FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) +FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) + +FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) +FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) + +FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) +FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) + +FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) +FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) + +FETCH_ATTRIB( R32G32B32A32_SNORM, 4, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32B32_SNORM, 3, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32_SNORM, 2, CVT_32_SNORM ) +FETCH_ATTRIB( R32_SNORM, 1, CVT_32_SNORM ) + +FETCH_ATTRIB( R16G16B16A16_USCALED, 4, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16B16_USCALED, 3, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16_USCALED, 2, CVT_16_USCALED ) +FETCH_ATTRIB( R16_USCALED, 1, CVT_16_USCALED ) + +FETCH_ATTRIB( R16G16B16A16_SSCALED, 4, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16B16_SSCALED, 3, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16_SSCALED, 2, CVT_16_SSCALED ) +FETCH_ATTRIB( R16_SSCALED, 1, CVT_16_SSCALED ) + +FETCH_ATTRIB( R16G16B16A16_UNORM, 4, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16B16_UNORM, 3, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16_UNORM, 2, CVT_16_UNORM ) +FETCH_ATTRIB( R16_UNORM, 1, CVT_16_UNORM ) + +FETCH_ATTRIB( R16G16B16A16_SNORM, 4, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16B16_SNORM, 3, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16_SNORM, 2, CVT_16_SNORM ) +FETCH_ATTRIB( R16_SNORM, 1, CVT_16_SNORM ) + +FETCH_ATTRIB( R8G8B8A8_USCALED, 4, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8B8_USCALED, 3, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8_USCALED, 2, CVT_8_USCALED ) +FETCH_ATTRIB( R8_USCALED, 1, CVT_8_USCALED ) + +FETCH_ATTRIB( R8G8B8A8_SSCALED, 4, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8B8_SSCALED, 3, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8_SSCALED, 2, CVT_8_SSCALED ) +FETCH_ATTRIB( R8_SSCALED, 1, CVT_8_SSCALED ) + +FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8B8_UNORM, 3, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8_UNORM, 2, CVT_8_UNORM ) +FETCH_ATTRIB( R8_UNORM, 1, CVT_8_UNORM ) + +FETCH_ATTRIB( R8G8B8A8_SNORM, 4, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8B8_SNORM, 3, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8_SNORM, 2, CVT_8_SNORM ) +FETCH_ATTRIB( R8_SNORM, 1, CVT_8_SNORM ) + +FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) +//FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) + + + +static spu_fetch_func get_fetch_func( enum pipe_format format ) +{ +#if 0 + { + char tmp[80]; + pf_sprint_name(tmp, format); + _mesa_printf("%s: %s\n", __FUNCTION__, tmp); + } +#endif + + switch (format) { + case PIPE_FORMAT_R64_FLOAT: + return fetch_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return fetch_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return fetch_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return fetch_R64G64B64A64_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + return fetch_R32_FLOAT; + case PIPE_FORMAT_R32G32_FLOAT: + return fetch_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return fetch_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return fetch_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return fetch_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return fetch_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return fetch_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return fetch_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return fetch_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return fetch_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return fetch_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return fetch_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return fetch_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return fetch_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return fetch_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return fetch_R32G32B32A32_SNORM; + + case PIPE_FORMAT_R32_SSCALED: + return fetch_R32_SSCALED; + case PIPE_FORMAT_R32G32_SSCALED: + return fetch_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return fetch_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return fetch_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return fetch_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return fetch_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return fetch_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return fetch_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return fetch_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return fetch_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return fetch_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return fetch_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return fetch_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return fetch_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return fetch_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return fetch_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return fetch_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return fetch_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return fetch_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return fetch_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return fetch_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return fetch_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return fetch_R8G8B8_UNORM; + case PIPE_FORMAT_R8G8B8A8_UNORM: + return fetch_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return fetch_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return fetch_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return fetch_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return fetch_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return fetch_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return fetch_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return fetch_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return fetch_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return fetch_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return fetch_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return fetch_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return fetch_R8G8B8A8_SSCALED; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + return fetch_A8R8G8B8_UNORM; + + case 0: + return NULL; /* not sure why this is needed */ + + default: + assert(0); + return NULL; + } +} + + +static void +transpose_4x4( float *out, const float *in ) +{ + /* This can be achieved in 12 sse instructions, plus the final + * stores I guess. This is probably a bit more than that - maybe + * 32 or so? + */ + out[0] = in[0]; out[1] = in[4]; out[2] = in[8]; out[3] = in[12]; + out[4] = in[1]; out[5] = in[5]; out[6] = in[9]; out[7] = in[13]; + out[8] = in[2]; out[9] = in[6]; out[10] = in[10]; out[11] = in[14]; + out[12] = in[3]; out[13] = in[7]; out[14] = in[11]; out[15] = in[15]; +} + + + +static void fetch_xyz_rgb( struct spu_vs_context *draw, + struct spu_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + assert(count <= 4); + +// _mesa_printf("%s\n", __FUNCTION__); + + /* loop over vertex attributes (vertex shader inputs) + */ + + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + + for (i = 0; i < 4; i++) { + { + const float *in = (const float *)(src[0] + elts[i] * pitch[0]); + float *out = &machine->Inputs[0].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[1] + elts[i] * pitch[1]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + } +} + + + + +static void fetch_xyz_rgb_st( struct spu_vs_context *draw, + struct spu_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + assert(count <= 4); + + /* loop over vertex attributes (vertex shader inputs) + */ + + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + + for (i = 0; i < 4; i++) { + { + const float *in = (const float *)(src[0] + elts[i] * pitch[0]); + float *out = &machine->Inputs[0].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[1] + elts[i] * pitch[1]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[2] + elts[i] * pitch[2]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = 0.0f; + out[12] = 1.0f; + } + } +} + + + + +/** + * Fetch vertex attributes for 'count' vertices. + */ +static void generic_vertex_fetch( struct spu_vs_context *draw, + struct spu_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + unsigned nr_attrs = draw->vertex_fetch.nr_attrs; + unsigned attr; + + assert(count <= 4); + +// _mesa_printf("%s %d\n", __FUNCTION__, count); + + /* loop over vertex attributes (vertex shader inputs) + */ + for (attr = 0; attr < nr_attrs; attr++) { + + const unsigned pitch = draw->vertex_fetch.pitch[attr]; + const ubyte *src = draw->vertex_fetch.src_ptr[attr]; + const spu_fetch_func fetch = draw->vertex_fetch.fetch[attr]; + unsigned i; + float p[4][4]; + + + /* Fetch four attributes for four vertices. + * + * Could fetch directly into AOS format, but this is meant to be + * a prototype for an sse implementation, which would have + * difficulties doing that. + */ + for (i = 0; i < count; i++) + fetch( src + elts[i] * pitch, p[i] ); + + /* Be nice and zero out any missing vertices: + */ + for (/* empty */; i < 4; i++) + p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; + + /* Transpose/swizzle into sse-friendly format. Currently + * assuming that all vertex shader inputs are float[4], but this + * isn't true -- if the vertex shader only wants tex0.xy, we + * could optimize for that. + * + * To do so fully without codegen would probably require an + * excessive number of fetch functions, but we could at least + * minimize the transpose step: + */ + transpose_4x4( (float *)&machine->Inputs[attr].xyzw[0].f[0], (float *)p ); + } +} + + +void spu_update_vertex_fetch( struct spu_vs_context *draw ) +{ + unsigned i; + + + for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { + draw->vertex_fetch.fetch[i] = + get_fetch_func(draw->vertex_fetch.format[i]); + } + + draw->vertex_fetch.fetch_func = generic_vertex_fetch; + + switch (draw->vertex_fetch.nr_attrs) { + case 2: + if (draw->vertex_fetch.format[0] == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_fetch.format[1] == PIPE_FORMAT_R32G32B32_FLOAT) + draw->vertex_fetch.fetch_func = fetch_xyz_rgb; + break; + case 3: + if (draw->vertex_fetch.format[0] == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_fetch.format[1] == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_fetch.format[2] == PIPE_FORMAT_R32G32_FLOAT) + draw->vertex_fetch.fetch_func = fetch_xyz_rgb_st; + break; + default: + break; + } +} diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.c b/src/mesa/pipe/cell/spu/spu_vertex_shader.c new file mode 100644 index 0000000000..e694ff729f --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.c @@ -0,0 +1,224 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + * Ian Romanick + */ + +#include "pipe/p_util.h" +#include "pipe/p_state.h" +#include "pipe/p_shader_tokens.h" +#include "spu_vertex_shader.h" +#include "spu_exec.h" +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_context.h" +#include "pipe/cell/common.h" + +#define DBG_VS 0 + + +static INLINE unsigned +compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<machine; + unsigned int j; + + ALIGN16_DECL(struct spu_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct spu_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); + + /* Consts does not require 16 byte alignment. */ + ASSERT_ALIGN16(draw->constants); + machine->Consts = (float (*)[4]) draw->constants; + + machine->Inputs = ALIGN16_ASSIGN(inputs); + machine->Outputs = ALIGN16_ASSIGN(outputs); + + spu_vertex_fetch( draw, machine, elts, count ); + + /* run shader */ + spu_exec_machine_run( machine ); + + + /* store machine results */ + for (j = 0; j < count; j++) { + unsigned slot; + float x, y, z, w; + + /* Handle attr[0] (position) specially: + * + * XXX: Computing the clipmask should be done in the vertex + * program as a set of DP4 instructions appended to the + * user-provided code. + */ + x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, + draw->nr_planes); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + +#if DBG_VS + printf("output[%d]win: %f %f %f %f\n", j, + vOut[j]->data[0][0], + vOut[j]->data[0][1], + vOut[j]->data[0][2], + vOut[j]->data[0][3]); +#endif + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; +#if DBG_VS + printf("output[%d][%d]: %f %f %f %f\n", j, slot, + vOut[j]->data[slot][0], + vOut[j]->data[slot][1], + vOut[j]->data[slot][2], + vOut[j]->data[slot][3]); +#endif + } + } /* loop over vertices */ +} + + +static void +spu_bind_vertex_shader(struct spu_vs_context *draw, + void *uniforms, + void *planes, + unsigned nr_planes, + unsigned num_outputs + ) +{ + draw->constants = (float (*)[4]) uniforms; + + (void) memcpy(draw->plane, planes, sizeof(float) * 4 * nr_planes); + draw->nr_planes = nr_planes; + draw->num_vs_outputs = num_outputs; + + /* specify the shader to interpret/execute */ + spu_exec_machine_init(&draw->machine, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/, + PIPE_SHADER_VERTEX); +} + + +void +spu_execute_vertex_shader(struct spu_vs_context *draw, + const struct cell_command_vs *vs) +{ + unsigned i; + unsigned j; + + draw->machine.Instructions = (struct tgsi_full_instruction *) + vs->shader.instructions; + draw->machine.NumInstructions = vs->shader.num_instructions; + + draw->machine.Declarations = (struct tgsi_full_declaration *) + vs->shader.declarations; + draw->machine.NumDeclarations = vs->shader.num_declarations; + + spu_bind_vertex_shader(draw, vs->shader.uniforms, + NULL, -1, + vs->shader.num_outputs); + + for (i = 0; i < vs->num_elts; i += 4) { + const unsigned batch_size = MIN2(vs->num_elts - i, 4); + unsigned elts[4]; + + for (j = 0; j < batch_size; j++) { + switch (vs->bytes_per_elt) { + case 1: elts[j] = ((unsigned char *) vs->elts)[i + j]; break; + case 2: elts[j] = ((unsigned short *)vs->elts)[i + j]; break; + case 4: elts[j] = ((unsigned int *) vs->elts)[i + j]; break; + } + } + + run_vertex_program(draw, elts, batch_size, + (struct vertex_header (*)[]) vs->vOut); + } +} diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.h b/src/mesa/pipe/cell/spu/spu_vertex_shader.h new file mode 100644 index 0000000000..c52f38fd02 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.h @@ -0,0 +1,61 @@ +#ifndef SPU_VERTEX_SHADER_H +#define SPU_VERTEX_SHADER_H + +#include "pipe/p_format.h" +#include "spu_exec.h" + +struct spu_vs_context; + +typedef void (*spu_fetch_func)(const void *ptr, float *attrib); +typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw, + struct spu_exec_machine *machine, + const unsigned *elts, + unsigned count ); + +struct spu_vs_context { + struct pipe_viewport_state viewport; + + struct { + const ubyte *src_ptr[PIPE_ATTRIB_MAX]; + unsigned pitch[PIPE_ATTRIB_MAX]; + enum pipe_format format[PIPE_ATTRIB_MAX]; + unsigned nr_attrs; + boolean dirty; + + spu_fetch_func fetch[PIPE_ATTRIB_MAX]; + spu_full_fetch_func fetch_func; + } vertex_fetch; + + /* Clip derived state: + */ + float plane[12][4]; + unsigned nr_planes; + + struct spu_exec_machine machine; + const float (*constants)[4]; + + unsigned num_vs_outputs; +}; + +extern void spu_update_vertex_fetch(struct spu_vs_context *draw); + +static INLINE void spu_vertex_fetch(struct spu_vs_context *draw, + struct spu_exec_machine *machine, + const unsigned *elts, + unsigned count) +{ + if (draw->vertex_fetch.dirty) { + spu_update_vertex_fetch(draw); + draw->vertex_fetch.dirty = 0; + } + + (*draw->vertex_fetch.fetch_func)(draw, machine, elts, count); +} + +struct cell_command_vs; + +extern void +spu_execute_vertex_shader(struct spu_vs_context *draw, + const struct cell_command_vs *vs); + +#endif /* SPU_VERTEX_SHADER_H */ -- cgit v1.2.3 From 33cac4824195337d9cf3dfda3fc1147c429ae43c Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 29 Jan 2008 10:56:53 -0800 Subject: Initial pass at instruction / declaration fetch --- src/mesa/pipe/cell/spu/spu_exec.c | 22 ++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_main.h | 1 + 2 files changed, 23 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c index 6888e97caf..f43278198e 100644 --- a/src/mesa/pipe/cell/spu/spu_exec.c +++ b/src/mesa/pipe/cell/spu/spu_exec.c @@ -50,6 +50,9 @@ * Brian Paul */ +#include +#include + #include "pipe/p_compiler.h" #include "pipe/p_state.h" #include "pipe/p_util.h" @@ -57,6 +60,7 @@ #include "pipe/tgsi/util/tgsi_parse.h" #include "pipe/tgsi/util/tgsi_util.h" #include "spu_exec.h" +#include "spu_main.h" #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 @@ -2329,12 +2333,30 @@ spu_exec_machine_run( struct spu_exec_machine *mach ) /* execute declarations (interpolants) */ for (i = 0; i < mach->NumDeclarations; i++) { + uint8_t buffer[sizeof(struct tgsi_full_declaration) + 32] ALIGN16_ATTRIB; + struct tgsi_full_declaration decl; + unsigned long decl_addr = (unsigned long) (mach->Declarations+i); + unsigned size = ((sizeof(decl) + (decl_addr & 0x0f) + 0x0f) & ~0x0f); + + mfc_get(buffer, decl_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); + wait_on_mask(1 << TAG_INSTRUCTION_FETCH); + + memcpy(& decl, buffer + (decl_addr & 0x0f), sizeof(decl)); exec_declaration( mach, mach->Declarations+i ); } /* execute instructions, until pc is set to -1 */ while (pc != -1) { + uint8_t buffer[sizeof(struct tgsi_full_instruction) + 32] ALIGN16_ATTRIB; + struct tgsi_full_instruction inst; + unsigned long inst_addr = (unsigned long) (mach->Instructions + pc); + unsigned size = ((sizeof(inst) + (inst_addr & 0x0f) + 0x0f) & ~0x0f); + assert(pc < mach->NumInstructions); + mfc_get(buffer, inst_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); + wait_on_mask(1 << TAG_INSTRUCTION_FETCH); + + memcpy(& inst, buffer + (inst_addr & 0x0f), sizeof(inst)); exec_instruction( mach, mach->Instructions + pc, &pc ); } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 73f9ed29d6..8be5268f52 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -96,6 +96,7 @@ extern boolean Debug; #define TAG_BATCH_BUFFER 17 #define TAG_MISC 18 #define TAG_TEXTURE_TILE 19 +#define TAG_INSTRUCTION_FETCH 20 -- cgit v1.2.3 From 13eec106881b846538bef13d694c9d2d9cf1ae6b Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 29 Jan 2008 11:28:06 -0800 Subject: Implement vertex fetch / vertex shader output write-back --- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 32 +++++++++++---- src/mesa/pipe/cell/spu/spu_vertex_shader.c | 62 +++++++++++++++--------------- src/mesa/pipe/draw/draw_context.c | 5 ++- 3 files changed, 58 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index b8f8c52eed..0192227d57 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -30,11 +30,13 @@ * Keith Whitwell */ +#include #include "pipe/p_util.h" #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" #include "spu_exec.h" #include "spu_vertex_shader.h" +#include "spu_main.h" #define DRAW_DBG 0 @@ -412,16 +414,18 @@ static void fetch_xyz_rgb_st( struct spu_vs_context *draw, /** * Fetch vertex attributes for 'count' vertices. */ -static void generic_vertex_fetch( struct spu_vs_context *draw, - struct spu_exec_machine *machine, - const unsigned *elts, - unsigned count ) +static void generic_vertex_fetch(struct spu_vs_context *draw, + struct spu_exec_machine *machine, + const unsigned *elts, + unsigned count) { unsigned nr_attrs = draw->vertex_fetch.nr_attrs; unsigned attr; assert(count <= 4); + wait_on_mask(1 << TAG_VERTEX_BUFFER); + // _mesa_printf("%s %d\n", __FUNCTION__, count); /* loop over vertex attributes (vertex shader inputs) @@ -441,13 +445,23 @@ static void generic_vertex_fetch( struct spu_vs_context *draw, * a prototype for an sse implementation, which would have * difficulties doing that. */ - for (i = 0; i < count; i++) - fetch( src + elts[i] * pitch, p[i] ); + for (i = 0; i < count; i++) { + uint8_t buffer[32 + (sizeof(float) * 4)] ALIGN16_ATTRIB; + const unsigned long addr = src + elts[i] * pitch; + const unsigned size = (sizeof(float) * 4) + (addr & 0x0f); + + mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); + wait_on_mask(1 << TAG_VERTEX_BUFFER); + + memcpy(& buffer, buffer + (addr & 0x0f), sizeof(float) * 4); + + fetch(buffer, p[i]); + } /* Be nice and zero out any missing vertices: */ for (/* empty */; i < 4; i++) - p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; + p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; /* Transpose/swizzle into sse-friendly format. Currently * assuming that all vertex shader inputs are float[4], but this @@ -475,6 +489,9 @@ void spu_update_vertex_fetch( struct spu_vs_context *draw ) draw->vertex_fetch.fetch_func = generic_vertex_fetch; + /* Disable the fast path because they don't use mfc_get yet. + */ +#if 0 switch (draw->vertex_fetch.nr_attrs) { case 2: if (draw->vertex_fetch.format[0] == PIPE_FORMAT_R32G32B32_FLOAT && @@ -490,4 +507,5 @@ void spu_update_vertex_fetch( struct spu_vs_context *draw ) default: break; } +#endif } diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.c b/src/mesa/pipe/cell/spu/spu_vertex_shader.c index e694ff729f..595f54b0eb 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.c @@ -32,6 +32,8 @@ * Ian Romanick */ +#include + #include "pipe/p_util.h" #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" @@ -40,9 +42,7 @@ #include "pipe/draw/draw_private.h" #include "pipe/draw/draw_context.h" #include "pipe/cell/common.h" - -#define DBG_VS 0 - +#include "spu_main.h" static INLINE unsigned compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) @@ -110,6 +110,12 @@ run_vertex_program(struct spu_vs_context *draw, for (j = 0; j < count; j++) { unsigned slot; float x, y, z, w; + unsigned char buffer[sizeof(struct vertex_header) + + MAX_VERTEX_SIZE] ALIGN16_ATTRIB; + struct vertex_header *const tmpOut = + (struct vertex_header *) buffer; + const unsigned vert_size = sizeof(struct vertex_header) + + (sizeof(float) * 4 * draw->num_vs_outputs); /* Handle attr[0] (position) specially: * @@ -117,14 +123,14 @@ run_vertex_program(struct spu_vs_context *draw, * program as a set of DP4 instructions appended to the * user-provided code. */ - x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + x = tmpOut->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = tmpOut->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = tmpOut->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = tmpOut->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, + tmpOut->clipmask = compute_clipmask(tmpOut->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; + tmpOut->edgeflag = 1; /* divide by w */ w = 1.0f / w; @@ -133,35 +139,27 @@ run_vertex_program(struct spu_vs_context *draw, z *= w; /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; - -#if DBG_VS - printf("output[%d]win: %f %f %f %f\n", j, - vOut[j]->data[0][0], - vOut[j]->data[0][1], - vOut[j]->data[0][2], - vOut[j]->data[0][3]); -#endif + tmpOut->data[0][0] = x * scale[0] + trans[0]; + tmpOut->data[0][1] = y * scale[1] + trans[1]; + tmpOut->data[0][2] = z * scale[2] + trans[2]; + tmpOut->data[0][3] = w; + /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. */ for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; -#if DBG_VS - printf("output[%d][%d]: %f %f %f %f\n", j, slot, - vOut[j]->data[slot][0], - vOut[j]->data[slot][1], - vOut[j]->data[slot][2], - vOut[j]->data[slot][3]); -#endif + tmpOut->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + tmpOut->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + tmpOut->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + tmpOut->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; } + + wait_on_mask(1 << TAG_VERTEX_BUFFER); + mfc_put(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0); + } /* loop over vertices */ + + wait_on_mask(1 << TAG_VERTEX_BUFFER); } diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index e8ca1f035b..711bcd02f6 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -71,10 +71,11 @@ struct draw_context *draw_create( void ) */ { uint i; - char *tmp = (char*) MALLOC( Elements(draw->vcache.vertex) * MAX_VERTEX_SIZE ); + const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; + char *tmp = align_malloc(Elements(draw->vcache.vertex) * size, 16); for (i = 0; i < Elements(draw->vcache.vertex); i++) - draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * MAX_VERTEX_SIZE); + draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * size); } draw->convert_wide_points = TRUE; -- cgit v1.2.3 From 3d13605ee5fc92a1e3d82f1dbbcb8342066d8af0 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 29 Jan 2008 11:43:04 -0800 Subject: Provide mechanism to hook in custom vertex shader cache flush function --- src/mesa/pipe/draw/draw_context.c | 2 ++ src/mesa/pipe/draw/draw_prim.c | 2 +- src/mesa/pipe/draw/draw_private.h | 5 +++++ 3 files changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 711bcd02f6..87f4969983 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -78,6 +78,8 @@ struct draw_context *draw_create( void ) draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * size); } + draw->shader_queue_flush = draw_vertex_shader_queue_flush; + draw->convert_wide_points = TRUE; draw->convert_wide_lines = TRUE; diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 41b3fddcc1..58400213d7 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -127,7 +127,7 @@ void draw_do_flush( struct draw_context *draw, unsigned flags ) if (flags >= DRAW_FLUSH_SHADER_QUEUE) { if (draw->vs.queue_nr) - draw_vertex_shader_queue_flush(draw); + (*draw->shader_queue_flush)(draw); if (flags >= DRAW_FLUSH_PRIM_QUEUE) { if (draw->pq.queue_nr) diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 21de400676..fea6d94ed8 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -240,6 +240,11 @@ struct draw_context unsigned queue_nr; } vs; + /** + * Run the vertex shader on all vertices in the vertex queue. + */ + void (*shader_queue_flush)(struct draw_context *draw); + /* Prim pipeline queue: */ struct { -- cgit v1.2.3 From fcf944177325cdf8bf6e4f1b70296c19476e2375 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 10:43:23 -0800 Subject: Pass ptr to local memory copy instead of main memory to exec_instruction This was essentially a cut-and-paste bug when the instruction fetcher was added. Also, the test for TGSI_PROCESSOR_FRAGMENT was moved outside the loop for exec_declaration. --- src/mesa/pipe/cell/spu/spu_exec.c | 22 ++++++++++++---------- 1 file changed, 12 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c index f43278198e..b3db6716d5 100644 --- a/src/mesa/pipe/cell/spu/spu_exec.c +++ b/src/mesa/pipe/cell/spu/spu_exec.c @@ -2332,17 +2332,19 @@ spu_exec_machine_run( struct spu_exec_machine *mach ) /* execute declarations (interpolants) */ - for (i = 0; i < mach->NumDeclarations; i++) { - uint8_t buffer[sizeof(struct tgsi_full_declaration) + 32] ALIGN16_ATTRIB; - struct tgsi_full_declaration decl; - unsigned long decl_addr = (unsigned long) (mach->Declarations+i); - unsigned size = ((sizeof(decl) + (decl_addr & 0x0f) + 0x0f) & ~0x0f); + if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { + for (i = 0; i < mach->NumDeclarations; i++) { + uint8_t buffer[sizeof(struct tgsi_full_declaration) + 32] ALIGN16_ATTRIB; + struct tgsi_full_declaration decl; + unsigned long decl_addr = (unsigned long) (mach->Declarations+i); + unsigned size = ((sizeof(decl) + (decl_addr & 0x0f) + 0x0f) & ~0x0f); - mfc_get(buffer, decl_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); - wait_on_mask(1 << TAG_INSTRUCTION_FETCH); + mfc_get(buffer, decl_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); + wait_on_mask(1 << TAG_INSTRUCTION_FETCH); - memcpy(& decl, buffer + (decl_addr & 0x0f), sizeof(decl)); - exec_declaration( mach, mach->Declarations+i ); + memcpy(& decl, buffer + (decl_addr & 0x0f), sizeof(decl)); + exec_declaration( mach, decl ); + } } /* execute instructions, until pc is set to -1 */ @@ -2357,7 +2359,7 @@ spu_exec_machine_run( struct spu_exec_machine *mach ) wait_on_mask(1 << TAG_INSTRUCTION_FETCH); memcpy(& inst, buffer + (inst_addr & 0x0f), sizeof(inst)); - exec_instruction( mach, mach->Instructions + pc, &pc ); + exec_instruction( mach, & inst, &pc ); } #if 0 -- cgit v1.2.3 From a0a707342a353024271f09cd52bd955d8df310a8 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 10:46:55 -0800 Subject: Missing amperstand in previous commit. Oops. --- src/mesa/pipe/cell/spu/spu_exec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c index b3db6716d5..85b5815cad 100644 --- a/src/mesa/pipe/cell/spu/spu_exec.c +++ b/src/mesa/pipe/cell/spu/spu_exec.c @@ -2343,7 +2343,7 @@ spu_exec_machine_run( struct spu_exec_machine *mach ) wait_on_mask(1 << TAG_INSTRUCTION_FETCH); memcpy(& decl, buffer + (decl_addr & 0x0f), sizeof(decl)); - exec_declaration( mach, decl ); + exec_declaration( mach, &decl ); } } -- cgit v1.2.3 From 708d699e0cebb2dfbca7b6639ee5b177dc8c4c61 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 12:59:09 -0800 Subject: Fetch uniforms from main memory. --- src/mesa/pipe/cell/spu/spu_exec.c | 21 ++++++++++++++++----- 1 file changed, 16 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c index 85b5815cad..78f7d0962f 100644 --- a/src/mesa/pipe/cell/spu/spu_exec.c +++ b/src/mesa/pipe/cell/spu/spu_exec.c @@ -791,12 +791,23 @@ fetch_src_file_channel( case TGSI_EXTSWIZZLE_Z: case TGSI_EXTSWIZZLE_W: switch( file ) { - case TGSI_FILE_CONSTANT: - chan->f[0] = mach->Consts[index->i[0]][swizzle]; - chan->f[1] = mach->Consts[index->i[1]][swizzle]; - chan->f[2] = mach->Consts[index->i[2]][swizzle]; - chan->f[3] = mach->Consts[index->i[3]][swizzle]; + case TGSI_FILE_CONSTANT: { + unsigned char buffer[32] ALIGN16_ATTRIB; + unsigned i; + + for (i = 0; i < 4; i++) { + const float *ptr = mach->Consts[index->i[i]]; + const uint64_t addr = (uint64_t)(uintptr_t) ptr; + const unsigned size = ((addr & 0x0f) == 0) ? 16 : 32; + + mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); + wait_on_mask(1 << TAG_VERTEX_BUFFER); + + (void) memcpy(& chan->f[i], &buffer[(addr & 0x0f) + + (sizeof(float) * swizzle)], sizeof(float)); + } break; + } case TGSI_FILE_INPUT: chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; -- cgit v1.2.3 From 7b27d9fd660c122fb2ec50007129d67e78814587 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 17:26:22 -0800 Subject: Fix size calculation in attribute fetch. --- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index 0192227d57..1e846868e3 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -446,14 +446,14 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, * difficulties doing that. */ for (i = 0; i < count; i++) { - uint8_t buffer[32 + (sizeof(float) * 4)] ALIGN16_ATTRIB; - const unsigned long addr = src + elts[i] * pitch; - const unsigned size = (sizeof(float) * 4) + (addr & 0x0f); + uint8_t buffer[32] ALIGN16_ATTRIB; + const unsigned long addr = src + (elts[i] * pitch); + const unsigned size = ((addr & 0x0f) == 0) ? 16 : 32; mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); wait_on_mask(1 << TAG_VERTEX_BUFFER); - memcpy(& buffer, buffer + (addr & 0x0f), sizeof(float) * 4); + memmove(& buffer, buffer + (addr & 0x0f), 16); fetch(buffer, p[i]); } -- cgit v1.2.3 From 334986114665df650649634b63184be6f1b9cd9b Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 17:28:48 -0800 Subject: Implement micro_pow and micro_sqrt Unimplemented micro ops get assertions for now. --- src/mesa/pipe/cell/spu/spu_exec.c | 43 ++++++++++++++++++++++++++++----------- 1 file changed, 31 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c index 78f7d0962f..168bada3bb 100644 --- a/src/mesa/pipe/cell/spu/spu_exec.c +++ b/src/mesa/pipe/cell/spu/spu_exec.c @@ -52,6 +52,8 @@ #include #include +#include +#include #include "pipe/p_compiler.h" #include "pipe/p_state.h" @@ -207,6 +209,7 @@ micro_ceil( union spu_exec_channel *dst, const union spu_exec_channel *src ) { + ASSERT(0); #if 0 dst->f[0] = (float) ceil( (double) src->f[0] ); dst->f[1] = (float) ceil( (double) src->f[1] ); @@ -220,6 +223,7 @@ micro_cos( union spu_exec_channel *dst, const union spu_exec_channel *src ) { + ASSERT(0); #if 0 dst->f[0] = (float) cos( (double) src->f[0] ); dst->f[1] = (float) cos( (double) src->f[1] ); @@ -307,6 +311,7 @@ micro_exp2( union spu_exec_channel *dst, const union spu_exec_channel *src) { + ASSERT(0); #if 0 dst->f[0] = (float) pow( 2.0, (double) src->f[0] ); dst->f[1] = (float) pow( 2.0, (double) src->f[1] ); @@ -342,6 +347,7 @@ micro_flr( union spu_exec_channel *dst, const union spu_exec_channel *src ) { + ASSERT(0); #if 0 dst->f[0] = (float) floor( (double) src->f[0] ); dst->f[1] = (float) floor( (double) src->f[1] ); @@ -355,6 +361,7 @@ micro_frc( union spu_exec_channel *dst, const union spu_exec_channel *src ) { + ASSERT(0); #if 0 dst->f[0] = src->f[0] - (float) floor( (double) src->f[0] ); dst->f[1] = src->f[1] - (float) floor( (double) src->f[1] ); @@ -393,6 +400,7 @@ micro_lg2( union spu_exec_channel *dst, const union spu_exec_channel *src ) { + ASSERT(0); #if 0 dst->f[0] = (float) log( (double) src->f[0] ) * 1.442695f; dst->f[1] = (float) log( (double) src->f[1] ) * 1.442695f; @@ -649,12 +657,18 @@ micro_pow( const union spu_exec_channel *src0, const union spu_exec_channel *src1 ) { -#if 0 - dst->f[0] = (float) pow( (double) src0->f[0], (double) src1->f[0] ); - dst->f[1] = (float) pow( (double) src0->f[1], (double) src1->f[1] ); - dst->f[2] = (float) pow( (double) src0->f[2], (double) src1->f[2] ); - dst->f[3] = (float) pow( (double) src0->f[3], (double) src1->f[3] ); -#endif + vec_float4 s0 = (vec_float4) { + src0->f[0], src0->f[1], src0->f[2], src0->f[3] + }; + vec_float4 s1 = (vec_float4) { + src1->f[0], src1->f[1], src1->f[2], src1->f[3] + }; + vec_float4 d = _powf4(s0, s1); + + dst->f[0] = spu_extract(d, 0); + dst->f[1] = spu_extract(d, 1); + dst->f[2] = spu_extract(d, 2); + dst->f[3] = spu_extract(d, 3); } static void @@ -662,6 +676,7 @@ micro_rnd( union spu_exec_channel *dst, const union spu_exec_channel *src ) { + ASSERT(0); #if 0 dst->f[0] = (float) floor( (double) (src->f[0] + 0.5f) ); dst->f[1] = (float) floor( (double) (src->f[1] + 0.5f) ); @@ -722,6 +737,7 @@ micro_sin( union spu_exec_channel *dst, const union spu_exec_channel *src ) { + ASSERT(0); #if 0 dst->f[0] = (float) sin( (double) src->f[0] ); dst->f[1] = (float) sin( (double) src->f[1] ); @@ -734,12 +750,15 @@ static void micro_sqrt( union spu_exec_channel *dst, const union spu_exec_channel *src ) { -#if 0 - dst->f[0] = (float) sqrt( (double) src->f[0] ); - dst->f[1] = (float) sqrt( (double) src->f[1] ); - dst->f[2] = (float) sqrt( (double) src->f[2] ); - dst->f[3] = (float) sqrt( (double) src->f[3] ); -#endif + vec_float4 s = (vec_float4) { + src->f[0], src->f[1], src->f[2], src->f[3] + }; + vec_float4 d = _sqrtf4(s); + + dst->f[0] = spu_extract(d, 0); + dst->f[1] = spu_extract(d, 1); + dst->f[2] = spu_extract(d, 2); + dst->f[3] = spu_extract(d, 3); } static void -- cgit v1.2.3 From 137cb72284a115d8f5ffadf2154b6f5eb5323a7d Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 19:24:40 -0800 Subject: Elts are always ints, pass vOut pointers in-line in command --- src/mesa/pipe/cell/common.h | 6 +++--- src/mesa/pipe/cell/spu/spu_vertex_shader.c | 14 ++------------ 2 files changed, 5 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 80a1425ec7..fbbdf728a1 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -144,13 +144,13 @@ struct cell_shader_info } ALIGN16_ATTRIB; +#define SPU_VERTS_PER_BATCH 64 struct cell_command_vs { struct cell_shader_info shader; - void *elts; unsigned num_elts; - unsigned bytes_per_elt; - void *vOut; + unsigned elts[SPU_VERTS_PER_BATCH]; + uint64_t vOut[SPU_VERTS_PER_BATCH]; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.c b/src/mesa/pipe/cell/spu/spu_vertex_shader.c index 595f54b0eb..82165501c5 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.c @@ -81,7 +81,7 @@ compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) static void run_vertex_program(struct spu_vs_context *draw, unsigned elts[4], unsigned count, - struct vertex_header *vOut[]) + const uint64_t *vOut) { struct spu_exec_machine *machine = &draw->machine; unsigned int j; @@ -206,17 +206,7 @@ spu_execute_vertex_shader(struct spu_vs_context *draw, for (i = 0; i < vs->num_elts; i += 4) { const unsigned batch_size = MIN2(vs->num_elts - i, 4); - unsigned elts[4]; - - for (j = 0; j < batch_size; j++) { - switch (vs->bytes_per_elt) { - case 1: elts[j] = ((unsigned char *) vs->elts)[i + j]; break; - case 2: elts[j] = ((unsigned short *)vs->elts)[i + j]; break; - case 4: elts[j] = ((unsigned int *) vs->elts)[i + j]; break; - } - } - run_vertex_program(draw, elts, batch_size, - (struct vertex_header (*)[]) vs->vOut); + run_vertex_program(draw, & vs->elts[i], batch_size, &vs->vOut[i]); } } -- cgit v1.2.3 From fb348c2cb16d0bc216d29889474972d5c14d0980 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 19:25:47 -0800 Subject: Set machine->Processor The default value is 0, which is TGSI_PROCESSOR_FRAGMENT...not correct for a vertex shader! --- src/mesa/pipe/cell/spu/spu_vertex_shader.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.c b/src/mesa/pipe/cell/spu/spu_vertex_shader.c index 82165501c5..125b2c3a43 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.c @@ -93,7 +93,8 @@ run_vertex_program(struct spu_vs_context *draw, assert(count <= 4); - /* Consts does not require 16 byte alignment. */ + machine->Processor = TGSI_PROCESSOR_VERTEX; + ASSERT_ALIGN16(draw->constants); machine->Consts = (float (*)[4]) draw->constants; -- cgit v1.2.3 From 193491cbd3ad2ad95243181c201da4640f3a29c2 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 19:30:15 -0800 Subject: Handle CELL_CMD_VS_EXECUTE *only* outside batch commands. --- src/mesa/pipe/cell/common.h | 3 ++- src/mesa/pipe/cell/spu/spu_main.c | 8 +++----- 2 files changed, 5 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index fbbdf728a1..a40cfb8210 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -133,7 +133,6 @@ struct cell_array_info struct cell_shader_info { - unsigned processor; unsigned num_outputs; void *declarations; @@ -147,6 +146,7 @@ struct cell_shader_info #define SPU_VERTS_PER_BATCH 64 struct cell_command_vs { + uint opcode; /**< CELL_CMD_VS_EXECUTE */ struct cell_shader_info shader; unsigned num_elts; unsigned elts[SPU_VERTS_PER_BATCH]; @@ -190,6 +190,7 @@ struct cell_command struct cell_command_framebuffer fb; struct cell_command_clear_surface clear; struct cell_command_render render; + struct cell_command_vs vs; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 9daa3ec735..7105c0f897 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -397,11 +397,6 @@ cmd_batch(uint opcode) cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); pos += (1 + sizeof(struct cell_array_info) / 4); break; - case CELL_CMD_VS_EXECUTE: - spu_execute_vertex_shader(&draw, - (struct cell_command_vs *) &buffer[pos+1]); - pos += (1 + sizeof(struct cell_command_vs) / 4); - break; default: printf("SPU %u: bad opcode: 0x%x\n", spu.init.id, buffer[pos]); ASSERT(0); @@ -470,6 +465,9 @@ main_loop(void) assert(pos_incr == 0); } break; + case CELL_CMD_VS_EXECUTE: + spu_execute_vertex_shader(&draw, &cmd.vs); + break; case CELL_CMD_BATCH: cmd_batch(opcode); break; -- cgit v1.2.3 From 10270fbe2d362fe8f27384b9a5423381e2882460 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 19:33:30 -0800 Subject: Correctly read / write vertex header from / to main memory --- src/mesa/pipe/cell/spu/spu_vertex_shader.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.c b/src/mesa/pipe/cell/spu/spu_vertex_shader.c index 125b2c3a43..ea5ffae6bc 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.c @@ -112,11 +112,16 @@ run_vertex_program(struct spu_vs_context *draw, unsigned slot; float x, y, z, w; unsigned char buffer[sizeof(struct vertex_header) - + MAX_VERTEX_SIZE] ALIGN16_ATTRIB; + + MAX_VERTEX_SIZE] ALIGN16_ATTRIB; struct vertex_header *const tmpOut = - (struct vertex_header *) buffer; - const unsigned vert_size = sizeof(struct vertex_header) - + (sizeof(float) * 4 * draw->num_vs_outputs); + (struct vertex_header *) buffer; + const unsigned vert_size = ROUNDUP16(sizeof(struct vertex_header) + + (sizeof(float) * 4 + * draw->num_vs_outputs)); + + mfc_get(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0); + wait_on_mask(1 << TAG_VERTEX_BUFFER); + /* Handle attr[0] (position) specially: * @@ -155,12 +160,8 @@ run_vertex_program(struct spu_vs_context *draw, tmpOut->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; } - wait_on_mask(1 << TAG_VERTEX_BUFFER); mfc_put(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0); - } /* loop over vertices */ - - wait_on_mask(1 << TAG_VERTEX_BUFFER); } -- cgit v1.2.3 From de949a471ed66f0e6db0819bf55b2ec74b7e4048 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 19:34:22 -0800 Subject: cell_array_info should not be 16-byte aligned Forcing cell_array_info to be 16-byte aligned makes it more difficult to stuff that state in batch commands. --- src/mesa/pipe/cell/common.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index a40cfb8210..533ad2cf6e 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -128,7 +128,7 @@ struct cell_array_info uint attr; /**< Attribute that this state if for. */ uint pitch; /**< Byte pitch from one entry to the next. */ enum pipe_format format; /**< Pipe format of each entry. */ -} ALIGN16_ATTRIB; +}; struct cell_shader_info -- cgit v1.2.3 From 9ad986b88763f6baefa73830dcd5762156ab9b20 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 19:40:24 -0800 Subject: Numerous small fixed to PPU-SPU vertex shader protocol --- src/mesa/pipe/cell/common.h | 19 ++++++++++++------- src/mesa/pipe/cell/spu/spu_vertex_shader.c | 27 ++++++++++++++++++++++----- 2 files changed, 34 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 533ad2cf6e..28b0c59a0a 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -124,10 +124,10 @@ struct cell_command_clear_surface */ struct cell_array_info { - void *base; /**< Base address of the 0th element. */ - uint attr; /**< Attribute that this state if for. */ - uint pitch; /**< Byte pitch from one entry to the next. */ - enum pipe_format format; /**< Pipe format of each entry. */ + uint64_t base; /**< Base address of the 0th element. */ + uint attr; /**< Attribute that this state if for. */ + uint pitch; /**< Byte pitch from one entry to the next. */ + uint format; /**< Pipe format of each entry. */ }; @@ -135,11 +135,13 @@ struct cell_shader_info { unsigned num_outputs; - void *declarations; + uint64_t declarations; unsigned num_declarations; - void *instructions; + uint64_t instructions; unsigned num_instructions; - void *uniforms; + uint64_t uniforms; + uint64_t immediates; + unsigned num_immediates; } ALIGN16_ATTRIB; @@ -151,6 +153,9 @@ struct cell_command_vs unsigned num_elts; unsigned elts[SPU_VERTS_PER_BATCH]; uint64_t vOut[SPU_VERTS_PER_BATCH]; + float plane[12][4]; + unsigned nr_planes; + unsigned nr_attrs; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.c b/src/mesa/pipe/cell/spu/spu_vertex_shader.c index ea5ffae6bc..c1cbbb6d1e 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.c @@ -187,12 +187,22 @@ spu_bind_vertex_shader(struct spu_vs_context *draw, } +unsigned char immediates[(sizeof(float) * 4 * TGSI_EXEC_NUM_IMMEDIATES) + 32] + ALIGN16_ATTRIB; + void spu_execute_vertex_shader(struct spu_vs_context *draw, - const struct cell_command_vs *vs) + const struct cell_command_vs *vs) { unsigned i; - unsigned j; + + const uint64_t immediate_addr = vs->shader.immediates; + const unsigned immediate_size = + ROUNDUP16((sizeof(float) * 4 * vs->shader.num_immediates) + + (immediate_addr & 0x0f)); + + mfc_get(immediates, immediate_addr & ~0x0f, immediate_size, + TAG_VERTEX_BUFFER, 0, 0); draw->machine.Instructions = (struct tgsi_full_instruction *) vs->shader.instructions; @@ -202,10 +212,17 @@ spu_execute_vertex_shader(struct spu_vs_context *draw, vs->shader.declarations; draw->machine.NumDeclarations = vs->shader.num_declarations; + draw->vertex_fetch.nr_attrs = vs->nr_attrs; + + wait_on_mask(1 << TAG_VERTEX_BUFFER); + + (void) memcpy(& draw->machine.Imms, &immediates[immediate_addr & 0x0f], + sizeof(float) * 4 * vs->shader.num_immediates); + spu_bind_vertex_shader(draw, vs->shader.uniforms, - NULL, -1, - vs->shader.num_outputs); - + vs->plane, vs->nr_planes, + vs->shader.num_outputs); + for (i = 0; i < vs->num_elts; i += 4) { const unsigned batch_size = MIN2(vs->num_elts - i, 4); -- cgit v1.2.3 From a89ee8a96db1ac7674a5ae82e518ce5c380d9195 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 20:10:45 -0800 Subject: Add driver_private field for drivers that hook shader_queue_flush. --- src/mesa/pipe/draw/draw_private.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index fea6d94ed8..7782db0477 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -259,6 +259,8 @@ struct draw_context #ifdef MESA_LLVM struct gallivm_cpu_engine *engine; #endif + + void *driver_private; }; -- cgit v1.2.3 From 5a6fd9393021b9476273b7831bcda2186c9324a1 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 20:12:00 -0800 Subject: Use SPUs for vertex shader processing --- src/mesa/pipe/cell/ppu/Makefile | 1 + src/mesa/pipe/cell/ppu/cell_context.c | 12 ++- src/mesa/pipe/cell/ppu/cell_context.h | 2 + src/mesa/pipe/cell/ppu/cell_vertex_shader.c | 118 ++++++++++++++++++++++++++++ 4 files changed, 132 insertions(+), 1 deletion(-) create mode 100644 src/mesa/pipe/cell/ppu/cell_vertex_shader.c (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile index e7f2562da7..50060f5cd3 100644 --- a/src/mesa/pipe/cell/ppu/Makefile +++ b/src/mesa/pipe/cell/ppu/Makefile @@ -34,6 +34,7 @@ SOURCES = \ cell_surface.c \ cell_texture.c \ cell_vbuf.c \ + cell_vertex_shader.c \ cell_winsys.c diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index e8020a49bc..4885cd0d2c 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -39,6 +39,7 @@ #include "pipe/p_winsys.h" #include "pipe/cell/common.h" #include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_private.h" #include "cell_clear.h" #include "cell_context.h" #include "cell_draw_arrays.h" @@ -156,6 +157,15 @@ cell_destroy_context( struct pipe_context *pipe ) } +static struct draw_context * +cell_draw_create(struct cell_context *cell) +{ + struct draw_context *draw = draw_create(); + + draw->shader_queue_flush = cell_vertex_shader_queue_flush; + draw->driver_private = cell; + return draw; +} struct pipe_context * @@ -242,7 +252,7 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell_init_surface_functions(cell); - cell->draw = draw_create(); + cell->draw = cell_draw_create(cell); cell_init_vbuf(cell); draw_set_rasterize_stage(cell->draw, cell->vbuf); diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h index 65b89518ad..3b63419b5e 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ b/src/mesa/pipe/cell/ppu/cell_context.h @@ -126,6 +126,8 @@ cell_context(struct pipe_context *pipe) extern struct pipe_context * cell_create_context(struct pipe_winsys *ws, struct cell_winsys *cws); +extern void +cell_vertex_shader_queue_flush(struct draw_context *draw); diff --git a/src/mesa/pipe/cell/ppu/cell_vertex_shader.c b/src/mesa/pipe/cell/ppu/cell_vertex_shader.c new file mode 100644 index 0000000000..aef329a902 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_vertex_shader.c @@ -0,0 +1,118 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file cell_vertex_shader.c + * Vertex shader interface routines for Cell. + * + * \author Ian Romanick + */ + +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +#include "cell_context.h" +#include "cell_draw_arrays.h" +#include "cell_spu.h" +#include "cell_batch.h" + +#include "pipe/cell/common.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_private.h" + +/** + * Run the vertex shader on all vertices in the vertex queue. + * Called by the draw module when the vertx cache needs to be flushed. + */ +void +cell_vertex_shader_queue_flush(struct draw_context *draw) +{ + struct cell_context *const cell = + (struct cell_context *) draw->driver_private; + struct cell_command_vs *const vs = &cell_global.command[0].vs; + unsigned *batch; + struct cell_array_info array_info; + unsigned i, j; + + assert(draw->vs.queue_nr != 0); + + /* XXX: do this on statechange: + */ + draw_update_vertex_fetch(draw); + + for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { + array_info.opcode = CELL_CMD_STATE_VS_ARRAY_INFO; + assert(draw->vertex_fetch.src_ptr[i] != NULL); + array_info.base = (uintptr_t) draw->vertex_fetch.src_ptr[i]; + array_info.attr = i; + array_info.pitch = draw->vertex_fetch.pitch[i]; + array_info.format = draw->vertex_element[i].src_format; + + cell_batch_append(cell, & array_info, sizeof(array_info)); + } + + batch = cell_batch_alloc(cell, sizeof(unsigned) + + sizeof(struct pipe_viewport_state)); + batch[0] = CELL_CMD_STATE_VIEWPORT; + (void) memcpy(&batch[1], &draw->viewport, + sizeof(struct pipe_viewport_state)); + + cell_batch_flush(cell); + + vs->opcode = CELL_CMD_VS_EXECUTE; + vs->shader.num_outputs = draw->num_vs_outputs; + vs->shader.declarations = (uintptr_t) draw->machine.Declarations; + vs->shader.num_declarations = draw->machine.NumDeclarations; + vs->shader.instructions = (uintptr_t) draw->machine.Instructions; + vs->shader.num_instructions = draw->machine.NumInstructions; + vs->shader.uniforms = (uintptr_t) draw->user.constants; + vs->shader.immediates = (uintptr_t) draw->machine.Imms; + vs->shader.num_immediates = draw->machine.ImmLimit / 4; + vs->nr_attrs = draw->vertex_fetch.nr_attrs; + + (void) memcpy(vs->plane, draw->plane, sizeof(draw->plane)); + vs->nr_planes = draw->nr_planes; + + for (i = 0; i < draw->vs.queue_nr; i += SPU_VERTS_PER_BATCH) { + const unsigned n = MIN2(SPU_VERTS_PER_BATCH, draw->vs.queue_nr - i); + + for (j = 0; j < n; j++) { + vs->elts[j] = draw->vs.queue[i + j].elt; + vs->vOut[j] = (uintptr_t) draw->vs.queue[i + j].dest; + } + + for (/* empty */; j < SPU_VERTS_PER_BATCH; j++) { + vs->elts[j] = vs->elts[0]; + vs->vOut[j] = vs->vOut[0]; + } + + vs->num_elts = n; + send_mbox_message(cell_global.spe_contexts[0], CELL_CMD_VS_EXECUTE); + + cell_flush_int(& cell->pipe, PIPE_FLUSH_WAIT); + } + + draw->vs.queue_nr = 0; +} -- cgit v1.2.3 From 62d11b98c4a4904b56fab153407f49619d6d331d Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 30 Jan 2008 20:14:14 -0800 Subject: I don't know why using uint64_t for "base" doesn't work. Ugh. --- src/mesa/pipe/cell/common.h | 5 +++-- src/mesa/pipe/cell/spu/spu_main.c | 4 ++-- 2 files changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 28b0c59a0a..05aeed83ab 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -124,11 +124,12 @@ struct cell_command_clear_surface */ struct cell_array_info { - uint64_t base; /**< Base address of the 0th element. */ + uint opcode; + uint base; /**< Base address of the 0th element. */ uint attr; /**< Attribute that this state if for. */ uint pitch; /**< Byte pitch from one entry to the next. */ uint format; /**< Pipe format of each entry. */ -}; +} ALIGN16_ATTRIB; struct cell_shader_info diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 7105c0f897..d6393048f5 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -394,8 +394,8 @@ cmd_batch(uint opcode) pos += (1 + sizeof(struct pipe_viewport_state) / 4); break; case CELL_CMD_STATE_VS_ARRAY_INFO: - cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); - pos += (1 + sizeof(struct cell_array_info) / 4); + cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos]); + pos += (sizeof(struct cell_array_info) / 4); break; default: printf("SPU %u: bad opcode: 0x%x\n", spu.init.id, buffer[pos]); -- cgit v1.2.3 From c42e6254cffb8ef480868e9c1942f73129fc4f80 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 31 Jan 2008 13:14:35 +0900 Subject: gallium: Add SCons as alternative build system for Gallium. --- src/mesa/SConscript | 435 +++++++++++++++++++++++++++ src/mesa/drivers/dri/SConscript | 48 +++ src/mesa/drivers/dri/intel_winsys/SConscript | 41 +++ src/mesa/pipe/SConscript | 9 + src/mesa/pipe/i915simple/SConscript | 29 ++ src/mesa/pipe/i965simple/SConscript | 55 ++++ src/mesa/pipe/softpipe/SConscript | 42 +++ 7 files changed, 659 insertions(+) create mode 100644 src/mesa/SConscript create mode 100644 src/mesa/drivers/dri/SConscript create mode 100644 src/mesa/drivers/dri/intel_winsys/SConscript create mode 100644 src/mesa/pipe/SConscript create mode 100644 src/mesa/pipe/i915simple/SConscript create mode 100644 src/mesa/pipe/i965simple/SConscript create mode 100644 src/mesa/pipe/softpipe/SConscript (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript new file mode 100644 index 0000000000..70a98f3129 --- /dev/null +++ b/src/mesa/SConscript @@ -0,0 +1,435 @@ +####################################################################### +# SConscript for mesa +# +# TODO: Split this into per-module SConscripts + + +Import('*') + + +####################################################################### +# Core sources + +MAIN_SOURCES = [ + 'main/api_arrayelt.c', + 'main/api_loopback.c', + 'main/api_noop.c', + 'main/api_validate.c', + 'main/accum.c', + 'main/attrib.c', + 'main/arrayobj.c', + 'main/blend.c', + 'main/bufferobj.c', + 'main/buffers.c', + 'main/clip.c', + 'main/colortab.c', + 'main/context.c', + 'main/convolve.c', + 'main/debug.c', + 'main/depth.c', + 'main/depthstencil.c', + 'main/dlist.c', + 'main/drawpix.c', + 'main/enable.c', + 'main/enums.c', + 'main/eval.c', + 'main/execmem.c', + 'main/extensions.c', + 'main/fbobject.c', + 'main/feedback.c', + 'main/ffvertex_prog.c', + 'main/fog.c', + 'main/framebuffer.c', + 'main/get.c', + 'main/getstring.c', + 'main/hash.c', + 'main/hint.c', + 'main/histogram.c', + 'main/image.c', + 'main/imports.c', + 'main/light.c', + 'main/lines.c', + 'main/matrix.c', + 'main/mipmap.c', + 'main/mm.c', + 'main/pixel.c', + 'main/points.c', + 'main/polygon.c', + 'main/queryobj.c', + 'main/rastpos.c', + 'main/rbadaptors.c', + 'main/renderbuffer.c', + 'main/shaders.c', + 'main/state.c', + 'main/stencil.c', + 'main/texcompress.c', + 'main/texcompress_s3tc.c', + 'main/texcompress_fxt1.c', + 'main/texenvprogram.c', + 'main/texformat.c', + 'main/teximage.c', + 'main/texobj.c', + 'main/texrender.c', + 'main/texstate.c', + 'main/texstore.c', + 'main/varray.c', + 'main/vtxfmt.c', +] + +GLAPI_SOURCES = [ + 'main/dispatch.c', + 'glapi/glapi.c', + 'glapi/glthread.c', +] + +MATH_SOURCES = [ + 'math/m_debug_clip.c', + 'math/m_debug_norm.c', + 'math/m_debug_xform.c', + 'math/m_eval.c', + 'math/m_matrix.c', + 'math/m_translate.c', + 'math/m_vector.c', + 'math/m_xform.c', +] + +VBO_SOURCES = [ + 'vbo/vbo_context.c', + 'vbo/vbo_exec.c', + 'vbo/vbo_exec_api.c', + 'vbo/vbo_exec_array.c', + 'vbo/vbo_exec_draw.c', + 'vbo/vbo_exec_eval.c', + 'vbo/vbo_rebase.c', + 'vbo/vbo_split.c', + 'vbo/vbo_split_copy.c', + 'vbo/vbo_split_inplace.c', + 'vbo/vbo_save.c', + 'vbo/vbo_save_api.c', + 'vbo/vbo_save_draw.c', + 'vbo/vbo_save_loopback.c', +] + +VF_SOURCES = [ + 'vf/vf.c', + 'vf/vf_generic.c', + 'vf/vf_sse.c', +] + +DRAW_SOURCES = [ + 'pipe/draw/draw_clip.c', + 'pipe/draw/draw_context.c', + 'pipe/draw/draw_cull.c', + 'pipe/draw/draw_debug.c', + 'pipe/draw/draw_flatshade.c', + 'pipe/draw/draw_offset.c', + 'pipe/draw/draw_prim.c', + 'pipe/draw/draw_stipple.c', + 'pipe/draw/draw_twoside.c', + 'pipe/draw/draw_unfilled.c', + 'pipe/draw/draw_validate.c', + 'pipe/draw/draw_vbuf.c', + 'pipe/draw/draw_vertex.c', + 'pipe/draw/draw_vertex_cache.c', + 'pipe/draw/draw_vertex_fetch.c', + 'pipe/draw/draw_vertex_shader.c', + 'pipe/draw/draw_vertex_shader_llvm.c', + 'pipe/draw/draw_vf.c', + 'pipe/draw/draw_vf_generic.c', + 'pipe/draw/draw_vf_sse.c', + 'pipe/draw/draw_wide_prims.c', +] + +TGSIEXEC_SOURCES = [ + 'pipe/tgsi/exec/tgsi_exec.c', + 'pipe/tgsi/exec/tgsi_sse2.c', +] + +TGSIUTIL_SOURCES = [ + 'pipe/tgsi/util/tgsi_build.c', + 'pipe/tgsi/util/tgsi_dump.c', + 'pipe/tgsi/util/tgsi_parse.c', + 'pipe/tgsi/util/tgsi_util.c', +] + +STATECACHE_SOURCES = [ + 'pipe/cso_cache/cso_hash.c', + 'pipe/cso_cache/cso_cache.c', +] + +PIPEUTIL_SOURCES = [ + 'pipe/util/p_tile.c', + 'pipe/util/p_util.c', +] + +STATETRACKER_SOURCES = [ + 'state_tracker/st_atom.c', + 'state_tracker/st_atom_blend.c', + 'state_tracker/st_atom_clip.c', + 'state_tracker/st_atom_constbuf.c', + 'state_tracker/st_atom_depth.c', + 'state_tracker/st_atom_framebuffer.c', + 'state_tracker/st_atom_pixeltransfer.c', + 'state_tracker/st_atom_sampler.c', + 'state_tracker/st_atom_scissor.c', + 'state_tracker/st_atom_shader.c', + 'state_tracker/st_atom_rasterizer.c', + 'state_tracker/st_atom_stipple.c', + 'state_tracker/st_atom_texture.c', + 'state_tracker/st_atom_viewport.c', + 'state_tracker/st_cb_accum.c', + 'state_tracker/st_cb_bufferobjects.c', + 'state_tracker/st_cb_clear.c', + 'state_tracker/st_cb_flush.c', + 'state_tracker/st_cb_drawpixels.c', + 'state_tracker/st_cb_fbo.c', + 'state_tracker/st_cb_feedback.c', + 'state_tracker/st_cb_program.c', + 'state_tracker/st_cb_queryobj.c', + 'state_tracker/st_cb_rasterpos.c', + 'state_tracker/st_cb_readpixels.c', + 'state_tracker/st_cb_strings.c', + 'state_tracker/st_cb_texture.c', + 'state_tracker/st_cache.c', + 'state_tracker/st_context.c', + 'state_tracker/st_debug.c', + 'state_tracker/st_draw.c', + 'state_tracker/st_extensions.c', + 'state_tracker/st_format.c', + 'state_tracker/st_framebuffer.c', + 'state_tracker/st_mesa_to_tgsi.c', + 'state_tracker/st_program.c', + 'state_tracker/st_texture.c', +] + +SHADER_SOURCES = [ + 'shader/arbprogparse.c', + 'shader/arbprogram.c', + 'shader/atifragshader.c', + 'shader/grammar/grammar_mesa.c', + 'shader/nvfragparse.c', + 'shader/nvprogram.c', + 'shader/nvvertparse.c', + 'shader/program.c', + 'shader/prog_cache.c', + 'shader/prog_debug.c', + 'shader/prog_execute.c', + 'shader/prog_instruction.c', + 'shader/prog_parameter.c', + 'shader/prog_print.c', + 'shader/prog_statevars.c', + 'shader/programopt.c', + 'shader/shader_api.c', +] + +SLANG_SOURCES = [ + 'shader/slang/slang_builtin.c', + 'shader/slang/slang_codegen.c', + 'shader/slang/slang_compile.c', + 'shader/slang/slang_compile_function.c', + 'shader/slang/slang_compile_operation.c', + 'shader/slang/slang_compile_struct.c', + 'shader/slang/slang_compile_variable.c', + 'shader/slang/slang_emit.c', + 'shader/slang/slang_ir.c', + 'shader/slang/slang_label.c', + 'shader/slang/slang_library_noise.c', + 'shader/slang/slang_link.c', + 'shader/slang/slang_log.c', + 'shader/slang/slang_mem.c', + 'shader/slang/slang_preprocess.c', + 'shader/slang/slang_print.c', + 'shader/slang/slang_simplify.c', + 'shader/slang/slang_storage.c', + 'shader/slang/slang_typeinfo.c', + 'shader/slang/slang_vartable.c', + 'shader/slang/slang_utility.c', +] + + +####################################################################### +# Assembly sources + +ASM_C_SOURCES = [ + 'x86/common_x86.c', + 'x86/x86.c', + 'x86/3dnow.c', + 'x86/sse.c', + 'x86/rtasm/x86sse.c', + 'sparc/sparc.c', + 'ppc/common_ppc.c', + 'x86-64/x86-64.c', +] + +X86_SOURCES = [ + 'x86/common_x86_asm.S', + 'x86/x86_xform2.S', + 'x86/x86_xform3.S', + 'x86/x86_xform4.S', + 'x86/x86_cliptest.S', + 'x86/mmx_blend.S', + 'x86/3dnow_xform1.S', + 'x86/3dnow_xform2.S', + 'x86/3dnow_xform3.S', + 'x86/3dnow_xform4.S', + 'x86/3dnow_normal.S', + 'x86/sse_xform1.S', + 'x86/sse_xform2.S', + 'x86/sse_xform3.S', + 'x86/sse_xform4.S', + 'x86/sse_normal.S', + 'x86/read_rgba_span_x86.S', +] + +X86_API = [ + 'x86/glapi_x86.S', +] + +X86_64_SOURCES = [ + 'x86-64/xform4.S', +] + +X86_64_API = [ + 'x86-64/glapi_x86-64.S', +] + +SPARC_SOURCES = [ + 'sparc/clip.S', + 'sparc/norm.S', + 'sparc/xform.S', +] + +SPARC_API = [ + 'sparc/glapi_sparc.S', +] + +if x86: + ASM_SOURCES = ASM_C_SOURCES + X86_SOURCES + API_SOURCES = X86_API +else: + ASM_SOURCES = [] + API_SOURCES = [] + + +####################################################################### +# Driver sources + + +X11_DRIVER_SOURCES = [ + 'pipe/xlib/glxapi.c', + 'pipe/xlib/fakeglx.c', + 'pipe/xlib/xfonts.c', + 'pipe/xlib/xm_api.c', + 'pipe/xlib/xm_winsys.c', + 'pipe/xlib/xm_winsys_aub.c', + 'pipe/xlib/brw_aub.c', +] + +OSMESA_DRIVER_SOURCES = [ + 'drivers/osmesa/osmesa.c', +] + +GLIDE_DRIVER_SOURCES = [ + 'drivers/glide/fxapi.c', + 'drivers/glide/fxdd.c', + 'drivers/glide/fxddspan.c', + 'drivers/glide/fxddtex.c', + 'drivers/glide/fxsetup.c', + 'drivers/glide/fxtexman.c', + 'drivers/glide/fxtris.c', + 'drivers/glide/fxvb.c', + 'drivers/glide/fxglidew.c', + 'drivers/glide/fxg.c', +] + +SVGA_DRIVER_SOURCES = [ + 'drivers/svga/svgamesa.c', + 'drivers/svga/svgamesa8.c', + 'drivers/svga/svgamesa15.c', + 'drivers/svga/svgamesa16.c', + 'drivers/svga/svgamesa24.c', + 'drivers/svga/svgamesa32.c', +] + +FBDEV_DRIVER_SOURCES = [ + 'drivers/fbdev/glfbdev.c', +] + + +### All the core C sources + +SOLO_SOURCES = \ + MAIN_SOURCES + \ + MATH_SOURCES + \ + VBO_SOURCES + \ + VF_SOURCES + \ + DRAW_SOURCES + \ + TGSIEXEC_SOURCES + \ + TGSIUTIL_SOURCES + \ + PIPEUTIL_SOURCES + \ + STATECACHE_SOURCES + \ + STATETRACKER_SOURCES + \ + SHADER_SOURCES + \ + ASM_SOURCES + \ + SLANG_SOURCES + +CORE_SOURCES = \ + GLAPI_SOURCES + API_SOURCES + \ + SOLO_SOURCES + +ALL_SOURCES = \ + GLAPI_SOURCES + API_SOURCES + \ + SOLO_SOURCES + \ + ASM_SOURCES + \ + X11_DRIVER_SOURCES + \ + FBDEV_DRIVER_SOURCES + \ + OSMESA_DRIVER_SOURCES + + +###################################################################### +# Gallium sources + +SConscript([ + 'pipe/SConscript', +]) + + +###################################################################### +# libGL + +if not dri: + STAND_ALONE_DRIVER_SOURCES = \ + CORE_SOURCES + \ + X11_DRIVER_SOURCES + + Import( + 'softpipe', + 'i915simple', + 'i965simple' + ) + + pipe_drivers = [ + softpipe, + i965simple + ] + + env.SharedLibrary( + target ='GL', + source = STAND_ALONE_DRIVER_SOURCES, + LIBS = [softpipe, i965simple] + env['LIBS'], + ) + + +###################################################################### +# Driver sources + +if dri: + mesa = env.ConvenienceLibrary( + target = 'mesa', + source = SOLO_SOURCES, + ) + env.Prepend(LIBS = [mesa]) + + SConscript([ + 'drivers/dri/SConscript', + ]) diff --git a/src/mesa/drivers/dri/SConscript b/src/mesa/drivers/dri/SConscript new file mode 100644 index 0000000000..d32bd08669 --- /dev/null +++ b/src/mesa/drivers/dri/SConscript @@ -0,0 +1,48 @@ +Import('*') + +drienv = env.Clone() + +drienv.Replace(CPPPATH = [ + '#src/mesa/drivers/dri/common', + '#include', + '#include/GL/internal', + '#src/mesa', + '#src/mesa/main', + '#src/mesa/glapi', + '#src/mesa/math', + '#src/mesa/transform', + '#src/mesa/shader', + '#src/mesa/swrast', + '#src/mesa/swrast_setup', + '#src/egl/main', + '#src/egl/drivers/dri', +]) + +drienv.ParseConfig('pkg-config --cflags --libs libdrm') + +COMMON_GALLIUM_SOURCES = [ + '../common/utils.c', + '../common/vblank.c', + '../common/dri_util.c', + '../common/xmlconfig.c', +] + +COMMON_BM_SOURCES = [ + '../common/dri_bufmgr.c', + '../common/dri_drmpool.c', +] + +Export([ + 'drienv', + 'COMMON_GALLIUM_SOURCES', + 'COMMON_BM_SOURCES', +]) + +# TODO: Installation +#install: $(LIBNAME) +# $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) +# $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) + +SConscript([ + 'intel_winsys/SConscript', +]) diff --git a/src/mesa/drivers/dri/intel_winsys/SConscript b/src/mesa/drivers/dri/intel_winsys/SConscript new file mode 100644 index 0000000000..a7cc10450e --- /dev/null +++ b/src/mesa/drivers/dri/intel_winsys/SConscript @@ -0,0 +1,41 @@ +Import('*') + +env = drienv.Clone() + +env.Append(CPPPATH = [ + '../intel', + 'server' +]) + +#MINIGLX_SOURCES = server/intel_dri.c + +pipe_drivers = [ + softpipe, + i915simple +] + +DRIVER_SOURCES = [ + 'intel_winsys_pipe.c', + 'intel_winsys_softpipe.c', + 'intel_winsys_i915.c', + 'intel_batchbuffer.c', + 'intel_swapbuffers.c', + 'intel_context.c', + 'intel_lock.c', + 'intel_screen.c', + 'intel_batchpool.c', +] + +sources = \ + COMMON_GALLIUM_SOURCES + \ + COMMON_BM_SOURCES + \ + DRIVER_SOURCES + +# DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ +# && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") + +env.SharedLibrary( + target ='i915tex_dri.so', + source = sources, + LIBS = pipe_drivers + env['LIBS'], +) \ No newline at end of file diff --git a/src/mesa/pipe/SConscript b/src/mesa/pipe/SConscript new file mode 100644 index 0000000000..d9c20e0100 --- /dev/null +++ b/src/mesa/pipe/SConscript @@ -0,0 +1,9 @@ +Import('*') + +#env = env.Clone() + +SConscript([ + 'softpipe/SConscript', + 'i915simple/SConscript', + 'i965simple/SConscript', +]) diff --git a/src/mesa/pipe/i915simple/SConscript b/src/mesa/pipe/i915simple/SConscript new file mode 100644 index 0000000000..f5fb96b995 --- /dev/null +++ b/src/mesa/pipe/i915simple/SConscript @@ -0,0 +1,29 @@ +Import('*') + +env = env.Clone() + +i915simple = env.ConvenienceLibrary( + target = 'i915simple', + source = [ + 'i915_blit.c', + 'i915_clear.c', + 'i915_context.c', + 'i915_debug.c', + 'i915_debug_fp.c', + 'i915_flush.c', + 'i915_fpc_emit.c', + 'i915_fpc_translate.c', + 'i915_prim_emit.c', + 'i915_prim_vbuf.c', + 'i915_state.c', + 'i915_state_derived.c', + 'i915_state_dynamic.c', + 'i915_state_emit.c', + 'i915_state_immediate.c', + 'i915_state_sampler.c', + 'i915_strings.c', + 'i915_surface.c', + 'i915_texture.c', + ]) + +Export('i915simple') diff --git a/src/mesa/pipe/i965simple/SConscript b/src/mesa/pipe/i965simple/SConscript new file mode 100644 index 0000000000..74621de84c --- /dev/null +++ b/src/mesa/pipe/i965simple/SConscript @@ -0,0 +1,55 @@ +Import('*') + +env = env.Clone() + +i965simple = env.ConvenienceLibrary( + target = 'i965simple', + source = [ + 'brw_blit.c', + 'brw_cc.c', + 'brw_clip.c', + 'brw_clip_line.c', + 'brw_clip_point.c', + 'brw_clip_state.c', + 'brw_clip_tri.c', + 'brw_clip_util.c', + 'brw_context.c', + 'brw_curbe.c', + 'brw_draw.c', + 'brw_draw_upload.c', + 'brw_eu.c', + 'brw_eu_debug.c', + 'brw_eu_emit.c', + 'brw_eu_util.c', + 'brw_flush.c', + 'brw_gs.c', + 'brw_gs_emit.c', + 'brw_gs_state.c', + 'brw_misc_state.c', + 'brw_sf.c', + 'brw_sf_emit.c', + 'brw_sf_state.c', + 'brw_shader_info.c', + 'brw_state.c', + 'brw_state_batch.c', + 'brw_state_cache.c', + 'brw_state_pool.c', + 'brw_state_upload.c', + 'brw_strings.c', + 'brw_surface.c', + 'brw_tex_layout.c', + 'brw_urb.c', + 'brw_util.c', + 'brw_vs.c', + 'brw_vs_emit.c', + 'brw_vs_state.c', + 'brw_wm.c', + 'brw_wm_decl.c', + 'brw_wm_glsl.c', + 'brw_wm_iz.c', + 'brw_wm_sampler_state.c', + 'brw_wm_state.c', + 'brw_wm_surface_state.c', + ]) + +Export('i965simple') diff --git a/src/mesa/pipe/softpipe/SConscript b/src/mesa/pipe/softpipe/SConscript new file mode 100644 index 0000000000..d581ee8d3c --- /dev/null +++ b/src/mesa/pipe/softpipe/SConscript @@ -0,0 +1,42 @@ +Import('*') + +env = env.Clone() + +softpipe = env.ConvenienceLibrary( + target = 'softpipe', + source = [ + 'sp_clear.c', + 'sp_context.c', + 'sp_draw_arrays.c', + 'sp_flush.c', + 'sp_prim_setup.c', + 'sp_prim_vbuf.c', + 'sp_quad_alpha_test.c', + 'sp_quad_blend.c', + 'sp_quad_bufloop.c', + 'sp_quad.c', + 'sp_quad_colormask.c', + 'sp_quad_coverage.c', + 'sp_quad_depth_test.c', + 'sp_quad_earlyz.c', + 'sp_quad_fs.c', + 'sp_quad_occlusion.c', + 'sp_quad_output.c', + 'sp_quad_stencil.c', + 'sp_quad_stipple.c', + 'sp_query.c', + 'sp_state_blend.c', + 'sp_state_clip.c', + 'sp_state_derived.c', + 'sp_state_fs.c', + 'sp_state_rasterizer.c', + 'sp_state_sampler.c', + 'sp_state_surface.c', + 'sp_state_vertex.c', + 'sp_surface.c', + 'sp_tex_sample.c', + 'sp_texture.c', + 'sp_tile_cache.c', + ]) + +Export('softpipe') \ No newline at end of file -- cgit v1.2.3 From 50d5f304ad9eba8fe95a6cedfc56cd7213b33fea Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 31 Jan 2008 14:26:39 +0900 Subject: gallium: Portability fixes. --- src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index a2657dac59..b6af7cdedc 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -367,7 +367,7 @@ struct mm_pb_manager }; -static inline struct mm_pb_manager * +static INLINE struct mm_pb_manager * mm_pb_manager(struct pb_manager *mgr) { assert(mgr); @@ -385,7 +385,7 @@ struct mm_buffer }; -static inline struct mm_buffer * +static INLINE struct mm_buffer * mm_buffer(struct pb_buffer *buf) { assert(buf); -- cgit v1.2.3 From e82c9b867cc18af5da7375871a685f98d1c1527d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 31 Jan 2008 11:57:15 +0000 Subject: tgsi: Use ESI instead of EBX as temp reg on non-win32 --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index f8660e7ad1..df0c698301 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -198,9 +198,15 @@ get_output_base( void ) static struct x86_reg get_temp_base( void ) { +#ifdef WIN32 return x86_make_reg( file_REG32, reg_BX ); +#else + return x86_make_reg( + file_REG32, + reg_SI ); +#endif } static struct x86_reg -- cgit v1.2.3 From 256486829f0bc2be7a986a6bdc08df5fc16b77d8 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 31 Jan 2008 08:12:47 -0700 Subject: Cell: set GALLIUM_CELL_VS env var to enable SPU-based vertex transformation --- src/mesa/pipe/cell/ppu/cell_context.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c index 4885cd0d2c..bbe1fd7a11 100644 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ b/src/mesa/pipe/cell/ppu/cell_context.c @@ -162,8 +162,12 @@ cell_draw_create(struct cell_context *cell) { struct draw_context *draw = draw_create(); - draw->shader_queue_flush = cell_vertex_shader_queue_flush; - draw->driver_private = cell; + if (getenv("GALLIUM_CELL_VS")) { + /* plug in SPU-based vertex transformation code */ + draw->shader_queue_flush = cell_vertex_shader_queue_flush; + draw->driver_private = cell; + } + return draw; } -- cgit v1.2.3 From cd53eb0db19daf1c9aac94011a54e902eb10fe75 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 31 Jan 2008 08:21:38 -0700 Subject: Cell: SIMD-ize const_coeff() --- src/mesa/pipe/cell/spu/spu_tri.c | 24 +++++++++--------------- 1 file changed, 9 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index e436e153ec..08b8bf0c9c 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -723,24 +723,18 @@ static boolean setup_sort_vertices(const struct vertex_header *v0, /** * Compute a0 for a constant-valued coefficient (GL_FLAT shading). - * The value value comes from vertex->data[slot][i]. - * The result will be put into setup.coef[slot].a0[i]. + * The value value comes from vertex->data[slot]. + * The result will be put into setup.coef[slot].a0. * \param slot which attribute slot - * \param i which component of the slot (0..3) */ -static void const_coeff(uint slot) +static INLINE void const_coeff(uint slot) { - uint i; - ASSERT(slot < PIPE_MAX_SHADER_INPUTS); - - for (i = 0; i < 4; i++) { - setup.coef[slot].dadx.f[i] = 0; - setup.coef[slot].dady.f[i] = 0; - - /* need provoking vertex info! - */ - setup.coef[slot].a0.f[i] = setup.vprovoke->data[slot][i]; - } + setup.coef[slot].dadx.v = (vector float) {0.0, 0.0, 0.0, 0.0}; + setup.coef[slot].dady.v = (vector float) {0.0, 0.0, 0.0, 0.0}; + setup.coef[slot].a0.f[0] = setup.vprovoke->data[slot][0]; + setup.coef[slot].a0.f[1] = setup.vprovoke->data[slot][1]; + setup.coef[slot].a0.f[2] = setup.vprovoke->data[slot][2]; + setup.coef[slot].a0.f[3] = setup.vprovoke->data[slot][3]; } -- cgit v1.2.3 From c36cdc61561fee21108f0a68ca661f0d8c7f5d94 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 31 Jan 2008 13:32:29 -0700 Subject: gallium: fix get/put typo regression This came from commit f3aa4de034b0d791ce2e38e8aeb3b3abdb4e3b50 on 1/22/08. Fixes strange Z buffer glitches seen in progs/glsl/texdemo1.c --- src/mesa/pipe/softpipe/sp_tile_cache.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index 451e157abf..ccf367a5e4 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -415,8 +415,8 @@ sp_get_cached_tile(struct softpipe_context *softpipe, /* put dirty tile back in framebuffer */ if (tc->depth_stencil) { pipe_put_tile_raw(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); } else { pipe_put_tile_rgba(pipe, ps, @@ -441,9 +441,9 @@ sp_get_cached_tile(struct softpipe_context *softpipe, else { /* get new tile data from surface */ if (tc->depth_stencil) { - pipe_put_tile_raw(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); + pipe_get_tile_raw(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); } else { pipe_get_tile_rgba(pipe, ps, -- cgit v1.2.3 From 9536314a6c99d0acc249180034865b5cfb927e9d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 31 Jan 2008 13:36:00 -0700 Subject: gallium: Fix z clear bug when TILE_CLEAR_OPTIMIZATION==0 --- src/mesa/pipe/softpipe/sp_clear.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c index 571f64b38d..8d295a30ca 100644 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ b/src/mesa/pipe/softpipe/sp_clear.c @@ -55,7 +55,9 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, if (ps == sp_tile_cache_get_surface(softpipe->zsbuf_cache)) { sp_tile_cache_clear(softpipe->zsbuf_cache, clearValue); +#if TILE_CLEAR_OPTIMIZATION return; +#endif } for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { -- cgit v1.2.3 From acb81374c1d476ebffbcfc8405db7fff6cc6d6c3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 31 Jan 2008 13:37:01 -0700 Subject: gallium: comments about fragment Z computation --- src/mesa/pipe/softpipe/sp_quad_fs.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index c9cc8afa0c..90691c6065 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -168,6 +168,11 @@ shade_quad( sizeof( quad->outputs.color ) ); } + /* + * XXX the following code for updating quad->outputs.depth + * isn't really needed if we did early z testing. + */ + /* store result Z */ if (qss->depthOutSlot >= 0) { /* output[slot] is new Z */ @@ -181,6 +186,10 @@ shade_quad( uint i; for (i = 0; i < 4; i++) { quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; + /* XXX not sure the above line is always correct. The following + * might be better: + quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i]; + */ } } -- cgit v1.2.3 From 86787043fae59869133180474cb09dac4f2e619a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 31 Jan 2008 14:05:04 -0700 Subject: Fix problem in mapping vertex program outputs (found with "spring" game engine) If the vertex program writes to an output that's not consumed by the fragment program, map the vp output to an unused slot. --- src/mesa/state_tracker/st_atom_shader.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 1ed9333556..9196918509 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -226,9 +226,11 @@ find_translated_vp(struct st_context *st, GLint fpInAttrib = vp_out_to_fp_in(outAttr); if (fpInAttrib >= 0) { GLuint fpInSlot = stfp->input_to_slot[fpInAttrib]; - GLuint vpOutSlot = stfp->fs->state.input_map[fpInSlot]; - xvp->output_to_slot[outAttr] = vpOutSlot; - numVpOuts++; + if (fpInSlot != ~0) { + GLuint vpOutSlot = stfp->fs->state.input_map[fpInSlot]; + xvp->output_to_slot[outAttr] = vpOutSlot; + numVpOuts++; + } } else if (outAttr == VERT_RESULT_PSIZ || outAttr == VERT_RESULT_BFC0 || @@ -247,7 +249,7 @@ find_translated_vp(struct st_context *st, * We could use this info to do dead code elimination in the * vertex program. */ - dummySlot = stfp->num_input_slots; + dummySlot = numVpOuts; /* Map vert program outputs that aren't used to the dummy slot */ for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { -- cgit v1.2.3 From 633e1133aeed04df650b97c8d25a041014fd6f5e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 31 Jan 2008 14:35:25 -0700 Subject: gallium: fix problem in which texcoords and varying vars got mapped to the same slot This fixes the glsl/bump.c and glsl/texdemo1.c programs --- src/mesa/state_tracker/st_program.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 1f1e6500e0..84a9094001 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -33,6 +33,7 @@ #include "main/imports.h" #include "main/mtypes.h" +#include "shader/prog_print.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -68,6 +69,7 @@ st_translate_vertex_program(struct st_context *st, struct pipe_shader_state vs; const struct cso_vertex_shader *cso; GLuint attr, i; + GLuint num_generic = 0; memset(&vs, 0, sizeof(vs)); @@ -117,7 +119,7 @@ st_translate_vertex_program(struct st_context *st, case VERT_ATTRIB_TEX6: case VERT_ATTRIB_TEX7: vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[slot] = attr - VERT_ATTRIB_TEX0; + vs.input_semantic_index[slot] = num_generic++; break; case VERT_ATTRIB_GENERIC0: case VERT_ATTRIB_GENERIC1: @@ -129,7 +131,7 @@ st_translate_vertex_program(struct st_context *st, case VERT_ATTRIB_GENERIC7: assert(attr < VERT_ATTRIB_MAX); vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[slot] = attr - VERT_ATTRIB_GENERIC0; + vs.input_semantic_index[slot] = num_generic++; break; default: assert(0); @@ -143,6 +145,7 @@ st_translate_vertex_program(struct st_context *st, vs.output_semantic_index[i] = 0; } + num_generic = 0; /* * Determine number of outputs, the (default) output register * mapping and the semantic information for each output. @@ -207,14 +210,14 @@ st_translate_vertex_program(struct st_context *st, case VERT_RESULT_TEX6: case VERT_RESULT_TEX7: vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[slot] = attr - VERT_RESULT_TEX0; + vs.output_semantic_index[slot] = num_generic++; break; case VERT_RESULT_VAR0: /* fall-through */ default: assert(attr - VERT_RESULT_VAR0 < MAX_VARYING); vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[slot] = attr - VERT_RESULT_VAR0; + vs.output_semantic_index[slot] = num_generic++; } } } @@ -258,6 +261,9 @@ st_translate_vertex_program(struct st_context *st, cso = st_cached_vs_state(st, &vs); stvp->cso = cso; + if (0) + _mesa_print_program(&stvp->Base.Base); + if (TGSI_DEBUG) tgsi_dump( tokensOut, 0 ); } @@ -286,6 +292,7 @@ st_translate_fragment_program(struct st_context *st, GLuint attr; const GLbitfield inputsRead = stfp->Base.Base.InputsRead; GLuint vslot = 0; + GLuint num_generic = 0; memset(&fs, 0, sizeof(fs)); @@ -338,14 +345,14 @@ st_translate_fragment_program(struct st_context *st, case FRAG_ATTRIB_TEX6: case FRAG_ATTRIB_TEX7: fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - fs.input_semantic_index[slot] = attr - FRAG_ATTRIB_TEX0; + fs.input_semantic_index[slot] = num_generic++; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; case FRAG_ATTRIB_VAR0: /* fall-through */ default: fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - fs.input_semantic_index[slot] = attr - FRAG_ATTRIB_VAR0; + fs.input_semantic_index[slot] = num_generic++; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; } } @@ -415,6 +422,9 @@ st_translate_fragment_program(struct st_context *st, cso = st_cached_fs_state(st, &fs); stfp->fs = cso; + if (0) + _mesa_print_program(&stfp->Base.Base); + if (TGSI_DEBUG) tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); -- cgit v1.2.3 From 635341ec5b06b3db453e88f44663d2ad711c3f7d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 31 Jan 2008 17:05:43 -0700 Subject: fix typo --- src/mesa/pipe/p_shader_tokens.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_shader_tokens.h b/src/mesa/pipe/p_shader_tokens.h index e9d1d66bda..3ce35310f6 100644 --- a/src/mesa/pipe/p_shader_tokens.h +++ b/src/mesa/pipe/p_shader_tokens.h @@ -626,7 +626,7 @@ struct tgsi_src_register_ext /* * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_SWZ, - * it should be cast to tgsi_src_register_ext_extswz. + * it should be cast to tgsi_src_register_ext_swz. * * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_MOD, * it should be cast to tgsi_src_register_ext_mod. -- cgit v1.2.3 From 20df285b14bc655d5429c7d2b82446204f9a1f2e Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 31 Jan 2008 17:22:07 -0800 Subject: Fix using "ccache ppu-gcc" for CC and fix parallel builds CC wasn't quoted in a couple places in src/mesa/Makefile. Also, the OSMesa link was missing a dependency. --- src/mesa/Makefile | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index b16d74bf49..720f1b2e02 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -125,24 +125,25 @@ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(LLVM_LIB) @ $(TOP)/bin/mklib -o $(GL_LIB) \ - -linker $(CC) \ + -linker "$(CC)" \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(LLVM_LIB) $(GL_LIB_DEPS) # Make the OSMesa library -$(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) $(OSMESA16_OBJECTS) +$(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) \ + $(OSMESA16_OBJECTS) $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) @ if [ "${DRIVER_DIRS}" = "osmesa" ] ; then \ $(TOP)/bin/mklib -o $(OSMESA_LIB) \ - -linker $(CC) \ + -linker "$(CC)" \ -major $(MESA_MAJOR) \ -minor $(MESA_MINOR) -patch $(MESA_TINY) \ -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ $(OSMESA_LIB_DEPS) $(OSMESA16_OBJECTS) ; \ else \ $(TOP)/bin/mklib -o $(OSMESA_LIB) \ - -linker $(CC) \ + -linker "$(CC)" \ -major $(MESA_MAJOR) \ -minor $(MESA_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ -- cgit v1.2.3 From b108bea6b44c1abc6d61e3e47096e5122de89cd1 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 09:27:57 -0700 Subject: Cell: store current tile status in cur_tile_status_c/z, add TILE_STATUS_GETTING --- src/mesa/pipe/cell/spu/spu_render.c | 36 ++++++++++++++++----- src/mesa/pipe/cell/spu/spu_tile.c | 1 + src/mesa/pipe/cell/spu/spu_tile.h | 8 +++-- src/mesa/pipe/cell/spu/spu_tri.c | 62 ++++++++++++++++++++++++++++++------- src/mesa/pipe/cell/spu/spu_tri.h | 2 +- 5 files changed, 87 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_render.c b/src/mesa/pipe/cell/spu/spu_render.c index f506095116..ca54a103bd 100644 --- a/src/mesa/pipe/cell/spu/spu_render.c +++ b/src/mesa/pipe/cell/spu/spu_render.c @@ -95,13 +95,15 @@ static INLINE void get_cz_tiles(uint tx, uint ty) { if (spu.depth_stencil.depth.enabled) { - if (tile_status_z[ty][tx] != TILE_STATUS_CLEAR) { + if (cur_tile_status_z != TILE_STATUS_CLEAR) { get_tile(tx, ty, &ztile, TAG_READ_TILE_Z, 1); + cur_tile_status_z = TILE_STATUS_GETTING; } } - if (tile_status[ty][tx] != TILE_STATUS_CLEAR) { + if (cur_tile_status_c != TILE_STATUS_CLEAR) { get_tile(tx, ty, &ctile, TAG_READ_TILE_COLOR, 0); + cur_tile_status_c = TILE_STATUS_GETTING; } } @@ -112,14 +114,24 @@ get_cz_tiles(uint tx, uint ty) static INLINE void put_cz_tiles(uint tx, uint ty) { - if (tile_status_z[ty][tx] == TILE_STATUS_DIRTY) { + if (cur_tile_status_z == TILE_STATUS_DIRTY) { + /* tile was modified and needs to be written back */ put_tile(tx, ty, &ztile, TAG_WRITE_TILE_Z, 1); - tile_status_z[ty][tx] = TILE_STATUS_DEFINED; + cur_tile_status_z = TILE_STATUS_DEFINED; + } + else if (cur_tile_status_z == TILE_STATUS_GETTING) { + /* tile was never used */ + cur_tile_status_z = TILE_STATUS_DEFINED; } - if (tile_status[ty][tx] == TILE_STATUS_DIRTY) { + if (cur_tile_status_c == TILE_STATUS_DIRTY) { + /* tile was modified and needs to be written back */ put_tile(tx, ty, &ctile, TAG_WRITE_TILE_COLOR, 0); - tile_status[ty][tx] = TILE_STATUS_DEFINED; + cur_tile_status_c = TILE_STATUS_DEFINED; + } + else if (cur_tile_status_c == TILE_STATUS_GETTING) { + /* tile was never used */ + cur_tile_status_c = TILE_STATUS_DEFINED; } } @@ -238,8 +250,13 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) if (!my_tile(tx, ty)) continue; + cur_tile_status_c = tile_status[ty][tx]; + cur_tile_status_z = tile_status_z[ty][tx]; + get_cz_tiles(tx, ty); + uint drawn = 0; + /* loop over tris */ for (j = 0; j < render->num_indexes; j += 3) { const float *v0, *v1, *v2; @@ -248,13 +265,18 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) v1 = (const float *) (vertices + indexes[j+1] * vertex_size); v2 = (const float *) (vertices + indexes[j+2] * vertex_size); - tri_draw(v0, v1, v2, tx, ty); + drawn += tri_draw(v0, v1, v2, tx, ty); } + //printf("SPU %u: drew %u of %u\n", spu.init.id, drawn, render->num_indexes/3); + /* write color/z tiles back to main framebuffer, if dirtied */ put_cz_tiles(tx, ty); wait_put_cz_tiles(); /* XXX seems unnecessary... */ + + tile_status[ty][tx] = cur_tile_status_c; + tile_status_z[ty][tx] = cur_tile_status_z; } if (Debug) diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c index ca1352f9f8..aea4785bc2 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.c +++ b/src/mesa/pipe/cell/spu/spu_tile.c @@ -37,6 +37,7 @@ tile_t ztile ALIGN16_ATTRIB; ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; +ubyte cur_tile_status_c, cur_tile_status_z; void diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h index 18d1b3c117..1f123a2b7b 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -51,12 +51,16 @@ extern tile_t ztile ALIGN16_ATTRIB; #define TILE_STATUS_CLEAR 1 -#define TILE_STATUS_DEFINED 2 /**< defined pixel data */ -#define TILE_STATUS_DIRTY 3 /**< modified, but not put back yet */ +#define TILE_STATUS_DEFINED 2 /**< defined in FB, but not in local store */ +#define TILE_STATUS_CLEAN 3 /**< in local store, but not changed */ +#define TILE_STATUS_DIRTY 4 /**< modified locally, but not put back yet */ +#define TILE_STATUS_GETTING 5 /**< mfc_get() called but not yet arrived */ extern ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; extern ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; +extern ubyte cur_tile_status_c, cur_tile_status_z; + void get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf); diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 08b8bf0c9c..a32878d917 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -299,16 +299,23 @@ do_depth_test(int x, int y, unsigned int mask) zvals.v = eval_z((float) x, (float) y); - if (tile_status_z[setup.ty][setup.tx] == TILE_STATUS_CLEAR) { + if (cur_tile_status_c == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ clear_z_tile(&ztile); + cur_tile_status_z = TILE_STATUS_DIRTY; } - else if (tile_status_z[setup.ty][setup.tx] != TILE_STATUS_DIRTY) { + +#if 0 + if (cur_tile_status_z == TILE_STATUS_CLEAR) { + /* now, _really_ clear the tile */ + clear_z_tile(&ztile); + } + else if (cur_tile_status_z != TILE_STATUS_DIRTY) { /* make sure we've got the tile from main mem */ wait_on_mask(1 << TAG_READ_TILE_Z); } - tile_status_z[setup.ty][setup.tx] = TILE_STATUS_DIRTY; - + cur_tile_status_z = TILE_STATUS_DIRTY; +#endif if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { zvals.v = spu_mul(zvals.v, zscale16.v); @@ -380,6 +387,9 @@ do_depth_test(int x, int y, unsigned int mask) } } + if (mask) + cur_tile_status_z = TILE_STATUS_DIRTY; + return mask; } @@ -397,15 +407,15 @@ do_depth_test_simd(int x, int y, vector unsigned int quadmask) zvals.v = eval_z((float) x, (float) y); - if (tile_status_z[setup.ty][setup.tx] == TILE_STATUS_CLEAR) { + if (cur_tile_status_z == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ clear_z_tile(&ztile); } - else if (tile_status_z[setup.ty][setup.tx] != TILE_STATUS_DIRTY) { + else if (cur_tile_status_z != TILE_STATUS_DIRTY) { /* make sure we've got the tile from main mem */ wait_on_mask(1 << TAG_READ_TILE_Z); } - tile_status_z[setup.ty][setup.tx] = TILE_STATUS_DIRTY; + cur_tile_status_z = TILE_STATUS_DIRTY; /* XXX fetch Z value sooner to hide latency here */ zmask = spu_cmpgt(ztile.f4[ix][iy].v, zvals.v); @@ -462,15 +472,23 @@ emit_quad( int x, int y, mask_t mask ) if (mask) #endif { - if (tile_status[setup.ty][setup.tx] == TILE_STATUS_CLEAR) { + if (cur_tile_status_c == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ clear_c_tile(&ctile); } - else if (tile_status[setup.ty][setup.tx] != TILE_STATUS_DIRTY) { + +#if 0 + if (cur_tile_status_c == TILE_STATUS_CLEAR) { + /* now, _really_ clear the tile */ + clear_c_tile(&ctile); + cur_tile_status_c = TILE_STATUS_DIRTY; + } + else if (cur_tile_status_c != TILE_STATUS_DIRTY) { /* make sure we've got the tile from main mem */ wait_on_mask(1 << TAG_READ_TILE_COLOR); } - tile_status[setup.ty][setup.tx] = TILE_STATUS_DIRTY; +#endif + cur_tile_status_c = TILE_STATUS_DIRTY; #if SIMD_Z if (spu_extract(mask, 0)) @@ -970,7 +988,7 @@ static void subtriangle( struct edge *eleft, * Draw triangle into tile at (tx, ty) (tile coords) * The tile data should have already been fetched. */ -void +boolean tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) { setup.tx = tx; @@ -985,7 +1003,7 @@ tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) if (!setup_sort_vertices((struct vertex_header *) v0, (struct vertex_header *) v1, (struct vertex_header *) v2)) { - return; /* totally clipped */ + return FALSE; /* totally clipped */ } setup_tri_coefficients(); @@ -999,6 +1017,24 @@ tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) /* init_constant_attribs( setup ); */ + if (cur_tile_status_c == TILE_STATUS_GETTING) { + /* wait for mfc_get() to complete */ + wait_on_mask(1 << TAG_READ_TILE_COLOR); + cur_tile_status_c = TILE_STATUS_CLEAN; + } + + ASSERT(cur_tile_status_c != TILE_STATUS_DEFINED); + + if (spu.depth_stencil.depth.enabled) { + if (cur_tile_status_z == TILE_STATUS_GETTING) { + /* wait for mfc_get() to complete */ + wait_on_mask(1 << TAG_READ_TILE_Z); + cur_tile_status_z = TILE_STATUS_CLEAN; + } + ASSERT(cur_tile_status_z != TILE_STATUS_DEFINED); + } + + if (setup.oneoverarea < 0.0) { /* emaj on left: */ @@ -1013,4 +1049,6 @@ tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) } flush_spans(); + + return TRUE; } diff --git a/src/mesa/pipe/cell/spu/spu_tri.h b/src/mesa/pipe/cell/spu/spu_tri.h index 86c42b6339..aa694dd7c9 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.h +++ b/src/mesa/pipe/cell/spu/spu_tri.h @@ -30,7 +30,7 @@ #define SPU_TRI_H -extern void +extern boolean tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty); -- cgit v1.2.3 From 59be082909de6021ec7d08476253bd4c9920e137 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 13:45:58 -0700 Subject: Cell: implement Z16 and Z32 testing with SIMD instructions. --- src/mesa/pipe/cell/spu/spu_tile.h | 3 +- src/mesa/pipe/cell/spu/spu_tri.c | 222 +++++-------------------------------- src/mesa/pipe/cell/spu/spu_ztest.h | 135 ++++++++++++++++++++++ 3 files changed, 163 insertions(+), 197 deletions(-) create mode 100644 src/mesa/pipe/cell/spu/spu_ztest.h (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h index 1f123a2b7b..4b1ef2a4c8 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -42,7 +42,8 @@ typedef union { ushort t16[TILE_SIZE][TILE_SIZE]; uint t32[TILE_SIZE][TILE_SIZE]; - float4 f4[TILE_SIZE/2][TILE_SIZE/2]; + vector unsigned short us8[TILE_SIZE/2][TILE_SIZE/4]; + vector unsigned int ui4[TILE_SIZE/2][TILE_SIZE/2]; } tile_t; diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index a32878d917..a26a4f098d 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -39,18 +39,11 @@ #include "spu_tile.h" #include "spu_tri.h" +#include "spu_ztest.h" -/* - * If SIMD_Z=1 the Z buffer is floating point and we use vector instructions - * to do Z testing/updating. - */ -#define SIMD_Z 0 -#if SIMD_Z +/** Masks are uint[4] vectors with each element being 0 or 0xffffffff */ typedef vector unsigned int mask_t; -#else -typedef uint mask_t; -#endif /** @@ -282,20 +275,11 @@ pack_colors(uint uicolors[4], const float4 fcolors[4]) } - -static unsigned int -do_depth_test(int x, int y, unsigned int mask) +static INLINE mask_t +do_depth_test(int x, int y, mask_t quadmask) { - static const float4 zscale16 - = {.f={65535.0, 65535.0, 65535.0, 65535.0}}; - static const float4 zscale32 - = {.f={(float)0xffffffff, - (float)0xffffffff, - (float)0xffffffff, - (float)0xffffffff}}; - int ix = x - setup.cliprect_minx; - int iy = y - setup.cliprect_miny; float4 zvals; + mask_t mask; zvals.v = eval_z((float) x, (float) y); @@ -305,129 +289,20 @@ do_depth_test(int x, int y, unsigned int mask) cur_tile_status_z = TILE_STATUS_DIRTY; } -#if 0 - if (cur_tile_status_z == TILE_STATUS_CLEAR) { - /* now, _really_ clear the tile */ - clear_z_tile(&ztile); - } - else if (cur_tile_status_z != TILE_STATUS_DIRTY) { - /* make sure we've got the tile from main mem */ - wait_on_mask(1 << TAG_READ_TILE_Z); - } - cur_tile_status_z = TILE_STATUS_DIRTY; -#endif - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - zvals.v = spu_mul(zvals.v, zscale16.v); - if (mask & MASK_TOP_LEFT) { - uint z = (uint) zvals.f[0]; - if (z < ztile.t16[iy][ix]) - ztile.t16[iy][ix] = z; - else - mask &= ~MASK_TOP_LEFT; - } - - if (mask & MASK_TOP_RIGHT) { - uint z = (uint) zvals.f[1]; - if (z < ztile.t16[iy][ix+1]) - ztile.t16[iy][ix+1] = z; - else - mask &= ~MASK_TOP_RIGHT; - } - - if (mask & MASK_BOTTOM_LEFT) { - uint z = (uint) zvals.f[2]; - if (z < ztile.t16[iy+1][ix]) - ztile.t16[iy+1][ix] = z; - else - mask &= ~MASK_BOTTOM_LEFT; - } - - if (mask & MASK_BOTTOM_RIGHT) { - uint z = (uint) zvals.f[3]; - if (z < ztile.t16[iy+1][ix+1]) - ztile.t16[iy+1][ix+1] = z; - else - mask &= ~MASK_BOTTOM_RIGHT; - } + int ix = (x - setup.cliprect_minx) / 4; + int iy = (y - setup.cliprect_miny) / 2; + mask = spu_z16_test_less(zvals.v, &ztile.us8[iy][ix], x>>1, quadmask); } else { - zvals.v = spu_mul(zvals.v, zscale32.v); - ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); - if (mask & MASK_TOP_LEFT) { - uint z = (uint) zvals.f[0]; - if (z < ztile.t32[iy][ix]) - ztile.t32[iy][ix] = z; - else - mask &= ~MASK_TOP_LEFT; - } - - if (mask & MASK_TOP_RIGHT) { - uint z = (uint) zvals.f[1]; - if (z < ztile.t32[iy][ix+1]) - ztile.t32[iy][ix+1] = z; - else - mask &= ~MASK_TOP_RIGHT; - } - - if (mask & MASK_BOTTOM_LEFT) { - uint z = (uint) zvals.f[2]; - if (z < ztile.t32[iy+1][ix]) - ztile.t32[iy+1][ix] = z; - else - mask &= ~MASK_BOTTOM_LEFT; - } - - if (mask & MASK_BOTTOM_RIGHT) { - uint z = (uint) zvals.f[3]; - if (z < ztile.t32[iy+1][ix+1]) - ztile.t32[iy+1][ix+1] = z; - else - mask &= ~MASK_BOTTOM_RIGHT; - } + int ix = (x - setup.cliprect_minx) / 2; + int iy = (y - setup.cliprect_miny) / 2; + mask = spu_z32_test_less(zvals.v, &ztile.ui4[iy][ix], quadmask); } - - if (mask) - cur_tile_status_z = TILE_STATUS_DIRTY; - return mask; } - - -static vector unsigned int -do_depth_test_simd(int x, int y, vector unsigned int quadmask) -{ - int ix = (x - setup.cliprect_minx) / 2; - int iy = (y - setup.cliprect_miny) / 2; - float4 zvals; - - vector unsigned int zmask; - - zvals.v = eval_z((float) x, (float) y); - - if (cur_tile_status_z == TILE_STATUS_CLEAR) { - /* now, _really_ clear the tile */ - clear_z_tile(&ztile); - } - else if (cur_tile_status_z != TILE_STATUS_DIRTY) { - /* make sure we've got the tile from main mem */ - wait_on_mask(1 << TAG_READ_TILE_Z); - } - cur_tile_status_z = TILE_STATUS_DIRTY; - - /* XXX fetch Z value sooner to hide latency here */ - zmask = spu_cmpgt(ztile.f4[ix][iy].v, zvals.v); - zmask = spu_and(zmask, quadmask); - - ztile.f4[ix][iy].v = spu_sel(ztile.f4[ix][iy].v, zvals.v, zmask); - //ztile.f4[ix][iy].v = spu_sel(zvals.v, ztile.f4[ix][iy].v, mask4); - - return zmask; -} - - /** * Emit a quad (pass to next stage). No clipping is done. */ @@ -461,36 +336,18 @@ emit_quad( int x, int y, mask_t mask ) } if (spu.depth_stencil.depth.enabled) { -#if SIMD_Z - mask = do_depth_test_simd(x, y, mask); -#else mask = do_depth_test(x, y, mask); -#endif } -#if !SIMD_Z - if (mask) -#endif - { - if (cur_tile_status_c == TILE_STATUS_CLEAR) { - /* now, _really_ clear the tile */ - clear_c_tile(&ctile); - } + /* If any bits in mask are set... */ + if (spu_extract(spu_orx(mask), 0)) { -#if 0 if (cur_tile_status_c == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ clear_c_tile(&ctile); - cur_tile_status_c = TILE_STATUS_DIRTY; } - else if (cur_tile_status_c != TILE_STATUS_DIRTY) { - /* make sure we've got the tile from main mem */ - wait_on_mask(1 << TAG_READ_TILE_COLOR); - } -#endif cur_tile_status_c = TILE_STATUS_DIRTY; -#if SIMD_Z if (spu_extract(mask, 0)) ctile.t32[iy][ix] = colors[QUAD_TOP_LEFT]; if (spu_extract(mask, 1)) @@ -499,20 +356,11 @@ emit_quad( int x, int y, mask_t mask ) ctile.t32[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; if (spu_extract(mask, 3)) ctile.t32[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; -#elif 0 + +#if 0 /* SIMD_Z with swizzled color buffer (someday) */ vector float icolors = *((vector float *) &colors); ctile.f4[iy/2][ix/2].v = spu_sel(ctile.f4[iy/2][ix/2].v, icolors, mask); - -#else - if (mask & MASK_TOP_LEFT) - ctile.t32[iy][ix] = colors[QUAD_TOP_LEFT]; - if (mask & MASK_TOP_RIGHT) - ctile.t32[iy][ix+1] = colors[QUAD_TOP_RIGHT]; - if (mask & MASK_BOTTOM_LEFT) - ctile.t32[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; - if (mask & MASK_BOTTOM_RIGHT) - ctile.t32[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; #endif } @@ -533,38 +381,20 @@ static INLINE int block( int x ) /** * Compute mask which indicates which pixels in the 2x2 quad are actually inside * the triangle's bounds. - * - * this is pretty nasty... may need to rework flush_spans again to - * fix it, if possible. + * The mask is a uint4 vector and each element will be 0 or 0xffffffff. */ -static mask_t calculate_mask( int x ) +static INLINE mask_t calculate_mask( int x ) { -#if SIMD_Z - uint m0, m1, m2, m3; - - m0 = (x >= setup.span.left[0] && x < setup.span.right[0]) * ~0; - m1 = (x+1 >= setup.span.left[0] && x+1 < setup.span.right[0]) * ~0; - m2 = (x >= setup.span.left[1] && x < setup.span.right[1]) * ~0; - m3 = (x+1 >= setup.span.left[1] && x+1 < setup.span.right[1]) * ~0; - - return (vector unsigned int) {m0, m1, m2, m3}; -#else - unsigned mask = 0x0; - - if (x >= setup.span.left[0] && x < setup.span.right[0]) - mask |= MASK_TOP_LEFT; - - if (x >= setup.span.left[1] && x < setup.span.right[1]) - mask |= MASK_BOTTOM_LEFT; - - if (x+1 >= setup.span.left[0] && x+1 < setup.span.right[0]) - mask |= MASK_TOP_RIGHT; - - if (x+1 >= setup.span.left[1] && x+1 < setup.span.right[1]) - mask |= MASK_BOTTOM_RIGHT; - + /* This is a little tricky. + * Use & instead of && to avoid branches. + * Use negation to convert true/false to ~0/0 values. + */ + mask_t mask; + mask = spu_insert(-((x >= setup.span.left[0]) & (x < setup.span.right[0])), mask, 0); + mask = spu_insert(-((x+1 >= setup.span.left[0]) & (x+1 < setup.span.right[0])), mask, 1); + mask = spu_insert(-((x >= setup.span.left[1]) & (x < setup.span.right[1])), mask, 2); + mask = spu_insert(-((x+1 >= setup.span.left[1]) & (x+1 < setup.span.right[1])), mask, 3); return mask; -#endif } diff --git a/src/mesa/pipe/cell/spu/spu_ztest.h b/src/mesa/pipe/cell/spu/spu_ztest.h new file mode 100644 index 0000000000..5fefb15176 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_ztest.h @@ -0,0 +1,135 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Zbuffer/depth test code. + */ + + +#ifndef SPU_ZTEST_H +#define SPU_ZTEST_H + + +#ifdef __SPU__ +#include +#endif + + + +/** + * Perform Z testing for a 16-bit/value Z buffer. + * + * \param zvals vector of four fragment zvalues as floats + * \param zbuf ptr to vector of ushort[8] zbuffer values. Note that this + * contains the Z values for 2 quads, 8 pixels. + * \param x x coordinate of quad (only lsbit is significant) + * \param inMask indicates which fragments in the quad are alive + * \return new mask indicating which fragments are alive after ztest + */ +static INLINE vector unsigned int +spu_z16_test_less(vector float zvals, vector unsigned short *zbuf, + uint x, vector unsigned int inMask) +{ +#define ZERO 0x80 + vector unsigned int zvals_ui4, zbuf_ui4, mask; + + /* convert floats to uints in [0, 65535] */ + zvals_ui4 = spu_convtu(zvals, 32); /* convert to [0, 2^32] */ + zvals_ui4 = spu_rlmask(zvals_ui4, -16); /* right shift 16 */ + + /* XXX this conditional could be removed with a bit of work */ + if (x & 1) { + /* convert zbuffer values from ushorts to uints */ + /* gather lower four ushorts */ + zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, + (vector unsigned int) *zbuf, + VEC_LITERAL(vector unsigned char, + ZERO, ZERO, 8, 9, ZERO, ZERO, 10, 11, + ZERO, ZERO, 12, 13, ZERO, ZERO, 14, 15)); + /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ + mask = spu_cmpgt(zbuf_ui4, zvals_ui4); + /* mask &= inMask */ + mask = spu_and(mask, inMask); + /* zbuf = mask ? zval : zbuf */ + zbuf_ui4 = spu_sel(zbuf_ui4, zvals_ui4, mask); + /* convert zbuffer values from uints back to ushorts, preserve lower 4 */ + *zbuf = (vector unsigned short) + spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, + VEC_LITERAL(vector unsigned char, + 16, 17, 18, 19, 20, 21, 22, 23, + 2, 3, 6, 7, 10, 11, 14, 15)); + } + else { + /* convert zbuffer values from ushorts to uints */ + /* gather upper four ushorts */ + zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, + (vector unsigned int) *zbuf, + VEC_LITERAL(vector unsigned char, + ZERO, ZERO, 0, 1, ZERO, ZERO, 2, 3, + ZERO, ZERO, 4, 5, ZERO, ZERO, 6, 7)); + /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ + mask = spu_cmpgt(zbuf_ui4, zvals_ui4); + /* mask &= inMask */ + mask = spu_and(mask, inMask); + /* zbuf = mask ? zval : zbuf */ + zbuf_ui4 = spu_sel(zbuf_ui4, zvals_ui4, mask); + /* convert zbuffer values from uints back to ushorts, preserve upper 4 */ + *zbuf = (vector unsigned short) + spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, + VEC_LITERAL(vector unsigned char, + 2, 3, 6, 7, 10, 11, 14, 15, + 24, 25, 26, 27, 28, 29, 30, 31)); + } + return mask; +#undef ZERO +} + + +/** + * As above, but Zbuffer values as 32-bit uints + */ +static INLINE vector unsigned int +spu_z32_test_less(vector float zvals, vector unsigned int *zbuf_ptr, + vector unsigned int inMask) +{ + vector unsigned int zvals_ui4, mask, zbuf = *zbuf_ptr; + + /* convert floats to uints in [0, 0xffffffff] */ + zvals_ui4 = spu_convtu(zvals, 32); + /* mask = (zbuf < zvals_ui4) ? ~0 : 0 */ + mask = spu_cmpgt(zbuf, zvals_ui4); + /* mask &= inMask */ + mask = spu_and(mask, inMask); + /* zbuf = mask ? zval : zbuf */ + *zbuf_ptr = spu_sel(zbuf, zvals_ui4, mask); + + return mask; +} + + +#endif /* SPU_ZTEST_H */ -- cgit v1.2.3 From c392cc8f1bcaaecc2cc723fc5550e5f6462602f3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 13:49:51 -0700 Subject: Cell: rename fields of the tile_t union --- src/mesa/pipe/cell/spu/spu_main.c | 8 ++++++++ src/mesa/pipe/cell/spu/spu_texture.c | 6 +++--- src/mesa/pipe/cell/spu/spu_tile.c | 4 ++-- src/mesa/pipe/cell/spu/spu_tile.h | 18 +++++------------- src/mesa/pipe/cell/spu/spu_tri.c | 8 ++++---- 5 files changed, 22 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index d6393048f5..7d6e910ad5 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -36,6 +36,7 @@ #include "spu_render.h" #include "spu_texture.h" #include "spu_tile.h" +//#include "spu_test.h" #include "spu_vertex_shader.h" #include "pipe/cell/common.h" #include "pipe/p_defines.h" @@ -495,6 +496,7 @@ one_time_init(void) } + /* In some versions of the SDK the SPE main takes 'unsigned long' as a * parameter. In others it takes 'unsigned long long'. Use a define to * select between the two. @@ -515,6 +517,8 @@ main(main_param_t speid, main_param_t argp) (void) speid; + ASSERT(sizeof(tile_t) == TILE_SIZE * TILE_SIZE * 4); + one_time_init(); if (Debug) @@ -528,6 +532,10 @@ main(main_param_t speid, main_param_t argp) 0 /* rid */); wait_on_mask( 1 << tag ); +#if 0 + if (spu.init.id==0) + spu_test_misc(); +#endif main_loop(); diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index 7a1ca097c0..c1dc6bfe90 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -97,10 +97,10 @@ get_tex_tile(uint i, uint j) spu.init.id, src, tex_tiles[pos].t32); #endif - ASSERT_ALIGN16(tex_tiles[pos].t32); + ASSERT_ALIGN16(tex_tiles[pos].ui); ASSERT_ALIGN16(src); - mfc_get(tex_tiles[pos].t32, /* dest */ + mfc_get(tex_tiles[pos].ui, /* dest */ (unsigned int) src, bytes_per_tile, /* size */ TAG_TEXTURE_TILE, @@ -134,6 +134,6 @@ sample_texture(float4 texcoord) uint i = (uint) (texcoord.f[0] * spu.texture.width) % spu.texture.width; uint j = (uint) (texcoord.f[1] * spu.texture.height) % spu.texture.height; uint pos = get_tex_tile(i, j); - uint texel = tex_tiles[pos].t32[j % TILE_SIZE][i % TILE_SIZE]; + uint texel = tex_tiles[pos].ui[j % TILE_SIZE][i % TILE_SIZE]; return texel; } diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c index aea4785bc2..fd65c2b49c 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.c +++ b/src/mesa/pipe/cell/spu/spu_tile.c @@ -56,7 +56,7 @@ get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf) printf("get_tile: dest: %p src: 0x%x size: %d\n", tile, (unsigned int) src, bytesPerTile); */ - mfc_get(tile->t32, /* dest in local memory */ + mfc_get(tile->ui, /* dest in local memory */ (unsigned int) src, /* src in main memory */ bytesPerTile, tag, @@ -82,7 +82,7 @@ put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf) spu.init.id, tile, (unsigned int) dst, bytesPerTile); */ - mfc_put((void *) tile->t32, /* src in local memory */ + mfc_put((void *) tile->ui, /* src in local memory */ (unsigned int) dst, /* dst in main memory */ bytesPerTile, tag, diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h index 4b1ef2a4c8..85a0d55807 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -40,8 +40,8 @@ typedef union { - ushort t16[TILE_SIZE][TILE_SIZE]; - uint t32[TILE_SIZE][TILE_SIZE]; + ushort us[TILE_SIZE][TILE_SIZE]; + uint ui[TILE_SIZE][TILE_SIZE]; vector unsigned short us8[TILE_SIZE/2][TILE_SIZE/4]; vector unsigned int ui4[TILE_SIZE/2][TILE_SIZE/2]; } tile_t; @@ -74,7 +74,7 @@ put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf); static INLINE void clear_c_tile(tile_t *ctile) { - memset32((uint*) ctile->t32, + memset32((uint*) ctile->ui, spu.fb.color_clear_value, TILE_SIZE * TILE_SIZE); } @@ -84,23 +84,15 @@ static INLINE void clear_z_tile(tile_t *ztile) { if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - memset16((ushort*) ztile->t16, + memset16((ushort*) ztile->us, spu.fb.depth_clear_value, TILE_SIZE * TILE_SIZE); } else { ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); -#if SIMD_Z - union fi z; - z.f = 1.0; - memset32((uint*) ztile->t32, - z.i,/*spu.fb.depth_clear_value,*/ - TILE_SIZE * TILE_SIZE); -#else - memset32((uint*) ztile->t32, + memset32((uint*) ztile->ui, spu.fb.depth_clear_value, TILE_SIZE * TILE_SIZE); -#endif } } diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index a26a4f098d..b04b6841c0 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -349,13 +349,13 @@ emit_quad( int x, int y, mask_t mask ) cur_tile_status_c = TILE_STATUS_DIRTY; if (spu_extract(mask, 0)) - ctile.t32[iy][ix] = colors[QUAD_TOP_LEFT]; + ctile.ui[iy][ix] = colors[QUAD_TOP_LEFT]; if (spu_extract(mask, 1)) - ctile.t32[iy][ix+1] = colors[QUAD_TOP_RIGHT]; + ctile.ui[iy][ix+1] = colors[QUAD_TOP_RIGHT]; if (spu_extract(mask, 2)) - ctile.t32[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; + ctile.ui[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; if (spu_extract(mask, 3)) - ctile.t32[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; + ctile.ui[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; #if 0 /* SIMD_Z with swizzled color buffer (someday) */ -- cgit v1.2.3 From 0e9a370ae2fa7a6d8bbc7d236e63dae1e3dcac37 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 14:02:22 -0700 Subject: Cell: move ztest before color interp/packing --- src/mesa/pipe/cell/spu/spu_tri.c | 43 ++++++++++++++++++++-------------------- 1 file changed, 21 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index b04b6841c0..ae8fd17cc6 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -316,24 +316,6 @@ emit_quad( int x, int y, mask_t mask ) setup.quad.mask = mask; sp->quad.first->run(sp->quad.first, &setup.quad); #else - /* Cell: "write" quad fragments to the tile by setting prim color */ - const int ix = x - setup.cliprect_minx; - const int iy = y - setup.cliprect_miny; - uint colors[4]; /* indexed by QUAD_x */ - - if (spu.texture.start) { - float4 texcoords[4]; - uint i; - eval_coeff(2, (float) x, (float) y, texcoords); - for (i = 0; i < 4; i++) { - colors[i] = sample_texture(texcoords[i]); - } - } - else { - float4 fcolors[4]; - eval_coeff(1, (float) x, (float) y, fcolors); - pack_colors(colors, fcolors); - } if (spu.depth_stencil.depth.enabled) { mask = do_depth_test(x, y, mask); @@ -341,6 +323,23 @@ emit_quad( int x, int y, mask_t mask ) /* If any bits in mask are set... */ if (spu_extract(spu_orx(mask), 0)) { + const int ix = x - setup.cliprect_minx; + const int iy = y - setup.cliprect_miny; + uint colors[4]; /* indexed by QUAD_x */ + + if (spu.texture.start) { + float4 texcoords[4]; + uint i; + eval_coeff(2, (float) x, (float) y, texcoords); + for (i = 0; i < 4; i++) { + colors[i] = sample_texture(texcoords[i]); + } + } + else { + float4 fcolors[4]; + eval_coeff(1, (float) x, (float) y, fcolors); + pack_colors(colors, fcolors); + } if (cur_tile_status_c == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ @@ -348,6 +347,7 @@ emit_quad( int x, int y, mask_t mask ) } cur_tile_status_c = TILE_STATUS_DIRTY; +#if 1 if (spu_extract(mask, 0)) ctile.ui[iy][ix] = colors[QUAD_TOP_LEFT]; if (spu_extract(mask, 1)) @@ -356,11 +356,10 @@ emit_quad( int x, int y, mask_t mask ) ctile.ui[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; if (spu_extract(mask, 3)) ctile.ui[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; - -#if 0 +#else /* SIMD_Z with swizzled color buffer (someday) */ - vector float icolors = *((vector float *) &colors); - ctile.f4[iy/2][ix/2].v = spu_sel(ctile.f4[iy/2][ix/2].v, icolors, mask); + vector unsigned int uicolors = *((vector unsigned int *) &colors); + ctile.ui4[iy/2][ix/2] = spu_sel(ctile.ui4[iy/2][ix/2], uicolors, mask); #endif } -- cgit v1.2.3 From b1a472bfb7df5ba273574e1799c5b8e85ca5f2d9 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 15:20:07 -0700 Subject: Cell: remove commands from top-level while loop which should only appear in batch buffers --- src/mesa/pipe/cell/spu/spu_main.c | 20 ++++---------------- 1 file changed, 4 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 7d6e910ad5..1760de02b7 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -447,34 +447,22 @@ main_loop(void) 0 /* rid */); wait_on_mask( 1 << tag ); + /* + * NOTE: most commands should be contained in a batch buffer + */ + switch (opcode & CELL_CMD_OPCODE_MASK) { case CELL_CMD_EXIT: if (Debug) printf("SPU %u: EXIT\n", spu.init.id); exitFlag = 1; break; - case CELL_CMD_STATE_FRAMEBUFFER: - cmd_state_framebuffer(&cmd.fb); - break; - case CELL_CMD_CLEAR_SURFACE: - cmd_clear_surface(&cmd.clear); - break; - case CELL_CMD_RENDER: - { - uint pos_incr; - cmd_render(&cmd.render, &pos_incr); - assert(pos_incr == 0); - } - break; case CELL_CMD_VS_EXECUTE: spu_execute_vertex_shader(&draw, &cmd.vs); break; case CELL_CMD_BATCH: cmd_batch(opcode); break; - case CELL_CMD_FINISH: - cmd_finish(); - break; default: printf("Bad opcode!\n"); } -- cgit v1.2.3 From 17305489f0d2a0681d4c0d4952957af517019ab6 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 15:24:00 -0700 Subject: Cell: deprecate some use of struct cell_command - it should go away completely Also, remove ALIGN16_ATTRIB from structs that no longer need it. --- src/mesa/pipe/cell/common.h | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 05aeed83ab..7e193f31be 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -105,7 +105,7 @@ struct cell_command_framebuffer int width, height; void *color_start, *depth_start; enum pipe_format color_format, depth_format; -} ALIGN16_ATTRIB; +}; /** @@ -116,7 +116,7 @@ struct cell_command_clear_surface uint opcode; uint surface; /**< Temporary: 0=color, 1=Z */ uint value; -} ALIGN16_ATTRIB; +}; /** @@ -173,7 +173,7 @@ struct cell_command_render uint dummy3; uint min_index; boolean inline_verts; -} ALIGN16_ATTRIB; +}; struct cell_command_release_verts @@ -191,11 +191,14 @@ struct cell_command_texture /** XXX unions don't seem to work */ +/* XXX this should go away; all commands should be placed in batch buffers */ struct cell_command { +#if 0 struct cell_command_framebuffer fb; struct cell_command_clear_surface clear; struct cell_command_render render; +#endif struct cell_command_vs vs; } ALIGN16_ATTRIB; -- cgit v1.2.3 From 42201d7574ebb1582563988820c248680081c42f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 15:33:53 -0700 Subject: Cell: rename/move global vars Put tile-related globals into spu_global struct. Rename c/ztile fields to be more consistant. --- src/mesa/pipe/cell/spu/spu_main.c | 28 +++++++++++++------------- src/mesa/pipe/cell/spu/spu_main.h | 32 +++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_render.c | 40 ++++++++++++++++++------------------- src/mesa/pipe/cell/spu/spu_tile.c | 11 +--------- src/mesa/pipe/cell/spu/spu_tile.h | 27 ------------------------- src/mesa/pipe/cell/spu/spu_tri.c | 38 +++++++++++++++++------------------ 6 files changed, 86 insertions(+), 90 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 1760de02b7..8e3987f6ef 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -92,24 +92,24 @@ really_clear_tiles(uint surfaceIndex) uint i; if (surfaceIndex == 0) { - clear_c_tile(&ctile); + clear_c_tile(&spu.ctile); for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { uint tx = i % spu.fb.width_tiles; uint ty = i / spu.fb.width_tiles; - if (tile_status[ty][tx] == TILE_STATUS_CLEAR) { - put_tile(tx, ty, &ctile, TAG_SURFACE_CLEAR, 0); + if (spu.ctile_status[ty][tx] == TILE_STATUS_CLEAR) { + put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); } } } else { - clear_z_tile(&ztile); + clear_z_tile(&spu.ztile); for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { uint tx = i % spu.fb.width_tiles; uint ty = i / spu.fb.width_tiles; - if (tile_status_z[ty][tx] == TILE_STATUS_CLEAR) - put_tile(tx, ty, &ctile, TAG_SURFACE_CLEAR, 1); + if (spu.ztile_status[ty][tx] == TILE_STATUS_CLEAR) + put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 1); } } @@ -133,11 +133,11 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) #if CLEAR_OPT /* set all tile's status to CLEAR */ if (clear->surface == 0) { - memset(tile_status, TILE_STATUS_CLEAR, sizeof(tile_status)); + memset(spu.ctile_status, TILE_STATUS_CLEAR, sizeof(spu.ctile_status)); spu.fb.color_clear_value = clear->value; } else { - memset(tile_status_z, TILE_STATUS_CLEAR, sizeof(tile_status_z)); + memset(spu.ztile_status, TILE_STATUS_CLEAR, sizeof(spu.ztile_status)); spu.fb.depth_clear_value = clear->value; } return; @@ -145,11 +145,11 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) if (clear->surface == 0) { spu.fb.color_clear_value = clear->value; - clear_c_tile(&ctile); + clear_c_tile(&spu.ctile); } else { spu.fb.depth_clear_value = clear->value; - clear_z_tile(&ztile); + clear_z_tile(&spu.ztile); } /* @@ -161,9 +161,9 @@ cmd_clear_surface(const struct cell_command_clear_surface *clear) uint tx = i % spu.fb.width_tiles; uint ty = i / spu.fb.width_tiles; if (clear->surface == 0) - put_tile(tx, ty, &ctile, TAG_SURFACE_CLEAR, 0); + put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); else - put_tile(tx, ty, &ztile, TAG_SURFACE_CLEAR, 1); + put_tile(tx, ty, &spu.ztile, TAG_SURFACE_CLEAR, 1); /* XXX we don't want this here, but it fixes bad tile results */ } @@ -478,8 +478,8 @@ main_loop(void) static void one_time_init(void) { - memset(tile_status, TILE_STATUS_DEFINED, sizeof(tile_status)); - memset(tile_status_z, TILE_STATUS_DEFINED, sizeof(tile_status_z)); + memset(spu.ctile_status, TILE_STATUS_DEFINED, sizeof(spu.ctile_status)); + memset(spu.ztile_status, TILE_STATUS_DEFINED, sizeof(spu.ztile_status)); invalidate_tex_cache(); } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 8be5268f52..cce5e70802 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -36,6 +36,11 @@ #include "pipe/p_state.h" + +#define MAX_WIDTH 1024 +#define MAX_HEIGHT 1024 + + typedef union { vector float v; @@ -43,6 +48,21 @@ typedef union } float4; +typedef union { + ushort us[TILE_SIZE][TILE_SIZE]; + uint ui[TILE_SIZE][TILE_SIZE]; + vector unsigned short us8[TILE_SIZE/2][TILE_SIZE/4]; + vector unsigned int ui4[TILE_SIZE/2][TILE_SIZE/2]; +} tile_t; + + +#define TILE_STATUS_CLEAR 1 +#define TILE_STATUS_DEFINED 2 /**< defined in FB, but not in local store */ +#define TILE_STATUS_CLEAN 3 /**< in local store, but not changed */ +#define TILE_STATUS_DIRTY 4 /**< modified locally, but not put back yet */ +#define TILE_STATUS_GETTING 5 /**< mfc_get() called but not yet arrived */ + + struct spu_framebuffer { void *color_start; /**< addr of color surface in main memory */ void *depth_start; /**< addr of depth surface in main memory */ @@ -75,6 +95,18 @@ struct spu_global /* XXX more state to come */ + + /** current color and Z tiles */ + tile_t ctile ALIGN16_ATTRIB; + tile_t ztile ALIGN16_ATTRIB; + + /** Current tiles' status */ + ubyte cur_ctile_status, cur_ztile_status; + + /** Status of all tiles in framebuffer */ + ubyte ctile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + ubyte ztile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/spu_render.c b/src/mesa/pipe/cell/spu/spu_render.c index ca54a103bd..ab711d67fe 100644 --- a/src/mesa/pipe/cell/spu/spu_render.c +++ b/src/mesa/pipe/cell/spu/spu_render.c @@ -95,15 +95,15 @@ static INLINE void get_cz_tiles(uint tx, uint ty) { if (spu.depth_stencil.depth.enabled) { - if (cur_tile_status_z != TILE_STATUS_CLEAR) { - get_tile(tx, ty, &ztile, TAG_READ_TILE_Z, 1); - cur_tile_status_z = TILE_STATUS_GETTING; + if (spu.cur_ztile_status != TILE_STATUS_CLEAR) { + get_tile(tx, ty, &spu.ztile, TAG_READ_TILE_Z, 1); + spu.cur_ztile_status = TILE_STATUS_GETTING; } } - if (cur_tile_status_c != TILE_STATUS_CLEAR) { - get_tile(tx, ty, &ctile, TAG_READ_TILE_COLOR, 0); - cur_tile_status_c = TILE_STATUS_GETTING; + if (spu.cur_ctile_status != TILE_STATUS_CLEAR) { + get_tile(tx, ty, &spu.ctile, TAG_READ_TILE_COLOR, 0); + spu.cur_ctile_status = TILE_STATUS_GETTING; } } @@ -114,24 +114,24 @@ get_cz_tiles(uint tx, uint ty) static INLINE void put_cz_tiles(uint tx, uint ty) { - if (cur_tile_status_z == TILE_STATUS_DIRTY) { + if (spu.cur_ztile_status == TILE_STATUS_DIRTY) { /* tile was modified and needs to be written back */ - put_tile(tx, ty, &ztile, TAG_WRITE_TILE_Z, 1); - cur_tile_status_z = TILE_STATUS_DEFINED; + put_tile(tx, ty, &spu.ztile, TAG_WRITE_TILE_Z, 1); + spu.cur_ztile_status = TILE_STATUS_DEFINED; } - else if (cur_tile_status_z == TILE_STATUS_GETTING) { + else if (spu.cur_ztile_status == TILE_STATUS_GETTING) { /* tile was never used */ - cur_tile_status_z = TILE_STATUS_DEFINED; + spu.cur_ztile_status = TILE_STATUS_DEFINED; } - if (cur_tile_status_c == TILE_STATUS_DIRTY) { + if (spu.cur_ctile_status == TILE_STATUS_DIRTY) { /* tile was modified and needs to be written back */ - put_tile(tx, ty, &ctile, TAG_WRITE_TILE_COLOR, 0); - cur_tile_status_c = TILE_STATUS_DEFINED; + put_tile(tx, ty, &spu.ctile, TAG_WRITE_TILE_COLOR, 0); + spu.cur_ctile_status = TILE_STATUS_DEFINED; } - else if (cur_tile_status_c == TILE_STATUS_GETTING) { + else if (spu.cur_ctile_status == TILE_STATUS_GETTING) { /* tile was never used */ - cur_tile_status_c = TILE_STATUS_DEFINED; + spu.cur_ctile_status = TILE_STATUS_DEFINED; } } @@ -250,8 +250,8 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) if (!my_tile(tx, ty)) continue; - cur_tile_status_c = tile_status[ty][tx]; - cur_tile_status_z = tile_status_z[ty][tx]; + spu.cur_ctile_status = spu.ctile_status[ty][tx]; + spu.cur_ztile_status = spu.ztile_status[ty][tx]; get_cz_tiles(tx, ty); @@ -275,8 +275,8 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) wait_put_cz_tiles(); /* XXX seems unnecessary... */ - tile_status[ty][tx] = cur_tile_status_c; - tile_status_z[ty][tx] = cur_tile_status_z; + spu.ctile_status[ty][tx] = spu.cur_ctile_status; + spu.ztile_status[ty][tx] = spu.cur_ztile_status; } if (Debug) diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c index fd65c2b49c..12dc246328 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.c +++ b/src/mesa/pipe/cell/spu/spu_tile.c @@ -28,16 +28,7 @@ #include "spu_tile.h" - - - -tile_t ctile ALIGN16_ATTRIB; -tile_t ztile ALIGN16_ATTRIB; - -ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; -ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - -ubyte cur_tile_status_c, cur_tile_status_z; +#include "spu_main.h" void diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h index 85a0d55807..e53340a55a 100644 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ b/src/mesa/pipe/cell/spu/spu_tile.h @@ -35,33 +35,6 @@ #include "pipe/cell/common.h" -#define MAX_WIDTH 1024 -#define MAX_HEIGHT 1024 - - -typedef union { - ushort us[TILE_SIZE][TILE_SIZE]; - uint ui[TILE_SIZE][TILE_SIZE]; - vector unsigned short us8[TILE_SIZE/2][TILE_SIZE/4]; - vector unsigned int ui4[TILE_SIZE/2][TILE_SIZE/2]; -} tile_t; - - -extern tile_t ctile ALIGN16_ATTRIB; -extern tile_t ztile ALIGN16_ATTRIB; - - -#define TILE_STATUS_CLEAR 1 -#define TILE_STATUS_DEFINED 2 /**< defined in FB, but not in local store */ -#define TILE_STATUS_CLEAN 3 /**< in local store, but not changed */ -#define TILE_STATUS_DIRTY 4 /**< modified locally, but not put back yet */ -#define TILE_STATUS_GETTING 5 /**< mfc_get() called but not yet arrived */ - -extern ubyte tile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; -extern ubyte tile_status_z[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - -extern ubyte cur_tile_status_c, cur_tile_status_z; - void get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf); diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index ae8fd17cc6..6f61a3d816 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -283,21 +283,21 @@ do_depth_test(int x, int y, mask_t quadmask) zvals.v = eval_z((float) x, (float) y); - if (cur_tile_status_c == TILE_STATUS_CLEAR) { + if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ - clear_z_tile(&ztile); - cur_tile_status_z = TILE_STATUS_DIRTY; + clear_z_tile(&spu.ztile); + spu.cur_ztile_status = TILE_STATUS_DIRTY; } if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { int ix = (x - setup.cliprect_minx) / 4; int iy = (y - setup.cliprect_miny) / 2; - mask = spu_z16_test_less(zvals.v, &ztile.us8[iy][ix], x>>1, quadmask); + mask = spu_z16_test_less(zvals.v, &spu.ztile.us8[iy][ix], x>>1, quadmask); } else { int ix = (x - setup.cliprect_minx) / 2; int iy = (y - setup.cliprect_miny) / 2; - mask = spu_z32_test_less(zvals.v, &ztile.ui4[iy][ix], quadmask); + mask = spu_z32_test_less(zvals.v, &spu.ztile.ui4[iy][ix], quadmask); } return mask; } @@ -341,25 +341,25 @@ emit_quad( int x, int y, mask_t mask ) pack_colors(colors, fcolors); } - if (cur_tile_status_c == TILE_STATUS_CLEAR) { + if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { /* now, _really_ clear the tile */ - clear_c_tile(&ctile); + clear_c_tile(&spu.ctile); } - cur_tile_status_c = TILE_STATUS_DIRTY; + spu.cur_ctile_status = TILE_STATUS_DIRTY; #if 1 if (spu_extract(mask, 0)) - ctile.ui[iy][ix] = colors[QUAD_TOP_LEFT]; + spu.ctile.ui[iy][ix] = colors[QUAD_TOP_LEFT]; if (spu_extract(mask, 1)) - ctile.ui[iy][ix+1] = colors[QUAD_TOP_RIGHT]; + spu.ctile.ui[iy][ix+1] = colors[QUAD_TOP_RIGHT]; if (spu_extract(mask, 2)) - ctile.ui[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; + spu.ctile.ui[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; if (spu_extract(mask, 3)) - ctile.ui[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; + spu.ctile.ui[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; #else /* SIMD_Z with swizzled color buffer (someday) */ vector unsigned int uicolors = *((vector unsigned int *) &colors); - ctile.ui4[iy/2][ix/2] = spu_sel(ctile.ui4[iy/2][ix/2], uicolors, mask); + spu.ctile.ui4[iy/2][ix/2] = spu_sel(spu.ctile.ui4[iy/2][ix/2], uicolors, mask); #endif } @@ -846,21 +846,21 @@ tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) /* init_constant_attribs( setup ); */ - if (cur_tile_status_c == TILE_STATUS_GETTING) { + if (spu.cur_ctile_status == TILE_STATUS_GETTING) { /* wait for mfc_get() to complete */ wait_on_mask(1 << TAG_READ_TILE_COLOR); - cur_tile_status_c = TILE_STATUS_CLEAN; + spu.cur_ctile_status = TILE_STATUS_CLEAN; } - ASSERT(cur_tile_status_c != TILE_STATUS_DEFINED); + ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED); if (spu.depth_stencil.depth.enabled) { - if (cur_tile_status_z == TILE_STATUS_GETTING) { + if (spu.cur_ztile_status == TILE_STATUS_GETTING) { /* wait for mfc_get() to complete */ wait_on_mask(1 << TAG_READ_TILE_Z); - cur_tile_status_z = TILE_STATUS_CLEAN; + spu.cur_ztile_status = TILE_STATUS_CLEAN; } - ASSERT(cur_tile_status_z != TILE_STATUS_DEFINED); + ASSERT(spu.cur_ztile_status != TILE_STATUS_DEFINED); } -- cgit v1.2.3 From d7c2eb0df47bd79291172727539b99331a3c6724 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 15:45:02 -0700 Subject: Cell: New color packing functions (A8R8G8B8 and B8G8R8A8) --- src/mesa/pipe/cell/spu/spu_colorpack.h | 60 ++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_tri.c | 22 +++++-------- 2 files changed, 69 insertions(+), 13 deletions(-) create mode 100644 src/mesa/pipe/cell/spu/spu_colorpack.h (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_colorpack.h b/src/mesa/pipe/cell/spu/spu_colorpack.h new file mode 100644 index 0000000000..56709bd9f3 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_colorpack.h @@ -0,0 +1,60 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + + +#ifndef SPU_COLORPACK_H +#define SPU_COLORPACK_H + + +#include +#include + + +static INLINE unsigned int +spu_pack_A8R8G8B8(vector float rgba) +{ + vector unsigned int out = spu_convtu(rgba, 32); + out = spu_shuffle(out, out, VEC_LITERAL(vector unsigned char, + 12, 0, 4, 8, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0)); + return spu_extract(out, 0); +} + + +static INLINE unsigned int +spu_pack_B8G8R8A8(vector float rgba) +{ + vector unsigned int out = spu_convtu(rgba, 32); + out = spu_shuffle(out, out, VEC_LITERAL(vector unsigned char, + 8, 4, 0, 12, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0)); + return spu_extract(out, 0); +} + + +#endif /* SPU_COLORPACK_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 6f61a3d816..c82ca51000 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -29,11 +29,10 @@ * Triangle rendering within a tile. */ -#include - #include "pipe/p_compiler.h" #include "pipe/p_format.h" #include "pipe/p_util.h" +#include "spu_colorpack.h" #include "spu_main.h" #include "spu_texture.h" #include "spu_tile.h" @@ -253,21 +252,18 @@ eval_z(float x, float y) static INLINE void pack_colors(uint uicolors[4], const float4 fcolors[4]) { - /* XXX grab the code for _pack_rgba8() and use the shuffle - * command to do the swizzling seen here. - */ switch (spu.fb.color_format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - uicolors[0] = _pack_rgba8(fcolors[0].f[3], fcolors[0].f[0], fcolors[0].f[1], fcolors[0].f[2]); - uicolors[1] = _pack_rgba8(fcolors[1].f[3], fcolors[1].f[0], fcolors[1].f[1], fcolors[1].f[2]); - uicolors[2] = _pack_rgba8(fcolors[2].f[3], fcolors[2].f[0], fcolors[2].f[1], fcolors[2].f[2]); - uicolors[3] = _pack_rgba8(fcolors[3].f[3], fcolors[0].f[0], fcolors[3].f[1], fcolors[3].f[2]); + uicolors[0] = spu_pack_A8R8G8B8(fcolors[0].v); + uicolors[1] = spu_pack_A8R8G8B8(fcolors[1].v); + uicolors[2] = spu_pack_A8R8G8B8(fcolors[2].v); + uicolors[3] = spu_pack_A8R8G8B8(fcolors[3].v); break; case PIPE_FORMAT_B8G8R8A8_UNORM: - uicolors[0] = _pack_rgba8(fcolors[0].f[2], fcolors[0].f[1], fcolors[0].f[0], fcolors[0].f[3]); - uicolors[1] = _pack_rgba8(fcolors[1].f[2], fcolors[1].f[1], fcolors[1].f[0], fcolors[1].f[3]); - uicolors[2] = _pack_rgba8(fcolors[2].f[2], fcolors[2].f[1], fcolors[2].f[0], fcolors[2].f[3]); - uicolors[3] = _pack_rgba8(fcolors[3].f[2], fcolors[3].f[1], fcolors[3].f[0], fcolors[3].f[3]); + uicolors[0] = spu_pack_B8G8R8A8(fcolors[0].v); + uicolors[1] = spu_pack_B8G8R8A8(fcolors[1].v); + uicolors[2] = spu_pack_B8G8R8A8(fcolors[2].v); + uicolors[3] = spu_pack_B8G8R8A8(fcolors[3].v); break; default: ASSERT(0); -- cgit v1.2.3 From 8bd566a9cb8bb01ef5ce9c526047bafc0fbf0aef Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 16:25:42 -0700 Subject: Cell: use global color_shuffle to remove a switch stmnt --- src/mesa/pipe/cell/spu/Makefile | 2 + src/mesa/pipe/cell/spu/spu_colorpack.h | 9 ++++ src/mesa/pipe/cell/spu/spu_main.c | 12 ++++++ src/mesa/pipe/cell/spu/spu_main.h | 3 ++ src/mesa/pipe/cell/spu/spu_tri.c | 76 ++++++++++++++-------------------- 5 files changed, 56 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index 2d031bfbc6..91a631b699 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -8,6 +8,8 @@ TOP = ../../../../.. include $(TOP)/configs/linux-cell +OPT_FLAGS=-g +OPT_FLAGS=-O3 PROG = g3d PROG_SPU = $(PROG)_spu diff --git a/src/mesa/pipe/cell/spu/spu_colorpack.h b/src/mesa/pipe/cell/spu/spu_colorpack.h index 56709bd9f3..9977a6ece0 100644 --- a/src/mesa/pipe/cell/spu/spu_colorpack.h +++ b/src/mesa/pipe/cell/spu/spu_colorpack.h @@ -57,4 +57,13 @@ spu_pack_B8G8R8A8(vector float rgba) } +static INLINE unsigned int +spu_pack_color_shuffle(vector float rgba, vector unsigned char shuffle) +{ + vector unsigned int out = spu_convtu(rgba, 32); + out = spu_shuffle(out, out, shuffle); + return spu_extract(out, 0); +} + + #endif /* SPU_COLORPACK_H */ diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 8e3987f6ef..ba4d180cc0 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -31,6 +31,7 @@ #include #include +#include #include "spu_main.h" #include "spu_render.h" @@ -217,6 +218,17 @@ cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) spu.fb.zsize = 2; else spu.fb.zsize = 0; + + if (spu.fb.color_format == PIPE_FORMAT_A8R8G8B8_UNORM) + spu.color_shuffle = VEC_LITERAL(vector unsigned char, + 12, 0, 4, 8, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0); + else if (spu.fb.color_format == PIPE_FORMAT_B8G8R8A8_UNORM) + spu.color_shuffle = VEC_LITERAL(vector unsigned char, + 8, 4, 0, 12, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0); + else + ASSERT(0); } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index cce5e70802..7a12715b0b 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -107,6 +107,9 @@ struct spu_global ubyte ctile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; ubyte ztile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + + /** for converting RGBA to PIPE_FORMAT_x colors */ + vector unsigned char color_shuffle; } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index c82ca51000..165e41a781 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -249,28 +249,6 @@ eval_z(float x, float y) } -static INLINE void -pack_colors(uint uicolors[4], const float4 fcolors[4]) -{ - switch (spu.fb.color_format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - uicolors[0] = spu_pack_A8R8G8B8(fcolors[0].v); - uicolors[1] = spu_pack_A8R8G8B8(fcolors[1].v); - uicolors[2] = spu_pack_A8R8G8B8(fcolors[2].v); - uicolors[3] = spu_pack_A8R8G8B8(fcolors[3].v); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - uicolors[0] = spu_pack_B8G8R8A8(fcolors[0].v); - uicolors[1] = spu_pack_B8G8R8A8(fcolors[1].v); - uicolors[2] = spu_pack_B8G8R8A8(fcolors[2].v); - uicolors[3] = spu_pack_B8G8R8A8(fcolors[3].v); - break; - default: - ASSERT(0); - } -} - - static INLINE mask_t do_depth_test(int x, int y, mask_t quadmask) { @@ -321,38 +299,44 @@ emit_quad( int x, int y, mask_t mask ) if (spu_extract(spu_orx(mask), 0)) { const int ix = x - setup.cliprect_minx; const int iy = y - setup.cliprect_miny; - uint colors[4]; /* indexed by QUAD_x */ + + if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { + /* now, _really_ clear the tile */ + clear_c_tile(&spu.ctile); + } + spu.cur_ctile_status = TILE_STATUS_DIRTY; if (spu.texture.start) { + /* texture mapping */ float4 texcoords[4]; - uint i; eval_coeff(2, (float) x, (float) y, texcoords); - for (i = 0; i < 4; i++) { - colors[i] = sample_texture(texcoords[i]); - } + + if (spu_extract(mask, 0)) + spu.ctile.ui[iy][ix] = sample_texture(texcoords[0]); + if (spu_extract(mask, 1)) + spu.ctile.ui[iy][ix+1] = sample_texture(texcoords[1]); + if (spu_extract(mask, 2)) + spu.ctile.ui[iy+1][ix] = sample_texture(texcoords[2]); + if (spu_extract(mask, 3)) + spu.ctile.ui[iy+1][ix+1] = sample_texture(texcoords[3]); } else { - float4 fcolors[4]; - eval_coeff(1, (float) x, (float) y, fcolors); - pack_colors(colors, fcolors); + /* simple shading */ + const vector unsigned char shuffle = spu.color_shuffle; + float4 colors[4]; + eval_coeff(1, (float) x, (float) y, colors); + + if (spu_extract(mask, 0)) + spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0].v, shuffle); + if (spu_extract(mask, 1)) + spu.ctile.ui[iy][ix+1] = spu_pack_color_shuffle(colors[1].v, shuffle); + if (spu_extract(mask, 2)) + spu.ctile.ui[iy+1][ix] = spu_pack_color_shuffle(colors[2].v, shuffle); + if (spu_extract(mask, 3)) + spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(colors[3].v, shuffle); } - if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { - /* now, _really_ clear the tile */ - clear_c_tile(&spu.ctile); - } - spu.cur_ctile_status = TILE_STATUS_DIRTY; - -#if 1 - if (spu_extract(mask, 0)) - spu.ctile.ui[iy][ix] = colors[QUAD_TOP_LEFT]; - if (spu_extract(mask, 1)) - spu.ctile.ui[iy][ix+1] = colors[QUAD_TOP_RIGHT]; - if (spu_extract(mask, 2)) - spu.ctile.ui[iy+1][ix] = colors[QUAD_BOTTOM_LEFT]; - if (spu_extract(mask, 3)) - spu.ctile.ui[iy+1][ix+1] = colors[QUAD_BOTTOM_RIGHT]; -#else +#if 0 /* SIMD_Z with swizzled color buffer (someday) */ vector unsigned int uicolors = *((vector unsigned int *) &colors); spu.ctile.ui4[iy/2][ix/2] = spu_sel(spu.ctile.ui4[iy/2][ix/2], uicolors, mask); -- cgit v1.2.3 From 44d32693562e2fb83572bd10e4d489a7cb6f74f3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 16:42:09 -0700 Subject: Cell: move some tile get/clear code Also, we weren't marking the ztile as dirty after ztesting, fixes gears glitches. --- src/mesa/pipe/cell/spu/spu_tri.c | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 165e41a781..f0758c42e7 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -257,12 +257,6 @@ do_depth_test(int x, int y, mask_t quadmask) zvals.v = eval_z((float) x, (float) y); - if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { - /* now, _really_ clear the tile */ - clear_z_tile(&spu.ztile); - spu.cur_ztile_status = TILE_STATUS_DIRTY; - } - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { int ix = (x - setup.cliprect_minx) / 4; int iy = (y - setup.cliprect_miny) / 2; @@ -273,6 +267,10 @@ do_depth_test(int x, int y, mask_t quadmask) int iy = (y - setup.cliprect_miny) / 2; mask = spu_z32_test_less(zvals.v, &spu.ztile.ui4[iy][ix], quadmask); } + + if (spu_extract(spu_orx(mask), 0)) + spu.cur_ztile_status = TILE_STATUS_DIRTY; + return mask; } @@ -300,10 +298,6 @@ emit_quad( int x, int y, mask_t mask ) const int ix = x - setup.cliprect_minx; const int iy = y - setup.cliprect_miny; - if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { - /* now, _really_ clear the tile */ - clear_c_tile(&spu.ctile); - } spu.cur_ctile_status = TILE_STATUS_DIRTY; if (spu.texture.start) { @@ -408,6 +402,18 @@ static void flush_spans( void ) return; } + + /* _really_ clear tiles now if needed */ + if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { + clear_c_tile(&spu.ctile); + spu.cur_ctile_status = TILE_STATUS_DIRTY; + } + if (spu.depth_stencil.depth.enabled && + spu.cur_ztile_status == TILE_STATUS_CLEAR) { + clear_z_tile(&spu.ztile); + spu.cur_ztile_status = TILE_STATUS_DIRTY; + } + /* XXX this loop could be moved into the above switch cases and * calculate_mask() could be simplified a bit... */ @@ -831,7 +837,6 @@ tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) wait_on_mask(1 << TAG_READ_TILE_COLOR); spu.cur_ctile_status = TILE_STATUS_CLEAN; } - ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED); if (spu.depth_stencil.depth.enabled) { -- cgit v1.2.3 From aa761b160520479efcf09d12ae4a161fc2f872f7 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 1 Feb 2008 16:54:46 -0700 Subject: Cell: comment about emit_quad() mask --- src/mesa/pipe/cell/spu/spu_tri.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index f0758c42e7..83bb247b22 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -277,6 +277,9 @@ do_depth_test(int x, int y, mask_t quadmask) /** * Emit a quad (pass to next stage). No clipping is done. + * Note: about 1/5 to 1/7 of the time, mask is zero and this function + * should be skipped. But adding the test for that slows things down + * overall. */ static INLINE void emit_quad( int x, int y, mask_t mask ) -- cgit v1.2.3 From 69cc19751dd0122116cab03d808d5a1f5d0ade84 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 08:45:33 -0700 Subject: Cell: insert some draw_flush() calls --- src/mesa/pipe/cell/ppu/cell_state_blend.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_blend.c b/src/mesa/pipe/cell/ppu/cell_state_blend.c index 34ae0128ea..2c19aa3971 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_blend.c +++ b/src/mesa/pipe/cell/ppu/cell_state_blend.c @@ -29,6 +29,7 @@ */ #include "pipe/p_util.h" +#include "pipe/draw/draw_context.h" #include "cell_context.h" #include "cell_state.h" @@ -49,6 +50,8 @@ cell_bind_blend_state(struct pipe_context *pipe, void *blend) { struct cell_context *cell = cell_context(pipe); + draw_flush(cell->draw); + cell->blend = (const struct pipe_blend_state *)blend; cell->dirty |= CELL_NEW_BLEND; @@ -68,6 +71,8 @@ cell_set_blend_color(struct pipe_context *pipe, { struct cell_context *cell = cell_context(pipe); + draw_flush(cell->draw); + cell->blend_color = *blend_color; cell->dirty |= CELL_NEW_BLEND; @@ -93,6 +98,8 @@ cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, { struct cell_context *cell = cell_context(pipe); + draw_flush(cell->draw); + cell->depth_stencil = (const struct pipe_depth_stencil_alpha_state *) depth_stencil; -- cgit v1.2.3 From 6023311c7ce336f727d7aa6d5266e88a55b88d36 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 08:46:44 -0700 Subject: Cell: clamp txmax, tymax in tile_bounding_box() Also, added some debug printfs --- src/mesa/pipe/cell/spu/spu_render.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_render.c b/src/mesa/pipe/cell/spu/spu_render.c index ab711d67fe..e8705eeeba 100644 --- a/src/mesa/pipe/cell/spu/spu_render.c +++ b/src/mesa/pipe/cell/spu/spu_render.c @@ -65,6 +65,10 @@ tile_bounding_box(const struct cell_command_render *render, *tymin = (uint) render->ymin / TILE_SIZE; txmax = (uint) render->xmax / TILE_SIZE; tymax = (uint) render->ymax / TILE_SIZE; + if (txmax >= spu.fb.width_tiles) + txmax = spu.fb.width_tiles-1; + if (tymax >= spu.fb.height_tiles) + tymax = spu.fb.height_tiles-1; *box_width_tiles = txmax - *txmin + 1; box_height_tiles = tymax - *tymin + 1; *box_num_tiles = *box_width_tiles * box_height_tiles; @@ -96,12 +100,14 @@ get_cz_tiles(uint tx, uint ty) { if (spu.depth_stencil.depth.enabled) { if (spu.cur_ztile_status != TILE_STATUS_CLEAR) { + //printf("SPU %u: getting Z tile %u, %u\n", spu.init.id, tx, ty); get_tile(tx, ty, &spu.ztile, TAG_READ_TILE_Z, 1); spu.cur_ztile_status = TILE_STATUS_GETTING; } } if (spu.cur_ctile_status != TILE_STATUS_CLEAR) { + //printf("SPU %u: getting C tile %u, %u\n", spu.init.id, tx, ty); get_tile(tx, ty, &spu.ctile, TAG_READ_TILE_COLOR, 0); spu.cur_ctile_status = TILE_STATUS_GETTING; } @@ -116,22 +122,26 @@ put_cz_tiles(uint tx, uint ty) { if (spu.cur_ztile_status == TILE_STATUS_DIRTY) { /* tile was modified and needs to be written back */ + //printf("SPU %u: put dirty Z tile %u, %u\n", spu.init.id, tx, ty); put_tile(tx, ty, &spu.ztile, TAG_WRITE_TILE_Z, 1); spu.cur_ztile_status = TILE_STATUS_DEFINED; } else if (spu.cur_ztile_status == TILE_STATUS_GETTING) { /* tile was never used */ spu.cur_ztile_status = TILE_STATUS_DEFINED; + //printf("SPU %u: put getting Z tile %u, %u\n", spu.init.id, tx, ty); } if (spu.cur_ctile_status == TILE_STATUS_DIRTY) { /* tile was modified and needs to be written back */ + //printf("SPU %u: put dirty C tile %u, %u\n", spu.init.id, tx, ty); put_tile(tx, ty, &spu.ctile, TAG_WRITE_TILE_COLOR, 0); spu.cur_ctile_status = TILE_STATUS_DEFINED; } else if (spu.cur_ctile_status == TILE_STATUS_GETTING) { /* tile was never used */ spu.cur_ctile_status = TILE_STATUS_DEFINED; + //printf("SPU %u: put getting C tile %u, %u\n", spu.init.id, tx, ty); } } -- cgit v1.2.3 From e967a5c746f340a76b27181b4ead1035101cece3 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 08:53:18 -0700 Subject: Cell: move tile clear code to flush_spans() --- src/mesa/pipe/cell/spu/spu_tri.c | 51 ++++++++++++++++++++++------------------ 1 file changed, 28 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 83bb247b22..3f46e75d7c 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -406,22 +406,44 @@ static void flush_spans( void ) } - /* _really_ clear tiles now if needed */ - if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { + /* OK, we're very likely to need the tile data now. + * clear or finish waiting if needed. + */ + if (spu.cur_ctile_status == TILE_STATUS_GETTING) { + /* wait for mfc_get() to complete */ + //printf("SPU: %u: waiting for ctile\n", spu.init.id); + wait_on_mask(1 << TAG_READ_TILE_COLOR); + spu.cur_ctile_status = TILE_STATUS_CLEAN; + } + else if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { + //printf("SPU %u: clearing C tile %u, %u\n", spu.init.id, setup.tx, setup.ty); clear_c_tile(&spu.ctile); spu.cur_ctile_status = TILE_STATUS_DIRTY; } - if (spu.depth_stencil.depth.enabled && - spu.cur_ztile_status == TILE_STATUS_CLEAR) { - clear_z_tile(&spu.ztile); - spu.cur_ztile_status = TILE_STATUS_DIRTY; + ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED); + + if (spu.depth_stencil.depth.enabled) { + if (spu.cur_ztile_status == TILE_STATUS_GETTING) { + /* wait for mfc_get() to complete */ + //printf("SPU: %u: waiting for ztile\n", spu.init.id); + wait_on_mask(1 << TAG_READ_TILE_Z); + spu.cur_ztile_status = TILE_STATUS_CLEAN; + } + else if (spu.cur_ztile_status == TILE_STATUS_CLEAR) { + //printf("SPU %u: clearing Z tile %u, %u\n", spu.init.id, setup.tx, setup.ty); + clear_z_tile(&spu.ztile); + spu.cur_ztile_status = TILE_STATUS_DIRTY; + } + ASSERT(spu.cur_ztile_status != TILE_STATUS_DEFINED); } /* XXX this loop could be moved into the above switch cases and * calculate_mask() could be simplified a bit... */ for (x = block(minleft); x <= block(maxright); x += 2) { +#if 1 emit_quad( x, setup.span.y, calculate_mask( x ) ); +#endif } setup.span.y = 0; @@ -835,23 +857,6 @@ tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) /* init_constant_attribs( setup ); */ - if (spu.cur_ctile_status == TILE_STATUS_GETTING) { - /* wait for mfc_get() to complete */ - wait_on_mask(1 << TAG_READ_TILE_COLOR); - spu.cur_ctile_status = TILE_STATUS_CLEAN; - } - ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED); - - if (spu.depth_stencil.depth.enabled) { - if (spu.cur_ztile_status == TILE_STATUS_GETTING) { - /* wait for mfc_get() to complete */ - wait_on_mask(1 << TAG_READ_TILE_Z); - spu.cur_ztile_status = TILE_STATUS_CLEAN; - } - ASSERT(spu.cur_ztile_status != TILE_STATUS_DEFINED); - } - - if (setup.oneoverarea < 0.0) { /* emaj on left: */ -- cgit v1.2.3 From 18105195a86b8294b578462febf47692832e8705 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 09:54:21 -0700 Subject: Cell: checkpoint: start to SIMD-ize texture sampling --- src/mesa/pipe/cell/spu/spu_main.c | 10 ++++++++++ src/mesa/pipe/cell/spu/spu_main.h | 4 ++++ src/mesa/pipe/cell/spu/spu_texture.c | 17 ++++++++++++++--- src/mesa/pipe/cell/spu/spu_texture.h | 2 +- src/mesa/pipe/cell/spu/spu_tri.c | 8 ++++---- 5 files changed, 33 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index ba4d180cc0..412661061a 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -263,6 +263,16 @@ cmd_state_texture(const struct cell_command_texture *texture) spu.init.id, texture->start, texture->width, texture->height); memcpy(&spu.texture, texture, sizeof(*texture)); + spu.tex_size = VEC_LITERAL(vector float, + spu.texture.width, + spu.texture.height, + 0.0, + 0.0); + spu.tex_size_mask = VEC_LITERAL(vector unsigned int, + spu.texture.width - 1, + spu.texture.height - 1, + 0, + 0); } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 7a12715b0b..02b62ee5cd 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -110,6 +110,10 @@ struct spu_global /** for converting RGBA to PIPE_FORMAT_x colors */ vector unsigned char color_shuffle; + + vector float tex_size; + vector unsigned int tex_size_mask; /**< == int(size - 1) */ + } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index c1dc6bfe90..1cf958806f 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -128,12 +128,23 @@ get_tex_tile(uint i, uint j) * XXX this is extremely primitive for now. */ uint -sample_texture(float4 texcoord) +sample_texture(vector float texcoord) { +#if 0 /* wrap/repeat */ - uint i = (uint) (texcoord.f[0] * spu.texture.width) % spu.texture.width; - uint j = (uint) (texcoord.f[1] * spu.texture.height) % spu.texture.height; + uint i = (uint) (spu_extract(texcoord, 0) * spu.texture.width) % spu.texture.width; + uint j = (uint) (spu_extract(texcoord, 1) * spu.texture.height) % spu.texture.height; uint pos = get_tex_tile(i, j); uint texel = tex_tiles[pos].ui[j % TILE_SIZE][i % TILE_SIZE]; return texel; +#else + vector float tc = spu_mul(texcoord, spu.tex_size); + vector unsigned int itc = spu_convtu(tc, 0); + itc = spu_and(itc, spu.tex_size_mask); + uint i = spu_extract(itc, 0); + uint j = spu_extract(itc, 1); + uint pos = get_tex_tile(i, j); + uint texel = tex_tiles[pos].ui[j % TILE_SIZE][i % TILE_SIZE]; + return texel; +#endif } diff --git a/src/mesa/pipe/cell/spu/spu_texture.h b/src/mesa/pipe/cell/spu/spu_texture.h index 938a42b549..5bc8e71879 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.h +++ b/src/mesa/pipe/cell/spu/spu_texture.h @@ -37,7 +37,7 @@ invalidate_tex_cache(void); extern uint -sample_texture(float4 texcoord); +sample_texture(vector float texcoord); #endif /* SPU_TEXTURE_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 3f46e75d7c..c148c75dd6 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -309,13 +309,13 @@ emit_quad( int x, int y, mask_t mask ) eval_coeff(2, (float) x, (float) y, texcoords); if (spu_extract(mask, 0)) - spu.ctile.ui[iy][ix] = sample_texture(texcoords[0]); + spu.ctile.ui[iy][ix] = sample_texture(texcoords[0].v); if (spu_extract(mask, 1)) - spu.ctile.ui[iy][ix+1] = sample_texture(texcoords[1]); + spu.ctile.ui[iy][ix+1] = sample_texture(texcoords[1].v); if (spu_extract(mask, 2)) - spu.ctile.ui[iy+1][ix] = sample_texture(texcoords[2]); + spu.ctile.ui[iy+1][ix] = sample_texture(texcoords[2].v); if (spu_extract(mask, 3)) - spu.ctile.ui[iy+1][ix+1] = sample_texture(texcoords[3]); + spu.ctile.ui[iy+1][ix+1] = sample_texture(texcoords[3].v); } else { /* simple shading */ -- cgit v1.2.3 From 703a8691553386242bf3d6662c314fc35b617194 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 11:02:47 -0700 Subject: Cell: SIMD-ize more of texture sampling --- src/mesa/pipe/cell/spu/spu_texture.c | 66 ++++++++++++++++-------------------- 1 file changed, 29 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index 1cf958806f..b52df970d0 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -26,6 +26,8 @@ **************************************************************************/ +#include + #include "pipe/p_compiler.h" #include "spu_main.h" #include "spu_texture.h" @@ -41,7 +43,7 @@ static tile_t tex_tiles[CACHE_SIZE] ALIGN16_ATTRIB; -static int tex_tile_x[CACHE_SIZE], tex_tile_y[CACHE_SIZE]; +static vector unsigned int tex_tile_xy[CACHE_SIZE]; @@ -53,20 +55,19 @@ invalidate_tex_cache(void) { /* XXX memset? */ uint i; - for (i = 0; i < CACHE_SIZE; i++) - tex_tile_x[i] = tex_tile_y[i] = -1; + for (i = 0; i < CACHE_SIZE; i++) { + tex_tile_xy[i] = VEC_LITERAL(vector unsigned int, ~0U, ~0U, ~0U, ~0U); + } } /** - * Return the cache pos/index which corresponds to texel (i,j) + * Return the cache pos/index which corresponds to tile (tx,ty) */ static INLINE uint -cache_pos(uint i, uint j) +cache_pos(vector unsigned int txty) { - uint tx = i / TILE_SIZE; - uint ty = j / TILE_SIZE; - uint pos = (tx + ty * 4) % CACHE_SIZE; + uint pos = (spu_extract(txty,0) + spu_extract(txty,1) * 4) % CACHE_SIZE; return pos; } @@ -76,26 +77,28 @@ cache_pos(uint i, uint j) * in the cache. */ static uint -get_tex_tile(uint i, uint j) +get_tex_tile(vector unsigned int ij) { - const int tx = i / TILE_SIZE; - const int ty = j / TILE_SIZE; - const uint pos = cache_pos(i, j); + /* tile address: tx,ty */ + const vector unsigned int txty = spu_rlmask(ij, -5); /* divide by 32 */ + const uint pos = cache_pos(txty); + + if ((spu_extract(tex_tile_xy[pos], 0) != spu_extract(txty, 0)) || + (spu_extract(tex_tile_xy[pos], 1) != spu_extract(txty, 1))) { - if (tex_tile_x[pos] != tx || tex_tile_y[pos] != ty) { /* texture cache miss, fetch tile from main memory */ const uint tiles_per_row = spu.texture.width / TILE_SIZE; const uint bytes_per_tile = sizeof(tile_t); const void *src = (const ubyte *) spu.texture.start - + (ty * tiles_per_row + tx) * bytes_per_tile; + + (spu_extract(txty,1) * tiles_per_row + spu_extract(txty,0)) * bytes_per_tile; printf("SPU %u: tex cache miss at %d, %d pos=%u old=%d,%d\n", - spu.init.id, tx, ty, pos, - tex_tile_x[pos], tex_tile_y[pos]); -#if 0 - printf("SPU %u: get tex tile from %p to %p\n", - spu.init.id, src, tex_tiles[pos].t32); -#endif + spu.init.id, + spu_extract(txty,0), + spu_extract(txty,1), + pos, + spu_extract(tex_tile_xy[pos],0), + spu_extract(tex_tile_xy[pos],1)); ASSERT_ALIGN16(tex_tiles[pos].ui); ASSERT_ALIGN16(src); @@ -109,8 +112,7 @@ get_tex_tile(uint i, uint j) wait_on_mask(1 << TAG_TEXTURE_TILE); - tex_tile_x[pos] = tx; - tex_tile_y[pos] = ty; + tex_tile_xy[pos] = txty; } else { #if 0 @@ -130,21 +132,11 @@ get_tex_tile(uint i, uint j) uint sample_texture(vector float texcoord) { -#if 0 - /* wrap/repeat */ - uint i = (uint) (spu_extract(texcoord, 0) * spu.texture.width) % spu.texture.width; - uint j = (uint) (spu_extract(texcoord, 1) * spu.texture.height) % spu.texture.height; - uint pos = get_tex_tile(i, j); - uint texel = tex_tiles[pos].ui[j % TILE_SIZE][i % TILE_SIZE]; - return texel; -#else vector float tc = spu_mul(texcoord, spu.tex_size); - vector unsigned int itc = spu_convtu(tc, 0); - itc = spu_and(itc, spu.tex_size_mask); - uint i = spu_extract(itc, 0); - uint j = spu_extract(itc, 1); - uint pos = get_tex_tile(i, j); - uint texel = tex_tiles[pos].ui[j % TILE_SIZE][i % TILE_SIZE]; + vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ + itc = spu_and(itc, spu.tex_size_mask); /* mask (GL_REPEAT) */ + vector unsigned int ij = spu_and(itc, TILE_SIZE-1); /* intra tile addr */ + uint pos = get_tex_tile(itc); + uint texel = tex_tiles[pos].ui[spu_extract(ij, 1)][spu_extract(ij, 0)]; return texel; -#endif } -- cgit v1.2.3 From 9a5074217fd3be8feff2be597bb124a2a3637d0a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 12:50:16 -0700 Subject: Cell: added spu_unpack_color(), spu_pack_R8G8B8A8() --- src/mesa/pipe/cell/spu/spu_colorpack.h | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_colorpack.h b/src/mesa/pipe/cell/spu/spu_colorpack.h index 9977a6ece0..0c93c06562 100644 --- a/src/mesa/pipe/cell/spu/spu_colorpack.h +++ b/src/mesa/pipe/cell/spu/spu_colorpack.h @@ -35,6 +35,17 @@ #include +static INLINE unsigned int +spu_pack_R8G8B8A8(vector float rgba) +{ + vector unsigned int out = spu_convtu(rgba, 32); + out = spu_shuffle(out, out, VEC_LITERAL(vector unsigned char, + 0, 4, 8, 12, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0)); + return spu_extract(out, 0); +} + + static INLINE unsigned int spu_pack_A8R8G8B8(vector float rgba) { @@ -66,4 +77,18 @@ spu_pack_color_shuffle(vector float rgba, vector unsigned char shuffle) } +static INLINE vector float +spu_unpack_color(uint color) +{ + vector unsigned int color_u4 = spu_splats(color); + color_u4 = spu_shuffle(color_u4, color_u4, + VEC_LITERAL(vector unsigned char, + 0, 0, 0, 0, + 5, 5, 5, 5, + 10, 10, 10, 10, + 15, 15, 15, 15)); + return spu_convtf(color_u4, 32); +} + + #endif /* SPU_COLORPACK_H */ -- cgit v1.2.3 From 0a45f7594870cb7296100fb5f5d5dc82888a467d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 12:50:42 -0700 Subject: Cell: implement basic bilinear texture sampler --- src/mesa/pipe/cell/spu/spu_texture.c | 67 ++++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_texture.h | 4 +++ 2 files changed, 71 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index b52df970d0..26a5eefc48 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -32,6 +32,7 @@ #include "spu_main.h" #include "spu_texture.h" #include "spu_tile.h" +#include "spu_colorpack.h" /** @@ -140,3 +141,69 @@ sample_texture(vector float texcoord) uint texel = tex_tiles[pos].ui[spu_extract(ij, 1)][spu_extract(ij, 0)]; return texel; } + + +uint +sample_texture_bilinear(vector float texcoord) +{ + static const vector unsigned int offset10 = {1, 0, 0, 0}; + static const vector unsigned int offset01 = {0, 1, 0, 0}; + + vector float tc = spu_mul(texcoord, spu.tex_size); + /* itcST */ + vector unsigned int itc00 = spu_convtu(tc, 0); /* convert to int */ + vector unsigned int itc01 = spu_add(itc00, offset01); + vector unsigned int itc10 = spu_add(itc00, offset10); + vector unsigned int itc11 = spu_add(itc10, offset01); + + itc00 = spu_and(itc00, spu.tex_size_mask); /* mask (GL_REPEAT) */ + itc01 = spu_and(itc01, spu.tex_size_mask); /* mask (GL_REPEAT) */ + itc10 = spu_and(itc10, spu.tex_size_mask); /* mask (GL_REPEAT) */ + itc11 = spu_and(itc11, spu.tex_size_mask); /* mask (GL_REPEAT) */ + + /* intra tile addr */ + vector unsigned int ij00 = spu_and(itc00, TILE_SIZE-1); + vector unsigned int ij01 = spu_and(itc01, TILE_SIZE-1); + vector unsigned int ij10 = spu_and(itc10, TILE_SIZE-1); + vector unsigned int ij11 = spu_and(itc11, TILE_SIZE-1); + + uint pos00 = get_tex_tile(itc00); + uint pos01 = get_tex_tile(itc01); + uint pos10 = get_tex_tile(itc10); + uint pos11 = get_tex_tile(itc11); + + vector float texel00 = spu_unpack_color(tex_tiles[pos00].ui[spu_extract(ij00, 1)][spu_extract(ij00, 0)]); + vector float texel01 = spu_unpack_color(tex_tiles[pos01].ui[spu_extract(ij01, 1)][spu_extract(ij01, 0)]); + vector float texel10 = spu_unpack_color(tex_tiles[pos10].ui[spu_extract(ij10, 1)][spu_extract(ij10, 0)]); + vector float texel11 = spu_unpack_color(tex_tiles[pos11].ui[spu_extract(ij11, 1)][spu_extract(ij11, 0)]); + + /* Compute weighting factors in [0,1] + * Multiply texcoord by 1024, AND with 1023, convert back to float. + */ + vector float tc1024 = spu_mul(tc, spu_splats(1024.0f)); + vector signed int itc1024 = spu_convts(tc1024, 0); + itc1024 = spu_and(itc1024, spu_splats((1 << 10) - 1)); + vector float weight = spu_convtf(itc1024, 10); + + /* smeared frac and 1-frac */ + vector float sfrac = spu_splats(spu_extract(weight, 0)); + vector float tfrac = spu_splats(spu_extract(weight, 1)); + vector float sfrac1 = spu_sub(spu_splats(1.0f), sfrac); + vector float tfrac1 = spu_sub(spu_splats(1.0f), tfrac); + + /* multiply the samples (colors) by the S/T weights */ + texel00 = spu_mul(spu_mul(texel00, sfrac1), tfrac1); + texel10 = spu_mul(spu_mul(texel10, sfrac ), tfrac1); + texel01 = spu_mul(spu_mul(texel01, sfrac1), tfrac ); + texel11 = spu_mul(spu_mul(texel11, sfrac ), tfrac ); + + /* compute sum of weighted samples */ + vector float texel_sum = spu_add(texel00, texel01); + texel_sum = spu_add(texel_sum, texel10); + texel_sum = spu_add(texel_sum, texel11); + + /* convert to uint color */ + uint texel = spu_pack_R8G8B8A8(texel_sum); + + return texel; +} diff --git a/src/mesa/pipe/cell/spu/spu_texture.h b/src/mesa/pipe/cell/spu/spu_texture.h index 5bc8e71879..25cbe9b3c6 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.h +++ b/src/mesa/pipe/cell/spu/spu_texture.h @@ -40,4 +40,8 @@ extern uint sample_texture(vector float texcoord); +extern uint +sample_texture_bilinear(vector float texcoord); + + #endif /* SPU_TEXTURE_H */ -- cgit v1.2.3 From ca1d2fc5f6fb138025f6848591e3494e4b881930 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 13:16:10 -0700 Subject: Cell: improved bilinear filtering avoid calling get_tex_tile() if all texels are in same tile --- src/mesa/pipe/cell/spu/spu_texture.c | 27 +++++++++++++++++++-------- 1 file changed, 19 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index 26a5eefc48..6e243f7fa3 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -150,16 +150,17 @@ sample_texture_bilinear(vector float texcoord) static const vector unsigned int offset01 = {0, 1, 0, 0}; vector float tc = spu_mul(texcoord, spu.tex_size); - /* itcST */ + /* integer texcoords S,T: */ vector unsigned int itc00 = spu_convtu(tc, 0); /* convert to int */ vector unsigned int itc01 = spu_add(itc00, offset01); vector unsigned int itc10 = spu_add(itc00, offset10); vector unsigned int itc11 = spu_add(itc10, offset01); - itc00 = spu_and(itc00, spu.tex_size_mask); /* mask (GL_REPEAT) */ - itc01 = spu_and(itc01, spu.tex_size_mask); /* mask (GL_REPEAT) */ - itc10 = spu_and(itc10, spu.tex_size_mask); /* mask (GL_REPEAT) */ - itc11 = spu_and(itc11, spu.tex_size_mask); /* mask (GL_REPEAT) */ + /* mask (GL_REPEAT) */ + itc00 = spu_and(itc00, spu.tex_size_mask); + itc01 = spu_and(itc01, spu.tex_size_mask); + itc10 = spu_and(itc10, spu.tex_size_mask); + itc11 = spu_and(itc11, spu.tex_size_mask); /* intra tile addr */ vector unsigned int ij00 = spu_and(itc00, TILE_SIZE-1); @@ -167,11 +168,21 @@ sample_texture_bilinear(vector float texcoord) vector unsigned int ij10 = spu_and(itc10, TILE_SIZE-1); vector unsigned int ij11 = spu_and(itc11, TILE_SIZE-1); + /* get tile cache positions */ uint pos00 = get_tex_tile(itc00); - uint pos01 = get_tex_tile(itc01); - uint pos10 = get_tex_tile(itc10); - uint pos11 = get_tex_tile(itc11); + uint pos01, pos10, pos11; + if ((spu_extract(ij00, 0) < TILE_SIZE-1) && + (spu_extract(ij00, 1) < TILE_SIZE-1)) { + /* all texels are in the same tile */ + pos01 = pos10 = pos11 = pos00; + } + else { + pos01 = get_tex_tile(itc01); + pos10 = get_tex_tile(itc10); + pos11 = get_tex_tile(itc11); + } + /* get texels from tiles and convert to float[4] */ vector float texel00 = spu_unpack_color(tex_tiles[pos00].ui[spu_extract(ij00, 1)][spu_extract(ij00, 0)]); vector float texel01 = spu_unpack_color(tex_tiles[pos01].ui[spu_extract(ij01, 1)][spu_extract(ij01, 0)]); vector float texel10 = spu_unpack_color(tex_tiles[pos10].ui[spu_extract(ij10, 1)][spu_extract(ij10, 0)]); -- cgit v1.2.3 From 8f924e4df06a5d45dda338e7a0a87308e48df57e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 13:23:07 -0700 Subject: Cell: choose bilinear vs. nearest filtering according to sampler state --- src/mesa/pipe/cell/spu/spu_main.c | 4 ++++ src/mesa/pipe/cell/spu/spu_main.h | 2 ++ src/mesa/pipe/cell/spu/spu_texture.c | 2 +- src/mesa/pipe/cell/spu/spu_texture.h | 2 +- src/mesa/pipe/cell/spu/spu_tri.c | 8 ++++---- 5 files changed, 12 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 412661061a..48e016fc8b 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -252,6 +252,10 @@ cmd_state_sampler(const struct pipe_sampler_state *state) spu.init.id); memcpy(&spu.sampler[0], state, sizeof(*state)); + if (spu.sampler[0].min_img_filter == PIPE_TEX_FILTER_LINEAR) + spu.sample_texture = sample_texture_bilinear; + else + spu.sample_texture = sample_texture_nearest; } diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index 02b62ee5cd..fb98b0d889 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -114,6 +114,8 @@ struct spu_global vector float tex_size; vector unsigned int tex_size_mask; /**< == int(size - 1) */ + uint (*sample_texture)(vector float texcoord); + } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index 6e243f7fa3..ecacf2ec88 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -131,7 +131,7 @@ get_tex_tile(vector unsigned int ij) * XXX this is extremely primitive for now. */ uint -sample_texture(vector float texcoord) +sample_texture_nearest(vector float texcoord) { vector float tc = spu_mul(texcoord, spu.tex_size); vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ diff --git a/src/mesa/pipe/cell/spu/spu_texture.h b/src/mesa/pipe/cell/spu/spu_texture.h index 25cbe9b3c6..0e000bfebf 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.h +++ b/src/mesa/pipe/cell/spu/spu_texture.h @@ -37,7 +37,7 @@ invalidate_tex_cache(void); extern uint -sample_texture(vector float texcoord); +sample_texture_nearest(vector float texcoord); extern uint diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index c148c75dd6..7b422f71a8 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -309,13 +309,13 @@ emit_quad( int x, int y, mask_t mask ) eval_coeff(2, (float) x, (float) y, texcoords); if (spu_extract(mask, 0)) - spu.ctile.ui[iy][ix] = sample_texture(texcoords[0].v); + spu.ctile.ui[iy][ix] = spu.sample_texture(texcoords[0].v); if (spu_extract(mask, 1)) - spu.ctile.ui[iy][ix+1] = sample_texture(texcoords[1].v); + spu.ctile.ui[iy][ix+1] = spu.sample_texture(texcoords[1].v); if (spu_extract(mask, 2)) - spu.ctile.ui[iy+1][ix] = sample_texture(texcoords[2].v); + spu.ctile.ui[iy+1][ix] = spu.sample_texture(texcoords[2].v); if (spu_extract(mask, 3)) - spu.ctile.ui[iy+1][ix+1] = sample_texture(texcoords[3].v); + spu.ctile.ui[iy+1][ix+1] = spu.sample_texture(texcoords[3].v); } else { /* simple shading */ -- cgit v1.2.3 From 7a1d01f2a0d8f0875a265e7d4e31e1348fd82677 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 15:06:10 -0700 Subject: Cell: emit blend state to SPUs --- src/mesa/pipe/cell/common.h | 3 ++- src/mesa/pipe/cell/ppu/cell_state_emit.c | 6 ++++++ src/mesa/pipe/cell/spu/spu_main.c | 17 +++++++++++++++++ src/mesa/pipe/cell/spu/spu_main.h | 1 + 4 files changed, 26 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index 7e193f31be..d861e82d33 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -85,7 +85,8 @@ #define CELL_CMD_STATE_VERTEX_INFO 14 #define CELL_CMD_STATE_VIEWPORT 15 #define CELL_CMD_STATE_VS_ARRAY_INFO 16 -#define CELL_CMD_VS_EXECUTE 17 +#define CELL_CMD_STATE_BLEND 17 +#define CELL_CMD_VS_EXECUTE 18 #define CELL_NUM_BUFFERS 4 diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c index 702184416b..3b2670f786 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.c +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.c @@ -61,6 +61,12 @@ cell_emit_state(struct cell_context *cell) fb->height = cell->framebuffer.cbufs[0]->height; } + if (cell->dirty & CELL_NEW_BLEND) { + emit_state_cmd(cell, CELL_CMD_STATE_BLEND, + cell->blend, + sizeof(struct pipe_blend_state)); + } + if (cell->dirty & CELL_NEW_DEPTH_STENCIL) { emit_state_cmd(cell, CELL_CMD_STATE_DEPTH_STENCIL, cell->depth_stencil, diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 48e016fc8b..9d8e6df0e3 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -232,6 +232,18 @@ cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) } +static void +cmd_state_blend(const struct pipe_blend_state *state) +{ + if (Debug) + printf("SPU %u: BLEND: ztest %d\n", + spu.init.id, + state->blend_enable); + + memcpy(&spu.blend, state, sizeof(*state)); +} + + static void cmd_state_depth_stencil(const struct pipe_depth_stencil_alpha_state *state) { @@ -398,6 +410,11 @@ cmd_batch(uint opcode) cmd_finish(); pos += 1; break; + case CELL_CMD_STATE_BLEND: + cmd_state_blend((struct pipe_blend_state *) + &buffer[pos+1]); + pos += (1 + sizeof(struct pipe_blend_state) / 4); + break; case CELL_CMD_STATE_DEPTH_STENCIL: cmd_state_depth_stencil((struct pipe_depth_stencil_alpha_state *) &buffer[pos+1]); diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index fb98b0d889..b22d563551 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -86,6 +86,7 @@ struct spu_global struct cell_init_info init; struct spu_framebuffer fb; + struct pipe_blend_state blend_stencil; struct pipe_depth_stencil_alpha_state depth_stencil; struct pipe_blend_state blend; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; -- cgit v1.2.3 From 168247d1caee28ef577ad4c3c4308451f1193062 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 15:10:35 -0700 Subject: Cell: replace float 4 with vector float in eval_coeff() --- src/mesa/pipe/cell/spu/spu_tri.c | 33 +++++++++++++++++---------------- 1 file changed, 17 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 7b422f71a8..199afa1aa6 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -32,6 +32,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_format.h" #include "pipe/p_util.h" +#include "spu_blend.h" #include "spu_colorpack.h" #include "spu_main.h" #include "spu_texture.h" @@ -206,14 +207,14 @@ clip_emit_quad(struct setup_stage *setup) * Eg: four colors will be compute. */ static INLINE void -eval_coeff(uint slot, float x, float y, float4 result[4]) +eval_coeff(uint slot, float x, float y, vector float result[4]) { switch (spu.vertex_info.interp_mode[slot]) { case INTERP_CONSTANT: result[QUAD_TOP_LEFT] = result[QUAD_TOP_RIGHT] = result[QUAD_BOTTOM_LEFT] = - result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0; + result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0.v; break; case INTERP_LINEAR: @@ -227,10 +228,10 @@ eval_coeff(uint slot, float x, float y, float4 result[4]) spu_add(spu_mul(spu_splats(x), dadx), spu_mul(spu_splats(y), dady))); - result[QUAD_TOP_LEFT].v = topLeft; - result[QUAD_TOP_RIGHT].v = spu_add(topLeft, dadx); - result[QUAD_BOTTOM_LEFT].v = spu_add(topLeft, dady); - result[QUAD_BOTTOM_RIGHT].v = spu_add(spu_add(topLeft, dadx), dady); + result[QUAD_TOP_LEFT] = topLeft; + result[QUAD_TOP_RIGHT] = spu_add(topLeft, dadx); + result[QUAD_BOTTOM_LEFT] = spu_add(topLeft, dady); + result[QUAD_BOTTOM_RIGHT] = spu_add(spu_add(topLeft, dadx), dady); } } } @@ -305,32 +306,32 @@ emit_quad( int x, int y, mask_t mask ) if (spu.texture.start) { /* texture mapping */ - float4 texcoords[4]; + vector float texcoords[4]; eval_coeff(2, (float) x, (float) y, texcoords); if (spu_extract(mask, 0)) - spu.ctile.ui[iy][ix] = spu.sample_texture(texcoords[0].v); + spu.ctile.ui[iy][ix] = spu.sample_texture(texcoords[0]); if (spu_extract(mask, 1)) - spu.ctile.ui[iy][ix+1] = spu.sample_texture(texcoords[1].v); + spu.ctile.ui[iy][ix+1] = spu.sample_texture(texcoords[1]); if (spu_extract(mask, 2)) - spu.ctile.ui[iy+1][ix] = spu.sample_texture(texcoords[2].v); + spu.ctile.ui[iy+1][ix] = spu.sample_texture(texcoords[2]); if (spu_extract(mask, 3)) - spu.ctile.ui[iy+1][ix+1] = spu.sample_texture(texcoords[3].v); + spu.ctile.ui[iy+1][ix+1] = spu.sample_texture(texcoords[3]); } else { /* simple shading */ const vector unsigned char shuffle = spu.color_shuffle; - float4 colors[4]; + vector float colors[4]; eval_coeff(1, (float) x, (float) y, colors); if (spu_extract(mask, 0)) - spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0].v, shuffle); + spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0], shuffle); if (spu_extract(mask, 1)) - spu.ctile.ui[iy][ix+1] = spu_pack_color_shuffle(colors[1].v, shuffle); + spu.ctile.ui[iy][ix+1] = spu_pack_color_shuffle(colors[1], shuffle); if (spu_extract(mask, 2)) - spu.ctile.ui[iy+1][ix] = spu_pack_color_shuffle(colors[2].v, shuffle); + spu.ctile.ui[iy+1][ix] = spu_pack_color_shuffle(colors[2], shuffle); if (spu_extract(mask, 3)) - spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(colors[3].v, shuffle); + spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(colors[3], shuffle); } #if 0 -- cgit v1.2.3 From bc1ad6bcbd5c63da9c10d0276c9d7535b6139437 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 15:17:50 -0700 Subject: Cell: some basic blending code --- src/mesa/pipe/cell/spu/Makefile | 1 + src/mesa/pipe/cell/spu/spu_blend.c | 62 ++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_blend.h | 37 +++++++++++++++++++++++ src/mesa/pipe/cell/spu/spu_tri.c | 5 +++ 4 files changed, 105 insertions(+) create mode 100644 src/mesa/pipe/cell/spu/spu_blend.c create mode 100644 src/mesa/pipe/cell/spu/spu_blend.h (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index 91a631b699..66f16cde9b 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -19,6 +19,7 @@ PROG_SPU_EMBED_O = $(PROG)_spu-embed.o SOURCES = \ spu_main.c \ + spu_blend.c \ spu_render.c \ spu_texture.c \ spu_tile.c \ diff --git a/src/mesa/pipe/cell/spu/spu_blend.c b/src/mesa/pipe/cell/spu/spu_blend.c new file mode 100644 index 0000000000..23ec0eeb45 --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_blend.c @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "spu_main.h" +#include "spu_blend.h" +#include "spu_colorpack.h" + + +void +blend_quad(uint itx, uint ity, vector float colors[4]) +{ + /* simple SRC_ALPHA, ONE_MINUS_SRC_ALPHA blending */ + vector float fbc00 = spu_unpack_color(spu.ctile.ui[ity][itx]); + vector float fbc01 = spu_unpack_color(spu.ctile.ui[ity][itx+1]); + vector float fbc10 = spu_unpack_color(spu.ctile.ui[ity+1][itx]); + vector float fbc11 = spu_unpack_color(spu.ctile.ui[ity+1][itx+1]); + + vector float alpha00 = spu_splats(spu_extract(colors[0], 3)); + vector float alpha01 = spu_splats(spu_extract(colors[1], 3)); + vector float alpha10 = spu_splats(spu_extract(colors[2], 3)); + vector float alpha11 = spu_splats(spu_extract(colors[3], 3)); + + vector float one_minus_alpha00 = spu_sub(spu_splats(1.0f), alpha00); + vector float one_minus_alpha01 = spu_sub(spu_splats(1.0f), alpha01); + vector float one_minus_alpha10 = spu_sub(spu_splats(1.0f), alpha10); + vector float one_minus_alpha11 = spu_sub(spu_splats(1.0f), alpha11); + + colors[0] = spu_add(spu_mul(colors[0], alpha00), + spu_mul(fbc00, one_minus_alpha00)); + colors[1] = spu_add(spu_mul(colors[1], alpha01), + spu_mul(fbc01, one_minus_alpha01)); + colors[2] = spu_add(spu_mul(colors[2], alpha10), + spu_mul(fbc10, one_minus_alpha10)); + colors[3] = spu_add(spu_mul(colors[3], alpha11), + spu_mul(fbc11, one_minus_alpha11)); +} + diff --git a/src/mesa/pipe/cell/spu/spu_blend.h b/src/mesa/pipe/cell/spu/spu_blend.h new file mode 100644 index 0000000000..2b594b578b --- /dev/null +++ b/src/mesa/pipe/cell/spu/spu_blend.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SPU_BLEND_H +#define SPU_BLEND_H + + +extern void +blend_quad(uint itx, uint ity, vector float colors[4]); + + +#endif /* SPU_BLEND_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 199afa1aa6..89aaca9a72 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -324,6 +324,11 @@ emit_quad( int x, int y, mask_t mask ) vector float colors[4]; eval_coeff(1, (float) x, (float) y, colors); +#if 0 + if (spu.blend.blend_enable) + blend_quad(ix % TILE_SIZE, iy % TILE_SIZE, colors); +#endif + if (spu_extract(mask, 0)) spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0], shuffle); if (spu_extract(mask, 1)) -- cgit v1.2.3 From 76c1a10eb121f040ef510124bf6aa24c4c5c3f8f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 16:04:50 -0700 Subject: Cell: fix typo --- src/mesa/pipe/cell/spu/spu_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 9d8e6df0e3..b0311db1aa 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -236,7 +236,7 @@ static void cmd_state_blend(const struct pipe_blend_state *state) { if (Debug) - printf("SPU %u: BLEND: ztest %d\n", + printf("SPU %u: BLEND: enabled %d\n", spu.init.id, state->blend_enable); -- cgit v1.2.3 From 5068b573c417bdb317e1938585bebfe931bda049 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 16:05:13 -0700 Subject: Cell: added spu_unpack_A8R8G8B8() --- src/mesa/pipe/cell/spu/spu_colorpack.h | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_colorpack.h b/src/mesa/pipe/cell/spu/spu_colorpack.h index 0c93c06562..57ea3525c2 100644 --- a/src/mesa/pipe/cell/spu/spu_colorpack.h +++ b/src/mesa/pipe/cell/spu/spu_colorpack.h @@ -91,4 +91,19 @@ spu_unpack_color(uint color) } +static INLINE vector float +spu_unpack_A8R8G8B8(uint color) +{ + vector unsigned int color_u4 = spu_splats(color); + color_u4 = spu_shuffle(color_u4, color_u4, + VEC_LITERAL(vector unsigned char, + 5, 5, 5, 5, + 10, 10, 10, 10, + 15, 15, 15, 15, + 0, 0, 0, 0)); + + return spu_convtf(color_u4, 32); +} + + #endif /* SPU_COLORPACK_H */ -- cgit v1.2.3 From efa8e03a6f3f7c27b019d20cca93bf7e624d7035 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 16:06:51 -0700 Subject: Cell: texture sampler functions always return vector float now Texture colors look the same now, regardless of X display/pixel format --- src/mesa/pipe/cell/spu/spu_main.h | 2 +- src/mesa/pipe/cell/spu/spu_texture.c | 19 ++++++++----------- src/mesa/pipe/cell/spu/spu_texture.h | 4 ++-- src/mesa/pipe/cell/spu/spu_tri.c | 36 ++++++++++++++++++------------------ 4 files changed, 29 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index b22d563551..cfd4d72729 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -115,7 +115,7 @@ struct spu_global vector float tex_size; vector unsigned int tex_size_mask; /**< == int(size - 1) */ - uint (*sample_texture)(vector float texcoord); + vector float (*sample_texture)(vector float texcoord); } ALIGN16_ATTRIB; diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index ecacf2ec88..9ee2b45e24 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -130,7 +130,7 @@ get_tex_tile(vector unsigned int ij) * Get texture sample at texcoord. * XXX this is extremely primitive for now. */ -uint +vector float sample_texture_nearest(vector float texcoord) { vector float tc = spu_mul(texcoord, spu.tex_size); @@ -139,11 +139,11 @@ sample_texture_nearest(vector float texcoord) vector unsigned int ij = spu_and(itc, TILE_SIZE-1); /* intra tile addr */ uint pos = get_tex_tile(itc); uint texel = tex_tiles[pos].ui[spu_extract(ij, 1)][spu_extract(ij, 0)]; - return texel; + return spu_unpack_A8R8G8B8(texel); } -uint +vector float sample_texture_bilinear(vector float texcoord) { static const vector unsigned int offset10 = {1, 0, 0, 0}; @@ -183,10 +183,10 @@ sample_texture_bilinear(vector float texcoord) } /* get texels from tiles and convert to float[4] */ - vector float texel00 = spu_unpack_color(tex_tiles[pos00].ui[spu_extract(ij00, 1)][spu_extract(ij00, 0)]); - vector float texel01 = spu_unpack_color(tex_tiles[pos01].ui[spu_extract(ij01, 1)][spu_extract(ij01, 0)]); - vector float texel10 = spu_unpack_color(tex_tiles[pos10].ui[spu_extract(ij10, 1)][spu_extract(ij10, 0)]); - vector float texel11 = spu_unpack_color(tex_tiles[pos11].ui[spu_extract(ij11, 1)][spu_extract(ij11, 0)]); + vector float texel00 = spu_unpack_A8R8G8B8(tex_tiles[pos00].ui[spu_extract(ij00, 1)][spu_extract(ij00, 0)]); + vector float texel01 = spu_unpack_A8R8G8B8(tex_tiles[pos01].ui[spu_extract(ij01, 1)][spu_extract(ij01, 0)]); + vector float texel10 = spu_unpack_A8R8G8B8(tex_tiles[pos10].ui[spu_extract(ij10, 1)][spu_extract(ij10, 0)]); + vector float texel11 = spu_unpack_A8R8G8B8(tex_tiles[pos11].ui[spu_extract(ij11, 1)][spu_extract(ij11, 0)]); /* Compute weighting factors in [0,1] * Multiply texcoord by 1024, AND with 1023, convert back to float. @@ -213,8 +213,5 @@ sample_texture_bilinear(vector float texcoord) texel_sum = spu_add(texel_sum, texel10); texel_sum = spu_add(texel_sum, texel11); - /* convert to uint color */ - uint texel = spu_pack_R8G8B8A8(texel_sum); - - return texel; + return texel_sum; } diff --git a/src/mesa/pipe/cell/spu/spu_texture.h b/src/mesa/pipe/cell/spu/spu_texture.h index 0e000bfebf..95eb87080f 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.h +++ b/src/mesa/pipe/cell/spu/spu_texture.h @@ -36,11 +36,11 @@ extern void invalidate_tex_cache(void); -extern uint +extern vector float sample_texture_nearest(vector float texcoord); -extern uint +extern vector float sample_texture_bilinear(vector float texcoord); diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 89aaca9a72..4c6de56eda 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -301,6 +301,8 @@ emit_quad( int x, int y, mask_t mask ) if (spu_extract(spu_orx(mask), 0)) { const int ix = x - setup.cliprect_minx; const int iy = y - setup.cliprect_miny; + const vector unsigned char shuffle = spu.color_shuffle; + vector float colors[4]; spu.cur_ctile_status = TILE_STATUS_DIRTY; @@ -310,34 +312,32 @@ emit_quad( int x, int y, mask_t mask ) eval_coeff(2, (float) x, (float) y, texcoords); if (spu_extract(mask, 0)) - spu.ctile.ui[iy][ix] = spu.sample_texture(texcoords[0]); + colors[0] = spu.sample_texture(texcoords[0]); if (spu_extract(mask, 1)) - spu.ctile.ui[iy][ix+1] = spu.sample_texture(texcoords[1]); + colors[1] = spu.sample_texture(texcoords[1]); if (spu_extract(mask, 2)) - spu.ctile.ui[iy+1][ix] = spu.sample_texture(texcoords[2]); + colors[2] = spu.sample_texture(texcoords[2]); if (spu_extract(mask, 3)) - spu.ctile.ui[iy+1][ix+1] = spu.sample_texture(texcoords[3]); + colors[3] = spu.sample_texture(texcoords[3]); } else { /* simple shading */ - const vector unsigned char shuffle = spu.color_shuffle; - vector float colors[4]; eval_coeff(1, (float) x, (float) y, colors); + } -#if 0 - if (spu.blend.blend_enable) - blend_quad(ix % TILE_SIZE, iy % TILE_SIZE, colors); +#if 1 + if (spu.blend.blend_enable) + blend_quad(ix % TILE_SIZE, iy % TILE_SIZE, colors); #endif - if (spu_extract(mask, 0)) - spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0], shuffle); - if (spu_extract(mask, 1)) - spu.ctile.ui[iy][ix+1] = spu_pack_color_shuffle(colors[1], shuffle); - if (spu_extract(mask, 2)) - spu.ctile.ui[iy+1][ix] = spu_pack_color_shuffle(colors[2], shuffle); - if (spu_extract(mask, 3)) - spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(colors[3], shuffle); - } + if (spu_extract(mask, 0)) + spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0], shuffle); + if (spu_extract(mask, 1)) + spu.ctile.ui[iy][ix+1] = spu_pack_color_shuffle(colors[1], shuffle); + if (spu_extract(mask, 2)) + spu.ctile.ui[iy+1][ix] = spu_pack_color_shuffle(colors[2], shuffle); + if (spu_extract(mask, 3)) + spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(colors[3], shuffle); #if 0 /* SIMD_Z with swizzled color buffer (someday) */ -- cgit v1.2.3 From 1a75464cdc12a1e83f1452707cd624c53f808308 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 16:48:00 -0700 Subject: Cell: fix small sampling error in sample_texture_bilinear() --- src/mesa/pipe/cell/spu/spu_texture.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index 9ee2b45e24..01ff33a857 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -150,6 +150,8 @@ sample_texture_bilinear(vector float texcoord) static const vector unsigned int offset01 = {0, 1, 0, 0}; vector float tc = spu_mul(texcoord, spu.tex_size); + tc = spu_add(tc, spu_splats(-0.5f)); /* half texel bias */ + /* integer texcoords S,T: */ vector unsigned int itc00 = spu_convtu(tc, 0); /* convert to int */ vector unsigned int itc01 = spu_add(itc00, offset01); -- cgit v1.2.3 From 93d061b217e31d27a1c54e50a14538e94f1404d6 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 16:48:36 -0700 Subject: Cell: move float4 typedef (temporary datatype) --- src/mesa/pipe/cell/spu/spu_main.h | 7 ------- src/mesa/pipe/cell/spu/spu_tri.c | 6 ++++++ 2 files changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h index cfd4d72729..1710a17512 100644 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ b/src/mesa/pipe/cell/spu/spu_main.h @@ -41,13 +41,6 @@ #define MAX_HEIGHT 1024 -typedef union -{ - vector float v; - float f[4]; -} float4; - - typedef union { ushort us[TILE_SIZE][TILE_SIZE]; uint ui[TILE_SIZE][TILE_SIZE]; diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 4c6de56eda..688c8646ab 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -45,6 +45,12 @@ /** Masks are uint[4] vectors with each element being 0 or 0xffffffff */ typedef vector unsigned int mask_t; +typedef union +{ + vector float v; + float f[4]; +} float4; + /** * Simplified types taken from other parts of Gallium -- cgit v1.2.3 From 7cbe5cf212d296c19ccf8e1b74d3a5b1bcb2d9e9 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 18:02:21 -0700 Subject: Cell: don't use VEC_LITERAL macro, doesn't work w/ SDK 3.0 --- src/mesa/pipe/cell/spu/spu_colorpack.h | 41 +++++++++++++++++----------------- src/mesa/pipe/cell/spu/spu_ztest.h | 24 ++++++++++---------- 2 files changed, 33 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_colorpack.h b/src/mesa/pipe/cell/spu/spu_colorpack.h index 57ea3525c2..e9fee8a3a6 100644 --- a/src/mesa/pipe/cell/spu/spu_colorpack.h +++ b/src/mesa/pipe/cell/spu/spu_colorpack.h @@ -31,7 +31,6 @@ #define SPU_COLORPACK_H -#include #include @@ -39,9 +38,11 @@ static INLINE unsigned int spu_pack_R8G8B8A8(vector float rgba) { vector unsigned int out = spu_convtu(rgba, 32); - out = spu_shuffle(out, out, VEC_LITERAL(vector unsigned char, - 0, 4, 8, 12, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0)); + + out = spu_shuffle(out, out, ((vector unsigned char) { + 0, 4, 8, 12, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0 }) ); + return spu_extract(out, 0); } @@ -50,9 +51,9 @@ static INLINE unsigned int spu_pack_A8R8G8B8(vector float rgba) { vector unsigned int out = spu_convtu(rgba, 32); - out = spu_shuffle(out, out, VEC_LITERAL(vector unsigned char, - 12, 0, 4, 8, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0)); + out = spu_shuffle(out, out, ((vector unsigned char) { + 12, 0, 4, 8, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0}) ); return spu_extract(out, 0); } @@ -61,9 +62,9 @@ static INLINE unsigned int spu_pack_B8G8R8A8(vector float rgba) { vector unsigned int out = spu_convtu(rgba, 32); - out = spu_shuffle(out, out, VEC_LITERAL(vector unsigned char, - 8, 4, 0, 12, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0)); + out = spu_shuffle(out, out, ((vector unsigned char) { + 8, 4, 0, 12, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0}) ); return spu_extract(out, 0); } @@ -82,11 +83,11 @@ spu_unpack_color(uint color) { vector unsigned int color_u4 = spu_splats(color); color_u4 = spu_shuffle(color_u4, color_u4, - VEC_LITERAL(vector unsigned char, - 0, 0, 0, 0, - 5, 5, 5, 5, - 10, 10, 10, 10, - 15, 15, 15, 15)); + ((vector unsigned char) { + 0, 0, 0, 0, + 5, 5, 5, 5, + 10, 10, 10, 10, + 15, 15, 15, 15}) ); return spu_convtf(color_u4, 32); } @@ -96,11 +97,11 @@ spu_unpack_A8R8G8B8(uint color) { vector unsigned int color_u4 = spu_splats(color); color_u4 = spu_shuffle(color_u4, color_u4, - VEC_LITERAL(vector unsigned char, - 5, 5, 5, 5, - 10, 10, 10, 10, - 15, 15, 15, 15, - 0, 0, 0, 0)); + ((vector unsigned char) { + 5, 5, 5, 5, + 10, 10, 10, 10, + 15, 15, 15, 15, + 0, 0, 0, 0}) ); return spu_convtf(color_u4, 32); } diff --git a/src/mesa/pipe/cell/spu/spu_ztest.h b/src/mesa/pipe/cell/spu/spu_ztest.h index 5fefb15176..ce8ad00339 100644 --- a/src/mesa/pipe/cell/spu/spu_ztest.h +++ b/src/mesa/pipe/cell/spu/spu_ztest.h @@ -68,9 +68,9 @@ spu_z16_test_less(vector float zvals, vector unsigned short *zbuf, /* gather lower four ushorts */ zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, (vector unsigned int) *zbuf, - VEC_LITERAL(vector unsigned char, - ZERO, ZERO, 8, 9, ZERO, ZERO, 10, 11, - ZERO, ZERO, 12, 13, ZERO, ZERO, 14, 15)); + ((vector unsigned char) { + ZERO, ZERO, 8, 9, ZERO, ZERO, 10, 11, + ZERO, ZERO, 12, 13, ZERO, ZERO, 14, 15})); /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ mask = spu_cmpgt(zbuf_ui4, zvals_ui4); /* mask &= inMask */ @@ -80,18 +80,18 @@ spu_z16_test_less(vector float zvals, vector unsigned short *zbuf, /* convert zbuffer values from uints back to ushorts, preserve lower 4 */ *zbuf = (vector unsigned short) spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, - VEC_LITERAL(vector unsigned char, - 16, 17, 18, 19, 20, 21, 22, 23, - 2, 3, 6, 7, 10, 11, 14, 15)); + ((vector unsigned char) { + 16, 17, 18, 19, 20, 21, 22, 23, + 2, 3, 6, 7, 10, 11, 14, 15})); } else { /* convert zbuffer values from ushorts to uints */ /* gather upper four ushorts */ zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, (vector unsigned int) *zbuf, - VEC_LITERAL(vector unsigned char, - ZERO, ZERO, 0, 1, ZERO, ZERO, 2, 3, - ZERO, ZERO, 4, 5, ZERO, ZERO, 6, 7)); + ((vector unsigned char) { + ZERO, ZERO, 0, 1, ZERO, ZERO, 2, 3, + ZERO, ZERO, 4, 5, ZERO, ZERO, 6, 7})); /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ mask = spu_cmpgt(zbuf_ui4, zvals_ui4); /* mask &= inMask */ @@ -101,9 +101,9 @@ spu_z16_test_less(vector float zvals, vector unsigned short *zbuf, /* convert zbuffer values from uints back to ushorts, preserve upper 4 */ *zbuf = (vector unsigned short) spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, - VEC_LITERAL(vector unsigned char, - 2, 3, 6, 7, 10, 11, 14, 15, - 24, 25, 26, 27, 28, 29, 30, 31)); + ((vector unsigned char) { + 2, 3, 6, 7, 10, 11, 14, 15, + 24, 25, 26, 27, 28, 29, 30, 31})); } return mask; #undef ZERO -- cgit v1.2.3 From 684d320ea2e7ec03d01275a544068cc6b45e1e9a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 18:03:05 -0700 Subject: Cell: don't use VEC_LITERAL macro, doesn't work w/ SDK 3.0 --- src/mesa/pipe/cell/spu/spu_texture.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c index 01ff33a857..3962aaa4a9 100644 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ b/src/mesa/pipe/cell/spu/spu_texture.c @@ -26,8 +26,6 @@ **************************************************************************/ -#include - #include "pipe/p_compiler.h" #include "spu_main.h" #include "spu_texture.h" @@ -57,7 +55,7 @@ invalidate_tex_cache(void) /* XXX memset? */ uint i; for (i = 0; i < CACHE_SIZE; i++) { - tex_tile_xy[i] = VEC_LITERAL(vector unsigned int, ~0U, ~0U, ~0U, ~0U); + tex_tile_xy[i] = ((vector unsigned int) { ~0U, ~0U, ~0U, ~0U }); } } -- cgit v1.2.3 From 5db1593c78192b764ad2ef7bdc5182d8ec4aed7c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 4 Feb 2008 18:05:37 -0700 Subject: Cell: fix some alignment issues by aligning commands to 8-byte boundaries Contributed by Ian Romanick. Also, temporarily disable inlined vertex buffers. They need to be 16-byte aligned... --- src/mesa/pipe/cell/common.h | 16 ++++---- src/mesa/pipe/cell/ppu/cell_batch.c | 4 +- src/mesa/pipe/cell/ppu/cell_flush.c | 2 +- src/mesa/pipe/cell/ppu/cell_state_emit.c | 3 +- src/mesa/pipe/cell/ppu/cell_vbuf.c | 4 +- src/mesa/pipe/cell/ppu/cell_vertex_shader.c | 22 ++++++----- src/mesa/pipe/cell/spu/spu_main.c | 58 +++++++++++++---------------- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 7 ++-- src/mesa/pipe/cell/spu/spu_vertex_shader.h | 2 +- 9 files changed, 57 insertions(+), 61 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index d861e82d33..cf8fc94ebf 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -57,6 +57,9 @@ /** round up value to next multiple of 4 */ #define ROUNDUP4(k) (((k) + 0x3) & ~0x3) +/** round up value to next multiple of 8 */ +#define ROUNDUP8(k) (((k) + 0x7) & ~0x7) + /** round up value to next multiple of 16 */ #define ROUNDUP16(k) (((k) + 0xf) & ~0xf) @@ -102,7 +105,7 @@ */ struct cell_command_framebuffer { - uint opcode; + uint64_t opcode; int width, height; void *color_start, *depth_start; enum pipe_format color_format, depth_format; @@ -114,7 +117,7 @@ struct cell_command_framebuffer */ struct cell_command_clear_surface { - uint opcode; + uint64_t opcode; uint surface; /**< Temporary: 0=color, 1=Z */ uint value; }; @@ -125,8 +128,7 @@ struct cell_command_clear_surface */ struct cell_array_info { - uint opcode; - uint base; /**< Base address of the 0th element. */ + uint64_t base; /**< Base address of the 0th element. */ uint attr; /**< Attribute that this state if for. */ uint pitch; /**< Byte pitch from one entry to the next. */ uint format; /**< Pipe format of each entry. */ @@ -150,7 +152,7 @@ struct cell_shader_info #define SPU_VERTS_PER_BATCH 64 struct cell_command_vs { - uint opcode; /**< CELL_CMD_VS_EXECUTE */ + uint64_t opcode; /**< CELL_CMD_VS_EXECUTE */ struct cell_shader_info shader; unsigned num_elts; unsigned elts[SPU_VERTS_PER_BATCH]; @@ -163,7 +165,7 @@ struct cell_command_vs struct cell_command_render { - uint opcode; /**< CELL_CMD_RENDER */ + uint64_t opcode; /**< CELL_CMD_RENDER */ uint prim_type; /**< PIPE_PRIM_x */ uint num_verts; uint vertex_size; /**< bytes per vertex */ @@ -179,7 +181,7 @@ struct cell_command_render struct cell_command_release_verts { - int opcode; /**< CELL_CMD_RELEASE_VERTS */ + uint64_t opcode; /**< CELL_CMD_RELEASE_VERTS */ uint vertex_buf; /**< in [0, CELL_NUM_BUFFERS-1] */ }; diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index 2d032fc902..2fb49711b2 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -136,7 +136,7 @@ cell_batch_append(struct cell_context *cell, const void *data, uint bytes) { uint size; - ASSERT(bytes % 4 == 0); + ASSERT(bytes % 8 == 0); ASSERT(bytes <= CELL_BUFFER_SIZE); ASSERT(cell->cur_batch >= 0); @@ -171,7 +171,7 @@ cell_batch_alloc(struct cell_context *cell, uint bytes) void *pos; uint size; - ASSERT(bytes % 4 == 0); + ASSERT(bytes % 8 == 0); ASSERT(bytes <= CELL_BUFFER_SIZE); assert(cell->cur_batch >= 0); diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c index cf4e676645..f62bc4650c 100644 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ b/src/mesa/pipe/cell/ppu/cell_flush.c @@ -59,7 +59,7 @@ cell_flush_int(struct pipe_context *pipe, unsigned flags) flushing = TRUE; if (flags & PIPE_FLUSH_WAIT) { - uint *cmd = (uint *) cell_batch_alloc(cell, sizeof(uint)); + uint64_t *cmd = (uint64_t *) cell_batch_alloc(cell, sizeof(uint64_t)); *cmd = CELL_CMD_FINISH; } diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c index 3b2670f786..5d2a786449 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.c +++ b/src/mesa/pipe/cell/ppu/cell_state_emit.c @@ -37,7 +37,8 @@ static void emit_state_cmd(struct cell_context *cell, uint cmd, const void *state, uint state_size) { - uint *dst = (uint *) cell_batch_alloc(cell, sizeof(uint) + state_size); + uint64_t *dst = (uint64_t *) + cell_batch_alloc(cell, ROUNDUP8(sizeof(uint64_t) + state_size)); *dst = cmd; memcpy(dst + 1, state, state_size); } diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index e63b34cf52..0fee61821a 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -40,7 +40,7 @@ /** Allow vertex data to be inlined after RENDER command */ -#define ALLOW_INLINE_VERTS 1 +#define ALLOW_INLINE_VERTS 0 /** @@ -197,7 +197,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, /* build/insert batch RENDER command */ { - const uint index_bytes = ROUNDUP4(nr_indices * 2); + const uint index_bytes = ROUNDUP8(nr_indices * 2); const uint vertex_bytes = nr_vertices * 4 * cell->vertex_info.size; const uint batch_size = sizeof(struct cell_command_render) diff --git a/src/mesa/pipe/cell/ppu/cell_vertex_shader.c b/src/mesa/pipe/cell/ppu/cell_vertex_shader.c index aef329a902..80dd500b34 100644 --- a/src/mesa/pipe/cell/ppu/cell_vertex_shader.c +++ b/src/mesa/pipe/cell/ppu/cell_vertex_shader.c @@ -52,8 +52,8 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) struct cell_context *const cell = (struct cell_context *) draw->driver_private; struct cell_command_vs *const vs = &cell_global.command[0].vs; - unsigned *batch; - struct cell_array_info array_info; + uint64_t *batch; + struct cell_array_info *array_info; unsigned i, j; assert(draw->vs.queue_nr != 0); @@ -63,17 +63,19 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) draw_update_vertex_fetch(draw); for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { - array_info.opcode = CELL_CMD_STATE_VS_ARRAY_INFO; - assert(draw->vertex_fetch.src_ptr[i] != NULL); - array_info.base = (uintptr_t) draw->vertex_fetch.src_ptr[i]; - array_info.attr = i; - array_info.pitch = draw->vertex_fetch.pitch[i]; - array_info.format = draw->vertex_element[i].src_format; + batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*array_info)); + + batch[0] = CELL_CMD_STATE_VS_ARRAY_INFO; - cell_batch_append(cell, & array_info, sizeof(array_info)); + array_info = (struct cell_array_info *) &batch[1]; + assert(draw->vertex_fetch.src_ptr[i] != NULL); + array_info->base = (uintptr_t) draw->vertex_fetch.src_ptr[i]; + array_info->attr = i; + array_info->pitch = draw->vertex_fetch.pitch[i]; + array_info->format = draw->vertex_element[i].src_format; } - batch = cell_batch_alloc(cell, sizeof(unsigned) + batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(struct pipe_viewport_state)); batch[0] = CELL_CMD_STATE_VIEWPORT; (void) memcpy(&batch[1], &draw->viewport, diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index b0311db1aa..4f126d5e5b 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -31,7 +31,6 @@ #include #include -#include #include "spu_main.h" #include "spu_render.h" @@ -220,13 +219,13 @@ cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) spu.fb.zsize = 0; if (spu.fb.color_format == PIPE_FORMAT_A8R8G8B8_UNORM) - spu.color_shuffle = VEC_LITERAL(vector unsigned char, - 12, 0, 4, 8, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0); + spu.color_shuffle = ((vector unsigned char) { + 12, 0, 4, 8, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0}); else if (spu.fb.color_format == PIPE_FORMAT_B8G8R8A8_UNORM) - spu.color_shuffle = VEC_LITERAL(vector unsigned char, - 8, 4, 0, 12, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0); + spu.color_shuffle = ((vector unsigned char) { + 8, 4, 0, 12, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0}); else ASSERT(0); } @@ -279,16 +278,10 @@ cmd_state_texture(const struct cell_command_texture *texture) spu.init.id, texture->start, texture->width, texture->height); memcpy(&spu.texture, texture, sizeof(*texture)); - spu.tex_size = VEC_LITERAL(vector float, - spu.texture.width, - spu.texture.height, - 0.0, - 0.0); - spu.tex_size_mask = VEC_LITERAL(vector unsigned int, - spu.texture.width - 1, - spu.texture.height - 1, - 0, - 0); + spu.tex_size = (vector float) + { spu.texture.width, spu.texture.height, 0.0, 0.0}; + spu.tex_size_mask = (vector unsigned int) + { spu.texture.width - 1, spu.texture.height - 1, 0, 0 }; } @@ -341,8 +334,8 @@ cmd_batch(uint opcode) { const uint buf = (opcode >> 8) & 0xff; uint size = (opcode >> 16); - uint buffer[CELL_BUFFER_SIZE / 4] ALIGN16_ATTRIB; - const uint usize = size / sizeof(uint); + uint64_t buffer[CELL_BUFFER_SIZE / 8] ALIGN16_ATTRIB; + const unsigned usize = size / sizeof(buffer[0]); uint pos; if (Debug) @@ -377,7 +370,7 @@ cmd_batch(uint opcode) struct cell_command_framebuffer *fb = (struct cell_command_framebuffer *) &buffer[pos]; cmd_state_framebuffer(fb); - pos += sizeof(*fb) / 4; + pos += sizeof(*fb) / 8; } break; case CELL_CMD_CLEAR_SURFACE: @@ -385,7 +378,7 @@ cmd_batch(uint opcode) struct cell_command_clear_surface *clr = (struct cell_command_clear_surface *) &buffer[pos]; cmd_clear_surface(clr); - pos += sizeof(*clr) / 4; + pos += sizeof(*clr) / 8; } break; case CELL_CMD_RENDER: @@ -394,7 +387,7 @@ cmd_batch(uint opcode) = (struct cell_command_render *) &buffer[pos]; uint pos_incr; cmd_render(render, &pos_incr); - pos += sizeof(*render) / 4 + pos_incr; + pos += sizeof(*render) / 8 + ((pos_incr + 1) / 2); } break; case CELL_CMD_RELEASE_VERTS: @@ -402,8 +395,7 @@ cmd_batch(uint opcode) struct cell_command_release_verts *release = (struct cell_command_release_verts *) &buffer[pos]; cmd_release_verts(release); - ASSERT(sizeof(*release) == 8); - pos += sizeof(*release) / 4; + pos += sizeof(*release) / 8; } break; case CELL_CMD_FINISH: @@ -413,36 +405,36 @@ cmd_batch(uint opcode) case CELL_CMD_STATE_BLEND: cmd_state_blend((struct pipe_blend_state *) &buffer[pos+1]); - pos += (1 + sizeof(struct pipe_blend_state) / 4); + pos += (1 + ROUNDUP8(sizeof(struct pipe_blend_state)) / 8); break; case CELL_CMD_STATE_DEPTH_STENCIL: cmd_state_depth_stencil((struct pipe_depth_stencil_alpha_state *) &buffer[pos+1]); - pos += (1 + sizeof(struct pipe_depth_stencil_alpha_state) / 4); + pos += (1 + ROUNDUP8(sizeof(struct pipe_depth_stencil_alpha_state)) / 8); break; case CELL_CMD_STATE_SAMPLER: cmd_state_sampler((struct pipe_sampler_state *) &buffer[pos+1]); - pos += (1 + sizeof(struct pipe_sampler_state) / 4); + pos += (1 + ROUNDUP8(sizeof(struct pipe_sampler_state)) / 8); break; case CELL_CMD_STATE_TEXTURE: cmd_state_texture((struct cell_command_texture *) &buffer[pos+1]); - pos += (1 + sizeof(struct cell_command_texture) / 4); + pos += (1 + ROUNDUP8(sizeof(struct cell_command_texture)) / 8); break; case CELL_CMD_STATE_VERTEX_INFO: cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]); - pos += (1 + sizeof(struct vertex_info) / 4); + pos += (1 + ROUNDUP8(sizeof(struct vertex_info)) / 8); break; case CELL_CMD_STATE_VIEWPORT: (void) memcpy(& draw.viewport, &buffer[pos+1], sizeof(struct pipe_viewport_state)); - pos += (1 + sizeof(struct pipe_viewport_state) / 4); + pos += (1 + ROUNDUP8(sizeof(struct pipe_viewport_state)) / 8); break; case CELL_CMD_STATE_VS_ARRAY_INFO: - cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos]); - pos += (sizeof(struct cell_array_info) / 4); + cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct cell_array_info)) / 8); break; default: - printf("SPU %u: bad opcode: 0x%x\n", spu.init.id, buffer[pos]); + printf("SPU %u: bad opcode: 0x%llx\n", spu.init.id, buffer[pos]); ASSERT(0); break; } diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index 1e846868e3..5b0f2a6470 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -431,9 +431,8 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, /* loop over vertex attributes (vertex shader inputs) */ for (attr = 0; attr < nr_attrs; attr++) { - - const unsigned pitch = draw->vertex_fetch.pitch[attr]; - const ubyte *src = draw->vertex_fetch.src_ptr[attr]; + const unsigned pitch = draw->vertex_fetch.pitch[attr]; + const uint64_t src = draw->vertex_fetch.src_ptr[attr]; const spu_fetch_func fetch = draw->vertex_fetch.fetch[attr]; unsigned i; float p[4][4]; @@ -447,7 +446,7 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, */ for (i = 0; i < count; i++) { uint8_t buffer[32] ALIGN16_ATTRIB; - const unsigned long addr = src + (elts[i] * pitch); + const uint64_t addr = src + (elts[i] * pitch); const unsigned size = ((addr & 0x0f) == 0) ? 16 : 32; mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.h b/src/mesa/pipe/cell/spu/spu_vertex_shader.h index c52f38fd02..b261ab44a2 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.h +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.h @@ -16,7 +16,7 @@ struct spu_vs_context { struct pipe_viewport_state viewport; struct { - const ubyte *src_ptr[PIPE_ATTRIB_MAX]; + uint64_t src_ptr[PIPE_ATTRIB_MAX]; unsigned pitch[PIPE_ATTRIB_MAX]; enum pipe_format format[PIPE_ATTRIB_MAX]; unsigned nr_attrs; -- cgit v1.2.3 From 45b18c51c0f49731cb8fc0144d678da5fa814992 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 5 Feb 2008 07:50:56 -0700 Subject: gallium: Use align_free to free aligned memory. --- src/mesa/pipe/draw/draw_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index 87f4969983..b15f57c824 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -106,7 +106,7 @@ void draw_destroy( struct draw_context *draw ) if (draw->pipeline.rasterize) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); tgsi_exec_machine_free_data(&draw->machine); - FREE( draw->vcache.vertex[0] ); /* Frees all the vertices. */ + align_free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ FREE( draw ); } -- cgit v1.2.3 From 8fc2355949b67cd99403c1184ce711a344877375 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 1 Feb 2008 14:58:38 -0800 Subject: Vectorize all micro ops Fold single instruction micro ops inline. Remove unused micro ops. --- src/mesa/pipe/cell/spu/spu_exec.c | 912 ++++++++++---------------------------- src/mesa/pipe/cell/spu/spu_exec.h | 1 + 2 files changed, 230 insertions(+), 683 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c index 168bada3bb..1ac9c031e3 100644 --- a/src/mesa/pipe/cell/spu/spu_exec.c +++ b/src/mesa/pipe/cell/spu/spu_exec.c @@ -52,8 +52,15 @@ #include #include -#include +#include +#include +#include +#include +#include #include +#include +#include +#include #include "pipe/p_compiler.h" #include "pipe/p_state.h" @@ -157,643 +164,175 @@ spu_exec_machine_init(struct spu_exec_machine *mach, } -static void -micro_abs( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - dst->f[0] = (float) fabs( (double) src->f[0] ); - dst->f[1] = (float) fabs( (double) src->f[1] ); - dst->f[2] = (float) fabs( (double) src->f[2] ); - dst->f[3] = (float) fabs( (double) src->f[3] ); -} - -static void -micro_add( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] + src1->f[0]; - dst->f[1] = src0->f[1] + src1->f[1]; - dst->f[2] = src0->f[2] + src1->f[2]; - dst->f[3] = src0->f[3] + src1->f[3]; -} - -static void -micro_iadd( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] + src1->i[0]; - dst->i[1] = src0->i[1] + src1->i[1]; - dst->i[2] = src0->i[2] + src1->i[2]; - dst->i[3] = src0->i[3] + src1->i[3]; -} - -static void -micro_and( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] & src1->u[0]; - dst->u[1] = src0->u[1] & src1->u[1]; - dst->u[2] = src0->u[2] & src1->u[2]; - dst->u[3] = src0->u[3] & src1->u[3]; -} - -static void -micro_ceil( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - ASSERT(0); -#if 0 - dst->f[0] = (float) ceil( (double) src->f[0] ); - dst->f[1] = (float) ceil( (double) src->f[1] ); - dst->f[2] = (float) ceil( (double) src->f[2] ); - dst->f[3] = (float) ceil( (double) src->f[3] ); -#endif -} - -static void -micro_cos( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - ASSERT(0); -#if 0 - dst->f[0] = (float) cos( (double) src->f[0] ); - dst->f[1] = (float) cos( (double) src->f[1] ); - dst->f[2] = (float) cos( (double) src->f[2] ); - dst->f[3] = (float) cos( (double) src->f[3] ); -#endif -} - -static void -micro_ddx( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_ddy( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_div( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] / src1->f[0]; - dst->f[1] = src0->f[1] / src1->f[1]; - dst->f[2] = src0->f[2] / src1->f[2]; - dst->f[3] = src0->f[3] / src1->f[3]; -} - -static void -micro_udiv( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] / src1->u[0]; - dst->u[1] = src0->u[1] / src1->u[1]; - dst->u[2] = src0->u[2] / src1->u[2]; - dst->u[3] = src0->u[3] / src1->u[3]; -} - -static void -micro_eq( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1, - const union spu_exec_channel *src2, - const union spu_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_ieq( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1, - const union spu_exec_channel *src2, - const union spu_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; -} - -static void -micro_exp2( - union spu_exec_channel *dst, - const union spu_exec_channel *src) -{ - ASSERT(0); -#if 0 - dst->f[0] = (float) pow( 2.0, (double) src->f[0] ); - dst->f[1] = (float) pow( 2.0, (double) src->f[1] ); - dst->f[2] = (float) pow( 2.0, (double) src->f[2] ); - dst->f[3] = (float) pow( 2.0, (double) src->f[3] ); -#endif -} - -static void -micro_f2it( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - dst->i[0] = (int) src->f[0]; - dst->i[1] = (int) src->f[1]; - dst->i[2] = (int) src->f[2]; - dst->i[3] = (int) src->f[3]; -} - -static void -micro_f2ut( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - dst->u[0] = (uint) src->f[0]; - dst->u[1] = (uint) src->f[1]; - dst->u[2] = (uint) src->f[2]; - dst->u[3] = (uint) src->f[3]; -} - -static void -micro_flr( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - ASSERT(0); -#if 0 - dst->f[0] = (float) floor( (double) src->f[0] ); - dst->f[1] = (float) floor( (double) src->f[1] ); - dst->f[2] = (float) floor( (double) src->f[2] ); - dst->f[3] = (float) floor( (double) src->f[3] ); -#endif -} - -static void -micro_frc( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - ASSERT(0); -#if 0 - dst->f[0] = src->f[0] - (float) floor( (double) src->f[0] ); - dst->f[1] = src->f[1] - (float) floor( (double) src->f[1] ); - dst->f[2] = src->f[2] - (float) floor( (double) src->f[2] ); - dst->f[3] = src->f[3] - (float) floor( (double) src->f[3] ); -#endif -} - -static void -micro_ge( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1, - const union spu_exec_channel *src2, - const union spu_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_i2f( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - dst->f[0] = (float) src->i[0]; - dst->f[1] = (float) src->i[1]; - dst->f[2] = (float) src->i[2]; - dst->f[3] = (float) src->i[3]; -} - -static void -micro_lg2( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - ASSERT(0); -#if 0 - dst->f[0] = (float) log( (double) src->f[0] ) * 1.442695f; - dst->f[1] = (float) log( (double) src->f[1] ) * 1.442695f; - dst->f[2] = (float) log( (double) src->f[2] ) * 1.442695f; - dst->f[3] = (float) log( (double) src->f[3] ) * 1.442695f; -#endif -} - -static void -micro_lt( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1, - const union spu_exec_channel *src2, - const union spu_exec_channel *src3 ) +static INLINE qword +micro_abs(qword src) { - dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; + return si_rotmi(si_shli(src, 1), -1); } -static void -micro_ilt( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1, - const union spu_exec_channel *src2, - const union spu_exec_channel *src3 ) +static INLINE qword +micro_ceil(qword src) { - dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; + return (qword) _ceilf4((vec_float4) src); } -static void -micro_ult( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1, - const union spu_exec_channel *src2, - const union spu_exec_channel *src3 ) +static INLINE qword +micro_cos(qword src) { - dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; + return (qword) _cosf4((vec_float4) src); } -static void -micro_max( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static const qword br_shuf = { + TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, + TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, + TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, + TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, + TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, + TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, + TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, + TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, +}; + +static const qword bl_shuf = { + TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, + TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, + TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, + TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, + TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, + TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, + TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, + TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, +}; + +static const qword tl_shuf = { + TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, + TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, + TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, + TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, + TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, + TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, + TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, + TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, +}; + +static qword +micro_ddx(qword src) { - dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; -} + qword bottom_right = si_shufb(src, src, br_shuf); + qword bottom_left = si_shufb(src, src, bl_shuf); -static void -micro_imax( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; + return si_fs(bottom_right, bottom_left); } -static void -micro_umax( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static qword +micro_ddy(qword src) { - dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; -} + qword top_left = si_shufb(src, src, tl_shuf); + qword bottom_left = si_shufb(src, src, bl_shuf); -static void -micro_min( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; + return si_fs(top_left, bottom_left); } -static void -micro_imin( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static INLINE qword +micro_div(qword src0, qword src1) { - dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; + return (qword) _divf4((vec_float4) src0, (vec_float4) src1); } -static void -micro_umin( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static qword +micro_flr(qword src) { - dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; + return (qword) _floorf4((vec_float4) src); } -static void -micro_umod( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static qword +micro_frc(qword src) { - dst->u[0] = src0->u[0] % src1->u[0]; - dst->u[1] = src0->u[1] % src1->u[1]; - dst->u[2] = src0->u[2] % src1->u[2]; - dst->u[3] = src0->u[3] % src1->u[3]; + return si_fs(src, (qword) _floorf4((vec_float4) src)); } -static void -micro_mul( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] * src1->f[0]; - dst->f[1] = src0->f[1] * src1->f[1]; - dst->f[2] = src0->f[2] * src1->f[2]; - dst->f[3] = src0->f[3] * src1->f[3]; -} - -static void -micro_imul( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] * src1->i[0]; - dst->i[1] = src0->i[1] * src1->i[1]; - dst->i[2] = src0->i[2] * src1->i[2]; - dst->i[3] = src0->i[3] * src1->i[3]; -} - -static void -micro_imul64( - union spu_exec_channel *dst0, - union spu_exec_channel *dst1, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst1->i[0] = src0->i[0] * src1->i[0]; - dst1->i[1] = src0->i[1] * src1->i[1]; - dst1->i[2] = src0->i[2] * src1->i[2]; - dst1->i[3] = src0->i[3] * src1->i[3]; - dst0->i[0] = 0; - dst0->i[1] = 0; - dst0->i[2] = 0; - dst0->i[3] = 0; -} - -static void -micro_umul64( - union spu_exec_channel *dst0, - union spu_exec_channel *dst1, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst1->u[0] = src0->u[0] * src1->u[0]; - dst1->u[1] = src0->u[1] * src1->u[1]; - dst1->u[2] = src0->u[2] * src1->u[2]; - dst1->u[3] = src0->u[3] * src1->u[3]; - dst0->u[0] = 0; - dst0->u[1] = 0; - dst0->u[2] = 0; - dst0->u[3] = 0; -} - -static void -micro_movc( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1, - const union spu_exec_channel *src2 ) -{ - dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; - dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; - dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; - dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; -} - -static void -micro_neg( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) +static INLINE qword +micro_ge(qword src0, qword src1) { - dst->f[0] = -src->f[0]; - dst->f[1] = -src->f[1]; - dst->f[2] = -src->f[2]; - dst->f[3] = -src->f[3]; + return si_or(si_fceq(src0, src1), si_fcgt(src0, src1)); } -static void -micro_ineg( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) +static qword +micro_lg2(qword src) { - dst->i[0] = -src->i[0]; - dst->i[1] = -src->i[1]; - dst->i[2] = -src->i[2]; - dst->i[3] = -src->i[3]; + return (qword) _log2f4((vec_float4) src); } -static void -micro_not( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) +static INLINE qword +micro_lt(qword src0, qword src1) { - dst->u[0] = ~src->u[0]; - dst->u[1] = ~src->u[1]; - dst->u[2] = ~src->u[2]; - dst->u[3] = ~src->u[3]; -} + const qword tmp = si_or(si_fceq(src0, src1), si_fcgt(src0, src1)); -static void -micro_or( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] | src1->u[0]; - dst->u[1] = src0->u[1] | src1->u[1]; - dst->u[2] = src0->u[2] | src1->u[2]; - dst->u[3] = src0->u[3] | src1->u[3]; + return si_xori(tmp, 0xff); } -static void -micro_pow( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static INLINE qword +micro_max(qword src0, qword src1) { - vec_float4 s0 = (vec_float4) { - src0->f[0], src0->f[1], src0->f[2], src0->f[3] - }; - vec_float4 s1 = (vec_float4) { - src1->f[0], src1->f[1], src1->f[2], src1->f[3] - }; - vec_float4 d = _powf4(s0, s1); - - dst->f[0] = spu_extract(d, 0); - dst->f[1] = spu_extract(d, 1); - dst->f[2] = spu_extract(d, 2); - dst->f[3] = spu_extract(d, 3); + return si_selb(src1, src0, si_fcgt(src0, src1)); } -static void -micro_rnd( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) +static INLINE qword +micro_min(qword src0, qword src1) { - ASSERT(0); -#if 0 - dst->f[0] = (float) floor( (double) (src->f[0] + 0.5f) ); - dst->f[1] = (float) floor( (double) (src->f[1] + 0.5f) ); - dst->f[2] = (float) floor( (double) (src->f[2] + 0.5f) ); - dst->f[3] = (float) floor( (double) (src->f[3] + 0.5f) ); -#endif + return si_selb(src0, src1, si_fcgt(src0, src1)); } -static void -micro_shl( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static qword +micro_neg(qword src) { - dst->i[0] = src0->i[0] << src1->i[0]; - dst->i[1] = src0->i[1] << src1->i[1]; - dst->i[2] = src0->i[2] << src1->i[2]; - dst->i[3] = src0->i[3] << src1->i[3]; + return si_xor(src, (qword) spu_splats(0x80000000)); } -static void -micro_ishr( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static qword +micro_set_sign(qword src) { - dst->i[0] = src0->i[0] >> src1->i[0]; - dst->i[1] = src0->i[1] >> src1->i[1]; - dst->i[2] = src0->i[2] >> src1->i[2]; - dst->i[3] = src0->i[3] >> src1->i[3]; + return si_or(src, (qword) spu_splats(0x80000000)); } -static void -micro_trunc( - union spu_exec_channel *dst, - const union spu_exec_channel *src0 ) +static qword +micro_pow(qword src0, qword src1) { - dst->f[0] = (float) (int) src0->f[0]; - dst->f[1] = (float) (int) src0->f[1]; - dst->f[2] = (float) (int) src0->f[2]; - dst->f[3] = (float) (int) src0->f[3]; + return (qword) _powf4((vec_float4) src0, (vec_float4) src1); } -static void -micro_ushr( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static qword +micro_rnd(qword src) { - dst->u[0] = src0->u[0] >> src1->u[0]; - dst->u[1] = src0->u[1] >> src1->u[1]; - dst->u[2] = src0->u[2] >> src1->u[2]; - dst->u[3] = src0->u[3] >> src1->u[3]; -} + const qword half = (qword) spu_splats(0.5f); -static void -micro_sin( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) -{ - ASSERT(0); -#if 0 - dst->f[0] = (float) sin( (double) src->f[0] ); - dst->f[1] = (float) sin( (double) src->f[1] ); - dst->f[2] = (float) sin( (double) src->f[2] ); - dst->f[3] = (float) sin( (double) src->f[3] ); -#endif + /* May be able to use _roundf4. There may be some difference, though. + */ + return (qword) _floorf4((vec_float4) si_fa(src, half)); } -static void -micro_sqrt( union spu_exec_channel *dst, - const union spu_exec_channel *src ) +static INLINE qword +micro_ishr(qword src0, qword src1) { - vec_float4 s = (vec_float4) { - src->f[0], src->f[1], src->f[2], src->f[3] - }; - vec_float4 d = _sqrtf4(s); - - dst->f[0] = spu_extract(d, 0); - dst->f[1] = spu_extract(d, 1); - dst->f[2] = spu_extract(d, 2); - dst->f[3] = spu_extract(d, 3); + return si_rotma(src0, si_sfi(src1, 0)); } -static void -micro_sub( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static qword +micro_trunc(qword src) { - dst->f[0] = src0->f[0] - src1->f[0]; - dst->f[1] = src0->f[1] - src1->f[1]; - dst->f[2] = src0->f[2] - src1->f[2]; - dst->f[3] = src0->f[3] - src1->f[3]; + return (qword) _truncf4((vec_float4) src); } -static void -micro_u2f( - union spu_exec_channel *dst, - const union spu_exec_channel *src ) +static qword +micro_sin(qword src) { - dst->f[0] = (float) src->u[0]; - dst->f[1] = (float) src->u[1]; - dst->f[2] = (float) src->u[2]; - dst->f[3] = (float) src->u[3]; + return (qword) _sinf4((vec_float4) src); } -static void -micro_xor( - union spu_exec_channel *dst, - const union spu_exec_channel *src0, - const union spu_exec_channel *src1 ) +static INLINE qword +micro_sqrt(qword src) { - dst->u[0] = src0->u[0] ^ src1->u[0]; - dst->u[1] = src0->u[1] ^ src1->u[1]; - dst->u[2] = src0->u[2] ^ src1->u[2]; - dst->u[3] = src0->u[3] ^ src1->u[3]; + return (qword) _sqrtf4((vec_float4) src); } static void @@ -983,16 +522,15 @@ fetch_source( switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { case TGSI_UTIL_SIGN_CLEAR: - micro_abs( chan, chan ); + chan->q = micro_abs(chan->q); break; case TGSI_UTIL_SIGN_SET: - micro_abs( chan, chan ); - micro_neg( chan, chan ); + chan->q = micro_set_sign(chan->q); break; case TGSI_UTIL_SIGN_TOGGLE: - micro_neg( chan, chan ); + chan->q = micro_neg(chan->q); break; case TGSI_UTIL_SIGN_KEEP: @@ -1000,7 +538,7 @@ fetch_source( } if (reg->SrcRegisterExtMod.Complement) { - micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan ); + chan->q = si_fs(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, chan->q); } } @@ -1051,8 +589,8 @@ store_dest( case TGSI_SAT_ZERO_ONE: /* XXX need to obey ExecMask here */ - micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); - micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); + dst->q = micro_max(chan->q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); + dst->q = micro_min(dst->q, mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q); break; case TGSI_SAT_MINUS_PLUS_ONE: @@ -1162,7 +700,7 @@ exec_tex(struct spu_exec_machine *mach, switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { case TGSI_EXTSWIZZLE_W: FETCH(&r[1], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[1] ); + r[0].q = micro_div(r[0].q, r[1].q); break; case TGSI_EXTSWIZZLE_ONE: @@ -1194,9 +732,9 @@ exec_tex(struct spu_exec_machine *mach, switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { case TGSI_EXTSWIZZLE_W: FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); + r[0].q = micro_div(r[0].q, r[3].q); + r[1].q = micro_div(r[1].q, r[3].q); + r[2].q = micro_div(r[2].q, r[3].q); break; case TGSI_EXTSWIZZLE_ONE: @@ -1228,9 +766,9 @@ exec_tex(struct spu_exec_machine *mach, switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { case TGSI_EXTSWIZZLE_W: FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); + r[0].q = micro_div(r[0].q, r[3].q); + r[1].q = micro_div(r[1].q, r[3].q); + r[2].q = micro_div(r[2].q, r[3].q); break; case TGSI_EXTSWIZZLE_ONE: @@ -1389,7 +927,7 @@ exec_instruction( case TGSI_OPCODE_ARL: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_f2it( &r[0], &r[0] ); + r[0].q = si_cflts(r[0].q, 0); STORE( &r[0], 0, chan_index ); } break; @@ -1409,22 +947,27 @@ exec_instruction( if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { FETCH( &r[0], 0, CHAN_X ); - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + r[0].q = micro_max(r[0].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); STORE( &r[0], 0, CHAN_Y ); } - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[1], 0, CHAN_Y ); - micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - - FETCH( &r[2], 0, CHAN_W ); - micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); - micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); - micro_pow( &r[1], &r[1], &r[2] ); - micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Z ); - } + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[1], 0, CHAN_Y ); + r[1].q = micro_max(r[1].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); + + FETCH( &r[2], 0, CHAN_W ); + r[2].q = micro_min(r[2].q, mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].q); + r[2].q = micro_max(r[2].q, mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].q); + r[1].q = micro_pow(r[1].q, r[2].q); + + /* r0 = (r0 > 0.0) ? r1 : 0.0 + */ + r[0].q = si_fcgt(r[0].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); + r[0].q = si_selb(mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q, r[1].q, + r[0].q); + STORE( &r[0], 0, CHAN_Z ); + } } if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { @@ -1435,7 +978,7 @@ exec_instruction( case TGSI_OPCODE_RCP: /* TGSI_OPCODE_RECIP */ FETCH( &r[0], 0, CHAN_X ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); } @@ -1444,8 +987,8 @@ exec_instruction( case TGSI_OPCODE_RSQ: /* TGSI_OPCODE_RECIPSQRT */ FETCH( &r[0], 0, CHAN_X ); - micro_sqrt( &r[0], &r[0] ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + r[0].q = micro_sqrt(r[0].q); + r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); } @@ -1465,7 +1008,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - micro_mul( &r[0], &r[0], &r[1] ); + r[0].q = si_fm(r[0].q, r[1].q); STORE(&r[0], 0, chan_index); } @@ -1475,7 +1018,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fa(r[0].q, r[1].q); STORE( &r[0], 0, chan_index ); } break; @@ -1484,17 +1027,16 @@ exec_instruction( /* TGSI_OPCODE_DOT3 */ FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); + r[0].q = si_fm(r[0].q, r[1].q); FETCH( &r[1], 0, CHAN_Y ); FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); + FETCH( &r[1], 0, CHAN_Z ); FETCH( &r[2], 1, CHAN_Z ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1506,25 +1048,22 @@ exec_instruction( FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); - micro_mul( &r[0], &r[0], &r[1] ); + r[0].q = si_fm(r[0].q, r[1].q); FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 1, CHAN_Y); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); FETCH(&r[1], 0, CHAN_Z); FETCH(&r[2], 1, CHAN_Z); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); FETCH(&r[1], 0, CHAN_W); FETCH(&r[2], 1, CHAN_W); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1539,7 +1078,7 @@ exec_instruction( if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { FETCH( &r[0], 0, CHAN_Y ); FETCH( &r[1], 1, CHAN_Y); - micro_mul( &r[0], &r[0], &r[1] ); + r[0].q = si_fm(r[0].q, r[1].q); STORE( &r[0], 0, CHAN_Y ); } @@ -1559,8 +1098,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - /* XXX use micro_min()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); + r[0].q = micro_min(r[0].q, r[1].q); STORE(&r[0], 0, chan_index); } @@ -1571,8 +1109,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - /* XXX use micro_max()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); + r[0].q = micro_max(r[0].q, r[1].q); STORE(&r[0], 0, chan_index ); } @@ -1583,7 +1120,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + + r[0].q = micro_ge(r[0].q, r[1].q); + r[0].q = si_xori(r[0].q, 0xff); + STORE( &r[0], 0, chan_index ); } break; @@ -1593,7 +1133,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + r[0].q = micro_ge(r[0].q, r[1].q); STORE( &r[0], 0, chan_index ); } break; @@ -1603,9 +1143,8 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_mul( &r[0], &r[0], &r[1] ); - FETCH( &r[1], 2, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); + FETCH( &r[2], 2, chan_index ); + r[0].q = si_fma(r[0].q, r[1].q, r[2].q); STORE( &r[0], 0, chan_index ); } break; @@ -1615,7 +1154,7 @@ exec_instruction( FETCH(&r[0], 0, chan_index); FETCH(&r[1], 1, chan_index); - micro_sub( &r[0], &r[0], &r[1] ); + r[0].q = si_fs(r[0].q, r[1].q); STORE(&r[0], 0, chan_index); } @@ -1628,9 +1167,8 @@ exec_instruction( FETCH(&r[1], 1, chan_index); FETCH(&r[2], 2, chan_index); - micro_sub( &r[1], &r[1], &r[2] ); - micro_mul( &r[0], &r[0], &r[1] ); - micro_add( &r[0], &r[0], &r[2] ); + r[1].q = si_fs(r[1].q, r[2].q); + r[0].q = si_fma(r[0].q, r[1].q, r[2].q); STORE(&r[0], 0, chan_index); } @@ -1661,7 +1199,7 @@ exec_instruction( /* TGSI_OPCODE_FRC */ FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_frc( &r[0], &r[0] ); + r[0].q = micro_frc(r[0].q); STORE( &r[0], 0, chan_index ); } break; @@ -1674,7 +1212,7 @@ exec_instruction( /* TGSI_OPCODE_FLR */ FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_flr( &r[0], &r[0] ); + r[0].q = micro_flr(r[0].q); STORE( &r[0], 0, chan_index ); } break; @@ -1682,7 +1220,7 @@ exec_instruction( case TGSI_OPCODE_ROUND: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_rnd( &r[0], &r[0] ); + r[0].q = micro_rnd(r[0].q); STORE( &r[0], 0, chan_index ); } break; @@ -1691,7 +1229,7 @@ exec_instruction( /* TGSI_OPCODE_EX2 */ FETCH(&r[0], 0, CHAN_X); - micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); + r[0].q = micro_pow(mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q, r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1701,7 +1239,7 @@ exec_instruction( case TGSI_OPCODE_LOGBASE2: /* TGSI_OPCODE_LG2 */ FETCH( &r[0], 0, CHAN_X ); - micro_lg2( &r[0], &r[0] ); + r[0].q = micro_lg2(r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); } @@ -1712,7 +1250,7 @@ exec_instruction( FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); - micro_pow( &r[0], &r[0], &r[1] ); + r[0].q = micro_pow(r[0].q, r[1].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1723,35 +1261,34 @@ exec_instruction( /* TGSI_OPCODE_XPD */ FETCH(&r[0], 0, CHAN_Y); FETCH(&r[1], 1, CHAN_Z); - - micro_mul( &r[2], &r[0], &r[1] ); - FETCH(&r[3], 0, CHAN_Z); FETCH(&r[4], 1, CHAN_Y); - micro_mul( &r[5], &r[3], &r[4] ); - micro_sub( &r[2], &r[2], &r[5] ); + /* r2 = (r0 * r1) - (r3 * r5) + */ + r[2].q = si_fm(r[3].q, r[5].q); + r[2].q = si_fms(r[0].q, r[1].q, r[2].q); if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { STORE( &r[2], 0, CHAN_X ); } FETCH(&r[2], 1, CHAN_X); - - micro_mul( &r[3], &r[3], &r[2] ); - FETCH(&r[5], 0, CHAN_X); - micro_mul( &r[1], &r[1], &r[5] ); - micro_sub( &r[3], &r[3], &r[1] ); + /* r3 = (r3 * r2) - (r1 * r5) + */ + r[1].q = si_fm(r[1].q, r[5].q); + r[3].q = si_fms(r[3].q, r[2].q, r[1].q); if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { STORE( &r[3], 0, CHAN_Y ); } - micro_mul( &r[5], &r[5], &r[4] ); - micro_mul( &r[0], &r[0], &r[2] ); - micro_sub( &r[5], &r[5], &r[0] ); + /* r5 = (r5 * r4) - (r0 * r2) + */ + r[0].q = si_fm(r[0].q, r[2].q); + r[5].q = si_fms(r[5].q, r[4].q, r[0].q); if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { STORE( &r[5], 0, CHAN_Z ); @@ -1770,7 +1307,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH(&r[0], 0, chan_index); - micro_abs( &r[0], &r[0] ); + r[0].q = micro_abs(r[0].q); STORE(&r[0], 0, chan_index); } @@ -1784,23 +1321,21 @@ exec_instruction( FETCH(&r[0], 0, CHAN_X); FETCH(&r[1], 1, CHAN_X); - micro_mul( &r[0], &r[0], &r[1] ); + r[0].q = si_fm(r[0].q, r[1].q); FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 1, CHAN_Y); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); FETCH(&r[1], 0, CHAN_Z); FETCH(&r[2], 1, CHAN_Z); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); FETCH(&r[1], 1, CHAN_W); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fa(r[0].q, r[1].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1810,7 +1345,7 @@ exec_instruction( case TGSI_OPCODE_COS: FETCH(&r[0], 0, CHAN_X); - micro_cos( &r[0], &r[0] ); + r[0].q = micro_cos(r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -1820,7 +1355,7 @@ exec_instruction( case TGSI_OPCODE_DDX: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_ddx( &r[0], &r[0] ); + r[0].q = micro_ddx(r[0].q); STORE( &r[0], 0, chan_index ); } break; @@ -1828,7 +1363,7 @@ exec_instruction( case TGSI_OPCODE_DDY: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_ddy( &r[0], &r[0] ); + r[0].q = micro_ddy(r[0].q); STORE( &r[0], 0, chan_index ); } break; @@ -1866,9 +1401,9 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], - &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], - &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + + r[0].q = si_fceq(r[0].q, r[1].q); + STORE( &r[0], 0, chan_index ); } break; @@ -1881,14 +1416,14 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + r[0].q = si_fcgt(r[0].q, r[1].q); STORE( &r[0], 0, chan_index ); } break; case TGSI_OPCODE_SIN: FETCH( &r[0], 0, CHAN_X ); - micro_sin( &r[0], &r[0] ); + r[0].q = micro_sin(r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); } @@ -1898,7 +1433,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + + r[0].q = si_fcgt(r[0].q, r[1].q); + r[0].q = si_xori(r[0].q, 0xff); + STORE( &r[0], 0, chan_index ); } break; @@ -1907,7 +1445,10 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + + r[0].q = si_fceq(r[0].q, r[1].q); + r[0].q = si_xori(r[0].q, 0xff); + STORE( &r[0], 0, chan_index ); } break; @@ -2038,7 +1579,11 @@ exec_instruction( FETCH(&r[1], 1, chan_index); FETCH(&r[2], 2, chan_index); - micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); + /* r0 = (r0 < 0.0) ? r1 : r2 + */ + r[3].q = si_xor(r[3].q, r[3].q); + r[0].q = micro_lt(r[0].q, r[3].q); + r[0].q = si_selb(r[1].q, r[2].q, r[0].q); STORE(&r[0], 0, chan_index); } @@ -2049,11 +1594,11 @@ exec_instruction( FETCH( &r[0], 0, CHAN_X ); } if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - micro_cos( &r[1], &r[0] ); + r[1].q = micro_cos(r[0].q); STORE( &r[1], 0, CHAN_X ); } if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - micro_sin( &r[1], &r[0] ); + r[1].q = micro_sin(r[0].q); STORE( &r[1], 0, CHAN_Y ); } if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { @@ -2075,12 +1620,11 @@ exec_instruction( case TGSI_OPCODE_DP2: FETCH( &r[0], 0, CHAN_X ); FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); + r[0].q = si_fm(r[0].q, r[1].q); FETCH( &r[1], 0, CHAN_Y ); FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); @@ -2152,7 +1696,7 @@ exec_instruction( case TGSI_OPCODE_CEIL: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_ceil( &r[0], &r[0] ); + r[0].q = micro_ceil(r[0].q); STORE( &r[0], 0, chan_index ); } break; @@ -2160,7 +1704,7 @@ exec_instruction( case TGSI_OPCODE_I2F: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_i2f( &r[0], &r[0] ); + r[0].q = si_csflt(r[0].q, 0); STORE( &r[0], 0, chan_index ); } break; @@ -2168,7 +1712,7 @@ exec_instruction( case TGSI_OPCODE_NOT: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_not( &r[0], &r[0] ); + r[0].q = si_xorbi(r[0].q, 0xff); STORE( &r[0], 0, chan_index ); } break; @@ -2176,7 +1720,7 @@ exec_instruction( case TGSI_OPCODE_TRUNC: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); - micro_trunc( &r[0], &r[0] ); + r[0].q = micro_trunc(r[0].q); STORE( &r[0], 0, chan_index ); } break; @@ -2185,7 +1729,9 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_shl( &r[0], &r[0], &r[1] ); + + r[0].q = si_shl(r[0].q, r[1].q); + STORE( &r[0], 0, chan_index ); } break; @@ -2194,7 +1740,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_ishr( &r[0], &r[0], &r[1] ); + r[0].q = micro_ishr(r[0].q, r[1].q); STORE( &r[0], 0, chan_index ); } break; @@ -2203,7 +1749,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_and( &r[0], &r[0], &r[1] ); + r[0].q = si_and(r[0].q, r[1].q); STORE( &r[0], 0, chan_index ); } break; @@ -2212,7 +1758,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_or( &r[0], &r[0], &r[1] ); + r[0].q = si_or(r[0].q, r[1].q); STORE( &r[0], 0, chan_index ); } break; @@ -2225,7 +1771,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_xor( &r[0], &r[0], &r[1] ); + r[0].q = si_xor(r[0].q, r[1].q); STORE( &r[0], 0, chan_index ); } break; diff --git a/src/mesa/pipe/cell/spu/spu_exec.h b/src/mesa/pipe/cell/spu/spu_exec.h index 89e422ba48..b4c7661ef6 100644 --- a/src/mesa/pipe/cell/spu/spu_exec.h +++ b/src/mesa/pipe/cell/spu/spu_exec.h @@ -43,6 +43,7 @@ union spu_exec_channel float f[QUAD_SIZE]; int i[QUAD_SIZE]; unsigned u[QUAD_SIZE]; + qword q; }; /** -- cgit v1.2.3 From 490a7b1c73babd528b6d883471a8636157c5853a Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 1 Feb 2008 17:12:20 -0800 Subject: Vectorize vertex puller --- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 186 +++++++++-------------------- src/mesa/pipe/cell/spu/spu_vertex_shader.h | 4 +- 2 files changed, 61 insertions(+), 129 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index 5b0f2a6470..4133fbba17 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -42,6 +42,8 @@ #define DRAW_DBG 0 +static const vec_float4 defaults = { 0.0, 0.0, 0.0, 1.0 }; + /** * Fetch a float[4] vertex attribute from memory, doing format/type * conversion as needed. @@ -50,19 +52,16 @@ * conversion, texture sampling etc. */ #define FETCH_ATTRIB( NAME, SZ, CVT ) \ -static void \ -fetch_##NAME(const void *ptr, float *attrib) \ +static qword \ +fetch_##NAME(const void *ptr) \ { \ - static const float defaults[4] = { 0,0,0,1 }; \ + vec_float4 attrib = defaults; \ int i; \ \ for (i = 0; i < SZ; i++) { \ - attrib[i] = CVT; \ - } \ - \ - for (; i < 4; i++) { \ - attrib[i] = defaults[i]; \ + attrib = spu_insert(CVT, attrib, i); \ } \ + return (qword) attrib; \ } #define CVT_64_FLOAT (float) ((double *) ptr)[i] @@ -309,106 +308,59 @@ static spu_fetch_func get_fetch_func( enum pipe_format format ) } -static void -transpose_4x4( float *out, const float *in ) -{ - /* This can be achieved in 12 sse instructions, plus the final - * stores I guess. This is probably a bit more than that - maybe - * 32 or so? - */ - out[0] = in[0]; out[1] = in[4]; out[2] = in[8]; out[3] = in[12]; - out[4] = in[1]; out[5] = in[5]; out[6] = in[9]; out[7] = in[13]; - out[8] = in[2]; out[9] = in[6]; out[10] = in[10]; out[11] = in[14]; - out[12] = in[3]; out[13] = in[7]; out[14] = in[11]; out[15] = in[15]; -} - - - -static void fetch_xyz_rgb( struct spu_vs_context *draw, - struct spu_exec_machine *machine, - const unsigned *elts, - unsigned count ) +void +spu_transpose_4x4(qword *out, const qword *in) { - assert(count <= 4); - -// _mesa_printf("%s\n", __FUNCTION__); - - /* loop over vertex attributes (vertex shader inputs) - */ - - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - - for (i = 0; i < 4; i++) { + static const qword masks[8] = { { - const float *in = (const float *)(src[0] + elts[i] * pitch[0]); - float *out = &machine->Inputs[0].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - + 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + }, { - const float *in = (const float *)(src[1] + elts[i] * pitch[1]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - } -} - - - - -static void fetch_xyz_rgb_st( struct spu_vs_context *draw, - struct spu_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - assert(count <= 4); - - /* loop over vertex attributes (vertex shader inputs) - */ - - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - - for (i = 0; i < 4; i++) { + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, + }, + + { + 0x04, 0x05, 0x06, 0x07, 0x14, 0x15, 0x16, 0x17, + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + }, + { + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + 0x04, 0x05, 0x06, 0x07, 0x14, 0x15, 0x16, 0x17, + }, + + { + 0x08, 0x09, 0x0a, 0x0b, 0x18, 0x19, 0x1a, 0x1b, + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + }, + { + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + 0x08, 0x09, 0x0a, 0x0b, 0x18, 0x19, 0x1a, 0x1b, + }, + + { + 0x0c, 0x0d, 0x0e, 0x0f, 0x1c, 0x1d, 0x1e, 0x1f, + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + }, { - const float *in = (const float *)(src[0] + elts[i] * pitch[0]); - float *out = &machine->Inputs[0].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + 0x0c, 0x0d, 0x0e, 0x0f, 0x1c, 0x1d, 0x1e, 0x1f, + }, + }; - { - const float *in = (const float *)(src[1] + elts[i] * pitch[1]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } + out[0] = si_shufb(in[0], in[1], masks[0]); + out[0] = si_or(out[0], si_shufb(in[2], in[3], masks[1])); - { - const float *in = (const float *)(src[2] + elts[i] * pitch[2]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = 0.0f; - out[12] = 1.0f; - } - } -} + out[1] = si_shufb(in[0], in[1], masks[2]); + out[1] = si_or(out[1], si_shufb(in[2], in[3], masks[3])); + out[2] = si_shufb(in[0], in[1], masks[4]); + out[2] = si_or(out[2], si_shufb(in[2], in[3], masks[5])); + out[3] = si_shufb(in[0], in[1], masks[6]); + out[3] = si_or(out[3], si_shufb(in[2], in[3], masks[7])); +} /** @@ -435,7 +387,7 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, const uint64_t src = draw->vertex_fetch.src_ptr[attr]; const spu_fetch_func fetch = draw->vertex_fetch.fetch[attr]; unsigned i; - float p[4][4]; + qword p[4]; /* Fetch four attributes for four vertices. @@ -452,17 +404,15 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); wait_on_mask(1 << TAG_VERTEX_BUFFER); - memmove(& buffer, buffer + (addr & 0x0f), 16); - - fetch(buffer, p[i]); + p[i] = (*fetch)(buffer + (addr & 0x0f)); } /* Be nice and zero out any missing vertices: */ for (/* empty */; i < 4; i++) - p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; - - /* Transpose/swizzle into sse-friendly format. Currently + p[i] = si_xor(p[i], p[i]); + + /* Transpose/swizzle into vector-friendly format. Currently * assuming that all vertex shader inputs are float[4], but this * isn't true -- if the vertex shader only wants tex0.xy, we * could optimize for that. @@ -471,7 +421,7 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, * excessive number of fetch functions, but we could at least * minimize the transpose step: */ - transpose_4x4( (float *)&machine->Inputs[attr].xyzw[0].f[0], (float *)p ); + spu_transpose_4x4(&machine->Inputs[attr].xyzw[0].q, p); } } @@ -487,24 +437,4 @@ void spu_update_vertex_fetch( struct spu_vs_context *draw ) } draw->vertex_fetch.fetch_func = generic_vertex_fetch; - - /* Disable the fast path because they don't use mfc_get yet. - */ -#if 0 - switch (draw->vertex_fetch.nr_attrs) { - case 2: - if (draw->vertex_fetch.format[0] == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_fetch.format[1] == PIPE_FORMAT_R32G32B32_FLOAT) - draw->vertex_fetch.fetch_func = fetch_xyz_rgb; - break; - case 3: - if (draw->vertex_fetch.format[0] == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_fetch.format[1] == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_fetch.format[2] == PIPE_FORMAT_R32G32_FLOAT) - draw->vertex_fetch.fetch_func = fetch_xyz_rgb_st; - break; - default: - break; - } -#endif } diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.h b/src/mesa/pipe/cell/spu/spu_vertex_shader.h index b261ab44a2..2435b7ddae 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.h +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.h @@ -6,7 +6,7 @@ struct spu_vs_context; -typedef void (*spu_fetch_func)(const void *ptr, float *attrib); +typedef qword (*spu_fetch_func)(const void *ptr); typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw, struct spu_exec_machine *machine, const unsigned *elts, @@ -39,6 +39,8 @@ struct spu_vs_context { extern void spu_update_vertex_fetch(struct spu_vs_context *draw); +extern void spu_transpose_4x4(qword *out, const qword *in); + static INLINE void spu_vertex_fetch(struct spu_vs_context *draw, struct spu_exec_machine *machine, const unsigned *elts, -- cgit v1.2.3 From e8a80c8627972537c595f06fb28cd383569e7ea0 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 1 Feb 2008 17:14:09 -0800 Subject: More semi-trivial vectorization in the shader VM --- src/mesa/pipe/cell/spu/spu_exec.c | 62 +++++++++++++++++---------------------- 1 file changed, 27 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c index 1ac9c031e3..1bd8687d41 100644 --- a/src/mesa/pipe/cell/spu/spu_exec.c +++ b/src/mesa/pipe/cell/spu/spu_exec.c @@ -70,6 +70,7 @@ #include "pipe/tgsi/util/tgsi_util.h" #include "spu_exec.h" #include "spu_main.h" +#include "spu_vertex_shader.h" #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 @@ -144,23 +145,27 @@ spu_exec_machine_init(struct spu_exec_machine *mach, struct spu_sampler *samplers, unsigned processor) { + qword zero; + qword not_zero; uint i; mach->Samplers = samplers; mach->Processor = processor; mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; + zero = si_xor(zero, zero); + not_zero = si_xori(zero, 0xff); + /* Setup constants. */ - for( i = 0; i < 4; i++ ) { - mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; - mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; - mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; - mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; - mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; - mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; - mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; - mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; - } + mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q = zero; + mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].q = not_zero; + mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].q = si_shli(not_zero, -1); + mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].q = si_shli(not_zero, 31); + + mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q = (qword) spu_splats(1.0f); + mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q = (qword) spu_splats(2.0f); + mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].q = (qword) spu_splats(128.0f); + mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].q = (qword) spu_splats(-128.0f); } @@ -459,25 +464,16 @@ fetch_source( &index2, &indir_index ); - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; + index.q = si_a(index.q, indir_index.q); } if( reg->SrcRegister.Dimension ) { switch( reg->SrcRegister.File ) { case TGSI_FILE_INPUT: - index.i[0] *= 17; - index.i[1] *= 17; - index.i[2] *= 17; - index.i[3] *= 17; + index.q = si_mpyi(index.q, 17); break; case TGSI_FILE_CONSTANT: - index.i[0] *= 4096; - index.i[1] *= 4096; - index.i[2] *= 4096; - index.i[3] *= 4096; + index.q = si_shli(index.q, 12); break; default: assert( 0 ); @@ -505,10 +501,7 @@ fetch_source( &index2, &indir_index ); - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; + index.q = si_a(index.q, indir_index.q); } } @@ -666,17 +659,16 @@ fetch_texel( struct spu_sampler *sampler, union spu_exec_channel *b, union spu_exec_channel *a ) { - uint j; - float rgba[NUM_CHANNELS][QUAD_SIZE]; + qword rgba[4]; + qword out[4]; - sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); + sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, (float *) rgba); - for (j = 0; j < 4; j++) { - r->f[j] = rgba[0][j]; - g->f[j] = rgba[1][j]; - b->f[j] = rgba[2][j]; - a->f[j] = rgba[3][j]; - } + spu_transpose_4x4(out, rgba); + r->q = out[0]; + g->q = out[1]; + b->q = out[2]; + a->q = out[3]; } -- cgit v1.2.3 From 45f4125fa83c4e43a01d44cb8eb2a4c97b72181f Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 4 Feb 2008 16:03:55 -0800 Subject: Add some debug messages --- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index 4133fbba17..cfa449e813 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -378,7 +378,10 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, wait_on_mask(1 << TAG_VERTEX_BUFFER); -// _mesa_printf("%s %d\n", __FUNCTION__, count); +#if DRAW_DBG + printf("SPU: %s count = %u, nr_attrs = %u\n", + __FUNCTION__, count, nr_attrs); +#endif /* loop over vertex attributes (vertex shader inputs) */ @@ -401,6 +404,9 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, const uint64_t addr = src + (elts[i] * pitch); const unsigned size = ((addr & 0x0f) == 0) ? 16 : 32; +#if DRAW_DBG + printf("SPU: fetching = 0x%llx\n", addr); +#endif mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); wait_on_mask(1 << TAG_VERTEX_BUFFER); -- cgit v1.2.3 From c9f98142b6a47825c49aea72a79c1be62c2b7d89 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 5 Feb 2008 09:43:52 -0800 Subject: Use _transpose_matrix4x4 from Cell SDK instead of my own version --- src/mesa/pipe/cell/spu/spu_exec.c | 3 +- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 59 ++---------------------------- src/mesa/pipe/cell/spu/spu_vertex_shader.h | 2 - 3 files changed, 5 insertions(+), 59 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c index 1bd8687d41..e51008b9b3 100644 --- a/src/mesa/pipe/cell/spu/spu_exec.c +++ b/src/mesa/pipe/cell/spu/spu_exec.c @@ -52,6 +52,7 @@ #include #include +#include #include #include #include @@ -664,7 +665,7 @@ fetch_texel( struct spu_sampler *sampler, sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, (float *) rgba); - spu_transpose_4x4(out, rgba); + _transpose_matrix4x4(out, rgba); r->q = out[0]; g->q = out[1]; b->q = out[2]; diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index cfa449e813..6e86a919ce 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -31,6 +31,8 @@ */ #include +#include + #include "pipe/p_util.h" #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" @@ -308,61 +310,6 @@ static spu_fetch_func get_fetch_func( enum pipe_format format ) } -void -spu_transpose_4x4(qword *out, const qword *in) -{ - static const qword masks[8] = { - { - 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - }, - { - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, - }, - - { - 0x04, 0x05, 0x06, 0x07, 0x14, 0x15, 0x16, 0x17, - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - }, - { - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - 0x04, 0x05, 0x06, 0x07, 0x14, 0x15, 0x16, 0x17, - }, - - { - 0x08, 0x09, 0x0a, 0x0b, 0x18, 0x19, 0x1a, 0x1b, - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - }, - { - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - 0x08, 0x09, 0x0a, 0x0b, 0x18, 0x19, 0x1a, 0x1b, - }, - - { - 0x0c, 0x0d, 0x0e, 0x0f, 0x1c, 0x1d, 0x1e, 0x1f, - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - }, - { - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - 0x0c, 0x0d, 0x0e, 0x0f, 0x1c, 0x1d, 0x1e, 0x1f, - }, - }; - - out[0] = si_shufb(in[0], in[1], masks[0]); - out[0] = si_or(out[0], si_shufb(in[2], in[3], masks[1])); - - out[1] = si_shufb(in[0], in[1], masks[2]); - out[1] = si_or(out[1], si_shufb(in[2], in[3], masks[3])); - - out[2] = si_shufb(in[0], in[1], masks[4]); - out[2] = si_or(out[2], si_shufb(in[2], in[3], masks[5])); - - out[3] = si_shufb(in[0], in[1], masks[6]); - out[3] = si_or(out[3], si_shufb(in[2], in[3], masks[7])); -} - - /** * Fetch vertex attributes for 'count' vertices. */ @@ -427,7 +374,7 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, * excessive number of fetch functions, but we could at least * minimize the transpose step: */ - spu_transpose_4x4(&machine->Inputs[attr].xyzw[0].q, p); + _transpose_matrix4x4(&machine->Inputs[attr].xyzw[0].q, p); } } diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.h b/src/mesa/pipe/cell/spu/spu_vertex_shader.h index 2435b7ddae..c96b93ff0a 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.h +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.h @@ -39,8 +39,6 @@ struct spu_vs_context { extern void spu_update_vertex_fetch(struct spu_vs_context *draw); -extern void spu_transpose_4x4(qword *out, const qword *in); - static INLINE void spu_vertex_fetch(struct spu_vs_context *draw, struct spu_exec_machine *machine, const unsigned *elts, -- cgit v1.2.3 From b0974420f4dab55d398f4015cf71a62fa643f713 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 14:21:01 -0700 Subject: Cell: added cell_batch_alloc_aligned() --- src/mesa/pipe/cell/ppu/cell_batch.c | 26 ++++++++++++++++++++------ src/mesa/pipe/cell/ppu/cell_batch.h | 4 ++++ 2 files changed, 24 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c index 2fb49711b2..f45e5f25b6 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ b/src/mesa/pipe/cell/ppu/cell_batch.c @@ -157,7 +157,7 @@ cell_batch_append(struct cell_context *cell, const void *data, uint bytes) size = 0; } - assert(size + bytes <= CELL_BUFFER_SIZE); + ASSERT(size + bytes <= CELL_BUFFER_SIZE); memcpy(cell->buffer[cell->cur_batch] + size, data, bytes); @@ -167,14 +167,22 @@ cell_batch_append(struct cell_context *cell, const void *data, uint bytes) void * cell_batch_alloc(struct cell_context *cell, uint bytes) +{ + return cell_batch_alloc_aligned(cell, bytes, 1); +} + + +void * +cell_batch_alloc_aligned(struct cell_context *cell, uint bytes, + uint alignment) { void *pos; - uint size; + uint size, padbytes; ASSERT(bytes % 8 == 0); ASSERT(bytes <= CELL_BUFFER_SIZE); - - assert(cell->cur_batch >= 0); + ASSERT(alignment > 0); + ASSERT(cell->cur_batch >= 0); #ifdef ASSERT { @@ -188,12 +196,18 @@ cell_batch_alloc(struct cell_context *cell, uint bytes) size = cell->buffer_size[cell->cur_batch]; - if (size + bytes > CELL_BUFFER_SIZE) { + padbytes = (alignment - (size % alignment)) % alignment; + + if (padbytes + size + bytes > CELL_BUFFER_SIZE) { cell_batch_flush(cell); size = 0; } + else { + size += padbytes; + } - assert(size + bytes <= CELL_BUFFER_SIZE); + ASSERT(size % alignment == 0); + ASSERT(size + bytes <= CELL_BUFFER_SIZE); pos = (void *) (cell->buffer[cell->cur_batch] + size); diff --git a/src/mesa/pipe/cell/ppu/cell_batch.h b/src/mesa/pipe/cell/ppu/cell_batch.h index f4f37314a4..a6eee0a8b1 100644 --- a/src/mesa/pipe/cell/ppu/cell_batch.h +++ b/src/mesa/pipe/cell/ppu/cell_batch.h @@ -50,5 +50,9 @@ cell_batch_append(struct cell_context *cell, const void *data, uint bytes); extern void * cell_batch_alloc(struct cell_context *cell, uint bytes); +extern void * +cell_batch_alloc_aligned(struct cell_context *cell, uint bytes, + uint alignment); + #endif /* CELL_BATCH_H */ -- cgit v1.2.3 From 2174890ed030bde8494b7f13b7090e27771695fa Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 14:21:48 -0700 Subject: Cell: remove dummy fields, update/add some comments --- src/mesa/pipe/cell/common.h | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h index cf8fc94ebf..4de514c358 100644 --- a/src/mesa/pipe/cell/common.h +++ b/src/mesa/pipe/cell/common.h @@ -51,7 +51,7 @@ /** for sanity checking */ #define ASSERT_ALIGN16(ptr) \ - assert((((unsigned long) (ptr)) & 0xf) == 0); + ASSERT((((unsigned long) (ptr)) & 0xf) == 0); /** round up value to next multiple of 4 */ @@ -105,7 +105,7 @@ */ struct cell_command_framebuffer { - uint64_t opcode; + uint64_t opcode; /**< CELL_CMD_FRAMEBUFFER */ int width, height; void *color_start, *depth_start; enum pipe_format color_format, depth_format; @@ -117,7 +117,7 @@ struct cell_command_framebuffer */ struct cell_command_clear_surface { - uint64_t opcode; + uint64_t opcode; /**< CELL_CMD_CLEAR_SURFACE */ uint surface; /**< Temporary: 0=color, 1=Z */ uint value; }; @@ -128,8 +128,8 @@ struct cell_command_clear_surface */ struct cell_array_info { - uint64_t base; /**< Base address of the 0th element. */ - uint attr; /**< Attribute that this state if for. */ + uint64_t base; /**< Base address of the 0th element. */ + uint attr; /**< Attribute that this state is for. */ uint pitch; /**< Byte pitch from one entry to the next. */ uint format; /**< Pipe format of each entry. */ } ALIGN16_ATTRIB; @@ -169,11 +169,9 @@ struct cell_command_render uint prim_type; /**< PIPE_PRIM_x */ uint num_verts; uint vertex_size; /**< bytes per vertex */ - uint dummy; /* XXX this dummy field works around a compiler bug */ uint num_indexes; uint vertex_buf; /**< which cell->buffer[] contains the vertex data */ - float xmin, dummy2, ymin, xmax, ymax; /* XXX another dummy field */ - uint dummy3; + float xmin, ymin, xmax, ymax; /* XXX another dummy field */ uint min_index; boolean inline_verts; }; -- cgit v1.2.3 From 4da82fd5c5e0a7535e30aa81f08dcbe1a26358b7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 14:23:34 -0700 Subject: Cell: re-enable inlined vertex buffers Vertex data must be on a 16-byte address/offset so SIMD operations will work properly in the SPU code. --- src/mesa/pipe/cell/ppu/cell_vbuf.c | 12 +++++------- src/mesa/pipe/cell/spu/spu_main.c | 3 ++- src/mesa/pipe/cell/spu/spu_render.c | 12 ++++++++---- 3 files changed, 15 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c index 0fee61821a..e9fafe492e 100644 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ b/src/mesa/pipe/cell/ppu/cell_vbuf.c @@ -40,7 +40,7 @@ /** Allow vertex data to be inlined after RENDER command */ -#define ALLOW_INLINE_VERTS 0 +#define ALLOW_INLINE_VERTS 1 /** @@ -199,9 +199,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, { const uint index_bytes = ROUNDUP8(nr_indices * 2); const uint vertex_bytes = nr_vertices * 4 * cell->vertex_info.size; - - const uint batch_size = sizeof(struct cell_command_render) - + index_bytes; + const uint batch_size = sizeof(struct cell_command_render) + index_bytes; struct cell_command_render *render = (struct cell_command_render *) @@ -223,9 +221,9 @@ cell_vbuf_draw(struct vbuf_render *vbr, render->num_verts = nr_vertices; if (ALLOW_INLINE_VERTS && min_index == 0 && - vertex_bytes <= cell_batch_free_space(cell)) { - /* vertex data inlined, after indices */ - void *dst = cell_batch_alloc(cell, vertex_bytes); + vertex_bytes + 16 <= cell_batch_free_space(cell)) { + /* vertex data inlined, after indices, at 16-byte boundary */ + void *dst = cell_batch_alloc_aligned(cell, vertex_bytes, 16); memcpy(dst, vertices, vertex_bytes); render->inline_verts = TRUE; render->vertex_buf = ~0; diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c index 4f126d5e5b..e375197fe6 100644 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ b/src/mesa/pipe/cell/spu/spu_main.c @@ -387,7 +387,7 @@ cmd_batch(uint opcode) = (struct cell_command_render *) &buffer[pos]; uint pos_incr; cmd_render(render, &pos_incr); - pos += sizeof(*render) / 8 + ((pos_incr + 1) / 2); + pos += pos_incr; } break; case CELL_CMD_RELEASE_VERTS: @@ -541,6 +541,7 @@ main(main_param_t speid, main_param_t argp) (void) speid; ASSERT(sizeof(tile_t) == TILE_SIZE * TILE_SIZE * 4); + ASSERT(sizeof(struct cell_command_render) % 8 == 0); one_time_init(); diff --git a/src/mesa/pipe/cell/spu/spu_render.c b/src/mesa/pipe/cell/spu/spu_render.c index e8705eeeba..932fb500b3 100644 --- a/src/mesa/pipe/cell/spu/spu_render.c +++ b/src/mesa/pipe/cell/spu/spu_render.c @@ -171,6 +171,7 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) ubyte vertex_data[CELL_BUFFER_SIZE] ALIGN16_ATTRIB; const uint vertex_size = render->vertex_size; /* in bytes */ /*const*/ uint total_vertex_bytes = render->num_verts * vertex_size; + uint index_bytes; const ubyte *vertices; const ushort *indexes; uint i, j; @@ -199,13 +200,16 @@ cmd_render(const struct cell_command_render *render, uint *pos_incr) /* indexes are right after the render command in the batch buffer */ indexes = (const ushort *) (render + 1); - *pos_incr = (render->num_indexes * 2 + 3) / 4; + index_bytes = ROUNDUP8(render->num_indexes * 2); + *pos_incr = index_bytes / 8 + sizeof(*render) / 8; if (render->inline_verts) { - /* Vertices are right after indexes in batch buffer */ - vertices = (const ubyte *) (render + 1) + *pos_incr * 4; - *pos_incr = *pos_incr + total_vertex_bytes / 4; + /* Vertices are after indexes in batch buffer at next 16-byte addr */ + vertices = (const ubyte *) render + (*pos_incr * 8); + vertices = (const ubyte *) align_pointer((void *) vertices, 16); + ASSERT_ALIGN16(vertices); + *pos_incr = ((vertices + total_vertex_bytes) - (ubyte *) render) / 8; } else { /* Begin DMA fetch of vertex buffer */ -- cgit v1.2.3 From 1730f7bad462ac7f29857b8b2347e38c1b6c9820 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 15:07:36 -0700 Subject: Cell: SIMD-ize tri_linear_coeff(), use vector float for vertex attributes in struct vertex_header --- src/mesa/pipe/cell/spu/spu_tri.c | 112 ++++++++++++++++++++++++++------------- 1 file changed, 75 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c index 688c8646ab..be9624cf7d 100644 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ b/src/mesa/pipe/cell/spu/spu_tri.c @@ -56,7 +56,7 @@ typedef union * Simplified types taken from other parts of Gallium */ struct vertex_header { - float data[0][4]; + vector float data[1]; }; @@ -476,6 +476,7 @@ static void print_vertex(const struct vertex_header *v) } #endif + static boolean setup_sort_vertices(const struct vertex_header *v0, const struct vertex_header *v1, const struct vertex_header *v2) @@ -492,9 +493,9 @@ static boolean setup_sort_vertices(const struct vertex_header *v0, /* determine bottom to top order of vertices */ { - float y0 = v0->data[0][1]; - float y1 = v1->data[0][1]; - float y2 = v2->data[0][1]; + float y0 = spu_extract(v0->data[0], 1); + float y1 = spu_extract(v1->data[0], 1); + float y2 = spu_extract(v2->data[0], 1); if (y0 <= y1) { if (y1 <= y2) { /* y0<=y1<=y2 */ @@ -538,25 +539,25 @@ static boolean setup_sort_vertices(const struct vertex_header *v0, } /* Check if triangle is completely outside the tile bounds */ - if (setup.vmin->data[0][1] > setup.cliprect_maxy) + if (spu_extract(setup.vmin->data[0], 1) > setup.cliprect_maxy) return FALSE; - if (setup.vmax->data[0][1] < setup.cliprect_miny) + if (spu_extract(setup.vmax->data[0], 1) < setup.cliprect_miny) return FALSE; - if (setup.vmin->data[0][0] < setup.cliprect_minx && - setup.vmid->data[0][0] < setup.cliprect_minx && - setup.vmax->data[0][0] < setup.cliprect_minx) + if (spu_extract(setup.vmin->data[0], 0) < setup.cliprect_minx && + spu_extract(setup.vmid->data[0], 0) < setup.cliprect_minx && + spu_extract(setup.vmax->data[0], 0) < setup.cliprect_minx) return FALSE; - if (setup.vmin->data[0][0] > setup.cliprect_maxx && - setup.vmid->data[0][0] > setup.cliprect_maxx && - setup.vmax->data[0][0] > setup.cliprect_maxx) + if (spu_extract(setup.vmin->data[0], 0) > setup.cliprect_maxx && + spu_extract(setup.vmid->data[0], 0) > setup.cliprect_maxx && + spu_extract(setup.vmax->data[0], 0) > setup.cliprect_maxx) return FALSE; - setup.ebot.dx = setup.vmid->data[0][0] - setup.vmin->data[0][0]; - setup.ebot.dy = setup.vmid->data[0][1] - setup.vmin->data[0][1]; - setup.emaj.dx = setup.vmax->data[0][0] - setup.vmin->data[0][0]; - setup.emaj.dy = setup.vmax->data[0][1] - setup.vmin->data[0][1]; - setup.etop.dx = setup.vmax->data[0][0] - setup.vmid->data[0][0]; - setup.etop.dy = setup.vmax->data[0][1] - setup.vmid->data[0][1]; + setup.ebot.dx = spu_extract(setup.vmid->data[0], 0) - spu_extract(setup.vmin->data[0], 0); + setup.ebot.dy = spu_extract(setup.vmid->data[0], 1) - spu_extract(setup.vmin->data[0], 1); + setup.emaj.dx = spu_extract(setup.vmax->data[0], 0) - spu_extract(setup.vmin->data[0], 0); + setup.emaj.dy = spu_extract(setup.vmax->data[0], 1) - spu_extract(setup.vmin->data[0], 1); + setup.etop.dx = spu_extract(setup.vmax->data[0], 0) - spu_extract(setup.vmid->data[0], 0); + setup.etop.dy = spu_extract(setup.vmax->data[0], 1) - spu_extract(setup.vmid->data[0], 1); /* * Compute triangle's area. Use 1/area to compute partial @@ -597,14 +598,12 @@ static boolean setup_sort_vertices(const struct vertex_header *v0, * The result will be put into setup.coef[slot].a0. * \param slot which attribute slot */ -static INLINE void const_coeff(uint slot) +static INLINE void +const_coeff(uint slot) { setup.coef[slot].dadx.v = (vector float) {0.0, 0.0, 0.0, 0.0}; setup.coef[slot].dady.v = (vector float) {0.0, 0.0, 0.0, 0.0}; - setup.coef[slot].a0.f[0] = setup.vprovoke->data[slot][0]; - setup.coef[slot].a0.f[1] = setup.vprovoke->data[slot][1]; - setup.coef[slot].a0.f[2] = setup.vprovoke->data[slot][2]; - setup.coef[slot].a0.f[3] = setup.vprovoke->data[slot][3]; + setup.coef[slot].a0.v = setup.vprovoke->data[slot]; } @@ -612,12 +611,19 @@ static INLINE void const_coeff(uint slot) * Compute a0, dadx and dady for a linearly interpolated coefficient, * for a triangle. */ -static void tri_linear_coeff( uint slot, uint firstComp, uint lastComp ) +static INLINE void +tri_linear_coeff(uint slot, uint firstComp, uint lastComp) { uint i; + const float *vmin_d = (float *) &setup.vmin->data[slot]; + const float *vmid_d = (float *) &setup.vmid->data[slot]; + const float *vmax_d = (float *) &setup.vmax->data[slot]; + const float x = spu_extract(setup.vmin->data[0], 0) - 0.5f; + const float y = spu_extract(setup.vmin->data[0], 1) - 0.5f; + for (i = firstComp; i < lastComp; i++) { - float botda = setup.vmid->data[slot][i] - setup.vmin->data[slot][i]; - float majda = setup.vmax->data[slot][i] - setup.vmin->data[slot][i]; + float botda = vmid_d[i] - vmin_d[i]; + float majda = vmax_d[i] - vmin_d[i]; float a = setup.ebot.dy * majda - botda * setup.emaj.dy; float b = setup.emaj.dx * botda - majda * setup.ebot.dx; @@ -638,9 +644,9 @@ static void tri_linear_coeff( uint slot, uint firstComp, uint lastComp ) * to define a0 as the sample at a pixel center somewhere near vmin * instead - i'll switch to this later. */ - setup.coef[slot].a0.f[i] = (setup.vmin->data[slot][i] - - (setup.coef[slot].dadx.f[i] * (setup.vmin->data[0][0] - 0.5f) + - setup.coef[slot].dady.f[i] * (setup.vmin->data[0][1] - 0.5f))); + setup.coef[slot].a0.f[i] = (vmin_d[i] - + (setup.coef[slot].dadx.f[i] * x + + setup.coef[slot].dady.f[i] * y)); } /* @@ -653,6 +659,37 @@ static void tri_linear_coeff( uint slot, uint firstComp, uint lastComp ) } +/** + * As above, but interp setup all four vector components. + */ +static INLINE void +tri_linear_coeff4(uint slot) +{ + const vector float vmin_d = setup.vmin->data[slot]; + const vector float vmid_d = setup.vmid->data[slot]; + const vector float vmax_d = setup.vmax->data[slot]; + const vector float xxxx = spu_splats(spu_extract(setup.vmin->data[0], 0) - 0.5f); + const vector float yyyy = spu_splats(spu_extract(setup.vmin->data[0], 1) - 0.5f); + + vector float botda = vmid_d - vmin_d; + vector float majda = vmax_d - vmin_d; + + vector float a = spu_sub(spu_mul(spu_splats(setup.ebot.dy), majda), + spu_mul(botda, spu_splats(setup.emaj.dy))); + vector float b = spu_sub(spu_mul(spu_splats(setup.emaj.dx), botda), + spu_mul(majda, spu_splats(setup.ebot.dx))); + + setup.coef[slot].dadx.v = spu_mul(a, spu_splats(setup.oneoverarea)); + setup.coef[slot].dady.v = spu_mul(b, spu_splats(setup.oneoverarea)); + + vector float tempx = spu_mul(setup.coef[slot].dadx.v, xxxx); + vector float tempy = spu_mul(setup.coef[slot].dady.v, yyyy); + + setup.coef[slot].a0.v = spu_sub(vmin_d, spu_add(tempx, tempy)); +} + + + #if 0 /** * Compute a0, dadx and dady for a perspective-corrected interpolant, @@ -710,17 +747,18 @@ static void setup_tri_coefficients(void) case INTERP_NONE: break; case INTERP_POS: - tri_linear_coeff(i, 2, 3); + /*tri_linear_coeff(i, 2, 3);*/ /* XXX interp W if PERSPECTIVE... */ + tri_linear_coeff4(i); break; case INTERP_CONSTANT: const_coeff(i); break; case INTERP_LINEAR: - tri_linear_coeff(i, 0, 4); + tri_linear_coeff4(i); break; case INTERP_PERSPECTIVE: - tri_linear_coeff(i, 0, 4); /* XXX temporary */ + tri_linear_coeff4(i); /* temporary */ break; default: ASSERT(0); @@ -738,12 +776,12 @@ static void setup_tri_coefficients(void) static void setup_tri_edges(void) { - float vmin_x = setup.vmin->data[0][0] + 0.5f; - float vmid_x = setup.vmid->data[0][0] + 0.5f; + float vmin_x = spu_extract(setup.vmin->data[0], 0) + 0.5f; + float vmid_x = spu_extract(setup.vmid->data[0], 0) + 0.5f; - float vmin_y = setup.vmin->data[0][1] - 0.5f; - float vmid_y = setup.vmid->data[0][1] - 0.5f; - float vmax_y = setup.vmax->data[0][1] - 0.5f; + float vmin_y = spu_extract(setup.vmin->data[0], 1) - 0.5f; + float vmid_y = spu_extract(setup.vmid->data[0], 1) - 0.5f; + float vmax_y = spu_extract(setup.vmax->data[0], 1) - 0.5f; setup.emaj.sy = CEILF(vmin_y); setup.emaj.lines = (int) CEILF(vmax_y - setup.emaj.sy); -- cgit v1.2.3 From e39fccc34c07a015d8713841a69037e32187dd6d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 15:12:18 -0700 Subject: Cell: remove accidentally added OPT_FLAGS lines --- src/mesa/pipe/cell/spu/Makefile | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile index 66f16cde9b..f202971d73 100644 --- a/src/mesa/pipe/cell/spu/Makefile +++ b/src/mesa/pipe/cell/spu/Makefile @@ -8,8 +8,6 @@ TOP = ../../../../.. include $(TOP)/configs/linux-cell -OPT_FLAGS=-g -OPT_FLAGS=-O3 PROG = g3d PROG_SPU = $(PROG)_spu -- cgit v1.2.3 From 5e2d0517b1b42c6f94fa69bf4e32a19d00fd519f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 4 Feb 2008 17:02:08 +0900 Subject: gallium: Portability guidelines. --- src/mesa/pipe/README.portability | 43 ++++++++++++++++++++++++++++++++++++++++ 1 file changed, 43 insertions(+) create mode 100644 src/mesa/pipe/README.portability (limited to 'src') diff --git a/src/mesa/pipe/README.portability b/src/mesa/pipe/README.portability new file mode 100644 index 0000000000..c70ca774da --- /dev/null +++ b/src/mesa/pipe/README.portability @@ -0,0 +1,43 @@ + CROSS-PLATFORM PORTABILITY GUIDELINES FOR GALLIUM3D + + += General Considerations = + +The state tracker and winsys driver support a rather limited number of +platforms. However, the pipe drivers are meant to run in a wide number of +platforms. Hence the pipe drivers, the auxiliary modules, and all public +headers in general, should stricly follow these guidelines to ensure + + += Compiler Support = + +* Include the p_compiler.h. + +* Don't use the 'inline' keyword, use the INLINE macro in p_compiler.h instead. + +* Cast explicitly when converting to integer types of smaller sizes. + +* Cast explicitly when converting between float, double and integral types. + +* Don't use named struct initializers. + +* Don't use variable number of macro arguments. Use static inline functions +instead. + + += Standard Library = + +* Avoid including standard library headers. Most standard library functions are +not available in Windows Kernel Mode. Use the appropriate p_*.h include. + +== Memory Allocation == + +* Use MALLOC, CALLOC, FREE instead of the malloc, calloc, free functions. + +* Use align_pointer() function defined in p_util.h for aligning pointers in a +portable way. + +== Debugging == + +TODO + -- cgit v1.2.3 From a31d289de6091987e6b9da6af1b2e56eb79a96fb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 6 Feb 2008 13:27:49 +0900 Subject: gallium: Cross-platform debugging helpers. --- src/mesa/SConscript | 1 + src/mesa/pipe/p_debug.h | 79 ++++++++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/util/p_debug.c | 70 +++++++++++++++++++++++++++++++++++++++ src/mesa/sources | 1 + 4 files changed, 151 insertions(+) create mode 100644 src/mesa/pipe/p_debug.h create mode 100644 src/mesa/pipe/util/p_debug.c (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index 70a98f3129..faf8c84872 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -158,6 +158,7 @@ STATECACHE_SOURCES = [ ] PIPEUTIL_SOURCES = [ + 'pipe/util/p_debug.c', 'pipe/util/p_tile.c', 'pipe/util/p_util.c', ] diff --git a/src/mesa/pipe/p_debug.h b/src/mesa/pipe/p_debug.h new file mode 100644 index 0000000000..b037eba2a3 --- /dev/null +++ b/src/mesa/pipe/p_debug.h @@ -0,0 +1,79 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Cross-platform debugging helpers. + * + * For now it just has assert and printf replacements, but it might be extended + * with stack trace reports and more advanced logging in the near future. + * + * @author Jose Fonseca + */ + +#ifndef P_DEBUG_H_ +#define P_DEBUG_H_ + +#ifdef __cplusplus +extern "C" { +#endif + + +#ifdef DBG +#ifndef DEBUG +#define DEBUG 1 +#endif +#else +#ifndef NDEBUG +#define NDEBUG 1 +#endif +#endif + + +void debug_printf(const char *format, ...); +void debug_assert_fail(const char *expr, const char *file, unsigned line); + +/** Assert macro */ +#ifdef DEBUG +#define debug_assert(expr) ((expr) ? (void)0 : debug_assert_fail(#expr, __FILE__, __LINE__)) +#else +#define debug_assert(expr) ((void)0) +#endif + + +#ifdef assert +#warning Standard C Library assert macro usage detected. +#undef assert +#endif +#define assert(expr) debug_assert(expr) + + +#ifdef __cplusplus +} +#endif + +#endif /* P_DEBUG_H_ */ diff --git a/src/mesa/pipe/util/p_debug.c b/src/mesa/pipe/util/p_debug.c new file mode 100644 index 0000000000..faa093e57c --- /dev/null +++ b/src/mesa/pipe/util/p_debug.c @@ -0,0 +1,70 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include + +#ifdef WIN32 +#include +#include +#else +#include +#include +#endif + +#include "pipe/p_debug.h" +#include "pipe/p_compiler.h" + + +void debug_printf(const char *format, ...) +{ + va_list ap; + va_start( ap, format ); +#ifdef WIN32 + EngDebugPrint("Gallium3D: ", (PCHAR)format, ap); +#else + vfprintf(stderr, format, ap); +#endif + va_end( ap ); +} + + +static INLINE debug_abort(void) +{ +#ifdef WIN32 + EngDebugBreak(); +#else + abort(); +#endif +} + + +void debug_assert_fail(const char *expr, const char *file, unsigned line) +{ + debug_printf("%s:%i: Assertion `%s' failed."); + debug_abort(); +} diff --git a/src/mesa/sources b/src/mesa/sources index e31d8cc466..c0087f76e6 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -195,6 +195,7 @@ STATECACHE_SOURCES = \ pipe/cso_cache/cso_cache.c PIPEUTIL_SOURCES = \ + pipe/util/p_debug.c \ pipe/util/p_tile.c \ pipe/util/p_util.c -- cgit v1.2.3 From 78bce9c2dcd45d1d8706bb9bab3b3a73943de990 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 6 Feb 2008 14:37:24 +0900 Subject: gallium: Add forgotten return type. --- src/mesa/pipe/util/p_debug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/util/p_debug.c b/src/mesa/pipe/util/p_debug.c index faa093e57c..9303c970cc 100644 --- a/src/mesa/pipe/util/p_debug.c +++ b/src/mesa/pipe/util/p_debug.c @@ -53,7 +53,7 @@ void debug_printf(const char *format, ...) } -static INLINE debug_abort(void) +static INLINE void debug_abort(void) { #ifdef WIN32 EngDebugBreak(); -- cgit v1.2.3 From 9791d7f64c5a58b9c1bf32d00c71e0e031f54f70 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 6 Feb 2008 14:37:49 +0900 Subject: gallium: Use p_debug.h instead of non-portable stdio.h/assert.h functions. --- src/mesa/pipe/draw/draw_prim.c | 18 +++---- src/mesa/pipe/draw/draw_vbuf.c | 63 ++++++++++++------------- src/mesa/pipe/draw/draw_vertex_fetch.c | 8 ++-- src/mesa/pipe/draw/draw_vertex_shader.c | 6 +-- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 4 +- src/mesa/pipe/draw/draw_vf.c | 4 +- src/mesa/pipe/draw/draw_vf_generic.c | 3 +- src/mesa/pipe/draw/draw_vf_sse.c | 6 +-- src/mesa/pipe/i915simple/i915_fpc_translate.c | 14 +++--- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 3 +- src/mesa/pipe/i915simple/i915_state_derived.c | 2 +- src/mesa/pipe/i915simple/i915_state_emit.c | 2 +- src/mesa/pipe/i915simple/i915_state_immediate.c | 2 +- src/mesa/pipe/i915simple/i915_state_sampler.c | 2 +- src/mesa/pipe/i965simple/brw_cc.c | 2 +- src/mesa/pipe/i965simple/brw_curbe.c | 6 +-- src/mesa/pipe/i965simple/brw_eu_debug.c | 10 ++-- src/mesa/pipe/i965simple/brw_eu_emit.c | 4 +- src/mesa/pipe/i965simple/brw_sf.c | 8 ++-- src/mesa/pipe/i965simple/brw_sf_emit.c | 14 +++--- src/mesa/pipe/i965simple/brw_state.c | 2 +- src/mesa/pipe/i965simple/brw_state_cache.c | 6 +-- src/mesa/pipe/i965simple/brw_state_pool.c | 4 +- src/mesa/pipe/i965simple/brw_urb.c | 6 +-- src/mesa/pipe/i965simple/brw_vs_emit.c | 2 +- src/mesa/pipe/i965simple/brw_wm.c | 2 +- src/mesa/pipe/i965simple/brw_wm_glsl.c | 2 +- src/mesa/pipe/i965simple/brw_wm_sampler_state.c | 2 +- src/mesa/pipe/p_compiler.h | 3 +- src/mesa/pipe/p_format.h | 3 ++ src/mesa/pipe/p_util.h | 4 -- src/mesa/pipe/pipebuffer/pb_buffer.h | 4 +- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 4 +- src/mesa/pipe/pipebuffer/pb_buffer_fenced.h | 2 +- src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 4 +- src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c | 4 +- src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 21 ++++----- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 8 ++-- src/mesa/pipe/softpipe/sp_prim_setup.c | 24 +++++----- src/mesa/pipe/softpipe/sp_quad_fs.c | 10 ++-- src/mesa/pipe/softpipe/sp_tile_cache.c | 4 +- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 6 +-- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 44 ++++++++--------- src/mesa/pipe/tgsi/util/tgsi_build.c | 1 + src/mesa/pipe/tgsi/util/tgsi_dump.c | 3 ++ src/mesa/pipe/tgsi/util/tgsi_parse.c | 1 + src/mesa/pipe/tgsi/util/tgsi_util.c | 1 + 47 files changed, 176 insertions(+), 182 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c index 58400213d7..51e2242719 100644 --- a/src/mesa/pipe/draw/draw_prim.c +++ b/src/mesa/pipe/draw/draw_prim.c @@ -30,6 +30,8 @@ * Keith Whitwell */ +#include "pipe/p_debug.h" + #include "draw_private.h" #include "draw_context.h" @@ -60,8 +62,8 @@ static void draw_prim_queue_flush( struct draw_context *draw ) unsigned i; if (0) - fprintf(stdout,"Flushing with %d prims, %d verts\n", - draw->pq.queue_nr, draw->vs.queue_nr); + debug_printf("Flushing with %d prims, %d verts\n", + draw->pq.queue_nr, draw->vs.queue_nr); assert (draw->pq.queue_nr != 0); @@ -120,9 +122,9 @@ static void draw_prim_queue_flush( struct draw_context *draw ) void draw_do_flush( struct draw_context *draw, unsigned flags ) { if (0) - fprintf(stdout,"Flushing with %d verts, %d prims\n", - draw->vs.queue_nr, - draw->pq.queue_nr ); + debug_printf("Flushing with %d verts, %d prims\n", + draw->vs.queue_nr, + draw->pq.queue_nr ); if (flags >= DRAW_FLUSH_SHADER_QUEUE) { @@ -157,11 +159,11 @@ static struct prim_header *get_queued_prim( struct draw_context *draw, unsigned nr_verts ) { if (!draw_vertex_cache_check_space( draw, nr_verts )) { -// fprintf(stderr, "v"); +// debug_printf("v"); draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE ); } else if (draw->pq.queue_nr == PRIM_QUEUE_LENGTH) { -// fprintf(stderr, "p"); +// debug_printf("p"); draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); } @@ -283,7 +285,7 @@ draw_prim( struct draw_context *draw, boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); -// _mesa_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); +// debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); switch (prim) { case PIPE_PRIM_POINTS: diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index ac03001d8f..be96c8fdeb 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -34,8 +34,7 @@ */ -#include - +#include "pipe/p_debug.h" #include "pipe/p_util.h" #include "draw_vbuf.h" @@ -125,55 +124,55 @@ dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) j = vinfo->src_index[i]; switch (vinfo->emit[i]) { case EMIT_OMIT: - fprintf(stderr, "EMIT_OMIT:"); + debug_printf("EMIT_OMIT:"); break; case EMIT_ALL: assert(i == 0); assert(j == 0); - fprintf(stderr, "EMIT_ALL:\t"); + debug_printf("EMIT_ALL:\t"); for(k = 0; k < vinfo->size*4; ++k) - fprintf(stderr, "%02x ", *data++); + debug_printf("%02x ", *data++); break; case EMIT_1F: - fprintf(stderr, "EMIT_1F:\t"); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + debug_printf("EMIT_1F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); break; case EMIT_1F_PSIZE: - fprintf(stderr, "EMIT_1F_PSIZE:\t"); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + debug_printf("EMIT_1F_PSIZE:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); break; case EMIT_2F: - fprintf(stderr, "EMIT_2F:\t"); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + debug_printf("EMIT_2F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); break; case EMIT_3F: - fprintf(stderr, "EMIT_3F:\t"); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + debug_printf("EMIT_3F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); data += sizeof(float); break; case EMIT_4F: - fprintf(stderr, "EMIT_4F:\t"); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); - fprintf(stderr, "%f ", *(float *)data); data += sizeof(float); + debug_printf("EMIT_4F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); break; case EMIT_4UB: - fprintf(stderr, "EMIT_4UB:\t"); - fprintf(stderr, "%u ", *data++); - fprintf(stderr, "%u ", *data++); - fprintf(stderr, "%u ", *data++); - fprintf(stderr, "%u ", *data++); + debug_printf("EMIT_4UB:\t"); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); break; default: assert(0); } - fprintf(stderr, "\n"); + debug_printf("\n"); } - fprintf(stderr, "\n"); + debug_printf("\n"); } #endif @@ -190,7 +189,7 @@ emit_vertex( struct vbuf_stage *vbuf, struct vertex_header *vertex ) { #if 0 - fprintf(stderr, "emit vertex %d to %p\n", + debug_printf("emit vertex %d to %p\n", vbuf->nr_vertices, vbuf->vertex_ptr); #endif @@ -198,7 +197,7 @@ emit_vertex( struct vbuf_stage *vbuf, if(vertex->vertex_id < vbuf->nr_vertices) return; else - fprintf(stderr, "Bad vertex id 0x%04x (>= 0x%04x)\n", + debug_printf("Bad vertex id 0x%04x (>= 0x%04x)\n", vertex->vertex_id, vbuf->nr_vertices); return; } @@ -269,9 +268,9 @@ emit_vertex( struct vbuf_stage *vbuf, static float data[256]; draw_vf_emit_vertex(vbuf->vf, vertex, data); if(memcmp((uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size, data, vbuf->vertex_size)) { - fprintf(stderr, "With VF:\n"); + debug_printf("With VF:\n"); dump_emitted_vertex(vbuf->vinfo, (uint8_t *)data); - fprintf(stderr, "Without VF:\n"); + debug_printf("Without VF:\n"); dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size); assert(0); } diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c index b23f487e74..e13df04605 100644 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ b/src/mesa/pipe/draw/draw_vertex_fetch.c @@ -162,7 +162,7 @@ static fetch_func get_fetch_func( enum pipe_format format ) { char tmp[80]; pf_sprint_name(tmp, format); - _mesa_printf("%s: %s\n", __FUNCTION__, tmp); + debug_printf("%s: %s\n", __FUNCTION__, tmp); } #endif @@ -332,7 +332,7 @@ static void fetch_xyz_rgb( struct draw_context *draw, assert(count <= 4); -// _mesa_printf("%s\n", __FUNCTION__); +// debug_printf("%s\n", __FUNCTION__); /* loop over vertex attributes (vertex shader inputs) */ @@ -421,7 +421,7 @@ static void generic_vertex_fetch( struct draw_context *draw, assert(count <= 4); -// _mesa_printf("%s %d\n", __FUNCTION__, count); +// debug_printf("%s %d\n", __FUNCTION__, count); /* loop over vertex attributes (vertex shader inputs) */ @@ -467,7 +467,7 @@ void draw_update_vertex_fetch( struct draw_context *draw ) { unsigned nr_attrs, i; -// _mesa_printf("%s\n", __FUNCTION__); +// debug_printf("%s\n", __FUNCTION__); /* this may happend during context init */ if (!draw->vertex_shader) diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index b851da845f..e6590eafcc 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -166,7 +166,7 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[0][3] = w; #if DBG_VS - printf("output[%d]win: %f %f %f %f\n", j, + debug_printf("output[%d]win: %f %f %f %f\n", j, vOut[j]->data[0][0], vOut[j]->data[0][1], vOut[j]->data[0][2], @@ -181,7 +181,7 @@ run_vertex_program(struct draw_context *draw, vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; #if DBG_VS - printf("output[%d][%d]: %f %f %f %f\n", j, slot, + debug_printf("output[%d][%d]: %f %f %f %f\n", j, slot, vOut[j]->data[slot][0], vOut[j]->data[slot][1], vOut[j]->data[slot][2], @@ -207,7 +207,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) */ draw_update_vertex_fetch( draw ); -// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); +// debug_printf( " q(%d) ", draw->vs.queue_nr ); #ifdef MESA_LLVM if (draw->vertex_shader->llvm_prog) { draw_vertex_shader_queue_flush_llvm(draw); diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c index 4228c4f388..63551c993e 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c @@ -152,7 +152,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) z = vOut->clip[2] = dests[0][2]; w = vOut->clip[3] = dests[0][3]; #if DBG - printf("output %d: %f %f %f %f\n", 0, x, y, z, w); + debug_printf("output %d: %f %f %f %f\n", 0, x, y, z, w); #endif vOut->clipmask = compute_clipmask(vOut->clip, draw->plane, draw->nr_planes); @@ -179,7 +179,7 @@ void draw_vertex_shader_queue_flush_llvm(struct draw_context *draw) vOut->data[slot][3] = dests[slot][3]; #if DBG - printf("output %d: %f %f %f %f\n", slot, + debug_printf("output %d: %f %f %f %f\n", slot, vOut->data[slot][0], vOut->data[slot][1], vOut->data[slot][2], diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index 0da8e59ad6..f23d7fcec5 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -168,7 +168,7 @@ draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, const unsigned format = map[i].format; if (format == DRAW_EMIT_PAD) { #if (DRAW_VF_DBG) - _mesa_printf("%d: pad %d, offset %d\n", i, + debug_printf("%d: pad %d, offset %d\n", i, map[i].offset, offset); #endif @@ -186,7 +186,7 @@ draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, memcpy(vf->attr[j].data, &map[i].data, vf->attr[j].vertattrsize); #if (DRAW_VF_DBG) - _mesa_printf("%d: %s, offset %d\n", i, + debug_printf("%d: %s, offset %d\n", i, draw_vf_format_info[format].name, vf->attr[j].vertoffset); #endif diff --git a/src/mesa/pipe/draw/draw_vf_generic.c b/src/mesa/pipe/draw/draw_vf_generic.c index 7f5f56ef9c..7a60a9db9c 100644 --- a/src/mesa/pipe/draw/draw_vf_generic.c +++ b/src/mesa/pipe/draw/draw_vf_generic.c @@ -27,9 +27,8 @@ */ -#include - #include "pipe/p_compiler.h" +#include "pipe/p_debug.h" #include "pipe/p_util.h" #include "draw_vf.h" diff --git a/src/mesa/pipe/draw/draw_vf_sse.c b/src/mesa/pipe/draw/draw_vf_sse.c index 1389e6cfb9..1ad2ae756d 100644 --- a/src/mesa/pipe/draw/draw_vf_sse.c +++ b/src/mesa/pipe/draw/draw_vf_sse.c @@ -453,7 +453,7 @@ static boolean build_vertex_emit( struct x86_program *p ) update_src_ptr(p, srcECX, vfESI, a); } else { - fprintf(stderr, "Can't emit 1ub %x %x %d\n", + debug_printf("Can't emit 1ub %x %x %d\n", a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); return FALSE; } @@ -499,7 +499,7 @@ static boolean build_vertex_emit( struct x86_program *p ) j++; /* NOTE: two attrs consumed */ } else { - fprintf(stderr, "Can't emit 3ub\n"); + debug_printf("Can't emit 3ub\n"); } return FALSE; /* add this later */ break; @@ -532,7 +532,7 @@ static boolean build_vertex_emit( struct x86_program *p ) update_src_ptr(p, srcECX, vfESI, a); break; default: - fprintf(stderr, "unknown a[%d].format %d\n", j, a->format); + debug_printf("unknown a[%d].format %d\n", j, a->format); return FALSE; /* catch any new opcodes */ } diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c index 0185512aeb..868f0c7e04 100644 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ b/src/mesa/pipe/i915simple/i915_fpc_translate.c @@ -100,7 +100,7 @@ negate(int reg, int x, int y, int z, int w) static void i915_use_passthrough_shader(struct i915_context *i915) { - fprintf(stderr, "**** Using i915 pass-through fragment shader\n"); + debug_printf("**** Using i915 pass-through fragment shader\n"); i915->current.program = (uint *) MALLOC(sizeof(passthrough)); if (i915->current.program) { @@ -119,12 +119,12 @@ i915_program_error(struct i915_fp_compile *p, const char *msg, ...) va_list args; char buffer[1024]; - fprintf(stderr, "i915_program_error: "); + debug_printf("i915_program_error: "); va_start( args, msg ); vsprintf( buffer, msg, args ); va_end( args ); - fprintf(stderr, buffer); - fprintf(stderr, "\n"); + debug_printf(buffer); + debug_printf("\n"); p->error = 1; } @@ -169,7 +169,7 @@ src_vector(struct i915_fp_compile *p, switch (sem_name) { case TGSI_SEMANTIC_POSITION: - fprintf(stderr, "SKIP SEM POS\n"); + debug_printf("SKIP SEM POS\n"); /* assert(p->wpos_tex != -1); src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); @@ -913,7 +913,7 @@ i915_translate_instructions(struct i915_fp_compile *p, ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - /*printf("FS Input DECL [%u] sem %u\n", ind, sem);*/ + /*debug_printf("FS Input DECL [%u] sem %u\n", ind, sem);*/ p->input_semantic_name[ind] = sem; p->input_semantic_index[ind] = semi; } @@ -924,7 +924,7 @@ i915_translate_instructions(struct i915_fp_compile *p, ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - /*printf("FS Output DECL [%u] sem %u\n", ind, sem);*/ + /*debug_printf("FS Output DECL [%u] sem %u\n", ind, sem);*/ p->output_semantic_name[ind] = sem; p->output_semantic_index[ind] = semi; } diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c index 39154b2488..e069773fd4 100644 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ b/src/mesa/pipe/i915simple/i915_prim_vbuf.c @@ -38,9 +38,8 @@ */ -#include - #include "pipe/draw/draw_vbuf.h" +#include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c index 62741e30f8..653983e4a9 100644 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ b/src/mesa/pipe/i915simple/i915_state_derived.c @@ -87,7 +87,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) } break; case TGSI_SEMANTIC_FOG: - fprintf(stderr, "i915 fogcoord not implemented yet\n"); + debug_printf("i915 fogcoord not implemented yet\n"); draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_PERSPECTIVE, src++); break; default: diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c index 657f523893..3339287f49 100644 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ b/src/mesa/pipe/i915simple/i915_state_emit.c @@ -107,7 +107,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) ) * 3/2; /* plus 50% margin */ #if 0 - fprintf (stderr, "i915_emit_hardware_state: %d dwords, %d relocs\n", dwords, relocs); + debug_printf("i915_emit_hardware_state: %d dwords, %d relocs\n", dwords, relocs); #endif if(!BEGIN_BATCH(dwords, relocs)) { diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c index 752d25f233..07031fc6c5 100644 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ b/src/mesa/pipe/i915simple/i915_state_immediate.c @@ -97,7 +97,7 @@ static void upload_S2S4(struct i915_context *i915) LIS2 = i915->current.vertex_info.hwfmt[1]; LIS4 = i915->current.vertex_info.hwfmt[0]; /* - printf("LIS2: 0x%x LIS4: 0x%x\n", LIS2, LIS4); + debug_printf("LIS2: 0x%x LIS4: 0x%x\n", LIS2, LIS4); */ assert(LIS4); /* should never be zero? */ } diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 59408b6ba0..0dbbc5241d 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -169,7 +169,7 @@ translate_texture_format(enum pipe_format pipeFormat) case PIPE_FORMAT_S8Z24_UNORM: return (MAPSURF_32BIT | MT_32BIT_xL824); default: - fprintf(stderr, "i915: translate_texture_format() bad image format %x\n", + debug_printf("i915: translate_texture_format() bad image format %x\n", pipeFormat); assert(0); return 0; diff --git a/src/mesa/pipe/i965simple/brw_cc.c b/src/mesa/pipe/i965simple/brw_cc.c index dcee731895..337e4f95f6 100644 --- a/src/mesa/pipe/i965simple/brw_cc.c +++ b/src/mesa/pipe/i965simple/brw_cc.c @@ -58,7 +58,7 @@ static int brw_translate_compare_func(int func) return BRW_COMPAREFUNCTION_ALWAYS; } - fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); + debug_printf("Unknown value in %s: %x\n", __FUNCTION__, func); return BRW_COMPAREFUNCTION_ALWAYS; } diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c index 2733eb4e75..52bbd525c1 100644 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ b/src/mesa/pipe/i965simple/brw_curbe.c @@ -273,10 +273,10 @@ static void upload_constant_buffer(struct brw_context *brw) if (1) { for (i = 0; i < sz; i+=4) - _mesa_printf("curbe %d.%d: %f %f %f %f\n", i/8, i&4, + debug_printf("curbe %d.%d: %f %f %f %f\n", i/8, i&4, buf[i+0], buf[i+1], buf[i+2], buf[i+3]); - _mesa_printf("last_buf %p buf %p sz %d/%d cmp %d\n", + debug_printf("last_buf %p buf %p sz %d/%d cmp %d\n", brw->curbe.last_buf, buf, bufsz, brw->curbe.last_bufsz, brw->curbe.last_buf ? memcmp(buf, brw->curbe.last_buf, bufsz) : -1); @@ -299,7 +299,7 @@ static void upload_constant_buffer(struct brw_context *brw) bufsz, 1 << 6, &brw->curbe.gs_offset)) { - _mesa_printf("out of GS memory for curbe\n"); + debug_printf("out of GS memory for curbe\n"); assert(0); return; } diff --git a/src/mesa/pipe/i965simple/brw_eu_debug.c b/src/mesa/pipe/i965simple/brw_eu_debug.c index be692f6502..4a94ddefa6 100644 --- a/src/mesa/pipe/i965simple/brw_eu_debug.c +++ b/src/mesa/pipe/i965simple/brw_eu_debug.c @@ -30,6 +30,8 @@ */ +#include "pipe/p_debug.h" + #include "brw_eu.h" void brw_print_reg( struct brw_reg hwreg ) @@ -52,7 +54,7 @@ void brw_print_reg( struct brw_reg hwreg ) "f" }; - _mesa_printf("%s%s", + debug_printf("%s%s", hwreg.abs ? "abs/" : "", hwreg.negate ? "-" : ""); @@ -63,17 +65,17 @@ void brw_print_reg( struct brw_reg hwreg ) hwreg.width == BRW_WIDTH_8 && hwreg.hstride == BRW_HORIZONTAL_STRIDE_1 && hwreg.type == BRW_REGISTER_TYPE_F) { - _mesa_printf("vec%d", hwreg.nr); + debug_printf("vec%d", hwreg.nr); } else if (hwreg.file == BRW_GENERAL_REGISTER_FILE && hwreg.vstride == BRW_VERTICAL_STRIDE_0 && hwreg.width == BRW_WIDTH_1 && hwreg.hstride == BRW_HORIZONTAL_STRIDE_0 && hwreg.type == BRW_REGISTER_TYPE_F) { - _mesa_printf("scl%d.%d", hwreg.nr, hwreg.subnr / 4); + debug_printf("scl%d.%d", hwreg.nr, hwreg.subnr / 4); } else { - _mesa_printf("%s%d.%d<%d;%d,%d>:%s", + debug_printf("%s%d.%d<%d;%d,%d>:%s", file[hwreg.file], hwreg.nr, hwreg.subnr / type_sz(hwreg.type), diff --git a/src/mesa/pipe/i965simple/brw_eu_emit.c b/src/mesa/pipe/i965simple/brw_eu_emit.c index 2423536dd1..400a80b6fb 100644 --- a/src/mesa/pipe/i965simple/brw_eu_emit.c +++ b/src/mesa/pipe/i965simple/brw_eu_emit.c @@ -953,7 +953,7 @@ void brw_SAMPLE(struct brw_compile *p, boolean need_stall = 0; if(writemask == 0) { -/* _mesa_printf("%s: zero writemask??\n", __FUNCTION__); */ +/* debug_printf("%s: zero writemask??\n", __FUNCTION__); */ return; } @@ -985,7 +985,7 @@ void brw_SAMPLE(struct brw_compile *p, if (newmask != writemask) { need_stall = 1; -/* _mesa_printf("need stall %x %x\n", newmask , writemask); */ +/* debug_printf("need stall %x %x\n", newmask , writemask); */ } else { struct brw_reg m1 = brw_message_reg(msg_reg_nr); diff --git a/src/mesa/pipe/i965simple/brw_sf.c b/src/mesa/pipe/i965simple/brw_sf.c index b89b2e4087..7c83b81c85 100644 --- a/src/mesa/pipe/i965simple/brw_sf.c +++ b/src/mesa/pipe/i965simple/brw_sf.c @@ -175,7 +175,7 @@ static void upload_sf_prog( struct brw_context *brw ) //int semantic = parse.FullToken.FullDeclaration.Semantic.SemanticName; //int semantic_index = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - fprintf(stderr, "fs input %d..%d interp mode %d\n", first, last, interp_mode); + debug_printf("fs input %d..%d interp mode %d\n", first, last, interp_mode); switch (interp_mode) { case TGSI_INTERPOLATE_CONSTANT: @@ -213,9 +213,9 @@ static void upload_sf_prog( struct brw_context *brw ) key.linear_mask |= 1; key.const_mask <<= 1; - fprintf(stderr, "key.persp_mask: %x\n", key.persp_mask); - fprintf(stderr, "key.linear_mask: %x\n", key.linear_mask); - fprintf(stderr, "key.const_mask: %x\n", key.const_mask); + debug_printf("key.persp_mask: %x\n", key.persp_mask); + debug_printf("key.linear_mask: %x\n", key.linear_mask); + debug_printf("key.const_mask: %x\n", key.const_mask); // key.do_point_sprite = brw->attribs.Point->PointSprite; diff --git a/src/mesa/pipe/i965simple/brw_sf_emit.c b/src/mesa/pipe/i965simple/brw_sf_emit.c index 6ff5254ff7..78d6fa5e9e 100644 --- a/src/mesa/pipe/i965simple/brw_sf_emit.c +++ b/src/mesa/pipe/i965simple/brw_sf_emit.c @@ -137,8 +137,8 @@ static boolean calculate_masks( struct brw_sf_compile *c, unsigned persp_mask = c->key.persp_mask; unsigned linear_mask = c->key.linear_mask; - fprintf(stderr, "persp_mask: %x\n", persp_mask); - fprintf(stderr, "linear_mask: %x\n", linear_mask); + debug_printf("persp_mask: %x\n", persp_mask); + debug_printf("linear_mask: %x\n", linear_mask); *pc_persp = 0; *pc_linear = 0; @@ -162,9 +162,9 @@ static boolean calculate_masks( struct brw_sf_compile *c, *pc_linear |= 0xf0; } - fprintf(stderr, "pc: %x\n", *pc); - fprintf(stderr, "pc_persp: %x\n", *pc_persp); - fprintf(stderr, "pc_linear: %x\n", *pc_linear); + debug_printf("pc: %x\n", *pc); + debug_printf("pc_persp: %x\n", *pc_persp); + debug_printf("pc_linear: %x\n", *pc_linear); return is_last_attr; @@ -177,7 +177,7 @@ void brw_emit_tri_setup( struct brw_sf_compile *c ) struct brw_compile *p = &c->func; unsigned i; - fprintf(stderr, "%s START ==============\n", __FUNCTION__); + debug_printf("%s START ==============\n", __FUNCTION__); c->nr_verts = 3; alloc_regs(c); @@ -250,7 +250,7 @@ void brw_emit_tri_setup( struct brw_sf_compile *c ) } } - fprintf(stderr, "%s DONE ==============\n", __FUNCTION__); + debug_printf("%s DONE ==============\n", __FUNCTION__); } diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c index daf14ff4ff..95dfce88e4 100644 --- a/src/mesa/pipe/i965simple/brw_state.c +++ b/src/mesa/pipe/i965simple/brw_state.c @@ -225,7 +225,7 @@ static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; brw->state.dirty.brw |= BRW_NEW_VS; - printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); + debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); } static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) diff --git a/src/mesa/pipe/i965simple/brw_state_cache.c b/src/mesa/pipe/i965simple/brw_state_cache.c index c5738733f4..b3a5124461 100644 --- a/src/mesa/pipe/i965simple/brw_state_cache.c +++ b/src/mesa/pipe/i965simple/brw_state_cache.c @@ -149,7 +149,7 @@ unsigned brw_upload_cache( struct brw_cache *cache, if (!brw_pool_alloc(cache->pool, data_size, 1 << 6, &offset)) { /* Should not be possible: */ - printf("brw_pool_alloc failed\n"); + debug_printf("brw_pool_alloc failed\n"); exit(1); } @@ -177,7 +177,7 @@ unsigned brw_upload_cache( struct brw_cache *cache, } if (BRW_DEBUG & DEBUG_STATE) - printf("upload %s: %d bytes to pool buffer %p offset %x\n", + debug_printf("upload %s: %d bytes to pool buffer %p offset %x\n", cache->name, data_size, (void*)cache->pool->buffer, @@ -416,7 +416,7 @@ void brw_clear_all_caches( struct brw_context *brw ) int i; if (BRW_DEBUG & DEBUG_STATE) - fprintf(stderr, "%s\n", __FUNCTION__); + debug_printf("%s\n", __FUNCTION__); for (i = 0; i < BRW_MAX_CACHE; i++) clear_cache(&brw->cache[i]); diff --git a/src/mesa/pipe/i965simple/brw_state_pool.c b/src/mesa/pipe/i965simple/brw_state_pool.c index 7c67f0ee25..f3174bfe0a 100644 --- a/src/mesa/pipe/i965simple/brw_state_pool.c +++ b/src/mesa/pipe/i965simple/brw_state_pool.c @@ -58,7 +58,7 @@ boolean brw_pool_alloc( struct brw_mem_pool *pool, size = align(size, 4); if (pool->offset + fixup + size >= pool->size) { - printf("%s failed\n", __FUNCTION__); + debug_printf("%s failed\n", __FUNCTION__); assert(0); exit(0); } @@ -74,7 +74,7 @@ static void brw_invalidate_pool( struct brw_mem_pool *pool ) { if (BRW_DEBUG & DEBUG_STATE) - printf("\n\n\n %s \n\n\n", __FUNCTION__); + debug_printf("\n\n\n %s \n\n\n", __FUNCTION__); pool->offset = 0; diff --git a/src/mesa/pipe/i965simple/brw_urb.c b/src/mesa/pipe/i965simple/brw_urb.c index b284526aa6..101a4367b9 100644 --- a/src/mesa/pipe/i965simple/brw_urb.c +++ b/src/mesa/pipe/i965simple/brw_urb.c @@ -120,18 +120,18 @@ static void recalculate_urb_fence( struct brw_context *brw ) * entries and the values for minimum nr of entries * provided above. */ - fprintf(stderr, "couldn't calculate URB layout!\n"); + debug_printf("couldn't calculate URB layout!\n"); exit(1); } if (BRW_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS)) - printf("URB CONSTRAINED\n"); + debug_printf("URB CONSTRAINED\n"); } else brw->urb.constrained = 0; if (BRW_DEBUG & DEBUG_URB) - printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n", + debug_printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n", brw->urb.vs_start, brw->urb.gs_start, brw->urb.clip_start, diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c index b32c233dd2..98915ba101 100644 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ b/src/mesa/pipe/i965simple/brw_vs_emit.c @@ -1228,7 +1228,7 @@ static void process_instruction(struct brw_vs_compile *c, case TGSI_OPCODE_ENDSUB: break; default: - printf("Unsupport opcode %d in vertex shader\n", inst->Instruction.Opcode); + debug_printf("Unsupport opcode %d in vertex shader\n", inst->Instruction.Opcode); break; } diff --git a/src/mesa/pipe/i965simple/brw_wm.c b/src/mesa/pipe/i965simple/brw_wm.c index 0ee0fbed51..539b170744 100644 --- a/src/mesa/pipe/i965simple/brw_wm.c +++ b/src/mesa/pipe/i965simple/brw_wm.c @@ -57,7 +57,7 @@ static void do_wm_prog( struct brw_context *brw, c->pixel_w = brw_null_reg(); - fprintf(stderr, "XXXXXXXX FP\n"); + debug_printf("XXXXXXXX FP\n"); brw_wm_glsl_emit(c); diff --git a/src/mesa/pipe/i965simple/brw_wm_glsl.c b/src/mesa/pipe/i965simple/brw_wm_glsl.c index f4b5c13c06..d95645d108 100644 --- a/src/mesa/pipe/i965simple/brw_wm_glsl.c +++ b/src/mesa/pipe/i965simple/brw_wm_glsl.c @@ -982,7 +982,7 @@ static void brw_wm_emit_instruction( struct brw_wm_compile *c, break; default: - _mesa_printf("unsupported IR in fragment shader %d\n", + debug_printf("unsupported IR in fragment shader %d\n", inst->Instruction.Opcode); } #if 0 diff --git a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c b/src/mesa/pipe/i965simple/brw_wm_sampler_state.c index cfb430eb09..de42ffc5b1 100644 --- a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_sampler_state.c @@ -71,7 +71,7 @@ static int intel_translate_shadow_compare_func(unsigned func) return COMPAREFUNC_NEVER; } - fprintf(stderr, "Unknown value in %s: %x\n", __FUNCTION__, func); + debug_printf("Unknown value in %s: %x\n", __FUNCTION__, func); return COMPAREFUNC_NEVER; } diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h index e939d9cd9b..30cd729c56 100644 --- a/src/mesa/pipe/p_compiler.h +++ b/src/mesa/pipe/p_compiler.h @@ -28,10 +28,9 @@ #ifndef P_COMPILER_H #define P_COMPILER_H -#include + #include #include -#include #if defined(_WIN32) && !defined(__WIN32__) diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h index 9f60cdbb04..c9ad324315 100644 --- a/src/mesa/pipe/p_format.h +++ b/src/mesa/pipe/p_format.h @@ -28,7 +28,10 @@ #ifndef PIPE_FORMAT_H #define PIPE_FORMAT_H +#include // for sprintf + #include "p_compiler.h" +#include "p_debug.h" /** * The PIPE_FORMAT is a 32-bit wide bitfield that encodes all the information diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 059528787d..4780ed7818 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -381,10 +381,6 @@ static INLINE int align(int value, int alignment) return (value + alignment - 1) & ~(alignment - 1); } -/* Convenient... - */ -extern void _mesa_printf(const char *str, ...); - /* util/p_util.c */ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.h b/src/mesa/pipe/pipebuffer/pb_buffer.h index 17551b3b50..97beb5f72a 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer.h @@ -44,10 +44,8 @@ #define PB_BUFFER_H_ -#include -#include - #include "pipe/p_compiler.h" +#include "pipe/p_debug.h" #include "pipe/p_state.h" #include "pipe/p_inlines.h" diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c index 4cf4222db9..f4fc3f6d71 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c @@ -34,12 +34,10 @@ */ -#include -#include - #include "linked_list.h" #include "p_compiler.h" +#include "p_debug.h" #include "p_winsys.h" #include "p_thread.h" #include "p_util.h" diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h index 09082a5390..c40b9c75e1 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h +++ b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h @@ -51,7 +51,7 @@ #define PB_BUFFER_FENCED_H_ -#include +#include "pipe/p_debug.h" struct pipe_winsys; diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c index 2151f1d691..c1b7759874 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c @@ -34,9 +34,7 @@ */ -#include -#include - +#include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pb_buffer.h" diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c index 3b341c64c2..c535d3276c 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c @@ -34,9 +34,7 @@ */ -#include -#include - +#include "p_debug.h" #include "p_util.h" #include "pb_buffer.h" diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c index b6af7cdedc..8b1b51c0e2 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c @@ -34,11 +34,10 @@ */ -#include - #include "linked_list.h" #include "p_defines.h" +#include "p_debug.h" #include "p_thread.h" #include "p_util.h" #include "pb_buffer.h" @@ -69,28 +68,28 @@ struct mem_block static void mmDumpMemInfo(const struct mem_block *heap) { - fprintf(stderr, "Memory heap %p:\n", (void *)heap); + debug_printf("Memory heap %p:\n", (void *)heap); if (heap == 0) { - fprintf(stderr, " heap == 0\n"); + debug_printf(" heap == 0\n"); } else { const struct mem_block *p; for(p = heap->next; p != heap; p = p->next) { - fprintf(stderr, " Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, p->free ? 'F':'.', p->reserved ? 'R':'.'); } - fprintf(stderr, "\nFree list:\n"); + debug_printf("\nFree list:\n"); for(p = heap->next_free; p != heap; p = p->next_free) { - fprintf(stderr, " FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, p->free ? 'F':'.', p->reserved ? 'R':'.'); } } - fprintf(stderr, "End of memory blocks\n"); + debug_printf("End of memory blocks\n"); } #endif @@ -308,11 +307,11 @@ mmFreeMem(struct mem_block *b) return 0; if (b->free) { - fprintf(stderr, "block already free\n"); + debug_printf("block already free\n"); return -1; } if (b->reserved) { - fprintf(stderr, "block is reserved\n"); + debug_printf("block is reserved\n"); return -1; } @@ -479,7 +478,7 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); if(!mm_buf->block) { - fprintf(stderr, "warning: heap full\n"); + debug_printf("warning: heap full\n"); #if 0 mmDumpMemInfo(mm->heap); #endif diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index f80c7e34c0..bcd4b3e257 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -35,12 +35,10 @@ */ -#include -#include - #include "linked_list.h" #include "p_compiler.h" +#include "p_debug.h" #include "p_thread.h" #include "p_defines.h" #include "p_util.h" @@ -178,7 +176,7 @@ pool_bufmgr_create_buffer(struct pb_manager *mgr, if (pool->numFree == 0) { _glthread_UNLOCK_MUTEX(pool->mutex); - fprintf(stderr, "warning: out of fixed size buffer objects\n"); + debug_printf("warning: out of fixed size buffer objects\n"); return NULL; } @@ -186,7 +184,7 @@ pool_bufmgr_create_buffer(struct pb_manager *mgr, if (item == &pool->free) { _glthread_UNLOCK_MUTEX(pool->mutex); - fprintf(stderr, "error: fixed size buffer pool corruption\n"); + debug_printf("error: fixed size buffer pool corruption\n"); return NULL; } diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index b17801d13d..7478b2336b 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -251,9 +251,9 @@ static void print_vertex(const struct setup_stage *setup, const struct vertex_header *v) { int i; - fprintf(stderr, "Vertex: (%p)\n", v); + debug_printf("Vertex: (%p)\n", v); for (i = 0; i < setup->quad.nr_attrs; i++) { - fprintf(stderr, " %d: %f %f %f %f\n", i, + debug_printf(" %d: %f %f %f %f\n", i, v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); } } @@ -267,7 +267,7 @@ static boolean setup_sort_vertices( struct setup_stage *setup, const struct vertex_header *v2 = prim->v[2]; #if DEBUG_VERTS - fprintf(stderr, "Triangle:\n"); + debug_printf("Triangle:\n"); print_vertex(setup, v0); print_vertex(setup, v1); print_vertex(setup, v2); @@ -345,7 +345,7 @@ static boolean setup_sort_vertices( struct setup_stage *setup, setup->oneoverarea = 1.0f / area; /* - _mesa_printf("%s one-over-area %f area %f det %f\n", + debug_printf("%s one-over-area %f area %f det %f\n", __FUNCTION__, setup->oneoverarea, area, prim->det ); */ } @@ -419,7 +419,7 @@ static void tri_linear_coeff( struct setup_stage *setup, dady * (setup->vmin->data[0][1] - 0.5f))); /* - _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", + debug_printf("attr[%d].%c: %f dx:%f dy:%f\n", slot, "xyzw"[i], setup->coef[slot].a0[i], setup->coef[slot].dadx[i], @@ -453,10 +453,10 @@ static void tri_persp_coeff( struct setup_stage *setup, float dady = b * setup->oneoverarea; /* - printf("tri persp %d,%d: %f %f %f\n", vertSlot, i, - setup->vmin->data[vertSlot][i], - setup->vmid->data[vertSlot][i], - setup->vmax->data[vertSlot][i] + debug_printf("tri persp %d,%d: %f %f %f\n", vertSlot, i, + setup->vmin->data[vertSlot][i], + setup->vmid->data[vertSlot][i], + setup->vmax->data[vertSlot][i] ); */ assert(i <= 3); @@ -619,7 +619,7 @@ static void subtriangle( struct setup_stage *setup, finish_y -= sy; /* - _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); */ for (y = start_y; y < finish_y; y++) { @@ -671,7 +671,7 @@ static void setup_tri( struct draw_stage *stage, struct setup_stage *setup = setup_stage( stage ); /* - _mesa_printf("%s\n", __FUNCTION__ ); + debug_printf("%s\n", __FUNCTION__ ); */ setup_sort_vertices( setup, prim ); @@ -1124,7 +1124,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) int ix, iy; /* - printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax); + debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax); */ for (iy = iymin; iy <= iymax; iy += 2) { uint rowMask = 0xf; diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 90691c6065..b5d7dfca1c 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -223,13 +223,13 @@ shade_quad_llvm(struct quad_stage *qs, inputs[2][0][1] = fy + 1.0f; inputs[3][0][1] = fy + 1.0f; #if DLLVM - printf("MASK = %d\n", quad->mask); + debug_printf("MASK = %d\n", quad->mask); #endif gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef); #if DLLVM for (int i = 0; i < 4; ++i) { for (int j = 0; j < 2; ++j) { - printf("IN(%d,%d) [%f %f %f %f]\n", i, j, + debug_printf("IN(%d,%d) [%f %f %f %f]\n", i, j, inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]); } } @@ -240,7 +240,7 @@ shade_quad_llvm(struct quad_stage *qs, softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], qss->samplers); #if DLLVM - printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n", + debug_printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n", dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]); #endif @@ -260,7 +260,7 @@ shade_quad_llvm(struct quad_stage *qs, } #if DLLVM for (int i = 0; i < QUAD_SIZE; ++i) { - printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, + debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, quad->outputs.color[0][i], quad->outputs.color[1][i], quad->outputs.color[2][i], @@ -284,7 +284,7 @@ shade_quad_llvm(struct quad_stage *qs, } } #if DLLVM - printf("D [%f, %f, %f, %f] mask = %d\n", + debug_printf("D [%f, %f, %f, %f] mask = %d\n", quad->outputs.depth[0], quad->outputs.depth[1], quad->outputs.depth[2], diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c index ccf367a5e4..1597361b82 100644 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ b/src/mesa/pipe/softpipe/sp_tile_cache.c @@ -341,7 +341,7 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, } } #if 0 - printf("num cleared: %u\n", numCleared); + debug_printf("num cleared: %u\n", numCleared); #endif } @@ -384,7 +384,7 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, #endif #if 0 - printf("flushed tiles in use: %d\n", inuse); + debug_printf("flushed tiles in use: %d\n", inuse); #endif } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index dcc39362a9..463ff0d9da 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -143,7 +143,7 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) k = tgsi_parse_init( &parse, mach->Tokens ); if (k != TGSI_PARSE_OK) { - fprintf(stderr, "Problem parsing!\n"); + debug_printf("Problem parsing!\n"); return; } @@ -249,7 +249,7 @@ tgsi_exec_machine_init( k = tgsi_parse_init (&parse, mach->Tokens); if (k != TGSI_PARSE_OK) { - fprintf( stderr, "Problem parsing!\n" ); + debug_printf( "Problem parsing!\n" ); return; } @@ -1236,7 +1236,7 @@ exec_tex(struct tgsi_exec_machine *mach, uint chan_index; float lodBias; - /* printf("Sampler %u unit %u\n", sampler, unit); */ + /* debug_printf("Sampler %u unit %u\n", sampler, unit); */ switch (inst->InstructionExtTexture.Texture) { case TGSI_TEXTURE_1D: diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index df0c698301..f2180082f1 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -48,28 +48,28 @@ _print_reg( case file_REG32: switch( reg.idx ) { case reg_AX: - printf( "EAX" ); + debug_printf( "EAX" ); break; case reg_CX: - printf( "ECX" ); + debug_printf( "ECX" ); break; case reg_DX: - printf( "EDX" ); + debug_printf( "EDX" ); break; case reg_BX: - printf( "EBX" ); + debug_printf( "EBX" ); break; case reg_SP: - printf( "ESP" ); + debug_printf( "ESP" ); break; case reg_BP: - printf( "EBP" ); + debug_printf( "EBP" ); break; case reg_SI: - printf( "ESI" ); + debug_printf( "ESI" ); break; case reg_DI: - printf( "EDI" ); + debug_printf( "EDI" ); break; } break; @@ -77,7 +77,7 @@ _print_reg( assert( 0 ); break; case file_XMM: - printf( "XMM%u", reg.idx ); + debug_printf( "XMM%u", reg.idx ); break; case file_x87: assert( 0 ); @@ -92,35 +92,35 @@ _fill( unsigned count = 10 - strlen( op ); while( count-- ) { - printf( " " ); + debug_printf( " " ); } } -#define DUMP_START() printf( "\nsse-dump start ----------------" ) -#define DUMP_END() printf( "\nsse-dump end ----------------\n" ) -#define DUMP( OP ) printf( "\n%s", OP ) +#define DUMP_START() debug_printf( "\nsse-dump start ----------------" ) +#define DUMP_END() debug_printf( "\nsse-dump end ----------------\n" ) +#define DUMP( OP ) debug_printf( "\n%s", OP ) #define DUMP_I( OP, I ) do {\ - printf( "\n%s", OP );\ + debug_printf( "\n%s", OP );\ _fill( OP );\ - printf( "%u", I ); } while( 0 ) + debug_printf( "%u", I ); } while( 0 ) #define DUMP_R( OP, R0 ) do {\ - printf( "\n%s", OP );\ + debug_printf( "\n%s", OP );\ _fill( OP );\ _print_reg( R0 ); } while( 0 ) #define DUMP_RR( OP, R0, R1 ) do {\ - printf( "\n%s", OP );\ + debug_printf( "\n%s", OP );\ _fill( OP );\ _print_reg( R0 );\ - printf( ", " );\ + debug_printf( ", " );\ _print_reg( R1 ); } while( 0 ) #define DUMP_RRI( OP, R0, R1, I ) do {\ - printf( "\n%s", OP );\ + debug_printf( "\n%s", OP );\ _fill( OP );\ _print_reg( R0 );\ - printf( ", " );\ + debug_printf( ", " );\ _print_reg( R1 );\ - printf( ", " );\ - printf( "%u", I ); } while( 0 ) + debug_printf( ", " );\ + debug_printf( "%u", I ); } while( 0 ) #else diff --git a/src/mesa/pipe/tgsi/util/tgsi_build.c b/src/mesa/pipe/tgsi/util/tgsi_build.c index 67f7d2c2c2..a00ff1c2a5 100644 --- a/src/mesa/pipe/tgsi/util/tgsi_build.c +++ b/src/mesa/pipe/tgsi/util/tgsi_build.c @@ -1,3 +1,4 @@ +#include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi_build.h" diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.c b/src/mesa/pipe/tgsi/util/tgsi_dump.c index cdbc0dbc9c..b5c54847e0 100644 --- a/src/mesa/pipe/tgsi/util/tgsi_dump.c +++ b/src/mesa/pipe/tgsi/util/tgsi_dump.c @@ -25,6 +25,9 @@ * **************************************************************************/ +#include + +#include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi_dump.h" diff --git a/src/mesa/pipe/tgsi/util/tgsi_parse.c b/src/mesa/pipe/tgsi/util/tgsi_parse.c index f0f8d44ac2..bf6b89ce56 100644 --- a/src/mesa/pipe/tgsi/util/tgsi_parse.c +++ b/src/mesa/pipe/tgsi/util/tgsi_parse.c @@ -25,6 +25,7 @@ * **************************************************************************/ +#include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi_parse.h" diff --git a/src/mesa/pipe/tgsi/util/tgsi_util.c b/src/mesa/pipe/tgsi/util/tgsi_util.c index 1e76b0f133..4cdd89182a 100644 --- a/src/mesa/pipe/tgsi/util/tgsi_util.c +++ b/src/mesa/pipe/tgsi/util/tgsi_util.c @@ -1,3 +1,4 @@ +#include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi_parse.h" -- cgit v1.2.3 From af6b12cc76b40c86f3b144a7f5cd3ef1278863d0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Feb 2008 01:07:49 +0900 Subject: gallium: Bring latest fixes. --- src/mesa/pipe/p_debug.h | 9 ++++++++- src/mesa/pipe/util/p_debug.c | 16 +++++++++++----- 2 files changed, 19 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_debug.h b/src/mesa/pipe/p_debug.h index b037eba2a3..2a11627b36 100644 --- a/src/mesa/pipe/p_debug.h +++ b/src/mesa/pipe/p_debug.h @@ -38,6 +38,10 @@ #ifndef P_DEBUG_H_ #define P_DEBUG_H_ + +#include + + #ifdef __cplusplus extern "C" { #endif @@ -55,8 +59,12 @@ extern "C" { void debug_printf(const char *format, ...); + +void debug_vprintf(const char *format, va_list ap); + void debug_assert_fail(const char *expr, const char *file, unsigned line); + /** Assert macro */ #ifdef DEBUG #define debug_assert(expr) ((expr) ? (void)0 : debug_assert_fail(#expr, __FILE__, __LINE__)) @@ -66,7 +74,6 @@ void debug_assert_fail(const char *expr, const char *file, unsigned line); #ifdef assert -#warning Standard C Library assert macro usage detected. #undef assert #endif #define assert(expr) debug_assert(expr) diff --git a/src/mesa/pipe/util/p_debug.c b/src/mesa/pipe/util/p_debug.c index 9303c970cc..b9607a6ba7 100644 --- a/src/mesa/pipe/util/p_debug.c +++ b/src/mesa/pipe/util/p_debug.c @@ -40,16 +40,22 @@ #include "pipe/p_compiler.h" -void debug_printf(const char *format, ...) +void debug_vprintf(const char *format, va_list ap) { - va_list ap; - va_start( ap, format ); #ifdef WIN32 EngDebugPrint("Gallium3D: ", (PCHAR)format, ap); #else vfprintf(stderr, format, ap); #endif - va_end( ap ); +} + + +void debug_printf(const char *format, ...) +{ + va_list ap; + va_start(ap, format); + debug_vprintf(format, ap); + va_end(ap); } @@ -65,6 +71,6 @@ static INLINE void debug_abort(void) void debug_assert_fail(const char *expr, const char *file, unsigned line) { - debug_printf("%s:%i: Assertion `%s' failed."); + debug_printf("%s:%i: Assertion `%s' failed.\n", file, line, expr); debug_abort(); } -- cgit v1.2.3 From 4650b35846e8e87fb0d74573a5f66452bb449b4b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 16:32:15 -0700 Subject: gallium: add bitmap/drawpixels texcoord bias support The state tracker will call pipe->get_paramf(PIPE_CAP_BITMAP_TEXCOORD_BIAS) to get a bias factor for adjusting the texcoords used in bitmap/drawpixels. This allows us to compensate for small differences in rasterization from one device to another. --- src/mesa/pipe/p_defines.h | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 7 +++++-- src/mesa/state_tracker/st_context.h | 2 ++ src/mesa/state_tracker/st_extensions.c | 3 +++ 4 files changed, 11 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h index 85adf2d61d..0bf53ecb79 100644 --- a/src/mesa/pipe/p_defines.h +++ b/src/mesa/pipe/p_defines.h @@ -265,6 +265,6 @@ enum pipe_texture_target { #define PIPE_CAP_MAX_POINT_WIDTH_AA 17 #define PIPE_CAP_MAX_TEXTURE_ANISOTROPY 18 #define PIPE_CAP_MAX_TEXTURE_LOD_BIAS 19 - +#define PIPE_CAP_BITMAP_TEXCOORD_BIAS 20 #endif diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 6b44cba2e4..34d420fcff 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -581,10 +581,13 @@ draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, GLfloat x1, GLfloat y1, const GLfloat *color, GLboolean invertTex) { + GLfloat bias = ctx->st->bitmap_texcoord_bias; GLfloat verts[4][3][4]; /* four verts, three attribs, XYZW */ GLuint i; - GLfloat sLeft = 0.0, sRight = 1.0; - GLfloat tTop = invertTex, tBot = 1.0 - tTop; + GLfloat xBias = bias / (x1-x0); + GLfloat yBias = bias / (y1-y0); + GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias; + GLfloat tTop = invertTex - yBias, tBot = 1.0 - tTop - yBias; /* upper-left */ verts[0][0][0] = x0; /* attr[0].x */ diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 2b6f8743f3..a756055898 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -143,6 +143,8 @@ struct st_context GLfloat polygon_offset_scale; /* ?? */ + GLfloat bitmap_texcoord_bias; + /** Mapping from VERT_RESULT_x to post-transformed vertex slot */ const GLuint *vertex_result_to_slot; diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 0157bdd6b3..97d28d77c4 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -106,6 +106,9 @@ void st_init_limits(struct st_context *st) c->MaxTextureLodBias = pipe->get_paramf(pipe, PIPE_CAP_MAX_TEXTURE_LOD_BIAS); + + st->bitmap_texcoord_bias + = pipe->get_paramf(pipe, PIPE_CAP_BITMAP_TEXCOORD_BIAS); } -- cgit v1.2.3 From 71984d76aae937274f6dd08c24f995d3c0c06357 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 17:46:31 -0700 Subject: clean-up --- src/mesa/state_tracker/st_atom_texture.c | 19 +++++++------------ 1 file changed, 7 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index fb21d29c40..b3859f18cb 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -36,7 +36,6 @@ #include "st_atom.h" #include "st_cb_texture.h" #include "pipe/p_context.h" -#include "pipe/p_defines.h" /** @@ -46,18 +45,14 @@ static void update_textures(struct st_context *st) { - GLuint s; - /* ST_NEW_FRAGMENT_PROGRAM */ struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current; + GLuint unit; - for (s = 0; s < st->ctx->Const.MaxTextureCoordUnits; s++) { - GLuint su = fprog->Base.SamplerUnits[s]; - - struct gl_texture_object *texObj - = st->ctx->Texture.Unit[su]._Current; - + for (unit = 0; unit < st->ctx->Const.MaxTextureCoordUnits; unit++) { + const GLuint su = fprog->Base.SamplerUnits[unit]; + struct gl_texture_object *texObj = st->ctx->Texture.Unit[su]._Current; struct pipe_texture *pt; if (texObj) { @@ -75,9 +70,9 @@ update_textures(struct st_context *st) * this table before being deleted, otherwise the pointer * comparison below could fail. */ - if (st->state.sampler_texture[s] != pt) { - st->state.sampler_texture[s] = pt; - st->pipe->set_sampler_texture(st->pipe, s, pt); + if (st->state.sampler_texture[unit] != pt) { + st->state.sampler_texture[unit] = pt; + st->pipe->set_sampler_texture(st->pipe, unit, pt); } } } -- cgit v1.2.3 From a4fbf096734efca2100aff41e988cce26ced5f6f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 17:55:16 -0700 Subject: comments, clean-ups, consts --- src/mesa/state_tracker/st_texture.c | 30 ++++++++++++++++++------------ src/mesa/state_tracker/st_texture.h | 8 ++++---- 2 files changed, 22 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 15cc458be8..741f36c2a7 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -59,6 +59,10 @@ target_to_target(GLenum target) } #endif + +/** + * Allocate a new pipe_texture object + */ struct pipe_texture * st_texture_create(struct st_context *st, enum pipe_texture_target target, @@ -100,17 +104,16 @@ st_texture_create(struct st_context *st, } - - -/* Can the image be pulled into a unified mipmap texture. This mirrors - * the completeness test in a lot of ways. +/** + * Check if a texture image be pulled into a unified mipmap texture. + * This mirrors the completeness test in a lot of ways. * * Not sure whether I want to pass gl_texture_image here. */ GLboolean -st_texture_match_image(struct pipe_texture *pt, - struct gl_texture_image *image, - GLuint face, GLuint level) +st_texture_match_image(const struct pipe_texture *pt, + const struct gl_texture_image *image, + GLuint face, GLuint level) { /* Images with borders are never pulled into mipmap textures. */ @@ -189,6 +192,7 @@ st_texture_image_map(struct st_context *st, struct st_texture_image *stImage, return pipe_surface_map(stImage->surface); } + void st_texture_image_unmap(struct st_texture_image *stImage) { @@ -201,7 +205,8 @@ st_texture_image_unmap(struct st_texture_image *stImage) -/* Upload data to a rectangular sub-region. Lots of choices how to do this: +/** + * Upload data to a rectangular sub-region. Lots of choices how to do this: * * - memcpy by span to current destination * - upload data as new buffer and blit @@ -261,13 +266,14 @@ st_texture_image_data(struct pipe_context *pipe, } } + /* Copy mipmap image between textures */ void st_texture_image_copy(struct pipe_context *pipe, - struct pipe_texture *dst, - GLuint face, GLuint level, - struct pipe_texture *src) + struct pipe_texture *dst, + GLuint face, GLuint level, + struct pipe_texture *src) { GLuint width = src->width[level]; GLuint height = src->height[level]; @@ -278,6 +284,7 @@ st_texture_image_copy(struct pipe_context *pipe, if (dst->compressed) height /= 4; + for (i = 0; i < depth; i++) { dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); src_surface = pipe->get_tex_surface(pipe, src, face, level, i); @@ -292,5 +299,4 @@ st_texture_image_copy(struct pipe_context *pipe, pipe_surface_reference(&dst_surface, NULL); pipe_surface_reference(&src_surface, NULL); } - } diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index d8b1bcad9d..0b87a494c3 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -47,11 +47,11 @@ st_texture_create(struct st_context *st, GLuint compress_byte); -/* Check if an image fits an existing texture +/* Check if an image fits into an existing texture object. */ extern GLboolean -st_texture_match_image(struct pipe_texture *pt, - struct gl_texture_image *image, +st_texture_match_image(const struct pipe_texture *pt, + const struct gl_texture_image *image, GLuint face, GLuint level); /* Return a pointer to an image within a texture. Return image stride as @@ -73,7 +73,7 @@ extern const GLuint * st_texture_depth_offsets(struct pipe_texture *pt, GLuint level); -/* Return the linear offset of an image relative to the start of its region: +/* Return the linear offset of an image relative to the start of its region. */ extern GLuint st_texture_image_offset(const struct pipe_texture *pt, -- cgit v1.2.3 From 105b3596be6c0644e3aaa0823ab7e27aa76fa909 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 18:01:18 -0700 Subject: added comment --- src/mesa/state_tracker/st_atom_texture.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index b3859f18cb..2a836d630b 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -59,6 +59,7 @@ update_textures(struct st_context *st) GLboolean flush, retval; retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); + /* XXX retval indicates whether there's a texture border */ pt = st_get_texobj_texture(texObj); } -- cgit v1.2.3 From afc54983370033b65e3a7cbb29bd9c87156f0881 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 18:13:28 -0700 Subject: clean-ups --- src/mesa/state_tracker/st_cb_texture.c | 67 +++++++++++++++------------------- 1 file changed, 29 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index ba0950e295..eee94baa20 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -76,13 +76,13 @@ struct st_texture_object - static INLINE struct st_texture_object * st_texture_object(struct gl_texture_object *obj) { return (struct st_texture_object *) obj; } + static INLINE struct st_texture_image * st_texture_image(struct gl_texture_image *img) { @@ -122,32 +122,28 @@ gl_target_to_pipe(GLenum target) } +/** + * Return nominal bytes per texel for a compressed format, 0 for non-compressed + * format. + */ static int compressed_num_bytes(GLuint mesaFormat) { - int bytes = 0; switch(mesaFormat) { - case MESA_FORMAT_RGB_FXT1: case MESA_FORMAT_RGBA_FXT1: case MESA_FORMAT_RGB_DXT1: case MESA_FORMAT_RGBA_DXT1: - bytes = 2; - break; - + return 2; case MESA_FORMAT_RGBA_DXT3: case MESA_FORMAT_RGBA_DXT5: - bytes = 4; + return 4; default: - break; + return 0; } - - return bytes; } - - static GLboolean st_IsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) { @@ -164,7 +160,6 @@ st_IsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) } - static struct gl_texture_image * st_NewTextureImage(GLcontext * ctx) { @@ -216,8 +211,6 @@ st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) } - - /* ================================================================ * From linux kernel i386 header files, copes with odd sizes better * than COPY_DWORDS would: @@ -302,7 +295,7 @@ logbase2(int n) static void guess_and_alloc_texture(struct st_context *st, struct st_texture_object *stObj, - struct st_texture_image *stImage) + const struct st_texture_image *stImage) { GLuint firstLevel; GLuint lastLevel; @@ -487,21 +480,18 @@ try_pbo_upload(GLcontext *ctx, - - - - static void st_TexImage(GLcontext * ctx, - GLint dims, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage, GLsizei imageSize, int compressed) + GLint dims, + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage, + GLsizei imageSize, int compressed) { struct st_texture_object *stObj = st_texture_object(texObj); struct st_texture_image *stImage = st_texture_image(texImage); @@ -524,7 +514,7 @@ st_TexImage(GLcontext * ctx, /* choose the texture format */ texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat, - format, type); + format, type); _mesa_set_fetch_functions(texImage, dims); @@ -536,7 +526,8 @@ st_TexImage(GLcontext * ctx, ctx->Driver.CompressedTextureSize(ctx, texImage->Width, texImage->Height, texImage->Depth, texImage->TexFormat->MesaFormat); - } else { + } + else { texelBytes = texImage->TexFormat->TexelBytes; /* Minimum pitch of 32 bytes */ @@ -669,7 +660,7 @@ st_TexImage(GLcontext * ctx, * conversion and copy: */ if (compressed) { - memcpy(texImage->Data, pixels, imageSize); + memcpy(texImage->Data, pixels, imageSize); } else { GLuint srcImageStride = _mesa_image_image_stride(unpack, width, height, @@ -1401,7 +1392,10 @@ copy_image_data_to_texture(struct st_context *st, } -/* +/** + * Called during state validation. When this function is finished, + * the texture object should be ready for rendering. + * \return GL_FALSE if a texture border is present, GL_TRUE otherwise */ GLboolean st_finalize_texture(GLcontext *ctx, @@ -1410,11 +1404,10 @@ st_finalize_texture(GLcontext *ctx, GLboolean *needFlush) { struct st_texture_object *stObj = st_texture_object(tObj); + const GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; int comp_byte = 0; int cpp; - GLuint face, i; - GLuint nr_faces = 0; struct st_texture_image *firstImage; *needFlush = GL_FALSE; @@ -1426,8 +1419,7 @@ st_finalize_texture(GLcontext *ctx, /* What levels must the texture include at a minimum? */ calculate_first_last_level(stObj); - firstImage = - st_texture_image(stObj->base.Image[0][stObj->firstLevel]); + firstImage = st_texture_image(stObj->base.Image[0][stObj->firstLevel]); /* Fallback case: */ @@ -1503,7 +1495,6 @@ st_finalize_texture(GLcontext *ctx, /* Pull in any images not in the object's texture: */ - nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; for (face = 0; face < nr_faces; face++) { for (i = stObj->firstLevel; i <= stObj->lastLevel; i++) { struct st_texture_image *stImage = -- cgit v1.2.3 From 4c2f3dbca940f289e67248682b84a3516d5a3031 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 18:15:03 -0700 Subject: Added ctx->Driver.GenerateMipmap() driver hook --- src/mesa/drivers/common/driverfuncs.c | 2 ++ src/mesa/main/dd.h | 7 +++++ src/mesa/main/fbobject.c | 2 +- src/mesa/main/texstore.c | 48 +++++++++++++++++------------------ 4 files changed, 34 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/common/driverfuncs.c b/src/mesa/drivers/common/driverfuncs.c index 33caf7dae1..b5b383b4e4 100644 --- a/src/mesa/drivers/common/driverfuncs.c +++ b/src/mesa/drivers/common/driverfuncs.c @@ -28,6 +28,7 @@ #include "buffers.h" #include "context.h" #include "framebuffer.h" +#include "mipmap.h" #include "program.h" #include "prog_execute.h" #include "queryobj.h" @@ -99,6 +100,7 @@ _mesa_init_driver_functions(struct dd_function_table *driver) driver->CopyTexSubImage1D = _swrast_copy_texsubimage1d; driver->CopyTexSubImage2D = _swrast_copy_texsubimage2d; driver->CopyTexSubImage3D = _swrast_copy_texsubimage3d; + driver->GenerateMipmap = _mesa_generate_mipmap; driver->TestProxyTexImage = _mesa_test_proxy_teximage; driver->CompressedTexImage1D = _mesa_store_compressed_teximage1d; driver->CompressedTexImage2D = _mesa_store_compressed_teximage2d; diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index 3bec3bd433..c2ef67ba6d 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -332,6 +332,13 @@ struct dd_function_table { GLint x, GLint y, GLsizei width, GLsizei height ); + /** + * Called by glGenerateMipmap() or when GL_GENERATE_MIPMAP_SGIS is enabled. + */ + void (*GenerateMipmap)(GLcontext *ctx, GLenum target, + const struct gl_texture_unit *texUnit, + struct gl_texture_object *texObj); + /** * Called by glTexImage[123]D when user specifies a proxy texture * target. diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 963e35d678..13cbd35424 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -1560,7 +1560,7 @@ _mesa_GenerateMipmapEXT(GLenum target) /* XXX this might not handle cube maps correctly */ _mesa_lock_texture(ctx, texObj); - _mesa_generate_mipmap(ctx, target, texUnit, texObj); + ctx->Driver.GenerateMipmap(ctx, target, texUnit, texObj); _mesa_unlock_texture(ctx, texObj); } diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index 30be65525e..26ca4f1bd5 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -2917,9 +2917,9 @@ _mesa_store_teximage1d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3003,9 +3003,9 @@ _mesa_store_teximage2d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3079,9 +3079,9 @@ _mesa_store_teximage3d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3127,9 +3127,9 @@ _mesa_store_texsubimage1d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3182,9 +3182,9 @@ _mesa_store_texsubimage2d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3237,9 +3237,9 @@ _mesa_store_texsubimage3d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3313,9 +3313,9 @@ _mesa_store_compressed_teximage2d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); } _mesa_unmap_teximage_pbo(ctx, &ctx->Unpack); @@ -3425,9 +3425,9 @@ _mesa_store_compressed_texsubimage2d(GLcontext *ctx, GLenum target, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, + &ctx->Texture.Unit[ctx->Texture.CurrentUnit], + texObj); } _mesa_unmap_teximage_pbo(ctx, &ctx->Unpack); -- cgit v1.2.3 From 2440ff74d69a8caf49b05a960b4c7e282a96565e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 5 Feb 2008 18:15:39 -0700 Subject: plug in ctx->Driver.GenerateMipmap function --- src/mesa/state_tracker/st_cb_texture.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index eee94baa20..15c5359360 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -30,6 +30,7 @@ #include "main/enums.h" #include "main/image.h" #include "main/macros.h" +#include "main/mipmap.h" #include "main/texcompress.h" #include "main/texformat.h" #include "main/teximage.h" @@ -1531,6 +1532,7 @@ st_init_texture_functions(struct dd_function_table *functions) functions->CopyTexSubImage1D = st_CopyTexSubImage1D; functions->CopyTexSubImage2D = st_CopyTexSubImage2D; functions->CopyTexSubImage3D = st_CopyTexSubImage3D; + functions->GenerateMipmap = _mesa_generate_mipmap; functions->GetTexImage = st_GetTexImage; -- cgit v1.2.3 From f52f5136e6eed23e55098681e5b082cc452136d6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 6 Feb 2008 08:44:19 -0700 Subject: clean-ups in guess_and_alloc_texture() --- src/mesa/state_tracker/st_cb_texture.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 15c5359360..8db4a804ab 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -284,7 +284,12 @@ logbase2(int n) } -/* Otherwise, store it in memory if (Border != 0) or (any dimension == +/** + * Allocate a pipe_texture object for the given st_texture_object using + * the given st_texture_image to guess the mipmap size/levels. + * + * [comments...] + * Otherwise, store it in memory if (Border != 0) or (any dimension == * 1). * * Otherwise, if max_level >= level >= min_level, create texture with @@ -303,11 +308,12 @@ guess_and_alloc_texture(struct st_context *st, GLuint width = stImage->base.Width; GLuint height = stImage->base.Height; GLuint depth = stImage->base.Depth; - GLuint l2width, l2height, l2depth; GLuint i, comp_byte = 0; DBG("%s\n", __FUNCTION__); + assert(!stObj->pt); + if (stImage->base.Border) return; @@ -349,15 +355,15 @@ guess_and_alloc_texture(struct st_context *st, lastLevel = firstLevel; } else { - l2width = logbase2(width); - l2height = logbase2(height); - l2depth = logbase2(depth); + GLuint l2width = logbase2(width); + GLuint l2height = logbase2(height); + GLuint l2depth = logbase2(depth); lastLevel = firstLevel + MAX2(MAX2(l2width, l2height), l2depth); } - assert(!stObj->pt); if (stImage->base.IsCompressed) comp_byte = compressed_num_bytes(stImage->base.TexFormat->MesaFormat); + stObj->pt = st_texture_create(st, gl_target_to_pipe(stObj->base.Target), st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat), -- cgit v1.2.3 From 31c98eafb043cbc82e5de206ceecc5888174b5e6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 6 Feb 2008 09:24:30 -0700 Subject: gallium: change pipe->texture_create() to operate like the CSO functions Now, pass in a template object and return a new object. --- src/mesa/pipe/cell/ppu/cell_texture.c | 31 ++++++++++++++--------------- src/mesa/pipe/cell/ppu/cell_texture.h | 5 +++-- src/mesa/pipe/i915simple/i915_texture.c | 17 ++++++++-------- src/mesa/pipe/i915simple/i915_texture.h | 5 +++-- src/mesa/pipe/i965simple/brw_tex_layout.c | 15 +++++++------- src/mesa/pipe/i965simple/brw_tex_layout.h | 4 ++-- src/mesa/pipe/p_context.h | 4 ++-- src/mesa/pipe/softpipe/sp_texture.c | 33 +++++++++++++++---------------- src/mesa/pipe/softpipe/sp_texture.h | 5 +++-- src/mesa/state_tracker/st_texture.c | 31 ++++++++++++----------------- 10 files changed, 73 insertions(+), 77 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_texture.c b/src/mesa/pipe/cell/ppu/cell_texture.c index 2cf6022939..df178d9ca2 100644 --- a/src/mesa/pipe/cell/ppu/cell_texture.c +++ b/src/mesa/pipe/cell/ppu/cell_texture.c @@ -79,31 +79,30 @@ cell_texture_layout(struct cell_texture * spt) } -void -cell_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) +struct pipe_texture * +cell_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) { - struct cell_texture *spt = REALLOC(*pt, sizeof(struct pipe_texture), - sizeof(struct cell_texture)); + struct cell_texture *spt = CALLOC_STRUCT(cell_texture); + if (!spt) + return NULL; - if (spt) { - memset(&spt->base + 1, 0, - sizeof(struct cell_texture) - sizeof(struct pipe_texture)); + spt->base = *templat; - cell_texture_layout(spt); + cell_texture_layout(spt); - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); - if (!spt->buffer) { - FREE(spt); - spt = NULL; - } + if (!spt->buffer) { + FREE(spt); + return NULL; } - *pt = &spt->base; + return &spt->base; } + void cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) { diff --git a/src/mesa/pipe/cell/ppu/cell_texture.h b/src/mesa/pipe/cell/ppu/cell_texture.h index bd434c8776..0264fed88e 100644 --- a/src/mesa/pipe/cell/ppu/cell_texture.h +++ b/src/mesa/pipe/cell/ppu/cell_texture.h @@ -60,8 +60,9 @@ cell_texture(struct pipe_texture *pt) -extern void -cell_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); +extern struct pipe_texture * +cell_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat); extern void cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c index 61944fe7d9..6faeab134a 100644 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -477,17 +477,17 @@ i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) return TRUE; } -void -i915_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) + +struct pipe_texture * +i915_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat) { - struct i915_texture *tex = REALLOC(*pt, sizeof(struct pipe_texture), - sizeof(struct i915_texture)); + struct i915_texture *tex = CALLOC_STRUCT(i915_texture); if (tex) { struct i915_context *i915 = i915_context(pipe); - memset(&tex->base + 1, 0, - sizeof(struct i915_texture) - sizeof(struct pipe_texture)); + tex->base = *templat; if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : i915_miptree_layout(pipe, tex)) @@ -498,13 +498,14 @@ i915_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) if (!tex->buffer) { FREE(tex); - tex = NULL; + return NULL; } } - *pt = &tex->base; + return &tex->base; } + void i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) { diff --git a/src/mesa/pipe/i915simple/i915_texture.h b/src/mesa/pipe/i915simple/i915_texture.h index 84a0502e81..330d111dc7 100644 --- a/src/mesa/pipe/i915simple/i915_texture.h +++ b/src/mesa/pipe/i915simple/i915_texture.h @@ -6,8 +6,9 @@ struct pipe_context; struct pipe_texture; -extern void -i915_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); +struct pipe_texture * +i915_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat); extern void i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.c b/src/mesa/pipe/i965simple/brw_tex_layout.c index b8b6b579e2..405fd1f794 100644 --- a/src/mesa/pipe/i965simple/brw_tex_layout.c +++ b/src/mesa/pipe/i965simple/brw_tex_layout.c @@ -299,15 +299,14 @@ static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture return TRUE; } -void -brw_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) + +struct pipe_texture * +brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) { - struct brw_texture *tex = REALLOC(*pt, sizeof(struct pipe_texture), - sizeof(struct brw_texture)); + struct brw_texture *tex = CALLOC_STRUCT(brw_texture); if (tex) { - memset(&tex->base + 1, 0, - sizeof(struct brw_texture) - sizeof(struct pipe_texture)); + tex->base = *templat; if (brw_miptree_layout(pipe, tex)) tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, @@ -317,11 +316,11 @@ brw_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) if (!tex->buffer) { FREE(tex); - tex = NULL; + return NULL; } } - *pt = &tex->base; + return &tex->base; } void diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.h b/src/mesa/pipe/i965simple/brw_tex_layout.h index 15e275058a..cfd6b1ef3a 100644 --- a/src/mesa/pipe/i965simple/brw_tex_layout.h +++ b/src/mesa/pipe/i965simple/brw_tex_layout.h @@ -6,8 +6,8 @@ struct pipe_context; struct pipe_texture; -extern void -brw_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); +extern struct pipe_texture * +brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat); extern void brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 0dda06c53b..92a1cd70c4 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -199,8 +199,8 @@ struct pipe_context { /* * Texture functions */ - void (*texture_create)(struct pipe_context *pipe, - struct pipe_texture **pt); + struct pipe_texture * (*texture_create)(struct pipe_context *pipe, + const struct pipe_texture *templat); void (*texture_release)(struct pipe_context *pipe, struct pipe_texture **pt); diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index 172234843d..fd2cc3dbbb 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -79,31 +79,30 @@ softpipe_texture_layout(struct softpipe_texture * spt) } -void -softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt) +struct pipe_texture * +softpipe_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat) { - struct softpipe_texture *spt = REALLOC(*pt, sizeof(struct pipe_texture), - sizeof(struct softpipe_texture)); - - if (spt) { - memset(&spt->base + 1, 0, - sizeof(struct softpipe_texture) - sizeof(struct pipe_texture)); + struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); + if (!spt) + return NULL; - softpipe_texture_layout(spt); + spt->base = *templat; - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); + softpipe_texture_layout(spt); - if (!spt->buffer) { - FREE(spt); - spt = NULL; - } + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); + if (!spt->buffer) { + FREE(spt); + return NULL; } - *pt = &spt->base; + return &spt->base; } + void softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) { diff --git a/src/mesa/pipe/softpipe/sp_texture.h b/src/mesa/pipe/softpipe/sp_texture.h index c6cf370351..fa646c0de9 100644 --- a/src/mesa/pipe/softpipe/sp_texture.h +++ b/src/mesa/pipe/softpipe/sp_texture.h @@ -55,8 +55,9 @@ softpipe_texture(struct pipe_texture *pt) -extern void -softpipe_texture_create(struct pipe_context *pipe, struct pipe_texture **pt); +extern struct pipe_texture * +softpipe_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat); extern void softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 741f36c2a7..844a9f80d8 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -74,7 +74,7 @@ st_texture_create(struct st_context *st, GLuint depth0, GLuint compress_byte) { - struct pipe_texture *pt = CALLOC_STRUCT(pipe_texture); + struct pipe_texture pt; assert(target <= PIPE_TEXTURE_CUBE); @@ -82,25 +82,20 @@ st_texture_create(struct st_context *st, _mesa_lookup_enum_by_nr(target), _mesa_lookup_enum_by_nr(format), first_level, last_level); - if (!pt) - return NULL; - assert(format); - pt->target = target; - pt->format = format; - pt->first_level = first_level; - pt->last_level = last_level; - pt->width[0] = width0; - pt->height[0] = height0; - pt->depth[0] = depth0; - pt->compressed = compress_byte ? 1 : 0; - pt->cpp = pt->compressed ? compress_byte : st_sizeof_format(format); - pt->refcount = 1; - - st->pipe->texture_create(st->pipe, &pt); - - return pt; + pt.target = target; + pt.format = format; + pt.first_level = first_level; + pt.last_level = last_level; + pt.width[0] = width0; + pt.height[0] = height0; + pt.depth[0] = depth0; + pt.compressed = compress_byte ? 1 : 0; + pt.cpp = pt.compressed ? compress_byte : st_sizeof_format(format); + pt.refcount = 1; + + return st->pipe->texture_create(st->pipe, &pt); } -- cgit v1.2.3 From c8af89cf722830ec16d594afd99d717aed71d44c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 6 Feb 2008 09:28:20 -0700 Subject: gallium: added mem_dup() --- src/mesa/pipe/p_util.h | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 4780ed7818..991ac447ba 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -183,6 +183,20 @@ align_free(void *ptr) +/** + * Duplicate of a block of memory + */ +static INLINE void * +mem_dup(const void *src, uint size) +{ + void *dup = malloc(size); + if (dup) + memcpy(dup, src, size); + return dup; +} + + + #define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) -- cgit v1.2.3 From f12d641ab2bafe20f876dddb90ada76c83732757 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 6 Feb 2008 09:29:00 -0700 Subject: gallium: use mem_dup() --- src/mesa/pipe/softpipe/sp_state_blend.c | 10 +++------- src/mesa/pipe/softpipe/sp_state_rasterizer.c | 7 ++----- src/mesa/pipe/softpipe/sp_state_sampler.c | 4 +--- 3 files changed, 6 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c index 160ca5cbc0..2d40d6bd8f 100644 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ b/src/mesa/pipe/softpipe/sp_state_blend.c @@ -32,13 +32,12 @@ #include "sp_context.h" #include "sp_state.h" + void * softpipe_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - struct pipe_blend_state *state = MALLOC( sizeof(struct pipe_blend_state) ); - memcpy(state, blend, sizeof(struct pipe_blend_state)); - return state; + return mem_dup(blend, sizeof(*blend)); } void softpipe_bind_blend_state( struct pipe_context *pipe, @@ -78,10 +77,7 @@ void * softpipe_create_depth_stencil_state(struct pipe_context *pipe, const struct pipe_depth_stencil_alpha_state *depth_stencil) { - struct pipe_depth_stencil_alpha_state *state = - MALLOC( sizeof(struct pipe_depth_stencil_alpha_state) ); - memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_alpha_state)); - return state; + return mem_dup(depth_stencil, sizeof(*depth_stencil)); } void diff --git a/src/mesa/pipe/softpipe/sp_state_rasterizer.c b/src/mesa/pipe/softpipe/sp_state_rasterizer.c index ce8fa4f2b8..53755099dd 100644 --- a/src/mesa/pipe/softpipe/sp_state_rasterizer.c +++ b/src/mesa/pipe/softpipe/sp_state_rasterizer.c @@ -35,12 +35,9 @@ void * softpipe_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup) + const struct pipe_rasterizer_state *rast) { - struct pipe_rasterizer_state *state = - MALLOC( sizeof(struct pipe_rasterizer_state) ); - memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); - return state; + return mem_dup(rast, sizeof(*rast)); } void softpipe_bind_rasterizer_state(struct pipe_context *pipe, diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 3842e71503..51b4b78287 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -40,9 +40,7 @@ void * softpipe_create_sampler_state(struct pipe_context *pipe, const struct pipe_sampler_state *sampler) { - struct pipe_sampler_state *state = MALLOC( sizeof(struct pipe_sampler_state) ); - memcpy(state, sampler, sizeof(struct pipe_sampler_state)); - return state; + return mem_dup(sampler, sizeof(*sampler)); } void -- cgit v1.2.3 From c0235d0a24da82304f7f23936c71032c0a9a7ce1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 6 Feb 2008 09:35:33 -0700 Subject: Cell: use mem_dup() --- src/mesa/pipe/cell/ppu/cell_state_blend.c | 9 ++------- src/mesa/pipe/cell/ppu/cell_state_sampler.c | 4 +--- 2 files changed, 3 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_blend.c b/src/mesa/pipe/cell/ppu/cell_state_blend.c index 2c19aa3971..4fc60548c8 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_blend.c +++ b/src/mesa/pipe/cell/ppu/cell_state_blend.c @@ -39,9 +39,7 @@ void * cell_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - struct pipe_blend_state *state = MALLOC(sizeof(struct pipe_blend_state)); - memcpy(state, blend, sizeof(struct pipe_blend_state)); - return state; + return mem_dup(blend, sizeof(*blend)); } @@ -85,10 +83,7 @@ void * cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, const struct pipe_depth_stencil_alpha_state *depth_stencil) { - struct pipe_depth_stencil_alpha_state *state = - MALLOC(sizeof(struct pipe_depth_stencil_alpha_state)); - memcpy(state, depth_stencil, sizeof(struct pipe_depth_stencil_alpha_state)); - return state; + return mem_dup(depth_stencil, sizeof(*depth_stencil)); } diff --git a/src/mesa/pipe/cell/ppu/cell_state_sampler.c b/src/mesa/pipe/cell/ppu/cell_state_sampler.c index 317f7603bb..ade6cc8338 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_sampler.c +++ b/src/mesa/pipe/cell/ppu/cell_state_sampler.c @@ -40,9 +40,7 @@ void * cell_create_sampler_state(struct pipe_context *pipe, const struct pipe_sampler_state *sampler) { - struct pipe_sampler_state *state = MALLOC( sizeof(struct pipe_sampler_state) ); - memcpy(state, sampler, sizeof(struct pipe_sampler_state)); - return state; + return mem_dup(sampler, sizeof(*sampler)); } void -- cgit v1.2.3 From 2ed6604f50bc06cfd44b429f476588381113db9c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 6 Feb 2008 09:48:37 -0700 Subject: gallium: #include p_debug.h since we use assert --- src/mesa/pipe/p_util.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 991ac447ba..469920efee 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -29,6 +29,7 @@ #define P_UTIL_H #include "p_compiler.h" +#include "p_debug.h" #include -- cgit v1.2.3 From c4e0d725dc9f18aed2babed344bb4e42df9e481f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 6 Feb 2008 09:48:56 -0700 Subject: Cell: silence unused var warnings --- src/mesa/pipe/cell/ppu/cell_state_fs.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_fs.c b/src/mesa/pipe/cell/ppu/cell_state_fs.c index 81c2ac14dd..96a52273b0 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_fs.c +++ b/src/mesa/pipe/cell/ppu/cell_state_fs.c @@ -45,7 +45,7 @@ void * cell_create_fs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { - struct cell_context *cell = cell_context(pipe); + /*struct cell_context *cell = cell_context(pipe);*/ struct cell_fragment_shader_state *state; state = CALLOC_STRUCT(cell_fragment_shader_state); @@ -94,8 +94,6 @@ cell_bind_fs_state(struct pipe_context *pipe, void *fs) void cell_delete_fs_state(struct pipe_context *pipe, void *fs) { - struct cell_context *cell = cell_context(pipe); - struct cell_fragment_shader_state *state = (struct cell_fragment_shader_state *) fs; -- cgit v1.2.3 From 28ecb986d958d52c9a996453e53418d1f98d7f75 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 7 Feb 2008 19:43:34 +0000 Subject: pipebuffer: Fix reversed assertion. --- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c index bcd4b3e257..04477a865a 100644 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c @@ -170,7 +170,7 @@ pool_bufmgr_create_buffer(struct pb_manager *mgr, struct list_head *item; assert(size == pool->bufSize); - assert(desc->alignment % pool->bufAlign == 0); + assert(pool->bufAlign % desc->alignment == 0); _glthread_LOCK_MUTEX(pool->mutex); -- cgit v1.2.3 From 909c703bfbf7404414befaa0a94b76d78ba3cb4c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 7 Feb 2008 19:44:42 +0000 Subject: tgsi: Fall back to interpreter instead of assert(0) on unimplemented SSE code. --- src/mesa/pipe/draw/draw_vertex_shader.c | 22 ++++++++++++++++------ src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 3 +-- src/mesa/x86/rtasm/x86sse.c | 1 + 3 files changed, 18 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index e6590eafcc..5ca93aa615 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -121,11 +121,16 @@ run_vertex_program(struct draw_context *draw, = (struct draw_vertex_shader *)draw->vertex_shader; codegen_function func = (codegen_function) x86_get_func( &shader->sse2_program ); - func( - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps ); + + if (func) + func( + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps ); + else + /* interpreter */ + tgsi_exec_machine_run( machine ); } else #endif @@ -269,7 +274,12 @@ draw_create_vertex_shader(struct draw_context *draw, struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; x86_init_func( &vs->sse2_program ); - tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, &vs->sse2_program ); + if (!tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, + &vs->sse2_program )) { + x86_release_func( (struct x86_function *) &vs->sse2_program ); + fprintf(stdout /*err*/, + "tgsi_emit_sse2() failed, falling back to interpreter\n"); + } } #endif diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index f2180082f1..40bacf8552 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -2254,8 +2254,7 @@ tgsi_emit_sse2( case TGSI_TOKEN_TYPE_IMMEDIATE: /* XXX implement this */ - assert(0); - break; + return 0; default: assert( 0 ); diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index 56c211eee0..f8da6e405f 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -1137,6 +1137,7 @@ void x86_init_func_size( struct x86_function *p, unsigned code_size ) void x86_release_func( struct x86_function *p ) { _mesa_exec_free(p->store); + p->store = NULL; } -- cgit v1.2.3 From 2b6a31bfda319975b728930f019175611145ebb9 Mon Sep 17 00:00:00 2001 From: Jerome Glisse Date: Fri, 8 Feb 2008 18:25:49 +0100 Subject: failover: several fixes to failover pipe module --- src/mesa/pipe/failover/fo_context.c | 2 + src/mesa/pipe/failover/fo_state.c | 79 ++++++++++++++++++++++++++----------- 2 files changed, 57 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c index cf6c9fed50..7ce4a7df17 100644 --- a/src/mesa/pipe/failover/fo_context.c +++ b/src/mesa/pipe/failover/fo_context.c @@ -114,6 +114,8 @@ struct pipe_context *failover_create( struct pipe_context *hw, if (failover == NULL) return NULL; + failover->hw = hw; + failover->sw = sw; failover->pipe.winsys = hw->winsys; failover->pipe.destroy = failover_destroy; failover->pipe.is_format_supported = hw->is_format_supported; diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c index fa700b9674..0fc5568da1 100644 --- a/src/mesa/pipe/failover/fo_state.c +++ b/src/mesa/pipe/failover/fo_state.c @@ -54,8 +54,8 @@ failover_create_blend_state( struct pipe_context *pipe, struct fo_state *state = malloc(sizeof(struct fo_state)); struct failover_context *failover = failover_context(pipe); - state->sw_state = failover->sw->create_blend_state(pipe, blend); - state->hw_state = failover->hw->create_blend_state(pipe, blend); + state->sw_state = failover->sw->create_blend_state(failover->sw, blend); + state->hw_state = failover->hw->create_blend_state(failover->hw, blend); return state; } @@ -68,6 +68,7 @@ failover_bind_blend_state( struct pipe_context *pipe, struct fo_state *state = (struct fo_state *)blend; failover->blend = state; failover->dirty |= FO_NEW_BLEND; + failover->sw->bind_blend_state( failover->sw, state->sw_state ); failover->hw->bind_blend_state( failover->hw, state->hw_state ); } @@ -78,8 +79,8 @@ failover_delete_blend_state( struct pipe_context *pipe, struct fo_state *state = (struct fo_state*)blend; struct failover_context *failover = failover_context(pipe); - failover->sw->delete_blend_state(pipe, state->sw_state); - failover->hw->delete_blend_state(pipe, state->hw_state); + failover->sw->delete_blend_state(failover->sw, state->sw_state); + failover->hw->delete_blend_state(failover->hw, state->hw_state); state->sw_state = 0; state->hw_state = 0; free(state); @@ -93,6 +94,7 @@ failover_set_blend_color( struct pipe_context *pipe, failover->blend_color = *blend_color; failover->dirty |= FO_NEW_BLEND_COLOR; + failover->sw->set_blend_color( failover->sw, blend_color ); failover->hw->set_blend_color( failover->hw, blend_color ); } @@ -104,6 +106,7 @@ failover_set_clip_state( struct pipe_context *pipe, failover->clip = *clip; failover->dirty |= FO_NEW_CLIP; + failover->sw->set_clip_state( failover->sw, clip ); failover->hw->set_clip_state( failover->hw, clip ); } @@ -115,8 +118,8 @@ failover_create_depth_stencil_state(struct pipe_context *pipe, struct fo_state *state = malloc(sizeof(struct fo_state)); struct failover_context *failover = failover_context(pipe); - state->sw_state = failover->sw->create_depth_stencil_alpha_state(pipe, templ); - state->hw_state = failover->hw->create_depth_stencil_alpha_state(pipe, templ); + state->sw_state = failover->sw->create_depth_stencil_alpha_state(failover->sw, templ); + state->hw_state = failover->hw->create_depth_stencil_alpha_state(failover->hw, templ); return state; } @@ -129,6 +132,7 @@ failover_bind_depth_stencil_state(struct pipe_context *pipe, struct fo_state *state = (struct fo_state *)depth_stencil; failover->depth_stencil = state; failover->dirty |= FO_NEW_DEPTH_STENCIL; + failover->sw->bind_depth_stencil_alpha_state(failover->sw, state->sw_state); failover->hw->bind_depth_stencil_alpha_state(failover->hw, state->hw_state); } @@ -139,8 +143,8 @@ failover_delete_depth_stencil_state(struct pipe_context *pipe, struct fo_state *state = (struct fo_state*)ds; struct failover_context *failover = failover_context(pipe); - failover->sw->delete_depth_stencil_alpha_state(pipe, state->sw_state); - failover->hw->delete_depth_stencil_alpha_state(pipe, state->hw_state); + failover->sw->delete_depth_stencil_alpha_state(failover->sw, state->sw_state); + failover->hw->delete_depth_stencil_alpha_state(failover->hw, state->hw_state); state->sw_state = 0; state->hw_state = 0; free(state); @@ -154,6 +158,7 @@ failover_set_framebuffer_state(struct pipe_context *pipe, failover->framebuffer = *framebuffer; failover->dirty |= FO_NEW_FRAMEBUFFER; + failover->sw->set_framebuffer_state( failover->sw, framebuffer ); failover->hw->set_framebuffer_state( failover->hw, framebuffer ); } @@ -165,8 +170,8 @@ failover_create_fs_state(struct pipe_context *pipe, struct fo_state *state = malloc(sizeof(struct fo_state)); struct failover_context *failover = failover_context(pipe); - state->sw_state = failover->sw->create_fs_state(pipe, templ); - state->hw_state = failover->hw->create_fs_state(pipe, templ); + state->sw_state = failover->sw->create_fs_state(failover->sw, templ); + state->hw_state = failover->hw->create_fs_state(failover->hw, templ); return state; } @@ -178,6 +183,7 @@ failover_bind_fs_state(struct pipe_context *pipe, void *fs) struct fo_state *state = (struct fo_state*)fs; failover->fragment_shader = state; failover->dirty |= FO_NEW_FRAGMENT_SHADER; + failover->sw->bind_fs_state(failover->sw, state->sw_state); failover->hw->bind_fs_state(failover->hw, state->hw_state); } @@ -188,8 +194,8 @@ failover_delete_fs_state(struct pipe_context *pipe, struct fo_state *state = (struct fo_state*)fs; struct failover_context *failover = failover_context(pipe); - failover->sw->delete_fs_state(pipe, state->sw_state); - failover->hw->delete_fs_state(pipe, state->hw_state); + failover->sw->delete_fs_state(failover->sw, state->sw_state); + failover->hw->delete_fs_state(failover->hw, state->hw_state); state->sw_state = 0; state->hw_state = 0; free(state); @@ -202,8 +208,8 @@ failover_create_vs_state(struct pipe_context *pipe, struct fo_state *state = malloc(sizeof(struct fo_state)); struct failover_context *failover = failover_context(pipe); - state->sw_state = failover->sw->create_vs_state(pipe, templ); - state->hw_state = failover->hw->create_vs_state(pipe, templ); + state->sw_state = failover->sw->create_vs_state(failover->sw, templ); + state->hw_state = failover->hw->create_vs_state(failover->hw, templ); return state; } @@ -217,6 +223,7 @@ failover_bind_vs_state(struct pipe_context *pipe, struct fo_state *state = (struct fo_state*)vs; failover->vertex_shader = state; failover->dirty |= FO_NEW_VERTEX_SHADER; + failover->sw->bind_vs_state(failover->sw, state->sw_state); failover->hw->bind_vs_state(failover->hw, state->hw_state); } @@ -227,8 +234,8 @@ failover_delete_vs_state(struct pipe_context *pipe, struct fo_state *state = (struct fo_state*)vs; struct failover_context *failover = failover_context(pipe); - failover->sw->delete_vs_state(pipe, state->sw_state); - failover->hw->delete_vs_state(pipe, state->hw_state); + failover->sw->delete_vs_state(failover->sw, state->sw_state); + failover->hw->delete_vs_state(failover->hw, state->hw_state); state->sw_state = 0; state->hw_state = 0; free(state); @@ -242,6 +249,7 @@ failover_set_polygon_stipple( struct pipe_context *pipe, failover->poly_stipple = *stipple; failover->dirty |= FO_NEW_STIPPLE; + failover->sw->set_polygon_stipple( failover->sw, stipple ); failover->hw->set_polygon_stipple( failover->hw, stipple ); } @@ -253,8 +261,8 @@ failover_create_rasterizer_state(struct pipe_context *pipe, struct fo_state *state = malloc(sizeof(struct fo_state)); struct failover_context *failover = failover_context(pipe); - state->sw_state = failover->sw->create_rasterizer_state(pipe, templ); - state->hw_state = failover->hw->create_rasterizer_state(pipe, templ); + state->sw_state = failover->sw->create_rasterizer_state(failover->sw, templ); + state->hw_state = failover->hw->create_rasterizer_state(failover->hw, templ); return state; } @@ -268,6 +276,7 @@ failover_bind_rasterizer_state(struct pipe_context *pipe, struct fo_state *state = (struct fo_state*)raster; failover->rasterizer = state; failover->dirty |= FO_NEW_RASTERIZER; + failover->sw->bind_rasterizer_state(failover->sw, state->sw_state); failover->hw->bind_rasterizer_state(failover->hw, state->hw_state); } @@ -278,8 +287,8 @@ failover_delete_rasterizer_state(struct pipe_context *pipe, struct fo_state *state = (struct fo_state*)raster; struct failover_context *failover = failover_context(pipe); - failover->sw->delete_rasterizer_state(pipe, state->sw_state); - failover->hw->delete_rasterizer_state(pipe, state->hw_state); + failover->sw->delete_rasterizer_state(failover->sw, state->sw_state); + failover->hw->delete_rasterizer_state(failover->hw, state->hw_state); state->sw_state = 0; state->hw_state = 0; free(state); @@ -294,6 +303,7 @@ failover_set_scissor_state( struct pipe_context *pipe, failover->scissor = *scissor; failover->dirty |= FO_NEW_SCISSOR; + failover->sw->set_scissor_state( failover->sw, scissor ); failover->hw->set_scissor_state( failover->hw, scissor ); } @@ -305,8 +315,8 @@ failover_create_sampler_state(struct pipe_context *pipe, struct fo_state *state = malloc(sizeof(struct fo_state)); struct failover_context *failover = failover_context(pipe); - state->sw_state = failover->sw->create_sampler_state(pipe, templ); - state->hw_state = failover->hw->create_sampler_state(pipe, templ); + state->sw_state = failover->sw->create_sampler_state(failover->sw, templ); + state->hw_state = failover->hw->create_sampler_state(failover->hw, templ); return state; } @@ -320,6 +330,8 @@ failover_bind_sampler_state(struct pipe_context *pipe, failover->sampler[unit] = state; failover->dirty |= FO_NEW_SAMPLER; failover->dirty_sampler |= (1<sw->bind_sampler_state(failover->sw, unit, + state->sw_state); failover->hw->bind_sampler_state(failover->hw, unit, state->hw_state); } @@ -330,8 +342,8 @@ failover_delete_sampler_state(struct pipe_context *pipe, void *sampler) struct fo_state *state = (struct fo_state*)sampler; struct failover_context *failover = failover_context(pipe); - failover->sw->delete_sampler_state(pipe, state->sw_state); - failover->hw->delete_sampler_state(pipe, state->hw_state); + failover->sw->delete_sampler_state(failover->sw, state->sw_state); + failover->hw->delete_sampler_state(failover->hw, state->hw_state); state->sw_state = 0; state->hw_state = 0; free(state); @@ -348,6 +360,7 @@ failover_set_sampler_texture(struct pipe_context *pipe, failover->texture[unit] = texture; failover->dirty |= FO_NEW_TEXTURE; failover->dirty_texture |= (1<sw->set_sampler_texture( failover->sw, unit, texture ); failover->hw->set_sampler_texture( failover->hw, unit, texture ); } @@ -360,6 +373,7 @@ failover_set_viewport_state( struct pipe_context *pipe, failover->viewport = *viewport; failover->dirty |= FO_NEW_VIEWPORT; + failover->sw->set_viewport_state( failover->sw, viewport ); failover->hw->set_viewport_state( failover->hw, viewport ); } @@ -374,6 +388,7 @@ failover_set_vertex_buffer(struct pipe_context *pipe, failover->vertex_buffer[unit] = *vertex_buffer; failover->dirty |= FO_NEW_VERTEX_BUFFER; failover->dirty_vertex_buffer |= (1<sw->set_vertex_buffer( failover->sw, unit, vertex_buffer ); failover->hw->set_vertex_buffer( failover->hw, unit, vertex_buffer ); } @@ -388,9 +403,24 @@ failover_set_vertex_element(struct pipe_context *pipe, failover->vertex_element[unit] = *vertex_element; failover->dirty |= FO_NEW_VERTEX_ELEMENT; failover->dirty_vertex_element |= (1<sw->set_vertex_element( failover->sw, unit, vertex_element ); failover->hw->set_vertex_element( failover->hw, unit, vertex_element ); } +void +failover_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct failover_context *failover = failover_context(pipe); + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + failover->sw->set_constant_buffer(failover->sw, shader, index, buf); + failover->hw->set_constant_buffer(failover->hw, shader, index, buf); +} + void failover_init_state_functions( struct failover_context *failover ) @@ -423,4 +453,5 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.set_viewport_state = failover_set_viewport_state; failover->pipe.set_vertex_buffer = failover_set_vertex_buffer; failover->pipe.set_vertex_element = failover_set_vertex_element; + failover->pipe.set_constant_buffer = failover_set_constant_buffer; } -- cgit v1.2.3 From 21e9396e650d23084bfeae0d2670b5ffcf731a85 Mon Sep 17 00:00:00 2001 From: Jerome Glisse Date: Fri, 8 Feb 2008 18:47:25 +0100 Subject: intel_winsys: remove leftover code --- src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c | 5 ----- 1 file changed, 5 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c index 910c0d2cc5..789a386500 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c @@ -224,11 +224,6 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, if(!surf->buffer) return -1; - if(ret) { - pipe_buffer_reference(winsys, &surf->buffer, NULL); - return ret; - } - return 0; } -- cgit v1.2.3 From c3395f4473c8fdf75d04c0dd72e687bc8d8127a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 14:45:58 -0700 Subject: Remove unused texunit parameter to ctx->Driver.GenerateMipmap() --- src/mesa/main/dd.h | 3 +-- src/mesa/main/fbobject.c | 2 +- src/mesa/main/mipmap.c | 1 - src/mesa/main/mipmap.h | 1 - src/mesa/main/texstore.c | 32 ++++++++------------------------ 5 files changed, 10 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index c2ef67ba6d..37ef2a865b 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -335,8 +335,7 @@ struct dd_function_table { /** * Called by glGenerateMipmap() or when GL_GENERATE_MIPMAP_SGIS is enabled. */ - void (*GenerateMipmap)(GLcontext *ctx, GLenum target, - const struct gl_texture_unit *texUnit, + void (*GenerateMipmap)(GLcontext *ctx, GLenum target, struct gl_texture_object *texObj); /** diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 13cbd35424..6a8cba4d8a 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -1560,7 +1560,7 @@ _mesa_GenerateMipmapEXT(GLenum target) /* XXX this might not handle cube maps correctly */ _mesa_lock_texture(ctx, texObj); - ctx->Driver.GenerateMipmap(ctx, target, texUnit, texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); _mesa_unlock_texture(ctx, texObj); } diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index 9f3db22b75..1e61829e8f 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -933,7 +933,6 @@ make_2d_stack_mipmap(const struct gl_texture_format *format, GLint border, */ void _mesa_generate_mipmap(GLcontext *ctx, GLenum target, - const struct gl_texture_unit *texUnit, struct gl_texture_object *texObj) { const struct gl_texture_image *srcImage; diff --git a/src/mesa/main/mipmap.h b/src/mesa/main/mipmap.h index df78603283..46e16902c8 100644 --- a/src/mesa/main/mipmap.h +++ b/src/mesa/main/mipmap.h @@ -30,7 +30,6 @@ extern void _mesa_generate_mipmap(GLcontext *ctx, GLenum target, - const struct gl_texture_unit *texUnit, struct gl_texture_object *texObj); diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index 26ca4f1bd5..a6a18910fc 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -2917,9 +2917,7 @@ _mesa_store_teximage1d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - ctx->Driver.GenerateMipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3003,9 +3001,7 @@ _mesa_store_teximage2d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - ctx->Driver.GenerateMipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3079,9 +3075,7 @@ _mesa_store_teximage3d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - ctx->Driver.GenerateMipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3127,9 +3121,7 @@ _mesa_store_texsubimage1d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - ctx->Driver.GenerateMipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3182,9 +3174,7 @@ _mesa_store_texsubimage2d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - ctx->Driver.GenerateMipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3237,9 +3227,7 @@ _mesa_store_texsubimage3d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - ctx->Driver.GenerateMipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } _mesa_unmap_teximage_pbo(ctx, packing); @@ -3313,9 +3301,7 @@ _mesa_store_compressed_teximage2d(GLcontext *ctx, GLenum target, GLint level, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - ctx->Driver.GenerateMipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } _mesa_unmap_teximage_pbo(ctx, &ctx->Unpack); @@ -3425,9 +3411,7 @@ _mesa_store_compressed_texsubimage2d(GLcontext *ctx, GLenum target, /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - ctx->Driver.GenerateMipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } _mesa_unmap_teximage_pbo(ctx, &ctx->Unpack); -- cgit v1.2.3 From 864abce57d3b81d0f92673472959b71e09c4f245 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 14:46:47 -0700 Subject: gallium: added draw_flush() call in softpipe_bind_sampler_state() --- src/mesa/pipe/softpipe/sp_state_sampler.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 51b4b78287..291bbc40ad 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -49,6 +49,8 @@ softpipe_bind_sampler_state(struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); + draw_flush(softpipe->draw); + assert(unit < PIPE_MAX_SAMPLERS); softpipe->sampler[unit] = (struct pipe_sampler_state *)sampler; -- cgit v1.2.3 From 0b64ee6960f9e099bc1a6ca6fa10720fee875b3a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 14:51:32 -0700 Subject: gallium: added inClipCoords param to st_draw_vertices() to indicate coord system of vertices Also, export st_make_passthrough_vertex_shader() from st_cb_drawpixels.c --- src/mesa/state_tracker/st_cb_clear.c | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 18 +++++++++--------- src/mesa/state_tracker/st_cb_drawpixels.h | 4 ++++ src/mesa/state_tracker/st_draw.c | 21 ++++++++++++--------- src/mesa/state_tracker/st_draw.h | 3 ++- 5 files changed, 28 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 0cd469c156..ab98b54bab 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -251,7 +251,7 @@ draw_quad(GLcontext *ctx, verts[i][1][3] = color[3]; } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2); + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, GL_FALSE); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 34d420fcff..07886e7982 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -355,8 +355,8 @@ make_fragment_shader_z(struct st_context *st) * Create a simple vertex shader that just passes through the * vertex position and texcoord (and optionally, color). */ -static struct st_vertex_program * -make_vertex_shader(struct st_context *st, GLboolean passColor) +struct st_vertex_program * +st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor) { /* only make programs once and re-use */ static struct st_vertex_program *progs[2] = { NULL, NULL }; @@ -572,7 +572,7 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, verts[i][1][3] = 1.0; /*Q*/ } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2); + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, GL_FALSE); } @@ -625,7 +625,7 @@ draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, verts[i][2][3] = 1.0; /*Q*/ } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 3); + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 3, GL_FALSE); } @@ -945,7 +945,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (format == GL_DEPTH_COMPONENT) { ps = st->state.framebuffer.zsbuf; stfp = make_fragment_shader_z(ctx->st); - stvp = make_vertex_shader(ctx->st, GL_TRUE); + stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE); color = ctx->Current.RasterColor; } else if (format == GL_STENCIL_INDEX) { @@ -956,7 +956,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, else { ps = st->state.framebuffer.cbufs[0]; stfp = combined_drawpix_fragment_program(ctx); - stvp = make_vertex_shader(ctx->st, GL_FALSE); + stvp = st_make_passthrough_vertex_shader(ctx->st, GL_FALSE); color = NULL; } @@ -1111,7 +1111,7 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct st_context *st = ctx->st; struct pipe_texture *pt; - stvp = make_vertex_shader(ctx->st, GL_TRUE); + stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE); stfp = combined_bitmap_fragment_program(ctx); st_validate_state(st); @@ -1229,13 +1229,13 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, rbRead = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); color = NULL; stfp = combined_drawpix_fragment_program(ctx); - stvp = make_vertex_shader(ctx->st, GL_FALSE); + stvp = st_make_passthrough_vertex_shader(ctx->st, GL_FALSE); } else { rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer); color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; stfp = make_fragment_shader_z(ctx->st); - stvp = make_vertex_shader(ctx->st, GL_TRUE); + stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE); } psRead = rbRead->surface; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h index 71ba487020..b8b906f06b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.h +++ b/src/mesa/state_tracker/st_cb_drawpixels.h @@ -30,6 +30,10 @@ #define ST_CB_DRAWPIXELS_H +extern struct st_vertex_program * +st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor); + + extern void st_init_drawpixels_functions(struct dd_function_table *functions); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index c9b8e78485..ae9f5c8b11 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -354,7 +354,8 @@ st_draw_vbo(GLcontext *ctx, void st_draw_vertices(GLcontext *ctx, unsigned prim, unsigned numVertex, float *verts, - unsigned numAttribs) + unsigned numAttribs, + GLboolean inClipCoords) { const float width = ctx->DrawBuffer->Width; const float height = ctx->DrawBuffer->Height; @@ -367,14 +368,16 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, assert(numAttribs > 0); - /* convert to clip coords */ - for (i = 0; i < numVertex; i++) { - float x = verts[i * numAttribs * 4 + 0]; - float y = verts[i * numAttribs * 4 + 1]; - x = x / width * 2.0 - 1.0; - y = y / height * 2.0 - 1.0; - verts[i * numAttribs * 4 + 0] = x; - verts[i * numAttribs * 4 + 1] = y; + if (!inClipCoords) { + /* convert to clip coords */ + for (i = 0; i < numVertex; i++) { + float x = verts[i * numAttribs * 4 + 0]; + float y = verts[i * numAttribs * 4 + 1]; + x = x / width * 2.0 - 1.0; + y = y / height * 2.0 - 1.0; + verts[i * numAttribs * 4 + 0] = x; + verts[i * numAttribs * 4 + 1] = y; + } } /* XXX create one-time */ diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h index 89ee790c57..171bde57e5 100644 --- a/src/mesa/state_tracker/st_draw.h +++ b/src/mesa/state_tracker/st_draw.h @@ -62,7 +62,8 @@ st_feedback_draw_vbo(GLcontext *ctx, void st_draw_vertices(GLcontext *ctx, unsigned prim, unsigned numVertex, float *verts, - unsigned numAttribs); + unsigned numAttribs, + GLboolean inClipCoords); #endif -- cgit v1.2.3 From 62abcb9aacc33218d0143a743c738435794b32a9 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 14:54:18 -0700 Subject: gallium: initial implemenation of auto mipmap generation in state tracker Use hardware rendering to compute/render mipmap levels. The fallback path (which will be used for non-renderable texture formats) isn't working yet. --- src/mesa/sources | 1 + src/mesa/state_tracker/st_cb_texture.c | 11 +- src/mesa/state_tracker/st_context.c | 2 + src/mesa/state_tracker/st_gen_mipmap.c | 362 +++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_gen_mipmap.h | 46 +++++ 5 files changed, 415 insertions(+), 7 deletions(-) create mode 100644 src/mesa/state_tracker/st_gen_mipmap.c create mode 100644 src/mesa/state_tracker/st_gen_mipmap.h (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index c0087f76e6..84492c91ac 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -234,6 +234,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_extensions.c \ state_tracker/st_format.c \ state_tracker/st_framebuffer.c \ + state_tracker/st_gen_mipmap.c \ state_tracker/st_mesa_to_tgsi.c \ state_tracker/st_program.c \ state_tracker/st_texture.c diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 8db4a804ab..3350254654 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -42,6 +42,7 @@ #include "state_tracker/st_cb_texture.h" #include "state_tracker/st_format.h" #include "state_tracker/st_texture.h" +#include "state_tracker/st_gen_mipmap.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -703,13 +704,9 @@ st_TexImage(GLcontext * ctx, texImage->Data = NULL; } -#if 0 - /* GL_SGIS_generate_mipmap -- this can be accelerated now. - */ +#if 01 if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } #endif } @@ -1538,7 +1535,7 @@ st_init_texture_functions(struct dd_function_table *functions) functions->CopyTexSubImage1D = st_CopyTexSubImage1D; functions->CopyTexSubImage2D = st_CopyTexSubImage2D; functions->CopyTexSubImage3D = st_CopyTexSubImage3D; - functions->GenerateMipmap = _mesa_generate_mipmap; + functions->GenerateMipmap = st_generate_mipmap; functions->GetTexImage = st_GetTexImage; diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 9c206c057a..bf4618bed8 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -49,6 +49,7 @@ #include "st_atom.h" #include "st_draw.h" #include "st_extensions.h" +#include "st_gen_mipmap.h" #include "st_program.h" #include "pipe/p_context.h" #include "pipe/p_winsys.h" @@ -96,6 +97,7 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st_init_atoms( st ); st_init_draw( st ); + st_init_generate_mipmap(st); /* we want all vertex data to be placed in buffer objects */ vbo_use_buffer_objects(ctx); diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c new file mode 100644 index 0000000000..16f9e4cd27 --- /dev/null +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -0,0 +1,362 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "main/imports.h" +#include "main/mipmap.h" +#include "main/teximage.h" + +#include "shader/prog_instruction.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/cso_cache/cso_cache.h" + +#include "st_context.h" +#include "st_draw.h" +#include "st_gen_mipmap.h" +#include "st_program.h" +#include "st_cb_texture.h" + + + +static void *blend_cso = NULL; +static void *depthstencil_cso = NULL; +static void *rasterizer_cso = NULL; +static void *sampler_cso = NULL; + +static struct st_fragment_program *stfp = NULL; +static struct st_vertex_program *stvp = NULL; + + + +static struct st_fragment_program * +make_tex_fragment_program(GLcontext *ctx) +{ + struct st_fragment_program *stfp; + struct gl_program *p; + GLuint ic = 0; + + p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + if (!p) + return NULL; + + p->NumInstructions = 2; + + p->Instructions = _mesa_alloc_instructions(p->NumInstructions); + if (!p->Instructions) { + ctx->Driver.DeleteProgram(ctx, p); + return NULL; + } + _mesa_init_instructions(p->Instructions, p->NumInstructions); + + /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ + p->Instructions[ic].Opcode = OPCODE_TEX; + p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; + p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + p->Instructions[ic].TexSrcUnit = 0; + p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; + + /* END; */ + p->Instructions[ic++].Opcode = OPCODE_END; + + assert(ic == p->NumInstructions); + + p->InputsRead = FRAG_BIT_TEX0; + p->OutputsWritten = (1 << FRAG_RESULT_COLR); + + stfp = (struct st_fragment_program *) p; + + st_translate_fragment_program(ctx->st, stfp, NULL, + stfp->tokens, ST_MAX_SHADER_TOKENS); + + return stfp; +} + + + + +/** + * one-time init for generate mipmap + * XXX Note: there may be other times we need no-op/simple state like this. + * In that case, some code refactoring would be good. + */ +void +st_init_generate_mipmap(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + struct pipe_blend_state blend; + struct pipe_rasterizer_state rasterizer; + struct pipe_sampler_state sampler; + struct pipe_depth_stencil_alpha_state depthstencil; + + assert(!blend_cso); + + memset(&blend, 0, sizeof(blend)); + blend.colormask = PIPE_MASK_RGBA; + blend_cso = pipe->create_blend_state(pipe, &blend); + + memset(&depthstencil, 0, sizeof(depthstencil)); + depthstencil_cso = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil); + + memset(&rasterizer, 0, sizeof(rasterizer)); + rasterizer_cso = pipe->create_rasterizer_state(pipe, &rasterizer); + + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.normalized_coords = 1; + sampler_cso = pipe->create_sampler_state(pipe, &sampler); + + stfp = make_tex_fragment_program(st->ctx); + stvp = st_make_passthrough_vertex_shader(st, GL_FALSE); +} + + +void +st_destroy_generate_mipmpap(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + + pipe->delete_blend_state(pipe, blend_cso); + pipe->delete_depth_stencil_alpha_state(pipe, depthstencil_cso); + pipe->delete_rasterizer_state(pipe, rasterizer_cso); + pipe->delete_sampler_state(pipe, sampler_cso); + + /* XXX free stfp, stvp */ + + blend_cso = NULL; + depthstencil_cso = NULL; + rasterizer_cso = NULL; + sampler_cso = NULL; +} + + +static void +simple_viewport(struct pipe_context *pipe, uint width, uint height) +{ + struct pipe_viewport_state vp; + + vp.scale[0] = 0.5 * width; + vp.scale[1] = -0.5 * height; + vp.scale[2] = 1.0; + vp.scale[3] = 1.0; + vp.translate[0] = 0.5 * width; + vp.translate[1] = 0.5 * height; + vp.translate[2] = 0.0; + vp.translate[3] = 0.0; + + pipe->set_viewport_state(pipe, &vp); +} + + + +/* + * Draw simple [-1,1]x[-1,1] quad + */ +static void +draw_quad(GLcontext *ctx) +{ + GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ + GLuint i; + GLfloat sLeft = 0.0, sRight = 1.0; + GLfloat tTop = 1.0, tBot = 0.0; + GLfloat x0 = -1.0, x1 = 1.0; + GLfloat y0 = -1.0, y1 = 1.0; + + /* upper-left */ + verts[0][0][0] = x0; /* attr[0].x */ + verts[0][0][1] = y0; /* attr[0].y */ + verts[0][1][0] = sLeft; /* attr[1].s */ + verts[0][1][1] = tTop; /* attr[1].t */ + + /* upper-right */ + verts[1][0][0] = x1; + verts[1][0][1] = y0; + verts[1][1][0] = sRight; + verts[1][1][1] = tTop; + + /* lower-right */ + verts[2][0][0] = x1; + verts[2][0][1] = y1; + verts[2][1][0] = sRight; + verts[2][1][1] = tBot; + + /* lower-left */ + verts[3][0][0] = x0; + verts[3][0][1] = y1; + verts[3][1][0] = sLeft; + verts[3][1][1] = tBot; + + /* same for all verts: */ + for (i = 0; i < 4; i++) { + verts[i][0][2] = 0.0; /*Z*/ + verts[i][0][3] = 1.0; /*W*/ + verts[i][1][2] = 0.0; /*R*/ + verts[i][1][3] = 1.0; /*Q*/ + } + + st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, GL_TRUE); +} + + + +/** + * Generate mipmap levels using hardware rendering. + * \return TRUE if successful, FALSE if not possible + */ +static boolean +st_render_mipmap(struct st_context *st, + struct pipe_texture *pt, + uint baseLevel, uint lastLevel) +{ + struct pipe_context *pipe = st->pipe; + struct pipe_framebuffer_state fb; + const uint face = 0, zslice = 0; + const uint first_level_save = pt->first_level; + uint dstLevel; + + /* check if we can render in the texture's format */ + if (!pipe->is_format_supported(pipe, pt->format, PIPE_SURFACE)) { + return FALSE; + } + + /* init framebuffer state */ + memset(&fb, 0, sizeof(fb)); + fb.num_cbufs = 1; + + /* bind CSOs */ + pipe->bind_blend_state(pipe, blend_cso); + pipe->bind_depth_stencil_alpha_state(pipe, depthstencil_cso); + pipe->bind_rasterizer_state(pipe, rasterizer_cso); + pipe->bind_sampler_state(pipe, 0, sampler_cso); + + /* bind shaders */ + pipe->bind_fs_state(pipe, stfp->fs->data); + pipe->bind_vs_state(pipe, stvp->cso->data); + + for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { + const uint srcLevel = dstLevel - 1; + + /* + * Setup framebuffer / dest surface + */ + fb.cbufs[0] = pipe->get_tex_surface(pipe, pt, face, dstLevel, zslice); + pipe->set_framebuffer_state(pipe, &fb); + + simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]); + + /* + * Setup src texture, override pt->first_level so we sample from + * the right mipmap level. + */ + pt->first_level = srcLevel; + pipe->set_sampler_texture(pipe, 0, pt); + + draw_quad(st->ctx); + } + + /* restore first_level */ + pt->first_level = first_level_save; + + /* restore pipe state */ + if (st->state.rasterizer) + pipe->bind_rasterizer_state(pipe, st->state.rasterizer->data); + if (st->state.fs) + pipe->bind_fs_state(pipe, st->state.fs->data); + if (st->state.vs) + pipe->bind_vs_state(pipe, st->state.vs->cso->data); + if (st->state.sampler[0]) + pipe->bind_sampler_state(pipe, 0, st->state.sampler[0]->data); + pipe->set_sampler_texture(pipe, 0, st->state.sampler_texture[0]); + pipe->set_viewport_state(pipe, &st->state.viewport); + + return TRUE; +} + + + +void +st_generate_mipmap(GLcontext *ctx, GLenum target, + struct gl_texture_object *texObj) +{ + struct st_context *st = ctx->st; + struct pipe_texture *pt = st_get_texobj_texture(texObj); + const uint baseLevel = texObj->BaseLevel; + const uint lastLevel = pt->last_level; + uint dstLevel; + + if (!st_render_mipmap(st, pt, baseLevel, lastLevel)) { + abort(); + /* XXX the following won't really work at this time */ + _mesa_generate_mipmap(ctx, target, texObj); + return; + } + + for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { + const uint srcLevel = dstLevel - 1; + const struct gl_texture_image *srcImage + = _mesa_get_tex_image(ctx, texObj, target, srcLevel); + struct gl_texture_image *dstImage; + struct st_texture_image *stImage; + uint dstWidth = pt->width[dstLevel]; + uint dstHeight = pt->height[dstLevel]; + uint dstDepth = pt->depth[dstLevel]; + uint border = srcImage->Border; + + + dstImage = _mesa_get_tex_image(ctx, texObj, target, dstLevel); + if (!dstImage) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "generating mipmaps"); + return; + } + + if (dstImage->ImageOffsets) + _mesa_free(dstImage->ImageOffsets); + + /* Free old image data */ + if (dstImage->Data) + ctx->Driver.FreeTexImageData(ctx, dstImage); + + /* initialize new image */ + _mesa_init_teximage_fields(ctx, target, dstImage, dstWidth, dstHeight, + dstDepth, border, srcImage->InternalFormat); + + dstImage->TexFormat = srcImage->TexFormat; + + stImage = (struct st_texture_image *) dstImage; + stImage->pt = pt; + } + +} diff --git a/src/mesa/state_tracker/st_gen_mipmap.h b/src/mesa/state_tracker/st_gen_mipmap.h new file mode 100644 index 0000000000..7668c1e44e --- /dev/null +++ b/src/mesa/state_tracker/st_gen_mipmap.h @@ -0,0 +1,46 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_GEN_MIPMAP_H +#define ST_GEN_MIPMAP_H + + +extern void +st_init_generate_mipmap(struct st_context *st); + + +extern void +st_destroy_generate_mipmpap(struct st_context *st); + + +extern void +st_generate_mipmap(GLcontext *ctx, GLenum target, + struct gl_texture_object *texObj); + + +#endif /* ST_GEN_MIPMAP_H */ -- cgit v1.2.3 From 9f9c3b752a82d216d4655d6dcd07361f66f8a4c8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 14:55:33 -0700 Subject: fix comment typos --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 463ff0d9da..336ae1c8b6 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -2010,7 +2010,7 @@ exec_instruction( case TGSI_OPCODE_TXB: /* Texture lookup with lod bias */ - /* src[0] = texcoord (src[0].w = load bias) */ + /* src[0] = texcoord (src[0].w = LOD bias) */ /* src[1] = sampler unit */ exec_tex(mach, inst, TRUE); break; @@ -2026,7 +2026,7 @@ exec_instruction( case TGSI_OPCODE_TXL: /* Texture lookup with explit LOD */ - /* src[0] = texcoord (src[0].w = load bias) */ + /* src[0] = texcoord (src[0].w = LOD) */ /* src[1] = sampler unit */ exec_tex(mach, inst, TRUE); break; -- cgit v1.2.3 From 479b5e9b5d9e0e387332c6fbeaffffa7612a0c52 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 14:56:38 -0700 Subject: gallium: include st_cb_drawpixels.h --- src/mesa/state_tracker/st_gen_mipmap.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 16f9e4cd27..a6ac9a55fb 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -40,6 +40,7 @@ #include "st_draw.h" #include "st_gen_mipmap.h" #include "st_program.h" +#include "st_cb_drawpixels.h" #include "st_cb_texture.h" -- cgit v1.2.3 From f625c6d1d07e3b285cf966a888aa534f33c36710 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 25 Jan 2008 06:36:35 -0500 Subject: add a stub of a lowering pass --- src/mesa/pipe/llvm/Makefile | 1 + src/mesa/pipe/llvm/gallivm.cpp | 2 ++ src/mesa/pipe/llvm/loweringpass.cpp | 17 +++++++++++++++++ src/mesa/pipe/llvm/loweringpass.h | 15 +++++++++++++++ 4 files changed, 35 insertions(+) create mode 100644 src/mesa/pipe/llvm/loweringpass.cpp create mode 100644 src/mesa/pipe/llvm/loweringpass.h (limited to 'src') diff --git a/src/mesa/pipe/llvm/Makefile b/src/mesa/pipe/llvm/Makefile index f655fb8340..1e3ae988df 100644 --- a/src/mesa/pipe/llvm/Makefile +++ b/src/mesa/pipe/llvm/Makefile @@ -8,6 +8,7 @@ LIBNAME = gallivm GALLIVM_SOURCES = \ gallivm.cpp \ instructions.cpp \ + loweringpass.cpp \ storage.cpp INC_SOURCES = gallivm_builtins.cpp llvm_base_shader.cpp diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index afa1446890..46e11c185a 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -34,6 +34,7 @@ #include "gallivm.h" #include "instructions.h" +#include "loweringpass.h" #include "storage.h" #include "pipe/p_context.h" @@ -95,6 +96,7 @@ using namespace llvm; static int GLOBAL_ID = 0; static inline void AddStandardCompilePasses(PassManager &PM) { + PM.add(new LoweringPass()); PM.add(createVerifierPass()); // Verify that input is correct PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp diff --git a/src/mesa/pipe/llvm/loweringpass.cpp b/src/mesa/pipe/llvm/loweringpass.cpp new file mode 100644 index 0000000000..556dbec366 --- /dev/null +++ b/src/mesa/pipe/llvm/loweringpass.cpp @@ -0,0 +1,17 @@ +#include "loweringpass.h" + +using namespace llvm; + +char LoweringPass::ID = 0; +RegisterPass X("lowering", "Lowering Pass"); + +LoweringPass::LoweringPass() + : ModulePass((intptr_t)&ID) +{ +} + +bool LoweringPass::runOnModule(Module &m) +{ + llvm::cerr << "Hello: " << m.getModuleIdentifier() << "\n"; + return false; +} diff --git a/src/mesa/pipe/llvm/loweringpass.h b/src/mesa/pipe/llvm/loweringpass.h new file mode 100644 index 0000000000..f62dcf6ba7 --- /dev/null +++ b/src/mesa/pipe/llvm/loweringpass.h @@ -0,0 +1,15 @@ +#ifndef LOWERINGPASS_H +#define LOWERINGPASS_H + +#include "llvm/Pass.h" +#include "llvm/Module.h" + +struct LoweringPass : public llvm::ModulePass +{ + static char ID; + LoweringPass(); + + virtual bool runOnModule(llvm::Module &m); +}; + +#endif -- cgit v1.2.3 From 7d69090e272d0d429f0ef7a733cebe1363383447 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 4 Feb 2008 10:07:02 -0500 Subject: rewrite the way we handle ir in llvm code introduce intermediate step gallivm_ir before compiling it down to the final llvm ir. --- src/mesa/pipe/cell/ppu/cell_state_fs.c | 2 +- src/mesa/pipe/draw/draw_vertex_shader.c | 27 +- src/mesa/pipe/llvm/Makefile | 1 + src/mesa/pipe/llvm/gallivm.cpp | 791 +++----------------------------- src/mesa/pipe/llvm/gallivm.h | 23 +- src/mesa/pipe/llvm/gallivm_p.h | 56 +++ src/mesa/pipe/llvm/llvm_base_shader.cpp | 435 ++++++++---------- src/mesa/pipe/llvm/llvm_entry.c | 14 +- src/mesa/pipe/llvm/tgsitollvm.cpp | 682 +++++++++++++++++++++++++++ src/mesa/pipe/llvm/tgsitollvm.h | 16 + src/mesa/pipe/softpipe/sp_state_fs.c | 2 +- 11 files changed, 1079 insertions(+), 970 deletions(-) create mode 100644 src/mesa/pipe/llvm/gallivm_p.h create mode 100644 src/mesa/pipe/llvm/tgsitollvm.cpp create mode 100644 src/mesa/pipe/llvm/tgsitollvm.h (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_state_fs.c b/src/mesa/pipe/cell/ppu/cell_state_fs.c index 96a52273b0..3f46a87d18 100644 --- a/src/mesa/pipe/cell/ppu/cell_state_fs.c +++ b/src/mesa/pipe/cell/ppu/cell_state_fs.c @@ -67,7 +67,7 @@ cell_create_fs_state(struct pipe_context *pipe, #endif #ifdef MESA_LLVM - state->llvm_prog = gallivm_from_tgsi(state->shader.tokens, GALLIVM_FS); + state->llvm_prog = 0; if (!gallivm_global_cpu_engine()) { gallivm_cpu_engine_create(state->llvm_prog); } diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 5ca93aa615..9567283ff5 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -113,7 +113,16 @@ run_vertex_program(struct draw_context *draw, draw->vertex_fetch.fetch_func( draw, machine, elts, count ); /* run shader */ -#if defined(__i386__) || defined(__386__) +#ifdef MESA_LLVM + if (1) { + struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; + gallivm_prog_exec(prog, + machine->Inputs, + machine->Outputs, + machine->Consts, + 12, 12, 12); + } else +#elif defined(__i386__) || defined(__386__) if (draw->use_sse) { /* SSE */ /* cast away const */ @@ -212,13 +221,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) */ draw_update_vertex_fetch( draw ); -// debug_printf( " q(%d) ", draw->vs.queue_nr ); -#ifdef MESA_LLVM - if (draw->vertex_shader->llvm_prog) { - draw_vertex_shader_queue_flush_llvm(draw); - return; - } -#endif +// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); /* run vertex shader on vertex cache entries, four per invokation */ for (i = 0; i < draw->vs.queue_nr; i += 4) { @@ -260,7 +263,13 @@ draw_create_vertex_shader(struct draw_context *draw, vs->state = shader; #ifdef MESA_LLVM - vs->llvm_prog = gallivm_from_tgsi(shader->tokens, GALLIVM_VS); + struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); + gallivm_ir_set_layout(ir, GALLIVM_SOA); + gallivm_ir_set_components(ir, 4); + gallivm_ir_fill_from_tgsi(ir, shader->tokens); + vs->llvm_prog = gallivm_ir_compile(ir); + gallivm_ir_delete(ir); + draw->engine = gallivm_global_cpu_engine(); if (!draw->engine) { draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); diff --git a/src/mesa/pipe/llvm/Makefile b/src/mesa/pipe/llvm/Makefile index 1e3ae988df..10ff7aacae 100644 --- a/src/mesa/pipe/llvm/Makefile +++ b/src/mesa/pipe/llvm/Makefile @@ -9,6 +9,7 @@ GALLIVM_SOURCES = \ gallivm.cpp \ instructions.cpp \ loweringpass.cpp \ + tgsitollvm.cpp \ storage.cpp INC_SOURCES = gallivm_builtins.cpp llvm_base_shader.cpp diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index 46e11c185a..cf9b0f6406 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -32,17 +32,17 @@ #ifdef MESA_LLVM #include "gallivm.h" +#include "gallivm_p.h" #include "instructions.h" #include "loweringpass.h" #include "storage.h" +#include "tgsitollvm.h" #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" + #include "pipe/tgsi/exec/tgsi_exec.h" -#include "pipe/tgsi/util/tgsi_util.h" -#include "pipe/tgsi/util/tgsi_build.h" #include "pipe/tgsi/util/tgsi_dump.h" #include @@ -64,38 +64,23 @@ #include #include #include +#include #include #include #include -struct gallivm_interpolate { - int attrib; - int chan; - int type; -}; - -struct gallivm_prog { - llvm::Module *module; - void *function; - int num_consts; - int id; - enum gallivm_shader_type type; - - struct gallivm_interpolate interpolators[32*4]; //FIXME: this might not be enough for some shaders - int num_interp; -}; - struct gallivm_cpu_engine { llvm::ExecutionEngine *engine; }; -using namespace llvm; -#include "llvm_base_shader.cpp" - static int GLOBAL_ID = 0; -static inline void AddStandardCompilePasses(PassManager &PM) { +using namespace llvm; + +static inline +void AddStandardCompilePasses(PassManager &PM) +{ PM.add(new LoweringPass()); PM.add(createVerifierPass()); // Verify that input is correct @@ -152,691 +137,16 @@ static inline void AddStandardCompilePasses(PassManager &PM) { PM.add(createConstantMergePass()); // Merge dup global constants } -static inline void -add_interpolator(struct gallivm_prog *prog, - struct gallivm_interpolate *interp) -{ - prog->interpolators[prog->num_interp] = *interp; - ++prog->num_interp; -} - -static void -translate_declaration(struct gallivm_prog *prog, - llvm::Module *module, - Storage *storage, - struct tgsi_full_declaration *decl, - struct tgsi_full_declaration *fd) -{ - if (decl->Declaration.File == TGSI_FILE_INPUT) { - unsigned first, last, mask; - uint interp_method; - - assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - /* Do not touch WPOS.xy */ - if (first == 0) { - mask &= ~TGSI_WRITEMASK_XY; - if (mask == TGSI_WRITEMASK_NONE) { - first++; - if (first > last) { - return; - } - } - } - - interp_method = decl->Interpolation.Interpolate; - - if (mask == TGSI_WRITEMASK_XYZW) { - unsigned i, j; - - for (i = first; i <= last; i++) { - for (j = 0; j < NUM_CHANNELS; j++) { - //interp( mach, i, j ); - struct gallivm_interpolate interp; - interp.type = interp_method; - interp.attrib = i; - interp.chan = j; - add_interpolator(prog, &interp); - } - } - } else { - unsigned i, j; - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - struct gallivm_interpolate interp; - interp.type = interp_method; - interp.attrib = i; - interp.chan = j; - add_interpolator(prog, &interp); - } - } - } - } - } -} - - -static void -translate_immediate(Storage *storage, - struct tgsi_full_immediate *imm) -{ - float vec[4]; - int i; - for (i = 0; i < imm->Immediate.Size - 1; ++i) { - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; - break; - default: - assert( 0 ); - } - } - storage->addImmediate(vec); -} - -static inline llvm::Value * -swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, - Storage *storage) -{ - int swizzle = 0; - int start = 1000; - const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + - TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; - for (int k = 0; k < 4; ++k) { - swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; - start /= 10; - } - if (swizzle != NO_SWIZZLE) { - /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ - val = storage->shuffleVector(val, swizzle); - } - return val; -} - -static void -translate_instruction(llvm::Module *module, - Storage *storage, - Instructions *instr, - struct tgsi_full_instruction *inst, - struct tgsi_full_instruction *fi, - unsigned instno) -{ - llvm::Value *inputs[4]; - inputs[0] = 0; - inputs[1] = 0; - inputs[2] = 0; - inputs[3] = 0; - - for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - llvm::Value *val = 0; - llvm::Value *indIdx = 0; - - if (src->SrcRegister.Indirect) { - indIdx = storage->addrElement(src->SrcRegisterInd.Index); - indIdx = storage->extractIndex(indIdx); - } - if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->constElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->inputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->tempElement(src->SrcRegister.Index); - } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->outputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->immediateElement(src->SrcRegister.Index); - } else { - fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); - return; - } - - inputs[i] = swizzleVector(val, src, storage); - } - - /*if (inputs[0]) - instr->printVector(inputs[0]); - if (inputs[1]) - instr->printVector(inputs[1]);*/ - llvm::Value *out = 0; - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: { - out = instr->arl(inputs[0]); - } - break; - case TGSI_OPCODE_MOV: { - out = inputs[0]; - } - break; - case TGSI_OPCODE_LIT: { - out = instr->lit(inputs[0]); - } - break; - case TGSI_OPCODE_RCP: { - out = instr->rcp(inputs[0]); - } - break; - case TGSI_OPCODE_RSQ: { - out = instr->rsq(inputs[0]); - } - break; - case TGSI_OPCODE_EXP: - break; - case TGSI_OPCODE_LOG: - break; - case TGSI_OPCODE_MUL: { - out = instr->mul(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_ADD: { - out = instr->add(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP3: { - out = instr->dp3(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP4: { - out = instr->dp4(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DST: { - out = instr->dst(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MIN: { - out = instr->min(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAX: { - out = instr->max(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SLT: { - out = instr->slt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SGE: { - out = instr->sge(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAD: { - out = instr->madd(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SUB: { - out = instr->sub(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_LERP: { - out = instr->lerp(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_CND: - break; - case TGSI_OPCODE_CND0: - break; - case TGSI_OPCODE_DOT2ADD: - break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { - out = instr->frc(inputs[0]); - } - break; - case TGSI_OPCODE_CLAMP: - break; - case TGSI_OPCODE_FLOOR: { - out = instr->floor(inputs[0]); - } - break; - case TGSI_OPCODE_ROUND: - break; - case TGSI_OPCODE_EXPBASE2: { - out = instr->ex2(inputs[0]); - } - break; - case TGSI_OPCODE_LOGBASE2: { - out = instr->lg2(inputs[0]); - } - break; - case TGSI_OPCODE_POWER: { - out = instr->pow(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_CROSSPRODUCT: { - out = instr->cross(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; - case TGSI_OPCODE_ABS: { - out = instr->abs(inputs[0]); - } - break; - case TGSI_OPCODE_RCC: - break; - case TGSI_OPCODE_DPH: { - out = instr->dph(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_COS: { - out = instr->cos(inputs[0]); - } - break; - case TGSI_OPCODE_DDX: - break; - case TGSI_OPCODE_DDY: - break; - case TGSI_OPCODE_KILP: { - out = instr->kilp(inputs[0]); - storage->setKilElement(out); - return; - } - break; - case TGSI_OPCODE_PK2H: - break; - case TGSI_OPCODE_PK2US: - break; - case TGSI_OPCODE_PK4B: - break; - case TGSI_OPCODE_PK4UB: - break; - case TGSI_OPCODE_RFL: - break; - case TGSI_OPCODE_SEQ: - break; - case TGSI_OPCODE_SFL: - break; - case TGSI_OPCODE_SGT: { - out = instr->sgt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SIN: { - out = instr->sin(inputs[0]); - } - break; - case TGSI_OPCODE_SLE: - break; - case TGSI_OPCODE_SNE: - break; - case TGSI_OPCODE_STR: - break; - case TGSI_OPCODE_TEX: - break; - case TGSI_OPCODE_TXD: - break; - case TGSI_OPCODE_UP2H: - break; - case TGSI_OPCODE_UP2US: - break; - case TGSI_OPCODE_UP4B: - break; - case TGSI_OPCODE_UP4UB: - break; - case TGSI_OPCODE_X2D: - break; - case TGSI_OPCODE_ARA: - break; - case TGSI_OPCODE_ARR: - break; - case TGSI_OPCODE_BRA: - break; - case TGSI_OPCODE_CAL: { - instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr()); - return; - } - break; - case TGSI_OPCODE_RET: { - instr->end(); - return; - } - break; - case TGSI_OPCODE_SSG: - break; - case TGSI_OPCODE_CMP: { - out = instr->cmp(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SCS: { - out = instr->scs(inputs[0]); - } - break; - case TGSI_OPCODE_TXB: - break; - case TGSI_OPCODE_NRM: - break; - case TGSI_OPCODE_DIV: - break; - case TGSI_OPCODE_DP2: - break; - case TGSI_OPCODE_TXL: - break; - case TGSI_OPCODE_BRK: { - instr->brk(); - return; - } - break; - case TGSI_OPCODE_IF: { - instr->ifop(inputs[0]); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_LOOP: - break; - case TGSI_OPCODE_REP: - break; - case TGSI_OPCODE_ELSE: { - instr->elseop(); - storage->setCurrentBlock(instr->currentBlock()); - return; //only state update - } - break; - case TGSI_OPCODE_ENDIF: { - instr->endif(); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_ENDLOOP: - break; - case TGSI_OPCODE_ENDREP: - break; - case TGSI_OPCODE_PUSHA: - break; - case TGSI_OPCODE_POPA: - break; - case TGSI_OPCODE_CEIL: - break; - case TGSI_OPCODE_I2F: - break; - case TGSI_OPCODE_NOT: - break; - case TGSI_OPCODE_TRUNC: { - out = instr->trunc(inputs[0]); - } - break; - case TGSI_OPCODE_SHL: - break; - case TGSI_OPCODE_SHR: - break; - case TGSI_OPCODE_AND: - break; - case TGSI_OPCODE_OR: - break; - case TGSI_OPCODE_MOD: - break; - case TGSI_OPCODE_XOR: - break; - case TGSI_OPCODE_SAD: - break; - case TGSI_OPCODE_TXF: - break; - case TGSI_OPCODE_TXQ: - break; - case TGSI_OPCODE_CONT: - break; - case TGSI_OPCODE_EMIT: - break; - case TGSI_OPCODE_ENDPRIM: - break; - case TGSI_OPCODE_BGNLOOP2: { - instr->beginLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_BGNSUB: { - instr->bgnSub(instno); - storage->setCurrentBlock(instr->currentBlock()); - storage->pushTemps(); - return; - } - break; - case TGSI_OPCODE_ENDLOOP2: { - instr->endLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_ENDSUB: { - instr->endSub(); - storage->setCurrentBlock(instr->currentBlock()); - storage->popArguments(); - storage->popTemps(); - return; - } - break; - case TGSI_OPCODE_NOISE1: - break; - case TGSI_OPCODE_NOISE2: - break; - case TGSI_OPCODE_NOISE3: - break; - case TGSI_OPCODE_NOISE4: - break; - case TGSI_OPCODE_NOP: - break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; - case TGSI_OPCODE_NRM4: - break; - case TGSI_OPCODE_CALLNZ: - break; - case TGSI_OPCODE_IFC: - break; - case TGSI_OPCODE_BREAKC: - break; - case TGSI_OPCODE_KIL: - break; - case TGSI_OPCODE_END: - instr->end(); - return; - break; - default: - fprintf(stderr, "ERROR: Unknown opcode %d\n", - inst->Instruction.Opcode); - assert(0); - break; - } - - if (!out) { - fprintf(stderr, "ERROR: unsupported opcode %d\n", - inst->Instruction.Opcode); - assert(!"Unsupported opcode"); - } - - /* # not sure if we need this */ - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - /*TXT( "_SAT" );*/ - break; - case TGSI_SAT_MINUS_PLUS_ONE: - /*TXT( "_SAT[-1,1]" );*/ - break; - default: - assert( 0 ); - } - - /* store results */ - for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else { - fprintf(stderr, "ERROR: unsupported LLVM destination!"); - assert(!"wrong destination"); - } - } -} - -static llvm::Module * -tgsi_to_llvm(struct gallivm_prog *prog, const struct tgsi_token *tokens) -{ - llvm::Module *mod = createBaseShader(); - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned instno = 0; - Function* shader = mod->getFunction("execute_shader"); - std::ostringstream stream; - if (prog->type == GALLIVM_VS) { - stream << "vs_shader"; - } else { - stream << "fs_shader"; - } - stream << prog->id; - std::string func_name = stream.str(); - shader->setName(func_name.c_str()); - - Function::arg_iterator args = shader->arg_begin(); - Value *ptr_INPUT = args++; - ptr_INPUT->setName("input"); - - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); - - tgsi_parse_init(&parse, tokens); - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - Storage storage(label_entry, ptr_INPUT); - Instructions instr(mod, shader, label_entry, &storage); - while(!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - translate_declaration(prog, mod, &storage, - &parse.FullToken.FullDeclaration, - &fd); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - translate_immediate(&storage, - &parse.FullToken.FullImmediate); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - translate_instruction(mod, &storage, &instr, - &parse.FullToken.FullInstruction, - &fi, instno); - ++instno; - break; - - default: - assert(0); - } - } - - tgsi_parse_free(&parse); - - prog->num_consts = storage.numConsts(); - return mod; -} - -/*! - Translates the TGSI tokens into LLVM format. Translated representation - is stored in the gallivm_prog and returned. - After calling this function the gallivm_prog can either be used with a custom - code generator to generate machine code for the GPU which the code generator - addresses or it can be jit compiled with gallivm_cpu_jit_compile and executed - with gallivm_prog_exec to run the module on the CPU. - */ -struct gallivm_prog * -gallivm_from_tgsi(const struct tgsi_token *tokens, enum gallivm_shader_type type) -{ - std::cout << "Creating llvm from: " <id = GLOBAL_ID; - gallivm->type = type; - tgsi_dump(tokens, 0); - - llvm::Module *mod = tgsi_to_llvm(gallivm, tokens); - gallivm->module = mod; - gallivm_prog_dump(gallivm, 0); - - /* Run optimization passes over it */ - PassManager passes; - passes.add(new TargetData(mod)); - AddStandardCompilePasses(passes); - passes.run(*mod); - - gallivm->module = mod; - - gallivm_prog_dump(gallivm, 0); - - return gallivm; -} - - void gallivm_prog_delete(struct gallivm_prog *prog) { - llvm::Module *mod = static_cast(prog->module); - delete mod; + delete prog->module; prog->module = 0; prog->function = 0; free(prog); } -typedef void (*vertex_shader_runner)(float (*ainputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], +typedef void (*vertex_shader_runner)(void *ainputs, + void *dests, float (*aconsts)[4], int num_vertices, int num_inputs, @@ -850,8 +160,8 @@ typedef void (*vertex_shader_runner)(float (*ainputs)[PIPE_MAX_SHADER_INPUTS][4] function. */ int gallivm_prog_exec(struct gallivm_prog *prog, - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, float (*consts)[4], int num_vertices, int num_inputs, @@ -943,18 +253,15 @@ int gallivm_fragment_shader_exec(struct gallivm_prog *prog, samplers); } -void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) +void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix) { - llvm::Module *mod; - if (!prog || !prog->module) + if (!ir || !ir->module) return; - mod = static_cast(prog->module); - if (file_prefix) { std::ostringstream stream; stream << file_prefix; - stream << prog->id; + stream << ir->id; stream << ".ll"; std::string name = stream.str(); std::ofstream out(name.c_str()); @@ -962,12 +269,12 @@ void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix) std::cerr<<"Can't open file : "<module); out.close(); } else { - const llvm::Module::FunctionListType &funcs = mod->getFunctionList(); + const llvm::Module::FunctionListType &funcs = ir->module->getFunctionList(); llvm::Module::FunctionListType::const_iterator itr; - std::cout<<"; ---------- Start shader "<id<id<id<id<id = GLOBAL_ID; + ir->type = type; + return ir; +} +void gallivm_ir_set_layout(struct gallivm_ir *ir, + enum gallivm_vector_layout layout) +{ + ir->layout = layout; +} +void gallivm_ir_set_components(struct gallivm_ir *ir, int num) +{ + ir->num_components = num; +} +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, + const struct tgsi_token *tokens) +{ + std::cout << "Creating llvm from: " <module = mod; + gallivm_ir_dump(ir, 0); +} + +void gallivm_ir_delete(struct gallivm_ir *ir) +{ + delete ir->module; + free(ir); +} + +struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) +{ + struct gallivm_prog *prog = + (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); + llvm::Module *mod = llvm::CloneModule(ir->module); + prog->num_consts = ir->num_consts; + memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators)); + prog->num_interp = ir->num_interp; + + /* Run optimization passes over it */ + PassManager passes; + passes.add(new TargetData(mod)); + AddStandardCompilePasses(passes); + passes.run(*mod); + prog->module = mod; + + return prog; +} + +#endif /* MESA_LLVM */ diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h index 4695de3127..f9f5d5ee74 100644 --- a/src/mesa/pipe/llvm/gallivm.h +++ b/src/mesa/pipe/llvm/gallivm.h @@ -43,21 +43,36 @@ extern "C" { struct tgsi_token; +struct gallivm_ir; struct gallivm_prog; struct gallivm_cpu_engine; struct tgsi_interp_coef; struct tgsi_sampler; +struct tgsi_exec_vector; enum gallivm_shader_type { GALLIVM_VS, GALLIVM_FS }; -struct gallivm_prog *gallivm_from_tgsi(const struct tgsi_token *tokens, enum gallivm_shader_type type); -void gallivm_prog_delete(struct gallivm_prog *prog); +enum gallivm_vector_layout { + GALLIVM_AOS, + GALLIVM_SOA +}; + +struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type); +void gallivm_ir_set_layout(struct gallivm_ir *prog, + enum gallivm_vector_layout layout); +void gallivm_ir_set_components(struct gallivm_ir *prog, int num); +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *prog, + const struct tgsi_token *tokens); +void gallivm_ir_delete(struct gallivm_ir *prog); + +struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir); + int gallivm_prog_exec(struct gallivm_prog *prog, - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, float (*consts)[4], int num_vertices, int num_inputs, diff --git a/src/mesa/pipe/llvm/gallivm_p.h b/src/mesa/pipe/llvm/gallivm_p.h new file mode 100644 index 0000000000..2c6e5e8f5f --- /dev/null +++ b/src/mesa/pipe/llvm/gallivm_p.h @@ -0,0 +1,56 @@ +#ifndef GALLIVM_P_H +#define GALLIVM_P_H + +#ifdef MESA_LLVM + +namespace llvm { + class Module; +} + +#if defined __cplusplus +extern "C" { +#endif + +enum gallivm_shader_type; +enum gallivm_vector_layout; + +struct gallivm_interpolate { + int attrib; + int chan; + int type; +}; + +struct gallivm_ir { + llvm::Module *module; + int id; + enum gallivm_shader_type type; + enum gallivm_vector_layout layout; + int num_components; + int num_consts; + + //FIXME: this might not be enough for some shaders + struct gallivm_interpolate interpolators[32*4]; + int num_interp; +}; + +struct gallivm_prog { + llvm::Module *module; + void *function; + + int id; + enum gallivm_shader_type type; + + int num_consts; + + //FIXME: this might not be enough for some shaders + struct gallivm_interpolate interpolators[32*4]; + int num_interp; +}; + +#endif /* MESA_LLVM */ + +#if defined __cplusplus +} // extern "C" +#endif + +#endif diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp index b574b550ae..90a25a440a 100644 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ b/src/mesa/pipe/llvm/llvm_base_shader.cpp @@ -104,8 +104,8 @@ Module* createBaseShader() { /*isVarArg=*/false); std::vectorFuncTy_18_args; - FuncTy_18_args.push_back(PointerTy_12); - FuncTy_18_args.push_back(PointerTy_12); + FuncTy_18_args.push_back(PointerTy_9); + FuncTy_18_args.push_back(PointerTy_9); FuncTy_18_args.push_back(PointerTy_7); FuncTy_18_args.push_back(IntegerType::get(32)); FuncTy_18_args.push_back(IntegerType::get(32)); @@ -526,14 +526,23 @@ Module* createBaseShader() { BasicBlock* label_forbody_preheader_i = new BasicBlock("forbody.preheader.i",func_run_vertex_shader,0); BasicBlock* label_forbody_i = new BasicBlock("forbody.i",func_run_vertex_shader,0); BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_vertex_shader,0); - BasicBlock* label_forbody_preheader_91 = new BasicBlock("forbody.preheader",func_run_vertex_shader,0); - BasicBlock* label_forbody_92 = new BasicBlock("forbody",func_run_vertex_shader,0); - BasicBlock* label_afterfor_93 = new BasicBlock("afterfor",func_run_vertex_shader,0); // Block entry (label_entry_90) AllocaInst* ptr_consts = new AllocaInst(ArrayTy_20, "consts", label_entry_90); AllocaInst* ptr_temps = new AllocaInst(ArrayTy_22, "temps", label_entry_90); AllocaInst* ptr_args = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_90); + std::vector ptr_tmp_indices; + ptr_tmp_indices.push_back(const_int32_29); + ptr_tmp_indices.push_back(const_int32_29); + Instruction* ptr_tmp = new GetElementPtrInst(ptr_args, ptr_tmp_indices.begin(), ptr_tmp_indices.end(), "tmp", label_entry_90); + CastInst* ptr_conv = new BitCastInst(ptr_results, PointerTy_0, "conv", label_entry_90); + StoreInst* void_91 = new StoreInst(ptr_conv, ptr_tmp, false, label_entry_90); + std::vector ptr_tmp2_indices; + ptr_tmp2_indices.push_back(const_int32_29); + ptr_tmp2_indices.push_back(const_int32_31); + Instruction* ptr_tmp2 = new GetElementPtrInst(ptr_args, ptr_tmp2_indices.begin(), ptr_tmp2_indices.end(), "tmp2", label_entry_90); + CastInst* ptr_conv4 = new BitCastInst(ptr_inputs, PointerTy_0, "conv4", label_entry_90); + StoreInst* void_92 = new StoreInst(ptr_conv4, ptr_tmp2, false, label_entry_90); ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_29, "cmp.i", label_entry_90); new BranchInst(label_forbody_preheader_i, label_from_consts_exit, int1_cmp_i, label_entry_90); @@ -544,17 +553,17 @@ Module* createBaseShader() { new BranchInst(label_forbody_i, label_forbody_preheader_i); // Block forbody.i (label_forbody_i) - Argument* fwdref_96 = new Argument(IntegerType::get(32)); + Argument* fwdref_95 = new Argument(IntegerType::get(32)); PHINode* int32_i_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i); int32_i_0_reg2mem_0_i->reserveOperandSpace(2); int32_i_0_reg2mem_0_i->addIncoming(const_int32_29, label_forbody_preheader_i); - int32_i_0_reg2mem_0_i->addIncoming(fwdref_96, label_forbody_i); + int32_i_0_reg2mem_0_i->addIncoming(fwdref_95, label_forbody_i); - Argument* fwdref_97 = new Argument(VectorTy_1); + Argument* fwdref_96 = new Argument(VectorTy_1); PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i); packed_vec_0_reg2mem_0_i->reserveOperandSpace(2); packed_vec_0_reg2mem_0_i->addIncoming(const_packed_32, label_forbody_preheader_i); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_97, label_forbody_i); + packed_vec_0_reg2mem_0_i->addIncoming(fwdref_96, label_forbody_i); std::vector ptr_arraydecay_i_indices; ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_i); @@ -584,80 +593,40 @@ Module* createBaseShader() { ptr_arrayidx34_i_indices.push_back(const_int32_29); ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i); Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody_i); - StoreInst* void_98 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i, false, label_forbody_i); - BinaryOperator* int32_indvar_next8 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i, const_int32_31, "indvar.next8", label_forbody_i); - ICmpInst* int1_exitcond9 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next8, int32_tmp10_i, "exitcond9", label_forbody_i); - new BranchInst(label_from_consts_exit, label_forbody_i, int1_exitcond9, label_forbody_i); + StoreInst* void_97 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i, false, label_forbody_i); + BinaryOperator* int32_indvar_next_98 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i, const_int32_31, "indvar.next", label_forbody_i); + ICmpInst* int1_exitcond_99 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_98, int32_tmp10_i, "exitcond", label_forbody_i); + new BranchInst(label_from_consts_exit, label_forbody_i, int1_exitcond_99, label_forbody_i); // Block from_consts.exit (label_from_consts_exit) - std::vector ptr_tmp2_indices; - ptr_tmp2_indices.push_back(const_int32_29); - ptr_tmp2_indices.push_back(const_int32_34); - Instruction* ptr_tmp2 = new GetElementPtrInst(ptr_args, ptr_tmp2_indices.begin(), ptr_tmp2_indices.end(), "tmp2", label_from_consts_exit); - std::vector ptr_arraydecay3_indices; - ptr_arraydecay3_indices.push_back(const_int32_29); - ptr_arraydecay3_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay3 = new GetElementPtrInst(ptr_consts, ptr_arraydecay3_indices.begin(), ptr_arraydecay3_indices.end(), "arraydecay3", label_from_consts_exit); - StoreInst* void_100 = new StoreInst(ptr_arraydecay3, ptr_tmp2, false, label_from_consts_exit); - std::vector ptr_tmp4_indices; - ptr_tmp4_indices.push_back(const_int32_29); - ptr_tmp4_indices.push_back(const_int32_33); - Instruction* ptr_tmp4 = new GetElementPtrInst(ptr_args, ptr_tmp4_indices.begin(), ptr_tmp4_indices.end(), "tmp4", label_from_consts_exit); - std::vector ptr_arraydecay5_indices; - ptr_arraydecay5_indices.push_back(const_int32_29); - ptr_arraydecay5_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay5 = new GetElementPtrInst(ptr_temps, ptr_arraydecay5_indices.begin(), ptr_arraydecay5_indices.end(), "arraydecay5", label_from_consts_exit); - StoreInst* void_101 = new StoreInst(ptr_arraydecay5, ptr_tmp4, false, label_from_consts_exit); - ICmpInst* int1_cmp_102 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_vertices, const_int32_29, "cmp", label_from_consts_exit); - new BranchInst(label_forbody_preheader_91, label_afterfor_93, int1_cmp_102, label_from_consts_exit); - - // Block forbody.preheader (label_forbody_preheader_91) - std::vector ptr_tmp8_indices; - ptr_tmp8_indices.push_back(const_int32_29); - ptr_tmp8_indices.push_back(const_int32_29); - Instruction* ptr_tmp8 = new GetElementPtrInst(ptr_args, ptr_tmp8_indices.begin(), ptr_tmp8_indices.end(), "tmp8", label_forbody_preheader_91); - std::vector ptr_tmp12_indices; - ptr_tmp12_indices.push_back(const_int32_29); - ptr_tmp12_indices.push_back(const_int32_31); - Instruction* ptr_tmp12 = new GetElementPtrInst(ptr_args, ptr_tmp12_indices.begin(), ptr_tmp12_indices.end(), "tmp12", label_forbody_preheader_91); - BinaryOperator* int32_tmp_104 = BinaryOperator::create(Instruction::Add, int32_num_vertices, const_int32_30, "tmp", label_forbody_preheader_91); - ICmpInst* int1_tmp6 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_104, const_int32_29, "tmp6", label_forbody_preheader_91); - SelectInst* int32_tmp7 = new SelectInst(int1_tmp6, const_int32_31, int32_num_vertices, "tmp7", label_forbody_preheader_91); - new BranchInst(label_forbody_92, label_forbody_preheader_91); - - // Block forbody (label_forbody_92) - Argument* fwdref_107 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_106 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_92); - int32_i_0_reg2mem_0_106->reserveOperandSpace(2); - int32_i_0_reg2mem_0_106->addIncoming(const_int32_29, label_forbody_preheader_91); - int32_i_0_reg2mem_0_106->addIncoming(fwdref_107, label_forbody_92); - - std::vector ptr_arraydecay11_108_indices; - ptr_arraydecay11_108_indices.push_back(int32_i_0_reg2mem_0_106); - ptr_arraydecay11_108_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11_108 = new GetElementPtrInst(ptr_results, ptr_arraydecay11_108_indices.begin(), ptr_arraydecay11_108_indices.end(), "arraydecay11", label_forbody_92); - StoreInst* void_109 = new StoreInst(ptr_arraydecay11_108, ptr_tmp8, false, label_forbody_92); - std::vector ptr_arraydecay16_indices; - ptr_arraydecay16_indices.push_back(int32_i_0_reg2mem_0_106); - ptr_arraydecay16_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay16 = new GetElementPtrInst(ptr_inputs, ptr_arraydecay16_indices.begin(), ptr_arraydecay16_indices.end(), "arraydecay16", label_forbody_92); - StoreInst* void_110 = new StoreInst(ptr_arraydecay16, ptr_tmp12, false, label_forbody_92); - CallInst* void_111 = new CallInst(func_execute_shader, ptr_args, "", label_forbody_92); - void_111->setCallingConv(CallingConv::C); - void_111->setTailCall(false);const ParamAttrsList *void_111_PAL = 0; - void_111->setParamAttrs(void_111_PAL); - - BinaryOperator* int32_indvar_next_112 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_106, const_int32_31, "indvar.next", label_forbody_92); - ICmpInst* int1_exitcond_113 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_112, int32_tmp7, "exitcond", label_forbody_92); - new BranchInst(label_afterfor_93, label_forbody_92, int1_exitcond_113, label_forbody_92); - - // Block afterfor (label_afterfor_93) - new ReturnInst(label_afterfor_93); + std::vector ptr_tmp7_indices; + ptr_tmp7_indices.push_back(const_int32_29); + ptr_tmp7_indices.push_back(const_int32_34); + Instruction* ptr_tmp7 = new GetElementPtrInst(ptr_args, ptr_tmp7_indices.begin(), ptr_tmp7_indices.end(), "tmp7", label_from_consts_exit); + std::vector ptr_arraydecay8_indices; + ptr_arraydecay8_indices.push_back(const_int32_29); + ptr_arraydecay8_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay8 = new GetElementPtrInst(ptr_consts, ptr_arraydecay8_indices.begin(), ptr_arraydecay8_indices.end(), "arraydecay8", label_from_consts_exit); + StoreInst* void_101 = new StoreInst(ptr_arraydecay8, ptr_tmp7, false, label_from_consts_exit); + std::vector ptr_tmp9_indices; + ptr_tmp9_indices.push_back(const_int32_29); + ptr_tmp9_indices.push_back(const_int32_33); + Instruction* ptr_tmp9 = new GetElementPtrInst(ptr_args, ptr_tmp9_indices.begin(), ptr_tmp9_indices.end(), "tmp9", label_from_consts_exit); + std::vector ptr_arraydecay10_indices; + ptr_arraydecay10_indices.push_back(const_int32_29); + ptr_arraydecay10_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_temps, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_from_consts_exit); + StoreInst* void_102 = new StoreInst(ptr_arraydecay10, ptr_tmp9, false, label_from_consts_exit); + CallInst* void_103 = new CallInst(func_execute_shader, ptr_args, "", label_from_consts_exit); + void_103->setCallingConv(CallingConv::C); + void_103->setTailCall(false);const ParamAttrsList *void_103_PAL = 0; + void_103->setParamAttrs(void_103_PAL); + + new ReturnInst(label_from_consts_exit); // Resolve Forward References - fwdref_107->replaceAllUsesWith(int32_indvar_next_112); delete fwdref_107; - fwdref_97->replaceAllUsesWith(packed_tmp31_i); delete fwdref_97; - fwdref_96->replaceAllUsesWith(int32_indvar_next8); delete fwdref_96; + fwdref_96->replaceAllUsesWith(packed_tmp31_i); delete fwdref_96; + fwdref_95->replaceAllUsesWith(int32_indvar_next_98); delete fwdref_95; } @@ -668,195 +637,195 @@ Module* createBaseShader() { float_x->setName("x"); Value* float_y = args++; float_y->setName("y"); - Value* ptr_results_116 = args++; - ptr_results_116->setName("results"); - Value* ptr_inputs_117 = args++; - ptr_inputs_117->setName("inputs"); - Value* int32_num_inputs_118 = args++; - int32_num_inputs_118->setName("num_inputs"); - Value* ptr_aconsts_119 = args++; - ptr_aconsts_119->setName("aconsts"); - Value* int32_num_consts_120 = args++; - int32_num_consts_120->setName("num_consts"); + Value* ptr_results_105 = args++; + ptr_results_105->setName("results"); + Value* ptr_inputs_106 = args++; + ptr_inputs_106->setName("inputs"); + Value* int32_num_inputs_107 = args++; + int32_num_inputs_107->setName("num_inputs"); + Value* ptr_aconsts_108 = args++; + ptr_aconsts_108->setName("aconsts"); + Value* int32_num_consts_109 = args++; + int32_num_consts_109->setName("num_consts"); Value* ptr_samplers = args++; ptr_samplers->setName("samplers"); - BasicBlock* label_entry_121 = new BasicBlock("entry",func_run_fragment_shader,0); - BasicBlock* label_forbody_preheader_i_122 = new BasicBlock("forbody.preheader.i",func_run_fragment_shader,0); - BasicBlock* label_forbody_i_123 = new BasicBlock("forbody.i",func_run_fragment_shader,0); - BasicBlock* label_from_consts_exit_124 = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); - - // Block entry (label_entry_121) - AllocaInst* ptr_consts_125 = new AllocaInst(ArrayTy_20, "consts", label_entry_121); - AllocaInst* ptr_temps_126 = new AllocaInst(ArrayTy_22, "temps", label_entry_121); - AllocaInst* ptr_args_127 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_121); - std::vector ptr_tmp_indices; - ptr_tmp_indices.push_back(const_int32_29); - ptr_tmp_indices.push_back(const_int32_35); - Instruction* ptr_tmp = new GetElementPtrInst(ptr_args_127, ptr_tmp_indices.begin(), ptr_tmp_indices.end(), "tmp", label_entry_121); - StoreInst* void_128 = new StoreInst(const_int32_29, ptr_tmp, false, label_entry_121); - ICmpInst* int1_cmp_i_129 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_120, const_int32_29, "cmp.i", label_entry_121); - new BranchInst(label_forbody_preheader_i_122, label_from_consts_exit_124, int1_cmp_i_129, label_entry_121); - - // Block forbody.preheader.i (label_forbody_preheader_i_122) - BinaryOperator* int32_tmp_i_131 = BinaryOperator::create(Instruction::Add, int32_num_consts_120, const_int32_30, "tmp.i", label_forbody_preheader_i_122); - ICmpInst* int1_tmp9_i_132 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_i_131, const_int32_29, "tmp9.i", label_forbody_preheader_i_122); - SelectInst* int32_tmp10_i_133 = new SelectInst(int1_tmp9_i_132, const_int32_31, int32_num_consts_120, "tmp10.i", label_forbody_preheader_i_122); - new BranchInst(label_forbody_i_123, label_forbody_preheader_i_122); - - // Block forbody.i (label_forbody_i_123) - Argument* fwdref_136 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_i_135 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i_123); - int32_i_0_reg2mem_0_i_135->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i_135->addIncoming(const_int32_29, label_forbody_preheader_i_122); - int32_i_0_reg2mem_0_i_135->addIncoming(fwdref_136, label_forbody_i_123); - - Argument* fwdref_138 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i_137 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i_123); - packed_vec_0_reg2mem_0_i_137->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_137->addIncoming(const_packed_32, label_forbody_preheader_i_122); - packed_vec_0_reg2mem_0_i_137->addIncoming(fwdref_138, label_forbody_i_123); - - std::vector ptr_arraydecay_i_139_indices; - ptr_arraydecay_i_139_indices.push_back(int32_i_0_reg2mem_0_i_135); - ptr_arraydecay_i_139_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay_i_139 = new GetElementPtrInst(ptr_aconsts_119, ptr_arraydecay_i_139_indices.begin(), ptr_arraydecay_i_139_indices.end(), "arraydecay.i", label_forbody_i_123); - LoadInst* float_tmp5_i_140 = new LoadInst(ptr_arraydecay_i_139, "tmp5.i", false, label_forbody_i_123); - InsertElementInst* packed_tmp7_i_141 = new InsertElementInst(packed_vec_0_reg2mem_0_i_137, float_tmp5_i_140, const_int32_29, "tmp7.i", label_forbody_i_123); - std::vector ptr_arrayidx12_i_142_indices; - ptr_arrayidx12_i_142_indices.push_back(int32_i_0_reg2mem_0_i_135); - ptr_arrayidx12_i_142_indices.push_back(const_int32_31); - Instruction* ptr_arrayidx12_i_142 = new GetElementPtrInst(ptr_aconsts_119, ptr_arrayidx12_i_142_indices.begin(), ptr_arrayidx12_i_142_indices.end(), "arrayidx12.i", label_forbody_i_123); - LoadInst* float_tmp13_i_143 = new LoadInst(ptr_arrayidx12_i_142, "tmp13.i", false, label_forbody_i_123); - InsertElementInst* packed_tmp15_i_144 = new InsertElementInst(packed_tmp7_i_141, float_tmp13_i_143, const_int32_31, "tmp15.i", label_forbody_i_123); - std::vector ptr_arrayidx20_i_145_indices; - ptr_arrayidx20_i_145_indices.push_back(int32_i_0_reg2mem_0_i_135); - ptr_arrayidx20_i_145_indices.push_back(const_int32_33); - Instruction* ptr_arrayidx20_i_145 = new GetElementPtrInst(ptr_aconsts_119, ptr_arrayidx20_i_145_indices.begin(), ptr_arrayidx20_i_145_indices.end(), "arrayidx20.i", label_forbody_i_123); - LoadInst* float_tmp21_i_146 = new LoadInst(ptr_arrayidx20_i_145, "tmp21.i", false, label_forbody_i_123); - InsertElementInst* packed_tmp23_i_147 = new InsertElementInst(packed_tmp15_i_144, float_tmp21_i_146, const_int32_33, "tmp23.i", label_forbody_i_123); - std::vector ptr_arrayidx28_i_148_indices; - ptr_arrayidx28_i_148_indices.push_back(int32_i_0_reg2mem_0_i_135); - ptr_arrayidx28_i_148_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx28_i_148 = new GetElementPtrInst(ptr_aconsts_119, ptr_arrayidx28_i_148_indices.begin(), ptr_arrayidx28_i_148_indices.end(), "arrayidx28.i", label_forbody_i_123); - LoadInst* float_tmp29_i_149 = new LoadInst(ptr_arrayidx28_i_148, "tmp29.i", false, label_forbody_i_123); - InsertElementInst* packed_tmp31_i_150 = new InsertElementInst(packed_tmp23_i_147, float_tmp29_i_149, const_int32_34, "tmp31.i", label_forbody_i_123); - std::vector ptr_arrayidx34_i_151_indices; - ptr_arrayidx34_i_151_indices.push_back(const_int32_29); - ptr_arrayidx34_i_151_indices.push_back(int32_i_0_reg2mem_0_i_135); - Instruction* ptr_arrayidx34_i_151 = new GetElementPtrInst(ptr_consts_125, ptr_arrayidx34_i_151_indices.begin(), ptr_arrayidx34_i_151_indices.end(), "arrayidx34.i", label_forbody_i_123); - StoreInst* void_152 = new StoreInst(packed_tmp31_i_150, ptr_arrayidx34_i_151, false, label_forbody_i_123); - BinaryOperator* int32_indvar_next7 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_135, const_int32_31, "indvar.next7", label_forbody_i_123); - ICmpInst* int1_exitcond8 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next7, int32_tmp10_i_133, "exitcond8", label_forbody_i_123); - new BranchInst(label_from_consts_exit_124, label_forbody_i_123, int1_exitcond8, label_forbody_i_123); - - // Block from_consts.exit (label_from_consts_exit_124) + BasicBlock* label_entry_110 = new BasicBlock("entry",func_run_fragment_shader,0); + BasicBlock* label_forbody_preheader_i_111 = new BasicBlock("forbody.preheader.i",func_run_fragment_shader,0); + BasicBlock* label_forbody_i_112 = new BasicBlock("forbody.i",func_run_fragment_shader,0); + BasicBlock* label_from_consts_exit_113 = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); + + // Block entry (label_entry_110) + AllocaInst* ptr_consts_114 = new AllocaInst(ArrayTy_20, "consts", label_entry_110); + AllocaInst* ptr_temps_115 = new AllocaInst(ArrayTy_22, "temps", label_entry_110); + AllocaInst* ptr_args_116 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_110); + std::vector ptr_tmp_117_indices; + ptr_tmp_117_indices.push_back(const_int32_29); + ptr_tmp_117_indices.push_back(const_int32_35); + Instruction* ptr_tmp_117 = new GetElementPtrInst(ptr_args_116, ptr_tmp_117_indices.begin(), ptr_tmp_117_indices.end(), "tmp", label_entry_110); + StoreInst* void_118 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_entry_110); + ICmpInst* int1_cmp_i_119 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_109, const_int32_29, "cmp.i", label_entry_110); + new BranchInst(label_forbody_preheader_i_111, label_from_consts_exit_113, int1_cmp_i_119, label_entry_110); + + // Block forbody.preheader.i (label_forbody_preheader_i_111) + BinaryOperator* int32_tmp_i_121 = BinaryOperator::create(Instruction::Add, int32_num_consts_109, const_int32_30, "tmp.i", label_forbody_preheader_i_111); + ICmpInst* int1_tmp9_i_122 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_i_121, const_int32_29, "tmp9.i", label_forbody_preheader_i_111); + SelectInst* int32_tmp10_i_123 = new SelectInst(int1_tmp9_i_122, const_int32_31, int32_num_consts_109, "tmp10.i", label_forbody_preheader_i_111); + new BranchInst(label_forbody_i_112, label_forbody_preheader_i_111); + + // Block forbody.i (label_forbody_i_112) + Argument* fwdref_126 = new Argument(IntegerType::get(32)); + PHINode* int32_i_0_reg2mem_0_i_125 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i_112); + int32_i_0_reg2mem_0_i_125->reserveOperandSpace(2); + int32_i_0_reg2mem_0_i_125->addIncoming(const_int32_29, label_forbody_preheader_i_111); + int32_i_0_reg2mem_0_i_125->addIncoming(fwdref_126, label_forbody_i_112); + + Argument* fwdref_128 = new Argument(VectorTy_1); + PHINode* packed_vec_0_reg2mem_0_i_127 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i_112); + packed_vec_0_reg2mem_0_i_127->reserveOperandSpace(2); + packed_vec_0_reg2mem_0_i_127->addIncoming(const_packed_32, label_forbody_preheader_i_111); + packed_vec_0_reg2mem_0_i_127->addIncoming(fwdref_128, label_forbody_i_112); + + std::vector ptr_arraydecay_i_129_indices; + ptr_arraydecay_i_129_indices.push_back(int32_i_0_reg2mem_0_i_125); + ptr_arraydecay_i_129_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay_i_129 = new GetElementPtrInst(ptr_aconsts_108, ptr_arraydecay_i_129_indices.begin(), ptr_arraydecay_i_129_indices.end(), "arraydecay.i", label_forbody_i_112); + LoadInst* float_tmp5_i_130 = new LoadInst(ptr_arraydecay_i_129, "tmp5.i", false, label_forbody_i_112); + InsertElementInst* packed_tmp7_i_131 = new InsertElementInst(packed_vec_0_reg2mem_0_i_127, float_tmp5_i_130, const_int32_29, "tmp7.i", label_forbody_i_112); + std::vector ptr_arrayidx12_i_132_indices; + ptr_arrayidx12_i_132_indices.push_back(int32_i_0_reg2mem_0_i_125); + ptr_arrayidx12_i_132_indices.push_back(const_int32_31); + Instruction* ptr_arrayidx12_i_132 = new GetElementPtrInst(ptr_aconsts_108, ptr_arrayidx12_i_132_indices.begin(), ptr_arrayidx12_i_132_indices.end(), "arrayidx12.i", label_forbody_i_112); + LoadInst* float_tmp13_i_133 = new LoadInst(ptr_arrayidx12_i_132, "tmp13.i", false, label_forbody_i_112); + InsertElementInst* packed_tmp15_i_134 = new InsertElementInst(packed_tmp7_i_131, float_tmp13_i_133, const_int32_31, "tmp15.i", label_forbody_i_112); + std::vector ptr_arrayidx20_i_135_indices; + ptr_arrayidx20_i_135_indices.push_back(int32_i_0_reg2mem_0_i_125); + ptr_arrayidx20_i_135_indices.push_back(const_int32_33); + Instruction* ptr_arrayidx20_i_135 = new GetElementPtrInst(ptr_aconsts_108, ptr_arrayidx20_i_135_indices.begin(), ptr_arrayidx20_i_135_indices.end(), "arrayidx20.i", label_forbody_i_112); + LoadInst* float_tmp21_i_136 = new LoadInst(ptr_arrayidx20_i_135, "tmp21.i", false, label_forbody_i_112); + InsertElementInst* packed_tmp23_i_137 = new InsertElementInst(packed_tmp15_i_134, float_tmp21_i_136, const_int32_33, "tmp23.i", label_forbody_i_112); + std::vector ptr_arrayidx28_i_138_indices; + ptr_arrayidx28_i_138_indices.push_back(int32_i_0_reg2mem_0_i_125); + ptr_arrayidx28_i_138_indices.push_back(const_int32_34); + Instruction* ptr_arrayidx28_i_138 = new GetElementPtrInst(ptr_aconsts_108, ptr_arrayidx28_i_138_indices.begin(), ptr_arrayidx28_i_138_indices.end(), "arrayidx28.i", label_forbody_i_112); + LoadInst* float_tmp29_i_139 = new LoadInst(ptr_arrayidx28_i_138, "tmp29.i", false, label_forbody_i_112); + InsertElementInst* packed_tmp31_i_140 = new InsertElementInst(packed_tmp23_i_137, float_tmp29_i_139, const_int32_34, "tmp31.i", label_forbody_i_112); + std::vector ptr_arrayidx34_i_141_indices; + ptr_arrayidx34_i_141_indices.push_back(const_int32_29); + ptr_arrayidx34_i_141_indices.push_back(int32_i_0_reg2mem_0_i_125); + Instruction* ptr_arrayidx34_i_141 = new GetElementPtrInst(ptr_consts_114, ptr_arrayidx34_i_141_indices.begin(), ptr_arrayidx34_i_141_indices.end(), "arrayidx34.i", label_forbody_i_112); + StoreInst* void_142 = new StoreInst(packed_tmp31_i_140, ptr_arrayidx34_i_141, false, label_forbody_i_112); + BinaryOperator* int32_indvar_next7 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_125, const_int32_31, "indvar.next7", label_forbody_i_112); + ICmpInst* int1_exitcond8 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next7, int32_tmp10_i_123, "exitcond8", label_forbody_i_112); + new BranchInst(label_from_consts_exit_113, label_forbody_i_112, int1_exitcond8, label_forbody_i_112); + + // Block from_consts.exit (label_from_consts_exit_113) std::vector ptr_tmp3_indices; ptr_tmp3_indices.push_back(const_int32_29); ptr_tmp3_indices.push_back(const_int32_34); - Instruction* ptr_tmp3 = new GetElementPtrInst(ptr_args_127, ptr_tmp3_indices.begin(), ptr_tmp3_indices.end(), "tmp3", label_from_consts_exit_124); + Instruction* ptr_tmp3 = new GetElementPtrInst(ptr_args_116, ptr_tmp3_indices.begin(), ptr_tmp3_indices.end(), "tmp3", label_from_consts_exit_113); std::vector ptr_arraydecay4_indices; ptr_arraydecay4_indices.push_back(const_int32_29); ptr_arraydecay4_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay4 = new GetElementPtrInst(ptr_consts_125, ptr_arraydecay4_indices.begin(), ptr_arraydecay4_indices.end(), "arraydecay4", label_from_consts_exit_124); - StoreInst* void_154 = new StoreInst(ptr_arraydecay4, ptr_tmp3, false, label_from_consts_exit_124); + Instruction* ptr_arraydecay4 = new GetElementPtrInst(ptr_consts_114, ptr_arraydecay4_indices.begin(), ptr_arraydecay4_indices.end(), "arraydecay4", label_from_consts_exit_113); + StoreInst* void_144 = new StoreInst(ptr_arraydecay4, ptr_tmp3, false, label_from_consts_exit_113); std::vector ptr_tmp5_indices; ptr_tmp5_indices.push_back(const_int32_29); ptr_tmp5_indices.push_back(const_int32_33); - Instruction* ptr_tmp5 = new GetElementPtrInst(ptr_args_127, ptr_tmp5_indices.begin(), ptr_tmp5_indices.end(), "tmp5", label_from_consts_exit_124); + Instruction* ptr_tmp5 = new GetElementPtrInst(ptr_args_116, ptr_tmp5_indices.begin(), ptr_tmp5_indices.end(), "tmp5", label_from_consts_exit_113); std::vector ptr_arraydecay6_indices; ptr_arraydecay6_indices.push_back(const_int32_29); ptr_arraydecay6_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay6 = new GetElementPtrInst(ptr_temps_126, ptr_arraydecay6_indices.begin(), ptr_arraydecay6_indices.end(), "arraydecay6", label_from_consts_exit_124); - StoreInst* void_155 = new StoreInst(ptr_arraydecay6, ptr_tmp5, false, label_from_consts_exit_124); - std::vector ptr_tmp8_156_indices; - ptr_tmp8_156_indices.push_back(const_int32_29); - ptr_tmp8_156_indices.push_back(const_int32_31); - Instruction* ptr_tmp8_156 = new GetElementPtrInst(ptr_args_127, ptr_tmp8_156_indices.begin(), ptr_tmp8_156_indices.end(), "tmp8", label_from_consts_exit_124); - std::vector ptr_tmp12_157_indices; - ptr_tmp12_157_indices.push_back(const_int32_29); - ptr_tmp12_157_indices.push_back(const_int32_29); - Instruction* ptr_tmp12_157 = new GetElementPtrInst(ptr_args_127, ptr_tmp12_157_indices.begin(), ptr_tmp12_157_indices.end(), "tmp12", label_from_consts_exit_124); - std::vector ptr_arraydecay11_158_indices; - ptr_arraydecay11_158_indices.push_back(const_int32_29); - ptr_arraydecay11_158_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11_158 = new GetElementPtrInst(ptr_inputs_117, ptr_arraydecay11_158_indices.begin(), ptr_arraydecay11_158_indices.end(), "arraydecay11", label_from_consts_exit_124); - StoreInst* void_159 = new StoreInst(ptr_arraydecay11_158, ptr_tmp8_156, false, label_from_consts_exit_124); - std::vector ptr_arraydecay16_160_indices; - ptr_arraydecay16_160_indices.push_back(const_int32_29); - ptr_arraydecay16_160_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay16_160 = new GetElementPtrInst(ptr_results_116, ptr_arraydecay16_160_indices.begin(), ptr_arraydecay16_160_indices.end(), "arraydecay16", label_from_consts_exit_124); - StoreInst* void_161 = new StoreInst(ptr_arraydecay16_160, ptr_tmp12_157, false, label_from_consts_exit_124); - StoreInst* void_162 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); - CallInst* void_163 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); - void_163->setCallingConv(CallingConv::C); - void_163->setTailCall(false);const ParamAttrsList *void_163_PAL = 0; - void_163->setParamAttrs(void_163_PAL); - - LoadInst* int32_tmp23 = new LoadInst(ptr_tmp, "tmp23", false, label_from_consts_exit_124); + Instruction* ptr_arraydecay6 = new GetElementPtrInst(ptr_temps_115, ptr_arraydecay6_indices.begin(), ptr_arraydecay6_indices.end(), "arraydecay6", label_from_consts_exit_113); + StoreInst* void_145 = new StoreInst(ptr_arraydecay6, ptr_tmp5, false, label_from_consts_exit_113); + std::vector ptr_tmp8_indices; + ptr_tmp8_indices.push_back(const_int32_29); + ptr_tmp8_indices.push_back(const_int32_31); + Instruction* ptr_tmp8 = new GetElementPtrInst(ptr_args_116, ptr_tmp8_indices.begin(), ptr_tmp8_indices.end(), "tmp8", label_from_consts_exit_113); + std::vector ptr_tmp12_indices; + ptr_tmp12_indices.push_back(const_int32_29); + ptr_tmp12_indices.push_back(const_int32_29); + Instruction* ptr_tmp12 = new GetElementPtrInst(ptr_args_116, ptr_tmp12_indices.begin(), ptr_tmp12_indices.end(), "tmp12", label_from_consts_exit_113); + std::vector ptr_arraydecay11_146_indices; + ptr_arraydecay11_146_indices.push_back(const_int32_29); + ptr_arraydecay11_146_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay11_146 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_146_indices.begin(), ptr_arraydecay11_146_indices.end(), "arraydecay11", label_from_consts_exit_113); + StoreInst* void_147 = new StoreInst(ptr_arraydecay11_146, ptr_tmp8, false, label_from_consts_exit_113); + std::vector ptr_arraydecay16_indices; + ptr_arraydecay16_indices.push_back(const_int32_29); + ptr_arraydecay16_indices.push_back(const_int32_29); + Instruction* ptr_arraydecay16 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_indices.begin(), ptr_arraydecay16_indices.end(), "arraydecay16", label_from_consts_exit_113); + StoreInst* void_148 = new StoreInst(ptr_arraydecay16, ptr_tmp12, false, label_from_consts_exit_113); + StoreInst* void_149 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113); + CallInst* void_150 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113); + void_150->setCallingConv(CallingConv::C); + void_150->setTailCall(false);const ParamAttrsList *void_150_PAL = 0; + void_150->setParamAttrs(void_150_PAL); + + LoadInst* int32_tmp23 = new LoadInst(ptr_tmp_117, "tmp23", false, label_from_consts_exit_113); std::vector ptr_arraydecay11_1_indices; ptr_arraydecay11_1_indices.push_back(const_int32_31); ptr_arraydecay11_1_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11_1 = new GetElementPtrInst(ptr_inputs_117, ptr_arraydecay11_1_indices.begin(), ptr_arraydecay11_1_indices.end(), "arraydecay11.1", label_from_consts_exit_124); - StoreInst* void_164 = new StoreInst(ptr_arraydecay11_1, ptr_tmp8_156, false, label_from_consts_exit_124); + Instruction* ptr_arraydecay11_1 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_1_indices.begin(), ptr_arraydecay11_1_indices.end(), "arraydecay11.1", label_from_consts_exit_113); + StoreInst* void_151 = new StoreInst(ptr_arraydecay11_1, ptr_tmp8, false, label_from_consts_exit_113); std::vector ptr_arraydecay16_1_indices; ptr_arraydecay16_1_indices.push_back(const_int32_31); ptr_arraydecay16_1_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay16_1 = new GetElementPtrInst(ptr_results_116, ptr_arraydecay16_1_indices.begin(), ptr_arraydecay16_1_indices.end(), "arraydecay16.1", label_from_consts_exit_124); - StoreInst* void_165 = new StoreInst(ptr_arraydecay16_1, ptr_tmp12_157, false, label_from_consts_exit_124); - StoreInst* void_166 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); - CallInst* void_167 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); - void_167->setCallingConv(CallingConv::C); - void_167->setTailCall(false);const ParamAttrsList *void_167_PAL = 0; - void_167->setParamAttrs(void_167_PAL); - - LoadInst* int32_tmp23_1 = new LoadInst(ptr_tmp, "tmp23.1", false, label_from_consts_exit_124); - BinaryOperator* int32_shl_1 = BinaryOperator::create(Instruction::Shl, int32_tmp23_1, const_int32_31, "shl.1", label_from_consts_exit_124); - BinaryOperator* int32_or_1 = BinaryOperator::create(Instruction::Or, int32_shl_1, int32_tmp23, "or.1", label_from_consts_exit_124); + Instruction* ptr_arraydecay16_1 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_1_indices.begin(), ptr_arraydecay16_1_indices.end(), "arraydecay16.1", label_from_consts_exit_113); + StoreInst* void_152 = new StoreInst(ptr_arraydecay16_1, ptr_tmp12, false, label_from_consts_exit_113); + StoreInst* void_153 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113); + CallInst* void_154 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113); + void_154->setCallingConv(CallingConv::C); + void_154->setTailCall(false);const ParamAttrsList *void_154_PAL = 0; + void_154->setParamAttrs(void_154_PAL); + + LoadInst* int32_tmp23_1 = new LoadInst(ptr_tmp_117, "tmp23.1", false, label_from_consts_exit_113); + BinaryOperator* int32_shl_1 = BinaryOperator::create(Instruction::Shl, int32_tmp23_1, const_int32_31, "shl.1", label_from_consts_exit_113); + BinaryOperator* int32_or_1 = BinaryOperator::create(Instruction::Or, int32_shl_1, int32_tmp23, "or.1", label_from_consts_exit_113); std::vector ptr_arraydecay11_2_indices; ptr_arraydecay11_2_indices.push_back(const_int32_33); ptr_arraydecay11_2_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11_2 = new GetElementPtrInst(ptr_inputs_117, ptr_arraydecay11_2_indices.begin(), ptr_arraydecay11_2_indices.end(), "arraydecay11.2", label_from_consts_exit_124); - StoreInst* void_168 = new StoreInst(ptr_arraydecay11_2, ptr_tmp8_156, false, label_from_consts_exit_124); + Instruction* ptr_arraydecay11_2 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_2_indices.begin(), ptr_arraydecay11_2_indices.end(), "arraydecay11.2", label_from_consts_exit_113); + StoreInst* void_155 = new StoreInst(ptr_arraydecay11_2, ptr_tmp8, false, label_from_consts_exit_113); std::vector ptr_arraydecay16_2_indices; ptr_arraydecay16_2_indices.push_back(const_int32_33); ptr_arraydecay16_2_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay16_2 = new GetElementPtrInst(ptr_results_116, ptr_arraydecay16_2_indices.begin(), ptr_arraydecay16_2_indices.end(), "arraydecay16.2", label_from_consts_exit_124); - StoreInst* void_169 = new StoreInst(ptr_arraydecay16_2, ptr_tmp12_157, false, label_from_consts_exit_124); - StoreInst* void_170 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); - CallInst* void_171 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); - void_171->setCallingConv(CallingConv::C); - void_171->setTailCall(false);const ParamAttrsList *void_171_PAL = 0; - void_171->setParamAttrs(void_171_PAL); - - LoadInst* int32_tmp23_2 = new LoadInst(ptr_tmp, "tmp23.2", false, label_from_consts_exit_124); - BinaryOperator* int32_shl_2 = BinaryOperator::create(Instruction::Shl, int32_tmp23_2, const_int32_33, "shl.2", label_from_consts_exit_124); - BinaryOperator* int32_or_2 = BinaryOperator::create(Instruction::Or, int32_shl_2, int32_or_1, "or.2", label_from_consts_exit_124); + Instruction* ptr_arraydecay16_2 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_2_indices.begin(), ptr_arraydecay16_2_indices.end(), "arraydecay16.2", label_from_consts_exit_113); + StoreInst* void_156 = new StoreInst(ptr_arraydecay16_2, ptr_tmp12, false, label_from_consts_exit_113); + StoreInst* void_157 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113); + CallInst* void_158 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113); + void_158->setCallingConv(CallingConv::C); + void_158->setTailCall(false);const ParamAttrsList *void_158_PAL = 0; + void_158->setParamAttrs(void_158_PAL); + + LoadInst* int32_tmp23_2 = new LoadInst(ptr_tmp_117, "tmp23.2", false, label_from_consts_exit_113); + BinaryOperator* int32_shl_2 = BinaryOperator::create(Instruction::Shl, int32_tmp23_2, const_int32_33, "shl.2", label_from_consts_exit_113); + BinaryOperator* int32_or_2 = BinaryOperator::create(Instruction::Or, int32_shl_2, int32_or_1, "or.2", label_from_consts_exit_113); std::vector ptr_arraydecay11_3_indices; ptr_arraydecay11_3_indices.push_back(const_int32_34); ptr_arraydecay11_3_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11_3 = new GetElementPtrInst(ptr_inputs_117, ptr_arraydecay11_3_indices.begin(), ptr_arraydecay11_3_indices.end(), "arraydecay11.3", label_from_consts_exit_124); - StoreInst* void_172 = new StoreInst(ptr_arraydecay11_3, ptr_tmp8_156, false, label_from_consts_exit_124); + Instruction* ptr_arraydecay11_3 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_3_indices.begin(), ptr_arraydecay11_3_indices.end(), "arraydecay11.3", label_from_consts_exit_113); + StoreInst* void_159 = new StoreInst(ptr_arraydecay11_3, ptr_tmp8, false, label_from_consts_exit_113); std::vector ptr_arraydecay16_3_indices; ptr_arraydecay16_3_indices.push_back(const_int32_34); ptr_arraydecay16_3_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay16_3 = new GetElementPtrInst(ptr_results_116, ptr_arraydecay16_3_indices.begin(), ptr_arraydecay16_3_indices.end(), "arraydecay16.3", label_from_consts_exit_124); - StoreInst* void_173 = new StoreInst(ptr_arraydecay16_3, ptr_tmp12_157, false, label_from_consts_exit_124); - StoreInst* void_174 = new StoreInst(const_int32_29, ptr_tmp, false, label_from_consts_exit_124); - CallInst* void_175 = new CallInst(func_execute_shader, ptr_args_127, "", label_from_consts_exit_124); - void_175->setCallingConv(CallingConv::C); - void_175->setTailCall(false);const ParamAttrsList *void_175_PAL = 0; - void_175->setParamAttrs(void_175_PAL); - - LoadInst* int32_tmp23_3 = new LoadInst(ptr_tmp, "tmp23.3", false, label_from_consts_exit_124); - BinaryOperator* int32_shl_3 = BinaryOperator::create(Instruction::Shl, int32_tmp23_3, const_int32_34, "shl.3", label_from_consts_exit_124); - BinaryOperator* int32_or_3 = BinaryOperator::create(Instruction::Or, int32_shl_3, int32_or_2, "or.3", label_from_consts_exit_124); - BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_or_3, const_int32_30, "neg", label_from_consts_exit_124); - new ReturnInst(int32_neg, label_from_consts_exit_124); + Instruction* ptr_arraydecay16_3 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_3_indices.begin(), ptr_arraydecay16_3_indices.end(), "arraydecay16.3", label_from_consts_exit_113); + StoreInst* void_160 = new StoreInst(ptr_arraydecay16_3, ptr_tmp12, false, label_from_consts_exit_113); + StoreInst* void_161 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113); + CallInst* void_162 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113); + void_162->setCallingConv(CallingConv::C); + void_162->setTailCall(false);const ParamAttrsList *void_162_PAL = 0; + void_162->setParamAttrs(void_162_PAL); + + LoadInst* int32_tmp23_3 = new LoadInst(ptr_tmp_117, "tmp23.3", false, label_from_consts_exit_113); + BinaryOperator* int32_shl_3 = BinaryOperator::create(Instruction::Shl, int32_tmp23_3, const_int32_34, "shl.3", label_from_consts_exit_113); + BinaryOperator* int32_or_3 = BinaryOperator::create(Instruction::Or, int32_shl_3, int32_or_2, "or.3", label_from_consts_exit_113); + BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_or_3, const_int32_30, "neg", label_from_consts_exit_113); + new ReturnInst(int32_neg, label_from_consts_exit_113); // Resolve Forward References - fwdref_138->replaceAllUsesWith(packed_tmp31_i_150); delete fwdref_138; - fwdref_136->replaceAllUsesWith(int32_indvar_next7); delete fwdref_136; + fwdref_128->replaceAllUsesWith(packed_tmp31_i_140); delete fwdref_128; + fwdref_126->replaceAllUsesWith(int32_indvar_next7); delete fwdref_126; } diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c index c3b34584e1..fa50b60e66 100644 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ b/src/mesa/pipe/llvm/llvm_entry.c @@ -86,8 +86,8 @@ struct ShaderInput extern void execute_shader(struct ShaderInput *input); -void run_vertex_shader(float4 (*inputs)[16], - float4 (*results)[16], +void run_vertex_shader(void *inputs, + void *results, float (*aconsts)[4], int num_vertices, int num_inputs, @@ -98,16 +98,16 @@ void run_vertex_shader(float4 (*inputs)[16], float4 temps[128];//MAX_PROGRAM_TEMPS struct ShaderInput args; + args.dests = results; + args.inputs = inputs; + /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", num_vertices, num_inputs, num_attribs, num_consts);*/ from_consts(consts, aconsts, num_consts); args.consts = consts; args.temps = temps; - for (int i = 0; i < num_vertices; ++i) { - args.dests = results[i]; - args.inputs = inputs[i]; - execute_shader(&args); - } + + execute_shader(&args); } diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp new file mode 100644 index 0000000000..eb9e1196f1 --- /dev/null +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -0,0 +1,682 @@ +#include "tgsitollvm.h" + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "storage.h" +#include "instructions.h" + +#include "pipe/p_shader_tokens.h" + +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/exec/tgsi_exec.h" +#include "pipe/tgsi/util/tgsi_util.h" +#include "pipe/tgsi/util/tgsi_build.h" +#include "pipe/tgsi/util/tgsi_dump.h" + + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + + +#include +#include +#include + +using namespace llvm; +#include "llvm_base_shader.cpp" + +static inline void +add_interpolator(struct gallivm_ir *ir, + struct gallivm_interpolate *interp) +{ + ir->interpolators[ir->num_interp] = *interp; + ++ir->num_interp; +} + +static void +translate_declaration(struct gallivm_ir *prog, + llvm::Module *module, + Storage *storage, + struct tgsi_full_declaration *decl, + struct tgsi_full_declaration *fd) +{ + if (decl->Declaration.File == TGSI_FILE_INPUT) { + unsigned first, last, mask; + uint interp_method; + + assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if (first == 0) { + mask &= ~TGSI_WRITEMASK_XY; + if (mask == TGSI_WRITEMASK_NONE) { + first++; + if (first > last) { + return; + } + } + } + + interp_method = decl->Interpolation.Interpolate; + + if (mask == TGSI_WRITEMASK_XYZW) { + unsigned i, j; + + for (i = first; i <= last; i++) { + for (j = 0; j < NUM_CHANNELS; j++) { + //interp( mach, i, j ); + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } else { + unsigned i, j; + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } + } + } +} + + +static void +translate_immediate(Storage *storage, + struct tgsi_full_immediate *imm) +{ + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert( 0 ); + } + } + storage->addImmediate(vec); +} + +static inline llvm::Value * +swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, + Storage *storage) +{ + int swizzle = 0; + int start = 1000; + const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + + TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; + for (int k = 0; k < 4; ++k) { + swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; + start /= 10; + } + if (swizzle != NO_SWIZZLE) { + /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ + val = storage->shuffleVector(val, swizzle); + } + return val; +} + +static void +translate_instruction(llvm::Module *module, + Storage *storage, + Instructions *instr, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi, + unsigned instno) +{ + llvm::Value *inputs[4]; + inputs[0] = 0; + inputs[1] = 0; + inputs[2] = 0; + inputs[3] = 0; + + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + llvm::Value *val = 0; + llvm::Value *indIdx = 0; + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + indIdx = storage->extractIndex(indIdx); + } + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->constElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->inputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->tempElement(src->SrcRegister.Index); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->outputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->immediateElement(src->SrcRegister.Index); + } else { + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); + return; + } + + inputs[i] = swizzleVector(val, src, storage); + } + + /*if (inputs[0]) + instr->printVector(inputs[0]); + if (inputs[1]) + instr->printVector(inputs[1]);*/ + llvm::Value *out = 0; + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); + } + break; + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } + break; + case TGSI_OPCODE_LIT: { + out = instr->lit(inputs[0]); + } + break; + case TGSI_OPCODE_RCP: { + out = instr->rcp(inputs[0]); + } + break; + case TGSI_OPCODE_RSQ: { + out = instr->rsq(inputs[0]); + } + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP3: { + out = instr->dp3(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP4: { + out = instr->dp4(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DST: { + out = instr->dst(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MIN: { + out = instr->min(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAX: { + out = instr->max(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SLT: { + out = instr->slt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SGE: { + out = instr->sge(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SUB: { + out = instr->sub(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_LERP: { + out = instr->lerp(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: { + out = instr->frc(inputs[0]); + } + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: { + out = instr->floor(inputs[0]); + } + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: { + out = instr->ex2(inputs[0]); + } + break; + case TGSI_OPCODE_LOGBASE2: { + out = instr->lg2(inputs[0]); + } + break; + case TGSI_OPCODE_POWER: { + out = instr->pow(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_CROSSPRODUCT: { + out = instr->cross(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: { + out = instr->abs(inputs[0]); + } + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: { + out = instr->dph(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_COS: { + out = instr->cos(inputs[0]); + } + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: { + out = instr->kilp(inputs[0]); + storage->setKilElement(out); + return; + } + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: { + out = instr->sgt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SIN: { + out = instr->sin(inputs[0]); + } + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: { + instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr()); + return; + } + break; + case TGSI_OPCODE_RET: { + instr->end(); + return; + } + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: { + out = instr->cmp(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SCS: { + out = instr->scs(inputs[0]); + } + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: { + instr->brk(); + return; + } + break; + case TGSI_OPCODE_IF: { + instr->ifop(inputs[0]); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: { + instr->elseop(); + storage->setCurrentBlock(instr->currentBlock()); + return; //only state update + } + break; + case TGSI_OPCODE_ENDIF: { + instr->endif(); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: { + out = instr->trunc(inputs[0]); + } + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: { + instr->beginLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_BGNSUB: { + instr->bgnSub(instno); + storage->setCurrentBlock(instr->currentBlock()); + storage->pushTemps(); + return; + } + break; + case TGSI_OPCODE_ENDLOOP2: { + instr->endLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_ENDSUB: { + instr->endSub(); + storage->setCurrentBlock(instr->currentBlock()); + storage->popArguments(); + storage->popTemps(); + return; + } + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + instr->end(); + return; + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + if (!out) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } + + /* # not sure if we need this */ + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + /*TXT( "_SAT" );*/ + break; + case TGSI_SAT_MINUS_PLUS_ONE: + /*TXT( "_SAT[-1,1]" );*/ + break; + default: + assert( 0 ); + } + + /* store results */ + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); + } + } +} + +llvm::Module * +tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) +{ + llvm::Module *mod = createBaseShader(); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned instno = 0; + Function* shader = mod->getFunction("execute_shader"); + std::ostringstream stream; + if (ir->type == GALLIVM_VS) { + stream << "vs_shader"; + } else { + stream << "fs_shader"; + } + stream << ir->id; + std::string func_name = stream.str(); + shader->setName(func_name.c_str()); + + Function::arg_iterator args = shader->arg_begin(); + Value *ptr_INPUT = args++; + ptr_INPUT->setName("input"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + + tgsi_parse_init(&parse, tokens); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + Storage storage(label_entry, ptr_INPUT); + Instructions instr(mod, shader, label_entry, &storage); + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declaration(ir, mod, &storage, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediate(&storage, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + translate_instruction(mod, &storage, &instr, + &parse.FullToken.FullInstruction, + &fi, instno); + ++instno; + break; + + default: + assert(0); + } + } + + tgsi_parse_free(&parse); + + ir->num_consts = storage.numConsts(); + return mod; +} diff --git a/src/mesa/pipe/llvm/tgsitollvm.h b/src/mesa/pipe/llvm/tgsitollvm.h new file mode 100644 index 0000000000..073ffb5749 --- /dev/null +++ b/src/mesa/pipe/llvm/tgsitollvm.h @@ -0,0 +1,16 @@ +#ifndef TGSITOLLVM_H +#define TGSITOLLVM_H + + +namespace llvm { + class Module; +} + +struct gallivm_ir; +struct tgsi_token; + + +llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, + const struct tgsi_token *tokens); + +#endif diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 1430be7869..598a70c827 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -61,7 +61,7 @@ softpipe_create_fs_state(struct pipe_context *pipe, } #ifdef MESA_LLVM - state->llvm_prog = gallivm_from_tgsi(state->shader.tokens, GALLIVM_FS); + state->llvm_prog = 0; if (!gallivm_global_cpu_engine()) { gallivm_cpu_engine_create(state->llvm_prog); } -- cgit v1.2.3 From 137edf75335910c9f15daacaf0ce3f4cbd43080c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 5 Feb 2008 03:09:24 -0500 Subject: rewrite the way cpu engine is handled --- src/mesa/pipe/draw/draw_vertex_shader.c | 9 +- src/mesa/pipe/llvm/Makefile | 1 + src/mesa/pipe/llvm/gallivm.cpp | 134 --------------------- src/mesa/pipe/llvm/gallivm.h | 37 +++--- src/mesa/pipe/llvm/gallivm_cpu.cpp | 204 ++++++++++++++++++++++++++++++++ 5 files changed, 227 insertions(+), 158 deletions(-) create mode 100644 src/mesa/pipe/llvm/gallivm_cpu.cpp (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 9567283ff5..574ac67057 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -116,11 +116,10 @@ run_vertex_program(struct draw_context *draw, #ifdef MESA_LLVM if (1) { struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; - gallivm_prog_exec(prog, - machine->Inputs, - machine->Outputs, - machine->Consts, - 12, 12, 12); + gallivm_cpu_vs_exec(prog, + machine->Inputs, + machine->Outputs, + machine->Consts); } else #elif defined(__i386__) || defined(__386__) if (draw->use_sse) { diff --git a/src/mesa/pipe/llvm/Makefile b/src/mesa/pipe/llvm/Makefile index 10ff7aacae..b333661c6f 100644 --- a/src/mesa/pipe/llvm/Makefile +++ b/src/mesa/pipe/llvm/Makefile @@ -7,6 +7,7 @@ LIBNAME = gallivm GALLIVM_SOURCES = \ gallivm.cpp \ + gallivm_cpu.cpp \ instructions.cpp \ loweringpass.cpp \ tgsitollvm.cpp \ diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index cf9b0f6406..cb9a88f0b8 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -70,10 +70,6 @@ #include #include -struct gallivm_cpu_engine { - llvm::ExecutionEngine *engine; -}; - static int GLOBAL_ID = 0; using namespace llvm; @@ -145,38 +141,6 @@ void gallivm_prog_delete(struct gallivm_prog *prog) free(prog); } -typedef void (*vertex_shader_runner)(void *ainputs, - void *dests, - float (*aconsts)[4], - int num_vertices, - int num_inputs, - int num_attribs, - int num_consts); - - -/*! - This function is used to execute the gallivm_prog in software. Before calling - this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile - function. - */ -int gallivm_prog_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - int num_vertices, - int num_inputs, - int num_attribs) -{ - vertex_shader_runner runner = reinterpret_cast(prog->function); - assert(runner); - runner(inputs, dests, consts, num_vertices, num_inputs, - num_attribs, prog->num_consts); - - return 0; -} - - - static inline void constant_interpolation(float (*inputs)[16][4], const struct tgsi_interp_coef *coefs, @@ -231,28 +195,6 @@ perspective_interpolation(float (*inputs)[16][4], } } -typedef int (*fragment_shader_runner)(float x, float y, - float (*dests)[16][4], - float (*inputs)[16][4], - int num_attribs, - float (*consts)[4], int num_consts, - struct tgsi_sampler *samplers); - -int gallivm_fragment_shader_exec(struct gallivm_prog *prog, - float fx, float fy, - float (*dests)[16][4], - float (*inputs)[16][4], - float (*consts)[4], - struct tgsi_sampler *samplers) -{ - fragment_shader_runner runner = reinterpret_cast(prog->function); - assert(runner); - - return runner(fx, fy, dests, inputs, prog->num_interp, - consts, prog->num_consts, - samplers); -} - void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix) { if (!ir || !ir->module) @@ -292,82 +234,6 @@ void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix) } -static struct gallivm_cpu_engine *CPU = 0; - -static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) -{ - llvm::Module *mod = prog->module; - llvm::Function *func = 0; - - switch (prog->type) { - case GALLIVM_VS: - func = mod->getFunction("run_vertex_shader"); - break; - case GALLIVM_FS: - func = mod->getFunction("run_fragment_shader"); - break; - default: - assert(!"Unknown shader type!"); - break; - } - return func; -} - -/*! - This function creates a CPU based execution engine for the given gallivm_prog. - gallivm_cpu_engine should be used as a singleton throughout the library. Before - executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile. - The gallivm_prog instance which is being passed to the constructor is being - automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile - with it again. - */ -struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) -{ - struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *) - calloc(1, sizeof(struct gallivm_cpu_engine)); - llvm::Module *mod = static_cast(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); - ee->DisableLazyCompilation(); - cpu->engine = ee; - - llvm::Function *func = func_for_shader(prog); - - prog->function = ee->getPointerToFunction(func); - CPU = cpu; - return cpu; -} - - -/*! - This function JIT compiles the given gallivm_prog with the given cpu based execution engine. - The reference to the generated machine code entry point will be stored - in the gallivm_prog program. After executing this function one can call gallivm_prog_exec - in order to execute the gallivm_prog on the CPU. - */ -void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) -{ - llvm::Module *mod = static_cast(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = cpu->engine; - assert(ee); - ee->DisableLazyCompilation(); - ee->addModuleProvider(mp); - - llvm::Function *func = func_for_shader(prog); - prog->function = ee->getPointerToFunction(func); -} - -void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) -{ - free(cpu); -} - -struct gallivm_cpu_engine * gallivm_global_cpu_engine() -{ - return CPU; -} - void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, float (*inputs)[16][4], const struct tgsi_interp_coef *coef) diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h index f9f5d5ee74..98eda56f81 100644 --- a/src/mesa/pipe/llvm/gallivm.h +++ b/src/mesa/pipe/llvm/gallivm.h @@ -61,28 +61,16 @@ enum gallivm_vector_layout { }; struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type); -void gallivm_ir_set_layout(struct gallivm_ir *prog, - enum gallivm_vector_layout layout); -void gallivm_ir_set_components(struct gallivm_ir *prog, int num); -void gallivm_ir_fill_from_tgsi(struct gallivm_ir *prog, - const struct tgsi_token *tokens); -void gallivm_ir_delete(struct gallivm_ir *prog); +void gallivm_ir_set_layout(struct gallivm_ir *prog, + enum gallivm_vector_layout layout); +void gallivm_ir_set_components(struct gallivm_ir *prog, int num); +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *prog, + const struct tgsi_token *tokens); +void gallivm_ir_delete(struct gallivm_ir *prog); + struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir); -int gallivm_prog_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - int num_vertices, - int num_inputs, - int num_attribs); -int gallivm_fragment_shader_exec(struct gallivm_prog *prog, - float x, float y, - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*consts)[4], - struct tgsi_sampler *samplers); void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], const struct tgsi_interp_coef *coefs); @@ -91,9 +79,20 @@ void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); struct gallivm_cpu_engine *gallivm_global_cpu_engine(); +int gallivm_cpu_vs_exec(struct gallivm_prog *prog, + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, + float (*consts)[4]); +int gallivm_cpu_fs_exec(struct gallivm_prog *prog, + float x, float y, + float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + float (*consts)[4], + struct tgsi_sampler *samplers); void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); + #endif /* MESA_LLVM */ #if defined __cplusplus diff --git a/src/mesa/pipe/llvm/gallivm_cpu.cpp b/src/mesa/pipe/llvm/gallivm_cpu.cpp new file mode 100644 index 0000000000..5f1268bf4f --- /dev/null +++ b/src/mesa/pipe/llvm/gallivm_cpu.cpp @@ -0,0 +1,204 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "instructions.h" +#include "loweringpass.h" +#include "storage.h" +#include "tgsitollvm.h" + +#include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" + +#include "pipe/tgsi/exec/tgsi_exec.h" +#include "pipe/tgsi/util/tgsi_dump.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +struct gallivm_cpu_engine { + llvm::ExecutionEngine *engine; +}; + +static struct gallivm_cpu_engine *CPU = 0; + +typedef int (*fragment_shader_runner)(float x, float y, + float (*dests)[16][4], + float (*inputs)[16][4], + int num_attribs, + float (*consts)[4], int num_consts, + struct tgsi_sampler *samplers); + +int gallivm_cpu_fs_exec(struct gallivm_prog *prog, + float fx, float fy, + float (*dests)[16][4], + float (*inputs)[16][4], + float (*consts)[4], + struct tgsi_sampler *samplers) +{ + fragment_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + + return runner(fx, fy, dests, inputs, prog->num_interp, + consts, prog->num_consts, + samplers); +} + +static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) +{ + llvm::Module *mod = prog->module; + llvm::Function *func = 0; + + switch (prog->type) { + case GALLIVM_VS: + func = mod->getFunction("run_vertex_shader"); + break; + case GALLIVM_FS: + func = mod->getFunction("run_fragment_shader"); + break; + default: + assert(!"Unknown shader type!"); + break; + } + return func; +} + +/*! + This function creates a CPU based execution engine for the given gallivm_prog. + gallivm_cpu_engine should be used as a singleton throughout the library. Before + executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile. + The gallivm_prog instance which is being passed to the constructor is being + automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile + with it again. + */ +struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) +{ + struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *) + calloc(1, sizeof(struct gallivm_cpu_engine)); + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); + ee->DisableLazyCompilation(); + cpu->engine = ee; + + llvm::Function *func = func_for_shader(prog); + + prog->function = ee->getPointerToFunction(func); + CPU = cpu; + return cpu; +} + + +/*! + This function JIT compiles the given gallivm_prog with the given cpu based execution engine. + The reference to the generated machine code entry point will be stored + in the gallivm_prog program. After executing this function one can call gallivm_prog_exec + in order to execute the gallivm_prog on the CPU. + */ +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) +{ + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = cpu->engine; + assert(ee); + /*FIXME : remove */ + ee->DisableLazyCompilation(); + ee->addModuleProvider(mp); + + llvm::Function *func = func_for_shader(prog); + prog->function = ee->getPointerToFunction(func); +} + +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) +{ + free(cpu); +} + +struct gallivm_cpu_engine * gallivm_global_cpu_engine() +{ + return CPU; +} + + +typedef void (*vertex_shader_runner)(void *ainputs, + void *dests, + float (*aconsts)[4], + int num_vertices, + int num_inputs, + int num_attribs, + int num_consts); + + +/*! + This function is used to execute the gallivm_prog in software. Before calling + this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile + function. + */ +int gallivm_cpu_vs_exec(struct gallivm_prog *prog, + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, + float (*consts)[4]) +{ + vertex_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + /*FIXME*/ + runner(inputs, dests, consts, 4, 4, 4, prog->num_consts); + + return 0; +} + +#endif -- cgit v1.2.3 From 1712a5380a12fe66fa03a281394abeca034f1a7c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 6 Feb 2008 06:12:34 -0500 Subject: disable llvm for fragment shaders for now --- src/mesa/pipe/softpipe/sp_state_fs.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 598a70c827..0b814fc284 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -62,11 +62,15 @@ softpipe_create_fs_state(struct pipe_context *pipe, #ifdef MESA_LLVM state->llvm_prog = 0; + +#if 0 if (!gallivm_global_cpu_engine()) { gallivm_cpu_engine_create(state->llvm_prog); } else gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); +#endif + #elif defined(__i386__) || defined(__386__) if (softpipe->use_sse) { x86_init_func( &state->sse2_program ); -- cgit v1.2.3 From e761161d014cde4e0a6e52631e9e4bc740b554f6 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 11 Feb 2008 09:43:59 -0500 Subject: start genering soa type code in llvm paths --- src/mesa/pipe/llvm/Makefile | 4 +- src/mesa/pipe/llvm/gallivm.cpp | 3 + src/mesa/pipe/llvm/gallivm.h | 8 +- src/mesa/pipe/llvm/instructionssoa.cpp | 26 ++ src/mesa/pipe/llvm/instructionssoa.h | 55 ++++ src/mesa/pipe/llvm/storagesoa.cpp | 117 ++++++++ src/mesa/pipe/llvm/storagesoa.h | 71 +++++ src/mesa/pipe/llvm/tgsitollvm.cpp | 506 ++++++++++++++++++++++++++++++++- src/mesa/pipe/llvm/tgsitollvm.h | 4 + 9 files changed, 782 insertions(+), 12 deletions(-) create mode 100644 src/mesa/pipe/llvm/instructionssoa.cpp create mode 100644 src/mesa/pipe/llvm/instructionssoa.h create mode 100644 src/mesa/pipe/llvm/storagesoa.cpp create mode 100644 src/mesa/pipe/llvm/storagesoa.h (limited to 'src') diff --git a/src/mesa/pipe/llvm/Makefile b/src/mesa/pipe/llvm/Makefile index b333661c6f..a0494ba966 100644 --- a/src/mesa/pipe/llvm/Makefile +++ b/src/mesa/pipe/llvm/Makefile @@ -11,7 +11,9 @@ GALLIVM_SOURCES = \ instructions.cpp \ loweringpass.cpp \ tgsitollvm.cpp \ - storage.cpp + storage.cpp \ + storagesoa.cpp \ + instructionssoa.cpp INC_SOURCES = gallivm_builtins.cpp llvm_base_shader.cpp diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index cb9a88f0b8..b99dc6db5b 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -288,6 +288,9 @@ void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, std::cout << "Creating llvm from: " <module = mod; gallivm_ir_dump(ir, 0); diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h index 98eda56f81..b104520cb7 100644 --- a/src/mesa/pipe/llvm/gallivm.h +++ b/src/mesa/pipe/llvm/gallivm.h @@ -61,12 +61,12 @@ enum gallivm_vector_layout { }; struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type); -void gallivm_ir_set_layout(struct gallivm_ir *prog, +void gallivm_ir_set_layout(struct gallivm_ir *ir, enum gallivm_vector_layout layout); -void gallivm_ir_set_components(struct gallivm_ir *prog, int num); -void gallivm_ir_fill_from_tgsi(struct gallivm_ir *prog, +void gallivm_ir_set_components(struct gallivm_ir *ir, int num); +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, const struct tgsi_token *tokens); -void gallivm_ir_delete(struct gallivm_ir *prog); +void gallivm_ir_delete(struct gallivm_ir *ir); struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir); diff --git a/src/mesa/pipe/llvm/instructionssoa.cpp b/src/mesa/pipe/llvm/instructionssoa.cpp new file mode 100644 index 0000000000..9ac4d8fbc7 --- /dev/null +++ b/src/mesa/pipe/llvm/instructionssoa.cpp @@ -0,0 +1,26 @@ +#include "instructionssoa.h" + +InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, + llvm::BasicBlock *block, StorageSoa *storage) +{ +} + +std::vector InstructionsSoa::add(const std::vector in1, + const std::vector in2) +{ + std::vector res(4); + + return res; +} + +std::vector InstructionsSoa::mul(const std::vector in1, + const std::vector in2) +{ + std::vector res(4); + + return res; +} + +void InstructionsSoa::end() +{ +} diff --git a/src/mesa/pipe/llvm/instructionssoa.h b/src/mesa/pipe/llvm/instructionssoa.h new file mode 100644 index 0000000000..0b6b41cf05 --- /dev/null +++ b/src/mesa/pipe/llvm/instructionssoa.h @@ -0,0 +1,55 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INSTRUCTIONSSOA_H +#define INSTRUCTIONSSOA_H + +#include + +namespace llvm { + class Module; + class Function; + class BasicBlock; + class Value; +} +class StorageSoa; + +class InstructionsSoa +{ +public: + InstructionsSoa(llvm::Module *mod, llvm::Function *func, + llvm::BasicBlock *block, StorageSoa *storage); + + std::vector add(const std::vector in1, + const std::vector in2); + std::vector mul(const std::vector in1, + const std::vector in2); + void end(); +}; + + +#endif diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp new file mode 100644 index 0000000000..b2aca3557a --- /dev/null +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -0,0 +1,117 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "storagesoa.h" + + +#include "pipe/p_shader_tokens.h" +#include +#include +#include + +#include +#include +#include +#include +#include + +using namespace llvm; + +StorageSoa::StorageSoa(llvm::BasicBlock *block, + llvm::Value *input, + llvm::Value *output, + llvm::Value *consts) +{ +} + +void StorageSoa::addImmediate(float *vec) +{ +} + +llvm::Value *StorageSoa::addrElement(int idx) const +{ + return 0; +} + +std::vector StorageSoa::inputElement(int idx, int swizzle, + llvm::Value *indIdx) +{ + std::vector res(4); + + return res; +} + +std::vector StorageSoa::constElement(int idx, int swizzle, + llvm::Value *indIdx) +{ + std::vector res(4); + + return res; +} + +std::vector StorageSoa::outputElement(int idx, int swizzle, + llvm::Value *indIdx) +{ + std::vector res(4); + + return res; +} + +std::vector StorageSoa::tempElement(int idx, int swizzle, + llvm::Value *indIdx) +{ + std::vector res(4); + + return res; +} + +std::vector StorageSoa::immediateElement(int idx, int swizzle) +{ + std::vector res(4); + + return res; +} + +llvm::Value * StorageSoa::extractIndex(llvm::Value *vec) +{ + return 0; +} + +void StorageSoa::storeOutput(int dstIdx, const std::vector &val, + int mask) +{ +} + +void StorageSoa::storeTemp(int idx, const std::vector &val, + int mask) +{ +} + +void StorageSoa::storeAddress(int idx, const std::vector &val, + int mask) +{ +} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h new file mode 100644 index 0000000000..551b0b9734 --- /dev/null +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -0,0 +1,71 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef STORAGESOA_H +#define STORAGESOA_H + +#include + +namespace llvm { + class BasicBlock; + class Constant; + class ConstantInt; + class LoadInst; + class Value; + class VectorType; +} + +class StorageSoa +{ +public: + StorageSoa(llvm::BasicBlock *block, + llvm::Value *input, + llvm::Value *output, + llvm::Value *consts); + + void addImmediate(float *vec); + + llvm::Value * addrElement(int idx) const; + + std::vector inputElement(int idx, int swizzle, llvm::Value *indIdx =0); + std::vector constElement(int idx, int swizzle, llvm::Value *indIdx =0); + std::vector outputElement(int idx, int swizzle, llvm::Value *indIdx =0); + std::vector tempElement(int idx, int swizzle, llvm::Value *indIdx =0); + std::vector immediateElement(int idx, int swizzle); + + llvm::Value *extractIndex(llvm::Value *vec); + + void storeOutput(int dstIdx, const std::vector &val, + int mask); + void storeTemp(int idx, const std::vector &val, + int mask); + void storeAddress(int idx, const std::vector &val, + int mask); + +}; + +#endif diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index eb9e1196f1..bc4df61071 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -5,6 +5,8 @@ #include "storage.h" #include "instructions.h" +#include "storagesoa.h" +#include "instructionssoa.h" #include "pipe/p_shader_tokens.h" @@ -112,6 +114,14 @@ translate_declaration(struct gallivm_ir *prog, } } +static void +translate_declarationir(struct gallivm_ir *, + llvm::Module *, + StorageSoa *, + struct tgsi_full_declaration *, + struct tgsi_full_declaration *) +{ +} static void translate_immediate(Storage *storage, @@ -120,29 +130,56 @@ translate_immediate(Storage *storage, float vec[4]; int i; for (i = 0; i < imm->Immediate.Size - 1; ++i) { - switch( imm->Immediate.DataType ) { + switch (imm->Immediate.DataType) { case TGSI_IMM_FLOAT32: vec[i] = imm->u.ImmediateFloat32[i].Float; break; default: - assert( 0 ); + assert(0); } } storage->addImmediate(vec); } -static inline llvm::Value * -swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, - Storage *storage) + +static void +translate_immediateir(StorageSoa *storage, + struct tgsi_full_immediate *imm) +{ + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert(0); + } + } + storage->addImmediate(vec); +} + +static inline int +swizzleInt(struct tgsi_full_src_register *src) { int swizzle = 0; int start = 1000; - const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + - TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; + for (int k = 0; k < 4; ++k) { swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; start /= 10; } + return swizzle; +} + +static inline llvm::Value * +swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, + Storage *storage) +{ + int swizzle = swizzleInt(src); + const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + + TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; if (swizzle != NO_SWIZZLE) { /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ val = storage->shuffleVector(val, swizzle); @@ -617,6 +654,392 @@ translate_instruction(llvm::Module *module, } } + +static void +translate_instructionir(llvm::Module *module, + StorageSoa *storage, + InstructionsSoa *instr, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi, + unsigned instno) +{ + std::vector< std::vector > inputs(inst->Instruction.NumSrcRegs); + + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + std::vector val; + llvm::Value *indIdx = 0; + int swizzle = swizzleInt(src); + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + indIdx = storage->extractIndex(indIdx); + } + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->constElement(src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->inputElement(src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->tempElement(src->SrcRegister.Index, swizzle); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->outputElement(src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->immediateElement(src->SrcRegister.Index, swizzle); + } else { + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); + return; + } + + inputs[i] = val; + } + + std::vector out(4); + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: { + } + break; + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } + break; + case TGSI_OPCODE_LIT: { + } + break; + case TGSI_OPCODE_RCP: { + } + break; + case TGSI_OPCODE_RSQ: { + } + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP3: { + } + break; + case TGSI_OPCODE_DP4: { + } + break; + case TGSI_OPCODE_DST: { + } + break; + case TGSI_OPCODE_MIN: { + } + break; + case TGSI_OPCODE_MAX: { + } + break; + case TGSI_OPCODE_SLT: { + } + break; + case TGSI_OPCODE_SGE: { + } + break; + case TGSI_OPCODE_MAD: { + } + break; + case TGSI_OPCODE_SUB: { + } + break; + case TGSI_OPCODE_LERP: { + } + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: { + } + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: { + } + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: { + } + break; + case TGSI_OPCODE_LOGBASE2: { + } + break; + case TGSI_OPCODE_POWER: { + } + break; + case TGSI_OPCODE_CROSSPRODUCT: { + } + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: { + } + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: { + } + break; + case TGSI_OPCODE_COS: { + } + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: { + } + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: { + } + break; + case TGSI_OPCODE_SIN: { + } + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: { + } + break; + case TGSI_OPCODE_RET: { + } + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: { + } + break; + case TGSI_OPCODE_SCS: { + } + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: { + } + break; + case TGSI_OPCODE_IF: { + } + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: { + } + break; + case TGSI_OPCODE_ENDIF: { + } + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: { + } + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: { + } + break; + case TGSI_OPCODE_BGNSUB: { + } + break; + case TGSI_OPCODE_ENDLOOP2: { + } + break; + case TGSI_OPCODE_ENDSUB: { + } + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + instr->end(); + return; + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + if (!out[0]) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } + + /* store results */ + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->storeOutput(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->storeTemp(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->storeAddress(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); + } + } +} + llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) { @@ -680,3 +1103,72 @@ tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) ir->num_consts = storage.numConsts(); return mod; } + +llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, + const struct tgsi_token *tokens) +{ + llvm::Module *mod = createBaseShader(); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned instno = 0; + Function* shader = mod->getFunction("execute_shader"); + std::ostringstream stream; + if (ir->type == GALLIVM_VS) { + stream << "vs_shader"; + } else { + stream << "fs_shader"; + } + stream << ir->id; + std::string func_name = stream.str(); + shader->setName(func_name.c_str()); + + Function::arg_iterator args = shader->arg_begin(); + Value *input = args++; + input->setName("input"); + Value *output = args++; + output->setName("output"); + Value *consts = args++; + consts->setName("consts"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + + tgsi_parse_init(&parse, tokens); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + StorageSoa storage(label_entry, input, output, consts); + InstructionsSoa instr(mod, shader, label_entry, &storage); + + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declarationir(ir, mod, &storage, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediateir(&storage, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + translate_instructionir(mod, &storage, &instr, + &parse.FullToken.FullInstruction, + &fi, instno); + ++instno; + break; + + default: + assert(0); + } + } + + tgsi_parse_free(&parse); + + return mod; +} diff --git a/src/mesa/pipe/llvm/tgsitollvm.h b/src/mesa/pipe/llvm/tgsitollvm.h index 073ffb5749..7ada04d629 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.h +++ b/src/mesa/pipe/llvm/tgsitollvm.h @@ -13,4 +13,8 @@ struct tgsi_token; llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens); + +llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, + const struct tgsi_token *tokens); + #endif -- cgit v1.2.3 From 084e6c92e088b443eacf6bf649d4a6ce6ede2386 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 11 Feb 2008 10:47:28 -0500 Subject: remove seperate llvm vs entry points they're not necessary anymore. we use the same paths as sse and tgsi code --- src/mesa/pipe/draw/draw_vertex_shader_llvm.c | 194 --------------------------- src/mesa/sources | 1 - 2 files changed, 195 deletions(-) delete mode 100644 src/mesa/pipe/draw/draw_vertex_shader_llvm.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c b/src/mesa/pipe/draw/draw_vertex_shader_llvm.c deleted file mode 100644 index 63551c993e..0000000000 --- a/src/mesa/pipe/draw/draw_vertex_shader_llvm.c +++ /dev/null @@ -1,194 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#include "pipe/p_util.h" -#include "draw_private.h" -#include "draw_context.h" - -#ifdef MESA_LLVM - -#include "pipe/llvm/gallivm.h" -#include "pipe/p_shader_tokens.h" - -#define DBG 0 - -static INLINE void -fetch_attrib4(const void *ptr, enum pipe_format format, float attrib[4]) -{ - /* defaults */ - attrib[1] = 0.0; - attrib[2] = 0.0; - attrib[3] = 1.0; - switch (format) { - case PIPE_FORMAT_R32G32B32A32_FLOAT: - attrib[3] = ((float *) ptr)[3]; - /* fall-through */ - case PIPE_FORMAT_R32G32B32_FLOAT: - attrib[2] = ((float *) ptr)[2]; - /* fall-through */ - case PIPE_FORMAT_R32G32_FLOAT: - attrib[1] = ((float *) ptr)[1]; - /* fall-through */ - case PIPE_FORMAT_R32_FLOAT: - attrib[0] = ((float *) ptr)[0]; - break; - default: - assert(0); - } -} - - -/** - * Fetch vertex attributes for 'count' vertices. - */ -static INLINE -void vertex_fetch(struct draw_context *draw, - const unsigned elt, - float (*inputs)[4]) -{ - uint attr; - - /* loop over vertex attributes (vertex shader inputs) */ - for (attr = 0; attr < draw->vertex_shader->state->num_inputs; attr++) { - - unsigned buf = draw->vertex_element[attr].vertex_buffer_index; - const void *src - = (const void *) ((const ubyte *) draw->user.vbuffer[buf] - + draw->vertex_buffer[buf].buffer_offset - + draw->vertex_element[attr].src_offset - + elt * draw->vertex_buffer[buf].pitch); - fetch_attrib4(src, draw->vertex_element[attr].src_format, inputs[attr]); - } -} - -static INLINE unsigned -compute_clipmask(const float *clip, const float (*plane)[4], unsigned nr) -{ - unsigned mask = 0; - unsigned i; - - for (i = 0; i < nr; i++) { - if (dot4(clip, plane[i]) < 0) - mask |= (1<user.constants; - struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - /* fetch the inputs */ - for (i = 0; i < draw->vs.queue_nr; ++i) { - unsigned elt = draw->vs.queue[i].elt; - dests[i] = draw->vs.queue[i].dest; - vertex_fetch(draw, elt, inputs[i]); - } - - /* batch execute the shaders on all the vertices */ - gallivm_prog_exec(prog, inputs, outputs, consts, - draw->vs.queue_nr, - draw->vertex_shader->state->num_inputs, - draw->vertex_shader->state->num_outputs); - - - /* store machine results */ - for (int i = 0; i < draw->vs.queue_nr; ++i) { - unsigned slot; - float x, y, z, w; - struct vertex_header *vOut = draw->vs.queue[i].dest; - float (*dests)[4] = outputs[i]; - - /* Handle attr[0] (position) specially: - * - * XXX: Computing the clipmask should be done in the vertex - * program as a set of DP4 instructions appended to the - * user-provided code. - */ - x = vOut->clip[0] = dests[0][0]; - y = vOut->clip[1] = dests[0][1]; - z = vOut->clip[2] = dests[0][2]; - w = vOut->clip[3] = dests[0][3]; -#if DBG - debug_printf("output %d: %f %f %f %f\n", 0, x, y, z, w); -#endif - - vOut->clipmask = compute_clipmask(vOut->clip, draw->plane, draw->nr_planes); - vOut->edgeflag = 1; - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut->data[0][0] = x * scale[0] + trans[0]; - vOut->data[0][1] = y * scale[1] + trans[1]; - vOut->data[0][2] = z * scale[2] + trans[2]; - vOut->data[0][3] = w; - - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut->data[slot][0] = dests[slot][0]; - vOut->data[slot][1] = dests[slot][1]; - vOut->data[slot][2] = dests[slot][2]; - vOut->data[slot][3] = dests[slot][3]; - -#if DBG - debug_printf("output %d: %f %f %f %f\n", slot, - vOut->data[slot][0], - vOut->data[slot][1], - vOut->data[slot][2], - vOut->data[slot][3]); -#endif - } - } /* loop over vertices */ - - draw->vs.queue_nr = 0; -} - -#endif /* MESA_LLVM */ diff --git a/src/mesa/sources b/src/mesa/sources index 84492c91ac..96ae3dbca0 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -174,7 +174,6 @@ DRAW_SOURCES = \ pipe/draw/draw_vertex_cache.c \ pipe/draw/draw_vertex_fetch.c \ pipe/draw/draw_vertex_shader.c \ - pipe/draw/draw_vertex_shader_llvm.c \ pipe/draw/draw_vf.c \ pipe/draw/draw_vf_generic.c \ pipe/draw/draw_vf_sse.c \ -- cgit v1.2.3 From aa9d9d9c40ae34d61b113ffc54bffd702138f72f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 16:35:44 -0700 Subject: checkpoint- remove dependencies on gl_texture_format to make code re-usable by state tracker --- src/mesa/main/mipmap.c | 424 +++++++++++++++++++++++++++++++------------------ 1 file changed, 266 insertions(+), 158 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index 1e61829e8f..013dc3752e 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -36,27 +36,205 @@ +static GLint +bytes_per_pixel(GLenum datatype, GLuint comps) +{ + GLint b = _mesa_sizeof_packed_type(datatype); + assert(b >= 0); + return b * comps; +} + + +static void +mesa_format_to_type_and_comps(const struct gl_texture_format *format, + GLenum *datatype, GLuint *comps) +{ + switch (format->MesaFormat) { + case MESA_FORMAT_RGBA8888: + case MESA_FORMAT_RGBA8888_REV: + case MESA_FORMAT_ARGB8888: + case MESA_FORMAT_ARGB8888_REV: + *datatype = CHAN_TYPE; + *comps = 4; + return; + case MESA_FORMAT_RGB888: + case MESA_FORMAT_BGR888: + *datatype = GL_UNSIGNED_BYTE; + *comps = 3; + return; + case MESA_FORMAT_RGB565: + case MESA_FORMAT_RGB565_REV: + *datatype = GL_UNSIGNED_SHORT_5_6_5; + *comps = 3; + return; + + case MESA_FORMAT_ARGB4444: + case MESA_FORMAT_ARGB4444_REV: + *datatype = GL_UNSIGNED_SHORT_4_4_4_4; + *comps = 4; + return; + + case MESA_FORMAT_ARGB1555: + case MESA_FORMAT_ARGB1555_REV: + *datatype = GL_UNSIGNED_SHORT_1_5_5_5_REV; + *comps = 3; + return; + + case MESA_FORMAT_AL88: + case MESA_FORMAT_AL88_REV: + *datatype = GL_UNSIGNED_BYTE; + *comps = 2; + return; + case MESA_FORMAT_RGB332: + *datatype = GL_UNSIGNED_BYTE_3_3_2; + *comps = 3; + return; + + case MESA_FORMAT_A8: + case MESA_FORMAT_L8: + case MESA_FORMAT_I8: + case MESA_FORMAT_CI8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 1; + return; + + case MESA_FORMAT_YCBCR: + case MESA_FORMAT_YCBCR_REV: + *datatype = GL_UNSIGNED_SHORT; + *comps = 2; + return; + + case MESA_FORMAT_Z24_S8: + *datatype = GL_UNSIGNED_INT; + *comps = 1; /* XXX OK? */ + return; + + case MESA_FORMAT_Z16: + *datatype = GL_UNSIGNED_SHORT; + *comps = 1; + return; + + case MESA_FORMAT_Z32: + *datatype = GL_UNSIGNED_INT; + *comps = 1; + return; + + case MESA_FORMAT_SRGB8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 3; + return; + case MESA_FORMAT_SRGBA8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 4; + return; + case MESA_FORMAT_SL8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 1; + return; + case MESA_FORMAT_SLA8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 2; + return; + + case MESA_FORMAT_RGB_FXT1: + case MESA_FORMAT_RGBA_FXT1: + case MESA_FORMAT_RGB_DXT1: + case MESA_FORMAT_RGBA_DXT1: + case MESA_FORMAT_RGBA_DXT3: + case MESA_FORMAT_RGBA_DXT5: + /* XXX generate error instead? */ + *datatype = GL_UNSIGNED_BYTE; + *comps = 0; + return; + + case MESA_FORMAT_RGBA: + *datatype = CHAN_TYPE; + *comps = 4; + return; + case MESA_FORMAT_RGB: + *datatype = CHAN_TYPE; + *comps = 3; + return; + case MESA_FORMAT_LUMINANCE_ALPHA: + *datatype = CHAN_TYPE; + *comps = 2; + return; + case MESA_FORMAT_ALPHA: + case MESA_FORMAT_LUMINANCE: + case MESA_FORMAT_INTENSITY: + *datatype = CHAN_TYPE; + *comps = 1; + return; + + case MESA_FORMAT_RGBA_FLOAT32: + *datatype = GL_FLOAT; + *comps = 4; + return; + case MESA_FORMAT_RGBA_FLOAT16: + *datatype = GL_HALF_FLOAT_ARB; + *comps = 4; + return; + case MESA_FORMAT_RGB_FLOAT32: + *datatype = GL_FLOAT; + *comps = 3; + return; + case MESA_FORMAT_RGB_FLOAT16: + *datatype = GL_HALF_FLOAT_ARB; + *comps = 3; + return; + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32: + *datatype = GL_FLOAT; + *comps = 2; + return; + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16: + *datatype = GL_HALF_FLOAT_ARB; + *comps = 2; + return; + case MESA_FORMAT_ALPHA_FLOAT32: + case MESA_FORMAT_LUMINANCE_FLOAT32: + case MESA_FORMAT_INTENSITY_FLOAT32: + *datatype = GL_FLOAT; + *comps = 1; + return; + case MESA_FORMAT_ALPHA_FLOAT16: + case MESA_FORMAT_LUMINANCE_FLOAT16: + case MESA_FORMAT_INTENSITY_FLOAT16: + *datatype = GL_HALF_FLOAT_ARB; + *comps = 1; + return; + + default: + _mesa_problem(NULL, "bad texture format in mesa_format_to_type_and_comps"); + *datatype = 0; + *comps = 1; + } +} + + /** * Average together two rows of a source image to produce a single new * row in the dest image. It's legal for the two source rows to point * to the same data. The source width must be equal to either the * dest width or two times the dest width. + * \param datatype GL_UNSIGNED_BYTE, GL_UNSIGNED_SHORT, GL_FLOAT, etc. + * \param comps number of components per pixel (1..4) */ static void -do_row(const struct gl_texture_format *format, GLint srcWidth, +do_row(GLenum datatype, GLuint comps, GLint srcWidth, const GLvoid *srcRowA, const GLvoid *srcRowB, GLint dstWidth, GLvoid *dstRow) { const GLuint k0 = (srcWidth == dstWidth) ? 0 : 1; const GLuint colStride = (srcWidth == dstWidth) ? 1 : 2; + ASSERT(comps >= 1); + ASSERT(comps <= 4); + /* This assertion is no longer valid with non-power-of-2 textures assert(srcWidth == dstWidth || srcWidth == 2 * dstWidth); */ - switch (format->MesaFormat) { - case MESA_FORMAT_RGBA: - { + if (datatype == CHAN_TYPE && comps == 4) { GLuint i, j, k; const GLchan (*rowA)[4] = (const GLchan (*)[4]) srcRowA; const GLchan (*rowB)[4] = (const GLchan (*)[4]) srcRowB; @@ -72,10 +250,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; } - } - return; - case MESA_FORMAT_RGB: - { + } + else if (datatype == CHAN_TYPE && comps == 3) { GLuint i, j, k; const GLchan (*rowA)[3] = (const GLchan (*)[3]) srcRowA; const GLchan (*rowB)[3] = (const GLchan (*)[3]) srcRowB; @@ -89,12 +265,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; } - } - return; - case MESA_FORMAT_ALPHA: - case MESA_FORMAT_LUMINANCE: - case MESA_FORMAT_INTENSITY: - { + } + else if (datatype == CHAN_TYPE && comps == 1) { GLuint i, j, k; const GLchan *rowA = (const GLchan *) srcRowA; const GLchan *rowB = (const GLchan *) srcRowB; @@ -103,10 +275,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, i++, j += colStride, k += colStride) { dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4; } - } - return; - case MESA_FORMAT_LUMINANCE_ALPHA: - { + } + else if (datatype == CHAN_TYPE && comps == 2) { GLuint i, j, k; const GLchan (*rowA)[2] = (const GLchan (*)[2]) srcRowA; const GLchan (*rowB)[2] = (const GLchan (*)[2]) srcRowB; @@ -118,10 +288,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; } - } - return; - case MESA_FORMAT_Z32: - { + } + else if (datatype == GL_UNSIGNED_INT && comps == 1) { GLuint i, j, k; const GLuint *rowA = (const GLuint *) srcRowA; const GLuint *rowB = (const GLuint *) srcRowB; @@ -130,10 +298,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, i++, j += colStride, k += colStride) { dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; } - } - return; - case MESA_FORMAT_Z16: - { + } + else if (datatype == GL_UNSIGNED_SHORT && comps == 1) { GLuint i, j, k; const GLushort *rowA = (const GLushort *) srcRowA; const GLushort *rowB = (const GLushort *) srcRowB; @@ -142,17 +308,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, i++, j += colStride, k += colStride) { dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4; } - } - return; - /* Begin hardware formats */ - case MESA_FORMAT_RGBA8888: - case MESA_FORMAT_RGBA8888_REV: - case MESA_FORMAT_ARGB8888: - case MESA_FORMAT_ARGB8888_REV: -#if FEATURE_EXT_texture_sRGB - case MESA_FORMAT_SRGBA8: -#endif - { + } + else if (datatype == GL_UNSIGNED_BYTE && comps == 4) { GLuint i, j, k; const GLubyte (*rowA)[4] = (const GLubyte (*)[4]) srcRowA; const GLubyte (*rowB)[4] = (const GLubyte (*)[4]) srcRowB; @@ -168,14 +325,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; } - } - return; - case MESA_FORMAT_RGB888: - case MESA_FORMAT_BGR888: -#if FEATURE_EXT_texture_sRGB - case MESA_FORMAT_SRGB8: -#endif - { + } + else if (datatype == GL_UNSIGNED_BYTE && comps == 3) { GLuint i, j, k; const GLubyte (*rowA)[3] = (const GLubyte (*)[3]) srcRowA; const GLubyte (*rowB)[3] = (const GLubyte (*)[3]) srcRowB; @@ -189,11 +340,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; } - } - return; - case MESA_FORMAT_RGB565: - case MESA_FORMAT_RGB565_REV: - { + } + else if (datatype == GL_UNSIGNED_SHORT_5_6_5 && comps == 3) { GLuint i, j, k; const GLushort *rowA = (const GLushort *) srcRowA; const GLushort *rowB = (const GLushort *) srcRowB; @@ -217,11 +365,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; dst[i] = (blue << 11) | (green << 5) | red; } - } - return; - case MESA_FORMAT_ARGB4444: - case MESA_FORMAT_ARGB4444_REV: - { + } + else if (datatype == GL_UNSIGNED_SHORT_4_4_4_4 && comps == 4) { GLuint i, j, k; const GLushort *rowA = (const GLushort *) srcRowA; const GLushort *rowB = (const GLushort *) srcRowB; @@ -250,11 +395,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, const GLint alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2; dst[i] = (alpha << 12) | (blue << 8) | (green << 4) | red; } - } - return; - case MESA_FORMAT_ARGB1555: - case MESA_FORMAT_ARGB1555_REV: /* XXX broken? */ - { + } + else if (datatype == GL_UNSIGNED_SHORT_1_5_5_5_REV && comps == 4) { GLuint i, j, k; const GLushort *rowA = (const GLushort *) srcRowA; const GLushort *rowB = (const GLushort *) srcRowB; @@ -283,14 +425,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, const GLint alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2; dst[i] = (alpha << 15) | (blue << 10) | (green << 5) | red; } - } - return; - case MESA_FORMAT_AL88: - case MESA_FORMAT_AL88_REV: -#if FEATURE_EXT_texture_sRGB - case MESA_FORMAT_SLA8: -#endif - { + } + else if (datatype == GL_UNSIGNED_BYTE && comps == 2) { GLuint i, j, k; const GLubyte (*rowA)[2] = (const GLubyte (*)[2]) srcRowA; const GLubyte (*rowB)[2] = (const GLubyte (*)[2]) srcRowB; @@ -302,10 +438,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) >> 2; } - } - return; - case MESA_FORMAT_RGB332: - { + } + else if (datatype == GL_UNSIGNED_BYTE_3_3_2 && comps == 3) { GLuint i, j, k; const GLubyte *rowA = (const GLubyte *) srcRowA; const GLubyte *rowB = (const GLubyte *) srcRowB; @@ -329,16 +463,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; dst[i] = (blue << 5) | (green << 2) | red; } - } - return; - case MESA_FORMAT_A8: - case MESA_FORMAT_L8: - case MESA_FORMAT_I8: - case MESA_FORMAT_CI8: -#if FEATURE_EXT_texture_sRGB - case MESA_FORMAT_SL8: -#endif - { + } + else if (datatype == GL_UNSIGNED_BYTE && comps == 1) { GLuint i, j, k; const GLubyte *rowA = (const GLubyte *) srcRowA; const GLubyte *rowB = (const GLubyte *) srcRowB; @@ -347,10 +473,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, i++, j += colStride, k += colStride) { dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) >> 2; } - } - return; - case MESA_FORMAT_RGBA_FLOAT32: - { + } + else if (datatype == GL_FLOAT && comps == 4) { GLuint i, j, k; const GLfloat (*rowA)[4] = (const GLfloat (*)[4]) srcRowA; const GLfloat (*rowB)[4] = (const GLfloat (*)[4]) srcRowB; @@ -366,10 +490,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) * 0.25F; } - } - return; - case MESA_FORMAT_RGBA_FLOAT16: - { + } + else if (datatype == GL_HALF_FLOAT_ARB && comps == 4) { GLuint i, j, k, comp; const GLhalfARB (*rowA)[4] = (const GLhalfARB (*)[4]) srcRowA; const GLhalfARB (*rowB)[4] = (const GLhalfARB (*)[4]) srcRowB; @@ -385,10 +507,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); } } - } - return; - case MESA_FORMAT_RGB_FLOAT32: - { + } + else if (datatype == GL_FLOAT && comps == 3) { GLuint i, j, k; const GLfloat (*rowA)[3] = (const GLfloat (*)[3]) srcRowA; const GLfloat (*rowB)[3] = (const GLfloat (*)[3]) srcRowB; @@ -402,10 +522,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) * 0.25F; } - } - return; - case MESA_FORMAT_RGB_FLOAT16: - { + } + else if (datatype == GL_HALF_FLOAT_ARB && comps == 3) { GLuint i, j, k, comp; const GLhalfARB (*rowA)[3] = (const GLhalfARB (*)[3]) srcRowA; const GLhalfARB (*rowB)[3] = (const GLhalfARB (*)[3]) srcRowB; @@ -421,10 +539,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); } } - } - return; - case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32: - { + } + else if (datatype == GL_FLOAT && comps == 2) { GLuint i, j, k; const GLfloat (*rowA)[2] = (const GLfloat (*)[2]) srcRowA; const GLfloat (*rowB)[2] = (const GLfloat (*)[2]) srcRowB; @@ -436,10 +552,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) * 0.25F; } - } - return; - case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16: - { + } + else if (datatype == GL_HALF_FLOAT_ARB && comps == 2) { GLuint i, j, k, comp; const GLhalfARB (*rowA)[2] = (const GLhalfARB (*)[2]) srcRowA; const GLhalfARB (*rowB)[2] = (const GLhalfARB (*)[2]) srcRowB; @@ -455,12 +569,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); } } - } - return; - case MESA_FORMAT_ALPHA_FLOAT32: - case MESA_FORMAT_LUMINANCE_FLOAT32: - case MESA_FORMAT_INTENSITY_FLOAT32: - { + } + else if (datatype == GL_FLOAT && comps == 1) { GLuint i, j, k; const GLfloat *rowA = (const GLfloat *) srcRowA; const GLfloat *rowB = (const GLfloat *) srcRowB; @@ -469,12 +579,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, i++, j += colStride, k += colStride) { dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) * 0.25F; } - } - return; - case MESA_FORMAT_ALPHA_FLOAT16: - case MESA_FORMAT_LUMINANCE_FLOAT16: - case MESA_FORMAT_INTENSITY_FLOAT16: - { + } + else if (datatype == GL_HALF_FLOAT_ARB && comps == 1) { GLuint i, j, k; const GLhalfARB *rowA = (const GLhalfARB *) srcRowA; const GLhalfARB *rowB = (const GLhalfARB *) srcRowB; @@ -488,10 +594,8 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, bk = _mesa_half_to_float(rowB[k]); dst[i] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); } - } - return; - - default: + } + else { _mesa_problem(NULL, "bad format in do_row()"); } } @@ -504,11 +608,11 @@ do_row(const struct gl_texture_format *format, GLint srcWidth, */ static void -make_1d_mipmap(const struct gl_texture_format *format, GLint border, +make_1d_mipmap(GLenum datatype, GLuint comps, GLint border, GLint srcWidth, const GLubyte *srcPtr, GLint dstWidth, GLubyte *dstPtr) { - const GLint bpt = format->TexelBytes; + const GLint bpt = bytes_per_pixel(datatype, comps); const GLubyte *src; GLubyte *dst; @@ -517,7 +621,7 @@ make_1d_mipmap(const struct gl_texture_format *format, GLint border, dst = dstPtr + border * bpt; /* we just duplicate the input row, kind of hack, saves code */ - do_row(format, srcWidth - 2 * border, src, src, + do_row(datatype, comps, srcWidth - 2 * border, src, src, dstWidth - 2 * border, dst); if (border) { @@ -535,11 +639,11 @@ make_1d_mipmap(const struct gl_texture_format *format, GLint border, * XXX need to use the tex image's row stride! */ static void -make_2d_mipmap(const struct gl_texture_format *format, GLint border, +make_2d_mipmap(GLenum datatype, GLuint comps, GLint border, GLint srcWidth, GLint srcHeight, const GLubyte *srcPtr, GLint dstWidth, GLint dstHeight, GLubyte *dstPtr) { - const GLint bpt = format->TexelBytes; + const GLint bpt = bytes_per_pixel(datatype, comps); const GLint srcWidthNB = srcWidth - 2 * border; /* sizes w/out border */ const GLint dstWidthNB = dstWidth - 2 * border; const GLint dstHeightNB = dstHeight - 2 * border; @@ -558,7 +662,7 @@ make_2d_mipmap(const struct gl_texture_format *format, GLint border, dst = dstPtr + border * ((dstWidth + 1) * bpt); for (row = 0; row < dstHeightNB; row++) { - do_row(format, srcWidthNB, srcA, srcB, + do_row(datatype, comps, srcWidthNB, srcA, srcB, dstWidthNB, dst); srcA += 2 * srcRowStride; srcB += 2 * srcRowStride; @@ -580,12 +684,12 @@ make_2d_mipmap(const struct gl_texture_format *format, GLint border, MEMCPY(dstPtr + (dstWidth * dstHeight - 1) * bpt, srcPtr + (srcWidth * srcHeight - 1) * bpt, bpt); /* lower border */ - do_row(format, srcWidthNB, + do_row(datatype, comps, srcWidthNB, srcPtr + bpt, srcPtr + bpt, dstWidthNB, dstPtr + bpt); /* upper border */ - do_row(format, srcWidthNB, + do_row(datatype, comps, srcWidthNB, srcPtr + (srcWidth * (srcHeight - 1) + 1) * bpt, srcPtr + (srcWidth * (srcHeight - 1) + 1) * bpt, dstWidthNB, @@ -603,11 +707,11 @@ make_2d_mipmap(const struct gl_texture_format *format, GLint border, else { /* average two src pixels each dest pixel */ for (row = 0; row < dstHeightNB; row += 2) { - do_row(format, 1, + do_row(datatype, comps, 1, srcPtr + (srcWidth * (row * 2 + 1)) * bpt, srcPtr + (srcWidth * (row * 2 + 2)) * bpt, 1, dstPtr + (dstWidth * row + 1) * bpt); - do_row(format, 1, + do_row(datatype, comps, 1, srcPtr + (srcWidth * (row * 2 + 1) + srcWidth - 1) * bpt, srcPtr + (srcWidth * (row * 2 + 2) + srcWidth - 1) * bpt, 1, dstPtr + (dstWidth * row + 1 + dstWidth - 1) * bpt); @@ -618,13 +722,13 @@ make_2d_mipmap(const struct gl_texture_format *format, GLint border, static void -make_3d_mipmap(const struct gl_texture_format *format, GLint border, +make_3d_mipmap(GLenum datatype, GLuint comps, GLint border, GLint srcWidth, GLint srcHeight, GLint srcDepth, const GLubyte *srcPtr, GLint dstWidth, GLint dstHeight, GLint dstDepth, GLubyte *dstPtr) { - const GLint bpt = format->TexelBytes; + const GLint bpt = bytes_per_pixel(datatype, comps); const GLint srcWidthNB = srcWidth - 2 * border; /* sizes w/out border */ const GLint srcDepthNB = srcDepth - 2 * border; const GLint dstWidthNB = dstWidth - 2 * border; @@ -694,13 +798,13 @@ make_3d_mipmap(const struct gl_texture_format *format, GLint border, for (row = 0; row < dstHeightNB; row++) { /* Average together two rows from first src image */ - do_row(format, srcWidthNB, srcImgARowA, srcImgARowB, + do_row(datatype, comps, srcWidthNB, srcImgARowA, srcImgARowB, srcWidthNB, tmpRowA); /* Average together two rows from second src image */ - do_row(format, srcWidthNB, srcImgBRowA, srcImgBRowB, + do_row(datatype, comps, srcWidthNB, srcImgBRowA, srcImgBRowB, srcWidthNB, tmpRowB); /* Average together the temp rows to make the final row */ - do_row(format, srcWidthNB, tmpRowA, tmpRowB, + do_row(datatype, comps, srcWidthNB, tmpRowA, tmpRowB, dstWidthNB, dstImgRow); /* advance to next rows */ srcImgARowA += bytesPerSrcRow + srcRowOffset; @@ -717,10 +821,10 @@ make_3d_mipmap(const struct gl_texture_format *format, GLint border, /* Luckily we can leverage the make_2d_mipmap() function here! */ if (border > 0) { /* do front border image */ - make_2d_mipmap(format, 1, srcWidth, srcHeight, srcPtr, + make_2d_mipmap(datatype, comps, 1, srcWidth, srcHeight, srcPtr, dstWidth, dstHeight, dstPtr); /* do back border image */ - make_2d_mipmap(format, 1, srcWidth, srcHeight, + make_2d_mipmap(datatype, comps, 1, srcWidth, srcHeight, srcPtr + bytesPerSrcImage * (srcDepth - 1), dstWidth, dstHeight, dstPtr + bytesPerDstImage * (dstDepth - 1)); @@ -768,28 +872,28 @@ make_3d_mipmap(const struct gl_texture_format *format, GLint border, /* do border along [img][row=0][col=0] */ src = srcPtr + (img * 2 + 1) * bytesPerSrcImage; dst = dstPtr + (img + 1) * bytesPerDstImage; - do_row(format, 1, src, src + srcImageOffset, 1, dst); + do_row(datatype, comps, 1, src, src + srcImageOffset, 1, dst); /* do border along [img][row=dstHeight-1][col=0] */ src = srcPtr + (img * 2 + 1) * bytesPerSrcImage + (srcHeight - 1) * bytesPerSrcRow; dst = dstPtr + (img + 1) * bytesPerDstImage + (dstHeight - 1) * bytesPerDstRow; - do_row(format, 1, src, src + srcImageOffset, 1, dst); + do_row(datatype, comps, 1, src, src + srcImageOffset, 1, dst); /* do border along [img][row=0][col=dstWidth-1] */ src = srcPtr + (img * 2 + 1) * bytesPerSrcImage + (srcWidth - 1) * bpt; dst = dstPtr + (img + 1) * bytesPerDstImage + (dstWidth - 1) * bpt; - do_row(format, 1, src, src + srcImageOffset, 1, dst); + do_row(datatype, comps, 1, src, src + srcImageOffset, 1, dst); /* do border along [img][row=dstHeight-1][col=dstWidth-1] */ src = srcPtr + (img * 2 + 1) * bytesPerSrcImage + (bytesPerSrcImage - bpt); dst = dstPtr + (img + 1) * bytesPerDstImage + (bytesPerDstImage - bpt); - do_row(format, 1, src, src + srcImageOffset, 1, dst); + do_row(datatype, comps, 1, src, src + srcImageOffset, 1, dst); } } } @@ -797,11 +901,11 @@ make_3d_mipmap(const struct gl_texture_format *format, GLint border, static void -make_1d_stack_mipmap(const struct gl_texture_format *format, GLint border, +make_1d_stack_mipmap(GLenum datatype, GLuint comps, GLint border, GLint srcWidth, const GLubyte *srcPtr, GLint dstWidth, GLint dstHeight, GLubyte *dstPtr) { - const GLint bpt = format->TexelBytes; + const GLint bpt = bytes_per_pixel(datatype, comps); const GLint srcWidthNB = srcWidth - 2 * border; /* sizes w/out border */ const GLint dstWidthNB = dstWidth - 2 * border; const GLint dstHeightNB = dstHeight - 2 * border; @@ -816,7 +920,7 @@ make_1d_stack_mipmap(const struct gl_texture_format *format, GLint border, dst = dstPtr + border * ((dstWidth + 1) * bpt); for (row = 0; row < dstHeightNB; row++) { - do_row(format, srcWidthNB, src, src, + do_row(datatype, comps, srcWidthNB, src, src, dstWidthNB, dst); src += srcRowStride; dst += dstRowStride; @@ -839,12 +943,12 @@ make_1d_stack_mipmap(const struct gl_texture_format *format, GLint border, * and \c make_2d_mipmap. */ static void -make_2d_stack_mipmap(const struct gl_texture_format *format, GLint border, +make_2d_stack_mipmap(GLenum datatype, GLuint comps, GLint border, GLint srcWidth, GLint srcHeight, const GLubyte *srcPtr, GLint dstWidth, GLint dstHeight, GLint dstDepth, GLubyte *dstPtr) { - const GLint bpt = format->TexelBytes; + const GLint bpt = bytes_per_pixel(datatype, comps); const GLint srcWidthNB = srcWidth - 2 * border; /* sizes w/out border */ const GLint dstWidthNB = dstWidth - 2 * border; const GLint dstHeightNB = dstHeight - 2 * border; @@ -866,7 +970,7 @@ make_2d_stack_mipmap(const struct gl_texture_format *format, GLint border, for (layer = 0; layer < dstDepthNB; layer++) { for (row = 0; row < dstHeightNB; row++) { - do_row(format, srcWidthNB, srcA, srcB, + do_row(datatype, comps, srcWidthNB, srcA, srcB, dstWidthNB, dst); srcA += 2 * srcRowStride; srcB += 2 * srcRowStride; @@ -888,12 +992,12 @@ make_2d_stack_mipmap(const struct gl_texture_format *format, GLint border, MEMCPY(dstPtr + (dstWidth * dstHeight - 1) * bpt, srcPtr + (srcWidth * srcHeight - 1) * bpt, bpt); /* lower border */ - do_row(format, srcWidthNB, + do_row(datatype, comps, srcWidthNB, srcPtr + bpt, srcPtr + bpt, dstWidthNB, dstPtr + bpt); /* upper border */ - do_row(format, srcWidthNB, + do_row(datatype, comps, srcWidthNB, srcPtr + (srcWidth * (srcHeight - 1) + 1) * bpt, srcPtr + (srcWidth * (srcHeight - 1) + 1) * bpt, dstWidthNB, @@ -911,11 +1015,11 @@ make_2d_stack_mipmap(const struct gl_texture_format *format, GLint border, else { /* average two src pixels each dest pixel */ for (row = 0; row < dstHeightNB; row += 2) { - do_row(format, 1, + do_row(datatype, comps, 1, srcPtr + (srcWidth * (row * 2 + 1)) * bpt, srcPtr + (srcWidth * (row * 2 + 2)) * bpt, 1, dstPtr + (dstWidth * row + 1) * bpt); - do_row(format, 1, + do_row(datatype, comps, 1, srcPtr + (srcWidth * (row * 2 + 1) + srcWidth - 1) * bpt, srcPtr + (srcWidth * (row * 2 + 2) + srcWidth - 1) * bpt, 1, dstPtr + (dstWidth * row + 1 + dstWidth - 1) * bpt); @@ -940,6 +1044,8 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target, const GLubyte *srcData = NULL; GLubyte *dstData = NULL; GLint level, maxLevels; + GLenum datatype; + GLuint comps; ASSERT(texObj); /* XXX choose cube map face here??? */ @@ -1002,6 +1108,8 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target, convertFormat = srcImage->TexFormat; } + mesa_format_to_type_and_comps(convertFormat, &datatype, &comps); + for (level = texObj->BaseLevel; level < texObj->MaxLevel && level < maxLevels - 1; level++) { /* generate image[level+1] from image[level] */ @@ -1118,7 +1226,7 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target, */ switch (target) { case GL_TEXTURE_1D: - make_1d_mipmap(convertFormat, border, + make_1d_mipmap(datatype, comps, border, srcWidth, srcData, dstWidth, dstData); break; @@ -1129,22 +1237,22 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target, case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - make_2d_mipmap(convertFormat, border, + make_2d_mipmap(datatype, comps, border, srcWidth, srcHeight, srcData, dstWidth, dstHeight, dstData); break; case GL_TEXTURE_3D: - make_3d_mipmap(convertFormat, border, + make_3d_mipmap(datatype, comps, border, srcWidth, srcHeight, srcDepth, srcData, dstWidth, dstHeight, dstDepth, dstData); break; case GL_TEXTURE_1D_ARRAY_EXT: - make_1d_stack_mipmap(convertFormat, border, + make_1d_stack_mipmap(datatype, comps, border, srcWidth, srcData, dstWidth, dstHeight, dstData); break; case GL_TEXTURE_2D_ARRAY_EXT: - make_2d_stack_mipmap(convertFormat, border, + make_2d_stack_mipmap(datatype, comps, border, srcWidth, srcHeight, srcData, dstWidth, dstHeight, dstDepth, dstData); break; -- cgit v1.2.3 From ed6e72e8556a961ad31c80ae3c0582878ce64bf3 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 16:38:28 -0700 Subject: checkpoint- consolidation in do_row() --- src/mesa/main/mipmap.c | 42 ++++++++++++++++-------------------------- 1 file changed, 16 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index 013dc3752e..c5f1c5bcbe 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -234,11 +234,11 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, assert(srcWidth == dstWidth || srcWidth == 2 * dstWidth); */ - if (datatype == CHAN_TYPE && comps == 4) { + if (datatype == GL_UNSIGNED_SHORT && comps == 4) { GLuint i, j, k; - const GLchan (*rowA)[4] = (const GLchan (*)[4]) srcRowA; - const GLchan (*rowB)[4] = (const GLchan (*)[4]) srcRowB; - GLchan (*dst)[4] = (GLchan (*)[4]) dstRow; + const GLushort (*rowA)[4] = (const GLushort (*)[4]) srcRowA; + const GLushort (*rowB)[4] = (const GLushort (*)[4]) srcRowB; + GLushort (*dst)[4] = (GLushort (*)[4]) dstRow; for (i = j = 0, k = k0; i < (GLuint) dstWidth; i++, j += colStride, k += colStride) { dst[i][0] = (rowA[j][0] + rowA[k][0] + @@ -251,11 +251,11 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, rowB[j][3] + rowB[k][3]) / 4; } } - else if (datatype == CHAN_TYPE && comps == 3) { + else if (datatype == GL_UNSIGNED_SHORT && comps == 3) { GLuint i, j, k; - const GLchan (*rowA)[3] = (const GLchan (*)[3]) srcRowA; - const GLchan (*rowB)[3] = (const GLchan (*)[3]) srcRowB; - GLchan (*dst)[3] = (GLchan (*)[3]) dstRow; + const GLushort (*rowA)[3] = (const GLushort (*)[3]) srcRowA; + const GLushort (*rowB)[3] = (const GLushort (*)[3]) srcRowB; + GLushort (*dst)[3] = (GLushort (*)[3]) dstRow; for (i = j = 0, k = k0; i < (GLuint) dstWidth; i++, j += colStride, k += colStride) { dst[i][0] = (rowA[j][0] + rowA[k][0] + @@ -266,21 +266,21 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, rowB[j][2] + rowB[k][2]) / 4; } } - else if (datatype == CHAN_TYPE && comps == 1) { + else if (datatype == GL_UNSIGNED_SHORT && comps == 1) { GLuint i, j, k; - const GLchan *rowA = (const GLchan *) srcRowA; - const GLchan *rowB = (const GLchan *) srcRowB; - GLchan *dst = (GLchan *) dstRow; + const GLushort *rowA = (const GLushort *) srcRowA; + const GLushort *rowB = (const GLushort *) srcRowB; + GLushort *dst = (GLushort *) dstRow; for (i = j = 0, k = k0; i < (GLuint) dstWidth; i++, j += colStride, k += colStride) { dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4; } } - else if (datatype == CHAN_TYPE && comps == 2) { + else if (datatype == GL_UNSIGNED_SHORT && comps == 2) { GLuint i, j, k; - const GLchan (*rowA)[2] = (const GLchan (*)[2]) srcRowA; - const GLchan (*rowB)[2] = (const GLchan (*)[2]) srcRowB; - GLchan (*dst)[2] = (GLchan (*)[2]) dstRow; + const GLushort (*rowA)[2] = (const GLushort (*)[2]) srcRowA; + const GLushort (*rowB)[2] = (const GLushort (*)[2]) srcRowB; + GLushort (*dst)[2] = (GLushort (*)[2]) dstRow; for (i = j = 0, k = k0; i < (GLuint) dstWidth; i++, j += colStride, k += colStride) { dst[i][0] = (rowA[j][0] + rowA[k][0] + @@ -299,16 +299,6 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; } } - else if (datatype == GL_UNSIGNED_SHORT && comps == 1) { - GLuint i, j, k; - const GLushort *rowA = (const GLushort *) srcRowA; - const GLushort *rowB = (const GLushort *) srcRowB; - GLushort *dst = (GLushort *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4; - } - } else if (datatype == GL_UNSIGNED_BYTE && comps == 4) { GLuint i, j, k; const GLubyte (*rowA)[4] = (const GLubyte (*)[4]) srcRowA; -- cgit v1.2.3 From 30dff589a447529232016f3e792378fd8fa67820 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 16:40:39 -0700 Subject: re-indent do_row() --- src/mesa/main/mipmap.c | 594 ++++++++++++++++++++++++------------------------- 1 file changed, 288 insertions(+), 306 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index c5f1c5bcbe..22c7530e83 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -235,355 +235,337 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, */ if (datatype == GL_UNSIGNED_SHORT && comps == 4) { - GLuint i, j, k; - const GLushort (*rowA)[4] = (const GLushort (*)[4]) srcRowA; - const GLushort (*rowB)[4] = (const GLushort (*)[4]) srcRowB; - GLushort (*dst)[4] = (GLushort (*)[4]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) / 4; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) / 4; - dst[i][2] = (rowA[j][2] + rowA[k][2] + - rowB[j][2] + rowB[k][2]) / 4; - dst[i][3] = (rowA[j][3] + rowA[k][3] + - rowB[j][3] + rowB[k][3]) / 4; - } + GLuint i, j, k; + const GLushort(*rowA)[4] = (const GLushort(*)[4]) srcRowA; + const GLushort(*rowB)[4] = (const GLushort(*)[4]) srcRowB; + GLushort(*dst)[4] = (GLushort(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; + } } else if (datatype == GL_UNSIGNED_SHORT && comps == 3) { - GLuint i, j, k; - const GLushort (*rowA)[3] = (const GLushort (*)[3]) srcRowA; - const GLushort (*rowB)[3] = (const GLushort (*)[3]) srcRowB; - GLushort (*dst)[3] = (GLushort (*)[3]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) / 4; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) / 4; - dst[i][2] = (rowA[j][2] + rowA[k][2] + - rowB[j][2] + rowB[k][2]) / 4; - } + GLuint i, j, k; + const GLushort(*rowA)[3] = (const GLushort(*)[3]) srcRowA; + const GLushort(*rowB)[3] = (const GLushort(*)[3]) srcRowB; + GLushort(*dst)[3] = (GLushort(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + } } else if (datatype == GL_UNSIGNED_SHORT && comps == 1) { - GLuint i, j, k; - const GLushort *rowA = (const GLushort *) srcRowA; - const GLushort *rowB = (const GLushort *) srcRowB; - GLushort *dst = (GLushort *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4; - } + GLuint i, j, k; + const GLushort *rowA = (const GLushort *) srcRowA; + const GLushort *rowB = (const GLushort *) srcRowB; + GLushort *dst = (GLushort *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4; + } } else if (datatype == GL_UNSIGNED_SHORT && comps == 2) { - GLuint i, j, k; - const GLushort (*rowA)[2] = (const GLushort (*)[2]) srcRowA; - const GLushort (*rowB)[2] = (const GLushort (*)[2]) srcRowB; - GLushort (*dst)[2] = (GLushort (*)[2]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) / 4; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) / 4; - } + GLuint i, j, k; + const GLushort(*rowA)[2] = (const GLushort(*)[2]) srcRowA; + const GLushort(*rowB)[2] = (const GLushort(*)[2]) srcRowB; + GLushort(*dst)[2] = (GLushort(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + } } else if (datatype == GL_UNSIGNED_INT && comps == 1) { - GLuint i, j, k; - const GLuint *rowA = (const GLuint *) srcRowA; - const GLuint *rowB = (const GLuint *) srcRowB; - GLfloat *dst = (GLfloat *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; - } + GLuint i, j, k; + const GLuint *rowA = (const GLuint *) srcRowA; + const GLuint *rowB = (const GLuint *) srcRowB; + GLfloat *dst = (GLfloat *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; + } } else if (datatype == GL_UNSIGNED_BYTE && comps == 4) { - GLuint i, j, k; - const GLubyte (*rowA)[4] = (const GLubyte (*)[4]) srcRowA; - const GLubyte (*rowB)[4] = (const GLubyte (*)[4]) srcRowB; - GLubyte (*dst)[4] = (GLubyte (*)[4]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) / 4; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) / 4; - dst[i][2] = (rowA[j][2] + rowA[k][2] + - rowB[j][2] + rowB[k][2]) / 4; - dst[i][3] = (rowA[j][3] + rowA[k][3] + - rowB[j][3] + rowB[k][3]) / 4; - } + GLuint i, j, k; + const GLubyte(*rowA)[4] = (const GLubyte(*)[4]) srcRowA; + const GLubyte(*rowB)[4] = (const GLubyte(*)[4]) srcRowB; + GLubyte(*dst)[4] = (GLubyte(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; + } } else if (datatype == GL_UNSIGNED_BYTE && comps == 3) { - GLuint i, j, k; - const GLubyte (*rowA)[3] = (const GLubyte (*)[3]) srcRowA; - const GLubyte (*rowB)[3] = (const GLubyte (*)[3]) srcRowB; - GLubyte (*dst)[3] = (GLubyte (*)[3]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) / 4; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) / 4; - dst[i][2] = (rowA[j][2] + rowA[k][2] + - rowB[j][2] + rowB[k][2]) / 4; - } + GLuint i, j, k; + const GLubyte(*rowA)[3] = (const GLubyte(*)[3]) srcRowA; + const GLubyte(*rowB)[3] = (const GLubyte(*)[3]) srcRowB; + GLubyte(*dst)[3] = (GLubyte(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + } } else if (datatype == GL_UNSIGNED_SHORT_5_6_5 && comps == 3) { - GLuint i, j, k; - const GLushort *rowA = (const GLushort *) srcRowA; - const GLushort *rowB = (const GLushort *) srcRowB; - GLushort *dst = (GLushort *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - const GLint rowAr0 = rowA[j] & 0x1f; - const GLint rowAr1 = rowA[k] & 0x1f; - const GLint rowBr0 = rowB[j] & 0x1f; - const GLint rowBr1 = rowB[k] & 0x1f; - const GLint rowAg0 = (rowA[j] >> 5) & 0x3f; - const GLint rowAg1 = (rowA[k] >> 5) & 0x3f; - const GLint rowBg0 = (rowB[j] >> 5) & 0x3f; - const GLint rowBg1 = (rowB[k] >> 5) & 0x3f; - const GLint rowAb0 = (rowA[j] >> 11) & 0x1f; - const GLint rowAb1 = (rowA[k] >> 11) & 0x1f; - const GLint rowBb0 = (rowB[j] >> 11) & 0x1f; - const GLint rowBb1 = (rowB[k] >> 11) & 0x1f; - const GLint red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; - const GLint green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; - const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; - dst[i] = (blue << 11) | (green << 5) | red; - } + GLuint i, j, k; + const GLushort *rowA = (const GLushort *) srcRowA; + const GLushort *rowB = (const GLushort *) srcRowB; + GLushort *dst = (GLushort *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + const GLint rowAr0 = rowA[j] & 0x1f; + const GLint rowAr1 = rowA[k] & 0x1f; + const GLint rowBr0 = rowB[j] & 0x1f; + const GLint rowBr1 = rowB[k] & 0x1f; + const GLint rowAg0 = (rowA[j] >> 5) & 0x3f; + const GLint rowAg1 = (rowA[k] >> 5) & 0x3f; + const GLint rowBg0 = (rowB[j] >> 5) & 0x3f; + const GLint rowBg1 = (rowB[k] >> 5) & 0x3f; + const GLint rowAb0 = (rowA[j] >> 11) & 0x1f; + const GLint rowAb1 = (rowA[k] >> 11) & 0x1f; + const GLint rowBb0 = (rowB[j] >> 11) & 0x1f; + const GLint rowBb1 = (rowB[k] >> 11) & 0x1f; + const GLint red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; + const GLint green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; + const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; + dst[i] = (blue << 11) | (green << 5) | red; + } } else if (datatype == GL_UNSIGNED_SHORT_4_4_4_4 && comps == 4) { - GLuint i, j, k; - const GLushort *rowA = (const GLushort *) srcRowA; - const GLushort *rowB = (const GLushort *) srcRowB; - GLushort *dst = (GLushort *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - const GLint rowAr0 = rowA[j] & 0xf; - const GLint rowAr1 = rowA[k] & 0xf; - const GLint rowBr0 = rowB[j] & 0xf; - const GLint rowBr1 = rowB[k] & 0xf; - const GLint rowAg0 = (rowA[j] >> 4) & 0xf; - const GLint rowAg1 = (rowA[k] >> 4) & 0xf; - const GLint rowBg0 = (rowB[j] >> 4) & 0xf; - const GLint rowBg1 = (rowB[k] >> 4) & 0xf; - const GLint rowAb0 = (rowA[j] >> 8) & 0xf; - const GLint rowAb1 = (rowA[k] >> 8) & 0xf; - const GLint rowBb0 = (rowB[j] >> 8) & 0xf; - const GLint rowBb1 = (rowB[k] >> 8) & 0xf; - const GLint rowAa0 = (rowA[j] >> 12) & 0xf; - const GLint rowAa1 = (rowA[k] >> 12) & 0xf; - const GLint rowBa0 = (rowB[j] >> 12) & 0xf; - const GLint rowBa1 = (rowB[k] >> 12) & 0xf; - const GLint red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; - const GLint green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; - const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; - const GLint alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2; - dst[i] = (alpha << 12) | (blue << 8) | (green << 4) | red; - } + GLuint i, j, k; + const GLushort *rowA = (const GLushort *) srcRowA; + const GLushort *rowB = (const GLushort *) srcRowB; + GLushort *dst = (GLushort *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + const GLint rowAr0 = rowA[j] & 0xf; + const GLint rowAr1 = rowA[k] & 0xf; + const GLint rowBr0 = rowB[j] & 0xf; + const GLint rowBr1 = rowB[k] & 0xf; + const GLint rowAg0 = (rowA[j] >> 4) & 0xf; + const GLint rowAg1 = (rowA[k] >> 4) & 0xf; + const GLint rowBg0 = (rowB[j] >> 4) & 0xf; + const GLint rowBg1 = (rowB[k] >> 4) & 0xf; + const GLint rowAb0 = (rowA[j] >> 8) & 0xf; + const GLint rowAb1 = (rowA[k] >> 8) & 0xf; + const GLint rowBb0 = (rowB[j] >> 8) & 0xf; + const GLint rowBb1 = (rowB[k] >> 8) & 0xf; + const GLint rowAa0 = (rowA[j] >> 12) & 0xf; + const GLint rowAa1 = (rowA[k] >> 12) & 0xf; + const GLint rowBa0 = (rowB[j] >> 12) & 0xf; + const GLint rowBa1 = (rowB[k] >> 12) & 0xf; + const GLint red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; + const GLint green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; + const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; + const GLint alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2; + dst[i] = (alpha << 12) | (blue << 8) | (green << 4) | red; + } } else if (datatype == GL_UNSIGNED_SHORT_1_5_5_5_REV && comps == 4) { - GLuint i, j, k; - const GLushort *rowA = (const GLushort *) srcRowA; - const GLushort *rowB = (const GLushort *) srcRowB; - GLushort *dst = (GLushort *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - const GLint rowAr0 = rowA[j] & 0x1f; - const GLint rowAr1 = rowA[k] & 0x1f; - const GLint rowBr0 = rowB[j] & 0x1f; - const GLint rowBr1 = rowB[k] & 0xf; - const GLint rowAg0 = (rowA[j] >> 5) & 0x1f; - const GLint rowAg1 = (rowA[k] >> 5) & 0x1f; - const GLint rowBg0 = (rowB[j] >> 5) & 0x1f; - const GLint rowBg1 = (rowB[k] >> 5) & 0x1f; - const GLint rowAb0 = (rowA[j] >> 10) & 0x1f; - const GLint rowAb1 = (rowA[k] >> 10) & 0x1f; - const GLint rowBb0 = (rowB[j] >> 10) & 0x1f; - const GLint rowBb1 = (rowB[k] >> 10) & 0x1f; - const GLint rowAa0 = (rowA[j] >> 15) & 0x1; - const GLint rowAa1 = (rowA[k] >> 15) & 0x1; - const GLint rowBa0 = (rowB[j] >> 15) & 0x1; - const GLint rowBa1 = (rowB[k] >> 15) & 0x1; - const GLint red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; - const GLint green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; - const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; - const GLint alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2; - dst[i] = (alpha << 15) | (blue << 10) | (green << 5) | red; - } + GLuint i, j, k; + const GLushort *rowA = (const GLushort *) srcRowA; + const GLushort *rowB = (const GLushort *) srcRowB; + GLushort *dst = (GLushort *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + const GLint rowAr0 = rowA[j] & 0x1f; + const GLint rowAr1 = rowA[k] & 0x1f; + const GLint rowBr0 = rowB[j] & 0x1f; + const GLint rowBr1 = rowB[k] & 0xf; + const GLint rowAg0 = (rowA[j] >> 5) & 0x1f; + const GLint rowAg1 = (rowA[k] >> 5) & 0x1f; + const GLint rowBg0 = (rowB[j] >> 5) & 0x1f; + const GLint rowBg1 = (rowB[k] >> 5) & 0x1f; + const GLint rowAb0 = (rowA[j] >> 10) & 0x1f; + const GLint rowAb1 = (rowA[k] >> 10) & 0x1f; + const GLint rowBb0 = (rowB[j] >> 10) & 0x1f; + const GLint rowBb1 = (rowB[k] >> 10) & 0x1f; + const GLint rowAa0 = (rowA[j] >> 15) & 0x1; + const GLint rowAa1 = (rowA[k] >> 15) & 0x1; + const GLint rowBa0 = (rowB[j] >> 15) & 0x1; + const GLint rowBa1 = (rowB[k] >> 15) & 0x1; + const GLint red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; + const GLint green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; + const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; + const GLint alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2; + dst[i] = (alpha << 15) | (blue << 10) | (green << 5) | red; + } } else if (datatype == GL_UNSIGNED_BYTE && comps == 2) { - GLuint i, j, k; - const GLubyte (*rowA)[2] = (const GLubyte (*)[2]) srcRowA; - const GLubyte (*rowB)[2] = (const GLubyte (*)[2]) srcRowB; - GLubyte (*dst)[2] = (GLubyte (*)[2]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) >> 2; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) >> 2; - } + GLuint i, j, k; + const GLubyte(*rowA)[2] = (const GLubyte(*)[2]) srcRowA; + const GLubyte(*rowB)[2] = (const GLubyte(*)[2]) srcRowB; + GLubyte(*dst)[2] = (GLubyte(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) >> 2; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) >> 2; + } } else if (datatype == GL_UNSIGNED_BYTE_3_3_2 && comps == 3) { - GLuint i, j, k; - const GLubyte *rowA = (const GLubyte *) srcRowA; - const GLubyte *rowB = (const GLubyte *) srcRowB; - GLubyte *dst = (GLubyte *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - const GLint rowAr0 = rowA[j] & 0x3; - const GLint rowAr1 = rowA[k] & 0x3; - const GLint rowBr0 = rowB[j] & 0x3; - const GLint rowBr1 = rowB[k] & 0x3; - const GLint rowAg0 = (rowA[j] >> 2) & 0x7; - const GLint rowAg1 = (rowA[k] >> 2) & 0x7; - const GLint rowBg0 = (rowB[j] >> 2) & 0x7; - const GLint rowBg1 = (rowB[k] >> 2) & 0x7; - const GLint rowAb0 = (rowA[j] >> 5) & 0x7; - const GLint rowAb1 = (rowA[k] >> 5) & 0x7; - const GLint rowBb0 = (rowB[j] >> 5) & 0x7; - const GLint rowBb1 = (rowB[k] >> 5) & 0x7; - const GLint red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; - const GLint green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; - const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; - dst[i] = (blue << 5) | (green << 2) | red; - } + GLuint i, j, k; + const GLubyte *rowA = (const GLubyte *) srcRowA; + const GLubyte *rowB = (const GLubyte *) srcRowB; + GLubyte *dst = (GLubyte *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + const GLint rowAr0 = rowA[j] & 0x3; + const GLint rowAr1 = rowA[k] & 0x3; + const GLint rowBr0 = rowB[j] & 0x3; + const GLint rowBr1 = rowB[k] & 0x3; + const GLint rowAg0 = (rowA[j] >> 2) & 0x7; + const GLint rowAg1 = (rowA[k] >> 2) & 0x7; + const GLint rowBg0 = (rowB[j] >> 2) & 0x7; + const GLint rowBg1 = (rowB[k] >> 2) & 0x7; + const GLint rowAb0 = (rowA[j] >> 5) & 0x7; + const GLint rowAb1 = (rowA[k] >> 5) & 0x7; + const GLint rowBb0 = (rowB[j] >> 5) & 0x7; + const GLint rowBb1 = (rowB[k] >> 5) & 0x7; + const GLint red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; + const GLint green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; + const GLint blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; + dst[i] = (blue << 5) | (green << 2) | red; + } } else if (datatype == GL_UNSIGNED_BYTE && comps == 1) { - GLuint i, j, k; - const GLubyte *rowA = (const GLubyte *) srcRowA; - const GLubyte *rowB = (const GLubyte *) srcRowB; - GLubyte *dst = (GLubyte *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) >> 2; - } + GLuint i, j, k; + const GLubyte *rowA = (const GLubyte *) srcRowA; + const GLubyte *rowB = (const GLubyte *) srcRowB; + GLubyte *dst = (GLubyte *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) >> 2; + } } else if (datatype == GL_FLOAT && comps == 4) { - GLuint i, j, k; - const GLfloat (*rowA)[4] = (const GLfloat (*)[4]) srcRowA; - const GLfloat (*rowB)[4] = (const GLfloat (*)[4]) srcRowB; - GLfloat (*dst)[4] = (GLfloat (*)[4]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) * 0.25F; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) * 0.25F; - dst[i][2] = (rowA[j][2] + rowA[k][2] + - rowB[j][2] + rowB[k][2]) * 0.25F; - dst[i][3] = (rowA[j][3] + rowA[k][3] + - rowB[j][3] + rowB[k][3]) * 0.25F; - } + GLuint i, j, k; + const GLfloat(*rowA)[4] = (const GLfloat(*)[4]) srcRowA; + const GLfloat(*rowB)[4] = (const GLfloat(*)[4]) srcRowB; + GLfloat(*dst)[4] = (GLfloat(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + + rowB[j][0] + rowB[k][0]) * 0.25F; + dst[i][1] = (rowA[j][1] + rowA[k][1] + + rowB[j][1] + rowB[k][1]) * 0.25F; + dst[i][2] = (rowA[j][2] + rowA[k][2] + + rowB[j][2] + rowB[k][2]) * 0.25F; + dst[i][3] = (rowA[j][3] + rowA[k][3] + + rowB[j][3] + rowB[k][3]) * 0.25F; + } } else if (datatype == GL_HALF_FLOAT_ARB && comps == 4) { - GLuint i, j, k, comp; - const GLhalfARB (*rowA)[4] = (const GLhalfARB (*)[4]) srcRowA; - const GLhalfARB (*rowB)[4] = (const GLhalfARB (*)[4]) srcRowB; - GLhalfARB (*dst)[4] = (GLhalfARB (*)[4]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - for (comp = 0; comp < 4; comp++) { - GLfloat aj, ak, bj, bk; - aj = _mesa_half_to_float(rowA[j][comp]); - ak = _mesa_half_to_float(rowA[k][comp]); - bj = _mesa_half_to_float(rowB[j][comp]); - bk = _mesa_half_to_float(rowB[k][comp]); - dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); - } + GLuint i, j, k, comp; + const GLhalfARB(*rowA)[4] = (const GLhalfARB(*)[4]) srcRowA; + const GLhalfARB(*rowB)[4] = (const GLhalfARB(*)[4]) srcRowB; + GLhalfARB(*dst)[4] = (GLhalfARB(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + for (comp = 0; comp < 4; comp++) { + GLfloat aj, ak, bj, bk; + aj = _mesa_half_to_float(rowA[j][comp]); + ak = _mesa_half_to_float(rowA[k][comp]); + bj = _mesa_half_to_float(rowB[j][comp]); + bk = _mesa_half_to_float(rowB[k][comp]); + dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); } + } } else if (datatype == GL_FLOAT && comps == 3) { - GLuint i, j, k; - const GLfloat (*rowA)[3] = (const GLfloat (*)[3]) srcRowA; - const GLfloat (*rowB)[3] = (const GLfloat (*)[3]) srcRowB; - GLfloat (*dst)[3] = (GLfloat (*)[3]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) * 0.25F; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) * 0.25F; - dst[i][2] = (rowA[j][2] + rowA[k][2] + - rowB[j][2] + rowB[k][2]) * 0.25F; - } + GLuint i, j, k; + const GLfloat(*rowA)[3] = (const GLfloat(*)[3]) srcRowA; + const GLfloat(*rowB)[3] = (const GLfloat(*)[3]) srcRowB; + GLfloat(*dst)[3] = (GLfloat(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + + rowB[j][0] + rowB[k][0]) * 0.25F; + dst[i][1] = (rowA[j][1] + rowA[k][1] + + rowB[j][1] + rowB[k][1]) * 0.25F; + dst[i][2] = (rowA[j][2] + rowA[k][2] + + rowB[j][2] + rowB[k][2]) * 0.25F; + } } else if (datatype == GL_HALF_FLOAT_ARB && comps == 3) { - GLuint i, j, k, comp; - const GLhalfARB (*rowA)[3] = (const GLhalfARB (*)[3]) srcRowA; - const GLhalfARB (*rowB)[3] = (const GLhalfARB (*)[3]) srcRowB; - GLhalfARB (*dst)[3] = (GLhalfARB (*)[3]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - for (comp = 0; comp < 3; comp++) { - GLfloat aj, ak, bj, bk; - aj = _mesa_half_to_float(rowA[j][comp]); - ak = _mesa_half_to_float(rowA[k][comp]); - bj = _mesa_half_to_float(rowB[j][comp]); - bk = _mesa_half_to_float(rowB[k][comp]); - dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); - } + GLuint i, j, k, comp; + const GLhalfARB(*rowA)[3] = (const GLhalfARB(*)[3]) srcRowA; + const GLhalfARB(*rowB)[3] = (const GLhalfARB(*)[3]) srcRowB; + GLhalfARB(*dst)[3] = (GLhalfARB(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + for (comp = 0; comp < 3; comp++) { + GLfloat aj, ak, bj, bk; + aj = _mesa_half_to_float(rowA[j][comp]); + ak = _mesa_half_to_float(rowA[k][comp]); + bj = _mesa_half_to_float(rowB[j][comp]); + bk = _mesa_half_to_float(rowB[k][comp]); + dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); } + } } else if (datatype == GL_FLOAT && comps == 2) { - GLuint i, j, k; - const GLfloat (*rowA)[2] = (const GLfloat (*)[2]) srcRowA; - const GLfloat (*rowB)[2] = (const GLfloat (*)[2]) srcRowB; - GLfloat (*dst)[2] = (GLfloat (*)[2]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) * 0.25F; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) * 0.25F; - } + GLuint i, j, k; + const GLfloat(*rowA)[2] = (const GLfloat(*)[2]) srcRowA; + const GLfloat(*rowB)[2] = (const GLfloat(*)[2]) srcRowB; + GLfloat(*dst)[2] = (GLfloat(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + + rowB[j][0] + rowB[k][0]) * 0.25F; + dst[i][1] = (rowA[j][1] + rowA[k][1] + + rowB[j][1] + rowB[k][1]) * 0.25F; + } } else if (datatype == GL_HALF_FLOAT_ARB && comps == 2) { - GLuint i, j, k, comp; - const GLhalfARB (*rowA)[2] = (const GLhalfARB (*)[2]) srcRowA; - const GLhalfARB (*rowB)[2] = (const GLhalfARB (*)[2]) srcRowB; - GLhalfARB (*dst)[2] = (GLhalfARB (*)[2]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - for (comp = 0; comp < 2; comp++) { - GLfloat aj, ak, bj, bk; - aj = _mesa_half_to_float(rowA[j][comp]); - ak = _mesa_half_to_float(rowA[k][comp]); - bj = _mesa_half_to_float(rowB[j][comp]); - bk = _mesa_half_to_float(rowB[k][comp]); - dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); - } + GLuint i, j, k, comp; + const GLhalfARB(*rowA)[2] = (const GLhalfARB(*)[2]) srcRowA; + const GLhalfARB(*rowB)[2] = (const GLhalfARB(*)[2]) srcRowB; + GLhalfARB(*dst)[2] = (GLhalfARB(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + for (comp = 0; comp < 2; comp++) { + GLfloat aj, ak, bj, bk; + aj = _mesa_half_to_float(rowA[j][comp]); + ak = _mesa_half_to_float(rowA[k][comp]); + bj = _mesa_half_to_float(rowB[j][comp]); + bk = _mesa_half_to_float(rowB[k][comp]); + dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); } + } } else if (datatype == GL_FLOAT && comps == 1) { - GLuint i, j, k; - const GLfloat *rowA = (const GLfloat *) srcRowA; - const GLfloat *rowB = (const GLfloat *) srcRowB; - GLfloat *dst = (GLfloat *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) * 0.25F; - } + GLuint i, j, k; + const GLfloat *rowA = (const GLfloat *) srcRowA; + const GLfloat *rowB = (const GLfloat *) srcRowB; + GLfloat *dst = (GLfloat *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) * 0.25F; + } } else if (datatype == GL_HALF_FLOAT_ARB && comps == 1) { - GLuint i, j, k; - const GLhalfARB *rowA = (const GLhalfARB *) srcRowA; - const GLhalfARB *rowB = (const GLhalfARB *) srcRowB; - GLhalfARB *dst = (GLhalfARB *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - GLfloat aj, ak, bj, bk; - aj = _mesa_half_to_float(rowA[j]); - ak = _mesa_half_to_float(rowA[k]); - bj = _mesa_half_to_float(rowB[j]); - bk = _mesa_half_to_float(rowB[k]); - dst[i] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); - } + GLuint i, j, k; + const GLhalfARB *rowA = (const GLhalfARB *) srcRowA; + const GLhalfARB *rowB = (const GLhalfARB *) srcRowB; + GLhalfARB *dst = (GLhalfARB *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + GLfloat aj, ak, bj, bk; + aj = _mesa_half_to_float(rowA[j]); + ak = _mesa_half_to_float(rowA[k]); + bj = _mesa_half_to_float(rowB[j]); + bk = _mesa_half_to_float(rowB[k]); + dst[i] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); + } } else { _mesa_problem(NULL, "bad format in do_row()"); -- cgit v1.2.3 From 9ddb7d794bac57e63f5eb247ad3c02ba6dfb87f9 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 16:42:50 -0700 Subject: reorder cases in do_row() --- src/mesa/main/mipmap.c | 341 +++++++++++++++++++++++++------------------------ 1 file changed, 173 insertions(+), 168 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index 22c7530e83..db8ab65401 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -234,7 +234,54 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, assert(srcWidth == dstWidth || srcWidth == 2 * dstWidth); */ - if (datatype == GL_UNSIGNED_SHORT && comps == 4) { + if (datatype == GL_UNSIGNED_BYTE && comps == 4) { + GLuint i, j, k; + const GLubyte(*rowA)[4] = (const GLubyte(*)[4]) srcRowA; + const GLubyte(*rowB)[4] = (const GLubyte(*)[4]) srcRowB; + GLubyte(*dst)[4] = (GLubyte(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; + } + } + else if (datatype == GL_UNSIGNED_BYTE && comps == 3) { + GLuint i, j, k; + const GLubyte(*rowA)[3] = (const GLubyte(*)[3]) srcRowA; + const GLubyte(*rowB)[3] = (const GLubyte(*)[3]) srcRowB; + GLubyte(*dst)[3] = (GLubyte(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + } + } + else if (datatype == GL_UNSIGNED_BYTE && comps == 2) { + GLuint i, j, k; + const GLubyte(*rowA)[2] = (const GLubyte(*)[2]) srcRowA; + const GLubyte(*rowB)[2] = (const GLubyte(*)[2]) srcRowB; + GLubyte(*dst)[2] = (GLubyte(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) >> 2; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) >> 2; + } + } + else if (datatype == GL_UNSIGNED_BYTE && comps == 1) { + GLuint i, j, k; + const GLubyte *rowA = (const GLubyte *) srcRowA; + const GLubyte *rowB = (const GLubyte *) srcRowB; + GLubyte *dst = (GLubyte *) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) >> 2; + } + } + + else if (datatype == GL_UNSIGNED_SHORT && comps == 4) { GLuint i, j, k; const GLushort(*rowA)[4] = (const GLushort(*)[4]) srcRowA; const GLushort(*rowB)[4] = (const GLushort(*)[4]) srcRowB; @@ -259,6 +306,17 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; } } + else if (datatype == GL_UNSIGNED_SHORT && comps == 2) { + GLuint i, j, k; + const GLushort(*rowA)[2] = (const GLushort(*)[2]) srcRowA; + const GLushort(*rowB)[2] = (const GLushort(*)[2]) srcRowB; + GLushort(*dst)[2] = (GLushort(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + } + } else if (datatype == GL_UNSIGNED_SHORT && comps == 1) { GLuint i, j, k; const GLushort *rowA = (const GLushort *) srcRowA; @@ -269,52 +327,141 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4; } } - else if (datatype == GL_UNSIGNED_SHORT && comps == 2) { + + else if (datatype == GL_FLOAT && comps == 4) { GLuint i, j, k; - const GLushort(*rowA)[2] = (const GLushort(*)[2]) srcRowA; - const GLushort(*rowB)[2] = (const GLushort(*)[2]) srcRowB; - GLushort(*dst)[2] = (GLushort(*)[2]) dstRow; + const GLfloat(*rowA)[4] = (const GLfloat(*)[4]) srcRowA; + const GLfloat(*rowB)[4] = (const GLfloat(*)[4]) srcRowB; + GLfloat(*dst)[4] = (GLfloat(*)[4]) dstRow; for (i = j = 0, k = k0; i < (GLuint) dstWidth; i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; - dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][0] = (rowA[j][0] + rowA[k][0] + + rowB[j][0] + rowB[k][0]) * 0.25F; + dst[i][1] = (rowA[j][1] + rowA[k][1] + + rowB[j][1] + rowB[k][1]) * 0.25F; + dst[i][2] = (rowA[j][2] + rowA[k][2] + + rowB[j][2] + rowB[k][2]) * 0.25F; + dst[i][3] = (rowA[j][3] + rowA[k][3] + + rowB[j][3] + rowB[k][3]) * 0.25F; } } - else if (datatype == GL_UNSIGNED_INT && comps == 1) { + else if (datatype == GL_FLOAT && comps == 3) { GLuint i, j, k; - const GLuint *rowA = (const GLuint *) srcRowA; - const GLuint *rowB = (const GLuint *) srcRowB; + const GLfloat(*rowA)[3] = (const GLfloat(*)[3]) srcRowA; + const GLfloat(*rowB)[3] = (const GLfloat(*)[3]) srcRowB; + GLfloat(*dst)[3] = (GLfloat(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + + rowB[j][0] + rowB[k][0]) * 0.25F; + dst[i][1] = (rowA[j][1] + rowA[k][1] + + rowB[j][1] + rowB[k][1]) * 0.25F; + dst[i][2] = (rowA[j][2] + rowA[k][2] + + rowB[j][2] + rowB[k][2]) * 0.25F; + } + } + else if (datatype == GL_FLOAT && comps == 2) { + GLuint i, j, k; + const GLfloat(*rowA)[2] = (const GLfloat(*)[2]) srcRowA; + const GLfloat(*rowB)[2] = (const GLfloat(*)[2]) srcRowB; + GLfloat(*dst)[2] = (GLfloat(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + + rowB[j][0] + rowB[k][0]) * 0.25F; + dst[i][1] = (rowA[j][1] + rowA[k][1] + + rowB[j][1] + rowB[k][1]) * 0.25F; + } + } + else if (datatype == GL_FLOAT && comps == 1) { + GLuint i, j, k; + const GLfloat *rowA = (const GLfloat *) srcRowA; + const GLfloat *rowB = (const GLfloat *) srcRowB; GLfloat *dst = (GLfloat *) dstRow; for (i = j = 0, k = k0; i < (GLuint) dstWidth; i++, j += colStride, k += colStride) { - dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; + dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) * 0.25F; + } + } + + else if (datatype == GL_HALF_FLOAT_ARB && comps == 4) { + GLuint i, j, k, comp; + const GLhalfARB(*rowA)[4] = (const GLhalfARB(*)[4]) srcRowA; + const GLhalfARB(*rowB)[4] = (const GLhalfARB(*)[4]) srcRowB; + GLhalfARB(*dst)[4] = (GLhalfARB(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + for (comp = 0; comp < 4; comp++) { + GLfloat aj, ak, bj, bk; + aj = _mesa_half_to_float(rowA[j][comp]); + ak = _mesa_half_to_float(rowA[k][comp]); + bj = _mesa_half_to_float(rowB[j][comp]); + bk = _mesa_half_to_float(rowB[k][comp]); + dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); + } + } + } + else if (datatype == GL_HALF_FLOAT_ARB && comps == 3) { + GLuint i, j, k, comp; + const GLhalfARB(*rowA)[3] = (const GLhalfARB(*)[3]) srcRowA; + const GLhalfARB(*rowB)[3] = (const GLhalfARB(*)[3]) srcRowB; + GLhalfARB(*dst)[3] = (GLhalfARB(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + for (comp = 0; comp < 3; comp++) { + GLfloat aj, ak, bj, bk; + aj = _mesa_half_to_float(rowA[j][comp]); + ak = _mesa_half_to_float(rowA[k][comp]); + bj = _mesa_half_to_float(rowB[j][comp]); + bk = _mesa_half_to_float(rowB[k][comp]); + dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); + } + } + } + else if (datatype == GL_HALF_FLOAT_ARB && comps == 2) { + GLuint i, j, k, comp; + const GLhalfARB(*rowA)[2] = (const GLhalfARB(*)[2]) srcRowA; + const GLhalfARB(*rowB)[2] = (const GLhalfARB(*)[2]) srcRowB; + GLhalfARB(*dst)[2] = (GLhalfARB(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (GLuint) dstWidth; + i++, j += colStride, k += colStride) { + for (comp = 0; comp < 2; comp++) { + GLfloat aj, ak, bj, bk; + aj = _mesa_half_to_float(rowA[j][comp]); + ak = _mesa_half_to_float(rowA[k][comp]); + bj = _mesa_half_to_float(rowB[j][comp]); + bk = _mesa_half_to_float(rowB[k][comp]); + dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); + } } } - else if (datatype == GL_UNSIGNED_BYTE && comps == 4) { + else if (datatype == GL_HALF_FLOAT_ARB && comps == 1) { GLuint i, j, k; - const GLubyte(*rowA)[4] = (const GLubyte(*)[4]) srcRowA; - const GLubyte(*rowB)[4] = (const GLubyte(*)[4]) srcRowB; - GLubyte(*dst)[4] = (GLubyte(*)[4]) dstRow; + const GLhalfARB *rowA = (const GLhalfARB *) srcRowA; + const GLhalfARB *rowB = (const GLhalfARB *) srcRowB; + GLhalfARB *dst = (GLhalfARB *) dstRow; for (i = j = 0, k = k0; i < (GLuint) dstWidth; i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; - dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; - dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; - dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; + GLfloat aj, ak, bj, bk; + aj = _mesa_half_to_float(rowA[j]); + ak = _mesa_half_to_float(rowA[k]); + bj = _mesa_half_to_float(rowB[j]); + bk = _mesa_half_to_float(rowB[k]); + dst[i] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); } } - else if (datatype == GL_UNSIGNED_BYTE && comps == 3) { + + else if (datatype == GL_UNSIGNED_INT && comps == 1) { GLuint i, j, k; - const GLubyte(*rowA)[3] = (const GLubyte(*)[3]) srcRowA; - const GLubyte(*rowB)[3] = (const GLubyte(*)[3]) srcRowB; - GLubyte(*dst)[3] = (GLubyte(*)[3]) dstRow; + const GLuint *rowA = (const GLuint *) srcRowA; + const GLuint *rowB = (const GLuint *) srcRowB; + GLfloat *dst = (GLfloat *) dstRow; for (i = j = 0, k = k0; i < (GLuint) dstWidth; i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; - dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; - dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; } } + else if (datatype == GL_UNSIGNED_SHORT_5_6_5 && comps == 3) { GLuint i, j, k; const GLushort *rowA = (const GLushort *) srcRowA; @@ -400,17 +547,6 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, dst[i] = (alpha << 15) | (blue << 10) | (green << 5) | red; } } - else if (datatype == GL_UNSIGNED_BYTE && comps == 2) { - GLuint i, j, k; - const GLubyte(*rowA)[2] = (const GLubyte(*)[2]) srcRowA; - const GLubyte(*rowB)[2] = (const GLubyte(*)[2]) srcRowB; - GLubyte(*dst)[2] = (GLubyte(*)[2]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) >> 2; - dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) >> 2; - } - } else if (datatype == GL_UNSIGNED_BYTE_3_3_2 && comps == 3) { GLuint i, j, k; const GLubyte *rowA = (const GLubyte *) srcRowA; @@ -436,137 +572,6 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, dst[i] = (blue << 5) | (green << 2) | red; } } - else if (datatype == GL_UNSIGNED_BYTE && comps == 1) { - GLuint i, j, k; - const GLubyte *rowA = (const GLubyte *) srcRowA; - const GLubyte *rowB = (const GLubyte *) srcRowB; - GLubyte *dst = (GLubyte *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) >> 2; - } - } - else if (datatype == GL_FLOAT && comps == 4) { - GLuint i, j, k; - const GLfloat(*rowA)[4] = (const GLfloat(*)[4]) srcRowA; - const GLfloat(*rowB)[4] = (const GLfloat(*)[4]) srcRowB; - GLfloat(*dst)[4] = (GLfloat(*)[4]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) * 0.25F; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) * 0.25F; - dst[i][2] = (rowA[j][2] + rowA[k][2] + - rowB[j][2] + rowB[k][2]) * 0.25F; - dst[i][3] = (rowA[j][3] + rowA[k][3] + - rowB[j][3] + rowB[k][3]) * 0.25F; - } - } - else if (datatype == GL_HALF_FLOAT_ARB && comps == 4) { - GLuint i, j, k, comp; - const GLhalfARB(*rowA)[4] = (const GLhalfARB(*)[4]) srcRowA; - const GLhalfARB(*rowB)[4] = (const GLhalfARB(*)[4]) srcRowB; - GLhalfARB(*dst)[4] = (GLhalfARB(*)[4]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - for (comp = 0; comp < 4; comp++) { - GLfloat aj, ak, bj, bk; - aj = _mesa_half_to_float(rowA[j][comp]); - ak = _mesa_half_to_float(rowA[k][comp]); - bj = _mesa_half_to_float(rowB[j][comp]); - bk = _mesa_half_to_float(rowB[k][comp]); - dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); - } - } - } - else if (datatype == GL_FLOAT && comps == 3) { - GLuint i, j, k; - const GLfloat(*rowA)[3] = (const GLfloat(*)[3]) srcRowA; - const GLfloat(*rowB)[3] = (const GLfloat(*)[3]) srcRowB; - GLfloat(*dst)[3] = (GLfloat(*)[3]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) * 0.25F; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) * 0.25F; - dst[i][2] = (rowA[j][2] + rowA[k][2] + - rowB[j][2] + rowB[k][2]) * 0.25F; - } - } - else if (datatype == GL_HALF_FLOAT_ARB && comps == 3) { - GLuint i, j, k, comp; - const GLhalfARB(*rowA)[3] = (const GLhalfARB(*)[3]) srcRowA; - const GLhalfARB(*rowB)[3] = (const GLhalfARB(*)[3]) srcRowB; - GLhalfARB(*dst)[3] = (GLhalfARB(*)[3]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - for (comp = 0; comp < 3; comp++) { - GLfloat aj, ak, bj, bk; - aj = _mesa_half_to_float(rowA[j][comp]); - ak = _mesa_half_to_float(rowA[k][comp]); - bj = _mesa_half_to_float(rowB[j][comp]); - bk = _mesa_half_to_float(rowB[k][comp]); - dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); - } - } - } - else if (datatype == GL_FLOAT && comps == 2) { - GLuint i, j, k; - const GLfloat(*rowA)[2] = (const GLfloat(*)[2]) srcRowA; - const GLfloat(*rowB)[2] = (const GLfloat(*)[2]) srcRowB; - GLfloat(*dst)[2] = (GLfloat(*)[2]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i][0] = (rowA[j][0] + rowA[k][0] + - rowB[j][0] + rowB[k][0]) * 0.25F; - dst[i][1] = (rowA[j][1] + rowA[k][1] + - rowB[j][1] + rowB[k][1]) * 0.25F; - } - } - else if (datatype == GL_HALF_FLOAT_ARB && comps == 2) { - GLuint i, j, k, comp; - const GLhalfARB(*rowA)[2] = (const GLhalfARB(*)[2]) srcRowA; - const GLhalfARB(*rowB)[2] = (const GLhalfARB(*)[2]) srcRowB; - GLhalfARB(*dst)[2] = (GLhalfARB(*)[2]) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - for (comp = 0; comp < 2; comp++) { - GLfloat aj, ak, bj, bk; - aj = _mesa_half_to_float(rowA[j][comp]); - ak = _mesa_half_to_float(rowA[k][comp]); - bj = _mesa_half_to_float(rowB[j][comp]); - bk = _mesa_half_to_float(rowB[k][comp]); - dst[i][comp] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); - } - } - } - else if (datatype == GL_FLOAT && comps == 1) { - GLuint i, j, k; - const GLfloat *rowA = (const GLfloat *) srcRowA; - const GLfloat *rowB = (const GLfloat *) srcRowB; - GLfloat *dst = (GLfloat *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) * 0.25F; - } - } - else if (datatype == GL_HALF_FLOAT_ARB && comps == 1) { - GLuint i, j, k; - const GLhalfARB *rowA = (const GLhalfARB *) srcRowA; - const GLhalfARB *rowB = (const GLhalfARB *) srcRowB; - GLhalfARB *dst = (GLhalfARB *) dstRow; - for (i = j = 0, k = k0; i < (GLuint) dstWidth; - i++, j += colStride, k += colStride) { - GLfloat aj, ak, bj, bk; - aj = _mesa_half_to_float(rowA[j]); - ak = _mesa_half_to_float(rowA[k]); - bj = _mesa_half_to_float(rowB[j]); - bk = _mesa_half_to_float(rowB[k]); - dst[i] = _mesa_float_to_half((aj + ak + bj + bk) * 0.25F); - } - } else { _mesa_problem(NULL, "bad format in do_row()"); } -- cgit v1.2.3 From 42eac65da45fb58bffdf94ab8f9860d8cee5b256 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 16:46:12 -0700 Subject: move _mesa_format_to_type_and_comps() to texformat.c --- src/mesa/main/mipmap.c | 168 +-------------------------------------------- src/mesa/main/texformat.c | 171 ++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/texformat.h | 6 ++ 3 files changed, 178 insertions(+), 167 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index db8ab65401..981da5dd89 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -45,172 +45,6 @@ bytes_per_pixel(GLenum datatype, GLuint comps) } -static void -mesa_format_to_type_and_comps(const struct gl_texture_format *format, - GLenum *datatype, GLuint *comps) -{ - switch (format->MesaFormat) { - case MESA_FORMAT_RGBA8888: - case MESA_FORMAT_RGBA8888_REV: - case MESA_FORMAT_ARGB8888: - case MESA_FORMAT_ARGB8888_REV: - *datatype = CHAN_TYPE; - *comps = 4; - return; - case MESA_FORMAT_RGB888: - case MESA_FORMAT_BGR888: - *datatype = GL_UNSIGNED_BYTE; - *comps = 3; - return; - case MESA_FORMAT_RGB565: - case MESA_FORMAT_RGB565_REV: - *datatype = GL_UNSIGNED_SHORT_5_6_5; - *comps = 3; - return; - - case MESA_FORMAT_ARGB4444: - case MESA_FORMAT_ARGB4444_REV: - *datatype = GL_UNSIGNED_SHORT_4_4_4_4; - *comps = 4; - return; - - case MESA_FORMAT_ARGB1555: - case MESA_FORMAT_ARGB1555_REV: - *datatype = GL_UNSIGNED_SHORT_1_5_5_5_REV; - *comps = 3; - return; - - case MESA_FORMAT_AL88: - case MESA_FORMAT_AL88_REV: - *datatype = GL_UNSIGNED_BYTE; - *comps = 2; - return; - case MESA_FORMAT_RGB332: - *datatype = GL_UNSIGNED_BYTE_3_3_2; - *comps = 3; - return; - - case MESA_FORMAT_A8: - case MESA_FORMAT_L8: - case MESA_FORMAT_I8: - case MESA_FORMAT_CI8: - *datatype = GL_UNSIGNED_BYTE; - *comps = 1; - return; - - case MESA_FORMAT_YCBCR: - case MESA_FORMAT_YCBCR_REV: - *datatype = GL_UNSIGNED_SHORT; - *comps = 2; - return; - - case MESA_FORMAT_Z24_S8: - *datatype = GL_UNSIGNED_INT; - *comps = 1; /* XXX OK? */ - return; - - case MESA_FORMAT_Z16: - *datatype = GL_UNSIGNED_SHORT; - *comps = 1; - return; - - case MESA_FORMAT_Z32: - *datatype = GL_UNSIGNED_INT; - *comps = 1; - return; - - case MESA_FORMAT_SRGB8: - *datatype = GL_UNSIGNED_BYTE; - *comps = 3; - return; - case MESA_FORMAT_SRGBA8: - *datatype = GL_UNSIGNED_BYTE; - *comps = 4; - return; - case MESA_FORMAT_SL8: - *datatype = GL_UNSIGNED_BYTE; - *comps = 1; - return; - case MESA_FORMAT_SLA8: - *datatype = GL_UNSIGNED_BYTE; - *comps = 2; - return; - - case MESA_FORMAT_RGB_FXT1: - case MESA_FORMAT_RGBA_FXT1: - case MESA_FORMAT_RGB_DXT1: - case MESA_FORMAT_RGBA_DXT1: - case MESA_FORMAT_RGBA_DXT3: - case MESA_FORMAT_RGBA_DXT5: - /* XXX generate error instead? */ - *datatype = GL_UNSIGNED_BYTE; - *comps = 0; - return; - - case MESA_FORMAT_RGBA: - *datatype = CHAN_TYPE; - *comps = 4; - return; - case MESA_FORMAT_RGB: - *datatype = CHAN_TYPE; - *comps = 3; - return; - case MESA_FORMAT_LUMINANCE_ALPHA: - *datatype = CHAN_TYPE; - *comps = 2; - return; - case MESA_FORMAT_ALPHA: - case MESA_FORMAT_LUMINANCE: - case MESA_FORMAT_INTENSITY: - *datatype = CHAN_TYPE; - *comps = 1; - return; - - case MESA_FORMAT_RGBA_FLOAT32: - *datatype = GL_FLOAT; - *comps = 4; - return; - case MESA_FORMAT_RGBA_FLOAT16: - *datatype = GL_HALF_FLOAT_ARB; - *comps = 4; - return; - case MESA_FORMAT_RGB_FLOAT32: - *datatype = GL_FLOAT; - *comps = 3; - return; - case MESA_FORMAT_RGB_FLOAT16: - *datatype = GL_HALF_FLOAT_ARB; - *comps = 3; - return; - case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32: - *datatype = GL_FLOAT; - *comps = 2; - return; - case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16: - *datatype = GL_HALF_FLOAT_ARB; - *comps = 2; - return; - case MESA_FORMAT_ALPHA_FLOAT32: - case MESA_FORMAT_LUMINANCE_FLOAT32: - case MESA_FORMAT_INTENSITY_FLOAT32: - *datatype = GL_FLOAT; - *comps = 1; - return; - case MESA_FORMAT_ALPHA_FLOAT16: - case MESA_FORMAT_LUMINANCE_FLOAT16: - case MESA_FORMAT_INTENSITY_FLOAT16: - *datatype = GL_HALF_FLOAT_ARB; - *comps = 1; - return; - - default: - _mesa_problem(NULL, "bad texture format in mesa_format_to_type_and_comps"); - *datatype = 0; - *comps = 1; - } -} - - /** * Average together two rows of a source image to produce a single new * row in the dest image. It's legal for the two source rows to point @@ -1085,7 +919,7 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target, convertFormat = srcImage->TexFormat; } - mesa_format_to_type_and_comps(convertFormat, &datatype, &comps); + _mesa_format_to_type_and_comps(convertFormat, &datatype, &comps); for (level = texObj->BaseLevel; level < texObj->MaxLevel && level < maxLevels - 1; level++) { diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c index acc268e622..88fbd8f07c 100644 --- a/src/mesa/main/texformat.c +++ b/src/mesa/main/texformat.c @@ -1569,3 +1569,174 @@ _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat, _mesa_problem(ctx, "unexpected format in _mesa_choose_tex_format()"); return NULL; } + + + +/** + * Return datatype and number of components per texel for the + * given gl_texture_format. + */ +void +_mesa_format_to_type_and_comps(const struct gl_texture_format *format, + GLenum *datatype, GLuint *comps) +{ + switch (format->MesaFormat) { + case MESA_FORMAT_RGBA8888: + case MESA_FORMAT_RGBA8888_REV: + case MESA_FORMAT_ARGB8888: + case MESA_FORMAT_ARGB8888_REV: + *datatype = CHAN_TYPE; + *comps = 4; + return; + case MESA_FORMAT_RGB888: + case MESA_FORMAT_BGR888: + *datatype = GL_UNSIGNED_BYTE; + *comps = 3; + return; + case MESA_FORMAT_RGB565: + case MESA_FORMAT_RGB565_REV: + *datatype = GL_UNSIGNED_SHORT_5_6_5; + *comps = 3; + return; + + case MESA_FORMAT_ARGB4444: + case MESA_FORMAT_ARGB4444_REV: + *datatype = GL_UNSIGNED_SHORT_4_4_4_4; + *comps = 4; + return; + + case MESA_FORMAT_ARGB1555: + case MESA_FORMAT_ARGB1555_REV: + *datatype = GL_UNSIGNED_SHORT_1_5_5_5_REV; + *comps = 3; + return; + + case MESA_FORMAT_AL88: + case MESA_FORMAT_AL88_REV: + *datatype = GL_UNSIGNED_BYTE; + *comps = 2; + return; + case MESA_FORMAT_RGB332: + *datatype = GL_UNSIGNED_BYTE_3_3_2; + *comps = 3; + return; + + case MESA_FORMAT_A8: + case MESA_FORMAT_L8: + case MESA_FORMAT_I8: + case MESA_FORMAT_CI8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 1; + return; + + case MESA_FORMAT_YCBCR: + case MESA_FORMAT_YCBCR_REV: + *datatype = GL_UNSIGNED_SHORT; + *comps = 2; + return; + + case MESA_FORMAT_Z24_S8: + *datatype = GL_UNSIGNED_INT; + *comps = 1; /* XXX OK? */ + return; + + case MESA_FORMAT_Z16: + *datatype = GL_UNSIGNED_SHORT; + *comps = 1; + return; + + case MESA_FORMAT_Z32: + *datatype = GL_UNSIGNED_INT; + *comps = 1; + return; + + case MESA_FORMAT_SRGB8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 3; + return; + case MESA_FORMAT_SRGBA8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 4; + return; + case MESA_FORMAT_SL8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 1; + return; + case MESA_FORMAT_SLA8: + *datatype = GL_UNSIGNED_BYTE; + *comps = 2; + return; + + case MESA_FORMAT_RGB_FXT1: + case MESA_FORMAT_RGBA_FXT1: + case MESA_FORMAT_RGB_DXT1: + case MESA_FORMAT_RGBA_DXT1: + case MESA_FORMAT_RGBA_DXT3: + case MESA_FORMAT_RGBA_DXT5: + /* XXX generate error instead? */ + *datatype = GL_UNSIGNED_BYTE; + *comps = 0; + return; + + case MESA_FORMAT_RGBA: + *datatype = CHAN_TYPE; + *comps = 4; + return; + case MESA_FORMAT_RGB: + *datatype = CHAN_TYPE; + *comps = 3; + return; + case MESA_FORMAT_LUMINANCE_ALPHA: + *datatype = CHAN_TYPE; + *comps = 2; + return; + case MESA_FORMAT_ALPHA: + case MESA_FORMAT_LUMINANCE: + case MESA_FORMAT_INTENSITY: + *datatype = CHAN_TYPE; + *comps = 1; + return; + + case MESA_FORMAT_RGBA_FLOAT32: + *datatype = GL_FLOAT; + *comps = 4; + return; + case MESA_FORMAT_RGBA_FLOAT16: + *datatype = GL_HALF_FLOAT_ARB; + *comps = 4; + return; + case MESA_FORMAT_RGB_FLOAT32: + *datatype = GL_FLOAT; + *comps = 3; + return; + case MESA_FORMAT_RGB_FLOAT16: + *datatype = GL_HALF_FLOAT_ARB; + *comps = 3; + return; + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32: + *datatype = GL_FLOAT; + *comps = 2; + return; + case MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16: + *datatype = GL_HALF_FLOAT_ARB; + *comps = 2; + return; + case MESA_FORMAT_ALPHA_FLOAT32: + case MESA_FORMAT_LUMINANCE_FLOAT32: + case MESA_FORMAT_INTENSITY_FLOAT32: + *datatype = GL_FLOAT; + *comps = 1; + return; + case MESA_FORMAT_ALPHA_FLOAT16: + case MESA_FORMAT_LUMINANCE_FLOAT16: + case MESA_FORMAT_INTENSITY_FLOAT16: + *datatype = GL_HALF_FLOAT_ARB; + *comps = 1; + return; + + default: + _mesa_problem(NULL, "bad format in _mesa_format_to_type_and_comps"); + *datatype = 0; + *comps = 1; + } +} diff --git a/src/mesa/main/texformat.h b/src/mesa/main/texformat.h index 55851db701..48f0fe99f2 100644 --- a/src/mesa/main/texformat.h +++ b/src/mesa/main/texformat.h @@ -239,4 +239,10 @@ extern const struct gl_texture_format * _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat, GLenum format, GLenum type ); + +extern void +_mesa_format_to_type_and_comps(const struct gl_texture_format *format, + GLenum *datatype, GLuint *comps); + + #endif -- cgit v1.2.3 From 3bae27fcc54187235a6962577580e838b012f9a5 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 8 Feb 2008 16:56:21 -0700 Subject: refactor code, export _mesa_generate_mipmap_level() --- src/mesa/main/mipmap.c | 98 ++++++++++++++++++++++++++++---------------------- src/mesa/main/mipmap.h | 11 ++++++ 2 files changed, 67 insertions(+), 42 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index 981da5dd89..ed7795aef9 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -841,6 +841,59 @@ make_2d_stack_mipmap(GLenum datatype, GLuint comps, GLint border, } +/** + * Down-sample a texture image to produce the next lower mipmap level. + */ +void +_mesa_generate_mipmap_level(GLenum target, + GLenum datatype, GLuint comps, + GLint border, + GLint srcWidth, GLint srcHeight, GLint srcDepth, + const GLubyte *srcData, + GLint dstWidth, GLint dstHeight, GLint dstDepth, + GLubyte *dstData) +{ + switch (target) { + case GL_TEXTURE_1D: + make_1d_mipmap(datatype, comps, border, + srcWidth, srcData, + dstWidth, dstData); + break; + case GL_TEXTURE_2D: + case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: + make_2d_mipmap(datatype, comps, border, + srcWidth, srcHeight, srcData, + dstWidth, dstHeight, dstData); + break; + case GL_TEXTURE_3D: + make_3d_mipmap(datatype, comps, border, + srcWidth, srcHeight, srcDepth, srcData, + dstWidth, dstHeight, dstDepth, dstData); + break; + case GL_TEXTURE_1D_ARRAY_EXT: + make_1d_stack_mipmap(datatype, comps, border, + srcWidth, srcData, + dstWidth, dstHeight, dstData); + break; + case GL_TEXTURE_2D_ARRAY_EXT: + make_2d_stack_mipmap(datatype, comps, border, + srcWidth, srcHeight, srcData, + dstWidth, dstHeight, dstDepth, dstData); + break; + case GL_TEXTURE_RECTANGLE_NV: + /* no mipmaps, do nothing */ + break; + default: + _mesa_problem(NULL, "bad target in _mesa_generate_mipmap_level"); + } +} + + /** * For GL_SGIX_generate_mipmap: * Generate a complete set of mipmaps from texObj's base-level image. @@ -1032,48 +1085,9 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target, dstData = (GLubyte *) dstImage->Data; } - /* - * We use simple 2x2 averaging to compute the next mipmap level. - */ - switch (target) { - case GL_TEXTURE_1D: - make_1d_mipmap(datatype, comps, border, - srcWidth, srcData, - dstWidth, dstData); - break; - case GL_TEXTURE_2D: - case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - make_2d_mipmap(datatype, comps, border, - srcWidth, srcHeight, srcData, - dstWidth, dstHeight, dstData); - break; - case GL_TEXTURE_3D: - make_3d_mipmap(datatype, comps, border, - srcWidth, srcHeight, srcDepth, srcData, - dstWidth, dstHeight, dstDepth, dstData); - break; - case GL_TEXTURE_1D_ARRAY_EXT: - make_1d_stack_mipmap(datatype, comps, border, - srcWidth, srcData, - dstWidth, dstHeight, dstData); - break; - case GL_TEXTURE_2D_ARRAY_EXT: - make_2d_stack_mipmap(datatype, comps, border, - srcWidth, srcHeight, srcData, - dstWidth, dstHeight, dstDepth, dstData); - break; - case GL_TEXTURE_RECTANGLE_NV: - /* no mipmaps, do nothing */ - break; - default: - _mesa_problem(ctx, "bad dimensions in _mesa_generate_mipmaps"); - return; - } + _mesa_generate_mipmap_level(target, datatype, comps, border, + srcWidth, srcHeight, srcDepth, srcData, + dstWidth, dstHeight, dstDepth, dstData); if (dstImage->IsCompressed) { GLubyte *temp; diff --git a/src/mesa/main/mipmap.h b/src/mesa/main/mipmap.h index 46e16902c8..b6491f5507 100644 --- a/src/mesa/main/mipmap.h +++ b/src/mesa/main/mipmap.h @@ -28,6 +28,17 @@ #include "mtypes.h" + +extern void +_mesa_generate_mipmap_level(GLenum target, + GLenum datatype, GLuint comps, + GLint border, + GLint srcWidth, GLint srcHeight, GLint srcDepth, + const GLubyte *srcData, + GLint dstWidth, GLint dstHeight, GLint dstDepth, + GLubyte *dstData); + + extern void _mesa_generate_mipmap(GLcontext *ctx, GLenum target, struct gl_texture_object *texObj); -- cgit v1.2.3 From 2e4077db16b5a55f1c3ee24ac7944bc05e0a7579 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 11 Feb 2008 08:18:43 -0700 Subject: gallium: fix fragment/vertex typo --- src/mesa/pipe/draw/draw_vertex_shader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 574ac67057..9c31df1e3e 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -304,7 +304,7 @@ draw_bind_vertex_shader(struct draw_context *draw, draw->vertex_shader = dvs; draw->num_vs_outputs = dvs->state->num_outputs; - /* specify the fragment program to interpret/execute */ + /* specify the vertex program to interpret/execute */ tgsi_exec_machine_init(&draw->machine, draw->vertex_shader->state->tokens, PIPE_MAX_SAMPLERS, -- cgit v1.2.3 From 789f2f15e867d6ef9f635d0ab59e213b35a2960c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 11 Feb 2008 08:56:18 -0700 Subject: gallium: update comment about buffer map flags --- src/mesa/pipe/p_winsys.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h index 95e3684008..1e81eebd78 100644 --- a/src/mesa/pipe/p_winsys.h +++ b/src/mesa/pipe/p_winsys.h @@ -112,7 +112,7 @@ struct pipe_winsys /** * Map the entire data store of a buffer object into the client's address. - * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. + * flags is bitmask of PIPE_BUFFER_USAGE_CPU_READ/WRITE flags. */ void *(*buffer_map)( struct pipe_winsys *sws, struct pipe_buffer *buf, -- cgit v1.2.3 From b52ce6341f877a0a1cfd3c50c07eeb1c156ae89e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 11 Feb 2008 09:33:28 -0700 Subject: added _mesa_tex_target_to_face() --- src/mesa/main/teximage.c | 20 +++++++++++++------- src/mesa/main/teximage.h | 5 +++++ 2 files changed, 18 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c index 09ec0d4553..5c96be9216 100644 --- a/src/mesa/main/teximage.c +++ b/src/mesa/main/teximage.c @@ -595,8 +595,12 @@ is_compressed_format(GLcontext *ctx, GLenum internalFormat) } -static GLuint -texture_face(GLenum target) +/** + * For cube map faces, return a face index in [0,5]. + * For other targets return 0; + */ +GLuint +_mesa_tex_target_to_face(GLenum target) { if (target >= GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB && target <= GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB) @@ -625,6 +629,7 @@ _mesa_set_tex_image(struct gl_texture_object *tObj, { ASSERT(tObj); ASSERT(texImage); + /* XXX simplify this with _mesa_tex_target_to_face() */ switch (target) { case GL_TEXTURE_1D: case GL_TEXTURE_2D: @@ -828,6 +833,7 @@ _mesa_select_tex_image(GLcontext *ctx, const struct gl_texture_object *texObj, if (level < 0 || level >= MAX_TEXTURE_LEVELS) return NULL; + /* XXX simplify this with _mesa_tex_target_to_face() */ switch (target) { case GL_TEXTURE_1D: case GL_PROXY_TEXTURE_1D: @@ -2424,7 +2430,7 @@ _mesa_TexImage1D( GLenum target, GLint level, GLint internalFormat, struct gl_texture_unit *texUnit; struct gl_texture_object *texObj; struct gl_texture_image *texImage; - const GLuint face = texture_face(target); + const GLuint face = _mesa_tex_target_to_face(target); if (texture_error_check(ctx, target, level, internalFormat, format, type, 1, postConvWidth, 1, 1, border)) { @@ -2527,7 +2533,7 @@ _mesa_TexImage2D( GLenum target, GLint level, GLint internalFormat, struct gl_texture_unit *texUnit; struct gl_texture_object *texObj; struct gl_texture_image *texImage; - const GLuint face = texture_face(target); + const GLuint face = _mesa_tex_target_to_face(target); if (texture_error_check(ctx, target, level, internalFormat, format, type, 2, postConvWidth, postConvHeight, @@ -2629,7 +2635,7 @@ _mesa_TexImage3D( GLenum target, GLint level, GLint internalFormat, struct gl_texture_unit *texUnit; struct gl_texture_object *texObj; struct gl_texture_image *texImage; - const GLuint face = texture_face(target); + const GLuint face = _mesa_tex_target_to_face(target); if (texture_error_check(ctx, target, level, (GLint) internalFormat, format, type, 3, width, height, depth, border)) { @@ -2897,7 +2903,7 @@ _mesa_CopyTexImage1D( GLenum target, GLint level, struct gl_texture_object *texObj; struct gl_texture_image *texImage; GLsizei postConvWidth = width; - const GLuint face = texture_face(target); + const GLuint face = _mesa_tex_target_to_face(target); GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); @@ -2960,7 +2966,7 @@ _mesa_CopyTexImage2D( GLenum target, GLint level, GLenum internalFormat, struct gl_texture_object *texObj; struct gl_texture_image *texImage; GLsizei postConvWidth = width, postConvHeight = height; - const GLuint face = texture_face(target); + const GLuint face = _mesa_tex_target_to_face(target); GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); diff --git a/src/mesa/main/teximage.h b/src/mesa/main/teximage.h index f2cad7eb2d..b718c0046d 100644 --- a/src/mesa/main/teximage.h +++ b/src/mesa/main/teximage.h @@ -107,6 +107,11 @@ _mesa_test_proxy_teximage(GLcontext *ctx, GLenum target, GLint level, GLint width, GLint height, GLint depth, GLint border); +extern GLuint +_mesa_tex_target_to_face(GLenum target); + + + /** * Lock a texture for updating. See also _mesa_lock_context_textures(). */ -- cgit v1.2.3 From cfe9e66f2bc596c43760911e7c1604bb32cdee28 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 11 Feb 2008 09:35:28 -0700 Subject: gallium: use _mesa_tex_target_to_face() --- src/mesa/state_tracker/st_cb_texture.c | 22 +--------------------- 1 file changed, 1 insertion(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 3350254654..992723afba 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -379,26 +379,6 @@ guess_and_alloc_texture(struct st_context *st, } - - -static GLuint -target_to_face(GLenum target) -{ - switch (target) { - case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: - return ((GLuint) target - (GLuint) GL_TEXTURE_CUBE_MAP_POSITIVE_X); - default: - return 0; - } -} - - - /* There are actually quite a few combinations this will work for, * more than what I've listed here. */ @@ -512,7 +492,7 @@ st_TexImage(GLcontext * ctx, DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); - stImage->face = target_to_face(target); + stImage->face = _mesa_tex_target_to_face(target); stImage->level = level; if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { -- cgit v1.2.3 From 212b27d33f94eeb25ba9cbc58f9e41295a29d2c9 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 11 Feb 2008 09:38:36 -0700 Subject: gallium: strip borders from textures passed to st_TexImage. Manipulate the unpack params to skip the border. Gallium drivers won't support texture borders. --- src/mesa/state_tracker/st_cb_texture.c | 53 ++++++++++++++++++++++++++++++++-- 1 file changed, 50 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 992723afba..7099ec33b9 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -467,6 +467,43 @@ try_pbo_upload(GLcontext *ctx, } +/** + * Adjust pixel unpack params and image dimensions to strip off the + * texture border. + * Gallium doesn't support texture borders. They've seldem been used + * and seldom been implemented correctly anyway. + * \param unpackNew returns the new pixel unpack parameters + */ +static void +strip_texture_border(GLint border, + GLint *width, GLint *height, GLint *depth, + const struct gl_pixelstore_attrib *unpack, + struct gl_pixelstore_attrib *unpackNew) +{ + assert(border > 0); /* sanity check */ + + *unpackNew = *unpack; + + if (unpackNew->RowLength == 0) + unpackNew->RowLength = *width; + + if (depth && unpackNew->ImageHeight == 0) + unpackNew->ImageHeight = *height; + + unpackNew->SkipPixels += border; + if (height) + unpackNew->SkipRows += border; + if (depth) + unpackNew->SkipImages += border; + + assert(*width >= 3); + *width = *width - 2 * border; + if (height && *height >= 3) + *height = *height - 2 * border; + if (depth && *depth >= 3) + *depth = *depth - 2 * border; +} + static void st_TexImage(GLcontext * ctx, @@ -483,15 +520,25 @@ st_TexImage(GLcontext * ctx, { struct st_texture_object *stObj = st_texture_object(texObj); struct st_texture_image *stImage = st_texture_image(texImage); - GLint postConvWidth = width; - GLint postConvHeight = height; + GLint postConvWidth, postConvHeight; GLint texelBytes, sizeInBytes; GLuint dstRowStride; - + struct gl_pixelstore_attrib unpackNB; DBG("%s target %s level %d %dx%dx%d border %d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), level, width, height, depth, border); + /* gallium does not support texture borders, strip it off */ + if (border) { + strip_texture_border(border, &width, &height, &depth, + unpack, &unpackNB); + unpack = &unpackNB; + border = 0; + } + + postConvWidth = width; + postConvHeight = height; + stImage->face = _mesa_tex_target_to_face(target); stImage->level = level; -- cgit v1.2.3 From e4026167d70f0aa0bc0cd2546ef9b41f7b167735 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 11 Feb 2008 09:42:02 -0700 Subject: gallium: implement software fallback for mipmap generation This is used when we can't render to the surface type of the texture (such as luminance/alpha). --- src/mesa/state_tracker/st_gen_mipmap.c | 68 ++++++++++++++++++++++++++++++---- 1 file changed, 60 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index a6ac9a55fb..5c00392af7 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -29,11 +29,14 @@ #include "main/imports.h" #include "main/mipmap.h" #include "main/teximage.h" +#include "main/texformat.h" #include "shader/prog_instruction.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" #include "pipe/cso_cache/cso_cache.h" #include "st_context.h" @@ -239,15 +242,18 @@ draw_quad(GLcontext *ctx) */ static boolean st_render_mipmap(struct st_context *st, + GLenum target, struct pipe_texture *pt, uint baseLevel, uint lastLevel) { struct pipe_context *pipe = st->pipe; struct pipe_framebuffer_state fb; - const uint face = 0, zslice = 0; + const uint face = _mesa_tex_target_to_face(target), zslice = 0; const uint first_level_save = pt->first_level; uint dstLevel; + assert(target != GL_TEXTURE_3D); /* not done yet */ + /* check if we can render in the texture's format */ if (!pipe->is_format_supported(pipe, pt->format, PIPE_SURFACE)) { return FALSE; @@ -307,6 +313,56 @@ st_render_mipmap(struct st_context *st, } +static void +fallback_generate_mipmap(GLcontext *ctx, GLenum target, + struct gl_texture_object *texObj) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct pipe_winsys *ws = pipe->winsys; + struct pipe_texture *pt = st_get_texobj_texture(texObj); + const uint baseLevel = texObj->BaseLevel; + const uint lastLevel = pt->last_level; + const uint face = _mesa_tex_target_to_face(target), zslice = 0; + uint dstLevel; + GLenum datatype; + GLuint comps; + + assert(target != GL_TEXTURE_3D); /* not done yet */ + + _mesa_format_to_type_and_comps(texObj->Image[face][baseLevel]->TexFormat, + &datatype, &comps); + + for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { + const uint srcLevel = dstLevel - 1; + struct pipe_surface *srcSurf, *dstSurf; + const ubyte *srcData; + ubyte *dstData; + + srcSurf = pipe->get_tex_surface(pipe, pt, face, srcLevel, zslice); + dstSurf = pipe->get_tex_surface(pipe, pt, face, dstLevel, zslice); + + srcData = (ubyte *) ws->buffer_map(ws, srcSurf->buffer, + PIPE_BUFFER_USAGE_CPU_READ) + + srcSurf->offset; + dstData = (ubyte *) ws->buffer_map(ws, dstSurf->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE) + + dstSurf->offset; + + _mesa_generate_mipmap_level(target, datatype, comps, + 0 /*border*/, + pt->width[srcLevel], pt->height[srcLevel], pt->depth[srcLevel], + srcData, + pt->width[dstLevel], pt->height[dstLevel], pt->depth[dstLevel], + dstData); + + ws->buffer_unmap(ws, srcSurf->buffer); + ws->buffer_unmap(ws, dstSurf->buffer); + + pipe_surface_reference(&srcSurf, NULL); + pipe_surface_reference(&dstSurf, NULL); + } +} + void st_generate_mipmap(GLcontext *ctx, GLenum target, @@ -318,13 +374,11 @@ st_generate_mipmap(GLcontext *ctx, GLenum target, const uint lastLevel = pt->last_level; uint dstLevel; - if (!st_render_mipmap(st, pt, baseLevel, lastLevel)) { - abort(); - /* XXX the following won't really work at this time */ - _mesa_generate_mipmap(ctx, target, texObj); - return; + if (!st_render_mipmap(st, target, pt, baseLevel, lastLevel)) { + fallback_generate_mipmap(ctx, target, texObj); } + /* Fill in the Mesa gl_texture_image fields */ for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; const struct gl_texture_image *srcImage @@ -336,7 +390,6 @@ st_generate_mipmap(GLcontext *ctx, GLenum target, uint dstDepth = pt->depth[dstLevel]; uint border = srcImage->Border; - dstImage = _mesa_get_tex_image(ctx, texObj, target, dstLevel); if (!dstImage) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "generating mipmaps"); @@ -359,5 +412,4 @@ st_generate_mipmap(GLcontext *ctx, GLenum target, stImage = (struct st_texture_image *) dstImage; stImage->pt = pt; } - } -- cgit v1.2.3 From c22d9152e33792ea58426c53bc9b96bf552b0b44 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 11 Feb 2008 09:43:30 -0700 Subject: code refactoring, new next_mipmap_level_size() function --- src/mesa/main/mipmap.c | 75 +++++++++++++++++++++++++++++++++----------------- 1 file changed, 50 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index ed7795aef9..721a227bb5 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -894,6 +894,51 @@ _mesa_generate_mipmap_level(GLenum target, } +/** + * compute next (level+1) image size + * \return GL_FALSE if no smaller size can be generated (eg. src is 1x1x1 size) + */ +static GLboolean +next_mipmap_level_size(GLenum target, GLint border, + GLint srcWidth, GLint srcHeight, GLint srcDepth, + GLint *dstWidth, GLint *dstHeight, GLint *dstDepth) +{ + if (srcWidth - 2 * border > 1) { + *dstWidth = (srcWidth - 2 * border) / 2 + 2 * border; + } + else { + *dstWidth = srcWidth; /* can't go smaller */ + } + + if ((srcHeight - 2 * border > 1) && + (target != GL_TEXTURE_1D_ARRAY_EXT)) { + *dstHeight = (srcHeight - 2 * border) / 2 + 2 * border; + } + else { + *dstHeight = srcHeight; /* can't go smaller */ + } + + if ((srcDepth - 2 * border > 1) && + (target != GL_TEXTURE_2D_ARRAY_EXT)) { + *dstDepth = (srcDepth - 2 * border) / 2 + 2 * border; + } + else { + *dstDepth = srcDepth; /* can't go smaller */ + } + + if (*dstWidth == srcWidth && + *dstHeight == srcHeight && + *dstDepth == srcDepth) { + return GL_FALSE; + } + else { + return GL_TRUE; + } +} + + + + /** * For GL_SGIX_generate_mipmap: * Generate a complete set of mipmaps from texObj's base-level image. @@ -982,6 +1027,7 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target, GLint srcWidth, srcHeight, srcDepth; GLint dstWidth, dstHeight, dstDepth; GLint border, bytesPerTexel; + GLboolean nextLevel; /* get src image parameters */ srcImage = _mesa_select_tex_image(ctx, texObj, target, level); @@ -991,31 +1037,10 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target, srcDepth = srcImage->Depth; border = srcImage->Border; - /* compute next (level+1) image size */ - if (srcWidth - 2 * border > 1) { - dstWidth = (srcWidth - 2 * border) / 2 + 2 * border; - } - else { - dstWidth = srcWidth; /* can't go smaller */ - } - if ((srcHeight - 2 * border > 1) && - (texObj->Target != GL_TEXTURE_1D_ARRAY_EXT)) { - dstHeight = (srcHeight - 2 * border) / 2 + 2 * border; - } - else { - dstHeight = srcHeight; /* can't go smaller */ - } - if ((srcDepth - 2 * border > 1) && - (texObj->Target != GL_TEXTURE_2D_ARRAY_EXT)) { - dstDepth = (srcDepth - 2 * border) / 2 + 2 * border; - } - else { - dstDepth = srcDepth; /* can't go smaller */ - } - - if (dstWidth == srcWidth && - dstHeight == srcHeight && - dstDepth == srcDepth) { + nextLevel = next_mipmap_level_size(target, border, + srcWidth, srcHeight, srcDepth, + &dstWidth, &dstHeight, &dstDepth); + if (!nextLevel) { /* all done */ if (srcImage->IsCompressed) { _mesa_free((void *) srcData); -- cgit v1.2.3 From 3d0fd8a6cf294d58df2d2ec4139192416a1a4078 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 11 Feb 2008 09:50:21 -0700 Subject: gallium: comments about mipmap gen --- src/mesa/state_tracker/st_gen_mipmap.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 5c00392af7..f6af37cfac 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -273,6 +273,10 @@ st_render_mipmap(struct st_context *st, pipe->bind_fs_state(pipe, stfp->fs->data); pipe->bind_vs_state(pipe, stvp->cso->data); + /* + * XXX for small mipmap levels, it may be faster to use the software + * fallback path... + */ for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; @@ -348,6 +352,7 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, PIPE_BUFFER_USAGE_CPU_WRITE) + dstSurf->offset; + /* XXX need to take stride/pitch info into account... */ _mesa_generate_mipmap_level(target, datatype, comps, 0 /*border*/, pt->width[srcLevel], pt->height[srcLevel], pt->depth[srcLevel], -- cgit v1.2.3 From b61b1a295b13a0ff2cf98c8d07e62147d71c08b9 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 11 Feb 2008 10:59:40 -0700 Subject: gallium: take pitch/stride into account in mipmap generation --- src/mesa/main/mipmap.c | 85 +++++++++++++++++++++------------- src/mesa/main/mipmap.h | 2 + src/mesa/state_tracker/st_gen_mipmap.c | 3 +- 3 files changed, 57 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index 721a227bb5..d3d1958951 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -447,23 +447,29 @@ make_1d_mipmap(GLenum datatype, GLuint comps, GLint border, /** - * XXX need to use the tex image's row stride! + * Strides are in bytes. If zero, it'll be computed as width * bpp. */ static void make_2d_mipmap(GLenum datatype, GLuint comps, GLint border, - GLint srcWidth, GLint srcHeight, const GLubyte *srcPtr, - GLint dstWidth, GLint dstHeight, GLubyte *dstPtr) + GLint srcWidth, GLint srcHeight, + GLint srcRowStride, const GLubyte *srcPtr, + GLint dstWidth, GLint dstHeight, + GLint dstRowStride, GLubyte *dstPtr) { const GLint bpt = bytes_per_pixel(datatype, comps); const GLint srcWidthNB = srcWidth - 2 * border; /* sizes w/out border */ const GLint dstWidthNB = dstWidth - 2 * border; const GLint dstHeightNB = dstHeight - 2 * border; - const GLint srcRowStride = bpt * srcWidth; - const GLint dstRowStride = bpt * dstWidth; const GLubyte *srcA, *srcB; GLubyte *dst; GLint row; + if (!srcRowStride) + srcRowStride = bpt * srcWidth; + + if (!dstRowStride) + dstRowStride = bpt * dstWidth; + /* Compute src and dst pointers, skipping any border */ srcA = srcPtr + border * ((srcWidth + 1) * bpt); if (srcHeight > 1) @@ -535,8 +541,10 @@ make_2d_mipmap(GLenum datatype, GLuint comps, GLint border, static void make_3d_mipmap(GLenum datatype, GLuint comps, GLint border, GLint srcWidth, GLint srcHeight, GLint srcDepth, + GLint srcRowStride, const GLubyte *srcPtr, GLint dstWidth, GLint dstHeight, GLint dstDepth, + GLint dstRowStride, GLubyte *dstPtr) { const GLint bpt = bytes_per_pixel(datatype, comps); @@ -548,7 +556,6 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border, GLvoid *tmpRowA, *tmpRowB; GLint img, row; GLint bytesPerSrcImage, bytesPerDstImage; - GLint bytesPerSrcRow, bytesPerDstRow; GLint srcImageOffset, srcRowOffset; (void) srcDepthNB; /* silence warnings */ @@ -566,8 +573,10 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border, bytesPerSrcImage = srcWidth * srcHeight * bpt; bytesPerDstImage = dstWidth * dstHeight * bpt; - bytesPerSrcRow = srcWidth * bpt; - bytesPerDstRow = dstWidth * bpt; + if (!srcRowStride) + srcRowStride = srcWidth * bpt; + if (!dstRowStride) + dstRowStride = dstWidth * bpt; /* Offset between adjacent src images to be averaged together */ srcImageOffset = (srcDepth == dstDepth) ? 0 : bytesPerSrcImage; @@ -591,13 +600,13 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border, for (img = 0; img < dstDepthNB; img++) { /* first source image pointer, skipping border */ const GLubyte *imgSrcA = srcPtr - + (bytesPerSrcImage + bytesPerSrcRow + border) * bpt * border + + (bytesPerSrcImage + srcRowStride + border) * bpt * border + img * (bytesPerSrcImage + srcImageOffset); /* second source image pointer, skipping border */ const GLubyte *imgSrcB = imgSrcA + srcImageOffset; /* address of the dest image, skipping border */ GLubyte *imgDst = dstPtr - + (bytesPerDstImage + bytesPerDstRow + border) * bpt * border + + (bytesPerDstImage + dstRowStride + border) * bpt * border + img * bytesPerDstImage; /* setup the four source row pointers and the dest row pointer */ @@ -618,11 +627,11 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border, do_row(datatype, comps, srcWidthNB, tmpRowA, tmpRowB, dstWidthNB, dstImgRow); /* advance to next rows */ - srcImgARowA += bytesPerSrcRow + srcRowOffset; - srcImgARowB += bytesPerSrcRow + srcRowOffset; - srcImgBRowA += bytesPerSrcRow + srcRowOffset; - srcImgBRowB += bytesPerSrcRow + srcRowOffset; - dstImgRow += bytesPerDstRow; + srcImgARowA += srcRowStride + srcRowOffset; + srcImgARowB += srcRowStride + srcRowOffset; + srcImgBRowA += srcRowStride + srcRowOffset; + srcImgBRowB += srcRowStride + srcRowOffset; + dstImgRow += dstRowStride; } } @@ -632,12 +641,14 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border, /* Luckily we can leverage the make_2d_mipmap() function here! */ if (border > 0) { /* do front border image */ - make_2d_mipmap(datatype, comps, 1, srcWidth, srcHeight, srcPtr, - dstWidth, dstHeight, dstPtr); + make_2d_mipmap(datatype, comps, 1, srcWidth, srcHeight, 0, srcPtr, + dstWidth, dstHeight, 0, dstPtr); /* do back border image */ make_2d_mipmap(datatype, comps, 1, srcWidth, srcHeight, + 0, srcPtr + bytesPerSrcImage * (srcDepth - 1), dstWidth, dstHeight, + 0, dstPtr + bytesPerDstImage * (dstDepth - 1)); /* do four remaining border edges that span the image slices */ if (srcDepth == dstDepth) { @@ -653,9 +664,9 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border, /* do border along [img][row=dstHeight-1][col=0] */ src = srcPtr + (img * 2 + 1) * bytesPerSrcImage - + (srcHeight - 1) * bytesPerSrcRow; + + (srcHeight - 1) * srcRowStride; dst = dstPtr + (img + 1) * bytesPerDstImage - + (dstHeight - 1) * bytesPerDstRow; + + (dstHeight - 1) * dstRowStride; MEMCPY(dst, src, bpt); /* do border along [img][row=0][col=dstWidth-1] */ @@ -687,9 +698,9 @@ make_3d_mipmap(GLenum datatype, GLuint comps, GLint border, /* do border along [img][row=dstHeight-1][col=0] */ src = srcPtr + (img * 2 + 1) * bytesPerSrcImage - + (srcHeight - 1) * bytesPerSrcRow; + + (srcHeight - 1) * srcRowStride; dst = dstPtr + (img + 1) * bytesPerDstImage - + (dstHeight - 1) * bytesPerDstRow; + + (dstHeight - 1) * dstRowStride; do_row(datatype, comps, 1, src, src + srcImageOffset, 1, dst); /* do border along [img][row=0][col=dstWidth-1] */ @@ -755,8 +766,11 @@ make_1d_stack_mipmap(GLenum datatype, GLuint comps, GLint border, */ static void make_2d_stack_mipmap(GLenum datatype, GLuint comps, GLint border, - GLint srcWidth, GLint srcHeight, const GLubyte *srcPtr, + GLint srcWidth, GLint srcHeight, + GLint srcRowStride, + const GLubyte *srcPtr, GLint dstWidth, GLint dstHeight, GLint dstDepth, + GLint dstRowStride, GLubyte *dstPtr) { const GLint bpt = bytes_per_pixel(datatype, comps); @@ -764,13 +778,17 @@ make_2d_stack_mipmap(GLenum datatype, GLuint comps, GLint border, const GLint dstWidthNB = dstWidth - 2 * border; const GLint dstHeightNB = dstHeight - 2 * border; const GLint dstDepthNB = dstDepth - 2 * border; - const GLint srcRowStride = bpt * srcWidth; - const GLint dstRowStride = bpt * dstWidth; const GLubyte *srcA, *srcB; GLubyte *dst; GLint layer; GLint row; + if (!srcRowStride) + srcRowStride = bpt * srcWidth; + + if (!dstRowStride) + dstRowStride = bpt * dstWidth; + /* Compute src and dst pointers, skipping any border */ srcA = srcPtr + border * ((srcWidth + 1) * bpt); if (srcHeight > 1) @@ -849,8 +867,10 @@ _mesa_generate_mipmap_level(GLenum target, GLenum datatype, GLuint comps, GLint border, GLint srcWidth, GLint srcHeight, GLint srcDepth, + GLint srcRowStride, const GLubyte *srcData, GLint dstWidth, GLint dstHeight, GLint dstDepth, + GLint dstRowStride, GLubyte *dstData) { switch (target) { @@ -867,13 +887,13 @@ _mesa_generate_mipmap_level(GLenum target, case GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB: case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB: make_2d_mipmap(datatype, comps, border, - srcWidth, srcHeight, srcData, - dstWidth, dstHeight, dstData); + srcWidth, srcHeight, srcRowStride, srcData, + dstWidth, dstHeight, dstRowStride, dstData); break; case GL_TEXTURE_3D: make_3d_mipmap(datatype, comps, border, - srcWidth, srcHeight, srcDepth, srcData, - dstWidth, dstHeight, dstDepth, dstData); + srcWidth, srcHeight, srcDepth, srcRowStride, srcData, + dstWidth, dstHeight, dstDepth, dstRowStride, dstData); break; case GL_TEXTURE_1D_ARRAY_EXT: make_1d_stack_mipmap(datatype, comps, border, @@ -882,8 +902,8 @@ _mesa_generate_mipmap_level(GLenum target, break; case GL_TEXTURE_2D_ARRAY_EXT: make_2d_stack_mipmap(datatype, comps, border, - srcWidth, srcHeight, srcData, - dstWidth, dstHeight, dstDepth, dstData); + srcWidth, srcHeight, srcRowStride, srcData, + dstWidth, dstHeight, dstDepth, dstRowStride, dstData); break; case GL_TEXTURE_RECTANGLE_NV: /* no mipmaps, do nothing */ @@ -1110,9 +1130,10 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target, dstData = (GLubyte *) dstImage->Data; } + /* Note, 0 indicates default row strides */ _mesa_generate_mipmap_level(target, datatype, comps, border, - srcWidth, srcHeight, srcDepth, srcData, - dstWidth, dstHeight, dstDepth, dstData); + srcWidth, srcHeight, srcDepth, 0, srcData, + dstWidth, dstHeight, dstDepth, 0, dstData); if (dstImage->IsCompressed) { GLubyte *temp; diff --git a/src/mesa/main/mipmap.h b/src/mesa/main/mipmap.h index b6491f5507..44ecdddb27 100644 --- a/src/mesa/main/mipmap.h +++ b/src/mesa/main/mipmap.h @@ -34,8 +34,10 @@ _mesa_generate_mipmap_level(GLenum target, GLenum datatype, GLuint comps, GLint border, GLint srcWidth, GLint srcHeight, GLint srcDepth, + GLint srcRowStride, const GLubyte *srcData, GLint dstWidth, GLint dstHeight, GLint dstDepth, + GLint dstRowStride, GLubyte *dstData); diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index f6af37cfac..c152c59905 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -352,12 +352,13 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, PIPE_BUFFER_USAGE_CPU_WRITE) + dstSurf->offset; - /* XXX need to take stride/pitch info into account... */ _mesa_generate_mipmap_level(target, datatype, comps, 0 /*border*/, pt->width[srcLevel], pt->height[srcLevel], pt->depth[srcLevel], + srcSurf->pitch * srcSurf->cpp, /* stride in bytes */ srcData, pt->width[dstLevel], pt->height[dstLevel], pt->depth[dstLevel], + dstSurf->pitch * dstSurf->cpp, /* stride in bytes */ dstData); ws->buffer_unmap(ws, srcSurf->buffer); -- cgit v1.2.3 From 4da1cdf78fa3b954840650fa46cf72da5daf149f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Feb 2008 14:53:25 -0700 Subject: gallium: clean-up, simplification of mipmapped textures Remove pipe_texture->first_level (always implicitly zero). This means there's never any unused mipmap levels at the top. In the state tracker, we no longer re-layout mipmapped textures if the MinLod/MaxLod texture parameters change. It's up to the driver to obey the pipe_sampler->min/max_lod clamps. --- src/mesa/pipe/i915simple/i915_state.c | 7 ++++ src/mesa/pipe/i915simple/i915_state_sampler.c | 2 +- src/mesa/pipe/i915simple/i915_texture.c | 30 +++++++-------- src/mesa/pipe/i965simple/brw_tex_layout.c | 8 ++-- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 2 +- src/mesa/pipe/p_state.h | 14 ++----- src/mesa/pipe/softpipe/sp_state_sampler.c | 4 ++ src/mesa/pipe/softpipe/sp_tex_sample.c | 18 +++------ src/mesa/pipe/softpipe/sp_texture.c | 4 +- src/mesa/state_tracker/st_cb_texture.c | 51 +++++++++---------------- src/mesa/state_tracker/st_gen_mipmap.c | 43 ++++++++++++--------- src/mesa/state_tracker/st_texture.c | 35 ++++++++++++----- src/mesa/state_tracker/st_texture.h | 6 +-- 13 files changed, 115 insertions(+), 109 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c index 950ea52d60..abd5571b88 100644 --- a/src/mesa/pipe/i915simple/i915_state.c +++ b/src/mesa/pipe/i915simple/i915_state.c @@ -250,6 +250,13 @@ i915_create_sampler_state(struct pipe_context *pipe, if (sampler->normalized_coords) cso->state[1] |= SS3_NORMALIZED_COORDS; + if (0) /* XXX not tested yet */ + { + int minlod = (int) (16.0 * sampler->min_lod); + minlod = CLAMP(minlod, 0, 16 * 11); + cso->state[1] |= (minlod << SS3_MIN_LOD_SHIFT); + } + { ubyte r = float_to_ubyte(sampler->border_color[0]); ubyte g = float_to_ubyte(sampler->border_color[1]); diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c index 0dbbc5241d..9c1a5bbbd6 100644 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ b/src/mesa/pipe/i915simple/i915_state_sampler.c @@ -185,7 +185,7 @@ i915_update_texture(struct i915_context *i915, uint unit, const struct pipe_texture *pt = &tex->base; uint format, pitch; const uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; - const uint num_levels = pt->last_level - pt->first_level; + const uint num_levels = pt->last_level; assert(tex); assert(width); diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c index 6faeab134a..6d37ae3d74 100644 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ b/src/mesa/pipe/i915simple/i915_texture.c @@ -118,11 +118,11 @@ i945_miptree_layout_2d( struct i915_texture *tex ) tex->pitch = pt->width[0]; /* May need to adjust pitch to accomodate the placement of - * the 2nd mipmap. This occurs when the alignment + * the 2nd mipmap level. This occurs when the alignment * constraints of mipmap placement push the right edge of the - * 2nd mipmap out past the width of its parent. + * 2nd mipmap level out past the width of its parent. */ - if (pt->first_level != pt->last_level) { + if (pt->last_level > 0) { unsigned mip1_width = align_int(minify(pt->width[0]), align_w) + minify(minify(pt->width[0])); @@ -136,7 +136,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) tex->pitch = align_int(tex->pitch * pt->cpp, 4) / pt->cpp; tex->total_height = 0; - for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + for (level = 0; level <= pt->last_level; level++) { unsigned img_height; i915_miptree_set_level_info(tex, level, 1, x, y, width, height, 1); @@ -152,9 +152,9 @@ i945_miptree_layout_2d( struct i915_texture *tex ) */ tex->total_height = MAX2(tex->total_height, y + img_height); - /* Layout_below: step right after second mipmap. + /* Layout_below: step right after second mipmap level. */ - if (level == pt->first_level + 1) { + if (level == 1) { x += align_int(width, align_w); } else { @@ -204,7 +204,7 @@ i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; tex->total_height = dim * 4; - for (level = pt->first_level; level <= pt->last_level; level++) { + for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_level_info(tex, level, 6, 0, 0, /*OLD: tex->pitch, tex->total_height,*/ @@ -219,7 +219,7 @@ i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) unsigned y = initial_offsets[face][1] * dim; unsigned d = dim; - for (level = pt->first_level; level <= pt->last_level; level++) { + for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_image_offset(tex, level, face, x, y); d >>= 1; x += step_offsets[face][0] * d; @@ -240,7 +240,7 @@ i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) /* XXX: hardware expects/requires 9 levels at minimum. */ - for (level = pt->first_level; level <= MAX2(8, pt->last_level); + for (level = 0; level <= MAX2(8, pt->last_level); level++) { i915_miptree_set_level_info(tex, level, depth, 0, tex->total_height, width, height, depth); @@ -256,7 +256,7 @@ i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) /* Fixup depth image_offsets: */ depth = pt->depth[0]; - for (level = pt->first_level; level <= pt->last_level; level++) { + for (level = 0; level <= pt->last_level; level++) { unsigned i; for (i = 0; i < depth; i++) i915_miptree_set_image_offset(tex, level, i, @@ -282,7 +282,7 @@ i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; tex->total_height = 0; - for (level = pt->first_level; level <= pt->last_level; level++) { + for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_level_info(tex, level, 1, 0, tex->total_height, width, height, 1); @@ -337,7 +337,7 @@ i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) /* Set all the levels to effectively occupy the whole rectangular region. */ - for (level = pt->first_level; level <= pt->last_level; level++) { + for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_level_info(tex, level, 6, 0, 0, lvlWidth, lvlHeight, 1); @@ -355,12 +355,12 @@ i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) y = tex->total_height - 4; x = (face - 4) * 8; } - else if (dim < 4 && (face > 0 || pt->first_level > 0)) { + else if (dim < 4 && (face > 0)) { y = tex->total_height - 4; x = face * 8; } - for (level = pt->first_level; level <= pt->last_level; level++) { + for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_image_offset(tex, level, face, x, y); d >>= 1; @@ -418,7 +418,7 @@ i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) pack_x_pitch = tex->pitch; pack_x_nr = 1; - for (level = pt->first_level; level <= pt->last_level; level++) { + for (level = 0; level <= pt->last_level; level++) { unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; int x = 0; int y = 0; diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.c b/src/mesa/pipe/i965simple/brw_tex_layout.c index 405fd1f794..90561f1307 100644 --- a/src/mesa/pipe/i965simple/brw_tex_layout.c +++ b/src/mesa/pipe/i965simple/brw_tex_layout.c @@ -146,7 +146,7 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) * constraints of mipmap placement push the right edge of the * 2nd mipmap out past the width of its parent. */ - if (pt->first_level != pt->last_level) { + if (pt->last_level > 0) { unsigned mip1_width; if (pt->compressed) { @@ -168,7 +168,7 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) tex->pitch = align(tex->pitch * pt->cpp, 4) / pt->cpp; tex->total_height = 0; - for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + for (level = 0; level <= pt->last_level; level++) { unsigned img_height; intel_miptree_set_level_info(tex, level, 1, x, y, width, @@ -187,7 +187,7 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) /* Layout_below: step right after second mipmap. */ - if (level == pt->first_level + 1) { + if (level == 1) { x += align(width, align_w); } else { @@ -234,7 +234,7 @@ static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture pack_x_pitch = tex->pitch; pack_x_nr = 1; - for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + for (level = 0; level <= pt->last_level; level++) { unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; int x = 0; int y = 0; diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c index cbb4f2efd3..d16d919bce 100644 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ b/src/mesa/pipe/i965simple/brw_wm_surface_state.c @@ -154,7 +154,7 @@ void brw_update_texture_surface( struct brw_context *brw, /* Updated in emit_reloc */ surf.ss1.base_addr = brw_buffer_offset( brw, tObj->buffer ); - surf.ss2.mip_count = tObj->base.last_level - tObj->base.first_level; + surf.ss2.mip_count = tObj->base.last_level; surf.ss2.width = tObj->base.width[0] - 1; surf.ss2.height = tObj->base.height[0] - 1; diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 83ca43f678..4d3a6b2f41 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -234,14 +234,9 @@ struct pipe_sampler_state unsigned compare_mode:1; /**< PIPE_TEX_COMPARE_x */ unsigned compare_func:3; /**< PIPE_FUNC_x */ unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */ - float shadow_ambient; /**< shadow test fail color/intensity */ - float min_lod; - float max_lod; - float lod_bias; -#if 0 /* need these? */ - int BaseLevel; /**< min mipmap level, OpenGL 1.2 */ - int MaxLevel; /**< max mipmap level, OpenGL 1.2 */ -#endif + float shadow_ambient; /**< shadow test fail color/intensity */ + float lod_bias; /**< LOD/lambda bias */ + float min_lod, max_lod; /**< LOD clamp range, after bias */ float border_color[4]; float max_anisotropy; }; @@ -277,8 +272,7 @@ struct pipe_texture enum pipe_texture_target target; /**< PIPE_TEXTURE_x */ enum pipe_format format; /**< PIPE_FORMAT_x */ - unsigned first_level; - unsigned last_level; + unsigned last_level; /**< Index of last mipmap level present/defined */ unsigned width[PIPE_MAX_TEXTURE_LEVELS]; unsigned height[PIPE_MAX_TEXTURE_LEVELS]; diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 291bbc40ad..6a5a643c89 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -34,6 +34,8 @@ #include "sp_state.h" #include "sp_texture.h" #include "sp_tile_cache.h" +#include "pipe/draw/draw_context.h" + void * @@ -73,6 +75,8 @@ softpipe_set_sampler_texture(struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); + draw_flush(softpipe->draw); + assert(unit < PIPE_MAX_SAMPLERS); softpipe->texture[unit] = softpipe_texture(texture); /* ptr, not struct */ diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c index 5e215c433a..325bdb86da 100644 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ b/src/mesa/pipe/softpipe/sp_tex_sample.c @@ -449,7 +449,6 @@ compute_lambda(struct tgsi_sampler *sampler, } lambda = LOG2(rho); - lambda += lodbias + sampler->state->lod_bias; lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); @@ -457,7 +456,6 @@ compute_lambda(struct tgsi_sampler *sampler, } - /** * Do several things here: * 1. Compute lambda from the texcoords, if needed @@ -477,7 +475,7 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { /* no mipmap selection needed */ *imgFilter = sampler->state->mag_img_filter; - *level0 = *level1 = sampler->texture->first_level; + *level0 = *level1 = (int) sampler->state->min_lod; } else { float lambda; @@ -492,7 +490,7 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, if (lambda < 0.0) { /* XXX threshold depends on the filter */ /* magnifying */ *imgFilter = sampler->state->mag_img_filter; - *level0 = *level1 = sampler->texture->first_level; + *level0 = *level1 = 0; } else { /* minifying */ @@ -503,19 +501,13 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, /* Nearest mipmap level */ const int lvl = (int) (lambda + 0.5); *level0 = - *level1 = CLAMP(lvl, - (int) sampler->texture->first_level, - (int) sampler->texture->last_level); + *level1 = CLAMP(lvl, 0, (int) sampler->texture->last_level); } else { /* Linear interpolation between mipmap levels */ const int lvl = (int) lambda; - *level0 = CLAMP(lvl, - (int) sampler->texture->first_level, - (int) sampler->texture->last_level); - *level1 = CLAMP(lvl + 1, - (int) sampler->texture->first_level, - (int) sampler->texture->last_level); + *level0 = CLAMP(lvl, 0, (int) sampler->texture->last_level); + *level1 = CLAMP(lvl + 1, 0, (int) sampler->texture->last_level); *levelBlend = FRAC(lambda); /* blending weight between levels */ } } diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c index fd2cc3dbbb..6de7a9b543 100644 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ b/src/mesa/pipe/softpipe/sp_texture.c @@ -61,7 +61,7 @@ softpipe_texture_layout(struct softpipe_texture * spt) spt->buffer_size = 0; - for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + for (level = 0; level <= pt->last_level; level++) { pt->width[level] = width; pt->height[level] = height; pt->depth[level] = depth; @@ -139,6 +139,8 @@ softpipe_get_tex_surface(struct pipe_context *pipe, struct softpipe_texture *spt = softpipe_texture(pt); struct pipe_surface *ps; + assert(level <= pt->last_level); + ps = pipe->winsys->surface_alloc(pipe->winsys); if (ps) { assert(ps->refcount); diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 7099ec33b9..f012b2f1a0 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -57,15 +57,10 @@ struct st_texture_object { struct gl_texture_object base; /* The "parent" object */ - /* The texture must include at least these levels once validated: + /* The texture must include at levels [0..lastLevel] once validated: */ - GLuint firstLevel; GLuint lastLevel; - /* Offset for firstLevel image: - */ - GLuint textureOffset; - /* On validation any active images held in main memory or in other * textures will be copied to this texture and the old storage freed. */ @@ -585,12 +580,12 @@ st_TexImage(GLcontext * ctx, _mesa_align_free(texImage->Data); } - /* If this is the only texture image in the texture, could call + /* If this is the only mipmap level in the texture, could call * bmBufferData with NULL data to free the old block and avoid * waiting on any outstanding fences. */ if (stObj->pt && - stObj->pt->first_level == level && + /*stObj->pt->first_level == level &&*/ stObj->pt->last_level == level && stObj->pt->target != PIPE_TEXTURE_CUBE && !st_texture_match_image(stObj->pt, &stImage->base, @@ -1363,13 +1358,8 @@ calculate_first_last_level(struct st_texture_object *stObj) firstLevel = lastLevel = tObj->BaseLevel; } else { - firstLevel = tObj->BaseLevel + (GLint) (tObj->MinLod + 0.5); - firstLevel = MAX2(firstLevel, tObj->BaseLevel); - lastLevel = tObj->BaseLevel + (GLint) (tObj->MaxLod + 0.5); - lastLevel = MAX2(lastLevel, tObj->BaseLevel); - lastLevel = MIN2(lastLevel, tObj->BaseLevel + baseImage->MaxLog2); - lastLevel = MIN2(lastLevel, tObj->MaxLevel); - lastLevel = MAX2(firstLevel, lastLevel); /* need at least one level */ + firstLevel = 0; + lastLevel = MIN2(tObj->MaxLevel - tObj->BaseLevel, baseImage->MaxLog2); } break; case GL_TEXTURE_RECTANGLE_NV: @@ -1380,8 +1370,6 @@ calculate_first_last_level(struct st_texture_object *stObj) return; } - /* save these values */ - stObj->firstLevel = firstLevel; stObj->lastLevel = lastLevel; } @@ -1389,15 +1377,16 @@ calculate_first_last_level(struct st_texture_object *stObj) static void copy_image_data_to_texture(struct st_context *st, struct st_texture_object *stObj, + GLuint dstLevel, struct st_texture_image *stImage) { if (stImage->pt) { /* Copy potentially with the blitter: */ st_texture_image_copy(st->pipe, - stObj->pt, /* dest texture */ - stImage->face, stImage->level, - stImage->pt /* src texture */ + stObj->pt, dstLevel, /* dest texture, level */ + stImage->pt, /* src texture */ + stImage->face ); st->pipe->texture_release(st->pipe, &stImage->pt); @@ -1438,7 +1427,7 @@ st_finalize_texture(GLcontext *ctx, const GLuint nr_faces = (stObj->base.Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1; int comp_byte = 0; int cpp; - GLuint face, i; + GLuint face; struct st_texture_image *firstImage; *needFlush = GL_FALSE; @@ -1450,7 +1439,7 @@ st_finalize_texture(GLcontext *ctx, /* What levels must the texture include at a minimum? */ calculate_first_last_level(stObj); - firstImage = st_texture_image(stObj->base.Image[0][stObj->firstLevel]); + firstImage = st_texture_image(stObj->base.Image[0][stObj->base.BaseLevel]); /* Fallback case: */ @@ -1469,7 +1458,6 @@ st_finalize_texture(GLcontext *ctx, */ if (firstImage->pt && firstImage->pt != stObj->pt && - firstImage->pt->first_level <= stObj->firstLevel && firstImage->pt->last_level >= stObj->lastLevel) { if (stObj->pt) @@ -1488,18 +1476,11 @@ st_finalize_texture(GLcontext *ctx, /* Check texture can hold all active levels. Check texture matches * target, imageFormat, etc. - * - * XXX: For some layouts (eg i945?), the test might have to be - * first_level == firstLevel, as the texture isn't valid except at the - * original start level. Hope to get around this by - * programming minLod, maxLod, baseLevel into the hardware and - * leaving the texture alone. */ if (stObj->pt && (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || stObj->pt->format != st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat) || - stObj->pt->first_level != stObj->firstLevel || stObj->pt->last_level != stObj->lastLevel || stObj->pt->width[0] != firstImage->base.Width || stObj->pt->height[0] != firstImage->base.Height || @@ -1516,7 +1497,7 @@ st_finalize_texture(GLcontext *ctx, stObj->pt = st_texture_create(ctx->st, gl_target_to_pipe(stObj->base.Target), st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat), - stObj->firstLevel, + 0, /* first level */ stObj->lastLevel, firstImage->base.Width, firstImage->base.Height, @@ -1527,14 +1508,16 @@ st_finalize_texture(GLcontext *ctx, /* Pull in any images not in the object's texture: */ for (face = 0; face < nr_faces; face++) { - for (i = stObj->firstLevel; i <= stObj->lastLevel; i++) { + GLuint level; + for (level = 0; level <= stObj->lastLevel; level++) { struct st_texture_image *stImage = - st_texture_image(stObj->base.Image[face][i]); + //st_texture_image(stObj->base.Image[face][level]); + st_texture_image(stObj->base.Image[face][stObj->base.BaseLevel + level]); /* Need to import images in main memory or held in other textures. */ if (stObj->pt != stImage->pt) { - copy_image_data_to_texture(ctx->st, stObj, stImage); + copy_image_data_to_texture(ctx->st, stObj, level, stImage); *needFlush = GL_TRUE; } } diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index c152c59905..fd7d8cefea 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -51,7 +51,6 @@ static void *blend_cso = NULL; static void *depthstencil_cso = NULL; static void *rasterizer_cso = NULL; -static void *sampler_cso = NULL; static struct st_fragment_program *stfp = NULL; static struct st_vertex_program *stvp = NULL; @@ -118,7 +117,6 @@ st_init_generate_mipmap(struct st_context *st) struct pipe_context *pipe = st->pipe; struct pipe_blend_state blend; struct pipe_rasterizer_state rasterizer; - struct pipe_sampler_state sampler; struct pipe_depth_stencil_alpha_state depthstencil; assert(!blend_cso); @@ -133,16 +131,6 @@ st_init_generate_mipmap(struct st_context *st) memset(&rasterizer, 0, sizeof(rasterizer)); rasterizer_cso = pipe->create_rasterizer_state(pipe, &rasterizer); - memset(&sampler, 0, sizeof(sampler)); - sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; - sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; - sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; - sampler.normalized_coords = 1; - sampler_cso = pipe->create_sampler_state(pipe, &sampler); - stfp = make_tex_fragment_program(st->ctx); stvp = st_make_passthrough_vertex_shader(st, GL_FALSE); } @@ -156,14 +144,12 @@ st_destroy_generate_mipmpap(struct st_context *st) pipe->delete_blend_state(pipe, blend_cso); pipe->delete_depth_stencil_alpha_state(pipe, depthstencil_cso); pipe->delete_rasterizer_state(pipe, rasterizer_cso); - pipe->delete_sampler_state(pipe, sampler_cso); /* XXX free stfp, stvp */ blend_cso = NULL; depthstencil_cso = NULL; rasterizer_cso = NULL; - sampler_cso = NULL; } @@ -248,8 +234,10 @@ st_render_mipmap(struct st_context *st, { struct pipe_context *pipe = st->pipe; struct pipe_framebuffer_state fb; + struct pipe_sampler_state sampler; + void *sampler_cso; const uint face = _mesa_tex_target_to_face(target), zslice = 0; - const uint first_level_save = pt->first_level; + /*const uint first_level_save = pt->first_level;*/ uint dstLevel; assert(target != GL_TEXTURE_3D); /* not done yet */ @@ -263,11 +251,21 @@ st_render_mipmap(struct st_context *st, memset(&fb, 0, sizeof(fb)); fb.num_cbufs = 1; + /* sampler state */ + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.normalized_coords = 1; + + /* bind CSOs */ pipe->bind_blend_state(pipe, blend_cso); pipe->bind_depth_stencil_alpha_state(pipe, depthstencil_cso); pipe->bind_rasterizer_state(pipe, rasterizer_cso); - pipe->bind_sampler_state(pipe, 0, sampler_cso); /* bind shaders */ pipe->bind_fs_state(pipe, stfp->fs->data); @@ -286,20 +284,29 @@ st_render_mipmap(struct st_context *st, fb.cbufs[0] = pipe->get_tex_surface(pipe, pt, face, dstLevel, zslice); pipe->set_framebuffer_state(pipe, &fb); + /* + * Setup sampler state + */ + sampler.min_lod = sampler.max_lod = srcLevel; + sampler_cso = pipe->create_sampler_state(pipe, &sampler); + pipe->bind_sampler_state(pipe, 0, sampler_cso); + simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]); /* * Setup src texture, override pt->first_level so we sample from * the right mipmap level. */ - pt->first_level = srcLevel; + /*pt->first_level = srcLevel;*/ pipe->set_sampler_texture(pipe, 0, pt); draw_quad(st->ctx); + + pipe->delete_sampler_state(pipe, sampler_cso); } /* restore first_level */ - pt->first_level = first_level_save; + /*pt->first_level = first_level_save;*/ /* restore pipe state */ if (st->state.rasterizer) diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 844a9f80d8..ff33da6f9e 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -62,6 +62,10 @@ target_to_target(GLenum target) /** * Allocate a new pipe_texture object + * width0, height0, depth0 are the dimensions of the level 0 image + * (the highest resolution). last_level indicates how many mipmap levels + * to allocate storage for. For non-mipmapped textures, this will be zero. + * XXX first_level obsolete */ struct pipe_texture * st_texture_create(struct st_context *st, @@ -84,9 +88,9 @@ st_texture_create(struct st_context *st, assert(format); + memset(&pt, 0, sizeof(pt)); pt.target = target; pt.format = format; - pt.first_level = first_level; pt.last_level = last_level; pt.width[0] = width0; pt.height[0] = height0; @@ -266,23 +270,36 @@ st_texture_image_data(struct pipe_context *pipe, */ void st_texture_image_copy(struct pipe_context *pipe, - struct pipe_texture *dst, - GLuint face, GLuint level, - struct pipe_texture *src) + struct pipe_texture *dst, GLuint dstLevel, + struct pipe_texture *src, + GLuint face) { - GLuint width = src->width[level]; - GLuint height = src->height[level]; - GLuint depth = src->depth[level]; + GLuint width = dst->width[dstLevel]; + GLuint height = dst->height[dstLevel]; + GLuint depth = dst->depth[dstLevel]; struct pipe_surface *src_surface; struct pipe_surface *dst_surface; GLuint i; + /* XXX this is a hack */ if (dst->compressed) height /= 4; for (i = 0; i < depth; i++) { - dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); - src_surface = pipe->get_tex_surface(pipe, src, face, level, i); + GLuint srcLevel; + + /* find src texture level of needed size */ + for (srcLevel = 0; srcLevel <= src->last_level; srcLevel++) { + if (src->width[srcLevel] == width && + src->height[srcLevel] == height) { + break; + } + } + assert(src->width[srcLevel] == width); + assert(src->height[srcLevel] == height); + + dst_surface = pipe->get_tex_surface(pipe, dst, face, dstLevel, i); + src_surface = pipe->get_tex_surface(pipe, src, face, srcLevel, i); pipe->surface_copy(pipe, dst_surface, diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 0b87a494c3..6c5f0930fa 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -98,9 +98,9 @@ st_texture_image_data(struct pipe_context *pipe, */ extern void st_texture_image_copy(struct pipe_context *pipe, - struct pipe_texture *dst, - GLuint face, GLuint level, - struct pipe_texture *src); + struct pipe_texture *dst, GLuint dstLevel, + struct pipe_texture *src, + GLuint face); #endif -- cgit v1.2.3 From 1d62a057bcb3ee4ef6ebedd93f62ed2e0d8061ba Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Feb 2008 14:55:51 -0700 Subject: gallium: remove unused first_level param from st_texture_create() --- src/mesa/state_tracker/st_cb_drawpixels.c | 6 +++--- src/mesa/state_tracker/st_cb_texture.c | 2 -- src/mesa/state_tracker/st_texture.c | 6 ++---- src/mesa/state_tracker/st_texture.h | 1 - 4 files changed, 5 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 07886e7982..475e23653e 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -471,7 +471,7 @@ make_texture(struct st_context *st, assert(pipeFormat); cpp = st_sizeof_format(pipeFormat); - pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, 0, width, height, + pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height, 1, 0); if (!pt) return NULL; @@ -1017,7 +1017,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, /** * Create a texture. */ - pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, 0, width, height, + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, width, height, 1, 0); if (!pt) return NULL; @@ -1241,7 +1241,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, psRead = rbRead->surface; format = psRead->format; - pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, 0, width, height, + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, width, height, 1, 0); if (!pt) return; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index f012b2f1a0..fab9eafc7f 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -363,7 +363,6 @@ guess_and_alloc_texture(struct st_context *st, stObj->pt = st_texture_create(st, gl_target_to_pipe(stObj->base.Target), st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat), - firstLevel, lastLevel, width, height, @@ -1497,7 +1496,6 @@ st_finalize_texture(GLcontext *ctx, stObj->pt = st_texture_create(ctx->st, gl_target_to_pipe(stObj->base.Target), st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat), - 0, /* first level */ stObj->lastLevel, firstImage->base.Width, firstImage->base.Height, diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index ff33da6f9e..2622d00953 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -65,13 +65,11 @@ target_to_target(GLenum target) * width0, height0, depth0 are the dimensions of the level 0 image * (the highest resolution). last_level indicates how many mipmap levels * to allocate storage for. For non-mipmapped textures, this will be zero. - * XXX first_level obsolete */ struct pipe_texture * st_texture_create(struct st_context *st, enum pipe_texture_target target, enum pipe_format format, - GLuint first_level, GLuint last_level, GLuint width0, GLuint height0, @@ -82,9 +80,9 @@ st_texture_create(struct st_context *st, assert(target <= PIPE_TEXTURE_CUBE); - DBG("%s target %s format %s level %d..%d\n", __FUNCTION__, + DBG("%s target %s format %s last_level %d\n", __FUNCTION__, _mesa_lookup_enum_by_nr(target), - _mesa_lookup_enum_by_nr(format), first_level, last_level); + _mesa_lookup_enum_by_nr(format), last_level); assert(format); diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 6c5f0930fa..72324cd9ab 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -39,7 +39,6 @@ extern struct pipe_texture * st_texture_create(struct st_context *st, enum pipe_texture_target target, enum pipe_format format, - GLuint first_level, GLuint last_level, GLuint width0, GLuint height0, -- cgit v1.2.3 From 6d53b8f42349c507e7184aa567a4b2a4fc7b037f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Feb 2008 16:06:51 -0700 Subject: gallium: move gen-mipmap global/static vars into st_context This fixes potential problems with multi-context programs. --- src/mesa/state_tracker/st_context.h | 9 ++++++++ src/mesa/state_tracker/st_gen_mipmap.c | 41 +++++++++++----------------------- 2 files changed, 22 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index a756055898..59d1590f05 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -166,6 +166,15 @@ struct st_context struct st_fragment_program *combined_prog; } bitmap; + /** For gen/render mipmap feature */ + struct { + void *blend_cso; + void *depthstencil_cso; + void *rasterizer_cso; + struct st_fragment_program *stfp; + struct st_vertex_program *stvp; + } gen_mipmap; + struct cso_cache *cache; }; diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index fd7d8cefea..b4a21fd7e2 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -48,15 +48,6 @@ -static void *blend_cso = NULL; -static void *depthstencil_cso = NULL; -static void *rasterizer_cso = NULL; - -static struct st_fragment_program *stfp = NULL; -static struct st_vertex_program *stvp = NULL; - - - static struct st_fragment_program * make_tex_fragment_program(GLcontext *ctx) { @@ -119,20 +110,18 @@ st_init_generate_mipmap(struct st_context *st) struct pipe_rasterizer_state rasterizer; struct pipe_depth_stencil_alpha_state depthstencil; - assert(!blend_cso); - memset(&blend, 0, sizeof(blend)); blend.colormask = PIPE_MASK_RGBA; - blend_cso = pipe->create_blend_state(pipe, &blend); + st->gen_mipmap.blend_cso = pipe->create_blend_state(pipe, &blend); memset(&depthstencil, 0, sizeof(depthstencil)); - depthstencil_cso = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil); + st->gen_mipmap.depthstencil_cso = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil); memset(&rasterizer, 0, sizeof(rasterizer)); - rasterizer_cso = pipe->create_rasterizer_state(pipe, &rasterizer); + st->gen_mipmap.rasterizer_cso = pipe->create_rasterizer_state(pipe, &rasterizer); - stfp = make_tex_fragment_program(st->ctx); - stvp = st_make_passthrough_vertex_shader(st, GL_FALSE); + st->gen_mipmap.stfp = make_tex_fragment_program(st->ctx); + st->gen_mipmap.stvp = st_make_passthrough_vertex_shader(st, GL_FALSE); } @@ -141,15 +130,11 @@ st_destroy_generate_mipmpap(struct st_context *st) { struct pipe_context *pipe = st->pipe; - pipe->delete_blend_state(pipe, blend_cso); - pipe->delete_depth_stencil_alpha_state(pipe, depthstencil_cso); - pipe->delete_rasterizer_state(pipe, rasterizer_cso); + pipe->delete_blend_state(pipe, st->gen_mipmap.blend_cso); + pipe->delete_depth_stencil_alpha_state(pipe, st->gen_mipmap.depthstencil_cso); + pipe->delete_rasterizer_state(pipe, st->gen_mipmap.rasterizer_cso); /* XXX free stfp, stvp */ - - blend_cso = NULL; - depthstencil_cso = NULL; - rasterizer_cso = NULL; } @@ -263,13 +248,13 @@ st_render_mipmap(struct st_context *st, /* bind CSOs */ - pipe->bind_blend_state(pipe, blend_cso); - pipe->bind_depth_stencil_alpha_state(pipe, depthstencil_cso); - pipe->bind_rasterizer_state(pipe, rasterizer_cso); + pipe->bind_blend_state(pipe, st->gen_mipmap.blend_cso); + pipe->bind_depth_stencil_alpha_state(pipe, st->gen_mipmap.depthstencil_cso); + pipe->bind_rasterizer_state(pipe, st->gen_mipmap.rasterizer_cso); /* bind shaders */ - pipe->bind_fs_state(pipe, stfp->fs->data); - pipe->bind_vs_state(pipe, stvp->cso->data); + pipe->bind_fs_state(pipe, st->gen_mipmap.stfp->fs->data); + pipe->bind_vs_state(pipe, st->gen_mipmap.stvp->cso->data); /* * XXX for small mipmap levels, it may be faster to use the software -- cgit v1.2.3 From 9677336845511be4852520d2e50f91f1df362f58 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Feb 2008 16:10:11 -0700 Subject: gallium: rename st_fragment_program's fs field to cso to match st_vertex_program --- src/mesa/state_tracker/st_atom_shader.c | 8 ++++---- src/mesa/state_tracker/st_cb_clear.c | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_cb_program.c | 4 ++-- src/mesa/state_tracker/st_gen_mipmap.c | 2 +- src/mesa/state_tracker/st_program.c | 2 +- src/mesa/state_tracker/st_program.h | 2 +- 7 files changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 9196918509..2c6ec8421b 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -158,7 +158,7 @@ find_translated_vp(struct st_context *st, /* * Translate fragment program if needed. */ - if (!stfp->fs) { + if (!stfp->cso) { GLuint inAttr, numIn = 0; for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) { @@ -179,7 +179,7 @@ find_translated_vp(struct st_context *st, stfp->input_to_slot, stfp->tokens, ST_MAX_SHADER_TOKENS); - assert(stfp->fs); + assert(stfp->cso); } @@ -227,7 +227,7 @@ find_translated_vp(struct st_context *st, if (fpInAttrib >= 0) { GLuint fpInSlot = stfp->input_to_slot[fpInAttrib]; if (fpInSlot != ~0) { - GLuint vpOutSlot = stfp->fs->state.input_map[fpInSlot]; + GLuint vpOutSlot = stfp->cso->state.input_map[fpInSlot]; xvp->output_to_slot[outAttr] = vpOutSlot; numVpOuts++; } @@ -300,7 +300,7 @@ update_linkage( struct st_context *st ) st->pipe->bind_vs_state(st->pipe, st->state.vs->cso->data); st->fp = stfp; - st->state.fs = stfp->fs; + st->state.fs = stfp->cso; st->pipe->bind_fs_state(st->pipe, st->state.fs->data); st->vertex_result_to_slot = xvp->output_to_slot; diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index ab98b54bab..410062e1e8 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -341,7 +341,7 @@ clear_with_quad(GLcontext *ctx, if (!stfp) { stfp = make_frag_shader(st); } - pipe->bind_fs_state(pipe, stfp->fs->data); + pipe->bind_fs_state(pipe, stfp->cso->data); } /* vertex shader state: color/position pass-through */ diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 475e23653e..3245a7488b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -665,7 +665,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, } /* fragment shader state: TEX lookup program */ - pipe->bind_fs_state(pipe, stfp->fs->data); + pipe->bind_fs_state(pipe, stfp->cso->data); /* vertex shader state: position + texcoord pass-through */ pipe->bind_vs_state(pipe, stvp->cso->data); diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index f1f33fb0dd..af3ee65504 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -168,10 +168,10 @@ static void st_program_string_notify( GLcontext *ctx, stfp->serialNo++; - if (stfp->fs) { + if (stfp->cso) { /* free the TGSI code */ // cso_delete(stfp->vs); - stfp->fs = NULL; + stfp->cso = NULL; } stfp->param_state = stfp->Base.Base.Parameters->StateFlags; diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index b4a21fd7e2..459941cca8 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -253,7 +253,7 @@ st_render_mipmap(struct st_context *st, pipe->bind_rasterizer_state(pipe, st->gen_mipmap.rasterizer_cso); /* bind shaders */ - pipe->bind_fs_state(pipe, st->gen_mipmap.stfp->fs->data); + pipe->bind_fs_state(pipe, st->gen_mipmap.stfp->cso->data); pipe->bind_vs_state(pipe, st->gen_mipmap.stvp->cso->data); /* diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 84a9094001..c8297baded 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -420,7 +420,7 @@ st_translate_fragment_program(struct st_context *st, fs.tokens = tokensOut; cso = st_cached_fs_state(st, &fs); - stfp->fs = cso; + stfp->cso = cso; if (0) _mesa_print_program(&stfp->Base.Base); diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index de02c3185f..ea1dde4a7a 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -62,7 +62,7 @@ struct st_fragment_program struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; /** Pointer to the corresponding cached shader */ - const struct cso_fragment_shader *fs; + const struct cso_fragment_shader *cso; GLuint param_state; -- cgit v1.2.3 From a34b43b3f4817268ef4b3f186203b5fbafc7214e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Feb 2008 16:37:49 -0700 Subject: gallium: added null ptr check --- src/mesa/state_tracker/st_cb_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index fab9eafc7f..c6fe928d47 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1514,7 +1514,7 @@ st_finalize_texture(GLcontext *ctx, /* Need to import images in main memory or held in other textures. */ - if (stObj->pt != stImage->pt) { + if (stImage && stObj->pt != stImage->pt) { copy_image_data_to_texture(ctx->st, stObj, level, stImage); *needFlush = GL_TRUE; } -- cgit v1.2.3 From 508b73289de5c4e16b31ae7267140efbb0f319a5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Feb 2008 16:38:56 -0700 Subject: gallium: remove dead code --- src/mesa/state_tracker/st_cb_texture.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index c6fe928d47..0ea367549b 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1509,7 +1509,6 @@ st_finalize_texture(GLcontext *ctx, GLuint level; for (level = 0; level <= stObj->lastLevel; level++) { struct st_texture_image *stImage = - //st_texture_image(stObj->base.Image[face][level]); st_texture_image(stObj->base.Image[face][stObj->base.BaseLevel + level]); /* Need to import images in main memory or held in other textures. -- cgit v1.2.3 From 48e223a0db85e83e180a49f3c3c7ea4101a86f40 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Feb 2008 17:07:27 -0700 Subject: gallium: clamp min_lod so it's never negative --- src/mesa/state_tracker/st_atom_sampler.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 6241e70b55..92263cb688 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -147,7 +147,7 @@ update_samplers(struct st_context *st) sampler.lod_bias = st->ctx->Texture.Unit[su].LodBias; #if 1 - sampler.min_lod = texobj->MinLod; + sampler.min_lod = (texobj->MinLod) < 0.0 ? 0.0 : texobj->MinLod; sampler.max_lod = texobj->MaxLod; #else /* min/max lod should really be as follows (untested). -- cgit v1.2.3 From 9af687c53037afe57e765f705906e12a86865c03 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Feb 2008 17:44:57 -0700 Subject: gallium: fix/finish glReadPixels(GL_DEPTH_COMPONENT). This fixes demos/reflect ('d' key) and tests/zreaddraw.c --- src/mesa/state_tracker/st_cb_readpixels.c | 99 ++++++++++++++++++++++++------- 1 file changed, 78 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index a1bbb3a831..868c5f3c5f 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -180,20 +180,13 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (!strb) return; + if (format == GL_RGBA && type == GL_FLOAT) { /* write tile(row) directly into user's buffer */ df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width, height, format, type, 0, 0); dfStride = width * 4; } -#if 0 - else if (format == GL_DEPTH_COMPONENT && type == GL_FLOAT) { - /* write tile(row) directly into user's buffer */ - df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width, - height, format, type, 0, 0); - dfStride = width; - } -#endif else { /* write tile(row) into temp row buffer */ df = (GLfloat *) temp; @@ -209,22 +202,86 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, yStep = 1; } - /* Do a row at a time to flip image data vertically */ - for (i = 0; i < height; i++) { - pipe_get_tile_rgba(pipe, strb->surface, x, y, width, 1, df); - y += yStep; - df += dfStride; - if (!dfStride) { - /* convert GLfloat to user's format/type */ - GLvoid *dst = _mesa_image_address2d(&clippedPacking, dest, width, - height, format, type, i, 0); + /* + * Copy pixels from pipe_surface to user memory + */ + { + /* dest of first pixel in client memory */ + GLubyte *dst = _mesa_image_address2d(&clippedPacking, dest, width, + height, format, type, 0, 0); + /* dest row stride */ + const GLint dstStride = _mesa_image_row_stride(&clippedPacking, width, + format, type); + + if (strb->surface->format == PIPE_FORMAT_S8Z24_UNORM) { if (format == GL_DEPTH_COMPONENT) { - _mesa_pack_depth_span(ctx, width, dst, type, - (GLfloat *) temp, &clippedPacking); + for (i = 0; i < height; i++) { + GLuint ztemp[MAX_WIDTH], j; + GLfloat zfloat[MAX_WIDTH]; + const double scale = 1.0 / ((1 << 24) - 1); + pipe_get_tile_raw(pipe, strb->surface, x, y, + width, 1, ztemp, 0); + y += yStep; + for (j = 0; j < width; j++) { + zfloat[j] = (float) (scale * (ztemp[j] & 0xffffff)); + } + _mesa_pack_depth_span(ctx, width, dst, type, + zfloat, &clippedPacking); + dst += dstStride; + } } else { - _mesa_pack_rgba_span_float(ctx, width, temp, format, type, dst, - &clippedPacking, transferOps); + /* untested, but simple: */ + assert(format == GL_DEPTH_STENCIL_EXT); + for (i = 0; i < height; i++) { + pipe_get_tile_raw(pipe, strb->surface, x, y, width, 1, dst, 0); + y += yStep; + dst += dstStride; + } + } + } + else if (strb->surface->format == PIPE_FORMAT_Z16_UNORM) { + for (i = 0; i < height; i++) { + GLshort ztemp[MAX_WIDTH], j; + GLfloat zfloat[MAX_WIDTH]; + const double scale = 1.0 / 0xffff; + pipe_get_tile_raw(pipe, strb->surface, x, y, width, 1, ztemp, 0); + y += yStep; + for (j = 0; j < width; j++) { + zfloat[j] = (float) (scale * ztemp[j]); + } + _mesa_pack_depth_span(ctx, width, dst, type, + zfloat, &clippedPacking); + dst += dstStride; + } + } + else if (strb->surface->format == PIPE_FORMAT_Z32_UNORM) { + for (i = 0; i < height; i++) { + GLuint ztemp[MAX_WIDTH], j; + GLfloat zfloat[MAX_WIDTH]; + const double scale = 1.0 / 0xffffffff; + pipe_get_tile_raw(pipe, strb->surface, x, y, width, 1, ztemp, 0); + y += yStep; + for (j = 0; j < width; j++) { + zfloat[j] = (float) (scale * ztemp[j]); + } + _mesa_pack_depth_span(ctx, width, dst, type, + zfloat, &clippedPacking); + dst += dstStride; + } + } + else { + /* RGBA format */ + /* Do a row at a time to flip image data vertically */ + for (i = 0; i < height; i++) { + pipe_get_tile_rgba(pipe, strb->surface, x, y, width, 1, df); + y += yStep; + df += dfStride; + if (!dfStride) { + _mesa_pack_rgba_span_float(ctx, width, temp, format, type, dst, + &clippedPacking, transferOps); + dst += dstStride; + } } } } -- cgit v1.2.3 From e20e89e48287808068086ec148920dd89495d813 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 12 Feb 2008 19:14:58 -0700 Subject: gallium: initial code for wide/AA line drawing --- src/mesa/pipe/draw/draw_wide_prims.c | 75 +++++++++++++++++++++++++++++++++--- 1 file changed, 69 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 9759e7e2e8..4c7e279b20 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -76,7 +76,6 @@ static void passthrough_tri( struct draw_stage *stage, /** * Draw a wide line by drawing a quad (two triangles). - * XXX still need line stipple. * XXX need to disable polygon stipple. */ static void wide_line( struct draw_stage *stage, @@ -103,12 +102,9 @@ static void wide_line( struct draw_stage *stage, /* * Draw wide line as a quad (two tris) by "stretching" the line along * X or Y. - * XXX For AA lines, the quad corners have to be computed in a - * more sophisticated way. + * We need to tweak coords in several ways to be conformant here. */ - /* need to tweak coords in several ways to be conformant here */ - if (dx > dy) { /* x-major line */ pos0[1] = pos0[1] - half_width - 0.25f; @@ -165,6 +161,70 @@ static void wide_line( struct draw_stage *stage, } +/** + * Draw a wide line by drawing a quad, using geometry which will + * fullfill GL's antialiased line requirements. + */ +static void wide_line_aa(struct draw_stage *stage, + struct prim_header *header) +{ + const struct wide_stage *wide = wide_stage(stage); + const float half_width = wide->half_line_width; + struct prim_header tri; + struct vertex_header *v[4]; + float *pos; + float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; + float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; + const float len = sqrt(dx * dx + dy * dy); + uint i; + + dx = dx * half_width / len; + dy = dy * half_width / len; + + /* allocate/dup new verts */ + for (i = 0; i < 4; i++) { + v[i] = dup_vert(stage, header->v[i/2], i); + } + + /* + * Quad for line from v0 to v1: + * + * 1 3 + * +-------------------------+ + * | | + * *v0 v1* + * | | + * +-------------------------+ + * 0 2 + */ + + pos = v[0]->data[0]; + pos[0] += dy; + pos[1] -= dx; + + pos = v[1]->data[0]; + pos[0] -= dy; + pos[1] += dx; + + pos = v[2]->data[0]; + pos[0] += dy; + pos[1] -= dx; + + pos = v[3]->data[0]; + pos[0] -= dy; + pos[1] += dx; + + tri.det = header->det; /* only the sign matters */ + + tri.v[0] = v[2]; tri.v[1] = v[1]; tri.v[2] = v[0]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[3]; tri.v[1] = v[1]; tri.v[2] = v[2]; + stage->next->tri( stage->next, &tri ); + +} + + /** * Set the vertex texcoords for sprite mode. * Coords may be left untouched or set to a right-side-up or upside-down @@ -319,7 +379,10 @@ static void wide_first_line( struct draw_stage *stage, wide->half_line_width = 0.5f * draw->rasterizer->line_width; if (draw->rasterizer->line_width != 1.0) { - wide->stage.line = wide_line; + if (draw->rasterizer->line_smooth) + wide->stage.line = wide_line_aa; + else + wide->stage.line = wide_line; } else { wide->stage.line = passthrough_line; -- cgit v1.2.3 From cad7bc74d69ee053452aa4bd20740dc79ad6eab6 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 12 Feb 2008 05:35:51 -0500 Subject: llvm ir code to fetch the elements --- src/mesa/pipe/llvm/storagesoa.cpp | 59 +++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/storagesoa.h | 17 +++++++++++ 2 files changed, 76 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index b2aca3557a..ff94307c85 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -45,6 +45,11 @@ StorageSoa::StorageSoa(llvm::BasicBlock *block, llvm::Value *input, llvm::Value *output, llvm::Value *consts) + : m_block(block), + m_input(input), + m_output(output), + m_consts(consts), + m_idx(0) { } @@ -62,6 +67,11 @@ std::vector StorageSoa::inputElement(int idx, int swizzle, { std::vector res(4); + res[0] = element(m_input, idx, 0); + res[1] = element(m_input, idx, 0); + res[2] = element(m_input, idx, 0); + res[3] = element(m_input, idx, 0); + return res; } @@ -78,6 +88,11 @@ std::vector StorageSoa::outputElement(int idx, int swizzle, { std::vector res(4); + res[0] = element(m_output, idx, 0); + res[1] = element(m_output, idx, 0); + res[2] = element(m_output, idx, 0); + res[3] = element(m_output, idx, 0); + return res; } @@ -115,3 +130,47 @@ void StorageSoa::storeAddress(int idx, const std::vector &val, int mask) { } + +llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, int index, + int channel) const +{ + std::vector indices; + indices.push_back(constantInt(index)); + indices.push_back(constantInt(0));//first element in the struct + indices.push_back(constantInt(channel)); + indices.push_back(constantInt(0));//f channel + indices.push_back(constantInt(0));//first ptr in the f channel + + GetElementPtrInst *getElem = new GetElementPtrInst(ptr, + indices.begin(), + indices.end(), + name("ptr"), + m_block); + return getElem; +} + +llvm::Value * StorageSoa::element(llvm::Value *ptr, int index, + int channel) const +{ + llvm::Value *res = elementPointer(ptr, index, channel); + LoadInst *load = new LoadInst(res, name("element"), false, m_block); + //load->setAlignment(8); + return load; +} + +const char * StorageSoa::name(const char *prefix) const +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::ConstantInt * StorageSoa::constantInt(int idx) const +{ + if (m_constInts.find(idx) != m_constInts.end()) { + return m_constInts[idx]; + } + ConstantInt *constInt = ConstantInt::get(APInt(32, idx)); + m_constInts[idx] = constInt; + return constInt; +} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index 551b0b9734..9d5609f539 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -29,6 +29,7 @@ #define STORAGESOA_H #include +#include namespace llvm { class BasicBlock; @@ -65,7 +66,23 @@ public: int mask); void storeAddress(int idx, const std::vector &val, int mask); +private: + llvm::Value *elementPointer(llvm::Value *ptr, int index, + int channel) const; + llvm::Value *element(llvm::Value *ptr, int index, + int channel) const; + const char *name(const char *prefix) const; + llvm::ConstantInt *constantInt(int) const; +private: + llvm::BasicBlock *m_block; + llvm::Value *m_input; + llvm::Value *m_output; + llvm::Value *m_consts; + + mutable std::map m_constInts; + mutable char m_name[32]; + mutable int m_idx; }; #endif -- cgit v1.2.3 From 3c3c1ff5cd19af23033e080d8f0b9b5ae8363f2e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 12 Feb 2008 23:08:42 -0500 Subject: stop generate llvm entry points entrypoints are useless because we use the same paths as all other code. also simplify llvm swizzling code --- src/mesa/pipe/llvm/gallivm.cpp | 4 +-- src/mesa/pipe/llvm/gallivm_p.h | 54 +++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/storage.cpp | 10 +++++--- src/mesa/pipe/llvm/storagesoa.cpp | 14 ++++++++++ src/mesa/pipe/llvm/tgsitollvm.cpp | 12 ++++----- 5 files changed, 82 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index b99dc6db5b..ab13be0915 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -289,9 +289,9 @@ void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, tgsi_dump(tokens, 0); - llvm::Module *irmod = tgsi_to_llvmir(ir, tokens); + llvm::Module *mod = tgsi_to_llvmir(ir, tokens); - llvm::Module *mod = tgsi_to_llvm(ir, tokens); + //llvm::Module *mod = tgsi_to_llvm(ir, tokens); ir->module = mod; gallivm_ir_dump(ir, 0); } diff --git a/src/mesa/pipe/llvm/gallivm_p.h b/src/mesa/pipe/llvm/gallivm_p.h index 2c6e5e8f5f..cfe7b1901b 100644 --- a/src/mesa/pipe/llvm/gallivm_p.h +++ b/src/mesa/pipe/llvm/gallivm_p.h @@ -3,6 +3,10 @@ #ifdef MESA_LLVM +#include "gallivm.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/p_compiler.h" + namespace llvm { class Module; } @@ -47,6 +51,56 @@ struct gallivm_prog { int num_interp; }; +static INLINE void gallivm_swizzle_components(int swizzle, + int *xc, int *yc, + int *zc, int *wc) +{ + int x = swizzle / 1000; swizzle -= x * 1000; + int y = swizzle / 100; swizzle -= y * 100; + int z = swizzle / 10; swizzle -= z * 10; + int w = swizzle; + + if (xc) *xc = x; + if (yc) *yc = y; + if (zc) *zc = z; + if (wc) *wc = w; +} + +static INLINE boolean gallivm_is_swizzle(int swizzle) +{ + const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + + TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; + return swizzle != NO_SWIZZLE; +} + +static INLINE int gallivm_x_swizzle(int swizzle) +{ + int x; + gallivm_swizzle_components(swizzle, &x, 0, 0, 0); + return x; +} + +static INLINE int gallivm_y_swizzle(int swizzle) +{ + int y; + gallivm_swizzle_components(swizzle, 0, &y, 0, 0); + return y; +} + +static INLINE int gallivm_z_swizzle(int swizzle) +{ + int z; + gallivm_swizzle_components(swizzle, 0, 0, &z, 0); + return z; +} + +static INLINE int gallivm_w_swizzle(int swizzle) +{ + int w; + gallivm_swizzle_components(swizzle, 0, 0, 0, &w); + return w; +} + #endif /* MESA_LLVM */ #if defined __cplusplus diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp index 08b9d29a24..c4326de8c5 100644 --- a/src/mesa/pipe/llvm/storage.cpp +++ b/src/mesa/pipe/llvm/storage.cpp @@ -33,6 +33,8 @@ #include "storage.h" +#include "gallivm_p.h" + #include "pipe/p_shader_tokens.h" #include #include @@ -82,10 +84,10 @@ llvm::Constant *Storage::shuffleMask(int vec) if (origVec == 0) { const_vec = Constant::getNullValue(m_intVecType); } else { - int x = vec / 1000; vec -= x * 1000; - int y = vec / 100; vec -= y * 100; - int z = vec / 10; vec -= z * 10; - int w = vec; + int x = gallivm_x_swizzle(vec); + int y = gallivm_y_swizzle(vec); + int z = gallivm_z_swizzle(vec); + int w = gallivm_w_swizzle(vec); std::vector elems; elems.push_back(constantInt(x)); elems.push_back(constantInt(y)); diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index ff94307c85..8f82989cc5 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -119,6 +119,20 @@ llvm::Value * StorageSoa::extractIndex(llvm::Value *vec) void StorageSoa::storeOutput(int dstIdx, const std::vector &val, int mask) { + if (mask != TGSI_WRITEMASK_XYZW) { + fprintf(stderr, "requires swizzle!!\n"); + assert(0); + } else { + llvm::Value *xChannel = elementPointer(m_output, dstIdx, 0); + llvm::Value *yChannel = elementPointer(m_output, dstIdx, 1); + llvm::Value *zChannel = elementPointer(m_output, dstIdx, 2); + llvm::Value *wChannel = elementPointer(m_output, dstIdx, 3); + + StoreInst *st = new StoreInst(val[0], xChannel, false, m_block); + st = new StoreInst(val[1], yChannel, false, m_block); + st = new StoreInst(val[2], zChannel, false, m_block); + st = new StoreInst(val[3], wChannel, false, m_block); + } } void StorageSoa::storeTemp(int idx, const std::vector &val, diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index bc4df61071..574e340f66 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -178,9 +178,8 @@ swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, Storage *storage) { int swizzle = swizzleInt(src); - const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + - TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; - if (swizzle != NO_SWIZZLE) { + + if (gallivm_is_swizzle(swizzle)) { /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ val = storage->shuffleVector(val, swizzle); } @@ -1107,12 +1106,11 @@ tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, const struct tgsi_token *tokens) { - llvm::Module *mod = createBaseShader(); + llvm::Module *mod = new Module("shader"); struct tgsi_parse_context parse; struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; unsigned instno = 0; - Function* shader = mod->getFunction("execute_shader"); std::ostringstream stream; if (ir->type == GALLIVM_VS) { stream << "vs_shader"; @@ -1121,7 +1119,9 @@ llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, } stream << ir->id; std::string func_name = stream.str(); - shader->setName(func_name.c_str()); + Function *shader = llvm::cast(mod->getOrInsertFunction( + func_name.c_str(), + (const llvm::FunctionType*)0)); Function::arg_iterator args = shader->arg_begin(); Value *input = args++; -- cgit v1.2.3 From 8b054cd6eb0b64264d9cb8b61ce9df5135664368 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 00:21:03 -0500 Subject: disable llvm code in fs --- src/mesa/pipe/softpipe/sp_quad_fs.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index b5d7dfca1c..3316858413 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -199,6 +199,7 @@ shade_quad( } } +#if 0 #ifdef MESA_LLVM #define DLLVM 0 static void @@ -297,6 +298,7 @@ shade_quad_llvm(struct quad_stage *qs, } } #endif /*MESA_LLVM*/ +#endif /** * Per-primitive (or per-begin?) setup -- cgit v1.2.3 From 9b6532f01b2e8a3ccc44d67d20a8f94f5de570e3 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 00:21:24 -0500 Subject: add functiontype for the llvm native vs entry point --- src/mesa/pipe/llvm/tgsitollvm.cpp | 38 +++++++++++++++++++++++++++++++++++--- 1 file changed, 35 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index 574e340f66..d4e9a21a13 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -46,6 +46,38 @@ using namespace llvm; #include "llvm_base_shader.cpp" +static inline FunctionType *vertexShaderFunctionType() +{ + //Function takes three arguments, + // the calling code has to make sure the types it will + // pass are castable to the following: + // [4 x <4 x float>] inputs, + // [4 x <4 x float>] output, + // [4 x [4 x float]] consts + std::vector funcArgs; + { + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorArray = ArrayType::get(vectorType, 4); + PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0); + + funcArgs.push_back(vectorArrayPtr);//inputs + funcArgs.push_back(vectorArrayPtr);//output + } + { + ArrayType *floatArray = ArrayType::get(Type::FloatTy, 4); + ArrayType *constsArray = ArrayType::get(floatArray, 4); + PointerType *constsArrayPtr = PointerType::get(constsArray, 0); + + funcArgs.push_back(constsArrayPtr);//consts + } + FunctionType *functionType = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/funcArgs, + /*isVarArg=*/false); + + return functionType; +} + static inline void add_interpolator(struct gallivm_ir *ir, struct gallivm_interpolate *interp) @@ -1121,13 +1153,13 @@ llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, std::string func_name = stream.str(); Function *shader = llvm::cast(mod->getOrInsertFunction( func_name.c_str(), - (const llvm::FunctionType*)0)); + vertexShaderFunctionType())); Function::arg_iterator args = shader->arg_begin(); Value *input = args++; - input->setName("input"); + input->setName("inputs"); Value *output = args++; - output->setName("output"); + output->setName("outputs"); Value *consts = args++; consts->setName("consts"); -- cgit v1.2.3 From 53c2963b84f7a8538d2a2c67e3fbb8a22f644c57 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 00:36:31 -0500 Subject: fix fetching input elements and generate a ret on end --- src/mesa/pipe/llvm/instructionssoa.cpp | 2 ++ src/mesa/pipe/llvm/instructionssoa.h | 4 ++++ src/mesa/pipe/llvm/storagesoa.cpp | 3 --- 3 files changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructionssoa.cpp b/src/mesa/pipe/llvm/instructionssoa.cpp index 9ac4d8fbc7..0e501ab08d 100644 --- a/src/mesa/pipe/llvm/instructionssoa.cpp +++ b/src/mesa/pipe/llvm/instructionssoa.cpp @@ -2,6 +2,7 @@ InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, StorageSoa *storage) + : m_builder(block) { } @@ -23,4 +24,5 @@ std::vector InstructionsSoa::mul(const std::vector i void InstructionsSoa::end() { + m_builder.CreateRetVoid(); } diff --git a/src/mesa/pipe/llvm/instructionssoa.h b/src/mesa/pipe/llvm/instructionssoa.h index 0b6b41cf05..233d363b90 100644 --- a/src/mesa/pipe/llvm/instructionssoa.h +++ b/src/mesa/pipe/llvm/instructionssoa.h @@ -28,6 +28,8 @@ #ifndef INSTRUCTIONSSOA_H #define INSTRUCTIONSSOA_H +#include + #include namespace llvm { @@ -49,6 +51,8 @@ public: std::vector mul(const std::vector in1, const std::vector in2); void end(); +private: + llvm::LLVMFoldingBuilder m_builder; }; diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index 8f82989cc5..786511204a 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -150,10 +150,7 @@ llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, int index, { std::vector indices; indices.push_back(constantInt(index)); - indices.push_back(constantInt(0));//first element in the struct indices.push_back(constantInt(channel)); - indices.push_back(constantInt(0));//f channel - indices.push_back(constantInt(0));//first ptr in the f channel GetElementPtrInst *getElem = new GetElementPtrInst(ptr, indices.begin(), -- cgit v1.2.3 From 4c8456264ca9c826165c41810b20dfbfbfb322e3 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 00:48:08 -0500 Subject: get mov working --- src/mesa/pipe/llvm/gallivm.cpp | 3 +++ src/mesa/pipe/llvm/gallivm_cpu.cpp | 4 ++-- src/mesa/pipe/llvm/storagesoa.cpp | 12 ++++++------ src/mesa/pipe/llvm/tgsitollvm.cpp | 2 +- 4 files changed, 12 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp index ab13be0915..da0105c2c9 100644 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ b/src/mesa/pipe/llvm/gallivm.cpp @@ -318,6 +318,9 @@ struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) passes.run(*mod); prog->module = mod; + std::cout << "After optimizations:"<dump(); + return prog; } diff --git a/src/mesa/pipe/llvm/gallivm_cpu.cpp b/src/mesa/pipe/llvm/gallivm_cpu.cpp index 5f1268bf4f..011cba55c2 100644 --- a/src/mesa/pipe/llvm/gallivm_cpu.cpp +++ b/src/mesa/pipe/llvm/gallivm_cpu.cpp @@ -105,10 +105,10 @@ static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) switch (prog->type) { case GALLIVM_VS: - func = mod->getFunction("run_vertex_shader"); + func = mod->getFunction("vs_shader"); break; case GALLIVM_FS: - func = mod->getFunction("run_fragment_shader"); + func = mod->getFunction("fs_shader"); break; default: assert(!"Unknown shader type!"); diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index 786511204a..ef14e29bfe 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -68,9 +68,9 @@ std::vector StorageSoa::inputElement(int idx, int swizzle, std::vector res(4); res[0] = element(m_input, idx, 0); - res[1] = element(m_input, idx, 0); - res[2] = element(m_input, idx, 0); - res[3] = element(m_input, idx, 0); + res[1] = element(m_input, idx, 1); + res[2] = element(m_input, idx, 2); + res[3] = element(m_input, idx, 3); return res; } @@ -89,9 +89,9 @@ std::vector StorageSoa::outputElement(int idx, int swizzle, std::vector res(4); res[0] = element(m_output, idx, 0); - res[1] = element(m_output, idx, 0); - res[2] = element(m_output, idx, 0); - res[3] = element(m_output, idx, 0); + res[1] = element(m_output, idx, 1); + res[2] = element(m_output, idx, 2); + res[3] = element(m_output, idx, 3); return res; } diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index d4e9a21a13..070c9a67f3 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -1149,7 +1149,7 @@ llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, } else { stream << "fs_shader"; } - stream << ir->id; + //stream << ir->id; std::string func_name = stream.str(); Function *shader = llvm::cast(mod->getOrInsertFunction( func_name.c_str(), -- cgit v1.2.3 From 135d2329de7721b2083aa5f38f8d66beb20c1181 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 03:18:37 -0500 Subject: implement mul --- src/mesa/pipe/llvm/instructionssoa.cpp | 15 ++++++++++++++- src/mesa/pipe/llvm/instructionssoa.h | 7 +++++++ 2 files changed, 21 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructionssoa.cpp b/src/mesa/pipe/llvm/instructionssoa.cpp index 0e501ab08d..f1c174a26e 100644 --- a/src/mesa/pipe/llvm/instructionssoa.cpp +++ b/src/mesa/pipe/llvm/instructionssoa.cpp @@ -2,7 +2,8 @@ InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, StorageSoa *storage) - : m_builder(block) + : m_builder(block), + m_idx(0) { } @@ -19,6 +20,11 @@ std::vector InstructionsSoa::mul(const std::vector i { std::vector res(4); + res[0] = m_builder.CreateMul(in1[0], in2[0], name("mul")); + res[1] = m_builder.CreateMul(in1[1], in2[1], name("mul")); + res[2] = m_builder.CreateMul(in1[2], in2[2], name("mul")); + res[3] = m_builder.CreateMul(in1[3], in2[3], name("mul")); + return res; } @@ -26,3 +32,10 @@ void InstructionsSoa::end() { m_builder.CreateRetVoid(); } + +const char * InstructionsSoa::name(const char *prefix) const +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} diff --git a/src/mesa/pipe/llvm/instructionssoa.h b/src/mesa/pipe/llvm/instructionssoa.h index 233d363b90..0195501584 100644 --- a/src/mesa/pipe/llvm/instructionssoa.h +++ b/src/mesa/pipe/llvm/instructionssoa.h @@ -51,8 +51,15 @@ public: std::vector mul(const std::vector in1, const std::vector in2); void end(); + +private: + const char * name(const char *prefix) const; private: llvm::LLVMFoldingBuilder m_builder; + +private: + mutable int m_idx; + mutable char m_name[32]; }; -- cgit v1.2.3 From e179d5bdd199e3747773f5b07efcf9a635c41089 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 04:37:40 -0500 Subject: implement add --- src/mesa/pipe/llvm/instructionssoa.cpp | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructionssoa.cpp b/src/mesa/pipe/llvm/instructionssoa.cpp index f1c174a26e..0c2032e56f 100644 --- a/src/mesa/pipe/llvm/instructionssoa.cpp +++ b/src/mesa/pipe/llvm/instructionssoa.cpp @@ -12,6 +12,11 @@ std::vector InstructionsSoa::add(const std::vector i { std::vector res(4); + res[0] = m_builder.CreateAdd(in1[0], in2[0], name("addx")); + res[1] = m_builder.CreateAdd(in1[1], in2[1], name("addy")); + res[2] = m_builder.CreateAdd(in1[2], in2[2], name("addz")); + res[3] = m_builder.CreateAdd(in1[3], in2[3], name("addw")); + return res; } @@ -20,10 +25,10 @@ std::vector InstructionsSoa::mul(const std::vector i { std::vector res(4); - res[0] = m_builder.CreateMul(in1[0], in2[0], name("mul")); - res[1] = m_builder.CreateMul(in1[1], in2[1], name("mul")); - res[2] = m_builder.CreateMul(in1[2], in2[2], name("mul")); - res[3] = m_builder.CreateMul(in1[3], in2[3], name("mul")); + res[0] = m_builder.CreateMul(in1[0], in2[0], name("mulx")); + res[1] = m_builder.CreateMul(in1[1], in2[1], name("muly")); + res[2] = m_builder.CreateMul(in1[2], in2[2], name("mulz")); + res[3] = m_builder.CreateMul(in1[3], in2[3], name("mulw")); return res; } -- cgit v1.2.3 From 4bb1a14d901fcddb25efeeff49c4dea8ca872f73 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 04:38:10 -0500 Subject: handle temporaries in llvm code generated paths --- src/mesa/pipe/draw/draw_vertex_shader.c | 3 ++- src/mesa/pipe/llvm/gallivm.h | 3 ++- src/mesa/pipe/llvm/gallivm_cpu.cpp | 10 ++++------ src/mesa/pipe/llvm/storagesoa.cpp | 23 ++++++++++++++++++++++- src/mesa/pipe/llvm/storagesoa.h | 4 +++- src/mesa/pipe/llvm/tgsitollvm.cpp | 32 +++++++++++++++++--------------- 6 files changed, 50 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index 9c31df1e3e..c824c1407e 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -119,7 +119,8 @@ run_vertex_program(struct draw_context *draw, gallivm_cpu_vs_exec(prog, machine->Inputs, machine->Outputs, - machine->Consts); + machine->Consts, + machine->Temps); } else #elif defined(__i386__) || defined(__386__) if (draw->use_sse) { diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h index b104520cb7..92da4bca7f 100644 --- a/src/mesa/pipe/llvm/gallivm.h +++ b/src/mesa/pipe/llvm/gallivm.h @@ -82,7 +82,8 @@ struct gallivm_cpu_engine *gallivm_global_cpu_engine(); int gallivm_cpu_vs_exec(struct gallivm_prog *prog, struct tgsi_exec_vector *inputs, struct tgsi_exec_vector *dests, - float (*consts)[4]); + float (*consts)[4], + struct tgsi_exec_vector *temps); int gallivm_cpu_fs_exec(struct gallivm_prog *prog, float x, float y, float (*dests)[PIPE_MAX_SHADER_INPUTS][4], diff --git a/src/mesa/pipe/llvm/gallivm_cpu.cpp b/src/mesa/pipe/llvm/gallivm_cpu.cpp index 011cba55c2..dc4d92a72a 100644 --- a/src/mesa/pipe/llvm/gallivm_cpu.cpp +++ b/src/mesa/pipe/llvm/gallivm_cpu.cpp @@ -177,10 +177,7 @@ struct gallivm_cpu_engine * gallivm_global_cpu_engine() typedef void (*vertex_shader_runner)(void *ainputs, void *dests, float (*aconsts)[4], - int num_vertices, - int num_inputs, - int num_attribs, - int num_consts); + void *temps); /*! @@ -191,12 +188,13 @@ typedef void (*vertex_shader_runner)(void *ainputs, int gallivm_cpu_vs_exec(struct gallivm_prog *prog, struct tgsi_exec_vector *inputs, struct tgsi_exec_vector *dests, - float (*consts)[4]) + float (*consts)[4], + struct tgsi_exec_vector *temps) { vertex_shader_runner runner = reinterpret_cast(prog->function); assert(runner); /*FIXME*/ - runner(inputs, dests, consts, 4, 4, 4, prog->num_consts); + runner(inputs, dests, consts, temps); return 0; } diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index ef14e29bfe..e09e9e8fe7 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -44,11 +44,13 @@ using namespace llvm; StorageSoa::StorageSoa(llvm::BasicBlock *block, llvm::Value *input, llvm::Value *output, - llvm::Value *consts) + llvm::Value *consts, + llvm::Value *temps) : m_block(block), m_input(input), m_output(output), m_consts(consts), + m_temps(temps), m_idx(0) { } @@ -101,6 +103,11 @@ std::vector StorageSoa::tempElement(int idx, int swizzle, { std::vector res(4); + res[0] = element(m_temps, idx, 0); + res[1] = element(m_temps, idx, 1); + res[2] = element(m_temps, idx, 2); + res[3] = element(m_temps, idx, 3); + return res; } @@ -138,6 +145,20 @@ void StorageSoa::storeOutput(int dstIdx, const std::vector &val, void StorageSoa::storeTemp(int idx, const std::vector &val, int mask) { + if (mask != TGSI_WRITEMASK_XYZW) { + fprintf(stderr, "requires swizzle!!\n"); + assert(0); + } else { + llvm::Value *xChannel = elementPointer(m_temps, idx, 0); + llvm::Value *yChannel = elementPointer(m_temps, idx, 1); + llvm::Value *zChannel = elementPointer(m_temps, idx, 2); + llvm::Value *wChannel = elementPointer(m_temps, idx, 3); + + StoreInst *st = new StoreInst(val[0], xChannel, false, m_block); + st = new StoreInst(val[1], yChannel, false, m_block); + st = new StoreInst(val[2], zChannel, false, m_block); + st = new StoreInst(val[3], wChannel, false, m_block); + } } void StorageSoa::storeAddress(int idx, const std::vector &val, diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index 9d5609f539..84db7726a7 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -46,7 +46,8 @@ public: StorageSoa(llvm::BasicBlock *block, llvm::Value *input, llvm::Value *output, - llvm::Value *consts); + llvm::Value *consts, + llvm::Value *temps); void addImmediate(float *vec); @@ -79,6 +80,7 @@ private: llvm::Value *m_input; llvm::Value *m_output; llvm::Value *m_consts; + llvm::Value *m_temps; mutable std::map m_constInts; mutable char m_name[32]; diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index 070c9a67f3..3497eebcd9 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -53,23 +53,23 @@ static inline FunctionType *vertexShaderFunctionType() // pass are castable to the following: // [4 x <4 x float>] inputs, // [4 x <4 x float>] output, - // [4 x [4 x float]] consts + // [4 x [4 x float]] consts, + // [4 x <4 x float>] temps + std::vector funcArgs; - { - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - ArrayType *vectorArray = ArrayType::get(vectorType, 4); - PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0); + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorArray = ArrayType::get(vectorType, 4); + PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0); - funcArgs.push_back(vectorArrayPtr);//inputs - funcArgs.push_back(vectorArrayPtr);//output - } - { - ArrayType *floatArray = ArrayType::get(Type::FloatTy, 4); - ArrayType *constsArray = ArrayType::get(floatArray, 4); - PointerType *constsArrayPtr = PointerType::get(constsArray, 0); + ArrayType *floatArray = ArrayType::get(Type::FloatTy, 4); + ArrayType *constsArray = ArrayType::get(floatArray, 4); + PointerType *constsArrayPtr = PointerType::get(constsArray, 0); + + funcArgs.push_back(vectorArrayPtr);//inputs + funcArgs.push_back(vectorArrayPtr);//output + funcArgs.push_back(constsArrayPtr);//consts + funcArgs.push_back(vectorArrayPtr);//temps - funcArgs.push_back(constsArrayPtr);//consts - } FunctionType *functionType = FunctionType::get( /*Result=*/Type::VoidTy, /*Params=*/funcArgs, @@ -1162,6 +1162,8 @@ llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, output->setName("outputs"); Value *consts = args++; consts->setName("consts"); + Value *temps = args++; + temps->setName("temps"); BasicBlock *label_entry = new BasicBlock("entry", shader, 0); @@ -1170,7 +1172,7 @@ llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); - StorageSoa storage(label_entry, input, output, consts); + StorageSoa storage(label_entry, input, output, consts, temps); InstructionsSoa instr(mod, shader, label_entry, &storage); while(!tgsi_parse_end_of_tokens(&parse)) { -- cgit v1.2.3 From e922adbe1d6c1764968377658ea92ae6de0585db Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 13 Feb 2008 11:39:58 +0000 Subject: gallium: pipe->surface_copy can flip the contents vertically when necessary. Fixes gears being upside down on the box in demos/gearbox. --- src/mesa/pipe/cell/ppu/cell_surface.c | 5 +++-- src/mesa/pipe/i915simple/i915_blit.c | 1 + src/mesa/pipe/i915simple/i915_blit.h | 1 + src/mesa/pipe/i915simple/i915_surface.c | 6 ++++-- src/mesa/pipe/i965simple/brw_blit.c | 1 + src/mesa/pipe/i965simple/brw_blit.h | 1 + src/mesa/pipe/i965simple/brw_surface.c | 6 ++++-- src/mesa/pipe/p_context.h | 1 + src/mesa/pipe/p_util.h | 2 +- src/mesa/pipe/softpipe/sp_surface.c | 5 +++-- src/mesa/pipe/util/p_util.c | 4 ++-- src/mesa/pipe/xlib/xm_api.c | 1 + src/mesa/state_tracker/st_cb_drawpixels.c | 1 + src/mesa/state_tracker/st_cb_texture.c | 3 +++ src/mesa/state_tracker/st_texture.c | 1 + 15 files changed, 28 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c index 6b7b918128..fca93e4742 100644 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ b/src/mesa/pipe/cell/ppu/cell_surface.c @@ -60,6 +60,7 @@ cell_surface_data(struct pipe_context *pipe, static void cell_surface_copy(struct pipe_context *pipe, + unsigned do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, @@ -74,8 +75,8 @@ cell_surface_copy(struct pipe_context *pipe, dstx, dsty, width, height, pipe_surface_map(src), - src->pitch, - srcx, srcy); + do_flip ? -src->pitch : src->pitch, + srcx, do_flip ? 1 - srcy - height : srcy); pipe_surface_unmap(src); pipe_surface_unmap(dst); diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/mesa/pipe/i915simple/i915_blit.c index d49876f970..db4671ff55 100644 --- a/src/mesa/pipe/i915simple/i915_blit.c +++ b/src/mesa/pipe/i915simple/i915_blit.c @@ -85,6 +85,7 @@ i915_fill_blit(struct i915_context *i915, void i915_copy_blit( struct i915_context *i915, + unsigned do_flip, unsigned cpp, short src_pitch, struct pipe_buffer *src_buffer, diff --git a/src/mesa/pipe/i915simple/i915_blit.h b/src/mesa/pipe/i915simple/i915_blit.h index d7a66be10a..6e5b44e124 100644 --- a/src/mesa/pipe/i915simple/i915_blit.h +++ b/src/mesa/pipe/i915simple/i915_blit.h @@ -31,6 +31,7 @@ #include "i915_context.h" extern void i915_copy_blit(struct i915_context *i915, + unsigned do_flip, unsigned cpp, short src_pitch, struct pipe_buffer *src_buffer, diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 1bdaba773f..6d4b8a0aa9 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -83,6 +83,7 @@ i915_get_tex_surface(struct pipe_context *pipe, */ static void i915_surface_copy(struct pipe_context *pipe, + unsigned do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, @@ -98,14 +99,15 @@ i915_surface_copy(struct pipe_context *pipe, dstx, dsty, width, height, pipe_surface_map(src), - src->pitch, - srcx, srcy); + do_flip ? -src->pitch : src->pitch, + srcx, do_flip ? 1 - srcy - height : srcy); pipe_surface_unmap(src); pipe_surface_unmap(dst); } else { i915_copy_blit( i915_context(pipe), + do_flip, dst->cpp, (short) src->pitch, src->buffer, src->offset, (short) dst->pitch, dst->buffer, dst->offset, diff --git a/src/mesa/pipe/i965simple/brw_blit.c b/src/mesa/pipe/i965simple/brw_blit.c index bbd366294f..8494f70493 100644 --- a/src/mesa/pipe/i965simple/brw_blit.c +++ b/src/mesa/pipe/i965simple/brw_blit.c @@ -111,6 +111,7 @@ static unsigned translate_raster_op(unsigned logicop) /* Copy BitBlt */ void brw_copy_blit(struct brw_context *brw, + unsigned do_flip, unsigned cpp, short src_pitch, struct pipe_buffer *src_buffer, diff --git a/src/mesa/pipe/i965simple/brw_blit.h b/src/mesa/pipe/i965simple/brw_blit.h index 7f17a70173..111c5d91d3 100644 --- a/src/mesa/pipe/i965simple/brw_blit.h +++ b/src/mesa/pipe/i965simple/brw_blit.h @@ -16,6 +16,7 @@ void brw_fill_blit(struct brw_context *intel, short w, short h, unsigned color); void brw_copy_blit(struct brw_context *intel, + unsigned do_flip, unsigned cpp, short src_pitch, struct pipe_buffer *src_buffer, diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c index eb7835836e..518845e4b2 100644 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ b/src/mesa/pipe/i965simple/brw_surface.c @@ -103,6 +103,7 @@ brw_surface_data(struct pipe_context *pipe, */ static void brw_surface_copy(struct pipe_context *pipe, + unsigned do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, @@ -118,14 +119,15 @@ brw_surface_copy(struct pipe_context *pipe, dstx, dsty, width, height, pipe_surface_map(src) + src->offset, - src->pitch, - srcx, srcy); + do_flip ? -src->pitch : src->pitch, + srcx, do_flip ? 1 - srcy - height : srcy); pipe_surface_unmap(src); pipe_surface_unmap(dst); } else { brw_copy_blit(brw_context(pipe), + do_flip, dst->cpp, (short) src->pitch, src->buffer, src->offset, FALSE, (short) dst->pitch, dst->buffer, dst->offset, FALSE, diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h index 92a1cd70c4..39f95695fb 100644 --- a/src/mesa/pipe/p_context.h +++ b/src/mesa/pipe/p_context.h @@ -178,6 +178,7 @@ struct pipe_context { */ void (*surface_copy)(struct pipe_context *pipe, + unsigned do_flip, /*<< flip surface contents vertically */ struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, /* don't make this const - diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index 469920efee..a8938a7e43 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -402,7 +402,7 @@ static INLINE int align(int value, int alignment) extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, unsigned dst_x, unsigned dst_y, unsigned width, unsigned height, const ubyte * src, - unsigned src_pitch, unsigned src_x, unsigned src_y); + int src_pitch, unsigned src_x, int src_y); #endif diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 5978ee48bd..5c6ed3b8d9 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -40,6 +40,7 @@ */ static void sp_surface_copy(struct pipe_context *pipe, + unsigned do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, @@ -53,8 +54,8 @@ sp_surface_copy(struct pipe_context *pipe, dstx, dsty, width, height, pipe_surface_map(src), - src->pitch, - srcx, srcy); + do_flip ? -src->pitch : src->pitch, + srcx, do_flip ? 1 - srcy - height : srcy); pipe_surface_unmap(src); pipe_surface_unmap(dst); diff --git a/src/mesa/pipe/util/p_util.c b/src/mesa/pipe/util/p_util.c index c4882b77d2..2a92f8e408 100644 --- a/src/mesa/pipe/util/p_util.c +++ b/src/mesa/pipe/util/p_util.c @@ -47,9 +47,9 @@ pipe_copy_rect(ubyte * dst, unsigned width, unsigned height, const ubyte * src, - unsigned src_pitch, + int src_pitch, unsigned src_x, - unsigned src_y) + int src_y) { unsigned i; diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c index 03985eab5a..e5fef1d7a8 100644 --- a/src/mesa/pipe/xlib/xm_api.c +++ b/src/mesa/pipe/xlib/xm_api.c @@ -1229,6 +1229,7 @@ void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height ) return; pipe->surface_copy(pipe, + FALSE, surf_front, x, y, /* dest */ surf_back, x, y, /* src */ width, height); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 3245a7488b..f13199a3c0 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1259,6 +1259,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, if (st->haveFramebufferSurfaces) { /* copy source framebuffer surface into mipmap/texture */ pipe->surface_copy(pipe, + FALSE, psTex, /* dest */ 0, 0, /* destx/y */ psRead, diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 0ea367549b..91a40288cc 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1132,6 +1132,7 @@ do_copy_texsubimage(GLcontext *ctx, struct pipe_context *pipe = ctx->st->pipe; struct pipe_surface *dest_surface; uint dest_format, src_format; + uint do_flip = FALSE; (void) texImage; @@ -1153,6 +1154,7 @@ do_copy_texsubimage(GLcontext *ctx, if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { srcY = strb->Base.Height - srcY - height; + do_flip = TRUE; } src_format = strb->surface->format; @@ -1190,6 +1192,7 @@ do_copy_texsubimage(GLcontext *ctx, #else pipe->surface_copy(pipe, + do_flip, /* dest */ dest_surface, destX, destY, diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 2622d00953..b86f416c9b 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -300,6 +300,7 @@ st_texture_image_copy(struct pipe_context *pipe, src_surface = pipe->get_tex_surface(pipe, src, face, srcLevel, i); pipe->surface_copy(pipe, + FALSE, dst_surface, 0, 0, /* destX, Y */ src_surface, -- cgit v1.2.3 From 7f342a20d2c7f59b8dd8daed21f3b44f5215a05a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 13 Feb 2008 10:02:59 +0000 Subject: tgsi: include more of the register info in debug dumps --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 40bacf8552..7d82a4b19b 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -44,6 +44,9 @@ static void _print_reg( struct x86_reg reg ) { + if (reg.mod != mod_REG) + debug_printf( "[" ); + switch( reg.file ) { case file_REG32: switch( reg.idx ) { @@ -83,6 +86,13 @@ _print_reg( assert( 0 ); break; } + + if (reg.mod == mod_DISP8 || + reg.mod == mod_DISP32) + debug_printf("+%d", reg.disp); + + if (reg.mod != mod_REG) + debug_printf( "]" ); } static void -- cgit v1.2.3 From 6046c54cc40d32d4c1a47c061494a37fadefd947 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 13 Feb 2008 11:35:54 +0000 Subject: x86: reallocate exec mem when we run out --- src/mesa/x86/rtasm/x86sse.c | 95 ++++++++++++++++++++++++++++----------------- src/mesa/x86/rtasm/x86sse.h | 1 + 2 files changed, 61 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index f8da6e405f..385fb84c01 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -12,55 +12,75 @@ static unsigned char *cptr( void (*label)() ) } +static void do_realloc( struct x86_function *p ) +{ + _mesa_printf("do_realloc %d %p\n", p->size, p->store); + + if (p->size == 0) { + p->size = 1024; + p->store = _mesa_exec_malloc(p->size); + p->csr = p->store; + } + else { + unsigned used = p->csr - p->store; + unsigned char *tmp = p->store; + p->size *= 2; + p->store = _mesa_exec_malloc(p->size); + memcpy(p->store, tmp, used); + p->csr = p->store + used; + _mesa_exec_free(tmp); + } +} + /* Emit bytes to the instruction stream: */ -static void emit_1b( struct x86_function *p, char b0 ) +static unsigned char *reserve( struct x86_function *p, int bytes ) { - *(char *)(p->csr++) = b0; + if (p->csr + bytes - p->store > p->size) + do_realloc(p); + + { + unsigned char *csr = p->csr; + p->csr += bytes; + return csr; + } } -static void emit_1i( struct x86_function *p, int i0 ) + + +static void emit_1b( struct x86_function *p, char b0 ) { - *(int *)(p->csr) = i0; - p->csr += 4; + char *csr = (char *)reserve(p, 1); + *csr = b0; } -static void disassem( struct x86_function *p, const char *fn ) +static void emit_1i( struct x86_function *p, int i0 ) { -#if DISASSEM && 0 - if (fn && fn != p->fn) { - _mesa_printf("0x%x: %s\n", p->csr, fn); - p->fn = fn; - } -#endif + int *icsr = (int *)reserve(p, sizeof(i0)); + *icsr = i0; } -static void emit_1ub_fn( struct x86_function *p, unsigned char b0, const char *fn ) +static void emit_1ub( struct x86_function *p, unsigned char b0 ) { - disassem(p, fn); - *(p->csr++) = b0; + unsigned char *csr = reserve(p, 1); + *csr++ = b0; } -static void emit_2ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, const char *fn ) +static void emit_2ub( struct x86_function *p, unsigned char b0, unsigned char b1 ) { - disassem(p, fn); - *(p->csr++) = b0; - *(p->csr++) = b1; + unsigned char *csr = reserve(p, 2); + *csr++ = b0; + *csr++ = b1; } -static void emit_3ub_fn( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2, const char *fn ) +static void emit_3ub( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2 ) { - disassem(p, fn); - *(p->csr++) = b0; - *(p->csr++) = b1; - *(p->csr++) = b2; + unsigned char *csr = reserve(p, 3); + *csr++ = b0; + *csr++ = b1; + *csr++ = b2; } -#define emit_1ub(p, b0) emit_1ub_fn(p, b0, __FUNCTION__) -#define emit_2ub(p, b0, b1) emit_2ub_fn(p, b0, b1, __FUNCTION__) -#define emit_3ub(p, b0, b1, b2) emit_3ub_fn(p, b0, b1, b2, __FUNCTION__) - - /* Build a modRM byte + possible displacement. No treatment of SIB * indexing. BZZT - no way to encode an absolute address. @@ -77,13 +97,13 @@ static void emit_modrm( struct x86_function *p, val |= reg.idx << 3; /* reg field */ val |= regmem.idx; /* r/m field */ - emit_1ub_fn(p, val, 0); + emit_1ub(p, val); /* Oh-oh we've stumbled into the SIB thing. */ if (regmem.file == file_REG32 && regmem.idx == reg_SP) { - emit_1ub_fn(p, 0x24, 0); /* simplistic! */ + emit_1ub(p, 0x24); /* simplistic! */ } switch (regmem.mod) { @@ -124,14 +144,14 @@ static void emit_op_modrm( struct x86_function *p, { switch (dst.mod) { case mod_REG: - emit_1ub_fn(p, op_dst_is_reg, 0); + emit_1ub(p, op_dst_is_reg); emit_modrm(p, dst, src); break; case mod_INDIRECT: case mod_DISP32: case mod_DISP8: assert(src.mod == mod_REG); - emit_1ub_fn(p, op_dst_is_mem, 0); + emit_1ub(p, op_dst_is_mem); emit_modrm(p, src, dst); break; default: @@ -1125,11 +1145,14 @@ struct x86_reg x86_fn_arg( struct x86_function *p, void x86_init_func( struct x86_function *p ) { - x86_init_func_size(p, 2048); + p->size = 0; + p->store = NULL; + p->csr = p->store; } void x86_init_func_size( struct x86_function *p, unsigned code_size ) { + p->size = code_size; p->store = _mesa_exec_malloc(code_size); p->csr = p->store; } @@ -1138,12 +1161,14 @@ void x86_release_func( struct x86_function *p ) { _mesa_exec_free(p->store); p->store = NULL; + p->csr = NULL; + p->size = 0; } void (*x86_get_func( struct x86_function *p ))(void) { - if (DISASSEM) + if (DISASSEM && p->store) _mesa_printf("disassemble %p %p\n", p->store, p->csr); return (void (*)(void)) (unsigned long) p->store; } diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h index 63b9a36392..d53b6d71a6 100644 --- a/src/mesa/x86/rtasm/x86sse.h +++ b/src/mesa/x86/rtasm/x86sse.h @@ -16,6 +16,7 @@ struct x86_reg { }; struct x86_function { + unsigned size; unsigned char *store; unsigned char *csr; unsigned stack_offset; -- cgit v1.2.3 From a3534a27bfc9827a12d83f7b6464af98424cf8d4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 13 Feb 2008 11:39:34 +0000 Subject: tgsi: fail gracefully when sse code can't emit shader --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 161 +++++++++++++++++++----------------- 1 file changed, 87 insertions(+), 74 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 7d82a4b19b..ecf4ca8d21 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -1436,7 +1436,7 @@ emit_cmp( } } -static void +static int emit_instruction( struct x86_function *func, struct tgsi_full_instruction *inst ) @@ -1547,11 +1547,11 @@ emit_instruction( break; case TGSI_OPCODE_EXP: - assert( 0 ); + return 0; break; case TGSI_OPCODE_LOG: - assert( 0 ); + return 0; break; case TGSI_OPCODE_MUL: @@ -1706,24 +1706,24 @@ emit_instruction( break; case TGSI_OPCODE_CND: - assert( 0 ); + return 0; break; case TGSI_OPCODE_CND0: - assert( 0 ); + return 0; break; case TGSI_OPCODE_DOT2ADD: /* TGSI_OPCODE_DP2A */ - assert( 0 ); + return 0; break; case TGSI_OPCODE_INDEX: - assert( 0 ); + return 0; break; case TGSI_OPCODE_NEGATE: - assert( 0 ); + return 0; break; case TGSI_OPCODE_FRAC: @@ -1736,7 +1736,7 @@ emit_instruction( break; case TGSI_OPCODE_CLAMP: - assert( 0 ); + return 0; break; case TGSI_OPCODE_FLOOR: @@ -1749,7 +1749,7 @@ emit_instruction( break; case TGSI_OPCODE_ROUND: - assert( 0 ); + return 0; break; case TGSI_OPCODE_EXPBASE2: @@ -1824,7 +1824,7 @@ emit_instruction( break; case TGSI_OPCODE_MULTIPLYMATRIX: - assert( 0 ); + return 0; break; case TGSI_OPCODE_ABS: @@ -1837,7 +1837,7 @@ emit_instruction( break; case TGSI_OPCODE_RCC: - assert( 0 ); + return 0; break; case TGSI_OPCODE_DPH: @@ -1868,11 +1868,11 @@ emit_instruction( break; case TGSI_OPCODE_DDX: - assert( 0 ); + return 0; break; case TGSI_OPCODE_DDY: - assert( 0 ); + return 0; break; case TGSI_OPCODE_KIL: @@ -1880,35 +1880,35 @@ emit_instruction( break; case TGSI_OPCODE_PK2H: - assert( 0 ); + return 0; break; case TGSI_OPCODE_PK2US: - assert( 0 ); + return 0; break; case TGSI_OPCODE_PK4B: - assert( 0 ); + return 0; break; case TGSI_OPCODE_PK4UB: - assert( 0 ); + return 0; break; case TGSI_OPCODE_RFL: - assert( 0 ); + return 0; break; case TGSI_OPCODE_SEQ: - assert( 0 ); + return 0; break; case TGSI_OPCODE_SFL: - assert( 0 ); + return 0; break; case TGSI_OPCODE_SGT: - assert( 0 ); + return 0; break; case TGSI_OPCODE_SIN: @@ -1920,15 +1920,15 @@ emit_instruction( break; case TGSI_OPCODE_SLE: - assert( 0 ); + return 0; break; case TGSI_OPCODE_SNE: - assert( 0 ); + return 0; break; case TGSI_OPCODE_STR: - assert( 0 ); + return 0; break; case TGSI_OPCODE_TEX: @@ -1943,43 +1943,43 @@ emit_instruction( break; case TGSI_OPCODE_TXD: - assert( 0 ); + return 0; break; case TGSI_OPCODE_UP2H: - assert( 0 ); + return 0; break; case TGSI_OPCODE_UP2US: - assert( 0 ); + return 0; break; case TGSI_OPCODE_UP4B: - assert( 0 ); + return 0; break; case TGSI_OPCODE_UP4UB: - assert( 0 ); + return 0; break; case TGSI_OPCODE_X2D: - assert( 0 ); + return 0; break; case TGSI_OPCODE_ARA: - assert( 0 ); + return 0; break; case TGSI_OPCODE_ARR: - assert( 0 ); + return 0; break; case TGSI_OPCODE_BRA: - assert( 0 ); + return 0; break; case TGSI_OPCODE_CAL: - assert( 0 ); + return 0; break; case TGSI_OPCODE_RET: @@ -1992,7 +1992,7 @@ emit_instruction( break; case TGSI_OPCODE_SSG: - assert( 0 ); + return 0; break; case TGSI_OPCODE_CMP: @@ -2021,132 +2021,134 @@ emit_instruction( break; case TGSI_OPCODE_TXB: - assert( 0 ); + return 0; break; case TGSI_OPCODE_NRM: - assert( 0 ); + return 0; break; case TGSI_OPCODE_DIV: - assert( 0 ); + return 0; break; case TGSI_OPCODE_DP2: - assert( 0 ); + return 0; break; case TGSI_OPCODE_TXL: - assert( 0 ); + return 0; break; case TGSI_OPCODE_BRK: - assert( 0 ); + return 0; break; case TGSI_OPCODE_IF: - assert( 0 ); + return 0; break; case TGSI_OPCODE_LOOP: - assert( 0 ); + return 0; break; case TGSI_OPCODE_REP: - assert( 0 ); + return 0; break; case TGSI_OPCODE_ELSE: - assert( 0 ); + return 0; break; case TGSI_OPCODE_ENDIF: - assert( 0 ); + return 0; break; case TGSI_OPCODE_ENDLOOP: - assert( 0 ); + return 0; break; case TGSI_OPCODE_ENDREP: - assert( 0 ); + return 0; break; case TGSI_OPCODE_PUSHA: - assert( 0 ); + return 0; break; case TGSI_OPCODE_POPA: - assert( 0 ); + return 0; break; case TGSI_OPCODE_CEIL: - assert( 0 ); + return 0; break; case TGSI_OPCODE_I2F: - assert( 0 ); + return 0; break; case TGSI_OPCODE_NOT: - assert( 0 ); + return 0; break; case TGSI_OPCODE_TRUNC: - assert( 0 ); + return 0; break; case TGSI_OPCODE_SHL: - assert( 0 ); + return 0; break; case TGSI_OPCODE_SHR: - assert( 0 ); + return 0; break; case TGSI_OPCODE_AND: - assert( 0 ); + return 0; break; case TGSI_OPCODE_OR: - assert( 0 ); + return 0; break; case TGSI_OPCODE_MOD: - assert( 0 ); + return 0; break; case TGSI_OPCODE_XOR: - assert( 0 ); + return 0; break; case TGSI_OPCODE_SAD: - assert( 0 ); + return 0; break; case TGSI_OPCODE_TXF: - assert( 0 ); + return 0; break; case TGSI_OPCODE_TXQ: - assert( 0 ); + return 0; break; case TGSI_OPCODE_CONT: - assert( 0 ); + return 0; break; case TGSI_OPCODE_EMIT: - assert( 0 ); + return 0; break; case TGSI_OPCODE_ENDPRIM: - assert( 0 ); + return 0; break; default: - assert( 0 ); + return 0; } + + return 1; } static void @@ -2212,6 +2214,7 @@ emit_declaration( default: assert( 0 ); + break; } } } @@ -2225,6 +2228,7 @@ tgsi_emit_sse2( struct x86_function *func ) { struct tgsi_parse_context parse; + unsigned ok = 1; DUMP_START(); @@ -2249,7 +2253,7 @@ tgsi_emit_sse2( tgsi_parse_init( &parse, tokens ); - while( !tgsi_parse_end_of_tokens( &parse ) ) { + while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { @@ -2257,17 +2261,26 @@ tgsi_emit_sse2( break; case TGSI_TOKEN_TYPE_INSTRUCTION: - emit_instruction( - func, - &parse.FullToken.FullInstruction ); + ok = emit_instruction( + func, + &parse.FullToken.FullInstruction ); + + if (!ok) { + debug_printf("failed to translate tgsi opcode %d\n", + parse.FullToken.FullInstruction.Instruction.Opcode ); + } break; case TGSI_TOKEN_TYPE_IMMEDIATE: /* XXX implement this */ - return 0; + ok = 0; + debug_printf("failed to emit immediate value\n"); + break; default: assert( 0 ); + ok = 0; + break; } } @@ -2275,7 +2288,7 @@ tgsi_emit_sse2( DUMP_END(); - return 1; + return ok; } /** -- cgit v1.2.3 From 8162d317d2f6f2dcc31f31c0c2d663c33dfee053 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 13 Feb 2008 12:35:16 +0000 Subject: x86: emit absolute calls, as reallocating exec mem breaks relative ones --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 5 ++++- src/mesa/x86/rtasm/x86sse.c | 13 +++++++++++++ src/mesa/x86/rtasm/x86sse.h | 3 ++- 3 files changed, 19 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index ecf4ca8d21..1e56e4afb6 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -328,8 +328,11 @@ emit_call( struct x86_function *func, void (* addr)() ) { + struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); + DUMP_I( "CALL", addr ); - x86_call( func, addr ); + x86_mov_reg_imm( func, ecx, (unsigned long) addr ); + x86_call( func, ecx ); } static void diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index 385fb84c01..e944d00f9e 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -278,11 +278,24 @@ void x86_jmp( struct x86_function *p, unsigned char *label) emit_1i(p, label - x86_get_label(p) - 4); } +#if 0 +/* This doesn't work once we start reallocating & copying the + * generated code on buffer fills, because the call is relative to the + * current pc. + */ void x86_call( struct x86_function *p, void (*label)()) { emit_1ub(p, 0xe8); emit_1i(p, cptr(label) - x86_get_label(p) - 4); } +#else +void x86_call( struct x86_function *p, struct x86_reg reg) +{ + emit_1ub(p, 0xff); + emit_modrm(p, reg, reg); +} +#endif + /* michal: * Temporary. As I need immediate operands, and dont want to mess with the codegen, diff --git a/src/mesa/x86/rtasm/x86sse.h b/src/mesa/x86/rtasm/x86sse.h index d53b6d71a6..c2aa416492 100644 --- a/src/mesa/x86/rtasm/x86sse.h +++ b/src/mesa/x86/rtasm/x86sse.h @@ -119,7 +119,8 @@ void x86_fixup_fwd_jump( struct x86_function *p, void x86_jmp( struct x86_function *p, unsigned char *label ); -void x86_call( struct x86_function *p, void (*label)() ); +/* void x86_call( struct x86_function *p, void (*label)() ); */ +void x86_call( struct x86_function *p, struct x86_reg reg); /* michal: * Temporary. As I need immediate operands, and dont want to mess with the codegen, -- cgit v1.2.3 From 66640c4b589db7b6b5edce7d297ae6623bfda9c1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 13 Feb 2008 12:37:42 +0000 Subject: x86: remove debug --- src/mesa/x86/rtasm/x86sse.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index e944d00f9e..1111e8db99 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -14,8 +14,6 @@ static unsigned char *cptr( void (*label)() ) static void do_realloc( struct x86_function *p ) { - _mesa_printf("do_realloc %d %p\n", p->size, p->store); - if (p->size == 0) { p->size = 1024; p->store = _mesa_exec_malloc(p->size); -- cgit v1.2.3 From c107d572104890c83647ad611d303fe1fae70e8f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 07:28:26 -0500 Subject: try to load the consts correctly --- src/mesa/pipe/llvm/storagesoa.cpp | 20 ++++++++++++++++++++ src/mesa/pipe/llvm/storagesoa.h | 2 ++ 2 files changed, 22 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index e09e9e8fe7..7292333a80 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -81,6 +81,15 @@ std::vector StorageSoa::constElement(int idx, int swizzle, llvm::Value *indIdx) { std::vector res(4); + llvm::Value *xChannel = elementPointer(m_consts, idx, 0); + llvm::Value *yChannel = elementPointer(m_consts, idx, 1); + llvm::Value *zChannel = elementPointer(m_consts, idx, 2); + llvm::Value *wChannel = elementPointer(m_consts, idx, 3); + + res[0] = alignedArrayLoad(xChannel); + res[1] = alignedArrayLoad(yChannel); + res[2] = alignedArrayLoad(zChannel); + res[3] = alignedArrayLoad(wChannel); return res; } @@ -206,3 +215,14 @@ llvm::ConstantInt * StorageSoa::constantInt(int idx) const m_constInts[idx] = constInt; return constInt; } + +llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + PointerType *vectorPtr = PointerType::get(vectorType, 0); + + CastInst *cast = new BitCastInst(val, vectorPtr, name("toVector"), m_block); + LoadInst *load = new LoadInst(cast, name("alignLoad"), false, m_block); + load->setAlignment(8); + return load; +} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index 84db7726a7..43b23951de 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -74,6 +74,8 @@ private: int channel) const; const char *name(const char *prefix) const; llvm::ConstantInt *constantInt(int) const; + llvm::Value *alignedArrayLoad(llvm::Value *val); + private: llvm::BasicBlock *m_block; -- cgit v1.2.3 From 7e51d2a9986231f4ca44a3643628645923e4d468 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 08:31:13 -0500 Subject: implement immediates and make them work --- src/mesa/pipe/llvm/storagesoa.cpp | 65 +++++++++++++++++++++++++++++++++++++++ src/mesa/pipe/llvm/storagesoa.h | 7 ++++- 2 files changed, 71 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index 7292333a80..7b758b1665 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -41,6 +41,7 @@ using namespace llvm; + StorageSoa::StorageSoa(llvm::BasicBlock *block, llvm::Value *input, llvm::Value *output, @@ -57,6 +58,28 @@ StorageSoa::StorageSoa(llvm::BasicBlock *block, void StorageSoa::addImmediate(float *vec) { + float vals[4]; //decompose into soa + + vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0]; + llvm::Value *xChannel = createConstGlobalVector(vals); + + vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1]; + llvm::Value *yChannel = createConstGlobalVector(vals); + + + vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2]; + llvm::Value *zChannel = createConstGlobalVector(vals); + + vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3]; + llvm::Value *wChannel = createConstGlobalVector(vals); + + std::vector res(4); + res[0] = xChannel; + res[1] = yChannel; + res[2] = zChannel; + res[3] = wChannel; + + m_immediates[m_immediates.size()] = res; } llvm::Value *StorageSoa::addrElement(int idx) const @@ -123,6 +146,12 @@ std::vector StorageSoa::tempElement(int idx, int swizzle, std::vector StorageSoa::immediateElement(int idx, int swizzle) { std::vector res(4); + res = m_immediates[idx]; + + res[0] = new LoadInst(res[0], name("immx"), false, m_block); + res[1] = new LoadInst(res[1], name("immx"), false, m_block); + res[2] = new LoadInst(res[2], name("immx"), false, m_block); + res[3] = new LoadInst(res[3], name("immx"), false, m_block); return res; } @@ -226,3 +255,39 @@ llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val) load->setAlignment(8); return load; } + +llvm::Module * StorageSoa::currentModule() const +{ + if (!m_block || !m_block->getParent()) + return 0; + + return m_block->getParent()->getParent(); +} + +llvm::Value * StorageSoa::createConstGlobalVector(float *vec) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + GlobalVariable *immediate = new GlobalVariable( + /*Type=*/vectorType, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name("immediate"), + currentModule()); + + std::vector immValues; + ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0])); + ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1])); + ConstantFP *constz = ConstantFP::get(Type::FloatTy, APFloat(vec[2])); + ConstantFP *constw = ConstantFP::get(Type::FloatTy, APFloat(vec[3])); + immValues.push_back(constx); + immValues.push_back(consty); + immValues.push_back(constz); + immValues.push_back(constw); + Constant *constVector = ConstantVector::get(vectorType, immValues); + + // Global Variable Definitions + immediate->setInitializer(constVector); + + return immediate; +} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index 43b23951de..2d07e836f4 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -38,6 +38,7 @@ namespace llvm { class LoadInst; class Value; class VectorType; + class Module; } class StorageSoa @@ -74,7 +75,9 @@ private: int channel) const; const char *name(const char *prefix) const; llvm::ConstantInt *constantInt(int) const; - llvm::Value *alignedArrayLoad(llvm::Value *val); + llvm::Value *alignedArrayLoad(llvm::Value *val); + llvm::Module *currentModule() const; + llvm::Value *createConstGlobalVector(float *vec); private: llvm::BasicBlock *m_block; @@ -84,6 +87,8 @@ private: llvm::Value *m_consts; llvm::Value *m_temps; + std::map > m_immediates; + mutable std::map m_constInts; mutable char m_name[32]; mutable int m_idx; -- cgit v1.2.3 From 12d5b078e8390cb2c598def31fd75260a54ace1a Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 08:46:05 -0500 Subject: get rid of the terrible auto-generated entry point it's not used anymore. --- src/mesa/pipe/llvm/Makefile | 6 +- src/mesa/pipe/llvm/llvm_base_shader.cpp | 833 -------------------------------- src/mesa/pipe/llvm/llvm_entry.c | 163 ------- src/mesa/pipe/llvm/tgsitollvm.cpp | 3 +- 4 files changed, 3 insertions(+), 1002 deletions(-) delete mode 100644 src/mesa/pipe/llvm/llvm_base_shader.cpp delete mode 100644 src/mesa/pipe/llvm/llvm_entry.c (limited to 'src') diff --git a/src/mesa/pipe/llvm/Makefile b/src/mesa/pipe/llvm/Makefile index a0494ba966..9c6e16d86b 100644 --- a/src/mesa/pipe/llvm/Makefile +++ b/src/mesa/pipe/llvm/Makefile @@ -15,7 +15,7 @@ GALLIVM_SOURCES = \ storagesoa.cpp \ instructionssoa.cpp -INC_SOURCES = gallivm_builtins.cpp llvm_base_shader.cpp +INC_SOURCES = gallivm_builtins.cpp CPP_SOURCES = \ $(GALLIVM_SOURCES) @@ -65,8 +65,6 @@ depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) gallivm_builtins.cpp: llvm_builtins.c clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins -llvm_base_shader.cpp: llvm_entry.c - clang --emit-llvm $< |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -o=$@ -f -funcname=createBaseShader # Emacs tags tags: @@ -77,7 +75,7 @@ tags: clean: -rm -f *.o */*.o *~ *.so *~ server/*.o -rm -f depend depend.bak - -rm -f gallivm_builtins.cpp llvm_base_shader.cpp + -rm -f gallivm_builtins.cpp symlinks: diff --git a/src/mesa/pipe/llvm/llvm_base_shader.cpp b/src/mesa/pipe/llvm/llvm_base_shader.cpp deleted file mode 100644 index 90a25a440a..0000000000 --- a/src/mesa/pipe/llvm/llvm_base_shader.cpp +++ /dev/null @@ -1,833 +0,0 @@ -// Generated by llvm2cpp - DO NOT MODIFY! - - -Module* createBaseShader() { - // Module Construction - Module* mod = new Module("Shader"); - mod->setDataLayout("e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"); - mod->setTargetTriple("i686-pc-linux-gnu"); - - // Type Definitions - std::vectorStructTy_struct_ShaderInput_fields; - VectorType* VectorTy_1 = VectorType::get(Type::FloatTy, 4); - - PointerType* PointerTy_0 = PointerType::get(VectorTy_1, 0); - - StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); - StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); - StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); - StructTy_struct_ShaderInput_fields.push_back(PointerTy_0); - StructTy_struct_ShaderInput_fields.push_back(IntegerType::get(32)); - StructType* StructTy_struct_ShaderInput = StructType::get(StructTy_struct_ShaderInput_fields, /*isPacked=*/false); - mod->addTypeName("struct.ShaderInput", StructTy_struct_ShaderInput); - - OpaqueType* OpaqueTy_struct_pipe_sampler_state = OpaqueType::get(); - mod->addTypeName("struct.pipe_sampler_state", OpaqueTy_struct_pipe_sampler_state); - - OpaqueType* OpaqueTy_struct_softpipe_tile_cache = OpaqueType::get(); - mod->addTypeName("struct.softpipe_tile_cache", OpaqueTy_struct_softpipe_tile_cache); - - std::vectorStructTy_struct_tgsi_sampler_fields; - PointerType* PointerTy_2 = PointerType::get(OpaqueTy_struct_pipe_sampler_state, 0); - - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_2); - std::vectorFuncTy_4_args; - PATypeHolder StructTy_struct_tgsi_sampler_fwd = OpaqueType::get(); - PointerType* PointerTy_5 = PointerType::get(StructTy_struct_tgsi_sampler_fwd, 0); - - FuncTy_4_args.push_back(PointerTy_5); - PointerType* PointerTy_6 = PointerType::get(Type::FloatTy, 0); - - FuncTy_4_args.push_back(PointerTy_6); - FuncTy_4_args.push_back(PointerTy_6); - FuncTy_4_args.push_back(PointerTy_6); - FuncTy_4_args.push_back(Type::FloatTy); - ArrayType* ArrayTy_8 = ArrayType::get(Type::FloatTy, 4); - - PointerType* PointerTy_7 = PointerType::get(ArrayTy_8, 0); - - FuncTy_4_args.push_back(PointerTy_7); - FunctionType* FuncTy_4 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_4_args, - /*isVarArg=*/false); - - PointerType* PointerTy_3 = PointerType::get(FuncTy_4, 0); - - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_3); - PointerType* PointerTy_9 = PointerType::get(IntegerType::get(8), 0); - - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_9); - PointerType* PointerTy_10 = PointerType::get(OpaqueTy_struct_softpipe_tile_cache, 0); - - StructTy_struct_tgsi_sampler_fields.push_back(PointerTy_10); - StructType* StructTy_struct_tgsi_sampler = StructType::get(StructTy_struct_tgsi_sampler_fields, /*isPacked=*/false); - mod->addTypeName("struct.tgsi_sampler", StructTy_struct_tgsi_sampler); - cast(StructTy_struct_tgsi_sampler_fwd.get())->refineAbstractTypeTo(StructTy_struct_tgsi_sampler); - StructTy_struct_tgsi_sampler = cast(StructTy_struct_tgsi_sampler_fwd.get()); - - - std::vectorFuncTy_11_args; - ArrayType* ArrayTy_13 = ArrayType::get(VectorTy_1, 16); - - PointerType* PointerTy_12 = PointerType::get(ArrayTy_13, 0); - - FuncTy_11_args.push_back(PointerTy_12); - ArrayType* ArrayTy_15 = ArrayType::get(ArrayTy_8, 16); - - PointerType* PointerTy_14 = PointerType::get(ArrayTy_15, 0); - - FuncTy_11_args.push_back(PointerTy_14); - FuncTy_11_args.push_back(IntegerType::get(32)); - FuncTy_11_args.push_back(IntegerType::get(32)); - FunctionType* FuncTy_11 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_11_args, - /*isVarArg=*/false); - - std::vectorFuncTy_16_args; - FuncTy_16_args.push_back(PointerTy_0); - FuncTy_16_args.push_back(PointerTy_7); - FuncTy_16_args.push_back(IntegerType::get(32)); - FunctionType* FuncTy_16 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_16_args, - /*isVarArg=*/false); - - std::vectorFuncTy_17_args; - FuncTy_17_args.push_back(PointerTy_7); - FuncTy_17_args.push_back(PointerTy_0); - FuncTy_17_args.push_back(IntegerType::get(32)); - FunctionType* FuncTy_17 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_17_args, - /*isVarArg=*/false); - - std::vectorFuncTy_18_args; - FuncTy_18_args.push_back(PointerTy_9); - FuncTy_18_args.push_back(PointerTy_9); - FuncTy_18_args.push_back(PointerTy_7); - FuncTy_18_args.push_back(IntegerType::get(32)); - FuncTy_18_args.push_back(IntegerType::get(32)); - FuncTy_18_args.push_back(IntegerType::get(32)); - FuncTy_18_args.push_back(IntegerType::get(32)); - FunctionType* FuncTy_18 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_18_args, - /*isVarArg=*/false); - - ArrayType* ArrayTy_20 = ArrayType::get(VectorTy_1, 32); - - PointerType* PointerTy_19 = PointerType::get(ArrayTy_20, 0); - - ArrayType* ArrayTy_22 = ArrayType::get(VectorTy_1, 128); - - PointerType* PointerTy_21 = PointerType::get(ArrayTy_22, 0); - - PointerType* PointerTy_23 = PointerType::get(StructTy_struct_ShaderInput, 0); - - PointerType* PointerTy_24 = PointerType::get(PointerTy_0, 0); - - std::vectorFuncTy_26_args; - FuncTy_26_args.push_back(PointerTy_23); - FunctionType* FuncTy_26 = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/FuncTy_26_args, - /*isVarArg=*/false); - - PointerType* PointerTy_25 = PointerType::get(FuncTy_26, 0); - - std::vectorFuncTy_27_args; - FuncTy_27_args.push_back(Type::FloatTy); - FuncTy_27_args.push_back(Type::FloatTy); - FuncTy_27_args.push_back(PointerTy_12); - FuncTy_27_args.push_back(PointerTy_12); - FuncTy_27_args.push_back(IntegerType::get(32)); - FuncTy_27_args.push_back(PointerTy_7); - FuncTy_27_args.push_back(IntegerType::get(32)); - FuncTy_27_args.push_back(PointerTy_5); - FunctionType* FuncTy_27 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_27_args, - /*isVarArg=*/false); - - PointerType* PointerTy_28 = PointerType::get(IntegerType::get(32), 0); - - - // Function Declarations - - Function* func_from_array = new Function( - /*Type=*/FuncTy_11, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"from_array", mod); - func_from_array->setCallingConv(CallingConv::C); - const ParamAttrsList *func_from_array_PAL = 0; - { - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_from_array_PAL = ParamAttrsList::get(Attrs); - - } - func_from_array->setParamAttrs(func_from_array_PAL); - - Function* func_from_consts = new Function( - /*Type=*/FuncTy_16, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"from_consts", mod); - func_from_consts->setCallingConv(CallingConv::C); - const ParamAttrsList *func_from_consts_PAL = 0; - { - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_from_consts_PAL = ParamAttrsList::get(Attrs); - - } - func_from_consts->setParamAttrs(func_from_consts_PAL); - - Function* func_to_array = new Function( - /*Type=*/FuncTy_17, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"to_array", mod); - func_to_array->setCallingConv(CallingConv::C); - const ParamAttrsList *func_to_array_PAL = 0; - { - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_to_array_PAL = ParamAttrsList::get(Attrs); - - } - func_to_array->setParamAttrs(func_to_array_PAL); - - Function* func_run_vertex_shader = new Function( - /*Type=*/FuncTy_18, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"run_vertex_shader", mod); - func_run_vertex_shader->setCallingConv(CallingConv::C); - const ParamAttrsList *func_run_vertex_shader_PAL = 0; - func_run_vertex_shader->setParamAttrs(func_run_vertex_shader_PAL); - - Function* func_execute_shader = new Function( - /*Type=*/FuncTy_26, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"execute_shader", mod); // (external, no body) - func_execute_shader->setCallingConv(CallingConv::C); - const ParamAttrsList *func_execute_shader_PAL = 0; - func_execute_shader->setParamAttrs(func_execute_shader_PAL); - - Function* func_run_fragment_shader = new Function( - /*Type=*/FuncTy_27, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"run_fragment_shader", mod); - func_run_fragment_shader->setCallingConv(CallingConv::C); - const ParamAttrsList *func_run_fragment_shader_PAL = 0; - func_run_fragment_shader->setParamAttrs(func_run_fragment_shader_PAL); - - // Global Variable Declarations - - - // Constant Definitions - Constant* const_int32_29 = Constant::getNullValue(IntegerType::get(32)); - ConstantInt* const_int32_30 = ConstantInt::get(APInt(32, "-1", 10)); - ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "1", 10)); - UndefValue* const_packed_32 = UndefValue::get(VectorTy_1); - ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "2", 10)); - ConstantInt* const_int32_34 = ConstantInt::get(APInt(32, "3", 10)); - ConstantInt* const_int32_35 = ConstantInt::get(APInt(32, "4", 10)); - - // Global Variable Definitions - - // Function Definitions - - // Function: from_array (func_from_array) - { - Function::arg_iterator args = func_from_array->arg_begin(); - Value* ptr_res = args++; - ptr_res->setName("res"); - Value* ptr_ainputs = args++; - ptr_ainputs->setName("ainputs"); - Value* int32_count = args++; - int32_count->setName("count"); - Value* int32_num_attribs = args++; - int32_num_attribs->setName("num_attribs"); - - BasicBlock* label_entry = new BasicBlock("entry",func_from_array,0); - BasicBlock* label_forcond2_preheader_split = new BasicBlock("forcond2.preheader.split",func_from_array,0); - BasicBlock* label_forcond2 = new BasicBlock("forcond2",func_from_array,0); - BasicBlock* label_forbody6 = new BasicBlock("forbody6",func_from_array,0); - BasicBlock* label_forinc57 = new BasicBlock("forinc57",func_from_array,0); - BasicBlock* label_afterfor60 = new BasicBlock("afterfor60",func_from_array,0); - - // Block entry (label_entry) - ICmpInst* int1_cmp = new ICmpInst(ICmpInst::ICMP_SGT, int32_count, const_int32_29, "cmp", label_entry); - ICmpInst* int1_cmp5 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs, const_int32_29, "cmp5", label_entry); - BinaryOperator* int1_bothcond = BinaryOperator::create(Instruction::And, int1_cmp, int1_cmp5, "bothcond", label_entry); - new BranchInst(label_forcond2_preheader_split, label_afterfor60, int1_bothcond, label_entry); - - // Block forcond2.preheader.split (label_forcond2_preheader_split) - BinaryOperator* int32_tmp21 = BinaryOperator::create(Instruction::Add, int32_count, const_int32_30, "tmp21", label_forcond2_preheader_split); - ICmpInst* int1_tmp22 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp21, const_int32_29, "tmp22", label_forcond2_preheader_split); - SelectInst* int32_tmp25 = new SelectInst(int1_tmp22, const_int32_31, int32_count, "tmp25", label_forcond2_preheader_split); - new BranchInst(label_forcond2, label_forcond2_preheader_split); - - // Block forcond2 (label_forcond2) - Argument* fwdref_38 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forcond2); - int32_i_0_reg2mem_0->reserveOperandSpace(2); - int32_i_0_reg2mem_0->addIncoming(const_int32_29, label_forcond2_preheader_split); - int32_i_0_reg2mem_0->addIncoming(fwdref_38, label_forinc57); - - Argument* fwdref_39 = new Argument(VectorTy_1); - PHINode* packed_vec_1_reg2mem_0 = new PHINode(VectorTy_1, "vec.1.reg2mem.0", label_forcond2); - packed_vec_1_reg2mem_0->reserveOperandSpace(2); - packed_vec_1_reg2mem_0->addIncoming(const_packed_32, label_forcond2_preheader_split); - packed_vec_1_reg2mem_0->addIncoming(fwdref_39, label_forinc57); - - BinaryOperator* int32_tmp = BinaryOperator::create(Instruction::Add, int32_num_attribs, const_int32_30, "tmp", label_forcond2); - ICmpInst* int1_tmp18 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp, const_int32_29, "tmp18", label_forcond2); - SelectInst* int32_tmp19 = new SelectInst(int1_tmp18, const_int32_31, int32_num_attribs, "tmp19", label_forcond2); - new BranchInst(label_forbody6, label_forcond2); - - // Block forbody6 (label_forbody6) - Argument* fwdref_41 = new Argument(IntegerType::get(32)); - PHINode* int32_j_0_reg2mem_0 = new PHINode(IntegerType::get(32), "j.0.reg2mem.0", label_forbody6); - int32_j_0_reg2mem_0->reserveOperandSpace(2); - int32_j_0_reg2mem_0->addIncoming(const_int32_29, label_forcond2); - int32_j_0_reg2mem_0->addIncoming(fwdref_41, label_forbody6); - - PHINode* packed_vec_0_reg2mem_0 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody6); - packed_vec_0_reg2mem_0->reserveOperandSpace(2); - packed_vec_0_reg2mem_0->addIncoming(packed_vec_1_reg2mem_0, label_forcond2); - packed_vec_0_reg2mem_0->addIncoming(fwdref_39, label_forbody6); - - std::vector ptr_arraydecay11_indices; - ptr_arraydecay11_indices.push_back(int32_i_0_reg2mem_0); - ptr_arraydecay11_indices.push_back(int32_j_0_reg2mem_0); - ptr_arraydecay11_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11 = new GetElementPtrInst(ptr_ainputs, ptr_arraydecay11_indices.begin(), ptr_arraydecay11_indices.end(), "arraydecay11", label_forbody6); - LoadInst* float_tmp13 = new LoadInst(ptr_arraydecay11, "tmp13", false, label_forbody6); - InsertElementInst* packed_tmp15 = new InsertElementInst(packed_vec_0_reg2mem_0, float_tmp13, const_int32_29, "tmp15", label_forbody6); - std::vector ptr_arrayidx23_indices; - ptr_arrayidx23_indices.push_back(int32_i_0_reg2mem_0); - ptr_arrayidx23_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx23_indices.push_back(const_int32_31); - Instruction* ptr_arrayidx23 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx23_indices.begin(), ptr_arrayidx23_indices.end(), "arrayidx23", label_forbody6); - LoadInst* float_tmp24 = new LoadInst(ptr_arrayidx23, "tmp24", false, label_forbody6); - InsertElementInst* packed_tmp26 = new InsertElementInst(packed_tmp15, float_tmp24, const_int32_31, "tmp26", label_forbody6); - std::vector ptr_arrayidx34_indices; - ptr_arrayidx34_indices.push_back(int32_i_0_reg2mem_0); - ptr_arrayidx34_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx34_indices.push_back(const_int32_33); - Instruction* ptr_arrayidx34 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx34_indices.begin(), ptr_arrayidx34_indices.end(), "arrayidx34", label_forbody6); - LoadInst* float_tmp35 = new LoadInst(ptr_arrayidx34, "tmp35", false, label_forbody6); - InsertElementInst* packed_tmp37 = new InsertElementInst(packed_tmp26, float_tmp35, const_int32_33, "tmp37", label_forbody6); - std::vector ptr_arrayidx45_indices; - ptr_arrayidx45_indices.push_back(int32_i_0_reg2mem_0); - ptr_arrayidx45_indices.push_back(int32_j_0_reg2mem_0); - ptr_arrayidx45_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx45 = new GetElementPtrInst(ptr_ainputs, ptr_arrayidx45_indices.begin(), ptr_arrayidx45_indices.end(), "arrayidx45", label_forbody6); - LoadInst* float_tmp46 = new LoadInst(ptr_arrayidx45, "tmp46", false, label_forbody6); - InsertElementInst* packed_tmp48 = new InsertElementInst(packed_tmp37, float_tmp46, const_int32_34, "tmp48", label_forbody6); - std::vector ptr_arrayidx54_indices; - ptr_arrayidx54_indices.push_back(int32_i_0_reg2mem_0); - ptr_arrayidx54_indices.push_back(int32_j_0_reg2mem_0); - Instruction* ptr_arrayidx54 = new GetElementPtrInst(ptr_res, ptr_arrayidx54_indices.begin(), ptr_arrayidx54_indices.end(), "arrayidx54", label_forbody6); - StoreInst* void_42 = new StoreInst(packed_tmp48, ptr_arrayidx54, false, label_forbody6); - BinaryOperator* int32_indvar_next = BinaryOperator::create(Instruction::Add, int32_j_0_reg2mem_0, const_int32_31, "indvar.next", label_forbody6); - ICmpInst* int1_exitcond = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next, int32_tmp19, "exitcond", label_forbody6); - new BranchInst(label_forinc57, label_forbody6, int1_exitcond, label_forbody6); - - // Block forinc57 (label_forinc57) - BinaryOperator* int32_indvar_next20 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0, const_int32_31, "indvar.next20", label_forinc57); - ICmpInst* int1_exitcond26 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next20, int32_tmp25, "exitcond26", label_forinc57); - new BranchInst(label_afterfor60, label_forcond2, int1_exitcond26, label_forinc57); - - // Block afterfor60 (label_afterfor60) - new ReturnInst(label_afterfor60); - - // Resolve Forward References - fwdref_39->replaceAllUsesWith(packed_tmp48); delete fwdref_39; - fwdref_41->replaceAllUsesWith(int32_indvar_next); delete fwdref_41; - fwdref_38->replaceAllUsesWith(int32_indvar_next20); delete fwdref_38; - - } - - // Function: from_consts (func_from_consts) - { - Function::arg_iterator args = func_from_consts->arg_begin(); - Value* ptr_res_46 = args++; - ptr_res_46->setName("res"); - Value* ptr_ainputs_47 = args++; - ptr_ainputs_47->setName("ainputs"); - Value* int32_count_48 = args++; - int32_count_48->setName("count"); - - BasicBlock* label_entry_49 = new BasicBlock("entry",func_from_consts,0); - BasicBlock* label_forbody_preheader = new BasicBlock("forbody.preheader",func_from_consts,0); - BasicBlock* label_forbody = new BasicBlock("forbody",func_from_consts,0); - BasicBlock* label_afterfor = new BasicBlock("afterfor",func_from_consts,0); - - // Block entry (label_entry_49) - ICmpInst* int1_cmp_50 = new ICmpInst(ICmpInst::ICMP_SGT, int32_count_48, const_int32_29, "cmp", label_entry_49); - new BranchInst(label_forbody_preheader, label_afterfor, int1_cmp_50, label_entry_49); - - // Block forbody.preheader (label_forbody_preheader) - BinaryOperator* int32_tmp_52 = BinaryOperator::create(Instruction::Add, int32_count_48, const_int32_30, "tmp", label_forbody_preheader); - ICmpInst* int1_tmp9 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_52, const_int32_29, "tmp9", label_forbody_preheader); - SelectInst* int32_tmp10 = new SelectInst(int1_tmp9, const_int32_31, int32_count_48, "tmp10", label_forbody_preheader); - new BranchInst(label_forbody, label_forbody_preheader); - - // Block forbody (label_forbody) - Argument* fwdref_55 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_54 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody); - int32_i_0_reg2mem_0_54->reserveOperandSpace(2); - int32_i_0_reg2mem_0_54->addIncoming(const_int32_29, label_forbody_preheader); - int32_i_0_reg2mem_0_54->addIncoming(fwdref_55, label_forbody); - - Argument* fwdref_57 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_56 = new PHINode(VectorTy_1, "vec.0.reg2mem.0", label_forbody); - packed_vec_0_reg2mem_0_56->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_56->addIncoming(const_packed_32, label_forbody_preheader); - packed_vec_0_reg2mem_0_56->addIncoming(fwdref_57, label_forbody); - - std::vector ptr_arraydecay_indices; - ptr_arraydecay_indices.push_back(int32_i_0_reg2mem_0_54); - ptr_arraydecay_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay = new GetElementPtrInst(ptr_ainputs_47, ptr_arraydecay_indices.begin(), ptr_arraydecay_indices.end(), "arraydecay", label_forbody); - LoadInst* float_tmp5 = new LoadInst(ptr_arraydecay, "tmp5", false, label_forbody); - InsertElementInst* packed_tmp7 = new InsertElementInst(packed_vec_0_reg2mem_0_56, float_tmp5, const_int32_29, "tmp7", label_forbody); - std::vector ptr_arrayidx12_indices; - ptr_arrayidx12_indices.push_back(int32_i_0_reg2mem_0_54); - ptr_arrayidx12_indices.push_back(const_int32_31); - Instruction* ptr_arrayidx12 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx12_indices.begin(), ptr_arrayidx12_indices.end(), "arrayidx12", label_forbody); - LoadInst* float_tmp13_58 = new LoadInst(ptr_arrayidx12, "tmp13", false, label_forbody); - InsertElementInst* packed_tmp15_59 = new InsertElementInst(packed_tmp7, float_tmp13_58, const_int32_31, "tmp15", label_forbody); - std::vector ptr_arrayidx20_indices; - ptr_arrayidx20_indices.push_back(int32_i_0_reg2mem_0_54); - ptr_arrayidx20_indices.push_back(const_int32_33); - Instruction* ptr_arrayidx20 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx20_indices.begin(), ptr_arrayidx20_indices.end(), "arrayidx20", label_forbody); - LoadInst* float_tmp21 = new LoadInst(ptr_arrayidx20, "tmp21", false, label_forbody); - InsertElementInst* packed_tmp23 = new InsertElementInst(packed_tmp15_59, float_tmp21, const_int32_33, "tmp23", label_forbody); - std::vector ptr_arrayidx28_indices; - ptr_arrayidx28_indices.push_back(int32_i_0_reg2mem_0_54); - ptr_arrayidx28_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx28 = new GetElementPtrInst(ptr_ainputs_47, ptr_arrayidx28_indices.begin(), ptr_arrayidx28_indices.end(), "arrayidx28", label_forbody); - LoadInst* float_tmp29 = new LoadInst(ptr_arrayidx28, "tmp29", false, label_forbody); - InsertElementInst* packed_tmp31 = new InsertElementInst(packed_tmp23, float_tmp29, const_int32_34, "tmp31", label_forbody); - GetElementPtrInst* ptr_arrayidx34_60 = new GetElementPtrInst(ptr_res_46, int32_i_0_reg2mem_0_54, "arrayidx34", label_forbody); - StoreInst* void_61 = new StoreInst(packed_tmp31, ptr_arrayidx34_60, false, label_forbody); - BinaryOperator* int32_indvar_next_62 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_54, const_int32_31, "indvar.next", label_forbody); - ICmpInst* int1_exitcond_63 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_62, int32_tmp10, "exitcond", label_forbody); - new BranchInst(label_afterfor, label_forbody, int1_exitcond_63, label_forbody); - - // Block afterfor (label_afterfor) - new ReturnInst(label_afterfor); - - // Resolve Forward References - fwdref_57->replaceAllUsesWith(packed_tmp31); delete fwdref_57; - fwdref_55->replaceAllUsesWith(int32_indvar_next_62); delete fwdref_55; - - } - - // Function: to_array (func_to_array) - { - Function::arg_iterator args = func_to_array->arg_begin(); - Value* ptr_dests = args++; - ptr_dests->setName("dests"); - Value* ptr_in = args++; - ptr_in->setName("in"); - Value* int32_num_attribs_66 = args++; - int32_num_attribs_66->setName("num_attribs"); - - BasicBlock* label_entry_67 = new BasicBlock("entry",func_to_array,0); - BasicBlock* label_forbody_preheader_68 = new BasicBlock("forbody.preheader",func_to_array,0); - BasicBlock* label_forbody_69 = new BasicBlock("forbody",func_to_array,0); - BasicBlock* label_afterfor_70 = new BasicBlock("afterfor",func_to_array,0); - - // Block entry (label_entry_67) - ICmpInst* int1_cmp_71 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_attribs_66, const_int32_29, "cmp", label_entry_67); - new BranchInst(label_forbody_preheader_68, label_afterfor_70, int1_cmp_71, label_entry_67); - - // Block forbody.preheader (label_forbody_preheader_68) - BinaryOperator* int32_tmp_73 = BinaryOperator::create(Instruction::Add, int32_num_attribs_66, const_int32_30, "tmp", label_forbody_preheader_68); - ICmpInst* int1_tmp9_74 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_73, const_int32_29, "tmp9", label_forbody_preheader_68); - SelectInst* int32_tmp10_75 = new SelectInst(int1_tmp9_74, const_int32_31, int32_num_attribs_66, "tmp10", label_forbody_preheader_68); - new BranchInst(label_forbody_69, label_forbody_preheader_68); - - // Block forbody (label_forbody_69) - Argument* fwdref_78 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_77 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0", label_forbody_69); - int32_i_0_reg2mem_0_77->reserveOperandSpace(2); - int32_i_0_reg2mem_0_77->addIncoming(const_int32_29, label_forbody_preheader_68); - int32_i_0_reg2mem_0_77->addIncoming(fwdref_78, label_forbody_69); - - std::vector ptr_arraydecay_79_indices; - ptr_arraydecay_79_indices.push_back(int32_i_0_reg2mem_0_77); - ptr_arraydecay_79_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay_79 = new GetElementPtrInst(ptr_dests, ptr_arraydecay_79_indices.begin(), ptr_arraydecay_79_indices.end(), "arraydecay", label_forbody_69); - GetElementPtrInst* ptr_arrayidx6 = new GetElementPtrInst(ptr_in, int32_i_0_reg2mem_0_77, "arrayidx6", label_forbody_69); - LoadInst* packed_tmp7_80 = new LoadInst(ptr_arrayidx6, "tmp7", false, label_forbody_69); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp7_80, const_int32_29, "tmp11", label_forbody_69); - StoreInst* void_81 = new StoreInst(float_tmp11, ptr_arraydecay_79, false, label_forbody_69); - std::vector ptr_arrayidx13_indices; - ptr_arrayidx13_indices.push_back(int32_i_0_reg2mem_0_77); - ptr_arrayidx13_indices.push_back(const_int32_31); - Instruction* ptr_arrayidx13 = new GetElementPtrInst(ptr_dests, ptr_arrayidx13_indices.begin(), ptr_arrayidx13_indices.end(), "arrayidx13", label_forbody_69); - ExtractElementInst* float_tmp15 = new ExtractElementInst(packed_tmp7_80, const_int32_31, "tmp15", label_forbody_69); - StoreInst* void_82 = new StoreInst(float_tmp15, ptr_arrayidx13, false, label_forbody_69); - std::vector ptr_arrayidx17_indices; - ptr_arrayidx17_indices.push_back(int32_i_0_reg2mem_0_77); - ptr_arrayidx17_indices.push_back(const_int32_33); - Instruction* ptr_arrayidx17 = new GetElementPtrInst(ptr_dests, ptr_arrayidx17_indices.begin(), ptr_arrayidx17_indices.end(), "arrayidx17", label_forbody_69); - ExtractElementInst* float_tmp19 = new ExtractElementInst(packed_tmp7_80, const_int32_33, "tmp19", label_forbody_69); - StoreInst* void_83 = new StoreInst(float_tmp19, ptr_arrayidx17, false, label_forbody_69); - std::vector ptr_arrayidx21_indices; - ptr_arrayidx21_indices.push_back(int32_i_0_reg2mem_0_77); - ptr_arrayidx21_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx21 = new GetElementPtrInst(ptr_dests, ptr_arrayidx21_indices.begin(), ptr_arrayidx21_indices.end(), "arrayidx21", label_forbody_69); - ExtractElementInst* float_tmp23 = new ExtractElementInst(packed_tmp7_80, const_int32_34, "tmp23", label_forbody_69); - StoreInst* void_84 = new StoreInst(float_tmp23, ptr_arrayidx21, false, label_forbody_69); - BinaryOperator* int32_indvar_next_85 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_77, const_int32_31, "indvar.next", label_forbody_69); - ICmpInst* int1_exitcond_86 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_85, int32_tmp10_75, "exitcond", label_forbody_69); - new BranchInst(label_afterfor_70, label_forbody_69, int1_exitcond_86, label_forbody_69); - - // Block afterfor (label_afterfor_70) - new ReturnInst(label_afterfor_70); - - // Resolve Forward References - fwdref_78->replaceAllUsesWith(int32_indvar_next_85); delete fwdref_78; - - } - - // Function: run_vertex_shader (func_run_vertex_shader) - { - Function::arg_iterator args = func_run_vertex_shader->arg_begin(); - Value* ptr_inputs = args++; - ptr_inputs->setName("inputs"); - Value* ptr_results = args++; - ptr_results->setName("results"); - Value* ptr_aconsts = args++; - ptr_aconsts->setName("aconsts"); - Value* int32_num_vertices = args++; - int32_num_vertices->setName("num_vertices"); - Value* int32_num_inputs = args++; - int32_num_inputs->setName("num_inputs"); - Value* int32_num_attribs_89 = args++; - int32_num_attribs_89->setName("num_attribs"); - Value* int32_num_consts = args++; - int32_num_consts->setName("num_consts"); - - BasicBlock* label_entry_90 = new BasicBlock("entry",func_run_vertex_shader,0); - BasicBlock* label_forbody_preheader_i = new BasicBlock("forbody.preheader.i",func_run_vertex_shader,0); - BasicBlock* label_forbody_i = new BasicBlock("forbody.i",func_run_vertex_shader,0); - BasicBlock* label_from_consts_exit = new BasicBlock("from_consts.exit",func_run_vertex_shader,0); - - // Block entry (label_entry_90) - AllocaInst* ptr_consts = new AllocaInst(ArrayTy_20, "consts", label_entry_90); - AllocaInst* ptr_temps = new AllocaInst(ArrayTy_22, "temps", label_entry_90); - AllocaInst* ptr_args = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_90); - std::vector ptr_tmp_indices; - ptr_tmp_indices.push_back(const_int32_29); - ptr_tmp_indices.push_back(const_int32_29); - Instruction* ptr_tmp = new GetElementPtrInst(ptr_args, ptr_tmp_indices.begin(), ptr_tmp_indices.end(), "tmp", label_entry_90); - CastInst* ptr_conv = new BitCastInst(ptr_results, PointerTy_0, "conv", label_entry_90); - StoreInst* void_91 = new StoreInst(ptr_conv, ptr_tmp, false, label_entry_90); - std::vector ptr_tmp2_indices; - ptr_tmp2_indices.push_back(const_int32_29); - ptr_tmp2_indices.push_back(const_int32_31); - Instruction* ptr_tmp2 = new GetElementPtrInst(ptr_args, ptr_tmp2_indices.begin(), ptr_tmp2_indices.end(), "tmp2", label_entry_90); - CastInst* ptr_conv4 = new BitCastInst(ptr_inputs, PointerTy_0, "conv4", label_entry_90); - StoreInst* void_92 = new StoreInst(ptr_conv4, ptr_tmp2, false, label_entry_90); - ICmpInst* int1_cmp_i = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts, const_int32_29, "cmp.i", label_entry_90); - new BranchInst(label_forbody_preheader_i, label_from_consts_exit, int1_cmp_i, label_entry_90); - - // Block forbody.preheader.i (label_forbody_preheader_i) - BinaryOperator* int32_tmp_i = BinaryOperator::create(Instruction::Add, int32_num_consts, const_int32_30, "tmp.i", label_forbody_preheader_i); - ICmpInst* int1_tmp9_i = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_i, const_int32_29, "tmp9.i", label_forbody_preheader_i); - SelectInst* int32_tmp10_i = new SelectInst(int1_tmp9_i, const_int32_31, int32_num_consts, "tmp10.i", label_forbody_preheader_i); - new BranchInst(label_forbody_i, label_forbody_preheader_i); - - // Block forbody.i (label_forbody_i) - Argument* fwdref_95 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_i = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i); - int32_i_0_reg2mem_0_i->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i->addIncoming(const_int32_29, label_forbody_preheader_i); - int32_i_0_reg2mem_0_i->addIncoming(fwdref_95, label_forbody_i); - - Argument* fwdref_96 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i); - packed_vec_0_reg2mem_0_i->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i->addIncoming(const_packed_32, label_forbody_preheader_i); - packed_vec_0_reg2mem_0_i->addIncoming(fwdref_96, label_forbody_i); - - std::vector ptr_arraydecay_i_indices; - ptr_arraydecay_i_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arraydecay_i_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay_i = new GetElementPtrInst(ptr_aconsts, ptr_arraydecay_i_indices.begin(), ptr_arraydecay_i_indices.end(), "arraydecay.i", label_forbody_i); - LoadInst* float_tmp5_i = new LoadInst(ptr_arraydecay_i, "tmp5.i", false, label_forbody_i); - InsertElementInst* packed_tmp7_i = new InsertElementInst(packed_vec_0_reg2mem_0_i, float_tmp5_i, const_int32_29, "tmp7.i", label_forbody_i); - std::vector ptr_arrayidx12_i_indices; - ptr_arrayidx12_i_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arrayidx12_i_indices.push_back(const_int32_31); - Instruction* ptr_arrayidx12_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx12_i_indices.begin(), ptr_arrayidx12_i_indices.end(), "arrayidx12.i", label_forbody_i); - LoadInst* float_tmp13_i = new LoadInst(ptr_arrayidx12_i, "tmp13.i", false, label_forbody_i); - InsertElementInst* packed_tmp15_i = new InsertElementInst(packed_tmp7_i, float_tmp13_i, const_int32_31, "tmp15.i", label_forbody_i); - std::vector ptr_arrayidx20_i_indices; - ptr_arrayidx20_i_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arrayidx20_i_indices.push_back(const_int32_33); - Instruction* ptr_arrayidx20_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx20_i_indices.begin(), ptr_arrayidx20_i_indices.end(), "arrayidx20.i", label_forbody_i); - LoadInst* float_tmp21_i = new LoadInst(ptr_arrayidx20_i, "tmp21.i", false, label_forbody_i); - InsertElementInst* packed_tmp23_i = new InsertElementInst(packed_tmp15_i, float_tmp21_i, const_int32_33, "tmp23.i", label_forbody_i); - std::vector ptr_arrayidx28_i_indices; - ptr_arrayidx28_i_indices.push_back(int32_i_0_reg2mem_0_i); - ptr_arrayidx28_i_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx28_i = new GetElementPtrInst(ptr_aconsts, ptr_arrayidx28_i_indices.begin(), ptr_arrayidx28_i_indices.end(), "arrayidx28.i", label_forbody_i); - LoadInst* float_tmp29_i = new LoadInst(ptr_arrayidx28_i, "tmp29.i", false, label_forbody_i); - InsertElementInst* packed_tmp31_i = new InsertElementInst(packed_tmp23_i, float_tmp29_i, const_int32_34, "tmp31.i", label_forbody_i); - std::vector ptr_arrayidx34_i_indices; - ptr_arrayidx34_i_indices.push_back(const_int32_29); - ptr_arrayidx34_i_indices.push_back(int32_i_0_reg2mem_0_i); - Instruction* ptr_arrayidx34_i = new GetElementPtrInst(ptr_consts, ptr_arrayidx34_i_indices.begin(), ptr_arrayidx34_i_indices.end(), "arrayidx34.i", label_forbody_i); - StoreInst* void_97 = new StoreInst(packed_tmp31_i, ptr_arrayidx34_i, false, label_forbody_i); - BinaryOperator* int32_indvar_next_98 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i, const_int32_31, "indvar.next", label_forbody_i); - ICmpInst* int1_exitcond_99 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next_98, int32_tmp10_i, "exitcond", label_forbody_i); - new BranchInst(label_from_consts_exit, label_forbody_i, int1_exitcond_99, label_forbody_i); - - // Block from_consts.exit (label_from_consts_exit) - std::vector ptr_tmp7_indices; - ptr_tmp7_indices.push_back(const_int32_29); - ptr_tmp7_indices.push_back(const_int32_34); - Instruction* ptr_tmp7 = new GetElementPtrInst(ptr_args, ptr_tmp7_indices.begin(), ptr_tmp7_indices.end(), "tmp7", label_from_consts_exit); - std::vector ptr_arraydecay8_indices; - ptr_arraydecay8_indices.push_back(const_int32_29); - ptr_arraydecay8_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay8 = new GetElementPtrInst(ptr_consts, ptr_arraydecay8_indices.begin(), ptr_arraydecay8_indices.end(), "arraydecay8", label_from_consts_exit); - StoreInst* void_101 = new StoreInst(ptr_arraydecay8, ptr_tmp7, false, label_from_consts_exit); - std::vector ptr_tmp9_indices; - ptr_tmp9_indices.push_back(const_int32_29); - ptr_tmp9_indices.push_back(const_int32_33); - Instruction* ptr_tmp9 = new GetElementPtrInst(ptr_args, ptr_tmp9_indices.begin(), ptr_tmp9_indices.end(), "tmp9", label_from_consts_exit); - std::vector ptr_arraydecay10_indices; - ptr_arraydecay10_indices.push_back(const_int32_29); - ptr_arraydecay10_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay10 = new GetElementPtrInst(ptr_temps, ptr_arraydecay10_indices.begin(), ptr_arraydecay10_indices.end(), "arraydecay10", label_from_consts_exit); - StoreInst* void_102 = new StoreInst(ptr_arraydecay10, ptr_tmp9, false, label_from_consts_exit); - CallInst* void_103 = new CallInst(func_execute_shader, ptr_args, "", label_from_consts_exit); - void_103->setCallingConv(CallingConv::C); - void_103->setTailCall(false);const ParamAttrsList *void_103_PAL = 0; - void_103->setParamAttrs(void_103_PAL); - - new ReturnInst(label_from_consts_exit); - - // Resolve Forward References - fwdref_96->replaceAllUsesWith(packed_tmp31_i); delete fwdref_96; - fwdref_95->replaceAllUsesWith(int32_indvar_next_98); delete fwdref_95; - - } - - // Function: run_fragment_shader (func_run_fragment_shader) - { - Function::arg_iterator args = func_run_fragment_shader->arg_begin(); - Value* float_x = args++; - float_x->setName("x"); - Value* float_y = args++; - float_y->setName("y"); - Value* ptr_results_105 = args++; - ptr_results_105->setName("results"); - Value* ptr_inputs_106 = args++; - ptr_inputs_106->setName("inputs"); - Value* int32_num_inputs_107 = args++; - int32_num_inputs_107->setName("num_inputs"); - Value* ptr_aconsts_108 = args++; - ptr_aconsts_108->setName("aconsts"); - Value* int32_num_consts_109 = args++; - int32_num_consts_109->setName("num_consts"); - Value* ptr_samplers = args++; - ptr_samplers->setName("samplers"); - - BasicBlock* label_entry_110 = new BasicBlock("entry",func_run_fragment_shader,0); - BasicBlock* label_forbody_preheader_i_111 = new BasicBlock("forbody.preheader.i",func_run_fragment_shader,0); - BasicBlock* label_forbody_i_112 = new BasicBlock("forbody.i",func_run_fragment_shader,0); - BasicBlock* label_from_consts_exit_113 = new BasicBlock("from_consts.exit",func_run_fragment_shader,0); - - // Block entry (label_entry_110) - AllocaInst* ptr_consts_114 = new AllocaInst(ArrayTy_20, "consts", label_entry_110); - AllocaInst* ptr_temps_115 = new AllocaInst(ArrayTy_22, "temps", label_entry_110); - AllocaInst* ptr_args_116 = new AllocaInst(StructTy_struct_ShaderInput, "args", label_entry_110); - std::vector ptr_tmp_117_indices; - ptr_tmp_117_indices.push_back(const_int32_29); - ptr_tmp_117_indices.push_back(const_int32_35); - Instruction* ptr_tmp_117 = new GetElementPtrInst(ptr_args_116, ptr_tmp_117_indices.begin(), ptr_tmp_117_indices.end(), "tmp", label_entry_110); - StoreInst* void_118 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_entry_110); - ICmpInst* int1_cmp_i_119 = new ICmpInst(ICmpInst::ICMP_SGT, int32_num_consts_109, const_int32_29, "cmp.i", label_entry_110); - new BranchInst(label_forbody_preheader_i_111, label_from_consts_exit_113, int1_cmp_i_119, label_entry_110); - - // Block forbody.preheader.i (label_forbody_preheader_i_111) - BinaryOperator* int32_tmp_i_121 = BinaryOperator::create(Instruction::Add, int32_num_consts_109, const_int32_30, "tmp.i", label_forbody_preheader_i_111); - ICmpInst* int1_tmp9_i_122 = new ICmpInst(ICmpInst::ICMP_SLT, int32_tmp_i_121, const_int32_29, "tmp9.i", label_forbody_preheader_i_111); - SelectInst* int32_tmp10_i_123 = new SelectInst(int1_tmp9_i_122, const_int32_31, int32_num_consts_109, "tmp10.i", label_forbody_preheader_i_111); - new BranchInst(label_forbody_i_112, label_forbody_preheader_i_111); - - // Block forbody.i (label_forbody_i_112) - Argument* fwdref_126 = new Argument(IntegerType::get(32)); - PHINode* int32_i_0_reg2mem_0_i_125 = new PHINode(IntegerType::get(32), "i.0.reg2mem.0.i", label_forbody_i_112); - int32_i_0_reg2mem_0_i_125->reserveOperandSpace(2); - int32_i_0_reg2mem_0_i_125->addIncoming(const_int32_29, label_forbody_preheader_i_111); - int32_i_0_reg2mem_0_i_125->addIncoming(fwdref_126, label_forbody_i_112); - - Argument* fwdref_128 = new Argument(VectorTy_1); - PHINode* packed_vec_0_reg2mem_0_i_127 = new PHINode(VectorTy_1, "vec.0.reg2mem.0.i", label_forbody_i_112); - packed_vec_0_reg2mem_0_i_127->reserveOperandSpace(2); - packed_vec_0_reg2mem_0_i_127->addIncoming(const_packed_32, label_forbody_preheader_i_111); - packed_vec_0_reg2mem_0_i_127->addIncoming(fwdref_128, label_forbody_i_112); - - std::vector ptr_arraydecay_i_129_indices; - ptr_arraydecay_i_129_indices.push_back(int32_i_0_reg2mem_0_i_125); - ptr_arraydecay_i_129_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay_i_129 = new GetElementPtrInst(ptr_aconsts_108, ptr_arraydecay_i_129_indices.begin(), ptr_arraydecay_i_129_indices.end(), "arraydecay.i", label_forbody_i_112); - LoadInst* float_tmp5_i_130 = new LoadInst(ptr_arraydecay_i_129, "tmp5.i", false, label_forbody_i_112); - InsertElementInst* packed_tmp7_i_131 = new InsertElementInst(packed_vec_0_reg2mem_0_i_127, float_tmp5_i_130, const_int32_29, "tmp7.i", label_forbody_i_112); - std::vector ptr_arrayidx12_i_132_indices; - ptr_arrayidx12_i_132_indices.push_back(int32_i_0_reg2mem_0_i_125); - ptr_arrayidx12_i_132_indices.push_back(const_int32_31); - Instruction* ptr_arrayidx12_i_132 = new GetElementPtrInst(ptr_aconsts_108, ptr_arrayidx12_i_132_indices.begin(), ptr_arrayidx12_i_132_indices.end(), "arrayidx12.i", label_forbody_i_112); - LoadInst* float_tmp13_i_133 = new LoadInst(ptr_arrayidx12_i_132, "tmp13.i", false, label_forbody_i_112); - InsertElementInst* packed_tmp15_i_134 = new InsertElementInst(packed_tmp7_i_131, float_tmp13_i_133, const_int32_31, "tmp15.i", label_forbody_i_112); - std::vector ptr_arrayidx20_i_135_indices; - ptr_arrayidx20_i_135_indices.push_back(int32_i_0_reg2mem_0_i_125); - ptr_arrayidx20_i_135_indices.push_back(const_int32_33); - Instruction* ptr_arrayidx20_i_135 = new GetElementPtrInst(ptr_aconsts_108, ptr_arrayidx20_i_135_indices.begin(), ptr_arrayidx20_i_135_indices.end(), "arrayidx20.i", label_forbody_i_112); - LoadInst* float_tmp21_i_136 = new LoadInst(ptr_arrayidx20_i_135, "tmp21.i", false, label_forbody_i_112); - InsertElementInst* packed_tmp23_i_137 = new InsertElementInst(packed_tmp15_i_134, float_tmp21_i_136, const_int32_33, "tmp23.i", label_forbody_i_112); - std::vector ptr_arrayidx28_i_138_indices; - ptr_arrayidx28_i_138_indices.push_back(int32_i_0_reg2mem_0_i_125); - ptr_arrayidx28_i_138_indices.push_back(const_int32_34); - Instruction* ptr_arrayidx28_i_138 = new GetElementPtrInst(ptr_aconsts_108, ptr_arrayidx28_i_138_indices.begin(), ptr_arrayidx28_i_138_indices.end(), "arrayidx28.i", label_forbody_i_112); - LoadInst* float_tmp29_i_139 = new LoadInst(ptr_arrayidx28_i_138, "tmp29.i", false, label_forbody_i_112); - InsertElementInst* packed_tmp31_i_140 = new InsertElementInst(packed_tmp23_i_137, float_tmp29_i_139, const_int32_34, "tmp31.i", label_forbody_i_112); - std::vector ptr_arrayidx34_i_141_indices; - ptr_arrayidx34_i_141_indices.push_back(const_int32_29); - ptr_arrayidx34_i_141_indices.push_back(int32_i_0_reg2mem_0_i_125); - Instruction* ptr_arrayidx34_i_141 = new GetElementPtrInst(ptr_consts_114, ptr_arrayidx34_i_141_indices.begin(), ptr_arrayidx34_i_141_indices.end(), "arrayidx34.i", label_forbody_i_112); - StoreInst* void_142 = new StoreInst(packed_tmp31_i_140, ptr_arrayidx34_i_141, false, label_forbody_i_112); - BinaryOperator* int32_indvar_next7 = BinaryOperator::create(Instruction::Add, int32_i_0_reg2mem_0_i_125, const_int32_31, "indvar.next7", label_forbody_i_112); - ICmpInst* int1_exitcond8 = new ICmpInst(ICmpInst::ICMP_EQ, int32_indvar_next7, int32_tmp10_i_123, "exitcond8", label_forbody_i_112); - new BranchInst(label_from_consts_exit_113, label_forbody_i_112, int1_exitcond8, label_forbody_i_112); - - // Block from_consts.exit (label_from_consts_exit_113) - std::vector ptr_tmp3_indices; - ptr_tmp3_indices.push_back(const_int32_29); - ptr_tmp3_indices.push_back(const_int32_34); - Instruction* ptr_tmp3 = new GetElementPtrInst(ptr_args_116, ptr_tmp3_indices.begin(), ptr_tmp3_indices.end(), "tmp3", label_from_consts_exit_113); - std::vector ptr_arraydecay4_indices; - ptr_arraydecay4_indices.push_back(const_int32_29); - ptr_arraydecay4_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay4 = new GetElementPtrInst(ptr_consts_114, ptr_arraydecay4_indices.begin(), ptr_arraydecay4_indices.end(), "arraydecay4", label_from_consts_exit_113); - StoreInst* void_144 = new StoreInst(ptr_arraydecay4, ptr_tmp3, false, label_from_consts_exit_113); - std::vector ptr_tmp5_indices; - ptr_tmp5_indices.push_back(const_int32_29); - ptr_tmp5_indices.push_back(const_int32_33); - Instruction* ptr_tmp5 = new GetElementPtrInst(ptr_args_116, ptr_tmp5_indices.begin(), ptr_tmp5_indices.end(), "tmp5", label_from_consts_exit_113); - std::vector ptr_arraydecay6_indices; - ptr_arraydecay6_indices.push_back(const_int32_29); - ptr_arraydecay6_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay6 = new GetElementPtrInst(ptr_temps_115, ptr_arraydecay6_indices.begin(), ptr_arraydecay6_indices.end(), "arraydecay6", label_from_consts_exit_113); - StoreInst* void_145 = new StoreInst(ptr_arraydecay6, ptr_tmp5, false, label_from_consts_exit_113); - std::vector ptr_tmp8_indices; - ptr_tmp8_indices.push_back(const_int32_29); - ptr_tmp8_indices.push_back(const_int32_31); - Instruction* ptr_tmp8 = new GetElementPtrInst(ptr_args_116, ptr_tmp8_indices.begin(), ptr_tmp8_indices.end(), "tmp8", label_from_consts_exit_113); - std::vector ptr_tmp12_indices; - ptr_tmp12_indices.push_back(const_int32_29); - ptr_tmp12_indices.push_back(const_int32_29); - Instruction* ptr_tmp12 = new GetElementPtrInst(ptr_args_116, ptr_tmp12_indices.begin(), ptr_tmp12_indices.end(), "tmp12", label_from_consts_exit_113); - std::vector ptr_arraydecay11_146_indices; - ptr_arraydecay11_146_indices.push_back(const_int32_29); - ptr_arraydecay11_146_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11_146 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_146_indices.begin(), ptr_arraydecay11_146_indices.end(), "arraydecay11", label_from_consts_exit_113); - StoreInst* void_147 = new StoreInst(ptr_arraydecay11_146, ptr_tmp8, false, label_from_consts_exit_113); - std::vector ptr_arraydecay16_indices; - ptr_arraydecay16_indices.push_back(const_int32_29); - ptr_arraydecay16_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay16 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_indices.begin(), ptr_arraydecay16_indices.end(), "arraydecay16", label_from_consts_exit_113); - StoreInst* void_148 = new StoreInst(ptr_arraydecay16, ptr_tmp12, false, label_from_consts_exit_113); - StoreInst* void_149 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113); - CallInst* void_150 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113); - void_150->setCallingConv(CallingConv::C); - void_150->setTailCall(false);const ParamAttrsList *void_150_PAL = 0; - void_150->setParamAttrs(void_150_PAL); - - LoadInst* int32_tmp23 = new LoadInst(ptr_tmp_117, "tmp23", false, label_from_consts_exit_113); - std::vector ptr_arraydecay11_1_indices; - ptr_arraydecay11_1_indices.push_back(const_int32_31); - ptr_arraydecay11_1_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11_1 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_1_indices.begin(), ptr_arraydecay11_1_indices.end(), "arraydecay11.1", label_from_consts_exit_113); - StoreInst* void_151 = new StoreInst(ptr_arraydecay11_1, ptr_tmp8, false, label_from_consts_exit_113); - std::vector ptr_arraydecay16_1_indices; - ptr_arraydecay16_1_indices.push_back(const_int32_31); - ptr_arraydecay16_1_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay16_1 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_1_indices.begin(), ptr_arraydecay16_1_indices.end(), "arraydecay16.1", label_from_consts_exit_113); - StoreInst* void_152 = new StoreInst(ptr_arraydecay16_1, ptr_tmp12, false, label_from_consts_exit_113); - StoreInst* void_153 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113); - CallInst* void_154 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113); - void_154->setCallingConv(CallingConv::C); - void_154->setTailCall(false);const ParamAttrsList *void_154_PAL = 0; - void_154->setParamAttrs(void_154_PAL); - - LoadInst* int32_tmp23_1 = new LoadInst(ptr_tmp_117, "tmp23.1", false, label_from_consts_exit_113); - BinaryOperator* int32_shl_1 = BinaryOperator::create(Instruction::Shl, int32_tmp23_1, const_int32_31, "shl.1", label_from_consts_exit_113); - BinaryOperator* int32_or_1 = BinaryOperator::create(Instruction::Or, int32_shl_1, int32_tmp23, "or.1", label_from_consts_exit_113); - std::vector ptr_arraydecay11_2_indices; - ptr_arraydecay11_2_indices.push_back(const_int32_33); - ptr_arraydecay11_2_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11_2 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_2_indices.begin(), ptr_arraydecay11_2_indices.end(), "arraydecay11.2", label_from_consts_exit_113); - StoreInst* void_155 = new StoreInst(ptr_arraydecay11_2, ptr_tmp8, false, label_from_consts_exit_113); - std::vector ptr_arraydecay16_2_indices; - ptr_arraydecay16_2_indices.push_back(const_int32_33); - ptr_arraydecay16_2_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay16_2 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_2_indices.begin(), ptr_arraydecay16_2_indices.end(), "arraydecay16.2", label_from_consts_exit_113); - StoreInst* void_156 = new StoreInst(ptr_arraydecay16_2, ptr_tmp12, false, label_from_consts_exit_113); - StoreInst* void_157 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113); - CallInst* void_158 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113); - void_158->setCallingConv(CallingConv::C); - void_158->setTailCall(false);const ParamAttrsList *void_158_PAL = 0; - void_158->setParamAttrs(void_158_PAL); - - LoadInst* int32_tmp23_2 = new LoadInst(ptr_tmp_117, "tmp23.2", false, label_from_consts_exit_113); - BinaryOperator* int32_shl_2 = BinaryOperator::create(Instruction::Shl, int32_tmp23_2, const_int32_33, "shl.2", label_from_consts_exit_113); - BinaryOperator* int32_or_2 = BinaryOperator::create(Instruction::Or, int32_shl_2, int32_or_1, "or.2", label_from_consts_exit_113); - std::vector ptr_arraydecay11_3_indices; - ptr_arraydecay11_3_indices.push_back(const_int32_34); - ptr_arraydecay11_3_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay11_3 = new GetElementPtrInst(ptr_inputs_106, ptr_arraydecay11_3_indices.begin(), ptr_arraydecay11_3_indices.end(), "arraydecay11.3", label_from_consts_exit_113); - StoreInst* void_159 = new StoreInst(ptr_arraydecay11_3, ptr_tmp8, false, label_from_consts_exit_113); - std::vector ptr_arraydecay16_3_indices; - ptr_arraydecay16_3_indices.push_back(const_int32_34); - ptr_arraydecay16_3_indices.push_back(const_int32_29); - Instruction* ptr_arraydecay16_3 = new GetElementPtrInst(ptr_results_105, ptr_arraydecay16_3_indices.begin(), ptr_arraydecay16_3_indices.end(), "arraydecay16.3", label_from_consts_exit_113); - StoreInst* void_160 = new StoreInst(ptr_arraydecay16_3, ptr_tmp12, false, label_from_consts_exit_113); - StoreInst* void_161 = new StoreInst(const_int32_29, ptr_tmp_117, false, label_from_consts_exit_113); - CallInst* void_162 = new CallInst(func_execute_shader, ptr_args_116, "", label_from_consts_exit_113); - void_162->setCallingConv(CallingConv::C); - void_162->setTailCall(false);const ParamAttrsList *void_162_PAL = 0; - void_162->setParamAttrs(void_162_PAL); - - LoadInst* int32_tmp23_3 = new LoadInst(ptr_tmp_117, "tmp23.3", false, label_from_consts_exit_113); - BinaryOperator* int32_shl_3 = BinaryOperator::create(Instruction::Shl, int32_tmp23_3, const_int32_34, "shl.3", label_from_consts_exit_113); - BinaryOperator* int32_or_3 = BinaryOperator::create(Instruction::Or, int32_shl_3, int32_or_2, "or.3", label_from_consts_exit_113); - BinaryOperator* int32_neg = BinaryOperator::create(Instruction::Xor, int32_or_3, const_int32_30, "neg", label_from_consts_exit_113); - new ReturnInst(int32_neg, label_from_consts_exit_113); - - // Resolve Forward References - fwdref_128->replaceAllUsesWith(packed_tmp31_i_140); delete fwdref_128; - fwdref_126->replaceAllUsesWith(int32_indvar_next7); delete fwdref_126; - - } - - return mod; -} diff --git a/src/mesa/pipe/llvm/llvm_entry.c b/src/mesa/pipe/llvm/llvm_entry.c deleted file mode 100644 index fa50b60e66..0000000000 --- a/src/mesa/pipe/llvm/llvm_entry.c +++ /dev/null @@ -1,163 +0,0 @@ -/* clang --emit-llvm llvm_entry.c |llvm-as |opt -std-compile-opts |llvm2cpp -for=Shader -gen-module -funcname=createBaseShader */ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -/* clang --emit-llvm llvm_builtins.c |llvm-as |opt -std-compile-opts |llvm-dis */ -typedef __attribute__(( ocu_vector_type(4) )) float float4; - -void from_array(float4 (*res)[16], float (*ainputs)[16][4], - int count, int num_attribs) -{ - for (int i = 0; i < count; ++i) { - for (int j = 0; j < num_attribs; ++j) { - float4 vec; - vec.x = ainputs[i][j][0]; - vec.y = ainputs[i][j][1]; - vec.z = ainputs[i][j][2]; - vec.w = ainputs[i][j][3]; - res[i][j] = vec; - } - } -} - -void from_consts(float4 *res, float (*ainputs)[4], - int count) -{ - for (int i = 0; i < count; ++i) { - float4 vec; - vec.x = ainputs[i][0]; - vec.y = ainputs[i][1]; - vec.z = ainputs[i][2]; - vec.w = ainputs[i][3]; - res[i] = vec; - } -} - -void to_array(float (*dests)[4], float4 *in, int num_attribs) -{ - for (int i = 0; i < num_attribs; ++i) { - float *rd = dests[i]; - float4 ri = in[i]; - rd[0] = ri.x; - rd[1] = ri.y; - rd[2] = ri.z; - rd[3] = ri.w; - } -} - - -struct ShaderInput -{ - float4 *dests; - float4 *inputs; - float4 *temps; - float4 *consts; - int kilmask; -}; - -extern void execute_shader(struct ShaderInput *input); - -void run_vertex_shader(void *inputs, - void *results, - float (*aconsts)[4], - int num_vertices, - int num_inputs, - int num_attribs, - int num_consts) -{ - float4 consts[32]; - float4 temps[128];//MAX_PROGRAM_TEMPS - - struct ShaderInput args; - args.dests = results; - args.inputs = inputs; - - /*printf("XXX LLVM run_vertex_shader vertices = %d, inputs = %d, attribs = %d, consts = %d\n", - num_vertices, num_inputs, num_attribs, num_consts);*/ - from_consts(consts, aconsts, num_consts); - args.consts = consts; - args.temps = temps; - - execute_shader(&args); -} - - -struct pipe_sampler_state; -struct softpipe_tile_cache; - -#define NUM_CHANNELS 4 /* R,G,B,A */ -#define QUAD_SIZE 4 /* 4 pixel/quad */ - -struct tgsi_sampler -{ - const struct pipe_sampler_state *state; - /** Get samples for four fragments in a quad */ - void (*get_samples)(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]); - void *pipe; /*XXX temporary*/ - struct softpipe_tile_cache *cache; -}; - - -int run_fragment_shader(float x, float y, - float4 (*results)[16], - float4 (*inputs)[16], - int num_inputs, - float (*aconsts)[4], - int num_consts, - struct tgsi_sampler *samplers) -{ - float4 consts[32]; - float4 temps[128];//MAX_PROGRAM_TEMPS - struct ShaderInput args; - int mask = 0; - args.kilmask = 0; - - from_consts(consts, aconsts, num_consts); - args.consts = consts; - args.temps = temps; - //printf("AAAAAAAAAAAAAAAAAAAAAAA FRAGMENT SHADER %f %f\n", x, y); - for (int i = 0; i < 4; ++i) { - args.inputs = inputs[i]; - args.dests = results[i]; - mask = args.kilmask; - args.kilmask = 0; - execute_shader(&args); - args.kilmask = mask | (args.kilmask << i); - } - return ~args.kilmask; -} - diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index 3497eebcd9..20fce9c9cc 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -44,7 +44,6 @@ #include using namespace llvm; -#include "llvm_base_shader.cpp" static inline FunctionType *vertexShaderFunctionType() { @@ -1074,7 +1073,7 @@ translate_instructionir(llvm::Module *module, llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) { - llvm::Module *mod = createBaseShader(); + llvm::Module *mod = new Module("shader"); struct tgsi_parse_context parse; struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; -- cgit v1.2.3 From 2e75f39bc4286cebb7330f54d7acf5b8f1d9777f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 09:26:56 -0500 Subject: make swizzling on incoming arguments work --- src/mesa/pipe/llvm/storagesoa.cpp | 54 ++++++++++++++++++++++++++++++--------- src/mesa/pipe/llvm/storagesoa.h | 21 ++++++++++----- src/mesa/pipe/llvm/tgsitollvm.cpp | 15 +++++++---- 3 files changed, 67 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index 7b758b1665..a65b5c14d9 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -27,6 +27,7 @@ #include "storagesoa.h" +#include "gallivm_p.h" #include "pipe/p_shader_tokens.h" #include @@ -87,8 +88,7 @@ llvm::Value *StorageSoa::addrElement(int idx) const return 0; } -std::vector StorageSoa::inputElement(int idx, int swizzle, - llvm::Value *indIdx) +std::vector StorageSoa::inputElement(int idx, llvm::Value *indIdx) { std::vector res(4); @@ -100,8 +100,7 @@ std::vector StorageSoa::inputElement(int idx, int swizzle, return res; } -std::vector StorageSoa::constElement(int idx, int swizzle, - llvm::Value *indIdx) +std::vector StorageSoa::constElement(int idx, llvm::Value *indIdx) { std::vector res(4); llvm::Value *xChannel = elementPointer(m_consts, idx, 0); @@ -117,8 +116,7 @@ std::vector StorageSoa::constElement(int idx, int swizzle, return res; } -std::vector StorageSoa::outputElement(int idx, int swizzle, - llvm::Value *indIdx) +std::vector StorageSoa::outputElement(int idx, llvm::Value *indIdx) { std::vector res(4); @@ -130,8 +128,7 @@ std::vector StorageSoa::outputElement(int idx, int swizzle, return res; } -std::vector StorageSoa::tempElement(int idx, int swizzle, - llvm::Value *indIdx) +std::vector StorageSoa::tempElement(int idx, llvm::Value *indIdx) { std::vector res(4); @@ -143,15 +140,15 @@ std::vector StorageSoa::tempElement(int idx, int swizzle, return res; } -std::vector StorageSoa::immediateElement(int idx, int swizzle) +std::vector StorageSoa::immediateElement(int idx) { std::vector res(4); res = m_immediates[idx]; res[0] = new LoadInst(res[0], name("immx"), false, m_block); - res[1] = new LoadInst(res[1], name("immx"), false, m_block); - res[2] = new LoadInst(res[2], name("immx"), false, m_block); - res[3] = new LoadInst(res[3], name("immx"), false, m_block); + res[1] = new LoadInst(res[1], name("immy"), false, m_block); + res[2] = new LoadInst(res[2], name("immz"), false, m_block); + res[3] = new LoadInst(res[3], name("immw"), false, m_block); return res; } @@ -291,3 +288,36 @@ llvm::Value * StorageSoa::createConstGlobalVector(float *vec) return immediate; } + +std::vector StorageSoa::argument(Argument type, int idx, int swizzle, + llvm::Value *indIdx ) +{ + std::vector val(4); + switch(type) { + case Input: + val = inputElement(idx, indIdx); + break; + case Output: + val = outputElement(idx, indIdx); + break; + case Temp: + val = tempElement(idx, indIdx); + break; + case Const: + val = constElement(idx, indIdx); + break; + case Immediate: + val = immediateElement(idx); + break; + } + if (!gallivm_is_swizzle(swizzle)) + return val; + + std::vector res(4); + + res[0] = val[gallivm_x_swizzle(swizzle)]; + res[1] = val[gallivm_y_swizzle(swizzle)]; + res[2] = val[gallivm_z_swizzle(swizzle)]; + res[3] = val[gallivm_w_swizzle(swizzle)]; + return res; +} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index 2d07e836f4..9443234c82 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -43,6 +43,14 @@ namespace llvm { class StorageSoa { +public: + enum Argument { + Input, + Output, + Temp, + Const, + Immediate + }; public: StorageSoa(llvm::BasicBlock *block, llvm::Value *input, @@ -50,16 +58,12 @@ public: llvm::Value *consts, llvm::Value *temps); + std::vector argument(Argument type, int idx, int swizzle, + llvm::Value *indIdx =0); void addImmediate(float *vec); llvm::Value * addrElement(int idx) const; - std::vector inputElement(int idx, int swizzle, llvm::Value *indIdx =0); - std::vector constElement(int idx, int swizzle, llvm::Value *indIdx =0); - std::vector outputElement(int idx, int swizzle, llvm::Value *indIdx =0); - std::vector tempElement(int idx, int swizzle, llvm::Value *indIdx =0); - std::vector immediateElement(int idx, int swizzle); - llvm::Value *extractIndex(llvm::Value *vec); void storeOutput(int dstIdx, const std::vector &val, @@ -79,6 +83,11 @@ private: llvm::Module *currentModule() const; llvm::Value *createConstGlobalVector(float *vec); + std::vector inputElement(int idx, llvm::Value *indIdx =0); + std::vector constElement(int idx, llvm::Value *indIdx =0); + std::vector outputElement(int idx, llvm::Value *indIdx =0); + std::vector tempElement(int idx, llvm::Value *indIdx =0); + std::vector immediateElement(int idx); private: llvm::BasicBlock *m_block; diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index 20fce9c9cc..10c417996a 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -706,15 +706,20 @@ translate_instructionir(llvm::Module *module, indIdx = storage->extractIndex(indIdx); } if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->constElement(src->SrcRegister.Index, swizzle, indIdx); + val = storage->argument(StorageSoa::Const, + src->SrcRegister.Index, swizzle, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->inputElement(src->SrcRegister.Index, swizzle, indIdx); + val = storage->argument(StorageSoa::Input, + src->SrcRegister.Index, swizzle, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->tempElement(src->SrcRegister.Index, swizzle); + val = storage->argument(StorageSoa::Temp, + src->SrcRegister.Index, swizzle); } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->outputElement(src->SrcRegister.Index, swizzle, indIdx); + val = storage->argument(StorageSoa::Output, + src->SrcRegister.Index, swizzle, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->immediateElement(src->SrcRegister.Index, swizzle); + val = storage->argument(StorageSoa::Immediate, + src->SrcRegister.Index, swizzle); } else { fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); return; -- cgit v1.2.3 From a856b399e6a46f2026006402bc6b9125bd23f9a9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 13 Feb 2008 18:29:29 +0000 Subject: x86: fix assignment in assert typo --- src/mesa/x86/rtasm/x86sse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index 1111e8db99..39c0e9b851 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -820,7 +820,7 @@ static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86 assert(0); } else if (dst.idx == 0) { - assert(arg.file = file_REG32); + assert(arg.file == file_REG32); emit_1ub(p, 0xd8); emit_modrm_noreg(p, argmem_noreg, arg); } -- cgit v1.2.3 From f3f7ff257370ff72dbc7a0ba05ed0a99ce67ebac Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 14 Feb 2008 12:14:46 +0000 Subject: tgsi: partially unbreak sse fragment shaders. more to do. --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 8 -------- 1 file changed, 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 1e56e4afb6..1d9e9a14ce 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -2169,14 +2169,6 @@ emit_declaration( last = decl->u.DeclarationRange.Last; mask = decl->Declaration.UsageMask; - /* Do not touch WPOS.xy */ - if( first == 0 ) { - mask &= ~TGSI_WRITEMASK_XY; - if( mask == TGSI_WRITEMASK_NONE ) { - first++; - } - } - for( i = first; i <= last; i++ ) { for( j = 0; j < NUM_CHANNELS; j++ ) { if( mask & (1 << j) ) { -- cgit v1.2.3 From 18e94e15c0ca881309ac5784075a04160cc1eeb6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 14 Feb 2008 13:36:21 +0000 Subject: gallium: Minor cleanups to bitfield sizes, etc. --- src/mesa/pipe/p_state.h | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h index 4d3a6b2f41..1082343e2f 100644 --- a/src/mesa/pipe/p_state.h +++ b/src/mesa/pipe/p_state.h @@ -253,7 +253,8 @@ struct pipe_surface unsigned status; /**< PIPE_SURFACE_STATUS_x */ unsigned clear_value; /**< may be temporary */ unsigned cpp; /**< bytes per pixel */ - unsigned width, height; + unsigned width; + unsigned height; unsigned pitch; /**< in pixels */ unsigned offset; /**< offset from start of buffer, in bytes */ unsigned refcount; @@ -272,13 +273,12 @@ struct pipe_texture enum pipe_texture_target target; /**< PIPE_TEXTURE_x */ enum pipe_format format; /**< PIPE_FORMAT_x */ - unsigned last_level; /**< Index of last mipmap level present/defined */ - unsigned width[PIPE_MAX_TEXTURE_LEVELS]; unsigned height[PIPE_MAX_TEXTURE_LEVELS]; unsigned depth[PIPE_MAX_TEXTURE_LEVELS]; - unsigned cpp; + unsigned cpp:8; + unsigned last_level:8; /**< Index of last mipmap level present/defined */ unsigned compressed:1; /* These are also refcounted: @@ -294,7 +294,7 @@ struct pipe_texture */ struct pipe_vertex_buffer { - unsigned pitch:11; /**< stride to same attrib in next vertex, in bytes */ + unsigned pitch; /**< stride to same attrib in next vertex, in bytes */ unsigned max_index; /**< number of vertices in this buffer */ unsigned buffer_offset; /**< offset to start of data in buffer, in bytes */ struct pipe_buffer *buffer; /**< the actual buffer */ @@ -307,13 +307,13 @@ struct pipe_vertex_buffer struct pipe_vertex_element { /** Offset of this attribute, in bytes, from the start of the vertex */ - unsigned src_offset:11; + unsigned src_offset; /** Which vertex_buffer (as given to pipe->set_vertex_buffer()) does * this attribute live in? */ - unsigned vertex_buffer_index:5; - unsigned nr_components:3; + unsigned vertex_buffer_index:8; + unsigned nr_components:8; enum pipe_format src_format; /**< PIPE_FORMAT_* */ }; -- cgit v1.2.3 From cf5ef20f436ac1d3efde2b7bff698decccb029e3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 14 Feb 2008 16:53:51 +0000 Subject: gallium: Cleanups related to clears. --- src/mesa/state_tracker/st_cb_clear.c | 117 +++++++++++++++++++++++++---------- 1 file changed, 85 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 410062e1e8..78baf772f4 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -272,6 +272,14 @@ clear_with_quad(GLcontext *ctx, const GLfloat x1 = ctx->DrawBuffer->_Xmax; const GLfloat y1 = ctx->DrawBuffer->_Ymax; + /* + printf("%s %s%s%s %f,%f %f,%f\n", __FUNCTION__, + color ? "color, " : "", + depth ? "depth, " : "", + stencil ? "stencil" : "", + x0, y0, + x1, y1); + */ /* blend state: RGBA masking */ { @@ -389,13 +397,44 @@ clear_with_quad(GLcontext *ctx, * Determine if we need to clear the depth buffer by drawing a quad. */ static INLINE GLboolean -check_clear_color_with_quad(GLcontext *ctx) +check_clear_color_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + const struct st_renderbuffer *strb = st_renderbuffer(rb); + + if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED) + return FALSE; + + if (ctx->Scissor.Enabled) + return TRUE; + + if (!ctx->Color.ColorMask[0] || + !ctx->Color.ColorMask[1] || + !ctx->Color.ColorMask[2] || + !ctx->Color.ColorMask[3]) + return TRUE; + + return FALSE; +} + + +static INLINE GLboolean +check_clear_depth_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) { - return !(ctx->Color.ColorMask[0] && - ctx->Color.ColorMask[1] && - ctx->Color.ColorMask[2] && - ctx->Color.ColorMask[3] && - !ctx->Scissor.Enabled); + const struct st_renderbuffer *strb = st_renderbuffer(rb); + const GLuint stencilMax = (1 << rb->StencilBits) - 1; + GLboolean maskStencil + = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax; + + if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED) + return FALSE; + + if (ctx->Scissor.Enabled) + return TRUE; + + if (maskStencil) + return TRUE; + + return FALSE; } @@ -407,10 +446,19 @@ check_clear_depth_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) { const struct st_renderbuffer *strb = st_renderbuffer(rb); const GLboolean isDS = is_depth_stencil_format(strb->surface->format); - return ctx->Scissor.Enabled - || (isDS && - strb->surface->status == PIPE_SURFACE_STATUS_DEFINED && - ctx->DrawBuffer->Visual.stencilBits > 0); + + if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED) + return FALSE; + + if (ctx->Scissor.Enabled) + return TRUE; + + if (isDS && + strb->surface->status == PIPE_SURFACE_STATUS_DEFINED && + ctx->DrawBuffer->Visual.stencilBits > 0) + return TRUE; + + return FALSE; } @@ -425,9 +473,27 @@ check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) const GLuint stencilMax = (1 << rb->StencilBits) - 1; const GLboolean maskStencil = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax; - return maskStencil - || ctx->Scissor.Enabled - || (isDS && ctx->DrawBuffer->Visual.depthBits > 0); + + if (strb->surface->status == PIPE_SURFACE_STATUS_UNDEFINED) + return FALSE; + + if (maskStencil) + return TRUE; + + if (ctx->Scissor.Enabled) + return TRUE; + + /* This is correct, but it is necessary to look at the depth clear + * value held in the surface when it comes time to issue the clear, + * rather than taking depth and stencil clear values from the + * current state. + */ + if (isDS && + strb->surface->status == PIPE_SURFACE_STATUS_DEFINED && + ctx->DrawBuffer->Visual.depthBits > 0) + return TRUE; + + return FALSE; } @@ -436,14 +502,10 @@ check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) static void clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { - struct st_renderbuffer *strb = st_renderbuffer(rb); - - if (ctx->Color.ColorMask[0] && - ctx->Color.ColorMask[1] && - ctx->Color.ColorMask[2] && - ctx->Color.ColorMask[3] && - !ctx->Scissor.Enabled) + if (!check_clear_color_with_quad( ctx, rb )) { + struct st_renderbuffer *strb = st_renderbuffer(rb); + /* clear whole buffer w/out masking */ GLuint clearValue = color_value(strb->surface->format, ctx->Color.ClearColor); @@ -482,14 +544,8 @@ static void clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { struct st_renderbuffer *strb = st_renderbuffer(rb); - const GLboolean isDS = is_depth_stencil_format(strb->surface->format); - const GLuint stencilMax = (1 << rb->StencilBits) - 1; - GLboolean maskStencil - = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax; - if (maskStencil || - ctx->Scissor.Enabled || - (isDS && ctx->DrawBuffer->Visual.depthBits > 0)) { + if (check_clear_stencil_with_quad(ctx, rb)) { /* masking or scissoring or combined depth/stencil buffer */ clear_with_quad(ctx, GL_FALSE, GL_FALSE, GL_TRUE); } @@ -505,13 +561,10 @@ static void clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { struct st_renderbuffer *strb = st_renderbuffer(rb); - const GLuint stencilMax = (1 << rb->StencilBits) - 1; - GLboolean maskStencil - = (ctx->Stencil.WriteMask[0] & stencilMax) != stencilMax; assert(is_depth_stencil_format(strb->surface->format)); - if (!maskStencil && !ctx->Scissor.Enabled) { + if (check_clear_depth_stencil_with_quad(ctx, rb)) { /* clear whole buffer w/out masking */ GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); @@ -520,7 +573,7 @@ clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) clearValue |= ctx->Stencil.Clear << 24; break; case PIPE_FORMAT_Z24S8_UNORM: - clearValue |= clearValue | ctx->Stencil.Clear; + clearValue |= ctx->Stencil.Clear; break; default: assert(0); -- cgit v1.2.3 From 0230c56ed0db16f007e4d1881554c7dbfa3ac3de Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 14 Feb 2008 17:42:52 +0000 Subject: intel_winsys: Call st_notify_swapbuffers_complete() after buffer swap. --- src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c index 454cd71f6c..56b86d6a63 100644 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c @@ -220,6 +220,7 @@ intelSwapBuffers(__DRIdrawablePrivate * dPriv) if (back_surf) { st_notify_swapbuffers(intel_fb->stfb); intelDisplaySurface(dPriv, back_surf, NULL); + st_notify_swapbuffers_complete(intel_fb->stfb); } } -- cgit v1.2.3 From c2903679856856e3758ceb744cd7d91af9e2eb45 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 11 Feb 2008 15:27:13 -0800 Subject: Vectorize a couple fetch functions These C-coded vectorized functions are expected to be short lived. They're basically a proof-of-concept for dynamically generated fetch routines. --- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 46 ++++++++++++++++++++++++++++--- 1 file changed, 42 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index 6e86a919ce..ec10bb99df 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -46,6 +46,48 @@ static const vec_float4 defaults = { 0.0, 0.0, 0.0, 1.0 }; +static INLINE qword +fetch_unaligned_qword(const void *ptr) +{ + const int shift = (unsigned)(ptr) & 0x0f; + const qword x = *(qword *)(ptr); + const qword y = *(qword *)(ptr + 16); + + return si_or((qword) spu_slqwbyte(x, shift), + (qword) spu_rlmaskqwbyte(y, shift - 16)); +} + +static qword +fetch_R32G32B32A32_FLOAT(const void *ptr) +{ + return fetch_unaligned_qword(ptr); +} + + +static qword +fetch_R32G32B32A32_USCALED(const void *ptr) +{ + return si_cuflt(fetch_unaligned_qword(ptr), 0); +} + + +static qword +fetch_R32G32B32A32_UNORM(const void *ptr) +{ + qword x = si_cuflt(fetch_unaligned_qword(ptr), 0); + vec_float4 scale = spu_splats(1.0f / 255.0f); + + return si_fm(x, (qword) scale); +} + + +static qword +fetch_R32G32B32A32_SSCALED(const void *ptr) +{ + return si_csflt(fetch_unaligned_qword(ptr), 0); +} + + /** * Fetch a float[4] vertex attribute from memory, doing format/type * conversion as needed. @@ -90,22 +132,18 @@ FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) -FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED ) FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM ) FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) -- cgit v1.2.3 From ca1a2da645bceb4500f1782cdfcb686db24ecba7 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 12 Feb 2008 11:16:38 -0800 Subject: Initial pass at vertex cache, more vertex fetch changes This is just another step towards dynamic generate of vertex fetch code. --- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 484 ++++++++++++++++++++++------- src/mesa/pipe/cell/spu/spu_vertex_shader.h | 3 +- 2 files changed, 368 insertions(+), 119 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index ec10bb99df..f6ffcae90e 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -40,41 +40,83 @@ #include "spu_vertex_shader.h" #include "spu_main.h" +#define CACHE_NAME attribute +#define CACHED_TYPE qword +#define CACHE_TYPE CACHE_TYPE_RO +#define CACHE_SET_TAGID(set) TAG_VERTEX_BUFFER +#define CACHE_LOG2NNWAY 2 +#define CACHE_LOG2NSETS 6 +#include + +/* Yes folks, this is ugly. + */ +#undef CACHE_NWAY +#undef CACHE_NSETS +#define CACHE_NAME attribute +#define CACHE_NWAY 4 +#define CACHE_NSETS (1U << 6) + #define DRAW_DBG 0 static const vec_float4 defaults = { 0.0, 0.0, 0.0, 1.0 }; -static INLINE qword -fetch_unaligned_qword(const void *ptr) +/** + * Fetch between 1 and 32 bytes from an unaligned address + */ +static INLINE void +fetch_unaligned(qword *dst, unsigned ea, unsigned size) { - const int shift = (unsigned)(ptr) & 0x0f; - const qword x = *(qword *)(ptr); - const qword y = *(qword *)(ptr + 16); + qword tmp[4]; + const int shift = ea & 0x0f; + const unsigned aligned_start_ea = ea & ~0x0f; + const unsigned aligned_end_ea = (ea + size) & ~0x0f; + const unsigned num_entries = ((aligned_end_ea - aligned_start_ea) / 16) + 1; + unsigned i; + + + if (shift == 0) { + /* Data is already aligned. Fetch directly into the destination buffer. + */ + for (i = 0; i < num_entries; i++) { + dst[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16)); + } + } else { + /* Fetch data from the cache to the local buffer. + */ + for (i = 0; i < num_entries; i++) { + tmp[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16)); + } + - return si_or((qword) spu_slqwbyte(x, shift), - (qword) spu_rlmaskqwbyte(y, shift - 16)); + /* Fix the alignment of the data and write to the destination buffer. + */ + for (i = 0; i < ((size + 15) / 16); i++) { + dst[i] = si_or((qword) spu_slqwbyte(tmp[i], shift), + (qword) spu_rlmaskqwbyte(tmp[i + 1], shift - 16)); + } + } } static qword -fetch_R32G32B32A32_FLOAT(const void *ptr) +fetch_R32G32B32A32_FLOAT(const qword *qw) { - return fetch_unaligned_qword(ptr); + return *qw; } static qword -fetch_R32G32B32A32_USCALED(const void *ptr) +fetch_R32G32B32A32_USCALED(const qword *qw) { - return si_cuflt(fetch_unaligned_qword(ptr), 0); + return si_cuflt(*qw, 0); } static qword -fetch_R32G32B32A32_UNORM(const void *ptr) +fetch_R32G32B32A32_UNORM(const qword *qw) { - qword x = si_cuflt(fetch_unaligned_qword(ptr), 0); + qword x = si_cuflt(*qw, 0); vec_float4 scale = spu_splats(1.0f / 255.0f); return si_fm(x, (qword) scale); @@ -82,12 +124,147 @@ fetch_R32G32B32A32_UNORM(const void *ptr) static qword -fetch_R32G32B32A32_SSCALED(const void *ptr) +fetch_R32G32B32A32_SSCALED(const qword *qw) +{ + return si_csflt(*qw, 0); +} + + +#define CVT_32_FLOAT(q) (*q) + +static INLINE qword +CVT_64_FLOAT(const qword *qw) +{ + qword shuf_first = (qword) { + 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + }; + + qword a = si_frds(qw[0]); + qword b = si_frds(si_rotqbyi(qw[0], 8)); + qword c = si_frds(qw[1]); + qword d = si_frds(si_rotqbyi(qw[1], 8)); + + qword ab = si_shufb(a, b, shuf_first); + qword cd = si_shufb(c, d, si_rotqbyi(shuf_first, 8)); + + return si_or(ab, cd); +} + + +static INLINE qword +CVT_8_USCALED(const qword *qw) +{ + qword shuffle = (qword) { + 0x00, 0x80, 0x80, 0x80, 0x01, 0x80, 0x80, 0x80, + 0x02, 0x80, 0x80, 0x80, 0x03, 0x80, 0x80, 0x80, + }; + + return si_cuflt(si_shufb(*qw, *qw, shuffle), 0); +} + + +static INLINE qword +CVT_16_USCALED(const qword *qw) +{ + qword shuffle = (qword) { + 0x00, 0x01, 0x80, 0x80, 0x02, 0x03, 0x80, 0x80, + 0x04, 0x05, 0x80, 0x80, 0x06, 0x07, 0x80, 0x80, + }; + + return si_cuflt(si_shufb(*qw, *qw, shuffle), 0); +} + + +static INLINE qword +CVT_32_USCALED(const qword *qw) +{ + return si_cuflt(*qw, 0); +} + +static INLINE qword +CVT_8_SSCALED(const qword *qw) { - return si_csflt(fetch_unaligned_qword(ptr), 0); + qword shuffle = (qword) { + 0x00, 0x80, 0x80, 0x80, 0x01, 0x80, 0x80, 0x80, + 0x02, 0x80, 0x80, 0x80, 0x03, 0x80, 0x80, 0x80, + }; + + return si_csflt(si_shufb(*qw, *qw, shuffle), 0); +} + + +static INLINE qword +CVT_16_SSCALED(const qword *qw) +{ + qword shuffle = (qword) { + 0x00, 0x01, 0x80, 0x80, 0x02, 0x03, 0x80, 0x80, + 0x04, 0x05, 0x80, 0x80, 0x06, 0x07, 0x80, 0x80, + }; + + return si_csflt(si_shufb(*qw, *qw, shuffle), 0); +} + + +static INLINE qword +CVT_32_SSCALED(const qword *qw) +{ + return si_csflt(*qw, 0); +} + + +static INLINE qword +CVT_8_UNORM(const qword *qw) +{ + const qword scale = (qword) spu_splats(1.0f / 255.0f); + return si_fm(CVT_8_USCALED(qw), scale); +} + + +static INLINE qword +CVT_16_UNORM(const qword *qw) +{ + const qword scale = (qword) spu_splats(1.0f / 65535.0f); + return si_fm(CVT_16_USCALED(qw), scale); } +static INLINE qword +CVT_32_UNORM(const qword *qw) +{ + const qword scale = (qword) spu_splats(1.0f / 4294967295.0f); + return si_fm(CVT_32_USCALED(qw), scale); +} + + +static INLINE qword +CVT_8_SNORM(const qword *qw) +{ + const qword scale = (qword) spu_splats(1.0f / 127.0f); + return si_fm(CVT_8_SSCALED(qw), scale); +} + + +static INLINE qword +CVT_16_SNORM(const qword *qw) +{ + const qword scale = (qword) spu_splats(1.0f / 32767.0f); + return si_fm(CVT_16_SSCALED(qw), scale); +} + + +static INLINE qword +CVT_32_SNORM(const qword *qw) +{ + const qword scale = (qword) spu_splats(1.0f / 2147483647.0f); + return si_fm(CVT_32_SSCALED(qw), scale); +} + +#define SZ_4 si_il(0U) +#define SZ_3 si_rotqmbyi(si_il(~0), -12) +#define SZ_2 si_rotqmbyi(si_il(~0), -8) +#define SZ_1 si_rotqmbyi(si_il(~0), -4) + /** * Fetch a float[4] vertex attribute from memory, doing format/type * conversion as needed. @@ -97,117 +274,84 @@ fetch_R32G32B32A32_SSCALED(const void *ptr) */ #define FETCH_ATTRIB( NAME, SZ, CVT ) \ static qword \ -fetch_##NAME(const void *ptr) \ +fetch_##NAME(const qword *qw) \ { \ - vec_float4 attrib = defaults; \ - int i; \ - \ - for (i = 0; i < SZ; i++) { \ - attrib = spu_insert(CVT, attrib, i); \ - } \ - return (qword) attrib; \ + qword expanded = CVT(qw); \ + return si_selb(expanded, (qword) defaults, SZ); \ } -#define CVT_64_FLOAT (float) ((double *) ptr)[i] -#define CVT_32_FLOAT ((float *) ptr)[i] - -#define CVT_8_USCALED (float) ((unsigned char *) ptr)[i] -#define CVT_16_USCALED (float) ((unsigned short *) ptr)[i] -#define CVT_32_USCALED (float) ((unsigned int *) ptr)[i] - -#define CVT_8_SSCALED (float) ((char *) ptr)[i] -#define CVT_16_SSCALED (float) ((short *) ptr)[i] -#define CVT_32_SSCALED (float) ((int *) ptr)[i] +FETCH_ATTRIB( R64G64B64A64_FLOAT, SZ_4, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64B64_FLOAT, SZ_3, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64_FLOAT, SZ_2, CVT_64_FLOAT ) +FETCH_ATTRIB( R64_FLOAT, SZ_1, CVT_64_FLOAT ) -#define CVT_8_UNORM (float) ((unsigned char *) ptr)[i] / 255.0f -#define CVT_16_UNORM (float) ((unsigned short *) ptr)[i] / 65535.0f -#define CVT_32_UNORM (float) ((unsigned int *) ptr)[i] / 4294967295.0f +FETCH_ATTRIB( R32G32B32_FLOAT, SZ_3, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32_FLOAT, SZ_2, CVT_32_FLOAT ) +FETCH_ATTRIB( R32_FLOAT, SZ_1, CVT_32_FLOAT ) -#define CVT_8_SNORM (float) ((char *) ptr)[i] / 127.0f -#define CVT_16_SNORM (float) ((short *) ptr)[i] / 32767.0f -#define CVT_32_SNORM (float) ((int *) ptr)[i] / 2147483647.0f +FETCH_ATTRIB( R32G32B32_USCALED, SZ_3, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32_USCALED, SZ_2, CVT_32_USCALED ) +FETCH_ATTRIB( R32_USCALED, SZ_1, CVT_32_USCALED ) -FETCH_ATTRIB( R64G64B64A64_FLOAT, 4, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) -FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) +FETCH_ATTRIB( R32G32B32_SSCALED, SZ_3, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32_SSCALED, SZ_2, CVT_32_SSCALED ) +FETCH_ATTRIB( R32_SSCALED, SZ_1, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) -FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32B32_UNORM, SZ_3, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32_UNORM, SZ_2, CVT_32_UNORM ) +FETCH_ATTRIB( R32_UNORM, SZ_1, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) -FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32B32A32_SNORM, SZ_4, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32B32_SNORM, SZ_3, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32_SNORM, SZ_2, CVT_32_SNORM ) +FETCH_ATTRIB( R32_SNORM, SZ_1, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) -FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) +FETCH_ATTRIB( R16G16B16A16_USCALED, SZ_4, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16B16_USCALED, SZ_3, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16_USCALED, SZ_2, CVT_16_USCALED ) +FETCH_ATTRIB( R16_USCALED, SZ_1, CVT_16_USCALED ) -FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) -FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) +FETCH_ATTRIB( R16G16B16A16_SSCALED, SZ_4, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16B16_SSCALED, SZ_3, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16_SSCALED, SZ_2, CVT_16_SSCALED ) +FETCH_ATTRIB( R16_SSCALED, SZ_1, CVT_16_SSCALED ) -FETCH_ATTRIB( R32G32B32A32_SNORM, 4, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32B32_SNORM, 3, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32_SNORM, 2, CVT_32_SNORM ) -FETCH_ATTRIB( R32_SNORM, 1, CVT_32_SNORM ) +FETCH_ATTRIB( R16G16B16A16_UNORM, SZ_4, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16B16_UNORM, SZ_3, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16_UNORM, SZ_2, CVT_16_UNORM ) +FETCH_ATTRIB( R16_UNORM, SZ_1, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16B16A16_USCALED, 4, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16B16_USCALED, 3, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16_USCALED, 2, CVT_16_USCALED ) -FETCH_ATTRIB( R16_USCALED, 1, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16B16A16_SNORM, SZ_4, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16B16_SNORM, SZ_3, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16_SNORM, SZ_2, CVT_16_SNORM ) +FETCH_ATTRIB( R16_SNORM, SZ_1, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16B16A16_SSCALED, 4, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16B16_SSCALED, 3, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16_SSCALED, 2, CVT_16_SSCALED ) -FETCH_ATTRIB( R16_SSCALED, 1, CVT_16_SSCALED ) +FETCH_ATTRIB( R8G8B8A8_USCALED, SZ_4, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8B8_USCALED, SZ_3, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8_USCALED, SZ_2, CVT_8_USCALED ) +FETCH_ATTRIB( R8_USCALED, SZ_1, CVT_8_USCALED ) -FETCH_ATTRIB( R16G16B16A16_UNORM, 4, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16B16_UNORM, 3, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16_UNORM, 2, CVT_16_UNORM ) -FETCH_ATTRIB( R16_UNORM, 1, CVT_16_UNORM ) +FETCH_ATTRIB( R8G8B8A8_SSCALED, SZ_4, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8B8_SSCALED, SZ_3, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8_SSCALED, SZ_2, CVT_8_SSCALED ) +FETCH_ATTRIB( R8_SSCALED, SZ_1, CVT_8_SSCALED ) -FETCH_ATTRIB( R16G16B16A16_SNORM, 4, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16B16_SNORM, 3, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16_SNORM, 2, CVT_16_SNORM ) -FETCH_ATTRIB( R16_SNORM, 1, CVT_16_SNORM ) +FETCH_ATTRIB( R8G8B8A8_UNORM, SZ_4, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8B8_UNORM, SZ_3, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8_UNORM, SZ_2, CVT_8_UNORM ) +FETCH_ATTRIB( R8_UNORM, SZ_1, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8B8A8_USCALED, 4, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8B8_USCALED, 3, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8_USCALED, 2, CVT_8_USCALED ) -FETCH_ATTRIB( R8_USCALED, 1, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8B8A8_SNORM, SZ_4, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8B8_SNORM, SZ_3, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8_SNORM, SZ_2, CVT_8_SNORM ) +FETCH_ATTRIB( R8_SNORM, SZ_1, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8B8A8_SSCALED, 4, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8B8_SSCALED, 3, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8_SSCALED, 2, CVT_8_SSCALED ) -FETCH_ATTRIB( R8_SSCALED, 1, CVT_8_SSCALED ) - -FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8B8_UNORM, 3, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8_UNORM, 2, CVT_8_UNORM ) -FETCH_ATTRIB( R8_UNORM, 1, CVT_8_UNORM ) - -FETCH_ATTRIB( R8G8B8A8_SNORM, 4, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8B8_SNORM, 3, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8_SNORM, 2, CVT_8_SNORM ) -FETCH_ATTRIB( R8_SNORM, 1, CVT_8_SNORM ) - -FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) -//FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) +FETCH_ATTRIB( A8R8G8B8_UNORM, SZ_4, CVT_8_UNORM ) static spu_fetch_func get_fetch_func( enum pipe_format format ) { -#if 0 - { - char tmp[80]; - pf_sprint_name(tmp, format); - _mesa_printf("%s: %s\n", __FUNCTION__, tmp); - } -#endif - switch (format) { case PIPE_FORMAT_R64_FLOAT: return fetch_R64_FLOAT; @@ -348,6 +492,96 @@ static spu_fetch_func get_fetch_func( enum pipe_format format ) } +static unsigned get_vertex_size( enum pipe_format format ) +{ + switch (format) { + case PIPE_FORMAT_R64_FLOAT: + return 8; + case PIPE_FORMAT_R64G64_FLOAT: + return 2 * 8; + case PIPE_FORMAT_R64G64B64_FLOAT: + return 3 * 8; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return 4 * 8; + + case PIPE_FORMAT_R32_SSCALED: + case PIPE_FORMAT_R32_SNORM: + case PIPE_FORMAT_R32_USCALED: + case PIPE_FORMAT_R32_UNORM: + case PIPE_FORMAT_R32_FLOAT: + return 4; + case PIPE_FORMAT_R32G32_SSCALED: + case PIPE_FORMAT_R32G32_SNORM: + case PIPE_FORMAT_R32G32_USCALED: + case PIPE_FORMAT_R32G32_UNORM: + case PIPE_FORMAT_R32G32_FLOAT: + return 2 * 4; + case PIPE_FORMAT_R32G32B32_SSCALED: + case PIPE_FORMAT_R32G32B32_SNORM: + case PIPE_FORMAT_R32G32B32_USCALED: + case PIPE_FORMAT_R32G32B32_UNORM: + case PIPE_FORMAT_R32G32B32_FLOAT: + return 3 * 4; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + case PIPE_FORMAT_R32G32B32A32_SNORM: + case PIPE_FORMAT_R32G32B32A32_USCALED: + case PIPE_FORMAT_R32G32B32A32_UNORM: + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return 4 * 4; + + case PIPE_FORMAT_R16_SSCALED: + case PIPE_FORMAT_R16_SNORM: + case PIPE_FORMAT_R16_UNORM: + case PIPE_FORMAT_R16_USCALED: + return 2; + case PIPE_FORMAT_R16G16_SSCALED: + case PIPE_FORMAT_R16G16_SNORM: + case PIPE_FORMAT_R16G16_USCALED: + case PIPE_FORMAT_R16G16_UNORM: + return 2 * 2; + case PIPE_FORMAT_R16G16B16_SSCALED: + case PIPE_FORMAT_R16G16B16_SNORM: + case PIPE_FORMAT_R16G16B16_USCALED: + case PIPE_FORMAT_R16G16B16_UNORM: + return 3 * 2; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + case PIPE_FORMAT_R16G16B16A16_SNORM: + case PIPE_FORMAT_R16G16B16A16_USCALED: + case PIPE_FORMAT_R16G16B16A16_UNORM: + return 4 * 2; + + case PIPE_FORMAT_R8_SSCALED: + case PIPE_FORMAT_R8_SNORM: + case PIPE_FORMAT_R8_USCALED: + case PIPE_FORMAT_R8_UNORM: + return 1; + case PIPE_FORMAT_R8G8_SSCALED: + case PIPE_FORMAT_R8G8_SNORM: + case PIPE_FORMAT_R8G8_USCALED: + case PIPE_FORMAT_R8G8_UNORM: + return 2 * 1; + case PIPE_FORMAT_R8G8B8_SSCALED: + case PIPE_FORMAT_R8G8B8_SNORM: + case PIPE_FORMAT_R8G8B8_USCALED: + case PIPE_FORMAT_R8G8B8_UNORM: + return 3 * 1; + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R8G8B8A8_SSCALED: + case PIPE_FORMAT_R8G8B8A8_SNORM: + case PIPE_FORMAT_R8G8B8A8_USCALED: + case PIPE_FORMAT_R8G8B8A8_UNORM: + return 4 * 1; + + case 0: + return 0; /* not sure why this is needed */ + + default: + assert(0); + return 0; + } +} + + /** * Fetch vertex attributes for 'count' vertices. */ @@ -361,8 +595,6 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, assert(count <= 4); - wait_on_mask(1 << TAG_VERTEX_BUFFER); - #if DRAW_DBG printf("SPU: %s count = %u, nr_attrs = %u\n", __FUNCTION__, count, nr_attrs); @@ -375,33 +607,40 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, const uint64_t src = draw->vertex_fetch.src_ptr[attr]; const spu_fetch_func fetch = draw->vertex_fetch.fetch[attr]; unsigned i; + unsigned idx; + const unsigned bytes_per_entry = draw->vertex_fetch.size[attr]; + const unsigned quads_per_entry = (bytes_per_entry + 15) / 16; qword p[4]; + qword in[2 * 4]; /* Fetch four attributes for four vertices. - * - * Could fetch directly into AOS format, but this is meant to be - * a prototype for an sse implementation, which would have - * difficulties doing that. */ + idx = 0; for (i = 0; i < count; i++) { - uint8_t buffer[32] ALIGN16_ATTRIB; const uint64_t addr = src + (elts[i] * pitch); - const unsigned size = ((addr & 0x0f) == 0) ? 16 : 32; #if DRAW_DBG printf("SPU: fetching = 0x%llx\n", addr); #endif - mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); - wait_on_mask(1 << TAG_VERTEX_BUFFER); - p[i] = (*fetch)(buffer + (addr & 0x0f)); + fetch_unaligned(& in[idx], addr, bytes_per_entry); + idx += quads_per_entry; } - /* Be nice and zero out any missing vertices: + /* Be nice and zero out any missing vertices. */ - for (/* empty */; i < 4; i++) - p[i] = si_xor(p[i], p[i]); + (void) memset(& in[idx], 0, (8 - idx) * sizeof(qword)); + + + /* Convert all 4 vertices to vectors of float. + */ + idx = 0; + for (i = 0; i < 4; i++) { + p[i] = (*fetch)(in + idx); + idx += quads_per_entry; + } + /* Transpose/swizzle into vector-friendly format. Currently * assuming that all vertex shader inputs are float[4], but this @@ -422,9 +661,18 @@ void spu_update_vertex_fetch( struct spu_vs_context *draw ) unsigned i; + /* Invalidate the vertex cache. + */ + for (i = 0; i < (CACHE_NWAY * CACHE_NSETS); i++) { + CACHELINE_CLEARVALID(i); + } + + for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { draw->vertex_fetch.fetch[i] = get_fetch_func(draw->vertex_fetch.format[i]); + draw->vertex_fetch.size[i] = + get_vertex_size(draw->vertex_fetch.format[i]); } draw->vertex_fetch.fetch_func = generic_vertex_fetch; diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.h b/src/mesa/pipe/cell/spu/spu_vertex_shader.h index c96b93ff0a..ea044e841d 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.h +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.h @@ -6,7 +6,7 @@ struct spu_vs_context; -typedef qword (*spu_fetch_func)(const void *ptr); +typedef qword (*spu_fetch_func)(const qword *qw); typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw, struct spu_exec_machine *machine, const unsigned *elts, @@ -18,6 +18,7 @@ struct spu_vs_context { struct { uint64_t src_ptr[PIPE_ATTRIB_MAX]; unsigned pitch[PIPE_ATTRIB_MAX]; + unsigned size[PIPE_ATTRIB_MAX]; enum pipe_format format[PIPE_ATTRIB_MAX]; unsigned nr_attrs; boolean dirty; -- cgit v1.2.3 From 125451b9f024ea5845eb6c1b3056bc1f1995cc55 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 12 Feb 2008 11:18:53 -0800 Subject: Remove open-coded fetch functions --- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 34 ++++--------------------------- 1 file changed, 4 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index f6ffcae90e..cbd389435e 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -99,36 +99,6 @@ fetch_unaligned(qword *dst, unsigned ea, unsigned size) } } -static qword -fetch_R32G32B32A32_FLOAT(const qword *qw) -{ - return *qw; -} - - -static qword -fetch_R32G32B32A32_USCALED(const qword *qw) -{ - return si_cuflt(*qw, 0); -} - - -static qword -fetch_R32G32B32A32_UNORM(const qword *qw) -{ - qword x = si_cuflt(*qw, 0); - vec_float4 scale = spu_splats(1.0f / 255.0f); - - return si_fm(x, (qword) scale); -} - - -static qword -fetch_R32G32B32A32_SSCALED(const qword *qw) -{ - return si_csflt(*qw, 0); -} - #define CVT_32_FLOAT(q) (*q) @@ -285,18 +255,22 @@ FETCH_ATTRIB( R64G64B64_FLOAT, SZ_3, CVT_64_FLOAT ) FETCH_ATTRIB( R64G64_FLOAT, SZ_2, CVT_64_FLOAT ) FETCH_ATTRIB( R64_FLOAT, SZ_1, CVT_64_FLOAT ) +FETCH_ATTRIB( R32G32B32A32_FLOAT, SZ_4, CVT_32_FLOAT ) FETCH_ATTRIB( R32G32B32_FLOAT, SZ_3, CVT_32_FLOAT ) FETCH_ATTRIB( R32G32_FLOAT, SZ_2, CVT_32_FLOAT ) FETCH_ATTRIB( R32_FLOAT, SZ_1, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32B32A32_USCALED, SZ_4, CVT_32_USCALED ) FETCH_ATTRIB( R32G32B32_USCALED, SZ_3, CVT_32_USCALED ) FETCH_ATTRIB( R32G32_USCALED, SZ_2, CVT_32_USCALED ) FETCH_ATTRIB( R32_USCALED, SZ_1, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32B32A32_SSCALED, SZ_4, CVT_32_SSCALED ) FETCH_ATTRIB( R32G32B32_SSCALED, SZ_3, CVT_32_SSCALED ) FETCH_ATTRIB( R32G32_SSCALED, SZ_2, CVT_32_SSCALED ) FETCH_ATTRIB( R32_SSCALED, SZ_1, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32B32A32_UNORM, SZ_4, CVT_32_UNORM ) FETCH_ATTRIB( R32G32B32_UNORM, SZ_3, CVT_32_UNORM ) FETCH_ATTRIB( R32G32_UNORM, SZ_2, CVT_32_UNORM ) FETCH_ATTRIB( R32_UNORM, SZ_1, CVT_32_UNORM ) -- cgit v1.2.3 From dd07e154d26c2c3ec248b7143eb67b6b4410246a Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 12 Feb 2008 11:29:34 -0800 Subject: Fetch routines convert and transpose all 4 vertices at once. --- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 152 +++++++++++++---------------- src/mesa/pipe/cell/spu/spu_vertex_shader.h | 2 +- 2 files changed, 71 insertions(+), 83 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index cbd389435e..3bbf9b7be4 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -100,7 +100,7 @@ fetch_unaligned(qword *dst, unsigned ea, unsigned size) } -#define CVT_32_FLOAT(q) (*q) +#define CVT_32_FLOAT(q) (*(q)) static INLINE qword CVT_64_FLOAT(const qword *qw) @@ -242,85 +242,90 @@ CVT_32_SNORM(const qword *qw) * This is probably needed/dupliocated elsewhere, eg format * conversion, texture sampling etc. */ -#define FETCH_ATTRIB( NAME, SZ, CVT ) \ -static qword \ -fetch_##NAME(const qword *qw) \ -{ \ - qword expanded = CVT(qw); \ - return si_selb(expanded, (qword) defaults, SZ); \ +#define FETCH_ATTRIB( NAME, SZ, CVT, N ) \ +static void \ +fetch_##NAME(qword *out, const qword *in) \ +{ \ + qword tmp[4]; \ + \ + tmp[0] = si_selb(CVT(in + (0 * N)), (qword) defaults, SZ); \ + tmp[1] = si_selb(CVT(in + (1 * N)), (qword) defaults, SZ); \ + tmp[2] = si_selb(CVT(in + (2 * N)), (qword) defaults, SZ); \ + tmp[3] = si_selb(CVT(in + (3 * N)), (qword) defaults, SZ); \ + _transpose_matrix4x4((vec_float4 *) out, (vec_float4 *) tmp); \ } -FETCH_ATTRIB( R64G64B64A64_FLOAT, SZ_4, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64B64_FLOAT, SZ_3, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64_FLOAT, SZ_2, CVT_64_FLOAT ) -FETCH_ATTRIB( R64_FLOAT, SZ_1, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64B64A64_FLOAT, SZ_4, CVT_64_FLOAT, 2 ) +FETCH_ATTRIB( R64G64B64_FLOAT, SZ_3, CVT_64_FLOAT, 2 ) +FETCH_ATTRIB( R64G64_FLOAT, SZ_2, CVT_64_FLOAT, 2 ) +FETCH_ATTRIB( R64_FLOAT, SZ_1, CVT_64_FLOAT, 2 ) -FETCH_ATTRIB( R32G32B32A32_FLOAT, SZ_4, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32B32_FLOAT, SZ_3, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32_FLOAT, SZ_2, CVT_32_FLOAT ) -FETCH_ATTRIB( R32_FLOAT, SZ_1, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32B32A32_FLOAT, SZ_4, CVT_32_FLOAT, 1 ) +FETCH_ATTRIB( R32G32B32_FLOAT, SZ_3, CVT_32_FLOAT, 1 ) +FETCH_ATTRIB( R32G32_FLOAT, SZ_2, CVT_32_FLOAT, 1 ) +FETCH_ATTRIB( R32_FLOAT, SZ_1, CVT_32_FLOAT, 1 ) -FETCH_ATTRIB( R32G32B32A32_USCALED, SZ_4, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32B32_USCALED, SZ_3, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32_USCALED, SZ_2, CVT_32_USCALED ) -FETCH_ATTRIB( R32_USCALED, SZ_1, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32B32A32_USCALED, SZ_4, CVT_32_USCALED, 1 ) +FETCH_ATTRIB( R32G32B32_USCALED, SZ_3, CVT_32_USCALED, 1 ) +FETCH_ATTRIB( R32G32_USCALED, SZ_2, CVT_32_USCALED, 1 ) +FETCH_ATTRIB( R32_USCALED, SZ_1, CVT_32_USCALED, 1 ) -FETCH_ATTRIB( R32G32B32A32_SSCALED, SZ_4, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32B32_SSCALED, SZ_3, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32_SSCALED, SZ_2, CVT_32_SSCALED ) -FETCH_ATTRIB( R32_SSCALED, SZ_1, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32B32A32_SSCALED, SZ_4, CVT_32_SSCALED, 1 ) +FETCH_ATTRIB( R32G32B32_SSCALED, SZ_3, CVT_32_SSCALED, 1 ) +FETCH_ATTRIB( R32G32_SSCALED, SZ_2, CVT_32_SSCALED, 1 ) +FETCH_ATTRIB( R32_SSCALED, SZ_1, CVT_32_SSCALED, 1 ) -FETCH_ATTRIB( R32G32B32A32_UNORM, SZ_4, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32B32_UNORM, SZ_3, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32_UNORM, SZ_2, CVT_32_UNORM ) -FETCH_ATTRIB( R32_UNORM, SZ_1, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32B32A32_UNORM, SZ_4, CVT_32_UNORM, 1 ) +FETCH_ATTRIB( R32G32B32_UNORM, SZ_3, CVT_32_UNORM, 1 ) +FETCH_ATTRIB( R32G32_UNORM, SZ_2, CVT_32_UNORM, 1 ) +FETCH_ATTRIB( R32_UNORM, SZ_1, CVT_32_UNORM, 1 ) -FETCH_ATTRIB( R32G32B32A32_SNORM, SZ_4, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32B32_SNORM, SZ_3, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32_SNORM, SZ_2, CVT_32_SNORM ) -FETCH_ATTRIB( R32_SNORM, SZ_1, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32B32A32_SNORM, SZ_4, CVT_32_SNORM, 1 ) +FETCH_ATTRIB( R32G32B32_SNORM, SZ_3, CVT_32_SNORM, 1 ) +FETCH_ATTRIB( R32G32_SNORM, SZ_2, CVT_32_SNORM, 1 ) +FETCH_ATTRIB( R32_SNORM, SZ_1, CVT_32_SNORM, 1 ) -FETCH_ATTRIB( R16G16B16A16_USCALED, SZ_4, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16B16_USCALED, SZ_3, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16_USCALED, SZ_2, CVT_16_USCALED ) -FETCH_ATTRIB( R16_USCALED, SZ_1, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16B16A16_USCALED, SZ_4, CVT_16_USCALED, 1 ) +FETCH_ATTRIB( R16G16B16_USCALED, SZ_3, CVT_16_USCALED, 1 ) +FETCH_ATTRIB( R16G16_USCALED, SZ_2, CVT_16_USCALED, 1 ) +FETCH_ATTRIB( R16_USCALED, SZ_1, CVT_16_USCALED, 1 ) -FETCH_ATTRIB( R16G16B16A16_SSCALED, SZ_4, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16B16_SSCALED, SZ_3, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16_SSCALED, SZ_2, CVT_16_SSCALED ) -FETCH_ATTRIB( R16_SSCALED, SZ_1, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16B16A16_SSCALED, SZ_4, CVT_16_SSCALED, 1 ) +FETCH_ATTRIB( R16G16B16_SSCALED, SZ_3, CVT_16_SSCALED, 1 ) +FETCH_ATTRIB( R16G16_SSCALED, SZ_2, CVT_16_SSCALED, 1 ) +FETCH_ATTRIB( R16_SSCALED, SZ_1, CVT_16_SSCALED, 1 ) -FETCH_ATTRIB( R16G16B16A16_UNORM, SZ_4, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16B16_UNORM, SZ_3, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16_UNORM, SZ_2, CVT_16_UNORM ) -FETCH_ATTRIB( R16_UNORM, SZ_1, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16B16A16_UNORM, SZ_4, CVT_16_UNORM, 1 ) +FETCH_ATTRIB( R16G16B16_UNORM, SZ_3, CVT_16_UNORM, 1 ) +FETCH_ATTRIB( R16G16_UNORM, SZ_2, CVT_16_UNORM, 1 ) +FETCH_ATTRIB( R16_UNORM, SZ_1, CVT_16_UNORM, 1 ) -FETCH_ATTRIB( R16G16B16A16_SNORM, SZ_4, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16B16_SNORM, SZ_3, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16_SNORM, SZ_2, CVT_16_SNORM ) -FETCH_ATTRIB( R16_SNORM, SZ_1, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16B16A16_SNORM, SZ_4, CVT_16_SNORM, 1 ) +FETCH_ATTRIB( R16G16B16_SNORM, SZ_3, CVT_16_SNORM, 1 ) +FETCH_ATTRIB( R16G16_SNORM, SZ_2, CVT_16_SNORM, 1 ) +FETCH_ATTRIB( R16_SNORM, SZ_1, CVT_16_SNORM, 1 ) -FETCH_ATTRIB( R8G8B8A8_USCALED, SZ_4, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8B8_USCALED, SZ_3, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8_USCALED, SZ_2, CVT_8_USCALED ) -FETCH_ATTRIB( R8_USCALED, SZ_1, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8B8A8_USCALED, SZ_4, CVT_8_USCALED, 1 ) +FETCH_ATTRIB( R8G8B8_USCALED, SZ_3, CVT_8_USCALED, 1 ) +FETCH_ATTRIB( R8G8_USCALED, SZ_2, CVT_8_USCALED, 1 ) +FETCH_ATTRIB( R8_USCALED, SZ_1, CVT_8_USCALED, 1 ) -FETCH_ATTRIB( R8G8B8A8_SSCALED, SZ_4, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8B8_SSCALED, SZ_3, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8_SSCALED, SZ_2, CVT_8_SSCALED ) -FETCH_ATTRIB( R8_SSCALED, SZ_1, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8B8A8_SSCALED, SZ_4, CVT_8_SSCALED, 1 ) +FETCH_ATTRIB( R8G8B8_SSCALED, SZ_3, CVT_8_SSCALED, 1 ) +FETCH_ATTRIB( R8G8_SSCALED, SZ_2, CVT_8_SSCALED, 1 ) +FETCH_ATTRIB( R8_SSCALED, SZ_1, CVT_8_SSCALED, 1 ) -FETCH_ATTRIB( R8G8B8A8_UNORM, SZ_4, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8B8_UNORM, SZ_3, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8_UNORM, SZ_2, CVT_8_UNORM ) -FETCH_ATTRIB( R8_UNORM, SZ_1, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8B8A8_UNORM, SZ_4, CVT_8_UNORM, 1 ) +FETCH_ATTRIB( R8G8B8_UNORM, SZ_3, CVT_8_UNORM, 1 ) +FETCH_ATTRIB( R8G8_UNORM, SZ_2, CVT_8_UNORM, 1 ) +FETCH_ATTRIB( R8_UNORM, SZ_1, CVT_8_UNORM, 1 ) -FETCH_ATTRIB( R8G8B8A8_SNORM, SZ_4, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8B8_SNORM, SZ_3, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8_SNORM, SZ_2, CVT_8_SNORM ) -FETCH_ATTRIB( R8_SNORM, SZ_1, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8B8A8_SNORM, SZ_4, CVT_8_SNORM, 1 ) +FETCH_ATTRIB( R8G8B8_SNORM, SZ_3, CVT_8_SNORM, 1 ) +FETCH_ATTRIB( R8G8_SNORM, SZ_2, CVT_8_SNORM, 1 ) +FETCH_ATTRIB( R8_SNORM, SZ_1, CVT_8_SNORM, 1 ) -FETCH_ATTRIB( A8R8G8B8_UNORM, SZ_4, CVT_8_UNORM ) +FETCH_ATTRIB( A8R8G8B8_UNORM, SZ_4, CVT_8_UNORM, 1 ) @@ -584,7 +589,6 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, unsigned idx; const unsigned bytes_per_entry = draw->vertex_fetch.size[attr]; const unsigned quads_per_entry = (bytes_per_entry + 15) / 16; - qword p[4]; qword in[2 * 4]; @@ -609,23 +613,7 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, /* Convert all 4 vertices to vectors of float. */ - idx = 0; - for (i = 0; i < 4; i++) { - p[i] = (*fetch)(in + idx); - idx += quads_per_entry; - } - - - /* Transpose/swizzle into vector-friendly format. Currently - * assuming that all vertex shader inputs are float[4], but this - * isn't true -- if the vertex shader only wants tex0.xy, we - * could optimize for that. - * - * To do so fully without codegen would probably require an - * excessive number of fetch functions, but we could at least - * minimize the transpose step: - */ - _transpose_matrix4x4(&machine->Inputs[attr].xyzw[0].q, p); + (*fetch)(&machine->Inputs[attr].xyzw[0].q, in); } } diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.h b/src/mesa/pipe/cell/spu/spu_vertex_shader.h index ea044e841d..8b37a239a4 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.h +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.h @@ -6,7 +6,7 @@ struct spu_vs_context; -typedef qword (*spu_fetch_func)(const qword *qw); +typedef void (*spu_fetch_func)(qword *out, const qword *in); typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw, struct spu_exec_machine *machine, const unsigned *elts, -- cgit v1.2.3 From 77a148862036bd6bf01f631ff44e455d50bcb8b9 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 14 Feb 2008 10:08:11 -0800 Subject: Real-time assembler for Cell SPE. --- src/mesa/ppc/rtasm/spe_asm.c | 351 +++++++++++++++++++++++++++++++++++++++++++ src/mesa/ppc/rtasm/spe_asm.h | 310 ++++++++++++++++++++++++++++++++++++++ src/mesa/sources | 1 + 3 files changed, 662 insertions(+) create mode 100644 src/mesa/ppc/rtasm/spe_asm.c create mode 100644 src/mesa/ppc/rtasm/spe_asm.h (limited to 'src') diff --git a/src/mesa/ppc/rtasm/spe_asm.c b/src/mesa/ppc/rtasm/spe_asm.c new file mode 100644 index 0000000000..b1851f05e7 --- /dev/null +++ b/src/mesa/ppc/rtasm/spe_asm.c @@ -0,0 +1,351 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file spe_asm.c + * Real-time assembly generation interface for Cell B.E. SPEs. + * + * \author Ian Romanick + */ +#ifdef GALLIUM_CELL +#include +#include +#include "spe_asm.h" + +/** + * SPE instruction types + * + * There are 6 primary instruction encodings used on the Cell's SPEs. Each of + * the following unions encodes one type. + * + * \bug + * If, at some point, we start generating SPE code from a little-endian host + * these unions will not work. + */ +/*@{*/ +/** + * Encode one output register with two input registers + */ +union spe_inst_RR { + uint32_t bits; + struct { + unsigned op:11; + unsigned rB:7; + unsigned rA:7; + unsigned rT:7; + } inst; +}; + + +/** + * Encode one output register with three input registers + */ +union spe_inst_RRR { + uint32_t bits; + struct { + unsigned op:4; + unsigned rT:7; + unsigned rB:7; + unsigned rA:7; + unsigned rC:7; + } inst; +}; + + +/** + * Encode one output register with one input reg. and a 7-bit signed immed + */ +union spe_inst_RI7 { + uint32_t bits; + struct { + unsigned op:11; + unsigned i7:7; + unsigned rA:7; + unsigned rT:7; + } inst; +}; + + +/** + * Encode one output register with one input reg. and a 10-bit signed immed + */ +union spe_inst_RI10 { + uint32_t bits; + struct { + unsigned op:8; + unsigned i10:10; + unsigned rA:7; + unsigned rT:7; + } inst; +}; + + +/** + * Encode one output register with a 16-bit signed immediate + */ +union spe_inst_RI16 { + uint32_t bits; + struct { + unsigned op:9; + unsigned i16:16; + unsigned rT:7; + } inst; +}; + + +/** + * Encode one output register with a 18-bit signed immediate + */ +union spe_inst_RI18 { + uint32_t bits; + struct { + unsigned op:7; + unsigned i18:18; + unsigned rT:7; + } inst; +}; +/*@}*/ + + +static void emit_RR(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, unsigned rB) +{ + union spe_inst_RR inst; + inst.inst.op = op; + inst.inst.rB = rB; + inst.inst.rA = rA; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + +static void emit_RRR(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, unsigned rB, unsigned rC) +{ + union spe_inst_RRR inst; + inst.inst.op = op; + inst.inst.rT = rT; + inst.inst.rB = rB; + inst.inst.rA = rA; + inst.inst.rC = rC; + *p->csr = inst.bits; + p->csr++; +} + + +static void emit_RI7(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, int imm) +{ + union spe_inst_RI7 inst; + inst.inst.op = op; + inst.inst.i7 = imm; + inst.inst.rA = rA; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + + +static void emit_RI10(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, int imm) +{ + union spe_inst_RI10 inst; + inst.inst.op = op; + inst.inst.i10 = imm; + inst.inst.rA = rA; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + +static void emit_RI16(struct spe_function *p, unsigned op, unsigned rT, + int imm) +{ + union spe_inst_RI16 inst; + inst.inst.op = op; + inst.inst.i16 = imm; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + +static void emit_RI18(struct spe_function *p, unsigned op, unsigned rT, + int imm) +{ + union spe_inst_RI18 inst; + inst.inst.op = op; + inst.inst.i18 = imm; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + + + +#define EMIT_(_name, _op) \ +void _name (struct spe_function *p, unsigned rT) \ +{ \ + emit_RR(p, _op, rT, 0, 0); \ +} + +#define EMIT_R(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA) \ +{ \ + emit_RR(p, _op, rT, rA, 0); \ +} + +#define EMIT_RR(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, unsigned rB) \ +{ \ + emit_RR(p, _op, rT, rA, rB); \ +} + +#define EMIT_RRR(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, unsigned rB, unsigned rC) \ +{ \ + emit_RRR(p, _op, rT, rA, rB, rC); \ +} + +#define EMIT_RI7(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ +{ \ + emit_RI7(p, _op, rT, rA, imm); \ +} + +#define EMIT_RI10(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ +{ \ + emit_RI10(p, _op, rT, rA, imm); \ +} + +#define EMIT_RI16(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, int imm) \ +{ \ + emit_RI16(p, _op, rT, imm); \ +} + +#define EMIT_RI18(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, int imm) \ +{ \ + emit_RI18(p, _op, rT, imm); \ +} + +#define EMIT_I16(_name, _op) \ +void _name (struct spe_function *p, int imm) \ +{ \ + emit_RI16(p, _op, 0, imm); \ +} + +#include "spe_asm.h" + + +/* + */ +void spe_init_func(struct spe_function *p, unsigned code_size) +{ + p->store = _mesa_align_malloc(code_size, 16); + p->csr = p->store; +} + + +void spe_release_func(struct spe_function *p) +{ + _mesa_align_free(p->store); + p->store = NULL; + p->csr = NULL; +} + + +void spu_bi(struct spe_function *p, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x1a8, 0, rA, (d << 5) | (e << 4)); +} + +void spu_iret(struct spe_function *p, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x1aa, 0, rA, (d << 5) | (e << 4)); +} + +void spu_bisled(struct spe_function *p, unsigned rT, unsigned rA, int d, + int e) +{ + emit_RI7(p, 0x1ab, rT, rA, (d << 5) | (e << 4)); +} + +void spu_bisl(struct spe_function *p, unsigned rT, unsigned rA, int d, + int e) +{ + emit_RI7(p, 0x1a9, rT, rA, (d << 5) | (e << 4)); +} + +void spu_biz(struct spe_function *p, unsigned rT, unsigned rA, int d, + int e) +{ + emit_RI7(p, 0x128, rT, rA, (d << 5) | (e << 4)); +} + +void spu_binz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x129, rT, rA, (d << 5) | (e << 4)); +} + +void spu_bihz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x12a, rT, rA, (d << 5) | (e << 4)); +} + +void spu_bihnz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x12b, rT, rA, (d << 5) | (e << 4)); +} + + +/* Hint-for-branch instructions + */ +#if 0 +hbr; +hbra; +hbrr; +#endif + + +/* Control instructions + */ +#if 0 +stop; +EMIT_RR (spu_stopd, 0x140); +EMIT_ (spu_lnop, 0x001); +EMIT_ (spu_nop, 0x201); +sync; +EMIT_ (spu_dsync, 0x003); +EMIT_R (spu_mfspr, 0x00c); +EMIT_R (spu_mtspr, 0x10c); +#endif + +#endif /* GALLIUM_CELL */ diff --git a/src/mesa/ppc/rtasm/spe_asm.h b/src/mesa/ppc/rtasm/spe_asm.h new file mode 100644 index 0000000000..1c7f145415 --- /dev/null +++ b/src/mesa/ppc/rtasm/spe_asm.h @@ -0,0 +1,310 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file spe_asm.h + * Real-time assembly generation interface for Cell B.E. SPEs. + * + * \author Ian Romanick + */ + +#ifndef SPE_ASM_H +#define SPE_ASM_H + +struct spe_function { + /** + * + */ + uint32_t *store; + uint32_t *csr; + const char *fn; +}; + +extern void spe_init_func(struct spe_function *p, unsigned code_size); +extern void spe_release_func(struct spe_function *p); + +#endif /* SPE_ASM_H */ + +#ifndef EMIT_ +#define EMIT_(name, _op) \ + extern void _name (struct spe_function *p, unsigned rT) +#define EMIT_R(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA) +#define EMIT_RR(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + unsigned rB) +#define EMIT_RRR(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + unsigned rB, unsigned rC) +#define EMIT_RI7(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + int imm) +#define EMIT_RI10(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + int imm) +#define EMIT_RI16(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, int imm) +#define EMIT_RI18(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, int imm) +#define EMIT_I16(_name, _op) \ + extern void _name (struct spe_function *p, int imm) +#define UNDEF_EMIT_MACROS +#endif /* EMIT_ */ + + +/* Memory load / store instructions + */ +EMIT_RI10(spu_ldq, 0x034); +EMIT_RR (spu_lqx, 0x1c4); +EMIT_RI16(spu_lqa, 0x061); +EMIT_RI16(spu_lqr, 0x067); +EMIT_RI10(spu_stqd, 0x024); +EMIT_RR (spu_stqx, 0x144); +EMIT_RI16(spu_stqa, 0x041); +EMIT_RI16(spu_stqr, 0x047); +EMIT_RI7 (spu_cbd, 0x1f4); +EMIT_RR (spu_cbx, 0x1d4); +EMIT_RI7 (spu_chd, 0x1f5); +EMIT_RI7 (spu_chx, 0x1d5); +EMIT_RI7 (spu_cwd, 0x1f6); +EMIT_RI7 (spu_cwx, 0x1d6); +EMIT_RI7 (spu_cdd, 0x1f7); +EMIT_RI7 (spu_cdx, 0x1d7); + + +/* Constant formation instructions + */ +EMIT_RI16(spu_ilh, 0x083); +EMIT_RI16(spu_ilhu, 0x082); +EMIT_RI16(spu_il, 0x081); +EMIT_RI18(spu_ila, 0x021); +EMIT_RI16(spu_iohl, 0x0c1); +EMIT_RI16(spu_fsmbi, 0x0c5); + + +/* Integer and logical instructions + */ +EMIT_RR (spe_ah, 0x0c8); +EMIT_RI10(spe_ahi, 0x01d); +EMIT_RR (spe_a, 0x0c0); +EMIT_RI10(spe_ai, 0x01c); +EMIT_RR (spe_sfh, 0x048); +EMIT_RI10(spe_sfhi, 0x00d); +EMIT_RR (spe_sf, 0x040); +EMIT_RI10(spe_sfi, 0x00c); +EMIT_RR (spe_addx, 0x340); +EMIT_RR (spu_cg, 0x0c2); +EMIT_RR (spu_cgx, 0x342); +EMIT_RR (spe_sfx, 0x341); +EMIT_RR (spu_bg, 0x042); +EMIT_RR (spu_bgx, 0x343); +EMIT_RR (spu_mpy, 0x3c4); +EMIT_RR (spu_mpyu, 0x3cc); +EMIT_RI10(spu_mpyi, 0x074); +EMIT_RI10(spu_mpyui, 0x075); +EMIT_RRR (spy_mpya, 0x00c); +EMIT_RR (spu_mpyh, 0x3c5); +EMIT_RR (spu_mpys, 0x3c7); +EMIT_RR (spu_mpyhh, 0x3c6); +EMIT_RR (spu_mpyhha, 0x346); +EMIT_RR (spu_mpyhhu, 0x3ce); +EMIT_RR (spu_mpyhhau, 0x34e); +EMIT_R (spe_clz, 0x2a5); +EMIT_R (spe_cntb, 0x2b4); +EMIT_R (spe_fsmb, 0x1b6); +EMIT_R (spe_fsmh, 0x1b5); +EMIT_R (spe_fsm, 0x1b4); +EMIT_R (spe_gbb, 0x1b2); +EMIT_R (spe_gbh, 0x1b1); +EMIT_R (spe_gb, 0x1b0); +EMIT_RR (spe_avgb, 0x0d3); +EMIT_RR (spe_absdb, 0x053); +EMIT_RR (spe_sumb, 0x253); +EMIT_R (spe_xsbh, 0x2b6); +EMIT_R (spe_xshw, 0x2ae); +EMIT_R (spe_xswd, 0x2a6); +EMIT_RR (spe_and, 0x0c1); +EMIT_RR (spe_andc, 0x2c1); +EMIT_RI10(spu_andbi, 0x016); +EMIT_RI10(spu_andhi, 0x015); +EMIT_RI10(spu_andi, 0x014); +EMIT_RR (spe_or, 0x041); +EMIT_RR (spe_orc, 0x2c9); +EMIT_RI10(spu_orbi, 0x006); +EMIT_RI10(spu_orhi, 0x005); +EMIT_RI10(spu_ori, 0x004); +EMIT_R (spu_orx, 0x1f0); +EMIT_RR (spu_xor, 0x241); +EMIT_RI10(spu_xorbi, 0x026); +EMIT_RI10(spu_xorhi, 0x025); +EMIT_RI10(spu_xori, 0x024); +EMIT_RR (spe_nand, 0x0c9); +EMIT_RR (spe_nor, 0x049); +EMIT_RR (spe_eqv, 0x249); +EMIT_RRR (spy_selb, 0x008); +EMIT_RRR (spy_shufb, 0x00b); + + +/* Shift and rotate instructions + */ +EMIT_RR (spe_shlh, 0x05f); +EMIT_RI7 (spe_shlhi, 0x07f); +EMIT_RR (spe_shl, 0x05b); +EMIT_RI7 (spe_shli, 0x07b); +EMIT_RR (spe_shlqbi, 0x1db); +EMIT_RI7 (spe_shlqbii, 0x1fb); +EMIT_RR (spe_shlqby, 0x1df); +EMIT_RI7 (spe_shlqbyi, 0x1ff); +EMIT_RR (spe_shlqbybi, 0x1cf); +EMIT_RR (spe_roth, 0x05c); +EMIT_RI7 (spe_rothi, 0x07c); +EMIT_RR (spe_rot, 0x058); +EMIT_RI7 (spe_roti, 0x078); +EMIT_RR (spe_rotqby, 0x1dc); +EMIT_RI7 (spe_rotqbyi, 0x1fc); +EMIT_RR (spe_rotqbybi, 0x1cc); +EMIT_RR (spe_rotqbi, 0x1d8); +EMIT_RI7 (spe_rotqbii, 0x1f8); +EMIT_RR (spe_rothm, 0x05d); +EMIT_RI7 (spe_rothmi, 0x07d); +EMIT_RR (spe_rotm, 0x059); +EMIT_RI7 (spe_rotmi, 0x079); +EMIT_RR (spe_rotqmby, 0x1dd); +EMIT_RI7 (spe_rotqmbyi, 0x1fd); +EMIT_RR (spe_rotqmbybi, 0x1cd); +EMIT_RR (spe_rotqmbi, 0x1c9); +EMIT_RI7 (spe_rotqmbii, 0x1f9); +EMIT_RR (spe_rotmah, 0x05e); +EMIT_RI7 (spe_rotmahi, 0x07e); +EMIT_RR (spe_rotma, 0x05a); +EMIT_RI7 (spe_rotmai, 0x07a); + + +/* Compare, branch, and halt instructions + */ +EMIT_RR (spe_heq, 0x3d8); +EMIT_RI10(spe_heqi, 0x07f); +EMIT_RR (spe_hgt, 0x258); +EMIT_RI10(spe_hgti, 0x04f); +EMIT_RR (spe_hlgt, 0x2d8); +EMIT_RI10(spe_hlgti, 0x05f); +EMIT_RR (spe_ceqb, 0x3d0); +EMIT_RI10(spe_ceqbi, 0x07e); +EMIT_RR (spe_ceqh, 0x3c8); +EMIT_RI10(spe_ceqhi, 0x07d); +EMIT_RR (spe_ceq, 0x3c0); +EMIT_RI10(spe_ceqi, 0x07c); +EMIT_RR (spe_cgtb, 0x250); +EMIT_RI10(spe_cgtbi, 0x04e); +EMIT_RR (spe_cgth, 0x248); +EMIT_RI10(spe_cgthi, 0x04d); +EMIT_RR (spe_cgt, 0x240); +EMIT_RI10(spe_cgti, 0x04c); +EMIT_RR (spe_clgtb, 0x2d0); +EMIT_RI10(spe_clgtbi, 0x05e); +EMIT_RR (spe_clgth, 0x2c8); +EMIT_RI10(spe_clgthi, 0x05d); +EMIT_RR (spe_clgt, 0x2c0); +EMIT_RI10(spe_clgti, 0x05c); +EMIT_I16 (spe_br, 0x064); +EMIT_I16 (spe_bra, 0x060); +EMIT_RI16(spu_brsl, 0x066); +EMIT_RI16(spu_brasl, 0x062); +EMIT_RI16(spu_brnz, 0x042); +EMIT_RI16(spu_brz, 0x040); +EMIT_RI16(spu_brhnz, 0x046); +EMIT_RI16(spu_brhz, 0x044); + +extern void spu_bi(struct spe_function *p, unsigned rA, int d, int e); +extern void spu_iret(struct spe_function *p, unsigned rA, int d, int e); +extern void spu_bisled(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spu_bisl(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spu_biz(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spu_binz(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spu_bihz(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spu_bihnz(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); + + +/* Floating-point instructions + */ +EMIT_RR (spu_fa, 0x2c4); +EMIT_RR (spu_dfa, 0x2cc); +EMIT_RR (spu_fs, 0x2c5); +EMIT_RR (spu_dfs, 0x2cd); +EMIT_RR (spu_fm, 0x2c6); +EMIT_RR (spu_dfm, 0x2ce); +EMIT_RRR (spu_fma, 0x00e); +EMIT_RR (spu_dfma, 0x35c); +EMIT_RRR (spu_fnms, 0x00d); +EMIT_RR (spu_dfnms, 0x35e); +EMIT_RRR (spu_fms, 0x00f); +EMIT_RR (spu_dfms, 0x35d); +EMIT_RR (spu_dfnma, 0x35f); +EMIT_R (spu_frest, 0x1b8); +EMIT_R (spu_frsqest, 0x1b9); +EMIT_RR (spu_fi, 0x3d4); +EMIT_RI7 (spu_csflt, 0x3da); +EMIT_RI7 (spu_cflts, 0x3d8); +EMIT_RI7 (spu_cuflt, 0x3db); +EMIT_RI7 (spu_cfltu, 0x3d9); +EMIT_R (spu_frds, 0x3b9); +EMIT_R (spu_fesd, 0x3b8); +EMIT_RR (spu_dfceq, 0x3c3); +EMIT_RR (spu_dfcmeq, 0x3cb); +EMIT_RR (spu_dfcgt, 0x2c3); +EMIT_RR (spu_dfcmgt, 0x2cb); +EMIT_RI7 (spu_dftsv, 0x3bf); +EMIT_RR (spu_fceq, 0x3c2); +EMIT_RR (spu_fcmeq, 0x3ca); +EMIT_RR (spu_fcgt, 0x2c2); +EMIT_RR (spu_fcmgt, 0x2ca); +EMIT_R (spu_fscrwr, 0x3ba); +EMIT_ (spu_fscrrd, 0x398); + + +/* Channel instructions + */ +EMIT_R (spu_rdch, 0x00d); +EMIT_R (spu_rdchcnt, 0x00f); +EMIT_R (spu_wrch, 0x10d); + + +#ifdef UNDEF_EMIT_MACROS +#undef EMIT_ +#undef EMIT_R +#undef EMIT_RR +#undef EMIT_RRR +#undef EMIT_RI7 +#undef EMIT_RI10 +#undef EMIT_RI16 +#undef EMIT_RI18 +#undef EMIT_I16 +#undef UNDEF_EMIT_MACROS +#endif /* EMIT_ */ diff --git a/src/mesa/sources b/src/mesa/sources index 96ae3dbca0..1165425183 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -288,6 +288,7 @@ ASM_C_SOURCES = \ x86/rtasm/x86sse.c \ sparc/sparc.c \ ppc/common_ppc.c \ + ppc/rtasm/spe_asm.c \ x86-64/x86-64.c X86_SOURCES = \ -- cgit v1.2.3 From f1257fd7944c5b62a07cecf0608e63ec7a947706 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 14 Feb 2008 11:55:19 -0800 Subject: Cell: Fix unintended breakage from commit 1d62a057bcb3ee4ef6ebedd93f62ed2e0d8061ba --- src/mesa/pipe/cell/ppu/cell_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_texture.c b/src/mesa/pipe/cell/ppu/cell_texture.c index df178d9ca2..c8ef36002f 100644 --- a/src/mesa/pipe/cell/ppu/cell_texture.c +++ b/src/mesa/pipe/cell/ppu/cell_texture.c @@ -61,7 +61,7 @@ cell_texture_layout(struct cell_texture * spt) spt->buffer_size = 0; - for ( level = pt->first_level ; level <= pt->last_level ; level++ ) { + for ( level = 0 ; level <= pt->last_level ; level++ ) { pt->width[level] = width; pt->height[level] = height; pt->depth[level] = depth; -- cgit v1.2.3 From 18fd3b757166c1c63284dc08f6dfd9e2061770be Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 14 Feb 2008 16:53:05 -0800 Subject: Cell: pass pointers to stored memory values Several routines use shuffle patterns that are stored in memory. For code gen, it is difficult to directly access the data segments. The routines have been modified to be passed a pointer to a global table of shuffle patterns. This *should* be the last change to this file before switching over to code gen. --- src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 160 +++++++++++++++++------------ src/mesa/pipe/cell/spu/spu_vertex_shader.h | 3 +- 2 files changed, 98 insertions(+), 65 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c index 3bbf9b7be4..45e3c26c00 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c @@ -59,8 +59,59 @@ #define DRAW_DBG 0 +static const qword fetch_shuffle_data[] = { + /* Shuffle used by CVT_64_FLOAT + */ + { + 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + }, + + /* Shuffle used by CVT_8_USCALED and CVT_8_SSCALED + */ + { + 0x00, 0x80, 0x80, 0x80, 0x01, 0x80, 0x80, 0x80, + 0x02, 0x80, 0x80, 0x80, 0x03, 0x80, 0x80, 0x80, + }, + + /* Shuffle used by CVT_16_USCALED and CVT_16_SSCALED + */ + { + 0x00, 0x01, 0x80, 0x80, 0x02, 0x03, 0x80, 0x80, + 0x04, 0x05, 0x80, 0x80, 0x06, 0x07, 0x80, 0x80, + }, + + /* High value shuffle used by trans4x4. + */ + { + 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, + 0x04, 0x05, 0x06, 0x07, 0x14, 0x15, 0x16, 0x17 + }, + + /* Low value shuffle used by trans4x4. + */ + { + 0x08, 0x09, 0x0A, 0x0B, 0x18, 0x19, 0x1A, 0x1B, + 0x0C, 0x0D, 0x0E, 0x0F, 0x1C, 0x1D, 0x1E, 0x1F + } +}; + + +static INLINE void +trans4x4(qword row0, qword row1, qword row2, qword row3, qword *out, + const qword *shuffle) +{ + qword t1 = si_shufb(row0, row2, shuffle[3]); + qword t2 = si_shufb(row0, row2, shuffle[4]); + qword t3 = si_shufb(row1, row3, shuffle[3]); + qword t4 = si_shufb(row1, row3, shuffle[4]); + + out[0] = si_shufb(t1, t3, shuffle[3]); + out[1] = si_shufb(t1, t3, shuffle[4]); + out[2] = si_shufb(t2, t4, shuffle[3]); + out[3] = si_shufb(t2, t4, shuffle[4]); +} -static const vec_float4 defaults = { 0.0, 0.0, 0.0, 1.0 }; /** * Fetch between 1 and 32 bytes from an unaligned address @@ -100,140 +151,117 @@ fetch_unaligned(qword *dst, unsigned ea, unsigned size) } -#define CVT_32_FLOAT(q) (*(q)) +#define CVT_32_FLOAT(q, s) (*(q)) static INLINE qword -CVT_64_FLOAT(const qword *qw) +CVT_64_FLOAT(const qword *qw, const qword *shuffle) { - qword shuf_first = (qword) { - 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - }; - qword a = si_frds(qw[0]); qword b = si_frds(si_rotqbyi(qw[0], 8)); qword c = si_frds(qw[1]); qword d = si_frds(si_rotqbyi(qw[1], 8)); - qword ab = si_shufb(a, b, shuf_first); - qword cd = si_shufb(c, d, si_rotqbyi(shuf_first, 8)); + qword ab = si_shufb(a, b, shuffle[0]); + qword cd = si_shufb(c, d, si_rotqbyi(shuffle[0], 8)); return si_or(ab, cd); } static INLINE qword -CVT_8_USCALED(const qword *qw) +CVT_8_USCALED(const qword *qw, const qword *shuffle) { - qword shuffle = (qword) { - 0x00, 0x80, 0x80, 0x80, 0x01, 0x80, 0x80, 0x80, - 0x02, 0x80, 0x80, 0x80, 0x03, 0x80, 0x80, 0x80, - }; - - return si_cuflt(si_shufb(*qw, *qw, shuffle), 0); + return si_cuflt(si_shufb(*qw, *qw, shuffle[1]), 0); } static INLINE qword -CVT_16_USCALED(const qword *qw) +CVT_16_USCALED(const qword *qw, const qword *shuffle) { - qword shuffle = (qword) { - 0x00, 0x01, 0x80, 0x80, 0x02, 0x03, 0x80, 0x80, - 0x04, 0x05, 0x80, 0x80, 0x06, 0x07, 0x80, 0x80, - }; - - return si_cuflt(si_shufb(*qw, *qw, shuffle), 0); + return si_cuflt(si_shufb(*qw, *qw, shuffle[2]), 0); } static INLINE qword -CVT_32_USCALED(const qword *qw) +CVT_32_USCALED(const qword *qw, const qword *shuffle) { + (void) shuffle; return si_cuflt(*qw, 0); } static INLINE qword -CVT_8_SSCALED(const qword *qw) +CVT_8_SSCALED(const qword *qw, const qword *shuffle) { - qword shuffle = (qword) { - 0x00, 0x80, 0x80, 0x80, 0x01, 0x80, 0x80, 0x80, - 0x02, 0x80, 0x80, 0x80, 0x03, 0x80, 0x80, 0x80, - }; - - return si_csflt(si_shufb(*qw, *qw, shuffle), 0); + return si_csflt(si_shufb(*qw, *qw, shuffle[1]), 0); } static INLINE qword -CVT_16_SSCALED(const qword *qw) +CVT_16_SSCALED(const qword *qw, const qword *shuffle) { - qword shuffle = (qword) { - 0x00, 0x01, 0x80, 0x80, 0x02, 0x03, 0x80, 0x80, - 0x04, 0x05, 0x80, 0x80, 0x06, 0x07, 0x80, 0x80, - }; - - return si_csflt(si_shufb(*qw, *qw, shuffle), 0); + return si_csflt(si_shufb(*qw, *qw, shuffle[2]), 0); } static INLINE qword -CVT_32_SSCALED(const qword *qw) +CVT_32_SSCALED(const qword *qw, const qword *shuffle) { + (void) shuffle; return si_csflt(*qw, 0); } static INLINE qword -CVT_8_UNORM(const qword *qw) +CVT_8_UNORM(const qword *qw, const qword *shuffle) { const qword scale = (qword) spu_splats(1.0f / 255.0f); - return si_fm(CVT_8_USCALED(qw), scale); + return si_fm(CVT_8_USCALED(qw, shuffle), scale); } static INLINE qword -CVT_16_UNORM(const qword *qw) +CVT_16_UNORM(const qword *qw, const qword *shuffle) { const qword scale = (qword) spu_splats(1.0f / 65535.0f); - return si_fm(CVT_16_USCALED(qw), scale); + return si_fm(CVT_16_USCALED(qw, shuffle), scale); } static INLINE qword -CVT_32_UNORM(const qword *qw) +CVT_32_UNORM(const qword *qw, const qword *shuffle) { const qword scale = (qword) spu_splats(1.0f / 4294967295.0f); - return si_fm(CVT_32_USCALED(qw), scale); + return si_fm(CVT_32_USCALED(qw, shuffle), scale); } static INLINE qword -CVT_8_SNORM(const qword *qw) +CVT_8_SNORM(const qword *qw, const qword *shuffle) { const qword scale = (qword) spu_splats(1.0f / 127.0f); - return si_fm(CVT_8_SSCALED(qw), scale); + return si_fm(CVT_8_SSCALED(qw, shuffle), scale); } static INLINE qword -CVT_16_SNORM(const qword *qw) +CVT_16_SNORM(const qword *qw, const qword *shuffle) { const qword scale = (qword) spu_splats(1.0f / 32767.0f); - return si_fm(CVT_16_SSCALED(qw), scale); + return si_fm(CVT_16_SSCALED(qw, shuffle), scale); } static INLINE qword -CVT_32_SNORM(const qword *qw) +CVT_32_SNORM(const qword *qw, const qword *shuffle) { const qword scale = (qword) spu_splats(1.0f / 2147483647.0f); - return si_fm(CVT_32_SSCALED(qw), scale); + return si_fm(CVT_32_SSCALED(qw, shuffle), scale); } #define SZ_4 si_il(0U) -#define SZ_3 si_rotqmbyi(si_il(~0), -12) -#define SZ_2 si_rotqmbyi(si_il(~0), -8) -#define SZ_1 si_rotqmbyi(si_il(~0), -4) +#define SZ_3 si_fsmbi(0x000f) +#define SZ_2 si_fsmbi(0x00ff) +#define SZ_1 si_fsmbi(0x0fff) /** * Fetch a float[4] vertex attribute from memory, doing format/type @@ -244,17 +272,19 @@ CVT_32_SNORM(const qword *qw) */ #define FETCH_ATTRIB( NAME, SZ, CVT, N ) \ static void \ -fetch_##NAME(qword *out, const qword *in) \ +fetch_##NAME(qword *out, const qword *in, qword defaults, \ + const qword *shuffle) \ { \ qword tmp[4]; \ \ - tmp[0] = si_selb(CVT(in + (0 * N)), (qword) defaults, SZ); \ - tmp[1] = si_selb(CVT(in + (1 * N)), (qword) defaults, SZ); \ - tmp[2] = si_selb(CVT(in + (2 * N)), (qword) defaults, SZ); \ - tmp[3] = si_selb(CVT(in + (3 * N)), (qword) defaults, SZ); \ - _transpose_matrix4x4((vec_float4 *) out, (vec_float4 *) tmp); \ + tmp[0] = si_selb(CVT(in + (0 * N), shuffle), defaults, SZ); \ + tmp[1] = si_selb(CVT(in + (1 * N), shuffle), defaults, SZ); \ + tmp[2] = si_selb(CVT(in + (2 * N), shuffle), defaults, SZ); \ + tmp[3] = si_selb(CVT(in + (3 * N), shuffle), defaults, SZ); \ + trans4x4(tmp[0], tmp[1], tmp[2], tmp[3], out, shuffle); \ } + FETCH_ATTRIB( R64G64B64A64_FLOAT, SZ_4, CVT_64_FLOAT, 2 ) FETCH_ATTRIB( R64G64B64_FLOAT, SZ_3, CVT_64_FLOAT, 2 ) FETCH_ATTRIB( R64G64_FLOAT, SZ_2, CVT_64_FLOAT, 2 ) @@ -582,6 +612,7 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, /* loop over vertex attributes (vertex shader inputs) */ for (attr = 0; attr < nr_attrs; attr++) { + const qword default_values = (qword)(vec_float4){ 0.0, 0.0, 0.0, 1.0 }; const unsigned pitch = draw->vertex_fetch.pitch[attr]; const uint64_t src = draw->vertex_fetch.src_ptr[attr]; const spu_fetch_func fetch = draw->vertex_fetch.fetch[attr]; @@ -602,8 +633,8 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, printf("SPU: fetching = 0x%llx\n", addr); #endif - fetch_unaligned(& in[idx], addr, bytes_per_entry); - idx += quads_per_entry; + fetch_unaligned(& in[idx], addr, bytes_per_entry); + idx += quads_per_entry; } /* Be nice and zero out any missing vertices. @@ -613,7 +644,8 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, /* Convert all 4 vertices to vectors of float. */ - (*fetch)(&machine->Inputs[attr].xyzw[0].q, in); + (*fetch)(&machine->Inputs[attr].xyzw[0].q, in, default_values, + fetch_shuffle_data); } } diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.h b/src/mesa/pipe/cell/spu/spu_vertex_shader.h index 8b37a239a4..b5bf31e67d 100644 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.h +++ b/src/mesa/pipe/cell/spu/spu_vertex_shader.h @@ -6,7 +6,8 @@ struct spu_vs_context; -typedef void (*spu_fetch_func)(qword *out, const qword *in); +typedef void (*spu_fetch_func)(qword *out, const qword *in, qword defaults, + const qword *shuffle_data); typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw, struct spu_exec_machine *machine, const unsigned *elts, -- cgit v1.2.3 From d4d9943b0144062e70372329ea46f5b41c6c270e Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 14 Feb 2008 17:06:13 -0800 Subject: s/spu_/spe_/g Ugh. --- src/mesa/ppc/rtasm/spe_asm.h | 198 +++++++++++++++++++++---------------------- 1 file changed, 99 insertions(+), 99 deletions(-) (limited to 'src') diff --git a/src/mesa/ppc/rtasm/spe_asm.h b/src/mesa/ppc/rtasm/spe_asm.h index 1c7f145415..e6cf1d35ba 100644 --- a/src/mesa/ppc/rtasm/spe_asm.h +++ b/src/mesa/ppc/rtasm/spe_asm.h @@ -75,32 +75,32 @@ extern void spe_release_func(struct spe_function *p); /* Memory load / store instructions */ -EMIT_RI10(spu_ldq, 0x034); -EMIT_RR (spu_lqx, 0x1c4); -EMIT_RI16(spu_lqa, 0x061); -EMIT_RI16(spu_lqr, 0x067); -EMIT_RI10(spu_stqd, 0x024); -EMIT_RR (spu_stqx, 0x144); -EMIT_RI16(spu_stqa, 0x041); -EMIT_RI16(spu_stqr, 0x047); -EMIT_RI7 (spu_cbd, 0x1f4); -EMIT_RR (spu_cbx, 0x1d4); -EMIT_RI7 (spu_chd, 0x1f5); -EMIT_RI7 (spu_chx, 0x1d5); -EMIT_RI7 (spu_cwd, 0x1f6); -EMIT_RI7 (spu_cwx, 0x1d6); -EMIT_RI7 (spu_cdd, 0x1f7); -EMIT_RI7 (spu_cdx, 0x1d7); +EMIT_RI10(spe_ldq, 0x034); +EMIT_RR (spe_lqx, 0x1c4); +EMIT_RI16(spe_lqa, 0x061); +EMIT_RI16(spe_lqr, 0x067); +EMIT_RI10(spe_stqd, 0x024); +EMIT_RR (spe_stqx, 0x144); +EMIT_RI16(spe_stqa, 0x041); +EMIT_RI16(spe_stqr, 0x047); +EMIT_RI7 (spe_cbd, 0x1f4); +EMIT_RR (spe_cbx, 0x1d4); +EMIT_RI7 (spe_chd, 0x1f5); +EMIT_RI7 (spe_chx, 0x1d5); +EMIT_RI7 (spe_cwd, 0x1f6); +EMIT_RI7 (spe_cwx, 0x1d6); +EMIT_RI7 (spe_cdd, 0x1f7); +EMIT_RI7 (spe_cdx, 0x1d7); /* Constant formation instructions */ -EMIT_RI16(spu_ilh, 0x083); -EMIT_RI16(spu_ilhu, 0x082); -EMIT_RI16(spu_il, 0x081); -EMIT_RI18(spu_ila, 0x021); -EMIT_RI16(spu_iohl, 0x0c1); -EMIT_RI16(spu_fsmbi, 0x0c5); +EMIT_RI16(spe_ilh, 0x083); +EMIT_RI16(spe_ilhu, 0x082); +EMIT_RI16(spe_il, 0x081); +EMIT_RI18(spe_ila, 0x021); +EMIT_RI16(spe_iohl, 0x0c1); +EMIT_RI16(spe_fsmbi, 0x0c5); /* Integer and logical instructions @@ -114,22 +114,22 @@ EMIT_RI10(spe_sfhi, 0x00d); EMIT_RR (spe_sf, 0x040); EMIT_RI10(spe_sfi, 0x00c); EMIT_RR (spe_addx, 0x340); -EMIT_RR (spu_cg, 0x0c2); -EMIT_RR (spu_cgx, 0x342); +EMIT_RR (spe_cg, 0x0c2); +EMIT_RR (spe_cgx, 0x342); EMIT_RR (spe_sfx, 0x341); -EMIT_RR (spu_bg, 0x042); -EMIT_RR (spu_bgx, 0x343); -EMIT_RR (spu_mpy, 0x3c4); -EMIT_RR (spu_mpyu, 0x3cc); -EMIT_RI10(spu_mpyi, 0x074); -EMIT_RI10(spu_mpyui, 0x075); +EMIT_RR (spe_bg, 0x042); +EMIT_RR (spe_bgx, 0x343); +EMIT_RR (spe_mpy, 0x3c4); +EMIT_RR (spe_mpyu, 0x3cc); +EMIT_RI10(spe_mpyi, 0x074); +EMIT_RI10(spe_mpyui, 0x075); EMIT_RRR (spy_mpya, 0x00c); -EMIT_RR (spu_mpyh, 0x3c5); -EMIT_RR (spu_mpys, 0x3c7); -EMIT_RR (spu_mpyhh, 0x3c6); -EMIT_RR (spu_mpyhha, 0x346); -EMIT_RR (spu_mpyhhu, 0x3ce); -EMIT_RR (spu_mpyhhau, 0x34e); +EMIT_RR (spe_mpyh, 0x3c5); +EMIT_RR (spe_mpys, 0x3c7); +EMIT_RR (spe_mpyhh, 0x3c6); +EMIT_RR (spe_mpyhha, 0x346); +EMIT_RR (spe_mpyhhu, 0x3ce); +EMIT_RR (spe_mpyhhau, 0x34e); EMIT_R (spe_clz, 0x2a5); EMIT_R (spe_cntb, 0x2b4); EMIT_R (spe_fsmb, 0x1b6); @@ -146,24 +146,24 @@ EMIT_R (spe_xshw, 0x2ae); EMIT_R (spe_xswd, 0x2a6); EMIT_RR (spe_and, 0x0c1); EMIT_RR (spe_andc, 0x2c1); -EMIT_RI10(spu_andbi, 0x016); -EMIT_RI10(spu_andhi, 0x015); -EMIT_RI10(spu_andi, 0x014); +EMIT_RI10(spe_andbi, 0x016); +EMIT_RI10(spe_andhi, 0x015); +EMIT_RI10(spe_andi, 0x014); EMIT_RR (spe_or, 0x041); EMIT_RR (spe_orc, 0x2c9); -EMIT_RI10(spu_orbi, 0x006); -EMIT_RI10(spu_orhi, 0x005); -EMIT_RI10(spu_ori, 0x004); -EMIT_R (spu_orx, 0x1f0); -EMIT_RR (spu_xor, 0x241); -EMIT_RI10(spu_xorbi, 0x026); -EMIT_RI10(spu_xorhi, 0x025); -EMIT_RI10(spu_xori, 0x024); +EMIT_RI10(spe_orbi, 0x006); +EMIT_RI10(spe_orhi, 0x005); +EMIT_RI10(spe_ori, 0x004); +EMIT_R (spe_orx, 0x1f0); +EMIT_RR (spe_xor, 0x241); +EMIT_RI10(spe_xorbi, 0x026); +EMIT_RI10(spe_xorhi, 0x025); +EMIT_RI10(spe_xori, 0x024); EMIT_RR (spe_nand, 0x0c9); EMIT_RR (spe_nor, 0x049); EMIT_RR (spe_eqv, 0x249); -EMIT_RRR (spy_selb, 0x008); -EMIT_RRR (spy_shufb, 0x00b); +EMIT_RRR (spe_selb, 0x008); +EMIT_RRR (spe_shufb, 0x00b); /* Shift and rotate instructions @@ -229,71 +229,71 @@ EMIT_RR (spe_clgt, 0x2c0); EMIT_RI10(spe_clgti, 0x05c); EMIT_I16 (spe_br, 0x064); EMIT_I16 (spe_bra, 0x060); -EMIT_RI16(spu_brsl, 0x066); -EMIT_RI16(spu_brasl, 0x062); -EMIT_RI16(spu_brnz, 0x042); -EMIT_RI16(spu_brz, 0x040); -EMIT_RI16(spu_brhnz, 0x046); -EMIT_RI16(spu_brhz, 0x044); +EMIT_RI16(spe_brsl, 0x066); +EMIT_RI16(spe_brasl, 0x062); +EMIT_RI16(spe_brnz, 0x042); +EMIT_RI16(spe_brz, 0x040); +EMIT_RI16(spe_brhnz, 0x046); +EMIT_RI16(spe_brhz, 0x044); -extern void spu_bi(struct spe_function *p, unsigned rA, int d, int e); -extern void spu_iret(struct spe_function *p, unsigned rA, int d, int e); -extern void spu_bisled(struct spe_function *p, unsigned rT, unsigned rA, +extern void spe_bi(struct spe_function *p, unsigned rA, int d, int e); +extern void spe_iret(struct spe_function *p, unsigned rA, int d, int e); +extern void spe_bisled(struct spe_function *p, unsigned rT, unsigned rA, int d, int e); -extern void spu_bisl(struct spe_function *p, unsigned rT, unsigned rA, +extern void spe_bisl(struct spe_function *p, unsigned rT, unsigned rA, int d, int e); -extern void spu_biz(struct spe_function *p, unsigned rT, unsigned rA, +extern void spe_biz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e); -extern void spu_binz(struct spe_function *p, unsigned rT, unsigned rA, +extern void spe_binz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e); -extern void spu_bihz(struct spe_function *p, unsigned rT, unsigned rA, +extern void spe_bihz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e); -extern void spu_bihnz(struct spe_function *p, unsigned rT, unsigned rA, +extern void spe_bihnz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e); /* Floating-point instructions */ -EMIT_RR (spu_fa, 0x2c4); -EMIT_RR (spu_dfa, 0x2cc); -EMIT_RR (spu_fs, 0x2c5); -EMIT_RR (spu_dfs, 0x2cd); -EMIT_RR (spu_fm, 0x2c6); -EMIT_RR (spu_dfm, 0x2ce); -EMIT_RRR (spu_fma, 0x00e); -EMIT_RR (spu_dfma, 0x35c); -EMIT_RRR (spu_fnms, 0x00d); -EMIT_RR (spu_dfnms, 0x35e); -EMIT_RRR (spu_fms, 0x00f); -EMIT_RR (spu_dfms, 0x35d); -EMIT_RR (spu_dfnma, 0x35f); -EMIT_R (spu_frest, 0x1b8); -EMIT_R (spu_frsqest, 0x1b9); -EMIT_RR (spu_fi, 0x3d4); -EMIT_RI7 (spu_csflt, 0x3da); -EMIT_RI7 (spu_cflts, 0x3d8); -EMIT_RI7 (spu_cuflt, 0x3db); -EMIT_RI7 (spu_cfltu, 0x3d9); -EMIT_R (spu_frds, 0x3b9); -EMIT_R (spu_fesd, 0x3b8); -EMIT_RR (spu_dfceq, 0x3c3); -EMIT_RR (spu_dfcmeq, 0x3cb); -EMIT_RR (spu_dfcgt, 0x2c3); -EMIT_RR (spu_dfcmgt, 0x2cb); -EMIT_RI7 (spu_dftsv, 0x3bf); -EMIT_RR (spu_fceq, 0x3c2); -EMIT_RR (spu_fcmeq, 0x3ca); -EMIT_RR (spu_fcgt, 0x2c2); -EMIT_RR (spu_fcmgt, 0x2ca); -EMIT_R (spu_fscrwr, 0x3ba); -EMIT_ (spu_fscrrd, 0x398); +EMIT_RR (spe_fa, 0x2c4); +EMIT_RR (spe_dfa, 0x2cc); +EMIT_RR (spe_fs, 0x2c5); +EMIT_RR (spe_dfs, 0x2cd); +EMIT_RR (spe_fm, 0x2c6); +EMIT_RR (spe_dfm, 0x2ce); +EMIT_RRR (spe_fma, 0x00e); +EMIT_RR (spe_dfma, 0x35c); +EMIT_RRR (spe_fnms, 0x00d); +EMIT_RR (spe_dfnms, 0x35e); +EMIT_RRR (spe_fms, 0x00f); +EMIT_RR (spe_dfms, 0x35d); +EMIT_RR (spe_dfnma, 0x35f); +EMIT_R (spe_frest, 0x1b8); +EMIT_R (spe_frsqest, 0x1b9); +EMIT_RR (spe_fi, 0x3d4); +EMIT_RI7 (spe_csflt, 0x3da); +EMIT_RI7 (spe_cflts, 0x3d8); +EMIT_RI7 (spe_cuflt, 0x3db); +EMIT_RI7 (spe_cfltu, 0x3d9); +EMIT_R (spe_frds, 0x3b9); +EMIT_R (spe_fesd, 0x3b8); +EMIT_RR (spe_dfceq, 0x3c3); +EMIT_RR (spe_dfcmeq, 0x3cb); +EMIT_RR (spe_dfcgt, 0x2c3); +EMIT_RR (spe_dfcmgt, 0x2cb); +EMIT_RI7 (spe_dftsv, 0x3bf); +EMIT_RR (spe_fceq, 0x3c2); +EMIT_RR (spe_fcmeq, 0x3ca); +EMIT_RR (spe_fcgt, 0x2c2); +EMIT_RR (spe_fcmgt, 0x2ca); +EMIT_R (spe_fscrwr, 0x3ba); +EMIT_ (spe_fscrrd, 0x398); /* Channel instructions */ -EMIT_R (spu_rdch, 0x00d); -EMIT_R (spu_rdchcnt, 0x00f); -EMIT_R (spu_wrch, 0x10d); +EMIT_R (spe_rdch, 0x00d); +EMIT_R (spe_rdchcnt, 0x00f); +EMIT_R (spe_wrch, 0x10d); #ifdef UNDEF_EMIT_MACROS -- cgit v1.2.3 From bbd7aabe2bc5743d11c77dd94665045b09791048 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 14 Feb 2008 17:42:34 -0800 Subject: More name typeo fixes. --- src/mesa/ppc/rtasm/spe_asm.c | 28 ++++++++++++++-------------- src/mesa/ppc/rtasm/spe_asm.h | 4 ++-- 2 files changed, 16 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/ppc/rtasm/spe_asm.c b/src/mesa/ppc/rtasm/spe_asm.c index b1851f05e7..f8aff9050b 100644 --- a/src/mesa/ppc/rtasm/spe_asm.c +++ b/src/mesa/ppc/rtasm/spe_asm.c @@ -282,45 +282,45 @@ void spe_release_func(struct spe_function *p) } -void spu_bi(struct spe_function *p, unsigned rA, int d, int e) +void spe_bi(struct spe_function *p, unsigned rA, int d, int e) { emit_RI7(p, 0x1a8, 0, rA, (d << 5) | (e << 4)); } -void spu_iret(struct spe_function *p, unsigned rA, int d, int e) +void spe_iret(struct spe_function *p, unsigned rA, int d, int e) { emit_RI7(p, 0x1aa, 0, rA, (d << 5) | (e << 4)); } -void spu_bisled(struct spe_function *p, unsigned rT, unsigned rA, int d, +void spe_bisled(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) { emit_RI7(p, 0x1ab, rT, rA, (d << 5) | (e << 4)); } -void spu_bisl(struct spe_function *p, unsigned rT, unsigned rA, int d, +void spe_bisl(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) { emit_RI7(p, 0x1a9, rT, rA, (d << 5) | (e << 4)); } -void spu_biz(struct spe_function *p, unsigned rT, unsigned rA, int d, +void spe_biz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) { emit_RI7(p, 0x128, rT, rA, (d << 5) | (e << 4)); } -void spu_binz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) +void spe_binz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) { emit_RI7(p, 0x129, rT, rA, (d << 5) | (e << 4)); } -void spu_bihz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) +void spe_bihz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) { emit_RI7(p, 0x12a, rT, rA, (d << 5) | (e << 4)); } -void spu_bihnz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) +void spe_bihnz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) { emit_RI7(p, 0x12b, rT, rA, (d << 5) | (e << 4)); } @@ -339,13 +339,13 @@ hbrr; */ #if 0 stop; -EMIT_RR (spu_stopd, 0x140); -EMIT_ (spu_lnop, 0x001); -EMIT_ (spu_nop, 0x201); +EMIT_RR (spe_stopd, 0x140); +EMIT_ (spe_lnop, 0x001); +EMIT_ (spe_nop, 0x201); sync; -EMIT_ (spu_dsync, 0x003); -EMIT_R (spu_mfspr, 0x00c); -EMIT_R (spu_mtspr, 0x10c); +EMIT_ (spe_dsync, 0x003); +EMIT_R (spe_mfspr, 0x00c); +EMIT_R (spe_mtspr, 0x10c); #endif #endif /* GALLIUM_CELL */ diff --git a/src/mesa/ppc/rtasm/spe_asm.h b/src/mesa/ppc/rtasm/spe_asm.h index e6cf1d35ba..9532669b8c 100644 --- a/src/mesa/ppc/rtasm/spe_asm.h +++ b/src/mesa/ppc/rtasm/spe_asm.h @@ -75,7 +75,7 @@ extern void spe_release_func(struct spe_function *p); /* Memory load / store instructions */ -EMIT_RI10(spe_ldq, 0x034); +EMIT_RI10(spe_lqd, 0x034); EMIT_RR (spe_lqx, 0x1c4); EMIT_RI16(spe_lqa, 0x061); EMIT_RI16(spe_lqr, 0x067); @@ -123,7 +123,7 @@ EMIT_RR (spe_mpy, 0x3c4); EMIT_RR (spe_mpyu, 0x3cc); EMIT_RI10(spe_mpyi, 0x074); EMIT_RI10(spe_mpyui, 0x075); -EMIT_RRR (spy_mpya, 0x00c); +EMIT_RRR (spe_mpya, 0x00c); EMIT_RR (spe_mpyh, 0x3c5); EMIT_RR (spe_mpys, 0x3c7); EMIT_RR (spe_mpyhh, 0x3c6); -- cgit v1.2.3 From e9c6c31651a0c1884633168adfd3ea6797fbdc60 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 13 Feb 2008 08:20:51 -0700 Subject: galllium: comments, minor clean-ups --- src/mesa/pipe/draw/draw_vertex.h | 7 ++++++- src/mesa/pipe/draw/draw_vf.c | 2 +- src/mesa/pipe/draw/draw_vf.h | 17 +++++++++++++++-- 3 files changed, 22 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h index dfc637b19b..267c74203b 100644 --- a/src/mesa/pipe/draw/draw_vertex.h +++ b/src/mesa/pipe/draw/draw_vertex.h @@ -25,7 +25,12 @@ * **************************************************************************/ -/* Author: +/** + * Post-transform vertex format info. The vertex_info struct is used by + * the draw_vbuf code to emit hardware-specific vertex layouts into hw + * vertex buffers. + * + * Author: * Brian Paul */ diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c index f23d7fcec5..dc3a5ecd21 100644 --- a/src/mesa/pipe/draw/draw_vf.c +++ b/src/mesa/pipe/draw/draw_vf.c @@ -370,7 +370,7 @@ void draw_vf_emit_vertex( struct draw_vertex_fetch *vf, unsigned j; for (j = 0; j < vf->attr_count; j++) { - if(!a[j].isconst) { + if (!a[j].isconst) { a[j].inputptr = (uint8_t *)&vertex->data[a[j].attrib][0]; a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ } diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h index e694b98675..011c8f0ff1 100644 --- a/src/mesa/pipe/draw/draw_vf.h +++ b/src/mesa/pipe/draw/draw_vf.h @@ -1,5 +1,5 @@ /* - * Copyright 2003 Tungsten Graphics, inc. + * Copyright 2008 Tungsten Graphics, inc. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -20,6 +20,17 @@ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * Vertex fetch/store/convert code. This functionality is used in two places: + * 1. Vertex fetch/convert - to grab vertex data from incoming vertex + * arrays and convert to format needed by vertex shaders. + * 2. Vertex store/emit - to convert simple float[][4] vertex attributes + * (which is the organization used throughout the draw/prim pipeline) to + * hardware-specific formats and emit into hardware vertex buffers. + * * * Authors: * Keith Whitwell @@ -33,7 +44,7 @@ #include "pipe/p_state.h" #include "draw_vertex.h" -#include "draw_private.h" // for vertex_header +#include "draw_private.h" /* for vertex_header */ enum draw_vf_attr_format { @@ -181,6 +192,7 @@ struct draw_vf_attr_type { unsigned offset; }; +/** XXX this could be moved into draw_vf.c */ struct draw_vf_fastpath { unsigned vertex_stride; unsigned attr_count; @@ -209,6 +221,7 @@ void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ); +/** XXX this type and function could probably be moved into draw_vf.c */ struct draw_vf_format_info { const char *name; draw_vf_insert_func insert[4]; -- cgit v1.2.3 From b08102a8f3ef558743f5f952c726ba2c28b6e82e Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 13 Feb 2008 10:25:38 -0700 Subject: gallium: rename draw_free_tmps->draw_free_temp_verts, draw_alloc_tmps->draw_alloc_temp_verts --- src/mesa/pipe/draw/draw_clip.c | 4 ++-- src/mesa/pipe/draw/draw_context.c | 4 ++-- src/mesa/pipe/draw/draw_cull.c | 4 ++-- src/mesa/pipe/draw/draw_flatshade.c | 4 ++-- src/mesa/pipe/draw/draw_offset.c | 4 ++-- src/mesa/pipe/draw/draw_private.h | 6 +++--- src/mesa/pipe/draw/draw_stipple.c | 2 +- src/mesa/pipe/draw/draw_twoside.c | 4 ++-- src/mesa/pipe/draw/draw_unfilled.c | 4 ++-- src/mesa/pipe/draw/draw_wide_prims.c | 4 ++-- 10 files changed, 20 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c index 61130c5600..e3051507ea 100644 --- a/src/mesa/pipe/draw/draw_clip.c +++ b/src/mesa/pipe/draw/draw_clip.c @@ -459,7 +459,7 @@ static void clip_reset_stipple_counter( struct draw_stage *stage ) static void clip_destroy( struct draw_stage *stage ) { - draw_free_tmps( stage ); + draw_free_temp_verts( stage ); FREE( stage ); } @@ -472,7 +472,7 @@ struct draw_stage *draw_clip_stage( struct draw_context *draw ) { struct clipper *clipper = CALLOC_STRUCT(clipper); - draw_alloc_tmps( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); + draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); clipper->stage.draw = draw; clipper->stage.point = clip_point; diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c index b15f57c824..4be3830316 100644 --- a/src/mesa/pipe/draw/draw_context.c +++ b/src/mesa/pipe/draw/draw_context.c @@ -242,7 +242,7 @@ draw_convert_wide_lines(struct draw_context *draw, boolean enable) /** * Allocate space for temporary post-transform vertices, such as for clipping. */ -void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) +void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) { assert(!stage->tmp); @@ -260,7 +260,7 @@ void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ) } -void draw_free_tmps( struct draw_stage *stage ) +void draw_free_temp_verts( struct draw_stage *stage ) { if (stage->tmp) { FREE( stage->tmp[0] ); diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c index 05c274e4dc..8177b0ac86 100644 --- a/src/mesa/pipe/draw/draw_cull.c +++ b/src/mesa/pipe/draw/draw_cull.c @@ -123,7 +123,7 @@ static void cull_reset_stipple_counter( struct draw_stage *stage ) static void cull_destroy( struct draw_stage *stage ) { - draw_free_tmps( stage ); + draw_free_temp_verts( stage ); FREE( stage ); } @@ -135,7 +135,7 @@ struct draw_stage *draw_cull_stage( struct draw_context *draw ) { struct cull_stage *cull = CALLOC_STRUCT(cull_stage); - draw_alloc_tmps( &cull->stage, 0 ); + draw_alloc_temp_verts( &cull->stage, 0 ); cull->stage.draw = draw; cull->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c index 1419f287d2..4398abbc60 100644 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ b/src/mesa/pipe/draw/draw_flatshade.c @@ -176,7 +176,7 @@ static void flatshade_reset_stipple_counter( struct draw_stage *stage ) static void flatshade_destroy( struct draw_stage *stage ) { - draw_free_tmps( stage ); + draw_free_temp_verts( stage ); FREE( stage ); } @@ -188,7 +188,7 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) { struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage); - draw_alloc_tmps( &flatshade->stage, 2 ); + draw_alloc_temp_verts( &flatshade->stage, 2 ); flatshade->stage.draw = draw; flatshade->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c index a2990ee8a8..dbc676deae 100644 --- a/src/mesa/pipe/draw/draw_offset.c +++ b/src/mesa/pipe/draw/draw_offset.c @@ -159,7 +159,7 @@ static void offset_reset_stipple_counter( struct draw_stage *stage ) static void offset_destroy( struct draw_stage *stage ) { - draw_free_tmps( stage ); + draw_free_temp_verts( stage ); FREE( stage ); } @@ -171,7 +171,7 @@ struct draw_stage *draw_offset_stage( struct draw_context *draw ) { struct offset_stage *offset = CALLOC_STRUCT(offset_stage); - draw_alloc_tmps( &offset->stage, 3 ); + draw_alloc_temp_verts( &offset->stage, 3 ); offset->stage.draw = draw; offset->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h index 7782db0477..b17eaaed65 100644 --- a/src/mesa/pipe/draw/draw_private.h +++ b/src/mesa/pipe/draw/draw_private.h @@ -276,9 +276,9 @@ extern struct draw_stage *draw_wide_stage( struct draw_context *context ); extern struct draw_stage *draw_validate_stage( struct draw_context *context ); -extern void draw_free_tmps( struct draw_stage *stage ); -extern void draw_reset_tmps( struct draw_stage *stage ); -extern void draw_alloc_tmps( struct draw_stage *stage, unsigned nr ); +extern void draw_free_temp_verts( struct draw_stage *stage ); + +extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); extern void draw_reset_vertex_ids( struct draw_context *draw ); diff --git a/src/mesa/pipe/draw/draw_stipple.c b/src/mesa/pipe/draw/draw_stipple.c index 9029101916..fb8b56e84c 100644 --- a/src/mesa/pipe/draw/draw_stipple.c +++ b/src/mesa/pipe/draw/draw_stipple.c @@ -223,7 +223,7 @@ struct draw_stage *draw_stipple_stage( struct draw_context *draw ) { struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); - draw_alloc_tmps( &stipple->stage, 2 ); + draw_alloc_temp_verts( &stipple->stage, 2 ); stipple->stage.draw = draw; stipple->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c index ad2aaf10bb..1c38957987 100644 --- a/src/mesa/pipe/draw/draw_twoside.c +++ b/src/mesa/pipe/draw/draw_twoside.c @@ -176,7 +176,7 @@ static void twoside_reset_stipple_counter( struct draw_stage *stage ) static void twoside_destroy( struct draw_stage *stage ) { - draw_free_tmps( stage ); + draw_free_temp_verts( stage ); FREE( stage ); } @@ -188,7 +188,7 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) { struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); - draw_alloc_tmps( &twoside->stage, 3 ); + draw_alloc_temp_verts( &twoside->stage, 3 ); twoside->stage.draw = draw; twoside->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index 364bda8b79..8bb9f3b558 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -176,7 +176,7 @@ static void unfilled_reset_stipple_counter( struct draw_stage *stage ) static void unfilled_destroy( struct draw_stage *stage ) { - draw_free_tmps( stage ); + draw_free_temp_verts( stage ); FREE( stage ); } @@ -188,7 +188,7 @@ struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) { struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); - draw_alloc_tmps( &unfilled->stage, 0 ); + draw_alloc_temp_verts( &unfilled->stage, 0 ); unfilled->stage.draw = draw; unfilled->stage.next = NULL; diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 4c7e279b20..163282b226 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -408,7 +408,7 @@ static void draw_reset_stipple_counter( struct draw_stage *stage ) static void wide_destroy( struct draw_stage *stage ) { - draw_free_tmps( stage ); + draw_free_temp_verts( stage ); FREE( stage ); } @@ -417,7 +417,7 @@ struct draw_stage *draw_wide_stage( struct draw_context *draw ) { struct wide_stage *wide = CALLOC_STRUCT(wide_stage); - draw_alloc_tmps( &wide->stage, 4 ); + draw_alloc_temp_verts( &wide->stage, 4 ); wide->stage.draw = draw; wide->stage.next = NULL; -- cgit v1.2.3 From 4f32c532376bc3394f8fce70f95156b49fcc4fec Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 14 Feb 2008 14:15:52 -0700 Subject: gallium: changes to polygon mode weren't detected in draw_unfilled stage. Need to reset stage->tri = unfilled_first_try in unfilled_flush() so that the front/back state is picked up. --- src/mesa/pipe/draw/draw_unfilled.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c index 8bb9f3b558..8777cfdfc8 100644 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ b/src/mesa/pipe/draw/draw_unfilled.c @@ -165,6 +165,8 @@ static void unfilled_flush( struct draw_stage *stage, unsigned flags ) { stage->next->flush( stage->next, flags ); + + stage->tri = unfilled_first_tri; } -- cgit v1.2.3 From 59cc1f4e62268cc73567d3e1cccbc5df9d116311 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 14 Feb 2008 14:20:51 -0700 Subject: gallium: replace "interpolate" terminology with "eval" to better reflect what's being done. --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 33 ++++++++++++++++++++++----------- 1 file changed, 22 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 336ae1c8b6..7801f95972 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -1346,9 +1346,12 @@ exec_tex(struct tgsi_exec_machine *mach, } - +/** + * Evaluate a constant-valued coefficient at the position of the + * current quad. + */ static void -constant_interpolation( +eval_constant_coef( struct tgsi_exec_machine *mach, unsigned attrib, unsigned chan ) @@ -1360,8 +1363,12 @@ constant_interpolation( } } +/** + * Evaluate a linear-valued coefficient at the position of the + * current quad. + */ static void -linear_interpolation( +eval_linear_coef( struct tgsi_exec_machine *mach, unsigned attrib, unsigned chan ) @@ -1377,8 +1384,12 @@ linear_interpolation( mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; } +/** + * Evaluate a perspective-valued coefficient at the position of the + * current quad. + */ static void -perspective_interpolation( +eval_perspective_coef( struct tgsi_exec_machine *mach, unsigned attrib, unsigned chan ) @@ -1397,7 +1408,7 @@ perspective_interpolation( } -typedef void (* interpolation_func)( +typedef void (* eval_coef_func)( struct tgsi_exec_machine *mach, unsigned attrib, unsigned chan ); @@ -1410,7 +1421,7 @@ exec_declaration( if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { if( decl->Declaration.File == TGSI_FILE_INPUT ) { unsigned first, last, mask; - interpolation_func interp; + eval_coef_func eval; assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); @@ -1420,15 +1431,15 @@ exec_declaration( switch( decl->Interpolation.Interpolate ) { case TGSI_INTERPOLATE_CONSTANT: - interp = constant_interpolation; + eval = eval_constant_coef; break; case TGSI_INTERPOLATE_LINEAR: - interp = linear_interpolation; + eval = eval_linear_coef; break; case TGSI_INTERPOLATE_PERSPECTIVE: - interp = perspective_interpolation; + eval = eval_perspective_coef; break; default: @@ -1440,7 +1451,7 @@ exec_declaration( for( i = first; i <= last; i++ ) { for( j = 0; j < NUM_CHANNELS; j++ ) { - interp( mach, i, j ); + eval( mach, i, j ); } } } @@ -1450,7 +1461,7 @@ exec_declaration( for( j = 0; j < NUM_CHANNELS; j++ ) { if( mask & (1 << j) ) { for( i = first; i <= last; i++ ) { - interp( mach, i, j ); + eval( mach, i, j ); } } } -- cgit v1.2.3 From 7a3e59d2363cba6dd2c1edc08d1afffd25b53ea3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 14 Feb 2008 14:22:22 -0700 Subject: gallium: fix some "instruction"/"declaration" mix-ups in tgsi_exec_prepare(). --- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 7801f95972..37e6007068 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -158,13 +158,13 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) if (numDeclarations == maxDeclarations) { declarations = REALLOC(declarations, maxDeclarations - * sizeof(struct tgsi_full_instruction), + * sizeof(struct tgsi_full_declaration), (maxDeclarations + 10) - * sizeof(struct tgsi_full_instruction)); + * sizeof(struct tgsi_full_declaration)); maxDeclarations += 10; } memcpy(declarations + numDeclarations, - &parse.FullToken.FullInstruction, + &parse.FullToken.FullDeclaration, sizeof(declarations[0])); numDeclarations++; break; -- cgit v1.2.3 From 5e345a653b0adc59487d786050abd01d4cb8b4ca Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 14 Feb 2008 14:48:35 -0700 Subject: gallium: call draw_flush() in softpipe_flush() Without this, we might not get any rendering at SwapBuffers time when using the vbuf path. --- src/mesa/pipe/softpipe/sp_flush.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index ec6bb4a0dc..fd3078d870 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -50,6 +50,8 @@ softpipe_flush( struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); uint i; + draw_flush(softpipe->draw); + /* - flush the quad pipeline * - flush the texture cache * - flush the render cache -- cgit v1.2.3 From 4c1403f667c6047a44ff494364725b3b7da82c68 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 14 Feb 2008 18:02:36 -0700 Subject: gallium: new tgsi_transform_shader() function Used to apply transformations to TGSI shaders, such as register search and replace. --- src/mesa/pipe/tgsi/util/tgsi_transform.c | 199 +++++++++++++++++++++++++++++++ src/mesa/pipe/tgsi/util/tgsi_transform.h | 93 +++++++++++++++ 2 files changed, 292 insertions(+) create mode 100644 src/mesa/pipe/tgsi/util/tgsi_transform.c create mode 100644 src/mesa/pipe/tgsi/util/tgsi_transform.h (limited to 'src') diff --git a/src/mesa/pipe/tgsi/util/tgsi_transform.c b/src/mesa/pipe/tgsi/util/tgsi_transform.c new file mode 100644 index 0000000000..357f77b05a --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_transform.c @@ -0,0 +1,199 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI program transformation utility. + * + * Authors: Brian Paul + */ + + +#include "tgsi_transform.h" + + + +static void +emit_instruction(struct tgsi_transform_context *ctx, + const struct tgsi_full_instruction *inst) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_instruction(inst, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + +static void +emit_declaration(struct tgsi_transform_context *ctx, + const struct tgsi_full_declaration *decl) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_declaration(decl, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + +static void +emit_immediate(struct tgsi_transform_context *ctx, + const struct tgsi_full_immediate *imm) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_immediate(imm, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + + +/** + * Apply user-defined transformations to the input shader to produce + * the output shader. + * For example, a register search-and-replace operation could be applied + * by defining a transform_instruction() callback that examined and changed + * the instruction src/dest regs. + * + * \return number of tokens emitted + */ +int +tgsi_transform_shader(const struct tgsi_token *tokens_in, + struct tgsi_token *tokens_out, + uint max_tokens_out, + struct tgsi_transform_context *ctx) +{ + uint procType; + + /* input shader */ + struct tgsi_parse_context parse; + + /* output shader */ + struct tgsi_processor *processor; + + + /** + ** callback context init + **/ + ctx->emit_instruction = emit_instruction; + ctx->emit_declaration = emit_declaration; + ctx->emit_immediate = emit_immediate; + ctx->tokens_out = tokens_out; + ctx->max_tokens_out = max_tokens_out; + + + /** + ** Setup to begin parsing input shader + **/ + if (tgsi_parse_init( &parse, tokens_in ) != TGSI_PARSE_OK) { + debug_printf("tgsi_parse_init() failed in tgsi_transform_shader()!\n"); + return -1; + } + procType = parse.FullHeader.Processor.Processor; + assert(procType == TGSI_PROCESSOR_FRAGMENT || + procType == TGSI_PROCESSOR_VERTEX || + procType == TGSI_PROCESSOR_GEOMETRY); + + + /** + ** Setup output shader + **/ + *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version(); + + ctx->header = (struct tgsi_header *) (tokens_out + 1); + *ctx->header = tgsi_build_header(); + + processor = (struct tgsi_processor *) (tokens_out + 2); + *processor = tgsi_build_processor( procType, ctx->header ); + + ctx->ti = 3; + + + /** + ** Loop over incoming program tokens/instructions + */ + while( !tgsi_parse_end_of_tokens( &parse ) ) { + + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *fullinst + = &parse.FullToken.FullInstruction; + + if (ctx->transform_instruction) + ctx->transform_instruction(ctx, fullinst); + else + ctx->emit_instruction(ctx, fullinst); + } + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *fulldecl + = &parse.FullToken.FullDeclaration; + + if (ctx->transform_declaration) + ctx->transform_declaration(ctx, fulldecl); + else + ctx->emit_declaration(ctx, fulldecl); + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *fullimm + = &parse.FullToken.FullImmediate; + + if (ctx->transform_immediate) + ctx->transform_immediate(ctx, fullimm); + else + ctx->emit_immediate(ctx, fullimm); + } + break; + + default: + assert( 0 ); + } + } + + if (ctx->epilog) { + ctx->epilog(ctx); + } + + tgsi_parse_free (&parse); + + return ctx->ti; +} diff --git a/src/mesa/pipe/tgsi/util/tgsi_transform.h b/src/mesa/pipe/tgsi/util/tgsi_transform.h new file mode 100644 index 0000000000..365d8c298c --- /dev/null +++ b/src/mesa/pipe/tgsi/util/tgsi_transform.h @@ -0,0 +1,93 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_TRANSFORM_H +#define TGSI_TRANSFORM_H + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/util/tgsi_build.h" + + + +/** + * Subclass this to add caller-specific data + */ +struct tgsi_transform_context +{ +/**** PUBLIC ***/ + + /** + * User-defined callbacks invoked per instruction. + */ + void (*transform_instruction)(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst); + + void (*transform_declaration)(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl); + + void (*transform_immediate)(struct tgsi_transform_context *ctx, + struct tgsi_full_immediate *imm); + + /** + * Called at end of input program to allow caller to append extra + * instructions. Return number of tokens emitted. + */ + void (*epilog)(struct tgsi_transform_context *ctx); + + +/*** PRIVATE ***/ + + /** + * These are setup by tgsi_transform_shader() and cannot be overridden. + * Meant to be called from in the above user callback functions. + */ + void (*emit_instruction)(struct tgsi_transform_context *ctx, + const struct tgsi_full_instruction *inst); + void (*emit_declaration)(struct tgsi_transform_context *ctx, + const struct tgsi_full_declaration *decl); + void (*emit_immediate)(struct tgsi_transform_context *ctx, + const struct tgsi_full_immediate *imm); + + struct tgsi_header *header; + uint max_tokens_out; + struct tgsi_token *tokens_out; + uint ti; +}; + + + +extern int +tgsi_transform_shader(const struct tgsi_token *tokens_in, + struct tgsi_token *tokens_out, + uint max_tokens_out, + struct tgsi_transform_context *ctx); + + +#endif /* TGSI_TRANSFORM_H */ -- cgit v1.2.3 From d2b14311d9da8d94056968173d6d5d11f1885d3b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 14 Feb 2008 18:03:38 -0700 Subject: gallium: minor function renaming --- src/mesa/pipe/draw/draw_wide_prims.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 163282b226..37b60a6da1 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -400,7 +400,7 @@ static void wide_flush( struct draw_stage *stage, unsigned flags ) } -static void draw_reset_stipple_counter( struct draw_stage *stage ) +static void wide_reset_stipple_counter( struct draw_stage *stage ) { stage->next->reset_stipple_counter( stage->next ); } @@ -425,7 +425,7 @@ struct draw_stage *draw_wide_stage( struct draw_context *draw ) wide->stage.line = wide_first_line; wide->stage.tri = passthrough_tri; wide->stage.flush = wide_flush; - wide->stage.reset_stipple_counter = draw_reset_stipple_counter; + wide->stage.reset_stipple_counter = wide_reset_stipple_counter; wide->stage.destroy = wide_destroy; return &wide->stage; -- cgit v1.2.3 From 1b6540b4b17969e9838facf5248fce34c9ff5c34 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 14 Feb 2008 18:53:18 -0700 Subject: gallium: include draw_context.h to silence warning --- src/mesa/pipe/softpipe/sp_flush.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c index fd3078d870..ced0d5d098 100644 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ b/src/mesa/pipe/softpipe/sp_flush.c @@ -31,6 +31,7 @@ #include "pipe/p_defines.h" +#include "pipe/draw/draw_context.h" #include "sp_flush.h" #include "sp_context.h" #include "sp_surface.h" -- cgit v1.2.3 From 3d81a956b9de709de17a98b93fead4d3307b2b99 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 14 Feb 2008 18:58:50 -0700 Subject: gallium: rearrange vertex info/layout validation Delay validation until someone really needs the vertex layout (vbuf alloc vertex buffer or point/line/tri setup/rendering). This will allow the vertex size to change depending on whether we're drawing points, lines or triangles. --- src/mesa/pipe/softpipe/sp_prim_setup.c | 21 ++-- src/mesa/pipe/softpipe/sp_prim_vbuf.c | 5 +- src/mesa/pipe/softpipe/sp_state.h | 10 ++ src/mesa/pipe/softpipe/sp_state_derived.c | 174 +++++++++++++++++++----------- 4 files changed, 134 insertions(+), 76 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c index 7478b2336b..2772048661 100644 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ b/src/mesa/pipe/softpipe/sp_prim_setup.c @@ -499,7 +499,7 @@ setup_fragcoord_coeff(struct setup_stage *setup) setup->coef[0].a0[2] = setup->posCoef.a0[2]; setup->coef[0].dadx[2] = setup->posCoef.dadx[2]; setup->coef[0].dady[2] = setup->posCoef.dady[2]; - /*w*/ + /*W*/ setup->coef[0].a0[3] = setup->posCoef.a0[3]; setup->coef[0].dadx[3] = setup->posCoef.dadx[3]; setup->coef[0].dady[3] = setup->posCoef.dady[3]; @@ -513,8 +513,9 @@ setup_fragcoord_coeff(struct setup_stage *setup) */ static void setup_tri_coefficients( struct setup_stage *setup ) { - const struct softpipe_context *softpipe = setup->softpipe; + struct softpipe_context *softpipe = setup->softpipe; const struct pipe_shader_state *fs = &softpipe->fs->shader; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); uint fragSlot; /* z and w are done by linear interpolation: @@ -525,10 +526,10 @@ static void setup_tri_coefficients( struct setup_stage *setup ) /* setup interpolation for all the remaining attributes: */ for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { - const uint vertSlot = softpipe->vertex_info.src_index[fragSlot]; + const uint vertSlot = vinfo->src_index[fragSlot]; uint j; - switch (softpipe->vertex_info.interp_mode[fragSlot]) { + switch (vinfo->interp_mode[fragSlot]) { case INTERP_CONSTANT: for (j = 0; j < NUM_CHANNELS; j++) const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); @@ -756,8 +757,9 @@ line_persp_coeff(struct setup_stage *setup, static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { - const struct softpipe_context *softpipe = setup->softpipe; + struct softpipe_context *softpipe = setup->softpipe; const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); uint fragSlot; /* use setup->vmin, vmax to point to vertices */ @@ -779,10 +781,10 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) /* setup interpolation for all the remaining attributes: */ for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { - const uint vertSlot = softpipe->vertex_info.src_index[fragSlot]; + const uint vertSlot = vinfo->src_index[fragSlot]; uint j; - switch (softpipe->vertex_info.interp_mode[fragSlot]) { + switch (vinfo->interp_mode[fragSlot]) { case INTERP_CONSTANT: for (j = 0; j < NUM_CHANNELS; j++) const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); @@ -978,6 +980,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth; const float x = v0->data[0][0]; /* Note: data[0] is always position */ const float y = v0->data[0][1]; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); uint fragSlot; /* For points, all interpolants are constant-valued. @@ -1003,10 +1006,10 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const_coeff(setup, &setup->posCoef, 0, 3); for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { - const uint vertSlot = softpipe->vertex_info.src_index[fragSlot]; + const uint vertSlot = vinfo->src_index[fragSlot]; uint j; - switch (softpipe->vertex_info.interp_mode[fragSlot]) { + switch (vinfo->interp_mode[fragSlot]) { case INTERP_CONSTANT: /* fall-through */ case INTERP_LINEAR: diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c index c9089e7eb2..7f71fdb6a9 100644 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ b/src/mesa/pipe/softpipe/sp_prim_vbuf.c @@ -37,6 +37,7 @@ #include "sp_context.h" +#include "sp_state.h" #include "sp_prim_vbuf.h" #include "pipe/draw/draw_context.h" #include "pipe/draw/draw_private.h" @@ -73,9 +74,7 @@ static const struct vertex_info * sp_vbuf_get_vertex_info(struct vbuf_render *vbr) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - /* XXX check for state changes? */ - assert(!cvbr->softpipe->dirty ); - return &cvbr->softpipe->vertex_info_vbuf; + return softpipe_get_vbuf_vertex_info(cvbr->softpipe); } diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index af955c1e17..b79db0d1f1 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -59,6 +59,8 @@ struct gallivm_prog; #endif +struct vertex_info; + /** Subclass of pipe_shader_state */ struct sp_fragment_shader_state { @@ -174,4 +176,12 @@ softpipe_map_texture_surfaces(struct softpipe_context *sp); void softpipe_unmap_texture_surfaces(struct softpipe_context *sp); + +struct vertex_info * +softpipe_get_vertex_info(struct softpipe_context *softpipe); + +struct vertex_info * +softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe); + + #endif diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 39c3e1afe1..9e4c35e696 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -49,82 +49,128 @@ find_vs_output(const struct pipe_shader_state *vs, } +/** + * Mark the current vertex layout as "invalid". + * We'll validate the vertex layout later, when we start to actually + * render a point or line or tri. + */ +static void +invalidate_vertex_layout(struct softpipe_context *softpipe) +{ + softpipe->vertex_info.num_attribs = 0; +} + /** - * Determine how to map vertex program outputs to fragment program inputs. - * Basically, this will be used when computing the triangle interpolation - * coefficients from the post-transform vertex attributes. + * The vertex info describes how to convert the post-transformed vertices + * (simple float[][4]) used by the 'draw' module into vertices for + * rasterization. + * + * This function validates the vertex layout and returns a pointer to a + * vertex_info object. */ -static void calculate_vertex_layout( struct softpipe_context *softpipe ) +struct vertex_info * +softpipe_get_vertex_info(struct softpipe_context *softpipe) { - const struct pipe_shader_state *vs = &softpipe->vs->shader; - const struct pipe_shader_state *fs = &softpipe->fs->shader; - const enum interp_mode colorInterp - = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &softpipe->vertex_info; - uint i; - if (softpipe->vbuf) { - /* if using the post-transform vertex buffer, tell draw_vbuf to - * simply emit the whole post-xform vertex as-is: - */ - struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; - vinfo_vbuf->num_attribs = 0; - draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); - vinfo_vbuf->size = 4 * vs->num_outputs + sizeof(struct vertex_header)/4; - } + if (vinfo->num_attribs == 0) { + /* compute vertex layout now */ + const struct pipe_shader_state *vs = &softpipe->vs->shader; + const struct pipe_shader_state *fs = &softpipe->fs->shader; + const enum interp_mode colorInterp + = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + uint i; + + if (softpipe->vbuf) { + /* if using the post-transform vertex buffer, tell draw_vbuf to + * simply emit the whole post-xform vertex as-is: + */ + struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; + vinfo_vbuf->num_attribs = 0; + draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); + vinfo_vbuf->size = 4 * vs->num_outputs + + sizeof(struct vertex_header) / 4; + } - /* - * Loop over fragment shader inputs, searching for the matching output - * from the vertex shader. - */ - vinfo->num_attribs = 0; - for (i = 0; i < fs->num_inputs; i++) { - int src; - switch (fs->input_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); - assert(src >= 0); - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); - break; - - case TGSI_SEMANTIC_COLOR: - src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, - fs->input_semantic_index[i]); - assert(src >= 0); - draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); - break; - - case TGSI_SEMANTIC_FOG: - src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); + /* + * Loop over fragment shader inputs, searching for the matching output + * from the vertex shader. + */ + vinfo->num_attribs = 0; + for (i = 0; i < fs->num_inputs; i++) { + int src; + switch (fs->input_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); + break; + + case TGSI_SEMANTIC_COLOR: + src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, + fs->input_semantic_index[i]); + if (src < 0) + src = 0; + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); + break; + + case TGSI_SEMANTIC_FOG: + src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); #if 1 - if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */ - src = 0; + if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */ + src = 0; #endif - assert(src >= 0); - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); - break; - - case TGSI_SEMANTIC_GENERIC: - /* this includes texcoords and varying vars */ - src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, - fs->input_semantic_index[i]); - assert(src >= 0); - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); - break; - - default: - assert(0); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); + break; + + case TGSI_SEMANTIC_GENERIC: + /* this includes texcoords and varying vars */ + src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, + fs->input_semantic_index[i]); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); + break; + + default: + assert(0); + } } - } - softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); - if (softpipe->psize_slot >= 0) { - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, - softpipe->psize_slot); + softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); + if (softpipe->psize_slot >= 0) { + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, + softpipe->psize_slot); + } + + draw_compute_vertex_size(vinfo); } - draw_compute_vertex_size(vinfo); + return vinfo; +} + + +/** + * Called from vbuf module. + * + * Note that there's actually two different vertex layouts in softpipe. + * + * The normal one is computed in softpipe_get_vertex_info() above and is + * used by the point/line/tri "setup" code. + * + * The other one (this one) is only used by the vbuf module (which is + * not normally used by default but used in testing). For the vbuf module, + * we basically want to pass-through the draw module's vertex layout as-is. + * When the softpipe vbuf code begins drawing, the normal vertex layout + * will come into play again. + */ +struct vertex_info * +softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe) +{ + (void) softpipe_get_vertex_info(softpipe); + return &softpipe->vertex_info_vbuf; } @@ -171,7 +217,7 @@ void softpipe_update_derived( struct softpipe_context *softpipe ) if (softpipe->dirty & (SP_NEW_RASTERIZER | SP_NEW_FS | SP_NEW_VS)) - calculate_vertex_layout( softpipe ); + invalidate_vertex_layout( softpipe ); if (softpipe->dirty & (SP_NEW_SCISSOR | SP_NEW_DEPTH_STENCIL_ALPHA | -- cgit v1.2.3 From 663f4aaae618a8f031fa1a6b5292ddc57698741c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 14 Feb 2008 19:18:09 -0700 Subject: gallium: remove some debug assertions in vertex format validation If a fragment shader references an input for which there's no vertex shader output (ex: texcoord3), use vertex output 0 by default. Basically, the attribute's value will be undefined. The shader writer should never rely on undefined fragment shader inputs anyway. --- src/mesa/pipe/softpipe/sp_state_derived.c | 23 +++++++++++------------ 1 file changed, 11 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c index 9e4c35e696..372597869f 100644 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ b/src/mesa/pipe/softpipe/sp_state_derived.c @@ -34,6 +34,15 @@ #include "sp_state.h" +/** + * Search vertex program's outputs to find a match for the given + * semantic name/index. Return the index of the output slot. + * + * Return 0 if not found. This will cause the fragment program to use + * vertex attrib 0 (position) in the cases where the fragment program + * attempts to use a missing vertex program output. This is an undefined + * condition that users shouldn't hit anyway. + */ static int find_vs_output(const struct pipe_shader_state *vs, uint semantic_name, @@ -45,7 +54,7 @@ find_vs_output(const struct pipe_shader_state *vs, vs->output_semantic_index[i] == semantic_index) return i; } - return -1; + return 0; } @@ -103,26 +112,17 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) switch (fs->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); - assert(src >= 0); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); break; case TGSI_SEMANTIC_COLOR: src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, fs->input_semantic_index[i]); - if (src < 0) - src = 0; - assert(src >= 0); draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); break; case TGSI_SEMANTIC_FOG: src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); -#if 1 - if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */ - src = 0; -#endif - assert(src >= 0); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; @@ -130,7 +130,6 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) /* this includes texcoords and varying vars */ src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, fs->input_semantic_index[i]); - assert(src >= 0); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; @@ -140,7 +139,7 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) } softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); - if (softpipe->psize_slot >= 0) { + if (softpipe->psize_slot > 0) { draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, softpipe->psize_slot); } -- cgit v1.2.3 From d8ae972fd067a2478d1b0b4a35fec47ac2c814f7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 12 Feb 2008 15:35:18 +0900 Subject: gallium: Fix MSVC compiler warnings. --- src/mesa/pipe/softpipe/sp_state_sampler.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c index 6a5a643c89..ea348c7e95 100644 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ b/src/mesa/pipe/softpipe/sp_state_sampler.c @@ -30,6 +30,10 @@ */ #include "pipe/p_util.h" + +#include "pipe/draw/draw_context.h" + +#include "sp_context.h" #include "sp_context.h" #include "sp_state.h" #include "sp_texture.h" -- cgit v1.2.3 From 3f1b712d0a47b440875e58429debd9a145192724 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 7 Feb 2008 10:26:08 +0100 Subject: gallium: Use MALLOC(). --- src/mesa/pipe/p_util.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h index a8938a7e43..d7da2801c9 100644 --- a/src/mesa/pipe/p_util.h +++ b/src/mesa/pipe/p_util.h @@ -185,12 +185,12 @@ align_free(void *ptr) /** - * Duplicate of a block of memory + * Duplicate a block of memory. */ static INLINE void * mem_dup(const void *src, uint size) { - void *dup = malloc(size); + void *dup = MALLOC(size); if (dup) memcpy(dup, src, size); return dup; -- cgit v1.2.3 From d3cd39493c1e776c1e21045920744a27de2787f1 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Feb 2008 17:51:41 +0100 Subject: gallium: Fix build on Windows. --- src/mesa/pipe/draw/draw_wide_prims.c | 2 +- src/mesa/pipe/i915simple/i915_surface.c | 2 +- src/mesa/pipe/softpipe/sp_surface.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c index 37b60a6da1..655774b155 100644 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ b/src/mesa/pipe/draw/draw_wide_prims.c @@ -175,7 +175,7 @@ static void wide_line_aa(struct draw_stage *stage, float *pos; float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; - const float len = sqrt(dx * dx + dy * dy); + const float len = (float) sqrt(dx * dx + dy * dy); uint i; dx = dx * half_width / len; diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c index 6d4b8a0aa9..de0cc5fe06 100644 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ b/src/mesa/pipe/i915simple/i915_surface.c @@ -99,7 +99,7 @@ i915_surface_copy(struct pipe_context *pipe, dstx, dsty, width, height, pipe_surface_map(src), - do_flip ? -src->pitch : src->pitch, + do_flip ? -(int) src->pitch : src->pitch, srcx, do_flip ? 1 - srcy - height : srcy); pipe_surface_unmap(src); diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c index 5c6ed3b8d9..8802ced187 100644 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ b/src/mesa/pipe/softpipe/sp_surface.c @@ -54,7 +54,7 @@ sp_surface_copy(struct pipe_context *pipe, dstx, dsty, width, height, pipe_surface_map(src), - do_flip ? -src->pitch : src->pitch, + do_flip ? -(int) src->pitch : src->pitch, srcx, do_flip ? 1 - srcy - height : srcy); pipe_surface_unmap(src); -- cgit v1.2.3 From 6a7820d31f9b697888ce9f9b193bacc86e646fe1 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 14 Feb 2008 17:36:47 +0100 Subject: gallium: Use align_free() to free aligned memories. --- src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c index c1b7759874..9e8244f909 100644 --- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c +++ b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c @@ -110,7 +110,7 @@ pb_malloc_buffer_create(size_t size, buf = CALLOC_STRUCT(malloc_buffer); if(!buf) return NULL; - + buf->base.base.refcount = 1; buf->base.base.alignment = desc->alignment; buf->base.base.usage = desc->usage; @@ -119,7 +119,7 @@ pb_malloc_buffer_create(size_t size, buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment); if(!buf->data) { - FREE(buf); + align_free(buf); return NULL; } -- cgit v1.2.3 From 742c5d3e1ba1455e6aca8454f4b7e146b27fbbe3 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 14 Feb 2008 20:48:40 +0100 Subject: gallium: Fix memory leak. --- src/mesa/pipe/draw/draw_stipple.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_stipple.c b/src/mesa/pipe/draw/draw_stipple.c index fb8b56e84c..7e0e2137b7 100644 --- a/src/mesa/pipe/draw/draw_stipple.c +++ b/src/mesa/pipe/draw/draw_stipple.c @@ -212,6 +212,7 @@ passthrough_tri(struct draw_stage *stage, struct prim_header *header) static void stipple_destroy( struct draw_stage *stage ) { + draw_free_tmps( stage ); FREE( stage ); } -- cgit v1.2.3 From 19780237ff0e6a89f31ecb9079781568bc2d3fdc Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 14 Feb 2008 21:03:12 +0100 Subject: gallium: Fix memory leak. --- src/mesa/pipe/draw/draw_vbuf.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c index be96c8fdeb..71ac73912b 100644 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ b/src/mesa/pipe/draw/draw_vbuf.c @@ -523,6 +523,9 @@ static void vbuf_destroy( struct draw_stage *stage ) if(vbuf->vf) draw_vf_destroy( vbuf->vf ); + if (vbuf->render) + vbuf->render->destroy( vbuf->render ); + FREE( stage ); } -- cgit v1.2.3 From 13d9e616f6ee253ecf99dbb67572c87b5dc9270f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 15 Feb 2008 01:11:15 -0500 Subject: a call was missed during the last rename --- src/mesa/pipe/draw/draw_stipple.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_stipple.c b/src/mesa/pipe/draw/draw_stipple.c index 7e0e2137b7..506f33512c 100644 --- a/src/mesa/pipe/draw/draw_stipple.c +++ b/src/mesa/pipe/draw/draw_stipple.c @@ -212,7 +212,7 @@ passthrough_tri(struct draw_stage *stage, struct prim_header *header) static void stipple_destroy( struct draw_stage *stage ) { - draw_free_tmps( stage ); + draw_free_temp_verts( stage ); FREE( stage ); } -- cgit v1.2.3 From d0364584bea6c57bb3ac8d616e677fb52b97ea98 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Feb 2008 22:53:00 -0500 Subject: implement swizzling on writes --- src/mesa/pipe/llvm/storagesoa.cpp | 87 ++++++++++++++++++++------------------- src/mesa/pipe/llvm/storagesoa.h | 14 +++---- src/mesa/pipe/llvm/tgsitollvm.cpp | 9 ++-- 3 files changed, 56 insertions(+), 54 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index a65b5c14d9..75e7a36bc2 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -30,6 +30,8 @@ #include "gallivm_p.h" #include "pipe/p_shader_tokens.h" +#include "pipe/p_debug.h" + #include #include #include @@ -158,49 +160,6 @@ llvm::Value * StorageSoa::extractIndex(llvm::Value *vec) return 0; } -void StorageSoa::storeOutput(int dstIdx, const std::vector &val, - int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - fprintf(stderr, "requires swizzle!!\n"); - assert(0); - } else { - llvm::Value *xChannel = elementPointer(m_output, dstIdx, 0); - llvm::Value *yChannel = elementPointer(m_output, dstIdx, 1); - llvm::Value *zChannel = elementPointer(m_output, dstIdx, 2); - llvm::Value *wChannel = elementPointer(m_output, dstIdx, 3); - - StoreInst *st = new StoreInst(val[0], xChannel, false, m_block); - st = new StoreInst(val[1], yChannel, false, m_block); - st = new StoreInst(val[2], zChannel, false, m_block); - st = new StoreInst(val[3], wChannel, false, m_block); - } -} - -void StorageSoa::storeTemp(int idx, const std::vector &val, - int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - fprintf(stderr, "requires swizzle!!\n"); - assert(0); - } else { - llvm::Value *xChannel = elementPointer(m_temps, idx, 0); - llvm::Value *yChannel = elementPointer(m_temps, idx, 1); - llvm::Value *zChannel = elementPointer(m_temps, idx, 2); - llvm::Value *wChannel = elementPointer(m_temps, idx, 3); - - StoreInst *st = new StoreInst(val[0], xChannel, false, m_block); - st = new StoreInst(val[1], yChannel, false, m_block); - st = new StoreInst(val[2], zChannel, false, m_block); - st = new StoreInst(val[3], wChannel, false, m_block); - } -} - -void StorageSoa::storeAddress(int idx, const std::vector &val, - int mask) -{ -} - llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, int index, int channel) const { @@ -309,6 +268,10 @@ std::vector StorageSoa::argument(Argument type, int idx, int swizz case Immediate: val = immediateElement(idx); break; + case Address: + debug_printf("Address not handled in the fetch phase!\n"); + assert(0); + break; } if (!gallivm_is_swizzle(swizzle)) return val; @@ -321,3 +284,41 @@ std::vector StorageSoa::argument(Argument type, int idx, int swizz res[3] = val[gallivm_w_swizzle(swizzle)]; return res; } + +void StorageSoa::store(Argument type, int idx, const std::vector &val, + int mask) +{ + llvm::Value *out = 0; + switch(type) { + case Output: + out = m_output; + break; + case Temp: + out = m_temps; + break; + case Input: + out = m_input; + break; + default: + debug_printf("Can't save output of this type: %d !\n", type); + assert(0); + break; + } + + if ((mask & TGSI_WRITEMASK_X)) { + llvm::Value *xChannel = elementPointer(out, idx, 0); + new StoreInst(val[0], xChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_Y)) { + llvm::Value *yChannel = elementPointer(out, idx, 1); + new StoreInst(val[1], yChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_Z)) { + llvm::Value *zChannel = elementPointer(out, idx, 2); + new StoreInst(val[2], zChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_W)) { + llvm::Value *wChannel = elementPointer(out, idx, 3); + new StoreInst(val[3], wChannel, false, m_block); + } +} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index 9443234c82..8880bf5ec6 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -49,7 +49,8 @@ public: Output, Temp, Const, - Immediate + Immediate, + Address }; public: StorageSoa(llvm::BasicBlock *block, @@ -58,20 +59,17 @@ public: llvm::Value *consts, llvm::Value *temps); + std::vector argument(Argument type, int idx, int swizzle, llvm::Value *indIdx =0); + void store(Argument type, int idx, const std::vector &val, + int mask); + void addImmediate(float *vec); llvm::Value * addrElement(int idx) const; llvm::Value *extractIndex(llvm::Value *vec); - - void storeOutput(int dstIdx, const std::vector &val, - int mask); - void storeTemp(int idx, const std::vector &val, - int mask); - void storeAddress(int idx, const std::vector &val, - int mask); private: llvm::Value *elementPointer(llvm::Value *ptr, int index, int channel) const; diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index 10c417996a..287a86c60c 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -1063,11 +1063,14 @@ translate_instructionir(llvm::Module *module, struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->storeOutput(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + storage->store(StorageSoa::Output, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->storeTemp(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + storage->store(StorageSoa::Temp, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->storeAddress(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + storage->store(StorageSoa::Address, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } else { fprintf(stderr, "ERROR: unsupported LLVM destination!"); assert(!"wrong destination"); -- cgit v1.2.3 From ae3375987fe9968f822442a0ce49b97f5f0a4070 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 14 Feb 2008 03:08:48 -0500 Subject: rename 'argument' to 'load' because that's what it does --- src/mesa/pipe/llvm/storagesoa.cpp | 4 ++-- src/mesa/pipe/llvm/storagesoa.h | 4 ++-- src/mesa/pipe/llvm/tgsitollvm.cpp | 20 ++++++++++---------- 3 files changed, 14 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index 75e7a36bc2..408b29417c 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -248,8 +248,8 @@ llvm::Value * StorageSoa::createConstGlobalVector(float *vec) return immediate; } -std::vector StorageSoa::argument(Argument type, int idx, int swizzle, - llvm::Value *indIdx ) +std::vector StorageSoa::load(Argument type, int idx, int swizzle, + llvm::Value *indIdx ) { std::vector val(4); switch(type) { diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index 8880bf5ec6..d59168f07d 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -60,8 +60,8 @@ public: llvm::Value *temps); - std::vector argument(Argument type, int idx, int swizzle, - llvm::Value *indIdx =0); + std::vector load(Argument type, int idx, int swizzle, + llvm::Value *indIdx =0); void store(Argument type, int idx, const std::vector &val, int mask); diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index 287a86c60c..0c2c3a9a0a 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -706,20 +706,20 @@ translate_instructionir(llvm::Module *module, indIdx = storage->extractIndex(indIdx); } if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->argument(StorageSoa::Const, - src->SrcRegister.Index, swizzle, indIdx); + val = storage->load(StorageSoa::Const, + src->SrcRegister.Index, swizzle, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->argument(StorageSoa::Input, - src->SrcRegister.Index, swizzle, indIdx); + val = storage->load(StorageSoa::Input, + src->SrcRegister.Index, swizzle, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->argument(StorageSoa::Temp, - src->SrcRegister.Index, swizzle); + val = storage->load(StorageSoa::Temp, + src->SrcRegister.Index, swizzle); } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->argument(StorageSoa::Output, - src->SrcRegister.Index, swizzle, indIdx); + val = storage->load(StorageSoa::Output, + src->SrcRegister.Index, swizzle, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->argument(StorageSoa::Immediate, - src->SrcRegister.Index, swizzle); + val = storage->load(StorageSoa::Immediate, + src->SrcRegister.Index, swizzle); } else { fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); return; -- cgit v1.2.3 From f70cc89dbc8c0f3e58d6cf8eca92e2df1677c86e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 14 Feb 2008 22:42:57 -0500 Subject: redo the way immediates are handled implement madd start implementing arl --- src/mesa/pipe/llvm/instructionssoa.cpp | 64 ++++++++++++++- src/mesa/pipe/llvm/instructionssoa.h | 9 ++- src/mesa/pipe/llvm/storagesoa.cpp | 142 +++++++++++++++++++++++---------- src/mesa/pipe/llvm/storagesoa.h | 15 +++- src/mesa/pipe/llvm/tgsitollvm.cpp | 17 ++-- 5 files changed, 190 insertions(+), 57 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructionssoa.cpp b/src/mesa/pipe/llvm/instructionssoa.cpp index 0c2032e56f..1e144393e0 100644 --- a/src/mesa/pipe/llvm/instructionssoa.cpp +++ b/src/mesa/pipe/llvm/instructionssoa.cpp @@ -1,12 +1,67 @@ #include "instructionssoa.h" +#include "storagesoa.h" + +#include + +using namespace llvm; + InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, StorageSoa *storage) : m_builder(block), + m_storage(storage), m_idx(0) { } +const char * InstructionsSoa::name(const char *prefix) const +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::Value * InstructionsSoa::vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + Constant *constVector = Constant::getNullValue(vectorType); + Value *res = m_builder.CreateInsertElement(constVector, x, + m_storage->constantInt(0), + name("vecx")); + res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), + name("vecxy")); + res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), + name("vecxyz")); + if (w) + res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), + name("vecxyzw")); + return res; +} + +std::vector InstructionsSoa::arl(const std::vector in) +{ + std::vector res(4); + + //Extract the first x (all 4 should be the same) + llvm::Value *x = m_builder.CreateExtractElement(in[0], + m_storage->constantInt(0), + name("extractX")); + //cast it to an unsigned int + x = m_builder.CreateFPToUI(x, IntegerType::get(32), name("xIntCast")); + + res[0] = x; + //only x is valid. the others shouldn't be necessary + /* + res[1] = Constant::getNullValue(m_floatVecType); + res[2] = Constant::getNullValue(m_floatVecType); + res[3] = Constant::getNullValue(m_floatVecType); + */ + + return res; +} + + std::vector InstructionsSoa::add(const std::vector in1, const std::vector in2) { @@ -38,9 +93,10 @@ void InstructionsSoa::end() m_builder.CreateRetVoid(); } -const char * InstructionsSoa::name(const char *prefix) const +std::vector InstructionsSoa::madd(const std::vector in1, + const std::vector in2, + const std::vector in3) { - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; + std::vector res = mul(in1, in2); + return add(res, in3); } diff --git a/src/mesa/pipe/llvm/instructionssoa.h b/src/mesa/pipe/llvm/instructionssoa.h index 0195501584..25ff4ac712 100644 --- a/src/mesa/pipe/llvm/instructionssoa.h +++ b/src/mesa/pipe/llvm/instructionssoa.h @@ -46,17 +46,24 @@ public: InstructionsSoa(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, StorageSoa *storage); + std::vector arl(const std::vector in); + std::vector add(const std::vector in1, const std::vector in2); + std::vector madd(const std::vector in1, + const std::vector in2, + const std::vector in3); std::vector mul(const std::vector in1, const std::vector in2); void end(); private: const char * name(const char *prefix) const; + llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w); private: llvm::LLVMFoldingBuilder m_builder; - + StorageSoa *m_storage; private: mutable int m_idx; mutable char m_name[32]; diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index 408b29417c..f8f4193fd7 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -55,39 +55,81 @@ StorageSoa::StorageSoa(llvm::BasicBlock *block, m_output(output), m_consts(consts), m_temps(temps), + m_immediates(0), m_idx(0) { } void StorageSoa::addImmediate(float *vec) { - float vals[4]; //decompose into soa - - vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0]; - llvm::Value *xChannel = createConstGlobalVector(vals); - - vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1]; - llvm::Value *yChannel = createConstGlobalVector(vals); + std::vector vals(4); + vals[0] = vec[0]; + vals[1] = vec[1]; + vals[2] = vec[2]; + vals[3] = vec[3]; + m_immediatesToFlush.push_back(vals); +} +void StorageSoa::declareImmediates() +{ + if (m_immediatesToFlush.empty()) + return; - vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2]; - llvm::Value *zChannel = createConstGlobalVector(vals); + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorChannels = ArrayType::get(vectorType, 4); + ArrayType *arrayType = ArrayType::get(vectorChannels, m_immediatesToFlush.size()); - vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3]; - llvm::Value *wChannel = createConstGlobalVector(vals); + m_immediates = new GlobalVariable( + /*Type=*/arrayType, + /*isConstant=*/false, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name("immediates"), + currentModule()); - std::vector res(4); - res[0] = xChannel; - res[1] = yChannel; - res[2] = zChannel; - res[3] = wChannel; + std::vector arrayVals; + for (unsigned int i = 0; i < m_immediatesToFlush.size(); ++i) { + std::vector vec = m_immediatesToFlush[i]; + std::vector vals(4); + std::vector channelArray; + + vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0]; + llvm::Constant *xChannel = createConstGlobalVector(vals); + + vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1]; + llvm::Constant *yChannel = createConstGlobalVector(vals); + + vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2]; + llvm::Constant *zChannel = createConstGlobalVector(vals); + + vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3]; + llvm::Constant *wChannel = createConstGlobalVector(vals); + channelArray.push_back(xChannel); + channelArray.push_back(yChannel); + channelArray.push_back(zChannel); + channelArray.push_back(wChannel); + Constant *constChannels = ConstantArray::get(vectorChannels, + channelArray); + arrayVals.push_back(constChannels); + } + Constant *constArray = ConstantArray::get(arrayType, arrayVals); + m_immediates->setInitializer(constArray); - m_immediates[m_immediates.size()] = res; + m_immediatesToFlush.clear(); } llvm::Value *StorageSoa::addrElement(int idx) const { - return 0; + std::map::const_iterator itr = m_addresses.find(idx); + if (itr == m_addresses.end()) { + debug_printf("Trying to access invalid shader 'address'\n"); + return 0; + } + llvm::Value * res = (*itr).second; + + res = new LoadInst(res, name("addr"), false, m_block); + + return res; } std::vector StorageSoa::inputElement(int idx, llvm::Value *indIdx) @@ -145,25 +187,21 @@ std::vector StorageSoa::tempElement(int idx, llvm::Value *indIdx) std::vector StorageSoa::immediateElement(int idx) { std::vector res(4); - res = m_immediates[idx]; - res[0] = new LoadInst(res[0], name("immx"), false, m_block); - res[1] = new LoadInst(res[1], name("immy"), false, m_block); - res[2] = new LoadInst(res[2], name("immz"), false, m_block); - res[3] = new LoadInst(res[3], name("immw"), false, m_block); + res[0] = element(m_immediates, idx, 0); + res[1] = element(m_immediates, idx, 1); + res[2] = element(m_immediates, idx, 2); + res[3] = element(m_immediates, idx, 3); return res; } -llvm::Value * StorageSoa::extractIndex(llvm::Value *vec) -{ - return 0; -} - llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, int index, int channel) const { std::vector indices; + if (m_immediates == ptr) + indices.push_back(constantInt(0)); indices.push_back(constantInt(index)); indices.push_back(constantInt(channel)); @@ -220,17 +258,9 @@ llvm::Module * StorageSoa::currentModule() const return m_block->getParent()->getParent(); } -llvm::Value * StorageSoa::createConstGlobalVector(float *vec) +llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector &vec) { VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - GlobalVariable *immediate = new GlobalVariable( - /*Type=*/vectorType, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name("immediate"), - currentModule()); - std::vector immValues; ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0])); ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1])); @@ -242,16 +272,15 @@ llvm::Value * StorageSoa::createConstGlobalVector(float *vec) immValues.push_back(constw); Constant *constVector = ConstantVector::get(vectorType, immValues); - // Global Variable Definitions - immediate->setInitializer(constVector); - - return immediate; + return constVector; } std::vector StorageSoa::load(Argument type, int idx, int swizzle, - llvm::Value *indIdx ) + llvm::Value *indIdx) { std::vector val(4); + debug_printf("XXXXXXXXX indIdx = %p\n", indIdx); + assert(!indIdx); switch(type) { case Input: val = inputElement(idx, indIdx); @@ -269,7 +298,7 @@ std::vector StorageSoa::load(Argument type, int idx, int swizzle, val = immediateElement(idx); break; case Address: - debug_printf("Address not handled in the fetch phase!\n"); + debug_printf("Address not handled in the load phase!\n"); assert(0); break; } @@ -299,6 +328,17 @@ void StorageSoa::store(Argument type, int idx, const std::vector & case Input: out = m_input; break; + case Address: { + llvm::Value *addr = m_addresses[idx]; + if (!addr) { + addAddress(idx); + addr = m_addresses[idx]; + assert(addr); + } + new StoreInst(val[0], addr, false, m_block); + return; + break; + } default: debug_printf("Can't save output of this type: %d !\n", type); assert(0); @@ -322,3 +362,19 @@ void StorageSoa::store(Argument type, int idx, const std::vector & new StoreInst(val[3], wChannel, false, m_block); } } + +void StorageSoa::addAddress(int idx) +{ + GlobalVariable *val = new GlobalVariable( + /*Type=*/IntegerType::get(32), + /*isConstant=*/false, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name("address"), + currentModule()); + //val->setInitializer(Constant::getNullValue(IntegerType::get(32))); + val->setInitializer(constantInt(1)); + + debug_printf("adding to %d\n", idx); + m_addresses[idx] = val; +} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index d59168f07d..ea5db2b427 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -29,12 +29,14 @@ #define STORAGESOA_H #include +#include #include namespace llvm { class BasicBlock; class Constant; class ConstantInt; + class GlobalVariable; class LoadInst; class Value; class VectorType; @@ -66,20 +68,22 @@ public: int mask); void addImmediate(float *vec); + void declareImmediates(); + + void addAddress(int idx); llvm::Value * addrElement(int idx) const; - llvm::Value *extractIndex(llvm::Value *vec); + llvm::ConstantInt *constantInt(int) const; private: llvm::Value *elementPointer(llvm::Value *ptr, int index, int channel) const; llvm::Value *element(llvm::Value *ptr, int index, int channel) const; const char *name(const char *prefix) const; - llvm::ConstantInt *constantInt(int) const; llvm::Value *alignedArrayLoad(llvm::Value *val); llvm::Module *currentModule() const; - llvm::Value *createConstGlobalVector(float *vec); + llvm::Constant *createConstGlobalVector(const std::vector &vec); std::vector inputElement(int idx, llvm::Value *indIdx =0); std::vector constElement(int idx, llvm::Value *indIdx =0); @@ -93,8 +97,11 @@ private: llvm::Value *m_output; llvm::Value *m_consts; llvm::Value *m_temps; + llvm::GlobalVariable *m_immediates; + + std::map m_addresses; - std::map > m_immediates; + std::vector > m_immediatesToFlush; mutable std::map m_constInts; mutable char m_name[32]; diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index 0c2c3a9a0a..071b7d112e 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -148,10 +148,14 @@ translate_declaration(struct gallivm_ir *prog, static void translate_declarationir(struct gallivm_ir *, llvm::Module *, - StorageSoa *, - struct tgsi_full_declaration *, + StorageSoa *storage, + struct tgsi_full_declaration *decl, struct tgsi_full_declaration *) { + if (decl->Declaration.File == TGSI_FILE_ADDRESS) { + int idx = decl->u.DeclarationRange.First; + storage->addAddress(idx); + } } static void @@ -703,7 +707,7 @@ translate_instructionir(llvm::Module *module, if (src->SrcRegister.Indirect) { indIdx = storage->addrElement(src->SrcRegisterInd.Index); - indIdx = storage->extractIndex(indIdx); + debug_printf("AAAAAAAAAAAAAAA INDIRECT %p\n", indIdx); } if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { val = storage->load(StorageSoa::Const, @@ -713,13 +717,13 @@ translate_instructionir(llvm::Module *module, src->SrcRegister.Index, swizzle, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { val = storage->load(StorageSoa::Temp, - src->SrcRegister.Index, swizzle); + src->SrcRegister.Index, swizzle, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { val = storage->load(StorageSoa::Output, src->SrcRegister.Index, swizzle, indIdx); } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { val = storage->load(StorageSoa::Immediate, - src->SrcRegister.Index, swizzle); + src->SrcRegister.Index, swizzle, indIdx); } else { fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); return; @@ -731,6 +735,7 @@ translate_instructionir(llvm::Module *module, std::vector out(4); switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); } break; case TGSI_OPCODE_MOV: { @@ -780,6 +785,7 @@ translate_instructionir(llvm::Module *module, } break; case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); } break; case TGSI_OPCODE_SUB: { @@ -1198,6 +1204,7 @@ llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, break; case TGSI_TOKEN_TYPE_INSTRUCTION: + storage.declareImmediates(); translate_instructionir(mod, &storage, &instr, &parse.FullToken.FullInstruction, &fi, instno); -- cgit v1.2.3 From cf51d5c4210d3301c96947e6e80e71c252bc04d1 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 14 Feb 2008 23:50:39 -0500 Subject: redo indirection make all load's respect indirection --- src/mesa/pipe/llvm/instructionssoa.cpp | 43 +++++++++++-- src/mesa/pipe/llvm/instructionssoa.h | 1 + src/mesa/pipe/llvm/storagesoa.cpp | 111 ++++++++++++++++++++++++++------- src/mesa/pipe/llvm/storagesoa.h | 6 +- 4 files changed, 130 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructionssoa.cpp b/src/mesa/pipe/llvm/instructionssoa.cpp index 1e144393e0..9c38c64bd1 100644 --- a/src/mesa/pipe/llvm/instructionssoa.cpp +++ b/src/mesa/pipe/llvm/instructionssoa.cpp @@ -43,14 +43,26 @@ std::vector InstructionsSoa::arl(const std::vector i { std::vector res(4); - //Extract the first x (all 4 should be the same) - llvm::Value *x = m_builder.CreateExtractElement(in[0], - m_storage->constantInt(0), - name("extractX")); + //Extract x's + llvm::Value *x1 = m_builder.CreateExtractElement(in[0], + m_storage->constantInt(0), + name("extractX")); + llvm::Value *x2 = m_builder.CreateExtractElement(in[0], + m_storage->constantInt(1), + name("extractX")); + llvm::Value *x3 = m_builder.CreateExtractElement(in[0], + m_storage->constantInt(2), + name("extractX")); + llvm::Value *x4 = m_builder.CreateExtractElement(in[0], + m_storage->constantInt(3), + name("extractX")); //cast it to an unsigned int - x = m_builder.CreateFPToUI(x, IntegerType::get(32), name("xIntCast")); + x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast")); + x2 = m_builder.CreateFPToUI(x2, IntegerType::get(32), name("x2IntCast")); + x3 = m_builder.CreateFPToUI(x3, IntegerType::get(32), name("x3IntCast")); + x4 = m_builder.CreateFPToUI(x4, IntegerType::get(32), name("x4IntCast")); - res[0] = x; + res[0] = vectorFromVals(x1, x2, x3, x4); //only x is valid. the others shouldn't be necessary /* res[1] = Constant::getNullValue(m_floatVecType); @@ -100,3 +112,22 @@ std::vector InstructionsSoa::madd(const std::vector std::vector res = mul(in1, in2); return add(res, in3); } + +std::vector InstructionsSoa::extractVector(llvm::Value *vector) +{ + std::vector res(4); + res[0] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(0), + name("extract1X")); + res[1] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(1), + name("extract2X")); + res[2] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(2), + name("extract3X")); + res[3] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(3), + name("extract4X")); + + return res; +} diff --git a/src/mesa/pipe/llvm/instructionssoa.h b/src/mesa/pipe/llvm/instructionssoa.h index 25ff4ac712..4169dcbb2e 100644 --- a/src/mesa/pipe/llvm/instructionssoa.h +++ b/src/mesa/pipe/llvm/instructionssoa.h @@ -57,6 +57,7 @@ public: const std::vector in2); void end(); + std::vector extractVector(llvm::Value *vector); private: const char * name(const char *prefix) const; llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index f8f4193fd7..314ffe62dc 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -136,10 +136,17 @@ std::vector StorageSoa::inputElement(int idx, llvm::Value *indIdx) { std::vector res(4); - res[0] = element(m_input, idx, 0); - res[1] = element(m_input, idx, 1); - res[2] = element(m_input, idx, 2); - res[3] = element(m_input, idx, 3); + if (!indIdx) { + res[0] = element(m_input, idx, 0); + res[1] = element(m_input, idx, 1); + res[2] = element(m_input, idx, 2); + res[3] = element(m_input, idx, 3); + } else { + res[0] = indirectElement(m_input, indIdx, 0); + res[1] = indirectElement(m_input, indIdx, 1); + res[2] = indirectElement(m_input, indIdx, 2); + res[3] = indirectElement(m_input, indIdx, 3); + } return res; } @@ -147,10 +154,18 @@ std::vector StorageSoa::inputElement(int idx, llvm::Value *indIdx) std::vector StorageSoa::constElement(int idx, llvm::Value *indIdx) { std::vector res(4); - llvm::Value *xChannel = elementPointer(m_consts, idx, 0); - llvm::Value *yChannel = elementPointer(m_consts, idx, 1); - llvm::Value *zChannel = elementPointer(m_consts, idx, 2); - llvm::Value *wChannel = elementPointer(m_consts, idx, 3); + llvm::Value *xChannel, *yChannel, *zChannel, *wChannel; + if (!indIdx) { + xChannel = elementPointer(m_consts, idx, 0); + yChannel = elementPointer(m_consts, idx, 1); + zChannel = elementPointer(m_consts, idx, 2); + wChannel = elementPointer(m_consts, idx, 3); + } else { + xChannel = indirectElementPointer(m_consts, indIdx, 0); + yChannel = indirectElementPointer(m_consts, indIdx, 1); + zChannel = indirectElementPointer(m_consts, indIdx, 2); + wChannel = indirectElementPointer(m_consts, indIdx, 3); + } res[0] = alignedArrayLoad(xChannel); res[1] = alignedArrayLoad(yChannel); @@ -164,10 +179,17 @@ std::vector StorageSoa::outputElement(int idx, llvm::Value *indIdx { std::vector res(4); - res[0] = element(m_output, idx, 0); - res[1] = element(m_output, idx, 1); - res[2] = element(m_output, idx, 2); - res[3] = element(m_output, idx, 3); + if (!indIdx) { + res[0] = element(m_output, idx, 0); + res[1] = element(m_output, idx, 1); + res[2] = element(m_output, idx, 2); + res[3] = element(m_output, idx, 3); + } else { + res[0] = indirectElement(m_output, indIdx, 0); + res[1] = indirectElement(m_output, indIdx, 1); + res[2] = indirectElement(m_output, indIdx, 2); + res[3] = indirectElement(m_output, indIdx, 3); + } return res; } @@ -176,22 +198,36 @@ std::vector StorageSoa::tempElement(int idx, llvm::Value *indIdx) { std::vector res(4); - res[0] = element(m_temps, idx, 0); - res[1] = element(m_temps, idx, 1); - res[2] = element(m_temps, idx, 2); - res[3] = element(m_temps, idx, 3); + if (!indIdx) { + res[0] = element(m_temps, idx, 0); + res[1] = element(m_temps, idx, 1); + res[2] = element(m_temps, idx, 2); + res[3] = element(m_temps, idx, 3); + } else { + res[0] = indirectElement(m_temps, indIdx, 0); + res[1] = indirectElement(m_temps, indIdx, 1); + res[2] = indirectElement(m_temps, indIdx, 2); + res[3] = indirectElement(m_temps, indIdx, 3); + } return res; } -std::vector StorageSoa::immediateElement(int idx) +std::vector StorageSoa::immediateElement(int idx, llvm::Value *indIdx) { std::vector res(4); - res[0] = element(m_immediates, idx, 0); - res[1] = element(m_immediates, idx, 1); - res[2] = element(m_immediates, idx, 2); - res[3] = element(m_immediates, idx, 3); + if (!indIdx) { + res[0] = element(m_immediates, idx, 0); + res[1] = element(m_immediates, idx, 1); + res[2] = element(m_immediates, idx, 2); + res[3] = element(m_immediates, idx, 3); + } else { + res[0] = indirectElement(m_immediates, indIdx, 0); + res[1] = indirectElement(m_immediates, indIdx, 1); + res[2] = indirectElement(m_immediates, indIdx, 2); + res[3] = indirectElement(m_immediates, indIdx, 3); + } return res; } @@ -295,7 +331,7 @@ std::vector StorageSoa::load(Argument type, int idx, int swizzle, val = constElement(idx, indIdx); break; case Immediate: - val = immediateElement(idx); + val = immediateElement(idx, indIdx); break; case Address: debug_printf("Address not handled in the load phase!\n"); @@ -365,16 +401,43 @@ void StorageSoa::store(Argument type, int idx, const std::vector & void StorageSoa::addAddress(int idx) { + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); GlobalVariable *val = new GlobalVariable( - /*Type=*/IntegerType::get(32), + /*Type=*/vectorType, /*isConstant=*/false, /*Linkage=*/GlobalValue::ExternalLinkage, /*Initializer=*/0, // has initializer, specified below /*Name=*/name("address"), currentModule()); //val->setInitializer(Constant::getNullValue(IntegerType::get(32))); - val->setInitializer(constantInt(1)); + //val->setInitializer(constantInt(1)); debug_printf("adding to %d\n", idx); m_addresses[idx] = val; } + +llvm::Value * StorageSoa::indirectElementPointer(llvm::Value *ptr, llvm::Value *indIdx, + int channel) const +{ + std::vector indices; + if (m_immediates == ptr) + indices.push_back(constantInt(0)); + indices.push_back(indIdx); + indices.push_back(constantInt(channel)); + + GetElementPtrInst *getElem = new GetElementPtrInst(ptr, + indices.begin(), + indices.end(), + name("ptr"), + m_block); + return getElem; +} + +llvm::Value * StorageSoa::indirectElement(llvm::Value *ptr, llvm::Value *indIdx, + int channel) const +{ + llvm::Value *res = indirectElementPointer(ptr, indIdx, channel); + LoadInst *load = new LoadInst(res, name("element"), false, m_block); + //load->setAlignment(8); + return load; +} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index ea5db2b427..ca8fee6340 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -80,6 +80,10 @@ private: int channel) const; llvm::Value *element(llvm::Value *ptr, int index, int channel) const; + llvm::Value *indirectElementPointer(llvm::Value *ptr, llvm::Value *indIdx, + int channel) const; + llvm::Value *indirectElement(llvm::Value *ptr, llvm::Value *indIdx, + int channel) const; const char *name(const char *prefix) const; llvm::Value *alignedArrayLoad(llvm::Value *val); llvm::Module *currentModule() const; @@ -89,7 +93,7 @@ private: std::vector constElement(int idx, llvm::Value *indIdx =0); std::vector outputElement(int idx, llvm::Value *indIdx =0); std::vector tempElement(int idx, llvm::Value *indIdx =0); - std::vector immediateElement(int idx); + std::vector immediateElement(int idx, llvm::Value *indIdx =0); private: llvm::BasicBlock *m_block; -- cgit v1.2.3 From 4593be34b2a6e494f0e476c8aa8e1d2633fffd47 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 15 Feb 2008 00:36:18 -0500 Subject: vastly simplify indexing --- src/mesa/pipe/llvm/instructionssoa.cpp | 14 +-- src/mesa/pipe/llvm/storagesoa.cpp | 156 +++++++++++---------------------- src/mesa/pipe/llvm/storagesoa.h | 18 ++-- src/mesa/pipe/llvm/tgsitollvm.cpp | 1 - 4 files changed, 59 insertions(+), 130 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/llvm/instructionssoa.cpp b/src/mesa/pipe/llvm/instructionssoa.cpp index 9c38c64bd1..a4d5046637 100644 --- a/src/mesa/pipe/llvm/instructionssoa.cpp +++ b/src/mesa/pipe/llvm/instructionssoa.cpp @@ -47,22 +47,10 @@ std::vector InstructionsSoa::arl(const std::vector i llvm::Value *x1 = m_builder.CreateExtractElement(in[0], m_storage->constantInt(0), name("extractX")); - llvm::Value *x2 = m_builder.CreateExtractElement(in[0], - m_storage->constantInt(1), - name("extractX")); - llvm::Value *x3 = m_builder.CreateExtractElement(in[0], - m_storage->constantInt(2), - name("extractX")); - llvm::Value *x4 = m_builder.CreateExtractElement(in[0], - m_storage->constantInt(3), - name("extractX")); //cast it to an unsigned int x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast")); - x2 = m_builder.CreateFPToUI(x2, IntegerType::get(32), name("x2IntCast")); - x3 = m_builder.CreateFPToUI(x3, IntegerType::get(32), name("x3IntCast")); - x4 = m_builder.CreateFPToUI(x4, IntegerType::get(32), name("x4IntCast")); - res[0] = vectorFromVals(x1, x2, x3, x4); + res[0] = x1;//vectorFromVals(x1, x2, x3, x4); //only x is valid. the others shouldn't be necessary /* res[1] = Constant::getNullValue(m_floatVecType); diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp index 314ffe62dc..ed0674a96f 100644 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ b/src/mesa/pipe/llvm/storagesoa.cpp @@ -132,40 +132,27 @@ llvm::Value *StorageSoa::addrElement(int idx) const return res; } -std::vector StorageSoa::inputElement(int idx, llvm::Value *indIdx) +std::vector StorageSoa::inputElement(llvm::Value *idx) { std::vector res(4); - if (!indIdx) { - res[0] = element(m_input, idx, 0); - res[1] = element(m_input, idx, 1); - res[2] = element(m_input, idx, 2); - res[3] = element(m_input, idx, 3); - } else { - res[0] = indirectElement(m_input, indIdx, 0); - res[1] = indirectElement(m_input, indIdx, 1); - res[2] = indirectElement(m_input, indIdx, 2); - res[3] = indirectElement(m_input, indIdx, 3); - } + res[0] = element(m_input, idx, 0); + res[1] = element(m_input, idx, 1); + res[2] = element(m_input, idx, 2); + res[3] = element(m_input, idx, 3); return res; } -std::vector StorageSoa::constElement(int idx, llvm::Value *indIdx) +std::vector StorageSoa::constElement(llvm::Value *idx) { std::vector res(4); llvm::Value *xChannel, *yChannel, *zChannel, *wChannel; - if (!indIdx) { - xChannel = elementPointer(m_consts, idx, 0); - yChannel = elementPointer(m_consts, idx, 1); - zChannel = elementPointer(m_consts, idx, 2); - wChannel = elementPointer(m_consts, idx, 3); - } else { - xChannel = indirectElementPointer(m_consts, indIdx, 0); - yChannel = indirectElementPointer(m_consts, indIdx, 1); - zChannel = indirectElementPointer(m_consts, indIdx, 2); - wChannel = indirectElementPointer(m_consts, indIdx, 3); - } + + xChannel = elementPointer(m_consts, idx, 0); + yChannel = elementPointer(m_consts, idx, 1); + zChannel = elementPointer(m_consts, idx, 2); + wChannel = elementPointer(m_consts, idx, 3); res[0] = alignedArrayLoad(xChannel); res[1] = alignedArrayLoad(yChannel); @@ -175,70 +162,49 @@ std::vector StorageSoa::constElement(int idx, llvm::Value *indIdx) return res; } -std::vector StorageSoa::outputElement(int idx, llvm::Value *indIdx) +std::vector StorageSoa::outputElement(llvm::Value *idx) { std::vector res(4); - if (!indIdx) { - res[0] = element(m_output, idx, 0); - res[1] = element(m_output, idx, 1); - res[2] = element(m_output, idx, 2); - res[3] = element(m_output, idx, 3); - } else { - res[0] = indirectElement(m_output, indIdx, 0); - res[1] = indirectElement(m_output, indIdx, 1); - res[2] = indirectElement(m_output, indIdx, 2); - res[3] = indirectElement(m_output, indIdx, 3); - } + res[0] = element(m_output, idx, 0); + res[1] = element(m_output, idx, 1); + res[2] = element(m_output, idx, 2); + res[3] = element(m_output, idx, 3); return res; } -std::vector StorageSoa::tempElement(int idx, llvm::Value *indIdx) +std::vector StorageSoa::tempElement(llvm::Value *idx) { std::vector res(4); - if (!indIdx) { - res[0] = element(m_temps, idx, 0); - res[1] = element(m_temps, idx, 1); - res[2] = element(m_temps, idx, 2); - res[3] = element(m_temps, idx, 3); - } else { - res[0] = indirectElement(m_temps, indIdx, 0); - res[1] = indirectElement(m_temps, indIdx, 1); - res[2] = indirectElement(m_temps, indIdx, 2); - res[3] = indirectElement(m_temps, indIdx, 3); - } + res[0] = element(m_temps, idx, 0); + res[1] = element(m_temps, idx, 1); + res[2] = element(m_temps, idx, 2); + res[3] = element(m_temps, idx, 3); return res; } -std::vector StorageSoa::immediateElement(int idx, llvm::Value *indIdx) +std::vector StorageSoa::immediateElement(llvm::Value *idx) { std::vector res(4); - if (!indIdx) { - res[0] = element(m_immediates, idx, 0); - res[1] = element(m_immediates, idx, 1); - res[2] = element(m_immediates, idx, 2); - res[3] = element(m_immediates, idx, 3); - } else { - res[0] = indirectElement(m_immediates, indIdx, 0); - res[1] = indirectElement(m_immediates, indIdx, 1); - res[2] = indirectElement(m_immediates, indIdx, 2); - res[3] = indirectElement(m_immediates, indIdx, 3); - } + res[0] = element(m_immediates, idx, 0); + res[1] = element(m_immediates, idx, 1); + res[2] = element(m_immediates, idx, 2); + res[3] = element(m_immediates, idx, 3); return res; } -llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, int index, +llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index, int channel) const { std::vector indices; if (m_immediates == ptr) indices.push_back(constantInt(0)); - indices.push_back(constantInt(index)); + indices.push_back(index); indices.push_back(constantInt(channel)); GetElementPtrInst *getElem = new GetElementPtrInst(ptr, @@ -249,7 +215,7 @@ llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, int index, return getElem; } -llvm::Value * StorageSoa::element(llvm::Value *ptr, int index, +llvm::Value * StorageSoa::element(llvm::Value *ptr, llvm::Value *index, int channel) const { llvm::Value *res = elementPointer(ptr, index, channel); @@ -315,23 +281,31 @@ std::vector StorageSoa::load(Argument type, int idx, int swizzle, llvm::Value *indIdx) { std::vector val(4); - debug_printf("XXXXXXXXX indIdx = %p\n", indIdx); - assert(!indIdx); + + //if we have an indirect index, always use that + // if not use the integer offset to create one + llvm::Value *realIndex = 0; + if (indIdx) + realIndex = indIdx; + else + realIndex = constantInt(idx); + debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx); + switch(type) { case Input: - val = inputElement(idx, indIdx); + val = inputElement(realIndex); break; case Output: - val = outputElement(idx, indIdx); + val = outputElement(realIndex); break; case Temp: - val = tempElement(idx, indIdx); + val = tempElement(realIndex); break; case Const: - val = constElement(idx, indIdx); + val = constElement(realIndex); break; case Immediate: - val = immediateElement(idx, indIdx); + val = immediateElement(realIndex); break; case Address: debug_printf("Address not handled in the load phase!\n"); @@ -380,64 +354,36 @@ void StorageSoa::store(Argument type, int idx, const std::vector & assert(0); break; } - + llvm::Value *realIndex = constantInt(idx); if ((mask & TGSI_WRITEMASK_X)) { - llvm::Value *xChannel = elementPointer(out, idx, 0); + llvm::Value *xChannel = elementPointer(out, realIndex, 0); new StoreInst(val[0], xChannel, false, m_block); } if ((mask & TGSI_WRITEMASK_Y)) { - llvm::Value *yChannel = elementPointer(out, idx, 1); + llvm::Value *yChannel = elementPointer(out, realIndex, 1); new StoreInst(val[1], yChannel, false, m_block); } if ((mask & TGSI_WRITEMASK_Z)) { - llvm::Value *zChannel = elementPointer(out, idx, 2); + llvm::Value *zChannel = elementPointer(out, realIndex, 2); new StoreInst(val[2], zChannel, false, m_block); } if ((mask & TGSI_WRITEMASK_W)) { - llvm::Value *wChannel = elementPointer(out, idx, 3); + llvm::Value *wChannel = elementPointer(out, realIndex, 3); new StoreInst(val[3], wChannel, false, m_block); } } void StorageSoa::addAddress(int idx) { - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); GlobalVariable *val = new GlobalVariable( - /*Type=*/vectorType, + /*Type=*/IntegerType::get(32), /*isConstant=*/false, /*Linkage=*/GlobalValue::ExternalLinkage, /*Initializer=*/0, // has initializer, specified below /*Name=*/name("address"), currentModule()); - //val->setInitializer(Constant::getNullValue(IntegerType::get(32))); - //val->setInitializer(constantInt(1)); + val->setInitializer(Constant::getNullValue(IntegerType::get(32))); debug_printf("adding to %d\n", idx); m_addresses[idx] = val; } - -llvm::Value * StorageSoa::indirectElementPointer(llvm::Value *ptr, llvm::Value *indIdx, - int channel) const -{ - std::vector indices; - if (m_immediates == ptr) - indices.push_back(constantInt(0)); - indices.push_back(indIdx); - indices.push_back(constantInt(channel)); - - GetElementPtrInst *getElem = new GetElementPtrInst(ptr, - indices.begin(), - indices.end(), - name("ptr"), - m_block); - return getElem; -} - -llvm::Value * StorageSoa::indirectElement(llvm::Value *ptr, llvm::Value *indIdx, - int channel) const -{ - llvm::Value *res = indirectElementPointer(ptr, indIdx, channel); - LoadInst *load = new LoadInst(res, name("element"), false, m_block); - //load->setAlignment(8); - return load; -} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h index ca8fee6340..6443351f27 100644 --- a/src/mesa/pipe/llvm/storagesoa.h +++ b/src/mesa/pipe/llvm/storagesoa.h @@ -76,24 +76,20 @@ public: llvm::ConstantInt *constantInt(int) const; private: - llvm::Value *elementPointer(llvm::Value *ptr, int index, + llvm::Value *elementPointer(llvm::Value *ptr, llvm::Value *indIdx, int channel) const; - llvm::Value *element(llvm::Value *ptr, int index, + llvm::Value *element(llvm::Value *ptr, llvm::Value *idx, int channel) const; - llvm::Value *indirectElementPointer(llvm::Value *ptr, llvm::Value *indIdx, - int channel) const; - llvm::Value *indirectElement(llvm::Value *ptr, llvm::Value *indIdx, - int channel) const; const char *name(const char *prefix) const; llvm::Value *alignedArrayLoad(llvm::Value *val); llvm::Module *currentModule() const; llvm::Constant *createConstGlobalVector(const std::vector &vec); - std::vector inputElement(int idx, llvm::Value *indIdx =0); - std::vector constElement(int idx, llvm::Value *indIdx =0); - std::vector outputElement(int idx, llvm::Value *indIdx =0); - std::vector tempElement(int idx, llvm::Value *indIdx =0); - std::vector immediateElement(int idx, llvm::Value *indIdx =0); + std::vector inputElement(llvm::Value *indIdx); + std::vector constElement(llvm::Value *indIdx); + std::vector outputElement(llvm::Value *indIdx); + std::vector tempElement(llvm::Value *indIdx); + std::vector immediateElement(llvm::Value *indIdx); private: llvm::BasicBlock *m_block; diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp index 071b7d112e..0de595e678 100644 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ b/src/mesa/pipe/llvm/tgsitollvm.cpp @@ -707,7 +707,6 @@ translate_instructionir(llvm::Module *module, if (src->SrcRegister.Indirect) { indIdx = storage->addrElement(src->SrcRegisterInd.Index); - debug_printf("AAAAAAAAAAAAAAA INDIRECT %p\n", indIdx); } if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { val = storage->load(StorageSoa::Const, -- cgit v1.2.3 From b642730be93149baa7556e5791393168ab396175 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Feb 2008 17:35:24 +0900 Subject: Code reorganization: move files into their places. This is in a separate commit to ensure renames are properly preserved. --- src/gallium/Makefile | 28 + src/gallium/Makefile.template | 63 + src/gallium/README.portability | 43 + src/gallium/SConscript | 9 + src/gallium/aux/cso_cache/cso_cache.c | 181 ++ src/gallium/aux/cso_cache/cso_cache.h | 107 + src/gallium/aux/cso_cache/cso_hash.c | 388 +++ src/gallium/aux/cso_cache/cso_hash.h | 62 + src/gallium/aux/draw/Makefile | 2 + src/gallium/aux/draw/draw_clip.c | 488 +++ src/gallium/aux/draw/draw_context.c | 293 ++ src/gallium/aux/draw/draw_context.h | 142 + src/gallium/aux/draw/draw_cull.c | 150 + src/gallium/aux/draw/draw_debug.c | 113 + src/gallium/aux/draw/draw_flatshade.c | 205 ++ src/gallium/aux/draw/draw_offset.c | 186 ++ src/gallium/aux/draw/draw_prim.c | 482 +++ src/gallium/aux/draw/draw_private.h | 346 +++ src/gallium/aux/draw/draw_stipple.c | 239 ++ src/gallium/aux/draw/draw_twoside.c | 203 ++ src/gallium/aux/draw/draw_unfilled.c | 206 ++ src/gallium/aux/draw/draw_validate.c | 185 ++ src/gallium/aux/draw/draw_vbuf.c | 570 ++++ src/gallium/aux/draw/draw_vbuf.h | 106 + src/gallium/aux/draw/draw_vertex.c | 79 + src/gallium/aux/draw/draw_vertex.h | 111 + src/gallium/aux/draw/draw_vertex_cache.c | 196 ++ src/gallium/aux/draw/draw_vertex_fetch.c | 510 ++++ src/gallium/aux/draw/draw_vertex_shader.c | 325 ++ src/gallium/aux/draw/draw_vf.c | 428 +++ src/gallium/aux/draw/draw_vf.h | 236 ++ src/gallium/aux/draw/draw_vf_generic.c | 585 ++++ src/gallium/aux/draw/draw_vf_sse.c | 614 ++++ src/gallium/aux/draw/draw_wide_prims.c | 432 +++ src/gallium/aux/llvm/Makefile | 83 + src/gallium/aux/llvm/gallivm.cpp | 327 ++ src/gallium/aux/llvm/gallivm.h | 103 + src/gallium/aux/llvm/gallivm_builtins.cpp | 567 ++++ src/gallium/aux/llvm/gallivm_cpu.cpp | 202 ++ src/gallium/aux/llvm/gallivm_p.h | 110 + src/gallium/aux/llvm/instructions.cpp | 889 ++++++ src/gallium/aux/llvm/instructions.h | 152 + src/gallium/aux/llvm/instructionssoa.cpp | 121 + src/gallium/aux/llvm/instructionssoa.h | 74 + src/gallium/aux/llvm/llvm_builtins.c | 115 + src/gallium/aux/llvm/loweringpass.cpp | 17 + src/gallium/aux/llvm/loweringpass.h | 15 + src/gallium/aux/llvm/storage.cpp | 364 +++ src/gallium/aux/llvm/storage.h | 133 + src/gallium/aux/llvm/storagesoa.cpp | 389 +++ src/gallium/aux/llvm/storagesoa.h | 111 + src/gallium/aux/llvm/tgsitollvm.cpp | 1221 ++++++++ src/gallium/aux/llvm/tgsitollvm.h | 20 + src/gallium/aux/pipebuffer/Makefile | 23 + src/gallium/aux/pipebuffer/linked_list.h | 91 + src/gallium/aux/pipebuffer/pb_buffer.h | 202 ++ src/gallium/aux/pipebuffer/pb_buffer_fenced.c | 299 ++ src/gallium/aux/pipebuffer/pb_buffer_fenced.h | 117 + src/gallium/aux/pipebuffer/pb_buffer_malloc.c | 127 + src/gallium/aux/pipebuffer/pb_bufmgr.h | 126 + src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c | 131 + src/gallium/aux/pipebuffer/pb_bufmgr_mm.c | 593 ++++ src/gallium/aux/pipebuffer/pb_bufmgr_pool.c | 288 ++ src/gallium/aux/pipebuffer/pb_winsys.c | 170 ++ src/gallium/aux/tgsi/Makefile | 3 + src/gallium/aux/tgsi/exec/Makefile | 3 + src/gallium/aux/tgsi/exec/tgsi_exec.c | 2485 +++++++++++++++ src/gallium/aux/tgsi/exec/tgsi_exec.h | 239 ++ src/gallium/aux/tgsi/exec/tgsi_sse2.c | 2378 +++++++++++++++ src/gallium/aux/tgsi/exec/tgsi_sse2.h | 26 + src/gallium/aux/tgsi/util/tgsi_build.c | 1371 +++++++++ src/gallium/aux/tgsi/util/tgsi_build.h | 320 ++ src/gallium/aux/tgsi/util/tgsi_dump.c | 1581 ++++++++++ src/gallium/aux/tgsi/util/tgsi_dump.h | 28 + src/gallium/aux/tgsi/util/tgsi_parse.c | 319 ++ src/gallium/aux/tgsi/util/tgsi_parse.h | 121 + src/gallium/aux/tgsi/util/tgsi_transform.c | 199 ++ src/gallium/aux/tgsi/util/tgsi_transform.h | 93 + src/gallium/aux/tgsi/util/tgsi_util.c | 274 ++ src/gallium/aux/tgsi/util/tgsi_util.h | 70 + src/gallium/aux/util/p_debug.c | 76 + src/gallium/aux/util/p_tile.c | 699 +++++ src/gallium/aux/util/p_tile.h | 81 + src/gallium/aux/util/p_util.c | 73 + src/gallium/drivers/cell/Makefile | 12 + src/gallium/drivers/cell/common.h | 220 ++ src/gallium/drivers/cell/ppu/Makefile | 76 + src/gallium/drivers/cell/ppu/cell_batch.c | 217 ++ src/gallium/drivers/cell/ppu/cell_batch.h | 58 + src/gallium/drivers/cell/ppu/cell_clear.c | 76 + src/gallium/drivers/cell/ppu/cell_clear.h | 43 + src/gallium/drivers/cell/ppu/cell_context.c | 287 ++ src/gallium/drivers/cell/ppu/cell_context.h | 135 + src/gallium/drivers/cell/ppu/cell_draw_arrays.c | 164 + src/gallium/drivers/cell/ppu/cell_draw_arrays.h | 42 + src/gallium/drivers/cell/ppu/cell_flush.c | 84 + src/gallium/drivers/cell/ppu/cell_flush.h | 38 + src/gallium/drivers/cell/ppu/cell_render.c | 210 ++ src/gallium/drivers/cell/ppu/cell_render.h | 39 + src/gallium/drivers/cell/ppu/cell_spu.c | 155 + src/gallium/drivers/cell/ppu/cell_spu.h | 82 + src/gallium/drivers/cell/ppu/cell_state.h | 115 + src/gallium/drivers/cell/ppu/cell_state_blend.c | 109 + src/gallium/drivers/cell/ppu/cell_state_clip.c | 84 + src/gallium/drivers/cell/ppu/cell_state_derived.c | 192 ++ src/gallium/drivers/cell/ppu/cell_state_emit.c | 103 + src/gallium/drivers/cell/ppu/cell_state_emit.h | 36 + src/gallium/drivers/cell/ppu/cell_state_fs.c | 171 ++ .../drivers/cell/ppu/cell_state_rasterizer.c | 106 + src/gallium/drivers/cell/ppu/cell_state_sampler.c | 84 + src/gallium/drivers/cell/ppu/cell_state_surface.c | 71 + src/gallium/drivers/cell/ppu/cell_state_vertex.c | 63 + src/gallium/drivers/cell/ppu/cell_surface.c | 179 ++ src/gallium/drivers/cell/ppu/cell_surface.h | 42 + src/gallium/drivers/cell/ppu/cell_texture.c | 252 ++ src/gallium/drivers/cell/ppu/cell_texture.h | 80 + src/gallium/drivers/cell/ppu/cell_vbuf.c | 294 ++ src/gallium/drivers/cell/ppu/cell_vbuf.h | 38 + src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 120 + src/gallium/drivers/cell/ppu/cell_winsys.c | 40 + src/gallium/drivers/cell/ppu/cell_winsys.h | 50 + src/gallium/drivers/cell/spu/Makefile | 72 + src/gallium/drivers/cell/spu/spu_blend.c | 62 + src/gallium/drivers/cell/spu/spu_blend.h | 37 + src/gallium/drivers/cell/spu/spu_colorpack.h | 110 + src/gallium/drivers/cell/spu/spu_exec.c | 1948 ++++++++++++ src/gallium/drivers/cell/spu/spu_exec.h | 172 ++ src/gallium/drivers/cell/spu/spu_main.c | 567 ++++ src/gallium/drivers/cell/spu/spu_main.h | 177 ++ src/gallium/drivers/cell/spu/spu_render.c | 301 ++ src/gallium/drivers/cell/spu/spu_render.h | 38 + src/gallium/drivers/cell/spu/spu_texture.c | 217 ++ src/gallium/drivers/cell/spu/spu_texture.h | 47 + src/gallium/drivers/cell/spu/spu_tile.c | 83 + src/gallium/drivers/cell/spu/spu_tile.h | 73 + src/gallium/drivers/cell/spu/spu_tri.c | 926 ++++++ src/gallium/drivers/cell/spu/spu_tri.h | 37 + src/gallium/drivers/cell/spu/spu_util.c | 165 + src/gallium/drivers/cell/spu/spu_vertex_fetch.c | 673 +++++ src/gallium/drivers/cell/spu/spu_vertex_shader.c | 231 ++ src/gallium/drivers/cell/spu/spu_vertex_shader.h | 63 + src/gallium/drivers/cell/spu/spu_ztest.h | 135 + src/gallium/drivers/failover/Makefile | 21 + src/gallium/drivers/failover/fo_context.c | 155 + src/gallium/drivers/failover/fo_context.h | 114 + src/gallium/drivers/failover/fo_state.c | 457 +++ src/gallium/drivers/failover/fo_state_emit.c | 137 + src/gallium/drivers/failover/fo_winsys.h | 45 + src/gallium/drivers/i915simple/Makefile | 38 + src/gallium/drivers/i915simple/SConscript | 29 + src/gallium/drivers/i915simple/i915_batch.h | 54 + src/gallium/drivers/i915simple/i915_blit.c | 162 + src/gallium/drivers/i915simple/i915_blit.h | 55 + src/gallium/drivers/i915simple/i915_clear.c | 47 + src/gallium/drivers/i915simple/i915_context.c | 320 ++ src/gallium/drivers/i915simple/i915_context.h | 304 ++ src/gallium/drivers/i915simple/i915_debug.c | 901 ++++++ src/gallium/drivers/i915simple/i915_debug.h | 117 + src/gallium/drivers/i915simple/i915_debug_fp.c | 366 +++ src/gallium/drivers/i915simple/i915_flush.c | 81 + src/gallium/drivers/i915simple/i915_fpc.h | 213 ++ src/gallium/drivers/i915simple/i915_fpc_emit.c | 375 +++ .../drivers/i915simple/i915_fpc_translate.c | 1135 +++++++ src/gallium/drivers/i915simple/i915_prim_emit.c | 215 ++ src/gallium/drivers/i915simple/i915_prim_vbuf.c | 254 ++ src/gallium/drivers/i915simple/i915_reg.h | 978 ++++++ src/gallium/drivers/i915simple/i915_state.c | 694 +++++ src/gallium/drivers/i915simple/i915_state.h | 50 + .../drivers/i915simple/i915_state_derived.c | 177 ++ .../drivers/i915simple/i915_state_dynamic.c | 308 ++ src/gallium/drivers/i915simple/i915_state_emit.c | 374 +++ .../drivers/i915simple/i915_state_immediate.c | 221 ++ .../drivers/i915simple/i915_state_inlines.h | 230 ++ .../drivers/i915simple/i915_state_sampler.c | 231 ++ src/gallium/drivers/i915simple/i915_strings.c | 83 + src/gallium/drivers/i915simple/i915_surface.c | 191 ++ src/gallium/drivers/i915simple/i915_texture.c | 536 ++++ src/gallium/drivers/i915simple/i915_texture.h | 17 + src/gallium/drivers/i915simple/i915_winsys.h | 115 + src/gallium/drivers/i965simple/Makefile | 66 + src/gallium/drivers/i965simple/SConscript | 55 + src/gallium/drivers/i965simple/brw_batch.h | 59 + src/gallium/drivers/i965simple/brw_blit.c | 218 ++ src/gallium/drivers/i965simple/brw_blit.h | 33 + src/gallium/drivers/i965simple/brw_cc.c | 269 ++ src/gallium/drivers/i965simple/brw_clip.c | 206 ++ src/gallium/drivers/i965simple/brw_clip.h | 170 ++ src/gallium/drivers/i965simple/brw_clip_line.c | 245 ++ src/gallium/drivers/i965simple/brw_clip_point.c | 47 + src/gallium/drivers/i965simple/brw_clip_state.c | 92 + src/gallium/drivers/i965simple/brw_clip_tri.c | 566 ++++ src/gallium/drivers/i965simple/brw_clip_unfilled.c | 477 +++ src/gallium/drivers/i965simple/brw_clip_util.c | 351 +++ src/gallium/drivers/i965simple/brw_context.c | 245 ++ src/gallium/drivers/i965simple/brw_context.h | 690 +++++ src/gallium/drivers/i965simple/brw_curbe.c | 368 +++ src/gallium/drivers/i965simple/brw_defines.h | 852 ++++++ src/gallium/drivers/i965simple/brw_draw.c | 239 ++ src/gallium/drivers/i965simple/brw_draw.h | 55 + src/gallium/drivers/i965simple/brw_draw_upload.c | 299 ++ src/gallium/drivers/i965simple/brw_eu.c | 130 + src/gallium/drivers/i965simple/brw_eu.h | 888 ++++++ src/gallium/drivers/i965simple/brw_eu_debug.c | 90 + src/gallium/drivers/i965simple/brw_eu_emit.c | 1080 +++++++ src/gallium/drivers/i965simple/brw_eu_util.c | 126 + src/gallium/drivers/i965simple/brw_flush.c | 80 + src/gallium/drivers/i965simple/brw_gs.c | 196 ++ src/gallium/drivers/i965simple/brw_gs.h | 75 + src/gallium/drivers/i965simple/brw_gs_emit.c | 148 + src/gallium/drivers/i965simple/brw_gs_state.c | 89 + src/gallium/drivers/i965simple/brw_misc_state.c | 486 +++ src/gallium/drivers/i965simple/brw_reg.h | 76 + src/gallium/drivers/i965simple/brw_sf.c | 351 +++ src/gallium/drivers/i965simple/brw_sf.h | 122 + src/gallium/drivers/i965simple/brw_sf_emit.c | 382 +++ src/gallium/drivers/i965simple/brw_sf_state.c | 180 ++ src/gallium/drivers/i965simple/brw_shader_info.c | 49 + src/gallium/drivers/i965simple/brw_state.c | 424 +++ src/gallium/drivers/i965simple/brw_state.h | 158 + src/gallium/drivers/i965simple/brw_state_batch.c | 113 + src/gallium/drivers/i965simple/brw_state_cache.c | 443 +++ src/gallium/drivers/i965simple/brw_state_pool.c | 137 + src/gallium/drivers/i965simple/brw_state_upload.c | 202 ++ src/gallium/drivers/i965simple/brw_strings.c | 72 + src/gallium/drivers/i965simple/brw_structs.h | 1348 +++++++++ src/gallium/drivers/i965simple/brw_surface.c | 210 ++ src/gallium/drivers/i965simple/brw_tex_layout.c | 353 +++ src/gallium/drivers/i965simple/brw_tex_layout.h | 15 + src/gallium/drivers/i965simple/brw_urb.c | 186 ++ src/gallium/drivers/i965simple/brw_util.c | 104 + src/gallium/drivers/i965simple/brw_util.h | 43 + src/gallium/drivers/i965simple/brw_vs.c | 120 + src/gallium/drivers/i965simple/brw_vs.h | 82 + src/gallium/drivers/i965simple/brw_vs_emit.c | 1332 ++++++++ src/gallium/drivers/i965simple/brw_vs_state.c | 102 + src/gallium/drivers/i965simple/brw_winsys.h | 205 ++ src/gallium/drivers/i965simple/brw_wm.c | 210 ++ src/gallium/drivers/i965simple/brw_wm.h | 142 + src/gallium/drivers/i965simple/brw_wm_decl.c | 383 +++ src/gallium/drivers/i965simple/brw_wm_glsl.c | 1079 +++++++ src/gallium/drivers/i965simple/brw_wm_iz.c | 214 ++ .../drivers/i965simple/brw_wm_sampler_state.c | 273 ++ src/gallium/drivers/i965simple/brw_wm_state.c | 194 ++ .../drivers/i965simple/brw_wm_surface_state.c | 304 ++ src/gallium/drivers/softpipe/Makefile | 50 + src/gallium/drivers/softpipe/SConscript | 42 + src/gallium/drivers/softpipe/sp_clear.c | 73 + src/gallium/drivers/softpipe/sp_clear.h | 43 + src/gallium/drivers/softpipe/sp_context.c | 333 ++ src/gallium/drivers/softpipe/sp_context.h | 152 + src/gallium/drivers/softpipe/sp_draw_arrays.c | 164 + src/gallium/drivers/softpipe/sp_flush.c | 76 + src/gallium/drivers/softpipe/sp_flush.h | 35 + src/gallium/drivers/softpipe/sp_headers.h | 82 + src/gallium/drivers/softpipe/sp_prim_setup.c | 1247 ++++++++ src/gallium/drivers/softpipe/sp_prim_setup.h | 79 + src/gallium/drivers/softpipe/sp_prim_vbuf.c | 221 ++ src/gallium/drivers/softpipe/sp_prim_vbuf.h | 38 + src/gallium/drivers/softpipe/sp_quad.c | 118 + src/gallium/drivers/softpipe/sp_quad.h | 70 + src/gallium/drivers/softpipe/sp_quad_alpha_test.c | 108 + src/gallium/drivers/softpipe/sp_quad_blend.c | 749 +++++ src/gallium/drivers/softpipe/sp_quad_bufloop.c | 72 + src/gallium/drivers/softpipe/sp_quad_colormask.c | 110 + src/gallium/drivers/softpipe/sp_quad_coverage.c | 88 + src/gallium/drivers/softpipe/sp_quad_depth_test.c | 276 ++ src/gallium/drivers/softpipe/sp_quad_earlyz.c | 88 + src/gallium/drivers/softpipe/sp_quad_fs.c | 390 +++ src/gallium/drivers/softpipe/sp_quad_occlusion.c | 85 + src/gallium/drivers/softpipe/sp_quad_output.c | 90 + src/gallium/drivers/softpipe/sp_quad_stencil.c | 352 +++ src/gallium/drivers/softpipe/sp_quad_stipple.c | 94 + src/gallium/drivers/softpipe/sp_query.c | 107 + src/gallium/drivers/softpipe/sp_query.h | 39 + src/gallium/drivers/softpipe/sp_state.h | 187 ++ src/gallium/drivers/softpipe/sp_state_blend.c | 98 + src/gallium/drivers/softpipe/sp_state_clip.c | 83 + src/gallium/drivers/softpipe/sp_state_derived.c | 235 ++ src/gallium/drivers/softpipe/sp_state_fs.c | 179 ++ src/gallium/drivers/softpipe/sp_state_rasterizer.c | 62 + src/gallium/drivers/softpipe/sp_state_sampler.c | 93 + src/gallium/drivers/softpipe/sp_state_surface.c | 109 + src/gallium/drivers/softpipe/sp_state_vertex.c | 64 + src/gallium/drivers/softpipe/sp_surface.c | 159 + src/gallium/drivers/softpipe/sp_surface.h | 42 + src/gallium/drivers/softpipe/sp_tex_sample.c | 916 ++++++ src/gallium/drivers/softpipe/sp_tex_sample.h | 17 + src/gallium/drivers/softpipe/sp_texture.c | 166 + src/gallium/drivers/softpipe/sp_texture.h | 71 + src/gallium/drivers/softpipe/sp_tile_cache.c | 585 ++++ src/gallium/drivers/softpipe/sp_tile_cache.h | 104 + src/gallium/drivers/softpipe/sp_winsys.h | 57 + src/gallium/include/pipe/p_compiler.h | 116 + src/gallium/include/pipe/p_context.h | 221 ++ src/gallium/include/pipe/p_debug.h | 86 + src/gallium/include/pipe/p_defines.h | 270 ++ src/gallium/include/pipe/p_format.h | 421 +++ src/gallium/include/pipe/p_inlines.h | 112 + src/gallium/include/pipe/p_shader_tokens.h | 806 +++++ src/gallium/include/pipe/p_state.h | 322 ++ src/gallium/include/pipe/p_thread.h | 54 + src/gallium/include/pipe/p_util.h | 408 +++ src/gallium/include/pipe/p_winsys.h | 160 + src/gallium/winsys/dri/intel/Makefile | 38 + src/gallium/winsys/dri/intel/SConscript | 41 + src/gallium/winsys/dri/intel/intel_batchbuffer.c | 357 +++ src/gallium/winsys/dri/intel/intel_batchbuffer.h | 149 + src/gallium/winsys/dri/intel/intel_batchpool.c | 424 +++ src/gallium/winsys/dri/intel/intel_batchpool.h | 37 + src/gallium/winsys/dri/intel/intel_context.c | 304 ++ src/gallium/winsys/dri/intel/intel_context.h | 158 + src/gallium/winsys/dri/intel/intel_lock.c | 102 + src/gallium/winsys/dri/intel/intel_reg.h | 53 + src/gallium/winsys/dri/intel/intel_screen.c | 537 ++++ src/gallium/winsys/dri/intel/intel_screen.h | 113 + src/gallium/winsys/dri/intel/intel_swapbuffers.c | 253 ++ src/gallium/winsys/dri/intel/intel_swapbuffers.h | 47 + src/gallium/winsys/dri/intel/intel_winsys.h | 73 + src/gallium/winsys/dri/intel/intel_winsys_i915.c | 154 + src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 302 ++ .../winsys/dri/intel/intel_winsys_softpipe.c | 81 + src/gallium/winsys/dri/intel/server/i830_common.h | 226 ++ src/gallium/winsys/dri/intel/server/i830_dri.h | 63 + src/gallium/winsys/dri/intel/server/intel.h | 331 ++ src/gallium/winsys/dri/intel/server/intel_dri.c | 1306 ++++++++ src/gallium/winsys/xlib/brw_aub.c | 392 +++ src/gallium/winsys/xlib/brw_aub.h | 114 + src/gallium/winsys/xlib/fakeglx.c | 3188 ++++++++++++++++++++ src/gallium/winsys/xlib/glxapi.c | 1408 +++++++++ src/gallium/winsys/xlib/glxapi.h | 228 ++ src/gallium/winsys/xlib/glxheader.h | 62 + src/gallium/winsys/xlib/realglx.c | 180 ++ src/gallium/winsys/xlib/realglx.h | 326 ++ src/gallium/winsys/xlib/xfonts.c | 377 +++ src/gallium/winsys/xlib/xfonts.h | 41 + src/gallium/winsys/xlib/xm_api.c | 1380 +++++++++ src/gallium/winsys/xlib/xm_image.c | 133 + src/gallium/winsys/xlib/xm_image.h | 77 + src/gallium/winsys/xlib/xm_winsys.c | 466 +++ src/gallium/winsys/xlib/xm_winsys_aub.c | 589 ++++ src/gallium/winsys/xlib/xm_winsys_aub.h | 67 + src/gallium/winsys/xlib/xmesaP.h | 176 ++ src/mesa/drivers/dri/intel_winsys/Makefile | 38 - src/mesa/drivers/dri/intel_winsys/SConscript | 41 - .../drivers/dri/intel_winsys/intel_batchbuffer.c | 357 --- .../drivers/dri/intel_winsys/intel_batchbuffer.h | 149 - .../drivers/dri/intel_winsys/intel_batchpool.c | 424 --- .../drivers/dri/intel_winsys/intel_batchpool.h | 37 - src/mesa/drivers/dri/intel_winsys/intel_context.c | 304 -- src/mesa/drivers/dri/intel_winsys/intel_context.h | 158 - src/mesa/drivers/dri/intel_winsys/intel_lock.c | 102 - src/mesa/drivers/dri/intel_winsys/intel_reg.h | 53 - src/mesa/drivers/dri/intel_winsys/intel_screen.c | 537 ---- src/mesa/drivers/dri/intel_winsys/intel_screen.h | 113 - .../drivers/dri/intel_winsys/intel_swapbuffers.c | 253 -- .../drivers/dri/intel_winsys/intel_swapbuffers.h | 47 - src/mesa/drivers/dri/intel_winsys/intel_winsys.h | 73 - .../drivers/dri/intel_winsys/intel_winsys_i915.c | 154 - .../drivers/dri/intel_winsys/intel_winsys_pipe.c | 302 -- .../dri/intel_winsys/intel_winsys_softpipe.c | 81 - .../drivers/dri/intel_winsys/server/i830_common.h | 226 -- .../drivers/dri/intel_winsys/server/i830_dri.h | 63 - src/mesa/drivers/dri/intel_winsys/server/intel.h | 331 -- .../drivers/dri/intel_winsys/server/intel_dri.c | 1306 -------- src/mesa/pipe/Makefile | 28 - src/mesa/pipe/Makefile.template | 63 - src/mesa/pipe/README.portability | 43 - src/mesa/pipe/SConscript | 9 - src/mesa/pipe/cell/Makefile | 12 - src/mesa/pipe/cell/common.h | 220 -- src/mesa/pipe/cell/ppu/Makefile | 76 - src/mesa/pipe/cell/ppu/cell_batch.c | 217 -- src/mesa/pipe/cell/ppu/cell_batch.h | 58 - src/mesa/pipe/cell/ppu/cell_clear.c | 76 - src/mesa/pipe/cell/ppu/cell_clear.h | 43 - src/mesa/pipe/cell/ppu/cell_context.c | 287 -- src/mesa/pipe/cell/ppu/cell_context.h | 135 - src/mesa/pipe/cell/ppu/cell_draw_arrays.c | 164 - src/mesa/pipe/cell/ppu/cell_draw_arrays.h | 42 - src/mesa/pipe/cell/ppu/cell_flush.c | 84 - src/mesa/pipe/cell/ppu/cell_flush.h | 38 - src/mesa/pipe/cell/ppu/cell_render.c | 210 -- src/mesa/pipe/cell/ppu/cell_render.h | 39 - src/mesa/pipe/cell/ppu/cell_spu.c | 155 - src/mesa/pipe/cell/ppu/cell_spu.h | 82 - src/mesa/pipe/cell/ppu/cell_state.h | 115 - src/mesa/pipe/cell/ppu/cell_state_blend.c | 109 - src/mesa/pipe/cell/ppu/cell_state_clip.c | 84 - src/mesa/pipe/cell/ppu/cell_state_derived.c | 192 -- src/mesa/pipe/cell/ppu/cell_state_emit.c | 103 - src/mesa/pipe/cell/ppu/cell_state_emit.h | 36 - src/mesa/pipe/cell/ppu/cell_state_fs.c | 171 -- src/mesa/pipe/cell/ppu/cell_state_rasterizer.c | 106 - src/mesa/pipe/cell/ppu/cell_state_sampler.c | 84 - src/mesa/pipe/cell/ppu/cell_state_surface.c | 71 - src/mesa/pipe/cell/ppu/cell_state_vertex.c | 63 - src/mesa/pipe/cell/ppu/cell_surface.c | 179 -- src/mesa/pipe/cell/ppu/cell_surface.h | 42 - src/mesa/pipe/cell/ppu/cell_texture.c | 252 -- src/mesa/pipe/cell/ppu/cell_texture.h | 80 - src/mesa/pipe/cell/ppu/cell_vbuf.c | 294 -- src/mesa/pipe/cell/ppu/cell_vbuf.h | 38 - src/mesa/pipe/cell/ppu/cell_vertex_shader.c | 120 - src/mesa/pipe/cell/ppu/cell_winsys.c | 40 - src/mesa/pipe/cell/ppu/cell_winsys.h | 50 - src/mesa/pipe/cell/spu/Makefile | 72 - src/mesa/pipe/cell/spu/spu_blend.c | 62 - src/mesa/pipe/cell/spu/spu_blend.h | 37 - src/mesa/pipe/cell/spu/spu_colorpack.h | 110 - src/mesa/pipe/cell/spu/spu_exec.c | 1948 ------------ src/mesa/pipe/cell/spu/spu_exec.h | 172 -- src/mesa/pipe/cell/spu/spu_main.c | 567 ---- src/mesa/pipe/cell/spu/spu_main.h | 177 -- src/mesa/pipe/cell/spu/spu_render.c | 301 -- src/mesa/pipe/cell/spu/spu_render.h | 38 - src/mesa/pipe/cell/spu/spu_texture.c | 217 -- src/mesa/pipe/cell/spu/spu_texture.h | 47 - src/mesa/pipe/cell/spu/spu_tile.c | 83 - src/mesa/pipe/cell/spu/spu_tile.h | 73 - src/mesa/pipe/cell/spu/spu_tri.c | 926 ------ src/mesa/pipe/cell/spu/spu_tri.h | 37 - src/mesa/pipe/cell/spu/spu_util.c | 165 - src/mesa/pipe/cell/spu/spu_vertex_fetch.c | 673 ----- src/mesa/pipe/cell/spu/spu_vertex_shader.c | 231 -- src/mesa/pipe/cell/spu/spu_vertex_shader.h | 63 - src/mesa/pipe/cell/spu/spu_ztest.h | 135 - src/mesa/pipe/cso_cache/cso_cache.c | 181 -- src/mesa/pipe/cso_cache/cso_cache.h | 107 - src/mesa/pipe/cso_cache/cso_hash.c | 388 --- src/mesa/pipe/cso_cache/cso_hash.h | 62 - src/mesa/pipe/draw/Makefile | 2 - src/mesa/pipe/draw/draw_clip.c | 488 --- src/mesa/pipe/draw/draw_context.c | 293 -- src/mesa/pipe/draw/draw_context.h | 142 - src/mesa/pipe/draw/draw_cull.c | 150 - src/mesa/pipe/draw/draw_debug.c | 113 - src/mesa/pipe/draw/draw_flatshade.c | 205 -- src/mesa/pipe/draw/draw_offset.c | 186 -- src/mesa/pipe/draw/draw_prim.c | 482 --- src/mesa/pipe/draw/draw_private.h | 346 --- src/mesa/pipe/draw/draw_stipple.c | 239 -- src/mesa/pipe/draw/draw_twoside.c | 203 -- src/mesa/pipe/draw/draw_unfilled.c | 206 -- src/mesa/pipe/draw/draw_validate.c | 185 -- src/mesa/pipe/draw/draw_vbuf.c | 570 ---- src/mesa/pipe/draw/draw_vbuf.h | 106 - src/mesa/pipe/draw/draw_vertex.c | 79 - src/mesa/pipe/draw/draw_vertex.h | 111 - src/mesa/pipe/draw/draw_vertex_cache.c | 196 -- src/mesa/pipe/draw/draw_vertex_fetch.c | 510 ---- src/mesa/pipe/draw/draw_vertex_shader.c | 325 -- src/mesa/pipe/draw/draw_vf.c | 428 --- src/mesa/pipe/draw/draw_vf.h | 236 -- src/mesa/pipe/draw/draw_vf_generic.c | 585 ---- src/mesa/pipe/draw/draw_vf_sse.c | 614 ---- src/mesa/pipe/draw/draw_wide_prims.c | 432 --- src/mesa/pipe/failover/Makefile | 21 - src/mesa/pipe/failover/fo_context.c | 155 - src/mesa/pipe/failover/fo_context.h | 114 - src/mesa/pipe/failover/fo_state.c | 457 --- src/mesa/pipe/failover/fo_state_emit.c | 137 - src/mesa/pipe/failover/fo_winsys.h | 45 - src/mesa/pipe/i915simple/Makefile | 38 - src/mesa/pipe/i915simple/SConscript | 29 - src/mesa/pipe/i915simple/i915_batch.h | 54 - src/mesa/pipe/i915simple/i915_blit.c | 162 - src/mesa/pipe/i915simple/i915_blit.h | 55 - src/mesa/pipe/i915simple/i915_clear.c | 47 - src/mesa/pipe/i915simple/i915_context.c | 320 -- src/mesa/pipe/i915simple/i915_context.h | 304 -- src/mesa/pipe/i915simple/i915_debug.c | 901 ------ src/mesa/pipe/i915simple/i915_debug.h | 117 - src/mesa/pipe/i915simple/i915_debug_fp.c | 366 --- src/mesa/pipe/i915simple/i915_flush.c | 81 - src/mesa/pipe/i915simple/i915_fpc.h | 213 -- src/mesa/pipe/i915simple/i915_fpc_emit.c | 375 --- src/mesa/pipe/i915simple/i915_fpc_translate.c | 1135 ------- src/mesa/pipe/i915simple/i915_prim_emit.c | 215 -- src/mesa/pipe/i915simple/i915_prim_vbuf.c | 254 -- src/mesa/pipe/i915simple/i915_reg.h | 978 ------ src/mesa/pipe/i915simple/i915_state.c | 694 ----- src/mesa/pipe/i915simple/i915_state.h | 50 - src/mesa/pipe/i915simple/i915_state_derived.c | 177 -- src/mesa/pipe/i915simple/i915_state_dynamic.c | 308 -- src/mesa/pipe/i915simple/i915_state_emit.c | 374 --- src/mesa/pipe/i915simple/i915_state_immediate.c | 221 -- src/mesa/pipe/i915simple/i915_state_inlines.h | 230 -- src/mesa/pipe/i915simple/i915_state_sampler.c | 231 -- src/mesa/pipe/i915simple/i915_strings.c | 83 - src/mesa/pipe/i915simple/i915_surface.c | 191 -- src/mesa/pipe/i915simple/i915_texture.c | 536 ---- src/mesa/pipe/i915simple/i915_texture.h | 17 - src/mesa/pipe/i915simple/i915_winsys.h | 115 - src/mesa/pipe/i965simple/Makefile | 66 - src/mesa/pipe/i965simple/SConscript | 55 - src/mesa/pipe/i965simple/brw_batch.h | 59 - src/mesa/pipe/i965simple/brw_blit.c | 218 -- src/mesa/pipe/i965simple/brw_blit.h | 33 - src/mesa/pipe/i965simple/brw_cc.c | 269 -- src/mesa/pipe/i965simple/brw_clip.c | 206 -- src/mesa/pipe/i965simple/brw_clip.h | 170 -- src/mesa/pipe/i965simple/brw_clip_line.c | 245 -- src/mesa/pipe/i965simple/brw_clip_point.c | 47 - src/mesa/pipe/i965simple/brw_clip_state.c | 92 - src/mesa/pipe/i965simple/brw_clip_tri.c | 566 ---- src/mesa/pipe/i965simple/brw_clip_unfilled.c | 477 --- src/mesa/pipe/i965simple/brw_clip_util.c | 351 --- src/mesa/pipe/i965simple/brw_context.c | 245 -- src/mesa/pipe/i965simple/brw_context.h | 690 ----- src/mesa/pipe/i965simple/brw_curbe.c | 368 --- src/mesa/pipe/i965simple/brw_defines.h | 852 ------ src/mesa/pipe/i965simple/brw_draw.c | 239 -- src/mesa/pipe/i965simple/brw_draw.h | 55 - src/mesa/pipe/i965simple/brw_draw_upload.c | 299 -- src/mesa/pipe/i965simple/brw_eu.c | 130 - src/mesa/pipe/i965simple/brw_eu.h | 888 ------ src/mesa/pipe/i965simple/brw_eu_debug.c | 90 - src/mesa/pipe/i965simple/brw_eu_emit.c | 1080 ------- src/mesa/pipe/i965simple/brw_eu_util.c | 126 - src/mesa/pipe/i965simple/brw_flush.c | 80 - src/mesa/pipe/i965simple/brw_gs.c | 196 -- src/mesa/pipe/i965simple/brw_gs.h | 75 - src/mesa/pipe/i965simple/brw_gs_emit.c | 148 - src/mesa/pipe/i965simple/brw_gs_state.c | 89 - src/mesa/pipe/i965simple/brw_misc_state.c | 486 --- src/mesa/pipe/i965simple/brw_reg.h | 76 - src/mesa/pipe/i965simple/brw_sf.c | 351 --- src/mesa/pipe/i965simple/brw_sf.h | 122 - src/mesa/pipe/i965simple/brw_sf_emit.c | 382 --- src/mesa/pipe/i965simple/brw_sf_state.c | 180 -- src/mesa/pipe/i965simple/brw_shader_info.c | 49 - src/mesa/pipe/i965simple/brw_state.c | 424 --- src/mesa/pipe/i965simple/brw_state.h | 158 - src/mesa/pipe/i965simple/brw_state_batch.c | 113 - src/mesa/pipe/i965simple/brw_state_cache.c | 443 --- src/mesa/pipe/i965simple/brw_state_pool.c | 137 - src/mesa/pipe/i965simple/brw_state_upload.c | 202 -- src/mesa/pipe/i965simple/brw_strings.c | 72 - src/mesa/pipe/i965simple/brw_structs.h | 1348 --------- src/mesa/pipe/i965simple/brw_surface.c | 210 -- src/mesa/pipe/i965simple/brw_tex_layout.c | 353 --- src/mesa/pipe/i965simple/brw_tex_layout.h | 15 - src/mesa/pipe/i965simple/brw_urb.c | 186 -- src/mesa/pipe/i965simple/brw_util.c | 104 - src/mesa/pipe/i965simple/brw_util.h | 43 - src/mesa/pipe/i965simple/brw_vs.c | 120 - src/mesa/pipe/i965simple/brw_vs.h | 82 - src/mesa/pipe/i965simple/brw_vs_emit.c | 1332 -------- src/mesa/pipe/i965simple/brw_vs_state.c | 102 - src/mesa/pipe/i965simple/brw_winsys.h | 205 -- src/mesa/pipe/i965simple/brw_wm.c | 210 -- src/mesa/pipe/i965simple/brw_wm.h | 142 - src/mesa/pipe/i965simple/brw_wm_decl.c | 383 --- src/mesa/pipe/i965simple/brw_wm_glsl.c | 1079 ------- src/mesa/pipe/i965simple/brw_wm_iz.c | 214 -- src/mesa/pipe/i965simple/brw_wm_sampler_state.c | 273 -- src/mesa/pipe/i965simple/brw_wm_state.c | 194 -- src/mesa/pipe/i965simple/brw_wm_surface_state.c | 304 -- src/mesa/pipe/llvm/Makefile | 83 - src/mesa/pipe/llvm/gallivm.cpp | 327 -- src/mesa/pipe/llvm/gallivm.h | 103 - src/mesa/pipe/llvm/gallivm_builtins.cpp | 567 ---- src/mesa/pipe/llvm/gallivm_cpu.cpp | 202 -- src/mesa/pipe/llvm/gallivm_p.h | 110 - src/mesa/pipe/llvm/instructions.cpp | 889 ------ src/mesa/pipe/llvm/instructions.h | 152 - src/mesa/pipe/llvm/instructionssoa.cpp | 121 - src/mesa/pipe/llvm/instructionssoa.h | 74 - src/mesa/pipe/llvm/llvm_builtins.c | 115 - src/mesa/pipe/llvm/loweringpass.cpp | 17 - src/mesa/pipe/llvm/loweringpass.h | 15 - src/mesa/pipe/llvm/storage.cpp | 364 --- src/mesa/pipe/llvm/storage.h | 133 - src/mesa/pipe/llvm/storagesoa.cpp | 389 --- src/mesa/pipe/llvm/storagesoa.h | 111 - src/mesa/pipe/llvm/tgsitollvm.cpp | 1221 -------- src/mesa/pipe/llvm/tgsitollvm.h | 20 - src/mesa/pipe/p_compiler.h | 116 - src/mesa/pipe/p_context.h | 221 -- src/mesa/pipe/p_debug.h | 86 - src/mesa/pipe/p_defines.h | 270 -- src/mesa/pipe/p_format.h | 421 --- src/mesa/pipe/p_inlines.h | 112 - src/mesa/pipe/p_shader_tokens.h | 806 ----- src/mesa/pipe/p_state.h | 322 -- src/mesa/pipe/p_thread.h | 54 - src/mesa/pipe/p_util.h | 408 --- src/mesa/pipe/p_winsys.h | 160 - src/mesa/pipe/pipebuffer/Makefile | 23 - src/mesa/pipe/pipebuffer/linked_list.h | 91 - src/mesa/pipe/pipebuffer/pb_buffer.h | 202 -- src/mesa/pipe/pipebuffer/pb_buffer_fenced.c | 299 -- src/mesa/pipe/pipebuffer/pb_buffer_fenced.h | 117 - src/mesa/pipe/pipebuffer/pb_buffer_malloc.c | 127 - src/mesa/pipe/pipebuffer/pb_bufmgr.h | 126 - src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c | 131 - src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c | 593 ---- src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c | 288 -- src/mesa/pipe/pipebuffer/pb_winsys.c | 170 -- src/mesa/pipe/softpipe/Makefile | 50 - src/mesa/pipe/softpipe/SConscript | 42 - src/mesa/pipe/softpipe/sp_clear.c | 73 - src/mesa/pipe/softpipe/sp_clear.h | 43 - src/mesa/pipe/softpipe/sp_context.c | 333 -- src/mesa/pipe/softpipe/sp_context.h | 152 - src/mesa/pipe/softpipe/sp_draw_arrays.c | 164 - src/mesa/pipe/softpipe/sp_flush.c | 76 - src/mesa/pipe/softpipe/sp_flush.h | 35 - src/mesa/pipe/softpipe/sp_headers.h | 82 - src/mesa/pipe/softpipe/sp_prim_setup.c | 1247 -------- src/mesa/pipe/softpipe/sp_prim_setup.h | 79 - src/mesa/pipe/softpipe/sp_prim_vbuf.c | 221 -- src/mesa/pipe/softpipe/sp_prim_vbuf.h | 38 - src/mesa/pipe/softpipe/sp_quad.c | 118 - src/mesa/pipe/softpipe/sp_quad.h | 70 - src/mesa/pipe/softpipe/sp_quad_alpha_test.c | 108 - src/mesa/pipe/softpipe/sp_quad_blend.c | 749 ----- src/mesa/pipe/softpipe/sp_quad_bufloop.c | 72 - src/mesa/pipe/softpipe/sp_quad_colormask.c | 110 - src/mesa/pipe/softpipe/sp_quad_coverage.c | 88 - src/mesa/pipe/softpipe/sp_quad_depth_test.c | 276 -- src/mesa/pipe/softpipe/sp_quad_earlyz.c | 88 - src/mesa/pipe/softpipe/sp_quad_fs.c | 390 --- src/mesa/pipe/softpipe/sp_quad_occlusion.c | 85 - src/mesa/pipe/softpipe/sp_quad_output.c | 90 - src/mesa/pipe/softpipe/sp_quad_stencil.c | 352 --- src/mesa/pipe/softpipe/sp_quad_stipple.c | 94 - src/mesa/pipe/softpipe/sp_query.c | 107 - src/mesa/pipe/softpipe/sp_query.h | 39 - src/mesa/pipe/softpipe/sp_state.h | 187 -- src/mesa/pipe/softpipe/sp_state_blend.c | 98 - src/mesa/pipe/softpipe/sp_state_clip.c | 83 - src/mesa/pipe/softpipe/sp_state_derived.c | 235 -- src/mesa/pipe/softpipe/sp_state_fs.c | 179 -- src/mesa/pipe/softpipe/sp_state_rasterizer.c | 62 - src/mesa/pipe/softpipe/sp_state_sampler.c | 93 - src/mesa/pipe/softpipe/sp_state_surface.c | 109 - src/mesa/pipe/softpipe/sp_state_vertex.c | 64 - src/mesa/pipe/softpipe/sp_surface.c | 159 - src/mesa/pipe/softpipe/sp_surface.h | 42 - src/mesa/pipe/softpipe/sp_tex_sample.c | 916 ------ src/mesa/pipe/softpipe/sp_tex_sample.h | 17 - src/mesa/pipe/softpipe/sp_texture.c | 166 - src/mesa/pipe/softpipe/sp_texture.h | 71 - src/mesa/pipe/softpipe/sp_tile_cache.c | 585 ---- src/mesa/pipe/softpipe/sp_tile_cache.h | 104 - src/mesa/pipe/softpipe/sp_winsys.h | 57 - src/mesa/pipe/tgsi/Makefile | 3 - src/mesa/pipe/tgsi/exec/Makefile | 3 - src/mesa/pipe/tgsi/exec/tgsi_exec.c | 2485 --------------- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 239 -- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 2378 --------------- src/mesa/pipe/tgsi/exec/tgsi_sse2.h | 26 - src/mesa/pipe/tgsi/util/tgsi_build.c | 1371 --------- src/mesa/pipe/tgsi/util/tgsi_build.h | 320 -- src/mesa/pipe/tgsi/util/tgsi_dump.c | 1581 ---------- src/mesa/pipe/tgsi/util/tgsi_dump.h | 28 - src/mesa/pipe/tgsi/util/tgsi_parse.c | 319 -- src/mesa/pipe/tgsi/util/tgsi_parse.h | 121 - src/mesa/pipe/tgsi/util/tgsi_transform.c | 199 -- src/mesa/pipe/tgsi/util/tgsi_transform.h | 93 - src/mesa/pipe/tgsi/util/tgsi_util.c | 274 -- src/mesa/pipe/tgsi/util/tgsi_util.h | 70 - src/mesa/pipe/util/p_debug.c | 76 - src/mesa/pipe/util/p_tile.c | 699 ----- src/mesa/pipe/util/p_tile.h | 81 - src/mesa/pipe/util/p_util.c | 73 - src/mesa/pipe/xlib/brw_aub.c | 392 --- src/mesa/pipe/xlib/brw_aub.h | 114 - src/mesa/pipe/xlib/fakeglx.c | 3188 -------------------- src/mesa/pipe/xlib/glxapi.c | 1408 --------- src/mesa/pipe/xlib/glxapi.h | 228 -- src/mesa/pipe/xlib/glxheader.h | 62 - src/mesa/pipe/xlib/realglx.c | 180 -- src/mesa/pipe/xlib/realglx.h | 326 -- src/mesa/pipe/xlib/xfonts.c | 377 --- src/mesa/pipe/xlib/xfonts.h | 41 - src/mesa/pipe/xlib/xm_api.c | 1380 --------- src/mesa/pipe/xlib/xm_image.c | 133 - src/mesa/pipe/xlib/xm_image.h | 77 - src/mesa/pipe/xlib/xm_winsys.c | 466 --- src/mesa/pipe/xlib/xm_winsys_aub.c | 589 ---- src/mesa/pipe/xlib/xm_winsys_aub.h | 67 - src/mesa/pipe/xlib/xmesaP.h | 176 -- 684 files changed, 90658 insertions(+), 90658 deletions(-) create mode 100644 src/gallium/Makefile create mode 100644 src/gallium/Makefile.template create mode 100644 src/gallium/README.portability create mode 100644 src/gallium/SConscript create mode 100644 src/gallium/aux/cso_cache/cso_cache.c create mode 100644 src/gallium/aux/cso_cache/cso_cache.h create mode 100644 src/gallium/aux/cso_cache/cso_hash.c create mode 100644 src/gallium/aux/cso_cache/cso_hash.h create mode 100644 src/gallium/aux/draw/Makefile create mode 100644 src/gallium/aux/draw/draw_clip.c create mode 100644 src/gallium/aux/draw/draw_context.c create mode 100644 src/gallium/aux/draw/draw_context.h create mode 100644 src/gallium/aux/draw/draw_cull.c create mode 100644 src/gallium/aux/draw/draw_debug.c create mode 100644 src/gallium/aux/draw/draw_flatshade.c create mode 100644 src/gallium/aux/draw/draw_offset.c create mode 100644 src/gallium/aux/draw/draw_prim.c create mode 100644 src/gallium/aux/draw/draw_private.h create mode 100644 src/gallium/aux/draw/draw_stipple.c create mode 100644 src/gallium/aux/draw/draw_twoside.c create mode 100644 src/gallium/aux/draw/draw_unfilled.c create mode 100644 src/gallium/aux/draw/draw_validate.c create mode 100644 src/gallium/aux/draw/draw_vbuf.c create mode 100644 src/gallium/aux/draw/draw_vbuf.h create mode 100644 src/gallium/aux/draw/draw_vertex.c create mode 100644 src/gallium/aux/draw/draw_vertex.h create mode 100644 src/gallium/aux/draw/draw_vertex_cache.c create mode 100644 src/gallium/aux/draw/draw_vertex_fetch.c create mode 100644 src/gallium/aux/draw/draw_vertex_shader.c create mode 100644 src/gallium/aux/draw/draw_vf.c create mode 100644 src/gallium/aux/draw/draw_vf.h create mode 100644 src/gallium/aux/draw/draw_vf_generic.c create mode 100644 src/gallium/aux/draw/draw_vf_sse.c create mode 100644 src/gallium/aux/draw/draw_wide_prims.c create mode 100644 src/gallium/aux/llvm/Makefile create mode 100644 src/gallium/aux/llvm/gallivm.cpp create mode 100644 src/gallium/aux/llvm/gallivm.h create mode 100644 src/gallium/aux/llvm/gallivm_builtins.cpp create mode 100644 src/gallium/aux/llvm/gallivm_cpu.cpp create mode 100644 src/gallium/aux/llvm/gallivm_p.h create mode 100644 src/gallium/aux/llvm/instructions.cpp create mode 100644 src/gallium/aux/llvm/instructions.h create mode 100644 src/gallium/aux/llvm/instructionssoa.cpp create mode 100644 src/gallium/aux/llvm/instructionssoa.h create mode 100644 src/gallium/aux/llvm/llvm_builtins.c create mode 100644 src/gallium/aux/llvm/loweringpass.cpp create mode 100644 src/gallium/aux/llvm/loweringpass.h create mode 100644 src/gallium/aux/llvm/storage.cpp create mode 100644 src/gallium/aux/llvm/storage.h create mode 100644 src/gallium/aux/llvm/storagesoa.cpp create mode 100644 src/gallium/aux/llvm/storagesoa.h create mode 100644 src/gallium/aux/llvm/tgsitollvm.cpp create mode 100644 src/gallium/aux/llvm/tgsitollvm.h create mode 100644 src/gallium/aux/pipebuffer/Makefile create mode 100644 src/gallium/aux/pipebuffer/linked_list.h create mode 100644 src/gallium/aux/pipebuffer/pb_buffer.h create mode 100644 src/gallium/aux/pipebuffer/pb_buffer_fenced.c create mode 100644 src/gallium/aux/pipebuffer/pb_buffer_fenced.h create mode 100644 src/gallium/aux/pipebuffer/pb_buffer_malloc.c create mode 100644 src/gallium/aux/pipebuffer/pb_bufmgr.h create mode 100644 src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c create mode 100644 src/gallium/aux/pipebuffer/pb_bufmgr_mm.c create mode 100644 src/gallium/aux/pipebuffer/pb_bufmgr_pool.c create mode 100644 src/gallium/aux/pipebuffer/pb_winsys.c create mode 100644 src/gallium/aux/tgsi/Makefile create mode 100644 src/gallium/aux/tgsi/exec/Makefile create mode 100644 src/gallium/aux/tgsi/exec/tgsi_exec.c create mode 100644 src/gallium/aux/tgsi/exec/tgsi_exec.h create mode 100755 src/gallium/aux/tgsi/exec/tgsi_sse2.c create mode 100755 src/gallium/aux/tgsi/exec/tgsi_sse2.h create mode 100644 src/gallium/aux/tgsi/util/tgsi_build.c create mode 100644 src/gallium/aux/tgsi/util/tgsi_build.h create mode 100644 src/gallium/aux/tgsi/util/tgsi_dump.c create mode 100644 src/gallium/aux/tgsi/util/tgsi_dump.h create mode 100644 src/gallium/aux/tgsi/util/tgsi_parse.c create mode 100644 src/gallium/aux/tgsi/util/tgsi_parse.h create mode 100644 src/gallium/aux/tgsi/util/tgsi_transform.c create mode 100644 src/gallium/aux/tgsi/util/tgsi_transform.h create mode 100644 src/gallium/aux/tgsi/util/tgsi_util.c create mode 100644 src/gallium/aux/tgsi/util/tgsi_util.h create mode 100644 src/gallium/aux/util/p_debug.c create mode 100644 src/gallium/aux/util/p_tile.c create mode 100644 src/gallium/aux/util/p_tile.h create mode 100644 src/gallium/aux/util/p_util.c create mode 100644 src/gallium/drivers/cell/Makefile create mode 100644 src/gallium/drivers/cell/common.h create mode 100644 src/gallium/drivers/cell/ppu/Makefile create mode 100644 src/gallium/drivers/cell/ppu/cell_batch.c create mode 100644 src/gallium/drivers/cell/ppu/cell_batch.h create mode 100644 src/gallium/drivers/cell/ppu/cell_clear.c create mode 100644 src/gallium/drivers/cell/ppu/cell_clear.h create mode 100644 src/gallium/drivers/cell/ppu/cell_context.c create mode 100644 src/gallium/drivers/cell/ppu/cell_context.h create mode 100644 src/gallium/drivers/cell/ppu/cell_draw_arrays.c create mode 100644 src/gallium/drivers/cell/ppu/cell_draw_arrays.h create mode 100644 src/gallium/drivers/cell/ppu/cell_flush.c create mode 100644 src/gallium/drivers/cell/ppu/cell_flush.h create mode 100644 src/gallium/drivers/cell/ppu/cell_render.c create mode 100644 src/gallium/drivers/cell/ppu/cell_render.h create mode 100644 src/gallium/drivers/cell/ppu/cell_spu.c create mode 100644 src/gallium/drivers/cell/ppu/cell_spu.h create mode 100644 src/gallium/drivers/cell/ppu/cell_state.h create mode 100644 src/gallium/drivers/cell/ppu/cell_state_blend.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_clip.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_derived.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_emit.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_emit.h create mode 100644 src/gallium/drivers/cell/ppu/cell_state_fs.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_rasterizer.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_sampler.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_surface.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_vertex.c create mode 100644 src/gallium/drivers/cell/ppu/cell_surface.c create mode 100644 src/gallium/drivers/cell/ppu/cell_surface.h create mode 100644 src/gallium/drivers/cell/ppu/cell_texture.c create mode 100644 src/gallium/drivers/cell/ppu/cell_texture.h create mode 100644 src/gallium/drivers/cell/ppu/cell_vbuf.c create mode 100644 src/gallium/drivers/cell/ppu/cell_vbuf.h create mode 100644 src/gallium/drivers/cell/ppu/cell_vertex_shader.c create mode 100644 src/gallium/drivers/cell/ppu/cell_winsys.c create mode 100644 src/gallium/drivers/cell/ppu/cell_winsys.h create mode 100644 src/gallium/drivers/cell/spu/Makefile create mode 100644 src/gallium/drivers/cell/spu/spu_blend.c create mode 100644 src/gallium/drivers/cell/spu/spu_blend.h create mode 100644 src/gallium/drivers/cell/spu/spu_colorpack.h create mode 100644 src/gallium/drivers/cell/spu/spu_exec.c create mode 100644 src/gallium/drivers/cell/spu/spu_exec.h create mode 100644 src/gallium/drivers/cell/spu/spu_main.c create mode 100644 src/gallium/drivers/cell/spu/spu_main.h create mode 100644 src/gallium/drivers/cell/spu/spu_render.c create mode 100644 src/gallium/drivers/cell/spu/spu_render.h create mode 100644 src/gallium/drivers/cell/spu/spu_texture.c create mode 100644 src/gallium/drivers/cell/spu/spu_texture.h create mode 100644 src/gallium/drivers/cell/spu/spu_tile.c create mode 100644 src/gallium/drivers/cell/spu/spu_tile.h create mode 100644 src/gallium/drivers/cell/spu/spu_tri.c create mode 100644 src/gallium/drivers/cell/spu/spu_tri.h create mode 100644 src/gallium/drivers/cell/spu/spu_util.c create mode 100644 src/gallium/drivers/cell/spu/spu_vertex_fetch.c create mode 100644 src/gallium/drivers/cell/spu/spu_vertex_shader.c create mode 100644 src/gallium/drivers/cell/spu/spu_vertex_shader.h create mode 100644 src/gallium/drivers/cell/spu/spu_ztest.h create mode 100644 src/gallium/drivers/failover/Makefile create mode 100644 src/gallium/drivers/failover/fo_context.c create mode 100644 src/gallium/drivers/failover/fo_context.h create mode 100644 src/gallium/drivers/failover/fo_state.c create mode 100644 src/gallium/drivers/failover/fo_state_emit.c create mode 100644 src/gallium/drivers/failover/fo_winsys.h create mode 100644 src/gallium/drivers/i915simple/Makefile create mode 100644 src/gallium/drivers/i915simple/SConscript create mode 100644 src/gallium/drivers/i915simple/i915_batch.h create mode 100644 src/gallium/drivers/i915simple/i915_blit.c create mode 100644 src/gallium/drivers/i915simple/i915_blit.h create mode 100644 src/gallium/drivers/i915simple/i915_clear.c create mode 100644 src/gallium/drivers/i915simple/i915_context.c create mode 100644 src/gallium/drivers/i915simple/i915_context.h create mode 100644 src/gallium/drivers/i915simple/i915_debug.c create mode 100644 src/gallium/drivers/i915simple/i915_debug.h create mode 100644 src/gallium/drivers/i915simple/i915_debug_fp.c create mode 100644 src/gallium/drivers/i915simple/i915_flush.c create mode 100644 src/gallium/drivers/i915simple/i915_fpc.h create mode 100644 src/gallium/drivers/i915simple/i915_fpc_emit.c create mode 100644 src/gallium/drivers/i915simple/i915_fpc_translate.c create mode 100644 src/gallium/drivers/i915simple/i915_prim_emit.c create mode 100644 src/gallium/drivers/i915simple/i915_prim_vbuf.c create mode 100644 src/gallium/drivers/i915simple/i915_reg.h create mode 100644 src/gallium/drivers/i915simple/i915_state.c create mode 100644 src/gallium/drivers/i915simple/i915_state.h create mode 100644 src/gallium/drivers/i915simple/i915_state_derived.c create mode 100644 src/gallium/drivers/i915simple/i915_state_dynamic.c create mode 100644 src/gallium/drivers/i915simple/i915_state_emit.c create mode 100644 src/gallium/drivers/i915simple/i915_state_immediate.c create mode 100644 src/gallium/drivers/i915simple/i915_state_inlines.h create mode 100644 src/gallium/drivers/i915simple/i915_state_sampler.c create mode 100644 src/gallium/drivers/i915simple/i915_strings.c create mode 100644 src/gallium/drivers/i915simple/i915_surface.c create mode 100644 src/gallium/drivers/i915simple/i915_texture.c create mode 100644 src/gallium/drivers/i915simple/i915_texture.h create mode 100644 src/gallium/drivers/i915simple/i915_winsys.h create mode 100644 src/gallium/drivers/i965simple/Makefile create mode 100644 src/gallium/drivers/i965simple/SConscript create mode 100644 src/gallium/drivers/i965simple/brw_batch.h create mode 100644 src/gallium/drivers/i965simple/brw_blit.c create mode 100644 src/gallium/drivers/i965simple/brw_blit.h create mode 100644 src/gallium/drivers/i965simple/brw_cc.c create mode 100644 src/gallium/drivers/i965simple/brw_clip.c create mode 100644 src/gallium/drivers/i965simple/brw_clip.h create mode 100644 src/gallium/drivers/i965simple/brw_clip_line.c create mode 100644 src/gallium/drivers/i965simple/brw_clip_point.c create mode 100644 src/gallium/drivers/i965simple/brw_clip_state.c create mode 100644 src/gallium/drivers/i965simple/brw_clip_tri.c create mode 100644 src/gallium/drivers/i965simple/brw_clip_unfilled.c create mode 100644 src/gallium/drivers/i965simple/brw_clip_util.c create mode 100644 src/gallium/drivers/i965simple/brw_context.c create mode 100644 src/gallium/drivers/i965simple/brw_context.h create mode 100644 src/gallium/drivers/i965simple/brw_curbe.c create mode 100644 src/gallium/drivers/i965simple/brw_defines.h create mode 100644 src/gallium/drivers/i965simple/brw_draw.c create mode 100644 src/gallium/drivers/i965simple/brw_draw.h create mode 100644 src/gallium/drivers/i965simple/brw_draw_upload.c create mode 100644 src/gallium/drivers/i965simple/brw_eu.c create mode 100644 src/gallium/drivers/i965simple/brw_eu.h create mode 100644 src/gallium/drivers/i965simple/brw_eu_debug.c create mode 100644 src/gallium/drivers/i965simple/brw_eu_emit.c create mode 100644 src/gallium/drivers/i965simple/brw_eu_util.c create mode 100644 src/gallium/drivers/i965simple/brw_flush.c create mode 100644 src/gallium/drivers/i965simple/brw_gs.c create mode 100644 src/gallium/drivers/i965simple/brw_gs.h create mode 100644 src/gallium/drivers/i965simple/brw_gs_emit.c create mode 100644 src/gallium/drivers/i965simple/brw_gs_state.c create mode 100644 src/gallium/drivers/i965simple/brw_misc_state.c create mode 100644 src/gallium/drivers/i965simple/brw_reg.h create mode 100644 src/gallium/drivers/i965simple/brw_sf.c create mode 100644 src/gallium/drivers/i965simple/brw_sf.h create mode 100644 src/gallium/drivers/i965simple/brw_sf_emit.c create mode 100644 src/gallium/drivers/i965simple/brw_sf_state.c create mode 100644 src/gallium/drivers/i965simple/brw_shader_info.c create mode 100644 src/gallium/drivers/i965simple/brw_state.c create mode 100644 src/gallium/drivers/i965simple/brw_state.h create mode 100644 src/gallium/drivers/i965simple/brw_state_batch.c create mode 100644 src/gallium/drivers/i965simple/brw_state_cache.c create mode 100644 src/gallium/drivers/i965simple/brw_state_pool.c create mode 100644 src/gallium/drivers/i965simple/brw_state_upload.c create mode 100644 src/gallium/drivers/i965simple/brw_strings.c create mode 100644 src/gallium/drivers/i965simple/brw_structs.h create mode 100644 src/gallium/drivers/i965simple/brw_surface.c create mode 100644 src/gallium/drivers/i965simple/brw_tex_layout.c create mode 100644 src/gallium/drivers/i965simple/brw_tex_layout.h create mode 100644 src/gallium/drivers/i965simple/brw_urb.c create mode 100644 src/gallium/drivers/i965simple/brw_util.c create mode 100644 src/gallium/drivers/i965simple/brw_util.h create mode 100644 src/gallium/drivers/i965simple/brw_vs.c create mode 100644 src/gallium/drivers/i965simple/brw_vs.h create mode 100644 src/gallium/drivers/i965simple/brw_vs_emit.c create mode 100644 src/gallium/drivers/i965simple/brw_vs_state.c create mode 100644 src/gallium/drivers/i965simple/brw_winsys.h create mode 100644 src/gallium/drivers/i965simple/brw_wm.c create mode 100644 src/gallium/drivers/i965simple/brw_wm.h create mode 100644 src/gallium/drivers/i965simple/brw_wm_decl.c create mode 100644 src/gallium/drivers/i965simple/brw_wm_glsl.c create mode 100644 src/gallium/drivers/i965simple/brw_wm_iz.c create mode 100644 src/gallium/drivers/i965simple/brw_wm_sampler_state.c create mode 100644 src/gallium/drivers/i965simple/brw_wm_state.c create mode 100644 src/gallium/drivers/i965simple/brw_wm_surface_state.c create mode 100644 src/gallium/drivers/softpipe/Makefile create mode 100644 src/gallium/drivers/softpipe/SConscript create mode 100644 src/gallium/drivers/softpipe/sp_clear.c create mode 100644 src/gallium/drivers/softpipe/sp_clear.h create mode 100644 src/gallium/drivers/softpipe/sp_context.c create mode 100644 src/gallium/drivers/softpipe/sp_context.h create mode 100644 src/gallium/drivers/softpipe/sp_draw_arrays.c create mode 100644 src/gallium/drivers/softpipe/sp_flush.c create mode 100644 src/gallium/drivers/softpipe/sp_flush.h create mode 100644 src/gallium/drivers/softpipe/sp_headers.h create mode 100644 src/gallium/drivers/softpipe/sp_prim_setup.c create mode 100644 src/gallium/drivers/softpipe/sp_prim_setup.h create mode 100644 src/gallium/drivers/softpipe/sp_prim_vbuf.c create mode 100644 src/gallium/drivers/softpipe/sp_prim_vbuf.h create mode 100644 src/gallium/drivers/softpipe/sp_quad.c create mode 100644 src/gallium/drivers/softpipe/sp_quad.h create mode 100644 src/gallium/drivers/softpipe/sp_quad_alpha_test.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_blend.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_bufloop.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_colormask.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_coverage.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_depth_test.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_earlyz.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_fs.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_occlusion.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_output.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_stencil.c create mode 100644 src/gallium/drivers/softpipe/sp_quad_stipple.c create mode 100644 src/gallium/drivers/softpipe/sp_query.c create mode 100644 src/gallium/drivers/softpipe/sp_query.h create mode 100644 src/gallium/drivers/softpipe/sp_state.h create mode 100644 src/gallium/drivers/softpipe/sp_state_blend.c create mode 100644 src/gallium/drivers/softpipe/sp_state_clip.c create mode 100644 src/gallium/drivers/softpipe/sp_state_derived.c create mode 100644 src/gallium/drivers/softpipe/sp_state_fs.c create mode 100644 src/gallium/drivers/softpipe/sp_state_rasterizer.c create mode 100644 src/gallium/drivers/softpipe/sp_state_sampler.c create mode 100644 src/gallium/drivers/softpipe/sp_state_surface.c create mode 100644 src/gallium/drivers/softpipe/sp_state_vertex.c create mode 100644 src/gallium/drivers/softpipe/sp_surface.c create mode 100644 src/gallium/drivers/softpipe/sp_surface.h create mode 100644 src/gallium/drivers/softpipe/sp_tex_sample.c create mode 100644 src/gallium/drivers/softpipe/sp_tex_sample.h create mode 100644 src/gallium/drivers/softpipe/sp_texture.c create mode 100644 src/gallium/drivers/softpipe/sp_texture.h create mode 100644 src/gallium/drivers/softpipe/sp_tile_cache.c create mode 100644 src/gallium/drivers/softpipe/sp_tile_cache.h create mode 100644 src/gallium/drivers/softpipe/sp_winsys.h create mode 100644 src/gallium/include/pipe/p_compiler.h create mode 100644 src/gallium/include/pipe/p_context.h create mode 100644 src/gallium/include/pipe/p_debug.h create mode 100644 src/gallium/include/pipe/p_defines.h create mode 100644 src/gallium/include/pipe/p_format.h create mode 100644 src/gallium/include/pipe/p_inlines.h create mode 100644 src/gallium/include/pipe/p_shader_tokens.h create mode 100644 src/gallium/include/pipe/p_state.h create mode 100644 src/gallium/include/pipe/p_thread.h create mode 100644 src/gallium/include/pipe/p_util.h create mode 100644 src/gallium/include/pipe/p_winsys.h create mode 100644 src/gallium/winsys/dri/intel/Makefile create mode 100644 src/gallium/winsys/dri/intel/SConscript create mode 100644 src/gallium/winsys/dri/intel/intel_batchbuffer.c create mode 100644 src/gallium/winsys/dri/intel/intel_batchbuffer.h create mode 100644 src/gallium/winsys/dri/intel/intel_batchpool.c create mode 100644 src/gallium/winsys/dri/intel/intel_batchpool.h create mode 100644 src/gallium/winsys/dri/intel/intel_context.c create mode 100644 src/gallium/winsys/dri/intel/intel_context.h create mode 100644 src/gallium/winsys/dri/intel/intel_lock.c create mode 100644 src/gallium/winsys/dri/intel/intel_reg.h create mode 100644 src/gallium/winsys/dri/intel/intel_screen.c create mode 100644 src/gallium/winsys/dri/intel/intel_screen.h create mode 100644 src/gallium/winsys/dri/intel/intel_swapbuffers.c create mode 100644 src/gallium/winsys/dri/intel/intel_swapbuffers.h create mode 100644 src/gallium/winsys/dri/intel/intel_winsys.h create mode 100644 src/gallium/winsys/dri/intel/intel_winsys_i915.c create mode 100644 src/gallium/winsys/dri/intel/intel_winsys_pipe.c create mode 100644 src/gallium/winsys/dri/intel/intel_winsys_softpipe.c create mode 100644 src/gallium/winsys/dri/intel/server/i830_common.h create mode 100644 src/gallium/winsys/dri/intel/server/i830_dri.h create mode 100644 src/gallium/winsys/dri/intel/server/intel.h create mode 100644 src/gallium/winsys/dri/intel/server/intel_dri.c create mode 100644 src/gallium/winsys/xlib/brw_aub.c create mode 100644 src/gallium/winsys/xlib/brw_aub.h create mode 100644 src/gallium/winsys/xlib/fakeglx.c create mode 100644 src/gallium/winsys/xlib/glxapi.c create mode 100644 src/gallium/winsys/xlib/glxapi.h create mode 100644 src/gallium/winsys/xlib/glxheader.h create mode 100644 src/gallium/winsys/xlib/realglx.c create mode 100644 src/gallium/winsys/xlib/realglx.h create mode 100644 src/gallium/winsys/xlib/xfonts.c create mode 100644 src/gallium/winsys/xlib/xfonts.h create mode 100644 src/gallium/winsys/xlib/xm_api.c create mode 100644 src/gallium/winsys/xlib/xm_image.c create mode 100644 src/gallium/winsys/xlib/xm_image.h create mode 100644 src/gallium/winsys/xlib/xm_winsys.c create mode 100644 src/gallium/winsys/xlib/xm_winsys_aub.c create mode 100644 src/gallium/winsys/xlib/xm_winsys_aub.h create mode 100644 src/gallium/winsys/xlib/xmesaP.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/Makefile delete mode 100644 src/mesa/drivers/dri/intel_winsys/SConscript delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_batchpool.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_batchpool.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_context.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_context.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_lock.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_reg.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_screen.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_screen.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_winsys.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c delete mode 100644 src/mesa/drivers/dri/intel_winsys/server/i830_common.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/server/i830_dri.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/server/intel.h delete mode 100644 src/mesa/drivers/dri/intel_winsys/server/intel_dri.c delete mode 100644 src/mesa/pipe/Makefile delete mode 100644 src/mesa/pipe/Makefile.template delete mode 100644 src/mesa/pipe/README.portability delete mode 100644 src/mesa/pipe/SConscript delete mode 100644 src/mesa/pipe/cell/Makefile delete mode 100644 src/mesa/pipe/cell/common.h delete mode 100644 src/mesa/pipe/cell/ppu/Makefile delete mode 100644 src/mesa/pipe/cell/ppu/cell_batch.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_batch.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_clear.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_clear.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_context.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_context.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_draw_arrays.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_draw_arrays.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_flush.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_flush.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_render.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_render.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_spu.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_spu.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_state.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_blend.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_clip.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_derived.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_emit.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_emit.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_fs.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_rasterizer.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_sampler.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_surface.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_state_vertex.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_surface.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_surface.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_texture.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_texture.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_vbuf.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_vbuf.h delete mode 100644 src/mesa/pipe/cell/ppu/cell_vertex_shader.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_winsys.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_winsys.h delete mode 100644 src/mesa/pipe/cell/spu/Makefile delete mode 100644 src/mesa/pipe/cell/spu/spu_blend.c delete mode 100644 src/mesa/pipe/cell/spu/spu_blend.h delete mode 100644 src/mesa/pipe/cell/spu/spu_colorpack.h delete mode 100644 src/mesa/pipe/cell/spu/spu_exec.c delete mode 100644 src/mesa/pipe/cell/spu/spu_exec.h delete mode 100644 src/mesa/pipe/cell/spu/spu_main.c delete mode 100644 src/mesa/pipe/cell/spu/spu_main.h delete mode 100644 src/mesa/pipe/cell/spu/spu_render.c delete mode 100644 src/mesa/pipe/cell/spu/spu_render.h delete mode 100644 src/mesa/pipe/cell/spu/spu_texture.c delete mode 100644 src/mesa/pipe/cell/spu/spu_texture.h delete mode 100644 src/mesa/pipe/cell/spu/spu_tile.c delete mode 100644 src/mesa/pipe/cell/spu/spu_tile.h delete mode 100644 src/mesa/pipe/cell/spu/spu_tri.c delete mode 100644 src/mesa/pipe/cell/spu/spu_tri.h delete mode 100644 src/mesa/pipe/cell/spu/spu_util.c delete mode 100644 src/mesa/pipe/cell/spu/spu_vertex_fetch.c delete mode 100644 src/mesa/pipe/cell/spu/spu_vertex_shader.c delete mode 100644 src/mesa/pipe/cell/spu/spu_vertex_shader.h delete mode 100644 src/mesa/pipe/cell/spu/spu_ztest.h delete mode 100644 src/mesa/pipe/cso_cache/cso_cache.c delete mode 100644 src/mesa/pipe/cso_cache/cso_cache.h delete mode 100644 src/mesa/pipe/cso_cache/cso_hash.c delete mode 100644 src/mesa/pipe/cso_cache/cso_hash.h delete mode 100644 src/mesa/pipe/draw/Makefile delete mode 100644 src/mesa/pipe/draw/draw_clip.c delete mode 100644 src/mesa/pipe/draw/draw_context.c delete mode 100644 src/mesa/pipe/draw/draw_context.h delete mode 100644 src/mesa/pipe/draw/draw_cull.c delete mode 100644 src/mesa/pipe/draw/draw_debug.c delete mode 100644 src/mesa/pipe/draw/draw_flatshade.c delete mode 100644 src/mesa/pipe/draw/draw_offset.c delete mode 100644 src/mesa/pipe/draw/draw_prim.c delete mode 100644 src/mesa/pipe/draw/draw_private.h delete mode 100644 src/mesa/pipe/draw/draw_stipple.c delete mode 100644 src/mesa/pipe/draw/draw_twoside.c delete mode 100644 src/mesa/pipe/draw/draw_unfilled.c delete mode 100644 src/mesa/pipe/draw/draw_validate.c delete mode 100644 src/mesa/pipe/draw/draw_vbuf.c delete mode 100644 src/mesa/pipe/draw/draw_vbuf.h delete mode 100644 src/mesa/pipe/draw/draw_vertex.c delete mode 100644 src/mesa/pipe/draw/draw_vertex.h delete mode 100644 src/mesa/pipe/draw/draw_vertex_cache.c delete mode 100644 src/mesa/pipe/draw/draw_vertex_fetch.c delete mode 100644 src/mesa/pipe/draw/draw_vertex_shader.c delete mode 100644 src/mesa/pipe/draw/draw_vf.c delete mode 100644 src/mesa/pipe/draw/draw_vf.h delete mode 100644 src/mesa/pipe/draw/draw_vf_generic.c delete mode 100644 src/mesa/pipe/draw/draw_vf_sse.c delete mode 100644 src/mesa/pipe/draw/draw_wide_prims.c delete mode 100644 src/mesa/pipe/failover/Makefile delete mode 100644 src/mesa/pipe/failover/fo_context.c delete mode 100644 src/mesa/pipe/failover/fo_context.h delete mode 100644 src/mesa/pipe/failover/fo_state.c delete mode 100644 src/mesa/pipe/failover/fo_state_emit.c delete mode 100644 src/mesa/pipe/failover/fo_winsys.h delete mode 100644 src/mesa/pipe/i915simple/Makefile delete mode 100644 src/mesa/pipe/i915simple/SConscript delete mode 100644 src/mesa/pipe/i915simple/i915_batch.h delete mode 100644 src/mesa/pipe/i915simple/i915_blit.c delete mode 100644 src/mesa/pipe/i915simple/i915_blit.h delete mode 100644 src/mesa/pipe/i915simple/i915_clear.c delete mode 100644 src/mesa/pipe/i915simple/i915_context.c delete mode 100644 src/mesa/pipe/i915simple/i915_context.h delete mode 100644 src/mesa/pipe/i915simple/i915_debug.c delete mode 100644 src/mesa/pipe/i915simple/i915_debug.h delete mode 100644 src/mesa/pipe/i915simple/i915_debug_fp.c delete mode 100644 src/mesa/pipe/i915simple/i915_flush.c delete mode 100644 src/mesa/pipe/i915simple/i915_fpc.h delete mode 100644 src/mesa/pipe/i915simple/i915_fpc_emit.c delete mode 100644 src/mesa/pipe/i915simple/i915_fpc_translate.c delete mode 100644 src/mesa/pipe/i915simple/i915_prim_emit.c delete mode 100644 src/mesa/pipe/i915simple/i915_prim_vbuf.c delete mode 100644 src/mesa/pipe/i915simple/i915_reg.h delete mode 100644 src/mesa/pipe/i915simple/i915_state.c delete mode 100644 src/mesa/pipe/i915simple/i915_state.h delete mode 100644 src/mesa/pipe/i915simple/i915_state_derived.c delete mode 100644 src/mesa/pipe/i915simple/i915_state_dynamic.c delete mode 100644 src/mesa/pipe/i915simple/i915_state_emit.c delete mode 100644 src/mesa/pipe/i915simple/i915_state_immediate.c delete mode 100644 src/mesa/pipe/i915simple/i915_state_inlines.h delete mode 100644 src/mesa/pipe/i915simple/i915_state_sampler.c delete mode 100644 src/mesa/pipe/i915simple/i915_strings.c delete mode 100644 src/mesa/pipe/i915simple/i915_surface.c delete mode 100644 src/mesa/pipe/i915simple/i915_texture.c delete mode 100644 src/mesa/pipe/i915simple/i915_texture.h delete mode 100644 src/mesa/pipe/i915simple/i915_winsys.h delete mode 100644 src/mesa/pipe/i965simple/Makefile delete mode 100644 src/mesa/pipe/i965simple/SConscript delete mode 100644 src/mesa/pipe/i965simple/brw_batch.h delete mode 100644 src/mesa/pipe/i965simple/brw_blit.c delete mode 100644 src/mesa/pipe/i965simple/brw_blit.h delete mode 100644 src/mesa/pipe/i965simple/brw_cc.c delete mode 100644 src/mesa/pipe/i965simple/brw_clip.c delete mode 100644 src/mesa/pipe/i965simple/brw_clip.h delete mode 100644 src/mesa/pipe/i965simple/brw_clip_line.c delete mode 100644 src/mesa/pipe/i965simple/brw_clip_point.c delete mode 100644 src/mesa/pipe/i965simple/brw_clip_state.c delete mode 100644 src/mesa/pipe/i965simple/brw_clip_tri.c delete mode 100644 src/mesa/pipe/i965simple/brw_clip_unfilled.c delete mode 100644 src/mesa/pipe/i965simple/brw_clip_util.c delete mode 100644 src/mesa/pipe/i965simple/brw_context.c delete mode 100644 src/mesa/pipe/i965simple/brw_context.h delete mode 100644 src/mesa/pipe/i965simple/brw_curbe.c delete mode 100644 src/mesa/pipe/i965simple/brw_defines.h delete mode 100644 src/mesa/pipe/i965simple/brw_draw.c delete mode 100644 src/mesa/pipe/i965simple/brw_draw.h delete mode 100644 src/mesa/pipe/i965simple/brw_draw_upload.c delete mode 100644 src/mesa/pipe/i965simple/brw_eu.c delete mode 100644 src/mesa/pipe/i965simple/brw_eu.h delete mode 100644 src/mesa/pipe/i965simple/brw_eu_debug.c delete mode 100644 src/mesa/pipe/i965simple/brw_eu_emit.c delete mode 100644 src/mesa/pipe/i965simple/brw_eu_util.c delete mode 100644 src/mesa/pipe/i965simple/brw_flush.c delete mode 100644 src/mesa/pipe/i965simple/brw_gs.c delete mode 100644 src/mesa/pipe/i965simple/brw_gs.h delete mode 100644 src/mesa/pipe/i965simple/brw_gs_emit.c delete mode 100644 src/mesa/pipe/i965simple/brw_gs_state.c delete mode 100644 src/mesa/pipe/i965simple/brw_misc_state.c delete mode 100644 src/mesa/pipe/i965simple/brw_reg.h delete mode 100644 src/mesa/pipe/i965simple/brw_sf.c delete mode 100644 src/mesa/pipe/i965simple/brw_sf.h delete mode 100644 src/mesa/pipe/i965simple/brw_sf_emit.c delete mode 100644 src/mesa/pipe/i965simple/brw_sf_state.c delete mode 100644 src/mesa/pipe/i965simple/brw_shader_info.c delete mode 100644 src/mesa/pipe/i965simple/brw_state.c delete mode 100644 src/mesa/pipe/i965simple/brw_state.h delete mode 100644 src/mesa/pipe/i965simple/brw_state_batch.c delete mode 100644 src/mesa/pipe/i965simple/brw_state_cache.c delete mode 100644 src/mesa/pipe/i965simple/brw_state_pool.c delete mode 100644 src/mesa/pipe/i965simple/brw_state_upload.c delete mode 100644 src/mesa/pipe/i965simple/brw_strings.c delete mode 100644 src/mesa/pipe/i965simple/brw_structs.h delete mode 100644 src/mesa/pipe/i965simple/brw_surface.c delete mode 100644 src/mesa/pipe/i965simple/brw_tex_layout.c delete mode 100644 src/mesa/pipe/i965simple/brw_tex_layout.h delete mode 100644 src/mesa/pipe/i965simple/brw_urb.c delete mode 100644 src/mesa/pipe/i965simple/brw_util.c delete mode 100644 src/mesa/pipe/i965simple/brw_util.h delete mode 100644 src/mesa/pipe/i965simple/brw_vs.c delete mode 100644 src/mesa/pipe/i965simple/brw_vs.h delete mode 100644 src/mesa/pipe/i965simple/brw_vs_emit.c delete mode 100644 src/mesa/pipe/i965simple/brw_vs_state.c delete mode 100644 src/mesa/pipe/i965simple/brw_winsys.h delete mode 100644 src/mesa/pipe/i965simple/brw_wm.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm.h delete mode 100644 src/mesa/pipe/i965simple/brw_wm_decl.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm_glsl.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm_iz.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm_sampler_state.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm_state.c delete mode 100644 src/mesa/pipe/i965simple/brw_wm_surface_state.c delete mode 100644 src/mesa/pipe/llvm/Makefile delete mode 100644 src/mesa/pipe/llvm/gallivm.cpp delete mode 100644 src/mesa/pipe/llvm/gallivm.h delete mode 100644 src/mesa/pipe/llvm/gallivm_builtins.cpp delete mode 100644 src/mesa/pipe/llvm/gallivm_cpu.cpp delete mode 100644 src/mesa/pipe/llvm/gallivm_p.h delete mode 100644 src/mesa/pipe/llvm/instructions.cpp delete mode 100644 src/mesa/pipe/llvm/instructions.h delete mode 100644 src/mesa/pipe/llvm/instructionssoa.cpp delete mode 100644 src/mesa/pipe/llvm/instructionssoa.h delete mode 100644 src/mesa/pipe/llvm/llvm_builtins.c delete mode 100644 src/mesa/pipe/llvm/loweringpass.cpp delete mode 100644 src/mesa/pipe/llvm/loweringpass.h delete mode 100644 src/mesa/pipe/llvm/storage.cpp delete mode 100644 src/mesa/pipe/llvm/storage.h delete mode 100644 src/mesa/pipe/llvm/storagesoa.cpp delete mode 100644 src/mesa/pipe/llvm/storagesoa.h delete mode 100644 src/mesa/pipe/llvm/tgsitollvm.cpp delete mode 100644 src/mesa/pipe/llvm/tgsitollvm.h delete mode 100644 src/mesa/pipe/p_compiler.h delete mode 100644 src/mesa/pipe/p_context.h delete mode 100644 src/mesa/pipe/p_debug.h delete mode 100644 src/mesa/pipe/p_defines.h delete mode 100644 src/mesa/pipe/p_format.h delete mode 100644 src/mesa/pipe/p_inlines.h delete mode 100644 src/mesa/pipe/p_shader_tokens.h delete mode 100644 src/mesa/pipe/p_state.h delete mode 100644 src/mesa/pipe/p_thread.h delete mode 100644 src/mesa/pipe/p_util.h delete mode 100644 src/mesa/pipe/p_winsys.h delete mode 100644 src/mesa/pipe/pipebuffer/Makefile delete mode 100644 src/mesa/pipe/pipebuffer/linked_list.h delete mode 100644 src/mesa/pipe/pipebuffer/pb_buffer.h delete mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_fenced.c delete mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_fenced.h delete mode 100644 src/mesa/pipe/pipebuffer/pb_buffer_malloc.c delete mode 100644 src/mesa/pipe/pipebuffer/pb_bufmgr.h delete mode 100644 src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c delete mode 100644 src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c delete mode 100644 src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c delete mode 100644 src/mesa/pipe/pipebuffer/pb_winsys.c delete mode 100644 src/mesa/pipe/softpipe/Makefile delete mode 100644 src/mesa/pipe/softpipe/SConscript delete mode 100644 src/mesa/pipe/softpipe/sp_clear.c delete mode 100644 src/mesa/pipe/softpipe/sp_clear.h delete mode 100644 src/mesa/pipe/softpipe/sp_context.c delete mode 100644 src/mesa/pipe/softpipe/sp_context.h delete mode 100644 src/mesa/pipe/softpipe/sp_draw_arrays.c delete mode 100644 src/mesa/pipe/softpipe/sp_flush.c delete mode 100644 src/mesa/pipe/softpipe/sp_flush.h delete mode 100644 src/mesa/pipe/softpipe/sp_headers.h delete mode 100644 src/mesa/pipe/softpipe/sp_prim_setup.c delete mode 100644 src/mesa/pipe/softpipe/sp_prim_setup.h delete mode 100644 src/mesa/pipe/softpipe/sp_prim_vbuf.c delete mode 100644 src/mesa/pipe/softpipe/sp_prim_vbuf.h delete mode 100644 src/mesa/pipe/softpipe/sp_quad.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad.h delete mode 100644 src/mesa/pipe/softpipe/sp_quad_alpha_test.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_blend.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_bufloop.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_colormask.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_coverage.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_depth_test.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_earlyz.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_fs.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_occlusion.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_output.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_stencil.c delete mode 100644 src/mesa/pipe/softpipe/sp_quad_stipple.c delete mode 100644 src/mesa/pipe/softpipe/sp_query.c delete mode 100644 src/mesa/pipe/softpipe/sp_query.h delete mode 100644 src/mesa/pipe/softpipe/sp_state.h delete mode 100644 src/mesa/pipe/softpipe/sp_state_blend.c delete mode 100644 src/mesa/pipe/softpipe/sp_state_clip.c delete mode 100644 src/mesa/pipe/softpipe/sp_state_derived.c delete mode 100644 src/mesa/pipe/softpipe/sp_state_fs.c delete mode 100644 src/mesa/pipe/softpipe/sp_state_rasterizer.c delete mode 100644 src/mesa/pipe/softpipe/sp_state_sampler.c delete mode 100644 src/mesa/pipe/softpipe/sp_state_surface.c delete mode 100644 src/mesa/pipe/softpipe/sp_state_vertex.c delete mode 100644 src/mesa/pipe/softpipe/sp_surface.c delete mode 100644 src/mesa/pipe/softpipe/sp_surface.h delete mode 100644 src/mesa/pipe/softpipe/sp_tex_sample.c delete mode 100644 src/mesa/pipe/softpipe/sp_tex_sample.h delete mode 100644 src/mesa/pipe/softpipe/sp_texture.c delete mode 100644 src/mesa/pipe/softpipe/sp_texture.h delete mode 100644 src/mesa/pipe/softpipe/sp_tile_cache.c delete mode 100644 src/mesa/pipe/softpipe/sp_tile_cache.h delete mode 100644 src/mesa/pipe/softpipe/sp_winsys.h delete mode 100644 src/mesa/pipe/tgsi/Makefile delete mode 100644 src/mesa/pipe/tgsi/exec/Makefile delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_exec.c delete mode 100644 src/mesa/pipe/tgsi/exec/tgsi_exec.h delete mode 100755 src/mesa/pipe/tgsi/exec/tgsi_sse2.c delete mode 100755 src/mesa/pipe/tgsi/exec/tgsi_sse2.h delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_build.c delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_build.h delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_dump.c delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_dump.h delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_parse.c delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_parse.h delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_transform.c delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_transform.h delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_util.c delete mode 100644 src/mesa/pipe/tgsi/util/tgsi_util.h delete mode 100644 src/mesa/pipe/util/p_debug.c delete mode 100644 src/mesa/pipe/util/p_tile.c delete mode 100644 src/mesa/pipe/util/p_tile.h delete mode 100644 src/mesa/pipe/util/p_util.c delete mode 100644 src/mesa/pipe/xlib/brw_aub.c delete mode 100644 src/mesa/pipe/xlib/brw_aub.h delete mode 100644 src/mesa/pipe/xlib/fakeglx.c delete mode 100644 src/mesa/pipe/xlib/glxapi.c delete mode 100644 src/mesa/pipe/xlib/glxapi.h delete mode 100644 src/mesa/pipe/xlib/glxheader.h delete mode 100644 src/mesa/pipe/xlib/realglx.c delete mode 100644 src/mesa/pipe/xlib/realglx.h delete mode 100644 src/mesa/pipe/xlib/xfonts.c delete mode 100644 src/mesa/pipe/xlib/xfonts.h delete mode 100644 src/mesa/pipe/xlib/xm_api.c delete mode 100644 src/mesa/pipe/xlib/xm_image.c delete mode 100644 src/mesa/pipe/xlib/xm_image.h delete mode 100644 src/mesa/pipe/xlib/xm_winsys.c delete mode 100644 src/mesa/pipe/xlib/xm_winsys_aub.c delete mode 100644 src/mesa/pipe/xlib/xm_winsys_aub.h delete mode 100644 src/mesa/pipe/xlib/xmesaP.h (limited to 'src') diff --git a/src/gallium/Makefile b/src/gallium/Makefile new file mode 100644 index 0000000000..d880d090c1 --- /dev/null +++ b/src/gallium/Makefile @@ -0,0 +1,28 @@ +TOP = ../../.. +include $(TOP)/configs/current + + +ifeq ($(CONFIG_NAME), linux-cell) +CELL_DIR = cell +endif + +ifeq ($(CONFIG_NAME), linux-llvm) +LLVM_DIR = llvm +endif + +SUBDIRS = softpipe i915simple i965simple failover pipebuffer $(CELL_DIR) $(LLVM_DIR) + + +default: subdirs + + +subdirs: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +clean: + rm -f `find . -name \*.[oa]` diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template new file mode 100644 index 0000000000..8e84f8eb2d --- /dev/null +++ b/src/gallium/Makefile.template @@ -0,0 +1,63 @@ +# -*-makefile-*- + + +# We still have a dependency on the "dri" buffer manager. Most likely +# the interface can be reused in non-dri environments, and also as a +# frontend to simpler memory managers. +# +COMMON_SOURCES = + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(CPP_SOURCES:.cpp=.o) \ + $(ASM_SOURCES:.S=.o) + + +### Include directories +INCLUDES = \ + -I. \ + -I$(TOP)/src/mesa/pipe \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include \ + $(DRIVER_INCLUDES) + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + + +##### TARGETS ##### + +default: depend symlinks $(LIBNAME) + + +$(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/mesa/pipe/Makefile.template + $(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS) + + +depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ + $(ASM_SOURCES) 2> /dev/null + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean:: + -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) + -rm -f depend depend.bak + + +include depend diff --git a/src/gallium/README.portability b/src/gallium/README.portability new file mode 100644 index 0000000000..c70ca774da --- /dev/null +++ b/src/gallium/README.portability @@ -0,0 +1,43 @@ + CROSS-PLATFORM PORTABILITY GUIDELINES FOR GALLIUM3D + + += General Considerations = + +The state tracker and winsys driver support a rather limited number of +platforms. However, the pipe drivers are meant to run in a wide number of +platforms. Hence the pipe drivers, the auxiliary modules, and all public +headers in general, should stricly follow these guidelines to ensure + + += Compiler Support = + +* Include the p_compiler.h. + +* Don't use the 'inline' keyword, use the INLINE macro in p_compiler.h instead. + +* Cast explicitly when converting to integer types of smaller sizes. + +* Cast explicitly when converting between float, double and integral types. + +* Don't use named struct initializers. + +* Don't use variable number of macro arguments. Use static inline functions +instead. + + += Standard Library = + +* Avoid including standard library headers. Most standard library functions are +not available in Windows Kernel Mode. Use the appropriate p_*.h include. + +== Memory Allocation == + +* Use MALLOC, CALLOC, FREE instead of the malloc, calloc, free functions. + +* Use align_pointer() function defined in p_util.h for aligning pointers in a +portable way. + +== Debugging == + +TODO + diff --git a/src/gallium/SConscript b/src/gallium/SConscript new file mode 100644 index 0000000000..d9c20e0100 --- /dev/null +++ b/src/gallium/SConscript @@ -0,0 +1,9 @@ +Import('*') + +#env = env.Clone() + +SConscript([ + 'softpipe/SConscript', + 'i915simple/SConscript', + 'i965simple/SConscript', +]) diff --git a/src/gallium/aux/cso_cache/cso_cache.c b/src/gallium/aux/cso_cache/cso_cache.c new file mode 100644 index 0000000000..9e77e0774d --- /dev/null +++ b/src/gallium/aux/cso_cache/cso_cache.c @@ -0,0 +1,181 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Zack Rusin + */ + +#include "cso_cache.h" +#include "cso_hash.h" + +#if 1 +static unsigned hash_key(const void *key, unsigned key_size) +{ + unsigned *ikey = (unsigned *)key; + unsigned hash = 0, i; + + assert(key_size % 4 == 0); + + /* I'm sure this can be improved on: + */ + for (i = 0; i < key_size/4; i++) + hash ^= ikey[i]; + + return hash; +} +#else +static unsigned hash_key(const unsigned char *p, int n) +{ + unsigned h = 0; + unsigned g; + + while (n--) { + h = (h << 4) + *p++; + if ((g = (h & 0xf0000000)) != 0) + h ^= g >> 23; + h &= ~g; + } + return h; +} +#endif + +unsigned cso_construct_key(void *item, int item_size) +{ + return hash_key((item), item_size); +} + +static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_type type) +{ + struct cso_hash *hash = 0; + + switch(type) { + case CSO_BLEND: + hash = sc->blend_hash; + break; + case CSO_SAMPLER: + hash = sc->sampler_hash; + break; + case CSO_DEPTH_STENCIL_ALPHA: + hash = sc->depth_stencil_hash; + break; + case CSO_RASTERIZER: + hash = sc->rasterizer_hash; + break; + case CSO_FRAGMENT_SHADER: + hash = sc->fs_hash; + break; + case CSO_VERTEX_SHADER: + hash = sc->vs_hash; + break; + } + + return hash; +} + +static int _cso_size_for_type(enum cso_cache_type type) +{ + switch(type) { + case CSO_BLEND: + return sizeof(struct pipe_blend_state); + case CSO_SAMPLER: + return sizeof(struct pipe_sampler_state); + case CSO_DEPTH_STENCIL_ALPHA: + return sizeof(struct pipe_depth_stencil_alpha_state); + case CSO_RASTERIZER: + return sizeof(struct pipe_rasterizer_state); + case CSO_FRAGMENT_SHADER: + return sizeof(struct pipe_shader_state); + case CSO_VERTEX_SHADER: + return sizeof(struct pipe_shader_state); + } + return 0; +} + +struct cso_hash_iter +cso_insert_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *state) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + return cso_hash_insert(hash, hash_key, state); +} + +struct cso_hash_iter +cso_find_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + + return cso_hash_find(hash, hash_key); +} + +struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *templ) +{ + struct cso_hash_iter iter = cso_find_state(sc, hash_key, type); + int size = _cso_size_for_type(type); + while (!cso_hash_iter_is_null(iter)) { + void *iter_data = cso_hash_iter_data(iter); + if (!memcmp(iter_data, templ, size)) + return iter; + iter = cso_hash_iter_next(iter); + } + return iter; +} + +void * cso_take_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + return cso_hash_take(hash, hash_key); +} + +struct cso_cache *cso_cache_create(void) +{ + struct cso_cache *sc = malloc(sizeof(struct cso_cache)); + + sc->blend_hash = cso_hash_create(); + sc->sampler_hash = cso_hash_create(); + sc->depth_stencil_hash = cso_hash_create(); + sc->rasterizer_hash = cso_hash_create(); + sc->fs_hash = cso_hash_create(); + sc->vs_hash = cso_hash_create(); + + return sc; +} + +void cso_cache_delete(struct cso_cache *sc) +{ + assert(sc); + cso_hash_delete(sc->blend_hash); + cso_hash_delete(sc->sampler_hash); + cso_hash_delete(sc->depth_stencil_hash); + cso_hash_delete(sc->rasterizer_hash); + cso_hash_delete(sc->fs_hash); + cso_hash_delete(sc->vs_hash); + free(sc); +} diff --git a/src/gallium/aux/cso_cache/cso_cache.h b/src/gallium/aux/cso_cache/cso_cache.h new file mode 100644 index 0000000000..116e2eaa2c --- /dev/null +++ b/src/gallium/aux/cso_cache/cso_cache.h @@ -0,0 +1,107 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#ifndef CSO_CACHE_H +#define CSO_CACHE_H + +#include "pipe/p_context.h" +#include "pipe/p_state.h" + + +struct cso_hash; + +struct cso_cache { + struct cso_hash *blend_hash; + struct cso_hash *depth_stencil_hash; + struct cso_hash *fs_hash; + struct cso_hash *vs_hash; + struct cso_hash *rasterizer_hash; + struct cso_hash *sampler_hash; +}; + +struct cso_blend { + struct pipe_blend_state state; + void *data; +}; + +struct cso_depth_stencil_alpha { + struct pipe_depth_stencil_alpha_state state; + void *data; +}; + +struct cso_rasterizer { + struct pipe_rasterizer_state state; + void *data; +}; + +struct cso_fragment_shader { + struct pipe_shader_state state; + void *data; +}; + +struct cso_vertex_shader { + struct pipe_shader_state state; + void *data; +}; + +struct cso_sampler { + struct pipe_sampler_state state; + void *data; +}; + + +enum cso_cache_type { + CSO_BLEND, + CSO_SAMPLER, + CSO_DEPTH_STENCIL_ALPHA, + CSO_RASTERIZER, + CSO_FRAGMENT_SHADER, + CSO_VERTEX_SHADER +}; + +unsigned cso_construct_key(void *item, int item_size); + +struct cso_cache *cso_cache_create(void); +void cso_cache_delete(struct cso_cache *sc); + +struct cso_hash_iter cso_insert_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *state); +struct cso_hash_iter cso_find_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type); +struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *templ); +void * cso_take_state(struct cso_cache *sc, unsigned hash_key, + enum cso_cache_type type); + +#endif diff --git a/src/gallium/aux/cso_cache/cso_hash.c b/src/gallium/aux/cso_cache/cso_hash.c new file mode 100644 index 0000000000..0338cb3b47 --- /dev/null +++ b/src/gallium/aux/cso_cache/cso_hash.c @@ -0,0 +1,388 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#include "cso_hash.h" + +#include +#include +#include +#include + +#define MAX(a, b) ((a > b) ? (a) : (b)) + +static const int MinNumBits = 4; + +static const unsigned char prime_deltas[] = { + 0, 0, 1, 3, 1, 5, 3, 3, 1, 9, 7, 5, 3, 9, 25, 3, + 1, 21, 3, 21, 7, 15, 9, 5, 3, 29, 15, 0, 0, 0, 0, 0 +}; + +static int primeForNumBits(int numBits) +{ + return (1 << numBits) + prime_deltas[numBits]; +} + +/* + Returns the smallest integer n such that + primeForNumBits(n) >= hint. +*/ +static int countBits(int hint) +{ + int numBits = 0; + int bits = hint; + + while (bits > 1) { + bits >>= 1; + numBits++; + } + + if (numBits >= (int)sizeof(prime_deltas)) { + numBits = sizeof(prime_deltas) - 1; + } else if (primeForNumBits(numBits) < hint) { + ++numBits; + } + return numBits; +} + +struct cso_node { + struct cso_node *next; + unsigned key; + void *value; +}; + +struct cso_hash_data { + struct cso_node *fakeNext; + struct cso_node **buckets; + int size; + int nodeSize; + short userNumBits; + short numBits; + int numBuckets; +}; + +struct cso_hash { + union { + struct cso_hash_data *d; + struct cso_node *e; + } data; +}; + +static void *cso_data_allocate_node(struct cso_hash_data *hash) +{ + return malloc(hash->nodeSize); +} + +static void cso_data_free_node(struct cso_node *node) +{ + /* XXX still a leak here. + * Need to cast value ptr to original cso type, then free the + * driver-specific data hanging off of it. For example: + struct cso_sampler *csamp = (struct cso_sampler *) node->value; + free(csamp->data); + */ + free(node->value); + free(node); +} + +static struct cso_node * +cso_hash_create_node(struct cso_hash *hash, + unsigned akey, void *avalue, + struct cso_node **anextNode) +{ + struct cso_node *node = cso_data_allocate_node(hash->data.d); + node->key = akey; + node->value = avalue; + + node->next = (struct cso_node*)(*anextNode); + *anextNode = node; + ++hash->data.d->size; + return node; +} + +static void cso_data_rehash(struct cso_hash_data *hash, int hint) +{ + if (hint < 0) { + hint = countBits(-hint); + if (hint < MinNumBits) + hint = MinNumBits; + hash->userNumBits = hint; + while (primeForNumBits(hint) < (hash->size >> 1)) + ++hint; + } else if (hint < MinNumBits) { + hint = MinNumBits; + } + + if (hash->numBits != hint) { + struct cso_node *e = (struct cso_node *)(hash); + struct cso_node **oldBuckets = hash->buckets; + int oldNumBuckets = hash->numBuckets; + int i = 0; + + hash->numBits = hint; + hash->numBuckets = primeForNumBits(hint); + hash->buckets = malloc(sizeof(struct cso_node*) * hash->numBuckets); + for (i = 0; i < hash->numBuckets; ++i) + hash->buckets[i] = e; + + for (i = 0; i < oldNumBuckets; ++i) { + struct cso_node *firstNode = oldBuckets[i]; + while (firstNode != e) { + unsigned h = firstNode->key; + struct cso_node *lastNode = firstNode; + while (lastNode->next != e && lastNode->next->key == h) + lastNode = lastNode->next; + + struct cso_node *afterLastNode = lastNode->next; + struct cso_node **beforeFirstNode = &hash->buckets[h % hash->numBuckets]; + while (*beforeFirstNode != e) + beforeFirstNode = &(*beforeFirstNode)->next; + lastNode->next = *beforeFirstNode; + *beforeFirstNode = firstNode; + firstNode = afterLastNode; + } + } + free(oldBuckets); + } +} + +static void cso_data_might_grow(struct cso_hash_data *hash) +{ + if (hash->size >= hash->numBuckets) + cso_data_rehash(hash, hash->numBits + 1); +} + +static void cso_data_has_shrunk(struct cso_hash_data *hash) +{ + if (hash->size <= (hash->numBuckets >> 3) && + hash->numBits > hash->userNumBits) { + int max = MAX(hash->numBits-2, hash->userNumBits); + cso_data_rehash(hash, max); + } +} + +static struct cso_node *cso_data_first_node(struct cso_hash_data *hash) +{ + struct cso_node *e = (struct cso_node *)(hash); + struct cso_node **bucket = hash->buckets; + int n = hash->numBuckets; + while (n--) { + if (*bucket != e) + return *bucket; + ++bucket; + } + return e; +} + +static struct cso_node **cso_hash_find_node(struct cso_hash *hash, unsigned akey) +{ + struct cso_node **node; + + if (hash->data.d->numBuckets) { + node = (struct cso_node **)(&hash->data.d->buckets[akey % hash->data.d->numBuckets]); + assert(*node == hash->data.e || (*node)->next); + while (*node != hash->data.e && (*node)->key != akey) + node = &(*node)->next; + } else { + node = (struct cso_node **)((const struct cso_node * const *)(&hash->data.e)); + } + return node; +} + +struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, + unsigned key, void *data) +{ + cso_data_might_grow(hash->data.d); + + struct cso_node **nextNode = cso_hash_find_node(hash, key); + struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); + struct cso_hash_iter iter = {hash, node}; + return iter; +} + +struct cso_hash * cso_hash_create(void) +{ + struct cso_hash *hash = malloc(sizeof(struct cso_hash)); + hash->data.d = malloc(sizeof(struct cso_hash_data)); + hash->data.d->fakeNext = 0; + hash->data.d->buckets = 0; + hash->data.d->size = 0; + hash->data.d->nodeSize = sizeof(struct cso_node); + hash->data.d->userNumBits = MinNumBits; + hash->data.d->numBits = 0; + hash->data.d->numBuckets = 0; + + return hash; +} + +void cso_hash_delete(struct cso_hash *hash) +{ + struct cso_node *e_for_x = (struct cso_node *)(hash->data.d); + struct cso_node **bucket = (struct cso_node **)(hash->data.d->buckets); + int n = hash->data.d->numBuckets; + while (n--) { + struct cso_node *cur = *bucket++; + while (cur != e_for_x) { + struct cso_node *next = cur->next; + cso_data_free_node(cur); + cur = next; + } + } + free(hash->data.d->buckets); + free(hash->data.d); + free(hash); +} + +struct cso_hash_iter cso_hash_find(struct cso_hash *hash, + unsigned key) +{ + struct cso_node **nextNode = cso_hash_find_node(hash, key); + struct cso_hash_iter iter = {hash, *nextNode}; + return iter; +} + +unsigned cso_hash_iter_key(struct cso_hash_iter iter) +{ + if (!iter.node || iter.hash->data.e == iter.node) + return 0; + return iter.node->key; +} + +void * cso_hash_iter_data(struct cso_hash_iter iter) +{ + if (!iter.node || iter.hash->data.e == iter.node) + return 0; + return iter.node->value; +} + +static struct cso_node *cso_hash_data_next(struct cso_node *node) +{ + union { + struct cso_node *next; + struct cso_node *e; + struct cso_hash_data *d; + } a; + a.next = node->next; + if (!a.next) { + fprintf(stderr, "iterating beyond the last element\n"); + return 0; + } + if (a.next->next) + return a.next; + + int start = (node->key % a.d->numBuckets) + 1; + struct cso_node **bucket = a.d->buckets + start; + int n = a.d->numBuckets - start; + while (n--) { + if (*bucket != a.e) + return *bucket; + ++bucket; + } + return a.e; +} + + +static struct cso_node *cso_hash_data_prev(struct cso_node *node) +{ + union { + struct cso_node *e; + struct cso_hash_data *d; + } a; + + a.e = node; + while (a.e->next) + a.e = a.e->next; + + int start; + if (node == a.e) + start = a.d->numBuckets - 1; + else + start = node->key % a.d->numBuckets; + + struct cso_node *sentinel = node; + struct cso_node **bucket = a.d->buckets + start; + while (start >= 0) { + if (*bucket != sentinel) { + struct cso_node *prev = *bucket; + while (prev->next != sentinel) + prev = prev->next; + return prev; + } + + sentinel = a.e; + --bucket; + --start; + } + fprintf(stderr, "iterating backward beyond first element\n"); + return a.e; +} + +struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter) +{ + struct cso_hash_iter next = {iter.hash, cso_hash_data_next(iter.node)}; + return next; +} + +int cso_hash_iter_is_null(struct cso_hash_iter iter) +{ + if (!iter.node || iter.node == iter.hash->data.e) + return 1; + return 0; +} + +void * cso_hash_take(struct cso_hash *hash, + unsigned akey) +{ + struct cso_node **node = cso_hash_find_node(hash, akey); + if (*node != hash->data.e) { + void *t = (*node)->value; + struct cso_node *next = (*node)->next; + cso_data_free_node(*node); + *node = next; + --hash->data.d->size; + cso_data_has_shrunk(hash->data.d); + return t; + } + return 0; +} + +struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter) +{ + struct cso_hash_iter prev = {iter.hash, + cso_hash_data_prev(iter.node)}; + return prev; +} + +struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash) +{ + struct cso_hash_iter iter = {hash, cso_data_first_node(hash->data.d)}; + return iter; +} diff --git a/src/gallium/aux/cso_cache/cso_hash.h b/src/gallium/aux/cso_cache/cso_hash.h new file mode 100644 index 0000000000..b4aa111860 --- /dev/null +++ b/src/gallium/aux/cso_cache/cso_hash.h @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#ifndef CSO_HASH_H +#define CSO_HASH_H + +struct cso_hash; +struct cso_node; + +struct cso_hash_iter { + struct cso_hash *hash; + struct cso_node *node; +}; + +struct cso_hash *cso_hash_create(void); +void cso_hash_delete(struct cso_hash *hash); + +struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, + void *data); +void *cso_hash_take(struct cso_hash *hash, unsigned key); + +struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash); +struct cso_hash_iter cso_hash_find(struct cso_hash *hash, unsigned key); + + +int cso_hash_iter_is_null(struct cso_hash_iter iter); +unsigned cso_hash_iter_key(struct cso_hash_iter iter); +void *cso_hash_iter_data(struct cso_hash_iter iter); + +struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter); +struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); + +#endif diff --git a/src/gallium/aux/draw/Makefile b/src/gallium/aux/draw/Makefile new file mode 100644 index 0000000000..451911a354 --- /dev/null +++ b/src/gallium/aux/draw/Makefile @@ -0,0 +1,2 @@ +default: + cd .. ; make diff --git a/src/gallium/aux/draw/draw_clip.c b/src/gallium/aux/draw/draw_clip.c new file mode 100644 index 0000000000..e3051507ea --- /dev/null +++ b/src/gallium/aux/draw/draw_clip.c @@ -0,0 +1,488 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Clipping stage + * + * \author Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + +#include "draw_context.h" +#include "draw_private.h" + + +#ifndef IS_NEGATIVE +#define IS_NEGATIVE(X) ((X) < 0.0) +#endif + +#ifndef DIFFERENT_SIGNS +#define DIFFERENT_SIGNS(x, y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F) +#endif + +#ifndef MAX_CLIPPED_VERTICES +#define MAX_CLIPPED_VERTICES ((2 * (6 + PIPE_MAX_CLIP_PLANES))+1) +#endif + + + +struct clipper { + struct draw_stage stage; /**< base class */ + + /* Basically duplicate some of the flatshading logic here: + */ + boolean flat; + uint num_color_attribs; + uint color_attribs[4]; /* front/back primary/secondary colors */ + + float (*plane)[4]; +}; + + +/* This is a bit confusing: + */ +static INLINE struct clipper *clipper_stage( struct draw_stage *stage ) +{ + return (struct clipper *)stage; +} + + +#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) + + +/* All attributes are float[4], so this is easy: + */ +static void interp_attr( float *fdst, + float t, + const float *fin, + const float *fout ) +{ + fdst[0] = LINTERP( t, fout[0], fin[0] ); + fdst[1] = LINTERP( t, fout[1], fin[1] ); + fdst[2] = LINTERP( t, fout[2], fin[2] ); + fdst[3] = LINTERP( t, fout[3], fin[3] ); +} + +static void copy_colors( struct draw_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + const struct clipper *clipper = clipper_stage(stage); + uint i; + for (i = 0; i < clipper->num_color_attribs; i++) { + const uint attr = clipper->color_attribs[i]; + COPY_4FV(dst->data[attr], src->data[attr]); + } +} + + + +/* Interpolate between two vertices to produce a third. + */ +static void interp( const struct clipper *clip, + struct vertex_header *dst, + float t, + const struct vertex_header *out, + const struct vertex_header *in ) +{ + const unsigned nr_attrs = clip->stage.draw->num_vs_outputs; + unsigned j; + + /* Vertex header. + */ + { + dst->clipmask = 0; + dst->edgeflag = 0; + dst->pad = 0; + dst->vertex_id = UNDEFINED_VERTEX_ID; + } + + /* Clip coordinates: interpolate normally + */ + { + interp_attr(dst->clip, t, in->clip, out->clip); + } + + /* Do the projective divide and insert window coordinates: + */ + { + const float *pos = dst->clip; + const float *scale = clip->stage.draw->viewport.scale; + const float *trans = clip->stage.draw->viewport.translate; + const float oow = 1.0f / pos[3]; + + dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; + dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; + dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; + dst->data[0][3] = oow; + } + + /* Other attributes + * Note: start at 1 to skip winpos (data[0]) since we just computed + * it above. + */ + for (j = 1; j < nr_attrs; j++) { + interp_attr(dst->data[j], t, in->data[j], out->data[j]); + } +} + + +static void emit_poly( struct draw_stage *stage, + struct vertex_header **inlist, + unsigned n, + const struct prim_header *origPrim) +{ + struct prim_header header; + unsigned i; + + /* later stages may need the determinant, but only the sign matters */ + header.det = origPrim->det; + + for (i = 2; i < n; i++) { + header.v[0] = inlist[i-1]; + header.v[1] = inlist[i]; + header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ + + { + unsigned tmp1 = header.v[1]->edgeflag; + unsigned tmp2 = header.v[2]->edgeflag; + + if (i != n-1) header.v[1]->edgeflag = 0; + if (i != 2) header.v[2]->edgeflag = 0; + + header.edgeflags = ((header.v[0]->edgeflag << 0) | + (header.v[1]->edgeflag << 1) | + (header.v[2]->edgeflag << 2)); + + stage->next->tri( stage->next, &header ); + + header.v[1]->edgeflag = tmp1; + header.v[2]->edgeflag = tmp2; + } + } +} + + + + +/* Clip a triangle against the viewport and user clip planes. + */ +static void +do_clip_tri( struct draw_stage *stage, + struct prim_header *header, + unsigned clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *a[MAX_CLIPPED_VERTICES]; + struct vertex_header *b[MAX_CLIPPED_VERTICES]; + struct vertex_header **inlist = a; + struct vertex_header **outlist = b; + unsigned tmpnr = 0; + unsigned n = 3; + unsigned i; + + inlist[0] = header->v[0]; + inlist[1] = header->v[1]; + inlist[2] = header->v[2]; + + while (clipmask && n >= 3) { + const unsigned plane_idx = ffs(clipmask)-1; + const float *plane = clipper->plane[plane_idx]; + struct vertex_header *vert_prev = inlist[0]; + float dp_prev = dot4( vert_prev->clip, plane ); + unsigned outcount = 0; + + clipmask &= ~(1<clip, plane ); + + if (!IS_NEGATIVE(dp_prev)) { + outlist[outcount++] = vert_prev; + } + + if (DIFFERENT_SIGNS(dp, dp_prev)) { + struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; + outlist[outcount++] = new_vert; + + if (IS_NEGATIVE(dp)) { + /* Going out of bounds. Avoid division by zero as we + * know dp != dp_prev from DIFFERENT_SIGNS, above. + */ + float t = dp / (dp - dp_prev); + interp( clipper, new_vert, t, vert, vert_prev ); + + /* Force edgeflag true in this case: + */ + new_vert->edgeflag = 1; + } else { + /* Coming back in. + */ + float t = dp_prev / (dp_prev - dp); + interp( clipper, new_vert, t, vert_prev, vert ); + + /* Copy starting vert's edgeflag: + */ + new_vert->edgeflag = vert_prev->edgeflag; + } + } + + vert_prev = vert; + dp_prev = dp; + } + + { + struct vertex_header **tmp = inlist; + inlist = outlist; + outlist = tmp; + n = outcount; + } + } + + /* If flat-shading, copy color to new provoking vertex. + */ + if (clipper->flat && inlist[0] != header->v[2]) { + if (1) { + inlist[0] = dup_vert(stage, inlist[0], tmpnr++); + } + + copy_colors(stage, inlist[0], header->v[2]); + } + + + + /* Emit the polygon as triangles to the setup stage: + */ + if (n >= 3) + emit_poly( stage, inlist, n, header ); +} + + +/* Clip a line against the viewport and user clip planes. + */ +static void +do_clip_line( struct draw_stage *stage, + struct prim_header *header, + unsigned clipmask ) +{ + const struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const float *pos0 = v0->clip; + const float *pos1 = v1->clip; + float t0 = 0.0F; + float t1 = 0.0F; + struct prim_header newprim; + + while (clipmask) { + const unsigned plane_idx = ffs(clipmask)-1; + const float *plane = clipper->plane[plane_idx]; + const float dp0 = dot4( pos0, plane ); + const float dp1 = dot4( pos1, plane ); + + if (dp1 < 0.0F) { + float t = dp1 / (dp1 - dp0); + t1 = MAX2(t1, t); + } + + if (dp0 < 0.0F) { + float t = dp0 / (dp0 - dp1); + t0 = MAX2(t0, t); + } + + if (t0 + t1 >= 1.0F) + return; /* discard */ + + clipmask &= ~(1 << plane_idx); /* turn off this plane's bit */ + } + + if (v0->clipmask) { + interp( clipper, stage->tmp[0], t0, v0, v1 ); + + if (clipper->flat) + copy_colors(stage, stage->tmp[0], v0); + + newprim.v[0] = stage->tmp[0]; + } + else { + newprim.v[0] = v0; + } + + if (v1->clipmask) { + interp( clipper, stage->tmp[1], t1, v1, v0 ); + newprim.v[1] = stage->tmp[1]; + } + else { + newprim.v[1] = v1; + } + + stage->next->line( stage->next, &newprim ); +} + + +static void +clip_point( struct draw_stage *stage, + struct prim_header *header ) +{ + if (header->v[0]->clipmask == 0) + stage->next->point( stage->next, header ); +} + + +static void +clip_line( struct draw_stage *stage, + struct prim_header *header ) +{ + unsigned clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask); + + if (clipmask == 0) { + /* no clipping needed */ + stage->next->line( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask) == 0) { + do_clip_line(stage, header, clipmask); + } + /* else, totally clipped */ +} + + +static void +clip_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + unsigned clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask | + header->v[2]->clipmask); + + if (clipmask == 0) { + /* no clipping needed */ + stage->next->tri( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask & + header->v[2]->clipmask) == 0) { + do_clip_tri(stage, header, clipmask); + } +} + +/* Update state. Could further delay this until we hit the first + * primitive that really requires clipping. + */ +static void +clip_init_state( struct draw_stage *stage ) +{ + struct clipper *clipper = clipper_stage( stage ); + + clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; + + if (clipper->flat) { + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + clipper->num_color_attribs = 0; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + clipper->color_attribs[clipper->num_color_attribs++] = i; + } + } + } + + stage->tri = clip_tri; + stage->line = clip_line; +} + + + +static void clip_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + clip_init_state( stage ); + stage->tri( stage, header ); +} + +static void clip_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + clip_init_state( stage ); + stage->line( stage, header ); +} + + +static void clip_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->tri = clip_first_tri; + stage->line = clip_first_line; + stage->next->flush( stage->next, flags ); +} + + +static void clip_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void clip_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Allocate a new clipper stage. + * \return pointer to new stage object + */ +struct draw_stage *draw_clip_stage( struct draw_context *draw ) +{ + struct clipper *clipper = CALLOC_STRUCT(clipper); + + draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); + + clipper->stage.draw = draw; + clipper->stage.point = clip_point; + clipper->stage.line = clip_first_line; + clipper->stage.tri = clip_first_tri; + clipper->stage.flush = clip_flush; + clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; + clipper->stage.destroy = clip_destroy; + + clipper->plane = draw->plane; + + return &clipper->stage; +} diff --git a/src/gallium/aux/draw/draw_context.c b/src/gallium/aux/draw/draw_context.c new file mode 100644 index 0000000000..4be3830316 --- /dev/null +++ b/src/gallium/aux/draw/draw_context.c @@ -0,0 +1,293 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "draw_context.h" +#include "draw_private.h" + + + +struct draw_context *draw_create( void ) +{ + struct draw_context *draw = CALLOC_STRUCT( draw_context ); + +#if defined(__i386__) || defined(__386__) + draw->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; +#else + draw->use_sse = FALSE; +#endif + + /* create pipeline stages */ + draw->pipeline.wide = draw_wide_stage( draw ); + draw->pipeline.stipple = draw_stipple_stage( draw ); + draw->pipeline.unfilled = draw_unfilled_stage( draw ); + draw->pipeline.twoside = draw_twoside_stage( draw ); + draw->pipeline.offset = draw_offset_stage( draw ); + draw->pipeline.clip = draw_clip_stage( draw ); + draw->pipeline.flatshade = draw_flatshade_stage( draw ); + draw->pipeline.cull = draw_cull_stage( draw ); + draw->pipeline.validate = draw_validate_stage( draw ); + draw->pipeline.first = draw->pipeline.validate; + + ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); + ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); + ASSIGN_4V( draw->plane[2], 0, -1, 0, 1 ); + ASSIGN_4V( draw->plane[3], 0, 1, 0, 1 ); + ASSIGN_4V( draw->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ + ASSIGN_4V( draw->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ + draw->nr_planes = 6; + + /* Statically allocate maximum sized vertices for the cache - could be cleverer... + */ + { + uint i; + const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; + char *tmp = align_malloc(Elements(draw->vcache.vertex) * size, 16); + + for (i = 0; i < Elements(draw->vcache.vertex); i++) + draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * size); + } + + draw->shader_queue_flush = draw_vertex_shader_queue_flush; + + draw->convert_wide_points = TRUE; + draw->convert_wide_lines = TRUE; + + draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ + + draw_vertex_cache_invalidate( draw ); + draw_set_mapped_element_buffer( draw, 0, NULL ); + + return draw; +} + + +void draw_destroy( struct draw_context *draw ) +{ + draw->pipeline.wide->destroy( draw->pipeline.wide ); + draw->pipeline.stipple->destroy( draw->pipeline.stipple ); + draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); + draw->pipeline.twoside->destroy( draw->pipeline.twoside ); + draw->pipeline.offset->destroy( draw->pipeline.offset ); + draw->pipeline.clip->destroy( draw->pipeline.clip ); + draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); + draw->pipeline.cull->destroy( draw->pipeline.cull ); + draw->pipeline.validate->destroy( draw->pipeline.validate ); + if (draw->pipeline.rasterize) + draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); + tgsi_exec_machine_free_data(&draw->machine); + align_free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ + FREE( draw ); +} + + + +void draw_flush( struct draw_context *draw ) +{ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); +} + + + +/** + * Register new primitive rasterization/rendering state. + * This causes the drawing pipeline to be rebuilt. + */ +void draw_set_rasterizer_state( struct draw_context *draw, + const struct pipe_rasterizer_state *raster ) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + draw->rasterizer = raster; +} + + +/** + * Plug in the primitive rendering/rasterization stage (which is the last + * stage in the drawing pipeline). + * This is provided by the device driver. + */ +void draw_set_rasterize_stage( struct draw_context *draw, + struct draw_stage *stage ) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + draw->pipeline.rasterize = stage; +} + + +/** + * Set the draw module's clipping state. + */ +void draw_set_clip_state( struct draw_context *draw, + const struct pipe_clip_state *clip ) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + assert(clip->nr <= PIPE_MAX_CLIP_PLANES); + memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); + draw->nr_planes = 6 + clip->nr; +} + + +/** + * Set the draw module's viewport state. + */ +void draw_set_viewport_state( struct draw_context *draw, + const struct pipe_viewport_state *viewport ) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->viewport = *viewport; /* struct copy */ +} + + + +void +draw_set_vertex_buffer(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_buffer *buffer) +{ + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); + assert(attr < PIPE_ATTRIB_MAX); + draw->vertex_buffer[attr] = *buffer; +} + + +void +draw_set_vertex_element(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_element *element) +{ + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); + assert(attr < PIPE_ATTRIB_MAX); + draw->vertex_element[attr] = *element; +} + + +/** + * Tell drawing context where to find mapped vertex buffers. + */ +void +draw_set_mapped_vertex_buffer(struct draw_context *draw, + unsigned attr, const void *buffer) +{ + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); + draw->user.vbuffer[attr] = buffer; +} + + +void +draw_set_mapped_constant_buffer(struct draw_context *draw, + const void *buffer) +{ + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); + draw->user.constants = buffer; +} + + +/** + * Tells the draw module whether to convert wide points (size != 1) + * into triangles. + */ +void +draw_convert_wide_points(struct draw_context *draw, boolean enable) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->convert_wide_points = enable; +} + + +/** + * Tells the draw module whether to convert wide lines (width != 1) + * into triangles. + */ +void +draw_convert_wide_lines(struct draw_context *draw, boolean enable) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->convert_wide_lines = enable; +} + + +/** + * Allocate space for temporary post-transform vertices, such as for clipping. + */ +void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) +{ + assert(!stage->tmp); + + stage->nr_tmps = nr; + + if (nr) { + ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); + unsigned i; + + stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } +} + + +void draw_free_temp_verts( struct draw_stage *stage ) +{ + if (stage->tmp) { + FREE( stage->tmp[0] ); + FREE( stage->tmp ); + stage->tmp = NULL; + } +} + + +boolean draw_use_sse(struct draw_context *draw) +{ + return (boolean) draw->use_sse; +} + + +void draw_reset_vertex_ids(struct draw_context *draw) +{ + struct draw_stage *stage = draw->pipeline.first; + + while (stage) { + unsigned i; + + for (i = 0; i < stage->nr_tmps; i++) + stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; + + stage = stage->next; + } + + draw_vertex_cache_reset_vertex_ids(draw); +} diff --git a/src/gallium/aux/draw/draw_context.h b/src/gallium/aux/draw/draw_context.h new file mode 100644 index 0000000000..ddeb184497 --- /dev/null +++ b/src/gallium/aux/draw/draw_context.h @@ -0,0 +1,142 @@ + +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Public interface into the drawing module. + */ + +/* Authors: Keith Whitwell + */ + + +#ifndef DRAW_CONTEXT_H +#define DRAW_CONTEXT_H + + +#include "pipe/p_state.h" + + +struct vertex_buffer; +struct vertex_info; +struct draw_context; +struct draw_stage; +struct draw_vertex_shader; + + +/** + * Clipmask flags + */ +/*@{*/ +#define CLIP_RIGHT_BIT 0x01 +#define CLIP_LEFT_BIT 0x02 +#define CLIP_TOP_BIT 0x04 +#define CLIP_BOTTOM_BIT 0x08 +#define CLIP_NEAR_BIT 0x10 +#define CLIP_FAR_BIT 0x20 +/*@}*/ + +/** + * Bitshift for each clip flag + */ +/*@{*/ +#define CLIP_RIGHT_SHIFT 0 +#define CLIP_LEFT_SHIFT 1 +#define CLIP_TOP_SHIFT 2 +#define CLIP_BOTTOM_SHIFT 3 +#define CLIP_NEAR_SHIFT 4 +#define CLIP_FAR_SHIFT 5 +/*@}*/ + + +struct draw_context *draw_create( void ); + +void draw_destroy( struct draw_context *draw ); + +void draw_set_viewport_state( struct draw_context *draw, + const struct pipe_viewport_state *viewport ); + +void draw_set_clip_state( struct draw_context *pipe, + const struct pipe_clip_state *clip ); + +void draw_set_rasterizer_state( struct draw_context *draw, + const struct pipe_rasterizer_state *raster ); + +void draw_set_rasterize_stage( struct draw_context *draw, + struct draw_stage *stage ); + +void draw_convert_wide_points(struct draw_context *draw, boolean enable); + +void draw_convert_wide_lines(struct draw_context *draw, boolean enable); + + +struct draw_vertex_shader * +draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader); +void draw_bind_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs); +void draw_delete_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs); + +boolean draw_use_sse(struct draw_context *draw); + +void draw_set_vertex_buffer(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_buffer *buffer); + +void draw_set_vertex_element(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_element *element); + +void draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ); + +void draw_set_mapped_vertex_buffer(struct draw_context *draw, + unsigned attr, const void *buffer); + +void draw_set_mapped_constant_buffer(struct draw_context *draw, + const void *buffer); + + +/*********************************************************************** + * draw_prim.c + */ + +void draw_arrays(struct draw_context *draw, unsigned prim, + unsigned start, unsigned count); + +void draw_flush(struct draw_context *draw); + +/*********************************************************************** + * draw_debug.c + */ +boolean draw_validate_prim( unsigned prim, unsigned length ); +unsigned draw_trim_prim( unsigned mode, unsigned count ); + + + +#endif /* DRAW_CONTEXT_H */ diff --git a/src/gallium/aux/draw/draw_cull.c b/src/gallium/aux/draw/draw_cull.c new file mode 100644 index 0000000000..8177b0ac86 --- /dev/null +++ b/src/gallium/aux/draw/draw_cull.c @@ -0,0 +1,150 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Drawing stage for polygon culling + */ + +/* Authors: Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + +struct cull_stage { + struct draw_stage stage; + unsigned winding; /**< which winding(s) to cull (one of PIPE_WINDING_x) */ +}; + + +static INLINE struct cull_stage *cull_stage( struct draw_stage *stage ) +{ + return (struct cull_stage *)stage; +} + + + + +static void cull_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + /* Window coords: */ + const float *v0 = header->v[0]->data[0]; + const float *v1 = header->v[1]->data[0]; + const float *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0] - v2[0]; + const float ey = v0[1] - v2[1]; + const float fx = v1[0] - v2[0]; + const float fy = v1[1] - v2[1]; + + /* det = cross(e,f).z */ + header->det = ex * fy - ey * fx; + + if (header->det != 0) { + /* if (det < 0 then Z points toward camera and triangle is + * counter-clockwise winding. + */ + unsigned winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; + + if ((winding & cull_stage(stage)->winding) == 0) { + /* triangle is not culled, pass to next stage */ + stage->next->tri( stage->next, header ); + } + } +} + +static void cull_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct cull_stage *cull = cull_stage(stage); + + cull->winding = stage->draw->rasterizer->cull_mode; + + stage->tri = cull_tri; + stage->tri( stage, header ); +} + + + +static void cull_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void cull_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void cull_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->tri = cull_first_tri; + stage->next->flush( stage->next, flags ); +} + +static void cull_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void cull_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create a new polygon culling stage. + */ +struct draw_stage *draw_cull_stage( struct draw_context *draw ) +{ + struct cull_stage *cull = CALLOC_STRUCT(cull_stage); + + draw_alloc_temp_verts( &cull->stage, 0 ); + + cull->stage.draw = draw; + cull->stage.next = NULL; + cull->stage.point = cull_point; + cull->stage.line = cull_line; + cull->stage.tri = cull_first_tri; + cull->stage.flush = cull_flush; + cull->stage.reset_stipple_counter = cull_reset_stipple_counter; + cull->stage.destroy = cull_destroy; + + return &cull->stage; +} diff --git a/src/gallium/aux/draw/draw_debug.c b/src/gallium/aux/draw/draw_debug.c new file mode 100644 index 0000000000..d6220b5f62 --- /dev/null +++ b/src/gallium/aux/draw/draw_debug.c @@ -0,0 +1,113 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "draw_private.h" +#include "draw_context.h" + + + +static void +draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) +{ + assert(prim >= PIPE_PRIM_POINTS); + assert(prim <= PIPE_PRIM_POLYGON); + + switch (prim) { + case PIPE_PRIM_POINTS: + *first = 1; + *incr = 1; + break; + case PIPE_PRIM_LINES: + *first = 2; + *incr = 2; + break; + case PIPE_PRIM_LINE_STRIP: + *first = 2; + *incr = 1; + break; + case PIPE_PRIM_LINE_LOOP: + *first = 2; + *incr = 1; + break; + case PIPE_PRIM_TRIANGLES: + *first = 3; + *incr = 3; + break; + case PIPE_PRIM_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + break; + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + *first = 3; + *incr = 1; + break; + case PIPE_PRIM_QUADS: + *first = 4; + *incr = 4; + break; + case PIPE_PRIM_QUAD_STRIP: + *first = 4; + *incr = 2; + break; + default: + assert(0); + *first = 1; + *incr = 1; + break; + } +} + + +unsigned +draw_trim_prim( unsigned mode, unsigned count ) +{ + unsigned length, first, incr; + + draw_prim_info( mode, &first, &incr ); + + if (count < first) + length = 0; + else + length = count - (count - first) % incr; + + return length; +} + + +boolean +draw_validate_prim( unsigned mode, unsigned count ) +{ + return (count > 0 && + count == draw_trim_prim( mode, count )); +} + diff --git a/src/gallium/aux/draw/draw_flatshade.c b/src/gallium/aux/draw/draw_flatshade.c new file mode 100644 index 0000000000..4398abbc60 --- /dev/null +++ b/src/gallium/aux/draw/draw_flatshade.c @@ -0,0 +1,205 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +/** subclass of draw_stage */ +struct flat_stage +{ + struct draw_stage stage; + + uint num_color_attribs; + uint color_attribs[4]; /* front/back primary/secondary colors */ +}; + + +static INLINE struct flat_stage * +flat_stage(struct draw_stage *stage) +{ + return (struct flat_stage *) stage; +} + + +/** Copy all the color attributes from 'src' vertex to 'dst' vertex */ +static INLINE void copy_colors( struct draw_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + const struct flat_stage *flat = flat_stage(stage); + uint i; + for (i = 0; i < flat->num_color_attribs; i++) { + const uint attr = flat->color_attribs[i]; + COPY_4FV(dst->data[attr], src->data[attr]); + } +} + + +/** Copy all the color attributes from src vertex to dst0 & dst1 vertices */ +static INLINE void copy_colors2( struct draw_stage *stage, + struct vertex_header *dst0, + struct vertex_header *dst1, + const struct vertex_header *src ) +{ + const struct flat_stage *flat = flat_stage(stage); + uint i; + for (i = 0; i < flat->num_color_attribs; i++) { + const uint attr = flat->color_attribs[i]; + COPY_4FV(dst0->data[attr], src->data[attr]); + COPY_4FV(dst1->data[attr], src->data[attr]); + } +} + + +/** + * Flatshade tri. Required for clipping and when unfilled tris are + * active, otherwise handled by hardware. + */ +static void flatshade_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = header->v[2]; + + copy_colors2(stage, tmp.v[0], tmp.v[1], tmp.v[2]); + + stage->next->tri( stage->next, &tmp ); +} + + +/** + * Flatshade line. Required for clipping. + */ +static void flatshade_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = header->v[1]; + + copy_colors(stage, tmp.v[0], tmp.v[1]); + + stage->next->line( stage->next, &tmp ); +} + + +static void flatshade_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void flatshade_init_state( struct draw_stage *stage ) +{ + struct flat_stage *flat = flat_stage(stage); + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + /* Find which vertex shader outputs are colors, make a list */ + flat->num_color_attribs = 0; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + flat->color_attribs[flat->num_color_attribs++] = i; + } + } + + stage->line = flatshade_line; + stage->tri = flatshade_tri; +} + +static void flatshade_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + flatshade_init_state( stage ); + stage->tri( stage, header ); +} + +static void flatshade_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + flatshade_init_state( stage ); + stage->line( stage, header ); +} + + +static void flatshade_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->tri = flatshade_first_tri; + stage->line = flatshade_first_line; + stage->next->flush( stage->next, flags ); +} + + +static void flatshade_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void flatshade_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create flatshading drawing stage. + */ +struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) +{ + struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage); + + draw_alloc_temp_verts( &flatshade->stage, 2 ); + + flatshade->stage.draw = draw; + flatshade->stage.next = NULL; + flatshade->stage.point = flatshade_point; + flatshade->stage.line = flatshade_first_line; + flatshade->stage.tri = flatshade_first_tri; + flatshade->stage.flush = flatshade_flush; + flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; + flatshade->stage.destroy = flatshade_destroy; + + return &flatshade->stage; +} + + diff --git a/src/gallium/aux/draw/draw_offset.c b/src/gallium/aux/draw/draw_offset.c new file mode 100644 index 0000000000..dbc676deae --- /dev/null +++ b/src/gallium/aux/draw/draw_offset.c @@ -0,0 +1,186 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief polygon offset state + * + * \author Keith Whitwell + * \author Brian Paul + */ + +#include "pipe/p_util.h" +#include "draw_private.h" + + + +struct offset_stage { + struct draw_stage stage; + + float scale; + float units; +}; + + + +static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) +{ + return (struct offset_stage *) stage; +} + + + + + +/** + * Offset tri Z. Some hardware can handle this, but not usually when + * doing unfilled rendering. + */ +static void do_offset_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + float inv_det = 1.0f / header->det; + + /* Window coords: + */ + float *v0 = header->v[0]->data[0]; + float *v1 = header->v[1]->data[0]; + float *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ + float ex = v0[0] - v2[0]; + float ey = v0[1] - v2[1]; + float ez = v0[2] - v2[2]; + float fx = v1[0] - v2[0]; + float fy = v1[1] - v2[1]; + float fz = v1[2] - v2[2]; + + /* (a,b) = cross(e,f).xy */ + float a = ey*fz - ez*fy; + float b = ez*fx - ex*fz; + + float dzdx = FABSF(a * inv_det); + float dzdy = FABSF(b * inv_det); + + float zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; + + /* + * Note: we're applying the offset and clamping per-vertex. + * Ideally, the offset is applied per-fragment prior to fragment shading. + */ + v0[2] = CLAMP(v0[2] + zoffset, 0.0f, 1.0f); + v1[2] = CLAMP(v1[2] + zoffset, 0.0f, 1.0f); + v2[2] = CLAMP(v2[2] + zoffset, 0.0f, 1.0f); + + stage->next->tri( stage->next, header ); +} + + +static void offset_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = dup_vert(stage, header->v[2], 2); + + do_offset_tri( stage, &tmp ); +} + + +static void offset_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ + + offset->units = stage->draw->rasterizer->offset_units * mrd; + offset->scale = stage->draw->rasterizer->offset_scale; + + stage->tri = offset_tri; + stage->tri( stage, header ); +} + + +static void offset_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void offset_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void offset_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->tri = offset_first_tri; + stage->next->flush( stage->next, flags ); +} + + +static void offset_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void offset_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create polygon offset drawing stage. + */ +struct draw_stage *draw_offset_stage( struct draw_context *draw ) +{ + struct offset_stage *offset = CALLOC_STRUCT(offset_stage); + + draw_alloc_temp_verts( &offset->stage, 3 ); + + offset->stage.draw = draw; + offset->stage.next = NULL; + offset->stage.point = offset_point; + offset->stage.line = offset_line; + offset->stage.tri = offset_first_tri; + offset->stage.flush = offset_flush; + offset->stage.reset_stipple_counter = offset_reset_stipple_counter; + offset->stage.destroy = offset_destroy; + + return &offset->stage; +} diff --git a/src/gallium/aux/draw/draw_prim.c b/src/gallium/aux/draw/draw_prim.c new file mode 100644 index 0000000000..51e2242719 --- /dev/null +++ b/src/gallium/aux/draw/draw_prim.c @@ -0,0 +1,482 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_debug.h" + +#include "draw_private.h" +#include "draw_context.h" + + + +#define RP_NONE 0 +#define RP_POINT 1 +#define RP_LINE 2 +#define RP_TRI 3 + + +static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { + RP_POINT, + RP_LINE, + RP_LINE, + RP_LINE, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI +}; + + +static void draw_prim_queue_flush( struct draw_context *draw ) +{ + unsigned i; + + if (0) + debug_printf("Flushing with %d prims, %d verts\n", + draw->pq.queue_nr, draw->vs.queue_nr); + + assert (draw->pq.queue_nr != 0); + + /* NOTE: we cannot save draw->pipeline->first in a local var because + * draw->pipeline->first is often changed by the first call to tri(), + * line(), etc. + */ + if (draw->rasterizer->line_stipple_enable) { + switch (draw->reduced_prim) { + case RP_TRI: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); + + draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); + } + break; + case RP_LINE: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); + + draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); + } + break; + case RP_POINT: + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); + break; + } + } + else { + switch (draw->reduced_prim) { + case RP_TRI: + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); + break; + case RP_LINE: + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); + break; + case RP_POINT: + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); + break; + } + } + + draw->pq.queue_nr = 0; + draw_vertex_cache_unreference( draw ); +} + + + +void draw_do_flush( struct draw_context *draw, unsigned flags ) +{ + if (0) + debug_printf("Flushing with %d verts, %d prims\n", + draw->vs.queue_nr, + draw->pq.queue_nr ); + + + if (flags >= DRAW_FLUSH_SHADER_QUEUE) { + if (draw->vs.queue_nr) + (*draw->shader_queue_flush)(draw); + + if (flags >= DRAW_FLUSH_PRIM_QUEUE) { + if (draw->pq.queue_nr) + draw_prim_queue_flush(draw); + + if (flags >= DRAW_FLUSH_VERTEX_CACHE) { + draw_vertex_cache_invalidate(draw); + + if (flags >= DRAW_FLUSH_STATE_CHANGE) { + draw->pipeline.first->flush( draw->pipeline.first, flags ); + draw->pipeline.first = draw->pipeline.validate; + draw->reduced_prim = ~0; + } + } + } + } +} + + + +/* Return a pointer to a freshly queued primitive header. Ensure that + * there is room in the vertex cache for a maximum of "nr_verts" new + * vertices. Flush primitive and/or vertex queues if necessary to + * make space. + */ +static struct prim_header *get_queued_prim( struct draw_context *draw, + unsigned nr_verts ) +{ + if (!draw_vertex_cache_check_space( draw, nr_verts )) { +// debug_printf("v"); + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE ); + } + else if (draw->pq.queue_nr == PRIM_QUEUE_LENGTH) { +// debug_printf("p"); + draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); + } + + assert(draw->pq.queue_nr < PRIM_QUEUE_LENGTH); + + return &draw->pq.queue[draw->pq.queue_nr++]; +} + + + +/** + * Add a point to the primitive queue. + * \param i0 index into user's vertex arrays + */ +static void do_point( struct draw_context *draw, + unsigned i0 ) +{ + struct prim_header *prim = get_queued_prim( draw, 1 ); + + prim->reset_line_stipple = 0; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->vcache.get_vertex( draw, i0 ); +} + + +/** + * Add a line to the primitive queue. + * \param i0 index into user's vertex arrays + * \param i1 index into user's vertex arrays + */ +static void do_line( struct draw_context *draw, + boolean reset_stipple, + unsigned i0, + unsigned i1 ) +{ + struct prim_header *prim = get_queued_prim( draw, 2 ); + + prim->reset_line_stipple = reset_stipple; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->vcache.get_vertex( draw, i0 ); + prim->v[1] = draw->vcache.get_vertex( draw, i1 ); +} + +/** + * Add a triangle to the primitive queue. + */ +static void do_triangle( struct draw_context *draw, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + + prim->reset_line_stipple = 1; + prim->edgeflags = ~0; + prim->pad = 0; + prim->v[0] = draw->vcache.get_vertex( draw, i0 ); + prim->v[1] = draw->vcache.get_vertex( draw, i1 ); + prim->v[2] = draw->vcache.get_vertex( draw, i2 ); +} + +static void do_ef_triangle( struct draw_context *draw, + boolean reset_stipple, + unsigned ef_mask, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + struct vertex_header *v0 = draw->vcache.get_vertex( draw, i0 ); + struct vertex_header *v1 = draw->vcache.get_vertex( draw, i1 ); + struct vertex_header *v2 = draw->vcache.get_vertex( draw, i2 ); + + prim->reset_line_stipple = reset_stipple; + + prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | + (v1->edgeflag << 1) | + (v2->edgeflag << 2)); + prim->pad = 0; + prim->v[0] = v0; + prim->v[1] = v1; + prim->v[2] = v2; +} + + +static void do_ef_quad( struct draw_context *draw, + unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3 ) +{ + const unsigned omitEdge2 = ~(1 << 1); + const unsigned omitEdge3 = ~(1 << 2); + do_ef_triangle( draw, 1, omitEdge2, v0, v1, v3 ); + do_ef_triangle( draw, 0, omitEdge3, v1, v2, v3 ); +} + +static void do_quad( struct draw_context *draw, + unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3 ) +{ + do_triangle( draw, v0, v1, v3 ); + do_triangle( draw, v1, v2, v3 ); +} + + +/** + * Main entrypoint to draw some number of points/lines/triangles + */ +static void +draw_prim( struct draw_context *draw, + unsigned prim, unsigned start, unsigned count ) +{ + unsigned i; + boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); + +// debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); + + switch (prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i ++) { + do_point( draw, + start + i ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) { + do_line( draw, + TRUE, + start + i + 0, + start + i + 1); + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, /* XXX: only if vb not split */ + start + i - 1, + start + i ); + } + + do_line( draw, + 0, + start + count - 1, + start + 0 ); + } + break; + + case PIPE_PRIM_LINE_STRIP: + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, + start + i - 1, + start + i ); + } + break; + + case PIPE_PRIM_TRIANGLES: + if (unfilled) { + for (i = 0; i+2 < count; i += 3) { + do_ef_triangle( draw, + 1, + ~0, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + else { + for (i = 0; i+2 < count; i += 3) { + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + do_triangle( draw, + start + i + 1, + start + i + 0, + start + i + 2 ); + } + else { + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + if (count >= 3) { + for (i = 0; i+2 < count; i++) { + do_triangle( draw, + start + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + + case PIPE_PRIM_QUADS: + if (unfilled) { + for (i = 0; i+3 < count; i += 4) { + do_ef_quad( draw, + start + i + 0, + start + i + 1, + start + i + 2, + start + i + 3); + } + } + else { + for (i = 0; i+3 < count; i += 4) { + do_quad( draw, + start + i + 0, + start + i + 1, + start + i + 2, + start + i + 3); + } + } + break; + + case PIPE_PRIM_QUAD_STRIP: + if (unfilled) { + for (i = 0; i+3 < count; i += 2) { + do_ef_quad( draw, + start + i + 2, + start + i + 0, + start + i + 1, + start + i + 3); + } + } + else { + for (i = 0; i+3 < count; i += 2) { + do_quad( draw, + start + i + 2, + start + i + 0, + start + i + 1, + start + i + 3); + } + } + break; + + case PIPE_PRIM_POLYGON: + if (unfilled) { + unsigned ef_mask = (1<<2) | (1<<0); + + for (i = 0; i+2 < count; i++) { + + if (i + 3 >= count) + ef_mask |= (1<<1); + + do_ef_triangle( draw, + i == 0, + ef_mask, + start + i + 1, + start + i + 2, + start + 0); + + ef_mask &= ~(1<<2); + } + } + else { + for (i = 0; i+2 < count; i++) { + do_triangle( draw, + start + i + 1, + start + i + 2, + start + 0); + } + } + break; + + default: + assert(0); + break; + } +} + + + + +/** + * Draw vertex arrays + * This is the main entrypoint into the drawing module. + * \param prim one of PIPE_PRIM_x + * \param start index of first vertex to draw + * \param count number of vertices to draw + */ +void +draw_arrays(struct draw_context *draw, unsigned prim, + unsigned start, unsigned count) +{ + if (reduced_prim[prim] != draw->reduced_prim) { + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->reduced_prim = reduced_prim[prim]; + } + + /* drawing done here: */ + draw_prim(draw, prim, start, count); +} + + diff --git a/src/gallium/aux/draw/draw_private.h b/src/gallium/aux/draw/draw_private.h new file mode 100644 index 0000000000..b17eaaed65 --- /dev/null +++ b/src/gallium/aux/draw/draw_private.h @@ -0,0 +1,346 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Private data structures, etc for the draw module. + */ + + +/** + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#ifndef DRAW_PRIVATE_H +#define DRAW_PRIVATE_H + + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" + +#include "x86/rtasm/x86sse.h" +#include "pipe/tgsi/exec/tgsi_exec.h" + + +struct gallivm_prog; +struct gallivm_cpu_engine; + +/** + * Basic vertex info. + * Carry some useful information around with the vertices in the prim pipe. + */ +struct vertex_header { + unsigned clipmask:12; + unsigned edgeflag:1; + unsigned pad:3; + unsigned vertex_id:16; + + float clip[4]; + + float data[][4]; /* Note variable size */ +}; + +/* NOTE: It should match vertex_id size above */ +#define UNDEFINED_VERTEX_ID 0xffff + +/* XXX This is too large */ +#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) + + + +/** + * Basic info for a point/line/triangle primitive. + */ +struct prim_header { + float det; /**< front/back face determinant */ + unsigned reset_line_stipple:1; + unsigned edgeflags:3; + unsigned pad:28; + struct vertex_header *v[3]; /**< 1 to 3 vertex pointers */ +}; + + + +struct draw_context; + +/** + * Base class for all primitive drawing stages. + */ +struct draw_stage +{ + struct draw_context *draw; /**< parent context */ + + struct draw_stage *next; /**< next stage in pipeline */ + + struct vertex_header **tmp; /**< temp vert storage, such as for clipping */ + unsigned nr_tmps; + + void (*point)( struct draw_stage *, + struct prim_header * ); + + void (*line)( struct draw_stage *, + struct prim_header * ); + + void (*tri)( struct draw_stage *, + struct prim_header * ); + + void (*flush)( struct draw_stage *, + unsigned flags ); + + void (*reset_stipple_counter)( struct draw_stage * ); + + void (*destroy)( struct draw_stage * ); +}; + + +#define PRIM_QUEUE_LENGTH 16 +#define VCACHE_SIZE 32 +#define VCACHE_OVERFLOW 4 +#define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ + +/** + * Private version of the compiled vertex_shader + */ +struct draw_vertex_shader { + const struct pipe_shader_state *state; +#if defined(__i386__) || defined(__386__) + struct x86_function sse2_program; +#endif +#ifdef MESA_LLVM + struct gallivm_prog *llvm_prog; +#endif +}; + + +/* Internal function for vertex fetch. + */ +typedef void (*fetch_func)(const void *ptr, float *attrib); +typedef void (*full_fetch_func)( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ); + + + +/** + * Private context for the drawing module. + */ +struct draw_context +{ + /** Drawing/primitive pipeline stages */ + struct { + struct draw_stage *first; /**< one of the following */ + + struct draw_stage *validate; + + /* stages (in logical order) */ + struct draw_stage *flatshade; + struct draw_stage *clip; + struct draw_stage *cull; + struct draw_stage *twoside; + struct draw_stage *offset; + struct draw_stage *unfilled; + struct draw_stage *stipple; + struct draw_stage *wide; + struct draw_stage *rasterize; + } pipeline; + + /* pipe state that we need: */ + const struct pipe_rasterizer_state *rasterizer; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + const struct draw_vertex_shader *vertex_shader; + + uint num_vs_outputs; /**< convenience, from vertex_shader */ + + /* user-space vertex data, buffers */ + struct { + /** vertex element/index buffer (ex: glDrawElements) */ + const void *elts; + /** bytes per index (0, 1, 2 or 4) */ + unsigned eltSize; + + /** vertex arrays */ + const void *vbuffer[PIPE_ATTRIB_MAX]; + + /** constant buffer (for vertex shader) */ + const void *constants; + } user; + + /* Clip derived state: + */ + float plane[12][4]; + unsigned nr_planes; + + boolean convert_wide_points; /**< convert wide points to tris? */ + boolean convert_wide_lines; /**< convert side lines to tris? */ + + unsigned reduced_prim; + + /** TGSI program interpreter runtime state */ + struct tgsi_exec_machine machine; + + /* Vertex fetch internal state + */ + struct { + const ubyte *src_ptr[PIPE_ATTRIB_MAX]; + unsigned pitch[PIPE_ATTRIB_MAX]; + fetch_func fetch[PIPE_ATTRIB_MAX]; + unsigned nr_attrs; + full_fetch_func fetch_func; + } vertex_fetch; + + /* Post-tnl vertex cache: + */ + struct { + unsigned referenced; /**< bitfield */ + unsigned idx[VCACHE_SIZE + VCACHE_OVERFLOW]; + struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; + unsigned overflow; + + /** To find space in the vertex cache: */ + struct vertex_header *(*get_vertex)( struct draw_context *draw, + unsigned i ); + } vcache; + + /* Vertex shader queue: + */ + struct { + struct { + unsigned elt; /**< index into the user's vertex arrays */ + struct vertex_header *dest; /**< points into vcache.vertex[] array */ + } queue[VS_QUEUE_LENGTH]; + unsigned queue_nr; + } vs; + + /** + * Run the vertex shader on all vertices in the vertex queue. + */ + void (*shader_queue_flush)(struct draw_context *draw); + + /* Prim pipeline queue: + */ + struct { + /* Need to queue up primitives until their vertices have been + * transformed by a vs queue flush. + */ + struct prim_header queue[PRIM_QUEUE_LENGTH]; + unsigned queue_nr; + } pq; + + int use_sse : 1; +#ifdef MESA_LLVM + struct gallivm_cpu_engine *engine; +#endif + + void *driver_private; +}; + + + +extern struct draw_stage *draw_unfilled_stage( struct draw_context *context ); +extern struct draw_stage *draw_twoside_stage( struct draw_context *context ); +extern struct draw_stage *draw_offset_stage( struct draw_context *context ); +extern struct draw_stage *draw_clip_stage( struct draw_context *context ); +extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); +extern struct draw_stage *draw_cull_stage( struct draw_context *context ); +extern struct draw_stage *draw_stipple_stage( struct draw_context *context ); +extern struct draw_stage *draw_wide_stage( struct draw_context *context ); +extern struct draw_stage *draw_validate_stage( struct draw_context *context ); + + +extern void draw_free_temp_verts( struct draw_stage *stage ); + +extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); + +extern void draw_reset_vertex_ids( struct draw_context *draw ); + + +extern int draw_vertex_cache_check_space( struct draw_context *draw, + unsigned nr_verts ); + +extern void draw_vertex_cache_invalidate( struct draw_context *draw ); +extern void draw_vertex_cache_unreference( struct draw_context *draw ); +extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); + + +extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); +#ifdef MESA_LLVM +extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw ); +#endif + +struct tgsi_exec_machine; + +extern void draw_update_vertex_fetch( struct draw_context *draw ); + + +#define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ +#define DRAW_FLUSH_PRIM_QUEUE 0x2 +#define DRAW_FLUSH_VERTEX_CACHE 0x4 +#define DRAW_FLUSH_STATE_CHANGE 0x8 +#define DRAW_FLUSH_BACKEND 0x10 + + +void draw_do_flush( struct draw_context *draw, unsigned flags ); + + + +/** + * Get a writeable copy of a vertex. + * \param stage drawing stage info + * \param vert the vertex to copy (source) + * \param idx index into stage's tmp[] array to put the copy (dest) + * \return pointer to the copied vertex + */ +static INLINE struct vertex_header * +dup_vert( struct draw_stage *stage, + const struct vertex_header *vert, + unsigned idx ) +{ + struct vertex_header *tmp = stage->tmp[idx]; + const uint vsize = sizeof(struct vertex_header) + + stage->draw->num_vs_outputs * 4 * sizeof(float); + memcpy(tmp, vert, vsize); + tmp->vertex_id = UNDEFINED_VERTEX_ID; + return tmp; +} + +static INLINE float +dot4(const float *a, const float *b) +{ + float result = (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); + + return result; +} + +#endif /* DRAW_PRIVATE_H */ diff --git a/src/gallium/aux/draw/draw_stipple.c b/src/gallium/aux/draw/draw_stipple.c new file mode 100644 index 0000000000..506f33512c --- /dev/null +++ b/src/gallium/aux/draw/draw_stipple.c @@ -0,0 +1,239 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +/* Implement line stipple by cutting lines up into smaller lines. + * There are hundreds of ways to implement line stipple, this is one + * choice that should work in all situations, requires no state + * manipulations, but with a penalty in terms of large amounts of + * generated geometry. + */ + + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +/** Subclass of draw_stage */ +struct stipple_stage { + struct draw_stage stage; + float counter; + uint pattern; + uint factor; +}; + + +static INLINE struct stipple_stage * +stipple_stage(struct draw_stage *stage) +{ + return (struct stipple_stage *) stage; +} + + +/** + * Compute interpolated vertex attributes for 'dst' at position 't' + * between 'v0' and 'v1'. + * XXX using linear interpolation for all attribs at this time. + */ +static void +screen_interp( struct draw_context *draw, + struct vertex_header *dst, + float t, + const struct vertex_header *v0, + const struct vertex_header *v1 ) +{ + uint attr; + for (attr = 0; attr < draw->num_vs_outputs; attr++) { + const float *val0 = v0->data[attr]; + const float *val1 = v1->data[attr]; + float *newv = dst->data[attr]; + uint i; + for (i = 0; i < 4; i++) { + newv[i] = val0[i] + t * (val1[i] - val0[i]); + } + } +} + + +static void +emit_segment(struct draw_stage *stage, struct prim_header *header, + float t0, float t1) +{ + struct vertex_header *v0new = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1new = dup_vert(stage, header->v[1], 1); + struct prim_header newprim = *header; + + if (t0 > 0.0) { + screen_interp( stage->draw, v0new, t0, header->v[0], header->v[1] ); + newprim.v[0] = v0new; + } + + if (t1 < 1.0) { + screen_interp( stage->draw, v1new, t1, header->v[0], header->v[1] ); + newprim.v[1] = v1new; + } + + stage->next->line( stage->next, &newprim ); +} + + +static INLINE unsigned +stipple_test(int counter, ushort pattern, int factor) +{ + int b = (counter / factor) & 0xf; + return (1 << b) & pattern; +} + + +static void +stipple_line(struct draw_stage *stage, struct prim_header *header) +{ + struct stipple_stage *stipple = stipple_stage(stage); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const float *pos0 = v0->data[0]; + const float *pos1 = v1->data[0]; + float start = 0; + int state = 0; + + float x0 = pos0[0]; + float x1 = pos1[0]; + float y0 = pos0[1]; + float y1 = pos1[1]; + + float dx = x0 > x1 ? x0 - x1 : x1 - x0; + float dy = y0 > y1 ? y0 - y1 : y1 - y0; + + float length = MAX2(dx, dy); + int i; + + /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table. + */ + for (i = 0; i < length; i++) { + int result = stipple_test( (int) stipple->counter+i, + (ushort) stipple->pattern, stipple->factor ); + if (result != state) { + /* changing from "off" to "on" or vice versa */ + if (state) { + if (start != i) { + /* finishing an "on" segment */ + emit_segment( stage, header, start / length, i / length ); + } + } + else { + /* starting an "on" segment */ + start = (float) i; + } + state = result; + } + } + + if (state && start < length) + emit_segment( stage, header, start / length, 1.0 ); + + stipple->counter += length; +} + + +static void +reset_stipple_counter(struct draw_stage *stage) +{ + struct stipple_stage *stipple = stipple_stage(stage); + stipple->counter = 0; + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +stipple_first_line(struct draw_stage *stage, + struct prim_header *header) +{ + struct stipple_stage *stipple = stipple_stage(stage); + struct draw_context *draw = stage->draw; + + stipple->pattern = draw->rasterizer->line_stipple_pattern; + stipple->factor = draw->rasterizer->line_stipple_factor + 1; + + stage->line = stipple_line; + stage->line( stage, header ); +} + + +static void +stipple_flush(struct draw_stage *stage, unsigned flags) +{ + stage->line = stipple_first_line; + stage->next->flush( stage->next, flags ); +} + + +static void +passthrough_point(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->point( stage->next, header ); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + +static void +stipple_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create line stippler stage + */ +struct draw_stage *draw_stipple_stage( struct draw_context *draw ) +{ + struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); + + draw_alloc_temp_verts( &stipple->stage, 2 ); + + stipple->stage.draw = draw; + stipple->stage.next = NULL; + stipple->stage.point = passthrough_point; + stipple->stage.line = stipple_first_line; + stipple->stage.tri = passthrough_tri; + stipple->stage.reset_stipple_counter = reset_stipple_counter; + stipple->stage.flush = stipple_flush; + stipple->stage.destroy = stipple_destroy; + + return &stipple->stage; +} diff --git a/src/gallium/aux/draw/draw_twoside.c b/src/gallium/aux/draw/draw_twoside.c new file mode 100644 index 0000000000..1c38957987 --- /dev/null +++ b/src/gallium/aux/draw/draw_twoside.c @@ -0,0 +1,203 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +struct twoside_stage { + struct draw_stage stage; + float sign; /**< +1 or -1 */ + uint attrib_front0, attrib_back0; + uint attrib_front1, attrib_back1; +}; + + +static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage ) +{ + return (struct twoside_stage *)stage; +} + + + + +/** + * Copy back color(s) to front color(s). + */ +static INLINE struct vertex_header * +copy_bfc( struct twoside_stage *twoside, + const struct vertex_header *v, + unsigned idx ) +{ + struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); + + if (twoside->attrib_back0) { + COPY_4FV(tmp->data[twoside->attrib_front0], + tmp->data[twoside->attrib_back0]); + } + if (twoside->attrib_back1) { + COPY_4FV(tmp->data[twoside->attrib_front1], + tmp->data[twoside->attrib_back1]); + } + + return tmp; +} + + +/* Twoside tri: + */ +static void twoside_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + if (header->det * twoside->sign < 0.0) { + /* this is a back-facing triangle */ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + /* copy back attribs to front attribs */ + tmp.v[0] = copy_bfc(twoside, header->v[0], 0); + tmp.v[1] = copy_bfc(twoside, header->v[1], 1); + tmp.v[2] = copy_bfc(twoside, header->v[2], 2); + + stage->next->tri( stage->next, &tmp ); + } + else { + stage->next->tri( stage->next, header ); + } +} + + +static void twoside_line( struct draw_stage *stage, + struct prim_header *header ) +{ + /* pass-through */ + stage->next->line( stage->next, header ); +} + + +static void twoside_point( struct draw_stage *stage, + struct prim_header *header ) +{ + /* pass-through */ + stage->next->point( stage->next, header ); +} + + +static void twoside_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + twoside->attrib_front0 = 0; + twoside->attrib_front1 = 0; + twoside->attrib_back0 = 0; + twoside->attrib_back1 = 0; + + /* Find which vertex shader outputs are front/back colors */ + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { + if (vs->output_semantic_index[i] == 0) + twoside->attrib_front0 = i; + else + twoside->attrib_front1 = i; + } + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + if (vs->output_semantic_index[i] == 0) + twoside->attrib_back0 = i; + else + twoside->attrib_back1 = i; + } + } + + if (!twoside->attrib_back0) + twoside->attrib_front0 = 0; + + if (!twoside->attrib_back1) + twoside->attrib_front1 = 0; + + /* + * We'll multiply the primitive's determinant by this sign to determine + * if the triangle is back-facing (negative). + * sign = -1 for CCW, +1 for CW + */ + twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; + + stage->tri = twoside_tri; + stage->tri( stage, header ); +} + + +static void twoside_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->tri = twoside_first_tri; + stage->next->flush( stage->next, flags ); +} + + +static void twoside_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void twoside_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create twoside pipeline stage. + */ +struct draw_stage *draw_twoside_stage( struct draw_context *draw ) +{ + struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); + + draw_alloc_temp_verts( &twoside->stage, 3 ); + + twoside->stage.draw = draw; + twoside->stage.next = NULL; + twoside->stage.point = twoside_point; + twoside->stage.line = twoside_line; + twoside->stage.tri = twoside_first_tri; + twoside->stage.flush = twoside_flush; + twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; + twoside->stage.destroy = twoside_destroy; + + return &twoside->stage; +} diff --git a/src/gallium/aux/draw/draw_unfilled.c b/src/gallium/aux/draw/draw_unfilled.c new file mode 100644 index 0000000000..8777cfdfc8 --- /dev/null +++ b/src/gallium/aux/draw/draw_unfilled.c @@ -0,0 +1,206 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Drawing stage for handling glPolygonMode(line/point). + * Convert triangles to points or lines as needed. + */ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + +struct unfilled_stage { + struct draw_stage stage; + + /** [0] = front face, [1] = back face. + * legal values: PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE, + * and PIPE_POLYGON_MODE_POINT, + */ + unsigned mode[2]; +}; + + +static INLINE struct unfilled_stage *unfilled_stage( struct draw_stage *stage ) +{ + return (struct unfilled_stage *)stage; +} + + + +static void point( struct draw_stage *stage, + struct vertex_header *v0 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + stage->next->point( stage->next, &tmp ); +} + +static void line( struct draw_stage *stage, + struct vertex_header *v0, + struct vertex_header *v1 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + tmp.v[1] = v1; + stage->next->line( stage->next, &tmp ); +} + + +static void points( struct draw_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (header->edgeflags & 0x1) point( stage, v0 ); + if (header->edgeflags & 0x2) point( stage, v1 ); + if (header->edgeflags & 0x4) point( stage, v2 ); +} + + +static void lines( struct draw_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + +#if 0 + assert(((header->edgeflags & 0x1) >> 0) == header->v[0]->edgeflag); + assert(((header->edgeflags & 0x2) >> 1) == header->v[1]->edgeflag); + assert(((header->edgeflags & 0x4) >> 2) == header->v[2]->edgeflag); +#endif + + if (header->edgeflags & 0x1) line( stage, v0, v1 ); + if (header->edgeflags & 0x2) line( stage, v1, v2 ); + if (header->edgeflags & 0x4) line( stage, v2, v0 ); +} + + +/* Unfilled tri: + * + * Note edgeflags in the vertex struct is not sufficient as we will + * need to manipulate them when decomposing primitives??? + */ +static void unfilled_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + unsigned mode = unfilled->mode[header->det >= 0.0]; + + switch (mode) { + case PIPE_POLYGON_MODE_FILL: + stage->next->tri( stage->next, header ); + break; + case PIPE_POLYGON_MODE_LINE: + lines( stage, header ); + break; + case PIPE_POLYGON_MODE_POINT: + points( stage, header ); + break; + default: + abort(); + } +} + + +static void unfilled_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + + unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */ + unfilled->mode[1] = stage->draw->rasterizer->fill_cw; /* back */ + + stage->tri = unfilled_tri; + stage->tri( stage, header ); +} + + +static void unfilled_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void unfilled_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void unfilled_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->next->flush( stage->next, flags ); + + stage->tri = unfilled_first_tri; +} + + +static void unfilled_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void unfilled_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create unfilled triangle stage. + */ +struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) +{ + struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); + + draw_alloc_temp_verts( &unfilled->stage, 0 ); + + unfilled->stage.draw = draw; + unfilled->stage.next = NULL; + unfilled->stage.tmp = NULL; + unfilled->stage.point = unfilled_point; + unfilled->stage.line = unfilled_line; + unfilled->stage.tri = unfilled_first_tri; + unfilled->stage.flush = unfilled_flush; + unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; + unfilled->stage.destroy = unfilled_destroy; + + return &unfilled->stage; +} diff --git a/src/gallium/aux/draw/draw_validate.c b/src/gallium/aux/draw/draw_validate.c new file mode 100644 index 0000000000..4375ebabbc --- /dev/null +++ b/src/gallium/aux/draw/draw_validate.c @@ -0,0 +1,185 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + + + + +/** + * Rebuild the rendering pipeline. + */ +static struct draw_stage *validate_pipeline( struct draw_stage *stage ) +{ + struct draw_context *draw = stage->draw; + struct draw_stage *next = draw->pipeline.rasterize; + int need_det = 0; + int precalc_flat = 0; + + /* Set the validate's next stage to the rasterize stage, so that it + * can be found later if needed for flushing. + */ + stage->next = next; + + /* + * NOTE: we build up the pipeline in end-to-start order. + * + * TODO: make the current primitive part of the state and build + * shorter pipelines for lines & points. + */ + + if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines) || + (draw->rasterizer->point_size != 1.0 && draw->convert_wide_points) || + draw->rasterizer->point_sprite) { + draw->pipeline.wide->next = next; + next = draw->pipeline.wide; + } + + if (draw->rasterizer->line_stipple_enable) { + draw->pipeline.stipple->next = next; + next = draw->pipeline.stipple; + precalc_flat = 1; /* only needed for lines really */ + } + + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { + draw->pipeline.unfilled->next = next; + next = draw->pipeline.unfilled; + precalc_flat = 1; /* only needed for triangles really */ + need_det = 1; + } + + if (draw->rasterizer->flatshade && precalc_flat) { + draw->pipeline.flatshade->next = next; + next = draw->pipeline.flatshade; + } + + if (draw->rasterizer->offset_cw || + draw->rasterizer->offset_ccw) { + draw->pipeline.offset->next = next; + next = draw->pipeline.offset; + need_det = 1; + } + + if (draw->rasterizer->light_twoside) { + draw->pipeline.twoside->next = next; + next = draw->pipeline.twoside; + need_det = 1; + } + + /* Always run the cull stage as we calculate determinant there + * also. + * + * This can actually be a win as culling out the triangles can lead + * to less work emitting vertices, smaller vertex buffers, etc. + * It's difficult to say whether this will be true in general. + */ + if (need_det || draw->rasterizer->cull_mode) { + draw->pipeline.cull->next = next; + next = draw->pipeline.cull; + } + + /* Clip stage + */ + if (!draw->rasterizer->bypass_clipping) + { + draw->pipeline.clip->next = next; + next = draw->pipeline.clip; + } + + + draw->pipeline.first = next; + return next; +} + +static void validate_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->tri( pipeline, header ); +} + +static void validate_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->line( pipeline, header ); +} + +static void validate_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->point( pipeline, header ); +} + +static void validate_reset_stipple_counter( struct draw_stage *stage ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->reset_stipple_counter( pipeline ); +} + +static void validate_flush( struct draw_stage *stage, + unsigned flags ) +{ + /* May need to pass a backend flush on to the rasterize stage. + */ + if (stage->next) + stage->next->flush( stage->next, flags ); +} + + +static void validate_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + +/** + * Create validate pipeline stage. + */ +struct draw_stage *draw_validate_stage( struct draw_context *draw ) +{ + struct draw_stage *stage = CALLOC_STRUCT(draw_stage); + + stage->draw = draw; + stage->next = NULL; + stage->point = validate_point; + stage->line = validate_line; + stage->tri = validate_tri; + stage->flush = validate_flush; + stage->reset_stipple_counter = validate_reset_stipple_counter; + stage->destroy = validate_destroy; + + return stage; +} diff --git a/src/gallium/aux/draw/draw_vbuf.c b/src/gallium/aux/draw/draw_vbuf.c new file mode 100644 index 0000000000..71ac73912b --- /dev/null +++ b/src/gallium/aux/draw/draw_vbuf.c @@ -0,0 +1,570 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Vertex buffer drawing stage. + * + * \author José Fonseca + * \author Keith Whitwell + */ + + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" + +#include "draw_vbuf.h" +#include "draw_private.h" +#include "draw_vertex.h" +#include "draw_vf.h" + + +/** + * Vertex buffer emit stage. + */ +struct vbuf_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct vbuf_render *render; + + const struct vertex_info *vinfo; + + /** Vertex size in bytes */ + unsigned vertex_size; + + struct draw_vertex_fetch *vf; + + /* FIXME: we have no guarantee that 'unsigned' is 32bit */ + + /** Vertices in hardware format */ + unsigned *vertices; + unsigned *vertex_ptr; + unsigned max_vertices; + unsigned nr_vertices; + + /** Indices */ + ushort *indices; + unsigned max_indices; + unsigned nr_indices; + + /** Pipe primitive */ + unsigned prim; +}; + + +/** + * Basically a cast wrapper. + */ +static INLINE struct vbuf_stage * +vbuf_stage( struct draw_stage *stage ) +{ + assert(stage); + return (struct vbuf_stage *)stage; +} + + +static void vbuf_flush_indices( struct vbuf_stage *vbuf ); +static void vbuf_flush_vertices( struct vbuf_stage *vbuf ); +static void vbuf_alloc_vertices( struct vbuf_stage *vbuf ); + + +static INLINE boolean +overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) +{ + unsigned long used = (unsigned long) ((char *)ptr - (char *)map); + return (used + bytes) > bufsz; +} + + +static INLINE void +check_space( struct vbuf_stage *vbuf, unsigned nr ) +{ + if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { + vbuf_flush_vertices(vbuf); + vbuf_alloc_vertices(vbuf); + } + + if (vbuf->nr_indices + nr > vbuf->max_indices ) + vbuf_flush_indices(vbuf); +} + + +#if 0 +static INLINE void +dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) +{ + assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); + unsigned i, j, k; + + for (i = 0; i < vinfo->num_attribs; i++) { + j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + debug_printf("EMIT_OMIT:"); + break; + case EMIT_ALL: + assert(i == 0); + assert(j == 0); + debug_printf("EMIT_ALL:\t"); + for(k = 0; k < vinfo->size*4; ++k) + debug_printf("%02x ", *data++); + break; + case EMIT_1F: + debug_printf("EMIT_1F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_1F_PSIZE: + debug_printf("EMIT_1F_PSIZE:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_2F: + debug_printf("EMIT_2F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_3F: + debug_printf("EMIT_3F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + data += sizeof(float); + break; + case EMIT_4F: + debug_printf("EMIT_4F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_4UB: + debug_printf("EMIT_4UB:\t"); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + break; + default: + assert(0); + } + debug_printf("\n"); + } + debug_printf("\n"); +} +#endif + + +/** + * Extract the needed fields from post-transformed vertex and emit + * a hardware(driver) vertex. + * Recall that the vertices are constructed by the 'draw' module and + * have a couple of slots at the beginning (1-dword header, 4-dword + * clip pos) that we ignore here. We only use the vertex->data[] fields. + */ +static INLINE void +emit_vertex( struct vbuf_stage *vbuf, + struct vertex_header *vertex ) +{ +#if 0 + debug_printf("emit vertex %d to %p\n", + vbuf->nr_vertices, vbuf->vertex_ptr); +#endif + + if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { + if(vertex->vertex_id < vbuf->nr_vertices) + return; + else + debug_printf("Bad vertex id 0x%04x (>= 0x%04x)\n", + vertex->vertex_id, vbuf->nr_vertices); + return; + } + + vertex->vertex_id = vbuf->nr_vertices++; + + if(!vbuf->vf) { + const struct vertex_info *vinfo = vbuf->vinfo; + uint i; + uint count = 0; /* for debug/sanity */ + + assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); + + for (i = 0; i < vinfo->num_attribs; i++) { + uint j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + /* no-op */ + break; + case EMIT_ALL: + /* just copy the whole vertex as-is to the vbuf */ + assert(i == 0); + assert(j == 0); + memcpy(vbuf->vertex_ptr, vertex, vinfo->size * 4); + vbuf->vertex_ptr += vinfo->size; + count += vinfo->size; + break; + case EMIT_1F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + count++; + break; + case EMIT_1F_PSIZE: + *vbuf->vertex_ptr++ = fui(vbuf->stage.draw->rasterizer->point_size); + count++; + break; + case EMIT_2F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + count += 2; + break; + case EMIT_3F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); + count += 3; + break; + case EMIT_4F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][3]); + count += 4; + break; + case EMIT_4UB: + *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), + float_to_ubyte( vertex->data[j][1] ), + float_to_ubyte( vertex->data[j][0] ), + float_to_ubyte( vertex->data[j][3] )); + count += 1; + break; + default: + assert(0); + } + } + assert(count == vinfo->size); +#if 0 + { + static float data[256]; + draw_vf_emit_vertex(vbuf->vf, vertex, data); + if(memcmp((uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size, data, vbuf->vertex_size)) { + debug_printf("With VF:\n"); + dump_emitted_vertex(vbuf->vinfo, (uint8_t *)data); + debug_printf("Without VF:\n"); + dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size); + assert(0); + } + } +#endif + } + else { + draw_vf_emit_vertex(vbuf->vf, vertex, vbuf->vertex_ptr); + + vbuf->vertex_ptr += vbuf->vertex_size/4; + } +} + + +static void +vbuf_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + check_space( vbuf, 3 ); + + for (i = 0; i < 3; i++) { + emit_vertex( vbuf, prim->v[i] ); + + vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; + } +} + + +static void +vbuf_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + check_space( vbuf, 2 ); + + for (i = 0; i < 2; i++) { + emit_vertex( vbuf, prim->v[i] ); + + vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; + } +} + + +static void +vbuf_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + check_space( vbuf, 1 ); + + emit_vertex( vbuf, prim->v[0] ); + + vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[0]->vertex_id; +} + + +/** + * Set the prim type for subsequent vertices. + * This may result in a new vertex size. The existing vbuffer (if any) + * will be flushed if needed and a new one allocated. + */ +static void +vbuf_set_prim( struct vbuf_stage *vbuf, uint newprim ) +{ + const struct vertex_info *vinfo; + unsigned vertex_size; + + assert(newprim == PIPE_PRIM_POINTS || + newprim == PIPE_PRIM_LINES || + newprim == PIPE_PRIM_TRIANGLES); + + vbuf->prim = newprim; + vbuf->render->set_primitive(vbuf->render, newprim); + + vinfo = vbuf->render->get_vertex_info(vbuf->render); + vertex_size = vinfo->size * sizeof(float); + + if (vertex_size != vbuf->vertex_size) + vbuf_flush_vertices(vbuf); + + vbuf->vinfo = vinfo; + vbuf->vertex_size = vertex_size; + if(vbuf->vf) + draw_vf_set_vertex_info(vbuf->vf, + vbuf->vinfo, + vbuf->stage.draw->rasterizer->point_size); + + if (!vbuf->vertices) + vbuf_alloc_vertices(vbuf); +} + + +static void +vbuf_first_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + stage->tri = vbuf_tri; + vbuf_set_prim(vbuf, PIPE_PRIM_TRIANGLES); + stage->tri( stage, prim ); +} + + +static void +vbuf_first_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + stage->line = vbuf_line; + vbuf_set_prim(vbuf, PIPE_PRIM_LINES); + stage->line( stage, prim ); +} + + +static void +vbuf_first_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + stage->point = vbuf_point; + vbuf_set_prim(vbuf, PIPE_PRIM_POINTS); + stage->point( stage, prim ); +} + + +static void +vbuf_flush_indices( struct vbuf_stage *vbuf ) +{ + if(!vbuf->nr_indices) + return; + + assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == + vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); + + switch(vbuf->prim) { + case PIPE_PRIM_POINTS: + break; + case PIPE_PRIM_LINES: + assert(vbuf->nr_indices % 2 == 0); + break; + case PIPE_PRIM_TRIANGLES: + assert(vbuf->nr_indices % 3 == 0); + break; + default: + assert(0); + } + + vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); + + vbuf->nr_indices = 0; + + /* don't need to reset point/line/tri functions */ +#if 0 + stage->point = vbuf_first_point; + stage->line = vbuf_first_line; + stage->tri = vbuf_first_tri; +#endif +} + + +/** + * Flush existing vertex buffer and allocate a new one. + * + * XXX: We separate flush-on-index-full and flush-on-vb-full, but may + * raise issues uploading vertices if the hardware wants to flush when + * we flush. + */ +static void +vbuf_flush_vertices( struct vbuf_stage *vbuf ) +{ + if(vbuf->vertices) { + vbuf_flush_indices(vbuf); + + /* Reset temporary vertices ids */ + if(vbuf->nr_vertices) + draw_reset_vertex_ids( vbuf->stage.draw ); + + /* Free the vertex buffer */ + vbuf->render->release_vertices(vbuf->render, + vbuf->vertices, + vbuf->vertex_size, + vbuf->nr_vertices); + vbuf->max_vertices = vbuf->nr_vertices = 0; + vbuf->vertex_ptr = vbuf->vertices = NULL; + + } +} + + +static void +vbuf_alloc_vertices( struct vbuf_stage *vbuf ) +{ + assert(!vbuf->nr_indices); + assert(!vbuf->vertices); + + /* Allocate a new vertex buffer */ + vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; + vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, + (ushort) vbuf->vertex_size, + (ushort) vbuf->max_vertices); + vbuf->vertex_ptr = vbuf->vertices; +} + + + +static void +vbuf_flush( struct draw_stage *stage, unsigned flags ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + + stage->point = vbuf_first_point; + stage->line = vbuf_first_line; + stage->tri = vbuf_first_tri; + + if (flags & DRAW_FLUSH_BACKEND) + vbuf_flush_vertices( vbuf ); +} + + +static void +vbuf_reset_stipple_counter( struct draw_stage *stage ) +{ + /* XXX: Need to do something here for hardware with linestipple. + */ + (void) stage; +} + + +static void vbuf_destroy( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + if(vbuf->indices) + align_free( vbuf->indices ); + + if(vbuf->vf) + draw_vf_destroy( vbuf->vf ); + + if (vbuf->render) + vbuf->render->destroy( vbuf->render ); + + FREE( stage ); +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *draw_vbuf_stage( struct draw_context *draw, + struct vbuf_render *render ) +{ + struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); + + if(!vbuf) + return NULL; + + vbuf->stage.draw = draw; + vbuf->stage.point = vbuf_first_point; + vbuf->stage.line = vbuf_first_line; + vbuf->stage.tri = vbuf_first_tri; + vbuf->stage.flush = vbuf_flush; + vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter; + vbuf->stage.destroy = vbuf_destroy; + + vbuf->render = render; + + assert(render->max_indices < UNDEFINED_VERTEX_ID); + vbuf->max_indices = render->max_indices; + vbuf->indices = (ushort *) + align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); + if(!vbuf->indices) + vbuf_destroy(&vbuf->stage); + + vbuf->vertices = NULL; + vbuf->vertex_ptr = vbuf->vertices; + + vbuf->prim = ~0; + + if(!GETENV("GALLIUM_NOVF")) + vbuf->vf = draw_vf_create(); + + return &vbuf->stage; +} diff --git a/src/gallium/aux/draw/draw_vbuf.h b/src/gallium/aux/draw/draw_vbuf.h new file mode 100644 index 0000000000..cfd2b9820c --- /dev/null +++ b/src/gallium/aux/draw/draw_vbuf.h @@ -0,0 +1,106 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Vertex buffer drawing stage. + * + * \author Keith Whitwell + * \author José Fonseca + */ + +#ifndef DRAW_VBUF_H_ +#define DRAW_VBUF_H_ + + +#include "pipe/p_util.h" + + +struct draw_context; +struct vertex_info; + + +/** + * Interface for hardware vertex buffer rendering. + */ +struct vbuf_render { + + /** + * Driver limits. May be tuned lower to improve cache hits on + * index list. + */ + unsigned max_indices; + unsigned max_vertex_buffer_bytes; + + /** + * Get the hardware vertex format. + * + * XXX: have this in draw_context instead? + */ + const struct vertex_info *(*get_vertex_info)( struct vbuf_render * ); + + /** + * Request a destination for vertices. + * Hardware renderers will use ttm memory, others will just malloc + * something. + */ + void *(*allocate_vertices)( struct vbuf_render *, + ushort vertex_size, + ushort nr_vertices ); + + /** + * Notify the renderer of the current primitive when it changes. + * Prim is restricted to TRIANGLES, LINES and POINTS. + */ + void (*set_primitive)( struct vbuf_render *, unsigned prim ); + + /** + * DrawElements, note indices are ushort: + */ + void (*draw)( struct vbuf_render *, + const ushort *indices, + uint nr_indices ); + + /** + * Called when vbuf is done with this set of vertices: + */ + void (*release_vertices)( struct vbuf_render *, + void *vertices, + unsigned vertex_size, + unsigned vertices_used ); + + void (*destroy)( struct vbuf_render * ); +}; + + + +struct draw_stage * +draw_vbuf_stage( struct draw_context *draw, + struct vbuf_render *render ); + + +#endif /*DRAW_VBUF_H_*/ diff --git a/src/gallium/aux/draw/draw_vertex.c b/src/gallium/aux/draw/draw_vertex.c new file mode 100644 index 0000000000..2d6592150f --- /dev/null +++ b/src/gallium/aux/draw/draw_vertex.c @@ -0,0 +1,79 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * Functions for specifying the post-transformation vertex layout. + * + * Author: + * Brian Paul + * Keith Whitwell + */ + + +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_vertex.h" + + +/** + * Compute the size of a vertex, in dwords/floats, to update the + * vinfo->size field. + */ +void +draw_compute_vertex_size(struct vertex_info *vinfo) +{ + uint i; + + vinfo->size = 0; + for (i = 0; i < vinfo->num_attribs; i++) { + switch (vinfo->emit[i]) { + case EMIT_OMIT: + break; + case EMIT_4UB: + /* fall-through */ + case EMIT_1F_PSIZE: + /* fall-through */ + case EMIT_1F: + vinfo->size += 1; + break; + case EMIT_2F: + vinfo->size += 2; + break; + case EMIT_3F: + vinfo->size += 3; + break; + case EMIT_4F: + vinfo->size += 4; + break; + case EMIT_ALL: + /* fall-through */ + default: + assert(0); + } + } + + assert(vinfo->size * 4 <= MAX_VERTEX_SIZE); +} diff --git a/src/gallium/aux/draw/draw_vertex.h b/src/gallium/aux/draw/draw_vertex.h new file mode 100644 index 0000000000..267c74203b --- /dev/null +++ b/src/gallium/aux/draw/draw_vertex.h @@ -0,0 +1,111 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Post-transform vertex format info. The vertex_info struct is used by + * the draw_vbuf code to emit hardware-specific vertex layouts into hw + * vertex buffers. + * + * Author: + * Brian Paul + */ + + +#ifndef DRAW_VERTEX_H +#define DRAW_VERTEX_H + + +#include "pipe/p_state.h" + + +/** + * Vertex attribute emit modes + */ +enum attrib_emit { + EMIT_OMIT, /**< don't emit the attribute */ + EMIT_ALL, /**< emit whole post-xform vertex, w/ header */ + EMIT_1F, + EMIT_1F_PSIZE, /**< insert constant point size */ + EMIT_2F, + EMIT_3F, + EMIT_4F, + EMIT_4UB /**< XXX may need variations for RGBA vs BGRA, etc */ +}; + + +/** + * Attribute interpolation mode + */ +enum interp_mode { + INTERP_NONE, /**< never interpolate vertex header info */ + INTERP_POS, /**< special case for frag position */ + INTERP_CONSTANT, + INTERP_LINEAR, + INTERP_PERSPECTIVE +}; + + +/** + * Information about hardware/rasterization vertex layout. + */ +struct vertex_info +{ + uint num_attribs; + uint hwfmt[4]; /**< hardware format info for this format */ + enum interp_mode interp_mode[PIPE_MAX_SHADER_INPUTS]; + enum attrib_emit emit[PIPE_MAX_SHADER_INPUTS]; /**< EMIT_x */ + uint src_index[PIPE_MAX_SHADER_INPUTS]; /**< map to post-xform attribs */ + uint size; /**< total vertex size in dwords */ +}; + + + +/** + * Add another attribute to the given vertex_info object. + * \param src_index indicates which post-transformed vertex attrib slot + * corresponds to this attribute. + * \return slot in which the attribute was added + */ +static INLINE uint +draw_emit_vertex_attr(struct vertex_info *vinfo, + enum attrib_emit emit, enum interp_mode interp, + uint src_index) +{ + const uint n = vinfo->num_attribs; + assert(n < PIPE_MAX_SHADER_INPUTS); + vinfo->emit[n] = emit; + vinfo->interp_mode[n] = interp; + vinfo->src_index[n] = src_index; + vinfo->num_attribs++; + return n; +} + + +extern void draw_compute_vertex_size(struct vertex_info *vinfo); + + +#endif /* DRAW_VERTEX_H */ diff --git a/src/gallium/aux/draw/draw_vertex_cache.c b/src/gallium/aux/draw/draw_vertex_cache.c new file mode 100644 index 0000000000..44427999cc --- /dev/null +++ b/src/gallium/aux/draw/draw_vertex_cache.c @@ -0,0 +1,196 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw_private.h" +#include "draw_context.h" + + +void draw_vertex_cache_invalidate( struct draw_context *draw ) +{ + assert(draw->pq.queue_nr == 0); + assert(draw->vs.queue_nr == 0); + assert(draw->vcache.referenced == 0); + + memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); +} + + +/** + * Check if vertex is in cache, otherwise add it. It won't go through + * VS yet, not until there is a flush operation or the VS queue fills up. + * + * Note that cache entries are basically just two pointers: the first + * an index into the user's vertex arrays, the second a location in + * the vertex shader cache for the post-transformed vertex. + * + * \return pointer to location of (post-transformed) vertex header in the cache + */ +static struct vertex_header *get_vertex( struct draw_context *draw, + unsigned i ) +{ + unsigned slot = (i + (i>>5)) % VCACHE_SIZE; + + assert(slot < 32); /* so we don't exceed the bitfield size below */ + + /* Cache miss? + */ + if (draw->vcache.idx[slot] != i) { + + /* If slot is in use, use the overflow area: + */ + if (draw->vcache.referenced & (1 << slot)) { + slot = VCACHE_SIZE + draw->vcache.overflow++; + } + + assert(slot < Elements(draw->vcache.idx)); + + draw->vcache.idx[slot] = i; + + /* Add to vertex shader queue: + */ + assert(draw->vs.queue_nr < VS_QUEUE_LENGTH); + draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; + draw->vs.queue[draw->vs.queue_nr].elt = i; + draw->vs.queue_nr++; + + /* Need to set the vertex's edge flag here. If we're being called + * by do_ef_triangle(), that function needs edge flag info! + */ + draw->vcache.vertex[slot]->clipmask = 0; + draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ + draw->vcache.vertex[slot]->pad = 0; + draw->vcache.vertex[slot]->vertex_id = UNDEFINED_VERTEX_ID; + } + + + /* primitive flushing may have cleared the bitfield but did not + * clear the idx[] array values. Set the bit now. This fixes a + * bug found when drawing long triangle fans. + */ + draw->vcache.referenced |= (1 << slot); + return draw->vcache.vertex[slot]; +} + + +static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const unsigned *elts = (const unsigned *) draw->user.elts; + return get_vertex( draw, elts[i] ); +} + + +static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const ushort *elts = (const ushort *) draw->user.elts; + return get_vertex( draw, elts[i] ); +} + + +static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const ubyte *elts = (const ubyte *) draw->user.elts; + return get_vertex( draw, elts[i] ); +} + + +void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) +{ + unsigned i; + + for (i = 0; i < Elements(draw->vcache.vertex); i++) + draw->vcache.vertex[i]->vertex_id = UNDEFINED_VERTEX_ID; +} + + +void draw_vertex_cache_unreference( struct draw_context *draw ) +{ + draw->vcache.referenced = 0; + draw->vcache.overflow = 0; +} + + +int draw_vertex_cache_check_space( struct draw_context *draw, + unsigned nr_verts ) +{ + if (draw->vcache.overflow + nr_verts < VCACHE_OVERFLOW) { + /* The vs queue is sized so that this can never happen: + */ + assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); + return TRUE; + } + else + return FALSE; +} + + + +/** + * Tell the drawing context about the index/element buffer to use + * (ala glDrawElements) + * If no element buffer is to be used (i.e. glDrawArrays) then this + * should be called with eltSize=0 and elements=NULL. + * + * \param draw the drawing context + * \param eltSize size of each element (1, 2 or 4 bytes) + * \param elements the element buffer ptr + */ +void +draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ) +{ +// draw_statechange( draw ); + + /* choose the get_vertex() function to use */ + switch (eltSize) { + case 0: + draw->vcache.get_vertex = get_vertex; + break; + case 1: + draw->vcache.get_vertex = get_ubyte_elt_vertex; + break; + case 2: + draw->vcache.get_vertex = get_ushort_elt_vertex; + break; + case 4: + draw->vcache.get_vertex = get_uint_elt_vertex; + break; + default: + assert(0); + } + draw->user.elts = elements; + draw->user.eltSize = eltSize; +} + diff --git a/src/gallium/aux/draw/draw_vertex_fetch.c b/src/gallium/aux/draw/draw_vertex_fetch.c new file mode 100644 index 0000000000..e13df04605 --- /dev/null +++ b/src/gallium/aux/draw/draw_vertex_fetch.c @@ -0,0 +1,510 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" +#include "draw_context.h" + + +#define DRAW_DBG 0 + + +/** + * Fetch a float[4] vertex attribute from memory, doing format/type + * conversion as needed. + * + * This is probably needed/dupliocated elsewhere, eg format + * conversion, texture sampling etc. + */ +#define FETCH_ATTRIB( NAME, SZ, CVT ) \ +static void \ +fetch_##NAME(const void *ptr, float *attrib) \ +{ \ + static const float defaults[4] = { 0,0,0,1 }; \ + int i; \ + \ + for (i = 0; i < SZ; i++) { \ + attrib[i] = CVT; \ + } \ + \ + for (; i < 4; i++) { \ + attrib[i] = defaults[i]; \ + } \ +} + +#define CVT_64_FLOAT (float) ((double *) ptr)[i] +#define CVT_32_FLOAT ((float *) ptr)[i] + +#define CVT_8_USCALED (float) ((unsigned char *) ptr)[i] +#define CVT_16_USCALED (float) ((unsigned short *) ptr)[i] +#define CVT_32_USCALED (float) ((unsigned int *) ptr)[i] + +#define CVT_8_SSCALED (float) ((char *) ptr)[i] +#define CVT_16_SSCALED (float) ((short *) ptr)[i] +#define CVT_32_SSCALED (float) ((int *) ptr)[i] + +#define CVT_8_UNORM (float) ((unsigned char *) ptr)[i] / 255.0f +#define CVT_16_UNORM (float) ((unsigned short *) ptr)[i] / 65535.0f +#define CVT_32_UNORM (float) ((unsigned int *) ptr)[i] / 4294967295.0f + +#define CVT_8_SNORM (float) ((char *) ptr)[i] / 127.0f +#define CVT_16_SNORM (float) ((short *) ptr)[i] / 32767.0f +#define CVT_32_SNORM (float) ((int *) ptr)[i] / 2147483647.0f + +FETCH_ATTRIB( R64G64B64A64_FLOAT, 4, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) +FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) + +FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) +FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) + +FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) +FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) + +FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) +FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) + +FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) +FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) + +FETCH_ATTRIB( R32G32B32A32_SNORM, 4, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32B32_SNORM, 3, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32_SNORM, 2, CVT_32_SNORM ) +FETCH_ATTRIB( R32_SNORM, 1, CVT_32_SNORM ) + +FETCH_ATTRIB( R16G16B16A16_USCALED, 4, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16B16_USCALED, 3, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16_USCALED, 2, CVT_16_USCALED ) +FETCH_ATTRIB( R16_USCALED, 1, CVT_16_USCALED ) + +FETCH_ATTRIB( R16G16B16A16_SSCALED, 4, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16B16_SSCALED, 3, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16_SSCALED, 2, CVT_16_SSCALED ) +FETCH_ATTRIB( R16_SSCALED, 1, CVT_16_SSCALED ) + +FETCH_ATTRIB( R16G16B16A16_UNORM, 4, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16B16_UNORM, 3, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16_UNORM, 2, CVT_16_UNORM ) +FETCH_ATTRIB( R16_UNORM, 1, CVT_16_UNORM ) + +FETCH_ATTRIB( R16G16B16A16_SNORM, 4, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16B16_SNORM, 3, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16_SNORM, 2, CVT_16_SNORM ) +FETCH_ATTRIB( R16_SNORM, 1, CVT_16_SNORM ) + +FETCH_ATTRIB( R8G8B8A8_USCALED, 4, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8B8_USCALED, 3, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8_USCALED, 2, CVT_8_USCALED ) +FETCH_ATTRIB( R8_USCALED, 1, CVT_8_USCALED ) + +FETCH_ATTRIB( R8G8B8A8_SSCALED, 4, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8B8_SSCALED, 3, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8_SSCALED, 2, CVT_8_SSCALED ) +FETCH_ATTRIB( R8_SSCALED, 1, CVT_8_SSCALED ) + +FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8B8_UNORM, 3, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8_UNORM, 2, CVT_8_UNORM ) +FETCH_ATTRIB( R8_UNORM, 1, CVT_8_UNORM ) + +FETCH_ATTRIB( R8G8B8A8_SNORM, 4, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8B8_SNORM, 3, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8_SNORM, 2, CVT_8_SNORM ) +FETCH_ATTRIB( R8_SNORM, 1, CVT_8_SNORM ) + +FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) +//FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) + + + +static fetch_func get_fetch_func( enum pipe_format format ) +{ +#if 0 + { + char tmp[80]; + pf_sprint_name(tmp, format); + debug_printf("%s: %s\n", __FUNCTION__, tmp); + } +#endif + + switch (format) { + case PIPE_FORMAT_R64_FLOAT: + return fetch_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return fetch_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return fetch_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return fetch_R64G64B64A64_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + return fetch_R32_FLOAT; + case PIPE_FORMAT_R32G32_FLOAT: + return fetch_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return fetch_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return fetch_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return fetch_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return fetch_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return fetch_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return fetch_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return fetch_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return fetch_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return fetch_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return fetch_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return fetch_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return fetch_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return fetch_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return fetch_R32G32B32A32_SNORM; + + case PIPE_FORMAT_R32_SSCALED: + return fetch_R32_SSCALED; + case PIPE_FORMAT_R32G32_SSCALED: + return fetch_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return fetch_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return fetch_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return fetch_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return fetch_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return fetch_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return fetch_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return fetch_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return fetch_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return fetch_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return fetch_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return fetch_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return fetch_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return fetch_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return fetch_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return fetch_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return fetch_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return fetch_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return fetch_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return fetch_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return fetch_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return fetch_R8G8B8_UNORM; + case PIPE_FORMAT_R8G8B8A8_UNORM: + return fetch_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return fetch_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return fetch_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return fetch_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return fetch_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return fetch_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return fetch_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return fetch_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return fetch_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return fetch_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return fetch_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return fetch_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return fetch_R8G8B8A8_SSCALED; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + return fetch_A8R8G8B8_UNORM; + + case 0: + return NULL; /* not sure why this is needed */ + + default: + assert(0); + return NULL; + } +} + + +static void +transpose_4x4( float *out, const float *in ) +{ + /* This can be achieved in 12 sse instructions, plus the final + * stores I guess. This is probably a bit more than that - maybe + * 32 or so? + */ + out[0] = in[0]; out[1] = in[4]; out[2] = in[8]; out[3] = in[12]; + out[4] = in[1]; out[5] = in[5]; out[6] = in[9]; out[7] = in[13]; + out[8] = in[2]; out[9] = in[6]; out[10] = in[10]; out[11] = in[14]; + out[12] = in[3]; out[13] = in[7]; out[14] = in[11]; out[15] = in[15]; +} + + + +static void fetch_xyz_rgb( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + + assert(count <= 4); + +// debug_printf("%s\n", __FUNCTION__); + + /* loop over vertex attributes (vertex shader inputs) + */ + + for (i = 0; i < 4; i++) { + { + const float *in = (const float *)(src[0] + elts[i] * pitch[0]); + float *out = &machine->Inputs[0].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[1] + elts[i] * pitch[1]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + } +} + + + + +static void fetch_xyz_rgb_st( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + + assert(count <= 4); + + /* loop over vertex attributes (vertex shader inputs) + */ + + for (i = 0; i < 4; i++) { + { + const float *in = (const float *)(src[0] + elts[i] * pitch[0]); + float *out = &machine->Inputs[0].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[1] + elts[i] * pitch[1]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[2] + elts[i] * pitch[2]); + float *out = &machine->Inputs[2].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = 0.0f; + out[12] = 1.0f; + } + } +} + + + + +/** + * Fetch vertex attributes for 'count' vertices. + */ +static void generic_vertex_fetch( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + unsigned nr_attrs = draw->vertex_fetch.nr_attrs; + unsigned attr; + + assert(count <= 4); + +// debug_printf("%s %d\n", __FUNCTION__, count); + + /* loop over vertex attributes (vertex shader inputs) + */ + for (attr = 0; attr < nr_attrs; attr++) { + + const unsigned pitch = draw->vertex_fetch.pitch[attr]; + const ubyte *src = draw->vertex_fetch.src_ptr[attr]; + const fetch_func fetch = draw->vertex_fetch.fetch[attr]; + unsigned i; + float p[4][4]; + + + /* Fetch four attributes for four vertices. + * + * Could fetch directly into AOS format, but this is meant to be + * a prototype for an sse implementation, which would have + * difficulties doing that. + */ + for (i = 0; i < count; i++) + fetch( src + elts[i] * pitch, p[i] ); + + /* Be nice and zero out any missing vertices: + */ + for ( ; i < 4; i++) + p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; + + /* Transpose/swizzle into sse-friendly format. Currently + * assuming that all vertex shader inputs are float[4], but this + * isn't true -- if the vertex shader only wants tex0.xy, we + * could optimize for that. + * + * To do so fully without codegen would probably require an + * excessive number of fetch functions, but we could at least + * minimize the transpose step: + */ + transpose_4x4( (float *)&machine->Inputs[attr].xyzw[0].f[0], (float *)p ); + } +} + + + +void draw_update_vertex_fetch( struct draw_context *draw ) +{ + unsigned nr_attrs, i; + +// debug_printf("%s\n", __FUNCTION__); + + /* this may happend during context init */ + if (!draw->vertex_shader) + return; + + nr_attrs = draw->vertex_shader->state->num_inputs; + + for (i = 0; i < nr_attrs; i++) { + unsigned buf = draw->vertex_element[i].vertex_buffer_index; + enum pipe_format format = draw->vertex_element[i].src_format; + + draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[i].src_offset; + + draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; + draw->vertex_fetch.fetch[i] = get_fetch_func( format ); + } + + draw->vertex_fetch.nr_attrs = nr_attrs; + + draw->vertex_fetch.fetch_func = generic_vertex_fetch; + + switch (nr_attrs) { + case 2: + if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT) + draw->vertex_fetch.fetch_func = fetch_xyz_rgb; + break; + case 3: + if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_element[2].src_format == PIPE_FORMAT_R32G32_FLOAT) + draw->vertex_fetch.fetch_func = fetch_xyz_rgb_st; + break; + default: + break; + } + +} diff --git a/src/gallium/aux/draw/draw_vertex_shader.c b/src/gallium/aux/draw/draw_vertex_shader.c new file mode 100644 index 0000000000..c824c1407e --- /dev/null +++ b/src/gallium/aux/draw/draw_vertex_shader.c @@ -0,0 +1,325 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#if defined(__i386__) || defined(__386__) +#include "pipe/tgsi/exec/tgsi_sse2.h" +#endif +#include "draw_private.h" +#include "draw_context.h" + +#include "x86/rtasm/x86sse.h" +#include "pipe/llvm/gallivm.h" + + +#define DBG_VS 0 + + +static INLINE unsigned +compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<machine; + unsigned int j; + + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); + assert(draw->vertex_shader->state->output_semantic_name[0] + == TGSI_SEMANTIC_POSITION); + + /* Consts does not require 16 byte alignment. */ + machine->Consts = (float (*)[4]) draw->user.constants; + + machine->Inputs = ALIGN16_ASSIGN(inputs); + machine->Outputs = ALIGN16_ASSIGN(outputs); + + draw->vertex_fetch.fetch_func( draw, machine, elts, count ); + + /* run shader */ +#ifdef MESA_LLVM + if (1) { + struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; + gallivm_cpu_vs_exec(prog, + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps); + } else +#elif defined(__i386__) || defined(__386__) + if (draw->use_sse) { + /* SSE */ + /* cast away const */ + struct draw_vertex_shader *shader + = (struct draw_vertex_shader *)draw->vertex_shader; + codegen_function func + = (codegen_function) x86_get_func( &shader->sse2_program ); + + if (func) + func( + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps ); + else + /* interpreter */ + tgsi_exec_machine_run( machine ); + } + else +#endif + { + /* interpreter */ + tgsi_exec_machine_run( machine ); + } + + /* store machine results */ + for (j = 0; j < count; j++) { + unsigned slot; + float x, y, z, w; + + /* Handle attr[0] (position) specially: + * + * XXX: Computing the clipmask should be done in the vertex + * program as a set of DP4 instructions appended to the + * user-provided code. + */ + x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + +#if DBG_VS + debug_printf("output[%d]win: %f %f %f %f\n", j, + vOut[j]->data[0][0], + vOut[j]->data[0][1], + vOut[j]->data[0][2], + vOut[j]->data[0][3]); +#endif + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; +#if DBG_VS + debug_printf("output[%d][%d]: %f %f %f %f\n", j, slot, + vOut[j]->data[slot][0], + vOut[j]->data[slot][1], + vOut[j]->data[slot][2], + vOut[j]->data[slot][3]); +#endif + } + } /* loop over vertices */ +} + + +/** + * Run the vertex shader on all vertices in the vertex queue. + * Called by the draw module when the vertx cache needs to be flushed. + */ +void +draw_vertex_shader_queue_flush(struct draw_context *draw) +{ + unsigned i; + + assert(draw->vs.queue_nr != 0); + + /* XXX: do this on statechange: + */ + draw_update_vertex_fetch( draw ); + +// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); + + /* run vertex shader on vertex cache entries, four per invokation */ + for (i = 0; i < draw->vs.queue_nr; i += 4) { + struct vertex_header *dests[4]; + unsigned elts[4]; + int j, n = MIN2(4, draw->vs.queue_nr - i); + + for (j = 0; j < n; j++) { + elts[j] = draw->vs.queue[i + j].elt; + dests[j] = draw->vs.queue[i + j].dest; + } + + for ( ; j < 4; j++) { + elts[j] = elts[0]; + dests[j] = dests[0]; + } + + assert(n > 0); + assert(n <= 4); + + run_vertex_program(draw, elts, n, dests); + } + + draw->vs.queue_nr = 0; +} + + +struct draw_vertex_shader * +draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader) +{ + struct draw_vertex_shader *vs; + + vs = CALLOC_STRUCT( draw_vertex_shader ); + if (vs == NULL) { + return NULL; + } + + vs->state = shader; + +#ifdef MESA_LLVM + struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); + gallivm_ir_set_layout(ir, GALLIVM_SOA); + gallivm_ir_set_components(ir, 4); + gallivm_ir_fill_from_tgsi(ir, shader->tokens); + vs->llvm_prog = gallivm_ir_compile(ir); + gallivm_ir_delete(ir); + + draw->engine = gallivm_global_cpu_engine(); + if (!draw->engine) { + draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); + } + else { + gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); + } +#elif defined(__i386__) || defined(__386__) + if (draw->use_sse) { + /* cast-away const */ + struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; + + x86_init_func( &vs->sse2_program ); + if (!tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, + &vs->sse2_program )) { + x86_release_func( (struct x86_function *) &vs->sse2_program ); + fprintf(stdout /*err*/, + "tgsi_emit_sse2() failed, falling back to interpreter\n"); + } + } +#endif + + return vs; +} + + +void +draw_bind_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + draw->vertex_shader = dvs; + draw->num_vs_outputs = dvs->state->num_outputs; + + /* specify the vertex program to interpret/execute */ + tgsi_exec_machine_init(&draw->machine, + draw->vertex_shader->state->tokens, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/ ); +} + + +void +draw_delete_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs) +{ +#if defined(__i386__) || defined(__386__) + x86_release_func( (struct x86_function *) &dvs->sse2_program ); +#endif + + FREE( dvs ); +} diff --git a/src/gallium/aux/draw/draw_vf.c b/src/gallium/aux/draw/draw_vf.c new file mode 100644 index 0000000000..dc3a5ecd21 --- /dev/null +++ b/src/gallium/aux/draw/draw_vf.c @@ -0,0 +1,428 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + + +#include + +#include "pipe/p_compiler.h" +#include "pipe/p_util.h" + +#include "draw_vf.h" + + +#define DRAW_VF_DBG 0 + + +/* TODO: remove this */ +extern void +_mesa_exec_free( void *addr ); + + +static boolean match_fastpath( struct draw_vertex_fetch *vf, + const struct draw_vf_fastpath *fp) +{ + unsigned j; + + if (vf->attr_count != fp->attr_count) + return FALSE; + + for (j = 0; j < vf->attr_count; j++) + if (vf->attr[j].format != fp->attr[j].format || + vf->attr[j].inputsize != fp->attr[j].size || + vf->attr[j].vertoffset != fp->attr[j].offset) + return FALSE; + + if (fp->match_strides) { + if (vf->vertex_stride != fp->vertex_stride) + return FALSE; + + for (j = 0; j < vf->attr_count; j++) + if (vf->attr[j].inputstride != fp->attr[j].stride) + return FALSE; + } + + return TRUE; +} + +static boolean search_fastpath_emit( struct draw_vertex_fetch *vf ) +{ + struct draw_vf_fastpath *fp = vf->fastpath; + + for ( ; fp ; fp = fp->next) { + if (match_fastpath(vf, fp)) { + vf->emit = fp->func; + return TRUE; + } + } + + return FALSE; +} + +void draw_vf_register_fastpath( struct draw_vertex_fetch *vf, + boolean match_strides ) +{ + struct draw_vf_fastpath *fastpath = CALLOC_STRUCT(draw_vf_fastpath); + unsigned i; + + fastpath->vertex_stride = vf->vertex_stride; + fastpath->attr_count = vf->attr_count; + fastpath->match_strides = match_strides; + fastpath->func = vf->emit; + fastpath->attr = (struct draw_vf_attr_type *) + MALLOC(vf->attr_count * sizeof(fastpath->attr[0])); + + for (i = 0; i < vf->attr_count; i++) { + fastpath->attr[i].format = vf->attr[i].format; + fastpath->attr[i].stride = vf->attr[i].inputstride; + fastpath->attr[i].size = vf->attr[i].inputsize; + fastpath->attr[i].offset = vf->attr[i].vertoffset; + } + + fastpath->next = vf->fastpath; + vf->fastpath = fastpath; +} + + + + +/*********************************************************************** + * Build codegen functions or return generic ones: + */ +static void choose_emit_func( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *dest) +{ + vf->emit = NULL; + + /* Does this match an existing (hardwired, codegen or known-bad) + * fastpath? + */ + if (search_fastpath_emit(vf)) { + /* Use this result. If it is null, then it is already known + * that the current state will fail for codegen and there is no + * point trying again. + */ + } + else if (vf->codegen_emit) { + vf->codegen_emit( vf ); + } + + if (!vf->emit) { + draw_vf_generate_hardwired_emit(vf); + } + + /* Otherwise use the generic version: + */ + if (!vf->emit) + vf->emit = draw_vf_generic_emit; + + vf->emit( vf, count, dest ); +} + + + + + +/*********************************************************************** + * Public entrypoints, mostly dispatch to the above: + */ + + + +static unsigned +draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, + const struct draw_vf_attr_map *map, + unsigned nr, + unsigned vertex_stride ) +{ + unsigned offset = 0; + unsigned i, j; + + assert(nr < PIPE_ATTRIB_MAX); + + for (j = 0, i = 0; i < nr; i++) { + const unsigned format = map[i].format; + if (format == DRAW_EMIT_PAD) { +#if (DRAW_VF_DBG) + debug_printf("%d: pad %d, offset %d\n", i, + map[i].offset, offset); +#endif + + offset += map[i].offset; + + } + else { + vf->attr[j].attrib = map[i].attrib; + vf->attr[j].format = format; + vf->attr[j].insert = draw_vf_format_info[format].insert; + vf->attr[j].vertattrsize = draw_vf_format_info[format].attrsize; + vf->attr[j].vertoffset = offset; + vf->attr[j].isconst = draw_vf_format_info[format].isconst; + if(vf->attr[j].isconst) + memcpy(vf->attr[j].data, &map[i].data, vf->attr[j].vertattrsize); + +#if (DRAW_VF_DBG) + debug_printf("%d: %s, offset %d\n", i, + draw_vf_format_info[format].name, + vf->attr[j].vertoffset); +#endif + + offset += draw_vf_format_info[format].attrsize; + j++; + } + } + + vf->attr_count = j; + vf->vertex_stride = vertex_stride ? vertex_stride : offset; + vf->emit = choose_emit_func; + + assert(vf->vertex_stride >= offset); + return vf->vertex_stride; +} + + +void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, + const struct vertex_info *vinfo, + float point_size ) +{ + unsigned i, j, k; + struct draw_vf_attr *a = vf->attr; + struct draw_vf_attr_map attrs[PIPE_MAX_SHADER_INPUTS]; + unsigned count = 0; /* for debug/sanity */ + unsigned nr_attrs = 0; + + for (i = 0; i < vinfo->num_attribs; i++) { + j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + /* no-op */ + break; + case EMIT_ALL: { + /* just copy the whole vertex as-is to the vbuf */ + unsigned s = vinfo->size; + assert(i == 0); + assert(j == 0); + /* copy the vertex header */ + /* XXX: we actually don't copy the header, just pad it */ + attrs[nr_attrs].attrib = 0; + attrs[nr_attrs].format = DRAW_EMIT_PAD; + attrs[nr_attrs].offset = offsetof(struct vertex_header, data); + s -= offsetof(struct vertex_header, data)/4; + count += offsetof(struct vertex_header, data)/4; + nr_attrs++; + /* copy the vertex data */ + for(k = 0; k < (s & ~0x3); k += 4) { + attrs[nr_attrs].attrib = k/4; + attrs[nr_attrs].format = DRAW_EMIT_4F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 4; + } + /* tail */ + /* XXX: actually, this shouldn't be needed */ + attrs[nr_attrs].attrib = k/4; + attrs[nr_attrs].offset = 0; + switch(s & 0x3) { + case 0: + break; + case 1: + attrs[nr_attrs].format = DRAW_EMIT_1F; + nr_attrs++; + count += 1; + break; + case 2: + attrs[nr_attrs].format = DRAW_EMIT_2F; + nr_attrs++; + count += 2; + break; + case 3: + attrs[nr_attrs].format = DRAW_EMIT_3F; + nr_attrs++; + count += 3; + break; + } + break; + } + case EMIT_1F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_1F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count++; + break; + case EMIT_1F_PSIZE: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_1F_CONST; + attrs[nr_attrs].offset = 0; + attrs[nr_attrs].data.f[0] = point_size; + nr_attrs++; + count++; + break; + case EMIT_2F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_2F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 2; + break; + case EMIT_3F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_3F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 3; + break; + case EMIT_4F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_4F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 4; + break; + case EMIT_4UB: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_4UB_4F_BGRA; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 1; + break; + default: + assert(0); + } + } + + assert(count == vinfo->size); + + draw_vf_set_vertex_attributes(vf, + attrs, + nr_attrs, + vinfo->size * sizeof(float) ); + + for (j = 0; j < vf->attr_count; j++) { + a[j].inputsize = 4; + a[j].do_insert = a[j].insert[4 - 1]; + if(a[j].isconst) { + a[j].inputptr = a[j].data; + a[j].inputstride = 0; + } + } +} + + +#if 0 +/* Set attribute pointers, adjusted for start position: + */ +void draw_vf_set_sources( struct draw_vertex_fetch *vf, + GLvector4f * const sources[], + unsigned start ) +{ + struct draw_vf_attr *a = vf->attr; + unsigned j; + + for (j = 0; j < vf->attr_count; j++) { + const GLvector4f *vptr = sources[a[j].attrib]; + + if ((a[j].inputstride != vptr->stride) || + (a[j].inputsize != vptr->size)) + vf->emit = choose_emit_func; + + a[j].inputstride = vptr->stride; + a[j].inputsize = vptr->size; + a[j].do_insert = a[j].insert[vptr->size - 1]; + a[j].inputptr = ((uint8_t *)vptr->data) + start * vptr->stride; + } +} +#endif + + +/** + * Emit a vertex to dest. + */ +void draw_vf_emit_vertex( struct draw_vertex_fetch *vf, + struct vertex_header *vertex, + void *dest ) +{ + struct draw_vf_attr *a = vf->attr; + unsigned j; + + for (j = 0; j < vf->attr_count; j++) { + if (!a[j].isconst) { + a[j].inputptr = (uint8_t *)&vertex->data[a[j].attrib][0]; + a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ + } + } + + vf->emit( vf, 1, (uint8_t*) dest ); +} + + + +struct draw_vertex_fetch *draw_vf_create( void ) +{ + struct draw_vertex_fetch *vf = CALLOC_STRUCT(draw_vertex_fetch); + unsigned i; + + for (i = 0; i < PIPE_ATTRIB_MAX; i++) + vf->attr[i].vf = vf; + + vf->identity[0] = 0.0; + vf->identity[1] = 0.0; + vf->identity[2] = 0.0; + vf->identity[3] = 1.0; + + vf->codegen_emit = NULL; + +#ifdef USE_SSE_ASM + if (!GETENV("GALLIUM_NO_CODEGEN")) + vf->codegen_emit = draw_vf_generate_sse_emit; +#endif + + return vf; +} + + +void draw_vf_destroy( struct draw_vertex_fetch *vf ) +{ + struct draw_vf_fastpath *fp, *tmp; + + for (fp = vf->fastpath ; fp ; fp = tmp) { + tmp = fp->next; + FREE(fp->attr); + + /* KW: At the moment, fp->func is constrained to be allocated by + * _mesa_exec_alloc(), as the hardwired fastpaths in + * t_vertex_generic.c are handled specially. It would be nice + * to unify them, but this probably won't change until this + * module gets another overhaul. + */ + //_mesa_exec_free((void *) fp->func); + FREE(fp); + } + + vf->fastpath = NULL; + FREE(vf); +} diff --git a/src/gallium/aux/draw/draw_vf.h b/src/gallium/aux/draw/draw_vf.h new file mode 100644 index 0000000000..011c8f0ff1 --- /dev/null +++ b/src/gallium/aux/draw/draw_vf.h @@ -0,0 +1,236 @@ +/* + * Copyright 2008 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * Vertex fetch/store/convert code. This functionality is used in two places: + * 1. Vertex fetch/convert - to grab vertex data from incoming vertex + * arrays and convert to format needed by vertex shaders. + * 2. Vertex store/emit - to convert simple float[][4] vertex attributes + * (which is the organization used throughout the draw/prim pipeline) to + * hardware-specific formats and emit into hardware vertex buffers. + * + * + * Authors: + * Keith Whitwell + */ + +#ifndef DRAW_VF_H +#define DRAW_VF_H + + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" + +#include "draw_vertex.h" +#include "draw_private.h" /* for vertex_header */ + + +enum draw_vf_attr_format { + DRAW_EMIT_1F, + DRAW_EMIT_2F, + DRAW_EMIT_3F, + DRAW_EMIT_4F, + DRAW_EMIT_3F_XYW, /**< for projective texture */ + DRAW_EMIT_1UB_1F, /**< for fog coordinate */ + DRAW_EMIT_3UB_3F_RGB, /**< for specular color */ + DRAW_EMIT_3UB_3F_BGR, /**< for specular color */ + DRAW_EMIT_4UB_4F_RGBA, /**< for color */ + DRAW_EMIT_4UB_4F_BGRA, /**< for color */ + DRAW_EMIT_4UB_4F_ARGB, /**< for color */ + DRAW_EMIT_4UB_4F_ABGR, /**< for color */ + DRAW_EMIT_1F_CONST, + DRAW_EMIT_2F_CONST, + DRAW_EMIT_3F_CONST, + DRAW_EMIT_4F_CONST, + DRAW_EMIT_PAD, /**< leave a hole of 'offset' bytes */ + DRAW_EMIT_MAX +}; + +struct draw_vf_attr_map +{ + /** Input attribute number */ + unsigned attrib; + + enum draw_vf_attr_format format; + + unsigned offset; + + /** + * Constant data for DRAW_EMIT_*_CONST + */ + union { + uint8_t ub[4]; + float f[4]; + } data; +}; + +struct draw_vertex_fetch; + + + +#if 0 +unsigned +draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, + const struct draw_vf_attr_map *map, + unsigned nr, + unsigned vertex_stride ); +#endif + +void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, + const struct vertex_info *vinfo, + float point_size ); + +#if 0 +void +draw_vf_set_sources( struct draw_vertex_fetch *vf, + GLvector4f * const attrib[], + unsigned start ); +#endif + +void +draw_vf_emit_vertex( struct draw_vertex_fetch *vf, + struct vertex_header *vertex, + void *dest ); + +struct draw_vertex_fetch * +draw_vf_create( void ); + +void +draw_vf_destroy( struct draw_vertex_fetch *vf ); + + + +/*********************************************************************** + * Internal functions and structs: + */ + +struct draw_vf_attr; + +typedef void (*draw_vf_extract_func)( const struct draw_vf_attr *a, + float *out, + const uint8_t *v ); + +typedef void (*draw_vf_insert_func)( const struct draw_vf_attr *a, + uint8_t *v, + const float *in ); + +typedef void (*draw_vf_emit_func)( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *dest ); + + + +/** + * Describes how to convert/move a vertex attribute from a vertex + * array to a vertex structure. + */ +struct draw_vf_attr +{ + struct draw_vertex_fetch *vf; + + unsigned format; + unsigned inputsize; + unsigned inputstride; + unsigned vertoffset; /**< position of the attrib in the vertex struct */ + + boolean isconst; /**< read from const data below */ + uint8_t data[16]; + + unsigned attrib; /**< which vertex attrib (0=position, etc) */ + unsigned vertattrsize; /**< size of the attribute in bytes */ + + uint8_t *inputptr; + const draw_vf_insert_func *insert; + draw_vf_insert_func do_insert; + draw_vf_extract_func extract; +}; + +struct draw_vertex_fetch +{ + struct draw_vf_attr attr[PIPE_ATTRIB_MAX]; + unsigned attr_count; + unsigned vertex_stride; + + draw_vf_emit_func emit; + + /* Parameters and constants for codegen: + */ + float identity[4]; + + struct draw_vf_fastpath *fastpath; + + void (*codegen_emit)( struct draw_vertex_fetch *vf ); +}; + + +struct draw_vf_attr_type { + unsigned format; + unsigned size; + unsigned stride; + unsigned offset; +}; + +/** XXX this could be moved into draw_vf.c */ +struct draw_vf_fastpath { + unsigned vertex_stride; + unsigned attr_count; + boolean match_strides; + + struct draw_vf_attr_type *attr; + + draw_vf_emit_func func; + struct draw_vf_fastpath *next; +}; + + +void +draw_vf_register_fastpath( struct draw_vertex_fetch *vtx, + boolean match_strides ); + +void +draw_vf_generic_emit( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *v ); + +void +draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ); + +void +draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ); + + +/** XXX this type and function could probably be moved into draw_vf.c */ +struct draw_vf_format_info { + const char *name; + draw_vf_insert_func insert[4]; + const unsigned attrsize; + const boolean isconst; +}; + +extern const struct draw_vf_format_info +draw_vf_format_info[DRAW_EMIT_MAX]; + + +#endif diff --git a/src/gallium/aux/draw/draw_vf_generic.c b/src/gallium/aux/draw/draw_vf_generic.c new file mode 100644 index 0000000000..7a60a9db9c --- /dev/null +++ b/src/gallium/aux/draw/draw_vf_generic.c @@ -0,0 +1,585 @@ + +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_util.h" + +#include "draw_vf.h" + + + +static INLINE void insert_4f_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = in[3]; +} + +static INLINE void insert_4f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = 1; +} + +static INLINE void insert_4f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; + out[3] = 1; +} + +static INLINE void insert_4f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; + out[3] = 1; +} + +static INLINE void insert_3f_xyw_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[3]; +} + +static INLINE void insert_3f_xyw_err( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + (void) a; (void) v; (void) in; + assert(0); +} + +static INLINE void insert_3f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; +} + +static INLINE void insert_3f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; +} + +static INLINE void insert_3f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; +} + + +static INLINE void insert_2f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; +} + +static INLINE void insert_2f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; +} + +static INLINE void insert_1f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; +} + +static INLINE void insert_null( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + (void) a; (void) v; (void) in; +} + +static INLINE void insert_4ub_4f_rgba_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); +} + +static INLINE void insert_4ub_4f_rgba_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_rgba_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[2] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_rgba_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + v[1] = 0; + v[2] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); +} + +static INLINE void insert_4ub_4f_bgra_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[0] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + v[1] = 0; + v[0] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); +} + +static INLINE void insert_4ub_4f_argb_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + v[3] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + v[2] = 0x00; + v[3] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); +} + +static INLINE void insert_4ub_4f_abgr_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + v[1] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + v[2] = 0x00; + v[1] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_3ub_3f_rgb_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); +} + +static INLINE void insert_3ub_3f_rgb_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[2] = 0; +} + +static INLINE void insert_3ub_3f_rgb_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + v[1] = 0; + v[2] = 0; +} + +static INLINE void insert_3ub_3f_bgr_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); +} + +static INLINE void insert_3ub_3f_bgr_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[0] = 0; +} + +static INLINE void insert_3ub_3f_bgr_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + v[1] = 0; + v[0] = 0; +} + + +static INLINE void insert_1ub_1f_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); +} + + +const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX] = +{ + { "1f", + { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, + sizeof(float), FALSE }, + + { "2f", + { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, + 2 * sizeof(float), FALSE }, + + { "3f", + { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, + 3 * sizeof(float), FALSE }, + + { "4f", + { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, + 4 * sizeof(float), FALSE }, + + { "3f_xyw", + { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, + insert_3f_xyw_4 }, + 3 * sizeof(float), FALSE }, + + { "1ub_1f", + { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 }, + sizeof(uint8_t), FALSE }, + + { "3ub_3f_rgb", + { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3, + insert_3ub_3f_rgb_3 }, + 3 * sizeof(uint8_t), FALSE }, + + { "3ub_3f_bgr", + { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3, + insert_3ub_3f_bgr_3 }, + 3 * sizeof(uint8_t), FALSE }, + + { "4ub_4f_rgba", + { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, + insert_4ub_4f_rgba_4 }, + 4 * sizeof(uint8_t), FALSE }, + + { "4ub_4f_bgra", + { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3, + insert_4ub_4f_bgra_4 }, + 4 * sizeof(uint8_t), FALSE }, + + { "4ub_4f_argb", + { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3, + insert_4ub_4f_argb_4 }, + 4 * sizeof(uint8_t), FALSE }, + + { "4ub_4f_abgr", + { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3, + insert_4ub_4f_abgr_4 }, + 4 * sizeof(uint8_t), FALSE }, + + { "1f_const", + { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, + sizeof(float), TRUE }, + + { "2f_const", + { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, + 2 * sizeof(float), TRUE }, + + { "3f_const", + { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, + 3 * sizeof(float), TRUE }, + + { "4f_const", + { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, + 4 * sizeof(float), TRUE }, + + { "pad", + { NULL, NULL, NULL, NULL }, + 0, FALSE }, + +}; + + + + +/*********************************************************************** + * Hardwired fastpaths for emitting whole vertices or groups of + * vertices + */ +#define EMIT5(NR, F0, F1, F2, F3, F4, NAME) \ +static void NAME( struct draw_vertex_fetch *vf, \ + unsigned count, \ + uint8_t *v ) \ +{ \ + struct draw_vf_attr *a = vf->attr; \ + unsigned i; \ + \ + for (i = 0 ; i < count ; i++, v += vf->vertex_stride) { \ + if (NR > 0) { \ + F0( &a[0], v + a[0].vertoffset, (float *)a[0].inputptr ); \ + a[0].inputptr += a[0].inputstride; \ + } \ + \ + if (NR > 1) { \ + F1( &a[1], v + a[1].vertoffset, (float *)a[1].inputptr ); \ + a[1].inputptr += a[1].inputstride; \ + } \ + \ + if (NR > 2) { \ + F2( &a[2], v + a[2].vertoffset, (float *)a[2].inputptr ); \ + a[2].inputptr += a[2].inputstride; \ + } \ + \ + if (NR > 3) { \ + F3( &a[3], v + a[3].vertoffset, (float *)a[3].inputptr ); \ + a[3].inputptr += a[3].inputstride; \ + } \ + \ + if (NR > 4) { \ + F4( &a[4], v + a[4].vertoffset, (float *)a[4].inputptr ); \ + a[4].inputptr += a[4].inputstride; \ + } \ + } \ +} + + +#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \ + insert_null, insert_null, NAME) + +#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \ + insert_null, NAME) + +#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \ + insert_null, NAME) + + +EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4) + +EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2) + +EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2) + + +/* Use the codegen paths to select one of a number of hardwired + * fastpaths. + */ +void draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ) +{ + draw_vf_emit_func func = NULL; + + /* Does it fit a hardwired fastpath? Help! this is growing out of + * control! + */ + switch (vf->attr_count) { + case 2: + if (vf->attr[0].do_insert == insert_3f_3 && + vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + func = emit_xyz3_rgba4; + } + break; + case 3: + if (vf->attr[2].do_insert == insert_2f_2) { + if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + if (vf->attr[0].do_insert == insert_4f_4) + func = emit_xyzw4_rgba4_st2; + } + } + break; + case 4: + if (vf->attr[2].do_insert == insert_2f_2 && + vf->attr[3].do_insert == insert_2f_2) { + if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + if (vf->attr[0].do_insert == insert_4f_4) + func = emit_xyzw4_rgba4_st2_st2; + } + } + break; + } + + vf->emit = func; +} + +/*********************************************************************** + * Generic (non-codegen) functions for whole vertices or groups of + * vertices + */ + +void draw_vf_generic_emit( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *v ) +{ + struct draw_vf_attr *a = vf->attr; + const unsigned attr_count = vf->attr_count; + const unsigned stride = vf->vertex_stride; + unsigned i, j; + + for (i = 0 ; i < count ; i++, v += stride) { + for (j = 0; j < attr_count; j++) { + float *in = (float *)a[j].inputptr; + a[j].inputptr += a[j].inputstride; + a[j].do_insert( &a[j], v + a[j].vertoffset, in ); + } + } +} + + diff --git a/src/gallium/aux/draw/draw_vf_sse.c b/src/gallium/aux/draw/draw_vf_sse.c new file mode 100644 index 0000000000..1ad2ae756d --- /dev/null +++ b/src/gallium/aux/draw/draw_vf_sse.c @@ -0,0 +1,614 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + + +#include "simple_list.h" + +#include "pipe/p_compiler.h" + +#include "draw_vf.h" + + +#if defined(USE_SSE_ASM) + +#include "x86/rtasm/x86sse.h" +#include "x86/common_x86_asm.h" + + +#define X 0 +#define Y 1 +#define Z 2 +#define W 3 + + +struct x86_program { + struct x86_function func; + + struct draw_vertex_fetch *vf; + boolean inputs_safe; + boolean outputs_safe; + boolean have_sse2; + + struct x86_reg identity; + struct x86_reg chan0; +}; + + +static struct x86_reg get_identity( struct x86_program *p ) +{ + return p->identity; +} + +static void emit_load4f_4( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movups(&p->func, dest, arg0); +} + +static void emit_load4f_3( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Have to jump through some hoops: + * + * c 0 0 0 + * c 0 0 1 + * 0 0 c 1 + * a b c 1 + */ + sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); + sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); + sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) ); + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load4f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Initialize from identity, then pull in low two words: + */ + sse_movups(&p->func, dest, get_identity(p)); + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load4f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Pull in low word, then swizzle in identity */ + sse_movss(&p->func, dest, arg0); + sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); +} + + + +static void emit_load3f_3( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Over-reads by 1 dword - potential SEGV if input is a vertex + * array. + */ + if (p->inputs_safe) { + sse_movups(&p->func, dest, arg0); + } + else { + /* c 0 0 0 + * c c c c + * a b c c + */ + sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); + sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X)); + sse_movlps(&p->func, dest, arg0); + } +} + +static void emit_load3f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_2(p, dest, arg0); +} + +static void emit_load3f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_1(p, dest, arg0); +} + +static void emit_load2f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load2f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_1(p, dest, arg0); +} + +static void emit_load1f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movss(&p->func, dest, arg0); +} + +static void (*load[4][4])( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) = { + { emit_load1f_1, + emit_load1f_1, + emit_load1f_1, + emit_load1f_1 }, + + { emit_load2f_1, + emit_load2f_2, + emit_load2f_2, + emit_load2f_2 }, + + { emit_load3f_1, + emit_load3f_2, + emit_load3f_3, + emit_load3f_3 }, + + { emit_load4f_1, + emit_load4f_2, + emit_load4f_3, + emit_load4f_4 } +}; + +static void emit_load( struct x86_program *p, + struct x86_reg dest, + unsigned sz, + struct x86_reg src, + unsigned src_sz) +{ + load[sz-1][src_sz-1](p, dest, src); +} + +static void emit_store4f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movups(&p->func, dest, arg0); +} + +static void emit_store3f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + if (p->outputs_safe) { + /* Emit the extra dword anyway. This may hurt writecombining, + * may cause other problems. + */ + sse_movups(&p->func, dest, arg0); + } + else { + /* Alternate strategy - emit two, shuffle, emit one. + */ + sse_movlps(&p->func, dest, arg0); + sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */ + sse_movss(&p->func, x86_make_disp(dest,8), arg0); + } +} + +static void emit_store2f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movlps(&p->func, dest, arg0); +} + +static void emit_store1f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movss(&p->func, dest, arg0); +} + + +static void (*store[4])( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) = +{ + emit_store1f, + emit_store2f, + emit_store3f, + emit_store4f +}; + +static void emit_store( struct x86_program *p, + struct x86_reg dest, + unsigned sz, + struct x86_reg temp ) + +{ + store[sz-1](p, dest, temp); +} + +static void emit_pack_store_4ub( struct x86_program *p, + struct x86_reg dest, + struct x86_reg temp ) +{ + /* Scale by 255.0 + */ + sse_mulps(&p->func, temp, p->chan0); + + if (p->have_sse2) { + sse2_cvtps2dq(&p->func, temp, temp); + sse2_packssdw(&p->func, temp, temp); + sse2_packuswb(&p->func, temp, temp); + sse_movss(&p->func, dest, temp); + } + else { + struct x86_reg mmx0 = x86_make_reg(file_MMX, 0); + struct x86_reg mmx1 = x86_make_reg(file_MMX, 1); + sse_cvtps2pi(&p->func, mmx0, temp); + sse_movhlps(&p->func, temp, temp); + sse_cvtps2pi(&p->func, mmx1, temp); + mmx_packssdw(&p->func, mmx0, mmx1); + mmx_packuswb(&p->func, mmx0, mmx0); + mmx_movd(&p->func, dest, mmx0); + } +} + +static int get_offset( const void *a, const void *b ) +{ + return (const char *)b - (const char *)a; +} + +/* Not much happens here. Eventually use this function to try and + * avoid saving/reloading the source pointers each vertex (if some of + * them can fit in registers). + */ +static void get_src_ptr( struct x86_program *p, + struct x86_reg srcREG, + struct x86_reg vfREG, + struct draw_vf_attr *a ) +{ + struct draw_vertex_fetch *vf = p->vf; + struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); + + /* Load current a[j].inputptr + */ + x86_mov(&p->func, srcREG, ptr_to_src); +} + +static void update_src_ptr( struct x86_program *p, + struct x86_reg srcREG, + struct x86_reg vfREG, + struct draw_vf_attr *a ) +{ + if (a->inputstride) { + struct draw_vertex_fetch *vf = p->vf; + struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); + + /* add a[j].inputstride (hardcoded value - could just as easily + * pull the stride value from memory each time). + */ + x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride)); + + /* save new value of a[j].inputptr + */ + x86_mov(&p->func, ptr_to_src, srcREG); + } +} + + +/* Lots of hardcoding + * + * EAX -- pointer to current output vertex + * ECX -- pointer to current attribute + * + */ +static boolean build_vertex_emit( struct x86_program *p ) +{ + struct draw_vertex_fetch *vf = p->vf; + unsigned j = 0; + + struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX); + struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX); + struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); + struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI); + struct x86_reg temp = x86_make_reg(file_XMM, 0); + uint8_t *fixup, *label; + + /* Push a few regs? + */ + x86_push(&p->func, countEBP); + x86_push(&p->func, vfESI); + + + /* Get vertex count, compare to zero + */ + x86_xor(&p->func, srcECX, srcECX); + x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2)); + x86_cmp(&p->func, countEBP, srcECX); + fixup = x86_jcc_forward(&p->func, cc_E); + + /* Initialize destination register. + */ + x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3)); + + /* Move argument 1 (vf) into a reg: + */ + x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1)); + + + /* always load, needed or not: + */ + sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0]))); + + /* Note address for loop jump */ + label = x86_get_label(&p->func); + + /* Emit code for each of the attributes. Currently routes + * everything through SSE registers, even when it might be more + * efficient to stick with regular old x86. No optimization or + * other tricks - enough new ground to cover here just getting + * things working. + */ + while (j < vf->attr_count) { + struct draw_vf_attr *a = &vf->attr[j]; + struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset); + + /* Now, load an XMM reg from src, perhaps transform, then save. + * Could be shortcircuited in specific cases: + */ + switch (a->format) { + case DRAW_EMIT_1F: + case DRAW_EMIT_1F_CONST: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 1, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_2F: + case DRAW_EMIT_2F_CONST: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 2, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_3F: + case DRAW_EMIT_3F_CONST: + /* Potentially the worst case - hardcode 2+1 copying: + */ + if (0) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + } + else { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 2, temp); + if (a->inputsize > 2) { + emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1); + emit_store(p, x86_make_disp(dest,8), 1, temp); + } + else { + sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p)); + } + update_src_ptr(p, srcECX, vfESI, a); + } + break; + case DRAW_EMIT_4F: + case DRAW_EMIT_4F_CONST: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 4, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_3F_XYW: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z)); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + + case DRAW_EMIT_1UB_1F: + /* Test for PAD3 + 1UB: + */ + if (j > 0 && + a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3) + { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X)); + emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */ + update_src_ptr(p, srcECX, vfESI, a); + } + else { + debug_printf("Can't emit 1ub %x %x %d\n", + a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); + return FALSE; + } + break; + case DRAW_EMIT_3UB_3F_RGB: + case DRAW_EMIT_3UB_3F_BGR: + /* Test for 3UB + PAD1: + */ + if (j == vf->attr_count - 1 || + a[1].vertoffset >= a->vertoffset + 4) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + if (a->format == DRAW_EMIT_3UB_3F_BGR) + sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + } + /* Test for 3UB + 1UB: + */ + else if (j < vf->attr_count - 1 && + a[1].format == DRAW_EMIT_1UB_1F && + a[1].vertoffset == a->vertoffset + 3) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + update_src_ptr(p, srcECX, vfESI, a); + + /* Make room for incoming value: + */ + sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); + + get_src_ptr(p, srcECX, vfESI, &a[1]); + emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize); + update_src_ptr(p, srcECX, vfESI, &a[1]); + + /* Rearrange and possibly do BGR conversion: + */ + if (a->format == DRAW_EMIT_3UB_3F_BGR) + sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); + else + sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X)); + + emit_pack_store_4ub(p, dest, temp); + j++; /* NOTE: two attrs consumed */ + } + else { + debug_printf("Can't emit 3ub\n"); + } + return FALSE; /* add this later */ + break; + + case DRAW_EMIT_4UB_4F_RGBA: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_4UB_4F_BGRA: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_4UB_4F_ARGB: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_4UB_4F_ABGR: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + default: + debug_printf("unknown a[%d].format %d\n", j, a->format); + return FALSE; /* catch any new opcodes */ + } + + /* Increment j by at least 1 - may have been incremented above also: + */ + j++; + } + + /* Next vertex: + */ + x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vf->vertex_stride)); + + /* decr count, loop if not zero + */ + x86_dec(&p->func, countEBP); + x86_test(&p->func, countEBP, countEBP); + x86_jcc(&p->func, cc_NZ, label); + + /* Exit mmx state? + */ + if (p->func.need_emms) + mmx_emms(&p->func); + + /* Land forward jump here: + */ + x86_fixup_fwd_jump(&p->func, fixup); + + /* Pop regs and return + */ + x86_pop(&p->func, x86_get_base_reg(vfESI)); + x86_pop(&p->func, countEBP); + x86_ret(&p->func); + + vf->emit = (draw_vf_emit_func)x86_get_func(&p->func); + return TRUE; +} + + + +void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) +{ + struct x86_program p; + + if (!cpu_has_xmm) { + vf->codegen_emit = NULL; + return; + } + + memset(&p, 0, sizeof(p)); + + p.vf = vf; + p.inputs_safe = 0; /* for now */ + p.outputs_safe = 1; /* for now */ + p.have_sse2 = cpu_has_xmm2; + p.identity = x86_make_reg(file_XMM, 6); + p.chan0 = x86_make_reg(file_XMM, 7); + + x86_init_func(&p.func); + + if (build_vertex_emit(&p)) { + draw_vf_register_fastpath( vf, TRUE ); + } + else { + /* Note the failure so that we don't keep trying to codegen an + * impossible state: + */ + draw_vf_register_fastpath( vf, FALSE ); + x86_release_func(&p.func); + } +} + +#else + +void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) +{ + /* Dummy version for when USE_SSE_ASM not defined */ +} + +#endif diff --git a/src/gallium/aux/draw/draw_wide_prims.c b/src/gallium/aux/draw/draw_wide_prims.c new file mode 100644 index 0000000000..655774b155 --- /dev/null +++ b/src/gallium/aux/draw/draw_wide_prims.c @@ -0,0 +1,432 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +struct wide_stage { + struct draw_stage stage; + + float half_line_width; + float half_point_size; + + uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; + uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; + uint num_texcoords; + + int psize_slot; +}; + + + +static INLINE struct wide_stage *wide_stage( struct draw_stage *stage ) +{ + return (struct wide_stage *)stage; +} + + +static void passthrough_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void passthrough_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line(stage->next, header); +} + +static void passthrough_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Draw a wide line by drawing a quad (two triangles). + * XXX need to disable polygon stipple. + */ +static void wide_line( struct draw_stage *stage, + struct prim_header *header ) +{ + const struct wide_stage *wide = wide_stage(stage); + const float half_width = wide->half_line_width; + + struct prim_header tri; + + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + const float dx = FABSF(pos0[0] - pos2[0]); + const float dy = FABSF(pos0[1] - pos2[1]); + + /* + * Draw wide line as a quad (two tris) by "stretching" the line along + * X or Y. + * We need to tweak coords in several ways to be conformant here. + */ + + if (dx > dy) { + /* x-major line */ + pos0[1] = pos0[1] - half_width - 0.25f; + pos1[1] = pos1[1] + half_width - 0.25f; + pos2[1] = pos2[1] - half_width - 0.25f; + pos3[1] = pos3[1] + half_width - 0.25f; + if (pos0[0] < pos2[0]) { + /* left to right line */ + pos0[0] -= 0.5f; + pos1[0] -= 0.5f; + pos2[0] -= 0.5f; + pos3[0] -= 0.5f; + } + else { + /* right to left line */ + pos0[0] += 0.5f; + pos1[0] += 0.5f; + pos2[0] += 0.5f; + pos3[0] += 0.5f; + } + } + else { + /* y-major line */ + pos0[0] = pos0[0] - half_width + 0.25f; + pos1[0] = pos1[0] + half_width + 0.25f; + pos2[0] = pos2[0] - half_width + 0.25f; + pos3[0] = pos3[0] + half_width + 0.25f; + if (pos0[1] < pos2[1]) { + /* top to bottom line */ + pos0[1] -= 0.5f; + pos1[1] -= 0.5f; + pos2[1] -= 0.5f; + pos3[1] -= 0.5f; + } + else { + /* bottom to top line */ + pos0[1] += 0.5f; + pos1[1] += 0.5f; + pos2[1] += 0.5f; + pos3[1] += 0.5f; + } + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +/** + * Draw a wide line by drawing a quad, using geometry which will + * fullfill GL's antialiased line requirements. + */ +static void wide_line_aa(struct draw_stage *stage, + struct prim_header *header) +{ + const struct wide_stage *wide = wide_stage(stage); + const float half_width = wide->half_line_width; + struct prim_header tri; + struct vertex_header *v[4]; + float *pos; + float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; + float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; + const float len = (float) sqrt(dx * dx + dy * dy); + uint i; + + dx = dx * half_width / len; + dy = dy * half_width / len; + + /* allocate/dup new verts */ + for (i = 0; i < 4; i++) { + v[i] = dup_vert(stage, header->v[i/2], i); + } + + /* + * Quad for line from v0 to v1: + * + * 1 3 + * +-------------------------+ + * | | + * *v0 v1* + * | | + * +-------------------------+ + * 0 2 + */ + + pos = v[0]->data[0]; + pos[0] += dy; + pos[1] -= dx; + + pos = v[1]->data[0]; + pos[0] -= dy; + pos[1] += dx; + + pos = v[2]->data[0]; + pos[0] += dy; + pos[1] -= dx; + + pos = v[3]->data[0]; + pos[0] -= dy; + pos[1] += dx; + + tri.det = header->det; /* only the sign matters */ + + tri.v[0] = v[2]; tri.v[1] = v[1]; tri.v[2] = v[0]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[3]; tri.v[1] = v[1]; tri.v[2] = v[2]; + stage->next->tri( stage->next, &tri ); + +} + + +/** + * Set the vertex texcoords for sprite mode. + * Coords may be left untouched or set to a right-side-up or upside-down + * orientation. + */ +static void set_texcoords(const struct wide_stage *wide, + struct vertex_header *v, const float tc[4]) +{ + uint i; + for (i = 0; i < wide->num_texcoords; i++) { + if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { + uint j = wide->texcoord_slot[i]; + v->data[j][0] = tc[0]; + if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) + v->data[j][1] = 1.0f - tc[1]; + else + v->data[j][1] = tc[1]; + v->data[j][2] = tc[2]; + v->data[j][3] = tc[3]; + } + } +} + + +/* If there are lots of sprite points (and why wouldn't there be?) it + * would probably be more sensible to change hardware setup to + * optimize this rather than doing the whole thing in software like + * this. + */ +static void wide_point( struct draw_stage *stage, + struct prim_header *header ) +{ + const struct wide_stage *wide = wide_stage(stage); + const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; + float half_size; + float left_adj, right_adj; + + struct prim_header tri; + + /* four dups of original vertex */ + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + /* point size is either per-vertex or fixed size */ + if (wide->psize_slot >= 0) { + half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; + } + else { + half_size = wide->half_point_size; + } + + left_adj = -half_size + 0.25f; + right_adj = half_size + 0.25f; + + pos0[0] += left_adj; + pos0[1] -= half_size; + + pos1[0] += left_adj; + pos1[1] += half_size; + + pos2[0] += right_adj; + pos2[1] -= half_size; + + pos3[0] += right_adj; + pos3[1] += half_size; + + if (sprite) { + static const float tex00[4] = { 0, 0, 0, 1 }; + static const float tex01[4] = { 0, 1, 0, 1 }; + static const float tex11[4] = { 1, 1, 0, 1 }; + static const float tex10[4] = { 1, 0, 0, 1 }; + set_texcoords( wide, v0, tex00 ); + set_texcoords( wide, v1, tex01 ); + set_texcoords( wide, v2, tex10 ); + set_texcoords( wide, v3, tex11 ); + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +static void wide_first_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct wide_stage *wide = wide_stage(stage); + struct draw_context *draw = stage->draw; + + wide->half_point_size = 0.5f * draw->rasterizer->point_size; + + if (draw->rasterizer->point_size != 1.0) { + stage->point = wide_point; + } + else { + stage->point = passthrough_point; + } + + if (draw->rasterizer->point_sprite) { + /* find vertex shader texcoord outputs */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i, j = 0; + for (i = 0; i < vs->state->num_outputs; i++) { + if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { + wide->texcoord_slot[j] = i; + wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; + j++; + } + } + wide->num_texcoords = j; + } + + wide->psize_slot = -1; + + if (draw->rasterizer->point_size_per_vertex) { + /* find PSIZ vertex output */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i; + for (i = 0; i < vs->state->num_outputs; i++) { + if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + wide->psize_slot = i; + break; + } + } + } + + stage->point( stage, header ); +} + + + +static void wide_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct wide_stage *wide = wide_stage(stage); + struct draw_context *draw = stage->draw; + + wide->half_line_width = 0.5f * draw->rasterizer->line_width; + + if (draw->rasterizer->line_width != 1.0) { + if (draw->rasterizer->line_smooth) + wide->stage.line = wide_line_aa; + else + wide->stage.line = wide_line; + } + else { + wide->stage.line = passthrough_line; + } + + stage->line( stage, header ); +} + + +static void wide_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->line = wide_first_line; + stage->point = wide_first_point; + stage->next->flush( stage->next, flags ); +} + + +static void wide_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void wide_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +struct draw_stage *draw_wide_stage( struct draw_context *draw ) +{ + struct wide_stage *wide = CALLOC_STRUCT(wide_stage); + + draw_alloc_temp_verts( &wide->stage, 4 ); + + wide->stage.draw = draw; + wide->stage.next = NULL; + wide->stage.point = wide_first_point; + wide->stage.line = wide_first_line; + wide->stage.tri = passthrough_tri; + wide->stage.flush = wide_flush; + wide->stage.reset_stipple_counter = wide_reset_stipple_counter; + wide->stage.destroy = wide_destroy; + + return &wide->stage; +} diff --git a/src/gallium/aux/llvm/Makefile b/src/gallium/aux/llvm/Makefile new file mode 100644 index 0000000000..9c6e16d86b --- /dev/null +++ b/src/gallium/aux/llvm/Makefile @@ -0,0 +1,83 @@ +# -*-makefile-*- +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = gallivm + + +GALLIVM_SOURCES = \ + gallivm.cpp \ + gallivm_cpu.cpp \ + instructions.cpp \ + loweringpass.cpp \ + tgsitollvm.cpp \ + storage.cpp \ + storagesoa.cpp \ + instructionssoa.cpp + +INC_SOURCES = gallivm_builtins.cpp + +CPP_SOURCES = \ + $(GALLIVM_SOURCES) + +C_SOURCES = +ASM_SOURCES = + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(CPP_SOURCES:.cpp=.o) \ + $(ASM_SOURCES:.S=.o) + +### Include directories +INCLUDES = \ + -I. \ + -I$(TOP)/src/mesa/pipe \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDES) $(LLVM_CXXFLAGS) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +##### TARGETS ##### + +default:: depend symlinks $(LIBNAME) + + +$(LIBNAME): $(OBJECTS) Makefile + $(TOP)/bin/mklib -o $@ -static $(OBJECTS) + + +depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ + $(ASM_SOURCES) $(INC_SOURCES) 2> /dev/null + + +gallivm_builtins.cpp: llvm_builtins.c + clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean: + -rm -f *.o */*.o *~ *.so *~ server/*.o + -rm -f depend depend.bak + -rm -f gallivm_builtins.cpp + +symlinks: + + +include depend diff --git a/src/gallium/aux/llvm/gallivm.cpp b/src/gallium/aux/llvm/gallivm.cpp new file mode 100644 index 0000000000..da0105c2c9 --- /dev/null +++ b/src/gallium/aux/llvm/gallivm.cpp @@ -0,0 +1,327 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "instructions.h" +#include "loweringpass.h" +#include "storage.h" +#include "tgsitollvm.h" + +#include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" + +#include "pipe/tgsi/exec/tgsi_exec.h" +#include "pipe/tgsi/util/tgsi_dump.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +static int GLOBAL_ID = 0; + +using namespace llvm; + +static inline +void AddStandardCompilePasses(PassManager &PM) +{ + PM.add(new LoweringPass()); + PM.add(createVerifierPass()); // Verify that input is correct + + PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp + + //PM.add(createStripSymbolsPass(true)); + + PM.add(createRaiseAllocationsPass()); // call %malloc -> malloc inst + PM.add(createCFGSimplificationPass()); // Clean up disgusting code + PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas + PM.add(createGlobalOptimizerPass()); // Optimize out global vars + PM.add(createGlobalDCEPass()); // Remove unused fns and globs + PM.add(createIPConstantPropagationPass());// IP Constant Propagation + PM.add(createDeadArgEliminationPass()); // Dead argument elimination + PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE + PM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE + + PM.add(createPruneEHPass()); // Remove dead EH info + + PM.add(createFunctionInliningPass()); // Inline small functions + PM.add(createArgumentPromotionPass()); // Scalarize uninlined fn args + + PM.add(createTailDuplicationPass()); // Simplify cfg by copying code + PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl. + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas + PM.add(createInstructionCombiningPass()); // Combine silly seq's + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createTailCallEliminationPass()); // Eliminate tail calls + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createReassociatePass()); // Reassociate expressions + PM.add(createLoopRotatePass()); + PM.add(createLICMPass()); // Hoist loop invariants + PM.add(createLoopUnswitchPass()); // Unswitch loops. + PM.add(createLoopIndexSplitPass()); // Index split loops. + PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc + PM.add(createIndVarSimplifyPass()); // Canonicalize indvars + PM.add(createLoopUnrollPass()); // Unroll small loops + PM.add(createInstructionCombiningPass()); // Clean up after the unroller + PM.add(createGVNPass()); // Remove redundancies + PM.add(createSCCPPass()); // Constant prop with SCCP + + // Run instcombine after redundancy elimination to exploit opportunities + // opened up by them. + PM.add(createInstructionCombiningPass()); + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createDeadStoreEliminationPass()); // Delete dead stores + PM.add(createAggressiveDCEPass()); // SSA based 'Aggressive DCE' + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createSimplifyLibCallsPass()); // Library Call Optimizations + PM.add(createDeadTypeEliminationPass()); // Eliminate dead types + PM.add(createConstantMergePass()); // Merge dup global constants +} + +void gallivm_prog_delete(struct gallivm_prog *prog) +{ + delete prog->module; + prog->module = 0; + prog->function = 0; + free(prog); +} + +static inline void +constant_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan) +{ + unsigned i; + + for (i = 0; i < QUAD_SIZE; ++i) { + inputs[i][attrib][chan] = coefs[attrib].a0[chan]; + } +} + +static inline void +linear_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = inputs[i][0][0]; + const float y = inputs[i][0][1]; + + inputs[i][attrib][chan] = + coefs[attrib].a0[chan] + + coefs[attrib].dadx[chan] * x + + coefs[attrib].dady[chan] * y; + } +} + +static inline void +perspective_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = inputs[i][0][0]; + const float y = inputs[i][0][1]; + /* WPOS.w here is really 1/w */ + const float w = 1.0f / inputs[i][0][3]; + assert(inputs[i][0][3] != 0.0); + + inputs[i][attrib][chan] = + (coefs[attrib].a0[chan] + + coefs[attrib].dadx[chan] * x + + coefs[attrib].dady[chan] * y) * w; + } +} + +void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix) +{ + if (!ir || !ir->module) + return; + + if (file_prefix) { + std::ostringstream stream; + stream << file_prefix; + stream << ir->id; + stream << ".ll"; + std::string name = stream.str(); + std::ofstream out(name.c_str()); + if (!out) { + std::cerr<<"Can't open file : "<module); + out.close(); + } else { + const llvm::Module::FunctionListType &funcs = ir->module->getFunctionList(); + llvm::Module::FunctionListType::const_iterator itr; + std::cout<<"; ---------- Start shader "<id<id<num_interp; ++i) { + const gallivm_interpolate &interp = prog->interpolators[i]; + switch (interp.type) { + case TGSI_INTERPOLATE_CONSTANT: + constant_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + case TGSI_INTERPOLATE_LINEAR: + linear_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + perspective_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + default: + assert( 0 ); + } + } +} + + +struct gallivm_ir * gallivm_ir_new(enum gallivm_shader_type type) +{ + struct gallivm_ir *ir = + (struct gallivm_ir *)calloc(1, sizeof(struct gallivm_ir)); + ++GLOBAL_ID; + ir->id = GLOBAL_ID; + ir->type = type; + + return ir; +} + +void gallivm_ir_set_layout(struct gallivm_ir *ir, + enum gallivm_vector_layout layout) +{ + ir->layout = layout; +} + +void gallivm_ir_set_components(struct gallivm_ir *ir, int num) +{ + ir->num_components = num; +} + +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, + const struct tgsi_token *tokens) +{ + std::cout << "Creating llvm from: " <module = mod; + gallivm_ir_dump(ir, 0); +} + +void gallivm_ir_delete(struct gallivm_ir *ir) +{ + delete ir->module; + free(ir); +} + +struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) +{ + struct gallivm_prog *prog = + (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); + llvm::Module *mod = llvm::CloneModule(ir->module); + prog->num_consts = ir->num_consts; + memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators)); + prog->num_interp = ir->num_interp; + + /* Run optimization passes over it */ + PassManager passes; + passes.add(new TargetData(mod)); + AddStandardCompilePasses(passes); + passes.run(*mod); + prog->module = mod; + + std::cout << "After optimizations:"<dump(); + + return prog; +} + +#endif /* MESA_LLVM */ diff --git a/src/gallium/aux/llvm/gallivm.h b/src/gallium/aux/llvm/gallivm.h new file mode 100644 index 0000000000..92da4bca7f --- /dev/null +++ b/src/gallium/aux/llvm/gallivm.h @@ -0,0 +1,103 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef GALLIVM_H +#define GALLIVM_H + +#if defined __cplusplus +extern "C" { +#endif + +#include "pipe/p_state.h" + +#ifdef MESA_LLVM + +struct tgsi_token; + +struct gallivm_ir; +struct gallivm_prog; +struct gallivm_cpu_engine; +struct tgsi_interp_coef; +struct tgsi_sampler; +struct tgsi_exec_vector; + +enum gallivm_shader_type { + GALLIVM_VS, + GALLIVM_FS +}; + +enum gallivm_vector_layout { + GALLIVM_AOS, + GALLIVM_SOA +}; + +struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type); +void gallivm_ir_set_layout(struct gallivm_ir *ir, + enum gallivm_vector_layout layout); +void gallivm_ir_set_components(struct gallivm_ir *ir, int num); +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, + const struct tgsi_token *tokens); +void gallivm_ir_delete(struct gallivm_ir *ir); + + +struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir); + +void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + const struct tgsi_interp_coef *coefs); +void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); + + +struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); +struct gallivm_cpu_engine *gallivm_global_cpu_engine(); +int gallivm_cpu_vs_exec(struct gallivm_prog *prog, + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, + float (*consts)[4], + struct tgsi_exec_vector *temps); +int gallivm_cpu_fs_exec(struct gallivm_prog *prog, + float x, float y, + float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + float (*consts)[4], + struct tgsi_sampler *samplers); +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); + + +#endif /* MESA_LLVM */ + +#if defined __cplusplus +} // extern "C" +#endif + +#endif diff --git a/src/gallium/aux/llvm/gallivm_builtins.cpp b/src/gallium/aux/llvm/gallivm_builtins.cpp new file mode 100644 index 0000000000..1796f0a177 --- /dev/null +++ b/src/gallium/aux/llvm/gallivm_builtins.cpp @@ -0,0 +1,567 @@ +// Generated by llvm2cpp - DO NOT MODIFY! + + +Module* createGallivmBuiltins(Module *mod) { + +mod->setModuleIdentifier("shader"); + +// Type Definitions +ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 25); + +PointerType* PointerTy_1 = PointerType::get(ArrayTy_0, 0); + +std::vectorFuncTy_2_args; +FuncTy_2_args.push_back(Type::FloatTy); +FuncTy_2_args.push_back(Type::FloatTy); +FunctionType* FuncTy_2 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_2_args, + /*isVarArg=*/false); + +PointerType* PointerTy_3 = PointerType::get(FuncTy_2, 0); + +VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); + +std::vectorFuncTy_5_args; +FuncTy_5_args.push_back(VectorTy_4); +FunctionType* FuncTy_5 = FunctionType::get( + /*Result=*/VectorTy_4, + /*Params=*/FuncTy_5_args, + /*isVarArg=*/false); + +std::vectorFuncTy_6_args; +FuncTy_6_args.push_back(VectorTy_4); +FuncTy_6_args.push_back(VectorTy_4); +FuncTy_6_args.push_back(VectorTy_4); +FunctionType* FuncTy_6 = FunctionType::get( + /*Result=*/VectorTy_4, + /*Params=*/FuncTy_6_args, + /*isVarArg=*/false); + +VectorType* VectorTy_7 = VectorType::get(IntegerType::get(32), 4); + +std::vectorFuncTy_9_args; +FunctionType* FuncTy_9 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_9_args, + /*isVarArg=*/true); + +PointerType* PointerTy_8 = PointerType::get(FuncTy_9, 0); + +PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8), 0); + +std::vectorFuncTy_12_args; +FuncTy_12_args.push_back(Type::FloatTy); +FunctionType* FuncTy_12 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_12_args, + /*isVarArg=*/false); + +PointerType* PointerTy_11 = PointerType::get(FuncTy_12, 0); + +std::vectorFuncTy_13_args; +FuncTy_13_args.push_back(VectorTy_4); +FunctionType* FuncTy_13 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_13_args, + /*isVarArg=*/false); + + +// Function Declarations + +Function* func_approx = new Function( + /*Type=*/FuncTy_2, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"approx", mod); +func_approx->setCallingConv(CallingConv::C); +const ParamAttrsList *func_approx_PAL = 0; +func_approx->setParamAttrs(func_approx_PAL); + +Function* func_powf = new Function( + /*Type=*/FuncTy_2, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"powf", mod); // (external, no body) +func_powf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_powf_PAL = 0; +func_powf->setParamAttrs(func_powf_PAL); + +Function* func_lit = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"lit", mod); +func_lit->setCallingConv(CallingConv::C); +const ParamAttrsList *func_lit_PAL = 0; +func_lit->setParamAttrs(func_lit_PAL); + +Function* func_cmp = new Function( + /*Type=*/FuncTy_6, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"cmp", mod); +func_cmp->setCallingConv(CallingConv::C); +const ParamAttrsList *func_cmp_PAL = 0; +{ + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_cmp_PAL = ParamAttrsList::get(Attrs); + +} +func_cmp->setParamAttrs(func_cmp_PAL); + +Function* func_vcos = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"vcos", mod); +func_vcos->setCallingConv(CallingConv::C); +const ParamAttrsList *func_vcos_PAL = 0; +func_vcos->setParamAttrs(func_vcos_PAL); + +Function* func_printf = new Function( + /*Type=*/FuncTy_9, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", mod); // (external, no body) +func_printf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_printf_PAL = 0; +func_printf->setParamAttrs(func_printf_PAL); + +Function* func_cosf = new Function( + /*Type=*/FuncTy_12, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"cosf", mod); // (external, no body) +func_cosf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_cosf_PAL = 0; +func_cosf->setParamAttrs(func_cosf_PAL); + +Function* func_scs = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"scs", mod); +func_scs->setCallingConv(CallingConv::C); +const ParamAttrsList *func_scs_PAL = 0; +func_scs->setParamAttrs(func_scs_PAL); + +Function* func_sinf = new Function( + /*Type=*/FuncTy_12, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"sinf", mod); // (external, no body) +func_sinf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_sinf_PAL = 0; +func_sinf->setParamAttrs(func_sinf_PAL); + +Function* func_vsin = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"vsin", mod); +func_vsin->setCallingConv(CallingConv::C); +const ParamAttrsList *func_vsin_PAL = 0; +func_vsin->setParamAttrs(func_vsin_PAL); + +Function* func_kilp = new Function( + /*Type=*/FuncTy_13, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"kilp", mod); +func_kilp->setCallingConv(CallingConv::C); +const ParamAttrsList *func_kilp_PAL = 0; +{ + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_kilp_PAL = ParamAttrsList::get(Attrs); + +} +func_kilp->setParamAttrs(func_kilp_PAL); + +// Global Variable Declarations + + +GlobalVariable* gvar_array__str = new GlobalVariable( +/*Type=*/ArrayTy_0, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str", +mod); + +GlobalVariable* gvar_array__str1 = new GlobalVariable( +/*Type=*/ArrayTy_0, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str1", +mod); + +// Constant Definitions +Constant* const_array_14 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); +Constant* const_array_15 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); +ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); +ConstantFP* const_float_17 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); +Constant* const_float_18 = Constant::getNullValue(Type::FloatTy); +Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); +std::vector const_packed_20_elems; +ConstantFP* const_float_21 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); +const_packed_20_elems.push_back(const_float_21); +UndefValue* const_float_22 = UndefValue::get(Type::FloatTy); +const_packed_20_elems.push_back(const_float_22); +const_packed_20_elems.push_back(const_float_22); +const_packed_20_elems.push_back(const_float_21); +Constant* const_packed_20 = ConstantVector::get(VectorTy_4, const_packed_20_elems); +ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); +ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "3", 10)); +ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "2", 10)); +std::vector const_packed_26_elems; +const_packed_26_elems.push_back(const_float_21); +const_packed_26_elems.push_back(const_float_18); +const_packed_26_elems.push_back(const_float_18); +const_packed_26_elems.push_back(const_float_21); +Constant* const_packed_26 = ConstantVector::get(VectorTy_4, const_packed_26_elems); +Constant* const_double_27 = Constant::getNullValue(Type::DoubleTy); +std::vector const_packed_28_elems; +const_packed_28_elems.push_back(const_int32_19); +ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "5", 10)); +const_packed_28_elems.push_back(const_int32_29); +const_packed_28_elems.push_back(const_int32_25); +const_packed_28_elems.push_back(const_int32_24); +Constant* const_packed_28 = ConstantVector::get(VectorTy_7, const_packed_28_elems); +std::vector const_packed_30_elems; +const_packed_30_elems.push_back(const_int32_19); +const_packed_30_elems.push_back(const_int32_23); +ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "6", 10)); +const_packed_30_elems.push_back(const_int32_31); +const_packed_30_elems.push_back(const_int32_24); +Constant* const_packed_30 = ConstantVector::get(VectorTy_7, const_packed_30_elems); +std::vector const_packed_32_elems; +const_packed_32_elems.push_back(const_int32_19); +const_packed_32_elems.push_back(const_int32_23); +const_packed_32_elems.push_back(const_int32_25); +ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "7", 10)); +const_packed_32_elems.push_back(const_int32_33); +Constant* const_packed_32 = ConstantVector::get(VectorTy_7, const_packed_32_elems); +std::vector const_ptr_34_indices; +const_ptr_34_indices.push_back(const_int32_19); +const_ptr_34_indices.push_back(const_int32_19); +Constant* const_ptr_34 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_34_indices[0], const_ptr_34_indices.size() ); +UndefValue* const_packed_35 = UndefValue::get(VectorTy_4); +std::vector const_ptr_36_indices; +const_ptr_36_indices.push_back(const_int32_19); +const_ptr_36_indices.push_back(const_int32_19); +Constant* const_ptr_36 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_36_indices[0], const_ptr_36_indices.size() ); + +// Global Variable Definitions +gvar_array__str->setInitializer(const_array_14); +gvar_array__str1->setInitializer(const_array_15); + +// Function Definitions + +// Function: approx (func_approx) +{ + Function::arg_iterator args = func_approx->arg_begin(); + Value* float_a = args++; + float_a->setName("a"); + Value* float_b = args++; + float_b->setName("b"); + + BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); + + // Block entry (label_entry) + FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_16, "cmp", label_entry); + SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_16, float_b, "b.addr.0", label_entry); + FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_17, "cmp3", label_entry); + SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_17, float_b_addr_0, "b.addr.1", label_entry); + FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_18, "cmp7", label_entry); + SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_18, float_a, "a.addr.0", label_entry); + std::vector float_call_params; + float_call_params.push_back(float_a_addr_0); + float_call_params.push_back(float_b_addr_1); + CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); + float_call->setCallingConv(CallingConv::C); + float_call->setTailCall(true);const ParamAttrsList *float_call_PAL = 0; + float_call->setParamAttrs(float_call_PAL); + + new ReturnInst(float_call, label_entry); + +} + +// Function: lit (func_lit) +{ + Function::arg_iterator args = func_lit->arg_begin(); + Value* packed_tmp = args++; + packed_tmp->setName("tmp"); + + BasicBlock* label_entry_38 = new BasicBlock("entry",func_lit,0); + BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); + BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); + + // Block entry (label_entry_38) + ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp6", label_entry_38); + FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp6, const_float_18, "cmp", label_entry_38); + new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_39, label_entry_38); + + // Block ifthen (label_ifthen) + InsertElementInst* packed_tmp10 = new InsertElementInst(const_packed_20, float_tmp6, const_int32_23, "tmp10", label_ifthen); + ExtractElementInst* float_tmp12 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp14", label_ifthen); + std::vector float_call_41_params; + float_call_41_params.push_back(float_tmp12); + float_call_41_params.push_back(float_tmp14); + CallInst* float_call_41 = new CallInst(func_approx, float_call_41_params.begin(), float_call_41_params.end(), "call", label_ifthen); + float_call_41->setCallingConv(CallingConv::C); + float_call_41->setTailCall(true);const ParamAttrsList *float_call_41_PAL = 0; + float_call_41->setParamAttrs(float_call_41_PAL); + + InsertElementInst* packed_tmp16 = new InsertElementInst(packed_tmp10, float_call_41, const_int32_25, "tmp16", label_ifthen); + new ReturnInst(packed_tmp16, label_ifthen); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock) + new ReturnInst(const_packed_26, label_UnifiedReturnBlock); + +} + +// Function: cmp (func_cmp) +{ + Function::arg_iterator args = func_cmp->arg_begin(); + Value* packed_tmp0 = args++; + packed_tmp0->setName("tmp0"); + Value* packed_tmp1 = args++; + packed_tmp1->setName("tmp1"); + Value* packed_tmp2 = args++; + packed_tmp2->setName("tmp2"); + + BasicBlock* label_entry_44 = new BasicBlock("entry",func_cmp,0); + BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); + BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); + BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); + BasicBlock* label_cond_cont34 = new BasicBlock("cond.cont34",func_cmp,0); + BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); + BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); + + // Block entry (label_entry_44) + ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_44); + CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_44); + FCmpInst* int1_cmp_45 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_44); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_44); + CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_44); + FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_44); + SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_45, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_44); + new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_44); + + // Block cond.?14 (label_cond__14) + ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_28, "tmp233", label_cond__14); + ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp254", label_cond__14); + CastInst* double_conv265 = new FPExtInst(float_tmp254, Type::DoubleTy, "conv265", label_cond__14); + FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_27, "cmp276", label_cond__14); + new BranchInst(label_cond__28, label_cond_cont34, int1_cmp276, label_cond__14); + + // Block cond.cont20 (label_cond_cont20) + ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_28, "tmp23", label_cond_cont20); + ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp25", label_cond_cont20); + CastInst* double_conv26 = new FPExtInst(float_tmp25, Type::DoubleTy, "conv26", label_cond_cont20); + FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_27, "cmp27", label_cond_cont20); + new BranchInst(label_cond__28, label_cond_cont34, int1_cmp27, label_cond_cont20); + + // Block cond.?28 (label_cond__28) + PHINode* packed_tmp23_reg2mem_0 = new PHINode(VectorTy_4, "tmp23.reg2mem.0", label_cond__28); + packed_tmp23_reg2mem_0->reserveOperandSpace(2); + packed_tmp23_reg2mem_0->addIncoming(packed_tmp233, label_cond__14); + packed_tmp23_reg2mem_0->addIncoming(packed_tmp23, label_cond_cont20); + + ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_30, "tmp378", label_cond__28); + ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp399", label_cond__28); + CastInst* double_conv4010 = new FPExtInst(float_tmp399, Type::DoubleTy, "conv4010", label_cond__28); + FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_27, "cmp4111", label_cond__28); + new BranchInst(label_cond__42, label_cond_cont48, int1_cmp4111, label_cond__28); + + // Block cond.cont34 (label_cond_cont34) + PHINode* packed_tmp23_reg2mem_1 = new PHINode(VectorTy_4, "tmp23.reg2mem.1", label_cond_cont34); + packed_tmp23_reg2mem_1->reserveOperandSpace(2); + packed_tmp23_reg2mem_1->addIncoming(packed_tmp233, label_cond__14); + packed_tmp23_reg2mem_1->addIncoming(packed_tmp23, label_cond_cont20); + + ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_30, "tmp37", label_cond_cont34); + ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp39", label_cond_cont34); + CastInst* double_conv40 = new FPExtInst(float_tmp39, Type::DoubleTy, "conv40", label_cond_cont34); + FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_27, "cmp41", label_cond_cont34); + new BranchInst(label_cond__42, label_cond_cont48, int1_cmp41, label_cond_cont34); + + // Block cond.?42 (label_cond__42) + PHINode* packed_tmp37_reg2mem_0 = new PHINode(VectorTy_4, "tmp37.reg2mem.0", label_cond__42); + packed_tmp37_reg2mem_0->reserveOperandSpace(2); + packed_tmp37_reg2mem_0->addIncoming(packed_tmp378, label_cond__28); + packed_tmp37_reg2mem_0->addIncoming(packed_tmp37, label_cond_cont34); + + ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_32, "tmp5113", label_cond__42); + new ReturnInst(packed_tmp5113, label_cond__42); + + // Block cond.cont48 (label_cond_cont48) + PHINode* packed_tmp37_reg2mem_1 = new PHINode(VectorTy_4, "tmp37.reg2mem.1", label_cond_cont48); + packed_tmp37_reg2mem_1->reserveOperandSpace(2); + packed_tmp37_reg2mem_1->addIncoming(packed_tmp378, label_cond__28); + packed_tmp37_reg2mem_1->addIncoming(packed_tmp37, label_cond_cont34); + + ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_32, "tmp51", label_cond_cont48); + new ReturnInst(packed_tmp51, label_cond_cont48); + +} + +// Function: vcos (func_vcos) +{ + Function::arg_iterator args = func_vcos->arg_begin(); + Value* packed_val = args++; + packed_val->setName("val"); + + BasicBlock* label_entry_53 = new BasicBlock("entry",func_vcos,0); + + // Block entry (label_entry_53) + ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_53); + CastInst* double_conv_54 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_53); + ExtractElementInst* float_tmp3_55 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_53); + CastInst* double_conv4 = new FPExtInst(float_tmp3_55, Type::DoubleTy, "conv4", label_entry_53); + ExtractElementInst* float_tmp6_56 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_53); + CastInst* double_conv7 = new FPExtInst(float_tmp6_56, Type::DoubleTy, "conv7", label_entry_53); + ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_53); + CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_53); + std::vector int32_call_params; + int32_call_params.push_back(const_ptr_34); + int32_call_params.push_back(double_conv_54); + int32_call_params.push_back(double_conv4); + int32_call_params.push_back(double_conv7); + int32_call_params.push_back(double_conv10); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_53); + int32_call->setCallingConv(CallingConv::C); + int32_call->setTailCall(true);const ParamAttrsList *int32_call_PAL = 0; + int32_call->setParamAttrs(int32_call_PAL); + + CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_53); + float_call13->setCallingConv(CallingConv::C); + float_call13->setTailCall(true);const ParamAttrsList *float_call13_PAL = 0; + float_call13->setParamAttrs(float_call13_PAL); + + InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_53); + CallInst* float_call18 = new CallInst(func_cosf, float_tmp1, "call18", label_entry_53); + float_call18->setCallingConv(CallingConv::C); + float_call18->setTailCall(true);const ParamAttrsList *float_call18_PAL = 0; + float_call18->setParamAttrs(float_call18_PAL); + + InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call18, const_int32_23, "tmp20", label_entry_53); + CallInst* float_call23 = new CallInst(func_cosf, float_tmp1, "call23", label_entry_53); + float_call23->setCallingConv(CallingConv::C); + float_call23->setTailCall(true);const ParamAttrsList *float_call23_PAL = 0; + float_call23->setParamAttrs(float_call23_PAL); + + InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call23, const_int32_25, "tmp25", label_entry_53); + CallInst* float_call28 = new CallInst(func_cosf, float_tmp1, "call28", label_entry_53); + float_call28->setCallingConv(CallingConv::C); + float_call28->setTailCall(true);const ParamAttrsList *float_call28_PAL = 0; + float_call28->setParamAttrs(float_call28_PAL); + + InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call28, const_int32_24, "tmp30", label_entry_53); + CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_53); + CastInst* double_conv36 = new FPExtInst(float_call18, Type::DoubleTy, "conv36", label_entry_53); + CastInst* double_conv39 = new FPExtInst(float_call23, Type::DoubleTy, "conv39", label_entry_53); + CastInst* double_conv42 = new FPExtInst(float_call28, Type::DoubleTy, "conv42", label_entry_53); + std::vector int32_call43_params; + int32_call43_params.push_back(const_ptr_36); + int32_call43_params.push_back(double_conv33); + int32_call43_params.push_back(double_conv36); + int32_call43_params.push_back(double_conv39); + int32_call43_params.push_back(double_conv42); + CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_53); + int32_call43->setCallingConv(CallingConv::C); + int32_call43->setTailCall(true);const ParamAttrsList *int32_call43_PAL = 0; + int32_call43->setParamAttrs(int32_call43_PAL); + + new ReturnInst(packed_tmp30, label_entry_53); + +} + +// Function: scs (func_scs) +{ + Function::arg_iterator args = func_scs->arg_begin(); + Value* packed_val_58 = args++; + packed_val_58->setName("val"); + + BasicBlock* label_entry_59 = new BasicBlock("entry",func_scs,0); + + // Block entry (label_entry_59) + ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_58, const_int32_19, "tmp2", label_entry_59); + CallInst* float_call_60 = new CallInst(func_cosf, float_tmp2, "call", label_entry_59); + float_call_60->setCallingConv(CallingConv::C); + float_call_60->setTailCall(true);const ParamAttrsList *float_call_60_PAL = 0; + float_call_60->setParamAttrs(float_call_60_PAL); + + InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_60, const_int32_19, "tmp5", label_entry_59); + CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_59); + float_call7->setCallingConv(CallingConv::C); + float_call7->setTailCall(true);const ParamAttrsList *float_call7_PAL = 0; + float_call7->setParamAttrs(float_call7_PAL); + + InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_59); + new ReturnInst(packed_tmp9, label_entry_59); + +} + +// Function: vsin (func_vsin) +{ + Function::arg_iterator args = func_vsin->arg_begin(); + Value* packed_val_62 = args++; + packed_val_62->setName("val"); + + BasicBlock* label_entry_63 = new BasicBlock("entry",func_vsin,0); + + // Block entry (label_entry_63) + ExtractElementInst* float_tmp2_64 = new ExtractElementInst(packed_val_62, const_int32_19, "tmp2", label_entry_63); + CallInst* float_call_65 = new CallInst(func_sinf, float_tmp2_64, "call", label_entry_63); + float_call_65->setCallingConv(CallingConv::C); + float_call_65->setTailCall(true);const ParamAttrsList *float_call_65_PAL = 0; + float_call_65->setParamAttrs(float_call_65_PAL); + + InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_65, const_int32_19, "tmp6", label_entry_63); + InsertElementInst* packed_tmp9_66 = new InsertElementInst(packed_tmp6, float_call_65, const_int32_23, "tmp9", label_entry_63); + InsertElementInst* packed_tmp12 = new InsertElementInst(packed_tmp9_66, float_call_65, const_int32_25, "tmp12", label_entry_63); + InsertElementInst* packed_tmp15_67 = new InsertElementInst(packed_tmp12, float_call_65, const_int32_24, "tmp15", label_entry_63); + new ReturnInst(packed_tmp15_67, label_entry_63); + +} + +// Function: kilp (func_kilp) +{ + Function::arg_iterator args = func_kilp->arg_begin(); + Value* packed_val_69 = args++; + packed_val_69->setName("val"); + + BasicBlock* label_entry_70 = new BasicBlock("entry",func_kilp,0); + BasicBlock* label_lor_rhs = new BasicBlock("lor_rhs",func_kilp,0); + BasicBlock* label_lor_rhs5 = new BasicBlock("lor_rhs5",func_kilp,0); + BasicBlock* label_lor_rhs11 = new BasicBlock("lor_rhs11",func_kilp,0); + BasicBlock* label_UnifiedReturnBlock_71 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); + + // Block entry (label_entry_70) + ExtractElementInst* float_tmp1_72 = new ExtractElementInst(packed_val_69, const_int32_19, "tmp1", label_entry_70); + FCmpInst* int1_cmp_73 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_72, const_float_18, "cmp", label_entry_70); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs, int1_cmp_73, label_entry_70); + + // Block lor_rhs (label_lor_rhs) + ExtractElementInst* float_tmp3_75 = new ExtractElementInst(packed_val_69, const_int32_23, "tmp3", label_lor_rhs); + FCmpInst* int1_cmp4 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_75, const_float_18, "cmp4", label_lor_rhs); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs5, int1_cmp4, label_lor_rhs); + + // Block lor_rhs5 (label_lor_rhs5) + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_val_69, const_int32_25, "tmp7", label_lor_rhs5); + FCmpInst* int1_cmp8 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp7, const_float_18, "cmp8", label_lor_rhs5); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs11, int1_cmp8, label_lor_rhs5); + + // Block lor_rhs11 (label_lor_rhs11) + ExtractElementInst* float_tmp13 = new ExtractElementInst(packed_val_69, const_int32_24, "tmp13", label_lor_rhs11); + FCmpInst* int1_cmp14 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp13, const_float_18, "cmp14", label_lor_rhs11); + CastInst* int32_retval = new ZExtInst(int1_cmp14, IntegerType::get(32), "retval", label_lor_rhs11); + new ReturnInst(int32_retval, label_lor_rhs11); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock_71) + new ReturnInst(const_int32_23, label_UnifiedReturnBlock_71); + +} + +return mod; + +} diff --git a/src/gallium/aux/llvm/gallivm_cpu.cpp b/src/gallium/aux/llvm/gallivm_cpu.cpp new file mode 100644 index 0000000000..dc4d92a72a --- /dev/null +++ b/src/gallium/aux/llvm/gallivm_cpu.cpp @@ -0,0 +1,202 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "instructions.h" +#include "loweringpass.h" +#include "storage.h" +#include "tgsitollvm.h" + +#include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" + +#include "pipe/tgsi/exec/tgsi_exec.h" +#include "pipe/tgsi/util/tgsi_dump.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +struct gallivm_cpu_engine { + llvm::ExecutionEngine *engine; +}; + +static struct gallivm_cpu_engine *CPU = 0; + +typedef int (*fragment_shader_runner)(float x, float y, + float (*dests)[16][4], + float (*inputs)[16][4], + int num_attribs, + float (*consts)[4], int num_consts, + struct tgsi_sampler *samplers); + +int gallivm_cpu_fs_exec(struct gallivm_prog *prog, + float fx, float fy, + float (*dests)[16][4], + float (*inputs)[16][4], + float (*consts)[4], + struct tgsi_sampler *samplers) +{ + fragment_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + + return runner(fx, fy, dests, inputs, prog->num_interp, + consts, prog->num_consts, + samplers); +} + +static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) +{ + llvm::Module *mod = prog->module; + llvm::Function *func = 0; + + switch (prog->type) { + case GALLIVM_VS: + func = mod->getFunction("vs_shader"); + break; + case GALLIVM_FS: + func = mod->getFunction("fs_shader"); + break; + default: + assert(!"Unknown shader type!"); + break; + } + return func; +} + +/*! + This function creates a CPU based execution engine for the given gallivm_prog. + gallivm_cpu_engine should be used as a singleton throughout the library. Before + executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile. + The gallivm_prog instance which is being passed to the constructor is being + automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile + with it again. + */ +struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) +{ + struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *) + calloc(1, sizeof(struct gallivm_cpu_engine)); + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); + ee->DisableLazyCompilation(); + cpu->engine = ee; + + llvm::Function *func = func_for_shader(prog); + + prog->function = ee->getPointerToFunction(func); + CPU = cpu; + return cpu; +} + + +/*! + This function JIT compiles the given gallivm_prog with the given cpu based execution engine. + The reference to the generated machine code entry point will be stored + in the gallivm_prog program. After executing this function one can call gallivm_prog_exec + in order to execute the gallivm_prog on the CPU. + */ +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) +{ + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = cpu->engine; + assert(ee); + /*FIXME : remove */ + ee->DisableLazyCompilation(); + ee->addModuleProvider(mp); + + llvm::Function *func = func_for_shader(prog); + prog->function = ee->getPointerToFunction(func); +} + +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) +{ + free(cpu); +} + +struct gallivm_cpu_engine * gallivm_global_cpu_engine() +{ + return CPU; +} + + +typedef void (*vertex_shader_runner)(void *ainputs, + void *dests, + float (*aconsts)[4], + void *temps); + + +/*! + This function is used to execute the gallivm_prog in software. Before calling + this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile + function. + */ +int gallivm_cpu_vs_exec(struct gallivm_prog *prog, + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, + float (*consts)[4], + struct tgsi_exec_vector *temps) +{ + vertex_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + /*FIXME*/ + runner(inputs, dests, consts, temps); + + return 0; +} + +#endif diff --git a/src/gallium/aux/llvm/gallivm_p.h b/src/gallium/aux/llvm/gallivm_p.h new file mode 100644 index 0000000000..cfe7b1901b --- /dev/null +++ b/src/gallium/aux/llvm/gallivm_p.h @@ -0,0 +1,110 @@ +#ifndef GALLIVM_P_H +#define GALLIVM_P_H + +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/p_compiler.h" + +namespace llvm { + class Module; +} + +#if defined __cplusplus +extern "C" { +#endif + +enum gallivm_shader_type; +enum gallivm_vector_layout; + +struct gallivm_interpolate { + int attrib; + int chan; + int type; +}; + +struct gallivm_ir { + llvm::Module *module; + int id; + enum gallivm_shader_type type; + enum gallivm_vector_layout layout; + int num_components; + int num_consts; + + //FIXME: this might not be enough for some shaders + struct gallivm_interpolate interpolators[32*4]; + int num_interp; +}; + +struct gallivm_prog { + llvm::Module *module; + void *function; + + int id; + enum gallivm_shader_type type; + + int num_consts; + + //FIXME: this might not be enough for some shaders + struct gallivm_interpolate interpolators[32*4]; + int num_interp; +}; + +static INLINE void gallivm_swizzle_components(int swizzle, + int *xc, int *yc, + int *zc, int *wc) +{ + int x = swizzle / 1000; swizzle -= x * 1000; + int y = swizzle / 100; swizzle -= y * 100; + int z = swizzle / 10; swizzle -= z * 10; + int w = swizzle; + + if (xc) *xc = x; + if (yc) *yc = y; + if (zc) *zc = z; + if (wc) *wc = w; +} + +static INLINE boolean gallivm_is_swizzle(int swizzle) +{ + const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + + TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; + return swizzle != NO_SWIZZLE; +} + +static INLINE int gallivm_x_swizzle(int swizzle) +{ + int x; + gallivm_swizzle_components(swizzle, &x, 0, 0, 0); + return x; +} + +static INLINE int gallivm_y_swizzle(int swizzle) +{ + int y; + gallivm_swizzle_components(swizzle, 0, &y, 0, 0); + return y; +} + +static INLINE int gallivm_z_swizzle(int swizzle) +{ + int z; + gallivm_swizzle_components(swizzle, 0, 0, &z, 0); + return z; +} + +static INLINE int gallivm_w_swizzle(int swizzle) +{ + int w; + gallivm_swizzle_components(swizzle, 0, 0, 0, &w); + return w; +} + +#endif /* MESA_LLVM */ + +#if defined __cplusplus +} // extern "C" +#endif + +#endif diff --git a/src/gallium/aux/llvm/instructions.cpp b/src/gallium/aux/llvm/instructions.cpp new file mode 100644 index 0000000000..55d39fa5f1 --- /dev/null +++ b/src/gallium/aux/llvm/instructions.cpp @@ -0,0 +1,889 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "instructions.h" + +#include "storage.h" + +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +using namespace llvm; + +#include "gallivm_builtins.cpp" + +static inline std::string createFuncName(int label) +{ + std::ostringstream stream; + stream << "function"; + stream << label; + return stream.str(); +} + +Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, + Storage *storage) + : m_mod(mod), m_func(func), m_builder(block), m_idx(0), + m_storage(storage) +{ + m_floatVecType = VectorType::get(Type::FloatTy, 4); + + m_llvmFSqrt = 0; + m_llvmFAbs = 0; + m_llvmPow = 0; + m_llvmFloor = 0; + m_llvmFlog = 0; + m_llvmLit = 0; + m_fmtPtr = 0; + + createGallivmBuiltins(m_mod); +} + +llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) +{ + return m_builder.CreateAdd(in1, in2, name("add")); +} + +llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3) +{ + Value *mulRes = mul(in1, in2); + return add(mulRes, in3); +} + +llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2) +{ + return m_builder.CreateMul(in1, in2, name("mul")); +} + +const char * Instructions::name(const char *prefix) +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + Value *x = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(0), + name("extractx")); + Value *y = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(1), + name("extracty")); + Value *z = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(2), + name("extractz")); + Value *xy = m_builder.CreateAdd(x, y,name("xy")); + Value *dot3 = m_builder.CreateAdd(xy, z, name("dot3")); + return vectorFromVals(dot3, dot3, dot3, dot3); +} + +llvm::Value *Instructions::callFSqrt(llvm::Value *val) +{ + if (!m_llvmFSqrt) { + // predeclare the intrinsic + std::vector fsqrtArgs; + fsqrtArgs.push_back(Type::FloatTy); + ParamAttrsList *fsqrtPal = 0; + FunctionType* fsqrtType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/fsqrtArgs, + /*isVarArg=*/false); + m_llvmFSqrt = new Function( + /*Type=*/fsqrtType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"llvm.sqrt.f32", m_mod); + m_llvmFSqrt->setCallingConv(CallingConv::C); + m_llvmFSqrt->setParamAttrs(fsqrtPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFSqrt, val, + name("sqrt")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::rsq(llvm::Value *in1) +{ + Value *x = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("extractx")); + Value *abs = callFAbs(x); + Value *sqrt = callFSqrt(abs); + + Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + sqrt, + name("rsqrt")); + return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt); +} + +llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w) +{ + Constant *const_vec = Constant::getNullValue(m_floatVecType); + Value *res = m_builder.CreateInsertElement(const_vec, x, + m_storage->constantInt(0), + name("vecx")); + res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), + name("vecxy")); + res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), + name("vecxyz")); + if (w) + res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), + name("vecxyzw")); + return res; +} + +llvm::Value *Instructions::callFAbs(llvm::Value *val) +{ + if (!m_llvmFAbs) { + // predeclare the intrinsic + std::vector fabsArgs; + fabsArgs.push_back(Type::FloatTy); + ParamAttrsList *fabsPal = 0; + FunctionType* fabsType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/fabsArgs, + /*isVarArg=*/false); + m_llvmFAbs = new Function( + /*Type=*/fabsType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"fabs", m_mod); + m_llvmFAbs->setCallingConv(CallingConv::C); + m_llvmFAbs->setParamAttrs(fabsPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFAbs, val, + name("fabs")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::lit(llvm::Value *in) +{ + if (!m_llvmLit) { + m_llvmLit = m_mod->getFunction("lit"); + } + CallInst *call = m_builder.CreateCall(m_llvmLit, in, name("litres")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) +{ + Value *res = m_builder.CreateSub(in1, in2, name("sub")); + return res; +} + +llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) +{ + if (!m_llvmPow) { + // predeclare the intrinsic + std::vector powArgs; + powArgs.push_back(Type::FloatTy); + powArgs.push_back(Type::FloatTy); + ParamAttrsList *powPal = 0; + FunctionType* powType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/powArgs, + /*isVarArg=*/false); + m_llvmPow = new Function( + /*Type=*/powType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"llvm.pow.f32", m_mod); + m_llvmPow->setCallingConv(CallingConv::C); + m_llvmPow->setParamAttrs(powPal); + } + std::vector params; + params.push_back(val1); + params.push_back(val2); + CallInst *call = m_builder.CreateCall(m_llvmPow, params.begin(), params.end(), + name("pow")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2) +{ + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); + llvm::Value *val = callPow(x1, x2); + return vectorFromVals(val, val, val, val); +} + +llvm::Value * Instructions::rcp(llvm::Value *in1) +{ + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *res = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + x1, name("rcp")); + return vectorFromVals(res, res, res, res); +} + +llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + std::vector vec = extractVector(mulRes); + Value *xy = m_builder.CreateAdd(vec[0], vec[1], name("xy")); + Value *xyz = m_builder.CreateAdd(xy, vec[2], name("xyz")); + Value *dot4 = m_builder.CreateAdd(xyz, vec[3], name("dot4")); + return vectorFromVals(dot4, dot4, dot4, dot4); +} + +llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + std::vector vec1 = extractVector(mulRes); + Value *xy = m_builder.CreateAdd(vec1[0], vec1[1], name("xy")); + Value *xyz = m_builder.CreateAdd(xy, vec1[2], name("xyz")); + Value *dph = m_builder.CreateAdd(xyz, vec1[3], name("dph")); + return vectorFromVals(dph, dph, dph, dph); +} + +llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) +{ + Value *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + Value *z = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z")); + Value *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + Value *w = m_builder.CreateExtractElement(in2, + m_storage->constantInt(3), + name("w")); + Value *ry = m_builder.CreateMul(y1, y2, name("tyuy")); + return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), + ry, z, w); +} + +llvm::Value * Instructions::ex2(llvm::Value *in) +{ + llvm::Value *val = callPow(ConstantFP::get(Type::FloatTy, APFloat(2.f)), + m_builder.CreateExtractElement( + in, m_storage->constantInt(0), + name("x1"))); + return vectorFromVals(val, val, val, val); +} + +llvm::Value * Instructions::callFloor(llvm::Value *val) +{ + if (!m_llvmFloor) { + // predeclare the intrinsic + std::vector floorArgs; + floorArgs.push_back(Type::FloatTy); + ParamAttrsList *floorPal = 0; + FunctionType* floorType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/floorArgs, + /*isVarArg=*/false); + m_llvmFloor = new Function( + /*Type=*/floorType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"floorf", m_mod); + m_llvmFloor->setCallingConv(CallingConv::C); + m_llvmFloor->setParamAttrs(floorPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFloor, val, + name("floorf")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::floor(llvm::Value *in) +{ + std::vector vec = extractVector(in); + return vectorFromVals(callFloor(vec[0]), callFloor(vec[1]), + callFloor(vec[2]), callFloor(vec[3])); +} + +llvm::Value * Instructions::arl(llvm::Value *in) +{ + return floor(in); +} + +llvm::Value * Instructions::frc(llvm::Value *in) +{ + llvm::Value *flr = floor(in); + return sub(in, flr); +} + +llvm::Value * Instructions::callFLog(llvm::Value *val) +{ + if (!m_llvmFlog) { + // predeclare the intrinsic + std::vector flogArgs; + flogArgs.push_back(Type::FloatTy); + ParamAttrsList *flogPal = 0; + FunctionType* flogType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/flogArgs, + /*isVarArg=*/false); + m_llvmFlog = new Function( + /*Type=*/flogType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"logf", m_mod); + m_llvmFlog->setCallingConv(CallingConv::C); + m_llvmFlog->setParamAttrs(flogPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFlog, val, + name("logf")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::lg2(llvm::Value *in) +{ + std::vector vec = extractVector(in); + llvm::Value *const_vec = constVector(1.442695f, 1.442695f, + 1.442695f, 1.442695f); + return mul(vectorFromVals(callFLog(vec[0]), callFLog(vec[1]), + callFLog(vec[2]), callFLog(vec[3])), const_vec); +} + +llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) +{ + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); + Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); + + Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); + Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); + + Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); + Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); + + Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); + Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); + + return vectorFromVals(selx, sely, selz, selw); +} + +llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) +{ + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], + name("xcmp")); + Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); + + Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], + name("ycmp")); + Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); + + Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], + name("zcmp")); + Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); + + Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], + name("wcmp")); + Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); + + return vectorFromVals(selx, sely, selz, selw); +} + +void Instructions::printVector(llvm::Value *val) +{ + static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A"; + + if (!m_fmtPtr) { + Constant *format = ConstantArray::get(frmt, true); + ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1); + GlobalVariable* globalFormat = new GlobalVariable( + /*Type=*/arrayTy, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name(".str"), + m_mod); + globalFormat->setInitializer(format); + + Constant* const_int0 = Constant::getNullValue(IntegerType::get(32)); + std::vector const_ptr_21_indices; + const_ptr_21_indices.push_back(const_int0); + const_ptr_21_indices.push_back(const_int0); + m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat, + &const_ptr_21_indices[0], const_ptr_21_indices.size()); + } + + Function *func_printf = m_mod->getFunction("printf"); + if (!func_printf) + func_printf = declarePrintf(); + assert(func_printf); + std::vector vec = extractVector(val); + Value *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx")); + Value *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy")); + Value *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz")); + Value *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw")); + std::vector params; + params.push_back(m_fmtPtr); + params.push_back(dx); + params.push_back(dy); + params.push_back(dz); + params.push_back(dw); + CallInst *call = m_builder.CreateCall(func_printf, params.begin(), params.end(), + name("printf")); + call->setCallingConv(CallingConv::C); + call->setTailCall(true); +} + +llvm::Function * Instructions::declarePrintf() +{ + std::vector args; + ParamAttrsList *params = 0; + FunctionType* funcTy = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/args, + /*isVarArg=*/true); + Function* func_printf = new Function( + /*Type=*/funcTy, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", m_mod); + func_printf->setCallingConv(CallingConv::C); + func_printf->setParamAttrs(params); + return func_printf; +} + + +llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], name("xcmp")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + + Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], name("ycmp")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + + Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], name("zcmp")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + + Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], name("wcmp")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + + return vectorFromVals(x, y, z, w); +} +llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOGE(vec1[0], vec2[0], name("xcmp")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + + Value *ycmp = m_builder.CreateFCmpOGE(vec1[1], vec2[1], name("ycmp")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + + Value *zcmp = m_builder.CreateFCmpOGE(vec1[2], vec2[2], name("zcmp")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + + Value *wcmp = m_builder.CreateFCmpOGE(vec1[3], vec2[3], name("wcmp")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + + return vectorFromVals(x, y, z, w); +} + + +llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + + Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + + Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + + Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + + return vectorFromVals(x, y, z, w); +} + +llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) +{ + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + Value *z1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z1")); + + Value *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); + Value *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + Value *z2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(2), + name("z2")); + Value *y1z2 = mul(y1, z2); + Value *z1y2 = mul(z1, y2); + + Value *z1x2 = mul(z1, x2); + Value *x1z2 = mul(x1, z2); + + Value *x1y2 = mul(x1, y2); + Value *y1x2 = mul(y1, x2); + + return vectorFromVals(sub(y1z2, z1y2), sub(z1x2, x1z2), sub(x1y2, y1x2)); +} + + +llvm::Value * Instructions::abs(llvm::Value *in) +{ + std::vector vec = extractVector(in); + Value *xabs = callFAbs(vec[0]); + Value *yabs = callFAbs(vec[1]); + Value *zabs = callFAbs(vec[2]); + Value *wabs = callFAbs(vec[3]); + return vectorFromVals(xabs, yabs, zabs, wabs); +} + +void Instructions::ifop(llvm::Value *in) +{ + BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); + BasicBlock *ifend = new BasicBlock(name("ifthenend"), m_func,0); + + //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0); + //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0); + //BasicBlock *wblock = new BasicBlock(name("wblock"), m_func,0); + + Constant *float0 = Constant::getNullValue(Type::FloatTy); + + Value *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0), + name("extractx")); + Value *xcmp = m_builder.CreateFCmpUNE(x, float0, name("xcmp")); + m_builder.CreateCondBr(xcmp, ifthen, ifend); + //m_builder.SetInsertPoint(yblock); + + m_builder.SetInsertPoint(ifthen); + m_ifStack.push(ifend); +} + +llvm::BasicBlock * Instructions::currentBlock() const +{ + return m_builder.GetInsertBlock(); +} + +void Instructions::elseop() +{ + assert(!m_ifStack.empty()); + BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); + m_builder.CreateBr(ifend); + m_builder.SetInsertPoint(m_ifStack.top()); + currentBlock()->setName(name("ifelse")); + m_ifStack.pop(); + m_ifStack.push(ifend); +} + +void Instructions::endif() +{ + assert(!m_ifStack.empty()); + m_builder.CreateBr(m_ifStack.top()); + m_builder.SetInsertPoint(m_ifStack.top()); + m_ifStack.pop(); +} + +llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3) +{ + llvm::Value *m = mul(in1, in2); + llvm::Value *vec1 = constVector(1.f, 1.f, 1.f, 1.f); + llvm::Value *s = sub(vec1, in1); + return add(m, mul(s, in3)); +} + +void Instructions::beginLoop() +{ + BasicBlock *begin = new BasicBlock(name("loop"), m_func,0); + BasicBlock *end = new BasicBlock(name("endloop"), m_func,0); + + m_builder.CreateBr(begin); + Loop loop; + loop.begin = begin; + loop.end = end; + m_builder.SetInsertPoint(begin); + m_loopStack.push(loop); +} + +void Instructions::endLoop() +{ + assert(!m_loopStack.empty()); + Loop loop = m_loopStack.top(); + m_builder.CreateBr(loop.begin); + loop.end->moveAfter(currentBlock()); + m_builder.SetInsertPoint(loop.end); + m_loopStack.pop(); +} + +void Instructions::brk() +{ + assert(!m_loopStack.empty()); + BasicBlock *unr = new BasicBlock(name("unreachable"), m_func,0); + m_builder.CreateBr(m_loopStack.top().end); + m_builder.SetInsertPoint(unr); +} + +llvm::Value * Instructions::trunc(llvm::Value *in) +{ + std::vector vec = extractVector(in); + Value *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32), + name("ftoix")); + Value *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32), + name("ftoiy")); + Value *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32), + name("ftoiz")); + Value *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32), + name("ftoiw")); + Value *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy, + name("fx")); + Value *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy, + name("fy")); + Value *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy, + name("fz")); + Value *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy, + name("fw")); + return vectorFromVals(fx, fy, fz, fw); +} + +void Instructions::end() +{ + m_builder.CreateRetVoid(); +} + +void Instructions::cal(int label, llvm::Value *input) +{ + std::vector params; + params.push_back(input); + llvm::Function *func = findFunction(label); + + m_builder.CreateCall(func, params.begin(), params.end()); +} + +llvm::Function * Instructions::declareFunc(int label) +{ + PointerType *vecPtr = PointerType::getUnqual(m_floatVecType); + std::vector args; + args.push_back(vecPtr); + args.push_back(vecPtr); + args.push_back(vecPtr); + args.push_back(vecPtr); + ParamAttrsList *params = 0; + FunctionType *funcType = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/args, + /*isVarArg=*/false); + std::string name = createFuncName(label); + Function *func = new Function( + /*Type=*/funcType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/name.c_str(), m_mod); + func->setCallingConv(CallingConv::C); + func->setParamAttrs(params); + return func; +} + +void Instructions::bgnSub(unsigned label) +{ + llvm::Function *func = findFunction(label); + + Function::arg_iterator args = func->arg_begin(); + Value *ptr_INPUT = args++; + ptr_INPUT->setName("INPUT"); + m_storage->pushArguments(ptr_INPUT); + + llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); + + m_func = func; + m_builder.SetInsertPoint(entry); +} + +void Instructions::endSub() +{ + m_func = 0; + m_builder.SetInsertPoint(0); +} + +llvm::Function * Instructions::findFunction(int label) +{ + llvm::Function *func = m_functions[label]; + if (!func) { + func = declareFunc(label); + m_functions[label] = func; + } + return func; +} + +llvm::Value * Instructions::constVector(float x, float y, float z, float w) +{ + std::vector vec(4); + vec[0] = ConstantFP::get(Type::FloatTy, APFloat(x)); + vec[1] = ConstantFP::get(Type::FloatTy, APFloat(y)); + vec[2] = ConstantFP::get(Type::FloatTy, APFloat(z)); + vec[3] = ConstantFP::get(Type::FloatTy, APFloat(w)); + return ConstantVector::get(m_floatVecType, vec); +} + + +std::vector Instructions::extractVector(llvm::Value *vec) +{ + std::vector elems(4); + elems[0] = m_builder.CreateExtractElement(vec, m_storage->constantInt(0), + name("x")); + elems[1] = m_builder.CreateExtractElement(vec, m_storage->constantInt(1), + name("y")); + elems[2] = m_builder.CreateExtractElement(vec, m_storage->constantInt(2), + name("z")); + elems[3] = m_builder.CreateExtractElement(vec, m_storage->constantInt(3), + name("w")); + return elems; +} + +llvm::Value * Instructions::cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3) +{ + llvm::Function *func = m_mod->getFunction("cmp"); + assert(func); + + std::vector params; + params.push_back(in1); + params.push_back(in2); + params.push_back(in3); + CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), name("cmpres")); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::cos(llvm::Value *in) +{ +#if 0 + llvm::Function *func = m_mod->getFunction("vcos"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("cosres")); + call->setTailCall(false); + return call; +#else + std::vector elems = extractVector(in); + Function *func = m_mod->getFunction("cosf"); + assert(func); + CallInst *cos = m_builder.CreateCall(func, elems[0], name("cosres")); + cos->setCallingConv(CallingConv::C); + cos->setTailCall(true); + return vectorFromVals(cos, cos, cos, cos); +#endif +} + +llvm::Value * Instructions::scs(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("scs"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("scsres")); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::kilp(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("kilp"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("kilpres")); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::sin(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("vsin"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("sinres")); + call->setTailCall(false); + return call; +} +#endif //MESA_LLVM + + diff --git a/src/gallium/aux/llvm/instructions.h b/src/gallium/aux/llvm/instructions.h new file mode 100644 index 0000000000..9ebc17dd8e --- /dev/null +++ b/src/gallium/aux/llvm/instructions.h @@ -0,0 +1,152 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef INSTRUCTIONS_H +#define INSTRUCTIONS_H + +#include +#include +#include +#include + +#include +#include + +namespace llvm { + class VectorType; + class Function; +} + +class Storage; + +class Instructions +{ +public: + Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, + Storage *storage); + + llvm::BasicBlock *currentBlock() const; + + llvm::Value *abs(llvm::Value *in1); + llvm::Value *arl(llvm::Value *in1); + llvm::Value *add(llvm::Value *in1, llvm::Value *in2); + void beginLoop(); + void bgnSub(unsigned); + void brk(); + void cal(int label, llvm::Value *input); + llvm::Value *cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); + llvm::Value *cos(llvm::Value *in); + llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); + void elseop(); + void endif(); + void endLoop(); + void end(); + void endSub(); + llvm::Value *ex2(llvm::Value *in); + llvm::Value *floor(llvm::Value *in); + llvm::Value *frc(llvm::Value *in); + void ifop(llvm::Value *in); + llvm::Value *kilp(llvm::Value *in); + llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3); + llvm::Value *lit(llvm::Value *in); + llvm::Value *lg2(llvm::Value *in); + llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in2); + llvm::Value *min(llvm::Value *in1, llvm::Value *in2); + llvm::Value *max(llvm::Value *in1, llvm::Value *in2); + llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); + llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); + llvm::Value *rcp(llvm::Value *in); + llvm::Value *rsq(llvm::Value *in); + llvm::Value *scs(llvm::Value *in); + llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sin(llvm::Value *in); + llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); + llvm::Value *trunc(llvm::Value *in); + + void printVector(llvm::Value *val); +private: + const char *name(const char *prefix); + + llvm::Value *callFAbs(llvm::Value *val); + llvm::Value *callFloor(llvm::Value *val); + llvm::Value *callFSqrt(llvm::Value *val); + llvm::Value *callFLog(llvm::Value *val); + llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2); + + llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w=0); + + llvm::Value *constVector(float x, float y, float z, float w); + + llvm::Function *declarePrintf(); + llvm::Function *declareFunc(int label); + + llvm::Function *findFunction(int label); + + std::vector extractVector(llvm::Value *vec); +private: + llvm::Module *m_mod; + llvm::Function *m_func; + char m_name[32]; + llvm::LLVMFoldingBuilder m_builder; + int m_idx; + + llvm::VectorType *m_floatVecType; + + llvm::Function *m_llvmFSqrt; + llvm::Function *m_llvmFAbs; + llvm::Function *m_llvmPow; + llvm::Function *m_llvmFloor; + llvm::Function *m_llvmFlog; + llvm::Function *m_llvmLit; + + llvm::Constant *m_fmtPtr; + + std::stack m_ifStack; + struct Loop { + llvm::BasicBlock *begin; + llvm::BasicBlock *end; + }; + std::stack m_loopStack; + std::map m_functions; + Storage *m_storage; +}; + +#endif diff --git a/src/gallium/aux/llvm/instructionssoa.cpp b/src/gallium/aux/llvm/instructionssoa.cpp new file mode 100644 index 0000000000..a4d5046637 --- /dev/null +++ b/src/gallium/aux/llvm/instructionssoa.cpp @@ -0,0 +1,121 @@ +#include "instructionssoa.h" + +#include "storagesoa.h" + +#include + +using namespace llvm; + +InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, + llvm::BasicBlock *block, StorageSoa *storage) + : m_builder(block), + m_storage(storage), + m_idx(0) +{ +} + +const char * InstructionsSoa::name(const char *prefix) const +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::Value * InstructionsSoa::vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + Constant *constVector = Constant::getNullValue(vectorType); + Value *res = m_builder.CreateInsertElement(constVector, x, + m_storage->constantInt(0), + name("vecx")); + res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), + name("vecxy")); + res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), + name("vecxyz")); + if (w) + res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), + name("vecxyzw")); + return res; +} + +std::vector InstructionsSoa::arl(const std::vector in) +{ + std::vector res(4); + + //Extract x's + llvm::Value *x1 = m_builder.CreateExtractElement(in[0], + m_storage->constantInt(0), + name("extractX")); + //cast it to an unsigned int + x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast")); + + res[0] = x1;//vectorFromVals(x1, x2, x3, x4); + //only x is valid. the others shouldn't be necessary + /* + res[1] = Constant::getNullValue(m_floatVecType); + res[2] = Constant::getNullValue(m_floatVecType); + res[3] = Constant::getNullValue(m_floatVecType); + */ + + return res; +} + + +std::vector InstructionsSoa::add(const std::vector in1, + const std::vector in2) +{ + std::vector res(4); + + res[0] = m_builder.CreateAdd(in1[0], in2[0], name("addx")); + res[1] = m_builder.CreateAdd(in1[1], in2[1], name("addy")); + res[2] = m_builder.CreateAdd(in1[2], in2[2], name("addz")); + res[3] = m_builder.CreateAdd(in1[3], in2[3], name("addw")); + + return res; +} + +std::vector InstructionsSoa::mul(const std::vector in1, + const std::vector in2) +{ + std::vector res(4); + + res[0] = m_builder.CreateMul(in1[0], in2[0], name("mulx")); + res[1] = m_builder.CreateMul(in1[1], in2[1], name("muly")); + res[2] = m_builder.CreateMul(in1[2], in2[2], name("mulz")); + res[3] = m_builder.CreateMul(in1[3], in2[3], name("mulw")); + + return res; +} + +void InstructionsSoa::end() +{ + m_builder.CreateRetVoid(); +} + +std::vector InstructionsSoa::madd(const std::vector in1, + const std::vector in2, + const std::vector in3) +{ + std::vector res = mul(in1, in2); + return add(res, in3); +} + +std::vector InstructionsSoa::extractVector(llvm::Value *vector) +{ + std::vector res(4); + res[0] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(0), + name("extract1X")); + res[1] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(1), + name("extract2X")); + res[2] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(2), + name("extract3X")); + res[3] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(3), + name("extract4X")); + + return res; +} diff --git a/src/gallium/aux/llvm/instructionssoa.h b/src/gallium/aux/llvm/instructionssoa.h new file mode 100644 index 0000000000..4169dcbb2e --- /dev/null +++ b/src/gallium/aux/llvm/instructionssoa.h @@ -0,0 +1,74 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INSTRUCTIONSSOA_H +#define INSTRUCTIONSSOA_H + +#include + +#include + +namespace llvm { + class Module; + class Function; + class BasicBlock; + class Value; +} +class StorageSoa; + +class InstructionsSoa +{ +public: + InstructionsSoa(llvm::Module *mod, llvm::Function *func, + llvm::BasicBlock *block, StorageSoa *storage); + + std::vector arl(const std::vector in); + + std::vector add(const std::vector in1, + const std::vector in2); + std::vector madd(const std::vector in1, + const std::vector in2, + const std::vector in3); + std::vector mul(const std::vector in1, + const std::vector in2); + void end(); + + std::vector extractVector(llvm::Value *vector); +private: + const char * name(const char *prefix) const; + llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w); +private: + llvm::LLVMFoldingBuilder m_builder; + StorageSoa *m_storage; +private: + mutable int m_idx; + mutable char m_name[32]; +}; + + +#endif diff --git a/src/gallium/aux/llvm/llvm_builtins.c b/src/gallium/aux/llvm/llvm_builtins.c new file mode 100644 index 0000000000..4f98d754ba --- /dev/null +++ b/src/gallium/aux/llvm/llvm_builtins.c @@ -0,0 +1,115 @@ +/*clang --emit-llvm llvm_builtins.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=gallivm_builtins.cpp -f -for=shader -funcname=createGallivmBuiltins*/ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +typedef __attribute__(( ocu_vector_type(4) )) float float4; + +extern float powf(float a, float b); + +inline float approx(float a, float b) +{ + if (b < -128.0f) b = -128.0f; + if (b > 128.0f) b = 128.0f; + if (a < 0) a = 0; + return powf(a, b); +} + +inline float4 lit(float4 tmp) +{ + float4 result; + result.x = 1.0; + result.w = 1.0; + if (tmp.x > 0) { + result.y = tmp.x; + result.z = approx(tmp.y, tmp.w); + } else { + result.y = 0; + result.z = 0; + } + return result; +} + +inline float4 cmp(float4 tmp0, float4 tmp1, float4 tmp2) +{ + float4 result; + + result.x = (tmp0.x < 0.0) ? tmp1.x : tmp2.x; + result.y = (tmp0.y < 0.0) ? tmp1.y : tmp2.y; + result.z = (tmp0.z < 0.0) ? tmp1.z : tmp2.z; + result.w = (tmp0.w < 0.0) ? tmp1.w : tmp2.w; + + return result; +} + +extern float cosf(float val); +extern float sinf(float val); + +inline float4 vcos(float4 val) +{ + float4 result; + printf("VEC IN is %f %f %f %f\n", val.x, val.y, val.z, val.w); + result.x = cosf(val.x); + result.y = cosf(val.x); + result.z = cosf(val.x); + result.w = cosf(val.x); + printf("VEC OUT is %f %f %f %f\n", result.x, result.y, result.z, result.w); + return result; +} + +inline float4 scs(float4 val) +{ + float4 result; + float tmp = val.x; + result.x = cosf(tmp); + result.y = sinf(tmp); + return result; +} + + +inline float4 vsin(float4 val) +{ + float4 result; + float tmp = val.x; + float res = sinf(tmp); + result.x = res; + result.y = res; + result.z = res; + result.w = res; + return result; +} + +inline int kilp(float4 val) +{ + if (val.x < 0 || val.y < 0 || val.z < 0 || val.w < 0) + return 1; + else + return 0; +} diff --git a/src/gallium/aux/llvm/loweringpass.cpp b/src/gallium/aux/llvm/loweringpass.cpp new file mode 100644 index 0000000000..556dbec366 --- /dev/null +++ b/src/gallium/aux/llvm/loweringpass.cpp @@ -0,0 +1,17 @@ +#include "loweringpass.h" + +using namespace llvm; + +char LoweringPass::ID = 0; +RegisterPass X("lowering", "Lowering Pass"); + +LoweringPass::LoweringPass() + : ModulePass((intptr_t)&ID) +{ +} + +bool LoweringPass::runOnModule(Module &m) +{ + llvm::cerr << "Hello: " << m.getModuleIdentifier() << "\n"; + return false; +} diff --git a/src/gallium/aux/llvm/loweringpass.h b/src/gallium/aux/llvm/loweringpass.h new file mode 100644 index 0000000000..f62dcf6ba7 --- /dev/null +++ b/src/gallium/aux/llvm/loweringpass.h @@ -0,0 +1,15 @@ +#ifndef LOWERINGPASS_H +#define LOWERINGPASS_H + +#include "llvm/Pass.h" +#include "llvm/Module.h" + +struct LoweringPass : public llvm::ModulePass +{ + static char ID; + LoweringPass(); + + virtual bool runOnModule(llvm::Module &m); +}; + +#endif diff --git a/src/gallium/aux/llvm/storage.cpp b/src/gallium/aux/llvm/storage.cpp new file mode 100644 index 0000000000..c4326de8c5 --- /dev/null +++ b/src/gallium/aux/llvm/storage.cpp @@ -0,0 +1,364 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "storage.h" + +#include "gallivm_p.h" + +#include "pipe/p_shader_tokens.h" +#include +#include +#include + +#include +#include +#include +#include +#include + +using namespace llvm; + +Storage::Storage(llvm::BasicBlock *block, llvm::Value *input) + : m_block(block), + m_INPUT(input), + m_addrs(32), + m_idx(0) +{ + m_floatVecType = VectorType::get(Type::FloatTy, 4); + m_intVecType = VectorType::get(IntegerType::get(32), 4); + + m_undefFloatVec = UndefValue::get(m_floatVecType); + m_undefIntVec = UndefValue::get(m_intVecType); + m_extSwizzleVec = 0; + + m_numConsts = 0; +} + +//can only build vectors with all members in the [0, 9] range +llvm::Constant *Storage::shuffleMask(int vec) +{ + if (!m_extSwizzleVec) { + std::vector elems; + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); + m_extSwizzleVec = ConstantVector::get(m_floatVecType, elems); + } + + if (m_intVecs.find(vec) != m_intVecs.end()) { + return m_intVecs[vec]; + } + int origVec = vec; + Constant* const_vec = 0; + if (origVec == 0) { + const_vec = Constant::getNullValue(m_intVecType); + } else { + int x = gallivm_x_swizzle(vec); + int y = gallivm_y_swizzle(vec); + int z = gallivm_z_swizzle(vec); + int w = gallivm_w_swizzle(vec); + std::vector elems; + elems.push_back(constantInt(x)); + elems.push_back(constantInt(y)); + elems.push_back(constantInt(z)); + elems.push_back(constantInt(w)); + const_vec = ConstantVector::get(m_intVecType, elems); + } + + m_intVecs[origVec] = const_vec; + return const_vec; +} + +llvm::ConstantInt *Storage::constantInt(int idx) +{ + if (m_constInts.find(idx) != m_constInts.end()) { + return m_constInts[idx]; + } + ConstantInt *const_int = ConstantInt::get(APInt(32, idx)); + m_constInts[idx] = const_int; + return const_int; +} + +llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) +{ + Value *val = element(InputsArg, idx, indIdx); + LoadInst *load = new LoadInst(val, name("input"), false, m_block); + load->setAlignment(8); + + return load; +} + +llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) +{ + m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; + + Value *elem = element(ConstsArg, idx, indIdx); + LoadInst *load = new LoadInst(elem, name("const"), false, m_block); + load->setAlignment(8); + return load; +} + +llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) +{ + Constant *mask = shuffleMask(shuffle); + ShuffleVectorInst *res = + new ShuffleVectorInst(vec, m_extSwizzleVec, mask, + name("shuffle"), m_block); + return res; +} + + +llvm::Value *Storage::tempElement(int idx, llvm::Value *indIdx) +{ + Value *elem = element(TempsArg, idx, indIdx); + + LoadInst *load = new LoadInst(elem, name("temp"), false, m_block); + load->setAlignment(8); + + return load; +} + +void Storage::setTempElement(int idx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = 0; + if (m_tempWriteMap[idx]) + templ = tempElement(idx); + val = maskWrite(val, mask, templ); + } + Value *elem = element(TempsArg, idx); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); + m_tempWriteMap[idx] = true; +} + +void Storage::setOutputElement(int dstIdx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = 0; + if (m_destWriteMap[dstIdx]) + templ = outputElement(dstIdx); + val = maskWrite(val, mask, templ); + } + + Value *elem = element(DestsArg, dstIdx); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); + m_destWriteMap[dstIdx] = true; +} + +llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ) +{ + llvm::Value *dst = templ; + if (!dst) + dst = Constant::getNullValue(m_floatVecType); + if ((mask & TGSI_WRITEMASK_X)) { + llvm::Value *x = new ExtractElementInst(src, unsigned(0), + name("x"), m_block); + dst = new InsertElementInst(dst, x, unsigned(0), + name("dstx"), m_block); + } + if ((mask & TGSI_WRITEMASK_Y)) { + llvm::Value *y = new ExtractElementInst(src, unsigned(1), + name("y"), m_block); + dst = new InsertElementInst(dst, y, unsigned(1), + name("dsty"), m_block); + } + if ((mask & TGSI_WRITEMASK_Z)) { + llvm::Value *z = new ExtractElementInst(src, unsigned(2), + name("z"), m_block); + dst = new InsertElementInst(dst, z, unsigned(2), + name("dstz"), m_block); + } + if ((mask & TGSI_WRITEMASK_W)) { + llvm::Value *w = new ExtractElementInst(src, unsigned(3), + name("w"), m_block); + dst = new InsertElementInst(dst, w, unsigned(3), + name("dstw"), m_block); + } + return dst; +} + +const char * Storage::name(const char *prefix) +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +int Storage::numConsts() const +{ + return m_numConsts; +} + +llvm::Value * Storage::addrElement(int idx) const +{ + Value *ret = m_addrs[idx]; + if (!ret) + return m_undefFloatVec; + return ret; +} + +void Storage::setAddrElement(int idx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = m_addrs[idx]; + val = maskWrite(val, mask, templ); + } + m_addrs[idx] = val; +} + +llvm::Value * Storage::extractIndex(llvm::Value *vec) +{ + llvm::Value *x = new ExtractElementInst(vec, unsigned(0), + name("x"), m_block); + return new FPToSIInst(x, IntegerType::get(32), name("intidx"), m_block); +} + +void Storage::setCurrentBlock(llvm::BasicBlock *block) +{ + m_block = block; +} + +llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx) +{ + Value *elem = element(DestsArg, idx, indIdx); + LoadInst *load = new LoadInst(elem, name("output"), false, m_block); + load->setAlignment(8); + + return load; +} + +llvm::Value * Storage::inputPtr() const +{ + return m_INPUT; +} + +void Storage::pushArguments(llvm::Value *input) +{ + m_argStack.push(m_INPUT); + + m_INPUT = input; +} + +void Storage::popArguments() +{ + m_INPUT = m_argStack.top(); + m_argStack.pop(); +} + +void Storage::pushTemps() +{ + m_extSwizzleVec = 0; +} + +void Storage::popTemps() +{ +} + +llvm::Value * Storage::immediateElement(int idx) +{ + return m_immediates[idx]; +} + +void Storage::addImmediate(float *val) +{ + std::vector vec(4); + vec[0] = ConstantFP::get(Type::FloatTy, APFloat(val[0])); + vec[1] = ConstantFP::get(Type::FloatTy, APFloat(val[1])); + vec[2] = ConstantFP::get(Type::FloatTy, APFloat(val[2])); + vec[3] = ConstantFP::get(Type::FloatTy, APFloat(val[3])); + m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); +} + + +llvm::Value * Storage::elemPtr(Args arg) +{ + std::vector indices; + indices.push_back(constantInt(0)); + indices.push_back(constantInt(static_cast(arg))); + GetElementPtrInst *getElem = new GetElementPtrInst(m_INPUT, + indices.begin(), + indices.end(), + name("input_ptr"), + m_block); + return new LoadInst(getElem, name("input_field"), false, m_block); +} + +llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx, + llvm::Value *indIdx ) +{ + GetElementPtrInst *getElem = 0; + + if (indIdx) { + getElem = new GetElementPtrInst(ptr, + BinaryOperator::create(Instruction::Add, + indIdx, + constantInt(idx), + name("add"), + m_block), + name("field"), + m_block); + } else { + getElem = new GetElementPtrInst(ptr, + constantInt(idx), + name("field"), + m_block); + } + return getElem; +} + +llvm::Value * Storage::element(Args arg, int idx, llvm::Value *indIdx ) +{ + Value *val = elemPtr(arg); + return elemIdx(val, idx, indIdx); +} + +void Storage::setKilElement(llvm::Value *val) +{ + std::vector indices; + indices.push_back(constantInt(0)); + indices.push_back(constantInt(static_cast(KilArg))); + GetElementPtrInst *elem = new GetElementPtrInst(m_INPUT, + indices.begin(), + indices.end(), + name("kil_ptr"), + m_block); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); +} + +#endif //MESA_LLVM + + diff --git a/src/gallium/aux/llvm/storage.h b/src/gallium/aux/llvm/storage.h new file mode 100644 index 0000000000..8574f7554e --- /dev/null +++ b/src/gallium/aux/llvm/storage.h @@ -0,0 +1,133 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef STORAGE_H +#define STORAGE_H + +#include +#include +#include +#include + +namespace llvm { + class BasicBlock; + class Constant; + class ConstantInt; + class LoadInst; + class Value; + class VectorType; +} + +class Storage +{ +public: + Storage(llvm::BasicBlock *block, + llvm::Value *input); + + llvm::Value *inputPtr() const; + + void setCurrentBlock(llvm::BasicBlock *block); + + llvm::ConstantInt *constantInt(int); + llvm::Constant *shuffleMask(int vec); + llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); + llvm::Value *constElement(int idx, llvm::Value *indIdx =0); + llvm::Value *outputElement(int idx, llvm::Value *indIdx =0); + llvm::Value *tempElement(int idx, llvm::Value *indIdx =0); + llvm::Value *immediateElement(int idx); + + void setOutputElement(int dstIdx, llvm::Value *val, int mask); + void setTempElement(int idx, llvm::Value *val, int mask); + + llvm::Value *addrElement(int idx) const; + void setAddrElement(int idx, llvm::Value *val, int mask); + + void setKilElement(llvm::Value *val); + + llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); + + llvm::Value *extractIndex(llvm::Value *vec); + + int numConsts() const; + + void pushArguments(llvm::Value *input); + void popArguments(); + void pushTemps(); + void popTemps(); + + void addImmediate(float *val); + +private: + llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); + const char *name(const char *prefix); + + enum Args { + DestsArg = 0, + InputsArg = 1, + TempsArg = 2, + ConstsArg = 3, + KilArg = 4 + }; + llvm::Value *elemPtr(Args arg); + llvm::Value *elemIdx(llvm::Value *ptr, int idx, + llvm::Value *indIdx = 0); + llvm::Value *element(Args arg, int idx, llvm::Value *indIdx = 0); + +private: + llvm::BasicBlock *m_block; + llvm::Value *m_INPUT; + + std::map m_constInts; + std::map m_intVecs; + std::vector m_addrs; + std::vector m_immediates; + + llvm::VectorType *m_floatVecType; + llvm::VectorType *m_intVecType; + + char m_name[32]; + int m_idx; + + int m_numConsts; + + std::map m_destWriteMap; + std::map m_tempWriteMap; + + llvm::Value *m_undefFloatVec; + llvm::Value *m_undefIntVec; + llvm::Value *m_extSwizzleVec; + + std::stack m_argStack; + std::stack > m_tempStack; +}; + +#endif diff --git a/src/gallium/aux/llvm/storagesoa.cpp b/src/gallium/aux/llvm/storagesoa.cpp new file mode 100644 index 0000000000..ed0674a96f --- /dev/null +++ b/src/gallium/aux/llvm/storagesoa.cpp @@ -0,0 +1,389 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "storagesoa.h" + +#include "gallivm_p.h" + +#include "pipe/p_shader_tokens.h" +#include "pipe/p_debug.h" + +#include +#include +#include + +#include +#include +#include +#include +#include + +using namespace llvm; + + +StorageSoa::StorageSoa(llvm::BasicBlock *block, + llvm::Value *input, + llvm::Value *output, + llvm::Value *consts, + llvm::Value *temps) + : m_block(block), + m_input(input), + m_output(output), + m_consts(consts), + m_temps(temps), + m_immediates(0), + m_idx(0) +{ +} + +void StorageSoa::addImmediate(float *vec) +{ + std::vector vals(4); + vals[0] = vec[0]; + vals[1] = vec[1]; + vals[2] = vec[2]; + vals[3] = vec[3]; + m_immediatesToFlush.push_back(vals); +} + +void StorageSoa::declareImmediates() +{ + if (m_immediatesToFlush.empty()) + return; + + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorChannels = ArrayType::get(vectorType, 4); + ArrayType *arrayType = ArrayType::get(vectorChannels, m_immediatesToFlush.size()); + + m_immediates = new GlobalVariable( + /*Type=*/arrayType, + /*isConstant=*/false, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name("immediates"), + currentModule()); + + std::vector arrayVals; + for (unsigned int i = 0; i < m_immediatesToFlush.size(); ++i) { + std::vector vec = m_immediatesToFlush[i]; + std::vector vals(4); + std::vector channelArray; + + vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0]; + llvm::Constant *xChannel = createConstGlobalVector(vals); + + vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1]; + llvm::Constant *yChannel = createConstGlobalVector(vals); + + vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2]; + llvm::Constant *zChannel = createConstGlobalVector(vals); + + vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3]; + llvm::Constant *wChannel = createConstGlobalVector(vals); + channelArray.push_back(xChannel); + channelArray.push_back(yChannel); + channelArray.push_back(zChannel); + channelArray.push_back(wChannel); + Constant *constChannels = ConstantArray::get(vectorChannels, + channelArray); + arrayVals.push_back(constChannels); + } + Constant *constArray = ConstantArray::get(arrayType, arrayVals); + m_immediates->setInitializer(constArray); + + m_immediatesToFlush.clear(); +} + +llvm::Value *StorageSoa::addrElement(int idx) const +{ + std::map::const_iterator itr = m_addresses.find(idx); + if (itr == m_addresses.end()) { + debug_printf("Trying to access invalid shader 'address'\n"); + return 0; + } + llvm::Value * res = (*itr).second; + + res = new LoadInst(res, name("addr"), false, m_block); + + return res; +} + +std::vector StorageSoa::inputElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_input, idx, 0); + res[1] = element(m_input, idx, 1); + res[2] = element(m_input, idx, 2); + res[3] = element(m_input, idx, 3); + + return res; +} + +std::vector StorageSoa::constElement(llvm::Value *idx) +{ + std::vector res(4); + llvm::Value *xChannel, *yChannel, *zChannel, *wChannel; + + xChannel = elementPointer(m_consts, idx, 0); + yChannel = elementPointer(m_consts, idx, 1); + zChannel = elementPointer(m_consts, idx, 2); + wChannel = elementPointer(m_consts, idx, 3); + + res[0] = alignedArrayLoad(xChannel); + res[1] = alignedArrayLoad(yChannel); + res[2] = alignedArrayLoad(zChannel); + res[3] = alignedArrayLoad(wChannel); + + return res; +} + +std::vector StorageSoa::outputElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_output, idx, 0); + res[1] = element(m_output, idx, 1); + res[2] = element(m_output, idx, 2); + res[3] = element(m_output, idx, 3); + + return res; +} + +std::vector StorageSoa::tempElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_temps, idx, 0); + res[1] = element(m_temps, idx, 1); + res[2] = element(m_temps, idx, 2); + res[3] = element(m_temps, idx, 3); + + return res; +} + +std::vector StorageSoa::immediateElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_immediates, idx, 0); + res[1] = element(m_immediates, idx, 1); + res[2] = element(m_immediates, idx, 2); + res[3] = element(m_immediates, idx, 3); + + return res; +} + +llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index, + int channel) const +{ + std::vector indices; + if (m_immediates == ptr) + indices.push_back(constantInt(0)); + indices.push_back(index); + indices.push_back(constantInt(channel)); + + GetElementPtrInst *getElem = new GetElementPtrInst(ptr, + indices.begin(), + indices.end(), + name("ptr"), + m_block); + return getElem; +} + +llvm::Value * StorageSoa::element(llvm::Value *ptr, llvm::Value *index, + int channel) const +{ + llvm::Value *res = elementPointer(ptr, index, channel); + LoadInst *load = new LoadInst(res, name("element"), false, m_block); + //load->setAlignment(8); + return load; +} + +const char * StorageSoa::name(const char *prefix) const +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::ConstantInt * StorageSoa::constantInt(int idx) const +{ + if (m_constInts.find(idx) != m_constInts.end()) { + return m_constInts[idx]; + } + ConstantInt *constInt = ConstantInt::get(APInt(32, idx)); + m_constInts[idx] = constInt; + return constInt; +} + +llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + PointerType *vectorPtr = PointerType::get(vectorType, 0); + + CastInst *cast = new BitCastInst(val, vectorPtr, name("toVector"), m_block); + LoadInst *load = new LoadInst(cast, name("alignLoad"), false, m_block); + load->setAlignment(8); + return load; +} + +llvm::Module * StorageSoa::currentModule() const +{ + if (!m_block || !m_block->getParent()) + return 0; + + return m_block->getParent()->getParent(); +} + +llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector &vec) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + std::vector immValues; + ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0])); + ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1])); + ConstantFP *constz = ConstantFP::get(Type::FloatTy, APFloat(vec[2])); + ConstantFP *constw = ConstantFP::get(Type::FloatTy, APFloat(vec[3])); + immValues.push_back(constx); + immValues.push_back(consty); + immValues.push_back(constz); + immValues.push_back(constw); + Constant *constVector = ConstantVector::get(vectorType, immValues); + + return constVector; +} + +std::vector StorageSoa::load(Argument type, int idx, int swizzle, + llvm::Value *indIdx) +{ + std::vector val(4); + + //if we have an indirect index, always use that + // if not use the integer offset to create one + llvm::Value *realIndex = 0; + if (indIdx) + realIndex = indIdx; + else + realIndex = constantInt(idx); + debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx); + + switch(type) { + case Input: + val = inputElement(realIndex); + break; + case Output: + val = outputElement(realIndex); + break; + case Temp: + val = tempElement(realIndex); + break; + case Const: + val = constElement(realIndex); + break; + case Immediate: + val = immediateElement(realIndex); + break; + case Address: + debug_printf("Address not handled in the load phase!\n"); + assert(0); + break; + } + if (!gallivm_is_swizzle(swizzle)) + return val; + + std::vector res(4); + + res[0] = val[gallivm_x_swizzle(swizzle)]; + res[1] = val[gallivm_y_swizzle(swizzle)]; + res[2] = val[gallivm_z_swizzle(swizzle)]; + res[3] = val[gallivm_w_swizzle(swizzle)]; + return res; +} + +void StorageSoa::store(Argument type, int idx, const std::vector &val, + int mask) +{ + llvm::Value *out = 0; + switch(type) { + case Output: + out = m_output; + break; + case Temp: + out = m_temps; + break; + case Input: + out = m_input; + break; + case Address: { + llvm::Value *addr = m_addresses[idx]; + if (!addr) { + addAddress(idx); + addr = m_addresses[idx]; + assert(addr); + } + new StoreInst(val[0], addr, false, m_block); + return; + break; + } + default: + debug_printf("Can't save output of this type: %d !\n", type); + assert(0); + break; + } + llvm::Value *realIndex = constantInt(idx); + if ((mask & TGSI_WRITEMASK_X)) { + llvm::Value *xChannel = elementPointer(out, realIndex, 0); + new StoreInst(val[0], xChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_Y)) { + llvm::Value *yChannel = elementPointer(out, realIndex, 1); + new StoreInst(val[1], yChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_Z)) { + llvm::Value *zChannel = elementPointer(out, realIndex, 2); + new StoreInst(val[2], zChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_W)) { + llvm::Value *wChannel = elementPointer(out, realIndex, 3); + new StoreInst(val[3], wChannel, false, m_block); + } +} + +void StorageSoa::addAddress(int idx) +{ + GlobalVariable *val = new GlobalVariable( + /*Type=*/IntegerType::get(32), + /*isConstant=*/false, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name("address"), + currentModule()); + val->setInitializer(Constant::getNullValue(IntegerType::get(32))); + + debug_printf("adding to %d\n", idx); + m_addresses[idx] = val; +} diff --git a/src/gallium/aux/llvm/storagesoa.h b/src/gallium/aux/llvm/storagesoa.h new file mode 100644 index 0000000000..6443351f27 --- /dev/null +++ b/src/gallium/aux/llvm/storagesoa.h @@ -0,0 +1,111 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef STORAGESOA_H +#define STORAGESOA_H + +#include +#include +#include + +namespace llvm { + class BasicBlock; + class Constant; + class ConstantInt; + class GlobalVariable; + class LoadInst; + class Value; + class VectorType; + class Module; +} + +class StorageSoa +{ +public: + enum Argument { + Input, + Output, + Temp, + Const, + Immediate, + Address + }; +public: + StorageSoa(llvm::BasicBlock *block, + llvm::Value *input, + llvm::Value *output, + llvm::Value *consts, + llvm::Value *temps); + + + std::vector load(Argument type, int idx, int swizzle, + llvm::Value *indIdx =0); + void store(Argument type, int idx, const std::vector &val, + int mask); + + void addImmediate(float *vec); + void declareImmediates(); + + void addAddress(int idx); + + llvm::Value * addrElement(int idx) const; + + llvm::ConstantInt *constantInt(int) const; +private: + llvm::Value *elementPointer(llvm::Value *ptr, llvm::Value *indIdx, + int channel) const; + llvm::Value *element(llvm::Value *ptr, llvm::Value *idx, + int channel) const; + const char *name(const char *prefix) const; + llvm::Value *alignedArrayLoad(llvm::Value *val); + llvm::Module *currentModule() const; + llvm::Constant *createConstGlobalVector(const std::vector &vec); + + std::vector inputElement(llvm::Value *indIdx); + std::vector constElement(llvm::Value *indIdx); + std::vector outputElement(llvm::Value *indIdx); + std::vector tempElement(llvm::Value *indIdx); + std::vector immediateElement(llvm::Value *indIdx); +private: + llvm::BasicBlock *m_block; + + llvm::Value *m_input; + llvm::Value *m_output; + llvm::Value *m_consts; + llvm::Value *m_temps; + llvm::GlobalVariable *m_immediates; + + std::map m_addresses; + + std::vector > m_immediatesToFlush; + + mutable std::map m_constInts; + mutable char m_name[32]; + mutable int m_idx; +}; + +#endif diff --git a/src/gallium/aux/llvm/tgsitollvm.cpp b/src/gallium/aux/llvm/tgsitollvm.cpp new file mode 100644 index 0000000000..0de595e678 --- /dev/null +++ b/src/gallium/aux/llvm/tgsitollvm.cpp @@ -0,0 +1,1221 @@ +#include "tgsitollvm.h" + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "storage.h" +#include "instructions.h" +#include "storagesoa.h" +#include "instructionssoa.h" + +#include "pipe/p_shader_tokens.h" + +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/exec/tgsi_exec.h" +#include "pipe/tgsi/util/tgsi_util.h" +#include "pipe/tgsi/util/tgsi_build.h" +#include "pipe/tgsi/util/tgsi_dump.h" + + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + + +#include +#include +#include + +using namespace llvm; + +static inline FunctionType *vertexShaderFunctionType() +{ + //Function takes three arguments, + // the calling code has to make sure the types it will + // pass are castable to the following: + // [4 x <4 x float>] inputs, + // [4 x <4 x float>] output, + // [4 x [4 x float]] consts, + // [4 x <4 x float>] temps + + std::vector funcArgs; + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorArray = ArrayType::get(vectorType, 4); + PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0); + + ArrayType *floatArray = ArrayType::get(Type::FloatTy, 4); + ArrayType *constsArray = ArrayType::get(floatArray, 4); + PointerType *constsArrayPtr = PointerType::get(constsArray, 0); + + funcArgs.push_back(vectorArrayPtr);//inputs + funcArgs.push_back(vectorArrayPtr);//output + funcArgs.push_back(constsArrayPtr);//consts + funcArgs.push_back(vectorArrayPtr);//temps + + FunctionType *functionType = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/funcArgs, + /*isVarArg=*/false); + + return functionType; +} + +static inline void +add_interpolator(struct gallivm_ir *ir, + struct gallivm_interpolate *interp) +{ + ir->interpolators[ir->num_interp] = *interp; + ++ir->num_interp; +} + +static void +translate_declaration(struct gallivm_ir *prog, + llvm::Module *module, + Storage *storage, + struct tgsi_full_declaration *decl, + struct tgsi_full_declaration *fd) +{ + if (decl->Declaration.File == TGSI_FILE_INPUT) { + unsigned first, last, mask; + uint interp_method; + + assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if (first == 0) { + mask &= ~TGSI_WRITEMASK_XY; + if (mask == TGSI_WRITEMASK_NONE) { + first++; + if (first > last) { + return; + } + } + } + + interp_method = decl->Interpolation.Interpolate; + + if (mask == TGSI_WRITEMASK_XYZW) { + unsigned i, j; + + for (i = first; i <= last; i++) { + for (j = 0; j < NUM_CHANNELS; j++) { + //interp( mach, i, j ); + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } else { + unsigned i, j; + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } + } + } +} + +static void +translate_declarationir(struct gallivm_ir *, + llvm::Module *, + StorageSoa *storage, + struct tgsi_full_declaration *decl, + struct tgsi_full_declaration *) +{ + if (decl->Declaration.File == TGSI_FILE_ADDRESS) { + int idx = decl->u.DeclarationRange.First; + storage->addAddress(idx); + } +} + +static void +translate_immediate(Storage *storage, + struct tgsi_full_immediate *imm) +{ + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert(0); + } + } + storage->addImmediate(vec); +} + + +static void +translate_immediateir(StorageSoa *storage, + struct tgsi_full_immediate *imm) +{ + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert(0); + } + } + storage->addImmediate(vec); +} + +static inline int +swizzleInt(struct tgsi_full_src_register *src) +{ + int swizzle = 0; + int start = 1000; + + for (int k = 0; k < 4; ++k) { + swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; + start /= 10; + } + return swizzle; +} + +static inline llvm::Value * +swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, + Storage *storage) +{ + int swizzle = swizzleInt(src); + + if (gallivm_is_swizzle(swizzle)) { + /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ + val = storage->shuffleVector(val, swizzle); + } + return val; +} + +static void +translate_instruction(llvm::Module *module, + Storage *storage, + Instructions *instr, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi, + unsigned instno) +{ + llvm::Value *inputs[4]; + inputs[0] = 0; + inputs[1] = 0; + inputs[2] = 0; + inputs[3] = 0; + + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + llvm::Value *val = 0; + llvm::Value *indIdx = 0; + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + indIdx = storage->extractIndex(indIdx); + } + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->constElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->inputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->tempElement(src->SrcRegister.Index); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->outputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->immediateElement(src->SrcRegister.Index); + } else { + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); + return; + } + + inputs[i] = swizzleVector(val, src, storage); + } + + /*if (inputs[0]) + instr->printVector(inputs[0]); + if (inputs[1]) + instr->printVector(inputs[1]);*/ + llvm::Value *out = 0; + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); + } + break; + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } + break; + case TGSI_OPCODE_LIT: { + out = instr->lit(inputs[0]); + } + break; + case TGSI_OPCODE_RCP: { + out = instr->rcp(inputs[0]); + } + break; + case TGSI_OPCODE_RSQ: { + out = instr->rsq(inputs[0]); + } + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP3: { + out = instr->dp3(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP4: { + out = instr->dp4(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DST: { + out = instr->dst(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MIN: { + out = instr->min(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAX: { + out = instr->max(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SLT: { + out = instr->slt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SGE: { + out = instr->sge(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SUB: { + out = instr->sub(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_LERP: { + out = instr->lerp(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: { + out = instr->frc(inputs[0]); + } + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: { + out = instr->floor(inputs[0]); + } + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: { + out = instr->ex2(inputs[0]); + } + break; + case TGSI_OPCODE_LOGBASE2: { + out = instr->lg2(inputs[0]); + } + break; + case TGSI_OPCODE_POWER: { + out = instr->pow(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_CROSSPRODUCT: { + out = instr->cross(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: { + out = instr->abs(inputs[0]); + } + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: { + out = instr->dph(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_COS: { + out = instr->cos(inputs[0]); + } + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: { + out = instr->kilp(inputs[0]); + storage->setKilElement(out); + return; + } + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: { + out = instr->sgt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SIN: { + out = instr->sin(inputs[0]); + } + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: { + instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr()); + return; + } + break; + case TGSI_OPCODE_RET: { + instr->end(); + return; + } + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: { + out = instr->cmp(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SCS: { + out = instr->scs(inputs[0]); + } + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: { + instr->brk(); + return; + } + break; + case TGSI_OPCODE_IF: { + instr->ifop(inputs[0]); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: { + instr->elseop(); + storage->setCurrentBlock(instr->currentBlock()); + return; //only state update + } + break; + case TGSI_OPCODE_ENDIF: { + instr->endif(); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: { + out = instr->trunc(inputs[0]); + } + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: { + instr->beginLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_BGNSUB: { + instr->bgnSub(instno); + storage->setCurrentBlock(instr->currentBlock()); + storage->pushTemps(); + return; + } + break; + case TGSI_OPCODE_ENDLOOP2: { + instr->endLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_ENDSUB: { + instr->endSub(); + storage->setCurrentBlock(instr->currentBlock()); + storage->popArguments(); + storage->popTemps(); + return; + } + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + instr->end(); + return; + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + if (!out) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } + + /* # not sure if we need this */ + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + /*TXT( "_SAT" );*/ + break; + case TGSI_SAT_MINUS_PLUS_ONE: + /*TXT( "_SAT[-1,1]" );*/ + break; + default: + assert( 0 ); + } + + /* store results */ + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); + } + } +} + + +static void +translate_instructionir(llvm::Module *module, + StorageSoa *storage, + InstructionsSoa *instr, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi, + unsigned instno) +{ + std::vector< std::vector > inputs(inst->Instruction.NumSrcRegs); + + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + std::vector val; + llvm::Value *indIdx = 0; + int swizzle = swizzleInt(src); + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + } + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->load(StorageSoa::Const, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->load(StorageSoa::Input, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->load(StorageSoa::Temp, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->load(StorageSoa::Output, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->load(StorageSoa::Immediate, + src->SrcRegister.Index, swizzle, indIdx); + } else { + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); + return; + } + + inputs[i] = val; + } + + std::vector out(4); + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); + } + break; + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } + break; + case TGSI_OPCODE_LIT: { + } + break; + case TGSI_OPCODE_RCP: { + } + break; + case TGSI_OPCODE_RSQ: { + } + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP3: { + } + break; + case TGSI_OPCODE_DP4: { + } + break; + case TGSI_OPCODE_DST: { + } + break; + case TGSI_OPCODE_MIN: { + } + break; + case TGSI_OPCODE_MAX: { + } + break; + case TGSI_OPCODE_SLT: { + } + break; + case TGSI_OPCODE_SGE: { + } + break; + case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SUB: { + } + break; + case TGSI_OPCODE_LERP: { + } + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: { + } + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: { + } + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: { + } + break; + case TGSI_OPCODE_LOGBASE2: { + } + break; + case TGSI_OPCODE_POWER: { + } + break; + case TGSI_OPCODE_CROSSPRODUCT: { + } + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: { + } + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: { + } + break; + case TGSI_OPCODE_COS: { + } + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: { + } + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: { + } + break; + case TGSI_OPCODE_SIN: { + } + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: { + } + break; + case TGSI_OPCODE_RET: { + } + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: { + } + break; + case TGSI_OPCODE_SCS: { + } + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: { + } + break; + case TGSI_OPCODE_IF: { + } + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: { + } + break; + case TGSI_OPCODE_ENDIF: { + } + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: { + } + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: { + } + break; + case TGSI_OPCODE_BGNSUB: { + } + break; + case TGSI_OPCODE_ENDLOOP2: { + } + break; + case TGSI_OPCODE_ENDSUB: { + } + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + instr->end(); + return; + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + if (!out[0]) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } + + /* store results */ + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->store(StorageSoa::Output, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->store(StorageSoa::Temp, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->store(StorageSoa::Address, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); + } + } +} + +llvm::Module * +tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) +{ + llvm::Module *mod = new Module("shader"); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned instno = 0; + Function* shader = mod->getFunction("execute_shader"); + std::ostringstream stream; + if (ir->type == GALLIVM_VS) { + stream << "vs_shader"; + } else { + stream << "fs_shader"; + } + stream << ir->id; + std::string func_name = stream.str(); + shader->setName(func_name.c_str()); + + Function::arg_iterator args = shader->arg_begin(); + Value *ptr_INPUT = args++; + ptr_INPUT->setName("input"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + + tgsi_parse_init(&parse, tokens); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + Storage storage(label_entry, ptr_INPUT); + Instructions instr(mod, shader, label_entry, &storage); + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declaration(ir, mod, &storage, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediate(&storage, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + translate_instruction(mod, &storage, &instr, + &parse.FullToken.FullInstruction, + &fi, instno); + ++instno; + break; + + default: + assert(0); + } + } + + tgsi_parse_free(&parse); + + ir->num_consts = storage.numConsts(); + return mod; +} + +llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, + const struct tgsi_token *tokens) +{ + llvm::Module *mod = new Module("shader"); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned instno = 0; + std::ostringstream stream; + if (ir->type == GALLIVM_VS) { + stream << "vs_shader"; + } else { + stream << "fs_shader"; + } + //stream << ir->id; + std::string func_name = stream.str(); + Function *shader = llvm::cast(mod->getOrInsertFunction( + func_name.c_str(), + vertexShaderFunctionType())); + + Function::arg_iterator args = shader->arg_begin(); + Value *input = args++; + input->setName("inputs"); + Value *output = args++; + output->setName("outputs"); + Value *consts = args++; + consts->setName("consts"); + Value *temps = args++; + temps->setName("temps"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + + tgsi_parse_init(&parse, tokens); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + StorageSoa storage(label_entry, input, output, consts, temps); + InstructionsSoa instr(mod, shader, label_entry, &storage); + + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declarationir(ir, mod, &storage, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediateir(&storage, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + storage.declareImmediates(); + translate_instructionir(mod, &storage, &instr, + &parse.FullToken.FullInstruction, + &fi, instno); + ++instno; + break; + + default: + assert(0); + } + } + + tgsi_parse_free(&parse); + + return mod; +} diff --git a/src/gallium/aux/llvm/tgsitollvm.h b/src/gallium/aux/llvm/tgsitollvm.h new file mode 100644 index 0000000000..7ada04d629 --- /dev/null +++ b/src/gallium/aux/llvm/tgsitollvm.h @@ -0,0 +1,20 @@ +#ifndef TGSITOLLVM_H +#define TGSITOLLVM_H + + +namespace llvm { + class Module; +} + +struct gallivm_ir; +struct tgsi_token; + + +llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, + const struct tgsi_token *tokens); + + +llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, + const struct tgsi_token *tokens); + +#endif diff --git a/src/gallium/aux/pipebuffer/Makefile b/src/gallium/aux/pipebuffer/Makefile new file mode 100644 index 0000000000..75764a9a18 --- /dev/null +++ b/src/gallium/aux/pipebuffer/Makefile @@ -0,0 +1,23 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = pipebuffer + +DRIVER_SOURCES = \ + pb_buffer_fenced.c \ + pb_buffer_malloc.c \ + pb_bufmgr_fenced.c \ + pb_bufmgr_mm.c \ + pb_bufmgr_pool.c \ + pb_winsys.c + +C_SOURCES = \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../Makefile.template + +symlinks: + diff --git a/src/gallium/aux/pipebuffer/linked_list.h b/src/gallium/aux/pipebuffer/linked_list.h new file mode 100644 index 0000000000..e99817fb13 --- /dev/null +++ b/src/gallium/aux/pipebuffer/linked_list.h @@ -0,0 +1,91 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND. USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + **************************************************************************/ + +/** + * \file + * List macros heavily inspired by the Linux kernel + * list handling. No list looping yet. + * + * Is not threadsafe, so common operations need to + * be protected using an external mutex. + */ + +#ifndef LINKED_LIST_H_ +#define LINKED_LIST_H_ + + +#include + + +struct list_head +{ + struct list_head *prev; + struct list_head *next; +}; + + +#define LIST_INITHEAD(__item) \ + do { \ + (__item)->prev = (__item); \ + (__item)->next = (__item); \ + } while (0) + +#define LIST_ADD(__item, __list) \ + do { \ + (__item)->prev = (__list); \ + (__item)->next = (__list)->next; \ + (__list)->next->prev = (__item); \ + (__list)->next = (__item); \ + } while (0) + +#define LIST_ADDTAIL(__item, __list) \ + do { \ + (__item)->next = (__list); \ + (__item)->prev = (__list)->prev; \ + (__list)->prev->next = (__item); \ + (__list)->prev = (__item); \ + } while(0) + +#define LIST_DEL(__item) \ + do { \ + (__item)->prev->next = (__item)->next; \ + (__item)->next->prev = (__item)->prev; \ + } while(0) + +#define LIST_DELINIT(__item) \ + do { \ + (__item)->prev->next = (__item)->next; \ + (__item)->next->prev = (__item)->prev; \ + (__item)->next = (__item); \ + (__item)->prev = (__item); \ + } while(0) + +#define LIST_ENTRY(__type, __item, __field) \ + ((__type *)(((char *)(__item)) - offsetof(__type, __field))) + + +#endif /*LINKED_LIST_H_*/ diff --git a/src/gallium/aux/pipebuffer/pb_buffer.h b/src/gallium/aux/pipebuffer/pb_buffer.h new file mode 100644 index 0000000000..97beb5f72a --- /dev/null +++ b/src/gallium/aux/pipebuffer/pb_buffer.h @@ -0,0 +1,202 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Generic code for buffers. + * + * Behind a pipe buffle handle there can be DMA buffers, client (or user) + * buffers, regular malloced buffers, etc. This file provides an abstract base + * buffer handle that allows the driver to cope with all those kinds of buffers + * in a more flexible way. + * + * There is no obligation of a winsys driver to use this library. And a pipe + * driver should be completly agnostic about it. + * + * \author José Fonseca + */ + +#ifndef PB_BUFFER_H_ +#define PB_BUFFER_H_ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" + + +struct pb_vtbl; + +/** + * Buffer description. + * + * Used when allocating the buffer. + */ +struct pb_desc +{ + unsigned alignment; + unsigned usage; +}; + + +/** + * Base class for all pb_* buffers. + */ +struct pb_buffer +{ + struct pipe_buffer base; + + /** + * Pointer to the virtual function table. + * + * Avoid accessing this table directly. Use the inline functions below + * instead to avoid mistakes. + */ + const struct pb_vtbl *vtbl; +}; + + +/** + * Virtual function table for the buffer storage operations. + * + * Note that creation is not done through this table. + */ +struct pb_vtbl +{ + void (*destroy)( struct pb_buffer *buf ); + + /** + * Map the entire data store of a buffer object into the client's address. + * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. + */ + void *(*map)( struct pb_buffer *buf, + unsigned flags ); + + void (*unmap)( struct pb_buffer *buf ); + + /** + * Get the base buffer and the offset. + * + * A buffer can be subdivided in smaller buffers. This method should return + * the underlaying buffer, and the relative offset. + * + * Buffers without an underlaying base buffer should return themselves, with + * a zero offset. + * + * Note that this will increase the reference count of the base buffer. + */ + void (*get_base_buffer)( struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset ); +}; + + +static INLINE struct pipe_buffer * +pb_pipe_buffer( struct pb_buffer *pbuf ) +{ + assert(pbuf); + return &pbuf->base; +} + + +static INLINE struct pb_buffer * +pb_buffer( struct pipe_buffer *buf ) +{ + assert(buf); + /* Could add a magic cookie check on debug builds. + */ + return (struct pb_buffer *)buf; +} + + +/* Accessor functions for pb->vtbl: + */ +static INLINE void * +pb_map(struct pb_buffer *buf, + unsigned flags) +{ + assert(buf); + return buf->vtbl->map(buf, flags); +} + + +static INLINE void +pb_unmap(struct pb_buffer *buf) +{ + assert(buf); + buf->vtbl->unmap(buf); +} + + +static INLINE void +pb_get_base_buffer( struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset ) +{ + buf->vtbl->get_base_buffer(buf, base_buf, offset); +} + + +static INLINE void +pb_destroy(struct pb_buffer *buf) +{ + assert(buf); + buf->vtbl->destroy(buf); +} + + +/* XXX: thread safety issues! + */ +static INLINE void +pb_reference(struct pb_buffer **dst, + struct pb_buffer *src) +{ + if (src) + src->base.refcount++; + + if (*dst && --(*dst)->base.refcount == 0) + pb_destroy( *dst ); + + *dst = src; +} + + +/** + * Malloc-based buffer to store data that can't be used by the graphics + * hardware. + */ +struct pb_buffer * +pb_malloc_buffer_create(size_t size, + const struct pb_desc *desc); + + +void +pb_init_winsys(struct pipe_winsys *winsys); + + +#endif /*PB_BUFFER_H_*/ diff --git a/src/gallium/aux/pipebuffer/pb_buffer_fenced.c b/src/gallium/aux/pipebuffer/pb_buffer_fenced.c new file mode 100644 index 0000000000..f4fc3f6d71 --- /dev/null +++ b/src/gallium/aux/pipebuffer/pb_buffer_fenced.c @@ -0,0 +1,299 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of fenced buffers. + * + * \author José Fonseca + * \author Thomas Hellström + */ + + +#include "linked_list.h" + +#include "p_compiler.h" +#include "p_debug.h" +#include "p_winsys.h" +#include "p_thread.h" +#include "p_util.h" + +#include "pb_buffer.h" +#include "pb_buffer_fenced.h" + +#ifndef __MSC__ +#include +#endif + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct fenced_buffer_list +{ + _glthread_Mutex mutex; + + struct pipe_winsys *winsys; + + size_t numDelayed; + size_t checkDelayed; + + struct list_head delayed; +}; + + +/** + * Wrapper around a pipe buffer which adds fencing and reference counting. + */ +struct fenced_buffer +{ + struct pb_buffer base; + + struct pb_buffer *buffer; + + struct pipe_fence_handle *fence; + + struct list_head head; + struct fenced_buffer_list *list; +}; + + +static INLINE struct fenced_buffer * +fenced_buffer(struct pb_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &fenced_buffer_vtbl); + return (struct fenced_buffer *)buf; +} + + + + +static void +_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait) +{ + struct pipe_winsys *winsys = fenced_list->winsys; + struct fenced_buffer *fenced_buf; + struct list_head *list, *prev; + int signaled = -1; + + list = fenced_list->delayed.next; + + if (fenced_list->numDelayed > 3) { + unsigned i; + + for (i = 0; i < fenced_list->numDelayed; i += 3) { + list = list->next; + } + } + + prev = list->prev; + for (; list != &fenced_list->delayed; list = prev, prev = list->prev) { + + fenced_buf = LIST_ENTRY(struct fenced_buffer, list, head); + + if (signaled != 0) { + if (wait) { + signaled = winsys->fence_finish(winsys, fenced_buf->fence, 0); + } + else { + signaled = winsys->fence_signalled(winsys, fenced_buf->fence, 0); + } + } + + if (signaled != 0) + /* XXX: we are assuming that buffers are freed in the same order they + * are fenced which may not always be true... + */ + break; + + winsys->fence_reference(winsys, &fenced_buf->fence, NULL); + + LIST_DEL(list); + fenced_list->numDelayed--; + + /* Do the delayed destroy: + */ + pb_reference(&fenced_buf->buffer, NULL); + FREE(fenced_buf); + } +} + + +static void +fenced_buffer_destroy(struct pb_buffer *buf) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + + if (fenced_buf->fence) { + LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); + fenced_list->numDelayed++; + } + else { + pb_reference(&fenced_buf->buffer, NULL); + FREE(fenced_buf); + } + + if ((fenced_list->numDelayed % fenced_list->checkDelayed) == 0) + _fenced_buffer_list_check_free(fenced_list, 0); +} + + +static void * +fenced_buffer_map(struct pb_buffer *buf, + unsigned flags) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + return pb_map(fenced_buf->buffer, flags); +} + + +static void +fenced_buffer_unmap(struct pb_buffer *buf) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + pb_unmap(fenced_buf->buffer); +} + + +static void +fenced_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + pb_get_base_buffer(fenced_buf->buffer, base_buf, offset); +} + + +const struct pb_vtbl +fenced_buffer_vtbl = { + fenced_buffer_destroy, + fenced_buffer_map, + fenced_buffer_unmap, + fenced_buffer_get_base_buffer +}; + + +struct pb_buffer * +fenced_buffer_create(struct fenced_buffer_list *fenced_list, + struct pb_buffer *buffer) +{ + struct fenced_buffer *buf; + + if(!buffer) + return NULL; + + buf = CALLOC_STRUCT(fenced_buffer); + if(!buf) + return NULL; + + buf->base.base.refcount = 1; + buf->base.base.alignment = buffer->base.alignment; + buf->base.base.usage = buffer->base.usage; + buf->base.base.size = buffer->base.size; + + buf->base.vtbl = &fenced_buffer_vtbl; + buf->buffer = buffer; + buf->list = fenced_list; + + return &buf->base; +} + + +void +buffer_fence(struct pb_buffer *buf, + struct pipe_fence_handle *fence) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + struct pipe_winsys *winsys = fenced_list->winsys; + + _glthread_LOCK_MUTEX(fenced_list->mutex); + winsys->fence_reference(winsys, &fenced_buf->fence, fence); + _glthread_UNLOCK_MUTEX(fenced_list->mutex); +} + + +struct fenced_buffer_list * +fenced_buffer_list_create(struct pipe_winsys *winsys) +{ + struct fenced_buffer_list *fenced_list; + + fenced_list = (struct fenced_buffer_list *)CALLOC(1, sizeof(*fenced_list)); + if (!fenced_list) + return NULL; + + fenced_list->winsys = winsys; + + LIST_INITHEAD(&fenced_list->delayed); + + fenced_list->numDelayed = 0; + + /* TODO: don't hard code this */ + fenced_list->checkDelayed = 5; + + _glthread_INIT_MUTEX(fenced_list->mutex); + + return fenced_list; +} + + +void +fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait) +{ + _glthread_LOCK_MUTEX(fenced_list->mutex); + _fenced_buffer_list_check_free(fenced_list, wait); + _glthread_UNLOCK_MUTEX(fenced_list->mutex); +} + + +void +fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) +{ + _glthread_LOCK_MUTEX(fenced_list->mutex); + + /* Wait on outstanding fences */ + while (fenced_list->numDelayed) { + _glthread_UNLOCK_MUTEX(fenced_list->mutex); + sched_yield(); + _fenced_buffer_list_check_free(fenced_list, 1); + _glthread_LOCK_MUTEX(fenced_list->mutex); + } + + _glthread_UNLOCK_MUTEX(fenced_list->mutex); + + FREE(fenced_list); +} + + diff --git a/src/gallium/aux/pipebuffer/pb_buffer_fenced.h b/src/gallium/aux/pipebuffer/pb_buffer_fenced.h new file mode 100644 index 0000000000..c40b9c75e1 --- /dev/null +++ b/src/gallium/aux/pipebuffer/pb_buffer_fenced.h @@ -0,0 +1,117 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer fencing. + * + * "Fenced buffers" is actually a misnomer. They should be referred as + * "fenceable buffers", i.e, buffers that can be fenced, but I couldn't find + * the word "fenceable" in the dictionary. + * + * A "fenced buffer" is a decorator around a normal buffer, which adds two + * special properties: + * - the ability for the destruction to be delayed by a fence; + * - reference counting. + * + * Usually DMA buffers have a life-time that will extend the life-time of its + * handle. The end-of-life is dictated by the fence signalling. + * + * Between the handle's destruction, and the fence signalling, the buffer is + * stored in a fenced buffer list. + * + * \author José Fonseca + */ + +#ifndef PB_BUFFER_FENCED_H_ +#define PB_BUFFER_FENCED_H_ + + +#include "pipe/p_debug.h" + + +struct pipe_winsys; +struct pipe_buffer; +struct pipe_fence_handle; + + +/** + * List of buffers which are awaiting fence signalling. + */ +struct fenced_buffer_list; + + +/** + * The fenced buffer's virtual function table. + * + * NOTE: Made public for debugging purposes. + */ +extern const struct pb_vtbl fenced_buffer_vtbl; + + +/** + * Create a fenced buffer list. + * + * See also fenced_bufmgr_create for a more convenient way to use this. + */ +struct fenced_buffer_list * +fenced_buffer_list_create(struct pipe_winsys *winsys); + + +/** + * Walk the fenced buffer list to check and free signalled buffers. + */ +void +fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait); + +void +fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list); + + +/** + * Wrap a buffer in a fenced buffer. + * + * NOTE: this will not increase the buffer reference count. + */ +struct pb_buffer * +fenced_buffer_create(struct fenced_buffer_list *fenced, + struct pb_buffer *buffer); + + +/** + * Set a buffer's fence. + * + * NOTE: Although it takes a generic pb_buffer argument, it will fail + * on everything but buffers returned by fenced_buffer_create. + */ +void +buffer_fence(struct pb_buffer *buf, + struct pipe_fence_handle *fence); + + +#endif /*PB_BUFFER_FENCED_H_*/ diff --git a/src/gallium/aux/pipebuffer/pb_buffer_malloc.c b/src/gallium/aux/pipebuffer/pb_buffer_malloc.c new file mode 100644 index 0000000000..9e8244f909 --- /dev/null +++ b/src/gallium/aux/pipebuffer/pb_buffer_malloc.c @@ -0,0 +1,127 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of malloc-based buffers to store data that can't be processed + * by the hardware. + * + * \author José Fonseca + */ + + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pb_buffer.h" + + +struct malloc_buffer +{ + struct pb_buffer base; + void *data; +}; + + +extern const struct pb_vtbl malloc_buffer_vtbl; + +static INLINE struct malloc_buffer * +malloc_buffer(struct pb_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &malloc_buffer_vtbl); + return (struct malloc_buffer *)buf; +} + + +static void +malloc_buffer_destroy(struct pb_buffer *buf) +{ + align_free(malloc_buffer(buf)->data); + FREE(buf); +} + + +static void * +malloc_buffer_map(struct pb_buffer *buf, + unsigned flags) +{ + return malloc_buffer(buf)->data; +} + + +static void +malloc_buffer_unmap(struct pb_buffer *buf) +{ + /* No-op */ +} + + +static void +malloc_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + *base_buf = buf; + *offset = 0; +} + + +const struct pb_vtbl +malloc_buffer_vtbl = { + malloc_buffer_destroy, + malloc_buffer_map, + malloc_buffer_unmap, + malloc_buffer_get_base_buffer +}; + + +struct pb_buffer * +pb_malloc_buffer_create(size_t size, + const struct pb_desc *desc) +{ + struct malloc_buffer *buf; + + /* TODO: do a single allocation */ + + buf = CALLOC_STRUCT(malloc_buffer); + if(!buf) + return NULL; + + buf->base.base.refcount = 1; + buf->base.base.alignment = desc->alignment; + buf->base.base.usage = desc->usage; + buf->base.base.size = size; + buf->base.vtbl = &malloc_buffer_vtbl; + + buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment); + if(!buf->data) { + align_free(buf); + return NULL; + } + + return &buf->base; +} diff --git a/src/gallium/aux/pipebuffer/pb_bufmgr.h b/src/gallium/aux/pipebuffer/pb_bufmgr.h new file mode 100644 index 0000000000..1ddf784c97 --- /dev/null +++ b/src/gallium/aux/pipebuffer/pb_bufmgr.h @@ -0,0 +1,126 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer management. + * + * A buffer manager does only one basic thing: it creates buffers. Actually, + * "buffer factory" would probably a more accurate description. + * + * You can chain buffer managers so that you can have a finer grained memory + * management and pooling. + * + * For example, for a simple batch buffer manager you would chain: + * - the native buffer manager, which provides DMA memory from the graphics + * memory space; + * - the pool buffer manager, which keep around a pool of equally sized buffers + * to avoid latency associated with the native buffer manager; + * - the fenced buffer manager, which will delay buffer destruction until the + * the moment the card finishing processing it. + * + * \author José Fonseca + */ + +#ifndef PB_BUFMGR_H_ +#define PB_BUFMGR_H_ + + +#include + + +struct pb_desc; +struct pipe_buffer; +struct pipe_winsys; + + +/** + * Abstract base class for all buffer managers. + */ +struct pb_manager +{ + /* XXX: we will likely need more allocation flags */ + struct pb_buffer * + (*create_buffer)( struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc); + + void + (*destroy)( struct pb_manager *mgr ); +}; + + +/** + * Static buffer pool manager. + * + * Manages the allocation of equally sized buffers. It does so by allocating + * a single big buffer and divide it equally sized buffers. + * + * It is meant to manage the allocation of batch buffer pools. + */ +struct pb_manager * +pool_bufmgr_create(struct pb_manager *provider, + size_t n, size_t size, + const struct pb_desc *desc); + + +/** + * Wraper around the old memory manager. + * + * It managers buffers of different sizes. It does so by allocating a buffer + * with the size of the heap, and then using the old mm memory manager to manage + * that heap. + */ +struct pb_manager * +mm_bufmgr_create(struct pb_manager *provider, + size_t size, size_t align2); + +/** + * Same as mm_bufmgr_create. + * + * Buffer will be release when the manager is destroyed. + */ +struct pb_manager * +mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, + size_t size, size_t align2); + + +/** + * Fenced buffer manager. + * + * This manager is just meant for convenience. It wraps the buffers returned + * by another manager in fenced buffers, so that + * + * NOTE: the buffer manager that provides the buffers will be destroyed + * at the same time. + */ +struct pb_manager * +fenced_bufmgr_create(struct pb_manager *provider, + struct pipe_winsys *winsys); + + +#endif /*PB_BUFMGR_H_*/ diff --git a/src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c new file mode 100644 index 0000000000..c535d3276c --- /dev/null +++ b/src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c @@ -0,0 +1,131 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * \file + * A buffer manager that wraps buffers in fenced buffers. + * + * \author José Fonseca + */ + + +#include "p_debug.h" +#include "p_util.h" + +#include "pb_buffer.h" +#include "pb_buffer_fenced.h" +#include "pb_bufmgr.h" + + +struct fenced_pb_manager +{ + struct pb_manager base; + + struct pb_manager *provider; + + struct fenced_buffer_list *fenced_list; +}; + + +static INLINE struct fenced_pb_manager * +fenced_pb_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct fenced_pb_manager *)mgr; +} + + +static struct pb_buffer * +fenced_bufmgr_create_buffer(struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc) +{ + struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); + struct pb_buffer *buf; + struct pb_buffer *fenced_buf; + + /* check for free buffers before allocating new ones */ + fenced_buffer_list_check_free(fenced_mgr->fenced_list, 0); + + buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); + if(!buf) { + /* try harder to get a buffer */ + fenced_buffer_list_check_free(fenced_mgr->fenced_list, 1); + + buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); + if(!buf) { + /* give up */ + return NULL; + } + } + + fenced_buf = fenced_buffer_create(fenced_mgr->fenced_list, buf); + if(!fenced_buf) { + assert(buf->base.refcount == 1); + pb_destroy(buf); + } + + return fenced_buf; +} + + +static void +fenced_bufmgr_destroy(struct pb_manager *mgr) +{ + struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); + + fenced_buffer_list_destroy(fenced_mgr->fenced_list); + + fenced_mgr->provider->destroy(fenced_mgr->provider); + + FREE(fenced_mgr); +} + + +struct pb_manager * +fenced_bufmgr_create(struct pb_manager *provider, + struct pipe_winsys *winsys) +{ + struct fenced_pb_manager *fenced_mgr; + + fenced_mgr = (struct fenced_pb_manager *)CALLOC(1, sizeof(*fenced_mgr)); + if (!fenced_mgr) + return NULL; + + fenced_mgr->base.destroy = fenced_bufmgr_destroy; + fenced_mgr->base.create_buffer = fenced_bufmgr_create_buffer; + + fenced_mgr->provider = provider; + fenced_mgr->fenced_list = fenced_buffer_list_create(winsys); + if(!fenced_mgr->fenced_list) { + FREE(fenced_mgr); + return NULL; + } + + return &fenced_mgr->base; +} diff --git a/src/gallium/aux/pipebuffer/pb_bufmgr_mm.c b/src/gallium/aux/pipebuffer/pb_bufmgr_mm.c new file mode 100644 index 0000000000..8b1b51c0e2 --- /dev/null +++ b/src/gallium/aux/pipebuffer/pb_bufmgr_mm.c @@ -0,0 +1,593 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 1999 Wittawat Yamwong + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer manager using the old texture memory manager. + * + * \author José Fonseca + */ + + +#include "linked_list.h" + +#include "p_defines.h" +#include "p_debug.h" +#include "p_thread.h" +#include "p_util.h" +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct mem_block +{ + struct mem_block *next, *prev; + struct mem_block *next_free, *prev_free; + struct mem_block *heap; + int ofs, size; + unsigned int free:1; + unsigned int reserved:1; +}; + + +#ifdef DEBUG +/** + * For debugging purposes. + */ +static void +mmDumpMemInfo(const struct mem_block *heap) +{ + debug_printf("Memory heap %p:\n", (void *)heap); + if (heap == 0) { + debug_printf(" heap == 0\n"); + } else { + const struct mem_block *p; + + for(p = heap->next; p != heap; p = p->next) { + debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + debug_printf("\nFree list:\n"); + + for(p = heap->next_free; p != heap; p = p->next_free) { + debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + } + debug_printf("End of memory blocks\n"); +} +#endif + + +/** + * input: total size in bytes + * return: a heap pointer if OK, NULL if error + */ +static struct mem_block * +mmInit(int ofs, int size) +{ + struct mem_block *heap, *block; + + if (size <= 0) + return NULL; + + heap = CALLOC_STRUCT(mem_block); + if (!heap) + return NULL; + + block = CALLOC_STRUCT(mem_block); + if (!block) { + FREE(heap); + return NULL; + } + + heap->next = block; + heap->prev = block; + heap->next_free = block; + heap->prev_free = block; + + block->heap = heap; + block->next = heap; + block->prev = heap; + block->next_free = heap; + block->prev_free = heap; + + block->ofs = ofs; + block->size = size; + block->free = 1; + + return heap; +} + + +static struct mem_block * +SliceBlock(struct mem_block *p, + int startofs, int size, + int reserved, int alignment) +{ + struct mem_block *newblock; + + /* break left [p, newblock, p->next], then p = newblock */ + if (startofs > p->ofs) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs; + newblock->size = p->size - (startofs - p->ofs); + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size -= newblock->size; + p = newblock; + } + + /* break right, also [p, newblock, p->next] */ + if (size < p->size) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs + size; + newblock->size = p->size - size; + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size = size; + } + + /* p = middle block */ + p->free = 0; + + /* Remove p from the free list: + */ + p->next_free->prev_free = p->prev_free; + p->prev_free->next_free = p->next_free; + + p->next_free = 0; + p->prev_free = 0; + + p->reserved = reserved; + return p; +} + + +/** + * Allocate 'size' bytes with 2^align2 bytes alignment, + * restrict the search to free memory after 'startSearch' + * depth and back buffers should be in different 4mb banks + * to get better page hits if possible + * input: size = size of block + * align2 = 2^align2 bytes alignment + * startSearch = linear offset from start of heap to begin search + * return: pointer to the allocated block, 0 if error + */ +static struct mem_block * +mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) +{ + struct mem_block *p; + const int mask = (1 << align2)-1; + int startofs = 0; + int endofs; + + if (!heap || align2 < 0 || size <= 0) + return NULL; + + for (p = heap->next_free; p != heap; p = p->next_free) { + assert(p->free); + + startofs = (p->ofs + mask) & ~mask; + if ( startofs < startSearch ) { + startofs = startSearch; + } + endofs = startofs+size; + if (endofs <= (p->ofs+p->size)) + break; + } + + if (p == heap) + return NULL; + + assert(p->free); + p = SliceBlock(p,startofs,size,0,mask+1); + + return p; +} + + +#if 0 +/** + * Free block starts at offset + * input: pointer to a heap, start offset + * return: pointer to a block + */ +static struct mem_block * +mmFindBlock(struct mem_block *heap, int start) +{ + struct mem_block *p; + + for (p = heap->next; p != heap; p = p->next) { + if (p->ofs == start) + return p; + } + + return NULL; +} +#endif + + +static INLINE int +Join2Blocks(struct mem_block *p) +{ + /* XXX there should be some assertions here */ + + /* NOTE: heap->free == 0 */ + + if (p->free && p->next->free) { + struct mem_block *q = p->next; + + assert(p->ofs + p->size == q->ofs); + p->size += q->size; + + p->next = q->next; + q->next->prev = p; + + q->next_free->prev_free = q->prev_free; + q->prev_free->next_free = q->next_free; + + FREE(q); + return 1; + } + return 0; +} + + +/** + * Free block starts at offset + * input: pointer to a block + * return: 0 if OK, -1 if error + */ +static int +mmFreeMem(struct mem_block *b) +{ + if (!b) + return 0; + + if (b->free) { + debug_printf("block already free\n"); + return -1; + } + if (b->reserved) { + debug_printf("block is reserved\n"); + return -1; + } + + b->free = 1; + b->next_free = b->heap->next_free; + b->prev_free = b->heap; + b->next_free->prev_free = b; + b->prev_free->next_free = b; + + Join2Blocks(b); + if (b->prev != b->heap) + Join2Blocks(b->prev); + + return 0; +} + + +/** + * destroy MM + */ +static void +mmDestroy(struct mem_block *heap) +{ + struct mem_block *p; + + if (!heap) + return; + + for (p = heap->next; p != heap; ) { + struct mem_block *next = p->next; + FREE(p); + p = next; + } + + FREE(heap); +} + + +struct mm_pb_manager +{ + struct pb_manager base; + + _glthread_Mutex mutex; + + size_t size; + struct mem_block *heap; + + size_t align2; + + struct pb_buffer *buffer; + void *map; +}; + + +static INLINE struct mm_pb_manager * +mm_pb_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct mm_pb_manager *)mgr; +} + + +struct mm_buffer +{ + struct pb_buffer base; + + struct mm_pb_manager *mgr; + + struct mem_block *block; +}; + + +static INLINE struct mm_buffer * +mm_buffer(struct pb_buffer *buf) +{ + assert(buf); + return (struct mm_buffer *)buf; +} + + +static void +mm_buffer_destroy(struct pb_buffer *buf) +{ + struct mm_buffer *mm_buf = mm_buffer(buf); + struct mm_pb_manager *mm = mm_buf->mgr; + + assert(buf->base.refcount == 0); + + _glthread_LOCK_MUTEX(mm->mutex); + mmFreeMem(mm_buf->block); + FREE(buf); + _glthread_UNLOCK_MUTEX(mm->mutex); +} + + +static void * +mm_buffer_map(struct pb_buffer *buf, + unsigned flags) +{ + struct mm_buffer *mm_buf = mm_buffer(buf); + struct mm_pb_manager *mm = mm_buf->mgr; + + return (unsigned char *) mm->map + mm_buf->block->ofs; +} + + +static void +mm_buffer_unmap(struct pb_buffer *buf) +{ + /* No-op */ +} + + +static void +mm_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + struct mm_buffer *mm_buf = mm_buffer(buf); + struct mm_pb_manager *mm = mm_buf->mgr; + pb_get_base_buffer(mm->buffer, base_buf, offset); + *offset += mm_buf->block->ofs; +} + + +static const struct pb_vtbl +mm_buffer_vtbl = { + mm_buffer_destroy, + mm_buffer_map, + mm_buffer_unmap, + mm_buffer_get_base_buffer +}; + + +static struct pb_buffer * +mm_bufmgr_create_buffer(struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc) +{ + struct mm_pb_manager *mm = mm_pb_manager(mgr); + struct mm_buffer *mm_buf; + + /* We don't handle alignments larger then the one initially setup */ + assert(desc->alignment % (1 << mm->align2) == 0); + if(desc->alignment % (1 << mm->align2)) + return NULL; + + _glthread_LOCK_MUTEX(mm->mutex); + + mm_buf = CALLOC_STRUCT(mm_buffer); + if (!mm_buf) { + _glthread_UNLOCK_MUTEX(mm->mutex); + return NULL; + } + + mm_buf->base.base.refcount = 1; + mm_buf->base.base.alignment = desc->alignment; + mm_buf->base.base.usage = desc->usage; + mm_buf->base.base.size = size; + + mm_buf->base.vtbl = &mm_buffer_vtbl; + + mm_buf->mgr = mm; + + mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); + if(!mm_buf->block) { + debug_printf("warning: heap full\n"); +#if 0 + mmDumpMemInfo(mm->heap); +#endif + + mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); + if(!mm_buf->block) { + assert(0); + FREE(mm_buf); + _glthread_UNLOCK_MUTEX(mm->mutex); + return NULL; + } + } + + /* Some sanity checks */ + assert(0 <= mm_buf->block->ofs && mm_buf->block->ofs < mm->size); + assert(size <= mm_buf->block->size && mm_buf->block->ofs + mm_buf->block->size <= mm->size); + + _glthread_UNLOCK_MUTEX(mm->mutex); + return SUPER(mm_buf); +} + + +static void +mm_bufmgr_destroy(struct pb_manager *mgr) +{ + struct mm_pb_manager *mm = mm_pb_manager(mgr); + + _glthread_LOCK_MUTEX(mm->mutex); + + mmDestroy(mm->heap); + + pb_unmap(mm->buffer); + pb_reference(&mm->buffer, NULL); + + _glthread_UNLOCK_MUTEX(mm->mutex); + + FREE(mgr); +} + + +struct pb_manager * +mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, + size_t size, size_t align2) +{ + struct mm_pb_manager *mm; + + if(!buffer) + return NULL; + + mm = CALLOC_STRUCT(mm_pb_manager); + if (!mm) + return NULL; + + mm->base.create_buffer = mm_bufmgr_create_buffer; + mm->base.destroy = mm_bufmgr_destroy; + + mm->size = size; + mm->align2 = align2; /* 64-byte alignment */ + + _glthread_INIT_MUTEX(mm->mutex); + + mm->buffer = buffer; + + mm->map = pb_map(mm->buffer, + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); + if(!mm->map) + goto failure; + + mm->heap = mmInit(0, size); + if (!mm->heap) + goto failure; + + return SUPER(mm); + +failure: +if(mm->heap) + mmDestroy(mm->heap); + if(mm->map) + pb_unmap(mm->buffer); + if(mm) + FREE(mm); + return NULL; +} + + +struct pb_manager * +mm_bufmgr_create(struct pb_manager *provider, + size_t size, size_t align2) +{ + struct pb_buffer *buffer; + struct pb_manager *mgr; + struct pb_desc desc; + + assert(provider); + assert(provider->create_buffer); + + memset(&desc, 0, sizeof(desc)); + desc.alignment = 1 << align2; + + buffer = provider->create_buffer(provider, size, &desc); + if (!buffer) + return NULL; + + mgr = mm_bufmgr_create_from_buffer(buffer, size, align2); + if (!mgr) { + pb_reference(&buffer, NULL); + return NULL; + } + + return mgr; +} diff --git a/src/gallium/aux/pipebuffer/pb_bufmgr_pool.c b/src/gallium/aux/pipebuffer/pb_bufmgr_pool.c new file mode 100644 index 0000000000..04477a865a --- /dev/null +++ b/src/gallium/aux/pipebuffer/pb_bufmgr_pool.c @@ -0,0 +1,288 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * \file + * Batch buffer pool management. + * + * \author José Fonseca + * \author Thomas Hellström + */ + + +#include "linked_list.h" + +#include "p_compiler.h" +#include "p_debug.h" +#include "p_thread.h" +#include "p_defines.h" +#include "p_util.h" + +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct pool_pb_manager +{ + struct pb_manager base; + + _glthread_Mutex mutex; + + size_t bufSize; + size_t bufAlign; + + size_t numFree; + size_t numTot; + + struct list_head free; + + struct pb_buffer *buffer; + void *map; + + struct pool_buffer *bufs; +}; + + +static INLINE struct pool_pb_manager * +pool_pb_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct pool_pb_manager *)mgr; +} + + +struct pool_buffer +{ + struct pb_buffer base; + + struct pool_pb_manager *mgr; + + struct list_head head; + + size_t start; +}; + + +static INLINE struct pool_buffer * +pool_buffer(struct pb_buffer *buf) +{ + assert(buf); + return (struct pool_buffer *)buf; +} + + + +static void +pool_buffer_destroy(struct pb_buffer *buf) +{ + struct pool_buffer *pool_buf = pool_buffer(buf); + struct pool_pb_manager *pool = pool_buf->mgr; + + assert(pool_buf->base.base.refcount == 0); + + _glthread_LOCK_MUTEX(pool->mutex); + LIST_ADD(&pool_buf->head, &pool->free); + pool->numFree++; + _glthread_UNLOCK_MUTEX(pool->mutex); +} + + +static void * +pool_buffer_map(struct pb_buffer *buf, unsigned flags) +{ + struct pool_buffer *pool_buf = pool_buffer(buf); + struct pool_pb_manager *pool = pool_buf->mgr; + void *map; + + _glthread_LOCK_MUTEX(pool->mutex); + map = (unsigned char *) pool->map + pool_buf->start; + _glthread_UNLOCK_MUTEX(pool->mutex); + return map; +} + + +static void +pool_buffer_unmap(struct pb_buffer *buf) +{ + /* No-op */ +} + + +static void +pool_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + struct pool_buffer *pool_buf = pool_buffer(buf); + struct pool_pb_manager *pool = pool_buf->mgr; + pb_get_base_buffer(pool->buffer, base_buf, offset); + *offset += pool_buf->start; +} + + +static const struct pb_vtbl +pool_buffer_vtbl = { + pool_buffer_destroy, + pool_buffer_map, + pool_buffer_unmap, + pool_buffer_get_base_buffer +}; + + +static struct pb_buffer * +pool_bufmgr_create_buffer(struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc) +{ + struct pool_pb_manager *pool = pool_pb_manager(mgr); + struct pool_buffer *pool_buf; + struct list_head *item; + + assert(size == pool->bufSize); + assert(pool->bufAlign % desc->alignment == 0); + + _glthread_LOCK_MUTEX(pool->mutex); + + if (pool->numFree == 0) { + _glthread_UNLOCK_MUTEX(pool->mutex); + debug_printf("warning: out of fixed size buffer objects\n"); + return NULL; + } + + item = pool->free.next; + + if (item == &pool->free) { + _glthread_UNLOCK_MUTEX(pool->mutex); + debug_printf("error: fixed size buffer pool corruption\n"); + return NULL; + } + + LIST_DEL(item); + --pool->numFree; + + _glthread_UNLOCK_MUTEX(pool->mutex); + + pool_buf = LIST_ENTRY(struct pool_buffer, item, head); + assert(pool_buf->base.base.refcount == 0); + pool_buf->base.base.refcount = 1; + pool_buf->base.base.alignment = desc->alignment; + pool_buf->base.base.usage = desc->usage; + + return SUPER(pool_buf); +} + + +static void +pool_bufmgr_destroy(struct pb_manager *mgr) +{ + struct pool_pb_manager *pool = pool_pb_manager(mgr); + _glthread_LOCK_MUTEX(pool->mutex); + + FREE(pool->bufs); + + pb_unmap(pool->buffer); + pb_reference(&pool->buffer, NULL); + + _glthread_UNLOCK_MUTEX(pool->mutex); + + FREE(mgr); +} + + +struct pb_manager * +pool_bufmgr_create(struct pb_manager *provider, + size_t numBufs, + size_t bufSize, + const struct pb_desc *desc) +{ + struct pool_pb_manager *pool; + struct pool_buffer *pool_buf; + size_t i; + + pool = (struct pool_pb_manager *)CALLOC(1, sizeof(*pool)); + if (!pool) + return NULL; + + pool->base.destroy = pool_bufmgr_destroy; + pool->base.create_buffer = pool_bufmgr_create_buffer; + + LIST_INITHEAD(&pool->free); + + pool->numTot = numBufs; + pool->numFree = numBufs; + pool->bufSize = bufSize; + pool->bufAlign = desc->alignment; + + _glthread_INIT_MUTEX(pool->mutex); + + pool->buffer = provider->create_buffer(provider, numBufs*bufSize, desc); + if (!pool->buffer) + goto failure; + + pool->map = pb_map(pool->buffer, + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); + if(!pool->map) + goto failure; + + pool->bufs = (struct pool_buffer *)CALLOC(numBufs, sizeof(*pool->bufs)); + if (!pool->bufs) + goto failure; + + pool_buf = pool->bufs; + for (i = 0; i < numBufs; ++i) { + pool_buf->base.base.refcount = 0; + pool_buf->base.base.alignment = 0; + pool_buf->base.base.usage = 0; + pool_buf->base.base.size = bufSize; + pool_buf->base.vtbl = &pool_buffer_vtbl; + pool_buf->mgr = pool; + pool_buf->start = i * bufSize; + LIST_ADDTAIL(&pool_buf->head, &pool->free); + pool_buf++; + } + + return SUPER(pool); + +failure: + if(pool->bufs) + FREE(pool->bufs); + if(pool->map) + pb_unmap(pool->buffer); + if(pool->buffer) + pb_reference(&pool->buffer, NULL); + if(pool) + FREE(pool); + return NULL; +} diff --git a/src/gallium/aux/pipebuffer/pb_winsys.c b/src/gallium/aux/pipebuffer/pb_winsys.c new file mode 100644 index 0000000000..978944091f --- /dev/null +++ b/src/gallium/aux/pipebuffer/pb_winsys.c @@ -0,0 +1,170 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of client buffer (also designated as "user buffers"), which + * are just state-tracker owned data masqueraded as buffers. + * + * \author José Fonseca + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" + +#include "pb_buffer.h" + + +/** + * User buffers are special buffers that initially reference memory + * held by the user but which may if necessary copy that memory into + * device memory behind the scenes, for submission to hardware. + * + * These are particularly useful when the referenced data is never + * submitted to hardware at all, in the particular case of software + * vertex processing. + */ +struct pb_user_buffer +{ + struct pb_buffer base; + void *data; +}; + + +extern const struct pb_vtbl pb_user_buffer_vtbl; + + +static INLINE struct pb_user_buffer * +pb_user_buffer(struct pb_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &pb_user_buffer_vtbl); + return (struct pb_user_buffer *)buf; +} + + +static void +pb_user_buffer_destroy(struct pb_buffer *buf) +{ + assert(buf); + FREE(buf); +} + + +static void * +pb_user_buffer_map(struct pb_buffer *buf, + unsigned flags) +{ + return pb_user_buffer(buf)->data; +} + + +static void +pb_user_buffer_unmap(struct pb_buffer *buf) +{ + /* No-op */ +} + + +static void +pb_user_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + *base_buf = buf; + *offset = 0; +} + + +const struct pb_vtbl +pb_user_buffer_vtbl = { + pb_user_buffer_destroy, + pb_user_buffer_map, + pb_user_buffer_unmap, + pb_user_buffer_get_base_buffer +}; + + +static struct pipe_buffer * +pb_winsys_user_buffer_create(struct pipe_winsys *winsys, + void *data, + unsigned bytes) +{ + struct pb_user_buffer *buf = CALLOC_STRUCT(pb_user_buffer); + + if(!buf) + return NULL; + + buf->base.base.refcount = 1; + buf->base.base.size = bytes; + buf->base.base.alignment = 0; + buf->base.base.usage = 0; + + buf->base.vtbl = &pb_user_buffer_vtbl; + buf->data = data; + + return &buf->base.base; +} + + +static void * +pb_winsys_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags) +{ + (void)winsys; + return pb_map(pb_buffer(buf), flags); +} + + +static void +pb_winsys_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + (void)winsys; + pb_unmap(pb_buffer(buf)); +} + + +static void +pb_winsys_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + (void)winsys; + pb_destroy(pb_buffer(buf)); +} + + +void +pb_init_winsys(struct pipe_winsys *winsys) +{ + winsys->user_buffer_create = pb_winsys_user_buffer_create; + winsys->buffer_map = pb_winsys_buffer_map; + winsys->buffer_unmap = pb_winsys_buffer_unmap; + winsys->buffer_destroy = pb_winsys_buffer_destroy; +} diff --git a/src/gallium/aux/tgsi/Makefile b/src/gallium/aux/tgsi/Makefile new file mode 100644 index 0000000000..12a8bd0409 --- /dev/null +++ b/src/gallium/aux/tgsi/Makefile @@ -0,0 +1,3 @@ +default: + cd ../.. ; make + diff --git a/src/gallium/aux/tgsi/exec/Makefile b/src/gallium/aux/tgsi/exec/Makefile new file mode 100644 index 0000000000..eb8b14e0e8 --- /dev/null +++ b/src/gallium/aux/tgsi/exec/Makefile @@ -0,0 +1,3 @@ +default: + cd ../../.. ; make + diff --git a/src/gallium/aux/tgsi/exec/tgsi_exec.c b/src/gallium/aux/tgsi/exec/tgsi_exec.c new file mode 100644 index 0000000000..37e6007068 --- /dev/null +++ b/src/gallium/aux/tgsi/exec/tgsi_exec.c @@ -0,0 +1,2485 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI interpretor/executor. + * + * Flow control information: + * + * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) + * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special + * care since a condition may be true for some quad components but false + * for other components. + * + * We basically execute all statements (even if they're in the part of + * an IF/ELSE clause that's "not taken") and use a special mask to + * control writing to destination registers. This is the ExecMask. + * See store_dest(). + * + * The ExecMask is computed from three other masks (CondMask, LoopMask and + * ContMask) which are controlled by the flow control instructions (namely: + * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). + * + * + * Authors: + * Michal Krol + * Brian Paul + */ + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/util/tgsi_util.h" +#include "tgsi_exec.h" + +#define TILE_TOP_LEFT 0 +#define TILE_TOP_RIGHT 1 +#define TILE_BOTTOM_LEFT 2 +#define TILE_BOTTOM_RIGHT 3 + +/* + * Shorthand locations of various utility registers (_I = Index, _C = Channel) + */ +#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I +#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C +#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I +#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C +#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I +#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C +#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I +#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C +#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I +#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C +#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I +#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C +#define TEMP_128_I TGSI_EXEC_TEMP_128_I +#define TEMP_128_C TGSI_EXEC_TEMP_128_C +#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I +#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C +#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I +#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C +#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I +#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C +#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I +#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +#define FOR_EACH_CHANNEL(CHAN)\ + for (CHAN = 0; CHAN < 4; CHAN++) + +#define IS_CHANNEL_ENABLED(INST, CHAN)\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IS_CHANNEL_ENABLED2(INST, CHAN)\ + ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) + +#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED2( INST, CHAN )) + + +/** The execution mask depends on the conditional mask and the loop mask */ +#define UPDATE_EXEC_MASK(MACH) \ + MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask + + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + + + +static void +tgsi_exec_prepare( struct tgsi_exec_machine *mach ) +{ + struct tgsi_exec_labels *labels = &mach->Labels; + struct tgsi_parse_context parse; + struct tgsi_full_instruction *instructions; + struct tgsi_full_declaration *declarations; + uint maxInstructions = 10, numInstructions = 0; + uint maxDeclarations = 10, numDeclarations = 0; + uint k; + uint instno = 0; + + mach->ImmLimit = 0; + labels->count = 0; + + declarations = (struct tgsi_full_declaration *) + MALLOC( maxDeclarations * sizeof(struct tgsi_full_declaration) ); + + instructions = (struct tgsi_full_instruction *) + MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) ); + + k = tgsi_parse_init( &parse, mach->Tokens ); + if (k != TGSI_PARSE_OK) { + debug_printf("Problem parsing!\n"); + return; + } + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + uint pointer = parse.Position; + uint i; + + tgsi_parse_token( &parse ); + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + /* save expanded declaration */ + if (numDeclarations == maxDeclarations) { + declarations = REALLOC(declarations, + maxDeclarations + * sizeof(struct tgsi_full_declaration), + (maxDeclarations + 10) + * sizeof(struct tgsi_full_declaration)); + maxDeclarations += 10; + } + memcpy(declarations + numDeclarations, + &parse.FullToken.FullDeclaration, + sizeof(declarations[0])); + numDeclarations++; + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; + assert( size % 4 == 0 ); + assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES ); + + for( i = 0; i < size; i++ ) { + mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } + mach->ImmLimit += size / 4; + } + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + assert( labels->count < 128 ); + + labels->labels[labels->count][0] = instno; + labels->labels[labels->count][1] = pointer; + labels->count++; + + /* save expanded instruction */ + if (numInstructions == maxInstructions) { + instructions = REALLOC(instructions, + maxInstructions + * sizeof(struct tgsi_full_instruction), + (maxInstructions + 10) + * sizeof(struct tgsi_full_instruction)); + maxInstructions += 10; + } + memcpy(instructions + numInstructions, + &parse.FullToken.FullInstruction, + sizeof(instructions[0])); + numInstructions++; + break; + + default: + assert( 0 ); + } + } + tgsi_parse_free (&parse); + + if (mach->Declarations) { + FREE( mach->Declarations ); + } + mach->Declarations = declarations; + mach->NumDeclarations = numDeclarations; + + if (mach->Instructions) { + FREE( mach->Instructions ); + } + mach->Instructions = instructions; + mach->NumInstructions = numInstructions; +} + + +/** + * Initialize machine state by expanding tokens to full instructions, + * allocating temporary storage, setting up constants, etc. + * After this, we can call tgsi_exec_machine_run() many times. + */ +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach, + const struct tgsi_token *tokens, + uint numSamplers, + struct tgsi_sampler *samplers) +{ + uint i, k; + struct tgsi_parse_context parse; + +#if 0 + tgsi_dump(tokens, 0); +#endif + + mach->Tokens = tokens; + + mach->Samplers = samplers; + + k = tgsi_parse_init (&parse, mach->Tokens); + if (k != TGSI_PARSE_OK) { + debug_printf( "Problem parsing!\n" ); + return; + } + + mach->Processor = parse.FullHeader.Processor.Processor; + tgsi_parse_free (&parse); + + mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); + mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; + + /* Setup constants. */ + for( i = 0; i < 4; i++ ) { + mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; + mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; + mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; + mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; + mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; + mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; + mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; + mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; + } + + tgsi_exec_prepare( mach ); +} + + +void +tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach) +{ + if (mach->Instructions) { + FREE(mach->Instructions); + mach->Instructions = NULL; + mach->NumInstructions = 0; + } + if (mach->Declarations) { + FREE(mach->Declarations); + mach->Declarations = NULL; + mach->NumDeclarations = 0; + } +} + + +static void +micro_abs( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) fabs( (double) src->f[0] ); + dst->f[1] = (float) fabs( (double) src->f[1] ); + dst->f[2] = (float) fabs( (double) src->f[2] ); + dst->f[3] = (float) fabs( (double) src->f[3] ); +} + +static void +micro_add( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] + src1->f[0]; + dst->f[1] = src0->f[1] + src1->f[1]; + dst->f[2] = src0->f[2] + src1->f[2]; + dst->f[3] = src0->f[3] + src1->f[3]; +} + +static void +micro_iadd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] + src1->i[0]; + dst->i[1] = src0->i[1] + src1->i[1]; + dst->i[2] = src0->i[2] + src1->i[2]; + dst->i[3] = src0->i[3] + src1->i[3]; +} + +static void +micro_and( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] & src1->u[0]; + dst->u[1] = src0->u[1] & src1->u[1]; + dst->u[2] = src0->u[2] & src1->u[2]; + dst->u[3] = src0->u[3] & src1->u[3]; +} + +static void +micro_ceil( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) ceil( (double) src->f[0] ); + dst->f[1] = (float) ceil( (double) src->f[1] ); + dst->f[2] = (float) ceil( (double) src->f[2] ); + dst->f[3] = (float) ceil( (double) src->f[3] ); +} + +static void +micro_cos( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) cos( (double) src->f[0] ); + dst->f[1] = (float) cos( (double) src->f[1] ); + dst->f[2] = (float) cos( (double) src->f[2] ); + dst->f[3] = (float) cos( (double) src->f[3] ); +} + +static void +micro_ddx( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_ddy( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_div( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] / src1->f[0]; + dst->f[1] = src0->f[1] / src1->f[1]; + dst->f[2] = src0->f[2] / src1->f[2]; + dst->f[3] = src0->f[3] / src1->f[3]; +} + +static void +micro_udiv( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] / src1->u[0]; + dst->u[1] = src0->u[1] / src1->u[1]; + dst->u[2] = src0->u[2] / src1->u[2]; + dst->u[3] = src0->u[3] / src1->u[3]; +} + +static void +micro_eq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ieq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_exp2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = (float) pow( 2.0, (double) src->f[0] ); + dst->f[1] = (float) pow( 2.0, (double) src->f[1] ); + dst->f[2] = (float) pow( 2.0, (double) src->f[2] ); + dst->f[3] = (float) pow( 2.0, (double) src->f[3] ); +} + +static void +micro_f2it( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = (int) src->f[0]; + dst->i[1] = (int) src->f[1]; + dst->i[2] = (int) src->f[2]; + dst->i[3] = (int) src->f[3]; +} + +static void +micro_f2ut( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = (uint) src->f[0]; + dst->u[1] = (uint) src->f[1]; + dst->u[2] = (uint) src->f[2]; + dst->u[3] = (uint) src->f[3]; +} + +static void +micro_flr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) floor( (double) src->f[0] ); + dst->f[1] = (float) floor( (double) src->f[1] ); + dst->f[2] = (float) floor( (double) src->f[2] ); + dst->f[3] = (float) floor( (double) src->f[3] ); +} + +static void +micro_frc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = src->f[0] - (float) floor( (double) src->f[0] ); + dst->f[1] = src->f[1] - (float) floor( (double) src->f[1] ); + dst->f[2] = src->f[2] - (float) floor( (double) src->f[2] ); + dst->f[3] = src->f[3] - (float) floor( (double) src->f[3] ); +} + +static void +micro_ge( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_i2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) src->i[0]; + dst->f[1] = (float) src->i[1]; + dst->f[2] = (float) src->i[2]; + dst->f[3] = (float) src->i[3]; +} + +static void +micro_lg2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) log( (double) src->f[0] ) * 1.442695f; + dst->f[1] = (float) log( (double) src->f[1] ) * 1.442695f; + dst->f[2] = (float) log( (double) src->f[2] ) * 1.442695f; + dst->f[3] = (float) log( (double) src->f[3] ) * 1.442695f; +} + +static void +micro_lt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ilt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_ult( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; +} + +static void +micro_max( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_min( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_umod( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] % src1->u[0]; + dst->u[1] = src0->u[1] % src1->u[1]; + dst->u[2] = src0->u[2] % src1->u[2]; + dst->u[3] = src0->u[3] % src1->u[3]; +} + +static void +micro_mul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] * src1->f[0]; + dst->f[1] = src0->f[1] * src1->f[1]; + dst->f[2] = src0->f[2] * src1->f[2]; + dst->f[3] = src0->f[3] * src1->f[3]; +} + +static void +micro_imul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] * src1->i[0]; + dst->i[1] = src0->i[1] * src1->i[1]; + dst->i[2] = src0->i[2] * src1->i[2]; + dst->i[3] = src0->i[3] * src1->i[3]; +} + +static void +micro_imul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->i[0] = src0->i[0] * src1->i[0]; + dst1->i[1] = src0->i[1] * src1->i[1]; + dst1->i[2] = src0->i[2] * src1->i[2]; + dst1->i[3] = src0->i[3] * src1->i[3]; + dst0->i[0] = 0; + dst0->i[1] = 0; + dst0->i[2] = 0; + dst0->i[3] = 0; +} + +static void +micro_umul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->u[0] = src0->u[0] * src1->u[0]; + dst1->u[1] = src0->u[1] * src1->u[1]; + dst1->u[2] = src0->u[2] * src1->u[2]; + dst1->u[3] = src0->u[3] * src1->u[3]; + dst0->u[0] = 0; + dst0->u[1] = 0; + dst0->u[2] = 0; + dst0->u[3] = 0; +} + +static void +micro_movc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2 ) +{ + dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; + dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; + dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; + dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; +} + +static void +micro_neg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = -src->f[0]; + dst->f[1] = -src->f[1]; + dst->f[2] = -src->f[2]; + dst->f[3] = -src->f[3]; +} + +static void +micro_ineg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = -src->i[0]; + dst->i[1] = -src->i[1]; + dst->i[2] = -src->i[2]; + dst->i[3] = -src->i[3]; +} + +static void +micro_not( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = ~src->u[0]; + dst->u[1] = ~src->u[1]; + dst->u[2] = ~src->u[2]; + dst->u[3] = ~src->u[3]; +} + +static void +micro_or( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] | src1->u[0]; + dst->u[1] = src0->u[1] | src1->u[1]; + dst->u[2] = src0->u[2] | src1->u[2]; + dst->u[3] = src0->u[3] | src1->u[3]; +} + +static void +micro_pow( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = (float) pow( (double) src0->f[0], (double) src1->f[0] ); + dst->f[1] = (float) pow( (double) src0->f[1], (double) src1->f[1] ); + dst->f[2] = (float) pow( (double) src0->f[2], (double) src1->f[2] ); + dst->f[3] = (float) pow( (double) src0->f[3], (double) src1->f[3] ); +} + +static void +micro_rnd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) floor( (double) (src->f[0] + 0.5f) ); + dst->f[1] = (float) floor( (double) (src->f[1] + 0.5f) ); + dst->f[2] = (float) floor( (double) (src->f[2] + 0.5f) ); + dst->f[3] = (float) floor( (double) (src->f[3] + 0.5f) ); +} + +static void +micro_shl( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] << src1->i[0]; + dst->i[1] = src0->i[1] << src1->i[1]; + dst->i[2] = src0->i[2] << src1->i[2]; + dst->i[3] = src0->i[3] << src1->i[3]; +} + +static void +micro_ishr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] >> src1->i[0]; + dst->i[1] = src0->i[1] >> src1->i[1]; + dst->i[2] = src0->i[2] >> src1->i[2]; + dst->i[3] = src0->i[3] >> src1->i[3]; +} + +static void +micro_trunc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0 ) +{ + dst->f[0] = (float) (int) src0->f[0]; + dst->f[1] = (float) (int) src0->f[1]; + dst->f[2] = (float) (int) src0->f[2]; + dst->f[3] = (float) (int) src0->f[3]; +} + +static void +micro_ushr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] >> src1->u[0]; + dst->u[1] = src0->u[1] >> src1->u[1]; + dst->u[2] = src0->u[2] >> src1->u[2]; + dst->u[3] = src0->u[3] >> src1->u[3]; +} + +static void +micro_sin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) sin( (double) src->f[0] ); + dst->f[1] = (float) sin( (double) src->f[1] ); + dst->f[2] = (float) sin( (double) src->f[2] ); + dst->f[3] = (float) sin( (double) src->f[3] ); +} + +static void +micro_sqrt( union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) sqrt( (double) src->f[0] ); + dst->f[1] = (float) sqrt( (double) src->f[1] ); + dst->f[2] = (float) sqrt( (double) src->f[2] ); + dst->f[3] = (float) sqrt( (double) src->f[3] ); +} + +static void +micro_sub( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] - src1->f[0]; + dst->f[1] = src0->f[1] - src1->f[1]; + dst->f[2] = src0->f[2] - src1->f[2]; + dst->f[3] = src0->f[3] - src1->f[3]; +} + +static void +micro_u2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) src->u[0]; + dst->f[1] = (float) src->u[1]; + dst->f[2] = (float) src->u[2]; + dst->f[3] = (float) src->u[3]; +} + +static void +micro_xor( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] ^ src1->u[0]; + dst->u[1] = src0->u[1] ^ src1->u[1]; + dst->u[2] = src0->u[2] ^ src1->u[2]; + dst->u[3] = src0->u[3] ^ src1->u[3]; +} + +static void +fetch_src_file_channel( + const struct tgsi_exec_machine *mach, + const uint file, + const uint swizzle, + const union tgsi_exec_channel *index, + union tgsi_exec_channel *chan ) +{ + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( file ) { + case TGSI_FILE_CONSTANT: + chan->f[0] = mach->Consts[index->i[0]][swizzle]; + chan->f[1] = mach->Consts[index->i[1]][swizzle]; + chan->f[2] = mach->Consts[index->i[2]][swizzle]; + chan->f[3] = mach->Consts[index->i[3]][swizzle]; + break; + + case TGSI_FILE_INPUT: + chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_TEMPORARY: + chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_IMMEDIATE: + assert( index->i[0] < (int) mach->ImmLimit ); + chan->f[0] = mach->Imms[index->i[0]][swizzle]; + assert( index->i[1] < (int) mach->ImmLimit ); + chan->f[1] = mach->Imms[index->i[1]][swizzle]; + assert( index->i[2] < (int) mach->ImmLimit ); + chan->f[2] = mach->Imms[index->i[2]][swizzle]; + assert( index->i[3] < (int) mach->ImmLimit ); + chan->f[3] = mach->Imms[index->i[3]][swizzle]; + break; + + case TGSI_FILE_ADDRESS: + chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_OUTPUT: + /* vertex/fragment output vars can be read too */ + chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; + break; + + case TGSI_EXTSWIZZLE_ONE: + *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; + break; + + default: + assert( 0 ); + } +} + +static void +fetch_source( + const struct tgsi_exec_machine *mach, + union tgsi_exec_channel *chan, + const struct tgsi_full_src_register *reg, + const uint chan_index ) +{ + union tgsi_exec_channel index; + uint swizzle; + + index.i[0] = + index.i[1] = + index.i[2] = + index.i[3] = reg->SrcRegister.Index; + + if (reg->SrcRegister.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + + if( reg->SrcRegister.Dimension ) { + switch( reg->SrcRegister.File ) { + case TGSI_FILE_INPUT: + index.i[0] *= 17; + index.i[1] *= 17; + index.i[2] *= 17; + index.i[3] *= 17; + break; + case TGSI_FILE_CONSTANT: + index.i[0] *= 4096; + index.i[1] *= 4096; + index.i[2] *= 4096; + index.i[3] *= 4096; + break; + default: + assert( 0 ); + } + + index.i[0] += reg->SrcRegisterDim.Index; + index.i[1] += reg->SrcRegisterDim.Index; + index.i[2] += reg->SrcRegisterDim.Index; + index.i[3] += reg->SrcRegisterDim.Index; + + if (reg->SrcRegisterDim.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterDimInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterDimInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + } + + swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + fetch_src_file_channel( + mach, + reg->SrcRegister.File, + swizzle, + &index, + chan ); + + switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { + case TGSI_UTIL_SIGN_CLEAR: + micro_abs( chan, chan ); + break; + + case TGSI_UTIL_SIGN_SET: + micro_abs( chan, chan ); + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } + + if (reg->SrcRegisterExtMod.Complement) { + micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan ); + } +} + +static void +store_dest( + struct tgsi_exec_machine *mach, + const union tgsi_exec_channel *chan, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + uint chan_index ) +{ + union tgsi_exec_channel *dst; + + switch( reg->DstRegister.File ) { + case TGSI_FILE_NULL: + return; + + case TGSI_FILE_OUTPUT: + dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + + reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_TEMPORARY: + dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_ADDRESS: + dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; + break; + + default: + assert( 0 ); + return; + } + + switch (inst->Instruction.Saturate) + { + case TGSI_SAT_NONE: + if (mach->ExecMask & 0x1) + dst->i[0] = chan->i[0]; + if (mach->ExecMask & 0x2) + dst->i[1] = chan->i[1]; + if (mach->ExecMask & 0x4) + dst->i[2] = chan->i[2]; + if (mach->ExecMask & 0x8) + dst->i[3] = chan->i[3]; + break; + + case TGSI_SAT_ZERO_ONE: + /* XXX need to obey ExecMask here */ + micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); + micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + + default: + assert( 0 ); + } +} + +#define FETCH(VAL,INDEX,CHAN)\ + fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) + +#define STORE(VAL,INDEX,CHAN)\ + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) + + +/** + * Execute ARB-style KIL which is predicated by a src register. + * Kill fragment if any of the four values is less than zero. + */ +static void +exec_kilp(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + uint uniquemask; + uint chan_index; + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + union tgsi_exec_channel r[1]; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + uint swizzle; + uint i; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle ( + &inst->FullSrcRegisters[0], + chan_index); + + /* check if the component has not been already tested */ + if (uniquemask & (1 << swizzle)) + continue; + uniquemask |= 1 << swizzle; + + FETCH(&r[0], 0, chan_index); + for (i = 0; i < 4; i++) + if (r[0].f[i] < 0.0f) + kilmask |= 1 << i; + } + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; +} + + +/* + * Fetch a texel using STR texture coordinates. + */ +static void +fetch_texel( struct tgsi_sampler *sampler, + const union tgsi_exec_channel *s, + const union tgsi_exec_channel *t, + const union tgsi_exec_channel *p, + float lodbias, /* XXX should be float[4] */ + union tgsi_exec_channel *r, + union tgsi_exec_channel *g, + union tgsi_exec_channel *b, + union tgsi_exec_channel *a ) +{ + uint j; + float rgba[NUM_CHANNELS][QUAD_SIZE]; + + sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); + + for (j = 0; j < 4; j++) { + r->f[j] = rgba[0][j]; + g->f[j] = rgba[1][j]; + b->f[j] = rgba[2][j]; + a->f[j] = rgba[3][j]; + } +} + + +static void +exec_tex(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + boolean biasLod) +{ + const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + union tgsi_exec_channel r[8]; + uint chan_index; + float lodBias; + + /* debug_printf("Sampler %u unit %u\n", sampler, unit); */ + + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: + + FETCH(&r[0], 0, CHAN_X); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[1], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[1] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[1], 0, CHAN_W); + lodBias = r[2].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ + &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ + break; + + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, /* inputs */ + &r[0], &r[1], &r[2], &r[3]); /* outputs */ + break; + + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, + &r[0], &r[1], &r[2], &r[3]); + break; + + default: + assert (0); + } + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } +} + + +/** + * Evaluate a constant-valued coefficient at the position of the + * current quad. + */ +static void +eval_constant_coef( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; + } +} + +/** + * Evaluate a linear-valued coefficient at the position of the + * current quad. + */ +static void +eval_linear_coef( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + mach->Inputs[attrib].xyzw[chan].f[0] = a0; + mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; + mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; + mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; +} + +/** + * Evaluate a perspective-valued coefficient at the position of the + * current quad. + */ +static void +eval_perspective_coef( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + const float *w = mach->QuadPos.xyzw[3].f; + /* divide by W here */ + mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; + mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; + mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; + mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; +} + + +typedef void (* eval_coef_func)( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ); + +static void +exec_declaration( + struct tgsi_exec_machine *mach, + const struct tgsi_full_declaration *decl ) +{ + if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + eval_coef_func eval; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + switch( decl->Interpolation.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + eval = eval_constant_coef; + break; + + case TGSI_INTERPOLATE_LINEAR: + eval = eval_linear_coef; + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + eval = eval_perspective_coef; + break; + + default: + assert( 0 ); + } + + if( mask == TGSI_WRITEMASK_XYZW ) { + unsigned i, j; + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + eval( mach, i, j ); + } + } + } + else { + unsigned i, j; + + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + eval( mach, i, j ); + } + } + } + } + } + } +} + +static void +exec_instruction( + struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + int *pc ) +{ + uint chan_index; + union tgsi_exec_channel r[8]; + + (*pc)++; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_f2it( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_X ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[1], 0, CHAN_Y ); + micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + + FETCH( &r[2], 0, CHAN_W ); + micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); + micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); + micro_pow( &r[1], &r[1], &r[2] ); + micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Z ); + } + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( &r[0], 0, CHAN_X ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( &r[0], 0, CHAN_X ); + micro_sqrt( &r[0], &r[0] ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + assert (0); + break; + + case TGSI_OPCODE_LOG: + assert (0); + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_mul( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Z ); + FETCH( &r[2], 1, CHAN_Z ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_W); + FETCH(&r[2], 1, CHAN_W); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + FETCH( &r[0], 0, CHAN_Y ); + FETCH( &r[1], 1, CHAN_Y); + micro_mul( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_Z ); + STORE( &r[0], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + FETCH( &r[0], 1, CHAN_W ); + STORE( &r[0], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + /* XXX use micro_min()?? */ + micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + /* XXX use micro_max()?? */ + micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); + + STORE(&r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_mul( &r[0], &r[0], &r[1] ); + FETCH( &r[1], 2, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_sub( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_sub( &r[1], &r[1], &r[2] ); + micro_mul( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_CND: + assert (0); + break; + + case TGSI_OPCODE_CND0: + assert (0); + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert (0); + break; + + case TGSI_OPCODE_INDEX: + assert (0); + break; + + case TGSI_OPCODE_NEGATE: + assert (0); + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_frc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + assert (0); + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_flr( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_rnd( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH(&r[0], 0, CHAN_X); + + micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( &r[0], 0, CHAN_X ); + micro_lg2( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_pow( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + FETCH(&r[0], 0, CHAN_Y); + FETCH(&r[1], 1, CHAN_Z); + + micro_mul( &r[2], &r[0], &r[1] ); + + FETCH(&r[3], 0, CHAN_Z); + FETCH(&r[4], 1, CHAN_Y); + + micro_mul( &r[5], &r[3], &r[4] ); + micro_sub( &r[2], &r[2], &r[5] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &r[2], 0, CHAN_X ); + } + + FETCH(&r[2], 1, CHAN_X); + + micro_mul( &r[3], &r[3], &r[2] ); + + FETCH(&r[5], 0, CHAN_X); + + micro_mul( &r[1], &r[1], &r[5] ); + micro_sub( &r[3], &r[3], &r[1] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + STORE( &r[3], 0, CHAN_Y ); + } + + micro_mul( &r[5], &r[5], &r[4] ); + micro_mul( &r[0], &r[0], &r[2] ); + micro_sub( &r[5], &r[5], &r[0] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( &r[5], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + assert (0); + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + + micro_abs( &r[0], &r[0] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_RCC: + assert (0); + break; + + case TGSI_OPCODE_DPH: + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 1, CHAN_W); + + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH(&r[0], 0, CHAN_X); + + micro_cos( &r[0], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddx( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDY: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddy( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_KILP: + exec_kilp (mach, inst); + break; + + case TGSI_OPCODE_KIL: + /* for enabled ExecMask bits, set the killed bit */ + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; + break; + + case TGSI_OPCODE_PK2H: + assert (0); + break; + + case TGSI_OPCODE_PK2US: + assert (0); + break; + + case TGSI_OPCODE_PK4B: + assert (0); + break; + + case TGSI_OPCODE_PK4UB: + assert (0); + break; + + case TGSI_OPCODE_RFL: + assert (0); + break; + + case TGSI_OPCODE_SEQ: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_eq( &r[0], &r[0], &r[1], + &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], + &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SFL: + assert (0); + break; + + case TGSI_OPCODE_SGT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SIN: + FETCH( &r[0], 0, CHAN_X ); + micro_sin( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SNE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_STR: + assert (0); + break; + + case TGSI_OPCODE_TEX: + /* simple texture lookup */ + /* src[0] = texcoord */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, FALSE); + break; + + case TGSI_OPCODE_TXB: + /* Texture lookup with lod bias */ + /* src[0] = texcoord (src[0].w = LOD bias) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE); + break; + + case TGSI_OPCODE_TXD: + /* Texture lookup with explict partial derivatives */ + /* src[0] = texcoord */ + /* src[1] = d[strq]/dx */ + /* src[2] = d[strq]/dy */ + /* src[3] = sampler unit */ + assert (0); + break; + + case TGSI_OPCODE_TXL: + /* Texture lookup with explit LOD */ + /* src[0] = texcoord (src[0].w = LOD) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE); + break; + + case TGSI_OPCODE_UP2H: + assert (0); + break; + + case TGSI_OPCODE_UP2US: + assert (0); + break; + + case TGSI_OPCODE_UP4B: + assert (0); + break; + + case TGSI_OPCODE_UP4UB: + assert (0); + break; + + case TGSI_OPCODE_X2D: + assert (0); + break; + + case TGSI_OPCODE_ARA: + assert (0); + break; + + case TGSI_OPCODE_ARR: + assert (0); + break; + + case TGSI_OPCODE_BRA: + assert (0); + break; + + case TGSI_OPCODE_CAL: + /* skip the call if no execution channels are enabled */ + if (mach->ExecMask) { + /* do the call */ + + /* push the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + + assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); + mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; + + /* note that PC was already incremented above */ + mach->CallStack[mach->CallStackTop++] = *pc; + *pc = inst->InstructionExtLabel.Label; + } + break; + + case TGSI_OPCODE_RET: + mach->FuncMask &= ~mach->ExecMask; + UPDATE_EXEC_MASK(mach); + + if (mach->ExecMask == 0x0) { + /* really return now (otherwise, keep executing */ + + if (mach->CallStackTop == 0) { + /* returning from main() */ + *pc = -1; + return; + } + *pc = mach->CallStack[--mach->CallStackTop]; + + /* pop the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->FuncStackTop > 0); + mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; + + UPDATE_EXEC_MASK(mach); + } + break; + + case TGSI_OPCODE_SSG: + assert (0); + break; + + case TGSI_OPCODE_CMP: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_SCS: + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( &r[0], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + micro_cos( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + micro_sin( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_Y ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_NRM: + assert (0); + break; + + case TGSI_OPCODE_DIV: + assert( 0 ); + break; + + case TGSI_OPCODE_DP2: + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_IF: + /* push CondMask */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + FETCH( &r[0], 0, CHAN_X ); + /* update CondMask */ + if( ! r[0].u[0] ) { + mach->CondMask &= ~0x1; + } + if( ! r[0].u[1] ) { + mach->CondMask &= ~0x2; + } + if( ! r[0].u[2] ) { + mach->CondMask &= ~0x4; + } + if( ! r[0].u[3] ) { + mach->CondMask &= ~0x8; + } + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ELSE */ + break; + + case TGSI_OPCODE_ELSE: + /* invert CondMask wrt previous mask */ + { + uint prevMask; + assert(mach->CondStackTop > 0); + prevMask = mach->CondStack[mach->CondStackTop - 1]; + mach->CondMask = ~mach->CondMask & prevMask; + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ENDIF */ + } + break; + + case TGSI_OPCODE_ENDIF: + /* pop CondMask */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_END: + /* halt execution */ + *pc = -1; + break; + + case TGSI_OPCODE_REP: + assert (0); + break; + + case TGSI_OPCODE_ENDREP: + assert (0); + break; + + case TGSI_OPCODE_PUSHA: + assert (0); + break; + + case TGSI_OPCODE_POPA: + assert (0); + break; + + case TGSI_OPCODE_CEIL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ceil( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_I2F: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_i2f( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_NOT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_not( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TRUNC: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_trunc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_shl( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ishr( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_AND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_and( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_OR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_or( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOD: + assert (0); + break; + + case TGSI_OPCODE_XOR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_xor( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SAD: + assert (0); + break; + + case TGSI_OPCODE_TXF: + assert (0); + break; + + case TGSI_OPCODE_TXQ: + assert (0); + break; + + case TGSI_OPCODE_EMIT: + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; + break; + + case TGSI_OPCODE_ENDPRIM: + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; + break; + + case TGSI_OPCODE_LOOP: + /* fall-through (for now) */ + case TGSI_OPCODE_BGNLOOP2: + /* push LoopMask and ContMasks */ + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + break; + + case TGSI_OPCODE_ENDLOOP: + /* fall-through (for now at least) */ + case TGSI_OPCODE_ENDLOOP2: + /* Restore ContMask, but don't pop */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; + if (mach->LoopMask) { + /* repeat loop: jump to instruction just past BGNLOOP */ + *pc = inst->InstructionExtLabel.Label + 1; + } + else { + /* exit loop: pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + /* pop ContMask */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + } + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BRK: + /* turn off loop channels for each enabled exec channel */ + mach->LoopMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_CONT: + /* turn off cont channels for each enabled exec channel */ + mach->ContMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BGNSUB: + /* no-op */ + break; + + case TGSI_OPCODE_ENDSUB: + /* no-op */ + break; + + case TGSI_OPCODE_NOISE1: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE2: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE3: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE4: + assert( 0 ); + break; + + case TGSI_OPCODE_NOP: + break; + + default: + assert( 0 ); + } +} + + +/** + * Run TGSI interpreter. + * \return bitmask of "alive" quad components + */ +uint +tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) +{ + uint i; + int pc = 0; + + mach->CondMask = 0xf; + mach->LoopMask = 0xf; + mach->ContMask = 0xf; + mach->FuncMask = 0xf; + mach->ExecMask = 0xf; + + mach->CondStackTop = 0; /* temporarily subvert this assertion */ + assert(mach->CondStackTop == 0); + assert(mach->LoopStackTop == 0); + assert(mach->ContStackTop == 0); + assert(mach->CallStackTop == 0); + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; + + if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; + mach->Primitives[0] = 0; + } + + + /* execute declarations (interpolants) */ + for (i = 0; i < mach->NumDeclarations; i++) { + exec_declaration( mach, mach->Declarations+i ); + } + + /* execute instructions, until pc is set to -1 */ + while (pc != -1) { + assert(pc < mach->NumInstructions); + exec_instruction( mach, mach->Instructions + pc, &pc ); + } + +#if 0 + /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ + if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { + /* + * Scale back depth component. + */ + for (i = 0; i < 4; i++) + mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; + } +#endif + + return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; +} + + diff --git a/src/gallium/aux/tgsi/exec/tgsi_exec.h b/src/gallium/aux/tgsi/exec/tgsi_exec.h new file mode 100644 index 0000000000..1fb66ee960 --- /dev/null +++ b/src/gallium/aux/tgsi/exec/tgsi_exec.h @@ -0,0 +1,239 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#if !defined TGSI_EXEC_H +#define TGSI_EXEC_H + +#include "pipe/p_compiler.h" + +#if defined __cplusplus +extern "C" { +#endif + +#define NUM_CHANNELS 4 /* R,G,B,A */ +#define QUAD_SIZE 4 /* 4 pixel/quad */ + +/** + * Registers may be treated as float, signed int or unsigned int. + */ +union tgsi_exec_channel +{ + float f[QUAD_SIZE]; + int i[QUAD_SIZE]; + unsigned u[QUAD_SIZE]; +}; + +/** + * A vector[RGBA] of channels[4 pixels] + */ +struct tgsi_exec_vector +{ + union tgsi_exec_channel xyzw[NUM_CHANNELS]; +}; + +/** + * For fragment programs, information for computing fragment input + * values from plane equation of the triangle/line. + */ +struct tgsi_interp_coef +{ + float a0[NUM_CHANNELS]; /* in an xyzw layout */ + float dadx[NUM_CHANNELS]; + float dady[NUM_CHANNELS]; +}; + + +struct softpipe_tile_cache; /**< Opaque to TGSI */ + +/** + * Information for sampling textures, which must be implemented + * by code outside the TGSI executor. + */ +struct tgsi_sampler +{ + const struct pipe_sampler_state *state; + struct pipe_texture *texture; + /** Get samples for four fragments in a quad */ + void (*get_samples)(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); + void *pipe; /*XXX temporary*/ + struct softpipe_tile_cache *cache; +}; + +/** + * For branching/calling subroutines. + */ +struct tgsi_exec_labels +{ + unsigned labels[128][2]; + unsigned count; +}; + +/* + * Locations of various utility registers (_I = Index, _C = Channel) + */ +#define TGSI_EXEC_TEMP_00000000_I 32 +#define TGSI_EXEC_TEMP_00000000_C 0 + +#define TGSI_EXEC_TEMP_7FFFFFFF_I 32 +#define TGSI_EXEC_TEMP_7FFFFFFF_C 1 + +#define TGSI_EXEC_TEMP_80000000_I 32 +#define TGSI_EXEC_TEMP_80000000_C 2 + +#define TGSI_EXEC_TEMP_FFFFFFFF_I 32 +#define TGSI_EXEC_TEMP_FFFFFFFF_C 3 + +#define TGSI_EXEC_TEMP_ONE_I 33 +#define TGSI_EXEC_TEMP_ONE_C 0 + +#define TGSI_EXEC_TEMP_TWO_I 33 +#define TGSI_EXEC_TEMP_TWO_C 1 + +#define TGSI_EXEC_TEMP_128_I 33 +#define TGSI_EXEC_TEMP_128_C 2 + +#define TGSI_EXEC_TEMP_MINUS_128_I 33 +#define TGSI_EXEC_TEMP_MINUS_128_C 3 + +#define TGSI_EXEC_TEMP_KILMASK_I 34 +#define TGSI_EXEC_TEMP_KILMASK_C 0 + +#define TGSI_EXEC_TEMP_OUTPUT_I 34 +#define TGSI_EXEC_TEMP_OUTPUT_C 1 + +#define TGSI_EXEC_TEMP_PRIMITIVE_I 34 +#define TGSI_EXEC_TEMP_PRIMITIVE_C 2 + +#define TGSI_EXEC_TEMP_R0 35 + +#define TGSI_EXEC_NUM_TEMPS (32 + 4) +#define TGSI_EXEC_NUM_ADDRS 1 +#define TGSI_EXEC_NUM_IMMEDIATES 256 + +#define TGSI_EXEC_MAX_COND_NESTING 10 +#define TGSI_EXEC_MAX_LOOP_NESTING 10 +#define TGSI_EXEC_MAX_CALL_NESTING 10 + +/** + * Run-time virtual machine state for executing TGSI shader. + */ +struct tgsi_exec_machine +{ + /* + * 32 program temporaries + * 4 internal temporaries + * 1 address + * 1 temporary of padding to align to 16 bytes + */ + struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS + 1]; + + /* + * This will point to _Temps after aligning to 16B boundary. + */ + struct tgsi_exec_vector *Temps; + struct tgsi_exec_vector *Addrs; + + struct tgsi_sampler *Samplers; + + float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; + unsigned ImmLimit; + float (*Consts)[4]; + struct tgsi_exec_vector *Inputs; + struct tgsi_exec_vector *Outputs; + const struct tgsi_token *Tokens; + unsigned Processor; + + /* GEOMETRY processor only. */ + unsigned *Primitives; + + /* FRAGMENT processor only. */ + const struct tgsi_interp_coef *InterpCoefs; + struct tgsi_exec_vector QuadPos; + + /* Conditional execution masks */ + uint CondMask; /**< For IF/ELSE/ENDIF */ + uint LoopMask; /**< For BGNLOOP/ENDLOOP */ + uint ContMask; /**< For loop CONT statements */ + uint FuncMask; /**< For function calls */ + uint ExecMask; /**< = CondMask & LoopMask */ + + /** Condition mask stack (for nested conditionals) */ + uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; + int CondStackTop; + + /** Loop mask stack (for nested loops) */ + uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopStackTop; + + /** Loop continue mask stack (see comments in tgsi_exec.c) */ + uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int ContStackTop; + + /** Function execution mask stack (for executing subroutine code) */ + uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; + int FuncStackTop; + + /** Function call stack for saving/restoring the program counter */ + uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; + int CallStackTop; + + struct tgsi_full_instruction *Instructions; + uint NumInstructions; + + struct tgsi_full_declaration *Declarations; + uint NumDeclarations; + + struct tgsi_exec_labels Labels; +}; + + +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach, + const struct tgsi_token *tokens, + unsigned numSamplers, + struct tgsi_sampler *samplers); + +uint +tgsi_exec_machine_run( + struct tgsi_exec_machine *mach ); + + +void +tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach); + + +#if defined __cplusplus +} /* extern "C" */ +#endif + +#endif /* TGSI_EXEC_H */ diff --git a/src/gallium/aux/tgsi/exec/tgsi_sse2.c b/src/gallium/aux/tgsi/exec/tgsi_sse2.c new file mode 100755 index 0000000000..1e56e4afb6 --- /dev/null +++ b/src/gallium/aux/tgsi/exec/tgsi_sse2.c @@ -0,0 +1,2378 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/util/tgsi_util.h" +#include "tgsi_exec.h" +#include "tgsi_sse2.h" + +#include "x86/rtasm/x86sse.h" + +#if defined(__i386__) || defined(__386__) + +#define DUMP_SSE 0 + +#if DUMP_SSE + +static void +_print_reg( + struct x86_reg reg ) +{ + if (reg.mod != mod_REG) + debug_printf( "[" ); + + switch( reg.file ) { + case file_REG32: + switch( reg.idx ) { + case reg_AX: + debug_printf( "EAX" ); + break; + case reg_CX: + debug_printf( "ECX" ); + break; + case reg_DX: + debug_printf( "EDX" ); + break; + case reg_BX: + debug_printf( "EBX" ); + break; + case reg_SP: + debug_printf( "ESP" ); + break; + case reg_BP: + debug_printf( "EBP" ); + break; + case reg_SI: + debug_printf( "ESI" ); + break; + case reg_DI: + debug_printf( "EDI" ); + break; + } + break; + case file_MMX: + assert( 0 ); + break; + case file_XMM: + debug_printf( "XMM%u", reg.idx ); + break; + case file_x87: + assert( 0 ); + break; + } + + if (reg.mod == mod_DISP8 || + reg.mod == mod_DISP32) + debug_printf("+%d", reg.disp); + + if (reg.mod != mod_REG) + debug_printf( "]" ); +} + +static void +_fill( + const char *op ) +{ + unsigned count = 10 - strlen( op ); + + while( count-- ) { + debug_printf( " " ); + } +} + +#define DUMP_START() debug_printf( "\nsse-dump start ----------------" ) +#define DUMP_END() debug_printf( "\nsse-dump end ----------------\n" ) +#define DUMP( OP ) debug_printf( "\n%s", OP ) +#define DUMP_I( OP, I ) do {\ + debug_printf( "\n%s", OP );\ + _fill( OP );\ + debug_printf( "%u", I ); } while( 0 ) +#define DUMP_R( OP, R0 ) do {\ + debug_printf( "\n%s", OP );\ + _fill( OP );\ + _print_reg( R0 ); } while( 0 ) +#define DUMP_RR( OP, R0, R1 ) do {\ + debug_printf( "\n%s", OP );\ + _fill( OP );\ + _print_reg( R0 );\ + debug_printf( ", " );\ + _print_reg( R1 ); } while( 0 ) +#define DUMP_RRI( OP, R0, R1, I ) do {\ + debug_printf( "\n%s", OP );\ + _fill( OP );\ + _print_reg( R0 );\ + debug_printf( ", " );\ + _print_reg( R1 );\ + debug_printf( ", " );\ + debug_printf( "%u", I ); } while( 0 ) + +#else + +#define DUMP_START() +#define DUMP_END() +#define DUMP( OP ) +#define DUMP_I( OP, I ) +#define DUMP_R( OP, R0 ) +#define DUMP_RR( OP, R0, R1 ) +#define DUMP_RRI( OP, R0, R1, I ) + +#endif + +#define FOR_EACH_CHANNEL( CHAN )\ + for( CHAN = 0; CHAN < 4; CHAN++ ) + +#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ + if( IS_DST0_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\ + FOR_EACH_CHANNEL( CHAN )\ + IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN ) + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +/** + * X86 utility functions. + */ + +static struct x86_reg +make_xmm( + unsigned xmm ) +{ + return x86_make_reg( + file_XMM, + (enum x86_reg_name) xmm ); +} + +/** + * X86 register mapping helpers. + */ + +static struct x86_reg +get_const_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_CX ); +} + +static struct x86_reg +get_input_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_AX ); +} + +static struct x86_reg +get_output_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_DX ); +} + +static struct x86_reg +get_temp_base( void ) +{ +#ifdef WIN32 + return x86_make_reg( + file_REG32, + reg_BX ); +#else + return x86_make_reg( + file_REG32, + reg_SI ); +#endif +} + +static struct x86_reg +get_coef_base( void ) +{ + return get_output_base(); +} + +/** + * Data access helpers. + */ + +static struct x86_reg +get_argument( + unsigned index ) +{ + return x86_make_disp( + x86_make_reg( file_REG32, reg_SP ), + (index + 1) * 4 ); +} + +static struct x86_reg +get_const( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_const_base(), + (vec * 4 + chan) * 4 ); +} + +static struct x86_reg +get_input( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_input_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_output( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_output_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_temp( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_temp_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_coef( + unsigned vec, + unsigned chan, + unsigned member ) +{ + return x86_make_disp( + get_coef_base(), + ((vec * 3 + member) * 4 + chan) * 4 ); +} + +/** + * X86 rtasm wrappers. + */ + +static void +emit_addps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ADDPS", dst, src ); + sse_addps( func, dst, src ); +} + +static void +emit_andnps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ANDNPS", dst, src ); + sse_andnps( func, dst, src ); +} + +static void +emit_andps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ANDPS", dst, src ); + sse_andps( func, dst, src ); +} + +static void +emit_call( + struct x86_function *func, + void (* addr)() ) +{ + struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); + + DUMP_I( "CALL", addr ); + x86_mov_reg_imm( func, ecx, (unsigned long) addr ); + x86_call( func, ecx ); +} + +static void +emit_cmpps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src, + enum sse_cc cc ) +{ + DUMP_RRI( "CMPPS", dst, src, cc ); + sse_cmpps( func, dst, src, cc ); +} + +static void +emit_cvttps2dq( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "CVTTPS2DQ", dst, src ); + sse2_cvttps2dq( func, dst, src ); +} + +static void +emit_maxps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MAXPS", dst, src ); + sse_maxps( func, dst, src ); +} + +static void +emit_minps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MINPS", dst, src ); + sse_minps( func, dst, src ); +} + +static void +emit_mov( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOV", dst, src ); + x86_mov( func, dst, src ); +} + +static void +emit_movaps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOVAPS", dst, src ); + sse_movaps( func, dst, src ); +} + +static void +emit_movss( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOVSS", dst, src ); + sse_movss( func, dst, src ); +} + +static void +emit_movups( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOVUPS", dst, src ); + sse_movups( func, dst, src ); +} + +static void +emit_mulps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MULPS", dst, src ); + sse_mulps( func, dst, src ); +} + +static void +emit_or( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "OR", dst, src ); + x86_or( func, dst, src ); +} + +static void +emit_orps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ORPS", dst, src ); + sse_orps( func, dst, src ); +} + +static void +emit_pmovmskb( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "PMOVMSKB", dst, src ); + sse_pmovmskb( func, dst, src ); +} + +static void +emit_pop( + struct x86_function *func, + struct x86_reg dst ) +{ + DUMP_R( "POP", dst ); + x86_pop( func, dst ); +} + +static void +emit_push( + struct x86_function *func, + struct x86_reg dst ) +{ + DUMP_R( "PUSH", dst ); + x86_push( func, dst ); +} + +static void +emit_rcpps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "RCPPS", dst, src ); + sse2_rcpps( func, dst, src ); +} + +#ifdef WIN32 +static void +emit_retw( + struct x86_function *func, + unsigned size ) +{ + DUMP_I( "RET", size ); + x86_retw( func, size ); +} +#else +static void +emit_ret( + struct x86_function *func ) +{ + DUMP( "RET" ); + x86_ret( func ); +} +#endif + +static void +emit_rsqrtps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "RSQRTPS", dst, src ); + sse_rsqrtps( func, dst, src ); +} + +static void +emit_shufps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src, + unsigned char shuf ) +{ + DUMP_RRI( "SHUFPS", dst, src, shuf ); + sse_shufps( func, dst, src, shuf ); +} + +static void +emit_subps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "SUBPS", dst, src ); + sse_subps( func, dst, src ); +} + +static void +emit_xorps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "XORPS", dst, src ); + sse_xorps( func, dst, src ); +} + +/** + * Data fetch helpers. + */ + +static void +emit_const( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movss( + func, + make_xmm( xmm ), + get_const( vec, chan ) ); + emit_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + +static void +emit_inputf( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movups( + func, + make_xmm( xmm ), + get_input( vec, chan ) ); +} + +static void +emit_output( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movups( + func, + get_output( vec, chan ), + make_xmm( xmm ) ); +} + +static void +emit_tempf( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movaps( + func, + make_xmm( xmm ), + get_temp( vec, chan ) ); +} + +static void +emit_coef( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan, + unsigned member ) +{ + emit_movss( + func, + make_xmm( xmm ), + get_coef( vec, chan, member ) ); + emit_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + +/** + * Data store helpers. + */ + +static void +emit_inputs( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movups( + func, + get_input( vec, chan ), + make_xmm( xmm ) ); +} + +static void +emit_temps( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movaps( + func, + get_temp( vec, chan ), + make_xmm( xmm ) ); +} + +static void +emit_addrs( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_temps( + func, + xmm, + vec + TGSI_EXEC_NUM_TEMPS, + chan ); +} + +/** + * Coefficent fetch helpers. + */ + +static void +emit_coef_a0( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 0 ); +} + +static void +emit_coef_dadx( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 1 ); +} + +static void +emit_coef_dady( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 2 ); +} + +/** + * Function call helpers. + */ + +static void +emit_push_gp( + struct x86_function *func ) +{ + emit_push( + func, + get_const_base() ); + emit_push( + func, + get_input_base() ); + emit_push( + func, + get_output_base() ); + + /* It is important on non-win32 platforms that temp base is pushed last. + */ + emit_push( + func, + get_temp_base() ); +} + +static void +emit_pop_gp( + struct x86_function *func ) +{ + /* Restore GP registers in a reverse order. + */ + emit_pop( + func, + get_temp_base() ); + emit_pop( + func, + get_output_base() ); + emit_pop( + func, + get_input_base() ); + emit_pop( + func, + get_const_base() ); +} + +static void +emit_func_call_dst( + struct x86_function *func, + unsigned xmm_dst, + void (*code)() ) +{ + emit_movaps( + func, + get_temp( TEMP_R0, 0 ), + make_xmm( xmm_dst ) ); + + emit_push_gp( + func ); + +#ifdef WIN32 + emit_push( + func, + get_temp( TEMP_R0, 0 ) ); +#endif + + emit_call( + func, + code ); + + emit_pop_gp( + func ); + + emit_movaps( + func, + make_xmm( xmm_dst ), + get_temp( TEMP_R0, 0 ) ); +} + +static void +emit_func_call_dst_src( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src, + void (*code)() ) +{ + emit_movaps( + func, + get_temp( TEMP_R0, 1 ), + make_xmm( xmm_src ) ); + + emit_func_call_dst( + func, + xmm_dst, + code ); +} + +/** + * Low-level instruction translators. + */ + +static void +emit_abs( + struct x86_function *func, + unsigned xmm ) +{ + emit_andps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_7FFFFFFF_I, + TGSI_EXEC_TEMP_7FFFFFFF_C ) ); +} + +static void +emit_add( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_addps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void XSTDCALL +cos4f( + float *store ) +{ +#ifdef WIN32 + store[0] = (float) cos( (double) store[0] ); + store[1] = (float) cos( (double) store[1] ); + store[2] = (float) cos( (double) store[2] ); + store[3] = (float) cos( (double) store[3] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = cosf( store[X + 0] ); + store[X + 1] = cosf( store[X + 1] ); + store[X + 2] = cosf( store[X + 2] ); + store[X + 3] = cosf( store[X + 3] ); +#endif +} + +static void +emit_cos( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + cos4f ); +} + +static void XSTDCALL +ex24f( + float *store ) +{ +#ifdef WIN32 + store[0] = (float) pow( 2.0, (double) store[0] ); + store[1] = (float) pow( 2.0, (double) store[1] ); + store[2] = (float) pow( 2.0, (double) store[2] ); + store[3] = (float) pow( 2.0, (double) store[3] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = powf( 2.0f, store[X + 0] ); + store[X + 1] = powf( 2.0f, store[X + 1] ); + store[X + 2] = powf( 2.0f, store[X + 2] ); + store[X + 3] = powf( 2.0f, store[X + 3] ); +#endif +} + +static void +emit_ex2( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + ex24f ); +} + +static void +emit_f2it( + struct x86_function *func, + unsigned xmm ) +{ + emit_cvttps2dq( + func, + make_xmm( xmm ), + make_xmm( xmm ) ); +} + +static void XSTDCALL +flr4f( + float *store ) +{ +#ifdef WIN32 + const unsigned X = 0; +#else + const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] = (float) floor( (double) store[X + 0] ); + store[X + 1] = (float) floor( (double) store[X + 1] ); + store[X + 2] = (float) floor( (double) store[X + 2] ); + store[X + 3] = (float) floor( (double) store[X + 3] ); +} + +static void +emit_flr( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + flr4f ); +} + +static void XSTDCALL +frc4f( + float *store ) +{ +#ifdef WIN32 + const unsigned X = 0; +#else + const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] -= (float) floor( (double) store[X + 0] ); + store[X + 1] -= (float) floor( (double) store[X + 1] ); + store[X + 2] -= (float) floor( (double) store[X + 2] ); + store[X + 3] -= (float) floor( (double) store[X + 3] ); +} + +static void +emit_frc( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + frc4f ); +} + +static void XSTDCALL +lg24f( + float *store ) +{ +#ifdef WIN32 + const unsigned X = 0; +#else + const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] = LOG2( store[X + 0] ); + store[X + 1] = LOG2( store[X + 1] ); + store[X + 2] = LOG2( store[X + 2] ); + store[X + 3] = LOG2( store[X + 3] ); +} + +static void +emit_lg2( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + lg24f ); +} + +static void +emit_MOV( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_movups( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_mul (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + emit_mulps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_neg( + struct x86_function *func, + unsigned xmm ) +{ + emit_xorps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); +} + +static void XSTDCALL +pow4f( + float *store ) +{ +#ifdef WIN32 + store[0] = (float) pow( (double) store[0], (double) store[4] ); + store[1] = (float) pow( (double) store[1], (double) store[5] ); + store[2] = (float) pow( (double) store[2], (double) store[6] ); + store[3] = (float) pow( (double) store[3], (double) store[7] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = powf( store[X + 0], store[X + 4] ); + store[X + 1] = powf( store[X + 1], store[X + 5] ); + store[X + 2] = powf( store[X + 2], store[X + 6] ); + store[X + 3] = powf( store[X + 3], store[X + 7] ); +#endif +} + +static void +emit_pow( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_func_call_dst_src( + func, + xmm_dst, + xmm_src, + pow4f ); +} + +static void +emit_rcp ( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_rcpps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_rsqrt( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_rsqrtps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_setsign( + struct x86_function *func, + unsigned xmm ) +{ + emit_orps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); +} + +static void XSTDCALL +sin4f( + float *store ) +{ +#ifdef WIN32 + store[0] = (float) sin( (double) store[0] ); + store[1] = (float) sin( (double) store[1] ); + store[2] = (float) sin( (double) store[2] ); + store[3] = (float) sin( (double) store[3] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = sinf( store[X + 0] ); + store[X + 1] = sinf( store[X + 1] ); + store[X + 2] = sinf( store[X + 2] ); + store[X + 3] = sinf( store[X + 3] ); +#endif +} + +static void +emit_sin (struct x86_function *func, + unsigned xmm_dst) +{ + emit_func_call_dst( + func, + xmm_dst, + sin4f ); +} + +static void +emit_sub( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_subps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +/** + * Register fetch. + */ + +static void +emit_fetch( + struct x86_function *func, + unsigned xmm, + const struct tgsi_full_src_register *reg, + const unsigned chan_index ) +{ + unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( reg->SrcRegister.File ) { + case TGSI_FILE_CONSTANT: + emit_const( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + case TGSI_FILE_INPUT: + emit_inputf( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + case TGSI_FILE_TEMPORARY: + emit_tempf( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + emit_tempf( + func, + xmm, + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ); + break; + + case TGSI_EXTSWIZZLE_ONE: + emit_tempf( + func, + xmm, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + break; + + default: + assert( 0 ); + } + + switch( tgsi_util_get_full_src_register_sign_mode( reg, chan_index ) ) { + case TGSI_UTIL_SIGN_CLEAR: + emit_abs( func, xmm ); + break; + + case TGSI_UTIL_SIGN_SET: + emit_setsign( func, xmm ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + emit_neg( func, xmm ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } +} + +#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\ + emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN ) + +/** + * Register store. + */ + +static void +emit_store( + struct x86_function *func, + unsigned xmm, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + unsigned chan_index ) +{ + switch( reg->DstRegister.File ) { + case TGSI_FILE_OUTPUT: + emit_output( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; + + case TGSI_FILE_TEMPORARY: + emit_temps( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; + + case TGSI_FILE_ADDRESS: + emit_addrs( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; + + default: + assert( 0 ); + } + + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + + case TGSI_SAT_ZERO_ONE: +// assert( 0 ); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + } +} + +#define STORE( FUNC, INST, XMM, INDEX, CHAN )\ + emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) + +/** + * High-level instruction translators. + */ + +static void +emit_kil( + struct x86_function *func, + const struct tgsi_full_src_register *reg ) +{ + unsigned uniquemask; + unsigned registers[4]; + unsigned nextregister = 0; + unsigned firstchan = ~0; + unsigned chan_index; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + FOR_EACH_CHANNEL( chan_index ) { + unsigned swizzle; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle( + reg, + chan_index ); + + /* check if the component has not been already tested */ + if( !(uniquemask & (1 << swizzle)) ) { + uniquemask |= 1 << swizzle; + + /* allocate register */ + registers[chan_index] = nextregister; + emit_fetch( + func, + nextregister, + reg, + chan_index ); + nextregister++; + + /* mark the first channel used */ + if( firstchan == ~0 ) { + firstchan = chan_index; + } + } + } + + emit_push( + func, + x86_make_reg( file_REG32, reg_AX ) ); + emit_push( + func, + x86_make_reg( file_REG32, reg_DX ) ); + + FOR_EACH_CHANNEL( chan_index ) { + if( uniquemask & (1 << chan_index) ) { + emit_cmpps( + func, + make_xmm( registers[chan_index] ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ), + cc_LessThan ); + + if( chan_index == firstchan ) { + emit_pmovmskb( + func, + x86_make_reg( file_REG32, reg_AX ), + make_xmm( registers[chan_index] ) ); + } + else { + emit_pmovmskb( + func, + x86_make_reg( file_REG32, reg_DX ), + make_xmm( registers[chan_index] ) ); + emit_or( + func, + x86_make_reg( file_REG32, reg_AX ), + x86_make_reg( file_REG32, reg_DX ) ); + } + } + } + + emit_or( + func, + get_temp( + TGSI_EXEC_TEMP_KILMASK_I, + TGSI_EXEC_TEMP_KILMASK_C ), + x86_make_reg( file_REG32, reg_AX ) ); + + emit_pop( + func, + x86_make_reg( file_REG32, reg_DX ) ); + emit_pop( + func, + x86_make_reg( file_REG32, reg_AX ) ); +} + +static void +emit_setcc( + struct x86_function *func, + struct tgsi_full_instruction *inst, + enum sse_cc cc ) +{ + unsigned chan_index; + + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_cmpps( + func, + make_xmm( 0 ), + make_xmm( 1 ), + cc ); + emit_andps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ) ); + STORE( func, *inst, 0, 0, chan_index ); + } +} + +static void +emit_cmp( + struct x86_function *func, + struct tgsi_full_instruction *inst ) +{ + unsigned chan_index; + + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_cmpps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ), + cc_LessThan ); + emit_andps( + func, + make_xmm( 1 ), + make_xmm( 0 ) ); + emit_andnps( + func, + make_xmm( 0 ), + make_xmm( 2 ) ); + emit_orps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } +} + +static int +emit_instruction( + struct x86_function *func, + struct tgsi_full_instruction *inst ) +{ + unsigned chan_index; + + switch( inst->Instruction.Opcode ) { + case TGSI_OPCODE_ARL: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_f2it( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C); + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + STORE( func, *inst, 0, 0, CHAN_X ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( func, *inst, 0, 0, CHAN_W ); + } + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_maxps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ) ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 1, 0, CHAN_Y ); + emit_maxps( + func, + make_xmm( 1 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ) ); + FETCH( func, *inst, 2, 0, CHAN_W ); + emit_minps( + func, + make_xmm( 2 ), + get_temp( + TGSI_EXEC_TEMP_128_I, + TGSI_EXEC_TEMP_128_C ) ); + emit_maxps( + func, + make_xmm( 2 ), + get_temp( + TGSI_EXEC_TEMP_MINUS_128_I, + TGSI_EXEC_TEMP_MINUS_128_C ) ); + emit_pow( func, 1, 2 ); + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_xorps( + func, + make_xmm( 2 ), + make_xmm( 2 ) ); + emit_cmpps( + func, + make_xmm( 2 ), + make_xmm( 0 ), + cc_LessThanEqual ); + emit_andps( + func, + make_xmm( 2 ), + make_xmm( 1 ) ); + STORE( func, *inst, 2, 0, CHAN_Z ); + } + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_rcp( func, 0, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_rsqrt( func, 0, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + return 0; + break; + + case TGSI_OPCODE_LOG: + return 0; + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_mul( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_add( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul(func, 1, 2 ); + emit_add(func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_W ); + FETCH( func, *inst, 2, 1, CHAN_W ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + FETCH( func, *inst, 1, 1, CHAN_Y ); + emit_mul( func, 0, 1 ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + FETCH( func, *inst, 0, 0, CHAN_Z ); + STORE( func, *inst, 0, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, 1, CHAN_W ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_minps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_maxps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + emit_setcc( func, inst, cc_LessThan ); + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + emit_setcc( func, inst, cc_NotLessThan ); + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_mul( func, 0, 1 ); + emit_add( func, 0, 2 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_sub( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_sub( func, 1, 2 ); + emit_mul( func, 0, 1 ); + emit_add( func, 0, 2 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CND: + return 0; + break; + + case TGSI_OPCODE_CND0: + return 0; + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + return 0; + break; + + case TGSI_OPCODE_INDEX: + return 0; + break; + + case TGSI_OPCODE_NEGATE: + return 0; + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_frc( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + return 0; + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_flr( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + return 0; + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_ex2( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_lg2( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_pow( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( func, *inst, 1, 1, CHAN_Z ); + FETCH( func, *inst, 3, 0, CHAN_Z ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + FETCH( func, *inst, 4, 1, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + emit_MOV( func, 2, 0 ); + emit_mul( func, 2, 1 ); + emit_MOV( func, 5, 3 ); + emit_mul( func, 5, 4 ); + emit_sub( func, 2, 5 ); + STORE( func, *inst, 2, 0, CHAN_X ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 2, 1, CHAN_X ); + FETCH( func, *inst, 5, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + emit_mul( func, 3, 2 ); + emit_mul( func, 1, 5 ); + emit_sub( func, 3, 1 ); + STORE( func, *inst, 3, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + emit_mul( func, 5, 4 ); + emit_mul( func, 0, 2 ); + emit_sub( func, 5, 0 ); + STORE( func, *inst, 5, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + return 0; + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_abs( func, 0) ; + + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_RCC: + return 0; + break; + + case TGSI_OPCODE_DPH: + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 1, CHAN_W ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_cos( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + return 0; + break; + + case TGSI_OPCODE_DDY: + return 0; + break; + + case TGSI_OPCODE_KIL: + emit_kil( func, &inst->FullSrcRegisters[0] ); + break; + + case TGSI_OPCODE_PK2H: + return 0; + break; + + case TGSI_OPCODE_PK2US: + return 0; + break; + + case TGSI_OPCODE_PK4B: + return 0; + break; + + case TGSI_OPCODE_PK4UB: + return 0; + break; + + case TGSI_OPCODE_RFL: + return 0; + break; + + case TGSI_OPCODE_SEQ: + return 0; + break; + + case TGSI_OPCODE_SFL: + return 0; + break; + + case TGSI_OPCODE_SGT: + return 0; + break; + + case TGSI_OPCODE_SIN: + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_sin( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + return 0; + break; + + case TGSI_OPCODE_SNE: + return 0; + break; + + case TGSI_OPCODE_STR: + return 0; + break; + + case TGSI_OPCODE_TEX: + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_TXD: + return 0; + break; + + case TGSI_OPCODE_UP2H: + return 0; + break; + + case TGSI_OPCODE_UP2US: + return 0; + break; + + case TGSI_OPCODE_UP4B: + return 0; + break; + + case TGSI_OPCODE_UP4UB: + return 0; + break; + + case TGSI_OPCODE_X2D: + return 0; + break; + + case TGSI_OPCODE_ARA: + return 0; + break; + + case TGSI_OPCODE_ARR: + return 0; + break; + + case TGSI_OPCODE_BRA: + return 0; + break; + + case TGSI_OPCODE_CAL: + return 0; + break; + + case TGSI_OPCODE_RET: + case TGSI_OPCODE_END: +#ifdef WIN32 + emit_retw( func, 16 ); +#else + emit_ret( func ); +#endif + break; + + case TGSI_OPCODE_SSG: + return 0; + break; + + case TGSI_OPCODE_CMP: + emit_cmp (func, inst); + break; + + case TGSI_OPCODE_SCS: + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_cos( func, 0 ); + STORE( func, *inst, 0, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + emit_sin( func, 0 ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + FETCH( func, *inst, 0, TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ); + STORE( func, *inst, 0, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_TXB: + return 0; + break; + + case TGSI_OPCODE_NRM: + return 0; + break; + + case TGSI_OPCODE_DIV: + return 0; + break; + + case TGSI_OPCODE_DP2: + return 0; + break; + + case TGSI_OPCODE_TXL: + return 0; + break; + + case TGSI_OPCODE_BRK: + return 0; + break; + + case TGSI_OPCODE_IF: + return 0; + break; + + case TGSI_OPCODE_LOOP: + return 0; + break; + + case TGSI_OPCODE_REP: + return 0; + break; + + case TGSI_OPCODE_ELSE: + return 0; + break; + + case TGSI_OPCODE_ENDIF: + return 0; + break; + + case TGSI_OPCODE_ENDLOOP: + return 0; + break; + + case TGSI_OPCODE_ENDREP: + return 0; + break; + + case TGSI_OPCODE_PUSHA: + return 0; + break; + + case TGSI_OPCODE_POPA: + return 0; + break; + + case TGSI_OPCODE_CEIL: + return 0; + break; + + case TGSI_OPCODE_I2F: + return 0; + break; + + case TGSI_OPCODE_NOT: + return 0; + break; + + case TGSI_OPCODE_TRUNC: + return 0; + break; + + case TGSI_OPCODE_SHL: + return 0; + break; + + case TGSI_OPCODE_SHR: + return 0; + break; + + case TGSI_OPCODE_AND: + return 0; + break; + + case TGSI_OPCODE_OR: + return 0; + break; + + case TGSI_OPCODE_MOD: + return 0; + break; + + case TGSI_OPCODE_XOR: + return 0; + break; + + case TGSI_OPCODE_SAD: + return 0; + break; + + case TGSI_OPCODE_TXF: + return 0; + break; + + case TGSI_OPCODE_TXQ: + return 0; + break; + + case TGSI_OPCODE_CONT: + return 0; + break; + + case TGSI_OPCODE_EMIT: + return 0; + break; + + case TGSI_OPCODE_ENDPRIM: + return 0; + break; + + default: + return 0; + } + + return 1; +} + +static void +emit_declaration( + struct x86_function *func, + struct tgsi_full_declaration *decl ) +{ + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + unsigned i, j; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if( first == 0 ) { + mask &= ~TGSI_WRITEMASK_XY; + if( mask == TGSI_WRITEMASK_NONE ) { + first++; + } + } + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + switch( decl->Interpolation.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + emit_coef_a0( func, 0, i, j ); + emit_inputs( func, 0, i, j ); + break; + + case TGSI_INTERPOLATE_LINEAR: + emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_coef_dadx( func, 1, i, j ); + emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_coef_dady( func, 3, i, j ); + emit_mul( func, 0, 1 ); /* x * dadx */ + emit_coef_a0( func, 4, i, j ); + emit_mul( func, 2, 3 ); /* y * dady */ + emit_add( func, 0, 4 ); /* x * dadx + a0 */ + emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ + emit_inputs( func, 0, i, j ); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_coef_dadx( func, 1, i, j ); + emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_coef_dady( func, 3, i, j ); + emit_mul( func, 0, 1 ); /* x * dadx */ + emit_inputf( func, 4, 0, TGSI_SWIZZLE_W ); + emit_coef_a0( func, 5, i, j ); + emit_rcp( func, 4, 4 ); /* 1.0 / w */ + emit_mul( func, 2, 3 ); /* y * dady */ + emit_add( func, 0, 5 ); /* x * dadx + a0 */ + emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ + emit_mul( func, 0, 4 ); /* (x * dadx + y * dady + a0) / w */ + emit_inputs( func, 0, i, j ); + break; + + default: + assert( 0 ); + break; + } + } + } + } + } +} + +unsigned +tgsi_emit_sse2( + struct tgsi_token *tokens, + struct x86_function *func ) +{ + struct tgsi_parse_context parse; + unsigned ok = 1; + + DUMP_START(); + + func->csr = func->store; + + emit_mov( + func, + get_input_base(), + get_argument( 0 ) ); + emit_mov( + func, + get_output_base(), + get_argument( 1 ) ); + emit_mov( + func, + get_const_base(), + get_argument( 2 ) ); + emit_mov( + func, + get_temp_base(), + get_argument( 3 ) ); + + tgsi_parse_init( &parse, tokens ); + + while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + ok = emit_instruction( + func, + &parse.FullToken.FullInstruction ); + + if (!ok) { + debug_printf("failed to translate tgsi opcode %d\n", + parse.FullToken.FullInstruction.Instruction.Opcode ); + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* XXX implement this */ + ok = 0; + debug_printf("failed to emit immediate value\n"); + break; + + default: + assert( 0 ); + ok = 0; + break; + } + } + + tgsi_parse_free( &parse ); + + DUMP_END(); + + return ok; +} + +/** + * Fragment shaders are responsible for interpolating shader inputs. Because on + * x86 we have only 4 GP registers, and here we have 5 shader arguments (input, + * output, const, temp and coef), the code is split into two phases -- + * DECLARATION and INSTRUCTION phase. + * GP register holding the output argument is aliased with the coeff argument, + * as outputs are not needed in the DECLARATION phase. + */ +unsigned +tgsi_emit_sse2_fs( + struct tgsi_token *tokens, + struct x86_function *func ) +{ + struct tgsi_parse_context parse; + boolean instruction_phase = FALSE; + + DUMP_START(); + + func->csr = func->store; + + /* DECLARATION phase, do not load output argument. */ + emit_mov( + func, + get_input_base(), + get_argument( 0 ) ); + emit_mov( + func, + get_const_base(), + get_argument( 2 ) ); + emit_mov( + func, + get_temp_base(), + get_argument( 3 ) ); + emit_mov( + func, + get_coef_base(), + get_argument( 4 ) ); + + tgsi_parse_init( &parse, tokens ); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + emit_declaration( + func, + &parse.FullToken.FullDeclaration ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + if( !instruction_phase ) { + /* INSTRUCTION phase, overwrite coeff with output. */ + instruction_phase = TRUE; + emit_mov( + func, + get_output_base(), + get_argument( 1 ) ); + } + emit_instruction( + func, + &parse.FullToken.FullInstruction ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* XXX implement this */ + assert(0); + break; + + default: + assert( 0 ); + } + } + + tgsi_parse_free( &parse ); + + DUMP_END(); + + return 1; +} + +#endif /* i386 */ diff --git a/src/gallium/aux/tgsi/exec/tgsi_sse2.h b/src/gallium/aux/tgsi/exec/tgsi_sse2.h new file mode 100755 index 0000000000..9bee371766 --- /dev/null +++ b/src/gallium/aux/tgsi/exec/tgsi_sse2.h @@ -0,0 +1,26 @@ +#if !defined TGSI_SSE2_H +#define TGSI_SSE2_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_token; +struct x86_function; + +unsigned +tgsi_emit_sse2( + struct tgsi_token *tokens, + struct x86_function *function ); + +unsigned +tgsi_emit_sse2_fs( + struct tgsi_token *tokens, + struct x86_function *function ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_SSE2_H + diff --git a/src/gallium/aux/tgsi/util/tgsi_build.c b/src/gallium/aux/tgsi/util/tgsi_build.c new file mode 100644 index 0000000000..a00ff1c2a5 --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_build.c @@ -0,0 +1,1371 @@ +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_build.h" +#include "tgsi_parse.h" + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ) +{ + struct tgsi_version version; + + version.MajorVersion = 1; + version.MinorVersion = 1; + version.Padding = 0; + + return version; +} + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ) +{ + struct tgsi_header header; + + header.HeaderSize = 1; + header.BodySize = 0; + + return header; +} + +static void +header_headersize_grow( struct tgsi_header *header ) +{ + assert( header->HeaderSize < 0xFF ); + assert( header->BodySize == 0 ); + + header->HeaderSize++; +} + +static void +header_bodysize_grow( struct tgsi_header *header ) +{ + assert( header->BodySize < 0xFFFFFF ); + + header->BodySize++; +} + +struct tgsi_processor +tgsi_default_processor( void ) +{ + struct tgsi_processor processor; + + processor.Processor = TGSI_PROCESSOR_FRAGMENT; + processor.Padding = 0; + + return processor; +} + +struct tgsi_processor +tgsi_build_processor( + unsigned type, + struct tgsi_header *header ) +{ + struct tgsi_processor processor; + + processor = tgsi_default_processor(); + processor.Processor = type; + + header_headersize_grow( header ); + + return processor; +} + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ) +{ + struct tgsi_declaration declaration; + + declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; + declaration.Size = 1; + declaration.File = TGSI_FILE_NULL; + declaration.Declare = TGSI_DECLARE_RANGE; + declaration.UsageMask = TGSI_WRITEMASK_XYZW; + declaration.Interpolate = 0; + declaration.Semantic = 0; + declaration.Padding = 0; + declaration.Extended = 0; + + return declaration; +} + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned declare, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ) +{ + struct tgsi_declaration declaration; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( declare <= TGSI_DECLARE_MASK ); + + declaration = tgsi_default_declaration(); + declaration.File = file; + declaration.Declare = declare; + declaration.UsageMask = usage_mask; + declaration.Interpolate = interpolate; + declaration.Semantic = semantic; + + header_bodysize_grow( header ); + + return declaration; +} + +static void +declaration_grow( + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + assert( declaration->Size < 0xFF ); + + declaration->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ) +{ + struct tgsi_full_declaration full_declaration; + + full_declaration.Declaration = tgsi_default_declaration(); + full_declaration.Interpolation = tgsi_default_declaration_interpolation(); + full_declaration.Semantic = tgsi_default_declaration_semantic(); + + return full_declaration; +} + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0; + struct tgsi_declaration *declaration; + + if( maxsize <= size ) + return 0; + declaration = (struct tgsi_declaration *) &tokens[size]; + size++; + + *declaration = tgsi_build_declaration( + full_decl->Declaration.File, + full_decl->Declaration.Declare, + full_decl->Declaration.UsageMask, + full_decl->Declaration.Interpolate, + full_decl->Declaration.Semantic, + header ); + + switch( full_decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + { + struct tgsi_declaration_range *dr; + + if( maxsize <= size ) + return 0; + dr = (struct tgsi_declaration_range *) &tokens[size]; + size++; + + *dr = tgsi_build_declaration_range( + full_decl->u.DeclarationRange.First, + full_decl->u.DeclarationRange.Last, + declaration, + header ); + break; + } + + case TGSI_DECLARE_MASK: + { + struct tgsi_declaration_mask *dm; + + if( maxsize <= size ) + return 0; + dm = (struct tgsi_declaration_mask *) &tokens[size]; + size++; + + *dm = tgsi_build_declaration_mask( + full_decl->u.DeclarationMask.Mask, + declaration, + header ); + break; + } + + default: + assert( 0 ); + } + + if( full_decl->Declaration.Interpolate ) { + struct tgsi_declaration_interpolation *di; + + if( maxsize <= size ) + return 0; + di = (struct tgsi_declaration_interpolation *) &tokens[size]; + size++; + + *di = tgsi_build_declaration_interpolation( + full_decl->Interpolation.Interpolate, + declaration, + header ); + } + + if( full_decl->Declaration.Semantic ) { + struct tgsi_declaration_semantic *ds; + + if( maxsize <= size ) + return 0; + ds = (struct tgsi_declaration_semantic *) &tokens[size]; + size++; + + *ds = tgsi_build_declaration_semantic( + full_decl->Semantic.SemanticName, + full_decl->Semantic.SemanticIndex, + declaration, + header ); + } + + return size; +} + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_range declaration_range; + + assert( last >= first ); + assert( last <= 0xFFFF ); + + declaration_range.First = first; + declaration_range.Last = last; + + declaration_grow( declaration, header ); + + return declaration_range; +} + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + unsigned mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_mask declaration_mask; + + declaration_mask.Mask = mask; + + declaration_grow( declaration, header ); + + return declaration_mask; +} + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ) +{ + struct tgsi_declaration_interpolation di; + + di.Interpolate = TGSI_INTERPOLATE_CONSTANT; + di.Padding = 0; + + return di; +} + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + unsigned interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_interpolation di; + + assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); + + di = tgsi_default_declaration_interpolation(); + di.Interpolate = interpolate; + + declaration_grow( declaration, header ); + + return di; +} + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ) +{ + struct tgsi_declaration_semantic ds; + + ds.SemanticName = TGSI_SEMANTIC_POSITION; + ds.SemanticIndex = 0; + ds.Padding = 0; + + return ds; +} + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_semantic ds; + + assert( semantic_name <= TGSI_SEMANTIC_COUNT ); + assert( semantic_index <= 0xFFFF ); + + ds = tgsi_default_declaration_semantic(); + ds.SemanticName = semantic_name; + ds.SemanticIndex = semantic_index; + + declaration_grow( declaration, header ); + + return ds; +} + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ) +{ + struct tgsi_immediate immediate; + + immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + immediate.Size = 1; + immediate.DataType = TGSI_IMM_FLOAT32; + immediate.Padding = 0; + immediate.Extended = 0; + + return immediate; +} + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ) +{ + struct tgsi_immediate immediate; + + immediate = tgsi_default_immediate(); + + header_bodysize_grow( header ); + + return immediate; +} + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ) +{ + struct tgsi_full_immediate fullimm; + + fullimm.Immediate = tgsi_default_immediate(); + fullimm.u.Pointer = (void *) 0; + + return fullimm; +} + +static void +immediate_grow( + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + assert( immediate->Size < 0xFF ); + + immediate->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + float value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + struct tgsi_immediate_float32 immediate_float32; + + immediate_float32.Float = value; + + immediate_grow( immediate, header ); + + return immediate_float32; +} + +unsigned +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0, i; + struct tgsi_immediate *immediate; + + if( maxsize <= size ) + return 0; + immediate = (struct tgsi_immediate *) &tokens[size]; + size++; + + *immediate = tgsi_build_immediate( header ); + + for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { + struct tgsi_immediate_float32 *if32; + + if( maxsize <= size ) + return 0; + if32 = (struct tgsi_immediate_float32 *) &tokens[size]; + size++; + + *if32 = tgsi_build_immediate_float32( + full_imm->u.ImmediateFloat32[i].Float, + immediate, + header ); + } + + return size; +} + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ) +{ + struct tgsi_instruction instruction; + + instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; + instruction.Size = 1; + instruction.Opcode = TGSI_OPCODE_MOV; + instruction.Saturate = TGSI_SAT_NONE; + instruction.NumDstRegs = 1; + instruction.NumSrcRegs = 1; + instruction.Padding = 0; + instruction.Extended = 0; + + return instruction; +} + +struct tgsi_instruction +tgsi_build_instruction( + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header ) +{ + struct tgsi_instruction instruction; + + assert (opcode <= TGSI_OPCODE_LAST); + assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); + assert (num_dst_regs <= 3); + assert (num_src_regs <= 15); + + instruction = tgsi_default_instruction(); + instruction.Opcode = opcode; + instruction.Saturate = saturate; + instruction.NumDstRegs = num_dst_regs; + instruction.NumSrcRegs = num_src_regs; + + header_bodysize_grow( header ); + + return instruction; +} + +static void +instruction_grow( + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + assert (instruction->Size < 0xFF); + + instruction->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ) +{ + struct tgsi_full_instruction full_instruction; + unsigned i; + + full_instruction.Instruction = tgsi_default_instruction(); + full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); + full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); + full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); + for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { + full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); + } + for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { + full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); + } + + return full_instruction; +} + +unsigned +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0; + unsigned i; + struct tgsi_instruction *instruction; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + instruction = (struct tgsi_instruction *) &tokens[size]; + size++; + + *instruction = tgsi_build_instruction( + full_inst->Instruction.Opcode, + full_inst->Instruction.Saturate, + full_inst->Instruction.NumDstRegs, + full_inst->Instruction.NumSrcRegs, + header ); + prev_token = (struct tgsi_token *) instruction; + + if( tgsi_compare_instruction_ext_nv( + full_inst->InstructionExtNv, + tgsi_default_instruction_ext_nv() ) ) { + struct tgsi_instruction_ext_nv *instruction_ext_nv; + + if( maxsize <= size ) + return 0; + instruction_ext_nv = + (struct tgsi_instruction_ext_nv *) &tokens[size]; + size++; + + *instruction_ext_nv = tgsi_build_instruction_ext_nv( + full_inst->InstructionExtNv.Precision, + full_inst->InstructionExtNv.CondDstIndex, + full_inst->InstructionExtNv.CondFlowIndex, + full_inst->InstructionExtNv.CondMask, + full_inst->InstructionExtNv.CondSwizzleX, + full_inst->InstructionExtNv.CondSwizzleY, + full_inst->InstructionExtNv.CondSwizzleZ, + full_inst->InstructionExtNv.CondSwizzleW, + full_inst->InstructionExtNv.CondDstUpdate, + full_inst->InstructionExtNv.CondFlowEnable, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_nv; + } + + if( tgsi_compare_instruction_ext_label( + full_inst->InstructionExtLabel, + tgsi_default_instruction_ext_label() ) ) { + struct tgsi_instruction_ext_label *instruction_ext_label; + + if( maxsize <= size ) + return 0; + instruction_ext_label = + (struct tgsi_instruction_ext_label *) &tokens[size]; + size++; + + *instruction_ext_label = tgsi_build_instruction_ext_label( + full_inst->InstructionExtLabel.Label, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_label; + } + + if( tgsi_compare_instruction_ext_texture( + full_inst->InstructionExtTexture, + tgsi_default_instruction_ext_texture() ) ) { + struct tgsi_instruction_ext_texture *instruction_ext_texture; + + if( maxsize <= size ) + return 0; + instruction_ext_texture = + (struct tgsi_instruction_ext_texture *) &tokens[size]; + size++; + + *instruction_ext_texture = tgsi_build_instruction_ext_texture( + full_inst->InstructionExtTexture.Texture, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_texture; + } + + for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { + const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; + struct tgsi_dst_register *dst_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + dst_register = (struct tgsi_dst_register *) &tokens[size]; + size++; + + *dst_register = tgsi_build_dst_register( + reg->DstRegister.File, + reg->DstRegister.WriteMask, + reg->DstRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register; + + if( tgsi_compare_dst_register_ext_concode( + reg->DstRegisterExtConcode, + tgsi_default_dst_register_ext_concode() ) ) { + struct tgsi_dst_register_ext_concode *dst_register_ext_concode; + + if( maxsize <= size ) + return 0; + dst_register_ext_concode = + (struct tgsi_dst_register_ext_concode *) &tokens[size]; + size++; + + *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( + reg->DstRegisterExtConcode.CondMask, + reg->DstRegisterExtConcode.CondSwizzleX, + reg->DstRegisterExtConcode.CondSwizzleY, + reg->DstRegisterExtConcode.CondSwizzleZ, + reg->DstRegisterExtConcode.CondSwizzleW, + reg->DstRegisterExtConcode.CondSrcIndex, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_concode; + } + + if( tgsi_compare_dst_register_ext_modulate( + reg->DstRegisterExtModulate, + tgsi_default_dst_register_ext_modulate() ) ) { + struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; + + if( maxsize <= size ) + return 0; + dst_register_ext_modulate = + (struct tgsi_dst_register_ext_modulate *) &tokens[size]; + size++; + + *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( + reg->DstRegisterExtModulate.Modulate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_modulate; + } + } + + for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { + const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; + struct tgsi_src_register *src_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + src_register = (struct tgsi_src_register *) &tokens[size]; + size++; + + *src_register = tgsi_build_src_register( + reg->SrcRegister.File, + reg->SrcRegister.SwizzleX, + reg->SrcRegister.SwizzleY, + reg->SrcRegister.SwizzleZ, + reg->SrcRegister.SwizzleW, + reg->SrcRegister.Negate, + reg->SrcRegister.Indirect, + reg->SrcRegister.Dimension, + reg->SrcRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register; + + if( tgsi_compare_src_register_ext_swz( + reg->SrcRegisterExtSwz, + tgsi_default_src_register_ext_swz() ) ) { + struct tgsi_src_register_ext_swz *src_register_ext_swz; + + if( maxsize <= size ) + return 0; + src_register_ext_swz = + (struct tgsi_src_register_ext_swz *) &tokens[size]; + size++; + + *src_register_ext_swz = tgsi_build_src_register_ext_swz( + reg->SrcRegisterExtSwz.ExtSwizzleX, + reg->SrcRegisterExtSwz.ExtSwizzleY, + reg->SrcRegisterExtSwz.ExtSwizzleZ, + reg->SrcRegisterExtSwz.ExtSwizzleW, + reg->SrcRegisterExtSwz.NegateX, + reg->SrcRegisterExtSwz.NegateY, + reg->SrcRegisterExtSwz.NegateZ, + reg->SrcRegisterExtSwz.NegateW, + reg->SrcRegisterExtSwz.ExtDivide, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_swz; + } + + if( tgsi_compare_src_register_ext_mod( + reg->SrcRegisterExtMod, + tgsi_default_src_register_ext_mod() ) ) { + struct tgsi_src_register_ext_mod *src_register_ext_mod; + + if( maxsize <= size ) + return 0; + src_register_ext_mod = + (struct tgsi_src_register_ext_mod *) &tokens[size]; + size++; + + *src_register_ext_mod = tgsi_build_src_register_ext_mod( + reg->SrcRegisterExtMod.Complement, + reg->SrcRegisterExtMod.Bias, + reg->SrcRegisterExtMod.Scale2X, + reg->SrcRegisterExtMod.Absolute, + reg->SrcRegisterExtMod.Negate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_mod; + } + + if( reg->SrcRegister.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterInd.File, + reg->SrcRegisterInd.SwizzleX, + reg->SrcRegisterInd.SwizzleY, + reg->SrcRegisterInd.SwizzleZ, + reg->SrcRegisterInd.SwizzleW, + reg->SrcRegisterInd.Negate, + reg->SrcRegisterInd.Indirect, + reg->SrcRegisterInd.Dimension, + reg->SrcRegisterInd.Index, + instruction, + header ); + } + + if( reg->SrcRegister.Dimension ) { + struct tgsi_dimension *dim; + + assert( !reg->SrcRegisterDim.Dimension ); + + if( maxsize <= size ) + return 0; + dim = (struct tgsi_dimension *) &tokens[size]; + size++; + + *dim = tgsi_build_dimension( + reg->SrcRegisterDim.Indirect, + reg->SrcRegisterDim.Index, + instruction, + header ); + + if( reg->SrcRegisterDim.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterDimInd.File, + reg->SrcRegisterDimInd.SwizzleX, + reg->SrcRegisterDimInd.SwizzleY, + reg->SrcRegisterDimInd.SwizzleZ, + reg->SrcRegisterDimInd.SwizzleW, + reg->SrcRegisterDimInd.Negate, + reg->SrcRegisterDimInd.Indirect, + reg->SrcRegisterDimInd.Dimension, + reg->SrcRegisterDimInd.Index, + instruction, + header ); + } + } + } + + return size; +} + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; + instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; + instruction_ext_nv.CondDstIndex = 0; + instruction_ext_nv.CondFlowIndex = 0; + instruction_ext_nv.CondMask = TGSI_CC_TR; + instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; + instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; + instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; + instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; + instruction_ext_nv.CondDstUpdate = 0; + instruction_ext_nv.CondFlowEnable = 0; + instruction_ext_nv.Padding = 0; + instruction_ext_nv.Extended = 0; + + return instruction_ext_nv; +} + +union token_u32 +{ + unsigned u32; +}; + +unsigned +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv = tgsi_default_instruction_ext_nv(); + instruction_ext_nv.Precision = precision; + instruction_ext_nv.CondDstIndex = cond_dst_index; + instruction_ext_nv.CondFlowIndex = cond_flow_index; + instruction_ext_nv.CondMask = cond_mask; + instruction_ext_nv.CondSwizzleX = cond_swizzle_x; + instruction_ext_nv.CondSwizzleY = cond_swizzle_y; + instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; + instruction_ext_nv.CondSwizzleW = cond_swizzle_w; + instruction_ext_nv.CondDstUpdate = cond_dst_update; + instruction_ext_nv.CondFlowEnable = cond_flow_update; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_nv; +} + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; + instruction_ext_label.Label = 0; + instruction_ext_label.Padding = 0; + instruction_ext_label.Extended = 0; + + return instruction_ext_label; +} + +unsigned +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + unsigned label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label = tgsi_default_instruction_ext_label(); + instruction_ext_label.Label = label; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_label; +} + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; + instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; + instruction_ext_texture.Padding = 0; + instruction_ext_texture.Extended = 0; + + return instruction_ext_texture; +} + +unsigned +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + unsigned texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture = tgsi_default_instruction_ext_texture(); + instruction_ext_texture.Texture = texture; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_texture; +} + +struct tgsi_src_register +tgsi_default_src_register( void ) +{ + struct tgsi_src_register src_register; + + src_register.File = TGSI_FILE_NULL; + src_register.SwizzleX = TGSI_SWIZZLE_X; + src_register.SwizzleY = TGSI_SWIZZLE_Y; + src_register.SwizzleZ = TGSI_SWIZZLE_Z; + src_register.SwizzleW = TGSI_SWIZZLE_W; + src_register.Negate = 0; + src_register.Indirect = 0; + src_register.Dimension = 0; + src_register.Index = 0; + src_register.Extended = 0; + + return src_register; +} + +struct tgsi_src_register +tgsi_build_src_register( + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register src_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( negate <= 1 ); + assert( index >= -0x8000 && index <= 0x7FFF ); + + src_register = tgsi_default_src_register(); + src_register.File = file; + src_register.SwizzleX = swizzle_x; + src_register.SwizzleY = swizzle_y; + src_register.SwizzleZ = swizzle_z; + src_register.SwizzleW = swizzle_w; + src_register.Negate = negate; + src_register.Indirect = indirect; + src_register.Dimension = dimension; + src_register.Index = index; + + instruction_grow( instruction, header ); + + return src_register; +} + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ) +{ + struct tgsi_full_src_register full_src_register; + + full_src_register.SrcRegister = tgsi_default_src_register(); + full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); + full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); + full_src_register.SrcRegisterInd = tgsi_default_src_register(); + full_src_register.SrcRegisterDim = tgsi_default_dimension(); + full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); + + return full_src_register; +} + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; + src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; + src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; + src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; + src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; + src_register_ext_swz.NegateX = 0; + src_register_ext_swz.NegateY = 0; + src_register_ext_swz.NegateZ = 0; + src_register_ext_swz.NegateW = 0; + src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; + src_register_ext_swz.Padding = 0; + src_register_ext_swz.Extended = 0; + + return src_register_ext_swz; +} + +unsigned +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + unsigned ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE ); + assert( negate_x <= 1 ); + assert( negate_y <= 1 ); + assert( negate_z <= 1 ); + assert( negate_w <= 1 ); + assert( ext_divide <= TGSI_EXTSWIZZLE_ONE ); + + src_register_ext_swz = tgsi_default_src_register_ext_swz(); + src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; + src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; + src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; + src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; + src_register_ext_swz.NegateX = negate_x; + src_register_ext_swz.NegateY = negate_y; + src_register_ext_swz.NegateZ = negate_z; + src_register_ext_swz.NegateW = negate_w; + src_register_ext_swz.ExtDivide = ext_divide; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_swz; +} + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; + src_register_ext_mod.Complement = 0; + src_register_ext_mod.Bias = 0; + src_register_ext_mod.Scale2X = 0; + src_register_ext_mod.Absolute = 0; + src_register_ext_mod.Negate = 0; + src_register_ext_mod.Padding = 0; + src_register_ext_mod.Extended = 0; + + return src_register_ext_mod; +} + +unsigned +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + assert( complement <= 1 ); + assert( bias <= 1 ); + assert( scale_2x <= 1 ); + assert( absolute <= 1 ); + assert( negate <= 1 ); + + src_register_ext_mod = tgsi_default_src_register_ext_mod(); + src_register_ext_mod.Complement = complement; + src_register_ext_mod.Bias = bias; + src_register_ext_mod.Scale2X = scale_2x; + src_register_ext_mod.Absolute = absolute; + src_register_ext_mod.Negate = negate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_mod; +} + +struct tgsi_dimension +tgsi_default_dimension( void ) +{ + struct tgsi_dimension dimension; + + dimension.Indirect = 0; + dimension.Dimension = 0; + dimension.Padding = 0; + dimension.Index = 0; + dimension.Extended = 0; + + return dimension; +} + +struct tgsi_dimension +tgsi_build_dimension( + unsigned indirect, + unsigned index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dimension dimension; + + dimension = tgsi_default_dimension(); + dimension.Indirect = indirect; + dimension.Index = index; + + instruction_grow( instruction, header ); + + return dimension; +} + +struct tgsi_dst_register +tgsi_default_dst_register( void ) +{ + struct tgsi_dst_register dst_register; + + dst_register.File = TGSI_FILE_NULL; + dst_register.WriteMask = TGSI_WRITEMASK_XYZW; + dst_register.Indirect = 0; + dst_register.Dimension = 0; + dst_register.Index = 0; + dst_register.Padding = 0; + dst_register.Extended = 0; + + return dst_register; +} + +struct tgsi_dst_register +tgsi_build_dst_register( + unsigned file, + unsigned mask, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register dst_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( mask <= TGSI_WRITEMASK_XYZW ); + assert( index >= -32768 && index <= 32767 ); + + dst_register = tgsi_default_dst_register(); + dst_register.File = file; + dst_register.WriteMask = mask; + dst_register.Index = index; + + instruction_grow( instruction, header ); + + return dst_register; +} + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ) +{ + struct tgsi_full_dst_register full_dst_register; + + full_dst_register.DstRegister = tgsi_default_dst_register(); + full_dst_register.DstRegisterExtConcode = + tgsi_default_dst_register_ext_concode(); + full_dst_register.DstRegisterExtModulate = + tgsi_default_dst_register_ext_modulate(); + + return full_dst_register; +} + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; + dst_register_ext_concode.CondMask = TGSI_CC_TR; + dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; + dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; + dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; + dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; + dst_register_ext_concode.CondSrcIndex = 0; + dst_register_ext_concode.Padding = 0; + dst_register_ext_concode.Extended = 0; + + return dst_register_ext_concode; +} + +unsigned +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + assert( cc <= TGSI_CC_FL ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( index >= -32768 && index <= 32767 ); + + dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); + dst_register_ext_concode.CondMask = cc; + dst_register_ext_concode.CondSwizzleX = swizzle_x; + dst_register_ext_concode.CondSwizzleY = swizzle_y; + dst_register_ext_concode.CondSwizzleZ = swizzle_z; + dst_register_ext_concode.CondSwizzleW = swizzle_w; + dst_register_ext_concode.CondSrcIndex = index; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_concode; +} + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; + dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; + dst_register_ext_modulate.Padding = 0; + dst_register_ext_modulate.Extended = 0; + + return dst_register_ext_modulate; +} + +unsigned +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + unsigned modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + assert( modulate <= TGSI_MODULATE_EIGHTH ); + + dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); + dst_register_ext_modulate.Modulate = modulate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_modulate; +} + diff --git a/src/gallium/aux/tgsi/util/tgsi_build.h b/src/gallium/aux/tgsi/util/tgsi_build.h new file mode 100644 index 0000000000..116c78abf3 --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_build.h @@ -0,0 +1,320 @@ +#if !defined TGSI_BUILD_H +#define TGSI_BUILD_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ); + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ); + +struct tgsi_processor +tgsi_default_processor( void ); + +struct tgsi_processor +tgsi_build_processor( + unsigned processor, + struct tgsi_header *header ); + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ); + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned declare, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ); + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ); + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + unsigned mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ); + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + unsigned interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ); + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ); + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ); + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ); + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + float value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ); + +unsigned +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ); + +struct tgsi_instruction +tgsi_build_instruction( + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header ); + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ); + +unsigned +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ); + +unsigned +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ); + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ); + +unsigned +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ); + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + unsigned label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ); + +unsigned +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ); + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + unsigned texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register +tgsi_default_src_register( void ); + +struct tgsi_src_register +tgsi_build_src_register( + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ); + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ); + +unsigned +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ); + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + unsigned ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ); + +unsigned +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ); + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dimension +tgsi_default_dimension( void ); + +struct tgsi_dimension +tgsi_build_dimension( + unsigned indirect, + unsigned index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register +tgsi_default_dst_register( void ); + +struct tgsi_dst_register +tgsi_build_dst_register( + unsigned file, + unsigned mask, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ); + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ); + +unsigned +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ); + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ); + +unsigned +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ); + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + unsigned modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_BUILD_H + diff --git a/src/gallium/aux/tgsi/util/tgsi_dump.c b/src/gallium/aux/tgsi/util/tgsi_dump.c new file mode 100644 index 0000000000..b5c54847e0 --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_dump.c @@ -0,0 +1,1581 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_dump.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" + +struct gen_dump +{ + unsigned tabs; + void (* write)( + struct gen_dump *dump, + const void *data, + unsigned size ); +}; + +struct text_dump +{ + struct gen_dump base; + char *text; + unsigned length; + unsigned capacity; +}; + +static void +_text_dump_write( + struct gen_dump *dump, + const void *data, + unsigned size ) +{ + struct text_dump *td = (struct text_dump *) dump; + unsigned new_length = td->length + size; + + if( new_length >= td->capacity ) { + unsigned new_capacity = td->capacity; + + do { + if( new_capacity == 0 ) { + new_capacity = 256; + } + else { + new_capacity *= 2; + } + } while( new_length >= new_capacity ); + td->text = (char *) REALLOC( + td->text, + td->capacity, + new_capacity ); + td->capacity = new_capacity; + } + memcpy( + &td->text[td->length], + data, + size ); + td->length = new_length; + td->text[td->length] = '\0'; +} + +struct file_dump +{ + struct gen_dump base; + FILE *file; +}; + +static void +_file_dump_write( + struct gen_dump *dump, + const void *data, + unsigned size ) +{ + struct file_dump *fd = (struct file_dump *) dump; + +#if 0 + fwrite( data, 1, size, fd->file ); +#else + { + unsigned i; + + for (i = 0; i < size; i++ ) { + fprintf( fd->file, "%c", ((const char *) data)[i] ); + } + } +#endif +} + +static void +gen_dump_str( + struct gen_dump *dump, + const char *str ) +{ + unsigned i; + size_t len = strlen( str ); + + for (i = 0; i < len; i++) { + dump->write( dump, &str[i], 1 ); + if (str[i] == '\n') { + unsigned i; + + for (i = 0; i < dump->tabs; i++) { + dump->write( dump, " ", 4 ); + } + } + } +} + +static void +gen_dump_chr( + struct gen_dump *dump, + const char chr ) +{ + dump->write( dump, &chr, 1 ); +} + +static void +gen_dump_uix( + struct gen_dump *dump, + const unsigned ui ) +{ + char str[36]; + + sprintf( str, "0x%x", ui ); + gen_dump_str( dump, str ); +} + +static void +gen_dump_uid( + struct gen_dump *dump, + const unsigned ui ) +{ + char str[16]; + + sprintf( str, "%u", ui ); + gen_dump_str( dump, str ); +} + +static void +gen_dump_sid( + struct gen_dump *dump, + const int si ) +{ + char str[16]; + + sprintf( str, "%d", si ); + gen_dump_str( dump, str ); +} + +static void +gen_dump_flt( + struct gen_dump *dump, + const float flt ) +{ + char str[48]; + + sprintf( str, "%10.4f", flt ); + gen_dump_str( dump, str ); +} + +static void +gen_dump_enum( + struct gen_dump *dump, + const unsigned e, + const char **enums, + const unsigned enums_count ) +{ + if (e >= enums_count) { + gen_dump_uid( dump, e ); + } + else { + gen_dump_str( dump, enums[e] ); + } +} + +static void +gen_dump_tab( + struct gen_dump *dump ) +{ + ++dump->tabs; +} + +static void +gen_dump_untab( + struct gen_dump *dump ) +{ + assert( dump->tabs > 0 ); + + --dump->tabs; +} + +#define TXT(S) gen_dump_str( dump, S ) +#define CHR(C) gen_dump_chr( dump, C ) +#define UIX(I) gen_dump_uix( dump, I ) +#define UID(I) gen_dump_uid( dump, I ) +#define SID(I) gen_dump_sid( dump, I ) +#define FLT(F) gen_dump_flt( dump, F ) +#define TAB() gen_dump_tab( dump ) +#define UNT() gen_dump_untab( dump ) +#define ENM(E,ENUMS) gen_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) + +static const char *TGSI_PROCESSOR_TYPES[] = +{ + "PROCESSOR_FRAGMENT", + "PROCESSOR_VERTEX", + "PROCESSOR_GEOMETRY" +}; + +static const char *TGSI_PROCESSOR_TYPES_SHORT[] = +{ + "FRAG", + "VERT", + "GEOM" +}; + +static const char *TGSI_TOKEN_TYPES[] = +{ + "TOKEN_TYPE_DECLARATION", + "TOKEN_TYPE_IMMEDIATE", + "TOKEN_TYPE_INSTRUCTION" +}; + +static const char *TGSI_FILES[] = +{ + "FILE_NULL", + "FILE_CONSTANT", + "FILE_INPUT", + "FILE_OUTPUT", + "FILE_TEMPORARY", + "FILE_SAMPLER", + "FILE_ADDRESS", + "FILE_IMMEDIATE" +}; + +static const char *TGSI_FILES_SHORT[] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + +static const char *TGSI_DECLARES[] = +{ + "DECLARE_RANGE", + "DECLARE_MASK" +}; + +static const char *TGSI_INTERPOLATES[] = +{ + "INTERPOLATE_CONSTANT", + "INTERPOLATE_LINEAR", + "INTERPOLATE_PERSPECTIVE", + "INTERPOLATE_ATTRIB" +}; + +static const char *TGSI_INTERPOLATES_SHORT[] = +{ + "CONSTANT", + "LINEAR", + "PERSPECTIVE", + "ATTRIB" +}; + +static const char *TGSI_SEMANTICS[] = +{ + "SEMANTIC_POSITION", + "SEMANTIC_COLOR", + "SEMANTIC_BCOLOR", + "SEMANTIC_FOG", + "SEMANTIC_PSIZE", + "SEMANTIC_GENERIC," +}; + +static const char *TGSI_SEMANTICS_SHORT[] = +{ + "POSITION", + "COLOR", + "BCOLOR", + "FOG", + "PSIZE", + "GENERIC", +}; + +static const char *TGSI_IMMS[] = +{ + "IMM_FLOAT32" +}; + +static const char *TGSI_IMMS_SHORT[] = +{ + "FLT32" +}; + +static const char *TGSI_OPCODES[] = +{ + "OPCODE_ARL", + "OPCODE_MOV", + "OPCODE_LIT", + "OPCODE_RCP", + "OPCODE_RSQ", + "OPCODE_EXP", + "OPCODE_LOG", + "OPCODE_MUL", + "OPCODE_ADD", + "OPCODE_DP3", + "OPCODE_DP4", + "OPCODE_DST", + "OPCODE_MIN", + "OPCODE_MAX", + "OPCODE_SLT", + "OPCODE_SGE", + "OPCODE_MAD", + "OPCODE_SUB", + "OPCODE_LERP", + "OPCODE_CND", + "OPCODE_CND0", + "OPCODE_DOT2ADD", + "OPCODE_INDEX", + "OPCODE_NEGATE", + "OPCODE_FRAC", + "OPCODE_CLAMP", + "OPCODE_FLOOR", + "OPCODE_ROUND", + "OPCODE_EXPBASE2", + "OPCODE_LOGBASE2", + "OPCODE_POWER", + "OPCODE_CROSSPRODUCT", + "OPCODE_MULTIPLYMATRIX", + "OPCODE_ABS", + "OPCODE_RCC", + "OPCODE_DPH", + "OPCODE_COS", + "OPCODE_DDX", + "OPCODE_DDY", + "OPCODE_KILP", + "OPCODE_PK2H", + "OPCODE_PK2US", + "OPCODE_PK4B", + "OPCODE_PK4UB", + "OPCODE_RFL", + "OPCODE_SEQ", + "OPCODE_SFL", + "OPCODE_SGT", + "OPCODE_SIN", + "OPCODE_SLE", + "OPCODE_SNE", + "OPCODE_STR", + "OPCODE_TEX", + "OPCODE_TXD", + "OPCODE_UP2H", + "OPCODE_UP2US", + "OPCODE_UP4B", + "OPCODE_UP4UB", + "OPCODE_X2D", + "OPCODE_ARA", + "OPCODE_ARR", + "OPCODE_BRA", + "OPCODE_CAL", + "OPCODE_RET", + "OPCODE_SSG", + "OPCODE_CMP", + "OPCODE_SCS", + "OPCODE_TXB", + "OPCODE_NRM", + "OPCODE_DIV", + "OPCODE_DP2", + "OPCODE_TXL", + "OPCODE_BRK", + "OPCODE_IF", + "OPCODE_LOOP", + "OPCODE_REP", + "OPCODE_ELSE", + "OPCODE_ENDIF", + "OPCODE_ENDLOOP", + "OPCODE_ENDREP", + "OPCODE_PUSHA", + "OPCODE_POPA", + "OPCODE_CEIL", + "OPCODE_I2F", + "OPCODE_NOT", + "OPCODE_TRUNC", + "OPCODE_SHL", + "OPCODE_SHR", + "OPCODE_AND", + "OPCODE_OR", + "OPCODE_MOD", + "OPCODE_XOR", + "OPCODE_SAD", + "OPCODE_TXF", + "OPCODE_TXQ", + "OPCODE_CONT", + "OPCODE_EMIT", + "OPCODE_ENDPRIM", + "OPCODE_BGNLOOP2", + "OPCODE_BGNSUB", + "OPCODE_ENDLOOP2", + "OPCODE_ENDSUB", + "OPCODE_NOISE1", + "OPCODE_NOISE2", + "OPCODE_NOISE3", + "OPCODE_NOISE4", + "OPCODE_NOP", + "OPCODE_TEXBEM", + "OPCODE_TEXBEML", + "OPCODE_TEXREG2AR", + "OPCODE_TEXM3X2PAD", + "OPCODE_TEXM3X2TEX", + "OPCODE_TEXM3X3PAD", + "OPCODE_TEXM3X3TEX", + "OPCODE_TEXM3X3SPEC", + "OPCODE_TEXM3X3VSPEC", + "OPCODE_TEXREG2GB", + "OPCODE_TEXREG2RGB", + "OPCODE_TEXDP3TEX", + "OPCODE_TEXDP3", + "OPCODE_TEXM3X3", + "OPCODE_TEXM3X2DEPTH", + "OPCODE_TEXDEPTH", + "OPCODE_BEM", + "OPCODE_M4X3", + "OPCODE_M3X4", + "OPCODE_M3X3", + "OPCODE_M3X2", + "OPCODE_NRM4", + "OPCODE_CALLNZ", + "OPCODE_IFC", + "OPCODE_BREAKC", + "OPCODE_KIL", + "OPCODE_END" +}; + +static const char *TGSI_OPCODES_SHORT[] = +{ + "ARL", + "MOV", + "LIT", + "RCP", + "RSQ", + "EXP", + "LOG", + "MUL", + "ADD", + "DP3", + "DP4", + "DST", + "MIN", + "MAX", + "SLT", + "SGE", + "MAD", + "SUB", + "LERP", + "CND", + "CND0", + "DOT2ADD", + "INDEX", + "NEGATE", + "FRAC", + "CLAMP", + "FLOOR", + "ROUND", + "EXPBASE2", + "LOGBASE2", + "POWER", + "CROSSPRODUCT", + "MULTIPLYMATRIX", + "ABS", + "RCC", + "DPH", + "COS", + "DDX", + "DDY", + "KILP", + "PK2H", + "PK2US", + "PK4B", + "PK4UB", + "RFL", + "SEQ", + "SFL", + "SGT", + "SIN", + "SLE", + "SNE", + "STR", + "TEX", + "TXD", + "UP2H", + "UP2US", + "UP4B", + "UP4UB", + "X2D", + "ARA", + "ARR", + "BRA", + "CAL", + "RET", + "SSG", + "CMP", + "SCS", + "TXB", + "NRM", + "DIV", + "DP2", + "TXL", + "BRK", + "IF", + "LOOP", + "REP", + "ELSE", + "ENDIF", + "ENDLOOP", + "ENDREP", + "PUSHA", + "POPA", + "CEIL", + "I2F", + "NOT", + "TRUNC", + "SHL", + "SHR", + "AND", + "OR", + "MOD", + "XOR", + "SAD", + "TXF", + "TXQ", + "CONT", + "EMIT", + "ENDPRIM", + "BGNLOOP2", + "BGNSUB", + "ENDLOOP2", + "ENDSUB", + "NOISE1", + "NOISE2", + "NOISE3", + "NOISE4", + "NOP", + "TEXBEM", + "TEXBEML", + "TEXREG2AR", + "TEXM3X2PAD", + "TEXM3X2TEX", + "TEXM3X3PAD", + "TEXM3X3TEX", + "TEXM3X3SPEC", + "TEXM3X3VSPEC", + "TEXREG2GB", + "TEXREG2RGB", + "TEXDP3TEX", + "TEXDP3", + "TEXM3X3", + "TEXM3X2DEPTH", + "TEXDEPTH", + "BEM", + "M4X3", + "M3X4", + "M3X3", + "M3X2", + "NRM4", + "CALLNZ", + "IFC", + "BREAKC", + "KIL", + "END" +}; + +static const char *TGSI_SATS[] = +{ + "SAT_NONE", + "SAT_ZERO_ONE", + "SAT_MINUS_PLUS_ONE" +}; + +static const char *TGSI_INSTRUCTION_EXTS[] = +{ + "INSTRUCTION_EXT_TYPE_NV", + "INSTRUCTION_EXT_TYPE_LABEL", + "INSTRUCTION_EXT_TYPE_TEXTURE" +}; + +static const char *TGSI_PRECISIONS[] = +{ + "PRECISION_DEFAULT", + "TGSI_PRECISION_FLOAT32", + "TGSI_PRECISION_FLOAT16", + "TGSI_PRECISION_FIXED12" +}; + +static const char *TGSI_CCS[] = +{ + "CC_GT", + "CC_EQ", + "CC_LT", + "CC_UN", + "CC_GE", + "CC_LE", + "CC_NE", + "CC_TR", + "CC_FL" +}; + +static const char *TGSI_SWIZZLES[] = +{ + "SWIZZLE_X", + "SWIZZLE_Y", + "SWIZZLE_Z", + "SWIZZLE_W" +}; + +static const char *TGSI_SWIZZLES_SHORT[] = +{ + "x", + "y", + "z", + "w" +}; + +static const char *TGSI_TEXTURES[] = +{ + "TEXTURE_UNKNOWN", + "TEXTURE_1D", + "TEXTURE_2D", + "TEXTURE_3D", + "TEXTURE_CUBE", + "TEXTURE_RECT", + "TEXTURE_SHADOW1D", + "TEXTURE_SHADOW2D", + "TEXTURE_SHADOWRECT" +}; + +static const char *TGSI_SRC_REGISTER_EXTS[] = +{ + "SRC_REGISTER_EXT_TYPE_SWZ", + "SRC_REGISTER_EXT_TYPE_MOD" +}; + +static const char *TGSI_EXTSWIZZLES[] = +{ + "EXTSWIZZLE_X", + "EXTSWIZZLE_Y", + "EXTSWIZZLE_Z", + "EXTSWIZZLE_W", + "EXTSWIZZLE_ZERO", + "EXTSWIZZLE_ONE" +}; + +static const char *TGSI_EXTSWIZZLES_SHORT[] = +{ + "x", + "y", + "z", + "w", + "0", + "1" +}; + +static const char *TGSI_WRITEMASKS[] = +{ + "0", + "WRITEMASK_X", + "WRITEMASK_Y", + "WRITEMASK_XY", + "WRITEMASK_Z", + "WRITEMASK_XZ", + "WRITEMASK_YZ", + "WRITEMASK_XYZ", + "WRITEMASK_W", + "WRITEMASK_XW", + "WRITEMASK_YW", + "WRITEMASK_XYW", + "WRITEMASK_ZW", + "WRITEMASK_XZW", + "WRITEMASK_YZW", + "WRITEMASK_XYZW" +}; + +static const char *TGSI_DST_REGISTER_EXTS[] = +{ + "DST_REGISTER_EXT_TYPE_CONDCODE", + "DST_REGISTER_EXT_TYPE_MODULATE" +}; + +static const char *TGSI_MODULATES[] = +{ + "MODULATE_1X", + "MODULATE_2X", + "MODULATE_4X", + "MODULATE_8X", + "MODULATE_HALF", + "MODULATE_QUARTER", + "MODULATE_EIGHTH" +}; + +static void +dump_declaration_short( + struct gen_dump *dump, + struct tgsi_full_declaration *decl ) +{ + TXT( "\nDCL " ); + ENM( decl->Declaration.File, TGSI_FILES_SHORT ); + + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + CHR( '[' ); + UID( decl->u.DeclarationRange.First ); + if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { + TXT( ".." ); + UID( decl->u.DeclarationRange.Last ); + } + CHR( ']' ); + break; + default: + assert( 0 ); + } + + if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + + if( decl->Declaration.Interpolate ) { + TXT( ", " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); + } + + if( decl->Declaration.Semantic ) { + TXT( ", " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); + CHR( '[' ); + UID( decl->Semantic.SemanticIndex ); + CHR( ']' ); + } +} + +static void +dump_declaration_verbose( + struct gen_dump *dump, + struct tgsi_full_declaration *decl, + unsigned ignored, + unsigned deflt, + struct tgsi_full_declaration *fd ) +{ + TXT( "\nFile : " ); + ENM( decl->Declaration.File, TGSI_FILES ); + TXT( "\nDeclare : " ); + ENM( decl->Declaration.Declare, TGSI_DECLARES ); + if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { + TXT( "\nUsageMask : " ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'X' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'Y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'Z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'W' ); + } + } + if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { + TXT( "\nInterpolate: " ); + UID( decl->Declaration.Interpolate ); + } + if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { + TXT( "\nSemantic : " ); + UID( decl->Declaration.Semantic ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Declaration.Padding ); + } + + CHR( '\n' ); + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + TXT( "\nFirst: " ); + UID( decl->u.DeclarationRange.First ); + TXT( "\nLast : " ); + UID( decl->u.DeclarationRange.Last ); + break; + + case TGSI_DECLARE_MASK: + TXT( "\nMask: " ); + UIX( decl->u.DeclarationMask.Mask ); + break; + + default: + assert( 0 ); + } + + if( decl->Declaration.Interpolate ) { + CHR( '\n' ); + TXT( "\nInterpolate: " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Interpolation.Padding ); + } + } + + if( decl->Declaration.Semantic ) { + CHR( '\n' ); + TXT( "\nSemanticName : " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); + TXT( "\nSemanticIndex: " ); + UID( decl->Semantic.SemanticIndex ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Semantic.Padding ); + } + } +} + +static void +dump_immediate_short( + struct gen_dump *dump, + struct tgsi_full_immediate *imm ) +{ + unsigned i; + + TXT( "\nIMM " ); + ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); + + TXT( " { " ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + + if( i < imm->Immediate.Size - 2 ) { + TXT( ", " ); + } + } + TXT( " }" ); +} + +static void +dump_immediate_verbose( + struct gen_dump *dump, + struct tgsi_full_immediate *imm, + unsigned ignored ) +{ + unsigned i; + + TXT( "\nDataType : " ); + ENM( imm->Immediate.DataType, TGSI_IMMS ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( imm->Immediate.Padding ); + } + + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + CHR( '\n' ); + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + TXT( "\nFloat: " ); + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + } +} + +static void +dump_instruction_short( + struct gen_dump *dump, + struct tgsi_full_instruction *inst, + unsigned instno ) +{ + unsigned i; + boolean first_reg = TRUE; + + CHR( '\n' ); + UID( instno ); + CHR( ':' ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); + + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + TXT( "_SAT" ); + break; + case TGSI_SAT_MINUS_PLUS_ONE: + TXT( "_SAT[-1,1]" ); + break; + default: + assert( 0 ); + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( dst->DstRegister.Index ); + CHR( ']' ); + + if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + + first_reg = FALSE; + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + if( src->SrcRegisterExtMod.Complement ) { + TXT( "(1 - " ); + } + if( src->SrcRegisterExtMod.Negate ) { + CHR( '-' ); + } + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + if( src->SrcRegister.Negate ) { + CHR( '-' ); + } + + ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( src->SrcRegister.Index ); + CHR( ']' ); + + if (src->SrcRegister.Extended) { + if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { + CHR( '.' ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); + } + } + else if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { + CHR( '.' ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); + } + + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + if( src->SrcRegisterExtMod.Complement ) { + CHR( ')' ); + } + + first_reg = FALSE; + } + + switch( inst->Instruction.Opcode ) { + case TGSI_OPCODE_IF: + case TGSI_OPCODE_ELSE: + case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_CAL: + TXT( " :" ); + UID( inst->InstructionExtLabel.Label ); + break; + } +} + +static void +dump_instruction_verbose( + struct gen_dump *dump, + struct tgsi_full_instruction *inst, + unsigned ignored, + unsigned deflt, + struct tgsi_full_instruction *fi ) +{ + unsigned i; + + TXT( "\nOpcode : " ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES ); + if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { + TXT( "\nSaturate : " ); + ENM( inst->Instruction.Saturate, TGSI_SATS ); + } + if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { + TXT( "\nNumDstRegs : " ); + UID( inst->Instruction.NumDstRegs ); + } + if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { + TXT( "\nNumSrcRegs : " ); + UID( inst->Instruction.NumSrcRegs ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->Instruction.Padding ); + } + + if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { + TXT( "\nPrecision : " ); + ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); + } + if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { + TXT( "\nCondDstIndex : " ); + UID( inst->InstructionExtNv.CondDstIndex ); + } + if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { + TXT( "\nCondFlowIndex : " ); + UID( inst->InstructionExtNv.CondFlowIndex ); + } + if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { + TXT( "\nCondMask : " ); + ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { + TXT( "\nCondSwizzleX : " ); + ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { + TXT( "\nCondSwizzleY : " ); + ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ : " ); + ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { + TXT( "\nCondSwizzleW : " ); + ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { + TXT( "\nCondDstUpdate : " ); + UID( inst->InstructionExtNv.CondDstUpdate ); + } + if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { + TXT( "\nCondFlowEnable: " ); + UID( inst->InstructionExtNv.CondFlowEnable ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtNv.Padding ); + if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { + TXT( "\nExtended : " ); + UID( inst->InstructionExtNv.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { + TXT( "\nLabel : " ); + UID( inst->InstructionExtLabel.Label ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtLabel.Padding ); + if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtLabel.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { + TXT( "\nTexture : " ); + ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtTexture.Padding ); + if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtTexture.Extended ); + } + } + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : " ); + ENM( dst->DstRegister.File, TGSI_FILES ); + if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { + TXT( "\nWriteMask: " ); + ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); + } + if( ignored ) { + if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( dst->DstRegister.Indirect ); + } + if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( dst->DstRegister.Dimension ); + } + } + if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { + TXT( "\nIndex : " ); + SID( dst->DstRegister.Index ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegister.Padding ); + if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegister.Extended ); + } + } + + if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { + TXT( "\nCondMask : " ); + ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { + TXT( "\nCondSwizzleX: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { + TXT( "\nCondSwizzleY: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { + TXT( "\nCondSwizzleW: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { + TXT( "\nCondSrcIndex: " ); + UID( dst->DstRegisterExtConcode.CondSrcIndex ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtConcode.Padding ); + if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegisterExtConcode.Extended ); + } + } + } + + if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { + TXT( "\nModulate: " ); + ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtModulate.Padding ); + if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { + TXT( "\nExtended: " ); + UID( dst->DstRegisterExtModulate.Extended ); + } + } + } + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : "); + ENM( src->SrcRegister.File, TGSI_FILES ); + if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { + TXT( "\nSwizzleX : " ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { + TXT( "\nSwizzleY : " ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { + TXT( "\nSwizzleZ : " ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { + TXT( "\nSwizzleW : " ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegister.Negate ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( src->SrcRegister.Indirect ); + } + if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( src->SrcRegister.Dimension ); + } + } + if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { + TXT( "\nIndex : " ); + SID( src->SrcRegister.Index ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegister.Extended ); + } + } + + if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { + TXT( "\nExtSwizzleX: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { + TXT( "\nExtSwizzleY: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { + TXT( "\nExtSwizzleZ: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { + TXT( "\nExtSwizzleW: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { + TXT( "\nNegateX : " ); + UID( src->SrcRegisterExtSwz.NegateX ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { + TXT( "\nNegateY : " ); + UID( src->SrcRegisterExtSwz.NegateY ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { + TXT( "\nNegateZ : " ); + UID( src->SrcRegisterExtSwz.NegateZ ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { + TXT( "\nNegateW : " ); + UID( src->SrcRegisterExtSwz.NegateW ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { + TXT( "\nExtDivide : " ); + ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtSwz.Padding ); + if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtSwz.Extended ); + } + } + } + + if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { + TXT( "\nComplement: " ); + UID( src->SrcRegisterExtMod.Complement ); + } + if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { + TXT( "\nBias : " ); + UID( src->SrcRegisterExtMod.Bias ); + } + if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { + TXT( "\nScale2X : " ); + UID( src->SrcRegisterExtMod.Scale2X ); + } + if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { + TXT( "\nAbsolute : " ); + UID( src->SrcRegisterExtMod.Absolute ); + } + if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegisterExtMod.Negate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtMod.Padding ); + if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtMod.Extended ); + } + } + } + } +} + +static void +dump_gen( + struct gen_dump *dump, + const struct tgsi_token *tokens, + unsigned flags ) +{ + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned verbose = flags & TGSI_DUMP_VERBOSE; + unsigned ignored = !(flags & TGSI_DUMP_NO_IGNORED); + unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); + unsigned instno = 0; + + dump->tabs = 0; + + /* sanity check */ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); + + tgsi_parse_init( &parse, tokens ); + + TXT( "tgsi-dump begin -----------------" ); + + CHR( '\n' ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); + CHR( ' ' ); + UID( parse.FullVersion.Version.MajorVersion ); + CHR( '.' ); + UID( parse.FullVersion.Version.MinorVersion ); + + if( verbose ) { + TXT( "\nMajorVersion: " ); + UID( parse.FullVersion.Version.MajorVersion ); + TXT( "\nMinorVersion: " ); + UID( parse.FullVersion.Version.MinorVersion ); + CHR( '\n' ); + + TXT( "\nHeaderSize: " ); + UID( parse.FullHeader.Header.HeaderSize ); + TXT( "\nBodySize : " ); + UID( parse.FullHeader.Header.BodySize ); + TXT( "\nProcessor : " ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + CHR( '\n' ); + } + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_short( + dump, + &parse.FullToken.FullDeclaration ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_short( + dump, + &parse.FullToken.FullImmediate ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_short( + dump, + &parse.FullToken.FullInstruction, + instno ); + instno++; + break; + + default: + assert( 0 ); + } + + if( verbose ) { + TXT( "\nType : " ); + ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); + if( ignored ) { + TXT( "\nSize : " ); + UID( parse.FullToken.Token.Size ); + if( deflt || parse.FullToken.Token.Extended ) { + TXT( "\nExtended : " ); + UID( parse.FullToken.Token.Extended ); + } + } + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_verbose( + dump, + &parse.FullToken.FullDeclaration, + ignored, + deflt, + &fd ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_verbose( + dump, + &parse.FullToken.FullImmediate, + ignored ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_verbose( + dump, + &parse.FullToken.FullInstruction, + ignored, + deflt, + &fi ); + break; + + default: + assert( 0 ); + } + + CHR( '\n' ); + } + } + + TXT( "\ntgsi-dump end -------------------\n" ); + + tgsi_parse_free( &parse ); +} + + +static void +sanity_checks(void) +{ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); + assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "END") == 0); +} + + +void +tgsi_dump( + const struct tgsi_token *tokens, + unsigned flags ) +{ + struct file_dump dump; + + sanity_checks(); + + dump.base.write = _file_dump_write; +#if 0 + { + static unsigned counter = 0; + char buffer[64]; + sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); + dump.file = fopen( buffer, "wt" ); + } +#else + dump.file = stderr; +#endif + + dump_gen( + &dump.base, + tokens, + flags ); + +#if 0 + fclose( dump.file ); +#endif +} + +void +tgsi_dump_str( + char **str, + const struct tgsi_token *tokens, + unsigned flags ) +{ + struct text_dump dump; + + dump.base.write = _text_dump_write; + dump.text = NULL; + dump.length = 0; + dump.capacity = 0; + + dump_gen( + &dump.base, + tokens, + flags ); + + *str = dump.text; +} diff --git a/src/gallium/aux/tgsi/util/tgsi_dump.h b/src/gallium/aux/tgsi/util/tgsi_dump.h new file mode 100644 index 0000000000..1adc9db251 --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_dump.h @@ -0,0 +1,28 @@ +#if !defined TGSI_DUMP_H +#define TGSI_DUMP_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +#define TGSI_DUMP_VERBOSE 1 +#define TGSI_DUMP_NO_IGNORED 2 +#define TGSI_DUMP_NO_DEFAULT 4 + +void +tgsi_dump( + const struct tgsi_token *tokens, + unsigned flags ); + +void +tgsi_dump_str( + char **str, + const struct tgsi_token *tokens, + unsigned flags ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_DUMP_H + diff --git a/src/gallium/aux/tgsi/util/tgsi_parse.c b/src/gallium/aux/tgsi/util/tgsi_parse.c new file mode 100644 index 0000000000..bf6b89ce56 --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_parse.c @@ -0,0 +1,319 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ) +{ + full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; +} + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ) +{ + if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { + FREE( full_token->FullImmediate.u.Pointer ); + } +} + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ) +{ + ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; + if( ctx->FullVersion.Version.MajorVersion > 1 ) { + return TGSI_PARSE_ERROR; + } + + ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; + if( ctx->FullHeader.Header.HeaderSize >= 2 ) { + ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; + } + else { + ctx->FullHeader.Processor = tgsi_default_processor(); + } + + ctx->Tokens = tokens; + ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; + + tgsi_full_token_init( &ctx->FullToken ); + + return TGSI_PARSE_OK; +} + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ) +{ + tgsi_full_token_free( &ctx->FullToken ); +} + +boolean +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ) +{ + return ctx->Position >= + 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; +} + +static void +next_token( + struct tgsi_parse_context *ctx, + void *token ) +{ + assert( !tgsi_parse_end_of_tokens( ctx ) ); + + *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; +} + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ) +{ + struct tgsi_token token; + unsigned i; + + tgsi_full_token_free( &ctx->FullToken ); + tgsi_full_token_init( &ctx->FullToken ); + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; + + *decl = tgsi_default_full_declaration(); + decl->Declaration = *(struct tgsi_declaration *) &token; + + switch( decl->Declaration.Type ) { + case TGSI_DECLARE_RANGE: + next_token( ctx, &decl->u.DeclarationRange ); + break; + + case TGSI_DECLARE_MASK: + next_token( ctx, &decl->u.DeclarationMask ); + break; + + default: + assert (0); + } + + if( decl->Declaration.Interpolate ) { + next_token( ctx, &decl->Interpolation ); + } + + if( decl->Declaration.Semantic ) { + next_token( ctx, &decl->Semantic ); + } + + break; + } + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; + + *imm = tgsi_default_full_immediate(); + imm->Immediate = *(struct tgsi_immediate *) &token; + + assert( !imm->Immediate.Extended ); + + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + imm->u.Pointer = MALLOC( + sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + next_token( ctx, &imm->u.ImmediateFloat32[i] ); + } + break; + + default: + assert( 0 ); + } + + break; + } + + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; + unsigned extended; + + *inst = tgsi_default_full_instruction(); + inst->Instruction = *(struct tgsi_instruction *) &token; + + extended = inst->Instruction.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_INSTRUCTION_EXT_TYPE_NV: + inst->InstructionExtNv = + *(struct tgsi_instruction_ext_nv *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_LABEL: + inst->InstructionExtLabel = + *(struct tgsi_instruction_ext_label *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: + inst->InstructionExtTexture = + *(struct tgsi_instruction_ext_texture *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); + assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); + + extended = inst->FullDstRegisters[i].DstRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: + inst->FullDstRegisters[i].DstRegisterExtConcode = + *(struct tgsi_dst_register_ext_concode *) &token; + break; + + case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: + inst->FullDstRegisters[i].DstRegisterExtModulate = + *(struct tgsi_dst_register_ext_modulate *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + } + + assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); + + extended = inst->FullSrcRegisters[i].SrcRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: + inst->FullSrcRegisters[i].SrcRegisterExtSwz = + *(struct tgsi_src_register_ext_swz *) &token; + break; + + case TGSI_SRC_REGISTER_EXT_TYPE_MOD: + inst->FullSrcRegisters[i].SrcRegisterExtMod = + *(struct tgsi_src_register_ext_mod *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + + if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); + + /* + * No support for multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); + + if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + } + } + + break; + } + + default: + assert( 0 ); + } +} + diff --git a/src/gallium/aux/tgsi/util/tgsi_parse.h b/src/gallium/aux/tgsi/util/tgsi_parse.h new file mode 100644 index 0000000000..9372da8d5d --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_parse.h @@ -0,0 +1,121 @@ +#if !defined TGSI_PARSE_H +#define TGSI_PARSE_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_full_version +{ + struct tgsi_version Version; +}; + +struct tgsi_full_header +{ + struct tgsi_header Header; + struct tgsi_processor Processor; +}; + +struct tgsi_full_dst_register +{ + struct tgsi_dst_register DstRegister; + struct tgsi_dst_register_ext_concode DstRegisterExtConcode; + struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; +}; + +struct tgsi_full_src_register +{ + struct tgsi_src_register SrcRegister; + struct tgsi_src_register_ext_swz SrcRegisterExtSwz; + struct tgsi_src_register_ext_mod SrcRegisterExtMod; + struct tgsi_src_register SrcRegisterInd; + struct tgsi_dimension SrcRegisterDim; + struct tgsi_src_register SrcRegisterDimInd; +}; + +struct tgsi_full_declaration +{ + struct tgsi_declaration Declaration; + union + { + struct tgsi_declaration_range DeclarationRange; + struct tgsi_declaration_mask DeclarationMask; + } u; + struct tgsi_declaration_interpolation Interpolation; + struct tgsi_declaration_semantic Semantic; +}; + +struct tgsi_full_immediate +{ + struct tgsi_immediate Immediate; + union + { + void *Pointer; + struct tgsi_immediate_float32 *ImmediateFloat32; + } u; +}; + +#define TGSI_FULL_MAX_DST_REGISTERS 2 +#define TGSI_FULL_MAX_SRC_REGISTERS 3 + +struct tgsi_full_instruction +{ + struct tgsi_instruction Instruction; + struct tgsi_instruction_ext_nv InstructionExtNv; + struct tgsi_instruction_ext_label InstructionExtLabel; + struct tgsi_instruction_ext_texture InstructionExtTexture; + struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; + struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; +}; + +union tgsi_full_token +{ + struct tgsi_token Token; + struct tgsi_full_declaration FullDeclaration; + struct tgsi_full_immediate FullImmediate; + struct tgsi_full_instruction FullInstruction; +}; + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ); + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ); + +struct tgsi_parse_context +{ + const struct tgsi_token *Tokens; + unsigned Position; + struct tgsi_full_version FullVersion; + struct tgsi_full_header FullHeader; + union tgsi_full_token FullToken; +}; + +#define TGSI_PARSE_OK 0 +#define TGSI_PARSE_ERROR 1 + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ); + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ); + +boolean +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ); + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_PARSE_H + diff --git a/src/gallium/aux/tgsi/util/tgsi_transform.c b/src/gallium/aux/tgsi/util/tgsi_transform.c new file mode 100644 index 0000000000..357f77b05a --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_transform.c @@ -0,0 +1,199 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI program transformation utility. + * + * Authors: Brian Paul + */ + + +#include "tgsi_transform.h" + + + +static void +emit_instruction(struct tgsi_transform_context *ctx, + const struct tgsi_full_instruction *inst) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_instruction(inst, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + +static void +emit_declaration(struct tgsi_transform_context *ctx, + const struct tgsi_full_declaration *decl) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_declaration(decl, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + +static void +emit_immediate(struct tgsi_transform_context *ctx, + const struct tgsi_full_immediate *imm) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_immediate(imm, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + + +/** + * Apply user-defined transformations to the input shader to produce + * the output shader. + * For example, a register search-and-replace operation could be applied + * by defining a transform_instruction() callback that examined and changed + * the instruction src/dest regs. + * + * \return number of tokens emitted + */ +int +tgsi_transform_shader(const struct tgsi_token *tokens_in, + struct tgsi_token *tokens_out, + uint max_tokens_out, + struct tgsi_transform_context *ctx) +{ + uint procType; + + /* input shader */ + struct tgsi_parse_context parse; + + /* output shader */ + struct tgsi_processor *processor; + + + /** + ** callback context init + **/ + ctx->emit_instruction = emit_instruction; + ctx->emit_declaration = emit_declaration; + ctx->emit_immediate = emit_immediate; + ctx->tokens_out = tokens_out; + ctx->max_tokens_out = max_tokens_out; + + + /** + ** Setup to begin parsing input shader + **/ + if (tgsi_parse_init( &parse, tokens_in ) != TGSI_PARSE_OK) { + debug_printf("tgsi_parse_init() failed in tgsi_transform_shader()!\n"); + return -1; + } + procType = parse.FullHeader.Processor.Processor; + assert(procType == TGSI_PROCESSOR_FRAGMENT || + procType == TGSI_PROCESSOR_VERTEX || + procType == TGSI_PROCESSOR_GEOMETRY); + + + /** + ** Setup output shader + **/ + *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version(); + + ctx->header = (struct tgsi_header *) (tokens_out + 1); + *ctx->header = tgsi_build_header(); + + processor = (struct tgsi_processor *) (tokens_out + 2); + *processor = tgsi_build_processor( procType, ctx->header ); + + ctx->ti = 3; + + + /** + ** Loop over incoming program tokens/instructions + */ + while( !tgsi_parse_end_of_tokens( &parse ) ) { + + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *fullinst + = &parse.FullToken.FullInstruction; + + if (ctx->transform_instruction) + ctx->transform_instruction(ctx, fullinst); + else + ctx->emit_instruction(ctx, fullinst); + } + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *fulldecl + = &parse.FullToken.FullDeclaration; + + if (ctx->transform_declaration) + ctx->transform_declaration(ctx, fulldecl); + else + ctx->emit_declaration(ctx, fulldecl); + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *fullimm + = &parse.FullToken.FullImmediate; + + if (ctx->transform_immediate) + ctx->transform_immediate(ctx, fullimm); + else + ctx->emit_immediate(ctx, fullimm); + } + break; + + default: + assert( 0 ); + } + } + + if (ctx->epilog) { + ctx->epilog(ctx); + } + + tgsi_parse_free (&parse); + + return ctx->ti; +} diff --git a/src/gallium/aux/tgsi/util/tgsi_transform.h b/src/gallium/aux/tgsi/util/tgsi_transform.h new file mode 100644 index 0000000000..365d8c298c --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_transform.h @@ -0,0 +1,93 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_TRANSFORM_H +#define TGSI_TRANSFORM_H + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/util/tgsi_build.h" + + + +/** + * Subclass this to add caller-specific data + */ +struct tgsi_transform_context +{ +/**** PUBLIC ***/ + + /** + * User-defined callbacks invoked per instruction. + */ + void (*transform_instruction)(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst); + + void (*transform_declaration)(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl); + + void (*transform_immediate)(struct tgsi_transform_context *ctx, + struct tgsi_full_immediate *imm); + + /** + * Called at end of input program to allow caller to append extra + * instructions. Return number of tokens emitted. + */ + void (*epilog)(struct tgsi_transform_context *ctx); + + +/*** PRIVATE ***/ + + /** + * These are setup by tgsi_transform_shader() and cannot be overridden. + * Meant to be called from in the above user callback functions. + */ + void (*emit_instruction)(struct tgsi_transform_context *ctx, + const struct tgsi_full_instruction *inst); + void (*emit_declaration)(struct tgsi_transform_context *ctx, + const struct tgsi_full_declaration *decl); + void (*emit_immediate)(struct tgsi_transform_context *ctx, + const struct tgsi_full_immediate *imm); + + struct tgsi_header *header; + uint max_tokens_out; + struct tgsi_token *tokens_out; + uint ti; +}; + + + +extern int +tgsi_transform_shader(const struct tgsi_token *tokens_in, + struct tgsi_token *tokens_out, + uint max_tokens_out, + struct tgsi_transform_context *ctx); + + +#endif /* TGSI_TRANSFORM_H */ diff --git a/src/gallium/aux/tgsi/util/tgsi_util.c b/src/gallium/aux/tgsi/util/tgsi_util.c new file mode 100644 index 0000000000..4cdd89182a --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_util.c @@ -0,0 +1,274 @@ +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" +#include "tgsi_util.h" + +union pointer_hack +{ + void *pointer; + unsigned long long uint64; +}; + +void * +tgsi_align_128bit( + void *unaligned ) +{ + union pointer_hack ph; + + ph.uint64 = 0; + ph.pointer = unaligned; + ph.uint64 = (ph.uint64 + 15) & ~15; + return ph.pointer; +} + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->SwizzleX; + case 1: + return reg->SwizzleY; + case 2: + return reg->SwizzleZ; + case 3: + return reg->SwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->ExtSwizzleX; + case 1: + return reg->ExtSwizzleY; + case 2: + return reg->ExtSwizzleZ; + case 3: + return reg->ExtSwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned swizzle; + + /* + * First, calculate the extended swizzle for a given channel. This will give + * us either a channel index into the simple swizzle or a constant 1 or 0. + */ + swizzle = tgsi_util_get_src_register_extswizzle( + ®->SrcRegisterExtSwz, + component ); + + assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); + assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); + assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); + assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); + + /* + * Second, calculate the simple swizzle for the unswizzled channel index. + * Leave the constants intact, they are not affected by the simple swizzle. + */ + if( swizzle <= TGSI_SWIZZLE_W ) { + swizzle = tgsi_util_get_src_register_swizzle( + ®->SrcRegister, + component ); + } + + return swizzle; +} + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + unsigned swizzle, + unsigned component ) +{ + switch( component ) { + case 0: + reg->SwizzleX = swizzle; + break; + case 1: + reg->SwizzleY = swizzle; + break; + case 2: + reg->SwizzleZ = swizzle; + break; + case 3: + reg->SwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + unsigned swizzle, + unsigned component ) +{ + switch( component ) { + case 0: + reg->ExtSwizzleX = swizzle; + break; + case 1: + reg->ExtSwizzleY = swizzle; + break; + case 2: + reg->ExtSwizzleZ = swizzle; + break; + case 3: + reg->ExtSwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->NegateX; + case 1: + return reg->NegateY; + case 2: + return reg->NegateZ; + case 3: + return reg->NegateW; + default: + assert( 0 ); + } + return 0; +} + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ) +{ + switch( component ) { + case 0: + reg->NegateX = negate; + break; + case 1: + reg->NegateY = negate; + break; + case 2: + reg->NegateZ = negate; + break; + case 3: + reg->NegateW = negate; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned sign_mode; + + if( reg->SrcRegisterExtMod.Absolute ) { + /* Consider only the post-abs negation. */ + + if( reg->SrcRegisterExtMod.Negate ) { + sign_mode = TGSI_UTIL_SIGN_SET; + } + else { + sign_mode = TGSI_UTIL_SIGN_CLEAR; + } + } + else { + /* Accumulate the three negations. */ + + unsigned negate; + + negate = reg->SrcRegister.Negate; + if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { + negate = !negate; + } + if( reg->SrcRegisterExtMod.Negate ) { + negate = !negate; + } + + if( negate ) { + sign_mode = TGSI_UTIL_SIGN_TOGGLE; + } + else { + sign_mode = TGSI_UTIL_SIGN_KEEP; + } + } + + return sign_mode; +} + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + unsigned sign_mode ) +{ + reg->SrcRegisterExtSwz.NegateX = 0; + reg->SrcRegisterExtSwz.NegateY = 0; + reg->SrcRegisterExtSwz.NegateZ = 0; + reg->SrcRegisterExtSwz.NegateW = 0; + + switch (sign_mode) + { + case TGSI_UTIL_SIGN_CLEAR: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_SET: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 1; + break; + + case TGSI_UTIL_SIGN_TOGGLE: + reg->SrcRegister.Negate = 1; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_KEEP: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + default: + assert( 0 ); + } +} + diff --git a/src/gallium/aux/tgsi/util/tgsi_util.h b/src/gallium/aux/tgsi/util/tgsi_util.h new file mode 100644 index 0000000000..ef14446f0e --- /dev/null +++ b/src/gallium/aux/tgsi/util/tgsi_util.h @@ -0,0 +1,70 @@ +#if !defined TGSI_UTIL_H +#define TGSI_UTIL_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +void * +tgsi_align_128bit( + void *unaligned ); + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component); + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + unsigned swizzle, + unsigned component ); + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + unsigned swizzle, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ); + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ); + +#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ +#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ +#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ +#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + unsigned sign_mode ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_UTIL_H + diff --git a/src/gallium/aux/util/p_debug.c b/src/gallium/aux/util/p_debug.c new file mode 100644 index 0000000000..b9607a6ba7 --- /dev/null +++ b/src/gallium/aux/util/p_debug.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include + +#ifdef WIN32 +#include +#include +#else +#include +#include +#endif + +#include "pipe/p_debug.h" +#include "pipe/p_compiler.h" + + +void debug_vprintf(const char *format, va_list ap) +{ +#ifdef WIN32 + EngDebugPrint("Gallium3D: ", (PCHAR)format, ap); +#else + vfprintf(stderr, format, ap); +#endif +} + + +void debug_printf(const char *format, ...) +{ + va_list ap; + va_start(ap, format); + debug_vprintf(format, ap); + va_end(ap); +} + + +static INLINE void debug_abort(void) +{ +#ifdef WIN32 + EngDebugBreak(); +#else + abort(); +#endif +} + + +void debug_assert_fail(const char *expr, const char *file, unsigned line) +{ + debug_printf("%s:%i: Assertion `%s' failed.\n", file, line, expr); + debug_abort(); +} diff --git a/src/gallium/aux/util/p_tile.c b/src/gallium/aux/util/p_tile.c new file mode 100644 index 0000000000..3f795a3898 --- /dev/null +++ b/src/gallium/aux/util/p_tile.c @@ -0,0 +1,699 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * RGBA/float tile get/put functions. + * Usable both by drivers and state trackers. + * Surfaces should already be in a mapped state. + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" + +#include "p_tile.h" + + + +/** + * Move raw block of pixels from surface to user memory. + * This should be usable by any hw driver that has mappable surfaces. + */ +void +pipe_get_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride) +{ + const uint cpp = ps->cpp; + const ubyte *pSrc; + const uint src_stride = ps->pitch * cpp; + ubyte *pDest; + uint i; + + if (dst_stride == 0) { + dst_stride = w * cpp; + } + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + pSrc = (const ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; + pDest = (ubyte *) p; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w * cpp); + pDest += dst_stride; + pSrc += src_stride; + } + + pipe_surface_unmap(ps); +} + + +/** + * Move raw block of pixels from user memory to surface. + * This should be usable by any hw driver that has mappable surfaces. + */ +void +pipe_put_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride) +{ + const uint cpp = ps->cpp; + const ubyte *pSrc; + const uint dst_stride = ps->pitch * cpp; + ubyte *pDest; + uint i; + + if (src_stride == 0) { + src_stride = w * cpp; + } + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + pSrc = (const ubyte *) p; + pDest = (ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w * cpp); + pDest += dst_stride; + pSrc += src_stride; + } + + pipe_surface_unmap(ps); +} + + + + +/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */ +#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) + +#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ + us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) + + + +/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ + +static void +a8r8g8b8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + } + p += dst_stride; + } +} + + +static void +a8r8g8b8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + *dst++ = (a << 24) | (r << 16) | (g << 8) | b; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ + +static void +b8g8r8a8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + } + p += dst_stride; + } +} + + +static void +b8g8r8a8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + *dst++ = (b << 24) | (g << 16) | (r << 8) | a; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ + +static void +a1r5g5b5_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); + pRow[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); + pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + pRow[3] = ((pixel >> 15) ) * 1.0f; + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ + +static void +a4r4g4b4_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); + pRow[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); + pRow[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); + pRow[3] = ((pixel >> 12) ) * (1.0f / 15.0f); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_R5G6B5_UNORM ***/ + +static void +r5g6b5_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); + pRow[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); + pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + pRow[3] = 1.0f; + } + p += dst_stride; + } +} + + +static void +r5g5b5_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); + uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); + uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); + *dst++ = (r << 11) | (g << 5) | (b); + } + p += src_stride; + } +} + + + +/*** PIPE_FORMAT_Z16_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z16_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const float scale = 1.0f / 65535.0f; + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = *src++ * scale; + } + p += dst_stride; + } +} + + + + +/*** PIPE_FORMAT_U_L8 ***/ + +static void +l8_get_tile_rgba(ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(*src); + pRow[3] = 1.0; + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_U_A8 ***/ + +static void +a8_get_tile_rgba(ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = 0.0; + pRow[3] = UBYTE_TO_FLOAT(*src); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ + +static void +r16g16b16a16_get_tile_rgba(short *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src += 4, pRow += 4) { + pRow[0] = SHORT_TO_FLOAT(src[0]); + pRow[1] = SHORT_TO_FLOAT(src[1]); + pRow[2] = SHORT_TO_FLOAT(src[2]); + pRow[3] = SHORT_TO_FLOAT(src[3]); + } + p += dst_stride; + } +} + + +static void +r16g16b16a16_put_tile_rgba(short *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, dst += 4, pRow += 4) { + UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); + UNCLAMPED_FLOAT_TO_SHORT(dst[1], pRow[1]); + UNCLAMPED_FLOAT_TO_SHORT(dst[2], pRow[2]); + UNCLAMPED_FLOAT_TO_SHORT(dst[3], pRow[3]); + } + p += src_stride; + } +} + + + +/*** PIPE_FORMAT_U_I8 ***/ + +static void +i8_get_tile_rgba(ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = UBYTE_TO_FLOAT(*src); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_U_A8_L8 ***/ + +static void +a8_l8_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + ushort p = *src++; + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(p & 0xff); + pRow[3] = UBYTE_TO_FLOAT(p >> 8); + } + p += dst_stride; + } +} + + + + +/*** PIPE_FORMAT_Z32_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z32_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const double scale = 1.0 / (double) 0xffffffff; + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (*src++ * scale); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_S8Z24_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +s8z24_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (scale * (*src++ & 0xffffff)); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_Z24S8_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +z24s8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (scale * (*src++ >> 8)); + } + p += dst_stride; + } +} + + +void +pipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p) +{ + unsigned dst_stride = w * 4; + void *packed; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + packed = MALLOC(h * w * ps->cpp); + + if (!packed) + return; + + pipe_get_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); + + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + a1r5g5b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_A4R4G4B4_UNORM: + a4r4g4b4_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g6b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_U_L8: + l8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_U_A8: + a8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_U_I8: + i8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_U_A8_L8: + a8_l8_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_get_tile_rgba((short *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_Z16_UNORM: + z16_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_Z32_UNORM: + z32_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_S8Z24_UNORM: + s8z24_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_Z24S8_UNORM: + z24s8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + default: + assert(0); + } + + FREE(packed); +} + + +void +pipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p) +{ + unsigned src_stride = w * 4; + void *packed; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + packed = MALLOC(h * w * ps->cpp); + + if (!packed) + return; + + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + /*a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + break; + case PIPE_FORMAT_U_L8: + /*l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_U_A8: + /*a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_U_I8: + /*i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_U_A8_L8: + /*a8_l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_Z16_UNORM: + /*z16_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_Z32_UNORM: + /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_S8Z24_UNORM: + /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_Z24S8_UNORM: + /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ + break; + default: + assert(0); + } + + pipe_put_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); + + FREE(packed); +} diff --git a/src/gallium/aux/util/p_tile.h b/src/gallium/aux/util/p_tile.h new file mode 100644 index 0000000000..cd8124bf11 --- /dev/null +++ b/src/gallium/aux/util/p_tile.h @@ -0,0 +1,81 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_TILE_H +#define P_TILE_H + +#include "pipe/p_compiler.h" + +struct pipe_context; +struct pipe_surface; + + +/** + * Clip tile against surface dims. + * \return TRUE if tile is totally clipped, FALSE otherwise + */ +static INLINE boolean +pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps) +{ + if (x >= ps->width) + return TRUE; + if (y >= ps->height) + return TRUE; + if (x + *w > ps->width) + *w = ps->width - x; + if (y + *h > ps->height) + *h = ps->height - y; + return FALSE; +} + + +extern void +pipe_get_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride); + +extern void +pipe_put_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride); + + +extern void +pipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p); + +extern void +pipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p); + +#endif diff --git a/src/gallium/aux/util/p_util.c b/src/gallium/aux/util/p_util.c new file mode 100644 index 0000000000..2a92f8e408 --- /dev/null +++ b/src/gallium/aux/util/p_util.c @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Miscellaneous utility functions. + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" + + +/** + * Copy 2D rect from one place to another. + * Position and sizes are in pixels. + */ +void +pipe_copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + int src_pitch, + unsigned src_x, + int src_y) +{ + unsigned i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * src_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } +} diff --git a/src/gallium/drivers/cell/Makefile b/src/gallium/drivers/cell/Makefile new file mode 100644 index 0000000000..47aef7b05f --- /dev/null +++ b/src/gallium/drivers/cell/Makefile @@ -0,0 +1,12 @@ +# Cell Gallium driver Makefile + + +default: + ( cd spu ; make ) + ( cd ppu ; make ) + + + +clean: + ( cd spu ; make clean ) + ( cd ppu ; make clean ) diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h new file mode 100644 index 0000000000..4de514c358 --- /dev/null +++ b/src/gallium/drivers/cell/common.h @@ -0,0 +1,220 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Types and tokens which are common to the SPU and PPU code. + */ + + +#ifndef CELL_COMMON_H +#define CELL_COMMON_H + +#include "pipe/p_compiler.h" +#include "pipe/p_util.h" +#include "pipe/p_format.h" + + +/** The standard assert macro doesn't seem to work reliably */ +#define ASSERT(x) \ + if (!(x)) { \ + ubyte *p = NULL; \ + fprintf(stderr, "%s:%d: %s(): assertion %s failed.\n", \ + __FILE__, __LINE__, __FUNCTION__, #x); \ + *p = 0; \ + exit(1); \ + } + + +/** for sanity checking */ +#define ASSERT_ALIGN16(ptr) \ + ASSERT((((unsigned long) (ptr)) & 0xf) == 0); + + +/** round up value to next multiple of 4 */ +#define ROUNDUP4(k) (((k) + 0x3) & ~0x3) + +/** round up value to next multiple of 8 */ +#define ROUNDUP8(k) (((k) + 0x7) & ~0x7) + +/** round up value to next multiple of 16 */ +#define ROUNDUP16(k) (((k) + 0xf) & ~0xf) + + +#define CELL_MAX_SPUS 6 + +#define TILE_SIZE 32 + + +/** + * The low byte of a mailbox word contains the command opcode. + * Remaining higher bytes are command specific. + */ +#define CELL_CMD_OPCODE_MASK 0xff + +#define CELL_CMD_EXIT 1 +#define CELL_CMD_CLEAR_SURFACE 2 +#define CELL_CMD_FINISH 3 +#define CELL_CMD_RENDER 4 +#define CELL_CMD_BATCH 5 +#define CELL_CMD_RELEASE_VERTS 6 +#define CELL_CMD_STATE_FRAMEBUFFER 10 +#define CELL_CMD_STATE_DEPTH_STENCIL 11 +#define CELL_CMD_STATE_SAMPLER 12 +#define CELL_CMD_STATE_TEXTURE 13 +#define CELL_CMD_STATE_VERTEX_INFO 14 +#define CELL_CMD_STATE_VIEWPORT 15 +#define CELL_CMD_STATE_VS_ARRAY_INFO 16 +#define CELL_CMD_STATE_BLEND 17 +#define CELL_CMD_VS_EXECUTE 18 + + +#define CELL_NUM_BUFFERS 4 +#define CELL_BUFFER_SIZE (4*1024) /**< 16KB would be the max */ + +#define CELL_BUFFER_STATUS_FREE 10 +#define CELL_BUFFER_STATUS_USED 20 + + + +/** + * Tell SPUs about the framebuffer size, location + */ +struct cell_command_framebuffer +{ + uint64_t opcode; /**< CELL_CMD_FRAMEBUFFER */ + int width, height; + void *color_start, *depth_start; + enum pipe_format color_format, depth_format; +}; + + +/** + * Clear framebuffer to the given value/color. + */ +struct cell_command_clear_surface +{ + uint64_t opcode; /**< CELL_CMD_CLEAR_SURFACE */ + uint surface; /**< Temporary: 0=color, 1=Z */ + uint value; +}; + + +/** + * Array info used by the vertex shader's vertex puller. + */ +struct cell_array_info +{ + uint64_t base; /**< Base address of the 0th element. */ + uint attr; /**< Attribute that this state is for. */ + uint pitch; /**< Byte pitch from one entry to the next. */ + uint format; /**< Pipe format of each entry. */ +} ALIGN16_ATTRIB; + + +struct cell_shader_info +{ + unsigned num_outputs; + + uint64_t declarations; + unsigned num_declarations; + uint64_t instructions; + unsigned num_instructions; + uint64_t uniforms; + uint64_t immediates; + unsigned num_immediates; +} ALIGN16_ATTRIB; + + +#define SPU_VERTS_PER_BATCH 64 +struct cell_command_vs +{ + uint64_t opcode; /**< CELL_CMD_VS_EXECUTE */ + struct cell_shader_info shader; + unsigned num_elts; + unsigned elts[SPU_VERTS_PER_BATCH]; + uint64_t vOut[SPU_VERTS_PER_BATCH]; + float plane[12][4]; + unsigned nr_planes; + unsigned nr_attrs; +} ALIGN16_ATTRIB; + + +struct cell_command_render +{ + uint64_t opcode; /**< CELL_CMD_RENDER */ + uint prim_type; /**< PIPE_PRIM_x */ + uint num_verts; + uint vertex_size; /**< bytes per vertex */ + uint num_indexes; + uint vertex_buf; /**< which cell->buffer[] contains the vertex data */ + float xmin, ymin, xmax, ymax; /* XXX another dummy field */ + uint min_index; + boolean inline_verts; +}; + + +struct cell_command_release_verts +{ + uint64_t opcode; /**< CELL_CMD_RELEASE_VERTS */ + uint vertex_buf; /**< in [0, CELL_NUM_BUFFERS-1] */ +}; + + +struct cell_command_texture +{ + void *start; /**< Address in main memory */ + uint width, height; +}; + + +/** XXX unions don't seem to work */ +/* XXX this should go away; all commands should be placed in batch buffers */ +struct cell_command +{ +#if 0 + struct cell_command_framebuffer fb; + struct cell_command_clear_surface clear; + struct cell_command_render render; +#endif + struct cell_command_vs vs; +} ALIGN16_ATTRIB; + + +/** This is the object passed to spe_create_thread() */ +struct cell_init_info +{ + unsigned id; + unsigned num_spus; + struct cell_command *cmd; + + /** Buffers for command batches, vertex/index data */ + ubyte *buffers[CELL_NUM_BUFFERS]; + uint *buffer_status; /**< points at cell_context->buffer_status */ +} ALIGN16_ATTRIB; + + +#endif /* CELL_COMMON_H */ diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile new file mode 100644 index 0000000000..50060f5cd3 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -0,0 +1,76 @@ +# Gallium3D Cell driver: PPU code + +# This makefile builds the g3dcell.a library which gets pulled into +# the main libGL.so library + + +TOP = ../../../../.. +include $(TOP)/configs/linux-cell + + +#PROG = gl4 + +CELL_LIB = libcell.a + +SPU_CODE_MODULE = ../spu/g3d_spu.a + + +SOURCES = \ + cell_batch.c \ + cell_clear.c \ + cell_context.c \ + cell_draw_arrays.c \ + cell_flush.c \ + cell_state_blend.c \ + cell_state_clip.c \ + cell_state_derived.c \ + cell_state_emit.c \ + cell_state_fs.c \ + cell_state_rasterizer.c \ + cell_state_sampler.c \ + cell_state_surface.c \ + cell_state_vertex.c \ + cell_spu.c \ + cell_surface.c \ + cell_texture.c \ + cell_vbuf.c \ + cell_vertex_shader.c \ + cell_winsys.c + + +OBJECTS = $(SOURCES:.c=.o) \ + +INCLUDE_DIRS = -I$(TOP)/src/mesa + + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + + + +default: $(CELL_LIB) + + +$(CELL_LIB): $(OBJECTS) $(SPU_CODE_MODULE) +# ar -ru $(CELL_LIB) $(OBJECTS) $(SPU_CODE_MODULE) + ar -ru $(CELL_LIB) $(OBJECTS) + +#$(PROG): $(PPU_OBJECTS) +# $(CC) -o $(PROG) $(PPU_OBJECTS) $(SPU_CODE_MODULE) $(PPU_LFLAGS) + + + +clean: + rm -f *.o *~ $(CELL_LIB) + + + +depend: $(SOURCES) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(SOURCES) 2> /dev/null + +include depend + + + diff --git a/src/gallium/drivers/cell/ppu/cell_batch.c b/src/gallium/drivers/cell/ppu/cell_batch.c new file mode 100644 index 0000000000..f45e5f25b6 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_batch.c @@ -0,0 +1,217 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "cell_context.h" +#include "cell_batch.h" +#include "cell_spu.h" + + + +uint +cell_get_empty_buffer(struct cell_context *cell) +{ + uint buf = 0, tries = 0; + + /* Find a buffer that's marked as free by all SPUs */ + while (1) { + uint spu, num_free = 0; + + for (spu = 0; spu < cell->num_spus; spu++) { + if (cell->buffer_status[spu][buf][0] == CELL_BUFFER_STATUS_FREE) { + num_free++; + + if (num_free == cell->num_spus) { + /* found a free buffer, now mark status as used */ + for (spu = 0; spu < cell->num_spus; spu++) { + cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED; + } + /* + printf("PPU: ALLOC BUFFER %u\n", buf); + */ + return buf; + } + } + else { + break; + } + } + + /* try next buf */ + buf = (buf + 1) % CELL_NUM_BUFFERS; + + tries++; + if (tries == 100) { + /* + printf("PPU WAITING for buffer...\n"); + */ + } + } +} + + +void +cell_batch_flush(struct cell_context *cell) +{ + static boolean flushing = FALSE; + uint batch = cell->cur_batch; + const uint size = cell->buffer_size[batch]; + uint spu, cmd_word; + + assert(!flushing); + + if (size == 0) + return; + + flushing = TRUE; + + assert(batch < CELL_NUM_BUFFERS); + + /* + printf("cell_batch_dispatch: buf %u at %p, size %u\n", + batch, &cell->batch_buffer[batch][0], size); + */ + + /* + * Build "BATCH" command and sent to all SPUs. + */ + cmd_word = CELL_CMD_BATCH | (batch << 8) | (size << 16); + + for (spu = 0; spu < cell->num_spus; spu++) { + assert(cell->buffer_status[spu][batch][0] == CELL_BUFFER_STATUS_USED); + send_mbox_message(cell_global.spe_contexts[spu], cmd_word); + } + + /* When the SPUs are done copying the buffer into their locals stores + * they'll write a BUFFER_STATUS_FREE message into the buffer_status[] + * array indicating that the PPU can re-use the buffer. + */ + + batch = cell_get_empty_buffer(cell); + + cell->buffer_size[batch] = 0; /* empty */ + cell->cur_batch = batch; + + flushing = FALSE; +} + + +uint +cell_batch_free_space(const struct cell_context *cell) +{ + uint free = CELL_BUFFER_SIZE - cell->buffer_size[cell->cur_batch]; + return free; +} + + +/** + * Append data to current batch. + */ +void +cell_batch_append(struct cell_context *cell, const void *data, uint bytes) +{ + uint size; + + ASSERT(bytes % 8 == 0); + ASSERT(bytes <= CELL_BUFFER_SIZE); + ASSERT(cell->cur_batch >= 0); + +#ifdef ASSERT + { + uint spu; + for (spu = 0; spu < cell->num_spus; spu++) { + ASSERT(cell->buffer_status[spu][cell->cur_batch][0] + == CELL_BUFFER_STATUS_USED); + } + } +#endif + + size = cell->buffer_size[cell->cur_batch]; + + if (size + bytes > CELL_BUFFER_SIZE) { + cell_batch_flush(cell); + size = 0; + } + + ASSERT(size + bytes <= CELL_BUFFER_SIZE); + + memcpy(cell->buffer[cell->cur_batch] + size, data, bytes); + + cell->buffer_size[cell->cur_batch] = size + bytes; +} + + +void * +cell_batch_alloc(struct cell_context *cell, uint bytes) +{ + return cell_batch_alloc_aligned(cell, bytes, 1); +} + + +void * +cell_batch_alloc_aligned(struct cell_context *cell, uint bytes, + uint alignment) +{ + void *pos; + uint size, padbytes; + + ASSERT(bytes % 8 == 0); + ASSERT(bytes <= CELL_BUFFER_SIZE); + ASSERT(alignment > 0); + ASSERT(cell->cur_batch >= 0); + +#ifdef ASSERT + { + uint spu; + for (spu = 0; spu < cell->num_spus; spu++) { + ASSERT(cell->buffer_status[spu][cell->cur_batch][0] + == CELL_BUFFER_STATUS_USED); + } + } +#endif + + size = cell->buffer_size[cell->cur_batch]; + + padbytes = (alignment - (size % alignment)) % alignment; + + if (padbytes + size + bytes > CELL_BUFFER_SIZE) { + cell_batch_flush(cell); + size = 0; + } + else { + size += padbytes; + } + + ASSERT(size % alignment == 0); + ASSERT(size + bytes <= CELL_BUFFER_SIZE); + + pos = (void *) (cell->buffer[cell->cur_batch] + size); + + cell->buffer_size[cell->cur_batch] = size + bytes; + + return pos; +} diff --git a/src/gallium/drivers/cell/ppu/cell_batch.h b/src/gallium/drivers/cell/ppu/cell_batch.h new file mode 100644 index 0000000000..a6eee0a8b1 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_batch.h @@ -0,0 +1,58 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_BATCH_H +#define CELL_BATCH_H + +#include "pipe/p_compiler.h" + + +struct cell_context; + + +extern uint +cell_get_empty_buffer(struct cell_context *cell); + +extern void +cell_batch_flush(struct cell_context *cell); + +extern uint +cell_batch_free_space(const struct cell_context *cell); + +extern void +cell_batch_append(struct cell_context *cell, const void *data, uint bytes); + +extern void * +cell_batch_alloc(struct cell_context *cell, uint bytes); + +extern void * +cell_batch_alloc_aligned(struct cell_context *cell, uint bytes, + uint alignment); + + +#endif /* CELL_BATCH_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_clear.c b/src/gallium/drivers/cell/ppu/cell_clear.c new file mode 100644 index 0000000000..07b908eec5 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_clear.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Authors + * Brian Paul + */ + +#include +#include +#include +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/cell/common.h" +#include "cell_clear.h" +#include "cell_context.h" +#include "cell_batch.h" +#include "cell_flush.h" +#include "cell_spu.h" + + +void +cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + struct cell_context *cell = cell_context(pipe); + uint surfIndex; + + if (cell->dirty) + cell_update_derived(cell); + + + if (!cell->cbuf_map[0]) + cell->cbuf_map[0] = pipe_surface_map(ps); + + if (ps == cell->framebuffer.zsbuf) { + surfIndex = 1; + } + else { + surfIndex = 0; + } + + + { + struct cell_command_clear_surface *clr + = (struct cell_command_clear_surface *) + cell_batch_alloc(cell, sizeof(*clr)); + clr->opcode = CELL_CMD_CLEAR_SURFACE; + clr->surface = surfIndex; + clr->value = clearValue; + } +} diff --git a/src/gallium/drivers/cell/ppu/cell_clear.h b/src/gallium/drivers/cell/ppu/cell_clear.h new file mode 100644 index 0000000000..ff47d43f4c --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_clear.h @@ -0,0 +1,43 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_CLEAR_H +#define CELL_CLEAR_H + + +struct pipe_context; +struct pipe_surface; + + +extern void +cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + + + +#endif /* CELL_CLEAR_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c new file mode 100644 index 0000000000..bbe1fd7a11 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -0,0 +1,287 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Authors + * Brian Paul + */ + + +#include + +#include "pipe/p_defines.h" +#include "pipe/p_format.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#include "pipe/cell/common.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_private.h" +#include "cell_clear.h" +#include "cell_context.h" +#include "cell_draw_arrays.h" +#include "cell_flush.h" +#include "cell_render.h" +#include "cell_state.h" +#include "cell_surface.h" +#include "cell_spu.h" +#include "cell_texture.h" +#include "cell_vbuf.h" + + + +static boolean +cell_is_format_supported( struct pipe_context *pipe, + enum pipe_format format, uint type ) +{ + /*struct cell_context *cell = cell_context( pipe );*/ + + switch (type) { + case PIPE_TEXTURE: + /* cell supports all texture formats, XXX for now anyway */ + return TRUE; + case PIPE_SURFACE: + /* cell supports all (off-screen) surface formats, XXX for now */ + return TRUE; + default: + assert(0); + return FALSE; + } +} + + +static int cell_get_param(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 1; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 1; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 1; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 12; /* max 2Kx2K */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 12; /* max 2Kx2K */ + default: + return 0; + } +} + +static float cell_get_paramf(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 0.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; /* arbitrary */ + + default: + return 0; + } +} + + +static const char * +cell_get_name( struct pipe_context *pipe ) +{ + return "Cell"; +} + +static const char * +cell_get_vendor( struct pipe_context *pipe ) +{ + return "Tungsten Graphics, Inc."; +} + + + +static void +cell_destroy_context( struct pipe_context *pipe ) +{ + struct cell_context *cell = cell_context(pipe); + + cell_spu_exit(cell); + + align_free(cell); +} + + +static struct draw_context * +cell_draw_create(struct cell_context *cell) +{ + struct draw_context *draw = draw_create(); + + if (getenv("GALLIUM_CELL_VS")) { + /* plug in SPU-based vertex transformation code */ + draw->shader_queue_flush = cell_vertex_shader_queue_flush; + draw->driver_private = cell; + } + + return draw; +} + + +struct pipe_context * +cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) +{ + struct cell_context *cell; + uint spu, buf; + + /* some fields need to be 16-byte aligned, so align the whole object */ + cell = (struct cell_context*) align_malloc(sizeof(struct cell_context), 16); + if (!cell) + return NULL; + + memset(cell, 0, sizeof(*cell)); + + cell->winsys = cws; + cell->pipe.winsys = winsys; + cell->pipe.destroy = cell_destroy_context; + + /* queries */ + cell->pipe.is_format_supported = cell_is_format_supported; + cell->pipe.get_name = cell_get_name; + cell->pipe.get_vendor = cell_get_vendor; + cell->pipe.get_param = cell_get_param; + cell->pipe.get_paramf = cell_get_paramf; + + + /* state setters */ + cell->pipe.create_blend_state = cell_create_blend_state; + cell->pipe.bind_blend_state = cell_bind_blend_state; + cell->pipe.delete_blend_state = cell_delete_blend_state; + + cell->pipe.create_sampler_state = cell_create_sampler_state; + cell->pipe.bind_sampler_state = cell_bind_sampler_state; + cell->pipe.delete_sampler_state = cell_delete_sampler_state; + + cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state; + cell->pipe.bind_depth_stencil_alpha_state = cell_bind_depth_stencil_alpha_state; + cell->pipe.delete_depth_stencil_alpha_state = cell_delete_depth_stencil_alpha_state; + + cell->pipe.create_rasterizer_state = cell_create_rasterizer_state; + cell->pipe.bind_rasterizer_state = cell_bind_rasterizer_state; + cell->pipe.delete_rasterizer_state = cell_delete_rasterizer_state; + + cell->pipe.create_fs_state = cell_create_fs_state; + cell->pipe.bind_fs_state = cell_bind_fs_state; + cell->pipe.delete_fs_state = cell_delete_fs_state; + + cell->pipe.create_vs_state = cell_create_vs_state; + cell->pipe.bind_vs_state = cell_bind_vs_state; + cell->pipe.delete_vs_state = cell_delete_vs_state; + + cell->pipe.set_blend_color = cell_set_blend_color; + cell->pipe.set_clip_state = cell_set_clip_state; + cell->pipe.set_constant_buffer = cell_set_constant_buffer; + + cell->pipe.set_framebuffer_state = cell_set_framebuffer_state; + + cell->pipe.set_polygon_stipple = cell_set_polygon_stipple; + cell->pipe.set_scissor_state = cell_set_scissor_state; + cell->pipe.set_viewport_state = cell_set_viewport_state; + + cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; + cell->pipe.set_vertex_element = cell_set_vertex_element; + + cell->pipe.draw_arrays = cell_draw_arrays; + cell->pipe.draw_elements = cell_draw_elements; + + cell->pipe.clear = cell_clear_surface; + cell->pipe.flush = cell_flush; + + /* textures */ + cell->pipe.texture_create = cell_texture_create; + cell->pipe.texture_release = cell_texture_release; + cell->pipe.get_tex_surface = cell_get_tex_surface; + + cell->pipe.set_sampler_texture = cell_set_sampler_texture; + +#if 0 + cell->pipe.begin_query = cell_begin_query; + cell->pipe.end_query = cell_end_query; + cell->pipe.wait_query = cell_wait_query; +#endif + + cell_init_surface_functions(cell); + + cell->draw = cell_draw_create(cell); + + cell_init_vbuf(cell); + draw_set_rasterize_stage(cell->draw, cell->vbuf); + + /* + * SPU stuff + */ + cell->num_spus = 6; + + cell_start_spus(cell); + + /* init command, vertex/index buffer info */ + for (buf = 0; buf < CELL_NUM_BUFFERS; buf++) { + cell->buffer_size[buf] = 0; + + /* init batch buffer status values, + * mark 0th buffer as used, rest as free. + */ + for (spu = 0; spu < cell->num_spus; spu++) { + if (buf == 0) + cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED; + else + cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_FREE; + } + } + + return &cell->pipe; +} diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h new file mode 100644 index 0000000000..3b63419b5e --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -0,0 +1,135 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_CONTEXT_H +#define CELL_CONTEXT_H + + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/draw/draw_vertex.h" +#include "pipe/draw/draw_vbuf.h" +#include "cell_winsys.h" +#include "pipe/cell/common.h" + + +struct cell_vbuf_render; + +struct cell_vertex_shader_state +{ + struct pipe_shader_state shader; + void *draw_data; +}; + + +struct cell_fragment_shader_state +{ + struct pipe_shader_state shader; + void *data; +}; + + +struct cell_context +{ + struct pipe_context pipe; + + struct cell_winsys *winsys; + + const struct pipe_blend_state *blend; + const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct pipe_depth_stencil_alpha_state *depth_stencil; + const struct pipe_rasterizer_state *rasterizer; + const struct cell_vertex_shader_state *vs; + const struct cell_fragment_shader_state *fs; + + struct pipe_blend_color blend_color; + struct pipe_clip_state clip; + struct pipe_constant_buffer constants[2]; + struct pipe_framebuffer_state framebuffer; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct cell_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + + ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS]; + ubyte *zsbuf_map; + + struct pipe_surface *tex_surf; + uint *tex_map; + + uint dirty; + + /** The primitive drawing context */ + struct draw_context *draw; + struct draw_stage *render_stage; + + /** For post-transformed vertex buffering: */ + struct cell_vbuf_render *vbuf_render; + struct draw_stage *vbuf; + + struct vertex_info vertex_info; + + /** Mapped constant buffers */ + void *mapped_constants[PIPE_SHADER_TYPES]; + + + uint num_spus; + + /** Buffers for command batches, vertex/index data */ + uint buffer_size[CELL_NUM_BUFFERS]; + ubyte buffer[CELL_NUM_BUFFERS][CELL_BUFFER_SIZE] ALIGN16_ATTRIB; + + int cur_batch; /**< which buffer is being filled w/ commands */ + + /** [4] to ensure 16-byte alignment for each status word */ + uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BUFFERS][4] ALIGN16_ATTRIB; + +}; + + + + +static INLINE struct cell_context * +cell_context(struct pipe_context *pipe) +{ + return (struct cell_context *) pipe; +} + + +extern struct pipe_context * +cell_create_context(struct pipe_winsys *ws, struct cell_winsys *cws); + +extern void +cell_vertex_shader_queue_flush(struct draw_context *draw); + + + + +#endif /* CELL_CONTEXT_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c new file mode 100644 index 0000000000..717cd8370f --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c @@ -0,0 +1,164 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +#include "cell_context.h" +#include "cell_draw_arrays.h" +#include "cell_state.h" + +#include "pipe/draw/draw_context.h" + + + +static void +cell_map_constant_buffers(struct cell_context *sp) +{ + struct pipe_winsys *ws = sp->pipe.winsys; + uint i; + for (i = 0; i < 2; i++) { + if (sp->constants[i].size) + sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + } + + draw_set_mapped_constant_buffer(sp->draw, + sp->mapped_constants[PIPE_SHADER_VERTEX]); +} + +static void +cell_unmap_constant_buffers(struct cell_context *sp) +{ + struct pipe_winsys *ws = sp->pipe.winsys; + uint i; + for (i = 0; i < 2; i++) { + if (sp->constants[i].size) + ws->buffer_unmap(ws, sp->constants[i].buffer); + sp->mapped_constants[i] = NULL; + } +} + + +boolean +cell_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count) +{ + return cell_draw_elements(pipe, NULL, 0, mode, start, count); +} + + + +/** + * Draw vertex arrays, with optional indexing. + * Basically, map the vertex buffers (and drawing surfaces), then hand off + * the drawing to the 'draw' module. + * + * XXX should the element buffer be specified/bound with a separate function? + */ +boolean +cell_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + struct cell_context *sp = cell_context(pipe); + struct draw_context *draw = sp->draw; + unsigned i; + + /* first, check that the primitive is not malformed. It is the + * state tracker's responsibility to do send only correctly formed + * primitives down. It currently isn't doing that though... + */ +#if 1 + count = draw_trim_prim( mode, count ); +#else + if (!draw_validate_prim( mode, count )) + assert(0); +#endif + + if (sp->dirty) + cell_update_derived( sp ); + +#if 0 + cell_map_surfaces(sp); +#endif + cell_map_constant_buffers(sp); + + /* + * Map vertex buffers + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (sp->vertex_buffer[i].buffer) { + void *buf = pipe->winsys->buffer_map(pipe->winsys, + sp->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); + } + } + /* Map index buffer, if present */ + if (indexBuffer) { + void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, + indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + } + else { + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + + /* draw! */ + draw_arrays(draw, mode, start, count); + + /* + * unmap vertex/index buffers - will cause draw module to flush + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (sp->vertex_buffer[i].buffer) { + draw_set_mapped_vertex_buffer(draw, i, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); + } + } + if (indexBuffer) { + draw_set_mapped_element_buffer(draw, 0, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + } + + /* Note: leave drawing surfaces mapped */ + cell_unmap_constant_buffers(sp); + + return TRUE; +} diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.h b/src/gallium/drivers/cell/ppu/cell_draw_arrays.h new file mode 100644 index 0000000000..d5df4aa05f --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_DRAW_ARRAYS_H +#define CELL_DRAW_ARRAYS_H + + +boolean cell_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count); + +boolean cell_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); + + + +#endif /* CELL_DRAW_ARRAYS_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_flush.c b/src/gallium/drivers/cell/ppu/cell_flush.c new file mode 100644 index 0000000000..f62bc4650c --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_flush.c @@ -0,0 +1,84 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "cell_context.h" +#include "cell_batch.h" +#include "cell_flush.h" +#include "cell_spu.h" +#include "cell_render.h" +#include "pipe/draw/draw_context.h" + + +void +cell_flush(struct pipe_context *pipe, unsigned flags) +{ + struct cell_context *cell = cell_context(pipe); + + if (flags & PIPE_FLUSH_SWAPBUFFERS) + flags |= PIPE_FLUSH_WAIT; + + draw_flush( cell->draw ); + cell_flush_int(pipe, flags); +} + + +/** internal flush */ +void +cell_flush_int(struct pipe_context *pipe, unsigned flags) +{ + static boolean flushing = FALSE; /* recursion catcher */ + struct cell_context *cell = cell_context(pipe); + uint i; + + ASSERT(!flushing); + flushing = TRUE; + + if (flags & PIPE_FLUSH_WAIT) { + uint64_t *cmd = (uint64_t *) cell_batch_alloc(cell, sizeof(uint64_t)); + *cmd = CELL_CMD_FINISH; + } + + cell_batch_flush(cell); + +#if 0 + /* Send CMD_FINISH to all SPUs */ + for (i = 0; i < cell->num_spus; i++) { + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FINISH); + } +#endif + + if (flags & PIPE_FLUSH_WAIT) { + /* Wait for ack */ + for (i = 0; i < cell->num_spus; i++) { + uint k = wait_mbox_message(cell_global.spe_contexts[i]); + assert(k == CELL_CMD_FINISH); + } + } + + flushing = FALSE; +} diff --git a/src/gallium/drivers/cell/ppu/cell_flush.h b/src/gallium/drivers/cell/ppu/cell_flush.h new file mode 100644 index 0000000000..eda351b1cb --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_flush.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_FLUSH +#define CELL_FLUSH + +extern void +cell_flush(struct pipe_context *pipe, unsigned flags); + +extern void +cell_flush_int(struct pipe_context *pipe, unsigned flags); + +#endif diff --git a/src/gallium/drivers/cell/ppu/cell_render.c b/src/gallium/drivers/cell/ppu/cell_render.c new file mode 100644 index 0000000000..4ab277a4b2 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_render.c @@ -0,0 +1,210 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Last stage of 'draw' pipeline: send tris to SPUs. + * \author Brian Paul + */ + +#include "cell_context.h" +#include "cell_render.h" +#include "cell_spu.h" +#include "pipe/p_util.h" +#include "pipe/draw/draw_private.h" + + +struct render_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct cell_context *cell; +}; + + +static INLINE struct render_stage * +render_stage(struct draw_stage *stage) +{ + return (struct render_stage *) stage; +} + + +static void render_begin( struct draw_stage *stage ) +{ +#if 0 + struct render_stage *render = render_stage(stage); + struct cell_context *sp = render->cell; + const struct pipe_shader_state *fs = &render->cell->fs->shader; + render->quad.nr_attrs = render->cell->nr_frag_attrs; + + render->firstFpInput = fs->input_semantic_name[0]; + + sp->quad.first->begin(sp->quad.first); +#endif +} + + +static void render_end( struct draw_stage *stage ) +{ +} + + +static void reset_stipple_counter( struct draw_stage *stage ) +{ + struct render_stage *render = render_stage(stage); + /*render->cell->line_stipple_counter = 0;*/ +} + + +static void +render_point(struct draw_stage *stage, struct prim_header *prim) +{ +} + + +static void +render_line(struct draw_stage *stage, struct prim_header *prim) +{ +} + + +/** Write a vertex into the prim buffer */ +static void +save_vertex(struct cell_prim_buffer *buf, uint pos, + const struct vertex_header *vert) +{ + uint attr, j; + + for (attr = 0; attr < 2; attr++) { + for (j = 0; j < 4; j++) { + buf->vertex[pos][attr][j] = vert->data[attr][j]; + } + } + + /* update bounding box */ + if (vert->data[0][0] < buf->xmin) + buf->xmin = vert->data[0][0]; + if (vert->data[0][0] > buf->xmax) + buf->xmax = vert->data[0][0]; + if (vert->data[0][1] < buf->ymin) + buf->ymin = vert->data[0][1]; + if (vert->data[0][1] > buf->ymax) + buf->ymax = vert->data[0][1]; +} + + +static void +render_tri(struct draw_stage *stage, struct prim_header *prim) +{ + struct render_stage *rs = render_stage(stage); + struct cell_context *cell = rs->cell; + struct cell_prim_buffer *buf = &cell->prim_buffer; + uint i; + + if (buf->num_verts + 3 > CELL_MAX_VERTS) { + cell_flush_prim_buffer(cell); + } + + i = buf->num_verts; + assert(i+2 <= CELL_MAX_VERTS); + save_vertex(buf, i+0, prim->v[0]); + save_vertex(buf, i+1, prim->v[1]); + save_vertex(buf, i+2, prim->v[2]); + buf->num_verts += 3; +} + + +/** + * Send the a RENDER command to all SPUs to have them render the prims + * in the current prim_buffer. + */ +void +cell_flush_prim_buffer(struct cell_context *cell) +{ + uint i; + + if (cell->prim_buffer.num_verts == 0) + return; + + for (i = 0; i < cell->num_spus; i++) { + struct cell_command_render *render = &cell_global.command[i].render; + render->prim_type = PIPE_PRIM_TRIANGLES; + render->num_verts = cell->prim_buffer.num_verts; + render->vertex_size = cell->vertex_info->size * 4; + render->xmin = cell->prim_buffer.xmin; + render->ymin = cell->prim_buffer.ymin; + render->xmax = cell->prim_buffer.xmax; + render->ymax = cell->prim_buffer.ymax; + render->vertex_data = &cell->prim_buffer.vertex; + ASSERT_ALIGN16(render->vertex_data); + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER); + } + + cell->prim_buffer.num_verts = 0; + + cell->prim_buffer.xmin = 1e100; + cell->prim_buffer.ymin = 1e100; + cell->prim_buffer.xmax = -1e100; + cell->prim_buffer.ymax = -1e100; + + /* XXX temporary, need to double-buffer the prim buffer until we get + * a real command buffer/list system. + */ + cell_flush(&cell->pipe, 0x0); +} + + + +static void render_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + +/** + * Create a new draw/render stage. This will be plugged into the + * draw module as the last pipeline stage. + */ +struct draw_stage *cell_draw_render_stage( struct cell_context *cell ) +{ + struct render_stage *render = CALLOC_STRUCT(render_stage); + + render->cell = cell; + render->stage.draw = cell->draw; + render->stage.begin = render_begin; + render->stage.point = render_point; + render->stage.line = render_line; + render->stage.tri = render_tri; + render->stage.end = render_end; + render->stage.reset_stipple_counter = reset_stipple_counter; + render->stage.destroy = render_destroy; + + /* + render->quad.coef = render->coef; + render->quad.posCoef = &render->posCoef; + */ + + return &render->stage; +} diff --git a/src/gallium/drivers/cell/ppu/cell_render.h b/src/gallium/drivers/cell/ppu/cell_render.h new file mode 100644 index 0000000000..826dcbafeb --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_render.h @@ -0,0 +1,39 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_RENDER_H +#define CELL_RENDER_H + +struct cell_context; +struct draw_stage; + +extern void +cell_flush_prim_buffer(struct cell_context *cell); + +extern struct draw_stage *cell_draw_render_stage( struct cell_context *cell ); + +#endif /* CELL_RENDER_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_spu.c b/src/gallium/drivers/cell/ppu/cell_spu.c new file mode 100644 index 0000000000..7c83a47e57 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_spu.c @@ -0,0 +1,155 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include + +#include "cell_spu.h" +#include "pipe/p_format.h" +#include "pipe/p_state.h" +#include "pipe/cell/common.h" + + +/* +helpful headers: +/opt/ibm/cell-sdk/prototype/src/include/ppu/cbe_mfc.h +*/ + + +struct cell_global_info cell_global; + + +/** + * Write a 1-word message to the given SPE mailbox. + */ +void +send_mbox_message(spe_context_ptr_t ctx, unsigned int msg) +{ + spe_in_mbox_write(ctx, &msg, 1, SPE_MBOX_ALL_BLOCKING); +} + + +/** + * Wait for a 1-word message to arrive in given mailbox. + */ +uint +wait_mbox_message(spe_context_ptr_t ctx) +{ + do { + unsigned data; + int count = spe_out_mbox_read(ctx, &data, 1); + + if (count == 1) { + return data; + } + + if (count < 0) { + /* error */ ; + } + } while (1); +} + + +static void *cell_thread_function(void *arg) +{ + struct cell_init_info *init = (struct cell_init_info *) arg; + unsigned entry = SPE_DEFAULT_ENTRY; + + ASSERT_ALIGN16(init); + + if (spe_context_run(cell_global.spe_contexts[init->id], &entry, 0, + init, NULL, NULL) < 0) { + fprintf(stderr, "spe_context_run() failed\n"); + exit(1); + } + + pthread_exit(NULL); +} + + +/** + * Create the SPU threads + */ +void +cell_start_spus(struct cell_context *cell) +{ + uint i, j; + + assert(cell->num_spus <= MAX_SPUS); + + ASSERT_ALIGN16(&cell_global.command[0]); + ASSERT_ALIGN16(&cell_global.command[1]); + + ASSERT_ALIGN16(&cell_global.inits[0]); + ASSERT_ALIGN16(&cell_global.inits[1]); + + for (i = 0; i < cell->num_spus; i++) { + cell_global.inits[i].id = i; + cell_global.inits[i].num_spus = cell->num_spus; + cell_global.inits[i].cmd = &cell_global.command[i]; + for (j = 0; j < CELL_NUM_BUFFERS; j++) { + cell_global.inits[i].buffers[j] = cell->buffer[j]; + } + cell_global.inits[i].buffer_status = &cell->buffer_status[0][0][0]; + + cell_global.spe_contexts[i] = spe_context_create(0, NULL); + if (!cell_global.spe_contexts[i]) { + fprintf(stderr, "spe_context_create() failed\n"); + exit(1); + } + + if (spe_program_load(cell_global.spe_contexts[i], &g3d_spu)) { + fprintf(stderr, "spe_program_load() failed\n"); + exit(1); + } + + pthread_create(&cell_global.spe_threads[i], NULL, &cell_thread_function, + &cell_global.inits[i]); + } +} + + +/** + * Tell all the SPUs to stop/exit. + */ +void +cell_spu_exit(struct cell_context *cell) +{ + uint i; + + for (i = 0; i < cell->num_spus; i++) { + send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_EXIT); + } + + /* wait for threads to exit */ + for (i = 0; i < cell->num_spus; i++) { + void *value; + pthread_join(cell_global.spe_threads[i], &value); + cell_global.spe_threads[i] = 0; + cell_global.spe_contexts[i] = 0; + } +} diff --git a/src/gallium/drivers/cell/ppu/cell_spu.h b/src/gallium/drivers/cell/ppu/cell_spu.h new file mode 100644 index 0000000000..19eff94f96 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_spu.h @@ -0,0 +1,82 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_SPU +#define CELL_SPU + + +#include +#include +#include "pipe/cell/common.h" + +#include "cell_context.h" + + +#define MAX_SPUS 8 + +/** + * Global vars, for now anyway. + */ +struct cell_global_info +{ + /** + * SPU/SPE handles, etc + */ + spe_context_ptr_t spe_contexts[MAX_SPUS]; + pthread_t spe_threads[MAX_SPUS]; + + /** + * Data sent to SPUs + */ + struct cell_init_info inits[MAX_SPUS]; + struct cell_command command[MAX_SPUS]; +}; + + +extern struct cell_global_info cell_global; + + +/** This is the handle for the actual SPE code */ +extern spe_program_handle_t g3d_spu; + + +extern void +send_mbox_message(spe_context_ptr_t ctx, unsigned int msg); + +extern uint +wait_mbox_message(spe_context_ptr_t ctx); + + +extern void +cell_start_spus(struct cell_context *cell); + + +extern void +cell_spu_exit(struct cell_context *cell); + + +#endif /* CELL_SPU */ diff --git a/src/gallium/drivers/cell/ppu/cell_state.h b/src/gallium/drivers/cell/ppu/cell_state.h new file mode 100644 index 0000000000..3a71ba14fa --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state.h @@ -0,0 +1,115 @@ + + +#ifndef CELL_STATE_H +#define CELL_STATE_H + + +#define CELL_NEW_VIEWPORT 0x1 +#define CELL_NEW_RASTERIZER 0x2 +#define CELL_NEW_FS 0x4 +#define CELL_NEW_BLEND 0x8 +#define CELL_NEW_CLIP 0x10 +#define CELL_NEW_SCISSOR 0x20 +#define CELL_NEW_STIPPLE 0x40 +#define CELL_NEW_FRAMEBUFFER 0x80 +#define CELL_NEW_ALPHA_TEST 0x100 +#define CELL_NEW_DEPTH_STENCIL 0x200 +#define CELL_NEW_SAMPLER 0x400 +#define CELL_NEW_TEXTURE 0x800 +#define CELL_NEW_VERTEX 0x1000 +#define CELL_NEW_VS 0x2000 +#define CELL_NEW_CONSTANTS 0x4000 +#define CELL_NEW_VERTEX_INFO 0x8000 + + + +extern void +cell_set_framebuffer_state( struct pipe_context *, + const struct pipe_framebuffer_state * ); + + + +extern void * +cell_create_blend_state(struct pipe_context *, const struct pipe_blend_state *); +extern void cell_bind_blend_state(struct pipe_context *, void *); +extern void cell_delete_blend_state(struct pipe_context *, void *); + +extern void cell_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ); + + +void * +cell_create_sampler_state(struct pipe_context *, + const struct pipe_sampler_state *); + +extern void +cell_bind_sampler_state(struct pipe_context *, unsigned, void *); + +extern void +cell_delete_sampler_state(struct pipe_context *, void *); + + +extern void * +cell_create_depth_stencil_alpha_state(struct pipe_context *, + const struct pipe_depth_stencil_alpha_state *); + +extern void +cell_bind_depth_stencil_alpha_state(struct pipe_context *, void *); + +extern void +cell_delete_depth_stencil_alpha_state(struct pipe_context *, void *); + + +void *cell_create_fs_state(struct pipe_context *, + const struct pipe_shader_state *); +void cell_bind_fs_state(struct pipe_context *, void *); +void cell_delete_fs_state(struct pipe_context *, void *); +void *cell_create_vs_state(struct pipe_context *, + const struct pipe_shader_state *); +void cell_bind_vs_state(struct pipe_context *, void *); +void cell_delete_vs_state(struct pipe_context *, void *); + + +void * +cell_create_rasterizer_state(struct pipe_context *, + const struct pipe_rasterizer_state *); +void cell_bind_rasterizer_state(struct pipe_context *, void *); +void cell_delete_rasterizer_state(struct pipe_context *, void *); + + +void cell_set_clip_state( struct pipe_context *, + const struct pipe_clip_state * ); + +void cell_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf); + +void cell_set_polygon_stipple( struct pipe_context *, + const struct pipe_poly_stipple * ); + +void +cell_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, + struct pipe_texture *texture); + +void cell_set_scissor_state( struct pipe_context *, + const struct pipe_scissor_state * ); + +void cell_set_texture_state( struct pipe_context *, + unsigned unit, struct pipe_texture * ); + +void cell_set_vertex_element(struct pipe_context *, + unsigned index, + const struct pipe_vertex_element *); + +void cell_set_vertex_buffer(struct pipe_context *, + unsigned index, + const struct pipe_vertex_buffer *); + +void cell_set_viewport_state( struct pipe_context *, + const struct pipe_viewport_state * ); + + +void cell_update_derived( struct cell_context *softpipe ); + +#endif diff --git a/src/gallium/drivers/cell/ppu/cell_state_blend.c b/src/gallium/drivers/cell/ppu/cell_state_blend.c new file mode 100644 index 0000000000..4fc60548c8 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_blend.c @@ -0,0 +1,109 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/draw/draw_context.h" +#include "cell_context.h" +#include "cell_state.h" + + + +void * +cell_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + return mem_dup(blend, sizeof(*blend)); +} + + +void +cell_bind_blend_state(struct pipe_context *pipe, void *blend) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + cell->blend = (const struct pipe_blend_state *)blend; + + cell->dirty |= CELL_NEW_BLEND; +} + + +void +cell_delete_blend_state(struct pipe_context *pipe, void *blend) +{ + FREE(blend); +} + + +void +cell_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *blend_color) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + cell->blend_color = *blend_color; + + cell->dirty |= CELL_NEW_BLEND; +} + + + + +void * +cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil) +{ + return mem_dup(depth_stencil, sizeof(*depth_stencil)); +} + + +void +cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + cell->depth_stencil + = (const struct pipe_depth_stencil_alpha_state *) depth_stencil; + + cell->dirty |= CELL_NEW_DEPTH_STENCIL; +} + + +void +cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth) +{ + FREE(depth); +} diff --git a/src/gallium/drivers/cell/ppu/cell_state_clip.c b/src/gallium/drivers/cell/ppu/cell_state_clip.c new file mode 100644 index 0000000000..4f43665941 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_clip.c @@ -0,0 +1,84 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "cell_context.h" +#include "cell_state.h" +#include "pipe/draw/draw_context.h" + + +void cell_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct cell_context *cell = cell_context(pipe); + + /* pass the clip state to the draw module */ + draw_set_clip_state(cell->draw, clip); +} + + + +/* Called when driver state tracker notices changes to the viewport + * matrix: + */ +void cell_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct cell_context *cell = cell_context(pipe); + + cell->viewport = *viewport; /* struct copy */ + cell->dirty |= CELL_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + draw_set_viewport_state(cell->draw, viewport); + + /* Using tnl/ and vf/ modules is temporary while getting started. + * Full pipe will have vertex shader, vertex fetch of its own. + */ +} + + +void cell_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct cell_context *cell = cell_context(pipe); + + memcpy( &cell->scissor, scissor, sizeof(*scissor) ); + cell->dirty |= CELL_NEW_SCISSOR; +} + + +void cell_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ + struct cell_context *cell = cell_context(pipe); + + memcpy( &cell->poly_stipple, stipple, sizeof(*stipple) ); + cell->dirty |= CELL_NEW_STIPPLE; +} diff --git a/src/gallium/drivers/cell/ppu/cell_state_derived.c b/src/gallium/drivers/cell/ppu/cell_state_derived.c new file mode 100644 index 0000000000..56daf5dfde --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_derived.c @@ -0,0 +1,192 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_vertex.h" +#include "cell_context.h" +#include "cell_batch.h" +#include "cell_state.h" +#include "cell_state_emit.h" + + +static int +find_vs_output(const struct pipe_shader_state *vs, + uint semantic_name, + uint semantic_index) +{ + uint i; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == semantic_name && + vs->output_semantic_index[i] == semantic_index) + return i; + } + return -1; +} + + +/** + * Determine how to map vertex program outputs to fragment program inputs. + * Basically, this will be used when computing the triangle interpolation + * coefficients from the post-transform vertex attributes. + */ +static void +calculate_vertex_layout( struct cell_context *cell ) +{ + const struct pipe_shader_state *vs = &cell->vs->shader; + const struct pipe_shader_state *fs = &cell->fs->shader; + const enum interp_mode colorInterp + = cell->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + struct vertex_info *vinfo = &cell->vertex_info; + uint i; + int src; + +#if 0 + if (cell->vbuf) { + /* if using the post-transform vertex buffer, tell draw_vbuf to + * simply emit the whole post-xform vertex as-is: + */ + struct vertex_info *vinfo_vbuf = &cell->vertex_info_vbuf; + vinfo_vbuf->num_attribs = 0; + draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); + vinfo_vbuf->size = 4 * vs->num_outputs + sizeof(struct vertex_header)/4; + } +#endif + + /* reset vinfo */ + vinfo->num_attribs = 0; + + /* we always want to emit vertex pos */ + src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); + + + /* + * Loop over fragment shader inputs, searching for the matching output + * from the vertex shader. + */ + for (i = 0; i < fs->num_inputs; i++) { + switch (fs->input_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + /* already done above */ + break; + + case TGSI_SEMANTIC_COLOR: + src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, + fs->input_semantic_index[i]); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); + break; + + case TGSI_SEMANTIC_FOG: + src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); +#if 1 + if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */ + src = 0; +#endif + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src); + break; + + case TGSI_SEMANTIC_GENERIC: + /* this includes texcoords and varying vars */ + src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, + fs->input_semantic_index[i]); + assert(src >= 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); + break; + + default: + assert(0); + } + } + + draw_compute_vertex_size(vinfo); + + /* XXX only signal this if format really changes */ + cell->dirty |= CELL_NEW_VERTEX_INFO; +} + + +#if 0 +/** + * Recompute cliprect from scissor bounds, scissor enable and surface size. + */ +static void +compute_cliprect(struct cell_context *sp) +{ + unsigned surfWidth, surfHeight; + + if (sp->framebuffer.num_cbufs > 0) { + surfWidth = sp->framebuffer.cbufs[0]->width; + surfHeight = sp->framebuffer.cbufs[0]->height; + } + else { + /* no surface? */ + surfWidth = sp->scissor.maxx; + surfHeight = sp->scissor.maxy; + } + + if (sp->rasterizer->scissor) { + /* clip to scissor rect */ + sp->cliprect.minx = MAX2(sp->scissor.minx, 0); + sp->cliprect.miny = MAX2(sp->scissor.miny, 0); + sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth); + sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight); + } + else { + /* clip to surface bounds */ + sp->cliprect.minx = 0; + sp->cliprect.miny = 0; + sp->cliprect.maxx = surfWidth; + sp->cliprect.maxy = surfHeight; + } +} +#endif + + + +void cell_update_derived( struct cell_context *cell ) +{ + if (cell->dirty & (CELL_NEW_RASTERIZER | + CELL_NEW_FS | + CELL_NEW_VS)) + calculate_vertex_layout( cell ); + +#if 0 + if (cell->dirty & (CELL_NEW_SCISSOR | + CELL_NEW_DEPTH_STENCIL_ALPHA | + CELL_NEW_FRAMEBUFFER)) + compute_cliprect(cell); +#endif + + cell_emit_state(cell); + + cell->dirty = 0; +} diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c new file mode 100644 index 0000000000..5d2a786449 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -0,0 +1,103 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "cell_context.h" +#include "cell_state.h" +#include "cell_state_emit.h" +#include "cell_batch.h" +#include "cell_texture.h" + + +static void +emit_state_cmd(struct cell_context *cell, uint cmd, + const void *state, uint state_size) +{ + uint64_t *dst = (uint64_t *) + cell_batch_alloc(cell, ROUNDUP8(sizeof(uint64_t) + state_size)); + *dst = cmd; + memcpy(dst + 1, state, state_size); +} + + + +void +cell_emit_state(struct cell_context *cell) +{ + if (cell->dirty & CELL_NEW_FRAMEBUFFER) { + struct pipe_surface *cbuf = cell->framebuffer.cbufs[0]; + struct pipe_surface *zbuf = cell->framebuffer.zsbuf; + struct cell_command_framebuffer *fb + = cell_batch_alloc(cell, sizeof(*fb)); + fb->opcode = CELL_CMD_STATE_FRAMEBUFFER; + fb->color_start = cell->cbuf_map[0]; + fb->color_format = cbuf->format; + fb->depth_start = cell->zsbuf_map; + fb->depth_format = zbuf ? zbuf->format : PIPE_FORMAT_NONE; + fb->width = cell->framebuffer.cbufs[0]->width; + fb->height = cell->framebuffer.cbufs[0]->height; + } + + if (cell->dirty & CELL_NEW_BLEND) { + emit_state_cmd(cell, CELL_CMD_STATE_BLEND, + cell->blend, + sizeof(struct pipe_blend_state)); + } + + if (cell->dirty & CELL_NEW_DEPTH_STENCIL) { + emit_state_cmd(cell, CELL_CMD_STATE_DEPTH_STENCIL, + cell->depth_stencil, + sizeof(struct pipe_depth_stencil_alpha_state)); + } + + if (cell->dirty & CELL_NEW_SAMPLER) { + emit_state_cmd(cell, CELL_CMD_STATE_SAMPLER, + cell->sampler[0], sizeof(struct pipe_sampler_state)); + } + + if (cell->dirty & CELL_NEW_TEXTURE) { + struct cell_command_texture texture; + if (cell->texture[0]) { + texture.start = cell->texture[0]->tiled_data; + texture.width = cell->texture[0]->base.width[0]; + texture.height = cell->texture[0]->base.height[0]; + } + else { + texture.start = NULL; + texture.width = 0; + texture.height = 0; + } + + emit_state_cmd(cell, CELL_CMD_STATE_TEXTURE, + &texture, sizeof(struct cell_command_texture)); + } + + if (cell->dirty & CELL_NEW_VERTEX_INFO) { + emit_state_cmd(cell, CELL_CMD_STATE_VERTEX_INFO, + &cell->vertex_info, sizeof(struct vertex_info)); + } +} diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.h b/src/gallium/drivers/cell/ppu/cell_state_emit.h new file mode 100644 index 0000000000..59f8affe8d --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.h @@ -0,0 +1,36 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_STATE_EMIT_H +#define CELL_STATE_EMIT_H + + +extern void +cell_emit_state(struct cell_context *cell); + + +#endif /* CELL_STATE_EMIT_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_state_fs.c b/src/gallium/drivers/cell/ppu/cell_state_fs.c new file mode 100644 index 0000000000..3f46a87d18 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_fs.c @@ -0,0 +1,171 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "pipe/draw/draw_context.h" +#if 0 +#include "pipe/p_shader_tokens.h" +#include "pipe/llvm/gallivm.h" +#include "pipe/tgsi/util/tgsi_dump.h" +#include "pipe/tgsi/exec/tgsi_sse2.h" +#endif + +#include "cell_context.h" +#include "cell_state.h" + + +void * +cell_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + /*struct cell_context *cell = cell_context(pipe);*/ + struct cell_fragment_shader_state *state; + + state = CALLOC_STRUCT(cell_fragment_shader_state); + if (!state) + return NULL; + + state->shader = *templ; + +#if 0 + if (cell->dump_fs) { + tgsi_dump(state->shader.tokens, 0); + } + +#if defined(__i386__) || defined(__386__) + if (cell->use_sse) { + x86_init_func( &state->sse2_program ); + tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); + } +#endif + +#ifdef MESA_LLVM + state->llvm_prog = 0; + if (!gallivm_global_cpu_engine()) { + gallivm_cpu_engine_create(state->llvm_prog); + } + else + gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); +#endif +#endif + + return state; +} + + +void +cell_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct cell_context *cell = cell_context(pipe); + + cell->fs = (struct cell_fragment_shader_state *) fs; + + cell->dirty |= CELL_NEW_FS; +} + + +void +cell_delete_fs_state(struct pipe_context *pipe, void *fs) +{ + struct cell_fragment_shader_state *state = + (struct cell_fragment_shader_state *) fs; + + FREE( state ); +} + + +void * +cell_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct cell_context *cell = cell_context(pipe); + struct cell_vertex_shader_state *state; + + state = CALLOC_STRUCT(cell_vertex_shader_state); + if (!state) + return NULL; + + state->shader = *templ; + + state->draw_data = draw_create_vertex_shader(cell->draw, &state->shader); + if (state->draw_data == NULL) { + FREE( state ); + return NULL; + } + + return state; +} + + +void +cell_bind_vs_state(struct pipe_context *pipe, void *vs) +{ + struct cell_context *cell = cell_context(pipe); + + cell->vs = (const struct cell_vertex_shader_state *) vs; + + draw_bind_vertex_shader(cell->draw, cell->vs->draw_data); + + cell->dirty |= CELL_NEW_VS; +} + + +void +cell_delete_vs_state(struct pipe_context *pipe, void *vs) +{ + struct cell_context *cell = cell_context(pipe); + + struct cell_vertex_shader_state *state = + (struct cell_vertex_shader_state *) vs; + + draw_delete_vertex_shader(cell->draw, state->draw_data); + FREE( state ); +} + + +void +cell_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct cell_context *cell = cell_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + /* note: reference counting */ + pipe_buffer_reference(ws, + &cell->constants[shader].buffer, + buf->buffer); + cell->constants[shader].size = buf->size; + + cell->dirty |= CELL_NEW_CONSTANTS; +} diff --git a/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c b/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c new file mode 100644 index 0000000000..d8128ece54 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c @@ -0,0 +1,106 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/draw/draw_context.h" +#include "cell_context.h" +#include "cell_state.h" + + + +struct spu_rasterizer_state +{ + unsigned flatshade:1; +#if 0 + unsigned light_twoside:1; + unsigned front_winding:2; /**< PIPE_WINDING_x */ + unsigned cull_mode:2; /**< PIPE_WINDING_x */ + unsigned fill_cw:2; /**< PIPE_POLYGON_MODE_x */ + unsigned fill_ccw:2; /**< PIPE_POLYGON_MODE_x */ + unsigned offset_cw:1; + unsigned offset_ccw:1; +#endif + unsigned scissor:1; + unsigned poly_smooth:1; + unsigned poly_stipple_enable:1; + unsigned point_smooth:1; +#if 0 + unsigned point_sprite:1; + unsigned point_size_per_vertex:1; /**< size computed in vertex shader */ +#endif + unsigned multisample:1; /* XXX maybe more ms state in future */ + unsigned line_smooth:1; + unsigned line_stipple_enable:1; + unsigned line_stipple_factor:8; /**< [1..256] actually */ + unsigned line_stipple_pattern:16; +#if 0 + unsigned bypass_clipping:1; +#endif + unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ + + float line_width; + float point_size; /**< used when no per-vertex size */ +#if 0 + float offset_units; + float offset_scale; + ubyte sprite_coord_mode[PIPE_MAX_SHADER_OUTPUTS]; /**< PIPE_SPRITE_COORD_ */ +#endif +}; + + + +void * +cell_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup) +{ + struct pipe_rasterizer_state *state + = MALLOC(sizeof(struct pipe_rasterizer_state)); + memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); + return state; +} + + +void +cell_bind_rasterizer_state(struct pipe_context *pipe, void *setup) +{ + struct cell_context *cell = cell_context(pipe); + + /* pass-through to draw module */ + draw_set_rasterizer_state(cell->draw, setup); + + cell->rasterizer = (struct pipe_rasterizer_state *)setup; + + cell->dirty |= CELL_NEW_RASTERIZER; +} + + +void +cell_delete_rasterizer_state(struct pipe_context *pipe, void *rasterizer) +{ + FREE(rasterizer); +} diff --git a/src/gallium/drivers/cell/ppu/cell_state_sampler.c b/src/gallium/drivers/cell/ppu/cell_state_sampler.c new file mode 100644 index 0000000000..ade6cc8338 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_sampler.c @@ -0,0 +1,84 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "pipe/draw/draw_context.h" +#include "cell_context.h" +#include "cell_state.h" +#include "cell_texture.h" + + +void * +cell_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + return mem_dup(sampler, sizeof(*sampler)); +} + +void +cell_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + assert(unit < PIPE_MAX_SAMPLERS); + cell->sampler[unit] = (struct pipe_sampler_state *)sampler; + + cell->dirty |= CELL_NEW_SAMPLER; +} + + +void +cell_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + FREE( sampler ); +} + + + +void +cell_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, + struct pipe_texture *texture) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + cell->texture[sampler] = texture; + + cell_update_texture_mapping(cell); + + cell->dirty |= CELL_NEW_TEXTURE; +} diff --git a/src/gallium/drivers/cell/ppu/cell_state_surface.c b/src/gallium/drivers/cell/ppu/cell_state_surface.c new file mode 100644 index 0000000000..287610b76b --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_surface.c @@ -0,0 +1,71 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_inlines.h" +#include "cell_context.h" +#include "cell_state.h" + + +void +cell_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct cell_context *cell = cell_context(pipe); + + if (1 /*memcmp(&cell->framebuffer, fb, sizeof(*fb))*/) { + struct pipe_surface *csurf = fb->cbufs[0]; + struct pipe_surface *zsurf = fb->zsbuf; + uint i; + + /* unmap old surfaces */ + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + if (cell->framebuffer.cbufs[i] && cell->cbuf_map[i]) { + pipe_surface_unmap(cell->framebuffer.cbufs[i]); + cell->cbuf_map[i] = NULL; + } + } + + if (cell->framebuffer.zsbuf && cell->zsbuf_map) { + pipe_surface_unmap(cell->framebuffer.zsbuf); + cell->zsbuf_map = NULL; + } + + /* update my state */ + cell->framebuffer = *fb; + + /* map new surfaces */ + if (csurf) + cell->cbuf_map[0] = pipe_surface_map(csurf); + + if (zsurf) + cell->zsbuf_map = pipe_surface_map(zsurf); + + cell->dirty |= CELL_NEW_FRAMEBUFFER; + } +} + diff --git a/src/gallium/drivers/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c new file mode 100644 index 0000000000..0f01e920f9 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + + +#include "cell_context.h" +#include "cell_state.h" + +#include "pipe/draw/draw_context.h" + + +void +cell_set_vertex_element(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_element *attrib) +{ + struct cell_context *cell = cell_context(pipe); + assert(index < PIPE_ATTRIB_MAX); + cell->vertex_element[index] = *attrib; /* struct copy */ + cell->dirty |= CELL_NEW_VERTEX; + + draw_set_vertex_element(cell->draw, index, attrib); +} + + +void +cell_set_vertex_buffer(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_buffer *buffer) +{ + struct cell_context *cell = cell_context(pipe); + assert(index < PIPE_ATTRIB_MAX); + cell->vertex_buffer[index] = *buffer; /* struct copy */ + cell->dirty |= CELL_NEW_VERTEX; + + draw_set_vertex_buffer(cell->draw, index, buffer); +} diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c new file mode 100644 index 0000000000..fca93e4742 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -0,0 +1,179 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "pipe/util/p_tile.h" +#include "cell_context.h" +#include "cell_surface.h" + + +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +static void +cell_surface_data(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + const void *src, unsigned src_pitch, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + pipe_surface_unmap(dst); +} + + +static void +cell_surface_copy(struct pipe_context *pipe, + unsigned do_flip, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + assert( dst->cpp == src->cpp ); + + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe_surface_map(src), + do_flip ? -src->pitch : src->pitch, + srcx, do_flip ? 1 - srcy - height : srcy); + + pipe_surface_unmap(src); + pipe_surface_unmap(dst); +} + + +static void * +get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) +{ + return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; +} + + +#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8)) + + +/** + * Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +static void +cell_surface_fill(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) +{ + unsigned i, j; + void *dst_map = pipe_surface_map(dst); + + assert(dst->pitch > 0); + assert(width <= dst->pitch); + + switch (dst->cpp) { + case 1: + { + ubyte *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: + { + ushort *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (ushort) value; + row += dst->pitch; + } + } + break; + case 4: + { + unsigned *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + case 8: + { + /* expand the 4-byte clear value to an 8-byte value */ + ushort *row = (ushort *) get_pointer(dst, dst_map, dstx, dsty); + ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); + ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); + ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); + ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff); + val0 = (val0 << 8) | val0; + val1 = (val1 << 8) | val1; + val2 = (val2 << 8) | val2; + val3 = (val3 << 8) | val3; + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) { + row[j*4+0] = val0; + row[j*4+1] = val1; + row[j*4+2] = val2; + row[j*4+3] = val3; + } + row += dst->pitch * 4; + } + } + break; + default: + assert(0); + break; + } + + pipe_surface_unmap( dst ); +} + + +void +cell_init_surface_functions(struct cell_context *cell) +{ + cell->pipe.surface_copy = cell_surface_copy; + cell->pipe.surface_fill = cell_surface_fill; +} diff --git a/src/gallium/drivers/cell/ppu/cell_surface.h b/src/gallium/drivers/cell/ppu/cell_surface.h new file mode 100644 index 0000000000..9e58f32944 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_surface.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef CELL_SURFACE_H +#define CELL_SURFACE_H + + +struct cell_context; + + +extern void +cell_init_surface_functions(struct cell_context *cell); + + +#endif /* SP_SURFACE_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c new file mode 100644 index 0000000000..c8ef36002f --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -0,0 +1,252 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "cell_context.h" +#include "cell_state.h" +#include "cell_texture.h" + + +/* Simple, maximally packed layout. + */ + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + + +static void +cell_texture_layout(struct cell_texture * spt) +{ + struct pipe_texture *pt = &spt->base; + unsigned level; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + + spt->buffer_size = 0; + + for ( level = 0 ; level <= pt->last_level ; level++ ) { + pt->width[level] = width; + pt->height[level] = height; + pt->depth[level] = depth; + + spt->level_offset[level] = spt->buffer_size; + + spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * + ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * + width * pt->cpp; + + width = minify(width); + height = minify(height); + depth = minify(depth); + } +} + + +struct pipe_texture * +cell_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) +{ + struct cell_texture *spt = CALLOC_STRUCT(cell_texture); + if (!spt) + return NULL; + + spt->base = *templat; + + cell_texture_layout(spt); + + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); + + if (!spt->buffer) { + FREE(spt); + return NULL; + } + + return &spt->base; +} + + +void +cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct cell_texture *spt = cell_texture(*pt); + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); + */ + + pipe_buffer_reference(pipe->winsys, &spt->buffer, NULL); + + FREE(spt); + } + *pt = NULL; +} + + +/** + * Called via pipe->get_tex_surface() + */ +struct pipe_surface * +cell_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct cell_texture *spt = cell_texture(pt); + struct pipe_surface *ps; + + ps = pipe->winsys->surface_alloc(pipe->winsys); + if (ps) { + assert(ps->refcount); + assert(ps->winsys); + pipe_buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = ps->width; + ps->offset = spt->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { + ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * + (pt->compressed ? ps->height/4 : ps->height) * + ps->width * ps->cpp; + } else { + assert(face == 0); + assert(zslice == 0); + } + } + return ps; +} + + + +static void +tile_copy_data(uint w, uint h, uint tile_size, uint *dst, const uint *src) +{ + const uint tile_size2 = tile_size * tile_size; + const uint h_t = h / tile_size, w_t = w / tile_size; + + uint it, jt; /* tile counters */ + uint i, j; /* intra-tile counters */ + + for (it = 0; it < h_t; it++) { + for (jt = 0; jt < w_t; jt++) { + /* fill in tile (i, j) */ + uint *tdst = dst + (it * w_t + jt) * tile_size2; + for (i = 0; i < tile_size; i++) { + for (j = 0; j < tile_size; j++) { + const uint srci = it * tile_size + i; + const uint srcj = jt * tile_size + j; + *tdst++ = src[srci * h + srcj]; + } + } + } + } +} + + + +/** + * Convert linear texture image data to tiled format for SPU usage. + */ +static void +cell_tile_texture(struct cell_context *cell, + struct cell_texture *texture) +{ + uint face = 0, level = 0, zslice = 0; + struct pipe_surface *surf; + const uint w = texture->base.width[0], h = texture->base.height[0]; + const uint *src; + + /* temporary restrictions: */ + assert(w >= TILE_SIZE); + assert(h >= TILE_SIZE); + assert(w % TILE_SIZE == 0); + assert(h % TILE_SIZE == 0); + + surf = cell_get_tex_surface(&cell->pipe, &texture->base, face, level, zslice); + ASSERT(surf); + + src = (const uint *) pipe_surface_map(surf); + + if (texture->tiled_data) { + align_free(texture->tiled_data); + } + texture->tiled_data = align_malloc(w * h * 4, 16); + + tile_copy_data(w, h, TILE_SIZE, texture->tiled_data, src); + + pipe_surface_unmap(surf); + + pipe_surface_reference(&surf, NULL); +} + + + +void +cell_update_texture_mapping(struct cell_context *cell) +{ + uint face = 0, level = 0, zslice = 0; + + if (cell->texture[0]) + cell_tile_texture(cell, cell->texture[0]); +#if 0 + if (cell->tex_surf && cell->tex_map) { + pipe_surface_unmap(cell->tex_surf); + cell->tex_map = NULL; + } + + /* XXX free old surface */ + + cell->tex_surf = cell_get_tex_surface(&cell->pipe, + &cell->texture[0]->base, + face, level, zslice); + + cell->tex_map = pipe_surface_map(cell->tex_surf); +#endif +} diff --git a/src/gallium/drivers/cell/ppu/cell_texture.h b/src/gallium/drivers/cell/ppu/cell_texture.h new file mode 100644 index 0000000000..0264fed88e --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_texture.h @@ -0,0 +1,80 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_TEXTURE_H +#define CELL_TEXTURE_H + + +struct pipe_context; +struct pipe_texture; + + +/** + * Subclass of pipe_texture + */ +struct cell_texture +{ + struct pipe_texture base; + + unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; + unsigned long buffer_size; + + void *tiled_data; /* XXX this may be temporary */ /*ALIGN16*/ +}; + + +/** cast wrapper */ +static INLINE struct cell_texture * +cell_texture(struct pipe_texture *pt) +{ + return (struct cell_texture *) pt; +} + + + +extern struct pipe_texture * +cell_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat); + +extern void +cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); + +extern struct pipe_surface * +cell_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice); + + +extern void +cell_update_texture_mapping(struct cell_context *cell); + + +#endif /* CELL_TEXTURE */ diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.c b/src/gallium/drivers/cell/ppu/cell_vbuf.c new file mode 100644 index 0000000000..e9fafe492e --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_vbuf.c @@ -0,0 +1,294 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Authors + * Brian Paul + */ + + +#include "cell_batch.h" +#include "cell_context.h" +#include "cell_flush.h" +#include "cell_spu.h" +#include "cell_vbuf.h" +#include "pipe/draw/draw_vbuf.h" + + +/** Allow vertex data to be inlined after RENDER command */ +#define ALLOW_INLINE_VERTS 1 + + +/** + * Subclass of vbuf_render because we need a cell_context pointer in + * a few places. + */ +struct cell_vbuf_render +{ + struct vbuf_render base; + struct cell_context *cell; + uint prim; /**< PIPE_PRIM_x */ + uint vertex_size; /**< in bytes */ + void *vertex_buffer; /**< just for debug, really */ + uint vertex_buf; /**< in [0, CELL_NUM_BUFFERS-1] */ +}; + + +/** cast wrapper */ +static struct cell_vbuf_render * +cell_vbuf_render(struct vbuf_render *vbr) +{ + return (struct cell_vbuf_render *) vbr; +} + + + +static const struct vertex_info * +cell_vbuf_get_vertex_info(struct vbuf_render *vbr) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + return &cvbr->cell->vertex_info; +} + + +static void * +cell_vbuf_allocate_vertices(struct vbuf_render *vbr, + ushort vertex_size, ushort nr_vertices) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + /*printf("Alloc verts %u * %u\n", vertex_size, nr_vertices);*/ + + assert(cvbr->vertex_buf == ~0); + cvbr->vertex_buf = cell_get_empty_buffer(cvbr->cell); + cvbr->vertex_buffer = cvbr->cell->buffer[cvbr->vertex_buf]; + cvbr->vertex_size = vertex_size; + return cvbr->vertex_buffer; +} + + +static void +cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, + unsigned vertex_size, unsigned vertices_used) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + struct cell_context *cell = cvbr->cell; + + /* + printf("%s vertex_buf = %u count = %u\n", + __FUNCTION__, cvbr->vertex_buf, vertices_used); + */ + + /* Tell SPUs they can release the vert buf */ + if (cvbr->vertex_buf != ~0U) { + struct cell_command_release_verts *release + = (struct cell_command_release_verts *) + cell_batch_alloc(cell, sizeof(struct cell_command_release_verts)); + release->opcode = CELL_CMD_RELEASE_VERTS; + release->vertex_buf = cvbr->vertex_buf; + } + + cvbr->vertex_buf = ~0; + cell_flush_int(&cell->pipe, 0x0); + + assert(vertices == cvbr->vertex_buffer); + cvbr->vertex_buffer = NULL; +} + + + +static void +cell_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + cvbr->prim = prim; + /*printf("cell_set_prim %u\n", prim);*/ +} + + +static void +cell_vbuf_draw(struct vbuf_render *vbr, + const ushort *indices, + uint nr_indices) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + struct cell_context *cell = cvbr->cell; + float xmin, ymin, xmax, ymax; + uint i; + uint nr_vertices = 0, min_index = ~0; + const void *vertices = cvbr->vertex_buffer; + const uint vertex_size = cvbr->vertex_size; + + for (i = 0; i < nr_indices; i++) { + if (indices[i] > nr_vertices) + nr_vertices = indices[i]; + if (indices[i] < min_index) + min_index = indices[i]; + } + nr_vertices++; + +#if 0 + /*if (min_index > 0)*/ + printf("%s min_index = %u\n", __FUNCTION__, min_index); +#endif + +#if 0 + printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u\n", + nr_indices, nr_vertices); + printf(" "); + for (i = 0; i < nr_indices; i += 3) { + printf("%u %u %u, ", indices[i+0], indices[i+1], indices[i+2]); + } + printf("\n"); +#elif 0 + printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u indexes = [%u %u %u ...]\n", + nr_indices, nr_vertices, + indices[0], indices[1], indices[2]); + printf("ind space = %u, vert space = %u, space = %u\n", + nr_indices * 2, + nr_vertices * 4 * cell->vertex_info.size, + cell_batch_free_space(cell)); +#endif + + /* compute x/y bounding box */ + xmin = ymin = 1e50; + xmax = ymax = -1e50; + for (i = min_index; i < nr_vertices; i++) { + const float *v = (float *) ((ubyte *) vertices + i * vertex_size); + if (v[0] < xmin) + xmin = v[0]; + if (v[0] > xmax) + xmax = v[0]; + if (v[1] < ymin) + ymin = v[1]; + if (v[1] > ymax) + ymax = v[1]; + } +#if 0 + printf("PPU Bounds %g, %g .. %g, %g\n", xmin, ymin, xmax, ymax); + fflush(stdout); +#endif + + if (cvbr->prim != PIPE_PRIM_TRIANGLES) + return; /* only render tris for now */ + + /* build/insert batch RENDER command */ + { + const uint index_bytes = ROUNDUP8(nr_indices * 2); + const uint vertex_bytes = nr_vertices * 4 * cell->vertex_info.size; + const uint batch_size = sizeof(struct cell_command_render) + index_bytes; + + struct cell_command_render *render + = (struct cell_command_render *) + cell_batch_alloc(cell, batch_size); + + render->opcode = CELL_CMD_RENDER; + render->prim_type = cvbr->prim; + + render->num_indexes = nr_indices; + render->min_index = min_index; + + /* append indices after render command */ + memcpy(render + 1, indices, nr_indices * 2); + + /* if there's room, append vertices after the indices, else leave + * vertices in the original/separate buffer. + */ + render->vertex_size = 4 * cell->vertex_info.size; + render->num_verts = nr_vertices; + if (ALLOW_INLINE_VERTS && + min_index == 0 && + vertex_bytes + 16 <= cell_batch_free_space(cell)) { + /* vertex data inlined, after indices, at 16-byte boundary */ + void *dst = cell_batch_alloc_aligned(cell, vertex_bytes, 16); + memcpy(dst, vertices, vertex_bytes); + render->inline_verts = TRUE; + render->vertex_buf = ~0; + } + else { + /* vertex data in separate buffer */ + render->inline_verts = FALSE; + ASSERT(cvbr->vertex_buf >= 0); + render->vertex_buf = cvbr->vertex_buf; + } + + render->xmin = xmin; + render->ymin = ymin; + render->xmax = xmax; + render->ymax = ymax; + } + +#if 0 + /* helpful for debug */ + cell_flush_int(&cell->pipe, PIPE_FLUSH_WAIT); +#endif +} + + +static void +cell_vbuf_destroy(struct vbuf_render *vbr) +{ + struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); + cvbr->cell->vbuf_render = NULL; + FREE(cvbr); +} + + +/** + * Initialize the post-transform vertex buffer information for the given + * context. + */ +void +cell_init_vbuf(struct cell_context *cell) +{ + assert(cell->draw); + + cell->vbuf_render = CALLOC_STRUCT(cell_vbuf_render); + + /* The max number of indexes is what can fix into a batch buffer, + * minus the render and release-verts commands. + */ + cell->vbuf_render->base.max_indices + = (CELL_BUFFER_SIZE + - sizeof(struct cell_command_render) + - sizeof(struct cell_command_release_verts)) + / sizeof(ushort); + cell->vbuf_render->base.max_vertex_buffer_bytes = CELL_BUFFER_SIZE; + + cell->vbuf_render->base.get_vertex_info = cell_vbuf_get_vertex_info; + cell->vbuf_render->base.allocate_vertices = cell_vbuf_allocate_vertices; + cell->vbuf_render->base.set_primitive = cell_vbuf_set_primitive; + cell->vbuf_render->base.draw = cell_vbuf_draw; + cell->vbuf_render->base.release_vertices = cell_vbuf_release_vertices; + cell->vbuf_render->base.destroy = cell_vbuf_destroy; + + cell->vbuf_render->cell = cell; +#if 1 + cell->vbuf_render->vertex_buf = ~0; +#endif + + cell->vbuf = draw_vbuf_stage(cell->draw, &cell->vbuf_render->base); +} diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.h b/src/gallium/drivers/cell/ppu/cell_vbuf.h new file mode 100644 index 0000000000..d265cbf770 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_vbuf.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef CELL_VBUF_H +#define CELL_VBUF_H + + +struct cell_context; + +extern void +cell_init_vbuf(struct cell_context *cell); + + +#endif /* CELL_VBUF_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c new file mode 100644 index 0000000000..80dd500b34 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -0,0 +1,120 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file cell_vertex_shader.c + * Vertex shader interface routines for Cell. + * + * \author Ian Romanick + */ + +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +#include "cell_context.h" +#include "cell_draw_arrays.h" +#include "cell_spu.h" +#include "cell_batch.h" + +#include "pipe/cell/common.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_private.h" + +/** + * Run the vertex shader on all vertices in the vertex queue. + * Called by the draw module when the vertx cache needs to be flushed. + */ +void +cell_vertex_shader_queue_flush(struct draw_context *draw) +{ + struct cell_context *const cell = + (struct cell_context *) draw->driver_private; + struct cell_command_vs *const vs = &cell_global.command[0].vs; + uint64_t *batch; + struct cell_array_info *array_info; + unsigned i, j; + + assert(draw->vs.queue_nr != 0); + + /* XXX: do this on statechange: + */ + draw_update_vertex_fetch(draw); + + for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { + batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*array_info)); + + batch[0] = CELL_CMD_STATE_VS_ARRAY_INFO; + + array_info = (struct cell_array_info *) &batch[1]; + assert(draw->vertex_fetch.src_ptr[i] != NULL); + array_info->base = (uintptr_t) draw->vertex_fetch.src_ptr[i]; + array_info->attr = i; + array_info->pitch = draw->vertex_fetch.pitch[i]; + array_info->format = draw->vertex_element[i].src_format; + } + + batch = cell_batch_alloc(cell, sizeof(batch[0]) + + sizeof(struct pipe_viewport_state)); + batch[0] = CELL_CMD_STATE_VIEWPORT; + (void) memcpy(&batch[1], &draw->viewport, + sizeof(struct pipe_viewport_state)); + + cell_batch_flush(cell); + + vs->opcode = CELL_CMD_VS_EXECUTE; + vs->shader.num_outputs = draw->num_vs_outputs; + vs->shader.declarations = (uintptr_t) draw->machine.Declarations; + vs->shader.num_declarations = draw->machine.NumDeclarations; + vs->shader.instructions = (uintptr_t) draw->machine.Instructions; + vs->shader.num_instructions = draw->machine.NumInstructions; + vs->shader.uniforms = (uintptr_t) draw->user.constants; + vs->shader.immediates = (uintptr_t) draw->machine.Imms; + vs->shader.num_immediates = draw->machine.ImmLimit / 4; + vs->nr_attrs = draw->vertex_fetch.nr_attrs; + + (void) memcpy(vs->plane, draw->plane, sizeof(draw->plane)); + vs->nr_planes = draw->nr_planes; + + for (i = 0; i < draw->vs.queue_nr; i += SPU_VERTS_PER_BATCH) { + const unsigned n = MIN2(SPU_VERTS_PER_BATCH, draw->vs.queue_nr - i); + + for (j = 0; j < n; j++) { + vs->elts[j] = draw->vs.queue[i + j].elt; + vs->vOut[j] = (uintptr_t) draw->vs.queue[i + j].dest; + } + + for (/* empty */; j < SPU_VERTS_PER_BATCH; j++) { + vs->elts[j] = vs->elts[0]; + vs->vOut[j] = vs->vOut[0]; + } + + vs->num_elts = n; + send_mbox_message(cell_global.spe_contexts[0], CELL_CMD_VS_EXECUTE); + + cell_flush_int(& cell->pipe, PIPE_FLUSH_WAIT); + } + + draw->vs.queue_nr = 0; +} diff --git a/src/gallium/drivers/cell/ppu/cell_winsys.c b/src/gallium/drivers/cell/ppu/cell_winsys.c new file mode 100644 index 0000000000..ebabce3c8f --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_winsys.c @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "cell_winsys.h" + + +struct cell_winsys * +cell_get_winsys(uint format) +{ + struct cell_winsys *cws = CALLOC_STRUCT(cell_winsys); + if (cws) + cws->preferredFormat = format; + return cws; +} diff --git a/src/gallium/drivers/cell/ppu/cell_winsys.h b/src/gallium/drivers/cell/ppu/cell_winsys.h new file mode 100644 index 0000000000..ae2af5696b --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_winsys.h @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_WINSYS_H +#define CELL_WINSYS_H + +#include "pipe/p_compiler.h" + + +/** + * Very simple winsys at this time. + * Will probably eventually add SPU control info. + */ +struct cell_winsys +{ + uint preferredFormat; +}; + + +extern struct cell_winsys * +cell_get_winsys(uint format); + + + +#endif diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile new file mode 100644 index 0000000000..f202971d73 --- /dev/null +++ b/src/gallium/drivers/cell/spu/Makefile @@ -0,0 +1,72 @@ +# Gallium3D Cell driver: SPU code + +# This makefile builds the g3d_spu.a file that's linked into the +# PPU code/library. + + +TOP = ../../../../.. +include $(TOP)/configs/linux-cell + + +PROG = g3d + +PROG_SPU = $(PROG)_spu +PROG_SPU_A = $(PROG)_spu.a +PROG_SPU_EMBED_O = $(PROG)_spu-embed.o + + +SOURCES = \ + spu_main.c \ + spu_blend.c \ + spu_render.c \ + spu_texture.c \ + spu_tile.c \ + spu_tri.c \ + spu_exec.c \ + spu_util.c \ + spu_vertex_fetch.c \ + spu_vertex_shader.c + +SPU_OBJECTS = $(SOURCES:.c=.o) \ + +SPU_ASM_OUT = $(SOURCES:.c=.s) \ + +INCLUDE_DIRS = -I$(TOP)/src/mesa + + +.c.o: + $(SPU_CC) $(SPU_CFLAGS) -c $< + +.c.s: + $(SPU_CC) $(SPU_CFLAGS) -S $< + + +# The .a file will be linked into the main/PPU executable +default: $(PROG_SPU_A) + +$(PROG_SPU_A): $(PROG_SPU_EMBED_O) + $(SPU_AR) $(SPU_AR_FLAGS) $(PROG_SPU_A) $(PROG_SPU_EMBED_O) + +$(PROG_SPU_EMBED_O): $(PROG_SPU) + $(SPU_EMBED) $(SPU_EMBED_FLAGS) $(PROG_SPU) $(PROG_SPU) $(PROG_SPU_EMBED_O) + +$(PROG_SPU): $(SPU_OBJECTS) + $(SPU_CC) -o $(PROG_SPU) $(SPU_OBJECTS) $(SPU_LFLAGS) + + + +asmfiles: $(SPU_ASM_OUT) + + +clean: + rm -f *~ *.o *.a *.d *.s $(PROG_SPU) + + + +depend: $(SOURCES) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(SOURCES) 2> /dev/null + +include depend + diff --git a/src/gallium/drivers/cell/spu/spu_blend.c b/src/gallium/drivers/cell/spu/spu_blend.c new file mode 100644 index 0000000000..23ec0eeb45 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_blend.c @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "spu_main.h" +#include "spu_blend.h" +#include "spu_colorpack.h" + + +void +blend_quad(uint itx, uint ity, vector float colors[4]) +{ + /* simple SRC_ALPHA, ONE_MINUS_SRC_ALPHA blending */ + vector float fbc00 = spu_unpack_color(spu.ctile.ui[ity][itx]); + vector float fbc01 = spu_unpack_color(spu.ctile.ui[ity][itx+1]); + vector float fbc10 = spu_unpack_color(spu.ctile.ui[ity+1][itx]); + vector float fbc11 = spu_unpack_color(spu.ctile.ui[ity+1][itx+1]); + + vector float alpha00 = spu_splats(spu_extract(colors[0], 3)); + vector float alpha01 = spu_splats(spu_extract(colors[1], 3)); + vector float alpha10 = spu_splats(spu_extract(colors[2], 3)); + vector float alpha11 = spu_splats(spu_extract(colors[3], 3)); + + vector float one_minus_alpha00 = spu_sub(spu_splats(1.0f), alpha00); + vector float one_minus_alpha01 = spu_sub(spu_splats(1.0f), alpha01); + vector float one_minus_alpha10 = spu_sub(spu_splats(1.0f), alpha10); + vector float one_minus_alpha11 = spu_sub(spu_splats(1.0f), alpha11); + + colors[0] = spu_add(spu_mul(colors[0], alpha00), + spu_mul(fbc00, one_minus_alpha00)); + colors[1] = spu_add(spu_mul(colors[1], alpha01), + spu_mul(fbc01, one_minus_alpha01)); + colors[2] = spu_add(spu_mul(colors[2], alpha10), + spu_mul(fbc10, one_minus_alpha10)); + colors[3] = spu_add(spu_mul(colors[3], alpha11), + spu_mul(fbc11, one_minus_alpha11)); +} + diff --git a/src/gallium/drivers/cell/spu/spu_blend.h b/src/gallium/drivers/cell/spu/spu_blend.h new file mode 100644 index 0000000000..2b594b578b --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_blend.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SPU_BLEND_H +#define SPU_BLEND_H + + +extern void +blend_quad(uint itx, uint ity, vector float colors[4]); + + +#endif /* SPU_BLEND_H */ diff --git a/src/gallium/drivers/cell/spu/spu_colorpack.h b/src/gallium/drivers/cell/spu/spu_colorpack.h new file mode 100644 index 0000000000..e9fee8a3a6 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_colorpack.h @@ -0,0 +1,110 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + + +#ifndef SPU_COLORPACK_H +#define SPU_COLORPACK_H + + +#include + + +static INLINE unsigned int +spu_pack_R8G8B8A8(vector float rgba) +{ + vector unsigned int out = spu_convtu(rgba, 32); + + out = spu_shuffle(out, out, ((vector unsigned char) { + 0, 4, 8, 12, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0 }) ); + + return spu_extract(out, 0); +} + + +static INLINE unsigned int +spu_pack_A8R8G8B8(vector float rgba) +{ + vector unsigned int out = spu_convtu(rgba, 32); + out = spu_shuffle(out, out, ((vector unsigned char) { + 12, 0, 4, 8, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0}) ); + return spu_extract(out, 0); +} + + +static INLINE unsigned int +spu_pack_B8G8R8A8(vector float rgba) +{ + vector unsigned int out = spu_convtu(rgba, 32); + out = spu_shuffle(out, out, ((vector unsigned char) { + 8, 4, 0, 12, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0}) ); + return spu_extract(out, 0); +} + + +static INLINE unsigned int +spu_pack_color_shuffle(vector float rgba, vector unsigned char shuffle) +{ + vector unsigned int out = spu_convtu(rgba, 32); + out = spu_shuffle(out, out, shuffle); + return spu_extract(out, 0); +} + + +static INLINE vector float +spu_unpack_color(uint color) +{ + vector unsigned int color_u4 = spu_splats(color); + color_u4 = spu_shuffle(color_u4, color_u4, + ((vector unsigned char) { + 0, 0, 0, 0, + 5, 5, 5, 5, + 10, 10, 10, 10, + 15, 15, 15, 15}) ); + return spu_convtf(color_u4, 32); +} + + +static INLINE vector float +spu_unpack_A8R8G8B8(uint color) +{ + vector unsigned int color_u4 = spu_splats(color); + color_u4 = spu_shuffle(color_u4, color_u4, + ((vector unsigned char) { + 5, 5, 5, 5, + 10, 10, 10, 10, + 15, 15, 15, 15, + 0, 0, 0, 0}) ); + + return spu_convtf(color_u4, 32); +} + + +#endif /* SPU_COLORPACK_H */ diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c new file mode 100644 index 0000000000..e51008b9b3 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -0,0 +1,1948 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI interpretor/executor. + * + * Flow control information: + * + * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) + * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special + * care since a condition may be true for some quad components but false + * for other components. + * + * We basically execute all statements (even if they're in the part of + * an IF/ELSE clause that's "not taken") and use a special mask to + * control writing to destination registers. This is the ExecMask. + * See store_dest(). + * + * The ExecMask is computed from three other masks (CondMask, LoopMask and + * ContMask) which are controlled by the flow control instructions (namely: + * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). + * + * + * Authors: + * Michal Krol + * Brian Paul + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +#include "pipe/tgsi/util/tgsi_util.h" +#include "spu_exec.h" +#include "spu_main.h" +#include "spu_vertex_shader.h" + +#define TILE_TOP_LEFT 0 +#define TILE_TOP_RIGHT 1 +#define TILE_BOTTOM_LEFT 2 +#define TILE_BOTTOM_RIGHT 3 + +/* + * Shorthand locations of various utility registers (_I = Index, _C = Channel) + */ +#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I +#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C +#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I +#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C +#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I +#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C +#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I +#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C +#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I +#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C +#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I +#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C +#define TEMP_128_I TGSI_EXEC_TEMP_128_I +#define TEMP_128_C TGSI_EXEC_TEMP_128_C +#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I +#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C +#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I +#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C +#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I +#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C +#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I +#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +#define FOR_EACH_CHANNEL(CHAN)\ + for (CHAN = 0; CHAN < 4; CHAN++) + +#define IS_CHANNEL_ENABLED(INST, CHAN)\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IS_CHANNEL_ENABLED2(INST, CHAN)\ + ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) + +#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED2( INST, CHAN )) + + +/** The execution mask depends on the conditional mask and the loop mask */ +#define UPDATE_EXEC_MASK(MACH) \ + MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask + + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + + + +/** + * Initialize machine state by expanding tokens to full instructions, + * allocating temporary storage, setting up constants, etc. + * After this, we can call spu_exec_machine_run() many times. + */ +void +spu_exec_machine_init(struct spu_exec_machine *mach, + uint numSamplers, + struct spu_sampler *samplers, + unsigned processor) +{ + qword zero; + qword not_zero; + uint i; + + mach->Samplers = samplers; + mach->Processor = processor; + mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; + + zero = si_xor(zero, zero); + not_zero = si_xori(zero, 0xff); + + /* Setup constants. */ + mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q = zero; + mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].q = not_zero; + mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].q = si_shli(not_zero, -1); + mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].q = si_shli(not_zero, 31); + + mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q = (qword) spu_splats(1.0f); + mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q = (qword) spu_splats(2.0f); + mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].q = (qword) spu_splats(128.0f); + mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].q = (qword) spu_splats(-128.0f); +} + + +static INLINE qword +micro_abs(qword src) +{ + return si_rotmi(si_shli(src, 1), -1); +} + +static INLINE qword +micro_ceil(qword src) +{ + return (qword) _ceilf4((vec_float4) src); +} + +static INLINE qword +micro_cos(qword src) +{ + return (qword) _cosf4((vec_float4) src); +} + +static const qword br_shuf = { + TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, + TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, + TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, + TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, + TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, + TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, + TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, + TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, +}; + +static const qword bl_shuf = { + TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, + TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, + TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, + TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, + TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, + TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, + TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, + TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, +}; + +static const qword tl_shuf = { + TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, + TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, + TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, + TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, + TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, + TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, + TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, + TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, +}; + +static qword +micro_ddx(qword src) +{ + qword bottom_right = si_shufb(src, src, br_shuf); + qword bottom_left = si_shufb(src, src, bl_shuf); + + return si_fs(bottom_right, bottom_left); +} + +static qword +micro_ddy(qword src) +{ + qword top_left = si_shufb(src, src, tl_shuf); + qword bottom_left = si_shufb(src, src, bl_shuf); + + return si_fs(top_left, bottom_left); +} + +static INLINE qword +micro_div(qword src0, qword src1) +{ + return (qword) _divf4((vec_float4) src0, (vec_float4) src1); +} + +static qword +micro_flr(qword src) +{ + return (qword) _floorf4((vec_float4) src); +} + +static qword +micro_frc(qword src) +{ + return si_fs(src, (qword) _floorf4((vec_float4) src)); +} + +static INLINE qword +micro_ge(qword src0, qword src1) +{ + return si_or(si_fceq(src0, src1), si_fcgt(src0, src1)); +} + +static qword +micro_lg2(qword src) +{ + return (qword) _log2f4((vec_float4) src); +} + +static INLINE qword +micro_lt(qword src0, qword src1) +{ + const qword tmp = si_or(si_fceq(src0, src1), si_fcgt(src0, src1)); + + return si_xori(tmp, 0xff); +} + +static INLINE qword +micro_max(qword src0, qword src1) +{ + return si_selb(src1, src0, si_fcgt(src0, src1)); +} + +static INLINE qword +micro_min(qword src0, qword src1) +{ + return si_selb(src0, src1, si_fcgt(src0, src1)); +} + +static qword +micro_neg(qword src) +{ + return si_xor(src, (qword) spu_splats(0x80000000)); +} + +static qword +micro_set_sign(qword src) +{ + return si_or(src, (qword) spu_splats(0x80000000)); +} + +static qword +micro_pow(qword src0, qword src1) +{ + return (qword) _powf4((vec_float4) src0, (vec_float4) src1); +} + +static qword +micro_rnd(qword src) +{ + const qword half = (qword) spu_splats(0.5f); + + /* May be able to use _roundf4. There may be some difference, though. + */ + return (qword) _floorf4((vec_float4) si_fa(src, half)); +} + +static INLINE qword +micro_ishr(qword src0, qword src1) +{ + return si_rotma(src0, si_sfi(src1, 0)); +} + +static qword +micro_trunc(qword src) +{ + return (qword) _truncf4((vec_float4) src); +} + +static qword +micro_sin(qword src) +{ + return (qword) _sinf4((vec_float4) src); +} + +static INLINE qword +micro_sqrt(qword src) +{ + return (qword) _sqrtf4((vec_float4) src); +} + +static void +fetch_src_file_channel( + const struct spu_exec_machine *mach, + const uint file, + const uint swizzle, + const union spu_exec_channel *index, + union spu_exec_channel *chan ) +{ + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( file ) { + case TGSI_FILE_CONSTANT: { + unsigned char buffer[32] ALIGN16_ATTRIB; + unsigned i; + + for (i = 0; i < 4; i++) { + const float *ptr = mach->Consts[index->i[i]]; + const uint64_t addr = (uint64_t)(uintptr_t) ptr; + const unsigned size = ((addr & 0x0f) == 0) ? 16 : 32; + + mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); + wait_on_mask(1 << TAG_VERTEX_BUFFER); + + (void) memcpy(& chan->f[i], &buffer[(addr & 0x0f) + + (sizeof(float) * swizzle)], sizeof(float)); + } + break; + } + + case TGSI_FILE_INPUT: + chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_TEMPORARY: + chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_IMMEDIATE: + assert( index->i[0] < (int) mach->ImmLimit ); + assert( index->i[1] < (int) mach->ImmLimit ); + assert( index->i[2] < (int) mach->ImmLimit ); + assert( index->i[3] < (int) mach->ImmLimit ); + + chan->f[0] = mach->Imms[index->i[0]][swizzle]; + chan->f[1] = mach->Imms[index->i[1]][swizzle]; + chan->f[2] = mach->Imms[index->i[2]][swizzle]; + chan->f[3] = mach->Imms[index->i[3]][swizzle]; + break; + + case TGSI_FILE_ADDRESS: + chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_OUTPUT: + /* vertex/fragment output vars can be read too */ + chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; + break; + + case TGSI_EXTSWIZZLE_ONE: + *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; + break; + + default: + assert( 0 ); + } +} + +static void +fetch_source( + const struct spu_exec_machine *mach, + union spu_exec_channel *chan, + const struct tgsi_full_src_register *reg, + const uint chan_index ) +{ + union spu_exec_channel index; + uint swizzle; + + index.i[0] = + index.i[1] = + index.i[2] = + index.i[3] = reg->SrcRegister.Index; + + if (reg->SrcRegister.Indirect) { + union spu_exec_channel index2; + union spu_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle(®->SrcRegisterInd, + CHAN_X); + fetch_src_file_channel( + mach, + reg->SrcRegisterInd.File, + swizzle, + &index2, + &indir_index ); + + index.q = si_a(index.q, indir_index.q); + } + + if( reg->SrcRegister.Dimension ) { + switch( reg->SrcRegister.File ) { + case TGSI_FILE_INPUT: + index.q = si_mpyi(index.q, 17); + break; + case TGSI_FILE_CONSTANT: + index.q = si_shli(index.q, 12); + break; + default: + assert( 0 ); + } + + index.i[0] += reg->SrcRegisterDim.Index; + index.i[1] += reg->SrcRegisterDim.Index; + index.i[2] += reg->SrcRegisterDim.Index; + index.i[3] += reg->SrcRegisterDim.Index; + + if (reg->SrcRegisterDim.Indirect) { + union spu_exec_channel index2; + union spu_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterDimInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterDimInd.File, + swizzle, + &index2, + &indir_index ); + + index.q = si_a(index.q, indir_index.q); + } + } + + swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + fetch_src_file_channel( + mach, + reg->SrcRegister.File, + swizzle, + &index, + chan ); + + switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { + case TGSI_UTIL_SIGN_CLEAR: + chan->q = micro_abs(chan->q); + break; + + case TGSI_UTIL_SIGN_SET: + chan->q = micro_set_sign(chan->q); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + chan->q = micro_neg(chan->q); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } + + if (reg->SrcRegisterExtMod.Complement) { + chan->q = si_fs(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, chan->q); + } +} + +static void +store_dest( + struct spu_exec_machine *mach, + const union spu_exec_channel *chan, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + uint chan_index ) +{ + union spu_exec_channel *dst; + + switch( reg->DstRegister.File ) { + case TGSI_FILE_NULL: + return; + + case TGSI_FILE_OUTPUT: + dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + + reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_TEMPORARY: + dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_ADDRESS: + dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; + break; + + default: + assert( 0 ); + return; + } + + switch (inst->Instruction.Saturate) + { + case TGSI_SAT_NONE: + if (mach->ExecMask & 0x1) + dst->i[0] = chan->i[0]; + if (mach->ExecMask & 0x2) + dst->i[1] = chan->i[1]; + if (mach->ExecMask & 0x4) + dst->i[2] = chan->i[2]; + if (mach->ExecMask & 0x8) + dst->i[3] = chan->i[3]; + break; + + case TGSI_SAT_ZERO_ONE: + /* XXX need to obey ExecMask here */ + dst->q = micro_max(chan->q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); + dst->q = micro_min(dst->q, mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + + default: + assert( 0 ); + } +} + +#define FETCH(VAL,INDEX,CHAN)\ + fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) + +#define STORE(VAL,INDEX,CHAN)\ + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) + + +/** + * Execute ARB-style KIL which is predicated by a src register. + * Kill fragment if any of the four values is less than zero. + */ +static void +exec_kilp(struct spu_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + uint uniquemask; + uint chan_index; + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + union spu_exec_channel r[1]; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + uint swizzle; + uint i; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle ( + &inst->FullSrcRegisters[0], + chan_index); + + /* check if the component has not been already tested */ + if (uniquemask & (1 << swizzle)) + continue; + uniquemask |= 1 << swizzle; + + FETCH(&r[0], 0, chan_index); + for (i = 0; i < 4; i++) + if (r[0].f[i] < 0.0f) + kilmask |= 1 << i; + } + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; +} + + +/* + * Fetch a texel using STR texture coordinates. + */ +static void +fetch_texel( struct spu_sampler *sampler, + const union spu_exec_channel *s, + const union spu_exec_channel *t, + const union spu_exec_channel *p, + float lodbias, /* XXX should be float[4] */ + union spu_exec_channel *r, + union spu_exec_channel *g, + union spu_exec_channel *b, + union spu_exec_channel *a ) +{ + qword rgba[4]; + qword out[4]; + + sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, (float *) rgba); + + _transpose_matrix4x4(out, rgba); + r->q = out[0]; + g->q = out[1]; + b->q = out[2]; + a->q = out[3]; +} + + +static void +exec_tex(struct spu_exec_machine *mach, + const struct tgsi_full_instruction *inst, + boolean biasLod) +{ + const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + union spu_exec_channel r[8]; + uint chan_index; + float lodBias; + + /* printf("Sampler %u unit %u\n", sampler, unit); */ + + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: + + FETCH(&r[0], 0, CHAN_X); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[1], 0, CHAN_W); + r[0].q = micro_div(r[0].q, r[1].q); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[1], 0, CHAN_W); + lodBias = r[2].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ + &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ + break; + + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + r[0].q = micro_div(r[0].q, r[3].q); + r[1].q = micro_div(r[1].q, r[3].q); + r[2].q = micro_div(r[2].q, r[3].q); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, /* inputs */ + &r[0], &r[1], &r[2], &r[3]); /* outputs */ + break; + + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + r[0].q = micro_div(r[0].q, r[3].q); + r[1].q = micro_div(r[1].q, r[3].q); + r[2].q = micro_div(r[2].q, r[3].q); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, + &r[0], &r[1], &r[2], &r[3]); + break; + + default: + assert (0); + } + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } +} + + + +static void +constant_interpolation( + struct spu_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; + } +} + +static void +linear_interpolation( + struct spu_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + mach->Inputs[attrib].xyzw[chan].f[0] = a0; + mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; + mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; + mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; +} + +static void +perspective_interpolation( + struct spu_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + const float *w = mach->QuadPos.xyzw[3].f; + /* divide by W here */ + mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; + mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; + mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; + mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; +} + + +typedef void (* interpolation_func)( + struct spu_exec_machine *mach, + unsigned attrib, + unsigned chan ); + +static void +exec_declaration(struct spu_exec_machine *mach, + const struct tgsi_full_declaration *decl) +{ + if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + interpolation_func interp; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + switch( decl->Interpolation.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + interp = constant_interpolation; + break; + + case TGSI_INTERPOLATE_LINEAR: + interp = linear_interpolation; + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + interp = perspective_interpolation; + break; + + default: + assert( 0 ); + } + + if( mask == TGSI_WRITEMASK_XYZW ) { + unsigned i, j; + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + interp( mach, i, j ); + } + } + } + else { + unsigned i, j; + + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + interp( mach, i, j ); + } + } + } + } + } + } +} + +static void +exec_instruction( + struct spu_exec_machine *mach, + const struct tgsi_full_instruction *inst, + int *pc ) +{ + uint chan_index; + union spu_exec_channel r[8]; + + (*pc)++; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = si_cflts(r[0].q, 0); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_X ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + r[0].q = micro_max(r[0].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[1], 0, CHAN_Y ); + r[1].q = micro_max(r[1].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); + + FETCH( &r[2], 0, CHAN_W ); + r[2].q = micro_min(r[2].q, mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].q); + r[2].q = micro_max(r[2].q, mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].q); + r[1].q = micro_pow(r[1].q, r[2].q); + + /* r0 = (r0 > 0.0) ? r1 : 0.0 + */ + r[0].q = si_fcgt(r[0].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); + r[0].q = si_selb(mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q, r[1].q, + r[0].q); + STORE( &r[0], 0, CHAN_Z ); + } + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( &r[0], 0, CHAN_X ); + r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( &r[0], 0, CHAN_X ); + r[0].q = micro_sqrt(r[0].q); + r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + assert (0); + break; + + case TGSI_OPCODE_LOG: + assert (0); + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + r[0].q = si_fm(r[0].q, r[1].q); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + r[0].q = si_fa(r[0].q, r[1].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + r[0].q = si_fm(r[0].q, r[1].q); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); + + + FETCH( &r[1], 0, CHAN_Z ); + FETCH( &r[2], 1, CHAN_Z ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + r[0].q = si_fm(r[0].q, r[1].q); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); + + FETCH(&r[1], 0, CHAN_W); + FETCH(&r[2], 1, CHAN_W); + + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + FETCH( &r[0], 0, CHAN_Y ); + FETCH( &r[1], 1, CHAN_Y); + r[0].q = si_fm(r[0].q, r[1].q); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_Z ); + STORE( &r[0], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + FETCH( &r[0], 1, CHAN_W ); + STORE( &r[0], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + r[0].q = micro_min(r[0].q, r[1].q); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + r[0].q = micro_max(r[0].q, r[1].q); + + STORE(&r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + + r[0].q = micro_ge(r[0].q, r[1].q); + r[0].q = si_xori(r[0].q, 0xff); + + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + r[0].q = micro_ge(r[0].q, r[1].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + FETCH( &r[2], 2, chan_index ); + r[0].q = si_fma(r[0].q, r[1].q, r[2].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + r[0].q = si_fs(r[0].q, r[1].q); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + r[1].q = si_fs(r[1].q, r[2].q); + r[0].q = si_fma(r[0].q, r[1].q, r[2].q); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_CND: + assert (0); + break; + + case TGSI_OPCODE_CND0: + assert (0); + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert (0); + break; + + case TGSI_OPCODE_INDEX: + assert (0); + break; + + case TGSI_OPCODE_NEGATE: + assert (0); + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = micro_frc(r[0].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + assert (0); + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = micro_flr(r[0].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = micro_rnd(r[0].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH(&r[0], 0, CHAN_X); + + r[0].q = micro_pow(mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q, r[0].q); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( &r[0], 0, CHAN_X ); + r[0].q = micro_lg2(r[0].q); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + r[0].q = micro_pow(r[0].q, r[1].q); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + FETCH(&r[0], 0, CHAN_Y); + FETCH(&r[1], 1, CHAN_Z); + FETCH(&r[3], 0, CHAN_Z); + FETCH(&r[4], 1, CHAN_Y); + + /* r2 = (r0 * r1) - (r3 * r5) + */ + r[2].q = si_fm(r[3].q, r[5].q); + r[2].q = si_fms(r[0].q, r[1].q, r[2].q); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &r[2], 0, CHAN_X ); + } + + FETCH(&r[2], 1, CHAN_X); + FETCH(&r[5], 0, CHAN_X); + + /* r3 = (r3 * r2) - (r1 * r5) + */ + r[1].q = si_fm(r[1].q, r[5].q); + r[3].q = si_fms(r[3].q, r[2].q, r[1].q); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + STORE( &r[3], 0, CHAN_Y ); + } + + /* r5 = (r5 * r4) - (r0 * r2) + */ + r[0].q = si_fm(r[0].q, r[2].q); + r[5].q = si_fms(r[5].q, r[4].q, r[0].q); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( &r[5], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + assert (0); + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + + r[0].q = micro_abs(r[0].q); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_RCC: + assert (0); + break; + + case TGSI_OPCODE_DPH: + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + r[0].q = si_fm(r[0].q, r[1].q); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); + + FETCH(&r[1], 1, CHAN_W); + + r[0].q = si_fa(r[0].q, r[1].q); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH(&r[0], 0, CHAN_X); + + r[0].q = micro_cos(r[0].q); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = micro_ddx(r[0].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDY: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = micro_ddy(r[0].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_KILP: + exec_kilp (mach, inst); + break; + + case TGSI_OPCODE_KIL: + /* for enabled ExecMask bits, set the killed bit */ + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; + break; + + case TGSI_OPCODE_PK2H: + assert (0); + break; + + case TGSI_OPCODE_PK2US: + assert (0); + break; + + case TGSI_OPCODE_PK4B: + assert (0); + break; + + case TGSI_OPCODE_PK4UB: + assert (0); + break; + + case TGSI_OPCODE_RFL: + assert (0); + break; + + case TGSI_OPCODE_SEQ: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + + r[0].q = si_fceq(r[0].q, r[1].q); + + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SFL: + assert (0); + break; + + case TGSI_OPCODE_SGT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + r[0].q = si_fcgt(r[0].q, r[1].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SIN: + FETCH( &r[0], 0, CHAN_X ); + r[0].q = micro_sin(r[0].q); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + + r[0].q = si_fcgt(r[0].q, r[1].q); + r[0].q = si_xori(r[0].q, 0xff); + + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SNE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + + r[0].q = si_fceq(r[0].q, r[1].q); + r[0].q = si_xori(r[0].q, 0xff); + + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_STR: + assert (0); + break; + + case TGSI_OPCODE_TEX: + /* simple texture lookup */ + /* src[0] = texcoord */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, FALSE); + break; + + case TGSI_OPCODE_TXB: + /* Texture lookup with lod bias */ + /* src[0] = texcoord (src[0].w = load bias) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE); + break; + + case TGSI_OPCODE_TXD: + /* Texture lookup with explict partial derivatives */ + /* src[0] = texcoord */ + /* src[1] = d[strq]/dx */ + /* src[2] = d[strq]/dy */ + /* src[3] = sampler unit */ + assert (0); + break; + + case TGSI_OPCODE_TXL: + /* Texture lookup with explit LOD */ + /* src[0] = texcoord (src[0].w = load bias) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE); + break; + + case TGSI_OPCODE_UP2H: + assert (0); + break; + + case TGSI_OPCODE_UP2US: + assert (0); + break; + + case TGSI_OPCODE_UP4B: + assert (0); + break; + + case TGSI_OPCODE_UP4UB: + assert (0); + break; + + case TGSI_OPCODE_X2D: + assert (0); + break; + + case TGSI_OPCODE_ARA: + assert (0); + break; + + case TGSI_OPCODE_ARR: + assert (0); + break; + + case TGSI_OPCODE_BRA: + assert (0); + break; + + case TGSI_OPCODE_CAL: + /* skip the call if no execution channels are enabled */ + if (mach->ExecMask) { + /* do the call */ + + /* push the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + + assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); + mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; + + /* note that PC was already incremented above */ + mach->CallStack[mach->CallStackTop++] = *pc; + *pc = inst->InstructionExtLabel.Label; + } + break; + + case TGSI_OPCODE_RET: + mach->FuncMask &= ~mach->ExecMask; + UPDATE_EXEC_MASK(mach); + + if (mach->ExecMask == 0x0) { + /* really return now (otherwise, keep executing */ + + if (mach->CallStackTop == 0) { + /* returning from main() */ + *pc = -1; + return; + } + *pc = mach->CallStack[--mach->CallStackTop]; + + /* pop the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->FuncStackTop > 0); + mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; + + UPDATE_EXEC_MASK(mach); + } + break; + + case TGSI_OPCODE_SSG: + assert (0); + break; + + case TGSI_OPCODE_CMP: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + /* r0 = (r0 < 0.0) ? r1 : r2 + */ + r[3].q = si_xor(r[3].q, r[3].q); + r[0].q = micro_lt(r[0].q, r[3].q); + r[0].q = si_selb(r[1].q, r[2].q, r[0].q); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_SCS: + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( &r[0], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + r[1].q = micro_cos(r[0].q); + STORE( &r[1], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + r[1].q = micro_sin(r[0].q); + STORE( &r[1], 0, CHAN_Y ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_NRM: + assert (0); + break; + + case TGSI_OPCODE_DIV: + assert( 0 ); + break; + + case TGSI_OPCODE_DP2: + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + r[0].q = si_fm(r[0].q, r[1].q); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + r[0].q = si_fma(r[1].q, r[2].q, r[0].q); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_IF: + /* push CondMask */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + FETCH( &r[0], 0, CHAN_X ); + /* update CondMask */ + if( ! r[0].u[0] ) { + mach->CondMask &= ~0x1; + } + if( ! r[0].u[1] ) { + mach->CondMask &= ~0x2; + } + if( ! r[0].u[2] ) { + mach->CondMask &= ~0x4; + } + if( ! r[0].u[3] ) { + mach->CondMask &= ~0x8; + } + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ELSE */ + break; + + case TGSI_OPCODE_ELSE: + /* invert CondMask wrt previous mask */ + { + uint prevMask; + assert(mach->CondStackTop > 0); + prevMask = mach->CondStack[mach->CondStackTop - 1]; + mach->CondMask = ~mach->CondMask & prevMask; + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ENDIF */ + } + break; + + case TGSI_OPCODE_ENDIF: + /* pop CondMask */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_END: + /* halt execution */ + *pc = -1; + break; + + case TGSI_OPCODE_REP: + assert (0); + break; + + case TGSI_OPCODE_ENDREP: + assert (0); + break; + + case TGSI_OPCODE_PUSHA: + assert (0); + break; + + case TGSI_OPCODE_POPA: + assert (0); + break; + + case TGSI_OPCODE_CEIL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = micro_ceil(r[0].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_I2F: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = si_csflt(r[0].q, 0); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_NOT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = si_xorbi(r[0].q, 0xff); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TRUNC: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + r[0].q = micro_trunc(r[0].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + + r[0].q = si_shl(r[0].q, r[1].q); + + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + r[0].q = micro_ishr(r[0].q, r[1].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_AND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + r[0].q = si_and(r[0].q, r[1].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_OR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + r[0].q = si_or(r[0].q, r[1].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOD: + assert (0); + break; + + case TGSI_OPCODE_XOR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + r[0].q = si_xor(r[0].q, r[1].q); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SAD: + assert (0); + break; + + case TGSI_OPCODE_TXF: + assert (0); + break; + + case TGSI_OPCODE_TXQ: + assert (0); + break; + + case TGSI_OPCODE_EMIT: + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; + break; + + case TGSI_OPCODE_ENDPRIM: + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; + break; + + case TGSI_OPCODE_LOOP: + /* fall-through (for now) */ + case TGSI_OPCODE_BGNLOOP2: + /* push LoopMask and ContMasks */ + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + break; + + case TGSI_OPCODE_ENDLOOP: + /* fall-through (for now at least) */ + case TGSI_OPCODE_ENDLOOP2: + /* Restore ContMask, but don't pop */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; + if (mach->LoopMask) { + /* repeat loop: jump to instruction just past BGNLOOP */ + *pc = inst->InstructionExtLabel.Label + 1; + } + else { + /* exit loop: pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + /* pop ContMask */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + } + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BRK: + /* turn off loop channels for each enabled exec channel */ + mach->LoopMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_CONT: + /* turn off cont channels for each enabled exec channel */ + mach->ContMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BGNSUB: + /* no-op */ + break; + + case TGSI_OPCODE_ENDSUB: + /* no-op */ + break; + + case TGSI_OPCODE_NOISE1: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE2: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE3: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE4: + assert( 0 ); + break; + + case TGSI_OPCODE_NOP: + break; + + default: + assert( 0 ); + } +} + + +/** + * Run TGSI interpreter. + * \return bitmask of "alive" quad components + */ +uint +spu_exec_machine_run( struct spu_exec_machine *mach ) +{ + uint i; + int pc = 0; + + mach->CondMask = 0xf; + mach->LoopMask = 0xf; + mach->ContMask = 0xf; + mach->FuncMask = 0xf; + mach->ExecMask = 0xf; + + mach->CondStackTop = 0; /* temporarily subvert this assertion */ + assert(mach->CondStackTop == 0); + assert(mach->LoopStackTop == 0); + assert(mach->ContStackTop == 0); + assert(mach->CallStackTop == 0); + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; + + if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; + mach->Primitives[0] = 0; + } + + + /* execute declarations (interpolants) */ + if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { + for (i = 0; i < mach->NumDeclarations; i++) { + uint8_t buffer[sizeof(struct tgsi_full_declaration) + 32] ALIGN16_ATTRIB; + struct tgsi_full_declaration decl; + unsigned long decl_addr = (unsigned long) (mach->Declarations+i); + unsigned size = ((sizeof(decl) + (decl_addr & 0x0f) + 0x0f) & ~0x0f); + + mfc_get(buffer, decl_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); + wait_on_mask(1 << TAG_INSTRUCTION_FETCH); + + memcpy(& decl, buffer + (decl_addr & 0x0f), sizeof(decl)); + exec_declaration( mach, &decl ); + } + } + + /* execute instructions, until pc is set to -1 */ + while (pc != -1) { + uint8_t buffer[sizeof(struct tgsi_full_instruction) + 32] ALIGN16_ATTRIB; + struct tgsi_full_instruction inst; + unsigned long inst_addr = (unsigned long) (mach->Instructions + pc); + unsigned size = ((sizeof(inst) + (inst_addr & 0x0f) + 0x0f) & ~0x0f); + + assert(pc < mach->NumInstructions); + mfc_get(buffer, inst_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); + wait_on_mask(1 << TAG_INSTRUCTION_FETCH); + + memcpy(& inst, buffer + (inst_addr & 0x0f), sizeof(inst)); + exec_instruction( mach, & inst, &pc ); + } + +#if 0 + /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ + if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { + /* + * Scale back depth component. + */ + for (i = 0; i < 4; i++) + mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; + } +#endif + + return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; +} + + diff --git a/src/gallium/drivers/cell/spu/spu_exec.h b/src/gallium/drivers/cell/spu/spu_exec.h new file mode 100644 index 0000000000..b4c7661ef6 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_exec.h @@ -0,0 +1,172 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#if !defined SPU_EXEC_H +#define SPU_EXEC_H + +#include "pipe/p_compiler.h" +#include "pipe/tgsi/exec/tgsi_exec.h" + +#if defined __cplusplus +extern "C" { +#endif + +/** + * Registers may be treated as float, signed int or unsigned int. + */ +union spu_exec_channel +{ + float f[QUAD_SIZE]; + int i[QUAD_SIZE]; + unsigned u[QUAD_SIZE]; + qword q; +}; + +/** + * A vector[RGBA] of channels[4 pixels] + */ +struct spu_exec_vector +{ + union spu_exec_channel xyzw[NUM_CHANNELS]; +}; + +/** + * For fragment programs, information for computing fragment input + * values from plane equation of the triangle/line. + */ +struct spu_interp_coef +{ + float a0[NUM_CHANNELS]; /* in an xyzw layout */ + float dadx[NUM_CHANNELS]; + float dady[NUM_CHANNELS]; +}; + + +struct softpipe_tile_cache; /**< Opaque to TGSI */ + +/** + * Information for sampling textures, which must be implemented + * by code outside the TGSI executor. + */ +struct spu_sampler +{ + const struct pipe_sampler_state *state; + struct pipe_texture *texture; + /** Get samples for four fragments in a quad */ + void (*get_samples)(struct spu_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); + void *pipe; /*XXX temporary*/ + struct softpipe_tile_cache *cache; +}; + + +/** + * Run-time virtual machine state for executing TGSI shader. + */ +struct spu_exec_machine +{ + /* + * 32 program temporaries + * 4 internal temporaries + * 1 address + */ + struct spu_exec_vector Temps[TGSI_EXEC_NUM_TEMPS + + TGSI_EXEC_NUM_ADDRS + 1] + ALIGN16_ATTRIB; + + struct spu_exec_vector *Addrs; + + struct spu_sampler *Samplers; + + float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; + unsigned ImmLimit; + float (*Consts)[4]; + struct spu_exec_vector *Inputs; + struct spu_exec_vector *Outputs; + unsigned Processor; + + /* GEOMETRY processor only. */ + unsigned *Primitives; + + /* FRAGMENT processor only. */ + const struct spu_interp_coef *InterpCoefs; + struct spu_exec_vector QuadPos; + + /* Conditional execution masks */ + uint CondMask; /**< For IF/ELSE/ENDIF */ + uint LoopMask; /**< For BGNLOOP/ENDLOOP */ + uint ContMask; /**< For loop CONT statements */ + uint FuncMask; /**< For function calls */ + uint ExecMask; /**< = CondMask & LoopMask */ + + /** Condition mask stack (for nested conditionals) */ + uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; + int CondStackTop; + + /** Loop mask stack (for nested loops) */ + uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopStackTop; + + /** Loop continue mask stack (see comments in tgsi_exec.c) */ + uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int ContStackTop; + + /** Function execution mask stack (for executing subroutine code) */ + uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; + int FuncStackTop; + + /** Function call stack for saving/restoring the program counter */ + uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; + int CallStackTop; + + struct tgsi_full_instruction *Instructions; + uint NumInstructions; + + struct tgsi_full_declaration *Declarations; + uint NumDeclarations; +}; + + +extern void +spu_exec_machine_init(struct spu_exec_machine *mach, + uint numSamplers, + struct spu_sampler *samplers, + unsigned processor); + +extern uint +spu_exec_machine_run( struct spu_exec_machine *mach ); + + +#if defined __cplusplus +} /* extern "C" */ +#endif + +#endif /* SPU_EXEC_H */ diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c new file mode 100644 index 0000000000..e375197fe6 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -0,0 +1,567 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/* main() for Cell SPU code */ + + +#include +#include + +#include "spu_main.h" +#include "spu_render.h" +#include "spu_texture.h" +#include "spu_tile.h" +//#include "spu_test.h" +#include "spu_vertex_shader.h" +#include "pipe/cell/common.h" +#include "pipe/p_defines.h" + + +/* +helpful headers: +/usr/lib/gcc/spu/4.1.1/include/spu_mfcio.h +/opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h +*/ + +boolean Debug = FALSE; + +struct spu_global spu; + +struct spu_vs_context draw; + +/** + * Tell the PPU that this SPU has finished copying a buffer to + * local store and that it may be reused by the PPU. + * This is done by writting a 16-byte batch-buffer-status block back into + * main memory (in cell_context->buffer_status[]). + */ +static void +release_buffer(uint buffer) +{ + /* Evidently, using less than a 16-byte status doesn't work reliably */ + static const uint status[4] ALIGN16_ATTRIB + = {CELL_BUFFER_STATUS_FREE, 0, 0, 0}; + + const uint index = 4 * (spu.init.id * CELL_NUM_BUFFERS + buffer); + uint *dst = spu.init.buffer_status + index; + + ASSERT(buffer < CELL_NUM_BUFFERS); + + mfc_put((void *) &status, /* src in local memory */ + (unsigned int) dst, /* dst in main memory */ + sizeof(status), /* size */ + TAG_MISC, /* tag is unimportant */ + 0, /* tid */ + 0 /* rid */); +} + + +/** + * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled + * tiles back to the main framebuffer. + */ +static void +really_clear_tiles(uint surfaceIndex) +{ + const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; + uint i; + + if (surfaceIndex == 0) { + clear_c_tile(&spu.ctile); + + for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { + uint tx = i % spu.fb.width_tiles; + uint ty = i / spu.fb.width_tiles; + if (spu.ctile_status[ty][tx] == TILE_STATUS_CLEAR) { + put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); + } + } + } + else { + clear_z_tile(&spu.ztile); + + for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { + uint tx = i % spu.fb.width_tiles; + uint ty = i / spu.fb.width_tiles; + if (spu.ztile_status[ty][tx] == TILE_STATUS_CLEAR) + put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 1); + } + } + +#if 0 + wait_on_mask(1 << TAG_SURFACE_CLEAR); +#endif +} + + +static void +cmd_clear_surface(const struct cell_command_clear_surface *clear) +{ + const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; + uint i; + + if (Debug) + printf("SPU %u: CLEAR SURF %u to 0x%08x\n", spu.init.id, + clear->surface, clear->value); + +#define CLEAR_OPT 1 +#if CLEAR_OPT + /* set all tile's status to CLEAR */ + if (clear->surface == 0) { + memset(spu.ctile_status, TILE_STATUS_CLEAR, sizeof(spu.ctile_status)); + spu.fb.color_clear_value = clear->value; + } + else { + memset(spu.ztile_status, TILE_STATUS_CLEAR, sizeof(spu.ztile_status)); + spu.fb.depth_clear_value = clear->value; + } + return; +#endif + + if (clear->surface == 0) { + spu.fb.color_clear_value = clear->value; + clear_c_tile(&spu.ctile); + } + else { + spu.fb.depth_clear_value = clear->value; + clear_z_tile(&spu.ztile); + } + + /* + printf("SPU: %s num=%d w=%d h=%d\n", + __FUNCTION__, num_tiles, spu.fb.width_tiles, spu.fb.height_tiles); + */ + + for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { + uint tx = i % spu.fb.width_tiles; + uint ty = i / spu.fb.width_tiles; + if (clear->surface == 0) + put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); + else + put_tile(tx, ty, &spu.ztile, TAG_SURFACE_CLEAR, 1); + /* XXX we don't want this here, but it fixes bad tile results */ + } + +#if 0 + wait_on_mask(1 << TAG_SURFACE_CLEAR); +#endif + + if (Debug) + printf("SPU %u: CLEAR SURF done\n", spu.init.id); +} + + +static void +cmd_release_verts(const struct cell_command_release_verts *release) +{ + if (Debug) + printf("SPU %u: RELEASE VERTS %u\n", + spu.init.id, release->vertex_buf); + ASSERT(release->vertex_buf != ~0U); + release_buffer(release->vertex_buf); +} + + +static void +cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) +{ + if (Debug) + printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", + spu.init.id, + cmd->width, + cmd->height, + cmd->color_start, + cmd->color_format, + cmd->depth_format); + + ASSERT_ALIGN16(cmd->color_start); + ASSERT_ALIGN16(cmd->depth_start); + + spu.fb.color_start = cmd->color_start; + spu.fb.depth_start = cmd->depth_start; + spu.fb.color_format = cmd->color_format; + spu.fb.depth_format = cmd->depth_format; + spu.fb.width = cmd->width; + spu.fb.height = cmd->height; + spu.fb.width_tiles = (spu.fb.width + TILE_SIZE - 1) / TILE_SIZE; + spu.fb.height_tiles = (spu.fb.height + TILE_SIZE - 1) / TILE_SIZE; + + if (spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM) + spu.fb.zsize = 4; + else if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) + spu.fb.zsize = 2; + else + spu.fb.zsize = 0; + + if (spu.fb.color_format == PIPE_FORMAT_A8R8G8B8_UNORM) + spu.color_shuffle = ((vector unsigned char) { + 12, 0, 4, 8, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0}); + else if (spu.fb.color_format == PIPE_FORMAT_B8G8R8A8_UNORM) + spu.color_shuffle = ((vector unsigned char) { + 8, 4, 0, 12, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0}); + else + ASSERT(0); +} + + +static void +cmd_state_blend(const struct pipe_blend_state *state) +{ + if (Debug) + printf("SPU %u: BLEND: enabled %d\n", + spu.init.id, + state->blend_enable); + + memcpy(&spu.blend, state, sizeof(*state)); +} + + +static void +cmd_state_depth_stencil(const struct pipe_depth_stencil_alpha_state *state) +{ + if (Debug) + printf("SPU %u: DEPTH_STENCIL: ztest %d\n", + spu.init.id, + state->depth.enabled); + + memcpy(&spu.depth_stencil, state, sizeof(*state)); +} + + +static void +cmd_state_sampler(const struct pipe_sampler_state *state) +{ + if (Debug) + printf("SPU %u: SAMPLER\n", + spu.init.id); + + memcpy(&spu.sampler[0], state, sizeof(*state)); + if (spu.sampler[0].min_img_filter == PIPE_TEX_FILTER_LINEAR) + spu.sample_texture = sample_texture_bilinear; + else + spu.sample_texture = sample_texture_nearest; +} + + +static void +cmd_state_texture(const struct cell_command_texture *texture) +{ + if (Debug) + printf("SPU %u: TEXTURE at %p size %u x %u\n", + spu.init.id, texture->start, texture->width, texture->height); + + memcpy(&spu.texture, texture, sizeof(*texture)); + spu.tex_size = (vector float) + { spu.texture.width, spu.texture.height, 0.0, 0.0}; + spu.tex_size_mask = (vector unsigned int) + { spu.texture.width - 1, spu.texture.height - 1, 0, 0 }; +} + + +static void +cmd_state_vertex_info(const struct vertex_info *vinfo) +{ + if (Debug) { + printf("SPU %u: VERTEX_INFO num_attribs=%u\n", spu.init.id, + vinfo->num_attribs); + } + ASSERT(vinfo->num_attribs >= 1); + ASSERT(vinfo->num_attribs <= 8); + memcpy(&spu.vertex_info, vinfo, sizeof(*vinfo)); +} + + +static void +cmd_state_vs_array_info(const struct cell_array_info *vs_info) +{ + const unsigned attr = vs_info->attr; + + ASSERT(attr < PIPE_ATTRIB_MAX); + draw.vertex_fetch.src_ptr[attr] = vs_info->base; + draw.vertex_fetch.pitch[attr] = vs_info->pitch; + draw.vertex_fetch.format[attr] = vs_info->format; + draw.vertex_fetch.dirty = 1; +} + + +static void +cmd_finish(void) +{ + if (Debug) + printf("SPU %u: FINISH\n", spu.init.id); + really_clear_tiles(0); + /* wait for all outstanding DMAs to finish */ + mfc_write_tag_mask(~0); + mfc_read_tag_status_all(); + /* send mbox message to PPU */ + spu_write_out_mbox(CELL_CMD_FINISH); +} + + +/** + * Execute a batch of commands + * The opcode param encodes the location of the buffer and its size. + */ +static void +cmd_batch(uint opcode) +{ + const uint buf = (opcode >> 8) & 0xff; + uint size = (opcode >> 16); + uint64_t buffer[CELL_BUFFER_SIZE / 8] ALIGN16_ATTRIB; + const unsigned usize = size / sizeof(buffer[0]); + uint pos; + + if (Debug) + printf("SPU %u: BATCH buffer %u, len %u, from %p\n", + spu.init.id, buf, size, spu.init.buffers[buf]); + + ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH); + + ASSERT_ALIGN16(spu.init.buffers[buf]); + + size = ROUNDUP16(size); + + ASSERT_ALIGN16(spu.init.buffers[buf]); + + mfc_get(buffer, /* dest */ + (unsigned int) spu.init.buffers[buf], /* src */ + size, + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); + + /* Tell PPU we're done copying the buffer to local store */ + if (Debug) + printf("SPU %u: release batch buf %u\n", spu.init.id, buf); + release_buffer(buf); + + for (pos = 0; pos < usize; /* no incr */) { + switch (buffer[pos]) { + case CELL_CMD_STATE_FRAMEBUFFER: + { + struct cell_command_framebuffer *fb + = (struct cell_command_framebuffer *) &buffer[pos]; + cmd_state_framebuffer(fb); + pos += sizeof(*fb) / 8; + } + break; + case CELL_CMD_CLEAR_SURFACE: + { + struct cell_command_clear_surface *clr + = (struct cell_command_clear_surface *) &buffer[pos]; + cmd_clear_surface(clr); + pos += sizeof(*clr) / 8; + } + break; + case CELL_CMD_RENDER: + { + struct cell_command_render *render + = (struct cell_command_render *) &buffer[pos]; + uint pos_incr; + cmd_render(render, &pos_incr); + pos += pos_incr; + } + break; + case CELL_CMD_RELEASE_VERTS: + { + struct cell_command_release_verts *release + = (struct cell_command_release_verts *) &buffer[pos]; + cmd_release_verts(release); + pos += sizeof(*release) / 8; + } + break; + case CELL_CMD_FINISH: + cmd_finish(); + pos += 1; + break; + case CELL_CMD_STATE_BLEND: + cmd_state_blend((struct pipe_blend_state *) + &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct pipe_blend_state)) / 8); + break; + case CELL_CMD_STATE_DEPTH_STENCIL: + cmd_state_depth_stencil((struct pipe_depth_stencil_alpha_state *) + &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct pipe_depth_stencil_alpha_state)) / 8); + break; + case CELL_CMD_STATE_SAMPLER: + cmd_state_sampler((struct pipe_sampler_state *) &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct pipe_sampler_state)) / 8); + break; + case CELL_CMD_STATE_TEXTURE: + cmd_state_texture((struct cell_command_texture *) &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct cell_command_texture)) / 8); + break; + case CELL_CMD_STATE_VERTEX_INFO: + cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct vertex_info)) / 8); + break; + case CELL_CMD_STATE_VIEWPORT: + (void) memcpy(& draw.viewport, &buffer[pos+1], + sizeof(struct pipe_viewport_state)); + pos += (1 + ROUNDUP8(sizeof(struct pipe_viewport_state)) / 8); + break; + case CELL_CMD_STATE_VS_ARRAY_INFO: + cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct cell_array_info)) / 8); + break; + default: + printf("SPU %u: bad opcode: 0x%llx\n", spu.init.id, buffer[pos]); + ASSERT(0); + break; + } + } + + if (Debug) + printf("SPU %u: BATCH complete\n", spu.init.id); +} + + +/** + * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. + */ +static void +main_loop(void) +{ + struct cell_command cmd; + int exitFlag = 0; + + if (Debug) + printf("SPU %u: Enter main loop\n", spu.init.id); + + ASSERT((sizeof(struct cell_command) & 0xf) == 0); + ASSERT_ALIGN16(&cmd); + + while (!exitFlag) { + unsigned opcode; + int tag = 0; + + if (Debug) + printf("SPU %u: Wait for cmd...\n", spu.init.id); + + /* read/wait from mailbox */ + opcode = (unsigned int) spu_read_in_mbox(); + + if (Debug) + printf("SPU %u: got cmd 0x%x\n", spu.init.id, opcode); + + /* command payload */ + mfc_get(&cmd, /* dest */ + (unsigned int) spu.init.cmd, /* src */ + sizeof(struct cell_command), /* bytes */ + tag, + 0, /* tid */ + 0 /* rid */); + wait_on_mask( 1 << tag ); + + /* + * NOTE: most commands should be contained in a batch buffer + */ + + switch (opcode & CELL_CMD_OPCODE_MASK) { + case CELL_CMD_EXIT: + if (Debug) + printf("SPU %u: EXIT\n", spu.init.id); + exitFlag = 1; + break; + case CELL_CMD_VS_EXECUTE: + spu_execute_vertex_shader(&draw, &cmd.vs); + break; + case CELL_CMD_BATCH: + cmd_batch(opcode); + break; + default: + printf("Bad opcode!\n"); + } + + } + + if (Debug) + printf("SPU %u: Exit main loop\n", spu.init.id); +} + + + +static void +one_time_init(void) +{ + memset(spu.ctile_status, TILE_STATUS_DEFINED, sizeof(spu.ctile_status)); + memset(spu.ztile_status, TILE_STATUS_DEFINED, sizeof(spu.ztile_status)); + invalidate_tex_cache(); +} + + + +/* In some versions of the SDK the SPE main takes 'unsigned long' as a + * parameter. In others it takes 'unsigned long long'. Use a define to + * select between the two. + */ +#ifdef SPU_MAIN_PARAM_LONG_LONG +typedef unsigned long long main_param_t; +#else +typedef unsigned long main_param_t; +#endif + +/** + * SPE entrypoint. + */ +int +main(main_param_t speid, main_param_t argp) +{ + int tag = 0; + + (void) speid; + + ASSERT(sizeof(tile_t) == TILE_SIZE * TILE_SIZE * 4); + ASSERT(sizeof(struct cell_command_render) % 8 == 0); + + one_time_init(); + + if (Debug) + printf("SPU: main() speid=%lu\n", speid); + + mfc_get(&spu.init, /* dest */ + (unsigned int) argp, /* src */ + sizeof(struct cell_init_info), /* bytes */ + tag, + 0, /* tid */ + 0 /* rid */); + wait_on_mask( 1 << tag ); + +#if 0 + if (spu.init.id==0) + spu_test_misc(); +#endif + + main_loop(); + + return 0; +} diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h new file mode 100644 index 0000000000..1710a17512 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -0,0 +1,177 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SPU_MAIN_H +#define SPU_MAIN_H + + +#include + +#include "pipe/cell/common.h" +#include "pipe/draw/draw_vertex.h" +#include "pipe/p_state.h" + + + +#define MAX_WIDTH 1024 +#define MAX_HEIGHT 1024 + + +typedef union { + ushort us[TILE_SIZE][TILE_SIZE]; + uint ui[TILE_SIZE][TILE_SIZE]; + vector unsigned short us8[TILE_SIZE/2][TILE_SIZE/4]; + vector unsigned int ui4[TILE_SIZE/2][TILE_SIZE/2]; +} tile_t; + + +#define TILE_STATUS_CLEAR 1 +#define TILE_STATUS_DEFINED 2 /**< defined in FB, but not in local store */ +#define TILE_STATUS_CLEAN 3 /**< in local store, but not changed */ +#define TILE_STATUS_DIRTY 4 /**< modified locally, but not put back yet */ +#define TILE_STATUS_GETTING 5 /**< mfc_get() called but not yet arrived */ + + +struct spu_framebuffer { + void *color_start; /**< addr of color surface in main memory */ + void *depth_start; /**< addr of depth surface in main memory */ + enum pipe_format color_format; + enum pipe_format depth_format; + uint width, height; /**< size in pixels */ + uint width_tiles, height_tiles; /**< width and height in tiles */ + + uint color_clear_value; + uint depth_clear_value; + + uint zsize; /**< 0, 2 or 4 bytes per Z */ +} ALIGN16_ATTRIB; + + +/** + * All SPU global/context state will be in singleton object of this type: + */ +struct spu_global +{ + struct cell_init_info init; + + struct spu_framebuffer fb; + struct pipe_blend_state blend_stencil; + struct pipe_depth_stencil_alpha_state depth_stencil; + struct pipe_blend_state blend; + struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; + struct cell_command_texture texture; + + struct vertex_info vertex_info; + + /* XXX more state to come */ + + + /** current color and Z tiles */ + tile_t ctile ALIGN16_ATTRIB; + tile_t ztile ALIGN16_ATTRIB; + + /** Current tiles' status */ + ubyte cur_ctile_status, cur_ztile_status; + + /** Status of all tiles in framebuffer */ + ubyte ctile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + ubyte ztile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; + + + /** for converting RGBA to PIPE_FORMAT_x colors */ + vector unsigned char color_shuffle; + + vector float tex_size; + vector unsigned int tex_size_mask; /**< == int(size - 1) */ + + vector float (*sample_texture)(vector float texcoord); + +} ALIGN16_ATTRIB; + + +extern struct spu_global spu; +extern boolean Debug; + + + + +/* DMA TAGS */ + +#define TAG_SURFACE_CLEAR 10 +#define TAG_VERTEX_BUFFER 11 +#define TAG_READ_TILE_COLOR 12 +#define TAG_READ_TILE_Z 13 +#define TAG_WRITE_TILE_COLOR 14 +#define TAG_WRITE_TILE_Z 15 +#define TAG_INDEX_BUFFER 16 +#define TAG_BATCH_BUFFER 17 +#define TAG_MISC 18 +#define TAG_TEXTURE_TILE 19 +#define TAG_INSTRUCTION_FETCH 20 + + + +static INLINE void +wait_on_mask(unsigned tagMask) +{ + mfc_write_tag_mask( tagMask ); + /* wait for completion of _any_ DMAs specified by tagMask */ + mfc_read_tag_status_any(); +} + + +static INLINE void +wait_on_mask_all(unsigned tagMask) +{ + mfc_write_tag_mask( tagMask ); + /* wait for completion of _any_ DMAs specified by tagMask */ + mfc_read_tag_status_all(); +} + + + + + +static INLINE void +memset16(ushort *d, ushort value, uint count) +{ + uint i; + for (i = 0; i < count; i++) + d[i] = value; +} + + +static INLINE void +memset32(uint *d, uint value, uint count) +{ + uint i; + for (i = 0; i < count; i++) + d[i] = value; +} + + +#endif /* SPU_MAIN_H */ diff --git a/src/gallium/drivers/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c new file mode 100644 index 0000000000..932fb500b3 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_render.c @@ -0,0 +1,301 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include +#include +#include + +#include "spu_main.h" +#include "spu_render.h" +#include "spu_tri.h" +#include "spu_tile.h" +#include "pipe/cell/common.h" + + + +/** + * Given a rendering command's bounding box (in pixels) compute the + * location of the corresponding screen tile bounding box. + */ +static INLINE void +tile_bounding_box(const struct cell_command_render *render, + uint *txmin, uint *tymin, + uint *box_num_tiles, uint *box_width_tiles) +{ +#if 0 + /* Debug: full-window bounding box */ + uint txmax = spu.fb.width_tiles - 1; + uint tymax = spu.fb.height_tiles - 1; + *txmin = 0; + *tymin = 0; + *box_num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; + *box_width_tiles = spu.fb.width_tiles; + (void) render; + (void) txmax; + (void) tymax; +#else + uint txmax, tymax, box_height_tiles; + + *txmin = (uint) render->xmin / TILE_SIZE; + *tymin = (uint) render->ymin / TILE_SIZE; + txmax = (uint) render->xmax / TILE_SIZE; + tymax = (uint) render->ymax / TILE_SIZE; + if (txmax >= spu.fb.width_tiles) + txmax = spu.fb.width_tiles-1; + if (tymax >= spu.fb.height_tiles) + tymax = spu.fb.height_tiles-1; + *box_width_tiles = txmax - *txmin + 1; + box_height_tiles = tymax - *tymin + 1; + *box_num_tiles = *box_width_tiles * box_height_tiles; +#endif +#if 0 + printf("SPU %u: bounds: %g, %g ... %g, %g\n", spu.init.id, + render->xmin, render->ymin, render->xmax, render->ymax); + printf("SPU %u: tiles: %u, %u .. %u, %u\n", + spu.init.id, *txmin, *tymin, txmax, tymax); + ASSERT(render->xmin <= render->xmax); + ASSERT(render->ymin <= render->ymax); +#endif +} + + +/** Check if the tile at (tx,ty) belongs to this SPU */ +static INLINE boolean +my_tile(uint tx, uint ty) +{ + return (spu.fb.width_tiles * ty + tx) % spu.init.num_spus == spu.init.id; +} + + +/** + * Start fetching non-clear color/Z tiles from main memory + */ +static INLINE void +get_cz_tiles(uint tx, uint ty) +{ + if (spu.depth_stencil.depth.enabled) { + if (spu.cur_ztile_status != TILE_STATUS_CLEAR) { + //printf("SPU %u: getting Z tile %u, %u\n", spu.init.id, tx, ty); + get_tile(tx, ty, &spu.ztile, TAG_READ_TILE_Z, 1); + spu.cur_ztile_status = TILE_STATUS_GETTING; + } + } + + if (spu.cur_ctile_status != TILE_STATUS_CLEAR) { + //printf("SPU %u: getting C tile %u, %u\n", spu.init.id, tx, ty); + get_tile(tx, ty, &spu.ctile, TAG_READ_TILE_COLOR, 0); + spu.cur_ctile_status = TILE_STATUS_GETTING; + } +} + + +/** + * Start putting dirty color/Z tiles back to main memory + */ +static INLINE void +put_cz_tiles(uint tx, uint ty) +{ + if (spu.cur_ztile_status == TILE_STATUS_DIRTY) { + /* tile was modified and needs to be written back */ + //printf("SPU %u: put dirty Z tile %u, %u\n", spu.init.id, tx, ty); + put_tile(tx, ty, &spu.ztile, TAG_WRITE_TILE_Z, 1); + spu.cur_ztile_status = TILE_STATUS_DEFINED; + } + else if (spu.cur_ztile_status == TILE_STATUS_GETTING) { + /* tile was never used */ + spu.cur_ztile_status = TILE_STATUS_DEFINED; + //printf("SPU %u: put getting Z tile %u, %u\n", spu.init.id, tx, ty); + } + + if (spu.cur_ctile_status == TILE_STATUS_DIRTY) { + /* tile was modified and needs to be written back */ + //printf("SPU %u: put dirty C tile %u, %u\n", spu.init.id, tx, ty); + put_tile(tx, ty, &spu.ctile, TAG_WRITE_TILE_COLOR, 0); + spu.cur_ctile_status = TILE_STATUS_DEFINED; + } + else if (spu.cur_ctile_status == TILE_STATUS_GETTING) { + /* tile was never used */ + spu.cur_ctile_status = TILE_STATUS_DEFINED; + //printf("SPU %u: put getting C tile %u, %u\n", spu.init.id, tx, ty); + } +} + + +/** + * Wait for 'put' of color/z tiles to complete. + */ +static INLINE void +wait_put_cz_tiles(void) +{ + wait_on_mask(1 << TAG_WRITE_TILE_COLOR); + if (spu.depth_stencil.depth.enabled) { + wait_on_mask(1 << TAG_WRITE_TILE_Z); + } +} + + +/** + * Render primitives + * \param pos_incr returns value indicating how may words to skip after + * this command in the batch buffer + */ +void +cmd_render(const struct cell_command_render *render, uint *pos_incr) +{ + /* we'll DMA into these buffers */ + ubyte vertex_data[CELL_BUFFER_SIZE] ALIGN16_ATTRIB; + const uint vertex_size = render->vertex_size; /* in bytes */ + /*const*/ uint total_vertex_bytes = render->num_verts * vertex_size; + uint index_bytes; + const ubyte *vertices; + const ushort *indexes; + uint i, j; + + + if (Debug) { + printf("SPU %u: RENDER prim %u, num_vert=%u num_ind=%u " + "inline_vert=%u\n", + spu.init.id, + render->prim_type, + render->num_verts, + render->num_indexes, + render->inline_verts); + + /* + printf(" bound: %g, %g .. %g, %g\n", + render->xmin, render->ymin, render->xmax, render->ymax); + */ + } + + ASSERT(sizeof(*render) % 4 == 0); + ASSERT(total_vertex_bytes % 16 == 0); + ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); + ASSERT(render->num_indexes % 3 == 0); + + + /* indexes are right after the render command in the batch buffer */ + indexes = (const ushort *) (render + 1); + index_bytes = ROUNDUP8(render->num_indexes * 2); + *pos_incr = index_bytes / 8 + sizeof(*render) / 8; + + + if (render->inline_verts) { + /* Vertices are after indexes in batch buffer at next 16-byte addr */ + vertices = (const ubyte *) render + (*pos_incr * 8); + vertices = (const ubyte *) align_pointer((void *) vertices, 16); + ASSERT_ALIGN16(vertices); + *pos_incr = ((vertices + total_vertex_bytes) - (ubyte *) render) / 8; + } + else { + /* Begin DMA fetch of vertex buffer */ + ubyte *src = spu.init.buffers[render->vertex_buf]; + ubyte *dest = vertex_data; + + /* skip vertex data we won't use */ +#if 01 + src += render->min_index * vertex_size; + dest += render->min_index * vertex_size; + total_vertex_bytes -= render->min_index * vertex_size; +#endif + ASSERT(total_vertex_bytes % 16 == 0); + ASSERT_ALIGN16(dest); + ASSERT_ALIGN16(src); + + mfc_get(dest, /* in vertex_data[] array */ + (unsigned int) src, /* src in main memory */ + total_vertex_bytes, /* size */ + TAG_VERTEX_BUFFER, + 0, /* tid */ + 0 /* rid */); + + vertices = vertex_data; + + wait_on_mask(1 << TAG_VERTEX_BUFFER); + } + + + /** + ** find tiles which intersect the prim bounding box + **/ + uint txmin, tymin, box_width_tiles, box_num_tiles; + tile_bounding_box(render, &txmin, &tymin, + &box_num_tiles, &box_width_tiles); + + + /* make sure any pending clears have completed */ + wait_on_mask(1 << TAG_SURFACE_CLEAR); /* XXX temporary */ + + + /** + ** loop over tiles, rendering tris + **/ + for (i = 0; i < box_num_tiles; i++) { + const uint tx = txmin + i % box_width_tiles; + const uint ty = tymin + i / box_width_tiles; + + ASSERT(tx < spu.fb.width_tiles); + ASSERT(ty < spu.fb.height_tiles); + + if (!my_tile(tx, ty)) + continue; + + spu.cur_ctile_status = spu.ctile_status[ty][tx]; + spu.cur_ztile_status = spu.ztile_status[ty][tx]; + + get_cz_tiles(tx, ty); + + uint drawn = 0; + + /* loop over tris */ + for (j = 0; j < render->num_indexes; j += 3) { + const float *v0, *v1, *v2; + + v0 = (const float *) (vertices + indexes[j+0] * vertex_size); + v1 = (const float *) (vertices + indexes[j+1] * vertex_size); + v2 = (const float *) (vertices + indexes[j+2] * vertex_size); + + drawn += tri_draw(v0, v1, v2, tx, ty); + } + + //printf("SPU %u: drew %u of %u\n", spu.init.id, drawn, render->num_indexes/3); + + /* write color/z tiles back to main framebuffer, if dirtied */ + put_cz_tiles(tx, ty); + + wait_put_cz_tiles(); /* XXX seems unnecessary... */ + + spu.ctile_status[ty][tx] = spu.cur_ctile_status; + spu.ztile_status[ty][tx] = spu.cur_ztile_status; + } + + if (Debug) + printf("SPU %u: RENDER done\n", + spu.init.id); +} + + diff --git a/src/gallium/drivers/cell/spu/spu_render.h b/src/gallium/drivers/cell/spu/spu_render.h new file mode 100644 index 0000000000..fbcdc5ec31 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_render.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SPU_RENDER_H +#define SPU_RENDER_H + +#include "pipe/cell/common.h" + +extern void +cmd_render(const struct cell_command_render *render, uint *pos_incr); + +#endif /* SPU_RENDER_H */ + diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c new file mode 100644 index 0000000000..3962aaa4a9 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -0,0 +1,217 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_compiler.h" +#include "spu_main.h" +#include "spu_texture.h" +#include "spu_tile.h" +#include "spu_colorpack.h" + + +/** + * Number of texture tiles to cache. + * Note that this will probably be the largest consumer of SPU local store/ + * memory for this driver! + */ +#define CACHE_SIZE 16 + +static tile_t tex_tiles[CACHE_SIZE] ALIGN16_ATTRIB; + +static vector unsigned int tex_tile_xy[CACHE_SIZE]; + + + +/** + * Mark all tex cache entries as invalid. + */ +void +invalidate_tex_cache(void) +{ + /* XXX memset? */ + uint i; + for (i = 0; i < CACHE_SIZE; i++) { + tex_tile_xy[i] = ((vector unsigned int) { ~0U, ~0U, ~0U, ~0U }); + } +} + + +/** + * Return the cache pos/index which corresponds to tile (tx,ty) + */ +static INLINE uint +cache_pos(vector unsigned int txty) +{ + uint pos = (spu_extract(txty,0) + spu_extract(txty,1) * 4) % CACHE_SIZE; + return pos; +} + + +/** + * Make sure the tile for texel (i,j) is present, return its position/index + * in the cache. + */ +static uint +get_tex_tile(vector unsigned int ij) +{ + /* tile address: tx,ty */ + const vector unsigned int txty = spu_rlmask(ij, -5); /* divide by 32 */ + const uint pos = cache_pos(txty); + + if ((spu_extract(tex_tile_xy[pos], 0) != spu_extract(txty, 0)) || + (spu_extract(tex_tile_xy[pos], 1) != spu_extract(txty, 1))) { + + /* texture cache miss, fetch tile from main memory */ + const uint tiles_per_row = spu.texture.width / TILE_SIZE; + const uint bytes_per_tile = sizeof(tile_t); + const void *src = (const ubyte *) spu.texture.start + + (spu_extract(txty,1) * tiles_per_row + spu_extract(txty,0)) * bytes_per_tile; + + printf("SPU %u: tex cache miss at %d, %d pos=%u old=%d,%d\n", + spu.init.id, + spu_extract(txty,0), + spu_extract(txty,1), + pos, + spu_extract(tex_tile_xy[pos],0), + spu_extract(tex_tile_xy[pos],1)); + + ASSERT_ALIGN16(tex_tiles[pos].ui); + ASSERT_ALIGN16(src); + + mfc_get(tex_tiles[pos].ui, /* dest */ + (unsigned int) src, + bytes_per_tile, /* size */ + TAG_TEXTURE_TILE, + 0, /* tid */ + 0 /* rid */); + + wait_on_mask(1 << TAG_TEXTURE_TILE); + + tex_tile_xy[pos] = txty; + } + else { +#if 0 + printf("SPU %u: tex cache HIT at %d, %d\n", + spu.init.id, tx, ty); +#endif + } + + return pos; +} + + +/** + * Get texture sample at texcoord. + * XXX this is extremely primitive for now. + */ +vector float +sample_texture_nearest(vector float texcoord) +{ + vector float tc = spu_mul(texcoord, spu.tex_size); + vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ + itc = spu_and(itc, spu.tex_size_mask); /* mask (GL_REPEAT) */ + vector unsigned int ij = spu_and(itc, TILE_SIZE-1); /* intra tile addr */ + uint pos = get_tex_tile(itc); + uint texel = tex_tiles[pos].ui[spu_extract(ij, 1)][spu_extract(ij, 0)]; + return spu_unpack_A8R8G8B8(texel); +} + + +vector float +sample_texture_bilinear(vector float texcoord) +{ + static const vector unsigned int offset10 = {1, 0, 0, 0}; + static const vector unsigned int offset01 = {0, 1, 0, 0}; + + vector float tc = spu_mul(texcoord, spu.tex_size); + tc = spu_add(tc, spu_splats(-0.5f)); /* half texel bias */ + + /* integer texcoords S,T: */ + vector unsigned int itc00 = spu_convtu(tc, 0); /* convert to int */ + vector unsigned int itc01 = spu_add(itc00, offset01); + vector unsigned int itc10 = spu_add(itc00, offset10); + vector unsigned int itc11 = spu_add(itc10, offset01); + + /* mask (GL_REPEAT) */ + itc00 = spu_and(itc00, spu.tex_size_mask); + itc01 = spu_and(itc01, spu.tex_size_mask); + itc10 = spu_and(itc10, spu.tex_size_mask); + itc11 = spu_and(itc11, spu.tex_size_mask); + + /* intra tile addr */ + vector unsigned int ij00 = spu_and(itc00, TILE_SIZE-1); + vector unsigned int ij01 = spu_and(itc01, TILE_SIZE-1); + vector unsigned int ij10 = spu_and(itc10, TILE_SIZE-1); + vector unsigned int ij11 = spu_and(itc11, TILE_SIZE-1); + + /* get tile cache positions */ + uint pos00 = get_tex_tile(itc00); + uint pos01, pos10, pos11; + if ((spu_extract(ij00, 0) < TILE_SIZE-1) && + (spu_extract(ij00, 1) < TILE_SIZE-1)) { + /* all texels are in the same tile */ + pos01 = pos10 = pos11 = pos00; + } + else { + pos01 = get_tex_tile(itc01); + pos10 = get_tex_tile(itc10); + pos11 = get_tex_tile(itc11); + } + + /* get texels from tiles and convert to float[4] */ + vector float texel00 = spu_unpack_A8R8G8B8(tex_tiles[pos00].ui[spu_extract(ij00, 1)][spu_extract(ij00, 0)]); + vector float texel01 = spu_unpack_A8R8G8B8(tex_tiles[pos01].ui[spu_extract(ij01, 1)][spu_extract(ij01, 0)]); + vector float texel10 = spu_unpack_A8R8G8B8(tex_tiles[pos10].ui[spu_extract(ij10, 1)][spu_extract(ij10, 0)]); + vector float texel11 = spu_unpack_A8R8G8B8(tex_tiles[pos11].ui[spu_extract(ij11, 1)][spu_extract(ij11, 0)]); + + /* Compute weighting factors in [0,1] + * Multiply texcoord by 1024, AND with 1023, convert back to float. + */ + vector float tc1024 = spu_mul(tc, spu_splats(1024.0f)); + vector signed int itc1024 = spu_convts(tc1024, 0); + itc1024 = spu_and(itc1024, spu_splats((1 << 10) - 1)); + vector float weight = spu_convtf(itc1024, 10); + + /* smeared frac and 1-frac */ + vector float sfrac = spu_splats(spu_extract(weight, 0)); + vector float tfrac = spu_splats(spu_extract(weight, 1)); + vector float sfrac1 = spu_sub(spu_splats(1.0f), sfrac); + vector float tfrac1 = spu_sub(spu_splats(1.0f), tfrac); + + /* multiply the samples (colors) by the S/T weights */ + texel00 = spu_mul(spu_mul(texel00, sfrac1), tfrac1); + texel10 = spu_mul(spu_mul(texel10, sfrac ), tfrac1); + texel01 = spu_mul(spu_mul(texel01, sfrac1), tfrac ); + texel11 = spu_mul(spu_mul(texel11, sfrac ), tfrac ); + + /* compute sum of weighted samples */ + vector float texel_sum = spu_add(texel00, texel01); + texel_sum = spu_add(texel_sum, texel10); + texel_sum = spu_add(texel_sum, texel11); + + return texel_sum; +} diff --git a/src/gallium/drivers/cell/spu/spu_texture.h b/src/gallium/drivers/cell/spu/spu_texture.h new file mode 100644 index 0000000000..95eb87080f --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_texture.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SPU_TEXTURE_H +#define SPU_TEXTURE_H + + +#include "pipe/p_compiler.h" + + +extern void +invalidate_tex_cache(void); + + +extern vector float +sample_texture_nearest(vector float texcoord); + + +extern vector float +sample_texture_bilinear(vector float texcoord); + + +#endif /* SPU_TEXTURE_H */ diff --git a/src/gallium/drivers/cell/spu/spu_tile.c b/src/gallium/drivers/cell/spu/spu_tile.c new file mode 100644 index 0000000000..12dc246328 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_tile.c @@ -0,0 +1,83 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + + +#include "spu_tile.h" +#include "spu_main.h" + + +void +get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf) +{ + const uint offset = ty * spu.fb.width_tiles + tx; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4); + const ubyte *src = zBuf ? spu.fb.depth_start : spu.fb.color_start; + + src += offset * bytesPerTile; + + ASSERT(tx < spu.fb.width_tiles); + ASSERT(ty < spu.fb.height_tiles); + ASSERT_ALIGN16(tile); + /* + printf("get_tile: dest: %p src: 0x%x size: %d\n", + tile, (unsigned int) src, bytesPerTile); + */ + mfc_get(tile->ui, /* dest in local memory */ + (unsigned int) src, /* src in main memory */ + bytesPerTile, + tag, + 0, /* tid */ + 0 /* rid */); +} + + +void +put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf) +{ + const uint offset = ty * spu.fb.width_tiles + tx; + const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4); + ubyte *dst = zBuf ? spu.fb.depth_start : spu.fb.color_start; + + dst += offset * bytesPerTile; + + ASSERT(tx < spu.fb.width_tiles); + ASSERT(ty < spu.fb.height_tiles); + ASSERT_ALIGN16(tile); + /* + printf("SPU %u: put_tile: src: %p dst: 0x%x size: %d\n", + spu.init.id, + tile, (unsigned int) dst, bytesPerTile); + */ + mfc_put((void *) tile->ui, /* src in local memory */ + (unsigned int) dst, /* dst in main memory */ + bytesPerTile, + tag, + 0, /* tid */ + 0 /* rid */); +} + diff --git a/src/gallium/drivers/cell/spu/spu_tile.h b/src/gallium/drivers/cell/spu/spu_tile.h new file mode 100644 index 0000000000..e53340a55a --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_tile.h @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SPU_TILE_H +#define SPU_TILE_H + + +#include +#include +#include "spu_main.h" +#include "pipe/cell/common.h" + + + +void +get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf); + +void +put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf); + + + +static INLINE void +clear_c_tile(tile_t *ctile) +{ + memset32((uint*) ctile->ui, + spu.fb.color_clear_value, + TILE_SIZE * TILE_SIZE); +} + + +static INLINE void +clear_z_tile(tile_t *ztile) +{ + if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + memset16((ushort*) ztile->us, + spu.fb.depth_clear_value, + TILE_SIZE * TILE_SIZE); + } + else { + ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); + memset32((uint*) ztile->ui, + spu.fb.depth_clear_value, + TILE_SIZE * TILE_SIZE); + } +} + + +#endif /* SPU_TILE_H */ diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c new file mode 100644 index 0000000000..be9624cf7d --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -0,0 +1,926 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Triangle rendering within a tile. + */ + +#include "pipe/p_compiler.h" +#include "pipe/p_format.h" +#include "pipe/p_util.h" +#include "spu_blend.h" +#include "spu_colorpack.h" +#include "spu_main.h" +#include "spu_texture.h" +#include "spu_tile.h" +#include "spu_tri.h" + +#include "spu_ztest.h" + + +/** Masks are uint[4] vectors with each element being 0 or 0xffffffff */ +typedef vector unsigned int mask_t; + +typedef union +{ + vector float v; + float f[4]; +} float4; + + +/** + * Simplified types taken from other parts of Gallium + */ +struct vertex_header { + vector float data[1]; +}; + + + +/* XXX fix this */ +#undef CEILF +#define CEILF(X) ((float) (int) ((X) + 0.99999)) + + +#define QUAD_TOP_LEFT 0 +#define QUAD_TOP_RIGHT 1 +#define QUAD_BOTTOM_LEFT 2 +#define QUAD_BOTTOM_RIGHT 3 +#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) +#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) +#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) +#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) +#define MASK_ALL 0xf + + +#define DEBUG_VERTS 0 + +/** + * Triangle edge info + */ +struct edge { + float dx; /**< X(v1) - X(v0), used only during setup */ + float dy; /**< Y(v1) - Y(v0), used only during setup */ + float dxdy; /**< dx/dy */ + float sx, sy; /**< first sample point coord */ + int lines; /**< number of lines on this edge */ +}; + + +struct interp_coef +{ + float4 a0; + float4 dadx; + float4 dady; +}; + + +/** + * Triangle setup info (derived from draw_stage). + * Also used for line drawing (taking some liberties). + */ +struct setup_stage { + + /* Vertices are just an array of floats making up each attribute in + * turn. Currently fixed at 4 floats, but should change in time. + * Codegen will help cope with this. + */ + const struct vertex_header *vmax; + const struct vertex_header *vmid; + const struct vertex_header *vmin; + const struct vertex_header *vprovoke; + + struct edge ebot; + struct edge etop; + struct edge emaj; + + float oneoverarea; + + uint tx, ty; + + int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; + +#if 0 + struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; +#else + struct interp_coef coef[PIPE_MAX_SHADER_INPUTS]; +#endif + +#if 0 + struct quad_header quad; +#endif + + struct { + int left[2]; /**< [0] = row0, [1] = row1 */ + int right[2]; + int y; + unsigned y_flags; + unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ + } span; +}; + + + +static struct setup_stage setup; + + + + +#if 0 +/** + * Basically a cast wrapper. + */ +static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) +{ + return (struct setup_stage *)stage; +} +#endif + +#if 0 +/** + * Clip setup.quad against the scissor/surface bounds. + */ +static INLINE void +quad_clip(struct setup_stage *setup) +{ + const struct pipe_scissor_state *cliprect = &setup.softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + + if (setup.quad.x0 >= maxx || + setup.quad.y0 >= maxy || + setup.quad.x0 + 1 < minx || + setup.quad.y0 + 1 < miny) { + /* totally clipped */ + setup.quad.mask = 0x0; + return; + } + if (setup.quad.x0 < minx) + setup.quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + if (setup.quad.y0 < miny) + setup.quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + if (setup.quad.x0 == maxx - 1) + setup.quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + if (setup.quad.y0 == maxy - 1) + setup.quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); +} +#endif + +#if 0 +/** + * Emit a quad (pass to next stage) with clipping. + */ +static INLINE void +clip_emit_quad(struct setup_stage *setup) +{ + quad_clip(setup); + if (setup.quad.mask) { + struct softpipe_context *sp = setup.softpipe; + sp->quad.first->run(sp->quad.first, &setup.quad); + } +} +#endif + +/** + * Evaluate attribute coefficients (plane equations) to compute + * attribute values for the four fragments in a quad. + * Eg: four colors will be compute. + */ +static INLINE void +eval_coeff(uint slot, float x, float y, vector float result[4]) +{ + switch (spu.vertex_info.interp_mode[slot]) { + case INTERP_CONSTANT: + result[QUAD_TOP_LEFT] = + result[QUAD_TOP_RIGHT] = + result[QUAD_BOTTOM_LEFT] = + result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0.v; + break; + + case INTERP_LINEAR: + /* fall-through, for now */ + default: + { + register vector float dadx = setup.coef[slot].dadx.v; + register vector float dady = setup.coef[slot].dady.v; + register vector float topLeft + = spu_add(setup.coef[slot].a0.v, + spu_add(spu_mul(spu_splats(x), dadx), + spu_mul(spu_splats(y), dady))); + + result[QUAD_TOP_LEFT] = topLeft; + result[QUAD_TOP_RIGHT] = spu_add(topLeft, dadx); + result[QUAD_BOTTOM_LEFT] = spu_add(topLeft, dady); + result[QUAD_BOTTOM_RIGHT] = spu_add(spu_add(topLeft, dadx), dady); + } + } +} + + +static INLINE vector float +eval_z(float x, float y) +{ + const uint slot = 0; + const float dzdx = setup.coef[slot].dadx.f[2]; + const float dzdy = setup.coef[slot].dady.f[2]; + const float topLeft = setup.coef[slot].a0.f[2] + x * dzdx + y * dzdy; + const vector float topLeftv = spu_splats(topLeft); + const vector float derivs = (vector float) { 0.0, dzdx, dzdy, dzdx + dzdy }; + return spu_add(topLeftv, derivs); +} + + +static INLINE mask_t +do_depth_test(int x, int y, mask_t quadmask) +{ + float4 zvals; + mask_t mask; + + zvals.v = eval_z((float) x, (float) y); + + if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + int ix = (x - setup.cliprect_minx) / 4; + int iy = (y - setup.cliprect_miny) / 2; + mask = spu_z16_test_less(zvals.v, &spu.ztile.us8[iy][ix], x>>1, quadmask); + } + else { + int ix = (x - setup.cliprect_minx) / 2; + int iy = (y - setup.cliprect_miny) / 2; + mask = spu_z32_test_less(zvals.v, &spu.ztile.ui4[iy][ix], quadmask); + } + + if (spu_extract(spu_orx(mask), 0)) + spu.cur_ztile_status = TILE_STATUS_DIRTY; + + return mask; +} + + +/** + * Emit a quad (pass to next stage). No clipping is done. + * Note: about 1/5 to 1/7 of the time, mask is zero and this function + * should be skipped. But adding the test for that slows things down + * overall. + */ +static INLINE void +emit_quad( int x, int y, mask_t mask ) +{ +#if 0 + struct softpipe_context *sp = setup.softpipe; + setup.quad.x0 = x; + setup.quad.y0 = y; + setup.quad.mask = mask; + sp->quad.first->run(sp->quad.first, &setup.quad); +#else + + if (spu.depth_stencil.depth.enabled) { + mask = do_depth_test(x, y, mask); + } + + /* If any bits in mask are set... */ + if (spu_extract(spu_orx(mask), 0)) { + const int ix = x - setup.cliprect_minx; + const int iy = y - setup.cliprect_miny; + const vector unsigned char shuffle = spu.color_shuffle; + vector float colors[4]; + + spu.cur_ctile_status = TILE_STATUS_DIRTY; + + if (spu.texture.start) { + /* texture mapping */ + vector float texcoords[4]; + eval_coeff(2, (float) x, (float) y, texcoords); + + if (spu_extract(mask, 0)) + colors[0] = spu.sample_texture(texcoords[0]); + if (spu_extract(mask, 1)) + colors[1] = spu.sample_texture(texcoords[1]); + if (spu_extract(mask, 2)) + colors[2] = spu.sample_texture(texcoords[2]); + if (spu_extract(mask, 3)) + colors[3] = spu.sample_texture(texcoords[3]); + } + else { + /* simple shading */ + eval_coeff(1, (float) x, (float) y, colors); + } + +#if 1 + if (spu.blend.blend_enable) + blend_quad(ix % TILE_SIZE, iy % TILE_SIZE, colors); +#endif + + if (spu_extract(mask, 0)) + spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0], shuffle); + if (spu_extract(mask, 1)) + spu.ctile.ui[iy][ix+1] = spu_pack_color_shuffle(colors[1], shuffle); + if (spu_extract(mask, 2)) + spu.ctile.ui[iy+1][ix] = spu_pack_color_shuffle(colors[2], shuffle); + if (spu_extract(mask, 3)) + spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(colors[3], shuffle); + +#if 0 + /* SIMD_Z with swizzled color buffer (someday) */ + vector unsigned int uicolors = *((vector unsigned int *) &colors); + spu.ctile.ui4[iy/2][ix/2] = spu_sel(spu.ctile.ui4[iy/2][ix/2], uicolors, mask); +#endif + } + +#endif +} + + +/** + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. + */ +static INLINE int block( int x ) +{ + return x & ~1; +} + + +/** + * Compute mask which indicates which pixels in the 2x2 quad are actually inside + * the triangle's bounds. + * The mask is a uint4 vector and each element will be 0 or 0xffffffff. + */ +static INLINE mask_t calculate_mask( int x ) +{ + /* This is a little tricky. + * Use & instead of && to avoid branches. + * Use negation to convert true/false to ~0/0 values. + */ + mask_t mask; + mask = spu_insert(-((x >= setup.span.left[0]) & (x < setup.span.right[0])), mask, 0); + mask = spu_insert(-((x+1 >= setup.span.left[0]) & (x+1 < setup.span.right[0])), mask, 1); + mask = spu_insert(-((x >= setup.span.left[1]) & (x < setup.span.right[1])), mask, 2); + mask = spu_insert(-((x+1 >= setup.span.left[1]) & (x+1 < setup.span.right[1])), mask, 3); + return mask; +} + + +/** + * Render a horizontal span of quads + */ +static void flush_spans( void ) +{ + int minleft, maxright; + int x; + + switch (setup.span.y_flags) { + case 0x3: + /* both odd and even lines written (both quad rows) */ + minleft = MIN2(setup.span.left[0], setup.span.left[1]); + maxright = MAX2(setup.span.right[0], setup.span.right[1]); + break; + + case 0x1: + /* only even line written (quad top row) */ + minleft = setup.span.left[0]; + maxright = setup.span.right[0]; + break; + + case 0x2: + /* only odd line written (quad bottom row) */ + minleft = setup.span.left[1]; + maxright = setup.span.right[1]; + break; + + default: + return; + } + + + /* OK, we're very likely to need the tile data now. + * clear or finish waiting if needed. + */ + if (spu.cur_ctile_status == TILE_STATUS_GETTING) { + /* wait for mfc_get() to complete */ + //printf("SPU: %u: waiting for ctile\n", spu.init.id); + wait_on_mask(1 << TAG_READ_TILE_COLOR); + spu.cur_ctile_status = TILE_STATUS_CLEAN; + } + else if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { + //printf("SPU %u: clearing C tile %u, %u\n", spu.init.id, setup.tx, setup.ty); + clear_c_tile(&spu.ctile); + spu.cur_ctile_status = TILE_STATUS_DIRTY; + } + ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED); + + if (spu.depth_stencil.depth.enabled) { + if (spu.cur_ztile_status == TILE_STATUS_GETTING) { + /* wait for mfc_get() to complete */ + //printf("SPU: %u: waiting for ztile\n", spu.init.id); + wait_on_mask(1 << TAG_READ_TILE_Z); + spu.cur_ztile_status = TILE_STATUS_CLEAN; + } + else if (spu.cur_ztile_status == TILE_STATUS_CLEAR) { + //printf("SPU %u: clearing Z tile %u, %u\n", spu.init.id, setup.tx, setup.ty); + clear_z_tile(&spu.ztile); + spu.cur_ztile_status = TILE_STATUS_DIRTY; + } + ASSERT(spu.cur_ztile_status != TILE_STATUS_DEFINED); + } + + /* XXX this loop could be moved into the above switch cases and + * calculate_mask() could be simplified a bit... + */ + for (x = block(minleft); x <= block(maxright); x += 2) { +#if 1 + emit_quad( x, setup.span.y, calculate_mask( x ) ); +#endif + } + + setup.span.y = 0; + setup.span.y_flags = 0; + setup.span.right[0] = 0; + setup.span.right[1] = 0; +} + +#if DEBUG_VERTS +static void print_vertex(const struct vertex_header *v) +{ + int i; + fprintf(stderr, "Vertex: (%p)\n", v); + for (i = 0; i < setup.quad.nr_attrs; i++) { + fprintf(stderr, " %d: %f %f %f %f\n", i, + v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); + } +} +#endif + + +static boolean setup_sort_vertices(const struct vertex_header *v0, + const struct vertex_header *v1, + const struct vertex_header *v2) +{ + +#if DEBUG_VERTS + fprintf(stderr, "Triangle:\n"); + print_vertex(v0); + print_vertex(v1); + print_vertex(v2); +#endif + + setup.vprovoke = v2; + + /* determine bottom to top order of vertices */ + { + float y0 = spu_extract(v0->data[0], 1); + float y1 = spu_extract(v1->data[0], 1); + float y2 = spu_extract(v2->data[0], 1); + if (y0 <= y1) { + if (y1 <= y2) { + /* y0<=y1<=y2 */ + setup.vmin = v0; + setup.vmid = v1; + setup.vmax = v2; + } + else if (y2 <= y0) { + /* y2<=y0<=y1 */ + setup.vmin = v2; + setup.vmid = v0; + setup.vmax = v1; + } + else { + /* y0<=y2<=y1 */ + setup.vmin = v0; + setup.vmid = v2; + setup.vmax = v1; + } + } + else { + if (y0 <= y2) { + /* y1<=y0<=y2 */ + setup.vmin = v1; + setup.vmid = v0; + setup.vmax = v2; + } + else if (y2 <= y1) { + /* y2<=y1<=y0 */ + setup.vmin = v2; + setup.vmid = v1; + setup.vmax = v0; + } + else { + /* y1<=y2<=y0 */ + setup.vmin = v1; + setup.vmid = v2; + setup.vmax = v0; + } + } + } + + /* Check if triangle is completely outside the tile bounds */ + if (spu_extract(setup.vmin->data[0], 1) > setup.cliprect_maxy) + return FALSE; + if (spu_extract(setup.vmax->data[0], 1) < setup.cliprect_miny) + return FALSE; + if (spu_extract(setup.vmin->data[0], 0) < setup.cliprect_minx && + spu_extract(setup.vmid->data[0], 0) < setup.cliprect_minx && + spu_extract(setup.vmax->data[0], 0) < setup.cliprect_minx) + return FALSE; + if (spu_extract(setup.vmin->data[0], 0) > setup.cliprect_maxx && + spu_extract(setup.vmid->data[0], 0) > setup.cliprect_maxx && + spu_extract(setup.vmax->data[0], 0) > setup.cliprect_maxx) + return FALSE; + + setup.ebot.dx = spu_extract(setup.vmid->data[0], 0) - spu_extract(setup.vmin->data[0], 0); + setup.ebot.dy = spu_extract(setup.vmid->data[0], 1) - spu_extract(setup.vmin->data[0], 1); + setup.emaj.dx = spu_extract(setup.vmax->data[0], 0) - spu_extract(setup.vmin->data[0], 0); + setup.emaj.dy = spu_extract(setup.vmax->data[0], 1) - spu_extract(setup.vmin->data[0], 1); + setup.etop.dx = spu_extract(setup.vmax->data[0], 0) - spu_extract(setup.vmid->data[0], 0); + setup.etop.dy = spu_extract(setup.vmax->data[0], 1) - spu_extract(setup.vmid->data[0], 1); + + /* + * Compute triangle's area. Use 1/area to compute partial + * derivatives of attributes later. + * + * The area will be the same as prim->det, but the sign may be + * different depending on how the vertices get sorted above. + * + * To determine whether the primitive is front or back facing we + * use the prim->det value because its sign is correct. + */ + { + const float area = (setup.emaj.dx * setup.ebot.dy - + setup.ebot.dx * setup.emaj.dy); + + setup.oneoverarea = 1.0f / area; + /* + _mesa_printf("%s one-over-area %f area %f det %f\n", + __FUNCTION__, setup.oneoverarea, area, prim->det ); + */ + } + +#if 0 + /* We need to know if this is a front or back-facing triangle for: + * - the GLSL gl_FrontFacing fragment attribute (bool) + * - two-sided stencil test + */ + setup.quad.facing = (prim->det > 0.0) ^ (setup.softpipe->rasterizer->front_winding == PIPE_WINDING_CW); +#endif + + return TRUE; +} + + +/** + * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + * The value value comes from vertex->data[slot]. + * The result will be put into setup.coef[slot].a0. + * \param slot which attribute slot + */ +static INLINE void +const_coeff(uint slot) +{ + setup.coef[slot].dadx.v = (vector float) {0.0, 0.0, 0.0, 0.0}; + setup.coef[slot].dady.v = (vector float) {0.0, 0.0, 0.0, 0.0}; + setup.coef[slot].a0.v = setup.vprovoke->data[slot]; +} + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a triangle. + */ +static INLINE void +tri_linear_coeff(uint slot, uint firstComp, uint lastComp) +{ + uint i; + const float *vmin_d = (float *) &setup.vmin->data[slot]; + const float *vmid_d = (float *) &setup.vmid->data[slot]; + const float *vmax_d = (float *) &setup.vmax->data[slot]; + const float x = spu_extract(setup.vmin->data[0], 0) - 0.5f; + const float y = spu_extract(setup.vmin->data[0], 1) - 0.5f; + + for (i = firstComp; i < lastComp; i++) { + float botda = vmid_d[i] - vmin_d[i]; + float majda = vmax_d[i] - vmin_d[i]; + float a = setup.ebot.dy * majda - botda * setup.emaj.dy; + float b = setup.emaj.dx * botda - majda * setup.ebot.dx; + + ASSERT(slot < PIPE_MAX_SHADER_INPUTS); + + setup.coef[slot].dadx.f[i] = a * setup.oneoverarea; + setup.coef[slot].dady.f[i] = b * setup.oneoverarea; + + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + setup.coef[slot].a0.f[i] = (vmin_d[i] - + (setup.coef[slot].dadx.f[i] * x + + setup.coef[slot].dady.f[i] * y)); + } + + /* + _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", + slot, "xyzw"[i], + setup.coef[slot].a0[i], + setup.coef[slot].dadx.f[i], + setup.coef[slot].dady.f[i]); + */ +} + + +/** + * As above, but interp setup all four vector components. + */ +static INLINE void +tri_linear_coeff4(uint slot) +{ + const vector float vmin_d = setup.vmin->data[slot]; + const vector float vmid_d = setup.vmid->data[slot]; + const vector float vmax_d = setup.vmax->data[slot]; + const vector float xxxx = spu_splats(spu_extract(setup.vmin->data[0], 0) - 0.5f); + const vector float yyyy = spu_splats(spu_extract(setup.vmin->data[0], 1) - 0.5f); + + vector float botda = vmid_d - vmin_d; + vector float majda = vmax_d - vmin_d; + + vector float a = spu_sub(spu_mul(spu_splats(setup.ebot.dy), majda), + spu_mul(botda, spu_splats(setup.emaj.dy))); + vector float b = spu_sub(spu_mul(spu_splats(setup.emaj.dx), botda), + spu_mul(majda, spu_splats(setup.ebot.dx))); + + setup.coef[slot].dadx.v = spu_mul(a, spu_splats(setup.oneoverarea)); + setup.coef[slot].dady.v = spu_mul(b, spu_splats(setup.oneoverarea)); + + vector float tempx = spu_mul(setup.coef[slot].dadx.v, xxxx); + vector float tempy = spu_mul(setup.coef[slot].dady.v, yyyy); + + setup.coef[slot].a0.v = spu_sub(vmin_d, spu_add(tempx, tempy)); +} + + + +#if 0 +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a triangle. + * We basically multiply the vertex value by 1/w before computing + * the plane coefficients (a0, dadx, dady). + * Later, when we compute the value at a particular fragment position we'll + * divide the interpolated value by the interpolated W at that fragment. + */ +static void tri_persp_coeff( unsigned slot, + unsigned i ) +{ + /* premultiply by 1/w: + */ + float mina = setup.vmin->data[slot][i] * setup.vmin->data[0][3]; + float mida = setup.vmid->data[slot][i] * setup.vmid->data[0][3]; + float maxa = setup.vmax->data[slot][i] * setup.vmax->data[0][3]; + + float botda = mida - mina; + float majda = maxa - mina; + float a = setup.ebot.dy * majda - botda * setup.emaj.dy; + float b = setup.emaj.dx * botda - majda * setup.ebot.dx; + + /* + printf("tri persp %d,%d: %f %f %f\n", slot, i, + setup.vmin->data[slot][i], + setup.vmid->data[slot][i], + setup.vmax->data[slot][i] + ); + */ + + assert(slot < PIPE_MAX_SHADER_INPUTS); + assert(i <= 3); + + setup.coef[slot].dadx.f[i] = a * setup.oneoverarea; + setup.coef[slot].dady.f[i] = b * setup.oneoverarea; + setup.coef[slot].a0.f[i] = (mina - + (setup.coef[slot].dadx.f[i] * (setup.vmin->data[0][0] - 0.5f) + + setup.coef[slot].dady.f[i] * (setup.vmin->data[0][1] - 0.5f))); +} +#endif + + +/** + * Compute the setup.coef[] array dadx, dady, a0 values. + * Must be called after setup.vmin,vmid,vmax,vprovoke are initialized. + */ +static void setup_tri_coefficients(void) +{ +#if 1 + uint i; + + for (i = 0; i < spu.vertex_info.num_attribs; i++) { + switch (spu.vertex_info.interp_mode[i]) { + case INTERP_NONE: + break; + case INTERP_POS: + /*tri_linear_coeff(i, 2, 3);*/ + /* XXX interp W if PERSPECTIVE... */ + tri_linear_coeff4(i); + break; + case INTERP_CONSTANT: + const_coeff(i); + break; + case INTERP_LINEAR: + tri_linear_coeff4(i); + break; + case INTERP_PERSPECTIVE: + tri_linear_coeff4(i); /* temporary */ + break; + default: + ASSERT(0); + } + } +#else + ASSERT(spu.vertex_info.interp_mode[0] == INTERP_POS); + ASSERT(spu.vertex_info.interp_mode[1] == INTERP_LINEAR || + spu.vertex_info.interp_mode[1] == INTERP_CONSTANT); + tri_linear_coeff(0, 2, 3); /* slot 0, z */ + tri_linear_coeff(1, 0, 4); /* slot 1, color */ +#endif +} + + +static void setup_tri_edges(void) +{ + float vmin_x = spu_extract(setup.vmin->data[0], 0) + 0.5f; + float vmid_x = spu_extract(setup.vmid->data[0], 0) + 0.5f; + + float vmin_y = spu_extract(setup.vmin->data[0], 1) - 0.5f; + float vmid_y = spu_extract(setup.vmid->data[0], 1) - 0.5f; + float vmax_y = spu_extract(setup.vmax->data[0], 1) - 0.5f; + + setup.emaj.sy = CEILF(vmin_y); + setup.emaj.lines = (int) CEILF(vmax_y - setup.emaj.sy); + setup.emaj.dxdy = setup.emaj.dx / setup.emaj.dy; + setup.emaj.sx = vmin_x + (setup.emaj.sy - vmin_y) * setup.emaj.dxdy; + + setup.etop.sy = CEILF(vmid_y); + setup.etop.lines = (int) CEILF(vmax_y - setup.etop.sy); + setup.etop.dxdy = setup.etop.dx / setup.etop.dy; + setup.etop.sx = vmid_x + (setup.etop.sy - vmid_y) * setup.etop.dxdy; + + setup.ebot.sy = CEILF(vmin_y); + setup.ebot.lines = (int) CEILF(vmid_y - setup.ebot.sy); + setup.ebot.dxdy = setup.ebot.dx / setup.ebot.dy; + setup.ebot.sx = vmin_x + (setup.ebot.sy - vmin_y) * setup.ebot.dxdy; +} + + +/** + * Render the upper or lower half of a triangle. + * Scissoring/cliprect is applied here too. + */ +static void subtriangle( struct edge *eleft, + struct edge *eright, + unsigned lines ) +{ + const int minx = setup.cliprect_minx; + const int maxx = setup.cliprect_maxx; + const int miny = setup.cliprect_miny; + const int maxy = setup.cliprect_maxy; + int y, start_y, finish_y; + int sy = (int)eleft->sy; + + ASSERT((int)eleft->sy == (int) eright->sy); + + /* clip top/bottom */ + start_y = sy; + finish_y = sy + lines; + + if (start_y < miny) + start_y = miny; + + if (finish_y > maxy) + finish_y = maxy; + + start_y -= sy; + finish_y -= sy; + + /* + _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + */ + + for (y = start_y; y < finish_y; y++) { + + /* avoid accumulating adds as floats don't have the precision to + * accurately iterate large triangle edges that way. luckily we + * can just multiply these days. + * + * this is all drowned out by the attribute interpolation anyway. + */ + int left = (int)(eleft->sx + y * eleft->dxdy); + int right = (int)(eright->sx + y * eright->dxdy); + + /* clip left/right */ + if (left < minx) + left = minx; + if (right > maxx) + right = maxx; + + if (left < right) { + int _y = sy + y; + if (block(_y) != setup.span.y) { + flush_spans(); + setup.span.y = block(_y); + } + + setup.span.left[_y&1] = left; + setup.span.right[_y&1] = right; + setup.span.y_flags |= 1<<(_y&1); + } + } + + + /* save the values so that emaj can be restarted: + */ + eleft->sx += lines * eleft->dxdy; + eright->sx += lines * eright->dxdy; + eleft->sy += lines; + eright->sy += lines; +} + + +/** + * Draw triangle into tile at (tx, ty) (tile coords) + * The tile data should have already been fetched. + */ +boolean +tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) +{ + setup.tx = tx; + setup.ty = ty; + + /* set clipping bounds to tile bounds */ + setup.cliprect_minx = tx * TILE_SIZE; + setup.cliprect_miny = ty * TILE_SIZE; + setup.cliprect_maxx = (tx + 1) * TILE_SIZE; + setup.cliprect_maxy = (ty + 1) * TILE_SIZE; + + if (!setup_sort_vertices((struct vertex_header *) v0, + (struct vertex_header *) v1, + (struct vertex_header *) v2)) { + return FALSE; /* totally clipped */ + } + + setup_tri_coefficients(); + setup_tri_edges(); + + setup.span.y = 0; + setup.span.y_flags = 0; + setup.span.right[0] = 0; + setup.span.right[1] = 0; + /* setup.span.z_mode = tri_z_mode( setup.ctx ); */ + + /* init_constant_attribs( setup ); */ + + if (setup.oneoverarea < 0.0) { + /* emaj on left: + */ + subtriangle( &setup.emaj, &setup.ebot, setup.ebot.lines ); + subtriangle( &setup.emaj, &setup.etop, setup.etop.lines ); + } + else { + /* emaj on right: + */ + subtriangle( &setup.ebot, &setup.emaj, setup.ebot.lines ); + subtriangle( &setup.etop, &setup.emaj, setup.etop.lines ); + } + + flush_spans(); + + return TRUE; +} diff --git a/src/gallium/drivers/cell/spu/spu_tri.h b/src/gallium/drivers/cell/spu/spu_tri.h new file mode 100644 index 0000000000..aa694dd7c9 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_tri.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SPU_TRI_H +#define SPU_TRI_H + + +extern boolean +tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty); + + +#endif /* SPU_TRI_H */ diff --git a/src/gallium/drivers/cell/spu/spu_util.c b/src/gallium/drivers/cell/spu/spu_util.c new file mode 100644 index 0000000000..ac373240c1 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_util.c @@ -0,0 +1,165 @@ +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" +//#include "tgsi_build.h" +#include "pipe/tgsi/util/tgsi_util.h" + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->SwizzleX; + case 1: + return reg->SwizzleY; + case 2: + return reg->SwizzleZ; + case 3: + return reg->SwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->ExtSwizzleX; + case 1: + return reg->ExtSwizzleY; + case 2: + return reg->ExtSwizzleZ; + case 3: + return reg->ExtSwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned swizzle; + + /* + * First, calculate the extended swizzle for a given channel. This will give + * us either a channel index into the simple swizzle or a constant 1 or 0. + */ + swizzle = tgsi_util_get_src_register_extswizzle( + ®->SrcRegisterExtSwz, + component ); + + assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); + assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); + assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); + assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); + + /* + * Second, calculate the simple swizzle for the unswizzled channel index. + * Leave the constants intact, they are not affected by the simple swizzle. + */ + if( swizzle <= TGSI_SWIZZLE_W ) { + swizzle = tgsi_util_get_src_register_swizzle( + ®->SrcRegister, + component ); + } + + return swizzle; +} + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->NegateX; + case 1: + return reg->NegateY; + case 2: + return reg->NegateZ; + case 3: + return reg->NegateW; + default: + assert( 0 ); + } + return 0; +} + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ) +{ + switch( component ) { + case 0: + reg->NegateX = negate; + break; + case 1: + reg->NegateY = negate; + break; + case 2: + reg->NegateZ = negate; + break; + case 3: + reg->NegateW = negate; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned sign_mode; + + if( reg->SrcRegisterExtMod.Absolute ) { + /* Consider only the post-abs negation. */ + + if( reg->SrcRegisterExtMod.Negate ) { + sign_mode = TGSI_UTIL_SIGN_SET; + } + else { + sign_mode = TGSI_UTIL_SIGN_CLEAR; + } + } + else { + /* Accumulate the three negations. */ + + unsigned negate; + + negate = reg->SrcRegister.Negate; + if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { + negate = !negate; + } + if( reg->SrcRegisterExtMod.Negate ) { + negate = !negate; + } + + if( negate ) { + sign_mode = TGSI_UTIL_SIGN_TOGGLE; + } + else { + sign_mode = TGSI_UTIL_SIGN_KEEP; + } + } + + return sign_mode; +} diff --git a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c new file mode 100644 index 0000000000..45e3c26c00 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c @@ -0,0 +1,673 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include +#include + +#include "pipe/p_util.h" +#include "pipe/p_state.h" +#include "pipe/p_shader_tokens.h" +#include "spu_exec.h" +#include "spu_vertex_shader.h" +#include "spu_main.h" + +#define CACHE_NAME attribute +#define CACHED_TYPE qword +#define CACHE_TYPE CACHE_TYPE_RO +#define CACHE_SET_TAGID(set) TAG_VERTEX_BUFFER +#define CACHE_LOG2NNWAY 2 +#define CACHE_LOG2NSETS 6 +#include + +/* Yes folks, this is ugly. + */ +#undef CACHE_NWAY +#undef CACHE_NSETS +#define CACHE_NAME attribute +#define CACHE_NWAY 4 +#define CACHE_NSETS (1U << 6) + + +#define DRAW_DBG 0 + +static const qword fetch_shuffle_data[] = { + /* Shuffle used by CVT_64_FLOAT + */ + { + 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, + 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, + }, + + /* Shuffle used by CVT_8_USCALED and CVT_8_SSCALED + */ + { + 0x00, 0x80, 0x80, 0x80, 0x01, 0x80, 0x80, 0x80, + 0x02, 0x80, 0x80, 0x80, 0x03, 0x80, 0x80, 0x80, + }, + + /* Shuffle used by CVT_16_USCALED and CVT_16_SSCALED + */ + { + 0x00, 0x01, 0x80, 0x80, 0x02, 0x03, 0x80, 0x80, + 0x04, 0x05, 0x80, 0x80, 0x06, 0x07, 0x80, 0x80, + }, + + /* High value shuffle used by trans4x4. + */ + { + 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, + 0x04, 0x05, 0x06, 0x07, 0x14, 0x15, 0x16, 0x17 + }, + + /* Low value shuffle used by trans4x4. + */ + { + 0x08, 0x09, 0x0A, 0x0B, 0x18, 0x19, 0x1A, 0x1B, + 0x0C, 0x0D, 0x0E, 0x0F, 0x1C, 0x1D, 0x1E, 0x1F + } +}; + + +static INLINE void +trans4x4(qword row0, qword row1, qword row2, qword row3, qword *out, + const qword *shuffle) +{ + qword t1 = si_shufb(row0, row2, shuffle[3]); + qword t2 = si_shufb(row0, row2, shuffle[4]); + qword t3 = si_shufb(row1, row3, shuffle[3]); + qword t4 = si_shufb(row1, row3, shuffle[4]); + + out[0] = si_shufb(t1, t3, shuffle[3]); + out[1] = si_shufb(t1, t3, shuffle[4]); + out[2] = si_shufb(t2, t4, shuffle[3]); + out[3] = si_shufb(t2, t4, shuffle[4]); +} + + +/** + * Fetch between 1 and 32 bytes from an unaligned address + */ +static INLINE void +fetch_unaligned(qword *dst, unsigned ea, unsigned size) +{ + qword tmp[4]; + const int shift = ea & 0x0f; + const unsigned aligned_start_ea = ea & ~0x0f; + const unsigned aligned_end_ea = (ea + size) & ~0x0f; + const unsigned num_entries = ((aligned_end_ea - aligned_start_ea) / 16) + 1; + unsigned i; + + + if (shift == 0) { + /* Data is already aligned. Fetch directly into the destination buffer. + */ + for (i = 0; i < num_entries; i++) { + dst[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16)); + } + } else { + /* Fetch data from the cache to the local buffer. + */ + for (i = 0; i < num_entries; i++) { + tmp[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16)); + } + + + /* Fix the alignment of the data and write to the destination buffer. + */ + for (i = 0; i < ((size + 15) / 16); i++) { + dst[i] = si_or((qword) spu_slqwbyte(tmp[i], shift), + (qword) spu_rlmaskqwbyte(tmp[i + 1], shift - 16)); + } + } +} + + +#define CVT_32_FLOAT(q, s) (*(q)) + +static INLINE qword +CVT_64_FLOAT(const qword *qw, const qword *shuffle) +{ + qword a = si_frds(qw[0]); + qword b = si_frds(si_rotqbyi(qw[0], 8)); + qword c = si_frds(qw[1]); + qword d = si_frds(si_rotqbyi(qw[1], 8)); + + qword ab = si_shufb(a, b, shuffle[0]); + qword cd = si_shufb(c, d, si_rotqbyi(shuffle[0], 8)); + + return si_or(ab, cd); +} + + +static INLINE qword +CVT_8_USCALED(const qword *qw, const qword *shuffle) +{ + return si_cuflt(si_shufb(*qw, *qw, shuffle[1]), 0); +} + + +static INLINE qword +CVT_16_USCALED(const qword *qw, const qword *shuffle) +{ + return si_cuflt(si_shufb(*qw, *qw, shuffle[2]), 0); +} + + +static INLINE qword +CVT_32_USCALED(const qword *qw, const qword *shuffle) +{ + (void) shuffle; + return si_cuflt(*qw, 0); +} + +static INLINE qword +CVT_8_SSCALED(const qword *qw, const qword *shuffle) +{ + return si_csflt(si_shufb(*qw, *qw, shuffle[1]), 0); +} + + +static INLINE qword +CVT_16_SSCALED(const qword *qw, const qword *shuffle) +{ + return si_csflt(si_shufb(*qw, *qw, shuffle[2]), 0); +} + + +static INLINE qword +CVT_32_SSCALED(const qword *qw, const qword *shuffle) +{ + (void) shuffle; + return si_csflt(*qw, 0); +} + + +static INLINE qword +CVT_8_UNORM(const qword *qw, const qword *shuffle) +{ + const qword scale = (qword) spu_splats(1.0f / 255.0f); + return si_fm(CVT_8_USCALED(qw, shuffle), scale); +} + + +static INLINE qword +CVT_16_UNORM(const qword *qw, const qword *shuffle) +{ + const qword scale = (qword) spu_splats(1.0f / 65535.0f); + return si_fm(CVT_16_USCALED(qw, shuffle), scale); +} + + +static INLINE qword +CVT_32_UNORM(const qword *qw, const qword *shuffle) +{ + const qword scale = (qword) spu_splats(1.0f / 4294967295.0f); + return si_fm(CVT_32_USCALED(qw, shuffle), scale); +} + + +static INLINE qword +CVT_8_SNORM(const qword *qw, const qword *shuffle) +{ + const qword scale = (qword) spu_splats(1.0f / 127.0f); + return si_fm(CVT_8_SSCALED(qw, shuffle), scale); +} + + +static INLINE qword +CVT_16_SNORM(const qword *qw, const qword *shuffle) +{ + const qword scale = (qword) spu_splats(1.0f / 32767.0f); + return si_fm(CVT_16_SSCALED(qw, shuffle), scale); +} + + +static INLINE qword +CVT_32_SNORM(const qword *qw, const qword *shuffle) +{ + const qword scale = (qword) spu_splats(1.0f / 2147483647.0f); + return si_fm(CVT_32_SSCALED(qw, shuffle), scale); +} + +#define SZ_4 si_il(0U) +#define SZ_3 si_fsmbi(0x000f) +#define SZ_2 si_fsmbi(0x00ff) +#define SZ_1 si_fsmbi(0x0fff) + +/** + * Fetch a float[4] vertex attribute from memory, doing format/type + * conversion as needed. + * + * This is probably needed/dupliocated elsewhere, eg format + * conversion, texture sampling etc. + */ +#define FETCH_ATTRIB( NAME, SZ, CVT, N ) \ +static void \ +fetch_##NAME(qword *out, const qword *in, qword defaults, \ + const qword *shuffle) \ +{ \ + qword tmp[4]; \ + \ + tmp[0] = si_selb(CVT(in + (0 * N), shuffle), defaults, SZ); \ + tmp[1] = si_selb(CVT(in + (1 * N), shuffle), defaults, SZ); \ + tmp[2] = si_selb(CVT(in + (2 * N), shuffle), defaults, SZ); \ + tmp[3] = si_selb(CVT(in + (3 * N), shuffle), defaults, SZ); \ + trans4x4(tmp[0], tmp[1], tmp[2], tmp[3], out, shuffle); \ +} + + +FETCH_ATTRIB( R64G64B64A64_FLOAT, SZ_4, CVT_64_FLOAT, 2 ) +FETCH_ATTRIB( R64G64B64_FLOAT, SZ_3, CVT_64_FLOAT, 2 ) +FETCH_ATTRIB( R64G64_FLOAT, SZ_2, CVT_64_FLOAT, 2 ) +FETCH_ATTRIB( R64_FLOAT, SZ_1, CVT_64_FLOAT, 2 ) + +FETCH_ATTRIB( R32G32B32A32_FLOAT, SZ_4, CVT_32_FLOAT, 1 ) +FETCH_ATTRIB( R32G32B32_FLOAT, SZ_3, CVT_32_FLOAT, 1 ) +FETCH_ATTRIB( R32G32_FLOAT, SZ_2, CVT_32_FLOAT, 1 ) +FETCH_ATTRIB( R32_FLOAT, SZ_1, CVT_32_FLOAT, 1 ) + +FETCH_ATTRIB( R32G32B32A32_USCALED, SZ_4, CVT_32_USCALED, 1 ) +FETCH_ATTRIB( R32G32B32_USCALED, SZ_3, CVT_32_USCALED, 1 ) +FETCH_ATTRIB( R32G32_USCALED, SZ_2, CVT_32_USCALED, 1 ) +FETCH_ATTRIB( R32_USCALED, SZ_1, CVT_32_USCALED, 1 ) + +FETCH_ATTRIB( R32G32B32A32_SSCALED, SZ_4, CVT_32_SSCALED, 1 ) +FETCH_ATTRIB( R32G32B32_SSCALED, SZ_3, CVT_32_SSCALED, 1 ) +FETCH_ATTRIB( R32G32_SSCALED, SZ_2, CVT_32_SSCALED, 1 ) +FETCH_ATTRIB( R32_SSCALED, SZ_1, CVT_32_SSCALED, 1 ) + +FETCH_ATTRIB( R32G32B32A32_UNORM, SZ_4, CVT_32_UNORM, 1 ) +FETCH_ATTRIB( R32G32B32_UNORM, SZ_3, CVT_32_UNORM, 1 ) +FETCH_ATTRIB( R32G32_UNORM, SZ_2, CVT_32_UNORM, 1 ) +FETCH_ATTRIB( R32_UNORM, SZ_1, CVT_32_UNORM, 1 ) + +FETCH_ATTRIB( R32G32B32A32_SNORM, SZ_4, CVT_32_SNORM, 1 ) +FETCH_ATTRIB( R32G32B32_SNORM, SZ_3, CVT_32_SNORM, 1 ) +FETCH_ATTRIB( R32G32_SNORM, SZ_2, CVT_32_SNORM, 1 ) +FETCH_ATTRIB( R32_SNORM, SZ_1, CVT_32_SNORM, 1 ) + +FETCH_ATTRIB( R16G16B16A16_USCALED, SZ_4, CVT_16_USCALED, 1 ) +FETCH_ATTRIB( R16G16B16_USCALED, SZ_3, CVT_16_USCALED, 1 ) +FETCH_ATTRIB( R16G16_USCALED, SZ_2, CVT_16_USCALED, 1 ) +FETCH_ATTRIB( R16_USCALED, SZ_1, CVT_16_USCALED, 1 ) + +FETCH_ATTRIB( R16G16B16A16_SSCALED, SZ_4, CVT_16_SSCALED, 1 ) +FETCH_ATTRIB( R16G16B16_SSCALED, SZ_3, CVT_16_SSCALED, 1 ) +FETCH_ATTRIB( R16G16_SSCALED, SZ_2, CVT_16_SSCALED, 1 ) +FETCH_ATTRIB( R16_SSCALED, SZ_1, CVT_16_SSCALED, 1 ) + +FETCH_ATTRIB( R16G16B16A16_UNORM, SZ_4, CVT_16_UNORM, 1 ) +FETCH_ATTRIB( R16G16B16_UNORM, SZ_3, CVT_16_UNORM, 1 ) +FETCH_ATTRIB( R16G16_UNORM, SZ_2, CVT_16_UNORM, 1 ) +FETCH_ATTRIB( R16_UNORM, SZ_1, CVT_16_UNORM, 1 ) + +FETCH_ATTRIB( R16G16B16A16_SNORM, SZ_4, CVT_16_SNORM, 1 ) +FETCH_ATTRIB( R16G16B16_SNORM, SZ_3, CVT_16_SNORM, 1 ) +FETCH_ATTRIB( R16G16_SNORM, SZ_2, CVT_16_SNORM, 1 ) +FETCH_ATTRIB( R16_SNORM, SZ_1, CVT_16_SNORM, 1 ) + +FETCH_ATTRIB( R8G8B8A8_USCALED, SZ_4, CVT_8_USCALED, 1 ) +FETCH_ATTRIB( R8G8B8_USCALED, SZ_3, CVT_8_USCALED, 1 ) +FETCH_ATTRIB( R8G8_USCALED, SZ_2, CVT_8_USCALED, 1 ) +FETCH_ATTRIB( R8_USCALED, SZ_1, CVT_8_USCALED, 1 ) + +FETCH_ATTRIB( R8G8B8A8_SSCALED, SZ_4, CVT_8_SSCALED, 1 ) +FETCH_ATTRIB( R8G8B8_SSCALED, SZ_3, CVT_8_SSCALED, 1 ) +FETCH_ATTRIB( R8G8_SSCALED, SZ_2, CVT_8_SSCALED, 1 ) +FETCH_ATTRIB( R8_SSCALED, SZ_1, CVT_8_SSCALED, 1 ) + +FETCH_ATTRIB( R8G8B8A8_UNORM, SZ_4, CVT_8_UNORM, 1 ) +FETCH_ATTRIB( R8G8B8_UNORM, SZ_3, CVT_8_UNORM, 1 ) +FETCH_ATTRIB( R8G8_UNORM, SZ_2, CVT_8_UNORM, 1 ) +FETCH_ATTRIB( R8_UNORM, SZ_1, CVT_8_UNORM, 1 ) + +FETCH_ATTRIB( R8G8B8A8_SNORM, SZ_4, CVT_8_SNORM, 1 ) +FETCH_ATTRIB( R8G8B8_SNORM, SZ_3, CVT_8_SNORM, 1 ) +FETCH_ATTRIB( R8G8_SNORM, SZ_2, CVT_8_SNORM, 1 ) +FETCH_ATTRIB( R8_SNORM, SZ_1, CVT_8_SNORM, 1 ) + +FETCH_ATTRIB( A8R8G8B8_UNORM, SZ_4, CVT_8_UNORM, 1 ) + + + +static spu_fetch_func get_fetch_func( enum pipe_format format ) +{ + switch (format) { + case PIPE_FORMAT_R64_FLOAT: + return fetch_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return fetch_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return fetch_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return fetch_R64G64B64A64_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + return fetch_R32_FLOAT; + case PIPE_FORMAT_R32G32_FLOAT: + return fetch_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return fetch_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return fetch_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return fetch_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return fetch_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return fetch_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return fetch_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return fetch_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return fetch_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return fetch_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return fetch_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return fetch_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return fetch_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return fetch_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return fetch_R32G32B32A32_SNORM; + + case PIPE_FORMAT_R32_SSCALED: + return fetch_R32_SSCALED; + case PIPE_FORMAT_R32G32_SSCALED: + return fetch_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return fetch_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return fetch_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return fetch_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return fetch_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return fetch_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return fetch_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return fetch_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return fetch_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return fetch_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return fetch_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return fetch_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return fetch_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return fetch_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return fetch_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return fetch_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return fetch_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return fetch_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return fetch_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return fetch_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return fetch_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return fetch_R8G8B8_UNORM; + case PIPE_FORMAT_R8G8B8A8_UNORM: + return fetch_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return fetch_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return fetch_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return fetch_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return fetch_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return fetch_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return fetch_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return fetch_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return fetch_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return fetch_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return fetch_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return fetch_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return fetch_R8G8B8A8_SSCALED; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + return fetch_A8R8G8B8_UNORM; + + case 0: + return NULL; /* not sure why this is needed */ + + default: + assert(0); + return NULL; + } +} + + +static unsigned get_vertex_size( enum pipe_format format ) +{ + switch (format) { + case PIPE_FORMAT_R64_FLOAT: + return 8; + case PIPE_FORMAT_R64G64_FLOAT: + return 2 * 8; + case PIPE_FORMAT_R64G64B64_FLOAT: + return 3 * 8; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return 4 * 8; + + case PIPE_FORMAT_R32_SSCALED: + case PIPE_FORMAT_R32_SNORM: + case PIPE_FORMAT_R32_USCALED: + case PIPE_FORMAT_R32_UNORM: + case PIPE_FORMAT_R32_FLOAT: + return 4; + case PIPE_FORMAT_R32G32_SSCALED: + case PIPE_FORMAT_R32G32_SNORM: + case PIPE_FORMAT_R32G32_USCALED: + case PIPE_FORMAT_R32G32_UNORM: + case PIPE_FORMAT_R32G32_FLOAT: + return 2 * 4; + case PIPE_FORMAT_R32G32B32_SSCALED: + case PIPE_FORMAT_R32G32B32_SNORM: + case PIPE_FORMAT_R32G32B32_USCALED: + case PIPE_FORMAT_R32G32B32_UNORM: + case PIPE_FORMAT_R32G32B32_FLOAT: + return 3 * 4; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + case PIPE_FORMAT_R32G32B32A32_SNORM: + case PIPE_FORMAT_R32G32B32A32_USCALED: + case PIPE_FORMAT_R32G32B32A32_UNORM: + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return 4 * 4; + + case PIPE_FORMAT_R16_SSCALED: + case PIPE_FORMAT_R16_SNORM: + case PIPE_FORMAT_R16_UNORM: + case PIPE_FORMAT_R16_USCALED: + return 2; + case PIPE_FORMAT_R16G16_SSCALED: + case PIPE_FORMAT_R16G16_SNORM: + case PIPE_FORMAT_R16G16_USCALED: + case PIPE_FORMAT_R16G16_UNORM: + return 2 * 2; + case PIPE_FORMAT_R16G16B16_SSCALED: + case PIPE_FORMAT_R16G16B16_SNORM: + case PIPE_FORMAT_R16G16B16_USCALED: + case PIPE_FORMAT_R16G16B16_UNORM: + return 3 * 2; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + case PIPE_FORMAT_R16G16B16A16_SNORM: + case PIPE_FORMAT_R16G16B16A16_USCALED: + case PIPE_FORMAT_R16G16B16A16_UNORM: + return 4 * 2; + + case PIPE_FORMAT_R8_SSCALED: + case PIPE_FORMAT_R8_SNORM: + case PIPE_FORMAT_R8_USCALED: + case PIPE_FORMAT_R8_UNORM: + return 1; + case PIPE_FORMAT_R8G8_SSCALED: + case PIPE_FORMAT_R8G8_SNORM: + case PIPE_FORMAT_R8G8_USCALED: + case PIPE_FORMAT_R8G8_UNORM: + return 2 * 1; + case PIPE_FORMAT_R8G8B8_SSCALED: + case PIPE_FORMAT_R8G8B8_SNORM: + case PIPE_FORMAT_R8G8B8_USCALED: + case PIPE_FORMAT_R8G8B8_UNORM: + return 3 * 1; + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R8G8B8A8_SSCALED: + case PIPE_FORMAT_R8G8B8A8_SNORM: + case PIPE_FORMAT_R8G8B8A8_USCALED: + case PIPE_FORMAT_R8G8B8A8_UNORM: + return 4 * 1; + + case 0: + return 0; /* not sure why this is needed */ + + default: + assert(0); + return 0; + } +} + + +/** + * Fetch vertex attributes for 'count' vertices. + */ +static void generic_vertex_fetch(struct spu_vs_context *draw, + struct spu_exec_machine *machine, + const unsigned *elts, + unsigned count) +{ + unsigned nr_attrs = draw->vertex_fetch.nr_attrs; + unsigned attr; + + assert(count <= 4); + +#if DRAW_DBG + printf("SPU: %s count = %u, nr_attrs = %u\n", + __FUNCTION__, count, nr_attrs); +#endif + + /* loop over vertex attributes (vertex shader inputs) + */ + for (attr = 0; attr < nr_attrs; attr++) { + const qword default_values = (qword)(vec_float4){ 0.0, 0.0, 0.0, 1.0 }; + const unsigned pitch = draw->vertex_fetch.pitch[attr]; + const uint64_t src = draw->vertex_fetch.src_ptr[attr]; + const spu_fetch_func fetch = draw->vertex_fetch.fetch[attr]; + unsigned i; + unsigned idx; + const unsigned bytes_per_entry = draw->vertex_fetch.size[attr]; + const unsigned quads_per_entry = (bytes_per_entry + 15) / 16; + qword in[2 * 4]; + + + /* Fetch four attributes for four vertices. + */ + idx = 0; + for (i = 0; i < count; i++) { + const uint64_t addr = src + (elts[i] * pitch); + +#if DRAW_DBG + printf("SPU: fetching = 0x%llx\n", addr); +#endif + + fetch_unaligned(& in[idx], addr, bytes_per_entry); + idx += quads_per_entry; + } + + /* Be nice and zero out any missing vertices. + */ + (void) memset(& in[idx], 0, (8 - idx) * sizeof(qword)); + + + /* Convert all 4 vertices to vectors of float. + */ + (*fetch)(&machine->Inputs[attr].xyzw[0].q, in, default_values, + fetch_shuffle_data); + } +} + + +void spu_update_vertex_fetch( struct spu_vs_context *draw ) +{ + unsigned i; + + + /* Invalidate the vertex cache. + */ + for (i = 0; i < (CACHE_NWAY * CACHE_NSETS); i++) { + CACHELINE_CLEARVALID(i); + } + + + for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { + draw->vertex_fetch.fetch[i] = + get_fetch_func(draw->vertex_fetch.format[i]); + draw->vertex_fetch.size[i] = + get_vertex_size(draw->vertex_fetch.format[i]); + } + + draw->vertex_fetch.fetch_func = generic_vertex_fetch; +} diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.c b/src/gallium/drivers/cell/spu/spu_vertex_shader.c new file mode 100644 index 0000000000..c1cbbb6d1e --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.c @@ -0,0 +1,231 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + * Ian Romanick + */ + +#include + +#include "pipe/p_util.h" +#include "pipe/p_state.h" +#include "pipe/p_shader_tokens.h" +#include "spu_vertex_shader.h" +#include "spu_exec.h" +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_context.h" +#include "pipe/cell/common.h" +#include "spu_main.h" + +static INLINE unsigned +compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<machine; + unsigned int j; + + ALIGN16_DECL(struct spu_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct spu_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); + + machine->Processor = TGSI_PROCESSOR_VERTEX; + + ASSERT_ALIGN16(draw->constants); + machine->Consts = (float (*)[4]) draw->constants; + + machine->Inputs = ALIGN16_ASSIGN(inputs); + machine->Outputs = ALIGN16_ASSIGN(outputs); + + spu_vertex_fetch( draw, machine, elts, count ); + + /* run shader */ + spu_exec_machine_run( machine ); + + + /* store machine results */ + for (j = 0; j < count; j++) { + unsigned slot; + float x, y, z, w; + unsigned char buffer[sizeof(struct vertex_header) + + MAX_VERTEX_SIZE] ALIGN16_ATTRIB; + struct vertex_header *const tmpOut = + (struct vertex_header *) buffer; + const unsigned vert_size = ROUNDUP16(sizeof(struct vertex_header) + + (sizeof(float) * 4 + * draw->num_vs_outputs)); + + mfc_get(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0); + wait_on_mask(1 << TAG_VERTEX_BUFFER); + + + /* Handle attr[0] (position) specially: + * + * XXX: Computing the clipmask should be done in the vertex + * program as a set of DP4 instructions appended to the + * user-provided code. + */ + x = tmpOut->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = tmpOut->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = tmpOut->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = tmpOut->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + tmpOut->clipmask = compute_clipmask(tmpOut->clip, draw->plane, + draw->nr_planes); + tmpOut->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + tmpOut->data[0][0] = x * scale[0] + trans[0]; + tmpOut->data[0][1] = y * scale[1] + trans[1]; + tmpOut->data[0][2] = z * scale[2] + trans[2]; + tmpOut->data[0][3] = w; + + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + tmpOut->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + tmpOut->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + tmpOut->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + tmpOut->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + + mfc_put(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0); + } /* loop over vertices */ +} + + +static void +spu_bind_vertex_shader(struct spu_vs_context *draw, + void *uniforms, + void *planes, + unsigned nr_planes, + unsigned num_outputs + ) +{ + draw->constants = (float (*)[4]) uniforms; + + (void) memcpy(draw->plane, planes, sizeof(float) * 4 * nr_planes); + draw->nr_planes = nr_planes; + draw->num_vs_outputs = num_outputs; + + /* specify the shader to interpret/execute */ + spu_exec_machine_init(&draw->machine, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/, + PIPE_SHADER_VERTEX); +} + + +unsigned char immediates[(sizeof(float) * 4 * TGSI_EXEC_NUM_IMMEDIATES) + 32] + ALIGN16_ATTRIB; + +void +spu_execute_vertex_shader(struct spu_vs_context *draw, + const struct cell_command_vs *vs) +{ + unsigned i; + + const uint64_t immediate_addr = vs->shader.immediates; + const unsigned immediate_size = + ROUNDUP16((sizeof(float) * 4 * vs->shader.num_immediates) + + (immediate_addr & 0x0f)); + + mfc_get(immediates, immediate_addr & ~0x0f, immediate_size, + TAG_VERTEX_BUFFER, 0, 0); + + draw->machine.Instructions = (struct tgsi_full_instruction *) + vs->shader.instructions; + draw->machine.NumInstructions = vs->shader.num_instructions; + + draw->machine.Declarations = (struct tgsi_full_declaration *) + vs->shader.declarations; + draw->machine.NumDeclarations = vs->shader.num_declarations; + + draw->vertex_fetch.nr_attrs = vs->nr_attrs; + + wait_on_mask(1 << TAG_VERTEX_BUFFER); + + (void) memcpy(& draw->machine.Imms, &immediates[immediate_addr & 0x0f], + sizeof(float) * 4 * vs->shader.num_immediates); + + spu_bind_vertex_shader(draw, vs->shader.uniforms, + vs->plane, vs->nr_planes, + vs->shader.num_outputs); + + for (i = 0; i < vs->num_elts; i += 4) { + const unsigned batch_size = MIN2(vs->num_elts - i, 4); + + run_vertex_program(draw, & vs->elts[i], batch_size, &vs->vOut[i]); + } +} diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.h b/src/gallium/drivers/cell/spu/spu_vertex_shader.h new file mode 100644 index 0000000000..b5bf31e67d --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.h @@ -0,0 +1,63 @@ +#ifndef SPU_VERTEX_SHADER_H +#define SPU_VERTEX_SHADER_H + +#include "pipe/p_format.h" +#include "spu_exec.h" + +struct spu_vs_context; + +typedef void (*spu_fetch_func)(qword *out, const qword *in, qword defaults, + const qword *shuffle_data); +typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw, + struct spu_exec_machine *machine, + const unsigned *elts, + unsigned count ); + +struct spu_vs_context { + struct pipe_viewport_state viewport; + + struct { + uint64_t src_ptr[PIPE_ATTRIB_MAX]; + unsigned pitch[PIPE_ATTRIB_MAX]; + unsigned size[PIPE_ATTRIB_MAX]; + enum pipe_format format[PIPE_ATTRIB_MAX]; + unsigned nr_attrs; + boolean dirty; + + spu_fetch_func fetch[PIPE_ATTRIB_MAX]; + spu_full_fetch_func fetch_func; + } vertex_fetch; + + /* Clip derived state: + */ + float plane[12][4]; + unsigned nr_planes; + + struct spu_exec_machine machine; + const float (*constants)[4]; + + unsigned num_vs_outputs; +}; + +extern void spu_update_vertex_fetch(struct spu_vs_context *draw); + +static INLINE void spu_vertex_fetch(struct spu_vs_context *draw, + struct spu_exec_machine *machine, + const unsigned *elts, + unsigned count) +{ + if (draw->vertex_fetch.dirty) { + spu_update_vertex_fetch(draw); + draw->vertex_fetch.dirty = 0; + } + + (*draw->vertex_fetch.fetch_func)(draw, machine, elts, count); +} + +struct cell_command_vs; + +extern void +spu_execute_vertex_shader(struct spu_vs_context *draw, + const struct cell_command_vs *vs); + +#endif /* SPU_VERTEX_SHADER_H */ diff --git a/src/gallium/drivers/cell/spu/spu_ztest.h b/src/gallium/drivers/cell/spu/spu_ztest.h new file mode 100644 index 0000000000..ce8ad00339 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_ztest.h @@ -0,0 +1,135 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Zbuffer/depth test code. + */ + + +#ifndef SPU_ZTEST_H +#define SPU_ZTEST_H + + +#ifdef __SPU__ +#include +#endif + + + +/** + * Perform Z testing for a 16-bit/value Z buffer. + * + * \param zvals vector of four fragment zvalues as floats + * \param zbuf ptr to vector of ushort[8] zbuffer values. Note that this + * contains the Z values for 2 quads, 8 pixels. + * \param x x coordinate of quad (only lsbit is significant) + * \param inMask indicates which fragments in the quad are alive + * \return new mask indicating which fragments are alive after ztest + */ +static INLINE vector unsigned int +spu_z16_test_less(vector float zvals, vector unsigned short *zbuf, + uint x, vector unsigned int inMask) +{ +#define ZERO 0x80 + vector unsigned int zvals_ui4, zbuf_ui4, mask; + + /* convert floats to uints in [0, 65535] */ + zvals_ui4 = spu_convtu(zvals, 32); /* convert to [0, 2^32] */ + zvals_ui4 = spu_rlmask(zvals_ui4, -16); /* right shift 16 */ + + /* XXX this conditional could be removed with a bit of work */ + if (x & 1) { + /* convert zbuffer values from ushorts to uints */ + /* gather lower four ushorts */ + zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, + (vector unsigned int) *zbuf, + ((vector unsigned char) { + ZERO, ZERO, 8, 9, ZERO, ZERO, 10, 11, + ZERO, ZERO, 12, 13, ZERO, ZERO, 14, 15})); + /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ + mask = spu_cmpgt(zbuf_ui4, zvals_ui4); + /* mask &= inMask */ + mask = spu_and(mask, inMask); + /* zbuf = mask ? zval : zbuf */ + zbuf_ui4 = spu_sel(zbuf_ui4, zvals_ui4, mask); + /* convert zbuffer values from uints back to ushorts, preserve lower 4 */ + *zbuf = (vector unsigned short) + spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, + ((vector unsigned char) { + 16, 17, 18, 19, 20, 21, 22, 23, + 2, 3, 6, 7, 10, 11, 14, 15})); + } + else { + /* convert zbuffer values from ushorts to uints */ + /* gather upper four ushorts */ + zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, + (vector unsigned int) *zbuf, + ((vector unsigned char) { + ZERO, ZERO, 0, 1, ZERO, ZERO, 2, 3, + ZERO, ZERO, 4, 5, ZERO, ZERO, 6, 7})); + /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ + mask = spu_cmpgt(zbuf_ui4, zvals_ui4); + /* mask &= inMask */ + mask = spu_and(mask, inMask); + /* zbuf = mask ? zval : zbuf */ + zbuf_ui4 = spu_sel(zbuf_ui4, zvals_ui4, mask); + /* convert zbuffer values from uints back to ushorts, preserve upper 4 */ + *zbuf = (vector unsigned short) + spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, + ((vector unsigned char) { + 2, 3, 6, 7, 10, 11, 14, 15, + 24, 25, 26, 27, 28, 29, 30, 31})); + } + return mask; +#undef ZERO +} + + +/** + * As above, but Zbuffer values as 32-bit uints + */ +static INLINE vector unsigned int +spu_z32_test_less(vector float zvals, vector unsigned int *zbuf_ptr, + vector unsigned int inMask) +{ + vector unsigned int zvals_ui4, mask, zbuf = *zbuf_ptr; + + /* convert floats to uints in [0, 0xffffffff] */ + zvals_ui4 = spu_convtu(zvals, 32); + /* mask = (zbuf < zvals_ui4) ? ~0 : 0 */ + mask = spu_cmpgt(zbuf, zvals_ui4); + /* mask &= inMask */ + mask = spu_and(mask, inMask); + /* zbuf = mask ? zval : zbuf */ + *zbuf_ptr = spu_sel(zbuf, zvals_ui4, mask); + + return mask; +} + + +#endif /* SPU_ZTEST_H */ diff --git a/src/gallium/drivers/failover/Makefile b/src/gallium/drivers/failover/Makefile new file mode 100644 index 0000000000..72d0895c74 --- /dev/null +++ b/src/gallium/drivers/failover/Makefile @@ -0,0 +1,21 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = failover + +DRIVER_SOURCES = \ + fo_state.c \ + fo_state_emit.c \ + fo_context.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c new file mode 100644 index 0000000000..7ce4a7df17 --- /dev/null +++ b/src/gallium/drivers/failover/fo_context.c @@ -0,0 +1,155 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_context.h" + +#include "fo_context.h" +#include "fo_winsys.h" + + + +static void failover_destroy( struct pipe_context *pipe ) +{ + struct failover_context *failover = failover_context( pipe ); + + free( failover ); +} + + + +static boolean failover_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) +{ + struct failover_context *failover = failover_context( pipe ); + + /* If there has been any statechange since last time, try hardware + * rendering again: + */ + if (failover->dirty) { + failover->mode = FO_HW; + } + + /* Try hardware: + */ + if (failover->mode == FO_HW) { + if (!failover->hw->draw_elements( failover->hw, + indexBuffer, + indexSize, + prim, + start, + count )) { + + failover->hw->flush( failover->hw, ~0 ); + failover->mode = FO_SW; + } + } + + /* Possibly try software: + */ + if (failover->mode == FO_SW) { + + if (failover->dirty) + failover_state_emit( failover ); + + failover->sw->draw_elements( failover->sw, + indexBuffer, + indexSize, + prim, + start, + count ); + + /* Be ready to switch back to hardware rendering without an + * intervening flush. Unlikely to be much performance impact to + * this: + */ + failover->sw->flush( failover->sw, ~0 ); + } + + return TRUE; +} + + +static boolean failover_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) +{ + return failover_draw_elements(pipe, NULL, 0, prim, start, count); +} + + + +struct pipe_context *failover_create( struct pipe_context *hw, + struct pipe_context *sw ) +{ + struct failover_context *failover = CALLOC_STRUCT(failover_context); + if (failover == NULL) + return NULL; + + failover->hw = hw; + failover->sw = sw; + failover->pipe.winsys = hw->winsys; + failover->pipe.destroy = failover_destroy; + failover->pipe.is_format_supported = hw->is_format_supported; + failover->pipe.get_name = hw->get_name; + failover->pipe.get_vendor = hw->get_vendor; + failover->pipe.get_param = hw->get_param; + failover->pipe.get_paramf = hw->get_paramf; + + failover->pipe.draw_arrays = failover_draw_arrays; + failover->pipe.draw_elements = failover_draw_elements; + failover->pipe.clear = hw->clear; + + /* No software occlusion fallback (or other optional functionality) + * at this point - if the hardware doesn't support it, don't + * advertise it to the application. + */ + failover->pipe.begin_query = hw->begin_query; + failover->pipe.end_query = hw->end_query; + + failover_init_state_functions( failover ); + +#if 0 + failover->pipe.surface_alloc = hw->surface_alloc; +#endif + failover->pipe.get_tex_surface = hw->get_tex_surface; + + failover->pipe.surface_copy = hw->surface_copy; + failover->pipe.surface_fill = hw->surface_fill; + failover->pipe.texture_create = hw->texture_create; + failover->pipe.texture_release = hw->texture_release; + failover->pipe.flush = hw->flush; + + failover->dirty = 0; + + return &failover->pipe; +} + diff --git a/src/gallium/drivers/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h new file mode 100644 index 0000000000..1dc87291c9 --- /dev/null +++ b/src/gallium/drivers/failover/fo_context.h @@ -0,0 +1,114 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef FO_CONTEXT_H +#define FO_CONTEXT_H + +#include "pipe/p_state.h" +#include "pipe/p_context.h" + + + +#define FO_NEW_VIEWPORT 0x1 +#define FO_NEW_RASTERIZER 0x2 +#define FO_NEW_FRAGMENT_SHADER 0x4 +#define FO_NEW_BLEND 0x8 +#define FO_NEW_CLIP 0x10 +#define FO_NEW_SCISSOR 0x20 +#define FO_NEW_STIPPLE 0x40 +#define FO_NEW_FRAMEBUFFER 0x80 +#define FO_NEW_ALPHA_TEST 0x100 +#define FO_NEW_DEPTH_STENCIL 0x200 +#define FO_NEW_SAMPLER 0x400 +#define FO_NEW_TEXTURE 0x800 +#define FO_NEW_VERTEX 0x2000 +#define FO_NEW_VERTEX_SHADER 0x4000 +#define FO_NEW_BLEND_COLOR 0x8000 +#define FO_NEW_CLEAR_COLOR 0x10000 +#define FO_NEW_VERTEX_BUFFER 0x20000 +#define FO_NEW_VERTEX_ELEMENT 0x40000 + + + +#define FO_HW 0 +#define FO_SW 1 + +struct fo_state { + void *sw_state; + void *hw_state; +}; +struct failover_context { + struct pipe_context pipe; /**< base class */ + + + /* The most recent drawing state as set by the driver: + */ + const struct fo_state *blend; + const struct fo_state *sampler[PIPE_MAX_SAMPLERS]; + const struct fo_state *depth_stencil; + const struct fo_state *rasterizer; + const struct fo_state *fragment_shader; + const struct fo_state *vertex_shader; + + struct pipe_blend_color blend_color; + struct pipe_clip_state clip; + struct pipe_framebuffer_state framebuffer; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + + unsigned dirty; + unsigned dirty_sampler; + unsigned dirty_texture; + unsigned dirty_vertex_buffer; + unsigned dirty_vertex_element; + + + unsigned mode; + struct pipe_context *hw; + struct pipe_context *sw; +}; + + + +void failover_init_state_functions( struct failover_context *failover ); +void failover_state_emit( struct failover_context *failover ); + +static INLINE struct failover_context * +failover_context( struct pipe_context *pipe ) +{ + return (struct failover_context *)pipe; +} + + +#endif /* FO_CONTEXT_H */ diff --git a/src/gallium/drivers/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c new file mode 100644 index 0000000000..0fc5568da1 --- /dev/null +++ b/src/gallium/drivers/failover/fo_state.c @@ -0,0 +1,457 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "fo_context.h" + + +/* This looks like a lot of work at the moment - we're keeping a + * duplicate copy of the state up-to-date. + * + * This can change in two ways: + * - With constant state objects we would only need to save a pointer, + * not the whole object. + * - By adding a callback in the state tracker to re-emit state. The + * state tracker knows the current state already and can re-emit it + * without additional complexity. + * + * This works as a proof-of-concept, but a final version will have + * lower overheads. + */ + + + +static void * +failover_create_blend_state( struct pipe_context *pipe, + const struct pipe_blend_state *blend ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_blend_state(failover->sw, blend); + state->hw_state = failover->hw->create_blend_state(failover->hw, blend); + + return state; +} + +static void +failover_bind_blend_state( struct pipe_context *pipe, + void *blend ) +{ + struct failover_context *failover = failover_context(pipe); + struct fo_state *state = (struct fo_state *)blend; + failover->blend = state; + failover->dirty |= FO_NEW_BLEND; + failover->sw->bind_blend_state( failover->sw, state->sw_state ); + failover->hw->bind_blend_state( failover->hw, state->hw_state ); +} + +static void +failover_delete_blend_state( struct pipe_context *pipe, + void *blend ) +{ + struct fo_state *state = (struct fo_state*)blend; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_blend_state(failover->sw, state->sw_state); + failover->hw->delete_blend_state(failover->hw, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); +} + +static void +failover_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->blend_color = *blend_color; + failover->dirty |= FO_NEW_BLEND_COLOR; + failover->sw->set_blend_color( failover->sw, blend_color ); + failover->hw->set_blend_color( failover->hw, blend_color ); +} + +static void +failover_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->clip = *clip; + failover->dirty |= FO_NEW_CLIP; + failover->sw->set_clip_state( failover->sw, clip ); + failover->hw->set_clip_state( failover->hw, clip ); +} + + +static void * +failover_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_depth_stencil_alpha_state(failover->sw, templ); + state->hw_state = failover->hw->create_depth_stencil_alpha_state(failover->hw, templ); + + return state; +} + +static void +failover_bind_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct failover_context *failover = failover_context(pipe); + struct fo_state *state = (struct fo_state *)depth_stencil; + failover->depth_stencil = state; + failover->dirty |= FO_NEW_DEPTH_STENCIL; + failover->sw->bind_depth_stencil_alpha_state(failover->sw, state->sw_state); + failover->hw->bind_depth_stencil_alpha_state(failover->hw, state->hw_state); +} + +static void +failover_delete_depth_stencil_state(struct pipe_context *pipe, + void *ds) +{ + struct fo_state *state = (struct fo_state*)ds; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_depth_stencil_alpha_state(failover->sw, state->sw_state); + failover->hw->delete_depth_stencil_alpha_state(failover->hw, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); +} + +static void +failover_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *framebuffer) +{ + struct failover_context *failover = failover_context(pipe); + + failover->framebuffer = *framebuffer; + failover->dirty |= FO_NEW_FRAMEBUFFER; + failover->sw->set_framebuffer_state( failover->sw, framebuffer ); + failover->hw->set_framebuffer_state( failover->hw, framebuffer ); +} + + +static void * +failover_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_fs_state(failover->sw, templ); + state->hw_state = failover->hw->create_fs_state(failover->hw, templ); + + return state; +} + +static void +failover_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct failover_context *failover = failover_context(pipe); + struct fo_state *state = (struct fo_state*)fs; + failover->fragment_shader = state; + failover->dirty |= FO_NEW_FRAGMENT_SHADER; + failover->sw->bind_fs_state(failover->sw, state->sw_state); + failover->hw->bind_fs_state(failover->hw, state->hw_state); +} + +static void +failover_delete_fs_state(struct pipe_context *pipe, + void *fs) +{ + struct fo_state *state = (struct fo_state*)fs; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_fs_state(failover->sw, state->sw_state); + failover->hw->delete_fs_state(failover->hw, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); +} + +static void * +failover_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_vs_state(failover->sw, templ); + state->hw_state = failover->hw->create_vs_state(failover->hw, templ); + + return state; +} + +static void +failover_bind_vs_state(struct pipe_context *pipe, + void *vs) +{ + struct failover_context *failover = failover_context(pipe); + + struct fo_state *state = (struct fo_state*)vs; + failover->vertex_shader = state; + failover->dirty |= FO_NEW_VERTEX_SHADER; + failover->sw->bind_vs_state(failover->sw, state->sw_state); + failover->hw->bind_vs_state(failover->hw, state->hw_state); +} + +static void +failover_delete_vs_state(struct pipe_context *pipe, + void *vs) +{ + struct fo_state *state = (struct fo_state*)vs; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_vs_state(failover->sw, state->sw_state); + failover->hw->delete_vs_state(failover->hw, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); +} + +static void +failover_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->poly_stipple = *stipple; + failover->dirty |= FO_NEW_STIPPLE; + failover->sw->set_polygon_stipple( failover->sw, stipple ); + failover->hw->set_polygon_stipple( failover->hw, stipple ); +} + + +static void * +failover_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_rasterizer_state(failover->sw, templ); + state->hw_state = failover->hw->create_rasterizer_state(failover->hw, templ); + + return state; +} + +static void +failover_bind_rasterizer_state(struct pipe_context *pipe, + void *raster) +{ + struct failover_context *failover = failover_context(pipe); + + struct fo_state *state = (struct fo_state*)raster; + failover->rasterizer = state; + failover->dirty |= FO_NEW_RASTERIZER; + failover->sw->bind_rasterizer_state(failover->sw, state->sw_state); + failover->hw->bind_rasterizer_state(failover->hw, state->hw_state); +} + +static void +failover_delete_rasterizer_state(struct pipe_context *pipe, + void *raster) +{ + struct fo_state *state = (struct fo_state*)raster; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_rasterizer_state(failover->sw, state->sw_state); + failover->hw->delete_rasterizer_state(failover->hw, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); +} + + +static void +failover_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->scissor = *scissor; + failover->dirty |= FO_NEW_SCISSOR; + failover->sw->set_scissor_state( failover->sw, scissor ); + failover->hw->set_scissor_state( failover->hw, scissor ); +} + + +static void * +failover_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *templ) +{ + struct fo_state *state = malloc(sizeof(struct fo_state)); + struct failover_context *failover = failover_context(pipe); + + state->sw_state = failover->sw->create_sampler_state(failover->sw, templ); + state->hw_state = failover->hw->create_sampler_state(failover->hw, templ); + + return state; +} + +static void +failover_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct failover_context *failover = failover_context(pipe); + struct fo_state *state = (struct fo_state*)sampler; + failover->sampler[unit] = state; + failover->dirty |= FO_NEW_SAMPLER; + failover->dirty_sampler |= (1<sw->bind_sampler_state(failover->sw, unit, + state->sw_state); + failover->hw->bind_sampler_state(failover->hw, unit, + state->hw_state); +} + +static void +failover_delete_sampler_state(struct pipe_context *pipe, void *sampler) +{ + struct fo_state *state = (struct fo_state*)sampler; + struct failover_context *failover = failover_context(pipe); + + failover->sw->delete_sampler_state(failover->sw, state->sw_state); + failover->hw->delete_sampler_state(failover->hw, state->hw_state); + state->sw_state = 0; + state->hw_state = 0; + free(state); +} + + +static void +failover_set_sampler_texture(struct pipe_context *pipe, + unsigned unit, + struct pipe_texture *texture) +{ + struct failover_context *failover = failover_context(pipe); + + failover->texture[unit] = texture; + failover->dirty |= FO_NEW_TEXTURE; + failover->dirty_texture |= (1<sw->set_sampler_texture( failover->sw, unit, texture ); + failover->hw->set_sampler_texture( failover->hw, unit, texture ); +} + + +static void +failover_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct failover_context *failover = failover_context(pipe); + + failover->viewport = *viewport; + failover->dirty |= FO_NEW_VIEWPORT; + failover->sw->set_viewport_state( failover->sw, viewport ); + failover->hw->set_viewport_state( failover->hw, viewport ); +} + + +static void +failover_set_vertex_buffer(struct pipe_context *pipe, + unsigned unit, + const struct pipe_vertex_buffer *vertex_buffer) +{ + struct failover_context *failover = failover_context(pipe); + + failover->vertex_buffer[unit] = *vertex_buffer; + failover->dirty |= FO_NEW_VERTEX_BUFFER; + failover->dirty_vertex_buffer |= (1<sw->set_vertex_buffer( failover->sw, unit, vertex_buffer ); + failover->hw->set_vertex_buffer( failover->hw, unit, vertex_buffer ); +} + + +static void +failover_set_vertex_element(struct pipe_context *pipe, + unsigned unit, + const struct pipe_vertex_element *vertex_element) +{ + struct failover_context *failover = failover_context(pipe); + + failover->vertex_element[unit] = *vertex_element; + failover->dirty |= FO_NEW_VERTEX_ELEMENT; + failover->dirty_vertex_element |= (1<sw->set_vertex_element( failover->sw, unit, vertex_element ); + failover->hw->set_vertex_element( failover->hw, unit, vertex_element ); +} + +void +failover_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct failover_context *failover = failover_context(pipe); + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + failover->sw->set_constant_buffer(failover->sw, shader, index, buf); + failover->hw->set_constant_buffer(failover->hw, shader, index, buf); +} + + +void +failover_init_state_functions( struct failover_context *failover ) +{ + failover->pipe.create_blend_state = failover_create_blend_state; + failover->pipe.bind_blend_state = failover_bind_blend_state; + failover->pipe.delete_blend_state = failover_delete_blend_state; + failover->pipe.create_sampler_state = failover_create_sampler_state; + failover->pipe.bind_sampler_state = failover_bind_sampler_state; + failover->pipe.delete_sampler_state = failover_delete_sampler_state; + failover->pipe.create_depth_stencil_alpha_state = failover_create_depth_stencil_state; + failover->pipe.bind_depth_stencil_alpha_state = failover_bind_depth_stencil_state; + failover->pipe.delete_depth_stencil_alpha_state = failover_delete_depth_stencil_state; + failover->pipe.create_rasterizer_state = failover_create_rasterizer_state; + failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state; + failover->pipe.delete_rasterizer_state = failover_delete_rasterizer_state; + failover->pipe.create_fs_state = failover_create_fs_state; + failover->pipe.bind_fs_state = failover_bind_fs_state; + failover->pipe.delete_fs_state = failover_delete_fs_state; + failover->pipe.create_vs_state = failover_create_vs_state; + failover->pipe.bind_vs_state = failover_bind_vs_state; + failover->pipe.delete_vs_state = failover_delete_vs_state; + + failover->pipe.set_blend_color = failover_set_blend_color; + failover->pipe.set_clip_state = failover_set_clip_state; + failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; + failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; + failover->pipe.set_scissor_state = failover_set_scissor_state; + failover->pipe.set_sampler_texture = failover_set_sampler_texture; + failover->pipe.set_viewport_state = failover_set_viewport_state; + failover->pipe.set_vertex_buffer = failover_set_vertex_buffer; + failover->pipe.set_vertex_element = failover_set_vertex_element; + failover->pipe.set_constant_buffer = failover_set_constant_buffer; +} diff --git a/src/gallium/drivers/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c new file mode 100644 index 0000000000..c663dd4947 --- /dev/null +++ b/src/gallium/drivers/failover/fo_state_emit.c @@ -0,0 +1,137 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "fo_context.h" + +/* This looks like a lot of work at the moment - we're keeping a + * duplicate copy of the state up-to-date. + * + * This can change in two ways: + * - With constant state objects we would only need to save a pointer, + * not the whole object. + * - By adding a callback in the state tracker to re-emit state. The + * state tracker knows the current state already and can re-emit it + * without additional complexity. + * + * This works as a proof-of-concept, but a final version will have + * lower overheads. + */ + + +/* Bring the software pipe uptodate with current state. + * + * With constant state objects we would probably just send all state + * to both rasterizers all the time??? + */ +void +failover_state_emit( struct failover_context *failover ) +{ + unsigned i; + + if (failover->dirty & FO_NEW_BLEND) + failover->sw->bind_blend_state( failover->sw, + failover->blend->sw_state ); + + if (failover->dirty & FO_NEW_BLEND_COLOR) + failover->sw->set_blend_color( failover->sw, &failover->blend_color ); + + if (failover->dirty & FO_NEW_CLIP) + failover->sw->set_clip_state( failover->sw, &failover->clip ); + + if (failover->dirty & FO_NEW_DEPTH_STENCIL) + failover->sw->bind_depth_stencil_alpha_state( failover->sw, + failover->depth_stencil->sw_state ); + + if (failover->dirty & FO_NEW_FRAMEBUFFER) + failover->sw->set_framebuffer_state( failover->sw, &failover->framebuffer ); + + if (failover->dirty & FO_NEW_FRAGMENT_SHADER) + failover->sw->bind_fs_state( failover->sw, + failover->fragment_shader->sw_state ); + + if (failover->dirty & FO_NEW_VERTEX_SHADER) + failover->sw->bind_vs_state( failover->sw, + failover->vertex_shader->sw_state ); + + if (failover->dirty & FO_NEW_STIPPLE) + failover->sw->set_polygon_stipple( failover->sw, &failover->poly_stipple ); + + if (failover->dirty & FO_NEW_RASTERIZER) + failover->sw->bind_rasterizer_state( failover->sw, + failover->rasterizer->sw_state ); + + if (failover->dirty & FO_NEW_SCISSOR) + failover->sw->set_scissor_state( failover->sw, &failover->scissor ); + + if (failover->dirty & FO_NEW_VIEWPORT) + failover->sw->set_viewport_state( failover->sw, &failover->viewport ); + + if (failover->dirty & FO_NEW_SAMPLER) { + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + if (failover->dirty_sampler & (1<sw->bind_sampler_state( failover->sw, i, + failover->sampler[i]->sw_state ); + } + } + } + + if (failover->dirty & FO_NEW_TEXTURE) { + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + if (failover->dirty_texture & (1<sw->set_sampler_texture( failover->sw, i, + failover->texture[i] ); + } + } + } + + if (failover->dirty & FO_NEW_VERTEX_BUFFER) { + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (failover->dirty_vertex_buffer & (1<sw->set_vertex_buffer( failover->sw, i, + &failover->vertex_buffer[i] ); + } + } + } + + if (failover->dirty & FO_NEW_VERTEX_ELEMENT) { + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (failover->dirty_vertex_element & (1<sw->set_vertex_element( failover->sw, i, + &failover->vertex_element[i] ); + } + } + } + + failover->dirty = 0; + failover->dirty_vertex_element = 0; + failover->dirty_vertex_buffer = 0; + failover->dirty_texture = 0; + failover->dirty_sampler = 0; +} diff --git a/src/gallium/drivers/failover/fo_winsys.h b/src/gallium/drivers/failover/fo_winsys.h new file mode 100644 index 0000000000..a8ce997a1f --- /dev/null +++ b/src/gallium/drivers/failover/fo_winsys.h @@ -0,0 +1,45 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef FO_WINSYS_H +#define FO_WINSYS_H + + +/* This is the interface that failover requires any window system + * hosting it to implement. This is the only include file in failover + * which is public. + */ + + +struct pipe_context; + + +struct pipe_context *failover_create( struct pipe_context *hw, + struct pipe_context *sw ); + + +#endif /* FO_WINSYS_H */ diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile new file mode 100644 index 0000000000..2f91de3afc --- /dev/null +++ b/src/gallium/drivers/i915simple/Makefile @@ -0,0 +1,38 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = i915simple + +DRIVER_SOURCES = \ + i915_blit.c \ + i915_clear.c \ + i915_flush.c \ + i915_context.c \ + i915_context.c \ + i915_debug.c \ + i915_debug_fp.c \ + i915_state.c \ + i915_state_immediate.c \ + i915_state_dynamic.c \ + i915_state_derived.c \ + i915_state_emit.c \ + i915_state_sampler.c \ + i915_strings.c \ + i915_prim_emit.c \ + i915_prim_vbuf.c \ + i915_texture.c \ + i915_fpc_emit.c \ + i915_fpc_translate.c \ + i915_surface.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/i915simple/SConscript b/src/gallium/drivers/i915simple/SConscript new file mode 100644 index 0000000000..f5fb96b995 --- /dev/null +++ b/src/gallium/drivers/i915simple/SConscript @@ -0,0 +1,29 @@ +Import('*') + +env = env.Clone() + +i915simple = env.ConvenienceLibrary( + target = 'i915simple', + source = [ + 'i915_blit.c', + 'i915_clear.c', + 'i915_context.c', + 'i915_debug.c', + 'i915_debug_fp.c', + 'i915_flush.c', + 'i915_fpc_emit.c', + 'i915_fpc_translate.c', + 'i915_prim_emit.c', + 'i915_prim_vbuf.c', + 'i915_state.c', + 'i915_state_derived.c', + 'i915_state_dynamic.c', + 'i915_state_emit.c', + 'i915_state_immediate.c', + 'i915_state_sampler.c', + 'i915_strings.c', + 'i915_surface.c', + 'i915_texture.c', + ]) + +Export('i915simple') diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h new file mode 100644 index 0000000000..fb88cd6db0 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_batch.h @@ -0,0 +1,54 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_BATCH_H +#define I915_BATCH_H + +#include "i915_winsys.h" +#include "i915_debug.h" + +#define BATCH_LOCALS + +#define BEGIN_BATCH( dwords, relocs ) \ + (i915->batch_start = i915->winsys->batch_start( i915->winsys, dwords, relocs )) + +#define OUT_BATCH( dword ) \ + i915->winsys->batch_dword( i915->winsys, dword ) + +#define OUT_RELOC( buf, flags, delta ) \ + i915->winsys->batch_reloc( i915->winsys, buf, flags, delta ) + +#define ADVANCE_BATCH() + +#define FLUSH_BATCH() do { \ + if (0) i915_dump_batchbuffer( i915 ); \ + i915->winsys->batch_flush( i915->winsys ); \ + i915->batch_start = NULL; \ + i915->hardware_dirty = ~0; \ +} while (0) + +#endif diff --git a/src/gallium/drivers/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c new file mode 100644 index 0000000000..db4671ff55 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_blit.c @@ -0,0 +1,162 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "i915_context.h" +#include "i915_winsys.h" +#include "i915_blit.h" +#include "i915_reg.h" +#include "i915_batch.h" + +#define FILE_DEBUG_FLAG DEBUG_BLIT + +void +i915_fill_blit(struct i915_context *i915, + unsigned cpp, + short dst_pitch, + struct pipe_buffer *dst_buffer, + unsigned dst_offset, + short x, short y, + short w, short h, + unsigned color) +{ + unsigned BR13, CMD; + BATCH_LOCALS; + + dst_pitch *= (short) cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); + CMD = XY_COLOR_BLT_CMD; + break; + case 4: + BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); + CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | + XY_COLOR_BLT_WRITE_RGB); + break; + default: + return; + } + +// DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", +// __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); + + + if (!BEGIN_BATCH(6, 1)) { + FLUSH_BATCH(); + assert(BEGIN_BATCH(6, 1)); + } + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((y << 16) | x); + OUT_BATCH(((y + h) << 16) | (x + w)); + OUT_RELOC( dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); + OUT_BATCH(color); + ADVANCE_BATCH(); +} + + +void +i915_copy_blit( struct i915_context *i915, + unsigned do_flip, + unsigned cpp, + short src_pitch, + struct pipe_buffer *src_buffer, + unsigned src_offset, + short dst_pitch, + struct pipe_buffer *dst_buffer, + unsigned dst_offset, + short src_x, short src_y, + short dst_x, short dst_y, + short w, short h ) +{ + unsigned CMD, BR13; + int dst_y2 = dst_y + h; + int dst_x2 = dst_x + w; + BATCH_LOCALS; + + + I915_DBG(i915, + "%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", + __FUNCTION__, + src_buffer, src_pitch, src_offset, src_x, src_y, + dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); + + src_pitch *= (short) cpp; + dst_pitch *= (short) cpp; + + switch (cpp) { + case 1: + case 2: + case 3: + BR13 = (((int) dst_pitch) & 0xffff) | + (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + break; + case 4: + BR13 = + (((int) dst_pitch) & 0xffff) | + (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = + (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + break; + default: + return; + } + + if (dst_y2 < dst_y || + dst_x2 < dst_x) { + return; + } + + /* Hardware can handle negative pitches but loses the ability to do + * proper overlapping blits in that case. We don't really have a + * need for either at this stage. + */ + assert (dst_pitch > 0 && src_pitch > 0); + + + if (!BEGIN_BATCH(8, 2)) { + FLUSH_BATCH(); + assert(BEGIN_BATCH(8, 2)); + } + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((dst_y << 16) | dst_x); + OUT_BATCH((dst_y2 << 16) | dst_x2); + OUT_RELOC(dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); + OUT_BATCH((src_y << 16) | src_x); + OUT_BATCH(((int) src_pitch & 0xffff)); + OUT_RELOC(src_buffer, I915_BUFFER_ACCESS_READ, src_offset); + ADVANCE_BATCH(); +} + + diff --git a/src/gallium/drivers/i915simple/i915_blit.h b/src/gallium/drivers/i915simple/i915_blit.h new file mode 100644 index 0000000000..6e5b44e124 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_blit.h @@ -0,0 +1,55 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_BLIT_H +#define I915_BLIT_H + +#include "i915_context.h" + +extern void i915_copy_blit(struct i915_context *i915, + unsigned do_flip, + unsigned cpp, + short src_pitch, + struct pipe_buffer *src_buffer, + unsigned src_offset, + short dst_pitch, + struct pipe_buffer *dst_buffer, + unsigned dst_offset, + short srcx, short srcy, + short dstx, short dsty, + short w, short h ); + +extern void i915_fill_blit(struct i915_context *i915, + unsigned cpp, + short dst_pitch, + struct pipe_buffer *dst_buffer, + unsigned dst_offset, + short x, short y, + short w, short h, unsigned color); + + +#endif diff --git a/src/gallium/drivers/i915simple/i915_clear.c b/src/gallium/drivers/i915simple/i915_clear.c new file mode 100644 index 0000000000..cde69daacc --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_clear.c @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + */ + + +#include "pipe/p_defines.h" +#include "i915_context.h" +#include "i915_state.h" + + +/** + * Clear the given surface to the specified value. + * No masking, no scissor (clear entire buffer). + */ +void +i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); +} diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c new file mode 100644 index 0000000000..497623a700 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -0,0 +1,320 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_context.h" +#include "i915_winsys.h" +#include "i915_state.h" +#include "i915_batch.h" +#include "i915_texture.h" +#include "i915_reg.h" + +#include "pipe/draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" + + +/** + * Query format support for creating a texture, drawing surface, etc. + * \param format the format to test + * \param type one of PIPE_TEXTURE, PIPE_SURFACE + */ +static boolean +i915_is_format_supported( struct pipe_context *pipe, + enum pipe_format format, uint type ) +{ + static const enum pipe_format tex_supported[] = { + PIPE_FORMAT_R8G8B8A8_UNORM, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_U_L8, + PIPE_FORMAT_U_A8, + PIPE_FORMAT_U_I8, + PIPE_FORMAT_U_A8_L8, + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_NONE /* list terminator */ + }; + static const enum pipe_format surface_supported[] = { + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_S8Z24_UNORM, + /*PIPE_FORMAT_R16G16B16A16_SNORM,*/ + PIPE_FORMAT_NONE /* list terminator */ + }; + const enum pipe_format *list; + uint i; + + switch (type) { + case PIPE_TEXTURE: + list = tex_supported; + break; + case PIPE_SURFACE: + list = surface_supported; + break; + default: + assert(0); + } + + for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) { + if (list[i] == format) + return TRUE; + } + + return FALSE; +} + + +static int +i915_get_param(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 11; /* max 1024x1024 */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 11; /* max 1024x1024 */ + default: + return 0; + } +} + + +static float +i915_get_paramf(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 7.5; + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 4.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; + + default: + return 0; + } +} + + +static void i915_destroy( struct pipe_context *pipe ) +{ + struct i915_context *i915 = i915_context( pipe ); + + draw_destroy( i915->draw ); + + FREE( i915 ); +} + + + + +static boolean +i915_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) +{ + struct i915_context *i915 = i915_context( pipe ); + struct draw_context *draw = i915->draw; + unsigned i; + + if (i915->dirty) + i915_update_derived( i915 ); + + /* + * Map vertex buffers + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (i915->vertex_buffer[i].buffer) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + i915->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); + } + } + /* Map index buffer, if present */ + if (indexBuffer) { + void *mapped_indexes + = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + } + else { + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + + draw_set_mapped_constant_buffer(draw, + i915->current.constants[PIPE_SHADER_VERTEX]); + + /* draw! */ + draw_arrays(i915->draw, prim, start, count); + + /* + * unmap vertex/index buffers + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (i915->vertex_buffer[i].buffer) { + pipe->winsys->buffer_unmap(pipe->winsys, i915->vertex_buffer[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); + } + } + if (indexBuffer) { + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + return TRUE; +} + + +static boolean i915_draw_arrays( struct pipe_context *pipe, + unsigned prim, unsigned start, unsigned count) +{ + return i915_draw_elements(pipe, NULL, 0, prim, start, count); +} + + + +struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, + struct i915_winsys *i915_winsys, + unsigned pci_id ) +{ + struct i915_context *i915; + unsigned is_i945 = 0; + + switch (pci_id) { + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + break; + + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + is_i945 = 1; + break; + + default: + pipe_winsys->printf(pipe_winsys, + "%s: unknown pci id 0x%x, cannot create context\n", + __FUNCTION__, pci_id); + return NULL; + } + + i915 = CALLOC_STRUCT(i915_context); + if (i915 == NULL) + return NULL; + + i915->winsys = i915_winsys; + i915->pipe.winsys = pipe_winsys; + + i915->pipe.destroy = i915_destroy; + i915->pipe.is_format_supported = i915_is_format_supported; + i915->pipe.get_param = i915_get_param; + i915->pipe.get_paramf = i915_get_paramf; + + i915->pipe.clear = i915_clear; + + + i915->pipe.draw_arrays = i915_draw_arrays; + i915->pipe.draw_elements = i915_draw_elements; + + /* + * Create drawing context and plug our rendering stage into it. + */ + i915->draw = draw_create(); + assert(i915->draw); + if (GETENV("I915_VBUF")) { + draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915)); + } + else { + draw_set_rasterize_stage(i915->draw, i915_draw_render_stage(i915)); + } + + i915_init_surface_functions(i915); + i915_init_state_functions(i915); + i915_init_flush_functions(i915); + i915_init_string_functions(i915); + + i915->pci_id = pci_id; + i915->flags.is_i945 = is_i945; + + i915->pipe.texture_create = i915_texture_create; + i915->pipe.texture_release = i915_texture_release; + + i915->dirty = ~0; + i915->hardware_dirty = ~0; + + /* Batch stream debugging is a bit hacked up at the moment: + */ + i915->batch_start = NULL; + + /* + * XXX we could plug GL selection/feedback into the drawing pipeline + * by specifying a different setup/render stage. + */ + + return &i915->pipe; +} + diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h new file mode 100644 index 0000000000..b4ea63c3e7 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -0,0 +1,304 @@ + /************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_CONTEXT_H +#define I915_CONTEXT_H + + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "pipe/draw/draw_vertex.h" + + +#define I915_TEX_UNITS 8 + +#define I915_DYNAMIC_MODES4 0 +#define I915_DYNAMIC_DEPTHSCALE_0 1 /* just the header */ +#define I915_DYNAMIC_DEPTHSCALE_1 2 +#define I915_DYNAMIC_IAB 3 +#define I915_DYNAMIC_BC_0 4 /* just the header */ +#define I915_DYNAMIC_BC_1 5 +#define I915_DYNAMIC_BFO_0 6 +#define I915_DYNAMIC_BFO_1 7 +#define I915_DYNAMIC_STP_0 8 +#define I915_DYNAMIC_STP_1 9 +#define I915_DYNAMIC_SC_ENA_0 10 +#define I915_DYNAMIC_SC_RECT_0 11 +#define I915_DYNAMIC_SC_RECT_1 12 +#define I915_DYNAMIC_SC_RECT_2 13 +#define I915_MAX_DYNAMIC 14 + + +#define I915_IMMEDIATE_S0 0 +#define I915_IMMEDIATE_S1 1 +#define I915_IMMEDIATE_S2 2 +#define I915_IMMEDIATE_S3 3 +#define I915_IMMEDIATE_S4 4 +#define I915_IMMEDIATE_S5 5 +#define I915_IMMEDIATE_S6 6 +#define I915_IMMEDIATE_S7 7 +#define I915_MAX_IMMEDIATE 8 + +/* These must mach the order of LI0_STATE_* bits, as they will be used + * to generate hardware packets: + */ +#define I915_CACHE_STATIC 0 +#define I915_CACHE_DYNAMIC 1 /* handled specially */ +#define I915_CACHE_SAMPLER 2 +#define I915_CACHE_MAP 3 +#define I915_CACHE_PROGRAM 4 +#define I915_CACHE_CONSTANTS 5 +#define I915_MAX_CACHE 6 + +#define I915_MAX_CONSTANT 32 + + + +struct i915_cache_context; + +/* Use to calculate differences between state emitted to hardware and + * current driver-calculated state. + */ +struct i915_state +{ + unsigned immediate[I915_MAX_IMMEDIATE]; + unsigned dynamic[I915_MAX_DYNAMIC]; + + float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; + /** number of constants passed in through a constant buffer */ + uint num_user_constants[PIPE_SHADER_TYPES]; + /** user constants, plus extra constants from shader translation */ + uint num_constants[PIPE_SHADER_TYPES]; + + uint *program; + uint program_len; + + /* texture sampler state */ + unsigned sampler[I915_TEX_UNITS][3]; + unsigned sampler_enable_flags; + unsigned sampler_enable_nr; + + /* texture image buffers */ + unsigned texbuffer[I915_TEX_UNITS][2]; + + /** Describes the current hardware vertex layout */ + struct vertex_info vertex_info; + + unsigned id; /* track lost context events */ +}; + +struct i915_blend_state { + unsigned iab; + unsigned modes4; + unsigned LIS5; + unsigned LIS6; +}; + +struct i915_depth_stencil_state { + unsigned stencil_modes4; + unsigned bfo[2]; + unsigned stencil_LIS5; + unsigned depth_LIS6; +}; + +struct i915_rasterizer_state { + int light_twoside : 1; + unsigned st; + enum interp_mode color_interp; + + unsigned LIS4; + unsigned LIS7; + unsigned sc[1]; + + const struct pipe_rasterizer_state *templ; + + union { float f; unsigned u; } ds[2]; +}; + +struct i915_sampler_state { + unsigned state[3]; + const struct pipe_sampler_state *templ; +}; + + +struct i915_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; +}; + +struct i915_context +{ + struct pipe_context pipe; + struct i915_winsys *winsys; + struct draw_context *draw; + + /* The most recent drawing state as set by the driver: + */ + const struct i915_blend_state *blend; + const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct i915_depth_stencil_state *depth_stencil; + const struct i915_rasterizer_state *rasterizer; + const struct pipe_shader_state *fs; + + struct pipe_blend_color blend_color; + struct pipe_clip_state clip; + struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; + struct pipe_framebuffer_state framebuffer; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct i915_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + + unsigned dirty; + + unsigned *batch_start; + + /** Vertex buffer */ + struct pipe_buffer *vbo; + + struct i915_state current; + unsigned hardware_dirty; + + unsigned debug; + unsigned pci_id; + + struct { + unsigned is_i945:1; + } flags; +}; + +/* A flag for each state_tracker state object: + */ +#define I915_NEW_VIEWPORT 0x1 +#define I915_NEW_RASTERIZER 0x2 +#define I915_NEW_FS 0x4 +#define I915_NEW_BLEND 0x8 +#define I915_NEW_CLIP 0x10 +#define I915_NEW_SCISSOR 0x20 +#define I915_NEW_STIPPLE 0x40 +#define I915_NEW_FRAMEBUFFER 0x80 +#define I915_NEW_ALPHA_TEST 0x100 +#define I915_NEW_DEPTH_STENCIL 0x200 +#define I915_NEW_SAMPLER 0x400 +#define I915_NEW_TEXTURE 0x800 +#define I915_NEW_CONSTANTS 0x1000 +#define I915_NEW_VBO 0x2000 + + +/* Driver's internally generated state flags: + */ +#define I915_NEW_VERTEX_FORMAT 0x10000 + + +/* Dirty flags for hardware emit + */ +#define I915_HW_STATIC (1<winsys->printf( stream->winsys, buffer ); + va_end( args ); +} + + +static boolean debug( struct debug_stream *stream, const char *name, unsigned len ) +{ + unsigned i; + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + + if (len == 0) { + PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]); + assert(0); + return FALSE; + } + + if (stream->print_addresses) + PRINTF(stream, "%08x: ", stream->offset); + + + PRINTF(stream, "%s (%d dwords):\n", name, len); + for (i = 0; i < len; i++) + PRINTF(stream, "\t0x%08x\n", ptr[i]); + PRINTF(stream, "\n"); + + stream->offset += len * sizeof(unsigned); + + return TRUE; +} + + +static const char *get_prim_name( unsigned val ) +{ + switch (val & PRIM3D_MASK) { + case PRIM3D_TRILIST: return "TRILIST"; break; + case PRIM3D_TRISTRIP: return "TRISTRIP"; break; + case PRIM3D_TRISTRIP_RVRSE: return "TRISTRIP_RVRSE"; break; + case PRIM3D_TRIFAN: return "TRIFAN"; break; + case PRIM3D_POLY: return "POLY"; break; + case PRIM3D_LINELIST: return "LINELIST"; break; + case PRIM3D_LINESTRIP: return "LINESTRIP"; break; + case PRIM3D_RECTLIST: return "RECTLIST"; break; + case PRIM3D_POINTLIST: return "POINTLIST"; break; + case PRIM3D_DIB: return "DIB"; break; + case PRIM3D_CLEAR_RECT: return "CLEAR_RECT"; break; + case PRIM3D_ZONE_INIT: return "ZONE_INIT"; break; + default: return "????"; break; + } +} + +static boolean debug_prim( struct debug_stream *stream, const char *name, + boolean dump_floats, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + const char *prim = get_prim_name( ptr[0] ); + unsigned i; + + + + PRINTF(stream, "%s %s (%d dwords):\n", name, prim, len); + PRINTF(stream, "\t0x%08x\n", ptr[0]); + for (i = 1; i < len; i++) { + if (dump_floats) + PRINTF(stream, "\t0x%08x // %f\n", ptr[i], *(float *)&ptr[i]); + else + PRINTF(stream, "\t0x%08x\n", ptr[i]); + } + + + PRINTF(stream, "\n"); + + stream->offset += len * sizeof(unsigned); + + return TRUE; +} + + + + +static boolean debug_program( struct debug_stream *stream, const char *name, unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + + if (len == 0) { + PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]); + assert(0); + return FALSE; + } + + if (stream->print_addresses) + PRINTF(stream, "%08x: ", stream->offset); + + PRINTF(stream, "%s (%d dwords):\n", name, len); + i915_disassemble_program( stream, ptr, len ); + + stream->offset += len * sizeof(unsigned); + return TRUE; +} + + +static boolean debug_chain( struct debug_stream *stream, const char *name, unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned old_offset = stream->offset + len * sizeof(unsigned); + unsigned i; + + PRINTF(stream, "%s (%d dwords):\n", name, len); + for (i = 0; i < len; i++) + PRINTF(stream, "\t0x%08x\n", ptr[i]); + + stream->offset = ptr[1] & ~0x3; + + if (stream->offset < old_offset) + PRINTF(stream, "\n... skipping backwards from 0x%x --> 0x%x ...\n\n", + old_offset, stream->offset ); + else + PRINTF(stream, "\n... skipping from 0x%x --> 0x%x ...\n\n", + old_offset, stream->offset ); + + + return TRUE; +} + + +static boolean debug_variable_length_prim( struct debug_stream *stream ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + const char *prim = get_prim_name( ptr[0] ); + unsigned i, len; + + ushort *idx = (ushort *)(ptr+1); + for (i = 0; idx[i] != 0xffff; i++) + ; + + len = 1+(i+2)/2; + + PRINTF(stream, "3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); + for (i = 0; i < len; i++) + PRINTF(stream, "\t0x%08x\n", ptr[i]); + PRINTF(stream, "\n"); + + stream->offset += len * sizeof(unsigned); + return TRUE; +} + + +static void +BITS( + struct debug_stream *stream, + unsigned dw, + unsigned hi, + unsigned lo, + const char *fmt, + ... ) +{ + va_list args; + char buffer[256]; + unsigned himask = ~0UL >> (31 - (hi)); + + PRINTF(stream, "\t\t "); + + va_start( args, fmt ); + vsprintf( buffer, fmt, args ); + stream->winsys->printf( stream->winsys, buffer ); + va_end( args ); + + PRINTF(stream, ": 0x%x\n", ((dw) & himask) >> (lo)); +} + +#define MBZ( dw, hi, lo) do { \ + unsigned x = (dw) >> (lo); \ + unsigned lomask = (1 << (lo)) - 1; \ + unsigned himask; \ + himask = (1UL << (hi)) - 1; \ + assert ((x & himask & ~lomask) == 0); \ +} while (0) + +static void +FLAG( + struct debug_stream *stream, + unsigned dw, + unsigned bit, + const char *fmt, + ... ) +{ + if (((dw) >> (bit)) & 1) { + va_list args; + char buffer[256]; + + PRINTF(stream, "\t\t "); + + va_start( args, fmt ); + vsprintf( buffer, fmt, args ); + stream->winsys->printf( stream->winsys, buffer ); + va_end( args ); + + PRINTF(stream, "\n"); + } +} + +static boolean debug_load_immediate( struct debug_stream *stream, + const char *name, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned bits = (ptr[0] >> 4) & 0xff; + unsigned j = 0; + + PRINTF(stream, "%s (%d dwords, flags: %x):\n", name, len, bits); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); + + if (bits & (1<<0)) { + PRINTF(stream, "\t LIS0: 0x%08x\n", ptr[j]); + PRINTF(stream, "\t vb address: 0x%08x\n", (ptr[j] & ~0x3)); + BITS(stream, ptr[j], 0, 0, "vb invalidate disable"); + j++; + } + if (bits & (1<<1)) { + PRINTF(stream, "\t LIS1: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 29, 24, "vb dword width"); + BITS(stream, ptr[j], 21, 16, "vb dword pitch"); + BITS(stream, ptr[j], 15, 0, "vb max index"); + j++; + } + if (bits & (1<<2)) { + int i; + PRINTF(stream, "\t LIS2: 0x%08x\n", ptr[j]); + for (i = 0; i < 8; i++) { + unsigned tc = (ptr[j] >> (i * 4)) & 0xf; + if (tc != 0xf) + BITS(stream, tc, 3, 0, "tex coord %d", i); + } + j++; + } + if (bits & (1<<3)) { + PRINTF(stream, "\t LIS3: 0x%08x\n", ptr[j]); + j++; + } + if (bits & (1<<4)) { + PRINTF(stream, "\t LIS4: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 23, "point width"); + BITS(stream, ptr[j], 22, 19, "line width"); + FLAG(stream, ptr[j], 18, "alpha flatshade"); + FLAG(stream, ptr[j], 17, "fog flatshade"); + FLAG(stream, ptr[j], 16, "spec flatshade"); + FLAG(stream, ptr[j], 15, "rgb flatshade"); + BITS(stream, ptr[j], 14, 13, "cull mode"); + FLAG(stream, ptr[j], 12, "vfmt: point width"); + FLAG(stream, ptr[j], 11, "vfmt: specular/fog"); + FLAG(stream, ptr[j], 10, "vfmt: rgba"); + FLAG(stream, ptr[j], 9, "vfmt: depth offset"); + BITS(stream, ptr[j], 8, 6, "vfmt: position (2==xyzw)"); + FLAG(stream, ptr[j], 5, "force dflt diffuse"); + FLAG(stream, ptr[j], 4, "force dflt specular"); + FLAG(stream, ptr[j], 3, "local depth offset enable"); + FLAG(stream, ptr[j], 2, "vfmt: fp32 fog coord"); + FLAG(stream, ptr[j], 1, "sprite point"); + FLAG(stream, ptr[j], 0, "antialiasing"); + j++; + } + if (bits & (1<<5)) { + PRINTF(stream, "\t LIS5: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 28, "rgba write disables"); + FLAG(stream, ptr[j], 27, "force dflt point width"); + FLAG(stream, ptr[j], 26, "last pixel enable"); + FLAG(stream, ptr[j], 25, "global z offset enable"); + FLAG(stream, ptr[j], 24, "fog enable"); + BITS(stream, ptr[j], 23, 16, "stencil ref"); + BITS(stream, ptr[j], 15, 13, "stencil test"); + BITS(stream, ptr[j], 12, 10, "stencil fail op"); + BITS(stream, ptr[j], 9, 7, "stencil pass z fail op"); + BITS(stream, ptr[j], 6, 4, "stencil pass z pass op"); + FLAG(stream, ptr[j], 3, "stencil write enable"); + FLAG(stream, ptr[j], 2, "stencil test enable"); + FLAG(stream, ptr[j], 1, "color dither enable"); + FLAG(stream, ptr[j], 0, "logiop enable"); + j++; + } + if (bits & (1<<6)) { + PRINTF(stream, "\t LIS6: 0x%08x\n", ptr[j]); + FLAG(stream, ptr[j], 31, "alpha test enable"); + BITS(stream, ptr[j], 30, 28, "alpha func"); + BITS(stream, ptr[j], 27, 20, "alpha ref"); + FLAG(stream, ptr[j], 19, "depth test enable"); + BITS(stream, ptr[j], 18, 16, "depth func"); + FLAG(stream, ptr[j], 15, "blend enable"); + BITS(stream, ptr[j], 14, 12, "blend func"); + BITS(stream, ptr[j], 11, 8, "blend src factor"); + BITS(stream, ptr[j], 7, 4, "blend dst factor"); + FLAG(stream, ptr[j], 3, "depth write enable"); + FLAG(stream, ptr[j], 2, "color write enable"); + BITS(stream, ptr[j], 1, 0, "provoking vertex"); + j++; + } + + + PRINTF(stream, "\n"); + + assert(j == len); + + stream->offset += len * sizeof(unsigned); + + return TRUE; +} + + + +static boolean debug_load_indirect( struct debug_stream *stream, + const char *name, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned bits = (ptr[0] >> 8) & 0x3f; + unsigned i, j = 0; + + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); + + for (i = 0; i < 6; i++) { + if (bits & (1<offset += len * sizeof(unsigned); + + return TRUE; +} + +static void BR13( struct debug_stream *stream, + unsigned val ) +{ + PRINTF(stream, "\t0x%08x\n", val); + FLAG(stream, val, 30, "clipping enable"); + BITS(stream, val, 25, 24, "color depth (3==32bpp)"); + BITS(stream, val, 23, 16, "raster op"); + BITS(stream, val, 15, 0, "dest pitch"); +} + + +static void BR22( struct debug_stream *stream, + unsigned val ) +{ + PRINTF(stream, "\t0x%08x\n", val); + BITS(stream, val, 31, 16, "dest y1"); + BITS(stream, val, 15, 0, "dest x1"); +} + +static void BR23( struct debug_stream *stream, + unsigned val ) +{ + PRINTF(stream, "\t0x%08x\n", val); + BITS(stream, val, 31, 16, "dest y2"); + BITS(stream, val, 15, 0, "dest x2"); +} + +static void BR09( struct debug_stream *stream, + unsigned val ) +{ + PRINTF(stream, "\t0x%08x -- dest address\n", val); +} + +static void BR26( struct debug_stream *stream, + unsigned val ) +{ + PRINTF(stream, "\t0x%08x\n", val); + BITS(stream, val, 31, 16, "src y1"); + BITS(stream, val, 15, 0, "src x1"); +} + +static void BR11( struct debug_stream *stream, + unsigned val ) +{ + PRINTF(stream, "\t0x%08x\n", val); + BITS(stream, val, 15, 0, "src pitch"); +} + +static void BR12( struct debug_stream *stream, + unsigned val ) +{ + PRINTF(stream, "\t0x%08x -- src address\n", val); +} + +static void BR16( struct debug_stream *stream, + unsigned val ) +{ + PRINTF(stream, "\t0x%08x -- color\n", val); +} + +static boolean debug_copy_blit( struct debug_stream *stream, + const char *name, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); + + BR13(stream, ptr[j++]); + BR22(stream, ptr[j++]); + BR23(stream, ptr[j++]); + BR09(stream, ptr[j++]); + BR26(stream, ptr[j++]); + BR11(stream, ptr[j++]); + BR12(stream, ptr[j++]); + + stream->offset += len * sizeof(unsigned); + assert(j == len); + return TRUE; +} + +static boolean debug_color_blit( struct debug_stream *stream, + const char *name, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); + + BR13(stream, ptr[j++]); + BR22(stream, ptr[j++]); + BR23(stream, ptr[j++]); + BR09(stream, ptr[j++]); + BR16(stream, ptr[j++]); + + stream->offset += len * sizeof(unsigned); + assert(j == len); + return TRUE; +} + +static boolean debug_modes4( struct debug_stream *stream, + const char *name, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 21, 18, "logicop func"); + FLAG(stream, ptr[j], 17, "stencil test mask modify-enable"); + FLAG(stream, ptr[j], 16, "stencil write mask modify-enable"); + BITS(stream, ptr[j], 15, 8, "stencil test mask"); + BITS(stream, ptr[j], 7, 0, "stencil write mask"); + j++; + + stream->offset += len * sizeof(unsigned); + assert(j == len); + return TRUE; +} + +static boolean debug_map_state( struct debug_stream *stream, + const char *name, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned j = 0; + + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); + + { + PRINTF(stream, "\t0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 15, 0, "map mask"); + j++; + } + + while (j < len) { + { + PRINTF(stream, "\t TMn.0: 0x%08x\n", ptr[j]); + PRINTF(stream, "\t map address: 0x%08x\n", (ptr[j] & ~0x3)); + FLAG(stream, ptr[j], 1, "vertical line stride"); + FLAG(stream, ptr[j], 0, "vertical line stride offset"); + j++; + } + + { + PRINTF(stream, "\t TMn.1: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 21, "height"); + BITS(stream, ptr[j], 20, 10, "width"); + BITS(stream, ptr[j], 9, 7, "surface format"); + BITS(stream, ptr[j], 6, 3, "texel format"); + FLAG(stream, ptr[j], 2, "use fence regs"); + FLAG(stream, ptr[j], 1, "tiled surface"); + FLAG(stream, ptr[j], 0, "tile walk ymajor"); + j++; + } + { + PRINTF(stream, "\t TMn.2: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 21, "dword pitch"); + BITS(stream, ptr[j], 20, 15, "cube face enables"); + BITS(stream, ptr[j], 14, 9, "max lod"); + FLAG(stream, ptr[j], 8, "mip layout right"); + BITS(stream, ptr[j], 7, 0, "depth"); + j++; + } + } + + stream->offset += len * sizeof(unsigned); + assert(j == len); + return TRUE; +} + +static boolean debug_sampler_state( struct debug_stream *stream, + const char *name, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned j = 0; + + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); + + { + PRINTF(stream, "\t0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 15, 0, "sampler mask"); + j++; + } + + while (j < len) { + { + PRINTF(stream, "\t TSn.0: 0x%08x\n", ptr[j]); + FLAG(stream, ptr[j], 31, "reverse gamma"); + FLAG(stream, ptr[j], 30, "planar to packed"); + FLAG(stream, ptr[j], 29, "yuv->rgb"); + BITS(stream, ptr[j], 28, 27, "chromakey index"); + BITS(stream, ptr[j], 26, 22, "base mip level"); + BITS(stream, ptr[j], 21, 20, "mip mode filter"); + BITS(stream, ptr[j], 19, 17, "mag mode filter"); + BITS(stream, ptr[j], 16, 14, "min mode filter"); + BITS(stream, ptr[j], 13, 5, "lod bias (s4.4)"); + FLAG(stream, ptr[j], 4, "shadow enable"); + FLAG(stream, ptr[j], 3, "max-aniso-4"); + BITS(stream, ptr[j], 2, 0, "shadow func"); + j++; + } + + { + PRINTF(stream, "\t TSn.1: 0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 31, 24, "min lod"); + MBZ( ptr[j], 23, 18 ); + FLAG(stream, ptr[j], 17, "kill pixel enable"); + FLAG(stream, ptr[j], 16, "keyed tex filter mode"); + FLAG(stream, ptr[j], 15, "chromakey enable"); + BITS(stream, ptr[j], 14, 12, "tcx wrap mode"); + BITS(stream, ptr[j], 11, 9, "tcy wrap mode"); + BITS(stream, ptr[j], 8, 6, "tcz wrap mode"); + FLAG(stream, ptr[j], 5, "normalized coords"); + BITS(stream, ptr[j], 4, 1, "map (surface) index"); + FLAG(stream, ptr[j], 0, "EAST deinterlacer enable"); + j++; + } + { + PRINTF(stream, "\t TSn.2: 0x%08x (default color)\n", ptr[j]); + j++; + } + } + + stream->offset += len * sizeof(unsigned); + assert(j == len); + return TRUE; +} + +static boolean debug_dest_vars( struct debug_stream *stream, + const char *name, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); + + { + PRINTF(stream, "\t0x%08x\n", ptr[j]); + FLAG(stream, ptr[j], 31, "early classic ztest"); + FLAG(stream, ptr[j], 30, "opengl tex default color"); + FLAG(stream, ptr[j], 29, "bypass iz"); + FLAG(stream, ptr[j], 28, "lod preclamp"); + BITS(stream, ptr[j], 27, 26, "dither pattern"); + FLAG(stream, ptr[j], 25, "linear gamma blend"); + FLAG(stream, ptr[j], 24, "debug dither"); + BITS(stream, ptr[j], 23, 20, "dstorg x"); + BITS(stream, ptr[j], 19, 16, "dstorg y"); + MBZ (ptr[j], 15, 15 ); + BITS(stream, ptr[j], 14, 12, "422 write select"); + BITS(stream, ptr[j], 11, 8, "cbuf format"); + BITS(stream, ptr[j], 3, 2, "zbuf format"); + FLAG(stream, ptr[j], 1, "vert line stride"); + FLAG(stream, ptr[j], 1, "vert line stride offset"); + j++; + } + + stream->offset += len * sizeof(unsigned); + assert(j == len); + return TRUE; +} + +static boolean debug_buf_info( struct debug_stream *stream, + const char *name, + unsigned len ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + int j = 0; + + PRINTF(stream, "%s (%d dwords):\n", name, len); + PRINTF(stream, "\t0x%08x\n", ptr[j++]); + + { + PRINTF(stream, "\t0x%08x\n", ptr[j]); + BITS(stream, ptr[j], 28, 28, "aux buffer id"); + BITS(stream, ptr[j], 27, 24, "buffer id (7=depth, 3=back)"); + FLAG(stream, ptr[j], 23, "use fence regs"); + FLAG(stream, ptr[j], 22, "tiled surface"); + FLAG(stream, ptr[j], 21, "tile walk ymajor"); + MBZ (ptr[j], 20, 14); + BITS(stream, ptr[j], 13, 2, "dword pitch"); + MBZ (ptr[j], 2, 0); + j++; + } + + PRINTF(stream, "\t0x%08x -- buffer base address\n", ptr[j++]); + + stream->offset += len * sizeof(unsigned); + assert(j == len); + return TRUE; +} + +static boolean i915_debug_packet( struct debug_stream *stream ) +{ + unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); + unsigned cmd = *ptr; + + switch (((cmd >> 29) & 0x7)) { + case 0x0: + switch ((cmd >> 23) & 0x3f) { + case 0x0: + return debug(stream, "MI_NOOP", 1); + case 0x3: + return debug(stream, "MI_WAIT_FOR_EVENT", 1); + case 0x4: + return debug(stream, "MI_FLUSH", 1); + case 0xA: + debug(stream, "MI_BATCH_BUFFER_END", 1); + return FALSE; + case 0x22: + return debug(stream, "MI_LOAD_REGISTER_IMM", 3); + case 0x31: + return debug_chain(stream, "MI_BATCH_BUFFER_START", 2); + default: + (void)debug(stream, "UNKNOWN 0x0 case!", 1); + assert(0); + break; + } + break; + case 0x1: + (void) debug(stream, "UNKNOWN 0x1 case!", 1); + assert(0); + break; + case 0x2: + switch ((cmd >> 22) & 0xff) { + case 0x50: + return debug_color_blit(stream, "XY_COLOR_BLT", (cmd & 0xff) + 2); + case 0x53: + return debug_copy_blit(stream, "XY_SRC_COPY_BLT", (cmd & 0xff) + 2); + default: + return debug(stream, "blit command", (cmd & 0xff) + 2); + } + break; + case 0x3: + switch ((cmd >> 24) & 0x1f) { + case 0x6: + return debug(stream, "3DSTATE_ANTI_ALIASING", 1); + case 0x7: + return debug(stream, "3DSTATE_RASTERIZATION_RULES", 1); + case 0x8: + return debug(stream, "3DSTATE_BACKFACE_STENCIL_OPS", 2); + case 0x9: + return debug(stream, "3DSTATE_BACKFACE_STENCIL_MASKS", 1); + case 0xb: + return debug(stream, "3DSTATE_INDEPENDENT_ALPHA_BLEND", 1); + case 0xc: + return debug(stream, "3DSTATE_MODES5", 1); + case 0xd: + return debug_modes4(stream, "3DSTATE_MODES4", 1); + case 0x15: + return debug(stream, "3DSTATE_FOG_COLOR", 1); + case 0x16: + return debug(stream, "3DSTATE_COORD_SET_BINDINGS", 1); + case 0x1c: + /* 3DState16NP */ + switch((cmd >> 19) & 0x1f) { + case 0x10: + return debug(stream, "3DSTATE_SCISSOR_ENABLE", 1); + case 0x11: + return debug(stream, "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE", 1); + default: + (void) debug(stream, "UNKNOWN 0x1c case!", 1); + assert(0); + break; + } + break; + case 0x1d: + /* 3DStateMW */ + switch ((cmd >> 16) & 0xff) { + case 0x0: + return debug_map_state(stream, "3DSTATE_MAP_STATE", (cmd & 0x1f) + 2); + case 0x1: + return debug_sampler_state(stream, "3DSTATE_SAMPLER_STATE", (cmd & 0x1f) + 2); + case 0x4: + return debug_load_immediate(stream, "3DSTATE_LOAD_STATE_IMMEDIATE", (cmd & 0xf) + 2); + case 0x5: + return debug_program(stream, "3DSTATE_PIXEL_SHADER_PROGRAM", (cmd & 0x1ff) + 2); + case 0x6: + return debug(stream, "3DSTATE_PIXEL_SHADER_CONSTANTS", (cmd & 0xff) + 2); + case 0x7: + return debug_load_indirect(stream, "3DSTATE_LOAD_INDIRECT", (cmd & 0xff) + 2); + case 0x80: + return debug(stream, "3DSTATE_DRAWING_RECTANGLE", (cmd & 0xffff) + 2); + case 0x81: + return debug(stream, "3DSTATE_SCISSOR_RECTANGLE", (cmd & 0xffff) + 2); + case 0x83: + return debug(stream, "3DSTATE_SPAN_STIPPLE", (cmd & 0xffff) + 2); + case 0x85: + return debug_dest_vars(stream, "3DSTATE_DEST_BUFFER_VARS", (cmd & 0xffff) + 2); + case 0x88: + return debug(stream, "3DSTATE_CONSTANT_BLEND_COLOR", (cmd & 0xffff) + 2); + case 0x89: + return debug(stream, "3DSTATE_FOG_MODE", (cmd & 0xffff) + 2); + case 0x8e: + return debug_buf_info(stream, "3DSTATE_BUFFER_INFO", (cmd & 0xffff) + 2); + case 0x97: + return debug(stream, "3DSTATE_DEPTH_OFFSET_SCALE", (cmd & 0xffff) + 2); + case 0x98: + return debug(stream, "3DSTATE_DEFAULT_Z", (cmd & 0xffff) + 2); + case 0x99: + return debug(stream, "3DSTATE_DEFAULT_DIFFUSE", (cmd & 0xffff) + 2); + case 0x9a: + return debug(stream, "3DSTATE_DEFAULT_SPECULAR", (cmd & 0xffff) + 2); + case 0x9c: + return debug(stream, "3DSTATE_CLEAR_PARAMETERS", (cmd & 0xffff) + 2); + default: + assert(0); + return 0; + } + break; + case 0x1e: + if (cmd & (1 << 23)) + return debug(stream, "???", (cmd & 0xffff) + 1); + else + return debug(stream, "", 1); + break; + case 0x1f: + if ((cmd & (1 << 23)) == 0) + return debug_prim(stream, "3DPRIM (inline)", 1, (cmd & 0x1ffff) + 2); + else if (cmd & (1 << 17)) + { + if ((cmd & 0xffff) == 0) + return debug_variable_length_prim(stream); + else + return debug_prim(stream, "3DPRIM (indexed)", 0, (((cmd & 0xffff) + 1) / 2) + 1); + } + else + return debug_prim(stream, "3DPRIM (indirect sequential)", 0, 2); + break; + default: + return debug(stream, "", 0); + } + default: + assert(0); + return 0; + } + + assert(0); + return 0; +} + + + +void +i915_dump_batchbuffer( struct i915_context *i915 ) +{ + struct debug_stream stream; + unsigned *start = i915->batch_start; + unsigned *end = i915->winsys->batch_start( i915->winsys, 0, 0 ); + unsigned long bytes = (unsigned long) (end - start) * 4; + boolean done = FALSE; + + stream.offset = 0; + stream.ptr = (char *)start; + stream.print_addresses = 0; + stream.winsys = i915->pipe.winsys; + + if (!start || !end) { + stream.winsys->printf( stream.winsys, "\n\nBATCH: ???\n"); + return; + } + + stream.winsys->printf( stream.winsys, "\n\nBATCH: (%d)\n", bytes / 4); + + while (!done && + stream.offset < bytes) + { + if (!i915_debug_packet( &stream )) + break; + + assert(stream.offset <= bytes && + stream.offset >= 0); + } + + stream.winsys->printf( stream.winsys, "END-BATCH\n\n\n"); +} + + diff --git a/src/gallium/drivers/i915simple/i915_debug.h b/src/gallium/drivers/i915simple/i915_debug.h new file mode 100644 index 0000000000..0bcd094233 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_debug.h @@ -0,0 +1,117 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef I915_DEBUG_H +#define I915_DEBUG_H + +#include + +struct i915_context; + +struct debug_stream +{ + unsigned offset; /* current gtt offset */ + char *ptr; /* pointer to gtt offset zero */ + char *end; /* pointer to gtt offset zero */ + unsigned print_addresses; + struct pipe_winsys *winsys; +}; + + +/* Internal functions + */ +void i915_disassemble_program(struct debug_stream *stream, + const unsigned *program, unsigned sz); + +void i915_print_ureg(const char *msg, unsigned ureg); + + +#define DEBUG_BATCH 0x1 +#define DEBUG_BLIT 0x2 +#define DEBUG_BUFFER 0x4 +#define DEBUG_CONSTANTS 0x8 +#define DEBUG_CONTEXT 0x10 +#define DEBUG_DRAW 0x20 +#define DEBUG_DYNAMIC 0x40 +#define DEBUG_FLUSH 0x80 +#define DEBUG_MAP 0x100 +#define DEBUG_PROGRAM 0x200 +#define DEBUG_REGIONS 0x400 +#define DEBUG_SAMPLER 0x800 +#define DEBUG_STATIC 0x1000 +#define DEBUG_SURFACE 0x2000 +#define DEBUG_WINSYS 0x4000 + +#include "pipe/p_compiler.h" + +#if defined(DEBUG) && defined(FILE_DEBUG_FLAG) + +#include "pipe/p_winsys.h" + +static INLINE void +I915_DBG( + struct i915_context *i915, + const char *fmt, + ... ) +{ + if ((i915)->debug & FILE_DEBUG_FLAG) { + va_list args; + char buffer[256]; + + va_start( args, fmt ); + vsprintf( buffer, fmt, args ); + i915->pipe.winsys->printf( i915->pipe.winsys, buffer ); + va_end( args ); + } +} + +#else + +static INLINE void +I915_DBG( + struct i915_context *i915, + const char *fmt, + ... ) +{ + (void) i915; + (void) fmt; +} + +#endif + + +void i915_dump_batchbuffer( struct i915_context *i915 ); + + + +void i915_debug_init( struct i915_context *i915 ); + + +#endif diff --git a/src/gallium/drivers/i915simple/i915_debug_fp.c b/src/gallium/drivers/i915simple/i915_debug_fp.c new file mode 100644 index 0000000000..ebfdb3d93c --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_debug_fp.c @@ -0,0 +1,366 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "i915_reg.h" +#include "i915_debug.h" +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" + + +static void +PRINTF( + struct debug_stream *stream, + const char *fmt, + ... ) +{ + va_list args; + char buffer[256]; + + va_start( args, fmt ); + vsprintf( buffer, fmt, args ); + stream->winsys->printf( stream->winsys, buffer ); + va_end( args ); +} + + +static const char *opcodes[0x20] = { + "NOP", + "ADD", + "MOV", + "MUL", + "MAD", + "DP2ADD", + "DP3", + "DP4", + "FRC", + "RCP", + "RSQ", + "EXP", + "LOG", + "CMP", + "MIN", + "MAX", + "FLR", + "MOD", + "TRC", + "SGE", + "SLT", + "TEXLD", + "TEXLDP", + "TEXLDB", + "TEXKILL", + "DCL", + "0x1a", + "0x1b", + "0x1c", + "0x1d", + "0x1e", + "0x1f", +}; + + +static const int args[0x20] = { + 0, /* 0 nop */ + 2, /* 1 add */ + 1, /* 2 mov */ + 2, /* 3 m ul */ + 3, /* 4 mad */ + 3, /* 5 dp2add */ + 2, /* 6 dp3 */ + 2, /* 7 dp4 */ + 1, /* 8 frc */ + 1, /* 9 rcp */ + 1, /* a rsq */ + 1, /* b exp */ + 1, /* c log */ + 3, /* d cmp */ + 2, /* e min */ + 2, /* f max */ + 1, /* 10 flr */ + 1, /* 11 mod */ + 1, /* 12 trc */ + 2, /* 13 sge */ + 2, /* 14 slt */ + 1, + 1, + 1, + 1, + 0, + 0, + 0, + 0, + 0, + 0, + 0, +}; + + +static const char *regname[0x8] = { + "R", + "T", + "CONST", + "S", + "OC", + "OD", + "U", + "UNKNOWN", +}; + +static void +print_reg_type_nr(struct debug_stream *stream, unsigned type, unsigned nr) +{ + switch (type) { + case REG_TYPE_T: + switch (nr) { + case T_DIFFUSE: + PRINTF(stream, "T_DIFFUSE"); + return; + case T_SPECULAR: + PRINTF(stream, "T_SPECULAR"); + return; + case T_FOG_W: + PRINTF(stream, "T_FOG_W"); + return; + default: + PRINTF(stream, "T_TEX%d", nr); + return; + } + case REG_TYPE_OC: + if (nr == 0) { + PRINTF(stream, "oC"); + return; + } + break; + case REG_TYPE_OD: + if (nr == 0) { + PRINTF(stream, "oD"); + return; + } + break; + default: + break; + } + + PRINTF(stream, "%s[%d]", regname[type], nr); +} + +#define REG_SWIZZLE_MASK 0x7777 +#define REG_NEGATE_MASK 0x8888 + +#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ + (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ + (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ + (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) + + +static void +print_reg_neg_swizzle(struct debug_stream *stream, unsigned reg) +{ + int i; + + if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && + (reg & REG_NEGATE_MASK) == 0) + return; + + PRINTF(stream, "."); + + for (i = 3; i >= 0; i--) { + if (reg & (1 << ((i * 4) + 3))) + PRINTF(stream, "-"); + + switch ((reg >> (i * 4)) & 0x7) { + case 0: + PRINTF(stream, "x"); + break; + case 1: + PRINTF(stream, "y"); + break; + case 2: + PRINTF(stream, "z"); + break; + case 3: + PRINTF(stream, "w"); + break; + case 4: + PRINTF(stream, "0"); + break; + case 5: + PRINTF(stream, "1"); + break; + default: + PRINTF(stream, "?"); + break; + } + } +} + + +static void +print_src_reg(struct debug_stream *stream, unsigned dword) +{ + unsigned nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; + unsigned type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(stream, type, nr); + print_reg_neg_swizzle(stream, dword); +} + + +static void +print_dest_reg(struct debug_stream *stream, unsigned dword) +{ + unsigned nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; + unsigned type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; + print_reg_type_nr(stream, type, nr); + if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) + return; + PRINTF(stream, "."); + if (dword & A0_DEST_CHANNEL_X) + PRINTF(stream, "x"); + if (dword & A0_DEST_CHANNEL_Y) + PRINTF(stream, "y"); + if (dword & A0_DEST_CHANNEL_Z) + PRINTF(stream, "z"); + if (dword & A0_DEST_CHANNEL_W) + PRINTF(stream, "w"); +} + + +#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) +#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) +#define GET_SRC2_REG(r) (r) + + +static void +print_arith_op(struct debug_stream *stream, + unsigned opcode, const unsigned * program) +{ + if (opcode != A0_NOP) { + print_dest_reg(stream, program[0]); + if (program[0] & A0_DEST_SATURATE) + PRINTF(stream, " = SATURATE "); + else + PRINTF(stream, " = "); + } + + PRINTF(stream, "%s ", opcodes[opcode]); + + print_src_reg(stream, GET_SRC0_REG(program[0], program[1])); + if (args[opcode] == 1) { + PRINTF(stream, "\n"); + return; + } + + PRINTF(stream, ", "); + print_src_reg(stream, GET_SRC1_REG(program[1], program[2])); + if (args[opcode] == 2) { + PRINTF(stream, "\n"); + return; + } + + PRINTF(stream, ", "); + print_src_reg(stream, GET_SRC2_REG(program[2])); + PRINTF(stream, "\n"); + return; +} + + +static void +print_tex_op(struct debug_stream *stream, + unsigned opcode, const unsigned * program) +{ + print_dest_reg(stream, program[0] | A0_DEST_CHANNEL_ALL); + PRINTF(stream, " = "); + + PRINTF(stream, "%s ", opcodes[opcode]); + + PRINTF(stream, "S[%d],", program[0] & T0_SAMPLER_NR_MASK); + + print_reg_type_nr(stream, + (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & + REG_TYPE_MASK, + (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); + PRINTF(stream, "\n"); +} + +static void +print_texkil_op(struct debug_stream *stream, + unsigned opcode, const unsigned * program) +{ + PRINTF(stream, "TEXKIL "); + + print_reg_type_nr(stream, + (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & + REG_TYPE_MASK, + (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); + PRINTF(stream, "\n"); +} + +static void +print_dcl_op(struct debug_stream *stream, + unsigned opcode, const unsigned * program) +{ + PRINTF(stream, "%s ", opcodes[opcode]); + print_dest_reg(stream, + program[0] | A0_DEST_CHANNEL_ALL); + PRINTF(stream, "\n"); +} + + +void +i915_disassemble_program(struct debug_stream *stream, + const unsigned * program, unsigned sz) +{ + unsigned size = program[0] & 0x1ff; + unsigned i; + + PRINTF(stream, "\t\tBEGIN\n"); + + assert(size + 2 == sz); + + program++; + for (i = 1; i < sz; i += 3, program += 3) { + unsigned opcode = program[0] & (0x1f << 24); + + PRINTF(stream, "\t\t"); + + if ((int) opcode >= A0_NOP && opcode <= A0_SLT) + print_arith_op(stream, opcode >> 24, program); + else if (opcode >= T0_TEXLD && opcode < T0_TEXKILL) + print_tex_op(stream, opcode >> 24, program); + else if (opcode == T0_TEXKILL) + print_texkil_op(stream, opcode >> 24, program); + else if (opcode == D0_DCL) + print_dcl_op(stream, opcode >> 24, program); + else + PRINTF(stream, "Unknown opcode 0x%x\n", opcode); + } + + PRINTF(stream, "\t\tEND\n\n"); +} + + diff --git a/src/gallium/drivers/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c new file mode 100644 index 0000000000..3c2069b827 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_flush.c @@ -0,0 +1,81 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "i915_context.h" +#include "i915_reg.h" +#include "i915_batch.h" + + +/** + * In future we may want a fence-like interface instead of finish. + */ +static void i915_flush( struct pipe_context *pipe, + unsigned flags ) +{ + struct i915_context *i915 = i915_context(pipe); + + /* Do we need to emit an MI_FLUSH command to flush the hardware + * caches? + */ + if (flags & (PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE)) { + unsigned flush = MI_FLUSH; + + if (!(flags & PIPE_FLUSH_RENDER_CACHE)) + flush |= INHIBIT_FLUSH_RENDER_CACHE; + + if (flags & PIPE_FLUSH_TEXTURE_CACHE) + flush |= FLUSH_MAP_CACHE; + + if (!BEGIN_BATCH(1, 0)) { + FLUSH_BATCH(); + assert(BEGIN_BATCH(1, 0)); + } + OUT_BATCH( flush ); + ADVANCE_BATCH(); + } + + /* If there are no flags, just flush pending commands to hardware: + */ + FLUSH_BATCH(); + + if (flags & PIPE_FLUSH_WAIT) { + i915->winsys->batch_finish(i915->winsys); + } +} + + + +void i915_init_flush_functions( struct i915_context *i915 ) +{ + i915->pipe.flush = i915_flush; +} diff --git a/src/gallium/drivers/i915simple/i915_fpc.h b/src/gallium/drivers/i915simple/i915_fpc.h new file mode 100644 index 0000000000..8c7b68aefb --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_fpc.h @@ -0,0 +1,213 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef I915_FPC_H +#define I915_FPC_H + +#include "pipe/p_util.h" + +#include "i915_context.h" +#include "i915_reg.h" + + + +#define I915_PROGRAM_SIZE 192 + + + +/** + * Program translation state + */ +struct i915_fp_compile { + const struct pipe_shader_state *shader; + + struct vertex_info *vertex_info; + + uint declarations[I915_PROGRAM_SIZE]; + uint program[I915_PROGRAM_SIZE]; + + uint input_semantic_name[PIPE_MAX_SHADER_INPUTS]; + uint input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + + uint output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; + uint output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; + + /** points into the i915->current.constants array: */ + float (*constants)[4]; + uint num_constants; + uint constant_flags[I915_MAX_CONSTANT]; /**< status of each constant */ + + uint *csr; /**< Cursor, points into program. */ + + uint *decl; /**< Cursor, points into declarations. */ + + uint decl_s; /**< flags for which s regs need to be decl'd */ + uint decl_t; /**< flags for which t regs need to be decl'd */ + + uint temp_flag; /**< Tracks temporary regs which are in use */ + uint utemp_flag; /**< Tracks TYPE_U temporary regs which are in use */ + + uint nr_tex_indirect; + uint nr_tex_insn; + uint nr_alu_insn; + uint nr_decl_insn; + + boolean error; /**< Set if i915_program_error() is called */ + uint wpos_tex; + uint NumNativeInstructions; + uint NumNativeAluInstructions; + uint NumNativeTexInstructions; + uint NumNativeTexIndirections; +}; + + +/* Having zero and one in here makes the definition of swizzle a lot + * easier. + */ +#define UREG_TYPE_SHIFT 29 +#define UREG_NR_SHIFT 24 +#define UREG_CHANNEL_X_NEGATE_SHIFT 23 +#define UREG_CHANNEL_X_SHIFT 20 +#define UREG_CHANNEL_Y_NEGATE_SHIFT 19 +#define UREG_CHANNEL_Y_SHIFT 16 +#define UREG_CHANNEL_Z_NEGATE_SHIFT 15 +#define UREG_CHANNEL_Z_SHIFT 12 +#define UREG_CHANNEL_W_NEGATE_SHIFT 11 +#define UREG_CHANNEL_W_SHIFT 8 +#define UREG_CHANNEL_ZERO_NEGATE_MBZ 5 +#define UREG_CHANNEL_ZERO_SHIFT 4 +#define UREG_CHANNEL_ONE_NEGATE_MBZ 1 +#define UREG_CHANNEL_ONE_SHIFT 0 + +#define UREG_BAD 0xffffffff /* not a valid ureg */ + +#define X SRC_X +#define Y SRC_Y +#define Z SRC_Z +#define W SRC_W +#define ZERO SRC_ZERO +#define ONE SRC_ONE + +/* Construct a ureg: + */ +#define UREG( type, nr ) (((type)<< UREG_TYPE_SHIFT) | \ + ((nr) << UREG_NR_SHIFT) | \ + (X << UREG_CHANNEL_X_SHIFT) | \ + (Y << UREG_CHANNEL_Y_SHIFT) | \ + (Z << UREG_CHANNEL_Z_SHIFT) | \ + (W << UREG_CHANNEL_W_SHIFT) | \ + (ZERO << UREG_CHANNEL_ZERO_SHIFT) | \ + (ONE << UREG_CHANNEL_ONE_SHIFT)) + +#define GET_CHANNEL_SRC( reg, channel ) ((reg<<(channel*4)) & (0xf<<20)) +#define CHANNEL_SRC( src, channel ) (src>>(channel*4)) + +#define GET_UREG_TYPE(reg) (((reg)>>UREG_TYPE_SHIFT)®_TYPE_MASK) +#define GET_UREG_NR(reg) (((reg)>>UREG_NR_SHIFT)®_NR_MASK) + + + +#define UREG_XYZW_CHANNEL_MASK 0x00ffff00 + +/* One neat thing about the UREG representation: + */ +static INLINE int +swizzle(int reg, uint x, uint y, uint z, uint w) +{ + assert(x <= SRC_ONE); + assert(y <= SRC_ONE); + assert(z <= SRC_ONE); + assert(w <= SRC_ONE); + return ((reg & ~UREG_XYZW_CHANNEL_MASK) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, x), 0) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, y), 1) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, z), 2) | + CHANNEL_SRC(GET_CHANNEL_SRC(reg, w), 3)); +} + + + +/*********************************************************************** + * Public interface for the compiler + */ +extern void i915_translate_fragment_program( struct i915_context *i915 ); + + + +extern uint i915_get_temp(struct i915_fp_compile *p); +extern uint i915_get_utemp(struct i915_fp_compile *p); +extern void i915_release_utemps(struct i915_fp_compile *p); + + +extern uint i915_emit_texld(struct i915_fp_compile *p, + uint dest, + uint destmask, + uint sampler, uint coord, uint op); + +extern uint i915_emit_arith(struct i915_fp_compile *p, + uint op, + uint dest, + uint mask, + uint saturate, + uint src0, uint src1, uint src2); + +extern uint i915_emit_decl(struct i915_fp_compile *p, + uint type, uint nr, uint d0_flags); + + +extern uint i915_emit_const1f(struct i915_fp_compile *p, float c0); + +extern uint i915_emit_const2f(struct i915_fp_compile *p, + float c0, float c1); + +extern uint i915_emit_const4fv(struct i915_fp_compile *p, + const float * c); + +extern uint i915_emit_const4f(struct i915_fp_compile *p, + float c0, float c1, + float c2, float c3); + + +/*====================================================================== + * i915_fpc_debug.c + */ +extern void i915_disassemble_program(const uint * program, uint sz); + + +/*====================================================================== + * i915_fpc_translate.c + */ + +extern void +i915_program_error(struct i915_fp_compile *p, const char *msg, ...); + +extern void +i915_translate_fragment_program(struct i915_context *i915); + + +#endif diff --git a/src/gallium/drivers/i915simple/i915_fpc_emit.c b/src/gallium/drivers/i915simple/i915_fpc_emit.c new file mode 100644 index 0000000000..74924ff0a1 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_fpc_emit.c @@ -0,0 +1,375 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_reg.h" +#include "i915_context.h" +#include "i915_fpc.h" + + +#define A0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) +#define D0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) +#define T0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) +#define A0_SRC0( reg ) (((reg)&UREG_MASK)>>UREG_A0_SRC0_SHIFT_LEFT) +#define A1_SRC0( reg ) (((reg)&UREG_MASK)<>UREG_A1_SRC1_SHIFT_LEFT) +#define A2_SRC1( reg ) (((reg)&UREG_MASK)<>UREG_A2_SRC2_SHIFT_LEFT) + +/* These are special, and don't have swizzle/negate bits. + */ +#define T0_SAMPLER( reg ) (GET_UREG_NR(reg)<temp_flag); + if (!bit) { + i915_program_error(p, "i915_get_temp: out of temporaries\n"); + return 0; + } + + p->temp_flag |= 1 << (bit - 1); + return UREG(REG_TYPE_R, (bit - 1)); +} + + +uint +i915_get_utemp(struct i915_fp_compile * p) +{ + int bit = ffs(~p->utemp_flag); + if (!bit) { + i915_program_error(p, "i915_get_utemp: out of temporaries\n"); + return 0; + } + + p->utemp_flag |= 1 << (bit - 1); + return UREG(REG_TYPE_U, (bit - 1)); +} + +void +i915_release_utemps(struct i915_fp_compile *p) +{ + p->utemp_flag = ~0x7; +} + + +uint +i915_emit_decl(struct i915_fp_compile *p, + uint type, uint nr, uint d0_flags) +{ + uint reg = UREG(type, nr); + + if (type == REG_TYPE_T) { + if (p->decl_t & (1 << nr)) + return reg; + + p->decl_t |= (1 << nr); + } + else if (type == REG_TYPE_S) { + if (p->decl_s & (1 << nr)) + return reg; + + p->decl_s |= (1 << nr); + } + else + return reg; + + *(p->decl++) = (D0_DCL | D0_DEST(reg) | d0_flags); + *(p->decl++) = D1_MBZ; + *(p->decl++) = D2_MBZ; + + p->nr_decl_insn++; + return reg; +} + +uint +i915_emit_arith(struct i915_fp_compile * p, + uint op, + uint dest, + uint mask, + uint saturate, uint src0, uint src1, uint src2) +{ + uint c[3]; + uint nr_const = 0; + + assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); + dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest)); + assert(dest); + + if (GET_UREG_TYPE(src0) == REG_TYPE_CONST) + c[nr_const++] = 0; + if (GET_UREG_TYPE(src1) == REG_TYPE_CONST) + c[nr_const++] = 1; + if (GET_UREG_TYPE(src2) == REG_TYPE_CONST) + c[nr_const++] = 2; + + /* Recursively call this function to MOV additional const values + * into temporary registers. Use utemp registers for this - + * currently shouldn't be possible to run out, but keep an eye on + * this. + */ + if (nr_const > 1) { + uint s[3], first, i, old_utemp_flag; + + s[0] = src0; + s[1] = src1; + s[2] = src2; + old_utemp_flag = p->utemp_flag; + + first = GET_UREG_NR(s[c[0]]); + for (i = 1; i < nr_const; i++) { + if (GET_UREG_NR(s[c[i]]) != first) { + uint tmp = i915_get_utemp(p); + + i915_emit_arith(p, A0_MOV, tmp, A0_DEST_CHANNEL_ALL, 0, + s[c[i]], 0, 0); + s[c[i]] = tmp; + } + } + + src0 = s[0]; + src1 = s[1]; + src2 = s[2]; + p->utemp_flag = old_utemp_flag; /* restore */ + } + + *(p->csr++) = (op | A0_DEST(dest) | mask | saturate | A0_SRC0(src0)); + *(p->csr++) = (A1_SRC0(src0) | A1_SRC1(src1)); + *(p->csr++) = (A2_SRC1(src1) | A2_SRC2(src2)); + + p->nr_alu_insn++; + return dest; +} + +uint i915_emit_texld( struct i915_fp_compile *p, + uint dest, + uint destmask, + uint sampler, + uint coord, + uint op ) +{ + uint k = UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord)); + if (coord != k) { + /* No real way to work around this in the general case - need to + * allocate and declare a new temporary register (a utemp won't + * do). Will fallback for now. + */ + i915_program_error(p, "Can't (yet) swizzle TEX arguments"); + assert(0); + return 0; + } + + /* Don't worry about saturate as we only support + */ + if (destmask != A0_DEST_CHANNEL_ALL) { + uint tmp = i915_get_utemp(p); + i915_emit_texld( p, tmp, A0_DEST_CHANNEL_ALL, sampler, coord, op ); + i915_emit_arith( p, A0_MOV, dest, destmask, 0, tmp, 0, 0 ); + return dest; + } + else { + assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); + assert(dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest))); + + if (GET_UREG_TYPE(coord) != REG_TYPE_T) { + p->nr_tex_indirect++; + } + + *(p->csr++) = (op | + T0_DEST( dest ) | + T0_SAMPLER( sampler )); + + *(p->csr++) = T1_ADDRESS_REG( coord ); + *(p->csr++) = T2_MBZ; + + p->nr_tex_insn++; + return dest; + } +} + + +uint +i915_emit_const1f(struct i915_fp_compile * p, float c0) +{ + unsigned reg, idx; + + if (c0 == 0.0) + return swizzle(UREG(REG_TYPE_R, 0), ZERO, ZERO, ZERO, ZERO); + if (c0 == 1.0) + return swizzle(UREG(REG_TYPE_R, 0), ONE, ONE, ONE, ONE); + + for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { + if (p->constant_flags[reg] == I915_CONSTFLAG_PARAM) + continue; + for (idx = 0; idx < 4; idx++) { + if (!(p->constant_flags[reg] & (1 << idx)) || + p->constants[reg][idx] == c0) { + p->constants[reg][idx] = c0; + p->constant_flags[reg] |= 1 << idx; + if (reg + 1 > p->num_constants) + p->num_constants = reg + 1; + return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE); + } + } + } + + i915_program_error(p, "i915_emit_const1f: out of constants\n"); + return 0; +} + +uint +i915_emit_const2f(struct i915_fp_compile * p, float c0, float c1) +{ + unsigned reg, idx; + + if (c0 == 0.0) + return swizzle(i915_emit_const1f(p, c1), ZERO, X, Z, W); + if (c0 == 1.0) + return swizzle(i915_emit_const1f(p, c1), ONE, X, Z, W); + + if (c1 == 0.0) + return swizzle(i915_emit_const1f(p, c0), X, ZERO, Z, W); + if (c1 == 1.0) + return swizzle(i915_emit_const1f(p, c0), X, ONE, Z, W); + + for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { + if (p->constant_flags[reg] == 0xf || + p->constant_flags[reg] == I915_CONSTFLAG_PARAM) + continue; + for (idx = 0; idx < 3; idx++) { + if (!(p->constant_flags[reg] & (3 << idx))) { + p->constants[reg][idx + 0] = c0; + p->constants[reg][idx + 1] = c1; + p->constant_flags[reg] |= 3 << idx; + if (reg + 1 > p->num_constants) + p->num_constants = reg + 1; + return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, ONE); + } + } + } + + i915_program_error(p, "i915_emit_const2f: out of constants\n"); + return 0; +} + + + +uint +i915_emit_const4f(struct i915_fp_compile * p, + float c0, float c1, float c2, float c3) +{ + unsigned reg; + + for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { + if (p->constant_flags[reg] == 0xf && + p->constants[reg][0] == c0 && + p->constants[reg][1] == c1 && + p->constants[reg][2] == c2 && + p->constants[reg][3] == c3) { + return UREG(REG_TYPE_CONST, reg); + } + else if (p->constant_flags[reg] == 0) { + + p->constants[reg][0] = c0; + p->constants[reg][1] = c1; + p->constants[reg][2] = c2; + p->constants[reg][3] = c3; + p->constant_flags[reg] = 0xf; + if (reg + 1 > p->num_constants) + p->num_constants = reg + 1; + return UREG(REG_TYPE_CONST, reg); + } + } + + i915_program_error(p, "i915_emit_const4f: out of constants\n"); + return 0; +} + + +uint +i915_emit_const4fv(struct i915_fp_compile * p, const float * c) +{ + return i915_emit_const4f(p, c[0], c[1], c[2], c[3]); +} + + +#if 00000/*UNUSED*/ +/* Reserve a slot in the constant file for a Mesa state parameter. + * These will later need to be tracked on statechanges, but that is + * done elsewhere. + */ +uint +i915_emit_param4fv(struct i915_fp_compile * p, const float * values) +{ + struct i915_fragment_program *fp = p->fp; + int i; + + for (i = 0; i < fp->nr_params; i++) { + if (fp->param[i].values == values) + return UREG(REG_TYPE_CONST, fp->param[i].reg); + } + + if (p->constants->nr_constants == I915_MAX_CONSTANT || + fp->nr_params == I915_MAX_CONSTANT) { + i915_program_error(p, "i915_emit_param4fv: out of constants\n"); + return 0; + } + + { + int reg = p->constants->nr_constants++; + int i = fp->nr_params++; + + assert (p->constant_flags[reg] == 0); + p->constant_flags[reg] = I915_CONSTFLAG_PARAM; + + fp->param[i].values = values; + fp->param[i].reg = reg; + + return UREG(REG_TYPE_CONST, reg); + } +} +#endif diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c new file mode 100644 index 0000000000..868f0c7e04 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -0,0 +1,1135 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include + +#include "i915_reg.h" +#include "i915_context.h" +#include "i915_fpc.h" + +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" + +#include "pipe/draw/draw_vertex.h" + + +/** + * Simple pass-through fragment shader to use when we don't have + * a real shader (or it fails to compile for some reason). + */ +static unsigned passthrough[] = +{ + _3DSTATE_PIXEL_SHADER_PROGRAM | ((2*3)-1), + + /* declare input color: + */ + (D0_DCL | + (REG_TYPE_T << D0_TYPE_SHIFT) | + (T_DIFFUSE << D0_NR_SHIFT) | + D0_CHANNEL_ALL), + 0, + 0, + + /* move to output color: + */ + (A0_MOV | + (REG_TYPE_OC << A0_DEST_TYPE_SHIFT) | + A0_DEST_CHANNEL_ALL | + (REG_TYPE_T << A0_SRC0_TYPE_SHIFT) | + (T_DIFFUSE << A0_SRC0_NR_SHIFT)), + 0x01230000, /* .xyzw */ + 0 +}; + + +/* 1, -1/3!, 1/5!, -1/7! */ +static const float sin_constants[4] = { 1.0, + -1.0f / (3 * 2 * 1), + 1.0f / (5 * 4 * 3 * 2 * 1), + -1.0f / (7 * 6 * 5 * 4 * 3 * 2 * 1) +}; + +/* 1, -1/2!, 1/4!, -1/6! */ +static const float cos_constants[4] = { 1.0, + -1.0f / (2 * 1), + 1.0f / (4 * 3 * 2 * 1), + -1.0f / (6 * 5 * 4 * 3 * 2 * 1) +}; + + + +/** + * component-wise negation of ureg + */ +static INLINE int +negate(int reg, int x, int y, int z, int w) +{ + /* Another neat thing about the UREG representation */ + return reg ^ (((x & 1) << UREG_CHANNEL_X_NEGATE_SHIFT) | + ((y & 1) << UREG_CHANNEL_Y_NEGATE_SHIFT) | + ((z & 1) << UREG_CHANNEL_Z_NEGATE_SHIFT) | + ((w & 1) << UREG_CHANNEL_W_NEGATE_SHIFT)); +} + + +static void +i915_use_passthrough_shader(struct i915_context *i915) +{ + debug_printf("**** Using i915 pass-through fragment shader\n"); + + i915->current.program = (uint *) MALLOC(sizeof(passthrough)); + if (i915->current.program) { + memcpy(i915->current.program, passthrough, sizeof(passthrough)); + i915->current.program_len = Elements(passthrough); + } + + i915->current.num_constants[PIPE_SHADER_FRAGMENT] = 0; + i915->current.num_user_constants[PIPE_SHADER_FRAGMENT] = 0; +} + + +void +i915_program_error(struct i915_fp_compile *p, const char *msg, ...) +{ + va_list args; + char buffer[1024]; + + debug_printf("i915_program_error: "); + va_start( args, msg ); + vsprintf( buffer, msg, args ); + va_end( args ); + debug_printf(buffer); + debug_printf("\n"); + + p->error = 1; +} + + + +/** + * Construct a ureg for the given source register. Will emit + * constants, apply swizzling and negation as needed. + */ +static uint +src_vector(struct i915_fp_compile *p, + const struct tgsi_full_src_register *source) +{ + uint index = source->SrcRegister.Index; + uint src, sem_name, sem_ind; + + switch (source->SrcRegister.File) { + case TGSI_FILE_TEMPORARY: + if (source->SrcRegister.Index >= I915_MAX_TEMPORARY) { + i915_program_error(p, "Exceeded max temporary reg"); + return 0; + } + src = UREG(REG_TYPE_R, index); + break; + case TGSI_FILE_INPUT: + /* XXX: Packing COL1, FOGC into a single attribute works for + * texenv programs, but will fail for real fragment programs + * that use these attributes and expect them to be a full 4 + * components wide. Could use a texcoord to pass these + * attributes if necessary, but that won't work in the general + * case. + * + * We also use a texture coordinate to pass wpos when possible. + */ + + /* use vertex format info to map a slot number to a VF attrib */ + assert(index < p->vertex_info->num_attribs); + + sem_name = p->input_semantic_name[index]; + sem_ind = p->input_semantic_index[index]; + + switch (sem_name) { + case TGSI_SEMANTIC_POSITION: + debug_printf("SKIP SEM POS\n"); + /* + assert(p->wpos_tex != -1); + src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); + */ + break; + case TGSI_SEMANTIC_COLOR: + if (sem_ind == 0) { + src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); + } + else { + /* secondary color */ + assert(sem_ind == 1); + src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); + src = swizzle(src, X, Y, Z, ONE); + } + break; + case TGSI_SEMANTIC_FOG: + src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); + src = swizzle(src, W, W, W, W); + break; + case TGSI_SEMANTIC_GENERIC: + /* usually a texcoord */ + src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + sem_ind, D0_CHANNEL_ALL); + break; + default: + i915_program_error(p, "Bad source->Index"); + return 0; + } + break; + + case TGSI_FILE_IMMEDIATE: + /* XXX unfinished - need to append immediates onto const buffer */ + /* fall-through */ + case TGSI_FILE_CONSTANT: + src = UREG(REG_TYPE_CONST, index); + break; + + default: + i915_program_error(p, "Bad source->File"); + return 0; + } + + if (source->SrcRegister.Extended) { + src = swizzle(src, + source->SrcRegisterExtSwz.ExtSwizzleX, + source->SrcRegisterExtSwz.ExtSwizzleY, + source->SrcRegisterExtSwz.ExtSwizzleZ, + source->SrcRegisterExtSwz.ExtSwizzleW); + } + else { + src = swizzle(src, + source->SrcRegister.SwizzleX, + source->SrcRegister.SwizzleY, + source->SrcRegister.SwizzleZ, + source->SrcRegister.SwizzleW); + } + + + /* There's both negate-all-components and per-component negation. + * Try to handle both here. + */ + { + int nx = source->SrcRegisterExtSwz.NegateX; + int ny = source->SrcRegisterExtSwz.NegateY; + int nz = source->SrcRegisterExtSwz.NegateZ; + int nw = source->SrcRegisterExtSwz.NegateW; + if (source->SrcRegister.Negate) { + nx = !nx; + ny = !ny; + nz = !nz; + nw = !nw; + } + src = negate(src, nx, ny, nz, nw); + } + + /* no abs() or post-abs negation */ +#if 0 + /* XXX assertions disabled to allow arbfplight.c to run */ + /* XXX enable these assertions, or fix things */ + assert(!source->SrcRegisterExtMod.Absolute); + assert(!source->SrcRegisterExtMod.Negate); +#endif + return src; +} + + +/** + * Construct a ureg for a destination register. + */ +static uint +get_result_vector(struct i915_fp_compile *p, + const struct tgsi_full_dst_register *dest) +{ + switch (dest->DstRegister.File) { + case TGSI_FILE_OUTPUT: + { + uint sem_name = p->output_semantic_name[dest->DstRegister.Index]; + switch (sem_name) { + case TGSI_SEMANTIC_POSITION: + return UREG(REG_TYPE_OD, 0); + case TGSI_SEMANTIC_COLOR: + return UREG(REG_TYPE_OC, 0); + default: + i915_program_error(p, "Bad inst->DstReg.Index/semantics"); + return 0; + } + } + case TGSI_FILE_TEMPORARY: + return UREG(REG_TYPE_R, dest->DstRegister.Index); + default: + i915_program_error(p, "Bad inst->DstReg.File"); + return 0; + } +} + + +/** + * Compute flags for saturation and writemask. + */ +static uint +get_result_flags(const struct tgsi_full_instruction *inst) +{ + const uint writeMask + = inst->FullDstRegisters[0].DstRegister.WriteMask; + uint flags = 0x0; + + if (inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE) + flags |= A0_DEST_SATURATE; + + if (writeMask & TGSI_WRITEMASK_X) + flags |= A0_DEST_CHANNEL_X; + if (writeMask & TGSI_WRITEMASK_Y) + flags |= A0_DEST_CHANNEL_Y; + if (writeMask & TGSI_WRITEMASK_Z) + flags |= A0_DEST_CHANNEL_Z; + if (writeMask & TGSI_WRITEMASK_W) + flags |= A0_DEST_CHANNEL_W; + + return flags; +} + + +/** + * Convert TGSI_TEXTURE_x token to DO_SAMPLE_TYPE_x token + */ +static uint +translate_tex_src_target(struct i915_fp_compile *p, uint tex) +{ + switch (tex) { + case TGSI_TEXTURE_1D: + return D0_SAMPLE_TYPE_2D; + case TGSI_TEXTURE_2D: + return D0_SAMPLE_TYPE_2D; + case TGSI_TEXTURE_RECT: + return D0_SAMPLE_TYPE_2D; + case TGSI_TEXTURE_3D: + return D0_SAMPLE_TYPE_VOLUME; + case TGSI_TEXTURE_CUBE: + return D0_SAMPLE_TYPE_CUBE; + default: + i915_program_error(p, "TexSrc type"); + return 0; + } +} + + +/** + * Generate texel lookup instruction. + */ +static void +emit_tex(struct i915_fp_compile *p, + const struct tgsi_full_instruction *inst, + uint opcode) +{ + uint texture = inst->InstructionExtTexture.Texture; + uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + uint tex = translate_tex_src_target( p, texture ); + uint sampler = i915_emit_decl(p, REG_TYPE_S, unit, tex); + uint coord = src_vector( p, &inst->FullSrcRegisters[0]); + + i915_emit_texld( p, + get_result_vector( p, &inst->FullDstRegisters[0] ), + get_result_flags( inst ), + sampler, + coord, + opcode); +} + + +/** + * Generate a simple arithmetic instruction + * \param opcode the i915 opcode + * \param numArgs the number of input/src arguments + */ +static void +emit_simple_arith(struct i915_fp_compile *p, + const struct tgsi_full_instruction *inst, + uint opcode, uint numArgs) +{ + uint arg1, arg2, arg3; + + assert(numArgs <= 3); + + arg1 = (numArgs < 1) ? 0 : src_vector( p, &inst->FullSrcRegisters[0] ); + arg2 = (numArgs < 2) ? 0 : src_vector( p, &inst->FullSrcRegisters[1] ); + arg3 = (numArgs < 3) ? 0 : src_vector( p, &inst->FullSrcRegisters[2] ); + + i915_emit_arith( p, + opcode, + get_result_vector( p, &inst->FullDstRegisters[0]), + get_result_flags( inst ), 0, + arg1, + arg2, + arg3 ); +} + +#ifndef M_PI +#define M_PI 3.14159265358979323846 +#endif + +/* + * Translate TGSI instruction to i915 instruction. + * + * Possible concerns: + * + * SIN, COS -- could use another taylor step? + * LIT -- results seem a little different to sw mesa + * LOG -- different to mesa on negative numbers, but this is conformant. + */ +static void +i915_translate_instruction(struct i915_fp_compile *p, + const struct tgsi_full_instruction *inst) +{ + uint writemask; + uint src0, src1, src2, flags; + uint tmp = 0; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + i915_emit_arith(p, + A0_MAX, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + src0, negate(src0, 1, 1, 1, 1), 0); + break; + + case TGSI_OPCODE_ADD: + emit_simple_arith(p, inst, A0_ADD, 2); + break; + + case TGSI_OPCODE_CMP: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + src2 = src_vector(p, &inst->FullSrcRegisters[2]); + i915_emit_arith(p, A0_CMP, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), + 0, src0, src2, src1); /* NOTE: order of src2, src1 */ + break; + + case TGSI_OPCODE_COS: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + src0, i915_emit_const1f(p, 1.0f / (float) (M_PI * 2.0)), 0); + + i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); + + /* By choosing different taylor constants, could get rid of this mul: + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + tmp, i915_emit_const1f(p, (float) (M_PI * 2.0)), 0); + + /* + * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 + * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, 1 + * t0 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 + * result = DP4 t0, cos_constants + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XY, 0, + swizzle(tmp, X, X, ONE, ONE), + swizzle(tmp, X, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XYZ, 0, + swizzle(tmp, X, Y, X, ONE), + swizzle(tmp, X, X, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XYZ, 0, + swizzle(tmp, X, X, Z, ONE), + swizzle(tmp, Z, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(tmp, ONE, Z, Y, X), + i915_emit_const4fv(p, cos_constants), 0); + break; + + case TGSI_OPCODE_DP3: + emit_simple_arith(p, inst, A0_DP3, 2); + break; + + case TGSI_OPCODE_DP4: + emit_simple_arith(p, inst, A0_DP4, 2); + break; + + case TGSI_OPCODE_DPH: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, Y, Z, ONE), src1, 0); + break; + + case TGSI_OPCODE_DST: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + + /* result[0] = 1 * 1; + * result[1] = a[1] * b[1]; + * result[2] = a[2] * 1; + * result[3] = 1 * b[3]; + */ + i915_emit_arith(p, + A0_MUL, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, ONE, Y, Z, ONE), + swizzle(src1, ONE, Y, ONE, W), 0); + break; + + case TGSI_OPCODE_END: + /* no-op */ + break; + + case TGSI_OPCODE_EX2: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + + i915_emit_arith(p, + A0_EXP, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_FLR: + emit_simple_arith(p, inst, A0_FLR, 1); + break; + + case TGSI_OPCODE_FRC: + emit_simple_arith(p, inst, A0_FRC, 1); + break; + + case TGSI_OPCODE_KIL: + /* unconditional kill */ + assert(0); /* not tested yet */ +#if 0 + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ + 0, src0, T0_TEXKILL); +#endif + break; + + case TGSI_OPCODE_KILP: + /* kill if src[0].x < 0 || src[0].y < 0 ... */ + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ + 0, src0, T0_TEXKILL); + break; + + case TGSI_OPCODE_LG2: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + + i915_emit_arith(p, + A0_LOG, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_LIT: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + /* tmp = max( a.xyzw, a.00zw ) + * XXX: Clamp tmp.w to -128..128 + * tmp.y = log(tmp.y) + * tmp.y = tmp.w * tmp.y + * tmp.y = exp(tmp.y) + * result = cmp (a.11-x1, a.1x01, a.1xy1 ) + */ + i915_emit_arith(p, A0_MAX, tmp, A0_DEST_CHANNEL_ALL, 0, + src0, swizzle(src0, ZERO, ZERO, Z, W), 0); + + i915_emit_arith(p, A0_LOG, tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, Y, Y, Y, Y), 0, 0); + + i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, ZERO, Y, ZERO, ZERO), + swizzle(tmp, ZERO, W, ZERO, ZERO), 0); + + i915_emit_arith(p, A0_EXP, tmp, A0_DEST_CHANNEL_Y, 0, + swizzle(tmp, Y, Y, Y, Y), 0, 0); + + i915_emit_arith(p, A0_CMP, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + negate(swizzle(tmp, ONE, ONE, X, ONE), 0, 0, 1, 0), + swizzle(tmp, ONE, X, ZERO, ONE), + swizzle(tmp, ONE, X, Y, ONE)); + + break; + + case TGSI_OPCODE_LRP: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + src2 = src_vector(p, &inst->FullSrcRegisters[2]); + flags = get_result_flags(inst); + tmp = i915_get_utemp(p); + + /* b*a + c*(1-a) + * + * b*a + c - ca + * + * tmp = b*a + c, + * result = (-c)*a + tmp + */ + i915_emit_arith(p, A0_MAD, tmp, + flags & A0_DEST_CHANNEL_ALL, 0, src1, src0, src2); + + i915_emit_arith(p, A0_MAD, + get_result_vector(p, &inst->FullDstRegisters[0]), + flags, 0, negate(src2, 1, 1, 1, 1), src0, tmp); + break; + + case TGSI_OPCODE_MAD: + emit_simple_arith(p, inst, A0_MAD, 3); + break; + + case TGSI_OPCODE_MAX: + emit_simple_arith(p, inst, A0_MAX, 2); + break; + + case TGSI_OPCODE_MIN: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + tmp = i915_get_utemp(p); + flags = get_result_flags(inst); + + i915_emit_arith(p, + A0_MAX, + tmp, flags & A0_DEST_CHANNEL_ALL, 0, + negate(src0, 1, 1, 1, 1), + negate(src1, 1, 1, 1, 1), 0); + + i915_emit_arith(p, + A0_MOV, + get_result_vector(p, &inst->FullDstRegisters[0]), + flags, 0, negate(tmp, 1, 1, 1, 1), 0, 0); + break; + + case TGSI_OPCODE_MOV: + /* aka TGSI_OPCODE_SWZ */ + emit_simple_arith(p, inst, A0_MOV, 1); + break; + + case TGSI_OPCODE_MUL: + emit_simple_arith(p, inst, A0_MUL, 2); + break; + + case TGSI_OPCODE_POW: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + tmp = i915_get_utemp(p); + flags = get_result_flags(inst); + + /* XXX: masking on intermediate values, here and elsewhere. + */ + i915_emit_arith(p, + A0_LOG, + tmp, A0_DEST_CHANNEL_X, 0, + swizzle(src0, X, X, X, X), 0, 0); + + i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, tmp, src1, 0); + + i915_emit_arith(p, + A0_EXP, + get_result_vector(p, &inst->FullDstRegisters[0]), + flags, 0, swizzle(tmp, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_RET: + /* XXX: no-op? */ + break; + + case TGSI_OPCODE_RCP: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + + i915_emit_arith(p, + A0_RCP, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_RSQ: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + + i915_emit_arith(p, + A0_RSQ, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, X, X, X, X), 0, 0); + break; + + case TGSI_OPCODE_SCS: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + /* + * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 + * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x + * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x + * scs.x = DP4 t1, sin_constants + * t1 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 + * scs.y = DP4 t1, cos_constants + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XY, 0, + swizzle(src0, X, X, ONE, ONE), + swizzle(src0, X, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, X, Y), + swizzle(tmp, X, X, ONE, ONE), 0); + + writemask = inst->FullDstRegisters[0].DstRegister.WriteMask; + + if (writemask & TGSI_WRITEMASK_Y) { + uint tmp1; + + if (writemask & TGSI_WRITEMASK_X) + tmp1 = i915_get_utemp(p); + else + tmp1 = tmp; + + i915_emit_arith(p, + A0_MUL, + tmp1, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, Y, W), + swizzle(tmp, X, Z, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + A0_DEST_CHANNEL_Y, 0, + swizzle(tmp1, W, Z, Y, X), + i915_emit_const4fv(p, sin_constants), 0); + } + + if (writemask & TGSI_WRITEMASK_X) { + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XYZ, 0, + swizzle(tmp, X, X, Z, ONE), + swizzle(tmp, Z, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + A0_DEST_CHANNEL_X, 0, + swizzle(tmp, ONE, Z, Y, X), + i915_emit_const4fv(p, cos_constants), 0); + } + break; + + case TGSI_OPCODE_SGE: + emit_simple_arith(p, inst, A0_SGE, 2); + break; + + case TGSI_OPCODE_SIN: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + tmp = i915_get_utemp(p); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + src0, i915_emit_const1f(p, 1.0f / (float) (M_PI * 2.0)), 0); + + i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); + + /* By choosing different taylor constants, could get rid of this mul: + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_X, 0, + tmp, i915_emit_const1f(p, (float) (M_PI * 2.0)), 0); + + /* + * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 + * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x + * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x + * result = DP4 t1.wzyx, sin_constants + */ + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_XY, 0, + swizzle(tmp, X, X, ONE, ONE), + swizzle(tmp, X, ONE, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, X, Y), + swizzle(tmp, X, X, ONE, ONE), 0); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(tmp, X, Y, Y, W), + swizzle(tmp, X, Z, ONE, ONE), 0); + + i915_emit_arith(p, + A0_DP4, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(tmp, W, Z, Y, X), + i915_emit_const4fv(p, sin_constants), 0); + break; + + case TGSI_OPCODE_SLT: + emit_simple_arith(p, inst, A0_SLT, 2); + break; + + case TGSI_OPCODE_SUB: + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + + i915_emit_arith(p, + A0_ADD, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + src0, negate(src1, 1, 1, 1, 1), 0); + break; + + case TGSI_OPCODE_TEX: + if (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide + == TGSI_EXTSWIZZLE_W) { + emit_tex(p, inst, T0_TEXLDP); + } + else { + emit_tex(p, inst, T0_TEXLD); + } + break; + + case TGSI_OPCODE_TXB: + emit_tex(p, inst, T0_TEXLDB); + break; + + case TGSI_OPCODE_XPD: + /* Cross product: + * result.x = src0.y * src1.z - src0.z * src1.y; + * result.y = src0.z * src1.x - src0.x * src1.z; + * result.z = src0.x * src1.y - src0.y * src1.x; + * result.w = undef; + */ + src0 = src_vector(p, &inst->FullSrcRegisters[0]); + src1 = src_vector(p, &inst->FullSrcRegisters[1]); + tmp = i915_get_utemp(p); + + i915_emit_arith(p, + A0_MUL, + tmp, A0_DEST_CHANNEL_ALL, 0, + swizzle(src0, Z, X, Y, ONE), + swizzle(src1, Y, Z, X, ONE), 0); + + i915_emit_arith(p, + A0_MAD, + get_result_vector(p, &inst->FullDstRegisters[0]), + get_result_flags(inst), 0, + swizzle(src0, Y, Z, X, ONE), + swizzle(src1, Z, X, Y, ONE), + negate(tmp, 1, 1, 1, 0)); + break; + + default: + i915_program_error(p, "bad opcode %d", inst->Instruction.Opcode); + return; + } + + i915_release_utemps(p); +} + + +/** + * Translate TGSI fragment shader into i915 hardware instructions. + * \param p the translation state + * \param tokens the TGSI token array + */ +static void +i915_translate_instructions(struct i915_fp_compile *p, + const struct tgsi_token *tokens) +{ + struct tgsi_parse_context parse; + + tgsi_parse_init( &parse, tokens ); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + if (parse.FullToken.FullDeclaration.Declaration.File + == TGSI_FILE_INPUT) { + /* save input register info for use in src_vector() */ + uint ind, sem, semi; + ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; + semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; + /*debug_printf("FS Input DECL [%u] sem %u\n", ind, sem);*/ + p->input_semantic_name[ind] = sem; + p->input_semantic_index[ind] = semi; + } + else if (parse.FullToken.FullDeclaration.Declaration.File + == TGSI_FILE_OUTPUT) { + /* save output register info for use in get_result_vector() */ + uint ind, sem, semi; + ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; + semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; + /*debug_printf("FS Output DECL [%u] sem %u\n", ind, sem);*/ + p->output_semantic_name[ind] = sem; + p->output_semantic_index[ind] = semi; + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* XXX append the immediate to the const buffer... */ + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + i915_translate_instruction(p, &parse.FullToken.FullInstruction); + break; + + default: + assert( 0 ); + } + + } /* while */ + + tgsi_parse_free (&parse); +} + + +static struct i915_fp_compile * +i915_init_compile(struct i915_context *i915, + const struct pipe_shader_state *fs) +{ + struct i915_fp_compile *p = CALLOC_STRUCT(i915_fp_compile); + + p->shader = i915->fs; + + p->vertex_info = &i915->current.vertex_info; + + /* new constants found during translation get appended after the + * user-provided constants. + */ + p->constants = i915->current.constants[PIPE_SHADER_FRAGMENT]; + p->num_constants = i915->current.num_user_constants[PIPE_SHADER_FRAGMENT]; + + p->nr_tex_indirect = 1; /* correct? */ + p->nr_tex_insn = 0; + p->nr_alu_insn = 0; + p->nr_decl_insn = 0; + + memset(p->constant_flags, 0, sizeof(p->constant_flags)); + + p->csr = p->program; + p->decl = p->declarations; + p->decl_s = 0; + p->decl_t = 0; + p->temp_flag = 0xffff000; + p->utemp_flag = ~0x7; + + p->wpos_tex = -1; + + /* initialize the first program word */ + *(p->decl++) = _3DSTATE_PIXEL_SHADER_PROGRAM; + + return p; +} + + +/* Copy compile results to the fragment program struct and destroy the + * compilation context. + */ +static void +i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) +{ + unsigned long program_size = (unsigned long) (p->csr - p->program); + unsigned long decl_size = (unsigned long) (p->decl - p->declarations); + + if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT) + i915_program_error(p, "Exceeded max nr indirect texture lookups"); + + if (p->nr_tex_insn > I915_MAX_TEX_INSN) + i915_program_error(p, "Exceeded max TEX instructions"); + + if (p->nr_alu_insn > I915_MAX_ALU_INSN) + i915_program_error(p, "Exceeded max ALU instructions"); + + if (p->nr_decl_insn > I915_MAX_DECL_INSN) + i915_program_error(p, "Exceeded max DECL instructions"); + + /* free old program, if present */ + if (i915->current.program) { + FREE(i915->current.program); + i915->current.program_len = 0; + } + + if (p->error) { + p->NumNativeInstructions = 0; + p->NumNativeAluInstructions = 0; + p->NumNativeTexInstructions = 0; + p->NumNativeTexIndirections = 0; + + i915_use_passthrough_shader(i915); + } + else { + p->NumNativeInstructions + = p->nr_alu_insn + p->nr_tex_insn + p->nr_decl_insn; + p->NumNativeAluInstructions = p->nr_alu_insn; + p->NumNativeTexInstructions = p->nr_tex_insn; + p->NumNativeTexIndirections = p->nr_tex_indirect; + + /* patch in the program length */ + p->declarations[0] |= program_size + decl_size - 2; + + /* Copy compilation results to fragment program struct: + */ + i915->current.program + = (uint *) MALLOC((program_size + decl_size) * sizeof(uint)); + if (i915->current.program) { + i915->current.program_len = program_size + decl_size; + + memcpy(i915->current.program, + p->declarations, + decl_size * sizeof(uint)); + + memcpy(i915->current.program + decl_size, + p->program, + program_size * sizeof(uint)); + } + + /* update number of constants */ + i915->current.num_constants[PIPE_SHADER_FRAGMENT] = p->num_constants; + assert(i915->current.num_constants[PIPE_SHADER_FRAGMENT] + >= i915->current.num_user_constants[PIPE_SHADER_FRAGMENT]); + } + + /* Release the compilation struct: + */ + FREE(p); +} + + +/** + * Find an unused texture coordinate slot to use for fragment WPOS. + * Update p->fp->wpos_tex with the result (-1 if no used texcoord slot is found). + */ +static void +i915_find_wpos_space(struct i915_fp_compile *p) +{ +#if 0 + const uint inputs + = p->shader->inputs_read | (1 << TGSI_ATTRIB_POS); /*XXX hack*/ + uint i; + + p->wpos_tex = -1; + + if (inputs & (1 << TGSI_ATTRIB_POS)) { + for (i = 0; i < I915_TEX_UNITS; i++) { + if ((inputs & (1 << (TGSI_ATTRIB_TEX0 + i))) == 0) { + p->wpos_tex = i; + return; + } + } + + i915_program_error(p, "No free texcoord for wpos value"); + } +#else + if (p->shader->input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { + /* frag shader using the fragment position input */ +#if 0 + assert(0); +#endif + } +#endif +} + + + + +/** + * Rather than trying to intercept and jiggle depth writes during + * emit, just move the value into its correct position at the end of + * the program: + */ +static void +i915_fixup_depth_write(struct i915_fp_compile *p) +{ + /* XXX assuming pos/depth is always in output[0] */ + if (p->shader->output_semantic_name[0] == TGSI_SEMANTIC_POSITION) { + const uint depth = UREG(REG_TYPE_OD, 0); + + i915_emit_arith(p, + A0_MOV, /* opcode */ + depth, /* dest reg */ + A0_DEST_CHANNEL_W, /* write mask */ + 0, /* saturate? */ + swizzle(depth, X, Y, Z, Z), /* src0 */ + 0, 0 /* src1, src2 */); + } +} + + +void +i915_translate_fragment_program( struct i915_context *i915 ) +{ + struct i915_fp_compile *p = i915_init_compile(i915, i915->fs); + const struct tgsi_token *tokens = i915->fs->tokens; + + i915_find_wpos_space(p); + + i915_translate_instructions(p, tokens); + i915_fixup_depth_write(p); + + i915_fini_compile(i915, p); +} diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c new file mode 100644 index 0000000000..c4a706c37d --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_prim_emit.c @@ -0,0 +1,215 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/draw/draw_private.h" +#include "pipe/p_util.h" + +#include "i915_context.h" +#include "i915_winsys.h" +#include "i915_reg.h" +#include "i915_state.h" +#include "i915_batch.h" + + + +/** + * Primitive emit to hardware. No support for vertex buffers or any + * nice fast paths. + */ +struct setup_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct i915_context *i915; +}; + + + +/** + * Basically a cast wrapper. + */ +static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) +{ + return (struct setup_stage *)stage; +} + + +/** + * Extract the needed fields from vertex_header and emit i915 dwords. + * Recall that the vertices are constructed by the 'draw' module and + * have a couple of slots at the beginning (1-dword header, 4-dword + * clip pos) that we ignore here. + */ +static INLINE void +emit_hw_vertex( struct i915_context *i915, + const struct vertex_header *vertex) +{ + const struct vertex_info *vinfo = &i915->current.vertex_info; + uint i; + uint count = 0; /* for debug/sanity */ + + for (i = 0; i < vinfo->num_attribs; i++) { + switch (vinfo->emit[i]) { + case EMIT_OMIT: + /* no-op */ + break; + case EMIT_1F: + OUT_BATCH( fui(vertex->data[i][0]) ); + count++; + break; + case EMIT_2F: + OUT_BATCH( fui(vertex->data[i][0]) ); + OUT_BATCH( fui(vertex->data[i][1]) ); + count += 2; + break; + case EMIT_3F: + OUT_BATCH( fui(vertex->data[i][0]) ); + OUT_BATCH( fui(vertex->data[i][1]) ); + OUT_BATCH( fui(vertex->data[i][2]) ); + count += 3; + break; + case EMIT_4F: + OUT_BATCH( fui(vertex->data[i][0]) ); + OUT_BATCH( fui(vertex->data[i][1]) ); + OUT_BATCH( fui(vertex->data[i][2]) ); + OUT_BATCH( fui(vertex->data[i][3]) ); + count += 4; + break; + case EMIT_4UB: + OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[i][2] ), + float_to_ubyte( vertex->data[i][1] ), + float_to_ubyte( vertex->data[i][0] ), + float_to_ubyte( vertex->data[i][3] )) ); + count += 1; + break; + default: + assert(0); + } + } + assert(count == vinfo->size); +} + + + +static INLINE void +emit_prim( struct draw_stage *stage, + struct prim_header *prim, + unsigned hwprim, + unsigned nr ) +{ + struct i915_context *i915 = setup_stage(stage)->i915; + unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ + unsigned i; + + assert(vertex_size >= 12); /* never smaller than 12 bytes */ + + if (i915->dirty) + i915_update_derived( i915 ); + + if (i915->hardware_dirty) + i915_emit_hardware_state( i915 ); + + if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) { + FLUSH_BATCH(); + + /* Make sure state is re-emitted after a flush: + */ + i915_update_derived( i915 ); + i915_emit_hardware_state( i915 ); + + if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) { + assert(0); + return; + } + } + + /* Emit each triangle as a single primitive. I told you this was + * simple. + */ + OUT_BATCH(_3DPRIMITIVE | + hwprim | + ((4 + vertex_size * nr)/4 - 2)); + + for (i = 0; i < nr; i++) + emit_hw_vertex(i915, prim->v[i]); +} + + +static void +setup_tri( struct draw_stage *stage, struct prim_header *prim ) +{ + emit_prim( stage, prim, PRIM3D_TRILIST, 3 ); +} + + +static void +setup_line(struct draw_stage *stage, struct prim_header *prim) +{ + emit_prim( stage, prim, PRIM3D_LINELIST, 2 ); +} + + +static void +setup_point(struct draw_stage *stage, struct prim_header *prim) +{ + emit_prim( stage, prim, PRIM3D_POINTLIST, 1 ); +} + + +static void setup_flush( struct draw_stage *stage, unsigned flags ) +{ +} + +static void reset_stipple_counter( struct draw_stage *stage ) +{ +} + +static void render_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + +/** + * Create a new primitive setup/render stage. This gets plugged into + * the 'draw' module's pipeline. + */ +struct draw_stage *i915_draw_render_stage( struct i915_context *i915 ) +{ + struct setup_stage *setup = CALLOC_STRUCT(setup_stage); + + setup->i915 = i915; + setup->stage.draw = i915->draw; + setup->stage.point = setup_point; + setup->stage.line = setup_line; + setup->stage.tri = setup_tri; + setup->stage.flush = setup_flush; + setup->stage.reset_stipple_counter = reset_stipple_counter; + setup->stage.destroy = render_destroy; + + return &setup->stage; +} diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c new file mode 100644 index 0000000000..e069773fd4 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -0,0 +1,254 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Build post-transformation, post-clipping vertex buffers and element + * lists by hooking into the end of the primitive pipeline and + * manipulating the vertex_id field in the vertex headers. + * + * XXX: work in progress + * + * \author José Fonseca + * \author Keith Whitwell + */ + + +#include "pipe/draw/draw_vbuf.h" +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" + +#include "i915_context.h" +#include "i915_reg.h" +#include "i915_winsys.h" +#include "i915_batch.h" +#include "i915_state.h" + + +/** + * Primitive renderer for i915. + */ +struct i915_vbuf_render { + struct vbuf_render base; + + struct i915_context *i915; + + /** Vertex size in bytes */ + unsigned vertex_size; + + /** Hardware primitive */ + unsigned hwprim; +}; + + +/** + * Basically a cast wrapper. + */ +static INLINE struct i915_vbuf_render * +i915_vbuf_render( struct vbuf_render *render ) +{ + assert(render); + return (struct i915_vbuf_render *)render; +} + + +static const struct vertex_info * +i915_vbuf_render_get_vertex_info( struct vbuf_render *render ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + return &i915->current.vertex_info; +} + + +static void * +i915_vbuf_render_allocate_vertices( struct vbuf_render *render, + ushort vertex_size, + ushort nr_vertices ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + struct pipe_winsys *winsys = i915->pipe.winsys; + size_t size = (size_t)vertex_size * (size_t)nr_vertices; + + /* FIXME: handle failure */ + assert(!i915->vbo); + i915->vbo = winsys->buffer_create(winsys, 64, I915_BUFFER_USAGE_LIT_VERTEX, + size); + + i915->dirty |= I915_NEW_VBO; + + return winsys->buffer_map(winsys, + i915->vbo, + PIPE_BUFFER_USAGE_CPU_WRITE); +} + + +static void +i915_vbuf_render_set_primitive( struct vbuf_render *render, + unsigned prim ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + + switch(prim) { + case PIPE_PRIM_POINTS: + i915_render->hwprim = PRIM3D_POINTLIST; + break; + case PIPE_PRIM_LINES: + i915_render->hwprim = PRIM3D_LINELIST; + break; + case PIPE_PRIM_TRIANGLES: + i915_render->hwprim = PRIM3D_TRILIST; + break; + default: + assert(0); + } +} + + +static void +i915_vbuf_render_draw( struct vbuf_render *render, + const ushort *indices, + uint nr_indices) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + unsigned i; + + assert(nr_indices); + + assert((i915->dirty & ~I915_NEW_VBO) == 0); + + if (i915->dirty) + i915_update_derived( i915 ); + + if (i915->hardware_dirty) + i915_emit_hardware_state( i915 ); + + if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { + FLUSH_BATCH(); + + /* Make sure state is re-emitted after a flush: + */ + i915_update_derived( i915 ); + i915_emit_hardware_state( i915 ); + + if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { + assert(0); + return; + } + } + + OUT_BATCH( _3DPRIMITIVE | + PRIM_INDIRECT | + i915_render->hwprim | + PRIM_INDIRECT_ELTS | + nr_indices ); + for (i = 0; i + 1 < nr_indices; i += 2) { + OUT_BATCH( indices[i] | + (indices[i + 1] << 16) ); + } + if (i < nr_indices) { + OUT_BATCH( indices[i] ); + } +} + + +static void +i915_vbuf_render_release_vertices( struct vbuf_render *render, + void *vertices, + unsigned vertex_size, + unsigned vertices_used ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + struct pipe_winsys *winsys = i915->pipe.winsys; + + assert(i915->vbo); + winsys->buffer_unmap(winsys, i915->vbo); + pipe_buffer_reference(winsys, &i915->vbo, NULL); +} + + +static void +i915_vbuf_render_destroy( struct vbuf_render *render ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + FREE(i915_render); +} + + +/** + * Create a new primitive render. + */ +static struct vbuf_render * +i915_vbuf_render_create( struct i915_context *i915 ) +{ + struct i915_vbuf_render *i915_render = CALLOC_STRUCT(i915_vbuf_render); + + i915_render->i915 = i915; + + i915_render->base.max_vertex_buffer_bytes = 128*1024; + + /* NOTE: it must be such that state and vertices indices fit in a single + * batch buffer. + */ + i915_render->base.max_indices = 16*1024; + + i915_render->base.get_vertex_info = i915_vbuf_render_get_vertex_info; + i915_render->base.allocate_vertices = i915_vbuf_render_allocate_vertices; + i915_render->base.set_primitive = i915_vbuf_render_set_primitive; + i915_render->base.draw = i915_vbuf_render_draw; + i915_render->base.release_vertices = i915_vbuf_render_release_vertices; + i915_render->base.destroy = i915_vbuf_render_destroy; + + return &i915_render->base; +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) +{ + struct vbuf_render *render; + struct draw_stage *stage; + + render = i915_vbuf_render_create(i915); + if(!render) + return NULL; + + stage = draw_vbuf_stage( i915->draw, render ); + if(!stage) { + render->destroy(render); + return NULL; + } + + return stage; +} diff --git a/src/gallium/drivers/i915simple/i915_reg.h b/src/gallium/drivers/i915simple/i915_reg.h new file mode 100644 index 0000000000..04620fec68 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_reg.h @@ -0,0 +1,978 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef I915_REG_H +#define I915_REG_H + + +#define I915_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value) + +#define CMD_3D (0x3<<29) + +#define PRIM3D_INLINE (CMD_3D | (0x1f<<24)) +#define PRIM3D_TRILIST (0x0<<18) +#define PRIM3D_TRISTRIP (0x1<<18) +#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) +#define PRIM3D_TRIFAN (0x3<<18) +#define PRIM3D_POLY (0x4<<18) +#define PRIM3D_LINELIST (0x5<<18) +#define PRIM3D_LINESTRIP (0x6<<18) +#define PRIM3D_RECTLIST (0x7<<18) +#define PRIM3D_POINTLIST (0x8<<18) +#define PRIM3D_DIB (0x9<<18) +#define PRIM3D_CLEAR_RECT (0xa<<18) +#define PRIM3D_ZONE_INIT (0xd<<18) +#define PRIM3D_MASK (0x1f<<18) + +/* p137 */ +#define _3DSTATE_AA_CMD (CMD_3D | (0x06<<24)) +#define AA_LINE_ECAAR_WIDTH_ENABLE (1<<16) +#define AA_LINE_ECAAR_WIDTH_0_5 0 +#define AA_LINE_ECAAR_WIDTH_1_0 (1<<14) +#define AA_LINE_ECAAR_WIDTH_2_0 (2<<14) +#define AA_LINE_ECAAR_WIDTH_4_0 (3<<14) +#define AA_LINE_REGION_WIDTH_ENABLE (1<<8) +#define AA_LINE_REGION_WIDTH_0_5 0 +#define AA_LINE_REGION_WIDTH_1_0 (1<<6) +#define AA_LINE_REGION_WIDTH_2_0 (2<<6) +#define AA_LINE_REGION_WIDTH_4_0 (3<<6) + +/* 3DSTATE_BACKFACE_STENCIL_OPS, p138*/ +#define _3DSTATE_BACKFACE_STENCIL_OPS (CMD_3D | (0x8<<24)) +#define BFO_ENABLE_STENCIL_REF (1<<23) +#define BFO_STENCIL_REF_SHIFT 15 +#define BFO_STENCIL_REF_MASK (0xff<<15) +#define BFO_ENABLE_STENCIL_FUNCS (1<<14) +#define BFO_STENCIL_TEST_SHIFT 11 +#define BFO_STENCIL_TEST_MASK (0x7<<11) +#define BFO_STENCIL_FAIL_SHIFT 8 +#define BFO_STENCIL_FAIL_MASK (0x7<<8) +#define BFO_STENCIL_PASS_Z_FAIL_SHIFT 5 +#define BFO_STENCIL_PASS_Z_FAIL_MASK (0x7<<5) +#define BFO_STENCIL_PASS_Z_PASS_SHIFT 2 +#define BFO_STENCIL_PASS_Z_PASS_MASK (0x7<<2) +#define BFO_ENABLE_STENCIL_TWO_SIDE (1<<1) +#define BFO_STENCIL_TWO_SIDE (1<<0) + + +/* 3DSTATE_BACKFACE_STENCIL_MASKS, p140 */ +#define _3DSTATE_BACKFACE_STENCIL_MASKS (CMD_3D | (0x9<<24)) +#define BFM_ENABLE_STENCIL_TEST_MASK (1<<17) +#define BFM_ENABLE_STENCIL_WRITE_MASK (1<<16) +#define BFM_STENCIL_TEST_MASK_SHIFT 8 +#define BFM_STENCIL_TEST_MASK_MASK (0xff<<8) +#define BFM_STENCIL_WRITE_MASK_SHIFT 0 +#define BFM_STENCIL_WRITE_MASK_MASK (0xff<<0) + + + +/* 3DSTATE_BIN_CONTROL p141 */ + +/* p143 */ +#define _3DSTATE_BUF_INFO_CMD (CMD_3D | (0x1d<<24) | (0x8e<<16) | 1) +/* Dword 1 */ +#define BUF_3D_ID_COLOR_BACK (0x3<<24) +#define BUF_3D_ID_DEPTH (0x7<<24) +#define BUF_3D_USE_FENCE (1<<23) +#define BUF_3D_TILED_SURFACE (1<<22) +#define BUF_3D_TILE_WALK_X 0 +#define BUF_3D_TILE_WALK_Y (1<<21) +#define BUF_3D_PITCH(x) (((x)/4)<<2) +/* Dword 2 */ +#define BUF_3D_ADDR(x) ((x) & ~0x3) + + +/* 3DSTATE_CHROMA_KEY */ + +/* 3DSTATE_CLEAR_PARAMETERS, p150 */ +#define _3DSTATE_CLEAR_PARAMETERS (CMD_3D | (0x1d<<24) | (0x9c<<16) | 5) +/* Dword 1 */ +#define CLEARPARAM_CLEAR_RECT (1 << 16) +#define CLEARPARAM_ZONE_INIT (0 << 16) +#define CLEARPARAM_WRITE_COLOR (1 << 2) +#define CLEARPARAM_WRITE_DEPTH (1 << 1) +#define CLEARPARAM_WRITE_STENCIL (1 << 0) + +/* 3DSTATE_CONSTANT_BLEND_COLOR, p153 */ +#define _3DSTATE_CONST_BLEND_COLOR_CMD (CMD_3D | (0x1d<<24) | (0x88<<16)) + + + +/* 3DSTATE_COORD_SET_BINDINGS, p154 */ +#define _3DSTATE_COORD_SET_BINDINGS (CMD_3D | (0x16<<24)) +#define CSB_TCB(iunit, eunit) ((eunit)<<(iunit*3)) + +/* p156 */ +#define _3DSTATE_DFLT_DIFFUSE_CMD (CMD_3D | (0x1d<<24) | (0x99<<16)) + +/* p157 */ +#define _3DSTATE_DFLT_SPEC_CMD (CMD_3D | (0x1d<<24) | (0x9a<<16)) + +/* p158 */ +#define _3DSTATE_DFLT_Z_CMD (CMD_3D | (0x1d<<24) | (0x98<<16)) + + +/* 3DSTATE_DEPTH_OFFSET_SCALE, p159 */ +#define _3DSTATE_DEPTH_OFFSET_SCALE (CMD_3D | (0x1d<<24) | (0x97<<16)) +/* scale in dword 1 */ + + +/* 3DSTATE_DEPTH_SUBRECT_DISABLE, p160 */ +#define _3DSTATE_DEPTH_SUBRECT_DISABLE (CMD_3D | (0x1c<<24) | (0x11<<19) | 0x2) + +/* p161 */ +#define _3DSTATE_DST_BUF_VARS_CMD (CMD_3D | (0x1d<<24) | (0x85<<16)) +/* Dword 1 */ +#define TEX_DEFAULT_COLOR_OGL (0<<30) +#define TEX_DEFAULT_COLOR_D3D (1<<30) +#define ZR_EARLY_DEPTH (1<<29) +#define LOD_PRECLAMP_OGL (1<<28) +#define LOD_PRECLAMP_D3D (0<<28) +#define DITHER_FULL_ALWAYS (0<<26) +#define DITHER_FULL_ON_FB_BLEND (1<<26) +#define DITHER_CLAMPED_ALWAYS (2<<26) +#define LINEAR_GAMMA_BLEND_32BPP (1<<25) +#define DEBUG_DISABLE_ENH_DITHER (1<<24) +#define DSTORG_HORT_BIAS(x) ((x)<<20) +#define DSTORG_VERT_BIAS(x) ((x)<<16) +#define COLOR_4_2_2_CHNL_WRT_ALL 0 +#define COLOR_4_2_2_CHNL_WRT_Y (1<<12) +#define COLOR_4_2_2_CHNL_WRT_CR (2<<12) +#define COLOR_4_2_2_CHNL_WRT_CB (3<<12) +#define COLOR_4_2_2_CHNL_WRT_CRCB (4<<12) +#define COLOR_BUF_8BIT 0 +#define COLOR_BUF_RGB555 (1<<8) +#define COLOR_BUF_RGB565 (2<<8) +#define COLOR_BUF_ARGB8888 (3<<8) +#define DEPTH_FRMT_16_FIXED 0 +#define DEPTH_FRMT_16_FLOAT (1<<2) +#define DEPTH_FRMT_24_FIXED_8_OTHER (2<<2) +#define VERT_LINE_STRIDE_1 (1<<1) +#define VERT_LINE_STRIDE_0 (0<<1) +#define VERT_LINE_STRIDE_OFS_1 1 +#define VERT_LINE_STRIDE_OFS_0 0 + +/* p166 */ +#define _3DSTATE_DRAW_RECT_CMD (CMD_3D|(0x1d<<24)|(0x80<<16)|3) +/* Dword 1 */ +#define DRAW_RECT_DIS_DEPTH_OFS (1<<30) +#define DRAW_DITHER_OFS_X(x) ((x)<<26) +#define DRAW_DITHER_OFS_Y(x) ((x)<<24) +/* Dword 2 */ +#define DRAW_YMIN(x) ((x)<<16) +#define DRAW_XMIN(x) (x) +/* Dword 3 */ +#define DRAW_YMAX(x) ((x)<<16) +#define DRAW_XMAX(x) (x) +/* Dword 4 */ +#define DRAW_YORG(x) ((x)<<16) +#define DRAW_XORG(x) (x) + + +/* 3DSTATE_FILTER_COEFFICIENTS_4X4, p170 */ + +/* 3DSTATE_FILTER_COEFFICIENTS_6X5, p172 */ + + +/* _3DSTATE_FOG_COLOR, p173 */ +#define _3DSTATE_FOG_COLOR_CMD (CMD_3D|(0x15<<24)) +#define FOG_COLOR_RED(x) ((x)<<16) +#define FOG_COLOR_GREEN(x) ((x)<<8) +#define FOG_COLOR_BLUE(x) (x) + +/* _3DSTATE_FOG_MODE, p174 */ +#define _3DSTATE_FOG_MODE_CMD (CMD_3D|(0x1d<<24)|(0x89<<16)|2) +/* Dword 1 */ +#define FMC1_FOGFUNC_MODIFY_ENABLE (1<<31) +#define FMC1_FOGFUNC_VERTEX (0<<28) +#define FMC1_FOGFUNC_PIXEL_EXP (1<<28) +#define FMC1_FOGFUNC_PIXEL_EXP2 (2<<28) +#define FMC1_FOGFUNC_PIXEL_LINEAR (3<<28) +#define FMC1_FOGFUNC_MASK (3<<28) +#define FMC1_FOGINDEX_MODIFY_ENABLE (1<<27) +#define FMC1_FOGINDEX_Z (0<<25) +#define FMC1_FOGINDEX_W (1<<25) +#define FMC1_C1_C2_MODIFY_ENABLE (1<<24) +#define FMC1_DENSITY_MODIFY_ENABLE (1<<23) +#define FMC1_C1_ONE (1<<13) +#define FMC1_C1_MASK (0xffff<<4) +/* Dword 2 */ +#define FMC2_C2_ONE (1<<16) +/* Dword 3 */ +#define FMC3_D_ONE (1<<16) + + + +/* _3DSTATE_INDEPENDENT_ALPHA_BLEND, p177 */ +#define _3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD (CMD_3D|(0x0b<<24)) +#define IAB_MODIFY_ENABLE (1<<23) +#define IAB_ENABLE (1<<22) +#define IAB_MODIFY_FUNC (1<<21) +#define IAB_FUNC_SHIFT 16 +#define IAB_MODIFY_SRC_FACTOR (1<<11) +#define IAB_SRC_FACTOR_SHIFT 6 +#define IAB_SRC_FACTOR_MASK (BLENDFACT_MASK<<6) +#define IAB_MODIFY_DST_FACTOR (1<<5) +#define IAB_DST_FACTOR_SHIFT 0 +#define IAB_DST_FACTOR_MASK (BLENDFACT_MASK<<0) + + +#define BLENDFUNC_ADD 0x0 +#define BLENDFUNC_SUBTRACT 0x1 +#define BLENDFUNC_REVERSE_SUBTRACT 0x2 +#define BLENDFUNC_MIN 0x3 +#define BLENDFUNC_MAX 0x4 +#define BLENDFUNC_MASK 0x7 + +/* 3DSTATE_LOAD_INDIRECT, p180 */ + +#define _3DSTATE_LOAD_INDIRECT (CMD_3D|(0x1d<<24)|(0x7<<16)) +#define LI0_STATE_STATIC_INDIRECT (0x01<<8) +#define LI0_STATE_DYNAMIC_INDIRECT (0x02<<8) +#define LI0_STATE_SAMPLER (0x04<<8) +#define LI0_STATE_MAP (0x08<<8) +#define LI0_STATE_PROGRAM (0x10<<8) +#define LI0_STATE_CONSTANTS (0x20<<8) + +#define SIS0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define SIS0_FORCE_LOAD (1<<1) +#define SIS0_BUFFER_VALID (1<<0) +#define SIS1_BUFFER_LENGTH(x) ((x)&0xff) + +#define DIS0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define DIS0_BUFFER_RESET (1<<1) +#define DIS0_BUFFER_VALID (1<<0) + +#define SSB0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define SSB0_FORCE_LOAD (1<<1) +#define SSB0_BUFFER_VALID (1<<0) +#define SSB1_BUFFER_LENGTH(x) ((x)&0xff) + +#define MSB0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define MSB0_FORCE_LOAD (1<<1) +#define MSB0_BUFFER_VALID (1<<0) +#define MSB1_BUFFER_LENGTH(x) ((x)&0xff) + +#define PSP0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define PSP0_FORCE_LOAD (1<<1) +#define PSP0_BUFFER_VALID (1<<0) +#define PSP1_BUFFER_LENGTH(x) ((x)&0xff) + +#define PSC0_BUFFER_ADDRESS(x) ((x)&~0x3) +#define PSC0_FORCE_LOAD (1<<1) +#define PSC0_BUFFER_VALID (1<<0) +#define PSC1_BUFFER_LENGTH(x) ((x)&0xff) + + + + + +/* _3DSTATE_RASTERIZATION_RULES */ +#define _3DSTATE_RASTER_RULES_CMD (CMD_3D|(0x07<<24)) +#define ENABLE_POINT_RASTER_RULE (1<<15) +#define OGL_POINT_RASTER_RULE (1<<13) +#define ENABLE_TEXKILL_3D_4D (1<<10) +#define TEXKILL_3D (0<<9) +#define TEXKILL_4D (1<<9) +#define ENABLE_LINE_STRIP_PROVOKE_VRTX (1<<8) +#define ENABLE_TRI_FAN_PROVOKE_VRTX (1<<5) +#define LINE_STRIP_PROVOKE_VRTX(x) ((x)<<6) +#define TRI_FAN_PROVOKE_VRTX(x) ((x)<<3) + +/* _3DSTATE_SCISSOR_ENABLE, p256 */ +#define _3DSTATE_SCISSOR_ENABLE_CMD (CMD_3D|(0x1c<<24)|(0x10<<19)) +#define ENABLE_SCISSOR_RECT ((1<<1) | 1) +#define DISABLE_SCISSOR_RECT (1<<1) + +/* _3DSTATE_SCISSOR_RECTANGLE_0, p257 */ +#define _3DSTATE_SCISSOR_RECT_0_CMD (CMD_3D|(0x1d<<24)|(0x81<<16)|1) +/* Dword 1 */ +#define SCISSOR_RECT_0_YMIN(x) ((x)<<16) +#define SCISSOR_RECT_0_XMIN(x) (x) +/* Dword 2 */ +#define SCISSOR_RECT_0_YMAX(x) ((x)<<16) +#define SCISSOR_RECT_0_XMAX(x) (x) + +/* p189 */ +#define _3DSTATE_LOAD_STATE_IMMEDIATE_1 ((0x3<<29)|(0x1d<<24)|(0x04<<16)) +#define I1_LOAD_S(n) (1<<(4+n)) + +#define S0_VB_OFFSET_MASK 0xffffffc +#define S0_AUTO_CACHE_INV_DISABLE (1<<0) + +#define S1_VERTEX_WIDTH_SHIFT 24 +#define S1_VERTEX_WIDTH_MASK (0x3f<<24) +#define S1_VERTEX_PITCH_SHIFT 16 +#define S1_VERTEX_PITCH_MASK (0x3f<<16) + +#define TEXCOORDFMT_2D 0x0 +#define TEXCOORDFMT_3D 0x1 +#define TEXCOORDFMT_4D 0x2 +#define TEXCOORDFMT_1D 0x3 +#define TEXCOORDFMT_2D_16 0x4 +#define TEXCOORDFMT_4D_16 0x5 +#define TEXCOORDFMT_NOT_PRESENT 0xf +#define S2_TEXCOORD_FMT0_MASK 0xf +#define S2_TEXCOORD_FMT1_SHIFT 4 +#define S2_TEXCOORD_FMT(unit, type) ((type)<<(unit*4)) +#define S2_TEXCOORD_NONE (~0) + +/* S3 not interesting */ + +#define S4_POINT_WIDTH_SHIFT 23 +#define S4_POINT_WIDTH_MASK (0x1ff<<23) +#define S4_LINE_WIDTH_SHIFT 19 +#define S4_LINE_WIDTH_ONE (0x2<<19) +#define S4_LINE_WIDTH_MASK (0xf<<19) +#define S4_FLATSHADE_ALPHA (1<<18) +#define S4_FLATSHADE_FOG (1<<17) +#define S4_FLATSHADE_SPECULAR (1<<16) +#define S4_FLATSHADE_COLOR (1<<15) +#define S4_CULLMODE_BOTH (0<<13) +#define S4_CULLMODE_NONE (1<<13) +#define S4_CULLMODE_CW (2<<13) +#define S4_CULLMODE_CCW (3<<13) +#define S4_CULLMODE_MASK (3<<13) +#define S4_VFMT_POINT_WIDTH (1<<12) +#define S4_VFMT_SPEC_FOG (1<<11) +#define S4_VFMT_COLOR (1<<10) +#define S4_VFMT_DEPTH_OFFSET (1<<9) +#define S4_VFMT_XYZ (1<<6) +#define S4_VFMT_XYZW (2<<6) +#define S4_VFMT_XY (3<<6) +#define S4_VFMT_XYW (4<<6) +#define S4_VFMT_XYZW_MASK (7<<6) +#define S4_FORCE_DEFAULT_DIFFUSE (1<<5) +#define S4_FORCE_DEFAULT_SPECULAR (1<<4) +#define S4_LOCAL_DEPTH_OFFSET_ENABLE (1<<3) +#define S4_VFMT_FOG_PARAM (1<<2) +#define S4_SPRITE_POINT_ENABLE (1<<1) +#define S4_LINE_ANTIALIAS_ENABLE (1<<0) + +#define S4_VFMT_MASK (S4_VFMT_POINT_WIDTH | \ + S4_VFMT_SPEC_FOG | \ + S4_VFMT_COLOR | \ + S4_VFMT_DEPTH_OFFSET | \ + S4_VFMT_XYZW_MASK | \ + S4_VFMT_FOG_PARAM) + + +#define S5_WRITEDISABLE_ALPHA (1<<31) +#define S5_WRITEDISABLE_RED (1<<30) +#define S5_WRITEDISABLE_GREEN (1<<29) +#define S5_WRITEDISABLE_BLUE (1<<28) +#define S5_WRITEDISABLE_MASK (0xf<<28) +#define S5_FORCE_DEFAULT_POINT_SIZE (1<<27) +#define S5_LAST_PIXEL_ENABLE (1<<26) +#define S5_GLOBAL_DEPTH_OFFSET_ENABLE (1<<25) +#define S5_FOG_ENABLE (1<<24) +#define S5_STENCIL_REF_SHIFT 16 +#define S5_STENCIL_REF_MASK (0xff<<16) +#define S5_STENCIL_TEST_FUNC_SHIFT 13 +#define S5_STENCIL_TEST_FUNC_MASK (0x7<<13) +#define S5_STENCIL_FAIL_SHIFT 10 +#define S5_STENCIL_FAIL_MASK (0x7<<10) +#define S5_STENCIL_PASS_Z_FAIL_SHIFT 7 +#define S5_STENCIL_PASS_Z_FAIL_MASK (0x7<<7) +#define S5_STENCIL_PASS_Z_PASS_SHIFT 4 +#define S5_STENCIL_PASS_Z_PASS_MASK (0x7<<4) +#define S5_STENCIL_WRITE_ENABLE (1<<3) +#define S5_STENCIL_TEST_ENABLE (1<<2) +#define S5_COLOR_DITHER_ENABLE (1<<1) +#define S5_LOGICOP_ENABLE (1<<0) + + +#define S6_ALPHA_TEST_ENABLE (1<<31) +#define S6_ALPHA_TEST_FUNC_SHIFT 28 +#define S6_ALPHA_TEST_FUNC_MASK (0x7<<28) +#define S6_ALPHA_REF_SHIFT 20 +#define S6_ALPHA_REF_MASK (0xff<<20) +#define S6_DEPTH_TEST_ENABLE (1<<19) +#define S6_DEPTH_TEST_FUNC_SHIFT 16 +#define S6_DEPTH_TEST_FUNC_MASK (0x7<<16) +#define S6_CBUF_BLEND_ENABLE (1<<15) +#define S6_CBUF_BLEND_FUNC_SHIFT 12 +#define S6_CBUF_BLEND_FUNC_MASK (0x7<<12) +#define S6_CBUF_SRC_BLEND_FACT_SHIFT 8 +#define S6_CBUF_SRC_BLEND_FACT_MASK (0xf<<8) +#define S6_CBUF_DST_BLEND_FACT_SHIFT 4 +#define S6_CBUF_DST_BLEND_FACT_MASK (0xf<<4) +#define S6_DEPTH_WRITE_ENABLE (1<<3) +#define S6_COLOR_WRITE_ENABLE (1<<2) +#define S6_TRISTRIP_PV_SHIFT 0 +#define S6_TRISTRIP_PV_MASK (0x3<<0) + +#define S7_DEPTH_OFFSET_CONST_MASK ~0 + + + +#define DST_BLND_FACT(f) ((f)<= 0.0) ? src1 : src2 */ +#define A0_MIN (0xe<<24) /* dst = (src0 < src1) ? src0 : src1 */ +#define A0_MAX (0xf<<24) /* dst = (src0 >= src1) ? src0 : src1 */ +#define A0_FLR (0x10<<24) /* dst = floor(src0) */ +#define A0_MOD (0x11<<24) /* dst = src0 fmod 1.0 */ +#define A0_TRC (0x12<<24) /* dst = int(src0) */ +#define A0_SGE (0x13<<24) /* dst = src0 >= src1 ? 1.0 : 0.0 */ +#define A0_SLT (0x14<<24) /* dst = src0 < src1 ? 1.0 : 0.0 */ +#define A0_DEST_SATURATE (1<<22) +#define A0_DEST_TYPE_SHIFT 19 +/* Allow: R, OC, OD, U */ +#define A0_DEST_NR_SHIFT 14 +/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ +#define A0_DEST_CHANNEL_X (1<<10) +#define A0_DEST_CHANNEL_Y (2<<10) +#define A0_DEST_CHANNEL_Z (4<<10) +#define A0_DEST_CHANNEL_W (8<<10) +#define A0_DEST_CHANNEL_ALL (0xf<<10) +#define A0_DEST_CHANNEL_SHIFT 10 +#define A0_SRC0_TYPE_SHIFT 7 +#define A0_SRC0_NR_SHIFT 2 + +#define A0_DEST_CHANNEL_XY (A0_DEST_CHANNEL_X|A0_DEST_CHANNEL_Y) +#define A0_DEST_CHANNEL_XYZ (A0_DEST_CHANNEL_XY|A0_DEST_CHANNEL_Z) + + +#define SRC_X 0 +#define SRC_Y 1 +#define SRC_Z 2 +#define SRC_W 3 +#define SRC_ZERO 4 +#define SRC_ONE 5 + +#define A1_SRC0_CHANNEL_X_NEGATE (1<<31) +#define A1_SRC0_CHANNEL_X_SHIFT 28 +#define A1_SRC0_CHANNEL_Y_NEGATE (1<<27) +#define A1_SRC0_CHANNEL_Y_SHIFT 24 +#define A1_SRC0_CHANNEL_Z_NEGATE (1<<23) +#define A1_SRC0_CHANNEL_Z_SHIFT 20 +#define A1_SRC0_CHANNEL_W_NEGATE (1<<19) +#define A1_SRC0_CHANNEL_W_SHIFT 16 +#define A1_SRC1_TYPE_SHIFT 13 +#define A1_SRC1_NR_SHIFT 8 +#define A1_SRC1_CHANNEL_X_NEGATE (1<<7) +#define A1_SRC1_CHANNEL_X_SHIFT 4 +#define A1_SRC1_CHANNEL_Y_NEGATE (1<<3) +#define A1_SRC1_CHANNEL_Y_SHIFT 0 + +#define A2_SRC1_CHANNEL_Z_NEGATE (1<<31) +#define A2_SRC1_CHANNEL_Z_SHIFT 28 +#define A2_SRC1_CHANNEL_W_NEGATE (1<<27) +#define A2_SRC1_CHANNEL_W_SHIFT 24 +#define A2_SRC2_TYPE_SHIFT 21 +#define A2_SRC2_NR_SHIFT 16 +#define A2_SRC2_CHANNEL_X_NEGATE (1<<15) +#define A2_SRC2_CHANNEL_X_SHIFT 12 +#define A2_SRC2_CHANNEL_Y_NEGATE (1<<11) +#define A2_SRC2_CHANNEL_Y_SHIFT 8 +#define A2_SRC2_CHANNEL_Z_NEGATE (1<<7) +#define A2_SRC2_CHANNEL_Z_SHIFT 4 +#define A2_SRC2_CHANNEL_W_NEGATE (1<<3) +#define A2_SRC2_CHANNEL_W_SHIFT 0 + + + +/* Texture instructions */ +#define T0_TEXLD (0x15<<24) /* Sample texture using predeclared + * sampler and address, and output + * filtered texel data to destination + * register */ +#define T0_TEXLDP (0x16<<24) /* Same as texld but performs a + * perspective divide of the texture + * coordinate .xyz values by .w before + * sampling. */ +#define T0_TEXLDB (0x17<<24) /* Same as texld but biases the + * computed LOD by w. Only S4.6 two's + * comp is used. This implies that a + * float to fixed conversion is + * done. */ +#define T0_TEXKILL (0x18<<24) /* Does not perform a sampling + * operation. Simply kills the pixel + * if any channel of the address + * register is < 0.0. */ +#define T0_DEST_TYPE_SHIFT 19 +/* Allow: R, OC, OD, U */ +/* Note: U (unpreserved) regs do not retain their values between + * phases (cannot be used for feedback) + * + * Note: oC and OD registers can only be used as the destination of a + * texture instruction once per phase (this is an implementation + * restriction). + */ +#define T0_DEST_NR_SHIFT 14 +/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ +#define T0_SAMPLER_NR_SHIFT 0 /* This field ignored for TEXKILL */ +#define T0_SAMPLER_NR_MASK (0xf<<0) + +#define T1_ADDRESS_REG_TYPE_SHIFT 24 /* Reg to use as texture coord */ +/* Allow R, T, OC, OD -- R, OC, OD are 'dependent' reads, new program phase */ +#define T1_ADDRESS_REG_NR_SHIFT 17 +#define T2_MBZ 0 + +/* Declaration instructions */ +#define D0_DCL (0x19<<24) /* Declare a t (interpolated attrib) + * register or an s (sampler) + * register. */ +#define D0_SAMPLE_TYPE_SHIFT 22 +#define D0_SAMPLE_TYPE_2D (0x0<<22) +#define D0_SAMPLE_TYPE_CUBE (0x1<<22) +#define D0_SAMPLE_TYPE_VOLUME (0x2<<22) +#define D0_SAMPLE_TYPE_MASK (0x3<<22) + +#define D0_TYPE_SHIFT 19 +/* Allow: T, S */ +#define D0_NR_SHIFT 14 +/* Allow T: 0..10, S: 0..15 */ +#define D0_CHANNEL_X (1<<10) +#define D0_CHANNEL_Y (2<<10) +#define D0_CHANNEL_Z (4<<10) +#define D0_CHANNEL_W (8<<10) +#define D0_CHANNEL_ALL (0xf<<10) +#define D0_CHANNEL_NONE (0<<10) + +#define D0_CHANNEL_XY (D0_CHANNEL_X|D0_CHANNEL_Y) +#define D0_CHANNEL_XYZ (D0_CHANNEL_XY|D0_CHANNEL_Z) + +/* I915 Errata: Do not allow (xz), (xw), (xzw) combinations for diffuse + * or specular declarations. + * + * For T dcls, only allow: (x), (xy), (xyz), (w), (xyzw) + * + * Must be zero for S (sampler) dcls + */ +#define D1_MBZ 0 +#define D2_MBZ 0 + + + +/* p207 */ +#define _3DSTATE_MAP_STATE (CMD_3D|(0x1d<<24)|(0x0<<16)) + +#define MS1_MAPMASK_SHIFT 0 +#define MS1_MAPMASK_MASK (0x8fff<<0) + +#define MS2_UNTRUSTED_SURFACE (1<<31) +#define MS2_ADDRESS_MASK 0xfffffffc +#define MS2_VERTICAL_LINE_STRIDE (1<<1) +#define MS2_VERTICAL_OFFSET (1<<1) + +#define MS3_HEIGHT_SHIFT 21 +#define MS3_WIDTH_SHIFT 10 +#define MS3_PALETTE_SELECT (1<<9) +#define MS3_MAPSURF_FORMAT_SHIFT 7 +#define MS3_MAPSURF_FORMAT_MASK (0x7<<7) +#define MAPSURF_8BIT (1<<7) +#define MAPSURF_16BIT (2<<7) +#define MAPSURF_32BIT (3<<7) +#define MAPSURF_422 (5<<7) +#define MAPSURF_COMPRESSED (6<<7) +#define MAPSURF_4BIT_INDEXED (7<<7) +#define MS3_MT_FORMAT_MASK (0x7 << 3) +#define MS3_MT_FORMAT_SHIFT 3 +#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ +#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ +#define MT_8BIT_L8 (1<<3) +#define MT_8BIT_A8 (4<<3) +#define MT_8BIT_MONO8 (5<<3) +#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ +#define MT_16BIT_ARGB1555 (1<<3) +#define MT_16BIT_ARGB4444 (2<<3) +#define MT_16BIT_AY88 (3<<3) +#define MT_16BIT_88DVDU (5<<3) +#define MT_16BIT_BUMP_655LDVDU (6<<3) +#define MT_16BIT_I16 (7<<3) +#define MT_16BIT_L16 (8<<3) +#define MT_16BIT_A16 (9<<3) +#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ +#define MT_32BIT_ABGR8888 (1<<3) +#define MT_32BIT_XRGB8888 (2<<3) +#define MT_32BIT_XBGR8888 (3<<3) +#define MT_32BIT_QWVU8888 (4<<3) +#define MT_32BIT_AXVU8888 (5<<3) +#define MT_32BIT_LXVU8888 (6<<3) +#define MT_32BIT_XLVU8888 (7<<3) +#define MT_32BIT_ARGB2101010 (8<<3) +#define MT_32BIT_ABGR2101010 (9<<3) +#define MT_32BIT_AWVU2101010 (0xA<<3) +#define MT_32BIT_GR1616 (0xB<<3) +#define MT_32BIT_VU1616 (0xC<<3) +#define MT_32BIT_xI824 (0xD<<3) +#define MT_32BIT_xA824 (0xE<<3) +#define MT_32BIT_xL824 (0xF<<3) +#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ +#define MT_422_YCRCB_NORMAL (1<<3) +#define MT_422_YCRCB_SWAPUV (2<<3) +#define MT_422_YCRCB_SWAPUVY (3<<3) +#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ +#define MT_COMPRESS_DXT2_3 (1<<3) +#define MT_COMPRESS_DXT4_5 (2<<3) +#define MT_COMPRESS_FXT1 (3<<3) +#define MT_COMPRESS_DXT1_RGB (4<<3) +#define MS3_USE_FENCE_REGS (1<<2) +#define MS3_TILED_SURFACE (1<<1) +#define MS3_TILE_WALK (1<<0) + +#define MS4_PITCH_SHIFT 21 +#define MS4_CUBE_FACE_ENA_NEGX (1<<20) +#define MS4_CUBE_FACE_ENA_POSX (1<<19) +#define MS4_CUBE_FACE_ENA_NEGY (1<<18) +#define MS4_CUBE_FACE_ENA_POSY (1<<17) +#define MS4_CUBE_FACE_ENA_NEGZ (1<<16) +#define MS4_CUBE_FACE_ENA_POSZ (1<<15) +#define MS4_CUBE_FACE_ENA_MASK (0x3f<<15) +#define MS4_MAX_LOD_SHIFT 9 +#define MS4_MAX_LOD_MASK (0x3f<<9) +#define MS4_MIP_LAYOUT_LEGACY (0<<8) +#define MS4_MIP_LAYOUT_BELOW_LPT (0<<8) +#define MS4_MIP_LAYOUT_RIGHT_LPT (1<<8) +#define MS4_VOLUME_DEPTH_SHIFT 0 +#define MS4_VOLUME_DEPTH_MASK (0xff<<0) + +/* p244 */ +#define _3DSTATE_SAMPLER_STATE (CMD_3D|(0x1d<<24)|(0x1<<16)) + +#define SS1_MAPMASK_SHIFT 0 +#define SS1_MAPMASK_MASK (0x8fff<<0) + +#define SS2_REVERSE_GAMMA_ENABLE (1<<31) +#define SS2_PACKED_TO_PLANAR_ENABLE (1<<30) +#define SS2_COLORSPACE_CONVERSION (1<<29) +#define SS2_CHROMAKEY_SHIFT 27 +#define SS2_BASE_MIP_LEVEL_SHIFT 22 +#define SS2_BASE_MIP_LEVEL_MASK (0x1f<<22) +#define SS2_MIP_FILTER_SHIFT 20 +#define SS2_MIP_FILTER_MASK (0x3<<20) +#define MIPFILTER_NONE 0 +#define MIPFILTER_NEAREST 1 +#define MIPFILTER_LINEAR 3 +#define SS2_MAG_FILTER_SHIFT 17 +#define SS2_MAG_FILTER_MASK (0x7<<17) +#define FILTER_NEAREST 0 +#define FILTER_LINEAR 1 +#define FILTER_ANISOTROPIC 2 +#define FILTER_4X4_1 3 +#define FILTER_4X4_2 4 +#define FILTER_4X4_FLAT 5 +#define FILTER_6X5_MONO 6 /* XXX - check */ +#define SS2_MIN_FILTER_SHIFT 14 +#define SS2_MIN_FILTER_MASK (0x7<<14) +#define SS2_LOD_BIAS_SHIFT 5 +#define SS2_LOD_BIAS_ONE (0x10<<5) +#define SS2_LOD_BIAS_MASK (0x1ff<<5) +/* Shadow requires: + * MT_X8{I,L,A}24 or MT_{I,L,A}16 texture format + * FILTER_4X4_x MIN and MAG filters + */ +#define SS2_SHADOW_ENABLE (1<<4) +#define SS2_MAX_ANISO_MASK (1<<3) +#define SS2_MAX_ANISO_2 (0<<3) +#define SS2_MAX_ANISO_4 (1<<3) +#define SS2_SHADOW_FUNC_SHIFT 0 +#define SS2_SHADOW_FUNC_MASK (0x7<<0) +/* SS2_SHADOW_FUNC values: see COMPAREFUNC_* */ + +#define SS3_MIN_LOD_SHIFT 24 +#define SS3_MIN_LOD_ONE (0x10<<24) +#define SS3_MIN_LOD_MASK (0xff<<24) +#define SS3_KILL_PIXEL_ENABLE (1<<17) +#define SS3_TCX_ADDR_MODE_SHIFT 12 +#define SS3_TCX_ADDR_MODE_MASK (0x7<<12) +#define TEXCOORDMODE_WRAP 0 +#define TEXCOORDMODE_MIRROR 1 +#define TEXCOORDMODE_CLAMP_EDGE 2 +#define TEXCOORDMODE_CUBE 3 +#define TEXCOORDMODE_CLAMP_BORDER 4 +#define TEXCOORDMODE_MIRROR_ONCE 5 +#define SS3_TCY_ADDR_MODE_SHIFT 9 +#define SS3_TCY_ADDR_MODE_MASK (0x7<<9) +#define SS3_TCZ_ADDR_MODE_SHIFT 6 +#define SS3_TCZ_ADDR_MODE_MASK (0x7<<6) +#define SS3_NORMALIZED_COORDS (1<<5) +#define SS3_TEXTUREMAP_INDEX_SHIFT 1 +#define SS3_TEXTUREMAP_INDEX_MASK (0xf<<1) +#define SS3_DEINTERLACER_ENABLE (1<<0) + +#define SS4_BORDER_COLOR_MASK (~0) + +/* 3DSTATE_SPAN_STIPPLE, p258 + */ +#define _3DSTATE_STIPPLE ((0x3<<29)|(0x1d<<24)|(0x83<<16)) +#define ST1_ENABLE (1<<16) +#define ST1_MASK (0xffff) + +#define _3DSTATE_DEFAULT_Z ((0x3<<29)|(0x1d<<24)|(0x98<<16)) +#define _3DSTATE_DEFAULT_DIFFUSE ((0x3<<29)|(0x1d<<24)|(0x99<<16)) +#define _3DSTATE_DEFAULT_SPECULAR ((0x3<<29)|(0x1d<<24)|(0x9a<<16)) + + +#define MI_FLUSH ((0<<29)|(4<<23)) +#define FLUSH_MAP_CACHE (1<<0) +#define INHIBIT_FLUSH_RENDER_CACHE (1<<2) + + +#define CMD_3D (0x3<<29) + + +#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) +#define PRIM_INDIRECT (1<<23) +#define PRIM_INLINE (0<<23) +#define PRIM_INDIRECT_SEQUENTIAL (0<<17) +#define PRIM_INDIRECT_ELTS (1<<17) + +#define PRIM3D_TRILIST (0x0<<18) +#define PRIM3D_TRISTRIP (0x1<<18) +#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) +#define PRIM3D_TRIFAN (0x3<<18) +#define PRIM3D_POLY (0x4<<18) +#define PRIM3D_LINELIST (0x5<<18) +#define PRIM3D_LINESTRIP (0x6<<18) +#define PRIM3D_RECTLIST (0x7<<18) +#define PRIM3D_POINTLIST (0x8<<18) +#define PRIM3D_DIB (0x9<<18) +#define PRIM3D_MASK (0x1f<<18) + +#define I915PACKCOLOR4444(r,g,b,a) \ + ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) + +#define I915PACKCOLOR1555(r,g,b,a) \ + ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ + ((a) ? 0x8000 : 0)) + +#define I915PACKCOLOR565(r,g,b) \ + ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) + +#define I915PACKCOLOR8888(r,g,b,a) \ + ((a<<24) | (r<<16) | (g<<8) | b) + + + + +#define BR00_BITBLT_CLIENT 0x40000000 +#define BR00_OP_COLOR_BLT 0x10000000 +#define BR00_OP_SRC_COPY_BLT 0x10C00000 +#define BR13_SOLID_PATTERN 0x80000000 + +#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) +#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) +#define XY_COLOR_BLT_WRITE_RGB (1<<20) + +#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) +#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) +#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) + +#define MI_WAIT_FOR_EVENT ((0x3<<23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +#define MI_BATCH_BUFFER (0x30<<23) +#define MI_BATCH_BUFFER_START (0x31<<23) +#define MI_BATCH_BUFFER_END (0xa<<23) + + + +#define COMPAREFUNC_ALWAYS 0 +#define COMPAREFUNC_NEVER 0x1 +#define COMPAREFUNC_LESS 0x2 +#define COMPAREFUNC_EQUAL 0x3 +#define COMPAREFUNC_LEQUAL 0x4 +#define COMPAREFUNC_GREATER 0x5 +#define COMPAREFUNC_NOTEQUAL 0x6 +#define COMPAREFUNC_GEQUAL 0x7 + +#define STENCILOP_KEEP 0 +#define STENCILOP_ZERO 0x1 +#define STENCILOP_REPLACE 0x2 +#define STENCILOP_INCRSAT 0x3 +#define STENCILOP_DECRSAT 0x4 +#define STENCILOP_INCR 0x5 +#define STENCILOP_DECR 0x6 +#define STENCILOP_INVERT 0x7 + +#define LOGICOP_CLEAR 0 +#define LOGICOP_NOR 0x1 +#define LOGICOP_AND_INV 0x2 +#define LOGICOP_COPY_INV 0x3 +#define LOGICOP_AND_RVRSE 0x4 +#define LOGICOP_INV 0x5 +#define LOGICOP_XOR 0x6 +#define LOGICOP_NAND 0x7 +#define LOGICOP_AND 0x8 +#define LOGICOP_EQUIV 0x9 +#define LOGICOP_NOOP 0xa +#define LOGICOP_OR_INV 0xb +#define LOGICOP_COPY 0xc +#define LOGICOP_OR_RVRSE 0xd +#define LOGICOP_OR 0xe +#define LOGICOP_SET 0xf + +#define BLENDFACT_ZERO 0x01 +#define BLENDFACT_ONE 0x02 +#define BLENDFACT_SRC_COLR 0x03 +#define BLENDFACT_INV_SRC_COLR 0x04 +#define BLENDFACT_SRC_ALPHA 0x05 +#define BLENDFACT_INV_SRC_ALPHA 0x06 +#define BLENDFACT_DST_ALPHA 0x07 +#define BLENDFACT_INV_DST_ALPHA 0x08 +#define BLENDFACT_DST_COLR 0x09 +#define BLENDFACT_INV_DST_COLR 0x0a +#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b +#define BLENDFACT_CONST_COLOR 0x0c +#define BLENDFACT_INV_CONST_COLOR 0x0d +#define BLENDFACT_CONST_ALPHA 0x0e +#define BLENDFACT_INV_CONST_ALPHA 0x0f +#define BLENDFACT_MASK 0x0f + +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GME 0x27AE +#define PCI_CHIP_G33_G 0x29C2 +#define PCI_CHIP_Q35_G 0x29B2 +#define PCI_CHIP_Q33_G 0x29D2 + + +#endif diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c new file mode 100644 index 0000000000..abd5571b88 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -0,0 +1,694 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + + +#include "pipe/draw/draw_context.h" +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" + +#include "i915_context.h" +#include "i915_reg.h" +#include "i915_state.h" +#include "i915_state_inlines.h" + + +/* The i915 (and related graphics cores) do not support GL_CLAMP. The + * Intel drivers for "other operating systems" implement GL_CLAMP as + * GL_CLAMP_TO_EDGE, so the same is done here. + */ +static unsigned +translate_wrap_mode(unsigned wrap) +{ + switch (wrap) { + case PIPE_TEX_WRAP_REPEAT: + return TEXCOORDMODE_WRAP; + case PIPE_TEX_WRAP_CLAMP: + return TEXCOORDMODE_CLAMP_EDGE; /* not quite correct */ + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return TEXCOORDMODE_CLAMP_EDGE; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return TEXCOORDMODE_CLAMP_BORDER; +// case PIPE_TEX_WRAP_MIRRORED_REPEAT: +// return TEXCOORDMODE_MIRROR; + default: + return TEXCOORDMODE_WRAP; + } +} + +static unsigned translate_img_filter( unsigned filter ) +{ + switch (filter) { + case PIPE_TEX_FILTER_NEAREST: + return FILTER_NEAREST; + case PIPE_TEX_FILTER_LINEAR: + return FILTER_LINEAR; + default: + assert(0); + return FILTER_NEAREST; + } +} + +static unsigned translate_mip_filter( unsigned filter ) +{ + switch (filter) { + case PIPE_TEX_MIPFILTER_NONE: + return MIPFILTER_NONE; + case PIPE_TEX_MIPFILTER_NEAREST: + return MIPFILTER_NEAREST; + case PIPE_TEX_FILTER_LINEAR: + return MIPFILTER_LINEAR; + default: + assert(0); + return MIPFILTER_NONE; + } +} + + +/* None of this state is actually used for anything yet. + */ +static void * +i915_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + struct i915_blend_state *cso_data = CALLOC_STRUCT( i915_blend_state ); + + { + unsigned eqRGB = blend->rgb_func; + unsigned srcRGB = blend->rgb_src_factor; + unsigned dstRGB = blend->rgb_dst_factor; + + unsigned eqA = blend->alpha_func; + unsigned srcA = blend->alpha_src_factor; + unsigned dstA = blend->alpha_dst_factor; + + /* Special handling for MIN/MAX filter modes handled at + * state_tracker level. + */ + + if (srcA != srcRGB || + dstA != dstRGB || + eqA != eqRGB) { + + cso_data->iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | + IAB_MODIFY_ENABLE | + IAB_ENABLE | + IAB_MODIFY_FUNC | + IAB_MODIFY_SRC_FACTOR | + IAB_MODIFY_DST_FACTOR | + SRC_ABLND_FACT(i915_translate_blend_factor(srcA)) | + DST_ABLND_FACT(i915_translate_blend_factor(dstA)) | + (i915_translate_blend_func(eqA) << IAB_FUNC_SHIFT)); + } + else { + cso_data->iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | + IAB_MODIFY_ENABLE | + 0); + } + } + + cso_data->modes4 |= (_3DSTATE_MODES_4_CMD | + ENABLE_LOGIC_OP_FUNC | + LOGIC_OP_FUNC(i915_translate_logic_op(blend->logicop_func))); + + if (blend->logicop_enable) + cso_data->LIS5 |= S5_LOGICOP_ENABLE; + + if (blend->dither) + cso_data->LIS5 |= S5_COLOR_DITHER_ENABLE; + + if ((blend->colormask & PIPE_MASK_R) == 0) + cso_data->LIS5 |= S5_WRITEDISABLE_RED; + + if ((blend->colormask & PIPE_MASK_G) == 0) + cso_data->LIS5 |= S5_WRITEDISABLE_GREEN; + + if ((blend->colormask & PIPE_MASK_B) == 0) + cso_data->LIS5 |= S5_WRITEDISABLE_BLUE; + + if ((blend->colormask & PIPE_MASK_A) == 0) + cso_data->LIS5 |= S5_WRITEDISABLE_ALPHA; + + if (blend->blend_enable) { + unsigned funcRGB = blend->rgb_func; + unsigned srcRGB = blend->rgb_src_factor; + unsigned dstRGB = blend->rgb_dst_factor; + + cso_data->LIS6 |= (S6_CBUF_BLEND_ENABLE | + SRC_BLND_FACT(i915_translate_blend_factor(srcRGB)) | + DST_BLND_FACT(i915_translate_blend_factor(dstRGB)) | + (i915_translate_blend_func(funcRGB) << S6_CBUF_BLEND_FUNC_SHIFT)); + } + + return cso_data; +} + +static void i915_bind_blend_state(struct pipe_context *pipe, + void *blend) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->blend = (struct i915_blend_state*)blend; + + i915->dirty |= I915_NEW_BLEND; +} + + +static void i915_delete_blend_state(struct pipe_context *pipe, void *blend) +{ + FREE(blend); +} + +static void i915_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->blend_color = *blend_color; + + i915->dirty |= I915_NEW_BLEND; +} + +static void * +i915_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + struct i915_sampler_state *cso = CALLOC_STRUCT( i915_sampler_state ); + const unsigned ws = sampler->wrap_s; + const unsigned wt = sampler->wrap_t; + const unsigned wr = sampler->wrap_r; + unsigned minFilt, magFilt; + unsigned mipFilt; + + cso->templ = sampler; + + mipFilt = translate_mip_filter(sampler->min_mip_filter); + if (sampler->max_anisotropy > 1.0) { + minFilt = FILTER_ANISOTROPIC; + magFilt = FILTER_ANISOTROPIC; + if (sampler->max_anisotropy > 2.0) { + cso->state[0] |= SS2_MAX_ANISO_4; + } + } + else { + minFilt = translate_img_filter( sampler->min_img_filter ); + magFilt = translate_img_filter( sampler->mag_img_filter ); + } + + { + int b = (int) (sampler->lod_bias * 16.0); + b = CLAMP(b, -256, 255); + cso->state[0] |= ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); + } + + /* Shadow: + */ + if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) + { + cso->state[0] |= (SS2_SHADOW_ENABLE | + i915_translate_compare_func(sampler->compare_func)); + + minFilt = FILTER_4X4_FLAT; + magFilt = FILTER_4X4_FLAT; + } + + cso->state[0] |= ((minFilt << SS2_MIN_FILTER_SHIFT) | + (mipFilt << SS2_MIP_FILTER_SHIFT) | + (magFilt << SS2_MAG_FILTER_SHIFT)); + + cso->state[1] |= + ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | + (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT)); + + if (sampler->normalized_coords) + cso->state[1] |= SS3_NORMALIZED_COORDS; + + if (0) /* XXX not tested yet */ + { + int minlod = (int) (16.0 * sampler->min_lod); + minlod = CLAMP(minlod, 0, 16 * 11); + cso->state[1] |= (minlod << SS3_MIN_LOD_SHIFT); + } + + { + ubyte r = float_to_ubyte(sampler->border_color[0]); + ubyte g = float_to_ubyte(sampler->border_color[1]); + ubyte b = float_to_ubyte(sampler->border_color[2]); + ubyte a = float_to_ubyte(sampler->border_color[3]); + cso->state[2] = I915PACKCOLOR8888(r, g, b, a); + } + return cso; +} + +static void i915_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct i915_context *i915 = i915_context(pipe); + + assert(unit < PIPE_MAX_SAMPLERS); + i915->sampler[unit] = (const struct i915_sampler_state*)sampler; + + i915->dirty |= I915_NEW_SAMPLER; +} + +static void i915_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + FREE(sampler); +} + + +/** XXX move someday? Or consolidate all these simple state setters + * into one file. + */ + +static void * +i915_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil) +{ + struct i915_depth_stencil_state *cso = CALLOC_STRUCT( i915_depth_stencil_state ); + + { + int testmask = depth_stencil->stencil[0].value_mask & 0xff; + int writemask = depth_stencil->stencil[0].write_mask & 0xff; + + cso->stencil_modes4 |= (_3DSTATE_MODES_4_CMD | + ENABLE_STENCIL_TEST_MASK | + STENCIL_TEST_MASK(testmask) | + ENABLE_STENCIL_WRITE_MASK | + STENCIL_WRITE_MASK(writemask)); + } + + if (depth_stencil->stencil[0].enabled) { + int test = i915_translate_compare_func(depth_stencil->stencil[0].func); + int fop = i915_translate_stencil_op(depth_stencil->stencil[0].fail_op); + int dfop = i915_translate_stencil_op(depth_stencil->stencil[0].zfail_op); + int dpop = i915_translate_stencil_op(depth_stencil->stencil[0].zpass_op); + int ref = depth_stencil->stencil[0].ref_value & 0xff; + + cso->stencil_LIS5 |= (S5_STENCIL_TEST_ENABLE | + S5_STENCIL_WRITE_ENABLE | + (ref << S5_STENCIL_REF_SHIFT) | + (test << S5_STENCIL_TEST_FUNC_SHIFT) | + (fop << S5_STENCIL_FAIL_SHIFT) | + (dfop << S5_STENCIL_PASS_Z_FAIL_SHIFT) | + (dpop << S5_STENCIL_PASS_Z_PASS_SHIFT)); + } + + if (depth_stencil->stencil[1].enabled) { + int test = i915_translate_compare_func(depth_stencil->stencil[1].func); + int fop = i915_translate_stencil_op(depth_stencil->stencil[1].fail_op); + int dfop = i915_translate_stencil_op(depth_stencil->stencil[1].zfail_op); + int dpop = i915_translate_stencil_op(depth_stencil->stencil[1].zpass_op); + int ref = depth_stencil->stencil[1].ref_value & 0xff; + int tmask = depth_stencil->stencil[1].value_mask & 0xff; + int wmask = depth_stencil->stencil[1].write_mask & 0xff; + + cso->bfo[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | + BFO_ENABLE_STENCIL_FUNCS | + BFO_ENABLE_STENCIL_TWO_SIDE | + BFO_ENABLE_STENCIL_REF | + BFO_STENCIL_TWO_SIDE | + (ref << BFO_STENCIL_REF_SHIFT) | + (test << BFO_STENCIL_TEST_SHIFT) | + (fop << BFO_STENCIL_FAIL_SHIFT) | + (dfop << BFO_STENCIL_PASS_Z_FAIL_SHIFT) | + (dpop << BFO_STENCIL_PASS_Z_PASS_SHIFT)); + + cso->bfo[1] = (_3DSTATE_BACKFACE_STENCIL_MASKS | + BFM_ENABLE_STENCIL_TEST_MASK | + BFM_ENABLE_STENCIL_WRITE_MASK | + (tmask << BFM_STENCIL_TEST_MASK_SHIFT) | + (wmask << BFM_STENCIL_WRITE_MASK_SHIFT)); + } + else { + /* This actually disables two-side stencil: The bit set is a + * modify-enable bit to indicate we are changing the two-side + * setting. Then there is a symbolic zero to show that we are + * setting the flag to zero/off. + */ + cso->bfo[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | + BFO_ENABLE_STENCIL_TWO_SIDE | + 0); + cso->bfo[1] = 0; + } + + if (depth_stencil->depth.enabled) { + int func = i915_translate_compare_func(depth_stencil->depth.func); + + cso->depth_LIS6 |= (S6_DEPTH_TEST_ENABLE | + (func << S6_DEPTH_TEST_FUNC_SHIFT)); + + if (depth_stencil->depth.writemask) + cso->depth_LIS6 |= S6_DEPTH_WRITE_ENABLE; + } + + if (depth_stencil->alpha.enabled) { + int test = i915_translate_compare_func(depth_stencil->alpha.func); + ubyte refByte = float_to_ubyte(depth_stencil->alpha.ref); + + cso->depth_LIS6 |= (S6_ALPHA_TEST_ENABLE | + (test << S6_ALPHA_TEST_FUNC_SHIFT) | + (((unsigned) refByte) << S6_ALPHA_REF_SHIFT)); + } + + return cso; +} + +static void i915_bind_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->depth_stencil = (const struct i915_depth_stencil_state *)depth_stencil; + + i915->dirty |= I915_NEW_DEPTH_STENCIL; +} + +static void i915_delete_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + FREE(depth_stencil); +} + + +static void i915_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct i915_context *i915 = i915_context(pipe); + + memcpy( &i915->scissor, scissor, sizeof(*scissor) ); + i915->dirty |= I915_NEW_SCISSOR; +} + + +static void i915_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ +} + + +static void * i915_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + return 0; +} + +static void i915_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->fs = (struct pipe_shader_state *)fs; + + i915->dirty |= I915_NEW_FS; +} + +static void i915_delete_fs_state(struct pipe_context *pipe, void *shader) +{ + /*do nothing*/ +} + +static void * +i915_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct i915_context *i915 = i915_context(pipe); + + /* just pass-through to draw module */ + return draw_create_vertex_shader(i915->draw, templ); +} + +static void i915_bind_vs_state(struct pipe_context *pipe, void *shader) +{ + struct i915_context *i915 = i915_context(pipe); + + /* just pass-through to draw module */ + draw_bind_vertex_shader(i915->draw, (struct draw_vertex_shader *) shader); +} + +static void i915_delete_vs_state(struct pipe_context *pipe, void *shader) +{ + struct i915_context *i915 = i915_context(pipe); + + /* just pass-through to draw module */ + draw_delete_vertex_shader(i915->draw, (struct draw_vertex_shader *) shader); +} + +static void i915_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct i915_context *i915 = i915_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + /* Make a copy of shader constants. + * During fragment program translation we may add additional + * constants to the array. + * + * We want to consider the situation where some user constants + * (ex: a material color) may change frequently but the shader program + * stays the same. In that case we should only be updating the first + * N constants, leaving any extras from shader translation alone. + */ + { + void *mapped; + if (buf->size && + (mapped = ws->buffer_map(ws, buf->buffer, + PIPE_BUFFER_USAGE_CPU_READ))) { + memcpy(i915->current.constants[shader], mapped, buf->size); + ws->buffer_unmap(ws, buf->buffer); + i915->current.num_user_constants[shader] + = buf->size / (4 * sizeof(float)); + } + else { + i915->current.num_user_constants[shader] = 0; + } + } + + i915->dirty |= I915_NEW_CONSTANTS; +} + + +static void i915_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, + struct pipe_texture *texture) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->texture[sampler] = (struct i915_texture*)texture; /* ptr, not struct */ + + i915->dirty |= I915_NEW_TEXTURE; +} + + + +static void i915_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->framebuffer = *fb; /* struct copy */ + + i915->dirty |= I915_NEW_FRAMEBUFFER; +} + + + +static void i915_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct i915_context *i915 = i915_context(pipe); + + draw_set_clip_state(i915->draw, clip); + + i915->dirty |= I915_NEW_CLIP; +} + + + +/* Called when driver state tracker notices changes to the viewport + * matrix: + */ +static void i915_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->viewport = *viewport; /* struct copy */ + + /* pass the viewport info to the draw module */ + draw_set_viewport_state(i915->draw, &i915->viewport); + + i915->dirty |= I915_NEW_VIEWPORT; +} + + +static void * +i915_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *rasterizer) +{ + struct i915_rasterizer_state *cso = CALLOC_STRUCT( i915_rasterizer_state ); + + cso->templ = rasterizer; + cso->color_interp = rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + cso->light_twoside = rasterizer->light_twoside; + cso->ds[0].u = _3DSTATE_DEPTH_OFFSET_SCALE; + cso->ds[1].f = rasterizer->offset_scale; + if (rasterizer->poly_stipple_enable) { + cso->st |= ST1_ENABLE; + } + + if (rasterizer->scissor) + cso->sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT; + else + cso->sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT; + + switch (rasterizer->cull_mode) { + case PIPE_WINDING_NONE: + cso->LIS4 |= S4_CULLMODE_NONE; + break; + case PIPE_WINDING_CW: + cso->LIS4 |= S4_CULLMODE_CW; + break; + case PIPE_WINDING_CCW: + cso->LIS4 |= S4_CULLMODE_CCW; + break; + case PIPE_WINDING_BOTH: + cso->LIS4 |= S4_CULLMODE_BOTH; + break; + } + + { + int line_width = CLAMP((int)(rasterizer->line_width * 2), 1, 0xf); + + cso->LIS4 |= line_width << S4_LINE_WIDTH_SHIFT; + + if (rasterizer->line_smooth) + cso->LIS4 |= S4_LINE_ANTIALIAS_ENABLE; + } + + { + int point_size = CLAMP((int) rasterizer->point_size, 1, 0xff); + + cso->LIS4 |= point_size << S4_POINT_WIDTH_SHIFT; + } + + if (rasterizer->flatshade) { + cso->LIS4 |= (S4_FLATSHADE_ALPHA | + S4_FLATSHADE_COLOR | + S4_FLATSHADE_SPECULAR); + } + + cso->LIS7 = fui( rasterizer->offset_units ); + + + return cso; +} + +static void i915_bind_rasterizer_state( struct pipe_context *pipe, + void *setup ) +{ + struct i915_context *i915 = i915_context(pipe); + + i915->rasterizer = (struct i915_rasterizer_state *)setup; + + /* pass-through to draw module */ + draw_set_rasterizer_state(i915->draw, i915->rasterizer->templ); + + i915->dirty |= I915_NEW_RASTERIZER; +} + +static void i915_delete_rasterizer_state(struct pipe_context *pipe, + void *setup) +{ + FREE(setup); +} + +static void i915_set_vertex_buffer( struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_buffer *buffer ) +{ + struct i915_context *i915 = i915_context(pipe); + i915->vertex_buffer[index] = *buffer; + /* pass-through to draw module */ + draw_set_vertex_buffer(i915->draw, index, buffer); +} + +static void i915_set_vertex_element( struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_element *element) +{ + struct i915_context *i915 = i915_context(pipe); + /* pass-through to draw module */ + draw_set_vertex_element(i915->draw, index, element); +} + + + +void +i915_init_state_functions( struct i915_context *i915 ) +{ + i915->pipe.create_blend_state = i915_create_blend_state; + i915->pipe.bind_blend_state = i915_bind_blend_state; + i915->pipe.delete_blend_state = i915_delete_blend_state; + + i915->pipe.create_sampler_state = i915_create_sampler_state; + i915->pipe.bind_sampler_state = i915_bind_sampler_state; + i915->pipe.delete_sampler_state = i915_delete_sampler_state; + + i915->pipe.create_depth_stencil_alpha_state = i915_create_depth_stencil_state; + i915->pipe.bind_depth_stencil_alpha_state = i915_bind_depth_stencil_state; + i915->pipe.delete_depth_stencil_alpha_state = i915_delete_depth_stencil_state; + + i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; + i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; + i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state; + i915->pipe.create_fs_state = i915_create_fs_state; + i915->pipe.bind_fs_state = i915_bind_fs_state; + i915->pipe.delete_fs_state = i915_delete_fs_state; + i915->pipe.create_vs_state = i915_create_vs_state; + i915->pipe.bind_vs_state = i915_bind_vs_state; + i915->pipe.delete_vs_state = i915_delete_vs_state; + + i915->pipe.set_blend_color = i915_set_blend_color; + i915->pipe.set_clip_state = i915_set_clip_state; + i915->pipe.set_constant_buffer = i915_set_constant_buffer; + i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; + + i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; + i915->pipe.set_scissor_state = i915_set_scissor_state; + i915->pipe.set_sampler_texture = i915_set_sampler_texture; + i915->pipe.set_viewport_state = i915_set_viewport_state; + i915->pipe.set_vertex_buffer = i915_set_vertex_buffer; + i915->pipe.set_vertex_element = i915_set_vertex_element; +} diff --git a/src/gallium/drivers/i915simple/i915_state.h b/src/gallium/drivers/i915simple/i915_state.h new file mode 100644 index 0000000000..86c6b0027d --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_state.h @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef I915_STATE_H +#define I915_STATE_H + +struct i915_context; + + +struct i915_tracked_state { + unsigned dirty; + void (*update)( struct i915_context * ); +}; + +void i915_update_immediate( struct i915_context *i915 ); +void i915_update_dynamic( struct i915_context *i915 ); +void i915_update_derived( struct i915_context *i915 ); +void i915_update_samplers( struct i915_context *i915 ); +void i915_update_textures(struct i915_context *i915); + +void i915_emit_hardware_state( struct i915_context *i915 ); + +#endif diff --git a/src/gallium/drivers/i915simple/i915_state_derived.c b/src/gallium/drivers/i915simple/i915_state_derived.c new file mode 100644 index 0000000000..653983e4a9 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_state_derived.c @@ -0,0 +1,177 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_vertex.h" +#include "i915_context.h" +#include "i915_state.h" +#include "i915_reg.h" +#include "i915_fpc.h" +#include "pipe/p_shader_tokens.h" + + +/** + * Determine which post-transform / pre-rasterization vertex attributes + * we need. + * Derived from: fs, setup states. + */ +static void calculate_vertex_layout( struct i915_context *i915 ) +{ + const struct pipe_shader_state *fs = i915->fs; + const enum interp_mode colorInterp = i915->rasterizer->color_interp; + struct vertex_info vinfo; + uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; + boolean needW = 0; + uint i; + boolean texCoords[8]; + uint src = 0; + + memset(texCoords, 0, sizeof(texCoords)); + memset(&vinfo, 0, sizeof(vinfo)); + + /* pos */ + draw_emit_vertex_attr(&vinfo, EMIT_3F, INTERP_LINEAR, src++); + /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ + + for (i = 0; i < fs->num_inputs; i++) { + switch (fs->input_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + break; + case TGSI_SEMANTIC_COLOR: + if (fs->input_semantic_index[i] == 0) { + front0 = draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src++); + vinfo.hwfmt[0] |= S4_VFMT_COLOR; + } + else { + assert(fs->input_semantic_index[i] == 1); + front1 = draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src++); + vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG; + } + break; + case TGSI_SEMANTIC_GENERIC: + /* usually a texcoord */ + { + const uint unit = fs->input_semantic_index[i]; + uint hwtc; + texCoords[unit] = TRUE; + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); + hwtc = TEXCOORDFMT_4D; + needW = TRUE; + vinfo.hwfmt[1] |= hwtc << (unit * 4); + } + break; + case TGSI_SEMANTIC_FOG: + debug_printf("i915 fogcoord not implemented yet\n"); + draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_PERSPECTIVE, src++); + break; + default: + assert(0); + } + + } + + /* finish up texcoord fields */ + for (i = 0; i < 8; i++) { + if (!texCoords[i]) { + const uint hwtc = TEXCOORDFMT_NOT_PRESENT; + vinfo.hwfmt[1] |= hwtc << (i* 4); + } + } + + /* go back and fill in the vertex position info now that we have needW */ + if (needW) { + vinfo.hwfmt[0] |= S4_VFMT_XYZW; + vinfo.emit[0] = EMIT_4F; + } + else { + vinfo.hwfmt[0] |= S4_VFMT_XYZ; + vinfo.emit[0] = EMIT_3F; + } + + /* Additional attributes required for setup: Just twosided + * lighting. Edgeflag is dealt with specially by setting bits in + * the vertex header. + */ + if (i915->rasterizer->light_twoside) { + if (front0) { + back0 = draw_emit_vertex_attr(&vinfo, EMIT_OMIT, colorInterp, src++); + } + if (back0) { + back1 = draw_emit_vertex_attr(&vinfo, EMIT_OMIT, colorInterp, src++); + } + } + + draw_compute_vertex_size(&vinfo); + + if (memcmp(&i915->current.vertex_info, &vinfo, sizeof(vinfo))) { + /* Need to set this flag so that the LIS2/4 registers get set. + * It also means the i915_update_immediate() function must be called + * after this one, in i915_update_derived(). + */ + i915->dirty |= I915_NEW_VERTEX_FORMAT; + + memcpy(&i915->current.vertex_info, &vinfo, sizeof(vinfo)); + } +} + + + + +/* Hopefully this will remain quite simple, otherwise need to pull in + * something like the state tracker mechanism. + */ +void i915_update_derived( struct i915_context *i915 ) +{ + if (i915->dirty & (I915_NEW_RASTERIZER | I915_NEW_FS)) + calculate_vertex_layout( i915 ); + + if (i915->dirty & (I915_NEW_SAMPLER | I915_NEW_TEXTURE)) + i915_update_samplers(i915); + + if (i915->dirty & I915_NEW_TEXTURE) + i915_update_textures(i915); + + if (i915->dirty) + i915_update_immediate( i915 ); + + if (i915->dirty) + i915_update_dynamic( i915 ); + + if (i915->dirty & I915_NEW_FS) { + i915_translate_fragment_program(i915); + i915->hardware_dirty |= I915_HW_PROGRAM; /* XXX right? */ + } + + /* HW emit currently references framebuffer state directly: + */ + if (i915->dirty & I915_NEW_FRAMEBUFFER) + i915->hardware_dirty |= I915_HW_STATIC; + + i915->dirty = 0; +} diff --git a/src/gallium/drivers/i915simple/i915_state_dynamic.c b/src/gallium/drivers/i915simple/i915_state_dynamic.c new file mode 100644 index 0000000000..8cfbdddd19 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_state_dynamic.c @@ -0,0 +1,308 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_batch.h" +#include "i915_state_inlines.h" +#include "i915_context.h" +#include "i915_reg.h" +#include "i915_state.h" +#include "pipe/p_util.h" + +#define FILE_DEBUG_FLAG DEBUG_STATE + +/* State that we have chosen to store in the DYNAMIC segment of the + * i915 indirect state mechanism. + * + * Can't cache these in the way we do the static state, as there is no + * start/size in the command packet, instead an 'end' value that gets + * incremented. + * + * Additionally, there seems to be a requirement to re-issue the full + * (active) state every time a 4kb boundary is crossed. + */ + +static INLINE void set_dynamic_indirect( struct i915_context *i915, + unsigned offset, + const unsigned *src, + unsigned dwords ) +{ + unsigned i; + + for (i = 0; i < dwords; i++) + i915->current.dynamic[offset + i] = src[i]; + + i915->hardware_dirty |= I915_HW_DYNAMIC; +} + + +/*********************************************************************** + * Modes4: stencil masks and logicop + */ +static void upload_MODES4( struct i915_context *i915 ) +{ + unsigned modes4 = 0; + + /* I915_NEW_STENCIL */ + modes4 |= i915->depth_stencil->stencil_modes4; + /* I915_NEW_BLEND */ + modes4 |= i915->blend->modes4; + + /* Always, so that we know when state is in-active: + */ + set_dynamic_indirect( i915, + I915_DYNAMIC_MODES4, + &modes4, + 1 ); +} + +const struct i915_tracked_state i915_upload_MODES4 = { + I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, + upload_MODES4 +}; + + + + +/*********************************************************************** + */ + +static void upload_BFO( struct i915_context *i915 ) +{ + set_dynamic_indirect( i915, + I915_DYNAMIC_BFO_0, + &(i915->depth_stencil->bfo[0]), + 2 ); +} + +const struct i915_tracked_state i915_upload_BFO = { + I915_NEW_DEPTH_STENCIL, + upload_BFO +}; + + +/*********************************************************************** + */ + + +static void upload_BLENDCOLOR( struct i915_context *i915 ) +{ + unsigned bc[2]; + + memset( bc, 0, sizeof(bc) ); + + /* I915_NEW_BLEND {_COLOR} + */ + { + const float *color = i915->blend_color.color; + + bc[0] = _3DSTATE_CONST_BLEND_COLOR_CMD; + bc[1] = pack_ui32_float4( color[0], + color[1], + color[2], + color[3] ); + } + + set_dynamic_indirect( i915, + I915_DYNAMIC_BC_0, + bc, + 2 ); +} + +const struct i915_tracked_state i915_upload_BLENDCOLOR = { + I915_NEW_BLEND, + upload_BLENDCOLOR +}; + +/*********************************************************************** + */ + + +static void upload_IAB( struct i915_context *i915 ) +{ + unsigned iab = i915->blend->iab; + + + set_dynamic_indirect( i915, + I915_DYNAMIC_IAB, + &iab, + 1 ); +} + +const struct i915_tracked_state i915_upload_IAB = { + I915_NEW_BLEND, + upload_IAB +}; + + +/*********************************************************************** + */ + + + +static void upload_DEPTHSCALE( struct i915_context *i915 ) +{ + set_dynamic_indirect( i915, + I915_DYNAMIC_DEPTHSCALE_0, + &(i915->rasterizer->ds[0].u), + 2 ); +} + +const struct i915_tracked_state i915_upload_DEPTHSCALE = { + I915_NEW_RASTERIZER, + upload_DEPTHSCALE +}; + + + +/*********************************************************************** + * Polygon stipple + * + * The i915 supports a 4x4 stipple natively, GL wants 32x32. + * Fortunately stipple is usually a repeating pattern. + * + * XXX: does stipple pattern need to be adjusted according to + * the window position? + * + * XXX: possibly need workaround for conform paths test. + */ + +static void upload_STIPPLE( struct i915_context *i915 ) +{ + unsigned st[2]; + + st[0] = _3DSTATE_STIPPLE; + st[1] = 0; + + /* I915_NEW_RASTERIZER + */ + st[1] |= i915->rasterizer->st; + + + /* I915_NEW_STIPPLE + */ + { + const ubyte *mask = (const ubyte *)i915->poly_stipple.stipple; + ubyte p[4]; + + p[0] = mask[12] & 0xf; + p[1] = mask[8] & 0xf; + p[2] = mask[4] & 0xf; + p[3] = mask[0] & 0xf; + + /* Not sure what to do about fallbacks, so for now just dont: + */ + st[1] |= ((p[0] << 0) | + (p[1] << 4) | + (p[2] << 8) | + (p[3] << 12)); + } + + + set_dynamic_indirect( i915, + I915_DYNAMIC_STP_0, + &st[0], + 2 ); +} + + +const struct i915_tracked_state i915_upload_STIPPLE = { + I915_NEW_RASTERIZER | I915_NEW_STIPPLE, + upload_STIPPLE +}; + + + +/*********************************************************************** + * Scissor. + */ +static void upload_SCISSOR_ENABLE( struct i915_context *i915 ) +{ + set_dynamic_indirect( i915, + I915_DYNAMIC_SC_ENA_0, + &(i915->rasterizer->sc[0]), + 1 ); +} + +const struct i915_tracked_state i915_upload_SCISSOR_ENABLE = { + I915_NEW_RASTERIZER, + upload_SCISSOR_ENABLE +}; + + + +static void upload_SCISSOR_RECT( struct i915_context *i915 ) +{ + unsigned x1 = i915->scissor.minx; + unsigned y1 = i915->scissor.miny; + unsigned x2 = i915->scissor.maxx; + unsigned y2 = i915->scissor.maxy; + unsigned sc[3]; + + sc[0] = _3DSTATE_SCISSOR_RECT_0_CMD; + sc[1] = (y1 << 16) | (x1 & 0xffff); + sc[2] = (y2 << 16) | (x2 & 0xffff); + + set_dynamic_indirect( i915, + I915_DYNAMIC_SC_RECT_0, + &sc[0], + 3 ); +} + + +const struct i915_tracked_state i915_upload_SCISSOR_RECT = { + I915_NEW_SCISSOR, + upload_SCISSOR_RECT +}; + + + + + + +static const struct i915_tracked_state *atoms[] = { + &i915_upload_MODES4, + &i915_upload_BFO, + &i915_upload_BLENDCOLOR, + &i915_upload_IAB, + &i915_upload_DEPTHSCALE, + &i915_upload_STIPPLE, + &i915_upload_SCISSOR_ENABLE, + &i915_upload_SCISSOR_RECT +}; + +/* These will be dynamic indirect state commands, but for now just end + * up on the batch buffer with everything else. + */ +void i915_update_dynamic( struct i915_context *i915 ) +{ + int i; + + for (i = 0; i < Elements(atoms); i++) + if (i915->dirty & atoms[i]->dirty) + atoms[i]->update( i915 ); +} + diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c new file mode 100644 index 0000000000..3339287f49 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -0,0 +1,374 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "i915_reg.h" +#include "i915_context.h" +#include "i915_winsys.h" +#include "i915_batch.h" +#include "i915_reg.h" + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + +static unsigned translate_format( enum pipe_format format ) +{ + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + return COLOR_BUF_ARGB8888; + case PIPE_FORMAT_R5G6B5_UNORM: + return COLOR_BUF_RGB565; + default: + assert(0); + return 0; + } +} + +static unsigned translate_depth_format( enum pipe_format zformat ) +{ + switch (zformat) { + case PIPE_FORMAT_S8Z24_UNORM: + return DEPTH_FRMT_24_FIXED_8_OTHER; + case PIPE_FORMAT_Z16_UNORM: + return DEPTH_FRMT_16_FIXED; + default: + assert(0); + return 0; + } +} + + +/** + * Examine framebuffer state to determine width, height. + */ +static boolean +framebuffer_size(const struct pipe_framebuffer_state *fb, + uint *width, uint *height) +{ + if (fb->cbufs[0]) { + *width = fb->cbufs[0]->width; + *height = fb->cbufs[0]->height; + return TRUE; + } + else if (fb->zsbuf) { + *width = fb->zsbuf->width; + *height = fb->zsbuf->height; + return TRUE; + } + else { + *width = *height = 0; + return FALSE; + } +} + + +/* Push the state into the sarea and/or texture memory. + */ +void +i915_emit_hardware_state(struct i915_context *i915 ) +{ + /* XXX: there must be an easier way */ + const unsigned dwords = ( 14 + + 7 + + I915_MAX_DYNAMIC + + 8 + + 2 + I915_TEX_UNITS*3 + + 2 + I915_TEX_UNITS*3 + + 2 + I915_MAX_CONSTANT*4 + + i915->current.program_len + + 6 + ) * 3/2; /* plus 50% margin */ + const unsigned relocs = ( I915_TEX_UNITS + + 3 + ) * 3/2; /* plus 50% margin */ + +#if 0 + debug_printf("i915_emit_hardware_state: %d dwords, %d relocs\n", dwords, relocs); +#endif + + if(!BEGIN_BATCH(dwords, relocs)) { + FLUSH_BATCH(); + assert(BEGIN_BATCH(dwords, relocs)); + } + + /* 14 dwords, 0 relocs */ + if (i915->hardware_dirty & I915_HW_INVARIENT) + { + OUT_BATCH(_3DSTATE_AA_CMD | + AA_LINE_ECAAR_WIDTH_ENABLE | + AA_LINE_ECAAR_WIDTH_1_0 | + AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0); + + OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); + OUT_BATCH(0); + + OUT_BATCH(_3DSTATE_DFLT_SPEC_CMD); + OUT_BATCH(0); + + OUT_BATCH(_3DSTATE_DFLT_Z_CMD); + OUT_BATCH(0); + + OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS | + CSB_TCB(0, 0) | + CSB_TCB(1, 1) | + CSB_TCB(2, 2) | + CSB_TCB(3, 3) | + CSB_TCB(4, 4) | + CSB_TCB(5, 5) | + CSB_TCB(6, 6) | + CSB_TCB(7, 7)); + + OUT_BATCH(_3DSTATE_RASTER_RULES_CMD | + ENABLE_POINT_RASTER_RULE | + OGL_POINT_RASTER_RULE | + ENABLE_LINE_STRIP_PROVOKE_VRTX | + ENABLE_TRI_FAN_PROVOKE_VRTX | + LINE_STRIP_PROVOKE_VRTX(1) | + TRI_FAN_PROVOKE_VRTX(2) | + ENABLE_TEXKILL_3D_4D | + TEXKILL_4D); + + /* Need to initialize this to zero. + */ + OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (0)); + OUT_BATCH(0); + + OUT_BATCH(_3DSTATE_DEPTH_SUBRECT_DISABLE); + + /* disable indirect state for now + */ + OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0); + OUT_BATCH(0); + } + + /* 7 dwords, 1 relocs */ + if (i915->hardware_dirty & I915_HW_IMMEDIATE) + { + OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | + I1_LOAD_S(0) | + I1_LOAD_S(1) | + I1_LOAD_S(2) | + I1_LOAD_S(4) | + I1_LOAD_S(5) | + I1_LOAD_S(6) | + (5)); + + if(i915->vbo) + OUT_RELOC(i915->vbo, + I915_BUFFER_ACCESS_READ, + i915->current.immediate[I915_IMMEDIATE_S0]); + else + /* FIXME: we should not do this */ + OUT_BATCH(0); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S1]); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S2]); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S4]); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S5]); + OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S6]); + } + + /* I915_MAX_DYNAMIC dwords, 0 relocs */ + if (i915->hardware_dirty & I915_HW_DYNAMIC) + { + int i; + for (i = 0; i < I915_MAX_DYNAMIC; i++) { + OUT_BATCH(i915->current.dynamic[i]); + } + } + + /* 8 dwords, 2 relocs */ + if (i915->hardware_dirty & I915_HW_STATIC) + { + struct pipe_surface *cbuf_surface = i915->framebuffer.cbufs[0]; + struct pipe_surface *depth_surface = i915->framebuffer.zsbuf; + + if (cbuf_surface) { + unsigned pitch = (cbuf_surface->pitch * cbuf_surface->cpp); + + OUT_BATCH(_3DSTATE_BUF_INFO_CMD); + + OUT_BATCH(BUF_3D_ID_COLOR_BACK | + BUF_3D_PITCH(pitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); + + OUT_RELOC(cbuf_surface->buffer, + I915_BUFFER_ACCESS_WRITE, + 0); + } + + /* What happens if no zbuf?? + */ + if (depth_surface) { + unsigned zpitch = (depth_surface->pitch * depth_surface->cpp); + + OUT_BATCH(_3DSTATE_BUF_INFO_CMD); + + OUT_BATCH(BUF_3D_ID_DEPTH | + BUF_3D_PITCH(zpitch) | /* pitch in bytes */ + BUF_3D_USE_FENCE); + + OUT_RELOC(depth_surface->buffer, + I915_BUFFER_ACCESS_WRITE, + 0); + } + + { + unsigned cformat, zformat = 0; + + if (cbuf_surface) + cformat = cbuf_surface->format; + else + cformat = PIPE_FORMAT_A8R8G8B8_UNORM; /* arbitrary */ + cformat = translate_format(cformat); + + if (depth_surface) + zformat = translate_depth_format( i915->framebuffer.zsbuf->format ); + + OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); + OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */ + DSTORG_VERT_BIAS(0x8) | /* .5 */ + LOD_PRECLAMP_OGL | + TEX_DEFAULT_COLOR_OGL | + cformat | + zformat ); + } + } + +#if 01 + /* texture images */ + /* 2 + I915_TEX_UNITS*3 dwords, I915_TEX_UNITS relocs */ + if (i915->hardware_dirty & (I915_HW_MAP | I915_HW_SAMPLER)) + { + /* XXX: we were refering to sampler state + * (current.sampler_enable_nr) below, but only checking + * I915_HW_MAP above. Should probably calculate the enabled + * flags separately - but there will be further rework of + * state so perhaps not necessary yet. + */ + const uint nr = i915->current.sampler_enable_nr; + if (nr) { + const uint enabled = i915->current.sampler_enable_flags; + uint unit; + uint count = 0; + OUT_BATCH(_3DSTATE_MAP_STATE | (3 * nr)); + OUT_BATCH(enabled); + for (unit = 0; unit < I915_TEX_UNITS; unit++) { + if (enabled & (1 << unit)) { + struct pipe_buffer *buf = + i915->texture[unit]->buffer; + uint offset = 0; + assert(buf); + + count++; + + OUT_RELOC(buf, + I915_BUFFER_ACCESS_READ, + offset); + OUT_BATCH(i915->current.texbuffer[unit][0]); /* MS3 */ + OUT_BATCH(i915->current.texbuffer[unit][1]); /* MS4 */ + } + } + assert(count == nr); + } + } +#endif + +#if 01 + /* samplers */ + /* 2 + I915_TEX_UNITS*3 dwords, 0 relocs */ + if (i915->hardware_dirty & I915_HW_SAMPLER) + { + if (i915->current.sampler_enable_nr) { + int i; + + OUT_BATCH( _3DSTATE_SAMPLER_STATE | + (3 * i915->current.sampler_enable_nr) ); + + OUT_BATCH( i915->current.sampler_enable_flags ); + + for (i = 0; i < I915_TEX_UNITS; i++) { + if (i915->current.sampler_enable_flags & (1<current.sampler[i][0] ); + OUT_BATCH( i915->current.sampler[i][1] ); + OUT_BATCH( i915->current.sampler[i][2] ); + } + } + } + } +#endif + + /* constants */ + /* 2 + I915_MAX_CONSTANT*4 dwords, 0 relocs */ + if (i915->hardware_dirty & I915_HW_PROGRAM) + { + const uint nr = i915->current.num_constants[PIPE_SHADER_FRAGMENT]; + assert(nr <= I915_MAX_CONSTANT); + if (nr > 0) { + const uint *c + = (const uint *) i915->current.constants[PIPE_SHADER_FRAGMENT]; + uint i; + OUT_BATCH( _3DSTATE_PIXEL_SHADER_CONSTANTS | (nr * 4) ); + OUT_BATCH( (1 << (nr - 1)) | ((1 << (nr - 1)) - 1) ); + for (i = 0; i < nr; i++) { + OUT_BATCH(*c++); + OUT_BATCH(*c++); + OUT_BATCH(*c++); + OUT_BATCH(*c++); + } + } + } + + /* Fragment program */ + /* i915->current.program_len dwords, 0 relocs */ + if (i915->hardware_dirty & I915_HW_PROGRAM) + { + uint i; + /* we should always have, at least, a pass-through program */ + assert(i915->current.program_len > 0); + for (i = 0; i < i915->current.program_len; i++) { + OUT_BATCH(i915->current.program[i]); + } + } + + /* drawing surface size */ + /* 6 dwords, 0 relocs */ + { + uint w, h; + boolean k = framebuffer_size(&i915->framebuffer, &w, &h); + assert(k); + + OUT_BATCH(_3DSTATE_DRAW_RECT_CMD); + OUT_BATCH(0); + OUT_BATCH(0); + OUT_BATCH(((w - 1) & 0xffff) | ((h - 1) << 16)); + OUT_BATCH(0); + OUT_BATCH(0); + } + + + i915->hardware_dirty = 0; +} diff --git a/src/gallium/drivers/i915simple/i915_state_immediate.c b/src/gallium/drivers/i915simple/i915_state_immediate.c new file mode 100644 index 0000000000..07031fc6c5 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_state_immediate.c @@ -0,0 +1,221 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "i915_state_inlines.h" +#include "i915_context.h" +#include "i915_state.h" +#include "i915_reg.h" +#include "p_util.h" + + +/* All state expressable with the LOAD_STATE_IMMEDIATE_1 packet. + * Would like to opportunistically recombine all these fragments into + * a single packet containing only what has changed, but for now emit + * as multiple packets. + */ + + + + +/*********************************************************************** + * S0,S1: Vertex buffer state. + */ +static void upload_S0S1(struct i915_context *i915) +{ + unsigned LIS0, LIS1; + + /* INTEL_NEW_VBO */ + /* TODO: re-use vertex buffers here? */ + LIS0 = 0; + + /* INTEL_NEW_VERTEX_SIZE -- do this where the vertex size is calculated! + */ + { + unsigned vertex_size = i915->current.vertex_info.size; + + LIS1 = ((vertex_size << 24) | + (vertex_size << 16)); + } + + /* INTEL_NEW_VBO */ + /* TODO: use a vertex generation number to track vbo changes */ + if (1 || + i915->current.immediate[I915_IMMEDIATE_S0] != LIS0 || + i915->current.immediate[I915_IMMEDIATE_S1] != LIS1) + { + i915->current.immediate[I915_IMMEDIATE_S0] = LIS0; + i915->current.immediate[I915_IMMEDIATE_S1] = LIS1; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + +const struct i915_tracked_state i915_upload_S0S1 = { + I915_NEW_VBO | I915_NEW_VERTEX_FORMAT, + upload_S0S1 +}; + + + + +/*********************************************************************** + * S4: Vertex format, rasterization state + */ +static void upload_S2S4(struct i915_context *i915) +{ + unsigned LIS2, LIS4; + + /* I915_NEW_VERTEX_FORMAT */ + { + LIS2 = i915->current.vertex_info.hwfmt[1]; + LIS4 = i915->current.vertex_info.hwfmt[0]; + /* + debug_printf("LIS2: 0x%x LIS4: 0x%x\n", LIS2, LIS4); + */ + assert(LIS4); /* should never be zero? */ + } + + LIS4 |= i915->rasterizer->LIS4; + + if (LIS2 != i915->current.immediate[I915_IMMEDIATE_S2] || + LIS4 != i915->current.immediate[I915_IMMEDIATE_S4]) { + + i915->current.immediate[I915_IMMEDIATE_S2] = LIS2; + i915->current.immediate[I915_IMMEDIATE_S4] = LIS4; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + + +const struct i915_tracked_state i915_upload_S2S4 = { + I915_NEW_RASTERIZER | I915_NEW_VERTEX_FORMAT, + upload_S2S4 +}; + + + +/*********************************************************************** + * + */ +static void upload_S5( struct i915_context *i915 ) +{ + unsigned LIS5 = 0; + + LIS5 |= i915->depth_stencil->stencil_LIS5; + + LIS5 |= i915->blend->LIS5; + +#if 0 + /* I915_NEW_RASTERIZER */ + if (i915->state.Polygon->OffsetFill) { + LIS5 |= S5_GLOBAL_DEPTH_OFFSET_ENABLE; + } +#endif + + + if (LIS5 != i915->current.immediate[I915_IMMEDIATE_S5]) { + i915->current.immediate[I915_IMMEDIATE_S5] = LIS5; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + +const struct i915_tracked_state i915_upload_S5 = { + (I915_NEW_DEPTH_STENCIL | I915_NEW_BLEND | I915_NEW_RASTERIZER), + upload_S5 +}; + + +/*********************************************************************** + */ +static void upload_S6( struct i915_context *i915 ) +{ + unsigned LIS6 = (S6_COLOR_WRITE_ENABLE | + (2 << S6_TRISTRIP_PV_SHIFT)); + + /* I915_NEW_BLEND + */ + LIS6 |= i915->blend->LIS6; + + /* I915_NEW_DEPTH + */ + LIS6 |= i915->depth_stencil->depth_LIS6; + + if (LIS6 != i915->current.immediate[I915_IMMEDIATE_S6]) { + i915->current.immediate[I915_IMMEDIATE_S6] = LIS6; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + +const struct i915_tracked_state i915_upload_S6 = { + I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, + upload_S6 +}; + + +/*********************************************************************** + */ +static void upload_S7( struct i915_context *i915 ) +{ + unsigned LIS7; + + /* I915_NEW_RASTERIZER + */ + LIS7 = i915->rasterizer->LIS7; + + if (LIS7 != i915->current.immediate[I915_IMMEDIATE_S7]) { + i915->current.immediate[I915_IMMEDIATE_S7] = LIS7; + i915->hardware_dirty |= I915_HW_IMMEDIATE; + } +} + +const struct i915_tracked_state i915_upload_S7 = { + I915_NEW_RASTERIZER, + upload_S7 +}; + + +static const struct i915_tracked_state *atoms[] = { + &i915_upload_S0S1, + &i915_upload_S2S4, + &i915_upload_S5, + &i915_upload_S6, + &i915_upload_S7 +}; + +/* + */ +void i915_update_immediate( struct i915_context *i915 ) +{ + int i; + + for (i = 0; i < Elements(atoms); i++) + if (i915->dirty & atoms[i]->dirty) + atoms[i]->update( i915 ); +} diff --git a/src/gallium/drivers/i915simple/i915_state_inlines.h b/src/gallium/drivers/i915simple/i915_state_inlines.h new file mode 100644 index 0000000000..0934ac79a4 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_state_inlines.h @@ -0,0 +1,230 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef I915_STATE_INLINES_H +#define I915_STATE_INLINES_H + +#include "p_compiler.h" +#include "p_defines.h" +#include "i915_reg.h" + + +static INLINE unsigned +i915_translate_compare_func(unsigned func) +{ + switch (func) { + case PIPE_FUNC_NEVER: + return COMPAREFUNC_NEVER; + case PIPE_FUNC_LESS: + return COMPAREFUNC_LESS; + case PIPE_FUNC_LEQUAL: + return COMPAREFUNC_LEQUAL; + case PIPE_FUNC_GREATER: + return COMPAREFUNC_GREATER; + case PIPE_FUNC_GEQUAL: + return COMPAREFUNC_GEQUAL; + case PIPE_FUNC_NOTEQUAL: + return COMPAREFUNC_NOTEQUAL; + case PIPE_FUNC_EQUAL: + return COMPAREFUNC_EQUAL; + case PIPE_FUNC_ALWAYS: + return COMPAREFUNC_ALWAYS; + default: + return COMPAREFUNC_ALWAYS; + } +} + +static INLINE unsigned +i915_translate_stencil_op(unsigned op) +{ + switch (op) { + case PIPE_STENCIL_OP_KEEP: + return STENCILOP_KEEP; + case PIPE_STENCIL_OP_ZERO: + return STENCILOP_ZERO; + case PIPE_STENCIL_OP_REPLACE: + return STENCILOP_REPLACE; + case PIPE_STENCIL_OP_INCR: + return STENCILOP_INCRSAT; + case PIPE_STENCIL_OP_DECR: + return STENCILOP_DECRSAT; + case PIPE_STENCIL_OP_INCR_WRAP: + return STENCILOP_INCR; + case PIPE_STENCIL_OP_DECR_WRAP: + return STENCILOP_DECR; + case PIPE_STENCIL_OP_INVERT: + return STENCILOP_INVERT; + default: + return STENCILOP_ZERO; + } +} + +static INLINE unsigned +i915_translate_blend_factor(unsigned factor) +{ + switch (factor) { + case PIPE_BLENDFACTOR_ZERO: + return BLENDFACT_ZERO; + case PIPE_BLENDFACTOR_SRC_ALPHA: + return BLENDFACT_SRC_ALPHA; + case PIPE_BLENDFACTOR_ONE: + return BLENDFACT_ONE; + case PIPE_BLENDFACTOR_SRC_COLOR: + return BLENDFACT_SRC_COLR; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + return BLENDFACT_INV_SRC_COLR; + case PIPE_BLENDFACTOR_DST_COLOR: + return BLENDFACT_DST_COLR; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + return BLENDFACT_INV_DST_COLR; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + return BLENDFACT_INV_SRC_ALPHA; + case PIPE_BLENDFACTOR_DST_ALPHA: + return BLENDFACT_DST_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + return BLENDFACT_INV_DST_ALPHA; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + return BLENDFACT_SRC_ALPHA_SATURATE; + case PIPE_BLENDFACTOR_CONST_COLOR: + return BLENDFACT_CONST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + return BLENDFACT_INV_CONST_COLOR; + case PIPE_BLENDFACTOR_CONST_ALPHA: + return BLENDFACT_CONST_ALPHA; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + return BLENDFACT_INV_CONST_ALPHA; + default: + return BLENDFACT_ZERO; + } +} + +static INLINE unsigned +i915_translate_blend_func(unsigned mode) +{ + switch (mode) { + case PIPE_BLEND_ADD: + return BLENDFUNC_ADD; + case PIPE_BLEND_MIN: + return BLENDFUNC_MIN; + case PIPE_BLEND_MAX: + return BLENDFUNC_MAX; + case PIPE_BLEND_SUBTRACT: + return BLENDFUNC_SUBTRACT; + case PIPE_BLEND_REVERSE_SUBTRACT: + return BLENDFUNC_REVERSE_SUBTRACT; + default: + return 0; + } +} + + +static INLINE unsigned +i915_translate_logic_op(unsigned opcode) +{ + switch (opcode) { + case PIPE_LOGICOP_CLEAR: + return LOGICOP_CLEAR; + case PIPE_LOGICOP_AND: + return LOGICOP_AND; + case PIPE_LOGICOP_AND_REVERSE: + return LOGICOP_AND_RVRSE; + case PIPE_LOGICOP_COPY: + return LOGICOP_COPY; + case PIPE_LOGICOP_COPY_INVERTED: + return LOGICOP_COPY_INV; + case PIPE_LOGICOP_AND_INVERTED: + return LOGICOP_AND_INV; + case PIPE_LOGICOP_NOOP: + return LOGICOP_NOOP; + case PIPE_LOGICOP_XOR: + return LOGICOP_XOR; + case PIPE_LOGICOP_OR: + return LOGICOP_OR; + case PIPE_LOGICOP_OR_INVERTED: + return LOGICOP_OR_INV; + case PIPE_LOGICOP_NOR: + return LOGICOP_NOR; + case PIPE_LOGICOP_EQUIV: + return LOGICOP_EQUIV; + case PIPE_LOGICOP_INVERT: + return LOGICOP_INV; + case PIPE_LOGICOP_OR_REVERSE: + return LOGICOP_OR_RVRSE; + case PIPE_LOGICOP_NAND: + return LOGICOP_NAND; + case PIPE_LOGICOP_SET: + return LOGICOP_SET; + default: + return LOGICOP_SET; + } +} + + + +static INLINE boolean i915_validate_vertices( unsigned hw_prim, unsigned nr ) +{ + boolean ok; + + switch (hw_prim) { + case PRIM3D_POINTLIST: + ok = (nr >= 1); + assert(ok); + break; + case PRIM3D_LINELIST: + ok = (nr >= 2) && (nr % 2) == 0; + assert(ok); + break; + case PRIM3D_LINESTRIP: + ok = (nr >= 2); + assert(ok); + break; + case PRIM3D_TRILIST: + ok = (nr >= 3) && (nr % 3) == 0; + assert(ok); + break; + case PRIM3D_TRISTRIP: + ok = (nr >= 3); + assert(ok); + break; + case PRIM3D_TRIFAN: + ok = (nr >= 3); + assert(ok); + break; + case PRIM3D_POLY: + ok = (nr >= 3); + assert(ok); + break; + default: + assert(0); + ok = 0; + break; + } + + return ok; +} + +#endif diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c new file mode 100644 index 0000000000..9c1a5bbbd6 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -0,0 +1,231 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_context.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" + +#include "i915_state_inlines.h" +#include "i915_context.h" +#include "i915_reg.h" +#include "i915_state.h" + + +/** + * Compute i915 texture sampling state. + * + * Recalculate all state from scratch. Perhaps not the most + * efficient, but this has gotten complex enough that we need + * something which is understandable and reliable. + * \param state returns the 3 words of compute state + */ +static void update_sampler(struct i915_context *i915, + uint unit, + const struct i915_sampler_state *sampler, + const struct i915_texture *tex, + unsigned state[3] ) +{ + const struct pipe_texture *pt = &tex->base; + + /* Need to do this after updating the maps, which call the + * intel_finalize_mipmap_tree and hence can update firstLevel: + */ + state[0] = sampler->state[0]; + state[1] = sampler->state[1]; + state[2] = sampler->state[2]; + + if (pt->format == PIPE_FORMAT_YCBCR || + pt->format == PIPE_FORMAT_YCBCR_REV) + state[0] |= SS2_COLORSPACE_CONVERSION; + + /* 3D textures don't seem to respect the border color. + * Fallback if there's ever a danger that they might refer to + * it. + * + * Effectively this means fallback on 3D clamp or + * clamp_to_border. + * + * XXX: Check if this is true on i945. + * XXX: Check if this bug got fixed in release silicon. + */ +#if 0 + { + const unsigned ws = sampler->templ->wrap_s; + const unsigned wt = sampler->templ->wrap_t; + const unsigned wr = sampler->templ->wrap_r; + if (pt->target == PIPE_TEXTURE_3D && + (sampler->templ->min_img_filter != PIPE_TEX_FILTER_NEAREST || + sampler->templ->mag_img_filter != PIPE_TEX_FILTER_NEAREST) && + (ws == PIPE_TEX_WRAP_CLAMP || + wt == PIPE_TEX_WRAP_CLAMP || + wr == PIPE_TEX_WRAP_CLAMP || + ws == PIPE_TEX_WRAP_CLAMP_TO_BORDER || + wt == PIPE_TEX_WRAP_CLAMP_TO_BORDER || + wr == PIPE_TEX_WRAP_CLAMP_TO_BORDER)) { + if (i915->strict_conformance) { + assert(0); + /* sampler->fallback = true; */ + /* TODO */ + } + } + } +#endif + + state[1] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT); +} + + +void i915_update_samplers( struct i915_context *i915 ) +{ + uint unit; + + i915->current.sampler_enable_nr = 0; + i915->current.sampler_enable_flags = 0x0; + + for (unit = 0; unit < I915_TEX_UNITS; unit++) { + /* determine unit enable/disable by looking for a bound texture */ + /* could also examine the fragment program? */ + if (i915->texture[unit]) { + update_sampler( i915, + unit, + i915->sampler[unit], /* sampler state */ + i915->texture[unit], /* texture */ + i915->current.sampler[unit] /* the result */ + ); + + i915->current.sampler_enable_nr++; + i915->current.sampler_enable_flags |= (1 << unit); + } + } + + i915->hardware_dirty |= I915_HW_SAMPLER; +} + + +static uint +translate_texture_format(enum pipe_format pipeFormat) +{ + switch (pipeFormat) { + case PIPE_FORMAT_U_L8: + return MAPSURF_8BIT | MT_8BIT_L8; + case PIPE_FORMAT_U_I8: + return MAPSURF_8BIT | MT_8BIT_I8; + case PIPE_FORMAT_U_A8: + return MAPSURF_8BIT | MT_8BIT_A8; + case PIPE_FORMAT_U_A8_L8: + return MAPSURF_16BIT | MT_16BIT_AY88; + case PIPE_FORMAT_R5G6B5_UNORM: + return MAPSURF_16BIT | MT_16BIT_RGB565; + case PIPE_FORMAT_A1R5G5B5_UNORM: + return MAPSURF_16BIT | MT_16BIT_ARGB1555; + case PIPE_FORMAT_A4R4G4B4_UNORM: + return MAPSURF_16BIT | MT_16BIT_ARGB4444; + case PIPE_FORMAT_A8R8G8B8_UNORM: + return MAPSURF_32BIT | MT_32BIT_ARGB8888; + case PIPE_FORMAT_YCBCR_REV: + return (MAPSURF_422 | MT_422_YCRCB_NORMAL); + case PIPE_FORMAT_YCBCR: + return (MAPSURF_422 | MT_422_YCRCB_SWAPY); +#if 0 + case PIPE_FORMAT_RGB_FXT1: + case PIPE_FORMAT_RGBA_FXT1: + return (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1); +#endif + case PIPE_FORMAT_Z16_UNORM: + return (MAPSURF_16BIT | MT_16BIT_L16); +#if 0 + case PIPE_FORMAT_RGBA_DXT1: + case PIPE_FORMAT_RGB_DXT1: + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1); + case PIPE_FORMAT_RGBA_DXT3: + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3); + case PIPE_FORMAT_RGBA_DXT5: + return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); +#endif + case PIPE_FORMAT_S8Z24_UNORM: + return (MAPSURF_32BIT | MT_32BIT_xL824); + default: + debug_printf("i915: translate_texture_format() bad image format %x\n", + pipeFormat); + assert(0); + return 0; + } +} + + +static void +i915_update_texture(struct i915_context *i915, uint unit, + uint state[6]) +{ + const struct i915_texture *tex = i915->texture[unit]; + const struct pipe_texture *pt = &tex->base; + uint format, pitch; + const uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; + const uint num_levels = pt->last_level; + + assert(tex); + assert(width); + assert(height); + assert(depth); + + format = translate_texture_format(pt->format); + pitch = tex->pitch * pt->cpp; + + assert(format); + assert(pitch); + + /* MS3 state */ + state[0] = + (((height - 1) << MS3_HEIGHT_SHIFT) + | ((width - 1) << MS3_WIDTH_SHIFT) + | format + | MS3_USE_FENCE_REGS); + + /* MS4 state */ + state[1] = + ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) + | MS4_CUBE_FACE_ENA_MASK + | ((num_levels * 4) << MS4_MAX_LOD_SHIFT) + | ((depth - 1) << MS4_VOLUME_DEPTH_SHIFT)); +} + + +void +i915_update_textures(struct i915_context *i915) +{ + uint unit; + + for (unit = 0; unit < I915_TEX_UNITS; unit++) { + /* determine unit enable/disable by looking for a bound texture */ + /* could also examine the fragment program? */ + if (i915->texture[unit]) { + i915_update_texture(i915, unit, i915->current.texbuffer[unit]); + } + } + + i915->hardware_dirty |= I915_HW_MAP; +} diff --git a/src/gallium/drivers/i915simple/i915_strings.c b/src/gallium/drivers/i915simple/i915_strings.c new file mode 100644 index 0000000000..c713bf7208 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_strings.c @@ -0,0 +1,83 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_context.h" +#include "i915_reg.h" + + +static const char *i915_get_vendor( struct pipe_context *pipe ) +{ + return "Tungsten Graphics, Inc."; +} + + +static const char *i915_get_name( struct pipe_context *pipe ) +{ + static char buffer[128]; + const char *chipset; + + switch (i915_context(pipe)->pci_id) { + case PCI_CHIP_I915_G: + chipset = "915G"; + break; + case PCI_CHIP_I915_GM: + chipset = "915GM"; + break; + case PCI_CHIP_I945_G: + chipset = "945G"; + break; + case PCI_CHIP_I945_GM: + chipset = "945GM"; + break; + case PCI_CHIP_I945_GME: + chipset = "945GME"; + break; + case PCI_CHIP_G33_G: + chipset = "G33"; + break; + case PCI_CHIP_Q35_G: + chipset = "Q35"; + break; + case PCI_CHIP_Q33_G: + chipset = "Q33"; + break; + default: + chipset = "unknown"; + break; + } + + sprintf(buffer, "pipe/i915 (chipset: %s)", chipset); + return buffer; +} + + +void +i915_init_string_functions(struct i915_context *i915) +{ + i915->pipe.get_name = i915_get_name; + i915->pipe.get_vendor = i915_get_vendor; +} diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c new file mode 100644 index 0000000000..de0cc5fe06 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -0,0 +1,191 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915_context.h" +#include "i915_blit.h" +#include "i915_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "pipe/util/p_tile.h" + + +/* + * XXX note: same as code in sp_surface.c + */ +static struct pipe_surface * +i915_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct i915_texture *tex = (struct i915_texture *)pt; + struct pipe_surface *ps; + unsigned offset; /* in bytes */ + + offset = tex->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE) { + offset += tex->image_offset[level][face] * pt->cpp; + } + else if (pt->target == PIPE_TEXTURE_3D) { + offset += tex->image_offset[level][zslice] * pt->cpp; + } + else { + assert(face == 0); + assert(zslice == 0); + } + + ps = pipe->winsys->surface_alloc(pipe->winsys); + if (ps) { + assert(ps->refcount); + assert(ps->winsys); + pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = tex->pitch; + ps->offset = offset; + } + return ps; +} + + + +/* Assumes all values are within bounds -- no checking at this level - + * do it higher up if required. + */ +static void +i915_surface_copy(struct pipe_context *pipe, + unsigned do_flip, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + assert( dst != src ); + assert( dst->cpp == src->cpp ); + + if (0) { + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe_surface_map(src), + do_flip ? -(int) src->pitch : src->pitch, + srcx, do_flip ? 1 - srcy - height : srcy); + + pipe_surface_unmap(src); + pipe_surface_unmap(dst); + } + else { + i915_copy_blit( i915_context(pipe), + do_flip, + dst->cpp, + (short) src->pitch, src->buffer, src->offset, + (short) dst->pitch, dst->buffer, dst->offset, + (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); + } +} + +/* Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +static void * +get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) +{ + return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; +} + + +static void +i915_surface_fill(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) +{ + if (0) { + unsigned i, j; + void *dst_map = pipe_surface_map(dst); + + switch (dst->cpp) { + case 1: { + ubyte *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: { + ushort *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (ushort) value; + row += dst->pitch; + } + } + break; + case 4: { + unsigned *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + default: + assert(0); + break; + } + + pipe_surface_unmap( dst ); + } + else { + i915_fill_blit( i915_context(pipe), + dst->cpp, + (short) dst->pitch, + dst->buffer, dst->offset, + (short) dstx, (short) dsty, + (short) width, (short) height, + value ); + } +} + + +void +i915_init_surface_functions(struct i915_context *i915) +{ + i915->pipe.get_tex_surface = i915_get_tex_surface; + + i915->pipe.surface_copy = i915_surface_copy; + i915->pipe.surface_fill = i915_surface_fill; +} diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c new file mode 100644 index 0000000000..6d37ae3d74 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -0,0 +1,536 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "pipe/p_state.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "i915_context.h" +#include "i915_texture.h" +#include "i915_debug.h" + + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + + + +static void +i915_miptree_set_level_info(struct i915_texture *tex, + unsigned level, + unsigned nr_images, + unsigned x, unsigned y, unsigned w, unsigned h, unsigned d) +{ + struct pipe_texture *pt = &tex->base; + + assert(level < PIPE_MAX_TEXTURE_LEVELS); + + pt->width[level] = w; + pt->height[level] = h; + pt->depth[level] = d; + + tex->level_offset[level] = (x + y * tex->pitch) * pt->cpp; + tex->nr_images[level] = nr_images; + + /* + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, tex->level_offset[level]); + */ + + /* Not sure when this would happen, but anyway: + */ + if (tex->image_offset[level]) { + FREE(tex->image_offset[level]); + tex->image_offset[level] = NULL; + } + + assert(nr_images); + assert(!tex->image_offset[level]); + + tex->image_offset[level] = (unsigned *) MALLOC(nr_images * sizeof(unsigned)); + tex->image_offset[level][0] = 0; +} + + +static void +i915_miptree_set_image_offset(struct i915_texture *tex, + unsigned level, unsigned img, unsigned x, unsigned y) +{ + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + + assert(img < tex->nr_images[level]); + + tex->image_offset[level][img] = (x + y * tex->pitch); + + /* + DBG("%s level %d img %d pos %d,%d image_offset %x\n", + __FUNCTION__, level, img, x, y, tex->image_offset[level][img]); + */ +} + + +static void +i945_miptree_layout_2d( struct i915_texture *tex ) +{ + struct pipe_texture *pt = &tex->base; + int align_h = 2, align_w = 4; + unsigned level; + unsigned x = 0; + unsigned y = 0; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + + tex->pitch = pt->width[0]; + + /* May need to adjust pitch to accomodate the placement of + * the 2nd mipmap level. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap level out past the width of its parent. + */ + if (pt->last_level > 0) { + unsigned mip1_width = align_int(minify(pt->width[0]), align_w) + + minify(minify(pt->width[0])); + + if (mip1_width > pt->width[0]) + tex->pitch = mip1_width; + } + + /* Pitch must be a whole number of dwords, even though we + * express it in texels. + */ + tex->pitch = align_int(tex->pitch * pt->cpp, 4) / pt->cpp; + tex->total_height = 0; + + for (level = 0; level <= pt->last_level; level++) { + unsigned img_height; + + i915_miptree_set_level_info(tex, level, 1, x, y, width, height, 1); + + if (pt->compressed) + img_height = MAX2(1, height/4); + else + img_height = align_int(height, align_h); + + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + tex->total_height = MAX2(tex->total_height, y + img_height); + + /* Layout_below: step right after second mipmap level. + */ + if (level == 1) { + x += align_int(width, align_w); + } + else { + y += img_height; + } + + width = minify(width); + height = minify(height); + } +} + + +static const int initial_offsets[6][2] = { + {0, 0}, + {0, 2}, + {1, 0}, + {1, 2}, + {1, 1}, + {1, 3} +}; + +static const int step_offsets[6][2] = { + {0, 2}, + {0, 2}, + {-1, 2}, + {-1, 2}, + {-1, 1}, + {-1, 1} +}; + + +static boolean +i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned level; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: { + const unsigned dim = pt->width[0]; + unsigned face; + unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* double pitch for cube layouts */ + tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; + tex->total_height = dim * 4; + + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_level_info(tex, level, 6, + 0, 0, + /*OLD: tex->pitch, tex->total_height,*/ + lvlWidth, lvlHeight, + 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + for (face = 0; face < 6; face++) { + unsigned x = initial_offsets[face][0] * dim; + unsigned y = initial_offsets[face][1] * dim; + unsigned d = dim; + + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_image_offset(tex, level, face, x, y); + d >>= 1; + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + } + } + break; + } + case PIPE_TEXTURE_3D:{ + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + unsigned stack_height = 0; + + /* Calculate the size of a single slice. + */ + tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; + + /* XXX: hardware expects/requires 9 levels at minimum. + */ + for (level = 0; level <= MAX2(8, pt->last_level); + level++) { + i915_miptree_set_level_info(tex, level, depth, 0, tex->total_height, + width, height, depth); + + + stack_height += MAX2(2, height); + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + + /* Fixup depth image_offsets: + */ + depth = pt->depth[0]; + for (level = 0; level <= pt->last_level; level++) { + unsigned i; + for (i = 0; i < depth; i++) + i915_miptree_set_image_offset(tex, level, i, + 0, i * stack_height); + + depth = minify(depth); + } + + + /* Multiply slice size by texture depth for total size. It's + * remarkable how wasteful of memory the i915 texture layouts + * are. They are largely fixed in the i945. + */ + tex->total_height = stack_height * pt->depth[0]; + break; + } + + default:{ + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned img_height; + + tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; + tex->total_height = 0; + + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_level_info(tex, level, 1, + 0, tex->total_height, + width, height, 1); + + if (pt->compressed) + img_height = MAX2(1, height / 4); + else + img_height = (MAX2(2, height) + 1) & ~1; + + tex->total_height += img_height; + + width = minify(width); + height = minify(height); + } + break; + } + } + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + tex->pitch, + tex->total_height, pt->cpp, tex->pitch * tex->total_height * pt->cpp); + */ + + return TRUE; +} + + +static boolean +i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned level; + + switch (pt->target) { + case PIPE_TEXTURE_CUBE:{ + const unsigned dim = pt->width[0]; + unsigned face; + unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; + + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* Depending on the size of the largest images, pitch can be + * determined either by the old-style packing of cubemap faces, + * or the final row of 4x4, 2x2 and 1x1 faces below this. + */ + if (dim > 32) + tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; + else + tex->pitch = 14 * 8; + + tex->total_height = dim * 4 + 4; + + /* Set all the levels to effectively occupy the whole rectangular region. + */ + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_level_info(tex, level, 6, + 0, 0, + lvlWidth, lvlHeight, 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + + + for (face = 0; face < 6; face++) { + unsigned x = initial_offsets[face][0] * dim; + unsigned y = initial_offsets[face][1] * dim; + unsigned d = dim; + + if (dim == 4 && face >= 4) { + y = tex->total_height - 4; + x = (face - 4) * 8; + } + else if (dim < 4 && (face > 0)) { + y = tex->total_height - 4; + x = face * 8; + } + + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_image_offset(tex, level, face, x, y); + + d >>= 1; + + switch (d) { + case 4: + switch (face) { + case PIPE_TEX_FACE_POS_X: + case PIPE_TEX_FACE_NEG_X: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + case PIPE_TEX_FACE_POS_Y: + case PIPE_TEX_FACE_NEG_Y: + y += 12; + x -= 8; + break; + case PIPE_TEX_FACE_POS_Z: + case PIPE_TEX_FACE_NEG_Z: + y = tex->total_height - 4; + x = (face - 4) * 8; + break; + } + + case 2: + y = tex->total_height - 4; + x = 16 + face * 8; + break; + + case 1: + x += 48; + break; + + default: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + } + } + } + break; + } + case PIPE_TEXTURE_3D:{ + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + unsigned pack_x_pitch, pack_x_nr; + unsigned pack_y_pitch; + unsigned level; + + tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; + tex->total_height = 0; + + pack_y_pitch = MAX2(pt->height[0], 2); + pack_x_pitch = tex->pitch; + pack_x_nr = 1; + + for (level = 0; level <= pt->last_level; level++) { + unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; + int x = 0; + int y = 0; + unsigned q, j; + + i915_miptree_set_level_info(tex, level, nr_images, + 0, tex->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + i915_miptree_set_image_offset(tex, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + tex->total_height += y; + + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= tex->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + } + + width = minify(width); + height = minify(height); + depth = minify(depth); + } + break; + } + + case PIPE_TEXTURE_1D: + case PIPE_TEXTURE_2D: +// case PIPE_TEXTURE_RECTANGLE: + i945_miptree_layout_2d(tex); + break; + default: + assert(0); + return FALSE; + } + + /* + DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + tex->pitch, + tex->total_height, pt->cpp, tex->pitch * tex->total_height * pt->cpp); + */ + + return TRUE; +} + + +struct pipe_texture * +i915_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat) +{ + struct i915_texture *tex = CALLOC_STRUCT(i915_texture); + + if (tex) { + struct i915_context *i915 = i915_context(pipe); + + tex->base = *templat; + + if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : + i915_miptree_layout(pipe, tex)) + tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, + PIPE_BUFFER_USAGE_PIXEL, + tex->pitch * tex->base.cpp * + tex->total_height); + + if (!tex->buffer) { + FREE(tex); + return NULL; + } + } + + return &tex->base; +} + + +void +i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct i915_texture *tex = (struct i915_texture *)*pt; + uint i; + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); + */ + + pipe_buffer_reference(pipe->winsys, &tex->buffer, NULL); + + for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) + if (tex->image_offset[i]) + FREE(tex->image_offset[i]); + + FREE(tex); + } + *pt = NULL; +} diff --git a/src/gallium/drivers/i915simple/i915_texture.h b/src/gallium/drivers/i915simple/i915_texture.h new file mode 100644 index 0000000000..330d111dc7 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_texture.h @@ -0,0 +1,17 @@ + +#ifndef I915_TEXTURE_H +#define I915_TEXTURE_H + +struct pipe_context; +struct pipe_texture; + + +struct pipe_texture * +i915_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat); + +extern void +i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); + + +#endif /* I915_TEXTURE_H */ diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h new file mode 100644 index 0000000000..fe49710852 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_winsys.h @@ -0,0 +1,115 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * This is the interface that i915simple requires any window system + * hosting it to implement. This is the only include file in i915simple + * which is public. + * + */ + +#ifndef I915_WINSYS_H +#define I915_WINSYS_H + + +#include "pipe/p_defines.h" + + +/* Pipe drivers are (meant to be!) independent of both GL and the + * window system. The window system provides a buffer manager and a + * set of additional hooks for things like command buffer submission, + * etc. + * + * There clearly has to be some agreement between the window system + * driver and the hardware driver about the format of command buffers, + * etc. + */ + +struct pipe_buffer; +struct pipe_winsys; + + +/** + * Additional winsys interface for i915simple. + * + * It is an over-simple batchbuffer mechanism. Will want to improve the + * performance of this, perhaps based on the cmdstream stuff. It + * would be pretty impossible to implement swz on top of this + * interface. + * + * Will also need additions/changes to implement static/dynamic + * indirect state. + */ +struct i915_winsys { + + /** + * Reserve space on batch buffer. + * + * Returns a null pointer if there is insufficient space in the batch buffer + * to hold the requested number of dwords and relocations. + * + * The number of dwords should also include the number of relocations. + */ + unsigned *(*batch_start)( struct i915_winsys *sws, + unsigned dwords, + unsigned relocs ); + + void (*batch_dword)( struct i915_winsys *sws, + unsigned dword ); + + /** + * Emit a relocation to a buffer. + * + * Used not only when the buffer addresses are not pinned, but also to + * ensure refered buffers will not be destroyed until the current batch + * buffer execution is finished. + * + * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and + * I915_BUFFER_ACCESS_READ macros. + */ + void (*batch_reloc)( struct i915_winsys *sws, + struct pipe_buffer *buf, + unsigned access_flags, + unsigned delta ); + + void (*batch_flush)( struct i915_winsys *sws ); + void (*batch_finish)( struct i915_winsys *sws ); +}; + +#define I915_BUFFER_ACCESS_WRITE 0x1 +#define I915_BUFFER_ACCESS_READ 0x2 + +#define I915_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) + + +struct pipe_context *i915_create( struct pipe_winsys *, + struct i915_winsys *, + unsigned pci_id ); + + +#endif diff --git a/src/gallium/drivers/i965simple/Makefile b/src/gallium/drivers/i965simple/Makefile new file mode 100644 index 0000000000..48c00ab50b --- /dev/null +++ b/src/gallium/drivers/i965simple/Makefile @@ -0,0 +1,66 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = i965simple + +DRIVER_SOURCES = \ + brw_blit.c \ + brw_flush.c \ + brw_strings.c \ + brw_surface.c \ + brw_cc.c \ + brw_clip.c \ + brw_clip_line.c \ + brw_clip_point.c \ + brw_clip_state.c \ + brw_clip_tri.c \ + brw_clip_util.c \ + brw_context.c \ + brw_curbe.c \ + brw_draw.c \ + brw_draw_upload.c \ + brw_eu.c \ + brw_eu_debug.c \ + brw_eu_emit.c \ + brw_eu_util.c \ + brw_gs.c \ + brw_gs_emit.c \ + brw_gs_state.c \ + brw_misc_state.c \ + brw_sf.c \ + brw_sf_emit.c \ + brw_sf_state.c \ + brw_shader_info.c \ + brw_state.c \ + brw_state_batch.c \ + brw_state_cache.c \ + brw_state_pool.c \ + brw_state_upload.c \ + brw_tex_layout.c \ + brw_urb.c \ + brw_util.c \ + brw_vs.c \ + brw_vs_emit.c \ + brw_vs_state.c \ + brw_wm.c \ + brw_wm_iz.c \ + brw_wm_decl.c \ + brw_wm_glsl.c \ + brw_wm_sampler_state.c \ + brw_wm_state.c \ + brw_wm_surface_state.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(COMMON_BM_SOURCES) \ + $(MINIGLX_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +DRIVER_DEFINES = -I. + +include ../Makefile.template + +symlinks: diff --git a/src/gallium/drivers/i965simple/SConscript b/src/gallium/drivers/i965simple/SConscript new file mode 100644 index 0000000000..74621de84c --- /dev/null +++ b/src/gallium/drivers/i965simple/SConscript @@ -0,0 +1,55 @@ +Import('*') + +env = env.Clone() + +i965simple = env.ConvenienceLibrary( + target = 'i965simple', + source = [ + 'brw_blit.c', + 'brw_cc.c', + 'brw_clip.c', + 'brw_clip_line.c', + 'brw_clip_point.c', + 'brw_clip_state.c', + 'brw_clip_tri.c', + 'brw_clip_util.c', + 'brw_context.c', + 'brw_curbe.c', + 'brw_draw.c', + 'brw_draw_upload.c', + 'brw_eu.c', + 'brw_eu_debug.c', + 'brw_eu_emit.c', + 'brw_eu_util.c', + 'brw_flush.c', + 'brw_gs.c', + 'brw_gs_emit.c', + 'brw_gs_state.c', + 'brw_misc_state.c', + 'brw_sf.c', + 'brw_sf_emit.c', + 'brw_sf_state.c', + 'brw_shader_info.c', + 'brw_state.c', + 'brw_state_batch.c', + 'brw_state_cache.c', + 'brw_state_pool.c', + 'brw_state_upload.c', + 'brw_strings.c', + 'brw_surface.c', + 'brw_tex_layout.c', + 'brw_urb.c', + 'brw_util.c', + 'brw_vs.c', + 'brw_vs_emit.c', + 'brw_vs_state.c', + 'brw_wm.c', + 'brw_wm_decl.c', + 'brw_wm_glsl.c', + 'brw_wm_iz.c', + 'brw_wm_sampler_state.c', + 'brw_wm_state.c', + 'brw_wm_surface_state.c', + ]) + +Export('i965simple') diff --git a/src/gallium/drivers/i965simple/brw_batch.h b/src/gallium/drivers/i965simple/brw_batch.h new file mode 100644 index 0000000000..5f5932a488 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_batch.h @@ -0,0 +1,59 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef BRW_BATCH_H +#define BRW_BATCH_H + +#include "brw_winsys.h" + +#define BATCH_LOCALS + +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 + +#define BEGIN_BATCH( dwords, relocs ) \ + brw->winsys->batch_start(brw->winsys, dwords, relocs) + +#define OUT_BATCH( dword ) \ + brw->winsys->batch_dword(brw->winsys, dword) + +#define OUT_RELOC( buf, flags, delta ) \ + brw->winsys->batch_reloc(brw->winsys, buf, flags, delta) + +#define ADVANCE_BATCH() \ + brw->winsys->batch_end( brw->winsys ) + +/* XXX: this is bogus - need proper handling for out-of-memory in batchbuffer. + */ +#define FLUSH_BATCH(fence) do { \ + brw->winsys->batch_flush(brw->winsys, fence); \ + brw->hardware_dirty = ~0; \ +} while (0) + +#define BRW_BATCH_STRUCT(brw, s) brw_batchbuffer_data( brw->winsys, (s), sizeof(*(s))) + +#endif diff --git a/src/gallium/drivers/i965simple/brw_blit.c b/src/gallium/drivers/i965simple/brw_blit.c new file mode 100644 index 0000000000..8494f70493 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_blit.c @@ -0,0 +1,218 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include +#include + +#include "brw_batch.h" +#include "brw_blit.h" +#include "brw_context.h" +#include "brw_reg.h" + +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +#define FILE_DEBUG_FLAG DEBUG_BLIT + +void brw_fill_blit(struct brw_context *brw, + unsigned cpp, + short dst_pitch, + struct pipe_buffer *dst_buffer, + unsigned dst_offset, + boolean dst_tiled, + short x, short y, + short w, short h, + unsigned color) +{ + unsigned BR13, CMD; + BATCH_LOCALS; + + dst_pitch *= cpp; + + switch(cpp) { + case 1: + case 2: + case 3: + BR13 = (0xF0 << 16) | (1<<24); + CMD = XY_COLOR_BLT_CMD; + break; + case 4: + BR13 = (0xF0 << 16) | (1<<24) | (1<<25); + CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; + break; + default: + return; + } + + if (dst_tiled) { + CMD |= XY_DST_TILED; + dst_pitch /= 4; + } + + BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH( CMD ); + OUT_BATCH( dst_pitch | BR13 ); + OUT_BATCH( (y << 16) | x ); + OUT_BATCH( ((y+h) << 16) | (x+w) ); + OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, dst_offset ); + OUT_BATCH( color ); + ADVANCE_BATCH(); +} + +static unsigned translate_raster_op(unsigned logicop) +{ + switch(logicop) { + case PIPE_LOGICOP_CLEAR: return 0x00; + case PIPE_LOGICOP_AND: return 0x88; + case PIPE_LOGICOP_AND_REVERSE: return 0x44; + case PIPE_LOGICOP_COPY: return 0xCC; + case PIPE_LOGICOP_AND_INVERTED: return 0x22; + case PIPE_LOGICOP_NOOP: return 0xAA; + case PIPE_LOGICOP_XOR: return 0x66; + case PIPE_LOGICOP_OR: return 0xEE; + case PIPE_LOGICOP_NOR: return 0x11; + case PIPE_LOGICOP_EQUIV: return 0x99; + case PIPE_LOGICOP_INVERT: return 0x55; + case PIPE_LOGICOP_OR_REVERSE: return 0xDD; + case PIPE_LOGICOP_COPY_INVERTED: return 0x33; + case PIPE_LOGICOP_OR_INVERTED: return 0xBB; + case PIPE_LOGICOP_NAND: return 0x77; + case PIPE_LOGICOP_SET: return 0xFF; + default: return 0; + } +} + + +/* Copy BitBlt + */ +void brw_copy_blit(struct brw_context *brw, + unsigned do_flip, + unsigned cpp, + short src_pitch, + struct pipe_buffer *src_buffer, + unsigned src_offset, + boolean src_tiled, + short dst_pitch, + struct pipe_buffer *dst_buffer, + unsigned dst_offset, + boolean dst_tiled, + short src_x, short src_y, + short dst_x, short dst_y, + short w, short h, + unsigned logic_op) +{ + unsigned CMD, BR13; + int dst_y2 = dst_y + h; + int dst_x2 = dst_x + w; + BATCH_LOCALS; + + + DBG("%s src:buf(%d)/%d %d,%d dst:buf(%d)/%d %d,%d sz:%dx%d op:%d\n", + __FUNCTION__, + src_buffer, src_pitch, src_x, src_y, + dst_buffer, dst_pitch, dst_x, dst_y, + w,h,logic_op); + + assert( logic_op - PIPE_LOGICOP_CLEAR >= 0 ); + assert( logic_op - PIPE_LOGICOP_CLEAR < 0x10 ); + + src_pitch *= cpp; + dst_pitch *= cpp; + + switch(cpp) { + case 1: + case 2: + case 3: + BR13 = (translate_raster_op(logic_op) << 16) | (1<<24); + CMD = XY_SRC_COPY_BLT_CMD; + break; + case 4: + BR13 = (translate_raster_op(logic_op) << 16) | (1<<24) | + (1<<25); + CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; + break; + default: + return; + } + + if (src_tiled) { + CMD |= XY_SRC_TILED; + src_pitch /= 4; + } + + if (dst_tiled) { + CMD |= XY_DST_TILED; + dst_pitch /= 4; + } + + if (dst_y2 < dst_y || + dst_x2 < dst_x) { + return; + } + + dst_pitch &= 0xffff; + src_pitch &= 0xffff; + + /* Initial y values don't seem to work with negative pitches. If + * we adjust the offsets manually (below), it seems to work fine. + * + * On the other hand, if we always adjust, the hardware doesn't + * know which blit directions to use, so overlapping copypixels get + * the wrong result. + */ + if (dst_pitch > 0 && src_pitch > 0) { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH( CMD ); + OUT_BATCH( dst_pitch | BR13 ); + OUT_BATCH( (dst_y << 16) | dst_x ); + OUT_BATCH( (dst_y2 << 16) | dst_x2 ); + OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, + dst_offset ); + OUT_BATCH( (src_y << 16) | src_x ); + OUT_BATCH( src_pitch ); + OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ, + src_offset ); + ADVANCE_BATCH(); + } + else { + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH( CMD ); + OUT_BATCH( (dst_pitch & 0xffff) | BR13 ); + OUT_BATCH( (0 << 16) | dst_x ); + OUT_BATCH( (h << 16) | dst_x2 ); + OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, + dst_offset + dst_y * dst_pitch ); + OUT_BATCH( (src_pitch & 0xffff) ); + OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ, + src_offset + src_y * src_pitch ); + ADVANCE_BATCH(); + } +} + + + diff --git a/src/gallium/drivers/i965simple/brw_blit.h b/src/gallium/drivers/i965simple/brw_blit.h new file mode 100644 index 0000000000..111c5d91d3 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_blit.h @@ -0,0 +1,33 @@ +#ifndef BRW_BLIT_H +#define BRW_BLIT_H + +#include "pipe/p_compiler.h" + +struct pipe_buffer; +struct brw_context; + +void brw_fill_blit(struct brw_context *intel, + unsigned cpp, + short dst_pitch, + struct pipe_buffer *dst_buffer, + unsigned dst_offset, + boolean dst_tiled, + short x, short y, + short w, short h, + unsigned color); +void brw_copy_blit(struct brw_context *intel, + unsigned do_flip, + unsigned cpp, + short src_pitch, + struct pipe_buffer *src_buffer, + unsigned src_offset, + boolean src_tiled, + short dst_pitch, + struct pipe_buffer *dst_buffer, + unsigned dst_offset, + boolean dst_tiled, + short src_x, short src_y, + short dst_x, short dst_y, + short w, short h, + unsigned logic_op); +#endif diff --git a/src/gallium/drivers/i965simple/brw_cc.c b/src/gallium/drivers/i965simple/brw_cc.c new file mode 100644 index 0000000000..337e4f95f6 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_cc.c @@ -0,0 +1,269 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_util.h" + + +static int brw_translate_compare_func(int func) +{ + switch(func) { + case PIPE_FUNC_NEVER: + return BRW_COMPAREFUNCTION_NEVER; + case PIPE_FUNC_LESS: + return BRW_COMPAREFUNCTION_LESS; + case PIPE_FUNC_LEQUAL: + return BRW_COMPAREFUNCTION_LEQUAL; + case PIPE_FUNC_GREATER: + return BRW_COMPAREFUNCTION_GREATER; + case PIPE_FUNC_GEQUAL: + return BRW_COMPAREFUNCTION_GEQUAL; + case PIPE_FUNC_NOTEQUAL: + return BRW_COMPAREFUNCTION_NOTEQUAL; + case PIPE_FUNC_EQUAL: + return BRW_COMPAREFUNCTION_EQUAL; + case PIPE_FUNC_ALWAYS: + return BRW_COMPAREFUNCTION_ALWAYS; + } + + debug_printf("Unknown value in %s: %x\n", __FUNCTION__, func); + return BRW_COMPAREFUNCTION_ALWAYS; +} + +static int brw_translate_stencil_op(int op) +{ + switch(op) { + case PIPE_STENCIL_OP_KEEP: + return BRW_STENCILOP_KEEP; + case PIPE_STENCIL_OP_ZERO: + return BRW_STENCILOP_ZERO; + case PIPE_STENCIL_OP_REPLACE: + return BRW_STENCILOP_REPLACE; + case PIPE_STENCIL_OP_INCR: + return BRW_STENCILOP_INCRSAT; + case PIPE_STENCIL_OP_DECR: + return BRW_STENCILOP_DECRSAT; + case PIPE_STENCIL_OP_INCR_WRAP: + return BRW_STENCILOP_INCR; + case PIPE_STENCIL_OP_DECR_WRAP: + return BRW_STENCILOP_DECR; + case PIPE_STENCIL_OP_INVERT: + return BRW_STENCILOP_INVERT; + default: + return BRW_STENCILOP_ZERO; + } +} + + +static int brw_translate_logic_op(int opcode) +{ + switch(opcode) { + case PIPE_LOGICOP_CLEAR: + return BRW_LOGICOPFUNCTION_CLEAR; + case PIPE_LOGICOP_AND: + return BRW_LOGICOPFUNCTION_AND; + case PIPE_LOGICOP_AND_REVERSE: + return BRW_LOGICOPFUNCTION_AND_REVERSE; + case PIPE_LOGICOP_COPY: + return BRW_LOGICOPFUNCTION_COPY; + case PIPE_LOGICOP_COPY_INVERTED: + return BRW_LOGICOPFUNCTION_COPY_INVERTED; + case PIPE_LOGICOP_AND_INVERTED: + return BRW_LOGICOPFUNCTION_AND_INVERTED; + case PIPE_LOGICOP_NOOP: + return BRW_LOGICOPFUNCTION_NOOP; + case PIPE_LOGICOP_XOR: + return BRW_LOGICOPFUNCTION_XOR; + case PIPE_LOGICOP_OR: + return BRW_LOGICOPFUNCTION_OR; + case PIPE_LOGICOP_OR_INVERTED: + return BRW_LOGICOPFUNCTION_OR_INVERTED; + case PIPE_LOGICOP_NOR: + return BRW_LOGICOPFUNCTION_NOR; + case PIPE_LOGICOP_EQUIV: + return BRW_LOGICOPFUNCTION_EQUIV; + case PIPE_LOGICOP_INVERT: + return BRW_LOGICOPFUNCTION_INVERT; + case PIPE_LOGICOP_OR_REVERSE: + return BRW_LOGICOPFUNCTION_OR_REVERSE; + case PIPE_LOGICOP_NAND: + return BRW_LOGICOPFUNCTION_NAND; + case PIPE_LOGICOP_SET: + return BRW_LOGICOPFUNCTION_SET; + default: + return BRW_LOGICOPFUNCTION_SET; + } +} + + +static void upload_cc_vp( struct brw_context *brw ) +{ + struct brw_cc_viewport ccv; + + memset(&ccv, 0, sizeof(ccv)); + + ccv.min_depth = 0.0; + ccv.max_depth = 1.0; + + brw->cc.vp_gs_offset = brw_cache_data( &brw->cache[BRW_CC_VP], &ccv ); +} + +const struct brw_tracked_state brw_cc_vp = { + .dirty = { + .brw = BRW_NEW_SCENE, + .cache = 0 + }, + .update = upload_cc_vp +}; + + +static void upload_cc_unit( struct brw_context *brw ) +{ + struct brw_cc_unit_state cc; + + memset(&cc, 0, sizeof(cc)); + + /* BRW_NEW_DEPTH_STENCIL */ + if (brw->attribs.DepthStencil->stencil[0].enabled) { + cc.cc0.stencil_enable = brw->attribs.DepthStencil->stencil[0].enabled; + cc.cc0.stencil_func = brw_translate_compare_func(brw->attribs.DepthStencil->stencil[0].func); + cc.cc0.stencil_fail_op = brw_translate_stencil_op(brw->attribs.DepthStencil->stencil[0].fail_op); + cc.cc0.stencil_pass_depth_fail_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil[0].zfail_op); + cc.cc0.stencil_pass_depth_pass_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil[0].zpass_op); + cc.cc1.stencil_ref = brw->attribs.DepthStencil->stencil[0].ref_value; + cc.cc1.stencil_write_mask = brw->attribs.DepthStencil->stencil[0].write_mask; + cc.cc1.stencil_test_mask = brw->attribs.DepthStencil->stencil[0].value_mask; + + if (brw->attribs.DepthStencil->stencil[1].enabled) { + cc.cc0.bf_stencil_enable = brw->attribs.DepthStencil->stencil[1].enabled; + cc.cc0.bf_stencil_func = brw_translate_compare_func( + brw->attribs.DepthStencil->stencil[1].func); + cc.cc0.bf_stencil_fail_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil[1].fail_op); + cc.cc0.bf_stencil_pass_depth_fail_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil[1].zfail_op); + cc.cc0.bf_stencil_pass_depth_pass_op = brw_translate_stencil_op( + brw->attribs.DepthStencil->stencil[1].zpass_op); + cc.cc1.bf_stencil_ref = brw->attribs.DepthStencil->stencil[1].ref_value; + cc.cc2.bf_stencil_write_mask = brw->attribs.DepthStencil->stencil[1].write_mask; + cc.cc2.bf_stencil_test_mask = brw->attribs.DepthStencil->stencil[1].value_mask; + } + + /* Not really sure about this: + */ + if (brw->attribs.DepthStencil->stencil[0].write_mask || + brw->attribs.DepthStencil->stencil[1].write_mask) + cc.cc0.stencil_write_enable = 1; + } + + /* BRW_NEW_BLEND */ + if (brw->attribs.Blend->logicop_enable) { + cc.cc2.logicop_enable = 1; + cc.cc5.logicop_func = brw_translate_logic_op( brw->attribs.Blend->logicop_func ); + } + else if (brw->attribs.Blend->blend_enable) { + int eqRGB = brw->attribs.Blend->rgb_func; + int eqA = brw->attribs.Blend->alpha_func; + int srcRGB = brw->attribs.Blend->rgb_src_factor; + int dstRGB = brw->attribs.Blend->rgb_dst_factor; + int srcA = brw->attribs.Blend->alpha_src_factor; + int dstA = brw->attribs.Blend->alpha_dst_factor; + + if (eqRGB == PIPE_BLEND_MIN || eqRGB == PIPE_BLEND_MAX) { + srcRGB = dstRGB = PIPE_BLENDFACTOR_ONE; + } + + if (eqA == PIPE_BLEND_MIN || eqA == PIPE_BLEND_MAX) { + srcA = dstA = PIPE_BLENDFACTOR_ONE; + } + + cc.cc6.dest_blend_factor = brw_translate_blend_factor(dstRGB); + cc.cc6.src_blend_factor = brw_translate_blend_factor(srcRGB); + cc.cc6.blend_function = brw_translate_blend_equation( eqRGB ); + + cc.cc5.ia_dest_blend_factor = brw_translate_blend_factor(dstA); + cc.cc5.ia_src_blend_factor = brw_translate_blend_factor(srcA); + cc.cc5.ia_blend_function = brw_translate_blend_equation( eqA ); + + cc.cc3.blend_enable = 1; + cc.cc3.ia_blend_enable = (srcA != srcRGB || + dstA != dstRGB || + eqA != eqRGB); + } + + /* BRW_NEW_ALPHATEST + */ + if (brw->attribs.DepthStencil->alpha.enabled) { + cc.cc3.alpha_test = 1; + cc.cc3.alpha_test_func = + brw_translate_compare_func(brw->attribs.DepthStencil->alpha.func); + + UNCLAMPED_FLOAT_TO_UBYTE(cc.cc7.alpha_ref.ub[0], + brw->attribs.DepthStencil->alpha.ref); + + cc.cc3.alpha_test_format = BRW_ALPHATEST_FORMAT_UNORM8; + } + + if (brw->attribs.Blend->dither) { + cc.cc5.dither_enable = 1; + cc.cc6.y_dither_offset = 0; + cc.cc6.x_dither_offset = 0; + } + + if (brw->attribs.DepthStencil->depth.enabled) { + cc.cc2.depth_test = brw->attribs.DepthStencil->depth.enabled; + cc.cc2.depth_test_function = brw_translate_compare_func(brw->attribs.DepthStencil->depth.func); + cc.cc2.depth_write_enable = brw->attribs.DepthStencil->depth.writemask; + } + + /* CACHE_NEW_CC_VP */ + cc.cc4.cc_viewport_state_offset = brw->cc.vp_gs_offset >> 5; + + if (BRW_DEBUG & DEBUG_STATS) + cc.cc5.statistics_enable = 1; + + brw->cc.state_gs_offset = brw_cache_data( &brw->cache[BRW_CC_UNIT], &cc ); +} + +const struct brw_tracked_state brw_cc_unit = { + .dirty = { + .brw = BRW_NEW_DEPTH_STENCIL | BRW_NEW_BLEND | BRW_NEW_ALPHA_TEST, + .cache = CACHE_NEW_CC_VP + }, + .update = upload_cc_unit +}; + diff --git a/src/gallium/drivers/i965simple/brw_clip.c b/src/gallium/drivers/i965simple/brw_clip.c new file mode 100644 index 0000000000..268124cc53 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_clip.c @@ -0,0 +1,206 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_state.h" +#include "brw_clip.h" + +#define FRONT_UNFILLED_BIT 0x1 +#define BACK_UNFILLED_BIT 0x2 + + +static void compile_clip_prog( struct brw_context *brw, + struct brw_clip_prog_key *key ) +{ + struct brw_clip_compile c; + const unsigned *program; + unsigned program_size; + unsigned delta; + unsigned i; + + memset(&c, 0, sizeof(c)); + + /* Begin the compilation: + */ + brw_init_compile(&c.func); + + c.func.single_program_flow = 1; + + c.key = *key; + + + /* Need to locate the two positions present in vertex + header. + * These are currently hardcoded: + */ + c.header_position_offset = ATTR_SIZE; + + for (i = 0, delta = REG_SIZE; i < PIPE_MAX_SHADER_OUTPUTS; i++) + if (c.key.attrs & (1<primitive) { + case PIPE_PRIM_TRIANGLES: +#if 0 + if (key->do_unfilled) + brw_emit_unfilled_clip( &c ); + else +#endif + brw_emit_tri_clip( &c ); + break; + case PIPE_PRIM_LINES: + brw_emit_line_clip( &c ); + break; + case PIPE_PRIM_POINTS: + brw_emit_point_clip( &c ); + break; + default: + assert(0); + return; + } + + + + /* get the program + */ + program = brw_get_program(&c.func, &program_size); + + /* Upload + */ + brw->clip.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_CLIP_PROG], + &c.key, + sizeof(c.key), + program, + program_size, + &c.prog_data, + &brw->clip.prog_data ); +} + + +static boolean search_cache( struct brw_context *brw, + struct brw_clip_prog_key *key ) +{ + return brw_search_cache(&brw->cache[BRW_CLIP_PROG], + key, sizeof(*key), + &brw->clip.prog_data, + &brw->clip.prog_gs_offset); +} + + + + +/* Calculate interpolants for triangle and line rasterization. + */ +static void upload_clip_prog(struct brw_context *brw) +{ + struct brw_clip_prog_key key; + + memset(&key, 0, sizeof(key)); + + /* Populate the key: + */ + /* BRW_NEW_REDUCED_PRIMITIVE */ + key.primitive = brw->reduced_primitive; + /* CACHE_NEW_VS_PROG */ + key.attrs = brw->vs.prog_data->outputs_written; + /* BRW_NEW_RASTER */ + key.do_flat_shading = (brw->attribs.Raster->flatshade); + /* BRW_NEW_CLIP */ + key.nr_userclip = brw->attribs.Clip.nr; /* XXX */ + +#if 0 + key.clip_mode = BRW_CLIPMODE_NORMAL; + + if (key.primitive == PIPE_PRIM_TRIANGLES) { + if (brw->attribs.Raster->cull_mode == PIPE_WINDING_BOTH) + key.clip_mode = BRW_CLIPMODE_REJECT_ALL; + else { + if (brw->attribs.Raster->fill_cw != PIPE_POLYGON_MODE_FILL || + brw->attribs.Raster->fill_ccw != PIPE_POLYGON_MODE_FILL) + key.do_unfilled = 1; + + /* Most cases the fixed function units will handle. Cases where + * one or more polygon faces are unfilled will require help: + */ + if (key.do_unfilled) { + key.clip_mode = BRW_CLIPMODE_CLIP_NON_REJECTED; + + if (brw->attribs.Raster->offset_cw || + brw->attribs.Raster->offset_ccw) { + key.offset_units = brw->attribs.Raster->offset_units; + key.offset_factor = brw->attribs.Raster->offset_scale; + } + key.fill_ccw = brw->attribs.Raster->fill_ccw; + key.fill_cw = brw->attribs.Raster->fill_cw; + key.offset_ccw = brw->attribs.Raster->offset_ccw; + key.offset_cw = brw->attribs.Raster->offset_cw; + if (brw->attribs.Raster->light_twoside && + key.fill_cw != CLIP_CULL) + key.copy_bfc_cw = 1; + } + } + } +#else + key.clip_mode = BRW_CLIPMODE_ACCEPT_ALL; +#endif + + if (!search_cache(brw, &key)) + compile_clip_prog( brw, &key ); +} + +const struct brw_tracked_state brw_clip_prog = { + .dirty = { + .brw = (BRW_NEW_RASTERIZER | + BRW_NEW_CLIP | + BRW_NEW_REDUCED_PRIMITIVE), + .cache = CACHE_NEW_VS_PROG + }, + .update = upload_clip_prog +}; diff --git a/src/gallium/drivers/i965simple/brw_clip.h b/src/gallium/drivers/i965simple/brw_clip.h new file mode 100644 index 0000000000..a89d08b791 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_clip.h @@ -0,0 +1,170 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#ifndef BRW_CLIP_H +#define BRW_CLIP_H + + +#include "brw_context.h" +#include "brw_eu.h" + +#define MAX_VERTS (3+6+6) + +/* Note that if unfilled primitives are being emitted, we have to fix + * up polygon offset and flatshading at this point: + */ +struct brw_clip_prog_key { + unsigned attrs:32; + unsigned primitive:4; + unsigned nr_userclip:3; + unsigned do_flat_shading:1; + unsigned do_unfilled:1; + unsigned fill_cw:2; /* includes cull information */ + unsigned fill_ccw:2; /* includes cull information */ + unsigned offset_cw:1; + unsigned offset_ccw:1; + unsigned pad0:17; + + unsigned copy_bfc_cw:1; + unsigned copy_bfc_ccw:1; + unsigned clip_mode:3; + unsigned pad1:27; + + float offset_factor; + float offset_units; +}; + + +#define CLIP_LINE 0 +#define CLIP_POINT 1 +#define CLIP_FILL 2 +#define CLIP_CULL 3 + + +#define PRIM_MASK (0x1f) + +struct brw_clip_compile { + struct brw_compile func; + struct brw_clip_prog_key key; + struct brw_clip_prog_data prog_data; + + struct { + struct brw_reg R0; + struct brw_reg vertex[MAX_VERTS]; + + struct brw_reg t; + struct brw_reg t0, t1; + struct brw_reg dp0, dp1; + + struct brw_reg dpPrev; + struct brw_reg dp; + struct brw_reg loopcount; + struct brw_reg nr_verts; + struct brw_reg planemask; + + struct brw_reg inlist; + struct brw_reg outlist; + struct brw_reg freelist; + + struct brw_reg dir; + struct brw_reg tmp0, tmp1; + struct brw_reg offset; + + struct brw_reg fixed_planes; + struct brw_reg plane_equation; + } reg; + + /* 3 different ways of expressing vertex size: + */ + unsigned nr_attrs; + unsigned nr_regs; + unsigned nr_bytes; + + unsigned first_tmp; + unsigned last_tmp; + + boolean need_direction; + + unsigned last_mrf; + + unsigned header_position_offset; + unsigned offset[PIPE_ATTRIB_MAX]; +}; + +#define ATTR_SIZE (4*4) + +/* Points are only culled, so no need for a clip routine, however it + * works out easier to have a dummy one. + */ +void brw_emit_unfilled_clip( struct brw_clip_compile *c ); +void brw_emit_tri_clip( struct brw_clip_compile *c ); +void brw_emit_line_clip( struct brw_clip_compile *c ); +void brw_emit_point_clip( struct brw_clip_compile *c ); + +/* brw_clip_tri.c, for use by the unfilled clip routine: + */ +void brw_clip_tri_init_vertices( struct brw_clip_compile *c ); +void brw_clip_tri_flat_shade( struct brw_clip_compile *c ); +void brw_clip_tri( struct brw_clip_compile *c ); +void brw_clip_tri_emit_polygon( struct brw_clip_compile *c ); +void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, + unsigned nr_verts ); + + +/* Utils: + */ + +void brw_clip_interp_vertex( struct brw_clip_compile *c, + struct brw_indirect dest_ptr, + struct brw_indirect v0_ptr, /* from */ + struct brw_indirect v1_ptr, /* to */ + struct brw_reg t0, + boolean force_edgeflag ); + +void brw_clip_init_planes( struct brw_clip_compile *c ); + +void brw_clip_emit_vue(struct brw_clip_compile *c, + struct brw_indirect vert, + boolean allocate, + boolean eot, + unsigned header); + +void brw_clip_kill_thread(struct brw_clip_compile *c); + +struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c ); +struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c ); + +void brw_clip_copy_colors( struct brw_clip_compile *c, + unsigned to, unsigned from ); + +void brw_clip_init_clipmask( struct brw_clip_compile *c ); + +#endif diff --git a/src/gallium/drivers/i965simple/brw_clip_line.c b/src/gallium/drivers/i965simple/brw_clip_line.c new file mode 100644 index 0000000000..75d9e5fcda --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_clip_line.c @@ -0,0 +1,245 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + + +static void brw_clip_line_alloc_regs( struct brw_clip_compile *c ) +{ + unsigned i = 0,j; + + /* Register usage is static, precompute here: + */ + c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; + + if (c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec4_grf(i, 0); + i += (6 + c->key.nr_userclip + 1) / 2; + + c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2; + } + else + c->prog_data.curb_read_length = 0; + + + /* Payload vertices plus space for more generated vertices: + */ + for (j = 0; j < 4; j++) { + c->reg.vertex[j] = brw_vec4_grf(i, 0); + i += c->nr_regs; + } + + c->reg.t = brw_vec1_grf(i, 0); + c->reg.t0 = brw_vec1_grf(i, 1); + c->reg.t1 = brw_vec1_grf(i, 2); + c->reg.planemask = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD); + c->reg.plane_equation = brw_vec4_grf(i, 4); + i++; + + c->reg.dp0 = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */ + c->reg.dp1 = brw_vec1_grf(i, 4); + i++; + + if (!c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec8_grf(i, 0); + i++; + } + + + c->first_tmp = i; + c->last_tmp = i; + + c->prog_data.urb_read_length = c->nr_regs; /* ? */ + c->prog_data.total_grf = i; +} + + + +/* Line clipping, more or less following the following algorithm: + * + * for (p=0;p t1) t1 = t; + * } else { + * float t = dp0 / (dp0 - dp1); + * if (t > t0) t0 = t; + * } + * + * if (t0 + t1 >= 1.0) + * return; + * } + * } + * + * interp( ctx, newvtx0, vtx0, vtx1, t0 ); + * interp( ctx, newvtx1, vtx1, vtx0, t1 ); + * + */ +static void clip_and_emit_line( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_indirect vtx0 = brw_indirect(0, 0); + struct brw_indirect vtx1 = brw_indirect(1, 0); + struct brw_indirect newvtx0 = brw_indirect(2, 0); + struct brw_indirect newvtx1 = brw_indirect(3, 0); + struct brw_indirect plane_ptr = brw_indirect(4, 0); + struct brw_instruction *plane_loop; + struct brw_instruction *plane_active; + struct brw_instruction *is_negative; + struct brw_instruction *is_neg2; + struct brw_instruction *not_culled; + struct brw_reg v1_null_ud = retype(vec1(brw_null_reg()), BRW_REGISTER_TYPE_UD); + + brw_MOV(p, get_addr_reg(vtx0), brw_address(c->reg.vertex[0])); + brw_MOV(p, get_addr_reg(vtx1), brw_address(c->reg.vertex[1])); + brw_MOV(p, get_addr_reg(newvtx0), brw_address(c->reg.vertex[2])); + brw_MOV(p, get_addr_reg(newvtx1), brw_address(c->reg.vertex[3])); + brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c)); + + /* Note: init t0, t1 together: + */ + brw_MOV(p, vec2(c->reg.t0), brw_imm_f(0)); + + brw_clip_init_planes(c); + brw_clip_init_clipmask(c); + + /* -ve rhw workaround */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), + brw_imm_ud(1<<20)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(0x3f)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + plane_loop = brw_DO(p, BRW_EXECUTE_1); + { + /* if (planemask & 1) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, v1_null_ud, c->reg.planemask, brw_imm_ud(1)); + + plane_active = brw_IF(p, BRW_EXECUTE_1); + { + if (c->key.nr_userclip) + brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0)); + else + brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0)); + +#if 0 + /* dp = DP4(vtx->position, plane) + */ + brw_DP4(p, vec4(c->reg.dp0), deref_4f(vtx0, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); + + /* if (IS_NEGATIVE(dp1)) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, vec4(c->reg.dp1), deref_4f(vtx1, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); +#else + #warning "disabled" +#endif + is_negative = brw_IF(p, BRW_EXECUTE_1); + { + brw_ADD(p, c->reg.t, c->reg.dp1, negate(c->reg.dp0)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp1); + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t1 ); + brw_MOV(p, c->reg.t1, c->reg.t); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + is_negative = brw_ELSE(p, is_negative); + { + /* Coming back in. We know that both cannot be negative + * because the line would have been culled in that case. + */ + + /* If both are positive, do nothing */ + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.dp0, brw_imm_f(0.0)); + is_neg2 = brw_IF(p, BRW_EXECUTE_1); + { + brw_ADD(p, c->reg.t, c->reg.dp0, negate(c->reg.dp1)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp0); + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t0 ); + brw_MOV(p, c->reg.t0, c->reg.t); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + brw_ENDIF(p, is_neg2); + } + brw_ENDIF(p, is_negative); + } + brw_ENDIF(p, plane_active); + + /* plane_ptr++; + */ + brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c)); + + /* while (planemask>>=1) != 0 + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1)); + } + brw_WHILE(p, plane_loop); + + brw_ADD(p, c->reg.t, c->reg.t0, c->reg.t1); + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.t, brw_imm_f(1.0)); + not_culled = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_interp_vertex(c, newvtx0, vtx0, vtx1, c->reg.t0, FALSE); + brw_clip_interp_vertex(c, newvtx1, vtx1, vtx0, c->reg.t1, FALSE); + + brw_clip_emit_vue(c, newvtx0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START); + brw_clip_emit_vue(c, newvtx1, 0, 1, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END); + } + brw_ENDIF(p, not_culled); + brw_clip_kill_thread(c); +} + + + +void brw_emit_line_clip( struct brw_clip_compile *c ) +{ + brw_clip_line_alloc_regs(c); + + if (c->key.do_flat_shading) + brw_clip_copy_colors(c, 0, 1); + + clip_and_emit_line(c); +} diff --git a/src/gallium/drivers/i965simple/brw_clip_point.c b/src/gallium/drivers/i965simple/brw_clip_point.c new file mode 100644 index 0000000000..6fce7210d1 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_clip_point.c @@ -0,0 +1,47 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + +/* Point clipping, nothing to do? + */ +void brw_emit_point_clip( struct brw_clip_compile *c ) +{ + /* Send an empty message to kill the thread: + */ + brw_clip_tri_alloc_regs(c, 0); + brw_clip_kill_thread(c); +} diff --git a/src/gallium/drivers/i965simple/brw_clip_state.c b/src/gallium/drivers/i965simple/brw_clip_state.c new file mode 100644 index 0000000000..ea5c05a279 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_clip_state.c @@ -0,0 +1,92 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "pipe/p_util.h" + + +static void upload_clip_unit( struct brw_context *brw ) +{ + struct brw_clip_unit_state clip; + + memset(&clip, 0, sizeof(clip)); + + /* CACHE_NEW_CLIP_PROG */ + clip.thread0.grf_reg_count = + align(brw->clip.prog_data->total_grf, 16) / 16 - 1; + clip.thread0.kernel_start_pointer = brw->clip.prog_gs_offset >> 6; + clip.thread3.urb_entry_read_length = brw->clip.prog_data->urb_read_length; + clip.thread3.const_urb_entry_read_length = brw->clip.prog_data->curb_read_length; + clip.clip5.clip_mode = brw->clip.prog_data->clip_mode; + + /* BRW_NEW_CURBE_OFFSETS */ + clip.thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2; + + /* BRW_NEW_URB_FENCE */ + clip.thread4.nr_urb_entries = brw->urb.nr_clip_entries; + clip.thread4.urb_entry_allocation_size = brw->urb.vsize - 1; + clip.thread4.max_threads = 1; /* 2 threads */ + + if (BRW_DEBUG & DEBUG_STATS) + clip.thread4.stats_enable = 1; + + /* CONSTANT */ + clip.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + clip.thread1.single_program_flow = 1; + clip.thread3.dispatch_grf_start_reg = 1; + clip.thread3.urb_entry_read_offset = 0; + clip.clip5.userclip_enable_flags = 0x7f; + clip.clip5.userclip_must_clip = 1; + clip.clip5.guard_band_enable = 0; + clip.clip5.viewport_z_clip_enable = 1; + clip.clip5.viewport_xy_clip_enable = 1; + clip.clip5.vertex_position_space = BRW_CLIP_NDCSPACE; + clip.clip5.api_mode = BRW_CLIP_API_OGL; + clip.clip6.clipper_viewport_state_ptr = 0; + clip.viewport_xmin = -1; + clip.viewport_xmax = 1; + clip.viewport_ymin = -1; + clip.viewport_ymax = 1; + + brw->clip.state_gs_offset = brw_cache_data( &brw->cache[BRW_CLIP_UNIT], &clip ); +} + + +const struct brw_tracked_state brw_clip_unit = { + .dirty = { + .brw = (BRW_NEW_CURBE_OFFSETS | + BRW_NEW_URB_FENCE), + .cache = CACHE_NEW_CLIP_PROG + }, + .update = upload_clip_unit +}; diff --git a/src/gallium/drivers/i965simple/brw_clip_tri.c b/src/gallium/drivers/i965simple/brw_clip_tri.c new file mode 100644 index 0000000000..c5da7b825e --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_clip_tri.c @@ -0,0 +1,566 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + +static struct brw_reg get_tmp( struct brw_clip_compile *c ) +{ + struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0); + + if (++c->last_tmp > c->prog_data.total_grf) + c->prog_data.total_grf = c->last_tmp; + + return tmp; +} + +static void release_tmps( struct brw_clip_compile *c ) +{ + c->last_tmp = c->first_tmp; +} + + +void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, + unsigned nr_verts ) +{ + unsigned i = 0,j; + + /* Register usage is static, precompute here: + */ + c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; + + if (c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec4_grf(i, 0); + i += (6 + c->key.nr_userclip + 1) / 2; + + c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2; + } + else + c->prog_data.curb_read_length = 0; + + + /* Payload vertices plus space for more generated vertices: + */ + for (j = 0; j < nr_verts; j++) { + c->reg.vertex[j] = brw_vec4_grf(i, 0); + i += c->nr_regs; + } + + if (c->nr_attrs & 1) { + for (j = 0; j < 3; j++) { + unsigned delta = c->nr_attrs*16 + 32; + brw_MOV(&c->func, byte_offset(c->reg.vertex[j], delta), brw_imm_f(0)); + } + } + + c->reg.t = brw_vec1_grf(i, 0); + c->reg.loopcount = retype(brw_vec1_grf(i, 1), BRW_REGISTER_TYPE_UD); + c->reg.nr_verts = retype(brw_vec1_grf(i, 2), BRW_REGISTER_TYPE_UD); + c->reg.planemask = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD); + c->reg.plane_equation = brw_vec4_grf(i, 4); + i++; + + c->reg.dpPrev = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */ + c->reg.dp = brw_vec1_grf(i, 4); + i++; + + c->reg.inlist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); + i++; + + c->reg.outlist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); + i++; + + c->reg.freelist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); + i++; + + if (!c->key.nr_userclip) { + c->reg.fixed_planes = brw_vec8_grf(i, 0); + i++; + } + + if (c->key.do_unfilled) { + c->reg.dir = brw_vec4_grf(i, 0); + c->reg.offset = brw_vec4_grf(i, 4); + i++; + c->reg.tmp0 = brw_vec4_grf(i, 0); + c->reg.tmp1 = brw_vec4_grf(i, 4); + i++; + } + + c->first_tmp = i; + c->last_tmp = i; + + c->prog_data.urb_read_length = c->nr_regs; /* ? */ + c->prog_data.total_grf = i; +} + + + +void brw_clip_tri_init_vertices( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */ + struct brw_instruction *is_rev; + + /* Initial list of indices for incoming vertexes: + */ + brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_EQ, + tmp0, + brw_imm_ud(_3DPRIM_TRISTRIP_REVERSE)); + + /* XXX: Is there an easier way to do this? Need to reverse every + * second tristrip element: Can ignore sometimes? + */ + is_rev = brw_IF(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[1]) ); + brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[0]) ); + if (c->need_direction) + brw_MOV(p, c->reg.dir, brw_imm_f(-1)); + } + is_rev = brw_ELSE(p, is_rev); + { + brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[0]) ); + brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[1]) ); + if (c->need_direction) + brw_MOV(p, c->reg.dir, brw_imm_f(1)); + } + brw_ENDIF(p, is_rev); + + brw_MOV(p, get_element(c->reg.inlist, 2), brw_address(c->reg.vertex[2]) ); + brw_MOV(p, brw_vec8_grf(c->reg.outlist.nr, 0), brw_imm_f(0)); + brw_MOV(p, c->reg.nr_verts, brw_imm_ud(3)); +} + + + +void brw_clip_tri_flat_shade( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *is_poly; + struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */ + + brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_EQ, + tmp0, + brw_imm_ud(_3DPRIM_POLYGON)); + + is_poly = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_copy_colors(c, 1, 0); + brw_clip_copy_colors(c, 2, 0); + } + is_poly = brw_ELSE(p, is_poly); + { + brw_clip_copy_colors(c, 0, 2); + brw_clip_copy_colors(c, 1, 2); + } + brw_ENDIF(p, is_poly); +} + + + +/* Use mesa's clipping algorithms, translated to GEN4 assembly. + */ +void brw_clip_tri( struct brw_clip_compile *c ) +{ +#if 0 + struct brw_compile *p = &c->func; + struct brw_indirect vtx = brw_indirect(0, 0); + struct brw_indirect vtxPrev = brw_indirect(1, 0); + struct brw_indirect vtxOut = brw_indirect(2, 0); + struct brw_indirect plane_ptr = brw_indirect(3, 0); + struct brw_indirect inlist_ptr = brw_indirect(4, 0); + struct brw_indirect outlist_ptr = brw_indirect(5, 0); + struct brw_indirect freelist_ptr = brw_indirect(6, 0); + struct brw_instruction *plane_loop; + struct brw_instruction *plane_active; + struct brw_instruction *vertex_loop; + struct brw_instruction *next_test; + struct brw_instruction *prev_test; + + brw_MOV(p, get_addr_reg(vtxPrev), brw_address(c->reg.vertex[2]) ); + brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c)); + brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist)); + brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist)); + + brw_MOV(p, get_addr_reg(freelist_ptr), brw_address(c->reg.vertex[3]) ); + + plane_loop = brw_DO(p, BRW_EXECUTE_1); + { + /* if (planemask & 1) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, vec1(brw_null_reg()), c->reg.planemask, brw_imm_ud(1)); + + plane_active = brw_IF(p, BRW_EXECUTE_1); + { + /* vtxOut = freelist_ptr++ + */ + brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(freelist_ptr) ); + brw_ADD(p, get_addr_reg(freelist_ptr), get_addr_reg(freelist_ptr), brw_imm_uw(c->nr_regs * REG_SIZE)); + + if (c->key.nr_userclip) + brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0)); + else + brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0)); + + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, c->reg.nr_verts, brw_imm_ud(0)); + + vertex_loop = brw_DO(p, BRW_EXECUTE_1); + { + /* vtx = *input_ptr; + */ + brw_MOV(p, get_addr_reg(vtx), deref_1uw(inlist_ptr, 0)); + + /* IS_NEGATIVE(prev) */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, vec4(c->reg.dpPrev), deref_4f(vtxPrev, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); + prev_test = brw_IF(p, BRW_EXECUTE_1); + { + /* IS_POSITIVE(next) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_GE); + brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); + next_test = brw_IF(p, BRW_EXECUTE_1); + { + + /* Coming back in. + */ + brw_ADD(p, c->reg.t, c->reg.dpPrev, negate(c->reg.dp)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dpPrev); + + /* If (vtxOut == 0) vtxOut = vtxPrev + */ + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) ); + brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtxPrev) ); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_clip_interp_vertex(c, vtxOut, vtxPrev, vtx, c->reg.t, FALSE); + + /* *outlist_ptr++ = vtxOut; + * nr_verts++; + * vtxOut = 0; + */ + brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut)); + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); + brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); + brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) ); + } + brw_ENDIF(p, next_test); + + } + prev_test = brw_ELSE(p, prev_test); + { + /* *outlist_ptr++ = vtxPrev; + * nr_verts++; + */ + brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxPrev)); + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); + brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); + + /* IS_NEGATIVE(next) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); + next_test = brw_IF(p, BRW_EXECUTE_1); + { + /* Going out of bounds. Avoid division by zero as we + * know dp != dpPrev from DIFFERENT_SIGNS, above. + */ + brw_ADD(p, c->reg.t, c->reg.dp, negate(c->reg.dpPrev)); + brw_math_invert(p, c->reg.t, c->reg.t); + brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp); + + /* If (vtxOut == 0) vtxOut = vtx + */ + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) ); + brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtx) ); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_clip_interp_vertex(c, vtxOut, vtx, vtxPrev, c->reg.t, TRUE); + + /* *outlist_ptr++ = vtxOut; + * nr_verts++; + * vtxOut = 0; + */ + brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut)); + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); + brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); + brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) ); + } + brw_ENDIF(p, next_test); + } + brw_ENDIF(p, prev_test); + + /* vtxPrev = vtx; + * inlist_ptr++; + */ + brw_MOV(p, get_addr_reg(vtxPrev), get_addr_reg(vtx)); + brw_ADD(p, get_addr_reg(inlist_ptr), get_addr_reg(inlist_ptr), brw_imm_uw(sizeof(short))); + + /* while (--loopcount != 0) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, vertex_loop); + + /* vtxPrev = *(outlist_ptr-1) OR: outlist[nr_verts-1] + * inlist = outlist + * inlist_ptr = &inlist[0] + * outlist_ptr = &outlist[0] + */ + brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_w(-2)); + brw_MOV(p, get_addr_reg(vtxPrev), deref_1uw(outlist_ptr, 0)); + brw_MOV(p, brw_vec8_grf(c->reg.inlist.nr, 0), brw_vec8_grf(c->reg.outlist.nr, 0)); + brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist)); + brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist)); + } + brw_ENDIF(p, plane_active); + + /* plane_ptr++; + */ + brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c)); + + /* nr_verts >= 3 + */ + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_GE, + c->reg.nr_verts, + brw_imm_ud(3)); + + /* && (planemask>>=1) != 0 + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1)); + } + brw_WHILE(p, plane_loop); +#else + #warning "disabled" +#endif +} + + + +void brw_clip_tri_emit_polygon(struct brw_clip_compile *c) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *loop, *if_insn; + + /* for (loopcount = nr_verts-2; loopcount > 0; loopcount--) + */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_G); + brw_ADD(p, + c->reg.loopcount, + c->reg.nr_verts, + brw_imm_d(-2)); + + if_insn = brw_IF(p, BRW_EXECUTE_1); + { + struct brw_indirect v0 = brw_indirect(0, 0); + struct brw_indirect vptr = brw_indirect(1, 0); + + brw_MOV(p, get_addr_reg(vptr), brw_address(c->reg.inlist)); + brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); + + brw_clip_emit_vue(c, v0, 1, 0, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_START)); + + brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2)); + brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_TRIFAN << 2)); + + brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2)); + brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); + + brw_clip_emit_vue(c, v0, 0, 1, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_END)); + } + brw_ENDIF(p, if_insn); +} + +static void do_clip_tri( struct brw_clip_compile *c ) +{ + brw_clip_init_planes(c); + + brw_clip_tri(c); +} + + +static void maybe_do_clip_tri( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *do_clip; + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0)); + do_clip = brw_IF(p, BRW_EXECUTE_1); + { + do_clip_tri(c); + } + brw_ENDIF(p, do_clip); +} + +static void brw_clip_test( struct brw_clip_compile *c ) +{ +#if 0 + struct brw_reg t = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + struct brw_reg t1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + struct brw_reg t2 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + struct brw_reg t3 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + + struct brw_reg v0 = get_tmp(c); + struct brw_reg v1 = get_tmp(c); + struct brw_reg v2 = get_tmp(c); + + struct brw_indirect vt0 = brw_indirect(0, 0); + struct brw_indirect vt1 = brw_indirect(1, 0); + struct brw_indirect vt2 = brw_indirect(2, 0); + + struct brw_compile *p = &c->func; + + brw_MOV(p, get_addr_reg(vt0), brw_address(c->reg.vertex[0])); + brw_MOV(p, get_addr_reg(vt1), brw_address(c->reg.vertex[1])); + brw_MOV(p, get_addr_reg(vt2), brw_address(c->reg.vertex[2])); + brw_MOV(p, v0, deref_4f(vt0, c->offset[VERT_RESULT_HPOS])); + brw_MOV(p, v1, deref_4f(vt1, c->offset[VERT_RESULT_HPOS])); + brw_MOV(p, v2, deref_4f(vt2, c->offset[VERT_RESULT_HPOS])); + + /* test nearz, xmin, ymin plane */ + brw_CMP(p, t1, BRW_CONDITIONAL_LE, negate(v0), get_element(v0, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t2, BRW_CONDITIONAL_LE, negate(v1), get_element(v1, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t3, BRW_CONDITIONAL_LE, negate(v2), get_element(v2, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_XOR(p, t, t1, t2); + brw_XOR(p, t1, t2, t3); + brw_OR(p, t, t, t1); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 0), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<5))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 1), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<3))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 2), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<1))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + /* test farz, xmax, ymax plane */ + brw_CMP(p, t1, BRW_CONDITIONAL_L, v0, get_element(v0, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t2, BRW_CONDITIONAL_L, v1, get_element(v1, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, t3, BRW_CONDITIONAL_L, v2, get_element(v2, 3)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_XOR(p, t, t1, t2); + brw_XOR(p, t1, t2, t3); + brw_OR(p, t, t, t1); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 0), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<4))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 1), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<2))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, + get_element(t, 2), brw_imm_ud(0)); + brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<0))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + release_tmps(c); +#else + #warning "disabled" +#endif +} + + +void brw_emit_tri_clip( struct brw_clip_compile *c ) +{ + struct brw_instruction *neg_rhw; + struct brw_compile *p = &c->func; + brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6); + brw_clip_tri_init_vertices(c); + brw_clip_init_clipmask(c); + + /* if -ve rhw workaround bit is set, + do cliptest */ + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), + brw_imm_ud(1<<20)); + neg_rhw = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_test(c); + } + brw_ENDIF(p, neg_rhw); + + /* Can't push into do_clip_tri because with polygon (or quad) + * flatshading, need to apply the flatshade here because we don't + * respect the PV when converting to trifan for emit: + */ + if (c->key.do_flat_shading) + brw_clip_tri_flat_shade(c); + + if (c->key.clip_mode == BRW_CLIPMODE_NORMAL) + do_clip_tri(c); + else + maybe_do_clip_tri(c); + + brw_clip_tri_emit_polygon(c); + + /* Send an empty message to kill the thread: + */ + brw_clip_kill_thread(c); +} diff --git a/src/gallium/drivers/i965simple/brw_clip_unfilled.c b/src/gallium/drivers/i965simple/brw_clip_unfilled.c new file mode 100644 index 0000000000..b774a76dd6 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_clip_unfilled.c @@ -0,0 +1,477 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + + +/* This is performed against the original triangles, so no indirection + * required: +BZZZT! + */ +static void compute_tri_direction( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg e = c->reg.tmp0; + struct brw_reg f = c->reg.tmp1; + struct brw_reg v0 = byte_offset(c->reg.vertex[0], c->offset[VERT_RESULT_HPOS]); + struct brw_reg v1 = byte_offset(c->reg.vertex[1], c->offset[VERT_RESULT_HPOS]); + struct brw_reg v2 = byte_offset(c->reg.vertex[2], c->offset[VERT_RESULT_HPOS]); + + + /* Calculate the vectors of two edges of the triangle: + */ + brw_ADD(p, e, v0, negate(v2)); + brw_ADD(p, f, v1, negate(v2)); + + /* Take their crossproduct: + */ + brw_set_access_mode(p, BRW_ALIGN_16); + brw_MUL(p, vec4(brw_null_reg()), brw_swizzle(e, 1,2,0,3), brw_swizzle(f,2,0,1,3)); + brw_MAC(p, vec4(e), negate(brw_swizzle(e, 2,0,1,3)), brw_swizzle(f,1,2,0,3)); + brw_set_access_mode(p, BRW_ALIGN_1); + + brw_MUL(p, c->reg.dir, c->reg.dir, vec4(e)); +} + + +static void cull_direction( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *ccw; + unsigned conditional; + + assert (!(c->key.fill_ccw == CLIP_CULL && + c->key.fill_cw == CLIP_CULL)); + + if (c->key.fill_ccw == CLIP_CULL) + conditional = BRW_CONDITIONAL_GE; + else + conditional = BRW_CONDITIONAL_L; + + brw_CMP(p, + vec1(brw_null_reg()), + conditional, + get_element(c->reg.dir, 2), + brw_imm_f(0)); + + ccw = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_kill_thread(c); + } + brw_ENDIF(p, ccw); +} + + + +static void copy_bfc( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *ccw; + unsigned conditional; + + /* Do we have any colors to copy? + */ + if (!(c->offset[VERT_RESULT_COL0] && c->offset[VERT_RESULT_BFC0]) && + !(c->offset[VERT_RESULT_COL1] && c->offset[VERT_RESULT_BFC1])) + return; + + /* In some wierd degnerate cases we can end up testing the + * direction twice, once for culling and once for bfc copying. Oh + * well, that's what you get for setting wierd GL state. + */ + if (c->key.copy_bfc_ccw) + conditional = BRW_CONDITIONAL_GE; + else + conditional = BRW_CONDITIONAL_L; + + brw_CMP(p, + vec1(brw_null_reg()), + conditional, + get_element(c->reg.dir, 2), + brw_imm_f(0)); + + ccw = brw_IF(p, BRW_EXECUTE_1); + { + unsigned i; + + for (i = 0; i < 3; i++) { + if (c->offset[VERT_RESULT_COL0] && c->offset[VERT_RESULT_BFC0]) + brw_MOV(p, + byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_COL0]), + byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_BFC0])); + + if (c->offset[VERT_RESULT_COL1] && c->offset[VERT_RESULT_BFC1]) + brw_MOV(p, + byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_COL1]), + byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_BFC1])); + } + } + brw_ENDIF(p, ccw); +} + + + + +/* + float iz = 1.0 / dir.z; + float ac = dir.x * iz; + float bc = dir.y * iz; + offset = ctx->Polygon.OffsetUnits * DEPTH_SCALE; + offset += MAX2( abs(ac), abs(bc) ) * ctx->Polygon.OffsetFactor; + offset *= MRD; +*/ +static void compute_offset( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg off = c->reg.offset; + struct brw_reg dir = c->reg.dir; + + brw_math_invert(p, get_element(off, 2), get_element(dir, 2)); + brw_MUL(p, vec2(off), dir, get_element(off, 2)); + + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_GE, + brw_abs(get_element(off, 0)), + brw_abs(get_element(off, 1))); + + brw_SEL(p, vec1(off), brw_abs(get_element(off, 0)), brw_abs(get_element(off, 1))); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_MUL(p, vec1(off), off, brw_imm_f(c->key.offset_factor)); + brw_ADD(p, vec1(off), off, brw_imm_f(c->key.offset_units)); +} + + +static void merge_edgeflags( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *is_poly; + struct brw_reg tmp0 = get_element_ud(c->reg.tmp0, 0); + + brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_EQ, + tmp0, + brw_imm_ud(_3DPRIM_POLYGON)); + + /* Get away with using reg.vertex because we know that this is not + * a _3DPRIM_TRISTRIP_REVERSE: + */ + is_poly = brw_IF(p, BRW_EXECUTE_1); + { + brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ); + brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<8)); + brw_MOV(p, byte_offset(c->reg.vertex[0], c->offset[VERT_RESULT_EDGE]), brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ); + brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<9)); + brw_MOV(p, byte_offset(c->reg.vertex[2], c->offset[VERT_RESULT_EDGE]), brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + brw_ENDIF(p, is_poly); +} + + + +static void apply_one_offset( struct brw_clip_compile *c, + struct brw_indirect vert ) +{ + struct brw_compile *p = &c->func; + struct brw_reg pos = deref_4f(vert, c->offset[VERT_RESULT_HPOS]); + struct brw_reg z = get_element(pos, 2); + + brw_ADD(p, z, z, vec1(c->reg.offset)); +} + + + +/*********************************************************************** + * Output clipped polygon as an unfilled primitive: + */ +static void emit_lines(struct brw_clip_compile *c, + boolean do_offset) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *loop; + struct brw_instruction *draw_edge; + struct brw_indirect v0 = brw_indirect(0, 0); + struct brw_indirect v1 = brw_indirect(1, 0); + struct brw_indirect v0ptr = brw_indirect(2, 0); + struct brw_indirect v1ptr = brw_indirect(3, 0); + + /* Need a seperate loop for offset: + */ + if (do_offset) { + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); + brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); + + apply_one_offset(c, v0); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_G); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); + } + + /* v1ptr = &inlist[nr_verts] + * *v1ptr = v0 + */ + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); + brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v0ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW)); + brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v1ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW)); + brw_MOV(p, deref_1uw(v1ptr, 0), deref_1uw(v0ptr, 0)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); + brw_MOV(p, get_addr_reg(v1), deref_1uw(v0ptr, 2)); + brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); + + /* draw edge if edgeflag != 0 */ + brw_CMP(p, + vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, + deref_1f(v0, c->offset[VERT_RESULT_EDGE]), + brw_imm_f(0)); + draw_edge = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START); + brw_clip_emit_vue(c, v1, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END); + } + brw_ENDIF(p, draw_edge); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); +} + + + +static void emit_points(struct brw_clip_compile *c, + boolean do_offset ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *loop; + struct brw_instruction *draw_point; + + struct brw_indirect v0 = brw_indirect(0, 0); + struct brw_indirect v0ptr = brw_indirect(2, 0); + + brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); + brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); + + loop = brw_DO(p, BRW_EXECUTE_1); + { + brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); + brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); + + /* draw if edgeflag != 0 + */ + brw_CMP(p, + vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, + deref_1f(v0, c->offset[VERT_RESULT_EDGE]), + brw_imm_f(0)); + draw_point = brw_IF(p, BRW_EXECUTE_1); + { + if (do_offset) + apply_one_offset(c, v0); + + brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END); + } + brw_ENDIF(p, draw_point); + + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); + } + brw_WHILE(p, loop); +} + + + + + + + +static void emit_primitives( struct brw_clip_compile *c, + unsigned mode, + boolean do_offset ) +{ + switch (mode) { + case CLIP_FILL: + brw_clip_tri_emit_polygon(c); + break; + + case CLIP_LINE: + emit_lines(c, do_offset); + break; + + case CLIP_POINT: + emit_points(c, do_offset); + break; + + case CLIP_CULL: + assert(0); + break; + } +} + + + +static void emit_unfilled_primitives( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *ccw; + + /* Direction culling has already been done. + */ + if (c->key.fill_ccw != c->key.fill_cw && + c->key.fill_ccw != CLIP_CULL && + c->key.fill_cw != CLIP_CULL) + { + brw_CMP(p, + vec1(brw_null_reg()), + BRW_CONDITIONAL_GE, + get_element(c->reg.dir, 2), + brw_imm_f(0)); + + ccw = brw_IF(p, BRW_EXECUTE_1); + { + emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw); + } + ccw = brw_ELSE(p, ccw); + { + emit_primitives(c, c->key.fill_cw, c->key.offset_cw); + } + brw_ENDIF(p, ccw); + } + else if (c->key.fill_cw != CLIP_CULL) { + emit_primitives(c, c->key.fill_cw, c->key.offset_cw); + } + else if (c->key.fill_ccw != CLIP_CULL) { + emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw); + } +} + + + + +static void check_nr_verts( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *if_insn; + + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.nr_verts, brw_imm_d(3)); + if_insn = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_kill_thread(c); + } + brw_ENDIF(p, if_insn); +} + + +void brw_emit_unfilled_clip( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *do_clip; + + + c->need_direction = ((c->key.offset_ccw || c->key.offset_cw) || + (c->key.fill_ccw != c->key.fill_cw) || + c->key.fill_ccw == CLIP_CULL || + c->key.fill_cw == CLIP_CULL || + c->key.copy_bfc_cw || + c->key.copy_bfc_ccw); + + brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6); + brw_clip_tri_init_vertices(c); + + assert(c->offset[VERT_RESULT_EDGE]); + + if (c->key.fill_ccw == CLIP_CULL && + c->key.fill_cw == CLIP_CULL) { + brw_clip_kill_thread(c); + return; + } + + merge_edgeflags(c); + + /* Need to use the inlist indirection here: + */ + if (c->need_direction) + compute_tri_direction(c); + + if (c->key.fill_ccw == CLIP_CULL || + c->key.fill_cw == CLIP_CULL) + cull_direction(c); + + if (c->key.offset_ccw || + c->key.offset_cw) + compute_offset(c); + + if (c->key.copy_bfc_ccw || + c->key.copy_bfc_cw) + copy_bfc(c); + + /* Need to do this whether we clip or not: + */ + if (c->key.do_flat_shading) + brw_clip_tri_flat_shade(c); + + brw_clip_init_clipmask(c); + brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0)); + do_clip = brw_IF(p, BRW_EXECUTE_1); + { + brw_clip_init_planes(c); + brw_clip_tri(c); + check_nr_verts(c); + } + brw_ENDIF(p, do_clip); + + emit_unfilled_primitives(c); + brw_clip_kill_thread(c); +} + + + diff --git a/src/gallium/drivers/i965simple/brw_clip_util.c b/src/gallium/drivers/i965simple/brw_clip_util.c new file mode 100644 index 0000000000..6d58ceafff --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_clip_util.c @@ -0,0 +1,351 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_clip.h" + + + + + +static struct brw_reg get_tmp( struct brw_clip_compile *c ) +{ + struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0); + + if (++c->last_tmp > c->prog_data.total_grf) + c->prog_data.total_grf = c->last_tmp; + + return tmp; +} + +static void release_tmp( struct brw_clip_compile *c, struct brw_reg tmp ) +{ + if (tmp.nr == c->last_tmp-1) + c->last_tmp--; +} + + +static struct brw_reg make_plane_ud(unsigned x, unsigned y, unsigned z, unsigned w) +{ + return brw_imm_ud((w<<24) | (z<<16) | (y<<8) | x); +} + + +void brw_clip_init_planes( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + + if (!c->key.nr_userclip) { + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 0), make_plane_ud( 0, 0, 0xff, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 1), make_plane_ud( 0, 0, 1, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 2), make_plane_ud( 0, 0xff, 0, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 3), make_plane_ud( 0, 1, 0, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 4), make_plane_ud(0xff, 0, 0, 1)); + brw_MOV(p, get_element_ud(c->reg.fixed_planes, 5), make_plane_ud( 1, 0, 0, 1)); + } +} + + + +#define W 3 + +/* Project 'pos' to screen space (or back again), overwrite with results: + */ +static void brw_clip_project_position(struct brw_clip_compile *c, struct brw_reg pos ) +{ + struct brw_compile *p = &c->func; + + /* calc rhw + */ + brw_math_invert(p, get_element(pos, W), get_element(pos, W)); + + /* value.xyz *= value.rhw + */ + brw_set_access_mode(p, BRW_ALIGN_16); + brw_MUL(p, brw_writemask(pos, TGSI_WRITEMASK_XYZ), pos, brw_swizzle1(pos, W)); + brw_set_access_mode(p, BRW_ALIGN_1); +} + + +static void brw_clip_project_vertex( struct brw_clip_compile *c, + struct brw_indirect vert_addr ) +{ +#if 0 + struct brw_compile *p = &c->func; + struct brw_reg tmp = get_tmp(c); + + /* Fixup position. Extract from the original vertex and re-project + * to screen space: + */ + brw_MOV(p, tmp, deref_4f(vert_addr, c->offset[VERT_RESULT_HPOS])); + brw_clip_project_position(c, tmp); + brw_MOV(p, deref_4f(vert_addr, c->header_position_offset), tmp); + + release_tmp(c, tmp); +#else + #warning "disabled" +#endif +} + + + + +/* Interpolate between two vertices and put the result into a0.0. + * Increment a0.0 accordingly. + */ +void brw_clip_interp_vertex( struct brw_clip_compile *c, + struct brw_indirect dest_ptr, + struct brw_indirect v0_ptr, /* from */ + struct brw_indirect v1_ptr, /* to */ + struct brw_reg t0, + boolean force_edgeflag) +{ +#if 0 + struct brw_compile *p = &c->func; + struct brw_reg tmp = get_tmp(c); + unsigned i; + + /* Just copy the vertex header: + */ + brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1); + + /* Iterate over each attribute (could be done in pairs?) + */ + for (i = 0; i < c->nr_attrs; i++) { + unsigned delta = i*16 + 32; + + if (delta == c->offset[VERT_RESULT_EDGE]) { + if (force_edgeflag) + brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(1)); + else + brw_MOV(p, deref_4f(dest_ptr, delta), deref_4f(v0_ptr, delta)); + } + else { + /* Interpolate: + * + * New = attr0 + t*attr1 - t*attr0 + */ + brw_MUL(p, + vec4(brw_null_reg()), + deref_4f(v1_ptr, delta), + t0); + + brw_MAC(p, + tmp, + negate(deref_4f(v0_ptr, delta)), + t0); + + brw_ADD(p, + deref_4f(dest_ptr, delta), + deref_4f(v0_ptr, delta), + tmp); + } + } + + if (i & 1) { + unsigned delta = i*16 + 32; + brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0)); + } + + release_tmp(c, tmp); + + /* Recreate the projected (NDC) coordinate in the new vertex + * header: + */ + brw_clip_project_vertex(c, dest_ptr ); +#else + #warning "disabled" +#endif +} + + + + +#define MAX_MRF 16 + +void brw_clip_emit_vue(struct brw_clip_compile *c, + struct brw_indirect vert, + boolean allocate, + boolean eot, + unsigned header) +{ + struct brw_compile *p = &c->func; + unsigned start = c->last_mrf; + + assert(!(allocate && eot)); + + /* Cycle through mrf regs - probably futile as we have to wait for + * the allocation response anyway. Also, the order this function + * is invoked doesn't correspond to the order the instructions will + * be executed, so it won't have any effect in many cases. + */ +#if 0 + if (start + c->nr_regs + 1 >= MAX_MRF) + start = 0; + + c->last_mrf = start + c->nr_regs + 1; +#endif + + /* Copy the vertex from vertn into m1..mN+1: + */ + brw_copy_from_indirect(p, brw_message_reg(start+1), vert, c->nr_regs); + + /* Overwrite PrimType and PrimStart in the message header, for + * each vertex in turn: + */ + brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header)); + + + /* Send each vertex as a seperate write to the urb. This + * is different to the concept in brw_sf_emit.c, where + * subsequent writes are used to build up a single urb + * entry. Each of these writes instantiates a seperate + * urb entry - (I think... what about 'allocate'?) + */ + brw_urb_WRITE(p, + allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), + start, + c->reg.R0, + allocate, + 1, /* used */ + c->nr_regs + 1, /* msg length */ + allocate ? 1 : 0, /* response_length */ + eot, /* eot */ + 1, /* writes_complete */ + 0, /* urb offset */ + BRW_URB_SWIZZLE_NONE); +} + + + +void brw_clip_kill_thread(struct brw_clip_compile *c) +{ + struct brw_compile *p = &c->func; + + /* Send an empty message to kill the thread and release any + * allocated urb entry: + */ + brw_urb_WRITE(p, + retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), + 0, + c->reg.R0, + 0, /* allocate */ + 0, /* used */ + 0, /* msg len */ + 0, /* response len */ + 1, /* eot */ + 1, /* writes complete */ + 0, + BRW_URB_SWIZZLE_NONE); +} + + + + +struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c ) +{ + return brw_address(c->reg.fixed_planes); +} + + +struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c ) +{ + if (c->key.nr_userclip) { + return brw_imm_uw(16); + } + else { + return brw_imm_uw(4); + } +} + + +/* If flatshading, distribute color from provoking vertex prior to + * clipping. + */ +void brw_clip_copy_colors( struct brw_clip_compile *c, + unsigned to, unsigned from ) +{ +#if 0 + struct brw_compile *p = &c->func; + + if (c->offset[VERT_RESULT_COL0]) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL0]), + byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL0])); + + if (c->offset[VERT_RESULT_COL1]) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL1]), + byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL1])); + + if (c->offset[VERT_RESULT_BFC0]) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC0]), + byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC0])); + + if (c->offset[VERT_RESULT_BFC1]) + brw_MOV(p, + byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC1]), + byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC1])); +#else + #warning "disabled" +#endif +} + + + +void brw_clip_init_clipmask( struct brw_clip_compile *c ) +{ + struct brw_compile *p = &c->func; + struct brw_reg incoming = get_element_ud(c->reg.R0, 2); + + /* Shift so that lowest outcode bit is rightmost: + */ + brw_SHR(p, c->reg.planemask, incoming, brw_imm_ud(26)); + + if (c->key.nr_userclip) { + struct brw_reg tmp = retype(vec1(get_tmp(c)), BRW_REGISTER_TYPE_UD); + + /* Rearrange userclip outcodes so that they come directly after + * the fixed plane bits. + */ + brw_AND(p, tmp, incoming, brw_imm_ud(0x3f<<14)); + brw_SHR(p, tmp, tmp, brw_imm_ud(8)); + brw_OR(p, c->reg.planemask, c->reg.planemask, tmp); + + release_tmp(c, tmp); + } +} + diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c new file mode 100644 index 0000000000..5e58701e91 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -0,0 +1,245 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_draw.h" +#include "brw_vs.h" +#include "brw_tex_layout.h" +#include "brw_winsys.h" + +#include "pipe/p_winsys.h" +#include "pipe/p_context.h" +#include "pipe/p_util.h" + +/*************************************** + * Mesa's Driver Functions + ***************************************/ + +#ifndef BRW_DEBUG +int BRW_DEBUG = (0); +#endif + +static void brw_destroy(struct pipe_context *pipe) +{ + struct brw_context *brw = brw_context(pipe); + + FREE(brw); +} + +static void brw_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + int x, y, w, h; + /* FIXME: corny... */ + + x = 0; + y = 0; + w = ps->width; + h = ps->height; + + pipe->surface_fill(pipe, ps, x, y, w, h, clearValue); +} + + +static int +brw_get_param(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 11; /* max 1024x1024 */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 11; /* max 1024x1024 */ + default: + return 0; + } +} + + +static float +brw_get_paramf(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 7.5; + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 4.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; + + default: + return 0; + } +} + +static boolean +brw_is_format_supported( struct pipe_context *pipe, + enum pipe_format format, uint type ) +{ +#if 0 + /* XXX: This is broken -- rewrite if still needed. */ + static const unsigned tex_supported[] = { + PIPE_FORMAT_U_R8_G8_B8_A8, + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_U_L8, + PIPE_FORMAT_U_A8, + PIPE_FORMAT_U_I8, + PIPE_FORMAT_U_L8_A8, + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + PIPE_FORMAT_S8_Z24, + }; + + + /* Actually a lot more than this - add later: + */ + static const unsigned render_supported[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + }; + + /* + */ + static const unsigned z_stencil_supported[] = { + PIPE_FORMAT_U_Z16, + PIPE_FORMAT_U_Z32, + PIPE_FORMAT_S8_Z24, + }; + + switch (type) { + case PIPE_RENDER_FORMAT: + *numFormats = Elements(render_supported); + return render_supported; + + case PIPE_TEX_FORMAT: + *numFormats = Elements(tex_supported); + return render_supported; + + case PIPE_Z_STENCIL_FORMAT: + *numFormats = Elements(render_supported); + return render_supported; + + default: + *numFormats = 0; + return NULL; + } +#else + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + return TRUE; + default: + return FALSE; + }; + return FALSE; +#endif +} + + + + +struct pipe_context *brw_create(struct pipe_winsys *pipe_winsys, + struct brw_winsys *brw_winsys, + unsigned pci_id) +{ + struct brw_context *brw; + + pipe_winsys->printf(pipe_winsys, + "%s: creating brw_context with pci id 0x%x\n", + __FUNCTION__, pci_id); + + brw = CALLOC_STRUCT(brw_context); + if (brw == NULL) + return NULL; + + brw->winsys = brw_winsys; + brw->pipe.winsys = pipe_winsys; + + brw->pipe.destroy = brw_destroy; + brw->pipe.is_format_supported = brw_is_format_supported; + brw->pipe.get_param = brw_get_param; + brw->pipe.get_paramf = brw_get_paramf; + brw->pipe.clear = brw_clear; + brw->pipe.texture_create = brw_texture_create; + brw->pipe.texture_release = brw_texture_release; + + brw_init_surface_functions(brw); + brw_init_state_functions(brw); + brw_init_flush_functions(brw); + brw_init_string_functions(brw); + brw_init_draw_functions( brw ); + + + brw_init_state( brw ); + + brw->pci_id = pci_id; + brw->dirty = ~0; + brw->hardware_dirty = ~0; + + memset(&brw->wm.bind, ~0, sizeof(brw->wm.bind)); + + return &brw->pipe; +} + diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h new file mode 100644 index 0000000000..65664d853d --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -0,0 +1,690 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRWCONTEXT_INC +#define BRWCONTEXT_INC + + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "brw_structs.h" +#include "brw_winsys.h" + + +/* Glossary: + * + * URB - uniform resource buffer. A mid-sized buffer which is + * partitioned between the fixed function units and used for passing + * values (vertices, primitives, constants) between them. + * + * CURBE - constant URB entry. An urb region (entry) used to hold + * constant values which the fixed function units can be instructed to + * preload into the GRF when spawining a thread. + * + * VUE - vertex URB entry. An urb entry holding a vertex and usually + * a vertex header. The header contains control information and + * things like primitive type, Begin/end flags and clip codes. + * + * PUE - primitive URB entry. An urb entry produced by the setup (SF) + * unit holding rasterization and interpolation parameters. + * + * GRF - general register file. One of several register files + * addressable by programmed threads. The inputs (r0, payload, curbe, + * urb) of the thread are preloaded to this area before the thread is + * spawned. The registers are individually 8 dwords wide and suitable + * for general usage. Registers holding thread input values are not + * special and may be overwritten. + * + * MRF - message register file. Threads communicate (and terminate) + * by sending messages. Message parameters are placed in contigous + * MRF registers. All program output is via these messages. URB + * entries are populated by sending a message to the shared URB + * function containing the new data, together with a control word, + * often an unmodified copy of R0. + * + * R0 - GRF register 0. Typically holds control information used when + * sending messages to other threads. + * + * EU or GEN4 EU: The name of the programmable subsystem of the + * i965 hardware. Threads are executed by the EU, the registers + * described above are part of the EU architecture. + * + * Fixed function units: + * + * CS - Command streamer. Notional first unit, little software + * interaction. Holds the URB entries used for constant data, ie the + * CURBEs. + * + * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of + * this unit is responsible for pulling vertices out of vertex buffers + * in vram and injecting them into the processing pipe as VUEs. If + * enabled, it first passes them to a VS thread which is a good place + * for the driver to implement any active vertex shader. + * + * GS - Geometry Shader. This corresponds to a new DX10 concept. If + * enabled, incoming strips etc are passed to GS threads in individual + * line/triangle/point units. The GS thread may perform arbitary + * computation and emit whatever primtives with whatever vertices it + * chooses. This makes GS an excellent place to implement GL's + * unfilled polygon modes, though of course it is capable of much + * more. Additionally, GS is used to translate away primitives not + * handled by latter units, including Quads and Lineloops. + * + * CS - Clipper. Mesa's clipping algorithms are imported to run on + * this unit. The fixed function part performs cliptesting against + * the 6 fixed clipplanes and makes descisions on whether or not the + * incoming primitive needs to be passed to a thread for clipping. + * User clip planes are handled via cooperation with the VS thread. + * + * SF - Strips Fans or Setup: Triangles are prepared for + * rasterization. Interpolation coefficients are calculated. + * Flatshading and two-side lighting usually performed here. + * + * WM - Windower. Interpolation of vertex attributes performed here. + * Fragment shader implemented here. SIMD aspects of EU taken full + * advantage of, as pixels are processed in blocks of 16. + * + * CC - Color Calculator. No EU threads associated with this unit. + * Handles blending and (presumably) depth and stencil testing. + */ + +#define BRW_MAX_CURBE (32*16) + +struct brw_context; +struct brw_winsys; + + +/* Raised when we receive new state across the pipe interface: + */ +#define BRW_NEW_VIEWPORT 0x1 +#define BRW_NEW_RASTERIZER 0x2 +#define BRW_NEW_FS 0x4 +#define BRW_NEW_BLEND 0x8 +#define BRW_NEW_CLIP 0x10 +#define BRW_NEW_SCISSOR 0x20 +#define BRW_NEW_STIPPLE 0x40 +#define BRW_NEW_FRAMEBUFFER 0x80 +#define BRW_NEW_ALPHA_TEST 0x100 +#define BRW_NEW_DEPTH_STENCIL 0x200 +#define BRW_NEW_SAMPLER 0x400 +#define BRW_NEW_TEXTURE 0x800 +#define BRW_NEW_CONSTANTS 0x1000 +#define BRW_NEW_VBO 0x2000 +#define BRW_NEW_VS 0x4000 + +/* Raised for other internal events: + */ +#define BRW_NEW_URB_FENCE 0x10000 +#define BRW_NEW_PSP 0x20000 +#define BRW_NEW_CURBE_OFFSETS 0x40000 +#define BRW_NEW_REDUCED_PRIMITIVE 0x80000 +#define BRW_NEW_PRIMITIVE 0x100000 +#define BRW_NEW_SCENE 0x200000 +#define BRW_NEW_SF_LINKAGE 0x400000 + +extern int BRW_DEBUG; + +#define DEBUG_TEXTURE 0x1 +#define DEBUG_STATE 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_PRIMS 0x8 +#define DEBUG_VERTS 0x10 +#define DEBUG_FALLBACKS 0x20 +#define DEBUG_VERBOSE 0x40 +#define DEBUG_DRI 0x80 +#define DEBUG_DMA 0x100 +#define DEBUG_SANITY 0x200 +#define DEBUG_SYNC 0x400 +#define DEBUG_SLEEP 0x800 +#define DEBUG_PIXEL 0x1000 +#define DEBUG_STATS 0x2000 +#define DEBUG_TILE 0x4000 +#define DEBUG_SINGLE_THREAD 0x8000 +#define DEBUG_WM 0x10000 +#define DEBUG_URB 0x20000 +#define DEBUG_VS 0x40000 +#define DEBUG_BATCH 0x80000 +#define DEBUG_BUFMGR 0x100000 +#define DEBUG_BLIT 0x200000 +#define DEBUG_REGION 0x400000 +#define DEBUG_MIPTREE 0x800000 + +#define DBG(...) do { \ + if (BRW_DEBUG & FILE_DEBUG_FLAG) \ + brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ +} while(0) + +#define PRINT(...) do { \ + brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ +} while(0) + +struct brw_state_flags { + unsigned cache; + unsigned brw; +}; + + +struct brw_shader_info { + int nr_regs[8]; /* TGSI_FILE_* */ +}; + + + +struct brw_vertex_program { + struct pipe_shader_state program; + struct brw_shader_info info; + int id; +}; + + + +struct brw_fragment_program { + struct pipe_shader_state program; + struct brw_shader_info info; + + boolean UsesDepth; + boolean UsesKill; + boolean ComputesDepth; + int id; +}; + + + + +struct pipe_setup_linkage { + struct { + unsigned vp_output:5; + unsigned interp_mode:4; + unsigned bf_vp_output:5; + } fp_input[PIPE_MAX_SHADER_INPUTS]; + + unsigned fp_input_count:5; + unsigned max_vp_output:5; +}; + + + +struct brw_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; +}; + +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ + +struct brw_wm_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + + unsigned first_curbe_grf; + unsigned total_grf; + unsigned total_scratch; + + /* Internally generated constants for the CURBE. These are loaded + * ahead of the data from the constant buffer. + */ + const float internal_const[8]; + unsigned nr_internal_consts; + unsigned max_const; + + boolean error; +}; + +struct brw_sf_prog_data { + unsigned urb_read_length; + unsigned total_grf; + + /* Each vertex may have upto 12 attributes, 4 components each, + * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 + * rows. + * + * Actually we use 4 for each, so call it 12 rows. + */ + unsigned urb_entry_size; +}; + +struct brw_clip_prog_data { + unsigned curb_read_length; /* user planes? */ + unsigned clip_mode; + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_gs_prog_data { + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_vs_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + unsigned total_grf; + unsigned outputs_written; + + unsigned inputs_read; + + unsigned max_const; + + float imm_buf[PIPE_MAX_CONSTANT][4]; + unsigned num_imm; + unsigned num_consts; + + /* Used for calculating urb partitions: + */ + unsigned urb_entry_size; +}; + + +#define BRW_MAX_TEX_UNIT 8 +#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 + +/* Create a fixed sized struct for caching binding tables: + */ +struct brw_surface_binding_table { + unsigned surf_ss_offset[BRW_WM_MAX_SURF]; +}; + + +struct brw_cache; + +struct brw_mem_pool { + struct pipe_buffer *buffer; + + unsigned size; + unsigned offset; /* offset of first free byte */ + + struct brw_context *brw; +}; + +struct brw_cache_item { + unsigned hash; + unsigned key_size; /* for variable-sized keys */ + const void *key; + + unsigned offset; /* offset within pool's buffer */ + unsigned data_size; + + struct brw_cache_item *next; +}; + + + +struct brw_cache { + unsigned id; + + const char *name; + + struct brw_context *brw; + struct brw_mem_pool *pool; + + struct brw_cache_item **items; + unsigned size, n_items; + + unsigned key_size; /* for fixed-size keys */ + unsigned aux_size; + + unsigned last_addr; /* offset of active item */ +}; + + + + +/* Considered adding a member to this struct to document which flags + * an update might raise so that ordering of the state atoms can be + * checked or derived at runtime. Dropped the idea in favor of having + * a debug mode where the state is monitored for flags which are + * raised that have already been tested against. + */ +struct brw_tracked_state { + struct brw_state_flags dirty; + void (*update)( struct brw_context *brw ); +}; + + +/* Flags for brw->state.cache. + */ +#define CACHE_NEW_CC_VP (1< 32. Wouldn't life + * be easier if C allowed arrays of packed elements? + */ +#define ATTRIB_BIT_DWORDS ((PIPE_ATTRIB_MAX+31)/32) + + + + +struct brw_vertex_info { + unsigned varying; /* varying:1[PIPE_ATTRIB_MAX] */ + unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_ATTRIB_MAX] */ +}; + + + + + +struct brw_context +{ + struct pipe_context pipe; + struct brw_winsys *winsys; + + unsigned primitive; + unsigned reduced_primitive; + + boolean emit_state_always; + + struct { + struct brw_state_flags dirty; + } state; + + + struct { + const struct pipe_blend_state *Blend; + const struct pipe_depth_stencil_alpha_state *DepthStencil; + const struct pipe_poly_stipple *PolygonStipple; + const struct pipe_rasterizer_state *Raster; + const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; + const struct brw_vertex_program *VertexProgram; + const struct brw_fragment_program *FragmentProgram; + + struct pipe_clip_state Clip; + struct pipe_blend_color BlendColor; + struct pipe_scissor_state Scissor; + struct pipe_viewport_state Viewport; + struct pipe_framebuffer_state FrameBuffer; + + const struct pipe_constant_buffer *Constants[2]; + const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; + } attribs; + + struct brw_mem_pool pool[BRW_MAX_POOL]; + struct brw_cache cache[BRW_MAX_CACHE]; + struct brw_cached_batch_item *cached_batch_items; + + struct { + + /* Arrays with buffer objects to copy non-bufferobj arrays into + * for upload: + */ + struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX]; + + struct brw_vertex_element_state inputs[PIPE_ATTRIB_MAX]; + +#define BRW_NR_UPLOAD_BUFS 17 +#define BRW_UPLOAD_INIT_SIZE (128*1024) + + /* Summary of size and varying of active arrays, so we can check + * for changes to this state: + */ + struct brw_vertex_info info; + } vb; + + + unsigned hardware_dirty; + unsigned dirty; + unsigned pci_id; + /* BRW_NEW_URB_ALLOCATIONS: + */ + struct { + unsigned vsize; /* vertex size plus header in urb registers */ + unsigned csize; /* constant buffer size in urb registers */ + unsigned sfsize; /* setup data size in urb registers */ + + boolean constrained; + + unsigned nr_vs_entries; + unsigned nr_gs_entries; + unsigned nr_clip_entries; + unsigned nr_sf_entries; + unsigned nr_cs_entries; + +/* unsigned vs_size; */ +/* unsigned gs_size; */ +/* unsigned clip_size; */ +/* unsigned sf_size; */ +/* unsigned cs_size; */ + + unsigned vs_start; + unsigned gs_start; + unsigned clip_start; + unsigned sf_start; + unsigned cs_start; + } urb; + + + /* BRW_NEW_CURBE_OFFSETS: + */ + struct { + unsigned wm_start; + unsigned wm_size; + unsigned clip_start; + unsigned clip_size; + unsigned vs_start; + unsigned vs_size; + unsigned total_size; + + unsigned gs_offset; + + float *last_buf; + unsigned last_bufsz; + } curbe; + + struct { + struct brw_vs_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } vs; + + struct { + struct brw_gs_prog_data *prog_data; + + boolean prog_active; + unsigned prog_gs_offset; + unsigned state_gs_offset; + } gs; + + struct { + struct brw_clip_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } clip; + + + struct { + struct brw_sf_prog_data *prog_data; + + struct pipe_setup_linkage linkage; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } sf; + + struct { + struct brw_wm_prog_data *prog_data; + +// struct brw_wm_compiler *compile_data; + + + /** + * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER + * cache + */ + struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; + + unsigned render_surf; + unsigned nr_surfaces; + + unsigned max_threads; + struct pipe_buffer *scratch_buffer; + unsigned scratch_buffer_size; + + unsigned sampler_count; + unsigned sampler_gs_offset; + + struct brw_surface_binding_table bind; + unsigned bind_ss_offset; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } wm; + + + struct { + unsigned vp_gs_offset; + unsigned state_gs_offset; + } cc; + + + /* Used to give every program string a unique id + */ + unsigned program_id; +}; + + +#define BRW_PACKCOLOR8888(r,g,b,a) ((r<<24) | (g<<16) | (b<<8) | a) + + +/*====================================================================== + * brw_vtbl.c + */ +void brw_do_flush( struct brw_context *brw, + unsigned flags ); + + +/*====================================================================== + * brw_state.c + */ +void brw_validate_state(struct brw_context *brw); +void brw_init_state(struct brw_context *brw); +void brw_destroy_state(struct brw_context *brw); + + +/*====================================================================== + * brw_tex.c + */ +void brwUpdateTextureState( struct brw_context *brw ); + + +/* brw_urb.c + */ +void brw_upload_urb_fence(struct brw_context *brw); + +void brw_upload_constant_buffer_state(struct brw_context *brw); + +void brw_init_surface_functions(struct brw_context *brw); +void brw_init_state_functions(struct brw_context *brw); +void brw_init_flush_functions(struct brw_context *brw); +void brw_init_string_functions(struct brw_context *brw); + +/*====================================================================== + * Inline conversion functions. These are better-typed than the + * macros used previously: + */ +static inline struct brw_context * +brw_context( struct pipe_context *ctx ) +{ + return (struct brw_context *)ctx; +} + +#endif + diff --git a/src/gallium/drivers/i965simple/brw_curbe.c b/src/gallium/drivers/i965simple/brw_curbe.c new file mode 100644 index 0000000000..52bbd525c1 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_curbe.c @@ -0,0 +1,368 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_state.h" +#include "brw_batch.h" +#include "brw_util.h" +#include "brw_wm.h" +#include "pipe/p_state.h" +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" + +#define FILE_DEBUG_FLAG DEBUG_FALLBACKS + +/* Partition the CURBE between the various users of constant values: + */ +static void calculate_curbe_offsets( struct brw_context *brw ) +{ + /* CACHE_NEW_WM_PROG */ + unsigned nr_fp_regs = align(brw->wm.prog_data->max_const, 16); + + /* BRW_NEW_VERTEX_PROGRAM */ + unsigned nr_vp_regs = align(brw->vs.prog_data->max_const, 16); + unsigned nr_clip_regs = 0; + unsigned total_regs; + +#if 0 + /* BRW_NEW_CLIP ? */ + if (brw->attribs.Transform->ClipPlanesEnabled) { + unsigned nr_planes = 6 + brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled); + nr_clip_regs = align(nr_planes * 4, 16); + } +#endif + + + total_regs = nr_fp_regs + nr_vp_regs + nr_clip_regs; + + /* This can happen - what to do? Probably rather than falling + * back, the best thing to do is emit programs which code the + * constants as immediate values. Could do this either as a static + * cap on WM and VS, or adaptively. + * + * Unfortunately, this is currently dependent on the results of the + * program generation process (in the case of wm), so this would + * introduce the need to re-generate programs in the event of a + * curbe allocation failure. + */ + /* Max size is 32 - just large enough to + * hold the 128 parameters allowed by + * the fragment and vertex program + * api's. It's not clear what happens + * when both VP and FP want to use 128 + * parameters, though. + */ + assert(total_regs <= 32); + + /* Lazy resize: + */ + if (nr_fp_regs > brw->curbe.wm_size || + nr_vp_regs > brw->curbe.vs_size || + nr_clip_regs != brw->curbe.clip_size || + (total_regs < brw->curbe.total_size / 4 && + brw->curbe.total_size > 16)) { + + unsigned reg = 0; + + /* Calculate a new layout: + */ + reg = 0; + brw->curbe.wm_start = reg; + brw->curbe.wm_size = nr_fp_regs; reg += nr_fp_regs; + brw->curbe.clip_start = reg; + brw->curbe.clip_size = nr_clip_regs; reg += nr_clip_regs; + brw->curbe.vs_start = reg; + brw->curbe.vs_size = nr_vp_regs; reg += nr_vp_regs; + brw->curbe.total_size = reg; + +#if 0 + if (0) + DBG("curbe wm %d+%d clip %d+%d vs %d+%d\n", + brw->curbe.wm_start, + brw->curbe.wm_size, + brw->curbe.clip_start, + brw->curbe.clip_size, + brw->curbe.vs_start, + brw->curbe.vs_size ); +#endif + + brw->state.dirty.brw |= BRW_NEW_CURBE_OFFSETS; + } +} + + +const struct brw_tracked_state brw_curbe_offsets = { + .dirty = { + .brw = (BRW_NEW_CLIP | + BRW_NEW_VS), + .cache = CACHE_NEW_WM_PROG + }, + .update = calculate_curbe_offsets +}; + + + +/* Define the number of curbes within CS's urb allocation. Multiple + * urb entries -> multiple curbes. These will be used by + * fixed-function hardware in a double-buffering scheme to avoid a + * pipeline stall each time the contents of the curbe is changed. + */ +void brw_upload_constant_buffer_state(struct brw_context *brw) +{ + struct brw_constant_buffer_state cbs; + memset(&cbs, 0, sizeof(cbs)); + + /* It appears that this is the state packet for the CS unit, ie. the + * urb entries detailed here are housed in the CS range from the + * URB_FENCE command. + */ + cbs.header.opcode = CMD_CONST_BUFFER_STATE; + cbs.header.length = sizeof(cbs)/4 - 2; + + /* BRW_NEW_URB_FENCE */ + cbs.bits0.nr_urb_entries = brw->urb.nr_cs_entries; + cbs.bits0.urb_entry_size = brw->urb.csize - 1; + + assert(brw->urb.nr_cs_entries); + BRW_CACHED_BATCH_STRUCT(brw, &cbs); +} + + +static float fixed_plane[6][4] = { + { 0, 0, -1, 1 }, + { 0, 0, 1, 1 }, + { 0, -1, 0, 1 }, + { 0, 1, 0, 1 }, + {-1, 0, 0, 1 }, + { 1, 0, 0, 1 } +}; + +/* Upload a new set of constants. Too much variability to go into the + * cache mechanism, but maybe would benefit from a comparison against + * the current uploaded set of constants. + */ +static void upload_constant_buffer(struct brw_context *brw) +{ + struct brw_mem_pool *pool = &brw->pool[BRW_GS_POOL]; + unsigned sz = brw->curbe.total_size; + unsigned bufsz = sz * sizeof(float); + float *buf; + unsigned i; + + + if (sz == 0) { + struct brw_constant_buffer cb; + cb.header.opcode = CMD_CONST_BUFFER; + cb.header.length = sizeof(cb)/4 - 2; + cb.header.valid = 0; + cb.bits0.buffer_length = 0; + cb.bits0.buffer_address = 0; + BRW_BATCH_STRUCT(brw, &cb); + + if (brw->curbe.last_buf) { + free(brw->curbe.last_buf); + brw->curbe.last_buf = NULL; + brw->curbe.last_bufsz = 0; + } + + return; + } + + buf = (float *)malloc(bufsz); + + memset(buf, 0, bufsz); + + if (brw->curbe.wm_size) { + unsigned offset = brw->curbe.wm_start * 16; + + /* First the constant buffer constants: + */ + + /* Then any internally generated constants: + */ + for (i = 0; i < brw->wm.prog_data->nr_internal_consts; i++) + buf[offset + i] = brw->wm.prog_data->internal_const[i]; + + assert(brw->wm.prog_data->max_const == + brw->wm.prog_data->nr_internal_consts); + } + + + /* The clipplanes are actually delivered to both CLIP and VS units. + * VS uses them to calculate the outcode bitmasks. + */ + if (brw->curbe.clip_size) { + unsigned offset = brw->curbe.clip_start * 16; + unsigned j; + + /* If any planes are going this way, send them all this way: + */ + for (i = 0; i < 6; i++) { + buf[offset + i * 4 + 0] = fixed_plane[i][0]; + buf[offset + i * 4 + 1] = fixed_plane[i][1]; + buf[offset + i * 4 + 2] = fixed_plane[i][2]; + buf[offset + i * 4 + 3] = fixed_plane[i][3]; + } + + /* Clip planes: BRW_NEW_CLIP: + */ + for (j = 0; j < brw->attribs.Clip.nr; j++) { + buf[offset + i * 4 + 0] = brw->attribs.Clip.ucp[j][0]; + buf[offset + i * 4 + 1] = brw->attribs.Clip.ucp[j][1]; + buf[offset + i * 4 + 2] = brw->attribs.Clip.ucp[j][2]; + buf[offset + i * 4 + 3] = brw->attribs.Clip.ucp[j][3]; + i++; + } + } + + + if (brw->curbe.vs_size) { + unsigned offset = brw->curbe.vs_start * 16; + /*unsigned nr = vp->max_const;*/ + const struct pipe_constant_buffer *cbuffer = brw->attribs.Constants[0]; + struct pipe_winsys *ws = brw->pipe.winsys; + /* FIXME: buffer size is num_consts + num_immediates */ + if (brw->vs.prog_data->num_consts) { + /* map the vertex constant buffer and copy to curbe: */ + void *data = ws->buffer_map(ws, cbuffer->buffer, 0); + /* FIXME: this is wrong. the cbuffer->size currently + * represents size of consts + immediates. so if we'll + * have both we'll copy over the end of the buffer + * with the subsequent memcpy */ + memcpy(&buf[offset], data, cbuffer->size); + ws->buffer_unmap(ws, cbuffer->buffer); + offset += cbuffer->size; + } + /*immediates*/ + if (brw->vs.prog_data->num_imm) { + memcpy(&buf[offset], brw->vs.prog_data->imm_buf, + brw->vs.prog_data->num_imm * 4 * sizeof(float)); + } + } + + if (1) { + for (i = 0; i < sz; i+=4) + debug_printf("curbe %d.%d: %f %f %f %f\n", i/8, i&4, + buf[i+0], buf[i+1], buf[i+2], buf[i+3]); + + debug_printf("last_buf %p buf %p sz %d/%d cmp %d\n", + brw->curbe.last_buf, buf, + bufsz, brw->curbe.last_bufsz, + brw->curbe.last_buf ? memcmp(buf, brw->curbe.last_buf, bufsz) : -1); + } + + if (brw->curbe.last_buf && + bufsz == brw->curbe.last_bufsz && + memcmp(buf, brw->curbe.last_buf, bufsz) == 0) { + free(buf); +/* return; */ + } + else { + if (brw->curbe.last_buf) + free(brw->curbe.last_buf); + brw->curbe.last_buf = buf; + brw->curbe.last_bufsz = bufsz; + + + if (!brw_pool_alloc(pool, + bufsz, + 1 << 6, + &brw->curbe.gs_offset)) { + debug_printf("out of GS memory for curbe\n"); + assert(0); + return; + } + + + /* Copy data to the buffer: + */ + brw->winsys->buffer_subdata_typed(brw->winsys, + pool->buffer, + brw->curbe.gs_offset, + bufsz, + buf, + BRW_CONSTANT_BUFFER ); + } + + /* TODO: only emit the constant_buffer packet when necessary, ie: + - contents have changed + - offset has changed + - hw requirements due to other packets emitted. + */ + { + struct brw_constant_buffer cb; + + memset(&cb, 0, sizeof(cb)); + + cb.header.opcode = CMD_CONST_BUFFER; + cb.header.length = sizeof(cb)/4 - 2; + cb.header.valid = 1; + cb.bits0.buffer_length = sz - 1; + cb.bits0.buffer_address = brw->curbe.gs_offset >> 6; + + /* Because this provokes an action (ie copy the constants into the + * URB), it shouldn't be shortcircuited if identical to the + * previous time - because eg. the urb destination may have + * changed, or the urb contents different to last time. + * + * Note that the data referred to is actually copied internally, + * not just used in place according to passed pointer. + * + * It appears that the CS unit takes care of using each available + * URB entry (Const URB Entry == CURBE) in turn, and issuing + * flushes as necessary when doublebuffering of CURBEs isn't + * possible. + */ + BRW_BATCH_STRUCT(brw, &cb); + } +} + +/* This tracked state is unique in that the state it monitors varies + * dynamically depending on the parameters tracked by the fragment and + * vertex programs. This is the template used as a starting point, + * each context will maintain a copy of this internally and update as + * required. + */ +const struct brw_tracked_state brw_constant_buffer = { + .dirty = { + .brw = (BRW_NEW_CLIP | + BRW_NEW_CONSTANTS | + BRW_NEW_URB_FENCE | /* Implicit - hardware requires this, not used above */ + BRW_NEW_PSP | /* Implicit - hardware requires this, not used above */ + BRW_NEW_CURBE_OFFSETS), + .cache = (CACHE_NEW_WM_PROG) + }, + .update = upload_constant_buffer +}; + diff --git a/src/gallium/drivers/i965simple/brw_defines.h b/src/gallium/drivers/i965simple/brw_defines.h new file mode 100644 index 0000000000..9379a397f6 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_defines.h @@ -0,0 +1,852 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_DEFINES_H +#define BRW_DEFINES_H + +/* + */ +#define MI_NOOP 0x00 +#define MI_USER_INTERRUPT 0x02 +#define MI_WAIT_FOR_EVENT 0x03 +#define MI_FLUSH 0x04 +#define MI_REPORT_HEAD 0x07 +#define MI_ARB_ON_OFF 0x08 +#define MI_BATCH_BUFFER_END 0x0A +#define MI_OVERLAY_FLIP 0x11 +#define MI_LOAD_SCAN_LINES_INCL 0x12 +#define MI_LOAD_SCAN_LINES_EXCL 0x13 +#define MI_DISPLAY_BUFFER_INFO 0x14 +#define MI_SET_CONTEXT 0x18 +#define MI_STORE_DATA_IMM 0x20 +#define MI_STORE_DATA_INDEX 0x21 +#define MI_LOAD_REGISTER_IMM 0x22 +#define MI_STORE_REGISTER_MEM 0x24 +#define MI_BATCH_BUFFER_START 0x31 + +#define MI_SYNCHRONOUS_FLIP 0x0 +#define MI_ASYNCHRONOUS_FLIP 0x1 + +#define MI_BUFFER_SECURE 0x0 +#define MI_BUFFER_NONSECURE 0x1 + +#define MI_ARBITRATE_AT_CHAIN_POINTS 0x0 +#define MI_ARBITRATE_BETWEEN_INSTS 0x1 +#define MI_NO_ARBITRATION 0x3 + +#define MI_CONDITION_CODE_WAIT_DISABLED 0x0 +#define MI_CONDITION_CODE_WAIT_0 0x1 +#define MI_CONDITION_CODE_WAIT_1 0x2 +#define MI_CONDITION_CODE_WAIT_2 0x3 +#define MI_CONDITION_CODE_WAIT_3 0x4 +#define MI_CONDITION_CODE_WAIT_4 0x5 + +#define MI_DISPLAY_PIPE_A 0x0 +#define MI_DISPLAY_PIPE_B 0x1 + +#define MI_DISPLAY_PLANE_A 0x0 +#define MI_DISPLAY_PLANE_B 0x1 +#define MI_DISPLAY_PLANE_C 0x2 + +#define MI_STANDARD_FLIP 0x0 +#define MI_ENQUEUE_FLIP_PERFORM_BASE_FRAME_NUMBER_LOAD 0x1 +#define MI_ENQUEUE_FLIP_TARGET_FRAME_NUMBER_RELATIVE 0x2 +#define MI_ENQUEUE_FLIP_ABSOLUTE_TARGET_FRAME_NUMBER 0x3 + +#define MI_PHYSICAL_ADDRESS 0x0 +#define MI_VIRTUAL_ADDRESS 0x1 + +#define MI_BUFFER_MEMORY_MAIN 0x0 +#define MI_BUFFER_MEMORY_GTT 0x2 +#define MI_BUFFER_MEMORY_PER_PROCESS_GTT 0x3 + +#define MI_FLIP_CONTINUE 0x0 +#define MI_FLIP_ON 0x1 +#define MI_FLIP_OFF 0x2 + +#define MI_UNTRUSTED_REGISTER_SPACE 0x0 +#define MI_TRUSTED_REGISTER_SPACE 0x1 + +/* 3D state: + */ +#define _3DOP_3DSTATE_PIPELINED 0x0 +#define _3DOP_3DSTATE_NONPIPELINED 0x1 +#define _3DOP_3DCONTROL 0x2 +#define _3DOP_3DPRIMITIVE 0x3 + +#define _3DSTATE_PIPELINED_POINTERS 0x00 +#define _3DSTATE_BINDING_TABLE_POINTERS 0x01 +#define _3DSTATE_VERTEX_BUFFERS 0x08 +#define _3DSTATE_VERTEX_ELEMENTS 0x09 +#define _3DSTATE_INDEX_BUFFER 0x0A +#define _3DSTATE_VF_STATISTICS 0x0B +#define _3DSTATE_DRAWING_RECTANGLE 0x00 +#define _3DSTATE_CONSTANT_COLOR 0x01 +#define _3DSTATE_SAMPLER_PALETTE_LOAD 0x02 +#define _3DSTATE_CHROMA_KEY 0x04 +#define _3DSTATE_DEPTH_BUFFER 0x05 +#define _3DSTATE_POLY_STIPPLE_OFFSET 0x06 +#define _3DSTATE_POLY_STIPPLE_PATTERN 0x07 +#define _3DSTATE_LINE_STIPPLE 0x08 +#define _3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP 0x09 +#define _3DCONTROL 0x00 +#define _3DPRIMITIVE 0x00 + +#define PIPE_CONTROL_NOWRITE 0x00 +#define PIPE_CONTROL_WRITEIMMEDIATE 0x01 +#define PIPE_CONTROL_WRITEDEPTH 0x02 +#define PIPE_CONTROL_WRITETIMESTAMP 0x03 + +#define PIPE_CONTROL_GTTWRITE_PROCESS_LOCAL 0x00 +#define PIPE_CONTROL_GTTWRITE_GLOBAL 0x01 + +#define _3DPRIM_POINTLIST 0x01 +#define _3DPRIM_LINELIST 0x02 +#define _3DPRIM_LINESTRIP 0x03 +#define _3DPRIM_TRILIST 0x04 +#define _3DPRIM_TRISTRIP 0x05 +#define _3DPRIM_TRIFAN 0x06 +#define _3DPRIM_QUADLIST 0x07 +#define _3DPRIM_QUADSTRIP 0x08 +#define _3DPRIM_LINELIST_ADJ 0x09 +#define _3DPRIM_LINESTRIP_ADJ 0x0A +#define _3DPRIM_TRILIST_ADJ 0x0B +#define _3DPRIM_TRISTRIP_ADJ 0x0C +#define _3DPRIM_TRISTRIP_REVERSE 0x0D +#define _3DPRIM_POLYGON 0x0E +#define _3DPRIM_RECTLIST 0x0F +#define _3DPRIM_LINELOOP 0x10 +#define _3DPRIM_POINTLIST_BF 0x11 +#define _3DPRIM_LINESTRIP_CONT 0x12 +#define _3DPRIM_LINESTRIP_BF 0x13 +#define _3DPRIM_LINESTRIP_CONT_BF 0x14 +#define _3DPRIM_TRIFAN_NOSTIPPLE 0x15 + +#define _3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL 0 +#define _3DPRIM_VERTEXBUFFER_ACCESS_RANDOM 1 + +#define BRW_ANISORATIO_2 0 +#define BRW_ANISORATIO_4 1 +#define BRW_ANISORATIO_6 2 +#define BRW_ANISORATIO_8 3 +#define BRW_ANISORATIO_10 4 +#define BRW_ANISORATIO_12 5 +#define BRW_ANISORATIO_14 6 +#define BRW_ANISORATIO_16 7 + +#define BRW_BLENDFACTOR_ONE 0x1 +#define BRW_BLENDFACTOR_SRC_COLOR 0x2 +#define BRW_BLENDFACTOR_SRC_ALPHA 0x3 +#define BRW_BLENDFACTOR_DST_ALPHA 0x4 +#define BRW_BLENDFACTOR_DST_COLOR 0x5 +#define BRW_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 +#define BRW_BLENDFACTOR_CONST_COLOR 0x7 +#define BRW_BLENDFACTOR_CONST_ALPHA 0x8 +#define BRW_BLENDFACTOR_SRC1_COLOR 0x9 +#define BRW_BLENDFACTOR_SRC1_ALPHA 0x0A +#define BRW_BLENDFACTOR_ZERO 0x11 +#define BRW_BLENDFACTOR_INV_SRC_COLOR 0x12 +#define BRW_BLENDFACTOR_INV_SRC_ALPHA 0x13 +#define BRW_BLENDFACTOR_INV_DST_ALPHA 0x14 +#define BRW_BLENDFACTOR_INV_DST_COLOR 0x15 +#define BRW_BLENDFACTOR_INV_CONST_COLOR 0x17 +#define BRW_BLENDFACTOR_INV_CONST_ALPHA 0x18 +#define BRW_BLENDFACTOR_INV_SRC1_COLOR 0x19 +#define BRW_BLENDFACTOR_INV_SRC1_ALPHA 0x1A + +#define BRW_BLENDFUNCTION_ADD 0 +#define BRW_BLENDFUNCTION_SUBTRACT 1 +#define BRW_BLENDFUNCTION_REVERSE_SUBTRACT 2 +#define BRW_BLENDFUNCTION_MIN 3 +#define BRW_BLENDFUNCTION_MAX 4 + +#define BRW_ALPHATEST_FORMAT_UNORM8 0 +#define BRW_ALPHATEST_FORMAT_FLOAT32 1 + +#define BRW_CHROMAKEY_KILL_ON_ANY_MATCH 0 +#define BRW_CHROMAKEY_REPLACE_BLACK 1 + +#define BRW_CLIP_API_OGL 0 +#define BRW_CLIP_API_DX 1 + +#define BRW_CLIPMODE_NORMAL 0 +#define BRW_CLIPMODE_CLIP_ALL 1 +#define BRW_CLIPMODE_CLIP_NON_REJECTED 2 +#define BRW_CLIPMODE_REJECT_ALL 3 +#define BRW_CLIPMODE_ACCEPT_ALL 4 + +#define BRW_CLIP_NDCSPACE 0 +#define BRW_CLIP_SCREENSPACE 1 + +#define BRW_COMPAREFUNCTION_ALWAYS 0 +#define BRW_COMPAREFUNCTION_NEVER 1 +#define BRW_COMPAREFUNCTION_LESS 2 +#define BRW_COMPAREFUNCTION_EQUAL 3 +#define BRW_COMPAREFUNCTION_LEQUAL 4 +#define BRW_COMPAREFUNCTION_GREATER 5 +#define BRW_COMPAREFUNCTION_NOTEQUAL 6 +#define BRW_COMPAREFUNCTION_GEQUAL 7 + +#define BRW_COVERAGE_PIXELS_HALF 0 +#define BRW_COVERAGE_PIXELS_1 1 +#define BRW_COVERAGE_PIXELS_2 2 +#define BRW_COVERAGE_PIXELS_4 3 + +#define BRW_CULLMODE_BOTH 0 +#define BRW_CULLMODE_NONE 1 +#define BRW_CULLMODE_FRONT 2 +#define BRW_CULLMODE_BACK 3 + +#define BRW_DEFAULTCOLOR_R8G8B8A8_UNORM 0 +#define BRW_DEFAULTCOLOR_R32G32B32A32_FLOAT 1 + +#define BRW_DEPTHFORMAT_D32_FLOAT_S8X24_UINT 0 +#define BRW_DEPTHFORMAT_D32_FLOAT 1 +#define BRW_DEPTHFORMAT_D24_UNORM_S8_UINT 2 +#define BRW_DEPTHFORMAT_D16_UNORM 5 + +#define BRW_FLOATING_POINT_IEEE_754 0 +#define BRW_FLOATING_POINT_NON_IEEE_754 1 + +#define BRW_FRONTWINDING_CW 0 +#define BRW_FRONTWINDING_CCW 1 + +#define BRW_SPRITE_POINT_ENABLE 16 + +#define BRW_INDEX_BYTE 0 +#define BRW_INDEX_WORD 1 +#define BRW_INDEX_DWORD 2 + +#define BRW_LOGICOPFUNCTION_CLEAR 0 +#define BRW_LOGICOPFUNCTION_NOR 1 +#define BRW_LOGICOPFUNCTION_AND_INVERTED 2 +#define BRW_LOGICOPFUNCTION_COPY_INVERTED 3 +#define BRW_LOGICOPFUNCTION_AND_REVERSE 4 +#define BRW_LOGICOPFUNCTION_INVERT 5 +#define BRW_LOGICOPFUNCTION_XOR 6 +#define BRW_LOGICOPFUNCTION_NAND 7 +#define BRW_LOGICOPFUNCTION_AND 8 +#define BRW_LOGICOPFUNCTION_EQUIV 9 +#define BRW_LOGICOPFUNCTION_NOOP 10 +#define BRW_LOGICOPFUNCTION_OR_INVERTED 11 +#define BRW_LOGICOPFUNCTION_COPY 12 +#define BRW_LOGICOPFUNCTION_OR_REVERSE 13 +#define BRW_LOGICOPFUNCTION_OR 14 +#define BRW_LOGICOPFUNCTION_SET 15 + +#define BRW_MAPFILTER_NEAREST 0x0 +#define BRW_MAPFILTER_LINEAR 0x1 +#define BRW_MAPFILTER_ANISOTROPIC 0x2 + +#define BRW_MIPFILTER_NONE 0 +#define BRW_MIPFILTER_NEAREST 1 +#define BRW_MIPFILTER_LINEAR 3 + +#define BRW_POLYGON_FRONT_FACING 0 +#define BRW_POLYGON_BACK_FACING 1 + +#define BRW_PREFILTER_ALWAYS 0x0 +#define BRW_PREFILTER_NEVER 0x1 +#define BRW_PREFILTER_LESS 0x2 +#define BRW_PREFILTER_EQUAL 0x3 +#define BRW_PREFILTER_LEQUAL 0x4 +#define BRW_PREFILTER_GREATER 0x5 +#define BRW_PREFILTER_NOTEQUAL 0x6 +#define BRW_PREFILTER_GEQUAL 0x7 + +#define BRW_PROVOKING_VERTEX_0 0 +#define BRW_PROVOKING_VERTEX_1 1 +#define BRW_PROVOKING_VERTEX_2 2 + +#define BRW_RASTRULE_UPPER_LEFT 0 +#define BRW_RASTRULE_UPPER_RIGHT 1 + +#define BRW_RENDERTARGET_CLAMPRANGE_UNORM 0 +#define BRW_RENDERTARGET_CLAMPRANGE_SNORM 1 +#define BRW_RENDERTARGET_CLAMPRANGE_FORMAT 2 + +#define BRW_STENCILOP_KEEP 0 +#define BRW_STENCILOP_ZERO 1 +#define BRW_STENCILOP_REPLACE 2 +#define BRW_STENCILOP_INCRSAT 3 +#define BRW_STENCILOP_DECRSAT 4 +#define BRW_STENCILOP_INCR 5 +#define BRW_STENCILOP_DECR 6 +#define BRW_STENCILOP_INVERT 7 + +#define BRW_SURFACE_MIPMAPLAYOUT_BELOW 0 +#define BRW_SURFACE_MIPMAPLAYOUT_RIGHT 1 + +#define BRW_SURFACEFORMAT_R32G32B32A32_FLOAT 0x000 +#define BRW_SURFACEFORMAT_R32G32B32A32_SINT 0x001 +#define BRW_SURFACEFORMAT_R32G32B32A32_UINT 0x002 +#define BRW_SURFACEFORMAT_R32G32B32A32_UNORM 0x003 +#define BRW_SURFACEFORMAT_R32G32B32A32_SNORM 0x004 +#define BRW_SURFACEFORMAT_R64G64_FLOAT 0x005 +#define BRW_SURFACEFORMAT_R32G32B32X32_FLOAT 0x006 +#define BRW_SURFACEFORMAT_R32G32B32A32_SSCALED 0x007 +#define BRW_SURFACEFORMAT_R32G32B32A32_USCALED 0x008 +#define BRW_SURFACEFORMAT_R32G32B32_FLOAT 0x040 +#define BRW_SURFACEFORMAT_R32G32B32_SINT 0x041 +#define BRW_SURFACEFORMAT_R32G32B32_UINT 0x042 +#define BRW_SURFACEFORMAT_R32G32B32_UNORM 0x043 +#define BRW_SURFACEFORMAT_R32G32B32_SNORM 0x044 +#define BRW_SURFACEFORMAT_R32G32B32_SSCALED 0x045 +#define BRW_SURFACEFORMAT_R32G32B32_USCALED 0x046 +#define BRW_SURFACEFORMAT_R16G16B16A16_UNORM 0x080 +#define BRW_SURFACEFORMAT_R16G16B16A16_SNORM 0x081 +#define BRW_SURFACEFORMAT_R16G16B16A16_SINT 0x082 +#define BRW_SURFACEFORMAT_R16G16B16A16_UINT 0x083 +#define BRW_SURFACEFORMAT_R16G16B16A16_FLOAT 0x084 +#define BRW_SURFACEFORMAT_R32G32_FLOAT 0x085 +#define BRW_SURFACEFORMAT_R32G32_SINT 0x086 +#define BRW_SURFACEFORMAT_R32G32_UINT 0x087 +#define BRW_SURFACEFORMAT_R32_FLOAT_X8X24_TYPELESS 0x088 +#define BRW_SURFACEFORMAT_X32_TYPELESS_G8X24_UINT 0x089 +#define BRW_SURFACEFORMAT_L32A32_FLOAT 0x08A +#define BRW_SURFACEFORMAT_R32G32_UNORM 0x08B +#define BRW_SURFACEFORMAT_R32G32_SNORM 0x08C +#define BRW_SURFACEFORMAT_R64_FLOAT 0x08D +#define BRW_SURFACEFORMAT_R16G16B16X16_UNORM 0x08E +#define BRW_SURFACEFORMAT_R16G16B16X16_FLOAT 0x08F +#define BRW_SURFACEFORMAT_A32X32_FLOAT 0x090 +#define BRW_SURFACEFORMAT_L32X32_FLOAT 0x091 +#define BRW_SURFACEFORMAT_I32X32_FLOAT 0x092 +#define BRW_SURFACEFORMAT_R16G16B16A16_SSCALED 0x093 +#define BRW_SURFACEFORMAT_R16G16B16A16_USCALED 0x094 +#define BRW_SURFACEFORMAT_R32G32_SSCALED 0x095 +#define BRW_SURFACEFORMAT_R32G32_USCALED 0x096 +#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM 0x0C0 +#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB 0x0C1 +#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM 0x0C2 +#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM_SRGB 0x0C3 +#define BRW_SURFACEFORMAT_R10G10B10A2_UINT 0x0C4 +#define BRW_SURFACEFORMAT_R10G10B10_SNORM_A2_UNORM 0x0C5 +#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM 0x0C7 +#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB 0x0C8 +#define BRW_SURFACEFORMAT_R8G8B8A8_SNORM 0x0C9 +#define BRW_SURFACEFORMAT_R8G8B8A8_SINT 0x0CA +#define BRW_SURFACEFORMAT_R8G8B8A8_UINT 0x0CB +#define BRW_SURFACEFORMAT_R16G16_UNORM 0x0CC +#define BRW_SURFACEFORMAT_R16G16_SNORM 0x0CD +#define BRW_SURFACEFORMAT_R16G16_SINT 0x0CE +#define BRW_SURFACEFORMAT_R16G16_UINT 0x0CF +#define BRW_SURFACEFORMAT_R16G16_FLOAT 0x0D0 +#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM 0x0D1 +#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM_SRGB 0x0D2 +#define BRW_SURFACEFORMAT_R11G11B10_FLOAT 0x0D3 +#define BRW_SURFACEFORMAT_R32_SINT 0x0D6 +#define BRW_SURFACEFORMAT_R32_UINT 0x0D7 +#define BRW_SURFACEFORMAT_R32_FLOAT 0x0D8 +#define BRW_SURFACEFORMAT_R24_UNORM_X8_TYPELESS 0x0D9 +#define BRW_SURFACEFORMAT_X24_TYPELESS_G8_UINT 0x0DA +#define BRW_SURFACEFORMAT_L16A16_UNORM 0x0DF +#define BRW_SURFACEFORMAT_I24X8_UNORM 0x0E0 +#define BRW_SURFACEFORMAT_L24X8_UNORM 0x0E1 +#define BRW_SURFACEFORMAT_A24X8_UNORM 0x0E2 +#define BRW_SURFACEFORMAT_I32_FLOAT 0x0E3 +#define BRW_SURFACEFORMAT_L32_FLOAT 0x0E4 +#define BRW_SURFACEFORMAT_A32_FLOAT 0x0E5 +#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM 0x0E9 +#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM_SRGB 0x0EA +#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM 0x0EB +#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM_SRGB 0x0EC +#define BRW_SURFACEFORMAT_R9G9B9E5_SHAREDEXP 0x0ED +#define BRW_SURFACEFORMAT_B10G10R10X2_UNORM 0x0EE +#define BRW_SURFACEFORMAT_L16A16_FLOAT 0x0F0 +#define BRW_SURFACEFORMAT_R32_UNORM 0x0F1 +#define BRW_SURFACEFORMAT_R32_SNORM 0x0F2 +#define BRW_SURFACEFORMAT_R10G10B10X2_USCALED 0x0F3 +#define BRW_SURFACEFORMAT_R8G8B8A8_SSCALED 0x0F4 +#define BRW_SURFACEFORMAT_R8G8B8A8_USCALED 0x0F5 +#define BRW_SURFACEFORMAT_R16G16_SSCALED 0x0F6 +#define BRW_SURFACEFORMAT_R16G16_USCALED 0x0F7 +#define BRW_SURFACEFORMAT_R32_SSCALED 0x0F8 +#define BRW_SURFACEFORMAT_R32_USCALED 0x0F9 +#define BRW_SURFACEFORMAT_B5G6R5_UNORM 0x100 +#define BRW_SURFACEFORMAT_B5G6R5_UNORM_SRGB 0x101 +#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM 0x102 +#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM_SRGB 0x103 +#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM 0x104 +#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM_SRGB 0x105 +#define BRW_SURFACEFORMAT_R8G8_UNORM 0x106 +#define BRW_SURFACEFORMAT_R8G8_SNORM 0x107 +#define BRW_SURFACEFORMAT_R8G8_SINT 0x108 +#define BRW_SURFACEFORMAT_R8G8_UINT 0x109 +#define BRW_SURFACEFORMAT_R16_UNORM 0x10A +#define BRW_SURFACEFORMAT_R16_SNORM 0x10B +#define BRW_SURFACEFORMAT_R16_SINT 0x10C +#define BRW_SURFACEFORMAT_R16_UINT 0x10D +#define BRW_SURFACEFORMAT_R16_FLOAT 0x10E +#define BRW_SURFACEFORMAT_I16_UNORM 0x111 +#define BRW_SURFACEFORMAT_L16_UNORM 0x112 +#define BRW_SURFACEFORMAT_A16_UNORM 0x113 +#define BRW_SURFACEFORMAT_L8A8_UNORM 0x114 +#define BRW_SURFACEFORMAT_I16_FLOAT 0x115 +#define BRW_SURFACEFORMAT_L16_FLOAT 0x116 +#define BRW_SURFACEFORMAT_A16_FLOAT 0x117 +#define BRW_SURFACEFORMAT_R5G5_SNORM_B6_UNORM 0x119 +#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM 0x11A +#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM_SRGB 0x11B +#define BRW_SURFACEFORMAT_R8G8_SSCALED 0x11C +#define BRW_SURFACEFORMAT_R8G8_USCALED 0x11D +#define BRW_SURFACEFORMAT_R16_SSCALED 0x11E +#define BRW_SURFACEFORMAT_R16_USCALED 0x11F +#define BRW_SURFACEFORMAT_R8_UNORM 0x140 +#define BRW_SURFACEFORMAT_R8_SNORM 0x141 +#define BRW_SURFACEFORMAT_R8_SINT 0x142 +#define BRW_SURFACEFORMAT_R8_UINT 0x143 +#define BRW_SURFACEFORMAT_A8_UNORM 0x144 +#define BRW_SURFACEFORMAT_I8_UNORM 0x145 +#define BRW_SURFACEFORMAT_L8_UNORM 0x146 +#define BRW_SURFACEFORMAT_P4A4_UNORM 0x147 +#define BRW_SURFACEFORMAT_A4P4_UNORM 0x148 +#define BRW_SURFACEFORMAT_R8_SSCALED 0x149 +#define BRW_SURFACEFORMAT_R8_USCALED 0x14A +#define BRW_SURFACEFORMAT_R1_UINT 0x181 +#define BRW_SURFACEFORMAT_YCRCB_NORMAL 0x182 +#define BRW_SURFACEFORMAT_YCRCB_SWAPUVY 0x183 +#define BRW_SURFACEFORMAT_BC1_UNORM 0x186 +#define BRW_SURFACEFORMAT_BC2_UNORM 0x187 +#define BRW_SURFACEFORMAT_BC3_UNORM 0x188 +#define BRW_SURFACEFORMAT_BC4_UNORM 0x189 +#define BRW_SURFACEFORMAT_BC5_UNORM 0x18A +#define BRW_SURFACEFORMAT_BC1_UNORM_SRGB 0x18B +#define BRW_SURFACEFORMAT_BC2_UNORM_SRGB 0x18C +#define BRW_SURFACEFORMAT_BC3_UNORM_SRGB 0x18D +#define BRW_SURFACEFORMAT_MONO8 0x18E +#define BRW_SURFACEFORMAT_YCRCB_SWAPUV 0x18F +#define BRW_SURFACEFORMAT_YCRCB_SWAPY 0x190 +#define BRW_SURFACEFORMAT_DXT1_RGB 0x191 +#define BRW_SURFACEFORMAT_FXT1 0x192 +#define BRW_SURFACEFORMAT_R8G8B8_UNORM 0x193 +#define BRW_SURFACEFORMAT_R8G8B8_SNORM 0x194 +#define BRW_SURFACEFORMAT_R8G8B8_SSCALED 0x195 +#define BRW_SURFACEFORMAT_R8G8B8_USCALED 0x196 +#define BRW_SURFACEFORMAT_R64G64B64A64_FLOAT 0x197 +#define BRW_SURFACEFORMAT_R64G64B64_FLOAT 0x198 +#define BRW_SURFACEFORMAT_BC4_SNORM 0x199 +#define BRW_SURFACEFORMAT_BC5_SNORM 0x19A +#define BRW_SURFACEFORMAT_R16G16B16_UNORM 0x19C +#define BRW_SURFACEFORMAT_R16G16B16_SNORM 0x19D +#define BRW_SURFACEFORMAT_R16G16B16_SSCALED 0x19E +#define BRW_SURFACEFORMAT_R16G16B16_USCALED 0x19F + +#define BRW_SURFACERETURNFORMAT_FLOAT32 0 +#define BRW_SURFACERETURNFORMAT_S1 1 + +#define BRW_SURFACE_1D 0 +#define BRW_SURFACE_2D 1 +#define BRW_SURFACE_3D 2 +#define BRW_SURFACE_CUBE 3 +#define BRW_SURFACE_BUFFER 4 +#define BRW_SURFACE_NULL 7 + +#define BRW_TEXCOORDMODE_WRAP 0 +#define BRW_TEXCOORDMODE_MIRROR 1 +#define BRW_TEXCOORDMODE_CLAMP 2 +#define BRW_TEXCOORDMODE_CUBE 3 +#define BRW_TEXCOORDMODE_CLAMP_BORDER 4 +#define BRW_TEXCOORDMODE_MIRROR_ONCE 5 + +#define BRW_THREAD_PRIORITY_NORMAL 0 +#define BRW_THREAD_PRIORITY_HIGH 1 + +#define BRW_TILEWALK_XMAJOR 0 +#define BRW_TILEWALK_YMAJOR 1 + +#define BRW_VERTEX_SUBPIXEL_PRECISION_8BITS 0 +#define BRW_VERTEX_SUBPIXEL_PRECISION_4BITS 1 + +#define BRW_VERTEXBUFFER_ACCESS_VERTEXDATA 0 +#define BRW_VERTEXBUFFER_ACCESS_INSTANCEDATA 1 + +#define BRW_VFCOMPONENT_NOSTORE 0 +#define BRW_VFCOMPONENT_STORE_SRC 1 +#define BRW_VFCOMPONENT_STORE_0 2 +#define BRW_VFCOMPONENT_STORE_1_FLT 3 +#define BRW_VFCOMPONENT_STORE_1_INT 4 +#define BRW_VFCOMPONENT_STORE_VID 5 +#define BRW_VFCOMPONENT_STORE_IID 6 +#define BRW_VFCOMPONENT_STORE_PID 7 + + + +/* Execution Unit (EU) defines + */ + +#define BRW_ALIGN_1 0 +#define BRW_ALIGN_16 1 + +#define BRW_ADDRESS_DIRECT 0 +#define BRW_ADDRESS_REGISTER_INDIRECT_REGISTER 1 + +#define BRW_CHANNEL_X 0 +#define BRW_CHANNEL_Y 1 +#define BRW_CHANNEL_Z 2 +#define BRW_CHANNEL_W 3 + +#define BRW_COMPRESSION_NONE 0 +#define BRW_COMPRESSION_2NDHALF 1 +#define BRW_COMPRESSION_COMPRESSED 2 + +#define BRW_CONDITIONAL_NONE 0 +#define BRW_CONDITIONAL_Z 1 +#define BRW_CONDITIONAL_NZ 2 +#define BRW_CONDITIONAL_EQ 1 /* Z */ +#define BRW_CONDITIONAL_NEQ 2 /* NZ */ +#define BRW_CONDITIONAL_G 3 +#define BRW_CONDITIONAL_GE 4 +#define BRW_CONDITIONAL_L 5 +#define BRW_CONDITIONAL_LE 6 +#define BRW_CONDITIONAL_C 7 +#define BRW_CONDITIONAL_O 8 + +#define BRW_DEBUG_NONE 0 +#define BRW_DEBUG_BREAKPOINT 1 + +#define BRW_DEPENDENCY_NORMAL 0 +#define BRW_DEPENDENCY_NOTCLEARED 1 +#define BRW_DEPENDENCY_NOTCHECKED 2 +#define BRW_DEPENDENCY_DISABLE 3 + +#define BRW_EXECUTE_1 0 +#define BRW_EXECUTE_2 1 +#define BRW_EXECUTE_4 2 +#define BRW_EXECUTE_8 3 +#define BRW_EXECUTE_16 4 +#define BRW_EXECUTE_32 5 + +#define BRW_HORIZONTAL_STRIDE_0 0 +#define BRW_HORIZONTAL_STRIDE_1 1 +#define BRW_HORIZONTAL_STRIDE_2 2 +#define BRW_HORIZONTAL_STRIDE_4 3 + +#define BRW_INSTRUCTION_NORMAL 0 +#define BRW_INSTRUCTION_SATURATE 1 + +#define BRW_MASK_ENABLE 0 +#define BRW_MASK_DISABLE 1 + +#define BRW_OPCODE_MOV 1 +#define BRW_OPCODE_SEL 2 +#define BRW_OPCODE_NOT 4 +#define BRW_OPCODE_AND 5 +#define BRW_OPCODE_OR 6 +#define BRW_OPCODE_XOR 7 +#define BRW_OPCODE_SHR 8 +#define BRW_OPCODE_SHL 9 +#define BRW_OPCODE_RSR 10 +#define BRW_OPCODE_RSL 11 +#define BRW_OPCODE_ASR 12 +#define BRW_OPCODE_CMP 16 +#define BRW_OPCODE_JMPI 32 +#define BRW_OPCODE_IF 34 +#define BRW_OPCODE_IFF 35 +#define BRW_OPCODE_ELSE 36 +#define BRW_OPCODE_ENDIF 37 +#define BRW_OPCODE_DO 38 +#define BRW_OPCODE_WHILE 39 +#define BRW_OPCODE_BREAK 40 +#define BRW_OPCODE_CONTINUE 41 +#define BRW_OPCODE_HALT 42 +#define BRW_OPCODE_MSAVE 44 +#define BRW_OPCODE_MRESTORE 45 +#define BRW_OPCODE_PUSH 46 +#define BRW_OPCODE_POP 47 +#define BRW_OPCODE_WAIT 48 +#define BRW_OPCODE_SEND 49 +#define BRW_OPCODE_ADD 64 +#define BRW_OPCODE_MUL 65 +#define BRW_OPCODE_AVG 66 +#define BRW_OPCODE_FRC 67 +#define BRW_OPCODE_RNDU 68 +#define BRW_OPCODE_RNDD 69 +#define BRW_OPCODE_RNDE 70 +#define BRW_OPCODE_RNDZ 71 +#define BRW_OPCODE_MAC 72 +#define BRW_OPCODE_MACH 73 +#define BRW_OPCODE_LZD 74 +#define BRW_OPCODE_SAD2 80 +#define BRW_OPCODE_SADA2 81 +#define BRW_OPCODE_DP4 84 +#define BRW_OPCODE_DPH 85 +#define BRW_OPCODE_DP3 86 +#define BRW_OPCODE_DP2 87 +#define BRW_OPCODE_DPA2 88 +#define BRW_OPCODE_LINE 89 +#define BRW_OPCODE_NOP 126 + +#define BRW_PREDICATE_NONE 0 +#define BRW_PREDICATE_NORMAL 1 +#define BRW_PREDICATE_ALIGN1_ANYV 2 +#define BRW_PREDICATE_ALIGN1_ALLV 3 +#define BRW_PREDICATE_ALIGN1_ANY2H 4 +#define BRW_PREDICATE_ALIGN1_ALL2H 5 +#define BRW_PREDICATE_ALIGN1_ANY4H 6 +#define BRW_PREDICATE_ALIGN1_ALL4H 7 +#define BRW_PREDICATE_ALIGN1_ANY8H 8 +#define BRW_PREDICATE_ALIGN1_ALL8H 9 +#define BRW_PREDICATE_ALIGN1_ANY16H 10 +#define BRW_PREDICATE_ALIGN1_ALL16H 11 +#define BRW_PREDICATE_ALIGN16_REPLICATE_X 2 +#define BRW_PREDICATE_ALIGN16_REPLICATE_Y 3 +#define BRW_PREDICATE_ALIGN16_REPLICATE_Z 4 +#define BRW_PREDICATE_ALIGN16_REPLICATE_W 5 +#define BRW_PREDICATE_ALIGN16_ANY4H 6 +#define BRW_PREDICATE_ALIGN16_ALL4H 7 + +#define BRW_ARCHITECTURE_REGISTER_FILE 0 +#define BRW_GENERAL_REGISTER_FILE 1 +#define BRW_MESSAGE_REGISTER_FILE 2 +#define BRW_IMMEDIATE_VALUE 3 + +#define BRW_REGISTER_TYPE_UD 0 +#define BRW_REGISTER_TYPE_D 1 +#define BRW_REGISTER_TYPE_UW 2 +#define BRW_REGISTER_TYPE_W 3 +#define BRW_REGISTER_TYPE_UB 4 +#define BRW_REGISTER_TYPE_B 5 +#define BRW_REGISTER_TYPE_VF 5 /* packed float vector, immediates only? */ +#define BRW_REGISTER_TYPE_HF 6 +#define BRW_REGISTER_TYPE_V 6 /* packed int vector, immediates only, uword dest only */ +#define BRW_REGISTER_TYPE_F 7 + +#define BRW_ARF_NULL 0x00 +#define BRW_ARF_ADDRESS 0x10 +#define BRW_ARF_ACCUMULATOR 0x20 +#define BRW_ARF_FLAG 0x30 +#define BRW_ARF_MASK 0x40 +#define BRW_ARF_MASK_STACK 0x50 +#define BRW_ARF_MASK_STACK_DEPTH 0x60 +#define BRW_ARF_STATE 0x70 +#define BRW_ARF_CONTROL 0x80 +#define BRW_ARF_NOTIFICATION_COUNT 0x90 +#define BRW_ARF_IP 0xA0 + +#define BRW_AMASK 0 +#define BRW_IMASK 1 +#define BRW_LMASK 2 +#define BRW_CMASK 3 + + + +#define BRW_THREAD_NORMAL 0 +#define BRW_THREAD_ATOMIC 1 +#define BRW_THREAD_SWITCH 2 + +#define BRW_VERTICAL_STRIDE_0 0 +#define BRW_VERTICAL_STRIDE_1 1 +#define BRW_VERTICAL_STRIDE_2 2 +#define BRW_VERTICAL_STRIDE_4 3 +#define BRW_VERTICAL_STRIDE_8 4 +#define BRW_VERTICAL_STRIDE_16 5 +#define BRW_VERTICAL_STRIDE_32 6 +#define BRW_VERTICAL_STRIDE_64 7 +#define BRW_VERTICAL_STRIDE_128 8 +#define BRW_VERTICAL_STRIDE_256 9 +#define BRW_VERTICAL_STRIDE_ONE_DIMENSIONAL 0xF + +#define BRW_WIDTH_1 0 +#define BRW_WIDTH_2 1 +#define BRW_WIDTH_4 2 +#define BRW_WIDTH_8 3 +#define BRW_WIDTH_16 4 + +#define BRW_STATELESS_BUFFER_BOUNDARY_1K 0 +#define BRW_STATELESS_BUFFER_BOUNDARY_2K 1 +#define BRW_STATELESS_BUFFER_BOUNDARY_4K 2 +#define BRW_STATELESS_BUFFER_BOUNDARY_8K 3 +#define BRW_STATELESS_BUFFER_BOUNDARY_16K 4 +#define BRW_STATELESS_BUFFER_BOUNDARY_32K 5 +#define BRW_STATELESS_BUFFER_BOUNDARY_64K 6 +#define BRW_STATELESS_BUFFER_BOUNDARY_128K 7 +#define BRW_STATELESS_BUFFER_BOUNDARY_256K 8 +#define BRW_STATELESS_BUFFER_BOUNDARY_512K 9 +#define BRW_STATELESS_BUFFER_BOUNDARY_1M 10 +#define BRW_STATELESS_BUFFER_BOUNDARY_2M 11 + +#define BRW_POLYGON_FACING_FRONT 0 +#define BRW_POLYGON_FACING_BACK 1 + +#define BRW_MESSAGE_TARGET_NULL 0 +#define BRW_MESSAGE_TARGET_MATH 1 +#define BRW_MESSAGE_TARGET_SAMPLER 2 +#define BRW_MESSAGE_TARGET_GATEWAY 3 +#define BRW_MESSAGE_TARGET_DATAPORT_READ 4 +#define BRW_MESSAGE_TARGET_DATAPORT_WRITE 5 +#define BRW_MESSAGE_TARGET_URB 6 +#define BRW_MESSAGE_TARGET_THREAD_SPAWNER 7 + +#define BRW_SAMPLER_RETURN_FORMAT_FLOAT32 0 +#define BRW_SAMPLER_RETURN_FORMAT_UINT32 2 +#define BRW_SAMPLER_RETURN_FORMAT_SINT32 3 + +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS 0 +#define BRW_SAMPLER_MESSAGE_SIMD8_KILLPIX 1 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD 1 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD 1 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS 2 +#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE 0 +#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_RESINFO 2 +#define BRW_SAMPLER_MESSAGE_SIMD8_RESINFO 2 +#define BRW_SAMPLER_MESSAGE_SIMD16_RESINFO 2 +#define BRW_SAMPLER_MESSAGE_SIMD4X2_LD 3 +#define BRW_SAMPLER_MESSAGE_SIMD8_LD 3 +#define BRW_SAMPLER_MESSAGE_SIMD16_LD 3 + +#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW 0 +#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDHIGH 1 +#define BRW_DATAPORT_OWORD_BLOCK_2_OWORDS 2 +#define BRW_DATAPORT_OWORD_BLOCK_4_OWORDS 3 +#define BRW_DATAPORT_OWORD_BLOCK_8_OWORDS 4 + +#define BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD 0 +#define BRW_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS 2 + +#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS 2 +#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS 3 + +#define BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ 0 +#define BRW_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 1 +#define BRW_DATAPORT_READ_MESSAGE_DWORD_BLOCK_READ 2 +#define BRW_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 3 + +#define BRW_DATAPORT_READ_TARGET_DATA_CACHE 0 +#define BRW_DATAPORT_READ_TARGET_RENDER_CACHE 1 +#define BRW_DATAPORT_READ_TARGET_SAMPLER_CACHE 2 + +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE 0 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED 1 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01 2 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23 3 +#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01 4 + +#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE 0 +#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE 1 +#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_BLOCK_WRITE 2 +#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE 3 +#define BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE 4 +#define BRW_DATAPORT_WRITE_MESSAGE_STREAMED_VERTEX_BUFFER_WRITE 5 +#define BRW_DATAPORT_WRITE_MESSAGE_FLUSH_RENDER_CACHE 7 + +#define BRW_MATH_FUNCTION_INV 1 +#define BRW_MATH_FUNCTION_LOG 2 +#define BRW_MATH_FUNCTION_EXP 3 +#define BRW_MATH_FUNCTION_SQRT 4 +#define BRW_MATH_FUNCTION_RSQ 5 +#define BRW_MATH_FUNCTION_SIN 6 /* was 7 */ +#define BRW_MATH_FUNCTION_COS 7 /* was 8 */ +#define BRW_MATH_FUNCTION_SINCOS 8 /* was 6 */ +#define BRW_MATH_FUNCTION_TAN 9 +#define BRW_MATH_FUNCTION_POW 10 +#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER 11 +#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT 12 +#define BRW_MATH_FUNCTION_INT_DIV_REMAINDER 13 + +#define BRW_MATH_INTEGER_UNSIGNED 0 +#define BRW_MATH_INTEGER_SIGNED 1 + +#define BRW_MATH_PRECISION_FULL 0 +#define BRW_MATH_PRECISION_PARTIAL 1 + +#define BRW_MATH_SATURATE_NONE 0 +#define BRW_MATH_SATURATE_SATURATE 1 + +#define BRW_MATH_DATA_VECTOR 0 +#define BRW_MATH_DATA_SCALAR 1 + +#define BRW_URB_OPCODE_WRITE 0 + +#define BRW_URB_SWIZZLE_NONE 0 +#define BRW_URB_SWIZZLE_INTERLEAVE 1 +#define BRW_URB_SWIZZLE_TRANSPOSE 2 + +#define BRW_SCRATCH_SPACE_SIZE_1K 0 +#define BRW_SCRATCH_SPACE_SIZE_2K 1 +#define BRW_SCRATCH_SPACE_SIZE_4K 2 +#define BRW_SCRATCH_SPACE_SIZE_8K 3 +#define BRW_SCRATCH_SPACE_SIZE_16K 4 +#define BRW_SCRATCH_SPACE_SIZE_32K 5 +#define BRW_SCRATCH_SPACE_SIZE_64K 6 +#define BRW_SCRATCH_SPACE_SIZE_128K 7 +#define BRW_SCRATCH_SPACE_SIZE_256K 8 +#define BRW_SCRATCH_SPACE_SIZE_512K 9 +#define BRW_SCRATCH_SPACE_SIZE_1M 10 +#define BRW_SCRATCH_SPACE_SIZE_2M 11 + + + + +#define CMD_URB_FENCE 0x6000 +#define CMD_CONST_BUFFER_STATE 0x6001 +#define CMD_CONST_BUFFER 0x6002 + +#define CMD_STATE_BASE_ADDRESS 0x6101 +#define CMD_STATE_INSN_POINTER 0x6102 +#define CMD_PIPELINE_SELECT 0x6104 + +#define CMD_PIPELINED_STATE_POINTERS 0x7800 +#define CMD_BINDING_TABLE_PTRS 0x7801 +#define CMD_VERTEX_BUFFER 0x7808 +#define CMD_VERTEX_ELEMENT 0x7809 +#define CMD_INDEX_BUFFER 0x780a +#define CMD_VF_STATISTICS 0x780b + +#define CMD_DRAW_RECT 0x7900 +#define CMD_BLEND_CONSTANT_COLOR 0x7901 +#define CMD_CHROMA_KEY 0x7904 +#define CMD_DEPTH_BUFFER 0x7905 +#define CMD_POLY_STIPPLE_OFFSET 0x7906 +#define CMD_POLY_STIPPLE_PATTERN 0x7907 +#define CMD_LINE_STIPPLE_PATTERN 0x7908 +#define CMD_GLOBAL_DEPTH_OFFSET_CLAMP 0x7909 + +#define CMD_PIPE_CONTROL 0x7a00 + +#define CMD_3D_PRIM 0x7b00 + +#define CMD_MI_FLUSH 0x0200 + + +/* Various values from the R0 vertex header: + */ +#define R02_PRIM_END 0x1 +#define R02_PRIM_START 0x2 + + + +#endif diff --git a/src/gallium/drivers/i965simple/brw_draw.c b/src/gallium/drivers/i965simple/brw_draw.c new file mode 100644 index 0000000000..7598e3dc8a --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_draw.c @@ -0,0 +1,239 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include + +#include "brw_batch.h" +#include "brw_draw.h" +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_state.h" + +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = { + _3DPRIM_POINTLIST, + _3DPRIM_LINELIST, + _3DPRIM_LINELOOP, + _3DPRIM_LINESTRIP, + _3DPRIM_TRILIST, + _3DPRIM_TRISTRIP, + _3DPRIM_TRIFAN, + _3DPRIM_QUADLIST, + _3DPRIM_QUADSTRIP, + _3DPRIM_POLYGON +}; + + +static const int reduced_prim[PIPE_PRIM_POLYGON+1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES +}; + + +/* When the primitive changes, set a state bit and re-validate. Not + * the nicest and would rather deal with this by having all the + * programs be immune to the active primitive (ie. cope with all + * possibilities). That may not be realistic however. + */ +static void brw_set_prim(struct brw_context *brw, int prim) +{ + PRINT("PRIM: %d\n", prim); + + /* Slight optimization to avoid the GS program when not needed: + */ + if (prim == PIPE_PRIM_QUAD_STRIP && + brw->attribs.Raster->flatshade && + brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_FILL && + brw->attribs.Raster->fill_ccw == PIPE_POLYGON_MODE_FILL) + prim = PIPE_PRIM_TRIANGLE_STRIP; + + if (prim != brw->primitive) { + brw->primitive = prim; + brw->state.dirty.brw |= BRW_NEW_PRIMITIVE; + + if (reduced_prim[prim] != brw->reduced_primitive) { + brw->reduced_primitive = reduced_prim[prim]; + brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE; + } + + brw_validate_state(brw); + } + +} + + +static unsigned trim(int prim, unsigned length) +{ + if (prim == PIPE_PRIM_QUAD_STRIP) + return length > 3 ? (length - length % 2) : 0; + else if (prim == PIPE_PRIM_QUADS) + return length - length % 4; + else + return length; +} + + + +static boolean brw_emit_prim( struct brw_context *brw, + boolean indexed, + unsigned start, + unsigned count ) + +{ + struct brw_3d_primitive prim_packet; + + if (BRW_DEBUG & DEBUG_PRIMS) + PRINT("PRIM: %d %d %d\n", brw->primitive, start, count); + + prim_packet.header.opcode = CMD_3D_PRIM; + prim_packet.header.length = sizeof(prim_packet)/4 - 2; + prim_packet.header.pad = 0; + prim_packet.header.topology = hw_prim[brw->primitive]; + prim_packet.header.indexed = indexed; + + prim_packet.verts_per_instance = trim(brw->primitive, count); + prim_packet.start_vert_location = start; + prim_packet.instance_count = 1; + prim_packet.start_instance_location = 0; + prim_packet.base_vert_location = 0; + + if (prim_packet.verts_per_instance == 0) + return TRUE; + + return brw_batchbuffer_data( brw->winsys, + &prim_packet, + sizeof(prim_packet) ); +} + + +/* May fail if out of video memory for texture or vbo upload, or on + * fallback conditions. + */ +static boolean brw_try_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *index_buffer, + unsigned index_size, + unsigned mode, + unsigned start, + unsigned count ) +{ + struct brw_context *brw = brw_context(pipe); + + /* Set the first primitive ahead of validate_state: + */ + brw_set_prim(brw, mode); + + /* Upload index, vertex data: + */ + if (index_buffer && + !brw_upload_indices( brw, index_buffer, index_size, start, count )) + return FALSE; + + if (!brw_upload_vertex_buffers(brw)) + return FALSE; + + if (!brw_upload_vertex_elements( brw )) + return FALSE; + + /* XXX: Need to separate validate and upload of state. + */ + if (brw->state.dirty.brw) + brw_validate_state( brw ); + + if (!brw_emit_prim(brw, index_buffer != NULL, + start, count)) + return FALSE; + + return TRUE; +} + + + +static boolean brw_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, + unsigned start, + unsigned count ) +{ + if (!brw_try_draw_elements( pipe, + indexBuffer, + indexSize, + mode, start, count )) + { + /* flush ? */ + + if (!brw_try_draw_elements( pipe, + indexBuffer, + indexSize, + mode, start, + count )) { + assert(0); + return FALSE; + } + } + + return TRUE; +} + + + +static boolean brw_draw_arrays( struct pipe_context *pipe, + unsigned mode, + unsigned start, + unsigned count ) +{ + if (!brw_try_draw_elements( pipe, NULL, 0, mode, start, count )) { + /* flush ? */ + + if (!brw_try_draw_elements( pipe, NULL, 0, mode, start, count )) { + assert(0); + return FALSE; + } + } + + return TRUE; +} + + + +void brw_init_draw_functions( struct brw_context *brw ) +{ + brw->pipe.draw_arrays = brw_draw_arrays; + brw->pipe.draw_elements = brw_draw_elements; +} + + diff --git a/src/gallium/drivers/i965simple/brw_draw.h b/src/gallium/drivers/i965simple/brw_draw.h new file mode 100644 index 0000000000..62fe0d5d0e --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_draw.h @@ -0,0 +1,55 @@ + /************************************************************************** + * + * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef BRW_DRAW_H +#define BRW_DRAW_H + +#include "pipe/p_context.h" + +struct brw_context; + + + +void brw_init_draw_functions( struct brw_context *brw ); + + +boolean brw_upload_vertices( struct brw_context *brw, + unsigned min_index, + unsigned max_index ); + +boolean brw_upload_indices(struct brw_context *brw, + const struct pipe_buffer *index_buffer, + int ib_size, int start, int count); + +boolean brw_upload_vertex_buffers( struct brw_context *brw ); +boolean brw_upload_vertex_elements( struct brw_context *brw ); + +unsigned brw_translate_surface_format( unsigned id ); + + + +#endif diff --git a/src/gallium/drivers/i965simple/brw_draw_upload.c b/src/gallium/drivers/i965simple/brw_draw_upload.c new file mode 100644 index 0000000000..aa85d93866 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_draw_upload.c @@ -0,0 +1,299 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include + +#include "brw_batch.h" +#include "brw_draw.h" +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_state.h" + +struct brw_array_state { + union header_union header; + + struct { + union { + struct { + unsigned pitch:11; + unsigned pad:15; + unsigned access_type:1; + unsigned vb_index:5; + } bits; + unsigned dword; + } vb0; + + struct pipe_buffer *buffer; + unsigned offset; + + unsigned max_index; + unsigned instance_data_step_rate; + + } vb[BRW_VBP_MAX]; +}; + + + +unsigned brw_translate_surface_format( unsigned id ) +{ + switch (id) { + case PIPE_FORMAT_R64_FLOAT: + return BRW_SURFACEFORMAT_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return BRW_SURFACEFORMAT_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return BRW_SURFACEFORMAT_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return BRW_SURFACEFORMAT_R64G64B64A64_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + return BRW_SURFACEFORMAT_R32_FLOAT; + case PIPE_FORMAT_R32G32_FLOAT: + return BRW_SURFACEFORMAT_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return BRW_SURFACEFORMAT_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return BRW_SURFACEFORMAT_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return BRW_SURFACEFORMAT_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return BRW_SURFACEFORMAT_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return BRW_SURFACEFORMAT_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return BRW_SURFACEFORMAT_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return BRW_SURFACEFORMAT_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return BRW_SURFACEFORMAT_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return BRW_SURFACEFORMAT_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return BRW_SURFACEFORMAT_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return BRW_SURFACEFORMAT_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return BRW_SURFACEFORMAT_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return BRW_SURFACEFORMAT_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return BRW_SURFACEFORMAT_R32G32B32A32_SNORM; + + case PIPE_FORMAT_R32_SSCALED: + return BRW_SURFACEFORMAT_R32_SSCALED; + case PIPE_FORMAT_R32G32_SSCALED: + return BRW_SURFACEFORMAT_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return BRW_SURFACEFORMAT_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return BRW_SURFACEFORMAT_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return BRW_SURFACEFORMAT_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return BRW_SURFACEFORMAT_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return BRW_SURFACEFORMAT_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return BRW_SURFACEFORMAT_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return BRW_SURFACEFORMAT_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return BRW_SURFACEFORMAT_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return BRW_SURFACEFORMAT_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return BRW_SURFACEFORMAT_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return BRW_SURFACEFORMAT_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return BRW_SURFACEFORMAT_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return BRW_SURFACEFORMAT_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return BRW_SURFACEFORMAT_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return BRW_SURFACEFORMAT_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return BRW_SURFACEFORMAT_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return BRW_SURFACEFORMAT_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return BRW_SURFACEFORMAT_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return BRW_SURFACEFORMAT_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return BRW_SURFACEFORMAT_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return BRW_SURFACEFORMAT_R8G8B8_UNORM; + case PIPE_FORMAT_R8G8B8A8_UNORM: + return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return BRW_SURFACEFORMAT_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return BRW_SURFACEFORMAT_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return BRW_SURFACEFORMAT_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return BRW_SURFACEFORMAT_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return BRW_SURFACEFORMAT_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return BRW_SURFACEFORMAT_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return BRW_SURFACEFORMAT_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return BRW_SURFACEFORMAT_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return BRW_SURFACEFORMAT_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return BRW_SURFACEFORMAT_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return BRW_SURFACEFORMAT_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return BRW_SURFACEFORMAT_R8G8B8A8_SSCALED; + + default: + assert(0); + return 0; + } +} + +static unsigned get_index_type(int type) +{ + switch (type) { + case 1: return BRW_INDEX_BYTE; + case 2: return BRW_INDEX_WORD; + case 4: return BRW_INDEX_DWORD; + default: assert(0); return 0; + } +} + + +boolean brw_upload_vertex_buffers( struct brw_context *brw ) +{ + struct brw_array_state vbp; + unsigned nr_enabled = 0; + unsigned i; + + memset(&vbp, 0, sizeof(vbp)); + + /* This is a hardware limit: + */ + + for (i = 0; i < BRW_VEP_MAX; i++) + { + if (brw->vb.vbo_array[i] == NULL) { + nr_enabled = i; + break; + } + + vbp.vb[i].vb0.bits.pitch = brw->vb.vbo_array[i]->pitch; + vbp.vb[i].vb0.bits.pad = 0; + vbp.vb[i].vb0.bits.access_type = BRW_VERTEXBUFFER_ACCESS_VERTEXDATA; + vbp.vb[i].vb0.bits.vb_index = i; + vbp.vb[i].offset = brw->vb.vbo_array[i]->buffer_offset; + vbp.vb[i].buffer = brw->vb.vbo_array[i]->buffer; + vbp.vb[i].max_index = brw->vb.vbo_array[i]->max_index; + } + + + vbp.header.bits.length = (1 + nr_enabled * 4) - 2; + vbp.header.bits.opcode = CMD_VERTEX_BUFFER; + + BEGIN_BATCH(vbp.header.bits.length+2, 0); + OUT_BATCH( vbp.header.dword ); + + for (i = 0; i < nr_enabled; i++) { + OUT_BATCH( vbp.vb[i].vb0.dword ); + OUT_RELOC( vbp.vb[i].buffer, PIPE_BUFFER_USAGE_GPU_READ, + vbp.vb[i].offset); + OUT_BATCH( vbp.vb[i].max_index ); + OUT_BATCH( vbp.vb[i].instance_data_step_rate ); + } + ADVANCE_BATCH(); + return TRUE; +} + + + +boolean brw_upload_vertex_elements( struct brw_context *brw ) +{ + struct brw_vertex_element_packet vep; + + unsigned i; + unsigned nr_enabled = brw->attribs.VertexProgram->program.num_inputs; + + memset(&vep, 0, sizeof(vep)); + + for (i = 0; i < nr_enabled; i++) + vep.ve[i] = brw->vb.inputs[i]; + + + vep.header.length = (1 + nr_enabled * sizeof(vep.ve[0])/4) - 2; + vep.header.opcode = CMD_VERTEX_ELEMENT; + brw_cached_batch_struct(brw, &vep, 4 + nr_enabled * sizeof(vep.ve[0])); + + return TRUE; +} + +boolean brw_upload_indices( struct brw_context *brw, + const struct pipe_buffer *index_buffer, + int ib_size, int start, int count) +{ + /* Emit the indexbuffer packet: + */ + { + struct brw_indexbuffer ib; + + memset(&ib, 0, sizeof(ib)); + + ib.header.bits.opcode = CMD_INDEX_BUFFER; + ib.header.bits.length = sizeof(ib)/4 - 2; + ib.header.bits.index_format = get_index_type(ib_size); + ib.header.bits.cut_index_enable = 0; + + + BEGIN_BATCH(4, 0); + OUT_BATCH( ib.header.dword ); + OUT_RELOC( index_buffer, PIPE_BUFFER_USAGE_GPU_READ, start); + OUT_RELOC( index_buffer, PIPE_BUFFER_USAGE_GPU_READ, start + count); + OUT_BATCH( 0 ); + ADVANCE_BATCH(); + } + return TRUE; +} diff --git a/src/gallium/drivers/i965simple/brw_eu.c b/src/gallium/drivers/i965simple/brw_eu.c new file mode 100644 index 0000000000..e2002d1821 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_eu.c @@ -0,0 +1,130 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" + + + +/* How does predicate control work when execution_size != 8? Do I + * need to test/set for 0xffff when execution_size is 16? + */ +void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value ) +{ + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + if (value != 0xff) { + if (value != p->flag_value) { + brw_push_insn_state(p); + brw_MOV(p, brw_flag_reg(), brw_imm_uw(value)); + p->flag_value = value; + brw_pop_insn_state(p); + } + + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + } +} + +void brw_set_predicate_control( struct brw_compile *p, unsigned pc ) +{ + p->current->header.predicate_control = pc; +} + +void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional ) +{ + p->current->header.destreg__conditonalmod = conditional; +} + +void brw_set_access_mode( struct brw_compile *p, unsigned access_mode ) +{ + p->current->header.access_mode = access_mode; +} + +void brw_set_compression_control( struct brw_compile *p, boolean compression_control ) +{ + p->current->header.compression_control = compression_control; +} + +void brw_set_mask_control( struct brw_compile *p, unsigned value ) +{ + p->current->header.mask_control = value; +} + +void brw_set_saturate( struct brw_compile *p, unsigned value ) +{ + p->current->header.saturate = value; +} + +void brw_push_insn_state( struct brw_compile *p ) +{ + assert(p->current != &p->stack[BRW_EU_MAX_INSN_STACK-1]); + memcpy(p->current+1, p->current, sizeof(struct brw_instruction)); + p->current++; +} + +void brw_pop_insn_state( struct brw_compile *p ) +{ + assert(p->current != p->stack); + p->current--; +} + + +/*********************************************************************** + */ +void brw_init_compile( struct brw_compile *p ) +{ + p->nr_insn = 0; + p->current = p->stack; + memset(p->current, 0, sizeof(p->current[0])); + + /* Some defaults? + */ + brw_set_mask_control(p, BRW_MASK_ENABLE); /* what does this do? */ + brw_set_saturate(p, 0); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_predicate_control_flag_value(p, 0xff); +} + + +const unsigned *brw_get_program( struct brw_compile *p, + unsigned *sz ) +{ + unsigned i; + + for (i = 0; i < 8; i++) + brw_NOP(p); + + *sz = p->nr_insn * sizeof(struct brw_instruction); + return (const unsigned *)p->store; +} + diff --git a/src/gallium/drivers/i965simple/brw_eu.h b/src/gallium/drivers/i965simple/brw_eu.h new file mode 100644 index 0000000000..23151ae9ed --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_eu.h @@ -0,0 +1,888 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_EU_H +#define BRW_EU_H + +#include "brw_structs.h" +#include "brw_defines.h" + +#include "pipe/p_compiler.h" +#include "pipe/p_shader_tokens.h" + +#define BRW_SWIZZLE4(a,b,c,d) (((a)<<0) | ((b)<<2) | ((c)<<4) | ((d)<<6)) +#define BRW_GET_SWZ(swz, idx) (((swz) >> ((idx)*2)) & 0x3) + +#define BRW_SWIZZLE_NOOP BRW_SWIZZLE4(0,1,2,3) +#define BRW_SWIZZLE_XYZW BRW_SWIZZLE4(0,1,2,3) +#define BRW_SWIZZLE_XXXX BRW_SWIZZLE4(0,0,0,0) +#define BRW_SWIZZLE_XYXY BRW_SWIZZLE4(0,1,0,1) + + +#define REG_SIZE (8*4) + + +/* These aren't hardware structs, just something useful for us to pass around: + * + * Align1 operation has a lot of control over input ranges. Used in + * WM programs to implement shaders decomposed into "channel serial" + * or "structure of array" form: + */ +struct brw_reg +{ + unsigned type:4; + unsigned file:2; + unsigned nr:8; + unsigned subnr:5; /* :1 in align16 */ + unsigned negate:1; /* source only */ + unsigned abs:1; /* source only */ + unsigned vstride:4; /* source only */ + unsigned width:3; /* src only, align1 only */ + unsigned hstride:2; /* src only, align1 only */ + unsigned address_mode:1; /* relative addressing, hopefully! */ + unsigned pad0:1; + + union { + struct { + unsigned swizzle:8; /* src only, align16 only */ + unsigned writemask:4; /* dest only, align16 only */ + int indirect_offset:10; /* relative addressing offset */ + unsigned pad1:10; /* two dwords total */ + } bits; + + float f; + int d; + unsigned ud; + } dw1; +}; + + +struct brw_indirect { + unsigned addr_subnr:4; + int addr_offset:10; + unsigned pad:18; +}; + + +#define BRW_EU_MAX_INSN_STACK 5 +#define BRW_EU_MAX_INSN 1200 + +struct brw_compile { + struct brw_instruction store[BRW_EU_MAX_INSN]; + unsigned nr_insn; + + /* Allow clients to push/pop instruction state: + */ + struct brw_instruction stack[BRW_EU_MAX_INSN_STACK]; + struct brw_instruction *current; + + unsigned flag_value; + boolean single_program_flow; +}; + + + +static __inline int type_sz( unsigned type ) +{ + switch( type ) { + case BRW_REGISTER_TYPE_UD: + case BRW_REGISTER_TYPE_D: + case BRW_REGISTER_TYPE_F: + return 4; + case BRW_REGISTER_TYPE_HF: + case BRW_REGISTER_TYPE_UW: + case BRW_REGISTER_TYPE_W: + return 2; + case BRW_REGISTER_TYPE_UB: + case BRW_REGISTER_TYPE_B: + return 1; + default: + return 0; + } +} + +static __inline struct brw_reg brw_reg( unsigned file, + unsigned nr, + unsigned subnr, + unsigned type, + unsigned vstride, + unsigned width, + unsigned hstride, + unsigned swizzle, + unsigned writemask) +{ + + struct brw_reg reg; + reg.type = type; + reg.file = file; + reg.nr = nr; + reg.subnr = subnr * type_sz(type); + reg.negate = 0; + reg.abs = 0; + reg.vstride = vstride; + reg.width = width; + reg.hstride = hstride; + reg.address_mode = BRW_ADDRESS_DIRECT; + reg.pad0 = 0; + + /* Could do better: If the reg is r5.3<0;1,0>, we probably want to + * set swizzle and writemask to W, as the lower bits of subnr will + * be lost when converted to align16. This is probably too much to + * keep track of as you'd want it adjusted by suboffset(), etc. + * Perhaps fix up when converting to align16? + */ + reg.dw1.bits.swizzle = swizzle; + reg.dw1.bits.writemask = writemask; + reg.dw1.bits.indirect_offset = 0; + reg.dw1.bits.pad1 = 0; + return reg; +} + +static __inline struct brw_reg brw_vec16_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_16, + BRW_WIDTH_16, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYZW, + TGSI_WRITEMASK_XYZW); +} + +static __inline struct brw_reg brw_vec8_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_8, + BRW_WIDTH_8, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYZW, + TGSI_WRITEMASK_XYZW); +} + + +static __inline struct brw_reg brw_vec4_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_4, + BRW_WIDTH_4, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYZW, + TGSI_WRITEMASK_XYZW); +} + + +static __inline struct brw_reg brw_vec2_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_2, + BRW_WIDTH_2, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XYXY, + TGSI_WRITEMASK_XY); +} + +static __inline struct brw_reg brw_vec1_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return brw_reg(file, + nr, + subnr, + BRW_REGISTER_TYPE_F, + BRW_VERTICAL_STRIDE_0, + BRW_WIDTH_1, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XXXX, + TGSI_WRITEMASK_X); +} + + +static __inline struct brw_reg retype( struct brw_reg reg, + unsigned type ) +{ + reg.type = type; + return reg; +} + +static __inline struct brw_reg suboffset( struct brw_reg reg, + unsigned delta ) +{ + reg.subnr += delta * type_sz(reg.type); + return reg; +} + + +static __inline struct brw_reg offset( struct brw_reg reg, + unsigned delta ) +{ + reg.nr += delta; + return reg; +} + + +static __inline struct brw_reg byte_offset( struct brw_reg reg, + unsigned bytes ) +{ + unsigned newoffset = reg.nr * REG_SIZE + reg.subnr + bytes; + reg.nr = newoffset / REG_SIZE; + reg.subnr = newoffset % REG_SIZE; + return reg; +} + + +static __inline struct brw_reg brw_uw16_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return suboffset(retype(brw_vec16_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); +} + +static __inline struct brw_reg brw_uw8_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return suboffset(retype(brw_vec8_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); +} + +static __inline struct brw_reg brw_uw1_reg( unsigned file, + unsigned nr, + unsigned subnr ) +{ + return suboffset(retype(brw_vec1_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); +} + +static __inline struct brw_reg brw_imm_reg( unsigned type ) +{ + return brw_reg( BRW_IMMEDIATE_VALUE, + 0, + 0, + type, + BRW_VERTICAL_STRIDE_0, + BRW_WIDTH_1, + BRW_HORIZONTAL_STRIDE_0, + 0, + 0); +} + +static __inline struct brw_reg brw_imm_f( float f ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_F); + imm.dw1.f = f; + return imm; +} + +static __inline struct brw_reg brw_imm_d( int d ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_D); + imm.dw1.d = d; + return imm; +} + +static __inline struct brw_reg brw_imm_ud( unsigned ud ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UD); + imm.dw1.ud = ud; + return imm; +} + +static __inline struct brw_reg brw_imm_uw( ushort uw ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UW); + imm.dw1.ud = uw; + return imm; +} + +static __inline struct brw_reg brw_imm_w( short w ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_W); + imm.dw1.d = w; + return imm; +} + +/* brw_imm_b and brw_imm_ub aren't supported by hardware - the type + * numbers alias with _V and _VF below: + */ + +/* Vector of eight signed half-byte values: + */ +static __inline struct brw_reg brw_imm_v( unsigned v ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_V); + imm.vstride = BRW_VERTICAL_STRIDE_0; + imm.width = BRW_WIDTH_8; + imm.hstride = BRW_HORIZONTAL_STRIDE_1; + imm.dw1.ud = v; + return imm; +} + +/* Vector of four 8-bit float values: + */ +static __inline struct brw_reg brw_imm_vf( unsigned v ) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); + imm.vstride = BRW_VERTICAL_STRIDE_0; + imm.width = BRW_WIDTH_4; + imm.hstride = BRW_HORIZONTAL_STRIDE_1; + imm.dw1.ud = v; + return imm; +} + +#define VF_ZERO 0x0 +#define VF_ONE 0x30 +#define VF_NEG (1<<7) + +static __inline struct brw_reg brw_imm_vf4( unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3) +{ + struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); + imm.vstride = BRW_VERTICAL_STRIDE_0; + imm.width = BRW_WIDTH_4; + imm.hstride = BRW_HORIZONTAL_STRIDE_1; + imm.dw1.ud = ((v0 << 0) | + (v1 << 8) | + (v2 << 16) | + (v3 << 24)); + return imm; +} + + +static __inline struct brw_reg brw_address( struct brw_reg reg ) +{ + return brw_imm_uw(reg.nr * REG_SIZE + reg.subnr); +} + + +static __inline struct brw_reg brw_vec1_grf( unsigned nr, + unsigned subnr ) +{ + return brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +static __inline struct brw_reg brw_vec8_grf( unsigned nr, + unsigned subnr ) +{ + return brw_vec8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +static __inline struct brw_reg brw_vec4_grf( unsigned nr, + unsigned subnr ) +{ + return brw_vec4_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + + +static __inline struct brw_reg brw_vec2_grf( unsigned nr, + unsigned subnr ) +{ + return brw_vec2_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +static __inline struct brw_reg brw_uw8_grf( unsigned nr, + unsigned subnr ) +{ + return brw_uw8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); +} + +static __inline struct brw_reg brw_null_reg( void ) +{ + return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_NULL, + 0); +} + +static __inline struct brw_reg brw_address_reg( unsigned subnr ) +{ + return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_ADDRESS, + subnr); +} + +/* If/else instructions break in align16 mode if writemask & swizzle + * aren't xyzw. This goes against the convention for other scalar + * regs: + */ +static __inline struct brw_reg brw_ip_reg( void ) +{ + return brw_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_IP, + 0, + BRW_REGISTER_TYPE_UD, + BRW_VERTICAL_STRIDE_4, /* ? */ + BRW_WIDTH_1, + BRW_HORIZONTAL_STRIDE_0, + BRW_SWIZZLE_XYZW, /* NOTE! */ + TGSI_WRITEMASK_XYZW); /* NOTE! */ +} + +static __inline struct brw_reg brw_acc_reg( void ) +{ + return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_ACCUMULATOR, + 0); +} + + +static __inline struct brw_reg brw_flag_reg( void ) +{ + return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_FLAG, + 0); +} + + +static __inline struct brw_reg brw_mask_reg( unsigned subnr ) +{ + return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, + BRW_ARF_MASK, + subnr); +} + +static __inline struct brw_reg brw_message_reg( unsigned nr ) +{ + return brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, + nr, + 0); +} + + + + +/* This is almost always called with a numeric constant argument, so + * make things easy to evaluate at compile time: + */ +static __inline unsigned cvt( unsigned val ) +{ + switch (val) { + case 0: return 0; + case 1: return 1; + case 2: return 2; + case 4: return 3; + case 8: return 4; + case 16: return 5; + case 32: return 6; + } + return 0; +} + +static __inline struct brw_reg stride( struct brw_reg reg, + unsigned vstride, + unsigned width, + unsigned hstride ) +{ + + reg.vstride = cvt(vstride); + reg.width = cvt(width) - 1; + reg.hstride = cvt(hstride); + return reg; +} + +static __inline struct brw_reg vec16( struct brw_reg reg ) +{ + return stride(reg, 16,16,1); +} + +static __inline struct brw_reg vec8( struct brw_reg reg ) +{ + return stride(reg, 8,8,1); +} + +static __inline struct brw_reg vec4( struct brw_reg reg ) +{ + return stride(reg, 4,4,1); +} + +static __inline struct brw_reg vec2( struct brw_reg reg ) +{ + return stride(reg, 2,2,1); +} + +static __inline struct brw_reg vec1( struct brw_reg reg ) +{ + return stride(reg, 0,1,0); +} + +static __inline struct brw_reg get_element( struct brw_reg reg, unsigned elt ) +{ + return vec1(suboffset(reg, elt)); +} + +static __inline struct brw_reg get_element_ud( struct brw_reg reg, unsigned elt ) +{ + return vec1(suboffset(retype(reg, BRW_REGISTER_TYPE_UD), elt)); +} + + +static __inline struct brw_reg brw_swizzle( struct brw_reg reg, + unsigned x, + unsigned y, + unsigned z, + unsigned w) +{ + reg.dw1.bits.swizzle = BRW_SWIZZLE4(BRW_GET_SWZ(reg.dw1.bits.swizzle, x), + BRW_GET_SWZ(reg.dw1.bits.swizzle, y), + BRW_GET_SWZ(reg.dw1.bits.swizzle, z), + BRW_GET_SWZ(reg.dw1.bits.swizzle, w)); + return reg; +} + + +static __inline struct brw_reg brw_swizzle1( struct brw_reg reg, + unsigned x ) +{ + return brw_swizzle(reg, x, x, x, x); +} + +static __inline struct brw_reg brw_writemask( struct brw_reg reg, + unsigned mask ) +{ + reg.dw1.bits.writemask &= mask; + return reg; +} + +static __inline struct brw_reg brw_set_writemask( struct brw_reg reg, + unsigned mask ) +{ + reg.dw1.bits.writemask = mask; + return reg; +} + +static __inline struct brw_reg negate( struct brw_reg reg ) +{ + reg.negate ^= 1; + return reg; +} + +static __inline struct brw_reg brw_abs( struct brw_reg reg ) +{ + reg.abs = 1; + return reg; +} + +/*********************************************************************** + */ +static __inline struct brw_reg brw_vec4_indirect( unsigned subnr, + int offset ) +{ + struct brw_reg reg = brw_vec4_grf(0, 0); + reg.subnr = subnr; + reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; + reg.dw1.bits.indirect_offset = offset; + return reg; +} + +static __inline struct brw_reg brw_vec1_indirect( unsigned subnr, + int offset ) +{ + struct brw_reg reg = brw_vec1_grf(0, 0); + reg.subnr = subnr; + reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; + reg.dw1.bits.indirect_offset = offset; + return reg; +} + +static __inline struct brw_reg deref_4f(struct brw_indirect ptr, int offset) +{ + return brw_vec4_indirect(ptr.addr_subnr, ptr.addr_offset + offset); +} + +static __inline struct brw_reg deref_1f(struct brw_indirect ptr, int offset) +{ + return brw_vec1_indirect(ptr.addr_subnr, ptr.addr_offset + offset); +} + +static __inline struct brw_reg deref_4b(struct brw_indirect ptr, int offset) +{ + return retype(deref_4f(ptr, offset), BRW_REGISTER_TYPE_B); +} + +static __inline struct brw_reg deref_1uw(struct brw_indirect ptr, int offset) +{ + return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UW); +} + +static __inline struct brw_reg deref_1ud(struct brw_indirect ptr, int offset) +{ + return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UD); +} + +static __inline struct brw_reg get_addr_reg(struct brw_indirect ptr) +{ + return brw_address_reg(ptr.addr_subnr); +} + +static __inline struct brw_indirect brw_indirect_offset( struct brw_indirect ptr, int offset ) +{ + ptr.addr_offset += offset; + return ptr; +} + +static __inline struct brw_indirect brw_indirect( unsigned addr_subnr, int offset ) +{ + struct brw_indirect ptr; + ptr.addr_subnr = addr_subnr; + ptr.addr_offset = offset; + ptr.pad = 0; + return ptr; +} + +static __inline struct brw_instruction *current_insn( struct brw_compile *p) +{ + return &p->store[p->nr_insn]; +} + +void brw_pop_insn_state( struct brw_compile *p ); +void brw_push_insn_state( struct brw_compile *p ); +void brw_set_mask_control( struct brw_compile *p, unsigned value ); +void brw_set_saturate( struct brw_compile *p, unsigned value ); +void brw_set_access_mode( struct brw_compile *p, unsigned access_mode ); +void brw_set_compression_control( struct brw_compile *p, boolean control ); +void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value ); +void brw_set_predicate_control( struct brw_compile *p, unsigned pc ); +void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional ); + +void brw_init_compile( struct brw_compile *p ); +const unsigned *brw_get_program( struct brw_compile *p, unsigned *sz ); + + +struct brw_instruction *brw_alu1( struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src ); + +struct brw_instruction *brw_alu2(struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1 ); + +/* Helpers for regular instructions: + */ +#define ALU1(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0); + +#define ALU2(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0, \ + struct brw_reg src1); + +ALU1(MOV) +ALU2(SEL) +ALU1(NOT) +ALU2(AND) +ALU2(OR) +ALU2(XOR) +ALU2(SHR) +ALU2(SHL) +ALU2(RSR) +ALU2(RSL) +ALU2(ASR) +ALU2(JMPI) +ALU2(ADD) +ALU2(MUL) +ALU1(FRC) +ALU1(RNDD) +ALU2(MAC) +ALU2(MACH) +ALU1(LZD) +ALU2(DP4) +ALU2(DPH) +ALU2(DP3) +ALU2(DP2) +ALU2(LINE) + +#undef ALU1 +#undef ALU2 + + + +/* Helpers for SEND instruction: + */ +void brw_urb_WRITE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + boolean allocate, + boolean used, + unsigned msg_length, + unsigned response_length, + boolean eot, + boolean writes_complete, + unsigned offset, + unsigned swizzle); + +void brw_fb_WRITE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + unsigned binding_table_index, + unsigned msg_length, + unsigned response_length, + boolean eot); + +void brw_SAMPLE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + unsigned binding_table_index, + unsigned sampler, + unsigned writemask, + unsigned msg_type, + unsigned response_length, + unsigned msg_length, + boolean eot); + +void brw_math_16( struct brw_compile *p, + struct brw_reg dest, + unsigned function, + unsigned saturate, + unsigned msg_reg_nr, + struct brw_reg src, + unsigned precision ); + +void brw_math( struct brw_compile *p, + struct brw_reg dest, + unsigned function, + unsigned saturate, + unsigned msg_reg_nr, + struct brw_reg src, + unsigned data_type, + unsigned precision ); + +void brw_dp_READ_16( struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + unsigned scratch_offset ); + +void brw_dp_WRITE_16( struct brw_compile *p, + struct brw_reg src, + unsigned msg_reg_nr, + unsigned scratch_offset ); + +/* If/else/endif. Works by manipulating the execution flags on each + * channel. + */ +struct brw_instruction *brw_IF(struct brw_compile *p, + unsigned execute_size); + +struct brw_instruction *brw_ELSE(struct brw_compile *p, + struct brw_instruction *if_insn); + +void brw_ENDIF(struct brw_compile *p, + struct brw_instruction *if_or_else_insn); + + +/* DO/WHILE loops: + */ +struct brw_instruction *brw_DO(struct brw_compile *p, + unsigned execute_size); + +struct brw_instruction *brw_WHILE(struct brw_compile *p, + struct brw_instruction *patch_insn); + +struct brw_instruction *brw_BREAK(struct brw_compile *p); +struct brw_instruction *brw_CONT(struct brw_compile *p); +/* Forward jumps: + */ +void brw_land_fwd_jump(struct brw_compile *p, + struct brw_instruction *jmp_insn); + + + +void brw_NOP(struct brw_compile *p); + +/* Special case: there is never a destination, execution size will be + * taken from src0: + */ +void brw_CMP(struct brw_compile *p, + struct brw_reg dest, + unsigned conditional, + struct brw_reg src0, + struct brw_reg src1); + +void brw_print_reg( struct brw_reg reg ); + + +/*********************************************************************** + * brw_eu_util.c: + */ + +void brw_copy_indirect_to_indirect(struct brw_compile *p, + struct brw_indirect dst_ptr, + struct brw_indirect src_ptr, + unsigned count); + +void brw_copy_from_indirect(struct brw_compile *p, + struct brw_reg dst, + struct brw_indirect ptr, + unsigned count); + +void brw_copy4(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + unsigned count); + +void brw_copy8(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + unsigned count); + +void brw_math_invert( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src); + +void brw_set_src1( struct brw_instruction *insn, + struct brw_reg reg ); +#endif diff --git a/src/gallium/drivers/i965simple/brw_eu_debug.c b/src/gallium/drivers/i965simple/brw_eu_debug.c new file mode 100644 index 0000000000..4a94ddefa6 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_eu_debug.c @@ -0,0 +1,90 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "pipe/p_debug.h" + +#include "brw_eu.h" + +void brw_print_reg( struct brw_reg hwreg ) +{ + static const char *file[] = { + "arf", + "grf", + "msg", + "imm" + }; + + static const char *type[] = { + "ud", + "d", + "uw", + "w", + "ub", + "vf", + "hf", + "f" + }; + + debug_printf("%s%s", + hwreg.abs ? "abs/" : "", + hwreg.negate ? "-" : ""); + + if (hwreg.file == BRW_GENERAL_REGISTER_FILE && + hwreg.nr % 2 == 0 && + hwreg.subnr == 0 && + hwreg.vstride == BRW_VERTICAL_STRIDE_8 && + hwreg.width == BRW_WIDTH_8 && + hwreg.hstride == BRW_HORIZONTAL_STRIDE_1 && + hwreg.type == BRW_REGISTER_TYPE_F) { + debug_printf("vec%d", hwreg.nr); + } + else if (hwreg.file == BRW_GENERAL_REGISTER_FILE && + hwreg.vstride == BRW_VERTICAL_STRIDE_0 && + hwreg.width == BRW_WIDTH_1 && + hwreg.hstride == BRW_HORIZONTAL_STRIDE_0 && + hwreg.type == BRW_REGISTER_TYPE_F) { + debug_printf("scl%d.%d", hwreg.nr, hwreg.subnr / 4); + } + else { + debug_printf("%s%d.%d<%d;%d,%d>:%s", + file[hwreg.file], + hwreg.nr, + hwreg.subnr / type_sz(hwreg.type), + hwreg.vstride ? (1<<(hwreg.vstride-1)) : 0, + 1< + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" + + + + +/*********************************************************************** + * Internal helper for constructing instructions + */ + +static void guess_execution_size( struct brw_instruction *insn, + struct brw_reg reg ) +{ + if (reg.width == BRW_WIDTH_8 && + insn->header.compression_control == BRW_COMPRESSION_COMPRESSED) + insn->header.execution_size = BRW_EXECUTE_16; + else + insn->header.execution_size = reg.width; /* note - definitions are compatible */ +} + + +static void brw_set_dest( struct brw_instruction *insn, + struct brw_reg dest ) +{ + insn->bits1.da1.dest_reg_file = dest.file; + insn->bits1.da1.dest_reg_type = dest.type; + insn->bits1.da1.dest_address_mode = dest.address_mode; + + if (dest.address_mode == BRW_ADDRESS_DIRECT) { + insn->bits1.da1.dest_reg_nr = dest.nr; + + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits1.da1.dest_subreg_nr = dest.subnr; + insn->bits1.da1.dest_horiz_stride = BRW_HORIZONTAL_STRIDE_1; + } + else { + insn->bits1.da16.dest_subreg_nr = dest.subnr / 16; + insn->bits1.da16.dest_writemask = dest.dw1.bits.writemask; + } + } + else { + insn->bits1.ia1.dest_subreg_nr = dest.subnr; + + /* These are different sizes in align1 vs align16: + */ + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits1.ia1.dest_indirect_offset = dest.dw1.bits.indirect_offset; + insn->bits1.ia1.dest_horiz_stride = BRW_HORIZONTAL_STRIDE_1; + } + else { + insn->bits1.ia16.dest_indirect_offset = dest.dw1.bits.indirect_offset; + } + } + + /* NEW: Set the execution size based on dest.width and + * insn->compression_control: + */ + guess_execution_size(insn, dest); +} + +static void brw_set_src0( struct brw_instruction *insn, + struct brw_reg reg ) +{ + assert(reg.file != BRW_MESSAGE_REGISTER_FILE); + + insn->bits1.da1.src0_reg_file = reg.file; + insn->bits1.da1.src0_reg_type = reg.type; + insn->bits2.da1.src0_abs = reg.abs; + insn->bits2.da1.src0_negate = reg.negate; + insn->bits2.da1.src0_address_mode = reg.address_mode; + + if (reg.file == BRW_IMMEDIATE_VALUE) { + insn->bits3.ud = reg.dw1.ud; + + /* Required to set some fields in src1 as well: + */ + insn->bits1.da1.src1_reg_file = 0; /* arf */ + insn->bits1.da1.src1_reg_type = reg.type; + } + else + { + if (reg.address_mode == BRW_ADDRESS_DIRECT) { + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits2.da1.src0_subreg_nr = reg.subnr; + insn->bits2.da1.src0_reg_nr = reg.nr; + } + else { + insn->bits2.da16.src0_subreg_nr = reg.subnr / 16; + insn->bits2.da16.src0_reg_nr = reg.nr; + } + } + else { + insn->bits2.ia1.src0_subreg_nr = reg.subnr; + + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits2.ia1.src0_indirect_offset = reg.dw1.bits.indirect_offset; + } + else { + insn->bits2.ia16.src0_subreg_nr = reg.dw1.bits.indirect_offset; + } + } + + if (insn->header.access_mode == BRW_ALIGN_1) { + if (reg.width == BRW_WIDTH_1 && + insn->header.execution_size == BRW_EXECUTE_1) { + insn->bits2.da1.src0_horiz_stride = BRW_HORIZONTAL_STRIDE_0; + insn->bits2.da1.src0_width = BRW_WIDTH_1; + insn->bits2.da1.src0_vert_stride = BRW_VERTICAL_STRIDE_0; + } + else { + insn->bits2.da1.src0_horiz_stride = reg.hstride; + insn->bits2.da1.src0_width = reg.width; + insn->bits2.da1.src0_vert_stride = reg.vstride; + } + } + else { + insn->bits2.da16.src0_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); + insn->bits2.da16.src0_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); + insn->bits2.da16.src0_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); + insn->bits2.da16.src0_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); + + /* This is an oddity of the fact we're using the same + * descriptions for registers in align_16 as align_1: + */ + if (reg.vstride == BRW_VERTICAL_STRIDE_8) + insn->bits2.da16.src0_vert_stride = BRW_VERTICAL_STRIDE_4; + else + insn->bits2.da16.src0_vert_stride = reg.vstride; + } + } +} + + +void brw_set_src1( struct brw_instruction *insn, + struct brw_reg reg ) +{ + assert(reg.file != BRW_MESSAGE_REGISTER_FILE); + + insn->bits1.da1.src1_reg_file = reg.file; + insn->bits1.da1.src1_reg_type = reg.type; + insn->bits3.da1.src1_abs = reg.abs; + insn->bits3.da1.src1_negate = reg.negate; + + /* Only src1 can be immediate in two-argument instructions. + */ + assert(insn->bits1.da1.src0_reg_file != BRW_IMMEDIATE_VALUE); + + if (reg.file == BRW_IMMEDIATE_VALUE) { + insn->bits3.ud = reg.dw1.ud; + } + else { + /* This is a hardware restriction, which may or may not be lifted + * in the future: + */ + assert (reg.address_mode == BRW_ADDRESS_DIRECT); + //assert (reg.file == BRW_GENERAL_REGISTER_FILE); + + if (insn->header.access_mode == BRW_ALIGN_1) { + insn->bits3.da1.src1_subreg_nr = reg.subnr; + insn->bits3.da1.src1_reg_nr = reg.nr; + } + else { + insn->bits3.da16.src1_subreg_nr = reg.subnr / 16; + insn->bits3.da16.src1_reg_nr = reg.nr; + } + + if (insn->header.access_mode == BRW_ALIGN_1) { + if (reg.width == BRW_WIDTH_1 && + insn->header.execution_size == BRW_EXECUTE_1) { + insn->bits3.da1.src1_horiz_stride = BRW_HORIZONTAL_STRIDE_0; + insn->bits3.da1.src1_width = BRW_WIDTH_1; + insn->bits3.da1.src1_vert_stride = BRW_VERTICAL_STRIDE_0; + } + else { + insn->bits3.da1.src1_horiz_stride = reg.hstride; + insn->bits3.da1.src1_width = reg.width; + insn->bits3.da1.src1_vert_stride = reg.vstride; + } + } + else { + insn->bits3.da16.src1_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); + insn->bits3.da16.src1_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); + insn->bits3.da16.src1_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); + insn->bits3.da16.src1_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); + + /* This is an oddity of the fact we're using the same + * descriptions for registers in align_16 as align_1: + */ + if (reg.vstride == BRW_VERTICAL_STRIDE_8) + insn->bits3.da16.src1_vert_stride = BRW_VERTICAL_STRIDE_4; + else + insn->bits3.da16.src1_vert_stride = reg.vstride; + } + } +} + + + +static void brw_set_math_message( struct brw_instruction *insn, + unsigned msg_length, + unsigned response_length, + unsigned function, + unsigned integer_type, + boolean low_precision, + boolean saturate, + unsigned dataType ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.math.function = function; + insn->bits3.math.int_type = integer_type; + insn->bits3.math.precision = low_precision; + insn->bits3.math.saturate = saturate; + insn->bits3.math.data_type = dataType; + insn->bits3.math.response_length = response_length; + insn->bits3.math.msg_length = msg_length; + insn->bits3.math.msg_target = BRW_MESSAGE_TARGET_MATH; + insn->bits3.math.end_of_thread = 0; +} + +static void brw_set_urb_message( struct brw_instruction *insn, + boolean allocate, + boolean used, + unsigned msg_length, + unsigned response_length, + boolean end_of_thread, + boolean complete, + unsigned offset, + unsigned swizzle_control ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.urb.opcode = 0; /* ? */ + insn->bits3.urb.offset = offset; + insn->bits3.urb.swizzle_control = swizzle_control; + insn->bits3.urb.allocate = allocate; + insn->bits3.urb.used = used; /* ? */ + insn->bits3.urb.complete = complete; + insn->bits3.urb.response_length = response_length; + insn->bits3.urb.msg_length = msg_length; + insn->bits3.urb.msg_target = BRW_MESSAGE_TARGET_URB; + insn->bits3.urb.end_of_thread = end_of_thread; +} + +static void brw_set_dp_write_message( struct brw_instruction *insn, + unsigned binding_table_index, + unsigned msg_control, + unsigned msg_type, + unsigned msg_length, + unsigned pixel_scoreboard_clear, + unsigned response_length, + unsigned end_of_thread ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.dp_write.binding_table_index = binding_table_index; + insn->bits3.dp_write.msg_control = msg_control; + insn->bits3.dp_write.pixel_scoreboard_clear = pixel_scoreboard_clear; + insn->bits3.dp_write.msg_type = msg_type; + insn->bits3.dp_write.send_commit_msg = 0; + insn->bits3.dp_write.response_length = response_length; + insn->bits3.dp_write.msg_length = msg_length; + insn->bits3.dp_write.msg_target = BRW_MESSAGE_TARGET_DATAPORT_WRITE; + insn->bits3.urb.end_of_thread = end_of_thread; +} + +static void brw_set_dp_read_message( struct brw_instruction *insn, + unsigned binding_table_index, + unsigned msg_control, + unsigned msg_type, + unsigned target_cache, + unsigned msg_length, + unsigned response_length, + unsigned end_of_thread ) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.dp_read.binding_table_index = binding_table_index; + insn->bits3.dp_read.msg_control = msg_control; + insn->bits3.dp_read.msg_type = msg_type; + insn->bits3.dp_read.target_cache = target_cache; + insn->bits3.dp_read.response_length = response_length; + insn->bits3.dp_read.msg_length = msg_length; + insn->bits3.dp_read.msg_target = BRW_MESSAGE_TARGET_DATAPORT_READ; + insn->bits3.dp_read.end_of_thread = end_of_thread; +} + +static void brw_set_sampler_message( struct brw_instruction *insn, + unsigned binding_table_index, + unsigned sampler, + unsigned msg_type, + unsigned response_length, + unsigned msg_length, + boolean eot) +{ + brw_set_src1(insn, brw_imm_d(0)); + + insn->bits3.sampler.binding_table_index = binding_table_index; + insn->bits3.sampler.sampler = sampler; + insn->bits3.sampler.msg_type = msg_type; + insn->bits3.sampler.return_format = BRW_SAMPLER_RETURN_FORMAT_FLOAT32; + insn->bits3.sampler.response_length = response_length; + insn->bits3.sampler.msg_length = msg_length; + insn->bits3.sampler.end_of_thread = eot; + insn->bits3.sampler.msg_target = BRW_MESSAGE_TARGET_SAMPLER; +} + + + +static struct brw_instruction *next_insn( struct brw_compile *p, + unsigned opcode ) +{ + struct brw_instruction *insn; + + assert(p->nr_insn + 1 < BRW_EU_MAX_INSN); + + insn = &p->store[p->nr_insn++]; + memcpy(insn, p->current, sizeof(*insn)); + + /* Reset this one-shot flag: + */ + + if (p->current->header.destreg__conditonalmod) { + p->current->header.destreg__conditonalmod = 0; + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + } + + insn->header.opcode = opcode; + return insn; +} + + +struct brw_instruction *brw_alu1( struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src ) +{ + struct brw_instruction *insn = next_insn(p, opcode); + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + return insn; +} + +struct brw_instruction *brw_alu2(struct brw_compile *p, + unsigned opcode, + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1 ) +{ + struct brw_instruction *insn = next_insn(p, opcode); + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, src1); + return insn; +} + + +/*********************************************************************** + * Convenience routines. + */ +#define ALU1(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0) \ +{ \ + return brw_alu1(p, BRW_OPCODE_##OP, dest, src0); \ +} + +#define ALU2(OP) \ +struct brw_instruction *brw_##OP(struct brw_compile *p, \ + struct brw_reg dest, \ + struct brw_reg src0, \ + struct brw_reg src1) \ +{ \ + return brw_alu2(p, BRW_OPCODE_##OP, dest, src0, src1); \ +} + + +ALU1(MOV) +ALU2(SEL) +ALU1(NOT) +ALU2(AND) +ALU2(OR) +ALU2(XOR) +ALU2(SHR) +ALU2(SHL) +ALU2(RSR) +ALU2(RSL) +ALU2(ASR) +ALU2(ADD) +ALU2(MUL) +ALU1(FRC) +ALU1(RNDD) +ALU2(MAC) +ALU2(MACH) +ALU1(LZD) +ALU2(DP4) +ALU2(DPH) +ALU2(DP3) +ALU2(DP2) +ALU2(LINE) + + + + +void brw_NOP(struct brw_compile *p) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_NOP); + brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src1(insn, brw_imm_ud(0x0)); +} + + + + + +/*********************************************************************** + * Comparisons, if/else/endif + */ + +struct brw_instruction *brw_JMPI(struct brw_compile *p, + struct brw_reg dest, + struct brw_reg src0, + struct brw_reg src1) +{ + struct brw_instruction *insn = brw_alu2(p, BRW_OPCODE_JMPI, dest, src0, src1); + + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + return insn; +} + +/* EU takes the value from the flag register and pushes it onto some + * sort of a stack (presumably merging with any flag value already on + * the stack). Within an if block, the flags at the top of the stack + * control execution on each channel of the unit, eg. on each of the + * 16 pixel values in our wm programs. + * + * When the matching 'else' instruction is reached (presumably by + * countdown of the instruction count patched in by our ELSE/ENDIF + * functions), the relevent flags are inverted. + * + * When the matching 'endif' instruction is reached, the flags are + * popped off. If the stack is now empty, normal execution resumes. + * + * No attempt is made to deal with stack overflow (14 elements?). + */ +struct brw_instruction *brw_IF(struct brw_compile *p, unsigned execute_size) +{ + struct brw_instruction *insn; + + if (p->single_program_flow) { + assert(execute_size == BRW_EXECUTE_1); + + insn = next_insn(p, BRW_OPCODE_ADD); + insn->header.predicate_inverse = 1; + } else { + insn = next_insn(p, BRW_OPCODE_IF); + } + + /* Override the defaults for this instruction: + */ + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.execution_size = execute_size; + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.predicate_control = BRW_PREDICATE_NORMAL; + insn->header.mask_control = BRW_MASK_ENABLE; + + p->current->header.predicate_control = BRW_PREDICATE_NONE; + + return insn; +} + + +struct brw_instruction *brw_ELSE(struct brw_compile *p, + struct brw_instruction *if_insn) +{ + struct brw_instruction *insn; + + if (p->single_program_flow) { + insn = next_insn(p, BRW_OPCODE_ADD); + } else { + insn = next_insn(p, BRW_OPCODE_ELSE); + } + + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = if_insn->header.execution_size; + insn->header.mask_control = BRW_MASK_ENABLE; + + /* Patch the if instruction to point at this instruction. + */ + if (p->single_program_flow) { + assert(if_insn->header.opcode == BRW_OPCODE_ADD); + + if_insn->bits3.ud = (insn - if_insn + 1) * 16; + } else { + assert(if_insn->header.opcode == BRW_OPCODE_IF); + + if_insn->bits3.if_else.jump_count = insn - if_insn; + if_insn->bits3.if_else.pop_count = 1; + if_insn->bits3.if_else.pad0 = 0; + } + + return insn; +} + +void brw_ENDIF(struct brw_compile *p, + struct brw_instruction *patch_insn) +{ + if (p->single_program_flow) { + /* In single program flow mode, there's no need to execute an ENDIF, + * since we don't need to do any stack operations, and if we're executing + * currently, we want to just continue executing. + */ + struct brw_instruction *next = &p->store[p->nr_insn]; + + assert(patch_insn->header.opcode == BRW_OPCODE_ADD); + + patch_insn->bits3.ud = (next - patch_insn) * 16; + } else { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_ENDIF); + + brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = patch_insn->header.execution_size; + insn->header.mask_control = BRW_MASK_ENABLE; + + assert(patch_insn->bits3.if_else.jump_count == 0); + + /* Patch the if or else instructions to point at this or the next + * instruction respectively. + */ + if (patch_insn->header.opcode == BRW_OPCODE_IF) { + /* Automagically turn it into an IFF: + */ + patch_insn->header.opcode = BRW_OPCODE_IFF; + patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1; + patch_insn->bits3.if_else.pop_count = 0; + patch_insn->bits3.if_else.pad0 = 0; + } else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) { + patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1; + patch_insn->bits3.if_else.pop_count = 1; + patch_insn->bits3.if_else.pad0 = 0; + } else { + assert(0); + } + + /* Also pop item off the stack in the endif instruction: + */ + insn->bits3.if_else.jump_count = 0; + insn->bits3.if_else.pop_count = 1; + insn->bits3.if_else.pad0 = 0; + } +} + +struct brw_instruction *brw_BREAK(struct brw_compile *p) +{ + struct brw_instruction *insn; + insn = next_insn(p, BRW_OPCODE_BREAK); + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = BRW_EXECUTE_8; + insn->header.mask_control = BRW_MASK_DISABLE; + insn->bits3.if_else.pad0 = 0; + return insn; +} + +struct brw_instruction *brw_CONT(struct brw_compile *p) +{ + struct brw_instruction *insn; + insn = next_insn(p, BRW_OPCODE_CONTINUE); + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = BRW_EXECUTE_8; + insn->header.mask_control = BRW_MASK_DISABLE; + insn->bits3.if_else.pad0 = 0; + return insn; +} + +/* DO/WHILE loop: + */ +struct brw_instruction *brw_DO(struct brw_compile *p, unsigned execute_size) +{ + if (p->single_program_flow) { + return &p->store[p->nr_insn]; + } else { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_DO); + + /* Override the defaults for this instruction: + */ + brw_set_dest(insn, brw_null_reg()); + brw_set_src0(insn, brw_null_reg()); + brw_set_src1(insn, brw_null_reg()); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.execution_size = execute_size; + insn->header.predicate_control = BRW_PREDICATE_NONE; + /* insn->header.mask_control = BRW_MASK_ENABLE; */ + insn->header.mask_control = BRW_MASK_DISABLE; + + return insn; + } +} + + + +struct brw_instruction *brw_WHILE(struct brw_compile *p, + struct brw_instruction *do_insn) +{ + struct brw_instruction *insn; + + if (p->single_program_flow) + insn = next_insn(p, BRW_OPCODE_ADD); + else + insn = next_insn(p, BRW_OPCODE_WHILE); + + brw_set_dest(insn, brw_ip_reg()); + brw_set_src0(insn, brw_ip_reg()); + brw_set_src1(insn, brw_imm_d(0x0)); + + insn->header.compression_control = BRW_COMPRESSION_NONE; + + if (p->single_program_flow) { + insn->header.execution_size = BRW_EXECUTE_1; + + insn->bits3.d = (do_insn - insn) * 16; + } else { + insn->header.execution_size = do_insn->header.execution_size; + + assert(do_insn->header.opcode == BRW_OPCODE_DO); + insn->bits3.if_else.jump_count = do_insn - insn; + insn->bits3.if_else.pop_count = 0; + insn->bits3.if_else.pad0 = 0; + } + +/* insn->header.mask_control = BRW_MASK_ENABLE; */ + + insn->header.mask_control = BRW_MASK_DISABLE; + p->current->header.predicate_control = BRW_PREDICATE_NONE; + return insn; +} + + +/* FORWARD JUMPS: + */ +void brw_land_fwd_jump(struct brw_compile *p, + struct brw_instruction *jmp_insn) +{ + struct brw_instruction *landing = &p->store[p->nr_insn]; + + assert(jmp_insn->header.opcode == BRW_OPCODE_JMPI); + assert(jmp_insn->bits1.da1.src1_reg_file = BRW_IMMEDIATE_VALUE); + + jmp_insn->bits3.ud = (landing - jmp_insn) - 1; +} + + + +/* To integrate with the above, it makes sense that the comparison + * instruction should populate the flag register. It might be simpler + * just to use the flag reg for most WM tasks? + */ +void brw_CMP(struct brw_compile *p, + struct brw_reg dest, + unsigned conditional, + struct brw_reg src0, + struct brw_reg src1) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_CMP); + + insn->header.destreg__conditonalmod = conditional; + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, src1); + +/* guess_execution_size(insn, src0); */ + + + /* Make it so that future instructions will use the computed flag + * value until brw_set_predicate_control_flag_value() is called + * again. + */ + if (dest.file == BRW_ARCHITECTURE_REGISTER_FILE && + dest.nr == 0) { + p->current->header.predicate_control = BRW_PREDICATE_NORMAL; + p->flag_value = 0xff; + } +} + + + +/*********************************************************************** + * Helpers for the various SEND message types: + */ + +/* Invert 8 values + */ +void brw_math( struct brw_compile *p, + struct brw_reg dest, + unsigned function, + unsigned saturate, + unsigned msg_reg_nr, + struct brw_reg src, + unsigned data_type, + unsigned precision ) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + unsigned msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1; + unsigned response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1; + + /* Example code doesn't set predicate_control for send + * instructions. + */ + insn->header.predicate_control = 0; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + brw_set_math_message(insn, + msg_length, response_length, + function, + BRW_MATH_INTEGER_UNSIGNED, + precision, + saturate, + data_type); +} + +/* Use 2 send instructions to invert 16 elements + */ +void brw_math_16( struct brw_compile *p, + struct brw_reg dest, + unsigned function, + unsigned saturate, + unsigned msg_reg_nr, + struct brw_reg src, + unsigned precision ) +{ + struct brw_instruction *insn; + unsigned msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1; + unsigned response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1; + + /* First instruction: + */ + brw_push_insn_state(p); + brw_set_predicate_control_flag_value(p, 0xff); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + insn = next_insn(p, BRW_OPCODE_SEND); + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + brw_set_math_message(insn, + msg_length, response_length, + function, + BRW_MATH_INTEGER_UNSIGNED, + precision, + saturate, + BRW_MATH_DATA_VECTOR); + + /* Second instruction: + */ + insn = next_insn(p, BRW_OPCODE_SEND); + insn->header.compression_control = BRW_COMPRESSION_2NDHALF; + insn->header.destreg__conditonalmod = msg_reg_nr+1; + + brw_set_dest(insn, offset(dest,1)); + brw_set_src0(insn, src); + brw_set_math_message(insn, + msg_length, response_length, + function, + BRW_MATH_INTEGER_UNSIGNED, + precision, + saturate, + BRW_MATH_DATA_VECTOR); + + brw_pop_insn_state(p); +} + + + + +void brw_dp_WRITE_16( struct brw_compile *p, + struct brw_reg src, + unsigned msg_reg_nr, + unsigned scratch_offset ) +{ + { + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + brw_MOV(p, + retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D), + brw_imm_d(scratch_offset)); + + brw_pop_insn_state(p); + } + + { + unsigned msg_length = 3; + struct brw_reg dest = retype(brw_null_reg(), BRW_REGISTER_TYPE_UW); + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src); + + brw_set_dp_write_message(insn, + 255, /* bti */ + BRW_DATAPORT_OWORD_BLOCK_4_OWORDS, /* msg_control */ + BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE, /* msg_type */ + msg_length, + 0, /* pixel scoreboard */ + 0, /* response_length */ + 0); /* eot */ + } + +} + + +void brw_dp_READ_16( struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + unsigned scratch_offset ) +{ + { + brw_push_insn_state(p); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_mask_control(p, BRW_MASK_DISABLE); + + brw_MOV(p, + retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D), + brw_imm_d(scratch_offset)); + + brw_pop_insn_state(p); + } + + { + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); /* UW? */ + brw_set_src0(insn, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW)); + + brw_set_dp_read_message(insn, + 255, /* bti */ + 3, /* msg_control */ + BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ + 1, /* target cache */ + 1, /* msg_length */ + 2, /* response_length */ + 0); /* eot */ + } +} + + +void brw_fb_WRITE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + unsigned binding_table_index, + unsigned msg_length, + unsigned response_length, + boolean eot) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + insn->header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_dp_write_message(insn, + binding_table_index, + BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE, /* msg_control */ + BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE, /* msg_type */ + msg_length, + 1, /* pixel scoreboard */ + response_length, + eot); +} + + + +void brw_SAMPLE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + unsigned binding_table_index, + unsigned sampler, + unsigned writemask, + unsigned msg_type, + unsigned response_length, + unsigned msg_length, + boolean eot) +{ + boolean need_stall = 0; + + if(writemask == 0) { +/* debug_printf("%s: zero writemask??\n", __FUNCTION__); */ + return; + } + + /* Hardware doesn't do destination dependency checking on send + * instructions properly. Add a workaround which generates the + * dependency by other means. In practice it seems like this bug + * only crops up for texture samples, and only where registers are + * written by the send and then written again later without being + * read in between. Luckily for us, we already track that + * information and use it to modify the writemask for the + * instruction, so that is a guide for whether a workaround is + * needed. + */ + if (writemask != TGSI_WRITEMASK_XYZW) { + unsigned dst_offset = 0; + unsigned i, newmask = 0, len = 0; + + for (i = 0; i < 4; i++) { + if (writemask & (1<header.predicate_control = 0; /* XXX */ + insn->header.compression_control = BRW_COMPRESSION_NONE; + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_sampler_message(insn, + binding_table_index, + sampler, + msg_type, + response_length, + msg_length, + eot); + } + + if (need_stall) + { + struct brw_reg reg = vec8(offset(dest, response_length-1)); + + /* mov (8) r9.0<1>:f r9.0<8;8,1>:f { Align1 } + */ + brw_push_insn_state(p); + brw_set_compression_control(p, FALSE); + brw_MOV(p, reg, reg); + brw_pop_insn_state(p); + } + +} + +/* All these variables are pretty confusing - we might be better off + * using bitmasks and macros for this, in the old style. Or perhaps + * just having the caller instantiate the fields in dword3 itself. + */ +void brw_urb_WRITE(struct brw_compile *p, + struct brw_reg dest, + unsigned msg_reg_nr, + struct brw_reg src0, + boolean allocate, + boolean used, + unsigned msg_length, + unsigned response_length, + boolean eot, + boolean writes_complete, + unsigned offset, + unsigned swizzle) +{ + struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); + + assert(msg_length < 16); + + brw_set_dest(insn, dest); + brw_set_src0(insn, src0); + brw_set_src1(insn, brw_imm_d(0)); + + insn->header.destreg__conditonalmod = msg_reg_nr; + + brw_set_urb_message(insn, + allocate, + used, + msg_length, + response_length, + eot, + writes_complete, + offset, + swizzle); +} + diff --git a/src/gallium/drivers/i965simple/brw_eu_util.c b/src/gallium/drivers/i965simple/brw_eu_util.c new file mode 100644 index 0000000000..3a65b141f0 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_eu_util.c @@ -0,0 +1,126 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_eu.h" + + +void brw_math_invert( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src) +{ + brw_math( p, + dst, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 0, + src, + BRW_MATH_PRECISION_FULL, + BRW_MATH_DATA_VECTOR ); +} + + + +void brw_copy4(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + unsigned count) +{ + unsigned i; + + dst = vec4(dst); + src = vec4(src); + + for (i = 0; i < count; i++) + { + unsigned delta = i*32; + brw_MOV(p, byte_offset(dst, delta), byte_offset(src, delta)); + brw_MOV(p, byte_offset(dst, delta+16), byte_offset(src, delta+16)); + } +} + + +void brw_copy8(struct brw_compile *p, + struct brw_reg dst, + struct brw_reg src, + unsigned count) +{ + unsigned i; + + dst = vec8(dst); + src = vec8(src); + + for (i = 0; i < count; i++) + { + unsigned delta = i*32; + brw_MOV(p, byte_offset(dst, delta), byte_offset(src, delta)); + } +} + + +void brw_copy_indirect_to_indirect(struct brw_compile *p, + struct brw_indirect dst_ptr, + struct brw_indirect src_ptr, + unsigned count) +{ + unsigned i; + + for (i = 0; i < count; i++) + { + unsigned delta = i*32; + brw_MOV(p, deref_4f(dst_ptr, delta), deref_4f(src_ptr, delta)); + brw_MOV(p, deref_4f(dst_ptr, delta+16), deref_4f(src_ptr, delta+16)); + } +} + + +void brw_copy_from_indirect(struct brw_compile *p, + struct brw_reg dst, + struct brw_indirect ptr, + unsigned count) +{ + unsigned i; + + dst = vec4(dst); + + for (i = 0; i < count; i++) + { + unsigned delta = i*32; + brw_MOV(p, byte_offset(dst, delta), deref_4f(ptr, delta)); + brw_MOV(p, byte_offset(dst, delta+16), deref_4f(ptr, delta+16)); + } +} + + + + diff --git a/src/gallium/drivers/i965simple/brw_flush.c b/src/gallium/drivers/i965simple/brw_flush.c new file mode 100644 index 0000000000..5216c680cf --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_flush.c @@ -0,0 +1,80 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_batch.h" + + +/** + * In future we may want a fence-like interface instead of finish. + */ +static void brw_flush( struct pipe_context *pipe, + unsigned flags ) +{ + struct brw_context *brw = brw_context(pipe); + struct pipe_fence_handle *fence; + + /* Do we need to emit an MI_FLUSH command to flush the hardware + * caches? + */ + if (flags & (PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE)) { + struct brw_mi_flush flush; + + memset(&flush, 0, sizeof(flush)); + flush.opcode = CMD_MI_FLUSH; + + if (!(flags & PIPE_FLUSH_RENDER_CACHE)) + flush.flags |= BRW_INHIBIT_FLUSH_RENDER_CACHE; + + if (flags & PIPE_FLUSH_TEXTURE_CACHE) + flush.flags |= BRW_FLUSH_READ_CACHE; + + BRW_BATCH_STRUCT(brw, &flush); + } + + /* If there are no flags, just flush pending commands to hardware: + */ + FLUSH_BATCH( &fence ); + + if (flags & PIPE_FLUSH_WAIT) { +// brw->winsys->wait_fence(brw->winsys, fence); + } +} + + + +void brw_init_flush_functions( struct brw_context *brw ) +{ + brw->pipe.flush = brw_flush; +} diff --git a/src/gallium/drivers/i965simple/brw_gs.c b/src/gallium/drivers/i965simple/brw_gs.c new file mode 100644 index 0000000000..de60868ccc --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_gs.c @@ -0,0 +1,196 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_state.h" +#include "brw_gs.h" + + + +static void compile_gs_prog( struct brw_context *brw, + struct brw_gs_prog_key *key ) +{ + struct brw_gs_compile c; + const unsigned *program; + unsigned program_size; + + memset(&c, 0, sizeof(c)); + + c.key = *key; + + /* Need to locate the two positions present in vertex + header. + * These are currently hardcoded: + */ + c.nr_attrs = brw_count_bits(c.key.attrs); + c.nr_regs = (c.nr_attrs + 1) / 2 + 1; /* are vertices packed, or reg-aligned? */ + c.nr_bytes = c.nr_regs * REG_SIZE; + + + /* Begin the compilation: + */ + brw_init_compile(&c.func); + + c.func.single_program_flow = 1; + + /* For some reason the thread is spawned with only 4 channels + * unmasked. + */ + brw_set_mask_control(&c.func, BRW_MASK_DISABLE); + + + /* Note that primitives which don't require a GS program have + * already been weeded out by this stage: + */ + switch (key->primitive) { + case PIPE_PRIM_QUADS: + brw_gs_quads( &c ); + break; + case PIPE_PRIM_QUAD_STRIP: + brw_gs_quad_strip( &c ); + break; + case PIPE_PRIM_LINE_LOOP: + brw_gs_lines( &c ); + break; + case PIPE_PRIM_LINES: + if (key->hint_gs_always) + brw_gs_lines( &c ); + else { + return; + } + break; + case PIPE_PRIM_TRIANGLES: + if (key->hint_gs_always) + brw_gs_tris( &c ); + else { + return; + } + break; + case PIPE_PRIM_POINTS: + if (key->hint_gs_always) + brw_gs_points( &c ); + else { + return; + } + break; + default: + return; + } + + /* get the program + */ + program = brw_get_program(&c.func, &program_size); + + /* Upload + */ + brw->gs.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_GS_PROG], + &c.key, + sizeof(c.key), + program, + program_size, + &c.prog_data, + &brw->gs.prog_data ); +} + + +static boolean search_cache( struct brw_context *brw, + struct brw_gs_prog_key *key ) +{ + return brw_search_cache(&brw->cache[BRW_GS_PROG], + key, sizeof(*key), + &brw->gs.prog_data, + &brw->gs.prog_gs_offset); +} + + +static const int gs_prim[PIPE_PRIM_POLYGON+1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINE_LOOP, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_QUADS, + PIPE_PRIM_QUAD_STRIP, + PIPE_PRIM_TRIANGLES +}; + +static void populate_key( struct brw_context *brw, + struct brw_gs_prog_key *key ) +{ + memset(key, 0, sizeof(*key)); + + /* CACHE_NEW_VS_PROG */ + key->attrs = brw->vs.prog_data->outputs_written; + + /* BRW_NEW_PRIMITIVE */ + key->primitive = gs_prim[brw->primitive]; + + key->hint_gs_always = 0; /* debug code? */ + + key->need_gs_prog = (key->hint_gs_always || + brw->primitive == PIPE_PRIM_QUADS || + brw->primitive == PIPE_PRIM_QUAD_STRIP || + brw->primitive == PIPE_PRIM_LINE_LOOP); +} + +/* Calculate interpolants for triangle and line rasterization. + */ +static void upload_gs_prog( struct brw_context *brw ) +{ + struct brw_gs_prog_key key; + + /* Populate the key: + */ + populate_key(brw, &key); + + if (brw->gs.prog_active != key.need_gs_prog) { + brw->state.dirty.cache |= CACHE_NEW_GS_PROG; + brw->gs.prog_active = key.need_gs_prog; + } + + if (brw->gs.prog_active) { + if (!search_cache(brw, &key)) + compile_gs_prog( brw, &key ); + } +} + + +const struct brw_tracked_state brw_gs_prog = { + .dirty = { + .brw = BRW_NEW_PRIMITIVE, + .cache = CACHE_NEW_VS_PROG + }, + .update = upload_gs_prog +}; diff --git a/src/gallium/drivers/i965simple/brw_gs.h b/src/gallium/drivers/i965simple/brw_gs.h new file mode 100644 index 0000000000..f09141c6aa --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_gs.h @@ -0,0 +1,75 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_GS_H +#define BRW_GS_H + + +#include "brw_context.h" +#include "brw_eu.h" + +#define MAX_GS_VERTS (4) + +struct brw_gs_prog_key { + unsigned attrs:32; + unsigned primitive:4; + unsigned hint_gs_always:1; + unsigned need_gs_prog:1; + unsigned pad:26; +}; + +struct brw_gs_compile { + struct brw_compile func; + struct brw_gs_prog_key key; + struct brw_gs_prog_data prog_data; + + struct { + struct brw_reg R0; + struct brw_reg vertex[MAX_GS_VERTS]; + } reg; + + /* 3 different ways of expressing vertex size: + */ + unsigned nr_attrs; + unsigned nr_regs; + unsigned nr_bytes; +}; + +#define ATTR_SIZE (4*4) + +void brw_gs_quads( struct brw_gs_compile *c ); +void brw_gs_quad_strip( struct brw_gs_compile *c ); +void brw_gs_tris( struct brw_gs_compile *c ); +void brw_gs_lines( struct brw_gs_compile *c ); +void brw_gs_points( struct brw_gs_compile *c ); + +#endif diff --git a/src/gallium/drivers/i965simple/brw_gs_emit.c b/src/gallium/drivers/i965simple/brw_gs_emit.c new file mode 100644 index 0000000000..c3cc90b10f --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_gs_emit.c @@ -0,0 +1,148 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_gs.h" + +static void brw_gs_alloc_regs( struct brw_gs_compile *c, + unsigned nr_verts ) +{ + unsigned i = 0,j; + + /* Register usage is static, precompute here: + */ + c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; + + /* Payload vertices plus space for more generated vertices: + */ + for (j = 0; j < nr_verts; j++) { + c->reg.vertex[j] = brw_vec4_grf(i, 0); + i += c->nr_regs; + } + + c->prog_data.urb_read_length = c->nr_regs; + c->prog_data.total_grf = i; +} + + +static void brw_gs_emit_vue(struct brw_gs_compile *c, + struct brw_reg vert, + boolean last, + unsigned header) +{ + struct brw_compile *p = &c->func; + boolean allocate = !last; + + /* Overwrite PrimType and PrimStart in the message header, for + * each vertex in turn: + */ + brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header)); + + /* Copy the vertex from vertn into m1..mN+1: + */ + brw_copy8(p, brw_message_reg(1), vert, c->nr_regs); + + /* Send each vertex as a seperate write to the urb. This is + * different to the concept in brw_sf_emit.c, where subsequent + * writes are used to build up a single urb entry. Each of these + * writes instantiates a seperate urb entry, and a new one must be + * allocated each time. + */ + brw_urb_WRITE(p, + allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), + 0, + c->reg.R0, + allocate, + 1, /* used */ + c->nr_regs + 1, /* msg length */ + allocate ? 1 : 0, /* response length */ + allocate ? 0 : 1, /* eot */ + 1, /* writes_complete */ + 0, /* urb offset */ + BRW_URB_SWIZZLE_NONE); +} + + + +void brw_gs_quads( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 4); + + /* Use polygons for correct edgeflag behaviour. Note that vertex 3 + * is the PV for quads, but vertex 0 for polygons: + */ + brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); +} + +void brw_gs_quad_strip( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 4); + + brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); +} + +void brw_gs_tris( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 3); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_TRILIST << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_TRILIST << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_TRILIST << 2) | R02_PRIM_END)); +} + +void brw_gs_lines( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 2); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_END)); +} + +void brw_gs_points( struct brw_gs_compile *c ) +{ + brw_gs_alloc_regs(c, 1); + brw_gs_emit_vue(c, c->reg.vertex[0], 1, ((_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END)); +} + + + + + + + + diff --git a/src/gallium/drivers/i965simple/brw_gs_state.c b/src/gallium/drivers/i965simple/brw_gs_state.c new file mode 100644 index 0000000000..3932e9e939 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_gs_state.c @@ -0,0 +1,89 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "pipe/p_util.h" + + + +static void upload_gs_unit( struct brw_context *brw ) +{ + struct brw_gs_unit_state gs; + + memset(&gs, 0, sizeof(gs)); + + /* CACHE_NEW_GS_PROG */ + if (brw->gs.prog_active) { + gs.thread0.grf_reg_count = + align(brw->gs.prog_data->total_grf, 16) / 16 - 1; + gs.thread0.kernel_start_pointer = brw->gs.prog_gs_offset >> 6; + gs.thread3.urb_entry_read_length = brw->gs.prog_data->urb_read_length; + } + else { + gs.thread0.grf_reg_count = 0; + gs.thread0.kernel_start_pointer = 0; + gs.thread3.urb_entry_read_length = 1; + } + + /* BRW_NEW_URB_FENCE */ + gs.thread4.nr_urb_entries = brw->urb.nr_gs_entries; + gs.thread4.urb_entry_allocation_size = brw->urb.vsize - 1; + + gs.thread4.max_threads = 0; /* Hardware requirement */ + + if (BRW_DEBUG & DEBUG_STATS) + gs.thread4.stats_enable = 1; + + /* CONSTANT */ + gs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + gs.thread1.single_program_flow = 1; + gs.thread3.dispatch_grf_start_reg = 1; + gs.thread3.const_urb_entry_read_offset = 0; + gs.thread3.const_urb_entry_read_length = 0; + gs.thread3.urb_entry_read_offset = 0; + + + brw->gs.state_gs_offset = brw_cache_data( &brw->cache[BRW_GS_UNIT], &gs ); +} + + +const struct brw_tracked_state brw_gs_unit = { + .dirty = { + .brw = (BRW_NEW_CURBE_OFFSETS | + BRW_NEW_URB_FENCE), + .cache = CACHE_NEW_GS_PROG + }, + .update = upload_gs_unit +}; diff --git a/src/gallium/drivers/i965simple/brw_misc_state.c b/src/gallium/drivers/i965simple/brw_misc_state.c new file mode 100644 index 0000000000..925049ecc1 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_misc_state.c @@ -0,0 +1,486 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_batch.h" +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + + + + + +/*********************************************************************** + * Blend color + */ + +static void upload_blend_constant_color(struct brw_context *brw) +{ + struct brw_blend_constant_color bcc; + + memset(&bcc, 0, sizeof(bcc)); + bcc.header.opcode = CMD_BLEND_CONSTANT_COLOR; + bcc.header.length = sizeof(bcc)/4-2; + bcc.blend_constant_color[0] = brw->attribs.BlendColor.color[0]; + bcc.blend_constant_color[1] = brw->attribs.BlendColor.color[1]; + bcc.blend_constant_color[2] = brw->attribs.BlendColor.color[2]; + bcc.blend_constant_color[3] = brw->attribs.BlendColor.color[3]; + + BRW_CACHED_BATCH_STRUCT(brw, &bcc); +} + + +const struct brw_tracked_state brw_blend_constant_color = { + .dirty = { + .brw = BRW_NEW_BLEND, + .cache = 0 + }, + .update = upload_blend_constant_color +}; + + +/*********************************************************************** + * Drawing rectangle + */ +static void upload_drawing_rect(struct brw_context *brw) +{ + struct brw_drawrect bdr; + + memset(&bdr, 0, sizeof(bdr)); + bdr.header.opcode = CMD_DRAW_RECT; + bdr.header.length = sizeof(bdr)/4 - 2; + bdr.xmin = 0; + bdr.ymin = 0; + bdr.xmax = brw->attribs.FrameBuffer.cbufs[0]->width; + bdr.ymax = brw->attribs.FrameBuffer.cbufs[0]->height; + bdr.xorg = 0; + bdr.yorg = 0; + + /* Can't use BRW_CACHED_BATCH_STRUCT because this is also emitted + * uncached in brw_draw.c: + */ + BRW_BATCH_STRUCT(brw, &bdr); +} + +const struct brw_tracked_state brw_drawing_rect = { + .dirty = { + .brw = BRW_NEW_SCENE, + .cache = 0 + }, + .update = upload_drawing_rect +}; + +/** + * Upload the binding table pointers, which point each stage's array of surface + * state pointers. + * + * The binding table pointers are relative to the surface state base address, + * which is the BRW_SS_POOL cache buffer. + */ +static void upload_binding_table_pointers(struct brw_context *brw) +{ + struct brw_binding_table_pointers btp; + memset(&btp, 0, sizeof(btp)); + + btp.header.opcode = CMD_BINDING_TABLE_PTRS; + btp.header.length = sizeof(btp)/4 - 2; + btp.vs = 0; + btp.gs = 0; + btp.clp = 0; + btp.sf = 0; + btp.wm = brw->wm.bind_ss_offset; + + BRW_CACHED_BATCH_STRUCT(brw, &btp); +} + +const struct brw_tracked_state brw_binding_table_pointers = { + .dirty = { + .brw = 0, + .cache = CACHE_NEW_SURF_BIND + }, + .update = upload_binding_table_pointers, +}; + + +/** + * Upload pointers to the per-stage state. + * + * The state pointers in this packet are all relative to the general state + * base address set by CMD_STATE_BASE_ADDRESS, which is the BRW_GS_POOL buffer. + */ +static void upload_pipelined_state_pointers(struct brw_context *brw ) +{ + struct brw_pipelined_state_pointers psp; + memset(&psp, 0, sizeof(psp)); + + psp.header.opcode = CMD_PIPELINED_STATE_POINTERS; + psp.header.length = sizeof(psp)/4 - 2; + + psp.vs.offset = brw->vs.state_gs_offset >> 5; + psp.sf.offset = brw->sf.state_gs_offset >> 5; + psp.wm.offset = brw->wm.state_gs_offset >> 5; + psp.cc.offset = brw->cc.state_gs_offset >> 5; + + /* GS gets turned on and off regularly. Need to re-emit URB fence + * after this occurs. + */ + if (brw->gs.prog_active) { + psp.gs.offset = brw->gs.state_gs_offset >> 5; + psp.gs.enable = 1; + } + + if (0) { + psp.clp.offset = brw->clip.state_gs_offset >> 5; + psp.clp.enable = 1; + } + + + if (BRW_CACHED_BATCH_STRUCT(brw, &psp)) + brw->state.dirty.brw |= BRW_NEW_PSP; +} + +const struct brw_tracked_state brw_pipelined_state_pointers = { + .dirty = { + .brw = 0, + .cache = (CACHE_NEW_VS_UNIT | + CACHE_NEW_GS_UNIT | + CACHE_NEW_GS_PROG | + CACHE_NEW_CLIP_UNIT | + CACHE_NEW_SF_UNIT | + CACHE_NEW_WM_UNIT | + CACHE_NEW_CC_UNIT) + }, + .update = upload_pipelined_state_pointers +}; + +static void upload_psp_urb_cbs(struct brw_context *brw ) +{ + upload_pipelined_state_pointers(brw); + brw_upload_urb_fence(brw); + brw_upload_constant_buffer_state(brw); +} + + +const struct brw_tracked_state brw_psp_urb_cbs = { + .dirty = { + .brw = BRW_NEW_URB_FENCE, + .cache = (CACHE_NEW_VS_UNIT | + CACHE_NEW_GS_UNIT | + CACHE_NEW_GS_PROG | + CACHE_NEW_CLIP_UNIT | + CACHE_NEW_SF_UNIT | + CACHE_NEW_WM_UNIT | + CACHE_NEW_CC_UNIT) + }, + .update = upload_psp_urb_cbs +}; + +/** + * Upload the depthbuffer offset and format. + * + * We have to do this per state validation as we need to emit the relocation + * in the batch buffer. + */ +static void upload_depthbuffer(struct brw_context *brw) +{ + struct pipe_surface *depth_surface = brw->attribs.FrameBuffer.zsbuf; + + BEGIN_BATCH(5, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (5 - 2)); + if (depth_surface == NULL) { + OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) | + (BRW_SURFACE_NULL << 29)); + OUT_BATCH(0); + OUT_BATCH(0); + OUT_BATCH(0); + } else { + unsigned int format; + + switch (depth_surface->cpp) { + case 2: + format = BRW_DEPTHFORMAT_D16_UNORM; + break; + case 4: + if (depth_surface->format == PIPE_FORMAT_Z32_FLOAT) + format = BRW_DEPTHFORMAT_D32_FLOAT; + else + format = BRW_DEPTHFORMAT_D24_UNORM_S8_UINT; + break; + default: + assert(0); + return; + } + + OUT_BATCH(((depth_surface->pitch * depth_surface->cpp) - 1) | + (format << 18) | + (BRW_TILEWALK_YMAJOR << 26) | +// (depth_surface->region->tiled << 27) | + (BRW_SURFACE_2D << 29)); + OUT_RELOC(depth_surface->buffer, + PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE, 0); + OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) | + ((depth_surface->pitch - 1) << 6) | + ((depth_surface->height - 1) << 19)); + OUT_BATCH(0); + } + ADVANCE_BATCH(); +} + +const struct brw_tracked_state brw_depthbuffer = { + .dirty = { + .brw = BRW_NEW_SCENE, + .cache = 0 + }, + .update = upload_depthbuffer, +}; + + + + +/*********************************************************************** + * Polygon stipple packet + */ + +static void upload_polygon_stipple(struct brw_context *brw) +{ + struct brw_polygon_stipple bps; + unsigned i; + + memset(&bps, 0, sizeof(bps)); + bps.header.opcode = CMD_POLY_STIPPLE_PATTERN; + bps.header.length = sizeof(bps)/4-2; + + /* XXX: state tracker should send *all* state down initially! + */ + if (brw->attribs.PolygonStipple) + for (i = 0; i < 32; i++) + bps.stipple[i] = brw->attribs.PolygonStipple->stipple[31 - i]; /* invert */ + + BRW_CACHED_BATCH_STRUCT(brw, &bps); +} + +const struct brw_tracked_state brw_polygon_stipple = { + .dirty = { + .brw = BRW_NEW_STIPPLE, + .cache = 0 + }, + .update = upload_polygon_stipple +}; + + +/*********************************************************************** + * Line stipple packet + */ + +static void upload_line_stipple(struct brw_context *brw) +{ + struct brw_line_stipple bls; + float tmp; + int tmpi; + + memset(&bls, 0, sizeof(bls)); + bls.header.opcode = CMD_LINE_STIPPLE_PATTERN; + bls.header.length = sizeof(bls)/4 - 2; + + bls.bits0.pattern = brw->attribs.Raster->line_stipple_pattern; + bls.bits1.repeat_count = brw->attribs.Raster->line_stipple_factor; + + tmp = 1.0 / (float) brw->attribs.Raster->line_stipple_factor; + tmpi = tmp * (1<<13); + + + bls.bits1.inverse_repeat_count = tmpi; + + BRW_CACHED_BATCH_STRUCT(brw, &bls); +} + +const struct brw_tracked_state brw_line_stipple = { + .dirty = { + .brw = BRW_NEW_STIPPLE, + .cache = 0 + }, + .update = upload_line_stipple +}; + + +/*********************************************************************** + * Misc constant state packets + */ + +static void upload_pipe_control(struct brw_context *brw) +{ + struct brw_pipe_control pc; + + return; + + memset(&pc, 0, sizeof(pc)); + + pc.header.opcode = CMD_PIPE_CONTROL; + pc.header.length = sizeof(pc)/4 - 2; + pc.header.post_sync_operation = PIPE_CONTROL_NOWRITE; + + pc.header.instruction_state_cache_flush_enable = 1; + + pc.bits1.dest_addr_type = PIPE_CONTROL_GTTWRITE_GLOBAL; + + BRW_BATCH_STRUCT(brw, &pc); +} + +const struct brw_tracked_state brw_pipe_control = { + .dirty = { + .brw = BRW_NEW_SCENE, + .cache = 0 + }, + .update = upload_pipe_control +}; + + +/*********************************************************************** + * Misc invarient state packets + */ + +static void upload_invarient_state( struct brw_context *brw ) +{ + { + struct brw_mi_flush flush; + + memset(&flush, 0, sizeof(flush)); + flush.opcode = CMD_MI_FLUSH; + flush.flags = BRW_FLUSH_STATE_CACHE | BRW_FLUSH_READ_CACHE; + BRW_BATCH_STRUCT(brw, &flush); + } + + { + /* 0x61040000 Pipeline Select */ + /* PipelineSelect : 0 */ + struct brw_pipeline_select ps; + + memset(&ps, 0, sizeof(ps)); + ps.header.opcode = CMD_PIPELINE_SELECT; + ps.header.pipeline_select = 0; + BRW_BATCH_STRUCT(brw, &ps); + } + + { + struct brw_global_depth_offset_clamp gdo; + memset(&gdo, 0, sizeof(gdo)); + + /* Disable depth offset clamping. + */ + gdo.header.opcode = CMD_GLOBAL_DEPTH_OFFSET_CLAMP; + gdo.header.length = sizeof(gdo)/4 - 2; + gdo.depth_offset_clamp = 0.0; + + BRW_BATCH_STRUCT(brw, &gdo); + } + + + /* 0x61020000 State Instruction Pointer */ + { + struct brw_system_instruction_pointer sip; + memset(&sip, 0, sizeof(sip)); + + sip.header.opcode = CMD_STATE_INSN_POINTER; + sip.header.length = 0; + sip.bits0.pad = 0; + sip.bits0.system_instruction_pointer = 0; + BRW_BATCH_STRUCT(brw, &sip); + } + + + { + struct brw_vf_statistics vfs; + memset(&vfs, 0, sizeof(vfs)); + + vfs.opcode = CMD_VF_STATISTICS; + if (BRW_DEBUG & DEBUG_STATS) + vfs.statistics_enable = 1; + + BRW_BATCH_STRUCT(brw, &vfs); + } + + + { + struct brw_polygon_stipple_offset bpso; + + memset(&bpso, 0, sizeof(bpso)); + bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET; + bpso.header.length = sizeof(bpso)/4-2; + bpso.bits0.x_offset = 0; + bpso.bits0.y_offset = 0; + + BRW_BATCH_STRUCT(brw, &bpso); + } +} + +const struct brw_tracked_state brw_invarient_state = { + .dirty = { + .brw = BRW_NEW_SCENE, + .cache = 0 + }, + .update = upload_invarient_state +}; + +/** + * Define the base addresses which some state is referenced from. + * + * This allows us to avoid having to emit relocations in many places for + * cached state, and instead emit pointers inside of large, mostly-static + * state pools. This comes at the expense of memory, and more expensive cache + * misses. + */ +static void upload_state_base_address( struct brw_context *brw ) +{ + /* Output the structure (brw_state_base_address) directly to the + * batchbuffer, so we can emit relocations inline. + */ + BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2)); + OUT_RELOC(brw->pool[BRW_GS_POOL].buffer, + PIPE_BUFFER_USAGE_GPU_READ, + 1); /* General state base address */ + OUT_RELOC(brw->pool[BRW_SS_POOL].buffer, + PIPE_BUFFER_USAGE_GPU_READ, + 1); /* Surface state base address */ + OUT_BATCH(1); /* Indirect object base address */ + OUT_BATCH(1); /* General state upper bound */ + OUT_BATCH(1); /* Indirect object upper bound */ + ADVANCE_BATCH(); +} + + +const struct brw_tracked_state brw_state_base_address = { + .dirty = { + .brw = BRW_NEW_SCENE, + .cache = 0 + }, + .update = upload_state_base_address +}; diff --git a/src/gallium/drivers/i965simple/brw_reg.h b/src/gallium/drivers/i965simple/brw_reg.h new file mode 100644 index 0000000000..9e885c3b3b --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_reg.h @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#define CMD_MI (0x0 << 29) +#define CMD_2D (0x2 << 29) +#define CMD_3D (0x3 << 29) + +#define MI_BATCH_BUFFER_END (CMD_MI | 0xA << 23) + +/* Stalls command execution waiting for the given events to have occurred. */ +#define MI_WAIT_FOR_EVENT (CMD_MI | (0x3 << 23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +/* Primitive dispatch on 830-945 */ +#define _3DPRIMITIVE (CMD_3D | (0x1f << 24)) +#define PRIM_INDIRECT (1<<23) +#define PRIM_INLINE (0<<23) +#define PRIM_INDIRECT_SEQUENTIAL (0<<17) +#define PRIM_INDIRECT_ELTS (1<<17) + +#define PRIM3D_TRILIST (0x0<<18) +#define PRIM3D_TRISTRIP (0x1<<18) +#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) +#define PRIM3D_TRIFAN (0x3<<18) +#define PRIM3D_POLY (0x4<<18) +#define PRIM3D_LINELIST (0x5<<18) +#define PRIM3D_LINESTRIP (0x6<<18) +#define PRIM3D_RECTLIST (0x7<<18) +#define PRIM3D_POINTLIST (0x8<<18) +#define PRIM3D_DIB (0x9<<18) +#define PRIM3D_MASK (0x1f<<18) + +#define XY_SETUP_BLT_CMD (CMD_2D | (0x01 << 22) | 6) + +#define XY_COLOR_BLT_CMD (CMD_2D | (0x50 << 22) | 4) + +#define XY_SRC_COPY_BLT_CMD (CMD_2D | (0x53 << 22) | 6) + +/* BR00 */ +#define XY_BLT_WRITE_ALPHA (1 << 21) +#define XY_BLT_WRITE_RGB (1 << 20) +#define XY_SRC_TILED (1 << 15) +#define XY_DST_TILED (1 << 11) + +/* BR13 */ +#define BR13_565 (0x1 << 24) +#define BR13_8888 (0x3 << 24) + +#define FENCE_LINEAR 0 +#define FENCE_XMAJOR 1 +#define FENCE_YMAJOR 2 diff --git a/src/gallium/drivers/i965simple/brw_sf.c b/src/gallium/drivers/i965simple/brw_sf.c new file mode 100644 index 0000000000..7c83b81c85 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_sf.c @@ -0,0 +1,351 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_sf.h" +#include "brw_state.h" +#include "tgsi/util/tgsi_parse.h" + + +static void compile_sf_prog( struct brw_context *brw, + struct brw_sf_prog_key *key ) +{ + struct brw_sf_compile c; + const unsigned *program; + unsigned program_size; + + memset(&c, 0, sizeof(c)); + + /* Begin the compilation: + */ + brw_init_compile(&c.func); + + c.key = *key; + + + c.nr_attrs = c.key.vp_output_count; + c.nr_attr_regs = (c.nr_attrs+1)/2; + + c.nr_setup_attrs = c.key.fp_input_count + 1; /* +1 for position */ + c.nr_setup_regs = (c.nr_setup_attrs+1)/2; + + c.prog_data.urb_read_length = c.nr_attr_regs; + c.prog_data.urb_entry_size = c.nr_setup_regs * 2; + + + /* Which primitive? Or all three? + */ + switch (key->primitive) { + case SF_TRIANGLES: + c.nr_verts = 3; + brw_emit_tri_setup( &c ); + break; + case SF_LINES: + c.nr_verts = 2; + brw_emit_line_setup( &c ); + break; + case SF_POINTS: + c.nr_verts = 1; + brw_emit_point_setup( &c ); + break; + + case SF_UNFILLED_TRIS: + default: + assert(0); + return; + } + + + + /* get the program + */ + program = brw_get_program(&c.func, &program_size); + + /* Upload + */ + brw->sf.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_SF_PROG], + &c.key, + sizeof(c.key), + program, + program_size, + &c.prog_data, + &brw->sf.prog_data ); +} + + +static boolean search_cache( struct brw_context *brw, + struct brw_sf_prog_key *key ) +{ + return brw_search_cache(&brw->cache[BRW_SF_PROG], + key, sizeof(*key), + &brw->sf.prog_data, + &brw->sf.prog_gs_offset); +} + + +/* Calculate interpolants for triangle and line rasterization. + */ +static void upload_sf_prog( struct brw_context *brw ) +{ + const struct brw_fragment_program *fs = brw->attribs.FragmentProgram; + struct brw_sf_prog_key key; + struct tgsi_parse_context parse; + int i, done = 0; + + + memset(&key, 0, sizeof(key)); + + /* Populate the key, noting state dependencies: + */ + /* CACHE_NEW_VS_PROG */ + key.vp_output_count = brw->vs.prog_data->outputs_written; + + /* BRW_NEW_FS */ + key.fp_input_count = brw->attribs.FragmentProgram->info.nr_regs[TGSI_FILE_INPUT]; + + + /* BRW_NEW_REDUCED_PRIMITIVE */ + switch (brw->reduced_primitive) { + case PIPE_PRIM_TRIANGLES: +// if (key.attrs & (1<program.tokens ); + while( !done && + !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) + { + int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + int interp_mode = parse.FullToken.FullDeclaration.Interpolation.Interpolate; + //int semantic = parse.FullToken.FullDeclaration.Semantic.SemanticName; + //int semantic_index = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; + + debug_printf("fs input %d..%d interp mode %d\n", first, last, interp_mode); + + switch (interp_mode) { + case TGSI_INTERPOLATE_CONSTANT: + for (i = first; i <= last; i++) + key.const_mask |= (1 << i); + break; + case TGSI_INTERPOLATE_LINEAR: + for (i = first; i <= last; i++) + key.linear_mask |= (1 << i); + break; + case TGSI_INTERPOLATE_PERSPECTIVE: + for (i = first; i <= last; i++) + key.persp_mask |= (1 << i); + break; + default: + break; + } + + /* Also need stuff for flat shading, twosided color. + */ + + } + break; + default: + done = 1; + break; + } + } + + /* Hack: Adjust for position. Optimize away when not required (ie + * for perspective interpolation). + */ + key.persp_mask <<= 1; + key.linear_mask <<= 1; + key.linear_mask |= 1; + key.const_mask <<= 1; + + debug_printf("key.persp_mask: %x\n", key.persp_mask); + debug_printf("key.linear_mask: %x\n", key.linear_mask); + debug_printf("key.const_mask: %x\n", key.const_mask); + + +// key.do_point_sprite = brw->attribs.Point->PointSprite; +// key.SpriteOrigin = brw->attribs.Point->SpriteOrigin; + +// key.do_flat_shading = (brw->attribs.Raster->flatshade); +// key.do_twoside_color = (brw->attribs.Light->Enabled && brw->attribs.Light->Model.TwoSide); + +// if (key.do_twoside_color) +// key.frontface_ccw = (brw->attribs.Polygon->FrontFace == GL_CCW); + + + if (!search_cache(brw, &key)) + compile_sf_prog( brw, &key ); +} + + +const struct brw_tracked_state brw_sf_prog = { + .dirty = { + .brw = (BRW_NEW_RASTERIZER | + BRW_NEW_REDUCED_PRIMITIVE | + BRW_NEW_VS | + BRW_NEW_FS), + .cache = 0, + }, + .update = upload_sf_prog +}; + + + +#if 0 +/* Build a struct like the one we'd like the state tracker to pass to + * us. + */ +static void update_sf_linkage( struct brw_context *brw ) +{ + const struct brw_vertex_program *vs = brw->attribs.VertexProgram; + const struct brw_fragment_program *fs = brw->attribs.FragmentProgram; + struct pipe_setup_linkage state; + struct tgsi_parse_context parse; + + int i, j; + int nr_vp_outputs = 0; + int done = 0; + + struct { + unsigned semantic:8; + unsigned semantic_index:16; + } fp_semantic[32], vp_semantic[32]; + + memset(&state, 0, sizeof(state)); + + state.fp_input_count = 0; + + + + + + + assert(state.fp_input_count == fs->program.num_inputs); + + + /* Then scan vp outputs + */ + done = 0; + tgsi_parse_init( &parse, vs->program.tokens ); + while( !done && + !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) + { + int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + + for (i = first; i < last; i++) { + vp_semantic[i].semantic = + parse.FullToken.FullDeclaration.Semantic.SemanticName; + vp_semantic[i].semantic_index = + parse.FullToken.FullDeclaration.Semantic.SemanticIndex; + } + + assert(last > nr_vp_outputs); + nr_vp_outputs = last; + } + break; + default: + done = 1; + break; + } + } + + + /* Now match based on semantic information. + */ + for (i = 0; i< state.fp_input_count; i++) { + for (j = 0; j < nr_vp_outputs; j++) { + if (fp_semantic[i].semantic == vp_semantic[j].semantic && + fp_semantic[i].semantic_index == vp_semantic[j].semantic_index) { + state.fp_input[i].vp_output = j; + } + } + if (fp_semantic[i].semantic == TGSI_SEMANTIC_COLOR) { + for (j = 0; j < nr_vp_outputs; j++) { + if (TGSI_SEMANTIC_BCOLOR == vp_semantic[j].semantic && + fp_semantic[i].semantic_index == vp_semantic[j].semantic_index) { + state.fp_input[i].bf_vp_output = j; + } + } + } + } + + if (memcmp(&brw->sf.linkage, &state, sizeof(state)) != 0) { + brw->sf.linkage = state; + brw->state.dirty.brw |= BRW_NEW_SF_LINKAGE; + } +} + + +const struct brw_tracked_state brw_sf_linkage = { + .dirty = { + .brw = (BRW_NEW_VS | + BRW_NEW_FS), + .cache = 0, + }, + .update = update_sf_linkage +}; + + +#endif diff --git a/src/gallium/drivers/i965simple/brw_sf.h b/src/gallium/drivers/i965simple/brw_sf.h new file mode 100644 index 0000000000..b7ada47560 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_sf.h @@ -0,0 +1,122 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_SF_H +#define BRW_SF_H + +#include "brw_context.h" +#include "brw_eu.h" + + +#define SF_POINTS 0 +#define SF_LINES 1 +#define SF_TRIANGLES 2 +#define SF_UNFILLED_TRIS 3 + + + +struct brw_sf_prog_key { + unsigned vp_output_count:5; + unsigned fp_input_count:5; + + unsigned primitive:2; + unsigned do_twoside_color:1; + unsigned do_flat_shading:1; + unsigned frontface_ccw:1; + unsigned do_point_sprite:1; + + /* Interpolation masks; + */ + unsigned linear_mask; + unsigned persp_mask; + unsigned const_mask; + + +// int SpriteOrigin; +}; + +struct brw_sf_point_tex { + boolean CoordReplace; +}; + +struct brw_sf_compile { + struct brw_compile func; + struct brw_sf_prog_key key; + struct brw_sf_prog_data prog_data; + + struct brw_reg pv; + struct brw_reg det; + struct brw_reg dx0; + struct brw_reg dx2; + struct brw_reg dy0; + struct brw_reg dy2; + + /* z and 1/w passed in seperately: + */ + struct brw_reg z[3]; + struct brw_reg inv_w[3]; + + /* The vertices: + */ + struct brw_reg vert[3]; + + /* Temporaries, allocated after last vertex reg. + */ + struct brw_reg inv_det; + struct brw_reg a1_sub_a0; + struct brw_reg a2_sub_a0; + struct brw_reg tmp; + + struct brw_reg m1Cx; + struct brw_reg m2Cy; + struct brw_reg m3C0; + + unsigned nr_verts; + unsigned nr_attrs; + unsigned nr_attr_regs; + unsigned nr_setup_attrs; + unsigned nr_setup_regs; +#if 0 + ubyte attr_to_idx[VERT_RESULT_MAX]; + ubyte idx_to_attr[VERT_RESULT_MAX]; + struct brw_sf_point_tex point_attrs[VERT_RESULT_MAX]; +#endif +}; + + +void brw_emit_tri_setup( struct brw_sf_compile *c ); +void brw_emit_line_setup( struct brw_sf_compile *c ); +void brw_emit_point_setup( struct brw_sf_compile *c ); +void brw_emit_point_sprite_setup( struct brw_sf_compile *c ); +void brw_emit_anyprim_setup( struct brw_sf_compile *c ); + +#endif diff --git a/src/gallium/drivers/i965simple/brw_sf_emit.c b/src/gallium/drivers/i965simple/brw_sf_emit.c new file mode 100644 index 0000000000..78d6fa5e9e --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_sf_emit.c @@ -0,0 +1,382 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_defines.h" +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_util.h" +#include "brw_sf.h" + + + +/*********************************************************************** + * Triangle setup. + */ + + +static void alloc_regs( struct brw_sf_compile *c ) +{ + unsigned reg, i; + + /* Values computed by fixed function unit: + */ + c->pv = retype(brw_vec1_grf(1, 1), BRW_REGISTER_TYPE_UD); + c->det = brw_vec1_grf(1, 2); + c->dx0 = brw_vec1_grf(1, 3); + c->dx2 = brw_vec1_grf(1, 4); + c->dy0 = brw_vec1_grf(1, 5); + c->dy2 = brw_vec1_grf(1, 6); + + /* z and 1/w passed in seperately: + */ + c->z[0] = brw_vec1_grf(2, 0); + c->inv_w[0] = brw_vec1_grf(2, 1); + c->z[1] = brw_vec1_grf(2, 2); + c->inv_w[1] = brw_vec1_grf(2, 3); + c->z[2] = brw_vec1_grf(2, 4); + c->inv_w[2] = brw_vec1_grf(2, 5); + + /* The vertices: + */ + reg = 3; + for (i = 0; i < c->nr_verts; i++) { + c->vert[i] = brw_vec8_grf(reg, 0); + reg += c->nr_attr_regs; + } + + /* Temporaries, allocated after last vertex reg. + */ + c->inv_det = brw_vec1_grf(reg, 0); reg++; + c->a1_sub_a0 = brw_vec8_grf(reg, 0); reg++; + c->a2_sub_a0 = brw_vec8_grf(reg, 0); reg++; + c->tmp = brw_vec8_grf(reg, 0); reg++; + + /* Note grf allocation: + */ + c->prog_data.total_grf = reg; + + + /* Outputs of this program - interpolation coefficients for + * rasterization: + */ + c->m1Cx = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 1, 0); + c->m2Cy = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 2, 0); + c->m3C0 = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 3, 0); +} + + +static void copy_z_inv_w( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned i; + + brw_push_insn_state(p); + + /* Copy both scalars with a single MOV: + */ + for (i = 0; i < c->nr_verts; i++) + brw_MOV(p, vec2(suboffset(c->vert[i], 2)), vec2(c->z[i])); + + brw_pop_insn_state(p); +} + + +static void invert_det( struct brw_sf_compile *c) +{ + brw_math(&c->func, + c->inv_det, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 0, + c->det, + BRW_MATH_DATA_SCALAR, + BRW_MATH_PRECISION_FULL); + +} + +#define NON_PERPECTIVE_ATTRS (FRAG_BIT_WPOS | \ + FRAG_BIT_COL0 | \ + FRAG_BIT_COL1) + +static boolean calculate_masks( struct brw_sf_compile *c, + unsigned reg, + ushort *pc, + ushort *pc_persp, + ushort *pc_linear) +{ + boolean is_last_attr = (reg == c->nr_setup_regs - 1); + unsigned persp_mask = c->key.persp_mask; + unsigned linear_mask = c->key.linear_mask; + + debug_printf("persp_mask: %x\n", persp_mask); + debug_printf("linear_mask: %x\n", linear_mask); + + *pc_persp = 0; + *pc_linear = 0; + *pc = 0xf; + + if (persp_mask & (1 << (reg*2))) + *pc_persp = 0xf; + + if (linear_mask & (1 << (reg*2))) + *pc_linear = 0xf; + + /* Maybe only processs one attribute on the final round: + */ + if (reg*2+1 < c->nr_setup_attrs) { + *pc |= 0xf0; + + if (persp_mask & (1 << (reg*2+1))) + *pc_persp |= 0xf0; + + if (linear_mask & (1 << (reg*2+1))) + *pc_linear |= 0xf0; + } + + debug_printf("pc: %x\n", *pc); + debug_printf("pc_persp: %x\n", *pc_persp); + debug_printf("pc_linear: %x\n", *pc_linear); + + + return is_last_attr; +} + + + +void brw_emit_tri_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned i; + + debug_printf("%s START ==============\n", __FUNCTION__); + + c->nr_verts = 3; + alloc_regs(c); + invert_det(c); + copy_z_inv_w(c); + + + for (i = 0; i < c->nr_setup_regs; i++) + { + /* Pair of incoming attributes: + */ + struct brw_reg a0 = offset(c->vert[0], i); + struct brw_reg a1 = offset(c->vert[1], i); + struct brw_reg a2 = offset(c->vert[2], i); + ushort pc = 0, pc_persp = 0, pc_linear = 0; + boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + brw_MUL(p, a1, a1, c->inv_w[1]); + brw_MUL(p, a2, a2, c->inv_w[2]); + } + + + /* Calculate coefficients for interpolated values: + */ + if (pc_linear) + { + brw_set_predicate_control_flag_value(p, pc_linear); + + brw_ADD(p, c->a1_sub_a0, a1, negate(a0)); + brw_ADD(p, c->a2_sub_a0, a2, negate(a0)); + + /* calculate dA/dx + */ + brw_MUL(p, brw_null_reg(), c->a1_sub_a0, c->dy2); + brw_MAC(p, c->tmp, c->a2_sub_a0, negate(c->dy0)); + brw_MUL(p, c->m1Cx, c->tmp, c->inv_det); + + /* calculate dA/dy + */ + brw_MUL(p, brw_null_reg(), c->a2_sub_a0, c->dx0); + brw_MAC(p, c->tmp, c->a1_sub_a0, negate(c->dx2)); + brw_MUL(p, c->m2Cy, c->tmp, c->inv_det); + } + + { + brw_set_predicate_control_flag_value(p, pc); + /* start point for interpolation + */ + brw_MOV(p, c->m3C0, a0); + + /* Copy m0..m3 to URB. m0 is implicitly copied from r0 in + * the send instruction: + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), /* r0, will be copied to m0 */ + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* offset */ + BRW_URB_SWIZZLE_TRANSPOSE); /* XXX: Swizzle control "SF to windower" */ + } + } + + debug_printf("%s DONE ==============\n", __FUNCTION__); + +} + + + +void brw_emit_line_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned i; + + + c->nr_verts = 2; + alloc_regs(c); + invert_det(c); + copy_z_inv_w(c); + + for (i = 0; i < c->nr_setup_regs; i++) + { + /* Pair of incoming attributes: + */ + struct brw_reg a0 = offset(c->vert[0], i); + struct brw_reg a1 = offset(c->vert[1], i); + ushort pc, pc_persp, pc_linear; + boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + brw_MUL(p, a1, a1, c->inv_w[1]); + } + + /* Calculate coefficients for position, color: + */ + if (pc_linear) { + brw_set_predicate_control_flag_value(p, pc_linear); + + brw_ADD(p, c->a1_sub_a0, a1, negate(a0)); + + brw_MUL(p, c->tmp, c->a1_sub_a0, c->dx0); + brw_MUL(p, c->m1Cx, c->tmp, c->inv_det); + + brw_MUL(p, c->tmp, c->a1_sub_a0, c->dy0); + brw_MUL(p, c->m2Cy, c->tmp, c->inv_det); + } + + { + brw_set_predicate_control_flag_value(p, pc); + + /* start point for interpolation + */ + brw_MOV(p, c->m3C0, a0); + + /* Copy m0..m3 to URB. + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* urb destination offset */ + BRW_URB_SWIZZLE_TRANSPOSE); + } + } +} + + +/* Points setup - several simplifications as all attributes are + * constant across the face of the point (point sprites excluded!) + */ +void brw_emit_point_setup( struct brw_sf_compile *c ) +{ + struct brw_compile *p = &c->func; + unsigned i; + + c->nr_verts = 1; + alloc_regs(c); + copy_z_inv_w(c); + + brw_MOV(p, c->m1Cx, brw_imm_ud(0)); /* zero - move out of loop */ + brw_MOV(p, c->m2Cy, brw_imm_ud(0)); /* zero - move out of loop */ + + for (i = 0; i < c->nr_setup_regs; i++) + { + struct brw_reg a0 = offset(c->vert[0], i); + ushort pc, pc_persp, pc_linear; + boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); + + if (pc_persp) + { + /* This seems odd as the values are all constant, but the + * fragment shader will be expecting it: + */ + brw_set_predicate_control_flag_value(p, pc_persp); + brw_MUL(p, a0, a0, c->inv_w[0]); + } + + + /* The delta values are always zero, just send the starting + * coordinate. Again, this is to fit in with the interpolation + * code in the fragment shader. + */ + { + brw_set_predicate_control_flag_value(p, pc); + + brw_MOV(p, c->m3C0, a0); /* constant value */ + + /* Copy m0..m3 to URB. + */ + brw_urb_WRITE(p, + brw_null_reg(), + 0, + brw_vec8_grf(0, 0), + 0, /* allocate */ + 1, /* used */ + 4, /* msg len */ + 0, /* response len */ + last, /* eot */ + last, /* writes complete */ + i*4, /* urb destination offset */ + BRW_URB_SWIZZLE_TRANSPOSE); + } + } +} diff --git a/src/gallium/drivers/i965simple/brw_sf_state.c b/src/gallium/drivers/i965simple/brw_sf_state.c new file mode 100644 index 0000000000..9acd3ea61b --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_sf_state.c @@ -0,0 +1,180 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "pipe/p_util.h" + +static void upload_sf_vp(struct brw_context *brw) +{ + struct brw_sf_viewport sfv; + + memset(&sfv, 0, sizeof(sfv)); + + + /* BRW_NEW_VIEWPORT */ + { + const float *scale = brw->attribs.Viewport.scale; + const float *trans = brw->attribs.Viewport.translate; + + sfv.viewport.m00 = scale[0]; + sfv.viewport.m11 = scale[1]; + sfv.viewport.m22 = scale[2]; + sfv.viewport.m30 = trans[0]; + sfv.viewport.m31 = trans[1]; + sfv.viewport.m32 = trans[2]; + } + + /* _NEW_SCISSOR */ + sfv.scissor.xmin = brw->attribs.Scissor.minx; + sfv.scissor.xmax = brw->attribs.Scissor.maxx - 1; + sfv.scissor.ymin = brw->attribs.Scissor.miny; + sfv.scissor.ymax = brw->attribs.Scissor.maxy - 1; + + brw->sf.vp_gs_offset = brw_cache_data( &brw->cache[BRW_SF_VP], &sfv ); +} + +const struct brw_tracked_state brw_sf_vp = { + .dirty = { + .brw = (BRW_NEW_SCISSOR | + BRW_NEW_VIEWPORT), + .cache = 0 + }, + .update = upload_sf_vp +}; + +static void upload_sf_unit( struct brw_context *brw ) +{ + struct brw_sf_unit_state sf; + memset(&sf, 0, sizeof(sf)); + + /* CACHE_NEW_SF_PROG */ + sf.thread0.grf_reg_count = align(brw->sf.prog_data->total_grf, 16) / 16 - 1; + sf.thread0.kernel_start_pointer = brw->sf.prog_gs_offset >> 6; + sf.thread3.urb_entry_read_length = brw->sf.prog_data->urb_read_length; + + sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + sf.thread3.dispatch_grf_start_reg = 3; + sf.thread3.urb_entry_read_offset = 1; + + /* BRW_NEW_URB_FENCE */ + sf.thread4.nr_urb_entries = brw->urb.nr_sf_entries; + sf.thread4.urb_entry_allocation_size = brw->urb.sfsize - 1; + sf.thread4.max_threads = MIN2(12, brw->urb.nr_sf_entries / 2) - 1; + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + sf.thread4.max_threads = 0; + + if (BRW_DEBUG & DEBUG_STATS) + sf.thread4.stats_enable = 1; + + /* CACHE_NEW_SF_VP */ + sf.sf5.sf_viewport_state_offset = brw->sf.vp_gs_offset >> 5; + sf.sf5.viewport_transform = 1; + + /* BRW_NEW_RASTER */ + if (brw->attribs.Raster->scissor) + sf.sf6.scissor = 1; + +#if 0 + if (brw->attribs.Polygon->FrontFace == GL_CCW) + sf.sf5.front_winding = BRW_FRONTWINDING_CCW; + else + sf.sf5.front_winding = BRW_FRONTWINDING_CW; + + + if (brw->attribs.Polygon->CullFlag) { + switch (brw->attribs.Polygon->CullFaceMode) { + case GL_FRONT: + sf.sf6.cull_mode = BRW_CULLMODE_FRONT; + break; + case GL_BACK: + sf.sf6.cull_mode = BRW_CULLMODE_BACK; + break; + case GL_FRONT_AND_BACK: + sf.sf6.cull_mode = BRW_CULLMODE_BOTH; + break; + default: + assert(0); + break; + } + } + else + sf.sf6.cull_mode = BRW_CULLMODE_NONE; +#else + sf.sf5.front_winding = BRW_FRONTWINDING_CCW; + sf.sf6.cull_mode = BRW_CULLMODE_NONE; +#endif + + sf.sf6.line_width = CLAMP(brw->attribs.Raster->line_width, 1.0, 5.0) * (1<<1); + + sf.sf6.line_endcap_aa_region_width = 1; + if (brw->attribs.Raster->line_smooth) + sf.sf6.aa_enable = 1; + else if (sf.sf6.line_width <= 0x2) + sf.sf6.line_width = 0; + + sf.sf6.point_rast_rule = 1; /* opengl conventions */ + + sf.sf7.sprite_point = brw->attribs.Raster->point_sprite; + sf.sf7.point_size = CLAMP(brw->attribs.Raster->line_width, 1.0, 255.0) * (1<<3); + sf.sf7.use_point_size_state = !brw->attribs.Raster->point_size_per_vertex; + + /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons: + */ + sf.sf7.trifan_pv = 2; + sf.sf7.linestrip_pv = 1; + sf.sf7.tristrip_pv = 2; + sf.sf7.line_last_pixel_enable = 0; + + /* Set bias for OpenGL rasterization rules: + */ + sf.sf6.dest_org_vbias = 0x8; + sf.sf6.dest_org_hbias = 0x8; + + brw->sf.state_gs_offset = brw_cache_data( &brw->cache[BRW_SF_UNIT], &sf ); +} + + +const struct brw_tracked_state brw_sf_unit = { + .dirty = { + .brw = (BRW_NEW_RASTERIZER | + BRW_NEW_URB_FENCE), + .cache = (CACHE_NEW_SF_VP | + CACHE_NEW_SF_PROG) + }, + .update = upload_sf_unit +}; + + diff --git a/src/gallium/drivers/i965simple/brw_shader_info.c b/src/gallium/drivers/i965simple/brw_shader_info.c new file mode 100644 index 0000000000..431b45466a --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_shader_info.c @@ -0,0 +1,49 @@ + +#include "brw_context.h" +#include "brw_state.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" + + + + +void brw_shader_info(const struct tgsi_token *tokens, + struct brw_shader_info *info ) +{ + struct tgsi_parse_context parse; + int done = 0; + + tgsi_parse_init( &parse, tokens ); + + while( !done && + !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; + unsigned last = decl->u.DeclarationRange.Last; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + // Broken by crazy wpos init: + //assert( info->nr_regs[decl->Declaration.File] <= last); + + info->nr_regs[decl->Declaration.File] = MAX2(info->nr_regs[decl->Declaration.File], + last+1); + break; + } + case TGSI_TOKEN_TYPE_IMMEDIATE: + case TGSI_TOKEN_TYPE_INSTRUCTION: + default: + done = 1; + break; + } + } + + tgsi_parse_free (&parse); + +} diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c new file mode 100644 index 0000000000..95dfce88e4 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -0,0 +1,424 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Zack Rusin + * Keith Whitwell + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_dump.h" + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_state.h" +#include "brw_draw.h" + + +#define DUP( TYPE, VAL ) \ +do { \ + struct TYPE *x = malloc(sizeof(*x)); \ + memcpy(x, VAL, sizeof(*x) ); \ + return x; \ +} while (0) + +/************************************************************************ + * Blend + */ +static void * +brw_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + DUP( pipe_blend_state, blend ); +} + +static void brw_bind_blend_state(struct pipe_context *pipe, + void *blend) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Blend = (struct pipe_blend_state*)blend; + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + + +static void brw_delete_blend_state(struct pipe_context *pipe, void *blend) +{ + free(blend); +} + +static void brw_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.BlendColor = *blend_color; + + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + +/************************************************************************ + * Sampler + */ + +static void * +brw_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + DUP( pipe_sampler_state, sampler ); +} + +static void brw_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Samplers[unit] = sampler; + brw->state.dirty.brw |= BRW_NEW_SAMPLER; +} + +static void brw_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + free(sampler); +} + + +/************************************************************************ + * Depth stencil + */ + +static void * +brw_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil) +{ + DUP( pipe_depth_stencil_alpha_state, depth_stencil ); +} + +static void brw_bind_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; + + brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; +} + +static void brw_delete_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + free(depth_stencil); +} + +/************************************************************************ + * Scissor + */ +static void brw_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct brw_context *brw = brw_context(pipe); + + memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) ); + brw->state.dirty.brw |= BRW_NEW_SCISSOR; +} + + +/************************************************************************ + * Stipple + */ + +static void brw_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ +} + + +/************************************************************************ + * Fragment shader + */ + +static void * brw_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); + + /* XXX: Do I have to duplicate the tokens as well?? + */ + brw_fp->program = *shader; + brw_fp->id = brw_context(pipe)->program_id++; + + brw_shader_info(shader->tokens, + &brw_fp->info); + + tgsi_dump(shader->tokens, 0); + + + return (void *)brw_fp; +} + +static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader; + brw->state.dirty.brw |= BRW_NEW_FS; +} + +static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) +{ + FREE(shader); +} + + +/************************************************************************ + * Vertex shader and other TNL state + */ + +static void *brw_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); + + /* XXX: Do I have to duplicate the tokens as well?? + */ + brw_vp->program = *shader; + brw_vp->id = brw_context(pipe)->program_id++; + brw_shader_info(shader->tokens, + &brw_vp->info); + + tgsi_dump(shader->tokens, 0); + + return (void *)brw_vp; +} + +static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; + brw->state.dirty.brw |= BRW_NEW_VS; + + debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); +} + +static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) +{ + FREE(shader); +} + + +static void brw_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Clip = *clip; +} + + +static void brw_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Viewport = *viewport; /* struct copy */ + brw->state.dirty.brw |= BRW_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + //draw_set_viewport_state(brw->draw, viewport); +} + + +static void brw_set_vertex_buffer( struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_buffer *buffer ) +{ + struct brw_context *brw = brw_context(pipe); + brw->vb.vbo_array[index] = buffer; +} + +static void brw_set_vertex_element(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_element *element) +{ + /* flush ? */ + struct brw_context *brw = brw_context(pipe); + + assert(index < PIPE_ATTRIB_MAX); + struct brw_vertex_element_state el; + memset(&el, 0, sizeof(el)); + + el.ve0.src_offset = element->src_offset; + el.ve0.src_format = brw_translate_surface_format(element->src_format); + el.ve0.valid = 1; + el.ve0.vertex_buffer_index = element->vertex_buffer_index; + + el.ve1.dst_offset = index * 4; + + el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; + + switch (element->nr_components) { + case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; + case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; + case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; + break; + } + + brw->vb.inputs[index] = el; +} + + + +/************************************************************************ + * Constant buffers + */ + +static void brw_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct brw_context *brw = brw_context(pipe); + + assert(buf == 0 || index == 0); + + brw->attribs.Constants[shader] = buf; + brw->state.dirty.brw |= BRW_NEW_CONSTANTS; +} + + +/************************************************************************ + * Texture surfaces + */ + + +static void brw_set_sampler_texture(struct pipe_context *pipe, + unsigned unit, + struct pipe_texture *texture) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Texture[unit] = (struct brw_texture*)texture; /* ptr, not struct */ + + brw->state.dirty.brw |= BRW_NEW_TEXTURE; +} + + +/************************************************************************ + * Render targets, etc + */ + +static void brw_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FrameBuffer = *fb; /* struct copy */ + + brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER; +} + + + +/************************************************************************ + * Rasterizer state + */ + +static void * +brw_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *rasterizer) +{ + DUP(pipe_rasterizer_state, rasterizer); +} + +static void brw_bind_rasterizer_state( struct pipe_context *pipe, + void *setup ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Raster = (struct pipe_rasterizer_state *)setup; + + /* Also pass-through to draw module: + */ + //draw_set_rasterizer_state(brw->draw, setup); + + brw->state.dirty.brw |= BRW_NEW_RASTERIZER; +} + +static void brw_delete_rasterizer_state(struct pipe_context *pipe, + void *setup) +{ + free(setup); +} + + + +void +brw_init_state_functions( struct brw_context *brw ) +{ + brw->pipe.create_blend_state = brw_create_blend_state; + brw->pipe.bind_blend_state = brw_bind_blend_state; + brw->pipe.delete_blend_state = brw_delete_blend_state; + + brw->pipe.create_sampler_state = brw_create_sampler_state; + brw->pipe.bind_sampler_state = brw_bind_sampler_state; + brw->pipe.delete_sampler_state = brw_delete_sampler_state; + + brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; + brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; + brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; + + brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; + brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; + brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state; + brw->pipe.create_fs_state = brw_create_fs_state; + brw->pipe.bind_fs_state = brw_bind_fs_state; + brw->pipe.delete_fs_state = brw_delete_fs_state; + brw->pipe.create_vs_state = brw_create_vs_state; + brw->pipe.bind_vs_state = brw_bind_vs_state; + brw->pipe.delete_vs_state = brw_delete_vs_state; + + brw->pipe.set_blend_color = brw_set_blend_color; + brw->pipe.set_clip_state = brw_set_clip_state; + brw->pipe.set_constant_buffer = brw_set_constant_buffer; + brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; + +// brw->pipe.set_feedback_state = brw_set_feedback_state; +// brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; + + brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; + brw->pipe.set_scissor_state = brw_set_scissor_state; + brw->pipe.set_sampler_texture = brw_set_sampler_texture; + brw->pipe.set_viewport_state = brw_set_viewport_state; + brw->pipe.set_vertex_buffer = brw_set_vertex_buffer; + brw->pipe.set_vertex_element = brw_set_vertex_element; +} diff --git a/src/gallium/drivers/i965simple/brw_state.h b/src/gallium/drivers/i965simple/brw_state.h new file mode 100644 index 0000000000..258e9a556e --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_state.h @@ -0,0 +1,158 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_STATE_H +#define BRW_STATE_H + +#include "brw_context.h" +#include "brw_winsys.h" + + +const struct brw_tracked_state brw_blend_constant_color; +const struct brw_tracked_state brw_cc_unit; +const struct brw_tracked_state brw_cc_vp; +const struct brw_tracked_state brw_clip_prog; +const struct brw_tracked_state brw_clip_unit; +const struct brw_tracked_state brw_constant_buffer_state; +const struct brw_tracked_state brw_constant_buffer; +const struct brw_tracked_state brw_curbe_offsets; +const struct brw_tracked_state brw_invarient_state; +const struct brw_tracked_state brw_gs_prog; +const struct brw_tracked_state brw_gs_unit; +const struct brw_tracked_state brw_drawing_rect; +const struct brw_tracked_state brw_line_stipple; +const struct brw_tracked_state brw_pipelined_state_pointers; +const struct brw_tracked_state brw_binding_table_pointers; +const struct brw_tracked_state brw_depthbuffer; +const struct brw_tracked_state brw_polygon_stipple_offset; +const struct brw_tracked_state brw_polygon_stipple; +const struct brw_tracked_state brw_program_parameters; +const struct brw_tracked_state brw_recalculate_urb_fence; +const struct brw_tracked_state brw_sf_prog; +const struct brw_tracked_state brw_sf_unit; +const struct brw_tracked_state brw_sf_vp; +const struct brw_tracked_state brw_state_base_address; +const struct brw_tracked_state brw_urb_fence; +const struct brw_tracked_state brw_vertex_state; +const struct brw_tracked_state brw_vs_prog; +const struct brw_tracked_state brw_vs_unit; +const struct brw_tracked_state brw_wm_prog; +const struct brw_tracked_state brw_wm_samplers; +const struct brw_tracked_state brw_wm_surfaces; +const struct brw_tracked_state brw_wm_unit; + +const struct brw_tracked_state brw_psp_urb_cbs; + +const struct brw_tracked_state brw_active_vertprog; +const struct brw_tracked_state brw_tnl_vertprog; +const struct brw_tracked_state brw_pipe_control; + +const struct brw_tracked_state brw_clear_surface_cache; +const struct brw_tracked_state brw_clear_batch_cache; + +/*********************************************************************** + * brw_state_cache.c + */ +unsigned brw_cache_data(struct brw_cache *cache, + const void *data ); + +unsigned brw_cache_data_sz(struct brw_cache *cache, + const void *data, + unsigned data_sz); + +unsigned brw_upload_cache( struct brw_cache *cache, + const void *key, + unsigned key_sz, + const void *data, + unsigned data_sz, + const void *aux, + void *aux_return ); + +boolean brw_search_cache( struct brw_cache *cache, + const void *key, + unsigned key_size, + void *aux_return, + unsigned *offset_return); + +void brw_init_caches( struct brw_context *brw ); +void brw_destroy_caches( struct brw_context *brw ); + +static inline struct pipe_buffer *brw_cache_buffer(struct brw_context *brw, + enum brw_cache_id id) +{ + return brw->cache[id].pool->buffer; +} + +/*********************************************************************** + * brw_state_batch.c + */ +#define BRW_CACHED_BATCH_STRUCT(brw, s) brw_cached_batch_struct( brw, (s), sizeof(*(s)) ) + +boolean brw_cached_batch_struct( struct brw_context *brw, + const void *data, + unsigned sz ); + +void brw_destroy_batch_cache( struct brw_context *brw ); + + +/*********************************************************************** + * brw_state_pool.c + */ +void brw_init_pools( struct brw_context *brw ); +void brw_destroy_pools( struct brw_context *brw ); + +boolean brw_pool_alloc( struct brw_mem_pool *pool, + unsigned size, + unsigned alignment, + unsigned *offset_return); + +void brw_pool_fence( struct brw_context *brw, + struct brw_mem_pool *pool, + unsigned fence ); + + +void brw_pool_check_wrap( struct brw_context *brw, + struct brw_mem_pool *pool ); + +void brw_clear_all_caches( struct brw_context *brw ); +void brw_invalidate_pools( struct brw_context *brw ); +void brw_clear_batch_cache_flush( struct brw_context *brw ); + + +/* brw_shader_info.c + */ + +void brw_shader_info(const struct tgsi_token *tokens, + struct brw_shader_info *info ); + + +#endif diff --git a/src/gallium/drivers/i965simple/brw_state_batch.c b/src/gallium/drivers/i965simple/brw_state_batch.c new file mode 100644 index 0000000000..35db76b594 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_state_batch.c @@ -0,0 +1,113 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_state.h" +#include "brw_winsys.h" + +#include "pipe/p_util.h" + +/* A facility similar to the data caching code above, which aims to + * prevent identical commands being issued repeatedly. + */ +boolean brw_cached_batch_struct( struct brw_context *brw, + const void *data, + unsigned sz ) +{ + struct brw_cached_batch_item *item = brw->cached_batch_items; + struct header *newheader = (struct header *)data; + + if (brw->emit_state_always) { + brw_batchbuffer_data(brw->winsys, data, sz); + return TRUE; + } + + while (item) { + if (item->header->opcode == newheader->opcode) { + if (item->sz == sz && memcmp(item->header, newheader, sz) == 0) + return FALSE; + if (item->sz != sz) { + FREE(item->header); + item->header = MALLOC(sz); + item->sz = sz; + } + goto emit; + } + item = item->next; + } + + assert(!item); + item = CALLOC_STRUCT(brw_cached_batch_item); + item->header = MALLOC(sz); + item->sz = sz; + item->next = brw->cached_batch_items; + brw->cached_batch_items = item; + +emit: + memcpy(item->header, newheader, sz); + brw_batchbuffer_data(brw->winsys, data, sz); + return TRUE; +} + +static void clear_batch_cache( struct brw_context *brw ) +{ + struct brw_cached_batch_item *item = brw->cached_batch_items; + + while (item) { + struct brw_cached_batch_item *next = item->next; + free((void *)item->header); + free(item); + item = next; + } + + brw->cached_batch_items = NULL; + + + brw_clear_all_caches(brw); + + brw_invalidate_pools(brw); +} + +void brw_clear_batch_cache_flush( struct brw_context *brw ) +{ + clear_batch_cache(brw); + +/* brw_do_flush(brw, BRW_FLUSH_STATE_CACHE|BRW_FLUSH_READ_CACHE); */ + + brw->state.dirty.brw |= ~0; + brw->state.dirty.cache |= ~0; +} + + + +void brw_destroy_batch_cache( struct brw_context *brw ) +{ + clear_batch_cache(brw); +} diff --git a/src/gallium/drivers/i965simple/brw_state_cache.c b/src/gallium/drivers/i965simple/brw_state_cache.c new file mode 100644 index 0000000000..b3a5124461 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_state_cache.c @@ -0,0 +1,443 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_state.h" + +#include "brw_wm.h" +#include "brw_vs.h" +#include "brw_clip.h" +#include "brw_sf.h" +#include "brw_gs.h" + +#include "pipe/p_util.h" + + + +/*********************************************************************** + * Check cache for uploaded version of struct, else upload new one. + * Fail when memory is exhausted. + * + * XXX: FIXME: Currently search is so slow it would be quicker to + * regenerate the data every time... + */ + +static unsigned hash_key( const void *key, unsigned key_size ) +{ + unsigned *ikey = (unsigned *)key; + unsigned hash = 0, i; + + assert(key_size % 4 == 0); + + /* I'm sure this can be improved on: + */ + for (i = 0; i < key_size/4; i++) + hash ^= ikey[i]; + + return hash; +} + +static struct brw_cache_item *search_cache( struct brw_cache *cache, + unsigned hash, + const void *key, + unsigned key_size) +{ + struct brw_cache_item *c; + + for (c = cache->items[hash % cache->size]; c; c = c->next) { + if (c->hash == hash && + c->key_size == key_size && + memcmp(c->key, key, key_size) == 0) + return c; + } + + return NULL; +} + + +static void rehash( struct brw_cache *cache ) +{ + struct brw_cache_item **items; + struct brw_cache_item *c, *next; + unsigned size, i; + + size = cache->size * 3; + items = (struct brw_cache_item**) MALLOC(size * sizeof(*items)); + memset(items, 0, size * sizeof(*items)); + + for (i = 0; i < cache->size; i++) + for (c = cache->items[i]; c; c = next) { + next = c->next; + c->next = items[c->hash % size]; + items[c->hash % size] = c; + } + + FREE(cache->items); + cache->items = items; + cache->size = size; +} + + +boolean brw_search_cache( struct brw_cache *cache, + const void *key, + unsigned key_size, + void *aux_return, + unsigned *offset_return) +{ + struct brw_cache_item *item; + unsigned addr = 0; + unsigned hash = hash_key(key, key_size); + + item = search_cache(cache, hash, key, key_size); + + if (item) { + if (aux_return) + *(void **)aux_return = (void *)((char *)item->key + item->key_size); + + *offset_return = addr = item->offset; + } + + if (item == NULL || addr != cache->last_addr) { + cache->brw->state.dirty.cache |= 1<id; + cache->last_addr = addr; + } + + return item != NULL; +} + +unsigned brw_upload_cache( struct brw_cache *cache, + const void *key, + unsigned key_size, + const void *data, + unsigned data_size, + const void *aux, + void *aux_return ) +{ + unsigned offset; + struct brw_cache_item *item = CALLOC_STRUCT(brw_cache_item); + unsigned hash = hash_key(key, key_size); + void *tmp = MALLOC(key_size + cache->aux_size); + + if (!brw_pool_alloc(cache->pool, data_size, 1 << 6, &offset)) { + /* Should not be possible: + */ + debug_printf("brw_pool_alloc failed\n"); + exit(1); + } + + memcpy(tmp, key, key_size); + + if (cache->aux_size) + memcpy(tmp+key_size, aux, cache->aux_size); + + item->key = tmp; + item->hash = hash; + item->key_size = key_size; + item->offset = offset; + item->data_size = data_size; + + if (++cache->n_items > cache->size * 1.5) + rehash(cache); + + hash %= cache->size; + item->next = cache->items[hash]; + cache->items[hash] = item; + + if (aux_return) { + assert(cache->aux_size); + *(void **)aux_return = (void *)((char *)item->key + item->key_size); + } + + if (BRW_DEBUG & DEBUG_STATE) + debug_printf("upload %s: %d bytes to pool buffer %p offset %x\n", + cache->name, + data_size, + (void*)cache->pool->buffer, + offset); + + /* Copy data to the buffer: + */ + cache->brw->winsys->buffer_subdata_typed(cache->brw->winsys, + cache->pool->buffer, + offset, + data_size, + data, + cache->id); + + cache->brw->state.dirty.cache |= 1<id; + cache->last_addr = offset; + + return offset; +} + +/* This doesn't really work with aux data. Use search/upload instead + */ +unsigned brw_cache_data_sz(struct brw_cache *cache, + const void *data, + unsigned data_size) +{ + unsigned addr; + + if (!brw_search_cache(cache, data, data_size, NULL, &addr)) { + addr = brw_upload_cache(cache, + data, data_size, + data, data_size, + NULL, NULL); + } + + return addr; +} + +unsigned brw_cache_data(struct brw_cache *cache, + const void *data) +{ + return brw_cache_data_sz(cache, data, cache->key_size); +} + +enum pool_type { + DW_SURFACE_STATE, + DW_GENERAL_STATE +}; + +static void brw_init_cache( struct brw_context *brw, + const char *name, + unsigned id, + unsigned key_size, + unsigned aux_size, + enum pool_type pool_type) +{ + struct brw_cache *cache = &brw->cache[id]; + cache->brw = brw; + cache->id = id; + cache->name = name; + cache->items = NULL; + + cache->size = 7; + cache->n_items = 0; + cache->items = (struct brw_cache_item **) + CALLOC(cache->size, sizeof(struct brw_cache_item)); + + + cache->key_size = key_size; + cache->aux_size = aux_size; + switch (pool_type) { + case DW_GENERAL_STATE: cache->pool = &brw->pool[BRW_GS_POOL]; break; + case DW_SURFACE_STATE: cache->pool = &brw->pool[BRW_SS_POOL]; break; + default: assert(0); break; + } +} + +void brw_init_caches( struct brw_context *brw ) +{ + + brw_init_cache(brw, + "CC_VP", + BRW_CC_VP, + sizeof(struct brw_cc_viewport), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "CC_UNIT", + BRW_CC_UNIT, + sizeof(struct brw_cc_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "WM_PROG", + BRW_WM_PROG, + sizeof(struct brw_wm_prog_key), + sizeof(struct brw_wm_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SAMPLER_DEFAULT_COLOR", + BRW_SAMPLER_DEFAULT_COLOR, + sizeof(struct brw_sampler_default_color), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SAMPLER", + BRW_SAMPLER, + 0, /* variable key/data size */ + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "WM_UNIT", + BRW_WM_UNIT, + sizeof(struct brw_wm_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SF_PROG", + BRW_SF_PROG, + sizeof(struct brw_sf_prog_key), + sizeof(struct brw_sf_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SF_VP", + BRW_SF_VP, + sizeof(struct brw_sf_viewport), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SF_UNIT", + BRW_SF_UNIT, + sizeof(struct brw_sf_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "VS_UNIT", + BRW_VS_UNIT, + sizeof(struct brw_vs_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "VS_PROG", + BRW_VS_PROG, + sizeof(struct brw_vs_prog_key), + sizeof(struct brw_vs_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "CLIP_UNIT", + BRW_CLIP_UNIT, + sizeof(struct brw_clip_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "CLIP_PROG", + BRW_CLIP_PROG, + sizeof(struct brw_clip_prog_key), + sizeof(struct brw_clip_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "GS_UNIT", + BRW_GS_UNIT, + sizeof(struct brw_gs_unit_state), + 0, + DW_GENERAL_STATE); + + brw_init_cache(brw, + "GS_PROG", + BRW_GS_PROG, + sizeof(struct brw_gs_prog_key), + sizeof(struct brw_gs_prog_data), + DW_GENERAL_STATE); + + brw_init_cache(brw, + "SS_SURFACE", + BRW_SS_SURFACE, + sizeof(struct brw_surface_state), + 0, + DW_SURFACE_STATE); + + brw_init_cache(brw, + "SS_SURF_BIND", + BRW_SS_SURF_BIND, + sizeof(struct brw_surface_binding_table), + 0, + DW_SURFACE_STATE); +} + + +/* When we lose hardware context, need to invalidate the surface cache + * as these structs must be explicitly re-uploaded. They are subject + * to fixup by the memory manager as they contain absolute agp + * offsets, so we need to ensure there is a fresh version of the + * struct available to receive the fixup. + * + * XXX: Need to ensure that there aren't two versions of a surface or + * bufferobj with different backing data active in the same buffer at + * once? Otherwise the cache could confuse them. Maybe better not to + * cache at all? + * + * --> Isn't this the same as saying need to ensure batch is flushed + * before new data is uploaded to an existing buffer? We + * already try to make sure of that. + */ +static void clear_cache( struct brw_cache *cache ) +{ + struct brw_cache_item *c, *next; + unsigned i; + + for (i = 0; i < cache->size; i++) { + for (c = cache->items[i]; c; c = next) { + next = c->next; + free((void *)c->key); + free(c); + } + cache->items[i] = NULL; + } + + cache->n_items = 0; +} + +void brw_clear_all_caches( struct brw_context *brw ) +{ + int i; + + if (BRW_DEBUG & DEBUG_STATE) + debug_printf("%s\n", __FUNCTION__); + + for (i = 0; i < BRW_MAX_CACHE; i++) + clear_cache(&brw->cache[i]); + + if (brw->curbe.last_buf) { + FREE(brw->curbe.last_buf); + brw->curbe.last_buf = NULL; + } + + brw->state.dirty.brw |= ~0; + brw->state.dirty.cache |= ~0; +} + + + + + +void brw_destroy_caches( struct brw_context *brw ) +{ + unsigned i; + + for (i = 0; i < BRW_MAX_CACHE; i++) + clear_cache(&brw->cache[i]); +} diff --git a/src/gallium/drivers/i965simple/brw_state_pool.c b/src/gallium/drivers/i965simple/brw_state_pool.c new file mode 100644 index 0000000000..f3174bfe0a --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_state_pool.c @@ -0,0 +1,137 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +/** @file brw_state_pool.c + * Implements the state pool allocator. + * + * For the 965, we create two state pools for state cache entries. Objects + * will be allocated into the pools depending on which state base address + * their pointer is relative to in other 965 state. + * + * The state pools are relatively simple: As objects are allocated, increment + * the offset to allocate space. When the pool is "full" (rather, close to + * full), we reset the pool and reset the state cache entries that point into + * the pool. + */ + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "brw_context.h" +#include "brw_state.h" + +boolean brw_pool_alloc( struct brw_mem_pool *pool, + unsigned size, + unsigned alignment, + unsigned *offset_return) +{ + unsigned fixup = align(pool->offset, alignment) - pool->offset; + + size = align(size, 4); + + if (pool->offset + fixup + size >= pool->size) { + debug_printf("%s failed\n", __FUNCTION__); + assert(0); + exit(0); + } + + pool->offset += fixup; + *offset_return = pool->offset; + pool->offset += size; + + return TRUE; +} + +static +void brw_invalidate_pool( struct brw_mem_pool *pool ) +{ + if (BRW_DEBUG & DEBUG_STATE) + debug_printf("\n\n\n %s \n\n\n", __FUNCTION__); + + pool->offset = 0; + + brw_clear_all_caches(pool->brw); +} + + +static void brw_init_pool( struct brw_context *brw, + unsigned pool_id, + unsigned size ) +{ + struct brw_mem_pool *pool = &brw->pool[pool_id]; + + pool->size = size; + pool->brw = brw; + + pool->buffer = brw->pipe.winsys->buffer_create(brw->pipe.winsys, + 4096, + 0 /* DRM_BO_FLAG_MEM_TT */, + size); +} + +static void brw_destroy_pool( struct brw_context *brw, + unsigned pool_id ) +{ + struct brw_mem_pool *pool = &brw->pool[pool_id]; + + pipe_buffer_reference( pool->brw->pipe.winsys, + &pool->buffer, + NULL ); +} + + +void brw_pool_check_wrap( struct brw_context *brw, + struct brw_mem_pool *pool ) +{ + if (pool->offset > (pool->size * 3) / 4) { + brw->state.dirty.brw |= BRW_NEW_SCENE; + } + +} + +void brw_init_pools( struct brw_context *brw ) +{ + brw_init_pool(brw, BRW_GS_POOL, 0x80000); + brw_init_pool(brw, BRW_SS_POOL, 0x80000); +} + +void brw_destroy_pools( struct brw_context *brw ) +{ + brw_destroy_pool(brw, BRW_GS_POOL); + brw_destroy_pool(brw, BRW_SS_POOL); +} + + +void brw_invalidate_pools( struct brw_context *brw ) +{ + brw_invalidate_pool(&brw->pool[BRW_GS_POOL]); + brw_invalidate_pool(&brw->pool[BRW_SS_POOL]); +} diff --git a/src/gallium/drivers/i965simple/brw_state_upload.c b/src/gallium/drivers/i965simple/brw_state_upload.c new file mode 100644 index 0000000000..e727601e1e --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_state_upload.c @@ -0,0 +1,202 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_state.h" + +#include "pipe/p_util.h" + +/* This is used to initialize brw->state.atoms[]. We could use this + * list directly except for a single atom, brw_constant_buffer, which + * has a .dirty value which changes according to the parameters of the + * current fragment and vertex programs, and so cannot be a static + * value. + */ +const struct brw_tracked_state *atoms[] = +{ + &brw_vs_prog, + &brw_gs_prog, + &brw_clip_prog, + &brw_sf_prog, + &brw_wm_prog, + + /* Once all the programs are done, we know how large urb entry + * sizes need to be and can decide if we need to change the urb + * layout. + */ + &brw_curbe_offsets, + &brw_recalculate_urb_fence, + + + &brw_cc_vp, + &brw_cc_unit, + + &brw_wm_surfaces, /* must do before samplers */ + &brw_wm_samplers, + + &brw_wm_unit, + &brw_sf_vp, + &brw_sf_unit, + &brw_vs_unit, /* always required, enabled or not */ + &brw_clip_unit, + &brw_gs_unit, + + /* Command packets: + */ + &brw_invarient_state, + &brw_state_base_address, + &brw_pipe_control, + + &brw_binding_table_pointers, + &brw_blend_constant_color, + + &brw_drawing_rect, + &brw_depthbuffer, + + &brw_polygon_stipple, + &brw_line_stipple, + + &brw_psp_urb_cbs, + + &brw_constant_buffer +}; + + +void brw_init_state( struct brw_context *brw ) +{ + brw_init_pools(brw); + brw_init_caches(brw); + + brw->state.dirty.brw = ~0; + brw->emit_state_always = 0; +} + + +void brw_destroy_state( struct brw_context *brw ) +{ + brw_destroy_caches(brw); + brw_destroy_batch_cache(brw); + brw_destroy_pools(brw); +} + +/*********************************************************************** + */ + +static boolean check_state( const struct brw_state_flags *a, + const struct brw_state_flags *b ) +{ + return ((a->brw & b->brw) || + (a->cache & b->cache)); +} + +static void accumulate_state( struct brw_state_flags *a, + const struct brw_state_flags *b ) +{ + a->brw |= b->brw; + a->cache |= b->cache; +} + + +static void xor_states( struct brw_state_flags *result, + const struct brw_state_flags *a, + const struct brw_state_flags *b ) +{ + result->brw = a->brw ^ b->brw; + result->cache = a->cache ^ b->cache; +} + + +/*********************************************************************** + * Emit all state: + */ +void brw_validate_state( struct brw_context *brw ) +{ + struct brw_state_flags *state = &brw->state.dirty; + unsigned i; + + if (brw->emit_state_always) + state->brw |= ~0; + + if (state->cache == 0 && + state->brw == 0) + return; + + if (brw->state.dirty.brw & BRW_NEW_SCENE) + brw_clear_batch_cache_flush(brw); + + if (BRW_DEBUG) { + /* Debug version which enforces various sanity checks on the + * state flags which are generated and checked to help ensure + * state atoms are ordered correctly in the list. + */ + struct brw_state_flags examined, prev; + memset(&examined, 0, sizeof(examined)); + prev = *state; + + for (i = 0; i < Elements(atoms); i++) { + const struct brw_tracked_state *atom = atoms[i]; + struct brw_state_flags generated; + + assert(atom->dirty.brw || + atom->dirty.cache); + assert(atom->update); + + if (check_state(state, &atom->dirty)) { + atom->update( brw ); + } + + accumulate_state(&examined, &atom->dirty); + + /* generated = (prev ^ state) + * if (examined & generated) + * fail; + */ + xor_states(&generated, &prev, state); + assert(!check_state(&examined, &generated)); + prev = *state; + } + } + else { + for (i = 0; i < Elements(atoms); i++) { + const struct brw_tracked_state *atom = atoms[i]; + + assert(atom->dirty.brw || + atom->dirty.cache); + assert(atom->update); + + if (check_state(state, &atom->dirty)) + atom->update( brw ); + } + } + + memset(state, 0, sizeof(*state)); +} diff --git a/src/gallium/drivers/i965simple/brw_strings.c b/src/gallium/drivers/i965simple/brw_strings.c new file mode 100644 index 0000000000..29a41ed1e9 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_strings.c @@ -0,0 +1,72 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "brw_context.h" +#include "brw_reg.h" + + +static const char *brw_get_vendor( struct pipe_context *pipe ) +{ + return "Tungsten Graphics, Inc."; +} + + +static const char *brw_get_name( struct pipe_context *pipe ) +{ + static char buffer[128]; + const char *chipset; + + switch (brw_context(pipe)->pci_id) { + case PCI_CHIP_I965_Q: + chipset = "Intel(R) 965Q"; + break; + case PCI_CHIP_I965_G: + case PCI_CHIP_I965_G_1: + chipset = "Intel(R) 965G"; + break; + case PCI_CHIP_I965_GM: + chipset = "Intel(R) 965GM"; + break; + case PCI_CHIP_I965_GME: + chipset = "Intel(R) 965GME/GLE"; + break; + default: + chipset = "unknown"; + break; + } + + sprintf(buffer, "pipe/i965 (chipset: %s)", chipset); + return buffer; +} + + +void +brw_init_string_functions(struct brw_context *brw) +{ + brw->pipe.get_name = brw_get_name; + brw->pipe.get_vendor = brw_get_vendor; +} diff --git a/src/gallium/drivers/i965simple/brw_structs.h b/src/gallium/drivers/i965simple/brw_structs.h new file mode 100644 index 0000000000..bbb087e95d --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_structs.h @@ -0,0 +1,1348 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_STRUCTS_H +#define BRW_STRUCTS_H + +#include "pipe/p_compiler.h" + +/* Command packets: + */ +struct header +{ + unsigned length:16; + unsigned opcode:16; +}; + + +union header_union +{ + struct header bits; + unsigned dword; +}; + +struct brw_3d_control +{ + struct + { + unsigned length:8; + unsigned notify_enable:1; + unsigned pad:3; + unsigned wc_flush_enable:1; + unsigned depth_stall_enable:1; + unsigned operation:2; + unsigned opcode:16; + } header; + + struct + { + unsigned pad:2; + unsigned dest_addr_type:1; + unsigned dest_addr:29; + } dest; + + unsigned dword2; + unsigned dword3; +}; + + +struct brw_3d_primitive +{ + struct + { + unsigned length:8; + unsigned pad:2; + unsigned topology:5; + unsigned indexed:1; + unsigned opcode:16; + } header; + + unsigned verts_per_instance; + unsigned start_vert_location; + unsigned instance_count; + unsigned start_instance_location; + unsigned base_vert_location; +}; + +/* These seem to be passed around as function args, so it works out + * better to keep them as #defines: + */ +#define BRW_FLUSH_READ_CACHE 0x1 +#define BRW_FLUSH_STATE_CACHE 0x2 +#define BRW_INHIBIT_FLUSH_RENDER_CACHE 0x4 +#define BRW_FLUSH_SNAPSHOT_COUNTERS 0x8 + +struct brw_mi_flush +{ + unsigned flags:4; + unsigned pad:12; + unsigned opcode:16; +}; + +struct brw_vf_statistics +{ + unsigned statistics_enable:1; + unsigned pad:15; + unsigned opcode:16; +}; + + + +struct brw_binding_table_pointers +{ + struct header header; + unsigned vs; + unsigned gs; + unsigned clp; + unsigned sf; + unsigned wm; +}; + + +struct brw_blend_constant_color +{ + struct header header; + float blend_constant_color[4]; +}; + + +struct brw_depthbuffer +{ + union header_union header; + + union { + struct { + unsigned pitch:18; + unsigned format:3; + unsigned pad:4; + unsigned depth_offset_disable:1; + unsigned tile_walk:1; + unsigned tiled_surface:1; + unsigned pad2:1; + unsigned surface_type:3; + } bits; + unsigned dword; + } dword1; + + unsigned dword2_base_addr; + + union { + struct { + unsigned pad:1; + unsigned mipmap_layout:1; + unsigned lod:4; + unsigned width:13; + unsigned height:13; + } bits; + unsigned dword; + } dword3; + + union { + struct { + unsigned pad:12; + unsigned min_array_element:9; + unsigned depth:11; + } bits; + unsigned dword; + } dword4; +}; + +struct brw_drawrect +{ + struct header header; + unsigned xmin:16; + unsigned ymin:16; + unsigned xmax:16; + unsigned ymax:16; + unsigned xorg:16; + unsigned yorg:16; +}; + + + + +struct brw_global_depth_offset_clamp +{ + struct header header; + float depth_offset_clamp; +}; + +struct brw_indexbuffer +{ + union { + struct + { + unsigned length:8; + unsigned index_format:2; + unsigned cut_index_enable:1; + unsigned pad:5; + unsigned opcode:16; + } bits; + unsigned dword; + + } header; + + unsigned buffer_start; + unsigned buffer_end; +}; + + +struct brw_line_stipple +{ + struct header header; + + struct + { + unsigned pattern:16; + unsigned pad:16; + } bits0; + + struct + { + unsigned repeat_count:9; + unsigned pad:7; + unsigned inverse_repeat_count:16; + } bits1; +}; + + +struct brw_pipelined_state_pointers +{ + struct header header; + + struct { + unsigned pad:5; + unsigned offset:27; + } vs; + + struct + { + unsigned enable:1; + unsigned pad:4; + unsigned offset:27; + } gs; + + struct + { + unsigned enable:1; + unsigned pad:4; + unsigned offset:27; + } clp; + + struct + { + unsigned pad:5; + unsigned offset:27; + } sf; + + struct + { + unsigned pad:5; + unsigned offset:27; + } wm; + + struct + { + unsigned pad:5; + unsigned offset:27; /* KW: check me! */ + } cc; +}; + + +struct brw_polygon_stipple_offset +{ + struct header header; + + struct { + unsigned y_offset:5; + unsigned pad:3; + unsigned x_offset:5; + unsigned pad0:19; + } bits0; +}; + + + +struct brw_polygon_stipple +{ + struct header header; + unsigned stipple[32]; +}; + + + +struct brw_pipeline_select +{ + struct + { + unsigned pipeline_select:1; + unsigned pad:15; + unsigned opcode:16; + } header; +}; + + +struct brw_pipe_control +{ + struct + { + unsigned length:8; + unsigned notify_enable:1; + unsigned pad:2; + unsigned instruction_state_cache_flush_enable:1; + unsigned write_cache_flush_enable:1; + unsigned depth_stall_enable:1; + unsigned post_sync_operation:2; + + unsigned opcode:16; + } header; + + struct + { + unsigned pad:2; + unsigned dest_addr_type:1; + unsigned dest_addr:29; + } bits1; + + unsigned data0; + unsigned data1; +}; + + +struct brw_urb_fence +{ + struct + { + unsigned length:8; + unsigned vs_realloc:1; + unsigned gs_realloc:1; + unsigned clp_realloc:1; + unsigned sf_realloc:1; + unsigned vfe_realloc:1; + unsigned cs_realloc:1; + unsigned pad:2; + unsigned opcode:16; + } header; + + struct + { + unsigned vs_fence:10; + unsigned gs_fence:10; + unsigned clp_fence:10; + unsigned pad:2; + } bits0; + + struct + { + unsigned sf_fence:10; + unsigned vf_fence:10; + unsigned cs_fence:10; + unsigned pad:2; + } bits1; +}; + +struct brw_constant_buffer_state /* previously brw_command_streamer */ +{ + struct header header; + + struct + { + unsigned nr_urb_entries:3; + unsigned pad:1; + unsigned urb_entry_size:5; + unsigned pad0:23; + } bits0; +}; + +struct brw_constant_buffer +{ + struct + { + unsigned length:8; + unsigned valid:1; + unsigned pad:7; + unsigned opcode:16; + } header; + + struct + { + unsigned buffer_length:6; + unsigned buffer_address:26; + } bits0; +}; + +struct brw_state_base_address +{ + struct header header; + + struct + { + unsigned modify_enable:1; + unsigned pad:4; + unsigned general_state_address:27; + } bits0; + + struct + { + unsigned modify_enable:1; + unsigned pad:4; + unsigned surface_state_address:27; + } bits1; + + struct + { + unsigned modify_enable:1; + unsigned pad:4; + unsigned indirect_object_state_address:27; + } bits2; + + struct + { + unsigned modify_enable:1; + unsigned pad:11; + unsigned general_state_upper_bound:20; + } bits3; + + struct + { + unsigned modify_enable:1; + unsigned pad:11; + unsigned indirect_object_state_upper_bound:20; + } bits4; +}; + +struct brw_state_prefetch +{ + struct header header; + + struct + { + unsigned prefetch_count:3; + unsigned pad:3; + unsigned prefetch_pointer:26; + } bits0; +}; + +struct brw_system_instruction_pointer +{ + struct header header; + + struct + { + unsigned pad:4; + unsigned system_instruction_pointer:28; + } bits0; +}; + + + + +/* State structs for the various fixed function units: + */ + + +struct thread0 +{ + unsigned pad0:1; + unsigned grf_reg_count:3; + unsigned pad1:2; + unsigned kernel_start_pointer:26; +}; + +struct thread1 +{ + unsigned ext_halt_exception_enable:1; + unsigned sw_exception_enable:1; + unsigned mask_stack_exception_enable:1; + unsigned timeout_exception_enable:1; + unsigned illegal_op_exception_enable:1; + unsigned pad0:3; + unsigned depth_coef_urb_read_offset:6; /* WM only */ + unsigned pad1:2; + unsigned floating_point_mode:1; + unsigned thread_priority:1; + unsigned binding_table_entry_count:8; + unsigned pad3:5; + unsigned single_program_flow:1; +}; + +struct thread2 +{ + unsigned per_thread_scratch_space:4; + unsigned pad0:6; + unsigned scratch_space_base_pointer:22; +}; + + +struct thread3 +{ + unsigned dispatch_grf_start_reg:4; + unsigned urb_entry_read_offset:6; + unsigned pad0:1; + unsigned urb_entry_read_length:6; + unsigned pad1:1; + unsigned const_urb_entry_read_offset:6; + unsigned pad2:1; + unsigned const_urb_entry_read_length:6; + unsigned pad3:1; +}; + + + +struct brw_clip_unit_state +{ + struct thread0 thread0; + struct + { + unsigned pad0:7; + unsigned sw_exception_enable:1; + unsigned pad1:3; + unsigned mask_stack_exception_enable:1; + unsigned pad2:1; + unsigned illegal_op_exception_enable:1; + unsigned pad3:2; + unsigned floating_point_mode:1; + unsigned thread_priority:1; + unsigned binding_table_entry_count:8; + unsigned pad4:5; + unsigned single_program_flow:1; + } thread1; + + struct thread2 thread2; + struct thread3 thread3; + + struct + { + unsigned pad0:9; + unsigned gs_output_stats:1; /* not always */ + unsigned stats_enable:1; + unsigned nr_urb_entries:7; + unsigned pad1:1; + unsigned urb_entry_allocation_size:5; + unsigned pad2:1; + unsigned max_threads:1; /* may be less */ + unsigned pad3:6; + } thread4; + + struct + { + unsigned pad0:13; + unsigned clip_mode:3; + unsigned userclip_enable_flags:8; + unsigned userclip_must_clip:1; + unsigned pad1:1; + unsigned guard_band_enable:1; + unsigned viewport_z_clip_enable:1; + unsigned viewport_xy_clip_enable:1; + unsigned vertex_position_space:1; + unsigned api_mode:1; + unsigned pad2:1; + } clip5; + + struct + { + unsigned pad0:5; + unsigned clipper_viewport_state_ptr:27; + } clip6; + + + float viewport_xmin; + float viewport_xmax; + float viewport_ymin; + float viewport_ymax; +}; + + + +struct brw_cc_unit_state +{ + struct + { + unsigned pad0:3; + unsigned bf_stencil_pass_depth_pass_op:3; + unsigned bf_stencil_pass_depth_fail_op:3; + unsigned bf_stencil_fail_op:3; + unsigned bf_stencil_func:3; + unsigned bf_stencil_enable:1; + unsigned pad1:2; + unsigned stencil_write_enable:1; + unsigned stencil_pass_depth_pass_op:3; + unsigned stencil_pass_depth_fail_op:3; + unsigned stencil_fail_op:3; + unsigned stencil_func:3; + unsigned stencil_enable:1; + } cc0; + + + struct + { + unsigned bf_stencil_ref:8; + unsigned stencil_write_mask:8; + unsigned stencil_test_mask:8; + unsigned stencil_ref:8; + } cc1; + + + struct + { + unsigned logicop_enable:1; + unsigned pad0:10; + unsigned depth_write_enable:1; + unsigned depth_test_function:3; + unsigned depth_test:1; + unsigned bf_stencil_write_mask:8; + unsigned bf_stencil_test_mask:8; + } cc2; + + + struct + { + unsigned pad0:8; + unsigned alpha_test_func:3; + unsigned alpha_test:1; + unsigned blend_enable:1; + unsigned ia_blend_enable:1; + unsigned pad1:1; + unsigned alpha_test_format:1; + unsigned pad2:16; + } cc3; + + struct + { + unsigned pad0:5; + unsigned cc_viewport_state_offset:27; + } cc4; + + struct + { + unsigned pad0:2; + unsigned ia_dest_blend_factor:5; + unsigned ia_src_blend_factor:5; + unsigned ia_blend_function:3; + unsigned statistics_enable:1; + unsigned logicop_func:4; + unsigned pad1:11; + unsigned dither_enable:1; + } cc5; + + struct + { + unsigned clamp_post_alpha_blend:1; + unsigned clamp_pre_alpha_blend:1; + unsigned clamp_range:2; + unsigned pad0:11; + unsigned y_dither_offset:2; + unsigned x_dither_offset:2; + unsigned dest_blend_factor:5; + unsigned src_blend_factor:5; + unsigned blend_function:3; + } cc6; + + struct { + union { + float f; + ubyte ub[4]; + } alpha_ref; + } cc7; +}; + + + +struct brw_sf_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct + { + unsigned pad0:10; + unsigned stats_enable:1; + unsigned nr_urb_entries:7; + unsigned pad1:1; + unsigned urb_entry_allocation_size:5; + unsigned pad2:1; + unsigned max_threads:6; + unsigned pad3:1; + } thread4; + + struct + { + unsigned front_winding:1; + unsigned viewport_transform:1; + unsigned pad0:3; + unsigned sf_viewport_state_offset:27; + } sf5; + + struct + { + unsigned pad0:9; + unsigned dest_org_vbias:4; + unsigned dest_org_hbias:4; + unsigned scissor:1; + unsigned disable_2x2_trifilter:1; + unsigned disable_zero_pix_trifilter:1; + unsigned point_rast_rule:2; + unsigned line_endcap_aa_region_width:2; + unsigned line_width:4; + unsigned fast_scissor_disable:1; + unsigned cull_mode:2; + unsigned aa_enable:1; + } sf6; + + struct + { + unsigned point_size:11; + unsigned use_point_size_state:1; + unsigned subpixel_precision:1; + unsigned sprite_point:1; + unsigned pad0:11; + unsigned trifan_pv:2; + unsigned linestrip_pv:2; + unsigned tristrip_pv:2; + unsigned line_last_pixel_enable:1; + } sf7; + +}; + + +struct brw_gs_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct + { + unsigned pad0:10; + unsigned stats_enable:1; + unsigned nr_urb_entries:7; + unsigned pad1:1; + unsigned urb_entry_allocation_size:5; + unsigned pad2:1; + unsigned max_threads:1; + unsigned pad3:6; + } thread4; + + struct + { + unsigned sampler_count:3; + unsigned pad0:2; + unsigned sampler_state_pointer:27; + } gs5; + + + struct + { + unsigned max_vp_index:4; + unsigned pad0:26; + unsigned reorder_enable:1; + unsigned pad1:1; + } gs6; +}; + + +struct brw_vs_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct + { + unsigned pad0:10; + unsigned stats_enable:1; + unsigned nr_urb_entries:7; + unsigned pad1:1; + unsigned urb_entry_allocation_size:5; + unsigned pad2:1; + unsigned max_threads:4; + unsigned pad3:3; + } thread4; + + struct + { + unsigned sampler_count:3; + unsigned pad0:2; + unsigned sampler_state_pointer:27; + } vs5; + + struct + { + unsigned vs_enable:1; + unsigned vert_cache_disable:1; + unsigned pad0:30; + } vs6; +}; + + +struct brw_wm_unit_state +{ + struct thread0 thread0; + struct thread1 thread1; + struct thread2 thread2; + struct thread3 thread3; + + struct { + unsigned stats_enable:1; + unsigned pad0:1; + unsigned sampler_count:3; + unsigned sampler_state_pointer:27; + } wm4; + + struct + { + unsigned enable_8_pix:1; + unsigned enable_16_pix:1; + unsigned enable_32_pix:1; + unsigned pad0:7; + unsigned legacy_global_depth_bias:1; + unsigned line_stipple:1; + unsigned depth_offset:1; + unsigned polygon_stipple:1; + unsigned line_aa_region_width:2; + unsigned line_endcap_aa_region_width:2; + unsigned early_depth_test:1; + unsigned thread_dispatch_enable:1; + unsigned program_uses_depth:1; + unsigned program_computes_depth:1; + unsigned program_uses_killpixel:1; + unsigned legacy_line_rast: 1; + unsigned pad1:1; + unsigned max_threads:6; + unsigned pad2:1; + } wm5; + + float global_depth_offset_constant; + float global_depth_offset_scale; +}; + +struct brw_sampler_default_color { + float color[4]; +}; + +struct brw_sampler_state +{ + + struct + { + unsigned shadow_function:3; + unsigned lod_bias:11; + unsigned min_filter:3; + unsigned mag_filter:3; + unsigned mip_filter:2; + unsigned base_level:5; + unsigned pad:1; + unsigned lod_preclamp:1; + unsigned default_color_mode:1; + unsigned pad0:1; + unsigned disable:1; + } ss0; + + struct + { + unsigned r_wrap_mode:3; + unsigned t_wrap_mode:3; + unsigned s_wrap_mode:3; + unsigned pad:3; + unsigned max_lod:10; + unsigned min_lod:10; + } ss1; + + + struct + { + unsigned pad:5; + unsigned default_color_pointer:27; + } ss2; + + struct + { + unsigned pad:19; + unsigned max_aniso:3; + unsigned chroma_key_mode:1; + unsigned chroma_key_index:2; + unsigned chroma_key_enable:1; + unsigned monochrome_filter_width:3; + unsigned monochrome_filter_height:3; + } ss3; +}; + + +struct brw_clipper_viewport +{ + float xmin; + float xmax; + float ymin; + float ymax; +}; + +struct brw_cc_viewport +{ + float min_depth; + float max_depth; +}; + +struct brw_sf_viewport +{ + struct { + float m00; + float m11; + float m22; + float m30; + float m31; + float m32; + } viewport; + + struct { + short xmin; + short ymin; + short xmax; + short ymax; + } scissor; +}; + +/* Documented in the subsystem/shared-functions/sampler chapter... + */ +struct brw_surface_state +{ + struct { + unsigned cube_pos_z:1; + unsigned cube_neg_z:1; + unsigned cube_pos_y:1; + unsigned cube_neg_y:1; + unsigned cube_pos_x:1; + unsigned cube_neg_x:1; + unsigned pad:4; + unsigned mipmap_layout_mode:1; + unsigned vert_line_stride_ofs:1; + unsigned vert_line_stride:1; + unsigned color_blend:1; + unsigned writedisable_blue:1; + unsigned writedisable_green:1; + unsigned writedisable_red:1; + unsigned writedisable_alpha:1; + unsigned surface_format:9; + unsigned data_return_format:1; + unsigned pad0:1; + unsigned surface_type:3; + } ss0; + + struct { + unsigned base_addr; + } ss1; + + struct { + unsigned pad:2; + unsigned mip_count:4; + unsigned width:13; + unsigned height:13; + } ss2; + + struct { + unsigned tile_walk:1; + unsigned tiled_surface:1; + unsigned pad:1; + unsigned pitch:18; + unsigned depth:11; + } ss3; + + struct { + unsigned pad:19; + unsigned min_array_elt:9; + unsigned min_lod:4; + } ss4; +}; + + + +struct brw_vertex_buffer_state +{ + struct { + unsigned pitch:11; + unsigned pad:15; + unsigned access_type:1; + unsigned vb_index:5; + } vb0; + + unsigned start_addr; + unsigned max_index; +#if 1 + unsigned instance_data_step_rate; /* not included for sequential/random vertices? */ +#endif +}; + +#define BRW_VBP_MAX 17 + +struct brw_vb_array_state { + struct header header; + struct brw_vertex_buffer_state vb[BRW_VBP_MAX]; +}; + + +struct brw_vertex_element_state +{ + struct + { + unsigned src_offset:11; + unsigned pad:5; + unsigned src_format:9; + unsigned pad0:1; + unsigned valid:1; + unsigned vertex_buffer_index:5; + } ve0; + + struct + { + unsigned dst_offset:8; + unsigned pad:8; + unsigned vfcomponent3:4; + unsigned vfcomponent2:4; + unsigned vfcomponent1:4; + unsigned vfcomponent0:4; + } ve1; +}; + +#define BRW_VEP_MAX 18 + +struct brw_vertex_element_packet { + struct header header; + struct brw_vertex_element_state ve[BRW_VEP_MAX]; /* note: less than _TNL_ATTRIB_MAX */ +}; + + +struct brw_urb_immediate { + unsigned opcode:4; + unsigned offset:6; + unsigned swizzle_control:2; + unsigned pad:1; + unsigned allocate:1; + unsigned used:1; + unsigned complete:1; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; +}; + +/* Instruction format for the execution units: + */ + +struct brw_instruction +{ + struct + { + unsigned opcode:7; + unsigned pad:1; + unsigned access_mode:1; + unsigned mask_control:1; + unsigned dependency_control:2; + unsigned compression_control:2; + unsigned thread_control:2; + unsigned predicate_control:4; + unsigned predicate_inverse:1; + unsigned execution_size:3; + unsigned destreg__conditonalmod:4; /* destreg - send, conditionalmod - others */ + unsigned pad0:2; + unsigned debug_control:1; + unsigned saturate:1; + } header; + + union { + struct + { + unsigned dest_reg_file:2; + unsigned dest_reg_type:3; + unsigned src0_reg_file:2; + unsigned src0_reg_type:3; + unsigned src1_reg_file:2; + unsigned src1_reg_type:3; + unsigned pad:1; + unsigned dest_subreg_nr:5; + unsigned dest_reg_nr:8; + unsigned dest_horiz_stride:2; + unsigned dest_address_mode:1; + } da1; + + struct + { + unsigned dest_reg_file:2; + unsigned dest_reg_type:3; + unsigned src0_reg_file:2; + unsigned src0_reg_type:3; + unsigned pad:6; + int dest_indirect_offset:10; /* offset against the deref'd address reg */ + unsigned dest_subreg_nr:3; /* subnr for the address reg a0.x */ + unsigned dest_horiz_stride:2; + unsigned dest_address_mode:1; + } ia1; + + struct + { + unsigned dest_reg_file:2; + unsigned dest_reg_type:3; + unsigned src0_reg_file:2; + unsigned src0_reg_type:3; + unsigned src1_reg_file:2; + unsigned src1_reg_type:3; + unsigned pad0:1; + unsigned dest_writemask:4; + unsigned dest_subreg_nr:1; + unsigned dest_reg_nr:8; + unsigned pad1:2; + unsigned dest_address_mode:1; + } da16; + + struct + { + unsigned dest_reg_file:2; + unsigned dest_reg_type:3; + unsigned src0_reg_file:2; + unsigned src0_reg_type:3; + unsigned pad0:6; + unsigned dest_writemask:4; + int dest_indirect_offset:6; + unsigned dest_subreg_nr:3; + unsigned pad1:2; + unsigned dest_address_mode:1; + } ia16; + } bits1; + + + union { + struct + { + unsigned src0_subreg_nr:5; + unsigned src0_reg_nr:8; + unsigned src0_abs:1; + unsigned src0_negate:1; + unsigned src0_address_mode:1; + unsigned src0_horiz_stride:2; + unsigned src0_width:3; + unsigned src0_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad:6; + } da1; + + struct + { + int src0_indirect_offset:10; + unsigned src0_subreg_nr:3; + unsigned src0_abs:1; + unsigned src0_negate:1; + unsigned src0_address_mode:1; + unsigned src0_horiz_stride:2; + unsigned src0_width:3; + unsigned src0_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad:6; + } ia1; + + struct + { + unsigned src0_swz_x:2; + unsigned src0_swz_y:2; + unsigned src0_subreg_nr:1; + unsigned src0_reg_nr:8; + unsigned src0_abs:1; + unsigned src0_negate:1; + unsigned src0_address_mode:1; + unsigned src0_swz_z:2; + unsigned src0_swz_w:2; + unsigned pad0:1; + unsigned src0_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad1:6; + } da16; + + struct + { + unsigned src0_swz_x:2; + unsigned src0_swz_y:2; + int src0_indirect_offset:6; + unsigned src0_subreg_nr:3; + unsigned src0_abs:1; + unsigned src0_negate:1; + unsigned src0_address_mode:1; + unsigned src0_swz_z:2; + unsigned src0_swz_w:2; + unsigned pad0:1; + unsigned src0_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad1:6; + } ia16; + + } bits2; + + union + { + struct + { + unsigned src1_subreg_nr:5; + unsigned src1_reg_nr:8; + unsigned src1_abs:1; + unsigned src1_negate:1; + unsigned pad:1; + unsigned src1_horiz_stride:2; + unsigned src1_width:3; + unsigned src1_vert_stride:4; + unsigned pad0:7; + } da1; + + struct + { + unsigned src1_swz_x:2; + unsigned src1_swz_y:2; + unsigned src1_subreg_nr:1; + unsigned src1_reg_nr:8; + unsigned src1_abs:1; + unsigned src1_negate:1; + unsigned pad0:1; + unsigned src1_swz_z:2; + unsigned src1_swz_w:2; + unsigned pad1:1; + unsigned src1_vert_stride:4; + unsigned pad2:7; + } da16; + + struct + { + int src1_indirect_offset:10; + unsigned src1_subreg_nr:3; + unsigned src1_abs:1; + unsigned src1_negate:1; + unsigned pad0:1; + unsigned src1_horiz_stride:2; + unsigned src1_width:3; + unsigned src1_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad1:6; + } ia1; + + struct + { + unsigned src1_swz_x:2; + unsigned src1_swz_y:2; + int src1_indirect_offset:6; + unsigned src1_subreg_nr:3; + unsigned src1_abs:1; + unsigned src1_negate:1; + unsigned pad0:1; + unsigned src1_swz_z:2; + unsigned src1_swz_w:2; + unsigned pad1:1; + unsigned src1_vert_stride:4; + unsigned flag_reg_nr:1; + unsigned pad2:6; + } ia16; + + + struct + { + int jump_count:16; /* note: signed */ + unsigned pop_count:4; + unsigned pad0:12; + } if_else; + + struct { + unsigned function:4; + unsigned int_type:1; + unsigned precision:1; + unsigned saturate:1; + unsigned data_type:1; + unsigned pad0:8; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } math; + + struct { + unsigned binding_table_index:8; + unsigned sampler:4; + unsigned return_format:2; + unsigned msg_type:2; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } sampler; + + struct brw_urb_immediate urb; + + struct { + unsigned binding_table_index:8; + unsigned msg_control:4; + unsigned msg_type:2; + unsigned target_cache:2; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } dp_read; + + struct { + unsigned binding_table_index:8; + unsigned msg_control:3; + unsigned pixel_scoreboard_clear:1; + unsigned msg_type:3; + unsigned send_commit_msg:1; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } dp_write; + + struct { + unsigned pad:16; + unsigned response_length:4; + unsigned msg_length:4; + unsigned msg_target:4; + unsigned pad1:3; + unsigned end_of_thread:1; + } generic; + + int d; + unsigned ud; + } bits3; +}; + + +#endif diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c new file mode 100644 index 0000000000..518845e4b2 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -0,0 +1,210 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "brw_blit.h" +#include "brw_context.h" +#include "brw_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "pipe/util/p_tile.h" + + +/* + * XXX note: same as code in sp_surface.c + */ +static struct pipe_surface * +brw_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct brw_texture *tex = (struct brw_texture *)pt; + struct pipe_surface *ps; + unsigned offset; /* in bytes */ + + offset = tex->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE) { + offset += tex->image_offset[level][face] * pt->cpp; + } + else if (pt->target == PIPE_TEXTURE_3D) { + offset += tex->image_offset[level][zslice] * pt->cpp; + } + else { + assert(face == 0); + assert(zslice == 0); + } + + ps = pipe->winsys->surface_alloc(pipe->winsys); + if (ps) { + assert(ps->format); + assert(ps->refcount); + pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = tex->pitch; + ps->offset = offset; + } + return ps; +} + + +/* Upload data to a rectangular sub-region. Lots of choices how to do this: + * + * - memcpy by span to current destination + * - upload data as new buffer and blit + * + * Currently always memcpy. + */ +static void +brw_surface_data(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + const void *src, unsigned src_pitch, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + pipe_copy_rect(pipe_surface_map(dst) + dst->offset, + dst->cpp, dst->pitch, + dstx, dsty, width, height, src, src_pitch, srcx, srcy); + + pipe_surface_unmap(dst); +} + + +/* Assumes all values are within bounds -- no checking at this level - + * do it higher up if required. + */ +static void +brw_surface_copy(struct pipe_context *pipe, + unsigned do_flip, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + assert(dst != src); + assert(dst->cpp == src->cpp); + + if (0) { + pipe_copy_rect(pipe_surface_map(dst) + dst->offset, + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe_surface_map(src) + src->offset, + do_flip ? -src->pitch : src->pitch, + srcx, do_flip ? 1 - srcy - height : srcy); + + pipe_surface_unmap(src); + pipe_surface_unmap(dst); + } + else { + brw_copy_blit(brw_context(pipe), + do_flip, + dst->cpp, + (short) src->pitch, src->buffer, src->offset, FALSE, + (short) dst->pitch, dst->buffer, dst->offset, FALSE, + (short) srcx, (short) srcy, (short) dstx, (short) dsty, + (short) width, (short) height, PIPE_LOGICOP_COPY); + } +} + +/* Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +static void * +get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) +{ + return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; +} + + +static void +brw_surface_fill(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) +{ + if (0) { + unsigned i, j; + void *dst_map = pipe_surface_map(dst); + + switch (dst->cpp) { + case 1: { + ubyte *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: { + ushort *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (ushort) value; + row += dst->pitch; + } + } + break; + case 4: { + unsigned *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + default: + assert(0); + break; + } + + pipe_surface_unmap( dst ); + } + else { + brw_fill_blit(brw_context(pipe), + dst->cpp, + (short) dst->pitch, + dst->buffer, dst->offset, FALSE, + (short) dstx, (short) dsty, + (short) width, (short) height, + value); + } +} + +void +brw_init_surface_functions(struct brw_context *brw) +{ + brw->pipe.get_tex_surface = brw_get_tex_surface; + brw->pipe.surface_copy = brw_surface_copy; + brw->pipe.surface_fill = brw_surface_fill; +} diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c new file mode 100644 index 0000000000..90561f1307 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -0,0 +1,353 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +/* Code to layout images in a mipmap tree for i965. + */ + +#include "brw_tex_layout.h" + +#include "pipe/p_state.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" + +#include "brw_context.h" + +#define FILE_DEBUG_FLAG DEBUG_TEXTURE + +#if 0 +unsigned intel_compressed_alignment(unsigned internalFormat) +{ + unsigned alignment = 4; + + switch (internalFormat) { + case GL_COMPRESSED_RGB_FXT1_3DFX: + case GL_COMPRESSED_RGBA_FXT1_3DFX: + alignment = 8; + break; + + default: + break; + } + + return alignment; +} +#endif + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + + +static boolean brw_miptree_layout(struct pipe_context *, struct brw_texture *); + +static void intel_miptree_set_image_offset(struct brw_texture *tex, + unsigned level, + unsigned img, + unsigned x, unsigned y) +{ + struct pipe_texture *pt = &tex->base; + if (img == 0 && level == 0) + assert(x == 0 && y == 0); + assert(img < tex->nr_images[level]); + + tex->image_offset[level][img] = (x + y * tex->pitch) * pt->cpp; +} + +static void intel_miptree_set_level_info(struct brw_texture *tex, + unsigned level, + unsigned nr_images, + unsigned x, unsigned y, + unsigned w, unsigned h, unsigned d) +{ + struct pipe_texture *pt = &tex->base; + + assert(level < PIPE_MAX_TEXTURE_LEVELS); + + pt->width[level] = w; + pt->height[level] = h; + pt->depth[level] = d; + + tex->level_offset[level] = (x + y * tex->pitch) * pt->cpp; + tex->nr_images[level] = nr_images; + + /* + DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + level, w, h, d, x, y, tex->level_offset[level]); + */ + + /* Not sure when this would happen, but anyway: + */ + if (tex->image_offset[level]) { + FREE(tex->image_offset[level]); + tex->image_offset[level] = NULL; + } + + assert(nr_images); + assert(!tex->image_offset[level]); + + tex->image_offset[level] = (unsigned *) MALLOC(nr_images * sizeof(unsigned)); + tex->image_offset[level][0] = 0; +} + +static void i945_miptree_layout_2d(struct brw_texture *tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned align_h = 2, align_w = 4; + unsigned level; + unsigned x = 0; + unsigned y = 0; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + + tex->pitch = pt->width[0]; + +#if 0 + if (pt->compressed) { + align_w = intel_compressed_alignment(pt->internal_format); + tex->pitch = ALIGN(pt->width[0], align_w); + } +#endif + + /* May need to adjust pitch to accomodate the placement of + * the 2nd mipmap. This occurs when the alignment + * constraints of mipmap placement push the right edge of the + * 2nd mipmap out past the width of its parent. + */ + if (pt->last_level > 0) { + unsigned mip1_width; + + if (pt->compressed) { + mip1_width = align(minify(pt->width[0]), align_w) + + align(minify(minify(pt->width[0])), align_w); + } else { + mip1_width = align(minify(pt->width[0]), align_w) + + minify(minify(pt->width[0])); + } + + if (mip1_width > tex->pitch) { + tex->pitch = mip1_width; + } + } + + /* Pitch must be a whole number of dwords, even though we + * express it in texels. + */ + tex->pitch = align(tex->pitch * pt->cpp, 4) / pt->cpp; + tex->total_height = 0; + + for (level = 0; level <= pt->last_level; level++) { + unsigned img_height; + + intel_miptree_set_level_info(tex, level, 1, x, y, width, + height, 1); + + if (pt->compressed) + img_height = MAX2(1, height/4); + else + img_height = align(height, align_h); + + + /* Because the images are packed better, the final offset + * might not be the maximal one: + */ + tex->total_height = MAX2(tex->total_height, y + img_height); + + /* Layout_below: step right after second mipmap. + */ + if (level == 1) { + x += align(width, align_w); + } + else { + y += img_height; + } + + width = minify(width); + height = minify(height); + } +} + +static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture *tex) +{ + struct pipe_texture *pt = &tex->base; + /* XXX: these vary depending on image format: + */ +/* int align_w = 4; */ + + switch (pt->target) { + case PIPE_TEXTURE_CUBE: + case PIPE_TEXTURE_3D: { + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + unsigned pack_x_pitch, pack_x_nr; + unsigned pack_y_pitch; + unsigned level; + unsigned align_h = 2; + unsigned align_w = 4; + + tex->total_height = 0; +#if 0 + if (pt->compressed) { + align_w = intel_compressed_alignment(pt->internal_format); + pt->pitch = align(width, align_w); + pack_y_pitch = (height + 3) / 4; + } else +#endif + { + tex->pitch = align(pt->width[0] * pt->cpp, 4) / pt->cpp; + pack_y_pitch = align(pt->height[0], align_h); + } + + pack_x_pitch = tex->pitch; + pack_x_nr = 1; + + for (level = 0; level <= pt->last_level; level++) { + unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; + int x = 0; + int y = 0; + uint q, j; + + intel_miptree_set_level_info(tex, level, nr_images, + 0, tex->total_height, + width, height, depth); + + for (q = 0; q < nr_images;) { + for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { + intel_miptree_set_image_offset(tex, level, q, x, y); + x += pack_x_pitch; + } + + x = 0; + y += pack_y_pitch; + } + + + tex->total_height += y; + width = minify(width); + height = minify(height); + depth = minify(depth); + + if (pt->compressed) { + pack_y_pitch = (height + 3) / 4; + + if (pack_x_pitch > align(width, align_w)) { + pack_x_pitch = align(width, align_w); + pack_x_nr <<= 1; + } + } else { + if (pack_x_pitch > 4) { + pack_x_pitch >>= 1; + pack_x_nr <<= 1; + assert(pack_x_pitch * pack_x_nr <= tex->pitch); + } + + if (pack_y_pitch > 2) { + pack_y_pitch >>= 1; + pack_y_pitch = align(pack_y_pitch, align_h); + } + } + + } + break; + } + + default: + i945_miptree_layout_2d(tex); + break; + } +#if 0 + PRINT("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, + pt->pitch, + pt->total_height, + pt->cpp, + pt->pitch * pt->total_height * pt->cpp ); +#endif + + return TRUE; +} + + +struct pipe_texture * +brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) +{ + struct brw_texture *tex = CALLOC_STRUCT(brw_texture); + + if (tex) { + tex->base = *templat; + + if (brw_miptree_layout(pipe, tex)) + tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, + PIPE_BUFFER_USAGE_PIXEL, + tex->pitch * tex->base.cpp * + tex->total_height); + + if (!tex->buffer) { + FREE(tex); + return NULL; + } + } + + return &tex->base; +} + +void +brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct brw_texture *tex = (struct brw_texture *)*pt; + uint i; + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); + */ + + pipe_buffer_reference(pipe->winsys, &tex->buffer, NULL); + + for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) + if (tex->image_offset[i]) + free(tex->image_offset[i]); + + free(tex); + } + *pt = NULL; +} diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.h b/src/gallium/drivers/i965simple/brw_tex_layout.h new file mode 100644 index 0000000000..cfd6b1ef3a --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_tex_layout.h @@ -0,0 +1,15 @@ +#ifndef BRW_TEX_LAYOUT_H +#define BRW_TEX_LAYOUT_H + +#include "pipe/p_compiler.h" + +struct pipe_context; +struct pipe_texture; + +extern struct pipe_texture * +brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat); + +extern void +brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); + +#endif diff --git a/src/gallium/drivers/i965simple/brw_urb.c b/src/gallium/drivers/i965simple/brw_urb.c new file mode 100644 index 0000000000..101a4367b9 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_urb.c @@ -0,0 +1,186 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +//#include "brw_state.h" +#include "brw_batch.h" +#include "brw_defines.h" + +#define VS 0 +#define GS 1 +#define CLP 2 +#define SF 3 +#define CS 4 + +/* XXX: Are the min_entry_size numbers useful? + * XXX: Verify min_nr_entries, esp for VS. + * XXX: Verify SF min_entry_size. + */ +static const struct { + unsigned min_nr_entries; + unsigned preferred_nr_entries; + unsigned min_entry_size; + unsigned max_entry_size; +} limits[CS+1] = { + { 8, 32, 1, 5 }, /* vs */ + { 4, 8, 1, 5 }, /* gs */ + { 6, 8, 1, 5 }, /* clp */ + { 1, 8, 1, 12 }, /* sf */ + { 1, 4, 1, 32 } /* cs */ +}; + + +static boolean check_urb_layout( struct brw_context *brw ) +{ + brw->urb.vs_start = 0; + brw->urb.gs_start = brw->urb.nr_vs_entries * brw->urb.vsize; + brw->urb.clip_start = brw->urb.gs_start + brw->urb.nr_gs_entries * brw->urb.vsize; + brw->urb.sf_start = brw->urb.clip_start + brw->urb.nr_clip_entries * brw->urb.vsize; + brw->urb.cs_start = brw->urb.sf_start + brw->urb.nr_sf_entries * brw->urb.sfsize; + + return brw->urb.cs_start + brw->urb.nr_cs_entries * brw->urb.csize <= 256; +} + +/* Most minimal update, forces re-emit of URB fence packet after GS + * unit turned on/off. + */ +static void recalculate_urb_fence( struct brw_context *brw ) +{ + unsigned csize = brw->curbe.total_size; + unsigned vsize = brw->vs.prog_data->urb_entry_size; + unsigned sfsize = brw->sf.prog_data->urb_entry_size; + + if (csize < limits[CS].min_entry_size) + csize = limits[CS].min_entry_size; + + if (vsize < limits[VS].min_entry_size) + vsize = limits[VS].min_entry_size; + + if (sfsize < limits[SF].min_entry_size) + sfsize = limits[SF].min_entry_size; + + if (brw->urb.vsize < vsize || + brw->urb.sfsize < sfsize || + brw->urb.csize < csize || + (brw->urb.constrained && (brw->urb.vsize > brw->urb.vsize || + brw->urb.sfsize > brw->urb.sfsize || + brw->urb.csize > brw->urb.csize))) { + + + brw->urb.csize = csize; + brw->urb.sfsize = sfsize; + brw->urb.vsize = vsize; + + brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries; + brw->urb.nr_gs_entries = limits[GS].preferred_nr_entries; + brw->urb.nr_clip_entries = limits[CLP].preferred_nr_entries; + brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries; + brw->urb.nr_cs_entries = limits[CS].preferred_nr_entries; + + if (!check_urb_layout(brw)) { + brw->urb.nr_vs_entries = limits[VS].min_nr_entries; + brw->urb.nr_gs_entries = limits[GS].min_nr_entries; + brw->urb.nr_clip_entries = limits[CLP].min_nr_entries; + brw->urb.nr_sf_entries = limits[SF].min_nr_entries; + brw->urb.nr_cs_entries = limits[CS].min_nr_entries; + + brw->urb.constrained = 1; + + if (!check_urb_layout(brw)) { + /* This is impossible, given the maximal sizes of urb + * entries and the values for minimum nr of entries + * provided above. + */ + debug_printf("couldn't calculate URB layout!\n"); + exit(1); + } + + if (BRW_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS)) + debug_printf("URB CONSTRAINED\n"); + } + else + brw->urb.constrained = 0; + + if (BRW_DEBUG & DEBUG_URB) + debug_printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n", + brw->urb.vs_start, + brw->urb.gs_start, + brw->urb.clip_start, + brw->urb.sf_start, + brw->urb.cs_start, + 256); + + brw->state.dirty.brw |= BRW_NEW_URB_FENCE; + } +} + + +const struct brw_tracked_state brw_recalculate_urb_fence = { + .dirty = { + .brw = BRW_NEW_CURBE_OFFSETS, + .cache = (CACHE_NEW_VS_PROG | + CACHE_NEW_SF_PROG) + }, + .update = recalculate_urb_fence +}; + + + + + +void brw_upload_urb_fence(struct brw_context *brw) +{ + struct brw_urb_fence uf; + memset(&uf, 0, sizeof(uf)); + + uf.header.opcode = CMD_URB_FENCE; + uf.header.length = sizeof(uf)/4-2; + uf.header.vs_realloc = 1; + uf.header.gs_realloc = 1; + uf.header.clp_realloc = 1; + uf.header.sf_realloc = 1; + uf.header.vfe_realloc = 1; + uf.header.cs_realloc = 1; + + /* The ordering below is correct, not the layout in the + * instruction. + * + * There are 256 urb reg pairs in total. + */ + uf.bits0.vs_fence = brw->urb.gs_start; + uf.bits0.gs_fence = brw->urb.clip_start; + uf.bits0.clp_fence = brw->urb.sf_start; + uf.bits1.sf_fence = brw->urb.cs_start; + uf.bits1.cs_fence = 256; + + BRW_BATCH_STRUCT(brw, &uf); +} diff --git a/src/gallium/drivers/i965simple/brw_util.c b/src/gallium/drivers/i965simple/brw_util.c new file mode 100644 index 0000000000..42391d7c8c --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_util.c @@ -0,0 +1,104 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_util.h" +#include "brw_defines.h" + +#include "pipe/p_defines.h" + +unsigned brw_count_bits( unsigned val ) +{ + unsigned i; + for (i = 0; val ; val >>= 1) + if (val & 1) + i++; + return i; +} + + +unsigned brw_translate_blend_equation( int mode ) +{ + switch (mode) { + case PIPE_BLEND_ADD: + return BRW_BLENDFUNCTION_ADD; + case PIPE_BLEND_MIN: + return BRW_BLENDFUNCTION_MIN; + case PIPE_BLEND_MAX: + return BRW_BLENDFUNCTION_MAX; + case PIPE_BLEND_SUBTRACT: + return BRW_BLENDFUNCTION_SUBTRACT; + case PIPE_BLEND_REVERSE_SUBTRACT: + return BRW_BLENDFUNCTION_REVERSE_SUBTRACT; + default: + assert(0); + return BRW_BLENDFUNCTION_ADD; + } +} + +unsigned brw_translate_blend_factor( int factor ) +{ + switch(factor) { + case PIPE_BLENDFACTOR_ZERO: + return BRW_BLENDFACTOR_ZERO; + case PIPE_BLENDFACTOR_SRC_ALPHA: + return BRW_BLENDFACTOR_SRC_ALPHA; + case PIPE_BLENDFACTOR_ONE: + return BRW_BLENDFACTOR_ONE; + case PIPE_BLENDFACTOR_SRC_COLOR: + return BRW_BLENDFACTOR_SRC_COLOR; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + return BRW_BLENDFACTOR_INV_SRC_COLOR; + case PIPE_BLENDFACTOR_DST_COLOR: + return BRW_BLENDFACTOR_DST_COLOR; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + return BRW_BLENDFACTOR_INV_DST_COLOR; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + return BRW_BLENDFACTOR_INV_SRC_ALPHA; + case PIPE_BLENDFACTOR_DST_ALPHA: + return BRW_BLENDFACTOR_DST_ALPHA; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + return BRW_BLENDFACTOR_INV_DST_ALPHA; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + return BRW_BLENDFACTOR_SRC_ALPHA_SATURATE; + case PIPE_BLENDFACTOR_CONST_COLOR: + return BRW_BLENDFACTOR_CONST_COLOR; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + return BRW_BLENDFACTOR_INV_CONST_COLOR; + case PIPE_BLENDFACTOR_CONST_ALPHA: + return BRW_BLENDFACTOR_CONST_ALPHA; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + return BRW_BLENDFACTOR_INV_CONST_ALPHA; + default: + assert(0); + return BRW_BLENDFACTOR_ZERO; + } +} diff --git a/src/gallium/drivers/i965simple/brw_util.h b/src/gallium/drivers/i965simple/brw_util.h new file mode 100644 index 0000000000..d60e5934db --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_util.h @@ -0,0 +1,43 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_UTIL_H +#define BRW_UTIL_H + +#include "pipe/p_state.h" + +extern unsigned brw_count_bits( unsigned val ); +extern unsigned brw_translate_blend_factor( int factor ); +extern unsigned brw_translate_blend_equation( int mode ); + + +#endif diff --git a/src/gallium/drivers/i965simple/brw_vs.c b/src/gallium/drivers/i965simple/brw_vs.c new file mode 100644 index 0000000000..738c6346d5 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_vs.c @@ -0,0 +1,120 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_vs.h" +#include "brw_util.h" +#include "brw_state.h" + + +static void do_vs_prog( struct brw_context *brw, + const struct brw_vertex_program *vp, + struct brw_vs_prog_key *key ) +{ + unsigned program_size; + const unsigned *program; + struct brw_vs_compile c; + + memset(&c, 0, sizeof(c)); + memcpy(&c.key, key, sizeof(*key)); + + brw_init_compile(&c.func); + c.vp = vp; + + c.prog_data.outputs_written = vp->program.num_outputs; + c.prog_data.inputs_read = vp->program.num_inputs; + +#if 0 + if (c.key.copy_edgeflag) { + c.prog_data.outputs_written |= 1<vs.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_VS_PROG], + &c.key, + sizeof(c.key), + program, + program_size, + &c.prog_data, + &brw->vs.prog_data); +} + + +static void brw_upload_vs_prog( struct brw_context *brw ) +{ + struct brw_vs_prog_key key; + const struct brw_vertex_program *vp = brw->attribs.VertexProgram; + + assert(vp); + + memset(&key, 0, sizeof(key)); + + /* Just upload the program verbatim for now. Always send it all + * the inputs it asks for, whether they are varying or not. + */ + key.program_string_id = vp->id; + key.nr_userclip = brw->attribs.Clip.nr; + key.copy_edgeflag = (brw->attribs.Raster->fill_cw != PIPE_POLYGON_MODE_FILL || + brw->attribs.Raster->fill_ccw != PIPE_POLYGON_MODE_FILL); + + /* Make an early check for the key. + */ + if (brw_search_cache(&brw->cache[BRW_VS_PROG], + &key, sizeof(key), + &brw->vs.prog_data, + &brw->vs.prog_gs_offset)) + return; + + do_vs_prog(brw, vp, &key); +} + + +/* See brw_vs.c: + */ +const struct brw_tracked_state brw_vs_prog = { + .dirty = { + .brw = BRW_NEW_VS, + .cache = 0 + }, + .update = brw_upload_vs_prog +}; diff --git a/src/gallium/drivers/i965simple/brw_vs.h b/src/gallium/drivers/i965simple/brw_vs.h new file mode 100644 index 0000000000..0e58f043b0 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_vs.h @@ -0,0 +1,82 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_VS_H +#define BRW_VS_H + + +#include "brw_context.h" +#include "brw_eu.h" + + +struct brw_vs_prog_key { + unsigned program_string_id; + unsigned nr_userclip:4; + unsigned copy_edgeflag:1; + unsigned know_w_is_one:1; + unsigned pad:26; +}; + + +struct brw_vs_compile { + struct brw_compile func; + struct brw_vs_prog_key key; + struct brw_vs_prog_data prog_data; + + struct brw_vertex_program *vp; + + unsigned nr_inputs; + + unsigned first_output; + unsigned nr_outputs; + + unsigned first_tmp; + unsigned last_tmp; + + struct brw_reg r0; + struct brw_reg r1; + struct brw_reg regs[12][128]; + struct brw_reg tmp; + struct brw_reg stack; + + struct { + boolean used_in_src; + struct brw_reg reg; + } output_regs[128]; + + struct brw_reg userplane[6]; + +}; + +void brw_vs_emit( struct brw_vs_compile *c ); + +#endif diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c new file mode 100644 index 0000000000..98915ba101 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_vs_emit.c @@ -0,0 +1,1332 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_context.h" +#include "brw_vs.h" + +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" + +struct brw_prog_info { + unsigned num_temps; + unsigned num_addrs; + unsigned num_consts; + + unsigned writes_psize; + + unsigned pos_idx; + unsigned result_edge_idx; + unsigned edge_flag_idx; + unsigned psize_idx; +}; + +/* Do things as simply as possible. Allocate and populate all regs + * ahead of time. + */ +static void brw_vs_alloc_regs( struct brw_vs_compile *c, + struct brw_prog_info *info ) +{ + unsigned i, reg = 0, mrf; + unsigned nr_params; + + /* r0 -- reserved as usual + */ + c->r0 = brw_vec8_grf(reg, 0); reg++; + + /* User clip planes from curbe: + */ + if (c->key.nr_userclip) { + for (i = 0; i < c->key.nr_userclip; i++) { + c->userplane[i] = stride( brw_vec4_grf(reg+3+i/2, (i%2) * 4), 0, 4, 1); + } + + /* Deal with curbe alignment: + */ + reg += ((6+c->key.nr_userclip+3)/4)*2; + } + + /* Vertex program parameters from curbe: + */ + nr_params = c->prog_data.max_const; + for (i = 0; i < nr_params; i++) { + c->regs[TGSI_FILE_CONSTANT][i] = stride(brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1); + } + reg += (nr_params+1)/2; + c->prog_data.curb_read_length = reg - 1; + + + + /* Allocate input regs: + */ + c->nr_inputs = c->vp->program.num_inputs; + for (i = 0; i < c->nr_inputs; i++) { + c->regs[TGSI_FILE_INPUT][i] = brw_vec8_grf(reg, 0); + reg++; + } + + + /* Allocate outputs: TODO: could organize the non-position outputs + * to go straight into message regs. + */ + c->nr_outputs = 0; + c->first_output = reg; + mrf = 4; + for (i = 0; i < c->vp->program.num_outputs; i++) { + c->nr_outputs++; +#if 0 + if (i == VERT_RESULT_HPOS) { + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); + reg++; + } + else if (i == VERT_RESULT_PSIZ) { + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); + reg++; + mrf++; /* just a placeholder? XXX fix later stages & remove this */ + } + else { + c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf); + mrf++; + } +#else + /*treat pos differently for now */ + if (i == info->pos_idx) { + c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); + reg++; + } else { + c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf); + mrf++; + } +#endif + } + + /* Allocate program temporaries: + */ + for (i = 0; i < info->num_temps; i++) { + c->regs[TGSI_FILE_TEMPORARY][i] = brw_vec8_grf(reg, 0); + reg++; + } + + /* Address reg(s). Don't try to use the internal address reg until + * deref time. + */ + for (i = 0; i < info->num_addrs; i++) { + c->regs[TGSI_FILE_ADDRESS][i] = brw_reg(BRW_GENERAL_REGISTER_FILE, + reg, + 0, + BRW_REGISTER_TYPE_D, + BRW_VERTICAL_STRIDE_8, + BRW_WIDTH_8, + BRW_HORIZONTAL_STRIDE_1, + BRW_SWIZZLE_XXXX, + TGSI_WRITEMASK_X); + reg++; + } + + for (i = 0; i < 128; i++) { + if (c->output_regs[i].used_in_src) { + c->output_regs[i].reg = brw_vec8_grf(reg, 0); + reg++; + } + } + + c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg, 0); + reg += 2; + + + /* Some opcodes need an internal temporary: + */ + c->first_tmp = reg; + c->last_tmp = reg; /* for allocation purposes */ + + /* Each input reg holds data from two vertices. The + * urb_read_length is the number of registers read from *each* + * vertex urb, so is half the amount: + */ + c->prog_data.urb_read_length = (c->nr_inputs+1)/2; + + c->prog_data.urb_entry_size = (c->nr_outputs+2+3)/4; + c->prog_data.total_grf = reg; +} + + +static struct brw_reg get_tmp( struct brw_vs_compile *c ) +{ + struct brw_reg tmp = brw_vec8_grf(c->last_tmp, 0); + + if (++c->last_tmp > c->prog_data.total_grf) + c->prog_data.total_grf = c->last_tmp; + + return tmp; +} + +static void release_tmp( struct brw_vs_compile *c, struct brw_reg tmp ) +{ + if (tmp.nr == c->last_tmp-1) + c->last_tmp--; +} + +static void release_tmps( struct brw_vs_compile *c ) +{ + c->last_tmp = c->first_tmp; +} + + +static void unalias1( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + void (*func)( struct brw_vs_compile *, + struct brw_reg, + struct brw_reg )) +{ + if (dst.file == arg0.file && dst.nr == arg0.nr) { + struct brw_compile *p = &c->func; + struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask); + func(c, tmp, arg0); + brw_MOV(p, dst, tmp); + } + else { + func(c, dst, arg0); + } +} + +static void unalias2( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + void (*func)( struct brw_vs_compile *, + struct brw_reg, + struct brw_reg, + struct brw_reg )) +{ + if ((dst.file == arg0.file && dst.nr == arg0.nr) || + (dst.file == arg1.file && dst.nr == arg1.nr)) { + struct brw_compile *p = &c->func; + struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask); + func(c, tmp, arg0, arg1); + brw_MOV(p, dst, tmp); + } + else { + func(c, dst, arg0, arg1); + } +} + +static void emit_sop( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + unsigned cond) +{ + brw_push_insn_state(p); + brw_CMP(p, brw_null_reg(), cond, arg0, arg1); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_MOV(p, dst, brw_imm_f(1.0f)); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, brw_imm_f(0.0f)); + brw_pop_insn_state(p); +} + +static void emit_seq( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_EQ); +} + +static void emit_sne( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_NEQ); +} +static void emit_slt( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_L); +} + +static void emit_sle( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_LE); +} + +static void emit_sgt( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_G); +} + +static void emit_sge( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_GE); +} + +static void emit_max( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1); + brw_SEL(p, dst, arg1, arg0); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +} + +static void emit_min( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1 ) +{ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1); + brw_SEL(p, dst, arg0, arg1); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +} + + +static void emit_math1( struct brw_vs_compile *c, + unsigned function, + struct brw_reg dst, + struct brw_reg arg0, + unsigned precision) +{ + /* There are various odd behaviours with SEND on the simulator. In + * addition there are documented issues with the fact that the GEN4 + * processor doesn't do dependency control properly on SEND + * results. So, on balance, this kludge to get around failures + * with writemasked math results looks like it might be necessary + * whether that turns out to be a simulator bug or not: + */ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + boolean need_tmp = (dst.dw1.bits.writemask != 0xf || + dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_math(p, + tmp, + function, + BRW_MATH_SATURATE_NONE, + 2, + arg0, + BRW_MATH_DATA_SCALAR, + precision); + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + +static void emit_math2( struct brw_vs_compile *c, + unsigned function, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1, + unsigned precision) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + boolean need_tmp = (dst.dw1.bits.writemask != 0xf || + dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_MOV(p, brw_message_reg(3), arg1); + + brw_math(p, + tmp, + function, + BRW_MATH_SATURATE_NONE, + 2, + arg0, + BRW_MATH_DATA_SCALAR, + precision); + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + + + +static void emit_exp_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_X) { + struct brw_reg tmp = get_tmp(c); + struct brw_reg tmp_d = retype(tmp, BRW_REGISTER_TYPE_D); + + /* tmp_d = floor(arg0.x) */ + brw_RNDD(p, tmp_d, brw_swizzle1(arg0, 0)); + + /* result[0] = 2.0 ^ tmp */ + + /* Adjust exponent for floating point: + * exp += 127 + */ + brw_ADD(p, brw_writemask(tmp_d, TGSI_WRITEMASK_X), tmp_d, brw_imm_d(127)); + + /* Install exponent and sign. + * Excess drops off the edge: + */ + brw_SHL(p, brw_writemask(retype(dst, BRW_REGISTER_TYPE_D), TGSI_WRITEMASK_X), + tmp_d, brw_imm_d(23)); + + release_tmp(c, tmp); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Y) { + /* result[1] = arg0.x - floor(arg0.x) */ + brw_FRC(p, brw_writemask(dst, TGSI_WRITEMASK_Y), brw_swizzle1(arg0, 0)); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) { + /* As with the LOG instruction, we might be better off just + * doing a taylor expansion here, seeing as we have to do all + * the prep work. + * + * If mathbox partial precision is too low, consider also: + * result[3] = result[0] * EXP(result[1]) + */ + emit_math1(c, + BRW_MATH_FUNCTION_EXP, + brw_writemask(dst, TGSI_WRITEMASK_Z), + brw_swizzle1(arg0, 0), + BRW_MATH_PRECISION_PARTIAL); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) { + /* result[3] = 1.0; */ + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_W), brw_imm_f(1)); + } +} + + +static void emit_log_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + struct brw_reg tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD); + struct brw_reg arg0_ud = retype(arg0, BRW_REGISTER_TYPE_UD); + boolean need_tmp = (dst.dw1.bits.writemask != 0xf || + dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) { + tmp = get_tmp(c); + tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD); + } + + /* Perform mant = frexpf(fabsf(x), &exp), adjust exp and mnt + * according to spec: + * + * These almost look likey they could be joined up, but not really + * practical: + * + * result[0].f = (x.i & ((1<<31)-1) >> 23) - 127 + * result[1].i = (x.i & ((1<<23)-1) + (127<<23) + */ + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_XZ) { + brw_AND(p, + brw_writemask(tmp_ud, TGSI_WRITEMASK_X), + brw_swizzle1(arg0_ud, 0), + brw_imm_ud((1U<<31)-1)); + + brw_SHR(p, + brw_writemask(tmp_ud, TGSI_WRITEMASK_X), + tmp_ud, + brw_imm_ud(23)); + + brw_ADD(p, + brw_writemask(tmp, TGSI_WRITEMASK_X), + retype(tmp_ud, BRW_REGISTER_TYPE_D), /* does it matter? */ + brw_imm_d(-127)); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_YZ) { + brw_AND(p, + brw_writemask(tmp_ud, TGSI_WRITEMASK_Y), + brw_swizzle1(arg0_ud, 0), + brw_imm_ud((1<<23)-1)); + + brw_OR(p, + brw_writemask(tmp_ud, TGSI_WRITEMASK_Y), + tmp_ud, + brw_imm_ud(127<<23)); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) { + /* result[2] = result[0] + LOG2(result[1]); */ + + /* Why bother? The above is just a hint how to do this with a + * taylor series. Maybe we *should* use a taylor series as by + * the time all the above has been done it's almost certainly + * quicker than calling the mathbox, even with low precision. + * + * Options are: + * - result[0] + mathbox.LOG2(result[1]) + * - mathbox.LOG2(arg0.x) + * - result[0] + inline_taylor_approx(result[1]) + */ + emit_math1(c, + BRW_MATH_FUNCTION_LOG, + brw_writemask(tmp, TGSI_WRITEMASK_Z), + brw_swizzle1(tmp, 1), + BRW_MATH_PRECISION_FULL); + + brw_ADD(p, + brw_writemask(tmp, TGSI_WRITEMASK_Z), + brw_swizzle1(tmp, 2), + brw_swizzle1(tmp, 0)); + } + + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) { + /* result[3] = 1.0; */ + brw_MOV(p, brw_writemask(tmp, TGSI_WRITEMASK_W), brw_imm_f(1)); + } + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + + + + +/* Need to unalias - consider swizzles: r0 = DST r0.xxxx r1 + */ +static void emit_dst_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0, + struct brw_reg arg1) +{ + struct brw_compile *p = &c->func; + + /* There must be a better way to do this: + */ + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_X) + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_X), brw_imm_f(1.0)); + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Y) + brw_MUL(p, brw_writemask(dst, TGSI_WRITEMASK_Y), arg0, arg1); + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_Z), arg0); + if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_W), arg1); +} + +static void emit_xpd( struct brw_compile *p, + struct brw_reg dst, + struct brw_reg t, + struct brw_reg u) +{ + brw_MUL(p, brw_null_reg(), brw_swizzle(t, 1,2,0,3), brw_swizzle(u,2,0,1,3)); + brw_MAC(p, dst, negate(brw_swizzle(t, 2,0,1,3)), brw_swizzle(u,1,2,0,3)); +} + + + +static void emit_lit_noalias( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + struct brw_instruction *if_insn; + struct brw_reg tmp = dst; + boolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_YZ), brw_imm_f(0)); + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_XW), brw_imm_f(1)); + + /* Need to use BRW_EXECUTE_8 and also do an 8-wide compare in order + * to get all channels active inside the IF. In the clipping code + * we run with NoMask, so it's not an option and we can use + * BRW_EXECUTE_1 for all comparisions. + */ + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,0), brw_imm_f(0)); + if_insn = brw_IF(p, BRW_EXECUTE_8); + { + brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_Y), brw_swizzle1(arg0,0)); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,1), brw_imm_f(0)); + brw_MOV(p, brw_writemask(tmp, TGSI_WRITEMASK_Z), brw_swizzle1(arg0,1)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + + emit_math2(c, + BRW_MATH_FUNCTION_POW, + brw_writemask(dst, TGSI_WRITEMASK_Z), + brw_swizzle1(tmp, 2), + brw_swizzle1(arg0, 3), + BRW_MATH_PRECISION_PARTIAL); + } + + brw_ENDIF(p, if_insn); +} + + + + + +/* TODO: relative addressing! + */ +static struct brw_reg get_reg( struct brw_vs_compile *c, + unsigned file, + unsigned index ) +{ + switch (file) { + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_INPUT: + case TGSI_FILE_OUTPUT: + assert(c->regs[file][index].nr != 0); + return c->regs[file][index]; + case TGSI_FILE_CONSTANT: + assert(c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_imm].nr != 0); + return c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_imm]; + case TGSI_FILE_IMMEDIATE: + assert(c->regs[TGSI_FILE_CONSTANT][index].nr != 0); + return c->regs[TGSI_FILE_CONSTANT][index]; + case TGSI_FILE_ADDRESS: + assert(index == 0); + return c->regs[file][index]; + + case TGSI_FILE_NULL: /* undef values */ + return brw_null_reg(); + + default: + assert(0); + return brw_null_reg(); + } +} + + + +static struct brw_reg deref( struct brw_vs_compile *c, + struct brw_reg arg, + int offset) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = vec4(get_tmp(c)); + struct brw_reg vp_address = retype(vec1(get_reg(c, TGSI_FILE_ADDRESS, 0)), BRW_REGISTER_TYPE_UW); + unsigned byte_offset = arg.nr * 32 + arg.subnr + offset * 16; + struct brw_reg indirect = brw_vec4_indirect(0,0); + + { + brw_push_insn_state(p); + brw_set_access_mode(p, BRW_ALIGN_1); + + /* This is pretty clunky - load the address register twice and + * fetch each 4-dword value in turn. There must be a way to do + * this in a single pass, but I couldn't get it to work. + */ + brw_ADD(p, brw_address_reg(0), vp_address, brw_imm_d(byte_offset)); + brw_MOV(p, tmp, indirect); + + brw_ADD(p, brw_address_reg(0), suboffset(vp_address, 8), brw_imm_d(byte_offset)); + brw_MOV(p, suboffset(tmp, 4), indirect); + + brw_pop_insn_state(p); + } + + return vec8(tmp); +} + + +static void emit_arl( struct brw_vs_compile *c, + struct brw_reg dst, + struct brw_reg arg0 ) +{ + struct brw_compile *p = &c->func; + struct brw_reg tmp = dst; + boolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE); + + if (need_tmp) + tmp = get_tmp(c); + + brw_RNDD(p, tmp, arg0); + brw_MUL(p, dst, tmp, brw_imm_d(16)); + + if (need_tmp) + release_tmp(c, tmp); +} + + +/* Will return mangled results for SWZ op. The emit_swz() function + * ignores this result and recalculates taking extended swizzles into + * account. + */ +static struct brw_reg get_arg( struct brw_vs_compile *c, + struct tgsi_src_register *src ) +{ + struct brw_reg reg; + + if (src->File == TGSI_FILE_NULL) + return brw_null_reg(); + +#if 0 + if (src->RelAddr) + reg = deref(c, c->regs[PROGRAM_STATE_VAR][0], src->Index); + else +#endif + reg = get_reg(c, src->File, src->Index); + + /* Convert 3-bit swizzle to 2-bit. + */ + reg.dw1.bits.swizzle = BRW_SWIZZLE4(src->SwizzleX, + src->SwizzleY, + src->SwizzleZ, + src->SwizzleW); + + /* Note this is ok for non-swizzle instructions: + */ + reg.negate = src->Negate ? 1 : 0; + + return reg; +} + + +static struct brw_reg get_dst( struct brw_vs_compile *c, + const struct tgsi_dst_register *dst ) +{ + struct brw_reg reg = get_reg(c, dst->File, dst->Index); + + reg.dw1.bits.writemask = dst->WriteMask; + + return reg; +} + + + + +static void emit_swz( struct brw_vs_compile *c, + struct brw_reg dst, + struct tgsi_src_register src ) +{ + struct brw_compile *p = &c->func; + unsigned zeros_mask = 0; + unsigned ones_mask = 0; + unsigned src_mask = 0; + ubyte src_swz[4]; + boolean need_tmp = (src.Negate && + dst.file != BRW_GENERAL_REGISTER_FILE); + struct brw_reg tmp = dst; + unsigned i; + + if (need_tmp) + tmp = get_tmp(c); + + for (i = 0; i < 4; i++) { + if (dst.dw1.bits.writemask & (1<regs[PROGRAM_STATE_VAR][0], src.Index); + else +#endif + arg0 = get_reg(c, src.File, src.Index); + + arg0 = brw_swizzle(arg0, + src_swz[0], src_swz[1], + src_swz[2], src_swz[3]); + + brw_MOV(p, brw_writemask(tmp, src_mask), arg0); + } + + if (zeros_mask) + brw_MOV(p, brw_writemask(tmp, zeros_mask), brw_imm_f(0)); + + if (ones_mask) + brw_MOV(p, brw_writemask(tmp, ones_mask), brw_imm_f(1)); + + if (src.Negate) + brw_MOV(p, brw_writemask(tmp, src.Negate), negate(tmp)); + + if (need_tmp) { + brw_MOV(p, dst, tmp); + release_tmp(c, tmp); + } +} + + + +/* Post-vertex-program processing. Send the results to the URB. + */ +static void emit_vertex_write( struct brw_vs_compile *c, struct brw_prog_info *info) +{ + struct brw_compile *p = &c->func; + struct brw_reg m0 = brw_message_reg(0); + struct brw_reg pos = c->regs[TGSI_FILE_OUTPUT][info->pos_idx]; + struct brw_reg ndc; + + if (c->key.copy_edgeflag) { + brw_MOV(p, + get_reg(c, TGSI_FILE_OUTPUT, info->result_edge_idx), + get_reg(c, TGSI_FILE_INPUT, info->edge_flag_idx)); + } + + + /* Build ndc coords? TODO: Shortcircuit when w is known to be one. + */ + if (!c->key.know_w_is_one) { + ndc = get_tmp(c); + emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL); + brw_MUL(p, brw_writemask(ndc, TGSI_WRITEMASK_XYZ), pos, ndc); + } + else { + ndc = pos; + } + + /* This includes the workaround for -ve rhw, so is no longer an + * optional step: + */ + if (info->writes_psize || + c->key.nr_userclip || + !c->key.know_w_is_one) + { + struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); + unsigned i; + + brw_MOV(p, header1, brw_imm_ud(0)); + + brw_set_access_mode(p, BRW_ALIGN_16); + + if (info->writes_psize) { + struct brw_reg psiz = c->regs[TGSI_FILE_OUTPUT][info->psize_idx]; + brw_MUL(p, brw_writemask(header1, TGSI_WRITEMASK_W), + brw_swizzle1(psiz, 0), brw_imm_f(1<<11)); + brw_AND(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, + brw_imm_ud(0x7ff<<8)); + } + + + for (i = 0; i < c->key.nr_userclip; i++) { + brw_set_conditionalmod(p, BRW_CONDITIONAL_L); + brw_DP4(p, brw_null_reg(), pos, c->userplane[i]); + brw_OR(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, brw_imm_ud(1<key.know_w_is_one) { + brw_CMP(p, + vec8(brw_null_reg()), + BRW_CONDITIONAL_L, + brw_swizzle1(ndc, 3), + brw_imm_f(0)); + + brw_OR(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, brw_imm_ud(1<<6)); + brw_MOV(p, ndc, brw_imm_f(0)); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + } + + brw_set_access_mode(p, BRW_ALIGN_1); /* why? */ + brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), header1); + brw_set_access_mode(p, BRW_ALIGN_16); + + release_tmp(c, header1); + } + else { + brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), brw_imm_ud(0)); + } + + + /* Emit the (interleaved) headers for the two vertices - an 8-reg + * of zeros followed by two sets of NDC coordinates: + */ + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, offset(m0, 2), ndc); + brw_MOV(p, offset(m0, 3), pos); + + + brw_urb_WRITE(p, + brw_null_reg(), /* dest */ + 0, /* starting mrf reg nr */ + c->r0, /* src */ + 0, /* allocate */ + 1, /* used */ + c->nr_outputs + 3, /* msg len */ + 0, /* response len */ + 1, /* eot */ + 1, /* writes complete */ + 0, /* urb destination offset */ + BRW_URB_SWIZZLE_INTERLEAVE); + +} + +static void +post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst ) +{ + struct tgsi_parse_context parse; + const struct tgsi_token *tokens = c->vp->program.tokens; + tgsi_parse_init(&parse, tokens); + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + if (parse.FullToken.Token.Type == TGSI_TOKEN_TYPE_INSTRUCTION) { +#if 0 + struct brw_instruction *brw_inst1, *brw_inst2; + const struct tgsi_full_instruction *inst1, *inst2; + int offset; + inst1 = &parse.FullToken.FullInstruction; + brw_inst1 = inst1->Data; + switch (inst1->Opcode) { + case TGSI_OPCODE_CAL: + case TGSI_OPCODE_BRA: + target_insn = inst1->BranchTarget; + inst2 = &c->vp->program.Base.Instructions[target_insn]; + brw_inst2 = inst2->Data; + offset = brw_inst2 - brw_inst1; + brw_set_src1(brw_inst1, brw_imm_d(offset*16)); + break; + case TGSI_OPCODE_END: + offset = end_inst - brw_inst1; + brw_set_src1(brw_inst1, brw_imm_d(offset*16)); + break; + default: + break; + } +#endif + } + } + tgsi_parse_free(&parse); +} + +static void process_declaration(const struct tgsi_full_declaration *decl, + struct brw_prog_info *info) +{ + int first = decl->u.DeclarationRange.First; + int last = decl->u.DeclarationRange.Last; + + assert (decl->Declaration.Declare != TGSI_DECLARE_MASK); + + switch(decl->Declaration.File) { + case TGSI_FILE_CONSTANT: + info->num_consts += last - first + 1; + break; + case TGSI_FILE_INPUT: { + } + break; + case TGSI_FILE_OUTPUT: { + assert(last == first); /* for now */ + if (decl->Declaration.Semantic) { + switch (decl->Semantic.SemanticName) { + case TGSI_SEMANTIC_POSITION: { + info->pos_idx = first; + } + break; + case TGSI_SEMANTIC_COLOR: + break; + case TGSI_SEMANTIC_BCOLOR: + break; + case TGSI_SEMANTIC_FOG: + break; + case TGSI_SEMANTIC_PSIZE: { + info->writes_psize = TRUE; + info->psize_idx = first; + } + break; + case TGSI_SEMANTIC_GENERIC: + break; + } + } + } + break; + case TGSI_FILE_TEMPORARY: { + info->num_temps += (last - first) + 1; + } + break; + case TGSI_FILE_SAMPLER: { + } + break; + case TGSI_FILE_ADDRESS: { + info->num_addrs += (last - first) + 1; + } + break; + case TGSI_FILE_IMMEDIATE: { + } + break; + case TGSI_FILE_NULL: { + } + break; + } +} + +static void process_instruction(struct brw_vs_compile *c, + struct tgsi_full_instruction *inst, + struct brw_prog_info *info) +{ + struct brw_reg args[3], dst; + struct brw_compile *p = &c->func; + struct brw_indirect stack_index = brw_indirect(0, 0); + unsigned i; + unsigned index; + unsigned file; + /*FIXME: might not be the only one*/ + const struct tgsi_dst_register *dst_reg = &inst->FullDstRegisters[0].DstRegister; + /* + struct brw_instruction *if_inst[MAX_IFSN]; + unsigned insn, if_insn = 0; + */ + + for (i = 0; i < 3; i++) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + index = src->SrcRegister.Index; + file = src->SrcRegister.File; + if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src) + args[i] = c->output_regs[index].reg; + else + args[i] = get_arg(c, &src->SrcRegister); + } + + /* Get dest regs. Note that it is possible for a reg to be both + * dst and arg, given the static allocation of registers. So + * care needs to be taken emitting multi-operation instructions. + */ + index = dst_reg->Index; + file = dst_reg->File; + if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src) + dst = c->output_regs[index].reg; + else + dst = get_dst(c, dst_reg); + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + brw_MOV(p, dst, brw_abs(args[0])); + break; + case TGSI_OPCODE_ADD: + brw_ADD(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DP3: + brw_DP3(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DP4: + brw_DP4(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DPH: + brw_DPH(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_DST: + unalias2(c, dst, args[0], args[1], emit_dst_noalias); + break; + case TGSI_OPCODE_EXP: + unalias1(c, dst, args[0], emit_exp_noalias); + break; + case TGSI_OPCODE_EX2: + emit_math1(c, BRW_MATH_FUNCTION_EXP, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_ARL: + emit_arl(c, dst, args[0]); + break; + case TGSI_OPCODE_FLR: + brw_RNDD(p, dst, args[0]); + break; + case TGSI_OPCODE_FRC: + brw_FRC(p, dst, args[0]); + break; + case TGSI_OPCODE_LOG: + unalias1(c, dst, args[0], emit_log_noalias); + break; + case TGSI_OPCODE_LG2: + emit_math1(c, BRW_MATH_FUNCTION_LOG, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_LIT: + unalias1(c, dst, args[0], emit_lit_noalias); + break; + case TGSI_OPCODE_MAD: + brw_MOV(p, brw_acc_reg(), args[2]); + brw_MAC(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MAX: + emit_max(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MIN: + emit_min(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_MOV: +#if 0 + case TGSI_OPCODE_SWZ: + /* The args[0] value can't be used here as it won't have + * correctly encoded the full swizzle: + */ + emit_swz(c, dst, inst->SrcReg[0] ); +#endif + brw_MOV(p, dst, args[0]); + break; + case TGSI_OPCODE_MUL: + brw_MUL(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_POW: + emit_math2(c, BRW_MATH_FUNCTION_POW, dst, args[0], args[1], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_RCP: + emit_math1(c, BRW_MATH_FUNCTION_INV, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + case TGSI_OPCODE_RSQ: + emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, args[0], BRW_MATH_PRECISION_FULL); + break; + + case TGSI_OPCODE_SEQ: + emit_seq(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SNE: + emit_sne(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SGE: + emit_sge(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SGT: + emit_sgt(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SLT: + emit_slt(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SLE: + emit_sle(p, dst, args[0], args[1]); + break; + case TGSI_OPCODE_SUB: + brw_ADD(p, dst, args[0], negate(args[1])); + break; + case TGSI_OPCODE_XPD: + emit_xpd(p, dst, args[0], args[1]); + break; +#if 0 + case TGSI_OPCODE_IF: + assert(if_insn < MAX_IFSN); + if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8); + break; + case TGSI_OPCODE_ELSE: + if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]); + break; + case TGSI_OPCODE_ENDIF: + assert(if_insn > 0); + brw_ENDIF(p, if_inst[--if_insn]); + break; + case TGSI_OPCODE_BRA: + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + brw_set_predicate_control_flag_value(p, 0xff); + break; + case TGSI_OPCODE_CAL: + brw_set_access_mode(p, BRW_ALIGN_1); + brw_ADD(p, deref_1uw(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(4)); + inst->Data = &p->store[p->nr_insn]; + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + break; +#endif + case TGSI_OPCODE_RET: +#if 0 + brw_ADD(p, get_addr_reg(stack_index), + get_addr_reg(stack_index), brw_imm_d(-4)); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, brw_ip_reg(), deref_1uw(stack_index, 0)); + brw_set_access_mode(p, BRW_ALIGN_16); +#else + /*brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));*/ +#endif + break; + case TGSI_OPCODE_END: + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + break; + case TGSI_OPCODE_BGNSUB: + case TGSI_OPCODE_ENDSUB: + break; + default: + debug_printf("Unsupport opcode %d in vertex shader\n", inst->Instruction.Opcode); + break; + } + + if (dst_reg->File == TGSI_FILE_OUTPUT + && dst_reg->Index != info->pos_idx + && c->output_regs[dst_reg->Index].used_in_src) + brw_MOV(p, get_dst(c, dst_reg), dst); + + release_tmps(c); +} + +/* Emit the fragment program instructions here. + */ +void brw_vs_emit(struct brw_vs_compile *c) +{ +#define MAX_IFSN 32 + struct brw_compile *p = &c->func; + struct brw_instruction *end_inst; + struct tgsi_parse_context parse; + struct brw_indirect stack_index = brw_indirect(0, 0); + const struct tgsi_token *tokens = c->vp->program.tokens; + struct brw_prog_info prog_info; + unsigned allocated_registers = 0; + memset(&prog_info, 0, sizeof(struct brw_prog_info)); + + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + brw_set_access_mode(p, BRW_ALIGN_16); + + tgsi_parse_init(&parse, tokens); + /* Message registers can't be read, so copy the output into GRF register + if they are used in source registers */ + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + unsigned i; + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_INSTRUCTION: { + const struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction; + for (i = 0; i < 3; ++i) { + const struct tgsi_src_register *src = &inst->FullSrcRegisters[i].SrcRegister; + unsigned index = src->Index; + unsigned file = src->File; + if (file == TGSI_FILE_OUTPUT) + c->output_regs[index].used_in_src = TRUE; + } + } + break; + default: + /* nothing */ + break; + } + } + tgsi_parse_free(&parse); + + tgsi_parse_init(&parse, tokens); + + while (!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: { + struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; + process_declaration(decl, &prog_info); + } + break; + case TGSI_TOKEN_TYPE_IMMEDIATE: { + struct tgsi_full_immediate *imm = &parse.FullToken.FullImmediate; + /*assert(imm->Immediate.Size == 4);*/ + c->prog_data.imm_buf[c->prog_data.num_imm][0] = imm->u.ImmediateFloat32[0].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][1] = imm->u.ImmediateFloat32[1].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][2] = imm->u.ImmediateFloat32[2].Float; + c->prog_data.imm_buf[c->prog_data.num_imm][3] = imm->u.ImmediateFloat32[3].Float; + c->prog_data.num_imm++; + } + break; + case TGSI_TOKEN_TYPE_INSTRUCTION: { + struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction; + if (!allocated_registers) { + /* first instruction (declerations finished). + * now that we know what vars are being used allocate + * registers for them.*/ + c->prog_data.num_consts = prog_info.num_consts; + c->prog_data.max_const = prog_info.num_consts + c->prog_data.num_imm; + brw_vs_alloc_regs(c, &prog_info); + + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); + brw_set_access_mode(p, BRW_ALIGN_16); + allocated_registers = 1; + } + process_instruction(c, inst, &prog_info); + } + break; + } + } + + end_inst = &p->store[p->nr_insn]; + emit_vertex_write(c, &prog_info); + post_vs_emit(c, end_inst); + tgsi_parse_free(&parse); + +} diff --git a/src/gallium/drivers/i965simple/brw_vs_state.c b/src/gallium/drivers/i965simple/brw_vs_state.c new file mode 100644 index 0000000000..c73469929c --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_vs_state.c @@ -0,0 +1,102 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + +#include "pipe/p_util.h" + +static void upload_vs_unit( struct brw_context *brw ) +{ + struct brw_vs_unit_state vs; + + memset(&vs, 0, sizeof(vs)); + + /* CACHE_NEW_VS_PROG */ + vs.thread0.kernel_start_pointer = brw->vs.prog_gs_offset >> 6; + vs.thread0.grf_reg_count = align(brw->vs.prog_data->total_grf, 16) / 16 - 1; + vs.thread3.urb_entry_read_length = brw->vs.prog_data->urb_read_length; + vs.thread3.const_urb_entry_read_length = brw->vs.prog_data->curb_read_length; + vs.thread3.dispatch_grf_start_reg = 1; + + + /* BRW_NEW_URB_FENCE */ + vs.thread4.nr_urb_entries = brw->urb.nr_vs_entries; + vs.thread4.urb_entry_allocation_size = brw->urb.vsize - 1; + vs.thread4.max_threads = MIN2( + MAX2(0, (brw->urb.nr_vs_entries - 6) / 2 - 1), + 15); + + + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + vs.thread4.max_threads = 0; + + /* BRW_NEW_CURBE_OFFSETS, _NEW_TRANSFORM */ + if (0 /*brw->attribs.Clip->ClipPlanesEnabled*/) { + /* Note that we read in the userclip planes as well, hence + * clip_start: + */ + vs.thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2; + } + else { + vs.thread3.const_urb_entry_read_offset = brw->curbe.vs_start * 2; + } + + vs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + vs.thread3.urb_entry_read_offset = 0; + + /* No samplers for ARB_vp programs: + */ + vs.vs5.sampler_count = 0; + + if (BRW_DEBUG & DEBUG_STATS) + vs.thread4.stats_enable = 1; + + /* Vertex program always enabled: + */ + vs.vs6.vs_enable = 1; + + brw->vs.state_gs_offset = brw_cache_data( &brw->cache[BRW_VS_UNIT], &vs ); +} + + +const struct brw_tracked_state brw_vs_unit = { + .dirty = { + .brw = (BRW_NEW_CLIP | + BRW_NEW_CURBE_OFFSETS | + BRW_NEW_URB_FENCE), + .cache = CACHE_NEW_VS_PROG + }, + .update = upload_vs_unit +}; diff --git a/src/gallium/drivers/i965simple/brw_winsys.h b/src/gallium/drivers/i965simple/brw_winsys.h new file mode 100644 index 0000000000..3523a58614 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_winsys.h @@ -0,0 +1,205 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * This is the interface that i965simple requires any window system + * hosting it to implement. This is the only include file in i965simple + * which is public. + * + */ + +#ifndef BRW_WINSYS_H +#define BRW_WINSYS_H + + +#include "pipe/p_defines.h" + + +/* Pipe drivers are (meant to be!) independent of both GL and the + * window system. The window system provides a buffer manager and a + * set of additional hooks for things like command buffer submission, + * etc. + * + * There clearly has to be some agreement between the window system + * driver and the hardware driver about the format of command buffers, + * etc. + */ + +struct pipe_buffer; +struct pipe_fence_handle; +struct pipe_winsys; + +/* The pipe driver currently understands the following chipsets: + */ +#define PCI_CHIP_I965_G 0x29A2 +#define PCI_CHIP_I965_Q 0x2992 +#define PCI_CHIP_I965_G_1 0x2982 +#define PCI_CHIP_I965_GM 0x2A02 +#define PCI_CHIP_I965_GME 0x2A12 + + +/* These are the names of all the state caches managed by the driver. + * + * When data is uploaded to a buffer with buffer_subdata, we use the + * special version of that function below so that information about + * what type of data this is can be passed to the winsys backend. + * That in turn allows the correct flags to be set in the aub file + * dump to allow human-readable file dumps later on. + */ + +enum brw_cache_id { + BRW_CC_VP, + BRW_CC_UNIT, + BRW_WM_PROG, + BRW_SAMPLER_DEFAULT_COLOR, + BRW_SAMPLER, + BRW_WM_UNIT, + BRW_SF_PROG, + BRW_SF_VP, + BRW_SF_UNIT, + BRW_VS_UNIT, + BRW_VS_PROG, + BRW_GS_UNIT, + BRW_GS_PROG, + BRW_CLIP_VP, + BRW_CLIP_UNIT, + BRW_CLIP_PROG, + BRW_SS_SURFACE, + BRW_SS_SURF_BIND, + + BRW_MAX_CACHE +}; + +#define BRW_CONSTANT_BUFFER BRW_MAX_CACHE + +/** + * Additional winsys interface for i965simple. + * + * It is an over-simple batchbuffer mechanism. Will want to improve the + * performance of this, perhaps based on the cmdstream stuff. It + * would be pretty impossible to implement swz on top of this + * interface. + * + * Will also need additions/changes to implement static/dynamic + * indirect state. + */ +struct brw_winsys { + + /** + * Reserve space on batch buffer. + * + * Returns a null pointer if there is insufficient space in the batch buffer + * to hold the requested number of dwords and relocations. + * + * The number of dwords should also include the number of relocations. + */ + unsigned *(*batch_start)(struct brw_winsys *sws, + unsigned dwords, + unsigned relocs); + + void (*batch_dword)(struct brw_winsys *sws, + unsigned dword); + + /** + * Emit a relocation to a buffer. + * + * Used not only when the buffer addresses are not pinned, but also to + * ensure refered buffers will not be destroyed until the current batch + * buffer execution is finished. + * + * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and + * I915_BUFFER_ACCESS_READ macros. + */ + void (*batch_reloc)(struct brw_winsys *sws, + struct pipe_buffer *buf, + unsigned access_flags, + unsigned delta); + + + /* Not used yet, but really want this: + */ + void (*batch_end)( struct brw_winsys *sws ); + + /** + * Flush the batch buffer. + * + * Fence argument must point to NULL or to a previous fence, and the caller + * must call fence_reference when done with the fence. + */ + void (*batch_flush)(struct brw_winsys *sws, + struct pipe_fence_handle **fence); + + + /* A version of buffer_subdata that includes information for the + * simulator: + */ + void (*buffer_subdata_typed)(struct brw_winsys *sws, + struct pipe_buffer *buf, + unsigned long offset, + unsigned long size, + const void *data, + unsigned data_type); + + + /* A cheat so we don't have to think about relocations in a couple + * of places yet: + */ + unsigned (*get_buffer_offset)( struct brw_winsys *sws, + struct pipe_buffer *buf, + unsigned flags ); + +}; + +#define BRW_BUFFER_ACCESS_WRITE 0x1 +#define BRW_BUFFER_ACCESS_READ 0x2 + +#define BRW_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) + + +struct pipe_context *brw_create(struct pipe_winsys *, + struct brw_winsys *, + unsigned pci_id); + +static inline boolean brw_batchbuffer_data(struct brw_winsys *winsys, + const void *data, + unsigned bytes) +{ + static const unsigned incr = sizeof(unsigned); + uint i; + const unsigned *udata = (const unsigned*)(data); + unsigned size = bytes/incr; + + winsys->batch_start(winsys, size, 0); + for (i = 0; i < size; ++i) { + winsys->batch_dword(winsys, udata[i]); + } + winsys->batch_end(winsys); + + return (i == size); +} +#endif diff --git a/src/gallium/drivers/i965simple/brw_wm.c b/src/gallium/drivers/i965simple/brw_wm.c new file mode 100644 index 0000000000..539b170744 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_wm.c @@ -0,0 +1,210 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_util.h" +#include "brw_wm.h" +#include "brw_eu.h" +#include "brw_state.h" +#include "pipe/p_util.h" + + + +static void do_wm_prog( struct brw_context *brw, + struct brw_fragment_program *fp, + struct brw_wm_prog_key *key) +{ + struct brw_wm_compile *c = CALLOC_STRUCT(brw_wm_compile); + const unsigned *program; + unsigned program_size; + + c->key = *key; + c->fp = fp; + + c->delta_xy[0] = brw_null_reg(); + c->delta_xy[1] = brw_null_reg(); + c->pixel_xy[0] = brw_null_reg(); + c->pixel_xy[1] = brw_null_reg(); + c->pixel_w = brw_null_reg(); + + + debug_printf("XXXXXXXX FP\n"); + + brw_wm_glsl_emit(c); + + /* get the program + */ + program = brw_get_program(&c->func, &program_size); + + /* + */ + brw->wm.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_WM_PROG], + &c->key, + sizeof(c->key), + program, + program_size, + &c->prog_data, + &brw->wm.prog_data ); + + FREE(c); +} + + + +static void brw_wm_populate_key( struct brw_context *brw, + struct brw_wm_prog_key *key ) +{ + /* BRW_NEW_FRAGMENT_PROGRAM */ + struct brw_fragment_program *fp = + (struct brw_fragment_program *)brw->attribs.FragmentProgram; + unsigned lookup = 0; + unsigned line_aa; + + memset(key, 0, sizeof(*key)); + + /* Build the index for table lookup + */ + /* BRW_NEW_DEPTH_STENCIL */ + if (fp->UsesKill || + brw->attribs.DepthStencil->alpha.enabled) + lookup |= IZ_PS_KILL_ALPHATEST_BIT; + + if (fp->ComputesDepth) + lookup |= IZ_PS_COMPUTES_DEPTH_BIT; + + if (brw->attribs.DepthStencil->depth.enabled) + lookup |= IZ_DEPTH_TEST_ENABLE_BIT; + + if (brw->attribs.DepthStencil->depth.enabled && + brw->attribs.DepthStencil->depth.writemask) /* ?? */ + lookup |= IZ_DEPTH_WRITE_ENABLE_BIT; + + if (brw->attribs.DepthStencil->stencil[0].enabled) { + lookup |= IZ_STENCIL_TEST_ENABLE_BIT; + + if (brw->attribs.DepthStencil->stencil[0].write_mask || + brw->attribs.DepthStencil->stencil[1].write_mask) + lookup |= IZ_STENCIL_WRITE_ENABLE_BIT; + } + + /* XXX: when should this be disabled? + */ + if (1) + lookup |= IZ_EARLY_DEPTH_TEST_BIT; + + + line_aa = AA_NEVER; + + /* _NEW_LINE, _NEW_POLYGON, BRW_NEW_REDUCED_PRIMITIVE */ + if (brw->attribs.Raster->line_smooth) { + if (brw->reduced_primitive == PIPE_PRIM_LINES) { + line_aa = AA_ALWAYS; + } + else if (brw->reduced_primitive == PIPE_PRIM_TRIANGLES) { + if (brw->attribs.Raster->fill_ccw == PIPE_POLYGON_MODE_LINE) { + line_aa = AA_SOMETIMES; + + if (brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_LINE || + (brw->attribs.Raster->cull_mode == PIPE_WINDING_CW)) + line_aa = AA_ALWAYS; + } + else if (brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_LINE) { + line_aa = AA_SOMETIMES; + + if (brw->attribs.Raster->cull_mode == PIPE_WINDING_CCW) + line_aa = AA_ALWAYS; + } + } + } + + brw_wm_lookup_iz(line_aa, + lookup, + key); + + +#if 0 + /* BRW_NEW_SAMPLER + * + * Not doing any of this at the moment: + */ + for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { + const struct pipe_sampler_state *unit = brw->attribs.Samplers[i]; + + if (unit) { + + if (unit->compare && + unit->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + key->shadowtex_mask |= 1<Image[0][t->BaseLevel]->InternalFormat == GL_YCBCR_MESA) + key->yuvtex_mask |= 1<program_string_id = fp->id; + +} + + +static void brw_upload_wm_prog( struct brw_context *brw ) +{ + struct brw_wm_prog_key key; + struct brw_fragment_program *fp = (struct brw_fragment_program *) + brw->attribs.FragmentProgram; + + brw_wm_populate_key(brw, &key); + + /* Make an early check for the key. + */ + if (brw_search_cache(&brw->cache[BRW_WM_PROG], + &key, sizeof(key), + &brw->wm.prog_data, + &brw->wm.prog_gs_offset)) + return; + + do_wm_prog(brw, fp, &key); +} + + +const struct brw_tracked_state brw_wm_prog = { + .dirty = { + .brw = (BRW_NEW_FS | + BRW_NEW_REDUCED_PRIMITIVE), + .cache = 0 + }, + .update = brw_upload_wm_prog +}; + diff --git a/src/gallium/drivers/i965simple/brw_wm.h b/src/gallium/drivers/i965simple/brw_wm.h new file mode 100644 index 0000000000..a1ac0f504a --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_wm.h @@ -0,0 +1,142 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRW_WM_H +#define BRW_WM_H + + +#include "brw_context.h" +#include "brw_eu.h" + +/* A big lookup table is used to figure out which and how many + * additional regs will inserted before the main payload in the WM + * program execution. These mainly relate to depth and stencil + * processing and the early-depth-test optimization. + */ +#define IZ_PS_KILL_ALPHATEST_BIT 0x1 +#define IZ_PS_COMPUTES_DEPTH_BIT 0x2 +#define IZ_DEPTH_WRITE_ENABLE_BIT 0x4 +#define IZ_DEPTH_TEST_ENABLE_BIT 0x8 +#define IZ_STENCIL_WRITE_ENABLE_BIT 0x10 +#define IZ_STENCIL_TEST_ENABLE_BIT 0x20 +#define IZ_EARLY_DEPTH_TEST_BIT 0x40 +#define IZ_BIT_MAX 0x80 + +#define AA_NEVER 0 +#define AA_SOMETIMES 1 +#define AA_ALWAYS 2 + +struct brw_wm_prog_key { + unsigned source_depth_reg:3; + unsigned aa_dest_stencil_reg:3; + unsigned dest_depth_reg:3; + unsigned nr_depth_regs:3; + unsigned shadowtex_mask:8; + unsigned computes_depth:1; /* could be derived from program string */ + unsigned source_depth_to_render_target:1; + unsigned runtime_check_aads_emit:1; + + unsigned yuvtex_mask:8; + + unsigned program_string_id; +}; + + + + + +#define PROGRAM_INTERNAL_PARAM +#define MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS 1024 /* 72 for GL_ARB_f_p */ +#define BRW_WM_MAX_INSN (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + PIPE_ATTRIB_MAX + 3) +#define BRW_WM_MAX_GRF 128 /* hardware limit */ +#define BRW_WM_MAX_VREG (BRW_WM_MAX_INSN * 4) +#define BRW_WM_MAX_REF (BRW_WM_MAX_INSN * 12) +#define BRW_WM_MAX_PARAM 256 +#define BRW_WM_MAX_CONST 256 +#define BRW_WM_MAX_KILLS MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS + +#define PAYLOAD_DEPTH (PIPE_ATTRIB_MAX) + +#define MAX_IFSN 32 +#define MAX_LOOP_DEPTH 32 + +struct brw_wm_compile { + struct brw_compile func; + struct brw_wm_prog_key key; + struct brw_wm_prog_data prog_data; /* result */ + + struct brw_fragment_program *fp; + + unsigned grf_limit; + unsigned max_wm_grf; + + + struct brw_reg pixel_xy[2]; + struct brw_reg delta_xy[2]; + struct brw_reg pixel_w; + + + struct brw_reg wm_regs[8][32][4]; + + struct brw_reg payload_depth[4]; + struct brw_reg payload_coef[16]; + + struct brw_reg emit_mask_reg; + + struct brw_instruction *if_inst[MAX_IFSN]; + int if_insn; + + struct brw_instruction *loop_inst[MAX_LOOP_DEPTH]; + int loop_insn; + + struct brw_instruction *inst0; + struct brw_instruction *inst1; + + struct brw_reg stack; + struct brw_indirect stack_index; + + unsigned reg_index; + + unsigned tmp_start; + unsigned tmp_index; +}; + + + +void brw_wm_lookup_iz( unsigned line_aa, + unsigned lookup, + struct brw_wm_prog_key *key ); + +void brw_wm_glsl_emit(struct brw_wm_compile *c); +void brw_wm_emit_decls(struct brw_wm_compile *c); + +#endif diff --git a/src/gallium/drivers/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c new file mode 100644 index 0000000000..b45a333a2e --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_wm_decl.c @@ -0,0 +1,383 @@ + +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_wm.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" + +static struct brw_reg alloc_tmp(struct brw_wm_compile *c) +{ + c->tmp_index++; + c->reg_index = MAX2(c->reg_index, c->tmp_start + c->tmp_index); + return brw_vec8_grf(c->tmp_start + c->tmp_index, 0); +} + +static void release_tmps(struct brw_wm_compile *c) +{ + c->tmp_index = 0; +} + + + +static int is_null( struct brw_reg reg ) +{ + return (reg.file == BRW_ARCHITECTURE_REGISTER_FILE && + reg.nr == BRW_ARF_NULL); +} + +static void emit_pixel_xy( struct brw_wm_compile *c ) +{ + if (is_null(c->pixel_xy[0])) { + + struct brw_compile *p = &c->func; + struct brw_reg r1_uw = retype(brw_vec1_grf(1, 0), BRW_REGISTER_TYPE_UW); + + c->pixel_xy[0] = vec8(retype(alloc_tmp(c), BRW_REGISTER_TYPE_UW)); + c->pixel_xy[1] = vec8(retype(alloc_tmp(c), BRW_REGISTER_TYPE_UW)); + + /* Calculate pixel centers by adding 1 or 0 to each of the + * micro-tile coordinates passed in r1. + */ + brw_ADD(p, + c->pixel_xy[0], + stride(suboffset(r1_uw, 4), 2, 4, 0), + brw_imm_v(0x10101010)); + + brw_ADD(p, + c->pixel_xy[1], + stride(suboffset(r1_uw, 5), 2, 4, 0), + brw_imm_v(0x11001100)); + } +} + + + + + + +static void emit_delta_xy( struct brw_wm_compile *c ) +{ + if (is_null(c->delta_xy[0])) { + struct brw_compile *p = &c->func; + struct brw_reg r1 = brw_vec1_grf(1, 0); + + emit_pixel_xy(c); + + c->delta_xy[0] = alloc_tmp(c); + c->delta_xy[1] = alloc_tmp(c); + + /* Calc delta X,Y by subtracting origin in r1 from the pixel + * centers. + */ + brw_ADD(p, + c->delta_xy[0], + retype(c->pixel_xy[0], BRW_REGISTER_TYPE_UW), + negate(r1)); + + brw_ADD(p, + c->delta_xy[1], + retype(c->pixel_xy[1], BRW_REGISTER_TYPE_UW), + negate(suboffset(r1,1))); + } +} + + + +#if 0 +static void emit_pixel_w( struct brw_wm_compile *c ) +{ + if (is_null(c->pixel_w)) { + struct brw_compile *p = &c->func; + + struct brw_reg interp_wpos = c->coef_wpos; + + c->pixel_w = alloc_tmp(c); + + emit_delta_xy(c); + + /* Calc 1/w - just linterp wpos[3] optimized by putting the + * result straight into a message reg. + */ + struct brw_reg interp3 = brw_vec1_grf(interp_wpos.nr+1, 4); + brw_LINE(p, brw_null_reg(), interp3, c->delta_xy[0]); + brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), c->delta_xy[1]); + + /* Calc w */ + brw_math_16( p, + c->pixel_w, + BRW_MATH_FUNCTION_INV, + BRW_MATH_SATURATE_NONE, + 2, + brw_null_reg(), + BRW_MATH_PRECISION_FULL); + } +} +#endif + + +static void emit_cinterp(struct brw_wm_compile *c, + int idx, + int mask ) +{ + struct brw_compile *p = &c->func; + struct brw_reg interp[4]; + struct brw_reg coef = c->payload_coef[idx]; + int i; + + interp[0] = brw_vec1_grf(coef.nr, 0); + interp[1] = brw_vec1_grf(coef.nr, 4); + interp[2] = brw_vec1_grf(coef.nr+1, 0); + interp[3] = brw_vec1_grf(coef.nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<wm_regs[TGSI_FILE_INPUT][idx][i]; + brw_MOV(p, dst, suboffset(interp[i],3)); + } + } +} + +static void emit_linterp(struct brw_wm_compile *c, + int idx, + int mask ) +{ + struct brw_compile *p = &c->func; + struct brw_reg interp[4]; + struct brw_reg coef = c->payload_coef[idx]; + int i; + + emit_delta_xy(c); + + interp[0] = brw_vec1_grf(coef.nr, 0); + interp[1] = brw_vec1_grf(coef.nr, 4); + interp[2] = brw_vec1_grf(coef.nr+1, 0); + interp[3] = brw_vec1_grf(coef.nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<wm_regs[TGSI_FILE_INPUT][idx][i]; + brw_LINE(p, brw_null_reg(), interp[i], c->delta_xy[0]); + brw_MAC(p, dst, suboffset(interp[i],1), c->delta_xy[1]); + } + } +} + +#if 0 +static void emit_pinterp(struct brw_wm_compile *c, + int idx, + int mask ) +{ + struct brw_compile *p = &c->func; + struct brw_reg interp[4]; + struct brw_reg coef = c->payload_coef[idx]; + int i; + + get_delta_xy(c); + get_pixel_w(c); + + interp[0] = brw_vec1_grf(coef.nr, 0); + interp[1] = brw_vec1_grf(coef.nr, 4); + interp[2] = brw_vec1_grf(coef.nr+1, 0); + interp[3] = brw_vec1_grf(coef.nr+1, 4); + + for(i = 0; i < 4; i++ ) { + if (mask & (1<delta_xy[0]); + brw_MAC(p, dst, suboffset(interp[i],1), c->delta_xy[1]); + brw_MUL(p, dst, dst, c->pixel_w); + } + } +} +#endif + + + +#if 0 +static void emit_wpos( ) +{ + struct prog_dst_register dst = dst_reg(PROGRAM_INPUT, idx); + struct tgsi_full_src_register interp = src_reg(PROGRAM_PAYLOAD, idx); + struct tgsi_full_src_register deltas = get_delta_xy(c); + struct tgsi_full_src_register arg2; + unsigned opcode; + + opcode = WM_LINTERP; + arg2 = src_undef(); + + /* Have to treat wpos.xy specially: + */ + emit_op(c, + WM_WPOSXY, + dst_mask(dst, WRITEMASK_XY), + 0, 0, 0, + get_pixel_xy(c), + src_undef(), + src_undef()); + + dst = dst_mask(dst, WRITEMASK_ZW); + + /* PROGRAM_INPUT.attr.xyzw = INTERP payload.interp[attr].x, deltas.xyw + */ + emit_op(c, + WM_LINTERP, + dst, + 0, 0, 0, + interp, + deltas, + arg2); +} +#endif + + + + +/* Perform register allocation: + * + * -- r0??? + * -- passthrough depth regs (and stencil/aa??) + * -- curbe ?? + * -- inputs (coefficients) + * + * Use a totally static register allocation. This will perform poorly + * but is an easy way to get started (again). + */ +static void prealloc_reg(struct brw_wm_compile *c) +{ + int i, j; + int nr_curbe_regs = 0; + + /* R0, then some depth related regs: + */ + for (i = 0; i < c->key.nr_depth_regs; i++) { + c->payload_depth[i] = brw_vec8_grf(i*2, 0); + c->reg_index += 2; + } + + + /* Then a copy of our part of the CURBE entry: + */ + { + int nr_constants = c->fp->info.nr_regs[TGSI_FILE_CONSTANT]; + int index = 0; + + c->prog_data.max_const = 4*nr_constants; + for (i = 0; i < nr_constants; i++) { + for (j = 0; j < 4; j++, index++) + c->wm_regs[TGSI_FILE_CONSTANT][i][j] = brw_vec1_grf(c->reg_index + index/8, + index%8); + } + + nr_curbe_regs = 2*((4*nr_constants+15)/16); + c->reg_index += nr_curbe_regs; + } + + /* Adjust for parameter coefficients for position, which are + * currently always provided. + */ +// c->position_coef[i] = brw_vec8_grf(c->reg_index, 0); + c->reg_index += 2; + + /* Next we receive the plane coefficients for parameter + * interpolation: + */ + for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_INPUT]; i++) { + c->payload_coef[i] = brw_vec8_grf(c->reg_index, 0); + c->reg_index += 2; + } + + c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; + c->prog_data.urb_read_length = (c->fp->program.num_inputs + 1) * 2; + c->prog_data.curb_read_length = nr_curbe_regs; + + /* That's the end of the payload, now we can start allocating registers. + */ + c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); + c->reg_index++; + + c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); + c->reg_index += 2; + + /* Now allocate room for the interpolated inputs and staging + * registers for the outputs: + */ + for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_INPUT]; i++) + for (j = 0; j < 4; j++) + c->wm_regs[TGSI_FILE_INPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); + + for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_OUTPUT]; i++) + for (j = 0; j < 4; j++) + c->wm_regs[TGSI_FILE_OUTPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); + + /* Beyond this we should only need registers for internal temporaries: + */ + c->tmp_start = c->reg_index; +} + + + + + +/* Need to interpolate fragment program inputs in as a preamble to the + * shader. A more sophisticated compiler would do this on demand, but + * we'll do it up front: + */ +void brw_wm_emit_decls(struct brw_wm_compile *c) +{ + struct tgsi_parse_context parse; + int done = 0; + + prealloc_reg(c); + + tgsi_parse_init( &parse, c->fp->program.tokens ); + + while( !done && + !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; + unsigned first = decl->u.DeclarationRange.First; + unsigned last = decl->u.DeclarationRange.Last; + unsigned mask = decl->Declaration.UsageMask; /* ? */ + unsigned i; + + if (decl->Declaration.File != TGSI_FILE_INPUT) + break; + + assert(decl->Declaration.Interpolate); + + for( i = first; i <= last; i++ ) { + switch (decl->Interpolation.Interpolate) { + case TGSI_INTERPOLATE_CONSTANT: + emit_cinterp(c, i, mask); + break; + + case TGSI_INTERPOLATE_LINEAR: + emit_linterp(c, i, mask); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + //emit_pinterp(c, i, mask); + emit_linterp(c, i, mask); + break; + } + } + break; + } + case TGSI_TOKEN_TYPE_IMMEDIATE: + case TGSI_TOKEN_TYPE_INSTRUCTION: + default: + done = 1; + break; + } + } + + tgsi_parse_free (&parse); + + release_tmps(c); +} diff --git a/src/gallium/drivers/i965simple/brw_wm_glsl.c b/src/gallium/drivers/i965simple/brw_wm_glsl.c new file mode 100644 index 0000000000..d95645d108 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_wm_glsl.c @@ -0,0 +1,1079 @@ + +#include "brw_context.h" +#include "brw_eu.h" +#include "brw_wm.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/tgsi/util/tgsi_parse.h" + + + +static int get_scalar_dst_index(struct tgsi_full_instruction *inst) +{ + struct tgsi_dst_register dst = inst->FullDstRegisters[0].DstRegister; + int i; + for (i = 0; i < 4; i++) + if (dst.WriteMask & (1<tmp_index++; + c->reg_index = MAX2(c->reg_index, c->tmp_index); + return brw_vec8_grf(c->tmp_start + c->tmp_index, 0); +} + +static void release_tmps(struct brw_wm_compile *c) +{ + c->tmp_index = 0; +} + + +static struct brw_reg +get_reg(struct brw_wm_compile *c, int file, int index, int component ) +{ + switch (file) { + case TGSI_FILE_NULL: + return brw_null_reg(); + + case TGSI_FILE_SAMPLER: + /* Should never get here: + */ + assert (0); + return brw_null_reg(); + + case TGSI_FILE_IMMEDIATE: + /* These need a different path: + */ + assert(0); + return brw_null_reg(); + + + case TGSI_FILE_CONSTANT: + case TGSI_FILE_INPUT: + case TGSI_FILE_OUTPUT: + case TGSI_FILE_TEMPORARY: + case TGSI_FILE_ADDRESS: + return c->wm_regs[file][index][component]; + + default: + assert(0); + return brw_null_reg(); + } +} + + +static struct brw_reg get_dst_reg(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst, + int component) +{ + return get_reg(c, + inst->FullDstRegisters[0].DstRegister.File, + inst->FullDstRegisters[0].DstRegister.Index, + component); +} + +static int get_swz( struct tgsi_src_register src, int index ) +{ + switch (index & 3) { + case 0: return src.SwizzleX; + case 1: return src.SwizzleY; + case 2: return src.SwizzleZ; + case 3: return src.SwizzleW; + default: return 0; + } +} + +static int get_ext_swz( struct tgsi_src_register_ext_swz src, int index ) +{ + switch (index & 3) { + case 0: return src.ExtSwizzleX; + case 1: return src.ExtSwizzleY; + case 2: return src.ExtSwizzleZ; + case 3: return src.ExtSwizzleW; + default: return 0; + } +} + +static struct brw_reg get_src_reg(struct brw_wm_compile *c, + struct tgsi_full_src_register *src, + int index) +{ + struct brw_reg reg; + int component = index; + int neg = 0; + int abs = 0; + + if (src->SrcRegister.Negate) + neg = 1; + + component = get_swz(src->SrcRegister, component); + + /* Yes, there are multiple negates: + */ + switch (component & 3) { + case 0: neg ^= src->SrcRegisterExtSwz.NegateX; break; + case 1: neg ^= src->SrcRegisterExtSwz.NegateY; break; + case 2: neg ^= src->SrcRegisterExtSwz.NegateZ; break; + case 3: neg ^= src->SrcRegisterExtSwz.NegateW; break; + } + + /* And multiple swizzles, fun isn't it: + */ + component = get_ext_swz(src->SrcRegisterExtSwz, component); + + /* Can't handle this, don't know if we need to: + */ + assert(src->SrcRegisterExtSwz.ExtDivide == TGSI_EXTSWIZZLE_ONE); + + /* Not handling indirect lookups yet: + */ + assert(src->SrcRegister.Indirect == 0); + + /* Don't know what dimension means: + */ + assert(src->SrcRegister.Dimension == 0); + + /* Will never handle any of this stuff: + */ + assert(src->SrcRegisterExtMod.Complement == 0); + assert(src->SrcRegisterExtMod.Bias == 0); + assert(src->SrcRegisterExtMod.Scale2X == 0); + + if (src->SrcRegisterExtMod.Absolute) + abs = 1; + + /* Another negate! This is a post-absolute negate, which we + * can't do. Need to clean the crap out of tgsi somehow. + */ + assert(src->SrcRegisterExtMod.Negate == 0); + + switch( component ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + reg = get_reg(c, + src->SrcRegister.File, + src->SrcRegister.Index, + component ); + + if (neg) + reg = negate(reg); + + if (abs) + reg = brw_abs(reg); + + break; + + /* XXX: this won't really work in the general case, but we know + * that the extended swizzle is only allowed in the SWZ + * instruction (right??), in which case using an immediate + * directly will work. + */ + case TGSI_EXTSWIZZLE_ZERO: + reg = brw_imm_f(0); + break; + + case TGSI_EXTSWIZZLE_ONE: + if (neg && !abs) + reg = brw_imm_f(-1.0); + else + reg = brw_imm_f(1.0); + break; + + default: + assert(0); + break; + } + + + return reg; +} + +static void emit_abs( struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + + int i; + struct brw_compile *p = &c->func; + brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + brw_MOV(p, dst, brw_abs(src)); /* NOTE */ + } + } + brw_set_saturate(p, 0); +} + + +static void emit_xpd(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + int i; + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + for (i = 0; i < 4; i++) { + unsigned i2 = (i+2)%3; + unsigned i1 = (i+1)%3; + if (mask & (1<FullSrcRegisters[0], i2)); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i1); + brw_MUL(p, brw_null_reg(), src0, src1); + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i1); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i2); + brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); + brw_MAC(p, dst, src0, src1); + brw_set_saturate(p, 0); + } + } + brw_set_saturate(p, 0); +} + +static void emit_dp3(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_reg src0[3], src1[3], dst; + int i; + struct brw_compile *p = &c->func; + for (i = 0; i < 3; i++) { + src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); + src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i); + } + + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MAC(p, dst, src0[2], src1[2]); + brw_set_saturate(p, 0); +} + +static void emit_dp4(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_reg src0[4], src1[4], dst; + int i; + struct brw_compile *p = &c->func; + for (i = 0; i < 4; i++) { + src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); + src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i); + } + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_MAC(p, brw_null_reg(), src0[2], src1[2]); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MAC(p, dst, src0[3], src1[3]); + brw_set_saturate(p, 0); +} + +static void emit_dph(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_reg src0[4], src1[4], dst; + int i; + struct brw_compile *p = &c->func; + for (i = 0; i < 4; i++) { + src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); + src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i); + } + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + brw_MUL(p, brw_null_reg(), src0[0], src1[0]); + brw_MAC(p, brw_null_reg(), src0[1], src1[1]); + brw_MAC(p, dst, src0[2], src1[2]); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_ADD(p, dst, src0[3], src1[3]); + brw_set_saturate(p, 0); +} + +static void emit_math1(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst, unsigned func) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; + + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + brw_MOV(p, brw_message_reg(2), src0); + brw_math(p, + dst, + func, + ((inst->Instruction.Saturate != TGSI_SAT_NONE) + ? BRW_MATH_SATURATE_SATURATE + : BRW_MATH_SATURATE_NONE), + 2, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); +} + + +static void emit_alu2(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst, + unsigned opcode) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, src1, dst; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + int i; + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + brw_alu2(p, opcode, dst, src0, src1); + } + } + brw_set_saturate(p, 0); +} + + +static void emit_alu1(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst, + unsigned opcode) +{ + struct brw_compile *p = &c->func; + struct brw_reg src0, dst; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + int i; + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + for (i = 0 ; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + brw_alu1(p, opcode, dst, src0); + } + } + if (inst->Instruction.Saturate != TGSI_SAT_NONE) + brw_set_saturate(p, 0); +} + + +static void emit_max(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg src0, src1, dst; + int i; + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MOV(p, dst, src0); + brw_set_saturate(p, 0); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src0, src1); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, src1); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + } + } + brw_pop_insn_state(p); +} + +static void emit_min(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg src0, src1, dst; + int i; + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MOV(p, dst, src0); + brw_set_saturate(p, 0); + + brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src1, src0); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, src1); + brw_set_saturate(p, 0); + brw_set_predicate_control_flag_value(p, 0xff); + } + } + brw_pop_insn_state(p); +} + +static void emit_pow(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_compile *p = &c->func; + struct brw_reg dst, src0, src1; + dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], 0); + + brw_MOV(p, brw_message_reg(2), src0); + brw_MOV(p, brw_message_reg(3), src1); + + brw_math(p, + dst, + BRW_MATH_FUNCTION_POW, + (inst->Instruction.Saturate != TGSI_SAT_NONE + ? BRW_MATH_SATURATE_SATURATE + : BRW_MATH_SATURATE_NONE), + 2, + brw_null_reg(), + BRW_MATH_DATA_VECTOR, + BRW_MATH_PRECISION_FULL); +} + +static void emit_lrp(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg dst, tmp1, tmp2, src0, src1, src2; + int i; + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + + if (src1.nr == dst.nr) { + tmp1 = alloc_tmp(c); + brw_MOV(p, tmp1, src1); + } else + tmp1 = src1; + + src2 = get_src_reg(c, &inst->FullSrcRegisters[2], i); + if (src2.nr == dst.nr) { + tmp2 = alloc_tmp(c); + brw_MOV(p, tmp2, src2); + } else + tmp2 = src2; + + brw_ADD(p, dst, negate(src0), brw_imm_f(1.0)); + brw_MUL(p, brw_null_reg(), dst, tmp2); + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_MAC(p, dst, src0, tmp1); + brw_set_saturate(p, 0); + } + release_tmps(c); + } +} + +static void emit_kil(struct brw_wm_compile *c) +{ + struct brw_compile *p = &c->func; + struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK + brw_AND(p, depth, c->emit_mask_reg, depth); + brw_pop_insn_state(p); +} + +static void emit_mad(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg dst, src0, src1, src2; + int i; + + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + src2 = get_src_reg(c, &inst->FullSrcRegisters[2], i); + brw_MUL(p, dst, src0, src1); + + brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); + brw_ADD(p, dst, dst, src2); + brw_set_saturate(p, 0); + } + } +} + +static void emit_sop(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst, unsigned cond) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg dst, src0, src1; + int i; + + brw_push_insn_state(p); + for (i = 0; i < 4; i++) { + if (mask & (1<FullSrcRegisters[0], i); + src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); + brw_CMP(p, brw_null_reg(), cond, src0, src1); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + brw_MOV(p, dst, brw_imm_f(0.0)); + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + brw_MOV(p, dst, brw_imm_f(1.0)); + } + } + brw_pop_insn_state(p); +} + + +static void emit_ddx(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_compile *p = &c->func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg interp[4]; + struct brw_reg dst; + struct brw_reg src0, w; + unsigned nr, i; + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); + w = get_src_reg(c, &inst->FullSrcRegisters[1], 3); + nr = src0.nr; + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; + struct brw_reg interp[4]; + struct brw_reg dst; + struct brw_reg src0, w; + unsigned nr, i; + + src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); + nr = src0.nr; + w = get_src_reg(c, &inst->FullSrcRegisters[1], 3); + interp[0] = brw_vec1_grf(nr, 0); + interp[1] = brw_vec1_grf(nr, 4); + interp[2] = brw_vec1_grf(nr+1, 0); + interp[3] = brw_vec1_grf(nr+1, 4); + brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); + for(i = 0; i < 4; i++ ) { + if (mask & (1<func; + struct brw_reg payload_reg = c->payload_depth[0]; + struct brw_reg dst[4], src[4]; + unsigned i; + for (i = 0; i < 4; i++) + dst[i] = get_dst_reg(c, inst, i); + for (i = 0; i < 4; i++) + src[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); + +#if 0 + switch (inst->TexSrcTarget) { + case TEXTURE_1D_INDEX: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), brw_imm_f(0)); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); + break; + case TEXTURE_2D_INDEX: + case TEXTURE_RECT_INDEX: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); + break; + default: + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), src[2]); + break; + } +#else + brw_MOV(p, brw_message_reg(2), src[0]); + brw_MOV(p, brw_message_reg(3), src[1]); + brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); +#endif + + brw_MOV(p, brw_message_reg(5), src[3]); + brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); + brw_SAMPLE(p, + retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), + 1, + retype(payload_reg, BRW_REGISTER_TYPE_UW), + inst->TexSrcUnit + 1, /* surface */ + inst->TexSrcUnit, /* sampler */ + inst->FullDstRegisters[0].DstRegister.WriteMask, + BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS, + 4, + 4, + 0); +#endif +} + +static void emit_tex(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ +#if 0 + struct brw_compile *p = &c->func; + struct brw_reg payload_reg = c->payload_depth[0]; + struct brw_reg dst[4], src[4]; + unsigned msg_len; + unsigned i, nr; + unsigned emit; + boolean shadow = (c->key.shadowtex_mask & (1<TexSrcUnit)) ? 1 : 0; + + for (i = 0; i < 4; i++) + dst[i] = get_dst_reg(c, inst, i); + for (i = 0; i < 4; i++) + src[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); + +#if 0 + switch (inst->TexSrcTarget) { + case TEXTURE_1D_INDEX: + emit = WRITEMASK_X; + nr = 1; + break; + case TEXTURE_2D_INDEX: + case TEXTURE_RECT_INDEX: + emit = WRITEMASK_XY; + nr = 2; + break; + default: + emit = WRITEMASK_XYZ; + nr = 3; + break; + } +#else + emit = WRITEMASK_XY; + nr = 2; +#endif + + msg_len = 1; + + for (i = 0; i < nr; i++) { + static const unsigned swz[4] = {0,1,2,2}; + if (emit & (1<TexSrcUnit + 1, /* surface */ + inst->TexSrcUnit, /* sampler */ + inst->FullDstRegisters[0].DstRegister.WriteMask, + BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE, + 4, + shadow ? 6 : 4, + 0); + + if (shadow) + brw_MOV(p, dst[3], brw_imm_f(1.0)); +#endif +} + + + + + + + + +static void emit_fb_write(struct brw_wm_compile *c, + struct tgsi_full_instruction *inst) +{ + struct brw_compile *p = &c->func; + int nr = 2; + int channel; + int base_reg = 0; + + // src0 = output color + // src1 = payload_depth[0] + // src2 = output depth + // dst = ??? + + + + /* Reserve a space for AA - may not be needed: + */ + if (c->key.aa_dest_stencil_reg) + nr += 1; + + { + brw_push_insn_state(p); + for (channel = 0; channel < 4; channel++) { + struct brw_reg src0 = c->wm_regs[TGSI_FILE_OUTPUT][0][channel]; + + /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ + /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ + brw_MOV(p, brw_message_reg(nr + channel), src0); + } + /* skip over the regs populated above: */ + nr += 8; + brw_pop_insn_state(p); + } + + + /* Pass through control information: + */ + /* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ + { + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ + brw_MOV(p, + brw_message_reg(base_reg + 1), + brw_vec8_grf(1, 0)); + brw_pop_insn_state(p); + } + + /* Send framebuffer write message: */ + brw_fb_WRITE(p, + retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW), + base_reg, + retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), + 0, /* render surface always 0 */ + nr, + 0, + 1); + +} + + +static void brw_wm_emit_instruction( struct brw_wm_compile *c, + struct tgsi_full_instruction *inst ) +{ + struct brw_compile *p = &c->func; + +#if 0 + if (inst->CondUpdate) + brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); + else + brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); +#else + brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); +#endif + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ABS: + emit_abs(c, inst); + break; + case TGSI_OPCODE_ADD: + emit_alu2(c, inst, BRW_OPCODE_ADD); + break; + case TGSI_OPCODE_SUB: + assert(0); +// emit_alu2(c, inst, BRW_OPCODE_SUB); + break; + case TGSI_OPCODE_FRC: + emit_alu1(c, inst, BRW_OPCODE_FRC); + break; + case TGSI_OPCODE_FLR: + assert(0); +// emit_alu1(c, inst, BRW_OPCODE_FLR); + break; + case TGSI_OPCODE_LRP: + emit_lrp(c, inst); + break; + case TGSI_OPCODE_INT: + emit_alu1(c, inst, BRW_OPCODE_RNDD); + break; + case TGSI_OPCODE_MOV: + emit_alu1(c, inst, BRW_OPCODE_MOV); + break; + case TGSI_OPCODE_DP3: + emit_dp3(c, inst); + break; + case TGSI_OPCODE_DP4: + emit_dp4(c, inst); + break; + case TGSI_OPCODE_XPD: + emit_xpd(c, inst); + break; + case TGSI_OPCODE_DPH: + emit_dph(c, inst); + break; + case TGSI_OPCODE_RCP: + emit_math1(c, inst, BRW_MATH_FUNCTION_INV); + break; + case TGSI_OPCODE_RSQ: + emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ); + break; + case TGSI_OPCODE_SIN: + emit_math1(c, inst, BRW_MATH_FUNCTION_SIN); + break; + case TGSI_OPCODE_COS: + emit_math1(c, inst, BRW_MATH_FUNCTION_COS); + break; + case TGSI_OPCODE_EX2: + emit_math1(c, inst, BRW_MATH_FUNCTION_EXP); + break; + case TGSI_OPCODE_LG2: + emit_math1(c, inst, BRW_MATH_FUNCTION_LOG); + break; + case TGSI_OPCODE_MAX: + emit_max(c, inst); + break; + case TGSI_OPCODE_MIN: + emit_min(c, inst); + break; + case TGSI_OPCODE_DDX: + emit_ddx(c, inst); + break; + case TGSI_OPCODE_DDY: + emit_ddy(c, inst); + break; + case TGSI_OPCODE_SLT: + emit_sop(c, inst, BRW_CONDITIONAL_L); + break; + case TGSI_OPCODE_SLE: + emit_sop(c, inst, BRW_CONDITIONAL_LE); + break; + case TGSI_OPCODE_SGT: + emit_sop(c, inst, BRW_CONDITIONAL_G); + break; + case TGSI_OPCODE_SGE: + emit_sop(c, inst, BRW_CONDITIONAL_GE); + break; + case TGSI_OPCODE_SEQ: + emit_sop(c, inst, BRW_CONDITIONAL_EQ); + break; + case TGSI_OPCODE_SNE: + emit_sop(c, inst, BRW_CONDITIONAL_NEQ); + break; + case TGSI_OPCODE_MUL: + emit_alu2(c, inst, BRW_OPCODE_MUL); + break; + case TGSI_OPCODE_POW: + emit_pow(c, inst); + break; + case TGSI_OPCODE_MAD: + emit_mad(c, inst); + break; + case TGSI_OPCODE_TEX: + emit_tex(c, inst); + break; + case TGSI_OPCODE_TXB: + emit_txb(c, inst); + break; + case TGSI_OPCODE_TEXKILL: + emit_kil(c); + break; + case TGSI_OPCODE_IF: + assert(c->if_insn < MAX_IFSN); + c->if_inst[c->if_insn++] = brw_IF(p, BRW_EXECUTE_8); + break; + case TGSI_OPCODE_ELSE: + c->if_inst[c->if_insn-1] = brw_ELSE(p, c->if_inst[c->if_insn-1]); + break; + case TGSI_OPCODE_ENDIF: + assert(c->if_insn > 0); + brw_ENDIF(p, c->if_inst[--c->if_insn]); + break; + case TGSI_OPCODE_BGNSUB: + case TGSI_OPCODE_ENDSUB: + break; + case TGSI_OPCODE_CAL: + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_ADD(p, deref_1ud(c->stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_ADD(p, + get_addr_reg(c->stack_index), + get_addr_reg(c->stack_index), brw_imm_d(4)); +// orig_inst = inst->Data; +// orig_inst->Data = &p->store[p->nr_insn]; + assert(0); + brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); + brw_pop_insn_state(p); + break; + + case TGSI_OPCODE_RET: +#if 0 + brw_push_insn_state(p); + brw_set_mask_control(p, BRW_MASK_DISABLE); + brw_ADD(p, + get_addr_reg(c->stack_index), + get_addr_reg(c->stack_index), brw_imm_d(-4)); + brw_set_access_mode(p, BRW_ALIGN_1); + brw_MOV(p, brw_ip_reg(), deref_1ud(c->stack_index, 0)); + brw_set_access_mode(p, BRW_ALIGN_16); + brw_pop_insn_state(p); +#else + emit_fb_write(c, inst); +#endif + + break; + case TGSI_OPCODE_LOOP: + c->loop_inst[c->loop_insn++] = brw_DO(p, BRW_EXECUTE_8); + break; + case TGSI_OPCODE_BRK: + brw_BREAK(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case TGSI_OPCODE_CONT: + brw_CONT(p); + brw_set_predicate_control(p, BRW_PREDICATE_NONE); + break; + case TGSI_OPCODE_ENDLOOP: + c->loop_insn--; + c->inst0 = c->inst1 = brw_WHILE(p, c->loop_inst[c->loop_insn]); + /* patch all the BREAK instructions from + last BEGINLOOP */ + while (c->inst0 > c->loop_inst[c->loop_insn]) { + c->inst0--; + if (c->inst0->header.opcode == BRW_OPCODE_BREAK) { + c->inst0->bits3.if_else.jump_count = c->inst1 - c->inst0 + 1; + c->inst0->bits3.if_else.pop_count = 0; + } else if (c->inst0->header.opcode == BRW_OPCODE_CONTINUE) { + c->inst0->bits3.if_else.jump_count = c->inst1 - c->inst0; + c->inst0->bits3.if_else.pop_count = 0; + } + } + break; + case TGSI_OPCODE_END: + emit_fb_write(c, inst); + break; + + default: + debug_printf("unsupported IR in fragment shader %d\n", + inst->Instruction.Opcode); + } +#if 0 + if (inst->CondUpdate) + brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); + else + brw_set_predicate_control(p, BRW_PREDICATE_NONE); +#endif +} + + + + + + +void brw_wm_glsl_emit(struct brw_wm_compile *c) +{ + struct tgsi_parse_context parse; + struct brw_compile *p = &c->func; + + brw_init_compile(&c->func); + brw_set_compression_control(p, BRW_COMPRESSION_NONE); + + c->reg_index = 0; + c->if_insn = 0; + c->loop_insn = 0; + c->stack_index = brw_indirect(0,0); + + /* Do static register allocation and parameter interpolation: + */ + brw_wm_emit_decls( c ); + + /* Emit the actual program. All done with very direct translation, + * hopefully we can improve on this shortly... + */ + brw_MOV(p, get_addr_reg(c->stack_index), brw_address(c->stack)); + + tgsi_parse_init( &parse, c->fp->program.tokens ); + + while( !tgsi_parse_end_of_tokens( &parse ) ) + { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + /* already done */ + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* not handled yet */ + assert(0); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + brw_wm_emit_instruction(c, &parse.FullToken.FullInstruction); + break; + + default: + assert( 0 ); + } + } + + tgsi_parse_free (&parse); + + /* Fix up call targets: + */ +#if 0 + { + unsigned nr_insns = c->fp->program.Base.NumInstructions; + unsigned insn, target_insn; + struct tgsi_full_instruction *inst1, *inst2; + struct brw_instruction *brw_inst1, *brw_inst2; + int offset; + for (insn = 0; insn < nr_insns; insn++) { + inst1 = &c->fp->program.Base.Instructions[insn]; + brw_inst1 = inst1->Data; + switch (inst1->Opcode) { + case TGSI_OPCODE_CAL: + target_insn = inst1->BranchTarget; + inst2 = &c->fp->program.Base.Instructions[target_insn]; + brw_inst2 = inst2->Data; + offset = brw_inst2 - brw_inst1; + brw_set_src1(brw_inst1, brw_imm_d(offset*16)); + break; + default: + break; + } + } + } +#endif + + c->prog_data.total_grf = c->reg_index; + c->prog_data.total_scratch = 0; +} diff --git a/src/gallium/drivers/i965simple/brw_wm_iz.c b/src/gallium/drivers/i965simple/brw_wm_iz.c new file mode 100644 index 0000000000..6c5f25bf39 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_wm_iz.c @@ -0,0 +1,214 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_wm.h" + + +#undef P /* prompted depth */ +#undef C /* computed */ +#undef N /* non-promoted? */ + +#define P 0 +#define C 1 +#define N 2 + +const struct { + unsigned mode:2; + unsigned sd_present:1; + unsigned sd_to_rt:1; + unsigned dd_present:1; + unsigned ds_present:1; +} wm_iz_table[IZ_BIT_MAX] = +{ + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 1, 1, 0, 0 }, + { C, 1, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 1, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 1, 1, 0, 0 }, + { C, 1, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 0, 1, 0, 0 }, + { C, 1, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 1, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 0, 0, 1 }, + { C, 0, 0, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 1, 1, 0, 1 }, + { C, 1, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 1, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 0, 0, 1 }, + { C, 0, 0, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 1, 1, 0, 1 }, + { C, 1, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 1, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { N, 0, 1, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 0 }, + { C, 0, 1, 1, 0 }, + { C, 0, 1, 1, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 1 }, + { N, 0, 1, 0, 1 }, + { N, 0, 1, 0, 1 }, + { P, 0, 0, 0, 0 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { N, 1, 1, 0, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { C, 0, 0, 0, 1 }, + { P, 0, 0, 0, 0 }, + { C, 0, 1, 0, 1 }, + { P, 0, 0, 0, 0 }, + { C, 1, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { C, 0, 1, 0, 1 }, + { P, 0, 0, 0, 0 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { P, 0, 0, 0, 0 }, + { C, 1, 1, 1, 1 }, + { C, 0, 1, 1, 1 }, + { C, 0, 1, 1, 1 } +}; + +void brw_wm_lookup_iz( unsigned line_aa, + unsigned lookup, + struct brw_wm_prog_key *key ) +{ + unsigned reg = 2; + + assert (lookup < IZ_BIT_MAX); + + if (lookup & IZ_PS_COMPUTES_DEPTH_BIT) + key->computes_depth = 1; + + if (wm_iz_table[lookup].sd_present) { + key->source_depth_reg = reg; + reg += 2; + } + + if (wm_iz_table[lookup].sd_to_rt) + key->source_depth_to_render_target = 1; + + if (wm_iz_table[lookup].ds_present || line_aa != AA_NEVER) { + key->aa_dest_stencil_reg = reg; + key->runtime_check_aads_emit = (!wm_iz_table[lookup].ds_present && + line_aa == AA_SOMETIMES); + reg++; + } + + if (wm_iz_table[lookup].dd_present) { + key->dest_depth_reg = reg; + reg+=2; + } + + key->nr_depth_regs = (reg+1)/2; +} + diff --git a/src/gallium/drivers/i965simple/brw_wm_sampler_state.c b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c new file mode 100644 index 0000000000..de42ffc5b1 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c @@ -0,0 +1,273 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + +#include "pipe/p_util.h" + + +#define COMPAREFUNC_ALWAYS 0 +#define COMPAREFUNC_NEVER 0x1 +#define COMPAREFUNC_LESS 0x2 +#define COMPAREFUNC_EQUAL 0x3 +#define COMPAREFUNC_LEQUAL 0x4 +#define COMPAREFUNC_GREATER 0x5 +#define COMPAREFUNC_NOTEQUAL 0x6 +#define COMPAREFUNC_GEQUAL 0x7 + +/* Samplers aren't strictly wm state from the hardware's perspective, + * but that is the only situation in which we use them in this driver. + */ + +static int intel_translate_shadow_compare_func(unsigned func) +{ + switch(func) { + case PIPE_FUNC_NEVER: + return COMPAREFUNC_ALWAYS; + case PIPE_FUNC_LESS: + return COMPAREFUNC_LEQUAL; + case PIPE_FUNC_LEQUAL: + return COMPAREFUNC_LESS; + case PIPE_FUNC_GREATER: + return COMPAREFUNC_GEQUAL; + case PIPE_FUNC_GEQUAL: + return COMPAREFUNC_GREATER; + case PIPE_FUNC_NOTEQUAL: + return COMPAREFUNC_EQUAL; + case PIPE_FUNC_EQUAL: + return COMPAREFUNC_NOTEQUAL; + case PIPE_FUNC_ALWAYS: + return COMPAREFUNC_NEVER; + } + + debug_printf("Unknown value in %s: %x\n", __FUNCTION__, func); + return COMPAREFUNC_NEVER; +} + +/* The brw (and related graphics cores) do not support GL_CLAMP. The + * Intel drivers for "other operating systems" implement GL_CLAMP as + * GL_CLAMP_TO_EDGE, so the same is done here. + */ +static unsigned translate_wrap_mode( int wrap ) +{ + switch( wrap ) { + case PIPE_TEX_WRAP_REPEAT: + return BRW_TEXCOORDMODE_WRAP; + case PIPE_TEX_WRAP_CLAMP: + return BRW_TEXCOORDMODE_CLAMP; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + return BRW_TEXCOORDMODE_CLAMP; /* conform likes it this way */ + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return BRW_TEXCOORDMODE_CLAMP_BORDER; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + return BRW_TEXCOORDMODE_MIRROR; + default: + return BRW_TEXCOORDMODE_WRAP; + } +} + + +static unsigned U_FIXED(float value, unsigned frac_bits) +{ + value *= (1<cache[BRW_SAMPLER_DEFAULT_COLOR], &sdc ); +} + + +/* + */ +static void brw_update_sampler_state( const struct pipe_sampler_state *pipe_sampler, + unsigned sdc_gs_offset, + struct brw_sampler_state *sampler) +{ + memset(sampler, 0, sizeof(*sampler)); + + switch (pipe_sampler->min_mip_filter) { + case PIPE_TEX_FILTER_NEAREST: + sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST; + break; + case PIPE_TEX_FILTER_LINEAR: + sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR; + break; + default: + break; + } + + switch (pipe_sampler->min_mip_filter) { + case PIPE_TEX_MIPFILTER_NEAREST: + sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST; + break; + case PIPE_TEX_MIPFILTER_LINEAR: + sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR; + break; + case PIPE_TEX_MIPFILTER_NONE: + sampler->ss0.mip_filter = BRW_MIPFILTER_NONE; + break; + default: + break; + } + /* Set Anisotropy: + */ + if (pipe_sampler->max_anisotropy > 1.0) { + sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC; + sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC; + + if (pipe_sampler->max_anisotropy > 2.0) { + sampler->ss3.max_aniso = MAX2((pipe_sampler->max_anisotropy - 2) / 2, + BRW_ANISORATIO_16); + } + } + else { + switch (pipe_sampler->mag_img_filter) { + case PIPE_TEX_FILTER_NEAREST: + sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST; + break; + case PIPE_TEX_FILTER_LINEAR: + sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR; + break; + default: + break; + } + } + + sampler->ss1.s_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_s); + sampler->ss1.r_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_r); + sampler->ss1.t_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_t); + + /* Fulsim complains if I don't do this. Hardware doesn't mind: + */ +#if 0 + if (texObj->Target == GL_TEXTURE_CUBE_MAP_ARB) { + sampler->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CUBE; + sampler->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CUBE; + sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CUBE; + } +#endif + + /* Set shadow function: + */ + if (pipe_sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + /* Shadowing is "enabled" by emitting a particular sampler + * message (sample_c). So need to recompile WM program when + * shadow comparison is enabled on each/any texture unit. + */ + sampler->ss0.shadow_function = intel_translate_shadow_compare_func(pipe_sampler->compare_func); + } + + /* Set LOD bias: + */ + sampler->ss0.lod_bias = S_FIXED(CLAMP(pipe_sampler->lod_bias, -16, 15), 6); + + sampler->ss0.lod_preclamp = 1; /* OpenGL mode */ + sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */ + + /* Set BaseMipLevel, MaxLOD, MinLOD: + * + * XXX: I don't think that using firstLevel, lastLevel works, + * because we always setup the surface state as if firstLevel == + * level zero. Probably have to subtract firstLevel from each of + * these: + */ + sampler->ss0.base_level = U_FIXED(0, 1); + + sampler->ss1.max_lod = U_FIXED(MIN2(MAX2(pipe_sampler->max_lod, 0), 13), 6); + sampler->ss1.min_lod = U_FIXED(MIN2(MAX2(pipe_sampler->min_lod, 0), 13), 6); + + sampler->ss2.default_color_pointer = sdc_gs_offset >> 5; +} + + + +/* All samplers must be uploaded in a single contiguous array, which + * complicates various things. However, this is still too confusing - + * FIXME: simplify all the different new texture state flags. + */ +static void upload_wm_samplers(struct brw_context *brw) +{ + unsigned unit; + unsigned sampler_count = 0; + + /* BRW_NEW_SAMPLER */ + for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { + /* determine unit enable/disable by looking for a bound texture */ + if (brw->attribs.Texture[unit]) { + const struct pipe_sampler_state *sampler = brw->attribs.Samplers[unit]; + unsigned sdc_gs_offset = upload_default_color(brw, sampler->border_color); + + brw_update_sampler_state(sampler, + sdc_gs_offset, + &brw->wm.sampler[unit]); + + sampler_count = unit + 1; + } + } + + if (brw->wm.sampler_count != sampler_count) { + brw->wm.sampler_count = sampler_count; + brw->state.dirty.cache |= CACHE_NEW_SAMPLER; + } + + brw->wm.sampler_gs_offset = 0; + + if (brw->wm.sampler_count) + brw->wm.sampler_gs_offset = + brw_cache_data_sz(&brw->cache[BRW_SAMPLER], + brw->wm.sampler, + sizeof(struct brw_sampler_state) * brw->wm.sampler_count); +} + +const struct brw_tracked_state brw_wm_samplers = { + .dirty = { + .brw = BRW_NEW_SAMPLER, + .cache = 0 + }, + .update = upload_wm_samplers +}; + diff --git a/src/gallium/drivers/i965simple/brw_wm_state.c b/src/gallium/drivers/i965simple/brw_wm_state.c new file mode 100644 index 0000000000..5ccd488842 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_wm_state.c @@ -0,0 +1,194 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" +#include "brw_wm.h" +#include "pipe/p_util.h" + +/*********************************************************************** + * WM unit - fragment programs and rasterization + */ +static void upload_wm_unit(struct brw_context *brw ) +{ + struct brw_wm_unit_state wm; + unsigned max_threads; + unsigned per_thread; + + if (BRW_DEBUG & DEBUG_SINGLE_THREAD) + max_threads = 0; + else + max_threads = 31; + + + memset(&wm, 0, sizeof(wm)); + + /* CACHE_NEW_WM_PROG */ + wm.thread0.grf_reg_count = align(brw->wm.prog_data->total_grf, 16) / 16 - 1; + wm.thread0.kernel_start_pointer = brw->wm.prog_gs_offset >> 6; + wm.thread3.dispatch_grf_start_reg = brw->wm.prog_data->first_curbe_grf; + wm.thread3.urb_entry_read_length = brw->wm.prog_data->urb_read_length; + wm.thread3.const_urb_entry_read_length = brw->wm.prog_data->curb_read_length; + + wm.wm5.max_threads = max_threads; + + per_thread = align(brw->wm.prog_data->total_scratch, 1024); + assert(per_thread <= 12 * 1024); + +#if 0 + if (brw->wm.prog_data->total_scratch) { + unsigned total = per_thread * (max_threads + 1); + + /* Scratch space -- just have to make sure there is sufficient + * allocated for the active program and current number of threads. + */ + brw->wm.scratch_buffer_size = total; + if (brw->wm.scratch_buffer && + brw->wm.scratch_buffer_size > brw->wm.scratch_buffer->size) { + dri_bo_unreference(brw->wm.scratch_buffer); + brw->wm.scratch_buffer = NULL; + } + if (!brw->wm.scratch_buffer) { + brw->wm.scratch_buffer = dri_bo_alloc(intel->intelScreen->bufmgr, + "wm scratch", + brw->wm.scratch_buffer_size, + 4096, DRM_BO_FLAG_MEM_TT); + } + } + /* XXX: Scratch buffers are not implemented correectly. + * + * The scratch offset to be programmed into wm is relative to the general + * state base address. However, using dri_bo_alloc/dri_bo_emit_reloc (or + * the previous bmGenBuffers scheme), we get an offset relative to the + * start of framebuffer. Even before then, it was broken in other ways, + * so just fail for now if we hit that path. + */ + assert(brw->wm.prog_data->total_scratch == 0); +#endif + + /* CACHE_NEW_SURFACE */ + wm.thread1.binding_table_entry_count = brw->wm.nr_surfaces; + + /* BRW_NEW_CURBE_OFFSETS */ + wm.thread3.const_urb_entry_read_offset = brw->curbe.wm_start * 2; + + wm.thread3.urb_entry_read_offset = 0; + wm.thread1.depth_coef_urb_read_offset = 1; + wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; + + /* CACHE_NEW_SAMPLER */ + wm.wm4.sampler_count = (brw->wm.sampler_count + 1) / 4; + wm.wm4.sampler_state_pointer = brw->wm.sampler_gs_offset >> 5; + + /* BRW_NEW_FRAGMENT_PROGRAM */ + { + const struct brw_fragment_program *fp = brw->attribs.FragmentProgram; + + if (fp->UsesDepth) + wm.wm5.program_uses_depth = 1; /* as far as we can tell */ + + if (fp->ComputesDepth) + wm.wm5.program_computes_depth = 1; + + /* BRW_NEW_ALPHA_TEST */ + if (fp->UsesKill || + brw->attribs.DepthStencil->alpha.enabled) + wm.wm5.program_uses_killpixel = 1; + + wm.wm5.enable_8_pix = 1; + } + + wm.wm5.thread_dispatch_enable = 1; /* AKA: color_write */ + wm.wm5.legacy_line_rast = 0; + wm.wm5.legacy_global_depth_bias = 0; + wm.wm5.early_depth_test = 1; /* never need to disable */ + wm.wm5.line_aa_region_width = 0; + wm.wm5.line_endcap_aa_region_width = 1; + + /* BRW_NEW_RASTERIZER */ + if (brw->attribs.Raster->poly_stipple_enable) + wm.wm5.polygon_stipple = 1; + +#if 0 + if (brw->attribs.Polygon->OffsetFill) { + wm.wm5.depth_offset = 1; + /* Something wierd going on with legacy_global_depth_bias, + * offset_constant, scaling and MRD. This value passes glean + * but gives some odd results elsewere (eg. the + * quad-offset-units test). + */ + wm.global_depth_offset_constant = brw->attribs.Polygon->OffsetUnits * 2; + + /* This is the only value that passes glean: + */ + wm.global_depth_offset_scale = brw->attribs.Polygon->OffsetFactor; + } +#endif + + if (brw->attribs.Raster->line_stipple_enable) { + wm.wm5.line_stipple = 1; + } + + if (BRW_DEBUG & DEBUG_STATS) + wm.wm4.stats_enable = 1; + + brw->wm.state_gs_offset = brw_cache_data( &brw->cache[BRW_WM_UNIT], &wm ); + + if (brw->wm.prog_data->total_scratch) { + /* + dri_emit_reloc(brw->cache[BRW_WM_UNIT].pool->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + (per_thread / 1024) - 1, + brw->wm.state_gs_offset + + ((char *)&wm.thread2 - (char *)&wm), + brw->wm.scratch_buffer); + */ + } else { + wm.thread2.scratch_space_base_pointer = 0; + } +} + +const struct brw_tracked_state brw_wm_unit = { + .dirty = { + .brw = (BRW_NEW_RASTERIZER | + BRW_NEW_ALPHA_TEST | + BRW_NEW_FS | + BRW_NEW_CURBE_OFFSETS), + + .cache = (CACHE_NEW_SURFACE | + CACHE_NEW_WM_PROG | + CACHE_NEW_SAMPLER) + }, + .update = upload_wm_unit +}; + diff --git a/src/gallium/drivers/i965simple/brw_wm_surface_state.c b/src/gallium/drivers/i965simple/brw_wm_surface_state.c new file mode 100644 index 0000000000..d16d919bce --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_wm_surface_state.c @@ -0,0 +1,304 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_context.h" +#include "brw_state.h" +#include "brw_defines.h" + +static unsigned translate_tex_target( enum pipe_texture_target target ) +{ + switch (target) { + case PIPE_TEXTURE_1D: + return BRW_SURFACE_1D; + + case PIPE_TEXTURE_2D: + return BRW_SURFACE_2D; + + case PIPE_TEXTURE_3D: + return BRW_SURFACE_3D; + + case PIPE_TEXTURE_CUBE: + return BRW_SURFACE_CUBE; + + default: + assert(0); + return 0; + } +} + +static unsigned translate_tex_format( enum pipe_format pipe_format ) +{ + switch( pipe_format ) { + case PIPE_FORMAT_U_L8: + return BRW_SURFACEFORMAT_L8_UNORM; + + case PIPE_FORMAT_U_I8: + return BRW_SURFACEFORMAT_I8_UNORM; + + case PIPE_FORMAT_U_A8: + return BRW_SURFACEFORMAT_A8_UNORM; + + case PIPE_FORMAT_U_A8_L8: + return BRW_SURFACEFORMAT_L8A8_UNORM; + + case PIPE_FORMAT_R8G8B8_UNORM: + assert(0); /* not supported for sampling */ + return BRW_SURFACEFORMAT_R8G8B8_UNORM; + + case PIPE_FORMAT_B8G8R8A8_UNORM: + return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + + case PIPE_FORMAT_R8G8B8A8_UNORM: + return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R5G6B5_UNORM: + return BRW_SURFACEFORMAT_B5G6R5_UNORM; + + case PIPE_FORMAT_A1R5G5B5_UNORM: + return BRW_SURFACEFORMAT_B5G5R5A1_UNORM; + + case PIPE_FORMAT_A4R4G4B4_UNORM: + return BRW_SURFACEFORMAT_B4G4R4A4_UNORM; + + case PIPE_FORMAT_YCBCR_REV: + return BRW_SURFACEFORMAT_YCRCB_NORMAL; + + case PIPE_FORMAT_YCBCR: + return BRW_SURFACEFORMAT_YCRCB_SWAPUVY; +#if 0 + case PIPE_FORMAT_RGB_FXT1: + case PIPE_FORMAT_RGBA_FXT1: + return BRW_SURFACEFORMAT_FXT1; +#endif + + case PIPE_FORMAT_Z16_UNORM: + return BRW_SURFACEFORMAT_I16_UNORM; +#if 0 + case PIPE_FORMAT_RGB_DXT1: + return BRW_SURFACEFORMAT_DXT1_RGB; + + case PIPE_FORMAT_RGBA_DXT1: + return BRW_SURFACEFORMAT_BC1_UNORM; + + case PIPE_FORMAT_RGBA_DXT3: + return BRW_SURFACEFORMAT_BC2_UNORM; + + case PIPE_FORMAT_RGBA_DXT5: + return BRW_SURFACEFORMAT_BC3_UNORM; + + case PIPE_FORMAT_SRGBA8: + return BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB; + case PIPE_FORMAT_SRGB_DXT1: + return BRW_SURFACEFORMAT_BC1_UNORM_SRGB; +#endif + + default: + assert(0); + return 0; + } +} + +static unsigned brw_buffer_offset(struct brw_context *brw, + struct pipe_buffer *buffer) +{ + return brw->winsys->get_buffer_offset(brw->winsys, + buffer, + 0); +} + +static +void brw_update_texture_surface( struct brw_context *brw, + unsigned unit ) +{ + const struct brw_texture *tObj = brw->attribs.Texture[unit]; + struct brw_surface_state surf; + + memset(&surf, 0, sizeof(surf)); + + surf.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW; + surf.ss0.surface_type = translate_tex_target(tObj->base.target); + surf.ss0.surface_format = translate_tex_format(tObj->base.format); + + /* This is ok for all textures with channel width 8bit or less: + */ +/* surf.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */ + + /* Updated in emit_reloc */ + surf.ss1.base_addr = brw_buffer_offset( brw, tObj->buffer ); + + surf.ss2.mip_count = tObj->base.last_level; + surf.ss2.width = tObj->base.width[0] - 1; + surf.ss2.height = tObj->base.height[0] - 1; + + surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; + surf.ss3.tiled_surface = 0; /* always zero */ + surf.ss3.pitch = tObj->pitch - 1; + surf.ss3.depth = tObj->base.depth[0] - 1; + + surf.ss4.min_lod = 0; + + if (tObj->base.target == PIPE_TEXTURE_CUBE) { + surf.ss0.cube_pos_x = 1; + surf.ss0.cube_pos_y = 1; + surf.ss0.cube_pos_z = 1; + surf.ss0.cube_neg_x = 1; + surf.ss0.cube_neg_y = 1; + surf.ss0.cube_neg_z = 1; + } + + brw->wm.bind.surf_ss_offset[unit + 1] = + brw_cache_data( &brw->cache[BRW_SS_SURFACE], &surf ); +} + + + +#define OFFSET(TYPE, FIELD) ( (unsigned)&(((TYPE *)0)->FIELD) ) + + +static void upload_wm_surfaces(struct brw_context *brw ) +{ + unsigned i; + + { + struct brw_surface_state surf; + + /* BRW_NEW_FRAMEBUFFER + */ + struct pipe_surface *pipe_surface = brw->attribs.FrameBuffer.cbufs[0];/*fixme*/ + + memset(&surf, 0, sizeof(surf)); + + if (pipe_surface != NULL) { + if (pipe_surface->cpp == 4) + surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + else + surf.ss0.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM; + + surf.ss0.surface_type = BRW_SURFACE_2D; + + surf.ss1.base_addr = brw_buffer_offset( brw, pipe_surface->buffer ); + + surf.ss2.width = pipe_surface->width - 1; + surf.ss2.height = pipe_surface->height - 1; + surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; + surf.ss3.tiled_surface = 0; + surf.ss3.pitch = (pipe_surface->pitch * pipe_surface->cpp) - 1; + } else { + surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; + surf.ss0.surface_type = BRW_SURFACE_NULL; + } + + /* BRW_NEW_BLEND */ + surf.ss0.color_blend = (!brw->attribs.Blend->logicop_enable && + brw->attribs.Blend->blend_enable); + + + surf.ss0.writedisable_red = !(brw->attribs.Blend->colormask & PIPE_MASK_R); + surf.ss0.writedisable_green = !(brw->attribs.Blend->colormask & PIPE_MASK_G); + surf.ss0.writedisable_blue = !(brw->attribs.Blend->colormask & PIPE_MASK_B); + surf.ss0.writedisable_alpha = !(brw->attribs.Blend->colormask & PIPE_MASK_A); + + + + + brw->wm.bind.surf_ss_offset[0] = brw_cache_data( &brw->cache[BRW_SS_SURFACE], &surf ); + + brw->wm.nr_surfaces = 1; + } + + + /* BRW_NEW_TEXTURE + */ + for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { + const struct brw_texture *texUnit = brw->attribs.Texture[i]; + + if (texUnit && + texUnit->base.refcount/*(texUnit->refcount > 0) == really used */) { + + brw_update_texture_surface(brw, i); + + brw->wm.nr_surfaces = i+2; + } + else { + brw->wm.bind.surf_ss_offset[i+1] = 0; + } + } + + brw->wm.bind_ss_offset = brw_cache_data( &brw->cache[BRW_SS_SURF_BIND], + &brw->wm.bind ); +} + + +/* KW: Will find a different way to acheive this, see for example the + * state caches with relocs in the i915 swz driver. + */ +#if 0 +static void emit_reloc_wm_surfaces(struct brw_context *brw) +{ + int unit; + + if (brw->state.draw_region != NULL) { + /* Emit framebuffer relocation */ + dri_emit_reloc(brw_cache_buffer(brw, BRW_SS_SURFACE), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, + brw->wm.bind.surf_ss_offset[0] + + offsetof(struct brw_surface_state, ss1), + brw->state.draw_region->buffer); + } + + /* Emit relocations for texture buffers */ + for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { + struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[unit]; + struct gl_texture_object *tObj = texUnit->_Current; + struct intel_texture_object *intelObj = intel_texture_object(tObj); + + if (texUnit->_ReallyEnabled && intelObj->mt != NULL) { + dri_emit_reloc(brw_cache_buffer(brw, BRW_SS_SURFACE), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + 0, + brw->wm.bind.surf_ss_offset[unit + 1] + + offsetof(struct brw_surface_state, ss1), + intelObj->mt->region->buffer); + } + } +} +#endif + +const struct brw_tracked_state brw_wm_surfaces = { + .dirty = { + .brw = (BRW_NEW_FRAMEBUFFER | + BRW_NEW_BLEND | + BRW_NEW_TEXTURE), + .cache = 0 + }, + .update = upload_wm_surfaces, +}; diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile new file mode 100644 index 0000000000..31438a882e --- /dev/null +++ b/src/gallium/drivers/softpipe/Makefile @@ -0,0 +1,50 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = softpipe + +DRIVER_SOURCES = \ + sp_clear.c \ + sp_flush.c \ + sp_query.c \ + sp_context.c \ + sp_draw_arrays.c \ + sp_prim_setup.c \ + sp_prim_vbuf.c \ + sp_quad.c \ + sp_quad_alpha_test.c \ + sp_quad_blend.c \ + sp_quad_bufloop.c \ + sp_quad_colormask.c \ + sp_quad_coverage.c \ + sp_quad_depth_test.c \ + sp_quad_earlyz.c \ + sp_quad_fs.c \ + sp_quad_occlusion.c \ + sp_quad_output.c \ + sp_quad_stencil.c \ + sp_quad_stipple.c \ + sp_state_blend.c \ + sp_state_clip.c \ + sp_state_derived.c \ + sp_state_fs.c \ + sp_state_sampler.c \ + sp_state_rasterizer.c \ + sp_state_surface.c \ + sp_state_vertex.c \ + sp_texture.c \ + sp_tex_sample.c \ + sp_tile_cache.c \ + sp_surface.c + +C_SOURCES = \ + $(COMMON_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript new file mode 100644 index 0000000000..d581ee8d3c --- /dev/null +++ b/src/gallium/drivers/softpipe/SConscript @@ -0,0 +1,42 @@ +Import('*') + +env = env.Clone() + +softpipe = env.ConvenienceLibrary( + target = 'softpipe', + source = [ + 'sp_clear.c', + 'sp_context.c', + 'sp_draw_arrays.c', + 'sp_flush.c', + 'sp_prim_setup.c', + 'sp_prim_vbuf.c', + 'sp_quad_alpha_test.c', + 'sp_quad_blend.c', + 'sp_quad_bufloop.c', + 'sp_quad.c', + 'sp_quad_colormask.c', + 'sp_quad_coverage.c', + 'sp_quad_depth_test.c', + 'sp_quad_earlyz.c', + 'sp_quad_fs.c', + 'sp_quad_occlusion.c', + 'sp_quad_output.c', + 'sp_quad_stencil.c', + 'sp_quad_stipple.c', + 'sp_query.c', + 'sp_state_blend.c', + 'sp_state_clip.c', + 'sp_state_derived.c', + 'sp_state_fs.c', + 'sp_state_rasterizer.c', + 'sp_state_sampler.c', + 'sp_state_surface.c', + 'sp_state_vertex.c', + 'sp_surface.c', + 'sp_tex_sample.c', + 'sp_texture.c', + 'sp_tile_cache.c', + ]) + +Export('softpipe') \ No newline at end of file diff --git a/src/gallium/drivers/softpipe/sp_clear.c b/src/gallium/drivers/softpipe/sp_clear.c new file mode 100644 index 0000000000..8d295a30ca --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_clear.c @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + */ + + +#include "pipe/p_defines.h" +#include "sp_clear.h" +#include "sp_context.h" +#include "sp_surface.h" +#include "sp_state.h" +#include "sp_tile_cache.h" + + +/** + * Clear the given surface to the specified value. + * No masking, no scissor (clear entire buffer). + */ +void +softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + uint i; + +#if 0 + softpipe_update_derived(softpipe); /* not needed?? */ +#endif + + if (ps == sp_tile_cache_get_surface(softpipe->zsbuf_cache)) { + sp_tile_cache_clear(softpipe->zsbuf_cache, clearValue); +#if TILE_CLEAR_OPTIMIZATION + return; +#endif + } + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[i])) { + sp_tile_cache_clear(softpipe->cbuf_cache[i], clearValue); + } + } + +#if !TILE_CLEAR_OPTIMIZATION + /* non-cached surface */ + pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); +#endif +} diff --git a/src/gallium/drivers/softpipe/sp_clear.h b/src/gallium/drivers/softpipe/sp_clear.h new file mode 100644 index 0000000000..a8ed1c4ecc --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_clear.h @@ -0,0 +1,43 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + */ + +#ifndef SP_CLEAR_H +#define SP_CLEAR_H + +#include "pipe/p_state.h" +struct pipe_context; + +extern void +softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, + unsigned clearValue); + + +#endif /* SP_CLEAR_H */ diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c new file mode 100644 index 0000000000..cea6b90104 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -0,0 +1,333 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#include "pipe/draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "sp_clear.h" +#include "sp_context.h" +#include "sp_flush.h" +#include "sp_prim_setup.h" +#include "sp_prim_vbuf.h" +#include "sp_state.h" +#include "sp_surface.h" +#include "sp_tile_cache.h" +#include "sp_texture.h" +#include "sp_winsys.h" +#include "sp_query.h" + + + +/** + * Query format support for creating a texture, drawing surface, etc. + * \param format the format to test + * \param type one of PIPE_TEXTURE, PIPE_SURFACE + */ +static boolean +softpipe_is_format_supported( struct pipe_context *pipe, + enum pipe_format format, uint type ) +{ + switch (type) { + case PIPE_TEXTURE: + /* softpipe supports all texture formats */ + return TRUE; + case PIPE_SURFACE: + /* softpipe supports all (off-screen) surface formats */ + return TRUE; + default: + assert(0); + return FALSE; + } +} + + +/** + * Map any drawing surfaces which aren't already mapped + */ +void +softpipe_map_surfaces(struct softpipe_context *sp) +{ + unsigned i; + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) { + sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); + } + + sp_tile_cache_map_surfaces(sp->zsbuf_cache); +} + + +/** + * Unmap any mapped drawing surfaces + */ +void +softpipe_unmap_surfaces(struct softpipe_context *sp) +{ + uint i; + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) + sp_flush_tile_cache(sp, sp->cbuf_cache[i]); + sp_flush_tile_cache(sp, sp->zsbuf_cache); + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) { + sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); + } + sp_tile_cache_unmap_surfaces(sp->zsbuf_cache); +} + + +static void softpipe_destroy( struct pipe_context *pipe ) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + struct pipe_winsys *ws = pipe->winsys; + uint i; + + draw_destroy( softpipe->draw ); + + softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); + softpipe->quad.earlyz->destroy( softpipe->quad.earlyz ); + softpipe->quad.shade->destroy( softpipe->quad.shade ); + softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); + softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); + softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); + softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); + softpipe->quad.coverage->destroy( softpipe->quad.coverage ); + softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); + softpipe->quad.blend->destroy( softpipe->quad.blend ); + softpipe->quad.colormask->destroy( softpipe->quad.colormask ); + softpipe->quad.output->destroy( softpipe->quad.output ); + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + sp_destroy_tile_cache(softpipe->cbuf_cache[i]); + sp_destroy_tile_cache(softpipe->zsbuf_cache); + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + sp_destroy_tile_cache(softpipe->tex_cache[i]); + + for (i = 0; i < Elements(softpipe->constants); i++) { + if (softpipe->constants[i].buffer) { + pipe_buffer_reference(ws, &softpipe->constants[i].buffer, NULL); + } + } + + FREE( softpipe ); +} + + +static const char *softpipe_get_name( struct pipe_context *pipe ) +{ + return "softpipe"; +} + +static const char *softpipe_get_vendor( struct pipe_context *pipe ) +{ + return "Tungsten Graphics, Inc."; +} + +static int softpipe_get_param(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 1; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 1; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 1; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 12; /* max 2Kx2K */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 12; /* max 2Kx2K */ + default: + return 0; + } +} + +static float softpipe_get_paramf(struct pipe_context *pipe, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 0.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; /* arbitrary */ + + default: + return 0; + } +} + +struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, + struct softpipe_winsys *softpipe_winsys ) +{ + struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); + uint i; + +#if defined(__i386__) || defined(__386__) + softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; +#else + softpipe->use_sse = FALSE; +#endif + + softpipe->dump_fs = GETENV( "GALLIUM_DUMP_FS" ) != NULL; + + softpipe->pipe.winsys = pipe_winsys; + softpipe->pipe.destroy = softpipe_destroy; + + /* queries */ + softpipe->pipe.is_format_supported = softpipe_is_format_supported; + softpipe->pipe.get_name = softpipe_get_name; + softpipe->pipe.get_vendor = softpipe_get_vendor; + softpipe->pipe.get_param = softpipe_get_param; + softpipe->pipe.get_paramf = softpipe_get_paramf; + + /* state setters */ + softpipe->pipe.create_blend_state = softpipe_create_blend_state; + softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; + softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; + + softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; + softpipe->pipe.bind_sampler_state = softpipe_bind_sampler_state; + softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; + + softpipe->pipe.create_depth_stencil_alpha_state = softpipe_create_depth_stencil_state; + softpipe->pipe.bind_depth_stencil_alpha_state = softpipe_bind_depth_stencil_state; + softpipe->pipe.delete_depth_stencil_alpha_state = softpipe_delete_depth_stencil_state; + + softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; + softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; + softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; + + softpipe->pipe.create_fs_state = softpipe_create_fs_state; + softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; + softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; + + softpipe->pipe.create_vs_state = softpipe_create_vs_state; + softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; + softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; + + softpipe->pipe.set_blend_color = softpipe_set_blend_color; + softpipe->pipe.set_clip_state = softpipe_set_clip_state; + softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; + softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; + softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; + softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; + softpipe->pipe.set_sampler_texture = softpipe_set_sampler_texture; + softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; + + softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; + softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; + + softpipe->pipe.draw_arrays = softpipe_draw_arrays; + softpipe->pipe.draw_elements = softpipe_draw_elements; + + softpipe->pipe.clear = softpipe_clear; + softpipe->pipe.flush = softpipe_flush; + + softpipe_init_query_funcs( softpipe ); + + /* textures */ + softpipe->pipe.texture_create = softpipe_texture_create; + softpipe->pipe.texture_release = softpipe_texture_release; + softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; + + /* + * Alloc caches for accessing drawing surfaces and textures. + * Must be before quad stage setup! + */ + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + softpipe->cbuf_cache[i] = sp_create_tile_cache(); + softpipe->zsbuf_cache = sp_create_tile_cache(); + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + softpipe->tex_cache[i] = sp_create_tile_cache(); + + + /* setup quad rendering stages */ + softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); + softpipe->quad.earlyz = sp_quad_earlyz_stage(softpipe); + softpipe->quad.shade = sp_quad_shade_stage(softpipe); + softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); + softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); + softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); + softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); + softpipe->quad.coverage = sp_quad_coverage_stage(softpipe); + softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); + softpipe->quad.blend = sp_quad_blend_stage(softpipe); + softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); + softpipe->quad.output = sp_quad_output_stage(softpipe); + + softpipe->winsys = softpipe_winsys; + + /* + * Create drawing context and plug our rendering stage into it. + */ + softpipe->draw = draw_create(); + assert(softpipe->draw); + softpipe->setup = sp_draw_render_stage(softpipe); + + if (GETENV( "SP_VBUF" ) != NULL) { + sp_init_vbuf(softpipe); + } + else { + draw_set_rasterize_stage(softpipe->draw, softpipe->setup); + } + + sp_init_surface_functions(softpipe); + + return &softpipe->pipe; +} diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h new file mode 100644 index 0000000000..aff8c2cc5d --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -0,0 +1,152 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef SP_CONTEXT_H +#define SP_CONTEXT_H + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" + +#include "pipe/draw/draw_vertex.h" + +#include "sp_quad.h" + + +struct softpipe_winsys; +struct softpipe_vbuf_render; +struct draw_context; +struct draw_stage; +struct softpipe_tile_cache; +struct sp_fragment_shader_state; +struct sp_vertex_shader_state; + + +struct softpipe_context { + struct pipe_context pipe; /**< base class */ + struct softpipe_winsys *winsys; /**< window system interface */ + + + /* The most recent drawing state as set by the driver: + */ + const struct pipe_blend_state *blend; + const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct pipe_depth_stencil_alpha_state *depth_stencil; + const struct pipe_rasterizer_state *rasterizer; + const struct sp_fragment_shader_state *fs; + const struct sp_vertex_shader_state *vs; + + struct pipe_blend_color blend_color; + struct pipe_clip_state clip; + struct pipe_constant_buffer constants[2]; + struct pipe_framebuffer_state framebuffer; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct softpipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + unsigned dirty; + + /* Counter for occlusion queries. Note this supports overlapping + * queries. + */ + uint64 occlusion_count; + + /* + * Mapped vertex buffers + */ + ubyte *mapped_vbuffer[PIPE_ATTRIB_MAX]; + + /** Mapped constant buffers */ + void *mapped_constants[PIPE_SHADER_TYPES]; + + /** Vertex format */ + struct vertex_info vertex_info; + struct vertex_info vertex_info_vbuf; + + int psize_slot; + +#if 0 + /* Stipple derived state: + */ + ubyte stipple_masks[16][16]; +#endif + + /** Derived from scissor and surface bounds: */ + struct pipe_scissor_state cliprect; + + unsigned line_stipple_counter; + + /** Software quad rendering pipeline */ + struct { + struct quad_stage *polygon_stipple; + struct quad_stage *earlyz; + struct quad_stage *shade; + struct quad_stage *alpha_test; + struct quad_stage *stencil_test; + struct quad_stage *depth_test; + struct quad_stage *occlusion; + struct quad_stage *coverage; + struct quad_stage *bufloop; + struct quad_stage *blend; + struct quad_stage *colormask; + struct quad_stage *output; + + struct quad_stage *first; /**< points to one of the above stages */ + } quad; + + /** The primitive drawing context */ + struct draw_context *draw; + struct draw_stage *setup; + struct draw_stage *vbuf; + struct softpipe_vbuf_render *vbuf_render; + + uint current_cbuf; /**< current color buffer being written to */ + + struct softpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS]; + struct softpipe_tile_cache *zsbuf_cache; + + struct softpipe_tile_cache *tex_cache[PIPE_MAX_SAMPLERS]; + + int use_sse : 1; + int dump_fs : 1; +}; + + + + +static INLINE struct softpipe_context * +softpipe_context( struct pipe_context *pipe ) +{ + return (struct softpipe_context *)pipe; +} + + +#endif /* SP_CONTEXT_H */ diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c new file mode 100644 index 0000000000..71a303a8b5 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -0,0 +1,164 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Brian Paul + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_winsys.h" + +#include "sp_context.h" +#include "sp_state.h" + +#include "pipe/draw/draw_context.h" + + + +static void +softpipe_map_constant_buffers(struct softpipe_context *sp) +{ + struct pipe_winsys *ws = sp->pipe.winsys; + uint i; + for (i = 0; i < 2; i++) { + if (sp->constants[i].size) + sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + } + + draw_set_mapped_constant_buffer(sp->draw, + sp->mapped_constants[PIPE_SHADER_VERTEX]); +} + +static void +softpipe_unmap_constant_buffers(struct softpipe_context *sp) +{ + struct pipe_winsys *ws = sp->pipe.winsys; + uint i; + for (i = 0; i < 2; i++) { + if (sp->constants[i].size) + ws->buffer_unmap(ws, sp->constants[i].buffer); + sp->mapped_constants[i] = NULL; + } +} + + +boolean +softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count) +{ + return softpipe_draw_elements(pipe, NULL, 0, mode, start, count); +} + + + +/** + * Draw vertex arrays, with optional indexing. + * Basically, map the vertex buffers (and drawing surfaces), then hand off + * the drawing to the 'draw' module. + * + * XXX should the element buffer be specified/bound with a separate function? + */ +boolean +softpipe_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + struct softpipe_context *sp = softpipe_context(pipe); + struct draw_context *draw = sp->draw; + unsigned i; + + /* first, check that the primitive is not malformed. It is the + * state tracker's responsibility to do send only correctly formed + * primitives down. It currently isn't doing that though... + */ +#if 1 + count = draw_trim_prim( mode, count ); +#else + if (!draw_validate_prim( mode, count )) + assert(0); +#endif + + + if (sp->dirty) + softpipe_update_derived( sp ); + + softpipe_map_surfaces(sp); + softpipe_map_constant_buffers(sp); + + /* + * Map vertex buffers + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (sp->vertex_buffer[i].buffer) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + sp->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); + } + } + /* Map index buffer, if present */ + if (indexBuffer) { + void *mapped_indexes + = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + } + else { + /* no index/element buffer */ + draw_set_mapped_element_buffer(draw, 0, NULL); + } + + + /* draw! */ + draw_arrays(draw, mode, start, count); + + /* + * unmap vertex/index buffers - will cause draw module to flush + */ + for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + if (sp->vertex_buffer[i].buffer) { + draw_set_mapped_vertex_buffer(draw, i, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); + } + } + if (indexBuffer) { + draw_set_mapped_element_buffer(draw, 0, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); + } + + + /* Note: leave drawing surfaces mapped */ + softpipe_unmap_constant_buffers(sp); + + return TRUE; +} diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c new file mode 100644 index 0000000000..ced0d5d098 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_flush.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + + +#include "pipe/p_defines.h" +#include "pipe/draw/draw_context.h" +#include "sp_flush.h" +#include "sp_context.h" +#include "sp_surface.h" +#include "sp_state.h" +#include "sp_tile_cache.h" +#include "sp_winsys.h" + + +/* There will be actual work to do here. In future we may want a + * fence-like interface instead of finish, and perhaps flush will take + * flags to indicate what type of flush is required. + */ +void +softpipe_flush( struct pipe_context *pipe, + unsigned flags ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + uint i; + + draw_flush(softpipe->draw); + + /* - flush the quad pipeline + * - flush the texture cache + * - flush the render cache + */ + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) + if (softpipe->cbuf_cache[i]) + sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]); + + if (softpipe->zsbuf_cache) + sp_flush_tile_cache(softpipe, softpipe->zsbuf_cache); + + /* Need this call for hardware buffers before swapbuffers. + * + * there should probably be another/different flush-type function + * that's called before swapbuffers because we don't always want + * to unmap surfaces when flushing. + */ + softpipe_unmap_surfaces(softpipe); +} + diff --git a/src/gallium/drivers/softpipe/sp_flush.h b/src/gallium/drivers/softpipe/sp_flush.h new file mode 100644 index 0000000000..34ec617866 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_flush.h @@ -0,0 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_FLUSH_H +#define SP_FLUSH_H + +struct pipe_context; + +void softpipe_flush(struct pipe_context *pipe, unsigned flags ); + +#endif diff --git a/src/gallium/drivers/softpipe/sp_headers.h b/src/gallium/drivers/softpipe/sp_headers.h new file mode 100644 index 0000000000..0ae31d8796 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_headers.h @@ -0,0 +1,82 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef SP_HEADERS_H +#define SP_HEADERS_H + +#include "pipe/tgsi/exec/tgsi_exec.h" + +#define PRIM_POINT 1 +#define PRIM_LINE 2 +#define PRIM_TRI 3 + + +/* The rasterizer generates 2x2 quads of fragment and feeds them to + * the current fp_machine (see below). + * Remember that Y=0=top with Y increasing down the window. + */ +#define QUAD_TOP_LEFT 0 +#define QUAD_TOP_RIGHT 1 +#define QUAD_BOTTOM_LEFT 2 +#define QUAD_BOTTOM_RIGHT 3 + +#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) +#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) +#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) +#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) +#define MASK_ALL 0xf + + +/** + * Encodes everything we need to know about a 2x2 pixel block. Uses + * "Channel-Serial" or "SoA" layout. + */ +struct quad_header { + int x0; + int y0; + unsigned mask:4; + unsigned facing:1; /**< Front (0) or back (1) facing? */ + unsigned prim:2; /**< PRIM_POINT, LINE, TRI */ + + struct { + float color[NUM_CHANNELS][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ + float depth[QUAD_SIZE]; + } outputs; + + float coverage[QUAD_SIZE]; /** fragment coverage for antialiasing */ + + const struct tgsi_interp_coef *coef; + const struct tgsi_interp_coef *posCoef; + + unsigned nr_attrs; +}; + + +#endif /* SP_HEADERS_H */ diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c new file mode 100644 index 0000000000..2772048661 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -0,0 +1,1247 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Primitive rasterization/rendering (points, lines, triangles) + * + * \author Keith Whitwell + * \author Brian Paul + */ + + +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "sp_state.h" +#include "sp_prim_setup.h" +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_vertex.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + +#define DEBUG_VERTS 0 + +/** + * Triangle edge info + */ +struct edge { + float dx; /**< X(v1) - X(v0), used only during setup */ + float dy; /**< Y(v1) - Y(v0), used only during setup */ + float dxdy; /**< dx/dy */ + float sx, sy; /**< first sample point coord */ + int lines; /**< number of lines on this edge */ +}; + + +/** + * Triangle setup info (derived from draw_stage). + * Also used for line drawing (taking some liberties). + */ +struct setup_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct softpipe_context *softpipe; + + /* Vertices are just an array of floats making up each attribute in + * turn. Currently fixed at 4 floats, but should change in time. + * Codegen will help cope with this. + */ + const struct vertex_header *vmax; + const struct vertex_header *vmid; + const struct vertex_header *vmin; + const struct vertex_header *vprovoke; + + struct edge ebot; + struct edge etop; + struct edge emaj; + + float oneoverarea; + + struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; + struct tgsi_interp_coef posCoef; /* For Z, W */ + struct quad_header quad; + + struct { + int left[2]; /**< [0] = row0, [1] = row1 */ + int right[2]; + int y; + unsigned y_flags; + unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ + } span; +}; + + + +/** + * Basically a cast wrapper. + */ +static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) +{ + return (struct setup_stage *)stage; +} + + +/** + * Clip setup->quad against the scissor/surface bounds. + */ +static INLINE void +quad_clip(struct setup_stage *setup) +{ + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + + if (setup->quad.x0 >= maxx || + setup->quad.y0 >= maxy || + setup->quad.x0 + 1 < minx || + setup->quad.y0 + 1 < miny) { + /* totally clipped */ + setup->quad.mask = 0x0; + return; + } + if (setup->quad.x0 < minx) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + if (setup->quad.y0 < miny) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + if (setup->quad.x0 == maxx - 1) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + if (setup->quad.y0 == maxy - 1) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); +} + + +/** + * Emit a quad (pass to next stage) with clipping. + */ +static INLINE void +clip_emit_quad(struct setup_stage *setup) +{ + quad_clip(setup); + if (setup->quad.mask) { + struct softpipe_context *sp = setup->softpipe; + sp->quad.first->run(sp->quad.first, &setup->quad); + } +} + + +/** + * Emit a quad (pass to next stage). No clipping is done. + */ +static INLINE void +emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) +{ + struct softpipe_context *sp = setup->softpipe; + setup->quad.x0 = x; + setup->quad.y0 = y; + setup->quad.mask = mask; + sp->quad.first->run(sp->quad.first, &setup->quad); +} + + +/** + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. + */ +static INLINE int block( int x ) +{ + return x & ~1; +} + + +/** + * Compute mask which indicates which pixels in the 2x2 quad are actually inside + * the triangle's bounds. + * + * this is pretty nasty... may need to rework flush_spans again to + * fix it, if possible. + */ +static unsigned calculate_mask( struct setup_stage *setup, int x ) +{ + unsigned mask = 0x0; + + if (x >= setup->span.left[0] && x < setup->span.right[0]) + mask |= MASK_TOP_LEFT; + + if (x >= setup->span.left[1] && x < setup->span.right[1]) + mask |= MASK_BOTTOM_LEFT; + + if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) + mask |= MASK_TOP_RIGHT; + + if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) + mask |= MASK_BOTTOM_RIGHT; + + return mask; +} + + +/** + * Render a horizontal span of quads + */ +static void flush_spans( struct setup_stage *setup ) +{ + int minleft, maxright; + int x; + + switch (setup->span.y_flags) { + case 0x3: + /* both odd and even lines written (both quad rows) */ + minleft = MIN2(setup->span.left[0], setup->span.left[1]); + maxright = MAX2(setup->span.right[0], setup->span.right[1]); + break; + + case 0x1: + /* only even line written (quad top row) */ + minleft = setup->span.left[0]; + maxright = setup->span.right[0]; + break; + + case 0x2: + /* only odd line written (quad bottom row) */ + minleft = setup->span.left[1]; + maxright = setup->span.right[1]; + break; + + default: + return; + } + + /* XXX this loop could be moved into the above switch cases and + * calculate_mask() could be simplified a bit... + */ + for (x = block(minleft); x <= block(maxright); x += 2) { + emit_quad( setup, x, setup->span.y, + calculate_mask( setup, x ) ); + } + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +} + +#if DEBUG_VERTS +static void print_vertex(const struct setup_stage *setup, + const struct vertex_header *v) +{ + int i; + debug_printf("Vertex: (%p)\n", v); + for (i = 0; i < setup->quad.nr_attrs; i++) { + debug_printf(" %d: %f %f %f %f\n", i, + v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); + } +} +#endif + +static boolean setup_sort_vertices( struct setup_stage *setup, + const struct prim_header *prim ) +{ + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + const struct vertex_header *v2 = prim->v[2]; + +#if DEBUG_VERTS + debug_printf("Triangle:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); + print_vertex(setup, v2); +#endif + + setup->vprovoke = v2; + + /* determine bottom to top order of vertices */ + { + float y0 = v0->data[0][1]; + float y1 = v1->data[0][1]; + float y2 = v2->data[0][1]; + if (y0 <= y1) { + if (y1 <= y2) { + /* y0<=y1<=y2 */ + setup->vmin = v0; + setup->vmid = v1; + setup->vmax = v2; + } + else if (y2 <= y0) { + /* y2<=y0<=y1 */ + setup->vmin = v2; + setup->vmid = v0; + setup->vmax = v1; + } + else { + /* y0<=y2<=y1 */ + setup->vmin = v0; + setup->vmid = v2; + setup->vmax = v1; + } + } + else { + if (y0 <= y2) { + /* y1<=y0<=y2 */ + setup->vmin = v1; + setup->vmid = v0; + setup->vmax = v2; + } + else if (y2 <= y1) { + /* y2<=y1<=y0 */ + setup->vmin = v2; + setup->vmid = v1; + setup->vmax = v0; + } + else { + /* y1<=y2<=y0 */ + setup->vmin = v1; + setup->vmid = v2; + setup->vmax = v0; + } + } + } + + setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; + setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; + setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; + + /* + * Compute triangle's area. Use 1/area to compute partial + * derivatives of attributes later. + * + * The area will be the same as prim->det, but the sign may be + * different depending on how the vertices get sorted above. + * + * To determine whether the primitive is front or back facing we + * use the prim->det value because its sign is correct. + */ + { + const float area = (setup->emaj.dx * setup->ebot.dy - + setup->ebot.dx * setup->emaj.dy); + + setup->oneoverarea = 1.0f / area; + /* + debug_printf("%s one-over-area %f area %f det %f\n", + __FUNCTION__, setup->oneoverarea, area, prim->det ); + */ + } + + /* We need to know if this is a front or back-facing triangle for: + * - the GLSL gl_FrontFacing fragment attribute (bool) + * - two-sided stencil test + */ + setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); + + return TRUE; +} + + +/** + * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + * The value value comes from vertex->data[slot][i]. + * The result will be put into setup->coef[slot].a0[i]. + * \param slot which attribute slot + * \param i which component of the slot (0..3) + */ +static void const_coeff( struct setup_stage *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + assert(i <= 3); + + coef->dadx[i] = 0; + coef->dady[i] = 0; + + /* need provoking vertex info! + */ + coef->a0[i] = setup->vprovoke->data[vertSlot][i]; +} + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a triangle. + */ +static void tri_linear_coeff( struct setup_stage *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + float botda = setup->vmid->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; + float majda = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float dadx = a * setup->oneoverarea; + float dady = b * setup->oneoverarea; + + assert(i <= 3); + + coef->dadx[i] = dadx; + coef->dady[i] = dady; + + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + coef->a0[i] = (setup->vmin->data[vertSlot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); + + /* + debug_printf("attr[%d].%c: %f dx:%f dy:%f\n", + slot, "xyzw"[i], + setup->coef[slot].a0[i], + setup->coef[slot].dadx[i], + setup->coef[slot].dady[i]); + */ +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a triangle. + * We basically multiply the vertex value by 1/w before computing + * the plane coefficients (a0, dadx, dady). + * Later, when we compute the value at a particular fragment position we'll + * divide the interpolated value by the interpolated W at that fragment. + */ +static void tri_persp_coeff( struct setup_stage *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + /* premultiply by 1/w (v->data[0][3] is always W): + */ + float mina = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; + float mida = setup->vmid->data[vertSlot][i] * setup->vmid->data[0][3]; + float maxa = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; + float botda = mida - mina; + float majda = maxa - mina; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float dadx = a * setup->oneoverarea; + float dady = b * setup->oneoverarea; + + /* + debug_printf("tri persp %d,%d: %f %f %f\n", vertSlot, i, + setup->vmin->data[vertSlot][i], + setup->vmid->data[vertSlot][i], + setup->vmax->data[vertSlot][i] + ); + */ + assert(i <= 3); + + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (mina - + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); +} + + +/** + * Special coefficient setup for gl_FragCoord. + * X and Y are trivial, though Y has to be inverted for OpenGL. + * Z and W are copied from posCoef which should have already been computed. + * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask. + */ +static void +setup_fragcoord_coeff(struct setup_stage *setup) +{ + /*X*/ + setup->coef[0].a0[0] = 0; + setup->coef[0].dadx[0] = 1.0; + setup->coef[0].dady[0] = 0.0; + /*Y*/ + if (setup->softpipe->rasterizer->origin_lower_left) { + /* y=0=bottom */ + const int winHeight = setup->softpipe->framebuffer.cbufs[0]->height; + setup->coef[0].a0[1] = (float) (winHeight - 1); + setup->coef[0].dady[1] = -1.0; + } + else { + /* y=0=top */ + setup->coef[0].a0[1] = 0.0; + setup->coef[0].dady[1] = 1.0; + } + setup->coef[0].dadx[1] = 0.0; + /*Z*/ + setup->coef[0].a0[2] = setup->posCoef.a0[2]; + setup->coef[0].dadx[2] = setup->posCoef.dadx[2]; + setup->coef[0].dady[2] = setup->posCoef.dady[2]; + /*W*/ + setup->coef[0].a0[3] = setup->posCoef.a0[3]; + setup->coef[0].dadx[3] = setup->posCoef.dadx[3]; + setup->coef[0].dady[3] = setup->posCoef.dady[3]; +} + + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. + */ +static void setup_tri_coefficients( struct setup_stage *setup ) +{ + struct softpipe_context *softpipe = setup->softpipe; + const struct pipe_shader_state *fs = &softpipe->fs->shader; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); + uint fragSlot; + + /* z and w are done by linear interpolation: + */ + tri_linear_coeff(setup, &setup->posCoef, 0, 2); + tri_linear_coeff(setup, &setup->posCoef, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { + const uint vertSlot = vinfo->src_index[fragSlot]; + uint j; + + switch (vinfo->interp_mode[fragSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + assert(fragSlot == 0); + setup_fragcoord_coeff(setup); + break; + default: + assert(0); + } + + if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } + } +} + + + +static void setup_tri_edges( struct setup_stage *setup ) +{ + float vmin_x = setup->vmin->data[0][0] + 0.5f; + float vmid_x = setup->vmid->data[0][0] + 0.5f; + + float vmin_y = setup->vmin->data[0][1] - 0.5f; + float vmid_y = setup->vmid->data[0][1] - 0.5f; + float vmax_y = setup->vmax->data[0][1] - 0.5f; + + setup->emaj.sy = CEILF(vmin_y); + setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); + setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; + setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; + + setup->etop.sy = CEILF(vmid_y); + setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); + setup->etop.dxdy = setup->etop.dx / setup->etop.dy; + setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; + + setup->ebot.sy = CEILF(vmin_y); + setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); + setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; + setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; +} + + +/** + * Render the upper or lower half of a triangle. + * Scissoring/cliprect is applied here too. + */ +static void subtriangle( struct setup_stage *setup, + struct edge *eleft, + struct edge *eright, + unsigned lines ) +{ + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + int y, start_y, finish_y; + int sy = (int)eleft->sy; + + assert((int)eleft->sy == (int) eright->sy); + + /* clip top/bottom */ + start_y = sy; + finish_y = sy + lines; + + if (start_y < miny) + start_y = miny; + + if (finish_y > maxy) + finish_y = maxy; + + start_y -= sy; + finish_y -= sy; + + /* + debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + */ + + for (y = start_y; y < finish_y; y++) { + + /* avoid accumulating adds as floats don't have the precision to + * accurately iterate large triangle edges that way. luckily we + * can just multiply these days. + * + * this is all drowned out by the attribute interpolation anyway. + */ + int left = (int)(eleft->sx + y * eleft->dxdy); + int right = (int)(eright->sx + y * eright->dxdy); + + /* clip left/right */ + if (left < minx) + left = minx; + if (right > maxx) + right = maxx; + + if (left < right) { + int _y = sy + y; + if (block(_y) != setup->span.y) { + flush_spans(setup); + setup->span.y = block(_y); + } + + setup->span.left[_y&1] = left; + setup->span.right[_y&1] = right; + setup->span.y_flags |= 1<<(_y&1); + } + } + + + /* save the values so that emaj can be restarted: + */ + eleft->sx += lines * eleft->dxdy; + eright->sx += lines * eright->dxdy; + eleft->sy += lines; + eright->sy += lines; +} + + +/** + * Do setup for triangle rasterization, then render the triangle. + */ +static void setup_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct setup_stage *setup = setup_stage( stage ); + + /* + debug_printf("%s\n", __FUNCTION__ ); + */ + + setup_sort_vertices( setup, prim ); + setup_tri_coefficients( setup ); + setup_tri_edges( setup ); + + setup->quad.prim = PRIM_TRI; + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; + /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ + + /* init_constant_attribs( setup ); */ + + if (setup->oneoverarea < 0.0) { + /* emaj on left: + */ + subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); + subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); + } + else { + /* emaj on right: + */ + subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); + subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); + } + + flush_spans( setup ); +} + + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a line. + */ +static void +line_linear_coeff(struct setup_stage *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + const float da = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; + const float dadx = da * setup->emaj.dx * setup->oneoverarea; + const float dady = da * setup->emaj.dy * setup->oneoverarea; + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (setup->vmin->data[vertSlot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a line. + */ +static void +line_persp_coeff(struct setup_stage *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + /* XXX double-check/verify this arithmetic */ + const float a0 = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; + const float a1 = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; + const float da = a1 - a0; + const float dadx = da * setup->emaj.dx * setup->oneoverarea; + const float dady = da * setup->emaj.dy * setup->oneoverarea; + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (setup->vmin->data[vertSlot][i] - + (dadx * (setup->vmin->data[0][0] - 0.5f) + + dady * (setup->vmin->data[0][1] - 0.5f))); +} + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmax are initialized. + */ +static INLINE void +setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) +{ + struct softpipe_context *softpipe = setup->softpipe; + const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); + uint fragSlot; + + /* use setup->vmin, vmax to point to vertices */ + setup->vprovoke = prim->v[1]; + setup->vmin = prim->v[0]; + setup->vmax = prim->v[1]; + + setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; + setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; + /* NOTE: this is not really 1/area */ + setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx + + setup->emaj.dy * setup->emaj.dy); + + /* z and w are done by linear interpolation: + */ + line_linear_coeff(setup, &setup->posCoef, 0, 2); + line_linear_coeff(setup, &setup->posCoef, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { + const uint vertSlot = vinfo->src_index[fragSlot]; + uint j; + + switch (vinfo->interp_mode[fragSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + assert(fragSlot == 0); + assert(0); /* XXX fix this: */ + setup_fragcoord_coeff(setup); + break; + default: + assert(0); + } + + if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } + } +} + + +/** + * Plot a pixel in a line segment. + */ +static INLINE void +plot(struct setup_stage *setup, int x, int y) +{ + const int iy = y & 1; + const int ix = x & 1; + const int quadX = x - ix; + const int quadY = y - iy; + const int mask = (1 << ix) << (2 * iy); + + if (quadX != setup->quad.x0 || + quadY != setup->quad.y0) + { + /* flush prev quad, start new quad */ + + if (setup->quad.x0 != -1) + clip_emit_quad(setup); + + setup->quad.x0 = quadX; + setup->quad.y0 = quadY; + setup->quad.mask = 0x0; + } + + setup->quad.mask |= mask; +} + + +/** + * Do setup for line rasterization, then render the line. + * Single-pixel width, no stipple, etc. We rely on the 'draw' module + * to handle stippling and wide lines. + */ +static void +setup_line(struct draw_stage *stage, struct prim_header *prim) +{ + const struct vertex_header *v0 = prim->v[0]; + const struct vertex_header *v1 = prim->v[1]; + struct setup_stage *setup = setup_stage( stage ); + int x0 = (int) v0->data[0][0]; + int x1 = (int) v1->data[0][0]; + int y0 = (int) v0->data[0][1]; + int y1 = (int) v1->data[0][1]; + int dx = x1 - x0; + int dy = y1 - y0; + int xstep, ystep; + + if (dx == 0 && dy == 0) + return; + + setup_line_coefficients(setup, prim); + + if (dx < 0) { + dx = -dx; /* make positive */ + xstep = -1; + } + else { + xstep = 1; + } + + if (dy < 0) { + dy = -dy; /* make positive */ + ystep = -1; + } + else { + ystep = 1; + } + + assert(dx >= 0); + assert(dy >= 0); + + setup->quad.x0 = setup->quad.y0 = -1; + setup->quad.mask = 0x0; + setup->quad.prim = PRIM_LINE; + /* XXX temporary: set coverage to 1.0 so the line appears + * if AA mode happens to be enabled. + */ + setup->quad.coverage[0] = + setup->quad.coverage[1] = + setup->quad.coverage[2] = + setup->quad.coverage[3] = 1.0; + + if (dx > dy) { + /*** X-major line ***/ + int i; + const int errorInc = dy + dy; + int error = errorInc - dx; + const int errorDec = error - dx; + + for (i = 0; i < dx; i++) { + plot(setup, x0, y0); + + x0 += xstep; + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + y0 += ystep; + } + } + } + else { + /*** Y-major line ***/ + int i; + const int errorInc = dx + dx; + int error = errorInc - dy; + const int errorDec = error - dy; + + for (i = 0; i < dy; i++) { + plot(setup, x0, y0); + + y0 += ystep; + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + x0 += xstep; + } + } + } + + /* draw final quad */ + if (setup->quad.mask) { + clip_emit_quad(setup); + } +} + + +static void +point_persp_coeff(struct setup_stage *setup, + const struct vertex_header *vert, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + assert(i <= 3); + coef->dadx[i] = 0.0F; + coef->dady[i] = 0.0F; + coef->a0[i] = vert->data[vertSlot][i] * vert->data[0][3]; +} + + +/** + * Do setup for point rasterization, then render the point. + * Round or square points... + * XXX could optimize a lot for 1-pixel points. + */ +static void +setup_point(struct draw_stage *stage, struct prim_header *prim) +{ + struct setup_stage *setup = setup_stage( stage ); + struct softpipe_context *softpipe = setup->softpipe; + const struct pipe_shader_state *fs = &softpipe->fs->shader; + const struct vertex_header *v0 = prim->v[0]; + const int sizeAttr = setup->softpipe->psize_slot; + const float size + = sizeAttr > 0 ? v0->data[sizeAttr][0] + : setup->softpipe->rasterizer->point_size; + const float halfSize = 0.5F * size; + const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth; + const float x = v0->data[0][0]; /* Note: data[0] is always position */ + const float y = v0->data[0][1]; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); + uint fragSlot; + + /* For points, all interpolants are constant-valued. + * However, for point sprites, we'll need to setup texcoords appropriately. + * XXX: which coefficients are the texcoords??? + * We may do point sprites as textured quads... + * + * KW: We don't know which coefficients are texcoords - ultimately + * the choice of what interpolation mode to use for each attribute + * should be determined by the fragment program, using + * per-attribute declaration statements that include interpolation + * mode as a parameter. So either the fragment program will have + * to be adjusted for pointsprite vs normal point behaviour, or + * otherwise a special interpolation mode will have to be defined + * which matches the required behaviour for point sprites. But - + * the latter is not a feature of normal hardware, and as such + * probably should be ruled out on that basis. + */ + setup->vprovoke = prim->v[0]; + + /* setup Z, W */ + const_coeff(setup, &setup->posCoef, 0, 2); + const_coeff(setup, &setup->posCoef, 0, 3); + + for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { + const uint vertSlot = vinfo->src_index[fragSlot]; + uint j; + + switch (vinfo->interp_mode[fragSlot]) { + case INTERP_CONSTANT: + /* fall-through */ + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + point_persp_coeff(setup, setup->vprovoke, + &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + assert(fragSlot == 0); + assert(0); /* XXX fix this: */ + setup_fragcoord_coeff(setup); + break; + default: + assert(0); + } + + if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } + } + + setup->quad.prim = PRIM_POINT; + + if (halfSize <= 0.5 && !round) { + /* special case for 1-pixel points */ + const int ix = ((int) x) & 1; + const int iy = ((int) y) & 1; + setup->quad.x0 = (int) x - ix; + setup->quad.y0 = (int) y - iy; + setup->quad.mask = (1 << ix) << (2 * iy); + clip_emit_quad(setup); + } + else { + if (round) { + /* rounded points */ + const int ixmin = block((int) (x - halfSize)); + const int ixmax = block((int) (x + halfSize)); + const int iymin = block((int) (y - halfSize)); + const int iymax = block((int) (y + halfSize)); + const float rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ + const float rmax = halfSize + 0.7071F; + const float rmin2 = MAX2(0.0F, rmin * rmin); + const float rmax2 = rmax * rmax; + const float cscale = 1.0F / (rmax2 - rmin2); + int ix, iy; + + for (iy = iymin; iy <= iymax; iy += 2) { + for (ix = ixmin; ix <= ixmax; ix += 2) { + float dx, dy, dist2, cover; + + setup->quad.mask = 0x0; + + dx = (ix + 0.5f) - x; + dy = (iy + 0.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_TOP_LEFT; + } + + dx = (ix + 1.5f) - x; + dy = (iy + 0.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_TOP_RIGHT; + } + + dx = (ix + 0.5f) - x; + dy = (iy + 1.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_BOTTOM_LEFT; + } + + dx = (ix + 1.5f) - x; + dy = (iy + 1.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_BOTTOM_RIGHT; + } + + if (setup->quad.mask) { + setup->quad.x0 = ix; + setup->quad.y0 = iy; + clip_emit_quad(setup); + } + } + } + } + else { + /* square points */ + const int xmin = (int) (x + 0.75 - halfSize); + const int ymin = (int) (y + 0.25 - halfSize); + const int xmax = xmin + (int) size; + const int ymax = ymin + (int) size; + /* XXX could apply scissor to xmin,ymin,xmax,ymax now */ + const int ixmin = block(xmin); + const int ixmax = block(xmax - 1); + const int iymin = block(ymin); + const int iymax = block(ymax - 1); + int ix, iy; + + /* + debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax); + */ + for (iy = iymin; iy <= iymax; iy += 2) { + uint rowMask = 0xf; + if (iy < ymin) { + /* above the top edge */ + rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + } + if (iy + 1 >= ymax) { + /* below the bottom edge */ + rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + } + + for (ix = ixmin; ix <= ixmax; ix += 2) { + uint mask = rowMask; + + if (ix < xmin) { + /* fragment is past left edge of point, turn off left bits */ + mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + } + if (ix + 1 >= xmax) { + /* past the right edge */ + mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + } + + setup->quad.mask = mask; + setup->quad.x0 = ix; + setup->quad.y0 = iy; + clip_emit_quad(setup); + } + } + } + } +} + + + +static void setup_begin( struct draw_stage *stage ) +{ + struct setup_stage *setup = setup_stage(stage); + struct softpipe_context *sp = setup->softpipe; + const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; + + setup->quad.nr_attrs = fs->num_inputs; + + sp->quad.first->begin(sp->quad.first); + + stage->point = setup_point; + stage->line = setup_line; + stage->tri = setup_tri; +} + + +static void setup_first_point( struct draw_stage *stage, + struct prim_header *header ) +{ + setup_begin(stage); + stage->point( stage, header ); +} + +static void setup_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + setup_begin(stage); + stage->line( stage, header ); +} + + +static void setup_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + setup_begin(stage); + stage->tri( stage, header ); +} + + + +static void setup_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->point = setup_first_point; + stage->line = setup_first_line; + stage->tri = setup_first_tri; +} + + +static void reset_stipple_counter( struct draw_stage *stage ) +{ +} + + +static void render_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + +/** + * Create a new primitive setup/render stage. + */ +struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) +{ + struct setup_stage *setup = CALLOC_STRUCT(setup_stage); + + setup->softpipe = softpipe; + setup->stage.draw = softpipe->draw; + setup->stage.point = setup_first_point; + setup->stage.line = setup_first_line; + setup->stage.tri = setup_first_tri; + setup->stage.flush = setup_flush; + setup->stage.reset_stipple_counter = reset_stipple_counter; + setup->stage.destroy = render_destroy; + + setup->quad.coef = setup->coef; + setup->quad.posCoef = &setup->posCoef; + + return &setup->stage; +} diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.h b/src/gallium/drivers/softpipe/sp_prim_setup.h new file mode 100644 index 0000000000..f3e8a79dd9 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_prim_setup.h @@ -0,0 +1,79 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SP_PRIM_SETUP_H +#define SP_PRIM_SETUP_H + + +/** + * vbuf is a special stage to gather the stream of triangles, lines, points + * together and reconstruct vertex buffers for hardware upload. + * + * First attempt, work in progress. + * + * TODO: + * - separate out vertex buffer building and primitive emit, ie >1 draw per vb. + * - tell vbuf stage how to build hw vertices directly + * - pass vbuf stage a buffer pointer for direct emit to agp/vram. + * + * + * + * Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + + +struct draw_stage; +struct softpipe_context; + + +typedef void (*vbuf_draw_func)( struct pipe_context *pipe, + unsigned prim, + const ushort *elements, + unsigned nr_elements, + const void *vertex_buffer, + unsigned nr_vertices ); + + +extern struct draw_stage * +sp_draw_render_stage( struct softpipe_context *softpipe ); + + +extern struct draw_stage * +sp_draw_vbuf_stage( struct draw_context *draw_context, + struct pipe_context *pipe, + vbuf_draw_func draw ); + + +#endif /* SP_PRIM_SETUP_H */ diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c new file mode 100644 index 0000000000..7f71fdb6a9 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -0,0 +1,221 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Post-transform vertex buffering. This is an optional part of the + * softpipe rendering pipeline. + * Probably not desired in general, but useful for testing/debuggin. + * Enabled/Disabled with SP_VBUF env var. + * + * Authors + * Brian Paul + */ + + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_prim_vbuf.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_private.h" +#include "pipe/draw/draw_vbuf.h" + + +#define SP_MAX_VBUF_INDEXES 1024 +#define SP_MAX_VBUF_SIZE 4096 + + +/** + * Subclass of vbuf_render. + */ +struct softpipe_vbuf_render +{ + struct vbuf_render base; + struct softpipe_context *softpipe; + uint prim; + uint vertex_size; + void *vertex_buffer; +}; + + +/** cast wrapper */ +static struct softpipe_vbuf_render * +softpipe_vbuf_render(struct vbuf_render *vbr) +{ + return (struct softpipe_vbuf_render *) vbr; +} + + + +static const struct vertex_info * +sp_vbuf_get_vertex_info(struct vbuf_render *vbr) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + return softpipe_get_vbuf_vertex_info(cvbr->softpipe); +} + + +static void * +sp_vbuf_allocate_vertices(struct vbuf_render *vbr, + ushort vertex_size, ushort nr_vertices) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + assert(!cvbr->vertex_buffer); + cvbr->vertex_buffer = align_malloc(vertex_size * nr_vertices, 16); + cvbr->vertex_size = vertex_size; + return cvbr->vertex_buffer; +} + + +static void +sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, + unsigned vertex_size, unsigned vertices_used) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + align_free(vertices); + assert(vertices == cvbr->vertex_buffer); + cvbr->vertex_buffer = NULL; +} + + +static void +sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + cvbr->prim = prim; +} + + +/** + * Recalculate prim's determinant. + * XXX is this needed? + */ +static void +calc_det(struct prim_header *header) +{ + /* Window coords: */ + const float *v0 = header->v[0]->data[0]; + const float *v1 = header->v[1]->data[0]; + const float *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0] - v2[0]; + const float ey = v0[1] - v2[1]; + const float fx = v1[0] - v2[0]; + const float fy = v1[1] - v2[1]; + + /* det = cross(e,f).z */ + header->det = ex * fy - ey * fx; +} + + +static void +sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + struct softpipe_context *softpipe = cvbr->softpipe; + struct draw_stage *setup = softpipe->setup; + struct prim_header prim; + unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); + unsigned i, j; + void *vertex_buffer = cvbr->vertex_buffer; + + prim.det = 0; + prim.reset_line_stipple = 0; + prim.edgeflags = 0; + prim.pad = 0; + + switch (cvbr->prim) { + case PIPE_PRIM_TRIANGLES: + for (i = 0; i < nr_indices; i += 3) { + for (j = 0; j < 3; j++) + prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + + indices[i+j] * vertex_size); + + calc_det(&prim); + setup->tri( setup, &prim ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i < nr_indices; i += 2) { + for (j = 0; j < 2; j++) + prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + + indices[i+j] * vertex_size); + + setup->line( setup, &prim ); + } + break; + + case PIPE_PRIM_POINTS: + for (i = 0; i < nr_indices; i++) { + prim.v[0] = (struct vertex_header *)((char *)vertex_buffer + + indices[i] * vertex_size); + setup->point( setup, &prim ); + } + break; + } + + setup->flush( setup, 0 ); +} + + +static void +sp_vbuf_destroy(struct vbuf_render *vbr) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + cvbr->softpipe->vbuf_render = NULL; + FREE(cvbr); +} + + +/** + * Initialize the post-transform vertex buffer information for the given + * context. + */ +void +sp_init_vbuf(struct softpipe_context *sp) +{ + assert(sp->draw); + + sp->vbuf_render = CALLOC_STRUCT(softpipe_vbuf_render); + + sp->vbuf_render->base.max_indices = SP_MAX_VBUF_INDEXES; + sp->vbuf_render->base.max_vertex_buffer_bytes = SP_MAX_VBUF_SIZE; + + sp->vbuf_render->base.get_vertex_info = sp_vbuf_get_vertex_info; + sp->vbuf_render->base.allocate_vertices = sp_vbuf_allocate_vertices; + sp->vbuf_render->base.set_primitive = sp_vbuf_set_primitive; + sp->vbuf_render->base.draw = sp_vbuf_draw; + sp->vbuf_render->base.release_vertices = sp_vbuf_release_vertices; + sp->vbuf_render->base.destroy = sp_vbuf_destroy; + + sp->vbuf_render->softpipe = sp; + + sp->vbuf = draw_vbuf_stage(sp->draw, &sp->vbuf_render->base); + + draw_set_rasterize_stage(sp->draw, sp->vbuf); +} diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.h b/src/gallium/drivers/softpipe/sp_prim_vbuf.h new file mode 100644 index 0000000000..1de9cc2a89 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.h @@ -0,0 +1,38 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_VBUF_H +#define SP_VBUF_H + + +struct softpipe_context; + +extern void +sp_init_vbuf(struct softpipe_context *softpipe); + + +#endif /* SP_VBUF_H */ diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad.c new file mode 100644 index 0000000000..6bd468a51c --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad.c @@ -0,0 +1,118 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "sp_context.h" +#include "sp_state.h" +#include "pipe/p_shader_tokens.h" + +static void +sp_push_quad_first( + struct softpipe_context *sp, + struct quad_stage *quad ) +{ + quad->next = sp->quad.first; + sp->quad.first = quad; +} + +static void +sp_build_depth_stencil( + struct softpipe_context *sp ) +{ + if (sp->depth_stencil->stencil[0].enabled || + sp->depth_stencil->stencil[1].enabled) { + sp_push_quad_first( sp, sp->quad.stencil_test ); + } + else if (sp->depth_stencil->depth.enabled && + sp->framebuffer.zsbuf) { + sp_push_quad_first( sp, sp->quad.depth_test ); + } +} + +void +sp_build_quad_pipeline(struct softpipe_context *sp) +{ + boolean early_depth_test = + sp->depth_stencil->depth.enabled && + sp->framebuffer.zsbuf && + !sp->depth_stencil->alpha.enabled && + sp->fs->shader.output_semantic_name[0] != TGSI_SEMANTIC_POSITION; + + /* build up the pipeline in reverse order... */ + + sp->quad.first = sp->quad.output; + + if (sp->blend->colormask != 0xf) { + sp_push_quad_first( sp, sp->quad.colormask ); + } + + if (sp->blend->blend_enable || + sp->blend->logicop_enable) { + sp_push_quad_first( sp, sp->quad.blend ); + } + + if (sp->framebuffer.num_cbufs == 1) { + /* the usual case: write to exactly one colorbuf */ + sp->current_cbuf = 0; + } + else { + /* insert bufloop stage */ + sp_push_quad_first( sp, sp->quad.bufloop ); + } + + if (sp->depth_stencil->depth.occlusion_count) { + sp_push_quad_first( sp, sp->quad.occlusion ); + } + + if (sp->rasterizer->poly_smooth || + sp->rasterizer->line_smooth || + sp->rasterizer->point_smooth) { + sp_push_quad_first( sp, sp->quad.coverage ); + } + + if (!early_depth_test) { + sp_build_depth_stencil( sp ); + } + + if (sp->depth_stencil->alpha.enabled) { + sp_push_quad_first( sp, sp->quad.alpha_test ); + } + + /* XXX always enable shader? */ + if (1) { + sp_push_quad_first( sp, sp->quad.shade ); + } + + if (early_depth_test) { + sp_build_depth_stencil( sp ); + sp_push_quad_first( sp, sp->quad.earlyz ); + } + + if (sp->rasterizer->poly_stipple_enable) { + sp_push_quad_first( sp, sp->quad.polygon_stipple ); + } +} diff --git a/src/gallium/drivers/softpipe/sp_quad.h b/src/gallium/drivers/softpipe/sp_quad.h new file mode 100644 index 0000000000..f1e0281764 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad.h @@ -0,0 +1,70 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef SP_QUAD_H +#define SP_QUAD_H + + +struct softpipe_context; +struct quad_header; + + +struct quad_stage { + struct softpipe_context *softpipe; + + struct quad_stage *next; + + void (*begin)(struct quad_stage *qs); + + /** the stage action */ + void (*run)(struct quad_stage *qs, struct quad_header *quad); + + void (*destroy)(struct quad_stage *qs); +}; + + +struct quad_stage *sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_earlyz_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); +struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); + +void sp_build_quad_pipeline(struct softpipe_context *sp); + +void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad); + +#endif /* SP_QUAD_H */ diff --git a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c new file mode 100644 index 0000000000..4ffeac35e1 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c @@ -0,0 +1,108 @@ + +/** + * quad alpha test + */ + +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" + + +static void +alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + const float ref = softpipe->depth_stencil->alpha.ref; + unsigned passMask = 0x0, j; + const float *aaaa = quad->outputs.color[3]; + + switch (softpipe->depth_stencil->alpha.func) { + case PIPE_FUNC_NEVER: + quad->mask = 0x0; + break; + case PIPE_FUNC_LESS: + /* + * If mask were an array [4] we could do this SIMD-style: + * passMask = (quad->outputs.color[3] <= vec4(ref)); + */ + for (j = 0; j < QUAD_SIZE; j++) { + if (aaaa[j] < ref) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_EQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (aaaa[j] == ref) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_LEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (aaaa[j] <= ref) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_GREATER: + for (j = 0; j < QUAD_SIZE; j++) { + if (aaaa[j] > ref) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_NOTEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (aaaa[j] != ref) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_GEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (aaaa[j] >= ref) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_ALWAYS: + passMask = MASK_ALL; + break; + default: + abort(); + } + + quad->mask &= passMask; + + if (quad->mask) + qs->next->run(qs->next, quad); +} + + +static void alpha_test_begin(struct quad_stage *qs) +{ + qs->next->begin(qs->next); +} + + +static void alpha_test_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +struct quad_stage * +sp_quad_alpha_test_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = alpha_test_begin; + stage->run = alpha_test_quad; + stage->destroy = alpha_test_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c new file mode 100644 index 0000000000..17f3ecd0b8 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -0,0 +1,749 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * quad blending + * \author Brian Paul + */ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_tile_cache.h" +#include "sp_quad.h" + + +#define VEC4_COPY(DST, SRC) \ +do { \ + DST[0] = SRC[0]; \ + DST[1] = SRC[1]; \ + DST[2] = SRC[2]; \ + DST[3] = SRC[3]; \ +} while(0) + +#define VEC4_SCALAR(DST, SRC) \ +do { \ + DST[0] = SRC; \ + DST[1] = SRC; \ + DST[2] = SRC; \ + DST[3] = SRC; \ +} while(0) + +#define VEC4_ADD(R, A, B) \ +do { \ + R[0] = A[0] + B[0]; \ + R[1] = A[1] + B[1]; \ + R[2] = A[2] + B[2]; \ + R[3] = A[3] + B[3]; \ +} while (0) + +#define VEC4_SUB(R, A, B) \ +do { \ + R[0] = A[0] - B[0]; \ + R[1] = A[1] - B[1]; \ + R[2] = A[2] - B[2]; \ + R[3] = A[3] - B[3]; \ +} while (0) + +#define VEC4_MUL(R, A, B) \ +do { \ + R[0] = A[0] * B[0]; \ + R[1] = A[1] * B[1]; \ + R[2] = A[2] * B[2]; \ + R[3] = A[3] * B[3]; \ +} while (0) + +#define VEC4_MIN(R, A, B) \ +do { \ + R[0] = (A[0] < B[0]) ? A[0] : B[0]; \ + R[1] = (A[1] < B[1]) ? A[1] : B[1]; \ + R[2] = (A[2] < B[2]) ? A[2] : B[2]; \ + R[3] = (A[3] < B[3]) ? A[3] : B[3]; \ +} while (0) + +#define VEC4_MAX(R, A, B) \ +do { \ + R[0] = (A[0] > B[0]) ? A[0] : B[0]; \ + R[1] = (A[1] > B[1]) ? A[1] : B[1]; \ + R[2] = (A[2] > B[2]) ? A[2] : B[2]; \ + R[3] = (A[3] > B[3]) ? A[3] : B[3]; \ +} while (0) + + + +static void +logicop_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + float dest[4][QUAD_SIZE]; + ubyte src[4][4], dst[4][4], res[4][4]; + uint *src4 = (uint *) src; + uint *dst4 = (uint *) dst; + uint *res4 = (uint *) res; + struct softpipe_cached_tile * + tile = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[softpipe->current_cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color; + uint i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } + } + + /* convert to ubyte */ + for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][0], dest[j][0]); /* P0 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][1], dest[j][1]); /* P1 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][2], dest[j][2]); /* P2 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][3], dest[j][3]); /* P3 */ + + UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quadColor[j][0]); /* P0 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quadColor[j][1]); /* P1 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quadColor[j][2]); /* P2 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quadColor[j][3]); /* P3 */ + } + + switch (softpipe->blend->logicop_func) { + case PIPE_LOGICOP_CLEAR: + for (j = 0; j < 4; j++) + res4[j] = 0; + break; + case PIPE_LOGICOP_NOR: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] | dst4[j]); + break; + case PIPE_LOGICOP_AND_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j] & dst4[j]; + break; + case PIPE_LOGICOP_COPY_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j]; + break; + case PIPE_LOGICOP_AND_REVERSE: + for (j = 0; j < 4; j++) + res4[j] = src4[j] & ~dst4[j]; + break; + case PIPE_LOGICOP_INVERT: + for (j = 0; j < 4; j++) + res4[j] = ~dst4[j]; + break; + case PIPE_LOGICOP_XOR: + for (j = 0; j < 4; j++) + res4[j] = dst4[j] ^ src4[j]; + break; + case PIPE_LOGICOP_NAND: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] & dst4[j]); + break; + case PIPE_LOGICOP_AND: + for (j = 0; j < 4; j++) + res4[j] = src4[j] & dst4[j]; + break; + case PIPE_LOGICOP_EQUIV: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] ^ dst4[j]); + break; + case PIPE_LOGICOP_NOOP: + for (j = 0; j < 4; j++) + res4[j] = dst4[j]; + break; + case PIPE_LOGICOP_OR_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j] | dst4[j]; + break; + case PIPE_LOGICOP_COPY: + for (j = 0; j < 4; j++) + res4[j] = src4[j]; + break; + case PIPE_LOGICOP_OR_REVERSE: + for (j = 0; j < 4; j++) + res4[j] = src4[j] | ~dst4[j]; + break; + case PIPE_LOGICOP_OR: + for (j = 0; j < 4; j++) + res4[j] = src4[j] | dst4[j]; + break; + case PIPE_LOGICOP_SET: + for (j = 0; j < 4; j++) + res4[j] = ~0; + break; + default: + assert(0); + } + + for (j = 0; j < 4; j++) { + quadColor[j][0] = UBYTE_TO_FLOAT(res[j][0]); + quadColor[j][1] = UBYTE_TO_FLOAT(res[j][1]); + quadColor[j][2] = UBYTE_TO_FLOAT(res[j][2]); + quadColor[j][3] = UBYTE_TO_FLOAT(res[j][3]); + } + + /* pass quad to next stage */ + qs->next->run(qs->next, quad); +} + + + + +static void +blend_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + static const float zero[4] = { 0, 0, 0, 0 }; + static const float one[4] = { 1, 1, 1, 1 }; + float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[softpipe->current_cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color; + uint i, j; + + if (softpipe->blend->logicop_enable) { + logicop_quad(qs, quad); + return; + } + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } + } + + /* + * Compute src/first term RGB + */ + switch (softpipe->blend->rgb_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[0], quadColor[0]); /* R */ + VEC4_COPY(source[1], quadColor[1]); /* G */ + VEC4_COPY(source[2], quadColor[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + VEC4_MUL(source[0], quadColor[0], quadColor[0]); /* R */ + VEC4_MUL(source[1], quadColor[1], quadColor[1]); /* G */ + VEC4_MUL(source[2], quadColor[2], quadColor[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const float *alpha = quadColor[3]; + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_DST_COLOR: + VEC4_MUL(source[0], quadColor[0], dest[0]); /* R */ + VEC4_MUL(source[1], quadColor[1], dest[1]); /* G */ + VEC4_MUL(source[2], quadColor[2], dest[2]); /* B */ + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + { + const float *alpha = dest[3]; + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + { + const float *alpha = quadColor[3]; + float diff[4]; + VEC4_SUB(diff, one, dest[3]); + VEC4_MIN(source[0], alpha, diff); /* R */ + VEC4_MIN(source[1], alpha, diff); /* G */ + VEC4_MIN(source[2], alpha, diff); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], comp); /* R */ + VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], comp); /* G */ + VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float alpha[4]; + VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_SRC1_ALPHA: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[0], zero); /* R */ + VEC4_COPY(source[1], zero); /* G */ + VEC4_COPY(source[2], zero); /* B */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, quadColor[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, dest[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, dest[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, dest[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + { + float inv_comp[4]; + /* R */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); + VEC4_MUL(source[0], quadColor[0], inv_comp); + /* G */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); + VEC4_MUL(source[1], quadColor[1], inv_comp); + /* B */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); + VEC4_MUL(source[2], quadColor[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_alpha[4]; + VEC4_SCALAR(inv_alpha, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + assert(0); /* to do */ + break; + default: + abort(); + } + + /* + * Compute src/first term A + */ + switch (softpipe->blend->alpha_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[3], quadColor[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const float *alpha = quadColor[3]; + VEC4_MUL(source[3], quadColor[3], alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + { + const float *alpha = quadColor[3]; + float diff[4]; + VEC4_SUB(diff, one, dest[3]); + VEC4_MIN(source[3], alpha, diff); /* A */ + } + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(source[3], quadColor[3], comp); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[3], zero); /* A */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, quadColor[3]); + VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, dest[3]); + VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + /* A */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(source[3], quadColor[3], inv_comp); + } + break; + default: + abort(); + } + + + /* + * Compute dest/second term RGB + */ + switch (softpipe->blend->rgb_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + VEC4_MUL(dest[0], dest[0], quadColor[0]); /* R */ + VEC4_MUL(dest[1], dest[1], quadColor[1]); /* G */ + VEC4_MUL(dest[2], dest[2], quadColor[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + VEC4_MUL(dest[0], dest[0], quadColor[3]); /* R * A */ + VEC4_MUL(dest[1], dest[1], quadColor[3]); /* G * A */ + VEC4_MUL(dest[2], dest[2], quadColor[3]); /* B * A */ + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */ + VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */ + VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */ + break; + case PIPE_BLENDFACTOR_DST_COLOR: + VEC4_MUL(dest[0], dest[0], dest[0]); /* R */ + VEC4_MUL(dest[1], dest[1], dest[1]); /* G */ + VEC4_MUL(dest[2], dest[2], dest[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* illegal */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ + VEC4_MUL(dest[0], dest[0], comp); /* R */ + VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ + VEC4_MUL(dest[1], dest[1], comp); /* G */ + VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ + VEC4_MUL(dest[2], dest[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(dest[0], dest[0], comp); /* R */ + VEC4_MUL(dest[1], dest[1], comp); /* G */ + VEC4_MUL(dest[2], dest[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[0], zero); /* R */ + VEC4_COPY(dest[1], zero); /* G */ + VEC4_COPY(dest[2], zero); /* B */ + break; + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: + /* XXX what are these? */ + assert(0); + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ + VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ + VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ + VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ + VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ + VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quadColor[3]); + VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ + VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ + VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quadColor[3]); /* A */ + VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ + VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ + VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[0]); /* R */ + VEC4_MUL(dest[0], dest[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, dest[1]); /* G */ + VEC4_MUL(dest[1], dest[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, dest[2]); /* B */ + VEC4_MUL(dest[2], dest[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + { + float inv_comp[4]; + /* R */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); + VEC4_MUL(dest[0], dest[0], inv_comp); + /* G */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); + VEC4_MUL(dest[1], dest[1], inv_comp); + /* B */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); + VEC4_MUL(dest[2], dest[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(dest[0], dest[0], inv_comp); + VEC4_MUL(dest[1], dest[1], inv_comp); + VEC4_MUL(dest[2], dest[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + /* XXX what are these? */ + assert(0); + break; + default: + assert(0); + } + + /* + * Compute dest/second term A + */ + switch (softpipe->blend->alpha_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_SRC_ALPHA: + VEC4_MUL(dest[3], dest[3], quadColor[3]); /* A * A */ + break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(dest[3], dest[3], dest[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* illegal */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(dest[3], dest[3], comp); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[3], zero); /* A */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quadColor[3]); + VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[3]); /* A */ + VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(dest[3], dest[3], inv_comp); + } + break; + default: + assert(0); + } + + /* + * Combine RGB terms + */ + switch (softpipe->blend->rgb_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quadColor[0], source[0], dest[0]); /* R */ + VEC4_ADD(quadColor[1], source[1], dest[1]); /* G */ + VEC4_ADD(quadColor[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quadColor[0], source[0], dest[0]); /* R */ + VEC4_SUB(quadColor[1], source[1], dest[1]); /* G */ + VEC4_SUB(quadColor[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quadColor[0], dest[0], source[0]); /* R */ + VEC4_SUB(quadColor[1], dest[1], source[1]); /* G */ + VEC4_SUB(quadColor[2], dest[2], source[2]); /* B */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quadColor[0], source[0], dest[0]); /* R */ + VEC4_MIN(quadColor[1], source[1], dest[1]); /* G */ + VEC4_MIN(quadColor[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quadColor[0], source[0], dest[0]); /* R */ + VEC4_MAX(quadColor[1], source[1], dest[1]); /* G */ + VEC4_MAX(quadColor[2], source[2], dest[2]); /* B */ + break; + default: + assert(0); + } + + /* + * Combine A terms + */ + switch (softpipe->blend->alpha_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quadColor[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quadColor[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quadColor[3], dest[3], source[3]); /* A */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quadColor[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quadColor[3], source[3], dest[3]); /* A */ + break; + default: + abort(); + } + + /* pass blended quad to next stage */ + qs->next->run(qs->next, quad); +} + + +static void blend_begin(struct quad_stage *qs) +{ + qs->next->begin(qs->next); +} + + +static void blend_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = blend_begin; + stage->run = blend_quad; + stage->destroy = blend_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_bufloop.c b/src/gallium/drivers/softpipe/sp_quad_bufloop.c new file mode 100644 index 0000000000..2ae4e22a7d --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_bufloop.c @@ -0,0 +1,72 @@ + +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" + + +/** + * Loop over colorbuffers, passing quad to next stage each time. + */ +static void +cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + float tmp[4][QUAD_SIZE]; + unsigned i; + + assert(sizeof(quad->outputs.color) == sizeof(tmp)); + assert(softpipe->framebuffer.num_cbufs <= PIPE_MAX_COLOR_BUFS); + + /* make copy of original colors since they can get modified + * by blending and masking. + * XXX we won't have to do this if the fragment program actually emits + * N separate colors and we're drawing to N color buffers (MRT). + * But if we emitted one color and glDrawBuffer(GL_FRONT_AND_BACK) is + * in effect, we need to save/restore colors like this. + */ + memcpy(tmp, quad->outputs.color, sizeof(tmp)); + + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { + /* set current cbuffer */ + softpipe->current_cbuf = i; + + /* pass blended quad to next stage */ + qs->next->run(qs->next, quad); + + /* restore quad's colors for next buffer */ + memcpy(quad->outputs.color, tmp, sizeof(tmp)); + } +} + + +static void cbuf_loop_begin(struct quad_stage *qs) +{ + qs->next->begin(qs->next); +} + + +static void cbuf_loop_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +/** + * Create the colorbuffer loop stage. + * This is used to implement multiple render targets and GL_FRONT_AND_BACK + * rendering. + */ +struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = cbuf_loop_begin; + stage->run = cbuf_loop_quad; + stage->destroy = cbuf_loop_destroy; + + return stage; +} + diff --git a/src/gallium/drivers/softpipe/sp_quad_colormask.c b/src/gallium/drivers/softpipe/sp_quad_colormask.c new file mode 100644 index 0000000000..1f09d900ca --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_colormask.c @@ -0,0 +1,110 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief quad colormask stage + * \author Brian Paul + */ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" +#include "sp_tile_cache.h" + + + +/** + * XXX colormask could be rolled into blending... + */ +static void +colormask_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + float dest[4][QUAD_SIZE]; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[softpipe->current_cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color; + uint i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } + } + + /* R */ + if (!(softpipe->blend->colormask & PIPE_MASK_R)) + COPY_4V(quadColor[0], dest[0]); + + /* G */ + if (!(softpipe->blend->colormask & PIPE_MASK_G)) + COPY_4V(quadColor[1], dest[1]); + + /* B */ + if (!(softpipe->blend->colormask & PIPE_MASK_B)) + COPY_4V(quadColor[2], dest[2]); + + /* A */ + if (!(softpipe->blend->colormask & PIPE_MASK_A)) + COPY_4V(quadColor[3], dest[3]); + + /* pass quad to next stage */ + qs->next->run(qs->next, quad); +} + + +static void colormask_begin(struct quad_stage *qs) +{ + qs->next->begin(qs->next); +} + + +static void colormask_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = colormask_begin; + stage->run = colormask_quad; + stage->destroy = colormask_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_coverage.c b/src/gallium/drivers/softpipe/sp_quad_coverage.c new file mode 100644 index 0000000000..b3d3fae22f --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_coverage.c @@ -0,0 +1,88 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * \brief Apply AA coverage to quad alpha valus + * \author Brian Paul + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" + + +/** + * Multiply quad's alpha values by the fragment coverage. + */ +static void +coverage_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + + if ((softpipe->rasterizer->poly_smooth && quad->prim == PRIM_TRI) || + (softpipe->rasterizer->line_smooth && quad->prim == PRIM_LINE) || + (softpipe->rasterizer->point_smooth && quad->prim == PRIM_POINT)) { + float (*quadColor)[4] = quad->outputs.color; + unsigned j; + for (j = 0; j < QUAD_SIZE; j++) { + assert(quad->coverage[j] >= 0.0); + assert(quad->coverage[j] <= 1.0); + quadColor[3][j] *= quad->coverage[j]; + } + } + + qs->next->run(qs->next, quad); +} + + +static void coverage_begin(struct quad_stage *qs) +{ + qs->next->begin(qs->next); +} + + +static void coverage_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = coverage_begin; + stage->run = coverage_quad; + stage->destroy = coverage_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c new file mode 100644 index 0000000000..a9a0754f27 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c @@ -0,0 +1,276 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Quad depth testing + */ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" +#include "sp_tile_cache.h" + + +/** + * Do depth testing for a quad. + * Not static since it's used by the stencil code. + */ + +/* + * To increase efficiency, we should probably have multiple versions + * of this function that are specifically for Z16, Z32 and FP Z buffers. + * Try to effectively do that with codegen... + */ + +void +sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + struct pipe_surface *ps = softpipe->framebuffer.zsbuf; + const enum pipe_format format = ps->format; + unsigned bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ + unsigned qzzzz[QUAD_SIZE]; /**< Z values from the quad */ + unsigned zmask = 0; + unsigned j; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->x0, quad->y0); + + assert(ps); /* shouldn't get here if there's no zbuffer */ + + /* + * Convert quad's float depth values to int depth values (qzzzz). + * If the Z buffer stores integer values, we _have_ to do the depth + * compares with integers (not floats). Otherwise, the float->int->float + * conversion of Z values (which isn't an identity function) will cause + * Z-fighting errors. + * + * Also, get the zbuffer values (bzzzz) from the cached tile. + */ + switch (format) { + case PIPE_FORMAT_Z16_UNORM: + { + float scale = 65535.0; + + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); + } + + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + bzzzz[j] = tile->data.depth16[y][x]; + } + } + break; + case PIPE_FORMAT_Z32_UNORM: + { + double scale = (double) (uint) ~0UL; + + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); + } + + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + bzzzz[j] = tile->data.depth32[y][x]; + } + } + break; + case PIPE_FORMAT_S8Z24_UNORM: + { + float scale = (float) ((1 << 24) - 1); + + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); + } + + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + bzzzz[j] = tile->data.depth32[y][x] & 0xffffff; + } + } + break; + case PIPE_FORMAT_Z24S8_UNORM: + { + float scale = (float) ((1 << 24) - 1); + + for (j = 0; j < QUAD_SIZE; j++) { + qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); + } + + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + bzzzz[j] = tile->data.depth32[y][x] >> 8; + } + } + break; + default: + assert(0); + } + + switch (softpipe->depth_stencil->depth.func) { + case PIPE_FUNC_NEVER: + /* zmask = 0 */ + break; + case PIPE_FUNC_LESS: + /* Note this is pretty much a single sse or cell instruction. + * Like this: quad->mask &= (quad->outputs.depth < zzzz); + */ + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] < bzzzz[j]) + zmask |= 1 << j; + } + break; + case PIPE_FUNC_EQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] == bzzzz[j]) + zmask |= 1 << j; + } + break; + case PIPE_FUNC_LEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] <= bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_GREATER: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] > bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_NOTEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] != bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_GEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (qzzzz[j] >= bzzzz[j]) + zmask |= (1 << j); + } + break; + case PIPE_FUNC_ALWAYS: + zmask = MASK_ALL; + break; + default: + abort(); + } + + quad->mask &= zmask; + + if (softpipe->depth_stencil->depth.writemask) { + + /* This is also efficient with sse / spe instructions: + */ + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + bzzzz[j] = qzzzz[j]; + } + } + + /* put updated Z values back into cached tile */ + switch (format) { + case PIPE_FORMAT_Z16_UNORM: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + tile->data.depth16[y][x] = (ushort) bzzzz[j]; + } + break; + case PIPE_FORMAT_Z32_UNORM: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + tile->data.depth32[y][x] = bzzzz[j]; + } + break; + case PIPE_FORMAT_S8Z24_UNORM: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + uint s8z24 = tile->data.depth32[y][x]; + s8z24 = (s8z24 & 0xff000000) | bzzzz[j]; + tile->data.depth32[y][x] = s8z24; + } + break; + case PIPE_FORMAT_Z24S8_UNORM: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + uint z24s8 = tile->data.depth32[y][x]; + z24s8 = (z24s8 & 0xff) | (bzzzz[j] << 24); + tile->data.depth32[y][x] = z24s8; + } + break; + default: + assert(0); + } + } +} + + +static void +depth_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ + sp_depth_test_quad(qs, quad); + + if (quad->mask) + qs->next->run(qs->next, quad); +} + + +static void depth_test_begin(struct quad_stage *qs) +{ + qs->next->begin(qs->next); +} + + +static void depth_test_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = depth_test_begin; + stage->run = depth_test_quad; + stage->destroy = depth_test_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_earlyz.c b/src/gallium/drivers/softpipe/sp_quad_earlyz.c new file mode 100644 index 0000000000..22ea99049f --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_earlyz.c @@ -0,0 +1,88 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Quad early-z testing + */ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "sp_headers.h" +#include "sp_quad.h" + + +/** + * All this stage does is compute the quad's Z values (which is normally + * done by the shading stage). + * The next stage will do the actual depth test. + */ +static void +earlyz_quad( + struct quad_stage *qs, + struct quad_header *quad ) +{ + const float fx = (float) quad->x0; + const float fy = (float) quad->y0; + const float dzdx = quad->posCoef->dadx[2]; + const float dzdy = quad->posCoef->dady[2]; + const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy; + + quad->outputs.depth[0] = z0; + quad->outputs.depth[1] = z0 + dzdx; + quad->outputs.depth[2] = z0 + dzdy; + quad->outputs.depth[3] = z0 + dzdx + dzdy; + + qs->next->run( qs->next, quad ); +} + +static void +earlyz_begin( + struct quad_stage *qs ) +{ + qs->next->begin( qs->next ); +} + +static void +earlyz_destroy( + struct quad_stage *qs ) +{ + FREE( qs ); +} + +struct quad_stage * +sp_quad_earlyz_stage( + struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT( quad_stage ); + + stage->softpipe = softpipe; + stage->begin = earlyz_begin; + stage->run = earlyz_quad; + stage->destroy = earlyz_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c new file mode 100644 index 0000000000..3316858413 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -0,0 +1,390 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "x86/rtasm/x86sse.h" + +#ifdef MESA_LLVM +#include "pipe/llvm/gallivm.h" +#endif + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "sp_texture.h" +#include "sp_tex_sample.h" + + +struct quad_shade_stage +{ + struct quad_stage stage; + struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; + struct tgsi_exec_machine machine; + struct tgsi_exec_vector *inputs, *outputs; + int colorOutSlot, depthOutSlot; +#ifdef MESA_LLVM + struct gallivm_prog *llvm_prog; +#endif +}; + + +/** cast wrapper */ +static INLINE struct quad_shade_stage * +quad_shade_stage(struct quad_stage *qs) +{ + return (struct quad_shade_stage *) qs; +} + + +/** + * Compute quad X,Y,Z,W for the four fragments in a quad. + * Note that we only need to "compute" X and Y for the upper-left fragment. + * We could do less work if we're not depth testing, or there's no + * perspective-corrected attributes, but that's seldom. + */ +static void +setup_pos_vector(const struct tgsi_interp_coef *coef, + float x, float y, + struct tgsi_exec_vector *quadpos) +{ + uint chan; + /* do X */ + quadpos->xyzw[0].f[0] = x; + /* do Y */ + quadpos->xyzw[1].f[0] = y; + /* do Z and W for all fragments in the quad */ + for (chan = 2; chan < 4; chan++) { + const float dadx = coef->dadx[chan]; + const float dady = coef->dady[chan]; + const float a0 = coef->a0[chan] + dadx * x + dady * y; + quadpos->xyzw[chan].f[0] = a0; + quadpos->xyzw[chan].f[1] = a0 + dadx; + quadpos->xyzw[chan].f[2] = a0 + dady; + quadpos->xyzw[chan].f[3] = a0 + dadx + dady; + } +} + + +typedef void (XSTDCALL *codegen_function)( + const struct tgsi_exec_vector *input, + struct tgsi_exec_vector *output, + float (*constant)[4], + struct tgsi_exec_vector *temporary, + const struct tgsi_interp_coef *coef +#if 0 + ,const struct tgsi_exec_vector *quadPos +#endif + ); + + +/** + * Execute fragment shader for the four fragments in the quad. + */ +static void +shade_quad( + struct quad_stage *qs, + struct quad_header *quad ) +{ + struct quad_shade_stage *qss = quad_shade_stage( qs ); + struct softpipe_context *softpipe = qs->softpipe; + struct tgsi_exec_machine *machine = &qss->machine; + + /* Consts do not require 16 byte alignment. */ + machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; + + machine->InterpCoefs = quad->coef; + + /* Compute X, Y, Z, W vals for this quad */ + setup_pos_vector(quad->posCoef, (float) quad->x0, (float) quad->y0, &machine->QuadPos); + + /* run shader */ +#if defined(__i386__) || defined(__386__) + if( softpipe->use_sse ) { + codegen_function func = (codegen_function) x86_get_func( &softpipe->fs->sse2_program ); + func( + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps, + machine->InterpCoefs +#if 0 + ,machine->QuadPos +#endif + ); + quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]); + } + else +#endif + { + quad->mask &= tgsi_exec_machine_run( machine ); + } + + /* store result color */ + if (qss->colorOutSlot >= 0) { + /* XXX need to handle multiple color outputs someday */ + assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] + == TGSI_SEMANTIC_COLOR); + memcpy( + quad->outputs.color, + &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); + } + + /* + * XXX the following code for updating quad->outputs.depth + * isn't really needed if we did early z testing. + */ + + /* store result Z */ + if (qss->depthOutSlot >= 0) { + /* output[slot] is new Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; + } + } + else { + /* copy input Z (which was interpolated by the executor) to output Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; + /* XXX not sure the above line is always correct. The following + * might be better: + quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i]; + */ + } + } + + /* shader may cull fragments */ + if( quad->mask ) { + qs->next->run( qs->next, quad ); + } +} + +#if 0 +#ifdef MESA_LLVM +#define DLLVM 0 +static void +shade_quad_llvm(struct quad_stage *qs, + struct quad_header *quad) +{ + struct quad_shade_stage *qss = quad_shade_stage(qs); + struct softpipe_context *softpipe = qs->softpipe; + float dests[4][16][4] ALIGN16_ATTRIB; + float inputs[4][16][4] ALIGN16_ATTRIB; + const float fx = (float) quad->x0; + const float fy = (float) quad->y0; + struct gallivm_prog *llvm = qss->llvm_prog; + + inputs[0][0][0] = fx; + inputs[1][0][0] = fx + 1.0f; + inputs[2][0][0] = fx; + inputs[3][0][0] = fx + 1.0f; + + inputs[0][0][1] = fy; + inputs[1][0][1] = fy; + inputs[2][0][1] = fy + 1.0f; + inputs[3][0][1] = fy + 1.0f; +#if DLLVM + debug_printf("MASK = %d\n", quad->mask); +#endif + gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef); +#if DLLVM + for (int i = 0; i < 4; ++i) { + for (int j = 0; j < 2; ++j) { + debug_printf("IN(%d,%d) [%f %f %f %f]\n", i, j, + inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]); + } + } +#endif + + quad->mask &= + gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs, + softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], + qss->samplers); +#if DLLVM + debug_printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n", + dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], + dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]); +#endif + + /* store result color */ + if (qss->colorOutSlot >= 0) { + unsigned i; + /* XXX need to handle multiple color outputs someday */ + assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] + == TGSI_SEMANTIC_COLOR); + for (i = 0; i < QUAD_SIZE; ++i) { + quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0]; + quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1]; + quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2]; + quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3]; + } + } +#if DLLVM + for (int i = 0; i < QUAD_SIZE; ++i) { + debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, + quad->outputs.color[0][i], + quad->outputs.color[1][i], + quad->outputs.color[2][i], + quad->outputs.color[3][i]); + } +#endif + + /* store result Z */ + if (qss->depthOutSlot >= 0) { + /* output[slot] is new Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = dests[i][0][2]; + } + } + else { + /* copy input Z (which was interpolated by the executor) to output Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = inputs[i][0][2]; + } + } +#if DLLVM + debug_printf("D [%f, %f, %f, %f] mask = %d\n", + quad->outputs.depth[0], + quad->outputs.depth[1], + quad->outputs.depth[2], + quad->outputs.depth[3], quad->mask); +#endif + + /* shader may cull fragments */ + if( quad->mask ) { + qs->next->run( qs->next, quad ); + } +} +#endif /*MESA_LLVM*/ +#endif + +/** + * Per-primitive (or per-begin?) setup + */ +static void shade_begin(struct quad_stage *qs) +{ + struct quad_shade_stage *qss = quad_shade_stage(qs); + struct softpipe_context *softpipe = qs->softpipe; + unsigned i; + + /* set TGSI sampler state that varies */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + qss->samplers[i].state = softpipe->sampler[i]; + qss->samplers[i].texture = &softpipe->texture[i]->base; + } + +#ifdef MESA_LLVM + qss->llvm_prog = softpipe->fs->llvm_prog; +#endif + /* XXX only do this if the fragment shader changes... */ + tgsi_exec_machine_init(&qss->machine, + softpipe->fs->shader.tokens, + PIPE_MAX_SAMPLERS, + qss->samplers ); + + /* find output slots for depth, color */ + qss->colorOutSlot = -1; + qss->depthOutSlot = -1; + for (i = 0; i < qss->stage.softpipe->fs->shader.num_outputs; i++) { + switch (qss->stage.softpipe->fs->shader.output_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + qss->depthOutSlot = i; + break; + case TGSI_SEMANTIC_COLOR: + qss->colorOutSlot = i; + break; + } + } + + qs->next->begin(qs->next); +} + + +static void shade_destroy(struct quad_stage *qs) +{ + struct quad_shade_stage *qss = (struct quad_shade_stage *) qs; + + tgsi_exec_machine_free_data(&qss->machine); + FREE( qss->inputs ); + FREE( qss->outputs ); + FREE( qs ); +} + + +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) +{ + struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); + uint i; + + /* allocate storage for program inputs/outputs, aligned to 16 bytes */ + qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); + qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); + qss->machine.Inputs = align16(qss->inputs); + qss->machine.Outputs = align16(qss->outputs); + + qss->stage.softpipe = softpipe; + qss->stage.begin = shade_begin; +#ifdef MESA_LLVM + /* disable until ported to accept + * x/y and soa layout + qss->stage.run = shade_quad_llvm; + */ + softpipe->use_sse = FALSE; + qss->stage.run = shade_quad; +#else + qss->stage.run = shade_quad; +#endif + qss->stage.destroy = shade_destroy; + + /* set TGSI sampler state that's constant */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + assert(softpipe->tex_cache[i]); + qss->samplers[i].get_samples = sp_get_samples; + qss->samplers[i].pipe = &softpipe->pipe; + qss->samplers[i].cache = softpipe->tex_cache[i]; + } + + return &qss->stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_occlusion.c b/src/gallium/drivers/softpipe/sp_quad_occlusion.c new file mode 100644 index 0000000000..54254df1f1 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_occlusion.c @@ -0,0 +1,85 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * \brief Quad occlusion counter stage + * \author Brian Paul + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" + +static unsigned count_bits( unsigned val ) +{ + unsigned i; + + for (i = 0; val ; val >>= 1) + i += (val & 1); + + return i; +} + +static void +occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + + softpipe->occlusion_count += count_bits(quad->mask); + + qs->next->run(qs->next, quad); +} + + +static void occlusion_begin(struct quad_stage *qs) +{ + qs->next->begin(qs->next); +} + + +static void occlusion_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = occlusion_begin; + stage->run = occlusion_count_quad; + stage->destroy = occlusion_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_output.c b/src/gallium/drivers/softpipe/sp_quad_output.c new file mode 100644 index 0000000000..cfe8f11808 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_output.c @@ -0,0 +1,90 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_quad.h" +#include "sp_tile_cache.h" + + +/** + * Write quad to framebuffer, taking mask into account. + * + * Note that surfaces support only full quad reads and writes. + */ +static void +output_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[softpipe->current_cbuf], + quad->x0, quad->y0); + /* in-tile pos: */ + const int itx = quad->x0 % TILE_SIZE; + const int ity = quad->y0 % TILE_SIZE; + float (*quadColor)[4] = quad->outputs.color; + int i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + int x = itx + (j & 1); + int y = ity + (j >> 1); + for (i = 0; i < 4; i++) { /* loop over color chans */ + tile->data.color[y][x][i] = quadColor[i][j]; + } + } + } +} + + +static void output_begin(struct quad_stage *qs) +{ + assert(qs->next == NULL); +} + + +static void output_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = output_begin; + stage->run = output_quad; + stage->destroy = output_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_stencil.c b/src/gallium/drivers/softpipe/sp_quad_stencil.c new file mode 100644 index 0000000000..92a0da0083 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_stencil.c @@ -0,0 +1,352 @@ + +/** + * \brief Quad stencil testing + */ + + +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_tile_cache.h" +#include "sp_quad.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" + + +/** Only 8-bit stencil supported */ +#define STENCIL_MAX 0xff + + +/** + * Do the basic stencil test (compare stencil buffer values against the + * reference value. + * + * \param stencilVals the stencil values from the stencil buffer + * \param func the stencil func (PIPE_FUNC_x) + * \param ref the stencil reference value + * \param valMask the stencil value mask indicating which bits of the stencil + * values and ref value are to be used. + * \return mask indicating which pixels passed the stencil test + */ +static unsigned +do_stencil_test(const ubyte stencilVals[QUAD_SIZE], unsigned func, + unsigned ref, unsigned valMask) +{ + unsigned passMask = 0x0; + unsigned j; + + ref &= valMask; + + switch (func) { + case PIPE_FUNC_NEVER: + /* passMask = 0x0 */ + break; + case PIPE_FUNC_LESS: + for (j = 0; j < QUAD_SIZE; j++) { + if (ref < (stencilVals[j] & valMask)) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_EQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (ref == (stencilVals[j] & valMask)) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_LEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (ref <= (stencilVals[j] & valMask)) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_GREATER: + for (j = 0; j < QUAD_SIZE; j++) { + if (ref > (stencilVals[j] & valMask)) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_NOTEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (ref != (stencilVals[j] & valMask)) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_GEQUAL: + for (j = 0; j < QUAD_SIZE; j++) { + if (ref >= (stencilVals[j] & valMask)) { + passMask |= (1 << j); + } + } + break; + case PIPE_FUNC_ALWAYS: + passMask = MASK_ALL; + break; + default: + assert(0); + } + + return passMask; +} + + +/** + * Apply the stencil operator to stencil values. + * + * \param stencilVals the stencil buffer values (read and written) + * \param mask indicates which pixels to update + * \param op the stencil operator (PIPE_STENCIL_OP_x) + * \param ref the stencil reference value + * \param wrtMask writemask controlling which bits are changed in the + * stencil values + */ +static void +apply_stencil_op(ubyte stencilVals[QUAD_SIZE], + unsigned mask, unsigned op, ubyte ref, ubyte wrtMask) +{ + unsigned j; + ubyte newstencil[QUAD_SIZE]; + + for (j = 0; j < QUAD_SIZE; j++) { + newstencil[j] = stencilVals[j]; + } + + switch (op) { + case PIPE_STENCIL_OP_KEEP: + /* no-op */ + break; + case PIPE_STENCIL_OP_ZERO: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + newstencil[j] = 0; + } + } + break; + case PIPE_STENCIL_OP_REPLACE: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + newstencil[j] = ref; + } + } + break; + case PIPE_STENCIL_OP_INCR: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + if (stencilVals[j] < STENCIL_MAX) { + newstencil[j] = stencilVals[j] + 1; + } + } + } + break; + case PIPE_STENCIL_OP_DECR: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + if (stencilVals[j] > 0) { + newstencil[j] = stencilVals[j] - 1; + } + } + } + break; + case PIPE_STENCIL_OP_INCR_WRAP: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + newstencil[j] = stencilVals[j] + 1; + } + } + break; + case PIPE_STENCIL_OP_DECR_WRAP: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + newstencil[j] = stencilVals[j] - 1; + } + } + break; + case PIPE_STENCIL_OP_INVERT: + for (j = 0; j < QUAD_SIZE; j++) { + if (mask & (1 << j)) { + newstencil[j] = ~stencilVals[j]; + } + } + break; + default: + assert(0); + } + + /* + * update the stencil values + */ + if (wrtMask != STENCIL_MAX) { + /* apply bit-wise stencil buffer writemask */ + for (j = 0; j < QUAD_SIZE; j++) { + stencilVals[j] = (wrtMask & newstencil[j]) | (~wrtMask & stencilVals[j]); + } + } + else { + for (j = 0; j < QUAD_SIZE; j++) { + stencilVals[j] = newstencil[j]; + } + } +} + + +/** + * Do stencil (and depth) testing. Stenciling depends on the outcome of + * depth testing. + */ +static void +stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) +{ + struct softpipe_context *softpipe = qs->softpipe; + struct pipe_surface *ps = softpipe->framebuffer.zsbuf; + unsigned func, zFailOp, zPassOp, failOp; + ubyte ref, wrtMask, valMask; + ubyte stencilVals[QUAD_SIZE]; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->x0, quad->y0); + uint j; + uint face = quad->facing; + + if (!softpipe->depth_stencil->stencil[1].enabled) { + /* single-sided stencil test, use front (face=0) state */ + face = 0; + } + + /* choose front or back face function, operator, etc */ + /* XXX we could do these initializations once per primitive */ + func = softpipe->depth_stencil->stencil[face].func; + failOp = softpipe->depth_stencil->stencil[face].fail_op; + zFailOp = softpipe->depth_stencil->stencil[face].zfail_op; + zPassOp = softpipe->depth_stencil->stencil[face].zpass_op; + ref = softpipe->depth_stencil->stencil[face].ref_value; + wrtMask = softpipe->depth_stencil->stencil[face].write_mask; + valMask = softpipe->depth_stencil->stencil[face].value_mask; + + assert(ps); /* shouldn't get here if there's no stencil buffer */ + + /* get stencil values from cached tile */ + switch (ps->format) { + case PIPE_FORMAT_S8Z24_UNORM: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + stencilVals[j] = tile->data.depth32[y][x] >> 24; + } + break; + case PIPE_FORMAT_Z24S8_UNORM: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + stencilVals[j] = tile->data.depth32[y][x] & 0xff; + } + break; + case PIPE_FORMAT_U_S8: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + stencilVals[j] = tile->data.stencil8[y][x]; + } + break; + default: + assert(0); + } + + /* do the stencil test first */ + { + unsigned passMask, failMask; + passMask = do_stencil_test(stencilVals, func, ref, valMask); + failMask = quad->mask & ~passMask; + quad->mask &= passMask; + + if (failOp != PIPE_STENCIL_OP_KEEP) { + apply_stencil_op(stencilVals, failMask, failOp, ref, wrtMask); + } + } + + if (quad->mask) { + /* now the pixels that passed the stencil test are depth tested */ + if (softpipe->depth_stencil->depth.enabled) { + const unsigned origMask = quad->mask; + + sp_depth_test_quad(qs, quad); /* quad->mask is updated */ + + /* update stencil buffer values according to z pass/fail result */ + if (zFailOp != PIPE_STENCIL_OP_KEEP) { + const unsigned failMask = origMask & ~quad->mask; + apply_stencil_op(stencilVals, failMask, zFailOp, ref, wrtMask); + } + + if (zPassOp != PIPE_STENCIL_OP_KEEP) { + const unsigned passMask = origMask & quad->mask; + apply_stencil_op(stencilVals, passMask, zPassOp, ref, wrtMask); + } + } + else { + /* no depth test, apply Zpass operator to stencil buffer values */ + apply_stencil_op(stencilVals, quad->mask, zPassOp, ref, wrtMask); + } + + } + + /* put new stencil values into cached tile */ + switch (ps->format) { + case PIPE_FORMAT_S8Z24_UNORM: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + uint s8z24 = tile->data.depth32[y][x]; + s8z24 = (stencilVals[j] << 24) | (s8z24 & 0xffffff); + tile->data.depth32[y][x] = s8z24; + } + break; + case PIPE_FORMAT_Z24S8_UNORM: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + uint z24s8 = tile->data.depth32[y][x]; + z24s8 = (z24s8 & 0xffffff00) | stencilVals[j]; + tile->data.depth32[y][x] = z24s8; + } + break; + case PIPE_FORMAT_U_S8: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + tile->data.stencil8[y][x] = stencilVals[j]; + } + break; + default: + assert(0); + } + + if (quad->mask) + qs->next->run(qs->next, quad); +} + + +static void stencil_begin(struct quad_stage *qs) +{ + qs->next->begin(qs->next); +} + + +static void stencil_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = stencil_begin; + stage->run = stencil_test_quad; + stage->destroy = stencil_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c new file mode 100644 index 0000000000..8660432259 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c @@ -0,0 +1,94 @@ + +/** + * quad polygon stipple stage + */ + +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" + + +/** + * Apply polygon stipple to quads produced by triangle rasterization + */ +static void +stipple_quad(struct quad_stage *qs, struct quad_header *quad) +{ + static const uint bit31 = 1 << 31; + static const uint bit30 = 1 << 30; + + if (quad->prim == PRIM_TRI) { + struct softpipe_context *softpipe = qs->softpipe; + /* need to invert Y to index into OpenGL's stipple pattern */ + int y0, y1; + uint stipple0, stipple1; + if (softpipe->rasterizer->origin_lower_left) { + y0 = softpipe->framebuffer.cbufs[0]->height - 1 - quad->y0; + y1 = y0 - 1; + } + else { + y0 = quad->y0; + y1 = y0 + 1; + } + stipple0 = softpipe->poly_stipple.stipple[y0 % 32]; + stipple1 = softpipe->poly_stipple.stipple[y1 % 32]; + +#if 1 + { + const int col0 = quad->x0 % 32; + if ((stipple0 & (bit31 >> col0)) == 0) + quad->mask &= ~MASK_TOP_LEFT; + + if ((stipple0 & (bit30 >> col0)) == 0) + quad->mask &= ~MASK_TOP_RIGHT; + + if ((stipple1 & (bit31 >> col0)) == 0) + quad->mask &= ~MASK_BOTTOM_LEFT; + + if ((stipple1 & (bit30 >> col0)) == 0) + quad->mask &= ~MASK_BOTTOM_RIGHT; + } +#else + /* We'd like to use this code, but we'd need to redefine + * MASK_TOP_LEFT to be (1 << 1) and MASK_TOP_RIGHT to be (1 << 0), + * and similarly for the BOTTOM bits. But that may have undesirable + * side effects elsewhere. + */ + const int col0 = 30 - (quad->x0 % 32); + quad->mask &= (((stipple0 >> col0) & 0x3) | + (((stipple1 >> col0) & 0x3) << 2)); +#endif + if (!quad->mask) + return; + } + + qs->next->run(qs->next, quad); +} + + +static void stipple_begin(struct quad_stage *qs) +{ + qs->next->begin(qs->next); +} + + +static void stipple_destroy(struct quad_stage *qs) +{ + FREE( qs ); +} + + +struct quad_stage * +sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ) +{ + struct quad_stage *stage = CALLOC_STRUCT(quad_stage); + + stage->softpipe = softpipe; + stage->begin = stipple_begin; + stage->run = stipple_quad; + stage->destroy = stipple_destroy; + + return stage; +} diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c new file mode 100644 index 0000000000..6a8a43aeda --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_query.c @@ -0,0 +1,107 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#include "pipe/draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_query.h" + +struct softpipe_query { + uint64 start; + uint64 end; +}; + + +static struct softpipe_query *softpipe_query( struct pipe_query *p ) +{ + return (struct softpipe_query *)p; +} + +static struct pipe_query * +softpipe_create_query(struct pipe_context *pipe, + unsigned type) +{ + assert(type == PIPE_QUERY_OCCLUSION_COUNTER); + return (struct pipe_query *)CALLOC_STRUCT( softpipe_query ); +} + + +static void +softpipe_destroy_query(struct pipe_context *pipe, struct pipe_query *q) +{ + FREE(q); +} + + +static void +softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + struct softpipe_query *sq = softpipe_query(q); + + sq->start = softpipe->occlusion_count; +} + + +static void +softpipe_end_query(struct pipe_context *pipe, struct pipe_query *q) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + struct softpipe_query *sq = softpipe_query(q); + + sq->end = softpipe->occlusion_count; +} + + +static boolean +softpipe_get_query_result(struct pipe_context *pipe, + struct pipe_query *q, + boolean wait, + uint64 *result ) +{ + struct softpipe_query *sq = softpipe_query(q); + *result = sq->end - sq->start; + return TRUE; +} + + +void softpipe_init_query_funcs(struct softpipe_context *softpipe ) +{ + softpipe->pipe.create_query = softpipe_create_query; + softpipe->pipe.destroy_query = softpipe_destroy_query; + softpipe->pipe.begin_query = softpipe_begin_query; + softpipe->pipe.end_query = softpipe_end_query; + softpipe->pipe.get_query_result = softpipe_get_query_result; +} + + diff --git a/src/gallium/drivers/softpipe/sp_query.h b/src/gallium/drivers/softpipe/sp_query.h new file mode 100644 index 0000000000..05060a4575 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_query.h @@ -0,0 +1,39 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#ifndef SP_QUERY_H +#define SP_QUERY_H + +struct softpipe_context; +extern void softpipe_init_query_funcs(struct softpipe_context * ); + + +#endif /* SP_QUERY_H */ diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h new file mode 100644 index 0000000000..b79db0d1f1 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -0,0 +1,187 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef SP_STATE_H +#define SP_STATE_H + +#include "pipe/p_state.h" + +#include "x86/rtasm/x86sse.h" + + +#define SP_NEW_VIEWPORT 0x1 +#define SP_NEW_RASTERIZER 0x2 +#define SP_NEW_FS 0x4 +#define SP_NEW_BLEND 0x8 +#define SP_NEW_CLIP 0x10 +#define SP_NEW_SCISSOR 0x20 +#define SP_NEW_STIPPLE 0x40 +#define SP_NEW_FRAMEBUFFER 0x80 +#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100 +#define SP_NEW_CONSTANTS 0x200 +#define SP_NEW_SAMPLER 0x400 +#define SP_NEW_TEXTURE 0x800 +#define SP_NEW_VERTEX 0x1000 +#define SP_NEW_VS 0x2000 +#define SP_NEW_QUERY 0x4000 + + + +#ifdef MESA_LLVM +struct gallivm_prog; +#endif + + +struct vertex_info; + + +/** Subclass of pipe_shader_state */ +struct sp_fragment_shader_state { + struct pipe_shader_state shader; +#if defined(__i386__) || defined(__386__) + struct x86_function sse2_program; +#endif +#ifdef MESA_LLVM + struct gallivm_prog *llvm_prog; +#endif +}; + + +/** Subclass of pipe_shader_state */ +struct sp_vertex_shader_state { + struct pipe_shader_state shader; + struct draw_vertex_shader *draw_data; +}; + + + +void * +softpipe_create_blend_state(struct pipe_context *, + const struct pipe_blend_state *); +void softpipe_bind_blend_state(struct pipe_context *, + void *); +void softpipe_delete_blend_state(struct pipe_context *, + void *); + +void * +softpipe_create_sampler_state(struct pipe_context *, + const struct pipe_sampler_state *); +void softpipe_bind_sampler_state(struct pipe_context *, unsigned, void *); +void softpipe_delete_sampler_state(struct pipe_context *, void *); + +void * +softpipe_create_depth_stencil_state(struct pipe_context *, + const struct pipe_depth_stencil_alpha_state *); +void softpipe_bind_depth_stencil_state(struct pipe_context *, void *); +void softpipe_delete_depth_stencil_state(struct pipe_context *, void *); + +void * +softpipe_create_rasterizer_state(struct pipe_context *, + const struct pipe_rasterizer_state *); +void softpipe_bind_rasterizer_state(struct pipe_context *, void *); +void softpipe_delete_rasterizer_state(struct pipe_context *, void *); + +void softpipe_set_framebuffer_state( struct pipe_context *, + const struct pipe_framebuffer_state * ); + +void softpipe_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ); + +void softpipe_set_clip_state( struct pipe_context *, + const struct pipe_clip_state * ); + +void softpipe_set_constant_buffer(struct pipe_context *, + uint shader, uint index, + const struct pipe_constant_buffer *buf); + +void *softpipe_create_fs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_fs_state(struct pipe_context *, void *); +void softpipe_delete_fs_state(struct pipe_context *, void *); +void *softpipe_create_vs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_vs_state(struct pipe_context *, void *); +void softpipe_delete_vs_state(struct pipe_context *, void *); + +void softpipe_set_polygon_stipple( struct pipe_context *, + const struct pipe_poly_stipple * ); + +void softpipe_set_scissor_state( struct pipe_context *, + const struct pipe_scissor_state * ); + +void softpipe_set_sampler_texture( struct pipe_context *, + unsigned unit, + struct pipe_texture * ); + +void softpipe_set_viewport_state( struct pipe_context *, + const struct pipe_viewport_state * ); + +void softpipe_set_vertex_element(struct pipe_context *, + unsigned index, + const struct pipe_vertex_element *); + +void softpipe_set_vertex_buffer(struct pipe_context *, + unsigned index, + const struct pipe_vertex_buffer *); + + +void softpipe_update_derived( struct softpipe_context *softpipe ); + + +boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count); + +boolean softpipe_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); + + +void +softpipe_map_surfaces(struct softpipe_context *sp); + +void +softpipe_unmap_surfaces(struct softpipe_context *sp); + +void +softpipe_map_texture_surfaces(struct softpipe_context *sp); + +void +softpipe_unmap_texture_surfaces(struct softpipe_context *sp); + + +struct vertex_info * +softpipe_get_vertex_info(struct softpipe_context *softpipe); + +struct vertex_info * +softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe); + + +#endif diff --git a/src/gallium/drivers/softpipe/sp_state_blend.c b/src/gallium/drivers/softpipe/sp_state_blend.c new file mode 100644 index 0000000000..2d40d6bd8f --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_state_blend.c @@ -0,0 +1,98 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_state.h" + + +void * +softpipe_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + return mem_dup(blend, sizeof(*blend)); +} + +void softpipe_bind_blend_state( struct pipe_context *pipe, + void *blend ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->blend = (const struct pipe_blend_state *)blend; + + softpipe->dirty |= SP_NEW_BLEND; +} + +void softpipe_delete_blend_state(struct pipe_context *pipe, + void *blend) +{ + FREE( blend ); +} + + +void softpipe_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->blend_color = *blend_color; + + softpipe->dirty |= SP_NEW_BLEND; +} + + +/** XXX move someday? Or consolidate all these simple state setters + * into one file. + */ + + +void * +softpipe_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil) +{ + return mem_dup(depth_stencil, sizeof(*depth_stencil)); +} + +void +softpipe_bind_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->depth_stencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; + + softpipe->dirty |= SP_NEW_DEPTH_STENCIL_ALPHA; +} + +void +softpipe_delete_depth_stencil_state(struct pipe_context *pipe, void *depth) +{ + FREE( depth ); +} diff --git a/src/gallium/drivers/softpipe/sp_state_clip.c b/src/gallium/drivers/softpipe/sp_state_clip.c new file mode 100644 index 0000000000..08c5f06d05 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_state_clip.c @@ -0,0 +1,83 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "sp_context.h" +#include "sp_state.h" +#include "pipe/draw/draw_context.h" + + +void softpipe_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + /* pass the clip state to the draw module */ + draw_set_clip_state(softpipe->draw, clip); +} + + + +/* Called when driver state tracker notices changes to the viewport + * matrix: + */ +void softpipe_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->viewport = *viewport; /* struct copy */ + softpipe->dirty |= SP_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + draw_set_viewport_state(softpipe->draw, viewport); + + /* Using tnl/ and vf/ modules is temporary while getting started. + * Full pipe will have vertex shader, vertex fetch of its own. + */ +} + + +void softpipe_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + memcpy( &softpipe->scissor, scissor, sizeof(*scissor) ); + softpipe->dirty |= SP_NEW_SCISSOR; +} + + +void softpipe_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + memcpy( &softpipe->poly_stipple, stipple, sizeof(*stipple) ); + softpipe->dirty |= SP_NEW_STIPPLE; +} diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c new file mode 100644 index 0000000000..372597869f --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -0,0 +1,235 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_vertex.h" +#include "pipe/draw/draw_private.h" +#include "sp_context.h" +#include "sp_state.h" + + +/** + * Search vertex program's outputs to find a match for the given + * semantic name/index. Return the index of the output slot. + * + * Return 0 if not found. This will cause the fragment program to use + * vertex attrib 0 (position) in the cases where the fragment program + * attempts to use a missing vertex program output. This is an undefined + * condition that users shouldn't hit anyway. + */ +static int +find_vs_output(const struct pipe_shader_state *vs, + uint semantic_name, + uint semantic_index) +{ + uint i; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == semantic_name && + vs->output_semantic_index[i] == semantic_index) + return i; + } + return 0; +} + + +/** + * Mark the current vertex layout as "invalid". + * We'll validate the vertex layout later, when we start to actually + * render a point or line or tri. + */ +static void +invalidate_vertex_layout(struct softpipe_context *softpipe) +{ + softpipe->vertex_info.num_attribs = 0; +} + + +/** + * The vertex info describes how to convert the post-transformed vertices + * (simple float[][4]) used by the 'draw' module into vertices for + * rasterization. + * + * This function validates the vertex layout and returns a pointer to a + * vertex_info object. + */ +struct vertex_info * +softpipe_get_vertex_info(struct softpipe_context *softpipe) +{ + struct vertex_info *vinfo = &softpipe->vertex_info; + + if (vinfo->num_attribs == 0) { + /* compute vertex layout now */ + const struct pipe_shader_state *vs = &softpipe->vs->shader; + const struct pipe_shader_state *fs = &softpipe->fs->shader; + const enum interp_mode colorInterp + = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; + uint i; + + if (softpipe->vbuf) { + /* if using the post-transform vertex buffer, tell draw_vbuf to + * simply emit the whole post-xform vertex as-is: + */ + struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; + vinfo_vbuf->num_attribs = 0; + draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); + vinfo_vbuf->size = 4 * vs->num_outputs + + sizeof(struct vertex_header) / 4; + } + + /* + * Loop over fragment shader inputs, searching for the matching output + * from the vertex shader. + */ + vinfo->num_attribs = 0; + for (i = 0; i < fs->num_inputs; i++) { + int src; + switch (fs->input_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); + break; + + case TGSI_SEMANTIC_COLOR: + src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, + fs->input_semantic_index[i]); + draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); + break; + + case TGSI_SEMANTIC_FOG: + src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); + break; + + case TGSI_SEMANTIC_GENERIC: + /* this includes texcoords and varying vars */ + src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, + fs->input_semantic_index[i]); + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); + break; + + default: + assert(0); + } + } + + softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); + if (softpipe->psize_slot > 0) { + draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, + softpipe->psize_slot); + } + + draw_compute_vertex_size(vinfo); + } + + return vinfo; +} + + +/** + * Called from vbuf module. + * + * Note that there's actually two different vertex layouts in softpipe. + * + * The normal one is computed in softpipe_get_vertex_info() above and is + * used by the point/line/tri "setup" code. + * + * The other one (this one) is only used by the vbuf module (which is + * not normally used by default but used in testing). For the vbuf module, + * we basically want to pass-through the draw module's vertex layout as-is. + * When the softpipe vbuf code begins drawing, the normal vertex layout + * will come into play again. + */ +struct vertex_info * +softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe) +{ + (void) softpipe_get_vertex_info(softpipe); + return &softpipe->vertex_info_vbuf; +} + + +/** + * Recompute cliprect from scissor bounds, scissor enable and surface size. + */ +static void +compute_cliprect(struct softpipe_context *sp) +{ + unsigned surfWidth, surfHeight; + + if (sp->framebuffer.num_cbufs > 0) { + surfWidth = sp->framebuffer.cbufs[0]->width; + surfHeight = sp->framebuffer.cbufs[0]->height; + } + else { + /* no surface? */ + surfWidth = sp->scissor.maxx; + surfHeight = sp->scissor.maxy; + } + + if (sp->rasterizer->scissor) { + /* clip to scissor rect */ + sp->cliprect.minx = MAX2(sp->scissor.minx, 0); + sp->cliprect.miny = MAX2(sp->scissor.miny, 0); + sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth); + sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight); + } + else { + /* clip to surface bounds */ + sp->cliprect.minx = 0; + sp->cliprect.miny = 0; + sp->cliprect.maxx = surfWidth; + sp->cliprect.maxy = surfHeight; + } +} + + +/* Hopefully this will remain quite simple, otherwise need to pull in + * something like the state tracker mechanism. + */ +void softpipe_update_derived( struct softpipe_context *softpipe ) +{ + if (softpipe->dirty & (SP_NEW_RASTERIZER | + SP_NEW_FS | + SP_NEW_VS)) + invalidate_vertex_layout( softpipe ); + + if (softpipe->dirty & (SP_NEW_SCISSOR | + SP_NEW_DEPTH_STENCIL_ALPHA | + SP_NEW_FRAMEBUFFER)) + compute_cliprect(softpipe); + + if (softpipe->dirty & (SP_NEW_BLEND | + SP_NEW_DEPTH_STENCIL_ALPHA | + SP_NEW_FRAMEBUFFER | + SP_NEW_RASTERIZER | + SP_NEW_FS | + SP_NEW_QUERY)) + sp_build_quad_pipeline(softpipe); + + softpipe->dirty = 0; +} diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c new file mode 100644 index 0000000000..0b814fc284 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -0,0 +1,179 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "sp_context.h" +#include "sp_state.h" + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "pipe/draw/draw_context.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/llvm/gallivm.h" +#include "pipe/tgsi/util/tgsi_dump.h" +#include "pipe/tgsi/exec/tgsi_sse2.h" + + +void * +softpipe_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + struct sp_fragment_shader_state *state; + + /* Decide whether we'll be codegenerating this shader and if so do + * that now. + */ + + state = CALLOC_STRUCT(sp_fragment_shader_state); + if (!state) + return NULL; + + state->shader = *templ; + + if (softpipe->dump_fs) { + tgsi_dump(state->shader.tokens, 0); + } + +#ifdef MESA_LLVM + state->llvm_prog = 0; + +#if 0 + if (!gallivm_global_cpu_engine()) { + gallivm_cpu_engine_create(state->llvm_prog); + } + else + gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); +#endif + +#elif defined(__i386__) || defined(__386__) + if (softpipe->use_sse) { + x86_init_func( &state->sse2_program ); + tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); + } +#endif + + return state; +} + + +void +softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->fs = (struct sp_fragment_shader_state *) fs; + + softpipe->dirty |= SP_NEW_FS; +} + + +void +softpipe_delete_fs_state(struct pipe_context *pipe, void *fs) +{ + struct sp_fragment_shader_state *state = fs; + +#if defined(__i386__) || defined(__386__) + x86_release_func( &state->sse2_program ); +#endif + + FREE( state ); +} + + +void * +softpipe_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + struct sp_vertex_shader_state *state; + + state = CALLOC_STRUCT(sp_vertex_shader_state); + if (state == NULL ) { + return NULL; + } + + state->shader = *templ; + + state->draw_data = draw_create_vertex_shader(softpipe->draw, + &state->shader); + if (state->draw_data == NULL) { + FREE( state ); + return NULL; + } + + return state; +} + + +void +softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + softpipe->vs = (const struct sp_vertex_shader_state *)vs; + + draw_bind_vertex_shader(softpipe->draw, softpipe->vs->draw_data); + + softpipe->dirty |= SP_NEW_VS; +} + + +void +softpipe_delete_vs_state(struct pipe_context *pipe, void *vs) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + struct sp_vertex_shader_state *state = + (struct sp_vertex_shader_state *)vs; + + draw_delete_vertex_shader(softpipe->draw, state->draw_data); + FREE( state ); +} + + + +void +softpipe_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + /* note: reference counting */ + pipe_buffer_reference(ws, + &softpipe->constants[shader].buffer, + buf->buffer); + softpipe->constants[shader].size = buf->size; + + softpipe->dirty |= SP_NEW_CONSTANTS; +} diff --git a/src/gallium/drivers/softpipe/sp_state_rasterizer.c b/src/gallium/drivers/softpipe/sp_state_rasterizer.c new file mode 100644 index 0000000000..53755099dd --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_state_rasterizer.c @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "sp_context.h" +#include "sp_state.h" +#include "pipe/draw/draw_context.h" + + + +void * +softpipe_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *rast) +{ + return mem_dup(rast, sizeof(*rast)); +} + +void softpipe_bind_rasterizer_state(struct pipe_context *pipe, + void *setup) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + /* pass-through to draw module */ + draw_set_rasterizer_state(softpipe->draw, setup); + + softpipe->rasterizer = (struct pipe_rasterizer_state *)setup; + + softpipe->dirty |= SP_NEW_RASTERIZER; +} + +void softpipe_delete_rasterizer_state(struct pipe_context *pipe, + void *rasterizer) +{ + FREE( rasterizer ); +} + + diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c new file mode 100644 index 0000000000..ea348c7e95 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -0,0 +1,93 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: + * Brian Paul + */ + +#include "pipe/p_util.h" + +#include "pipe/draw/draw_context.h" + +#include "sp_context.h" +#include "sp_context.h" +#include "sp_state.h" +#include "sp_texture.h" +#include "sp_tile_cache.h" +#include "pipe/draw/draw_context.h" + + + +void * +softpipe_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + return mem_dup(sampler, sizeof(*sampler)); +} + +void +softpipe_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + draw_flush(softpipe->draw); + + assert(unit < PIPE_MAX_SAMPLERS); + softpipe->sampler[unit] = (struct pipe_sampler_state *)sampler; + + softpipe->dirty |= SP_NEW_SAMPLER; +} + + +void +softpipe_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + FREE( sampler ); +} + + +void +softpipe_set_sampler_texture(struct pipe_context *pipe, + unsigned unit, + struct pipe_texture *texture) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + + draw_flush(softpipe->draw); + + assert(unit < PIPE_MAX_SAMPLERS); + softpipe->texture[unit] = softpipe_texture(texture); /* ptr, not struct */ + + sp_tile_cache_set_texture(softpipe->tex_cache[unit], texture); + + softpipe->dirty |= SP_NEW_TEXTURE; +} + + + diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c new file mode 100644 index 0000000000..e2c6893e9f --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_state_surface.c @@ -0,0 +1,109 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ +#include "p_inlines.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_surface.h" +#include "sp_tile_cache.h" + + +/** + * XXX this might get moved someday + * Set the framebuffer surface info: color buffers, zbuffer, stencil buffer. + * Here, we flush the old surfaces and update the tile cache to point to the new + * surfaces. + */ +void +softpipe_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct softpipe_context *sp = softpipe_context(pipe); + uint i; + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + /* check if changing cbuf */ + if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) { + /* flush old */ + sp_flush_tile_cache(sp, sp->cbuf_cache[i]); + + /* assign new */ + sp->framebuffer.cbufs[i] = fb->cbufs[i]; + + /* update cache */ + sp_tile_cache_set_surface(sp->cbuf_cache[i], fb->cbufs[i]); + } + } + + sp->framebuffer.num_cbufs = fb->num_cbufs; + + /* zbuf changing? */ + if (sp->framebuffer.zsbuf != fb->zsbuf) { + /* flush old */ + sp_flush_tile_cache(sp, sp->zsbuf_cache); + + /* assign new */ + sp->framebuffer.zsbuf = fb->zsbuf; + + /* update cache */ + sp_tile_cache_set_surface(sp->zsbuf_cache, fb->zsbuf); + } + +#if 0 + /* XXX combined depth/stencil here */ + + /* sbuf changing? */ + if (sp->framebuffer.sbuf != fb->sbuf) { + /* flush old */ + sp_flush_tile_cache(sp, sp->sbuf_cache_sep); + + /* assign new */ + sp->framebuffer.sbuf = fb->sbuf; + + /* update cache */ + if (fb->sbuf != fb->zbuf) { + /* separate stencil buf */ + sp->sbuf_cache = sp->sbuf_cache_sep; + sp_tile_cache_set_surface(sp->sbuf_cache, fb->sbuf); + } + else { + /* combined depth/stencil */ + sp->sbuf_cache = sp->zbuf_cache; + sp_tile_cache_set_surface(sp->sbuf_cache, fb->sbuf); + } + } +#endif + + sp->dirty |= SP_NEW_FRAMEBUFFER; +} + + + + diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c new file mode 100644 index 0000000000..09ff540ccf --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_state_vertex.c @@ -0,0 +1,64 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_surface.h" + +#include "pipe/draw/draw_context.h" + + +void +softpipe_set_vertex_element(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_element *attrib) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + assert(index < PIPE_ATTRIB_MAX); + softpipe->vertex_element[index] = *attrib; /* struct copy */ + softpipe->dirty |= SP_NEW_VERTEX; + + draw_set_vertex_element(softpipe->draw, index, attrib); +} + + +void +softpipe_set_vertex_buffer(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_buffer *buffer) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + assert(index < PIPE_ATTRIB_MAX); + softpipe->vertex_buffer[index] = *buffer; /* struct copy */ + softpipe->dirty |= SP_NEW_VERTEX; + + draw_set_vertex_buffer(softpipe->draw, index, buffer); +} diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c new file mode 100644 index 0000000000..8802ced187 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -0,0 +1,159 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "pipe/util/p_tile.h" +#include "sp_context.h" +#include "sp_surface.h" + + + +/* Assumes all values are within bounds -- no checking at this level - + * do it higher up if required. + */ +static void +sp_surface_copy(struct pipe_context *pipe, + unsigned do_flip, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, unsigned width, unsigned height) +{ + assert( dst->cpp == src->cpp ); + + pipe_copy_rect(pipe_surface_map(dst), + dst->cpp, + dst->pitch, + dstx, dsty, + width, height, + pipe_surface_map(src), + do_flip ? -(int) src->pitch : src->pitch, + srcx, do_flip ? 1 - srcy - height : srcy); + + pipe_surface_unmap(src); + pipe_surface_unmap(dst); +} + + +static void * +get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) +{ + return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; +} + + +#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8)) + + +/** + * Fill a rectangular sub-region. Need better logic about when to + * push buffers into AGP - will currently do so whenever possible. + */ +static void +sp_surface_fill(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, unsigned value) +{ + unsigned i, j; + void *dst_map = pipe_surface_map(dst); + + assert(dst->pitch > 0); + assert(width <= dst->pitch); + + + switch (dst->cpp) { + case 1: + { + ubyte *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + memset(row, value, width); + row += dst->pitch; + } + } + break; + case 2: + { + ushort *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = (ushort) value; + row += dst->pitch; + } + } + break; + case 4: + { + unsigned *row = get_pointer(dst, dst_map, dstx, dsty); + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) + row[j] = value; + row += dst->pitch; + } + } + break; + case 8: + { + /* expand the 4-byte clear value to an 8-byte value */ + ushort *row = (ushort *) get_pointer(dst, dst_map, dstx, dsty); + ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); + ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); + ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); + ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff); + val0 = (val0 << 8) | val0; + val1 = (val1 << 8) | val1; + val2 = (val2 << 8) | val2; + val3 = (val3 << 8) | val3; + for (i = 0; i < height; i++) { + for (j = 0; j < width; j++) { + row[j*4+0] = val0; + row[j*4+1] = val1; + row[j*4+2] = val2; + row[j*4+3] = val3; + } + row += dst->pitch * 4; + } + } + break; + default: + assert(0); + break; + } + + pipe_surface_unmap( dst ); +} + + +void +sp_init_surface_functions(struct softpipe_context *sp) +{ + sp->pipe.surface_copy = sp_surface_copy; + sp->pipe.surface_fill = sp_surface_fill; +} diff --git a/src/gallium/drivers/softpipe/sp_surface.h b/src/gallium/drivers/softpipe/sp_surface.h new file mode 100644 index 0000000000..22de3ba43f --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_surface.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef SP_SURFACE_H +#define SP_SURFACE_H + + +struct softpipe_context; + + +extern void +sp_init_surface_functions(struct softpipe_context *sp); + + +#endif /* SP_SURFACE_H */ diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c new file mode 100644 index 0000000000..325bdb86da --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -0,0 +1,916 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Texture sampling + * + * Authors: + * Brian Paul + */ + +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_surface.h" +#include "sp_tex_sample.h" +#include "sp_tile_cache.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/tgsi/exec/tgsi_exec.h" + + +/* + * Note, the FRAC macro has to work perfectly. Otherwise you'll sometimes + * see 1-pixel bands of improperly weighted linear-filtered textures. + * The tests/texwrap.c demo is a good test. + * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0. + * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x). + */ +#define FRAC(f) ((f) - ifloor(f)) + + +/** + * Linear interpolation macro + */ +#define LERP(T, A, B) ( (A) + (T) * ((B) - (A)) ) + + +/** + * Do 2D/biliner interpolation of float values. + * v00, v10, v01 and v11 are typically four texture samples in a square/box. + * a and b are the horizontal and vertical interpolants. + * It's important that this function is inlined when compiled with + * optimization! If we find that's not true on some systems, convert + * to a macro. + */ +static INLINE float +lerp_2d(float a, float b, + float v00, float v10, float v01, float v11) +{ + const float temp0 = LERP(a, v00, v10); + const float temp1 = LERP(a, v01, v11); + return LERP(b, temp0, temp1); +} + + +/** + * If A is a signed integer, A % B doesn't give the right value for A < 0 + * (in terms of texture repeat). Just casting to unsigned fixes that. + */ +#define REMAINDER(A, B) ((unsigned) (A) % (unsigned) (B)) + + +/** + * Apply texture coord wrapping mode and return integer texture index. + * \param wrapMode PIPE_TEX_WRAP_x + * \param s the texcoord + * \param size the texture image size + * \return integer texture index + */ +static INLINE int +nearest_texcoord(unsigned wrapMode, float s, unsigned size) +{ + int i; + switch (wrapMode) { + case PIPE_TEX_WRAP_REPEAT: + /* s limited to [0,1) */ + /* i limited to [0,size-1] */ + i = ifloor(s * size); + i = REMAINDER(i, size); + return i; + case PIPE_TEX_WRAP_CLAMP: + /* s limited to [0,1] */ + /* i limited to [0,size-1] */ + if (s <= 0.0F) + i = 0; + else if (s >= 1.0F) + i = size - 1; + else + i = ifloor(s * size); + return i; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + { + /* s limited to [min,max] */ + /* i limited to [0, size-1] */ + const float min = 1.0F / (2.0F * size); + const float max = 1.0F - min; + if (s < min) + i = 0; + else if (s > max) + i = size - 1; + else + i = ifloor(s * size); + } + return i; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + { + /* s limited to [min,max] */ + /* i limited to [-1, size] */ + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; + if (s <= min) + i = -1; + else if (s >= max) + i = size; + else + i = ifloor(s * size); + } + return i; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + { + const float min = 1.0F / (2.0F * size); + const float max = 1.0F - min; + const int flr = ifloor(s); + float u; + if (flr & 1) + u = 1.0F - (s - (float) flr); + else + u = s - (float) flr; + if (u < min) + i = 0; + else if (u > max) + i = size - 1; + else + i = ifloor(u * size); + } + return i; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + { + /* s limited to [0,1] */ + /* i limited to [0,size-1] */ + const float u = FABSF(s); + if (u <= 0.0F) + i = 0; + else if (u >= 1.0F) + i = size - 1; + else + i = ifloor(u * size); + } + return i; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + { + /* s limited to [min,max] */ + /* i limited to [0, size-1] */ + const float min = 1.0F / (2.0F * size); + const float max = 1.0F - min; + const float u = FABSF(s); + if (u < min) + i = 0; + else if (u > max) + i = size - 1; + else + i = ifloor(u * size); + } + return i; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + { + /* s limited to [min,max] */ + /* i limited to [0, size-1] */ + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; + const float u = FABSF(s); + if (u < min) + i = -1; + else if (u > max) + i = size; + else + i = ifloor(u * size); + } + return i; + default: + assert(0); + return 0; + } +} + + +/** + * Used to compute texel locations for linear sampling. + * \param wrapMode PIPE_TEX_WRAP_x + * \param s the texcoord + * \param size the texture image size + * \param i0 returns first texture index + * \param i1 returns second texture index (usually *i0 + 1) + * \param a returns blend factor/weight between texture indexes + */ +static INLINE void +linear_texcoord(unsigned wrapMode, float s, unsigned size, + int *i0, int *i1, float *a) +{ + float u; + switch (wrapMode) { + case PIPE_TEX_WRAP_REPEAT: + u = s * size - 0.5F; + *i0 = REMAINDER(ifloor(u), size); + *i1 = REMAINDER(*i0 + 1, size); + break; + case PIPE_TEX_WRAP_CLAMP: + if (s <= 0.0F) + u = 0.0F; + else if (s >= 1.0F) + u = (float) size; + else + u = s * size; + u -= 0.5F; + *i0 = ifloor(u); + *i1 = *i0 + 1; + break; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + if (s <= 0.0F) + u = 0.0F; + else if (s >= 1.0F) + u = (float) size; + else + u = s * size; + u -= 0.5F; + *i0 = ifloor(u); + *i1 = *i0 + 1; + if (*i0 < 0) + *i0 = 0; + if (*i1 >= (int) size) + *i1 = size - 1; + break; + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + { + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; + if (s <= min) + u = min * size; + else if (s >= max) + u = max * size; + else + u = s * size; + u -= 0.5F; + *i0 = ifloor(u); + *i1 = *i0 + 1; + } + break; + case PIPE_TEX_WRAP_MIRROR_REPEAT: + { + const int flr = ifloor(s); + if (flr & 1) + u = 1.0F - (s - (float) flr); + else + u = s - (float) flr; + u = (u * size) - 0.5F; + *i0 = ifloor(u); + *i1 = *i0 + 1; + if (*i0 < 0) + *i0 = 0; + if (*i1 >= (int) size) + *i1 = size - 1; + } + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP: + u = FABSF(s); + if (u >= 1.0F) + u = (float) size; + else + u *= size; + u -= 0.5F; + *i0 = ifloor(u); + *i1 = *i0 + 1; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: + u = FABSF(s); + if (u >= 1.0F) + u = (float) size; + else + u *= size; + u -= 0.5F; + *i0 = ifloor(u); + *i1 = *i0 + 1; + if (*i0 < 0) + *i0 = 0; + if (*i1 >= (int) size) + *i1 = size - 1; + break; + case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: + { + const float min = -1.0F / (2.0F * size); + const float max = 1.0F - min; + u = FABSF(s); + if (u <= min) + u = min * size; + else if (u >= max) + u = max * size; + else + u *= size; + u -= 0.5F; + *i0 = ifloor(u); + *i1 = *i0 + 1; + } + break; + default: + assert(0); + } + *a = FRAC(u); +} + + +static unsigned +choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) +{ + /* + major axis + direction target sc tc ma + ---------- ------------------------------- --- --- --- + +rx TEXTURE_CUBE_MAP_POSITIVE_X_EXT -rz -ry rx + -rx TEXTURE_CUBE_MAP_NEGATIVE_X_EXT +rz -ry rx + +ry TEXTURE_CUBE_MAP_POSITIVE_Y_EXT +rx +rz ry + -ry TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT +rx -rz ry + +rz TEXTURE_CUBE_MAP_POSITIVE_Z_EXT +rx -ry rz + -rz TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT -rx -ry rz + */ + const float arx = FABSF(rx), ary = FABSF(ry), arz = FABSF(rz); + unsigned face; + float sc, tc, ma; + + if (arx > ary && arx > arz) { + if (rx >= 0.0F) { + face = PIPE_TEX_FACE_POS_X; + sc = -rz; + tc = -ry; + ma = arx; + } + else { + face = PIPE_TEX_FACE_NEG_X; + sc = rz; + tc = -ry; + ma = arx; + } + } + else if (ary > arx && ary > arz) { + if (ry >= 0.0F) { + face = PIPE_TEX_FACE_POS_Y; + sc = rx; + tc = rz; + ma = ary; + } + else { + face = PIPE_TEX_FACE_NEG_Y; + sc = rx; + tc = -rz; + ma = ary; + } + } + else { + if (rz > 0.0F) { + face = PIPE_TEX_FACE_POS_Z; + sc = rx; + tc = -ry; + ma = arz; + } + else { + face = PIPE_TEX_FACE_NEG_Z; + sc = -rx; + tc = -ry; + ma = arz; + } + } + + *newS = ( sc / ma + 1.0F ) * 0.5F; + *newT = ( tc / ma + 1.0F ) * 0.5F; + + return face; +} + + +/** + * Examine the quad's texture coordinates to compute the partial + * derivatives w.r.t X and Y, then compute lambda (level of detail). + * + * This is only done for fragment shaders, not vertex shaders. + */ +static float +compute_lambda(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias) +{ + float rho, lambda; + + assert(s); + { + float dsdx = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]; + float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; + dsdx = FABSF(dsdx); + dsdy = FABSF(dsdy); + rho = MAX2(dsdx, dsdy); + if (sampler->state->normalized_coords) + rho *= sampler->texture->width[0]; + } + if (t) { + float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; + float dtdy = t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]; + float max; + dtdx = FABSF(dtdx); + dtdy = FABSF(dtdy); + max = MAX2(dtdx, dtdy); + if (sampler->state->normalized_coords) + max *= sampler->texture->height[0]; + rho = MAX2(rho, max); + } + if (p) { + float dpdx = p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT]; + float dpdy = p[QUAD_TOP_LEFT] - p[QUAD_BOTTOM_LEFT]; + float max; + dpdx = FABSF(dpdx); + dpdy = FABSF(dpdy); + max = MAX2(dpdx, dpdy); + if (sampler->state->normalized_coords) + max *= sampler->texture->depth[0]; + rho = MAX2(rho, max); + } + + lambda = LOG2(rho); + lambda += lodbias + sampler->state->lod_bias; + lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); + + return lambda; +} + + +/** + * Do several things here: + * 1. Compute lambda from the texcoords, if needed + * 2. Determine if we're minifying or magnifying + * 3. If minifying, choose mipmap levels + * 4. Return image filter to use within mipmap images + */ +static void +choose_mipmap_levels(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + unsigned *level0, unsigned *level1, float *levelBlend, + unsigned *imgFilter) +{ + if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { + /* no mipmap selection needed */ + *imgFilter = sampler->state->mag_img_filter; + *level0 = *level1 = (int) sampler->state->min_lod; + } + else { + float lambda; + + if (1) + /* fragment shader */ + lambda = compute_lambda(sampler, s, t, p, lodbias); + else + /* vertex shader */ + lambda = lodbias; /* not really a bias, but absolute LOD */ + + if (lambda < 0.0) { /* XXX threshold depends on the filter */ + /* magnifying */ + *imgFilter = sampler->state->mag_img_filter; + *level0 = *level1 = 0; + } + else { + /* minifying */ + *imgFilter = sampler->state->min_img_filter; + + /* choose mipmap level(s) and compute the blend factor between them */ + if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NEAREST) { + /* Nearest mipmap level */ + const int lvl = (int) (lambda + 0.5); + *level0 = + *level1 = CLAMP(lvl, 0, (int) sampler->texture->last_level); + } + else { + /* Linear interpolation between mipmap levels */ + const int lvl = (int) lambda; + *level0 = CLAMP(lvl, 0, (int) sampler->texture->last_level); + *level1 = CLAMP(lvl + 1, 0, (int) sampler->texture->last_level); + *levelBlend = FRAC(lambda); /* blending weight between levels */ + } + } + } +} + + +/** + * Get a texel from a texture, using the texture tile cache. + * + * \param face the cube face in 0..5 + * \param level the mipmap level + * \param x the x coord of texel within 2D image + * \param y the y coord of texel within 2D image + * \param z which slice of a 3D texture + * \param rgba the quad to put the texel/color into + * \param j which element of the rgba quad to write to + * + * XXX maybe move this into sp_tile_cache.c and merge with the + * sp_get_cached_tile_tex() function. Also, get 4 texels instead of 1... + */ +static void +get_texel(struct tgsi_sampler *sampler, + unsigned face, unsigned level, int x, int y, int z, + float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j) +{ + const int tx = x % TILE_SIZE; + const int ty = y % TILE_SIZE; + const struct softpipe_cached_tile *tile + = sp_get_cached_tile_tex(sampler->pipe, sampler->cache, + x, y, z, face, level); + rgba[0][j] = tile->data.color[ty][tx][0]; + rgba[1][j] = tile->data.color[ty][tx][1]; + rgba[2][j] = tile->data.color[ty][tx][2]; + rgba[3][j] = tile->data.color[ty][tx][3]; +} + + +/** + * Compare texcoord 'p' (aka R) against texture value 'rgba[0]' + * When we sampled the depth texture, the depth value was put into all + * RGBA channels. We look at the red channel here. + */ +static INLINE void +shadow_compare(uint compare_func, + float rgba[NUM_CHANNELS][QUAD_SIZE], + const float p[QUAD_SIZE], + uint j) +{ + int k; + switch (compare_func) { + case PIPE_FUNC_LESS: + k = p[j] < rgba[0][j]; + break; + case PIPE_FUNC_LEQUAL: + k = p[j] <= rgba[0][j]; + break; + case PIPE_FUNC_GREATER: + k = p[j] > rgba[0][j]; + break; + case PIPE_FUNC_GEQUAL: + k = p[j] >= rgba[0][j]; + break; + case PIPE_FUNC_EQUAL: + k = p[j] == rgba[0][j]; + break; + case PIPE_FUNC_NOTEQUAL: + k = p[j] != rgba[0][j]; + break; + case PIPE_FUNC_ALWAYS: + k = 1; + break; + case PIPE_FUNC_NEVER: + k = 0; + break; + default: + assert(0); + } + + rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k; +} + + +/** + * Common code for sampling 1D/2D/cube textures. + * Could probably extend for 3D... + */ +static void +sp_get_samples_2d_common(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE], + const unsigned faces[4]) +{ + const uint compare_func = sampler->state->compare_func; + unsigned level0, level1, j, imgFilter; + int width, height; + float levelBlend; + + choose_mipmap_levels(sampler, s, t, p, lodbias, + &level0, &level1, &levelBlend, &imgFilter); + + if (sampler->state->normalized_coords) { + width = sampler->texture->width[level0]; + height = sampler->texture->height[level0]; + } + else { + width = height = 1; + } + + assert(width > 0); + + switch (imgFilter) { + case PIPE_TEX_FILTER_NEAREST: + for (j = 0; j < QUAD_SIZE; j++) { + int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); + int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); + get_texel(sampler, faces[j], level0, x, y, 0, rgba, j); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(compare_func, rgba, p, j); + } + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + unsigned c; + x = x / 2; + y = y / 2; + get_texel(sampler, faces[j], level1, x, y, 0, rgba2, j); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ + shadow_compare(compare_func, rgba2, p, j); + } + + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); + } + } + } + break; + case PIPE_TEX_FILTER_LINEAR: + for (j = 0; j < QUAD_SIZE; j++) { + float tx[4][4], a, b; + int x0, y0, x1, y1, c; + linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &a); + linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &b); + get_texel(sampler, faces[j], level0, x0, y0, 0, tx, 0); + get_texel(sampler, faces[j], level0, x1, y0, 0, tx, 1); + get_texel(sampler, faces[j], level0, x0, y1, 0, tx, 2); + get_texel(sampler, faces[j], level0, x1, y1, 0, tx, 3); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(compare_func, tx, p, 0); + shadow_compare(compare_func, tx, p, 1); + shadow_compare(compare_func, tx, p, 2); + shadow_compare(compare_func, tx, p, 3); + } + + for (c = 0; c < 4; c++) { + rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + } + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + x0 = x0 / 2; + y0 = y0 / 2; + x1 = x1 / 2; + y1 = y1 / 2; + get_texel(sampler, faces[j], level1, x0, y0, 0, tx, 0); + get_texel(sampler, faces[j], level1, x1, y0, 0, tx, 1); + get_texel(sampler, faces[j], level1, x0, y1, 0, tx, 2); + get_texel(sampler, faces[j], level1, x1, y1, 0, tx, 3); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ + shadow_compare(compare_func, tx, p, 0); + shadow_compare(compare_func, tx, p, 1); + shadow_compare(compare_func, tx, p, 2); + shadow_compare(compare_func, tx, p, 3); + } + + for (c = 0; c < 4; c++) { + rgba2[c][j] = lerp_2d(a, b, + tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + } + + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); + } + } + } + break; + default: + assert(0); + } +} + + +static void +sp_get_samples_1d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + static const unsigned faces[4] = {0, 0, 0, 0}; + static const float tzero[4] = {0, 0, 0, 0}; + sp_get_samples_2d_common(sampler, s, tzero, NULL, lodbias, rgba, faces); +} + + +static void +sp_get_samples_2d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + static const unsigned faces[4] = {0, 0, 0, 0}; + sp_get_samples_2d_common(sampler, s, t, p, lodbias, rgba, faces); +} + + +static void +sp_get_samples_3d(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + /* get/map pipe_surfaces corresponding to 3D tex slices */ + unsigned level0, level1, j, imgFilter; + int width, height, depth; + float levelBlend; + const uint face = 0; + + choose_mipmap_levels(sampler, s, t, p, lodbias, + &level0, &level1, &levelBlend, &imgFilter); + + if (sampler->state->normalized_coords) { + width = sampler->texture->width[level0]; + height = sampler->texture->height[level0]; + depth = sampler->texture->depth[level0]; + } + else { + width = height = depth = 1; + } + + assert(width > 0); + assert(height > 0); + assert(depth > 0); + + switch (imgFilter) { + case PIPE_TEX_FILTER_NEAREST: + for (j = 0; j < QUAD_SIZE; j++) { + int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); + int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); + int z = nearest_texcoord(sampler->state->wrap_r, p[j], depth); + get_texel(sampler, face, level0, x, y, z, rgba, j); + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + unsigned c; + x /= 2; + y /= 2; + z /= 2; + get_texel(sampler, face, level1, x, y, z, rgba2, j); + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = LERP(levelBlend, rgba2[c][j], rgba[c][j]); + } + } + } + break; + case PIPE_TEX_FILTER_LINEAR: + for (j = 0; j < QUAD_SIZE; j++) { + float texel0[4][4], texel1[4][4]; + float xw, yw, zw; /* interpolation weights */ + int x0, x1, y0, y1, z0, z1, c; + linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &xw); + linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &yw); + linear_texcoord(sampler->state->wrap_r, p[j], depth, &z0, &z1, &zw); + get_texel(sampler, face, level0, x0, y0, z0, texel0, 0); + get_texel(sampler, face, level0, x1, y0, z0, texel0, 1); + get_texel(sampler, face, level0, x0, y1, z0, texel0, 2); + get_texel(sampler, face, level0, x1, y1, z0, texel0, 3); + get_texel(sampler, face, level0, x0, y0, z1, texel1, 0); + get_texel(sampler, face, level0, x1, y0, z1, texel1, 1); + get_texel(sampler, face, level0, x0, y1, z1, texel1, 2); + get_texel(sampler, face, level0, x1, y1, z1, texel1, 3); + + /* 3D lerp */ + for (c = 0; c < 4; c++) { + float ctemp0[4][4], ctemp1[4][4]; + ctemp0[c][j] = lerp_2d(xw, yw, + texel0[c][0], texel0[c][1], + texel0[c][2], texel0[c][3]); + ctemp1[c][j] = lerp_2d(xw, yw, + texel1[c][0], texel1[c][1], + texel1[c][2], texel1[c][3]); + rgba[c][j] = LERP(zw, ctemp0[c][j], ctemp1[c][j]); + } + + if (level0 != level1) { + /* get texels from second mipmap level and blend */ + float rgba2[4][4]; + x0 /= 2; + y0 /= 2; + z0 /= 2; + x1 /= 2; + y1 /= 2; + z1 /= 2; + get_texel(sampler, face, level1, x0, y0, z0, texel0, 0); + get_texel(sampler, face, level1, x1, y0, z0, texel0, 1); + get_texel(sampler, face, level1, x0, y1, z0, texel0, 2); + get_texel(sampler, face, level1, x1, y1, z0, texel0, 3); + get_texel(sampler, face, level1, x0, y0, z1, texel1, 0); + get_texel(sampler, face, level1, x1, y0, z1, texel1, 1); + get_texel(sampler, face, level1, x0, y1, z1, texel1, 2); + get_texel(sampler, face, level1, x1, y1, z1, texel1, 3); + + /* 3D lerp */ + for (c = 0; c < 4; c++) { + float ctemp0[4][4], ctemp1[4][4]; + ctemp0[c][j] = lerp_2d(xw, yw, + texel0[c][0], texel0[c][1], + texel0[c][2], texel0[c][3]); + ctemp1[c][j] = lerp_2d(xw, yw, + texel1[c][0], texel1[c][1], + texel1[c][2], texel1[c][3]); + rgba2[c][j] = LERP(zw, ctemp0[c][j], ctemp1[c][j]); + } + + /* blend mipmap levels */ + for (c = 0; c < NUM_CHANNELS; c++) { + rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); + } + } + } + break; + default: + assert(0); + } +} + + +static void +sp_get_samples_cube(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + unsigned faces[QUAD_SIZE], j; + float ssss[4], tttt[4]; + for (j = 0; j < QUAD_SIZE; j++) { + faces[j] = choose_cube_face(s[j], t[j], p[j], ssss + j, tttt + j); + } + sp_get_samples_2d_common(sampler, ssss, tttt, NULL, lodbias, rgba, faces); +} + + +/** + * Called via tgsi_sampler::get_samples() + * Use the sampler's state setting to get a filtered RGBA value + * from the sampler's texture. + * + * XXX we can implement many versions of this function, each + * tightly coded for a specific combination of sampler state + * (nearest + repeat), (bilinear mipmap + clamp), etc. + * + * The update_samplers() function in st_atom_sampler.c could create + * a new tgsi_sampler object for each state combo it finds.... + */ +void +sp_get_samples(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + if (!sampler->texture) + return; + + switch (sampler->texture->target) { + case PIPE_TEXTURE_1D: + sp_get_samples_1d(sampler, s, t, p, lodbias, rgba); + break; + case PIPE_TEXTURE_2D: + sp_get_samples_2d(sampler, s, t, p, lodbias, rgba); + break; + case PIPE_TEXTURE_3D: + sp_get_samples_3d(sampler, s, t, p, lodbias, rgba); + break; + case PIPE_TEXTURE_CUBE: + sp_get_samples_cube(sampler, s, t, p, lodbias, rgba); + break; + default: + assert(0); + } +} + diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.h b/src/gallium/drivers/softpipe/sp_tex_sample.h new file mode 100644 index 0000000000..404bfd0c36 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_tex_sample.h @@ -0,0 +1,17 @@ +#ifndef SP_TEX_SAMPLE_H +#define SP_TEX_SAMPLE_H + + +struct tgsi_sampler; + + +extern void +sp_get_samples(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); + + +#endif /* SP_TEX_SAMPLE_H */ diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c new file mode 100644 index 0000000000..6de7a9b543 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -0,0 +1,166 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_texture.h" + + +/* Simple, maximally packed layout. + */ + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + + +static void +softpipe_texture_layout(struct softpipe_texture * spt) +{ + struct pipe_texture *pt = &spt->base; + unsigned level; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + + spt->buffer_size = 0; + + for (level = 0; level <= pt->last_level; level++) { + pt->width[level] = width; + pt->height[level] = height; + pt->depth[level] = depth; + + spt->level_offset[level] = spt->buffer_size; + + spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * + ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * + width * pt->cpp; + + width = minify(width); + height = minify(height); + depth = minify(depth); + } +} + + +struct pipe_texture * +softpipe_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat) +{ + struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); + if (!spt) + return NULL; + + spt->base = *templat; + + softpipe_texture_layout(spt); + + spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); + if (!spt->buffer) { + FREE(spt); + return NULL; + } + + return &spt->base; +} + + +void +softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct softpipe_texture *spt = softpipe_texture(*pt); + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); + */ + + pipe_buffer_reference(pipe->winsys, &spt->buffer, NULL); + + FREE(spt); + } + *pt = NULL; +} + + +/** + * Called via pipe->get_tex_surface() + */ +struct pipe_surface * +softpipe_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct softpipe_texture *spt = softpipe_texture(pt); + struct pipe_surface *ps; + + assert(level <= pt->last_level); + + ps = pipe->winsys->surface_alloc(pipe->winsys); + if (ps) { + assert(ps->refcount); + assert(ps->winsys); + pipe_buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = ps->width; + ps->offset = spt->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { + ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * + (pt->compressed ? ps->height/4 : ps->height) * + ps->width * ps->cpp; + } else { + assert(face == 0); + assert(zslice == 0); + } + } + return ps; +} diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h new file mode 100644 index 0000000000..fa646c0de9 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -0,0 +1,71 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_TEXTURE_H +#define SP_TEXTURE_H + + +struct pipe_context; +struct pipe_texture; + + +struct softpipe_texture +{ + struct pipe_texture base; + + unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; + unsigned long buffer_size; +}; + + +/** cast wrapper */ +static INLINE struct softpipe_texture * +softpipe_texture(struct pipe_texture *pt) +{ + return (struct softpipe_texture *) pt; +} + + + +extern struct pipe_texture * +softpipe_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat); + +extern void +softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); + +extern struct pipe_surface * +softpipe_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice); + + +#endif /* SP_TEXTURE */ diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c new file mode 100644 index 0000000000..1597361b82 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -0,0 +1,585 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Framebuffer/surface tile caching. + * + * Author: + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/util/p_tile.h" +#include "sp_context.h" +#include "sp_surface.h" +#include "sp_tile_cache.h" + +#define NUM_ENTRIES 30 + + +/** XXX move these */ +#define MAX_WIDTH 2048 +#define MAX_HEIGHT 2048 + + +struct softpipe_tile_cache +{ + struct pipe_surface *surface; /**< the surface we're caching */ + void *surface_map; + struct pipe_texture *texture; /**< if caching a texture */ + struct softpipe_cached_tile entries[NUM_ENTRIES]; + uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; + float clear_color[4]; + uint clear_val; + boolean depth_stencil; /** Is the surface a depth/stencil format? */ + + struct pipe_surface *tex_surf; + void *tex_surf_map; + int tex_face, tex_level, tex_z; + + struct softpipe_cached_tile tile; /**< scratch tile for clears */ +}; + + +/** + * Return the position in the cache for the tile that contains win pos (x,y). + * We currently use a direct mapped cache so this is like a hack key. + * At some point we should investige something more sophisticated, like + * a LRU replacement policy. + */ +#define CACHE_POS(x, y) \ + (((x) / TILE_SIZE + ((y) / TILE_SIZE) * 5) % NUM_ENTRIES) + + + +/** + * Is the tile at (x,y) in cleared state? + */ +static INLINE uint +is_clear_flag_set(const uint *bitvec, int x, int y) +{ + int pos, bit; + x /= TILE_SIZE; + y /= TILE_SIZE; + pos = y * (MAX_WIDTH / TILE_SIZE) + x; + assert(pos / 32 < (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32); + bit = bitvec[pos / 32] & (1 << (pos & 31)); + return bit; +} + + +/** + * Mark the tile at (x,y) as not cleared. + */ +static INLINE void +clear_clear_flag(uint *bitvec, int x, int y) +{ + int pos; + x /= TILE_SIZE; + y /= TILE_SIZE; + pos = y * (MAX_WIDTH / TILE_SIZE) + x; + assert(pos / 32 < (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32); + bitvec[pos / 32] &= ~(1 << (pos & 31)); +} + + +struct softpipe_tile_cache * +sp_create_tile_cache(void) +{ + struct softpipe_tile_cache *tc; + uint pos; + + tc = CALLOC_STRUCT( softpipe_tile_cache ); + if (tc) { + for (pos = 0; pos < NUM_ENTRIES; pos++) { + tc->entries[pos].x = + tc->entries[pos].y = -1; + } + } + return tc; +} + + +void +sp_destroy_tile_cache(struct softpipe_tile_cache *tc) +{ + uint pos; + + for (pos = 0; pos < NUM_ENTRIES; pos++) { + //assert(tc->entries[pos].x < 0); + } + if (tc->surface) { + pipe_surface_reference(&tc->surface, NULL); + } + if (tc->tex_surf) { + pipe_surface_reference(&tc->tex_surf, NULL); + } + + FREE( tc ); +} + + +/** + * Specify the surface to cache. + */ +void +sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, + struct pipe_surface *ps) +{ + assert(!tc->texture); + + if (tc->surface_map) { + /*assert(tc->surface != ps);*/ + pipe_surface_unmap(tc->surface); + } + + pipe_surface_reference(&tc->surface, ps); + + if (ps) { + if (tc->surface_map) + tc->surface_map = pipe_surface_map(ps); + + tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || + ps->format == PIPE_FORMAT_Z16_UNORM || + ps->format == PIPE_FORMAT_Z32_UNORM || + ps->format == PIPE_FORMAT_U_S8); + } +} + + +/** + * Return the surface being cached. + */ +struct pipe_surface * +sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) +{ + return tc->surface; +} + + +void +sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc) +{ + if (tc->surface && !tc->surface_map) + tc->surface_map = pipe_surface_map(tc->surface); + + if (tc->tex_surf && !tc->tex_surf_map) + tc->tex_surf_map = pipe_surface_map(tc->tex_surf); +} + + +void +sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc) +{ + if (tc->surface_map) { + pipe_surface_unmap(tc->surface); + tc->surface_map = NULL; + } + + if (tc->tex_surf_map) { + pipe_surface_unmap(tc->tex_surf); + tc->tex_surf_map = NULL; + } +} + + +/** + * Specify the texture to cache. + */ +void +sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, + struct pipe_texture *texture) +{ + uint i; + + assert(!tc->surface); + + tc->texture = texture; + + if (tc->tex_surf_map) { + pipe_surface_unmap(tc->tex_surf); + tc->tex_surf_map = NULL; + } + pipe_surface_reference(&tc->tex_surf, NULL); + + /* mark as entries as invalid/empty */ + /* XXX we should try to avoid this when the teximage hasn't changed */ + for (i = 0; i < NUM_ENTRIES; i++) { + tc->entries[i].x = -1; + } + + tc->tex_face = -1; /* any invalid value here */ +} + + +/** + * Set pixels in a tile to the given clear color/value, float. + */ +static void +clear_tile_rgba(struct softpipe_cached_tile *tile, + enum pipe_format format, + const float clear_value[4]) +{ + if (clear_value[0] == 0.0 && + clear_value[1] == 0.0 && + clear_value[2] == 0.0 && + clear_value[3] == 0.0) { + memset(tile->data.color, 0, sizeof(tile->data.color)); + } + else { + uint i, j; + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.color[i][j][0] = clear_value[0]; + tile->data.color[i][j][1] = clear_value[1]; + tile->data.color[i][j][2] = clear_value[2]; + tile->data.color[i][j][3] = clear_value[3]; + } + } + } +} + + +/** + * Set a tile to a solid value/color. + */ +static void +clear_tile(struct softpipe_cached_tile *tile, + enum pipe_format format, + uint clear_value) +{ + uint i, j; + + switch (pf_get_size(format)) { + case 1: + memset(tile->data.any, 0, TILE_SIZE * TILE_SIZE); + break; + case 2: + if (clear_value == 0) { + memset(tile->data.any, 0, 2 * TILE_SIZE * TILE_SIZE); + } + else { + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.depth16[i][j] = (ushort) clear_value; + } + } + } + break; + case 4: + if (clear_value == 0) { + memset(tile->data.any, 0, 4 * TILE_SIZE * TILE_SIZE); + } + else { + for (i = 0; i < TILE_SIZE; i++) { + for (j = 0; j < TILE_SIZE; j++) { + tile->data.color32[i][j] = clear_value; + } + } + } + break; + default: + assert(0); + } +} + + +/** + * Actually clear the tiles which were flagged as being in a clear state. + */ +static void +sp_tile_cache_flush_clear(struct pipe_context *pipe, + struct softpipe_tile_cache *tc) +{ + struct pipe_surface *ps = tc->surface; + const uint w = tc->surface->width; + const uint h = tc->surface->height; + uint x, y; + uint numCleared = 0; + + /* clear the scratch tile to the clear value */ + clear_tile(&tc->tile, ps->format, tc->clear_val); + + /* push the tile to all positions marked as clear */ + for (y = 0; y < h; y += TILE_SIZE) { + for (x = 0; x < w; x += TILE_SIZE) { + if (is_clear_flag_set(tc->clear_flags, x, y)) { + pipe_put_tile_raw(pipe, ps, + x, y, TILE_SIZE, TILE_SIZE, + tc->tile.data.color32, 0/*STRIDE*/); + + /* do this? */ + clear_clear_flag(tc->clear_flags, x, y); + + numCleared++; + } + } + } +#if 0 + debug_printf("num cleared: %u\n", numCleared); +#endif +} + + +/** + * Flush the tile cache: write all dirty tiles back to the surface. + * any tiles "flagged" as cleared will be "really" cleared. + */ +void +sp_flush_tile_cache(struct softpipe_context *softpipe, + struct softpipe_tile_cache *tc) +{ + struct pipe_context *pipe = &softpipe->pipe; + struct pipe_surface *ps = tc->surface; + int inuse = 0, pos; + + if (!ps || !ps->buffer) + return; + + for (pos = 0; pos < NUM_ENTRIES; pos++) { + struct softpipe_cached_tile *tile = tc->entries + pos; + if (tile->x >= 0) { + if (tc->depth_stencil) { + pipe_put_tile_raw(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); + } + else { + pipe_put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } + tile->x = tile->y = -1; /* mark as empty */ + inuse++; + } + } + +#if TILE_CLEAR_OPTIMIZATION + sp_tile_cache_flush_clear(&softpipe->pipe, tc); +#endif + +#if 0 + debug_printf("flushed tiles in use: %d\n", inuse); +#endif +} + + +/** + * Get a tile from the cache. + * \param x, y position of tile, in pixels + */ +struct softpipe_cached_tile * +sp_get_cached_tile(struct softpipe_context *softpipe, + struct softpipe_tile_cache *tc, int x, int y) +{ + struct pipe_context *pipe = &softpipe->pipe; + struct pipe_surface *ps = tc->surface; + + /* tile pos in framebuffer: */ + const int tile_x = x & ~(TILE_SIZE - 1); + const int tile_y = y & ~(TILE_SIZE - 1); + + /* cache pos/entry: */ + const int pos = CACHE_POS(x, y); + struct softpipe_cached_tile *tile = tc->entries + pos; + + if (tile_x != tile->x || + tile_y != tile->y) { + + if (tile->x != -1) { + /* put dirty tile back in framebuffer */ + if (tc->depth_stencil) { + pipe_put_tile_raw(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); + } + else { + pipe_put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } + } + + tile->x = tile_x; + tile->y = tile_y; + + if (is_clear_flag_set(tc->clear_flags, x, y)) { + /* don't get tile from framebuffer, just clear it */ + if (tc->depth_stencil) { + clear_tile(tile, ps->format, tc->clear_val); + } + else { + clear_tile_rgba(tile, ps->format, tc->clear_color); + } + clear_clear_flag(tc->clear_flags, x, y); + } + else { + /* get new tile data from surface */ + if (tc->depth_stencil) { + pipe_get_tile_raw(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); + } + else { + pipe_get_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } + } + } + + return tile; +} + + +/** + * Given the texture face, level, zslice, x and y values, compute + * the cache entry position/index where we'd hope to find the + * cached texture tile. + * This is basically a direct-map cache. + * XXX There's probably lots of ways in which we can improve this. + */ +static INLINE uint +tex_cache_pos(int x, int y, int z, int face, int level) +{ + uint entry = x + y * 5 + z * 4 + face + level; + return entry % NUM_ENTRIES; +} + + +/** + * Similar to sp_get_cached_tile() but for textures. + * Tiles are read-only and indexed with more params. + */ +const struct softpipe_cached_tile * +sp_get_cached_tile_tex(struct pipe_context *pipe, + struct softpipe_tile_cache *tc, int x, int y, int z, + int face, int level) +{ + /* tile pos in framebuffer: */ + const int tile_x = x & ~(TILE_SIZE - 1); + const int tile_y = y & ~(TILE_SIZE - 1); + /* cache pos/entry: */ + const uint pos = tex_cache_pos(x / TILE_SIZE, y / TILE_SIZE, z, + face, level); + struct softpipe_cached_tile *tile = tc->entries + pos; + + if (tile_x != tile->x || + tile_y != tile->y || + z != tile->z || + face != tile->face || + level != tile->level) { + /* cache miss */ + + /* check if we need to get a new surface */ + if (!tc->tex_surf || + tc->tex_face != face || + tc->tex_level != level || + tc->tex_z != z) { + /* get new surface (view into texture) */ + + if (tc->tex_surf_map) + pipe_surface_unmap(tc->tex_surf); + + tc->tex_surf = pipe->get_tex_surface(pipe, tc->texture, face, level, z); + tc->tex_surf_map = pipe_surface_map(tc->tex_surf); + + tc->tex_face = face; + tc->tex_level = level; + tc->tex_z = z; + } + + /* get tile from the surface (view into texture) */ + pipe_get_tile_rgba(pipe, tc->tex_surf, + tile_x, tile_y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + tile->x = tile_x; + tile->y = tile_y; + tile->z = z; + tile->face = face; + tile->level = level; + } + + return tile; +} + + +/** + * When a whole surface is being cleared to a value we can avoid + * fetching tiles above. + * Save the color and set a 'clearflag' for each tile of the screen. + */ +void +sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue) +{ + uint r, g, b, a; + uint pos; + + tc->clear_val = clearValue; + + switch (tc->surface->format) { + case PIPE_FORMAT_R8G8B8A8_UNORM: + r = (clearValue >> 24) & 0xff; + g = (clearValue >> 16) & 0xff; + b = (clearValue >> 8) & 0xff; + a = (clearValue ) & 0xff; + break; + case PIPE_FORMAT_A8R8G8B8_UNORM: + r = (clearValue >> 16) & 0xff; + g = (clearValue >> 8) & 0xff; + b = (clearValue ) & 0xff; + a = (clearValue >> 24) & 0xff; + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + r = (clearValue >> 8) & 0xff; + g = (clearValue >> 16) & 0xff; + b = (clearValue >> 24) & 0xff; + a = (clearValue ) & 0xff; + break; + default: + r = g = b = a = 0; + } + + tc->clear_color[0] = r / 255.0f; + tc->clear_color[1] = g / 255.0f; + tc->clear_color[2] = b / 255.0f; + tc->clear_color[3] = a / 255.0f; + +#if TILE_CLEAR_OPTIMIZATION + /* set flags to indicate all the tiles are cleared */ + memset(tc->clear_flags, 255, sizeof(tc->clear_flags)); +#else + /* disable the optimization */ + memset(tc->clear_flags, 0, sizeof(tc->clear_flags)); +#endif + + for (pos = 0; pos < NUM_ENTRIES; pos++) { + struct softpipe_cached_tile *tile = tc->entries + pos; + tile->x = tile->y = -1; + } +} diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.h b/src/gallium/drivers/softpipe/sp_tile_cache.h new file mode 100644 index 0000000000..7fd1081286 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_tile_cache.h @@ -0,0 +1,104 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef SP_TILE_CACHE_H +#define SP_TILE_CACHE_H + +#define TILE_CLEAR_OPTIMIZATION 1 + + +#include "pipe/p_compiler.h" + + +struct softpipe_context; +struct softpipe_tile_cache; + + +/** + * Cache tile size (width and height). This needs to be a power of two. + */ +#define TILE_SIZE 64 + + + +struct softpipe_cached_tile +{ + int x, y; /**< pos of tile in window coords */ + int z, face, level; /**< Extra texture indexes */ + union { + float color[TILE_SIZE][TILE_SIZE][4]; + uint color32[TILE_SIZE][TILE_SIZE]; + uint depth32[TILE_SIZE][TILE_SIZE]; + ushort depth16[TILE_SIZE][TILE_SIZE]; + ubyte stencil8[TILE_SIZE][TILE_SIZE]; + ubyte any[1]; + } data; +}; + + +extern struct softpipe_tile_cache * +sp_create_tile_cache(void); + +extern void +sp_destroy_tile_cache(struct softpipe_tile_cache *tc); + +extern void +sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, + struct pipe_surface *sps); + +extern struct pipe_surface * +sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); + +extern void +sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc); + +extern void +sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc); + +extern void +sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, + struct pipe_texture *texture); + +extern void +sp_flush_tile_cache(struct softpipe_context *softpipe, + struct softpipe_tile_cache *tc); + +extern void +sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue); + +extern struct softpipe_cached_tile * +sp_get_cached_tile(struct softpipe_context *softpipe, + struct softpipe_tile_cache *tc, int x, int y); + +extern const struct softpipe_cached_tile * +sp_get_cached_tile_tex(struct pipe_context *pipe, + struct softpipe_tile_cache *tc, int x, int y, int z, + int face, int level); + + +#endif /* SP_TILE_CACHE_H */ + diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h new file mode 100644 index 0000000000..d6b379f58c --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_winsys.h @@ -0,0 +1,57 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* This is the interface that softpipe requires any window system + * hosting it to implement. This is the only include file in softpipe + * which is public. + */ + + +#ifndef SP_WINSYS_H +#define SP_WINSYS_H + + +#include "pipe/p_compiler.h" /* for boolean */ + +enum pipe_format; + +struct softpipe_winsys { + /** test if the given format is supported for front/back color bufs */ + boolean (*is_format_supported)( struct softpipe_winsys *sws, + enum pipe_format format ); + +}; + +struct pipe_winsys; +struct pipe_context; + + +struct pipe_context *softpipe_create( struct pipe_winsys *, + struct softpipe_winsys * ); + + +#endif /* SP_WINSYS_H */ diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h new file mode 100644 index 0000000000..30cd729c56 --- /dev/null +++ b/src/gallium/include/pipe/p_compiler.h @@ -0,0 +1,116 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_COMPILER_H +#define P_COMPILER_H + + +#include +#include + + +#if defined(_WIN32) && !defined(__WIN32__) +#define __WIN32__ +#endif + +#if defined(_MSC_VER) && !defined(__MSC__) +#define __MSC__ +#endif + + +typedef unsigned int uint; +typedef unsigned char ubyte; +typedef unsigned char boolean; +typedef unsigned short ushort; +typedef unsigned long long uint64; + + +#if defined(__MSC__) + +typedef char int8_t; +typedef unsigned char uint8_t; +typedef short int16_t; +typedef unsigned short uint16_t; +typedef long int32_t; +typedef unsigned long uint32_t; +typedef long long int64_t; +typedef unsigned long long uint64_t; + +#if defined(_WIN64) +typedef unsigned __int64 uintptr_t; +#else +typedef unsigned int uintptr_t; +#endif + +#else +#include +#endif + + +#define TRUE 1 +#define FALSE 0 + + +/* Function inlining */ +#if defined(__GNUC__) +# define INLINE __inline__ +#elif defined(__MSC__) +# define INLINE __inline +#elif defined(__ICL) +# define INLINE __inline +#elif defined(__INTEL_COMPILER) +# define INLINE inline +#elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100) +# define INLINE __inline +#else +# define INLINE +#endif + + +#if defined __GNUC__ +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___aligned[SIZE] __attribute__(( aligned( 16 ) )) +#define ALIGN16_ASSIGN(NAME) NAME##___aligned +#define ALIGN16_ATTRIB __attribute__(( aligned( 16 ) )) +#else +#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___unaligned[SIZE + 1] +#define ALIGN16_ASSIGN(NAME) align16(NAME##___unaligned) +#define ALIGN16_ATTRIB +#endif + + + +/** For calling code-gen'd functions */ +#if !defined(XSTDCALL) +#if defined(WIN32) +#define XSTDCALL __stdcall +#else +#define XSTDCALL +#endif +#endif + + +#endif /* P_COMPILER_H */ diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h new file mode 100644 index 0000000000..39f95695fb --- /dev/null +++ b/src/gallium/include/pipe/p_context.h @@ -0,0 +1,221 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef PIPE_CONTEXT_H +#define PIPE_CONTEXT_H + +#include "p_state.h" + + +struct pipe_state_cache; + +/* Opaque driver handles: + */ +struct pipe_query; + +/** + * Gallium rendering context. Basically: + * - state setting functions + * - VBO drawing functions + * - surface functions + * - device queries + */ +struct pipe_context { + struct pipe_winsys *winsys; + + void *priv; /** context private data (for DRI for example) */ + + void (*destroy)( struct pipe_context * ); + + /* + * Queries + */ + /** type is one of PIPE_SURFACE, PIPE_TEXTURE, etc. */ + boolean (*is_format_supported)( struct pipe_context *pipe, + enum pipe_format format, uint type ); + + const char *(*get_name)( struct pipe_context *pipe ); + + const char *(*get_vendor)( struct pipe_context *pipe ); + + int (*get_param)( struct pipe_context *pipe, int param ); + float (*get_paramf)( struct pipe_context *pipe, int param ); + + + /* + * Drawing. + * Return false on fallbacks (temporary??) + */ + boolean (*draw_arrays)( struct pipe_context *pipe, + unsigned mode, unsigned start, unsigned count); + + boolean (*draw_elements)( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); + + + /** + * Query objects + */ + struct pipe_query *(*create_query)( struct pipe_context *pipe, + unsigned query_type ); + + void (*destroy_query)(struct pipe_context *pipe, + struct pipe_query *q); + + void (*begin_query)(struct pipe_context *pipe, struct pipe_query *q); + void (*end_query)(struct pipe_context *pipe, struct pipe_query *q); + + boolean (*get_query_result)(struct pipe_context *pipe, + struct pipe_query *q, + boolean wait, + uint64 *result); + + /* + * State functions + */ + void * (*create_blend_state)(struct pipe_context *, + const struct pipe_blend_state *); + void (*bind_blend_state)(struct pipe_context *, void *); + void (*delete_blend_state)(struct pipe_context *, void *); + + void * (*create_sampler_state)(struct pipe_context *, + const struct pipe_sampler_state *); + void (*bind_sampler_state)(struct pipe_context *, unsigned unit, void *); + void (*delete_sampler_state)(struct pipe_context *, void *); + + void * (*create_rasterizer_state)(struct pipe_context *, + const struct pipe_rasterizer_state *); + void (*bind_rasterizer_state)(struct pipe_context *, void *); + void (*delete_rasterizer_state)(struct pipe_context *, void *); + + void * (*create_depth_stencil_alpha_state)(struct pipe_context *, + const struct pipe_depth_stencil_alpha_state *); + void (*bind_depth_stencil_alpha_state)(struct pipe_context *, void *); + void (*delete_depth_stencil_alpha_state)(struct pipe_context *, void *); + + void * (*create_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*bind_fs_state)(struct pipe_context *, void *); + void (*delete_fs_state)(struct pipe_context *, void *); + + void * (*create_vs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*bind_vs_state)(struct pipe_context *, void *); + void (*delete_vs_state)(struct pipe_context *, void *); + + /* The following look more properties than states. + * maybe combine a few of them into states or pass them + * in the bind calls to the state */ + void (*set_blend_color)( struct pipe_context *, + const struct pipe_blend_color * ); + + void (*set_clip_state)( struct pipe_context *, + const struct pipe_clip_state * ); + + void (*set_constant_buffer)( struct pipe_context *, + uint shader, uint index, + const struct pipe_constant_buffer *buf ); + + void (*set_framebuffer_state)( struct pipe_context *, + const struct pipe_framebuffer_state * ); + + void (*set_polygon_stipple)( struct pipe_context *, + const struct pipe_poly_stipple * ); + + void (*set_scissor_state)( struct pipe_context *, + const struct pipe_scissor_state * ); + + + /* Currently a sampler is constrained to sample from a single texture: + */ + void (*set_sampler_texture)( struct pipe_context *, + unsigned sampler, + struct pipe_texture * ); + + void (*set_viewport_state)( struct pipe_context *, + const struct pipe_viewport_state * ); + + /* + * Vertex arrays + */ + void (*set_vertex_buffer)( struct pipe_context *, + unsigned index, + const struct pipe_vertex_buffer * ); + + void (*set_vertex_element)( struct pipe_context *, + unsigned index, + const struct pipe_vertex_element * ); + + + /* + * Surface functions + */ + + void (*surface_copy)(struct pipe_context *pipe, + unsigned do_flip, /*<< flip surface contents vertically */ + struct pipe_surface *dest, + unsigned destx, unsigned desty, + struct pipe_surface *src, /* don't make this const - + need to map/unmap */ + unsigned srcx, unsigned srcy, + unsigned width, unsigned height); + + void (*surface_fill)(struct pipe_context *pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, + unsigned value); + + void (*clear)(struct pipe_context *pipe, + struct pipe_surface *ps, + unsigned clearValue); + + + /* + * Texture functions + */ + struct pipe_texture * (*texture_create)(struct pipe_context *pipe, + const struct pipe_texture *templat); + + void (*texture_release)(struct pipe_context *pipe, + struct pipe_texture **pt); + + /** Get a surface which is a "view" into a texture */ + struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, + struct pipe_texture *texture, + unsigned face, unsigned level, + unsigned zslice); + + /* Flush rendering: + */ + void (*flush)( struct pipe_context *pipe, + unsigned flags ); +}; + +#endif /* PIPE_CONTEXT_H */ diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h new file mode 100644 index 0000000000..2a11627b36 --- /dev/null +++ b/src/gallium/include/pipe/p_debug.h @@ -0,0 +1,86 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Cross-platform debugging helpers. + * + * For now it just has assert and printf replacements, but it might be extended + * with stack trace reports and more advanced logging in the near future. + * + * @author Jose Fonseca + */ + +#ifndef P_DEBUG_H_ +#define P_DEBUG_H_ + + +#include + + +#ifdef __cplusplus +extern "C" { +#endif + + +#ifdef DBG +#ifndef DEBUG +#define DEBUG 1 +#endif +#else +#ifndef NDEBUG +#define NDEBUG 1 +#endif +#endif + + +void debug_printf(const char *format, ...); + +void debug_vprintf(const char *format, va_list ap); + +void debug_assert_fail(const char *expr, const char *file, unsigned line); + + +/** Assert macro */ +#ifdef DEBUG +#define debug_assert(expr) ((expr) ? (void)0 : debug_assert_fail(#expr, __FILE__, __LINE__)) +#else +#define debug_assert(expr) ((void)0) +#endif + + +#ifdef assert +#undef assert +#endif +#define assert(expr) debug_assert(expr) + + +#ifdef __cplusplus +} +#endif + +#endif /* P_DEBUG_H_ */ diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h new file mode 100644 index 0000000000..0bf53ecb79 --- /dev/null +++ b/src/gallium/include/pipe/p_defines.h @@ -0,0 +1,270 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef PIPE_DEFINES_H +#define PIPE_DEFINES_H + +#include "p_format.h" + +#define PIPE_BLENDFACTOR_ONE 0x1 +#define PIPE_BLENDFACTOR_SRC_COLOR 0x2 +#define PIPE_BLENDFACTOR_SRC_ALPHA 0x3 +#define PIPE_BLENDFACTOR_DST_ALPHA 0x4 +#define PIPE_BLENDFACTOR_DST_COLOR 0x5 +#define PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 +#define PIPE_BLENDFACTOR_CONST_COLOR 0x7 +#define PIPE_BLENDFACTOR_CONST_ALPHA 0x8 +#define PIPE_BLENDFACTOR_SRC1_COLOR 0x9 +#define PIPE_BLENDFACTOR_SRC1_ALPHA 0x0A +#define PIPE_BLENDFACTOR_ZERO 0x11 +#define PIPE_BLENDFACTOR_INV_SRC_COLOR 0x12 +#define PIPE_BLENDFACTOR_INV_SRC_ALPHA 0x13 +#define PIPE_BLENDFACTOR_INV_DST_ALPHA 0x14 +#define PIPE_BLENDFACTOR_INV_DST_COLOR 0x15 +#define PIPE_BLENDFACTOR_INV_CONST_COLOR 0x17 +#define PIPE_BLENDFACTOR_INV_CONST_ALPHA 0x18 +#define PIPE_BLENDFACTOR_INV_SRC1_COLOR 0x19 +#define PIPE_BLENDFACTOR_INV_SRC1_ALPHA 0x1A + +#define PIPE_BLEND_ADD 0 +#define PIPE_BLEND_SUBTRACT 1 +#define PIPE_BLEND_REVERSE_SUBTRACT 2 +#define PIPE_BLEND_MIN 3 +#define PIPE_BLEND_MAX 4 + +#define PIPE_LOGICOP_CLEAR 0 +#define PIPE_LOGICOP_NOR 1 +#define PIPE_LOGICOP_AND_INVERTED 2 +#define PIPE_LOGICOP_COPY_INVERTED 3 +#define PIPE_LOGICOP_AND_REVERSE 4 +#define PIPE_LOGICOP_INVERT 5 +#define PIPE_LOGICOP_XOR 6 +#define PIPE_LOGICOP_NAND 7 +#define PIPE_LOGICOP_AND 8 +#define PIPE_LOGICOP_EQUIV 9 +#define PIPE_LOGICOP_NOOP 10 +#define PIPE_LOGICOP_OR_INVERTED 11 +#define PIPE_LOGICOP_COPY 12 +#define PIPE_LOGICOP_OR_REVERSE 13 +#define PIPE_LOGICOP_OR 14 +#define PIPE_LOGICOP_SET 15 + +#define PIPE_MASK_R 0x1 +#define PIPE_MASK_G 0x2 +#define PIPE_MASK_B 0x4 +#define PIPE_MASK_A 0x8 +#define PIPE_MASK_RGBA 0xf + + +/** + * Inequality functions. Used for depth test, stencil compare, alpha + * test, shadow compare, etc. + */ +#define PIPE_FUNC_NEVER 0 +#define PIPE_FUNC_LESS 1 +#define PIPE_FUNC_EQUAL 2 +#define PIPE_FUNC_LEQUAL 3 +#define PIPE_FUNC_GREATER 4 +#define PIPE_FUNC_NOTEQUAL 5 +#define PIPE_FUNC_GEQUAL 6 +#define PIPE_FUNC_ALWAYS 7 + +/** Polygon fill mode */ +#define PIPE_POLYGON_MODE_FILL 0 +#define PIPE_POLYGON_MODE_LINE 1 +#define PIPE_POLYGON_MODE_POINT 2 + +/** Polygon front/back window, also for culling */ +#define PIPE_WINDING_NONE 0 +#define PIPE_WINDING_CW 1 +#define PIPE_WINDING_CCW 2 +#define PIPE_WINDING_BOTH (PIPE_WINDING_CW | PIPE_WINDING_CCW) + +/** Stencil ops */ +#define PIPE_STENCIL_OP_KEEP 0 +#define PIPE_STENCIL_OP_ZERO 1 +#define PIPE_STENCIL_OP_REPLACE 2 +#define PIPE_STENCIL_OP_INCR 3 +#define PIPE_STENCIL_OP_DECR 4 +#define PIPE_STENCIL_OP_INCR_WRAP 5 +#define PIPE_STENCIL_OP_DECR_WRAP 6 +#define PIPE_STENCIL_OP_INVERT 7 + +/** Texture types */ +enum pipe_texture_target { + PIPE_TEXTURE_1D = 0, + PIPE_TEXTURE_2D = 1, + PIPE_TEXTURE_3D = 2, + PIPE_TEXTURE_CUBE = 3 +}; + +#define PIPE_TEX_FACE_POS_X 0 +#define PIPE_TEX_FACE_NEG_X 1 +#define PIPE_TEX_FACE_POS_Y 2 +#define PIPE_TEX_FACE_NEG_Y 3 +#define PIPE_TEX_FACE_POS_Z 4 +#define PIPE_TEX_FACE_NEG_Z 5 + +#define PIPE_TEX_WRAP_REPEAT 0 +#define PIPE_TEX_WRAP_CLAMP 1 +#define PIPE_TEX_WRAP_CLAMP_TO_EDGE 2 +#define PIPE_TEX_WRAP_CLAMP_TO_BORDER 3 +#define PIPE_TEX_WRAP_MIRROR_REPEAT 4 +#define PIPE_TEX_WRAP_MIRROR_CLAMP 5 +#define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE 6 +#define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER 7 + +/* Between mipmaps, ie mipfilter + */ +#define PIPE_TEX_MIPFILTER_NEAREST 0 +#define PIPE_TEX_MIPFILTER_LINEAR 1 +#define PIPE_TEX_MIPFILTER_NONE 2 + +/* Within a mipmap, ie min/mag filter + */ +#define PIPE_TEX_FILTER_NEAREST 0 +#define PIPE_TEX_FILTER_LINEAR 1 +//#define PIPE_TEX_FILTER_ANISO 2 + + +#define PIPE_TEX_COMPARE_NONE 0 +#define PIPE_TEX_COMPARE_R_TO_TEXTURE 1 + +#define PIPE_TEX_FACE_POS_X 0 +#define PIPE_TEX_FACE_NEG_X 1 +#define PIPE_TEX_FACE_POS_Y 2 +#define PIPE_TEX_FACE_NEG_Y 3 +#define PIPE_TEX_FACE_POS_Z 4 +#define PIPE_TEX_FACE_NEG_Z 5 +#define PIPE_TEX_FACE_MAX 6 + +/** + * Surfaces, textures, etc. (others may be added) + */ +#define PIPE_TEXTURE 1 +#define PIPE_SURFACE 2 /**< user-created surfaces */ + + +/** + * Surface status + */ +#define PIPE_SURFACE_STATUS_UNDEFINED 0 +#define PIPE_SURFACE_STATUS_DEFINED 1 +#define PIPE_SURFACE_STATUS_CLEAR 2 + + +/** + * Buffer usage flags + */ +#define PIPE_BUFFER_USAGE_CPU_READ (1 << 0) +#define PIPE_BUFFER_USAGE_CPU_WRITE (1 << 1) +#define PIPE_BUFFER_USAGE_GPU_READ (1 << 2) +#define PIPE_BUFFER_USAGE_GPU_WRITE (1 << 3) +#define PIPE_BUFFER_USAGE_PIXEL (1 << 4) +#define PIPE_BUFFER_USAGE_VERTEX (1 << 5) +#define PIPE_BUFFER_USAGE_INDEX (1 << 6) +#define PIPE_BUFFER_USAGE_CONSTANT (1 << 7) +/** Pipe driver custam usage flags should be greater or equal to this value */ +#define PIPE_BUFFER_USAGE_CUSTOM (1 << 16) + + +/** + * Flush types: + */ +#define PIPE_FLUSH_RENDER_CACHE 0x1 +#define PIPE_FLUSH_TEXTURE_CACHE 0x2 +#define PIPE_FLUSH_WAIT 0x4 +#define PIPE_FLUSH_SWAPBUFFERS 0x8 + + +/** + * Shaders + */ +#define PIPE_SHADER_VERTEX 0 +#define PIPE_SHADER_FRAGMENT 1 +#define PIPE_SHADER_TYPES 2 + + +/** + * Primitive types: + */ +#define PIPE_PRIM_POINTS 0 +#define PIPE_PRIM_LINES 1 +#define PIPE_PRIM_LINE_LOOP 2 +#define PIPE_PRIM_LINE_STRIP 3 +#define PIPE_PRIM_TRIANGLES 4 +#define PIPE_PRIM_TRIANGLE_STRIP 5 +#define PIPE_PRIM_TRIANGLE_FAN 6 +#define PIPE_PRIM_QUADS 7 +#define PIPE_PRIM_QUAD_STRIP 8 +#define PIPE_PRIM_POLYGON 9 + + +/** + * Query object types + */ +#define PIPE_QUERY_OCCLUSION_COUNTER 0 +#define PIPE_QUERY_PRIMITIVES_GENERATED 1 +#define PIPE_QUERY_PRIMITIVES_EMITTED 2 +#define PIPE_QUERY_TYPES 3 + + +/** + * Point sprite coord modes + */ +#define PIPE_SPRITE_COORD_NONE 0 +#define PIPE_SPRITE_COORD_UPPER_LEFT 1 +#define PIPE_SPRITE_COORD_LOWER_LEFT 2 + + +/** + * Implementation capabilities/limits + * Passed to pipe->get_param() + * XXX this will need some fine tuning... + */ +#define PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS 1 +#define PIPE_CAP_NPOT_TEXTURES 2 +#define PIPE_CAP_TWO_SIDED_STENCIL 3 +#define PIPE_CAP_GLSL 4 /* XXX need something better */ +#define PIPE_CAP_S3TC 5 +#define PIPE_CAP_ANISOTROPIC_FILTER 6 +#define PIPE_CAP_POINT_SPRITE 7 +#define PIPE_CAP_MAX_RENDER_TARGETS 8 +#define PIPE_CAP_OCCLUSION_QUERY 9 +#define PIPE_CAP_TEXTURE_SHADOW_MAP 10 +#define PIPE_CAP_MAX_TEXTURE_2D_LEVELS 11 +#define PIPE_CAP_MAX_TEXTURE_3D_LEVELS 12 +#define PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS 13 +#define PIPE_CAP_MAX_LINE_WIDTH 14 +#define PIPE_CAP_MAX_LINE_WIDTH_AA 15 +#define PIPE_CAP_MAX_POINT_WIDTH 16 +#define PIPE_CAP_MAX_POINT_WIDTH_AA 17 +#define PIPE_CAP_MAX_TEXTURE_ANISOTROPY 18 +#define PIPE_CAP_MAX_TEXTURE_LOD_BIAS 19 +#define PIPE_CAP_BITMAP_TEXCOORD_BIAS 20 + +#endif diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h new file mode 100644 index 0000000000..c9ad324315 --- /dev/null +++ b/src/gallium/include/pipe/p_format.h @@ -0,0 +1,421 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef PIPE_FORMAT_H +#define PIPE_FORMAT_H + +#include // for sprintf + +#include "p_compiler.h" +#include "p_debug.h" + +/** + * The PIPE_FORMAT is a 32-bit wide bitfield that encodes all the information + * needed to uniquely describe a pixel format. + */ + +/** + * Possible format layouts -- occupy first 2 bits. The interpretation of + * the remaining 30 bits depends on a particual format layout. + */ +#define PIPE_FORMAT_LAYOUT_RGBAZS 0 +#define PIPE_FORMAT_LAYOUT_YCBCR 1 + +static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ +{ + return f & 0x3; +} + +/** + * RGBAZS Format Layout. + */ + +/** + * Format component selectors for RGBAZS layout. + */ +#define PIPE_FORMAT_COMP_R 0 +#define PIPE_FORMAT_COMP_G 1 +#define PIPE_FORMAT_COMP_B 2 +#define PIPE_FORMAT_COMP_A 3 +#define PIPE_FORMAT_COMP_0 4 +#define PIPE_FORMAT_COMP_1 5 +#define PIPE_FORMAT_COMP_Z 6 +#define PIPE_FORMAT_COMP_S 7 + +/** + * Format types for RGBAZS layout. + */ +#define PIPE_FORMAT_TYPE_UNKNOWN 0 +#define PIPE_FORMAT_TYPE_FLOAT 1 +#define PIPE_FORMAT_TYPE_UNORM 2 +#define PIPE_FORMAT_TYPE_SNORM 3 +#define PIPE_FORMAT_TYPE_USCALED 4 +#define PIPE_FORMAT_TYPE_SSCALED 5 + +/** + * Because the destination vector is assumed to be RGBA FLOAT, we + * need to know how to swizzle and expand components from the source + * vector. + * Let's take U_A1_R5_G5_B5 as an example. X swizzle is A, X size + * is 1 bit and type is UNORM. So we take the most significant bit + * from source vector, convert 0 to 0.0 and 1 to 1.0 and save it + * in the last component of the destination RGBA component. + * Next, Y swizzle is R, Y size is 5 and type is UNORM. We normalize + * those 5 bits into [0.0; 1.0] range and put it into second + * component of the destination vector. Rinse and repeat for + * components Z and W. + * If any of size fields is zero, it means the source format contains + * less than four components. + * If any swizzle is 0 or 1, the corresponding destination component + * should be filled with 0.0 and 1.0, respectively. + */ +typedef uint pipe_format_rgbazs_t; + +static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) +{ + return (f >> shift) & mask; +} + +/* XXX: The bit layout needs to be revised, can't currently encode 10-bit components. */ + +#define pf_swizzle_x(f) pf_get(f, 2, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_y(f) pf_get(f, 5, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_z(f) pf_get(f, 8, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_w(f) pf_get(f, 11, 0x7) /**< PIPE_FORMAT_COMP_ */ +#define pf_swizzle_xyzw(f,i) pf_get(f, 2+((i)*3), 0x7) +#define pf_size_x(f) pf_get(f, 14, 0x7) /**< Size of X */ +#define pf_size_y(f) pf_get(f, 17, 0x7) /**< Size of Y */ +#define pf_size_z(f) pf_get(f, 20, 0x7) /**< Size of Z */ +#define pf_size_w(f) pf_get(f, 23, 0x7) /**< Size of W */ +#define pf_size_xyzw(f,i) pf_get(f, 14+((i)*3), 0x7) +#define pf_exp8(f) pf_get(f, 26, 0x3) /**< Scale size by 8 ^ exp8 */ +#define pf_type(f) pf_get(f, 28, 0xf) /**< PIPE_FORMAT_TYPE_ */ + +/** + * Helper macro to encode the above structure into a 32-bit value. + */ +#define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP8, TYPE ) (\ + (PIPE_FORMAT_LAYOUT_RGBAZS << 0) |\ + ((SWZ) << 2) |\ + ((SIZEX) << 14) |\ + ((SIZEY) << 17) |\ + ((SIZEZ) << 20) |\ + ((SIZEW) << 23) |\ + ((EXP8) << 26) |\ + ((TYPE) << 28) ) + +/** + * Helper macro to encode the swizzle part of the structure above. + */ +#define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 0) | ((SWZY) << 3) | ((SWZZ) << 6) | ((SWZW) << 9)) + +/** + * Shorthand macro for RGBAZS layout with component sizes in 1-bit units. + */ +#define _PIPE_FORMAT_RGBAZS_1( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE ) + +/** + * Shorthand macro for RGBAZS layout with component sizes in 8-bit units. + */ +#define _PIPE_FORMAT_RGBAZS_8( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) + +/** + * Shorthand macro for RGBAZS layout with component sizes in 64-bit units. + */ +#define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 2, TYPE ) + +/** + * Shorthand macro for common format swizzles. + */ +#define _PIPE_FORMAT_R000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_RG00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) +#define _PIPE_FORMAT_BGRA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R ) +#define _PIPE_FORMAT_RRR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) +#define _PIPE_FORMAT_RRRR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R ) +#define _PIPE_FORMAT_RRRG _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G ) +#define _PIPE_FORMAT_Z000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_ZS00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) + +/** + * YCBCR Format Layout. + */ + +/** + * This only contains a flag that indicates whether the format is reversed or + * not. + */ +typedef uint pipe_format_ycbcr_t; + +/** + * Helper macro to encode the above structure into a 32-bit value. + */ +#define _PIPE_FORMAT_YCBCR( REV ) (\ + (PIPE_FORMAT_LAYOUT_YCBCR << 0) |\ + ((REV) << 2) ) + +static INLINE uint pf_rev(pipe_format_ycbcr_t f) +{ + return (f >> 2) & 0x1; +} + +/** + * Texture/surface image formats (preliminary) + */ + +/* KW: Added lots of surface formats to support vertex element layout + * definitions, and eventually render-to-vertex-buffer. Could + * consider making float/int/uint/scaled/normalized a separate + * parameter, but on the other hand there are special cases like + * z24s8, compressed textures, ycbcr, etc that won't fit that model. + */ + +enum pipe_format { + PIPE_FORMAT_NONE = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_0000, 0, 0, 0, 0, PIPE_FORMAT_TYPE_UNKNOWN ), + PIPE_FORMAT_A8R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_B8G8R8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_A1R5G5B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_A4R4G4B4_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_U_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ + PIPE_FORMAT_U_A8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ + PIPE_FORMAT_U_I8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ + PIPE_FORMAT_U_A8_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha, luminance */ + PIPE_FORMAT_YCBCR = _PIPE_FORMAT_YCBCR( 0 ), + PIPE_FORMAT_YCBCR_REV = _PIPE_FORMAT_YCBCR( 1 ), + PIPE_FORMAT_Z16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_Z32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_Z32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_S8Z24_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_Z24S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte stencil */ + PIPE_FORMAT_R64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R64G64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R64G64B64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R64G64B64A64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32G32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32G32B32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32G32B32A32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32G32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32G32B32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32G32B32A32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32G32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32G32B32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32G32B32A32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32G32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32G32B32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32G32B32A32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32G32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32G32B32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32G32B32A32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16G16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16G16B16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16G16B16A16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16G16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16G16B16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16G16B16A16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16G16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16G16B16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16G16B16A16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16G16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16G16B16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16G16B16A16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8G8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8G8B8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8G8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8G8B8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8G8B8A8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8G8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8G8B8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8G8B8A8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8B8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8B8A8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) +}; + + +/** + * XXX should remove this, but S8_UNORM is a poor name + */ +#define PIPE_FORMAT_U_S8 PIPE_FORMAT_S8_UNORM + + +/** + * Builds pipe format name from format token. + */ +static INLINE char *pf_sprint_name( char *str, uint format ) +{ + strcpy( str, "PIPE_FORMAT_" ); + switch (pf_layout( format )) { + case PIPE_FORMAT_LAYOUT_RGBAZS: { + pipe_format_rgbazs_t rgbazs = (pipe_format_rgbazs_t) format; + uint i; + uint scale = 1 << (pf_exp8( rgbazs ) * 3); + + for (i = 0; i < 4; i++) { + uint size = pf_size_xyzw( rgbazs, i ); + + if (size == 0) { + break; + } + switch (pf_swizzle_xyzw( rgbazs, i )) { + case PIPE_FORMAT_COMP_R: + strcat( str, "R" ); + break; + case PIPE_FORMAT_COMP_G: + strcat( str, "G" ); + break; + case PIPE_FORMAT_COMP_B: + strcat( str, "B" ); + break; + case PIPE_FORMAT_COMP_A: + strcat( str, "A" ); + break; + case PIPE_FORMAT_COMP_0: + strcat( str, "0" ); + break; + case PIPE_FORMAT_COMP_1: + strcat( str, "1" ); + break; + case PIPE_FORMAT_COMP_Z: + strcat( str, "Z" ); + break; + case PIPE_FORMAT_COMP_S: + strcat( str, "S" ); + break; + } + sprintf( &str[strlen( str )], "%u", size * scale ); + } + if (i != 0) { + strcat( str, "_" ); + } + switch (pf_type( rgbazs )) { + case PIPE_FORMAT_TYPE_UNKNOWN: + strcat( str, "NONE" ); + break; + case PIPE_FORMAT_TYPE_FLOAT: + strcat( str, "FLOAT" ); + break; + case PIPE_FORMAT_TYPE_UNORM: + strcat( str, "UNORM" ); + break; + case PIPE_FORMAT_TYPE_SNORM: + strcat( str, "SNORM" ); + break; + case PIPE_FORMAT_TYPE_USCALED: + strcat( str, "USCALED" ); + break; + case PIPE_FORMAT_TYPE_SSCALED: + strcat( str, "SSCALED" ); + break; + } + } + break; + case PIPE_FORMAT_LAYOUT_YCBCR: { + pipe_format_ycbcr_t ycbcr = (pipe_format_ycbcr_t) format; + + strcat( str, "YCBCR" ); + if (pf_rev( ycbcr )) { + strcat( str, "_REV" ); + } + } + break; + } + return str; +} + +static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) +{ + uint size; + + if (pf_swizzle_x(format) == comp) { + size = pf_size_x(format); + } + else if (pf_swizzle_y(format) == comp) { + size = pf_size_y(format); + } + else if (pf_swizzle_z(format) == comp) { + size = pf_size_z(format); + } + else if (pf_swizzle_w(format) == comp) { + size = pf_size_w(format); + } + else { + size = 0; + } + return size << (pf_exp8(format) * 3); +} + +static INLINE uint pf_get_bits( enum pipe_format format ) +{ + if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { + return + pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_S ); + } + else { + assert( pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR ); + + /* TODO */ + assert( 0 ); + return 0; + } +} + +static INLINE uint pf_get_size( enum pipe_format format ) { + assert(pf_get_bits(format) % 8 == 0); + return pf_get_bits(format) / 8; +} + +#endif diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h new file mode 100644 index 0000000000..ebf6ed86bc --- /dev/null +++ b/src/gallium/include/pipe/p_inlines.h @@ -0,0 +1,112 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_INLINES_H +#define P_INLINES_H + +#include "p_context.h" +#include "p_defines.h" +#include "p_winsys.h" + + +static INLINE void * +pipe_surface_map(struct pipe_surface *surface) +{ + return (char *)surface->winsys->buffer_map( surface->winsys, surface->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_CPU_READ ) + + surface->offset; +} + +static INLINE void +pipe_surface_unmap(struct pipe_surface *surface) +{ + surface->winsys->buffer_unmap( surface->winsys, surface->buffer ); +} + +/** + * Set 'ptr' to point to 'surf' and update reference counting. + * The old thing pointed to, if any, will be unreferenced first. + * 'surf' may be NULL. + */ +static INLINE void +pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) +{ + /* bump the refcount first */ + if (surf) + surf->refcount++; + + if (*ptr /* && --(*ptr)->refcount == 0 */) { + struct pipe_winsys *winsys = (*ptr)->winsys; + winsys->surface_release(winsys, ptr); + assert(!*ptr); + } + + *ptr = surf; +} + + +/* XXX: thread safety issues! + */ +static INLINE void +pipe_buffer_reference(struct pipe_winsys *winsys, + struct pipe_buffer **ptr, + struct pipe_buffer *buf) +{ + if (buf) + buf->refcount++; + + if (*ptr && --(*ptr)->refcount == 0) + winsys->buffer_destroy( winsys, *ptr ); + + *ptr = buf; +} + + + +/** + * \sa pipe_surface_reference + */ +static INLINE void +pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, + struct pipe_texture *pt) +{ + assert(ptr); + + if (pt) + pt->refcount++; + + if (*ptr) { + pipe->texture_release(pipe, ptr); + assert(!*ptr); + } + + *ptr = pt; +} + + +#endif /* P_INLINES_H */ diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h new file mode 100644 index 0000000000..3ce35310f6 --- /dev/null +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -0,0 +1,806 @@ +#if !defined TGSI_TOKEN_H +#define TGSI_TOKEN_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_version +{ + unsigned MajorVersion : 8; + unsigned MinorVersion : 8; + unsigned Padding : 16; +}; + +struct tgsi_header +{ + unsigned HeaderSize : 8; + unsigned BodySize : 24; +}; + +#define TGSI_PROCESSOR_FRAGMENT 0 +#define TGSI_PROCESSOR_VERTEX 1 +#define TGSI_PROCESSOR_GEOMETRY 2 + +struct tgsi_processor +{ + unsigned Processor : 4; /* TGSI_PROCESSOR_ */ + unsigned Padding : 28; +}; + +#define TGSI_TOKEN_TYPE_DECLARATION 0 +#define TGSI_TOKEN_TYPE_IMMEDIATE 1 +#define TGSI_TOKEN_TYPE_INSTRUCTION 2 + +struct tgsi_token +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_ */ + unsigned Size : 8; /* UINT */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ +}; + +#define TGSI_FILE_NULL 0 +#define TGSI_FILE_CONSTANT 1 +#define TGSI_FILE_INPUT 2 +#define TGSI_FILE_OUTPUT 3 +#define TGSI_FILE_TEMPORARY 4 +#define TGSI_FILE_SAMPLER 5 +#define TGSI_FILE_ADDRESS 6 +#define TGSI_FILE_IMMEDIATE 7 + +#define TGSI_DECLARE_RANGE 0 +#define TGSI_DECLARE_MASK 1 + +#define TGSI_WRITEMASK_NONE 0x00 +#define TGSI_WRITEMASK_X 0x01 +#define TGSI_WRITEMASK_Y 0x02 +#define TGSI_WRITEMASK_XY 0x03 +#define TGSI_WRITEMASK_Z 0x04 +#define TGSI_WRITEMASK_XZ 0x05 +#define TGSI_WRITEMASK_YZ 0x06 +#define TGSI_WRITEMASK_XYZ 0x07 +#define TGSI_WRITEMASK_W 0x08 +#define TGSI_WRITEMASK_XW 0x09 +#define TGSI_WRITEMASK_YW 0x0A +#define TGSI_WRITEMASK_XYW 0x0B +#define TGSI_WRITEMASK_ZW 0x0C +#define TGSI_WRITEMASK_XZW 0x0D +#define TGSI_WRITEMASK_YZW 0x0E +#define TGSI_WRITEMASK_XYZW 0x0F + +struct tgsi_declaration +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ + unsigned Size : 8; /* UINT */ + unsigned File : 4; /* one of TGSI_FILE_x */ + unsigned Declare : 4; /* one of TGSI_DECLARE_x */ + unsigned UsageMask : 4; /* bitmask of TGSI_WRITEMASK_x flags */ + unsigned Interpolate : 1; /* BOOL, any interpolation info? */ + unsigned Semantic : 1; /* BOOL, any semantic info? */ + unsigned Padding : 5; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_declaration_range +{ + unsigned First : 16; /* UINT */ + unsigned Last : 16; /* UINT */ +}; + +struct tgsi_declaration_mask +{ + unsigned Mask : 32; /* UINT */ +}; + +#define TGSI_INTERPOLATE_CONSTANT 0 +#define TGSI_INTERPOLATE_LINEAR 1 +#define TGSI_INTERPOLATE_PERSPECTIVE 2 + +struct tgsi_declaration_interpolation +{ + unsigned Interpolate : 4; /* TGSI_INTERPOLATE_ */ + unsigned Padding : 28; +}; + +#define TGSI_SEMANTIC_POSITION 0 +#define TGSI_SEMANTIC_COLOR 1 +#define TGSI_SEMANTIC_BCOLOR 2 /**< back-face color */ +#define TGSI_SEMANTIC_FOG 3 +#define TGSI_SEMANTIC_PSIZE 4 +#define TGSI_SEMANTIC_GENERIC 5 +#define TGSI_SEMANTIC_NORMAL 6 +#define TGSI_SEMANTIC_COUNT 7 /**< number of semantic values */ + +struct tgsi_declaration_semantic +{ + unsigned SemanticName : 8; /* one of TGSI_SEMANTIC_ */ + unsigned SemanticIndex : 16; /* UINT */ + unsigned Padding : 8; +}; + +#define TGSI_IMM_FLOAT32 0 + +struct tgsi_immediate +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_IMMEDIATE */ + unsigned Size : 8; /* UINT */ + unsigned DataType : 4; /* TGSI_IMM_ */ + unsigned Padding : 15; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_immediate_float32 +{ + float Float; +}; + +/* + * GL_NV_vertex_program + */ +#define TGSI_OPCODE_ARL 0 +#define TGSI_OPCODE_MOV 1 +#define TGSI_OPCODE_LIT 2 +#define TGSI_OPCODE_RCP 3 +#define TGSI_OPCODE_RSQ 4 +#define TGSI_OPCODE_EXP 5 +#define TGSI_OPCODE_LOG 6 +#define TGSI_OPCODE_MUL 7 +#define TGSI_OPCODE_ADD 8 +#define TGSI_OPCODE_DP3 9 +#define TGSI_OPCODE_DP4 10 +#define TGSI_OPCODE_DST 11 +#define TGSI_OPCODE_MIN 12 +#define TGSI_OPCODE_MAX 13 +#define TGSI_OPCODE_SLT 14 +#define TGSI_OPCODE_SGE 15 +#define TGSI_OPCODE_MAD 16 + +/* + * GL_ATI_fragment_shader + */ +#define TGSI_OPCODE_SUB 17 +#define TGSI_OPCODE_DOT3 TGSI_OPCODE_DP3 +#define TGSI_OPCODE_DOT4 TGSI_OPCODE_DP4 +#define TGSI_OPCODE_LERP 18 +#define TGSI_OPCODE_CND 19 +#define TGSI_OPCODE_CND0 20 +#define TGSI_OPCODE_DOT2ADD 21 + +/* + * GL_EXT_vertex_shader + */ +#define TGSI_OPCODE_INDEX 22 +#define TGSI_OPCODE_NEGATE 23 +#define TGSI_OPCODE_MADD TGSI_OPCODE_MAD +#define TGSI_OPCODE_FRAC 24 +#define TGSI_OPCODE_SETGE TGSI_OPCODE_SGE +#define TGSI_OPCODE_SETLT TGSI_OPCODE_SLT +#define TGSI_OPCODE_CLAMP 25 +#define TGSI_OPCODE_FLOOR 26 +#define TGSI_OPCODE_ROUND 27 +#define TGSI_OPCODE_EXPBASE2 28 +#define TGSI_OPCODE_LOGBASE2 29 +#define TGSI_OPCODE_POWER 30 +#define TGSI_OPCODE_RECIP TGSI_OPCODE_RCP +#define TGSI_OPCODE_RECIPSQRT TGSI_OPCODE_RSQ +#define TGSI_OPCODE_CROSSPRODUCT 31 +#define TGSI_OPCODE_MULTIPLYMATRIX 32 + +/* + * GL_NV_vertex_program1_1 + */ +#define TGSI_OPCODE_ABS 33 +#define TGSI_OPCODE_RCC 34 +#define TGSI_OPCODE_DPH 35 + +/* + * GL_NV_fragment_program + */ +#define TGSI_OPCODE_COS 36 +#define TGSI_OPCODE_DDX 37 +#define TGSI_OPCODE_DDY 38 +#define TGSI_OPCODE_EX2 TGSI_OPCODE_EXPBASE2 +#define TGSI_OPCODE_FLR TGSI_OPCODE_FLOOR +#define TGSI_OPCODE_FRC TGSI_OPCODE_FRAC +#define TGSI_OPCODE_KILP 39 /* predicated kill */ +#define TGSI_OPCODE_LG2 TGSI_OPCODE_LOGBASE2 +#define TGSI_OPCODE_LRP TGSI_OPCODE_LERP +#define TGSI_OPCODE_PK2H 40 +#define TGSI_OPCODE_PK2US 41 +#define TGSI_OPCODE_PK4B 42 +#define TGSI_OPCODE_PK4UB 43 +#define TGSI_OPCODE_POW TGSI_OPCODE_POWER +#define TGSI_OPCODE_RFL 44 +#define TGSI_OPCODE_SEQ 45 +#define TGSI_OPCODE_SFL 46 +#define TGSI_OPCODE_SGT 47 +#define TGSI_OPCODE_SIN 48 +#define TGSI_OPCODE_SLE 49 +#define TGSI_OPCODE_SNE 50 +#define TGSI_OPCODE_STR 51 +#define TGSI_OPCODE_TEX 52 +#define TGSI_OPCODE_TXD 53 +#define TGSI_OPCODE_UP2H 54 +#define TGSI_OPCODE_UP2US 55 +#define TGSI_OPCODE_UP4B 56 +#define TGSI_OPCODE_UP4UB 57 +#define TGSI_OPCODE_X2D 58 + +/* + * GL_NV_vertex_program2 + */ +#define TGSI_OPCODE_ARA 59 +#define TGSI_OPCODE_ARR 60 +#define TGSI_OPCODE_BRA 61 +#define TGSI_OPCODE_CAL 62 +#define TGSI_OPCODE_RET 63 +#define TGSI_OPCODE_SSG 64 + +/* + * GL_ARB_vertex_program + */ +#define TGSI_OPCODE_SWZ TGSI_OPCODE_MOV +#define TGSI_OPCODE_XPD TGSI_OPCODE_CROSSPRODUCT + +/* + * GL_ARB_fragment_program + */ +#define TGSI_OPCODE_CMP 65 +#define TGSI_OPCODE_SCS 66 +#define TGSI_OPCODE_TXB 67 + +/* + * GL_NV_fragment_program_option + */ +/* No new opcode */ + +/* + * GL_NV_fragment_program2 + */ +#define TGSI_OPCODE_NRM 68 +#define TGSI_OPCODE_DIV 69 +#define TGSI_OPCODE_DP2 70 +#define TGSI_OPCODE_DP2A TGSI_OPCODE_DOT2ADD +#define TGSI_OPCODE_TXL 71 +#define TGSI_OPCODE_BRK 72 +#define TGSI_OPCODE_IF 73 +#define TGSI_OPCODE_LOOP 74 +#define TGSI_OPCODE_REP 75 +#define TGSI_OPCODE_ELSE 76 +#define TGSI_OPCODE_ENDIF 77 +#define TGSI_OPCODE_ENDLOOP 78 +#define TGSI_OPCODE_ENDREP 79 + +/* + * GL_NV_vertex_program2_option + */ + +/* + * GL_NV_vertex_program3 + */ +#define TGSI_OPCODE_PUSHA 80 +#define TGSI_OPCODE_POPA 81 + +/* + * GL_NV_gpu_program4 + */ +#define TGSI_OPCODE_CEIL 82 +#define TGSI_OPCODE_I2F 83 +#define TGSI_OPCODE_NOT 84 +#define TGSI_OPCODE_TRUNC 85 +#define TGSI_OPCODE_SHL 86 +#define TGSI_OPCODE_SHR 87 +#define TGSI_OPCODE_AND 88 +#define TGSI_OPCODE_OR 89 +#define TGSI_OPCODE_MOD 90 +#define TGSI_OPCODE_XOR 91 +#define TGSI_OPCODE_SAD 92 +#define TGSI_OPCODE_TXF 93 +#define TGSI_OPCODE_TXQ 94 +#define TGSI_OPCODE_CONT 95 + +/* + * GL_NV_vertex_program4 + */ +/* Same as GL_NV_gpu_program4 */ + +/* + * GL_NV_fragment_program4 + */ +/* Same as GL_NV_gpu_program4 */ + +/* + * GL_NV_geometry_program4 + */ +/* Same as GL_NV_gpu_program4 */ +#define TGSI_OPCODE_EMIT 96 +#define TGSI_OPCODE_ENDPRIM 97 + +/* + * GLSL + */ +#define TGSI_OPCODE_BGNLOOP2 98 +#define TGSI_OPCODE_BGNSUB 99 +#define TGSI_OPCODE_ENDLOOP2 100 +#define TGSI_OPCODE_ENDSUB 101 +#define TGSI_OPCODE_INT TGSI_OPCODE_TRUNC +#define TGSI_OPCODE_NOISE1 102 +#define TGSI_OPCODE_NOISE2 103 +#define TGSI_OPCODE_NOISE3 104 +#define TGSI_OPCODE_NOISE4 105 +#define TGSI_OPCODE_NOP 106 + +/* + * ps_1_1 + */ +#define TGSI_OPCODE_TEXCOORD TGSI_OPCODE_NOP +#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL +#define TGSI_OPCODE_TEXBEM 107 +#define TGSI_OPCODE_TEXBEML 108 +#define TGSI_OPCODE_TEXREG2AR 109 +#define TGSI_OPCODE_TEXM3X2PAD 110 +#define TGSI_OPCODE_TEXM3X2TEX 111 +#define TGSI_OPCODE_TEXM3X3PAD 112 +#define TGSI_OPCODE_TEXM3X3TEX 113 +#define TGSI_OPCODE_TEXM3X3SPEC 114 +#define TGSI_OPCODE_TEXM3X3VSPEC 115 + +/* + * ps_1_2 + */ +#define TGSI_OPCODE_TEXREG2GB 116 +#define TGSI_OPCODE_TEXREG2RGB 117 +#define TGSI_OPCODE_TEXDP3TEX 118 +#define TGSI_OPCODE_TEXDP3 119 +#define TGSI_OPCODE_TEXM3X3 120 +/* CMP - use TGSI_OPCODE_CND0 */ + +/* + * ps_1_3 + */ +#define TGSI_OPCODE_TEXM3X2DEPTH 121 +/* CMP - use TGSI_OPCODE_CND0 */ + +/* + * ps_1_4 + */ +#define TGSI_OPCODE_TEXCRD TGSI_OPCODE_TEXCOORD +#define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX +#define TGSI_OPCODE_TEXDEPTH 122 +#define TGSI_OPCODE_BEM 123 + +/* + * ps_2_0 + */ +#define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX +#define TGSI_OPCODE_M4X3 124 +#define TGSI_OPCODE_M3X4 125 +#define TGSI_OPCODE_M3X3 126 +#define TGSI_OPCODE_M3X2 127 +#define TGSI_OPCODE_CRS TGSI_OPCODE_XPD +#define TGSI_OPCODE_NRM4 128 +#define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS +#define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB +#define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A + +/* + * ps_2_x + */ +#define TGSI_OPCODE_CALL TGSI_OPCODE_CAL +#define TGSI_OPCODE_CALLNZ 129 +#define TGSI_OPCODE_IFC 130 +#define TGSI_OPCODE_BREAK TGSI_OPCODE_BRK +#define TGSI_OPCODE_BREAKC 131 +#define TGSI_OPCODE_DSX TGSI_OPCODE_DDX +#define TGSI_OPCODE_DSY TGSI_OPCODE_DDY +#define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD + +/* + * vs_1_1 + */ +#define TGSI_OPCODE_EXPP TGSI_OPCODE_EXP +#define TGSI_OPCODE_LOGP TGSI_OPCODE_LG2 + +/* + * vs_2_0 + */ +#define TGSI_OPCODE_SGN TGSI_OPCODE_SSG +#define TGSI_OPCODE_MOVA TGSI_OPCODE_ARR + +/* + * vs_2_x + */ + +#define TGSI_OPCODE_KIL 132 /* unpredicated kill */ +#define TGSI_OPCODE_END 133 /* aka HALT */ + +#define TGSI_OPCODE_LAST 134 + +#define TGSI_SAT_NONE 0 /* do not saturate */ +#define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ +#define TGSI_SAT_MINUS_PLUS_ONE 2 /* clamp to [-1,1] */ + +/* + * Opcode is the operation code to execute. A given operation defines the + * semantics how the source registers (if any) are interpreted and what is + * written to the destination registers (if any) as a result of execution. + * + * NumDstRegs and NumSrcRegs is the number of destination and source registers, + * respectively. For a given operation code, those numbers are fixed and are + * present here only for convenience. + * + * If Extended is TRUE, it is now executed. + * + * Saturate controls how are final results in destination registers modified. + */ + +struct tgsi_instruction +{ + unsigned Type : 4; /* TGSI_TOKEN_TYPE_INSTRUCTION */ + unsigned Size : 8; /* UINT */ + unsigned Opcode : 8; /* TGSI_OPCODE_ */ + unsigned Saturate : 2; /* TGSI_SAT_ */ + unsigned NumDstRegs : 2; /* UINT */ + unsigned NumSrcRegs : 4; /* UINT */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows. + * + * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow. + * + * Then, tgsi_instruction::NumSrcRegs of tgsi_src_register follow. + * + * tgsi_instruction::Size contains the total number of words that make the + * instruction, including the instruction word. + */ + +#define TGSI_INSTRUCTION_EXT_TYPE_NV 0 +#define TGSI_INSTRUCTION_EXT_TYPE_LABEL 1 +#define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE 2 +#define TGSI_INSTRUCTION_EXT_TYPE_PREDICATE 3 + +struct tgsi_instruction_ext +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_NV, it should + * be cast to tgsi_instruction_ext_nv. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_LABEL, it + * should be cast to tgsi_instruction_ext_label. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it + * should be cast to tgsi_instruction_ext_texture. + * + * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_PREDICATE, it + * should be cast to tgsi_instruction_ext_predicate. + * + * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext + * follows. + */ + +#define TGSI_PRECISION_DEFAULT 0 +#define TGSI_PRECISION_FLOAT32 1 +#define TGSI_PRECISION_FLOAT16 2 +#define TGSI_PRECISION_FIXED12 3 + +#define TGSI_CC_GT 0 +#define TGSI_CC_EQ 1 +#define TGSI_CC_LT 2 +#define TGSI_CC_UN 3 +#define TGSI_CC_GE 4 +#define TGSI_CC_LE 5 +#define TGSI_CC_NE 6 +#define TGSI_CC_TR 7 +#define TGSI_CC_FL 8 + +#define TGSI_SWIZZLE_X 0 +#define TGSI_SWIZZLE_Y 1 +#define TGSI_SWIZZLE_Z 2 +#define TGSI_SWIZZLE_W 3 + +/* + * Precision controls the precision at which the operation should be executed. + * + * CondDstUpdate enables condition code register writes. When this field is + * TRUE, CondDstIndex specifies the index of the condition code register to + * update. + * + * CondFlowEnable enables conditional execution of the operation. When this + * field is TRUE, CondFlowIndex specifies the index of the condition code + * register to test against CondMask with component swizzle controled by + * CondSwizzleX, CondSwizzleY, CondSwizzleZ and CondSwizzleW. If the test fails, + * the operation is not executed. + */ + +struct tgsi_instruction_ext_nv +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_NV */ + unsigned Precision : 4; /* TGSI_PRECISION_ */ + unsigned CondDstIndex : 4; /* UINT */ + unsigned CondFlowIndex : 4; /* UINT */ + unsigned CondMask : 4; /* TGSI_CC_ */ + unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned CondDstUpdate : 1; /* BOOL */ + unsigned CondFlowEnable : 1; /* BOOL */ + unsigned Padding : 1; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_instruction_ext_label +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ + unsigned Label : 24; /* UINT */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ +}; + +#define TGSI_TEXTURE_UNKNOWN 0 +#define TGSI_TEXTURE_1D 1 +#define TGSI_TEXTURE_2D 2 +#define TGSI_TEXTURE_3D 3 +#define TGSI_TEXTURE_CUBE 4 +#define TGSI_TEXTURE_RECT 5 +#define TGSI_TEXTURE_SHADOW1D 6 +#define TGSI_TEXTURE_SHADOW2D 7 +#define TGSI_TEXTURE_SHADOWRECT 8 + +struct tgsi_instruction_ext_texture +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */ + unsigned Texture : 8; /* TGSI_TEXTURE_ */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_instruction_ext_predicate +{ + unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ + unsigned PredDstIndex : 4; /* UINT */ + unsigned PredWriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Padding : 19; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * File specifies the register array to access. + * + * Index specifies the element number of a register in the register file. + * + * If Indirect is TRUE, Index should be offset by the X component of a source + * register that follows. The register can be now fetched into local storage + * for further processing. + * + * If Negate is TRUE, all components of the fetched register are negated. + * + * The fetched register components are swizzled according to SwizzleX, SwizzleY, + * SwizzleZ and SwizzleW. + * + * If Extended is TRUE, any further modifications to the source register are + * made to this temporary storage. + */ + +struct tgsi_src_register +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned Negate : 1; /* BOOL */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_src_register::Extended is TRUE, tgsi_src_register_ext follows. + * + * Then, if tgsi_src_register::Indirect is TRUE, another tgsi_src_register + * follows. + * + * Then, if tgsi_src_register::Dimension is TRUE, tgsi_dimension follows. + */ + +#define TGSI_SRC_REGISTER_EXT_TYPE_SWZ 0 +#define TGSI_SRC_REGISTER_EXT_TYPE_MOD 1 + +struct tgsi_src_register_ext +{ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_SWZ, + * it should be cast to tgsi_src_register_ext_swz. + * + * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_MOD, + * it should be cast to tgsi_src_register_ext_mod. + * + * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext + * follows. + */ + +#define TGSI_EXTSWIZZLE_X TGSI_SWIZZLE_X +#define TGSI_EXTSWIZZLE_Y TGSI_SWIZZLE_Y +#define TGSI_EXTSWIZZLE_Z TGSI_SWIZZLE_Z +#define TGSI_EXTSWIZZLE_W TGSI_SWIZZLE_W +#define TGSI_EXTSWIZZLE_ZERO 4 +#define TGSI_EXTSWIZZLE_ONE 5 + +/* + * ExtSwizzleX, ExtSwizzleY, ExtSwizzleZ and ExtSwizzleW swizzle the source + * register in an extended manner. + * + * NegateX, NegateY, NegateZ and NegateW negate individual components of the + * source register. + * + * ExtDivide specifies which component is used to divide all components of the + * source register. + */ + +struct tgsi_src_register_ext_swz +{ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */ + unsigned ExtSwizzleX : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleY : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleZ : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned ExtSwizzleW : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned NegateX : 1; /* BOOL */ + unsigned NegateY : 1; /* BOOL */ + unsigned NegateZ : 1; /* BOOL */ + unsigned NegateW : 1; /* BOOL */ + unsigned ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned Padding : 3; + unsigned Extended : 1; /* BOOL */ +}; + +/** + * Extra src register modifiers + * + * If Complement is TRUE, the source register is modified by subtracting it + * from 1.0. + * + * If Bias is TRUE, the source register is modified by subtracting 0.5 from it. + * + * If Scale2X is TRUE, the source register is modified by multiplying it by 2.0. + * + * If Absolute is TRUE, the source register is modified by removing the sign. + * + * If Negate is TRUE, the source register is modified by negating it. + */ + +struct tgsi_src_register_ext_mod +{ + unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */ + unsigned Complement : 1; /* BOOL */ + unsigned Bias : 1; /* BOOL */ + unsigned Scale2X : 1; /* BOOL */ + unsigned Absolute : 1; /* BOOL */ + unsigned Negate : 1; /* BOOL */ + unsigned Padding : 22; + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_dimension +{ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + unsigned Padding : 13; + int Index : 16; /* SINT */ + unsigned Extended : 1; /* BOOL */ +}; + +struct tgsi_dst_register +{ + unsigned File : 4; /* TGSI_FILE_ */ + unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ + unsigned Indirect : 1; /* BOOL */ + unsigned Dimension : 1; /* BOOL */ + int Index : 16; /* SINT */ + unsigned Padding : 5; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * If tgsi_dst_register::Extended is TRUE, tgsi_dst_register_ext follows. + * + * Then, if tgsi_dst_register::Indirect is TRUE, tgsi_src_register follows. + */ + +#define TGSI_DST_REGISTER_EXT_TYPE_CONDCODE 0 +#define TGSI_DST_REGISTER_EXT_TYPE_MODULATE 1 +#define TGSI_DST_REGISTER_EXT_TYPE_PREDICATE 2 + +struct tgsi_dst_register_ext +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_ */ + unsigned Padding : 27; + unsigned Extended : 1; /* BOOL */ +}; + +/** + * Extra destination register modifiers + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_CONDCODE, + * it should be cast to tgsi_dst_register_ext_condcode. + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_MODULATE, + * it should be cast to tgsi_dst_register_ext_modulate. + * + * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_PREDICATE, + * it should be cast to tgsi_dst_register_ext_predicate. + * + * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext + * follows. + */ +struct tgsi_dst_register_ext_concode +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ + unsigned CondMask : 4; /* TGSI_CC_ */ + unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned CondSrcIndex : 4; /* UINT */ + unsigned Padding : 11; + unsigned Extended : 1; /* BOOL */ +}; + +#define TGSI_MODULATE_1X 0 +#define TGSI_MODULATE_2X 1 +#define TGSI_MODULATE_4X 2 +#define TGSI_MODULATE_8X 3 +#define TGSI_MODULATE_HALF 4 +#define TGSI_MODULATE_QUARTER 5 +#define TGSI_MODULATE_EIGHTH 6 + +struct tgsi_dst_register_ext_modulate +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */ + unsigned Modulate : 4; /* TGSI_MODULATE_ */ + unsigned Padding : 23; + unsigned Extended : 1; /* BOOL */ +}; + +/* + * Currently, the following constraints apply. + * + * - PredSwizzleXYZW is either set to identity or replicate. + * - PredSrcIndex is 0. + */ + +struct tgsi_dst_register_ext_predicate +{ + unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */ + unsigned PredSwizzleX : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleY : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleZ : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSwizzleW : 2; /* TGSI_SWIZZLE_ */ + unsigned PredSrcIndex : 4; /* UINT */ + unsigned Negate : 1; /* BOOL */ + unsigned Padding : 14; + unsigned Extended : 1; /* BOOL */ +}; + + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_TOKEN_H + diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h new file mode 100644 index 0000000000..4d3a6b2f41 --- /dev/null +++ b/src/gallium/include/pipe/p_state.h @@ -0,0 +1,322 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Abstract graphics pipe state objects. + * + * Basic notes: + * 1. Want compact representations, so we use bitfields. + * 2. Put bitfields before other (GLfloat) fields. + */ + + +#ifndef PIPE_STATE_H +#define PIPE_STATE_H + +#include "p_compiler.h" +#include "p_defines.h" +#include "p_format.h" + +/** + * Implementation limits + */ +#define PIPE_MAX_SAMPLERS 8 +#define PIPE_MAX_CLIP_PLANES 6 +#define PIPE_MAX_CONSTANT 32 +#define PIPE_ATTRIB_MAX 32 +#define PIPE_MAX_COLOR_BUFS 8 +#define PIPE_MAX_TEXTURE_LEVELS 16 +#define PIPE_MAX_FEEDBACK_ATTRIBS 16 +#define PIPE_MAX_SHADER_INPUTS 16 +#define PIPE_MAX_SHADER_OUTPUTS 16 + + +/* fwd decls */ +struct pipe_surface; +struct pipe_winsys; + + + +/** + * The driver will certainly subclass this to include actual memory + * management information. + */ +struct pipe_buffer { + unsigned alignment; + unsigned usage; + unsigned size; + + /** Reference count */ + unsigned refcount; +}; + + + + +/** + * Primitive (point/line/tri) rasterization info + */ +struct pipe_rasterizer_state +{ + unsigned flatshade:1; + unsigned light_twoside:1; + unsigned front_winding:2; /**< PIPE_WINDING_x */ + unsigned cull_mode:2; /**< PIPE_WINDING_x */ + unsigned fill_cw:2; /**< PIPE_POLYGON_MODE_x */ + unsigned fill_ccw:2; /**< PIPE_POLYGON_MODE_x */ + unsigned offset_cw:1; + unsigned offset_ccw:1; + unsigned scissor:1; + unsigned poly_smooth:1; + unsigned poly_stipple_enable:1; + unsigned point_smooth:1; + unsigned point_sprite:1; + unsigned point_size_per_vertex:1; /**< size computed in vertex shader */ + unsigned multisample:1; /* XXX maybe more ms state in future */ + unsigned line_smooth:1; + unsigned line_stipple_enable:1; + unsigned line_stipple_factor:8; /**< [1..256] actually */ + unsigned line_stipple_pattern:16; + unsigned bypass_clipping:1; + unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ + + float line_width; + float point_size; /**< used when no per-vertex size */ + float offset_units; + float offset_scale; + ubyte sprite_coord_mode[PIPE_MAX_SHADER_OUTPUTS]; /**< PIPE_SPRITE_COORD_ */ +}; + + +struct pipe_poly_stipple { + unsigned stipple[32]; +}; + + +struct pipe_viewport_state { + float scale[4]; + float translate[4]; +}; + + +struct pipe_scissor_state { + unsigned minx:16; + unsigned miny:16; + unsigned maxx:16; + unsigned maxy:16; +}; + + +struct pipe_clip_state { + float ucp[PIPE_MAX_CLIP_PLANES][4]; + unsigned nr; +}; + + +/** + * Constants for vertex/fragment shaders + */ +struct pipe_constant_buffer { + struct pipe_buffer *buffer; + unsigned size; /** in bytes */ +}; + + +struct pipe_shader_state { + const struct tgsi_token *tokens; + ubyte num_inputs; + ubyte num_outputs; + ubyte input_map[PIPE_MAX_SHADER_INPUTS]; /* XXX this may be temporary */ + ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ + ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ + ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; +}; + + +struct pipe_depth_stencil_alpha_state +{ + struct { + unsigned enabled:1; /**< depth test enabled? */ + unsigned writemask:1; /**< allow depth buffer writes? */ + unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ + unsigned occlusion_count:1; /**< do occlusion counting? */ + } depth; + struct { + unsigned enabled:1; + unsigned func:3; /**< PIPE_FUNC_x */ + unsigned fail_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned zpass_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned zfail_op:3; /**< PIPE_STENCIL_OP_x */ + ubyte ref_value; + ubyte value_mask; + ubyte write_mask; + } stencil[2]; /**< [0] = front, [1] = back */ + struct { + unsigned enabled:1; + unsigned func:3; /**< PIPE_FUNC_x */ + float ref; /**< reference value */ + } alpha; +}; + + +struct pipe_blend_state { + unsigned blend_enable:1; + + unsigned rgb_func:3; /**< PIPE_BLEND_x */ + unsigned rgb_src_factor:5; /**< PIPE_BLENDFACTOR_x */ + unsigned rgb_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ + + unsigned alpha_func:3; /**< PIPE_BLEND_x */ + unsigned alpha_src_factor:5; /**< PIPE_BLENDFACTOR_x */ + unsigned alpha_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ + + unsigned logicop_enable:1; + unsigned logicop_func:4; /**< PIPE_LOGICOP_x */ + + unsigned colormask:4; /**< bitmask of PIPE_MASK_R/G/B/A */ + unsigned dither:1; +}; + + +struct pipe_blend_color { + float color[4]; +}; + + +struct pipe_framebuffer_state +{ + /** multiple colorbuffers for multiple render targets */ + unsigned num_cbufs; + struct pipe_surface *cbufs[PIPE_MAX_COLOR_BUFS]; + + struct pipe_surface *zsbuf; /**< Z/stencil buffer */ +}; + + +/** + * Texture sampler state. + */ +struct pipe_sampler_state +{ + unsigned wrap_s:3; /**< PIPE_TEX_WRAP_x */ + unsigned wrap_t:3; /**< PIPE_TEX_WRAP_x */ + unsigned wrap_r:3; /**< PIPE_TEX_WRAP_x */ + unsigned min_img_filter:2; /**< PIPE_TEX_FILTER_x */ + unsigned min_mip_filter:2; /**< PIPE_TEX_MIPFILTER_x */ + unsigned mag_img_filter:2; /**< PIPE_TEX_FILTER_x */ + unsigned compare:1; /**< shadow/depth compare enabled? */ + unsigned compare_mode:1; /**< PIPE_TEX_COMPARE_x */ + unsigned compare_func:3; /**< PIPE_FUNC_x */ + unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */ + float shadow_ambient; /**< shadow test fail color/intensity */ + float lod_bias; /**< LOD/lambda bias */ + float min_lod, max_lod; /**< LOD clamp range, after bias */ + float border_color[4]; + float max_anisotropy; +}; + + +/** + * 2D surface. This is basically a view into a memory buffer. + * May be a renderbuffer, texture mipmap level, etc. + */ +struct pipe_surface +{ + struct pipe_buffer *buffer; /**< driver private buffer handle */ + enum pipe_format format; /**< PIPE_FORMAT_x */ + unsigned status; /**< PIPE_SURFACE_STATUS_x */ + unsigned clear_value; /**< may be temporary */ + unsigned cpp; /**< bytes per pixel */ + unsigned width, height; + unsigned pitch; /**< in pixels */ + unsigned offset; /**< offset from start of buffer, in bytes */ + unsigned refcount; + struct pipe_winsys *winsys; /**< winsys which owns/created the surface */ +}; + + +/** + * Texture. Represents one or several texture images on one or several mipmap + * levels. + */ +struct pipe_texture +{ + /* Effectively the key: + */ + enum pipe_texture_target target; /**< PIPE_TEXTURE_x */ + enum pipe_format format; /**< PIPE_FORMAT_x */ + + unsigned last_level; /**< Index of last mipmap level present/defined */ + + unsigned width[PIPE_MAX_TEXTURE_LEVELS]; + unsigned height[PIPE_MAX_TEXTURE_LEVELS]; + unsigned depth[PIPE_MAX_TEXTURE_LEVELS]; + unsigned cpp; + + unsigned compressed:1; + + /* These are also refcounted: + */ + unsigned refcount; +}; + + +/** + * A vertex buffer. Typically, all the vertex data/attributes for + * drawing something will be in one buffer. But it's also possible, for + * example, to put colors in one buffer and texcoords in another. + */ +struct pipe_vertex_buffer +{ + unsigned pitch:11; /**< stride to same attrib in next vertex, in bytes */ + unsigned max_index; /**< number of vertices in this buffer */ + unsigned buffer_offset; /**< offset to start of data in buffer, in bytes */ + struct pipe_buffer *buffer; /**< the actual buffer */ +}; + + +/** + * Information to describe a vertex attribute (position, color, etc) + */ +struct pipe_vertex_element +{ + /** Offset of this attribute, in bytes, from the start of the vertex */ + unsigned src_offset:11; + + /** Which vertex_buffer (as given to pipe->set_vertex_buffer()) does + * this attribute live in? + */ + unsigned vertex_buffer_index:5; + unsigned nr_components:3; + + enum pipe_format src_format; /**< PIPE_FORMAT_* */ +}; + + +#endif diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h new file mode 100644 index 0000000000..cd432c547c --- /dev/null +++ b/src/gallium/include/pipe/p_thread.h @@ -0,0 +1,54 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_THREAD_H +#define P_THREAD_H + +#include "p_compiler.h" + +/* + * XXX: We should come up with a system-independent thread definitions. + * XXX: Patching glthread defs for now. + */ + +#ifndef __MSC__ + +#include "glapi/glthread.h" + +#else /* __MSC__ */ + +typedef int _glthread_Mutex; + +#define _glthread_INIT_MUTEX( M ) ((void) (M)) +#define _glthread_LOCK_MUTEX( M ) ((void) (M)) +#define _glthread_UNLOCK_MUTEX( M ) ((void) (M)) + +#define sched_yield() ((void) 0) + +#endif /* __MSC__ */ + +#endif /* P_THREAD_H */ diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h new file mode 100644 index 0000000000..d7da2801c9 --- /dev/null +++ b/src/gallium/include/pipe/p_util.h @@ -0,0 +1,408 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_UTIL_H +#define P_UTIL_H + +#include "p_compiler.h" +#include "p_debug.h" +#include + + +#ifdef WIN32 + +#ifdef __cplusplus +extern "C" +{ +#endif + +void * __stdcall +EngAllocMem( + unsigned long Flags, + unsigned long MemSize, + unsigned long Tag ); + +void __stdcall +EngFreeMem( + void *Mem ); + +#ifdef __cplusplus +} +#endif + +static INLINE void * +MALLOC( unsigned size ) +{ + return EngAllocMem( 0, size, 'D3AG' ); +} + +static INLINE void * +CALLOC( unsigned count, unsigned size ) +{ + void *ptr = MALLOC( count * size ); + if( ptr ) { + memset( ptr, 0, count * size ); + } + return ptr; +} + +static INLINE void +FREE( void *ptr ) +{ + if( ptr ) { + EngFreeMem( ptr ); + } +} + +static INLINE void * +REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) +{ + void *new_ptr; + if( new_size <= old_size ) { + return old_ptr; + } + new_ptr = MALLOC( new_size ); + if( new_ptr ) { + memcpy( new_ptr, old_ptr, old_size ); + } + FREE( old_ptr ); + return new_ptr; +} + +#define GETENV( X ) NULL + +#else /* WIN32 */ + +#define MALLOC( SIZE ) malloc( SIZE ) + +#define CALLOC( COUNT, SIZE ) calloc( COUNT, SIZE ) + +#define FREE( PTR ) free( PTR ) + +#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) + +#define GETENV( X ) getenv( X ) + +#endif /* WIN32 */ + +#define MALLOC_STRUCT(T) (struct T *) MALLOC(sizeof(struct T)) + +#define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) + + +/** + * Return a pointer aligned to next multiple of N bytes. + */ +static INLINE void * +align_pointer( void *unaligned, uint alignment ) +{ + if (sizeof(void *) == 64) { + union { + void *p; + uint64 u; + } pu; + pu.p = unaligned; + pu.u = (pu.u + alignment - 1) & ~(uint64) (alignment - 1); + return pu.p; + } + else { + /* 32-bit pointers */ + union { + void *p; + uint u; + } pu; + pu.p = unaligned; + pu.u = (pu.u + alignment - 1) & ~(alignment - 1); + return pu.p; + } +} + +/** + * Return memory on given byte alignment + */ +static INLINE void * +align_malloc(size_t bytes, uint alignment) +{ +#if defined(HAVE_POSIX_MEMALIGN) + void *mem; + (void) posix_memalign(& mem, alignment, bytes); + return mem; +#else + char *ptr, *buf; + + assert( alignment > 0 ); + + ptr = (char *) MALLOC(bytes + alignment + sizeof(void *)); + if (!ptr) + return NULL; + + buf = (char *) align_pointer( ptr + sizeof(void *), alignment ); + *(char **)(buf - sizeof(void *)) = ptr; + + return buf; +#endif /* defined(HAVE_POSIX_MEMALIGN) */ +} + +/** + * Free memory returned by align_malloc(). + */ +static INLINE void +align_free(void *ptr) +{ +#if defined(HAVE_POSIX_MEMALIGN) + FREE(ptr); +#else + void **cubbyHole = (void **) ((char *) ptr - sizeof(void *)); + void *realAddr = *cubbyHole; + FREE(realAddr); +#endif /* defined(HAVE_POSIX_MEMALIGN) */ +} + + + +/** + * Duplicate a block of memory. + */ +static INLINE void * +mem_dup(const void *src, uint size) +{ + void *dup = MALLOC(size); + if (dup) + memcpy(dup, src, size); + return dup; +} + + + +#define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) +#define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) +#define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) + +#define Elements(x) sizeof(x)/sizeof(*(x)) +#define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER)) + +/** + * Return a pointer aligned to next multiple of 16 bytes. + */ +static INLINE void * +align16( void *unaligned ) +{ + return align_pointer( unaligned, 16 ); +} + + +static INLINE int align_int(int x, int align) +{ + return (x + align - 1) & ~(align - 1); +} + + + +#if defined(__MSC__) && defined(__WIN32__) +static INLINE unsigned ffs( unsigned u ) +{ + unsigned i; + + if( u == 0 ) { + return 0; + } + + __asm bsf eax, [u] + __asm inc eax + __asm mov [i], eax + + return i; +} +#endif + +union fi { + float f; + int i; + unsigned ui; +}; + +#define UBYTE_TO_FLOAT( ub ) ((float)(ub) / 255.0F) + +#define IEEE_0996 0x3f7f0000 /* 0.996 or so */ + +/* This function/macro is sensitive to precision. Test very carefully + * if you change it! + */ +#define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \ + do { \ + union fi __tmp; \ + __tmp.f = (F); \ + if (__tmp.i < 0) \ + UB = (ubyte) 0; \ + else if (__tmp.i >= IEEE_0996) \ + UB = (ubyte) 255; \ + else { \ + __tmp.f = __tmp.f * (255.0f/256.0f) + 32768.0f; \ + UB = (ubyte) __tmp.i; \ + } \ + } while (0) + + + +static INLINE unsigned pack_ub4( unsigned char b0, + unsigned char b1, + unsigned char b2, + unsigned char b3 ) +{ + return ((((unsigned int)b0) << 0) | + (((unsigned int)b1) << 8) | + (((unsigned int)b2) << 16) | + (((unsigned int)b3) << 24)); +} + +static INLINE unsigned fui( float f ) +{ + union fi fi; + fi.f = f; + return fi.ui; +} + +static INLINE unsigned char float_to_ubyte( float f ) +{ + unsigned char ub; + UNCLAMPED_FLOAT_TO_UBYTE(ub, f); + return ub; +} + +static INLINE unsigned pack_ui32_float4( float a, + float b, + float c, + float d ) +{ + return pack_ub4( float_to_ubyte(a), + float_to_ubyte(b), + float_to_ubyte(c), + float_to_ubyte(d) ); +} + +#define COPY_4V( DST, SRC ) \ +do { \ + (DST)[0] = (SRC)[0]; \ + (DST)[1] = (SRC)[1]; \ + (DST)[2] = (SRC)[2]; \ + (DST)[3] = (SRC)[3]; \ +} while (0) + + +#define COPY_4FV( DST, SRC ) COPY_4V(DST, SRC) + + +#define ASSIGN_4V( DST, V0, V1, V2, V3 ) \ +do { \ + (DST)[0] = (V0); \ + (DST)[1] = (V1); \ + (DST)[2] = (V2); \ + (DST)[3] = (V3); \ +} while (0) + + +static INLINE int ifloor(float f) +{ + int ai, bi; + double af, bf; + union fi u; + + af = (3 << 22) + 0.5 + (double)f; + bf = (3 << 22) + 0.5 - (double)f; + u.f = (float) af; ai = u.i; + u.f = (float) bf; bi = u.i; + return (ai - bi) >> 1; +} + + +#if defined(__GNUC__) && defined(__i386__) +static INLINE int iround(float f) +{ + int r; + __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st"); + return r; +} +#elif defined(__MSC__) && defined(__WIN32__) +static INLINE int iround(float f) +{ + int r; + _asm { + fld f + fistp r + } + return r; +} +#else +#define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F))) +#endif + + +/* Could maybe have an inline version of this? + */ +#if defined(__GNUC__) +#define FABSF(x) fabsf(x) +#else +#define FABSF(x) ((float) fabs(x)) +#endif + +/* Pretty fast, and accurate. + * Based on code from http://www.flipcode.com/totd/ + */ +static INLINE float LOG2(float val) +{ + union fi num; + int log_2; + + num.f = val; + log_2 = ((num.i >> 23) & 255) - 128; + num.i &= ~(255 << 23); + num.i += 127 << 23; + num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3; + return num.f + log_2; +} + +#if defined(__GNUC__) +#define CEILF(x) ceilf(x) +#else +#define CEILF(x) ((float) ceil(x)) +#endif + +static INLINE int align(int value, int alignment) +{ + return (value + alignment - 1) & ~(alignment - 1); +} + + +/* util/p_util.c + */ +extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, + unsigned dst_x, unsigned dst_y, unsigned width, + unsigned height, const ubyte * src, + int src_pitch, unsigned src_x, int src_y); + + +#endif diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h new file mode 100644 index 0000000000..1e81eebd78 --- /dev/null +++ b/src/gallium/include/pipe/p_winsys.h @@ -0,0 +1,160 @@ + /************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_WINSYS_H +#define P_WINSYS_H + + +#include "p_format.h" + +/** + * \file + * This is the interface that Gallium3D requires any window system + * hosting it to implement. This is the only include file in Gallium3D + * which is public. + */ + + +/** Opaque type */ +struct pipe_fence_handle; + +struct pipe_surface; + + +/** + * Gallium3D drivers are (meant to be!) independent of both GL and the + * window system. The window system provides a buffer manager and a + * set of additional hooks for things like command buffer submission, + * etc. + * + * There clearly has to be some agreement between the window system + * driver and the hardware driver about the format of command buffers, + * etc. + */ +struct pipe_winsys +{ + /** Returns name of this winsys interface */ + const char *(*get_name)( struct pipe_winsys *sws ); + + /** + * Do any special operations to ensure frontbuffer contents are + * displayed, eg copy fake frontbuffer. + */ + void (*flush_frontbuffer)( struct pipe_winsys *sws, + struct pipe_surface *surf, + void *context_private ); + + /** Debug output */ + void (*printf)( struct pipe_winsys *sws, + const char *, ... ); + + + /** allocate a new surface (no context dependency) */ + struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws); + + /** + * Allocate storage for a pipe_surface. + * Returns 0 if succeeds. + */ + int (*surface_alloc_storage)(struct pipe_winsys *ws, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags); + + void (*surface_release)(struct pipe_winsys *ws, struct pipe_surface **s); + + + /** + * Buffer management. Buffer attributes are mostly fixed over its lifetime. + * + * Remember that gallium gets to choose the interface it needs, and the + * window systems must then implement that interface (rather than the + * other way around...). + * + * usage is a bitmask of PIPE_BUFFER_USAGE_PIXEL/VERTEX/INDEX/CONSTANT. This + * usage argument is only an optimization hint, not a guarantee, therefore + * proper behavior must be observed in all circumstances. + */ + struct pipe_buffer *(*buffer_create)( struct pipe_winsys *sws, + unsigned alignment, + unsigned usage, + unsigned size ); + + /** Create a buffer that wraps user-space data */ + struct pipe_buffer *(*user_buffer_create)(struct pipe_winsys *sws, + void *ptr, + unsigned bytes); + + /** + * Map the entire data store of a buffer object into the client's address. + * flags is bitmask of PIPE_BUFFER_USAGE_CPU_READ/WRITE flags. + */ + void *(*buffer_map)( struct pipe_winsys *sws, + struct pipe_buffer *buf, + unsigned usage ); + + void (*buffer_unmap)( struct pipe_winsys *sws, + struct pipe_buffer *buf ); + + void (*buffer_destroy)( struct pipe_winsys *sws, + struct pipe_buffer *buf ); + + + /** Set ptr = fence, with reference counting */ + void (*fence_reference)( struct pipe_winsys *sws, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence ); + + /** + * Checks whether the fence has been signalled. + * + * The meaning of flag is pipe-driver specific. + * + * Returns zero if it has. + */ + int (*fence_signalled)( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ); + + /** + * Wait for the fence to finish. + * + * The meaning of flag is pipe-driver specific. + * + * Returns zero on success. + */ + int (*fence_finish)( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ); + + +}; + + + +#endif /* P_WINSYS_H */ diff --git a/src/gallium/winsys/dri/intel/Makefile b/src/gallium/winsys/dri/intel/Makefile new file mode 100644 index 0000000000..9ae0f01325 --- /dev/null +++ b/src/gallium/winsys/dri/intel/Makefile @@ -0,0 +1,38 @@ + +TOP = ../../../../.. +include $(TOP)/configs/current + +LIBNAME = i915tex_dri.so + +MINIGLX_SOURCES = server/intel_dri.c + +PIPE_DRIVERS = \ + $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a \ + $(TOP)/src/mesa/pipe/i915simple/libi915simple.a + +DRIVER_SOURCES = \ + intel_winsys_pipe.c \ + intel_winsys_softpipe.c \ + intel_winsys_i915.c \ + intel_batchbuffer.c \ + intel_swapbuffers.c \ + intel_context.c \ + intel_lock.c \ + intel_screen.c \ + intel_batchpool.c + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(COMMON_BM_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") + +include ../Makefile.template + +intel_tex_layout.o: ../intel/intel_tex_layout.c + +symlinks: diff --git a/src/gallium/winsys/dri/intel/SConscript b/src/gallium/winsys/dri/intel/SConscript new file mode 100644 index 0000000000..a7cc10450e --- /dev/null +++ b/src/gallium/winsys/dri/intel/SConscript @@ -0,0 +1,41 @@ +Import('*') + +env = drienv.Clone() + +env.Append(CPPPATH = [ + '../intel', + 'server' +]) + +#MINIGLX_SOURCES = server/intel_dri.c + +pipe_drivers = [ + softpipe, + i915simple +] + +DRIVER_SOURCES = [ + 'intel_winsys_pipe.c', + 'intel_winsys_softpipe.c', + 'intel_winsys_i915.c', + 'intel_batchbuffer.c', + 'intel_swapbuffers.c', + 'intel_context.c', + 'intel_lock.c', + 'intel_screen.c', + 'intel_batchpool.c', +] + +sources = \ + COMMON_GALLIUM_SOURCES + \ + COMMON_BM_SOURCES + \ + DRIVER_SOURCES + +# DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ +# && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") + +env.SharedLibrary( + target ='i915tex_dri.so', + source = sources, + LIBS = pipe_drivers + env['LIBS'], +) \ No newline at end of file diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c new file mode 100644 index 0000000000..49e04d81ec --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.c @@ -0,0 +1,357 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include +#include "intel_batchbuffer.h" +#include "intel_context.h" +#include "intel_screen.h" +#include "intel_reg.h" +#include "drm.h" + +/* Relocations in kernel space: + * - pass dma buffer seperately + * - memory manager knows how to patch + * - pass list of dependent buffers + * - pass relocation list + * + * Either: + * - get back an offset for buffer to fire + * - memory manager knows how to fire buffer + * + * Really want the buffer to be AGP and pinned. + * + */ + +/* Cliprect fence: The highest fence protecting a dma buffer + * containing explicit cliprect information. Like the old drawable + * lock but irq-driven. X server must wait for this fence to expire + * before changing cliprects [and then doing sw rendering?]. For + * other dma buffers, the scheduler will grab current cliprect info + * and mix into buffer. X server must hold the lock while changing + * cliprects??? Make per-drawable. Need cliprects in shared memory + * -- beats storing them with every cmd buffer in the queue. + * + * ==> X server must wait for this fence to expire before touching the + * framebuffer with new cliprects. + * + * ==> Cliprect-dependent buffers associated with a + * cliprect-timestamp. All of the buffers associated with a timestamp + * must go to hardware before any buffer with a newer timestamp. + * + * ==> Dma should be queued per-drawable for correct X/GL + * synchronization. Or can fences be used for this? + * + * Applies to: Blit operations, metaops, X server operations -- X + * server automatically waits on its own dma to complete before + * modifying cliprects ??? + */ + +static void +intel_dump_batchbuffer(uint offset, uint * ptr, uint count) +{ + int i; + printf("\n\n\nSTART BATCH (%d dwords):\n", count / 4); + for (i = 0; i < count / 4; i += 1) + printf("\t0x%08x\n", ptr[i]); + printf("END BATCH\n\n\n"); +} + + +void +intel_batchbuffer_reset(struct intel_batchbuffer *batch) +{ + int i; + + if (batch->map) { + driBOUnmap(batch->buffer); + batch->map = NULL; + } + + /* + * Get a new, free batchbuffer. + */ + batch->size = BATCH_SZ; + driBOData(batch->buffer, batch->size, NULL, 0); + + driBOResetList(&batch->list); + + /* + * Unreference buffers previously on the relocation list. + */ + for (i = 0; i < batch->nr_relocs; i++) { + struct buffer_reloc *r = &batch->reloc[i]; + driBOUnReference(r->buf); + } + + batch->list_count = 0; + batch->nr_relocs = 0; + batch->flags = 0; + + /* + * We don't refcount the batchbuffer itself since we can't destroy it + * while it's on the list. + */ + + driBOAddListItem(&batch->list, batch->buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE); + + + batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); + batch->ptr = batch->map; +} + + +/*====================================================================== + * Public functions + */ +struct intel_batchbuffer * +intel_batchbuffer_alloc(struct intel_context *intel) +{ + struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); + + batch->intel = intel; + + driGenBuffers(intel->intelScreen->batchPool, "batchbuffer", 1, + &batch->buffer, 4096, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); + batch->last_fence = NULL; + driBOCreateList(20, &batch->list); + intel_batchbuffer_reset(batch); + return batch; +} + + +void +intel_batchbuffer_free(struct intel_batchbuffer *batch) +{ + if (batch->last_fence) { + driFenceFinish(batch->last_fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, + GL_FALSE); + driFenceUnReference(batch->last_fence); + batch->last_fence = NULL; + } + if (batch->map) { + driBOUnmap(batch->buffer); + batch->map = NULL; + } + driBOUnReference(batch->buffer); + batch->buffer = NULL; + free(batch); +} + + +static void +intel_batch_ioctl(struct intel_context *intel, + uint start_offset, uint used, boolean allow_unlock) +{ + drmI830BatchBuffer batch; + + batch.start = start_offset; + batch.used = used; + batch.cliprects = NULL; /* unused */ + batch.num_cliprects = 0; + batch.DR1 = 0; + batch.DR4 = 0; /* still need this ? */ + + DBG(IOCTL, "%s: 0x%x..0x%x DR4: %x cliprects: %d\n", + __FUNCTION__, + batch.start, + batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); + + if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, + sizeof(batch))) { + printf("DRM_I830_BATCHBUFFER: %d\n", -errno); + UNLOCK_HARDWARE(intel); + exit(1); + } +} + + +/* TODO: Push this whole function into bufmgr. + */ +static void +do_flush_locked(struct intel_batchbuffer *batch, + uint used, boolean allow_unlock) +{ + uint *ptr; + uint i, fenceFlags; + struct _DriFenceObject *fo; + + driBOValidateList(batch->intel->driFd, &batch->list); + + /* Apply the relocations. This nasty map indicates to me that the + * whole task should be done internally by the memory manager, and + * that dma buffers probably need to be pinned within agp space. + */ + ptr = (uint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, + DRM_BO_HINT_ALLOW_UNFENCED_MAP); + + for (i = 0; i < batch->nr_relocs; i++) { + struct buffer_reloc *r = &batch->reloc[i]; + + ptr[r->offset / 4] = driBOOffset(r->buf) + r->delta; + } + + if (0) + intel_dump_batchbuffer(0, ptr, used); + + driBOUnmap(batch->buffer); + batch->map = NULL; + + intel_batch_ioctl(batch->intel, + driBOOffset(batch->buffer), + used, allow_unlock); + + /* + * Kernel fencing. The flags tells the kernel that we've + * programmed an MI_FLUSH. + */ + fenceFlags = DRM_I915_FENCE_FLAG_FLUSHED; + fo = driFenceBuffers(batch->intel->driFd, "Batch fence", fenceFlags); + + /* + * User space fencing. + */ + driBOFence(batch->buffer, fo); + + if (driFenceType(fo) == DRM_FENCE_TYPE_EXE) { + /* + * Oops. We only validated a batch buffer. This means we + * didn't do any proper rendering. Discard this fence object. + */ + driFenceUnReference(fo); + } + else { + driFenceUnReference(batch->last_fence); + batch->last_fence = fo; + for (i = 0; i < batch->nr_relocs; i++) { + struct buffer_reloc *r = &batch->reloc[i]; + driBOFence(r->buf, fo); + } + } +} + + +struct _DriFenceObject * +intel_batchbuffer_flush(struct intel_batchbuffer *batch) +{ + struct intel_context *intel = batch->intel; + uint used = batch->ptr - batch->map; + const boolean was_locked = intel->locked; + + if (used == 0) + return batch->last_fence; + +#define MI_FLUSH ((0 << 29) | (4 << 23)) + + /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a + * performance drain that we would like to avoid. + */ + if (used & 4) { + ((int *) batch->ptr)[0] = MI_FLUSH; + ((int *) batch->ptr)[1] = 0; + ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->ptr)[0] = MI_FLUSH; + ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; + used += 8; + } + + driBOUnmap(batch->buffer); + batch->ptr = NULL; + batch->map = NULL; + + /* TODO: Just pass the relocation list and dma buffer up to the + * kernel. + */ + if (!was_locked) + LOCK_HARDWARE(intel); + + do_flush_locked(batch, used, GL_FALSE); + + if (!was_locked) + UNLOCK_HARDWARE(intel); + + /* Reset the buffer: + */ + intel_batchbuffer_reset(batch); + return batch->last_fence; +} + + +void +intel_batchbuffer_finish(struct intel_batchbuffer *batch) +{ + struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); + if (fence) { + driFenceReference(fence); + driFenceFinish(fence, + DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, + GL_FALSE); + driFenceUnReference(fence); + } +} + + +/* This is the only way buffers get added to the validate list. + */ +boolean +intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, + struct _DriBufferObject *buffer, + uint flags, uint mask, uint delta) +{ + assert(batch->nr_relocs < MAX_RELOCS); + + driBOAddListItem(&batch->list, buffer, flags, mask); + + { + struct buffer_reloc *r = &batch->reloc[batch->nr_relocs++]; + driBOReference(buffer); + r->buf = buffer; + r->offset = batch->ptr - batch->map; + r->delta = delta; + *(uint *) batch->ptr = 0x12345678; + } + + batch->ptr += 4; + return GL_TRUE; +} + + +void +intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, uint bytes, uint flags) +{ + assert((bytes & 3) == 0); + intel_batchbuffer_require_space(batch, bytes, flags); + memcpy(batch->ptr, data, bytes); + batch->ptr += bytes; +} diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.h b/src/gallium/winsys/dri/intel/intel_batchbuffer.h new file mode 100644 index 0000000000..82feafa21f --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.h @@ -0,0 +1,149 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BATCHBUFFER_H +#define INTEL_BATCHBUFFER_H + +#include "pipe/p_compiler.h" +#include "dri_bufmgr.h" + +struct intel_context; + +#define BATCH_SZ 16384 +#define BATCH_RESERVED 16 + +#define MAX_RELOCS 4096 + +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 + +struct buffer_reloc +{ + struct _DriBufferObject *buf; + uint offset; + uint delta; /* not needed? */ +}; + +struct intel_batchbuffer +{ + struct bufmgr *bm; + struct intel_context *intel; + + struct _DriBufferObject *buffer; + struct _DriFenceObject *last_fence; + uint flags; + + drmBOList list; + uint list_count; + ubyte *map; + ubyte *ptr; + + struct buffer_reloc reloc[MAX_RELOCS]; + uint nr_relocs; + uint size; +}; + +struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context *intel); + +void intel_batchbuffer_free(struct intel_batchbuffer *batch); + + +void intel_batchbuffer_finish(struct intel_batchbuffer *batch); + +struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer + *batch); + +void intel_batchbuffer_reset(struct intel_batchbuffer *batch); + + +/* Unlike bmBufferData, this currently requires the buffer be mapped. + * Consider it a convenience function wrapping multiple + * intel_buffer_dword() calls. + */ +void intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, uint bytes, uint flags); + +void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, + uint bytes); + +boolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, + struct _DriBufferObject *buffer, + uint flags, + uint mask, uint offset); + +/* Inline functions - might actually be better off with these + * non-inlined. Certainly better off switching all command packets to + * be passed as structs rather than dwords, but that's a little bit of + * work... + */ +static INLINE uint +intel_batchbuffer_space(struct intel_batchbuffer *batch) +{ + return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); +} + + +static INLINE void +intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, uint dword) +{ + assert(batch->map); + assert(intel_batchbuffer_space(batch) >= 4); + *(uint *) (batch->ptr) = dword; + batch->ptr += 4; +} + +static INLINE void +intel_batchbuffer_require_space(struct intel_batchbuffer *batch, + uint sz, uint flags) +{ + assert(sz < batch->size - 8); + if (intel_batchbuffer_space(batch) < sz || + (batch->flags != 0 && flags != 0 && batch->flags != flags)) + intel_batchbuffer_flush(batch); + + batch->flags |= flags; +} + +/* Here are the crusty old macros, to be removed: + */ +#define BATCH_LOCALS + +#define BEGIN_BATCH(n, flags) do { \ + intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ +} while (0) + +#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) + +#define OUT_RELOC(buf,flags,mask,delta) do { \ + assert((delta) >= 0); \ + intel_batchbuffer_emit_reloc(intel->batch, buf, flags, mask, delta); \ +} while (0) + +#define ADVANCE_BATCH() do { } while(0) + + +#endif diff --git a/src/gallium/winsys/dri/intel/intel_batchpool.c b/src/gallium/winsys/dri/intel/intel_batchpool.c new file mode 100644 index 0000000000..33b56817f6 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_batchpool.c @@ -0,0 +1,424 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +/** + * XXX NOTE: there are no intel dependencies in this file. + * Rename to dri_batchpool.c? + */ + +#include +#include +#include +#include "imports.h" +#include "glthread.h" +#include "dri_bufpool.h" +#include "dri_bufmgr.h" +#include "intel_batchpool.h" + + +typedef struct +{ + drmMMListHead head; + struct _BPool *parent; + struct _DriFenceObject *fence; + unsigned long start; + int unfenced; + int mapped; +} BBuf; + +typedef struct _BPool +{ + _glthread_Mutex mutex; + unsigned long bufSize; + unsigned poolSize; + unsigned numFree; + unsigned numTot; + unsigned numDelayed; + unsigned checkDelayed; + drmMMListHead free; + drmMMListHead delayed; + drmMMListHead head; + drmBO kernelBO; + void *virtual; + BBuf *bufs; +} BPool; + + +static BPool * +createBPool(int fd, unsigned long bufSize, unsigned numBufs, unsigned flags, + unsigned checkDelayed) +{ + BPool *p = (BPool *) malloc(sizeof(*p)); + BBuf *buf; + int i; + + if (!p) + return NULL; + + p->bufs = (BBuf *) malloc(numBufs * sizeof(*p->bufs)); + if (!p->bufs) { + free(p); + return NULL; + } + + DRMINITLISTHEAD(&p->free); + DRMINITLISTHEAD(&p->head); + DRMINITLISTHEAD(&p->delayed); + + p->numTot = numBufs; + p->numFree = numBufs; + p->bufSize = bufSize; + p->numDelayed = 0; + p->checkDelayed = checkDelayed; + + _glthread_INIT_MUTEX(p->mutex); + + if (drmBOCreate(fd, 0, numBufs * bufSize, 0, NULL, drm_bo_type_dc, + flags, DRM_BO_HINT_DONT_FENCE, &p->kernelBO)) { + free(p->bufs); + free(p); + return NULL; + } + if (drmBOMap(fd, &p->kernelBO, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, + &p->virtual)) { + drmBODestroy(fd, &p->kernelBO); + free(p->bufs); + free(p); + return NULL; + } + + /* + * We unmap the buffer so that we can validate it later. Note that this is + * just a synchronizing operation. The buffer will have a virtual mapping + * until it is destroyed. + */ + + drmBOUnmap(fd, &p->kernelBO); + + buf = p->bufs; + for (i = 0; i < numBufs; ++i) { + buf->parent = p; + buf->fence = NULL; + buf->start = i * bufSize; + buf->mapped = 0; + buf->unfenced = 0; + DRMLISTADDTAIL(&buf->head, &p->free); + buf++; + } + + return p; +} + + +static void +pool_checkFree(BPool * p, int wait) +{ + drmMMListHead *list, *prev; + BBuf *buf; + int signaled = 0; + int i; + + list = p->delayed.next; + + if (p->numDelayed > 3) { + for (i = 0; i < p->numDelayed; i += 3) { + list = list->next; + } + } + + prev = list->prev; + for (; list != &p->delayed; list = prev, prev = list->prev) { + + buf = DRMLISTENTRY(BBuf, list, head); + + if (!signaled) { + if (wait) { + driFenceFinish(buf->fence, DRM_FENCE_TYPE_EXE, 1); + signaled = 1; + } + else { + signaled = driFenceSignaled(buf->fence, DRM_FENCE_TYPE_EXE); + } + } + + if (!signaled) + break; + + driFenceUnReference(buf->fence); + buf->fence = NULL; + DRMLISTDEL(list); + p->numDelayed--; + DRMLISTADD(list, &p->free); + p->numFree++; + } +} + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, unsigned flags, unsigned hint, + unsigned alignment) +{ + BPool *p = (BPool *) pool->data; + + drmMMListHead *item; + + if (alignment && (alignment != 4096)) + return NULL; + + _glthread_LOCK_MUTEX(p->mutex); + + if (p->numFree == 0) + pool_checkFree(p, GL_TRUE); + + if (p->numFree == 0) { + fprintf(stderr, "Out of fixed size buffer objects\n"); + BM_CKFATAL(-ENOMEM); + } + + item = p->free.next; + + if (item == &p->free) { + fprintf(stderr, "Fixed size buffer pool corruption\n"); + } + + DRMLISTDEL(item); + --p->numFree; + + _glthread_UNLOCK_MUTEX(p->mutex); + return (void *) DRMLISTENTRY(BBuf, item, head); +} + + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + _glthread_LOCK_MUTEX(p->mutex); + + if (buf->fence) { + DRMLISTADDTAIL(&buf->head, &p->delayed); + p->numDelayed++; + } + else { + buf->unfenced = 0; + DRMLISTADD(&buf->head, &p->free); + p->numFree++; + } + + if ((p->numDelayed % p->checkDelayed) == 0) + pool_checkFree(p, 0); + + _glthread_UNLOCK_MUTEX(p->mutex); + return 0; +} + + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, void **virtual) +{ + + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + _glthread_LOCK_MUTEX(p->mutex); + + /* + * Currently Mesa doesn't have any condition variables to resolve this + * cleanly in a multithreading environment. + * We bail out instead. + */ + + if (buf->mapped) { + fprintf(stderr, "Trying to map already mapped buffer object\n"); + BM_CKFATAL(-EINVAL); + } + +#if 0 + if (buf->unfenced && !(hint & DRM_BO_HINT_ALLOW_UNFENCED_MAP)) { + fprintf(stderr, "Trying to map an unfenced buffer object 0x%08x" + " 0x%08x %d\n", hint, flags, buf->start); + BM_CKFATAL(-EINVAL); + } + +#endif + + if (buf->fence) { + _glthread_UNLOCK_MUTEX(p->mutex); + return -EBUSY; + } + + buf->mapped = GL_TRUE; + *virtual = (unsigned char *) p->virtual + buf->start; + _glthread_UNLOCK_MUTEX(p->mutex); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, int lazy) +{ + BBuf *buf = (BBuf *) private; + driFenceFinish(buf->fence, 0x0, lazy); + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + + buf->mapped = 0; + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + return p->kernelBO.offset + buf->start; +} + +static unsigned +pool_flags(struct _DriBufferPool *pool, void *private) +{ + BPool *p = (BPool *) pool->data; + + return p->kernelBO.flags; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + BPool *p = (BPool *) pool->data; + + return p->bufSize; +} + + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + _glthread_LOCK_MUTEX(p->mutex); + if (buf->fence) { + driFenceUnReference(buf->fence); + } + buf->fence = fence; + buf->unfenced = 0; + driFenceReference(buf->fence); + _glthread_UNLOCK_MUTEX(p->mutex); + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + + return &p->kernelBO; +} + +static int +pool_validate(struct _DriBufferPool *pool, void *private) +{ + BBuf *buf = (BBuf *) private; + BPool *p = buf->parent; + _glthread_LOCK_MUTEX(p->mutex); + buf->unfenced = GL_TRUE; + _glthread_UNLOCK_MUTEX(p->mutex); + return 0; +} + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + BPool *p = (BPool *) pool->data; + + /* + * Wait on outstanding fences. + */ + + _glthread_LOCK_MUTEX(p->mutex); + while ((p->numFree < p->numTot) && p->numDelayed) { + _glthread_UNLOCK_MUTEX(p->mutex); + sched_yield(); + pool_checkFree(p, GL_TRUE); + _glthread_LOCK_MUTEX(p->mutex); + } + + drmBODestroy(pool->fd, &p->kernelBO); + free(p->bufs); + _glthread_UNLOCK_MUTEX(p->mutex); + free(p); + free(pool); +} + + +struct _DriBufferPool * +driBatchPoolInit(int fd, unsigned flags, + unsigned long bufSize, + unsigned numBufs, unsigned checkDelayed) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + if (!pool) + return NULL; + + pool->data = createBPool(fd, bufSize, numBufs, flags, checkDelayed); + if (!pool->data) + return NULL; + + pool->fd = fd; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = &pool_validate; + pool->waitIdle = &pool_waitIdle; + pool->setstatic = NULL; + pool->takeDown = &pool_takedown; + return pool; +} diff --git a/src/gallium/winsys/dri/intel/intel_batchpool.h b/src/gallium/winsys/dri/intel/intel_batchpool.h new file mode 100644 index 0000000000..f6a95723bc --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_batchpool.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_BATCHPOOL_H +#define INTEL_BATCHPOOL_H + +extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, + unsigned long bufSize, + unsigned numBufs, + unsigned checkDelayed); + + +#endif /* INTEL_BATCHPOOL_H */ diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c new file mode 100644 index 0000000000..c033f2a592 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -0,0 +1,304 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "i830_dri.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_swapbuffers.h" +#include "intel_winsys.h" +#include "intel_batchbuffer.h" + +#include "state_tracker/st_public.h" +#include "pipe/p_defines.h" +#include "pipe/p_context.h" + +#include "utils.h" + + +#ifdef DEBUG +int __intel_debug = 0; +#endif + + +#define need_GL_ARB_multisample +#define need_GL_ARB_point_parameters +#define need_GL_ARB_texture_compression +#define need_GL_ARB_vertex_buffer_object +#define need_GL_ARB_vertex_program +#define need_GL_ARB_window_pos +#define need_GL_EXT_blend_color +#define need_GL_EXT_blend_equation_separate +#define need_GL_EXT_blend_func_separate +#define need_GL_EXT_blend_minmax +#define need_GL_EXT_cull_vertex +#define need_GL_EXT_fog_coord +#define need_GL_EXT_framebuffer_object +#define need_GL_EXT_multi_draw_arrays +#define need_GL_EXT_secondary_color +#define need_GL_NV_vertex_program +#include "extension_helper.h" + + +/** + * Extension strings exported by the intel driver. + * + * \note + * It appears that ARB_texture_env_crossbar has "disappeared" compared to the + * old i830-specific driver. + */ +const struct dri_extension card_extensions[] = { + {"GL_ARB_multisample", GL_ARB_multisample_functions}, + {"GL_ARB_multitexture", NULL}, + {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_texture_border_clamp", NULL}, + {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, + {"GL_ARB_texture_cube_map", NULL}, + {"GL_ARB_texture_env_add", NULL}, + {"GL_ARB_texture_env_combine", NULL}, + {"GL_ARB_texture_env_dot3", NULL}, + {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_rectangle", NULL}, + {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, + {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, + {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, + {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, + {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_subtract", NULL}, + {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, + {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, + {"GL_EXT_packed_depth_stencil", NULL}, + {"GL_EXT_pixel_buffer_object", NULL}, + {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, + {"GL_EXT_stencil_wrap", NULL}, + {"GL_EXT_texture_edge_clamp", NULL}, + {"GL_EXT_texture_env_combine", NULL}, + {"GL_EXT_texture_env_dot3", NULL}, + {"GL_EXT_texture_filter_anisotropic", NULL}, + {"GL_EXT_texture_lod_bias", NULL}, + {"GL_3DFX_texture_compression_FXT1", NULL}, + {"GL_APPLE_client_storage", NULL}, + {"GL_MESA_pack_invert", NULL}, + {"GL_MESA_ycbcr_texture", NULL}, + {"GL_NV_blend_square", NULL}, + {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, + {"GL_NV_vertex_program1_1", NULL}, + {"GL_SGIS_generate_mipmap", NULL }, + {NULL, NULL} +}; + + + +#ifdef DEBUG +static const struct dri_debug_control debug_control[] = { + {"ioctl", DEBUG_IOCTL}, + {"bat", DEBUG_BATCH}, + {"lock", DEBUG_LOCK}, + {"swap", DEBUG_SWAP}, + {NULL, 0} +}; +#endif + + + +GLboolean +intelCreateContext(const __GLcontextModes * visual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) +{ + struct intel_context *intel = CALLOC_STRUCT(intel_context); + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + struct intel_screen *intelScreen = intel_screen(sPriv); + drmI830Sarea *saPriv = intelScreen->sarea; + int fthrottle_mode; + GLboolean havePools; + struct pipe_context *pipe; + struct st_context *st_share = NULL; + + if (sharedContextPrivate) { + st_share = ((struct intel_context *) sharedContextPrivate)->st; + } + + driContextPriv->driverPrivate = intel; + intel->intelScreen = intelScreen; + intel->driScreen = sPriv; + intel->sarea = saPriv; + + driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, + intel->driScreen->myNum, "i915"); + + + /* + * memory pools + */ + DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + havePools = intelCreatePools(sPriv); + DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + if (!havePools) + return GL_FALSE; + + + /* Dri stuff */ + intel->hHWContext = driContextPriv->hHWContext; + intel->driFd = sPriv->fd; + intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; + + fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); + intel->iw.irq_seq = -1; + intel->irqsEmitted = 0; + + intel->batch = intel_batchbuffer_alloc(intel); + intel->last_swap_fence = NULL; + intel->first_swap_fence = NULL; + +#ifdef DEBUG + __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); +#endif + + /* + * Pipe-related setup + */ + if (!getenv("INTEL_HW")) { + pipe = intel_create_softpipe( intel, intelScreen->winsys ); + } + else { + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + pipe = intel_create_i915simple( intel, intelScreen->winsys ); + break; + default: + fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", + intel->intelScreen->deviceID, __FUNCTION__); + + pipe = intel_create_softpipe( intel, intelScreen->winsys ); + break; + } + } + + pipe->priv = intel; + + intel->st = st_create_context(pipe, visual, st_share); + + return GL_TRUE; +} + + +void +intelDestroyContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = intel_context(driContextPriv); + + assert(intel); /* should never be null */ + if (intel) { + st_flush(intel->st, PIPE_FLUSH_WAIT); + + intel_batchbuffer_free(intel->batch); + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + if (intel->first_swap_fence) { + driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(intel->first_swap_fence); + intel->first_swap_fence = NULL; + } + + if (intel->intelScreen->dummyContext == intel) + intel->intelScreen->dummyContext = NULL; + + st_destroy_context(intel->st); + free(intel); + } +} + + +GLboolean +intelUnbindContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = intel_context(driContextPriv); + st_flush(intel->st, 0x0); + /* XXX make_current(NULL)? */ + return GL_TRUE; +} + + +GLboolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) +{ + if (driContextPriv) { + struct intel_context *intel = intel_context(driContextPriv); + struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv); + struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv); + + assert(draw_fb->stfb); + assert(read_fb->stfb); + + /* This is for situations in which we need a rendering context but + * there may not be any currently bound. + */ + intel->intelScreen->dummyContext = intel; + + st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); + + if ((intel->driDrawable != driDrawPriv) || + (intel->lastStamp != driDrawPriv->lastStamp)) { + intel->driDrawable = driDrawPriv; + intelUpdateWindowSize(driDrawPriv); + intel->lastStamp = driDrawPriv->lastStamp; + } + + /* The size of the draw buffer will have been updated above. + * If the readbuffer is a different window, check/update its size now. + */ + if (driReadPriv != driDrawPriv) { + intelUpdateWindowSize(driReadPriv); + } + + } + else { + st_make_current(NULL, NULL, NULL); + } + + return GL_TRUE; +} diff --git a/src/gallium/winsys/dri/intel/intel_context.h b/src/gallium/winsys/dri/intel/intel_context.h new file mode 100644 index 0000000000..b01370c049 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_context.h @@ -0,0 +1,158 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_CONTEXT_H +#define INTEL_CONTEXT_H + + +#include "drm.h" +#include "intel_screen.h" +#include "i915_drm.h" + + +struct pipe_context; +struct intel_context; +struct _DriBufferObject; +struct st_context; + + +#define INTEL_MAX_FIXUP 64 + +/** + * Intel rendering context, contains a state tracker and intel-specific info. + */ +struct intel_context +{ + struct st_context *st; + + struct _DriFenceObject *last_swap_fence; + struct _DriFenceObject *first_swap_fence; + + struct intel_batchbuffer *batch; + + boolean locked; + char *prevLockFile; + int prevLockLine; + + uint irqsEmitted; + drm_i915_irq_wait_t iw; + + drm_context_t hHWContext; + drmLock *driHwLock; + int driFd; + + __DRIdrawablePrivate *driDrawable; + __DRIscreenPrivate *driScreen; + struct intel_screen *intelScreen; + drmI830Sarea *sarea; + + uint lastStamp; + + /** + * Configuration cache + */ + driOptionCache optionCache; +}; + + + +/** + * Intel framebuffer. + */ +struct intel_framebuffer +{ + struct st_framebuffer *stfb; + + /* other fields TBD */ + int other; +}; + + + + +/* These are functions now: + */ +void LOCK_HARDWARE( struct intel_context *intel ); +void UNLOCK_HARDWARE( struct intel_context *intel ); + +extern char *__progname; + + + +/* ================================================================ + * Debugging: + */ +#ifdef DEBUG +extern int __intel_debug; + +#define DEBUG_SWAP 0x1 +#define DEBUG_LOCK 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_BATCH 0x8 + +#define DBG(flag, ...) do { \ + if (__intel_debug & (DEBUG_##flag)) \ + printf(__VA_ARGS__); \ +} while(0) + +#else +#define DBG(flag, ...) +#endif + + + +#define PCI_CHIP_845_G 0x2562 +#define PCI_CHIP_I830_M 0x3577 +#define PCI_CHIP_I855_GM 0x3582 +#define PCI_CHIP_I865_G 0x2572 +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GME 0x27AE +#define PCI_CHIP_G33_G 0x29C2 +#define PCI_CHIP_Q35_G 0x29B2 +#define PCI_CHIP_Q33_G 0x29D2 + + +/** Cast wrapper */ +static INLINE struct intel_context * +intel_context(__DRIcontextPrivate *driContextPriv) +{ + return (struct intel_context *) driContextPriv->driverPrivate; +} + + +/** Cast wrapper */ +static INLINE struct intel_framebuffer * +intel_framebuffer(__DRIdrawablePrivate * driDrawPriv) +{ + return (struct intel_framebuffer *) driDrawPriv->driverPrivate; +} + + +#endif diff --git a/src/gallium/winsys/dri/intel/intel_lock.c b/src/gallium/winsys/dri/intel/intel_lock.c new file mode 100644 index 0000000000..70aa7ea5f4 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_lock.c @@ -0,0 +1,102 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "main/glheader.h" +#include "glapi/glthread.h" +#include +#include "state_tracker/st_public.h" +#include "intel_context.h" +#include "i830_dri.h" + + + +_glthread_DECLARE_STATIC_MUTEX( lockMutex ); + + +static void +intelContendedLock(struct intel_context *intel, uint flags) +{ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + __DRIscreenPrivate *sPriv = intel->driScreen; + struct intel_screen *intelScreen = intel_screen(sPriv); + drmI830Sarea *sarea = intel->sarea; + + drmGetLock(intel->driFd, intel->hHWContext, flags); + + DBG(LOCK, "%s - got contended lock\n", __progname); + + /* If the window moved, may need to set a new cliprect now. + * + * NOTE: This releases and regains the hw lock, so all state + * checking must be done *after* this call: + */ + if (dPriv) + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + + if (sarea->width != intelScreen->front.width || + sarea->height != intelScreen->front.height) { + + intelUpdateScreenRotation(sPriv, sarea); + } +} + + +/* Lock the hardware and validate our state. + */ +void LOCK_HARDWARE( struct intel_context *intel ) +{ + char __ret = 0; + + _glthread_LOCK_MUTEX(lockMutex); + assert(!intel->locked); + + DRM_CAS(intel->driHwLock, intel->hHWContext, + (DRM_LOCK_HELD|intel->hHWContext), __ret); + + if (__ret) + intelContendedLock( intel, 0 ); + + DBG(LOCK, "%s - locked\n", __progname); + + intel->locked = 1; +} + + +/* Unlock the hardware using the global current context + */ +void UNLOCK_HARDWARE( struct intel_context *intel ) +{ + assert(intel->locked); + intel->locked = 0; + + DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); + + _glthread_UNLOCK_MUTEX(lockMutex); + + DBG(LOCK, "%s - unlocked\n", __progname); +} diff --git a/src/gallium/winsys/dri/intel/intel_reg.h b/src/gallium/winsys/dri/intel/intel_reg.h new file mode 100644 index 0000000000..f37c24fda9 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_reg.h @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef _INTEL_REG_H_ +#define _INTEL_REG_H_ + + +#define BR00_BITBLT_CLIENT 0x40000000 +#define BR00_OP_COLOR_BLT 0x10000000 +#define BR00_OP_SRC_COPY_BLT 0x10C00000 +#define BR13_SOLID_PATTERN 0x80000000 + +#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) +#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) +#define XY_COLOR_BLT_WRITE_RGB (1<<20) + +#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) +#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) +#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) + +#define MI_WAIT_FOR_EVENT ((0x3<<23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +#define MI_BATCH_BUFFER_END (0xA<<23) + + +#endif diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c new file mode 100644 index 0000000000..9e31c013a9 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -0,0 +1,537 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "utils.h" +#include "vblank.h" +#include "xmlpool.h" + +#include "intel_context.h" +#include "intel_screen.h" +#include "intel_batchbuffer.h" +#include "intel_batchpool.h" +#include "intel_swapbuffers.h" +#include "intel_winsys.h" + +#include "i830_dri.h" +#include "dri_bufpool.h" + +#include "pipe/p_context.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_cb_fbo.h" + + + +PUBLIC const char __driConfigOptions[] = + DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE + DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) + DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) + DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY + DRI_CONF_FORCE_S3TC_ENABLE(false) + DRI_CONF_ALLOW_LARGE_TEXTURES(1) + DRI_CONF_SECTION_END DRI_CONF_END; + +const uint __driNConfigOptions = 4; + +#ifdef USE_NEW_INTERFACE +static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; +#endif /*USE_NEW_INTERFACE */ + +extern const struct dri_extension card_extensions[]; + + + + +static void +intelPrintDRIInfo(struct intel_screen * intelScreen, + __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) +{ + fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", + intelScreen->front.size, intelScreen->front.offset, + intelScreen->front.pitch); + fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); +} + + +#if 0 +static void +intelPrintSAREA(const drmI830Sarea * sarea) +{ + fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, + sarea->height); + fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); + fprintf(stderr, + "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->front_offset, sarea->front_size, + (unsigned) sarea->front_handle); + fprintf(stderr, + "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->back_offset, sarea->back_size, + (unsigned) sarea->back_handle); + fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->depth_offset, sarea->depth_size, + (unsigned) sarea->depth_handle); + fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); + fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); + fprintf(stderr, + "SAREA: rotated offset: 0x%08x size: 0x%x\n", + sarea->rotated_offset, sarea->rotated_size); + fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); +} +#endif + + +/** + * Use the information in the sarea to update the screen parameters + * related to screen rotation. Needs to be called locked. + */ +void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) +{ + struct intel_screen *intelScreen = intel_screen(sPriv); + + if (intelScreen->front.map) { + drmUnmap(intelScreen->front.map, intelScreen->front.size); + intelScreen->front.map = NULL; + } + + if (intelScreen->front.buffer) + driDeleteBuffers(1, &intelScreen->front.buffer); + + intelScreen->front.width = sarea->width; + intelScreen->front.height = sarea->height; + intelScreen->front.offset = sarea->front_offset; + intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; + intelScreen->front.size = sarea->front_size; + intelScreen->front.handle = sarea->front_handle; + + assert( sarea->front_size >= + intelScreen->front.pitch * intelScreen->front.height ); + + if (!sarea->front_handle) + return; + + if (drmMap(sPriv->fd, + sarea->front_handle, + intelScreen->front.size, + (drmAddress *) & intelScreen->front.map) != 0) { + fprintf(stderr, "drmMap(frontbuffer) failed!\n"); + return; + } + + if (intelScreen->staticPool) { + driGenBuffers(intelScreen->staticPool, "static region", 1, + &intelScreen->front.buffer, 64, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); + + driBOSetStatic(intelScreen->front.buffer, + intelScreen->front.offset, + intelScreen->front.pitch * intelScreen->front.height, + intelScreen->front.map, 0); + } +} + + +boolean +intelCreatePools(__DRIscreenPrivate * sPriv) +{ + unsigned batchPoolSize = 1024*1024; + struct intel_screen *intelScreen = intel_screen(sPriv); + + if (intelScreen->havePools) + return GL_TRUE; + + intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); + if (!intelScreen->staticPool) + return GL_FALSE; + + batchPoolSize /= BATCH_SZ; + intelScreen->batchPool = driBatchPoolInit(sPriv->fd, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT | + DRM_BO_FLAG_MEM_LOCAL, + BATCH_SZ, + batchPoolSize, 5); + if (!intelScreen->batchPool) { + fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); + return GL_FALSE; + } + + intelScreen->havePools = GL_TRUE; + + intelUpdateScreenRotation(sPriv, intelScreen->sarea); + + return GL_TRUE; +} + + +static boolean +intelInitDriver(__DRIscreenPrivate * sPriv) +{ + struct intel_screen *intelScreen; + I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; + + PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = + (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> + getProcAddress("glxEnableExtension")); + void *const psc = sPriv->psc->screenConfigs; + + if (sPriv->devPrivSize != sizeof(I830DRIRec)) { + fprintf(stderr, + "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); + return GL_FALSE; + } + + /* Allocate the private area */ + intelScreen = CALLOC_STRUCT(intel_screen); + if (!intelScreen) + return GL_FALSE; + + /* parse information in __driConfigOptions */ + driParseOptionInfo(&intelScreen->optionCache, + __driConfigOptions, __driNConfigOptions); + + sPriv->private = (void *) intelScreen; + + intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + + gDRIPriv->sarea_priv_offset); + intelScreen->deviceID = gDRIPriv->deviceID; + intelScreen->front.cpp = gDRIPriv->cpp; + intelScreen->drmMinor = sPriv->drmMinor; + + assert(gDRIPriv->bitsPerPixel == 16 || + gDRIPriv->bitsPerPixel == 32); + + intelUpdateScreenRotation(sPriv, intelScreen->sarea); + + if (0) + intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); + + if (glx_enable_extension != NULL) { + (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); + (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); + (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); + } + + intelScreen->winsys = intel_create_pipe_winsys(sPriv->fd); + + return GL_TRUE; +} + + +static void +intelDestroyScreen(__DRIscreenPrivate * sPriv) +{ + struct intel_screen *intelScreen = intel_screen(sPriv); + + /* intelUnmapScreenRegions(intelScreen); */ + + if (intelScreen->havePools) { + driPoolTakeDown(intelScreen->staticPool); + driPoolTakeDown(intelScreen->batchPool); + } + FREE(intelScreen); + sPriv->private = NULL; +} + + +/** + * This is called when we need to set up GL rendering to a new X window. + */ +static boolean +intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, + __DRIdrawablePrivate * driDrawPriv, + const __GLcontextModes * visual, boolean isPixmap) +{ + if (isPixmap) { + return GL_FALSE; /* not implemented */ + } + else { + enum pipe_format colorFormat, depthFormat, stencilFormat; + struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); + + if (!intelfb) + return GL_FALSE; + + if (visual->redBits == 5) + colorFormat = PIPE_FORMAT_R5G6B5_UNORM; + else + colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; + + if (visual->depthBits == 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (visual->depthBits == 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_NONE; + + if (visual->stencilBits == 8) + stencilFormat = PIPE_FORMAT_S8Z24_UNORM; + else + stencilFormat = PIPE_FORMAT_NONE; + + intelfb->stfb = st_create_framebuffer(visual, + colorFormat, + depthFormat, + stencilFormat, + driDrawPriv->w, + driDrawPriv->h, + (void*) intelfb); + if (!intelfb->stfb) { + free(intelfb); + return GL_FALSE; + } + + driDrawPriv->driverPrivate = (void *) intelfb; + return GL_TRUE; + } +} + +static void +intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) +{ + struct intel_framebuffer *intelfb = intel_framebuffer(driDrawPriv); + assert(intelfb->stfb); + st_unreference_framebuffer(&intelfb->stfb); + free(intelfb); +} + + +/** + * Get information about previous buffer swaps. + */ +static int +intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) +{ + if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) + || (sInfo == NULL)) { + return -1; + } + + return 0; +} + + +static void +intelSetTexOffset(__DRIcontext *pDRICtx, int texname, + unsigned long long offset, int depth, uint pitch) +{ + abort(); +#if 0 + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct st_texture_object *stObj = st_texture_object(tObj); + + if (!stObj) + return; + + if (stObj->pt) + st->pipe->texture_release(intel->st->pipe, &stObj->pt); + + stObj->imageOverride = GL_TRUE; + stObj->depthOverride = depth; + stObj->pitchOverride = pitch; + + if (offset) + stObj->textureOffset = offset; +#endif +} + + +static const struct __DriverAPIRec intelAPI = { + .InitDriver = intelInitDriver, + .DestroyScreen = intelDestroyScreen, + .CreateContext = intelCreateContext, + .DestroyContext = intelDestroyContext, + .CreateBuffer = intelCreateBuffer, + .DestroyBuffer = intelDestroyBuffer, + .SwapBuffers = intelSwapBuffers, + .MakeCurrent = intelMakeCurrent, + .UnbindContext = intelUnbindContext, + .GetSwapInfo = intelGetSwapInfo, + .GetMSC = driGetMSC32, + .WaitForMSC = driWaitForMSC32, + .WaitForSBC = NULL, + .SwapBuffersMSC = NULL, + .CopySubBuffer = intelCopySubBuffer, + .setTexOffset = intelSetTexOffset, +}; + + +static __GLcontextModes * +intelFillInModes(unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, boolean have_back_buffer) +{ + __GLcontextModes *modes; + __GLcontextModes *m; + unsigned num_modes; + unsigned depth_buffer_factor; + unsigned back_buffer_factor; + GLenum fb_format; + GLenum fb_type; + + /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't + * support pageflipping at all. + */ + static const GLenum back_buffer_modes[] = { + GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML + }; + + u_int8_t depth_bits_array[3]; + u_int8_t stencil_bits_array[3]; + + + depth_bits_array[0] = 0; + depth_bits_array[1] = depth_bits; + depth_bits_array[2] = depth_bits; + + /* Just like with the accumulation buffer, always provide some modes + * with a stencil buffer. It will be a sw fallback, but some apps won't + * care about that. + */ + stencil_bits_array[0] = 0; + stencil_bits_array[1] = 0; + if (depth_bits == 24) + stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; + + stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; + + depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; + back_buffer_factor = (have_back_buffer) ? 3 : 1; + + num_modes = depth_buffer_factor * back_buffer_factor * 4; + + if (pixel_bits == 16) { + fb_format = GL_RGB; + fb_type = GL_UNSIGNED_SHORT_5_6_5; + } + else { + fb_format = GL_BGRA; + fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; + } + + modes = + (*dri_interface->createContextModes) (num_modes, + sizeof(__GLcontextModes)); + m = modes; + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_TRUE_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_DIRECT_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + + /* Mark the visual as slow if there are "fake" stencil bits. + */ + for (m = modes; m != NULL; m = m->next) { + if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { + m->visualRating = GLX_SLOW_CONFIG; + } + } + + return modes; +} + + +/** + * This is the bootstrap function for the driver. libGL supplies all of the + * requisite information about the system, and the driver initializes itself. + * This routine also fills in the linked list pointed to by \c driver_modes + * with the \c __GLcontextModes that the driver can support for windows or + * pbuffers. + * + * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on + * failure. + */ +PUBLIC void * +__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, + __DRIscreen * psc, + const __GLcontextModes * modes, + const __DRIversion * ddx_version, + const __DRIversion * dri_version, + const __DRIversion * drm_version, + const __DRIframebuffer * frame_buffer, + drmAddress pSAREA, int fd, + int internal_api_version, + const __DRIinterfaceMethods * interface, + __GLcontextModes ** driver_modes) +{ + __DRIscreenPrivate *psp; + static const __DRIversion ddx_expected = { 1, 7, 0 }; + static const __DRIversion dri_expected = { 4, 0, 0 }; + static const __DRIversion drm_expected = { 1, 7, 0 }; + + dri_interface = interface; + + if (!driCheckDriDdxDrmVersions2("i915", + dri_version, &dri_expected, + ddx_version, &ddx_expected, + drm_version, &drm_expected)) { + return NULL; + } + + psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, + ddx_version, dri_version, drm_version, + frame_buffer, pSAREA, fd, + internal_api_version, &intelAPI); + + if (psp != NULL) { + I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; + *driver_modes = intelFillInModes(dri_priv->cpp * 8, + (dri_priv->cpp == 2) ? 16 : 24, + (dri_priv->cpp == 2) ? 0 : 8, 1); + + /* Calling driInitExtensions here, with a NULL context pointer, + * does not actually enable the extensions. It just makes sure + * that all the dispatch offsets for all the extensions that + * *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create + * is called, but we can't enable the extensions until we have a + * context pointer. + * + * Hello chicken. Hello egg. How are you two today? + */ + driInitExtensions(NULL, card_extensions, GL_FALSE); + } + + return (void *) psp; +} + diff --git a/src/gallium/winsys/dri/intel/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h new file mode 100644 index 0000000000..3396f9e564 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_screen.h @@ -0,0 +1,113 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _INTEL_SCREEN_H_ +#define _INTEL_SCREEN_H_ + +#include "dri_util.h" +#include "i830_common.h" +#include "xmlconfig.h" +#include "dri_bufpool.h" + +#include "pipe/p_compiler.h" + + +struct intel_screen +{ + struct { + drm_handle_t handle; + + /* We create a static dri buffer for the frontbuffer. + */ + struct _DriBufferObject *buffer; + + char *map; /* memory map */ + int offset; /* from start of video mem, in bytes */ + int pitch; /* row stride, in bytes */ + int width; + int height; + int size; + int cpp; /* for front and back buffers */ + } front; + + int deviceID; + int drmMinor; + + drmI830Sarea *sarea; + + /** + * Configuration cache with default values for all contexts + */ + driOptionCache optionCache; + + struct _DriBufferPool *batchPool; + struct _DriBufferPool *staticPool; /** for the X screen/framebuffer */ + boolean havePools; + + /** + * Temporary(?) context to use for SwapBuffers or other situations in + * which we need a rendering context, but none is currently bound. + */ + struct intel_context *dummyContext; + + struct pipe_winsys *winsys; +}; + + + +/** cast wrapper */ +static INLINE struct intel_screen * +intel_screen(__DRIscreenPrivate *sPriv) +{ + return (struct intel_screen *) sPriv->private; +} + + +extern void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); + + +extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); + +extern boolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); + +extern boolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv); + + +extern boolean +intelCreatePools(__DRIscreenPrivate *sPriv); + +extern boolean +intelCreateContext(const __GLcontextModes * visual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); + + +#endif diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c new file mode 100644 index 0000000000..56b86d6a63 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -0,0 +1,253 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_swapbuffers.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" +#include "intel_winsys.h" + +#include "pipe/p_context.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_cb_fbo.h" + + +/** + * Display a colorbuffer surface in an X window. + * Used for SwapBuffers and flushing front buffer rendering. + * + * \param dPriv the window/drawable to display into + * \param surf the surface to display + * \param rect optional subrect of surface to display (may be NULL). + */ +void +intelDisplaySurface(__DRIdrawablePrivate *dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t *rect) +{ + struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); + struct intel_context *intel = intelScreen->dummyContext; + + DBG(SWAP, "%s\n", __FUNCTION__); + + if (!intel) { + /* XXX this is where some kind of extra/meta context could be useful */ + return; + } + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); + driFenceUnReference(intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + * current context (which is what intelScreenContext should return) might + * not get a contended lock and thus cliprects not updated (tests/manywin) + */ + if (intel_context(dPriv->driContextPriv) != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + + + if (dPriv && dPriv->numClipRects) { + const int srcWidth = surf->width; + const int srcHeight = surf->height; + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; + const int cpp = intelScreen->front.cpp; + const int srcpitch = surf->pitch; + int BR13, CMD; + int i; + + ASSERT(surf->buffer); + ASSERT(surf->cpp == cpp); + + DBG(SWAP, "screen pitch %d src surface pitch %d\n", + pitch, surf->pitch); + + if (cpp == 2) { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + } + else { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + } + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + drm_clip_rect_t sbox; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->front.width || + pbox->y2 > intelScreen->front.height) { + /* invalid cliprect, skip it */ + continue; + } + + box = *pbox; + + if (rect) { + /* intersect cliprect with user-provided src rect */ + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > srcWidth) + box.x2 = srcWidth + box.x1; + if (box.y2 - box.y1 > srcHeight) + box.y2 = srcHeight + box.y1; + + DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + + assert(box.x1 < box.x2); + assert(box.y1 < box.y2); + + /* XXX this could be done with pipe->surface_copy() */ + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((box.y1 << 16) | box.x1); + OUT_BATCH((box.y2 << 16) | box.x2); + + OUT_RELOC(intelScreen->front.buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + OUT_BATCH((sbox.y1 << 16) | sbox.x1); + OUT_BATCH((srcpitch * cpp) & 0xffff); + OUT_RELOC(dri_bo(surf->buffer), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + ADVANCE_BATCH(); + } + + if (intel->first_swap_fence) + driFenceUnReference(intel->first_swap_fence); + intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); + driFenceReference(intel->first_swap_fence); + } + + UNLOCK_HARDWARE(intel); + + if (intel->lastStamp != dPriv->lastStamp) { + intelUpdateWindowSize(dPriv); + intel->lastStamp = dPriv->lastStamp; + } +} + + + +/** + * This will be called whenever the currently bound window is moved/resized. + */ +void +intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) +{ + struct intel_framebuffer *intelfb = intel_framebuffer(dPriv); + assert(intelfb->stfb); + st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); +} + + + +void +intelSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); + struct pipe_surface *back_surf; + + assert(intel_fb); + assert(intel_fb->stfb); + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { + st_notify_swapbuffers(intel_fb->stfb); + intelDisplaySurface(dPriv, back_surf, NULL); + st_notify_swapbuffers_complete(intel_fb->stfb); + } +} + + +/** + * Called via glXCopySubBufferMESA() to copy a subrect of the back + * buffer to the front buffer/screen. + */ +void +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) +{ + struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); + struct pipe_surface *back_surf; + + assert(intel_fb); + assert(intel_fb->stfb); + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { + drm_clip_rect_t rect; + rect.x1 = x; + rect.y1 = y; + rect.x2 = w; + rect.y2 = h; + + st_notify_swapbuffers(intel_fb->stfb); + intelDisplaySurface(dPriv, back_surf, &rect); + } +} diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.h b/src/gallium/winsys/dri/intel/intel_swapbuffers.h new file mode 100644 index 0000000000..7ae5fd15a5 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_SWAPBUFFERS_H +#define INTEL_SWAPBUFFERS_H + + +struct pipe_surface; + + +extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t * rect); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); + +extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, + int x, int y, int w, int h); + +extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); + + +#endif /* INTEL_SWAPBUFFERS_H */ diff --git a/src/gallium/winsys/dri/intel/intel_winsys.h b/src/gallium/winsys/dri/intel/intel_winsys.h new file mode 100644 index 0000000000..ffc40782be --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_winsys.h @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_WINSYS_H +#define INTEL_WINSYS_H + +#include "pipe/p_state.h" + +struct intel_context; +struct pipe_context; +struct pipe_winsys; +struct pipe_buffer; +struct _DriBufferObject; + +struct pipe_winsys * +intel_create_pipe_winsys( int fd ); + +void +intel_destroy_pipe_winsys( struct pipe_winsys *winsys ); + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel, + struct pipe_winsys *winsys ); + +struct pipe_context * +intel_create_i915simple( struct intel_context *intel, + struct pipe_winsys *winsys ); + + +struct intel_buffer { + struct pipe_buffer base; + struct _DriBufferObject *driBO; +}; + +static INLINE struct intel_buffer * +intel_buffer( struct pipe_buffer *buf ) +{ + return (struct intel_buffer *)buf; +} + +static INLINE struct _DriBufferObject * +dri_bo( struct pipe_buffer *buf ) +{ + return intel_buffer(buf)->driBO; +} + + + +#endif diff --git a/src/gallium/winsys/dri/intel/intel_winsys_i915.c b/src/gallium/winsys/dri/intel/intel_winsys_i915.c new file mode 100644 index 0000000000..1ba6a9e1b2 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_winsys_i915.c @@ -0,0 +1,154 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +#include "dri_bufpool.h" +#include "dri_bufmgr.h" + +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_winsys.h" + +#include "pipe/p_util.h" +#include "pipe/i915simple/i915_winsys.h" + + +struct intel_i915_winsys { + struct i915_winsys winsys; /**< batch buffer funcs */ + struct intel_context *intel; +}; + + +/* Turn a i915simple winsys into an intel/i915simple winsys: + */ +static inline struct intel_i915_winsys * +intel_i915_winsys( struct i915_winsys *sws ) +{ + return (struct intel_i915_winsys *)sws; +} + + +/* Simple batchbuffer interface: + */ + +static unsigned *intel_i915_batch_start( struct i915_winsys *sws, + unsigned dwords, + unsigned relocs ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + /* XXX: check relocs. + */ + if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { + /* XXX: Hmm, the driver can't really do much with this pointer: + */ + return (unsigned *)intel->batch->ptr; + } + else + return NULL; +} + +static void intel_i915_batch_dword( struct i915_winsys *sws, + unsigned dword ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + intel_batchbuffer_emit_dword( intel->batch, dword ); +} + +static void intel_i915_batch_reloc( struct i915_winsys *sws, + struct pipe_buffer *buf, + unsigned access_flags, + unsigned delta ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + unsigned flags = DRM_BO_FLAG_MEM_TT; + unsigned mask = DRM_BO_MASK_MEM; + + if (access_flags & I915_BUFFER_ACCESS_WRITE) { + flags |= DRM_BO_FLAG_WRITE; + mask |= DRM_BO_FLAG_WRITE; + } + + if (access_flags & I915_BUFFER_ACCESS_READ) { + flags |= DRM_BO_FLAG_READ; + mask |= DRM_BO_FLAG_READ; + } + + intel_batchbuffer_emit_reloc( intel->batch, + dri_bo( buf ), + flags, mask, + delta ); +} + + + +static void intel_i915_batch_flush( struct i915_winsys *sws ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + intel_batchbuffer_flush( intel->batch ); +// if (0) intel_i915_batch_wait_idle( sws ); +} + + +static void intel_i915_batch_finish( struct i915_winsys *sws ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + intel_batchbuffer_finish( intel->batch ); +} + + +/** + * Create i915 hardware rendering context. + */ +struct pipe_context * +intel_create_i915simple( struct intel_context *intel, + struct pipe_winsys *winsys ) +{ + struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); + + /* Fill in this struct with callbacks that i915simple will need to + * communicate with the window system, buffer manager, etc. + */ + iws->winsys.batch_start = intel_i915_batch_start; + iws->winsys.batch_dword = intel_i915_batch_dword; + iws->winsys.batch_reloc = intel_i915_batch_reloc; + iws->winsys.batch_flush = intel_i915_batch_flush; + iws->winsys.batch_finish = intel_i915_batch_finish; + iws->intel = intel; + + /* Create the i915simple context: + */ + return i915_create( winsys, + &iws->winsys, + intel->intelScreen->deviceID ); +} diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c new file mode 100644 index 0000000000..789a386500 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -0,0 +1,302 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +#include "dri_bufpool.h" +#include "dri_bufmgr.h" + +#include "intel_context.h" +#include "intel_winsys.h" +#include "intel_swapbuffers.h" +#include "intel_batchbuffer.h" + +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" + + + +struct intel_pipe_winsys { + struct pipe_winsys winsys; + struct _DriBufferPool *regionPool; +}; + + + +/* Turn a pipe winsys into an intel/pipe winsys: + */ +static inline struct intel_pipe_winsys * +intel_pipe_winsys( struct pipe_winsys *winsys ) +{ + return (struct intel_pipe_winsys *)winsys; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void *intel_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags ) +{ + unsigned drm_flags = 0; + + if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) + drm_flags |= DRM_BO_FLAG_WRITE; + + if (flags & PIPE_BUFFER_USAGE_CPU_READ) + drm_flags |= DRM_BO_FLAG_READ; + + return driBOMap( dri_bo(buf), drm_flags, 0 ); +} + +static void intel_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + driBOUnmap( dri_bo(buf) ); +} + + +static void +intel_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + driBOUnReference( dri_bo(buf) ); +} + + +/* Pipe has no concept of pools. We choose the tex/region pool + * for all buffers. + * Grabs the hardware lock! + */ +static struct pipe_buffer * +intel_buffer_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned usage, + unsigned size ) +{ + struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); + struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + unsigned flags = 0; + + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + + if (usage & (PIPE_BUFFER_USAGE_VERTEX /*| IWS_BUFFER_USAGE_LOCAL*/)) { + flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; + } else { + flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; + } + + if (usage & PIPE_BUFFER_USAGE_GPU_READ) + flags |= DRM_BO_FLAG_READ; + + if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) + flags |= DRM_BO_FLAG_WRITE; + + /* drm complains if we don't set any read/write flags. + */ + if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) + flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + +#if 0 + if (flags & IWS_BUFFER_USAGE_EXE) + flags |= DRM_BO_FLAG_EXE; + + if (usage & IWS_BUFFER_USAGE_CACHED) + flags |= DRM_BO_FLAG_CACHED; +#endif + + driGenBuffers( iws->regionPool, + "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); + + driBOData( buffer->driBO, size, NULL, 0 ); + + return &buffer->base; +} + + +static struct pipe_buffer * +intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) +{ + struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); + struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + + driGenUserBuffer( iws->regionPool, + "pipe user buffer", &buffer->driBO, ptr, bytes); + + return &buffer->base; +} + + +/* The state tracker (should!) keep track of whether the fake + * frontbuffer has been touched by any rendering since the last time + * we copied its contents to the real frontbuffer. Our task is easy: + */ +static void +intel_flush_frontbuffer( struct pipe_winsys *winsys, + struct pipe_surface *surf, + void *context_private) +{ + struct intel_context *intel = (struct intel_context *) context_private; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelDisplaySurface(dPriv, surf, NULL); +} + + +static struct pipe_surface * +intel_i915_surface_alloc(struct pipe_winsys *winsys) +{ + struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); + if (surf) { + surf->refcount = 1; + surf->winsys = winsys; + } + return surf; +} + + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + +/** + * Copied from xm_winsys.c + */ +static int +intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags) +{ + const unsigned alignment = 64; + int ret; + + surf->width = width; + surf->height = height; + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = round_up(width, alignment / surf->cpp); + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->cpp * height); + if(!surf->buffer) + return -1; + + return 0; +} + + +static void +intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + surf->refcount--; + if (surf->refcount == 0) { + if (surf->buffer) + pipe_buffer_reference(winsys, &surf->buffer, NULL); + free(surf); + } + *s = NULL; +} + + + +static void +intel_printf( struct pipe_winsys *winsys, const char *fmtString, ... ) +{ + va_list args; + va_start( args, fmtString ); + vfprintf(stderr, fmtString, args); + va_end( args ); +} + +static const char * +intel_get_name( struct pipe_winsys *winsys ) +{ + return "Intel/DRI/ttm"; +} + + +struct pipe_winsys * +intel_create_pipe_winsys( int fd ) +{ + struct intel_pipe_winsys *iws = CALLOC_STRUCT( intel_pipe_winsys ); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + * + * Pipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + iws->winsys.buffer_create = intel_buffer_create; + iws->winsys.user_buffer_create = intel_user_buffer_create; + iws->winsys.buffer_map = intel_buffer_map; + iws->winsys.buffer_unmap = intel_buffer_unmap; + iws->winsys.buffer_destroy = intel_buffer_destroy; + iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; + iws->winsys.printf = intel_printf; + iws->winsys.get_name = intel_get_name; + iws->winsys.surface_alloc = intel_i915_surface_alloc; + iws->winsys.surface_alloc_storage = intel_i915_surface_alloc_storage; + iws->winsys.surface_release = intel_i915_surface_release; + + if (fd) + iws->regionPool = driDRMPoolInit(fd); + + return &iws->winsys; +} + + +void +intel_destroy_pipe_winsys( struct pipe_winsys *winsys ) +{ + struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + if (iws->regionPool) { + driPoolTakeDown(iws->regionPool); + } + free(iws); +} + diff --git a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c new file mode 100644 index 0000000000..cec3437831 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c @@ -0,0 +1,81 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include "intel_context.h" +#include "intel_winsys.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_format.h" +#include "pipe/softpipe/sp_winsys.h" + + +struct intel_softpipe_winsys { + struct softpipe_winsys sws; + struct intel_context *intel; +}; + +/** + * Return list of surface formats supported by this driver. + */ +static boolean +intel_is_format_supported(struct softpipe_winsys *sws, + enum pipe_format format) +{ + switch(format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + return TRUE; + default: + return FALSE; + } +} + + +/** + * Create rendering context which uses software rendering. + */ +struct pipe_context * +intel_create_softpipe( struct intel_context *intel, + struct pipe_winsys *winsys ) +{ + struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); + + /* Fill in this struct with callbacks that softpipe will need to + * communicate with the window system, buffer manager, etc. + */ + isws->sws.is_format_supported = intel_is_format_supported; + isws->intel = intel; + + /* Create the softpipe context: + */ + return softpipe_create( winsys, &isws->sws ); +} diff --git a/src/gallium/winsys/dri/intel/server/i830_common.h b/src/gallium/winsys/dri/intel/server/i830_common.h new file mode 100644 index 0000000000..d4d58886ce --- /dev/null +++ b/src/gallium/winsys/dri/intel/server/i830_common.h @@ -0,0 +1,226 @@ +/************************************************************************** + +Copyright 2001 VA Linux Systems Inc., Fremont, California. +Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas. + +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the "Software"), +to deal in the Software without restriction, including without limitation +on the rights to use, copy, modify, merge, publish, distribute, sub +license, and/or sell copies of the Software, and to permit persons to whom +the Software is furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice (including the next +paragraph) shall be included in all copies or substantial portions of the +Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL +ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, +DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR +OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE +USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ + +#ifndef _I830_COMMON_H_ +#define _I830_COMMON_H_ + + +#define I830_NR_TEX_REGIONS 255 /* maximum due to use of chars for next/prev */ +#define I830_LOG_MIN_TEX_REGION_SIZE 14 + + +/* Driver specific DRM command indices + * NOTE: these are not OS specific, but they are driver specific + */ +#define DRM_I830_INIT 0x00 +#define DRM_I830_FLUSH 0x01 +#define DRM_I830_FLIP 0x02 +#define DRM_I830_BATCHBUFFER 0x03 +#define DRM_I830_IRQ_EMIT 0x04 +#define DRM_I830_IRQ_WAIT 0x05 +#define DRM_I830_GETPARAM 0x06 +#define DRM_I830_SETPARAM 0x07 +#define DRM_I830_ALLOC 0x08 +#define DRM_I830_FREE 0x09 +#define DRM_I830_INIT_HEAP 0x0a +#define DRM_I830_CMDBUFFER 0x0b +#define DRM_I830_DESTROY_HEAP 0x0c +#define DRM_I830_SET_VBLANK_PIPE 0x0d +#define DRM_I830_GET_VBLANK_PIPE 0x0e + +typedef struct { + enum { + I830_INIT_DMA = 0x01, + I830_CLEANUP_DMA = 0x02, + I830_RESUME_DMA = 0x03 + } func; + unsigned int mmio_offset; + int sarea_priv_offset; + unsigned int ring_start; + unsigned int ring_end; + unsigned int ring_size; + unsigned int front_offset; + unsigned int back_offset; + unsigned int depth_offset; + unsigned int w; + unsigned int h; + unsigned int pitch; + unsigned int pitch_bits; + unsigned int back_pitch; + unsigned int depth_pitch; + unsigned int cpp; + unsigned int chipset; +} drmI830Init; + +typedef struct { + drmTextureRegion texList[I830_NR_TEX_REGIONS+1]; + int last_upload; /* last time texture was uploaded */ + int last_enqueue; /* last time a buffer was enqueued */ + int last_dispatch; /* age of the most recently dispatched buffer */ + int ctxOwner; /* last context to upload state */ + int texAge; + int pf_enabled; /* is pageflipping allowed? */ + int pf_active; + int pf_current_page; /* which buffer is being displayed? */ + int perf_boxes; /* performance boxes to be displayed */ + int width, height; /* screen size in pixels */ + + drm_handle_t front_handle; + int front_offset; + int front_size; + + drm_handle_t back_handle; + int back_offset; + int back_size; + + drm_handle_t depth_handle; + int depth_offset; + int depth_size; + + drm_handle_t tex_handle; + int tex_offset; + int tex_size; + int log_tex_granularity; + int pitch; + int rotation; /* 0, 90, 180 or 270 */ + int rotated_offset; + int rotated_size; + int rotated_pitch; + int virtualX, virtualY; + + unsigned int front_tiled; + unsigned int back_tiled; + unsigned int depth_tiled; + unsigned int rotated_tiled; + unsigned int rotated2_tiled; + + int pipeA_x; + int pipeA_y; + int pipeA_w; + int pipeA_h; + int pipeB_x; + int pipeB_y; + int pipeB_w; + int pipeB_h; + + /* Triple buffering */ + drm_handle_t third_handle; + int third_offset; + int third_size; + unsigned int third_tiled; +} drmI830Sarea; + +/* Flags for perf_boxes + */ +#define I830_BOX_RING_EMPTY 0x1 /* populated by kernel */ +#define I830_BOX_FLIP 0x2 /* populated by kernel */ +#define I830_BOX_WAIT 0x4 /* populated by kernel & client */ +#define I830_BOX_TEXTURE_LOAD 0x8 /* populated by kernel */ +#define I830_BOX_LOST_CONTEXT 0x10 /* populated by client */ + + +typedef struct { + int start; /* agp offset */ + int used; /* nr bytes in use */ + int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ + int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ + int num_cliprects; /* mulitpass with multiple cliprects? */ + drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ +} drmI830BatchBuffer; + +typedef struct { + char *buf; /* agp offset */ + int sz; /* nr bytes in use */ + int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ + int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ + int num_cliprects; /* mulitpass with multiple cliprects? */ + drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ +} drmI830CmdBuffer; + +typedef struct { + int *irq_seq; +} drmI830IrqEmit; + +typedef struct { + int irq_seq; +} drmI830IrqWait; + +typedef struct { + int param; + int *value; +} drmI830GetParam; + +#define I830_PARAM_IRQ_ACTIVE 1 +#define I830_PARAM_ALLOW_BATCHBUFFER 2 + +typedef struct { + int param; + int value; +} drmI830SetParam; + +#define I830_SETPARAM_USE_MI_BATCHBUFFER_START 1 +#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY 2 +#define I830_SETPARAM_ALLOW_BATCHBUFFER 3 + + +/* A memory manager for regions of shared memory: + */ +#define I830_MEM_REGION_AGP 1 + +typedef struct { + int region; + int alignment; + int size; + int *region_offset; /* offset from start of fb or agp */ +} drmI830MemAlloc; + +typedef struct { + int region; + int region_offset; +} drmI830MemFree; + +typedef struct { + int region; + int size; + int start; +} drmI830MemInitHeap; + +typedef struct { + int region; +} drmI830MemDestroyHeap; + +#define DRM_I830_VBLANK_PIPE_A 1 +#define DRM_I830_VBLANK_PIPE_B 2 + +typedef struct { + int pipe; +} drmI830VBlankPipe; + +#endif /* _I830_DRM_H_ */ diff --git a/src/gallium/winsys/dri/intel/server/i830_dri.h b/src/gallium/winsys/dri/intel/server/i830_dri.h new file mode 100644 index 0000000000..c2a3af8cbf --- /dev/null +++ b/src/gallium/winsys/dri/intel/server/i830_dri.h @@ -0,0 +1,63 @@ +/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ + +#ifndef _I830_DRI_H +#define _I830_DRI_H + +#include "xf86drm.h" +#include "i830_common.h" + +#define I830_MAX_DRAWABLES 256 + +#define I830_MAJOR_VERSION 1 +#define I830_MINOR_VERSION 7 +#define I830_PATCHLEVEL 2 + +#define I830_REG_SIZE 0x80000 + +typedef struct _I830DRIRec { + drm_handle_t regs; + drmSize regsSize; + + drmSize unused1; /* backbufferSize */ + drm_handle_t unused2; /* backbuffer */ + + drmSize unused3; /* depthbufferSize */ + drm_handle_t unused4; /* depthbuffer */ + + drmSize unused5; /* rotatedSize */ + drm_handle_t unused6; /* rotatedbuffer */ + + drm_handle_t unused7; /* textures */ + int unused8; /* textureSize */ + + drm_handle_t unused9; /* agp_buffers */ + drmSize unused10; /* agp_buf_size */ + + int deviceID; + int width; + int height; + int mem; + int cpp; + int bitsPerPixel; + + int unused11[8]; /* was front/back/depth/rotated offset/pitch */ + + int unused12; /* logTextureGranularity */ + int unused13; /* textureOffset */ + + int irq; + int sarea_priv_offset; +} I830DRIRec, *I830DRIPtr; + +typedef struct { + /* Nothing here yet */ + int dummy; +} I830ConfigPrivRec, *I830ConfigPrivPtr; + +typedef struct { + /* Nothing here yet */ + int dummy; +} I830DRIContextRec, *I830DRIContextPtr; + + +#endif diff --git a/src/gallium/winsys/dri/intel/server/intel.h b/src/gallium/winsys/dri/intel/server/intel.h new file mode 100644 index 0000000000..6ea72499c1 --- /dev/null +++ b/src/gallium/winsys/dri/intel/server/intel.h @@ -0,0 +1,331 @@ +#ifndef _INTEL_H_ +#define _INTEL_H_ + +#include "xf86drm.h" /* drm_handle_t, etc */ + +/* Intel */ +#ifndef PCI_CHIP_I810 +#define PCI_CHIP_I810 0x7121 +#define PCI_CHIP_I810_DC100 0x7123 +#define PCI_CHIP_I810_E 0x7125 +#define PCI_CHIP_I815 0x1132 +#define PCI_CHIP_I810_BRIDGE 0x7120 +#define PCI_CHIP_I810_DC100_BRIDGE 0x7122 +#define PCI_CHIP_I810_E_BRIDGE 0x7124 +#define PCI_CHIP_I815_BRIDGE 0x1130 +#endif + +#define PCI_CHIP_845_G 0x2562 +#define PCI_CHIP_I830_M 0x3577 + +#ifndef PCI_CHIP_I855_GM +#define PCI_CHIP_I855_GM 0x3582 +#define PCI_CHIP_I855_GM_BRIDGE 0x3580 +#endif + +#ifndef PCI_CHIP_I865_G +#define PCI_CHIP_I865_G 0x2572 +#define PCI_CHIP_I865_G_BRIDGE 0x2570 +#endif + +#ifndef PCI_CHIP_I915_G +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_G_BRIDGE 0x2580 +#endif + +#ifndef PCI_CHIP_I915_GM +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I915_GM_BRIDGE 0x2590 +#endif + +#ifndef PCI_CHIP_E7221_G +#define PCI_CHIP_E7221_G 0x258A +/* Same as I915_G_BRIDGE */ +#define PCI_CHIP_E7221_G_BRIDGE 0x2580 +#endif + +#ifndef PCI_CHIP_I945_G +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_G_BRIDGE 0x2770 +#endif + +#ifndef PCI_CHIP_I945_GM +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GM_BRIDGE 0x27A0 +#endif + +#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 || \ + pI810->Chipset == PCI_CHIP_I810_DC100 || \ + pI810->Chipset == PCI_CHIP_I810_E) +#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815) +#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M) +#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G) +#define IS_I85X(pI810) (pI810->Chipset == PCI_CHIP_I855_GM) +#define IS_I852(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME)) +#define IS_I855(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME)) +#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G) + +#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G) +#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM) +#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G) +#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM) +#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810)) + +#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810)) + +#define I830_GMCH_CTRL 0x52 + +#define I830_GMCH_MEM_MASK 0x1 +#define I830_GMCH_MEM_64M 0x1 +#define I830_GMCH_MEM_128M 0 + +#define I830_GMCH_GMS_MASK 0x70 +#define I830_GMCH_GMS_DISABLED 0x00 +#define I830_GMCH_GMS_LOCAL 0x10 +#define I830_GMCH_GMS_STOLEN_512 0x20 +#define I830_GMCH_GMS_STOLEN_1024 0x30 +#define I830_GMCH_GMS_STOLEN_8192 0x40 + +#define I855_GMCH_GMS_MASK (0x7 << 4) +#define I855_GMCH_GMS_DISABLED 0x00 +#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) +#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) +#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) +#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) +#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) +#define I915G_GMCH_GMS_STOLEN_48M (0x6 << 4) +#define I915G_GMCH_GMS_STOLEN_64M (0x7 << 4) + +typedef unsigned char Bool; +#define TRUE 1 +#define FALSE 0 + +#define PIPE_NONE 0<<0 +#define PIPE_CRT 1<<0 +#define PIPE_TV 1<<1 +#define PIPE_DFP 1<<2 +#define PIPE_LFP 1<<3 +#define PIPE_CRT2 1<<4 +#define PIPE_TV2 1<<5 +#define PIPE_DFP2 1<<6 +#define PIPE_LFP2 1<<7 + +typedef struct _I830MemPool *I830MemPoolPtr; +typedef struct _I830MemRange *I830MemRangePtr; +typedef struct _I830MemRange { + long Start; + long End; + long Size; + unsigned long Physical; + unsigned long Offset; /* Offset of AGP-allocated portion */ + unsigned long Alignment; + drm_handle_t Key; + unsigned long Pitch; // add pitch + I830MemPoolPtr Pool; +} I830MemRange; + +typedef struct _I830MemPool { + I830MemRange Total; + I830MemRange Free; + I830MemRange Fixed; + I830MemRange Allocated; +} I830MemPool; + +typedef struct { + int tail_mask; + I830MemRange mem; + unsigned char *virtual_start; + int head; + int tail; + int space; +} I830RingBuffer; + +typedef struct _I830Rec { + unsigned char *MMIOBase; + unsigned char *FbBase; + int cpp; + uint32_t aper_size; + unsigned int bios_version; + + /* These are set in PreInit and never changed. */ + long FbMapSize; + long TotalVideoRam; + I830MemRange StolenMemory; /* pre-allocated memory */ + long BIOSMemorySize; /* min stolen pool size */ + int BIOSMemSizeLoc; + + /* These change according to what has been allocated. */ + long FreeMemory; + I830MemRange MemoryAperture; + I830MemPool StolenPool; + long allocatedMemory; + + /* Regions allocated either from the above pools, or from agpgart. */ + /* for single and dual head configurations */ + I830MemRange FrontBuffer; + I830MemRange FrontBuffer2; + I830MemRange Scratch; + I830MemRange Scratch2; + + I830RingBuffer *LpRing; + + I830MemRange BackBuffer; + I830MemRange DepthBuffer; + I830MemRange TexMem; + int TexGranularity; + I830MemRange ContextMem; + int drmMinor; + Bool have3DWindows; + + Bool NeedRingBufferLow; + Bool allowPageFlip; + Bool disableTiling; + + int Chipset; + unsigned long LinearAddr; + unsigned long MMIOAddr; + + drmSize registerSize; /**< \brief MMIO register map size */ + drm_handle_t registerHandle; /**< \brief MMIO register map handle */ + // IOADDRESS ioBase; + int irq; /**< \brief IRQ number */ + int GttBound; + + drm_handle_t ring_map; + unsigned int Fence[8]; + +} I830Rec; + +/* + * 12288 is set as the maximum, chosen because it is enough for + * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare. + */ +#define I830_MAXIMUM_VBIOS_MEM 12288 +#define I830_DEFAULT_VIDEOMEM_2D (MB(32) / 1024) +#define I830_DEFAULT_VIDEOMEM_3D (MB(64) / 1024) + +/* Flags for memory allocation function */ +#define FROM_ANYWHERE 0x00000000 +#define FROM_POOL_ONLY 0x00000001 +#define FROM_NEW_ONLY 0x00000002 +#define FROM_MASK 0x0000000f + +#define ALLOCATE_AT_TOP 0x00000010 +#define ALLOCATE_AT_BOTTOM 0x00000020 +#define FORCE_GAPS 0x00000040 + +#define NEED_PHYSICAL_ADDR 0x00000100 +#define ALIGN_BOTH_ENDS 0x00000200 +#define FORCE_LOW 0x00000400 + +#define ALLOC_NO_TILING 0x00001000 +#define ALLOC_INITIAL 0x00002000 + +#define ALLOCATE_DRY_RUN 0x80000000 + +/* Chipset registers for VIDEO BIOS memory RW access */ +#define _855_DRAM_RW_CONTROL 0x58 +#define _845_DRAM_RW_CONTROL 0x90 +#define DRAM_WRITE 0x33330000 + +#define KB(x) ((x) * 1024) +#define MB(x) ((x) * KB(1024)) + +#define GTT_PAGE_SIZE KB(4) +#define ROUND_TO(x, y) (((x) + (y) - 1) / (y) * (y)) +#define ROUND_DOWN_TO(x, y) ((x) / (y) * (y)) +#define ROUND_TO_PAGE(x) ROUND_TO((x), GTT_PAGE_SIZE) +#define ROUND_TO_MB(x) ROUND_TO((x), MB(1)) +#define PRIMARY_RINGBUFFER_SIZE KB(128) + + +/* Ring buffer registers, p277, overview p19 + */ +#define LP_RING 0x2030 +#define HP_RING 0x2040 + +#define RING_TAIL 0x00 +#define TAIL_ADDR 0x000FFFF8 +#define I830_TAIL_MASK 0x001FFFF8 + +#define RING_HEAD 0x04 +#define HEAD_WRAP_COUNT 0xFFE00000 +#define HEAD_WRAP_ONE 0x00200000 +#define HEAD_ADDR 0x001FFFFC +#define I830_HEAD_MASK 0x001FFFFC + +#define RING_START 0x08 +#define START_ADDR 0x03FFFFF8 +#define I830_RING_START_MASK 0xFFFFF000 + +#define RING_LEN 0x0C +#define RING_NR_PAGES 0x001FF000 +#define I830_RING_NR_PAGES 0x001FF000 +#define RING_REPORT_MASK 0x00000006 +#define RING_REPORT_64K 0x00000002 +#define RING_REPORT_128K 0x00000004 +#define RING_NO_REPORT 0x00000000 +#define RING_VALID_MASK 0x00000001 +#define RING_VALID 0x00000001 +#define RING_INVALID 0x00000000 + + +/* Fence/Tiling ranges [0..7] + */ +#define FENCE 0x2000 +#define FENCE_NR 8 + +#define I915G_FENCE_START_MASK 0x0ff00000 + +#define I830_FENCE_START_MASK 0x07f80000 + +#define FENCE_START_MASK 0x03F80000 +#define FENCE_X_MAJOR 0x00000000 +#define FENCE_Y_MAJOR 0x00001000 +#define FENCE_SIZE_MASK 0x00000700 +#define FENCE_SIZE_512K 0x00000000 +#define FENCE_SIZE_1M 0x00000100 +#define FENCE_SIZE_2M 0x00000200 +#define FENCE_SIZE_4M 0x00000300 +#define FENCE_SIZE_8M 0x00000400 +#define FENCE_SIZE_16M 0x00000500 +#define FENCE_SIZE_32M 0x00000600 +#define FENCE_SIZE_64M 0x00000700 +#define I915G_FENCE_SIZE_1M 0x00000000 +#define I915G_FENCE_SIZE_2M 0x00000100 +#define I915G_FENCE_SIZE_4M 0x00000200 +#define I915G_FENCE_SIZE_8M 0x00000300 +#define I915G_FENCE_SIZE_16M 0x00000400 +#define I915G_FENCE_SIZE_32M 0x00000500 +#define I915G_FENCE_SIZE_64M 0x00000600 +#define I915G_FENCE_SIZE_128M 0x00000700 +#define FENCE_PITCH_1 0x00000000 +#define FENCE_PITCH_2 0x00000010 +#define FENCE_PITCH_4 0x00000020 +#define FENCE_PITCH_8 0x00000030 +#define FENCE_PITCH_16 0x00000040 +#define FENCE_PITCH_32 0x00000050 +#define FENCE_PITCH_64 0x00000060 +#define FENCE_VALID 0x00000001 + +#include + +# define MMIO_IN8(base, offset) \ + *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) +# define MMIO_IN32(base, offset) \ + read_MMIO_LE32(base, offset) +# define MMIO_OUT8(base, offset, val) \ + *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val) +# define MMIO_OUT32(base, offset, val) \ + *(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val) + + + /* Memory mapped register access macros */ +#define INREG8(addr) MMIO_IN8(MMIO, addr) +#define INREG(addr) MMIO_IN32(MMIO, addr) +#define OUTREG8(addr, val) MMIO_OUT8(MMIO, addr, val) +#define OUTREG(addr, val) MMIO_OUT32(MMIO, addr, val) + +#define DSPABASE 0x70184 + +#endif diff --git a/src/gallium/winsys/dri/intel/server/intel_dri.c b/src/gallium/winsys/dri/intel/server/intel_dri.c new file mode 100644 index 0000000000..e49c4214ad --- /dev/null +++ b/src/gallium/winsys/dri/intel/server/intel_dri.c @@ -0,0 +1,1306 @@ +/** + * \file server/intel_dri.c + * \brief File to perform the device-specific initialization tasks typically + * done in the X server. + * + * Here they are converted to run in the client (or perhaps a standalone + * process), and to work with the frame buffer device rather than the X + * server infrastructure. + * + * Copyright (C) 2006 Dave Airlie (airlied@linux.ie) + + Permission is hereby granted, free of charge, to any person obtaining a + copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sub license, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial portions + of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR + ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +*/ + +#include +#include +#include +#include +#include + +#include "driver.h" +#include "drm.h" + +#include "intel.h" +#include "i830_dri.h" + +#include "memops.h" +#include "pciaccess.h" + +static size_t drm_page_size; +static int nextTile = 0; +#define xf86DrvMsg(...) do {} while(0) + +static const int pitches[] = { + 128 * 8, + 128 * 16, + 128 * 32, + 128 * 64, + 0 +}; + +static Bool I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea); + +static unsigned long +GetBestTileAlignment(unsigned long size) +{ + unsigned long i; + + for (i = KB(512); i < size; i <<= 1) + ; + + if (i > MB(64)) + i = MB(64); + + return i; +} + +static void SetFenceRegs(const DRIDriverContext *ctx, I830Rec *pI830) +{ + int i; + unsigned char *MMIO = ctx->MMIOAddress; + + for (i = 0; i < 8; i++) { + OUTREG(FENCE + i * 4, pI830->Fence[i]); + // if (I810_DEBUG & DEBUG_VERBOSE_VGA) + fprintf(stderr,"Fence Register : %x\n", pI830->Fence[i]); + } +} + +/* Tiled memory is good... really, really good... + * + * Need to make it less likely that we miss out on this - probably + * need to move the frontbuffer away from the 'guarenteed' alignment + * of the first memory segment, or perhaps allocate a discontigous + * framebuffer to get more alignment 'sweet spots'. + */ +static void +SetFence(const DRIDriverContext *ctx, I830Rec *pI830, + int nr, unsigned int start, unsigned int pitch, + unsigned int size) +{ + unsigned int val; + unsigned int fence_mask = 0; + unsigned int fence_pitch; + + if (nr < 0 || nr > 7) { + fprintf(stderr, + "SetFence: fence %d out of range\n",nr); + return; + } + + pI830->Fence[nr] = 0; + + if (IS_I9XX(pI830)) + fence_mask = ~I915G_FENCE_START_MASK; + else + fence_mask = ~I830_FENCE_START_MASK; + + if (start & fence_mask) { + fprintf(stderr, + "SetFence: %d: start (0x%08x) is not %s aligned\n", + nr, start, (IS_I9XX(pI830)) ? "1MB" : "512k"); + return; + } + + if (start % size) { + fprintf(stderr, + "SetFence: %d: start (0x%08x) is not size (%dk) aligned\n", + nr, start, size / 1024); + return; + } + + if (pitch & 127) { + fprintf(stderr, + "SetFence: %d: pitch (%d) not a multiple of 128 bytes\n", + nr, pitch); + return; + } + + val = (start | FENCE_X_MAJOR | FENCE_VALID); + + if (IS_I9XX(pI830)) { + switch (size) { + case MB(1): + val |= I915G_FENCE_SIZE_1M; + break; + case MB(2): + val |= I915G_FENCE_SIZE_2M; + break; + case MB(4): + val |= I915G_FENCE_SIZE_4M; + break; + case MB(8): + val |= I915G_FENCE_SIZE_8M; + break; + case MB(16): + val |= I915G_FENCE_SIZE_16M; + break; + case MB(32): + val |= I915G_FENCE_SIZE_32M; + break; + case MB(64): + val |= I915G_FENCE_SIZE_64M; + break; + default: + fprintf(stderr, + "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); + return; + } + } else { + switch (size) { + case KB(512): + val |= FENCE_SIZE_512K; + break; + case MB(1): + val |= FENCE_SIZE_1M; + break; + case MB(2): + val |= FENCE_SIZE_2M; + break; + case MB(4): + val |= FENCE_SIZE_4M; + break; + case MB(8): + val |= FENCE_SIZE_8M; + break; + case MB(16): + val |= FENCE_SIZE_16M; + break; + case MB(32): + val |= FENCE_SIZE_32M; + break; + case MB(64): + val |= FENCE_SIZE_64M; + break; + default: + fprintf(stderr, + "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); + return; + } + } + + if (IS_I9XX(pI830)) + fence_pitch = pitch / 512; + else + fence_pitch = pitch / 128; + + switch (fence_pitch) { + case 1: + val |= FENCE_PITCH_1; + break; + case 2: + val |= FENCE_PITCH_2; + break; + case 4: + val |= FENCE_PITCH_4; + break; + case 8: + val |= FENCE_PITCH_8; + break; + case 16: + val |= FENCE_PITCH_16; + break; + case 32: + val |= FENCE_PITCH_32; + break; + case 64: + val |= FENCE_PITCH_64; + break; + default: + fprintf(stderr, + "SetFence: %d: illegal pitch (%d)\n", nr, pitch); + return; + } + + pI830->Fence[nr] = val; +} + +static Bool +MakeTiles(const DRIDriverContext *ctx, I830Rec *pI830, I830MemRange *pMem) +{ + int pitch, ntiles, i; + + pitch = pMem->Pitch * ctx->cpp; + /* + * Simply try to break the region up into at most four pieces of size + * equal to the alignment. + */ + ntiles = ROUND_TO(pMem->Size, pMem->Alignment) / pMem->Alignment; + if (ntiles >= 4) { + return FALSE; + } + + for (i = 0; i < ntiles; i++, nextTile++) { + SetFence(ctx, pI830, nextTile, pMem->Start + i * pMem->Alignment, + pitch, pMem->Alignment); + } + return TRUE; +} + +static void I830SetupMemoryTiling(const DRIDriverContext *ctx, I830Rec *pI830) +{ + int i; + + /* Clear out */ + for (i = 0; i < 8; i++) + pI830->Fence[i] = 0; + + nextTile = 0; + + if (pI830->BackBuffer.Alignment >= KB(512)) { + if (MakeTiles(ctx, pI830, &(pI830->BackBuffer))) { + fprintf(stderr, + "Activating tiled memory for the back buffer.\n"); + } else { + fprintf(stderr, + "MakeTiles failed for the back buffer.\n"); + pI830->allowPageFlip = FALSE; + } + } + + if (pI830->DepthBuffer.Alignment >= KB(512)) { + if (MakeTiles(ctx, pI830, &(pI830->DepthBuffer))) { + fprintf(stderr, + "Activating tiled memory for the depth buffer.\n"); + } else { + fprintf(stderr, + "MakeTiles failed for the depth buffer.\n"); + } + } + + return; +} + +static int I830DetectMemory(const DRIDriverContext *ctx, I830Rec *pI830) +{ + struct pci_device host_bridge, ig_dev; + uint32_t gmch_ctrl; + int memsize = 0; + int range; + uint32_t aper_size; + uint32_t membase2 = 0; + + memset(&host_bridge, 0, sizeof(host_bridge)); + memset(&ig_dev, 0, sizeof(ig_dev)); + + ig_dev.dev = 2; + + pci_device_cfg_read_u32(&host_bridge, &gmch_ctrl, I830_GMCH_CTRL); + + if (IS_I830(pI830) || IS_845G(pI830)) { + if ((gmch_ctrl & I830_GMCH_MEM_MASK) == I830_GMCH_MEM_128M) { + aper_size = 0x80000000; + } else { + aper_size = 0x40000000; + } + } else { + if (IS_I9XX(pI830)) { + int ret; + ret = pci_device_cfg_read_u32(&ig_dev, &membase2, 0x18); + if (membase2 & 0x08000000) + aper_size = 0x8000000; + else + aper_size = 0x10000000; + + fprintf(stderr,"aper size is %08X %08x %d\n", aper_size, membase2, ret); + } else + aper_size = 0x8000000; + } + + pI830->aper_size = aper_size; + + + /* We need to reduce the stolen size, by the GTT and the popup. + * The GTT varying according the the FbMapSize and the popup is 4KB */ + range = (ctx->shared.fbSize / (1024*1024)) + 4; + + if (IS_I85X(pI830) || IS_I865G(pI830) || IS_I9XX(pI830)) { + switch (gmch_ctrl & I830_GMCH_GMS_MASK) { + case I855_GMCH_GMS_STOLEN_1M: + memsize = MB(1) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_4M: + memsize = MB(4) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_8M: + memsize = MB(8) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_16M: + memsize = MB(16) - KB(range); + break; + case I855_GMCH_GMS_STOLEN_32M: + memsize = MB(32) - KB(range); + break; + case I915G_GMCH_GMS_STOLEN_48M: + if (IS_I9XX(pI830)) + memsize = MB(48) - KB(range); + break; + case I915G_GMCH_GMS_STOLEN_64M: + if (IS_I9XX(pI830)) + memsize = MB(64) - KB(range); + break; + } + } else { + switch (gmch_ctrl & I830_GMCH_GMS_MASK) { + case I830_GMCH_GMS_STOLEN_512: + memsize = KB(512) - KB(range); + break; + case I830_GMCH_GMS_STOLEN_1024: + memsize = MB(1) - KB(range); + break; + case I830_GMCH_GMS_STOLEN_8192: + memsize = MB(8) - KB(range); + break; + case I830_GMCH_GMS_LOCAL: + memsize = 0; + xf86DrvMsg(pScrn->scrnIndex, X_WARNING, + "Local memory found, but won't be used.\n"); + break; + } + } + if (memsize > 0) { + fprintf(stderr, + "detected %d kB stolen memory.\n", memsize / 1024); + } else { + fprintf(stderr, + "no video memory detected.\n"); + } + return memsize; +} + +static int AgpInit(const DRIDriverContext *ctx, I830Rec *info) +{ + unsigned long mode = 0x4; + + if (drmAgpAcquire(ctx->drmFD) < 0) { + fprintf(stderr, "[gart] AGP not available\n"); + return 0; + } + + if (drmAgpEnable(ctx->drmFD, mode) < 0) { + fprintf(stderr, "[gart] AGP not enabled\n"); + drmAgpRelease(ctx->drmFD); + return 0; + } + else + fprintf(stderr, "[gart] AGP enabled at %dx\n", ctx->agpmode); + + return 1; +} + +/* + * Allocate memory from the given pool. Grow the pool if needed and if + * possible. + */ +static unsigned long +AllocFromPool(const DRIDriverContext *ctx, I830Rec *pI830, + I830MemRange *result, I830MemPool *pool, + long size, unsigned long alignment, int flags) +{ + long needed, start, end; + + if (!result || !pool || !size) + return 0; + + /* Calculate how much space is needed. */ + if (alignment <= GTT_PAGE_SIZE) + needed = size; + else { + start = ROUND_TO(pool->Free.Start, alignment); + end = ROUND_TO(start + size, alignment); + needed = end - pool->Free.Start; + } + if (needed > pool->Free.Size) { + return 0; + } + + result->Start = ROUND_TO(pool->Free.Start, alignment); + pool->Free.Start += needed; + result->End = pool->Free.Start; + + pool->Free.Size = pool->Free.End - pool->Free.Start; + result->Size = result->End - result->Start; + result->Pool = pool; + result->Alignment = alignment; + return needed; +} + +static unsigned long AllocFromAGP(const DRIDriverContext *ctx, I830Rec *pI830, long size, unsigned long alignment, I830MemRange *result) +{ + unsigned long start, end; + unsigned long newApStart, newApEnd; + int ret; + if (!result || !size) + return 0; + + if (!alignment) + alignment = 4; + + start = ROUND_TO(pI830->MemoryAperture.Start, alignment); + end = ROUND_TO(start + size, alignment); + newApStart = end; + newApEnd = pI830->MemoryAperture.End; + + ret=drmAgpAlloc(ctx->drmFD, size, 0, &(result->Physical), (drm_handle_t *)&(result->Key)); + + if (ret) + { + fprintf(stderr,"drmAgpAlloc failed %d\n", ret); + return 0; + } + pI830->allocatedMemory += size; + pI830->MemoryAperture.Start = newApStart; + pI830->MemoryAperture.End = newApEnd; + pI830->MemoryAperture.Size = newApEnd - newApStart; + // pI830->FreeMemory -= size; + result->Start = start; + result->End = start + size; + result->Size = size; + result->Offset = start; + result->Alignment = alignment; + result->Pool = NULL; + + return size; +} + +unsigned long +I830AllocVidMem(const DRIDriverContext *ctx, I830Rec *pI830, + I830MemRange *result, I830MemPool *pool, long size, + unsigned long alignment, int flags) +{ + unsigned long ret; + + if (!result) + return 0; + + /* Make sure these are initialised. */ + result->Size = 0; + result->Key = -1; + + if (!size) { + return 0; + } + + if (pool->Free.Size < size) { + ret = AllocFromAGP(ctx, pI830, size, alignment, result); + } + else { + ret = AllocFromPool(ctx, pI830, result, pool, size, alignment, flags); + if (ret == 0) + ret = AllocFromAGP(ctx, pI830, size, alignment, result); + } + return ret; +} + +static Bool BindAgpRange(const DRIDriverContext *ctx, I830MemRange *mem) +{ + if (!mem) + return FALSE; + + if (mem->Key == -1) + return TRUE; + + return !drmAgpBind(ctx->drmFD, mem->Key, mem->Offset); +} + +/* simple memory allocation routines needed */ +/* put ring buffer in low memory */ +/* need to allocate front, back, depth buffers aligned correctly, + allocate ring buffer, +*/ + +/* */ +static Bool +I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned long size, ret; + unsigned long lines, lineSize, align; + + /* allocate ring buffer */ + memset(pI830->LpRing, 0, sizeof(I830RingBuffer)); + pI830->LpRing->mem.Key = -1; + + size = PRIMARY_RINGBUFFER_SIZE; + + ret = I830AllocVidMem(ctx, pI830, &pI830->LpRing->mem, &pI830->StolenPool, size, 0x1000, 0); + + if (ret != size) + { + fprintf(stderr,"unable to allocate ring buffer %ld\n", ret); + return FALSE; + } + + pI830->LpRing->tail_mask = pI830->LpRing->mem.Size - 1; + + + /* allocate front buffer */ + memset(&(pI830->FrontBuffer), 0, sizeof(pI830->FrontBuffer)); + pI830->FrontBuffer.Key = -1; + pI830->FrontBuffer.Pitch = ctx->shared.virtualWidth; + + align = KB(512); + + lineSize = ctx->shared.virtualWidth * ctx->cpp; + lines = (ctx->shared.virtualHeight + 15) / 16 * 16; + size = lineSize * lines; + size = ROUND_TO_PAGE(size); + + align = GetBestTileAlignment(size); + + ret = I830AllocVidMem(ctx, pI830, &pI830->FrontBuffer, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate front buffer %ld\n", ret); + return FALSE; + } + + memset(&(pI830->BackBuffer), 0, sizeof(pI830->BackBuffer)); + pI830->BackBuffer.Key = -1; + pI830->BackBuffer.Pitch = ctx->shared.virtualWidth; + + ret = I830AllocVidMem(ctx, pI830, &pI830->BackBuffer, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate back buffer %ld\n", ret); + return FALSE; + } + + memset(&(pI830->DepthBuffer), 0, sizeof(pI830->DepthBuffer)); + pI830->DepthBuffer.Key = -1; + pI830->DepthBuffer.Pitch = ctx->shared.virtualWidth; + + ret = I830AllocVidMem(ctx, pI830, &pI830->DepthBuffer, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate depth buffer %ld\n", ret); + return FALSE; + } + + memset(&(pI830->ContextMem), 0, sizeof(pI830->ContextMem)); + pI830->ContextMem.Key = -1; + size = KB(32); + + ret = I830AllocVidMem(ctx, pI830, &pI830->ContextMem, &pI830->StolenPool, size, align, 0); + if (ret < size) + { + fprintf(stderr,"unable to allocate context buffer %ld\n", ret); + return FALSE; + } + +#if 0 + memset(&(pI830->TexMem), 0, sizeof(pI830->TexMem)); + pI830->TexMem.Key = -1; + + size = 32768 * 1024; + ret = AllocFromAGP(ctx, pI830, size, align, &pI830->TexMem); + if (ret < size) + { + fprintf(stderr,"unable to allocate texture memory %ld\n", ret); + return FALSE; + } +#endif + + return TRUE; +} + +static Bool +I830BindMemory(const DRIDriverContext *ctx, I830Rec *pI830) +{ + if (!BindAgpRange(ctx, &pI830->LpRing->mem)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->FrontBuffer)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->BackBuffer)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->DepthBuffer)) + return FALSE; + if (!BindAgpRange(ctx, &pI830->ContextMem)) + return FALSE; +#if 0 + if (!BindAgpRange(ctx, &pI830->TexMem)) + return FALSE; +#endif + return TRUE; +} + +static void SetupDRIMM(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned long aperEnd = ROUND_DOWN_TO(pI830->aper_size, GTT_PAGE_SIZE) / GTT_PAGE_SIZE; + unsigned long aperStart = ROUND_TO(pI830->aper_size - KB(32768), GTT_PAGE_SIZE) / GTT_PAGE_SIZE; + + fprintf(stderr, "aper size is %08X\n", ctx->shared.fbSize); + if (drmMMInit(ctx->drmFD, aperStart, aperEnd - aperStart, DRM_BO_MEM_TT)) { + fprintf(stderr, + "DRM MM Initialization Failed\n"); + } else { + fprintf(stderr, + "DRM MM Initialized at offset 0x%lx length %d page\n", aperStart, aperEnd-aperStart); + } + +} + +static Bool +I830CleanupDma(const DRIDriverContext *ctx) +{ + drmI830Init info; + + memset(&info, 0, sizeof(drmI830Init)); + info.func = I830_CLEANUP_DMA; + + if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, + &info, sizeof(drmI830Init))) { + fprintf(stderr, "I830 Dma Cleanup Failed\n"); + return FALSE; + } + + return TRUE; +} + +static Bool +I830InitDma(const DRIDriverContext *ctx, I830Rec *pI830) +{ + I830RingBuffer *ring = pI830->LpRing; + drmI830Init info; + + memset(&info, 0, sizeof(drmI830Init)); + info.func = I830_INIT_DMA; + + info.ring_start = ring->mem.Start + pI830->LinearAddr; + info.ring_end = ring->mem.End + pI830->LinearAddr; + info.ring_size = ring->mem.Size; + + info.mmio_offset = (unsigned int)ctx->MMIOStart; + + info.sarea_priv_offset = sizeof(drm_sarea_t); + + info.front_offset = pI830->FrontBuffer.Start; + info.back_offset = pI830->BackBuffer.Start; + info.depth_offset = pI830->DepthBuffer.Start; + info.w = ctx->shared.virtualWidth; + info.h = ctx->shared.virtualHeight; + info.pitch = ctx->shared.virtualWidth; + info.back_pitch = pI830->BackBuffer.Pitch; + info.depth_pitch = pI830->DepthBuffer.Pitch; + info.cpp = ctx->cpp; + + if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, + &info, sizeof(drmI830Init))) { + fprintf(stderr, + "I830 Dma Initialization Failed\n"); + return FALSE; + } + + return TRUE; +} + +static int I830CheckDRMVersion( const DRIDriverContext *ctx, + I830Rec *pI830 ) +{ + drmVersionPtr version; + + version = drmGetVersion(ctx->drmFD); + + if (version) { + int req_minor, req_patch; + + req_minor = 4; + req_patch = 0; + + if (version->version_major != 1 || + version->version_minor < req_minor || + (version->version_minor == req_minor && + version->version_patchlevel < req_patch)) { + /* Incompatible drm version */ + fprintf(stderr, + "[dri] I830DRIScreenInit failed because of a version " + "mismatch.\n" + "[dri] i915.o kernel module version is %d.%d.%d " + "but version 1.%d.%d or newer is needed.\n" + "[dri] Disabling DRI.\n", + version->version_major, + version->version_minor, + version->version_patchlevel, + req_minor, + req_patch); + drmFreeVersion(version); + return 0; + } + + pI830->drmMinor = version->version_minor; + drmFreeVersion(version); + } + return 1; +} + +static void +I830SetRingRegs(const DRIDriverContext *ctx, I830Rec *pI830) +{ + unsigned int itemp; + unsigned char *MMIO = ctx->MMIOAddress; + + OUTREG(LP_RING + RING_LEN, 0); + OUTREG(LP_RING + RING_TAIL, 0); + OUTREG(LP_RING + RING_HEAD, 0); + + if ((long)(pI830->LpRing->mem.Start & I830_RING_START_MASK) != + pI830->LpRing->mem.Start) { + fprintf(stderr, + "I830SetRingRegs: Ring buffer start (%lx) violates its " + "mask (%x)\n", pI830->LpRing->mem.Start, I830_RING_START_MASK); + } + /* Don't care about the old value. Reserved bits must be zero anyway. */ + itemp = pI830->LpRing->mem.Start & I830_RING_START_MASK; + OUTREG(LP_RING + RING_START, itemp); + + if (((pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES) != + pI830->LpRing->mem.Size - 4096) { + fprintf(stderr, + "I830SetRingRegs: Ring buffer size - 4096 (%lx) violates its " + "mask (%x)\n", pI830->LpRing->mem.Size - 4096, + I830_RING_NR_PAGES); + } + /* Don't care about the old value. Reserved bits must be zero anyway. */ + itemp = (pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES; + itemp |= (RING_NO_REPORT | RING_VALID); + OUTREG(LP_RING + RING_LEN, itemp); + + pI830->LpRing->head = INREG(LP_RING + RING_HEAD) & I830_HEAD_MASK; + pI830->LpRing->tail = INREG(LP_RING + RING_TAIL); + pI830->LpRing->space = pI830->LpRing->head - (pI830->LpRing->tail + 8); + if (pI830->LpRing->space < 0) + pI830->LpRing->space += pI830->LpRing->mem.Size; + + SetFenceRegs(ctx, pI830); + + /* RESET THE DISPLAY PIPE TO POINT TO THE FRONTBUFFER - hacky + hacky hacky */ + OUTREG(DSPABASE, pI830->FrontBuffer.Start + pI830->LinearAddr); + +} + +static Bool +I830SetParam(const DRIDriverContext *ctx, int param, int value) +{ + drmI830SetParam sp; + + memset(&sp, 0, sizeof(sp)); + sp.param = param; + sp.value = value; + + if (drmCommandWrite(ctx->drmFD, DRM_I830_SETPARAM, &sp, sizeof(sp))) { + fprintf(stderr, "I830 SetParam Failed\n"); + return FALSE; + } + + return TRUE; +} + +static Bool +I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ + fprintf(stderr, + "[drm] Mapping front buffer\n"); + + if (drmAddMap(ctx->drmFD, + (drm_handle_t)(sarea->front_offset + pI830->LinearAddr), + sarea->front_size, + DRM_FRAME_BUFFER, /*DRM_AGP,*/ + 0, + &sarea->front_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(front_handle) failed. Disabling DRI\n"); + return FALSE; + } + ctx->shared.hFrameBuffer = sarea->front_handle; + ctx->shared.fbSize = sarea->front_size; + fprintf(stderr, "[drm] Front Buffer = 0x%08x\n", + sarea->front_handle); + + if (drmAddMap(ctx->drmFD, + (drm_handle_t)(sarea->back_offset), + sarea->back_size, DRM_AGP, 0, + &sarea->back_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(back_handle) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] Back Buffer = 0x%08x\n", + sarea->back_handle); + + if (drmAddMap(ctx->drmFD, + (drm_handle_t)sarea->depth_offset, + sarea->depth_size, DRM_AGP, 0, + &sarea->depth_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(depth_handle) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] Depth Buffer = 0x%08x\n", + sarea->depth_handle); + +#if 0 + if (drmAddMap(ctx->drmFD, + (drm_handle_t)sarea->tex_offset, + sarea->tex_size, DRM_AGP, 0, + &sarea->tex_handle) < 0) { + fprintf(stderr, + "[drm] drmAddMap(tex_handle) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] textures = 0x%08x\n", + sarea->tex_handle); +#endif + return TRUE; +} + + +static void +I830DRIUnmapScreenRegions(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ +#if 1 + if (sarea->front_handle) { + drmRmMap(ctx->drmFD, sarea->front_handle); + sarea->front_handle = 0; + } +#endif + if (sarea->back_handle) { + drmRmMap(ctx->drmFD, sarea->back_handle); + sarea->back_handle = 0; + } + if (sarea->depth_handle) { + drmRmMap(ctx->drmFD, sarea->depth_handle); + sarea->depth_handle = 0; + } + if (sarea->tex_handle) { + drmRmMap(ctx->drmFD, sarea->tex_handle); + sarea->tex_handle = 0; + } +} + +static Bool +I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ + if (drmAddMap(ctx->drmFD, + (drm_handle_t)pI830->LpRing->mem.Start, + pI830->LpRing->mem.Size, DRM_AGP, 0, + &pI830->ring_map) < 0) { + fprintf(stderr, + "[drm] drmAddMap(ring_map) failed. Disabling DRI\n"); + return FALSE; + } + fprintf(stderr, "[drm] ring buffer = 0x%08x\n", + pI830->ring_map); + + if (I830InitDma(ctx, pI830) == FALSE) { + return FALSE; + } + + /* init to zero to be safe */ + + I830DRIMapScreenRegions(ctx, pI830, sarea); + SetupDRIMM(ctx, pI830); + + if (ctx->pciDevice != PCI_CHIP_845_G && + ctx->pciDevice != PCI_CHIP_I830_M) { + I830SetParam(ctx, I830_SETPARAM_USE_MI_BATCHBUFFER_START, 1 ); + } + + /* Okay now initialize the dma engine */ + { + pI830->irq = drmGetInterruptFromBusID(ctx->drmFD, + ctx->pciBus, + ctx->pciDevice, + ctx->pciFunc); + + if (drmCtlInstHandler(ctx->drmFD, pI830->irq)) { + fprintf(stderr, + "[drm] failure adding irq handler\n"); + pI830->irq = 0; + return FALSE; + } + else + fprintf(stderr, + "[drm] dma control initialized, using IRQ %d\n", + pI830->irq); + } + + fprintf(stderr, "[dri] visual configs initialized\n"); + + return TRUE; +} + +static Bool +I830ClearScreen(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) +{ + /* need to drmMap front and back buffers and zero them */ + drmAddress map_addr; + int ret; + + ret = drmMap(ctx->drmFD, + sarea->front_handle, + sarea->front_size, + &map_addr); + + if (ret) + { + fprintf(stderr, "Unable to map front buffer\n"); + return FALSE; + } + + drimemsetio((char *)map_addr, + 0, + sarea->front_size); + drmUnmap(map_addr, sarea->front_size); + + + ret = drmMap(ctx->drmFD, + sarea->back_handle, + sarea->back_size, + &map_addr); + + if (ret) + { + fprintf(stderr, "Unable to map back buffer\n"); + return FALSE; + } + + drimemsetio((char *)map_addr, + 0, + sarea->back_size); + drmUnmap(map_addr, sarea->back_size); + + return TRUE; +} + +static Bool +I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) + +{ + I830DRIPtr pI830DRI; + drmI830Sarea *pSAREAPriv; + int err; + + drm_page_size = getpagesize(); + + pI830->registerSize = ctx->MMIOSize; + /* This is a hack for now. We have to have more than a 4k page here + * because of the size of the state. However, the state should be + * in a per-context mapping. This will be added in the Mesa 3.5 port + * of the I830 driver. + */ + ctx->shared.SAREASize = SAREA_MAX; + + /* Note that drmOpen will try to load the kernel module, if needed. */ + ctx->drmFD = drmOpen("i915", NULL ); + if (ctx->drmFD < 0) { + fprintf(stderr, "[drm] drmOpen failed\n"); + return 0; + } + + if ((err = drmSetBusid(ctx->drmFD, ctx->pciBusID)) < 0) { + fprintf(stderr, "[drm] drmSetBusid failed (%d, %s), %s\n", + ctx->drmFD, ctx->pciBusID, strerror(-err)); + return 0; + } + + if (drmAddMap( ctx->drmFD, + 0, + ctx->shared.SAREASize, + DRM_SHM, + DRM_CONTAINS_LOCK, + &ctx->shared.hSAREA) < 0) + { + fprintf(stderr, "[drm] drmAddMap failed\n"); + return 0; + } + + fprintf(stderr, "[drm] added %d byte SAREA at 0x%08x\n", + ctx->shared.SAREASize, ctx->shared.hSAREA); + + if (drmMap( ctx->drmFD, + ctx->shared.hSAREA, + ctx->shared.SAREASize, + (drmAddressPtr)(&ctx->pSAREA)) < 0) + { + fprintf(stderr, "[drm] drmMap failed\n"); + return 0; + + } + + memset(ctx->pSAREA, 0, ctx->shared.SAREASize); + fprintf(stderr, "[drm] mapped SAREA 0x%08x to %p, size %d\n", + ctx->shared.hSAREA, ctx->pSAREA, ctx->shared.SAREASize); + + + if (drmAddMap(ctx->drmFD, + ctx->MMIOStart, + ctx->MMIOSize, + DRM_REGISTERS, + DRM_READ_ONLY, + &pI830->registerHandle) < 0) { + fprintf(stderr, "[drm] drmAddMap mmio failed\n"); + return 0; + } + fprintf(stderr, + "[drm] register handle = 0x%08x\n", pI830->registerHandle); + + + if (!I830CheckDRMVersion(ctx, pI830)) { + return FALSE; + } + + /* Create a 'server' context so we can grab the lock for + * initialization ioctls. + */ + if ((err = drmCreateContext(ctx->drmFD, &ctx->serverContext)) != 0) { + fprintf(stderr, "%s: drmCreateContext failed %d\n", __FUNCTION__, err); + return 0; + } + + DRM_LOCK(ctx->drmFD, ctx->pSAREA, ctx->serverContext, 0); + + /* Initialize the SAREA private data structure */ + pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + + sizeof(drm_sarea_t)); + memset(pSAREAPriv, 0, sizeof(*pSAREAPriv)); + + pI830->StolenMemory.Size = I830DetectMemory(ctx, pI830); + pI830->StolenMemory.Start = 0; + pI830->StolenMemory.End = pI830->StolenMemory.Size; + + pI830->MemoryAperture.Start = pI830->StolenMemory.End; + pI830->MemoryAperture.End = KB(40000); + pI830->MemoryAperture.Size = pI830->MemoryAperture.End - pI830->MemoryAperture.Start; + + pI830->StolenPool.Fixed = pI830->StolenMemory; + pI830->StolenPool.Total = pI830->StolenMemory; + pI830->StolenPool.Free = pI830->StolenPool.Total; + pI830->FreeMemory = pI830->StolenPool.Total.Size; + + if (!AgpInit(ctx, pI830)) + return FALSE; + + if (I830AllocateMemory(ctx, pI830) == FALSE) + { + return FALSE; + } + + if (I830BindMemory(ctx, pI830) == FALSE) + { + return FALSE; + } + + pSAREAPriv->rotated_offset = -1; + pSAREAPriv->rotated_size = 0; + pSAREAPriv->rotated_pitch = ctx->shared.virtualWidth; + + pSAREAPriv->front_offset = pI830->FrontBuffer.Start; + pSAREAPriv->front_size = pI830->FrontBuffer.Size; + pSAREAPriv->width = ctx->shared.virtualWidth; + pSAREAPriv->height = ctx->shared.virtualHeight; + pSAREAPriv->pitch = ctx->shared.virtualWidth; + pSAREAPriv->virtualX = ctx->shared.virtualWidth; + pSAREAPriv->virtualY = ctx->shared.virtualHeight; + pSAREAPriv->back_offset = pI830->BackBuffer.Start; + pSAREAPriv->back_size = pI830->BackBuffer.Size; + pSAREAPriv->depth_offset = pI830->DepthBuffer.Start; + pSAREAPriv->depth_size = pI830->DepthBuffer.Size; +#if 0 + pSAREAPriv->tex_offset = pI830->TexMem.Start; + pSAREAPriv->tex_size = pI830->TexMem.Size; +#endif + pSAREAPriv->log_tex_granularity = pI830->TexGranularity; + + ctx->driverClientMsg = malloc(sizeof(I830DRIRec)); + ctx->driverClientMsgSize = sizeof(I830DRIRec); + pI830DRI = (I830DRIPtr)ctx->driverClientMsg; + pI830DRI->deviceID = pI830->Chipset; + pI830DRI->regsSize = I830_REG_SIZE; + pI830DRI->width = ctx->shared.virtualWidth; + pI830DRI->height = ctx->shared.virtualHeight; + pI830DRI->mem = ctx->shared.fbSize; + pI830DRI->cpp = ctx->cpp; + + pI830DRI->bitsPerPixel = ctx->bpp; + pI830DRI->sarea_priv_offset = sizeof(drm_sarea_t); + + err = I830DRIDoMappings(ctx, pI830, pSAREAPriv); + if (err == FALSE) + return FALSE; + + I830SetupMemoryTiling(ctx, pI830); + + /* Quick hack to clear the front & back buffers. Could also use + * the clear ioctl to do this, but would need to setup hw state + * first. + */ + I830ClearScreen(ctx, pI830, pSAREAPriv); + + I830SetRingRegs(ctx, pI830); + + return TRUE; +} + + +/** + * \brief Validate the fbdev mode. + * + * \param ctx display handle. + * + * \return one on success, or zero on failure. + * + * Saves some registers and returns 1. + * + * \sa radeonValidateMode(). + */ +static int i830ValidateMode( const DRIDriverContext *ctx ) +{ + return 1; +} + +/** + * \brief Examine mode returned by fbdev. + * + * \param ctx display handle. + * + * \return one on success, or zero on failure. + * + * Restores registers that fbdev has clobbered and returns 1. + * + * \sa i810ValidateMode(). + */ +static int i830PostValidateMode( const DRIDriverContext *ctx ) +{ + I830Rec *pI830 = ctx->driverPrivate; + + I830SetRingRegs(ctx, pI830); + return 1; +} + + +/** + * \brief Initialize the framebuffer device mode + * + * \param ctx display handle. + * + * \return one on success, or zero on failure. + * + * Fills in \p info with some default values and some information from \p ctx + * and then calls I810ScreenInit() for the screen initialization. + * + * Before exiting clears the framebuffer memory accessing it directly. + */ +static int i830InitFBDev( DRIDriverContext *ctx ) +{ + I830Rec *pI830 = calloc(1, sizeof(I830Rec)); + int i; + + { + int dummy = ctx->shared.virtualWidth; + + switch (ctx->bpp / 8) { + case 1: dummy = (ctx->shared.virtualWidth + 127) & ~127; break; + case 2: dummy = (ctx->shared.virtualWidth + 31) & ~31; break; + case 3: + case 4: dummy = (ctx->shared.virtualWidth + 15) & ~15; break; + } + + ctx->shared.virtualWidth = dummy; + ctx->shared.Width = ctx->shared.virtualWidth; + } + + + for (i = 0; pitches[i] != 0; i++) { + if (pitches[i] >= ctx->shared.virtualWidth) { + ctx->shared.virtualWidth = pitches[i]; + break; + } + } + + ctx->driverPrivate = (void *)pI830; + + pI830->LpRing = calloc(1, sizeof(I830RingBuffer)); + pI830->Chipset = ctx->chipset; + pI830->LinearAddr = ctx->FBStart; + + if (!I830ScreenInit( ctx, pI830 )) + return 0; + + + return 1; +} + + +/** + * \brief The screen is being closed, so clean up any state and free any + * resources used by the DRI. + * + * \param ctx display handle. + * + * Unmaps the SAREA, closes the DRM device file descriptor and frees the driver + * private data. + */ +static void i830HaltFBDev( DRIDriverContext *ctx ) +{ + drmI830Sarea *pSAREAPriv; + I830Rec *pI830 = ctx->driverPrivate; + + if (pI830->irq) { + drmCtlUninstHandler(ctx->drmFD); + pI830->irq = 0; } + + I830CleanupDma(ctx); + + pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + + sizeof(drm_sarea_t)); + + I830DRIUnmapScreenRegions(ctx, pI830, pSAREAPriv); + drmUnmap( ctx->pSAREA, ctx->shared.SAREASize ); + drmClose(ctx->drmFD); + + if (ctx->driverPrivate) { + free(ctx->driverPrivate); + ctx->driverPrivate = 0; + } +} + + +extern void i810NotifyFocus( int ); + +/** + * \brief Exported driver interface for Mini GLX. + * + * \sa DRIDriverRec. + */ +const struct DRIDriverRec __driDriver = { + i830ValidateMode, + i830PostValidateMode, + i830InitFBDev, + i830HaltFBDev, + NULL,//I830EngineShutdown, + NULL, //I830EngineRestore, +#ifndef _EMBEDDED + 0, +#else + i810NotifyFocus, +#endif +}; diff --git a/src/gallium/winsys/xlib/brw_aub.c b/src/gallium/winsys/xlib/brw_aub.c new file mode 100644 index 0000000000..541d50c6e4 --- /dev/null +++ b/src/gallium/winsys/xlib/brw_aub.c @@ -0,0 +1,392 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#include "brw_aub.h" +#include "pipe/p_context.h" +#include "pipe/p_state.h" +#include "imports.h" +//#include "intel_winsys.h" + + +struct brw_aubfile { + FILE *file; + unsigned next_free_page; +}; + + +extern char *__progname; + + +struct aub_file_header { + unsigned int instruction_type; + unsigned int pad0:16; + unsigned int minor:8; + unsigned int major:8; + unsigned char application[8*4]; + unsigned int day:8; + unsigned int month:8; + unsigned int year:16; + unsigned int timezone:8; + unsigned int second:8; + unsigned int minute:8; + unsigned int hour:8; + unsigned int comment_length:16; + unsigned int pad1:16; +}; + +struct aub_block_header { + unsigned int instruction_type; + unsigned int operation:8; + unsigned int type:8; + unsigned int address_space:8; + unsigned int pad0:8; + unsigned int general_state_type:8; + unsigned int surface_state_type:8; + unsigned int pad1:16; + unsigned int address; + unsigned int length; +}; + +struct aub_dump_bmp { + unsigned int instruction_type; + unsigned int xmin:16; + unsigned int ymin:16; + unsigned int pitch:16; + unsigned int bpp:8; + unsigned int format:8; + unsigned int xsize:16; + unsigned int ysize:16; + unsigned int addr; + unsigned int unknown; +}; + +enum bh_operation { + BH_COMMENT, + BH_DATA_WRITE, + BH_COMMAND_WRITE, + BH_MMI0_WRITE32, + BH_END_SCENE, + BH_CONFIG_MEMORY_MAP, + BH_MAX_OPERATION +}; + +enum command_write_type { + CW_HWB_RING = 1, + CW_PRIMARY_RING_A, + CW_PRIMARY_RING_B, /* XXX - disagreement with listaub! */ + CW_PRIMARY_RING_C, + CW_MAX_TYPE +}; + +enum memory_map_type { + MM_DEFAULT, + MM_DYNAMIC, + MM_MAX_TYPE +}; + +enum address_space { + ADDR_GTT, + ADDR_LOCAL, + ADDR_MAIN, + ADDR_MAX +}; + + +#define AUB_FILE_HEADER 0xe085000b +#define AUB_BLOCK_HEADER 0xe0c10003 +#define AUB_DUMP_BMP 0xe09e0004 + +/* Registers to control page table + */ +#define PGETBL_CTL 0x2020 +#define PGETBL_ENABLED 0x1 + +#define NR_GTT_ENTRIES 65536 /* 256 mb */ + +#define FAIL \ +do { \ + fprintf(stderr, "failed to write aub data at %s/%d\n", __FUNCTION__, __LINE__); \ + exit(1); \ +} while (0) + + +/* Emit the headers at the top of each aubfile. Initialize the GTT. + */ +static void init_aubfile( FILE *aub_file ) +{ + struct aub_file_header fh; + struct aub_block_header bh; + unsigned int data; + + static int nr; + + nr++; + + /* Emit the aub header: + */ + memset(&fh, 0, sizeof(fh)); + + fh.instruction_type = AUB_FILE_HEADER; + fh.minor = 0x0; + fh.major = 0x7; + memcpy(fh.application, __progname, sizeof(fh.application)); + fh.day = (nr>>24) & 0xff; + fh.month = 0x0; + fh.year = 0x0; + fh.timezone = 0x0; + fh.second = nr & 0xff; + fh.minute = (nr>>8) & 0xff; + fh.hour = (nr>>16) & 0xff; + fh.comment_length = 0x0; + + if (fwrite(&fh, sizeof(fh), 1, aub_file) < 0) + FAIL; + + /* Setup the GTT starting at main memory address zero (!): + */ + memset(&bh, 0, sizeof(bh)); + + bh.instruction_type = AUB_BLOCK_HEADER; + bh.operation = BH_MMI0_WRITE32; + bh.type = 0x0; + bh.address_space = ADDR_GTT; /* ??? */ + bh.general_state_type = 0x0; + bh.surface_state_type = 0x0; + bh.address = PGETBL_CTL; + bh.length = 0x4; + + if (fwrite(&bh, sizeof(bh), 1, aub_file) < 0) + FAIL; + + data = 0x0 | PGETBL_ENABLED; + + if (fwrite(&data, sizeof(data), 1, aub_file) < 0) + FAIL; +} + + +static void init_aub_gtt( struct brw_aubfile *aubfile, + unsigned start_offset, + unsigned size ) +{ + FILE *aub_file = aubfile->file; + struct aub_block_header bh; + unsigned int i; + + assert(start_offset + size < NR_GTT_ENTRIES * 4096); + + + memset(&bh, 0, sizeof(bh)); + + bh.instruction_type = AUB_BLOCK_HEADER; + bh.operation = BH_DATA_WRITE; + bh.type = 0x0; + bh.address_space = ADDR_MAIN; + bh.general_state_type = 0x0; + bh.surface_state_type = 0x0; + bh.address = start_offset / 4096 * 4; + bh.length = size / 4096 * 4; + + if (fwrite(&bh, sizeof(bh), 1, aub_file) < 0) + FAIL; + + for (i = 0; i < size / 4096; i++) { + unsigned data = aubfile->next_free_page | 1; + + aubfile->next_free_page += 4096; + + if (fwrite(&data, sizeof(data), 1, aub_file) < 0) + FAIL; + } + +} + +static void write_block_header( FILE *aub_file, + struct aub_block_header *bh, + const unsigned *data, + unsigned sz ) +{ + sz = (sz + 3) & ~3; + + if (fwrite(bh, sizeof(*bh), 1, aub_file) < 0) + FAIL; + + if (fwrite(data, sz, 1, aub_file) < 0) + FAIL; + + fflush(aub_file); +} + + +static void write_dump_bmp( FILE *aub_file, + struct aub_dump_bmp *db ) +{ + if (fwrite(db, sizeof(*db), 1, aub_file) < 0) + FAIL; + + fflush(aub_file); +} + + + +void brw_aub_gtt_data( struct brw_aubfile *aubfile, + unsigned offset, + const void *data, + unsigned sz, + unsigned type, + unsigned state_type ) +{ + struct aub_block_header bh; + + bh.instruction_type = AUB_BLOCK_HEADER; + bh.operation = BH_DATA_WRITE; + bh.type = type; + bh.address_space = ADDR_GTT; + bh.pad0 = 0; + + if (type == DW_GENERAL_STATE) { + bh.general_state_type = state_type; + bh.surface_state_type = 0; + } + else { + bh.general_state_type = 0; + bh.surface_state_type = state_type; + } + + bh.pad1 = 0; + bh.address = offset; + bh.length = sz; + + write_block_header(aubfile->file, &bh, data, sz); +} + + + +void brw_aub_gtt_cmds( struct brw_aubfile *aubfile, + unsigned offset, + const void *data, + unsigned sz ) +{ + struct aub_block_header bh; + unsigned type = CW_PRIMARY_RING_A; + + + bh.instruction_type = AUB_BLOCK_HEADER; + bh.operation = BH_COMMAND_WRITE; + bh.type = type; + bh.address_space = ADDR_GTT; + bh.pad0 = 0; + bh.general_state_type = 0; + bh.surface_state_type = 0; + bh.pad1 = 0; + bh.address = offset; + bh.length = sz; + + write_block_header(aubfile->file, &bh, data, sz); +} + +void brw_aub_dump_bmp( struct brw_aubfile *aubfile, + struct pipe_surface *surface, + unsigned gtt_offset ) +{ + struct aub_dump_bmp db; + unsigned format; + + if (surface->cpp == 4) + format = 0x7; + else + format = 0x3; + + db.instruction_type = AUB_DUMP_BMP; + db.xmin = 0; + db.ymin = 0; + db.format = format; + db.bpp = surface->cpp * 8; + db.pitch = surface->pitch; + db.xsize = surface->width; + db.ysize = surface->height; + db.addr = gtt_offset; + db.unknown = /* surface->tiled ? 0x4 : */ 0x0; + + write_dump_bmp(aubfile->file, &db); +} + + + +struct brw_aubfile *brw_aubfile_create( void ) +{ + struct brw_aubfile *aubfile = CALLOC_STRUCT(brw_aubfile); + char filename[80]; + int val; + static int i = 0; + + i++; + + if (_mesa_getenv("INTEL_AUBFILE")) { + val = snprintf(filename, sizeof(filename), "%s%d.aub", _mesa_getenv("INTEL_AUBFILE"), i%4); + _mesa_printf("--> Aub file: %s\n", filename); + aubfile->file = fopen(filename, "w"); + } + else { + val = snprintf(filename, sizeof(filename), "%s.aub", __progname); + if (val < 0 || val > sizeof(filename)) + strcpy(filename, "default.aub"); + + _mesa_printf("--> Aub file: %s\n", filename); + aubfile->file = fopen(filename, "w"); + } + + if (!aubfile->file) { + _mesa_printf("couldn't open aubfile\n"); + exit(1); + } + + init_aubfile(aubfile->file); + + /* The GTT is located starting address zero in main memory. Pages + * to populate the gtt start after this point. + */ + aubfile->next_free_page = (NR_GTT_ENTRIES * 4 + 4095) & ~4095; + + /* More or less correspond with all the agp regions mapped by the + * driver: + */ + init_aub_gtt(aubfile, 0, 4096*4); + init_aub_gtt(aubfile, AUB_BUF_START, AUB_BUF_SIZE); + + return aubfile; +} + +void brw_aub_destroy( struct brw_aubfile *aubfile ) +{ + fclose(aubfile->file); + FREE(aubfile); +} diff --git a/src/gallium/winsys/xlib/brw_aub.h b/src/gallium/winsys/xlib/brw_aub.h new file mode 100644 index 0000000000..f5c60c7be2 --- /dev/null +++ b/src/gallium/winsys/xlib/brw_aub.h @@ -0,0 +1,114 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + +#ifndef BRW_AUB_H +#define BRW_AUB_H + +/* We set up this region, buffers may be allocated here: + */ +#define AUB_BUF_START (4096*4) +#define AUB_BUF_SIZE (8*1024*1024) + +struct intel_context; +struct pipe_surface; + +struct brw_aubfile *brw_aubfile_create( void ); + +void brw_aub_destroy( struct brw_aubfile *aubfile ); + +void brw_aub_gtt_data( struct brw_aubfile *aubfile, + unsigned offset, + const void *data, + unsigned sz, + unsigned type, + unsigned state_type ); + +void brw_aub_gtt_cmds( struct brw_aubfile *aubfile, + unsigned offset, + const void *data, + unsigned sz ); + +void brw_aub_dump_bmp( struct brw_aubfile *aubfile, + struct pipe_surface *surface, + unsigned gtt_offset ); + + +enum data_write_type { + DW_NOTYPE, + DW_BATCH_BUFFER, + DW_BIN_BUFFER, + DW_BIN_POINTER_LIST, + DW_SLOW_STATE_BUFFER, + DW_VERTEX_BUFFER, + DW_2D_MAP, + DW_CUBE_MAP, + DW_INDIRECT_STATE_BUFFER, + DW_VOLUME_MAP, + DW_1D_MAP, + DW_CONSTANT_BUFFER, + DW_CONSTANT_URB_ENTRY, + DW_INDEX_BUFFER, + DW_GENERAL_STATE, + DW_SURFACE_STATE, + DW_MEDIA_OBJECT_INDIRECT_DATA, + DW_MAX_TYPE +}; + +enum data_write_general_state_type { + DWGS_NOTYPE, + DWGS_VERTEX_SHADER_STATE, + DWGS_GEOMETRY_SHADER_STATE , + DWGS_CLIPPER_STATE, + DWGS_STRIPS_FANS_STATE, + DWGS_WINDOWER_IZ_STATE, + DWGS_COLOR_CALC_STATE, + DWGS_CLIPPER_VIEWPORT_STATE, /* was 0x7 */ + DWGS_STRIPS_FANS_VIEWPORT_STATE, + DWGS_COLOR_CALC_VIEWPORT_STATE, /* was 0x9 */ + DWGS_SAMPLER_STATE, + DWGS_KERNEL_INSTRUCTIONS, + DWGS_SCRATCH_SPACE, + DWGS_SAMPLER_DEFAULT_COLOR, + DWGS_INTERFACE_DESCRIPTOR, + DWGS_VLD_STATE, + DWGS_VFE_STATE, + DWGS_MAX_TYPE +}; + +enum data_write_surface_state_type { + DWSS_NOTYPE, + DWSS_BINDING_TABLE_STATE, + DWSS_SURFACE_STATE, + DWSS_MAX_TYPE +}; + + +#endif diff --git a/src/gallium/winsys/xlib/fakeglx.c b/src/gallium/winsys/xlib/fakeglx.c new file mode 100644 index 0000000000..902a755075 --- /dev/null +++ b/src/gallium/winsys/xlib/fakeglx.c @@ -0,0 +1,3188 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* + * This is an emulation of the GLX API which allows Mesa/GLX-based programs + * to run on X servers which do not have the real GLX extension. + * + * Thanks to the contributors: + * + * Initial version: Philip Brown (phil@bolthole.com) + * Better glXGetConfig() support: Armin Liebchen (liebchen@asylum.cs.utah.edu) + * Further visual-handling refinements: Wolfram Gloger + * (wmglo@Dent.MED.Uni-Muenchen.DE). + * + * Notes: + * Don't be fooled, stereo isn't supported yet. + */ + + + +#include "glxheader.h" +#include "glxapi.h" +#include "GL/xmesa.h" +#include "context.h" +#include "config.h" +#include "macros.h" +#include "imports.h" +#include "mtypes.h" +#include "version.h" +#include "xfonts.h" +#include "xmesaP.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_public.h" + + +#ifdef __VMS +#define _mesa_sprintf sprintf +#endif + +/* This indicates the client-side GLX API and GLX encoder version. */ +#define CLIENT_MAJOR_VERSION 1 +#define CLIENT_MINOR_VERSION 4 /* but don't have 1.3's pbuffers, etc yet */ + +/* This indicates the server-side GLX decoder version. + * GLX 1.4 indicates OpenGL 1.3 support + */ +#define SERVER_MAJOR_VERSION 1 +#define SERVER_MINOR_VERSION 4 + +/* This is appended onto the glXGetClient/ServerString version strings. */ +#define MESA_GLX_VERSION "Mesa " MESA_VERSION_STRING + +/* Who implemented this GLX? */ +#define VENDOR "Brian Paul" + +#define EXTENSIONS \ + "GLX_MESA_set_3dfx_mode " \ + "GLX_MESA_copy_sub_buffer " \ + "GLX_MESA_pixmap_colormap " \ + "GLX_MESA_release_buffers " \ + "GLX_ARB_get_proc_address " \ + "GLX_EXT_texture_from_pixmap " \ + "GLX_EXT_visual_info " \ + "GLX_EXT_visual_rating " \ + /*"GLX_SGI_video_sync "*/ \ + "GLX_SGIX_fbconfig " \ + "GLX_SGIX_pbuffer " + +/* + * Our fake GLX context will contain a "real" GLX context and an XMesa context. + * + * Note that a pointer to a __GLXcontext is a pointer to a fake_glx_context, + * and vice versa. + * + * We really just need this structure in order to make the libGL functions + * glXGetCurrentContext(), glXGetCurrentDrawable() and glXGetCurrentDisplay() + * work correctly. + */ +struct fake_glx_context { + __GLXcontext glxContext; /* this MUST be first! */ + XMesaContext xmesaContext; +}; + + + +/**********************************************************************/ +/*** GLX Visual Code ***/ +/**********************************************************************/ + +#define DONT_CARE -1 + + +static XMesaVisual *VisualTable = NULL; +static int NumVisuals = 0; + + +/* + * This struct and some code fragments borrowed + * from Mark Kilgard's GLUT library. + */ +typedef struct _OverlayInfo { + /* Avoid 64-bit portability problems by being careful to use + longs due to the way XGetWindowProperty is specified. Note + that these parameters are passed as CARD32s over X + protocol. */ + unsigned long overlay_visual; + long transparent_type; + long value; + long layer; +} OverlayInfo; + + + +/* Macro to handle c_class vs class field name in XVisualInfo struct */ +#if defined(__cplusplus) || defined(c_plusplus) +#define CLASS c_class +#else +#define CLASS class +#endif + + + +/* + * Test if the given XVisualInfo is usable for Mesa rendering. + */ +static GLboolean +is_usable_visual( XVisualInfo *vinfo ) +{ + switch (vinfo->CLASS) { + case StaticGray: + case GrayScale: + /* Any StaticGray/GrayScale visual works in RGB or CI mode */ + return GL_TRUE; + case StaticColor: + case PseudoColor: + /* Any StaticColor/PseudoColor visual of at least 4 bits */ + if (vinfo->depth>=4) { + return GL_TRUE; + } + else { + return GL_FALSE; + } + case TrueColor: + case DirectColor: + /* Any depth of TrueColor or DirectColor works in RGB mode */ + return GL_TRUE; + default: + /* This should never happen */ + return GL_FALSE; + } +} + + + +/** + * Get an array OverlayInfo records for specified screen. + * \param dpy the display + * \param screen screen number + * \param numOverlays returns numver of OverlayInfo records + * \return pointer to OverlayInfo array, free with XFree() + */ +static OverlayInfo * +GetOverlayInfo(Display *dpy, int screen, int *numOverlays) +{ + Atom overlayVisualsAtom; + Atom actualType; + Status status; + unsigned char *ovInfo; + unsigned long sizeData, bytesLeft; + int actualFormat; + + /* + * The SERVER_OVERLAY_VISUALS property on the root window contains + * a list of overlay visuals. Get that list now. + */ + overlayVisualsAtom = XInternAtom(dpy,"SERVER_OVERLAY_VISUALS", True); + if (overlayVisualsAtom == None) { + return 0; + } + + status = XGetWindowProperty(dpy, RootWindow(dpy, screen), + overlayVisualsAtom, 0L, (long) 10000, False, + overlayVisualsAtom, &actualType, &actualFormat, + &sizeData, &bytesLeft, + &ovInfo); + + if (status != Success || actualType != overlayVisualsAtom || + actualFormat != 32 || sizeData < 4) { + /* something went wrong */ + XFree((void *) ovInfo); + *numOverlays = 0; + return NULL; + } + + *numOverlays = sizeData / 4; + return (OverlayInfo *) ovInfo; +} + + + +/** + * Return the level (overlay, normal, underlay) of a given XVisualInfo. + * Input: dpy - the X display + * vinfo - the XVisualInfo to test + * Return: level of the visual: + * 0 = normal planes + * >0 = overlay planes + * <0 = underlay planes + */ +static int +level_of_visual( Display *dpy, XVisualInfo *vinfo ) +{ + OverlayInfo *overlay_info; + int numOverlaysPerScreen, i; + + overlay_info = GetOverlayInfo(dpy, vinfo->screen, &numOverlaysPerScreen); + if (!overlay_info) { + return 0; + } + + /* search the overlay visual list for the visual ID of interest */ + for (i = 0; i < numOverlaysPerScreen; i++) { + const OverlayInfo *ov = overlay_info + i; + if (ov->overlay_visual == vinfo->visualid) { + /* found the visual */ + if (/*ov->transparent_type==1 &&*/ ov->layer!=0) { + int level = ov->layer; + XFree((void *) overlay_info); + return level; + } + else { + XFree((void *) overlay_info); + return 0; + } + } + } + + /* The visual ID was not found in the overlay list. */ + XFree((void *) overlay_info); + return 0; +} + + + + +/* + * Given an XVisualInfo and RGB, Double, and Depth buffer flags, save the + * configuration in our list of GLX visuals. + */ +static XMesaVisual +save_glx_visual( Display *dpy, XVisualInfo *vinfo, + GLboolean rgbFlag, GLboolean alphaFlag, GLboolean dbFlag, + GLboolean stereoFlag, + GLint depth_size, GLint stencil_size, + GLint accumRedSize, GLint accumGreenSize, + GLint accumBlueSize, GLint accumAlphaSize, + GLint level, GLint numAuxBuffers ) +{ + GLboolean ximageFlag = GL_TRUE; + XMesaVisual xmvis; + GLint i; + GLboolean comparePointers; + + if (dbFlag) { + /* Check if the MESA_BACK_BUFFER env var is set */ + char *backbuffer = _mesa_getenv("MESA_BACK_BUFFER"); + if (backbuffer) { + if (backbuffer[0]=='p' || backbuffer[0]=='P') { + ximageFlag = GL_FALSE; + } + else if (backbuffer[0]=='x' || backbuffer[0]=='X') { + ximageFlag = GL_TRUE; + } + else { + _mesa_warning(NULL, "Mesa: invalid value for MESA_BACK_BUFFER environment variable, using an XImage."); + } + } + } + + if (stereoFlag) { + /* stereo not supported */ + return NULL; + } + + /* Comparing IDs uses less memory but sometimes fails. */ + /* XXX revisit this after 3.0 is finished. */ + if (_mesa_getenv("MESA_GLX_VISUAL_HACK")) + comparePointers = GL_TRUE; + else + comparePointers = GL_FALSE; + + /* Force the visual to have an alpha channel */ + if (rgbFlag && _mesa_getenv("MESA_GLX_FORCE_ALPHA")) + alphaFlag = GL_TRUE; + + /* First check if a matching visual is already in the list */ + for (i=0; idisplay == dpy + && v->mesa_visual.level == level + && v->mesa_visual.numAuxBuffers == numAuxBuffers + && v->ximage_flag == ximageFlag + && v->mesa_visual.rgbMode == rgbFlag + && v->mesa_visual.doubleBufferMode == dbFlag + && v->mesa_visual.stereoMode == stereoFlag + && (v->mesa_visual.alphaBits > 0) == alphaFlag + && (v->mesa_visual.depthBits >= depth_size || depth_size == 0) + && (v->mesa_visual.stencilBits >= stencil_size || stencil_size == 0) + && (v->mesa_visual.accumRedBits >= accumRedSize || accumRedSize == 0) + && (v->mesa_visual.accumGreenBits >= accumGreenSize || accumGreenSize == 0) + && (v->mesa_visual.accumBlueBits >= accumBlueSize || accumBlueSize == 0) + && (v->mesa_visual.accumAlphaBits >= accumAlphaSize || accumAlphaSize == 0)) { + /* now either compare XVisualInfo pointers or visual IDs */ + if ((!comparePointers && v->visinfo->visualid == vinfo->visualid) + || (comparePointers && v->vishandle == vinfo)) { + return v; + } + } + } + + /* Create a new visual and add it to the list. */ + + xmvis = XMesaCreateVisual( dpy, vinfo, rgbFlag, alphaFlag, dbFlag, + stereoFlag, ximageFlag, + depth_size, stencil_size, + accumRedSize, accumBlueSize, + accumBlueSize, accumAlphaSize, 0, level, + GLX_NONE_EXT ); + if (xmvis) { + /* Save a copy of the pointer now so we can find this visual again + * if we need to search for it in find_glx_visual(). + */ + xmvis->vishandle = vinfo; + /* Allocate more space for additional visual */ + VisualTable = (XMesaVisual *) _mesa_realloc( VisualTable, + sizeof(XMesaVisual) * NumVisuals, + sizeof(XMesaVisual) * (NumVisuals + 1)); + /* add xmvis to the list */ + VisualTable[NumVisuals] = xmvis; + NumVisuals++; + /* XXX minor hack, because XMesaCreateVisual doesn't support an + * aux buffers parameter. + */ + xmvis->mesa_visual.numAuxBuffers = numAuxBuffers; + } + return xmvis; +} + + +/** + * Return the default number of bits for the Z buffer. + * If defined, use the MESA_GLX_DEPTH_BITS env var value. + * Otherwise, use the DEFAULT_SOFTWARE_DEPTH_BITS constant. + * XXX probably do the same thing for stencil, accum, etc. + */ +static GLint +default_depth_bits(void) +{ + int zBits; + const char *zEnv = _mesa_getenv("MESA_GLX_DEPTH_BITS"); + if (zEnv) + zBits = _mesa_atoi(zEnv); + else + zBits = DEFAULT_SOFTWARE_DEPTH_BITS; + return zBits; +} + +static GLint +default_alpha_bits(void) +{ + int aBits; + const char *aEnv = _mesa_getenv("MESA_GLX_ALPHA_BITS"); + if (aEnv) + aBits = _mesa_atoi(aEnv); + else + aBits = 0; + return aBits; +} + +static GLint +default_accum_bits(void) +{ + return 16; +} + + + +/* + * Create a GLX visual from a regular XVisualInfo. + * This is called when Fake GLX is given an XVisualInfo which wasn't + * returned by glXChooseVisual. Since this is the first time we're + * considering this visual we'll take a guess at reasonable values + * for depth buffer size, stencil size, accum size, etc. + * This is the best we can do with a client-side emulation of GLX. + */ +static XMesaVisual +create_glx_visual( Display *dpy, XVisualInfo *visinfo ) +{ + int vislevel; + GLint zBits = 24; /*default_depth_bits();*/ + GLint accBits = default_accum_bits(); + GLboolean alphaFlag = default_alpha_bits() > 0; + + vislevel = level_of_visual( dpy, visinfo ); + if (vislevel) { + /* Configure this visual as a CI, single-buffered overlay */ + return save_glx_visual( dpy, visinfo, + GL_FALSE, /* rgb */ + GL_FALSE, /* alpha */ + GL_FALSE, /* double */ + GL_FALSE, /* stereo */ + 0, /* depth bits */ + 0, /* stencil bits */ + 0,0,0,0, /* accum bits */ + vislevel, /* level */ + 0 /* numAux */ + ); + } + else if (is_usable_visual( visinfo )) { + if (_mesa_getenv("MESA_GLX_FORCE_CI")) { + /* Configure this visual as a COLOR INDEX visual. */ + return save_glx_visual( dpy, visinfo, + GL_FALSE, /* rgb */ + GL_FALSE, /* alpha */ + GL_TRUE, /* double */ + GL_FALSE, /* stereo */ + zBits, + STENCIL_BITS, + 0, 0, 0, 0, /* accum bits */ + 0, /* level */ + 0 /* numAux */ + ); + } + else { + /* Configure this visual as RGB, double-buffered, depth-buffered. */ + /* This is surely wrong for some people's needs but what else */ + /* can be done? They should use glXChooseVisual(). */ + return save_glx_visual( dpy, visinfo, + GL_TRUE, /* rgb */ + alphaFlag, /* alpha */ + GL_TRUE, /* double */ + GL_FALSE, /* stereo */ + zBits, + STENCIL_BITS, + accBits, /* r */ + accBits, /* g */ + accBits, /* b */ + accBits, /* a */ + 0, /* level */ + 0 /* numAux */ + ); + } + } + else { + _mesa_warning(NULL, "Mesa: error in glXCreateContext: bad visual\n"); + return NULL; + } +} + + + +/* + * Find the GLX visual associated with an XVisualInfo. + */ +static XMesaVisual +find_glx_visual( Display *dpy, XVisualInfo *vinfo ) +{ + int i; + + /* try to match visual id */ + for (i=0;idisplay==dpy + && VisualTable[i]->visinfo->visualid == vinfo->visualid) { + return VisualTable[i]; + } + } + + /* if that fails, try to match pointers */ + for (i=0;idisplay==dpy && VisualTable[i]->vishandle==vinfo) { + return VisualTable[i]; + } + } + + return NULL; +} + + + +/** + * Return the transparent pixel value for a GLX visual. + * Input: glxvis - the glx_visual + * Return: a pixel value or -1 if no transparent pixel + */ +static int +transparent_pixel( XMesaVisual glxvis ) +{ + Display *dpy = glxvis->display; + XVisualInfo *vinfo = glxvis->visinfo; + OverlayInfo *overlay_info; + int numOverlaysPerScreen, i; + + overlay_info = GetOverlayInfo(dpy, vinfo->screen, &numOverlaysPerScreen); + if (!overlay_info) { + return -1; + } + + for (i = 0; i < numOverlaysPerScreen; i++) { + const OverlayInfo *ov = overlay_info + i; + if (ov->overlay_visual == vinfo->visualid) { + /* found it! */ + if (ov->transparent_type == 0) { + /* type 0 indicates no transparency */ + XFree((void *) overlay_info); + return -1; + } + else { + /* ov->value is the transparent pixel */ + XFree((void *) overlay_info); + return ov->value; + } + } + } + + /* The visual ID was not found in the overlay list. */ + XFree((void *) overlay_info); + return -1; +} + + + +/** + * Try to get an X visual which matches the given arguments. + */ +static XVisualInfo * +get_visual( Display *dpy, int scr, unsigned int depth, int xclass ) +{ + XVisualInfo temp, *vis; + long mask; + int n; + unsigned int default_depth; + int default_class; + + mask = VisualScreenMask | VisualDepthMask | VisualClassMask; + temp.screen = scr; + temp.depth = depth; + temp.CLASS = xclass; + + default_depth = DefaultDepth(dpy,scr); + default_class = DefaultVisual(dpy,scr)->CLASS; + + if (depth==default_depth && xclass==default_class) { + /* try to get root window's visual */ + temp.visualid = DefaultVisual(dpy,scr)->visualid; + mask |= VisualIDMask; + } + + vis = XGetVisualInfo( dpy, mask, &temp, &n ); + + /* In case bits/pixel > 24, make sure color channels are still <=8 bits. + * An SGI Infinite Reality system, for example, can have 30bpp pixels: + * 10 bits per color channel. Mesa's limited to a max of 8 bits/channel. + */ + if (vis && depth > 24 && (xclass==TrueColor || xclass==DirectColor)) { + if (_mesa_bitcount((GLuint) vis->red_mask ) <= 8 && + _mesa_bitcount((GLuint) vis->green_mask) <= 8 && + _mesa_bitcount((GLuint) vis->blue_mask ) <= 8) { + return vis; + } + else { + XFree((void *) vis); + return NULL; + } + } + + return vis; +} + + + +/* + * Retrieve the value of the given environment variable and find + * the X visual which matches it. + * Input: dpy - the display + * screen - the screen number + * varname - the name of the environment variable + * Return: an XVisualInfo pointer to NULL if error. + */ +static XVisualInfo * +get_env_visual(Display *dpy, int scr, const char *varname) +{ + char value[100], type[100]; + int depth, xclass = -1; + XVisualInfo *vis; + + if (!_mesa_getenv( varname )) { + return NULL; + } + + _mesa_strncpy( value, _mesa_getenv(varname), 100 ); + value[99] = 0; + + sscanf( value, "%s %d", type, &depth ); + + if (_mesa_strcmp(type,"TrueColor")==0) xclass = TrueColor; + else if (_mesa_strcmp(type,"DirectColor")==0) xclass = DirectColor; + else if (_mesa_strcmp(type,"PseudoColor")==0) xclass = PseudoColor; + else if (_mesa_strcmp(type,"StaticColor")==0) xclass = StaticColor; + else if (_mesa_strcmp(type,"GrayScale")==0) xclass = GrayScale; + else if (_mesa_strcmp(type,"StaticGray")==0) xclass = StaticGray; + + if (xclass>-1 && depth>0) { + vis = get_visual( dpy, scr, depth, xclass ); + if (vis) { + return vis; + } + } + + _mesa_warning(NULL, "GLX unable to find visual class=%s, depth=%d.", + type, depth); + + return NULL; +} + + + +/* + * Select an X visual which satisfies the RGBA/CI flag and minimum depth. + * Input: dpy, screen - X display and screen number + * rgba - GL_TRUE = RGBA mode, GL_FALSE = CI mode + * min_depth - minimum visual depth + * preferred_class - preferred GLX visual class or DONT_CARE + * Return: pointer to an XVisualInfo or NULL. + */ +static XVisualInfo * +choose_x_visual( Display *dpy, int screen, GLboolean rgba, int min_depth, + int preferred_class ) +{ + XVisualInfo *vis; + int xclass, visclass = 0; + int depth; + + if (rgba) { + Atom hp_cr_maps = XInternAtom(dpy, "_HP_RGB_SMOOTH_MAP_LIST", True); + /* First see if the MESA_RGB_VISUAL env var is defined */ + vis = get_env_visual( dpy, screen, "MESA_RGB_VISUAL" ); + if (vis) { + return vis; + } + /* Otherwise, search for a suitable visual */ + if (preferred_class==DONT_CARE) { + for (xclass=0;xclass<6;xclass++) { + switch (xclass) { + case 0: visclass = TrueColor; break; + case 1: visclass = DirectColor; break; + case 2: visclass = PseudoColor; break; + case 3: visclass = StaticColor; break; + case 4: visclass = GrayScale; break; + case 5: visclass = StaticGray; break; + } + if (min_depth==0) { + /* start with shallowest */ + for (depth=0;depth<=32;depth++) { + if (visclass==TrueColor && depth==8 && !hp_cr_maps) { + /* Special case: try to get 8-bit PseudoColor before */ + /* 8-bit TrueColor */ + vis = get_visual( dpy, screen, 8, PseudoColor ); + if (vis) { + return vis; + } + } + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + else { + /* start with deepest */ + for (depth=32;depth>=min_depth;depth--) { + if (visclass==TrueColor && depth==8 && !hp_cr_maps) { + /* Special case: try to get 8-bit PseudoColor before */ + /* 8-bit TrueColor */ + vis = get_visual( dpy, screen, 8, PseudoColor ); + if (vis) { + return vis; + } + } + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + } + } + else { + /* search for a specific visual class */ + switch (preferred_class) { + case GLX_TRUE_COLOR_EXT: visclass = TrueColor; break; + case GLX_DIRECT_COLOR_EXT: visclass = DirectColor; break; + case GLX_PSEUDO_COLOR_EXT: visclass = PseudoColor; break; + case GLX_STATIC_COLOR_EXT: visclass = StaticColor; break; + case GLX_GRAY_SCALE_EXT: visclass = GrayScale; break; + case GLX_STATIC_GRAY_EXT: visclass = StaticGray; break; + default: return NULL; + } + if (min_depth==0) { + /* start with shallowest */ + for (depth=0;depth<=32;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + else { + /* start with deepest */ + for (depth=32;depth>=min_depth;depth--) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + } + } + else { + /* First see if the MESA_CI_VISUAL env var is defined */ + vis = get_env_visual( dpy, screen, "MESA_CI_VISUAL" ); + if (vis) { + return vis; + } + /* Otherwise, search for a suitable visual, starting with shallowest */ + if (preferred_class==DONT_CARE) { + for (xclass=0;xclass<4;xclass++) { + switch (xclass) { + case 0: visclass = PseudoColor; break; + case 1: visclass = StaticColor; break; + case 2: visclass = GrayScale; break; + case 3: visclass = StaticGray; break; + } + /* try 8-bit up through 16-bit */ + for (depth=8;depth<=16;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + /* try min_depth up to 8-bit */ + for (depth=min_depth;depth<8;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + } + else { + /* search for a specific visual class */ + switch (preferred_class) { + case GLX_TRUE_COLOR_EXT: visclass = TrueColor; break; + case GLX_DIRECT_COLOR_EXT: visclass = DirectColor; break; + case GLX_PSEUDO_COLOR_EXT: visclass = PseudoColor; break; + case GLX_STATIC_COLOR_EXT: visclass = StaticColor; break; + case GLX_GRAY_SCALE_EXT: visclass = GrayScale; break; + case GLX_STATIC_GRAY_EXT: visclass = StaticGray; break; + default: return NULL; + } + /* try 8-bit up through 16-bit */ + for (depth=8;depth<=16;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + /* try min_depth up to 8-bit */ + for (depth=min_depth;depth<8;depth++) { + vis = get_visual( dpy, screen, depth, visclass ); + if (vis) { + return vis; + } + } + } + } + + /* didn't find a visual */ + return NULL; +} + + + +/* + * Find the deepest X over/underlay visual of at least min_depth. + * Input: dpy, screen - X display and screen number + * level - the over/underlay level + * trans_type - transparent pixel type: GLX_NONE_EXT, + * GLX_TRANSPARENT_RGB_EXT, GLX_TRANSPARENT_INDEX_EXT, + * or DONT_CARE + * trans_value - transparent pixel value or DONT_CARE + * min_depth - minimum visual depth + * preferred_class - preferred GLX visual class or DONT_CARE + * Return: pointer to an XVisualInfo or NULL. + */ +static XVisualInfo * +choose_x_overlay_visual( Display *dpy, int scr, GLboolean rgbFlag, + int level, int trans_type, int trans_value, + int min_depth, int preferred_class ) +{ + OverlayInfo *overlay_info; + int numOverlaysPerScreen; + int i; + XVisualInfo *deepvis; + int deepest; + + /*DEBUG int tt, tv; */ + + switch (preferred_class) { + case GLX_TRUE_COLOR_EXT: preferred_class = TrueColor; break; + case GLX_DIRECT_COLOR_EXT: preferred_class = DirectColor; break; + case GLX_PSEUDO_COLOR_EXT: preferred_class = PseudoColor; break; + case GLX_STATIC_COLOR_EXT: preferred_class = StaticColor; break; + case GLX_GRAY_SCALE_EXT: preferred_class = GrayScale; break; + case GLX_STATIC_GRAY_EXT: preferred_class = StaticGray; break; + default: preferred_class = DONT_CARE; + } + + overlay_info = GetOverlayInfo(dpy, scr, &numOverlaysPerScreen); + if (!overlay_info) { + return NULL; + } + + /* Search for the deepest overlay which satisifies all criteria. */ + deepest = min_depth; + deepvis = NULL; + + for (i = 0; i < numOverlaysPerScreen; i++) { + const OverlayInfo *ov = overlay_info + i; + XVisualInfo *vislist, vistemplate; + int count; + + if (ov->layer!=level) { + /* failed overlay level criteria */ + continue; + } + if (!(trans_type==DONT_CARE + || (trans_type==GLX_TRANSPARENT_INDEX_EXT + && ov->transparent_type>0) + || (trans_type==GLX_NONE_EXT && ov->transparent_type==0))) { + /* failed transparent pixel type criteria */ + continue; + } + if (trans_value!=DONT_CARE && trans_value!=ov->value) { + /* failed transparent pixel value criteria */ + continue; + } + + /* get XVisualInfo and check the depth */ + vistemplate.visualid = ov->overlay_visual; + vistemplate.screen = scr; + vislist = XGetVisualInfo( dpy, VisualIDMask | VisualScreenMask, + &vistemplate, &count ); + + if (count!=1) { + /* something went wrong */ + continue; + } + if (preferred_class!=DONT_CARE && preferred_class!=vislist->CLASS) { + /* wrong visual class */ + continue; + } + + /* if RGB was requested, make sure we have True/DirectColor */ + if (rgbFlag && vislist->CLASS != TrueColor + && vislist->CLASS != DirectColor) + continue; + + /* if CI was requested, make sure we have a color indexed visual */ + if (!rgbFlag + && (vislist->CLASS == TrueColor || vislist->CLASS == DirectColor)) + continue; + + if (deepvis==NULL || vislist->depth > deepest) { + /* YES! found a satisfactory visual */ + if (deepvis) { + XFree( deepvis ); + } + deepest = vislist->depth; + deepvis = vislist; + /* DEBUG tt = ov->transparent_type;*/ + /* DEBUG tv = ov->value; */ + } + } + +/*DEBUG + if (deepvis) { + printf("chose 0x%x: layer=%d depth=%d trans_type=%d trans_value=%d\n", + deepvis->visualid, level, deepvis->depth, tt, tv ); + } +*/ + return deepvis; +} + + +/**********************************************************************/ +/*** Display-related functions ***/ +/**********************************************************************/ + + +/** + * Free all XMesaVisuals which are associated with the given display. + */ +static void +destroy_visuals_on_display(Display *dpy) +{ + int i; + for (i = 0; i < NumVisuals; i++) { + if (VisualTable[i]->display == dpy) { + /* remove this visual */ + int j; + free(VisualTable[i]); + for (j = i; j < NumVisuals - 1; j++) + VisualTable[j] = VisualTable[j + 1]; + NumVisuals--; + } + } +} + + +/** + * Called from XCloseDisplay() to let us free our display-related data. + */ +static int +close_display_callback(Display *dpy, XExtCodes *codes) +{ + destroy_visuals_on_display(dpy); + xmesa_destroy_buffers_on_display(dpy); + return 0; +} + + +/** + * Look for the named extension on given display and return a pointer + * to the _XExtension data, or NULL if extension not found. + */ +static _XExtension * +lookup_extension(Display *dpy, const char *extName) +{ + _XExtension *ext; + for (ext = dpy->ext_procs; ext; ext = ext->next) { + if (ext->name && strcmp(ext->name, extName) == 0) { + return ext; + } + } + return NULL; +} + + +/** + * Whenever we're given a new Display pointer, call this function to + * register our close_display_callback function. + */ +static void +register_with_display(Display *dpy) +{ + const char *extName = "MesaGLX"; + _XExtension *ext; + + ext = lookup_extension(dpy, extName); + if (!ext) { + XExtCodes *c = XAddExtension(dpy); + ext = dpy->ext_procs; /* new extension is at head of list */ + assert(c->extension == ext->codes.extension); + ext->name = _mesa_strdup(extName); + ext->close_display = close_display_callback; + } +} + + +/**********************************************************************/ +/*** Begin Fake GLX API Functions ***/ +/**********************************************************************/ + + +/** + * Helper used by glXChooseVisual and glXChooseFBConfig. + * The fbConfig parameter must be GL_FALSE for the former and GL_TRUE for + * the later. + * In either case, the attribute list is terminated with the value 'None'. + */ +static XMesaVisual +choose_visual( Display *dpy, int screen, const int *list, GLboolean fbConfig ) +{ + const GLboolean rgbModeDefault = fbConfig; + const int *parselist; + XVisualInfo *vis; + int min_ci = 0; + int min_red=0, min_green=0, min_blue=0; + GLboolean rgb_flag = rgbModeDefault; + GLboolean alpha_flag = GL_FALSE; + GLboolean double_flag = GL_FALSE; + GLboolean stereo_flag = GL_FALSE; + GLint depth_size = 0; + GLint stencil_size = 0; + GLint accumRedSize = 0; + GLint accumGreenSize = 0; + GLint accumBlueSize = 0; + GLint accumAlphaSize = 0; + int level = 0; + int visual_type = DONT_CARE; + int trans_type = DONT_CARE; + int trans_value = DONT_CARE; + GLint caveat = DONT_CARE; + XMesaVisual xmvis = NULL; + int desiredVisualID = -1; + int numAux = 0; + + parselist = list; + + while (*parselist) { + + switch (*parselist) { + case GLX_USE_GL: + if (fbConfig) { + /* invalid token */ + return NULL; + } + else { + /* skip */ + parselist++; + } + break; + case GLX_BUFFER_SIZE: + parselist++; + min_ci = *parselist++; + break; + case GLX_LEVEL: + parselist++; + level = *parselist++; + break; + case GLX_RGBA: + if (fbConfig) { + /* invalid token */ + return NULL; + } + else { + rgb_flag = GL_TRUE; + parselist++; + } + break; + case GLX_DOUBLEBUFFER: + parselist++; + if (fbConfig) { + double_flag = *parselist++; + } + else { + double_flag = GL_TRUE; + } + break; + case GLX_STEREO: + parselist++; + if (fbConfig) { + stereo_flag = *parselist++; + } + else { + stereo_flag = GL_TRUE; + } + break; + case GLX_AUX_BUFFERS: + parselist++; + numAux = *parselist++; + if (numAux > MAX_AUX_BUFFERS) + return NULL; + break; + case GLX_RED_SIZE: + parselist++; + min_red = *parselist++; + break; + case GLX_GREEN_SIZE: + parselist++; + min_green = *parselist++; + break; + case GLX_BLUE_SIZE: + parselist++; + min_blue = *parselist++; + break; + case GLX_ALPHA_SIZE: + parselist++; + { + GLint size = *parselist++; + alpha_flag = size ? GL_TRUE : GL_FALSE; + } + break; + case GLX_DEPTH_SIZE: + parselist++; + depth_size = *parselist++; + break; + case GLX_STENCIL_SIZE: + parselist++; + stencil_size = *parselist++; + break; + case GLX_ACCUM_RED_SIZE: + parselist++; + { + GLint size = *parselist++; + accumRedSize = MAX2( accumRedSize, size ); + } + break; + case GLX_ACCUM_GREEN_SIZE: + parselist++; + { + GLint size = *parselist++; + accumGreenSize = MAX2( accumGreenSize, size ); + } + break; + case GLX_ACCUM_BLUE_SIZE: + parselist++; + { + GLint size = *parselist++; + accumBlueSize = MAX2( accumBlueSize, size ); + } + break; + case GLX_ACCUM_ALPHA_SIZE: + parselist++; + { + GLint size = *parselist++; + accumAlphaSize = MAX2( accumAlphaSize, size ); + } + break; + + /* + * GLX_EXT_visual_info extension + */ + case GLX_X_VISUAL_TYPE_EXT: + parselist++; + visual_type = *parselist++; + break; + case GLX_TRANSPARENT_TYPE_EXT: + parselist++; + trans_type = *parselist++; + break; + case GLX_TRANSPARENT_INDEX_VALUE_EXT: + parselist++; + trans_value = *parselist++; + break; + case GLX_TRANSPARENT_RED_VALUE_EXT: + case GLX_TRANSPARENT_GREEN_VALUE_EXT: + case GLX_TRANSPARENT_BLUE_VALUE_EXT: + case GLX_TRANSPARENT_ALPHA_VALUE_EXT: + /* ignore */ + parselist++; + parselist++; + break; + + /* + * GLX_EXT_visual_info extension + */ + case GLX_VISUAL_CAVEAT_EXT: + parselist++; + caveat = *parselist++; /* ignored for now */ + break; + + /* + * GLX_ARB_multisample + */ + case GLX_SAMPLE_BUFFERS_ARB: + /* ms not supported */ + return NULL; + case GLX_SAMPLES_ARB: + /* ms not supported */ + return NULL; + + /* + * FBConfig attribs. + */ + case GLX_RENDER_TYPE: + if (!fbConfig) + return NULL; + parselist++; + if (*parselist == GLX_RGBA_BIT) { + rgb_flag = GL_TRUE; + } + else if (*parselist == GLX_COLOR_INDEX_BIT) { + rgb_flag = GL_FALSE; + } + else if (*parselist == 0) { + rgb_flag = GL_TRUE; + } + parselist++; + break; + case GLX_DRAWABLE_TYPE: + if (!fbConfig) + return NULL; + parselist++; + if (*parselist & ~(GLX_WINDOW_BIT | GLX_PIXMAP_BIT | GLX_PBUFFER_BIT)) { + return NULL; /* bad bit */ + } + parselist++; + break; + case GLX_FBCONFIG_ID: + if (!fbConfig) + return NULL; + parselist++; + desiredVisualID = *parselist++; + break; + case GLX_X_RENDERABLE: + if (!fbConfig) + return NULL; + parselist += 2; + /* ignore */ + break; + +#ifdef GLX_EXT_texture_from_pixmap + case GLX_BIND_TO_TEXTURE_RGB_EXT: + parselist++; /*skip*/ + break; + case GLX_BIND_TO_TEXTURE_RGBA_EXT: + parselist++; /*skip*/ + break; + case GLX_BIND_TO_MIPMAP_TEXTURE_EXT: + parselist++; /*skip*/ + break; + case GLX_BIND_TO_TEXTURE_TARGETS_EXT: + parselist++; + if (*parselist & ~(GLX_TEXTURE_1D_BIT_EXT | + GLX_TEXTURE_2D_BIT_EXT | + GLX_TEXTURE_RECTANGLE_BIT_EXT)) { + /* invalid bit */ + return NULL; + } + break; + case GLX_Y_INVERTED_EXT: + parselist++; /*skip*/ + break; +#endif + + case None: + /* end of list */ + break; + + default: + /* undefined attribute */ + _mesa_warning(NULL, "unexpected attrib 0x%x in choose_visual()", + *parselist); + return NULL; + } + } + + (void) caveat; + + /* + * Since we're only simulating the GLX extension this function will never + * find any real GL visuals. Instead, all we can do is try to find an RGB + * or CI visual of appropriate depth. Other requested attributes such as + * double buffering, depth buffer, etc. will be associated with the X + * visual and stored in the VisualTable[]. + */ + if (desiredVisualID != -1) { + /* try to get a specific visual, by visualID */ + XVisualInfo temp; + int n; + temp.visualid = desiredVisualID; + temp.screen = screen; + vis = XGetVisualInfo(dpy, VisualIDMask | VisualScreenMask, &temp, &n); + if (vis) { + /* give the visual some useful GLX attributes */ + double_flag = GL_TRUE; + if (vis->depth > 8) + rgb_flag = GL_TRUE; + depth_size = 24; /*default_depth_bits();*/ + stencil_size = STENCIL_BITS; + /* XXX accum??? */ + } + } + else if (level==0) { + /* normal color planes */ + if (rgb_flag) { + /* Get an RGB visual */ + int min_rgb = min_red + min_green + min_blue; + if (min_rgb>1 && min_rgb<8) { + /* a special case to be sure we can get a monochrome visual */ + min_rgb = 1; + } + vis = choose_x_visual( dpy, screen, rgb_flag, min_rgb, visual_type ); + } + else { + /* Get a color index visual */ + vis = choose_x_visual( dpy, screen, rgb_flag, min_ci, visual_type ); + accumRedSize = accumGreenSize = accumBlueSize = accumAlphaSize = 0; + } + } + else { + /* over/underlay planes */ + if (rgb_flag) { + /* rgba overlay */ + int min_rgb = min_red + min_green + min_blue; + if (min_rgb>1 && min_rgb<8) { + /* a special case to be sure we can get a monochrome visual */ + min_rgb = 1; + } + vis = choose_x_overlay_visual( dpy, screen, rgb_flag, level, + trans_type, trans_value, min_rgb, visual_type ); + } + else { + /* color index overlay */ + vis = choose_x_overlay_visual( dpy, screen, rgb_flag, level, + trans_type, trans_value, min_ci, visual_type ); + } + } + + if (vis) { + /* Note: we're not exactly obeying the glXChooseVisual rules here. + * When GLX_DEPTH_SIZE = 1 is specified we're supposed to choose the + * largest depth buffer size, which is 32bits/value. Instead, we + * return 16 to maintain performance with earlier versions of Mesa. + */ + if (stencil_size > 0) + depth_size = 24; /* if Z and stencil, always use 24+8 format */ + else if (depth_size > 24) + depth_size = 32; + else if (depth_size > 16) + depth_size = 24; + else if (depth_size > 0) { + depth_size = default_depth_bits(); + } + + if (!alpha_flag) { + alpha_flag = default_alpha_bits() > 0; + } + + /* we only support one size of stencil and accum buffers. */ + if (stencil_size > 0) + stencil_size = STENCIL_BITS; + if (accumRedSize > 0 || accumGreenSize > 0 || accumBlueSize > 0 || + accumAlphaSize > 0) { + accumRedSize = + accumGreenSize = + accumBlueSize = default_accum_bits(); + accumAlphaSize = alpha_flag ? accumRedSize : 0; + } + + xmvis = save_glx_visual( dpy, vis, rgb_flag, alpha_flag, double_flag, + stereo_flag, depth_size, stencil_size, + accumRedSize, accumGreenSize, + accumBlueSize, accumAlphaSize, level, numAux ); + } + + return xmvis; +} + + +static XVisualInfo * +Fake_glXChooseVisual( Display *dpy, int screen, int *list ) +{ + XMesaVisual xmvis; + + /* register ourselves as an extension on this display */ + register_with_display(dpy); + + xmvis = choose_visual(dpy, screen, list, GL_FALSE); + if (xmvis) { +#if 0 + return xmvis->vishandle; +#else + /* create a new vishandle - the cached one may be stale */ + xmvis->vishandle = (XVisualInfo *) _mesa_malloc(sizeof(XVisualInfo)); + if (xmvis->vishandle) { + _mesa_memcpy(xmvis->vishandle, xmvis->visinfo, sizeof(XVisualInfo)); + } + return xmvis->vishandle; +#endif + } + else + return NULL; +} + + +static GLXContext +Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo, + GLXContext share_list, Bool direct ) +{ + XMesaVisual xmvis; + struct fake_glx_context *glxCtx; + struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list; + + if (!dpy || !visinfo) + return 0; + + glxCtx = CALLOC_STRUCT(fake_glx_context); + if (!glxCtx) + return 0; + + /* deallocate unused windows/buffers */ +#if 0 + XMesaGarbageCollect(); +#endif + + xmvis = find_glx_visual( dpy, visinfo ); + if (!xmvis) { + /* This visual wasn't found with glXChooseVisual() */ + xmvis = create_glx_visual( dpy, visinfo ); + if (!xmvis) { + /* unusable visual */ + _mesa_free(glxCtx); + return NULL; + } + } + + glxCtx->xmesaContext = XMesaCreateContext(xmvis, + shareCtx ? shareCtx->xmesaContext : NULL); + if (!glxCtx->xmesaContext) { + _mesa_free(glxCtx); + return NULL; + } + + glxCtx->glxContext.isDirect = GL_FALSE; + glxCtx->glxContext.currentDpy = dpy; + glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ + + assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); + + return (GLXContext) glxCtx; +} + + +/* XXX these may have to be removed due to thread-safety issues. */ +static GLXContext MakeCurrent_PrevContext = 0; +static GLXDrawable MakeCurrent_PrevDrawable = 0; +static GLXDrawable MakeCurrent_PrevReadable = 0; +static XMesaBuffer MakeCurrent_PrevDrawBuffer = 0; +static XMesaBuffer MakeCurrent_PrevReadBuffer = 0; + + +/* GLX 1.3 and later */ +static Bool +Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, + GLXDrawable read, GLXContext ctx ) +{ + struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + + if (ctx && draw && read) { + XMesaBuffer drawBuffer, readBuffer; + XMesaContext xmctx = glxCtx->xmesaContext; + + /* Find the XMesaBuffer which corresponds to the GLXDrawable 'draw' */ + if (ctx == MakeCurrent_PrevContext + && draw == MakeCurrent_PrevDrawable) { + drawBuffer = MakeCurrent_PrevDrawBuffer; + } + else { + drawBuffer = XMesaFindBuffer( dpy, draw ); + } + if (!drawBuffer) { + /* drawable must be a new window! */ + drawBuffer = XMesaCreateWindowBuffer( xmctx->xm_visual, draw ); + if (!drawBuffer) { + /* Out of memory, or context/drawable depth mismatch */ + return False; + } +#ifdef FX + FXcreateContext( xmctx->xm_visual, draw, xmctx, drawBuffer ); +#endif + } + + /* Find the XMesaBuffer which corresponds to the GLXDrawable 'read' */ + if (ctx == MakeCurrent_PrevContext + && read == MakeCurrent_PrevReadable) { + readBuffer = MakeCurrent_PrevReadBuffer; + } + else { + readBuffer = XMesaFindBuffer( dpy, read ); + } + if (!readBuffer) { + /* drawable must be a new window! */ + readBuffer = XMesaCreateWindowBuffer( xmctx->xm_visual, read ); + if (!readBuffer) { + /* Out of memory, or context/drawable depth mismatch */ + return False; + } +#ifdef FX + FXcreateContext( xmctx->xm_visual, read, xmctx, readBuffer ); +#endif + } + + MakeCurrent_PrevContext = ctx; + MakeCurrent_PrevDrawable = draw; + MakeCurrent_PrevReadable = read; + MakeCurrent_PrevDrawBuffer = drawBuffer; + MakeCurrent_PrevReadBuffer = readBuffer; + + /* Now make current! */ + if (XMesaMakeCurrent2(xmctx, drawBuffer, readBuffer)) { + ((__GLXcontext *) ctx)->currentDpy = dpy; + ((__GLXcontext *) ctx)->currentDrawable = draw; + ((__GLXcontext *) ctx)->currentReadable = read; + return True; + } + else { + return False; + } + } + else if (!ctx && !draw && !read) { + /* release current context w/out assigning new one. */ + XMesaMakeCurrent( NULL, NULL ); + MakeCurrent_PrevContext = 0; + MakeCurrent_PrevDrawable = 0; + MakeCurrent_PrevReadable = 0; + MakeCurrent_PrevDrawBuffer = 0; + MakeCurrent_PrevReadBuffer = 0; + return True; + } + else { + /* The args must either all be non-zero or all zero. + * This is an error. + */ + return False; + } +} + + +static Bool +Fake_glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx ) +{ + return Fake_glXMakeContextCurrent( dpy, drawable, drawable, ctx ); +} + + +static GLXPixmap +Fake_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ) +{ + XMesaVisual v; + XMesaBuffer b; + + v = find_glx_visual( dpy, visinfo ); + if (!v) { + v = create_glx_visual( dpy, visinfo ); + if (!v) { + /* unusable visual */ + return 0; + } + } + + b = XMesaCreatePixmapBuffer( v, pixmap, 0 ); + if (!b) { + return 0; + } + return b->drawable; +} + + +/*** GLX_MESA_pixmap_colormap ***/ + +static GLXPixmap +Fake_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, + Pixmap pixmap, Colormap cmap ) +{ + XMesaVisual v; + XMesaBuffer b; + + v = find_glx_visual( dpy, visinfo ); + if (!v) { + v = create_glx_visual( dpy, visinfo ); + if (!v) { + /* unusable visual */ + return 0; + } + } + + b = XMesaCreatePixmapBuffer( v, pixmap, cmap ); + if (!b) { + return 0; + } + return b->drawable; +} + + +static void +Fake_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, pixmap); + if (b) { + XMesaDestroyBuffer(b); + } + else if (_mesa_getenv("MESA_DEBUG")) { + _mesa_warning(NULL, "Mesa: glXDestroyGLXPixmap: invalid pixmap\n"); + } +} + + +static void +Fake_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, + unsigned long mask ) +{ + struct fake_glx_context *fakeSrc = (struct fake_glx_context *) src; + struct fake_glx_context *fakeDst = (struct fake_glx_context *) dst; + XMesaContext xm_src = fakeSrc->xmesaContext; + XMesaContext xm_dst = fakeDst->xmesaContext; + (void) dpy; + if (MakeCurrent_PrevContext == src) { + _mesa_Flush(); + } + st_copy_context_state( xm_src->st, xm_dst->st, (GLuint) mask ); +} + + +static Bool +Fake_glXQueryExtension( Display *dpy, int *errorb, int *event ) +{ + /* Mesa's GLX isn't really an X extension but we try to act like one. */ + (void) dpy; + (void) errorb; + (void) event; + return True; +} + + +extern void _kw_ungrab_all( Display *dpy ); +void _kw_ungrab_all( Display *dpy ) +{ + XUngrabPointer( dpy, CurrentTime ); + XUngrabKeyboard( dpy, CurrentTime ); +} + + +static void +Fake_glXDestroyContext( Display *dpy, GLXContext ctx ) +{ + struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + (void) dpy; + MakeCurrent_PrevContext = 0; + MakeCurrent_PrevDrawable = 0; + MakeCurrent_PrevReadable = 0; + MakeCurrent_PrevDrawBuffer = 0; + MakeCurrent_PrevReadBuffer = 0; + XMesaDestroyContext( glxCtx->xmesaContext ); + XMesaGarbageCollect(); + _mesa_free(glxCtx); +} + + +static Bool +Fake_glXIsDirect( Display *dpy, GLXContext ctx ) +{ + (void) dpy; + (void) ctx; + return False; +} + + + +static void +Fake_glXSwapBuffers( Display *dpy, GLXDrawable drawable ) +{ + XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable ); + + if (buffer) { + XMesaSwapBuffers(buffer); + } + else if (_mesa_getenv("MESA_DEBUG")) { + _mesa_warning(NULL, "glXSwapBuffers: invalid drawable 0x%x\n", + (int) drawable); + } +} + + + +/*** GLX_MESA_copy_sub_buffer ***/ + +static void +Fake_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable, + int x, int y, int width, int height ) +{ + XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable ); + if (buffer) { + XMesaCopySubBuffer(buffer, x, y, width, height); + } + else if (_mesa_getenv("MESA_DEBUG")) { + _mesa_warning(NULL, "Mesa: glXCopySubBufferMESA: invalid drawable\n"); + } +} + + +static Bool +Fake_glXQueryVersion( Display *dpy, int *maj, int *min ) +{ + (void) dpy; + /* Return GLX version, not Mesa version */ + assert(CLIENT_MAJOR_VERSION == SERVER_MAJOR_VERSION); + *maj = CLIENT_MAJOR_VERSION; + *min = MIN2( CLIENT_MINOR_VERSION, SERVER_MINOR_VERSION ); + return True; +} + + +/* + * Query the GLX attributes of the given XVisualInfo. + */ +static int +get_config( XMesaVisual xmvis, int attrib, int *value, GLboolean fbconfig ) +{ + ASSERT(xmvis); + switch(attrib) { + case GLX_USE_GL: + if (fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = (int) True; + return 0; + case GLX_BUFFER_SIZE: + *value = xmvis->visinfo->depth; + return 0; + case GLX_LEVEL: + *value = xmvis->mesa_visual.level; + return 0; + case GLX_RGBA: + if (fbconfig) + return GLX_BAD_ATTRIBUTE; + if (xmvis->mesa_visual.rgbMode) { + *value = True; + } + else { + *value = False; + } + return 0; + case GLX_DOUBLEBUFFER: + *value = (int) xmvis->mesa_visual.doubleBufferMode; + return 0; + case GLX_STEREO: + *value = (int) xmvis->mesa_visual.stereoMode; + return 0; + case GLX_AUX_BUFFERS: + *value = xmvis->mesa_visual.numAuxBuffers; + return 0; + case GLX_RED_SIZE: + *value = xmvis->mesa_visual.redBits; + return 0; + case GLX_GREEN_SIZE: + *value = xmvis->mesa_visual.greenBits; + return 0; + case GLX_BLUE_SIZE: + *value = xmvis->mesa_visual.blueBits; + return 0; + case GLX_ALPHA_SIZE: + *value = xmvis->mesa_visual.alphaBits; + return 0; + case GLX_DEPTH_SIZE: + *value = xmvis->mesa_visual.depthBits; + return 0; + case GLX_STENCIL_SIZE: + *value = xmvis->mesa_visual.stencilBits; + return 0; + case GLX_ACCUM_RED_SIZE: + *value = xmvis->mesa_visual.accumRedBits; + return 0; + case GLX_ACCUM_GREEN_SIZE: + *value = xmvis->mesa_visual.accumGreenBits; + return 0; + case GLX_ACCUM_BLUE_SIZE: + *value = xmvis->mesa_visual.accumBlueBits; + return 0; + case GLX_ACCUM_ALPHA_SIZE: + *value = xmvis->mesa_visual.accumAlphaBits; + return 0; + + /* + * GLX_EXT_visual_info extension + */ + case GLX_X_VISUAL_TYPE_EXT: + switch (xmvis->visinfo->CLASS) { + case StaticGray: *value = GLX_STATIC_GRAY_EXT; return 0; + case GrayScale: *value = GLX_GRAY_SCALE_EXT; return 0; + case StaticColor: *value = GLX_STATIC_GRAY_EXT; return 0; + case PseudoColor: *value = GLX_PSEUDO_COLOR_EXT; return 0; + case TrueColor: *value = GLX_TRUE_COLOR_EXT; return 0; + case DirectColor: *value = GLX_DIRECT_COLOR_EXT; return 0; + } + return 0; + case GLX_TRANSPARENT_TYPE_EXT: + if (xmvis->mesa_visual.level==0) { + /* normal planes */ + *value = GLX_NONE_EXT; + } + else if (xmvis->mesa_visual.level>0) { + /* overlay */ + if (xmvis->mesa_visual.rgbMode) { + *value = GLX_TRANSPARENT_RGB_EXT; + } + else { + *value = GLX_TRANSPARENT_INDEX_EXT; + } + } + else if (xmvis->mesa_visual.level<0) { + /* underlay */ + *value = GLX_NONE_EXT; + } + return 0; + case GLX_TRANSPARENT_INDEX_VALUE_EXT: + { + int pixel = transparent_pixel( xmvis ); + if (pixel>=0) { + *value = pixel; + } + /* else undefined */ + } + return 0; + case GLX_TRANSPARENT_RED_VALUE_EXT: + /* undefined */ + return 0; + case GLX_TRANSPARENT_GREEN_VALUE_EXT: + /* undefined */ + return 0; + case GLX_TRANSPARENT_BLUE_VALUE_EXT: + /* undefined */ + return 0; + case GLX_TRANSPARENT_ALPHA_VALUE_EXT: + /* undefined */ + return 0; + + /* + * GLX_EXT_visual_info extension + */ + case GLX_VISUAL_CAVEAT_EXT: + /* test for zero, just in case */ + if (xmvis->mesa_visual.visualRating > 0) + *value = xmvis->mesa_visual.visualRating; + else + *value = GLX_NONE_EXT; + return 0; + + /* + * GLX_ARB_multisample + */ + case GLX_SAMPLE_BUFFERS_ARB: + *value = 0; + return 0; + case GLX_SAMPLES_ARB: + *value = 0; + return 0; + + /* + * For FBConfigs: + */ + case GLX_SCREEN_EXT: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = xmvis->visinfo->screen; + break; + case GLX_DRAWABLE_TYPE: /*SGIX too */ + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = GLX_WINDOW_BIT | GLX_PIXMAP_BIT | GLX_PBUFFER_BIT; + break; + case GLX_RENDER_TYPE_SGIX: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + if (xmvis->mesa_visual.rgbMode) + *value = GLX_RGBA_BIT; + else + *value = GLX_COLOR_INDEX_BIT; + break; + case GLX_X_RENDERABLE_SGIX: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = True; /* XXX really? */ + break; + case GLX_FBCONFIG_ID_SGIX: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = xmvis->visinfo->visualid; + break; + case GLX_MAX_PBUFFER_WIDTH: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + /* XXX or MAX_WIDTH? */ + *value = DisplayWidth(xmvis->display, xmvis->visinfo->screen); + break; + case GLX_MAX_PBUFFER_HEIGHT: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = DisplayHeight(xmvis->display, xmvis->visinfo->screen); + break; + case GLX_MAX_PBUFFER_PIXELS: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = DisplayWidth(xmvis->display, xmvis->visinfo->screen) * + DisplayHeight(xmvis->display, xmvis->visinfo->screen); + break; + case GLX_VISUAL_ID: + if (!fbconfig) + return GLX_BAD_ATTRIBUTE; + *value = xmvis->visinfo->visualid; + break; + +#ifdef GLX_EXT_texture_from_pixmap + case GLX_BIND_TO_TEXTURE_RGB_EXT: + *value = True; /*XXX*/ + break; + case GLX_BIND_TO_TEXTURE_RGBA_EXT: + /* XXX review */ + *value = xmvis->mesa_visual.alphaBits > 0 ? True : False; + break; + case GLX_BIND_TO_MIPMAP_TEXTURE_EXT: + *value = True; /*XXX*/ + break; + case GLX_BIND_TO_TEXTURE_TARGETS_EXT: + *value = (GLX_TEXTURE_1D_BIT_EXT | + GLX_TEXTURE_2D_BIT_EXT | + GLX_TEXTURE_RECTANGLE_BIT_EXT); /*XXX*/ + break; + case GLX_Y_INVERTED_EXT: + *value = True; /*XXX*/ + break; +#endif + + default: + return GLX_BAD_ATTRIBUTE; + } + return Success; +} + + +static int +Fake_glXGetConfig( Display *dpy, XVisualInfo *visinfo, + int attrib, int *value ) +{ + XMesaVisual xmvis; + int k; + if (!dpy || !visinfo) + return GLX_BAD_ATTRIBUTE; + + xmvis = find_glx_visual( dpy, visinfo ); + if (!xmvis) { + /* this visual wasn't obtained with glXChooseVisual */ + xmvis = create_glx_visual( dpy, visinfo ); + if (!xmvis) { + /* this visual can't be used for GL rendering */ + if (attrib==GLX_USE_GL) { + *value = (int) False; + return 0; + } + else { + return GLX_BAD_VISUAL; + } + } + } + + k = get_config(xmvis, attrib, value, GL_FALSE); + return k; +} + + +static void +Fake_glXWaitGL( void ) +{ + XMesaContext xmesa = XMesaGetCurrentContext(); + XMesaFlush( xmesa ); +} + + + +static void +Fake_glXWaitX( void ) +{ + XMesaContext xmesa = XMesaGetCurrentContext(); + XMesaFlush( xmesa ); +} + + +static const char * +get_extensions( void ) +{ +#ifdef FX + const char *fx = _mesa_getenv("MESA_GLX_FX"); + if (fx && fx[0] != 'd') { + return EXTENSIONS; + } +#endif + return EXTENSIONS + 23; /* skip "GLX_MESA_set_3dfx_mode" */ +} + + + +/* GLX 1.1 and later */ +static const char * +Fake_glXQueryExtensionsString( Display *dpy, int screen ) +{ + (void) dpy; + (void) screen; + return get_extensions(); +} + + + +/* GLX 1.1 and later */ +static const char * +Fake_glXQueryServerString( Display *dpy, int screen, int name ) +{ + static char version[1000]; + _mesa_sprintf(version, "%d.%d %s", + SERVER_MAJOR_VERSION, SERVER_MINOR_VERSION, MESA_GLX_VERSION); + + (void) dpy; + (void) screen; + + switch (name) { + case GLX_EXTENSIONS: + return get_extensions(); + case GLX_VENDOR: + return VENDOR; + case GLX_VERSION: + return version; + default: + return NULL; + } +} + + + +/* GLX 1.1 and later */ +static const char * +Fake_glXGetClientString( Display *dpy, int name ) +{ + static char version[1000]; + _mesa_sprintf(version, "%d.%d %s", CLIENT_MAJOR_VERSION, + CLIENT_MINOR_VERSION, MESA_GLX_VERSION); + + (void) dpy; + + switch (name) { + case GLX_EXTENSIONS: + return get_extensions(); + case GLX_VENDOR: + return VENDOR; + case GLX_VERSION: + return version; + default: + return NULL; + } +} + + + +/* + * GLX 1.3 and later + */ + + +static int +Fake_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config, + int attribute, int *value ) +{ + XMesaVisual v = (XMesaVisual) config; + (void) dpy; + (void) config; + + if (!dpy || !config || !value) + return -1; + + return get_config(v, attribute, value, GL_TRUE); +} + + +static GLXFBConfig * +Fake_glXGetFBConfigs( Display *dpy, int screen, int *nelements ) +{ + XVisualInfo *visuals, visTemplate; + const long visMask = VisualScreenMask; + int i; + + /* Get list of all X visuals */ + visTemplate.screen = screen; + visuals = XGetVisualInfo(dpy, visMask, &visTemplate, nelements); + if (*nelements > 0) { + XMesaVisual *results; + results = (XMesaVisual *) _mesa_malloc(*nelements * sizeof(XMesaVisual)); + if (!results) { + *nelements = 0; + return NULL; + } + for (i = 0; i < *nelements; i++) { + results[i] = create_glx_visual(dpy, visuals + i); + } + return (GLXFBConfig *) results; + } + return NULL; +} + + +static GLXFBConfig * +Fake_glXChooseFBConfig( Display *dpy, int screen, + const int *attribList, int *nitems ) +{ + XMesaVisual xmvis; + + if (!attribList || !attribList[0]) { + /* return list of all configs (per GLX_SGIX_fbconfig spec) */ + return Fake_glXGetFBConfigs(dpy, screen, nitems); + } + + xmvis = choose_visual(dpy, screen, attribList, GL_TRUE); + if (xmvis) { + GLXFBConfig *config = (GLXFBConfig *) _mesa_malloc(sizeof(XMesaVisual)); + if (!config) { + *nitems = 0; + return NULL; + } + *nitems = 1; + config[0] = (GLXFBConfig) xmvis; + return (GLXFBConfig *) config; + } + else { + *nitems = 0; + return NULL; + } +} + + +static XVisualInfo * +Fake_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config ) +{ + if (dpy && config) { + XMesaVisual xmvis = (XMesaVisual) config; +#if 0 + return xmvis->vishandle; +#else + /* create a new vishandle - the cached one may be stale */ + xmvis->vishandle = (XVisualInfo *) _mesa_malloc(sizeof(XVisualInfo)); + if (xmvis->vishandle) { + _mesa_memcpy(xmvis->vishandle, xmvis->visinfo, sizeof(XVisualInfo)); + } + return xmvis->vishandle; +#endif + } + else { + return NULL; + } +} + + +static GLXWindow +Fake_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win, + const int *attribList ) +{ + XMesaVisual xmvis = (XMesaVisual) config; + XMesaBuffer xmbuf; + if (!xmvis) + return 0; + + xmbuf = XMesaCreateWindowBuffer(xmvis, win); + if (!xmbuf) + return 0; + +#ifdef FX + /* XXX this will segfault if actually called */ + FXcreateContext(xmvis, win, NULL, xmbuf); +#endif + + (void) dpy; + (void) attribList; /* Ignored in GLX 1.3 */ + + return win; /* A hack for now */ +} + + +static void +Fake_glXDestroyWindow( Display *dpy, GLXWindow window ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, (XMesaDrawable) window); + if (b) + XMesaDestroyBuffer(b); + /* don't destroy X window */ +} + + +/* XXX untested */ +static GLXPixmap +Fake_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, + const int *attribList ) +{ + XMesaVisual v = (XMesaVisual) config; + XMesaBuffer b; + const int *attr; + int target = 0, format = 0, mipmap = 0; + int value; + + if (!dpy || !config || !pixmap) + return 0; + + for (attr = attribList; *attr; attr++) { + switch (*attr) { + case GLX_TEXTURE_FORMAT_EXT: + attr++; + switch (*attr) { + case GLX_TEXTURE_FORMAT_NONE_EXT: + case GLX_TEXTURE_FORMAT_RGB_EXT: + case GLX_TEXTURE_FORMAT_RGBA_EXT: + format = *attr; + break; + default: + /* error */ + return 0; + } + break; + case GLX_TEXTURE_TARGET_EXT: + attr++; + switch (*attr) { + case GLX_TEXTURE_1D_EXT: + case GLX_TEXTURE_2D_EXT: + case GLX_TEXTURE_RECTANGLE_EXT: + target = *attr; + break; + default: + /* error */ + return 0; + } + break; + case GLX_MIPMAP_TEXTURE_EXT: + attr++; + if (*attr) + mipmap = 1; + break; + default: + /* error */ + return 0; + } + } + + if (format == GLX_TEXTURE_FORMAT_RGB_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_RGB_EXT, + &value, GL_TRUE) != Success + || !value) { + return 0; /* error! */ + } + } + else if (format == GLX_TEXTURE_FORMAT_RGBA_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_RGBA_EXT, + &value, GL_TRUE) != Success + || !value) { + return 0; /* error! */ + } + } + if (mipmap) { + if (get_config(v, GLX_BIND_TO_MIPMAP_TEXTURE_EXT, + &value, GL_TRUE) != Success + || !value) { + return 0; /* error! */ + } + } + if (target == GLX_TEXTURE_1D_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT, + &value, GL_TRUE) != Success + || (value & GLX_TEXTURE_1D_BIT_EXT) == 0) { + return 0; /* error! */ + } + } + else if (target == GLX_TEXTURE_2D_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT, + &value, GL_TRUE) != Success + || (value & GLX_TEXTURE_2D_BIT_EXT) == 0) { + return 0; /* error! */ + } + } + if (target == GLX_TEXTURE_RECTANGLE_EXT) { + if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT, + &value, GL_TRUE) != Success + || (value & GLX_TEXTURE_RECTANGLE_BIT_EXT) == 0) { + return 0; /* error! */ + } + } + + if (format || target || mipmap) { + /* texture from pixmap */ + b = XMesaCreatePixmapTextureBuffer(v, pixmap, 0, format, target, mipmap); + } + else { + b = XMesaCreatePixmapBuffer( v, pixmap, 0 ); + } + if (!b) { + return 0; + } + + return pixmap; +} + + +static void +Fake_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, (XMesaDrawable)pixmap); + if (b) + XMesaDestroyBuffer(b); + /* don't destroy X pixmap */ +} + + +static GLXPbuffer +Fake_glXCreatePbuffer( Display *dpy, GLXFBConfig config, + const int *attribList ) +{ + XMesaVisual xmvis = (XMesaVisual) config; + XMesaBuffer xmbuf; + const int *attrib; + int width = 0, height = 0; + GLboolean useLargest = GL_FALSE, preserveContents = GL_FALSE; + + (void) dpy; + + for (attrib = attribList; *attrib; attrib++) { + switch (*attrib) { + case GLX_PBUFFER_WIDTH: + attrib++; + width = *attrib; + break; + case GLX_PBUFFER_HEIGHT: + attrib++; + height = *attrib; + break; + case GLX_PRESERVED_CONTENTS: + attrib++; + preserveContents = *attrib; /* ignored */ + break; + case GLX_LARGEST_PBUFFER: + attrib++; + useLargest = *attrib; /* ignored */ + break; + default: + return 0; + } + } + + /* not used at this time */ + (void) useLargest; + (void) preserveContents; + + if (width == 0 || height == 0) + return 0; + + xmbuf = XMesaCreatePBuffer( xmvis, 0, width, height); + /* A GLXPbuffer handle must be an X Drawable because that's what + * glXMakeCurrent takes. + */ + if (xmbuf) + return (GLXPbuffer) xmbuf->drawable; + else + return 0; +} + + +static void +Fake_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, pbuf); + if (b) { + XMesaDestroyBuffer(b); + } +} + + +static void +Fake_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, + unsigned int *value ) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, draw); + if (!xmbuf) + return; + + switch (attribute) { + case GLX_WIDTH: + *value = xmesa_buffer_width(xmbuf); + break; + case GLX_HEIGHT: + *value = xmesa_buffer_width(xmbuf); + break; + case GLX_PRESERVED_CONTENTS: + *value = True; + break; + case GLX_LARGEST_PBUFFER: + *value = xmesa_buffer_width(xmbuf) * xmesa_buffer_height(xmbuf); + break; + case GLX_FBCONFIG_ID: + *value = xmbuf->xm_visual->visinfo->visualid; + return; +#ifdef GLX_EXT_texture_from_pixmap + case GLX_TEXTURE_FORMAT_EXT: + *value = xmbuf->TextureFormat; + break; + case GLX_TEXTURE_TARGET_EXT: + *value = xmbuf->TextureTarget; + break; + case GLX_MIPMAP_TEXTURE_EXT: + *value = xmbuf->TextureMipmap; + break; +#endif + + default: + return; /* raise BadValue error */ + } +} + + +static GLXContext +Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config, + int renderType, GLXContext shareList, Bool direct ) +{ + struct fake_glx_context *glxCtx; + struct fake_glx_context *shareCtx = (struct fake_glx_context *) shareList; + XMesaVisual xmvis = (XMesaVisual) config; + + if (!dpy || !config || + (renderType != GLX_RGBA_TYPE && renderType != GLX_COLOR_INDEX_TYPE)) + return 0; + + glxCtx = CALLOC_STRUCT(fake_glx_context); + if (!glxCtx) + return 0; + + /* deallocate unused windows/buffers */ + XMesaGarbageCollect(); + + glxCtx->xmesaContext = XMesaCreateContext(xmvis, + shareCtx ? shareCtx->xmesaContext : NULL); + if (!glxCtx->xmesaContext) { + _mesa_free(glxCtx); + return NULL; + } + + glxCtx->glxContext.isDirect = GL_FALSE; + glxCtx->glxContext.currentDpy = dpy; + glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ + + assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); + + return (GLXContext) glxCtx; +} + + +static int +Fake_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value ) +{ + struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + XMesaContext xmctx = glxCtx->xmesaContext; + + (void) dpy; + (void) ctx; + + switch (attribute) { + case GLX_FBCONFIG_ID: + *value = xmctx->xm_visual->visinfo->visualid; + break; + case GLX_RENDER_TYPE: + if (xmctx->xm_visual->mesa_visual.rgbMode) + *value = GLX_RGBA_BIT; + else + *value = GLX_COLOR_INDEX_BIT; + break; + case GLX_SCREEN: + *value = 0; + return Success; + default: + return GLX_BAD_ATTRIBUTE; + } + return 0; +} + + +static void +Fake_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask ) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); + if (xmbuf) + xmbuf->selectedEvents = mask; +} + + +static void +Fake_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable, + unsigned long *mask ) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); + if (xmbuf) + *mask = xmbuf->selectedEvents; + else + *mask = 0; +} + + + +/*** GLX_SGI_swap_control ***/ + +static int +Fake_glXSwapIntervalSGI(int interval) +{ + (void) interval; + return 0; +} + + + +/*** GLX_SGI_video_sync ***/ + +static unsigned int FrameCounter = 0; + +static int +Fake_glXGetVideoSyncSGI(unsigned int *count) +{ + /* this is a bogus implementation */ + *count = FrameCounter++; + return 0; +} + +static int +Fake_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) +{ + if (divisor <= 0 || remainder < 0) + return GLX_BAD_VALUE; + /* this is a bogus implementation */ + FrameCounter++; + while (FrameCounter % divisor != remainder) + FrameCounter++; + *count = FrameCounter; + return 0; +} + + + +/*** GLX_SGI_make_current_read ***/ + +static Bool +Fake_glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) +{ + return Fake_glXMakeContextCurrent( dpy, draw, read, ctx ); +} + +/* not used +static GLXDrawable +Fake_glXGetCurrentReadDrawableSGI(void) +{ + return 0; +} +*/ + + +/*** GLX_SGIX_video_source ***/ +#if defined(_VL_H) + +static GLXVideoSourceSGIX +Fake_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode) +{ + (void) dpy; + (void) screen; + (void) server; + (void) path; + (void) nodeClass; + (void) drainNode; + return 0; +} + +static void +Fake_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) +{ + (void) dpy; + (void) src; +} + +#endif + + +/*** GLX_EXT_import_context ***/ + +static void +Fake_glXFreeContextEXT(Display *dpy, GLXContext context) +{ + (void) dpy; + (void) context; +} + +static GLXContextID +Fake_glXGetContextIDEXT(const GLXContext context) +{ + (void) context; + return 0; +} + +static GLXContext +Fake_glXImportContextEXT(Display *dpy, GLXContextID contextID) +{ + (void) dpy; + (void) contextID; + return 0; +} + +static int +Fake_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value) +{ + (void) dpy; + (void) context; + (void) attribute; + (void) value; + return 0; +} + + + +/*** GLX_SGIX_fbconfig ***/ + +static int +Fake_glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value) +{ + return Fake_glXGetFBConfigAttrib(dpy, config, attribute, value); +} + +static GLXFBConfigSGIX * +Fake_glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements) +{ + return (GLXFBConfig *) Fake_glXChooseFBConfig(dpy, screen, attrib_list, nelements); +} + + +static GLXPixmap +Fake_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap) +{ + XMesaVisual xmvis = (XMesaVisual) config; + XMesaBuffer xmbuf = XMesaCreatePixmapBuffer(xmvis, pixmap, 0); + return xmbuf->drawable; /* need to return an X ID */ +} + + +static GLXContext +Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct) +{ + XMesaVisual xmvis = (XMesaVisual) config; + struct fake_glx_context *glxCtx; + struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list; + + glxCtx = CALLOC_STRUCT(fake_glx_context); + if (!glxCtx) + return 0; + + /* deallocate unused windows/buffers */ + XMesaGarbageCollect(); + + glxCtx->xmesaContext = XMesaCreateContext(xmvis, + shareCtx ? shareCtx->xmesaContext : NULL); + if (!glxCtx->xmesaContext) { + _mesa_free(glxCtx); + return NULL; + } + + glxCtx->glxContext.isDirect = GL_FALSE; + glxCtx->glxContext.currentDpy = dpy; + glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ + + assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); + + return (GLXContext) glxCtx; +} + + +static XVisualInfo * +Fake_glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config) +{ + return Fake_glXGetVisualFromFBConfig(dpy, config); +} + + +static GLXFBConfigSGIX +Fake_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) +{ + XMesaVisual xmvis = find_glx_visual(dpy, vis); + if (!xmvis) { + /* This visual wasn't found with glXChooseVisual() */ + xmvis = create_glx_visual(dpy, vis); + } + + return (GLXFBConfigSGIX) xmvis; +} + + + +/*** GLX_SGIX_pbuffer ***/ + +static GLXPbufferSGIX +Fake_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, + unsigned int width, unsigned int height, + int *attribList) +{ + XMesaVisual xmvis = (XMesaVisual) config; + XMesaBuffer xmbuf; + const int *attrib; + GLboolean useLargest = GL_FALSE, preserveContents = GL_FALSE; + + (void) dpy; + + for (attrib = attribList; attrib && *attrib; attrib++) { + switch (*attrib) { + case GLX_PRESERVED_CONTENTS_SGIX: + attrib++; + preserveContents = *attrib; /* ignored */ + break; + case GLX_LARGEST_PBUFFER_SGIX: + attrib++; + useLargest = *attrib; /* ignored */ + break; + default: + return 0; + } + } + + /* not used at this time */ + (void) useLargest; + (void) preserveContents; + + xmbuf = XMesaCreatePBuffer( xmvis, 0, width, height); + /* A GLXPbuffer handle must be an X Drawable because that's what + * glXMakeCurrent takes. + */ + return (GLXPbuffer) xmbuf->drawable; +} + + +static void +Fake_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf); + if (xmbuf) { + XMesaDestroyBuffer(xmbuf); + } +} + + +static int +Fake_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value) +{ + const XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf); + + if (!xmbuf) { + /* Generate GLXBadPbufferSGIX for bad pbuffer */ + return 0; + } + + switch (attribute) { + case GLX_PRESERVED_CONTENTS_SGIX: + *value = True; + break; + case GLX_LARGEST_PBUFFER_SGIX: + *value = xmesa_buffer_width(xmbuf) * xmesa_buffer_height(xmbuf); + break; + case GLX_WIDTH_SGIX: + *value = xmesa_buffer_width(xmbuf); + break; + case GLX_HEIGHT_SGIX: + *value = xmesa_buffer_height(xmbuf); + break; + case GLX_EVENT_MASK_SGIX: + *value = 0; /* XXX might be wrong */ + break; + default: + *value = 0; + } + return 0; +} + + +static void +Fake_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); + if (xmbuf) { + /* Note: we'll never generate clobber events */ + xmbuf->selectedEvents = mask; + } +} + + +static void +Fake_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask) +{ + XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); + if (xmbuf) { + *mask = xmbuf->selectedEvents; + } + else { + *mask = 0; + } +} + + + +/*** GLX_SGI_cushion ***/ + +static void +Fake_glXCushionSGI(Display *dpy, Window win, float cushion) +{ + (void) dpy; + (void) win; + (void) cushion; +} + + + +/*** GLX_SGIX_video_resize ***/ + +static int +Fake_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) window; + return 0; +} + +static int +Fake_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) x; + (void) y; + (void) w; + (void) h; + return 0; +} + +static int +Fake_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) x; + (void) y; + (void) w; + (void) h; + return 0; +} + +static int +Fake_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) dx; + (void) dy; + (void) dw; + (void) dh; + return 0; +} + +static int +Fake_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype) +{ + (void) dpy; + (void) screen; + (void) channel; + (void) synctype; + return 0; +} + + + +/*** GLX_SGIX_dmbuffer **/ + +#if defined(_DM_BUFFER_H_) +static Bool +Fake_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer) +{ + (void) dpy; + (void) pbuffer; + (void) params; + (void) dmbuffer; + return False; +} +#endif + + +/*** GLX_SGIX_swap_group ***/ + +static void +Fake_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member) +{ + (void) dpy; + (void) drawable; + (void) member; +} + + + +/*** GLX_SGIX_swap_barrier ***/ + +static void +Fake_glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier) +{ + (void) dpy; + (void) drawable; + (void) barrier; +} + +static Bool +Fake_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) +{ + (void) dpy; + (void) screen; + (void) max; + return False; +} + + + +/*** GLX_SUN_get_transparent_index ***/ + +static Status +Fake_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent) +{ + (void) dpy; + (void) overlay; + (void) underlay; + (void) pTransparent; + return 0; +} + + + +/*** GLX_MESA_release_buffers ***/ + +/* + * Release the depth, stencil, accum buffers attached to a GLXDrawable + * (a window or pixmap) prior to destroying the GLXDrawable. + */ +static Bool +Fake_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d ) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, d); + if (b) { + XMesaDestroyBuffer(b); + return True; + } + return False; +} + + + +/*** GLX_MESA_set_3dfx_mode ***/ + +static Bool +Fake_glXSet3DfxModeMESA( int mode ) +{ + return XMesaSetFXmode( mode ); +} + + + +/*** GLX_NV_vertex_array range ***/ +static void * +Fake_glXAllocateMemoryNV( GLsizei size, + GLfloat readFrequency, + GLfloat writeFrequency, + GLfloat priority ) +{ + (void) size; + (void) readFrequency; + (void) writeFrequency; + (void) priority; + return NULL; +} + + +static void +Fake_glXFreeMemoryNV( GLvoid *pointer ) +{ + (void) pointer; +} + + +/*** GLX_MESA_agp_offset ***/ + +static GLuint +Fake_glXGetAGPOffsetMESA( const GLvoid *pointer ) +{ + (void) pointer; + return ~0; +} + + +/*** GLX_EXT_texture_from_pixmap ***/ + +static void +Fake_glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, + const int *attrib_list) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, drawable); + if (b) + XMesaBindTexImage(dpy, b, buffer, attrib_list); +} + +static void +Fake_glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer) +{ + XMesaBuffer b = XMesaFindBuffer(dpy, drawable); + if (b) + XMesaReleaseTexImage(dpy, b, buffer); +} + + +/* silence warning */ +extern struct _glxapi_table *_mesa_GetGLXDispatchTable(void); + + +/** + * Create a new GLX API dispatch table with its function pointers + * initialized to point to Mesa's "fake" GLX API functions. + * Note: there's a similar function (_real_GetGLXDispatchTable) that + * returns a new dispatch table with all pointers initalized to point + * to "real" GLX functions (which understand GLX wire protocol, etc). + */ +struct _glxapi_table * +_mesa_GetGLXDispatchTable(void) +{ + static struct _glxapi_table glx; + + /* be sure our dispatch table size <= libGL's table */ + { + GLuint size = sizeof(struct _glxapi_table) / sizeof(void *); + (void) size; + assert(_glxapi_get_dispatch_table_size() >= size); + } + + /* initialize the whole table to no-ops */ + _glxapi_set_no_op_table(&glx); + + /* now initialize the table with the functions I implement */ + glx.ChooseVisual = Fake_glXChooseVisual; + glx.CopyContext = Fake_glXCopyContext; + glx.CreateContext = Fake_glXCreateContext; + glx.CreateGLXPixmap = Fake_glXCreateGLXPixmap; + glx.DestroyContext = Fake_glXDestroyContext; + glx.DestroyGLXPixmap = Fake_glXDestroyGLXPixmap; + glx.GetConfig = Fake_glXGetConfig; + /*glx.GetCurrentContext = Fake_glXGetCurrentContext;*/ + /*glx.GetCurrentDrawable = Fake_glXGetCurrentDrawable;*/ + glx.IsDirect = Fake_glXIsDirect; + glx.MakeCurrent = Fake_glXMakeCurrent; + glx.QueryExtension = Fake_glXQueryExtension; + glx.QueryVersion = Fake_glXQueryVersion; + glx.SwapBuffers = Fake_glXSwapBuffers; + glx.UseXFont = Fake_glXUseXFont; + glx.WaitGL = Fake_glXWaitGL; + glx.WaitX = Fake_glXWaitX; + + /*** GLX_VERSION_1_1 ***/ + glx.GetClientString = Fake_glXGetClientString; + glx.QueryExtensionsString = Fake_glXQueryExtensionsString; + glx.QueryServerString = Fake_glXQueryServerString; + + /*** GLX_VERSION_1_2 ***/ + /*glx.GetCurrentDisplay = Fake_glXGetCurrentDisplay;*/ + + /*** GLX_VERSION_1_3 ***/ + glx.ChooseFBConfig = Fake_glXChooseFBConfig; + glx.CreateNewContext = Fake_glXCreateNewContext; + glx.CreatePbuffer = Fake_glXCreatePbuffer; + glx.CreatePixmap = Fake_glXCreatePixmap; + glx.CreateWindow = Fake_glXCreateWindow; + glx.DestroyPbuffer = Fake_glXDestroyPbuffer; + glx.DestroyPixmap = Fake_glXDestroyPixmap; + glx.DestroyWindow = Fake_glXDestroyWindow; + /*glx.GetCurrentReadDrawable = Fake_glXGetCurrentReadDrawable;*/ + glx.GetFBConfigAttrib = Fake_glXGetFBConfigAttrib; + glx.GetFBConfigs = Fake_glXGetFBConfigs; + glx.GetSelectedEvent = Fake_glXGetSelectedEvent; + glx.GetVisualFromFBConfig = Fake_glXGetVisualFromFBConfig; + glx.MakeContextCurrent = Fake_glXMakeContextCurrent; + glx.QueryContext = Fake_glXQueryContext; + glx.QueryDrawable = Fake_glXQueryDrawable; + glx.SelectEvent = Fake_glXSelectEvent; + + /*** GLX_SGI_swap_control ***/ + glx.SwapIntervalSGI = Fake_glXSwapIntervalSGI; + + /*** GLX_SGI_video_sync ***/ + glx.GetVideoSyncSGI = Fake_glXGetVideoSyncSGI; + glx.WaitVideoSyncSGI = Fake_glXWaitVideoSyncSGI; + + /*** GLX_SGI_make_current_read ***/ + glx.MakeCurrentReadSGI = Fake_glXMakeCurrentReadSGI; + /*glx.GetCurrentReadDrawableSGI = Fake_glXGetCurrentReadDrawableSGI;*/ + +/*** GLX_SGIX_video_source ***/ +#if defined(_VL_H) + glx.CreateGLXVideoSourceSGIX = Fake_glXCreateGLXVideoSourceSGIX; + glx.DestroyGLXVideoSourceSGIX = Fake_glXDestroyGLXVideoSourceSGIX; +#endif + + /*** GLX_EXT_import_context ***/ + glx.FreeContextEXT = Fake_glXFreeContextEXT; + glx.GetContextIDEXT = Fake_glXGetContextIDEXT; + /*glx.GetCurrentDisplayEXT = Fake_glXGetCurrentDisplayEXT;*/ + glx.ImportContextEXT = Fake_glXImportContextEXT; + glx.QueryContextInfoEXT = Fake_glXQueryContextInfoEXT; + + /*** GLX_SGIX_fbconfig ***/ + glx.GetFBConfigAttribSGIX = Fake_glXGetFBConfigAttribSGIX; + glx.ChooseFBConfigSGIX = Fake_glXChooseFBConfigSGIX; + glx.CreateGLXPixmapWithConfigSGIX = Fake_glXCreateGLXPixmapWithConfigSGIX; + glx.CreateContextWithConfigSGIX = Fake_glXCreateContextWithConfigSGIX; + glx.GetVisualFromFBConfigSGIX = Fake_glXGetVisualFromFBConfigSGIX; + glx.GetFBConfigFromVisualSGIX = Fake_glXGetFBConfigFromVisualSGIX; + + /*** GLX_SGIX_pbuffer ***/ + glx.CreateGLXPbufferSGIX = Fake_glXCreateGLXPbufferSGIX; + glx.DestroyGLXPbufferSGIX = Fake_glXDestroyGLXPbufferSGIX; + glx.QueryGLXPbufferSGIX = Fake_glXQueryGLXPbufferSGIX; + glx.SelectEventSGIX = Fake_glXSelectEventSGIX; + glx.GetSelectedEventSGIX = Fake_glXGetSelectedEventSGIX; + + /*** GLX_SGI_cushion ***/ + glx.CushionSGI = Fake_glXCushionSGI; + + /*** GLX_SGIX_video_resize ***/ + glx.BindChannelToWindowSGIX = Fake_glXBindChannelToWindowSGIX; + glx.ChannelRectSGIX = Fake_glXChannelRectSGIX; + glx.QueryChannelRectSGIX = Fake_glXQueryChannelRectSGIX; + glx.QueryChannelDeltasSGIX = Fake_glXQueryChannelDeltasSGIX; + glx.ChannelRectSyncSGIX = Fake_glXChannelRectSyncSGIX; + + /*** GLX_SGIX_dmbuffer **/ +#if defined(_DM_BUFFER_H_) + glx.AssociateDMPbufferSGIX = NULL; +#endif + + /*** GLX_SGIX_swap_group ***/ + glx.JoinSwapGroupSGIX = Fake_glXJoinSwapGroupSGIX; + + /*** GLX_SGIX_swap_barrier ***/ + glx.BindSwapBarrierSGIX = Fake_glXBindSwapBarrierSGIX; + glx.QueryMaxSwapBarriersSGIX = Fake_glXQueryMaxSwapBarriersSGIX; + + /*** GLX_SUN_get_transparent_index ***/ + glx.GetTransparentIndexSUN = Fake_glXGetTransparentIndexSUN; + + /*** GLX_MESA_copy_sub_buffer ***/ + glx.CopySubBufferMESA = Fake_glXCopySubBufferMESA; + + /*** GLX_MESA_release_buffers ***/ + glx.ReleaseBuffersMESA = Fake_glXReleaseBuffersMESA; + + /*** GLX_MESA_pixmap_colormap ***/ + glx.CreateGLXPixmapMESA = Fake_glXCreateGLXPixmapMESA; + + /*** GLX_MESA_set_3dfx_mode ***/ + glx.Set3DfxModeMESA = Fake_glXSet3DfxModeMESA; + + /*** GLX_NV_vertex_array_range ***/ + glx.AllocateMemoryNV = Fake_glXAllocateMemoryNV; + glx.FreeMemoryNV = Fake_glXFreeMemoryNV; + + /*** GLX_MESA_agp_offset ***/ + glx.GetAGPOffsetMESA = Fake_glXGetAGPOffsetMESA; + + /*** GLX_EXT_texture_from_pixmap ***/ + glx.BindTexImageEXT = Fake_glXBindTexImageEXT; + glx.ReleaseTexImageEXT = Fake_glXReleaseTexImageEXT; + + return &glx; +} diff --git a/src/gallium/winsys/xlib/glxapi.c b/src/gallium/winsys/xlib/glxapi.c new file mode 100644 index 0000000000..c2ccce6f52 --- /dev/null +++ b/src/gallium/winsys/xlib/glxapi.c @@ -0,0 +1,1408 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* + * This is the GLX API dispatcher. Calls to the glX* functions are + * either routed to the real GLX encoders or to Mesa's pseudo-GLX functions. + * See the glxapi.h file for more details. + */ + + +#include +#include +#include +#include +#include "main/glheader.h" +#include "glapi/glapi.h" +#include "glxapi.h" + + +extern struct _glxapi_table *_real_GetGLXDispatchTable(void); +extern struct _glxapi_table *_mesa_GetGLXDispatchTable(void); + + +struct display_dispatch { + Display *Dpy; + struct _glxapi_table *Table; + struct display_dispatch *Next; +}; + +static struct display_dispatch *DispatchList = NULL; + + +/* Display -> Dispatch caching */ +static Display *prevDisplay = NULL; +static struct _glxapi_table *prevTable = NULL; + + +static struct _glxapi_table * +get_dispatch(Display *dpy) +{ + if (!dpy) + return NULL; + + /* search list of display/dispatch pairs for this display */ + { + const struct display_dispatch *d = DispatchList; + while (d) { + if (d->Dpy == dpy) { + prevDisplay = dpy; + prevTable = d->Table; + return d->Table; /* done! */ + } + d = d->Next; + } + } + + /* A new display, determine if we should use real GLX + * or Mesa's pseudo-GLX. + */ + { + struct _glxapi_table *t = _mesa_GetGLXDispatchTable(); + + if (t) { + struct display_dispatch *d; + d = (struct display_dispatch *) malloc(sizeof(struct display_dispatch)); + if (d) { + d->Dpy = dpy; + d->Table = t; + /* insert at head of list */ + d->Next = DispatchList; + DispatchList = d; + /* update cache */ + prevDisplay = dpy; + prevTable = t; + return t; + } + } + } + + /* If we get here that means we can't use real GLX on this display + * and the Mesa pseudo-GLX software renderer wasn't compiled in. + * Or, we ran out of memory! + */ + return NULL; +} + + +/* Don't use the GET_DISPATCH defined in glthread.h */ +#undef GET_DISPATCH + +#define GET_DISPATCH(DPY, TABLE) \ + if (DPY == prevDisplay) { \ + TABLE = prevTable; \ + } \ + else if (!DPY) { \ + TABLE = NULL; \ + } \ + else { \ + TABLE = get_dispatch(DPY); \ + } + + + + +/** + * GLX API current context. + */ +#if defined(GLX_USE_TLS) +PUBLIC __thread void * CurrentContext + __attribute__((tls_model("initial-exec"))); +#elif defined(THREADS) +static _glthread_TSD ContextTSD; /**< Per-thread context pointer */ +#else +static GLXContext CurrentContext = 0; +#endif + + +static void +SetCurrentContext(GLXContext c) +{ +#if defined(GLX_USE_TLS) + CurrentContext = c; +#elif defined(THREADS) + _glthread_SetTSD(&ContextTSD, c); +#else + CurrentContext = c; +#endif +} + + +/* + * GLX API entrypoints + */ + +/*** GLX_VERSION_1_0 ***/ + +XVisualInfo PUBLIC * +glXChooseVisual(Display *dpy, int screen, int *list) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->ChooseVisual)(dpy, screen, list); +} + + +void PUBLIC +glXCopyContext(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->CopyContext)(dpy, src, dst, mask); +} + + +GLXContext PUBLIC +glXCreateContext(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateContext)(dpy, visinfo, shareList, direct); +} + + +GLXPixmap PUBLIC +glXCreateGLXPixmap(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXPixmap)(dpy, visinfo, pixmap); +} + + +void PUBLIC +glXDestroyContext(Display *dpy, GLXContext ctx) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + if (glXGetCurrentContext() == ctx) + SetCurrentContext(NULL); + (t->DestroyContext)(dpy, ctx); +} + + +void PUBLIC +glXDestroyGLXPixmap(Display *dpy, GLXPixmap pixmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyGLXPixmap)(dpy, pixmap); +} + + +int PUBLIC +glXGetConfig(Display *dpy, XVisualInfo *visinfo, int attrib, int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return GLX_NO_EXTENSION; + return (t->GetConfig)(dpy, visinfo, attrib, value); +} + + +GLXContext PUBLIC +glXGetCurrentContext(void) +{ +#if defined(GLX_USE_TLS) + return CurrentContext; +#elif defined(THREADS) + return (GLXContext) _glthread_GetTSD(&ContextTSD); +#else + return CurrentContext; +#endif +} + + +GLXDrawable PUBLIC +glXGetCurrentDrawable(void) +{ + __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); + return gc ? gc->currentDrawable : 0; +} + + +Bool PUBLIC +glXIsDirect(Display *dpy, GLXContext ctx) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->IsDirect)(dpy, ctx); +} + + +Bool PUBLIC +glXMakeCurrent(Display *dpy, GLXDrawable drawable, GLXContext ctx) +{ + Bool b; + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) { + return False; + } + b = (*t->MakeCurrent)(dpy, drawable, ctx); + if (b) { + SetCurrentContext(ctx); + } + return b; +} + + +Bool PUBLIC +glXQueryExtension(Display *dpy, int *errorb, int *event) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->QueryExtension)(dpy, errorb, event); +} + + +Bool PUBLIC +glXQueryVersion(Display *dpy, int *maj, int *min) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->QueryVersion)(dpy, maj, min); +} + + +void PUBLIC +glXSwapBuffers(Display *dpy, GLXDrawable drawable) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->SwapBuffers)(dpy, drawable); +} + + +void PUBLIC +glXUseXFont(Font font, int first, int count, int listBase) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->UseXFont)(font, first, count, listBase); +} + + +void PUBLIC +glXWaitGL(void) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->WaitGL)(); +} + + +void PUBLIC +glXWaitX(void) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->WaitX)(); +} + + + +/*** GLX_VERSION_1_1 ***/ + +const char PUBLIC * +glXGetClientString(Display *dpy, int name) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->GetClientString)(dpy, name); +} + + +const char PUBLIC * +glXQueryExtensionsString(Display *dpy, int screen) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->QueryExtensionsString)(dpy, screen); +} + + +const char PUBLIC * +glXQueryServerString(Display *dpy, int screen, int name) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->QueryServerString)(dpy, screen, name); +} + + +/*** GLX_VERSION_1_2 ***/ + +Display PUBLIC * +glXGetCurrentDisplay(void) +{ + /* Same code as in libGL's glxext.c */ + __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); + if (NULL == gc) return NULL; + return gc->currentDpy; +} + + + +/*** GLX_VERSION_1_3 ***/ + +GLXFBConfig PUBLIC * +glXChooseFBConfig(Display *dpy, int screen, const int *attribList, int *nitems) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ChooseFBConfig)(dpy, screen, attribList, nitems); +} + + +GLXContext PUBLIC +glXCreateNewContext(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateNewContext)(dpy, config, renderType, shareList, direct); +} + + +GLXPbuffer PUBLIC +glXCreatePbuffer(Display *dpy, GLXFBConfig config, const int *attribList) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreatePbuffer)(dpy, config, attribList); +} + + +GLXPixmap PUBLIC +glXCreatePixmap(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreatePixmap)(dpy, config, pixmap, attribList); +} + + +GLXWindow PUBLIC +glXCreateWindow(Display *dpy, GLXFBConfig config, Window win, const int *attribList) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateWindow)(dpy, config, win, attribList); +} + + +void PUBLIC +glXDestroyPbuffer(Display *dpy, GLXPbuffer pbuf) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyPbuffer)(dpy, pbuf); +} + + +void PUBLIC +glXDestroyPixmap(Display *dpy, GLXPixmap pixmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyPixmap)(dpy, pixmap); +} + + +void PUBLIC +glXDestroyWindow(Display *dpy, GLXWindow window) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyWindow)(dpy, window); +} + + +GLXDrawable PUBLIC +glXGetCurrentReadDrawable(void) +{ + __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); + return gc ? gc->currentReadable : 0; +} + + +int PUBLIC +glXGetFBConfigAttrib(Display *dpy, GLXFBConfig config, int attribute, int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return GLX_NO_EXTENSION; + return (t->GetFBConfigAttrib)(dpy, config, attribute, value); +} + + +GLXFBConfig PUBLIC * +glXGetFBConfigs(Display *dpy, int screen, int *nelements) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->GetFBConfigs)(dpy, screen, nelements); +} + +void PUBLIC +glXGetSelectedEvent(Display *dpy, GLXDrawable drawable, unsigned long *mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->GetSelectedEvent)(dpy, drawable, mask); +} + + +XVisualInfo PUBLIC * +glXGetVisualFromFBConfig(Display *dpy, GLXFBConfig config) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->GetVisualFromFBConfig)(dpy, config); +} + + +Bool PUBLIC +glXMakeContextCurrent(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) +{ + Bool b; + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + b = (t->MakeContextCurrent)(dpy, draw, read, ctx); + if (b) { + SetCurrentContext(ctx); + } + return b; +} + + +int PUBLIC +glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + assert(t); + if (!t) + return 0; /* XXX correct? */ + return (t->QueryContext)(dpy, ctx, attribute, value); +} + + +void PUBLIC +glXQueryDrawable(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->QueryDrawable)(dpy, draw, attribute, value); +} + + +void PUBLIC +glXSelectEvent(Display *dpy, GLXDrawable drawable, unsigned long mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->SelectEvent)(dpy, drawable, mask); +} + + + +/*** GLX_SGI_swap_control ***/ + +int PUBLIC +glXSwapIntervalSGI(int interval) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->SwapIntervalSGI)(interval); +} + + + +/*** GLX_SGI_video_sync ***/ + +int PUBLIC +glXGetVideoSyncSGI(unsigned int *count) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t || !glXGetCurrentContext()) + return GLX_BAD_CONTEXT; + return (t->GetVideoSyncSGI)(count); +} + +int PUBLIC +glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t || !glXGetCurrentContext()) + return GLX_BAD_CONTEXT; + return (t->WaitVideoSyncSGI)(divisor, remainder, count); +} + + + +/*** GLX_SGI_make_current_read ***/ + +Bool PUBLIC +glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->MakeCurrentReadSGI)(dpy, draw, read, ctx); +} + +GLXDrawable PUBLIC +glXGetCurrentReadDrawableSGI(void) +{ + return glXGetCurrentReadDrawable(); +} + + +#if defined(_VL_H) + +GLXVideoSourceSGIX PUBLIC +glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXVideoSourceSGIX)(dpy, screen, server, path, nodeClass, drainNode); +} + +void PUBLIC +glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->DestroyGLXVideoSourceSGIX)(dpy, src); +} + +#endif + + +/*** GLX_EXT_import_context ***/ + +void PUBLIC +glXFreeContextEXT(Display *dpy, GLXContext context) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->FreeContextEXT)(dpy, context); +} + +GLXContextID PUBLIC +glXGetContextIDEXT(const GLXContext context) +{ + return ((__GLXcontext *) context)->xid; +} + +Display PUBLIC * +glXGetCurrentDisplayEXT(void) +{ + return glXGetCurrentDisplay(); +} + +GLXContext PUBLIC +glXImportContextEXT(Display *dpy, GLXContextID contextID) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ImportContextEXT)(dpy, contextID); +} + +int PUBLIC +glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute,int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; /* XXX ok? */ + return (t->QueryContextInfoEXT)(dpy, context, attribute, value); +} + + + +/*** GLX_SGIX_fbconfig ***/ + +int PUBLIC +glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->GetFBConfigAttribSGIX)(dpy, config, attribute, value); +} + +GLXFBConfigSGIX PUBLIC * +glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ChooseFBConfigSGIX)(dpy, screen, attrib_list, nelements); +} + +GLXPixmap PUBLIC +glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXPixmapWithConfigSGIX)(dpy, config, pixmap); +} + +GLXContext PUBLIC +glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateContextWithConfigSGIX)(dpy, config, render_type, share_list, direct); +} + +XVisualInfo PUBLIC * +glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->GetVisualFromFBConfigSGIX)(dpy, config); +} + +GLXFBConfigSGIX PUBLIC +glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->GetFBConfigFromVisualSGIX)(dpy, vis); +} + + + +/*** GLX_SGIX_pbuffer ***/ + +GLXPbufferSGIX PUBLIC +glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attrib_list) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXPbufferSGIX)(dpy, config, width, height, attrib_list); +} + +void PUBLIC +glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->DestroyGLXPbufferSGIX)(dpy, pbuf); +} + +int PUBLIC +glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->QueryGLXPbufferSGIX)(dpy, pbuf, attribute, value); +} + +void PUBLIC +glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->SelectEventSGIX)(dpy, drawable, mask); +} + +void PUBLIC +glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->GetSelectedEventSGIX)(dpy, drawable, mask); +} + + + +/*** GLX_SGI_cushion ***/ + +void PUBLIC +glXCushionSGI(Display *dpy, Window win, float cushion) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->CushionSGI)(dpy, win, cushion); +} + + + +/*** GLX_SGIX_video_resize ***/ + +int PUBLIC +glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->BindChannelToWindowSGIX)(dpy, screen, channel, window); +} + +int PUBLIC +glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ChannelRectSGIX)(dpy, screen, channel, x, y, w, h); +} + +int PUBLIC +glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->QueryChannelRectSGIX)(dpy, screen, channel, x, y, w, h); +} + +int PUBLIC +glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->QueryChannelDeltasSGIX)(dpy, screen, channel, dx, dy, dw, dh); +} + +int PUBLIC +glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->ChannelRectSyncSGIX)(dpy, screen, channel, synctype); +} + + + +#if defined(_DM_BUFFER_H_) + +Bool PUBLIC +glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->AssociateDMPbufferSGIX)(dpy, pbuffer, params, dmbuffer); +} + +#endif + + +/*** GLX_SGIX_swap_group ***/ + +void PUBLIC +glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (*t->JoinSwapGroupSGIX)(dpy, drawable, member); +} + + +/*** GLX_SGIX_swap_barrier ***/ + +void PUBLIC +glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (*t->BindSwapBarrierSGIX)(dpy, drawable, barrier); +} + +Bool PUBLIC +glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (*t->QueryMaxSwapBarriersSGIX)(dpy, screen, max); +} + + + +/*** GLX_SUN_get_transparent_index ***/ + +Status PUBLIC +glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (*t->GetTransparentIndexSUN)(dpy, overlay, underlay, pTransparent); +} + + + +/*** GLX_MESA_copy_sub_buffer ***/ + +void PUBLIC +glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->CopySubBufferMESA)(dpy, drawable, x, y, width, height); +} + + + +/*** GLX_MESA_release_buffers ***/ + +Bool PUBLIC +glXReleaseBuffersMESA(Display *dpy, Window w) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->ReleaseBuffersMESA)(dpy, w); +} + + + +/*** GLX_MESA_pixmap_colormap ***/ + +GLXPixmap PUBLIC +glXCreateGLXPixmapMESA(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (!t) + return 0; + return (t->CreateGLXPixmapMESA)(dpy, visinfo, pixmap, cmap); +} + + + +/*** GLX_MESA_set_3dfx_mode ***/ + +Bool PUBLIC +glXSet3DfxModeMESA(int mode) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return False; + return (t->Set3DfxModeMESA)(mode); +} + + + +/*** GLX_NV_vertex_array_range ***/ + +void PUBLIC * +glXAllocateMemoryNV( GLsizei size, + GLfloat readFrequency, + GLfloat writeFrequency, + GLfloat priority ) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return NULL; + return (t->AllocateMemoryNV)(size, readFrequency, writeFrequency, priority); +} + + +void PUBLIC +glXFreeMemoryNV( GLvoid *pointer ) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return; + (t->FreeMemoryNV)(pointer); +} + + + + +/*** GLX_MESA_agp_offset */ + +GLuint PUBLIC +glXGetAGPOffsetMESA( const GLvoid *pointer ) +{ + struct _glxapi_table *t; + Display *dpy = glXGetCurrentDisplay(); + GET_DISPATCH(dpy, t); + if (!t) + return ~0; + return (t->GetAGPOffsetMESA)(pointer); +} + + +/*** GLX_MESA_allocate_memory */ + +void * +glXAllocateMemoryMESA(Display *dpy, int scrn, size_t size, + float readfreq, float writefreq, float priority) +{ + /* dummy */ + return NULL; +} + +void +glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer) +{ + /* dummy */ +} + + +GLuint +glXGetMemoryOffsetMESA(Display *dpy, int scrn, const void *pointer) +{ + /* dummy */ + return 0; +} + + +/*** GLX_EXT_texture_from_pixmap */ + +void +glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, + const int *attrib_list) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (t) + t->BindTexImageEXT(dpy, drawable, buffer, attrib_list); +} + +void +glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer) +{ + struct _glxapi_table *t; + GET_DISPATCH(dpy, t); + if (t) + t->ReleaseTexImageEXT(dpy, drawable, buffer); +} + + +/**********************************************************************/ +/* GLX API management functions */ +/**********************************************************************/ + + +const char * +_glxapi_get_version(void) +{ + return "1.3"; +} + + +/* + * Return array of extension strings. + */ +const char ** +_glxapi_get_extensions(void) +{ + static const char *extensions[] = { +#ifdef GLX_EXT_import_context + "GLX_EXT_import_context", +#endif +#ifdef GLX_SGI_video_sync + "GLX_SGI_video_sync", +#endif +#ifdef GLX_MESA_copy_sub_buffer + "GLX_MESA_copy_sub_buffer", +#endif +#ifdef GLX_MESA_release_buffers + "GLX_MESA_release_buffers", +#endif +#ifdef GLX_MESA_pixmap_colormap + "GLX_MESA_pixmap_colormap", +#endif +#ifdef GLX_MESA_set_3dfx_mode + "GLX_MESA_set_3dfx_mode", +#endif +#ifdef GLX_SGIX_fbconfig + "GLX_SGIX_fbconfig", +#endif +#ifdef GLX_SGIX_pbuffer + "GLX_SGIX_pbuffer", +#endif +#ifdef GLX_EXT_texture_from_pixmap + "GLX_EXT_texture_from_pixmap", +#endif + NULL + }; + return extensions; +} + + +/* + * Return size of the GLX dispatch table, in entries, not bytes. + */ +GLuint +_glxapi_get_dispatch_table_size(void) +{ + return sizeof(struct _glxapi_table) / sizeof(void *); +} + + +static int +generic_no_op_func(void) +{ + return 0; +} + + +/* + * Initialize all functions in given dispatch table to be no-ops + */ +void +_glxapi_set_no_op_table(struct _glxapi_table *t) +{ + typedef int (*nop_func)(void); + nop_func *dispatch = (nop_func *) t; + GLuint n = _glxapi_get_dispatch_table_size(); + GLuint i; + for (i = 0; i < n; i++) { + dispatch[i] = generic_no_op_func; + } +} + + +struct name_address_pair { + const char *Name; + __GLXextFuncPtr Address; +}; + +static struct name_address_pair GLX_functions[] = { + /*** GLX_VERSION_1_0 ***/ + { "glXChooseVisual", (__GLXextFuncPtr) glXChooseVisual }, + { "glXCopyContext", (__GLXextFuncPtr) glXCopyContext }, + { "glXCreateContext", (__GLXextFuncPtr) glXCreateContext }, + { "glXCreateGLXPixmap", (__GLXextFuncPtr) glXCreateGLXPixmap }, + { "glXDestroyContext", (__GLXextFuncPtr) glXDestroyContext }, + { "glXDestroyGLXPixmap", (__GLXextFuncPtr) glXDestroyGLXPixmap }, + { "glXGetConfig", (__GLXextFuncPtr) glXGetConfig }, + { "glXGetCurrentContext", (__GLXextFuncPtr) glXGetCurrentContext }, + { "glXGetCurrentDrawable", (__GLXextFuncPtr) glXGetCurrentDrawable }, + { "glXIsDirect", (__GLXextFuncPtr) glXIsDirect }, + { "glXMakeCurrent", (__GLXextFuncPtr) glXMakeCurrent }, + { "glXQueryExtension", (__GLXextFuncPtr) glXQueryExtension }, + { "glXQueryVersion", (__GLXextFuncPtr) glXQueryVersion }, + { "glXSwapBuffers", (__GLXextFuncPtr) glXSwapBuffers }, + { "glXUseXFont", (__GLXextFuncPtr) glXUseXFont }, + { "glXWaitGL", (__GLXextFuncPtr) glXWaitGL }, + { "glXWaitX", (__GLXextFuncPtr) glXWaitX }, + + /*** GLX_VERSION_1_1 ***/ + { "glXGetClientString", (__GLXextFuncPtr) glXGetClientString }, + { "glXQueryExtensionsString", (__GLXextFuncPtr) glXQueryExtensionsString }, + { "glXQueryServerString", (__GLXextFuncPtr) glXQueryServerString }, + + /*** GLX_VERSION_1_2 ***/ + { "glXGetCurrentDisplay", (__GLXextFuncPtr) glXGetCurrentDisplay }, + + /*** GLX_VERSION_1_3 ***/ + { "glXChooseFBConfig", (__GLXextFuncPtr) glXChooseFBConfig }, + { "glXCreateNewContext", (__GLXextFuncPtr) glXCreateNewContext }, + { "glXCreatePbuffer", (__GLXextFuncPtr) glXCreatePbuffer }, + { "glXCreatePixmap", (__GLXextFuncPtr) glXCreatePixmap }, + { "glXCreateWindow", (__GLXextFuncPtr) glXCreateWindow }, + { "glXDestroyPbuffer", (__GLXextFuncPtr) glXDestroyPbuffer }, + { "glXDestroyPixmap", (__GLXextFuncPtr) glXDestroyPixmap }, + { "glXDestroyWindow", (__GLXextFuncPtr) glXDestroyWindow }, + { "glXGetCurrentReadDrawable", (__GLXextFuncPtr) glXGetCurrentReadDrawable }, + { "glXGetFBConfigAttrib", (__GLXextFuncPtr) glXGetFBConfigAttrib }, + { "glXGetFBConfigs", (__GLXextFuncPtr) glXGetFBConfigs }, + { "glXGetSelectedEvent", (__GLXextFuncPtr) glXGetSelectedEvent }, + { "glXGetVisualFromFBConfig", (__GLXextFuncPtr) glXGetVisualFromFBConfig }, + { "glXMakeContextCurrent", (__GLXextFuncPtr) glXMakeContextCurrent }, + { "glXQueryContext", (__GLXextFuncPtr) glXQueryContext }, + { "glXQueryDrawable", (__GLXextFuncPtr) glXQueryDrawable }, + { "glXSelectEvent", (__GLXextFuncPtr) glXSelectEvent }, + + /*** GLX_VERSION_1_4 ***/ + { "glXGetProcAddress", (__GLXextFuncPtr) glXGetProcAddress }, + + /*** GLX_SGI_swap_control ***/ + { "glXSwapIntervalSGI", (__GLXextFuncPtr) glXSwapIntervalSGI }, + + /*** GLX_SGI_video_sync ***/ + { "glXGetVideoSyncSGI", (__GLXextFuncPtr) glXGetVideoSyncSGI }, + { "glXWaitVideoSyncSGI", (__GLXextFuncPtr) glXWaitVideoSyncSGI }, + + /*** GLX_SGI_make_current_read ***/ + { "glXMakeCurrentReadSGI", (__GLXextFuncPtr) glXMakeCurrentReadSGI }, + { "glXGetCurrentReadDrawableSGI", (__GLXextFuncPtr) glXGetCurrentReadDrawableSGI }, + + /*** GLX_SGIX_video_source ***/ +#if defined(_VL_H) + { "glXCreateGLXVideoSourceSGIX", (__GLXextFuncPtr) glXCreateGLXVideoSourceSGIX }, + { "glXDestroyGLXVideoSourceSGIX", (__GLXextFuncPtr) glXDestroyGLXVideoSourceSGIX }, +#endif + + /*** GLX_EXT_import_context ***/ + { "glXFreeContextEXT", (__GLXextFuncPtr) glXFreeContextEXT }, + { "glXGetContextIDEXT", (__GLXextFuncPtr) glXGetContextIDEXT }, + { "glXGetCurrentDisplayEXT", (__GLXextFuncPtr) glXGetCurrentDisplayEXT }, + { "glXImportContextEXT", (__GLXextFuncPtr) glXImportContextEXT }, + { "glXQueryContextInfoEXT", (__GLXextFuncPtr) glXQueryContextInfoEXT }, + + /*** GLX_SGIX_fbconfig ***/ + { "glXGetFBConfigAttribSGIX", (__GLXextFuncPtr) glXGetFBConfigAttribSGIX }, + { "glXChooseFBConfigSGIX", (__GLXextFuncPtr) glXChooseFBConfigSGIX }, + { "glXCreateGLXPixmapWithConfigSGIX", (__GLXextFuncPtr) glXCreateGLXPixmapWithConfigSGIX }, + { "glXCreateContextWithConfigSGIX", (__GLXextFuncPtr) glXCreateContextWithConfigSGIX }, + { "glXGetVisualFromFBConfigSGIX", (__GLXextFuncPtr) glXGetVisualFromFBConfigSGIX }, + { "glXGetFBConfigFromVisualSGIX", (__GLXextFuncPtr) glXGetFBConfigFromVisualSGIX }, + + /*** GLX_SGIX_pbuffer ***/ + { "glXCreateGLXPbufferSGIX", (__GLXextFuncPtr) glXCreateGLXPbufferSGIX }, + { "glXDestroyGLXPbufferSGIX", (__GLXextFuncPtr) glXDestroyGLXPbufferSGIX }, + { "glXQueryGLXPbufferSGIX", (__GLXextFuncPtr) glXQueryGLXPbufferSGIX }, + { "glXSelectEventSGIX", (__GLXextFuncPtr) glXSelectEventSGIX }, + { "glXGetSelectedEventSGIX", (__GLXextFuncPtr) glXGetSelectedEventSGIX }, + + /*** GLX_SGI_cushion ***/ + { "glXCushionSGI", (__GLXextFuncPtr) glXCushionSGI }, + + /*** GLX_SGIX_video_resize ***/ + { "glXBindChannelToWindowSGIX", (__GLXextFuncPtr) glXBindChannelToWindowSGIX }, + { "glXChannelRectSGIX", (__GLXextFuncPtr) glXChannelRectSGIX }, + { "glXQueryChannelRectSGIX", (__GLXextFuncPtr) glXQueryChannelRectSGIX }, + { "glXQueryChannelDeltasSGIX", (__GLXextFuncPtr) glXQueryChannelDeltasSGIX }, + { "glXChannelRectSyncSGIX", (__GLXextFuncPtr) glXChannelRectSyncSGIX }, + + /*** GLX_SGIX_dmbuffer **/ +#if defined(_DM_BUFFER_H_) + { "glXAssociateDMPbufferSGIX", (__GLXextFuncPtr) glXAssociateDMPbufferSGIX }, +#endif + + /*** GLX_SGIX_swap_group ***/ + { "glXJoinSwapGroupSGIX", (__GLXextFuncPtr) glXJoinSwapGroupSGIX }, + + /*** GLX_SGIX_swap_barrier ***/ + { "glXBindSwapBarrierSGIX", (__GLXextFuncPtr) glXBindSwapBarrierSGIX }, + { "glXQueryMaxSwapBarriersSGIX", (__GLXextFuncPtr) glXQueryMaxSwapBarriersSGIX }, + + /*** GLX_SUN_get_transparent_index ***/ + { "glXGetTransparentIndexSUN", (__GLXextFuncPtr) glXGetTransparentIndexSUN }, + + /*** GLX_MESA_copy_sub_buffer ***/ + { "glXCopySubBufferMESA", (__GLXextFuncPtr) glXCopySubBufferMESA }, + + /*** GLX_MESA_pixmap_colormap ***/ + { "glXCreateGLXPixmapMESA", (__GLXextFuncPtr) glXCreateGLXPixmapMESA }, + + /*** GLX_MESA_release_buffers ***/ + { "glXReleaseBuffersMESA", (__GLXextFuncPtr) glXReleaseBuffersMESA }, + + /*** GLX_MESA_set_3dfx_mode ***/ + { "glXSet3DfxModeMESA", (__GLXextFuncPtr) glXSet3DfxModeMESA }, + + /*** GLX_ARB_get_proc_address ***/ + { "glXGetProcAddressARB", (__GLXextFuncPtr) glXGetProcAddressARB }, + + /*** GLX_NV_vertex_array_range ***/ + { "glXAllocateMemoryNV", (__GLXextFuncPtr) glXAllocateMemoryNV }, + { "glXFreeMemoryNV", (__GLXextFuncPtr) glXFreeMemoryNV }, + + /*** GLX_MESA_agp_offset ***/ + { "glXGetAGPOffsetMESA", (__GLXextFuncPtr) glXGetAGPOffsetMESA }, + + /*** GLX_MESA_allocate_memory ***/ + { "glXAllocateMemoryMESA", (__GLXextFuncPtr) glXAllocateMemoryMESA }, + { "glXFreeMemoryMESA", (__GLXextFuncPtr) glXFreeMemoryMESA }, + { "glXGetMemoryOffsetMESA", (__GLXextFuncPtr) glXGetMemoryOffsetMESA }, + + /*** GLX_EXT_texture_from_pixmap ***/ + { "glXBindTexImageEXT", (__GLXextFuncPtr) glXBindTexImageEXT }, + { "glXReleaseTexImageEXT", (__GLXextFuncPtr) glXReleaseTexImageEXT }, + + { NULL, NULL } /* end of list */ +}; + + + +/* + * Return address of named glX function, or NULL if not found. + */ +__GLXextFuncPtr +_glxapi_get_proc_address(const char *funcName) +{ + GLuint i; + for (i = 0; GLX_functions[i].Name; i++) { + if (strcmp(GLX_functions[i].Name, funcName) == 0) + return GLX_functions[i].Address; + } + return NULL; +} + + + +/* + * This function does not get dispatched through the dispatch table + * since it's really a "meta" function. + */ +__GLXextFuncPtr +glXGetProcAddressARB(const GLubyte *procName) +{ + __GLXextFuncPtr f; + + f = _glxapi_get_proc_address((const char *) procName); + if (f) { + return f; + } + + f = (__GLXextFuncPtr) _glapi_get_proc_address((const char *) procName); + return f; +} + + +/* GLX 1.4 */ +void (*glXGetProcAddress(const GLubyte *procName))() +{ + return glXGetProcAddressARB(procName); +} diff --git a/src/gallium/winsys/xlib/glxapi.h b/src/gallium/winsys/xlib/glxapi.h new file mode 100644 index 0000000000..37de81e55a --- /dev/null +++ b/src/gallium/winsys/xlib/glxapi.h @@ -0,0 +1,228 @@ +/* + * Mesa 3-D graphics library + * Version: 6.3 + * + * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef _glxapi_h_ +#define _glxapi_h_ + + +#define GLX_GLXEXT_PROTOTYPES +#include "GL/glx.h" + + +/* The GLX API dispatcher (i.e. this code) is being built into stand-alone + * Mesa. We don't know anything about XFree86 or real GLX so we define a + * minimal __GLXContextRec here so some of the functions in this file can + * work properly. + */ +typedef struct __GLXcontextRec { + Display *currentDpy; + GLboolean isDirect; + GLXDrawable currentDrawable; + GLXDrawable currentReadable; + XID xid; +} __GLXcontext; + + +/* + * Almost all the GLX API functions get routed through this dispatch table. + * The exceptions are the glXGetCurrentXXX() functions. + * + * This dispatch table allows multiple GLX client-side modules to coexist. + * Specifically, a real GLX library (like SGI's or the Utah GLX) and Mesa's + * pseudo-GLX can be present at the same time. The former being used on + * GLX-enabled X servers and the later on non-GLX X servers. + * + * Red Hat has been using this since Red Hat Linux 7.0 (I think). + * This'll be a standard feature in XFree86 4.3. It basically allows one + * libGL to do both DRI-rendering and "fake GLX" rendering to X displays + * that lack the GLX extension. + */ +struct _glxapi_table { + /*** GLX_VERSION_1_0 ***/ + XVisualInfo *(*ChooseVisual)(Display *dpy, int screen, int *list); + void (*CopyContext)(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask); + GLXContext (*CreateContext)(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct); + GLXPixmap (*CreateGLXPixmap)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap); + void (*DestroyContext)(Display *dpy, GLXContext ctx); + void (*DestroyGLXPixmap)(Display *dpy, GLXPixmap pixmap); + int (*GetConfig)(Display *dpy, XVisualInfo *visinfo, int attrib, int *value); + /*GLXContext (*GetCurrentContext)(void);*/ + /*GLXDrawable (*GetCurrentDrawable)(void);*/ + Bool (*IsDirect)(Display *dpy, GLXContext ctx); + Bool (*MakeCurrent)(Display *dpy, GLXDrawable drawable, GLXContext ctx); + Bool (*QueryExtension)(Display *dpy, int *errorb, int *event); + Bool (*QueryVersion)(Display *dpy, int *maj, int *min); + void (*SwapBuffers)(Display *dpy, GLXDrawable drawable); + void (*UseXFont)(Font font, int first, int count, int listBase); + void (*WaitGL)(void); + void (*WaitX)(void); + + /*** GLX_VERSION_1_1 ***/ + const char *(*GetClientString)(Display *dpy, int name); + const char *(*QueryExtensionsString)(Display *dpy, int screen); + const char *(*QueryServerString)(Display *dpy, int screen, int name); + + /*** GLX_VERSION_1_2 ***/ + /*Display *(*GetCurrentDisplay)(void);*/ + + /*** GLX_VERSION_1_3 ***/ + GLXFBConfig *(*ChooseFBConfig)(Display *dpy, int screen, const int *attribList, int *nitems); + GLXContext (*CreateNewContext)(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct); + GLXPbuffer (*CreatePbuffer)(Display *dpy, GLXFBConfig config, const int *attribList); + GLXPixmap (*CreatePixmap)(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList); + GLXWindow (*CreateWindow)(Display *dpy, GLXFBConfig config, Window win, const int *attribList); + void (*DestroyPbuffer)(Display *dpy, GLXPbuffer pbuf); + void (*DestroyPixmap)(Display *dpy, GLXPixmap pixmap); + void (*DestroyWindow)(Display *dpy, GLXWindow window); + /*GLXDrawable (*GetCurrentReadDrawable)(void);*/ + int (*GetFBConfigAttrib)(Display *dpy, GLXFBConfig config, int attribute, int *value); + GLXFBConfig *(*GetFBConfigs)(Display *dpy, int screen, int *nelements); + void (*GetSelectedEvent)(Display *dpy, GLXDrawable drawable, unsigned long *mask); + XVisualInfo *(*GetVisualFromFBConfig)(Display *dpy, GLXFBConfig config); + Bool (*MakeContextCurrent)(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx); + int (*QueryContext)(Display *dpy, GLXContext ctx, int attribute, int *value); + void (*QueryDrawable)(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value); + void (*SelectEvent)(Display *dpy, GLXDrawable drawable, unsigned long mask); + + /*** GLX_SGI_swap_control ***/ + int (*SwapIntervalSGI)(int); + + /*** GLX_SGI_video_sync ***/ + int (*GetVideoSyncSGI)(unsigned int *count); + int (*WaitVideoSyncSGI)(int divisor, int remainder, unsigned int *count); + + /*** GLX_SGI_make_current_read ***/ + Bool (*MakeCurrentReadSGI)(Display *, GLXDrawable, GLXDrawable, GLXContext); + /*GLXDrawable (*GetCurrentReadDrawableSGI)(void);*/ + + /*** GLX_SGIX_video_source (needs video library) ***/ +#if defined(_VL_H_) + GLXVideoSourceSGIX (*CreateGLXVideoSourceSGIX)(Display *, int, VLServer, VLPath, int, VLNode); + void (*DestroyGLXVideoSourceSGIX)(Display *, GLXVideoSourceSGIX); +#else + void *CreateGLXVideoSourceSGIX; + void *DestroyGLXVideoSourceSGIX; +#endif + + /*** GLX_EXT_import_context ***/ + void (*FreeContextEXT)(Display *dpy, GLXContext context); + GLXContextID (*GetContextIDEXT)(const GLXContext context); + /*Display *(*GetCurrentDisplayEXT)(void);*/ + GLXContext (*ImportContextEXT)(Display *dpy, GLXContextID contextID); + int (*QueryContextInfoEXT)(Display *dpy, GLXContext context, int attribute,int *value); + + /*** GLX_SGIX_fbconfig ***/ + int (*GetFBConfigAttribSGIX)(Display *, GLXFBConfigSGIX, int, int *); + GLXFBConfigSGIX * (*ChooseFBConfigSGIX)(Display *, int, int *, int *); + GLXPixmap (*CreateGLXPixmapWithConfigSGIX)(Display *, GLXFBConfigSGIX, Pixmap); + GLXContext (*CreateContextWithConfigSGIX)(Display *, GLXFBConfigSGIX, int, GLXContext, Bool); + XVisualInfo * (*GetVisualFromFBConfigSGIX)(Display *, GLXFBConfigSGIX); + GLXFBConfigSGIX (*GetFBConfigFromVisualSGIX)(Display *, XVisualInfo *); + + /*** GLX_SGIX_pbuffer ***/ + GLXPbufferSGIX (*CreateGLXPbufferSGIX)(Display *, GLXFBConfigSGIX, unsigned int, unsigned int, int *); + void (*DestroyGLXPbufferSGIX)(Display *, GLXPbufferSGIX); + int (*QueryGLXPbufferSGIX)(Display *, GLXPbufferSGIX, int, unsigned int *); + void (*SelectEventSGIX)(Display *, GLXDrawable, unsigned long); + void (*GetSelectedEventSGIX)(Display *, GLXDrawable, unsigned long *); + + /*** GLX_SGI_cushion ***/ + void (*CushionSGI)(Display *, Window, float); + + /*** GLX_SGIX_video_resize ***/ + int (*BindChannelToWindowSGIX)(Display *, int, int, Window); + int (*ChannelRectSGIX)(Display *, int, int, int, int, int, int); + int (*QueryChannelRectSGIX)(Display *, int, int, int *, int *, int *, int *); + int (*QueryChannelDeltasSGIX)(Display *, int, int, int *, int *, int *, int *); + int (*ChannelRectSyncSGIX)(Display *, int, int, GLenum); + + /*** GLX_SGIX_dmbuffer (needs dmedia library) ***/ +#if defined (_DM_BUFFER_H_) + Bool (*AssociateDMPbufferSGIX)(Display *, GLXPbufferSGIX, DMparams *, DMbuffer); +#else + void *AssociciateDMPbufferSGIX; +#endif + + /*** GLX_SGIX_swap_group ***/ + void (*JoinSwapGroupSGIX)(Display *, GLXDrawable, GLXDrawable); + + /*** GLX_SGIX_swap_barrier ***/ + void (*BindSwapBarrierSGIX)(Display *, GLXDrawable, int); + Bool (*QueryMaxSwapBarriersSGIX)(Display *, int, int *); + + /*** GLX_SUN_get_transparent_index ***/ + Status (*GetTransparentIndexSUN)(Display *, Window, Window, long *); + + /*** GLX_MESA_copy_sub_buffer ***/ + void (*CopySubBufferMESA)(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height); + + /*** GLX_MESA_release_buffers ***/ + Bool (*ReleaseBuffersMESA)(Display *dpy, Window w); + + /*** GLX_MESA_pixmap_colormap ***/ + GLXPixmap (*CreateGLXPixmapMESA)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap); + + /*** GLX_MESA_set_3dfx_mode ***/ + Bool (*Set3DfxModeMESA)(int mode); + + /*** GLX_NV_vertex_array_range ***/ + void * (*AllocateMemoryNV)( GLsizei size, + GLfloat readFrequency, + GLfloat writeFrequency, + GLfloat priority ); + void (*FreeMemoryNV)( GLvoid *pointer ); + + /*** GLX_MESA_agp_offset ***/ + GLuint (*GetAGPOffsetMESA)( const GLvoid *pointer ); + + /*** GLX_EXT_texture_from_pixmap ***/ + void (*BindTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer, + const int *attrib_list); + void (*ReleaseTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer); +}; + + + +extern const char * +_glxapi_get_version(void); + + +extern const char ** +_glxapi_get_extensions(void); + + +extern GLuint +_glxapi_get_dispatch_table_size(void); + + +extern void +_glxapi_set_no_op_table(struct _glxapi_table *t); + + +extern __GLXextFuncPtr +_glxapi_get_proc_address(const char *funcName); + + +#endif diff --git a/src/gallium/winsys/xlib/glxheader.h b/src/gallium/winsys/xlib/glxheader.h new file mode 100644 index 0000000000..a402191f13 --- /dev/null +++ b/src/gallium/winsys/xlib/glxheader.h @@ -0,0 +1,62 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5.1 + * + * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef GLX_HEADER_H +#define GLX_HEADER_H + +#ifdef __VMS +#include +#endif + +#include "glheader.h" + +#ifdef XFree86Server + +# include "resource.h" +# include "windowstr.h" + +#else + +# include +# include +# include +# ifdef USE_XSHM /* was SHM */ +# include +# include +# include +# endif +# include +# include + +#endif + + + +/* this silences a compiler warning on several systems */ +struct timespec; +struct itimerspec; + + +#endif /*GLX_HEADER*/ diff --git a/src/gallium/winsys/xlib/realglx.c b/src/gallium/winsys/xlib/realglx.c new file mode 100644 index 0000000000..30adb7465b --- /dev/null +++ b/src/gallium/winsys/xlib/realglx.c @@ -0,0 +1,180 @@ + +/* + * Mesa 3-D graphics library + * Version: 5.1 + * + * Copyright (C) 1999-2002 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#include +#include +#include "realglx.h" +#include "glxapi.h" + + +struct _glxapi_table * +_real_GetGLXDispatchTable(void) +{ + static struct _glxapi_table glx; + + /* be sure our dispatch table size <= libGL's table */ + { + GLuint size = sizeof(struct _glxapi_table) / sizeof(void *); + (void) size; + assert(_glxapi_get_dispatch_table_size() >= size); + } + + /* initialize the whole table to no-ops */ + _glxapi_set_no_op_table(&glx); + + /* now initialize the table with the functions I implement */ + + /*** GLX_VERSION_1_0 ***/ + glx.ChooseVisual = _real_glXChooseVisual; + glx.CopyContext = _real_glXCopyContext; + glx.CreateContext = _real_glXCreateContext; + glx.CreateGLXPixmap = _real_glXCreateGLXPixmap; + glx.DestroyContext = _real_glXDestroyContext; + glx.DestroyGLXPixmap = _real_glXDestroyGLXPixmap; + glx.GetConfig = _real_glXGetConfig; + /*glx.GetCurrentContext = _real_glXGetCurrentContext;*/ + /*glx.GetCurrentDrawable = _real_glXGetCurrentDrawable;*/ + glx.IsDirect = _real_glXIsDirect; + glx.MakeCurrent = _real_glXMakeCurrent; + glx.QueryExtension = _real_glXQueryExtension; + glx.QueryVersion = _real_glXQueryVersion; + glx.SwapBuffers = _real_glXSwapBuffers; + glx.UseXFont = _real_glXUseXFont; + glx.WaitGL = _real_glXWaitGL; + glx.WaitX = _real_glXWaitX; + + /*** GLX_VERSION_1_1 ***/ + glx.GetClientString = _real_glXGetClientString; + glx.QueryExtensionsString = _real_glXQueryExtensionsString; + glx.QueryServerString = _real_glXQueryServerString; + + /*** GLX_VERSION_1_2 ***/ + /*glx.GetCurrentDisplay = _real_glXGetCurrentDisplay;*/ + + /*** GLX_VERSION_1_3 ***/ + glx.ChooseFBConfig = _real_glXChooseFBConfig; + glx.CreateNewContext = _real_glXCreateNewContext; + glx.CreatePbuffer = _real_glXCreatePbuffer; + glx.CreatePixmap = _real_glXCreatePixmap; + glx.CreateWindow = _real_glXCreateWindow; + glx.DestroyPbuffer = _real_glXDestroyPbuffer; + glx.DestroyPixmap = _real_glXDestroyPixmap; + glx.DestroyWindow = _real_glXDestroyWindow; + /*glx.GetCurrentReadDrawable = _real_glXGetCurrentReadDrawable;*/ + glx.GetFBConfigAttrib = _real_glXGetFBConfigAttrib; + glx.GetFBConfigs = _real_glXGetFBConfigs; + glx.GetSelectedEvent = _real_glXGetSelectedEvent; + glx.GetVisualFromFBConfig = _real_glXGetVisualFromFBConfig; + glx.MakeContextCurrent = _real_glXMakeContextCurrent; + glx.QueryContext = _real_glXQueryContext; + glx.QueryDrawable = _real_glXQueryDrawable; + glx.SelectEvent = _real_glXSelectEvent; + + /*** GLX_SGI_swap_control ***/ + glx.SwapIntervalSGI = _real_glXSwapIntervalSGI; + + /*** GLX_SGI_video_sync ***/ + glx.GetVideoSyncSGI = _real_glXGetVideoSyncSGI; + glx.WaitVideoSyncSGI = _real_glXWaitVideoSyncSGI; + + /*** GLX_SGI_make_current_read ***/ + glx.MakeCurrentReadSGI = _real_glXMakeCurrentReadSGI; + /*glx.GetCurrentReadDrawableSGI = _real_glXGetCurrentReadDrawableSGI;*/ + +#if defined(_VL_H) + /*** GLX_SGIX_video_source ***/ + glx.CreateGLXVideoSourceSGIX = _real_glXCreateGLXVideoSourceSGIX; + glx.DestroyGLXVideoSourceSGIX = _real_glXDestroyGLXVideoSourceSGIX; +#endif + + /*** GLX_EXT_import_context ***/ + glx.FreeContextEXT = _real_glXFreeContextEXT; + /*glx.GetContextIDEXT = _real_glXGetContextIDEXT;*/ + /*glx.GetCurrentDisplayEXT = _real_glXGetCurrentDisplayEXT;*/ + glx.ImportContextEXT = _real_glXImportContextEXT; + glx.QueryContextInfoEXT = _real_glXQueryContextInfoEXT; + + /*** GLX_SGIX_fbconfig ***/ + glx.GetFBConfigAttribSGIX = _real_glXGetFBConfigAttribSGIX; + glx.ChooseFBConfigSGIX = _real_glXChooseFBConfigSGIX; + glx.CreateGLXPixmapWithConfigSGIX = _real_glXCreateGLXPixmapWithConfigSGIX; + glx.CreateContextWithConfigSGIX = _real_glXCreateContextWithConfigSGIX; + glx.GetVisualFromFBConfigSGIX = _real_glXGetVisualFromFBConfigSGIX; + glx.GetFBConfigFromVisualSGIX = _real_glXGetFBConfigFromVisualSGIX; + + /*** GLX_SGIX_pbuffer ***/ + glx.CreateGLXPbufferSGIX = _real_glXCreateGLXPbufferSGIX; + glx.DestroyGLXPbufferSGIX = _real_glXDestroyGLXPbufferSGIX; + glx.QueryGLXPbufferSGIX = _real_glXQueryGLXPbufferSGIX; + glx.SelectEventSGIX = _real_glXSelectEventSGIX; + glx.GetSelectedEventSGIX = _real_glXGetSelectedEventSGIX; + + /*** GLX_SGI_cushion ***/ + glx.CushionSGI = _real_glXCushionSGI; + + /*** GLX_SGIX_video_resize ***/ + glx.BindChannelToWindowSGIX = _real_glXBindChannelToWindowSGIX; + glx.ChannelRectSGIX = _real_glXChannelRectSGIX; + glx.QueryChannelRectSGIX = _real_glXQueryChannelRectSGIX; + glx.QueryChannelDeltasSGIX = _real_glXQueryChannelDeltasSGIX; + glx.ChannelRectSyncSGIX = _real_glXChannelRectSyncSGIX; + +#if defined(_DM_BUFFER_H_) + /*** (GLX_SGIX_dmbuffer ***/ + glx.AssociateDMPbufferSGIX = NULL; +#endif + + /*** GLX_SGIX_swap_group ***/ + glx.JoinSwapGroupSGIX = _real_glXJoinSwapGroupSGIX; + + /*** GLX_SGIX_swap_barrier ***/ + glx.BindSwapBarrierSGIX = _real_glXBindSwapBarrierSGIX; + glx.QueryMaxSwapBarriersSGIX = _real_glXQueryMaxSwapBarriersSGIX; + + /*** GLX_SUN_get_transparent_index ***/ + glx.GetTransparentIndexSUN = _real_glXGetTransparentIndexSUN; + + /*** GLX_MESA_copy_sub_buffer ***/ + glx.CopySubBufferMESA = _real_glXCopySubBufferMESA; + + /*** GLX_MESA_release_buffers ***/ + glx.ReleaseBuffersMESA = _real_glXReleaseBuffersMESA; + + /*** GLX_MESA_pixmap_colormap ***/ + glx.CreateGLXPixmapMESA = _real_glXCreateGLXPixmapMESA; + + /*** GLX_MESA_set_3dfx_mode ***/ + glx.Set3DfxModeMESA = _real_glXSet3DfxModeMESA; + + /*** GLX_NV_vertex_array_range ***/ + glx.AllocateMemoryNV = _real_glXAllocateMemoryNV; + glx.FreeMemoryNV = _real_glXFreeMemoryNV; + + /*** GLX_MESA_agp_offset ***/ + glx.GetAGPOffsetMESA = _real_glXGetAGPOffsetMESA; + + return &glx; +} diff --git a/src/gallium/winsys/xlib/realglx.h b/src/gallium/winsys/xlib/realglx.h new file mode 100644 index 0000000000..150129db68 --- /dev/null +++ b/src/gallium/winsys/xlib/realglx.h @@ -0,0 +1,326 @@ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef REALGLX_H +#define REALGLX_H + + +extern struct _glxapi_table * +_real_GetGLXDispatchTable(void); + + +/* + * Basically just need these to prevent compiler warnings. + */ + + +extern XVisualInfo * +_real_glXChooseVisual( Display *dpy, int screen, int *list ); + +extern GLXContext +_real_glXCreateContext( Display *dpy, XVisualInfo *visinfo, + GLXContext share_list, Bool direct ); + +extern GLXPixmap +_real_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ); + +extern GLXPixmap +_real_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, + Pixmap pixmap, Colormap cmap ); + +extern void +_real_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap ); + +extern void +_real_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, + unsigned long mask ); + +extern Bool +_real_glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx ); + +extern Bool +_real_glXQueryExtension( Display *dpy, int *errorb, int *event ); + +extern void +_real_glXDestroyContext( Display *dpy, GLXContext ctx ); + +extern Bool +_real_glXIsDirect( Display *dpy, GLXContext ctx ); + +extern void +_real_glXSwapBuffers( Display *dpy, GLXDrawable drawable ); + +extern void +_real_glXUseXFont( Font font, int first, int count, int listbase ); + +extern Bool +_real_glXQueryVersion( Display *dpy, int *maj, int *min ); + +extern int +_real_glXGetConfig( Display *dpy, XVisualInfo *visinfo, + int attrib, int *value ); + +extern void +_real_glXWaitGL( void ); + + +extern void +_real_glXWaitX( void ); + +/* GLX 1.1 and later */ +extern const char * +_real_glXQueryExtensionsString( Display *dpy, int screen ); + +/* GLX 1.1 and later */ +extern const char * +_real_glXQueryServerString( Display *dpy, int screen, int name ); + +/* GLX 1.1 and later */ +extern const char * +_real_glXGetClientString( Display *dpy, int name ); + + +/* + * GLX 1.3 and later + */ + +extern GLXFBConfig * +_real_glXChooseFBConfig( Display *dpy, int screen, + const int *attribList, int *nitems ); + +extern int +_real_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config, + int attribute, int *value ); + +extern GLXFBConfig * +_real_glXGetFBConfigs( Display *dpy, int screen, int *nelements ); + +extern XVisualInfo * +_real_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config ); + +extern GLXWindow +_real_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win, + const int *attribList ); + +extern void +_real_glXDestroyWindow( Display *dpy, GLXWindow window ); + +extern GLXPixmap +_real_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, + const int *attribList ); + +extern void +_real_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap ); + +extern GLXPbuffer +_real_glXCreatePbuffer( Display *dpy, GLXFBConfig config, + const int *attribList ); + +extern void +_real_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf ); + +extern void +_real_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, + unsigned int *value ); + +extern GLXContext +_real_glXCreateNewContext( Display *dpy, GLXFBConfig config, + int renderType, GLXContext shareList, Bool direct ); + + +extern Bool +_real_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, + GLXDrawable read, GLXContext ctx ); + +extern int +_real_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value ); + +extern void +_real_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask ); + +extern void +_real_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable, + unsigned long *mask ); + +#ifdef GLX_SGI_swap_control +extern int +_real_glXSwapIntervalSGI(int interval); +#endif + + +#ifdef GLX_SGI_video_sync +extern int +_real_glXGetVideoSyncSGI(unsigned int *count); + +extern int +_real_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count); +#endif + + +#ifdef GLX_SGI_make_current_read +extern Bool +_real_glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx); + +extern GLXDrawable +_real_glXGetCurrentReadDrawableSGI(void); +#endif + +#if defined(_VL_H) && defined(GLX_SGIX_video_source) +extern GLXVideoSourceSGIX +_real_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode); + +extern void +_real_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src); +#endif + +#ifdef GLX_EXT_import_context +extern void +_real_glXFreeContextEXT(Display *dpy, GLXContext context); + +extern GLXContextID +_real_glXGetContextIDEXT(const GLXContext context); + +extern Display * +_real_glXGetCurrentDisplayEXT(void); + +extern GLXContext +_real_glXImportContextEXT(Display *dpy, GLXContextID contextID); + +extern int +_real_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value); +#endif + +#ifdef GLX_SGIX_fbconfig +extern int +_real_glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value); + +extern GLXFBConfigSGIX * +_real_glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements); + +extern GLXPixmap +_real_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap); + +extern GLXContext +_real_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct); + +extern XVisualInfo * +_real_glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config); + +extern GLXFBConfigSGIX +_real_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis); +#endif + +#ifdef GLX_SGIX_pbuffer +extern GLXPbufferSGIX +_real_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attrib_list); + +extern void +_real_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf); + +extern int +_real_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value); + +extern void +_real_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask); + +extern void +_real_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask); +#endif + +#ifdef GLX_SGI_cushion +extern void +_real_glXCushionSGI(Display *dpy, Window win, float cushion); +#endif + +#ifdef GLX_SGIX_video_resize +extern int +_real_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window); + +extern int +_real_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h); + +extern int +_real_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h); + +extern int +_real_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh); + +extern int +_real_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype); +#endif + +#if defined(_DM_BUFFER_H_) && defined(GLX_SGIX_dmbuffer) +extern Bool +_real_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer); +#endif + +#ifdef GLX_SGIX_swap_group +extern void +_real_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member); +#endif + +#ifdef GLX_SGIX_swap_barrier +extern void +_real_glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier); + +extern Bool +_real_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max); +#endif + +#ifdef GLX_SUN_get_transparent_index +extern Status +_real_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent); +#endif + +#ifdef GLX_MESA_release_buffers +extern Bool +_real_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d ); +#endif + +#ifdef GLX_MESA_set_3dfx_mode +extern Bool +_real_glXSet3DfxModeMESA( int mode ); +#endif + +#ifdef GLX_NV_vertex_array_range +extern void * +_real_glXAllocateMemoryNV(GLsizei size, GLfloat readfreq, GLfloat writefreq, GLfloat priority); +extern void +_real_glXFreeMemoryNV(GLvoid *pointer); +#endif + +#ifdef GLX_MESA_agp_offset +extern GLuint +_real_glXGetAGPOffsetMESA(const GLvoid *pointer); +#endif + +#ifdef GLX_MESA_copy_sub_buffer +extern void +_real_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable, + int x, int y, int width, int height ); +#endif + +#endif /* REALGLX_H */ diff --git a/src/gallium/winsys/xlib/xfonts.c b/src/gallium/winsys/xlib/xfonts.c new file mode 100644 index 0000000000..d72c600bd1 --- /dev/null +++ b/src/gallium/winsys/xlib/xfonts.c @@ -0,0 +1,377 @@ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2000 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* xfonts.c -- glXUseXFont() for Mesa written by + * Copyright (C) 1995 Thorsten.Ohl @ Physik.TH-Darmstadt.de + */ + +#ifdef __VMS +#include +#endif + +#include "glxheader.h" +#include "context.h" +#include "imports.h" +#include "xfonts.h" + + +/* Some debugging info. */ + +#ifdef DEBUG +#undef _R +#undef _G +#undef _B +#include + +int debug_xfonts = 0; + +static void +dump_char_struct(XCharStruct * ch, char *prefix) +{ + printf("%slbearing = %d, rbearing = %d, width = %d\n", + prefix, ch->lbearing, ch->rbearing, ch->width); + printf("%sascent = %d, descent = %d, attributes = %u\n", + prefix, ch->ascent, ch->descent, (unsigned int) ch->attributes); +} + +static void +dump_font_struct(XFontStruct * font) +{ + printf("ascent = %d, descent = %d\n", font->ascent, font->descent); + printf("char_or_byte2 = (%u,%u)\n", + font->min_char_or_byte2, font->max_char_or_byte2); + printf("byte1 = (%u,%u)\n", font->min_byte1, font->max_byte1); + printf("all_chars_exist = %s\n", font->all_chars_exist ? "True" : "False"); + printf("default_char = %c (\\%03o)\n", + (char) (isprint(font->default_char) ? font->default_char : ' '), + font->default_char); + dump_char_struct(&font->min_bounds, "min> "); + dump_char_struct(&font->max_bounds, "max> "); +#if 0 + for (c = font->min_char_or_byte2; c <= font->max_char_or_byte2; c++) { + char prefix[8]; + sprintf(prefix, "%d> ", c); + dump_char_struct(&font->per_char[c], prefix); + } +#endif +} + +static void +dump_bitmap(unsigned int width, unsigned int height, GLubyte * bitmap) +{ + unsigned int x, y; + + printf(" "); + for (x = 0; x < 8 * width; x++) + printf("%o", 7 - (x % 8)); + putchar('\n'); + for (y = 0; y < height; y++) { + printf("%3o:", y); + for (x = 0; x < 8 * width; x++) + putchar((bitmap[width * (height - y - 1) + x / 8] & (1 << (7 - (x % + 8)))) + ? '*' : '.'); + printf(" "); + for (x = 0; x < width; x++) + printf("0x%02x, ", bitmap[width * (height - y - 1) + x]); + putchar('\n'); + } +} +#endif /* DEBUG */ + + +/* Implementation. */ + +/* Fill a BITMAP with a character C from thew current font + in the graphics context GC. WIDTH is the width in bytes + and HEIGHT is the height in bits. + + Note that the generated bitmaps must be used with + + glPixelStorei (GL_UNPACK_SWAP_BYTES, GL_FALSE); + glPixelStorei (GL_UNPACK_LSB_FIRST, GL_FALSE); + glPixelStorei (GL_UNPACK_ROW_LENGTH, 0); + glPixelStorei (GL_UNPACK_SKIP_ROWS, 0); + glPixelStorei (GL_UNPACK_SKIP_PIXELS, 0); + glPixelStorei (GL_UNPACK_ALIGNMENT, 1); + + Possible optimizations: + + * use only one reusable pixmap with the maximum dimensions. + * draw the entire font into a single pixmap (careful with + proportional fonts!). +*/ + + +/* + * Generate OpenGL-compatible bitmap. + */ +static void +fill_bitmap(Display * dpy, Window win, GC gc, + unsigned int width, unsigned int height, + int x0, int y0, unsigned int c, GLubyte * bitmap) +{ + XImage *image; + unsigned int x, y; + Pixmap pixmap; + XChar2b char2b; + + pixmap = XCreatePixmap(dpy, win, 8 * width, height, 1); + XSetForeground(dpy, gc, 0); + XFillRectangle(dpy, pixmap, gc, 0, 0, 8 * width, height); + XSetForeground(dpy, gc, 1); + + char2b.byte1 = (c >> 8) & 0xff; + char2b.byte2 = (c & 0xff); + + XDrawString16(dpy, pixmap, gc, x0, y0, &char2b, 1); + + image = XGetImage(dpy, pixmap, 0, 0, 8 * width, height, 1, XYPixmap); + if (image) { + /* Fill the bitmap (X11 and OpenGL are upside down wrt each other). */ + for (y = 0; y < height; y++) + for (x = 0; x < 8 * width; x++) + if (XGetPixel(image, x, y)) + bitmap[width * (height - y - 1) + x / 8] |= + (1 << (7 - (x % 8))); + XDestroyImage(image); + } + + XFreePixmap(dpy, pixmap); +} + +/* + * determine if a given glyph is valid and return the + * corresponding XCharStruct. + */ +static XCharStruct * +isvalid(XFontStruct * fs, unsigned int which) +{ + unsigned int rows, pages; + unsigned int byte1 = 0, byte2 = 0; + int i, valid = 1; + + rows = fs->max_byte1 - fs->min_byte1 + 1; + pages = fs->max_char_or_byte2 - fs->min_char_or_byte2 + 1; + + if (rows == 1) { + /* "linear" fonts */ + if ((fs->min_char_or_byte2 > which) || (fs->max_char_or_byte2 < which)) + valid = 0; + } + else { + /* "matrix" fonts */ + byte2 = which & 0xff; + byte1 = which >> 8; + if ((fs->min_char_or_byte2 > byte2) || + (fs->max_char_or_byte2 < byte2) || + (fs->min_byte1 > byte1) || (fs->max_byte1 < byte1)) + valid = 0; + } + + if (valid) { + if (fs->per_char) { + if (rows == 1) { + /* "linear" fonts */ + return (fs->per_char + (which - fs->min_char_or_byte2)); + } + else { + /* "matrix" fonts */ + i = ((byte1 - fs->min_byte1) * pages) + + (byte2 - fs->min_char_or_byte2); + return (fs->per_char + i); + } + } + else { + return (&fs->min_bounds); + } + } + return (NULL); +} + + +void +Fake_glXUseXFont(Font font, int first, int count, int listbase) +{ + Display *dpy; + Window win; + Pixmap pixmap; + GC gc; + XGCValues values; + unsigned long valuemask; + XFontStruct *fs; + GLint swapbytes, lsbfirst, rowlength; + GLint skiprows, skippixels, alignment; + unsigned int max_width, max_height, max_bm_width, max_bm_height; + GLubyte *bm; + int i; + + dpy = glXGetCurrentDisplay(); + if (!dpy) + return; /* I guess glXMakeCurrent wasn't called */ + win = RootWindow(dpy, DefaultScreen(dpy)); + + fs = XQueryFont(dpy, font); + if (!fs) { + _mesa_error(NULL, GL_INVALID_VALUE, + "Couldn't get font structure information"); + return; + } + + /* Allocate a bitmap that can fit all characters. */ + max_width = fs->max_bounds.rbearing - fs->min_bounds.lbearing; + max_height = fs->max_bounds.ascent + fs->max_bounds.descent; + max_bm_width = (max_width + 7) / 8; + max_bm_height = max_height; + + bm = (GLubyte *) MALLOC((max_bm_width * max_bm_height) * sizeof(GLubyte)); + if (!bm) { + XFreeFontInfo(NULL, fs, 1); + _mesa_error(NULL, GL_OUT_OF_MEMORY, + "Couldn't allocate bitmap in glXUseXFont()"); + return; + } + +#if 0 + /* get the page info */ + pages = fs->max_char_or_byte2 - fs->min_char_or_byte2 + 1; + firstchar = (fs->min_byte1 << 8) + fs->min_char_or_byte2; + lastchar = (fs->max_byte1 << 8) + fs->max_char_or_byte2; + rows = fs->max_byte1 - fs->min_byte1 + 1; + unsigned int first_char, last_char, pages, rows; +#endif + + /* Save the current packing mode for bitmaps. */ + glGetIntegerv(GL_UNPACK_SWAP_BYTES, &swapbytes); + glGetIntegerv(GL_UNPACK_LSB_FIRST, &lsbfirst); + glGetIntegerv(GL_UNPACK_ROW_LENGTH, &rowlength); + glGetIntegerv(GL_UNPACK_SKIP_ROWS, &skiprows); + glGetIntegerv(GL_UNPACK_SKIP_PIXELS, &skippixels); + glGetIntegerv(GL_UNPACK_ALIGNMENT, &alignment); + + /* Enforce a standard packing mode which is compatible with + fill_bitmap() from above. This is actually the default mode, + except for the (non)alignment. */ + glPixelStorei(GL_UNPACK_SWAP_BYTES, GL_FALSE); + glPixelStorei(GL_UNPACK_LSB_FIRST, GL_FALSE); + glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); + glPixelStorei(GL_UNPACK_SKIP_ROWS, 0); + glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0); + glPixelStorei(GL_UNPACK_ALIGNMENT, 1); + + pixmap = XCreatePixmap(dpy, win, 10, 10, 1); + values.foreground = BlackPixel(dpy, DefaultScreen(dpy)); + values.background = WhitePixel(dpy, DefaultScreen(dpy)); + values.font = fs->fid; + valuemask = GCForeground | GCBackground | GCFont; + gc = XCreateGC(dpy, pixmap, valuemask, &values); + XFreePixmap(dpy, pixmap); + +#ifdef DEBUG + if (debug_xfonts) + dump_font_struct(fs); +#endif + + for (i = 0; i < count; i++) { + unsigned int width, height, bm_width, bm_height; + GLfloat x0, y0, dx, dy; + XCharStruct *ch; + int x, y; + unsigned int c = first + i; + int list = listbase + i; + int valid; + + /* check on index validity and get the bounds */ + ch = isvalid(fs, c); + if (!ch) { + ch = &fs->max_bounds; + valid = 0; + } + else { + valid = 1; + } + +#ifdef DEBUG + if (debug_xfonts) { + char s[7]; + sprintf(s, isprint(c) ? "%c> " : "\\%03o> ", c); + dump_char_struct(ch, s); + } +#endif + + /* glBitmap()' parameters: + straight from the glXUseXFont(3) manpage. */ + width = ch->rbearing - ch->lbearing; + height = ch->ascent + ch->descent; + x0 = -ch->lbearing; + y0 = ch->descent - 0; /* XXX used to subtract 1 here */ + /* but that caused a conformace failure */ + dx = ch->width; + dy = 0; + + /* X11's starting point. */ + x = -ch->lbearing; + y = ch->ascent; + + /* Round the width to a multiple of eight. We will use this also + for the pixmap for capturing the X11 font. This is slightly + inefficient, but it makes the OpenGL part real easy. */ + bm_width = (width + 7) / 8; + bm_height = height; + + glNewList(list, GL_COMPILE); + if (valid && (bm_width > 0) && (bm_height > 0)) { + + MEMSET(bm, '\0', bm_width * bm_height); + fill_bitmap(dpy, win, gc, bm_width, bm_height, x, y, c, bm); + + glBitmap(width, height, x0, y0, dx, dy, bm); +#ifdef DEBUG + if (debug_xfonts) { + printf("width/height = %u/%u\n", width, height); + printf("bm_width/bm_height = %u/%u\n", bm_width, bm_height); + dump_bitmap(bm_width, bm_height, bm); + } +#endif + } + else { + glBitmap(0, 0, 0.0, 0.0, dx, dy, NULL); + } + glEndList(); + } + + FREE(bm); + XFreeFontInfo(NULL, fs, 1); + XFreeGC(dpy, gc); + + /* Restore saved packing modes. */ + glPixelStorei(GL_UNPACK_SWAP_BYTES, swapbytes); + glPixelStorei(GL_UNPACK_LSB_FIRST, lsbfirst); + glPixelStorei(GL_UNPACK_ROW_LENGTH, rowlength); + glPixelStorei(GL_UNPACK_SKIP_ROWS, skiprows); + glPixelStorei(GL_UNPACK_SKIP_PIXELS, skippixels); + glPixelStorei(GL_UNPACK_ALIGNMENT, alignment); +} diff --git a/src/gallium/winsys/xlib/xfonts.h b/src/gallium/winsys/xlib/xfonts.h new file mode 100644 index 0000000000..e36f42f817 --- /dev/null +++ b/src/gallium/winsys/xlib/xfonts.h @@ -0,0 +1,41 @@ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2000 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef XFONTS_H +#define XFONTS_H + +#ifdef __VMS +#include +#endif + +#include + + +extern void Fake_glXUseXFont( Font font, int first, int count, int listbase ); + + +#endif + diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c new file mode 100644 index 0000000000..e5fef1d7a8 --- /dev/null +++ b/src/gallium/winsys/xlib/xm_api.c @@ -0,0 +1,1380 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file xm_api.c + * + * All the XMesa* API functions. + * + * + * NOTES: + * + * The window coordinate system origin (0,0) is in the lower-left corner + * of the window. X11's window coordinate origin is in the upper-left + * corner of the window. Therefore, most drawing functions in this + * file have to flip Y coordinates. + * + * Define USE_XSHM in the Makefile with -DUSE_XSHM if you want to compile + * in support for the MIT Shared Memory extension. If enabled, when you + * use an Ximage for the back buffer in double buffered mode, the "swap" + * operation will be faster. You must also link with -lXext. + * + * Byte swapping: If the Mesa host and the X display use a different + * byte order then there's some trickiness to be aware of when using + * XImages. The byte ordering used for the XImage is that of the X + * display, not the Mesa host. + * The color-to-pixel encoding for True/DirectColor must be done + * according to the display's visual red_mask, green_mask, and blue_mask. + * If XPutPixel is used to put a pixel into an XImage then XPutPixel will + * do byte swapping if needed. If one wants to directly "poke" the pixel + * into the XImage's buffer then the pixel must be byte swapped first. + * + */ + +#ifdef __CYGWIN__ +#undef WIN32 +#undef __WIN32__ +#endif + +#include "glxheader.h" +#include "GL/xmesa.h" +#include "xmesaP.h" +#include "main/context.h" +#include "main/framebuffer.h" +#include "glapi/glthread.h" + +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_context.h" + +#include "xm_winsys_aub.h" + +/** + * Global X driver lock + */ +_glthread_Mutex _xmesa_lock; + + +int xmesa_mode; + + +/**********************************************************************/ +/***** X Utility Functions *****/ +/**********************************************************************/ + + +/** + * Return the host's byte order as LSBFirst or MSBFirst ala X. + */ +#ifndef XFree86Server +static int host_byte_order( void ) +{ + int i = 1; + char *cptr = (char *) &i; + return (*cptr==1) ? LSBFirst : MSBFirst; +} +#endif + + +/** + * Check if the X Shared Memory extension is available. + * Return: 0 = not available + * 1 = shared XImage support available + * 2 = shared Pixmap support available also + */ +static int check_for_xshm( XMesaDisplay *display ) +{ +#if defined(USE_XSHM) && !defined(XFree86Server) + int major, minor, ignore; + Bool pixmaps; + + if (XQueryExtension( display, "MIT-SHM", &ignore, &ignore, &ignore )) { + if (XShmQueryVersion( display, &major, &minor, &pixmaps )==True) { + return (pixmaps==True) ? 2 : 1; + } + else { + return 0; + } + } + else { + return 0; + } +#else + /* No XSHM support */ + return 0; +#endif +} + + +/** + * Return the true number of bits per pixel for XImages. + * For example, if we request a 24-bit deep visual we may actually need/get + * 32bpp XImages. This function returns the appropriate bpp. + * Input: dpy - the X display + * visinfo - desribes the visual to be used for XImages + * Return: true number of bits per pixel for XImages + */ +static int +bits_per_pixel( XMesaVisual xmv ) +{ +#ifdef XFree86Server + const int depth = xmv->nplanes; + int i; + assert(depth > 0); + for (i = 0; i < screenInfo.numPixmapFormats; i++) { + if (screenInfo.formats[i].depth == depth) + return screenInfo.formats[i].bitsPerPixel; + } + return depth; /* should never get here, but this should be safe */ +#else + XMesaDisplay *dpy = xmv->display; + XMesaVisualInfo visinfo = xmv->visinfo; + XMesaImage *img; + int bitsPerPixel; + /* Create a temporary XImage */ + img = XCreateImage( dpy, visinfo->visual, visinfo->depth, + ZPixmap, 0, /*format, offset*/ + (char*) MALLOC(8), /*data*/ + 1, 1, /*width, height*/ + 32, /*bitmap_pad*/ + 0 /*bytes_per_line*/ + ); + assert(img); + /* grab the bits/pixel value */ + bitsPerPixel = img->bits_per_pixel; + /* free the XImage */ + _mesa_free( img->data ); + img->data = NULL; + XMesaDestroyImage( img ); + return bitsPerPixel; +#endif +} + + + +/* + * Determine if a given X window ID is valid (window exists). + * Do this by calling XGetWindowAttributes() for the window and + * checking if we catch an X error. + * Input: dpy - the display + * win - the window to check for existance + * Return: GL_TRUE - window exists + * GL_FALSE - window doesn't exist + */ +#ifndef XFree86Server +static GLboolean WindowExistsFlag; + +static int window_exists_err_handler( XMesaDisplay* dpy, XErrorEvent* xerr ) +{ + (void) dpy; + if (xerr->error_code == BadWindow) { + WindowExistsFlag = GL_FALSE; + } + return 0; +} + +static GLboolean window_exists( XMesaDisplay *dpy, Window win ) +{ + XWindowAttributes wa; + int (*old_handler)( XMesaDisplay*, XErrorEvent* ); + WindowExistsFlag = GL_TRUE; + old_handler = XSetErrorHandler(window_exists_err_handler); + XGetWindowAttributes( dpy, win, &wa ); /* dummy request */ + XSetErrorHandler(old_handler); + return WindowExistsFlag; +} + +static Status +get_drawable_size( XMesaDisplay *dpy, Drawable d, uint *width, uint *height ) +{ + Window root; + Status stat; + int xpos, ypos; + unsigned int w, h, bw, depth; + stat = XGetGeometry(dpy, d, &root, &xpos, &ypos, &w, &h, &bw, &depth); + *width = w; + *height = h; + return stat; +} +#endif + + +/** + * Return the size of the window (or pixmap) that corresponds to the + * given XMesaBuffer. + * \param width returns width in pixels + * \param height returns height in pixels + */ +static void +xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b, + GLuint *width, GLuint *height) +{ +#ifdef XFree86Server + *width = MIN2(b->drawable->width, MAX_WIDTH); + *height = MIN2(b->drawable->height, MAX_HEIGHT); +#else + Status stat; + + _glthread_LOCK_MUTEX(_xmesa_lock); + XSync(b->xm_visual->display, 0); /* added for Chromium */ + stat = get_drawable_size(dpy, b->drawable, width, height); + _glthread_UNLOCK_MUTEX(_xmesa_lock); + + if (!stat) { + /* probably querying a window that's recently been destroyed */ + _mesa_warning(NULL, "XGetGeometry failed!\n"); + *width = *height = 1; + } +#endif +} + + +/** + * Choose the pixel format for the given visual. + * This will tell the gallium driver how to pack pixel data into + * drawing surfaces. + */ +static GLuint +choose_pixel_format(XMesaVisual v) +{ + if ( GET_REDMASK(v) == 0x0000ff + && GET_GREENMASK(v) == 0x00ff00 + && GET_BLUEMASK(v) == 0xff0000 + && v->BitsPerPixel == 32) { + if (CHECK_BYTE_ORDER(v)) { + /* no byteswapping needed */ + return 0 /* PIXEL_FORMAT_U_A8_B8_G8_R8 */; + } + else { + return PIPE_FORMAT_R8G8B8A8_UNORM; + } + } + else if ( GET_REDMASK(v) == 0xff0000 + && GET_GREENMASK(v) == 0x00ff00 + && GET_BLUEMASK(v) == 0x0000ff + && v->BitsPerPixel == 32) { + if (CHECK_BYTE_ORDER(v)) { + /* no byteswapping needed */ + return PIPE_FORMAT_A8R8G8B8_UNORM; + } + else { + return PIPE_FORMAT_B8G8R8A8_UNORM; + } + } + else if ( GET_REDMASK(v) == 0xf800 + && GET_GREENMASK(v) == 0x07e0 + && GET_BLUEMASK(v) == 0x001f + && CHECK_BYTE_ORDER(v) + && v->BitsPerPixel == 16) { + /* 5-6-5 RGB */ + return PIPE_FORMAT_R5G6B5_UNORM; + } + + assert(0); + return 0; +} + + + +/**********************************************************************/ +/***** Linked list of XMesaBuffers *****/ +/**********************************************************************/ + +XMesaBuffer XMesaBufferList = NULL; + + +/** + * Allocate a new XMesaBuffer object which corresponds to the given drawable. + * Note that XMesaBuffer is derived from GLframebuffer. + * The new XMesaBuffer will not have any size (Width=Height=0). + * + * \param d the corresponding X drawable (window or pixmap) + * \param type either WINDOW, PIXMAP or PBUFFER, describing d + * \param vis the buffer's visual + * \param cmap the window's colormap, if known. + * \return new XMesaBuffer or NULL if any problem + */ +static XMesaBuffer +create_xmesa_buffer(XMesaDrawable d, BufferType type, + XMesaVisual vis, XMesaColormap cmap) +{ + XMesaBuffer b; + GLframebuffer *fb; + enum pipe_format colorFormat, depthFormat, stencilFormat; + uint width, height; + + ASSERT(type == WINDOW || type == PIXMAP || type == PBUFFER); + + b = (XMesaBuffer) CALLOC_STRUCT(xmesa_buffer); + if (!b) + return NULL; + + b->drawable = d; + + b->xm_visual = vis; + b->type = type; + b->cmap = cmap; + + /* determine PIPE_FORMATs for buffers */ + colorFormat = choose_pixel_format(vis); + + if (vis->mesa_visual.depthBits == 0) + depthFormat = PIPE_FORMAT_NONE; + else if (vis->mesa_visual.depthBits <= 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (vis->mesa_visual.depthBits <= 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_Z32_UNORM; + + if (vis->mesa_visual.stencilBits == 8) { + if (depthFormat == PIPE_FORMAT_S8Z24_UNORM) + stencilFormat = depthFormat; + else + stencilFormat = PIPE_FORMAT_S8_UNORM; + } + else { + stencilFormat = PIPE_FORMAT_NONE; + } + + + get_drawable_size(vis->display, d, &width, &height); + + /* + * Create framebuffer, but we'll plug in our own renderbuffers below. + */ + b->stfb = st_create_framebuffer(&vis->mesa_visual, + colorFormat, depthFormat, stencilFormat, + width, height, + (void *) b); + fb = &b->stfb->Base; + + /* + * Create scratch XImage for xmesa_display_surface() + */ + b->tempImage = XCreateImage(vis->display, + vis->visinfo->visual, + vis->visinfo->depth, + ZPixmap, 0, /* format, offset */ + NULL, /* data */ + 0, 0, /* size */ + 32, /* bitmap_pad */ + 0); /* bytes_per_line */ + + /* GLX_EXT_texture_from_pixmap */ + b->TextureTarget = 0; + b->TextureFormat = GLX_TEXTURE_FORMAT_NONE_EXT; + b->TextureMipmap = 0; + + /* insert buffer into linked list */ + b->Next = XMesaBufferList; + XMesaBufferList = b; + + return b; +} + + +/** + * Find an XMesaBuffer by matching X display and colormap but NOT matching + * the notThis buffer. + */ +XMesaBuffer +xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis) +{ + XMesaBuffer b; + for (b = XMesaBufferList; b; b = b->Next) { + if (b->xm_visual->display == dpy && + b->cmap == cmap && + b != notThis) { + return b; + } + } + return NULL; +} + + +/** + * Remove buffer from linked list, delete if no longer referenced. + */ +static void +xmesa_free_buffer(XMesaBuffer buffer) +{ + XMesaBuffer prev = NULL, b; + + for (b = XMesaBufferList; b; b = b->Next) { + if (b == buffer) { + struct gl_framebuffer *fb = &buffer->stfb->Base; + + /* unlink buffer from list */ + if (prev) + prev->Next = buffer->Next; + else + XMesaBufferList = buffer->Next; + + /* mark as delete pending */ + fb->DeletePending = GL_TRUE; + + /* Since the X window for the XMesaBuffer is going away, we don't + * want to dereference this pointer in the future. + */ + b->drawable = 0; + + buffer->tempImage->data = NULL; + XDestroyImage(buffer->tempImage); + + /* Unreference. If count = zero we'll really delete the buffer */ + _mesa_unreference_framebuffer(&fb); + + XFreeGC(b->xm_visual->display, b->gc); + + free(buffer); + + return; + } + /* continue search */ + prev = b; + } + /* buffer not found in XMesaBufferList */ + _mesa_problem(NULL,"xmesa_free_buffer() - buffer not found\n"); +} + + + +/**********************************************************************/ +/***** Misc Private Functions *****/ +/**********************************************************************/ + + +/** + * When a context is bound for the first time, we can finally finish + * initializing the context's visual and buffer information. + * \param v the XMesaVisual to initialize + * \param b the XMesaBuffer to initialize (may be NULL) + * \param rgb_flag TRUE = RGBA mode, FALSE = color index mode + * \param window the window/pixmap we're rendering into + * \param cmap the colormap associated with the window/pixmap + * \return GL_TRUE=success, GL_FALSE=failure + */ +static GLboolean +initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, + GLboolean rgb_flag, XMesaDrawable window, + XMesaColormap cmap) +{ +#ifdef XFree86Server + int client = (window) ? CLIENT_ID(window->id) : 0; +#endif + + ASSERT(!b || b->xm_visual == v); + + /* Save true bits/pixel */ + v->BitsPerPixel = bits_per_pixel(v); + assert(v->BitsPerPixel > 0); + + if (rgb_flag == GL_FALSE) { + /* COLOR-INDEXED WINDOW: not supported*/ + return GL_FALSE; + } + else { + /* RGB WINDOW: + * We support RGB rendering into almost any kind of visual. + */ + const int xclass = v->mesa_visual.visualType; + if (xclass != GLX_TRUE_COLOR && xclass == !GLX_DIRECT_COLOR) { + _mesa_warning(NULL, + "XMesa: RGB mode rendering not supported in given visual.\n"); + return GL_FALSE; + } + v->mesa_visual.indexBits = 0; + } + + /* + * If MESA_INFO env var is set print out some debugging info + * which can help Brian figure out what's going on when a user + * reports bugs. + */ + if (_mesa_getenv("MESA_INFO")) { + _mesa_printf("X/Mesa visual = %p\n", (void *) v); + _mesa_printf("X/Mesa level = %d\n", v->mesa_visual.level); + _mesa_printf("X/Mesa depth = %d\n", GET_VISUAL_DEPTH(v)); + _mesa_printf("X/Mesa bits per pixel = %d\n", v->BitsPerPixel); + } + + if (b && window) { + /* these should have been set in create_xmesa_buffer */ + ASSERT(b->drawable == window); + + /* Setup for single/double buffering */ + if (v->mesa_visual.doubleBufferMode) { + /* Double buffered */ + b->shm = check_for_xshm( v->display ); + } + + /* X11 graphics context */ +#ifdef XFree86Server + b->gc = CreateScratchGC(v->display, window->depth); +#else + b->gc = XCreateGC( v->display, window, 0, NULL ); +#endif + XMesaSetFunction( v->display, b->gc, GXcopy ); + } + + return GL_TRUE; +} + + + +#define NUM_VISUAL_TYPES 6 + +/** + * Convert an X visual type to a GLX visual type. + * + * \param visualType X visual type (i.e., \c TrueColor, \c StaticGray, etc.) + * to be converted. + * \return If \c visualType is a valid X visual type, a GLX visual type will + * be returned. Otherwise \c GLX_NONE will be returned. + * + * \note + * This code was lifted directly from lib/GL/glx/glcontextmodes.c in the + * DRI CVS tree. + */ +static GLint +xmesa_convert_from_x_visual_type( int visualType ) +{ + static const int glx_visual_types[ NUM_VISUAL_TYPES ] = { + GLX_STATIC_GRAY, GLX_GRAY_SCALE, + GLX_STATIC_COLOR, GLX_PSEUDO_COLOR, + GLX_TRUE_COLOR, GLX_DIRECT_COLOR + }; + + return ( (unsigned) visualType < NUM_VISUAL_TYPES ) + ? glx_visual_types[ visualType ] : GLX_NONE; +} + + +/**********************************************************************/ +/***** Public Functions *****/ +/**********************************************************************/ + + +/* + * Create a new X/Mesa visual. + * Input: display - X11 display + * visinfo - an XVisualInfo pointer + * rgb_flag - GL_TRUE = RGB mode, + * GL_FALSE = color index mode + * alpha_flag - alpha buffer requested? + * db_flag - GL_TRUE = double-buffered, + * GL_FALSE = single buffered + * stereo_flag - stereo visual? + * ximage_flag - GL_TRUE = use an XImage for back buffer, + * GL_FALSE = use an off-screen pixmap for back buffer + * depth_size - requested bits/depth values, or zero + * stencil_size - requested bits/stencil values, or zero + * accum_red_size - requested bits/red accum values, or zero + * accum_green_size - requested bits/green accum values, or zero + * accum_blue_size - requested bits/blue accum values, or zero + * accum_alpha_size - requested bits/alpha accum values, or zero + * num_samples - number of samples/pixel if multisampling, or zero + * level - visual level, usually 0 + * visualCaveat - ala the GLX extension, usually GLX_NONE + * Return; a new XMesaVisual or 0 if error. + */ +PUBLIC +XMesaVisual XMesaCreateVisual( XMesaDisplay *display, + XMesaVisualInfo visinfo, + GLboolean rgb_flag, + GLboolean alpha_flag, + GLboolean db_flag, + GLboolean stereo_flag, + GLboolean ximage_flag, + GLint depth_size, + GLint stencil_size, + GLint accum_red_size, + GLint accum_green_size, + GLint accum_blue_size, + GLint accum_alpha_size, + GLint num_samples, + GLint level, + GLint visualCaveat ) +{ + XMesaVisual v; + GLint red_bits, green_bits, blue_bits, alpha_bits; + +#ifndef XFree86Server + /* For debugging only */ + if (_mesa_getenv("MESA_XSYNC")) { + /* This makes debugging X easier. + * In your debugger, set a breakpoint on _XError to stop when an + * X protocol error is generated. + */ + XSynchronize( display, 1 ); + } +#endif + + v = (XMesaVisual) CALLOC_STRUCT(xmesa_visual); + if (!v) { + return NULL; + } + + v->display = display; + + /* Save a copy of the XVisualInfo struct because the user may X_mesa_free() + * the struct but we may need some of the information contained in it + * at a later time. + */ +#ifndef XFree86Server + v->visinfo = (XVisualInfo *) MALLOC(sizeof(*visinfo)); + if(!v->visinfo) { + _mesa_free(v); + return NULL; + } + MEMCPY(v->visinfo, visinfo, sizeof(*visinfo)); +#endif + + v->ximage_flag = ximage_flag; + +#ifdef XFree86Server + /* We could calculate these values by ourselves. nplanes is either the sum + * of the red, green, and blue bits or the number index bits. + * ColormapEntries is either (1U << index_bits) or + * (1U << max(redBits, greenBits, blueBits)). + */ + assert(visinfo->nplanes > 0); + v->nplanes = visinfo->nplanes; + v->ColormapEntries = visinfo->ColormapEntries; + + v->mesa_visual.redMask = visinfo->redMask; + v->mesa_visual.greenMask = visinfo->greenMask; + v->mesa_visual.blueMask = visinfo->blueMask; + v->mesa_visual.visualID = visinfo->vid; + v->mesa_visual.screen = 0; /* FIXME: What should be done here? */ +#else + v->mesa_visual.redMask = visinfo->red_mask; + v->mesa_visual.greenMask = visinfo->green_mask; + v->mesa_visual.blueMask = visinfo->blue_mask; + v->mesa_visual.visualID = visinfo->visualid; + v->mesa_visual.screen = visinfo->screen; +#endif + +#if defined(XFree86Server) || !(defined(__cplusplus) || defined(c_plusplus)) + v->mesa_visual.visualType = xmesa_convert_from_x_visual_type(visinfo->class); +#else + v->mesa_visual.visualType = xmesa_convert_from_x_visual_type(visinfo->c_class); +#endif + + v->mesa_visual.visualRating = visualCaveat; + + if (alpha_flag) + v->mesa_visual.alphaBits = 8; + + (void) initialize_visual_and_buffer( v, NULL, rgb_flag, 0, 0 ); + + { + const int xclass = v->mesa_visual.visualType; + if (xclass == GLX_TRUE_COLOR || xclass == GLX_DIRECT_COLOR) { + red_bits = _mesa_bitcount(GET_REDMASK(v)); + green_bits = _mesa_bitcount(GET_GREENMASK(v)); + blue_bits = _mesa_bitcount(GET_BLUEMASK(v)); + } + else { + /* this is an approximation */ + int depth; + depth = GET_VISUAL_DEPTH(v); + red_bits = depth / 3; + depth -= red_bits; + green_bits = depth / 2; + depth -= green_bits; + blue_bits = depth; + alpha_bits = 0; + assert( red_bits + green_bits + blue_bits == GET_VISUAL_DEPTH(v) ); + } + alpha_bits = v->mesa_visual.alphaBits; + } + + _mesa_initialize_visual( &v->mesa_visual, + rgb_flag, db_flag, stereo_flag, + red_bits, green_bits, + blue_bits, alpha_bits, + v->mesa_visual.indexBits, + depth_size, + stencil_size, + accum_red_size, accum_green_size, + accum_blue_size, accum_alpha_size, + 0 ); + + /* XXX minor hack */ + v->mesa_visual.level = level; + return v; +} + + +PUBLIC +void XMesaDestroyVisual( XMesaVisual v ) +{ +#ifndef XFree86Server + _mesa_free(v->visinfo); +#endif + _mesa_free(v); +} + + + +/** + * Create a new XMesaContext. + * \param v the XMesaVisual + * \param share_list another XMesaContext with which to share display + * lists or NULL if no sharing is wanted. + * \return an XMesaContext or NULL if error. + */ +PUBLIC +XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) +{ + static GLboolean firstTime = GL_TRUE; + struct pipe_context *pipe; + XMesaContext c; + GLcontext *mesaCtx; + uint pf; + + if (firstTime) { + _glthread_INIT_MUTEX(_xmesa_lock); + firstTime = GL_FALSE; + } + + /* Note: the XMesaContext contains a Mesa GLcontext struct (inheritance) */ + c = (XMesaContext) CALLOC_STRUCT(xmesa_context); + if (!c) + return NULL; + + pf = choose_pixel_format(v); + assert(pf); + + if (!getenv("XM_AUB")) { + xmesa_mode = XMESA_SOFTPIPE; + pipe = xmesa_create_pipe_context( c, pf ); + } + else { + xmesa_mode = XMESA_AUB; + pipe = xmesa_create_i965simple( xmesa_get_pipe_winsys_aub() ); + } + + c->st = st_create_context(pipe, &v->mesa_visual, + share_list ? share_list->st : NULL); + mesaCtx = c->st->ctx; + c->st->ctx->DriverCtx = c; + +#if 00 + _mesa_enable_sw_extensions(mesaCtx); + _mesa_enable_1_3_extensions(mesaCtx); + _mesa_enable_1_4_extensions(mesaCtx); + _mesa_enable_1_5_extensions(mesaCtx); + _mesa_enable_2_0_extensions(mesaCtx); +#endif + +#ifdef XFree86Server + /* If we're running in the X server, do bounds checking to prevent + * segfaults and server crashes! + */ + mesaCtx->Const.CheckArrayBounds = GL_TRUE; +#endif + + /* finish up xmesa context initializations */ + c->xm_visual = v; + c->xm_buffer = NULL; /* set later by XMesaMakeCurrent */ + + c->st->haveFramebufferSurfaces = GL_TRUE; + + return c; +} + + + +PUBLIC +void XMesaDestroyContext( XMesaContext c ) +{ + st_destroy_context(c->st); + _mesa_free(c); +} + + + +/** + * Private function for creating an XMesaBuffer which corresponds to an + * X window or pixmap. + * \param v the window's XMesaVisual + * \param w the window we're wrapping + * \return new XMesaBuffer or NULL if error + */ +PUBLIC XMesaBuffer +XMesaCreateWindowBuffer(XMesaVisual v, XMesaWindow w) +{ +#ifndef XFree86Server + XWindowAttributes attr; +#endif + XMesaBuffer b; + XMesaColormap cmap; + int depth; + + assert(v); + assert(w); + + /* Check that window depth matches visual depth */ +#ifdef XFree86Server + depth = ((XMesaDrawable)w)->depth; +#else + XGetWindowAttributes( v->display, w, &attr ); + depth = attr.depth; +#endif + if (GET_VISUAL_DEPTH(v) != depth) { + _mesa_warning(NULL, "XMesaCreateWindowBuffer: depth mismatch between visual (%d) and window (%d)!\n", + GET_VISUAL_DEPTH(v), depth); + return NULL; + } + + /* Find colormap */ +#ifdef XFree86Server + cmap = (ColormapPtr)LookupIDByType(wColormap(w), RT_COLORMAP); +#else + if (attr.colormap) { + cmap = attr.colormap; + } + else { + _mesa_warning(NULL, "Window %u has no colormap!\n", (unsigned int) w); + /* this is weird, a window w/out a colormap!? */ + /* OK, let's just allocate a new one and hope for the best */ + cmap = XCreateColormap(v->display, w, attr.visual, AllocNone); + } +#endif + + b = create_xmesa_buffer((XMesaDrawable) w, WINDOW, v, cmap); + if (!b) + return NULL; + + if (!initialize_visual_and_buffer( v, b, v->mesa_visual.rgbMode, + (XMesaDrawable) w, cmap )) { + xmesa_free_buffer(b); + return NULL; + } + + return b; +} + + + +/** + * Create a new XMesaBuffer from an X pixmap. + * + * \param v the XMesaVisual + * \param p the pixmap + * \param cmap the colormap, may be 0 if using a \c GLX_TRUE_COLOR or + * \c GLX_DIRECT_COLOR visual for the pixmap + * \returns new XMesaBuffer or NULL if error + */ +PUBLIC XMesaBuffer +XMesaCreatePixmapBuffer(XMesaVisual v, XMesaPixmap p, XMesaColormap cmap) +{ + XMesaBuffer b; + + assert(v); + + b = create_xmesa_buffer((XMesaDrawable) p, PIXMAP, v, cmap); + if (!b) + return NULL; + + if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode, + (XMesaDrawable) p, cmap)) { + xmesa_free_buffer(b); + return NULL; + } + + return b; +} + + +/** + * For GLX_EXT_texture_from_pixmap + */ +XMesaBuffer +XMesaCreatePixmapTextureBuffer(XMesaVisual v, XMesaPixmap p, + XMesaColormap cmap, + int format, int target, int mipmap) +{ + GET_CURRENT_CONTEXT(ctx); + XMesaBuffer b; + GLuint width, height; + + assert(v); + + b = create_xmesa_buffer((XMesaDrawable) p, PIXMAP, v, cmap); + if (!b) + return NULL; + + /* get pixmap size, update framebuffer/renderbuffer dims */ + xmesa_get_window_size(v->display, b, &width, &height); + _mesa_resize_framebuffer(NULL, &(b->stfb->Base), width, height); + + if (target == 0) { + /* examine dims */ + if (ctx->Extensions.ARB_texture_non_power_of_two) { + target = GLX_TEXTURE_2D_EXT; + } + else if ( _mesa_bitcount(width) == 1 + && _mesa_bitcount(height) == 1) { + /* power of two size */ + if (height == 1) { + target = GLX_TEXTURE_1D_EXT; + } + else { + target = GLX_TEXTURE_2D_EXT; + } + } + else if (ctx->Extensions.NV_texture_rectangle) { + target = GLX_TEXTURE_RECTANGLE_EXT; + } + else { + /* non power of two textures not supported */ + XMesaDestroyBuffer(b); + return 0; + } + } + + b->TextureTarget = target; + b->TextureFormat = format; + b->TextureMipmap = mipmap; + + if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode, + (XMesaDrawable) p, cmap)) { + xmesa_free_buffer(b); + return NULL; + } + + return b; +} + + + +XMesaBuffer +XMesaCreatePBuffer(XMesaVisual v, XMesaColormap cmap, + unsigned int width, unsigned int height) +{ +#ifndef XFree86Server + XMesaWindow root; + XMesaDrawable drawable; /* X Pixmap Drawable */ + XMesaBuffer b; + + /* allocate pixmap for front buffer */ + root = RootWindow( v->display, v->visinfo->screen ); + drawable = XCreatePixmap(v->display, root, width, height, + v->visinfo->depth); + if (!drawable) + return NULL; + + b = create_xmesa_buffer(drawable, PBUFFER, v, cmap); + if (!b) + return NULL; + + if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode, + drawable, cmap)) { + xmesa_free_buffer(b); + return NULL; + } + + return b; +#else + return 0; +#endif +} + + + +/* + * Deallocate an XMesaBuffer structure and all related info. + */ +PUBLIC void +XMesaDestroyBuffer(XMesaBuffer b) +{ + xmesa_free_buffer(b); +} + + +/** + * Query the current window size and update the corresponding GLframebuffer + * and all attached renderbuffers. + * Called when: + * 1. the first time a buffer is bound to a context. + * 2. from the XMesaResizeBuffers() API function. + * 3. SwapBuffers. XXX probabaly from xm_flush_frontbuffer() too... + * Note: it's possible (and legal) for xmctx to be NULL. That can happen + * when resizing a buffer when no rendering context is bound. + */ +void +xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer) +{ + GLuint width, height; + xmesa_get_window_size(drawBuffer->xm_visual->display, drawBuffer, &width, &height); + st_resize_framebuffer(drawBuffer->stfb, width, height); +} + + +/* + * Bind buffer b to context c and make c the current rendering context. + */ +GLboolean XMesaMakeCurrent( XMesaContext c, XMesaBuffer b ) +{ + return XMesaMakeCurrent2( c, b, b ); +} + + +/* + * Bind buffer b to context c and make c the current rendering context. + */ +PUBLIC +GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, + XMesaBuffer readBuffer ) +{ + if (c) { + if (!drawBuffer || !readBuffer) + return GL_FALSE; /* must specify buffers! */ + +#if 0 + /* XXX restore this optimization */ + if (&(c->mesa) == _mesa_get_current_context() + && c->mesa.DrawBuffer == &drawBuffer->mesa_buffer + && c->mesa.ReadBuffer == &readBuffer->mesa_buffer + && xmesa_buffer(c->mesa.DrawBuffer)->wasCurrent) { + /* same context and buffer, do nothing */ + return GL_TRUE; + } +#endif + + c->xm_buffer = drawBuffer; + + /* Call this periodically to detect when the user has begun using + * GL rendering from multiple threads. + */ + _glapi_check_multithread(); + + st_make_current(c->st, drawBuffer->stfb, readBuffer->stfb); + + xmesa_check_and_update_buffer_size(c, drawBuffer); + if (readBuffer != drawBuffer) + xmesa_check_and_update_buffer_size(c, readBuffer); + + /* Solution to Stephane Rehel's problem with glXReleaseBuffersMESA(): */ + drawBuffer->wasCurrent = GL_TRUE; + } + else { + /* Detach */ + st_make_current( NULL, NULL, NULL ); + } + return GL_TRUE; +} + + +/* + * Unbind the context c from its buffer. + */ +GLboolean XMesaUnbindContext( XMesaContext c ) +{ + /* A no-op for XFree86 integration purposes */ + return GL_TRUE; +} + + +XMesaContext XMesaGetCurrentContext( void ) +{ + GET_CURRENT_CONTEXT(ctx); + if (ctx) { + XMesaContext xmesa = xmesa_context(ctx); + return xmesa; + } + else { + return 0; + } +} + + +XMesaBuffer XMesaGetCurrentBuffer( void ) +{ + GET_CURRENT_CONTEXT(ctx); + if (ctx) { + XMesaBuffer xmbuf = xmesa_buffer(ctx->DrawBuffer); + return xmbuf; + } + else { + return 0; + } +} + + +/* New in Mesa 3.1 */ +XMesaBuffer XMesaGetCurrentReadBuffer( void ) +{ + GET_CURRENT_CONTEXT(ctx); + if (ctx) { + return xmesa_buffer(ctx->ReadBuffer); + } + else { + return 0; + } +} + + +#ifdef XFree86Server +PUBLIC +GLboolean XMesaForceCurrent(XMesaContext c) +{ + if (c) { + _glapi_set_dispatch(c->mesa.CurrentDispatch); + + if (&(c->mesa) != _mesa_get_current_context()) { + _mesa_make_current(&c->mesa, c->mesa.DrawBuffer, c->mesa.ReadBuffer); + } + } + else { + _mesa_make_current(NULL, NULL, NULL); + } + return GL_TRUE; +} + + +PUBLIC +GLboolean XMesaLoseCurrent(XMesaContext c) +{ + (void) c; + _mesa_make_current(NULL, NULL, NULL); + return GL_TRUE; +} + + +PUBLIC +GLboolean XMesaCopyContext( XMesaContext xm_src, XMesaContext xm_dst, GLuint mask ) +{ + _mesa_copy_context(&xm_src->mesa, &xm_dst->mesa, mask); + return GL_TRUE; +} +#endif /* XFree86Server */ + + +#ifndef FX +GLboolean XMesaSetFXmode( GLint mode ) +{ + (void) mode; + return GL_FALSE; +} +#endif + + + +/* + * Copy the back buffer to the front buffer. If there's no back buffer + * this is a no-op. + */ +PUBLIC +void XMesaSwapBuffers( XMesaBuffer b ) +{ + struct pipe_surface *surf; + + /* If we're swapping the buffer associated with the current context + * we have to flush any pending rendering commands first. + */ + st_notify_swapbuffers(b->stfb); + + surf = st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT); + if (surf) { + if (xmesa_mode == XMESA_AUB) + xmesa_display_aub( surf ); + else + xmesa_display_surface(b, surf); + } + + xmesa_check_and_update_buffer_size(NULL, b); +} + + + +/* + * Copy sub-region of back buffer to front buffer + */ +void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height ) +{ + struct pipe_surface *surf_front + = st_get_framebuffer_surface(b->stfb, ST_SURFACE_FRONT_LEFT); + struct pipe_surface *surf_back + = st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT); + struct pipe_context *pipe = NULL; /* XXX fix */ + + if (!surf_front || !surf_back) + return; + + pipe->surface_copy(pipe, + FALSE, + surf_front, x, y, /* dest */ + surf_back, x, y, /* src */ + width, height); +} + + + +/* + * Return the depth buffer associated with an XMesaBuffer. + * Input: b - the XMesa buffer handle + * Output: width, height - size of buffer in pixels + * bytesPerValue - bytes per depth value (2 or 4) + * buffer - pointer to depth buffer values + * Return: GL_TRUE or GL_FALSE to indicate success or failure. + */ +GLboolean XMesaGetDepthBuffer( XMesaBuffer b, GLint *width, GLint *height, + GLint *bytesPerValue, void **buffer ) +{ + *width = 0; + *height = 0; + *bytesPerValue = 0; + *buffer = 0; + return GL_FALSE; +} + + +void XMesaFlush( XMesaContext c ) +{ + if (c && c->xm_visual->display) { +#ifdef XFree86Server + /* NOT_NEEDED */ +#else + XSync( c->xm_visual->display, False ); +#endif + } +} + + + +const char *XMesaGetString( XMesaContext c, int name ) +{ + (void) c; + if (name==XMESA_VERSION) { + return "5.0"; + } + else if (name==XMESA_EXTENSIONS) { + return ""; + } + else { + return NULL; + } +} + + + +XMesaBuffer XMesaFindBuffer( XMesaDisplay *dpy, XMesaDrawable d ) +{ + XMesaBuffer b; + for (b=XMesaBufferList; b; b=b->Next) { + if (b->drawable == d && b->xm_visual->display == dpy) { + return b; + } + } + return NULL; +} + + +/** + * Free/destroy all XMesaBuffers associated with given display. + */ +void xmesa_destroy_buffers_on_display(XMesaDisplay *dpy) +{ + XMesaBuffer b, next; + for (b = XMesaBufferList; b; b = next) { + next = b->Next; + if (b->xm_visual->display == dpy) { + xmesa_free_buffer(b); + } + } +} + + +/* + * Look for XMesaBuffers whose X window has been destroyed. + * Deallocate any such XMesaBuffers. + */ +void XMesaGarbageCollect( void ) +{ + XMesaBuffer b, next; + for (b=XMesaBufferList; b; b=next) { + next = b->Next; + if (b->xm_visual && + b->xm_visual->display && + b->drawable && + b->type == WINDOW) { +#ifdef XFree86Server + /* NOT_NEEDED */ +#else + XSync(b->xm_visual->display, False); + if (!window_exists( b->xm_visual->display, b->drawable )) { + /* found a dead window, free the ancillary info */ + XMesaDestroyBuffer( b ); + } +#endif + } + } +} + + +unsigned long XMesaDitherColor( XMesaContext xmesa, GLint x, GLint y, + GLfloat red, GLfloat green, + GLfloat blue, GLfloat alpha ) +{ + /* no longer supported */ + return 0; +} + + +/* + * This is typically called when the window size changes and we need + * to reallocate the buffer's back/depth/stencil/accum buffers. + */ +PUBLIC void +XMesaResizeBuffers( XMesaBuffer b ) +{ + GET_CURRENT_CONTEXT(ctx); + XMesaContext xmctx = xmesa_context(ctx); + if (!xmctx) + return; + xmesa_check_and_update_buffer_size(xmctx, b); +} + + + + +PUBLIC void +XMesaBindTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer, + const int *attrib_list) +{ +} + + + +PUBLIC void +XMesaReleaseTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer) +{ +} + diff --git a/src/gallium/winsys/xlib/xm_image.c b/src/gallium/winsys/xlib/xm_image.c new file mode 100644 index 0000000000..087b4e4c3a --- /dev/null +++ b/src/gallium/winsys/xlib/xm_image.c @@ -0,0 +1,133 @@ +/************************************************************************** + +Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sub license, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial portions +of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR +ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Kevin E. Martin + * Brian Paul + */ + +#include +#include + +#include "glxheader.h" +#include "xmesaP.h" + +#ifdef XFree86Server + +#ifdef ROUNDUP +#undef ROUNDUP +#endif + +#define ROUNDUP(nbytes, pad) ((((nbytes) + ((pad)-1)) / (pad)) * ((pad)>>3)) + +XMesaImage *XMesaCreateImage(int bitsPerPixel, int width, int height, char *data) +{ + XMesaImage *image; + + image = (XMesaImage *)xalloc(sizeof(XMesaImage)); + + if (image) { + image->width = width; + image->height = height; + image->data = data; + /* Always pad to 32 bits */ + image->bytes_per_line = ROUNDUP((bitsPerPixel * width), 32); + image->bits_per_pixel = bitsPerPixel; + } + + return image; +} + +void XMesaDestroyImage(XMesaImage *image) +{ + if (image->data) + free(image->data); + xfree(image); +} + +unsigned long XMesaGetPixel(XMesaImage *image, int x, int y) +{ + CARD8 *row = (CARD8 *)(image->data + y*image->bytes_per_line); + CARD8 *i8; + CARD16 *i16; + CARD32 *i32; + switch (image->bits_per_pixel) { + case 8: + i8 = (CARD8 *)row; + return i8[x]; + break; + case 15: + case 16: + i16 = (CARD16 *)row; + return i16[x]; + break; + case 24: /* WARNING: architecture specific code */ + i8 = (CARD8 *)row; + return (((CARD32)i8[x*3]) | + (((CARD32)i8[x*3+1])<<8) | + (((CARD32)i8[x*3+2])<<16)); + break; + case 32: + i32 = (CARD32 *)row; + return i32[x]; + break; + } + return 0; +} + +#ifndef XMESA_USE_PUTPIXEL_MACRO +void XMesaPutPixel(XMesaImage *image, int x, int y, unsigned long pixel) +{ + CARD8 *row = (CARD8 *)(image->data + y*image->bytes_per_line); + CARD8 *i8; + CARD16 *i16; + CARD32 *i32; + switch (image->bits_per_pixel) { + case 8: + i8 = (CARD8 *)row; + i8[x] = (CARD8)pixel; + break; + case 15: + case 16: + i16 = (CARD16 *)row; + i16[x] = (CARD16)pixel; + break; + case 24: /* WARNING: architecture specific code */ + i8 = (CARD8 *)__row; + i8[x*3] = (CARD8)(p); + i8[x*3+1] = (CARD8)(p>>8); + i8[x*3+2] = (CARD8)(p>>16); + case 32: + i32 = (CARD32 *)row; + i32[x] = (CARD32)pixel; + break; + } +} +#endif + +#endif /* XFree86Server */ diff --git a/src/gallium/winsys/xlib/xm_image.h b/src/gallium/winsys/xlib/xm_image.h new file mode 100644 index 0000000000..2a5e0f3777 --- /dev/null +++ b/src/gallium/winsys/xlib/xm_image.h @@ -0,0 +1,77 @@ +/************************************************************************** + +Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the +"Software"), to deal in the Software without restriction, including +without limitation the rights to use, copy, modify, merge, publish, +distribute, sub license, and/or sell copies of the Software, and to +permit persons to whom the Software is furnished to do so, subject to +the following conditions: + +The above copyright notice and this permission notice (including the +next paragraph) shall be included in all copies or substantial portions +of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR +ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + +/* + * Authors: + * Kevin E. Martin + * Brian Paul + */ + +#ifndef _XM_IMAGE_H_ +#define _XM_IMAGE_H_ + +#define XMESA_USE_PUTPIXEL_MACRO + +extern XMesaImage *XMesaCreateImage(int bitsPerPixel, int width, int height, + char *data); +extern void XMesaDestroyImage(XMesaImage *image); +extern unsigned long XMesaGetPixel(XMesaImage *image, int x, int y); +#ifdef XMESA_USE_PUTPIXEL_MACRO +#define XMesaPutPixel(__i,__x,__y,__p) \ +{ \ + CARD8 *__row = (CARD8 *)(__i->data + __y*__i->bytes_per_line); \ + CARD8 *__i8; \ + CARD16 *__i16; \ + CARD32 *__i32; \ + switch (__i->bits_per_pixel) { \ + case 8: \ + __i8 = (CARD8 *)__row; \ + __i8[__x] = (CARD8)__p; \ + break; \ + case 15: \ + case 16: \ + __i16 = (CARD16 *)__row; \ + __i16[__x] = (CARD16)__p; \ + break; \ + case 24: /* WARNING: architecture specific code */ \ + __i8 = (CARD8 *)__row; \ + __i8[__x*3] = (CARD8)(__p); \ + __i8[__x*3+1] = (CARD8)(__p>>8); \ + __i8[__x*3+2] = (CARD8)(__p>>16); \ + break; \ + case 32: \ + __i32 = (CARD32 *)__row; \ + __i32[__x] = (CARD32)__p; \ + break; \ + } \ +} +#else +extern void XMesaPutPixel(XMesaImage *image, int x, int y, + unsigned long pixel); +#endif + +#endif /* _XM_IMAGE_H_ */ diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c new file mode 100644 index 0000000000..c3cd22eea3 --- /dev/null +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -0,0 +1,466 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "glxheader.h" +#include "xmesaP.h" + +#include "pipe/p_winsys.h" +#include "pipe/p_format.h" +#include "pipe/p_context.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/softpipe/sp_winsys.h" + +#ifdef GALLIUM_CELL +#include "pipe/cell/ppu/cell_context.h" +#include "pipe/cell/ppu/cell_winsys.h" +#else +#define TILE_SIZE 32 /* avoid compilation errors */ +#endif + +#include "xm_winsys_aub.h" + + +/** + * Low-level OS/window system memory buffer + */ +struct xm_buffer +{ + struct pipe_buffer base; + boolean userBuffer; /** Is this a user-space buffer? */ + void *data; + void *mapped; +}; + + +struct xmesa_surface +{ + struct pipe_surface surface; + + int tileSize; +}; + + +/** + * Derived from softpipe_winsys. + * We just need one extra field which indicates the pixel format to use for + * drawing surfaces so that we're compatible with the XVisual/window format. + */ +struct xmesa_softpipe_winsys +{ + struct softpipe_winsys spws; + enum pipe_format pixelformat; +}; + + + +/** Cast wrapper */ +static INLINE struct xmesa_surface * +xmesa_surface(struct pipe_surface *ps) +{ +// assert(0); + return (struct xmesa_surface *) ps; +} + +/** cast wrapper */ +static INLINE struct xmesa_softpipe_winsys * +xmesa_softpipe_winsys(struct softpipe_winsys *spws) +{ + return (struct xmesa_softpipe_winsys *) spws; +} + +/** + * Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static INLINE struct xm_buffer * +xm_buffer( struct pipe_buffer *buf ) +{ + return (struct xm_buffer *)buf; +} + + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void * +xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buf, + unsigned flags) +{ + struct xm_buffer *xm_buf = xm_buffer(buf); + xm_buf->mapped = xm_buf->data; + return xm_buf->mapped; +} + +static void +xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) +{ + struct xm_buffer *xm_buf = xm_buffer(buf); + xm_buf->mapped = NULL; +} + +static void +xm_buffer_destroy(struct pipe_winsys *pws, + struct pipe_buffer *buf) +{ + struct xm_buffer *oldBuf = xm_buffer(buf); + + if (oldBuf->data) { + if (!oldBuf->userBuffer) + align_free(oldBuf->data); + oldBuf->data = NULL; + } + + free(oldBuf); +} + + +/** + * Display a surface that's in a tiled configuration. That is, all the + * pixels for a TILE_SIZExTILE_SIZE block are contiguous in memory. + */ +static void +xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) +{ + XImage *ximage = b->tempImage; + struct xm_buffer *xm_buf = xm_buffer(surf->buffer); + const uint tilesPerRow = (surf->width + TILE_SIZE - 1) / TILE_SIZE; + uint x, y; + + /* check that the XImage has been previously initialized */ + assert(ximage->format); + assert(ximage->bitmap_unit); + + /* update XImage's fields */ + ximage->width = TILE_SIZE; + ximage->height = TILE_SIZE; + ximage->bytes_per_line = TILE_SIZE * 4; + + for (y = 0; y < surf->height; y += TILE_SIZE) { + for (x = 0; x < surf->width; x += TILE_SIZE) { + int dx = x; + int dy = y; + int tx = x / TILE_SIZE; + int ty = y / TILE_SIZE; + int offset = ty * tilesPerRow + tx; + + offset *= 4 * TILE_SIZE * TILE_SIZE; + + ximage->data = (char *) xm_buf->data + offset; + + XPutImage(b->xm_visual->display, b->drawable, b->gc, + ximage, 0, 0, dx, dy, TILE_SIZE, TILE_SIZE); + } + } +} + + +/** + * Display/copy the image in the surface into the X window specified + * by the XMesaBuffer. + */ +void +xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) +{ + XImage *ximage = b->tempImage; + struct xm_buffer *xm_buf = xm_buffer(surf->buffer); + const struct xmesa_surface *xm_surf + = xmesa_surface((struct pipe_surface *) surf); + + if (xm_surf->tileSize) { + xmesa_display_surface_tiled(b, surf); + return; + } + + /* check that the XImage has been previously initialized */ + assert(ximage->format); + assert(ximage->bitmap_unit); + + /* update XImage's fields */ + ximage->width = surf->width; + ximage->height = surf->height; + ximage->bytes_per_line = surf->pitch * (ximage->bits_per_pixel / 8); + ximage->data = xm_buf->data; + + /* display image in Window */ + XPutImage(b->xm_visual->display, b->drawable, b->gc, + ximage, 0, 0, 0, 0, surf->width, surf->height); +} + + +static void +xm_flush_frontbuffer(struct pipe_winsys *pws, + struct pipe_surface *surf, + void *context_private) +{ + /* The Xlib driver's front color surfaces are actually X Windows so + * this flush is a no-op. + * If we instead did front buffer rendering to a temporary XImage, + * this would be the place to copy the Ximage to the on-screen Window. + */ + XMesaContext xmctx = (XMesaContext) context_private; + xmesa_display_surface(xmctx->xm_buffer, surf); +} + + + +static void +xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) +{ + va_list args; + va_start( args, fmtString ); + vfprintf(stderr, fmtString, args); + va_end( args ); +} + + +static const char * +xm_get_name(struct pipe_winsys *pws) +{ + return "Xlib"; +} + + +static struct pipe_buffer * +xm_buffer_create(struct pipe_winsys *pws, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + + /* align to 16-byte multiple for Cell */ + buffer->data = align_malloc(size, max(alignment, 16)); + + return &buffer->base; +} + + +/** + * Create buffer which wraps user-space data. + */ +static struct pipe_buffer * +xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) +{ + struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); + buffer->base.refcount = 1; + buffer->base.size = bytes; + buffer->userBuffer = TRUE; + buffer->data = ptr; + + return &buffer->base; +} + + + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + +static int +xm_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags) +{ + const unsigned alignment = 64; + + surf->width = width; + surf->height = height; + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = round_up(width, alignment / surf->cpp); + +#ifdef GALLIUM_CELL /* XXX a bit of a hack */ + height = round_up(height, TILE_SIZE); +#endif + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->cpp * height); + if(!surf->buffer) + return -1; + + return 0; +} + + +/** + * Called via pipe->surface_alloc() to create new surfaces (textures, + * renderbuffers, etc. + */ +static struct pipe_surface * +xm_surface_alloc(struct pipe_winsys *ws) +{ + struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + + assert(ws); + + xms->surface.refcount = 1; + xms->surface.winsys = ws; + +#ifdef GALLIUM_CELL + if (!getenv("GALLIUM_NOCELL")) { + xms->tileSize = 32; /** probably temporary */ + } +#endif + + return &xms->surface; +} + + + +static void +xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + surf->refcount--; + if (surf->refcount == 0) { + if (surf->buffer) + pipe_buffer_reference(winsys, &surf->buffer, NULL); + free(surf); + } + *s = NULL; +} + + + +/** + * Return pointer to a pipe_winsys object. + * For Xlib, this is a singleton object. + * Nothing special for the Xlib driver so no subclassing or anything. + */ +struct pipe_winsys * +xmesa_get_pipe_winsys_aub(void) +{ + static struct pipe_winsys *ws = NULL; + + if (!ws && getenv("XM_AUB")) { + ws = xmesa_create_pipe_winsys_aub(); + } + else if (!ws) { + ws = CALLOC_STRUCT(pipe_winsys); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + */ + ws->buffer_create = xm_buffer_create; + ws->user_buffer_create = xm_user_buffer_create; + ws->buffer_map = xm_buffer_map; + ws->buffer_unmap = xm_buffer_unmap; + ws->buffer_destroy = xm_buffer_destroy; + + ws->surface_alloc = xm_surface_alloc; + ws->surface_alloc_storage = xm_surface_alloc_storage; + ws->surface_release = xm_surface_release; + + ws->flush_frontbuffer = xm_flush_frontbuffer; + ws->printf = xm_printf; + ws->get_name = xm_get_name; + } + + return ws; +} + + +/** + * Called via softpipe_winsys->is_format_supported(). + * This function is only called to test formats for front/back color surfaces. + * The winsys being queried will have been created at glXCreateContext + * time, with a pixel format corresponding to the context's visual. + */ +static boolean +xmesa_is_format_supported(struct softpipe_winsys *sws, + enum pipe_format format) +{ + struct xmesa_softpipe_winsys *xmws = xmesa_softpipe_winsys(sws); + return (format == xmws->pixelformat); +} + + +/** + * Return pointer to a softpipe_winsys object. + */ +static struct softpipe_winsys * +xmesa_get_softpipe_winsys(uint pixelformat) +{ + struct xmesa_softpipe_winsys *xmws + = CALLOC_STRUCT(xmesa_softpipe_winsys); + if (!xmws) + return NULL; + + xmws->spws.is_format_supported = xmesa_is_format_supported; + xmws->pixelformat = pixelformat; + + return &xmws->spws; +} + + +struct pipe_context * +xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) +{ + struct pipe_winsys *pws = xmesa_get_pipe_winsys_aub(); + struct pipe_context *pipe; + +#ifdef GALLIUM_CELL + if (!getenv("GALLIUM_NOCELL")) { + struct cell_winsys *cws = cell_get_winsys(pixelformat); + pipe = cell_create_context(pws, cws); + if (pipe) + pipe->priv = xmesa; + return pipe; + } + else +#endif + { + struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(pixelformat); + pipe = softpipe_create( pws, spws ); + if (pipe) + pipe->priv = xmesa; + + return pipe; + } +} diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c new file mode 100644 index 0000000000..bf41570257 --- /dev/null +++ b/src/gallium/winsys/xlib/xm_winsys_aub.c @@ -0,0 +1,589 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/* + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#include "glxheader.h" +#include "xmesaP.h" + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/i965simple/brw_winsys.h" +#include "brw_aub.h" +#include "xm_winsys_aub.h" + + + +struct aub_buffer { + char *data; + unsigned offset; + unsigned size; + unsigned refcount; + unsigned map_count; + boolean dump_on_unmap; +}; + + + +struct aub_pipe_winsys { + struct pipe_winsys winsys; + + struct brw_aubfile *aubfile; + + /* This is simple, isn't it: + */ + char *pool; + unsigned size; + unsigned used; +}; + + +/* Turn a pipe winsys into an aub/pipe winsys: + */ +static inline struct aub_pipe_winsys * +aub_pipe_winsys( struct pipe_winsys *winsys ) +{ + return (struct aub_pipe_winsys *)winsys; +} + + + +static INLINE struct aub_buffer * +aub_bo( struct pipe_buffer *bo ) +{ + return (struct aub_buffer *)bo; +} + +static INLINE struct pipe_buffer * +pipe_bo( struct aub_buffer *bo ) +{ + return (struct pipe_buffer *)bo; +} + + + + +static void *aub_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags ) +{ + struct aub_buffer *sbo = aub_bo(buf); + + assert(sbo->data); + + if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) + sbo->dump_on_unmap = 1; + + sbo->map_count++; + return sbo->data; +} + +static void aub_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + struct aub_buffer *sbo = aub_bo(buf); + + sbo->map_count--; + + if (sbo->map_count == 0 && + sbo->dump_on_unmap) { + + sbo->dump_on_unmap = 0; + + brw_aub_gtt_data( iws->aubfile, + sbo->offset, + sbo->data, + sbo->size, + 0, + 0); + } +} + + +static void +aub_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + free(buf); +} + + +void xmesa_buffer_subdata_aub(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned long offset, + unsigned long size, + const void *data, + unsigned aub_type, + unsigned aub_sub_type) +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + struct aub_buffer *sbo = aub_bo(buf); + + assert(sbo->size > offset + size); + memcpy(sbo->data + offset, data, size); + + brw_aub_gtt_data( iws->aubfile, + sbo->offset + offset, + sbo->data + offset, + size, + aub_type, + aub_sub_type ); +} + +void xmesa_commands_aub(struct pipe_winsys *winsys, + unsigned *cmds, + unsigned nr_dwords) +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + unsigned size = nr_dwords * 4; + + assert(iws->used + size < iws->size); + + brw_aub_gtt_cmds( iws->aubfile, + AUB_BUF_START + iws->used, + cmds, + nr_dwords * sizeof(int) ); + + iws->used += align(size, 4096); +} + + +static struct aub_pipe_winsys *global_winsys = NULL; + +void xmesa_display_aub( /* struct pipe_winsys *winsys, */ + struct pipe_surface *surface ) +{ +// struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + brw_aub_dump_bmp( global_winsys->aubfile, + surface, + aub_bo(surface->buffer)->offset ); +} + + + +/* Pipe has no concept of pools. We choose the tex/region pool + * for all buffers. + */ +static struct pipe_buffer * +aub_buffer_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer); + + sbo->refcount = 1; + + /* Could reuse buffers that are not referenced in current + * batchbuffer. Can't do that atm, so always reallocate: + */ + assert(iws->used + size < iws->size); + sbo->data = iws->pool + iws->used; + sbo->offset = AUB_BUF_START + iws->used; + iws->used += align(size, 4096); + + sbo->size = size; + + return pipe_bo(sbo); +} + + +static struct pipe_buffer * +aub_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) +{ + struct aub_buffer *sbo; + + /* Lets hope this is meant for upload, not as a result! + */ + sbo = aub_bo(aub_buffer_create( winsys, 0, 0, 0 )); + + sbo->data = ptr; + sbo->size = bytes; + + return pipe_bo(sbo); +} + + +/* The state tracker (should!) keep track of whether the fake + * frontbuffer has been touched by any rendering since the last time + * we copied its contents to the real frontbuffer. Our task is easy: + */ +static void +aub_flush_frontbuffer( struct pipe_winsys *winsys, + struct pipe_surface *surf, + void *context_private) +{ + xmesa_display_aub( surf ); +} + +static struct pipe_surface * +aub_i915_surface_alloc(struct pipe_winsys *winsys) +{ + struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); + if (surf) { + surf->refcount = 1; + surf->winsys = winsys; + } + return surf; +} + + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + +static int +aub_i915_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags) +{ + const unsigned alignment = 64; + + surf->width = width; + surf->height = height; + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = round_up(width, alignment / surf->cpp); + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->cpp * height); + if(!surf->buffer) + return -1; + + return 0; +} + +static void +aub_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + surf->refcount--; + if (surf->refcount == 0) { + if (surf->buffer) + pipe_buffer_reference(winsys, &surf->buffer, NULL); + free(surf); + } + *s = NULL; +} + + + +static void +aub_printf( struct pipe_winsys *winsys, const char *fmtString, ... ) +{ + va_list args; + va_start( args, fmtString ); + vfprintf(stderr, fmtString, args); + va_end( args ); +} + +static const char * +aub_get_name( struct pipe_winsys *winsys ) +{ + return "Aub/xlib"; +} + +struct pipe_winsys * +xmesa_create_pipe_winsys_aub( void ) +{ + struct aub_pipe_winsys *iws = CALLOC_STRUCT( aub_pipe_winsys ); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + * + * Pipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + iws->winsys.buffer_create = aub_buffer_create; + iws->winsys.user_buffer_create = aub_user_buffer_create; + iws->winsys.buffer_map = aub_buffer_map; + iws->winsys.buffer_unmap = aub_buffer_unmap; + iws->winsys.buffer_destroy = aub_buffer_destroy; + iws->winsys.flush_frontbuffer = aub_flush_frontbuffer; + iws->winsys.printf = aub_printf; + iws->winsys.get_name = aub_get_name; + + iws->winsys.surface_alloc = aub_i915_surface_alloc; + iws->winsys.surface_alloc_storage = aub_i915_surface_alloc_storage; + iws->winsys.surface_release = aub_i915_surface_release; + + iws->aubfile = brw_aubfile_create(); + iws->size = AUB_BUF_SIZE; + iws->pool = malloc(AUB_BUF_SIZE); + + /* HACK: static copy of this pointer: + */ + assert(global_winsys == NULL); + global_winsys = iws; + + return &iws->winsys; +} + + +void +xmesa_destroy_pipe_winsys_aub( struct pipe_winsys *winsys ) + +{ + struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); + brw_aub_destroy(iws->aubfile); + free(iws->pool); + free(iws); +} + + + + + + + +#define IWS_BATCHBUFFER_SIZE 1024 + +struct aub_brw_winsys { + struct brw_winsys winsys; /**< batch buffer funcs */ + struct aub_context *aub; + + struct pipe_winsys *pipe_winsys; + + unsigned batch_data[IWS_BATCHBUFFER_SIZE]; + unsigned batch_nr; + unsigned batch_size; + unsigned batch_alloc; +}; + + +/* Turn a i965simple winsys into an aub/i965simple winsys: + */ +static inline struct aub_brw_winsys * +aub_brw_winsys( struct brw_winsys *sws ) +{ + return (struct aub_brw_winsys *)sws; +} + + +/* Simple batchbuffer interface: + */ + +static unsigned *aub_i965_batch_start( struct brw_winsys *sws, + unsigned dwords, + unsigned relocs ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + + if (iws->batch_size < iws->batch_nr + dwords) + return NULL; + + iws->batch_alloc = iws->batch_nr + dwords; + return (void *)1; /* not a valid pointer! */ +} + +static void aub_i965_batch_dword( struct brw_winsys *sws, + unsigned dword ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + + assert(iws->batch_nr < iws->batch_alloc); + iws->batch_data[iws->batch_nr++] = dword; +} + +static void aub_i965_batch_reloc( struct brw_winsys *sws, + struct pipe_buffer *buf, + unsigned access_flags, + unsigned delta ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + + assert(iws->batch_nr < iws->batch_alloc); + iws->batch_data[iws->batch_nr++] = aub_bo(buf)->offset + delta; +} + +static unsigned aub_i965_get_buffer_offset( struct brw_winsys *sws, + struct pipe_buffer *buf, + unsigned access_flags ) +{ + return aub_bo(buf)->offset; +} + +static void aub_i965_batch_end( struct brw_winsys *sws ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + + assert(iws->batch_nr <= iws->batch_alloc); + iws->batch_alloc = 0; +} + +static void aub_i965_batch_flush( struct brw_winsys *sws, + struct pipe_fence_handle **fence ) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(sws); + assert(iws->batch_nr <= iws->batch_size); + + if (iws->batch_nr) { + xmesa_commands_aub( iws->pipe_winsys, + iws->batch_data, + iws->batch_nr ); + } + + iws->batch_nr = 0; +} + + + +static void aub_i965_buffer_subdata_typed(struct brw_winsys *winsys, + struct pipe_buffer *buf, + unsigned long offset, + unsigned long size, + const void *data, + unsigned data_type) +{ + struct aub_brw_winsys *iws = aub_brw_winsys(winsys); + unsigned aub_type = DW_GENERAL_STATE; + unsigned aub_sub_type; + + switch (data_type) { + case BRW_CC_VP: + aub_sub_type = DWGS_COLOR_CALC_VIEWPORT_STATE; + break; + case BRW_CC_UNIT: + aub_sub_type = DWGS_COLOR_CALC_STATE; + break; + case BRW_WM_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_SAMPLER_DEFAULT_COLOR: + aub_sub_type = DWGS_SAMPLER_DEFAULT_COLOR; + break; + case BRW_SAMPLER: + aub_sub_type = DWGS_SAMPLER_STATE; + break; + case BRW_WM_UNIT: + aub_sub_type = DWGS_WINDOWER_IZ_STATE; + break; + case BRW_SF_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_SF_VP: + aub_sub_type = DWGS_STRIPS_FANS_VIEWPORT_STATE; + break; + case BRW_SF_UNIT: + aub_sub_type = DWGS_STRIPS_FANS_STATE; + break; + case BRW_VS_UNIT: + aub_sub_type = DWGS_VERTEX_SHADER_STATE; + break; + case BRW_VS_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_GS_UNIT: + aub_sub_type = DWGS_GEOMETRY_SHADER_STATE; + break; + case BRW_GS_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_CLIP_VP: + aub_sub_type = DWGS_CLIPPER_VIEWPORT_STATE; + break; + case BRW_CLIP_UNIT: + aub_sub_type = DWGS_CLIPPER_STATE; + break; + case BRW_CLIP_PROG: + aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; + break; + case BRW_SS_SURFACE: + aub_type = DW_SURFACE_STATE; + aub_sub_type = DWSS_SURFACE_STATE; + break; + case BRW_SS_SURF_BIND: + aub_type = DW_SURFACE_STATE; + aub_sub_type = DWSS_BINDING_TABLE_STATE; + break; + case BRW_CONSTANT_BUFFER: + aub_type = DW_CONSTANT_URB_ENTRY; + aub_sub_type = 0; + break; + + default: + assert(0); + break; + } + + xmesa_buffer_subdata_aub( iws->pipe_winsys, + buf, + offset, + size, + data, + aub_type, + aub_sub_type ); +} + +/** + * Create i965 hardware rendering context. + */ +struct pipe_context * +xmesa_create_i965simple( struct pipe_winsys *winsys ) +{ + struct aub_brw_winsys *iws = CALLOC_STRUCT( aub_brw_winsys ); + + /* Fill in this struct with callbacks that i965simple will need to + * communicate with the window system, buffer manager, etc. + */ + iws->winsys.batch_start = aub_i965_batch_start; + iws->winsys.batch_dword = aub_i965_batch_dword; + iws->winsys.batch_reloc = aub_i965_batch_reloc; + iws->winsys.batch_end = aub_i965_batch_end; + iws->winsys.batch_flush = aub_i965_batch_flush; + iws->winsys.buffer_subdata_typed = aub_i965_buffer_subdata_typed; + iws->winsys.get_buffer_offset = aub_i965_get_buffer_offset; + + iws->pipe_winsys = winsys; + + iws->batch_size = IWS_BATCHBUFFER_SIZE; + + /* Create the i965simple context: + */ + return brw_create( winsys, + &iws->winsys, + 0 ); +} diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.h b/src/gallium/winsys/xlib/xm_winsys_aub.h new file mode 100644 index 0000000000..7bee199116 --- /dev/null +++ b/src/gallium/winsys/xlib/xm_winsys_aub.h @@ -0,0 +1,67 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef AUB_WINSYS_H +#define AUB_WINSYS_H + +struct pipe_context; +struct pipe_winsys; +struct pipe_buffer; +struct pipe_surface; + +struct pipe_winsys * +xmesa_create_pipe_winsys_aub( void ); + +void +xmesa_destroy_pipe_winsys_aub( struct pipe_winsys *winsys ); + + + +struct pipe_context * +xmesa_create_i965simple( struct pipe_winsys *winsys ); + + + +void xmesa_buffer_subdata_aub(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned long offset, + unsigned long size, + const void *data, + unsigned aub_type, + unsigned aub_sub_type); + +void xmesa_commands_aub(struct pipe_winsys *winsys, + unsigned *cmds, + unsigned nr_dwords); + + +void xmesa_display_aub( /* struct pipe_winsys *winsys, */ + struct pipe_surface *surface ); + +struct pipe_winsys *xmesa_get_pipe_winsys_aub(void); + +#endif diff --git a/src/gallium/winsys/xlib/xmesaP.h b/src/gallium/winsys/xlib/xmesaP.h new file mode 100644 index 0000000000..fa8d1f14b9 --- /dev/null +++ b/src/gallium/winsys/xlib/xmesaP.h @@ -0,0 +1,176 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef XMESAP_H +#define XMESAP_H + + +#include "GL/xmesa.h" +#include "mtypes.h" +#ifdef XFree86Server +#include "xm_image.h" +#endif + +#include "state_tracker/st_context.h" +#include "state_tracker/st_public.h" + + +extern _glthread_Mutex _xmesa_lock; + +extern XMesaBuffer XMesaBufferList; + +/* + */ +#define XMESA_SOFTPIPE 1 +#define XMESA_AUB 2 +extern int xmesa_mode; + + +/** + * Visual inforation, derived from GLvisual. + * Basically corresponds to an XVisualInfo. + */ +struct xmesa_visual { + GLvisual mesa_visual; /* Device independent visual parameters */ + XMesaDisplay *display; /* The X11 display */ +#ifdef XFree86Server + GLint ColormapEntries; + GLint nplanes; +#else + XMesaVisualInfo visinfo; /* X's visual info (pointer to private copy) */ + XVisualInfo *vishandle; /* Only used in fakeglx.c */ +#endif + GLint BitsPerPixel; /* True bits per pixel for XImages */ + + GLboolean ximage_flag; /* Use XImage for back buffer (not pixmap)? */ +}; + + +/** + * Context info, derived from st_context. + * Basically corresponds to a GLXContext. + */ +struct xmesa_context { + struct st_context *st; + XMesaVisual xm_visual; /** pixel format info */ + XMesaBuffer xm_buffer; /** current drawbuffer */ +}; + + +/** + * Types of X/GLX drawables we might render into. + */ +typedef enum { + WINDOW, /* An X window */ + GLXWINDOW, /* GLX window */ + PIXMAP, /* GLX pixmap */ + PBUFFER /* GLX Pbuffer */ +} BufferType; + + +/** + * Framebuffer information, derived from. + * Basically corresponds to a GLXDrawable. + */ +struct xmesa_buffer { + struct st_framebuffer *stfb; + + GLboolean wasCurrent; /* was ever the current buffer? */ + XMesaVisual xm_visual; /* the X/Mesa visual */ + XMesaDrawable drawable; /* Usually the X window ID */ + XMesaColormap cmap; /* the X colormap */ + BufferType type; /* window, pixmap, pbuffer or glxwindow */ + + XMesaImage *tempImage; + unsigned long selectedEvents;/* for pbuffers only */ + + GLuint shm; /* X Shared Memory extension status: */ + /* 0 = not available */ + /* 1 = XImage support available */ + /* 2 = Pixmap support available too */ +#if defined(USE_XSHM) && !defined(XFree86Server) + XShmSegmentInfo shminfo; +#endif + + XMesaGC gc; /* scratch GC for span, line, tri drawing */ + + /* GLX_EXT_texture_from_pixmap */ + GLint TextureTarget; /** GLX_TEXTURE_1D_EXT, for example */ + GLint TextureFormat; /** GLX_TEXTURE_FORMAT_RGB_EXT, for example */ + GLint TextureMipmap; /** 0 or 1 */ + + struct xmesa_buffer *Next; /* Linked list pointer: */ +}; + + + +/** cast wrapper */ +static INLINE XMesaContext +xmesa_context(GLcontext *ctx) +{ + return (XMesaContext) ctx->DriverCtx; +} + + +/** cast wrapper */ +static INLINE XMesaBuffer +xmesa_buffer(GLframebuffer *fb) +{ + struct st_framebuffer *stfb = (struct st_framebuffer *) fb; + return (XMesaBuffer) st_framebuffer_private(stfb); +} + + +extern void +xmesa_delete_framebuffer(struct gl_framebuffer *fb); + +extern XMesaBuffer +xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis); + +extern void +xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer); + +extern void +xmesa_destroy_buffers_on_display(XMesaDisplay *dpy); + +extern struct pipe_context * +xmesa_create_pipe_context(XMesaContext xm, uint pixelformat); + +static INLINE GLuint +xmesa_buffer_width(XMesaBuffer b) +{ + return b->stfb->Base.Width; +} + +static INLINE GLuint +xmesa_buffer_height(XMesaBuffer b) +{ + return b->stfb->Base.Height; +} + +extern void +xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf); + +#endif diff --git a/src/mesa/drivers/dri/intel_winsys/Makefile b/src/mesa/drivers/dri/intel_winsys/Makefile deleted file mode 100644 index 9ae0f01325..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/Makefile +++ /dev/null @@ -1,38 +0,0 @@ - -TOP = ../../../../.. -include $(TOP)/configs/current - -LIBNAME = i915tex_dri.so - -MINIGLX_SOURCES = server/intel_dri.c - -PIPE_DRIVERS = \ - $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a \ - $(TOP)/src/mesa/pipe/i915simple/libi915simple.a - -DRIVER_SOURCES = \ - intel_winsys_pipe.c \ - intel_winsys_softpipe.c \ - intel_winsys_i915.c \ - intel_batchbuffer.c \ - intel_swapbuffers.c \ - intel_context.c \ - intel_lock.c \ - intel_screen.c \ - intel_batchpool.c - -C_SOURCES = \ - $(COMMON_GALLIUM_SOURCES) \ - $(COMMON_BM_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") - -include ../Makefile.template - -intel_tex_layout.o: ../intel/intel_tex_layout.c - -symlinks: diff --git a/src/mesa/drivers/dri/intel_winsys/SConscript b/src/mesa/drivers/dri/intel_winsys/SConscript deleted file mode 100644 index a7cc10450e..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/SConscript +++ /dev/null @@ -1,41 +0,0 @@ -Import('*') - -env = drienv.Clone() - -env.Append(CPPPATH = [ - '../intel', - 'server' -]) - -#MINIGLX_SOURCES = server/intel_dri.c - -pipe_drivers = [ - softpipe, - i915simple -] - -DRIVER_SOURCES = [ - 'intel_winsys_pipe.c', - 'intel_winsys_softpipe.c', - 'intel_winsys_i915.c', - 'intel_batchbuffer.c', - 'intel_swapbuffers.c', - 'intel_context.c', - 'intel_lock.c', - 'intel_screen.c', - 'intel_batchpool.c', -] - -sources = \ - COMMON_GALLIUM_SOURCES + \ - COMMON_BM_SOURCES + \ - DRIVER_SOURCES - -# DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ -# && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") - -env.SharedLibrary( - target ='i915tex_dri.so', - source = sources, - LIBS = pipe_drivers + env['LIBS'], -) \ No newline at end of file diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c deleted file mode 100644 index 49e04d81ec..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.c +++ /dev/null @@ -1,357 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include -#include "intel_batchbuffer.h" -#include "intel_context.h" -#include "intel_screen.h" -#include "intel_reg.h" -#include "drm.h" - -/* Relocations in kernel space: - * - pass dma buffer seperately - * - memory manager knows how to patch - * - pass list of dependent buffers - * - pass relocation list - * - * Either: - * - get back an offset for buffer to fire - * - memory manager knows how to fire buffer - * - * Really want the buffer to be AGP and pinned. - * - */ - -/* Cliprect fence: The highest fence protecting a dma buffer - * containing explicit cliprect information. Like the old drawable - * lock but irq-driven. X server must wait for this fence to expire - * before changing cliprects [and then doing sw rendering?]. For - * other dma buffers, the scheduler will grab current cliprect info - * and mix into buffer. X server must hold the lock while changing - * cliprects??? Make per-drawable. Need cliprects in shared memory - * -- beats storing them with every cmd buffer in the queue. - * - * ==> X server must wait for this fence to expire before touching the - * framebuffer with new cliprects. - * - * ==> Cliprect-dependent buffers associated with a - * cliprect-timestamp. All of the buffers associated with a timestamp - * must go to hardware before any buffer with a newer timestamp. - * - * ==> Dma should be queued per-drawable for correct X/GL - * synchronization. Or can fences be used for this? - * - * Applies to: Blit operations, metaops, X server operations -- X - * server automatically waits on its own dma to complete before - * modifying cliprects ??? - */ - -static void -intel_dump_batchbuffer(uint offset, uint * ptr, uint count) -{ - int i; - printf("\n\n\nSTART BATCH (%d dwords):\n", count / 4); - for (i = 0; i < count / 4; i += 1) - printf("\t0x%08x\n", ptr[i]); - printf("END BATCH\n\n\n"); -} - - -void -intel_batchbuffer_reset(struct intel_batchbuffer *batch) -{ - int i; - - if (batch->map) { - driBOUnmap(batch->buffer); - batch->map = NULL; - } - - /* - * Get a new, free batchbuffer. - */ - batch->size = BATCH_SZ; - driBOData(batch->buffer, batch->size, NULL, 0); - - driBOResetList(&batch->list); - - /* - * Unreference buffers previously on the relocation list. - */ - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - driBOUnReference(r->buf); - } - - batch->list_count = 0; - batch->nr_relocs = 0; - batch->flags = 0; - - /* - * We don't refcount the batchbuffer itself since we can't destroy it - * while it's on the list. - */ - - driBOAddListItem(&batch->list, batch->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE); - - - batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); - batch->ptr = batch->map; -} - - -/*====================================================================== - * Public functions - */ -struct intel_batchbuffer * -intel_batchbuffer_alloc(struct intel_context *intel) -{ - struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); - - batch->intel = intel; - - driGenBuffers(intel->intelScreen->batchPool, "batchbuffer", 1, - &batch->buffer, 4096, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); - batch->last_fence = NULL; - driBOCreateList(20, &batch->list); - intel_batchbuffer_reset(batch); - return batch; -} - - -void -intel_batchbuffer_free(struct intel_batchbuffer *batch) -{ - if (batch->last_fence) { - driFenceFinish(batch->last_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, - GL_FALSE); - driFenceUnReference(batch->last_fence); - batch->last_fence = NULL; - } - if (batch->map) { - driBOUnmap(batch->buffer); - batch->map = NULL; - } - driBOUnReference(batch->buffer); - batch->buffer = NULL; - free(batch); -} - - -static void -intel_batch_ioctl(struct intel_context *intel, - uint start_offset, uint used, boolean allow_unlock) -{ - drmI830BatchBuffer batch; - - batch.start = start_offset; - batch.used = used; - batch.cliprects = NULL; /* unused */ - batch.num_cliprects = 0; - batch.DR1 = 0; - batch.DR4 = 0; /* still need this ? */ - - DBG(IOCTL, "%s: 0x%x..0x%x DR4: %x cliprects: %d\n", - __FUNCTION__, - batch.start, - batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); - - if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, - sizeof(batch))) { - printf("DRM_I830_BATCHBUFFER: %d\n", -errno); - UNLOCK_HARDWARE(intel); - exit(1); - } -} - - -/* TODO: Push this whole function into bufmgr. - */ -static void -do_flush_locked(struct intel_batchbuffer *batch, - uint used, boolean allow_unlock) -{ - uint *ptr; - uint i, fenceFlags; - struct _DriFenceObject *fo; - - driBOValidateList(batch->intel->driFd, &batch->list); - - /* Apply the relocations. This nasty map indicates to me that the - * whole task should be done internally by the memory manager, and - * that dma buffers probably need to be pinned within agp space. - */ - ptr = (uint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, - DRM_BO_HINT_ALLOW_UNFENCED_MAP); - - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - - ptr[r->offset / 4] = driBOOffset(r->buf) + r->delta; - } - - if (0) - intel_dump_batchbuffer(0, ptr, used); - - driBOUnmap(batch->buffer); - batch->map = NULL; - - intel_batch_ioctl(batch->intel, - driBOOffset(batch->buffer), - used, allow_unlock); - - /* - * Kernel fencing. The flags tells the kernel that we've - * programmed an MI_FLUSH. - */ - fenceFlags = DRM_I915_FENCE_FLAG_FLUSHED; - fo = driFenceBuffers(batch->intel->driFd, "Batch fence", fenceFlags); - - /* - * User space fencing. - */ - driBOFence(batch->buffer, fo); - - if (driFenceType(fo) == DRM_FENCE_TYPE_EXE) { - /* - * Oops. We only validated a batch buffer. This means we - * didn't do any proper rendering. Discard this fence object. - */ - driFenceUnReference(fo); - } - else { - driFenceUnReference(batch->last_fence); - batch->last_fence = fo; - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - driBOFence(r->buf, fo); - } - } -} - - -struct _DriFenceObject * -intel_batchbuffer_flush(struct intel_batchbuffer *batch) -{ - struct intel_context *intel = batch->intel; - uint used = batch->ptr - batch->map; - const boolean was_locked = intel->locked; - - if (used == 0) - return batch->last_fence; - -#define MI_FLUSH ((0 << 29) | (4 << 23)) - - /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a - * performance drain that we would like to avoid. - */ - if (used & 4) { - ((int *) batch->ptr)[0] = MI_FLUSH; - ((int *) batch->ptr)[1] = 0; - ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->ptr)[0] = MI_FLUSH; - ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; - used += 8; - } - - driBOUnmap(batch->buffer); - batch->ptr = NULL; - batch->map = NULL; - - /* TODO: Just pass the relocation list and dma buffer up to the - * kernel. - */ - if (!was_locked) - LOCK_HARDWARE(intel); - - do_flush_locked(batch, used, GL_FALSE); - - if (!was_locked) - UNLOCK_HARDWARE(intel); - - /* Reset the buffer: - */ - intel_batchbuffer_reset(batch); - return batch->last_fence; -} - - -void -intel_batchbuffer_finish(struct intel_batchbuffer *batch) -{ - struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); - if (fence) { - driFenceReference(fence); - driFenceFinish(fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, - GL_FALSE); - driFenceUnReference(fence); - } -} - - -/* This is the only way buffers get added to the validate list. - */ -boolean -intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, - struct _DriBufferObject *buffer, - uint flags, uint mask, uint delta) -{ - assert(batch->nr_relocs < MAX_RELOCS); - - driBOAddListItem(&batch->list, buffer, flags, mask); - - { - struct buffer_reloc *r = &batch->reloc[batch->nr_relocs++]; - driBOReference(buffer); - r->buf = buffer; - r->offset = batch->ptr - batch->map; - r->delta = delta; - *(uint *) batch->ptr = 0x12345678; - } - - batch->ptr += 4; - return GL_TRUE; -} - - -void -intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, uint bytes, uint flags) -{ - assert((bytes & 3) == 0); - intel_batchbuffer_require_space(batch, bytes, flags); - memcpy(batch->ptr, data, bytes); - batch->ptr += bytes; -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h b/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h deleted file mode 100644 index 82feafa21f..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchbuffer.h +++ /dev/null @@ -1,149 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BATCHBUFFER_H -#define INTEL_BATCHBUFFER_H - -#include "pipe/p_compiler.h" -#include "dri_bufmgr.h" - -struct intel_context; - -#define BATCH_SZ 16384 -#define BATCH_RESERVED 16 - -#define MAX_RELOCS 4096 - -#define INTEL_BATCH_NO_CLIPRECTS 0x1 -#define INTEL_BATCH_CLIPRECTS 0x2 - -struct buffer_reloc -{ - struct _DriBufferObject *buf; - uint offset; - uint delta; /* not needed? */ -}; - -struct intel_batchbuffer -{ - struct bufmgr *bm; - struct intel_context *intel; - - struct _DriBufferObject *buffer; - struct _DriFenceObject *last_fence; - uint flags; - - drmBOList list; - uint list_count; - ubyte *map; - ubyte *ptr; - - struct buffer_reloc reloc[MAX_RELOCS]; - uint nr_relocs; - uint size; -}; - -struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context *intel); - -void intel_batchbuffer_free(struct intel_batchbuffer *batch); - - -void intel_batchbuffer_finish(struct intel_batchbuffer *batch); - -struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer - *batch); - -void intel_batchbuffer_reset(struct intel_batchbuffer *batch); - - -/* Unlike bmBufferData, this currently requires the buffer be mapped. - * Consider it a convenience function wrapping multiple - * intel_buffer_dword() calls. - */ -void intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, uint bytes, uint flags); - -void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, - uint bytes); - -boolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, - struct _DriBufferObject *buffer, - uint flags, - uint mask, uint offset); - -/* Inline functions - might actually be better off with these - * non-inlined. Certainly better off switching all command packets to - * be passed as structs rather than dwords, but that's a little bit of - * work... - */ -static INLINE uint -intel_batchbuffer_space(struct intel_batchbuffer *batch) -{ - return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); -} - - -static INLINE void -intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, uint dword) -{ - assert(batch->map); - assert(intel_batchbuffer_space(batch) >= 4); - *(uint *) (batch->ptr) = dword; - batch->ptr += 4; -} - -static INLINE void -intel_batchbuffer_require_space(struct intel_batchbuffer *batch, - uint sz, uint flags) -{ - assert(sz < batch->size - 8); - if (intel_batchbuffer_space(batch) < sz || - (batch->flags != 0 && flags != 0 && batch->flags != flags)) - intel_batchbuffer_flush(batch); - - batch->flags |= flags; -} - -/* Here are the crusty old macros, to be removed: - */ -#define BATCH_LOCALS - -#define BEGIN_BATCH(n, flags) do { \ - intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ -} while (0) - -#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) - -#define OUT_RELOC(buf,flags,mask,delta) do { \ - assert((delta) >= 0); \ - intel_batchbuffer_emit_reloc(intel->batch, buf, flags, mask, delta); \ -} while (0) - -#define ADVANCE_BATCH() do { } while(0) - - -#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c deleted file mode 100644 index 33b56817f6..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.c +++ /dev/null @@ -1,424 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -/** - * XXX NOTE: there are no intel dependencies in this file. - * Rename to dri_batchpool.c? - */ - -#include -#include -#include -#include "imports.h" -#include "glthread.h" -#include "dri_bufpool.h" -#include "dri_bufmgr.h" -#include "intel_batchpool.h" - - -typedef struct -{ - drmMMListHead head; - struct _BPool *parent; - struct _DriFenceObject *fence; - unsigned long start; - int unfenced; - int mapped; -} BBuf; - -typedef struct _BPool -{ - _glthread_Mutex mutex; - unsigned long bufSize; - unsigned poolSize; - unsigned numFree; - unsigned numTot; - unsigned numDelayed; - unsigned checkDelayed; - drmMMListHead free; - drmMMListHead delayed; - drmMMListHead head; - drmBO kernelBO; - void *virtual; - BBuf *bufs; -} BPool; - - -static BPool * -createBPool(int fd, unsigned long bufSize, unsigned numBufs, unsigned flags, - unsigned checkDelayed) -{ - BPool *p = (BPool *) malloc(sizeof(*p)); - BBuf *buf; - int i; - - if (!p) - return NULL; - - p->bufs = (BBuf *) malloc(numBufs * sizeof(*p->bufs)); - if (!p->bufs) { - free(p); - return NULL; - } - - DRMINITLISTHEAD(&p->free); - DRMINITLISTHEAD(&p->head); - DRMINITLISTHEAD(&p->delayed); - - p->numTot = numBufs; - p->numFree = numBufs; - p->bufSize = bufSize; - p->numDelayed = 0; - p->checkDelayed = checkDelayed; - - _glthread_INIT_MUTEX(p->mutex); - - if (drmBOCreate(fd, 0, numBufs * bufSize, 0, NULL, drm_bo_type_dc, - flags, DRM_BO_HINT_DONT_FENCE, &p->kernelBO)) { - free(p->bufs); - free(p); - return NULL; - } - if (drmBOMap(fd, &p->kernelBO, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, - &p->virtual)) { - drmBODestroy(fd, &p->kernelBO); - free(p->bufs); - free(p); - return NULL; - } - - /* - * We unmap the buffer so that we can validate it later. Note that this is - * just a synchronizing operation. The buffer will have a virtual mapping - * until it is destroyed. - */ - - drmBOUnmap(fd, &p->kernelBO); - - buf = p->bufs; - for (i = 0; i < numBufs; ++i) { - buf->parent = p; - buf->fence = NULL; - buf->start = i * bufSize; - buf->mapped = 0; - buf->unfenced = 0; - DRMLISTADDTAIL(&buf->head, &p->free); - buf++; - } - - return p; -} - - -static void -pool_checkFree(BPool * p, int wait) -{ - drmMMListHead *list, *prev; - BBuf *buf; - int signaled = 0; - int i; - - list = p->delayed.next; - - if (p->numDelayed > 3) { - for (i = 0; i < p->numDelayed; i += 3) { - list = list->next; - } - } - - prev = list->prev; - for (; list != &p->delayed; list = prev, prev = list->prev) { - - buf = DRMLISTENTRY(BBuf, list, head); - - if (!signaled) { - if (wait) { - driFenceFinish(buf->fence, DRM_FENCE_TYPE_EXE, 1); - signaled = 1; - } - else { - signaled = driFenceSignaled(buf->fence, DRM_FENCE_TYPE_EXE); - } - } - - if (!signaled) - break; - - driFenceUnReference(buf->fence); - buf->fence = NULL; - DRMLISTDEL(list); - p->numDelayed--; - DRMLISTADD(list, &p->free); - p->numFree++; - } -} - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, unsigned flags, unsigned hint, - unsigned alignment) -{ - BPool *p = (BPool *) pool->data; - - drmMMListHead *item; - - if (alignment && (alignment != 4096)) - return NULL; - - _glthread_LOCK_MUTEX(p->mutex); - - if (p->numFree == 0) - pool_checkFree(p, GL_TRUE); - - if (p->numFree == 0) { - fprintf(stderr, "Out of fixed size buffer objects\n"); - BM_CKFATAL(-ENOMEM); - } - - item = p->free.next; - - if (item == &p->free) { - fprintf(stderr, "Fixed size buffer pool corruption\n"); - } - - DRMLISTDEL(item); - --p->numFree; - - _glthread_UNLOCK_MUTEX(p->mutex); - return (void *) DRMLISTENTRY(BBuf, item, head); -} - - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - - if (buf->fence) { - DRMLISTADDTAIL(&buf->head, &p->delayed); - p->numDelayed++; - } - else { - buf->unfenced = 0; - DRMLISTADD(&buf->head, &p->free); - p->numFree++; - } - - if ((p->numDelayed % p->checkDelayed) == 0) - pool_checkFree(p, 0); - - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, void **virtual) -{ - - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - - /* - * Currently Mesa doesn't have any condition variables to resolve this - * cleanly in a multithreading environment. - * We bail out instead. - */ - - if (buf->mapped) { - fprintf(stderr, "Trying to map already mapped buffer object\n"); - BM_CKFATAL(-EINVAL); - } - -#if 0 - if (buf->unfenced && !(hint & DRM_BO_HINT_ALLOW_UNFENCED_MAP)) { - fprintf(stderr, "Trying to map an unfenced buffer object 0x%08x" - " 0x%08x %d\n", hint, flags, buf->start); - BM_CKFATAL(-EINVAL); - } - -#endif - - if (buf->fence) { - _glthread_UNLOCK_MUTEX(p->mutex); - return -EBUSY; - } - - buf->mapped = GL_TRUE; - *virtual = (unsigned char *) p->virtual + buf->start; - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, int lazy) -{ - BBuf *buf = (BBuf *) private; - driFenceFinish(buf->fence, 0x0, lazy); - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - - buf->mapped = 0; - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - return p->kernelBO.offset + buf->start; -} - -static unsigned -pool_flags(struct _DriBufferPool *pool, void *private) -{ - BPool *p = (BPool *) pool->data; - - return p->kernelBO.flags; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - BPool *p = (BPool *) pool->data; - - return p->bufSize; -} - - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - if (buf->fence) { - driFenceUnReference(buf->fence); - } - buf->fence = fence; - buf->unfenced = 0; - driFenceReference(buf->fence); - _glthread_UNLOCK_MUTEX(p->mutex); - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - return &p->kernelBO; -} - -static int -pool_validate(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - _glthread_LOCK_MUTEX(p->mutex); - buf->unfenced = GL_TRUE; - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - BPool *p = (BPool *) pool->data; - - /* - * Wait on outstanding fences. - */ - - _glthread_LOCK_MUTEX(p->mutex); - while ((p->numFree < p->numTot) && p->numDelayed) { - _glthread_UNLOCK_MUTEX(p->mutex); - sched_yield(); - pool_checkFree(p, GL_TRUE); - _glthread_LOCK_MUTEX(p->mutex); - } - - drmBODestroy(pool->fd, &p->kernelBO); - free(p->bufs); - _glthread_UNLOCK_MUTEX(p->mutex); - free(p); - free(pool); -} - - -struct _DriBufferPool * -driBatchPoolInit(int fd, unsigned flags, - unsigned long bufSize, - unsigned numBufs, unsigned checkDelayed) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - if (!pool) - return NULL; - - pool->data = createBPool(fd, bufSize, numBufs, flags, checkDelayed); - if (!pool->data) - return NULL; - - pool->fd = fd; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = &pool_validate; - pool->waitIdle = &pool_waitIdle; - pool->setstatic = NULL; - pool->takeDown = &pool_takedown; - return pool; -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.h b/src/mesa/drivers/dri/intel_winsys/intel_batchpool.h deleted file mode 100644 index f6a95723bc..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_batchpool.h +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BATCHPOOL_H -#define INTEL_BATCHPOOL_H - -extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, - unsigned long bufSize, - unsigned numBufs, - unsigned checkDelayed); - - -#endif /* INTEL_BATCHPOOL_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.c b/src/mesa/drivers/dri/intel_winsys/intel_context.c deleted file mode 100644 index c033f2a592..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.c +++ /dev/null @@ -1,304 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "i830_dri.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_swapbuffers.h" -#include "intel_winsys.h" -#include "intel_batchbuffer.h" - -#include "state_tracker/st_public.h" -#include "pipe/p_defines.h" -#include "pipe/p_context.h" - -#include "utils.h" - - -#ifdef DEBUG -int __intel_debug = 0; -#endif - - -#define need_GL_ARB_multisample -#define need_GL_ARB_point_parameters -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_buffer_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_window_pos -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_equation_separate -#define need_GL_EXT_blend_func_separate -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_cull_vertex -#define need_GL_EXT_fog_coord -#define need_GL_EXT_framebuffer_object -#define need_GL_EXT_multi_draw_arrays -#define need_GL_EXT_secondary_color -#define need_GL_NV_vertex_program -#include "extension_helper.h" - - -/** - * Extension strings exported by the intel driver. - * - * \note - * It appears that ARB_texture_env_crossbar has "disappeared" compared to the - * old i830-specific driver. - */ -const struct dri_extension card_extensions[] = { - {"GL_ARB_multisample", GL_ARB_multisample_functions}, - {"GL_ARB_multitexture", NULL}, - {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, - {"GL_ARB_texture_border_clamp", NULL}, - {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, - {"GL_ARB_texture_cube_map", NULL}, - {"GL_ARB_texture_env_add", NULL}, - {"GL_ARB_texture_env_combine", NULL}, - {"GL_ARB_texture_env_dot3", NULL}, - {"GL_ARB_texture_mirrored_repeat", NULL}, - {"GL_ARB_texture_rectangle", NULL}, - {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, - {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, - {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, - {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, - {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, - {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_subtract", NULL}, - {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, - {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, - {"GL_EXT_packed_depth_stencil", NULL}, - {"GL_EXT_pixel_buffer_object", NULL}, - {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, - {"GL_EXT_stencil_wrap", NULL}, - {"GL_EXT_texture_edge_clamp", NULL}, - {"GL_EXT_texture_env_combine", NULL}, - {"GL_EXT_texture_env_dot3", NULL}, - {"GL_EXT_texture_filter_anisotropic", NULL}, - {"GL_EXT_texture_lod_bias", NULL}, - {"GL_3DFX_texture_compression_FXT1", NULL}, - {"GL_APPLE_client_storage", NULL}, - {"GL_MESA_pack_invert", NULL}, - {"GL_MESA_ycbcr_texture", NULL}, - {"GL_NV_blend_square", NULL}, - {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, - {"GL_NV_vertex_program1_1", NULL}, - {"GL_SGIS_generate_mipmap", NULL }, - {NULL, NULL} -}; - - - -#ifdef DEBUG -static const struct dri_debug_control debug_control[] = { - {"ioctl", DEBUG_IOCTL}, - {"bat", DEBUG_BATCH}, - {"lock", DEBUG_LOCK}, - {"swap", DEBUG_SWAP}, - {NULL, 0} -}; -#endif - - - -GLboolean -intelCreateContext(const __GLcontextModes * visual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate) -{ - struct intel_context *intel = CALLOC_STRUCT(intel_context); - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - struct intel_screen *intelScreen = intel_screen(sPriv); - drmI830Sarea *saPriv = intelScreen->sarea; - int fthrottle_mode; - GLboolean havePools; - struct pipe_context *pipe; - struct st_context *st_share = NULL; - - if (sharedContextPrivate) { - st_share = ((struct intel_context *) sharedContextPrivate)->st; - } - - driContextPriv->driverPrivate = intel; - intel->intelScreen = intelScreen; - intel->driScreen = sPriv; - intel->sarea = saPriv; - - driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, - intel->driScreen->myNum, "i915"); - - - /* - * memory pools - */ - DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - havePools = intelCreatePools(sPriv); - DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - if (!havePools) - return GL_FALSE; - - - /* Dri stuff */ - intel->hHWContext = driContextPriv->hHWContext; - intel->driFd = sPriv->fd; - intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; - - fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->iw.irq_seq = -1; - intel->irqsEmitted = 0; - - intel->batch = intel_batchbuffer_alloc(intel); - intel->last_swap_fence = NULL; - intel->first_swap_fence = NULL; - -#ifdef DEBUG - __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); -#endif - - /* - * Pipe-related setup - */ - if (!getenv("INTEL_HW")) { - pipe = intel_create_softpipe( intel, intelScreen->winsys ); - } - else { - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - pipe = intel_create_i915simple( intel, intelScreen->winsys ); - break; - default: - fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", - intel->intelScreen->deviceID, __FUNCTION__); - - pipe = intel_create_softpipe( intel, intelScreen->winsys ); - break; - } - } - - pipe->priv = intel; - - intel->st = st_create_context(pipe, visual, st_share); - - return GL_TRUE; -} - - -void -intelDestroyContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = intel_context(driContextPriv); - - assert(intel); /* should never be null */ - if (intel) { - st_flush(intel->st, PIPE_FLUSH_WAIT); - - intel_batchbuffer_free(intel->batch); - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - if (intel->first_swap_fence) { - driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->first_swap_fence); - intel->first_swap_fence = NULL; - } - - if (intel->intelScreen->dummyContext == intel) - intel->intelScreen->dummyContext = NULL; - - st_destroy_context(intel->st); - free(intel); - } -} - - -GLboolean -intelUnbindContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = intel_context(driContextPriv); - st_flush(intel->st, 0x0); - /* XXX make_current(NULL)? */ - return GL_TRUE; -} - - -GLboolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv) -{ - if (driContextPriv) { - struct intel_context *intel = intel_context(driContextPriv); - struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv); - struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv); - - assert(draw_fb->stfb); - assert(read_fb->stfb); - - /* This is for situations in which we need a rendering context but - * there may not be any currently bound. - */ - intel->intelScreen->dummyContext = intel; - - st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); - - if ((intel->driDrawable != driDrawPriv) || - (intel->lastStamp != driDrawPriv->lastStamp)) { - intel->driDrawable = driDrawPriv; - intelUpdateWindowSize(driDrawPriv); - intel->lastStamp = driDrawPriv->lastStamp; - } - - /* The size of the draw buffer will have been updated above. - * If the readbuffer is a different window, check/update its size now. - */ - if (driReadPriv != driDrawPriv) { - intelUpdateWindowSize(driReadPriv); - } - - } - else { - st_make_current(NULL, NULL, NULL); - } - - return GL_TRUE; -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_context.h b/src/mesa/drivers/dri/intel_winsys/intel_context.h deleted file mode 100644 index b01370c049..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_context.h +++ /dev/null @@ -1,158 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_CONTEXT_H -#define INTEL_CONTEXT_H - - -#include "drm.h" -#include "intel_screen.h" -#include "i915_drm.h" - - -struct pipe_context; -struct intel_context; -struct _DriBufferObject; -struct st_context; - - -#define INTEL_MAX_FIXUP 64 - -/** - * Intel rendering context, contains a state tracker and intel-specific info. - */ -struct intel_context -{ - struct st_context *st; - - struct _DriFenceObject *last_swap_fence; - struct _DriFenceObject *first_swap_fence; - - struct intel_batchbuffer *batch; - - boolean locked; - char *prevLockFile; - int prevLockLine; - - uint irqsEmitted; - drm_i915_irq_wait_t iw; - - drm_context_t hHWContext; - drmLock *driHwLock; - int driFd; - - __DRIdrawablePrivate *driDrawable; - __DRIscreenPrivate *driScreen; - struct intel_screen *intelScreen; - drmI830Sarea *sarea; - - uint lastStamp; - - /** - * Configuration cache - */ - driOptionCache optionCache; -}; - - - -/** - * Intel framebuffer. - */ -struct intel_framebuffer -{ - struct st_framebuffer *stfb; - - /* other fields TBD */ - int other; -}; - - - - -/* These are functions now: - */ -void LOCK_HARDWARE( struct intel_context *intel ); -void UNLOCK_HARDWARE( struct intel_context *intel ); - -extern char *__progname; - - - -/* ================================================================ - * Debugging: - */ -#ifdef DEBUG -extern int __intel_debug; - -#define DEBUG_SWAP 0x1 -#define DEBUG_LOCK 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_BATCH 0x8 - -#define DBG(flag, ...) do { \ - if (__intel_debug & (DEBUG_##flag)) \ - printf(__VA_ARGS__); \ -} while(0) - -#else -#define DBG(flag, ...) -#endif - - - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GME 0x27AE -#define PCI_CHIP_G33_G 0x29C2 -#define PCI_CHIP_Q35_G 0x29B2 -#define PCI_CHIP_Q33_G 0x29D2 - - -/** Cast wrapper */ -static INLINE struct intel_context * -intel_context(__DRIcontextPrivate *driContextPriv) -{ - return (struct intel_context *) driContextPriv->driverPrivate; -} - - -/** Cast wrapper */ -static INLINE struct intel_framebuffer * -intel_framebuffer(__DRIdrawablePrivate * driDrawPriv) -{ - return (struct intel_framebuffer *) driDrawPriv->driverPrivate; -} - - -#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_lock.c b/src/mesa/drivers/dri/intel_winsys/intel_lock.c deleted file mode 100644 index 70aa7ea5f4..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_lock.c +++ /dev/null @@ -1,102 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "main/glheader.h" -#include "glapi/glthread.h" -#include -#include "state_tracker/st_public.h" -#include "intel_context.h" -#include "i830_dri.h" - - - -_glthread_DECLARE_STATIC_MUTEX( lockMutex ); - - -static void -intelContendedLock(struct intel_context *intel, uint flags) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - __DRIscreenPrivate *sPriv = intel->driScreen; - struct intel_screen *intelScreen = intel_screen(sPriv); - drmI830Sarea *sarea = intel->sarea; - - drmGetLock(intel->driFd, intel->hHWContext, flags); - - DBG(LOCK, "%s - got contended lock\n", __progname); - - /* If the window moved, may need to set a new cliprect now. - * - * NOTE: This releases and regains the hw lock, so all state - * checking must be done *after* this call: - */ - if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - - if (sarea->width != intelScreen->front.width || - sarea->height != intelScreen->front.height) { - - intelUpdateScreenRotation(sPriv, sarea); - } -} - - -/* Lock the hardware and validate our state. - */ -void LOCK_HARDWARE( struct intel_context *intel ) -{ - char __ret = 0; - - _glthread_LOCK_MUTEX(lockMutex); - assert(!intel->locked); - - DRM_CAS(intel->driHwLock, intel->hHWContext, - (DRM_LOCK_HELD|intel->hHWContext), __ret); - - if (__ret) - intelContendedLock( intel, 0 ); - - DBG(LOCK, "%s - locked\n", __progname); - - intel->locked = 1; -} - - -/* Unlock the hardware using the global current context - */ -void UNLOCK_HARDWARE( struct intel_context *intel ) -{ - assert(intel->locked); - intel->locked = 0; - - DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); - - _glthread_UNLOCK_MUTEX(lockMutex); - - DBG(LOCK, "%s - unlocked\n", __progname); -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_reg.h b/src/mesa/drivers/dri/intel_winsys/intel_reg.h deleted file mode 100644 index f37c24fda9..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_reg.h +++ /dev/null @@ -1,53 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef _INTEL_REG_H_ -#define _INTEL_REG_H_ - - -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 - -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) - -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) - -#define MI_WAIT_FOR_EVENT ((0x3<<23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) - -#define MI_BATCH_BUFFER_END (0xA<<23) - - -#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.c b/src/mesa/drivers/dri/intel_winsys/intel_screen.c deleted file mode 100644 index 9e31c013a9..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.c +++ /dev/null @@ -1,537 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "utils.h" -#include "vblank.h" -#include "xmlpool.h" - -#include "intel_context.h" -#include "intel_screen.h" -#include "intel_batchbuffer.h" -#include "intel_batchpool.h" -#include "intel_swapbuffers.h" -#include "intel_winsys.h" - -#include "i830_dri.h" -#include "dri_bufpool.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_cb_fbo.h" - - - -PUBLIC const char __driConfigOptions[] = - DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE - DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) - DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) - DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY - DRI_CONF_FORCE_S3TC_ENABLE(false) - DRI_CONF_ALLOW_LARGE_TEXTURES(1) - DRI_CONF_SECTION_END DRI_CONF_END; - -const uint __driNConfigOptions = 4; - -#ifdef USE_NEW_INTERFACE -static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; -#endif /*USE_NEW_INTERFACE */ - -extern const struct dri_extension card_extensions[]; - - - - -static void -intelPrintDRIInfo(struct intel_screen * intelScreen, - __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) -{ - fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", - intelScreen->front.size, intelScreen->front.offset, - intelScreen->front.pitch); - fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); -} - - -#if 0 -static void -intelPrintSAREA(const drmI830Sarea * sarea) -{ - fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, - sarea->height); - fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); - fprintf(stderr, - "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->front_offset, sarea->front_size, - (unsigned) sarea->front_handle); - fprintf(stderr, - "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->back_offset, sarea->back_size, - (unsigned) sarea->back_handle); - fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->depth_offset, sarea->depth_size, - (unsigned) sarea->depth_handle); - fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); - fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); - fprintf(stderr, - "SAREA: rotated offset: 0x%08x size: 0x%x\n", - sarea->rotated_offset, sarea->rotated_size); - fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); -} -#endif - - -/** - * Use the information in the sarea to update the screen parameters - * related to screen rotation. Needs to be called locked. - */ -void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) -{ - struct intel_screen *intelScreen = intel_screen(sPriv); - - if (intelScreen->front.map) { - drmUnmap(intelScreen->front.map, intelScreen->front.size); - intelScreen->front.map = NULL; - } - - if (intelScreen->front.buffer) - driDeleteBuffers(1, &intelScreen->front.buffer); - - intelScreen->front.width = sarea->width; - intelScreen->front.height = sarea->height; - intelScreen->front.offset = sarea->front_offset; - intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; - intelScreen->front.size = sarea->front_size; - intelScreen->front.handle = sarea->front_handle; - - assert( sarea->front_size >= - intelScreen->front.pitch * intelScreen->front.height ); - - if (!sarea->front_handle) - return; - - if (drmMap(sPriv->fd, - sarea->front_handle, - intelScreen->front.size, - (drmAddress *) & intelScreen->front.map) != 0) { - fprintf(stderr, "drmMap(frontbuffer) failed!\n"); - return; - } - - if (intelScreen->staticPool) { - driGenBuffers(intelScreen->staticPool, "static region", 1, - &intelScreen->front.buffer, 64, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - - driBOSetStatic(intelScreen->front.buffer, - intelScreen->front.offset, - intelScreen->front.pitch * intelScreen->front.height, - intelScreen->front.map, 0); - } -} - - -boolean -intelCreatePools(__DRIscreenPrivate * sPriv) -{ - unsigned batchPoolSize = 1024*1024; - struct intel_screen *intelScreen = intel_screen(sPriv); - - if (intelScreen->havePools) - return GL_TRUE; - - intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); - if (!intelScreen->staticPool) - return GL_FALSE; - - batchPoolSize /= BATCH_SZ; - intelScreen->batchPool = driBatchPoolInit(sPriv->fd, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT | - DRM_BO_FLAG_MEM_LOCAL, - BATCH_SZ, - batchPoolSize, 5); - if (!intelScreen->batchPool) { - fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); - return GL_FALSE; - } - - intelScreen->havePools = GL_TRUE; - - intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - return GL_TRUE; -} - - -static boolean -intelInitDriver(__DRIscreenPrivate * sPriv) -{ - struct intel_screen *intelScreen; - I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; - - PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = - (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> - getProcAddress("glxEnableExtension")); - void *const psc = sPriv->psc->screenConfigs; - - if (sPriv->devPrivSize != sizeof(I830DRIRec)) { - fprintf(stderr, - "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); - return GL_FALSE; - } - - /* Allocate the private area */ - intelScreen = CALLOC_STRUCT(intel_screen); - if (!intelScreen) - return GL_FALSE; - - /* parse information in __driConfigOptions */ - driParseOptionInfo(&intelScreen->optionCache, - __driConfigOptions, __driNConfigOptions); - - sPriv->private = (void *) intelScreen; - - intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + - gDRIPriv->sarea_priv_offset); - intelScreen->deviceID = gDRIPriv->deviceID; - intelScreen->front.cpp = gDRIPriv->cpp; - intelScreen->drmMinor = sPriv->drmMinor; - - assert(gDRIPriv->bitsPerPixel == 16 || - gDRIPriv->bitsPerPixel == 32); - - intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - if (0) - intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); - - if (glx_enable_extension != NULL) { - (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); - (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); - (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); - } - - intelScreen->winsys = intel_create_pipe_winsys(sPriv->fd); - - return GL_TRUE; -} - - -static void -intelDestroyScreen(__DRIscreenPrivate * sPriv) -{ - struct intel_screen *intelScreen = intel_screen(sPriv); - - /* intelUnmapScreenRegions(intelScreen); */ - - if (intelScreen->havePools) { - driPoolTakeDown(intelScreen->staticPool); - driPoolTakeDown(intelScreen->batchPool); - } - FREE(intelScreen); - sPriv->private = NULL; -} - - -/** - * This is called when we need to set up GL rendering to a new X window. - */ -static boolean -intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, - __DRIdrawablePrivate * driDrawPriv, - const __GLcontextModes * visual, boolean isPixmap) -{ - if (isPixmap) { - return GL_FALSE; /* not implemented */ - } - else { - enum pipe_format colorFormat, depthFormat, stencilFormat; - struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); - - if (!intelfb) - return GL_FALSE; - - if (visual->redBits == 5) - colorFormat = PIPE_FORMAT_R5G6B5_UNORM; - else - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (visual->depthBits == 16) - depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (visual->depthBits == 24) - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - else - depthFormat = PIPE_FORMAT_NONE; - - if (visual->stencilBits == 8) - stencilFormat = PIPE_FORMAT_S8Z24_UNORM; - else - stencilFormat = PIPE_FORMAT_NONE; - - intelfb->stfb = st_create_framebuffer(visual, - colorFormat, - depthFormat, - stencilFormat, - driDrawPriv->w, - driDrawPriv->h, - (void*) intelfb); - if (!intelfb->stfb) { - free(intelfb); - return GL_FALSE; - } - - driDrawPriv->driverPrivate = (void *) intelfb; - return GL_TRUE; - } -} - -static void -intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) -{ - struct intel_framebuffer *intelfb = intel_framebuffer(driDrawPriv); - assert(intelfb->stfb); - st_unreference_framebuffer(&intelfb->stfb); - free(intelfb); -} - - -/** - * Get information about previous buffer swaps. - */ -static int -intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) -{ - if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) - || (sInfo == NULL)) { - return -1; - } - - return 0; -} - - -static void -intelSetTexOffset(__DRIcontext *pDRICtx, int texname, - unsigned long long offset, int depth, uint pitch) -{ - abort(); -#if 0 - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct st_texture_object *stObj = st_texture_object(tObj); - - if (!stObj) - return; - - if (stObj->pt) - st->pipe->texture_release(intel->st->pipe, &stObj->pt); - - stObj->imageOverride = GL_TRUE; - stObj->depthOverride = depth; - stObj->pitchOverride = pitch; - - if (offset) - stObj->textureOffset = offset; -#endif -} - - -static const struct __DriverAPIRec intelAPI = { - .InitDriver = intelInitDriver, - .DestroyScreen = intelDestroyScreen, - .CreateContext = intelCreateContext, - .DestroyContext = intelDestroyContext, - .CreateBuffer = intelCreateBuffer, - .DestroyBuffer = intelDestroyBuffer, - .SwapBuffers = intelSwapBuffers, - .MakeCurrent = intelMakeCurrent, - .UnbindContext = intelUnbindContext, - .GetSwapInfo = intelGetSwapInfo, - .GetMSC = driGetMSC32, - .WaitForMSC = driWaitForMSC32, - .WaitForSBC = NULL, - .SwapBuffersMSC = NULL, - .CopySubBuffer = intelCopySubBuffer, - .setTexOffset = intelSetTexOffset, -}; - - -static __GLcontextModes * -intelFillInModes(unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, boolean have_back_buffer) -{ - __GLcontextModes *modes; - __GLcontextModes *m; - unsigned num_modes; - unsigned depth_buffer_factor; - unsigned back_buffer_factor; - GLenum fb_format; - GLenum fb_type; - - /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't - * support pageflipping at all. - */ - static const GLenum back_buffer_modes[] = { - GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML - }; - - u_int8_t depth_bits_array[3]; - u_int8_t stencil_bits_array[3]; - - - depth_bits_array[0] = 0; - depth_bits_array[1] = depth_bits; - depth_bits_array[2] = depth_bits; - - /* Just like with the accumulation buffer, always provide some modes - * with a stencil buffer. It will be a sw fallback, but some apps won't - * care about that. - */ - stencil_bits_array[0] = 0; - stencil_bits_array[1] = 0; - if (depth_bits == 24) - stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; - - stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; - - depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; - back_buffer_factor = (have_back_buffer) ? 3 : 1; - - num_modes = depth_buffer_factor * back_buffer_factor * 4; - - if (pixel_bits == 16) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; - } - - modes = - (*dri_interface->createContextModes) (num_modes, - sizeof(__GLcontextModes)); - m = modes; - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_TRUE_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_DIRECT_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - - /* Mark the visual as slow if there are "fake" stencil bits. - */ - for (m = modes; m != NULL; m = m->next) { - if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { - m->visualRating = GLX_SLOW_CONFIG; - } - } - - return modes; -} - - -/** - * This is the bootstrap function for the driver. libGL supplies all of the - * requisite information about the system, and the driver initializes itself. - * This routine also fills in the linked list pointed to by \c driver_modes - * with the \c __GLcontextModes that the driver can support for windows or - * pbuffers. - * - * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on - * failure. - */ -PUBLIC void * -__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, - __DRIscreen * psc, - const __GLcontextModes * modes, - const __DRIversion * ddx_version, - const __DRIversion * dri_version, - const __DRIversion * drm_version, - const __DRIframebuffer * frame_buffer, - drmAddress pSAREA, int fd, - int internal_api_version, - const __DRIinterfaceMethods * interface, - __GLcontextModes ** driver_modes) -{ - __DRIscreenPrivate *psp; - static const __DRIversion ddx_expected = { 1, 7, 0 }; - static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 7, 0 }; - - dri_interface = interface; - - if (!driCheckDriDdxDrmVersions2("i915", - dri_version, &dri_expected, - ddx_version, &ddx_expected, - drm_version, &drm_expected)) { - return NULL; - } - - psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, - ddx_version, dri_version, drm_version, - frame_buffer, pSAREA, fd, - internal_api_version, &intelAPI); - - if (psp != NULL) { - I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; - *driver_modes = intelFillInModes(dri_priv->cpp * 8, - (dri_priv->cpp == 2) ? 16 : 24, - (dri_priv->cpp == 2) ? 0 : 8, 1); - - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create - * is called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions(NULL, card_extensions, GL_FALSE); - } - - return (void *) psp; -} - diff --git a/src/mesa/drivers/dri/intel_winsys/intel_screen.h b/src/mesa/drivers/dri/intel_winsys/intel_screen.h deleted file mode 100644 index 3396f9e564..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_screen.h +++ /dev/null @@ -1,113 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef _INTEL_SCREEN_H_ -#define _INTEL_SCREEN_H_ - -#include "dri_util.h" -#include "i830_common.h" -#include "xmlconfig.h" -#include "dri_bufpool.h" - -#include "pipe/p_compiler.h" - - -struct intel_screen -{ - struct { - drm_handle_t handle; - - /* We create a static dri buffer for the frontbuffer. - */ - struct _DriBufferObject *buffer; - - char *map; /* memory map */ - int offset; /* from start of video mem, in bytes */ - int pitch; /* row stride, in bytes */ - int width; - int height; - int size; - int cpp; /* for front and back buffers */ - } front; - - int deviceID; - int drmMinor; - - drmI830Sarea *sarea; - - /** - * Configuration cache with default values for all contexts - */ - driOptionCache optionCache; - - struct _DriBufferPool *batchPool; - struct _DriBufferPool *staticPool; /** for the X screen/framebuffer */ - boolean havePools; - - /** - * Temporary(?) context to use for SwapBuffers or other situations in - * which we need a rendering context, but none is currently bound. - */ - struct intel_context *dummyContext; - - struct pipe_winsys *winsys; -}; - - - -/** cast wrapper */ -static INLINE struct intel_screen * -intel_screen(__DRIscreenPrivate *sPriv) -{ - return (struct intel_screen *) sPriv->private; -} - - -extern void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); - - -extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); - -extern boolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); - -extern boolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); - - -extern boolean -intelCreatePools(__DRIscreenPrivate *sPriv); - -extern boolean -intelCreateContext(const __GLcontextModes * visual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - - -#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c deleted file mode 100644 index 56b86d6a63..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.c +++ /dev/null @@ -1,253 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_swapbuffers.h" -#include "intel_batchbuffer.h" -#include "intel_reg.h" -#include "intel_winsys.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_cb_fbo.h" - - -/** - * Display a colorbuffer surface in an X window. - * Used for SwapBuffers and flushing front buffer rendering. - * - * \param dPriv the window/drawable to display into - * \param surf the surface to display - * \param rect optional subrect of surface to display (may be NULL). - */ -void -intelDisplaySurface(__DRIdrawablePrivate *dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t *rect) -{ - struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - struct intel_context *intel = intelScreen->dummyContext; - - DBG(SWAP, "%s\n", __FUNCTION__); - - if (!intel) { - /* XXX this is where some kind of extra/meta context could be useful */ - return; - } - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); - driFenceUnReference(intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - * current context (which is what intelScreenContext should return) might - * not get a contended lock and thus cliprects not updated (tests/manywin) - */ - if (intel_context(dPriv->driContextPriv) != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - - if (dPriv && dPriv->numClipRects) { - const int srcWidth = surf->width; - const int srcHeight = surf->height; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; - const int cpp = intelScreen->front.cpp; - const int srcpitch = surf->pitch; - int BR13, CMD; - int i; - - ASSERT(surf->buffer); - ASSERT(surf->cpp == cpp); - - DBG(SWAP, "screen pitch %d src surface pitch %d\n", - pitch, surf->pitch); - - if (cpp == 2) { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - } - else { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - } - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) { - /* invalid cliprect, skip it */ - continue; - } - - box = *pbox; - - if (rect) { - /* intersect cliprect with user-provided src rect */ - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > srcWidth) - box.x2 = srcWidth + box.x1; - if (box.y2 - box.y1 > srcHeight) - box.y2 = srcHeight + box.y1; - - DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - assert(box.x1 < box.x2); - assert(box.y1 < box.y2); - - /* XXX this could be done with pipe->surface_copy() */ - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((box.y1 << 16) | box.x1); - OUT_BATCH((box.y2 << 16) | box.x2); - - OUT_RELOC(intelScreen->front.buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((sbox.y1 << 16) | sbox.x1); - OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(dri_bo(surf->buffer), - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - ADVANCE_BATCH(); - } - - if (intel->first_swap_fence) - driFenceUnReference(intel->first_swap_fence); - intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); - driFenceReference(intel->first_swap_fence); - } - - UNLOCK_HARDWARE(intel); - - if (intel->lastStamp != dPriv->lastStamp) { - intelUpdateWindowSize(dPriv); - intel->lastStamp = dPriv->lastStamp; - } -} - - - -/** - * This will be called whenever the currently bound window is moved/resized. - */ -void -intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) -{ - struct intel_framebuffer *intelfb = intel_framebuffer(dPriv); - assert(intelfb->stfb); - st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); -} - - - -void -intelSwapBuffers(__DRIdrawablePrivate * dPriv) -{ - struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; - - assert(intel_fb); - assert(intel_fb->stfb); - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, NULL); - st_notify_swapbuffers_complete(intel_fb->stfb); - } -} - - -/** - * Called via glXCopySubBufferMESA() to copy a subrect of the back - * buffer to the front buffer/screen. - */ -void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) -{ - struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; - - assert(intel_fb); - assert(intel_fb->stfb); - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - drm_clip_rect_t rect; - rect.x1 = x; - rect.y1 = y; - rect.x2 = w; - rect.y2 = h; - - st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, &rect); - } -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h b/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h deleted file mode 100644 index 7ae5fd15a5..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_swapbuffers.h +++ /dev/null @@ -1,47 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_SWAPBUFFERS_H -#define INTEL_SWAPBUFFERS_H - - -struct pipe_surface; - - -extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t * rect); - -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, - int x, int y, int w, int h); - -extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); - - -#endif /* INTEL_SWAPBUFFERS_H */ diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h b/src/mesa/drivers/dri/intel_winsys/intel_winsys.h deleted file mode 100644 index ffc40782be..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys.h +++ /dev/null @@ -1,73 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_WINSYS_H -#define INTEL_WINSYS_H - -#include "pipe/p_state.h" - -struct intel_context; -struct pipe_context; -struct pipe_winsys; -struct pipe_buffer; -struct _DriBufferObject; - -struct pipe_winsys * -intel_create_pipe_winsys( int fd ); - -void -intel_destroy_pipe_winsys( struct pipe_winsys *winsys ); - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel, - struct pipe_winsys *winsys ); - -struct pipe_context * -intel_create_i915simple( struct intel_context *intel, - struct pipe_winsys *winsys ); - - -struct intel_buffer { - struct pipe_buffer base; - struct _DriBufferObject *driBO; -}; - -static INLINE struct intel_buffer * -intel_buffer( struct pipe_buffer *buf ) -{ - return (struct intel_buffer *)buf; -} - -static INLINE struct _DriBufferObject * -dri_bo( struct pipe_buffer *buf ) -{ - return intel_buffer(buf)->driBO; -} - - - -#endif diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c deleted file mode 100644 index 1ba6a9e1b2..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_i915.c +++ /dev/null @@ -1,154 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include -#include "dri_bufpool.h" -#include "dri_bufmgr.h" - -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_winsys.h" - -#include "pipe/p_util.h" -#include "pipe/i915simple/i915_winsys.h" - - -struct intel_i915_winsys { - struct i915_winsys winsys; /**< batch buffer funcs */ - struct intel_context *intel; -}; - - -/* Turn a i915simple winsys into an intel/i915simple winsys: - */ -static inline struct intel_i915_winsys * -intel_i915_winsys( struct i915_winsys *sws ) -{ - return (struct intel_i915_winsys *)sws; -} - - -/* Simple batchbuffer interface: - */ - -static unsigned *intel_i915_batch_start( struct i915_winsys *sws, - unsigned dwords, - unsigned relocs ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - - /* XXX: check relocs. - */ - if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { - /* XXX: Hmm, the driver can't really do much with this pointer: - */ - return (unsigned *)intel->batch->ptr; - } - else - return NULL; -} - -static void intel_i915_batch_dword( struct i915_winsys *sws, - unsigned dword ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - intel_batchbuffer_emit_dword( intel->batch, dword ); -} - -static void intel_i915_batch_reloc( struct i915_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - unsigned flags = DRM_BO_FLAG_MEM_TT; - unsigned mask = DRM_BO_MASK_MEM; - - if (access_flags & I915_BUFFER_ACCESS_WRITE) { - flags |= DRM_BO_FLAG_WRITE; - mask |= DRM_BO_FLAG_WRITE; - } - - if (access_flags & I915_BUFFER_ACCESS_READ) { - flags |= DRM_BO_FLAG_READ; - mask |= DRM_BO_FLAG_READ; - } - - intel_batchbuffer_emit_reloc( intel->batch, - dri_bo( buf ), - flags, mask, - delta ); -} - - - -static void intel_i915_batch_flush( struct i915_winsys *sws ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - - intel_batchbuffer_flush( intel->batch ); -// if (0) intel_i915_batch_wait_idle( sws ); -} - - -static void intel_i915_batch_finish( struct i915_winsys *sws ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - intel_batchbuffer_finish( intel->batch ); -} - - -/** - * Create i915 hardware rendering context. - */ -struct pipe_context * -intel_create_i915simple( struct intel_context *intel, - struct pipe_winsys *winsys ) -{ - struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); - - /* Fill in this struct with callbacks that i915simple will need to - * communicate with the window system, buffer manager, etc. - */ - iws->winsys.batch_start = intel_i915_batch_start; - iws->winsys.batch_dword = intel_i915_batch_dword; - iws->winsys.batch_reloc = intel_i915_batch_reloc; - iws->winsys.batch_flush = intel_i915_batch_flush; - iws->winsys.batch_finish = intel_i915_batch_finish; - iws->intel = intel; - - /* Create the i915simple context: - */ - return i915_create( winsys, - &iws->winsys, - intel->intelScreen->deviceID ); -} diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c deleted file mode 100644 index 789a386500..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_pipe.c +++ /dev/null @@ -1,302 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include -#include "dri_bufpool.h" -#include "dri_bufmgr.h" - -#include "intel_context.h" -#include "intel_winsys.h" -#include "intel_swapbuffers.h" -#include "intel_batchbuffer.h" - -#include "pipe/p_winsys.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" - - - -struct intel_pipe_winsys { - struct pipe_winsys winsys; - struct _DriBufferPool *regionPool; -}; - - - -/* Turn a pipe winsys into an intel/pipe winsys: - */ -static inline struct intel_pipe_winsys * -intel_pipe_winsys( struct pipe_winsys *winsys ) -{ - return (struct intel_pipe_winsys *)winsys; -} - - -/* Most callbacks map direcly onto dri_bufmgr operations: - */ -static void *intel_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned flags ) -{ - unsigned drm_flags = 0; - - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - drm_flags |= DRM_BO_FLAG_WRITE; - - if (flags & PIPE_BUFFER_USAGE_CPU_READ) - drm_flags |= DRM_BO_FLAG_READ; - - return driBOMap( dri_bo(buf), drm_flags, 0 ); -} - -static void intel_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnmap( dri_bo(buf) ); -} - - -static void -intel_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnReference( dri_bo(buf) ); -} - - -/* Pipe has no concept of pools. We choose the tex/region pool - * for all buffers. - * Grabs the hardware lock! - */ -static struct pipe_buffer * -intel_buffer_create(struct pipe_winsys *winsys, - unsigned alignment, - unsigned usage, - unsigned size ) -{ - struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); - struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); - unsigned flags = 0; - - buffer->base.refcount = 1; - buffer->base.alignment = alignment; - buffer->base.usage = usage; - buffer->base.size = size; - - if (usage & (PIPE_BUFFER_USAGE_VERTEX /*| IWS_BUFFER_USAGE_LOCAL*/)) { - flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; - } else { - flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; - } - - if (usage & PIPE_BUFFER_USAGE_GPU_READ) - flags |= DRM_BO_FLAG_READ; - - if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) - flags |= DRM_BO_FLAG_WRITE; - - /* drm complains if we don't set any read/write flags. - */ - if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) - flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; - -#if 0 - if (flags & IWS_BUFFER_USAGE_EXE) - flags |= DRM_BO_FLAG_EXE; - - if (usage & IWS_BUFFER_USAGE_CACHED) - flags |= DRM_BO_FLAG_CACHED; -#endif - - driGenBuffers( iws->regionPool, - "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); - - driBOData( buffer->driBO, size, NULL, 0 ); - - return &buffer->base; -} - - -static struct pipe_buffer * -intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) -{ - struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); - struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); - - driGenUserBuffer( iws->regionPool, - "pipe user buffer", &buffer->driBO, ptr, bytes); - - return &buffer->base; -} - - -/* The state tracker (should!) keep track of whether the fake - * frontbuffer has been touched by any rendering since the last time - * we copied its contents to the real frontbuffer. Our task is easy: - */ -static void -intel_flush_frontbuffer( struct pipe_winsys *winsys, - struct pipe_surface *surf, - void *context_private) -{ - struct intel_context *intel = (struct intel_context *) context_private; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelDisplaySurface(dPriv, surf, NULL); -} - - -static struct pipe_surface * -intel_i915_surface_alloc(struct pipe_winsys *winsys) -{ - struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); - if (surf) { - surf->refcount = 1; - surf->winsys = winsys; - } - return surf; -} - - -/** - * Round n up to next multiple. - */ -static INLINE unsigned -round_up(unsigned n, unsigned multiple) -{ - return (n + multiple - 1) & ~(multiple - 1); -} - -/** - * Copied from xm_winsys.c - */ -static int -intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, - struct pipe_surface *surf, - unsigned width, unsigned height, - enum pipe_format format, - unsigned flags) -{ - const unsigned alignment = 64; - int ret; - - surf->width = width; - surf->height = height; - surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = round_up(width, alignment / surf->cpp); - - assert(!surf->buffer); - surf->buffer = winsys->buffer_create(winsys, alignment, - PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * height); - if(!surf->buffer) - return -1; - - return 0; -} - - -static void -intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) -{ - struct pipe_surface *surf = *s; - surf->refcount--; - if (surf->refcount == 0) { - if (surf->buffer) - pipe_buffer_reference(winsys, &surf->buffer, NULL); - free(surf); - } - *s = NULL; -} - - - -static void -intel_printf( struct pipe_winsys *winsys, const char *fmtString, ... ) -{ - va_list args; - va_start( args, fmtString ); - vfprintf(stderr, fmtString, args); - va_end( args ); -} - -static const char * -intel_get_name( struct pipe_winsys *winsys ) -{ - return "Intel/DRI/ttm"; -} - - -struct pipe_winsys * -intel_create_pipe_winsys( int fd ) -{ - struct intel_pipe_winsys *iws = CALLOC_STRUCT( intel_pipe_winsys ); - - /* Fill in this struct with callbacks that pipe will need to - * communicate with the window system, buffer manager, etc. - * - * Pipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate _DriBufferObject. - */ - iws->winsys.buffer_create = intel_buffer_create; - iws->winsys.user_buffer_create = intel_user_buffer_create; - iws->winsys.buffer_map = intel_buffer_map; - iws->winsys.buffer_unmap = intel_buffer_unmap; - iws->winsys.buffer_destroy = intel_buffer_destroy; - iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; - iws->winsys.printf = intel_printf; - iws->winsys.get_name = intel_get_name; - iws->winsys.surface_alloc = intel_i915_surface_alloc; - iws->winsys.surface_alloc_storage = intel_i915_surface_alloc_storage; - iws->winsys.surface_release = intel_i915_surface_release; - - if (fd) - iws->regionPool = driDRMPoolInit(fd); - - return &iws->winsys; -} - - -void -intel_destroy_pipe_winsys( struct pipe_winsys *winsys ) -{ - struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); - if (iws->regionPool) { - driPoolTakeDown(iws->regionPool); - } - free(iws); -} - diff --git a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c b/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c deleted file mode 100644 index cec3437831..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/intel_winsys_softpipe.c +++ /dev/null @@ -1,81 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include "intel_context.h" -#include "intel_winsys.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_format.h" -#include "pipe/softpipe/sp_winsys.h" - - -struct intel_softpipe_winsys { - struct softpipe_winsys sws; - struct intel_context *intel; -}; - -/** - * Return list of surface formats supported by this driver. - */ -static boolean -intel_is_format_supported(struct softpipe_winsys *sws, - enum pipe_format format) -{ - switch(format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - return TRUE; - default: - return FALSE; - } -} - - -/** - * Create rendering context which uses software rendering. - */ -struct pipe_context * -intel_create_softpipe( struct intel_context *intel, - struct pipe_winsys *winsys ) -{ - struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); - - /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. - */ - isws->sws.is_format_supported = intel_is_format_supported; - isws->intel = intel; - - /* Create the softpipe context: - */ - return softpipe_create( winsys, &isws->sws ); -} diff --git a/src/mesa/drivers/dri/intel_winsys/server/i830_common.h b/src/mesa/drivers/dri/intel_winsys/server/i830_common.h deleted file mode 100644 index d4d58886ce..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/server/i830_common.h +++ /dev/null @@ -1,226 +0,0 @@ -/************************************************************************** - -Copyright 2001 VA Linux Systems Inc., Fremont, California. -Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas. - -All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining a -copy of this software and associated documentation files (the "Software"), -to deal in the Software without restriction, including without limitation -on the rights to use, copy, modify, merge, publish, distribute, sub -license, and/or sell copies of the Software, and to permit persons to whom -the Software is furnished to do so, subject to the following conditions: - -The above copyright notice and this permission notice (including the next -paragraph) shall be included in all copies or substantial portions of the -Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL -ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, -DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR -OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE -USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ - -#ifndef _I830_COMMON_H_ -#define _I830_COMMON_H_ - - -#define I830_NR_TEX_REGIONS 255 /* maximum due to use of chars for next/prev */ -#define I830_LOG_MIN_TEX_REGION_SIZE 14 - - -/* Driver specific DRM command indices - * NOTE: these are not OS specific, but they are driver specific - */ -#define DRM_I830_INIT 0x00 -#define DRM_I830_FLUSH 0x01 -#define DRM_I830_FLIP 0x02 -#define DRM_I830_BATCHBUFFER 0x03 -#define DRM_I830_IRQ_EMIT 0x04 -#define DRM_I830_IRQ_WAIT 0x05 -#define DRM_I830_GETPARAM 0x06 -#define DRM_I830_SETPARAM 0x07 -#define DRM_I830_ALLOC 0x08 -#define DRM_I830_FREE 0x09 -#define DRM_I830_INIT_HEAP 0x0a -#define DRM_I830_CMDBUFFER 0x0b -#define DRM_I830_DESTROY_HEAP 0x0c -#define DRM_I830_SET_VBLANK_PIPE 0x0d -#define DRM_I830_GET_VBLANK_PIPE 0x0e - -typedef struct { - enum { - I830_INIT_DMA = 0x01, - I830_CLEANUP_DMA = 0x02, - I830_RESUME_DMA = 0x03 - } func; - unsigned int mmio_offset; - int sarea_priv_offset; - unsigned int ring_start; - unsigned int ring_end; - unsigned int ring_size; - unsigned int front_offset; - unsigned int back_offset; - unsigned int depth_offset; - unsigned int w; - unsigned int h; - unsigned int pitch; - unsigned int pitch_bits; - unsigned int back_pitch; - unsigned int depth_pitch; - unsigned int cpp; - unsigned int chipset; -} drmI830Init; - -typedef struct { - drmTextureRegion texList[I830_NR_TEX_REGIONS+1]; - int last_upload; /* last time texture was uploaded */ - int last_enqueue; /* last time a buffer was enqueued */ - int last_dispatch; /* age of the most recently dispatched buffer */ - int ctxOwner; /* last context to upload state */ - int texAge; - int pf_enabled; /* is pageflipping allowed? */ - int pf_active; - int pf_current_page; /* which buffer is being displayed? */ - int perf_boxes; /* performance boxes to be displayed */ - int width, height; /* screen size in pixels */ - - drm_handle_t front_handle; - int front_offset; - int front_size; - - drm_handle_t back_handle; - int back_offset; - int back_size; - - drm_handle_t depth_handle; - int depth_offset; - int depth_size; - - drm_handle_t tex_handle; - int tex_offset; - int tex_size; - int log_tex_granularity; - int pitch; - int rotation; /* 0, 90, 180 or 270 */ - int rotated_offset; - int rotated_size; - int rotated_pitch; - int virtualX, virtualY; - - unsigned int front_tiled; - unsigned int back_tiled; - unsigned int depth_tiled; - unsigned int rotated_tiled; - unsigned int rotated2_tiled; - - int pipeA_x; - int pipeA_y; - int pipeA_w; - int pipeA_h; - int pipeB_x; - int pipeB_y; - int pipeB_w; - int pipeB_h; - - /* Triple buffering */ - drm_handle_t third_handle; - int third_offset; - int third_size; - unsigned int third_tiled; -} drmI830Sarea; - -/* Flags for perf_boxes - */ -#define I830_BOX_RING_EMPTY 0x1 /* populated by kernel */ -#define I830_BOX_FLIP 0x2 /* populated by kernel */ -#define I830_BOX_WAIT 0x4 /* populated by kernel & client */ -#define I830_BOX_TEXTURE_LOAD 0x8 /* populated by kernel */ -#define I830_BOX_LOST_CONTEXT 0x10 /* populated by client */ - - -typedef struct { - int start; /* agp offset */ - int used; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830BatchBuffer; - -typedef struct { - char *buf; /* agp offset */ - int sz; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830CmdBuffer; - -typedef struct { - int *irq_seq; -} drmI830IrqEmit; - -typedef struct { - int irq_seq; -} drmI830IrqWait; - -typedef struct { - int param; - int *value; -} drmI830GetParam; - -#define I830_PARAM_IRQ_ACTIVE 1 -#define I830_PARAM_ALLOW_BATCHBUFFER 2 - -typedef struct { - int param; - int value; -} drmI830SetParam; - -#define I830_SETPARAM_USE_MI_BATCHBUFFER_START 1 -#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY 2 -#define I830_SETPARAM_ALLOW_BATCHBUFFER 3 - - -/* A memory manager for regions of shared memory: - */ -#define I830_MEM_REGION_AGP 1 - -typedef struct { - int region; - int alignment; - int size; - int *region_offset; /* offset from start of fb or agp */ -} drmI830MemAlloc; - -typedef struct { - int region; - int region_offset; -} drmI830MemFree; - -typedef struct { - int region; - int size; - int start; -} drmI830MemInitHeap; - -typedef struct { - int region; -} drmI830MemDestroyHeap; - -#define DRM_I830_VBLANK_PIPE_A 1 -#define DRM_I830_VBLANK_PIPE_B 2 - -typedef struct { - int pipe; -} drmI830VBlankPipe; - -#endif /* _I830_DRM_H_ */ diff --git a/src/mesa/drivers/dri/intel_winsys/server/i830_dri.h b/src/mesa/drivers/dri/intel_winsys/server/i830_dri.h deleted file mode 100644 index c2a3af8cbf..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/server/i830_dri.h +++ /dev/null @@ -1,63 +0,0 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ - -#ifndef _I830_DRI_H -#define _I830_DRI_H - -#include "xf86drm.h" -#include "i830_common.h" - -#define I830_MAX_DRAWABLES 256 - -#define I830_MAJOR_VERSION 1 -#define I830_MINOR_VERSION 7 -#define I830_PATCHLEVEL 2 - -#define I830_REG_SIZE 0x80000 - -typedef struct _I830DRIRec { - drm_handle_t regs; - drmSize regsSize; - - drmSize unused1; /* backbufferSize */ - drm_handle_t unused2; /* backbuffer */ - - drmSize unused3; /* depthbufferSize */ - drm_handle_t unused4; /* depthbuffer */ - - drmSize unused5; /* rotatedSize */ - drm_handle_t unused6; /* rotatedbuffer */ - - drm_handle_t unused7; /* textures */ - int unused8; /* textureSize */ - - drm_handle_t unused9; /* agp_buffers */ - drmSize unused10; /* agp_buf_size */ - - int deviceID; - int width; - int height; - int mem; - int cpp; - int bitsPerPixel; - - int unused11[8]; /* was front/back/depth/rotated offset/pitch */ - - int unused12; /* logTextureGranularity */ - int unused13; /* textureOffset */ - - int irq; - int sarea_priv_offset; -} I830DRIRec, *I830DRIPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830ConfigPrivRec, *I830ConfigPrivPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830DRIContextRec, *I830DRIContextPtr; - - -#endif diff --git a/src/mesa/drivers/dri/intel_winsys/server/intel.h b/src/mesa/drivers/dri/intel_winsys/server/intel.h deleted file mode 100644 index 6ea72499c1..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/server/intel.h +++ /dev/null @@ -1,331 +0,0 @@ -#ifndef _INTEL_H_ -#define _INTEL_H_ - -#include "xf86drm.h" /* drm_handle_t, etc */ - -/* Intel */ -#ifndef PCI_CHIP_I810 -#define PCI_CHIP_I810 0x7121 -#define PCI_CHIP_I810_DC100 0x7123 -#define PCI_CHIP_I810_E 0x7125 -#define PCI_CHIP_I815 0x1132 -#define PCI_CHIP_I810_BRIDGE 0x7120 -#define PCI_CHIP_I810_DC100_BRIDGE 0x7122 -#define PCI_CHIP_I810_E_BRIDGE 0x7124 -#define PCI_CHIP_I815_BRIDGE 0x1130 -#endif - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 - -#ifndef PCI_CHIP_I855_GM -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I855_GM_BRIDGE 0x3580 -#endif - -#ifndef PCI_CHIP_I865_G -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I865_G_BRIDGE 0x2570 -#endif - -#ifndef PCI_CHIP_I915_G -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I915_GM -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I915_GM_BRIDGE 0x2590 -#endif - -#ifndef PCI_CHIP_E7221_G -#define PCI_CHIP_E7221_G 0x258A -/* Same as I915_G_BRIDGE */ -#define PCI_CHIP_E7221_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I945_G -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_G_BRIDGE 0x2770 -#endif - -#ifndef PCI_CHIP_I945_GM -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GM_BRIDGE 0x27A0 -#endif - -#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 || \ - pI810->Chipset == PCI_CHIP_I810_DC100 || \ - pI810->Chipset == PCI_CHIP_I810_E) -#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815) -#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M) -#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G) -#define IS_I85X(pI810) (pI810->Chipset == PCI_CHIP_I855_GM) -#define IS_I852(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME)) -#define IS_I855(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME)) -#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G) - -#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G) -#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM) -#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G) -#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM) -#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810)) - -#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810)) - -#define I830_GMCH_CTRL 0x52 - -#define I830_GMCH_MEM_MASK 0x1 -#define I830_GMCH_MEM_64M 0x1 -#define I830_GMCH_MEM_128M 0 - -#define I830_GMCH_GMS_MASK 0x70 -#define I830_GMCH_GMS_DISABLED 0x00 -#define I830_GMCH_GMS_LOCAL 0x10 -#define I830_GMCH_GMS_STOLEN_512 0x20 -#define I830_GMCH_GMS_STOLEN_1024 0x30 -#define I830_GMCH_GMS_STOLEN_8192 0x40 - -#define I855_GMCH_GMS_MASK (0x7 << 4) -#define I855_GMCH_GMS_DISABLED 0x00 -#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) -#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) -#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) -#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) -#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) -#define I915G_GMCH_GMS_STOLEN_48M (0x6 << 4) -#define I915G_GMCH_GMS_STOLEN_64M (0x7 << 4) - -typedef unsigned char Bool; -#define TRUE 1 -#define FALSE 0 - -#define PIPE_NONE 0<<0 -#define PIPE_CRT 1<<0 -#define PIPE_TV 1<<1 -#define PIPE_DFP 1<<2 -#define PIPE_LFP 1<<3 -#define PIPE_CRT2 1<<4 -#define PIPE_TV2 1<<5 -#define PIPE_DFP2 1<<6 -#define PIPE_LFP2 1<<7 - -typedef struct _I830MemPool *I830MemPoolPtr; -typedef struct _I830MemRange *I830MemRangePtr; -typedef struct _I830MemRange { - long Start; - long End; - long Size; - unsigned long Physical; - unsigned long Offset; /* Offset of AGP-allocated portion */ - unsigned long Alignment; - drm_handle_t Key; - unsigned long Pitch; // add pitch - I830MemPoolPtr Pool; -} I830MemRange; - -typedef struct _I830MemPool { - I830MemRange Total; - I830MemRange Free; - I830MemRange Fixed; - I830MemRange Allocated; -} I830MemPool; - -typedef struct { - int tail_mask; - I830MemRange mem; - unsigned char *virtual_start; - int head; - int tail; - int space; -} I830RingBuffer; - -typedef struct _I830Rec { - unsigned char *MMIOBase; - unsigned char *FbBase; - int cpp; - uint32_t aper_size; - unsigned int bios_version; - - /* These are set in PreInit and never changed. */ - long FbMapSize; - long TotalVideoRam; - I830MemRange StolenMemory; /* pre-allocated memory */ - long BIOSMemorySize; /* min stolen pool size */ - int BIOSMemSizeLoc; - - /* These change according to what has been allocated. */ - long FreeMemory; - I830MemRange MemoryAperture; - I830MemPool StolenPool; - long allocatedMemory; - - /* Regions allocated either from the above pools, or from agpgart. */ - /* for single and dual head configurations */ - I830MemRange FrontBuffer; - I830MemRange FrontBuffer2; - I830MemRange Scratch; - I830MemRange Scratch2; - - I830RingBuffer *LpRing; - - I830MemRange BackBuffer; - I830MemRange DepthBuffer; - I830MemRange TexMem; - int TexGranularity; - I830MemRange ContextMem; - int drmMinor; - Bool have3DWindows; - - Bool NeedRingBufferLow; - Bool allowPageFlip; - Bool disableTiling; - - int Chipset; - unsigned long LinearAddr; - unsigned long MMIOAddr; - - drmSize registerSize; /**< \brief MMIO register map size */ - drm_handle_t registerHandle; /**< \brief MMIO register map handle */ - // IOADDRESS ioBase; - int irq; /**< \brief IRQ number */ - int GttBound; - - drm_handle_t ring_map; - unsigned int Fence[8]; - -} I830Rec; - -/* - * 12288 is set as the maximum, chosen because it is enough for - * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare. - */ -#define I830_MAXIMUM_VBIOS_MEM 12288 -#define I830_DEFAULT_VIDEOMEM_2D (MB(32) / 1024) -#define I830_DEFAULT_VIDEOMEM_3D (MB(64) / 1024) - -/* Flags for memory allocation function */ -#define FROM_ANYWHERE 0x00000000 -#define FROM_POOL_ONLY 0x00000001 -#define FROM_NEW_ONLY 0x00000002 -#define FROM_MASK 0x0000000f - -#define ALLOCATE_AT_TOP 0x00000010 -#define ALLOCATE_AT_BOTTOM 0x00000020 -#define FORCE_GAPS 0x00000040 - -#define NEED_PHYSICAL_ADDR 0x00000100 -#define ALIGN_BOTH_ENDS 0x00000200 -#define FORCE_LOW 0x00000400 - -#define ALLOC_NO_TILING 0x00001000 -#define ALLOC_INITIAL 0x00002000 - -#define ALLOCATE_DRY_RUN 0x80000000 - -/* Chipset registers for VIDEO BIOS memory RW access */ -#define _855_DRAM_RW_CONTROL 0x58 -#define _845_DRAM_RW_CONTROL 0x90 -#define DRAM_WRITE 0x33330000 - -#define KB(x) ((x) * 1024) -#define MB(x) ((x) * KB(1024)) - -#define GTT_PAGE_SIZE KB(4) -#define ROUND_TO(x, y) (((x) + (y) - 1) / (y) * (y)) -#define ROUND_DOWN_TO(x, y) ((x) / (y) * (y)) -#define ROUND_TO_PAGE(x) ROUND_TO((x), GTT_PAGE_SIZE) -#define ROUND_TO_MB(x) ROUND_TO((x), MB(1)) -#define PRIMARY_RINGBUFFER_SIZE KB(128) - - -/* Ring buffer registers, p277, overview p19 - */ -#define LP_RING 0x2030 -#define HP_RING 0x2040 - -#define RING_TAIL 0x00 -#define TAIL_ADDR 0x000FFFF8 -#define I830_TAIL_MASK 0x001FFFF8 - -#define RING_HEAD 0x04 -#define HEAD_WRAP_COUNT 0xFFE00000 -#define HEAD_WRAP_ONE 0x00200000 -#define HEAD_ADDR 0x001FFFFC -#define I830_HEAD_MASK 0x001FFFFC - -#define RING_START 0x08 -#define START_ADDR 0x03FFFFF8 -#define I830_RING_START_MASK 0xFFFFF000 - -#define RING_LEN 0x0C -#define RING_NR_PAGES 0x001FF000 -#define I830_RING_NR_PAGES 0x001FF000 -#define RING_REPORT_MASK 0x00000006 -#define RING_REPORT_64K 0x00000002 -#define RING_REPORT_128K 0x00000004 -#define RING_NO_REPORT 0x00000000 -#define RING_VALID_MASK 0x00000001 -#define RING_VALID 0x00000001 -#define RING_INVALID 0x00000000 - - -/* Fence/Tiling ranges [0..7] - */ -#define FENCE 0x2000 -#define FENCE_NR 8 - -#define I915G_FENCE_START_MASK 0x0ff00000 - -#define I830_FENCE_START_MASK 0x07f80000 - -#define FENCE_START_MASK 0x03F80000 -#define FENCE_X_MAJOR 0x00000000 -#define FENCE_Y_MAJOR 0x00001000 -#define FENCE_SIZE_MASK 0x00000700 -#define FENCE_SIZE_512K 0x00000000 -#define FENCE_SIZE_1M 0x00000100 -#define FENCE_SIZE_2M 0x00000200 -#define FENCE_SIZE_4M 0x00000300 -#define FENCE_SIZE_8M 0x00000400 -#define FENCE_SIZE_16M 0x00000500 -#define FENCE_SIZE_32M 0x00000600 -#define FENCE_SIZE_64M 0x00000700 -#define I915G_FENCE_SIZE_1M 0x00000000 -#define I915G_FENCE_SIZE_2M 0x00000100 -#define I915G_FENCE_SIZE_4M 0x00000200 -#define I915G_FENCE_SIZE_8M 0x00000300 -#define I915G_FENCE_SIZE_16M 0x00000400 -#define I915G_FENCE_SIZE_32M 0x00000500 -#define I915G_FENCE_SIZE_64M 0x00000600 -#define I915G_FENCE_SIZE_128M 0x00000700 -#define FENCE_PITCH_1 0x00000000 -#define FENCE_PITCH_2 0x00000010 -#define FENCE_PITCH_4 0x00000020 -#define FENCE_PITCH_8 0x00000030 -#define FENCE_PITCH_16 0x00000040 -#define FENCE_PITCH_32 0x00000050 -#define FENCE_PITCH_64 0x00000060 -#define FENCE_VALID 0x00000001 - -#include - -# define MMIO_IN8(base, offset) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) -# define MMIO_IN32(base, offset) \ - read_MMIO_LE32(base, offset) -# define MMIO_OUT8(base, offset, val) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val) -# define MMIO_OUT32(base, offset, val) \ - *(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val) - - - /* Memory mapped register access macros */ -#define INREG8(addr) MMIO_IN8(MMIO, addr) -#define INREG(addr) MMIO_IN32(MMIO, addr) -#define OUTREG8(addr, val) MMIO_OUT8(MMIO, addr, val) -#define OUTREG(addr, val) MMIO_OUT32(MMIO, addr, val) - -#define DSPABASE 0x70184 - -#endif diff --git a/src/mesa/drivers/dri/intel_winsys/server/intel_dri.c b/src/mesa/drivers/dri/intel_winsys/server/intel_dri.c deleted file mode 100644 index e49c4214ad..0000000000 --- a/src/mesa/drivers/dri/intel_winsys/server/intel_dri.c +++ /dev/null @@ -1,1306 +0,0 @@ -/** - * \file server/intel_dri.c - * \brief File to perform the device-specific initialization tasks typically - * done in the X server. - * - * Here they are converted to run in the client (or perhaps a standalone - * process), and to work with the frame buffer device rather than the X - * server infrastructure. - * - * Copyright (C) 2006 Dave Airlie (airlied@linux.ie) - - Permission is hereby granted, free of charge, to any person obtaining a - copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sub license, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial portions - of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR - ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. -*/ - -#include -#include -#include -#include -#include - -#include "driver.h" -#include "drm.h" - -#include "intel.h" -#include "i830_dri.h" - -#include "memops.h" -#include "pciaccess.h" - -static size_t drm_page_size; -static int nextTile = 0; -#define xf86DrvMsg(...) do {} while(0) - -static const int pitches[] = { - 128 * 8, - 128 * 16, - 128 * 32, - 128 * 64, - 0 -}; - -static Bool I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea); - -static unsigned long -GetBestTileAlignment(unsigned long size) -{ - unsigned long i; - - for (i = KB(512); i < size; i <<= 1) - ; - - if (i > MB(64)) - i = MB(64); - - return i; -} - -static void SetFenceRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - unsigned char *MMIO = ctx->MMIOAddress; - - for (i = 0; i < 8; i++) { - OUTREG(FENCE + i * 4, pI830->Fence[i]); - // if (I810_DEBUG & DEBUG_VERBOSE_VGA) - fprintf(stderr,"Fence Register : %x\n", pI830->Fence[i]); - } -} - -/* Tiled memory is good... really, really good... - * - * Need to make it less likely that we miss out on this - probably - * need to move the frontbuffer away from the 'guarenteed' alignment - * of the first memory segment, or perhaps allocate a discontigous - * framebuffer to get more alignment 'sweet spots'. - */ -static void -SetFence(const DRIDriverContext *ctx, I830Rec *pI830, - int nr, unsigned int start, unsigned int pitch, - unsigned int size) -{ - unsigned int val; - unsigned int fence_mask = 0; - unsigned int fence_pitch; - - if (nr < 0 || nr > 7) { - fprintf(stderr, - "SetFence: fence %d out of range\n",nr); - return; - } - - pI830->Fence[nr] = 0; - - if (IS_I9XX(pI830)) - fence_mask = ~I915G_FENCE_START_MASK; - else - fence_mask = ~I830_FENCE_START_MASK; - - if (start & fence_mask) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not %s aligned\n", - nr, start, (IS_I9XX(pI830)) ? "1MB" : "512k"); - return; - } - - if (start % size) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not size (%dk) aligned\n", - nr, start, size / 1024); - return; - } - - if (pitch & 127) { - fprintf(stderr, - "SetFence: %d: pitch (%d) not a multiple of 128 bytes\n", - nr, pitch); - return; - } - - val = (start | FENCE_X_MAJOR | FENCE_VALID); - - if (IS_I9XX(pI830)) { - switch (size) { - case MB(1): - val |= I915G_FENCE_SIZE_1M; - break; - case MB(2): - val |= I915G_FENCE_SIZE_2M; - break; - case MB(4): - val |= I915G_FENCE_SIZE_4M; - break; - case MB(8): - val |= I915G_FENCE_SIZE_8M; - break; - case MB(16): - val |= I915G_FENCE_SIZE_16M; - break; - case MB(32): - val |= I915G_FENCE_SIZE_32M; - break; - case MB(64): - val |= I915G_FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } else { - switch (size) { - case KB(512): - val |= FENCE_SIZE_512K; - break; - case MB(1): - val |= FENCE_SIZE_1M; - break; - case MB(2): - val |= FENCE_SIZE_2M; - break; - case MB(4): - val |= FENCE_SIZE_4M; - break; - case MB(8): - val |= FENCE_SIZE_8M; - break; - case MB(16): - val |= FENCE_SIZE_16M; - break; - case MB(32): - val |= FENCE_SIZE_32M; - break; - case MB(64): - val |= FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } - - if (IS_I9XX(pI830)) - fence_pitch = pitch / 512; - else - fence_pitch = pitch / 128; - - switch (fence_pitch) { - case 1: - val |= FENCE_PITCH_1; - break; - case 2: - val |= FENCE_PITCH_2; - break; - case 4: - val |= FENCE_PITCH_4; - break; - case 8: - val |= FENCE_PITCH_8; - break; - case 16: - val |= FENCE_PITCH_16; - break; - case 32: - val |= FENCE_PITCH_32; - break; - case 64: - val |= FENCE_PITCH_64; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal pitch (%d)\n", nr, pitch); - return; - } - - pI830->Fence[nr] = val; -} - -static Bool -MakeTiles(const DRIDriverContext *ctx, I830Rec *pI830, I830MemRange *pMem) -{ - int pitch, ntiles, i; - - pitch = pMem->Pitch * ctx->cpp; - /* - * Simply try to break the region up into at most four pieces of size - * equal to the alignment. - */ - ntiles = ROUND_TO(pMem->Size, pMem->Alignment) / pMem->Alignment; - if (ntiles >= 4) { - return FALSE; - } - - for (i = 0; i < ntiles; i++, nextTile++) { - SetFence(ctx, pI830, nextTile, pMem->Start + i * pMem->Alignment, - pitch, pMem->Alignment); - } - return TRUE; -} - -static void I830SetupMemoryTiling(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - - /* Clear out */ - for (i = 0; i < 8; i++) - pI830->Fence[i] = 0; - - nextTile = 0; - - if (pI830->BackBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->BackBuffer))) { - fprintf(stderr, - "Activating tiled memory for the back buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the back buffer.\n"); - pI830->allowPageFlip = FALSE; - } - } - - if (pI830->DepthBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->DepthBuffer))) { - fprintf(stderr, - "Activating tiled memory for the depth buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the depth buffer.\n"); - } - } - - return; -} - -static int I830DetectMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - struct pci_device host_bridge, ig_dev; - uint32_t gmch_ctrl; - int memsize = 0; - int range; - uint32_t aper_size; - uint32_t membase2 = 0; - - memset(&host_bridge, 0, sizeof(host_bridge)); - memset(&ig_dev, 0, sizeof(ig_dev)); - - ig_dev.dev = 2; - - pci_device_cfg_read_u32(&host_bridge, &gmch_ctrl, I830_GMCH_CTRL); - - if (IS_I830(pI830) || IS_845G(pI830)) { - if ((gmch_ctrl & I830_GMCH_MEM_MASK) == I830_GMCH_MEM_128M) { - aper_size = 0x80000000; - } else { - aper_size = 0x40000000; - } - } else { - if (IS_I9XX(pI830)) { - int ret; - ret = pci_device_cfg_read_u32(&ig_dev, &membase2, 0x18); - if (membase2 & 0x08000000) - aper_size = 0x8000000; - else - aper_size = 0x10000000; - - fprintf(stderr,"aper size is %08X %08x %d\n", aper_size, membase2, ret); - } else - aper_size = 0x8000000; - } - - pI830->aper_size = aper_size; - - - /* We need to reduce the stolen size, by the GTT and the popup. - * The GTT varying according the the FbMapSize and the popup is 4KB */ - range = (ctx->shared.fbSize / (1024*1024)) + 4; - - if (IS_I85X(pI830) || IS_I865G(pI830) || IS_I9XX(pI830)) { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I855_GMCH_GMS_STOLEN_1M: - memsize = MB(1) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_4M: - memsize = MB(4) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_8M: - memsize = MB(8) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_16M: - memsize = MB(16) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_32M: - memsize = MB(32) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_48M: - if (IS_I9XX(pI830)) - memsize = MB(48) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_64M: - if (IS_I9XX(pI830)) - memsize = MB(64) - KB(range); - break; - } - } else { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I830_GMCH_GMS_STOLEN_512: - memsize = KB(512) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_1024: - memsize = MB(1) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_8192: - memsize = MB(8) - KB(range); - break; - case I830_GMCH_GMS_LOCAL: - memsize = 0; - xf86DrvMsg(pScrn->scrnIndex, X_WARNING, - "Local memory found, but won't be used.\n"); - break; - } - } - if (memsize > 0) { - fprintf(stderr, - "detected %d kB stolen memory.\n", memsize / 1024); - } else { - fprintf(stderr, - "no video memory detected.\n"); - } - return memsize; -} - -static int AgpInit(const DRIDriverContext *ctx, I830Rec *info) -{ - unsigned long mode = 0x4; - - if (drmAgpAcquire(ctx->drmFD) < 0) { - fprintf(stderr, "[gart] AGP not available\n"); - return 0; - } - - if (drmAgpEnable(ctx->drmFD, mode) < 0) { - fprintf(stderr, "[gart] AGP not enabled\n"); - drmAgpRelease(ctx->drmFD); - return 0; - } - else - fprintf(stderr, "[gart] AGP enabled at %dx\n", ctx->agpmode); - - return 1; -} - -/* - * Allocate memory from the given pool. Grow the pool if needed and if - * possible. - */ -static unsigned long -AllocFromPool(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, - long size, unsigned long alignment, int flags) -{ - long needed, start, end; - - if (!result || !pool || !size) - return 0; - - /* Calculate how much space is needed. */ - if (alignment <= GTT_PAGE_SIZE) - needed = size; - else { - start = ROUND_TO(pool->Free.Start, alignment); - end = ROUND_TO(start + size, alignment); - needed = end - pool->Free.Start; - } - if (needed > pool->Free.Size) { - return 0; - } - - result->Start = ROUND_TO(pool->Free.Start, alignment); - pool->Free.Start += needed; - result->End = pool->Free.Start; - - pool->Free.Size = pool->Free.End - pool->Free.Start; - result->Size = result->End - result->Start; - result->Pool = pool; - result->Alignment = alignment; - return needed; -} - -static unsigned long AllocFromAGP(const DRIDriverContext *ctx, I830Rec *pI830, long size, unsigned long alignment, I830MemRange *result) -{ - unsigned long start, end; - unsigned long newApStart, newApEnd; - int ret; - if (!result || !size) - return 0; - - if (!alignment) - alignment = 4; - - start = ROUND_TO(pI830->MemoryAperture.Start, alignment); - end = ROUND_TO(start + size, alignment); - newApStart = end; - newApEnd = pI830->MemoryAperture.End; - - ret=drmAgpAlloc(ctx->drmFD, size, 0, &(result->Physical), (drm_handle_t *)&(result->Key)); - - if (ret) - { - fprintf(stderr,"drmAgpAlloc failed %d\n", ret); - return 0; - } - pI830->allocatedMemory += size; - pI830->MemoryAperture.Start = newApStart; - pI830->MemoryAperture.End = newApEnd; - pI830->MemoryAperture.Size = newApEnd - newApStart; - // pI830->FreeMemory -= size; - result->Start = start; - result->End = start + size; - result->Size = size; - result->Offset = start; - result->Alignment = alignment; - result->Pool = NULL; - - return size; -} - -unsigned long -I830AllocVidMem(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, long size, - unsigned long alignment, int flags) -{ - unsigned long ret; - - if (!result) - return 0; - - /* Make sure these are initialised. */ - result->Size = 0; - result->Key = -1; - - if (!size) { - return 0; - } - - if (pool->Free.Size < size) { - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - else { - ret = AllocFromPool(ctx, pI830, result, pool, size, alignment, flags); - if (ret == 0) - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - return ret; -} - -static Bool BindAgpRange(const DRIDriverContext *ctx, I830MemRange *mem) -{ - if (!mem) - return FALSE; - - if (mem->Key == -1) - return TRUE; - - return !drmAgpBind(ctx->drmFD, mem->Key, mem->Offset); -} - -/* simple memory allocation routines needed */ -/* put ring buffer in low memory */ -/* need to allocate front, back, depth buffers aligned correctly, - allocate ring buffer, -*/ - -/* */ -static Bool -I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned long size, ret; - unsigned long lines, lineSize, align; - - /* allocate ring buffer */ - memset(pI830->LpRing, 0, sizeof(I830RingBuffer)); - pI830->LpRing->mem.Key = -1; - - size = PRIMARY_RINGBUFFER_SIZE; - - ret = I830AllocVidMem(ctx, pI830, &pI830->LpRing->mem, &pI830->StolenPool, size, 0x1000, 0); - - if (ret != size) - { - fprintf(stderr,"unable to allocate ring buffer %ld\n", ret); - return FALSE; - } - - pI830->LpRing->tail_mask = pI830->LpRing->mem.Size - 1; - - - /* allocate front buffer */ - memset(&(pI830->FrontBuffer), 0, sizeof(pI830->FrontBuffer)); - pI830->FrontBuffer.Key = -1; - pI830->FrontBuffer.Pitch = ctx->shared.virtualWidth; - - align = KB(512); - - lineSize = ctx->shared.virtualWidth * ctx->cpp; - lines = (ctx->shared.virtualHeight + 15) / 16 * 16; - size = lineSize * lines; - size = ROUND_TO_PAGE(size); - - align = GetBestTileAlignment(size); - - ret = I830AllocVidMem(ctx, pI830, &pI830->FrontBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate front buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->BackBuffer), 0, sizeof(pI830->BackBuffer)); - pI830->BackBuffer.Key = -1; - pI830->BackBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->BackBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate back buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->DepthBuffer), 0, sizeof(pI830->DepthBuffer)); - pI830->DepthBuffer.Key = -1; - pI830->DepthBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->DepthBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate depth buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->ContextMem), 0, sizeof(pI830->ContextMem)); - pI830->ContextMem.Key = -1; - size = KB(32); - - ret = I830AllocVidMem(ctx, pI830, &pI830->ContextMem, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate context buffer %ld\n", ret); - return FALSE; - } - -#if 0 - memset(&(pI830->TexMem), 0, sizeof(pI830->TexMem)); - pI830->TexMem.Key = -1; - - size = 32768 * 1024; - ret = AllocFromAGP(ctx, pI830, size, align, &pI830->TexMem); - if (ret < size) - { - fprintf(stderr,"unable to allocate texture memory %ld\n", ret); - return FALSE; - } -#endif - - return TRUE; -} - -static Bool -I830BindMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - if (!BindAgpRange(ctx, &pI830->LpRing->mem)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->FrontBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->BackBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->DepthBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->ContextMem)) - return FALSE; -#if 0 - if (!BindAgpRange(ctx, &pI830->TexMem)) - return FALSE; -#endif - return TRUE; -} - -static void SetupDRIMM(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned long aperEnd = ROUND_DOWN_TO(pI830->aper_size, GTT_PAGE_SIZE) / GTT_PAGE_SIZE; - unsigned long aperStart = ROUND_TO(pI830->aper_size - KB(32768), GTT_PAGE_SIZE) / GTT_PAGE_SIZE; - - fprintf(stderr, "aper size is %08X\n", ctx->shared.fbSize); - if (drmMMInit(ctx->drmFD, aperStart, aperEnd - aperStart, DRM_BO_MEM_TT)) { - fprintf(stderr, - "DRM MM Initialization Failed\n"); - } else { - fprintf(stderr, - "DRM MM Initialized at offset 0x%lx length %d page\n", aperStart, aperEnd-aperStart); - } - -} - -static Bool -I830CleanupDma(const DRIDriverContext *ctx) -{ - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_CLEANUP_DMA; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, "I830 Dma Cleanup Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830InitDma(const DRIDriverContext *ctx, I830Rec *pI830) -{ - I830RingBuffer *ring = pI830->LpRing; - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_INIT_DMA; - - info.ring_start = ring->mem.Start + pI830->LinearAddr; - info.ring_end = ring->mem.End + pI830->LinearAddr; - info.ring_size = ring->mem.Size; - - info.mmio_offset = (unsigned int)ctx->MMIOStart; - - info.sarea_priv_offset = sizeof(drm_sarea_t); - - info.front_offset = pI830->FrontBuffer.Start; - info.back_offset = pI830->BackBuffer.Start; - info.depth_offset = pI830->DepthBuffer.Start; - info.w = ctx->shared.virtualWidth; - info.h = ctx->shared.virtualHeight; - info.pitch = ctx->shared.virtualWidth; - info.back_pitch = pI830->BackBuffer.Pitch; - info.depth_pitch = pI830->DepthBuffer.Pitch; - info.cpp = ctx->cpp; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, - "I830 Dma Initialization Failed\n"); - return FALSE; - } - - return TRUE; -} - -static int I830CheckDRMVersion( const DRIDriverContext *ctx, - I830Rec *pI830 ) -{ - drmVersionPtr version; - - version = drmGetVersion(ctx->drmFD); - - if (version) { - int req_minor, req_patch; - - req_minor = 4; - req_patch = 0; - - if (version->version_major != 1 || - version->version_minor < req_minor || - (version->version_minor == req_minor && - version->version_patchlevel < req_patch)) { - /* Incompatible drm version */ - fprintf(stderr, - "[dri] I830DRIScreenInit failed because of a version " - "mismatch.\n" - "[dri] i915.o kernel module version is %d.%d.%d " - "but version 1.%d.%d or newer is needed.\n" - "[dri] Disabling DRI.\n", - version->version_major, - version->version_minor, - version->version_patchlevel, - req_minor, - req_patch); - drmFreeVersion(version); - return 0; - } - - pI830->drmMinor = version->version_minor; - drmFreeVersion(version); - } - return 1; -} - -static void -I830SetRingRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned int itemp; - unsigned char *MMIO = ctx->MMIOAddress; - - OUTREG(LP_RING + RING_LEN, 0); - OUTREG(LP_RING + RING_TAIL, 0); - OUTREG(LP_RING + RING_HEAD, 0); - - if ((long)(pI830->LpRing->mem.Start & I830_RING_START_MASK) != - pI830->LpRing->mem.Start) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer start (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Start, I830_RING_START_MASK); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = pI830->LpRing->mem.Start & I830_RING_START_MASK; - OUTREG(LP_RING + RING_START, itemp); - - if (((pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES) != - pI830->LpRing->mem.Size - 4096) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer size - 4096 (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Size - 4096, - I830_RING_NR_PAGES); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = (pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES; - itemp |= (RING_NO_REPORT | RING_VALID); - OUTREG(LP_RING + RING_LEN, itemp); - - pI830->LpRing->head = INREG(LP_RING + RING_HEAD) & I830_HEAD_MASK; - pI830->LpRing->tail = INREG(LP_RING + RING_TAIL); - pI830->LpRing->space = pI830->LpRing->head - (pI830->LpRing->tail + 8); - if (pI830->LpRing->space < 0) - pI830->LpRing->space += pI830->LpRing->mem.Size; - - SetFenceRegs(ctx, pI830); - - /* RESET THE DISPLAY PIPE TO POINT TO THE FRONTBUFFER - hacky - hacky hacky */ - OUTREG(DSPABASE, pI830->FrontBuffer.Start + pI830->LinearAddr); - -} - -static Bool -I830SetParam(const DRIDriverContext *ctx, int param, int value) -{ - drmI830SetParam sp; - - memset(&sp, 0, sizeof(sp)); - sp.param = param; - sp.value = value; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_SETPARAM, &sp, sizeof(sp))) { - fprintf(stderr, "I830 SetParam Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - fprintf(stderr, - "[drm] Mapping front buffer\n"); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->front_offset + pI830->LinearAddr), - sarea->front_size, - DRM_FRAME_BUFFER, /*DRM_AGP,*/ - 0, - &sarea->front_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(front_handle) failed. Disabling DRI\n"); - return FALSE; - } - ctx->shared.hFrameBuffer = sarea->front_handle; - ctx->shared.fbSize = sarea->front_size; - fprintf(stderr, "[drm] Front Buffer = 0x%08x\n", - sarea->front_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->back_offset), - sarea->back_size, DRM_AGP, 0, - &sarea->back_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(back_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Back Buffer = 0x%08x\n", - sarea->back_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->depth_offset, - sarea->depth_size, DRM_AGP, 0, - &sarea->depth_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(depth_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Depth Buffer = 0x%08x\n", - sarea->depth_handle); - -#if 0 - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->tex_offset, - sarea->tex_size, DRM_AGP, 0, - &sarea->tex_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(tex_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] textures = 0x%08x\n", - sarea->tex_handle); -#endif - return TRUE; -} - - -static void -I830DRIUnmapScreenRegions(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ -#if 1 - if (sarea->front_handle) { - drmRmMap(ctx->drmFD, sarea->front_handle); - sarea->front_handle = 0; - } -#endif - if (sarea->back_handle) { - drmRmMap(ctx->drmFD, sarea->back_handle); - sarea->back_handle = 0; - } - if (sarea->depth_handle) { - drmRmMap(ctx->drmFD, sarea->depth_handle); - sarea->depth_handle = 0; - } - if (sarea->tex_handle) { - drmRmMap(ctx->drmFD, sarea->tex_handle); - sarea->tex_handle = 0; - } -} - -static Bool -I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - if (drmAddMap(ctx->drmFD, - (drm_handle_t)pI830->LpRing->mem.Start, - pI830->LpRing->mem.Size, DRM_AGP, 0, - &pI830->ring_map) < 0) { - fprintf(stderr, - "[drm] drmAddMap(ring_map) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] ring buffer = 0x%08x\n", - pI830->ring_map); - - if (I830InitDma(ctx, pI830) == FALSE) { - return FALSE; - } - - /* init to zero to be safe */ - - I830DRIMapScreenRegions(ctx, pI830, sarea); - SetupDRIMM(ctx, pI830); - - if (ctx->pciDevice != PCI_CHIP_845_G && - ctx->pciDevice != PCI_CHIP_I830_M) { - I830SetParam(ctx, I830_SETPARAM_USE_MI_BATCHBUFFER_START, 1 ); - } - - /* Okay now initialize the dma engine */ - { - pI830->irq = drmGetInterruptFromBusID(ctx->drmFD, - ctx->pciBus, - ctx->pciDevice, - ctx->pciFunc); - - if (drmCtlInstHandler(ctx->drmFD, pI830->irq)) { - fprintf(stderr, - "[drm] failure adding irq handler\n"); - pI830->irq = 0; - return FALSE; - } - else - fprintf(stderr, - "[drm] dma control initialized, using IRQ %d\n", - pI830->irq); - } - - fprintf(stderr, "[dri] visual configs initialized\n"); - - return TRUE; -} - -static Bool -I830ClearScreen(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - /* need to drmMap front and back buffers and zero them */ - drmAddress map_addr; - int ret; - - ret = drmMap(ctx->drmFD, - sarea->front_handle, - sarea->front_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map front buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->front_size); - drmUnmap(map_addr, sarea->front_size); - - - ret = drmMap(ctx->drmFD, - sarea->back_handle, - sarea->back_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map back buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->back_size); - drmUnmap(map_addr, sarea->back_size); - - return TRUE; -} - -static Bool -I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) - -{ - I830DRIPtr pI830DRI; - drmI830Sarea *pSAREAPriv; - int err; - - drm_page_size = getpagesize(); - - pI830->registerSize = ctx->MMIOSize; - /* This is a hack for now. We have to have more than a 4k page here - * because of the size of the state. However, the state should be - * in a per-context mapping. This will be added in the Mesa 3.5 port - * of the I830 driver. - */ - ctx->shared.SAREASize = SAREA_MAX; - - /* Note that drmOpen will try to load the kernel module, if needed. */ - ctx->drmFD = drmOpen("i915", NULL ); - if (ctx->drmFD < 0) { - fprintf(stderr, "[drm] drmOpen failed\n"); - return 0; - } - - if ((err = drmSetBusid(ctx->drmFD, ctx->pciBusID)) < 0) { - fprintf(stderr, "[drm] drmSetBusid failed (%d, %s), %s\n", - ctx->drmFD, ctx->pciBusID, strerror(-err)); - return 0; - } - - if (drmAddMap( ctx->drmFD, - 0, - ctx->shared.SAREASize, - DRM_SHM, - DRM_CONTAINS_LOCK, - &ctx->shared.hSAREA) < 0) - { - fprintf(stderr, "[drm] drmAddMap failed\n"); - return 0; - } - - fprintf(stderr, "[drm] added %d byte SAREA at 0x%08x\n", - ctx->shared.SAREASize, ctx->shared.hSAREA); - - if (drmMap( ctx->drmFD, - ctx->shared.hSAREA, - ctx->shared.SAREASize, - (drmAddressPtr)(&ctx->pSAREA)) < 0) - { - fprintf(stderr, "[drm] drmMap failed\n"); - return 0; - - } - - memset(ctx->pSAREA, 0, ctx->shared.SAREASize); - fprintf(stderr, "[drm] mapped SAREA 0x%08x to %p, size %d\n", - ctx->shared.hSAREA, ctx->pSAREA, ctx->shared.SAREASize); - - - if (drmAddMap(ctx->drmFD, - ctx->MMIOStart, - ctx->MMIOSize, - DRM_REGISTERS, - DRM_READ_ONLY, - &pI830->registerHandle) < 0) { - fprintf(stderr, "[drm] drmAddMap mmio failed\n"); - return 0; - } - fprintf(stderr, - "[drm] register handle = 0x%08x\n", pI830->registerHandle); - - - if (!I830CheckDRMVersion(ctx, pI830)) { - return FALSE; - } - - /* Create a 'server' context so we can grab the lock for - * initialization ioctls. - */ - if ((err = drmCreateContext(ctx->drmFD, &ctx->serverContext)) != 0) { - fprintf(stderr, "%s: drmCreateContext failed %d\n", __FUNCTION__, err); - return 0; - } - - DRM_LOCK(ctx->drmFD, ctx->pSAREA, ctx->serverContext, 0); - - /* Initialize the SAREA private data structure */ - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - memset(pSAREAPriv, 0, sizeof(*pSAREAPriv)); - - pI830->StolenMemory.Size = I830DetectMemory(ctx, pI830); - pI830->StolenMemory.Start = 0; - pI830->StolenMemory.End = pI830->StolenMemory.Size; - - pI830->MemoryAperture.Start = pI830->StolenMemory.End; - pI830->MemoryAperture.End = KB(40000); - pI830->MemoryAperture.Size = pI830->MemoryAperture.End - pI830->MemoryAperture.Start; - - pI830->StolenPool.Fixed = pI830->StolenMemory; - pI830->StolenPool.Total = pI830->StolenMemory; - pI830->StolenPool.Free = pI830->StolenPool.Total; - pI830->FreeMemory = pI830->StolenPool.Total.Size; - - if (!AgpInit(ctx, pI830)) - return FALSE; - - if (I830AllocateMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - if (I830BindMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - pSAREAPriv->rotated_offset = -1; - pSAREAPriv->rotated_size = 0; - pSAREAPriv->rotated_pitch = ctx->shared.virtualWidth; - - pSAREAPriv->front_offset = pI830->FrontBuffer.Start; - pSAREAPriv->front_size = pI830->FrontBuffer.Size; - pSAREAPriv->width = ctx->shared.virtualWidth; - pSAREAPriv->height = ctx->shared.virtualHeight; - pSAREAPriv->pitch = ctx->shared.virtualWidth; - pSAREAPriv->virtualX = ctx->shared.virtualWidth; - pSAREAPriv->virtualY = ctx->shared.virtualHeight; - pSAREAPriv->back_offset = pI830->BackBuffer.Start; - pSAREAPriv->back_size = pI830->BackBuffer.Size; - pSAREAPriv->depth_offset = pI830->DepthBuffer.Start; - pSAREAPriv->depth_size = pI830->DepthBuffer.Size; -#if 0 - pSAREAPriv->tex_offset = pI830->TexMem.Start; - pSAREAPriv->tex_size = pI830->TexMem.Size; -#endif - pSAREAPriv->log_tex_granularity = pI830->TexGranularity; - - ctx->driverClientMsg = malloc(sizeof(I830DRIRec)); - ctx->driverClientMsgSize = sizeof(I830DRIRec); - pI830DRI = (I830DRIPtr)ctx->driverClientMsg; - pI830DRI->deviceID = pI830->Chipset; - pI830DRI->regsSize = I830_REG_SIZE; - pI830DRI->width = ctx->shared.virtualWidth; - pI830DRI->height = ctx->shared.virtualHeight; - pI830DRI->mem = ctx->shared.fbSize; - pI830DRI->cpp = ctx->cpp; - - pI830DRI->bitsPerPixel = ctx->bpp; - pI830DRI->sarea_priv_offset = sizeof(drm_sarea_t); - - err = I830DRIDoMappings(ctx, pI830, pSAREAPriv); - if (err == FALSE) - return FALSE; - - I830SetupMemoryTiling(ctx, pI830); - - /* Quick hack to clear the front & back buffers. Could also use - * the clear ioctl to do this, but would need to setup hw state - * first. - */ - I830ClearScreen(ctx, pI830, pSAREAPriv); - - I830SetRingRegs(ctx, pI830); - - return TRUE; -} - - -/** - * \brief Validate the fbdev mode. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Saves some registers and returns 1. - * - * \sa radeonValidateMode(). - */ -static int i830ValidateMode( const DRIDriverContext *ctx ) -{ - return 1; -} - -/** - * \brief Examine mode returned by fbdev. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Restores registers that fbdev has clobbered and returns 1. - * - * \sa i810ValidateMode(). - */ -static int i830PostValidateMode( const DRIDriverContext *ctx ) -{ - I830Rec *pI830 = ctx->driverPrivate; - - I830SetRingRegs(ctx, pI830); - return 1; -} - - -/** - * \brief Initialize the framebuffer device mode - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Fills in \p info with some default values and some information from \p ctx - * and then calls I810ScreenInit() for the screen initialization. - * - * Before exiting clears the framebuffer memory accessing it directly. - */ -static int i830InitFBDev( DRIDriverContext *ctx ) -{ - I830Rec *pI830 = calloc(1, sizeof(I830Rec)); - int i; - - { - int dummy = ctx->shared.virtualWidth; - - switch (ctx->bpp / 8) { - case 1: dummy = (ctx->shared.virtualWidth + 127) & ~127; break; - case 2: dummy = (ctx->shared.virtualWidth + 31) & ~31; break; - case 3: - case 4: dummy = (ctx->shared.virtualWidth + 15) & ~15; break; - } - - ctx->shared.virtualWidth = dummy; - ctx->shared.Width = ctx->shared.virtualWidth; - } - - - for (i = 0; pitches[i] != 0; i++) { - if (pitches[i] >= ctx->shared.virtualWidth) { - ctx->shared.virtualWidth = pitches[i]; - break; - } - } - - ctx->driverPrivate = (void *)pI830; - - pI830->LpRing = calloc(1, sizeof(I830RingBuffer)); - pI830->Chipset = ctx->chipset; - pI830->LinearAddr = ctx->FBStart; - - if (!I830ScreenInit( ctx, pI830 )) - return 0; - - - return 1; -} - - -/** - * \brief The screen is being closed, so clean up any state and free any - * resources used by the DRI. - * - * \param ctx display handle. - * - * Unmaps the SAREA, closes the DRM device file descriptor and frees the driver - * private data. - */ -static void i830HaltFBDev( DRIDriverContext *ctx ) -{ - drmI830Sarea *pSAREAPriv; - I830Rec *pI830 = ctx->driverPrivate; - - if (pI830->irq) { - drmCtlUninstHandler(ctx->drmFD); - pI830->irq = 0; } - - I830CleanupDma(ctx); - - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - - I830DRIUnmapScreenRegions(ctx, pI830, pSAREAPriv); - drmUnmap( ctx->pSAREA, ctx->shared.SAREASize ); - drmClose(ctx->drmFD); - - if (ctx->driverPrivate) { - free(ctx->driverPrivate); - ctx->driverPrivate = 0; - } -} - - -extern void i810NotifyFocus( int ); - -/** - * \brief Exported driver interface for Mini GLX. - * - * \sa DRIDriverRec. - */ -const struct DRIDriverRec __driDriver = { - i830ValidateMode, - i830PostValidateMode, - i830InitFBDev, - i830HaltFBDev, - NULL,//I830EngineShutdown, - NULL, //I830EngineRestore, -#ifndef _EMBEDDED - 0, -#else - i810NotifyFocus, -#endif -}; diff --git a/src/mesa/pipe/Makefile b/src/mesa/pipe/Makefile deleted file mode 100644 index d880d090c1..0000000000 --- a/src/mesa/pipe/Makefile +++ /dev/null @@ -1,28 +0,0 @@ -TOP = ../../.. -include $(TOP)/configs/current - - -ifeq ($(CONFIG_NAME), linux-cell) -CELL_DIR = cell -endif - -ifeq ($(CONFIG_NAME), linux-llvm) -LLVM_DIR = llvm -endif - -SUBDIRS = softpipe i915simple i965simple failover pipebuffer $(CELL_DIR) $(LLVM_DIR) - - -default: subdirs - - -subdirs: - @for dir in $(SUBDIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ - fi \ - done - - -clean: - rm -f `find . -name \*.[oa]` diff --git a/src/mesa/pipe/Makefile.template b/src/mesa/pipe/Makefile.template deleted file mode 100644 index 8e84f8eb2d..0000000000 --- a/src/mesa/pipe/Makefile.template +++ /dev/null @@ -1,63 +0,0 @@ -# -*-makefile-*- - - -# We still have a dependency on the "dri" buffer manager. Most likely -# the interface can be reused in non-dri environments, and also as a -# frontend to simpler memory managers. -# -COMMON_SOURCES = - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(CPP_SOURCES:.cpp=.o) \ - $(ASM_SOURCES:.S=.o) - - -### Include directories -INCLUDES = \ - -I. \ - -I$(TOP)/src/mesa/pipe \ - -I$(TOP)/src/mesa \ - -I$(TOP)/include \ - $(DRIVER_INCLUDES) - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.cpp.o: - $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - - -##### TARGETS ##### - -default: depend symlinks $(LIBNAME) - - -$(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/mesa/pipe/Makefile.template - $(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS) - - -depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ - $(ASM_SOURCES) 2> /dev/null - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean:: - -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) - -rm -f depend depend.bak - - -include depend diff --git a/src/mesa/pipe/README.portability b/src/mesa/pipe/README.portability deleted file mode 100644 index c70ca774da..0000000000 --- a/src/mesa/pipe/README.portability +++ /dev/null @@ -1,43 +0,0 @@ - CROSS-PLATFORM PORTABILITY GUIDELINES FOR GALLIUM3D - - -= General Considerations = - -The state tracker and winsys driver support a rather limited number of -platforms. However, the pipe drivers are meant to run in a wide number of -platforms. Hence the pipe drivers, the auxiliary modules, and all public -headers in general, should stricly follow these guidelines to ensure - - -= Compiler Support = - -* Include the p_compiler.h. - -* Don't use the 'inline' keyword, use the INLINE macro in p_compiler.h instead. - -* Cast explicitly when converting to integer types of smaller sizes. - -* Cast explicitly when converting between float, double and integral types. - -* Don't use named struct initializers. - -* Don't use variable number of macro arguments. Use static inline functions -instead. - - -= Standard Library = - -* Avoid including standard library headers. Most standard library functions are -not available in Windows Kernel Mode. Use the appropriate p_*.h include. - -== Memory Allocation == - -* Use MALLOC, CALLOC, FREE instead of the malloc, calloc, free functions. - -* Use align_pointer() function defined in p_util.h for aligning pointers in a -portable way. - -== Debugging == - -TODO - diff --git a/src/mesa/pipe/SConscript b/src/mesa/pipe/SConscript deleted file mode 100644 index d9c20e0100..0000000000 --- a/src/mesa/pipe/SConscript +++ /dev/null @@ -1,9 +0,0 @@ -Import('*') - -#env = env.Clone() - -SConscript([ - 'softpipe/SConscript', - 'i915simple/SConscript', - 'i965simple/SConscript', -]) diff --git a/src/mesa/pipe/cell/Makefile b/src/mesa/pipe/cell/Makefile deleted file mode 100644 index 47aef7b05f..0000000000 --- a/src/mesa/pipe/cell/Makefile +++ /dev/null @@ -1,12 +0,0 @@ -# Cell Gallium driver Makefile - - -default: - ( cd spu ; make ) - ( cd ppu ; make ) - - - -clean: - ( cd spu ; make clean ) - ( cd ppu ; make clean ) diff --git a/src/mesa/pipe/cell/common.h b/src/mesa/pipe/cell/common.h deleted file mode 100644 index 4de514c358..0000000000 --- a/src/mesa/pipe/cell/common.h +++ /dev/null @@ -1,220 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Types and tokens which are common to the SPU and PPU code. - */ - - -#ifndef CELL_COMMON_H -#define CELL_COMMON_H - -#include "pipe/p_compiler.h" -#include "pipe/p_util.h" -#include "pipe/p_format.h" - - -/** The standard assert macro doesn't seem to work reliably */ -#define ASSERT(x) \ - if (!(x)) { \ - ubyte *p = NULL; \ - fprintf(stderr, "%s:%d: %s(): assertion %s failed.\n", \ - __FILE__, __LINE__, __FUNCTION__, #x); \ - *p = 0; \ - exit(1); \ - } - - -/** for sanity checking */ -#define ASSERT_ALIGN16(ptr) \ - ASSERT((((unsigned long) (ptr)) & 0xf) == 0); - - -/** round up value to next multiple of 4 */ -#define ROUNDUP4(k) (((k) + 0x3) & ~0x3) - -/** round up value to next multiple of 8 */ -#define ROUNDUP8(k) (((k) + 0x7) & ~0x7) - -/** round up value to next multiple of 16 */ -#define ROUNDUP16(k) (((k) + 0xf) & ~0xf) - - -#define CELL_MAX_SPUS 6 - -#define TILE_SIZE 32 - - -/** - * The low byte of a mailbox word contains the command opcode. - * Remaining higher bytes are command specific. - */ -#define CELL_CMD_OPCODE_MASK 0xff - -#define CELL_CMD_EXIT 1 -#define CELL_CMD_CLEAR_SURFACE 2 -#define CELL_CMD_FINISH 3 -#define CELL_CMD_RENDER 4 -#define CELL_CMD_BATCH 5 -#define CELL_CMD_RELEASE_VERTS 6 -#define CELL_CMD_STATE_FRAMEBUFFER 10 -#define CELL_CMD_STATE_DEPTH_STENCIL 11 -#define CELL_CMD_STATE_SAMPLER 12 -#define CELL_CMD_STATE_TEXTURE 13 -#define CELL_CMD_STATE_VERTEX_INFO 14 -#define CELL_CMD_STATE_VIEWPORT 15 -#define CELL_CMD_STATE_VS_ARRAY_INFO 16 -#define CELL_CMD_STATE_BLEND 17 -#define CELL_CMD_VS_EXECUTE 18 - - -#define CELL_NUM_BUFFERS 4 -#define CELL_BUFFER_SIZE (4*1024) /**< 16KB would be the max */ - -#define CELL_BUFFER_STATUS_FREE 10 -#define CELL_BUFFER_STATUS_USED 20 - - - -/** - * Tell SPUs about the framebuffer size, location - */ -struct cell_command_framebuffer -{ - uint64_t opcode; /**< CELL_CMD_FRAMEBUFFER */ - int width, height; - void *color_start, *depth_start; - enum pipe_format color_format, depth_format; -}; - - -/** - * Clear framebuffer to the given value/color. - */ -struct cell_command_clear_surface -{ - uint64_t opcode; /**< CELL_CMD_CLEAR_SURFACE */ - uint surface; /**< Temporary: 0=color, 1=Z */ - uint value; -}; - - -/** - * Array info used by the vertex shader's vertex puller. - */ -struct cell_array_info -{ - uint64_t base; /**< Base address of the 0th element. */ - uint attr; /**< Attribute that this state is for. */ - uint pitch; /**< Byte pitch from one entry to the next. */ - uint format; /**< Pipe format of each entry. */ -} ALIGN16_ATTRIB; - - -struct cell_shader_info -{ - unsigned num_outputs; - - uint64_t declarations; - unsigned num_declarations; - uint64_t instructions; - unsigned num_instructions; - uint64_t uniforms; - uint64_t immediates; - unsigned num_immediates; -} ALIGN16_ATTRIB; - - -#define SPU_VERTS_PER_BATCH 64 -struct cell_command_vs -{ - uint64_t opcode; /**< CELL_CMD_VS_EXECUTE */ - struct cell_shader_info shader; - unsigned num_elts; - unsigned elts[SPU_VERTS_PER_BATCH]; - uint64_t vOut[SPU_VERTS_PER_BATCH]; - float plane[12][4]; - unsigned nr_planes; - unsigned nr_attrs; -} ALIGN16_ATTRIB; - - -struct cell_command_render -{ - uint64_t opcode; /**< CELL_CMD_RENDER */ - uint prim_type; /**< PIPE_PRIM_x */ - uint num_verts; - uint vertex_size; /**< bytes per vertex */ - uint num_indexes; - uint vertex_buf; /**< which cell->buffer[] contains the vertex data */ - float xmin, ymin, xmax, ymax; /* XXX another dummy field */ - uint min_index; - boolean inline_verts; -}; - - -struct cell_command_release_verts -{ - uint64_t opcode; /**< CELL_CMD_RELEASE_VERTS */ - uint vertex_buf; /**< in [0, CELL_NUM_BUFFERS-1] */ -}; - - -struct cell_command_texture -{ - void *start; /**< Address in main memory */ - uint width, height; -}; - - -/** XXX unions don't seem to work */ -/* XXX this should go away; all commands should be placed in batch buffers */ -struct cell_command -{ -#if 0 - struct cell_command_framebuffer fb; - struct cell_command_clear_surface clear; - struct cell_command_render render; -#endif - struct cell_command_vs vs; -} ALIGN16_ATTRIB; - - -/** This is the object passed to spe_create_thread() */ -struct cell_init_info -{ - unsigned id; - unsigned num_spus; - struct cell_command *cmd; - - /** Buffers for command batches, vertex/index data */ - ubyte *buffers[CELL_NUM_BUFFERS]; - uint *buffer_status; /**< points at cell_context->buffer_status */ -} ALIGN16_ATTRIB; - - -#endif /* CELL_COMMON_H */ diff --git a/src/mesa/pipe/cell/ppu/Makefile b/src/mesa/pipe/cell/ppu/Makefile deleted file mode 100644 index 50060f5cd3..0000000000 --- a/src/mesa/pipe/cell/ppu/Makefile +++ /dev/null @@ -1,76 +0,0 @@ -# Gallium3D Cell driver: PPU code - -# This makefile builds the g3dcell.a library which gets pulled into -# the main libGL.so library - - -TOP = ../../../../.. -include $(TOP)/configs/linux-cell - - -#PROG = gl4 - -CELL_LIB = libcell.a - -SPU_CODE_MODULE = ../spu/g3d_spu.a - - -SOURCES = \ - cell_batch.c \ - cell_clear.c \ - cell_context.c \ - cell_draw_arrays.c \ - cell_flush.c \ - cell_state_blend.c \ - cell_state_clip.c \ - cell_state_derived.c \ - cell_state_emit.c \ - cell_state_fs.c \ - cell_state_rasterizer.c \ - cell_state_sampler.c \ - cell_state_surface.c \ - cell_state_vertex.c \ - cell_spu.c \ - cell_surface.c \ - cell_texture.c \ - cell_vbuf.c \ - cell_vertex_shader.c \ - cell_winsys.c - - -OBJECTS = $(SOURCES:.c=.o) \ - -INCLUDE_DIRS = -I$(TOP)/src/mesa - - -.c.o: - $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ - - - -default: $(CELL_LIB) - - -$(CELL_LIB): $(OBJECTS) $(SPU_CODE_MODULE) -# ar -ru $(CELL_LIB) $(OBJECTS) $(SPU_CODE_MODULE) - ar -ru $(CELL_LIB) $(OBJECTS) - -#$(PROG): $(PPU_OBJECTS) -# $(CC) -o $(PROG) $(PPU_OBJECTS) $(SPU_CODE_MODULE) $(PPU_LFLAGS) - - - -clean: - rm -f *.o *~ $(CELL_LIB) - - - -depend: $(SOURCES) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(SOURCES) 2> /dev/null - -include depend - - - diff --git a/src/mesa/pipe/cell/ppu/cell_batch.c b/src/mesa/pipe/cell/ppu/cell_batch.c deleted file mode 100644 index f45e5f25b6..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_batch.c +++ /dev/null @@ -1,217 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "cell_context.h" -#include "cell_batch.h" -#include "cell_spu.h" - - - -uint -cell_get_empty_buffer(struct cell_context *cell) -{ - uint buf = 0, tries = 0; - - /* Find a buffer that's marked as free by all SPUs */ - while (1) { - uint spu, num_free = 0; - - for (spu = 0; spu < cell->num_spus; spu++) { - if (cell->buffer_status[spu][buf][0] == CELL_BUFFER_STATUS_FREE) { - num_free++; - - if (num_free == cell->num_spus) { - /* found a free buffer, now mark status as used */ - for (spu = 0; spu < cell->num_spus; spu++) { - cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED; - } - /* - printf("PPU: ALLOC BUFFER %u\n", buf); - */ - return buf; - } - } - else { - break; - } - } - - /* try next buf */ - buf = (buf + 1) % CELL_NUM_BUFFERS; - - tries++; - if (tries == 100) { - /* - printf("PPU WAITING for buffer...\n"); - */ - } - } -} - - -void -cell_batch_flush(struct cell_context *cell) -{ - static boolean flushing = FALSE; - uint batch = cell->cur_batch; - const uint size = cell->buffer_size[batch]; - uint spu, cmd_word; - - assert(!flushing); - - if (size == 0) - return; - - flushing = TRUE; - - assert(batch < CELL_NUM_BUFFERS); - - /* - printf("cell_batch_dispatch: buf %u at %p, size %u\n", - batch, &cell->batch_buffer[batch][0], size); - */ - - /* - * Build "BATCH" command and sent to all SPUs. - */ - cmd_word = CELL_CMD_BATCH | (batch << 8) | (size << 16); - - for (spu = 0; spu < cell->num_spus; spu++) { - assert(cell->buffer_status[spu][batch][0] == CELL_BUFFER_STATUS_USED); - send_mbox_message(cell_global.spe_contexts[spu], cmd_word); - } - - /* When the SPUs are done copying the buffer into their locals stores - * they'll write a BUFFER_STATUS_FREE message into the buffer_status[] - * array indicating that the PPU can re-use the buffer. - */ - - batch = cell_get_empty_buffer(cell); - - cell->buffer_size[batch] = 0; /* empty */ - cell->cur_batch = batch; - - flushing = FALSE; -} - - -uint -cell_batch_free_space(const struct cell_context *cell) -{ - uint free = CELL_BUFFER_SIZE - cell->buffer_size[cell->cur_batch]; - return free; -} - - -/** - * Append data to current batch. - */ -void -cell_batch_append(struct cell_context *cell, const void *data, uint bytes) -{ - uint size; - - ASSERT(bytes % 8 == 0); - ASSERT(bytes <= CELL_BUFFER_SIZE); - ASSERT(cell->cur_batch >= 0); - -#ifdef ASSERT - { - uint spu; - for (spu = 0; spu < cell->num_spus; spu++) { - ASSERT(cell->buffer_status[spu][cell->cur_batch][0] - == CELL_BUFFER_STATUS_USED); - } - } -#endif - - size = cell->buffer_size[cell->cur_batch]; - - if (size + bytes > CELL_BUFFER_SIZE) { - cell_batch_flush(cell); - size = 0; - } - - ASSERT(size + bytes <= CELL_BUFFER_SIZE); - - memcpy(cell->buffer[cell->cur_batch] + size, data, bytes); - - cell->buffer_size[cell->cur_batch] = size + bytes; -} - - -void * -cell_batch_alloc(struct cell_context *cell, uint bytes) -{ - return cell_batch_alloc_aligned(cell, bytes, 1); -} - - -void * -cell_batch_alloc_aligned(struct cell_context *cell, uint bytes, - uint alignment) -{ - void *pos; - uint size, padbytes; - - ASSERT(bytes % 8 == 0); - ASSERT(bytes <= CELL_BUFFER_SIZE); - ASSERT(alignment > 0); - ASSERT(cell->cur_batch >= 0); - -#ifdef ASSERT - { - uint spu; - for (spu = 0; spu < cell->num_spus; spu++) { - ASSERT(cell->buffer_status[spu][cell->cur_batch][0] - == CELL_BUFFER_STATUS_USED); - } - } -#endif - - size = cell->buffer_size[cell->cur_batch]; - - padbytes = (alignment - (size % alignment)) % alignment; - - if (padbytes + size + bytes > CELL_BUFFER_SIZE) { - cell_batch_flush(cell); - size = 0; - } - else { - size += padbytes; - } - - ASSERT(size % alignment == 0); - ASSERT(size + bytes <= CELL_BUFFER_SIZE); - - pos = (void *) (cell->buffer[cell->cur_batch] + size); - - cell->buffer_size[cell->cur_batch] = size + bytes; - - return pos; -} diff --git a/src/mesa/pipe/cell/ppu/cell_batch.h b/src/mesa/pipe/cell/ppu/cell_batch.h deleted file mode 100644 index a6eee0a8b1..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_batch.h +++ /dev/null @@ -1,58 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef CELL_BATCH_H -#define CELL_BATCH_H - -#include "pipe/p_compiler.h" - - -struct cell_context; - - -extern uint -cell_get_empty_buffer(struct cell_context *cell); - -extern void -cell_batch_flush(struct cell_context *cell); - -extern uint -cell_batch_free_space(const struct cell_context *cell); - -extern void -cell_batch_append(struct cell_context *cell, const void *data, uint bytes); - -extern void * -cell_batch_alloc(struct cell_context *cell, uint bytes); - -extern void * -cell_batch_alloc_aligned(struct cell_context *cell, uint bytes, - uint alignment); - - -#endif /* CELL_BATCH_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_clear.c b/src/mesa/pipe/cell/ppu/cell_clear.c deleted file mode 100644 index 07b908eec5..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_clear.c +++ /dev/null @@ -1,76 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Authors - * Brian Paul - */ - -#include -#include -#include -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "pipe/cell/common.h" -#include "cell_clear.h" -#include "cell_context.h" -#include "cell_batch.h" -#include "cell_flush.h" -#include "cell_spu.h" - - -void -cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue) -{ - struct cell_context *cell = cell_context(pipe); - uint surfIndex; - - if (cell->dirty) - cell_update_derived(cell); - - - if (!cell->cbuf_map[0]) - cell->cbuf_map[0] = pipe_surface_map(ps); - - if (ps == cell->framebuffer.zsbuf) { - surfIndex = 1; - } - else { - surfIndex = 0; - } - - - { - struct cell_command_clear_surface *clr - = (struct cell_command_clear_surface *) - cell_batch_alloc(cell, sizeof(*clr)); - clr->opcode = CELL_CMD_CLEAR_SURFACE; - clr->surface = surfIndex; - clr->value = clearValue; - } -} diff --git a/src/mesa/pipe/cell/ppu/cell_clear.h b/src/mesa/pipe/cell/ppu/cell_clear.h deleted file mode 100644 index ff47d43f4c..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_clear.h +++ /dev/null @@ -1,43 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef CELL_CLEAR_H -#define CELL_CLEAR_H - - -struct pipe_context; -struct pipe_surface; - - -extern void -cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue); - - - -#endif /* CELL_CLEAR_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_context.c b/src/mesa/pipe/cell/ppu/cell_context.c deleted file mode 100644 index bbe1fd7a11..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_context.c +++ /dev/null @@ -1,287 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Authors - * Brian Paul - */ - - -#include - -#include "pipe/p_defines.h" -#include "pipe/p_format.h" -#include "pipe/p_util.h" -#include "pipe/p_winsys.h" -#include "pipe/cell/common.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" -#include "cell_clear.h" -#include "cell_context.h" -#include "cell_draw_arrays.h" -#include "cell_flush.h" -#include "cell_render.h" -#include "cell_state.h" -#include "cell_surface.h" -#include "cell_spu.h" -#include "cell_texture.h" -#include "cell_vbuf.h" - - - -static boolean -cell_is_format_supported( struct pipe_context *pipe, - enum pipe_format format, uint type ) -{ - /*struct cell_context *cell = cell_context( pipe );*/ - - switch (type) { - case PIPE_TEXTURE: - /* cell supports all texture formats, XXX for now anyway */ - return TRUE; - case PIPE_SURFACE: - /* cell supports all (off-screen) surface formats, XXX for now */ - return TRUE; - default: - assert(0); - return FALSE; - } -} - - -static int cell_get_param(struct pipe_context *pipe, int param) -{ - switch (param) { - case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; - case PIPE_CAP_NPOT_TEXTURES: - return 1; - case PIPE_CAP_TWO_SIDED_STENCIL: - return 1; - case PIPE_CAP_GLSL: - return 1; - case PIPE_CAP_S3TC: - return 0; - case PIPE_CAP_ANISOTROPIC_FILTER: - return 0; - case PIPE_CAP_POINT_SPRITE: - return 1; - case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; - case PIPE_CAP_OCCLUSION_QUERY: - return 1; - case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 1; - case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 12; /* max 2Kx2K */ - case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: - return 8; /* max 128x128x128 */ - case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 12; /* max 2Kx2K */ - default: - return 0; - } -} - -static float cell_get_paramf(struct pipe_context *pipe, int param) -{ - switch (param) { - case PIPE_CAP_MAX_LINE_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_LINE_WIDTH_AA: - return 255.0; /* arbitrary */ - - case PIPE_CAP_MAX_POINT_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_POINT_WIDTH_AA: - return 255.0; /* arbitrary */ - - case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 0.0; - - case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: - return 16.0; /* arbitrary */ - - default: - return 0; - } -} - - -static const char * -cell_get_name( struct pipe_context *pipe ) -{ - return "Cell"; -} - -static const char * -cell_get_vendor( struct pipe_context *pipe ) -{ - return "Tungsten Graphics, Inc."; -} - - - -static void -cell_destroy_context( struct pipe_context *pipe ) -{ - struct cell_context *cell = cell_context(pipe); - - cell_spu_exit(cell); - - align_free(cell); -} - - -static struct draw_context * -cell_draw_create(struct cell_context *cell) -{ - struct draw_context *draw = draw_create(); - - if (getenv("GALLIUM_CELL_VS")) { - /* plug in SPU-based vertex transformation code */ - draw->shader_queue_flush = cell_vertex_shader_queue_flush; - draw->driver_private = cell; - } - - return draw; -} - - -struct pipe_context * -cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) -{ - struct cell_context *cell; - uint spu, buf; - - /* some fields need to be 16-byte aligned, so align the whole object */ - cell = (struct cell_context*) align_malloc(sizeof(struct cell_context), 16); - if (!cell) - return NULL; - - memset(cell, 0, sizeof(*cell)); - - cell->winsys = cws; - cell->pipe.winsys = winsys; - cell->pipe.destroy = cell_destroy_context; - - /* queries */ - cell->pipe.is_format_supported = cell_is_format_supported; - cell->pipe.get_name = cell_get_name; - cell->pipe.get_vendor = cell_get_vendor; - cell->pipe.get_param = cell_get_param; - cell->pipe.get_paramf = cell_get_paramf; - - - /* state setters */ - cell->pipe.create_blend_state = cell_create_blend_state; - cell->pipe.bind_blend_state = cell_bind_blend_state; - cell->pipe.delete_blend_state = cell_delete_blend_state; - - cell->pipe.create_sampler_state = cell_create_sampler_state; - cell->pipe.bind_sampler_state = cell_bind_sampler_state; - cell->pipe.delete_sampler_state = cell_delete_sampler_state; - - cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state; - cell->pipe.bind_depth_stencil_alpha_state = cell_bind_depth_stencil_alpha_state; - cell->pipe.delete_depth_stencil_alpha_state = cell_delete_depth_stencil_alpha_state; - - cell->pipe.create_rasterizer_state = cell_create_rasterizer_state; - cell->pipe.bind_rasterizer_state = cell_bind_rasterizer_state; - cell->pipe.delete_rasterizer_state = cell_delete_rasterizer_state; - - cell->pipe.create_fs_state = cell_create_fs_state; - cell->pipe.bind_fs_state = cell_bind_fs_state; - cell->pipe.delete_fs_state = cell_delete_fs_state; - - cell->pipe.create_vs_state = cell_create_vs_state; - cell->pipe.bind_vs_state = cell_bind_vs_state; - cell->pipe.delete_vs_state = cell_delete_vs_state; - - cell->pipe.set_blend_color = cell_set_blend_color; - cell->pipe.set_clip_state = cell_set_clip_state; - cell->pipe.set_constant_buffer = cell_set_constant_buffer; - - cell->pipe.set_framebuffer_state = cell_set_framebuffer_state; - - cell->pipe.set_polygon_stipple = cell_set_polygon_stipple; - cell->pipe.set_scissor_state = cell_set_scissor_state; - cell->pipe.set_viewport_state = cell_set_viewport_state; - - cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; - cell->pipe.set_vertex_element = cell_set_vertex_element; - - cell->pipe.draw_arrays = cell_draw_arrays; - cell->pipe.draw_elements = cell_draw_elements; - - cell->pipe.clear = cell_clear_surface; - cell->pipe.flush = cell_flush; - - /* textures */ - cell->pipe.texture_create = cell_texture_create; - cell->pipe.texture_release = cell_texture_release; - cell->pipe.get_tex_surface = cell_get_tex_surface; - - cell->pipe.set_sampler_texture = cell_set_sampler_texture; - -#if 0 - cell->pipe.begin_query = cell_begin_query; - cell->pipe.end_query = cell_end_query; - cell->pipe.wait_query = cell_wait_query; -#endif - - cell_init_surface_functions(cell); - - cell->draw = cell_draw_create(cell); - - cell_init_vbuf(cell); - draw_set_rasterize_stage(cell->draw, cell->vbuf); - - /* - * SPU stuff - */ - cell->num_spus = 6; - - cell_start_spus(cell); - - /* init command, vertex/index buffer info */ - for (buf = 0; buf < CELL_NUM_BUFFERS; buf++) { - cell->buffer_size[buf] = 0; - - /* init batch buffer status values, - * mark 0th buffer as used, rest as free. - */ - for (spu = 0; spu < cell->num_spus; spu++) { - if (buf == 0) - cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_USED; - else - cell->buffer_status[spu][buf][0] = CELL_BUFFER_STATUS_FREE; - } - } - - return &cell->pipe; -} diff --git a/src/mesa/pipe/cell/ppu/cell_context.h b/src/mesa/pipe/cell/ppu/cell_context.h deleted file mode 100644 index 3b63419b5e..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_context.h +++ /dev/null @@ -1,135 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef CELL_CONTEXT_H -#define CELL_CONTEXT_H - - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/draw/draw_vertex.h" -#include "pipe/draw/draw_vbuf.h" -#include "cell_winsys.h" -#include "pipe/cell/common.h" - - -struct cell_vbuf_render; - -struct cell_vertex_shader_state -{ - struct pipe_shader_state shader; - void *draw_data; -}; - - -struct cell_fragment_shader_state -{ - struct pipe_shader_state shader; - void *data; -}; - - -struct cell_context -{ - struct pipe_context pipe; - - struct cell_winsys *winsys; - - const struct pipe_blend_state *blend; - const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct pipe_depth_stencil_alpha_state *depth_stencil; - const struct pipe_rasterizer_state *rasterizer; - const struct cell_vertex_shader_state *vs; - const struct cell_fragment_shader_state *fs; - - struct pipe_blend_color blend_color; - struct pipe_clip_state clip; - struct pipe_constant_buffer constants[2]; - struct pipe_framebuffer_state framebuffer; - struct pipe_poly_stipple poly_stipple; - struct pipe_scissor_state scissor; - struct cell_texture *texture[PIPE_MAX_SAMPLERS]; - struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - - ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS]; - ubyte *zsbuf_map; - - struct pipe_surface *tex_surf; - uint *tex_map; - - uint dirty; - - /** The primitive drawing context */ - struct draw_context *draw; - struct draw_stage *render_stage; - - /** For post-transformed vertex buffering: */ - struct cell_vbuf_render *vbuf_render; - struct draw_stage *vbuf; - - struct vertex_info vertex_info; - - /** Mapped constant buffers */ - void *mapped_constants[PIPE_SHADER_TYPES]; - - - uint num_spus; - - /** Buffers for command batches, vertex/index data */ - uint buffer_size[CELL_NUM_BUFFERS]; - ubyte buffer[CELL_NUM_BUFFERS][CELL_BUFFER_SIZE] ALIGN16_ATTRIB; - - int cur_batch; /**< which buffer is being filled w/ commands */ - - /** [4] to ensure 16-byte alignment for each status word */ - uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BUFFERS][4] ALIGN16_ATTRIB; - -}; - - - - -static INLINE struct cell_context * -cell_context(struct pipe_context *pipe) -{ - return (struct cell_context *) pipe; -} - - -extern struct pipe_context * -cell_create_context(struct pipe_winsys *ws, struct cell_winsys *cws); - -extern void -cell_vertex_shader_queue_flush(struct draw_context *draw); - - - - -#endif /* CELL_CONTEXT_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c b/src/mesa/pipe/cell/ppu/cell_draw_arrays.c deleted file mode 100644 index 717cd8370f..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_draw_arrays.c +++ /dev/null @@ -1,164 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Brian Paul - * Keith Whitwell - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_context.h" -#include "pipe/p_winsys.h" - -#include "cell_context.h" -#include "cell_draw_arrays.h" -#include "cell_state.h" - -#include "pipe/draw/draw_context.h" - - - -static void -cell_map_constant_buffers(struct cell_context *sp) -{ - struct pipe_winsys *ws = sp->pipe.winsys; - uint i; - for (i = 0; i < 2; i++) { - if (sp->constants[i].size) - sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - } - - draw_set_mapped_constant_buffer(sp->draw, - sp->mapped_constants[PIPE_SHADER_VERTEX]); -} - -static void -cell_unmap_constant_buffers(struct cell_context *sp) -{ - struct pipe_winsys *ws = sp->pipe.winsys; - uint i; - for (i = 0; i < 2; i++) { - if (sp->constants[i].size) - ws->buffer_unmap(ws, sp->constants[i].buffer); - sp->mapped_constants[i] = NULL; - } -} - - -boolean -cell_draw_arrays(struct pipe_context *pipe, unsigned mode, - unsigned start, unsigned count) -{ - return cell_draw_elements(pipe, NULL, 0, mode, start, count); -} - - - -/** - * Draw vertex arrays, with optional indexing. - * Basically, map the vertex buffers (and drawing surfaces), then hand off - * the drawing to the 'draw' module. - * - * XXX should the element buffer be specified/bound with a separate function? - */ -boolean -cell_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count) -{ - struct cell_context *sp = cell_context(pipe); - struct draw_context *draw = sp->draw; - unsigned i; - - /* first, check that the primitive is not malformed. It is the - * state tracker's responsibility to do send only correctly formed - * primitives down. It currently isn't doing that though... - */ -#if 1 - count = draw_trim_prim( mode, count ); -#else - if (!draw_validate_prim( mode, count )) - assert(0); -#endif - - if (sp->dirty) - cell_update_derived( sp ); - -#if 0 - cell_map_surfaces(sp); -#endif - cell_map_constant_buffers(sp); - - /* - * Map vertex buffers - */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { - if (sp->vertex_buffer[i].buffer) { - void *buf = pipe->winsys->buffer_map(pipe->winsys, - sp->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_vertex_buffer(draw, i, buf); - } - } - /* Map index buffer, if present */ - if (indexBuffer) { - void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, - indexBuffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); - } - else { - /* no index/element buffer */ - draw_set_mapped_element_buffer(draw, 0, NULL); - } - - - /* draw! */ - draw_arrays(draw, mode, start, count); - - /* - * unmap vertex/index buffers - will cause draw module to flush - */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { - if (sp->vertex_buffer[i].buffer) { - draw_set_mapped_vertex_buffer(draw, i, NULL); - pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); - } - } - if (indexBuffer) { - draw_set_mapped_element_buffer(draw, 0, NULL); - pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); - } - - /* Note: leave drawing surfaces mapped */ - cell_unmap_constant_buffers(sp); - - return TRUE; -} diff --git a/src/mesa/pipe/cell/ppu/cell_draw_arrays.h b/src/mesa/pipe/cell/ppu/cell_draw_arrays.h deleted file mode 100644 index d5df4aa05f..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_draw_arrays.h +++ /dev/null @@ -1,42 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef CELL_DRAW_ARRAYS_H -#define CELL_DRAW_ARRAYS_H - - -boolean cell_draw_arrays(struct pipe_context *pipe, unsigned mode, - unsigned start, unsigned count); - -boolean cell_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count); - - - -#endif /* CELL_DRAW_ARRAYS_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_flush.c b/src/mesa/pipe/cell/ppu/cell_flush.c deleted file mode 100644 index f62bc4650c..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_flush.c +++ /dev/null @@ -1,84 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "cell_context.h" -#include "cell_batch.h" -#include "cell_flush.h" -#include "cell_spu.h" -#include "cell_render.h" -#include "pipe/draw/draw_context.h" - - -void -cell_flush(struct pipe_context *pipe, unsigned flags) -{ - struct cell_context *cell = cell_context(pipe); - - if (flags & PIPE_FLUSH_SWAPBUFFERS) - flags |= PIPE_FLUSH_WAIT; - - draw_flush( cell->draw ); - cell_flush_int(pipe, flags); -} - - -/** internal flush */ -void -cell_flush_int(struct pipe_context *pipe, unsigned flags) -{ - static boolean flushing = FALSE; /* recursion catcher */ - struct cell_context *cell = cell_context(pipe); - uint i; - - ASSERT(!flushing); - flushing = TRUE; - - if (flags & PIPE_FLUSH_WAIT) { - uint64_t *cmd = (uint64_t *) cell_batch_alloc(cell, sizeof(uint64_t)); - *cmd = CELL_CMD_FINISH; - } - - cell_batch_flush(cell); - -#if 0 - /* Send CMD_FINISH to all SPUs */ - for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_FINISH); - } -#endif - - if (flags & PIPE_FLUSH_WAIT) { - /* Wait for ack */ - for (i = 0; i < cell->num_spus; i++) { - uint k = wait_mbox_message(cell_global.spe_contexts[i]); - assert(k == CELL_CMD_FINISH); - } - } - - flushing = FALSE; -} diff --git a/src/mesa/pipe/cell/ppu/cell_flush.h b/src/mesa/pipe/cell/ppu/cell_flush.h deleted file mode 100644 index eda351b1cb..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_flush.h +++ /dev/null @@ -1,38 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef CELL_FLUSH -#define CELL_FLUSH - -extern void -cell_flush(struct pipe_context *pipe, unsigned flags); - -extern void -cell_flush_int(struct pipe_context *pipe, unsigned flags); - -#endif diff --git a/src/mesa/pipe/cell/ppu/cell_render.c b/src/mesa/pipe/cell/ppu/cell_render.c deleted file mode 100644 index 4ab277a4b2..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_render.c +++ /dev/null @@ -1,210 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Last stage of 'draw' pipeline: send tris to SPUs. - * \author Brian Paul - */ - -#include "cell_context.h" -#include "cell_render.h" -#include "cell_spu.h" -#include "pipe/p_util.h" -#include "pipe/draw/draw_private.h" - - -struct render_stage { - struct draw_stage stage; /**< This must be first (base class) */ - - struct cell_context *cell; -}; - - -static INLINE struct render_stage * -render_stage(struct draw_stage *stage) -{ - return (struct render_stage *) stage; -} - - -static void render_begin( struct draw_stage *stage ) -{ -#if 0 - struct render_stage *render = render_stage(stage); - struct cell_context *sp = render->cell; - const struct pipe_shader_state *fs = &render->cell->fs->shader; - render->quad.nr_attrs = render->cell->nr_frag_attrs; - - render->firstFpInput = fs->input_semantic_name[0]; - - sp->quad.first->begin(sp->quad.first); -#endif -} - - -static void render_end( struct draw_stage *stage ) -{ -} - - -static void reset_stipple_counter( struct draw_stage *stage ) -{ - struct render_stage *render = render_stage(stage); - /*render->cell->line_stipple_counter = 0;*/ -} - - -static void -render_point(struct draw_stage *stage, struct prim_header *prim) -{ -} - - -static void -render_line(struct draw_stage *stage, struct prim_header *prim) -{ -} - - -/** Write a vertex into the prim buffer */ -static void -save_vertex(struct cell_prim_buffer *buf, uint pos, - const struct vertex_header *vert) -{ - uint attr, j; - - for (attr = 0; attr < 2; attr++) { - for (j = 0; j < 4; j++) { - buf->vertex[pos][attr][j] = vert->data[attr][j]; - } - } - - /* update bounding box */ - if (vert->data[0][0] < buf->xmin) - buf->xmin = vert->data[0][0]; - if (vert->data[0][0] > buf->xmax) - buf->xmax = vert->data[0][0]; - if (vert->data[0][1] < buf->ymin) - buf->ymin = vert->data[0][1]; - if (vert->data[0][1] > buf->ymax) - buf->ymax = vert->data[0][1]; -} - - -static void -render_tri(struct draw_stage *stage, struct prim_header *prim) -{ - struct render_stage *rs = render_stage(stage); - struct cell_context *cell = rs->cell; - struct cell_prim_buffer *buf = &cell->prim_buffer; - uint i; - - if (buf->num_verts + 3 > CELL_MAX_VERTS) { - cell_flush_prim_buffer(cell); - } - - i = buf->num_verts; - assert(i+2 <= CELL_MAX_VERTS); - save_vertex(buf, i+0, prim->v[0]); - save_vertex(buf, i+1, prim->v[1]); - save_vertex(buf, i+2, prim->v[2]); - buf->num_verts += 3; -} - - -/** - * Send the a RENDER command to all SPUs to have them render the prims - * in the current prim_buffer. - */ -void -cell_flush_prim_buffer(struct cell_context *cell) -{ - uint i; - - if (cell->prim_buffer.num_verts == 0) - return; - - for (i = 0; i < cell->num_spus; i++) { - struct cell_command_render *render = &cell_global.command[i].render; - render->prim_type = PIPE_PRIM_TRIANGLES; - render->num_verts = cell->prim_buffer.num_verts; - render->vertex_size = cell->vertex_info->size * 4; - render->xmin = cell->prim_buffer.xmin; - render->ymin = cell->prim_buffer.ymin; - render->xmax = cell->prim_buffer.xmax; - render->ymax = cell->prim_buffer.ymax; - render->vertex_data = &cell->prim_buffer.vertex; - ASSERT_ALIGN16(render->vertex_data); - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_RENDER); - } - - cell->prim_buffer.num_verts = 0; - - cell->prim_buffer.xmin = 1e100; - cell->prim_buffer.ymin = 1e100; - cell->prim_buffer.xmax = -1e100; - cell->prim_buffer.ymax = -1e100; - - /* XXX temporary, need to double-buffer the prim buffer until we get - * a real command buffer/list system. - */ - cell_flush(&cell->pipe, 0x0); -} - - - -static void render_destroy( struct draw_stage *stage ) -{ - FREE( stage ); -} - - -/** - * Create a new draw/render stage. This will be plugged into the - * draw module as the last pipeline stage. - */ -struct draw_stage *cell_draw_render_stage( struct cell_context *cell ) -{ - struct render_stage *render = CALLOC_STRUCT(render_stage); - - render->cell = cell; - render->stage.draw = cell->draw; - render->stage.begin = render_begin; - render->stage.point = render_point; - render->stage.line = render_line; - render->stage.tri = render_tri; - render->stage.end = render_end; - render->stage.reset_stipple_counter = reset_stipple_counter; - render->stage.destroy = render_destroy; - - /* - render->quad.coef = render->coef; - render->quad.posCoef = &render->posCoef; - */ - - return &render->stage; -} diff --git a/src/mesa/pipe/cell/ppu/cell_render.h b/src/mesa/pipe/cell/ppu/cell_render.h deleted file mode 100644 index 826dcbafeb..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_render.h +++ /dev/null @@ -1,39 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef CELL_RENDER_H -#define CELL_RENDER_H - -struct cell_context; -struct draw_stage; - -extern void -cell_flush_prim_buffer(struct cell_context *cell); - -extern struct draw_stage *cell_draw_render_stage( struct cell_context *cell ); - -#endif /* CELL_RENDER_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_spu.c b/src/mesa/pipe/cell/ppu/cell_spu.c deleted file mode 100644 index 7c83a47e57..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_spu.c +++ /dev/null @@ -1,155 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include - -#include "cell_spu.h" -#include "pipe/p_format.h" -#include "pipe/p_state.h" -#include "pipe/cell/common.h" - - -/* -helpful headers: -/opt/ibm/cell-sdk/prototype/src/include/ppu/cbe_mfc.h -*/ - - -struct cell_global_info cell_global; - - -/** - * Write a 1-word message to the given SPE mailbox. - */ -void -send_mbox_message(spe_context_ptr_t ctx, unsigned int msg) -{ - spe_in_mbox_write(ctx, &msg, 1, SPE_MBOX_ALL_BLOCKING); -} - - -/** - * Wait for a 1-word message to arrive in given mailbox. - */ -uint -wait_mbox_message(spe_context_ptr_t ctx) -{ - do { - unsigned data; - int count = spe_out_mbox_read(ctx, &data, 1); - - if (count == 1) { - return data; - } - - if (count < 0) { - /* error */ ; - } - } while (1); -} - - -static void *cell_thread_function(void *arg) -{ - struct cell_init_info *init = (struct cell_init_info *) arg; - unsigned entry = SPE_DEFAULT_ENTRY; - - ASSERT_ALIGN16(init); - - if (spe_context_run(cell_global.spe_contexts[init->id], &entry, 0, - init, NULL, NULL) < 0) { - fprintf(stderr, "spe_context_run() failed\n"); - exit(1); - } - - pthread_exit(NULL); -} - - -/** - * Create the SPU threads - */ -void -cell_start_spus(struct cell_context *cell) -{ - uint i, j; - - assert(cell->num_spus <= MAX_SPUS); - - ASSERT_ALIGN16(&cell_global.command[0]); - ASSERT_ALIGN16(&cell_global.command[1]); - - ASSERT_ALIGN16(&cell_global.inits[0]); - ASSERT_ALIGN16(&cell_global.inits[1]); - - for (i = 0; i < cell->num_spus; i++) { - cell_global.inits[i].id = i; - cell_global.inits[i].num_spus = cell->num_spus; - cell_global.inits[i].cmd = &cell_global.command[i]; - for (j = 0; j < CELL_NUM_BUFFERS; j++) { - cell_global.inits[i].buffers[j] = cell->buffer[j]; - } - cell_global.inits[i].buffer_status = &cell->buffer_status[0][0][0]; - - cell_global.spe_contexts[i] = spe_context_create(0, NULL); - if (!cell_global.spe_contexts[i]) { - fprintf(stderr, "spe_context_create() failed\n"); - exit(1); - } - - if (spe_program_load(cell_global.spe_contexts[i], &g3d_spu)) { - fprintf(stderr, "spe_program_load() failed\n"); - exit(1); - } - - pthread_create(&cell_global.spe_threads[i], NULL, &cell_thread_function, - &cell_global.inits[i]); - } -} - - -/** - * Tell all the SPUs to stop/exit. - */ -void -cell_spu_exit(struct cell_context *cell) -{ - uint i; - - for (i = 0; i < cell->num_spus; i++) { - send_mbox_message(cell_global.spe_contexts[i], CELL_CMD_EXIT); - } - - /* wait for threads to exit */ - for (i = 0; i < cell->num_spus; i++) { - void *value; - pthread_join(cell_global.spe_threads[i], &value); - cell_global.spe_threads[i] = 0; - cell_global.spe_contexts[i] = 0; - } -} diff --git a/src/mesa/pipe/cell/ppu/cell_spu.h b/src/mesa/pipe/cell/ppu/cell_spu.h deleted file mode 100644 index 19eff94f96..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_spu.h +++ /dev/null @@ -1,82 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef CELL_SPU -#define CELL_SPU - - -#include -#include -#include "pipe/cell/common.h" - -#include "cell_context.h" - - -#define MAX_SPUS 8 - -/** - * Global vars, for now anyway. - */ -struct cell_global_info -{ - /** - * SPU/SPE handles, etc - */ - spe_context_ptr_t spe_contexts[MAX_SPUS]; - pthread_t spe_threads[MAX_SPUS]; - - /** - * Data sent to SPUs - */ - struct cell_init_info inits[MAX_SPUS]; - struct cell_command command[MAX_SPUS]; -}; - - -extern struct cell_global_info cell_global; - - -/** This is the handle for the actual SPE code */ -extern spe_program_handle_t g3d_spu; - - -extern void -send_mbox_message(spe_context_ptr_t ctx, unsigned int msg); - -extern uint -wait_mbox_message(spe_context_ptr_t ctx); - - -extern void -cell_start_spus(struct cell_context *cell); - - -extern void -cell_spu_exit(struct cell_context *cell); - - -#endif /* CELL_SPU */ diff --git a/src/mesa/pipe/cell/ppu/cell_state.h b/src/mesa/pipe/cell/ppu/cell_state.h deleted file mode 100644 index 3a71ba14fa..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state.h +++ /dev/null @@ -1,115 +0,0 @@ - - -#ifndef CELL_STATE_H -#define CELL_STATE_H - - -#define CELL_NEW_VIEWPORT 0x1 -#define CELL_NEW_RASTERIZER 0x2 -#define CELL_NEW_FS 0x4 -#define CELL_NEW_BLEND 0x8 -#define CELL_NEW_CLIP 0x10 -#define CELL_NEW_SCISSOR 0x20 -#define CELL_NEW_STIPPLE 0x40 -#define CELL_NEW_FRAMEBUFFER 0x80 -#define CELL_NEW_ALPHA_TEST 0x100 -#define CELL_NEW_DEPTH_STENCIL 0x200 -#define CELL_NEW_SAMPLER 0x400 -#define CELL_NEW_TEXTURE 0x800 -#define CELL_NEW_VERTEX 0x1000 -#define CELL_NEW_VS 0x2000 -#define CELL_NEW_CONSTANTS 0x4000 -#define CELL_NEW_VERTEX_INFO 0x8000 - - - -extern void -cell_set_framebuffer_state( struct pipe_context *, - const struct pipe_framebuffer_state * ); - - - -extern void * -cell_create_blend_state(struct pipe_context *, const struct pipe_blend_state *); -extern void cell_bind_blend_state(struct pipe_context *, void *); -extern void cell_delete_blend_state(struct pipe_context *, void *); - -extern void cell_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ); - - -void * -cell_create_sampler_state(struct pipe_context *, - const struct pipe_sampler_state *); - -extern void -cell_bind_sampler_state(struct pipe_context *, unsigned, void *); - -extern void -cell_delete_sampler_state(struct pipe_context *, void *); - - -extern void * -cell_create_depth_stencil_alpha_state(struct pipe_context *, - const struct pipe_depth_stencil_alpha_state *); - -extern void -cell_bind_depth_stencil_alpha_state(struct pipe_context *, void *); - -extern void -cell_delete_depth_stencil_alpha_state(struct pipe_context *, void *); - - -void *cell_create_fs_state(struct pipe_context *, - const struct pipe_shader_state *); -void cell_bind_fs_state(struct pipe_context *, void *); -void cell_delete_fs_state(struct pipe_context *, void *); -void *cell_create_vs_state(struct pipe_context *, - const struct pipe_shader_state *); -void cell_bind_vs_state(struct pipe_context *, void *); -void cell_delete_vs_state(struct pipe_context *, void *); - - -void * -cell_create_rasterizer_state(struct pipe_context *, - const struct pipe_rasterizer_state *); -void cell_bind_rasterizer_state(struct pipe_context *, void *); -void cell_delete_rasterizer_state(struct pipe_context *, void *); - - -void cell_set_clip_state( struct pipe_context *, - const struct pipe_clip_state * ); - -void cell_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf); - -void cell_set_polygon_stipple( struct pipe_context *, - const struct pipe_poly_stipple * ); - -void -cell_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, - struct pipe_texture *texture); - -void cell_set_scissor_state( struct pipe_context *, - const struct pipe_scissor_state * ); - -void cell_set_texture_state( struct pipe_context *, - unsigned unit, struct pipe_texture * ); - -void cell_set_vertex_element(struct pipe_context *, - unsigned index, - const struct pipe_vertex_element *); - -void cell_set_vertex_buffer(struct pipe_context *, - unsigned index, - const struct pipe_vertex_buffer *); - -void cell_set_viewport_state( struct pipe_context *, - const struct pipe_viewport_state * ); - - -void cell_update_derived( struct cell_context *softpipe ); - -#endif diff --git a/src/mesa/pipe/cell/ppu/cell_state_blend.c b/src/mesa/pipe/cell/ppu/cell_state_blend.c deleted file mode 100644 index 4fc60548c8..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_blend.c +++ /dev/null @@ -1,109 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/draw/draw_context.h" -#include "cell_context.h" -#include "cell_state.h" - - - -void * -cell_create_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend) -{ - return mem_dup(blend, sizeof(*blend)); -} - - -void -cell_bind_blend_state(struct pipe_context *pipe, void *blend) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - cell->blend = (const struct pipe_blend_state *)blend; - - cell->dirty |= CELL_NEW_BLEND; -} - - -void -cell_delete_blend_state(struct pipe_context *pipe, void *blend) -{ - FREE(blend); -} - - -void -cell_set_blend_color(struct pipe_context *pipe, - const struct pipe_blend_color *blend_color) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - cell->blend_color = *blend_color; - - cell->dirty |= CELL_NEW_BLEND; -} - - - - -void * -cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) -{ - return mem_dup(depth_stencil, sizeof(*depth_stencil)); -} - - -void -cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, - void *depth_stencil) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - cell->depth_stencil - = (const struct pipe_depth_stencil_alpha_state *) depth_stencil; - - cell->dirty |= CELL_NEW_DEPTH_STENCIL; -} - - -void -cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth) -{ - FREE(depth); -} diff --git a/src/mesa/pipe/cell/ppu/cell_state_clip.c b/src/mesa/pipe/cell/ppu/cell_state_clip.c deleted file mode 100644 index 4f43665941..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_clip.c +++ /dev/null @@ -1,84 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "cell_context.h" -#include "cell_state.h" -#include "pipe/draw/draw_context.h" - - -void cell_set_clip_state( struct pipe_context *pipe, - const struct pipe_clip_state *clip ) -{ - struct cell_context *cell = cell_context(pipe); - - /* pass the clip state to the draw module */ - draw_set_clip_state(cell->draw, clip); -} - - - -/* Called when driver state tracker notices changes to the viewport - * matrix: - */ -void cell_set_viewport_state( struct pipe_context *pipe, - const struct pipe_viewport_state *viewport ) -{ - struct cell_context *cell = cell_context(pipe); - - cell->viewport = *viewport; /* struct copy */ - cell->dirty |= CELL_NEW_VIEWPORT; - - /* pass the viewport info to the draw module */ - draw_set_viewport_state(cell->draw, viewport); - - /* Using tnl/ and vf/ modules is temporary while getting started. - * Full pipe will have vertex shader, vertex fetch of its own. - */ -} - - -void cell_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct cell_context *cell = cell_context(pipe); - - memcpy( &cell->scissor, scissor, sizeof(*scissor) ); - cell->dirty |= CELL_NEW_SCISSOR; -} - - -void cell_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ - struct cell_context *cell = cell_context(pipe); - - memcpy( &cell->poly_stipple, stipple, sizeof(*stipple) ); - cell->dirty |= CELL_NEW_STIPPLE; -} diff --git a/src/mesa/pipe/cell/ppu/cell_state_derived.c b/src/mesa/pipe/cell/ppu/cell_state_derived.c deleted file mode 100644 index 56daf5dfde..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_derived.c +++ /dev/null @@ -1,192 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_vertex.h" -#include "cell_context.h" -#include "cell_batch.h" -#include "cell_state.h" -#include "cell_state_emit.h" - - -static int -find_vs_output(const struct pipe_shader_state *vs, - uint semantic_name, - uint semantic_index) -{ - uint i; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == semantic_name && - vs->output_semantic_index[i] == semantic_index) - return i; - } - return -1; -} - - -/** - * Determine how to map vertex program outputs to fragment program inputs. - * Basically, this will be used when computing the triangle interpolation - * coefficients from the post-transform vertex attributes. - */ -static void -calculate_vertex_layout( struct cell_context *cell ) -{ - const struct pipe_shader_state *vs = &cell->vs->shader; - const struct pipe_shader_state *fs = &cell->fs->shader; - const enum interp_mode colorInterp - = cell->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; - struct vertex_info *vinfo = &cell->vertex_info; - uint i; - int src; - -#if 0 - if (cell->vbuf) { - /* if using the post-transform vertex buffer, tell draw_vbuf to - * simply emit the whole post-xform vertex as-is: - */ - struct vertex_info *vinfo_vbuf = &cell->vertex_info_vbuf; - vinfo_vbuf->num_attribs = 0; - draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); - vinfo_vbuf->size = 4 * vs->num_outputs + sizeof(struct vertex_header)/4; - } -#endif - - /* reset vinfo */ - vinfo->num_attribs = 0; - - /* we always want to emit vertex pos */ - src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); - assert(src >= 0); - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); - - - /* - * Loop over fragment shader inputs, searching for the matching output - * from the vertex shader. - */ - for (i = 0; i < fs->num_inputs; i++) { - switch (fs->input_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - /* already done above */ - break; - - case TGSI_SEMANTIC_COLOR: - src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, - fs->input_semantic_index[i]); - assert(src >= 0); - draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); - break; - - case TGSI_SEMANTIC_FOG: - src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); -#if 1 - if (src < 0) /* XXX temp hack, try demos/fogcoord.c with this */ - src = 0; -#endif - assert(src >= 0); - draw_emit_vertex_attr(vinfo, EMIT_1F, INTERP_PERSPECTIVE, src); - break; - - case TGSI_SEMANTIC_GENERIC: - /* this includes texcoords and varying vars */ - src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, - fs->input_semantic_index[i]); - assert(src >= 0); - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); - break; - - default: - assert(0); - } - } - - draw_compute_vertex_size(vinfo); - - /* XXX only signal this if format really changes */ - cell->dirty |= CELL_NEW_VERTEX_INFO; -} - - -#if 0 -/** - * Recompute cliprect from scissor bounds, scissor enable and surface size. - */ -static void -compute_cliprect(struct cell_context *sp) -{ - unsigned surfWidth, surfHeight; - - if (sp->framebuffer.num_cbufs > 0) { - surfWidth = sp->framebuffer.cbufs[0]->width; - surfHeight = sp->framebuffer.cbufs[0]->height; - } - else { - /* no surface? */ - surfWidth = sp->scissor.maxx; - surfHeight = sp->scissor.maxy; - } - - if (sp->rasterizer->scissor) { - /* clip to scissor rect */ - sp->cliprect.minx = MAX2(sp->scissor.minx, 0); - sp->cliprect.miny = MAX2(sp->scissor.miny, 0); - sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth); - sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight); - } - else { - /* clip to surface bounds */ - sp->cliprect.minx = 0; - sp->cliprect.miny = 0; - sp->cliprect.maxx = surfWidth; - sp->cliprect.maxy = surfHeight; - } -} -#endif - - - -void cell_update_derived( struct cell_context *cell ) -{ - if (cell->dirty & (CELL_NEW_RASTERIZER | - CELL_NEW_FS | - CELL_NEW_VS)) - calculate_vertex_layout( cell ); - -#if 0 - if (cell->dirty & (CELL_NEW_SCISSOR | - CELL_NEW_DEPTH_STENCIL_ALPHA | - CELL_NEW_FRAMEBUFFER)) - compute_cliprect(cell); -#endif - - cell_emit_state(cell); - - cell->dirty = 0; -} diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.c b/src/mesa/pipe/cell/ppu/cell_state_emit.c deleted file mode 100644 index 5d2a786449..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.c +++ /dev/null @@ -1,103 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_util.h" -#include "cell_context.h" -#include "cell_state.h" -#include "cell_state_emit.h" -#include "cell_batch.h" -#include "cell_texture.h" - - -static void -emit_state_cmd(struct cell_context *cell, uint cmd, - const void *state, uint state_size) -{ - uint64_t *dst = (uint64_t *) - cell_batch_alloc(cell, ROUNDUP8(sizeof(uint64_t) + state_size)); - *dst = cmd; - memcpy(dst + 1, state, state_size); -} - - - -void -cell_emit_state(struct cell_context *cell) -{ - if (cell->dirty & CELL_NEW_FRAMEBUFFER) { - struct pipe_surface *cbuf = cell->framebuffer.cbufs[0]; - struct pipe_surface *zbuf = cell->framebuffer.zsbuf; - struct cell_command_framebuffer *fb - = cell_batch_alloc(cell, sizeof(*fb)); - fb->opcode = CELL_CMD_STATE_FRAMEBUFFER; - fb->color_start = cell->cbuf_map[0]; - fb->color_format = cbuf->format; - fb->depth_start = cell->zsbuf_map; - fb->depth_format = zbuf ? zbuf->format : PIPE_FORMAT_NONE; - fb->width = cell->framebuffer.cbufs[0]->width; - fb->height = cell->framebuffer.cbufs[0]->height; - } - - if (cell->dirty & CELL_NEW_BLEND) { - emit_state_cmd(cell, CELL_CMD_STATE_BLEND, - cell->blend, - sizeof(struct pipe_blend_state)); - } - - if (cell->dirty & CELL_NEW_DEPTH_STENCIL) { - emit_state_cmd(cell, CELL_CMD_STATE_DEPTH_STENCIL, - cell->depth_stencil, - sizeof(struct pipe_depth_stencil_alpha_state)); - } - - if (cell->dirty & CELL_NEW_SAMPLER) { - emit_state_cmd(cell, CELL_CMD_STATE_SAMPLER, - cell->sampler[0], sizeof(struct pipe_sampler_state)); - } - - if (cell->dirty & CELL_NEW_TEXTURE) { - struct cell_command_texture texture; - if (cell->texture[0]) { - texture.start = cell->texture[0]->tiled_data; - texture.width = cell->texture[0]->base.width[0]; - texture.height = cell->texture[0]->base.height[0]; - } - else { - texture.start = NULL; - texture.width = 0; - texture.height = 0; - } - - emit_state_cmd(cell, CELL_CMD_STATE_TEXTURE, - &texture, sizeof(struct cell_command_texture)); - } - - if (cell->dirty & CELL_NEW_VERTEX_INFO) { - emit_state_cmd(cell, CELL_CMD_STATE_VERTEX_INFO, - &cell->vertex_info, sizeof(struct vertex_info)); - } -} diff --git a/src/mesa/pipe/cell/ppu/cell_state_emit.h b/src/mesa/pipe/cell/ppu/cell_state_emit.h deleted file mode 100644 index 59f8affe8d..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_emit.h +++ /dev/null @@ -1,36 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef CELL_STATE_EMIT_H -#define CELL_STATE_EMIT_H - - -extern void -cell_emit_state(struct cell_context *cell); - - -#endif /* CELL_STATE_EMIT_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_state_fs.c b/src/mesa/pipe/cell/ppu/cell_state_fs.c deleted file mode 100644 index 3f46a87d18..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_fs.c +++ /dev/null @@ -1,171 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" -#include "pipe/draw/draw_context.h" -#if 0 -#include "pipe/p_shader_tokens.h" -#include "pipe/llvm/gallivm.h" -#include "pipe/tgsi/util/tgsi_dump.h" -#include "pipe/tgsi/exec/tgsi_sse2.h" -#endif - -#include "cell_context.h" -#include "cell_state.h" - - -void * -cell_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - /*struct cell_context *cell = cell_context(pipe);*/ - struct cell_fragment_shader_state *state; - - state = CALLOC_STRUCT(cell_fragment_shader_state); - if (!state) - return NULL; - - state->shader = *templ; - -#if 0 - if (cell->dump_fs) { - tgsi_dump(state->shader.tokens, 0); - } - -#if defined(__i386__) || defined(__386__) - if (cell->use_sse) { - x86_init_func( &state->sse2_program ); - tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); - } -#endif - -#ifdef MESA_LLVM - state->llvm_prog = 0; - if (!gallivm_global_cpu_engine()) { - gallivm_cpu_engine_create(state->llvm_prog); - } - else - gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); -#endif -#endif - - return state; -} - - -void -cell_bind_fs_state(struct pipe_context *pipe, void *fs) -{ - struct cell_context *cell = cell_context(pipe); - - cell->fs = (struct cell_fragment_shader_state *) fs; - - cell->dirty |= CELL_NEW_FS; -} - - -void -cell_delete_fs_state(struct pipe_context *pipe, void *fs) -{ - struct cell_fragment_shader_state *state = - (struct cell_fragment_shader_state *) fs; - - FREE( state ); -} - - -void * -cell_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - struct cell_context *cell = cell_context(pipe); - struct cell_vertex_shader_state *state; - - state = CALLOC_STRUCT(cell_vertex_shader_state); - if (!state) - return NULL; - - state->shader = *templ; - - state->draw_data = draw_create_vertex_shader(cell->draw, &state->shader); - if (state->draw_data == NULL) { - FREE( state ); - return NULL; - } - - return state; -} - - -void -cell_bind_vs_state(struct pipe_context *pipe, void *vs) -{ - struct cell_context *cell = cell_context(pipe); - - cell->vs = (const struct cell_vertex_shader_state *) vs; - - draw_bind_vertex_shader(cell->draw, cell->vs->draw_data); - - cell->dirty |= CELL_NEW_VS; -} - - -void -cell_delete_vs_state(struct pipe_context *pipe, void *vs) -{ - struct cell_context *cell = cell_context(pipe); - - struct cell_vertex_shader_state *state = - (struct cell_vertex_shader_state *) vs; - - draw_delete_vertex_shader(cell->draw, state->draw_data); - FREE( state ); -} - - -void -cell_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) -{ - struct cell_context *cell = cell_context(pipe); - struct pipe_winsys *ws = pipe->winsys; - - assert(shader < PIPE_SHADER_TYPES); - assert(index == 0); - - /* note: reference counting */ - pipe_buffer_reference(ws, - &cell->constants[shader].buffer, - buf->buffer); - cell->constants[shader].size = buf->size; - - cell->dirty |= CELL_NEW_CONSTANTS; -} diff --git a/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c b/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c deleted file mode 100644 index d8128ece54..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_rasterizer.c +++ /dev/null @@ -1,106 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/draw/draw_context.h" -#include "cell_context.h" -#include "cell_state.h" - - - -struct spu_rasterizer_state -{ - unsigned flatshade:1; -#if 0 - unsigned light_twoside:1; - unsigned front_winding:2; /**< PIPE_WINDING_x */ - unsigned cull_mode:2; /**< PIPE_WINDING_x */ - unsigned fill_cw:2; /**< PIPE_POLYGON_MODE_x */ - unsigned fill_ccw:2; /**< PIPE_POLYGON_MODE_x */ - unsigned offset_cw:1; - unsigned offset_ccw:1; -#endif - unsigned scissor:1; - unsigned poly_smooth:1; - unsigned poly_stipple_enable:1; - unsigned point_smooth:1; -#if 0 - unsigned point_sprite:1; - unsigned point_size_per_vertex:1; /**< size computed in vertex shader */ -#endif - unsigned multisample:1; /* XXX maybe more ms state in future */ - unsigned line_smooth:1; - unsigned line_stipple_enable:1; - unsigned line_stipple_factor:8; /**< [1..256] actually */ - unsigned line_stipple_pattern:16; -#if 0 - unsigned bypass_clipping:1; -#endif - unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ - - float line_width; - float point_size; /**< used when no per-vertex size */ -#if 0 - float offset_units; - float offset_scale; - ubyte sprite_coord_mode[PIPE_MAX_SHADER_OUTPUTS]; /**< PIPE_SPRITE_COORD_ */ -#endif -}; - - - -void * -cell_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup) -{ - struct pipe_rasterizer_state *state - = MALLOC(sizeof(struct pipe_rasterizer_state)); - memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); - return state; -} - - -void -cell_bind_rasterizer_state(struct pipe_context *pipe, void *setup) -{ - struct cell_context *cell = cell_context(pipe); - - /* pass-through to draw module */ - draw_set_rasterizer_state(cell->draw, setup); - - cell->rasterizer = (struct pipe_rasterizer_state *)setup; - - cell->dirty |= CELL_NEW_RASTERIZER; -} - - -void -cell_delete_rasterizer_state(struct pipe_context *pipe, void *rasterizer) -{ - FREE(rasterizer); -} diff --git a/src/mesa/pipe/cell/ppu/cell_state_sampler.c b/src/mesa/pipe/cell/ppu/cell_state_sampler.c deleted file mode 100644 index ade6cc8338..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_sampler.c +++ /dev/null @@ -1,84 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: - * Brian Paul - */ - -#include "pipe/p_util.h" -#include "pipe/draw/draw_context.h" -#include "cell_context.h" -#include "cell_state.h" -#include "cell_texture.h" - - -void * -cell_create_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) -{ - return mem_dup(sampler, sizeof(*sampler)); -} - -void -cell_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - assert(unit < PIPE_MAX_SAMPLERS); - cell->sampler[unit] = (struct pipe_sampler_state *)sampler; - - cell->dirty |= CELL_NEW_SAMPLER; -} - - -void -cell_delete_sampler_state(struct pipe_context *pipe, - void *sampler) -{ - FREE( sampler ); -} - - - -void -cell_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, - struct pipe_texture *texture) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - cell->texture[sampler] = texture; - - cell_update_texture_mapping(cell); - - cell->dirty |= CELL_NEW_TEXTURE; -} diff --git a/src/mesa/pipe/cell/ppu/cell_state_surface.c b/src/mesa/pipe/cell/ppu/cell_state_surface.c deleted file mode 100644 index 287610b76b..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_surface.c +++ /dev/null @@ -1,71 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "pipe/p_inlines.h" -#include "cell_context.h" -#include "cell_state.h" - - -void -cell_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb) -{ - struct cell_context *cell = cell_context(pipe); - - if (1 /*memcmp(&cell->framebuffer, fb, sizeof(*fb))*/) { - struct pipe_surface *csurf = fb->cbufs[0]; - struct pipe_surface *zsurf = fb->zsbuf; - uint i; - - /* unmap old surfaces */ - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { - if (cell->framebuffer.cbufs[i] && cell->cbuf_map[i]) { - pipe_surface_unmap(cell->framebuffer.cbufs[i]); - cell->cbuf_map[i] = NULL; - } - } - - if (cell->framebuffer.zsbuf && cell->zsbuf_map) { - pipe_surface_unmap(cell->framebuffer.zsbuf); - cell->zsbuf_map = NULL; - } - - /* update my state */ - cell->framebuffer = *fb; - - /* map new surfaces */ - if (csurf) - cell->cbuf_map[0] = pipe_surface_map(csurf); - - if (zsurf) - cell->zsbuf_map = pipe_surface_map(zsurf); - - cell->dirty |= CELL_NEW_FRAMEBUFFER; - } -} - diff --git a/src/mesa/pipe/cell/ppu/cell_state_vertex.c b/src/mesa/pipe/cell/ppu/cell_state_vertex.c deleted file mode 100644 index 0f01e920f9..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_state_vertex.c +++ /dev/null @@ -1,63 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - - -#include "cell_context.h" -#include "cell_state.h" - -#include "pipe/draw/draw_context.h" - - -void -cell_set_vertex_element(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_element *attrib) -{ - struct cell_context *cell = cell_context(pipe); - assert(index < PIPE_ATTRIB_MAX); - cell->vertex_element[index] = *attrib; /* struct copy */ - cell->dirty |= CELL_NEW_VERTEX; - - draw_set_vertex_element(cell->draw, index, attrib); -} - - -void -cell_set_vertex_buffer(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer) -{ - struct cell_context *cell = cell_context(pipe); - assert(index < PIPE_ATTRIB_MAX); - cell->vertex_buffer[index] = *buffer; /* struct copy */ - cell->dirty |= CELL_NEW_VERTEX; - - draw_set_vertex_buffer(cell->draw, index, buffer); -} diff --git a/src/mesa/pipe/cell/ppu/cell_surface.c b/src/mesa/pipe/cell/ppu/cell_surface.c deleted file mode 100644 index fca93e4742..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_surface.c +++ /dev/null @@ -1,179 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" -#include "pipe/util/p_tile.h" -#include "cell_context.h" -#include "cell_surface.h" - - -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -static void -cell_surface_data(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - const void *src, unsigned src_pitch, - unsigned srcx, unsigned srcy, - unsigned width, unsigned height) -{ - pipe_copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - pipe_surface_unmap(dst); -} - - -static void -cell_surface_copy(struct pipe_context *pipe, - unsigned do_flip, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - struct pipe_surface *src, - unsigned srcx, unsigned srcy, - unsigned width, unsigned height) -{ - assert( dst->cpp == src->cpp ); - - pipe_copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - pipe_surface_map(src), - do_flip ? -src->pitch : src->pitch, - srcx, do_flip ? 1 - srcy - height : srcy); - - pipe_surface_unmap(src); - pipe_surface_unmap(dst); -} - - -static void * -get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) -{ - return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; -} - - -#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8)) - - -/** - * Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static void -cell_surface_fill(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - unsigned width, unsigned height, unsigned value) -{ - unsigned i, j; - void *dst_map = pipe_surface_map(dst); - - assert(dst->pitch > 0); - assert(width <= dst->pitch); - - switch (dst->cpp) { - case 1: - { - ubyte *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; - case 2: - { - ushort *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; - case 4: - { - unsigned *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - break; - case 8: - { - /* expand the 4-byte clear value to an 8-byte value */ - ushort *row = (ushort *) get_pointer(dst, dst_map, dstx, dsty); - ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); - ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); - ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); - ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff); - val0 = (val0 << 8) | val0; - val1 = (val1 << 8) | val1; - val2 = (val2 << 8) | val2; - val3 = (val3 << 8) | val3; - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) { - row[j*4+0] = val0; - row[j*4+1] = val1; - row[j*4+2] = val2; - row[j*4+3] = val3; - } - row += dst->pitch * 4; - } - } - break; - default: - assert(0); - break; - } - - pipe_surface_unmap( dst ); -} - - -void -cell_init_surface_functions(struct cell_context *cell) -{ - cell->pipe.surface_copy = cell_surface_copy; - cell->pipe.surface_fill = cell_surface_fill; -} diff --git a/src/mesa/pipe/cell/ppu/cell_surface.h b/src/mesa/pipe/cell/ppu/cell_surface.h deleted file mode 100644 index 9e58f32944..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_surface.h +++ /dev/null @@ -1,42 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef CELL_SURFACE_H -#define CELL_SURFACE_H - - -struct cell_context; - - -extern void -cell_init_surface_functions(struct cell_context *cell); - - -#endif /* SP_SURFACE_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_texture.c b/src/mesa/pipe/cell/ppu/cell_texture.c deleted file mode 100644 index c8ef36002f..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_texture.c +++ /dev/null @@ -1,252 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Michel Dänzer - */ - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "pipe/p_winsys.h" - -#include "cell_context.h" -#include "cell_state.h" -#include "cell_texture.h" - - -/* Simple, maximally packed layout. - */ - -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - - -static void -cell_texture_layout(struct cell_texture * spt) -{ - struct pipe_texture *pt = &spt->base; - unsigned level; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - - spt->buffer_size = 0; - - for ( level = 0 ; level <= pt->last_level ; level++ ) { - pt->width[level] = width; - pt->height[level] = height; - pt->depth[level] = depth; - - spt->level_offset[level] = spt->buffer_size; - - spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * - ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * - width * pt->cpp; - - width = minify(width); - height = minify(height); - depth = minify(depth); - } -} - - -struct pipe_texture * -cell_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) -{ - struct cell_texture *spt = CALLOC_STRUCT(cell_texture); - if (!spt) - return NULL; - - spt->base = *templat; - - cell_texture_layout(spt); - - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); - - if (!spt->buffer) { - FREE(spt); - return NULL; - } - - return &spt->base; -} - - -void -cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) -{ - if (!*pt) - return; - - /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); - */ - if (--(*pt)->refcount <= 0) { - struct cell_texture *spt = cell_texture(*pt); - - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); - */ - - pipe_buffer_reference(pipe->winsys, &spt->buffer, NULL); - - FREE(spt); - } - *pt = NULL; -} - - -/** - * Called via pipe->get_tex_surface() - */ -struct pipe_surface * -cell_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - struct cell_texture *spt = cell_texture(pt); - struct pipe_surface *ps; - - ps = pipe->winsys->surface_alloc(pipe->winsys); - if (ps) { - assert(ps->refcount); - assert(ps->winsys); - pipe_buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); - ps->format = pt->format; - ps->cpp = pt->cpp; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->pitch = ps->width; - ps->offset = spt->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { - ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - (pt->compressed ? ps->height/4 : ps->height) * - ps->width * ps->cpp; - } else { - assert(face == 0); - assert(zslice == 0); - } - } - return ps; -} - - - -static void -tile_copy_data(uint w, uint h, uint tile_size, uint *dst, const uint *src) -{ - const uint tile_size2 = tile_size * tile_size; - const uint h_t = h / tile_size, w_t = w / tile_size; - - uint it, jt; /* tile counters */ - uint i, j; /* intra-tile counters */ - - for (it = 0; it < h_t; it++) { - for (jt = 0; jt < w_t; jt++) { - /* fill in tile (i, j) */ - uint *tdst = dst + (it * w_t + jt) * tile_size2; - for (i = 0; i < tile_size; i++) { - for (j = 0; j < tile_size; j++) { - const uint srci = it * tile_size + i; - const uint srcj = jt * tile_size + j; - *tdst++ = src[srci * h + srcj]; - } - } - } - } -} - - - -/** - * Convert linear texture image data to tiled format for SPU usage. - */ -static void -cell_tile_texture(struct cell_context *cell, - struct cell_texture *texture) -{ - uint face = 0, level = 0, zslice = 0; - struct pipe_surface *surf; - const uint w = texture->base.width[0], h = texture->base.height[0]; - const uint *src; - - /* temporary restrictions: */ - assert(w >= TILE_SIZE); - assert(h >= TILE_SIZE); - assert(w % TILE_SIZE == 0); - assert(h % TILE_SIZE == 0); - - surf = cell_get_tex_surface(&cell->pipe, &texture->base, face, level, zslice); - ASSERT(surf); - - src = (const uint *) pipe_surface_map(surf); - - if (texture->tiled_data) { - align_free(texture->tiled_data); - } - texture->tiled_data = align_malloc(w * h * 4, 16); - - tile_copy_data(w, h, TILE_SIZE, texture->tiled_data, src); - - pipe_surface_unmap(surf); - - pipe_surface_reference(&surf, NULL); -} - - - -void -cell_update_texture_mapping(struct cell_context *cell) -{ - uint face = 0, level = 0, zslice = 0; - - if (cell->texture[0]) - cell_tile_texture(cell, cell->texture[0]); -#if 0 - if (cell->tex_surf && cell->tex_map) { - pipe_surface_unmap(cell->tex_surf); - cell->tex_map = NULL; - } - - /* XXX free old surface */ - - cell->tex_surf = cell_get_tex_surface(&cell->pipe, - &cell->texture[0]->base, - face, level, zslice); - - cell->tex_map = pipe_surface_map(cell->tex_surf); -#endif -} diff --git a/src/mesa/pipe/cell/ppu/cell_texture.h b/src/mesa/pipe/cell/ppu/cell_texture.h deleted file mode 100644 index 0264fed88e..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_texture.h +++ /dev/null @@ -1,80 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef CELL_TEXTURE_H -#define CELL_TEXTURE_H - - -struct pipe_context; -struct pipe_texture; - - -/** - * Subclass of pipe_texture - */ -struct cell_texture -{ - struct pipe_texture base; - - unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; - unsigned long buffer_size; - - void *tiled_data; /* XXX this may be temporary */ /*ALIGN16*/ -}; - - -/** cast wrapper */ -static INLINE struct cell_texture * -cell_texture(struct pipe_texture *pt) -{ - return (struct cell_texture *) pt; -} - - - -extern struct pipe_texture * -cell_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat); - -extern void -cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); - -extern struct pipe_surface * -cell_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice); - - -extern void -cell_update_texture_mapping(struct cell_context *cell); - - -#endif /* CELL_TEXTURE */ diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.c b/src/mesa/pipe/cell/ppu/cell_vbuf.c deleted file mode 100644 index e9fafe492e..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.c +++ /dev/null @@ -1,294 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Authors - * Brian Paul - */ - - -#include "cell_batch.h" -#include "cell_context.h" -#include "cell_flush.h" -#include "cell_spu.h" -#include "cell_vbuf.h" -#include "pipe/draw/draw_vbuf.h" - - -/** Allow vertex data to be inlined after RENDER command */ -#define ALLOW_INLINE_VERTS 1 - - -/** - * Subclass of vbuf_render because we need a cell_context pointer in - * a few places. - */ -struct cell_vbuf_render -{ - struct vbuf_render base; - struct cell_context *cell; - uint prim; /**< PIPE_PRIM_x */ - uint vertex_size; /**< in bytes */ - void *vertex_buffer; /**< just for debug, really */ - uint vertex_buf; /**< in [0, CELL_NUM_BUFFERS-1] */ -}; - - -/** cast wrapper */ -static struct cell_vbuf_render * -cell_vbuf_render(struct vbuf_render *vbr) -{ - return (struct cell_vbuf_render *) vbr; -} - - - -static const struct vertex_info * -cell_vbuf_get_vertex_info(struct vbuf_render *vbr) -{ - struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); - return &cvbr->cell->vertex_info; -} - - -static void * -cell_vbuf_allocate_vertices(struct vbuf_render *vbr, - ushort vertex_size, ushort nr_vertices) -{ - struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); - /*printf("Alloc verts %u * %u\n", vertex_size, nr_vertices);*/ - - assert(cvbr->vertex_buf == ~0); - cvbr->vertex_buf = cell_get_empty_buffer(cvbr->cell); - cvbr->vertex_buffer = cvbr->cell->buffer[cvbr->vertex_buf]; - cvbr->vertex_size = vertex_size; - return cvbr->vertex_buffer; -} - - -static void -cell_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, - unsigned vertex_size, unsigned vertices_used) -{ - struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); - struct cell_context *cell = cvbr->cell; - - /* - printf("%s vertex_buf = %u count = %u\n", - __FUNCTION__, cvbr->vertex_buf, vertices_used); - */ - - /* Tell SPUs they can release the vert buf */ - if (cvbr->vertex_buf != ~0U) { - struct cell_command_release_verts *release - = (struct cell_command_release_verts *) - cell_batch_alloc(cell, sizeof(struct cell_command_release_verts)); - release->opcode = CELL_CMD_RELEASE_VERTS; - release->vertex_buf = cvbr->vertex_buf; - } - - cvbr->vertex_buf = ~0; - cell_flush_int(&cell->pipe, 0x0); - - assert(vertices == cvbr->vertex_buffer); - cvbr->vertex_buffer = NULL; -} - - - -static void -cell_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) -{ - struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); - cvbr->prim = prim; - /*printf("cell_set_prim %u\n", prim);*/ -} - - -static void -cell_vbuf_draw(struct vbuf_render *vbr, - const ushort *indices, - uint nr_indices) -{ - struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); - struct cell_context *cell = cvbr->cell; - float xmin, ymin, xmax, ymax; - uint i; - uint nr_vertices = 0, min_index = ~0; - const void *vertices = cvbr->vertex_buffer; - const uint vertex_size = cvbr->vertex_size; - - for (i = 0; i < nr_indices; i++) { - if (indices[i] > nr_vertices) - nr_vertices = indices[i]; - if (indices[i] < min_index) - min_index = indices[i]; - } - nr_vertices++; - -#if 0 - /*if (min_index > 0)*/ - printf("%s min_index = %u\n", __FUNCTION__, min_index); -#endif - -#if 0 - printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u\n", - nr_indices, nr_vertices); - printf(" "); - for (i = 0; i < nr_indices; i += 3) { - printf("%u %u %u, ", indices[i+0], indices[i+1], indices[i+2]); - } - printf("\n"); -#elif 0 - printf("cell_vbuf_draw() nr_indices = %u nr_verts = %u indexes = [%u %u %u ...]\n", - nr_indices, nr_vertices, - indices[0], indices[1], indices[2]); - printf("ind space = %u, vert space = %u, space = %u\n", - nr_indices * 2, - nr_vertices * 4 * cell->vertex_info.size, - cell_batch_free_space(cell)); -#endif - - /* compute x/y bounding box */ - xmin = ymin = 1e50; - xmax = ymax = -1e50; - for (i = min_index; i < nr_vertices; i++) { - const float *v = (float *) ((ubyte *) vertices + i * vertex_size); - if (v[0] < xmin) - xmin = v[0]; - if (v[0] > xmax) - xmax = v[0]; - if (v[1] < ymin) - ymin = v[1]; - if (v[1] > ymax) - ymax = v[1]; - } -#if 0 - printf("PPU Bounds %g, %g .. %g, %g\n", xmin, ymin, xmax, ymax); - fflush(stdout); -#endif - - if (cvbr->prim != PIPE_PRIM_TRIANGLES) - return; /* only render tris for now */ - - /* build/insert batch RENDER command */ - { - const uint index_bytes = ROUNDUP8(nr_indices * 2); - const uint vertex_bytes = nr_vertices * 4 * cell->vertex_info.size; - const uint batch_size = sizeof(struct cell_command_render) + index_bytes; - - struct cell_command_render *render - = (struct cell_command_render *) - cell_batch_alloc(cell, batch_size); - - render->opcode = CELL_CMD_RENDER; - render->prim_type = cvbr->prim; - - render->num_indexes = nr_indices; - render->min_index = min_index; - - /* append indices after render command */ - memcpy(render + 1, indices, nr_indices * 2); - - /* if there's room, append vertices after the indices, else leave - * vertices in the original/separate buffer. - */ - render->vertex_size = 4 * cell->vertex_info.size; - render->num_verts = nr_vertices; - if (ALLOW_INLINE_VERTS && - min_index == 0 && - vertex_bytes + 16 <= cell_batch_free_space(cell)) { - /* vertex data inlined, after indices, at 16-byte boundary */ - void *dst = cell_batch_alloc_aligned(cell, vertex_bytes, 16); - memcpy(dst, vertices, vertex_bytes); - render->inline_verts = TRUE; - render->vertex_buf = ~0; - } - else { - /* vertex data in separate buffer */ - render->inline_verts = FALSE; - ASSERT(cvbr->vertex_buf >= 0); - render->vertex_buf = cvbr->vertex_buf; - } - - render->xmin = xmin; - render->ymin = ymin; - render->xmax = xmax; - render->ymax = ymax; - } - -#if 0 - /* helpful for debug */ - cell_flush_int(&cell->pipe, PIPE_FLUSH_WAIT); -#endif -} - - -static void -cell_vbuf_destroy(struct vbuf_render *vbr) -{ - struct cell_vbuf_render *cvbr = cell_vbuf_render(vbr); - cvbr->cell->vbuf_render = NULL; - FREE(cvbr); -} - - -/** - * Initialize the post-transform vertex buffer information for the given - * context. - */ -void -cell_init_vbuf(struct cell_context *cell) -{ - assert(cell->draw); - - cell->vbuf_render = CALLOC_STRUCT(cell_vbuf_render); - - /* The max number of indexes is what can fix into a batch buffer, - * minus the render and release-verts commands. - */ - cell->vbuf_render->base.max_indices - = (CELL_BUFFER_SIZE - - sizeof(struct cell_command_render) - - sizeof(struct cell_command_release_verts)) - / sizeof(ushort); - cell->vbuf_render->base.max_vertex_buffer_bytes = CELL_BUFFER_SIZE; - - cell->vbuf_render->base.get_vertex_info = cell_vbuf_get_vertex_info; - cell->vbuf_render->base.allocate_vertices = cell_vbuf_allocate_vertices; - cell->vbuf_render->base.set_primitive = cell_vbuf_set_primitive; - cell->vbuf_render->base.draw = cell_vbuf_draw; - cell->vbuf_render->base.release_vertices = cell_vbuf_release_vertices; - cell->vbuf_render->base.destroy = cell_vbuf_destroy; - - cell->vbuf_render->cell = cell; -#if 1 - cell->vbuf_render->vertex_buf = ~0; -#endif - - cell->vbuf = draw_vbuf_stage(cell->draw, &cell->vbuf_render->base); -} diff --git a/src/mesa/pipe/cell/ppu/cell_vbuf.h b/src/mesa/pipe/cell/ppu/cell_vbuf.h deleted file mode 100644 index d265cbf770..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_vbuf.h +++ /dev/null @@ -1,38 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef CELL_VBUF_H -#define CELL_VBUF_H - - -struct cell_context; - -extern void -cell_init_vbuf(struct cell_context *cell); - - -#endif /* CELL_VBUF_H */ diff --git a/src/mesa/pipe/cell/ppu/cell_vertex_shader.c b/src/mesa/pipe/cell/ppu/cell_vertex_shader.c deleted file mode 100644 index 80dd500b34..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_vertex_shader.c +++ /dev/null @@ -1,120 +0,0 @@ -/* - * (C) Copyright IBM Corporation 2008 - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file cell_vertex_shader.c - * Vertex shader interface routines for Cell. - * - * \author Ian Romanick - */ - -#include "pipe/p_defines.h" -#include "pipe/p_context.h" -#include "pipe/p_winsys.h" - -#include "cell_context.h" -#include "cell_draw_arrays.h" -#include "cell_spu.h" -#include "cell_batch.h" - -#include "pipe/cell/common.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" - -/** - * Run the vertex shader on all vertices in the vertex queue. - * Called by the draw module when the vertx cache needs to be flushed. - */ -void -cell_vertex_shader_queue_flush(struct draw_context *draw) -{ - struct cell_context *const cell = - (struct cell_context *) draw->driver_private; - struct cell_command_vs *const vs = &cell_global.command[0].vs; - uint64_t *batch; - struct cell_array_info *array_info; - unsigned i, j; - - assert(draw->vs.queue_nr != 0); - - /* XXX: do this on statechange: - */ - draw_update_vertex_fetch(draw); - - for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { - batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*array_info)); - - batch[0] = CELL_CMD_STATE_VS_ARRAY_INFO; - - array_info = (struct cell_array_info *) &batch[1]; - assert(draw->vertex_fetch.src_ptr[i] != NULL); - array_info->base = (uintptr_t) draw->vertex_fetch.src_ptr[i]; - array_info->attr = i; - array_info->pitch = draw->vertex_fetch.pitch[i]; - array_info->format = draw->vertex_element[i].src_format; - } - - batch = cell_batch_alloc(cell, sizeof(batch[0]) - + sizeof(struct pipe_viewport_state)); - batch[0] = CELL_CMD_STATE_VIEWPORT; - (void) memcpy(&batch[1], &draw->viewport, - sizeof(struct pipe_viewport_state)); - - cell_batch_flush(cell); - - vs->opcode = CELL_CMD_VS_EXECUTE; - vs->shader.num_outputs = draw->num_vs_outputs; - vs->shader.declarations = (uintptr_t) draw->machine.Declarations; - vs->shader.num_declarations = draw->machine.NumDeclarations; - vs->shader.instructions = (uintptr_t) draw->machine.Instructions; - vs->shader.num_instructions = draw->machine.NumInstructions; - vs->shader.uniforms = (uintptr_t) draw->user.constants; - vs->shader.immediates = (uintptr_t) draw->machine.Imms; - vs->shader.num_immediates = draw->machine.ImmLimit / 4; - vs->nr_attrs = draw->vertex_fetch.nr_attrs; - - (void) memcpy(vs->plane, draw->plane, sizeof(draw->plane)); - vs->nr_planes = draw->nr_planes; - - for (i = 0; i < draw->vs.queue_nr; i += SPU_VERTS_PER_BATCH) { - const unsigned n = MIN2(SPU_VERTS_PER_BATCH, draw->vs.queue_nr - i); - - for (j = 0; j < n; j++) { - vs->elts[j] = draw->vs.queue[i + j].elt; - vs->vOut[j] = (uintptr_t) draw->vs.queue[i + j].dest; - } - - for (/* empty */; j < SPU_VERTS_PER_BATCH; j++) { - vs->elts[j] = vs->elts[0]; - vs->vOut[j] = vs->vOut[0]; - } - - vs->num_elts = n; - send_mbox_message(cell_global.spe_contexts[0], CELL_CMD_VS_EXECUTE); - - cell_flush_int(& cell->pipe, PIPE_FLUSH_WAIT); - } - - draw->vs.queue_nr = 0; -} diff --git a/src/mesa/pipe/cell/ppu/cell_winsys.c b/src/mesa/pipe/cell/ppu/cell_winsys.c deleted file mode 100644 index ebabce3c8f..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_winsys.c +++ /dev/null @@ -1,40 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "pipe/p_util.h" -#include "cell_winsys.h" - - -struct cell_winsys * -cell_get_winsys(uint format) -{ - struct cell_winsys *cws = CALLOC_STRUCT(cell_winsys); - if (cws) - cws->preferredFormat = format; - return cws; -} diff --git a/src/mesa/pipe/cell/ppu/cell_winsys.h b/src/mesa/pipe/cell/ppu/cell_winsys.h deleted file mode 100644 index ae2af5696b..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_winsys.h +++ /dev/null @@ -1,50 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef CELL_WINSYS_H -#define CELL_WINSYS_H - -#include "pipe/p_compiler.h" - - -/** - * Very simple winsys at this time. - * Will probably eventually add SPU control info. - */ -struct cell_winsys -{ - uint preferredFormat; -}; - - -extern struct cell_winsys * -cell_get_winsys(uint format); - - - -#endif diff --git a/src/mesa/pipe/cell/spu/Makefile b/src/mesa/pipe/cell/spu/Makefile deleted file mode 100644 index f202971d73..0000000000 --- a/src/mesa/pipe/cell/spu/Makefile +++ /dev/null @@ -1,72 +0,0 @@ -# Gallium3D Cell driver: SPU code - -# This makefile builds the g3d_spu.a file that's linked into the -# PPU code/library. - - -TOP = ../../../../.. -include $(TOP)/configs/linux-cell - - -PROG = g3d - -PROG_SPU = $(PROG)_spu -PROG_SPU_A = $(PROG)_spu.a -PROG_SPU_EMBED_O = $(PROG)_spu-embed.o - - -SOURCES = \ - spu_main.c \ - spu_blend.c \ - spu_render.c \ - spu_texture.c \ - spu_tile.c \ - spu_tri.c \ - spu_exec.c \ - spu_util.c \ - spu_vertex_fetch.c \ - spu_vertex_shader.c - -SPU_OBJECTS = $(SOURCES:.c=.o) \ - -SPU_ASM_OUT = $(SOURCES:.c=.s) \ - -INCLUDE_DIRS = -I$(TOP)/src/mesa - - -.c.o: - $(SPU_CC) $(SPU_CFLAGS) -c $< - -.c.s: - $(SPU_CC) $(SPU_CFLAGS) -S $< - - -# The .a file will be linked into the main/PPU executable -default: $(PROG_SPU_A) - -$(PROG_SPU_A): $(PROG_SPU_EMBED_O) - $(SPU_AR) $(SPU_AR_FLAGS) $(PROG_SPU_A) $(PROG_SPU_EMBED_O) - -$(PROG_SPU_EMBED_O): $(PROG_SPU) - $(SPU_EMBED) $(SPU_EMBED_FLAGS) $(PROG_SPU) $(PROG_SPU) $(PROG_SPU_EMBED_O) - -$(PROG_SPU): $(SPU_OBJECTS) - $(SPU_CC) -o $(PROG_SPU) $(SPU_OBJECTS) $(SPU_LFLAGS) - - - -asmfiles: $(SPU_ASM_OUT) - - -clean: - rm -f *~ *.o *.a *.d *.s $(PROG_SPU) - - - -depend: $(SOURCES) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDE_DIRS) $(SOURCES) 2> /dev/null - -include depend - diff --git a/src/mesa/pipe/cell/spu/spu_blend.c b/src/mesa/pipe/cell/spu/spu_blend.c deleted file mode 100644 index 23ec0eeb45..0000000000 --- a/src/mesa/pipe/cell/spu/spu_blend.c +++ /dev/null @@ -1,62 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "spu_main.h" -#include "spu_blend.h" -#include "spu_colorpack.h" - - -void -blend_quad(uint itx, uint ity, vector float colors[4]) -{ - /* simple SRC_ALPHA, ONE_MINUS_SRC_ALPHA blending */ - vector float fbc00 = spu_unpack_color(spu.ctile.ui[ity][itx]); - vector float fbc01 = spu_unpack_color(spu.ctile.ui[ity][itx+1]); - vector float fbc10 = spu_unpack_color(spu.ctile.ui[ity+1][itx]); - vector float fbc11 = spu_unpack_color(spu.ctile.ui[ity+1][itx+1]); - - vector float alpha00 = spu_splats(spu_extract(colors[0], 3)); - vector float alpha01 = spu_splats(spu_extract(colors[1], 3)); - vector float alpha10 = spu_splats(spu_extract(colors[2], 3)); - vector float alpha11 = spu_splats(spu_extract(colors[3], 3)); - - vector float one_minus_alpha00 = spu_sub(spu_splats(1.0f), alpha00); - vector float one_minus_alpha01 = spu_sub(spu_splats(1.0f), alpha01); - vector float one_minus_alpha10 = spu_sub(spu_splats(1.0f), alpha10); - vector float one_minus_alpha11 = spu_sub(spu_splats(1.0f), alpha11); - - colors[0] = spu_add(spu_mul(colors[0], alpha00), - spu_mul(fbc00, one_minus_alpha00)); - colors[1] = spu_add(spu_mul(colors[1], alpha01), - spu_mul(fbc01, one_minus_alpha01)); - colors[2] = spu_add(spu_mul(colors[2], alpha10), - spu_mul(fbc10, one_minus_alpha10)); - colors[3] = spu_add(spu_mul(colors[3], alpha11), - spu_mul(fbc11, one_minus_alpha11)); -} - diff --git a/src/mesa/pipe/cell/spu/spu_blend.h b/src/mesa/pipe/cell/spu/spu_blend.h deleted file mode 100644 index 2b594b578b..0000000000 --- a/src/mesa/pipe/cell/spu/spu_blend.h +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef SPU_BLEND_H -#define SPU_BLEND_H - - -extern void -blend_quad(uint itx, uint ity, vector float colors[4]); - - -#endif /* SPU_BLEND_H */ diff --git a/src/mesa/pipe/cell/spu/spu_colorpack.h b/src/mesa/pipe/cell/spu/spu_colorpack.h deleted file mode 100644 index e9fee8a3a6..0000000000 --- a/src/mesa/pipe/cell/spu/spu_colorpack.h +++ /dev/null @@ -1,110 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - - -#ifndef SPU_COLORPACK_H -#define SPU_COLORPACK_H - - -#include - - -static INLINE unsigned int -spu_pack_R8G8B8A8(vector float rgba) -{ - vector unsigned int out = spu_convtu(rgba, 32); - - out = spu_shuffle(out, out, ((vector unsigned char) { - 0, 4, 8, 12, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0 }) ); - - return spu_extract(out, 0); -} - - -static INLINE unsigned int -spu_pack_A8R8G8B8(vector float rgba) -{ - vector unsigned int out = spu_convtu(rgba, 32); - out = spu_shuffle(out, out, ((vector unsigned char) { - 12, 0, 4, 8, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0}) ); - return spu_extract(out, 0); -} - - -static INLINE unsigned int -spu_pack_B8G8R8A8(vector float rgba) -{ - vector unsigned int out = spu_convtu(rgba, 32); - out = spu_shuffle(out, out, ((vector unsigned char) { - 8, 4, 0, 12, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0}) ); - return spu_extract(out, 0); -} - - -static INLINE unsigned int -spu_pack_color_shuffle(vector float rgba, vector unsigned char shuffle) -{ - vector unsigned int out = spu_convtu(rgba, 32); - out = spu_shuffle(out, out, shuffle); - return spu_extract(out, 0); -} - - -static INLINE vector float -spu_unpack_color(uint color) -{ - vector unsigned int color_u4 = spu_splats(color); - color_u4 = spu_shuffle(color_u4, color_u4, - ((vector unsigned char) { - 0, 0, 0, 0, - 5, 5, 5, 5, - 10, 10, 10, 10, - 15, 15, 15, 15}) ); - return spu_convtf(color_u4, 32); -} - - -static INLINE vector float -spu_unpack_A8R8G8B8(uint color) -{ - vector unsigned int color_u4 = spu_splats(color); - color_u4 = spu_shuffle(color_u4, color_u4, - ((vector unsigned char) { - 5, 5, 5, 5, - 10, 10, 10, 10, - 15, 15, 15, 15, - 0, 0, 0, 0}) ); - - return spu_convtf(color_u4, 32); -} - - -#endif /* SPU_COLORPACK_H */ diff --git a/src/mesa/pipe/cell/spu/spu_exec.c b/src/mesa/pipe/cell/spu/spu_exec.c deleted file mode 100644 index e51008b9b3..0000000000 --- a/src/mesa/pipe/cell/spu/spu_exec.c +++ /dev/null @@ -1,1948 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * TGSI interpretor/executor. - * - * Flow control information: - * - * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) - * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special - * care since a condition may be true for some quad components but false - * for other components. - * - * We basically execute all statements (even if they're in the part of - * an IF/ELSE clause that's "not taken") and use a special mask to - * control writing to destination registers. This is the ExecMask. - * See store_dest(). - * - * The ExecMask is computed from three other masks (CondMask, LoopMask and - * ContMask) which are controlled by the flow control instructions (namely: - * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). - * - * - * Authors: - * Michal Krol - * Brian Paul - */ - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include "pipe/p_compiler.h" -#include "pipe/p_state.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/util/tgsi_util.h" -#include "spu_exec.h" -#include "spu_main.h" -#include "spu_vertex_shader.h" - -#define TILE_TOP_LEFT 0 -#define TILE_TOP_RIGHT 1 -#define TILE_BOTTOM_LEFT 2 -#define TILE_BOTTOM_RIGHT 3 - -/* - * Shorthand locations of various utility registers (_I = Index, _C = Channel) - */ -#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I -#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C -#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I -#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C -#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I -#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C -#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I -#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C -#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I -#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C -#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I -#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C -#define TEMP_128_I TGSI_EXEC_TEMP_128_I -#define TEMP_128_C TGSI_EXEC_TEMP_128_C -#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I -#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C -#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I -#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C -#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I -#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C -#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I -#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C -#define TEMP_R0 TGSI_EXEC_TEMP_R0 - -#define FOR_EACH_CHANNEL(CHAN)\ - for (CHAN = 0; CHAN < 4; CHAN++) - -#define IS_CHANNEL_ENABLED(INST, CHAN)\ - ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) - -#define IS_CHANNEL_ENABLED2(INST, CHAN)\ - ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) - -#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED( INST, CHAN )) - -#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED2( INST, CHAN )) - - -/** The execution mask depends on the conditional mask and the loop mask */ -#define UPDATE_EXEC_MASK(MACH) \ - MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask - - -#define CHAN_X 0 -#define CHAN_Y 1 -#define CHAN_Z 2 -#define CHAN_W 3 - - - -/** - * Initialize machine state by expanding tokens to full instructions, - * allocating temporary storage, setting up constants, etc. - * After this, we can call spu_exec_machine_run() many times. - */ -void -spu_exec_machine_init(struct spu_exec_machine *mach, - uint numSamplers, - struct spu_sampler *samplers, - unsigned processor) -{ - qword zero; - qword not_zero; - uint i; - - mach->Samplers = samplers; - mach->Processor = processor; - mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; - - zero = si_xor(zero, zero); - not_zero = si_xori(zero, 0xff); - - /* Setup constants. */ - mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q = zero; - mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].q = not_zero; - mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].q = si_shli(not_zero, -1); - mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].q = si_shli(not_zero, 31); - - mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q = (qword) spu_splats(1.0f); - mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q = (qword) spu_splats(2.0f); - mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].q = (qword) spu_splats(128.0f); - mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].q = (qword) spu_splats(-128.0f); -} - - -static INLINE qword -micro_abs(qword src) -{ - return si_rotmi(si_shli(src, 1), -1); -} - -static INLINE qword -micro_ceil(qword src) -{ - return (qword) _ceilf4((vec_float4) src); -} - -static INLINE qword -micro_cos(qword src) -{ - return (qword) _cosf4((vec_float4) src); -} - -static const qword br_shuf = { - TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, - TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, - TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, - TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, - TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, - TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, - TILE_BOTTOM_RIGHT + 0, TILE_BOTTOM_RIGHT + 1, - TILE_BOTTOM_RIGHT + 2, TILE_BOTTOM_RIGHT + 3, -}; - -static const qword bl_shuf = { - TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, - TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, - TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, - TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, - TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, - TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, - TILE_BOTTOM_LEFT + 0, TILE_BOTTOM_LEFT + 1, - TILE_BOTTOM_LEFT + 2, TILE_BOTTOM_LEFT + 3, -}; - -static const qword tl_shuf = { - TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, - TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, - TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, - TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, - TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, - TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, - TILE_TOP_LEFT + 0, TILE_TOP_LEFT + 1, - TILE_TOP_LEFT + 2, TILE_TOP_LEFT + 3, -}; - -static qword -micro_ddx(qword src) -{ - qword bottom_right = si_shufb(src, src, br_shuf); - qword bottom_left = si_shufb(src, src, bl_shuf); - - return si_fs(bottom_right, bottom_left); -} - -static qword -micro_ddy(qword src) -{ - qword top_left = si_shufb(src, src, tl_shuf); - qword bottom_left = si_shufb(src, src, bl_shuf); - - return si_fs(top_left, bottom_left); -} - -static INLINE qword -micro_div(qword src0, qword src1) -{ - return (qword) _divf4((vec_float4) src0, (vec_float4) src1); -} - -static qword -micro_flr(qword src) -{ - return (qword) _floorf4((vec_float4) src); -} - -static qword -micro_frc(qword src) -{ - return si_fs(src, (qword) _floorf4((vec_float4) src)); -} - -static INLINE qword -micro_ge(qword src0, qword src1) -{ - return si_or(si_fceq(src0, src1), si_fcgt(src0, src1)); -} - -static qword -micro_lg2(qword src) -{ - return (qword) _log2f4((vec_float4) src); -} - -static INLINE qword -micro_lt(qword src0, qword src1) -{ - const qword tmp = si_or(si_fceq(src0, src1), si_fcgt(src0, src1)); - - return si_xori(tmp, 0xff); -} - -static INLINE qword -micro_max(qword src0, qword src1) -{ - return si_selb(src1, src0, si_fcgt(src0, src1)); -} - -static INLINE qword -micro_min(qword src0, qword src1) -{ - return si_selb(src0, src1, si_fcgt(src0, src1)); -} - -static qword -micro_neg(qword src) -{ - return si_xor(src, (qword) spu_splats(0x80000000)); -} - -static qword -micro_set_sign(qword src) -{ - return si_or(src, (qword) spu_splats(0x80000000)); -} - -static qword -micro_pow(qword src0, qword src1) -{ - return (qword) _powf4((vec_float4) src0, (vec_float4) src1); -} - -static qword -micro_rnd(qword src) -{ - const qword half = (qword) spu_splats(0.5f); - - /* May be able to use _roundf4. There may be some difference, though. - */ - return (qword) _floorf4((vec_float4) si_fa(src, half)); -} - -static INLINE qword -micro_ishr(qword src0, qword src1) -{ - return si_rotma(src0, si_sfi(src1, 0)); -} - -static qword -micro_trunc(qword src) -{ - return (qword) _truncf4((vec_float4) src); -} - -static qword -micro_sin(qword src) -{ - return (qword) _sinf4((vec_float4) src); -} - -static INLINE qword -micro_sqrt(qword src) -{ - return (qword) _sqrtf4((vec_float4) src); -} - -static void -fetch_src_file_channel( - const struct spu_exec_machine *mach, - const uint file, - const uint swizzle, - const union spu_exec_channel *index, - union spu_exec_channel *chan ) -{ - switch( swizzle ) { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - switch( file ) { - case TGSI_FILE_CONSTANT: { - unsigned char buffer[32] ALIGN16_ATTRIB; - unsigned i; - - for (i = 0; i < 4; i++) { - const float *ptr = mach->Consts[index->i[i]]; - const uint64_t addr = (uint64_t)(uintptr_t) ptr; - const unsigned size = ((addr & 0x0f) == 0) ? 16 : 32; - - mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); - wait_on_mask(1 << TAG_VERTEX_BUFFER); - - (void) memcpy(& chan->f[i], &buffer[(addr & 0x0f) - + (sizeof(float) * swizzle)], sizeof(float)); - } - break; - } - - case TGSI_FILE_INPUT: - chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_TEMPORARY: - chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_IMMEDIATE: - assert( index->i[0] < (int) mach->ImmLimit ); - assert( index->i[1] < (int) mach->ImmLimit ); - assert( index->i[2] < (int) mach->ImmLimit ); - assert( index->i[3] < (int) mach->ImmLimit ); - - chan->f[0] = mach->Imms[index->i[0]][swizzle]; - chan->f[1] = mach->Imms[index->i[1]][swizzle]; - chan->f[2] = mach->Imms[index->i[2]][swizzle]; - chan->f[3] = mach->Imms[index->i[3]][swizzle]; - break; - - case TGSI_FILE_ADDRESS: - chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_OUTPUT: - /* vertex/fragment output vars can be read too */ - chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; - break; - - default: - assert( 0 ); - } - break; - - case TGSI_EXTSWIZZLE_ZERO: - *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; - break; - - case TGSI_EXTSWIZZLE_ONE: - *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; - break; - - default: - assert( 0 ); - } -} - -static void -fetch_source( - const struct spu_exec_machine *mach, - union spu_exec_channel *chan, - const struct tgsi_full_src_register *reg, - const uint chan_index ) -{ - union spu_exec_channel index; - uint swizzle; - - index.i[0] = - index.i[1] = - index.i[2] = - index.i[3] = reg->SrcRegister.Index; - - if (reg->SrcRegister.Indirect) { - union spu_exec_channel index2; - union spu_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle(®->SrcRegisterInd, - CHAN_X); - fetch_src_file_channel( - mach, - reg->SrcRegisterInd.File, - swizzle, - &index2, - &indir_index ); - - index.q = si_a(index.q, indir_index.q); - } - - if( reg->SrcRegister.Dimension ) { - switch( reg->SrcRegister.File ) { - case TGSI_FILE_INPUT: - index.q = si_mpyi(index.q, 17); - break; - case TGSI_FILE_CONSTANT: - index.q = si_shli(index.q, 12); - break; - default: - assert( 0 ); - } - - index.i[0] += reg->SrcRegisterDim.Index; - index.i[1] += reg->SrcRegisterDim.Index; - index.i[2] += reg->SrcRegisterDim.Index; - index.i[3] += reg->SrcRegisterDim.Index; - - if (reg->SrcRegisterDim.Indirect) { - union spu_exec_channel index2; - union spu_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterDimInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); - fetch_src_file_channel( - mach, - reg->SrcRegisterDimInd.File, - swizzle, - &index2, - &indir_index ); - - index.q = si_a(index.q, indir_index.q); - } - } - - swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); - fetch_src_file_channel( - mach, - reg->SrcRegister.File, - swizzle, - &index, - chan ); - - switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { - case TGSI_UTIL_SIGN_CLEAR: - chan->q = micro_abs(chan->q); - break; - - case TGSI_UTIL_SIGN_SET: - chan->q = micro_set_sign(chan->q); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - chan->q = micro_neg(chan->q); - break; - - case TGSI_UTIL_SIGN_KEEP: - break; - } - - if (reg->SrcRegisterExtMod.Complement) { - chan->q = si_fs(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, chan->q); - } -} - -static void -store_dest( - struct spu_exec_machine *mach, - const union spu_exec_channel *chan, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - uint chan_index ) -{ - union spu_exec_channel *dst; - - switch( reg->DstRegister.File ) { - case TGSI_FILE_NULL: - return; - - case TGSI_FILE_OUTPUT: - dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] - + reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_TEMPORARY: - dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_ADDRESS: - dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; - break; - - default: - assert( 0 ); - return; - } - - switch (inst->Instruction.Saturate) - { - case TGSI_SAT_NONE: - if (mach->ExecMask & 0x1) - dst->i[0] = chan->i[0]; - if (mach->ExecMask & 0x2) - dst->i[1] = chan->i[1]; - if (mach->ExecMask & 0x4) - dst->i[2] = chan->i[2]; - if (mach->ExecMask & 0x8) - dst->i[3] = chan->i[3]; - break; - - case TGSI_SAT_ZERO_ONE: - /* XXX need to obey ExecMask here */ - dst->q = micro_max(chan->q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); - dst->q = micro_min(dst->q, mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q); - break; - - case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); - break; - - default: - assert( 0 ); - } -} - -#define FETCH(VAL,INDEX,CHAN)\ - fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) - -#define STORE(VAL,INDEX,CHAN)\ - store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) - - -/** - * Execute ARB-style KIL which is predicated by a src register. - * Kill fragment if any of the four values is less than zero. - */ -static void -exec_kilp(struct spu_exec_machine *mach, - const struct tgsi_full_instruction *inst) -{ - uint uniquemask; - uint chan_index; - uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ - union spu_exec_channel r[1]; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - for (chan_index = 0; chan_index < 4; chan_index++) - { - uint swizzle; - uint i; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle ( - &inst->FullSrcRegisters[0], - chan_index); - - /* check if the component has not been already tested */ - if (uniquemask & (1 << swizzle)) - continue; - uniquemask |= 1 << swizzle; - - FETCH(&r[0], 0, chan_index); - for (i = 0; i < 4; i++) - if (r[0].f[i] < 0.0f) - kilmask |= 1 << i; - } - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; -} - - -/* - * Fetch a texel using STR texture coordinates. - */ -static void -fetch_texel( struct spu_sampler *sampler, - const union spu_exec_channel *s, - const union spu_exec_channel *t, - const union spu_exec_channel *p, - float lodbias, /* XXX should be float[4] */ - union spu_exec_channel *r, - union spu_exec_channel *g, - union spu_exec_channel *b, - union spu_exec_channel *a ) -{ - qword rgba[4]; - qword out[4]; - - sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, (float *) rgba); - - _transpose_matrix4x4(out, rgba); - r->q = out[0]; - g->q = out[1]; - b->q = out[2]; - a->q = out[3]; -} - - -static void -exec_tex(struct spu_exec_machine *mach, - const struct tgsi_full_instruction *inst, - boolean biasLod) -{ - const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; - union spu_exec_channel r[8]; - uint chan_index; - float lodBias; - - /* printf("Sampler %u unit %u\n", sampler, unit); */ - - switch (inst->InstructionExtTexture.Texture) { - case TGSI_TEXTURE_1D: - - FETCH(&r[0], 0, CHAN_X); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[1], 0, CHAN_W); - r[0].q = micro_div(r[0].q, r[1].q); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - if (biasLod) { - FETCH(&r[1], 0, CHAN_W); - lodBias = r[2].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ - &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ - break; - - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - r[0].q = micro_div(r[0].q, r[3].q); - r[1].q = micro_div(r[1].q, r[3].q); - r[2].q = micro_div(r[2].q, r[3].q); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, /* inputs */ - &r[0], &r[1], &r[2], &r[3]); /* outputs */ - break; - - case TGSI_TEXTURE_3D: - case TGSI_TEXTURE_CUBE: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - r[0].q = micro_div(r[0].q, r[3].q); - r[1].q = micro_div(r[1].q, r[3].q); - r[2].q = micro_div(r[2].q, r[3].q); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, - &r[0], &r[1], &r[2], &r[3]); - break; - - default: - assert (0); - } - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[chan_index], 0, chan_index ); - } -} - - - -static void -constant_interpolation( - struct spu_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; - } -} - -static void -linear_interpolation( - struct spu_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - const float x = mach->QuadPos.xyzw[0].f[0]; - const float y = mach->QuadPos.xyzw[1].f[0]; - const float dadx = mach->InterpCoefs[attrib].dadx[chan]; - const float dady = mach->InterpCoefs[attrib].dady[chan]; - const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; - mach->Inputs[attrib].xyzw[chan].f[0] = a0; - mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; - mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; - mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; -} - -static void -perspective_interpolation( - struct spu_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - const float x = mach->QuadPos.xyzw[0].f[0]; - const float y = mach->QuadPos.xyzw[1].f[0]; - const float dadx = mach->InterpCoefs[attrib].dadx[chan]; - const float dady = mach->InterpCoefs[attrib].dady[chan]; - const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; - const float *w = mach->QuadPos.xyzw[3].f; - /* divide by W here */ - mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; - mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; - mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; - mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; -} - - -typedef void (* interpolation_func)( - struct spu_exec_machine *mach, - unsigned attrib, - unsigned chan ); - -static void -exec_declaration(struct spu_exec_machine *mach, - const struct tgsi_full_declaration *decl) -{ - if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { - if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask; - interpolation_func interp; - - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - switch( decl->Interpolation.Interpolate ) { - case TGSI_INTERPOLATE_CONSTANT: - interp = constant_interpolation; - break; - - case TGSI_INTERPOLATE_LINEAR: - interp = linear_interpolation; - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - interp = perspective_interpolation; - break; - - default: - assert( 0 ); - } - - if( mask == TGSI_WRITEMASK_XYZW ) { - unsigned i, j; - - for( i = first; i <= last; i++ ) { - for( j = 0; j < NUM_CHANNELS; j++ ) { - interp( mach, i, j ); - } - } - } - else { - unsigned i, j; - - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - interp( mach, i, j ); - } - } - } - } - } - } -} - -static void -exec_instruction( - struct spu_exec_machine *mach, - const struct tgsi_full_instruction *inst, - int *pc ) -{ - uint chan_index; - union spu_exec_channel r[8]; - - (*pc)++; - - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = si_cflts(r[0].q, 0); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LIT: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_X ); - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - r[0].q = micro_max(r[0].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[1], 0, CHAN_Y ); - r[1].q = micro_max(r[1].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); - - FETCH( &r[2], 0, CHAN_W ); - r[2].q = micro_min(r[2].q, mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].q); - r[2].q = micro_max(r[2].q, mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].q); - r[1].q = micro_pow(r[1].q, r[2].q); - - /* r0 = (r0 > 0.0) ? r1 : 0.0 - */ - r[0].q = si_fcgt(r[0].q, mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q); - r[0].q = si_selb(mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q, r[1].q, - r[0].q); - STORE( &r[0], 0, CHAN_Z ); - } - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH( &r[0], 0, CHAN_X ); - r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH( &r[0], 0, CHAN_X ); - r[0].q = micro_sqrt(r[0].q); - r[0].q = micro_div(mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].q, r[0].q); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXP: - assert (0); - break; - - case TGSI_OPCODE_LOG: - assert (0); - break; - - case TGSI_OPCODE_MUL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - r[0].q = si_fm(r[0].q, r[1].q); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_ADD: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - r[0].q = si_fa(r[0].q, r[1].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - r[0].q = si_fm(r[0].q, r[1].q); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - r[0].q = si_fma(r[1].q, r[2].q, r[0].q); - - - FETCH( &r[1], 0, CHAN_Z ); - FETCH( &r[2], 1, CHAN_Z ); - r[0].q = si_fma(r[1].q, r[2].q, r[0].q); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - r[0].q = si_fm(r[0].q, r[1].q); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - r[0].q = si_fma(r[1].q, r[2].q, r[0].q); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - r[0].q = si_fma(r[1].q, r[2].q, r[0].q); - - FETCH(&r[1], 0, CHAN_W); - FETCH(&r[2], 1, CHAN_W); - - r[0].q = si_fma(r[1].q, r[2].q, r[0].q); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DST: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - FETCH( &r[0], 0, CHAN_Y ); - FETCH( &r[1], 1, CHAN_Y); - r[0].q = si_fm(r[0].q, r[1].q); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_Z ); - STORE( &r[0], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - FETCH( &r[0], 1, CHAN_W ); - STORE( &r[0], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MIN: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - r[0].q = micro_min(r[0].q, r[1].q); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_MAX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - r[0].q = micro_max(r[0].q, r[1].q); - - STORE(&r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - - r[0].q = micro_ge(r[0].q, r[1].q); - r[0].q = si_xori(r[0].q, 0xff); - - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - r[0].q = micro_ge(r[0].q, r[1].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - FETCH( &r[2], 2, chan_index ); - r[0].q = si_fma(r[0].q, r[1].q, r[2].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SUB: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - r[0].q = si_fs(r[0].q, r[1].q); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - r[1].q = si_fs(r[1].q, r[2].q); - r[0].q = si_fma(r[0].q, r[1].q, r[2].q); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_CND: - assert (0); - break; - - case TGSI_OPCODE_CND0: - assert (0); - break; - - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - assert (0); - break; - - case TGSI_OPCODE_INDEX: - assert (0); - break; - - case TGSI_OPCODE_NEGATE: - assert (0); - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = micro_frc(r[0].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CLAMP: - assert (0); - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = micro_flr(r[0].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_ROUND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = micro_rnd(r[0].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ - FETCH(&r[0], 0, CHAN_X); - - r[0].q = micro_pow(mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].q, r[0].q); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ - FETCH( &r[0], 0, CHAN_X ); - r[0].q = micro_lg2(r[0].q); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - r[0].q = micro_pow(r[0].q, r[1].q); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - FETCH(&r[0], 0, CHAN_Y); - FETCH(&r[1], 1, CHAN_Z); - FETCH(&r[3], 0, CHAN_Z); - FETCH(&r[4], 1, CHAN_Y); - - /* r2 = (r0 * r1) - (r3 * r5) - */ - r[2].q = si_fm(r[3].q, r[5].q); - r[2].q = si_fms(r[0].q, r[1].q, r[2].q); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &r[2], 0, CHAN_X ); - } - - FETCH(&r[2], 1, CHAN_X); - FETCH(&r[5], 0, CHAN_X); - - /* r3 = (r3 * r2) - (r1 * r5) - */ - r[1].q = si_fm(r[1].q, r[5].q); - r[3].q = si_fms(r[3].q, r[2].q, r[1].q); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - STORE( &r[3], 0, CHAN_Y ); - } - - /* r5 = (r5 * r4) - (r0 * r2) - */ - r[0].q = si_fm(r[0].q, r[2].q); - r[5].q = si_fms(r[5].q, r[4].q, r[0].q); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - STORE( &r[5], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MULTIPLYMATRIX: - assert (0); - break; - - case TGSI_OPCODE_ABS: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - - r[0].q = micro_abs(r[0].q); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_RCC: - assert (0); - break; - - case TGSI_OPCODE_DPH: - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - r[0].q = si_fm(r[0].q, r[1].q); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - r[0].q = si_fma(r[1].q, r[2].q, r[0].q); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - r[0].q = si_fma(r[1].q, r[2].q, r[0].q); - - FETCH(&r[1], 1, CHAN_W); - - r[0].q = si_fa(r[0].q, r[1].q); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_COS: - FETCH(&r[0], 0, CHAN_X); - - r[0].q = micro_cos(r[0].q); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = micro_ddx(r[0].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDY: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = micro_ddy(r[0].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_KILP: - exec_kilp (mach, inst); - break; - - case TGSI_OPCODE_KIL: - /* for enabled ExecMask bits, set the killed bit */ - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; - break; - - case TGSI_OPCODE_PK2H: - assert (0); - break; - - case TGSI_OPCODE_PK2US: - assert (0); - break; - - case TGSI_OPCODE_PK4B: - assert (0); - break; - - case TGSI_OPCODE_PK4UB: - assert (0); - break; - - case TGSI_OPCODE_RFL: - assert (0); - break; - - case TGSI_OPCODE_SEQ: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - - r[0].q = si_fceq(r[0].q, r[1].q); - - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SFL: - assert (0); - break; - - case TGSI_OPCODE_SGT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - r[0].q = si_fcgt(r[0].q, r[1].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SIN: - FETCH( &r[0], 0, CHAN_X ); - r[0].q = micro_sin(r[0].q); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - - r[0].q = si_fcgt(r[0].q, r[1].q); - r[0].q = si_xori(r[0].q, 0xff); - - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SNE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - - r[0].q = si_fceq(r[0].q, r[1].q); - r[0].q = si_xori(r[0].q, 0xff); - - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_STR: - assert (0); - break; - - case TGSI_OPCODE_TEX: - /* simple texture lookup */ - /* src[0] = texcoord */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, FALSE); - break; - - case TGSI_OPCODE_TXB: - /* Texture lookup with lod bias */ - /* src[0] = texcoord (src[0].w = load bias) */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); - break; - - case TGSI_OPCODE_TXD: - /* Texture lookup with explict partial derivatives */ - /* src[0] = texcoord */ - /* src[1] = d[strq]/dx */ - /* src[2] = d[strq]/dy */ - /* src[3] = sampler unit */ - assert (0); - break; - - case TGSI_OPCODE_TXL: - /* Texture lookup with explit LOD */ - /* src[0] = texcoord (src[0].w = load bias) */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); - break; - - case TGSI_OPCODE_UP2H: - assert (0); - break; - - case TGSI_OPCODE_UP2US: - assert (0); - break; - - case TGSI_OPCODE_UP4B: - assert (0); - break; - - case TGSI_OPCODE_UP4UB: - assert (0); - break; - - case TGSI_OPCODE_X2D: - assert (0); - break; - - case TGSI_OPCODE_ARA: - assert (0); - break; - - case TGSI_OPCODE_ARR: - assert (0); - break; - - case TGSI_OPCODE_BRA: - assert (0); - break; - - case TGSI_OPCODE_CAL: - /* skip the call if no execution channels are enabled */ - if (mach->ExecMask) { - /* do the call */ - - /* push the Cond, Loop, Cont stacks */ - assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); - mach->CondStack[mach->CondStackTop++] = mach->CondMask; - assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; - assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; - - assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); - mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; - - /* note that PC was already incremented above */ - mach->CallStack[mach->CallStackTop++] = *pc; - *pc = inst->InstructionExtLabel.Label; - } - break; - - case TGSI_OPCODE_RET: - mach->FuncMask &= ~mach->ExecMask; - UPDATE_EXEC_MASK(mach); - - if (mach->ExecMask == 0x0) { - /* really return now (otherwise, keep executing */ - - if (mach->CallStackTop == 0) { - /* returning from main() */ - *pc = -1; - return; - } - *pc = mach->CallStack[--mach->CallStackTop]; - - /* pop the Cond, Loop, Cont stacks */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; - assert(mach->FuncStackTop > 0); - mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; - - UPDATE_EXEC_MASK(mach); - } - break; - - case TGSI_OPCODE_SSG: - assert (0); - break; - - case TGSI_OPCODE_CMP: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - /* r0 = (r0 < 0.0) ? r1 : r2 - */ - r[3].q = si_xor(r[3].q, r[3].q); - r[0].q = micro_lt(r[0].q, r[3].q); - r[0].q = si_selb(r[1].q, r[2].q, r[0].q); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_SCS: - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( &r[0], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - r[1].q = micro_cos(r[0].q); - STORE( &r[1], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - r[1].q = micro_sin(r[0].q); - STORE( &r[1], 0, CHAN_Y ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_NRM: - assert (0); - break; - - case TGSI_OPCODE_DIV: - assert( 0 ); - break; - - case TGSI_OPCODE_DP2: - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - r[0].q = si_fm(r[0].q, r[1].q); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - r[0].q = si_fma(r[1].q, r[2].q, r[0].q); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_IF: - /* push CondMask */ - assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); - mach->CondStack[mach->CondStackTop++] = mach->CondMask; - FETCH( &r[0], 0, CHAN_X ); - /* update CondMask */ - if( ! r[0].u[0] ) { - mach->CondMask &= ~0x1; - } - if( ! r[0].u[1] ) { - mach->CondMask &= ~0x2; - } - if( ! r[0].u[2] ) { - mach->CondMask &= ~0x4; - } - if( ! r[0].u[3] ) { - mach->CondMask &= ~0x8; - } - UPDATE_EXEC_MASK(mach); - /* Todo: If CondMask==0, jump to ELSE */ - break; - - case TGSI_OPCODE_ELSE: - /* invert CondMask wrt previous mask */ - { - uint prevMask; - assert(mach->CondStackTop > 0); - prevMask = mach->CondStack[mach->CondStackTop - 1]; - mach->CondMask = ~mach->CondMask & prevMask; - UPDATE_EXEC_MASK(mach); - /* Todo: If CondMask==0, jump to ENDIF */ - } - break; - - case TGSI_OPCODE_ENDIF: - /* pop CondMask */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_END: - /* halt execution */ - *pc = -1; - break; - - case TGSI_OPCODE_REP: - assert (0); - break; - - case TGSI_OPCODE_ENDREP: - assert (0); - break; - - case TGSI_OPCODE_PUSHA: - assert (0); - break; - - case TGSI_OPCODE_POPA: - assert (0); - break; - - case TGSI_OPCODE_CEIL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = micro_ceil(r[0].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_I2F: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = si_csflt(r[0].q, 0); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_NOT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = si_xorbi(r[0].q, 0xff); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_TRUNC: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - r[0].q = micro_trunc(r[0].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SHL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - - r[0].q = si_shl(r[0].q, r[1].q); - - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SHR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - r[0].q = micro_ishr(r[0].q, r[1].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_AND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - r[0].q = si_and(r[0].q, r[1].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_OR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - r[0].q = si_or(r[0].q, r[1].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOD: - assert (0); - break; - - case TGSI_OPCODE_XOR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - r[0].q = si_xor(r[0].q, r[1].q); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SAD: - assert (0); - break; - - case TGSI_OPCODE_TXF: - assert (0); - break; - - case TGSI_OPCODE_TXQ: - assert (0); - break; - - case TGSI_OPCODE_EMIT: - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; - break; - - case TGSI_OPCODE_ENDPRIM: - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; - break; - - case TGSI_OPCODE_LOOP: - /* fall-through (for now) */ - case TGSI_OPCODE_BGNLOOP2: - /* push LoopMask and ContMasks */ - assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; - assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; - break; - - case TGSI_OPCODE_ENDLOOP: - /* fall-through (for now at least) */ - case TGSI_OPCODE_ENDLOOP2: - /* Restore ContMask, but don't pop */ - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; - if (mach->LoopMask) { - /* repeat loop: jump to instruction just past BGNLOOP */ - *pc = inst->InstructionExtLabel.Label + 1; - } - else { - /* exit loop: pop LoopMask */ - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - /* pop ContMask */ - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; - } - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_BRK: - /* turn off loop channels for each enabled exec channel */ - mach->LoopMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_CONT: - /* turn off cont channels for each enabled exec channel */ - mach->ContMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_BGNSUB: - /* no-op */ - break; - - case TGSI_OPCODE_ENDSUB: - /* no-op */ - break; - - case TGSI_OPCODE_NOISE1: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE2: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE3: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE4: - assert( 0 ); - break; - - case TGSI_OPCODE_NOP: - break; - - default: - assert( 0 ); - } -} - - -/** - * Run TGSI interpreter. - * \return bitmask of "alive" quad components - */ -uint -spu_exec_machine_run( struct spu_exec_machine *mach ) -{ - uint i; - int pc = 0; - - mach->CondMask = 0xf; - mach->LoopMask = 0xf; - mach->ContMask = 0xf; - mach->FuncMask = 0xf; - mach->ExecMask = 0xf; - - mach->CondStackTop = 0; /* temporarily subvert this assertion */ - assert(mach->CondStackTop == 0); - assert(mach->LoopStackTop == 0); - assert(mach->ContStackTop == 0); - assert(mach->CallStackTop == 0); - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; - - if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; - mach->Primitives[0] = 0; - } - - - /* execute declarations (interpolants) */ - if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { - for (i = 0; i < mach->NumDeclarations; i++) { - uint8_t buffer[sizeof(struct tgsi_full_declaration) + 32] ALIGN16_ATTRIB; - struct tgsi_full_declaration decl; - unsigned long decl_addr = (unsigned long) (mach->Declarations+i); - unsigned size = ((sizeof(decl) + (decl_addr & 0x0f) + 0x0f) & ~0x0f); - - mfc_get(buffer, decl_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); - wait_on_mask(1 << TAG_INSTRUCTION_FETCH); - - memcpy(& decl, buffer + (decl_addr & 0x0f), sizeof(decl)); - exec_declaration( mach, &decl ); - } - } - - /* execute instructions, until pc is set to -1 */ - while (pc != -1) { - uint8_t buffer[sizeof(struct tgsi_full_instruction) + 32] ALIGN16_ATTRIB; - struct tgsi_full_instruction inst; - unsigned long inst_addr = (unsigned long) (mach->Instructions + pc); - unsigned size = ((sizeof(inst) + (inst_addr & 0x0f) + 0x0f) & ~0x0f); - - assert(pc < mach->NumInstructions); - mfc_get(buffer, inst_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); - wait_on_mask(1 << TAG_INSTRUCTION_FETCH); - - memcpy(& inst, buffer + (inst_addr & 0x0f), sizeof(inst)); - exec_instruction( mach, & inst, &pc ); - } - -#if 0 - /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ - if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { - /* - * Scale back depth component. - */ - for (i = 0; i < 4; i++) - mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; - } -#endif - - return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; -} - - diff --git a/src/mesa/pipe/cell/spu/spu_exec.h b/src/mesa/pipe/cell/spu/spu_exec.h deleted file mode 100644 index b4c7661ef6..0000000000 --- a/src/mesa/pipe/cell/spu/spu_exec.h +++ /dev/null @@ -1,172 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#if !defined SPU_EXEC_H -#define SPU_EXEC_H - -#include "pipe/p_compiler.h" -#include "pipe/tgsi/exec/tgsi_exec.h" - -#if defined __cplusplus -extern "C" { -#endif - -/** - * Registers may be treated as float, signed int or unsigned int. - */ -union spu_exec_channel -{ - float f[QUAD_SIZE]; - int i[QUAD_SIZE]; - unsigned u[QUAD_SIZE]; - qword q; -}; - -/** - * A vector[RGBA] of channels[4 pixels] - */ -struct spu_exec_vector -{ - union spu_exec_channel xyzw[NUM_CHANNELS]; -}; - -/** - * For fragment programs, information for computing fragment input - * values from plane equation of the triangle/line. - */ -struct spu_interp_coef -{ - float a0[NUM_CHANNELS]; /* in an xyzw layout */ - float dadx[NUM_CHANNELS]; - float dady[NUM_CHANNELS]; -}; - - -struct softpipe_tile_cache; /**< Opaque to TGSI */ - -/** - * Information for sampling textures, which must be implemented - * by code outside the TGSI executor. - */ -struct spu_sampler -{ - const struct pipe_sampler_state *state; - struct pipe_texture *texture; - /** Get samples for four fragments in a quad */ - void (*get_samples)(struct spu_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]); - void *pipe; /*XXX temporary*/ - struct softpipe_tile_cache *cache; -}; - - -/** - * Run-time virtual machine state for executing TGSI shader. - */ -struct spu_exec_machine -{ - /* - * 32 program temporaries - * 4 internal temporaries - * 1 address - */ - struct spu_exec_vector Temps[TGSI_EXEC_NUM_TEMPS - + TGSI_EXEC_NUM_ADDRS + 1] - ALIGN16_ATTRIB; - - struct spu_exec_vector *Addrs; - - struct spu_sampler *Samplers; - - float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; - unsigned ImmLimit; - float (*Consts)[4]; - struct spu_exec_vector *Inputs; - struct spu_exec_vector *Outputs; - unsigned Processor; - - /* GEOMETRY processor only. */ - unsigned *Primitives; - - /* FRAGMENT processor only. */ - const struct spu_interp_coef *InterpCoefs; - struct spu_exec_vector QuadPos; - - /* Conditional execution masks */ - uint CondMask; /**< For IF/ELSE/ENDIF */ - uint LoopMask; /**< For BGNLOOP/ENDLOOP */ - uint ContMask; /**< For loop CONT statements */ - uint FuncMask; /**< For function calls */ - uint ExecMask; /**< = CondMask & LoopMask */ - - /** Condition mask stack (for nested conditionals) */ - uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; - int CondStackTop; - - /** Loop mask stack (for nested loops) */ - uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; - int LoopStackTop; - - /** Loop continue mask stack (see comments in tgsi_exec.c) */ - uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; - int ContStackTop; - - /** Function execution mask stack (for executing subroutine code) */ - uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; - int FuncStackTop; - - /** Function call stack for saving/restoring the program counter */ - uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; - int CallStackTop; - - struct tgsi_full_instruction *Instructions; - uint NumInstructions; - - struct tgsi_full_declaration *Declarations; - uint NumDeclarations; -}; - - -extern void -spu_exec_machine_init(struct spu_exec_machine *mach, - uint numSamplers, - struct spu_sampler *samplers, - unsigned processor); - -extern uint -spu_exec_machine_run( struct spu_exec_machine *mach ); - - -#if defined __cplusplus -} /* extern "C" */ -#endif - -#endif /* SPU_EXEC_H */ diff --git a/src/mesa/pipe/cell/spu/spu_main.c b/src/mesa/pipe/cell/spu/spu_main.c deleted file mode 100644 index e375197fe6..0000000000 --- a/src/mesa/pipe/cell/spu/spu_main.c +++ /dev/null @@ -1,567 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/* main() for Cell SPU code */ - - -#include -#include - -#include "spu_main.h" -#include "spu_render.h" -#include "spu_texture.h" -#include "spu_tile.h" -//#include "spu_test.h" -#include "spu_vertex_shader.h" -#include "pipe/cell/common.h" -#include "pipe/p_defines.h" - - -/* -helpful headers: -/usr/lib/gcc/spu/4.1.1/include/spu_mfcio.h -/opt/ibm/cell-sdk/prototype/sysroot/usr/include/libmisc.h -*/ - -boolean Debug = FALSE; - -struct spu_global spu; - -struct spu_vs_context draw; - -/** - * Tell the PPU that this SPU has finished copying a buffer to - * local store and that it may be reused by the PPU. - * This is done by writting a 16-byte batch-buffer-status block back into - * main memory (in cell_context->buffer_status[]). - */ -static void -release_buffer(uint buffer) -{ - /* Evidently, using less than a 16-byte status doesn't work reliably */ - static const uint status[4] ALIGN16_ATTRIB - = {CELL_BUFFER_STATUS_FREE, 0, 0, 0}; - - const uint index = 4 * (spu.init.id * CELL_NUM_BUFFERS + buffer); - uint *dst = spu.init.buffer_status + index; - - ASSERT(buffer < CELL_NUM_BUFFERS); - - mfc_put((void *) &status, /* src in local memory */ - (unsigned int) dst, /* dst in main memory */ - sizeof(status), /* size */ - TAG_MISC, /* tag is unimportant */ - 0, /* tid */ - 0 /* rid */); -} - - -/** - * For tiles whose status is TILE_STATUS_CLEAR, write solid-filled - * tiles back to the main framebuffer. - */ -static void -really_clear_tiles(uint surfaceIndex) -{ - const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; - uint i; - - if (surfaceIndex == 0) { - clear_c_tile(&spu.ctile); - - for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { - uint tx = i % spu.fb.width_tiles; - uint ty = i / spu.fb.width_tiles; - if (spu.ctile_status[ty][tx] == TILE_STATUS_CLEAR) { - put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); - } - } - } - else { - clear_z_tile(&spu.ztile); - - for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { - uint tx = i % spu.fb.width_tiles; - uint ty = i / spu.fb.width_tiles; - if (spu.ztile_status[ty][tx] == TILE_STATUS_CLEAR) - put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 1); - } - } - -#if 0 - wait_on_mask(1 << TAG_SURFACE_CLEAR); -#endif -} - - -static void -cmd_clear_surface(const struct cell_command_clear_surface *clear) -{ - const uint num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; - uint i; - - if (Debug) - printf("SPU %u: CLEAR SURF %u to 0x%08x\n", spu.init.id, - clear->surface, clear->value); - -#define CLEAR_OPT 1 -#if CLEAR_OPT - /* set all tile's status to CLEAR */ - if (clear->surface == 0) { - memset(spu.ctile_status, TILE_STATUS_CLEAR, sizeof(spu.ctile_status)); - spu.fb.color_clear_value = clear->value; - } - else { - memset(spu.ztile_status, TILE_STATUS_CLEAR, sizeof(spu.ztile_status)); - spu.fb.depth_clear_value = clear->value; - } - return; -#endif - - if (clear->surface == 0) { - spu.fb.color_clear_value = clear->value; - clear_c_tile(&spu.ctile); - } - else { - spu.fb.depth_clear_value = clear->value; - clear_z_tile(&spu.ztile); - } - - /* - printf("SPU: %s num=%d w=%d h=%d\n", - __FUNCTION__, num_tiles, spu.fb.width_tiles, spu.fb.height_tiles); - */ - - for (i = spu.init.id; i < num_tiles; i += spu.init.num_spus) { - uint tx = i % spu.fb.width_tiles; - uint ty = i / spu.fb.width_tiles; - if (clear->surface == 0) - put_tile(tx, ty, &spu.ctile, TAG_SURFACE_CLEAR, 0); - else - put_tile(tx, ty, &spu.ztile, TAG_SURFACE_CLEAR, 1); - /* XXX we don't want this here, but it fixes bad tile results */ - } - -#if 0 - wait_on_mask(1 << TAG_SURFACE_CLEAR); -#endif - - if (Debug) - printf("SPU %u: CLEAR SURF done\n", spu.init.id); -} - - -static void -cmd_release_verts(const struct cell_command_release_verts *release) -{ - if (Debug) - printf("SPU %u: RELEASE VERTS %u\n", - spu.init.id, release->vertex_buf); - ASSERT(release->vertex_buf != ~0U); - release_buffer(release->vertex_buf); -} - - -static void -cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) -{ - if (Debug) - printf("SPU %u: FRAMEBUFFER: %d x %d at %p, cformat 0x%x zformat 0x%x\n", - spu.init.id, - cmd->width, - cmd->height, - cmd->color_start, - cmd->color_format, - cmd->depth_format); - - ASSERT_ALIGN16(cmd->color_start); - ASSERT_ALIGN16(cmd->depth_start); - - spu.fb.color_start = cmd->color_start; - spu.fb.depth_start = cmd->depth_start; - spu.fb.color_format = cmd->color_format; - spu.fb.depth_format = cmd->depth_format; - spu.fb.width = cmd->width; - spu.fb.height = cmd->height; - spu.fb.width_tiles = (spu.fb.width + TILE_SIZE - 1) / TILE_SIZE; - spu.fb.height_tiles = (spu.fb.height + TILE_SIZE - 1) / TILE_SIZE; - - if (spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM) - spu.fb.zsize = 4; - else if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) - spu.fb.zsize = 2; - else - spu.fb.zsize = 0; - - if (spu.fb.color_format == PIPE_FORMAT_A8R8G8B8_UNORM) - spu.color_shuffle = ((vector unsigned char) { - 12, 0, 4, 8, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0}); - else if (spu.fb.color_format == PIPE_FORMAT_B8G8R8A8_UNORM) - spu.color_shuffle = ((vector unsigned char) { - 8, 4, 0, 12, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0}); - else - ASSERT(0); -} - - -static void -cmd_state_blend(const struct pipe_blend_state *state) -{ - if (Debug) - printf("SPU %u: BLEND: enabled %d\n", - spu.init.id, - state->blend_enable); - - memcpy(&spu.blend, state, sizeof(*state)); -} - - -static void -cmd_state_depth_stencil(const struct pipe_depth_stencil_alpha_state *state) -{ - if (Debug) - printf("SPU %u: DEPTH_STENCIL: ztest %d\n", - spu.init.id, - state->depth.enabled); - - memcpy(&spu.depth_stencil, state, sizeof(*state)); -} - - -static void -cmd_state_sampler(const struct pipe_sampler_state *state) -{ - if (Debug) - printf("SPU %u: SAMPLER\n", - spu.init.id); - - memcpy(&spu.sampler[0], state, sizeof(*state)); - if (spu.sampler[0].min_img_filter == PIPE_TEX_FILTER_LINEAR) - spu.sample_texture = sample_texture_bilinear; - else - spu.sample_texture = sample_texture_nearest; -} - - -static void -cmd_state_texture(const struct cell_command_texture *texture) -{ - if (Debug) - printf("SPU %u: TEXTURE at %p size %u x %u\n", - spu.init.id, texture->start, texture->width, texture->height); - - memcpy(&spu.texture, texture, sizeof(*texture)); - spu.tex_size = (vector float) - { spu.texture.width, spu.texture.height, 0.0, 0.0}; - spu.tex_size_mask = (vector unsigned int) - { spu.texture.width - 1, spu.texture.height - 1, 0, 0 }; -} - - -static void -cmd_state_vertex_info(const struct vertex_info *vinfo) -{ - if (Debug) { - printf("SPU %u: VERTEX_INFO num_attribs=%u\n", spu.init.id, - vinfo->num_attribs); - } - ASSERT(vinfo->num_attribs >= 1); - ASSERT(vinfo->num_attribs <= 8); - memcpy(&spu.vertex_info, vinfo, sizeof(*vinfo)); -} - - -static void -cmd_state_vs_array_info(const struct cell_array_info *vs_info) -{ - const unsigned attr = vs_info->attr; - - ASSERT(attr < PIPE_ATTRIB_MAX); - draw.vertex_fetch.src_ptr[attr] = vs_info->base; - draw.vertex_fetch.pitch[attr] = vs_info->pitch; - draw.vertex_fetch.format[attr] = vs_info->format; - draw.vertex_fetch.dirty = 1; -} - - -static void -cmd_finish(void) -{ - if (Debug) - printf("SPU %u: FINISH\n", spu.init.id); - really_clear_tiles(0); - /* wait for all outstanding DMAs to finish */ - mfc_write_tag_mask(~0); - mfc_read_tag_status_all(); - /* send mbox message to PPU */ - spu_write_out_mbox(CELL_CMD_FINISH); -} - - -/** - * Execute a batch of commands - * The opcode param encodes the location of the buffer and its size. - */ -static void -cmd_batch(uint opcode) -{ - const uint buf = (opcode >> 8) & 0xff; - uint size = (opcode >> 16); - uint64_t buffer[CELL_BUFFER_SIZE / 8] ALIGN16_ATTRIB; - const unsigned usize = size / sizeof(buffer[0]); - uint pos; - - if (Debug) - printf("SPU %u: BATCH buffer %u, len %u, from %p\n", - spu.init.id, buf, size, spu.init.buffers[buf]); - - ASSERT((opcode & CELL_CMD_OPCODE_MASK) == CELL_CMD_BATCH); - - ASSERT_ALIGN16(spu.init.buffers[buf]); - - size = ROUNDUP16(size); - - ASSERT_ALIGN16(spu.init.buffers[buf]); - - mfc_get(buffer, /* dest */ - (unsigned int) spu.init.buffers[buf], /* src */ - size, - TAG_BATCH_BUFFER, - 0, /* tid */ - 0 /* rid */); - wait_on_mask(1 << TAG_BATCH_BUFFER); - - /* Tell PPU we're done copying the buffer to local store */ - if (Debug) - printf("SPU %u: release batch buf %u\n", spu.init.id, buf); - release_buffer(buf); - - for (pos = 0; pos < usize; /* no incr */) { - switch (buffer[pos]) { - case CELL_CMD_STATE_FRAMEBUFFER: - { - struct cell_command_framebuffer *fb - = (struct cell_command_framebuffer *) &buffer[pos]; - cmd_state_framebuffer(fb); - pos += sizeof(*fb) / 8; - } - break; - case CELL_CMD_CLEAR_SURFACE: - { - struct cell_command_clear_surface *clr - = (struct cell_command_clear_surface *) &buffer[pos]; - cmd_clear_surface(clr); - pos += sizeof(*clr) / 8; - } - break; - case CELL_CMD_RENDER: - { - struct cell_command_render *render - = (struct cell_command_render *) &buffer[pos]; - uint pos_incr; - cmd_render(render, &pos_incr); - pos += pos_incr; - } - break; - case CELL_CMD_RELEASE_VERTS: - { - struct cell_command_release_verts *release - = (struct cell_command_release_verts *) &buffer[pos]; - cmd_release_verts(release); - pos += sizeof(*release) / 8; - } - break; - case CELL_CMD_FINISH: - cmd_finish(); - pos += 1; - break; - case CELL_CMD_STATE_BLEND: - cmd_state_blend((struct pipe_blend_state *) - &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct pipe_blend_state)) / 8); - break; - case CELL_CMD_STATE_DEPTH_STENCIL: - cmd_state_depth_stencil((struct pipe_depth_stencil_alpha_state *) - &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct pipe_depth_stencil_alpha_state)) / 8); - break; - case CELL_CMD_STATE_SAMPLER: - cmd_state_sampler((struct pipe_sampler_state *) &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct pipe_sampler_state)) / 8); - break; - case CELL_CMD_STATE_TEXTURE: - cmd_state_texture((struct cell_command_texture *) &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct cell_command_texture)) / 8); - break; - case CELL_CMD_STATE_VERTEX_INFO: - cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct vertex_info)) / 8); - break; - case CELL_CMD_STATE_VIEWPORT: - (void) memcpy(& draw.viewport, &buffer[pos+1], - sizeof(struct pipe_viewport_state)); - pos += (1 + ROUNDUP8(sizeof(struct pipe_viewport_state)) / 8); - break; - case CELL_CMD_STATE_VS_ARRAY_INFO: - cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct cell_array_info)) / 8); - break; - default: - printf("SPU %u: bad opcode: 0x%llx\n", spu.init.id, buffer[pos]); - ASSERT(0); - break; - } - } - - if (Debug) - printf("SPU %u: BATCH complete\n", spu.init.id); -} - - -/** - * Temporary/simple main loop for SPEs: Get a command, execute it, repeat. - */ -static void -main_loop(void) -{ - struct cell_command cmd; - int exitFlag = 0; - - if (Debug) - printf("SPU %u: Enter main loop\n", spu.init.id); - - ASSERT((sizeof(struct cell_command) & 0xf) == 0); - ASSERT_ALIGN16(&cmd); - - while (!exitFlag) { - unsigned opcode; - int tag = 0; - - if (Debug) - printf("SPU %u: Wait for cmd...\n", spu.init.id); - - /* read/wait from mailbox */ - opcode = (unsigned int) spu_read_in_mbox(); - - if (Debug) - printf("SPU %u: got cmd 0x%x\n", spu.init.id, opcode); - - /* command payload */ - mfc_get(&cmd, /* dest */ - (unsigned int) spu.init.cmd, /* src */ - sizeof(struct cell_command), /* bytes */ - tag, - 0, /* tid */ - 0 /* rid */); - wait_on_mask( 1 << tag ); - - /* - * NOTE: most commands should be contained in a batch buffer - */ - - switch (opcode & CELL_CMD_OPCODE_MASK) { - case CELL_CMD_EXIT: - if (Debug) - printf("SPU %u: EXIT\n", spu.init.id); - exitFlag = 1; - break; - case CELL_CMD_VS_EXECUTE: - spu_execute_vertex_shader(&draw, &cmd.vs); - break; - case CELL_CMD_BATCH: - cmd_batch(opcode); - break; - default: - printf("Bad opcode!\n"); - } - - } - - if (Debug) - printf("SPU %u: Exit main loop\n", spu.init.id); -} - - - -static void -one_time_init(void) -{ - memset(spu.ctile_status, TILE_STATUS_DEFINED, sizeof(spu.ctile_status)); - memset(spu.ztile_status, TILE_STATUS_DEFINED, sizeof(spu.ztile_status)); - invalidate_tex_cache(); -} - - - -/* In some versions of the SDK the SPE main takes 'unsigned long' as a - * parameter. In others it takes 'unsigned long long'. Use a define to - * select between the two. - */ -#ifdef SPU_MAIN_PARAM_LONG_LONG -typedef unsigned long long main_param_t; -#else -typedef unsigned long main_param_t; -#endif - -/** - * SPE entrypoint. - */ -int -main(main_param_t speid, main_param_t argp) -{ - int tag = 0; - - (void) speid; - - ASSERT(sizeof(tile_t) == TILE_SIZE * TILE_SIZE * 4); - ASSERT(sizeof(struct cell_command_render) % 8 == 0); - - one_time_init(); - - if (Debug) - printf("SPU: main() speid=%lu\n", speid); - - mfc_get(&spu.init, /* dest */ - (unsigned int) argp, /* src */ - sizeof(struct cell_init_info), /* bytes */ - tag, - 0, /* tid */ - 0 /* rid */); - wait_on_mask( 1 << tag ); - -#if 0 - if (spu.init.id==0) - spu_test_misc(); -#endif - - main_loop(); - - return 0; -} diff --git a/src/mesa/pipe/cell/spu/spu_main.h b/src/mesa/pipe/cell/spu/spu_main.h deleted file mode 100644 index 1710a17512..0000000000 --- a/src/mesa/pipe/cell/spu/spu_main.h +++ /dev/null @@ -1,177 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SPU_MAIN_H -#define SPU_MAIN_H - - -#include - -#include "pipe/cell/common.h" -#include "pipe/draw/draw_vertex.h" -#include "pipe/p_state.h" - - - -#define MAX_WIDTH 1024 -#define MAX_HEIGHT 1024 - - -typedef union { - ushort us[TILE_SIZE][TILE_SIZE]; - uint ui[TILE_SIZE][TILE_SIZE]; - vector unsigned short us8[TILE_SIZE/2][TILE_SIZE/4]; - vector unsigned int ui4[TILE_SIZE/2][TILE_SIZE/2]; -} tile_t; - - -#define TILE_STATUS_CLEAR 1 -#define TILE_STATUS_DEFINED 2 /**< defined in FB, but not in local store */ -#define TILE_STATUS_CLEAN 3 /**< in local store, but not changed */ -#define TILE_STATUS_DIRTY 4 /**< modified locally, but not put back yet */ -#define TILE_STATUS_GETTING 5 /**< mfc_get() called but not yet arrived */ - - -struct spu_framebuffer { - void *color_start; /**< addr of color surface in main memory */ - void *depth_start; /**< addr of depth surface in main memory */ - enum pipe_format color_format; - enum pipe_format depth_format; - uint width, height; /**< size in pixels */ - uint width_tiles, height_tiles; /**< width and height in tiles */ - - uint color_clear_value; - uint depth_clear_value; - - uint zsize; /**< 0, 2 or 4 bytes per Z */ -} ALIGN16_ATTRIB; - - -/** - * All SPU global/context state will be in singleton object of this type: - */ -struct spu_global -{ - struct cell_init_info init; - - struct spu_framebuffer fb; - struct pipe_blend_state blend_stencil; - struct pipe_depth_stencil_alpha_state depth_stencil; - struct pipe_blend_state blend; - struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; - struct cell_command_texture texture; - - struct vertex_info vertex_info; - - /* XXX more state to come */ - - - /** current color and Z tiles */ - tile_t ctile ALIGN16_ATTRIB; - tile_t ztile ALIGN16_ATTRIB; - - /** Current tiles' status */ - ubyte cur_ctile_status, cur_ztile_status; - - /** Status of all tiles in framebuffer */ - ubyte ctile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - ubyte ztile_status[MAX_HEIGHT/TILE_SIZE][MAX_WIDTH/TILE_SIZE] ALIGN16_ATTRIB; - - - /** for converting RGBA to PIPE_FORMAT_x colors */ - vector unsigned char color_shuffle; - - vector float tex_size; - vector unsigned int tex_size_mask; /**< == int(size - 1) */ - - vector float (*sample_texture)(vector float texcoord); - -} ALIGN16_ATTRIB; - - -extern struct spu_global spu; -extern boolean Debug; - - - - -/* DMA TAGS */ - -#define TAG_SURFACE_CLEAR 10 -#define TAG_VERTEX_BUFFER 11 -#define TAG_READ_TILE_COLOR 12 -#define TAG_READ_TILE_Z 13 -#define TAG_WRITE_TILE_COLOR 14 -#define TAG_WRITE_TILE_Z 15 -#define TAG_INDEX_BUFFER 16 -#define TAG_BATCH_BUFFER 17 -#define TAG_MISC 18 -#define TAG_TEXTURE_TILE 19 -#define TAG_INSTRUCTION_FETCH 20 - - - -static INLINE void -wait_on_mask(unsigned tagMask) -{ - mfc_write_tag_mask( tagMask ); - /* wait for completion of _any_ DMAs specified by tagMask */ - mfc_read_tag_status_any(); -} - - -static INLINE void -wait_on_mask_all(unsigned tagMask) -{ - mfc_write_tag_mask( tagMask ); - /* wait for completion of _any_ DMAs specified by tagMask */ - mfc_read_tag_status_all(); -} - - - - - -static INLINE void -memset16(ushort *d, ushort value, uint count) -{ - uint i; - for (i = 0; i < count; i++) - d[i] = value; -} - - -static INLINE void -memset32(uint *d, uint value, uint count) -{ - uint i; - for (i = 0; i < count; i++) - d[i] = value; -} - - -#endif /* SPU_MAIN_H */ diff --git a/src/mesa/pipe/cell/spu/spu_render.c b/src/mesa/pipe/cell/spu/spu_render.c deleted file mode 100644 index 932fb500b3..0000000000 --- a/src/mesa/pipe/cell/spu/spu_render.c +++ /dev/null @@ -1,301 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include -#include -#include - -#include "spu_main.h" -#include "spu_render.h" -#include "spu_tri.h" -#include "spu_tile.h" -#include "pipe/cell/common.h" - - - -/** - * Given a rendering command's bounding box (in pixels) compute the - * location of the corresponding screen tile bounding box. - */ -static INLINE void -tile_bounding_box(const struct cell_command_render *render, - uint *txmin, uint *tymin, - uint *box_num_tiles, uint *box_width_tiles) -{ -#if 0 - /* Debug: full-window bounding box */ - uint txmax = spu.fb.width_tiles - 1; - uint tymax = spu.fb.height_tiles - 1; - *txmin = 0; - *tymin = 0; - *box_num_tiles = spu.fb.width_tiles * spu.fb.height_tiles; - *box_width_tiles = spu.fb.width_tiles; - (void) render; - (void) txmax; - (void) tymax; -#else - uint txmax, tymax, box_height_tiles; - - *txmin = (uint) render->xmin / TILE_SIZE; - *tymin = (uint) render->ymin / TILE_SIZE; - txmax = (uint) render->xmax / TILE_SIZE; - tymax = (uint) render->ymax / TILE_SIZE; - if (txmax >= spu.fb.width_tiles) - txmax = spu.fb.width_tiles-1; - if (tymax >= spu.fb.height_tiles) - tymax = spu.fb.height_tiles-1; - *box_width_tiles = txmax - *txmin + 1; - box_height_tiles = tymax - *tymin + 1; - *box_num_tiles = *box_width_tiles * box_height_tiles; -#endif -#if 0 - printf("SPU %u: bounds: %g, %g ... %g, %g\n", spu.init.id, - render->xmin, render->ymin, render->xmax, render->ymax); - printf("SPU %u: tiles: %u, %u .. %u, %u\n", - spu.init.id, *txmin, *tymin, txmax, tymax); - ASSERT(render->xmin <= render->xmax); - ASSERT(render->ymin <= render->ymax); -#endif -} - - -/** Check if the tile at (tx,ty) belongs to this SPU */ -static INLINE boolean -my_tile(uint tx, uint ty) -{ - return (spu.fb.width_tiles * ty + tx) % spu.init.num_spus == spu.init.id; -} - - -/** - * Start fetching non-clear color/Z tiles from main memory - */ -static INLINE void -get_cz_tiles(uint tx, uint ty) -{ - if (spu.depth_stencil.depth.enabled) { - if (spu.cur_ztile_status != TILE_STATUS_CLEAR) { - //printf("SPU %u: getting Z tile %u, %u\n", spu.init.id, tx, ty); - get_tile(tx, ty, &spu.ztile, TAG_READ_TILE_Z, 1); - spu.cur_ztile_status = TILE_STATUS_GETTING; - } - } - - if (spu.cur_ctile_status != TILE_STATUS_CLEAR) { - //printf("SPU %u: getting C tile %u, %u\n", spu.init.id, tx, ty); - get_tile(tx, ty, &spu.ctile, TAG_READ_TILE_COLOR, 0); - spu.cur_ctile_status = TILE_STATUS_GETTING; - } -} - - -/** - * Start putting dirty color/Z tiles back to main memory - */ -static INLINE void -put_cz_tiles(uint tx, uint ty) -{ - if (spu.cur_ztile_status == TILE_STATUS_DIRTY) { - /* tile was modified and needs to be written back */ - //printf("SPU %u: put dirty Z tile %u, %u\n", spu.init.id, tx, ty); - put_tile(tx, ty, &spu.ztile, TAG_WRITE_TILE_Z, 1); - spu.cur_ztile_status = TILE_STATUS_DEFINED; - } - else if (spu.cur_ztile_status == TILE_STATUS_GETTING) { - /* tile was never used */ - spu.cur_ztile_status = TILE_STATUS_DEFINED; - //printf("SPU %u: put getting Z tile %u, %u\n", spu.init.id, tx, ty); - } - - if (spu.cur_ctile_status == TILE_STATUS_DIRTY) { - /* tile was modified and needs to be written back */ - //printf("SPU %u: put dirty C tile %u, %u\n", spu.init.id, tx, ty); - put_tile(tx, ty, &spu.ctile, TAG_WRITE_TILE_COLOR, 0); - spu.cur_ctile_status = TILE_STATUS_DEFINED; - } - else if (spu.cur_ctile_status == TILE_STATUS_GETTING) { - /* tile was never used */ - spu.cur_ctile_status = TILE_STATUS_DEFINED; - //printf("SPU %u: put getting C tile %u, %u\n", spu.init.id, tx, ty); - } -} - - -/** - * Wait for 'put' of color/z tiles to complete. - */ -static INLINE void -wait_put_cz_tiles(void) -{ - wait_on_mask(1 << TAG_WRITE_TILE_COLOR); - if (spu.depth_stencil.depth.enabled) { - wait_on_mask(1 << TAG_WRITE_TILE_Z); - } -} - - -/** - * Render primitives - * \param pos_incr returns value indicating how may words to skip after - * this command in the batch buffer - */ -void -cmd_render(const struct cell_command_render *render, uint *pos_incr) -{ - /* we'll DMA into these buffers */ - ubyte vertex_data[CELL_BUFFER_SIZE] ALIGN16_ATTRIB; - const uint vertex_size = render->vertex_size; /* in bytes */ - /*const*/ uint total_vertex_bytes = render->num_verts * vertex_size; - uint index_bytes; - const ubyte *vertices; - const ushort *indexes; - uint i, j; - - - if (Debug) { - printf("SPU %u: RENDER prim %u, num_vert=%u num_ind=%u " - "inline_vert=%u\n", - spu.init.id, - render->prim_type, - render->num_verts, - render->num_indexes, - render->inline_verts); - - /* - printf(" bound: %g, %g .. %g, %g\n", - render->xmin, render->ymin, render->xmax, render->ymax); - */ - } - - ASSERT(sizeof(*render) % 4 == 0); - ASSERT(total_vertex_bytes % 16 == 0); - ASSERT(render->prim_type == PIPE_PRIM_TRIANGLES); - ASSERT(render->num_indexes % 3 == 0); - - - /* indexes are right after the render command in the batch buffer */ - indexes = (const ushort *) (render + 1); - index_bytes = ROUNDUP8(render->num_indexes * 2); - *pos_incr = index_bytes / 8 + sizeof(*render) / 8; - - - if (render->inline_verts) { - /* Vertices are after indexes in batch buffer at next 16-byte addr */ - vertices = (const ubyte *) render + (*pos_incr * 8); - vertices = (const ubyte *) align_pointer((void *) vertices, 16); - ASSERT_ALIGN16(vertices); - *pos_incr = ((vertices + total_vertex_bytes) - (ubyte *) render) / 8; - } - else { - /* Begin DMA fetch of vertex buffer */ - ubyte *src = spu.init.buffers[render->vertex_buf]; - ubyte *dest = vertex_data; - - /* skip vertex data we won't use */ -#if 01 - src += render->min_index * vertex_size; - dest += render->min_index * vertex_size; - total_vertex_bytes -= render->min_index * vertex_size; -#endif - ASSERT(total_vertex_bytes % 16 == 0); - ASSERT_ALIGN16(dest); - ASSERT_ALIGN16(src); - - mfc_get(dest, /* in vertex_data[] array */ - (unsigned int) src, /* src in main memory */ - total_vertex_bytes, /* size */ - TAG_VERTEX_BUFFER, - 0, /* tid */ - 0 /* rid */); - - vertices = vertex_data; - - wait_on_mask(1 << TAG_VERTEX_BUFFER); - } - - - /** - ** find tiles which intersect the prim bounding box - **/ - uint txmin, tymin, box_width_tiles, box_num_tiles; - tile_bounding_box(render, &txmin, &tymin, - &box_num_tiles, &box_width_tiles); - - - /* make sure any pending clears have completed */ - wait_on_mask(1 << TAG_SURFACE_CLEAR); /* XXX temporary */ - - - /** - ** loop over tiles, rendering tris - **/ - for (i = 0; i < box_num_tiles; i++) { - const uint tx = txmin + i % box_width_tiles; - const uint ty = tymin + i / box_width_tiles; - - ASSERT(tx < spu.fb.width_tiles); - ASSERT(ty < spu.fb.height_tiles); - - if (!my_tile(tx, ty)) - continue; - - spu.cur_ctile_status = spu.ctile_status[ty][tx]; - spu.cur_ztile_status = spu.ztile_status[ty][tx]; - - get_cz_tiles(tx, ty); - - uint drawn = 0; - - /* loop over tris */ - for (j = 0; j < render->num_indexes; j += 3) { - const float *v0, *v1, *v2; - - v0 = (const float *) (vertices + indexes[j+0] * vertex_size); - v1 = (const float *) (vertices + indexes[j+1] * vertex_size); - v2 = (const float *) (vertices + indexes[j+2] * vertex_size); - - drawn += tri_draw(v0, v1, v2, tx, ty); - } - - //printf("SPU %u: drew %u of %u\n", spu.init.id, drawn, render->num_indexes/3); - - /* write color/z tiles back to main framebuffer, if dirtied */ - put_cz_tiles(tx, ty); - - wait_put_cz_tiles(); /* XXX seems unnecessary... */ - - spu.ctile_status[ty][tx] = spu.cur_ctile_status; - spu.ztile_status[ty][tx] = spu.cur_ztile_status; - } - - if (Debug) - printf("SPU %u: RENDER done\n", - spu.init.id); -} - - diff --git a/src/mesa/pipe/cell/spu/spu_render.h b/src/mesa/pipe/cell/spu/spu_render.h deleted file mode 100644 index fbcdc5ec31..0000000000 --- a/src/mesa/pipe/cell/spu/spu_render.h +++ /dev/null @@ -1,38 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef SPU_RENDER_H -#define SPU_RENDER_H - -#include "pipe/cell/common.h" - -extern void -cmd_render(const struct cell_command_render *render, uint *pos_incr); - -#endif /* SPU_RENDER_H */ - diff --git a/src/mesa/pipe/cell/spu/spu_texture.c b/src/mesa/pipe/cell/spu/spu_texture.c deleted file mode 100644 index 3962aaa4a9..0000000000 --- a/src/mesa/pipe/cell/spu/spu_texture.c +++ /dev/null @@ -1,217 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "pipe/p_compiler.h" -#include "spu_main.h" -#include "spu_texture.h" -#include "spu_tile.h" -#include "spu_colorpack.h" - - -/** - * Number of texture tiles to cache. - * Note that this will probably be the largest consumer of SPU local store/ - * memory for this driver! - */ -#define CACHE_SIZE 16 - -static tile_t tex_tiles[CACHE_SIZE] ALIGN16_ATTRIB; - -static vector unsigned int tex_tile_xy[CACHE_SIZE]; - - - -/** - * Mark all tex cache entries as invalid. - */ -void -invalidate_tex_cache(void) -{ - /* XXX memset? */ - uint i; - for (i = 0; i < CACHE_SIZE; i++) { - tex_tile_xy[i] = ((vector unsigned int) { ~0U, ~0U, ~0U, ~0U }); - } -} - - -/** - * Return the cache pos/index which corresponds to tile (tx,ty) - */ -static INLINE uint -cache_pos(vector unsigned int txty) -{ - uint pos = (spu_extract(txty,0) + spu_extract(txty,1) * 4) % CACHE_SIZE; - return pos; -} - - -/** - * Make sure the tile for texel (i,j) is present, return its position/index - * in the cache. - */ -static uint -get_tex_tile(vector unsigned int ij) -{ - /* tile address: tx,ty */ - const vector unsigned int txty = spu_rlmask(ij, -5); /* divide by 32 */ - const uint pos = cache_pos(txty); - - if ((spu_extract(tex_tile_xy[pos], 0) != spu_extract(txty, 0)) || - (spu_extract(tex_tile_xy[pos], 1) != spu_extract(txty, 1))) { - - /* texture cache miss, fetch tile from main memory */ - const uint tiles_per_row = spu.texture.width / TILE_SIZE; - const uint bytes_per_tile = sizeof(tile_t); - const void *src = (const ubyte *) spu.texture.start - + (spu_extract(txty,1) * tiles_per_row + spu_extract(txty,0)) * bytes_per_tile; - - printf("SPU %u: tex cache miss at %d, %d pos=%u old=%d,%d\n", - spu.init.id, - spu_extract(txty,0), - spu_extract(txty,1), - pos, - spu_extract(tex_tile_xy[pos],0), - spu_extract(tex_tile_xy[pos],1)); - - ASSERT_ALIGN16(tex_tiles[pos].ui); - ASSERT_ALIGN16(src); - - mfc_get(tex_tiles[pos].ui, /* dest */ - (unsigned int) src, - bytes_per_tile, /* size */ - TAG_TEXTURE_TILE, - 0, /* tid */ - 0 /* rid */); - - wait_on_mask(1 << TAG_TEXTURE_TILE); - - tex_tile_xy[pos] = txty; - } - else { -#if 0 - printf("SPU %u: tex cache HIT at %d, %d\n", - spu.init.id, tx, ty); -#endif - } - - return pos; -} - - -/** - * Get texture sample at texcoord. - * XXX this is extremely primitive for now. - */ -vector float -sample_texture_nearest(vector float texcoord) -{ - vector float tc = spu_mul(texcoord, spu.tex_size); - vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ - itc = spu_and(itc, spu.tex_size_mask); /* mask (GL_REPEAT) */ - vector unsigned int ij = spu_and(itc, TILE_SIZE-1); /* intra tile addr */ - uint pos = get_tex_tile(itc); - uint texel = tex_tiles[pos].ui[spu_extract(ij, 1)][spu_extract(ij, 0)]; - return spu_unpack_A8R8G8B8(texel); -} - - -vector float -sample_texture_bilinear(vector float texcoord) -{ - static const vector unsigned int offset10 = {1, 0, 0, 0}; - static const vector unsigned int offset01 = {0, 1, 0, 0}; - - vector float tc = spu_mul(texcoord, spu.tex_size); - tc = spu_add(tc, spu_splats(-0.5f)); /* half texel bias */ - - /* integer texcoords S,T: */ - vector unsigned int itc00 = spu_convtu(tc, 0); /* convert to int */ - vector unsigned int itc01 = spu_add(itc00, offset01); - vector unsigned int itc10 = spu_add(itc00, offset10); - vector unsigned int itc11 = spu_add(itc10, offset01); - - /* mask (GL_REPEAT) */ - itc00 = spu_and(itc00, spu.tex_size_mask); - itc01 = spu_and(itc01, spu.tex_size_mask); - itc10 = spu_and(itc10, spu.tex_size_mask); - itc11 = spu_and(itc11, spu.tex_size_mask); - - /* intra tile addr */ - vector unsigned int ij00 = spu_and(itc00, TILE_SIZE-1); - vector unsigned int ij01 = spu_and(itc01, TILE_SIZE-1); - vector unsigned int ij10 = spu_and(itc10, TILE_SIZE-1); - vector unsigned int ij11 = spu_and(itc11, TILE_SIZE-1); - - /* get tile cache positions */ - uint pos00 = get_tex_tile(itc00); - uint pos01, pos10, pos11; - if ((spu_extract(ij00, 0) < TILE_SIZE-1) && - (spu_extract(ij00, 1) < TILE_SIZE-1)) { - /* all texels are in the same tile */ - pos01 = pos10 = pos11 = pos00; - } - else { - pos01 = get_tex_tile(itc01); - pos10 = get_tex_tile(itc10); - pos11 = get_tex_tile(itc11); - } - - /* get texels from tiles and convert to float[4] */ - vector float texel00 = spu_unpack_A8R8G8B8(tex_tiles[pos00].ui[spu_extract(ij00, 1)][spu_extract(ij00, 0)]); - vector float texel01 = spu_unpack_A8R8G8B8(tex_tiles[pos01].ui[spu_extract(ij01, 1)][spu_extract(ij01, 0)]); - vector float texel10 = spu_unpack_A8R8G8B8(tex_tiles[pos10].ui[spu_extract(ij10, 1)][spu_extract(ij10, 0)]); - vector float texel11 = spu_unpack_A8R8G8B8(tex_tiles[pos11].ui[spu_extract(ij11, 1)][spu_extract(ij11, 0)]); - - /* Compute weighting factors in [0,1] - * Multiply texcoord by 1024, AND with 1023, convert back to float. - */ - vector float tc1024 = spu_mul(tc, spu_splats(1024.0f)); - vector signed int itc1024 = spu_convts(tc1024, 0); - itc1024 = spu_and(itc1024, spu_splats((1 << 10) - 1)); - vector float weight = spu_convtf(itc1024, 10); - - /* smeared frac and 1-frac */ - vector float sfrac = spu_splats(spu_extract(weight, 0)); - vector float tfrac = spu_splats(spu_extract(weight, 1)); - vector float sfrac1 = spu_sub(spu_splats(1.0f), sfrac); - vector float tfrac1 = spu_sub(spu_splats(1.0f), tfrac); - - /* multiply the samples (colors) by the S/T weights */ - texel00 = spu_mul(spu_mul(texel00, sfrac1), tfrac1); - texel10 = spu_mul(spu_mul(texel10, sfrac ), tfrac1); - texel01 = spu_mul(spu_mul(texel01, sfrac1), tfrac ); - texel11 = spu_mul(spu_mul(texel11, sfrac ), tfrac ); - - /* compute sum of weighted samples */ - vector float texel_sum = spu_add(texel00, texel01); - texel_sum = spu_add(texel_sum, texel10); - texel_sum = spu_add(texel_sum, texel11); - - return texel_sum; -} diff --git a/src/mesa/pipe/cell/spu/spu_texture.h b/src/mesa/pipe/cell/spu/spu_texture.h deleted file mode 100644 index 95eb87080f..0000000000 --- a/src/mesa/pipe/cell/spu/spu_texture.h +++ /dev/null @@ -1,47 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SPU_TEXTURE_H -#define SPU_TEXTURE_H - - -#include "pipe/p_compiler.h" - - -extern void -invalidate_tex_cache(void); - - -extern vector float -sample_texture_nearest(vector float texcoord); - - -extern vector float -sample_texture_bilinear(vector float texcoord); - - -#endif /* SPU_TEXTURE_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tile.c b/src/mesa/pipe/cell/spu/spu_tile.c deleted file mode 100644 index 12dc246328..0000000000 --- a/src/mesa/pipe/cell/spu/spu_tile.c +++ /dev/null @@ -1,83 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - - -#include "spu_tile.h" -#include "spu_main.h" - - -void -get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf) -{ - const uint offset = ty * spu.fb.width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4); - const ubyte *src = zBuf ? spu.fb.depth_start : spu.fb.color_start; - - src += offset * bytesPerTile; - - ASSERT(tx < spu.fb.width_tiles); - ASSERT(ty < spu.fb.height_tiles); - ASSERT_ALIGN16(tile); - /* - printf("get_tile: dest: %p src: 0x%x size: %d\n", - tile, (unsigned int) src, bytesPerTile); - */ - mfc_get(tile->ui, /* dest in local memory */ - (unsigned int) src, /* src in main memory */ - bytesPerTile, - tag, - 0, /* tid */ - 0 /* rid */); -} - - -void -put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf) -{ - const uint offset = ty * spu.fb.width_tiles + tx; - const uint bytesPerTile = TILE_SIZE * TILE_SIZE * (zBuf ? spu.fb.zsize : 4); - ubyte *dst = zBuf ? spu.fb.depth_start : spu.fb.color_start; - - dst += offset * bytesPerTile; - - ASSERT(tx < spu.fb.width_tiles); - ASSERT(ty < spu.fb.height_tiles); - ASSERT_ALIGN16(tile); - /* - printf("SPU %u: put_tile: src: %p dst: 0x%x size: %d\n", - spu.init.id, - tile, (unsigned int) dst, bytesPerTile); - */ - mfc_put((void *) tile->ui, /* src in local memory */ - (unsigned int) dst, /* dst in main memory */ - bytesPerTile, - tag, - 0, /* tid */ - 0 /* rid */); -} - diff --git a/src/mesa/pipe/cell/spu/spu_tile.h b/src/mesa/pipe/cell/spu/spu_tile.h deleted file mode 100644 index e53340a55a..0000000000 --- a/src/mesa/pipe/cell/spu/spu_tile.h +++ /dev/null @@ -1,73 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SPU_TILE_H -#define SPU_TILE_H - - -#include -#include -#include "spu_main.h" -#include "pipe/cell/common.h" - - - -void -get_tile(uint tx, uint ty, tile_t *tile, int tag, int zBuf); - -void -put_tile(uint tx, uint ty, const tile_t *tile, int tag, int zBuf); - - - -static INLINE void -clear_c_tile(tile_t *ctile) -{ - memset32((uint*) ctile->ui, - spu.fb.color_clear_value, - TILE_SIZE * TILE_SIZE); -} - - -static INLINE void -clear_z_tile(tile_t *ztile) -{ - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - memset16((ushort*) ztile->us, - spu.fb.depth_clear_value, - TILE_SIZE * TILE_SIZE); - } - else { - ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); - memset32((uint*) ztile->ui, - spu.fb.depth_clear_value, - TILE_SIZE * TILE_SIZE); - } -} - - -#endif /* SPU_TILE_H */ diff --git a/src/mesa/pipe/cell/spu/spu_tri.c b/src/mesa/pipe/cell/spu/spu_tri.c deleted file mode 100644 index be9624cf7d..0000000000 --- a/src/mesa/pipe/cell/spu/spu_tri.c +++ /dev/null @@ -1,926 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Triangle rendering within a tile. - */ - -#include "pipe/p_compiler.h" -#include "pipe/p_format.h" -#include "pipe/p_util.h" -#include "spu_blend.h" -#include "spu_colorpack.h" -#include "spu_main.h" -#include "spu_texture.h" -#include "spu_tile.h" -#include "spu_tri.h" - -#include "spu_ztest.h" - - -/** Masks are uint[4] vectors with each element being 0 or 0xffffffff */ -typedef vector unsigned int mask_t; - -typedef union -{ - vector float v; - float f[4]; -} float4; - - -/** - * Simplified types taken from other parts of Gallium - */ -struct vertex_header { - vector float data[1]; -}; - - - -/* XXX fix this */ -#undef CEILF -#define CEILF(X) ((float) (int) ((X) + 0.99999)) - - -#define QUAD_TOP_LEFT 0 -#define QUAD_TOP_RIGHT 1 -#define QUAD_BOTTOM_LEFT 2 -#define QUAD_BOTTOM_RIGHT 3 -#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) -#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) -#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) -#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) -#define MASK_ALL 0xf - - -#define DEBUG_VERTS 0 - -/** - * Triangle edge info - */ -struct edge { - float dx; /**< X(v1) - X(v0), used only during setup */ - float dy; /**< Y(v1) - Y(v0), used only during setup */ - float dxdy; /**< dx/dy */ - float sx, sy; /**< first sample point coord */ - int lines; /**< number of lines on this edge */ -}; - - -struct interp_coef -{ - float4 a0; - float4 dadx; - float4 dady; -}; - - -/** - * Triangle setup info (derived from draw_stage). - * Also used for line drawing (taking some liberties). - */ -struct setup_stage { - - /* Vertices are just an array of floats making up each attribute in - * turn. Currently fixed at 4 floats, but should change in time. - * Codegen will help cope with this. - */ - const struct vertex_header *vmax; - const struct vertex_header *vmid; - const struct vertex_header *vmin; - const struct vertex_header *vprovoke; - - struct edge ebot; - struct edge etop; - struct edge emaj; - - float oneoverarea; - - uint tx, ty; - - int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; - -#if 0 - struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; -#else - struct interp_coef coef[PIPE_MAX_SHADER_INPUTS]; -#endif - -#if 0 - struct quad_header quad; -#endif - - struct { - int left[2]; /**< [0] = row0, [1] = row1 */ - int right[2]; - int y; - unsigned y_flags; - unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ - } span; -}; - - - -static struct setup_stage setup; - - - - -#if 0 -/** - * Basically a cast wrapper. - */ -static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) -{ - return (struct setup_stage *)stage; -} -#endif - -#if 0 -/** - * Clip setup.quad against the scissor/surface bounds. - */ -static INLINE void -quad_clip(struct setup_stage *setup) -{ - const struct pipe_scissor_state *cliprect = &setup.softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; - - if (setup.quad.x0 >= maxx || - setup.quad.y0 >= maxy || - setup.quad.x0 + 1 < minx || - setup.quad.y0 + 1 < miny) { - /* totally clipped */ - setup.quad.mask = 0x0; - return; - } - if (setup.quad.x0 < minx) - setup.quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup.quad.y0 < miny) - setup.quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - if (setup.quad.x0 == maxx - 1) - setup.quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup.quad.y0 == maxy - 1) - setup.quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); -} -#endif - -#if 0 -/** - * Emit a quad (pass to next stage) with clipping. - */ -static INLINE void -clip_emit_quad(struct setup_stage *setup) -{ - quad_clip(setup); - if (setup.quad.mask) { - struct softpipe_context *sp = setup.softpipe; - sp->quad.first->run(sp->quad.first, &setup.quad); - } -} -#endif - -/** - * Evaluate attribute coefficients (plane equations) to compute - * attribute values for the four fragments in a quad. - * Eg: four colors will be compute. - */ -static INLINE void -eval_coeff(uint slot, float x, float y, vector float result[4]) -{ - switch (spu.vertex_info.interp_mode[slot]) { - case INTERP_CONSTANT: - result[QUAD_TOP_LEFT] = - result[QUAD_TOP_RIGHT] = - result[QUAD_BOTTOM_LEFT] = - result[QUAD_BOTTOM_RIGHT] = setup.coef[slot].a0.v; - break; - - case INTERP_LINEAR: - /* fall-through, for now */ - default: - { - register vector float dadx = setup.coef[slot].dadx.v; - register vector float dady = setup.coef[slot].dady.v; - register vector float topLeft - = spu_add(setup.coef[slot].a0.v, - spu_add(spu_mul(spu_splats(x), dadx), - spu_mul(spu_splats(y), dady))); - - result[QUAD_TOP_LEFT] = topLeft; - result[QUAD_TOP_RIGHT] = spu_add(topLeft, dadx); - result[QUAD_BOTTOM_LEFT] = spu_add(topLeft, dady); - result[QUAD_BOTTOM_RIGHT] = spu_add(spu_add(topLeft, dadx), dady); - } - } -} - - -static INLINE vector float -eval_z(float x, float y) -{ - const uint slot = 0; - const float dzdx = setup.coef[slot].dadx.f[2]; - const float dzdy = setup.coef[slot].dady.f[2]; - const float topLeft = setup.coef[slot].a0.f[2] + x * dzdx + y * dzdy; - const vector float topLeftv = spu_splats(topLeft); - const vector float derivs = (vector float) { 0.0, dzdx, dzdy, dzdx + dzdy }; - return spu_add(topLeftv, derivs); -} - - -static INLINE mask_t -do_depth_test(int x, int y, mask_t quadmask) -{ - float4 zvals; - mask_t mask; - - zvals.v = eval_z((float) x, (float) y); - - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - int ix = (x - setup.cliprect_minx) / 4; - int iy = (y - setup.cliprect_miny) / 2; - mask = spu_z16_test_less(zvals.v, &spu.ztile.us8[iy][ix], x>>1, quadmask); - } - else { - int ix = (x - setup.cliprect_minx) / 2; - int iy = (y - setup.cliprect_miny) / 2; - mask = spu_z32_test_less(zvals.v, &spu.ztile.ui4[iy][ix], quadmask); - } - - if (spu_extract(spu_orx(mask), 0)) - spu.cur_ztile_status = TILE_STATUS_DIRTY; - - return mask; -} - - -/** - * Emit a quad (pass to next stage). No clipping is done. - * Note: about 1/5 to 1/7 of the time, mask is zero and this function - * should be skipped. But adding the test for that slows things down - * overall. - */ -static INLINE void -emit_quad( int x, int y, mask_t mask ) -{ -#if 0 - struct softpipe_context *sp = setup.softpipe; - setup.quad.x0 = x; - setup.quad.y0 = y; - setup.quad.mask = mask; - sp->quad.first->run(sp->quad.first, &setup.quad); -#else - - if (spu.depth_stencil.depth.enabled) { - mask = do_depth_test(x, y, mask); - } - - /* If any bits in mask are set... */ - if (spu_extract(spu_orx(mask), 0)) { - const int ix = x - setup.cliprect_minx; - const int iy = y - setup.cliprect_miny; - const vector unsigned char shuffle = spu.color_shuffle; - vector float colors[4]; - - spu.cur_ctile_status = TILE_STATUS_DIRTY; - - if (spu.texture.start) { - /* texture mapping */ - vector float texcoords[4]; - eval_coeff(2, (float) x, (float) y, texcoords); - - if (spu_extract(mask, 0)) - colors[0] = spu.sample_texture(texcoords[0]); - if (spu_extract(mask, 1)) - colors[1] = spu.sample_texture(texcoords[1]); - if (spu_extract(mask, 2)) - colors[2] = spu.sample_texture(texcoords[2]); - if (spu_extract(mask, 3)) - colors[3] = spu.sample_texture(texcoords[3]); - } - else { - /* simple shading */ - eval_coeff(1, (float) x, (float) y, colors); - } - -#if 1 - if (spu.blend.blend_enable) - blend_quad(ix % TILE_SIZE, iy % TILE_SIZE, colors); -#endif - - if (spu_extract(mask, 0)) - spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0], shuffle); - if (spu_extract(mask, 1)) - spu.ctile.ui[iy][ix+1] = spu_pack_color_shuffle(colors[1], shuffle); - if (spu_extract(mask, 2)) - spu.ctile.ui[iy+1][ix] = spu_pack_color_shuffle(colors[2], shuffle); - if (spu_extract(mask, 3)) - spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(colors[3], shuffle); - -#if 0 - /* SIMD_Z with swizzled color buffer (someday) */ - vector unsigned int uicolors = *((vector unsigned int *) &colors); - spu.ctile.ui4[iy/2][ix/2] = spu_sel(spu.ctile.ui4[iy/2][ix/2], uicolors, mask); -#endif - } - -#endif -} - - -/** - * Given an X or Y coordinate, return the block/quad coordinate that it - * belongs to. - */ -static INLINE int block( int x ) -{ - return x & ~1; -} - - -/** - * Compute mask which indicates which pixels in the 2x2 quad are actually inside - * the triangle's bounds. - * The mask is a uint4 vector and each element will be 0 or 0xffffffff. - */ -static INLINE mask_t calculate_mask( int x ) -{ - /* This is a little tricky. - * Use & instead of && to avoid branches. - * Use negation to convert true/false to ~0/0 values. - */ - mask_t mask; - mask = spu_insert(-((x >= setup.span.left[0]) & (x < setup.span.right[0])), mask, 0); - mask = spu_insert(-((x+1 >= setup.span.left[0]) & (x+1 < setup.span.right[0])), mask, 1); - mask = spu_insert(-((x >= setup.span.left[1]) & (x < setup.span.right[1])), mask, 2); - mask = spu_insert(-((x+1 >= setup.span.left[1]) & (x+1 < setup.span.right[1])), mask, 3); - return mask; -} - - -/** - * Render a horizontal span of quads - */ -static void flush_spans( void ) -{ - int minleft, maxright; - int x; - - switch (setup.span.y_flags) { - case 0x3: - /* both odd and even lines written (both quad rows) */ - minleft = MIN2(setup.span.left[0], setup.span.left[1]); - maxright = MAX2(setup.span.right[0], setup.span.right[1]); - break; - - case 0x1: - /* only even line written (quad top row) */ - minleft = setup.span.left[0]; - maxright = setup.span.right[0]; - break; - - case 0x2: - /* only odd line written (quad bottom row) */ - minleft = setup.span.left[1]; - maxright = setup.span.right[1]; - break; - - default: - return; - } - - - /* OK, we're very likely to need the tile data now. - * clear or finish waiting if needed. - */ - if (spu.cur_ctile_status == TILE_STATUS_GETTING) { - /* wait for mfc_get() to complete */ - //printf("SPU: %u: waiting for ctile\n", spu.init.id); - wait_on_mask(1 << TAG_READ_TILE_COLOR); - spu.cur_ctile_status = TILE_STATUS_CLEAN; - } - else if (spu.cur_ctile_status == TILE_STATUS_CLEAR) { - //printf("SPU %u: clearing C tile %u, %u\n", spu.init.id, setup.tx, setup.ty); - clear_c_tile(&spu.ctile); - spu.cur_ctile_status = TILE_STATUS_DIRTY; - } - ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED); - - if (spu.depth_stencil.depth.enabled) { - if (spu.cur_ztile_status == TILE_STATUS_GETTING) { - /* wait for mfc_get() to complete */ - //printf("SPU: %u: waiting for ztile\n", spu.init.id); - wait_on_mask(1 << TAG_READ_TILE_Z); - spu.cur_ztile_status = TILE_STATUS_CLEAN; - } - else if (spu.cur_ztile_status == TILE_STATUS_CLEAR) { - //printf("SPU %u: clearing Z tile %u, %u\n", spu.init.id, setup.tx, setup.ty); - clear_z_tile(&spu.ztile); - spu.cur_ztile_status = TILE_STATUS_DIRTY; - } - ASSERT(spu.cur_ztile_status != TILE_STATUS_DEFINED); - } - - /* XXX this loop could be moved into the above switch cases and - * calculate_mask() could be simplified a bit... - */ - for (x = block(minleft); x <= block(maxright); x += 2) { -#if 1 - emit_quad( x, setup.span.y, calculate_mask( x ) ); -#endif - } - - setup.span.y = 0; - setup.span.y_flags = 0; - setup.span.right[0] = 0; - setup.span.right[1] = 0; -} - -#if DEBUG_VERTS -static void print_vertex(const struct vertex_header *v) -{ - int i; - fprintf(stderr, "Vertex: (%p)\n", v); - for (i = 0; i < setup.quad.nr_attrs; i++) { - fprintf(stderr, " %d: %f %f %f %f\n", i, - v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); - } -} -#endif - - -static boolean setup_sort_vertices(const struct vertex_header *v0, - const struct vertex_header *v1, - const struct vertex_header *v2) -{ - -#if DEBUG_VERTS - fprintf(stderr, "Triangle:\n"); - print_vertex(v0); - print_vertex(v1); - print_vertex(v2); -#endif - - setup.vprovoke = v2; - - /* determine bottom to top order of vertices */ - { - float y0 = spu_extract(v0->data[0], 1); - float y1 = spu_extract(v1->data[0], 1); - float y2 = spu_extract(v2->data[0], 1); - if (y0 <= y1) { - if (y1 <= y2) { - /* y0<=y1<=y2 */ - setup.vmin = v0; - setup.vmid = v1; - setup.vmax = v2; - } - else if (y2 <= y0) { - /* y2<=y0<=y1 */ - setup.vmin = v2; - setup.vmid = v0; - setup.vmax = v1; - } - else { - /* y0<=y2<=y1 */ - setup.vmin = v0; - setup.vmid = v2; - setup.vmax = v1; - } - } - else { - if (y0 <= y2) { - /* y1<=y0<=y2 */ - setup.vmin = v1; - setup.vmid = v0; - setup.vmax = v2; - } - else if (y2 <= y1) { - /* y2<=y1<=y0 */ - setup.vmin = v2; - setup.vmid = v1; - setup.vmax = v0; - } - else { - /* y1<=y2<=y0 */ - setup.vmin = v1; - setup.vmid = v2; - setup.vmax = v0; - } - } - } - - /* Check if triangle is completely outside the tile bounds */ - if (spu_extract(setup.vmin->data[0], 1) > setup.cliprect_maxy) - return FALSE; - if (spu_extract(setup.vmax->data[0], 1) < setup.cliprect_miny) - return FALSE; - if (spu_extract(setup.vmin->data[0], 0) < setup.cliprect_minx && - spu_extract(setup.vmid->data[0], 0) < setup.cliprect_minx && - spu_extract(setup.vmax->data[0], 0) < setup.cliprect_minx) - return FALSE; - if (spu_extract(setup.vmin->data[0], 0) > setup.cliprect_maxx && - spu_extract(setup.vmid->data[0], 0) > setup.cliprect_maxx && - spu_extract(setup.vmax->data[0], 0) > setup.cliprect_maxx) - return FALSE; - - setup.ebot.dx = spu_extract(setup.vmid->data[0], 0) - spu_extract(setup.vmin->data[0], 0); - setup.ebot.dy = spu_extract(setup.vmid->data[0], 1) - spu_extract(setup.vmin->data[0], 1); - setup.emaj.dx = spu_extract(setup.vmax->data[0], 0) - spu_extract(setup.vmin->data[0], 0); - setup.emaj.dy = spu_extract(setup.vmax->data[0], 1) - spu_extract(setup.vmin->data[0], 1); - setup.etop.dx = spu_extract(setup.vmax->data[0], 0) - spu_extract(setup.vmid->data[0], 0); - setup.etop.dy = spu_extract(setup.vmax->data[0], 1) - spu_extract(setup.vmid->data[0], 1); - - /* - * Compute triangle's area. Use 1/area to compute partial - * derivatives of attributes later. - * - * The area will be the same as prim->det, but the sign may be - * different depending on how the vertices get sorted above. - * - * To determine whether the primitive is front or back facing we - * use the prim->det value because its sign is correct. - */ - { - const float area = (setup.emaj.dx * setup.ebot.dy - - setup.ebot.dx * setup.emaj.dy); - - setup.oneoverarea = 1.0f / area; - /* - _mesa_printf("%s one-over-area %f area %f det %f\n", - __FUNCTION__, setup.oneoverarea, area, prim->det ); - */ - } - -#if 0 - /* We need to know if this is a front or back-facing triangle for: - * - the GLSL gl_FrontFacing fragment attribute (bool) - * - two-sided stencil test - */ - setup.quad.facing = (prim->det > 0.0) ^ (setup.softpipe->rasterizer->front_winding == PIPE_WINDING_CW); -#endif - - return TRUE; -} - - -/** - * Compute a0 for a constant-valued coefficient (GL_FLAT shading). - * The value value comes from vertex->data[slot]. - * The result will be put into setup.coef[slot].a0. - * \param slot which attribute slot - */ -static INLINE void -const_coeff(uint slot) -{ - setup.coef[slot].dadx.v = (vector float) {0.0, 0.0, 0.0, 0.0}; - setup.coef[slot].dady.v = (vector float) {0.0, 0.0, 0.0, 0.0}; - setup.coef[slot].a0.v = setup.vprovoke->data[slot]; -} - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a triangle. - */ -static INLINE void -tri_linear_coeff(uint slot, uint firstComp, uint lastComp) -{ - uint i; - const float *vmin_d = (float *) &setup.vmin->data[slot]; - const float *vmid_d = (float *) &setup.vmid->data[slot]; - const float *vmax_d = (float *) &setup.vmax->data[slot]; - const float x = spu_extract(setup.vmin->data[0], 0) - 0.5f; - const float y = spu_extract(setup.vmin->data[0], 1) - 0.5f; - - for (i = firstComp; i < lastComp; i++) { - float botda = vmid_d[i] - vmin_d[i]; - float majda = vmax_d[i] - vmin_d[i]; - float a = setup.ebot.dy * majda - botda * setup.emaj.dy; - float b = setup.emaj.dx * botda - majda * setup.ebot.dx; - - ASSERT(slot < PIPE_MAX_SHADER_INPUTS); - - setup.coef[slot].dadx.f[i] = a * setup.oneoverarea; - setup.coef[slot].dady.f[i] = b * setup.oneoverarea; - - /* calculate a0 as the value which would be sampled for the - * fragment at (0,0), taking into account that we want to sample at - * pixel centers, in other words (0.5, 0.5). - * - * this is neat but unfortunately not a good way to do things for - * triangles with very large values of dadx or dady as it will - * result in the subtraction and re-addition from a0 of a very - * large number, which means we'll end up loosing a lot of the - * fractional bits and precision from a0. the way to fix this is - * to define a0 as the sample at a pixel center somewhere near vmin - * instead - i'll switch to this later. - */ - setup.coef[slot].a0.f[i] = (vmin_d[i] - - (setup.coef[slot].dadx.f[i] * x + - setup.coef[slot].dady.f[i] * y)); - } - - /* - _mesa_printf("attr[%d].%c: %f dx:%f dy:%f\n", - slot, "xyzw"[i], - setup.coef[slot].a0[i], - setup.coef[slot].dadx.f[i], - setup.coef[slot].dady.f[i]); - */ -} - - -/** - * As above, but interp setup all four vector components. - */ -static INLINE void -tri_linear_coeff4(uint slot) -{ - const vector float vmin_d = setup.vmin->data[slot]; - const vector float vmid_d = setup.vmid->data[slot]; - const vector float vmax_d = setup.vmax->data[slot]; - const vector float xxxx = spu_splats(spu_extract(setup.vmin->data[0], 0) - 0.5f); - const vector float yyyy = spu_splats(spu_extract(setup.vmin->data[0], 1) - 0.5f); - - vector float botda = vmid_d - vmin_d; - vector float majda = vmax_d - vmin_d; - - vector float a = spu_sub(spu_mul(spu_splats(setup.ebot.dy), majda), - spu_mul(botda, spu_splats(setup.emaj.dy))); - vector float b = spu_sub(spu_mul(spu_splats(setup.emaj.dx), botda), - spu_mul(majda, spu_splats(setup.ebot.dx))); - - setup.coef[slot].dadx.v = spu_mul(a, spu_splats(setup.oneoverarea)); - setup.coef[slot].dady.v = spu_mul(b, spu_splats(setup.oneoverarea)); - - vector float tempx = spu_mul(setup.coef[slot].dadx.v, xxxx); - vector float tempy = spu_mul(setup.coef[slot].dady.v, yyyy); - - setup.coef[slot].a0.v = spu_sub(vmin_d, spu_add(tempx, tempy)); -} - - - -#if 0 -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a triangle. - * We basically multiply the vertex value by 1/w before computing - * the plane coefficients (a0, dadx, dady). - * Later, when we compute the value at a particular fragment position we'll - * divide the interpolated value by the interpolated W at that fragment. - */ -static void tri_persp_coeff( unsigned slot, - unsigned i ) -{ - /* premultiply by 1/w: - */ - float mina = setup.vmin->data[slot][i] * setup.vmin->data[0][3]; - float mida = setup.vmid->data[slot][i] * setup.vmid->data[0][3]; - float maxa = setup.vmax->data[slot][i] * setup.vmax->data[0][3]; - - float botda = mida - mina; - float majda = maxa - mina; - float a = setup.ebot.dy * majda - botda * setup.emaj.dy; - float b = setup.emaj.dx * botda - majda * setup.ebot.dx; - - /* - printf("tri persp %d,%d: %f %f %f\n", slot, i, - setup.vmin->data[slot][i], - setup.vmid->data[slot][i], - setup.vmax->data[slot][i] - ); - */ - - assert(slot < PIPE_MAX_SHADER_INPUTS); - assert(i <= 3); - - setup.coef[slot].dadx.f[i] = a * setup.oneoverarea; - setup.coef[slot].dady.f[i] = b * setup.oneoverarea; - setup.coef[slot].a0.f[i] = (mina - - (setup.coef[slot].dadx.f[i] * (setup.vmin->data[0][0] - 0.5f) + - setup.coef[slot].dady.f[i] * (setup.vmin->data[0][1] - 0.5f))); -} -#endif - - -/** - * Compute the setup.coef[] array dadx, dady, a0 values. - * Must be called after setup.vmin,vmid,vmax,vprovoke are initialized. - */ -static void setup_tri_coefficients(void) -{ -#if 1 - uint i; - - for (i = 0; i < spu.vertex_info.num_attribs; i++) { - switch (spu.vertex_info.interp_mode[i]) { - case INTERP_NONE: - break; - case INTERP_POS: - /*tri_linear_coeff(i, 2, 3);*/ - /* XXX interp W if PERSPECTIVE... */ - tri_linear_coeff4(i); - break; - case INTERP_CONSTANT: - const_coeff(i); - break; - case INTERP_LINEAR: - tri_linear_coeff4(i); - break; - case INTERP_PERSPECTIVE: - tri_linear_coeff4(i); /* temporary */ - break; - default: - ASSERT(0); - } - } -#else - ASSERT(spu.vertex_info.interp_mode[0] == INTERP_POS); - ASSERT(spu.vertex_info.interp_mode[1] == INTERP_LINEAR || - spu.vertex_info.interp_mode[1] == INTERP_CONSTANT); - tri_linear_coeff(0, 2, 3); /* slot 0, z */ - tri_linear_coeff(1, 0, 4); /* slot 1, color */ -#endif -} - - -static void setup_tri_edges(void) -{ - float vmin_x = spu_extract(setup.vmin->data[0], 0) + 0.5f; - float vmid_x = spu_extract(setup.vmid->data[0], 0) + 0.5f; - - float vmin_y = spu_extract(setup.vmin->data[0], 1) - 0.5f; - float vmid_y = spu_extract(setup.vmid->data[0], 1) - 0.5f; - float vmax_y = spu_extract(setup.vmax->data[0], 1) - 0.5f; - - setup.emaj.sy = CEILF(vmin_y); - setup.emaj.lines = (int) CEILF(vmax_y - setup.emaj.sy); - setup.emaj.dxdy = setup.emaj.dx / setup.emaj.dy; - setup.emaj.sx = vmin_x + (setup.emaj.sy - vmin_y) * setup.emaj.dxdy; - - setup.etop.sy = CEILF(vmid_y); - setup.etop.lines = (int) CEILF(vmax_y - setup.etop.sy); - setup.etop.dxdy = setup.etop.dx / setup.etop.dy; - setup.etop.sx = vmid_x + (setup.etop.sy - vmid_y) * setup.etop.dxdy; - - setup.ebot.sy = CEILF(vmin_y); - setup.ebot.lines = (int) CEILF(vmid_y - setup.ebot.sy); - setup.ebot.dxdy = setup.ebot.dx / setup.ebot.dy; - setup.ebot.sx = vmin_x + (setup.ebot.sy - vmin_y) * setup.ebot.dxdy; -} - - -/** - * Render the upper or lower half of a triangle. - * Scissoring/cliprect is applied here too. - */ -static void subtriangle( struct edge *eleft, - struct edge *eright, - unsigned lines ) -{ - const int minx = setup.cliprect_minx; - const int maxx = setup.cliprect_maxx; - const int miny = setup.cliprect_miny; - const int maxy = setup.cliprect_maxy; - int y, start_y, finish_y; - int sy = (int)eleft->sy; - - ASSERT((int)eleft->sy == (int) eright->sy); - - /* clip top/bottom */ - start_y = sy; - finish_y = sy + lines; - - if (start_y < miny) - start_y = miny; - - if (finish_y > maxy) - finish_y = maxy; - - start_y -= sy; - finish_y -= sy; - - /* - _mesa_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); - */ - - for (y = start_y; y < finish_y; y++) { - - /* avoid accumulating adds as floats don't have the precision to - * accurately iterate large triangle edges that way. luckily we - * can just multiply these days. - * - * this is all drowned out by the attribute interpolation anyway. - */ - int left = (int)(eleft->sx + y * eleft->dxdy); - int right = (int)(eright->sx + y * eright->dxdy); - - /* clip left/right */ - if (left < minx) - left = minx; - if (right > maxx) - right = maxx; - - if (left < right) { - int _y = sy + y; - if (block(_y) != setup.span.y) { - flush_spans(); - setup.span.y = block(_y); - } - - setup.span.left[_y&1] = left; - setup.span.right[_y&1] = right; - setup.span.y_flags |= 1<<(_y&1); - } - } - - - /* save the values so that emaj can be restarted: - */ - eleft->sx += lines * eleft->dxdy; - eright->sx += lines * eright->dxdy; - eleft->sy += lines; - eright->sy += lines; -} - - -/** - * Draw triangle into tile at (tx, ty) (tile coords) - * The tile data should have already been fetched. - */ -boolean -tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) -{ - setup.tx = tx; - setup.ty = ty; - - /* set clipping bounds to tile bounds */ - setup.cliprect_minx = tx * TILE_SIZE; - setup.cliprect_miny = ty * TILE_SIZE; - setup.cliprect_maxx = (tx + 1) * TILE_SIZE; - setup.cliprect_maxy = (ty + 1) * TILE_SIZE; - - if (!setup_sort_vertices((struct vertex_header *) v0, - (struct vertex_header *) v1, - (struct vertex_header *) v2)) { - return FALSE; /* totally clipped */ - } - - setup_tri_coefficients(); - setup_tri_edges(); - - setup.span.y = 0; - setup.span.y_flags = 0; - setup.span.right[0] = 0; - setup.span.right[1] = 0; - /* setup.span.z_mode = tri_z_mode( setup.ctx ); */ - - /* init_constant_attribs( setup ); */ - - if (setup.oneoverarea < 0.0) { - /* emaj on left: - */ - subtriangle( &setup.emaj, &setup.ebot, setup.ebot.lines ); - subtriangle( &setup.emaj, &setup.etop, setup.etop.lines ); - } - else { - /* emaj on right: - */ - subtriangle( &setup.ebot, &setup.emaj, setup.ebot.lines ); - subtriangle( &setup.etop, &setup.emaj, setup.etop.lines ); - } - - flush_spans(); - - return TRUE; -} diff --git a/src/mesa/pipe/cell/spu/spu_tri.h b/src/mesa/pipe/cell/spu/spu_tri.h deleted file mode 100644 index aa694dd7c9..0000000000 --- a/src/mesa/pipe/cell/spu/spu_tri.h +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef SPU_TRI_H -#define SPU_TRI_H - - -extern boolean -tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty); - - -#endif /* SPU_TRI_H */ diff --git a/src/mesa/pipe/cell/spu/spu_util.c b/src/mesa/pipe/cell/spu/spu_util.c deleted file mode 100644 index ac373240c1..0000000000 --- a/src/mesa/pipe/cell/spu/spu_util.c +++ /dev/null @@ -1,165 +0,0 @@ -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -//#include "tgsi_build.h" -#include "pipe/tgsi/util/tgsi_util.h" - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->SwizzleX; - case 1: - return reg->SwizzleY; - case 2: - return reg->SwizzleZ; - case 3: - return reg->SwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->ExtSwizzleX; - case 1: - return reg->ExtSwizzleY; - case 2: - return reg->ExtSwizzleZ; - case 3: - return reg->ExtSwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned swizzle; - - /* - * First, calculate the extended swizzle for a given channel. This will give - * us either a channel index into the simple swizzle or a constant 1 or 0. - */ - swizzle = tgsi_util_get_src_register_extswizzle( - ®->SrcRegisterExtSwz, - component ); - - assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); - assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); - assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); - assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); - - /* - * Second, calculate the simple swizzle for the unswizzled channel index. - * Leave the constants intact, they are not affected by the simple swizzle. - */ - if( swizzle <= TGSI_SWIZZLE_W ) { - swizzle = tgsi_util_get_src_register_swizzle( - ®->SrcRegister, - component ); - } - - return swizzle; -} - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->NegateX; - case 1: - return reg->NegateY; - case 2: - return reg->NegateZ; - case 3: - return reg->NegateW; - default: - assert( 0 ); - } - return 0; -} - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ) -{ - switch( component ) { - case 0: - reg->NegateX = negate; - break; - case 1: - reg->NegateY = negate; - break; - case 2: - reg->NegateZ = negate; - break; - case 3: - reg->NegateW = negate; - break; - default: - assert( 0 ); - } -} - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned sign_mode; - - if( reg->SrcRegisterExtMod.Absolute ) { - /* Consider only the post-abs negation. */ - - if( reg->SrcRegisterExtMod.Negate ) { - sign_mode = TGSI_UTIL_SIGN_SET; - } - else { - sign_mode = TGSI_UTIL_SIGN_CLEAR; - } - } - else { - /* Accumulate the three negations. */ - - unsigned negate; - - negate = reg->SrcRegister.Negate; - if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { - negate = !negate; - } - if( reg->SrcRegisterExtMod.Negate ) { - negate = !negate; - } - - if( negate ) { - sign_mode = TGSI_UTIL_SIGN_TOGGLE; - } - else { - sign_mode = TGSI_UTIL_SIGN_KEEP; - } - } - - return sign_mode; -} diff --git a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c b/src/mesa/pipe/cell/spu/spu_vertex_fetch.c deleted file mode 100644 index 45e3c26c00..0000000000 --- a/src/mesa/pipe/cell/spu/spu_vertex_fetch.c +++ /dev/null @@ -1,673 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include -#include - -#include "pipe/p_util.h" -#include "pipe/p_state.h" -#include "pipe/p_shader_tokens.h" -#include "spu_exec.h" -#include "spu_vertex_shader.h" -#include "spu_main.h" - -#define CACHE_NAME attribute -#define CACHED_TYPE qword -#define CACHE_TYPE CACHE_TYPE_RO -#define CACHE_SET_TAGID(set) TAG_VERTEX_BUFFER -#define CACHE_LOG2NNWAY 2 -#define CACHE_LOG2NSETS 6 -#include - -/* Yes folks, this is ugly. - */ -#undef CACHE_NWAY -#undef CACHE_NSETS -#define CACHE_NAME attribute -#define CACHE_NWAY 4 -#define CACHE_NSETS (1U << 6) - - -#define DRAW_DBG 0 - -static const qword fetch_shuffle_data[] = { - /* Shuffle used by CVT_64_FLOAT - */ - { - 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, - 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, 0x80, - }, - - /* Shuffle used by CVT_8_USCALED and CVT_8_SSCALED - */ - { - 0x00, 0x80, 0x80, 0x80, 0x01, 0x80, 0x80, 0x80, - 0x02, 0x80, 0x80, 0x80, 0x03, 0x80, 0x80, 0x80, - }, - - /* Shuffle used by CVT_16_USCALED and CVT_16_SSCALED - */ - { - 0x00, 0x01, 0x80, 0x80, 0x02, 0x03, 0x80, 0x80, - 0x04, 0x05, 0x80, 0x80, 0x06, 0x07, 0x80, 0x80, - }, - - /* High value shuffle used by trans4x4. - */ - { - 0x00, 0x01, 0x02, 0x03, 0x10, 0x11, 0x12, 0x13, - 0x04, 0x05, 0x06, 0x07, 0x14, 0x15, 0x16, 0x17 - }, - - /* Low value shuffle used by trans4x4. - */ - { - 0x08, 0x09, 0x0A, 0x0B, 0x18, 0x19, 0x1A, 0x1B, - 0x0C, 0x0D, 0x0E, 0x0F, 0x1C, 0x1D, 0x1E, 0x1F - } -}; - - -static INLINE void -trans4x4(qword row0, qword row1, qword row2, qword row3, qword *out, - const qword *shuffle) -{ - qword t1 = si_shufb(row0, row2, shuffle[3]); - qword t2 = si_shufb(row0, row2, shuffle[4]); - qword t3 = si_shufb(row1, row3, shuffle[3]); - qword t4 = si_shufb(row1, row3, shuffle[4]); - - out[0] = si_shufb(t1, t3, shuffle[3]); - out[1] = si_shufb(t1, t3, shuffle[4]); - out[2] = si_shufb(t2, t4, shuffle[3]); - out[3] = si_shufb(t2, t4, shuffle[4]); -} - - -/** - * Fetch between 1 and 32 bytes from an unaligned address - */ -static INLINE void -fetch_unaligned(qword *dst, unsigned ea, unsigned size) -{ - qword tmp[4]; - const int shift = ea & 0x0f; - const unsigned aligned_start_ea = ea & ~0x0f; - const unsigned aligned_end_ea = (ea + size) & ~0x0f; - const unsigned num_entries = ((aligned_end_ea - aligned_start_ea) / 16) + 1; - unsigned i; - - - if (shift == 0) { - /* Data is already aligned. Fetch directly into the destination buffer. - */ - for (i = 0; i < num_entries; i++) { - dst[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16)); - } - } else { - /* Fetch data from the cache to the local buffer. - */ - for (i = 0; i < num_entries; i++) { - tmp[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16)); - } - - - /* Fix the alignment of the data and write to the destination buffer. - */ - for (i = 0; i < ((size + 15) / 16); i++) { - dst[i] = si_or((qword) spu_slqwbyte(tmp[i], shift), - (qword) spu_rlmaskqwbyte(tmp[i + 1], shift - 16)); - } - } -} - - -#define CVT_32_FLOAT(q, s) (*(q)) - -static INLINE qword -CVT_64_FLOAT(const qword *qw, const qword *shuffle) -{ - qword a = si_frds(qw[0]); - qword b = si_frds(si_rotqbyi(qw[0], 8)); - qword c = si_frds(qw[1]); - qword d = si_frds(si_rotqbyi(qw[1], 8)); - - qword ab = si_shufb(a, b, shuffle[0]); - qword cd = si_shufb(c, d, si_rotqbyi(shuffle[0], 8)); - - return si_or(ab, cd); -} - - -static INLINE qword -CVT_8_USCALED(const qword *qw, const qword *shuffle) -{ - return si_cuflt(si_shufb(*qw, *qw, shuffle[1]), 0); -} - - -static INLINE qword -CVT_16_USCALED(const qword *qw, const qword *shuffle) -{ - return si_cuflt(si_shufb(*qw, *qw, shuffle[2]), 0); -} - - -static INLINE qword -CVT_32_USCALED(const qword *qw, const qword *shuffle) -{ - (void) shuffle; - return si_cuflt(*qw, 0); -} - -static INLINE qword -CVT_8_SSCALED(const qword *qw, const qword *shuffle) -{ - return si_csflt(si_shufb(*qw, *qw, shuffle[1]), 0); -} - - -static INLINE qword -CVT_16_SSCALED(const qword *qw, const qword *shuffle) -{ - return si_csflt(si_shufb(*qw, *qw, shuffle[2]), 0); -} - - -static INLINE qword -CVT_32_SSCALED(const qword *qw, const qword *shuffle) -{ - (void) shuffle; - return si_csflt(*qw, 0); -} - - -static INLINE qword -CVT_8_UNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 255.0f); - return si_fm(CVT_8_USCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_16_UNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 65535.0f); - return si_fm(CVT_16_USCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_32_UNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 4294967295.0f); - return si_fm(CVT_32_USCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_8_SNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 127.0f); - return si_fm(CVT_8_SSCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_16_SNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 32767.0f); - return si_fm(CVT_16_SSCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_32_SNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 2147483647.0f); - return si_fm(CVT_32_SSCALED(qw, shuffle), scale); -} - -#define SZ_4 si_il(0U) -#define SZ_3 si_fsmbi(0x000f) -#define SZ_2 si_fsmbi(0x00ff) -#define SZ_1 si_fsmbi(0x0fff) - -/** - * Fetch a float[4] vertex attribute from memory, doing format/type - * conversion as needed. - * - * This is probably needed/dupliocated elsewhere, eg format - * conversion, texture sampling etc. - */ -#define FETCH_ATTRIB( NAME, SZ, CVT, N ) \ -static void \ -fetch_##NAME(qword *out, const qword *in, qword defaults, \ - const qword *shuffle) \ -{ \ - qword tmp[4]; \ - \ - tmp[0] = si_selb(CVT(in + (0 * N), shuffle), defaults, SZ); \ - tmp[1] = si_selb(CVT(in + (1 * N), shuffle), defaults, SZ); \ - tmp[2] = si_selb(CVT(in + (2 * N), shuffle), defaults, SZ); \ - tmp[3] = si_selb(CVT(in + (3 * N), shuffle), defaults, SZ); \ - trans4x4(tmp[0], tmp[1], tmp[2], tmp[3], out, shuffle); \ -} - - -FETCH_ATTRIB( R64G64B64A64_FLOAT, SZ_4, CVT_64_FLOAT, 2 ) -FETCH_ATTRIB( R64G64B64_FLOAT, SZ_3, CVT_64_FLOAT, 2 ) -FETCH_ATTRIB( R64G64_FLOAT, SZ_2, CVT_64_FLOAT, 2 ) -FETCH_ATTRIB( R64_FLOAT, SZ_1, CVT_64_FLOAT, 2 ) - -FETCH_ATTRIB( R32G32B32A32_FLOAT, SZ_4, CVT_32_FLOAT, 1 ) -FETCH_ATTRIB( R32G32B32_FLOAT, SZ_3, CVT_32_FLOAT, 1 ) -FETCH_ATTRIB( R32G32_FLOAT, SZ_2, CVT_32_FLOAT, 1 ) -FETCH_ATTRIB( R32_FLOAT, SZ_1, CVT_32_FLOAT, 1 ) - -FETCH_ATTRIB( R32G32B32A32_USCALED, SZ_4, CVT_32_USCALED, 1 ) -FETCH_ATTRIB( R32G32B32_USCALED, SZ_3, CVT_32_USCALED, 1 ) -FETCH_ATTRIB( R32G32_USCALED, SZ_2, CVT_32_USCALED, 1 ) -FETCH_ATTRIB( R32_USCALED, SZ_1, CVT_32_USCALED, 1 ) - -FETCH_ATTRIB( R32G32B32A32_SSCALED, SZ_4, CVT_32_SSCALED, 1 ) -FETCH_ATTRIB( R32G32B32_SSCALED, SZ_3, CVT_32_SSCALED, 1 ) -FETCH_ATTRIB( R32G32_SSCALED, SZ_2, CVT_32_SSCALED, 1 ) -FETCH_ATTRIB( R32_SSCALED, SZ_1, CVT_32_SSCALED, 1 ) - -FETCH_ATTRIB( R32G32B32A32_UNORM, SZ_4, CVT_32_UNORM, 1 ) -FETCH_ATTRIB( R32G32B32_UNORM, SZ_3, CVT_32_UNORM, 1 ) -FETCH_ATTRIB( R32G32_UNORM, SZ_2, CVT_32_UNORM, 1 ) -FETCH_ATTRIB( R32_UNORM, SZ_1, CVT_32_UNORM, 1 ) - -FETCH_ATTRIB( R32G32B32A32_SNORM, SZ_4, CVT_32_SNORM, 1 ) -FETCH_ATTRIB( R32G32B32_SNORM, SZ_3, CVT_32_SNORM, 1 ) -FETCH_ATTRIB( R32G32_SNORM, SZ_2, CVT_32_SNORM, 1 ) -FETCH_ATTRIB( R32_SNORM, SZ_1, CVT_32_SNORM, 1 ) - -FETCH_ATTRIB( R16G16B16A16_USCALED, SZ_4, CVT_16_USCALED, 1 ) -FETCH_ATTRIB( R16G16B16_USCALED, SZ_3, CVT_16_USCALED, 1 ) -FETCH_ATTRIB( R16G16_USCALED, SZ_2, CVT_16_USCALED, 1 ) -FETCH_ATTRIB( R16_USCALED, SZ_1, CVT_16_USCALED, 1 ) - -FETCH_ATTRIB( R16G16B16A16_SSCALED, SZ_4, CVT_16_SSCALED, 1 ) -FETCH_ATTRIB( R16G16B16_SSCALED, SZ_3, CVT_16_SSCALED, 1 ) -FETCH_ATTRIB( R16G16_SSCALED, SZ_2, CVT_16_SSCALED, 1 ) -FETCH_ATTRIB( R16_SSCALED, SZ_1, CVT_16_SSCALED, 1 ) - -FETCH_ATTRIB( R16G16B16A16_UNORM, SZ_4, CVT_16_UNORM, 1 ) -FETCH_ATTRIB( R16G16B16_UNORM, SZ_3, CVT_16_UNORM, 1 ) -FETCH_ATTRIB( R16G16_UNORM, SZ_2, CVT_16_UNORM, 1 ) -FETCH_ATTRIB( R16_UNORM, SZ_1, CVT_16_UNORM, 1 ) - -FETCH_ATTRIB( R16G16B16A16_SNORM, SZ_4, CVT_16_SNORM, 1 ) -FETCH_ATTRIB( R16G16B16_SNORM, SZ_3, CVT_16_SNORM, 1 ) -FETCH_ATTRIB( R16G16_SNORM, SZ_2, CVT_16_SNORM, 1 ) -FETCH_ATTRIB( R16_SNORM, SZ_1, CVT_16_SNORM, 1 ) - -FETCH_ATTRIB( R8G8B8A8_USCALED, SZ_4, CVT_8_USCALED, 1 ) -FETCH_ATTRIB( R8G8B8_USCALED, SZ_3, CVT_8_USCALED, 1 ) -FETCH_ATTRIB( R8G8_USCALED, SZ_2, CVT_8_USCALED, 1 ) -FETCH_ATTRIB( R8_USCALED, SZ_1, CVT_8_USCALED, 1 ) - -FETCH_ATTRIB( R8G8B8A8_SSCALED, SZ_4, CVT_8_SSCALED, 1 ) -FETCH_ATTRIB( R8G8B8_SSCALED, SZ_3, CVT_8_SSCALED, 1 ) -FETCH_ATTRIB( R8G8_SSCALED, SZ_2, CVT_8_SSCALED, 1 ) -FETCH_ATTRIB( R8_SSCALED, SZ_1, CVT_8_SSCALED, 1 ) - -FETCH_ATTRIB( R8G8B8A8_UNORM, SZ_4, CVT_8_UNORM, 1 ) -FETCH_ATTRIB( R8G8B8_UNORM, SZ_3, CVT_8_UNORM, 1 ) -FETCH_ATTRIB( R8G8_UNORM, SZ_2, CVT_8_UNORM, 1 ) -FETCH_ATTRIB( R8_UNORM, SZ_1, CVT_8_UNORM, 1 ) - -FETCH_ATTRIB( R8G8B8A8_SNORM, SZ_4, CVT_8_SNORM, 1 ) -FETCH_ATTRIB( R8G8B8_SNORM, SZ_3, CVT_8_SNORM, 1 ) -FETCH_ATTRIB( R8G8_SNORM, SZ_2, CVT_8_SNORM, 1 ) -FETCH_ATTRIB( R8_SNORM, SZ_1, CVT_8_SNORM, 1 ) - -FETCH_ATTRIB( A8R8G8B8_UNORM, SZ_4, CVT_8_UNORM, 1 ) - - - -static spu_fetch_func get_fetch_func( enum pipe_format format ) -{ - switch (format) { - case PIPE_FORMAT_R64_FLOAT: - return fetch_R64_FLOAT; - case PIPE_FORMAT_R64G64_FLOAT: - return fetch_R64G64_FLOAT; - case PIPE_FORMAT_R64G64B64_FLOAT: - return fetch_R64G64B64_FLOAT; - case PIPE_FORMAT_R64G64B64A64_FLOAT: - return fetch_R64G64B64A64_FLOAT; - - case PIPE_FORMAT_R32_FLOAT: - return fetch_R32_FLOAT; - case PIPE_FORMAT_R32G32_FLOAT: - return fetch_R32G32_FLOAT; - case PIPE_FORMAT_R32G32B32_FLOAT: - return fetch_R32G32B32_FLOAT; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return fetch_R32G32B32A32_FLOAT; - - case PIPE_FORMAT_R32_UNORM: - return fetch_R32_UNORM; - case PIPE_FORMAT_R32G32_UNORM: - return fetch_R32G32_UNORM; - case PIPE_FORMAT_R32G32B32_UNORM: - return fetch_R32G32B32_UNORM; - case PIPE_FORMAT_R32G32B32A32_UNORM: - return fetch_R32G32B32A32_UNORM; - - case PIPE_FORMAT_R32_USCALED: - return fetch_R32_USCALED; - case PIPE_FORMAT_R32G32_USCALED: - return fetch_R32G32_USCALED; - case PIPE_FORMAT_R32G32B32_USCALED: - return fetch_R32G32B32_USCALED; - case PIPE_FORMAT_R32G32B32A32_USCALED: - return fetch_R32G32B32A32_USCALED; - - case PIPE_FORMAT_R32_SNORM: - return fetch_R32_SNORM; - case PIPE_FORMAT_R32G32_SNORM: - return fetch_R32G32_SNORM; - case PIPE_FORMAT_R32G32B32_SNORM: - return fetch_R32G32B32_SNORM; - case PIPE_FORMAT_R32G32B32A32_SNORM: - return fetch_R32G32B32A32_SNORM; - - case PIPE_FORMAT_R32_SSCALED: - return fetch_R32_SSCALED; - case PIPE_FORMAT_R32G32_SSCALED: - return fetch_R32G32_SSCALED; - case PIPE_FORMAT_R32G32B32_SSCALED: - return fetch_R32G32B32_SSCALED; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - return fetch_R32G32B32A32_SSCALED; - - case PIPE_FORMAT_R16_UNORM: - return fetch_R16_UNORM; - case PIPE_FORMAT_R16G16_UNORM: - return fetch_R16G16_UNORM; - case PIPE_FORMAT_R16G16B16_UNORM: - return fetch_R16G16B16_UNORM; - case PIPE_FORMAT_R16G16B16A16_UNORM: - return fetch_R16G16B16A16_UNORM; - - case PIPE_FORMAT_R16_USCALED: - return fetch_R16_USCALED; - case PIPE_FORMAT_R16G16_USCALED: - return fetch_R16G16_USCALED; - case PIPE_FORMAT_R16G16B16_USCALED: - return fetch_R16G16B16_USCALED; - case PIPE_FORMAT_R16G16B16A16_USCALED: - return fetch_R16G16B16A16_USCALED; - - case PIPE_FORMAT_R16_SNORM: - return fetch_R16_SNORM; - case PIPE_FORMAT_R16G16_SNORM: - return fetch_R16G16_SNORM; - case PIPE_FORMAT_R16G16B16_SNORM: - return fetch_R16G16B16_SNORM; - case PIPE_FORMAT_R16G16B16A16_SNORM: - return fetch_R16G16B16A16_SNORM; - - case PIPE_FORMAT_R16_SSCALED: - return fetch_R16_SSCALED; - case PIPE_FORMAT_R16G16_SSCALED: - return fetch_R16G16_SSCALED; - case PIPE_FORMAT_R16G16B16_SSCALED: - return fetch_R16G16B16_SSCALED; - case PIPE_FORMAT_R16G16B16A16_SSCALED: - return fetch_R16G16B16A16_SSCALED; - - case PIPE_FORMAT_R8_UNORM: - return fetch_R8_UNORM; - case PIPE_FORMAT_R8G8_UNORM: - return fetch_R8G8_UNORM; - case PIPE_FORMAT_R8G8B8_UNORM: - return fetch_R8G8B8_UNORM; - case PIPE_FORMAT_R8G8B8A8_UNORM: - return fetch_R8G8B8A8_UNORM; - - case PIPE_FORMAT_R8_USCALED: - return fetch_R8_USCALED; - case PIPE_FORMAT_R8G8_USCALED: - return fetch_R8G8_USCALED; - case PIPE_FORMAT_R8G8B8_USCALED: - return fetch_R8G8B8_USCALED; - case PIPE_FORMAT_R8G8B8A8_USCALED: - return fetch_R8G8B8A8_USCALED; - - case PIPE_FORMAT_R8_SNORM: - return fetch_R8_SNORM; - case PIPE_FORMAT_R8G8_SNORM: - return fetch_R8G8_SNORM; - case PIPE_FORMAT_R8G8B8_SNORM: - return fetch_R8G8B8_SNORM; - case PIPE_FORMAT_R8G8B8A8_SNORM: - return fetch_R8G8B8A8_SNORM; - - case PIPE_FORMAT_R8_SSCALED: - return fetch_R8_SSCALED; - case PIPE_FORMAT_R8G8_SSCALED: - return fetch_R8G8_SSCALED; - case PIPE_FORMAT_R8G8B8_SSCALED: - return fetch_R8G8B8_SSCALED; - case PIPE_FORMAT_R8G8B8A8_SSCALED: - return fetch_R8G8B8A8_SSCALED; - - case PIPE_FORMAT_A8R8G8B8_UNORM: - return fetch_A8R8G8B8_UNORM; - - case 0: - return NULL; /* not sure why this is needed */ - - default: - assert(0); - return NULL; - } -} - - -static unsigned get_vertex_size( enum pipe_format format ) -{ - switch (format) { - case PIPE_FORMAT_R64_FLOAT: - return 8; - case PIPE_FORMAT_R64G64_FLOAT: - return 2 * 8; - case PIPE_FORMAT_R64G64B64_FLOAT: - return 3 * 8; - case PIPE_FORMAT_R64G64B64A64_FLOAT: - return 4 * 8; - - case PIPE_FORMAT_R32_SSCALED: - case PIPE_FORMAT_R32_SNORM: - case PIPE_FORMAT_R32_USCALED: - case PIPE_FORMAT_R32_UNORM: - case PIPE_FORMAT_R32_FLOAT: - return 4; - case PIPE_FORMAT_R32G32_SSCALED: - case PIPE_FORMAT_R32G32_SNORM: - case PIPE_FORMAT_R32G32_USCALED: - case PIPE_FORMAT_R32G32_UNORM: - case PIPE_FORMAT_R32G32_FLOAT: - return 2 * 4; - case PIPE_FORMAT_R32G32B32_SSCALED: - case PIPE_FORMAT_R32G32B32_SNORM: - case PIPE_FORMAT_R32G32B32_USCALED: - case PIPE_FORMAT_R32G32B32_UNORM: - case PIPE_FORMAT_R32G32B32_FLOAT: - return 3 * 4; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - case PIPE_FORMAT_R32G32B32A32_SNORM: - case PIPE_FORMAT_R32G32B32A32_USCALED: - case PIPE_FORMAT_R32G32B32A32_UNORM: - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return 4 * 4; - - case PIPE_FORMAT_R16_SSCALED: - case PIPE_FORMAT_R16_SNORM: - case PIPE_FORMAT_R16_UNORM: - case PIPE_FORMAT_R16_USCALED: - return 2; - case PIPE_FORMAT_R16G16_SSCALED: - case PIPE_FORMAT_R16G16_SNORM: - case PIPE_FORMAT_R16G16_USCALED: - case PIPE_FORMAT_R16G16_UNORM: - return 2 * 2; - case PIPE_FORMAT_R16G16B16_SSCALED: - case PIPE_FORMAT_R16G16B16_SNORM: - case PIPE_FORMAT_R16G16B16_USCALED: - case PIPE_FORMAT_R16G16B16_UNORM: - return 3 * 2; - case PIPE_FORMAT_R16G16B16A16_SSCALED: - case PIPE_FORMAT_R16G16B16A16_SNORM: - case PIPE_FORMAT_R16G16B16A16_USCALED: - case PIPE_FORMAT_R16G16B16A16_UNORM: - return 4 * 2; - - case PIPE_FORMAT_R8_SSCALED: - case PIPE_FORMAT_R8_SNORM: - case PIPE_FORMAT_R8_USCALED: - case PIPE_FORMAT_R8_UNORM: - return 1; - case PIPE_FORMAT_R8G8_SSCALED: - case PIPE_FORMAT_R8G8_SNORM: - case PIPE_FORMAT_R8G8_USCALED: - case PIPE_FORMAT_R8G8_UNORM: - return 2 * 1; - case PIPE_FORMAT_R8G8B8_SSCALED: - case PIPE_FORMAT_R8G8B8_SNORM: - case PIPE_FORMAT_R8G8B8_USCALED: - case PIPE_FORMAT_R8G8B8_UNORM: - return 3 * 1; - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R8G8B8A8_SSCALED: - case PIPE_FORMAT_R8G8B8A8_SNORM: - case PIPE_FORMAT_R8G8B8A8_USCALED: - case PIPE_FORMAT_R8G8B8A8_UNORM: - return 4 * 1; - - case 0: - return 0; /* not sure why this is needed */ - - default: - assert(0); - return 0; - } -} - - -/** - * Fetch vertex attributes for 'count' vertices. - */ -static void generic_vertex_fetch(struct spu_vs_context *draw, - struct spu_exec_machine *machine, - const unsigned *elts, - unsigned count) -{ - unsigned nr_attrs = draw->vertex_fetch.nr_attrs; - unsigned attr; - - assert(count <= 4); - -#if DRAW_DBG - printf("SPU: %s count = %u, nr_attrs = %u\n", - __FUNCTION__, count, nr_attrs); -#endif - - /* loop over vertex attributes (vertex shader inputs) - */ - for (attr = 0; attr < nr_attrs; attr++) { - const qword default_values = (qword)(vec_float4){ 0.0, 0.0, 0.0, 1.0 }; - const unsigned pitch = draw->vertex_fetch.pitch[attr]; - const uint64_t src = draw->vertex_fetch.src_ptr[attr]; - const spu_fetch_func fetch = draw->vertex_fetch.fetch[attr]; - unsigned i; - unsigned idx; - const unsigned bytes_per_entry = draw->vertex_fetch.size[attr]; - const unsigned quads_per_entry = (bytes_per_entry + 15) / 16; - qword in[2 * 4]; - - - /* Fetch four attributes for four vertices. - */ - idx = 0; - for (i = 0; i < count; i++) { - const uint64_t addr = src + (elts[i] * pitch); - -#if DRAW_DBG - printf("SPU: fetching = 0x%llx\n", addr); -#endif - - fetch_unaligned(& in[idx], addr, bytes_per_entry); - idx += quads_per_entry; - } - - /* Be nice and zero out any missing vertices. - */ - (void) memset(& in[idx], 0, (8 - idx) * sizeof(qword)); - - - /* Convert all 4 vertices to vectors of float. - */ - (*fetch)(&machine->Inputs[attr].xyzw[0].q, in, default_values, - fetch_shuffle_data); - } -} - - -void spu_update_vertex_fetch( struct spu_vs_context *draw ) -{ - unsigned i; - - - /* Invalidate the vertex cache. - */ - for (i = 0; i < (CACHE_NWAY * CACHE_NSETS); i++) { - CACHELINE_CLEARVALID(i); - } - - - for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { - draw->vertex_fetch.fetch[i] = - get_fetch_func(draw->vertex_fetch.format[i]); - draw->vertex_fetch.size[i] = - get_vertex_size(draw->vertex_fetch.format[i]); - } - - draw->vertex_fetch.fetch_func = generic_vertex_fetch; -} diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.c b/src/mesa/pipe/cell/spu/spu_vertex_shader.c deleted file mode 100644 index c1cbbb6d1e..0000000000 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.c +++ /dev/null @@ -1,231 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - * Ian Romanick - */ - -#include - -#include "pipe/p_util.h" -#include "pipe/p_state.h" -#include "pipe/p_shader_tokens.h" -#include "spu_vertex_shader.h" -#include "spu_exec.h" -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_context.h" -#include "pipe/cell/common.h" -#include "spu_main.h" - -static INLINE unsigned -compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) -{ - unsigned mask = 0; - unsigned i; - - /* Do the hardwired planes first: - */ - if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; - if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; - if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; - if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; - if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; - if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; - - /* Followed by any remaining ones: - */ - for (i = 6; i < nr; i++) { - if (dot4(clip, plane[i]) < 0) - mask |= (1<machine; - unsigned int j; - - ALIGN16_DECL(struct spu_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct spu_exec_vector, outputs, PIPE_ATTRIB_MAX); - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - - assert(count <= 4); - - machine->Processor = TGSI_PROCESSOR_VERTEX; - - ASSERT_ALIGN16(draw->constants); - machine->Consts = (float (*)[4]) draw->constants; - - machine->Inputs = ALIGN16_ASSIGN(inputs); - machine->Outputs = ALIGN16_ASSIGN(outputs); - - spu_vertex_fetch( draw, machine, elts, count ); - - /* run shader */ - spu_exec_machine_run( machine ); - - - /* store machine results */ - for (j = 0; j < count; j++) { - unsigned slot; - float x, y, z, w; - unsigned char buffer[sizeof(struct vertex_header) - + MAX_VERTEX_SIZE] ALIGN16_ATTRIB; - struct vertex_header *const tmpOut = - (struct vertex_header *) buffer; - const unsigned vert_size = ROUNDUP16(sizeof(struct vertex_header) - + (sizeof(float) * 4 - * draw->num_vs_outputs)); - - mfc_get(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0); - wait_on_mask(1 << TAG_VERTEX_BUFFER); - - - /* Handle attr[0] (position) specially: - * - * XXX: Computing the clipmask should be done in the vertex - * program as a set of DP4 instructions appended to the - * user-provided code. - */ - x = tmpOut->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = tmpOut->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = tmpOut->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = tmpOut->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - - tmpOut->clipmask = compute_clipmask(tmpOut->clip, draw->plane, - draw->nr_planes); - tmpOut->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - tmpOut->data[0][0] = x * scale[0] + trans[0]; - tmpOut->data[0][1] = y * scale[1] + trans[1]; - tmpOut->data[0][2] = z * scale[2] + trans[2]; - tmpOut->data[0][3] = w; - - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - tmpOut->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - tmpOut->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - tmpOut->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - tmpOut->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; - } - - mfc_put(tmpOut, vOut[j], vert_size, TAG_VERTEX_BUFFER, 0, 0); - } /* loop over vertices */ -} - - -static void -spu_bind_vertex_shader(struct spu_vs_context *draw, - void *uniforms, - void *planes, - unsigned nr_planes, - unsigned num_outputs - ) -{ - draw->constants = (float (*)[4]) uniforms; - - (void) memcpy(draw->plane, planes, sizeof(float) * 4 * nr_planes); - draw->nr_planes = nr_planes; - draw->num_vs_outputs = num_outputs; - - /* specify the shader to interpret/execute */ - spu_exec_machine_init(&draw->machine, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/, - PIPE_SHADER_VERTEX); -} - - -unsigned char immediates[(sizeof(float) * 4 * TGSI_EXEC_NUM_IMMEDIATES) + 32] - ALIGN16_ATTRIB; - -void -spu_execute_vertex_shader(struct spu_vs_context *draw, - const struct cell_command_vs *vs) -{ - unsigned i; - - const uint64_t immediate_addr = vs->shader.immediates; - const unsigned immediate_size = - ROUNDUP16((sizeof(float) * 4 * vs->shader.num_immediates) - + (immediate_addr & 0x0f)); - - mfc_get(immediates, immediate_addr & ~0x0f, immediate_size, - TAG_VERTEX_BUFFER, 0, 0); - - draw->machine.Instructions = (struct tgsi_full_instruction *) - vs->shader.instructions; - draw->machine.NumInstructions = vs->shader.num_instructions; - - draw->machine.Declarations = (struct tgsi_full_declaration *) - vs->shader.declarations; - draw->machine.NumDeclarations = vs->shader.num_declarations; - - draw->vertex_fetch.nr_attrs = vs->nr_attrs; - - wait_on_mask(1 << TAG_VERTEX_BUFFER); - - (void) memcpy(& draw->machine.Imms, &immediates[immediate_addr & 0x0f], - sizeof(float) * 4 * vs->shader.num_immediates); - - spu_bind_vertex_shader(draw, vs->shader.uniforms, - vs->plane, vs->nr_planes, - vs->shader.num_outputs); - - for (i = 0; i < vs->num_elts; i += 4) { - const unsigned batch_size = MIN2(vs->num_elts - i, 4); - - run_vertex_program(draw, & vs->elts[i], batch_size, &vs->vOut[i]); - } -} diff --git a/src/mesa/pipe/cell/spu/spu_vertex_shader.h b/src/mesa/pipe/cell/spu/spu_vertex_shader.h deleted file mode 100644 index b5bf31e67d..0000000000 --- a/src/mesa/pipe/cell/spu/spu_vertex_shader.h +++ /dev/null @@ -1,63 +0,0 @@ -#ifndef SPU_VERTEX_SHADER_H -#define SPU_VERTEX_SHADER_H - -#include "pipe/p_format.h" -#include "spu_exec.h" - -struct spu_vs_context; - -typedef void (*spu_fetch_func)(qword *out, const qword *in, qword defaults, - const qword *shuffle_data); -typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw, - struct spu_exec_machine *machine, - const unsigned *elts, - unsigned count ); - -struct spu_vs_context { - struct pipe_viewport_state viewport; - - struct { - uint64_t src_ptr[PIPE_ATTRIB_MAX]; - unsigned pitch[PIPE_ATTRIB_MAX]; - unsigned size[PIPE_ATTRIB_MAX]; - enum pipe_format format[PIPE_ATTRIB_MAX]; - unsigned nr_attrs; - boolean dirty; - - spu_fetch_func fetch[PIPE_ATTRIB_MAX]; - spu_full_fetch_func fetch_func; - } vertex_fetch; - - /* Clip derived state: - */ - float plane[12][4]; - unsigned nr_planes; - - struct spu_exec_machine machine; - const float (*constants)[4]; - - unsigned num_vs_outputs; -}; - -extern void spu_update_vertex_fetch(struct spu_vs_context *draw); - -static INLINE void spu_vertex_fetch(struct spu_vs_context *draw, - struct spu_exec_machine *machine, - const unsigned *elts, - unsigned count) -{ - if (draw->vertex_fetch.dirty) { - spu_update_vertex_fetch(draw); - draw->vertex_fetch.dirty = 0; - } - - (*draw->vertex_fetch.fetch_func)(draw, machine, elts, count); -} - -struct cell_command_vs; - -extern void -spu_execute_vertex_shader(struct spu_vs_context *draw, - const struct cell_command_vs *vs); - -#endif /* SPU_VERTEX_SHADER_H */ diff --git a/src/mesa/pipe/cell/spu/spu_ztest.h b/src/mesa/pipe/cell/spu/spu_ztest.h deleted file mode 100644 index ce8ad00339..0000000000 --- a/src/mesa/pipe/cell/spu/spu_ztest.h +++ /dev/null @@ -1,135 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * Zbuffer/depth test code. - */ - - -#ifndef SPU_ZTEST_H -#define SPU_ZTEST_H - - -#ifdef __SPU__ -#include -#endif - - - -/** - * Perform Z testing for a 16-bit/value Z buffer. - * - * \param zvals vector of four fragment zvalues as floats - * \param zbuf ptr to vector of ushort[8] zbuffer values. Note that this - * contains the Z values for 2 quads, 8 pixels. - * \param x x coordinate of quad (only lsbit is significant) - * \param inMask indicates which fragments in the quad are alive - * \return new mask indicating which fragments are alive after ztest - */ -static INLINE vector unsigned int -spu_z16_test_less(vector float zvals, vector unsigned short *zbuf, - uint x, vector unsigned int inMask) -{ -#define ZERO 0x80 - vector unsigned int zvals_ui4, zbuf_ui4, mask; - - /* convert floats to uints in [0, 65535] */ - zvals_ui4 = spu_convtu(zvals, 32); /* convert to [0, 2^32] */ - zvals_ui4 = spu_rlmask(zvals_ui4, -16); /* right shift 16 */ - - /* XXX this conditional could be removed with a bit of work */ - if (x & 1) { - /* convert zbuffer values from ushorts to uints */ - /* gather lower four ushorts */ - zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, - (vector unsigned int) *zbuf, - ((vector unsigned char) { - ZERO, ZERO, 8, 9, ZERO, ZERO, 10, 11, - ZERO, ZERO, 12, 13, ZERO, ZERO, 14, 15})); - /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ - mask = spu_cmpgt(zbuf_ui4, zvals_ui4); - /* mask &= inMask */ - mask = spu_and(mask, inMask); - /* zbuf = mask ? zval : zbuf */ - zbuf_ui4 = spu_sel(zbuf_ui4, zvals_ui4, mask); - /* convert zbuffer values from uints back to ushorts, preserve lower 4 */ - *zbuf = (vector unsigned short) - spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, - ((vector unsigned char) { - 16, 17, 18, 19, 20, 21, 22, 23, - 2, 3, 6, 7, 10, 11, 14, 15})); - } - else { - /* convert zbuffer values from ushorts to uints */ - /* gather upper four ushorts */ - zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, - (vector unsigned int) *zbuf, - ((vector unsigned char) { - ZERO, ZERO, 0, 1, ZERO, ZERO, 2, 3, - ZERO, ZERO, 4, 5, ZERO, ZERO, 6, 7})); - /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ - mask = spu_cmpgt(zbuf_ui4, zvals_ui4); - /* mask &= inMask */ - mask = spu_and(mask, inMask); - /* zbuf = mask ? zval : zbuf */ - zbuf_ui4 = spu_sel(zbuf_ui4, zvals_ui4, mask); - /* convert zbuffer values from uints back to ushorts, preserve upper 4 */ - *zbuf = (vector unsigned short) - spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, - ((vector unsigned char) { - 2, 3, 6, 7, 10, 11, 14, 15, - 24, 25, 26, 27, 28, 29, 30, 31})); - } - return mask; -#undef ZERO -} - - -/** - * As above, but Zbuffer values as 32-bit uints - */ -static INLINE vector unsigned int -spu_z32_test_less(vector float zvals, vector unsigned int *zbuf_ptr, - vector unsigned int inMask) -{ - vector unsigned int zvals_ui4, mask, zbuf = *zbuf_ptr; - - /* convert floats to uints in [0, 0xffffffff] */ - zvals_ui4 = spu_convtu(zvals, 32); - /* mask = (zbuf < zvals_ui4) ? ~0 : 0 */ - mask = spu_cmpgt(zbuf, zvals_ui4); - /* mask &= inMask */ - mask = spu_and(mask, inMask); - /* zbuf = mask ? zval : zbuf */ - *zbuf_ptr = spu_sel(zbuf, zvals_ui4, mask); - - return mask; -} - - -#endif /* SPU_ZTEST_H */ diff --git a/src/mesa/pipe/cso_cache/cso_cache.c b/src/mesa/pipe/cso_cache/cso_cache.c deleted file mode 100644 index 9e77e0774d..0000000000 --- a/src/mesa/pipe/cso_cache/cso_cache.c +++ /dev/null @@ -1,181 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Zack Rusin - */ - -#include "cso_cache.h" -#include "cso_hash.h" - -#if 1 -static unsigned hash_key(const void *key, unsigned key_size) -{ - unsigned *ikey = (unsigned *)key; - unsigned hash = 0, i; - - assert(key_size % 4 == 0); - - /* I'm sure this can be improved on: - */ - for (i = 0; i < key_size/4; i++) - hash ^= ikey[i]; - - return hash; -} -#else -static unsigned hash_key(const unsigned char *p, int n) -{ - unsigned h = 0; - unsigned g; - - while (n--) { - h = (h << 4) + *p++; - if ((g = (h & 0xf0000000)) != 0) - h ^= g >> 23; - h &= ~g; - } - return h; -} -#endif - -unsigned cso_construct_key(void *item, int item_size) -{ - return hash_key((item), item_size); -} - -static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_type type) -{ - struct cso_hash *hash = 0; - - switch(type) { - case CSO_BLEND: - hash = sc->blend_hash; - break; - case CSO_SAMPLER: - hash = sc->sampler_hash; - break; - case CSO_DEPTH_STENCIL_ALPHA: - hash = sc->depth_stencil_hash; - break; - case CSO_RASTERIZER: - hash = sc->rasterizer_hash; - break; - case CSO_FRAGMENT_SHADER: - hash = sc->fs_hash; - break; - case CSO_VERTEX_SHADER: - hash = sc->vs_hash; - break; - } - - return hash; -} - -static int _cso_size_for_type(enum cso_cache_type type) -{ - switch(type) { - case CSO_BLEND: - return sizeof(struct pipe_blend_state); - case CSO_SAMPLER: - return sizeof(struct pipe_sampler_state); - case CSO_DEPTH_STENCIL_ALPHA: - return sizeof(struct pipe_depth_stencil_alpha_state); - case CSO_RASTERIZER: - return sizeof(struct pipe_rasterizer_state); - case CSO_FRAGMENT_SHADER: - return sizeof(struct pipe_shader_state); - case CSO_VERTEX_SHADER: - return sizeof(struct pipe_shader_state); - } - return 0; -} - -struct cso_hash_iter -cso_insert_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *state) -{ - struct cso_hash *hash = _cso_hash_for_type(sc, type); - return cso_hash_insert(hash, hash_key, state); -} - -struct cso_hash_iter -cso_find_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type) -{ - struct cso_hash *hash = _cso_hash_for_type(sc, type); - - return cso_hash_find(hash, hash_key); -} - -struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *templ) -{ - struct cso_hash_iter iter = cso_find_state(sc, hash_key, type); - int size = _cso_size_for_type(type); - while (!cso_hash_iter_is_null(iter)) { - void *iter_data = cso_hash_iter_data(iter); - if (!memcmp(iter_data, templ, size)) - return iter; - iter = cso_hash_iter_next(iter); - } - return iter; -} - -void * cso_take_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type) -{ - struct cso_hash *hash = _cso_hash_for_type(sc, type); - return cso_hash_take(hash, hash_key); -} - -struct cso_cache *cso_cache_create(void) -{ - struct cso_cache *sc = malloc(sizeof(struct cso_cache)); - - sc->blend_hash = cso_hash_create(); - sc->sampler_hash = cso_hash_create(); - sc->depth_stencil_hash = cso_hash_create(); - sc->rasterizer_hash = cso_hash_create(); - sc->fs_hash = cso_hash_create(); - sc->vs_hash = cso_hash_create(); - - return sc; -} - -void cso_cache_delete(struct cso_cache *sc) -{ - assert(sc); - cso_hash_delete(sc->blend_hash); - cso_hash_delete(sc->sampler_hash); - cso_hash_delete(sc->depth_stencil_hash); - cso_hash_delete(sc->rasterizer_hash); - cso_hash_delete(sc->fs_hash); - cso_hash_delete(sc->vs_hash); - free(sc); -} diff --git a/src/mesa/pipe/cso_cache/cso_cache.h b/src/mesa/pipe/cso_cache/cso_cache.h deleted file mode 100644 index 116e2eaa2c..0000000000 --- a/src/mesa/pipe/cso_cache/cso_cache.h +++ /dev/null @@ -1,107 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#ifndef CSO_CACHE_H -#define CSO_CACHE_H - -#include "pipe/p_context.h" -#include "pipe/p_state.h" - - -struct cso_hash; - -struct cso_cache { - struct cso_hash *blend_hash; - struct cso_hash *depth_stencil_hash; - struct cso_hash *fs_hash; - struct cso_hash *vs_hash; - struct cso_hash *rasterizer_hash; - struct cso_hash *sampler_hash; -}; - -struct cso_blend { - struct pipe_blend_state state; - void *data; -}; - -struct cso_depth_stencil_alpha { - struct pipe_depth_stencil_alpha_state state; - void *data; -}; - -struct cso_rasterizer { - struct pipe_rasterizer_state state; - void *data; -}; - -struct cso_fragment_shader { - struct pipe_shader_state state; - void *data; -}; - -struct cso_vertex_shader { - struct pipe_shader_state state; - void *data; -}; - -struct cso_sampler { - struct pipe_sampler_state state; - void *data; -}; - - -enum cso_cache_type { - CSO_BLEND, - CSO_SAMPLER, - CSO_DEPTH_STENCIL_ALPHA, - CSO_RASTERIZER, - CSO_FRAGMENT_SHADER, - CSO_VERTEX_SHADER -}; - -unsigned cso_construct_key(void *item, int item_size); - -struct cso_cache *cso_cache_create(void); -void cso_cache_delete(struct cso_cache *sc); - -struct cso_hash_iter cso_insert_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *state); -struct cso_hash_iter cso_find_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type); -struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *templ); -void * cso_take_state(struct cso_cache *sc, unsigned hash_key, - enum cso_cache_type type); - -#endif diff --git a/src/mesa/pipe/cso_cache/cso_hash.c b/src/mesa/pipe/cso_cache/cso_hash.c deleted file mode 100644 index 0338cb3b47..0000000000 --- a/src/mesa/pipe/cso_cache/cso_hash.c +++ /dev/null @@ -1,388 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#include "cso_hash.h" - -#include -#include -#include -#include - -#define MAX(a, b) ((a > b) ? (a) : (b)) - -static const int MinNumBits = 4; - -static const unsigned char prime_deltas[] = { - 0, 0, 1, 3, 1, 5, 3, 3, 1, 9, 7, 5, 3, 9, 25, 3, - 1, 21, 3, 21, 7, 15, 9, 5, 3, 29, 15, 0, 0, 0, 0, 0 -}; - -static int primeForNumBits(int numBits) -{ - return (1 << numBits) + prime_deltas[numBits]; -} - -/* - Returns the smallest integer n such that - primeForNumBits(n) >= hint. -*/ -static int countBits(int hint) -{ - int numBits = 0; - int bits = hint; - - while (bits > 1) { - bits >>= 1; - numBits++; - } - - if (numBits >= (int)sizeof(prime_deltas)) { - numBits = sizeof(prime_deltas) - 1; - } else if (primeForNumBits(numBits) < hint) { - ++numBits; - } - return numBits; -} - -struct cso_node { - struct cso_node *next; - unsigned key; - void *value; -}; - -struct cso_hash_data { - struct cso_node *fakeNext; - struct cso_node **buckets; - int size; - int nodeSize; - short userNumBits; - short numBits; - int numBuckets; -}; - -struct cso_hash { - union { - struct cso_hash_data *d; - struct cso_node *e; - } data; -}; - -static void *cso_data_allocate_node(struct cso_hash_data *hash) -{ - return malloc(hash->nodeSize); -} - -static void cso_data_free_node(struct cso_node *node) -{ - /* XXX still a leak here. - * Need to cast value ptr to original cso type, then free the - * driver-specific data hanging off of it. For example: - struct cso_sampler *csamp = (struct cso_sampler *) node->value; - free(csamp->data); - */ - free(node->value); - free(node); -} - -static struct cso_node * -cso_hash_create_node(struct cso_hash *hash, - unsigned akey, void *avalue, - struct cso_node **anextNode) -{ - struct cso_node *node = cso_data_allocate_node(hash->data.d); - node->key = akey; - node->value = avalue; - - node->next = (struct cso_node*)(*anextNode); - *anextNode = node; - ++hash->data.d->size; - return node; -} - -static void cso_data_rehash(struct cso_hash_data *hash, int hint) -{ - if (hint < 0) { - hint = countBits(-hint); - if (hint < MinNumBits) - hint = MinNumBits; - hash->userNumBits = hint; - while (primeForNumBits(hint) < (hash->size >> 1)) - ++hint; - } else if (hint < MinNumBits) { - hint = MinNumBits; - } - - if (hash->numBits != hint) { - struct cso_node *e = (struct cso_node *)(hash); - struct cso_node **oldBuckets = hash->buckets; - int oldNumBuckets = hash->numBuckets; - int i = 0; - - hash->numBits = hint; - hash->numBuckets = primeForNumBits(hint); - hash->buckets = malloc(sizeof(struct cso_node*) * hash->numBuckets); - for (i = 0; i < hash->numBuckets; ++i) - hash->buckets[i] = e; - - for (i = 0; i < oldNumBuckets; ++i) { - struct cso_node *firstNode = oldBuckets[i]; - while (firstNode != e) { - unsigned h = firstNode->key; - struct cso_node *lastNode = firstNode; - while (lastNode->next != e && lastNode->next->key == h) - lastNode = lastNode->next; - - struct cso_node *afterLastNode = lastNode->next; - struct cso_node **beforeFirstNode = &hash->buckets[h % hash->numBuckets]; - while (*beforeFirstNode != e) - beforeFirstNode = &(*beforeFirstNode)->next; - lastNode->next = *beforeFirstNode; - *beforeFirstNode = firstNode; - firstNode = afterLastNode; - } - } - free(oldBuckets); - } -} - -static void cso_data_might_grow(struct cso_hash_data *hash) -{ - if (hash->size >= hash->numBuckets) - cso_data_rehash(hash, hash->numBits + 1); -} - -static void cso_data_has_shrunk(struct cso_hash_data *hash) -{ - if (hash->size <= (hash->numBuckets >> 3) && - hash->numBits > hash->userNumBits) { - int max = MAX(hash->numBits-2, hash->userNumBits); - cso_data_rehash(hash, max); - } -} - -static struct cso_node *cso_data_first_node(struct cso_hash_data *hash) -{ - struct cso_node *e = (struct cso_node *)(hash); - struct cso_node **bucket = hash->buckets; - int n = hash->numBuckets; - while (n--) { - if (*bucket != e) - return *bucket; - ++bucket; - } - return e; -} - -static struct cso_node **cso_hash_find_node(struct cso_hash *hash, unsigned akey) -{ - struct cso_node **node; - - if (hash->data.d->numBuckets) { - node = (struct cso_node **)(&hash->data.d->buckets[akey % hash->data.d->numBuckets]); - assert(*node == hash->data.e || (*node)->next); - while (*node != hash->data.e && (*node)->key != akey) - node = &(*node)->next; - } else { - node = (struct cso_node **)((const struct cso_node * const *)(&hash->data.e)); - } - return node; -} - -struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, - unsigned key, void *data) -{ - cso_data_might_grow(hash->data.d); - - struct cso_node **nextNode = cso_hash_find_node(hash, key); - struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); - struct cso_hash_iter iter = {hash, node}; - return iter; -} - -struct cso_hash * cso_hash_create(void) -{ - struct cso_hash *hash = malloc(sizeof(struct cso_hash)); - hash->data.d = malloc(sizeof(struct cso_hash_data)); - hash->data.d->fakeNext = 0; - hash->data.d->buckets = 0; - hash->data.d->size = 0; - hash->data.d->nodeSize = sizeof(struct cso_node); - hash->data.d->userNumBits = MinNumBits; - hash->data.d->numBits = 0; - hash->data.d->numBuckets = 0; - - return hash; -} - -void cso_hash_delete(struct cso_hash *hash) -{ - struct cso_node *e_for_x = (struct cso_node *)(hash->data.d); - struct cso_node **bucket = (struct cso_node **)(hash->data.d->buckets); - int n = hash->data.d->numBuckets; - while (n--) { - struct cso_node *cur = *bucket++; - while (cur != e_for_x) { - struct cso_node *next = cur->next; - cso_data_free_node(cur); - cur = next; - } - } - free(hash->data.d->buckets); - free(hash->data.d); - free(hash); -} - -struct cso_hash_iter cso_hash_find(struct cso_hash *hash, - unsigned key) -{ - struct cso_node **nextNode = cso_hash_find_node(hash, key); - struct cso_hash_iter iter = {hash, *nextNode}; - return iter; -} - -unsigned cso_hash_iter_key(struct cso_hash_iter iter) -{ - if (!iter.node || iter.hash->data.e == iter.node) - return 0; - return iter.node->key; -} - -void * cso_hash_iter_data(struct cso_hash_iter iter) -{ - if (!iter.node || iter.hash->data.e == iter.node) - return 0; - return iter.node->value; -} - -static struct cso_node *cso_hash_data_next(struct cso_node *node) -{ - union { - struct cso_node *next; - struct cso_node *e; - struct cso_hash_data *d; - } a; - a.next = node->next; - if (!a.next) { - fprintf(stderr, "iterating beyond the last element\n"); - return 0; - } - if (a.next->next) - return a.next; - - int start = (node->key % a.d->numBuckets) + 1; - struct cso_node **bucket = a.d->buckets + start; - int n = a.d->numBuckets - start; - while (n--) { - if (*bucket != a.e) - return *bucket; - ++bucket; - } - return a.e; -} - - -static struct cso_node *cso_hash_data_prev(struct cso_node *node) -{ - union { - struct cso_node *e; - struct cso_hash_data *d; - } a; - - a.e = node; - while (a.e->next) - a.e = a.e->next; - - int start; - if (node == a.e) - start = a.d->numBuckets - 1; - else - start = node->key % a.d->numBuckets; - - struct cso_node *sentinel = node; - struct cso_node **bucket = a.d->buckets + start; - while (start >= 0) { - if (*bucket != sentinel) { - struct cso_node *prev = *bucket; - while (prev->next != sentinel) - prev = prev->next; - return prev; - } - - sentinel = a.e; - --bucket; - --start; - } - fprintf(stderr, "iterating backward beyond first element\n"); - return a.e; -} - -struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter) -{ - struct cso_hash_iter next = {iter.hash, cso_hash_data_next(iter.node)}; - return next; -} - -int cso_hash_iter_is_null(struct cso_hash_iter iter) -{ - if (!iter.node || iter.node == iter.hash->data.e) - return 1; - return 0; -} - -void * cso_hash_take(struct cso_hash *hash, - unsigned akey) -{ - struct cso_node **node = cso_hash_find_node(hash, akey); - if (*node != hash->data.e) { - void *t = (*node)->value; - struct cso_node *next = (*node)->next; - cso_data_free_node(*node); - *node = next; - --hash->data.d->size; - cso_data_has_shrunk(hash->data.d); - return t; - } - return 0; -} - -struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter) -{ - struct cso_hash_iter prev = {iter.hash, - cso_hash_data_prev(iter.node)}; - return prev; -} - -struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash) -{ - struct cso_hash_iter iter = {hash, cso_data_first_node(hash->data.d)}; - return iter; -} diff --git a/src/mesa/pipe/cso_cache/cso_hash.h b/src/mesa/pipe/cso_cache/cso_hash.h deleted file mode 100644 index b4aa111860..0000000000 --- a/src/mesa/pipe/cso_cache/cso_hash.h +++ /dev/null @@ -1,62 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#ifndef CSO_HASH_H -#define CSO_HASH_H - -struct cso_hash; -struct cso_node; - -struct cso_hash_iter { - struct cso_hash *hash; - struct cso_node *node; -}; - -struct cso_hash *cso_hash_create(void); -void cso_hash_delete(struct cso_hash *hash); - -struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, - void *data); -void *cso_hash_take(struct cso_hash *hash, unsigned key); - -struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash); -struct cso_hash_iter cso_hash_find(struct cso_hash *hash, unsigned key); - - -int cso_hash_iter_is_null(struct cso_hash_iter iter); -unsigned cso_hash_iter_key(struct cso_hash_iter iter); -void *cso_hash_iter_data(struct cso_hash_iter iter); - -struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter); -struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); - -#endif diff --git a/src/mesa/pipe/draw/Makefile b/src/mesa/pipe/draw/Makefile deleted file mode 100644 index 451911a354..0000000000 --- a/src/mesa/pipe/draw/Makefile +++ /dev/null @@ -1,2 +0,0 @@ -default: - cd .. ; make diff --git a/src/mesa/pipe/draw/draw_clip.c b/src/mesa/pipe/draw/draw_clip.c deleted file mode 100644 index e3051507ea..0000000000 --- a/src/mesa/pipe/draw/draw_clip.c +++ /dev/null @@ -1,488 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Clipping stage - * - * \author Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" - -#include "draw_context.h" -#include "draw_private.h" - - -#ifndef IS_NEGATIVE -#define IS_NEGATIVE(X) ((X) < 0.0) -#endif - -#ifndef DIFFERENT_SIGNS -#define DIFFERENT_SIGNS(x, y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F) -#endif - -#ifndef MAX_CLIPPED_VERTICES -#define MAX_CLIPPED_VERTICES ((2 * (6 + PIPE_MAX_CLIP_PLANES))+1) -#endif - - - -struct clipper { - struct draw_stage stage; /**< base class */ - - /* Basically duplicate some of the flatshading logic here: - */ - boolean flat; - uint num_color_attribs; - uint color_attribs[4]; /* front/back primary/secondary colors */ - - float (*plane)[4]; -}; - - -/* This is a bit confusing: - */ -static INLINE struct clipper *clipper_stage( struct draw_stage *stage ) -{ - return (struct clipper *)stage; -} - - -#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) - - -/* All attributes are float[4], so this is easy: - */ -static void interp_attr( float *fdst, - float t, - const float *fin, - const float *fout ) -{ - fdst[0] = LINTERP( t, fout[0], fin[0] ); - fdst[1] = LINTERP( t, fout[1], fin[1] ); - fdst[2] = LINTERP( t, fout[2], fin[2] ); - fdst[3] = LINTERP( t, fout[3], fin[3] ); -} - -static void copy_colors( struct draw_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - const struct clipper *clipper = clipper_stage(stage); - uint i; - for (i = 0; i < clipper->num_color_attribs; i++) { - const uint attr = clipper->color_attribs[i]; - COPY_4FV(dst->data[attr], src->data[attr]); - } -} - - - -/* Interpolate between two vertices to produce a third. - */ -static void interp( const struct clipper *clip, - struct vertex_header *dst, - float t, - const struct vertex_header *out, - const struct vertex_header *in ) -{ - const unsigned nr_attrs = clip->stage.draw->num_vs_outputs; - unsigned j; - - /* Vertex header. - */ - { - dst->clipmask = 0; - dst->edgeflag = 0; - dst->pad = 0; - dst->vertex_id = UNDEFINED_VERTEX_ID; - } - - /* Clip coordinates: interpolate normally - */ - { - interp_attr(dst->clip, t, in->clip, out->clip); - } - - /* Do the projective divide and insert window coordinates: - */ - { - const float *pos = dst->clip; - const float *scale = clip->stage.draw->viewport.scale; - const float *trans = clip->stage.draw->viewport.translate; - const float oow = 1.0f / pos[3]; - - dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; - dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; - dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; - dst->data[0][3] = oow; - } - - /* Other attributes - * Note: start at 1 to skip winpos (data[0]) since we just computed - * it above. - */ - for (j = 1; j < nr_attrs; j++) { - interp_attr(dst->data[j], t, in->data[j], out->data[j]); - } -} - - -static void emit_poly( struct draw_stage *stage, - struct vertex_header **inlist, - unsigned n, - const struct prim_header *origPrim) -{ - struct prim_header header; - unsigned i; - - /* later stages may need the determinant, but only the sign matters */ - header.det = origPrim->det; - - for (i = 2; i < n; i++) { - header.v[0] = inlist[i-1]; - header.v[1] = inlist[i]; - header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ - - { - unsigned tmp1 = header.v[1]->edgeflag; - unsigned tmp2 = header.v[2]->edgeflag; - - if (i != n-1) header.v[1]->edgeflag = 0; - if (i != 2) header.v[2]->edgeflag = 0; - - header.edgeflags = ((header.v[0]->edgeflag << 0) | - (header.v[1]->edgeflag << 1) | - (header.v[2]->edgeflag << 2)); - - stage->next->tri( stage->next, &header ); - - header.v[1]->edgeflag = tmp1; - header.v[2]->edgeflag = tmp2; - } - } -} - - - - -/* Clip a triangle against the viewport and user clip planes. - */ -static void -do_clip_tri( struct draw_stage *stage, - struct prim_header *header, - unsigned clipmask ) -{ - struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *a[MAX_CLIPPED_VERTICES]; - struct vertex_header *b[MAX_CLIPPED_VERTICES]; - struct vertex_header **inlist = a; - struct vertex_header **outlist = b; - unsigned tmpnr = 0; - unsigned n = 3; - unsigned i; - - inlist[0] = header->v[0]; - inlist[1] = header->v[1]; - inlist[2] = header->v[2]; - - while (clipmask && n >= 3) { - const unsigned plane_idx = ffs(clipmask)-1; - const float *plane = clipper->plane[plane_idx]; - struct vertex_header *vert_prev = inlist[0]; - float dp_prev = dot4( vert_prev->clip, plane ); - unsigned outcount = 0; - - clipmask &= ~(1<clip, plane ); - - if (!IS_NEGATIVE(dp_prev)) { - outlist[outcount++] = vert_prev; - } - - if (DIFFERENT_SIGNS(dp, dp_prev)) { - struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; - outlist[outcount++] = new_vert; - - if (IS_NEGATIVE(dp)) { - /* Going out of bounds. Avoid division by zero as we - * know dp != dp_prev from DIFFERENT_SIGNS, above. - */ - float t = dp / (dp - dp_prev); - interp( clipper, new_vert, t, vert, vert_prev ); - - /* Force edgeflag true in this case: - */ - new_vert->edgeflag = 1; - } else { - /* Coming back in. - */ - float t = dp_prev / (dp_prev - dp); - interp( clipper, new_vert, t, vert_prev, vert ); - - /* Copy starting vert's edgeflag: - */ - new_vert->edgeflag = vert_prev->edgeflag; - } - } - - vert_prev = vert; - dp_prev = dp; - } - - { - struct vertex_header **tmp = inlist; - inlist = outlist; - outlist = tmp; - n = outcount; - } - } - - /* If flat-shading, copy color to new provoking vertex. - */ - if (clipper->flat && inlist[0] != header->v[2]) { - if (1) { - inlist[0] = dup_vert(stage, inlist[0], tmpnr++); - } - - copy_colors(stage, inlist[0], header->v[2]); - } - - - - /* Emit the polygon as triangles to the setup stage: - */ - if (n >= 3) - emit_poly( stage, inlist, n, header ); -} - - -/* Clip a line against the viewport and user clip planes. - */ -static void -do_clip_line( struct draw_stage *stage, - struct prim_header *header, - unsigned clipmask ) -{ - const struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - const float *pos0 = v0->clip; - const float *pos1 = v1->clip; - float t0 = 0.0F; - float t1 = 0.0F; - struct prim_header newprim; - - while (clipmask) { - const unsigned plane_idx = ffs(clipmask)-1; - const float *plane = clipper->plane[plane_idx]; - const float dp0 = dot4( pos0, plane ); - const float dp1 = dot4( pos1, plane ); - - if (dp1 < 0.0F) { - float t = dp1 / (dp1 - dp0); - t1 = MAX2(t1, t); - } - - if (dp0 < 0.0F) { - float t = dp0 / (dp0 - dp1); - t0 = MAX2(t0, t); - } - - if (t0 + t1 >= 1.0F) - return; /* discard */ - - clipmask &= ~(1 << plane_idx); /* turn off this plane's bit */ - } - - if (v0->clipmask) { - interp( clipper, stage->tmp[0], t0, v0, v1 ); - - if (clipper->flat) - copy_colors(stage, stage->tmp[0], v0); - - newprim.v[0] = stage->tmp[0]; - } - else { - newprim.v[0] = v0; - } - - if (v1->clipmask) { - interp( clipper, stage->tmp[1], t1, v1, v0 ); - newprim.v[1] = stage->tmp[1]; - } - else { - newprim.v[1] = v1; - } - - stage->next->line( stage->next, &newprim ); -} - - -static void -clip_point( struct draw_stage *stage, - struct prim_header *header ) -{ - if (header->v[0]->clipmask == 0) - stage->next->point( stage->next, header ); -} - - -static void -clip_line( struct draw_stage *stage, - struct prim_header *header ) -{ - unsigned clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask); - - if (clipmask == 0) { - /* no clipping needed */ - stage->next->line( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask) == 0) { - do_clip_line(stage, header, clipmask); - } - /* else, totally clipped */ -} - - -static void -clip_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - unsigned clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask | - header->v[2]->clipmask); - - if (clipmask == 0) { - /* no clipping needed */ - stage->next->tri( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask & - header->v[2]->clipmask) == 0) { - do_clip_tri(stage, header, clipmask); - } -} - -/* Update state. Could further delay this until we hit the first - * primitive that really requires clipping. - */ -static void -clip_init_state( struct draw_stage *stage ) -{ - struct clipper *clipper = clipper_stage( stage ); - - clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; - - if (clipper->flat) { - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; - uint i; - - clipper->num_color_attribs = 0; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || - vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - clipper->color_attribs[clipper->num_color_attribs++] = i; - } - } - } - - stage->tri = clip_tri; - stage->line = clip_line; -} - - - -static void clip_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - clip_init_state( stage ); - stage->tri( stage, header ); -} - -static void clip_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - clip_init_state( stage ); - stage->line( stage, header ); -} - - -static void clip_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->tri = clip_first_tri; - stage->line = clip_first_line; - stage->next->flush( stage->next, flags ); -} - - -static void clip_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void clip_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Allocate a new clipper stage. - * \return pointer to new stage object - */ -struct draw_stage *draw_clip_stage( struct draw_context *draw ) -{ - struct clipper *clipper = CALLOC_STRUCT(clipper); - - draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); - - clipper->stage.draw = draw; - clipper->stage.point = clip_point; - clipper->stage.line = clip_first_line; - clipper->stage.tri = clip_first_tri; - clipper->stage.flush = clip_flush; - clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; - clipper->stage.destroy = clip_destroy; - - clipper->plane = draw->plane; - - return &clipper->stage; -} diff --git a/src/mesa/pipe/draw/draw_context.c b/src/mesa/pipe/draw/draw_context.c deleted file mode 100644 index 4be3830316..0000000000 --- a/src/mesa/pipe/draw/draw_context.c +++ /dev/null @@ -1,293 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "draw_context.h" -#include "draw_private.h" - - - -struct draw_context *draw_create( void ) -{ - struct draw_context *draw = CALLOC_STRUCT( draw_context ); - -#if defined(__i386__) || defined(__386__) - draw->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; -#else - draw->use_sse = FALSE; -#endif - - /* create pipeline stages */ - draw->pipeline.wide = draw_wide_stage( draw ); - draw->pipeline.stipple = draw_stipple_stage( draw ); - draw->pipeline.unfilled = draw_unfilled_stage( draw ); - draw->pipeline.twoside = draw_twoside_stage( draw ); - draw->pipeline.offset = draw_offset_stage( draw ); - draw->pipeline.clip = draw_clip_stage( draw ); - draw->pipeline.flatshade = draw_flatshade_stage( draw ); - draw->pipeline.cull = draw_cull_stage( draw ); - draw->pipeline.validate = draw_validate_stage( draw ); - draw->pipeline.first = draw->pipeline.validate; - - ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); - ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); - ASSIGN_4V( draw->plane[2], 0, -1, 0, 1 ); - ASSIGN_4V( draw->plane[3], 0, 1, 0, 1 ); - ASSIGN_4V( draw->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ - ASSIGN_4V( draw->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ - draw->nr_planes = 6; - - /* Statically allocate maximum sized vertices for the cache - could be cleverer... - */ - { - uint i; - const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; - char *tmp = align_malloc(Elements(draw->vcache.vertex) * size, 16); - - for (i = 0; i < Elements(draw->vcache.vertex); i++) - draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * size); - } - - draw->shader_queue_flush = draw_vertex_shader_queue_flush; - - draw->convert_wide_points = TRUE; - draw->convert_wide_lines = TRUE; - - draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ - - draw_vertex_cache_invalidate( draw ); - draw_set_mapped_element_buffer( draw, 0, NULL ); - - return draw; -} - - -void draw_destroy( struct draw_context *draw ) -{ - draw->pipeline.wide->destroy( draw->pipeline.wide ); - draw->pipeline.stipple->destroy( draw->pipeline.stipple ); - draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); - draw->pipeline.twoside->destroy( draw->pipeline.twoside ); - draw->pipeline.offset->destroy( draw->pipeline.offset ); - draw->pipeline.clip->destroy( draw->pipeline.clip ); - draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); - draw->pipeline.cull->destroy( draw->pipeline.cull ); - draw->pipeline.validate->destroy( draw->pipeline.validate ); - if (draw->pipeline.rasterize) - draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); - tgsi_exec_machine_free_data(&draw->machine); - align_free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ - FREE( draw ); -} - - - -void draw_flush( struct draw_context *draw ) -{ - draw_do_flush( draw, DRAW_FLUSH_BACKEND ); -} - - - -/** - * Register new primitive rasterization/rendering state. - * This causes the drawing pipeline to be rebuilt. - */ -void draw_set_rasterizer_state( struct draw_context *draw, - const struct pipe_rasterizer_state *raster ) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - - draw->rasterizer = raster; -} - - -/** - * Plug in the primitive rendering/rasterization stage (which is the last - * stage in the drawing pipeline). - * This is provided by the device driver. - */ -void draw_set_rasterize_stage( struct draw_context *draw, - struct draw_stage *stage ) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - - draw->pipeline.rasterize = stage; -} - - -/** - * Set the draw module's clipping state. - */ -void draw_set_clip_state( struct draw_context *draw, - const struct pipe_clip_state *clip ) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - - assert(clip->nr <= PIPE_MAX_CLIP_PLANES); - memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); - draw->nr_planes = 6 + clip->nr; -} - - -/** - * Set the draw module's viewport state. - */ -void draw_set_viewport_state( struct draw_context *draw, - const struct pipe_viewport_state *viewport ) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->viewport = *viewport; /* struct copy */ -} - - - -void -draw_set_vertex_buffer(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_buffer *buffer) -{ - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - assert(attr < PIPE_ATTRIB_MAX); - draw->vertex_buffer[attr] = *buffer; -} - - -void -draw_set_vertex_element(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_element *element) -{ - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - assert(attr < PIPE_ATTRIB_MAX); - draw->vertex_element[attr] = *element; -} - - -/** - * Tell drawing context where to find mapped vertex buffers. - */ -void -draw_set_mapped_vertex_buffer(struct draw_context *draw, - unsigned attr, const void *buffer) -{ - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - draw->user.vbuffer[attr] = buffer; -} - - -void -draw_set_mapped_constant_buffer(struct draw_context *draw, - const void *buffer) -{ - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - draw->user.constants = buffer; -} - - -/** - * Tells the draw module whether to convert wide points (size != 1) - * into triangles. - */ -void -draw_convert_wide_points(struct draw_context *draw, boolean enable) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->convert_wide_points = enable; -} - - -/** - * Tells the draw module whether to convert wide lines (width != 1) - * into triangles. - */ -void -draw_convert_wide_lines(struct draw_context *draw, boolean enable) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->convert_wide_lines = enable; -} - - -/** - * Allocate space for temporary post-transform vertices, such as for clipping. - */ -void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) -{ - assert(!stage->tmp); - - stage->nr_tmps = nr; - - if (nr) { - ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); - unsigned i; - - stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); - } -} - - -void draw_free_temp_verts( struct draw_stage *stage ) -{ - if (stage->tmp) { - FREE( stage->tmp[0] ); - FREE( stage->tmp ); - stage->tmp = NULL; - } -} - - -boolean draw_use_sse(struct draw_context *draw) -{ - return (boolean) draw->use_sse; -} - - -void draw_reset_vertex_ids(struct draw_context *draw) -{ - struct draw_stage *stage = draw->pipeline.first; - - while (stage) { - unsigned i; - - for (i = 0; i < stage->nr_tmps; i++) - stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; - - stage = stage->next; - } - - draw_vertex_cache_reset_vertex_ids(draw); -} diff --git a/src/mesa/pipe/draw/draw_context.h b/src/mesa/pipe/draw/draw_context.h deleted file mode 100644 index ddeb184497..0000000000 --- a/src/mesa/pipe/draw/draw_context.h +++ /dev/null @@ -1,142 +0,0 @@ - -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Public interface into the drawing module. - */ - -/* Authors: Keith Whitwell - */ - - -#ifndef DRAW_CONTEXT_H -#define DRAW_CONTEXT_H - - -#include "pipe/p_state.h" - - -struct vertex_buffer; -struct vertex_info; -struct draw_context; -struct draw_stage; -struct draw_vertex_shader; - - -/** - * Clipmask flags - */ -/*@{*/ -#define CLIP_RIGHT_BIT 0x01 -#define CLIP_LEFT_BIT 0x02 -#define CLIP_TOP_BIT 0x04 -#define CLIP_BOTTOM_BIT 0x08 -#define CLIP_NEAR_BIT 0x10 -#define CLIP_FAR_BIT 0x20 -/*@}*/ - -/** - * Bitshift for each clip flag - */ -/*@{*/ -#define CLIP_RIGHT_SHIFT 0 -#define CLIP_LEFT_SHIFT 1 -#define CLIP_TOP_SHIFT 2 -#define CLIP_BOTTOM_SHIFT 3 -#define CLIP_NEAR_SHIFT 4 -#define CLIP_FAR_SHIFT 5 -/*@}*/ - - -struct draw_context *draw_create( void ); - -void draw_destroy( struct draw_context *draw ); - -void draw_set_viewport_state( struct draw_context *draw, - const struct pipe_viewport_state *viewport ); - -void draw_set_clip_state( struct draw_context *pipe, - const struct pipe_clip_state *clip ); - -void draw_set_rasterizer_state( struct draw_context *draw, - const struct pipe_rasterizer_state *raster ); - -void draw_set_rasterize_stage( struct draw_context *draw, - struct draw_stage *stage ); - -void draw_convert_wide_points(struct draw_context *draw, boolean enable); - -void draw_convert_wide_lines(struct draw_context *draw, boolean enable); - - -struct draw_vertex_shader * -draw_create_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader); -void draw_bind_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs); -void draw_delete_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs); - -boolean draw_use_sse(struct draw_context *draw); - -void draw_set_vertex_buffer(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_buffer *buffer); - -void draw_set_vertex_element(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_element *element); - -void draw_set_mapped_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ); - -void draw_set_mapped_vertex_buffer(struct draw_context *draw, - unsigned attr, const void *buffer); - -void draw_set_mapped_constant_buffer(struct draw_context *draw, - const void *buffer); - - -/*********************************************************************** - * draw_prim.c - */ - -void draw_arrays(struct draw_context *draw, unsigned prim, - unsigned start, unsigned count); - -void draw_flush(struct draw_context *draw); - -/*********************************************************************** - * draw_debug.c - */ -boolean draw_validate_prim( unsigned prim, unsigned length ); -unsigned draw_trim_prim( unsigned mode, unsigned count ); - - - -#endif /* DRAW_CONTEXT_H */ diff --git a/src/mesa/pipe/draw/draw_cull.c b/src/mesa/pipe/draw/draw_cull.c deleted file mode 100644 index 8177b0ac86..0000000000 --- a/src/mesa/pipe/draw/draw_cull.c +++ /dev/null @@ -1,150 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Drawing stage for polygon culling - */ - -/* Authors: Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "draw_private.h" - - -struct cull_stage { - struct draw_stage stage; - unsigned winding; /**< which winding(s) to cull (one of PIPE_WINDING_x) */ -}; - - -static INLINE struct cull_stage *cull_stage( struct draw_stage *stage ) -{ - return (struct cull_stage *)stage; -} - - - - -static void cull_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - /* Window coords: */ - const float *v0 = header->v[0]->data[0]; - const float *v1 = header->v[1]->data[0]; - const float *v2 = header->v[2]->data[0]; - - /* edge vectors e = v0 - v2, f = v1 - v2 */ - const float ex = v0[0] - v2[0]; - const float ey = v0[1] - v2[1]; - const float fx = v1[0] - v2[0]; - const float fy = v1[1] - v2[1]; - - /* det = cross(e,f).z */ - header->det = ex * fy - ey * fx; - - if (header->det != 0) { - /* if (det < 0 then Z points toward camera and triangle is - * counter-clockwise winding. - */ - unsigned winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; - - if ((winding & cull_stage(stage)->winding) == 0) { - /* triangle is not culled, pass to next stage */ - stage->next->tri( stage->next, header ); - } - } -} - -static void cull_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct cull_stage *cull = cull_stage(stage); - - cull->winding = stage->draw->rasterizer->cull_mode; - - stage->tri = cull_tri; - stage->tri( stage, header ); -} - - - -static void cull_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void cull_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void cull_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->tri = cull_first_tri; - stage->next->flush( stage->next, flags ); -} - -static void cull_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void cull_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create a new polygon culling stage. - */ -struct draw_stage *draw_cull_stage( struct draw_context *draw ) -{ - struct cull_stage *cull = CALLOC_STRUCT(cull_stage); - - draw_alloc_temp_verts( &cull->stage, 0 ); - - cull->stage.draw = draw; - cull->stage.next = NULL; - cull->stage.point = cull_point; - cull->stage.line = cull_line; - cull->stage.tri = cull_first_tri; - cull->stage.flush = cull_flush; - cull->stage.reset_stipple_counter = cull_reset_stipple_counter; - cull->stage.destroy = cull_destroy; - - return &cull->stage; -} diff --git a/src/mesa/pipe/draw/draw_debug.c b/src/mesa/pipe/draw/draw_debug.c deleted file mode 100644 index d6220b5f62..0000000000 --- a/src/mesa/pipe/draw/draw_debug.c +++ /dev/null @@ -1,113 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "draw_private.h" -#include "draw_context.h" - - - -static void -draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) -{ - assert(prim >= PIPE_PRIM_POINTS); - assert(prim <= PIPE_PRIM_POLYGON); - - switch (prim) { - case PIPE_PRIM_POINTS: - *first = 1; - *incr = 1; - break; - case PIPE_PRIM_LINES: - *first = 2; - *incr = 2; - break; - case PIPE_PRIM_LINE_STRIP: - *first = 2; - *incr = 1; - break; - case PIPE_PRIM_LINE_LOOP: - *first = 2; - *incr = 1; - break; - case PIPE_PRIM_TRIANGLES: - *first = 3; - *incr = 3; - break; - case PIPE_PRIM_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - break; - case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: - *first = 3; - *incr = 1; - break; - case PIPE_PRIM_QUADS: - *first = 4; - *incr = 4; - break; - case PIPE_PRIM_QUAD_STRIP: - *first = 4; - *incr = 2; - break; - default: - assert(0); - *first = 1; - *incr = 1; - break; - } -} - - -unsigned -draw_trim_prim( unsigned mode, unsigned count ) -{ - unsigned length, first, incr; - - draw_prim_info( mode, &first, &incr ); - - if (count < first) - length = 0; - else - length = count - (count - first) % incr; - - return length; -} - - -boolean -draw_validate_prim( unsigned mode, unsigned count ) -{ - return (count > 0 && - count == draw_trim_prim( mode, count )); -} - diff --git a/src/mesa/pipe/draw/draw_flatshade.c b/src/mesa/pipe/draw/draw_flatshade.c deleted file mode 100644 index 4398abbc60..0000000000 --- a/src/mesa/pipe/draw/draw_flatshade.c +++ /dev/null @@ -1,205 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -/** subclass of draw_stage */ -struct flat_stage -{ - struct draw_stage stage; - - uint num_color_attribs; - uint color_attribs[4]; /* front/back primary/secondary colors */ -}; - - -static INLINE struct flat_stage * -flat_stage(struct draw_stage *stage) -{ - return (struct flat_stage *) stage; -} - - -/** Copy all the color attributes from 'src' vertex to 'dst' vertex */ -static INLINE void copy_colors( struct draw_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - const struct flat_stage *flat = flat_stage(stage); - uint i; - for (i = 0; i < flat->num_color_attribs; i++) { - const uint attr = flat->color_attribs[i]; - COPY_4FV(dst->data[attr], src->data[attr]); - } -} - - -/** Copy all the color attributes from src vertex to dst0 & dst1 vertices */ -static INLINE void copy_colors2( struct draw_stage *stage, - struct vertex_header *dst0, - struct vertex_header *dst1, - const struct vertex_header *src ) -{ - const struct flat_stage *flat = flat_stage(stage); - uint i; - for (i = 0; i < flat->num_color_attribs; i++) { - const uint attr = flat->color_attribs[i]; - COPY_4FV(dst0->data[attr], src->data[attr]); - COPY_4FV(dst1->data[attr], src->data[attr]); - } -} - - -/** - * Flatshade tri. Required for clipping and when unfilled tris are - * active, otherwise handled by hardware. - */ -static void flatshade_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = header->v[2]; - - copy_colors2(stage, tmp.v[0], tmp.v[1], tmp.v[2]); - - stage->next->tri( stage->next, &tmp ); -} - - -/** - * Flatshade line. Required for clipping. - */ -static void flatshade_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = header->v[1]; - - copy_colors(stage, tmp.v[0], tmp.v[1]); - - stage->next->line( stage->next, &tmp ); -} - - -static void flatshade_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void flatshade_init_state( struct draw_stage *stage ) -{ - struct flat_stage *flat = flat_stage(stage); - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; - uint i; - - /* Find which vertex shader outputs are colors, make a list */ - flat->num_color_attribs = 0; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || - vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - flat->color_attribs[flat->num_color_attribs++] = i; - } - } - - stage->line = flatshade_line; - stage->tri = flatshade_tri; -} - -static void flatshade_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - flatshade_init_state( stage ); - stage->tri( stage, header ); -} - -static void flatshade_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - flatshade_init_state( stage ); - stage->line( stage, header ); -} - - -static void flatshade_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->tri = flatshade_first_tri; - stage->line = flatshade_first_line; - stage->next->flush( stage->next, flags ); -} - - -static void flatshade_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void flatshade_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create flatshading drawing stage. - */ -struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) -{ - struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage); - - draw_alloc_temp_verts( &flatshade->stage, 2 ); - - flatshade->stage.draw = draw; - flatshade->stage.next = NULL; - flatshade->stage.point = flatshade_point; - flatshade->stage.line = flatshade_first_line; - flatshade->stage.tri = flatshade_first_tri; - flatshade->stage.flush = flatshade_flush; - flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; - flatshade->stage.destroy = flatshade_destroy; - - return &flatshade->stage; -} - - diff --git a/src/mesa/pipe/draw/draw_offset.c b/src/mesa/pipe/draw/draw_offset.c deleted file mode 100644 index dbc676deae..0000000000 --- a/src/mesa/pipe/draw/draw_offset.c +++ /dev/null @@ -1,186 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief polygon offset state - * - * \author Keith Whitwell - * \author Brian Paul - */ - -#include "pipe/p_util.h" -#include "draw_private.h" - - - -struct offset_stage { - struct draw_stage stage; - - float scale; - float units; -}; - - - -static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) -{ - return (struct offset_stage *) stage; -} - - - - - -/** - * Offset tri Z. Some hardware can handle this, but not usually when - * doing unfilled rendering. - */ -static void do_offset_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct offset_stage *offset = offset_stage(stage); - float inv_det = 1.0f / header->det; - - /* Window coords: - */ - float *v0 = header->v[0]->data[0]; - float *v1 = header->v[1]->data[0]; - float *v2 = header->v[2]->data[0]; - - /* edge vectors e = v0 - v2, f = v1 - v2 */ - float ex = v0[0] - v2[0]; - float ey = v0[1] - v2[1]; - float ez = v0[2] - v2[2]; - float fx = v1[0] - v2[0]; - float fy = v1[1] - v2[1]; - float fz = v1[2] - v2[2]; - - /* (a,b) = cross(e,f).xy */ - float a = ey*fz - ez*fy; - float b = ez*fx - ex*fz; - - float dzdx = FABSF(a * inv_det); - float dzdy = FABSF(b * inv_det); - - float zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; - - /* - * Note: we're applying the offset and clamping per-vertex. - * Ideally, the offset is applied per-fragment prior to fragment shading. - */ - v0[2] = CLAMP(v0[2] + zoffset, 0.0f, 1.0f); - v1[2] = CLAMP(v1[2] + zoffset, 0.0f, 1.0f); - v2[2] = CLAMP(v2[2] + zoffset, 0.0f, 1.0f); - - stage->next->tri( stage->next, header ); -} - - -static void offset_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = dup_vert(stage, header->v[2], 2); - - do_offset_tri( stage, &tmp ); -} - - -static void offset_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct offset_stage *offset = offset_stage(stage); - float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ - - offset->units = stage->draw->rasterizer->offset_units * mrd; - offset->scale = stage->draw->rasterizer->offset_scale; - - stage->tri = offset_tri; - stage->tri( stage, header ); -} - - -static void offset_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void offset_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void offset_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->tri = offset_first_tri; - stage->next->flush( stage->next, flags ); -} - - -static void offset_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void offset_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create polygon offset drawing stage. - */ -struct draw_stage *draw_offset_stage( struct draw_context *draw ) -{ - struct offset_stage *offset = CALLOC_STRUCT(offset_stage); - - draw_alloc_temp_verts( &offset->stage, 3 ); - - offset->stage.draw = draw; - offset->stage.next = NULL; - offset->stage.point = offset_point; - offset->stage.line = offset_line; - offset->stage.tri = offset_first_tri; - offset->stage.flush = offset_flush; - offset->stage.reset_stipple_counter = offset_reset_stipple_counter; - offset->stage.destroy = offset_destroy; - - return &offset->stage; -} diff --git a/src/mesa/pipe/draw/draw_prim.c b/src/mesa/pipe/draw/draw_prim.c deleted file mode 100644 index 51e2242719..0000000000 --- a/src/mesa/pipe/draw/draw_prim.c +++ /dev/null @@ -1,482 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_debug.h" - -#include "draw_private.h" -#include "draw_context.h" - - - -#define RP_NONE 0 -#define RP_POINT 1 -#define RP_LINE 2 -#define RP_TRI 3 - - -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { - RP_POINT, - RP_LINE, - RP_LINE, - RP_LINE, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI -}; - - -static void draw_prim_queue_flush( struct draw_context *draw ) -{ - unsigned i; - - if (0) - debug_printf("Flushing with %d prims, %d verts\n", - draw->pq.queue_nr, draw->vs.queue_nr); - - assert (draw->pq.queue_nr != 0); - - /* NOTE: we cannot save draw->pipeline->first in a local var because - * draw->pipeline->first is often changed by the first call to tri(), - * line(), etc. - */ - if (draw->rasterizer->line_stipple_enable) { - switch (draw->reduced_prim) { - case RP_TRI: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - - draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); - } - break; - case RP_LINE: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - - draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); - } - break; - case RP_POINT: - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); - break; - } - } - else { - switch (draw->reduced_prim) { - case RP_TRI: - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); - break; - case RP_LINE: - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); - break; - case RP_POINT: - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); - break; - } - } - - draw->pq.queue_nr = 0; - draw_vertex_cache_unreference( draw ); -} - - - -void draw_do_flush( struct draw_context *draw, unsigned flags ) -{ - if (0) - debug_printf("Flushing with %d verts, %d prims\n", - draw->vs.queue_nr, - draw->pq.queue_nr ); - - - if (flags >= DRAW_FLUSH_SHADER_QUEUE) { - if (draw->vs.queue_nr) - (*draw->shader_queue_flush)(draw); - - if (flags >= DRAW_FLUSH_PRIM_QUEUE) { - if (draw->pq.queue_nr) - draw_prim_queue_flush(draw); - - if (flags >= DRAW_FLUSH_VERTEX_CACHE) { - draw_vertex_cache_invalidate(draw); - - if (flags >= DRAW_FLUSH_STATE_CHANGE) { - draw->pipeline.first->flush( draw->pipeline.first, flags ); - draw->pipeline.first = draw->pipeline.validate; - draw->reduced_prim = ~0; - } - } - } - } -} - - - -/* Return a pointer to a freshly queued primitive header. Ensure that - * there is room in the vertex cache for a maximum of "nr_verts" new - * vertices. Flush primitive and/or vertex queues if necessary to - * make space. - */ -static struct prim_header *get_queued_prim( struct draw_context *draw, - unsigned nr_verts ) -{ - if (!draw_vertex_cache_check_space( draw, nr_verts )) { -// debug_printf("v"); - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE ); - } - else if (draw->pq.queue_nr == PRIM_QUEUE_LENGTH) { -// debug_printf("p"); - draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); - } - - assert(draw->pq.queue_nr < PRIM_QUEUE_LENGTH); - - return &draw->pq.queue[draw->pq.queue_nr++]; -} - - - -/** - * Add a point to the primitive queue. - * \param i0 index into user's vertex arrays - */ -static void do_point( struct draw_context *draw, - unsigned i0 ) -{ - struct prim_header *prim = get_queued_prim( draw, 1 ); - - prim->reset_line_stipple = 0; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->vcache.get_vertex( draw, i0 ); -} - - -/** - * Add a line to the primitive queue. - * \param i0 index into user's vertex arrays - * \param i1 index into user's vertex arrays - */ -static void do_line( struct draw_context *draw, - boolean reset_stipple, - unsigned i0, - unsigned i1 ) -{ - struct prim_header *prim = get_queued_prim( draw, 2 ); - - prim->reset_line_stipple = reset_stipple; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->vcache.get_vertex( draw, i0 ); - prim->v[1] = draw->vcache.get_vertex( draw, i1 ); -} - -/** - * Add a triangle to the primitive queue. - */ -static void do_triangle( struct draw_context *draw, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - - prim->reset_line_stipple = 1; - prim->edgeflags = ~0; - prim->pad = 0; - prim->v[0] = draw->vcache.get_vertex( draw, i0 ); - prim->v[1] = draw->vcache.get_vertex( draw, i1 ); - prim->v[2] = draw->vcache.get_vertex( draw, i2 ); -} - -static void do_ef_triangle( struct draw_context *draw, - boolean reset_stipple, - unsigned ef_mask, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - struct vertex_header *v0 = draw->vcache.get_vertex( draw, i0 ); - struct vertex_header *v1 = draw->vcache.get_vertex( draw, i1 ); - struct vertex_header *v2 = draw->vcache.get_vertex( draw, i2 ); - - prim->reset_line_stipple = reset_stipple; - - prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | - (v1->edgeflag << 1) | - (v2->edgeflag << 2)); - prim->pad = 0; - prim->v[0] = v0; - prim->v[1] = v1; - prim->v[2] = v2; -} - - -static void do_ef_quad( struct draw_context *draw, - unsigned v0, - unsigned v1, - unsigned v2, - unsigned v3 ) -{ - const unsigned omitEdge2 = ~(1 << 1); - const unsigned omitEdge3 = ~(1 << 2); - do_ef_triangle( draw, 1, omitEdge2, v0, v1, v3 ); - do_ef_triangle( draw, 0, omitEdge3, v1, v2, v3 ); -} - -static void do_quad( struct draw_context *draw, - unsigned v0, - unsigned v1, - unsigned v2, - unsigned v3 ) -{ - do_triangle( draw, v0, v1, v3 ); - do_triangle( draw, v1, v2, v3 ); -} - - -/** - * Main entrypoint to draw some number of points/lines/triangles - */ -static void -draw_prim( struct draw_context *draw, - unsigned prim, unsigned start, unsigned count ) -{ - unsigned i; - boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); - -// debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - - switch (prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i ++) { - do_point( draw, - start + i ); - } - break; - - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) { - do_line( draw, - TRUE, - start + i + 0, - start + i + 1); - } - break; - - case PIPE_PRIM_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, /* XXX: only if vb not split */ - start + i - 1, - start + i ); - } - - do_line( draw, - 0, - start + count - 1, - start + 0 ); - } - break; - - case PIPE_PRIM_LINE_STRIP: - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, - start + i - 1, - start + i ); - } - break; - - case PIPE_PRIM_TRIANGLES: - if (unfilled) { - for (i = 0; i+2 < count; i += 3) { - do_ef_triangle( draw, - 1, - ~0, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - else { - for (i = 0; i+2 < count; i += 3) { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - do_triangle( draw, - start + i + 1, - start + i + 0, - start + i + 2 ); - } - else { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - case PIPE_PRIM_TRIANGLE_FAN: - if (count >= 3) { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - - case PIPE_PRIM_QUADS: - if (unfilled) { - for (i = 0; i+3 < count; i += 4) { - do_ef_quad( draw, - start + i + 0, - start + i + 1, - start + i + 2, - start + i + 3); - } - } - else { - for (i = 0; i+3 < count; i += 4) { - do_quad( draw, - start + i + 0, - start + i + 1, - start + i + 2, - start + i + 3); - } - } - break; - - case PIPE_PRIM_QUAD_STRIP: - if (unfilled) { - for (i = 0; i+3 < count; i += 2) { - do_ef_quad( draw, - start + i + 2, - start + i + 0, - start + i + 1, - start + i + 3); - } - } - else { - for (i = 0; i+3 < count; i += 2) { - do_quad( draw, - start + i + 2, - start + i + 0, - start + i + 1, - start + i + 3); - } - } - break; - - case PIPE_PRIM_POLYGON: - if (unfilled) { - unsigned ef_mask = (1<<2) | (1<<0); - - for (i = 0; i+2 < count; i++) { - - if (i + 3 >= count) - ef_mask |= (1<<1); - - do_ef_triangle( draw, - i == 0, - ef_mask, - start + i + 1, - start + i + 2, - start + 0); - - ef_mask &= ~(1<<2); - } - } - else { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + i + 1, - start + i + 2, - start + 0); - } - } - break; - - default: - assert(0); - break; - } -} - - - - -/** - * Draw vertex arrays - * This is the main entrypoint into the drawing module. - * \param prim one of PIPE_PRIM_x - * \param start index of first vertex to draw - * \param count number of vertices to draw - */ -void -draw_arrays(struct draw_context *draw, unsigned prim, - unsigned start, unsigned count) -{ - if (reduced_prim[prim] != draw->reduced_prim) { - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->reduced_prim = reduced_prim[prim]; - } - - /* drawing done here: */ - draw_prim(draw, prim, start, count); -} - - diff --git a/src/mesa/pipe/draw/draw_private.h b/src/mesa/pipe/draw/draw_private.h deleted file mode 100644 index b17eaaed65..0000000000 --- a/src/mesa/pipe/draw/draw_private.h +++ /dev/null @@ -1,346 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Private data structures, etc for the draw module. - */ - - -/** - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#ifndef DRAW_PRIVATE_H -#define DRAW_PRIVATE_H - - -#include "pipe/p_state.h" -#include "pipe/p_defines.h" - -#include "x86/rtasm/x86sse.h" -#include "pipe/tgsi/exec/tgsi_exec.h" - - -struct gallivm_prog; -struct gallivm_cpu_engine; - -/** - * Basic vertex info. - * Carry some useful information around with the vertices in the prim pipe. - */ -struct vertex_header { - unsigned clipmask:12; - unsigned edgeflag:1; - unsigned pad:3; - unsigned vertex_id:16; - - float clip[4]; - - float data[][4]; /* Note variable size */ -}; - -/* NOTE: It should match vertex_id size above */ -#define UNDEFINED_VERTEX_ID 0xffff - -/* XXX This is too large */ -#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) - - - -/** - * Basic info for a point/line/triangle primitive. - */ -struct prim_header { - float det; /**< front/back face determinant */ - unsigned reset_line_stipple:1; - unsigned edgeflags:3; - unsigned pad:28; - struct vertex_header *v[3]; /**< 1 to 3 vertex pointers */ -}; - - - -struct draw_context; - -/** - * Base class for all primitive drawing stages. - */ -struct draw_stage -{ - struct draw_context *draw; /**< parent context */ - - struct draw_stage *next; /**< next stage in pipeline */ - - struct vertex_header **tmp; /**< temp vert storage, such as for clipping */ - unsigned nr_tmps; - - void (*point)( struct draw_stage *, - struct prim_header * ); - - void (*line)( struct draw_stage *, - struct prim_header * ); - - void (*tri)( struct draw_stage *, - struct prim_header * ); - - void (*flush)( struct draw_stage *, - unsigned flags ); - - void (*reset_stipple_counter)( struct draw_stage * ); - - void (*destroy)( struct draw_stage * ); -}; - - -#define PRIM_QUEUE_LENGTH 16 -#define VCACHE_SIZE 32 -#define VCACHE_OVERFLOW 4 -#define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ - -/** - * Private version of the compiled vertex_shader - */ -struct draw_vertex_shader { - const struct pipe_shader_state *state; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif -#ifdef MESA_LLVM - struct gallivm_prog *llvm_prog; -#endif -}; - - -/* Internal function for vertex fetch. - */ -typedef void (*fetch_func)(const void *ptr, float *attrib); -typedef void (*full_fetch_func)( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ); - - - -/** - * Private context for the drawing module. - */ -struct draw_context -{ - /** Drawing/primitive pipeline stages */ - struct { - struct draw_stage *first; /**< one of the following */ - - struct draw_stage *validate; - - /* stages (in logical order) */ - struct draw_stage *flatshade; - struct draw_stage *clip; - struct draw_stage *cull; - struct draw_stage *twoside; - struct draw_stage *offset; - struct draw_stage *unfilled; - struct draw_stage *stipple; - struct draw_stage *wide; - struct draw_stage *rasterize; - } pipeline; - - /* pipe state that we need: */ - const struct pipe_rasterizer_state *rasterizer; - struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - const struct draw_vertex_shader *vertex_shader; - - uint num_vs_outputs; /**< convenience, from vertex_shader */ - - /* user-space vertex data, buffers */ - struct { - /** vertex element/index buffer (ex: glDrawElements) */ - const void *elts; - /** bytes per index (0, 1, 2 or 4) */ - unsigned eltSize; - - /** vertex arrays */ - const void *vbuffer[PIPE_ATTRIB_MAX]; - - /** constant buffer (for vertex shader) */ - const void *constants; - } user; - - /* Clip derived state: - */ - float plane[12][4]; - unsigned nr_planes; - - boolean convert_wide_points; /**< convert wide points to tris? */ - boolean convert_wide_lines; /**< convert side lines to tris? */ - - unsigned reduced_prim; - - /** TGSI program interpreter runtime state */ - struct tgsi_exec_machine machine; - - /* Vertex fetch internal state - */ - struct { - const ubyte *src_ptr[PIPE_ATTRIB_MAX]; - unsigned pitch[PIPE_ATTRIB_MAX]; - fetch_func fetch[PIPE_ATTRIB_MAX]; - unsigned nr_attrs; - full_fetch_func fetch_func; - } vertex_fetch; - - /* Post-tnl vertex cache: - */ - struct { - unsigned referenced; /**< bitfield */ - unsigned idx[VCACHE_SIZE + VCACHE_OVERFLOW]; - struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; - unsigned overflow; - - /** To find space in the vertex cache: */ - struct vertex_header *(*get_vertex)( struct draw_context *draw, - unsigned i ); - } vcache; - - /* Vertex shader queue: - */ - struct { - struct { - unsigned elt; /**< index into the user's vertex arrays */ - struct vertex_header *dest; /**< points into vcache.vertex[] array */ - } queue[VS_QUEUE_LENGTH]; - unsigned queue_nr; - } vs; - - /** - * Run the vertex shader on all vertices in the vertex queue. - */ - void (*shader_queue_flush)(struct draw_context *draw); - - /* Prim pipeline queue: - */ - struct { - /* Need to queue up primitives until their vertices have been - * transformed by a vs queue flush. - */ - struct prim_header queue[PRIM_QUEUE_LENGTH]; - unsigned queue_nr; - } pq; - - int use_sse : 1; -#ifdef MESA_LLVM - struct gallivm_cpu_engine *engine; -#endif - - void *driver_private; -}; - - - -extern struct draw_stage *draw_unfilled_stage( struct draw_context *context ); -extern struct draw_stage *draw_twoside_stage( struct draw_context *context ); -extern struct draw_stage *draw_offset_stage( struct draw_context *context ); -extern struct draw_stage *draw_clip_stage( struct draw_context *context ); -extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); -extern struct draw_stage *draw_cull_stage( struct draw_context *context ); -extern struct draw_stage *draw_stipple_stage( struct draw_context *context ); -extern struct draw_stage *draw_wide_stage( struct draw_context *context ); -extern struct draw_stage *draw_validate_stage( struct draw_context *context ); - - -extern void draw_free_temp_verts( struct draw_stage *stage ); - -extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); - -extern void draw_reset_vertex_ids( struct draw_context *draw ); - - -extern int draw_vertex_cache_check_space( struct draw_context *draw, - unsigned nr_verts ); - -extern void draw_vertex_cache_invalidate( struct draw_context *draw ); -extern void draw_vertex_cache_unreference( struct draw_context *draw ); -extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); - - -extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); -#ifdef MESA_LLVM -extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw ); -#endif - -struct tgsi_exec_machine; - -extern void draw_update_vertex_fetch( struct draw_context *draw ); - - -#define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ -#define DRAW_FLUSH_PRIM_QUEUE 0x2 -#define DRAW_FLUSH_VERTEX_CACHE 0x4 -#define DRAW_FLUSH_STATE_CHANGE 0x8 -#define DRAW_FLUSH_BACKEND 0x10 - - -void draw_do_flush( struct draw_context *draw, unsigned flags ); - - - -/** - * Get a writeable copy of a vertex. - * \param stage drawing stage info - * \param vert the vertex to copy (source) - * \param idx index into stage's tmp[] array to put the copy (dest) - * \return pointer to the copied vertex - */ -static INLINE struct vertex_header * -dup_vert( struct draw_stage *stage, - const struct vertex_header *vert, - unsigned idx ) -{ - struct vertex_header *tmp = stage->tmp[idx]; - const uint vsize = sizeof(struct vertex_header) - + stage->draw->num_vs_outputs * 4 * sizeof(float); - memcpy(tmp, vert, vsize); - tmp->vertex_id = UNDEFINED_VERTEX_ID; - return tmp; -} - -static INLINE float -dot4(const float *a, const float *b) -{ - float result = (a[0]*b[0] + - a[1]*b[1] + - a[2]*b[2] + - a[3]*b[3]); - - return result; -} - -#endif /* DRAW_PRIVATE_H */ diff --git a/src/mesa/pipe/draw/draw_stipple.c b/src/mesa/pipe/draw/draw_stipple.c deleted file mode 100644 index 506f33512c..0000000000 --- a/src/mesa/pipe/draw/draw_stipple.c +++ /dev/null @@ -1,239 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -/* Implement line stipple by cutting lines up into smaller lines. - * There are hundreds of ways to implement line stipple, this is one - * choice that should work in all situations, requires no state - * manipulations, but with a penalty in terms of large amounts of - * generated geometry. - */ - - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -/** Subclass of draw_stage */ -struct stipple_stage { - struct draw_stage stage; - float counter; - uint pattern; - uint factor; -}; - - -static INLINE struct stipple_stage * -stipple_stage(struct draw_stage *stage) -{ - return (struct stipple_stage *) stage; -} - - -/** - * Compute interpolated vertex attributes for 'dst' at position 't' - * between 'v0' and 'v1'. - * XXX using linear interpolation for all attribs at this time. - */ -static void -screen_interp( struct draw_context *draw, - struct vertex_header *dst, - float t, - const struct vertex_header *v0, - const struct vertex_header *v1 ) -{ - uint attr; - for (attr = 0; attr < draw->num_vs_outputs; attr++) { - const float *val0 = v0->data[attr]; - const float *val1 = v1->data[attr]; - float *newv = dst->data[attr]; - uint i; - for (i = 0; i < 4; i++) { - newv[i] = val0[i] + t * (val1[i] - val0[i]); - } - } -} - - -static void -emit_segment(struct draw_stage *stage, struct prim_header *header, - float t0, float t1) -{ - struct vertex_header *v0new = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1new = dup_vert(stage, header->v[1], 1); - struct prim_header newprim = *header; - - if (t0 > 0.0) { - screen_interp( stage->draw, v0new, t0, header->v[0], header->v[1] ); - newprim.v[0] = v0new; - } - - if (t1 < 1.0) { - screen_interp( stage->draw, v1new, t1, header->v[0], header->v[1] ); - newprim.v[1] = v1new; - } - - stage->next->line( stage->next, &newprim ); -} - - -static INLINE unsigned -stipple_test(int counter, ushort pattern, int factor) -{ - int b = (counter / factor) & 0xf; - return (1 << b) & pattern; -} - - -static void -stipple_line(struct draw_stage *stage, struct prim_header *header) -{ - struct stipple_stage *stipple = stipple_stage(stage); - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - const float *pos0 = v0->data[0]; - const float *pos1 = v1->data[0]; - float start = 0; - int state = 0; - - float x0 = pos0[0]; - float x1 = pos1[0]; - float y0 = pos0[1]; - float y1 = pos1[1]; - - float dx = x0 > x1 ? x0 - x1 : x1 - x0; - float dy = y0 > y1 ? y0 - y1 : y1 - y0; - - float length = MAX2(dx, dy); - int i; - - /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table. - */ - for (i = 0; i < length; i++) { - int result = stipple_test( (int) stipple->counter+i, - (ushort) stipple->pattern, stipple->factor ); - if (result != state) { - /* changing from "off" to "on" or vice versa */ - if (state) { - if (start != i) { - /* finishing an "on" segment */ - emit_segment( stage, header, start / length, i / length ); - } - } - else { - /* starting an "on" segment */ - start = (float) i; - } - state = result; - } - } - - if (state && start < length) - emit_segment( stage, header, start / length, 1.0 ); - - stipple->counter += length; -} - - -static void -reset_stipple_counter(struct draw_stage *stage) -{ - struct stipple_stage *stipple = stipple_stage(stage); - stipple->counter = 0; - stage->next->reset_stipple_counter( stage->next ); -} - - -static void -stipple_first_line(struct draw_stage *stage, - struct prim_header *header) -{ - struct stipple_stage *stipple = stipple_stage(stage); - struct draw_context *draw = stage->draw; - - stipple->pattern = draw->rasterizer->line_stipple_pattern; - stipple->factor = draw->rasterizer->line_stipple_factor + 1; - - stage->line = stipple_line; - stage->line( stage, header ); -} - - -static void -stipple_flush(struct draw_stage *stage, unsigned flags) -{ - stage->line = stipple_first_line; - stage->next->flush( stage->next, flags ); -} - - -static void -passthrough_point(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->point( stage->next, header ); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} - - -static void -stipple_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create line stippler stage - */ -struct draw_stage *draw_stipple_stage( struct draw_context *draw ) -{ - struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); - - draw_alloc_temp_verts( &stipple->stage, 2 ); - - stipple->stage.draw = draw; - stipple->stage.next = NULL; - stipple->stage.point = passthrough_point; - stipple->stage.line = stipple_first_line; - stipple->stage.tri = passthrough_tri; - stipple->stage.reset_stipple_counter = reset_stipple_counter; - stipple->stage.flush = stipple_flush; - stipple->stage.destroy = stipple_destroy; - - return &stipple->stage; -} diff --git a/src/mesa/pipe/draw/draw_twoside.c b/src/mesa/pipe/draw/draw_twoside.c deleted file mode 100644 index 1c38957987..0000000000 --- a/src/mesa/pipe/draw/draw_twoside.c +++ /dev/null @@ -1,203 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -struct twoside_stage { - struct draw_stage stage; - float sign; /**< +1 or -1 */ - uint attrib_front0, attrib_back0; - uint attrib_front1, attrib_back1; -}; - - -static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage ) -{ - return (struct twoside_stage *)stage; -} - - - - -/** - * Copy back color(s) to front color(s). - */ -static INLINE struct vertex_header * -copy_bfc( struct twoside_stage *twoside, - const struct vertex_header *v, - unsigned idx ) -{ - struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); - - if (twoside->attrib_back0) { - COPY_4FV(tmp->data[twoside->attrib_front0], - tmp->data[twoside->attrib_back0]); - } - if (twoside->attrib_back1) { - COPY_4FV(tmp->data[twoside->attrib_front1], - tmp->data[twoside->attrib_back1]); - } - - return tmp; -} - - -/* Twoside tri: - */ -static void twoside_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - - if (header->det * twoside->sign < 0.0) { - /* this is a back-facing triangle */ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - /* copy back attribs to front attribs */ - tmp.v[0] = copy_bfc(twoside, header->v[0], 0); - tmp.v[1] = copy_bfc(twoside, header->v[1], 1); - tmp.v[2] = copy_bfc(twoside, header->v[2], 2); - - stage->next->tri( stage->next, &tmp ); - } - else { - stage->next->tri( stage->next, header ); - } -} - - -static void twoside_line( struct draw_stage *stage, - struct prim_header *header ) -{ - /* pass-through */ - stage->next->line( stage->next, header ); -} - - -static void twoside_point( struct draw_stage *stage, - struct prim_header *header ) -{ - /* pass-through */ - stage->next->point( stage->next, header ); -} - - -static void twoside_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; - uint i; - - twoside->attrib_front0 = 0; - twoside->attrib_front1 = 0; - twoside->attrib_back0 = 0; - twoside->attrib_back1 = 0; - - /* Find which vertex shader outputs are front/back colors */ - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { - if (vs->output_semantic_index[i] == 0) - twoside->attrib_front0 = i; - else - twoside->attrib_front1 = i; - } - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - if (vs->output_semantic_index[i] == 0) - twoside->attrib_back0 = i; - else - twoside->attrib_back1 = i; - } - } - - if (!twoside->attrib_back0) - twoside->attrib_front0 = 0; - - if (!twoside->attrib_back1) - twoside->attrib_front1 = 0; - - /* - * We'll multiply the primitive's determinant by this sign to determine - * if the triangle is back-facing (negative). - * sign = -1 for CCW, +1 for CW - */ - twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; - - stage->tri = twoside_tri; - stage->tri( stage, header ); -} - - -static void twoside_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->tri = twoside_first_tri; - stage->next->flush( stage->next, flags ); -} - - -static void twoside_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void twoside_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create twoside pipeline stage. - */ -struct draw_stage *draw_twoside_stage( struct draw_context *draw ) -{ - struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); - - draw_alloc_temp_verts( &twoside->stage, 3 ); - - twoside->stage.draw = draw; - twoside->stage.next = NULL; - twoside->stage.point = twoside_point; - twoside->stage.line = twoside_line; - twoside->stage.tri = twoside_first_tri; - twoside->stage.flush = twoside_flush; - twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; - twoside->stage.destroy = twoside_destroy; - - return &twoside->stage; -} diff --git a/src/mesa/pipe/draw/draw_unfilled.c b/src/mesa/pipe/draw/draw_unfilled.c deleted file mode 100644 index 8777cfdfc8..0000000000 --- a/src/mesa/pipe/draw/draw_unfilled.c +++ /dev/null @@ -1,206 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Drawing stage for handling glPolygonMode(line/point). - * Convert triangles to points or lines as needed. - */ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "draw_private.h" - - -struct unfilled_stage { - struct draw_stage stage; - - /** [0] = front face, [1] = back face. - * legal values: PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE, - * and PIPE_POLYGON_MODE_POINT, - */ - unsigned mode[2]; -}; - - -static INLINE struct unfilled_stage *unfilled_stage( struct draw_stage *stage ) -{ - return (struct unfilled_stage *)stage; -} - - - -static void point( struct draw_stage *stage, - struct vertex_header *v0 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - stage->next->point( stage->next, &tmp ); -} - -static void line( struct draw_stage *stage, - struct vertex_header *v0, - struct vertex_header *v1 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - tmp.v[1] = v1; - stage->next->line( stage->next, &tmp ); -} - - -static void points( struct draw_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - - if (header->edgeflags & 0x1) point( stage, v0 ); - if (header->edgeflags & 0x2) point( stage, v1 ); - if (header->edgeflags & 0x4) point( stage, v2 ); -} - - -static void lines( struct draw_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - -#if 0 - assert(((header->edgeflags & 0x1) >> 0) == header->v[0]->edgeflag); - assert(((header->edgeflags & 0x2) >> 1) == header->v[1]->edgeflag); - assert(((header->edgeflags & 0x4) >> 2) == header->v[2]->edgeflag); -#endif - - if (header->edgeflags & 0x1) line( stage, v0, v1 ); - if (header->edgeflags & 0x2) line( stage, v1, v2 ); - if (header->edgeflags & 0x4) line( stage, v2, v0 ); -} - - -/* Unfilled tri: - * - * Note edgeflags in the vertex struct is not sufficient as we will - * need to manipulate them when decomposing primitives??? - */ -static void unfilled_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - unsigned mode = unfilled->mode[header->det >= 0.0]; - - switch (mode) { - case PIPE_POLYGON_MODE_FILL: - stage->next->tri( stage->next, header ); - break; - case PIPE_POLYGON_MODE_LINE: - lines( stage, header ); - break; - case PIPE_POLYGON_MODE_POINT: - points( stage, header ); - break; - default: - abort(); - } -} - - -static void unfilled_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - - unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */ - unfilled->mode[1] = stage->draw->rasterizer->fill_cw; /* back */ - - stage->tri = unfilled_tri; - stage->tri( stage, header ); -} - - -static void unfilled_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void unfilled_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void unfilled_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->next->flush( stage->next, flags ); - - stage->tri = unfilled_first_tri; -} - - -static void unfilled_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void unfilled_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create unfilled triangle stage. - */ -struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) -{ - struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); - - draw_alloc_temp_verts( &unfilled->stage, 0 ); - - unfilled->stage.draw = draw; - unfilled->stage.next = NULL; - unfilled->stage.tmp = NULL; - unfilled->stage.point = unfilled_point; - unfilled->stage.line = unfilled_line; - unfilled->stage.tri = unfilled_first_tri; - unfilled->stage.flush = unfilled_flush; - unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; - unfilled->stage.destroy = unfilled_destroy; - - return &unfilled->stage; -} diff --git a/src/mesa/pipe/draw/draw_validate.c b/src/mesa/pipe/draw/draw_validate.c deleted file mode 100644 index 4375ebabbc..0000000000 --- a/src/mesa/pipe/draw/draw_validate.c +++ /dev/null @@ -1,185 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "draw_private.h" - - - - - -/** - * Rebuild the rendering pipeline. - */ -static struct draw_stage *validate_pipeline( struct draw_stage *stage ) -{ - struct draw_context *draw = stage->draw; - struct draw_stage *next = draw->pipeline.rasterize; - int need_det = 0; - int precalc_flat = 0; - - /* Set the validate's next stage to the rasterize stage, so that it - * can be found later if needed for flushing. - */ - stage->next = next; - - /* - * NOTE: we build up the pipeline in end-to-start order. - * - * TODO: make the current primitive part of the state and build - * shorter pipelines for lines & points. - */ - - if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines) || - (draw->rasterizer->point_size != 1.0 && draw->convert_wide_points) || - draw->rasterizer->point_sprite) { - draw->pipeline.wide->next = next; - next = draw->pipeline.wide; - } - - if (draw->rasterizer->line_stipple_enable) { - draw->pipeline.stipple->next = next; - next = draw->pipeline.stipple; - precalc_flat = 1; /* only needed for lines really */ - } - - if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { - draw->pipeline.unfilled->next = next; - next = draw->pipeline.unfilled; - precalc_flat = 1; /* only needed for triangles really */ - need_det = 1; - } - - if (draw->rasterizer->flatshade && precalc_flat) { - draw->pipeline.flatshade->next = next; - next = draw->pipeline.flatshade; - } - - if (draw->rasterizer->offset_cw || - draw->rasterizer->offset_ccw) { - draw->pipeline.offset->next = next; - next = draw->pipeline.offset; - need_det = 1; - } - - if (draw->rasterizer->light_twoside) { - draw->pipeline.twoside->next = next; - next = draw->pipeline.twoside; - need_det = 1; - } - - /* Always run the cull stage as we calculate determinant there - * also. - * - * This can actually be a win as culling out the triangles can lead - * to less work emitting vertices, smaller vertex buffers, etc. - * It's difficult to say whether this will be true in general. - */ - if (need_det || draw->rasterizer->cull_mode) { - draw->pipeline.cull->next = next; - next = draw->pipeline.cull; - } - - /* Clip stage - */ - if (!draw->rasterizer->bypass_clipping) - { - draw->pipeline.clip->next = next; - next = draw->pipeline.clip; - } - - - draw->pipeline.first = next; - return next; -} - -static void validate_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->tri( pipeline, header ); -} - -static void validate_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->line( pipeline, header ); -} - -static void validate_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->point( pipeline, header ); -} - -static void validate_reset_stipple_counter( struct draw_stage *stage ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->reset_stipple_counter( pipeline ); -} - -static void validate_flush( struct draw_stage *stage, - unsigned flags ) -{ - /* May need to pass a backend flush on to the rasterize stage. - */ - if (stage->next) - stage->next->flush( stage->next, flags ); -} - - -static void validate_destroy( struct draw_stage *stage ) -{ - FREE( stage ); -} - - -/** - * Create validate pipeline stage. - */ -struct draw_stage *draw_validate_stage( struct draw_context *draw ) -{ - struct draw_stage *stage = CALLOC_STRUCT(draw_stage); - - stage->draw = draw; - stage->next = NULL; - stage->point = validate_point; - stage->line = validate_line; - stage->tri = validate_tri; - stage->flush = validate_flush; - stage->reset_stipple_counter = validate_reset_stipple_counter; - stage->destroy = validate_destroy; - - return stage; -} diff --git a/src/mesa/pipe/draw/draw_vbuf.c b/src/mesa/pipe/draw/draw_vbuf.c deleted file mode 100644 index 71ac73912b..0000000000 --- a/src/mesa/pipe/draw/draw_vbuf.c +++ /dev/null @@ -1,570 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Vertex buffer drawing stage. - * - * \author José Fonseca - * \author Keith Whitwell - */ - - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" - -#include "draw_vbuf.h" -#include "draw_private.h" -#include "draw_vertex.h" -#include "draw_vf.h" - - -/** - * Vertex buffer emit stage. - */ -struct vbuf_stage { - struct draw_stage stage; /**< This must be first (base class) */ - - struct vbuf_render *render; - - const struct vertex_info *vinfo; - - /** Vertex size in bytes */ - unsigned vertex_size; - - struct draw_vertex_fetch *vf; - - /* FIXME: we have no guarantee that 'unsigned' is 32bit */ - - /** Vertices in hardware format */ - unsigned *vertices; - unsigned *vertex_ptr; - unsigned max_vertices; - unsigned nr_vertices; - - /** Indices */ - ushort *indices; - unsigned max_indices; - unsigned nr_indices; - - /** Pipe primitive */ - unsigned prim; -}; - - -/** - * Basically a cast wrapper. - */ -static INLINE struct vbuf_stage * -vbuf_stage( struct draw_stage *stage ) -{ - assert(stage); - return (struct vbuf_stage *)stage; -} - - -static void vbuf_flush_indices( struct vbuf_stage *vbuf ); -static void vbuf_flush_vertices( struct vbuf_stage *vbuf ); -static void vbuf_alloc_vertices( struct vbuf_stage *vbuf ); - - -static INLINE boolean -overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) -{ - unsigned long used = (unsigned long) ((char *)ptr - (char *)map); - return (used + bytes) > bufsz; -} - - -static INLINE void -check_space( struct vbuf_stage *vbuf, unsigned nr ) -{ - if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { - vbuf_flush_vertices(vbuf); - vbuf_alloc_vertices(vbuf); - } - - if (vbuf->nr_indices + nr > vbuf->max_indices ) - vbuf_flush_indices(vbuf); -} - - -#if 0 -static INLINE void -dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) -{ - assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); - unsigned i, j, k; - - for (i = 0; i < vinfo->num_attribs; i++) { - j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - debug_printf("EMIT_OMIT:"); - break; - case EMIT_ALL: - assert(i == 0); - assert(j == 0); - debug_printf("EMIT_ALL:\t"); - for(k = 0; k < vinfo->size*4; ++k) - debug_printf("%02x ", *data++); - break; - case EMIT_1F: - debug_printf("EMIT_1F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_1F_PSIZE: - debug_printf("EMIT_1F_PSIZE:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_2F: - debug_printf("EMIT_2F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_3F: - debug_printf("EMIT_3F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - data += sizeof(float); - break; - case EMIT_4F: - debug_printf("EMIT_4F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_4UB: - debug_printf("EMIT_4UB:\t"); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - break; - default: - assert(0); - } - debug_printf("\n"); - } - debug_printf("\n"); -} -#endif - - -/** - * Extract the needed fields from post-transformed vertex and emit - * a hardware(driver) vertex. - * Recall that the vertices are constructed by the 'draw' module and - * have a couple of slots at the beginning (1-dword header, 4-dword - * clip pos) that we ignore here. We only use the vertex->data[] fields. - */ -static INLINE void -emit_vertex( struct vbuf_stage *vbuf, - struct vertex_header *vertex ) -{ -#if 0 - debug_printf("emit vertex %d to %p\n", - vbuf->nr_vertices, vbuf->vertex_ptr); -#endif - - if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { - if(vertex->vertex_id < vbuf->nr_vertices) - return; - else - debug_printf("Bad vertex id 0x%04x (>= 0x%04x)\n", - vertex->vertex_id, vbuf->nr_vertices); - return; - } - - vertex->vertex_id = vbuf->nr_vertices++; - - if(!vbuf->vf) { - const struct vertex_info *vinfo = vbuf->vinfo; - uint i; - uint count = 0; /* for debug/sanity */ - - assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); - - for (i = 0; i < vinfo->num_attribs; i++) { - uint j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; - case EMIT_ALL: - /* just copy the whole vertex as-is to the vbuf */ - assert(i == 0); - assert(j == 0); - memcpy(vbuf->vertex_ptr, vertex, vinfo->size * 4); - vbuf->vertex_ptr += vinfo->size; - count += vinfo->size; - break; - case EMIT_1F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - count++; - break; - case EMIT_1F_PSIZE: - *vbuf->vertex_ptr++ = fui(vbuf->stage.draw->rasterizer->point_size); - count++; - break; - case EMIT_2F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - count += 2; - break; - case EMIT_3F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); - count += 3; - break; - case EMIT_4F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][3]); - count += 4; - break; - case EMIT_4UB: - *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), - float_to_ubyte( vertex->data[j][1] ), - float_to_ubyte( vertex->data[j][0] ), - float_to_ubyte( vertex->data[j][3] )); - count += 1; - break; - default: - assert(0); - } - } - assert(count == vinfo->size); -#if 0 - { - static float data[256]; - draw_vf_emit_vertex(vbuf->vf, vertex, data); - if(memcmp((uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size, data, vbuf->vertex_size)) { - debug_printf("With VF:\n"); - dump_emitted_vertex(vbuf->vinfo, (uint8_t *)data); - debug_printf("Without VF:\n"); - dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size); - assert(0); - } - } -#endif - } - else { - draw_vf_emit_vertex(vbuf->vf, vertex, vbuf->vertex_ptr); - - vbuf->vertex_ptr += vbuf->vertex_size/4; - } -} - - -static void -vbuf_tri( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - check_space( vbuf, 3 ); - - for (i = 0; i < 3; i++) { - emit_vertex( vbuf, prim->v[i] ); - - vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; - } -} - - -static void -vbuf_line( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - check_space( vbuf, 2 ); - - for (i = 0; i < 2; i++) { - emit_vertex( vbuf, prim->v[i] ); - - vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; - } -} - - -static void -vbuf_point( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - check_space( vbuf, 1 ); - - emit_vertex( vbuf, prim->v[0] ); - - vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[0]->vertex_id; -} - - -/** - * Set the prim type for subsequent vertices. - * This may result in a new vertex size. The existing vbuffer (if any) - * will be flushed if needed and a new one allocated. - */ -static void -vbuf_set_prim( struct vbuf_stage *vbuf, uint newprim ) -{ - const struct vertex_info *vinfo; - unsigned vertex_size; - - assert(newprim == PIPE_PRIM_POINTS || - newprim == PIPE_PRIM_LINES || - newprim == PIPE_PRIM_TRIANGLES); - - vbuf->prim = newprim; - vbuf->render->set_primitive(vbuf->render, newprim); - - vinfo = vbuf->render->get_vertex_info(vbuf->render); - vertex_size = vinfo->size * sizeof(float); - - if (vertex_size != vbuf->vertex_size) - vbuf_flush_vertices(vbuf); - - vbuf->vinfo = vinfo; - vbuf->vertex_size = vertex_size; - if(vbuf->vf) - draw_vf_set_vertex_info(vbuf->vf, - vbuf->vinfo, - vbuf->stage.draw->rasterizer->point_size); - - if (!vbuf->vertices) - vbuf_alloc_vertices(vbuf); -} - - -static void -vbuf_first_tri( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - stage->tri = vbuf_tri; - vbuf_set_prim(vbuf, PIPE_PRIM_TRIANGLES); - stage->tri( stage, prim ); -} - - -static void -vbuf_first_line( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - stage->line = vbuf_line; - vbuf_set_prim(vbuf, PIPE_PRIM_LINES); - stage->line( stage, prim ); -} - - -static void -vbuf_first_point( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - stage->point = vbuf_point; - vbuf_set_prim(vbuf, PIPE_PRIM_POINTS); - stage->point( stage, prim ); -} - - -static void -vbuf_flush_indices( struct vbuf_stage *vbuf ) -{ - if(!vbuf->nr_indices) - return; - - assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == - vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); - - switch(vbuf->prim) { - case PIPE_PRIM_POINTS: - break; - case PIPE_PRIM_LINES: - assert(vbuf->nr_indices % 2 == 0); - break; - case PIPE_PRIM_TRIANGLES: - assert(vbuf->nr_indices % 3 == 0); - break; - default: - assert(0); - } - - vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); - - vbuf->nr_indices = 0; - - /* don't need to reset point/line/tri functions */ -#if 0 - stage->point = vbuf_first_point; - stage->line = vbuf_first_line; - stage->tri = vbuf_first_tri; -#endif -} - - -/** - * Flush existing vertex buffer and allocate a new one. - * - * XXX: We separate flush-on-index-full and flush-on-vb-full, but may - * raise issues uploading vertices if the hardware wants to flush when - * we flush. - */ -static void -vbuf_flush_vertices( struct vbuf_stage *vbuf ) -{ - if(vbuf->vertices) { - vbuf_flush_indices(vbuf); - - /* Reset temporary vertices ids */ - if(vbuf->nr_vertices) - draw_reset_vertex_ids( vbuf->stage.draw ); - - /* Free the vertex buffer */ - vbuf->render->release_vertices(vbuf->render, - vbuf->vertices, - vbuf->vertex_size, - vbuf->nr_vertices); - vbuf->max_vertices = vbuf->nr_vertices = 0; - vbuf->vertex_ptr = vbuf->vertices = NULL; - - } -} - - -static void -vbuf_alloc_vertices( struct vbuf_stage *vbuf ) -{ - assert(!vbuf->nr_indices); - assert(!vbuf->vertices); - - /* Allocate a new vertex buffer */ - vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; - vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, - (ushort) vbuf->vertex_size, - (ushort) vbuf->max_vertices); - vbuf->vertex_ptr = vbuf->vertices; -} - - - -static void -vbuf_flush( struct draw_stage *stage, unsigned flags ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - - stage->point = vbuf_first_point; - stage->line = vbuf_first_line; - stage->tri = vbuf_first_tri; - - if (flags & DRAW_FLUSH_BACKEND) - vbuf_flush_vertices( vbuf ); -} - - -static void -vbuf_reset_stipple_counter( struct draw_stage *stage ) -{ - /* XXX: Need to do something here for hardware with linestipple. - */ - (void) stage; -} - - -static void vbuf_destroy( struct draw_stage *stage ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - if(vbuf->indices) - align_free( vbuf->indices ); - - if(vbuf->vf) - draw_vf_destroy( vbuf->vf ); - - if (vbuf->render) - vbuf->render->destroy( vbuf->render ); - - FREE( stage ); -} - - -/** - * Create a new primitive vbuf/render stage. - */ -struct draw_stage *draw_vbuf_stage( struct draw_context *draw, - struct vbuf_render *render ) -{ - struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); - - if(!vbuf) - return NULL; - - vbuf->stage.draw = draw; - vbuf->stage.point = vbuf_first_point; - vbuf->stage.line = vbuf_first_line; - vbuf->stage.tri = vbuf_first_tri; - vbuf->stage.flush = vbuf_flush; - vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter; - vbuf->stage.destroy = vbuf_destroy; - - vbuf->render = render; - - assert(render->max_indices < UNDEFINED_VERTEX_ID); - vbuf->max_indices = render->max_indices; - vbuf->indices = (ushort *) - align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); - if(!vbuf->indices) - vbuf_destroy(&vbuf->stage); - - vbuf->vertices = NULL; - vbuf->vertex_ptr = vbuf->vertices; - - vbuf->prim = ~0; - - if(!GETENV("GALLIUM_NOVF")) - vbuf->vf = draw_vf_create(); - - return &vbuf->stage; -} diff --git a/src/mesa/pipe/draw/draw_vbuf.h b/src/mesa/pipe/draw/draw_vbuf.h deleted file mode 100644 index cfd2b9820c..0000000000 --- a/src/mesa/pipe/draw/draw_vbuf.h +++ /dev/null @@ -1,106 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Vertex buffer drawing stage. - * - * \author Keith Whitwell - * \author José Fonseca - */ - -#ifndef DRAW_VBUF_H_ -#define DRAW_VBUF_H_ - - -#include "pipe/p_util.h" - - -struct draw_context; -struct vertex_info; - - -/** - * Interface for hardware vertex buffer rendering. - */ -struct vbuf_render { - - /** - * Driver limits. May be tuned lower to improve cache hits on - * index list. - */ - unsigned max_indices; - unsigned max_vertex_buffer_bytes; - - /** - * Get the hardware vertex format. - * - * XXX: have this in draw_context instead? - */ - const struct vertex_info *(*get_vertex_info)( struct vbuf_render * ); - - /** - * Request a destination for vertices. - * Hardware renderers will use ttm memory, others will just malloc - * something. - */ - void *(*allocate_vertices)( struct vbuf_render *, - ushort vertex_size, - ushort nr_vertices ); - - /** - * Notify the renderer of the current primitive when it changes. - * Prim is restricted to TRIANGLES, LINES and POINTS. - */ - void (*set_primitive)( struct vbuf_render *, unsigned prim ); - - /** - * DrawElements, note indices are ushort: - */ - void (*draw)( struct vbuf_render *, - const ushort *indices, - uint nr_indices ); - - /** - * Called when vbuf is done with this set of vertices: - */ - void (*release_vertices)( struct vbuf_render *, - void *vertices, - unsigned vertex_size, - unsigned vertices_used ); - - void (*destroy)( struct vbuf_render * ); -}; - - - -struct draw_stage * -draw_vbuf_stage( struct draw_context *draw, - struct vbuf_render *render ); - - -#endif /*DRAW_VBUF_H_*/ diff --git a/src/mesa/pipe/draw/draw_vertex.c b/src/mesa/pipe/draw/draw_vertex.c deleted file mode 100644 index 2d6592150f..0000000000 --- a/src/mesa/pipe/draw/draw_vertex.c +++ /dev/null @@ -1,79 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* - * Functions for specifying the post-transformation vertex layout. - * - * Author: - * Brian Paul - * Keith Whitwell - */ - - -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vertex.h" - - -/** - * Compute the size of a vertex, in dwords/floats, to update the - * vinfo->size field. - */ -void -draw_compute_vertex_size(struct vertex_info *vinfo) -{ - uint i; - - vinfo->size = 0; - for (i = 0; i < vinfo->num_attribs; i++) { - switch (vinfo->emit[i]) { - case EMIT_OMIT: - break; - case EMIT_4UB: - /* fall-through */ - case EMIT_1F_PSIZE: - /* fall-through */ - case EMIT_1F: - vinfo->size += 1; - break; - case EMIT_2F: - vinfo->size += 2; - break; - case EMIT_3F: - vinfo->size += 3; - break; - case EMIT_4F: - vinfo->size += 4; - break; - case EMIT_ALL: - /* fall-through */ - default: - assert(0); - } - } - - assert(vinfo->size * 4 <= MAX_VERTEX_SIZE); -} diff --git a/src/mesa/pipe/draw/draw_vertex.h b/src/mesa/pipe/draw/draw_vertex.h deleted file mode 100644 index 267c74203b..0000000000 --- a/src/mesa/pipe/draw/draw_vertex.h +++ /dev/null @@ -1,111 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Post-transform vertex format info. The vertex_info struct is used by - * the draw_vbuf code to emit hardware-specific vertex layouts into hw - * vertex buffers. - * - * Author: - * Brian Paul - */ - - -#ifndef DRAW_VERTEX_H -#define DRAW_VERTEX_H - - -#include "pipe/p_state.h" - - -/** - * Vertex attribute emit modes - */ -enum attrib_emit { - EMIT_OMIT, /**< don't emit the attribute */ - EMIT_ALL, /**< emit whole post-xform vertex, w/ header */ - EMIT_1F, - EMIT_1F_PSIZE, /**< insert constant point size */ - EMIT_2F, - EMIT_3F, - EMIT_4F, - EMIT_4UB /**< XXX may need variations for RGBA vs BGRA, etc */ -}; - - -/** - * Attribute interpolation mode - */ -enum interp_mode { - INTERP_NONE, /**< never interpolate vertex header info */ - INTERP_POS, /**< special case for frag position */ - INTERP_CONSTANT, - INTERP_LINEAR, - INTERP_PERSPECTIVE -}; - - -/** - * Information about hardware/rasterization vertex layout. - */ -struct vertex_info -{ - uint num_attribs; - uint hwfmt[4]; /**< hardware format info for this format */ - enum interp_mode interp_mode[PIPE_MAX_SHADER_INPUTS]; - enum attrib_emit emit[PIPE_MAX_SHADER_INPUTS]; /**< EMIT_x */ - uint src_index[PIPE_MAX_SHADER_INPUTS]; /**< map to post-xform attribs */ - uint size; /**< total vertex size in dwords */ -}; - - - -/** - * Add another attribute to the given vertex_info object. - * \param src_index indicates which post-transformed vertex attrib slot - * corresponds to this attribute. - * \return slot in which the attribute was added - */ -static INLINE uint -draw_emit_vertex_attr(struct vertex_info *vinfo, - enum attrib_emit emit, enum interp_mode interp, - uint src_index) -{ - const uint n = vinfo->num_attribs; - assert(n < PIPE_MAX_SHADER_INPUTS); - vinfo->emit[n] = emit; - vinfo->interp_mode[n] = interp; - vinfo->src_index[n] = src_index; - vinfo->num_attribs++; - return n; -} - - -extern void draw_compute_vertex_size(struct vertex_info *vinfo); - - -#endif /* DRAW_VERTEX_H */ diff --git a/src/mesa/pipe/draw/draw_vertex_cache.c b/src/mesa/pipe/draw/draw_vertex_cache.c deleted file mode 100644 index 44427999cc..0000000000 --- a/src/mesa/pipe/draw/draw_vertex_cache.c +++ /dev/null @@ -1,196 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "draw_private.h" -#include "draw_context.h" - - -void draw_vertex_cache_invalidate( struct draw_context *draw ) -{ - assert(draw->pq.queue_nr == 0); - assert(draw->vs.queue_nr == 0); - assert(draw->vcache.referenced == 0); - - memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); -} - - -/** - * Check if vertex is in cache, otherwise add it. It won't go through - * VS yet, not until there is a flush operation or the VS queue fills up. - * - * Note that cache entries are basically just two pointers: the first - * an index into the user's vertex arrays, the second a location in - * the vertex shader cache for the post-transformed vertex. - * - * \return pointer to location of (post-transformed) vertex header in the cache - */ -static struct vertex_header *get_vertex( struct draw_context *draw, - unsigned i ) -{ - unsigned slot = (i + (i>>5)) % VCACHE_SIZE; - - assert(slot < 32); /* so we don't exceed the bitfield size below */ - - /* Cache miss? - */ - if (draw->vcache.idx[slot] != i) { - - /* If slot is in use, use the overflow area: - */ - if (draw->vcache.referenced & (1 << slot)) { - slot = VCACHE_SIZE + draw->vcache.overflow++; - } - - assert(slot < Elements(draw->vcache.idx)); - - draw->vcache.idx[slot] = i; - - /* Add to vertex shader queue: - */ - assert(draw->vs.queue_nr < VS_QUEUE_LENGTH); - draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; - draw->vs.queue[draw->vs.queue_nr].elt = i; - draw->vs.queue_nr++; - - /* Need to set the vertex's edge flag here. If we're being called - * by do_ef_triangle(), that function needs edge flag info! - */ - draw->vcache.vertex[slot]->clipmask = 0; - draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ - draw->vcache.vertex[slot]->pad = 0; - draw->vcache.vertex[slot]->vertex_id = UNDEFINED_VERTEX_ID; - } - - - /* primitive flushing may have cleared the bitfield but did not - * clear the idx[] array values. Set the bit now. This fixes a - * bug found when drawing long triangle fans. - */ - draw->vcache.referenced |= (1 << slot); - return draw->vcache.vertex[slot]; -} - - -static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const unsigned *elts = (const unsigned *) draw->user.elts; - return get_vertex( draw, elts[i] ); -} - - -static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const ushort *elts = (const ushort *) draw->user.elts; - return get_vertex( draw, elts[i] ); -} - - -static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const ubyte *elts = (const ubyte *) draw->user.elts; - return get_vertex( draw, elts[i] ); -} - - -void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) -{ - unsigned i; - - for (i = 0; i < Elements(draw->vcache.vertex); i++) - draw->vcache.vertex[i]->vertex_id = UNDEFINED_VERTEX_ID; -} - - -void draw_vertex_cache_unreference( struct draw_context *draw ) -{ - draw->vcache.referenced = 0; - draw->vcache.overflow = 0; -} - - -int draw_vertex_cache_check_space( struct draw_context *draw, - unsigned nr_verts ) -{ - if (draw->vcache.overflow + nr_verts < VCACHE_OVERFLOW) { - /* The vs queue is sized so that this can never happen: - */ - assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); - return TRUE; - } - else - return FALSE; -} - - - -/** - * Tell the drawing context about the index/element buffer to use - * (ala glDrawElements) - * If no element buffer is to be used (i.e. glDrawArrays) then this - * should be called with eltSize=0 and elements=NULL. - * - * \param draw the drawing context - * \param eltSize size of each element (1, 2 or 4 bytes) - * \param elements the element buffer ptr - */ -void -draw_set_mapped_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ) -{ -// draw_statechange( draw ); - - /* choose the get_vertex() function to use */ - switch (eltSize) { - case 0: - draw->vcache.get_vertex = get_vertex; - break; - case 1: - draw->vcache.get_vertex = get_ubyte_elt_vertex; - break; - case 2: - draw->vcache.get_vertex = get_ushort_elt_vertex; - break; - case 4: - draw->vcache.get_vertex = get_uint_elt_vertex; - break; - default: - assert(0); - } - draw->user.elts = elements; - draw->user.eltSize = eltSize; -} - diff --git a/src/mesa/pipe/draw/draw_vertex_fetch.c b/src/mesa/pipe/draw/draw_vertex_fetch.c deleted file mode 100644 index e13df04605..0000000000 --- a/src/mesa/pipe/draw/draw_vertex_fetch.c +++ /dev/null @@ -1,510 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" -#include "draw_context.h" - - -#define DRAW_DBG 0 - - -/** - * Fetch a float[4] vertex attribute from memory, doing format/type - * conversion as needed. - * - * This is probably needed/dupliocated elsewhere, eg format - * conversion, texture sampling etc. - */ -#define FETCH_ATTRIB( NAME, SZ, CVT ) \ -static void \ -fetch_##NAME(const void *ptr, float *attrib) \ -{ \ - static const float defaults[4] = { 0,0,0,1 }; \ - int i; \ - \ - for (i = 0; i < SZ; i++) { \ - attrib[i] = CVT; \ - } \ - \ - for (; i < 4; i++) { \ - attrib[i] = defaults[i]; \ - } \ -} - -#define CVT_64_FLOAT (float) ((double *) ptr)[i] -#define CVT_32_FLOAT ((float *) ptr)[i] - -#define CVT_8_USCALED (float) ((unsigned char *) ptr)[i] -#define CVT_16_USCALED (float) ((unsigned short *) ptr)[i] -#define CVT_32_USCALED (float) ((unsigned int *) ptr)[i] - -#define CVT_8_SSCALED (float) ((char *) ptr)[i] -#define CVT_16_SSCALED (float) ((short *) ptr)[i] -#define CVT_32_SSCALED (float) ((int *) ptr)[i] - -#define CVT_8_UNORM (float) ((unsigned char *) ptr)[i] / 255.0f -#define CVT_16_UNORM (float) ((unsigned short *) ptr)[i] / 65535.0f -#define CVT_32_UNORM (float) ((unsigned int *) ptr)[i] / 4294967295.0f - -#define CVT_8_SNORM (float) ((char *) ptr)[i] / 127.0f -#define CVT_16_SNORM (float) ((short *) ptr)[i] / 32767.0f -#define CVT_32_SNORM (float) ((int *) ptr)[i] / 2147483647.0f - -FETCH_ATTRIB( R64G64B64A64_FLOAT, 4, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) -FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) - -FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) -FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) - -FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) -FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) - -FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) -FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) - -FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) -FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) - -FETCH_ATTRIB( R32G32B32A32_SNORM, 4, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32B32_SNORM, 3, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32_SNORM, 2, CVT_32_SNORM ) -FETCH_ATTRIB( R32_SNORM, 1, CVT_32_SNORM ) - -FETCH_ATTRIB( R16G16B16A16_USCALED, 4, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16B16_USCALED, 3, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16_USCALED, 2, CVT_16_USCALED ) -FETCH_ATTRIB( R16_USCALED, 1, CVT_16_USCALED ) - -FETCH_ATTRIB( R16G16B16A16_SSCALED, 4, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16B16_SSCALED, 3, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16_SSCALED, 2, CVT_16_SSCALED ) -FETCH_ATTRIB( R16_SSCALED, 1, CVT_16_SSCALED ) - -FETCH_ATTRIB( R16G16B16A16_UNORM, 4, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16B16_UNORM, 3, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16_UNORM, 2, CVT_16_UNORM ) -FETCH_ATTRIB( R16_UNORM, 1, CVT_16_UNORM ) - -FETCH_ATTRIB( R16G16B16A16_SNORM, 4, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16B16_SNORM, 3, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16_SNORM, 2, CVT_16_SNORM ) -FETCH_ATTRIB( R16_SNORM, 1, CVT_16_SNORM ) - -FETCH_ATTRIB( R8G8B8A8_USCALED, 4, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8B8_USCALED, 3, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8_USCALED, 2, CVT_8_USCALED ) -FETCH_ATTRIB( R8_USCALED, 1, CVT_8_USCALED ) - -FETCH_ATTRIB( R8G8B8A8_SSCALED, 4, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8B8_SSCALED, 3, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8_SSCALED, 2, CVT_8_SSCALED ) -FETCH_ATTRIB( R8_SSCALED, 1, CVT_8_SSCALED ) - -FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8B8_UNORM, 3, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8_UNORM, 2, CVT_8_UNORM ) -FETCH_ATTRIB( R8_UNORM, 1, CVT_8_UNORM ) - -FETCH_ATTRIB( R8G8B8A8_SNORM, 4, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8B8_SNORM, 3, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8_SNORM, 2, CVT_8_SNORM ) -FETCH_ATTRIB( R8_SNORM, 1, CVT_8_SNORM ) - -FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) -//FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) - - - -static fetch_func get_fetch_func( enum pipe_format format ) -{ -#if 0 - { - char tmp[80]; - pf_sprint_name(tmp, format); - debug_printf("%s: %s\n", __FUNCTION__, tmp); - } -#endif - - switch (format) { - case PIPE_FORMAT_R64_FLOAT: - return fetch_R64_FLOAT; - case PIPE_FORMAT_R64G64_FLOAT: - return fetch_R64G64_FLOAT; - case PIPE_FORMAT_R64G64B64_FLOAT: - return fetch_R64G64B64_FLOAT; - case PIPE_FORMAT_R64G64B64A64_FLOAT: - return fetch_R64G64B64A64_FLOAT; - - case PIPE_FORMAT_R32_FLOAT: - return fetch_R32_FLOAT; - case PIPE_FORMAT_R32G32_FLOAT: - return fetch_R32G32_FLOAT; - case PIPE_FORMAT_R32G32B32_FLOAT: - return fetch_R32G32B32_FLOAT; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return fetch_R32G32B32A32_FLOAT; - - case PIPE_FORMAT_R32_UNORM: - return fetch_R32_UNORM; - case PIPE_FORMAT_R32G32_UNORM: - return fetch_R32G32_UNORM; - case PIPE_FORMAT_R32G32B32_UNORM: - return fetch_R32G32B32_UNORM; - case PIPE_FORMAT_R32G32B32A32_UNORM: - return fetch_R32G32B32A32_UNORM; - - case PIPE_FORMAT_R32_USCALED: - return fetch_R32_USCALED; - case PIPE_FORMAT_R32G32_USCALED: - return fetch_R32G32_USCALED; - case PIPE_FORMAT_R32G32B32_USCALED: - return fetch_R32G32B32_USCALED; - case PIPE_FORMAT_R32G32B32A32_USCALED: - return fetch_R32G32B32A32_USCALED; - - case PIPE_FORMAT_R32_SNORM: - return fetch_R32_SNORM; - case PIPE_FORMAT_R32G32_SNORM: - return fetch_R32G32_SNORM; - case PIPE_FORMAT_R32G32B32_SNORM: - return fetch_R32G32B32_SNORM; - case PIPE_FORMAT_R32G32B32A32_SNORM: - return fetch_R32G32B32A32_SNORM; - - case PIPE_FORMAT_R32_SSCALED: - return fetch_R32_SSCALED; - case PIPE_FORMAT_R32G32_SSCALED: - return fetch_R32G32_SSCALED; - case PIPE_FORMAT_R32G32B32_SSCALED: - return fetch_R32G32B32_SSCALED; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - return fetch_R32G32B32A32_SSCALED; - - case PIPE_FORMAT_R16_UNORM: - return fetch_R16_UNORM; - case PIPE_FORMAT_R16G16_UNORM: - return fetch_R16G16_UNORM; - case PIPE_FORMAT_R16G16B16_UNORM: - return fetch_R16G16B16_UNORM; - case PIPE_FORMAT_R16G16B16A16_UNORM: - return fetch_R16G16B16A16_UNORM; - - case PIPE_FORMAT_R16_USCALED: - return fetch_R16_USCALED; - case PIPE_FORMAT_R16G16_USCALED: - return fetch_R16G16_USCALED; - case PIPE_FORMAT_R16G16B16_USCALED: - return fetch_R16G16B16_USCALED; - case PIPE_FORMAT_R16G16B16A16_USCALED: - return fetch_R16G16B16A16_USCALED; - - case PIPE_FORMAT_R16_SNORM: - return fetch_R16_SNORM; - case PIPE_FORMAT_R16G16_SNORM: - return fetch_R16G16_SNORM; - case PIPE_FORMAT_R16G16B16_SNORM: - return fetch_R16G16B16_SNORM; - case PIPE_FORMAT_R16G16B16A16_SNORM: - return fetch_R16G16B16A16_SNORM; - - case PIPE_FORMAT_R16_SSCALED: - return fetch_R16_SSCALED; - case PIPE_FORMAT_R16G16_SSCALED: - return fetch_R16G16_SSCALED; - case PIPE_FORMAT_R16G16B16_SSCALED: - return fetch_R16G16B16_SSCALED; - case PIPE_FORMAT_R16G16B16A16_SSCALED: - return fetch_R16G16B16A16_SSCALED; - - case PIPE_FORMAT_R8_UNORM: - return fetch_R8_UNORM; - case PIPE_FORMAT_R8G8_UNORM: - return fetch_R8G8_UNORM; - case PIPE_FORMAT_R8G8B8_UNORM: - return fetch_R8G8B8_UNORM; - case PIPE_FORMAT_R8G8B8A8_UNORM: - return fetch_R8G8B8A8_UNORM; - - case PIPE_FORMAT_R8_USCALED: - return fetch_R8_USCALED; - case PIPE_FORMAT_R8G8_USCALED: - return fetch_R8G8_USCALED; - case PIPE_FORMAT_R8G8B8_USCALED: - return fetch_R8G8B8_USCALED; - case PIPE_FORMAT_R8G8B8A8_USCALED: - return fetch_R8G8B8A8_USCALED; - - case PIPE_FORMAT_R8_SNORM: - return fetch_R8_SNORM; - case PIPE_FORMAT_R8G8_SNORM: - return fetch_R8G8_SNORM; - case PIPE_FORMAT_R8G8B8_SNORM: - return fetch_R8G8B8_SNORM; - case PIPE_FORMAT_R8G8B8A8_SNORM: - return fetch_R8G8B8A8_SNORM; - - case PIPE_FORMAT_R8_SSCALED: - return fetch_R8_SSCALED; - case PIPE_FORMAT_R8G8_SSCALED: - return fetch_R8G8_SSCALED; - case PIPE_FORMAT_R8G8B8_SSCALED: - return fetch_R8G8B8_SSCALED; - case PIPE_FORMAT_R8G8B8A8_SSCALED: - return fetch_R8G8B8A8_SSCALED; - - case PIPE_FORMAT_A8R8G8B8_UNORM: - return fetch_A8R8G8B8_UNORM; - - case 0: - return NULL; /* not sure why this is needed */ - - default: - assert(0); - return NULL; - } -} - - -static void -transpose_4x4( float *out, const float *in ) -{ - /* This can be achieved in 12 sse instructions, plus the final - * stores I guess. This is probably a bit more than that - maybe - * 32 or so? - */ - out[0] = in[0]; out[1] = in[4]; out[2] = in[8]; out[3] = in[12]; - out[4] = in[1]; out[5] = in[5]; out[6] = in[9]; out[7] = in[13]; - out[8] = in[2]; out[9] = in[6]; out[10] = in[10]; out[11] = in[14]; - out[12] = in[3]; out[13] = in[7]; out[14] = in[11]; out[15] = in[15]; -} - - - -static void fetch_xyz_rgb( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - - assert(count <= 4); - -// debug_printf("%s\n", __FUNCTION__); - - /* loop over vertex attributes (vertex shader inputs) - */ - - for (i = 0; i < 4; i++) { - { - const float *in = (const float *)(src[0] + elts[i] * pitch[0]); - float *out = &machine->Inputs[0].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - - { - const float *in = (const float *)(src[1] + elts[i] * pitch[1]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - } -} - - - - -static void fetch_xyz_rgb_st( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - - assert(count <= 4); - - /* loop over vertex attributes (vertex shader inputs) - */ - - for (i = 0; i < 4; i++) { - { - const float *in = (const float *)(src[0] + elts[i] * pitch[0]); - float *out = &machine->Inputs[0].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - - { - const float *in = (const float *)(src[1] + elts[i] * pitch[1]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - - { - const float *in = (const float *)(src[2] + elts[i] * pitch[2]); - float *out = &machine->Inputs[2].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = 0.0f; - out[12] = 1.0f; - } - } -} - - - - -/** - * Fetch vertex attributes for 'count' vertices. - */ -static void generic_vertex_fetch( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - unsigned nr_attrs = draw->vertex_fetch.nr_attrs; - unsigned attr; - - assert(count <= 4); - -// debug_printf("%s %d\n", __FUNCTION__, count); - - /* loop over vertex attributes (vertex shader inputs) - */ - for (attr = 0; attr < nr_attrs; attr++) { - - const unsigned pitch = draw->vertex_fetch.pitch[attr]; - const ubyte *src = draw->vertex_fetch.src_ptr[attr]; - const fetch_func fetch = draw->vertex_fetch.fetch[attr]; - unsigned i; - float p[4][4]; - - - /* Fetch four attributes for four vertices. - * - * Could fetch directly into AOS format, but this is meant to be - * a prototype for an sse implementation, which would have - * difficulties doing that. - */ - for (i = 0; i < count; i++) - fetch( src + elts[i] * pitch, p[i] ); - - /* Be nice and zero out any missing vertices: - */ - for ( ; i < 4; i++) - p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; - - /* Transpose/swizzle into sse-friendly format. Currently - * assuming that all vertex shader inputs are float[4], but this - * isn't true -- if the vertex shader only wants tex0.xy, we - * could optimize for that. - * - * To do so fully without codegen would probably require an - * excessive number of fetch functions, but we could at least - * minimize the transpose step: - */ - transpose_4x4( (float *)&machine->Inputs[attr].xyzw[0].f[0], (float *)p ); - } -} - - - -void draw_update_vertex_fetch( struct draw_context *draw ) -{ - unsigned nr_attrs, i; - -// debug_printf("%s\n", __FUNCTION__); - - /* this may happend during context init */ - if (!draw->vertex_shader) - return; - - nr_attrs = draw->vertex_shader->state->num_inputs; - - for (i = 0; i < nr_attrs; i++) { - unsigned buf = draw->vertex_element[i].vertex_buffer_index; - enum pipe_format format = draw->vertex_element[i].src_format; - - draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + - draw->vertex_buffer[buf].buffer_offset + - draw->vertex_element[i].src_offset; - - draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; - draw->vertex_fetch.fetch[i] = get_fetch_func( format ); - } - - draw->vertex_fetch.nr_attrs = nr_attrs; - - draw->vertex_fetch.fetch_func = generic_vertex_fetch; - - switch (nr_attrs) { - case 2: - if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT) - draw->vertex_fetch.fetch_func = fetch_xyz_rgb; - break; - case 3: - if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[2].src_format == PIPE_FORMAT_R32G32_FLOAT) - draw->vertex_fetch.fetch_func = fetch_xyz_rgb_st; - break; - default: - break; - } - -} diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c deleted file mode 100644 index c824c1407e..0000000000 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ /dev/null @@ -1,325 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#if defined(__i386__) || defined(__386__) -#include "pipe/tgsi/exec/tgsi_sse2.h" -#endif -#include "draw_private.h" -#include "draw_context.h" - -#include "x86/rtasm/x86sse.h" -#include "pipe/llvm/gallivm.h" - - -#define DBG_VS 0 - - -static INLINE unsigned -compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) -{ - unsigned mask = 0; - unsigned i; - - /* Do the hardwired planes first: - */ - if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; - if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; - if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; - if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; - if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; - if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; - - /* Followed by any remaining ones: - */ - for (i = 6; i < nr; i++) { - if (dot4(clip, plane[i]) < 0) - mask |= (1<machine; - unsigned int j; - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - - assert(count <= 4); - assert(draw->vertex_shader->state->output_semantic_name[0] - == TGSI_SEMANTIC_POSITION); - - /* Consts does not require 16 byte alignment. */ - machine->Consts = (float (*)[4]) draw->user.constants; - - machine->Inputs = ALIGN16_ASSIGN(inputs); - machine->Outputs = ALIGN16_ASSIGN(outputs); - - draw->vertex_fetch.fetch_func( draw, machine, elts, count ); - - /* run shader */ -#ifdef MESA_LLVM - if (1) { - struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; - gallivm_cpu_vs_exec(prog, - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps); - } else -#elif defined(__i386__) || defined(__386__) - if (draw->use_sse) { - /* SSE */ - /* cast away const */ - struct draw_vertex_shader *shader - = (struct draw_vertex_shader *)draw->vertex_shader; - codegen_function func - = (codegen_function) x86_get_func( &shader->sse2_program ); - - if (func) - func( - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps ); - else - /* interpreter */ - tgsi_exec_machine_run( machine ); - } - else -#endif - { - /* interpreter */ - tgsi_exec_machine_run( machine ); - } - - /* store machine results */ - for (j = 0; j < count; j++) { - unsigned slot; - float x, y, z, w; - - /* Handle attr[0] (position) specially: - * - * XXX: Computing the clipmask should be done in the vertex - * program as a set of DP4 instructions appended to the - * user-provided code. - */ - x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; - -#if DBG_VS - debug_printf("output[%d]win: %f %f %f %f\n", j, - vOut[j]->data[0][0], - vOut[j]->data[0][1], - vOut[j]->data[0][2], - vOut[j]->data[0][3]); -#endif - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; -#if DBG_VS - debug_printf("output[%d][%d]: %f %f %f %f\n", j, slot, - vOut[j]->data[slot][0], - vOut[j]->data[slot][1], - vOut[j]->data[slot][2], - vOut[j]->data[slot][3]); -#endif - } - } /* loop over vertices */ -} - - -/** - * Run the vertex shader on all vertices in the vertex queue. - * Called by the draw module when the vertx cache needs to be flushed. - */ -void -draw_vertex_shader_queue_flush(struct draw_context *draw) -{ - unsigned i; - - assert(draw->vs.queue_nr != 0); - - /* XXX: do this on statechange: - */ - draw_update_vertex_fetch( draw ); - -// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); - - /* run vertex shader on vertex cache entries, four per invokation */ - for (i = 0; i < draw->vs.queue_nr; i += 4) { - struct vertex_header *dests[4]; - unsigned elts[4]; - int j, n = MIN2(4, draw->vs.queue_nr - i); - - for (j = 0; j < n; j++) { - elts[j] = draw->vs.queue[i + j].elt; - dests[j] = draw->vs.queue[i + j].dest; - } - - for ( ; j < 4; j++) { - elts[j] = elts[0]; - dests[j] = dests[0]; - } - - assert(n > 0); - assert(n <= 4); - - run_vertex_program(draw, elts, n, dests); - } - - draw->vs.queue_nr = 0; -} - - -struct draw_vertex_shader * -draw_create_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader) -{ - struct draw_vertex_shader *vs; - - vs = CALLOC_STRUCT( draw_vertex_shader ); - if (vs == NULL) { - return NULL; - } - - vs->state = shader; - -#ifdef MESA_LLVM - struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); - gallivm_ir_set_layout(ir, GALLIVM_SOA); - gallivm_ir_set_components(ir, 4); - gallivm_ir_fill_from_tgsi(ir, shader->tokens); - vs->llvm_prog = gallivm_ir_compile(ir); - gallivm_ir_delete(ir); - - draw->engine = gallivm_global_cpu_engine(); - if (!draw->engine) { - draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); - } - else { - gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); - } -#elif defined(__i386__) || defined(__386__) - if (draw->use_sse) { - /* cast-away const */ - struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; - - x86_init_func( &vs->sse2_program ); - if (!tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, - &vs->sse2_program )) { - x86_release_func( (struct x86_function *) &vs->sse2_program ); - fprintf(stdout /*err*/, - "tgsi_emit_sse2() failed, falling back to interpreter\n"); - } - } -#endif - - return vs; -} - - -void -draw_bind_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - - draw->vertex_shader = dvs; - draw->num_vs_outputs = dvs->state->num_outputs; - - /* specify the vertex program to interpret/execute */ - tgsi_exec_machine_init(&draw->machine, - draw->vertex_shader->state->tokens, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/ ); -} - - -void -draw_delete_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs) -{ -#if defined(__i386__) || defined(__386__) - x86_release_func( (struct x86_function *) &dvs->sse2_program ); -#endif - - FREE( dvs ); -} diff --git a/src/mesa/pipe/draw/draw_vf.c b/src/mesa/pipe/draw/draw_vf.c deleted file mode 100644 index dc3a5ecd21..0000000000 --- a/src/mesa/pipe/draw/draw_vf.c +++ /dev/null @@ -1,428 +0,0 @@ -/* - * Copyright 2003 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: - * Keith Whitwell - */ - - -#include - -#include "pipe/p_compiler.h" -#include "pipe/p_util.h" - -#include "draw_vf.h" - - -#define DRAW_VF_DBG 0 - - -/* TODO: remove this */ -extern void -_mesa_exec_free( void *addr ); - - -static boolean match_fastpath( struct draw_vertex_fetch *vf, - const struct draw_vf_fastpath *fp) -{ - unsigned j; - - if (vf->attr_count != fp->attr_count) - return FALSE; - - for (j = 0; j < vf->attr_count; j++) - if (vf->attr[j].format != fp->attr[j].format || - vf->attr[j].inputsize != fp->attr[j].size || - vf->attr[j].vertoffset != fp->attr[j].offset) - return FALSE; - - if (fp->match_strides) { - if (vf->vertex_stride != fp->vertex_stride) - return FALSE; - - for (j = 0; j < vf->attr_count; j++) - if (vf->attr[j].inputstride != fp->attr[j].stride) - return FALSE; - } - - return TRUE; -} - -static boolean search_fastpath_emit( struct draw_vertex_fetch *vf ) -{ - struct draw_vf_fastpath *fp = vf->fastpath; - - for ( ; fp ; fp = fp->next) { - if (match_fastpath(vf, fp)) { - vf->emit = fp->func; - return TRUE; - } - } - - return FALSE; -} - -void draw_vf_register_fastpath( struct draw_vertex_fetch *vf, - boolean match_strides ) -{ - struct draw_vf_fastpath *fastpath = CALLOC_STRUCT(draw_vf_fastpath); - unsigned i; - - fastpath->vertex_stride = vf->vertex_stride; - fastpath->attr_count = vf->attr_count; - fastpath->match_strides = match_strides; - fastpath->func = vf->emit; - fastpath->attr = (struct draw_vf_attr_type *) - MALLOC(vf->attr_count * sizeof(fastpath->attr[0])); - - for (i = 0; i < vf->attr_count; i++) { - fastpath->attr[i].format = vf->attr[i].format; - fastpath->attr[i].stride = vf->attr[i].inputstride; - fastpath->attr[i].size = vf->attr[i].inputsize; - fastpath->attr[i].offset = vf->attr[i].vertoffset; - } - - fastpath->next = vf->fastpath; - vf->fastpath = fastpath; -} - - - - -/*********************************************************************** - * Build codegen functions or return generic ones: - */ -static void choose_emit_func( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *dest) -{ - vf->emit = NULL; - - /* Does this match an existing (hardwired, codegen or known-bad) - * fastpath? - */ - if (search_fastpath_emit(vf)) { - /* Use this result. If it is null, then it is already known - * that the current state will fail for codegen and there is no - * point trying again. - */ - } - else if (vf->codegen_emit) { - vf->codegen_emit( vf ); - } - - if (!vf->emit) { - draw_vf_generate_hardwired_emit(vf); - } - - /* Otherwise use the generic version: - */ - if (!vf->emit) - vf->emit = draw_vf_generic_emit; - - vf->emit( vf, count, dest ); -} - - - - - -/*********************************************************************** - * Public entrypoints, mostly dispatch to the above: - */ - - - -static unsigned -draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, - const struct draw_vf_attr_map *map, - unsigned nr, - unsigned vertex_stride ) -{ - unsigned offset = 0; - unsigned i, j; - - assert(nr < PIPE_ATTRIB_MAX); - - for (j = 0, i = 0; i < nr; i++) { - const unsigned format = map[i].format; - if (format == DRAW_EMIT_PAD) { -#if (DRAW_VF_DBG) - debug_printf("%d: pad %d, offset %d\n", i, - map[i].offset, offset); -#endif - - offset += map[i].offset; - - } - else { - vf->attr[j].attrib = map[i].attrib; - vf->attr[j].format = format; - vf->attr[j].insert = draw_vf_format_info[format].insert; - vf->attr[j].vertattrsize = draw_vf_format_info[format].attrsize; - vf->attr[j].vertoffset = offset; - vf->attr[j].isconst = draw_vf_format_info[format].isconst; - if(vf->attr[j].isconst) - memcpy(vf->attr[j].data, &map[i].data, vf->attr[j].vertattrsize); - -#if (DRAW_VF_DBG) - debug_printf("%d: %s, offset %d\n", i, - draw_vf_format_info[format].name, - vf->attr[j].vertoffset); -#endif - - offset += draw_vf_format_info[format].attrsize; - j++; - } - } - - vf->attr_count = j; - vf->vertex_stride = vertex_stride ? vertex_stride : offset; - vf->emit = choose_emit_func; - - assert(vf->vertex_stride >= offset); - return vf->vertex_stride; -} - - -void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, - const struct vertex_info *vinfo, - float point_size ) -{ - unsigned i, j, k; - struct draw_vf_attr *a = vf->attr; - struct draw_vf_attr_map attrs[PIPE_MAX_SHADER_INPUTS]; - unsigned count = 0; /* for debug/sanity */ - unsigned nr_attrs = 0; - - for (i = 0; i < vinfo->num_attribs; i++) { - j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; - case EMIT_ALL: { - /* just copy the whole vertex as-is to the vbuf */ - unsigned s = vinfo->size; - assert(i == 0); - assert(j == 0); - /* copy the vertex header */ - /* XXX: we actually don't copy the header, just pad it */ - attrs[nr_attrs].attrib = 0; - attrs[nr_attrs].format = DRAW_EMIT_PAD; - attrs[nr_attrs].offset = offsetof(struct vertex_header, data); - s -= offsetof(struct vertex_header, data)/4; - count += offsetof(struct vertex_header, data)/4; - nr_attrs++; - /* copy the vertex data */ - for(k = 0; k < (s & ~0x3); k += 4) { - attrs[nr_attrs].attrib = k/4; - attrs[nr_attrs].format = DRAW_EMIT_4F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 4; - } - /* tail */ - /* XXX: actually, this shouldn't be needed */ - attrs[nr_attrs].attrib = k/4; - attrs[nr_attrs].offset = 0; - switch(s & 0x3) { - case 0: - break; - case 1: - attrs[nr_attrs].format = DRAW_EMIT_1F; - nr_attrs++; - count += 1; - break; - case 2: - attrs[nr_attrs].format = DRAW_EMIT_2F; - nr_attrs++; - count += 2; - break; - case 3: - attrs[nr_attrs].format = DRAW_EMIT_3F; - nr_attrs++; - count += 3; - break; - } - break; - } - case EMIT_1F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_1F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count++; - break; - case EMIT_1F_PSIZE: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_1F_CONST; - attrs[nr_attrs].offset = 0; - attrs[nr_attrs].data.f[0] = point_size; - nr_attrs++; - count++; - break; - case EMIT_2F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_2F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 2; - break; - case EMIT_3F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_3F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 3; - break; - case EMIT_4F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_4F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 4; - break; - case EMIT_4UB: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_4UB_4F_BGRA; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 1; - break; - default: - assert(0); - } - } - - assert(count == vinfo->size); - - draw_vf_set_vertex_attributes(vf, - attrs, - nr_attrs, - vinfo->size * sizeof(float) ); - - for (j = 0; j < vf->attr_count; j++) { - a[j].inputsize = 4; - a[j].do_insert = a[j].insert[4 - 1]; - if(a[j].isconst) { - a[j].inputptr = a[j].data; - a[j].inputstride = 0; - } - } -} - - -#if 0 -/* Set attribute pointers, adjusted for start position: - */ -void draw_vf_set_sources( struct draw_vertex_fetch *vf, - GLvector4f * const sources[], - unsigned start ) -{ - struct draw_vf_attr *a = vf->attr; - unsigned j; - - for (j = 0; j < vf->attr_count; j++) { - const GLvector4f *vptr = sources[a[j].attrib]; - - if ((a[j].inputstride != vptr->stride) || - (a[j].inputsize != vptr->size)) - vf->emit = choose_emit_func; - - a[j].inputstride = vptr->stride; - a[j].inputsize = vptr->size; - a[j].do_insert = a[j].insert[vptr->size - 1]; - a[j].inputptr = ((uint8_t *)vptr->data) + start * vptr->stride; - } -} -#endif - - -/** - * Emit a vertex to dest. - */ -void draw_vf_emit_vertex( struct draw_vertex_fetch *vf, - struct vertex_header *vertex, - void *dest ) -{ - struct draw_vf_attr *a = vf->attr; - unsigned j; - - for (j = 0; j < vf->attr_count; j++) { - if (!a[j].isconst) { - a[j].inputptr = (uint8_t *)&vertex->data[a[j].attrib][0]; - a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ - } - } - - vf->emit( vf, 1, (uint8_t*) dest ); -} - - - -struct draw_vertex_fetch *draw_vf_create( void ) -{ - struct draw_vertex_fetch *vf = CALLOC_STRUCT(draw_vertex_fetch); - unsigned i; - - for (i = 0; i < PIPE_ATTRIB_MAX; i++) - vf->attr[i].vf = vf; - - vf->identity[0] = 0.0; - vf->identity[1] = 0.0; - vf->identity[2] = 0.0; - vf->identity[3] = 1.0; - - vf->codegen_emit = NULL; - -#ifdef USE_SSE_ASM - if (!GETENV("GALLIUM_NO_CODEGEN")) - vf->codegen_emit = draw_vf_generate_sse_emit; -#endif - - return vf; -} - - -void draw_vf_destroy( struct draw_vertex_fetch *vf ) -{ - struct draw_vf_fastpath *fp, *tmp; - - for (fp = vf->fastpath ; fp ; fp = tmp) { - tmp = fp->next; - FREE(fp->attr); - - /* KW: At the moment, fp->func is constrained to be allocated by - * _mesa_exec_alloc(), as the hardwired fastpaths in - * t_vertex_generic.c are handled specially. It would be nice - * to unify them, but this probably won't change until this - * module gets another overhaul. - */ - //_mesa_exec_free((void *) fp->func); - FREE(fp); - } - - vf->fastpath = NULL; - FREE(vf); -} diff --git a/src/mesa/pipe/draw/draw_vf.h b/src/mesa/pipe/draw/draw_vf.h deleted file mode 100644 index 011c8f0ff1..0000000000 --- a/src/mesa/pipe/draw/draw_vf.h +++ /dev/null @@ -1,236 +0,0 @@ -/* - * Copyright 2008 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/** - * Vertex fetch/store/convert code. This functionality is used in two places: - * 1. Vertex fetch/convert - to grab vertex data from incoming vertex - * arrays and convert to format needed by vertex shaders. - * 2. Vertex store/emit - to convert simple float[][4] vertex attributes - * (which is the organization used throughout the draw/prim pipeline) to - * hardware-specific formats and emit into hardware vertex buffers. - * - * - * Authors: - * Keith Whitwell - */ - -#ifndef DRAW_VF_H -#define DRAW_VF_H - - -#include "pipe/p_compiler.h" -#include "pipe/p_state.h" - -#include "draw_vertex.h" -#include "draw_private.h" /* for vertex_header */ - - -enum draw_vf_attr_format { - DRAW_EMIT_1F, - DRAW_EMIT_2F, - DRAW_EMIT_3F, - DRAW_EMIT_4F, - DRAW_EMIT_3F_XYW, /**< for projective texture */ - DRAW_EMIT_1UB_1F, /**< for fog coordinate */ - DRAW_EMIT_3UB_3F_RGB, /**< for specular color */ - DRAW_EMIT_3UB_3F_BGR, /**< for specular color */ - DRAW_EMIT_4UB_4F_RGBA, /**< for color */ - DRAW_EMIT_4UB_4F_BGRA, /**< for color */ - DRAW_EMIT_4UB_4F_ARGB, /**< for color */ - DRAW_EMIT_4UB_4F_ABGR, /**< for color */ - DRAW_EMIT_1F_CONST, - DRAW_EMIT_2F_CONST, - DRAW_EMIT_3F_CONST, - DRAW_EMIT_4F_CONST, - DRAW_EMIT_PAD, /**< leave a hole of 'offset' bytes */ - DRAW_EMIT_MAX -}; - -struct draw_vf_attr_map -{ - /** Input attribute number */ - unsigned attrib; - - enum draw_vf_attr_format format; - - unsigned offset; - - /** - * Constant data for DRAW_EMIT_*_CONST - */ - union { - uint8_t ub[4]; - float f[4]; - } data; -}; - -struct draw_vertex_fetch; - - - -#if 0 -unsigned -draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, - const struct draw_vf_attr_map *map, - unsigned nr, - unsigned vertex_stride ); -#endif - -void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, - const struct vertex_info *vinfo, - float point_size ); - -#if 0 -void -draw_vf_set_sources( struct draw_vertex_fetch *vf, - GLvector4f * const attrib[], - unsigned start ); -#endif - -void -draw_vf_emit_vertex( struct draw_vertex_fetch *vf, - struct vertex_header *vertex, - void *dest ); - -struct draw_vertex_fetch * -draw_vf_create( void ); - -void -draw_vf_destroy( struct draw_vertex_fetch *vf ); - - - -/*********************************************************************** - * Internal functions and structs: - */ - -struct draw_vf_attr; - -typedef void (*draw_vf_extract_func)( const struct draw_vf_attr *a, - float *out, - const uint8_t *v ); - -typedef void (*draw_vf_insert_func)( const struct draw_vf_attr *a, - uint8_t *v, - const float *in ); - -typedef void (*draw_vf_emit_func)( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *dest ); - - - -/** - * Describes how to convert/move a vertex attribute from a vertex - * array to a vertex structure. - */ -struct draw_vf_attr -{ - struct draw_vertex_fetch *vf; - - unsigned format; - unsigned inputsize; - unsigned inputstride; - unsigned vertoffset; /**< position of the attrib in the vertex struct */ - - boolean isconst; /**< read from const data below */ - uint8_t data[16]; - - unsigned attrib; /**< which vertex attrib (0=position, etc) */ - unsigned vertattrsize; /**< size of the attribute in bytes */ - - uint8_t *inputptr; - const draw_vf_insert_func *insert; - draw_vf_insert_func do_insert; - draw_vf_extract_func extract; -}; - -struct draw_vertex_fetch -{ - struct draw_vf_attr attr[PIPE_ATTRIB_MAX]; - unsigned attr_count; - unsigned vertex_stride; - - draw_vf_emit_func emit; - - /* Parameters and constants for codegen: - */ - float identity[4]; - - struct draw_vf_fastpath *fastpath; - - void (*codegen_emit)( struct draw_vertex_fetch *vf ); -}; - - -struct draw_vf_attr_type { - unsigned format; - unsigned size; - unsigned stride; - unsigned offset; -}; - -/** XXX this could be moved into draw_vf.c */ -struct draw_vf_fastpath { - unsigned vertex_stride; - unsigned attr_count; - boolean match_strides; - - struct draw_vf_attr_type *attr; - - draw_vf_emit_func func; - struct draw_vf_fastpath *next; -}; - - -void -draw_vf_register_fastpath( struct draw_vertex_fetch *vtx, - boolean match_strides ); - -void -draw_vf_generic_emit( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *v ); - -void -draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ); - -void -draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ); - - -/** XXX this type and function could probably be moved into draw_vf.c */ -struct draw_vf_format_info { - const char *name; - draw_vf_insert_func insert[4]; - const unsigned attrsize; - const boolean isconst; -}; - -extern const struct draw_vf_format_info -draw_vf_format_info[DRAW_EMIT_MAX]; - - -#endif diff --git a/src/mesa/pipe/draw/draw_vf_generic.c b/src/mesa/pipe/draw/draw_vf_generic.c deleted file mode 100644 index 7a60a9db9c..0000000000 --- a/src/mesa/pipe/draw/draw_vf_generic.c +++ /dev/null @@ -1,585 +0,0 @@ - -/* - * Copyright 2003 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: - * Keith Whitwell - */ - - -#include "pipe/p_compiler.h" -#include "pipe/p_debug.h" -#include "pipe/p_util.h" - -#include "draw_vf.h" - - - -static INLINE void insert_4f_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; - out[3] = in[3]; -} - -static INLINE void insert_4f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; - out[3] = 1; -} - -static INLINE void insert_4f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = 0; - out[3] = 1; -} - -static INLINE void insert_4f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = 0; - out[2] = 0; - out[3] = 1; -} - -static INLINE void insert_3f_xyw_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[3]; -} - -static INLINE void insert_3f_xyw_err( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - (void) a; (void) v; (void) in; - assert(0); -} - -static INLINE void insert_3f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; -} - -static INLINE void insert_3f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = 0; -} - -static INLINE void insert_3f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = 0; - out[2] = 0; -} - - -static INLINE void insert_2f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; -} - -static INLINE void insert_2f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = 0; -} - -static INLINE void insert_1f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; -} - -static INLINE void insert_null( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - (void) a; (void) v; (void) in; -} - -static INLINE void insert_4ub_4f_rgba_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); -} - -static INLINE void insert_4ub_4f_rgba_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_rgba_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[2] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_rgba_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - v[1] = 0; - v[2] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_bgra_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); -} - -static INLINE void insert_4ub_4f_bgra_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_bgra_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[0] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_bgra_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - v[1] = 0; - v[0] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_argb_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); -} - -static INLINE void insert_4ub_4f_argb_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_argb_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - v[3] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_argb_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - v[2] = 0x00; - v[3] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_abgr_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); -} - -static INLINE void insert_4ub_4f_abgr_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_abgr_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - v[1] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_abgr_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - v[2] = 0x00; - v[1] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_3ub_3f_rgb_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); -} - -static INLINE void insert_3ub_3f_rgb_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[2] = 0; -} - -static INLINE void insert_3ub_3f_rgb_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - v[1] = 0; - v[2] = 0; -} - -static INLINE void insert_3ub_3f_bgr_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); -} - -static INLINE void insert_3ub_3f_bgr_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[0] = 0; -} - -static INLINE void insert_3ub_3f_bgr_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - v[1] = 0; - v[0] = 0; -} - - -static INLINE void insert_1ub_1f_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); -} - - -const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX] = -{ - { "1f", - { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, - sizeof(float), FALSE }, - - { "2f", - { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, - 2 * sizeof(float), FALSE }, - - { "3f", - { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, - 3 * sizeof(float), FALSE }, - - { "4f", - { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, - 4 * sizeof(float), FALSE }, - - { "3f_xyw", - { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, - insert_3f_xyw_4 }, - 3 * sizeof(float), FALSE }, - - { "1ub_1f", - { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 }, - sizeof(uint8_t), FALSE }, - - { "3ub_3f_rgb", - { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3, - insert_3ub_3f_rgb_3 }, - 3 * sizeof(uint8_t), FALSE }, - - { "3ub_3f_bgr", - { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3, - insert_3ub_3f_bgr_3 }, - 3 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_rgba", - { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, - insert_4ub_4f_rgba_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_bgra", - { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3, - insert_4ub_4f_bgra_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_argb", - { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3, - insert_4ub_4f_argb_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_abgr", - { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3, - insert_4ub_4f_abgr_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "1f_const", - { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, - sizeof(float), TRUE }, - - { "2f_const", - { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, - 2 * sizeof(float), TRUE }, - - { "3f_const", - { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, - 3 * sizeof(float), TRUE }, - - { "4f_const", - { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, - 4 * sizeof(float), TRUE }, - - { "pad", - { NULL, NULL, NULL, NULL }, - 0, FALSE }, - -}; - - - - -/*********************************************************************** - * Hardwired fastpaths for emitting whole vertices or groups of - * vertices - */ -#define EMIT5(NR, F0, F1, F2, F3, F4, NAME) \ -static void NAME( struct draw_vertex_fetch *vf, \ - unsigned count, \ - uint8_t *v ) \ -{ \ - struct draw_vf_attr *a = vf->attr; \ - unsigned i; \ - \ - for (i = 0 ; i < count ; i++, v += vf->vertex_stride) { \ - if (NR > 0) { \ - F0( &a[0], v + a[0].vertoffset, (float *)a[0].inputptr ); \ - a[0].inputptr += a[0].inputstride; \ - } \ - \ - if (NR > 1) { \ - F1( &a[1], v + a[1].vertoffset, (float *)a[1].inputptr ); \ - a[1].inputptr += a[1].inputstride; \ - } \ - \ - if (NR > 2) { \ - F2( &a[2], v + a[2].vertoffset, (float *)a[2].inputptr ); \ - a[2].inputptr += a[2].inputstride; \ - } \ - \ - if (NR > 3) { \ - F3( &a[3], v + a[3].vertoffset, (float *)a[3].inputptr ); \ - a[3].inputptr += a[3].inputstride; \ - } \ - \ - if (NR > 4) { \ - F4( &a[4], v + a[4].vertoffset, (float *)a[4].inputptr ); \ - a[4].inputptr += a[4].inputstride; \ - } \ - } \ -} - - -#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \ - insert_null, insert_null, NAME) - -#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \ - insert_null, NAME) - -#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \ - insert_null, NAME) - - -EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4) - -EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2) - -EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2) - - -/* Use the codegen paths to select one of a number of hardwired - * fastpaths. - */ -void draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ) -{ - draw_vf_emit_func func = NULL; - - /* Does it fit a hardwired fastpath? Help! this is growing out of - * control! - */ - switch (vf->attr_count) { - case 2: - if (vf->attr[0].do_insert == insert_3f_3 && - vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - func = emit_xyz3_rgba4; - } - break; - case 3: - if (vf->attr[2].do_insert == insert_2f_2) { - if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - if (vf->attr[0].do_insert == insert_4f_4) - func = emit_xyzw4_rgba4_st2; - } - } - break; - case 4: - if (vf->attr[2].do_insert == insert_2f_2 && - vf->attr[3].do_insert == insert_2f_2) { - if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - if (vf->attr[0].do_insert == insert_4f_4) - func = emit_xyzw4_rgba4_st2_st2; - } - } - break; - } - - vf->emit = func; -} - -/*********************************************************************** - * Generic (non-codegen) functions for whole vertices or groups of - * vertices - */ - -void draw_vf_generic_emit( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *v ) -{ - struct draw_vf_attr *a = vf->attr; - const unsigned attr_count = vf->attr_count; - const unsigned stride = vf->vertex_stride; - unsigned i, j; - - for (i = 0 ; i < count ; i++, v += stride) { - for (j = 0; j < attr_count; j++) { - float *in = (float *)a[j].inputptr; - a[j].inputptr += a[j].inputstride; - a[j].do_insert( &a[j], v + a[j].vertoffset, in ); - } - } -} - - diff --git a/src/mesa/pipe/draw/draw_vf_sse.c b/src/mesa/pipe/draw/draw_vf_sse.c deleted file mode 100644 index 1ad2ae756d..0000000000 --- a/src/mesa/pipe/draw/draw_vf_sse.c +++ /dev/null @@ -1,614 +0,0 @@ -/* - * Copyright 2003 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: - * Keith Whitwell - */ - - -#include "simple_list.h" - -#include "pipe/p_compiler.h" - -#include "draw_vf.h" - - -#if defined(USE_SSE_ASM) - -#include "x86/rtasm/x86sse.h" -#include "x86/common_x86_asm.h" - - -#define X 0 -#define Y 1 -#define Z 2 -#define W 3 - - -struct x86_program { - struct x86_function func; - - struct draw_vertex_fetch *vf; - boolean inputs_safe; - boolean outputs_safe; - boolean have_sse2; - - struct x86_reg identity; - struct x86_reg chan0; -}; - - -static struct x86_reg get_identity( struct x86_program *p ) -{ - return p->identity; -} - -static void emit_load4f_4( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movups(&p->func, dest, arg0); -} - -static void emit_load4f_3( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Have to jump through some hoops: - * - * c 0 0 0 - * c 0 0 1 - * 0 0 c 1 - * a b c 1 - */ - sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); - sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); - sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) ); - sse_movlps(&p->func, dest, arg0); -} - -static void emit_load4f_2( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Initialize from identity, then pull in low two words: - */ - sse_movups(&p->func, dest, get_identity(p)); - sse_movlps(&p->func, dest, arg0); -} - -static void emit_load4f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Pull in low word, then swizzle in identity */ - sse_movss(&p->func, dest, arg0); - sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); -} - - - -static void emit_load3f_3( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Over-reads by 1 dword - potential SEGV if input is a vertex - * array. - */ - if (p->inputs_safe) { - sse_movups(&p->func, dest, arg0); - } - else { - /* c 0 0 0 - * c c c c - * a b c c - */ - sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); - sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X)); - sse_movlps(&p->func, dest, arg0); - } -} - -static void emit_load3f_2( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - emit_load4f_2(p, dest, arg0); -} - -static void emit_load3f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - emit_load4f_1(p, dest, arg0); -} - -static void emit_load2f_2( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movlps(&p->func, dest, arg0); -} - -static void emit_load2f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - emit_load4f_1(p, dest, arg0); -} - -static void emit_load1f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movss(&p->func, dest, arg0); -} - -static void (*load[4][4])( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) = { - { emit_load1f_1, - emit_load1f_1, - emit_load1f_1, - emit_load1f_1 }, - - { emit_load2f_1, - emit_load2f_2, - emit_load2f_2, - emit_load2f_2 }, - - { emit_load3f_1, - emit_load3f_2, - emit_load3f_3, - emit_load3f_3 }, - - { emit_load4f_1, - emit_load4f_2, - emit_load4f_3, - emit_load4f_4 } -}; - -static void emit_load( struct x86_program *p, - struct x86_reg dest, - unsigned sz, - struct x86_reg src, - unsigned src_sz) -{ - load[sz-1][src_sz-1](p, dest, src); -} - -static void emit_store4f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movups(&p->func, dest, arg0); -} - -static void emit_store3f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - if (p->outputs_safe) { - /* Emit the extra dword anyway. This may hurt writecombining, - * may cause other problems. - */ - sse_movups(&p->func, dest, arg0); - } - else { - /* Alternate strategy - emit two, shuffle, emit one. - */ - sse_movlps(&p->func, dest, arg0); - sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */ - sse_movss(&p->func, x86_make_disp(dest,8), arg0); - } -} - -static void emit_store2f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movlps(&p->func, dest, arg0); -} - -static void emit_store1f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movss(&p->func, dest, arg0); -} - - -static void (*store[4])( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) = -{ - emit_store1f, - emit_store2f, - emit_store3f, - emit_store4f -}; - -static void emit_store( struct x86_program *p, - struct x86_reg dest, - unsigned sz, - struct x86_reg temp ) - -{ - store[sz-1](p, dest, temp); -} - -static void emit_pack_store_4ub( struct x86_program *p, - struct x86_reg dest, - struct x86_reg temp ) -{ - /* Scale by 255.0 - */ - sse_mulps(&p->func, temp, p->chan0); - - if (p->have_sse2) { - sse2_cvtps2dq(&p->func, temp, temp); - sse2_packssdw(&p->func, temp, temp); - sse2_packuswb(&p->func, temp, temp); - sse_movss(&p->func, dest, temp); - } - else { - struct x86_reg mmx0 = x86_make_reg(file_MMX, 0); - struct x86_reg mmx1 = x86_make_reg(file_MMX, 1); - sse_cvtps2pi(&p->func, mmx0, temp); - sse_movhlps(&p->func, temp, temp); - sse_cvtps2pi(&p->func, mmx1, temp); - mmx_packssdw(&p->func, mmx0, mmx1); - mmx_packuswb(&p->func, mmx0, mmx0); - mmx_movd(&p->func, dest, mmx0); - } -} - -static int get_offset( const void *a, const void *b ) -{ - return (const char *)b - (const char *)a; -} - -/* Not much happens here. Eventually use this function to try and - * avoid saving/reloading the source pointers each vertex (if some of - * them can fit in registers). - */ -static void get_src_ptr( struct x86_program *p, - struct x86_reg srcREG, - struct x86_reg vfREG, - struct draw_vf_attr *a ) -{ - struct draw_vertex_fetch *vf = p->vf; - struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); - - /* Load current a[j].inputptr - */ - x86_mov(&p->func, srcREG, ptr_to_src); -} - -static void update_src_ptr( struct x86_program *p, - struct x86_reg srcREG, - struct x86_reg vfREG, - struct draw_vf_attr *a ) -{ - if (a->inputstride) { - struct draw_vertex_fetch *vf = p->vf; - struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); - - /* add a[j].inputstride (hardcoded value - could just as easily - * pull the stride value from memory each time). - */ - x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride)); - - /* save new value of a[j].inputptr - */ - x86_mov(&p->func, ptr_to_src, srcREG); - } -} - - -/* Lots of hardcoding - * - * EAX -- pointer to current output vertex - * ECX -- pointer to current attribute - * - */ -static boolean build_vertex_emit( struct x86_program *p ) -{ - struct draw_vertex_fetch *vf = p->vf; - unsigned j = 0; - - struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX); - struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX); - struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); - struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI); - struct x86_reg temp = x86_make_reg(file_XMM, 0); - uint8_t *fixup, *label; - - /* Push a few regs? - */ - x86_push(&p->func, countEBP); - x86_push(&p->func, vfESI); - - - /* Get vertex count, compare to zero - */ - x86_xor(&p->func, srcECX, srcECX); - x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2)); - x86_cmp(&p->func, countEBP, srcECX); - fixup = x86_jcc_forward(&p->func, cc_E); - - /* Initialize destination register. - */ - x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3)); - - /* Move argument 1 (vf) into a reg: - */ - x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1)); - - - /* always load, needed or not: - */ - sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0]))); - - /* Note address for loop jump */ - label = x86_get_label(&p->func); - - /* Emit code for each of the attributes. Currently routes - * everything through SSE registers, even when it might be more - * efficient to stick with regular old x86. No optimization or - * other tricks - enough new ground to cover here just getting - * things working. - */ - while (j < vf->attr_count) { - struct draw_vf_attr *a = &vf->attr[j]; - struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset); - - /* Now, load an XMM reg from src, perhaps transform, then save. - * Could be shortcircuited in specific cases: - */ - switch (a->format) { - case DRAW_EMIT_1F: - case DRAW_EMIT_1F_CONST: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 1, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_2F: - case DRAW_EMIT_2F_CONST: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 2, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_3F: - case DRAW_EMIT_3F_CONST: - /* Potentially the worst case - hardcode 2+1 copying: - */ - if (0) { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 3, temp); - update_src_ptr(p, srcECX, vfESI, a); - } - else { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 2, temp); - if (a->inputsize > 2) { - emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1); - emit_store(p, x86_make_disp(dest,8), 1, temp); - } - else { - sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p)); - } - update_src_ptr(p, srcECX, vfESI, a); - } - break; - case DRAW_EMIT_4F: - case DRAW_EMIT_4F_CONST: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 4, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_3F_XYW: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z)); - emit_store(p, dest, 3, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - - case DRAW_EMIT_1UB_1F: - /* Test for PAD3 + 1UB: - */ - if (j > 0 && - a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3) - { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X)); - emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */ - update_src_ptr(p, srcECX, vfESI, a); - } - else { - debug_printf("Can't emit 1ub %x %x %d\n", - a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); - return FALSE; - } - break; - case DRAW_EMIT_3UB_3F_RGB: - case DRAW_EMIT_3UB_3F_BGR: - /* Test for 3UB + PAD1: - */ - if (j == vf->attr_count - 1 || - a[1].vertoffset >= a->vertoffset + 4) { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - if (a->format == DRAW_EMIT_3UB_3F_BGR) - sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - } - /* Test for 3UB + 1UB: - */ - else if (j < vf->attr_count - 1 && - a[1].format == DRAW_EMIT_1UB_1F && - a[1].vertoffset == a->vertoffset + 3) { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - update_src_ptr(p, srcECX, vfESI, a); - - /* Make room for incoming value: - */ - sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); - - get_src_ptr(p, srcECX, vfESI, &a[1]); - emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize); - update_src_ptr(p, srcECX, vfESI, &a[1]); - - /* Rearrange and possibly do BGR conversion: - */ - if (a->format == DRAW_EMIT_3UB_3F_BGR) - sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); - else - sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X)); - - emit_pack_store_4ub(p, dest, temp); - j++; /* NOTE: two attrs consumed */ - } - else { - debug_printf("Can't emit 3ub\n"); - } - return FALSE; /* add this later */ - break; - - case DRAW_EMIT_4UB_4F_RGBA: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4UB_4F_BGRA: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4UB_4F_ARGB: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4UB_4F_ABGR: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - default: - debug_printf("unknown a[%d].format %d\n", j, a->format); - return FALSE; /* catch any new opcodes */ - } - - /* Increment j by at least 1 - may have been incremented above also: - */ - j++; - } - - /* Next vertex: - */ - x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vf->vertex_stride)); - - /* decr count, loop if not zero - */ - x86_dec(&p->func, countEBP); - x86_test(&p->func, countEBP, countEBP); - x86_jcc(&p->func, cc_NZ, label); - - /* Exit mmx state? - */ - if (p->func.need_emms) - mmx_emms(&p->func); - - /* Land forward jump here: - */ - x86_fixup_fwd_jump(&p->func, fixup); - - /* Pop regs and return - */ - x86_pop(&p->func, x86_get_base_reg(vfESI)); - x86_pop(&p->func, countEBP); - x86_ret(&p->func); - - vf->emit = (draw_vf_emit_func)x86_get_func(&p->func); - return TRUE; -} - - - -void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) -{ - struct x86_program p; - - if (!cpu_has_xmm) { - vf->codegen_emit = NULL; - return; - } - - memset(&p, 0, sizeof(p)); - - p.vf = vf; - p.inputs_safe = 0; /* for now */ - p.outputs_safe = 1; /* for now */ - p.have_sse2 = cpu_has_xmm2; - p.identity = x86_make_reg(file_XMM, 6); - p.chan0 = x86_make_reg(file_XMM, 7); - - x86_init_func(&p.func); - - if (build_vertex_emit(&p)) { - draw_vf_register_fastpath( vf, TRUE ); - } - else { - /* Note the failure so that we don't keep trying to codegen an - * impossible state: - */ - draw_vf_register_fastpath( vf, FALSE ); - x86_release_func(&p.func); - } -} - -#else - -void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) -{ - /* Dummy version for when USE_SSE_ASM not defined */ -} - -#endif diff --git a/src/mesa/pipe/draw/draw_wide_prims.c b/src/mesa/pipe/draw/draw_wide_prims.c deleted file mode 100644 index 655774b155..0000000000 --- a/src/mesa/pipe/draw/draw_wide_prims.c +++ /dev/null @@ -1,432 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -struct wide_stage { - struct draw_stage stage; - - float half_line_width; - float half_point_size; - - uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; - uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; - uint num_texcoords; - - int psize_slot; -}; - - - -static INLINE struct wide_stage *wide_stage( struct draw_stage *stage ) -{ - return (struct wide_stage *)stage; -} - - -static void passthrough_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void passthrough_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line(stage->next, header); -} - -static void passthrough_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->tri(stage->next, header); -} - - -/** - * Draw a wide line by drawing a quad (two triangles). - * XXX need to disable polygon stipple. - */ -static void wide_line( struct draw_stage *stage, - struct prim_header *header ) -{ - const struct wide_stage *wide = wide_stage(stage); - const float half_width = wide->half_line_width; - - struct prim_header tri; - - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - const float dx = FABSF(pos0[0] - pos2[0]); - const float dy = FABSF(pos0[1] - pos2[1]); - - /* - * Draw wide line as a quad (two tris) by "stretching" the line along - * X or Y. - * We need to tweak coords in several ways to be conformant here. - */ - - if (dx > dy) { - /* x-major line */ - pos0[1] = pos0[1] - half_width - 0.25f; - pos1[1] = pos1[1] + half_width - 0.25f; - pos2[1] = pos2[1] - half_width - 0.25f; - pos3[1] = pos3[1] + half_width - 0.25f; - if (pos0[0] < pos2[0]) { - /* left to right line */ - pos0[0] -= 0.5f; - pos1[0] -= 0.5f; - pos2[0] -= 0.5f; - pos3[0] -= 0.5f; - } - else { - /* right to left line */ - pos0[0] += 0.5f; - pos1[0] += 0.5f; - pos2[0] += 0.5f; - pos3[0] += 0.5f; - } - } - else { - /* y-major line */ - pos0[0] = pos0[0] - half_width + 0.25f; - pos1[0] = pos1[0] + half_width + 0.25f; - pos2[0] = pos2[0] - half_width + 0.25f; - pos3[0] = pos3[0] + half_width + 0.25f; - if (pos0[1] < pos2[1]) { - /* top to bottom line */ - pos0[1] -= 0.5f; - pos1[1] -= 0.5f; - pos2[1] -= 0.5f; - pos3[1] -= 0.5f; - } - else { - /* bottom to top line */ - pos0[1] += 0.5f; - pos1[1] += 0.5f; - pos2[1] += 0.5f; - pos3[1] += 0.5f; - } - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -/** - * Draw a wide line by drawing a quad, using geometry which will - * fullfill GL's antialiased line requirements. - */ -static void wide_line_aa(struct draw_stage *stage, - struct prim_header *header) -{ - const struct wide_stage *wide = wide_stage(stage); - const float half_width = wide->half_line_width; - struct prim_header tri; - struct vertex_header *v[4]; - float *pos; - float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; - float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; - const float len = (float) sqrt(dx * dx + dy * dy); - uint i; - - dx = dx * half_width / len; - dy = dy * half_width / len; - - /* allocate/dup new verts */ - for (i = 0; i < 4; i++) { - v[i] = dup_vert(stage, header->v[i/2], i); - } - - /* - * Quad for line from v0 to v1: - * - * 1 3 - * +-------------------------+ - * | | - * *v0 v1* - * | | - * +-------------------------+ - * 0 2 - */ - - pos = v[0]->data[0]; - pos[0] += dy; - pos[1] -= dx; - - pos = v[1]->data[0]; - pos[0] -= dy; - pos[1] += dx; - - pos = v[2]->data[0]; - pos[0] += dy; - pos[1] -= dx; - - pos = v[3]->data[0]; - pos[0] -= dy; - pos[1] += dx; - - tri.det = header->det; /* only the sign matters */ - - tri.v[0] = v[2]; tri.v[1] = v[1]; tri.v[2] = v[0]; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v[3]; tri.v[1] = v[1]; tri.v[2] = v[2]; - stage->next->tri( stage->next, &tri ); - -} - - -/** - * Set the vertex texcoords for sprite mode. - * Coords may be left untouched or set to a right-side-up or upside-down - * orientation. - */ -static void set_texcoords(const struct wide_stage *wide, - struct vertex_header *v, const float tc[4]) -{ - uint i; - for (i = 0; i < wide->num_texcoords; i++) { - if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { - uint j = wide->texcoord_slot[i]; - v->data[j][0] = tc[0]; - if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) - v->data[j][1] = 1.0f - tc[1]; - else - v->data[j][1] = tc[1]; - v->data[j][2] = tc[2]; - v->data[j][3] = tc[3]; - } - } -} - - -/* If there are lots of sprite points (and why wouldn't there be?) it - * would probably be more sensible to change hardware setup to - * optimize this rather than doing the whole thing in software like - * this. - */ -static void wide_point( struct draw_stage *stage, - struct prim_header *header ) -{ - const struct wide_stage *wide = wide_stage(stage); - const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; - float half_size; - float left_adj, right_adj; - - struct prim_header tri; - - /* four dups of original vertex */ - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - /* point size is either per-vertex or fixed size */ - if (wide->psize_slot >= 0) { - half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; - } - else { - half_size = wide->half_point_size; - } - - left_adj = -half_size + 0.25f; - right_adj = half_size + 0.25f; - - pos0[0] += left_adj; - pos0[1] -= half_size; - - pos1[0] += left_adj; - pos1[1] += half_size; - - pos2[0] += right_adj; - pos2[1] -= half_size; - - pos3[0] += right_adj; - pos3[1] += half_size; - - if (sprite) { - static const float tex00[4] = { 0, 0, 0, 1 }; - static const float tex01[4] = { 0, 1, 0, 1 }; - static const float tex11[4] = { 1, 1, 0, 1 }; - static const float tex10[4] = { 1, 0, 0, 1 }; - set_texcoords( wide, v0, tex00 ); - set_texcoords( wide, v1, tex01 ); - set_texcoords( wide, v2, tex10 ); - set_texcoords( wide, v3, tex11 ); - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -static void wide_first_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct wide_stage *wide = wide_stage(stage); - struct draw_context *draw = stage->draw; - - wide->half_point_size = 0.5f * draw->rasterizer->point_size; - - if (draw->rasterizer->point_size != 1.0) { - stage->point = wide_point; - } - else { - stage->point = passthrough_point; - } - - if (draw->rasterizer->point_sprite) { - /* find vertex shader texcoord outputs */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i, j = 0; - for (i = 0; i < vs->state->num_outputs; i++) { - if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { - wide->texcoord_slot[j] = i; - wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; - j++; - } - } - wide->num_texcoords = j; - } - - wide->psize_slot = -1; - - if (draw->rasterizer->point_size_per_vertex) { - /* find PSIZ vertex output */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i; - for (i = 0; i < vs->state->num_outputs; i++) { - if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { - wide->psize_slot = i; - break; - } - } - } - - stage->point( stage, header ); -} - - - -static void wide_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct wide_stage *wide = wide_stage(stage); - struct draw_context *draw = stage->draw; - - wide->half_line_width = 0.5f * draw->rasterizer->line_width; - - if (draw->rasterizer->line_width != 1.0) { - if (draw->rasterizer->line_smooth) - wide->stage.line = wide_line_aa; - else - wide->stage.line = wide_line; - } - else { - wide->stage.line = passthrough_line; - } - - stage->line( stage, header ); -} - - -static void wide_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->line = wide_first_line; - stage->point = wide_first_point; - stage->next->flush( stage->next, flags ); -} - - -static void wide_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void wide_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -struct draw_stage *draw_wide_stage( struct draw_context *draw ) -{ - struct wide_stage *wide = CALLOC_STRUCT(wide_stage); - - draw_alloc_temp_verts( &wide->stage, 4 ); - - wide->stage.draw = draw; - wide->stage.next = NULL; - wide->stage.point = wide_first_point; - wide->stage.line = wide_first_line; - wide->stage.tri = passthrough_tri; - wide->stage.flush = wide_flush; - wide->stage.reset_stipple_counter = wide_reset_stipple_counter; - wide->stage.destroy = wide_destroy; - - return &wide->stage; -} diff --git a/src/mesa/pipe/failover/Makefile b/src/mesa/pipe/failover/Makefile deleted file mode 100644 index 72d0895c74..0000000000 --- a/src/mesa/pipe/failover/Makefile +++ /dev/null @@ -1,21 +0,0 @@ - -TOP = ../../../.. -include $(TOP)/configs/current - -LIBNAME = failover - -DRIVER_SOURCES = \ - fo_state.c \ - fo_state_emit.c \ - fo_context.c - -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -include ../Makefile.template - -symlinks: - diff --git a/src/mesa/pipe/failover/fo_context.c b/src/mesa/pipe/failover/fo_context.c deleted file mode 100644 index 7ce4a7df17..0000000000 --- a/src/mesa/pipe/failover/fo_context.c +++ /dev/null @@ -1,155 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "pipe/p_defines.h" -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" -#include "pipe/p_context.h" - -#include "fo_context.h" -#include "fo_winsys.h" - - - -static void failover_destroy( struct pipe_context *pipe ) -{ - struct failover_context *failover = failover_context( pipe ); - - free( failover ); -} - - - -static boolean failover_draw_elements( struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned prim, unsigned start, unsigned count) -{ - struct failover_context *failover = failover_context( pipe ); - - /* If there has been any statechange since last time, try hardware - * rendering again: - */ - if (failover->dirty) { - failover->mode = FO_HW; - } - - /* Try hardware: - */ - if (failover->mode == FO_HW) { - if (!failover->hw->draw_elements( failover->hw, - indexBuffer, - indexSize, - prim, - start, - count )) { - - failover->hw->flush( failover->hw, ~0 ); - failover->mode = FO_SW; - } - } - - /* Possibly try software: - */ - if (failover->mode == FO_SW) { - - if (failover->dirty) - failover_state_emit( failover ); - - failover->sw->draw_elements( failover->sw, - indexBuffer, - indexSize, - prim, - start, - count ); - - /* Be ready to switch back to hardware rendering without an - * intervening flush. Unlikely to be much performance impact to - * this: - */ - failover->sw->flush( failover->sw, ~0 ); - } - - return TRUE; -} - - -static boolean failover_draw_arrays( struct pipe_context *pipe, - unsigned prim, unsigned start, unsigned count) -{ - return failover_draw_elements(pipe, NULL, 0, prim, start, count); -} - - - -struct pipe_context *failover_create( struct pipe_context *hw, - struct pipe_context *sw ) -{ - struct failover_context *failover = CALLOC_STRUCT(failover_context); - if (failover == NULL) - return NULL; - - failover->hw = hw; - failover->sw = sw; - failover->pipe.winsys = hw->winsys; - failover->pipe.destroy = failover_destroy; - failover->pipe.is_format_supported = hw->is_format_supported; - failover->pipe.get_name = hw->get_name; - failover->pipe.get_vendor = hw->get_vendor; - failover->pipe.get_param = hw->get_param; - failover->pipe.get_paramf = hw->get_paramf; - - failover->pipe.draw_arrays = failover_draw_arrays; - failover->pipe.draw_elements = failover_draw_elements; - failover->pipe.clear = hw->clear; - - /* No software occlusion fallback (or other optional functionality) - * at this point - if the hardware doesn't support it, don't - * advertise it to the application. - */ - failover->pipe.begin_query = hw->begin_query; - failover->pipe.end_query = hw->end_query; - - failover_init_state_functions( failover ); - -#if 0 - failover->pipe.surface_alloc = hw->surface_alloc; -#endif - failover->pipe.get_tex_surface = hw->get_tex_surface; - - failover->pipe.surface_copy = hw->surface_copy; - failover->pipe.surface_fill = hw->surface_fill; - failover->pipe.texture_create = hw->texture_create; - failover->pipe.texture_release = hw->texture_release; - failover->pipe.flush = hw->flush; - - failover->dirty = 0; - - return &failover->pipe; -} - diff --git a/src/mesa/pipe/failover/fo_context.h b/src/mesa/pipe/failover/fo_context.h deleted file mode 100644 index 1dc87291c9..0000000000 --- a/src/mesa/pipe/failover/fo_context.h +++ /dev/null @@ -1,114 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef FO_CONTEXT_H -#define FO_CONTEXT_H - -#include "pipe/p_state.h" -#include "pipe/p_context.h" - - - -#define FO_NEW_VIEWPORT 0x1 -#define FO_NEW_RASTERIZER 0x2 -#define FO_NEW_FRAGMENT_SHADER 0x4 -#define FO_NEW_BLEND 0x8 -#define FO_NEW_CLIP 0x10 -#define FO_NEW_SCISSOR 0x20 -#define FO_NEW_STIPPLE 0x40 -#define FO_NEW_FRAMEBUFFER 0x80 -#define FO_NEW_ALPHA_TEST 0x100 -#define FO_NEW_DEPTH_STENCIL 0x200 -#define FO_NEW_SAMPLER 0x400 -#define FO_NEW_TEXTURE 0x800 -#define FO_NEW_VERTEX 0x2000 -#define FO_NEW_VERTEX_SHADER 0x4000 -#define FO_NEW_BLEND_COLOR 0x8000 -#define FO_NEW_CLEAR_COLOR 0x10000 -#define FO_NEW_VERTEX_BUFFER 0x20000 -#define FO_NEW_VERTEX_ELEMENT 0x40000 - - - -#define FO_HW 0 -#define FO_SW 1 - -struct fo_state { - void *sw_state; - void *hw_state; -}; -struct failover_context { - struct pipe_context pipe; /**< base class */ - - - /* The most recent drawing state as set by the driver: - */ - const struct fo_state *blend; - const struct fo_state *sampler[PIPE_MAX_SAMPLERS]; - const struct fo_state *depth_stencil; - const struct fo_state *rasterizer; - const struct fo_state *fragment_shader; - const struct fo_state *vertex_shader; - - struct pipe_blend_color blend_color; - struct pipe_clip_state clip; - struct pipe_framebuffer_state framebuffer; - struct pipe_poly_stipple poly_stipple; - struct pipe_scissor_state scissor; - struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; - struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - - unsigned dirty; - unsigned dirty_sampler; - unsigned dirty_texture; - unsigned dirty_vertex_buffer; - unsigned dirty_vertex_element; - - - unsigned mode; - struct pipe_context *hw; - struct pipe_context *sw; -}; - - - -void failover_init_state_functions( struct failover_context *failover ); -void failover_state_emit( struct failover_context *failover ); - -static INLINE struct failover_context * -failover_context( struct pipe_context *pipe ) -{ - return (struct failover_context *)pipe; -} - - -#endif /* FO_CONTEXT_H */ diff --git a/src/mesa/pipe/failover/fo_state.c b/src/mesa/pipe/failover/fo_state.c deleted file mode 100644 index 0fc5568da1..0000000000 --- a/src/mesa/pipe/failover/fo_state.c +++ /dev/null @@ -1,457 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "fo_context.h" - - -/* This looks like a lot of work at the moment - we're keeping a - * duplicate copy of the state up-to-date. - * - * This can change in two ways: - * - With constant state objects we would only need to save a pointer, - * not the whole object. - * - By adding a callback in the state tracker to re-emit state. The - * state tracker knows the current state already and can re-emit it - * without additional complexity. - * - * This works as a proof-of-concept, but a final version will have - * lower overheads. - */ - - - -static void * -failover_create_blend_state( struct pipe_context *pipe, - const struct pipe_blend_state *blend ) -{ - struct fo_state *state = malloc(sizeof(struct fo_state)); - struct failover_context *failover = failover_context(pipe); - - state->sw_state = failover->sw->create_blend_state(failover->sw, blend); - state->hw_state = failover->hw->create_blend_state(failover->hw, blend); - - return state; -} - -static void -failover_bind_blend_state( struct pipe_context *pipe, - void *blend ) -{ - struct failover_context *failover = failover_context(pipe); - struct fo_state *state = (struct fo_state *)blend; - failover->blend = state; - failover->dirty |= FO_NEW_BLEND; - failover->sw->bind_blend_state( failover->sw, state->sw_state ); - failover->hw->bind_blend_state( failover->hw, state->hw_state ); -} - -static void -failover_delete_blend_state( struct pipe_context *pipe, - void *blend ) -{ - struct fo_state *state = (struct fo_state*)blend; - struct failover_context *failover = failover_context(pipe); - - failover->sw->delete_blend_state(failover->sw, state->sw_state); - failover->hw->delete_blend_state(failover->hw, state->hw_state); - state->sw_state = 0; - state->hw_state = 0; - free(state); -} - -static void -failover_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ) -{ - struct failover_context *failover = failover_context(pipe); - - failover->blend_color = *blend_color; - failover->dirty |= FO_NEW_BLEND_COLOR; - failover->sw->set_blend_color( failover->sw, blend_color ); - failover->hw->set_blend_color( failover->hw, blend_color ); -} - -static void -failover_set_clip_state( struct pipe_context *pipe, - const struct pipe_clip_state *clip ) -{ - struct failover_context *failover = failover_context(pipe); - - failover->clip = *clip; - failover->dirty |= FO_NEW_CLIP; - failover->sw->set_clip_state( failover->sw, clip ); - failover->hw->set_clip_state( failover->hw, clip ); -} - - -static void * -failover_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *templ) -{ - struct fo_state *state = malloc(sizeof(struct fo_state)); - struct failover_context *failover = failover_context(pipe); - - state->sw_state = failover->sw->create_depth_stencil_alpha_state(failover->sw, templ); - state->hw_state = failover->hw->create_depth_stencil_alpha_state(failover->hw, templ); - - return state; -} - -static void -failover_bind_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - struct failover_context *failover = failover_context(pipe); - struct fo_state *state = (struct fo_state *)depth_stencil; - failover->depth_stencil = state; - failover->dirty |= FO_NEW_DEPTH_STENCIL; - failover->sw->bind_depth_stencil_alpha_state(failover->sw, state->sw_state); - failover->hw->bind_depth_stencil_alpha_state(failover->hw, state->hw_state); -} - -static void -failover_delete_depth_stencil_state(struct pipe_context *pipe, - void *ds) -{ - struct fo_state *state = (struct fo_state*)ds; - struct failover_context *failover = failover_context(pipe); - - failover->sw->delete_depth_stencil_alpha_state(failover->sw, state->sw_state); - failover->hw->delete_depth_stencil_alpha_state(failover->hw, state->hw_state); - state->sw_state = 0; - state->hw_state = 0; - free(state); -} - -static void -failover_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *framebuffer) -{ - struct failover_context *failover = failover_context(pipe); - - failover->framebuffer = *framebuffer; - failover->dirty |= FO_NEW_FRAMEBUFFER; - failover->sw->set_framebuffer_state( failover->sw, framebuffer ); - failover->hw->set_framebuffer_state( failover->hw, framebuffer ); -} - - -static void * -failover_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - struct fo_state *state = malloc(sizeof(struct fo_state)); - struct failover_context *failover = failover_context(pipe); - - state->sw_state = failover->sw->create_fs_state(failover->sw, templ); - state->hw_state = failover->hw->create_fs_state(failover->hw, templ); - - return state; -} - -static void -failover_bind_fs_state(struct pipe_context *pipe, void *fs) -{ - struct failover_context *failover = failover_context(pipe); - struct fo_state *state = (struct fo_state*)fs; - failover->fragment_shader = state; - failover->dirty |= FO_NEW_FRAGMENT_SHADER; - failover->sw->bind_fs_state(failover->sw, state->sw_state); - failover->hw->bind_fs_state(failover->hw, state->hw_state); -} - -static void -failover_delete_fs_state(struct pipe_context *pipe, - void *fs) -{ - struct fo_state *state = (struct fo_state*)fs; - struct failover_context *failover = failover_context(pipe); - - failover->sw->delete_fs_state(failover->sw, state->sw_state); - failover->hw->delete_fs_state(failover->hw, state->hw_state); - state->sw_state = 0; - state->hw_state = 0; - free(state); -} - -static void * -failover_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - struct fo_state *state = malloc(sizeof(struct fo_state)); - struct failover_context *failover = failover_context(pipe); - - state->sw_state = failover->sw->create_vs_state(failover->sw, templ); - state->hw_state = failover->hw->create_vs_state(failover->hw, templ); - - return state; -} - -static void -failover_bind_vs_state(struct pipe_context *pipe, - void *vs) -{ - struct failover_context *failover = failover_context(pipe); - - struct fo_state *state = (struct fo_state*)vs; - failover->vertex_shader = state; - failover->dirty |= FO_NEW_VERTEX_SHADER; - failover->sw->bind_vs_state(failover->sw, state->sw_state); - failover->hw->bind_vs_state(failover->hw, state->hw_state); -} - -static void -failover_delete_vs_state(struct pipe_context *pipe, - void *vs) -{ - struct fo_state *state = (struct fo_state*)vs; - struct failover_context *failover = failover_context(pipe); - - failover->sw->delete_vs_state(failover->sw, state->sw_state); - failover->hw->delete_vs_state(failover->hw, state->hw_state); - state->sw_state = 0; - state->hw_state = 0; - free(state); -} - -static void -failover_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ - struct failover_context *failover = failover_context(pipe); - - failover->poly_stipple = *stipple; - failover->dirty |= FO_NEW_STIPPLE; - failover->sw->set_polygon_stipple( failover->sw, stipple ); - failover->hw->set_polygon_stipple( failover->hw, stipple ); -} - - -static void * -failover_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *templ) -{ - struct fo_state *state = malloc(sizeof(struct fo_state)); - struct failover_context *failover = failover_context(pipe); - - state->sw_state = failover->sw->create_rasterizer_state(failover->sw, templ); - state->hw_state = failover->hw->create_rasterizer_state(failover->hw, templ); - - return state; -} - -static void -failover_bind_rasterizer_state(struct pipe_context *pipe, - void *raster) -{ - struct failover_context *failover = failover_context(pipe); - - struct fo_state *state = (struct fo_state*)raster; - failover->rasterizer = state; - failover->dirty |= FO_NEW_RASTERIZER; - failover->sw->bind_rasterizer_state(failover->sw, state->sw_state); - failover->hw->bind_rasterizer_state(failover->hw, state->hw_state); -} - -static void -failover_delete_rasterizer_state(struct pipe_context *pipe, - void *raster) -{ - struct fo_state *state = (struct fo_state*)raster; - struct failover_context *failover = failover_context(pipe); - - failover->sw->delete_rasterizer_state(failover->sw, state->sw_state); - failover->hw->delete_rasterizer_state(failover->hw, state->hw_state); - state->sw_state = 0; - state->hw_state = 0; - free(state); -} - - -static void -failover_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct failover_context *failover = failover_context(pipe); - - failover->scissor = *scissor; - failover->dirty |= FO_NEW_SCISSOR; - failover->sw->set_scissor_state( failover->sw, scissor ); - failover->hw->set_scissor_state( failover->hw, scissor ); -} - - -static void * -failover_create_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *templ) -{ - struct fo_state *state = malloc(sizeof(struct fo_state)); - struct failover_context *failover = failover_context(pipe); - - state->sw_state = failover->sw->create_sampler_state(failover->sw, templ); - state->hw_state = failover->hw->create_sampler_state(failover->hw, templ); - - return state; -} - -static void -failover_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) -{ - struct failover_context *failover = failover_context(pipe); - struct fo_state *state = (struct fo_state*)sampler; - failover->sampler[unit] = state; - failover->dirty |= FO_NEW_SAMPLER; - failover->dirty_sampler |= (1<sw->bind_sampler_state(failover->sw, unit, - state->sw_state); - failover->hw->bind_sampler_state(failover->hw, unit, - state->hw_state); -} - -static void -failover_delete_sampler_state(struct pipe_context *pipe, void *sampler) -{ - struct fo_state *state = (struct fo_state*)sampler; - struct failover_context *failover = failover_context(pipe); - - failover->sw->delete_sampler_state(failover->sw, state->sw_state); - failover->hw->delete_sampler_state(failover->hw, state->hw_state); - state->sw_state = 0; - state->hw_state = 0; - free(state); -} - - -static void -failover_set_sampler_texture(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) -{ - struct failover_context *failover = failover_context(pipe); - - failover->texture[unit] = texture; - failover->dirty |= FO_NEW_TEXTURE; - failover->dirty_texture |= (1<sw->set_sampler_texture( failover->sw, unit, texture ); - failover->hw->set_sampler_texture( failover->hw, unit, texture ); -} - - -static void -failover_set_viewport_state( struct pipe_context *pipe, - const struct pipe_viewport_state *viewport ) -{ - struct failover_context *failover = failover_context(pipe); - - failover->viewport = *viewport; - failover->dirty |= FO_NEW_VIEWPORT; - failover->sw->set_viewport_state( failover->sw, viewport ); - failover->hw->set_viewport_state( failover->hw, viewport ); -} - - -static void -failover_set_vertex_buffer(struct pipe_context *pipe, - unsigned unit, - const struct pipe_vertex_buffer *vertex_buffer) -{ - struct failover_context *failover = failover_context(pipe); - - failover->vertex_buffer[unit] = *vertex_buffer; - failover->dirty |= FO_NEW_VERTEX_BUFFER; - failover->dirty_vertex_buffer |= (1<sw->set_vertex_buffer( failover->sw, unit, vertex_buffer ); - failover->hw->set_vertex_buffer( failover->hw, unit, vertex_buffer ); -} - - -static void -failover_set_vertex_element(struct pipe_context *pipe, - unsigned unit, - const struct pipe_vertex_element *vertex_element) -{ - struct failover_context *failover = failover_context(pipe); - - failover->vertex_element[unit] = *vertex_element; - failover->dirty |= FO_NEW_VERTEX_ELEMENT; - failover->dirty_vertex_element |= (1<sw->set_vertex_element( failover->sw, unit, vertex_element ); - failover->hw->set_vertex_element( failover->hw, unit, vertex_element ); -} - -void -failover_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) -{ - struct failover_context *failover = failover_context(pipe); - - assert(shader < PIPE_SHADER_TYPES); - assert(index == 0); - - failover->sw->set_constant_buffer(failover->sw, shader, index, buf); - failover->hw->set_constant_buffer(failover->hw, shader, index, buf); -} - - -void -failover_init_state_functions( struct failover_context *failover ) -{ - failover->pipe.create_blend_state = failover_create_blend_state; - failover->pipe.bind_blend_state = failover_bind_blend_state; - failover->pipe.delete_blend_state = failover_delete_blend_state; - failover->pipe.create_sampler_state = failover_create_sampler_state; - failover->pipe.bind_sampler_state = failover_bind_sampler_state; - failover->pipe.delete_sampler_state = failover_delete_sampler_state; - failover->pipe.create_depth_stencil_alpha_state = failover_create_depth_stencil_state; - failover->pipe.bind_depth_stencil_alpha_state = failover_bind_depth_stencil_state; - failover->pipe.delete_depth_stencil_alpha_state = failover_delete_depth_stencil_state; - failover->pipe.create_rasterizer_state = failover_create_rasterizer_state; - failover->pipe.bind_rasterizer_state = failover_bind_rasterizer_state; - failover->pipe.delete_rasterizer_state = failover_delete_rasterizer_state; - failover->pipe.create_fs_state = failover_create_fs_state; - failover->pipe.bind_fs_state = failover_bind_fs_state; - failover->pipe.delete_fs_state = failover_delete_fs_state; - failover->pipe.create_vs_state = failover_create_vs_state; - failover->pipe.bind_vs_state = failover_bind_vs_state; - failover->pipe.delete_vs_state = failover_delete_vs_state; - - failover->pipe.set_blend_color = failover_set_blend_color; - failover->pipe.set_clip_state = failover_set_clip_state; - failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; - failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; - failover->pipe.set_scissor_state = failover_set_scissor_state; - failover->pipe.set_sampler_texture = failover_set_sampler_texture; - failover->pipe.set_viewport_state = failover_set_viewport_state; - failover->pipe.set_vertex_buffer = failover_set_vertex_buffer; - failover->pipe.set_vertex_element = failover_set_vertex_element; - failover->pipe.set_constant_buffer = failover_set_constant_buffer; -} diff --git a/src/mesa/pipe/failover/fo_state_emit.c b/src/mesa/pipe/failover/fo_state_emit.c deleted file mode 100644 index c663dd4947..0000000000 --- a/src/mesa/pipe/failover/fo_state_emit.c +++ /dev/null @@ -1,137 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "fo_context.h" - -/* This looks like a lot of work at the moment - we're keeping a - * duplicate copy of the state up-to-date. - * - * This can change in two ways: - * - With constant state objects we would only need to save a pointer, - * not the whole object. - * - By adding a callback in the state tracker to re-emit state. The - * state tracker knows the current state already and can re-emit it - * without additional complexity. - * - * This works as a proof-of-concept, but a final version will have - * lower overheads. - */ - - -/* Bring the software pipe uptodate with current state. - * - * With constant state objects we would probably just send all state - * to both rasterizers all the time??? - */ -void -failover_state_emit( struct failover_context *failover ) -{ - unsigned i; - - if (failover->dirty & FO_NEW_BLEND) - failover->sw->bind_blend_state( failover->sw, - failover->blend->sw_state ); - - if (failover->dirty & FO_NEW_BLEND_COLOR) - failover->sw->set_blend_color( failover->sw, &failover->blend_color ); - - if (failover->dirty & FO_NEW_CLIP) - failover->sw->set_clip_state( failover->sw, &failover->clip ); - - if (failover->dirty & FO_NEW_DEPTH_STENCIL) - failover->sw->bind_depth_stencil_alpha_state( failover->sw, - failover->depth_stencil->sw_state ); - - if (failover->dirty & FO_NEW_FRAMEBUFFER) - failover->sw->set_framebuffer_state( failover->sw, &failover->framebuffer ); - - if (failover->dirty & FO_NEW_FRAGMENT_SHADER) - failover->sw->bind_fs_state( failover->sw, - failover->fragment_shader->sw_state ); - - if (failover->dirty & FO_NEW_VERTEX_SHADER) - failover->sw->bind_vs_state( failover->sw, - failover->vertex_shader->sw_state ); - - if (failover->dirty & FO_NEW_STIPPLE) - failover->sw->set_polygon_stipple( failover->sw, &failover->poly_stipple ); - - if (failover->dirty & FO_NEW_RASTERIZER) - failover->sw->bind_rasterizer_state( failover->sw, - failover->rasterizer->sw_state ); - - if (failover->dirty & FO_NEW_SCISSOR) - failover->sw->set_scissor_state( failover->sw, &failover->scissor ); - - if (failover->dirty & FO_NEW_VIEWPORT) - failover->sw->set_viewport_state( failover->sw, &failover->viewport ); - - if (failover->dirty & FO_NEW_SAMPLER) { - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - if (failover->dirty_sampler & (1<sw->bind_sampler_state( failover->sw, i, - failover->sampler[i]->sw_state ); - } - } - } - - if (failover->dirty & FO_NEW_TEXTURE) { - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - if (failover->dirty_texture & (1<sw->set_sampler_texture( failover->sw, i, - failover->texture[i] ); - } - } - } - - if (failover->dirty & FO_NEW_VERTEX_BUFFER) { - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { - if (failover->dirty_vertex_buffer & (1<sw->set_vertex_buffer( failover->sw, i, - &failover->vertex_buffer[i] ); - } - } - } - - if (failover->dirty & FO_NEW_VERTEX_ELEMENT) { - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { - if (failover->dirty_vertex_element & (1<sw->set_vertex_element( failover->sw, i, - &failover->vertex_element[i] ); - } - } - } - - failover->dirty = 0; - failover->dirty_vertex_element = 0; - failover->dirty_vertex_buffer = 0; - failover->dirty_texture = 0; - failover->dirty_sampler = 0; -} diff --git a/src/mesa/pipe/failover/fo_winsys.h b/src/mesa/pipe/failover/fo_winsys.h deleted file mode 100644 index a8ce997a1f..0000000000 --- a/src/mesa/pipe/failover/fo_winsys.h +++ /dev/null @@ -1,45 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef FO_WINSYS_H -#define FO_WINSYS_H - - -/* This is the interface that failover requires any window system - * hosting it to implement. This is the only include file in failover - * which is public. - */ - - -struct pipe_context; - - -struct pipe_context *failover_create( struct pipe_context *hw, - struct pipe_context *sw ); - - -#endif /* FO_WINSYS_H */ diff --git a/src/mesa/pipe/i915simple/Makefile b/src/mesa/pipe/i915simple/Makefile deleted file mode 100644 index 2f91de3afc..0000000000 --- a/src/mesa/pipe/i915simple/Makefile +++ /dev/null @@ -1,38 +0,0 @@ - -TOP = ../../../.. -include $(TOP)/configs/current - -LIBNAME = i915simple - -DRIVER_SOURCES = \ - i915_blit.c \ - i915_clear.c \ - i915_flush.c \ - i915_context.c \ - i915_context.c \ - i915_debug.c \ - i915_debug_fp.c \ - i915_state.c \ - i915_state_immediate.c \ - i915_state_dynamic.c \ - i915_state_derived.c \ - i915_state_emit.c \ - i915_state_sampler.c \ - i915_strings.c \ - i915_prim_emit.c \ - i915_prim_vbuf.c \ - i915_texture.c \ - i915_fpc_emit.c \ - i915_fpc_translate.c \ - i915_surface.c - -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -include ../Makefile.template - -symlinks: - diff --git a/src/mesa/pipe/i915simple/SConscript b/src/mesa/pipe/i915simple/SConscript deleted file mode 100644 index f5fb96b995..0000000000 --- a/src/mesa/pipe/i915simple/SConscript +++ /dev/null @@ -1,29 +0,0 @@ -Import('*') - -env = env.Clone() - -i915simple = env.ConvenienceLibrary( - target = 'i915simple', - source = [ - 'i915_blit.c', - 'i915_clear.c', - 'i915_context.c', - 'i915_debug.c', - 'i915_debug_fp.c', - 'i915_flush.c', - 'i915_fpc_emit.c', - 'i915_fpc_translate.c', - 'i915_prim_emit.c', - 'i915_prim_vbuf.c', - 'i915_state.c', - 'i915_state_derived.c', - 'i915_state_dynamic.c', - 'i915_state_emit.c', - 'i915_state_immediate.c', - 'i915_state_sampler.c', - 'i915_strings.c', - 'i915_surface.c', - 'i915_texture.c', - ]) - -Export('i915simple') diff --git a/src/mesa/pipe/i915simple/i915_batch.h b/src/mesa/pipe/i915simple/i915_batch.h deleted file mode 100644 index fb88cd6db0..0000000000 --- a/src/mesa/pipe/i915simple/i915_batch.h +++ /dev/null @@ -1,54 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef I915_BATCH_H -#define I915_BATCH_H - -#include "i915_winsys.h" -#include "i915_debug.h" - -#define BATCH_LOCALS - -#define BEGIN_BATCH( dwords, relocs ) \ - (i915->batch_start = i915->winsys->batch_start( i915->winsys, dwords, relocs )) - -#define OUT_BATCH( dword ) \ - i915->winsys->batch_dword( i915->winsys, dword ) - -#define OUT_RELOC( buf, flags, delta ) \ - i915->winsys->batch_reloc( i915->winsys, buf, flags, delta ) - -#define ADVANCE_BATCH() - -#define FLUSH_BATCH() do { \ - if (0) i915_dump_batchbuffer( i915 ); \ - i915->winsys->batch_flush( i915->winsys ); \ - i915->batch_start = NULL; \ - i915->hardware_dirty = ~0; \ -} while (0) - -#endif diff --git a/src/mesa/pipe/i915simple/i915_blit.c b/src/mesa/pipe/i915simple/i915_blit.c deleted file mode 100644 index db4671ff55..0000000000 --- a/src/mesa/pipe/i915simple/i915_blit.c +++ /dev/null @@ -1,162 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "i915_context.h" -#include "i915_winsys.h" -#include "i915_blit.h" -#include "i915_reg.h" -#include "i915_batch.h" - -#define FILE_DEBUG_FLAG DEBUG_BLIT - -void -i915_fill_blit(struct i915_context *i915, - unsigned cpp, - short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - short x, short y, - short w, short h, - unsigned color) -{ - unsigned BR13, CMD; - BATCH_LOCALS; - - dst_pitch *= (short) cpp; - - switch (cpp) { - case 1: - case 2: - case 3: - BR13 = dst_pitch | (0xF0 << 16) | (1 << 24); - CMD = XY_COLOR_BLT_CMD; - break; - case 4: - BR13 = dst_pitch | (0xF0 << 16) | (1 << 24) | (1 << 25); - CMD = (XY_COLOR_BLT_CMD | XY_COLOR_BLT_WRITE_ALPHA | - XY_COLOR_BLT_WRITE_RGB); - break; - default: - return; - } - -// DBG("%s dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", -// __FUNCTION__, dst_buffer, dst_pitch, dst_offset, x, y, w, h); - - - if (!BEGIN_BATCH(6, 1)) { - FLUSH_BATCH(); - assert(BEGIN_BATCH(6, 1)); - } - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((y << 16) | x); - OUT_BATCH(((y + h) << 16) | (x + w)); - OUT_RELOC( dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); - OUT_BATCH(color); - ADVANCE_BATCH(); -} - - -void -i915_copy_blit( struct i915_context *i915, - unsigned do_flip, - unsigned cpp, - short src_pitch, - struct pipe_buffer *src_buffer, - unsigned src_offset, - short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - short src_x, short src_y, - short dst_x, short dst_y, - short w, short h ) -{ - unsigned CMD, BR13; - int dst_y2 = dst_y + h; - int dst_x2 = dst_x + w; - BATCH_LOCALS; - - - I915_DBG(i915, - "%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n", - __FUNCTION__, - src_buffer, src_pitch, src_offset, src_x, src_y, - dst_buffer, dst_pitch, dst_offset, dst_x, dst_y, w, h); - - src_pitch *= (short) cpp; - dst_pitch *= (short) cpp; - - switch (cpp) { - case 1: - case 2: - case 3: - BR13 = (((int) dst_pitch) & 0xffff) | - (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - break; - case 4: - BR13 = - (((int) dst_pitch) & 0xffff) | - (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = - (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - break; - default: - return; - } - - if (dst_y2 < dst_y || - dst_x2 < dst_x) { - return; - } - - /* Hardware can handle negative pitches but loses the ability to do - * proper overlapping blits in that case. We don't really have a - * need for either at this stage. - */ - assert (dst_pitch > 0 && src_pitch > 0); - - - if (!BEGIN_BATCH(8, 2)) { - FLUSH_BATCH(); - assert(BEGIN_BATCH(8, 2)); - } - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((dst_y << 16) | dst_x); - OUT_BATCH((dst_y2 << 16) | dst_x2); - OUT_RELOC(dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); - OUT_BATCH((src_y << 16) | src_x); - OUT_BATCH(((int) src_pitch & 0xffff)); - OUT_RELOC(src_buffer, I915_BUFFER_ACCESS_READ, src_offset); - ADVANCE_BATCH(); -} - - diff --git a/src/mesa/pipe/i915simple/i915_blit.h b/src/mesa/pipe/i915simple/i915_blit.h deleted file mode 100644 index 6e5b44e124..0000000000 --- a/src/mesa/pipe/i915simple/i915_blit.h +++ /dev/null @@ -1,55 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef I915_BLIT_H -#define I915_BLIT_H - -#include "i915_context.h" - -extern void i915_copy_blit(struct i915_context *i915, - unsigned do_flip, - unsigned cpp, - short src_pitch, - struct pipe_buffer *src_buffer, - unsigned src_offset, - short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - short srcx, short srcy, - short dstx, short dsty, - short w, short h ); - -extern void i915_fill_blit(struct i915_context *i915, - unsigned cpp, - short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - short x, short y, - short w, short h, unsigned color); - - -#endif diff --git a/src/mesa/pipe/i915simple/i915_clear.c b/src/mesa/pipe/i915simple/i915_clear.c deleted file mode 100644 index cde69daacc..0000000000 --- a/src/mesa/pipe/i915simple/i915_clear.c +++ /dev/null @@ -1,47 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Brian Paul - */ - - -#include "pipe/p_defines.h" -#include "i915_context.h" -#include "i915_state.h" - - -/** - * Clear the given surface to the specified value. - * No masking, no scissor (clear entire buffer). - */ -void -i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue) -{ - pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); -} diff --git a/src/mesa/pipe/i915simple/i915_context.c b/src/mesa/pipe/i915simple/i915_context.c deleted file mode 100644 index 497623a700..0000000000 --- a/src/mesa/pipe/i915simple/i915_context.c +++ /dev/null @@ -1,320 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915_context.h" -#include "i915_winsys.h" -#include "i915_state.h" -#include "i915_batch.h" -#include "i915_texture.h" -#include "i915_reg.h" - -#include "pipe/draw/draw_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" - - -/** - * Query format support for creating a texture, drawing surface, etc. - * \param format the format to test - * \param type one of PIPE_TEXTURE, PIPE_SURFACE - */ -static boolean -i915_is_format_supported( struct pipe_context *pipe, - enum pipe_format format, uint type ) -{ - static const enum pipe_format tex_supported[] = { - PIPE_FORMAT_R8G8B8A8_UNORM, - PIPE_FORMAT_A8R8G8B8_UNORM, - PIPE_FORMAT_R5G6B5_UNORM, - PIPE_FORMAT_U_L8, - PIPE_FORMAT_U_A8, - PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_A8_L8, - PIPE_FORMAT_YCBCR, - PIPE_FORMAT_YCBCR_REV, - PIPE_FORMAT_S8Z24_UNORM, - PIPE_FORMAT_NONE /* list terminator */ - }; - static const enum pipe_format surface_supported[] = { - PIPE_FORMAT_A8R8G8B8_UNORM, - PIPE_FORMAT_R5G6B5_UNORM, - PIPE_FORMAT_S8Z24_UNORM, - /*PIPE_FORMAT_R16G16B16A16_SNORM,*/ - PIPE_FORMAT_NONE /* list terminator */ - }; - const enum pipe_format *list; - uint i; - - switch (type) { - case PIPE_TEXTURE: - list = tex_supported; - break; - case PIPE_SURFACE: - list = surface_supported; - break; - default: - assert(0); - } - - for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) { - if (list[i] == format) - return TRUE; - } - - return FALSE; -} - - -static int -i915_get_param(struct pipe_context *pipe, int param) -{ - switch (param) { - case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; - case PIPE_CAP_NPOT_TEXTURES: - return 1; - case PIPE_CAP_TWO_SIDED_STENCIL: - return 1; - case PIPE_CAP_GLSL: - return 0; - case PIPE_CAP_S3TC: - return 0; - case PIPE_CAP_ANISOTROPIC_FILTER: - return 0; - case PIPE_CAP_POINT_SPRITE: - return 0; - case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; - case PIPE_CAP_OCCLUSION_QUERY: - return 0; - case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 1; - case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 11; /* max 1024x1024 */ - case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: - return 8; /* max 128x128x128 */ - case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 11; /* max 1024x1024 */ - default: - return 0; - } -} - - -static float -i915_get_paramf(struct pipe_context *pipe, int param) -{ - switch (param) { - case PIPE_CAP_MAX_LINE_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_LINE_WIDTH_AA: - return 7.5; - - case PIPE_CAP_MAX_POINT_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_POINT_WIDTH_AA: - return 255.0; - - case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 4.0; - - case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: - return 16.0; - - default: - return 0; - } -} - - -static void i915_destroy( struct pipe_context *pipe ) -{ - struct i915_context *i915 = i915_context( pipe ); - - draw_destroy( i915->draw ); - - FREE( i915 ); -} - - - - -static boolean -i915_draw_elements( struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned prim, unsigned start, unsigned count) -{ - struct i915_context *i915 = i915_context( pipe ); - struct draw_context *draw = i915->draw; - unsigned i; - - if (i915->dirty) - i915_update_derived( i915 ); - - /* - * Map vertex buffers - */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { - if (i915->vertex_buffer[i].buffer) { - void *buf - = pipe->winsys->buffer_map(pipe->winsys, - i915->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_vertex_buffer(draw, i, buf); - } - } - /* Map index buffer, if present */ - if (indexBuffer) { - void *mapped_indexes - = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); - } - else { - /* no index/element buffer */ - draw_set_mapped_element_buffer(draw, 0, NULL); - } - - - draw_set_mapped_constant_buffer(draw, - i915->current.constants[PIPE_SHADER_VERTEX]); - - /* draw! */ - draw_arrays(i915->draw, prim, start, count); - - /* - * unmap vertex/index buffers - */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { - if (i915->vertex_buffer[i].buffer) { - pipe->winsys->buffer_unmap(pipe->winsys, i915->vertex_buffer[i].buffer); - draw_set_mapped_vertex_buffer(draw, i, NULL); - } - } - if (indexBuffer) { - pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); - draw_set_mapped_element_buffer(draw, 0, NULL); - } - - return TRUE; -} - - -static boolean i915_draw_arrays( struct pipe_context *pipe, - unsigned prim, unsigned start, unsigned count) -{ - return i915_draw_elements(pipe, NULL, 0, prim, start, count); -} - - - -struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, - struct i915_winsys *i915_winsys, - unsigned pci_id ) -{ - struct i915_context *i915; - unsigned is_i945 = 0; - - switch (pci_id) { - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - break; - - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - is_i945 = 1; - break; - - default: - pipe_winsys->printf(pipe_winsys, - "%s: unknown pci id 0x%x, cannot create context\n", - __FUNCTION__, pci_id); - return NULL; - } - - i915 = CALLOC_STRUCT(i915_context); - if (i915 == NULL) - return NULL; - - i915->winsys = i915_winsys; - i915->pipe.winsys = pipe_winsys; - - i915->pipe.destroy = i915_destroy; - i915->pipe.is_format_supported = i915_is_format_supported; - i915->pipe.get_param = i915_get_param; - i915->pipe.get_paramf = i915_get_paramf; - - i915->pipe.clear = i915_clear; - - - i915->pipe.draw_arrays = i915_draw_arrays; - i915->pipe.draw_elements = i915_draw_elements; - - /* - * Create drawing context and plug our rendering stage into it. - */ - i915->draw = draw_create(); - assert(i915->draw); - if (GETENV("I915_VBUF")) { - draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915)); - } - else { - draw_set_rasterize_stage(i915->draw, i915_draw_render_stage(i915)); - } - - i915_init_surface_functions(i915); - i915_init_state_functions(i915); - i915_init_flush_functions(i915); - i915_init_string_functions(i915); - - i915->pci_id = pci_id; - i915->flags.is_i945 = is_i945; - - i915->pipe.texture_create = i915_texture_create; - i915->pipe.texture_release = i915_texture_release; - - i915->dirty = ~0; - i915->hardware_dirty = ~0; - - /* Batch stream debugging is a bit hacked up at the moment: - */ - i915->batch_start = NULL; - - /* - * XXX we could plug GL selection/feedback into the drawing pipeline - * by specifying a different setup/render stage. - */ - - return &i915->pipe; -} - diff --git a/src/mesa/pipe/i915simple/i915_context.h b/src/mesa/pipe/i915simple/i915_context.h deleted file mode 100644 index b4ea63c3e7..0000000000 --- a/src/mesa/pipe/i915simple/i915_context.h +++ /dev/null @@ -1,304 +0,0 @@ - /************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef I915_CONTEXT_H -#define I915_CONTEXT_H - - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" - -#include "pipe/draw/draw_vertex.h" - - -#define I915_TEX_UNITS 8 - -#define I915_DYNAMIC_MODES4 0 -#define I915_DYNAMIC_DEPTHSCALE_0 1 /* just the header */ -#define I915_DYNAMIC_DEPTHSCALE_1 2 -#define I915_DYNAMIC_IAB 3 -#define I915_DYNAMIC_BC_0 4 /* just the header */ -#define I915_DYNAMIC_BC_1 5 -#define I915_DYNAMIC_BFO_0 6 -#define I915_DYNAMIC_BFO_1 7 -#define I915_DYNAMIC_STP_0 8 -#define I915_DYNAMIC_STP_1 9 -#define I915_DYNAMIC_SC_ENA_0 10 -#define I915_DYNAMIC_SC_RECT_0 11 -#define I915_DYNAMIC_SC_RECT_1 12 -#define I915_DYNAMIC_SC_RECT_2 13 -#define I915_MAX_DYNAMIC 14 - - -#define I915_IMMEDIATE_S0 0 -#define I915_IMMEDIATE_S1 1 -#define I915_IMMEDIATE_S2 2 -#define I915_IMMEDIATE_S3 3 -#define I915_IMMEDIATE_S4 4 -#define I915_IMMEDIATE_S5 5 -#define I915_IMMEDIATE_S6 6 -#define I915_IMMEDIATE_S7 7 -#define I915_MAX_IMMEDIATE 8 - -/* These must mach the order of LI0_STATE_* bits, as they will be used - * to generate hardware packets: - */ -#define I915_CACHE_STATIC 0 -#define I915_CACHE_DYNAMIC 1 /* handled specially */ -#define I915_CACHE_SAMPLER 2 -#define I915_CACHE_MAP 3 -#define I915_CACHE_PROGRAM 4 -#define I915_CACHE_CONSTANTS 5 -#define I915_MAX_CACHE 6 - -#define I915_MAX_CONSTANT 32 - - - -struct i915_cache_context; - -/* Use to calculate differences between state emitted to hardware and - * current driver-calculated state. - */ -struct i915_state -{ - unsigned immediate[I915_MAX_IMMEDIATE]; - unsigned dynamic[I915_MAX_DYNAMIC]; - - float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; - /** number of constants passed in through a constant buffer */ - uint num_user_constants[PIPE_SHADER_TYPES]; - /** user constants, plus extra constants from shader translation */ - uint num_constants[PIPE_SHADER_TYPES]; - - uint *program; - uint program_len; - - /* texture sampler state */ - unsigned sampler[I915_TEX_UNITS][3]; - unsigned sampler_enable_flags; - unsigned sampler_enable_nr; - - /* texture image buffers */ - unsigned texbuffer[I915_TEX_UNITS][2]; - - /** Describes the current hardware vertex layout */ - struct vertex_info vertex_info; - - unsigned id; /* track lost context events */ -}; - -struct i915_blend_state { - unsigned iab; - unsigned modes4; - unsigned LIS5; - unsigned LIS6; -}; - -struct i915_depth_stencil_state { - unsigned stencil_modes4; - unsigned bfo[2]; - unsigned stencil_LIS5; - unsigned depth_LIS6; -}; - -struct i915_rasterizer_state { - int light_twoside : 1; - unsigned st; - enum interp_mode color_interp; - - unsigned LIS4; - unsigned LIS7; - unsigned sc[1]; - - const struct pipe_rasterizer_state *templ; - - union { float f; unsigned u; } ds[2]; -}; - -struct i915_sampler_state { - unsigned state[3]; - const struct pipe_sampler_state *templ; -}; - - -struct i915_texture { - struct pipe_texture base; - - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ - unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; -}; - -struct i915_context -{ - struct pipe_context pipe; - struct i915_winsys *winsys; - struct draw_context *draw; - - /* The most recent drawing state as set by the driver: - */ - const struct i915_blend_state *blend; - const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct i915_depth_stencil_state *depth_stencil; - const struct i915_rasterizer_state *rasterizer; - const struct pipe_shader_state *fs; - - struct pipe_blend_color blend_color; - struct pipe_clip_state clip; - struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; - struct pipe_framebuffer_state framebuffer; - struct pipe_poly_stipple poly_stipple; - struct pipe_scissor_state scissor; - struct i915_texture *texture[PIPE_MAX_SAMPLERS]; - struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - - unsigned dirty; - - unsigned *batch_start; - - /** Vertex buffer */ - struct pipe_buffer *vbo; - - struct i915_state current; - unsigned hardware_dirty; - - unsigned debug; - unsigned pci_id; - - struct { - unsigned is_i945:1; - } flags; -}; - -/* A flag for each state_tracker state object: - */ -#define I915_NEW_VIEWPORT 0x1 -#define I915_NEW_RASTERIZER 0x2 -#define I915_NEW_FS 0x4 -#define I915_NEW_BLEND 0x8 -#define I915_NEW_CLIP 0x10 -#define I915_NEW_SCISSOR 0x20 -#define I915_NEW_STIPPLE 0x40 -#define I915_NEW_FRAMEBUFFER 0x80 -#define I915_NEW_ALPHA_TEST 0x100 -#define I915_NEW_DEPTH_STENCIL 0x200 -#define I915_NEW_SAMPLER 0x400 -#define I915_NEW_TEXTURE 0x800 -#define I915_NEW_CONSTANTS 0x1000 -#define I915_NEW_VBO 0x2000 - - -/* Driver's internally generated state flags: - */ -#define I915_NEW_VERTEX_FORMAT 0x10000 - - -/* Dirty flags for hardware emit - */ -#define I915_HW_STATIC (1<winsys->printf( stream->winsys, buffer ); - va_end( args ); -} - - -static boolean debug( struct debug_stream *stream, const char *name, unsigned len ) -{ - unsigned i; - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - - if (len == 0) { - PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]); - assert(0); - return FALSE; - } - - if (stream->print_addresses) - PRINTF(stream, "%08x: ", stream->offset); - - - PRINTF(stream, "%s (%d dwords):\n", name, len); - for (i = 0; i < len; i++) - PRINTF(stream, "\t0x%08x\n", ptr[i]); - PRINTF(stream, "\n"); - - stream->offset += len * sizeof(unsigned); - - return TRUE; -} - - -static const char *get_prim_name( unsigned val ) -{ - switch (val & PRIM3D_MASK) { - case PRIM3D_TRILIST: return "TRILIST"; break; - case PRIM3D_TRISTRIP: return "TRISTRIP"; break; - case PRIM3D_TRISTRIP_RVRSE: return "TRISTRIP_RVRSE"; break; - case PRIM3D_TRIFAN: return "TRIFAN"; break; - case PRIM3D_POLY: return "POLY"; break; - case PRIM3D_LINELIST: return "LINELIST"; break; - case PRIM3D_LINESTRIP: return "LINESTRIP"; break; - case PRIM3D_RECTLIST: return "RECTLIST"; break; - case PRIM3D_POINTLIST: return "POINTLIST"; break; - case PRIM3D_DIB: return "DIB"; break; - case PRIM3D_CLEAR_RECT: return "CLEAR_RECT"; break; - case PRIM3D_ZONE_INIT: return "ZONE_INIT"; break; - default: return "????"; break; - } -} - -static boolean debug_prim( struct debug_stream *stream, const char *name, - boolean dump_floats, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - const char *prim = get_prim_name( ptr[0] ); - unsigned i; - - - - PRINTF(stream, "%s %s (%d dwords):\n", name, prim, len); - PRINTF(stream, "\t0x%08x\n", ptr[0]); - for (i = 1; i < len; i++) { - if (dump_floats) - PRINTF(stream, "\t0x%08x // %f\n", ptr[i], *(float *)&ptr[i]); - else - PRINTF(stream, "\t0x%08x\n", ptr[i]); - } - - - PRINTF(stream, "\n"); - - stream->offset += len * sizeof(unsigned); - - return TRUE; -} - - - - -static boolean debug_program( struct debug_stream *stream, const char *name, unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - - if (len == 0) { - PRINTF(stream, "Error - zero length packet (0x%08x)\n", stream->ptr[0]); - assert(0); - return FALSE; - } - - if (stream->print_addresses) - PRINTF(stream, "%08x: ", stream->offset); - - PRINTF(stream, "%s (%d dwords):\n", name, len); - i915_disassemble_program( stream, ptr, len ); - - stream->offset += len * sizeof(unsigned); - return TRUE; -} - - -static boolean debug_chain( struct debug_stream *stream, const char *name, unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - unsigned old_offset = stream->offset + len * sizeof(unsigned); - unsigned i; - - PRINTF(stream, "%s (%d dwords):\n", name, len); - for (i = 0; i < len; i++) - PRINTF(stream, "\t0x%08x\n", ptr[i]); - - stream->offset = ptr[1] & ~0x3; - - if (stream->offset < old_offset) - PRINTF(stream, "\n... skipping backwards from 0x%x --> 0x%x ...\n\n", - old_offset, stream->offset ); - else - PRINTF(stream, "\n... skipping from 0x%x --> 0x%x ...\n\n", - old_offset, stream->offset ); - - - return TRUE; -} - - -static boolean debug_variable_length_prim( struct debug_stream *stream ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - const char *prim = get_prim_name( ptr[0] ); - unsigned i, len; - - ushort *idx = (ushort *)(ptr+1); - for (i = 0; idx[i] != 0xffff; i++) - ; - - len = 1+(i+2)/2; - - PRINTF(stream, "3DPRIM, %s variable length %d indicies (%d dwords):\n", prim, i, len); - for (i = 0; i < len; i++) - PRINTF(stream, "\t0x%08x\n", ptr[i]); - PRINTF(stream, "\n"); - - stream->offset += len * sizeof(unsigned); - return TRUE; -} - - -static void -BITS( - struct debug_stream *stream, - unsigned dw, - unsigned hi, - unsigned lo, - const char *fmt, - ... ) -{ - va_list args; - char buffer[256]; - unsigned himask = ~0UL >> (31 - (hi)); - - PRINTF(stream, "\t\t "); - - va_start( args, fmt ); - vsprintf( buffer, fmt, args ); - stream->winsys->printf( stream->winsys, buffer ); - va_end( args ); - - PRINTF(stream, ": 0x%x\n", ((dw) & himask) >> (lo)); -} - -#define MBZ( dw, hi, lo) do { \ - unsigned x = (dw) >> (lo); \ - unsigned lomask = (1 << (lo)) - 1; \ - unsigned himask; \ - himask = (1UL << (hi)) - 1; \ - assert ((x & himask & ~lomask) == 0); \ -} while (0) - -static void -FLAG( - struct debug_stream *stream, - unsigned dw, - unsigned bit, - const char *fmt, - ... ) -{ - if (((dw) >> (bit)) & 1) { - va_list args; - char buffer[256]; - - PRINTF(stream, "\t\t "); - - va_start( args, fmt ); - vsprintf( buffer, fmt, args ); - stream->winsys->printf( stream->winsys, buffer ); - va_end( args ); - - PRINTF(stream, "\n"); - } -} - -static boolean debug_load_immediate( struct debug_stream *stream, - const char *name, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - unsigned bits = (ptr[0] >> 4) & 0xff; - unsigned j = 0; - - PRINTF(stream, "%s (%d dwords, flags: %x):\n", name, len, bits); - PRINTF(stream, "\t0x%08x\n", ptr[j++]); - - if (bits & (1<<0)) { - PRINTF(stream, "\t LIS0: 0x%08x\n", ptr[j]); - PRINTF(stream, "\t vb address: 0x%08x\n", (ptr[j] & ~0x3)); - BITS(stream, ptr[j], 0, 0, "vb invalidate disable"); - j++; - } - if (bits & (1<<1)) { - PRINTF(stream, "\t LIS1: 0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 29, 24, "vb dword width"); - BITS(stream, ptr[j], 21, 16, "vb dword pitch"); - BITS(stream, ptr[j], 15, 0, "vb max index"); - j++; - } - if (bits & (1<<2)) { - int i; - PRINTF(stream, "\t LIS2: 0x%08x\n", ptr[j]); - for (i = 0; i < 8; i++) { - unsigned tc = (ptr[j] >> (i * 4)) & 0xf; - if (tc != 0xf) - BITS(stream, tc, 3, 0, "tex coord %d", i); - } - j++; - } - if (bits & (1<<3)) { - PRINTF(stream, "\t LIS3: 0x%08x\n", ptr[j]); - j++; - } - if (bits & (1<<4)) { - PRINTF(stream, "\t LIS4: 0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 31, 23, "point width"); - BITS(stream, ptr[j], 22, 19, "line width"); - FLAG(stream, ptr[j], 18, "alpha flatshade"); - FLAG(stream, ptr[j], 17, "fog flatshade"); - FLAG(stream, ptr[j], 16, "spec flatshade"); - FLAG(stream, ptr[j], 15, "rgb flatshade"); - BITS(stream, ptr[j], 14, 13, "cull mode"); - FLAG(stream, ptr[j], 12, "vfmt: point width"); - FLAG(stream, ptr[j], 11, "vfmt: specular/fog"); - FLAG(stream, ptr[j], 10, "vfmt: rgba"); - FLAG(stream, ptr[j], 9, "vfmt: depth offset"); - BITS(stream, ptr[j], 8, 6, "vfmt: position (2==xyzw)"); - FLAG(stream, ptr[j], 5, "force dflt diffuse"); - FLAG(stream, ptr[j], 4, "force dflt specular"); - FLAG(stream, ptr[j], 3, "local depth offset enable"); - FLAG(stream, ptr[j], 2, "vfmt: fp32 fog coord"); - FLAG(stream, ptr[j], 1, "sprite point"); - FLAG(stream, ptr[j], 0, "antialiasing"); - j++; - } - if (bits & (1<<5)) { - PRINTF(stream, "\t LIS5: 0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 31, 28, "rgba write disables"); - FLAG(stream, ptr[j], 27, "force dflt point width"); - FLAG(stream, ptr[j], 26, "last pixel enable"); - FLAG(stream, ptr[j], 25, "global z offset enable"); - FLAG(stream, ptr[j], 24, "fog enable"); - BITS(stream, ptr[j], 23, 16, "stencil ref"); - BITS(stream, ptr[j], 15, 13, "stencil test"); - BITS(stream, ptr[j], 12, 10, "stencil fail op"); - BITS(stream, ptr[j], 9, 7, "stencil pass z fail op"); - BITS(stream, ptr[j], 6, 4, "stencil pass z pass op"); - FLAG(stream, ptr[j], 3, "stencil write enable"); - FLAG(stream, ptr[j], 2, "stencil test enable"); - FLAG(stream, ptr[j], 1, "color dither enable"); - FLAG(stream, ptr[j], 0, "logiop enable"); - j++; - } - if (bits & (1<<6)) { - PRINTF(stream, "\t LIS6: 0x%08x\n", ptr[j]); - FLAG(stream, ptr[j], 31, "alpha test enable"); - BITS(stream, ptr[j], 30, 28, "alpha func"); - BITS(stream, ptr[j], 27, 20, "alpha ref"); - FLAG(stream, ptr[j], 19, "depth test enable"); - BITS(stream, ptr[j], 18, 16, "depth func"); - FLAG(stream, ptr[j], 15, "blend enable"); - BITS(stream, ptr[j], 14, 12, "blend func"); - BITS(stream, ptr[j], 11, 8, "blend src factor"); - BITS(stream, ptr[j], 7, 4, "blend dst factor"); - FLAG(stream, ptr[j], 3, "depth write enable"); - FLAG(stream, ptr[j], 2, "color write enable"); - BITS(stream, ptr[j], 1, 0, "provoking vertex"); - j++; - } - - - PRINTF(stream, "\n"); - - assert(j == len); - - stream->offset += len * sizeof(unsigned); - - return TRUE; -} - - - -static boolean debug_load_indirect( struct debug_stream *stream, - const char *name, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - unsigned bits = (ptr[0] >> 8) & 0x3f; - unsigned i, j = 0; - - PRINTF(stream, "%s (%d dwords):\n", name, len); - PRINTF(stream, "\t0x%08x\n", ptr[j++]); - - for (i = 0; i < 6; i++) { - if (bits & (1<offset += len * sizeof(unsigned); - - return TRUE; -} - -static void BR13( struct debug_stream *stream, - unsigned val ) -{ - PRINTF(stream, "\t0x%08x\n", val); - FLAG(stream, val, 30, "clipping enable"); - BITS(stream, val, 25, 24, "color depth (3==32bpp)"); - BITS(stream, val, 23, 16, "raster op"); - BITS(stream, val, 15, 0, "dest pitch"); -} - - -static void BR22( struct debug_stream *stream, - unsigned val ) -{ - PRINTF(stream, "\t0x%08x\n", val); - BITS(stream, val, 31, 16, "dest y1"); - BITS(stream, val, 15, 0, "dest x1"); -} - -static void BR23( struct debug_stream *stream, - unsigned val ) -{ - PRINTF(stream, "\t0x%08x\n", val); - BITS(stream, val, 31, 16, "dest y2"); - BITS(stream, val, 15, 0, "dest x2"); -} - -static void BR09( struct debug_stream *stream, - unsigned val ) -{ - PRINTF(stream, "\t0x%08x -- dest address\n", val); -} - -static void BR26( struct debug_stream *stream, - unsigned val ) -{ - PRINTF(stream, "\t0x%08x\n", val); - BITS(stream, val, 31, 16, "src y1"); - BITS(stream, val, 15, 0, "src x1"); -} - -static void BR11( struct debug_stream *stream, - unsigned val ) -{ - PRINTF(stream, "\t0x%08x\n", val); - BITS(stream, val, 15, 0, "src pitch"); -} - -static void BR12( struct debug_stream *stream, - unsigned val ) -{ - PRINTF(stream, "\t0x%08x -- src address\n", val); -} - -static void BR16( struct debug_stream *stream, - unsigned val ) -{ - PRINTF(stream, "\t0x%08x -- color\n", val); -} - -static boolean debug_copy_blit( struct debug_stream *stream, - const char *name, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - int j = 0; - - PRINTF(stream, "%s (%d dwords):\n", name, len); - PRINTF(stream, "\t0x%08x\n", ptr[j++]); - - BR13(stream, ptr[j++]); - BR22(stream, ptr[j++]); - BR23(stream, ptr[j++]); - BR09(stream, ptr[j++]); - BR26(stream, ptr[j++]); - BR11(stream, ptr[j++]); - BR12(stream, ptr[j++]); - - stream->offset += len * sizeof(unsigned); - assert(j == len); - return TRUE; -} - -static boolean debug_color_blit( struct debug_stream *stream, - const char *name, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - int j = 0; - - PRINTF(stream, "%s (%d dwords):\n", name, len); - PRINTF(stream, "\t0x%08x\n", ptr[j++]); - - BR13(stream, ptr[j++]); - BR22(stream, ptr[j++]); - BR23(stream, ptr[j++]); - BR09(stream, ptr[j++]); - BR16(stream, ptr[j++]); - - stream->offset += len * sizeof(unsigned); - assert(j == len); - return TRUE; -} - -static boolean debug_modes4( struct debug_stream *stream, - const char *name, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - int j = 0; - - PRINTF(stream, "%s (%d dwords):\n", name, len); - PRINTF(stream, "\t0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 21, 18, "logicop func"); - FLAG(stream, ptr[j], 17, "stencil test mask modify-enable"); - FLAG(stream, ptr[j], 16, "stencil write mask modify-enable"); - BITS(stream, ptr[j], 15, 8, "stencil test mask"); - BITS(stream, ptr[j], 7, 0, "stencil write mask"); - j++; - - stream->offset += len * sizeof(unsigned); - assert(j == len); - return TRUE; -} - -static boolean debug_map_state( struct debug_stream *stream, - const char *name, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - unsigned j = 0; - - PRINTF(stream, "%s (%d dwords):\n", name, len); - PRINTF(stream, "\t0x%08x\n", ptr[j++]); - - { - PRINTF(stream, "\t0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 15, 0, "map mask"); - j++; - } - - while (j < len) { - { - PRINTF(stream, "\t TMn.0: 0x%08x\n", ptr[j]); - PRINTF(stream, "\t map address: 0x%08x\n", (ptr[j] & ~0x3)); - FLAG(stream, ptr[j], 1, "vertical line stride"); - FLAG(stream, ptr[j], 0, "vertical line stride offset"); - j++; - } - - { - PRINTF(stream, "\t TMn.1: 0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 31, 21, "height"); - BITS(stream, ptr[j], 20, 10, "width"); - BITS(stream, ptr[j], 9, 7, "surface format"); - BITS(stream, ptr[j], 6, 3, "texel format"); - FLAG(stream, ptr[j], 2, "use fence regs"); - FLAG(stream, ptr[j], 1, "tiled surface"); - FLAG(stream, ptr[j], 0, "tile walk ymajor"); - j++; - } - { - PRINTF(stream, "\t TMn.2: 0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 31, 21, "dword pitch"); - BITS(stream, ptr[j], 20, 15, "cube face enables"); - BITS(stream, ptr[j], 14, 9, "max lod"); - FLAG(stream, ptr[j], 8, "mip layout right"); - BITS(stream, ptr[j], 7, 0, "depth"); - j++; - } - } - - stream->offset += len * sizeof(unsigned); - assert(j == len); - return TRUE; -} - -static boolean debug_sampler_state( struct debug_stream *stream, - const char *name, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - unsigned j = 0; - - PRINTF(stream, "%s (%d dwords):\n", name, len); - PRINTF(stream, "\t0x%08x\n", ptr[j++]); - - { - PRINTF(stream, "\t0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 15, 0, "sampler mask"); - j++; - } - - while (j < len) { - { - PRINTF(stream, "\t TSn.0: 0x%08x\n", ptr[j]); - FLAG(stream, ptr[j], 31, "reverse gamma"); - FLAG(stream, ptr[j], 30, "planar to packed"); - FLAG(stream, ptr[j], 29, "yuv->rgb"); - BITS(stream, ptr[j], 28, 27, "chromakey index"); - BITS(stream, ptr[j], 26, 22, "base mip level"); - BITS(stream, ptr[j], 21, 20, "mip mode filter"); - BITS(stream, ptr[j], 19, 17, "mag mode filter"); - BITS(stream, ptr[j], 16, 14, "min mode filter"); - BITS(stream, ptr[j], 13, 5, "lod bias (s4.4)"); - FLAG(stream, ptr[j], 4, "shadow enable"); - FLAG(stream, ptr[j], 3, "max-aniso-4"); - BITS(stream, ptr[j], 2, 0, "shadow func"); - j++; - } - - { - PRINTF(stream, "\t TSn.1: 0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 31, 24, "min lod"); - MBZ( ptr[j], 23, 18 ); - FLAG(stream, ptr[j], 17, "kill pixel enable"); - FLAG(stream, ptr[j], 16, "keyed tex filter mode"); - FLAG(stream, ptr[j], 15, "chromakey enable"); - BITS(stream, ptr[j], 14, 12, "tcx wrap mode"); - BITS(stream, ptr[j], 11, 9, "tcy wrap mode"); - BITS(stream, ptr[j], 8, 6, "tcz wrap mode"); - FLAG(stream, ptr[j], 5, "normalized coords"); - BITS(stream, ptr[j], 4, 1, "map (surface) index"); - FLAG(stream, ptr[j], 0, "EAST deinterlacer enable"); - j++; - } - { - PRINTF(stream, "\t TSn.2: 0x%08x (default color)\n", ptr[j]); - j++; - } - } - - stream->offset += len * sizeof(unsigned); - assert(j == len); - return TRUE; -} - -static boolean debug_dest_vars( struct debug_stream *stream, - const char *name, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - int j = 0; - - PRINTF(stream, "%s (%d dwords):\n", name, len); - PRINTF(stream, "\t0x%08x\n", ptr[j++]); - - { - PRINTF(stream, "\t0x%08x\n", ptr[j]); - FLAG(stream, ptr[j], 31, "early classic ztest"); - FLAG(stream, ptr[j], 30, "opengl tex default color"); - FLAG(stream, ptr[j], 29, "bypass iz"); - FLAG(stream, ptr[j], 28, "lod preclamp"); - BITS(stream, ptr[j], 27, 26, "dither pattern"); - FLAG(stream, ptr[j], 25, "linear gamma blend"); - FLAG(stream, ptr[j], 24, "debug dither"); - BITS(stream, ptr[j], 23, 20, "dstorg x"); - BITS(stream, ptr[j], 19, 16, "dstorg y"); - MBZ (ptr[j], 15, 15 ); - BITS(stream, ptr[j], 14, 12, "422 write select"); - BITS(stream, ptr[j], 11, 8, "cbuf format"); - BITS(stream, ptr[j], 3, 2, "zbuf format"); - FLAG(stream, ptr[j], 1, "vert line stride"); - FLAG(stream, ptr[j], 1, "vert line stride offset"); - j++; - } - - stream->offset += len * sizeof(unsigned); - assert(j == len); - return TRUE; -} - -static boolean debug_buf_info( struct debug_stream *stream, - const char *name, - unsigned len ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - int j = 0; - - PRINTF(stream, "%s (%d dwords):\n", name, len); - PRINTF(stream, "\t0x%08x\n", ptr[j++]); - - { - PRINTF(stream, "\t0x%08x\n", ptr[j]); - BITS(stream, ptr[j], 28, 28, "aux buffer id"); - BITS(stream, ptr[j], 27, 24, "buffer id (7=depth, 3=back)"); - FLAG(stream, ptr[j], 23, "use fence regs"); - FLAG(stream, ptr[j], 22, "tiled surface"); - FLAG(stream, ptr[j], 21, "tile walk ymajor"); - MBZ (ptr[j], 20, 14); - BITS(stream, ptr[j], 13, 2, "dword pitch"); - MBZ (ptr[j], 2, 0); - j++; - } - - PRINTF(stream, "\t0x%08x -- buffer base address\n", ptr[j++]); - - stream->offset += len * sizeof(unsigned); - assert(j == len); - return TRUE; -} - -static boolean i915_debug_packet( struct debug_stream *stream ) -{ - unsigned *ptr = (unsigned *)(stream->ptr + stream->offset); - unsigned cmd = *ptr; - - switch (((cmd >> 29) & 0x7)) { - case 0x0: - switch ((cmd >> 23) & 0x3f) { - case 0x0: - return debug(stream, "MI_NOOP", 1); - case 0x3: - return debug(stream, "MI_WAIT_FOR_EVENT", 1); - case 0x4: - return debug(stream, "MI_FLUSH", 1); - case 0xA: - debug(stream, "MI_BATCH_BUFFER_END", 1); - return FALSE; - case 0x22: - return debug(stream, "MI_LOAD_REGISTER_IMM", 3); - case 0x31: - return debug_chain(stream, "MI_BATCH_BUFFER_START", 2); - default: - (void)debug(stream, "UNKNOWN 0x0 case!", 1); - assert(0); - break; - } - break; - case 0x1: - (void) debug(stream, "UNKNOWN 0x1 case!", 1); - assert(0); - break; - case 0x2: - switch ((cmd >> 22) & 0xff) { - case 0x50: - return debug_color_blit(stream, "XY_COLOR_BLT", (cmd & 0xff) + 2); - case 0x53: - return debug_copy_blit(stream, "XY_SRC_COPY_BLT", (cmd & 0xff) + 2); - default: - return debug(stream, "blit command", (cmd & 0xff) + 2); - } - break; - case 0x3: - switch ((cmd >> 24) & 0x1f) { - case 0x6: - return debug(stream, "3DSTATE_ANTI_ALIASING", 1); - case 0x7: - return debug(stream, "3DSTATE_RASTERIZATION_RULES", 1); - case 0x8: - return debug(stream, "3DSTATE_BACKFACE_STENCIL_OPS", 2); - case 0x9: - return debug(stream, "3DSTATE_BACKFACE_STENCIL_MASKS", 1); - case 0xb: - return debug(stream, "3DSTATE_INDEPENDENT_ALPHA_BLEND", 1); - case 0xc: - return debug(stream, "3DSTATE_MODES5", 1); - case 0xd: - return debug_modes4(stream, "3DSTATE_MODES4", 1); - case 0x15: - return debug(stream, "3DSTATE_FOG_COLOR", 1); - case 0x16: - return debug(stream, "3DSTATE_COORD_SET_BINDINGS", 1); - case 0x1c: - /* 3DState16NP */ - switch((cmd >> 19) & 0x1f) { - case 0x10: - return debug(stream, "3DSTATE_SCISSOR_ENABLE", 1); - case 0x11: - return debug(stream, "3DSTATE_DEPTH_SUBRECTANGLE_DISABLE", 1); - default: - (void) debug(stream, "UNKNOWN 0x1c case!", 1); - assert(0); - break; - } - break; - case 0x1d: - /* 3DStateMW */ - switch ((cmd >> 16) & 0xff) { - case 0x0: - return debug_map_state(stream, "3DSTATE_MAP_STATE", (cmd & 0x1f) + 2); - case 0x1: - return debug_sampler_state(stream, "3DSTATE_SAMPLER_STATE", (cmd & 0x1f) + 2); - case 0x4: - return debug_load_immediate(stream, "3DSTATE_LOAD_STATE_IMMEDIATE", (cmd & 0xf) + 2); - case 0x5: - return debug_program(stream, "3DSTATE_PIXEL_SHADER_PROGRAM", (cmd & 0x1ff) + 2); - case 0x6: - return debug(stream, "3DSTATE_PIXEL_SHADER_CONSTANTS", (cmd & 0xff) + 2); - case 0x7: - return debug_load_indirect(stream, "3DSTATE_LOAD_INDIRECT", (cmd & 0xff) + 2); - case 0x80: - return debug(stream, "3DSTATE_DRAWING_RECTANGLE", (cmd & 0xffff) + 2); - case 0x81: - return debug(stream, "3DSTATE_SCISSOR_RECTANGLE", (cmd & 0xffff) + 2); - case 0x83: - return debug(stream, "3DSTATE_SPAN_STIPPLE", (cmd & 0xffff) + 2); - case 0x85: - return debug_dest_vars(stream, "3DSTATE_DEST_BUFFER_VARS", (cmd & 0xffff) + 2); - case 0x88: - return debug(stream, "3DSTATE_CONSTANT_BLEND_COLOR", (cmd & 0xffff) + 2); - case 0x89: - return debug(stream, "3DSTATE_FOG_MODE", (cmd & 0xffff) + 2); - case 0x8e: - return debug_buf_info(stream, "3DSTATE_BUFFER_INFO", (cmd & 0xffff) + 2); - case 0x97: - return debug(stream, "3DSTATE_DEPTH_OFFSET_SCALE", (cmd & 0xffff) + 2); - case 0x98: - return debug(stream, "3DSTATE_DEFAULT_Z", (cmd & 0xffff) + 2); - case 0x99: - return debug(stream, "3DSTATE_DEFAULT_DIFFUSE", (cmd & 0xffff) + 2); - case 0x9a: - return debug(stream, "3DSTATE_DEFAULT_SPECULAR", (cmd & 0xffff) + 2); - case 0x9c: - return debug(stream, "3DSTATE_CLEAR_PARAMETERS", (cmd & 0xffff) + 2); - default: - assert(0); - return 0; - } - break; - case 0x1e: - if (cmd & (1 << 23)) - return debug(stream, "???", (cmd & 0xffff) + 1); - else - return debug(stream, "", 1); - break; - case 0x1f: - if ((cmd & (1 << 23)) == 0) - return debug_prim(stream, "3DPRIM (inline)", 1, (cmd & 0x1ffff) + 2); - else if (cmd & (1 << 17)) - { - if ((cmd & 0xffff) == 0) - return debug_variable_length_prim(stream); - else - return debug_prim(stream, "3DPRIM (indexed)", 0, (((cmd & 0xffff) + 1) / 2) + 1); - } - else - return debug_prim(stream, "3DPRIM (indirect sequential)", 0, 2); - break; - default: - return debug(stream, "", 0); - } - default: - assert(0); - return 0; - } - - assert(0); - return 0; -} - - - -void -i915_dump_batchbuffer( struct i915_context *i915 ) -{ - struct debug_stream stream; - unsigned *start = i915->batch_start; - unsigned *end = i915->winsys->batch_start( i915->winsys, 0, 0 ); - unsigned long bytes = (unsigned long) (end - start) * 4; - boolean done = FALSE; - - stream.offset = 0; - stream.ptr = (char *)start; - stream.print_addresses = 0; - stream.winsys = i915->pipe.winsys; - - if (!start || !end) { - stream.winsys->printf( stream.winsys, "\n\nBATCH: ???\n"); - return; - } - - stream.winsys->printf( stream.winsys, "\n\nBATCH: (%d)\n", bytes / 4); - - while (!done && - stream.offset < bytes) - { - if (!i915_debug_packet( &stream )) - break; - - assert(stream.offset <= bytes && - stream.offset >= 0); - } - - stream.winsys->printf( stream.winsys, "END-BATCH\n\n\n"); -} - - diff --git a/src/mesa/pipe/i915simple/i915_debug.h b/src/mesa/pipe/i915simple/i915_debug.h deleted file mode 100644 index 0bcd094233..0000000000 --- a/src/mesa/pipe/i915simple/i915_debug.h +++ /dev/null @@ -1,117 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef I915_DEBUG_H -#define I915_DEBUG_H - -#include - -struct i915_context; - -struct debug_stream -{ - unsigned offset; /* current gtt offset */ - char *ptr; /* pointer to gtt offset zero */ - char *end; /* pointer to gtt offset zero */ - unsigned print_addresses; - struct pipe_winsys *winsys; -}; - - -/* Internal functions - */ -void i915_disassemble_program(struct debug_stream *stream, - const unsigned *program, unsigned sz); - -void i915_print_ureg(const char *msg, unsigned ureg); - - -#define DEBUG_BATCH 0x1 -#define DEBUG_BLIT 0x2 -#define DEBUG_BUFFER 0x4 -#define DEBUG_CONSTANTS 0x8 -#define DEBUG_CONTEXT 0x10 -#define DEBUG_DRAW 0x20 -#define DEBUG_DYNAMIC 0x40 -#define DEBUG_FLUSH 0x80 -#define DEBUG_MAP 0x100 -#define DEBUG_PROGRAM 0x200 -#define DEBUG_REGIONS 0x400 -#define DEBUG_SAMPLER 0x800 -#define DEBUG_STATIC 0x1000 -#define DEBUG_SURFACE 0x2000 -#define DEBUG_WINSYS 0x4000 - -#include "pipe/p_compiler.h" - -#if defined(DEBUG) && defined(FILE_DEBUG_FLAG) - -#include "pipe/p_winsys.h" - -static INLINE void -I915_DBG( - struct i915_context *i915, - const char *fmt, - ... ) -{ - if ((i915)->debug & FILE_DEBUG_FLAG) { - va_list args; - char buffer[256]; - - va_start( args, fmt ); - vsprintf( buffer, fmt, args ); - i915->pipe.winsys->printf( i915->pipe.winsys, buffer ); - va_end( args ); - } -} - -#else - -static INLINE void -I915_DBG( - struct i915_context *i915, - const char *fmt, - ... ) -{ - (void) i915; - (void) fmt; -} - -#endif - - -void i915_dump_batchbuffer( struct i915_context *i915 ); - - - -void i915_debug_init( struct i915_context *i915 ); - - -#endif diff --git a/src/mesa/pipe/i915simple/i915_debug_fp.c b/src/mesa/pipe/i915simple/i915_debug_fp.c deleted file mode 100644 index ebfdb3d93c..0000000000 --- a/src/mesa/pipe/i915simple/i915_debug_fp.c +++ /dev/null @@ -1,366 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "i915_reg.h" -#include "i915_debug.h" -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" - - -static void -PRINTF( - struct debug_stream *stream, - const char *fmt, - ... ) -{ - va_list args; - char buffer[256]; - - va_start( args, fmt ); - vsprintf( buffer, fmt, args ); - stream->winsys->printf( stream->winsys, buffer ); - va_end( args ); -} - - -static const char *opcodes[0x20] = { - "NOP", - "ADD", - "MOV", - "MUL", - "MAD", - "DP2ADD", - "DP3", - "DP4", - "FRC", - "RCP", - "RSQ", - "EXP", - "LOG", - "CMP", - "MIN", - "MAX", - "FLR", - "MOD", - "TRC", - "SGE", - "SLT", - "TEXLD", - "TEXLDP", - "TEXLDB", - "TEXKILL", - "DCL", - "0x1a", - "0x1b", - "0x1c", - "0x1d", - "0x1e", - "0x1f", -}; - - -static const int args[0x20] = { - 0, /* 0 nop */ - 2, /* 1 add */ - 1, /* 2 mov */ - 2, /* 3 m ul */ - 3, /* 4 mad */ - 3, /* 5 dp2add */ - 2, /* 6 dp3 */ - 2, /* 7 dp4 */ - 1, /* 8 frc */ - 1, /* 9 rcp */ - 1, /* a rsq */ - 1, /* b exp */ - 1, /* c log */ - 3, /* d cmp */ - 2, /* e min */ - 2, /* f max */ - 1, /* 10 flr */ - 1, /* 11 mod */ - 1, /* 12 trc */ - 2, /* 13 sge */ - 2, /* 14 slt */ - 1, - 1, - 1, - 1, - 0, - 0, - 0, - 0, - 0, - 0, - 0, -}; - - -static const char *regname[0x8] = { - "R", - "T", - "CONST", - "S", - "OC", - "OD", - "U", - "UNKNOWN", -}; - -static void -print_reg_type_nr(struct debug_stream *stream, unsigned type, unsigned nr) -{ - switch (type) { - case REG_TYPE_T: - switch (nr) { - case T_DIFFUSE: - PRINTF(stream, "T_DIFFUSE"); - return; - case T_SPECULAR: - PRINTF(stream, "T_SPECULAR"); - return; - case T_FOG_W: - PRINTF(stream, "T_FOG_W"); - return; - default: - PRINTF(stream, "T_TEX%d", nr); - return; - } - case REG_TYPE_OC: - if (nr == 0) { - PRINTF(stream, "oC"); - return; - } - break; - case REG_TYPE_OD: - if (nr == 0) { - PRINTF(stream, "oD"); - return; - } - break; - default: - break; - } - - PRINTF(stream, "%s[%d]", regname[type], nr); -} - -#define REG_SWIZZLE_MASK 0x7777 -#define REG_NEGATE_MASK 0x8888 - -#define REG_SWIZZLE_XYZW ((SRC_X << A2_SRC2_CHANNEL_X_SHIFT) | \ - (SRC_Y << A2_SRC2_CHANNEL_Y_SHIFT) | \ - (SRC_Z << A2_SRC2_CHANNEL_Z_SHIFT) | \ - (SRC_W << A2_SRC2_CHANNEL_W_SHIFT)) - - -static void -print_reg_neg_swizzle(struct debug_stream *stream, unsigned reg) -{ - int i; - - if ((reg & REG_SWIZZLE_MASK) == REG_SWIZZLE_XYZW && - (reg & REG_NEGATE_MASK) == 0) - return; - - PRINTF(stream, "."); - - for (i = 3; i >= 0; i--) { - if (reg & (1 << ((i * 4) + 3))) - PRINTF(stream, "-"); - - switch ((reg >> (i * 4)) & 0x7) { - case 0: - PRINTF(stream, "x"); - break; - case 1: - PRINTF(stream, "y"); - break; - case 2: - PRINTF(stream, "z"); - break; - case 3: - PRINTF(stream, "w"); - break; - case 4: - PRINTF(stream, "0"); - break; - case 5: - PRINTF(stream, "1"); - break; - default: - PRINTF(stream, "?"); - break; - } - } -} - - -static void -print_src_reg(struct debug_stream *stream, unsigned dword) -{ - unsigned nr = (dword >> A2_SRC2_NR_SHIFT) & REG_NR_MASK; - unsigned type = (dword >> A2_SRC2_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(stream, type, nr); - print_reg_neg_swizzle(stream, dword); -} - - -static void -print_dest_reg(struct debug_stream *stream, unsigned dword) -{ - unsigned nr = (dword >> A0_DEST_NR_SHIFT) & REG_NR_MASK; - unsigned type = (dword >> A0_DEST_TYPE_SHIFT) & REG_TYPE_MASK; - print_reg_type_nr(stream, type, nr); - if ((dword & A0_DEST_CHANNEL_ALL) == A0_DEST_CHANNEL_ALL) - return; - PRINTF(stream, "."); - if (dword & A0_DEST_CHANNEL_X) - PRINTF(stream, "x"); - if (dword & A0_DEST_CHANNEL_Y) - PRINTF(stream, "y"); - if (dword & A0_DEST_CHANNEL_Z) - PRINTF(stream, "z"); - if (dword & A0_DEST_CHANNEL_W) - PRINTF(stream, "w"); -} - - -#define GET_SRC0_REG(r0, r1) ((r0<<14)|(r1>>A1_SRC0_CHANNEL_W_SHIFT)) -#define GET_SRC1_REG(r0, r1) ((r0<<8)|(r1>>A2_SRC1_CHANNEL_W_SHIFT)) -#define GET_SRC2_REG(r) (r) - - -static void -print_arith_op(struct debug_stream *stream, - unsigned opcode, const unsigned * program) -{ - if (opcode != A0_NOP) { - print_dest_reg(stream, program[0]); - if (program[0] & A0_DEST_SATURATE) - PRINTF(stream, " = SATURATE "); - else - PRINTF(stream, " = "); - } - - PRINTF(stream, "%s ", opcodes[opcode]); - - print_src_reg(stream, GET_SRC0_REG(program[0], program[1])); - if (args[opcode] == 1) { - PRINTF(stream, "\n"); - return; - } - - PRINTF(stream, ", "); - print_src_reg(stream, GET_SRC1_REG(program[1], program[2])); - if (args[opcode] == 2) { - PRINTF(stream, "\n"); - return; - } - - PRINTF(stream, ", "); - print_src_reg(stream, GET_SRC2_REG(program[2])); - PRINTF(stream, "\n"); - return; -} - - -static void -print_tex_op(struct debug_stream *stream, - unsigned opcode, const unsigned * program) -{ - print_dest_reg(stream, program[0] | A0_DEST_CHANNEL_ALL); - PRINTF(stream, " = "); - - PRINTF(stream, "%s ", opcodes[opcode]); - - PRINTF(stream, "S[%d],", program[0] & T0_SAMPLER_NR_MASK); - - print_reg_type_nr(stream, - (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & - REG_TYPE_MASK, - (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); - PRINTF(stream, "\n"); -} - -static void -print_texkil_op(struct debug_stream *stream, - unsigned opcode, const unsigned * program) -{ - PRINTF(stream, "TEXKIL "); - - print_reg_type_nr(stream, - (program[1] >> T1_ADDRESS_REG_TYPE_SHIFT) & - REG_TYPE_MASK, - (program[1] >> T1_ADDRESS_REG_NR_SHIFT) & REG_NR_MASK); - PRINTF(stream, "\n"); -} - -static void -print_dcl_op(struct debug_stream *stream, - unsigned opcode, const unsigned * program) -{ - PRINTF(stream, "%s ", opcodes[opcode]); - print_dest_reg(stream, - program[0] | A0_DEST_CHANNEL_ALL); - PRINTF(stream, "\n"); -} - - -void -i915_disassemble_program(struct debug_stream *stream, - const unsigned * program, unsigned sz) -{ - unsigned size = program[0] & 0x1ff; - unsigned i; - - PRINTF(stream, "\t\tBEGIN\n"); - - assert(size + 2 == sz); - - program++; - for (i = 1; i < sz; i += 3, program += 3) { - unsigned opcode = program[0] & (0x1f << 24); - - PRINTF(stream, "\t\t"); - - if ((int) opcode >= A0_NOP && opcode <= A0_SLT) - print_arith_op(stream, opcode >> 24, program); - else if (opcode >= T0_TEXLD && opcode < T0_TEXKILL) - print_tex_op(stream, opcode >> 24, program); - else if (opcode == T0_TEXKILL) - print_texkil_op(stream, opcode >> 24, program); - else if (opcode == D0_DCL) - print_dcl_op(stream, opcode >> 24, program); - else - PRINTF(stream, "Unknown opcode 0x%x\n", opcode); - } - - PRINTF(stream, "\t\tEND\n\n"); -} - - diff --git a/src/mesa/pipe/i915simple/i915_flush.c b/src/mesa/pipe/i915simple/i915_flush.c deleted file mode 100644 index 3c2069b827..0000000000 --- a/src/mesa/pipe/i915simple/i915_flush.c +++ /dev/null @@ -1,81 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - - -#include "pipe/p_defines.h" -#include "i915_context.h" -#include "i915_reg.h" -#include "i915_batch.h" - - -/** - * In future we may want a fence-like interface instead of finish. - */ -static void i915_flush( struct pipe_context *pipe, - unsigned flags ) -{ - struct i915_context *i915 = i915_context(pipe); - - /* Do we need to emit an MI_FLUSH command to flush the hardware - * caches? - */ - if (flags & (PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE)) { - unsigned flush = MI_FLUSH; - - if (!(flags & PIPE_FLUSH_RENDER_CACHE)) - flush |= INHIBIT_FLUSH_RENDER_CACHE; - - if (flags & PIPE_FLUSH_TEXTURE_CACHE) - flush |= FLUSH_MAP_CACHE; - - if (!BEGIN_BATCH(1, 0)) { - FLUSH_BATCH(); - assert(BEGIN_BATCH(1, 0)); - } - OUT_BATCH( flush ); - ADVANCE_BATCH(); - } - - /* If there are no flags, just flush pending commands to hardware: - */ - FLUSH_BATCH(); - - if (flags & PIPE_FLUSH_WAIT) { - i915->winsys->batch_finish(i915->winsys); - } -} - - - -void i915_init_flush_functions( struct i915_context *i915 ) -{ - i915->pipe.flush = i915_flush; -} diff --git a/src/mesa/pipe/i915simple/i915_fpc.h b/src/mesa/pipe/i915simple/i915_fpc.h deleted file mode 100644 index 8c7b68aefb..0000000000 --- a/src/mesa/pipe/i915simple/i915_fpc.h +++ /dev/null @@ -1,213 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef I915_FPC_H -#define I915_FPC_H - -#include "pipe/p_util.h" - -#include "i915_context.h" -#include "i915_reg.h" - - - -#define I915_PROGRAM_SIZE 192 - - - -/** - * Program translation state - */ -struct i915_fp_compile { - const struct pipe_shader_state *shader; - - struct vertex_info *vertex_info; - - uint declarations[I915_PROGRAM_SIZE]; - uint program[I915_PROGRAM_SIZE]; - - uint input_semantic_name[PIPE_MAX_SHADER_INPUTS]; - uint input_semantic_index[PIPE_MAX_SHADER_INPUTS]; - - uint output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; - uint output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; - - /** points into the i915->current.constants array: */ - float (*constants)[4]; - uint num_constants; - uint constant_flags[I915_MAX_CONSTANT]; /**< status of each constant */ - - uint *csr; /**< Cursor, points into program. */ - - uint *decl; /**< Cursor, points into declarations. */ - - uint decl_s; /**< flags for which s regs need to be decl'd */ - uint decl_t; /**< flags for which t regs need to be decl'd */ - - uint temp_flag; /**< Tracks temporary regs which are in use */ - uint utemp_flag; /**< Tracks TYPE_U temporary regs which are in use */ - - uint nr_tex_indirect; - uint nr_tex_insn; - uint nr_alu_insn; - uint nr_decl_insn; - - boolean error; /**< Set if i915_program_error() is called */ - uint wpos_tex; - uint NumNativeInstructions; - uint NumNativeAluInstructions; - uint NumNativeTexInstructions; - uint NumNativeTexIndirections; -}; - - -/* Having zero and one in here makes the definition of swizzle a lot - * easier. - */ -#define UREG_TYPE_SHIFT 29 -#define UREG_NR_SHIFT 24 -#define UREG_CHANNEL_X_NEGATE_SHIFT 23 -#define UREG_CHANNEL_X_SHIFT 20 -#define UREG_CHANNEL_Y_NEGATE_SHIFT 19 -#define UREG_CHANNEL_Y_SHIFT 16 -#define UREG_CHANNEL_Z_NEGATE_SHIFT 15 -#define UREG_CHANNEL_Z_SHIFT 12 -#define UREG_CHANNEL_W_NEGATE_SHIFT 11 -#define UREG_CHANNEL_W_SHIFT 8 -#define UREG_CHANNEL_ZERO_NEGATE_MBZ 5 -#define UREG_CHANNEL_ZERO_SHIFT 4 -#define UREG_CHANNEL_ONE_NEGATE_MBZ 1 -#define UREG_CHANNEL_ONE_SHIFT 0 - -#define UREG_BAD 0xffffffff /* not a valid ureg */ - -#define X SRC_X -#define Y SRC_Y -#define Z SRC_Z -#define W SRC_W -#define ZERO SRC_ZERO -#define ONE SRC_ONE - -/* Construct a ureg: - */ -#define UREG( type, nr ) (((type)<< UREG_TYPE_SHIFT) | \ - ((nr) << UREG_NR_SHIFT) | \ - (X << UREG_CHANNEL_X_SHIFT) | \ - (Y << UREG_CHANNEL_Y_SHIFT) | \ - (Z << UREG_CHANNEL_Z_SHIFT) | \ - (W << UREG_CHANNEL_W_SHIFT) | \ - (ZERO << UREG_CHANNEL_ZERO_SHIFT) | \ - (ONE << UREG_CHANNEL_ONE_SHIFT)) - -#define GET_CHANNEL_SRC( reg, channel ) ((reg<<(channel*4)) & (0xf<<20)) -#define CHANNEL_SRC( src, channel ) (src>>(channel*4)) - -#define GET_UREG_TYPE(reg) (((reg)>>UREG_TYPE_SHIFT)®_TYPE_MASK) -#define GET_UREG_NR(reg) (((reg)>>UREG_NR_SHIFT)®_NR_MASK) - - - -#define UREG_XYZW_CHANNEL_MASK 0x00ffff00 - -/* One neat thing about the UREG representation: - */ -static INLINE int -swizzle(int reg, uint x, uint y, uint z, uint w) -{ - assert(x <= SRC_ONE); - assert(y <= SRC_ONE); - assert(z <= SRC_ONE); - assert(w <= SRC_ONE); - return ((reg & ~UREG_XYZW_CHANNEL_MASK) | - CHANNEL_SRC(GET_CHANNEL_SRC(reg, x), 0) | - CHANNEL_SRC(GET_CHANNEL_SRC(reg, y), 1) | - CHANNEL_SRC(GET_CHANNEL_SRC(reg, z), 2) | - CHANNEL_SRC(GET_CHANNEL_SRC(reg, w), 3)); -} - - - -/*********************************************************************** - * Public interface for the compiler - */ -extern void i915_translate_fragment_program( struct i915_context *i915 ); - - - -extern uint i915_get_temp(struct i915_fp_compile *p); -extern uint i915_get_utemp(struct i915_fp_compile *p); -extern void i915_release_utemps(struct i915_fp_compile *p); - - -extern uint i915_emit_texld(struct i915_fp_compile *p, - uint dest, - uint destmask, - uint sampler, uint coord, uint op); - -extern uint i915_emit_arith(struct i915_fp_compile *p, - uint op, - uint dest, - uint mask, - uint saturate, - uint src0, uint src1, uint src2); - -extern uint i915_emit_decl(struct i915_fp_compile *p, - uint type, uint nr, uint d0_flags); - - -extern uint i915_emit_const1f(struct i915_fp_compile *p, float c0); - -extern uint i915_emit_const2f(struct i915_fp_compile *p, - float c0, float c1); - -extern uint i915_emit_const4fv(struct i915_fp_compile *p, - const float * c); - -extern uint i915_emit_const4f(struct i915_fp_compile *p, - float c0, float c1, - float c2, float c3); - - -/*====================================================================== - * i915_fpc_debug.c - */ -extern void i915_disassemble_program(const uint * program, uint sz); - - -/*====================================================================== - * i915_fpc_translate.c - */ - -extern void -i915_program_error(struct i915_fp_compile *p, const char *msg, ...); - -extern void -i915_translate_fragment_program(struct i915_context *i915); - - -#endif diff --git a/src/mesa/pipe/i915simple/i915_fpc_emit.c b/src/mesa/pipe/i915simple/i915_fpc_emit.c deleted file mode 100644 index 74924ff0a1..0000000000 --- a/src/mesa/pipe/i915simple/i915_fpc_emit.c +++ /dev/null @@ -1,375 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915_reg.h" -#include "i915_context.h" -#include "i915_fpc.h" - - -#define A0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) -#define D0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) -#define T0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) -#define A0_SRC0( reg ) (((reg)&UREG_MASK)>>UREG_A0_SRC0_SHIFT_LEFT) -#define A1_SRC0( reg ) (((reg)&UREG_MASK)<>UREG_A1_SRC1_SHIFT_LEFT) -#define A2_SRC1( reg ) (((reg)&UREG_MASK)<>UREG_A2_SRC2_SHIFT_LEFT) - -/* These are special, and don't have swizzle/negate bits. - */ -#define T0_SAMPLER( reg ) (GET_UREG_NR(reg)<temp_flag); - if (!bit) { - i915_program_error(p, "i915_get_temp: out of temporaries\n"); - return 0; - } - - p->temp_flag |= 1 << (bit - 1); - return UREG(REG_TYPE_R, (bit - 1)); -} - - -uint -i915_get_utemp(struct i915_fp_compile * p) -{ - int bit = ffs(~p->utemp_flag); - if (!bit) { - i915_program_error(p, "i915_get_utemp: out of temporaries\n"); - return 0; - } - - p->utemp_flag |= 1 << (bit - 1); - return UREG(REG_TYPE_U, (bit - 1)); -} - -void -i915_release_utemps(struct i915_fp_compile *p) -{ - p->utemp_flag = ~0x7; -} - - -uint -i915_emit_decl(struct i915_fp_compile *p, - uint type, uint nr, uint d0_flags) -{ - uint reg = UREG(type, nr); - - if (type == REG_TYPE_T) { - if (p->decl_t & (1 << nr)) - return reg; - - p->decl_t |= (1 << nr); - } - else if (type == REG_TYPE_S) { - if (p->decl_s & (1 << nr)) - return reg; - - p->decl_s |= (1 << nr); - } - else - return reg; - - *(p->decl++) = (D0_DCL | D0_DEST(reg) | d0_flags); - *(p->decl++) = D1_MBZ; - *(p->decl++) = D2_MBZ; - - p->nr_decl_insn++; - return reg; -} - -uint -i915_emit_arith(struct i915_fp_compile * p, - uint op, - uint dest, - uint mask, - uint saturate, uint src0, uint src1, uint src2) -{ - uint c[3]; - uint nr_const = 0; - - assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); - dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest)); - assert(dest); - - if (GET_UREG_TYPE(src0) == REG_TYPE_CONST) - c[nr_const++] = 0; - if (GET_UREG_TYPE(src1) == REG_TYPE_CONST) - c[nr_const++] = 1; - if (GET_UREG_TYPE(src2) == REG_TYPE_CONST) - c[nr_const++] = 2; - - /* Recursively call this function to MOV additional const values - * into temporary registers. Use utemp registers for this - - * currently shouldn't be possible to run out, but keep an eye on - * this. - */ - if (nr_const > 1) { - uint s[3], first, i, old_utemp_flag; - - s[0] = src0; - s[1] = src1; - s[2] = src2; - old_utemp_flag = p->utemp_flag; - - first = GET_UREG_NR(s[c[0]]); - for (i = 1; i < nr_const; i++) { - if (GET_UREG_NR(s[c[i]]) != first) { - uint tmp = i915_get_utemp(p); - - i915_emit_arith(p, A0_MOV, tmp, A0_DEST_CHANNEL_ALL, 0, - s[c[i]], 0, 0); - s[c[i]] = tmp; - } - } - - src0 = s[0]; - src1 = s[1]; - src2 = s[2]; - p->utemp_flag = old_utemp_flag; /* restore */ - } - - *(p->csr++) = (op | A0_DEST(dest) | mask | saturate | A0_SRC0(src0)); - *(p->csr++) = (A1_SRC0(src0) | A1_SRC1(src1)); - *(p->csr++) = (A2_SRC1(src1) | A2_SRC2(src2)); - - p->nr_alu_insn++; - return dest; -} - -uint i915_emit_texld( struct i915_fp_compile *p, - uint dest, - uint destmask, - uint sampler, - uint coord, - uint op ) -{ - uint k = UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord)); - if (coord != k) { - /* No real way to work around this in the general case - need to - * allocate and declare a new temporary register (a utemp won't - * do). Will fallback for now. - */ - i915_program_error(p, "Can't (yet) swizzle TEX arguments"); - assert(0); - return 0; - } - - /* Don't worry about saturate as we only support - */ - if (destmask != A0_DEST_CHANNEL_ALL) { - uint tmp = i915_get_utemp(p); - i915_emit_texld( p, tmp, A0_DEST_CHANNEL_ALL, sampler, coord, op ); - i915_emit_arith( p, A0_MOV, dest, destmask, 0, tmp, 0, 0 ); - return dest; - } - else { - assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); - assert(dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest))); - - if (GET_UREG_TYPE(coord) != REG_TYPE_T) { - p->nr_tex_indirect++; - } - - *(p->csr++) = (op | - T0_DEST( dest ) | - T0_SAMPLER( sampler )); - - *(p->csr++) = T1_ADDRESS_REG( coord ); - *(p->csr++) = T2_MBZ; - - p->nr_tex_insn++; - return dest; - } -} - - -uint -i915_emit_const1f(struct i915_fp_compile * p, float c0) -{ - unsigned reg, idx; - - if (c0 == 0.0) - return swizzle(UREG(REG_TYPE_R, 0), ZERO, ZERO, ZERO, ZERO); - if (c0 == 1.0) - return swizzle(UREG(REG_TYPE_R, 0), ONE, ONE, ONE, ONE); - - for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == I915_CONSTFLAG_PARAM) - continue; - for (idx = 0; idx < 4; idx++) { - if (!(p->constant_flags[reg] & (1 << idx)) || - p->constants[reg][idx] == c0) { - p->constants[reg][idx] = c0; - p->constant_flags[reg] |= 1 << idx; - if (reg + 1 > p->num_constants) - p->num_constants = reg + 1; - return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE); - } - } - } - - i915_program_error(p, "i915_emit_const1f: out of constants\n"); - return 0; -} - -uint -i915_emit_const2f(struct i915_fp_compile * p, float c0, float c1) -{ - unsigned reg, idx; - - if (c0 == 0.0) - return swizzle(i915_emit_const1f(p, c1), ZERO, X, Z, W); - if (c0 == 1.0) - return swizzle(i915_emit_const1f(p, c1), ONE, X, Z, W); - - if (c1 == 0.0) - return swizzle(i915_emit_const1f(p, c0), X, ZERO, Z, W); - if (c1 == 1.0) - return swizzle(i915_emit_const1f(p, c0), X, ONE, Z, W); - - for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == 0xf || - p->constant_flags[reg] == I915_CONSTFLAG_PARAM) - continue; - for (idx = 0; idx < 3; idx++) { - if (!(p->constant_flags[reg] & (3 << idx))) { - p->constants[reg][idx + 0] = c0; - p->constants[reg][idx + 1] = c1; - p->constant_flags[reg] |= 3 << idx; - if (reg + 1 > p->num_constants) - p->num_constants = reg + 1; - return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, ONE); - } - } - } - - i915_program_error(p, "i915_emit_const2f: out of constants\n"); - return 0; -} - - - -uint -i915_emit_const4f(struct i915_fp_compile * p, - float c0, float c1, float c2, float c3) -{ - unsigned reg; - - for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == 0xf && - p->constants[reg][0] == c0 && - p->constants[reg][1] == c1 && - p->constants[reg][2] == c2 && - p->constants[reg][3] == c3) { - return UREG(REG_TYPE_CONST, reg); - } - else if (p->constant_flags[reg] == 0) { - - p->constants[reg][0] = c0; - p->constants[reg][1] = c1; - p->constants[reg][2] = c2; - p->constants[reg][3] = c3; - p->constant_flags[reg] = 0xf; - if (reg + 1 > p->num_constants) - p->num_constants = reg + 1; - return UREG(REG_TYPE_CONST, reg); - } - } - - i915_program_error(p, "i915_emit_const4f: out of constants\n"); - return 0; -} - - -uint -i915_emit_const4fv(struct i915_fp_compile * p, const float * c) -{ - return i915_emit_const4f(p, c[0], c[1], c[2], c[3]); -} - - -#if 00000/*UNUSED*/ -/* Reserve a slot in the constant file for a Mesa state parameter. - * These will later need to be tracked on statechanges, but that is - * done elsewhere. - */ -uint -i915_emit_param4fv(struct i915_fp_compile * p, const float * values) -{ - struct i915_fragment_program *fp = p->fp; - int i; - - for (i = 0; i < fp->nr_params; i++) { - if (fp->param[i].values == values) - return UREG(REG_TYPE_CONST, fp->param[i].reg); - } - - if (p->constants->nr_constants == I915_MAX_CONSTANT || - fp->nr_params == I915_MAX_CONSTANT) { - i915_program_error(p, "i915_emit_param4fv: out of constants\n"); - return 0; - } - - { - int reg = p->constants->nr_constants++; - int i = fp->nr_params++; - - assert (p->constant_flags[reg] == 0); - p->constant_flags[reg] = I915_CONSTFLAG_PARAM; - - fp->param[i].values = values; - fp->param[i].reg = reg; - - return UREG(REG_TYPE_CONST, reg); - } -} -#endif diff --git a/src/mesa/pipe/i915simple/i915_fpc_translate.c b/src/mesa/pipe/i915simple/i915_fpc_translate.c deleted file mode 100644 index 868f0c7e04..0000000000 --- a/src/mesa/pipe/i915simple/i915_fpc_translate.c +++ /dev/null @@ -1,1135 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include - -#include "i915_reg.h" -#include "i915_context.h" -#include "i915_fpc.h" - -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" - -#include "pipe/draw/draw_vertex.h" - - -/** - * Simple pass-through fragment shader to use when we don't have - * a real shader (or it fails to compile for some reason). - */ -static unsigned passthrough[] = -{ - _3DSTATE_PIXEL_SHADER_PROGRAM | ((2*3)-1), - - /* declare input color: - */ - (D0_DCL | - (REG_TYPE_T << D0_TYPE_SHIFT) | - (T_DIFFUSE << D0_NR_SHIFT) | - D0_CHANNEL_ALL), - 0, - 0, - - /* move to output color: - */ - (A0_MOV | - (REG_TYPE_OC << A0_DEST_TYPE_SHIFT) | - A0_DEST_CHANNEL_ALL | - (REG_TYPE_T << A0_SRC0_TYPE_SHIFT) | - (T_DIFFUSE << A0_SRC0_NR_SHIFT)), - 0x01230000, /* .xyzw */ - 0 -}; - - -/* 1, -1/3!, 1/5!, -1/7! */ -static const float sin_constants[4] = { 1.0, - -1.0f / (3 * 2 * 1), - 1.0f / (5 * 4 * 3 * 2 * 1), - -1.0f / (7 * 6 * 5 * 4 * 3 * 2 * 1) -}; - -/* 1, -1/2!, 1/4!, -1/6! */ -static const float cos_constants[4] = { 1.0, - -1.0f / (2 * 1), - 1.0f / (4 * 3 * 2 * 1), - -1.0f / (6 * 5 * 4 * 3 * 2 * 1) -}; - - - -/** - * component-wise negation of ureg - */ -static INLINE int -negate(int reg, int x, int y, int z, int w) -{ - /* Another neat thing about the UREG representation */ - return reg ^ (((x & 1) << UREG_CHANNEL_X_NEGATE_SHIFT) | - ((y & 1) << UREG_CHANNEL_Y_NEGATE_SHIFT) | - ((z & 1) << UREG_CHANNEL_Z_NEGATE_SHIFT) | - ((w & 1) << UREG_CHANNEL_W_NEGATE_SHIFT)); -} - - -static void -i915_use_passthrough_shader(struct i915_context *i915) -{ - debug_printf("**** Using i915 pass-through fragment shader\n"); - - i915->current.program = (uint *) MALLOC(sizeof(passthrough)); - if (i915->current.program) { - memcpy(i915->current.program, passthrough, sizeof(passthrough)); - i915->current.program_len = Elements(passthrough); - } - - i915->current.num_constants[PIPE_SHADER_FRAGMENT] = 0; - i915->current.num_user_constants[PIPE_SHADER_FRAGMENT] = 0; -} - - -void -i915_program_error(struct i915_fp_compile *p, const char *msg, ...) -{ - va_list args; - char buffer[1024]; - - debug_printf("i915_program_error: "); - va_start( args, msg ); - vsprintf( buffer, msg, args ); - va_end( args ); - debug_printf(buffer); - debug_printf("\n"); - - p->error = 1; -} - - - -/** - * Construct a ureg for the given source register. Will emit - * constants, apply swizzling and negation as needed. - */ -static uint -src_vector(struct i915_fp_compile *p, - const struct tgsi_full_src_register *source) -{ - uint index = source->SrcRegister.Index; - uint src, sem_name, sem_ind; - - switch (source->SrcRegister.File) { - case TGSI_FILE_TEMPORARY: - if (source->SrcRegister.Index >= I915_MAX_TEMPORARY) { - i915_program_error(p, "Exceeded max temporary reg"); - return 0; - } - src = UREG(REG_TYPE_R, index); - break; - case TGSI_FILE_INPUT: - /* XXX: Packing COL1, FOGC into a single attribute works for - * texenv programs, but will fail for real fragment programs - * that use these attributes and expect them to be a full 4 - * components wide. Could use a texcoord to pass these - * attributes if necessary, but that won't work in the general - * case. - * - * We also use a texture coordinate to pass wpos when possible. - */ - - /* use vertex format info to map a slot number to a VF attrib */ - assert(index < p->vertex_info->num_attribs); - - sem_name = p->input_semantic_name[index]; - sem_ind = p->input_semantic_index[index]; - - switch (sem_name) { - case TGSI_SEMANTIC_POSITION: - debug_printf("SKIP SEM POS\n"); - /* - assert(p->wpos_tex != -1); - src = i915_emit_decl(p, REG_TYPE_T, p->wpos_tex, D0_CHANNEL_ALL); - */ - break; - case TGSI_SEMANTIC_COLOR: - if (sem_ind == 0) { - src = i915_emit_decl(p, REG_TYPE_T, T_DIFFUSE, D0_CHANNEL_ALL); - } - else { - /* secondary color */ - assert(sem_ind == 1); - src = i915_emit_decl(p, REG_TYPE_T, T_SPECULAR, D0_CHANNEL_XYZ); - src = swizzle(src, X, Y, Z, ONE); - } - break; - case TGSI_SEMANTIC_FOG: - src = i915_emit_decl(p, REG_TYPE_T, T_FOG_W, D0_CHANNEL_W); - src = swizzle(src, W, W, W, W); - break; - case TGSI_SEMANTIC_GENERIC: - /* usually a texcoord */ - src = i915_emit_decl(p, REG_TYPE_T, T_TEX0 + sem_ind, D0_CHANNEL_ALL); - break; - default: - i915_program_error(p, "Bad source->Index"); - return 0; - } - break; - - case TGSI_FILE_IMMEDIATE: - /* XXX unfinished - need to append immediates onto const buffer */ - /* fall-through */ - case TGSI_FILE_CONSTANT: - src = UREG(REG_TYPE_CONST, index); - break; - - default: - i915_program_error(p, "Bad source->File"); - return 0; - } - - if (source->SrcRegister.Extended) { - src = swizzle(src, - source->SrcRegisterExtSwz.ExtSwizzleX, - source->SrcRegisterExtSwz.ExtSwizzleY, - source->SrcRegisterExtSwz.ExtSwizzleZ, - source->SrcRegisterExtSwz.ExtSwizzleW); - } - else { - src = swizzle(src, - source->SrcRegister.SwizzleX, - source->SrcRegister.SwizzleY, - source->SrcRegister.SwizzleZ, - source->SrcRegister.SwizzleW); - } - - - /* There's both negate-all-components and per-component negation. - * Try to handle both here. - */ - { - int nx = source->SrcRegisterExtSwz.NegateX; - int ny = source->SrcRegisterExtSwz.NegateY; - int nz = source->SrcRegisterExtSwz.NegateZ; - int nw = source->SrcRegisterExtSwz.NegateW; - if (source->SrcRegister.Negate) { - nx = !nx; - ny = !ny; - nz = !nz; - nw = !nw; - } - src = negate(src, nx, ny, nz, nw); - } - - /* no abs() or post-abs negation */ -#if 0 - /* XXX assertions disabled to allow arbfplight.c to run */ - /* XXX enable these assertions, or fix things */ - assert(!source->SrcRegisterExtMod.Absolute); - assert(!source->SrcRegisterExtMod.Negate); -#endif - return src; -} - - -/** - * Construct a ureg for a destination register. - */ -static uint -get_result_vector(struct i915_fp_compile *p, - const struct tgsi_full_dst_register *dest) -{ - switch (dest->DstRegister.File) { - case TGSI_FILE_OUTPUT: - { - uint sem_name = p->output_semantic_name[dest->DstRegister.Index]; - switch (sem_name) { - case TGSI_SEMANTIC_POSITION: - return UREG(REG_TYPE_OD, 0); - case TGSI_SEMANTIC_COLOR: - return UREG(REG_TYPE_OC, 0); - default: - i915_program_error(p, "Bad inst->DstReg.Index/semantics"); - return 0; - } - } - case TGSI_FILE_TEMPORARY: - return UREG(REG_TYPE_R, dest->DstRegister.Index); - default: - i915_program_error(p, "Bad inst->DstReg.File"); - return 0; - } -} - - -/** - * Compute flags for saturation and writemask. - */ -static uint -get_result_flags(const struct tgsi_full_instruction *inst) -{ - const uint writeMask - = inst->FullDstRegisters[0].DstRegister.WriteMask; - uint flags = 0x0; - - if (inst->Instruction.Saturate == TGSI_SAT_ZERO_ONE) - flags |= A0_DEST_SATURATE; - - if (writeMask & TGSI_WRITEMASK_X) - flags |= A0_DEST_CHANNEL_X; - if (writeMask & TGSI_WRITEMASK_Y) - flags |= A0_DEST_CHANNEL_Y; - if (writeMask & TGSI_WRITEMASK_Z) - flags |= A0_DEST_CHANNEL_Z; - if (writeMask & TGSI_WRITEMASK_W) - flags |= A0_DEST_CHANNEL_W; - - return flags; -} - - -/** - * Convert TGSI_TEXTURE_x token to DO_SAMPLE_TYPE_x token - */ -static uint -translate_tex_src_target(struct i915_fp_compile *p, uint tex) -{ - switch (tex) { - case TGSI_TEXTURE_1D: - return D0_SAMPLE_TYPE_2D; - case TGSI_TEXTURE_2D: - return D0_SAMPLE_TYPE_2D; - case TGSI_TEXTURE_RECT: - return D0_SAMPLE_TYPE_2D; - case TGSI_TEXTURE_3D: - return D0_SAMPLE_TYPE_VOLUME; - case TGSI_TEXTURE_CUBE: - return D0_SAMPLE_TYPE_CUBE; - default: - i915_program_error(p, "TexSrc type"); - return 0; - } -} - - -/** - * Generate texel lookup instruction. - */ -static void -emit_tex(struct i915_fp_compile *p, - const struct tgsi_full_instruction *inst, - uint opcode) -{ - uint texture = inst->InstructionExtTexture.Texture; - uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; - uint tex = translate_tex_src_target( p, texture ); - uint sampler = i915_emit_decl(p, REG_TYPE_S, unit, tex); - uint coord = src_vector( p, &inst->FullSrcRegisters[0]); - - i915_emit_texld( p, - get_result_vector( p, &inst->FullDstRegisters[0] ), - get_result_flags( inst ), - sampler, - coord, - opcode); -} - - -/** - * Generate a simple arithmetic instruction - * \param opcode the i915 opcode - * \param numArgs the number of input/src arguments - */ -static void -emit_simple_arith(struct i915_fp_compile *p, - const struct tgsi_full_instruction *inst, - uint opcode, uint numArgs) -{ - uint arg1, arg2, arg3; - - assert(numArgs <= 3); - - arg1 = (numArgs < 1) ? 0 : src_vector( p, &inst->FullSrcRegisters[0] ); - arg2 = (numArgs < 2) ? 0 : src_vector( p, &inst->FullSrcRegisters[1] ); - arg3 = (numArgs < 3) ? 0 : src_vector( p, &inst->FullSrcRegisters[2] ); - - i915_emit_arith( p, - opcode, - get_result_vector( p, &inst->FullDstRegisters[0]), - get_result_flags( inst ), 0, - arg1, - arg2, - arg3 ); -} - -#ifndef M_PI -#define M_PI 3.14159265358979323846 -#endif - -/* - * Translate TGSI instruction to i915 instruction. - * - * Possible concerns: - * - * SIN, COS -- could use another taylor step? - * LIT -- results seem a little different to sw mesa - * LOG -- different to mesa on negative numbers, but this is conformant. - */ -static void -i915_translate_instruction(struct i915_fp_compile *p, - const struct tgsi_full_instruction *inst) -{ - uint writemask; - uint src0, src1, src2, flags; - uint tmp = 0; - - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ABS: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - i915_emit_arith(p, - A0_MAX, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - src0, negate(src0, 1, 1, 1, 1), 0); - break; - - case TGSI_OPCODE_ADD: - emit_simple_arith(p, inst, A0_ADD, 2); - break; - - case TGSI_OPCODE_CMP: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - src1 = src_vector(p, &inst->FullSrcRegisters[1]); - src2 = src_vector(p, &inst->FullSrcRegisters[2]); - i915_emit_arith(p, A0_CMP, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), - 0, src0, src2, src1); /* NOTE: order of src2, src1 */ - break; - - case TGSI_OPCODE_COS: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - tmp = i915_get_utemp(p); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - src0, i915_emit_const1f(p, 1.0f / (float) (M_PI * 2.0)), 0); - - i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); - - /* By choosing different taylor constants, could get rid of this mul: - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, i915_emit_const1f(p, (float) (M_PI * 2.0)), 0); - - /* - * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 - * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, 1 - * t0 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 - * result = DP4 t0, cos_constants - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XY, 0, - swizzle(tmp, X, X, ONE, ONE), - swizzle(tmp, X, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XYZ, 0, - swizzle(tmp, X, Y, X, ONE), - swizzle(tmp, X, X, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XYZ, 0, - swizzle(tmp, X, X, Z, ONE), - swizzle(tmp, Z, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - swizzle(tmp, ONE, Z, Y, X), - i915_emit_const4fv(p, cos_constants), 0); - break; - - case TGSI_OPCODE_DP3: - emit_simple_arith(p, inst, A0_DP3, 2); - break; - - case TGSI_OPCODE_DP4: - emit_simple_arith(p, inst, A0_DP4, 2); - break; - - case TGSI_OPCODE_DPH: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - src1 = src_vector(p, &inst->FullSrcRegisters[1]); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - swizzle(src0, X, Y, Z, ONE), src1, 0); - break; - - case TGSI_OPCODE_DST: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - src1 = src_vector(p, &inst->FullSrcRegisters[1]); - - /* result[0] = 1 * 1; - * result[1] = a[1] * b[1]; - * result[2] = a[2] * 1; - * result[3] = 1 * b[3]; - */ - i915_emit_arith(p, - A0_MUL, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - swizzle(src0, ONE, Y, Z, ONE), - swizzle(src1, ONE, Y, ONE, W), 0); - break; - - case TGSI_OPCODE_END: - /* no-op */ - break; - - case TGSI_OPCODE_EX2: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - - i915_emit_arith(p, - A0_EXP, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - swizzle(src0, X, X, X, X), 0, 0); - break; - - case TGSI_OPCODE_FLR: - emit_simple_arith(p, inst, A0_FLR, 1); - break; - - case TGSI_OPCODE_FRC: - emit_simple_arith(p, inst, A0_FRC, 1); - break; - - case TGSI_OPCODE_KIL: - /* unconditional kill */ - assert(0); /* not tested yet */ -#if 0 - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - tmp = i915_get_utemp(p); - - i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ - 0, src0, T0_TEXKILL); -#endif - break; - - case TGSI_OPCODE_KILP: - /* kill if src[0].x < 0 || src[0].y < 0 ... */ - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - tmp = i915_get_utemp(p); - - i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ - 0, src0, T0_TEXKILL); - break; - - case TGSI_OPCODE_LG2: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - - i915_emit_arith(p, - A0_LOG, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - swizzle(src0, X, X, X, X), 0, 0); - break; - - case TGSI_OPCODE_LIT: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - tmp = i915_get_utemp(p); - - /* tmp = max( a.xyzw, a.00zw ) - * XXX: Clamp tmp.w to -128..128 - * tmp.y = log(tmp.y) - * tmp.y = tmp.w * tmp.y - * tmp.y = exp(tmp.y) - * result = cmp (a.11-x1, a.1x01, a.1xy1 ) - */ - i915_emit_arith(p, A0_MAX, tmp, A0_DEST_CHANNEL_ALL, 0, - src0, swizzle(src0, ZERO, ZERO, Z, W), 0); - - i915_emit_arith(p, A0_LOG, tmp, A0_DEST_CHANNEL_Y, 0, - swizzle(tmp, Y, Y, Y, Y), 0, 0); - - i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_Y, 0, - swizzle(tmp, ZERO, Y, ZERO, ZERO), - swizzle(tmp, ZERO, W, ZERO, ZERO), 0); - - i915_emit_arith(p, A0_EXP, tmp, A0_DEST_CHANNEL_Y, 0, - swizzle(tmp, Y, Y, Y, Y), 0, 0); - - i915_emit_arith(p, A0_CMP, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - negate(swizzle(tmp, ONE, ONE, X, ONE), 0, 0, 1, 0), - swizzle(tmp, ONE, X, ZERO, ONE), - swizzle(tmp, ONE, X, Y, ONE)); - - break; - - case TGSI_OPCODE_LRP: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - src1 = src_vector(p, &inst->FullSrcRegisters[1]); - src2 = src_vector(p, &inst->FullSrcRegisters[2]); - flags = get_result_flags(inst); - tmp = i915_get_utemp(p); - - /* b*a + c*(1-a) - * - * b*a + c - ca - * - * tmp = b*a + c, - * result = (-c)*a + tmp - */ - i915_emit_arith(p, A0_MAD, tmp, - flags & A0_DEST_CHANNEL_ALL, 0, src1, src0, src2); - - i915_emit_arith(p, A0_MAD, - get_result_vector(p, &inst->FullDstRegisters[0]), - flags, 0, negate(src2, 1, 1, 1, 1), src0, tmp); - break; - - case TGSI_OPCODE_MAD: - emit_simple_arith(p, inst, A0_MAD, 3); - break; - - case TGSI_OPCODE_MAX: - emit_simple_arith(p, inst, A0_MAX, 2); - break; - - case TGSI_OPCODE_MIN: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - src1 = src_vector(p, &inst->FullSrcRegisters[1]); - tmp = i915_get_utemp(p); - flags = get_result_flags(inst); - - i915_emit_arith(p, - A0_MAX, - tmp, flags & A0_DEST_CHANNEL_ALL, 0, - negate(src0, 1, 1, 1, 1), - negate(src1, 1, 1, 1, 1), 0); - - i915_emit_arith(p, - A0_MOV, - get_result_vector(p, &inst->FullDstRegisters[0]), - flags, 0, negate(tmp, 1, 1, 1, 1), 0, 0); - break; - - case TGSI_OPCODE_MOV: - /* aka TGSI_OPCODE_SWZ */ - emit_simple_arith(p, inst, A0_MOV, 1); - break; - - case TGSI_OPCODE_MUL: - emit_simple_arith(p, inst, A0_MUL, 2); - break; - - case TGSI_OPCODE_POW: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - src1 = src_vector(p, &inst->FullSrcRegisters[1]); - tmp = i915_get_utemp(p); - flags = get_result_flags(inst); - - /* XXX: masking on intermediate values, here and elsewhere. - */ - i915_emit_arith(p, - A0_LOG, - tmp, A0_DEST_CHANNEL_X, 0, - swizzle(src0, X, X, X, X), 0, 0); - - i915_emit_arith(p, A0_MUL, tmp, A0_DEST_CHANNEL_X, 0, tmp, src1, 0); - - i915_emit_arith(p, - A0_EXP, - get_result_vector(p, &inst->FullDstRegisters[0]), - flags, 0, swizzle(tmp, X, X, X, X), 0, 0); - break; - - case TGSI_OPCODE_RET: - /* XXX: no-op? */ - break; - - case TGSI_OPCODE_RCP: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - - i915_emit_arith(p, - A0_RCP, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - swizzle(src0, X, X, X, X), 0, 0); - break; - - case TGSI_OPCODE_RSQ: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - - i915_emit_arith(p, - A0_RSQ, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - swizzle(src0, X, X, X, X), 0, 0); - break; - - case TGSI_OPCODE_SCS: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - tmp = i915_get_utemp(p); - - /* - * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 - * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x - * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x - * scs.x = DP4 t1, sin_constants - * t1 = MUL t0.xxz1 t0.z111 ; x^6 x^4 x^2 1 - * scs.y = DP4 t1, cos_constants - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XY, 0, - swizzle(src0, X, X, ONE, ONE), - swizzle(src0, X, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X, Y, X, Y), - swizzle(tmp, X, X, ONE, ONE), 0); - - writemask = inst->FullDstRegisters[0].DstRegister.WriteMask; - - if (writemask & TGSI_WRITEMASK_Y) { - uint tmp1; - - if (writemask & TGSI_WRITEMASK_X) - tmp1 = i915_get_utemp(p); - else - tmp1 = tmp; - - i915_emit_arith(p, - A0_MUL, - tmp1, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X, Y, Y, W), - swizzle(tmp, X, Z, ONE, ONE), 0); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, &inst->FullDstRegisters[0]), - A0_DEST_CHANNEL_Y, 0, - swizzle(tmp1, W, Z, Y, X), - i915_emit_const4fv(p, sin_constants), 0); - } - - if (writemask & TGSI_WRITEMASK_X) { - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XYZ, 0, - swizzle(tmp, X, X, Z, ONE), - swizzle(tmp, Z, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, &inst->FullDstRegisters[0]), - A0_DEST_CHANNEL_X, 0, - swizzle(tmp, ONE, Z, Y, X), - i915_emit_const4fv(p, cos_constants), 0); - } - break; - - case TGSI_OPCODE_SGE: - emit_simple_arith(p, inst, A0_SGE, 2); - break; - - case TGSI_OPCODE_SIN: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - tmp = i915_get_utemp(p); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - src0, i915_emit_const1f(p, 1.0f / (float) (M_PI * 2.0)), 0); - - i915_emit_arith(p, A0_MOD, tmp, A0_DEST_CHANNEL_X, 0, tmp, 0, 0); - - /* By choosing different taylor constants, could get rid of this mul: - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_X, 0, - tmp, i915_emit_const1f(p, (float) (M_PI * 2.0)), 0); - - /* - * t0.xy = MUL x.xx11, x.x1111 ; x^2, x, 1, 1 - * t0 = MUL t0.xyxy t0.xx11 ; x^4, x^3, x^2, x - * t1 = MUL t0.xyyw t0.yz11 ; x^7 x^5 x^3 x - * result = DP4 t1.wzyx, sin_constants - */ - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_XY, 0, - swizzle(tmp, X, X, ONE, ONE), - swizzle(tmp, X, ONE, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X, Y, X, Y), - swizzle(tmp, X, X, ONE, ONE), 0); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(tmp, X, Y, Y, W), - swizzle(tmp, X, Z, ONE, ONE), 0); - - i915_emit_arith(p, - A0_DP4, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - swizzle(tmp, W, Z, Y, X), - i915_emit_const4fv(p, sin_constants), 0); - break; - - case TGSI_OPCODE_SLT: - emit_simple_arith(p, inst, A0_SLT, 2); - break; - - case TGSI_OPCODE_SUB: - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - src1 = src_vector(p, &inst->FullSrcRegisters[1]); - - i915_emit_arith(p, - A0_ADD, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - src0, negate(src1, 1, 1, 1, 1), 0); - break; - - case TGSI_OPCODE_TEX: - if (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide - == TGSI_EXTSWIZZLE_W) { - emit_tex(p, inst, T0_TEXLDP); - } - else { - emit_tex(p, inst, T0_TEXLD); - } - break; - - case TGSI_OPCODE_TXB: - emit_tex(p, inst, T0_TEXLDB); - break; - - case TGSI_OPCODE_XPD: - /* Cross product: - * result.x = src0.y * src1.z - src0.z * src1.y; - * result.y = src0.z * src1.x - src0.x * src1.z; - * result.z = src0.x * src1.y - src0.y * src1.x; - * result.w = undef; - */ - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - src1 = src_vector(p, &inst->FullSrcRegisters[1]); - tmp = i915_get_utemp(p); - - i915_emit_arith(p, - A0_MUL, - tmp, A0_DEST_CHANNEL_ALL, 0, - swizzle(src0, Z, X, Y, ONE), - swizzle(src1, Y, Z, X, ONE), 0); - - i915_emit_arith(p, - A0_MAD, - get_result_vector(p, &inst->FullDstRegisters[0]), - get_result_flags(inst), 0, - swizzle(src0, Y, Z, X, ONE), - swizzle(src1, Z, X, Y, ONE), - negate(tmp, 1, 1, 1, 0)); - break; - - default: - i915_program_error(p, "bad opcode %d", inst->Instruction.Opcode); - return; - } - - i915_release_utemps(p); -} - - -/** - * Translate TGSI fragment shader into i915 hardware instructions. - * \param p the translation state - * \param tokens the TGSI token array - */ -static void -i915_translate_instructions(struct i915_fp_compile *p, - const struct tgsi_token *tokens) -{ - struct tgsi_parse_context parse; - - tgsi_parse_init( &parse, tokens ); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - if (parse.FullToken.FullDeclaration.Declaration.File - == TGSI_FILE_INPUT) { - /* save input register info for use in src_vector() */ - uint ind, sem, semi; - ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; - semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - /*debug_printf("FS Input DECL [%u] sem %u\n", ind, sem);*/ - p->input_semantic_name[ind] = sem; - p->input_semantic_index[ind] = semi; - } - else if (parse.FullToken.FullDeclaration.Declaration.File - == TGSI_FILE_OUTPUT) { - /* save output register info for use in get_result_vector() */ - uint ind, sem, semi; - ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; - semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - /*debug_printf("FS Output DECL [%u] sem %u\n", ind, sem);*/ - p->output_semantic_name[ind] = sem; - p->output_semantic_index[ind] = semi; - } - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - /* XXX append the immediate to the const buffer... */ - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - i915_translate_instruction(p, &parse.FullToken.FullInstruction); - break; - - default: - assert( 0 ); - } - - } /* while */ - - tgsi_parse_free (&parse); -} - - -static struct i915_fp_compile * -i915_init_compile(struct i915_context *i915, - const struct pipe_shader_state *fs) -{ - struct i915_fp_compile *p = CALLOC_STRUCT(i915_fp_compile); - - p->shader = i915->fs; - - p->vertex_info = &i915->current.vertex_info; - - /* new constants found during translation get appended after the - * user-provided constants. - */ - p->constants = i915->current.constants[PIPE_SHADER_FRAGMENT]; - p->num_constants = i915->current.num_user_constants[PIPE_SHADER_FRAGMENT]; - - p->nr_tex_indirect = 1; /* correct? */ - p->nr_tex_insn = 0; - p->nr_alu_insn = 0; - p->nr_decl_insn = 0; - - memset(p->constant_flags, 0, sizeof(p->constant_flags)); - - p->csr = p->program; - p->decl = p->declarations; - p->decl_s = 0; - p->decl_t = 0; - p->temp_flag = 0xffff000; - p->utemp_flag = ~0x7; - - p->wpos_tex = -1; - - /* initialize the first program word */ - *(p->decl++) = _3DSTATE_PIXEL_SHADER_PROGRAM; - - return p; -} - - -/* Copy compile results to the fragment program struct and destroy the - * compilation context. - */ -static void -i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) -{ - unsigned long program_size = (unsigned long) (p->csr - p->program); - unsigned long decl_size = (unsigned long) (p->decl - p->declarations); - - if (p->nr_tex_indirect > I915_MAX_TEX_INDIRECT) - i915_program_error(p, "Exceeded max nr indirect texture lookups"); - - if (p->nr_tex_insn > I915_MAX_TEX_INSN) - i915_program_error(p, "Exceeded max TEX instructions"); - - if (p->nr_alu_insn > I915_MAX_ALU_INSN) - i915_program_error(p, "Exceeded max ALU instructions"); - - if (p->nr_decl_insn > I915_MAX_DECL_INSN) - i915_program_error(p, "Exceeded max DECL instructions"); - - /* free old program, if present */ - if (i915->current.program) { - FREE(i915->current.program); - i915->current.program_len = 0; - } - - if (p->error) { - p->NumNativeInstructions = 0; - p->NumNativeAluInstructions = 0; - p->NumNativeTexInstructions = 0; - p->NumNativeTexIndirections = 0; - - i915_use_passthrough_shader(i915); - } - else { - p->NumNativeInstructions - = p->nr_alu_insn + p->nr_tex_insn + p->nr_decl_insn; - p->NumNativeAluInstructions = p->nr_alu_insn; - p->NumNativeTexInstructions = p->nr_tex_insn; - p->NumNativeTexIndirections = p->nr_tex_indirect; - - /* patch in the program length */ - p->declarations[0] |= program_size + decl_size - 2; - - /* Copy compilation results to fragment program struct: - */ - i915->current.program - = (uint *) MALLOC((program_size + decl_size) * sizeof(uint)); - if (i915->current.program) { - i915->current.program_len = program_size + decl_size; - - memcpy(i915->current.program, - p->declarations, - decl_size * sizeof(uint)); - - memcpy(i915->current.program + decl_size, - p->program, - program_size * sizeof(uint)); - } - - /* update number of constants */ - i915->current.num_constants[PIPE_SHADER_FRAGMENT] = p->num_constants; - assert(i915->current.num_constants[PIPE_SHADER_FRAGMENT] - >= i915->current.num_user_constants[PIPE_SHADER_FRAGMENT]); - } - - /* Release the compilation struct: - */ - FREE(p); -} - - -/** - * Find an unused texture coordinate slot to use for fragment WPOS. - * Update p->fp->wpos_tex with the result (-1 if no used texcoord slot is found). - */ -static void -i915_find_wpos_space(struct i915_fp_compile *p) -{ -#if 0 - const uint inputs - = p->shader->inputs_read | (1 << TGSI_ATTRIB_POS); /*XXX hack*/ - uint i; - - p->wpos_tex = -1; - - if (inputs & (1 << TGSI_ATTRIB_POS)) { - for (i = 0; i < I915_TEX_UNITS; i++) { - if ((inputs & (1 << (TGSI_ATTRIB_TEX0 + i))) == 0) { - p->wpos_tex = i; - return; - } - } - - i915_program_error(p, "No free texcoord for wpos value"); - } -#else - if (p->shader->input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { - /* frag shader using the fragment position input */ -#if 0 - assert(0); -#endif - } -#endif -} - - - - -/** - * Rather than trying to intercept and jiggle depth writes during - * emit, just move the value into its correct position at the end of - * the program: - */ -static void -i915_fixup_depth_write(struct i915_fp_compile *p) -{ - /* XXX assuming pos/depth is always in output[0] */ - if (p->shader->output_semantic_name[0] == TGSI_SEMANTIC_POSITION) { - const uint depth = UREG(REG_TYPE_OD, 0); - - i915_emit_arith(p, - A0_MOV, /* opcode */ - depth, /* dest reg */ - A0_DEST_CHANNEL_W, /* write mask */ - 0, /* saturate? */ - swizzle(depth, X, Y, Z, Z), /* src0 */ - 0, 0 /* src1, src2 */); - } -} - - -void -i915_translate_fragment_program( struct i915_context *i915 ) -{ - struct i915_fp_compile *p = i915_init_compile(i915, i915->fs); - const struct tgsi_token *tokens = i915->fs->tokens; - - i915_find_wpos_space(p); - - i915_translate_instructions(p, tokens); - i915_fixup_depth_write(p); - - i915_fini_compile(i915, p); -} diff --git a/src/mesa/pipe/i915simple/i915_prim_emit.c b/src/mesa/pipe/i915simple/i915_prim_emit.c deleted file mode 100644 index c4a706c37d..0000000000 --- a/src/mesa/pipe/i915simple/i915_prim_emit.c +++ /dev/null @@ -1,215 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "pipe/draw/draw_private.h" -#include "pipe/p_util.h" - -#include "i915_context.h" -#include "i915_winsys.h" -#include "i915_reg.h" -#include "i915_state.h" -#include "i915_batch.h" - - - -/** - * Primitive emit to hardware. No support for vertex buffers or any - * nice fast paths. - */ -struct setup_stage { - struct draw_stage stage; /**< This must be first (base class) */ - - struct i915_context *i915; -}; - - - -/** - * Basically a cast wrapper. - */ -static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) -{ - return (struct setup_stage *)stage; -} - - -/** - * Extract the needed fields from vertex_header and emit i915 dwords. - * Recall that the vertices are constructed by the 'draw' module and - * have a couple of slots at the beginning (1-dword header, 4-dword - * clip pos) that we ignore here. - */ -static INLINE void -emit_hw_vertex( struct i915_context *i915, - const struct vertex_header *vertex) -{ - const struct vertex_info *vinfo = &i915->current.vertex_info; - uint i; - uint count = 0; /* for debug/sanity */ - - for (i = 0; i < vinfo->num_attribs; i++) { - switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; - case EMIT_1F: - OUT_BATCH( fui(vertex->data[i][0]) ); - count++; - break; - case EMIT_2F: - OUT_BATCH( fui(vertex->data[i][0]) ); - OUT_BATCH( fui(vertex->data[i][1]) ); - count += 2; - break; - case EMIT_3F: - OUT_BATCH( fui(vertex->data[i][0]) ); - OUT_BATCH( fui(vertex->data[i][1]) ); - OUT_BATCH( fui(vertex->data[i][2]) ); - count += 3; - break; - case EMIT_4F: - OUT_BATCH( fui(vertex->data[i][0]) ); - OUT_BATCH( fui(vertex->data[i][1]) ); - OUT_BATCH( fui(vertex->data[i][2]) ); - OUT_BATCH( fui(vertex->data[i][3]) ); - count += 4; - break; - case EMIT_4UB: - OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[i][2] ), - float_to_ubyte( vertex->data[i][1] ), - float_to_ubyte( vertex->data[i][0] ), - float_to_ubyte( vertex->data[i][3] )) ); - count += 1; - break; - default: - assert(0); - } - } - assert(count == vinfo->size); -} - - - -static INLINE void -emit_prim( struct draw_stage *stage, - struct prim_header *prim, - unsigned hwprim, - unsigned nr ) -{ - struct i915_context *i915 = setup_stage(stage)->i915; - unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ - unsigned i; - - assert(vertex_size >= 12); /* never smaller than 12 bytes */ - - if (i915->dirty) - i915_update_derived( i915 ); - - if (i915->hardware_dirty) - i915_emit_hardware_state( i915 ); - - if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) { - FLUSH_BATCH(); - - /* Make sure state is re-emitted after a flush: - */ - i915_update_derived( i915 ); - i915_emit_hardware_state( i915 ); - - if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) { - assert(0); - return; - } - } - - /* Emit each triangle as a single primitive. I told you this was - * simple. - */ - OUT_BATCH(_3DPRIMITIVE | - hwprim | - ((4 + vertex_size * nr)/4 - 2)); - - for (i = 0; i < nr; i++) - emit_hw_vertex(i915, prim->v[i]); -} - - -static void -setup_tri( struct draw_stage *stage, struct prim_header *prim ) -{ - emit_prim( stage, prim, PRIM3D_TRILIST, 3 ); -} - - -static void -setup_line(struct draw_stage *stage, struct prim_header *prim) -{ - emit_prim( stage, prim, PRIM3D_LINELIST, 2 ); -} - - -static void -setup_point(struct draw_stage *stage, struct prim_header *prim) -{ - emit_prim( stage, prim, PRIM3D_POINTLIST, 1 ); -} - - -static void setup_flush( struct draw_stage *stage, unsigned flags ) -{ -} - -static void reset_stipple_counter( struct draw_stage *stage ) -{ -} - -static void render_destroy( struct draw_stage *stage ) -{ - FREE( stage ); -} - - -/** - * Create a new primitive setup/render stage. This gets plugged into - * the 'draw' module's pipeline. - */ -struct draw_stage *i915_draw_render_stage( struct i915_context *i915 ) -{ - struct setup_stage *setup = CALLOC_STRUCT(setup_stage); - - setup->i915 = i915; - setup->stage.draw = i915->draw; - setup->stage.point = setup_point; - setup->stage.line = setup_line; - setup->stage.tri = setup_tri; - setup->stage.flush = setup_flush; - setup->stage.reset_stipple_counter = reset_stipple_counter; - setup->stage.destroy = render_destroy; - - return &setup->stage; -} diff --git a/src/mesa/pipe/i915simple/i915_prim_vbuf.c b/src/mesa/pipe/i915simple/i915_prim_vbuf.c deleted file mode 100644 index e069773fd4..0000000000 --- a/src/mesa/pipe/i915simple/i915_prim_vbuf.c +++ /dev/null @@ -1,254 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Build post-transformation, post-clipping vertex buffers and element - * lists by hooking into the end of the primitive pipeline and - * manipulating the vertex_id field in the vertex headers. - * - * XXX: work in progress - * - * \author José Fonseca - * \author Keith Whitwell - */ - - -#include "pipe/draw/draw_vbuf.h" -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" - -#include "i915_context.h" -#include "i915_reg.h" -#include "i915_winsys.h" -#include "i915_batch.h" -#include "i915_state.h" - - -/** - * Primitive renderer for i915. - */ -struct i915_vbuf_render { - struct vbuf_render base; - - struct i915_context *i915; - - /** Vertex size in bytes */ - unsigned vertex_size; - - /** Hardware primitive */ - unsigned hwprim; -}; - - -/** - * Basically a cast wrapper. - */ -static INLINE struct i915_vbuf_render * -i915_vbuf_render( struct vbuf_render *render ) -{ - assert(render); - return (struct i915_vbuf_render *)render; -} - - -static const struct vertex_info * -i915_vbuf_render_get_vertex_info( struct vbuf_render *render ) -{ - struct i915_vbuf_render *i915_render = i915_vbuf_render(render); - struct i915_context *i915 = i915_render->i915; - return &i915->current.vertex_info; -} - - -static void * -i915_vbuf_render_allocate_vertices( struct vbuf_render *render, - ushort vertex_size, - ushort nr_vertices ) -{ - struct i915_vbuf_render *i915_render = i915_vbuf_render(render); - struct i915_context *i915 = i915_render->i915; - struct pipe_winsys *winsys = i915->pipe.winsys; - size_t size = (size_t)vertex_size * (size_t)nr_vertices; - - /* FIXME: handle failure */ - assert(!i915->vbo); - i915->vbo = winsys->buffer_create(winsys, 64, I915_BUFFER_USAGE_LIT_VERTEX, - size); - - i915->dirty |= I915_NEW_VBO; - - return winsys->buffer_map(winsys, - i915->vbo, - PIPE_BUFFER_USAGE_CPU_WRITE); -} - - -static void -i915_vbuf_render_set_primitive( struct vbuf_render *render, - unsigned prim ) -{ - struct i915_vbuf_render *i915_render = i915_vbuf_render(render); - - switch(prim) { - case PIPE_PRIM_POINTS: - i915_render->hwprim = PRIM3D_POINTLIST; - break; - case PIPE_PRIM_LINES: - i915_render->hwprim = PRIM3D_LINELIST; - break; - case PIPE_PRIM_TRIANGLES: - i915_render->hwprim = PRIM3D_TRILIST; - break; - default: - assert(0); - } -} - - -static void -i915_vbuf_render_draw( struct vbuf_render *render, - const ushort *indices, - uint nr_indices) -{ - struct i915_vbuf_render *i915_render = i915_vbuf_render(render); - struct i915_context *i915 = i915_render->i915; - unsigned i; - - assert(nr_indices); - - assert((i915->dirty & ~I915_NEW_VBO) == 0); - - if (i915->dirty) - i915_update_derived( i915 ); - - if (i915->hardware_dirty) - i915_emit_hardware_state( i915 ); - - if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { - FLUSH_BATCH(); - - /* Make sure state is re-emitted after a flush: - */ - i915_update_derived( i915 ); - i915_emit_hardware_state( i915 ); - - if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { - assert(0); - return; - } - } - - OUT_BATCH( _3DPRIMITIVE | - PRIM_INDIRECT | - i915_render->hwprim | - PRIM_INDIRECT_ELTS | - nr_indices ); - for (i = 0; i + 1 < nr_indices; i += 2) { - OUT_BATCH( indices[i] | - (indices[i + 1] << 16) ); - } - if (i < nr_indices) { - OUT_BATCH( indices[i] ); - } -} - - -static void -i915_vbuf_render_release_vertices( struct vbuf_render *render, - void *vertices, - unsigned vertex_size, - unsigned vertices_used ) -{ - struct i915_vbuf_render *i915_render = i915_vbuf_render(render); - struct i915_context *i915 = i915_render->i915; - struct pipe_winsys *winsys = i915->pipe.winsys; - - assert(i915->vbo); - winsys->buffer_unmap(winsys, i915->vbo); - pipe_buffer_reference(winsys, &i915->vbo, NULL); -} - - -static void -i915_vbuf_render_destroy( struct vbuf_render *render ) -{ - struct i915_vbuf_render *i915_render = i915_vbuf_render(render); - FREE(i915_render); -} - - -/** - * Create a new primitive render. - */ -static struct vbuf_render * -i915_vbuf_render_create( struct i915_context *i915 ) -{ - struct i915_vbuf_render *i915_render = CALLOC_STRUCT(i915_vbuf_render); - - i915_render->i915 = i915; - - i915_render->base.max_vertex_buffer_bytes = 128*1024; - - /* NOTE: it must be such that state and vertices indices fit in a single - * batch buffer. - */ - i915_render->base.max_indices = 16*1024; - - i915_render->base.get_vertex_info = i915_vbuf_render_get_vertex_info; - i915_render->base.allocate_vertices = i915_vbuf_render_allocate_vertices; - i915_render->base.set_primitive = i915_vbuf_render_set_primitive; - i915_render->base.draw = i915_vbuf_render_draw; - i915_render->base.release_vertices = i915_vbuf_render_release_vertices; - i915_render->base.destroy = i915_vbuf_render_destroy; - - return &i915_render->base; -} - - -/** - * Create a new primitive vbuf/render stage. - */ -struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) -{ - struct vbuf_render *render; - struct draw_stage *stage; - - render = i915_vbuf_render_create(i915); - if(!render) - return NULL; - - stage = draw_vbuf_stage( i915->draw, render ); - if(!stage) { - render->destroy(render); - return NULL; - } - - return stage; -} diff --git a/src/mesa/pipe/i915simple/i915_reg.h b/src/mesa/pipe/i915simple/i915_reg.h deleted file mode 100644 index 04620fec68..0000000000 --- a/src/mesa/pipe/i915simple/i915_reg.h +++ /dev/null @@ -1,978 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef I915_REG_H -#define I915_REG_H - - -#define I915_SET_FIELD( var, mask, value ) (var &= ~(mask), var |= value) - -#define CMD_3D (0x3<<29) - -#define PRIM3D_INLINE (CMD_3D | (0x1f<<24)) -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_CLEAR_RECT (0xa<<18) -#define PRIM3D_ZONE_INIT (0xd<<18) -#define PRIM3D_MASK (0x1f<<18) - -/* p137 */ -#define _3DSTATE_AA_CMD (CMD_3D | (0x06<<24)) -#define AA_LINE_ECAAR_WIDTH_ENABLE (1<<16) -#define AA_LINE_ECAAR_WIDTH_0_5 0 -#define AA_LINE_ECAAR_WIDTH_1_0 (1<<14) -#define AA_LINE_ECAAR_WIDTH_2_0 (2<<14) -#define AA_LINE_ECAAR_WIDTH_4_0 (3<<14) -#define AA_LINE_REGION_WIDTH_ENABLE (1<<8) -#define AA_LINE_REGION_WIDTH_0_5 0 -#define AA_LINE_REGION_WIDTH_1_0 (1<<6) -#define AA_LINE_REGION_WIDTH_2_0 (2<<6) -#define AA_LINE_REGION_WIDTH_4_0 (3<<6) - -/* 3DSTATE_BACKFACE_STENCIL_OPS, p138*/ -#define _3DSTATE_BACKFACE_STENCIL_OPS (CMD_3D | (0x8<<24)) -#define BFO_ENABLE_STENCIL_REF (1<<23) -#define BFO_STENCIL_REF_SHIFT 15 -#define BFO_STENCIL_REF_MASK (0xff<<15) -#define BFO_ENABLE_STENCIL_FUNCS (1<<14) -#define BFO_STENCIL_TEST_SHIFT 11 -#define BFO_STENCIL_TEST_MASK (0x7<<11) -#define BFO_STENCIL_FAIL_SHIFT 8 -#define BFO_STENCIL_FAIL_MASK (0x7<<8) -#define BFO_STENCIL_PASS_Z_FAIL_SHIFT 5 -#define BFO_STENCIL_PASS_Z_FAIL_MASK (0x7<<5) -#define BFO_STENCIL_PASS_Z_PASS_SHIFT 2 -#define BFO_STENCIL_PASS_Z_PASS_MASK (0x7<<2) -#define BFO_ENABLE_STENCIL_TWO_SIDE (1<<1) -#define BFO_STENCIL_TWO_SIDE (1<<0) - - -/* 3DSTATE_BACKFACE_STENCIL_MASKS, p140 */ -#define _3DSTATE_BACKFACE_STENCIL_MASKS (CMD_3D | (0x9<<24)) -#define BFM_ENABLE_STENCIL_TEST_MASK (1<<17) -#define BFM_ENABLE_STENCIL_WRITE_MASK (1<<16) -#define BFM_STENCIL_TEST_MASK_SHIFT 8 -#define BFM_STENCIL_TEST_MASK_MASK (0xff<<8) -#define BFM_STENCIL_WRITE_MASK_SHIFT 0 -#define BFM_STENCIL_WRITE_MASK_MASK (0xff<<0) - - - -/* 3DSTATE_BIN_CONTROL p141 */ - -/* p143 */ -#define _3DSTATE_BUF_INFO_CMD (CMD_3D | (0x1d<<24) | (0x8e<<16) | 1) -/* Dword 1 */ -#define BUF_3D_ID_COLOR_BACK (0x3<<24) -#define BUF_3D_ID_DEPTH (0x7<<24) -#define BUF_3D_USE_FENCE (1<<23) -#define BUF_3D_TILED_SURFACE (1<<22) -#define BUF_3D_TILE_WALK_X 0 -#define BUF_3D_TILE_WALK_Y (1<<21) -#define BUF_3D_PITCH(x) (((x)/4)<<2) -/* Dword 2 */ -#define BUF_3D_ADDR(x) ((x) & ~0x3) - - -/* 3DSTATE_CHROMA_KEY */ - -/* 3DSTATE_CLEAR_PARAMETERS, p150 */ -#define _3DSTATE_CLEAR_PARAMETERS (CMD_3D | (0x1d<<24) | (0x9c<<16) | 5) -/* Dword 1 */ -#define CLEARPARAM_CLEAR_RECT (1 << 16) -#define CLEARPARAM_ZONE_INIT (0 << 16) -#define CLEARPARAM_WRITE_COLOR (1 << 2) -#define CLEARPARAM_WRITE_DEPTH (1 << 1) -#define CLEARPARAM_WRITE_STENCIL (1 << 0) - -/* 3DSTATE_CONSTANT_BLEND_COLOR, p153 */ -#define _3DSTATE_CONST_BLEND_COLOR_CMD (CMD_3D | (0x1d<<24) | (0x88<<16)) - - - -/* 3DSTATE_COORD_SET_BINDINGS, p154 */ -#define _3DSTATE_COORD_SET_BINDINGS (CMD_3D | (0x16<<24)) -#define CSB_TCB(iunit, eunit) ((eunit)<<(iunit*3)) - -/* p156 */ -#define _3DSTATE_DFLT_DIFFUSE_CMD (CMD_3D | (0x1d<<24) | (0x99<<16)) - -/* p157 */ -#define _3DSTATE_DFLT_SPEC_CMD (CMD_3D | (0x1d<<24) | (0x9a<<16)) - -/* p158 */ -#define _3DSTATE_DFLT_Z_CMD (CMD_3D | (0x1d<<24) | (0x98<<16)) - - -/* 3DSTATE_DEPTH_OFFSET_SCALE, p159 */ -#define _3DSTATE_DEPTH_OFFSET_SCALE (CMD_3D | (0x1d<<24) | (0x97<<16)) -/* scale in dword 1 */ - - -/* 3DSTATE_DEPTH_SUBRECT_DISABLE, p160 */ -#define _3DSTATE_DEPTH_SUBRECT_DISABLE (CMD_3D | (0x1c<<24) | (0x11<<19) | 0x2) - -/* p161 */ -#define _3DSTATE_DST_BUF_VARS_CMD (CMD_3D | (0x1d<<24) | (0x85<<16)) -/* Dword 1 */ -#define TEX_DEFAULT_COLOR_OGL (0<<30) -#define TEX_DEFAULT_COLOR_D3D (1<<30) -#define ZR_EARLY_DEPTH (1<<29) -#define LOD_PRECLAMP_OGL (1<<28) -#define LOD_PRECLAMP_D3D (0<<28) -#define DITHER_FULL_ALWAYS (0<<26) -#define DITHER_FULL_ON_FB_BLEND (1<<26) -#define DITHER_CLAMPED_ALWAYS (2<<26) -#define LINEAR_GAMMA_BLEND_32BPP (1<<25) -#define DEBUG_DISABLE_ENH_DITHER (1<<24) -#define DSTORG_HORT_BIAS(x) ((x)<<20) -#define DSTORG_VERT_BIAS(x) ((x)<<16) -#define COLOR_4_2_2_CHNL_WRT_ALL 0 -#define COLOR_4_2_2_CHNL_WRT_Y (1<<12) -#define COLOR_4_2_2_CHNL_WRT_CR (2<<12) -#define COLOR_4_2_2_CHNL_WRT_CB (3<<12) -#define COLOR_4_2_2_CHNL_WRT_CRCB (4<<12) -#define COLOR_BUF_8BIT 0 -#define COLOR_BUF_RGB555 (1<<8) -#define COLOR_BUF_RGB565 (2<<8) -#define COLOR_BUF_ARGB8888 (3<<8) -#define DEPTH_FRMT_16_FIXED 0 -#define DEPTH_FRMT_16_FLOAT (1<<2) -#define DEPTH_FRMT_24_FIXED_8_OTHER (2<<2) -#define VERT_LINE_STRIDE_1 (1<<1) -#define VERT_LINE_STRIDE_0 (0<<1) -#define VERT_LINE_STRIDE_OFS_1 1 -#define VERT_LINE_STRIDE_OFS_0 0 - -/* p166 */ -#define _3DSTATE_DRAW_RECT_CMD (CMD_3D|(0x1d<<24)|(0x80<<16)|3) -/* Dword 1 */ -#define DRAW_RECT_DIS_DEPTH_OFS (1<<30) -#define DRAW_DITHER_OFS_X(x) ((x)<<26) -#define DRAW_DITHER_OFS_Y(x) ((x)<<24) -/* Dword 2 */ -#define DRAW_YMIN(x) ((x)<<16) -#define DRAW_XMIN(x) (x) -/* Dword 3 */ -#define DRAW_YMAX(x) ((x)<<16) -#define DRAW_XMAX(x) (x) -/* Dword 4 */ -#define DRAW_YORG(x) ((x)<<16) -#define DRAW_XORG(x) (x) - - -/* 3DSTATE_FILTER_COEFFICIENTS_4X4, p170 */ - -/* 3DSTATE_FILTER_COEFFICIENTS_6X5, p172 */ - - -/* _3DSTATE_FOG_COLOR, p173 */ -#define _3DSTATE_FOG_COLOR_CMD (CMD_3D|(0x15<<24)) -#define FOG_COLOR_RED(x) ((x)<<16) -#define FOG_COLOR_GREEN(x) ((x)<<8) -#define FOG_COLOR_BLUE(x) (x) - -/* _3DSTATE_FOG_MODE, p174 */ -#define _3DSTATE_FOG_MODE_CMD (CMD_3D|(0x1d<<24)|(0x89<<16)|2) -/* Dword 1 */ -#define FMC1_FOGFUNC_MODIFY_ENABLE (1<<31) -#define FMC1_FOGFUNC_VERTEX (0<<28) -#define FMC1_FOGFUNC_PIXEL_EXP (1<<28) -#define FMC1_FOGFUNC_PIXEL_EXP2 (2<<28) -#define FMC1_FOGFUNC_PIXEL_LINEAR (3<<28) -#define FMC1_FOGFUNC_MASK (3<<28) -#define FMC1_FOGINDEX_MODIFY_ENABLE (1<<27) -#define FMC1_FOGINDEX_Z (0<<25) -#define FMC1_FOGINDEX_W (1<<25) -#define FMC1_C1_C2_MODIFY_ENABLE (1<<24) -#define FMC1_DENSITY_MODIFY_ENABLE (1<<23) -#define FMC1_C1_ONE (1<<13) -#define FMC1_C1_MASK (0xffff<<4) -/* Dword 2 */ -#define FMC2_C2_ONE (1<<16) -/* Dword 3 */ -#define FMC3_D_ONE (1<<16) - - - -/* _3DSTATE_INDEPENDENT_ALPHA_BLEND, p177 */ -#define _3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD (CMD_3D|(0x0b<<24)) -#define IAB_MODIFY_ENABLE (1<<23) -#define IAB_ENABLE (1<<22) -#define IAB_MODIFY_FUNC (1<<21) -#define IAB_FUNC_SHIFT 16 -#define IAB_MODIFY_SRC_FACTOR (1<<11) -#define IAB_SRC_FACTOR_SHIFT 6 -#define IAB_SRC_FACTOR_MASK (BLENDFACT_MASK<<6) -#define IAB_MODIFY_DST_FACTOR (1<<5) -#define IAB_DST_FACTOR_SHIFT 0 -#define IAB_DST_FACTOR_MASK (BLENDFACT_MASK<<0) - - -#define BLENDFUNC_ADD 0x0 -#define BLENDFUNC_SUBTRACT 0x1 -#define BLENDFUNC_REVERSE_SUBTRACT 0x2 -#define BLENDFUNC_MIN 0x3 -#define BLENDFUNC_MAX 0x4 -#define BLENDFUNC_MASK 0x7 - -/* 3DSTATE_LOAD_INDIRECT, p180 */ - -#define _3DSTATE_LOAD_INDIRECT (CMD_3D|(0x1d<<24)|(0x7<<16)) -#define LI0_STATE_STATIC_INDIRECT (0x01<<8) -#define LI0_STATE_DYNAMIC_INDIRECT (0x02<<8) -#define LI0_STATE_SAMPLER (0x04<<8) -#define LI0_STATE_MAP (0x08<<8) -#define LI0_STATE_PROGRAM (0x10<<8) -#define LI0_STATE_CONSTANTS (0x20<<8) - -#define SIS0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define SIS0_FORCE_LOAD (1<<1) -#define SIS0_BUFFER_VALID (1<<0) -#define SIS1_BUFFER_LENGTH(x) ((x)&0xff) - -#define DIS0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define DIS0_BUFFER_RESET (1<<1) -#define DIS0_BUFFER_VALID (1<<0) - -#define SSB0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define SSB0_FORCE_LOAD (1<<1) -#define SSB0_BUFFER_VALID (1<<0) -#define SSB1_BUFFER_LENGTH(x) ((x)&0xff) - -#define MSB0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define MSB0_FORCE_LOAD (1<<1) -#define MSB0_BUFFER_VALID (1<<0) -#define MSB1_BUFFER_LENGTH(x) ((x)&0xff) - -#define PSP0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define PSP0_FORCE_LOAD (1<<1) -#define PSP0_BUFFER_VALID (1<<0) -#define PSP1_BUFFER_LENGTH(x) ((x)&0xff) - -#define PSC0_BUFFER_ADDRESS(x) ((x)&~0x3) -#define PSC0_FORCE_LOAD (1<<1) -#define PSC0_BUFFER_VALID (1<<0) -#define PSC1_BUFFER_LENGTH(x) ((x)&0xff) - - - - - -/* _3DSTATE_RASTERIZATION_RULES */ -#define _3DSTATE_RASTER_RULES_CMD (CMD_3D|(0x07<<24)) -#define ENABLE_POINT_RASTER_RULE (1<<15) -#define OGL_POINT_RASTER_RULE (1<<13) -#define ENABLE_TEXKILL_3D_4D (1<<10) -#define TEXKILL_3D (0<<9) -#define TEXKILL_4D (1<<9) -#define ENABLE_LINE_STRIP_PROVOKE_VRTX (1<<8) -#define ENABLE_TRI_FAN_PROVOKE_VRTX (1<<5) -#define LINE_STRIP_PROVOKE_VRTX(x) ((x)<<6) -#define TRI_FAN_PROVOKE_VRTX(x) ((x)<<3) - -/* _3DSTATE_SCISSOR_ENABLE, p256 */ -#define _3DSTATE_SCISSOR_ENABLE_CMD (CMD_3D|(0x1c<<24)|(0x10<<19)) -#define ENABLE_SCISSOR_RECT ((1<<1) | 1) -#define DISABLE_SCISSOR_RECT (1<<1) - -/* _3DSTATE_SCISSOR_RECTANGLE_0, p257 */ -#define _3DSTATE_SCISSOR_RECT_0_CMD (CMD_3D|(0x1d<<24)|(0x81<<16)|1) -/* Dword 1 */ -#define SCISSOR_RECT_0_YMIN(x) ((x)<<16) -#define SCISSOR_RECT_0_XMIN(x) (x) -/* Dword 2 */ -#define SCISSOR_RECT_0_YMAX(x) ((x)<<16) -#define SCISSOR_RECT_0_XMAX(x) (x) - -/* p189 */ -#define _3DSTATE_LOAD_STATE_IMMEDIATE_1 ((0x3<<29)|(0x1d<<24)|(0x04<<16)) -#define I1_LOAD_S(n) (1<<(4+n)) - -#define S0_VB_OFFSET_MASK 0xffffffc -#define S0_AUTO_CACHE_INV_DISABLE (1<<0) - -#define S1_VERTEX_WIDTH_SHIFT 24 -#define S1_VERTEX_WIDTH_MASK (0x3f<<24) -#define S1_VERTEX_PITCH_SHIFT 16 -#define S1_VERTEX_PITCH_MASK (0x3f<<16) - -#define TEXCOORDFMT_2D 0x0 -#define TEXCOORDFMT_3D 0x1 -#define TEXCOORDFMT_4D 0x2 -#define TEXCOORDFMT_1D 0x3 -#define TEXCOORDFMT_2D_16 0x4 -#define TEXCOORDFMT_4D_16 0x5 -#define TEXCOORDFMT_NOT_PRESENT 0xf -#define S2_TEXCOORD_FMT0_MASK 0xf -#define S2_TEXCOORD_FMT1_SHIFT 4 -#define S2_TEXCOORD_FMT(unit, type) ((type)<<(unit*4)) -#define S2_TEXCOORD_NONE (~0) - -/* S3 not interesting */ - -#define S4_POINT_WIDTH_SHIFT 23 -#define S4_POINT_WIDTH_MASK (0x1ff<<23) -#define S4_LINE_WIDTH_SHIFT 19 -#define S4_LINE_WIDTH_ONE (0x2<<19) -#define S4_LINE_WIDTH_MASK (0xf<<19) -#define S4_FLATSHADE_ALPHA (1<<18) -#define S4_FLATSHADE_FOG (1<<17) -#define S4_FLATSHADE_SPECULAR (1<<16) -#define S4_FLATSHADE_COLOR (1<<15) -#define S4_CULLMODE_BOTH (0<<13) -#define S4_CULLMODE_NONE (1<<13) -#define S4_CULLMODE_CW (2<<13) -#define S4_CULLMODE_CCW (3<<13) -#define S4_CULLMODE_MASK (3<<13) -#define S4_VFMT_POINT_WIDTH (1<<12) -#define S4_VFMT_SPEC_FOG (1<<11) -#define S4_VFMT_COLOR (1<<10) -#define S4_VFMT_DEPTH_OFFSET (1<<9) -#define S4_VFMT_XYZ (1<<6) -#define S4_VFMT_XYZW (2<<6) -#define S4_VFMT_XY (3<<6) -#define S4_VFMT_XYW (4<<6) -#define S4_VFMT_XYZW_MASK (7<<6) -#define S4_FORCE_DEFAULT_DIFFUSE (1<<5) -#define S4_FORCE_DEFAULT_SPECULAR (1<<4) -#define S4_LOCAL_DEPTH_OFFSET_ENABLE (1<<3) -#define S4_VFMT_FOG_PARAM (1<<2) -#define S4_SPRITE_POINT_ENABLE (1<<1) -#define S4_LINE_ANTIALIAS_ENABLE (1<<0) - -#define S4_VFMT_MASK (S4_VFMT_POINT_WIDTH | \ - S4_VFMT_SPEC_FOG | \ - S4_VFMT_COLOR | \ - S4_VFMT_DEPTH_OFFSET | \ - S4_VFMT_XYZW_MASK | \ - S4_VFMT_FOG_PARAM) - - -#define S5_WRITEDISABLE_ALPHA (1<<31) -#define S5_WRITEDISABLE_RED (1<<30) -#define S5_WRITEDISABLE_GREEN (1<<29) -#define S5_WRITEDISABLE_BLUE (1<<28) -#define S5_WRITEDISABLE_MASK (0xf<<28) -#define S5_FORCE_DEFAULT_POINT_SIZE (1<<27) -#define S5_LAST_PIXEL_ENABLE (1<<26) -#define S5_GLOBAL_DEPTH_OFFSET_ENABLE (1<<25) -#define S5_FOG_ENABLE (1<<24) -#define S5_STENCIL_REF_SHIFT 16 -#define S5_STENCIL_REF_MASK (0xff<<16) -#define S5_STENCIL_TEST_FUNC_SHIFT 13 -#define S5_STENCIL_TEST_FUNC_MASK (0x7<<13) -#define S5_STENCIL_FAIL_SHIFT 10 -#define S5_STENCIL_FAIL_MASK (0x7<<10) -#define S5_STENCIL_PASS_Z_FAIL_SHIFT 7 -#define S5_STENCIL_PASS_Z_FAIL_MASK (0x7<<7) -#define S5_STENCIL_PASS_Z_PASS_SHIFT 4 -#define S5_STENCIL_PASS_Z_PASS_MASK (0x7<<4) -#define S5_STENCIL_WRITE_ENABLE (1<<3) -#define S5_STENCIL_TEST_ENABLE (1<<2) -#define S5_COLOR_DITHER_ENABLE (1<<1) -#define S5_LOGICOP_ENABLE (1<<0) - - -#define S6_ALPHA_TEST_ENABLE (1<<31) -#define S6_ALPHA_TEST_FUNC_SHIFT 28 -#define S6_ALPHA_TEST_FUNC_MASK (0x7<<28) -#define S6_ALPHA_REF_SHIFT 20 -#define S6_ALPHA_REF_MASK (0xff<<20) -#define S6_DEPTH_TEST_ENABLE (1<<19) -#define S6_DEPTH_TEST_FUNC_SHIFT 16 -#define S6_DEPTH_TEST_FUNC_MASK (0x7<<16) -#define S6_CBUF_BLEND_ENABLE (1<<15) -#define S6_CBUF_BLEND_FUNC_SHIFT 12 -#define S6_CBUF_BLEND_FUNC_MASK (0x7<<12) -#define S6_CBUF_SRC_BLEND_FACT_SHIFT 8 -#define S6_CBUF_SRC_BLEND_FACT_MASK (0xf<<8) -#define S6_CBUF_DST_BLEND_FACT_SHIFT 4 -#define S6_CBUF_DST_BLEND_FACT_MASK (0xf<<4) -#define S6_DEPTH_WRITE_ENABLE (1<<3) -#define S6_COLOR_WRITE_ENABLE (1<<2) -#define S6_TRISTRIP_PV_SHIFT 0 -#define S6_TRISTRIP_PV_MASK (0x3<<0) - -#define S7_DEPTH_OFFSET_CONST_MASK ~0 - - - -#define DST_BLND_FACT(f) ((f)<= 0.0) ? src1 : src2 */ -#define A0_MIN (0xe<<24) /* dst = (src0 < src1) ? src0 : src1 */ -#define A0_MAX (0xf<<24) /* dst = (src0 >= src1) ? src0 : src1 */ -#define A0_FLR (0x10<<24) /* dst = floor(src0) */ -#define A0_MOD (0x11<<24) /* dst = src0 fmod 1.0 */ -#define A0_TRC (0x12<<24) /* dst = int(src0) */ -#define A0_SGE (0x13<<24) /* dst = src0 >= src1 ? 1.0 : 0.0 */ -#define A0_SLT (0x14<<24) /* dst = src0 < src1 ? 1.0 : 0.0 */ -#define A0_DEST_SATURATE (1<<22) -#define A0_DEST_TYPE_SHIFT 19 -/* Allow: R, OC, OD, U */ -#define A0_DEST_NR_SHIFT 14 -/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ -#define A0_DEST_CHANNEL_X (1<<10) -#define A0_DEST_CHANNEL_Y (2<<10) -#define A0_DEST_CHANNEL_Z (4<<10) -#define A0_DEST_CHANNEL_W (8<<10) -#define A0_DEST_CHANNEL_ALL (0xf<<10) -#define A0_DEST_CHANNEL_SHIFT 10 -#define A0_SRC0_TYPE_SHIFT 7 -#define A0_SRC0_NR_SHIFT 2 - -#define A0_DEST_CHANNEL_XY (A0_DEST_CHANNEL_X|A0_DEST_CHANNEL_Y) -#define A0_DEST_CHANNEL_XYZ (A0_DEST_CHANNEL_XY|A0_DEST_CHANNEL_Z) - - -#define SRC_X 0 -#define SRC_Y 1 -#define SRC_Z 2 -#define SRC_W 3 -#define SRC_ZERO 4 -#define SRC_ONE 5 - -#define A1_SRC0_CHANNEL_X_NEGATE (1<<31) -#define A1_SRC0_CHANNEL_X_SHIFT 28 -#define A1_SRC0_CHANNEL_Y_NEGATE (1<<27) -#define A1_SRC0_CHANNEL_Y_SHIFT 24 -#define A1_SRC0_CHANNEL_Z_NEGATE (1<<23) -#define A1_SRC0_CHANNEL_Z_SHIFT 20 -#define A1_SRC0_CHANNEL_W_NEGATE (1<<19) -#define A1_SRC0_CHANNEL_W_SHIFT 16 -#define A1_SRC1_TYPE_SHIFT 13 -#define A1_SRC1_NR_SHIFT 8 -#define A1_SRC1_CHANNEL_X_NEGATE (1<<7) -#define A1_SRC1_CHANNEL_X_SHIFT 4 -#define A1_SRC1_CHANNEL_Y_NEGATE (1<<3) -#define A1_SRC1_CHANNEL_Y_SHIFT 0 - -#define A2_SRC1_CHANNEL_Z_NEGATE (1<<31) -#define A2_SRC1_CHANNEL_Z_SHIFT 28 -#define A2_SRC1_CHANNEL_W_NEGATE (1<<27) -#define A2_SRC1_CHANNEL_W_SHIFT 24 -#define A2_SRC2_TYPE_SHIFT 21 -#define A2_SRC2_NR_SHIFT 16 -#define A2_SRC2_CHANNEL_X_NEGATE (1<<15) -#define A2_SRC2_CHANNEL_X_SHIFT 12 -#define A2_SRC2_CHANNEL_Y_NEGATE (1<<11) -#define A2_SRC2_CHANNEL_Y_SHIFT 8 -#define A2_SRC2_CHANNEL_Z_NEGATE (1<<7) -#define A2_SRC2_CHANNEL_Z_SHIFT 4 -#define A2_SRC2_CHANNEL_W_NEGATE (1<<3) -#define A2_SRC2_CHANNEL_W_SHIFT 0 - - - -/* Texture instructions */ -#define T0_TEXLD (0x15<<24) /* Sample texture using predeclared - * sampler and address, and output - * filtered texel data to destination - * register */ -#define T0_TEXLDP (0x16<<24) /* Same as texld but performs a - * perspective divide of the texture - * coordinate .xyz values by .w before - * sampling. */ -#define T0_TEXLDB (0x17<<24) /* Same as texld but biases the - * computed LOD by w. Only S4.6 two's - * comp is used. This implies that a - * float to fixed conversion is - * done. */ -#define T0_TEXKILL (0x18<<24) /* Does not perform a sampling - * operation. Simply kills the pixel - * if any channel of the address - * register is < 0.0. */ -#define T0_DEST_TYPE_SHIFT 19 -/* Allow: R, OC, OD, U */ -/* Note: U (unpreserved) regs do not retain their values between - * phases (cannot be used for feedback) - * - * Note: oC and OD registers can only be used as the destination of a - * texture instruction once per phase (this is an implementation - * restriction). - */ -#define T0_DEST_NR_SHIFT 14 -/* Allow R: 0..15, OC,OD: 0..0, U: 0..2 */ -#define T0_SAMPLER_NR_SHIFT 0 /* This field ignored for TEXKILL */ -#define T0_SAMPLER_NR_MASK (0xf<<0) - -#define T1_ADDRESS_REG_TYPE_SHIFT 24 /* Reg to use as texture coord */ -/* Allow R, T, OC, OD -- R, OC, OD are 'dependent' reads, new program phase */ -#define T1_ADDRESS_REG_NR_SHIFT 17 -#define T2_MBZ 0 - -/* Declaration instructions */ -#define D0_DCL (0x19<<24) /* Declare a t (interpolated attrib) - * register or an s (sampler) - * register. */ -#define D0_SAMPLE_TYPE_SHIFT 22 -#define D0_SAMPLE_TYPE_2D (0x0<<22) -#define D0_SAMPLE_TYPE_CUBE (0x1<<22) -#define D0_SAMPLE_TYPE_VOLUME (0x2<<22) -#define D0_SAMPLE_TYPE_MASK (0x3<<22) - -#define D0_TYPE_SHIFT 19 -/* Allow: T, S */ -#define D0_NR_SHIFT 14 -/* Allow T: 0..10, S: 0..15 */ -#define D0_CHANNEL_X (1<<10) -#define D0_CHANNEL_Y (2<<10) -#define D0_CHANNEL_Z (4<<10) -#define D0_CHANNEL_W (8<<10) -#define D0_CHANNEL_ALL (0xf<<10) -#define D0_CHANNEL_NONE (0<<10) - -#define D0_CHANNEL_XY (D0_CHANNEL_X|D0_CHANNEL_Y) -#define D0_CHANNEL_XYZ (D0_CHANNEL_XY|D0_CHANNEL_Z) - -/* I915 Errata: Do not allow (xz), (xw), (xzw) combinations for diffuse - * or specular declarations. - * - * For T dcls, only allow: (x), (xy), (xyz), (w), (xyzw) - * - * Must be zero for S (sampler) dcls - */ -#define D1_MBZ 0 -#define D2_MBZ 0 - - - -/* p207 */ -#define _3DSTATE_MAP_STATE (CMD_3D|(0x1d<<24)|(0x0<<16)) - -#define MS1_MAPMASK_SHIFT 0 -#define MS1_MAPMASK_MASK (0x8fff<<0) - -#define MS2_UNTRUSTED_SURFACE (1<<31) -#define MS2_ADDRESS_MASK 0xfffffffc -#define MS2_VERTICAL_LINE_STRIDE (1<<1) -#define MS2_VERTICAL_OFFSET (1<<1) - -#define MS3_HEIGHT_SHIFT 21 -#define MS3_WIDTH_SHIFT 10 -#define MS3_PALETTE_SELECT (1<<9) -#define MS3_MAPSURF_FORMAT_SHIFT 7 -#define MS3_MAPSURF_FORMAT_MASK (0x7<<7) -#define MAPSURF_8BIT (1<<7) -#define MAPSURF_16BIT (2<<7) -#define MAPSURF_32BIT (3<<7) -#define MAPSURF_422 (5<<7) -#define MAPSURF_COMPRESSED (6<<7) -#define MAPSURF_4BIT_INDEXED (7<<7) -#define MS3_MT_FORMAT_MASK (0x7 << 3) -#define MS3_MT_FORMAT_SHIFT 3 -#define MT_4BIT_IDX_ARGB8888 (7<<3) /* SURFACE_4BIT_INDEXED */ -#define MT_8BIT_I8 (0<<3) /* SURFACE_8BIT */ -#define MT_8BIT_L8 (1<<3) -#define MT_8BIT_A8 (4<<3) -#define MT_8BIT_MONO8 (5<<3) -#define MT_16BIT_RGB565 (0<<3) /* SURFACE_16BIT */ -#define MT_16BIT_ARGB1555 (1<<3) -#define MT_16BIT_ARGB4444 (2<<3) -#define MT_16BIT_AY88 (3<<3) -#define MT_16BIT_88DVDU (5<<3) -#define MT_16BIT_BUMP_655LDVDU (6<<3) -#define MT_16BIT_I16 (7<<3) -#define MT_16BIT_L16 (8<<3) -#define MT_16BIT_A16 (9<<3) -#define MT_32BIT_ARGB8888 (0<<3) /* SURFACE_32BIT */ -#define MT_32BIT_ABGR8888 (1<<3) -#define MT_32BIT_XRGB8888 (2<<3) -#define MT_32BIT_XBGR8888 (3<<3) -#define MT_32BIT_QWVU8888 (4<<3) -#define MT_32BIT_AXVU8888 (5<<3) -#define MT_32BIT_LXVU8888 (6<<3) -#define MT_32BIT_XLVU8888 (7<<3) -#define MT_32BIT_ARGB2101010 (8<<3) -#define MT_32BIT_ABGR2101010 (9<<3) -#define MT_32BIT_AWVU2101010 (0xA<<3) -#define MT_32BIT_GR1616 (0xB<<3) -#define MT_32BIT_VU1616 (0xC<<3) -#define MT_32BIT_xI824 (0xD<<3) -#define MT_32BIT_xA824 (0xE<<3) -#define MT_32BIT_xL824 (0xF<<3) -#define MT_422_YCRCB_SWAPY (0<<3) /* SURFACE_422 */ -#define MT_422_YCRCB_NORMAL (1<<3) -#define MT_422_YCRCB_SWAPUV (2<<3) -#define MT_422_YCRCB_SWAPUVY (3<<3) -#define MT_COMPRESS_DXT1 (0<<3) /* SURFACE_COMPRESSED */ -#define MT_COMPRESS_DXT2_3 (1<<3) -#define MT_COMPRESS_DXT4_5 (2<<3) -#define MT_COMPRESS_FXT1 (3<<3) -#define MT_COMPRESS_DXT1_RGB (4<<3) -#define MS3_USE_FENCE_REGS (1<<2) -#define MS3_TILED_SURFACE (1<<1) -#define MS3_TILE_WALK (1<<0) - -#define MS4_PITCH_SHIFT 21 -#define MS4_CUBE_FACE_ENA_NEGX (1<<20) -#define MS4_CUBE_FACE_ENA_POSX (1<<19) -#define MS4_CUBE_FACE_ENA_NEGY (1<<18) -#define MS4_CUBE_FACE_ENA_POSY (1<<17) -#define MS4_CUBE_FACE_ENA_NEGZ (1<<16) -#define MS4_CUBE_FACE_ENA_POSZ (1<<15) -#define MS4_CUBE_FACE_ENA_MASK (0x3f<<15) -#define MS4_MAX_LOD_SHIFT 9 -#define MS4_MAX_LOD_MASK (0x3f<<9) -#define MS4_MIP_LAYOUT_LEGACY (0<<8) -#define MS4_MIP_LAYOUT_BELOW_LPT (0<<8) -#define MS4_MIP_LAYOUT_RIGHT_LPT (1<<8) -#define MS4_VOLUME_DEPTH_SHIFT 0 -#define MS4_VOLUME_DEPTH_MASK (0xff<<0) - -/* p244 */ -#define _3DSTATE_SAMPLER_STATE (CMD_3D|(0x1d<<24)|(0x1<<16)) - -#define SS1_MAPMASK_SHIFT 0 -#define SS1_MAPMASK_MASK (0x8fff<<0) - -#define SS2_REVERSE_GAMMA_ENABLE (1<<31) -#define SS2_PACKED_TO_PLANAR_ENABLE (1<<30) -#define SS2_COLORSPACE_CONVERSION (1<<29) -#define SS2_CHROMAKEY_SHIFT 27 -#define SS2_BASE_MIP_LEVEL_SHIFT 22 -#define SS2_BASE_MIP_LEVEL_MASK (0x1f<<22) -#define SS2_MIP_FILTER_SHIFT 20 -#define SS2_MIP_FILTER_MASK (0x3<<20) -#define MIPFILTER_NONE 0 -#define MIPFILTER_NEAREST 1 -#define MIPFILTER_LINEAR 3 -#define SS2_MAG_FILTER_SHIFT 17 -#define SS2_MAG_FILTER_MASK (0x7<<17) -#define FILTER_NEAREST 0 -#define FILTER_LINEAR 1 -#define FILTER_ANISOTROPIC 2 -#define FILTER_4X4_1 3 -#define FILTER_4X4_2 4 -#define FILTER_4X4_FLAT 5 -#define FILTER_6X5_MONO 6 /* XXX - check */ -#define SS2_MIN_FILTER_SHIFT 14 -#define SS2_MIN_FILTER_MASK (0x7<<14) -#define SS2_LOD_BIAS_SHIFT 5 -#define SS2_LOD_BIAS_ONE (0x10<<5) -#define SS2_LOD_BIAS_MASK (0x1ff<<5) -/* Shadow requires: - * MT_X8{I,L,A}24 or MT_{I,L,A}16 texture format - * FILTER_4X4_x MIN and MAG filters - */ -#define SS2_SHADOW_ENABLE (1<<4) -#define SS2_MAX_ANISO_MASK (1<<3) -#define SS2_MAX_ANISO_2 (0<<3) -#define SS2_MAX_ANISO_4 (1<<3) -#define SS2_SHADOW_FUNC_SHIFT 0 -#define SS2_SHADOW_FUNC_MASK (0x7<<0) -/* SS2_SHADOW_FUNC values: see COMPAREFUNC_* */ - -#define SS3_MIN_LOD_SHIFT 24 -#define SS3_MIN_LOD_ONE (0x10<<24) -#define SS3_MIN_LOD_MASK (0xff<<24) -#define SS3_KILL_PIXEL_ENABLE (1<<17) -#define SS3_TCX_ADDR_MODE_SHIFT 12 -#define SS3_TCX_ADDR_MODE_MASK (0x7<<12) -#define TEXCOORDMODE_WRAP 0 -#define TEXCOORDMODE_MIRROR 1 -#define TEXCOORDMODE_CLAMP_EDGE 2 -#define TEXCOORDMODE_CUBE 3 -#define TEXCOORDMODE_CLAMP_BORDER 4 -#define TEXCOORDMODE_MIRROR_ONCE 5 -#define SS3_TCY_ADDR_MODE_SHIFT 9 -#define SS3_TCY_ADDR_MODE_MASK (0x7<<9) -#define SS3_TCZ_ADDR_MODE_SHIFT 6 -#define SS3_TCZ_ADDR_MODE_MASK (0x7<<6) -#define SS3_NORMALIZED_COORDS (1<<5) -#define SS3_TEXTUREMAP_INDEX_SHIFT 1 -#define SS3_TEXTUREMAP_INDEX_MASK (0xf<<1) -#define SS3_DEINTERLACER_ENABLE (1<<0) - -#define SS4_BORDER_COLOR_MASK (~0) - -/* 3DSTATE_SPAN_STIPPLE, p258 - */ -#define _3DSTATE_STIPPLE ((0x3<<29)|(0x1d<<24)|(0x83<<16)) -#define ST1_ENABLE (1<<16) -#define ST1_MASK (0xffff) - -#define _3DSTATE_DEFAULT_Z ((0x3<<29)|(0x1d<<24)|(0x98<<16)) -#define _3DSTATE_DEFAULT_DIFFUSE ((0x3<<29)|(0x1d<<24)|(0x99<<16)) -#define _3DSTATE_DEFAULT_SPECULAR ((0x3<<29)|(0x1d<<24)|(0x9a<<16)) - - -#define MI_FLUSH ((0<<29)|(4<<23)) -#define FLUSH_MAP_CACHE (1<<0) -#define INHIBIT_FLUSH_RENDER_CACHE (1<<2) - - -#define CMD_3D (0x3<<29) - - -#define _3DPRIMITIVE ((0x3<<29)|(0x1f<<24)) -#define PRIM_INDIRECT (1<<23) -#define PRIM_INLINE (0<<23) -#define PRIM_INDIRECT_SEQUENTIAL (0<<17) -#define PRIM_INDIRECT_ELTS (1<<17) - -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_MASK (0x1f<<18) - -#define I915PACKCOLOR4444(r,g,b,a) \ - ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) - -#define I915PACKCOLOR1555(r,g,b,a) \ - ((((r) & 0xf8) << 7) | (((g) & 0xf8) << 2) | (((b) & 0xf8) >> 3) | \ - ((a) ? 0x8000 : 0)) - -#define I915PACKCOLOR565(r,g,b) \ - ((((r) & 0xf8) << 8) | (((g) & 0xfc) << 3) | (((b) & 0xf8) >> 3)) - -#define I915PACKCOLOR8888(r,g,b,a) \ - ((a<<24) | (r<<16) | (g<<8) | b) - - - - -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 - -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) - -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) - -#define MI_WAIT_FOR_EVENT ((0x3<<23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) - -#define MI_BATCH_BUFFER (0x30<<23) -#define MI_BATCH_BUFFER_START (0x31<<23) -#define MI_BATCH_BUFFER_END (0xa<<23) - - - -#define COMPAREFUNC_ALWAYS 0 -#define COMPAREFUNC_NEVER 0x1 -#define COMPAREFUNC_LESS 0x2 -#define COMPAREFUNC_EQUAL 0x3 -#define COMPAREFUNC_LEQUAL 0x4 -#define COMPAREFUNC_GREATER 0x5 -#define COMPAREFUNC_NOTEQUAL 0x6 -#define COMPAREFUNC_GEQUAL 0x7 - -#define STENCILOP_KEEP 0 -#define STENCILOP_ZERO 0x1 -#define STENCILOP_REPLACE 0x2 -#define STENCILOP_INCRSAT 0x3 -#define STENCILOP_DECRSAT 0x4 -#define STENCILOP_INCR 0x5 -#define STENCILOP_DECR 0x6 -#define STENCILOP_INVERT 0x7 - -#define LOGICOP_CLEAR 0 -#define LOGICOP_NOR 0x1 -#define LOGICOP_AND_INV 0x2 -#define LOGICOP_COPY_INV 0x3 -#define LOGICOP_AND_RVRSE 0x4 -#define LOGICOP_INV 0x5 -#define LOGICOP_XOR 0x6 -#define LOGICOP_NAND 0x7 -#define LOGICOP_AND 0x8 -#define LOGICOP_EQUIV 0x9 -#define LOGICOP_NOOP 0xa -#define LOGICOP_OR_INV 0xb -#define LOGICOP_COPY 0xc -#define LOGICOP_OR_RVRSE 0xd -#define LOGICOP_OR 0xe -#define LOGICOP_SET 0xf - -#define BLENDFACT_ZERO 0x01 -#define BLENDFACT_ONE 0x02 -#define BLENDFACT_SRC_COLR 0x03 -#define BLENDFACT_INV_SRC_COLR 0x04 -#define BLENDFACT_SRC_ALPHA 0x05 -#define BLENDFACT_INV_SRC_ALPHA 0x06 -#define BLENDFACT_DST_ALPHA 0x07 -#define BLENDFACT_INV_DST_ALPHA 0x08 -#define BLENDFACT_DST_COLR 0x09 -#define BLENDFACT_INV_DST_COLR 0x0a -#define BLENDFACT_SRC_ALPHA_SATURATE 0x0b -#define BLENDFACT_CONST_COLOR 0x0c -#define BLENDFACT_INV_CONST_COLOR 0x0d -#define BLENDFACT_CONST_ALPHA 0x0e -#define BLENDFACT_INV_CONST_ALPHA 0x0f -#define BLENDFACT_MASK 0x0f - -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GME 0x27AE -#define PCI_CHIP_G33_G 0x29C2 -#define PCI_CHIP_Q35_G 0x29B2 -#define PCI_CHIP_Q33_G 0x29D2 - - -#endif diff --git a/src/mesa/pipe/i915simple/i915_state.c b/src/mesa/pipe/i915simple/i915_state.c deleted file mode 100644 index abd5571b88..0000000000 --- a/src/mesa/pipe/i915simple/i915_state.c +++ /dev/null @@ -1,694 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - - -#include "pipe/draw/draw_context.h" -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" - -#include "i915_context.h" -#include "i915_reg.h" -#include "i915_state.h" -#include "i915_state_inlines.h" - - -/* The i915 (and related graphics cores) do not support GL_CLAMP. The - * Intel drivers for "other operating systems" implement GL_CLAMP as - * GL_CLAMP_TO_EDGE, so the same is done here. - */ -static unsigned -translate_wrap_mode(unsigned wrap) -{ - switch (wrap) { - case PIPE_TEX_WRAP_REPEAT: - return TEXCOORDMODE_WRAP; - case PIPE_TEX_WRAP_CLAMP: - return TEXCOORDMODE_CLAMP_EDGE; /* not quite correct */ - case PIPE_TEX_WRAP_CLAMP_TO_EDGE: - return TEXCOORDMODE_CLAMP_EDGE; - case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - return TEXCOORDMODE_CLAMP_BORDER; -// case PIPE_TEX_WRAP_MIRRORED_REPEAT: -// return TEXCOORDMODE_MIRROR; - default: - return TEXCOORDMODE_WRAP; - } -} - -static unsigned translate_img_filter( unsigned filter ) -{ - switch (filter) { - case PIPE_TEX_FILTER_NEAREST: - return FILTER_NEAREST; - case PIPE_TEX_FILTER_LINEAR: - return FILTER_LINEAR; - default: - assert(0); - return FILTER_NEAREST; - } -} - -static unsigned translate_mip_filter( unsigned filter ) -{ - switch (filter) { - case PIPE_TEX_MIPFILTER_NONE: - return MIPFILTER_NONE; - case PIPE_TEX_MIPFILTER_NEAREST: - return MIPFILTER_NEAREST; - case PIPE_TEX_FILTER_LINEAR: - return MIPFILTER_LINEAR; - default: - assert(0); - return MIPFILTER_NONE; - } -} - - -/* None of this state is actually used for anything yet. - */ -static void * -i915_create_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend) -{ - struct i915_blend_state *cso_data = CALLOC_STRUCT( i915_blend_state ); - - { - unsigned eqRGB = blend->rgb_func; - unsigned srcRGB = blend->rgb_src_factor; - unsigned dstRGB = blend->rgb_dst_factor; - - unsigned eqA = blend->alpha_func; - unsigned srcA = blend->alpha_src_factor; - unsigned dstA = blend->alpha_dst_factor; - - /* Special handling for MIN/MAX filter modes handled at - * state_tracker level. - */ - - if (srcA != srcRGB || - dstA != dstRGB || - eqA != eqRGB) { - - cso_data->iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | - IAB_MODIFY_ENABLE | - IAB_ENABLE | - IAB_MODIFY_FUNC | - IAB_MODIFY_SRC_FACTOR | - IAB_MODIFY_DST_FACTOR | - SRC_ABLND_FACT(i915_translate_blend_factor(srcA)) | - DST_ABLND_FACT(i915_translate_blend_factor(dstA)) | - (i915_translate_blend_func(eqA) << IAB_FUNC_SHIFT)); - } - else { - cso_data->iab = (_3DSTATE_INDEPENDENT_ALPHA_BLEND_CMD | - IAB_MODIFY_ENABLE | - 0); - } - } - - cso_data->modes4 |= (_3DSTATE_MODES_4_CMD | - ENABLE_LOGIC_OP_FUNC | - LOGIC_OP_FUNC(i915_translate_logic_op(blend->logicop_func))); - - if (blend->logicop_enable) - cso_data->LIS5 |= S5_LOGICOP_ENABLE; - - if (blend->dither) - cso_data->LIS5 |= S5_COLOR_DITHER_ENABLE; - - if ((blend->colormask & PIPE_MASK_R) == 0) - cso_data->LIS5 |= S5_WRITEDISABLE_RED; - - if ((blend->colormask & PIPE_MASK_G) == 0) - cso_data->LIS5 |= S5_WRITEDISABLE_GREEN; - - if ((blend->colormask & PIPE_MASK_B) == 0) - cso_data->LIS5 |= S5_WRITEDISABLE_BLUE; - - if ((blend->colormask & PIPE_MASK_A) == 0) - cso_data->LIS5 |= S5_WRITEDISABLE_ALPHA; - - if (blend->blend_enable) { - unsigned funcRGB = blend->rgb_func; - unsigned srcRGB = blend->rgb_src_factor; - unsigned dstRGB = blend->rgb_dst_factor; - - cso_data->LIS6 |= (S6_CBUF_BLEND_ENABLE | - SRC_BLND_FACT(i915_translate_blend_factor(srcRGB)) | - DST_BLND_FACT(i915_translate_blend_factor(dstRGB)) | - (i915_translate_blend_func(funcRGB) << S6_CBUF_BLEND_FUNC_SHIFT)); - } - - return cso_data; -} - -static void i915_bind_blend_state(struct pipe_context *pipe, - void *blend) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->blend = (struct i915_blend_state*)blend; - - i915->dirty |= I915_NEW_BLEND; -} - - -static void i915_delete_blend_state(struct pipe_context *pipe, void *blend) -{ - FREE(blend); -} - -static void i915_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->blend_color = *blend_color; - - i915->dirty |= I915_NEW_BLEND; -} - -static void * -i915_create_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) -{ - struct i915_sampler_state *cso = CALLOC_STRUCT( i915_sampler_state ); - const unsigned ws = sampler->wrap_s; - const unsigned wt = sampler->wrap_t; - const unsigned wr = sampler->wrap_r; - unsigned minFilt, magFilt; - unsigned mipFilt; - - cso->templ = sampler; - - mipFilt = translate_mip_filter(sampler->min_mip_filter); - if (sampler->max_anisotropy > 1.0) { - minFilt = FILTER_ANISOTROPIC; - magFilt = FILTER_ANISOTROPIC; - if (sampler->max_anisotropy > 2.0) { - cso->state[0] |= SS2_MAX_ANISO_4; - } - } - else { - minFilt = translate_img_filter( sampler->min_img_filter ); - magFilt = translate_img_filter( sampler->mag_img_filter ); - } - - { - int b = (int) (sampler->lod_bias * 16.0); - b = CLAMP(b, -256, 255); - cso->state[0] |= ((b << SS2_LOD_BIAS_SHIFT) & SS2_LOD_BIAS_MASK); - } - - /* Shadow: - */ - if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) - { - cso->state[0] |= (SS2_SHADOW_ENABLE | - i915_translate_compare_func(sampler->compare_func)); - - minFilt = FILTER_4X4_FLAT; - magFilt = FILTER_4X4_FLAT; - } - - cso->state[0] |= ((minFilt << SS2_MIN_FILTER_SHIFT) | - (mipFilt << SS2_MIP_FILTER_SHIFT) | - (magFilt << SS2_MAG_FILTER_SHIFT)); - - cso->state[1] |= - ((translate_wrap_mode(ws) << SS3_TCX_ADDR_MODE_SHIFT) | - (translate_wrap_mode(wt) << SS3_TCY_ADDR_MODE_SHIFT) | - (translate_wrap_mode(wr) << SS3_TCZ_ADDR_MODE_SHIFT)); - - if (sampler->normalized_coords) - cso->state[1] |= SS3_NORMALIZED_COORDS; - - if (0) /* XXX not tested yet */ - { - int minlod = (int) (16.0 * sampler->min_lod); - minlod = CLAMP(minlod, 0, 16 * 11); - cso->state[1] |= (minlod << SS3_MIN_LOD_SHIFT); - } - - { - ubyte r = float_to_ubyte(sampler->border_color[0]); - ubyte g = float_to_ubyte(sampler->border_color[1]); - ubyte b = float_to_ubyte(sampler->border_color[2]); - ubyte a = float_to_ubyte(sampler->border_color[3]); - cso->state[2] = I915PACKCOLOR8888(r, g, b, a); - } - return cso; -} - -static void i915_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) -{ - struct i915_context *i915 = i915_context(pipe); - - assert(unit < PIPE_MAX_SAMPLERS); - i915->sampler[unit] = (const struct i915_sampler_state*)sampler; - - i915->dirty |= I915_NEW_SAMPLER; -} - -static void i915_delete_sampler_state(struct pipe_context *pipe, - void *sampler) -{ - FREE(sampler); -} - - -/** XXX move someday? Or consolidate all these simple state setters - * into one file. - */ - -static void * -i915_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) -{ - struct i915_depth_stencil_state *cso = CALLOC_STRUCT( i915_depth_stencil_state ); - - { - int testmask = depth_stencil->stencil[0].value_mask & 0xff; - int writemask = depth_stencil->stencil[0].write_mask & 0xff; - - cso->stencil_modes4 |= (_3DSTATE_MODES_4_CMD | - ENABLE_STENCIL_TEST_MASK | - STENCIL_TEST_MASK(testmask) | - ENABLE_STENCIL_WRITE_MASK | - STENCIL_WRITE_MASK(writemask)); - } - - if (depth_stencil->stencil[0].enabled) { - int test = i915_translate_compare_func(depth_stencil->stencil[0].func); - int fop = i915_translate_stencil_op(depth_stencil->stencil[0].fail_op); - int dfop = i915_translate_stencil_op(depth_stencil->stencil[0].zfail_op); - int dpop = i915_translate_stencil_op(depth_stencil->stencil[0].zpass_op); - int ref = depth_stencil->stencil[0].ref_value & 0xff; - - cso->stencil_LIS5 |= (S5_STENCIL_TEST_ENABLE | - S5_STENCIL_WRITE_ENABLE | - (ref << S5_STENCIL_REF_SHIFT) | - (test << S5_STENCIL_TEST_FUNC_SHIFT) | - (fop << S5_STENCIL_FAIL_SHIFT) | - (dfop << S5_STENCIL_PASS_Z_FAIL_SHIFT) | - (dpop << S5_STENCIL_PASS_Z_PASS_SHIFT)); - } - - if (depth_stencil->stencil[1].enabled) { - int test = i915_translate_compare_func(depth_stencil->stencil[1].func); - int fop = i915_translate_stencil_op(depth_stencil->stencil[1].fail_op); - int dfop = i915_translate_stencil_op(depth_stencil->stencil[1].zfail_op); - int dpop = i915_translate_stencil_op(depth_stencil->stencil[1].zpass_op); - int ref = depth_stencil->stencil[1].ref_value & 0xff; - int tmask = depth_stencil->stencil[1].value_mask & 0xff; - int wmask = depth_stencil->stencil[1].write_mask & 0xff; - - cso->bfo[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | - BFO_ENABLE_STENCIL_FUNCS | - BFO_ENABLE_STENCIL_TWO_SIDE | - BFO_ENABLE_STENCIL_REF | - BFO_STENCIL_TWO_SIDE | - (ref << BFO_STENCIL_REF_SHIFT) | - (test << BFO_STENCIL_TEST_SHIFT) | - (fop << BFO_STENCIL_FAIL_SHIFT) | - (dfop << BFO_STENCIL_PASS_Z_FAIL_SHIFT) | - (dpop << BFO_STENCIL_PASS_Z_PASS_SHIFT)); - - cso->bfo[1] = (_3DSTATE_BACKFACE_STENCIL_MASKS | - BFM_ENABLE_STENCIL_TEST_MASK | - BFM_ENABLE_STENCIL_WRITE_MASK | - (tmask << BFM_STENCIL_TEST_MASK_SHIFT) | - (wmask << BFM_STENCIL_WRITE_MASK_SHIFT)); - } - else { - /* This actually disables two-side stencil: The bit set is a - * modify-enable bit to indicate we are changing the two-side - * setting. Then there is a symbolic zero to show that we are - * setting the flag to zero/off. - */ - cso->bfo[0] = (_3DSTATE_BACKFACE_STENCIL_OPS | - BFO_ENABLE_STENCIL_TWO_SIDE | - 0); - cso->bfo[1] = 0; - } - - if (depth_stencil->depth.enabled) { - int func = i915_translate_compare_func(depth_stencil->depth.func); - - cso->depth_LIS6 |= (S6_DEPTH_TEST_ENABLE | - (func << S6_DEPTH_TEST_FUNC_SHIFT)); - - if (depth_stencil->depth.writemask) - cso->depth_LIS6 |= S6_DEPTH_WRITE_ENABLE; - } - - if (depth_stencil->alpha.enabled) { - int test = i915_translate_compare_func(depth_stencil->alpha.func); - ubyte refByte = float_to_ubyte(depth_stencil->alpha.ref); - - cso->depth_LIS6 |= (S6_ALPHA_TEST_ENABLE | - (test << S6_ALPHA_TEST_FUNC_SHIFT) | - (((unsigned) refByte) << S6_ALPHA_REF_SHIFT)); - } - - return cso; -} - -static void i915_bind_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->depth_stencil = (const struct i915_depth_stencil_state *)depth_stencil; - - i915->dirty |= I915_NEW_DEPTH_STENCIL; -} - -static void i915_delete_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - FREE(depth_stencil); -} - - -static void i915_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct i915_context *i915 = i915_context(pipe); - - memcpy( &i915->scissor, scissor, sizeof(*scissor) ); - i915->dirty |= I915_NEW_SCISSOR; -} - - -static void i915_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ -} - - -static void * i915_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - return 0; -} - -static void i915_bind_fs_state(struct pipe_context *pipe, void *fs) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->fs = (struct pipe_shader_state *)fs; - - i915->dirty |= I915_NEW_FS; -} - -static void i915_delete_fs_state(struct pipe_context *pipe, void *shader) -{ - /*do nothing*/ -} - -static void * -i915_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - struct i915_context *i915 = i915_context(pipe); - - /* just pass-through to draw module */ - return draw_create_vertex_shader(i915->draw, templ); -} - -static void i915_bind_vs_state(struct pipe_context *pipe, void *shader) -{ - struct i915_context *i915 = i915_context(pipe); - - /* just pass-through to draw module */ - draw_bind_vertex_shader(i915->draw, (struct draw_vertex_shader *) shader); -} - -static void i915_delete_vs_state(struct pipe_context *pipe, void *shader) -{ - struct i915_context *i915 = i915_context(pipe); - - /* just pass-through to draw module */ - draw_delete_vertex_shader(i915->draw, (struct draw_vertex_shader *) shader); -} - -static void i915_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) -{ - struct i915_context *i915 = i915_context(pipe); - struct pipe_winsys *ws = pipe->winsys; - - assert(shader < PIPE_SHADER_TYPES); - assert(index == 0); - - /* Make a copy of shader constants. - * During fragment program translation we may add additional - * constants to the array. - * - * We want to consider the situation where some user constants - * (ex: a material color) may change frequently but the shader program - * stays the same. In that case we should only be updating the first - * N constants, leaving any extras from shader translation alone. - */ - { - void *mapped; - if (buf->size && - (mapped = ws->buffer_map(ws, buf->buffer, - PIPE_BUFFER_USAGE_CPU_READ))) { - memcpy(i915->current.constants[shader], mapped, buf->size); - ws->buffer_unmap(ws, buf->buffer); - i915->current.num_user_constants[shader] - = buf->size / (4 * sizeof(float)); - } - else { - i915->current.num_user_constants[shader] = 0; - } - } - - i915->dirty |= I915_NEW_CONSTANTS; -} - - -static void i915_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, - struct pipe_texture *texture) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->texture[sampler] = (struct i915_texture*)texture; /* ptr, not struct */ - - i915->dirty |= I915_NEW_TEXTURE; -} - - - -static void i915_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->framebuffer = *fb; /* struct copy */ - - i915->dirty |= I915_NEW_FRAMEBUFFER; -} - - - -static void i915_set_clip_state( struct pipe_context *pipe, - const struct pipe_clip_state *clip ) -{ - struct i915_context *i915 = i915_context(pipe); - - draw_set_clip_state(i915->draw, clip); - - i915->dirty |= I915_NEW_CLIP; -} - - - -/* Called when driver state tracker notices changes to the viewport - * matrix: - */ -static void i915_set_viewport_state( struct pipe_context *pipe, - const struct pipe_viewport_state *viewport ) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->viewport = *viewport; /* struct copy */ - - /* pass the viewport info to the draw module */ - draw_set_viewport_state(i915->draw, &i915->viewport); - - i915->dirty |= I915_NEW_VIEWPORT; -} - - -static void * -i915_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *rasterizer) -{ - struct i915_rasterizer_state *cso = CALLOC_STRUCT( i915_rasterizer_state ); - - cso->templ = rasterizer; - cso->color_interp = rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; - cso->light_twoside = rasterizer->light_twoside; - cso->ds[0].u = _3DSTATE_DEPTH_OFFSET_SCALE; - cso->ds[1].f = rasterizer->offset_scale; - if (rasterizer->poly_stipple_enable) { - cso->st |= ST1_ENABLE; - } - - if (rasterizer->scissor) - cso->sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | ENABLE_SCISSOR_RECT; - else - cso->sc[0] = _3DSTATE_SCISSOR_ENABLE_CMD | DISABLE_SCISSOR_RECT; - - switch (rasterizer->cull_mode) { - case PIPE_WINDING_NONE: - cso->LIS4 |= S4_CULLMODE_NONE; - break; - case PIPE_WINDING_CW: - cso->LIS4 |= S4_CULLMODE_CW; - break; - case PIPE_WINDING_CCW: - cso->LIS4 |= S4_CULLMODE_CCW; - break; - case PIPE_WINDING_BOTH: - cso->LIS4 |= S4_CULLMODE_BOTH; - break; - } - - { - int line_width = CLAMP((int)(rasterizer->line_width * 2), 1, 0xf); - - cso->LIS4 |= line_width << S4_LINE_WIDTH_SHIFT; - - if (rasterizer->line_smooth) - cso->LIS4 |= S4_LINE_ANTIALIAS_ENABLE; - } - - { - int point_size = CLAMP((int) rasterizer->point_size, 1, 0xff); - - cso->LIS4 |= point_size << S4_POINT_WIDTH_SHIFT; - } - - if (rasterizer->flatshade) { - cso->LIS4 |= (S4_FLATSHADE_ALPHA | - S4_FLATSHADE_COLOR | - S4_FLATSHADE_SPECULAR); - } - - cso->LIS7 = fui( rasterizer->offset_units ); - - - return cso; -} - -static void i915_bind_rasterizer_state( struct pipe_context *pipe, - void *setup ) -{ - struct i915_context *i915 = i915_context(pipe); - - i915->rasterizer = (struct i915_rasterizer_state *)setup; - - /* pass-through to draw module */ - draw_set_rasterizer_state(i915->draw, i915->rasterizer->templ); - - i915->dirty |= I915_NEW_RASTERIZER; -} - -static void i915_delete_rasterizer_state(struct pipe_context *pipe, - void *setup) -{ - FREE(setup); -} - -static void i915_set_vertex_buffer( struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer ) -{ - struct i915_context *i915 = i915_context(pipe); - i915->vertex_buffer[index] = *buffer; - /* pass-through to draw module */ - draw_set_vertex_buffer(i915->draw, index, buffer); -} - -static void i915_set_vertex_element( struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_element *element) -{ - struct i915_context *i915 = i915_context(pipe); - /* pass-through to draw module */ - draw_set_vertex_element(i915->draw, index, element); -} - - - -void -i915_init_state_functions( struct i915_context *i915 ) -{ - i915->pipe.create_blend_state = i915_create_blend_state; - i915->pipe.bind_blend_state = i915_bind_blend_state; - i915->pipe.delete_blend_state = i915_delete_blend_state; - - i915->pipe.create_sampler_state = i915_create_sampler_state; - i915->pipe.bind_sampler_state = i915_bind_sampler_state; - i915->pipe.delete_sampler_state = i915_delete_sampler_state; - - i915->pipe.create_depth_stencil_alpha_state = i915_create_depth_stencil_state; - i915->pipe.bind_depth_stencil_alpha_state = i915_bind_depth_stencil_state; - i915->pipe.delete_depth_stencil_alpha_state = i915_delete_depth_stencil_state; - - i915->pipe.create_rasterizer_state = i915_create_rasterizer_state; - i915->pipe.bind_rasterizer_state = i915_bind_rasterizer_state; - i915->pipe.delete_rasterizer_state = i915_delete_rasterizer_state; - i915->pipe.create_fs_state = i915_create_fs_state; - i915->pipe.bind_fs_state = i915_bind_fs_state; - i915->pipe.delete_fs_state = i915_delete_fs_state; - i915->pipe.create_vs_state = i915_create_vs_state; - i915->pipe.bind_vs_state = i915_bind_vs_state; - i915->pipe.delete_vs_state = i915_delete_vs_state; - - i915->pipe.set_blend_color = i915_set_blend_color; - i915->pipe.set_clip_state = i915_set_clip_state; - i915->pipe.set_constant_buffer = i915_set_constant_buffer; - i915->pipe.set_framebuffer_state = i915_set_framebuffer_state; - - i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; - i915->pipe.set_scissor_state = i915_set_scissor_state; - i915->pipe.set_sampler_texture = i915_set_sampler_texture; - i915->pipe.set_viewport_state = i915_set_viewport_state; - i915->pipe.set_vertex_buffer = i915_set_vertex_buffer; - i915->pipe.set_vertex_element = i915_set_vertex_element; -} diff --git a/src/mesa/pipe/i915simple/i915_state.h b/src/mesa/pipe/i915simple/i915_state.h deleted file mode 100644 index 86c6b0027d..0000000000 --- a/src/mesa/pipe/i915simple/i915_state.h +++ /dev/null @@ -1,50 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef I915_STATE_H -#define I915_STATE_H - -struct i915_context; - - -struct i915_tracked_state { - unsigned dirty; - void (*update)( struct i915_context * ); -}; - -void i915_update_immediate( struct i915_context *i915 ); -void i915_update_dynamic( struct i915_context *i915 ); -void i915_update_derived( struct i915_context *i915 ); -void i915_update_samplers( struct i915_context *i915 ); -void i915_update_textures(struct i915_context *i915); - -void i915_emit_hardware_state( struct i915_context *i915 ); - -#endif diff --git a/src/mesa/pipe/i915simple/i915_state_derived.c b/src/mesa/pipe/i915simple/i915_state_derived.c deleted file mode 100644 index 653983e4a9..0000000000 --- a/src/mesa/pipe/i915simple/i915_state_derived.c +++ /dev/null @@ -1,177 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "pipe/p_util.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_vertex.h" -#include "i915_context.h" -#include "i915_state.h" -#include "i915_reg.h" -#include "i915_fpc.h" -#include "pipe/p_shader_tokens.h" - - -/** - * Determine which post-transform / pre-rasterization vertex attributes - * we need. - * Derived from: fs, setup states. - */ -static void calculate_vertex_layout( struct i915_context *i915 ) -{ - const struct pipe_shader_state *fs = i915->fs; - const enum interp_mode colorInterp = i915->rasterizer->color_interp; - struct vertex_info vinfo; - uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; - boolean needW = 0; - uint i; - boolean texCoords[8]; - uint src = 0; - - memset(texCoords, 0, sizeof(texCoords)); - memset(&vinfo, 0, sizeof(vinfo)); - - /* pos */ - draw_emit_vertex_attr(&vinfo, EMIT_3F, INTERP_LINEAR, src++); - /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ - - for (i = 0; i < fs->num_inputs; i++) { - switch (fs->input_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - break; - case TGSI_SEMANTIC_COLOR: - if (fs->input_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src++); - vinfo.hwfmt[0] |= S4_VFMT_COLOR; - } - else { - assert(fs->input_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src++); - vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG; - } - break; - case TGSI_SEMANTIC_GENERIC: - /* usually a texcoord */ - { - const uint unit = fs->input_semantic_index[i]; - uint hwtc; - texCoords[unit] = TRUE; - draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); - hwtc = TEXCOORDFMT_4D; - needW = TRUE; - vinfo.hwfmt[1] |= hwtc << (unit * 4); - } - break; - case TGSI_SEMANTIC_FOG: - debug_printf("i915 fogcoord not implemented yet\n"); - draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_PERSPECTIVE, src++); - break; - default: - assert(0); - } - - } - - /* finish up texcoord fields */ - for (i = 0; i < 8; i++) { - if (!texCoords[i]) { - const uint hwtc = TEXCOORDFMT_NOT_PRESENT; - vinfo.hwfmt[1] |= hwtc << (i* 4); - } - } - - /* go back and fill in the vertex position info now that we have needW */ - if (needW) { - vinfo.hwfmt[0] |= S4_VFMT_XYZW; - vinfo.emit[0] = EMIT_4F; - } - else { - vinfo.hwfmt[0] |= S4_VFMT_XYZ; - vinfo.emit[0] = EMIT_3F; - } - - /* Additional attributes required for setup: Just twosided - * lighting. Edgeflag is dealt with specially by setting bits in - * the vertex header. - */ - if (i915->rasterizer->light_twoside) { - if (front0) { - back0 = draw_emit_vertex_attr(&vinfo, EMIT_OMIT, colorInterp, src++); - } - if (back0) { - back1 = draw_emit_vertex_attr(&vinfo, EMIT_OMIT, colorInterp, src++); - } - } - - draw_compute_vertex_size(&vinfo); - - if (memcmp(&i915->current.vertex_info, &vinfo, sizeof(vinfo))) { - /* Need to set this flag so that the LIS2/4 registers get set. - * It also means the i915_update_immediate() function must be called - * after this one, in i915_update_derived(). - */ - i915->dirty |= I915_NEW_VERTEX_FORMAT; - - memcpy(&i915->current.vertex_info, &vinfo, sizeof(vinfo)); - } -} - - - - -/* Hopefully this will remain quite simple, otherwise need to pull in - * something like the state tracker mechanism. - */ -void i915_update_derived( struct i915_context *i915 ) -{ - if (i915->dirty & (I915_NEW_RASTERIZER | I915_NEW_FS)) - calculate_vertex_layout( i915 ); - - if (i915->dirty & (I915_NEW_SAMPLER | I915_NEW_TEXTURE)) - i915_update_samplers(i915); - - if (i915->dirty & I915_NEW_TEXTURE) - i915_update_textures(i915); - - if (i915->dirty) - i915_update_immediate( i915 ); - - if (i915->dirty) - i915_update_dynamic( i915 ); - - if (i915->dirty & I915_NEW_FS) { - i915_translate_fragment_program(i915); - i915->hardware_dirty |= I915_HW_PROGRAM; /* XXX right? */ - } - - /* HW emit currently references framebuffer state directly: - */ - if (i915->dirty & I915_NEW_FRAMEBUFFER) - i915->hardware_dirty |= I915_HW_STATIC; - - i915->dirty = 0; -} diff --git a/src/mesa/pipe/i915simple/i915_state_dynamic.c b/src/mesa/pipe/i915simple/i915_state_dynamic.c deleted file mode 100644 index 8cfbdddd19..0000000000 --- a/src/mesa/pipe/i915simple/i915_state_dynamic.c +++ /dev/null @@ -1,308 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915_batch.h" -#include "i915_state_inlines.h" -#include "i915_context.h" -#include "i915_reg.h" -#include "i915_state.h" -#include "pipe/p_util.h" - -#define FILE_DEBUG_FLAG DEBUG_STATE - -/* State that we have chosen to store in the DYNAMIC segment of the - * i915 indirect state mechanism. - * - * Can't cache these in the way we do the static state, as there is no - * start/size in the command packet, instead an 'end' value that gets - * incremented. - * - * Additionally, there seems to be a requirement to re-issue the full - * (active) state every time a 4kb boundary is crossed. - */ - -static INLINE void set_dynamic_indirect( struct i915_context *i915, - unsigned offset, - const unsigned *src, - unsigned dwords ) -{ - unsigned i; - - for (i = 0; i < dwords; i++) - i915->current.dynamic[offset + i] = src[i]; - - i915->hardware_dirty |= I915_HW_DYNAMIC; -} - - -/*********************************************************************** - * Modes4: stencil masks and logicop - */ -static void upload_MODES4( struct i915_context *i915 ) -{ - unsigned modes4 = 0; - - /* I915_NEW_STENCIL */ - modes4 |= i915->depth_stencil->stencil_modes4; - /* I915_NEW_BLEND */ - modes4 |= i915->blend->modes4; - - /* Always, so that we know when state is in-active: - */ - set_dynamic_indirect( i915, - I915_DYNAMIC_MODES4, - &modes4, - 1 ); -} - -const struct i915_tracked_state i915_upload_MODES4 = { - I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, - upload_MODES4 -}; - - - - -/*********************************************************************** - */ - -static void upload_BFO( struct i915_context *i915 ) -{ - set_dynamic_indirect( i915, - I915_DYNAMIC_BFO_0, - &(i915->depth_stencil->bfo[0]), - 2 ); -} - -const struct i915_tracked_state i915_upload_BFO = { - I915_NEW_DEPTH_STENCIL, - upload_BFO -}; - - -/*********************************************************************** - */ - - -static void upload_BLENDCOLOR( struct i915_context *i915 ) -{ - unsigned bc[2]; - - memset( bc, 0, sizeof(bc) ); - - /* I915_NEW_BLEND {_COLOR} - */ - { - const float *color = i915->blend_color.color; - - bc[0] = _3DSTATE_CONST_BLEND_COLOR_CMD; - bc[1] = pack_ui32_float4( color[0], - color[1], - color[2], - color[3] ); - } - - set_dynamic_indirect( i915, - I915_DYNAMIC_BC_0, - bc, - 2 ); -} - -const struct i915_tracked_state i915_upload_BLENDCOLOR = { - I915_NEW_BLEND, - upload_BLENDCOLOR -}; - -/*********************************************************************** - */ - - -static void upload_IAB( struct i915_context *i915 ) -{ - unsigned iab = i915->blend->iab; - - - set_dynamic_indirect( i915, - I915_DYNAMIC_IAB, - &iab, - 1 ); -} - -const struct i915_tracked_state i915_upload_IAB = { - I915_NEW_BLEND, - upload_IAB -}; - - -/*********************************************************************** - */ - - - -static void upload_DEPTHSCALE( struct i915_context *i915 ) -{ - set_dynamic_indirect( i915, - I915_DYNAMIC_DEPTHSCALE_0, - &(i915->rasterizer->ds[0].u), - 2 ); -} - -const struct i915_tracked_state i915_upload_DEPTHSCALE = { - I915_NEW_RASTERIZER, - upload_DEPTHSCALE -}; - - - -/*********************************************************************** - * Polygon stipple - * - * The i915 supports a 4x4 stipple natively, GL wants 32x32. - * Fortunately stipple is usually a repeating pattern. - * - * XXX: does stipple pattern need to be adjusted according to - * the window position? - * - * XXX: possibly need workaround for conform paths test. - */ - -static void upload_STIPPLE( struct i915_context *i915 ) -{ - unsigned st[2]; - - st[0] = _3DSTATE_STIPPLE; - st[1] = 0; - - /* I915_NEW_RASTERIZER - */ - st[1] |= i915->rasterizer->st; - - - /* I915_NEW_STIPPLE - */ - { - const ubyte *mask = (const ubyte *)i915->poly_stipple.stipple; - ubyte p[4]; - - p[0] = mask[12] & 0xf; - p[1] = mask[8] & 0xf; - p[2] = mask[4] & 0xf; - p[3] = mask[0] & 0xf; - - /* Not sure what to do about fallbacks, so for now just dont: - */ - st[1] |= ((p[0] << 0) | - (p[1] << 4) | - (p[2] << 8) | - (p[3] << 12)); - } - - - set_dynamic_indirect( i915, - I915_DYNAMIC_STP_0, - &st[0], - 2 ); -} - - -const struct i915_tracked_state i915_upload_STIPPLE = { - I915_NEW_RASTERIZER | I915_NEW_STIPPLE, - upload_STIPPLE -}; - - - -/*********************************************************************** - * Scissor. - */ -static void upload_SCISSOR_ENABLE( struct i915_context *i915 ) -{ - set_dynamic_indirect( i915, - I915_DYNAMIC_SC_ENA_0, - &(i915->rasterizer->sc[0]), - 1 ); -} - -const struct i915_tracked_state i915_upload_SCISSOR_ENABLE = { - I915_NEW_RASTERIZER, - upload_SCISSOR_ENABLE -}; - - - -static void upload_SCISSOR_RECT( struct i915_context *i915 ) -{ - unsigned x1 = i915->scissor.minx; - unsigned y1 = i915->scissor.miny; - unsigned x2 = i915->scissor.maxx; - unsigned y2 = i915->scissor.maxy; - unsigned sc[3]; - - sc[0] = _3DSTATE_SCISSOR_RECT_0_CMD; - sc[1] = (y1 << 16) | (x1 & 0xffff); - sc[2] = (y2 << 16) | (x2 & 0xffff); - - set_dynamic_indirect( i915, - I915_DYNAMIC_SC_RECT_0, - &sc[0], - 3 ); -} - - -const struct i915_tracked_state i915_upload_SCISSOR_RECT = { - I915_NEW_SCISSOR, - upload_SCISSOR_RECT -}; - - - - - - -static const struct i915_tracked_state *atoms[] = { - &i915_upload_MODES4, - &i915_upload_BFO, - &i915_upload_BLENDCOLOR, - &i915_upload_IAB, - &i915_upload_DEPTHSCALE, - &i915_upload_STIPPLE, - &i915_upload_SCISSOR_ENABLE, - &i915_upload_SCISSOR_RECT -}; - -/* These will be dynamic indirect state commands, but for now just end - * up on the batch buffer with everything else. - */ -void i915_update_dynamic( struct i915_context *i915 ) -{ - int i; - - for (i = 0; i < Elements(atoms); i++) - if (i915->dirty & atoms[i]->dirty) - atoms[i]->update( i915 ); -} - diff --git a/src/mesa/pipe/i915simple/i915_state_emit.c b/src/mesa/pipe/i915simple/i915_state_emit.c deleted file mode 100644 index 3339287f49..0000000000 --- a/src/mesa/pipe/i915simple/i915_state_emit.c +++ /dev/null @@ -1,374 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "i915_reg.h" -#include "i915_context.h" -#include "i915_winsys.h" -#include "i915_batch.h" -#include "i915_reg.h" - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" - -static unsigned translate_format( enum pipe_format format ) -{ - switch (format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - return COLOR_BUF_ARGB8888; - case PIPE_FORMAT_R5G6B5_UNORM: - return COLOR_BUF_RGB565; - default: - assert(0); - return 0; - } -} - -static unsigned translate_depth_format( enum pipe_format zformat ) -{ - switch (zformat) { - case PIPE_FORMAT_S8Z24_UNORM: - return DEPTH_FRMT_24_FIXED_8_OTHER; - case PIPE_FORMAT_Z16_UNORM: - return DEPTH_FRMT_16_FIXED; - default: - assert(0); - return 0; - } -} - - -/** - * Examine framebuffer state to determine width, height. - */ -static boolean -framebuffer_size(const struct pipe_framebuffer_state *fb, - uint *width, uint *height) -{ - if (fb->cbufs[0]) { - *width = fb->cbufs[0]->width; - *height = fb->cbufs[0]->height; - return TRUE; - } - else if (fb->zsbuf) { - *width = fb->zsbuf->width; - *height = fb->zsbuf->height; - return TRUE; - } - else { - *width = *height = 0; - return FALSE; - } -} - - -/* Push the state into the sarea and/or texture memory. - */ -void -i915_emit_hardware_state(struct i915_context *i915 ) -{ - /* XXX: there must be an easier way */ - const unsigned dwords = ( 14 + - 7 + - I915_MAX_DYNAMIC + - 8 + - 2 + I915_TEX_UNITS*3 + - 2 + I915_TEX_UNITS*3 + - 2 + I915_MAX_CONSTANT*4 + - i915->current.program_len + - 6 - ) * 3/2; /* plus 50% margin */ - const unsigned relocs = ( I915_TEX_UNITS + - 3 - ) * 3/2; /* plus 50% margin */ - -#if 0 - debug_printf("i915_emit_hardware_state: %d dwords, %d relocs\n", dwords, relocs); -#endif - - if(!BEGIN_BATCH(dwords, relocs)) { - FLUSH_BATCH(); - assert(BEGIN_BATCH(dwords, relocs)); - } - - /* 14 dwords, 0 relocs */ - if (i915->hardware_dirty & I915_HW_INVARIENT) - { - OUT_BATCH(_3DSTATE_AA_CMD | - AA_LINE_ECAAR_WIDTH_ENABLE | - AA_LINE_ECAAR_WIDTH_1_0 | - AA_LINE_REGION_WIDTH_ENABLE | AA_LINE_REGION_WIDTH_1_0); - - OUT_BATCH(_3DSTATE_DFLT_DIFFUSE_CMD); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_DFLT_SPEC_CMD); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_DFLT_Z_CMD); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_COORD_SET_BINDINGS | - CSB_TCB(0, 0) | - CSB_TCB(1, 1) | - CSB_TCB(2, 2) | - CSB_TCB(3, 3) | - CSB_TCB(4, 4) | - CSB_TCB(5, 5) | - CSB_TCB(6, 6) | - CSB_TCB(7, 7)); - - OUT_BATCH(_3DSTATE_RASTER_RULES_CMD | - ENABLE_POINT_RASTER_RULE | - OGL_POINT_RASTER_RULE | - ENABLE_LINE_STRIP_PROVOKE_VRTX | - ENABLE_TRI_FAN_PROVOKE_VRTX | - LINE_STRIP_PROVOKE_VRTX(1) | - TRI_FAN_PROVOKE_VRTX(2) | - ENABLE_TEXKILL_3D_4D | - TEXKILL_4D); - - /* Need to initialize this to zero. - */ - OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | I1_LOAD_S(3) | (0)); - OUT_BATCH(0); - - OUT_BATCH(_3DSTATE_DEPTH_SUBRECT_DISABLE); - - /* disable indirect state for now - */ - OUT_BATCH(_3DSTATE_LOAD_INDIRECT | 0); - OUT_BATCH(0); - } - - /* 7 dwords, 1 relocs */ - if (i915->hardware_dirty & I915_HW_IMMEDIATE) - { - OUT_BATCH(_3DSTATE_LOAD_STATE_IMMEDIATE_1 | - I1_LOAD_S(0) | - I1_LOAD_S(1) | - I1_LOAD_S(2) | - I1_LOAD_S(4) | - I1_LOAD_S(5) | - I1_LOAD_S(6) | - (5)); - - if(i915->vbo) - OUT_RELOC(i915->vbo, - I915_BUFFER_ACCESS_READ, - i915->current.immediate[I915_IMMEDIATE_S0]); - else - /* FIXME: we should not do this */ - OUT_BATCH(0); - OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S1]); - OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S2]); - OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S4]); - OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S5]); - OUT_BATCH(i915->current.immediate[I915_IMMEDIATE_S6]); - } - - /* I915_MAX_DYNAMIC dwords, 0 relocs */ - if (i915->hardware_dirty & I915_HW_DYNAMIC) - { - int i; - for (i = 0; i < I915_MAX_DYNAMIC; i++) { - OUT_BATCH(i915->current.dynamic[i]); - } - } - - /* 8 dwords, 2 relocs */ - if (i915->hardware_dirty & I915_HW_STATIC) - { - struct pipe_surface *cbuf_surface = i915->framebuffer.cbufs[0]; - struct pipe_surface *depth_surface = i915->framebuffer.zsbuf; - - if (cbuf_surface) { - unsigned pitch = (cbuf_surface->pitch * cbuf_surface->cpp); - - OUT_BATCH(_3DSTATE_BUF_INFO_CMD); - - OUT_BATCH(BUF_3D_ID_COLOR_BACK | - BUF_3D_PITCH(pitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - - OUT_RELOC(cbuf_surface->buffer, - I915_BUFFER_ACCESS_WRITE, - 0); - } - - /* What happens if no zbuf?? - */ - if (depth_surface) { - unsigned zpitch = (depth_surface->pitch * depth_surface->cpp); - - OUT_BATCH(_3DSTATE_BUF_INFO_CMD); - - OUT_BATCH(BUF_3D_ID_DEPTH | - BUF_3D_PITCH(zpitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); - - OUT_RELOC(depth_surface->buffer, - I915_BUFFER_ACCESS_WRITE, - 0); - } - - { - unsigned cformat, zformat = 0; - - if (cbuf_surface) - cformat = cbuf_surface->format; - else - cformat = PIPE_FORMAT_A8R8G8B8_UNORM; /* arbitrary */ - cformat = translate_format(cformat); - - if (depth_surface) - zformat = translate_depth_format( i915->framebuffer.zsbuf->format ); - - OUT_BATCH(_3DSTATE_DST_BUF_VARS_CMD); - OUT_BATCH(DSTORG_HORT_BIAS(0x8) | /* .5 */ - DSTORG_VERT_BIAS(0x8) | /* .5 */ - LOD_PRECLAMP_OGL | - TEX_DEFAULT_COLOR_OGL | - cformat | - zformat ); - } - } - -#if 01 - /* texture images */ - /* 2 + I915_TEX_UNITS*3 dwords, I915_TEX_UNITS relocs */ - if (i915->hardware_dirty & (I915_HW_MAP | I915_HW_SAMPLER)) - { - /* XXX: we were refering to sampler state - * (current.sampler_enable_nr) below, but only checking - * I915_HW_MAP above. Should probably calculate the enabled - * flags separately - but there will be further rework of - * state so perhaps not necessary yet. - */ - const uint nr = i915->current.sampler_enable_nr; - if (nr) { - const uint enabled = i915->current.sampler_enable_flags; - uint unit; - uint count = 0; - OUT_BATCH(_3DSTATE_MAP_STATE | (3 * nr)); - OUT_BATCH(enabled); - for (unit = 0; unit < I915_TEX_UNITS; unit++) { - if (enabled & (1 << unit)) { - struct pipe_buffer *buf = - i915->texture[unit]->buffer; - uint offset = 0; - assert(buf); - - count++; - - OUT_RELOC(buf, - I915_BUFFER_ACCESS_READ, - offset); - OUT_BATCH(i915->current.texbuffer[unit][0]); /* MS3 */ - OUT_BATCH(i915->current.texbuffer[unit][1]); /* MS4 */ - } - } - assert(count == nr); - } - } -#endif - -#if 01 - /* samplers */ - /* 2 + I915_TEX_UNITS*3 dwords, 0 relocs */ - if (i915->hardware_dirty & I915_HW_SAMPLER) - { - if (i915->current.sampler_enable_nr) { - int i; - - OUT_BATCH( _3DSTATE_SAMPLER_STATE | - (3 * i915->current.sampler_enable_nr) ); - - OUT_BATCH( i915->current.sampler_enable_flags ); - - for (i = 0; i < I915_TEX_UNITS; i++) { - if (i915->current.sampler_enable_flags & (1<current.sampler[i][0] ); - OUT_BATCH( i915->current.sampler[i][1] ); - OUT_BATCH( i915->current.sampler[i][2] ); - } - } - } - } -#endif - - /* constants */ - /* 2 + I915_MAX_CONSTANT*4 dwords, 0 relocs */ - if (i915->hardware_dirty & I915_HW_PROGRAM) - { - const uint nr = i915->current.num_constants[PIPE_SHADER_FRAGMENT]; - assert(nr <= I915_MAX_CONSTANT); - if (nr > 0) { - const uint *c - = (const uint *) i915->current.constants[PIPE_SHADER_FRAGMENT]; - uint i; - OUT_BATCH( _3DSTATE_PIXEL_SHADER_CONSTANTS | (nr * 4) ); - OUT_BATCH( (1 << (nr - 1)) | ((1 << (nr - 1)) - 1) ); - for (i = 0; i < nr; i++) { - OUT_BATCH(*c++); - OUT_BATCH(*c++); - OUT_BATCH(*c++); - OUT_BATCH(*c++); - } - } - } - - /* Fragment program */ - /* i915->current.program_len dwords, 0 relocs */ - if (i915->hardware_dirty & I915_HW_PROGRAM) - { - uint i; - /* we should always have, at least, a pass-through program */ - assert(i915->current.program_len > 0); - for (i = 0; i < i915->current.program_len; i++) { - OUT_BATCH(i915->current.program[i]); - } - } - - /* drawing surface size */ - /* 6 dwords, 0 relocs */ - { - uint w, h; - boolean k = framebuffer_size(&i915->framebuffer, &w, &h); - assert(k); - - OUT_BATCH(_3DSTATE_DRAW_RECT_CMD); - OUT_BATCH(0); - OUT_BATCH(0); - OUT_BATCH(((w - 1) & 0xffff) | ((h - 1) << 16)); - OUT_BATCH(0); - OUT_BATCH(0); - } - - - i915->hardware_dirty = 0; -} diff --git a/src/mesa/pipe/i915simple/i915_state_immediate.c b/src/mesa/pipe/i915simple/i915_state_immediate.c deleted file mode 100644 index 07031fc6c5..0000000000 --- a/src/mesa/pipe/i915simple/i915_state_immediate.c +++ /dev/null @@ -1,221 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "i915_state_inlines.h" -#include "i915_context.h" -#include "i915_state.h" -#include "i915_reg.h" -#include "p_util.h" - - -/* All state expressable with the LOAD_STATE_IMMEDIATE_1 packet. - * Would like to opportunistically recombine all these fragments into - * a single packet containing only what has changed, but for now emit - * as multiple packets. - */ - - - - -/*********************************************************************** - * S0,S1: Vertex buffer state. - */ -static void upload_S0S1(struct i915_context *i915) -{ - unsigned LIS0, LIS1; - - /* INTEL_NEW_VBO */ - /* TODO: re-use vertex buffers here? */ - LIS0 = 0; - - /* INTEL_NEW_VERTEX_SIZE -- do this where the vertex size is calculated! - */ - { - unsigned vertex_size = i915->current.vertex_info.size; - - LIS1 = ((vertex_size << 24) | - (vertex_size << 16)); - } - - /* INTEL_NEW_VBO */ - /* TODO: use a vertex generation number to track vbo changes */ - if (1 || - i915->current.immediate[I915_IMMEDIATE_S0] != LIS0 || - i915->current.immediate[I915_IMMEDIATE_S1] != LIS1) - { - i915->current.immediate[I915_IMMEDIATE_S0] = LIS0; - i915->current.immediate[I915_IMMEDIATE_S1] = LIS1; - i915->hardware_dirty |= I915_HW_IMMEDIATE; - } -} - -const struct i915_tracked_state i915_upload_S0S1 = { - I915_NEW_VBO | I915_NEW_VERTEX_FORMAT, - upload_S0S1 -}; - - - - -/*********************************************************************** - * S4: Vertex format, rasterization state - */ -static void upload_S2S4(struct i915_context *i915) -{ - unsigned LIS2, LIS4; - - /* I915_NEW_VERTEX_FORMAT */ - { - LIS2 = i915->current.vertex_info.hwfmt[1]; - LIS4 = i915->current.vertex_info.hwfmt[0]; - /* - debug_printf("LIS2: 0x%x LIS4: 0x%x\n", LIS2, LIS4); - */ - assert(LIS4); /* should never be zero? */ - } - - LIS4 |= i915->rasterizer->LIS4; - - if (LIS2 != i915->current.immediate[I915_IMMEDIATE_S2] || - LIS4 != i915->current.immediate[I915_IMMEDIATE_S4]) { - - i915->current.immediate[I915_IMMEDIATE_S2] = LIS2; - i915->current.immediate[I915_IMMEDIATE_S4] = LIS4; - i915->hardware_dirty |= I915_HW_IMMEDIATE; - } -} - - -const struct i915_tracked_state i915_upload_S2S4 = { - I915_NEW_RASTERIZER | I915_NEW_VERTEX_FORMAT, - upload_S2S4 -}; - - - -/*********************************************************************** - * - */ -static void upload_S5( struct i915_context *i915 ) -{ - unsigned LIS5 = 0; - - LIS5 |= i915->depth_stencil->stencil_LIS5; - - LIS5 |= i915->blend->LIS5; - -#if 0 - /* I915_NEW_RASTERIZER */ - if (i915->state.Polygon->OffsetFill) { - LIS5 |= S5_GLOBAL_DEPTH_OFFSET_ENABLE; - } -#endif - - - if (LIS5 != i915->current.immediate[I915_IMMEDIATE_S5]) { - i915->current.immediate[I915_IMMEDIATE_S5] = LIS5; - i915->hardware_dirty |= I915_HW_IMMEDIATE; - } -} - -const struct i915_tracked_state i915_upload_S5 = { - (I915_NEW_DEPTH_STENCIL | I915_NEW_BLEND | I915_NEW_RASTERIZER), - upload_S5 -}; - - -/*********************************************************************** - */ -static void upload_S6( struct i915_context *i915 ) -{ - unsigned LIS6 = (S6_COLOR_WRITE_ENABLE | - (2 << S6_TRISTRIP_PV_SHIFT)); - - /* I915_NEW_BLEND - */ - LIS6 |= i915->blend->LIS6; - - /* I915_NEW_DEPTH - */ - LIS6 |= i915->depth_stencil->depth_LIS6; - - if (LIS6 != i915->current.immediate[I915_IMMEDIATE_S6]) { - i915->current.immediate[I915_IMMEDIATE_S6] = LIS6; - i915->hardware_dirty |= I915_HW_IMMEDIATE; - } -} - -const struct i915_tracked_state i915_upload_S6 = { - I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, - upload_S6 -}; - - -/*********************************************************************** - */ -static void upload_S7( struct i915_context *i915 ) -{ - unsigned LIS7; - - /* I915_NEW_RASTERIZER - */ - LIS7 = i915->rasterizer->LIS7; - - if (LIS7 != i915->current.immediate[I915_IMMEDIATE_S7]) { - i915->current.immediate[I915_IMMEDIATE_S7] = LIS7; - i915->hardware_dirty |= I915_HW_IMMEDIATE; - } -} - -const struct i915_tracked_state i915_upload_S7 = { - I915_NEW_RASTERIZER, - upload_S7 -}; - - -static const struct i915_tracked_state *atoms[] = { - &i915_upload_S0S1, - &i915_upload_S2S4, - &i915_upload_S5, - &i915_upload_S6, - &i915_upload_S7 -}; - -/* - */ -void i915_update_immediate( struct i915_context *i915 ) -{ - int i; - - for (i = 0; i < Elements(atoms); i++) - if (i915->dirty & atoms[i]->dirty) - atoms[i]->update( i915 ); -} diff --git a/src/mesa/pipe/i915simple/i915_state_inlines.h b/src/mesa/pipe/i915simple/i915_state_inlines.h deleted file mode 100644 index 0934ac79a4..0000000000 --- a/src/mesa/pipe/i915simple/i915_state_inlines.h +++ /dev/null @@ -1,230 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef I915_STATE_INLINES_H -#define I915_STATE_INLINES_H - -#include "p_compiler.h" -#include "p_defines.h" -#include "i915_reg.h" - - -static INLINE unsigned -i915_translate_compare_func(unsigned func) -{ - switch (func) { - case PIPE_FUNC_NEVER: - return COMPAREFUNC_NEVER; - case PIPE_FUNC_LESS: - return COMPAREFUNC_LESS; - case PIPE_FUNC_LEQUAL: - return COMPAREFUNC_LEQUAL; - case PIPE_FUNC_GREATER: - return COMPAREFUNC_GREATER; - case PIPE_FUNC_GEQUAL: - return COMPAREFUNC_GEQUAL; - case PIPE_FUNC_NOTEQUAL: - return COMPAREFUNC_NOTEQUAL; - case PIPE_FUNC_EQUAL: - return COMPAREFUNC_EQUAL; - case PIPE_FUNC_ALWAYS: - return COMPAREFUNC_ALWAYS; - default: - return COMPAREFUNC_ALWAYS; - } -} - -static INLINE unsigned -i915_translate_stencil_op(unsigned op) -{ - switch (op) { - case PIPE_STENCIL_OP_KEEP: - return STENCILOP_KEEP; - case PIPE_STENCIL_OP_ZERO: - return STENCILOP_ZERO; - case PIPE_STENCIL_OP_REPLACE: - return STENCILOP_REPLACE; - case PIPE_STENCIL_OP_INCR: - return STENCILOP_INCRSAT; - case PIPE_STENCIL_OP_DECR: - return STENCILOP_DECRSAT; - case PIPE_STENCIL_OP_INCR_WRAP: - return STENCILOP_INCR; - case PIPE_STENCIL_OP_DECR_WRAP: - return STENCILOP_DECR; - case PIPE_STENCIL_OP_INVERT: - return STENCILOP_INVERT; - default: - return STENCILOP_ZERO; - } -} - -static INLINE unsigned -i915_translate_blend_factor(unsigned factor) -{ - switch (factor) { - case PIPE_BLENDFACTOR_ZERO: - return BLENDFACT_ZERO; - case PIPE_BLENDFACTOR_SRC_ALPHA: - return BLENDFACT_SRC_ALPHA; - case PIPE_BLENDFACTOR_ONE: - return BLENDFACT_ONE; - case PIPE_BLENDFACTOR_SRC_COLOR: - return BLENDFACT_SRC_COLR; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - return BLENDFACT_INV_SRC_COLR; - case PIPE_BLENDFACTOR_DST_COLOR: - return BLENDFACT_DST_COLR; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - return BLENDFACT_INV_DST_COLR; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - return BLENDFACT_INV_SRC_ALPHA; - case PIPE_BLENDFACTOR_DST_ALPHA: - return BLENDFACT_DST_ALPHA; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - return BLENDFACT_INV_DST_ALPHA; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - return BLENDFACT_SRC_ALPHA_SATURATE; - case PIPE_BLENDFACTOR_CONST_COLOR: - return BLENDFACT_CONST_COLOR; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - return BLENDFACT_INV_CONST_COLOR; - case PIPE_BLENDFACTOR_CONST_ALPHA: - return BLENDFACT_CONST_ALPHA; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - return BLENDFACT_INV_CONST_ALPHA; - default: - return BLENDFACT_ZERO; - } -} - -static INLINE unsigned -i915_translate_blend_func(unsigned mode) -{ - switch (mode) { - case PIPE_BLEND_ADD: - return BLENDFUNC_ADD; - case PIPE_BLEND_MIN: - return BLENDFUNC_MIN; - case PIPE_BLEND_MAX: - return BLENDFUNC_MAX; - case PIPE_BLEND_SUBTRACT: - return BLENDFUNC_SUBTRACT; - case PIPE_BLEND_REVERSE_SUBTRACT: - return BLENDFUNC_REVERSE_SUBTRACT; - default: - return 0; - } -} - - -static INLINE unsigned -i915_translate_logic_op(unsigned opcode) -{ - switch (opcode) { - case PIPE_LOGICOP_CLEAR: - return LOGICOP_CLEAR; - case PIPE_LOGICOP_AND: - return LOGICOP_AND; - case PIPE_LOGICOP_AND_REVERSE: - return LOGICOP_AND_RVRSE; - case PIPE_LOGICOP_COPY: - return LOGICOP_COPY; - case PIPE_LOGICOP_COPY_INVERTED: - return LOGICOP_COPY_INV; - case PIPE_LOGICOP_AND_INVERTED: - return LOGICOP_AND_INV; - case PIPE_LOGICOP_NOOP: - return LOGICOP_NOOP; - case PIPE_LOGICOP_XOR: - return LOGICOP_XOR; - case PIPE_LOGICOP_OR: - return LOGICOP_OR; - case PIPE_LOGICOP_OR_INVERTED: - return LOGICOP_OR_INV; - case PIPE_LOGICOP_NOR: - return LOGICOP_NOR; - case PIPE_LOGICOP_EQUIV: - return LOGICOP_EQUIV; - case PIPE_LOGICOP_INVERT: - return LOGICOP_INV; - case PIPE_LOGICOP_OR_REVERSE: - return LOGICOP_OR_RVRSE; - case PIPE_LOGICOP_NAND: - return LOGICOP_NAND; - case PIPE_LOGICOP_SET: - return LOGICOP_SET; - default: - return LOGICOP_SET; - } -} - - - -static INLINE boolean i915_validate_vertices( unsigned hw_prim, unsigned nr ) -{ - boolean ok; - - switch (hw_prim) { - case PRIM3D_POINTLIST: - ok = (nr >= 1); - assert(ok); - break; - case PRIM3D_LINELIST: - ok = (nr >= 2) && (nr % 2) == 0; - assert(ok); - break; - case PRIM3D_LINESTRIP: - ok = (nr >= 2); - assert(ok); - break; - case PRIM3D_TRILIST: - ok = (nr >= 3) && (nr % 3) == 0; - assert(ok); - break; - case PRIM3D_TRISTRIP: - ok = (nr >= 3); - assert(ok); - break; - case PRIM3D_TRIFAN: - ok = (nr >= 3); - assert(ok); - break; - case PRIM3D_POLY: - ok = (nr >= 3); - assert(ok); - break; - default: - assert(0); - ok = 0; - break; - } - - return ok; -} - -#endif diff --git a/src/mesa/pipe/i915simple/i915_state_sampler.c b/src/mesa/pipe/i915simple/i915_state_sampler.c deleted file mode 100644 index 9c1a5bbbd6..0000000000 --- a/src/mesa/pipe/i915simple/i915_state_sampler.c +++ /dev/null @@ -1,231 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_context.h" -#include "pipe/p_state.h" -#include "pipe/p_util.h" - -#include "i915_state_inlines.h" -#include "i915_context.h" -#include "i915_reg.h" -#include "i915_state.h" - - -/** - * Compute i915 texture sampling state. - * - * Recalculate all state from scratch. Perhaps not the most - * efficient, but this has gotten complex enough that we need - * something which is understandable and reliable. - * \param state returns the 3 words of compute state - */ -static void update_sampler(struct i915_context *i915, - uint unit, - const struct i915_sampler_state *sampler, - const struct i915_texture *tex, - unsigned state[3] ) -{ - const struct pipe_texture *pt = &tex->base; - - /* Need to do this after updating the maps, which call the - * intel_finalize_mipmap_tree and hence can update firstLevel: - */ - state[0] = sampler->state[0]; - state[1] = sampler->state[1]; - state[2] = sampler->state[2]; - - if (pt->format == PIPE_FORMAT_YCBCR || - pt->format == PIPE_FORMAT_YCBCR_REV) - state[0] |= SS2_COLORSPACE_CONVERSION; - - /* 3D textures don't seem to respect the border color. - * Fallback if there's ever a danger that they might refer to - * it. - * - * Effectively this means fallback on 3D clamp or - * clamp_to_border. - * - * XXX: Check if this is true on i945. - * XXX: Check if this bug got fixed in release silicon. - */ -#if 0 - { - const unsigned ws = sampler->templ->wrap_s; - const unsigned wt = sampler->templ->wrap_t; - const unsigned wr = sampler->templ->wrap_r; - if (pt->target == PIPE_TEXTURE_3D && - (sampler->templ->min_img_filter != PIPE_TEX_FILTER_NEAREST || - sampler->templ->mag_img_filter != PIPE_TEX_FILTER_NEAREST) && - (ws == PIPE_TEX_WRAP_CLAMP || - wt == PIPE_TEX_WRAP_CLAMP || - wr == PIPE_TEX_WRAP_CLAMP || - ws == PIPE_TEX_WRAP_CLAMP_TO_BORDER || - wt == PIPE_TEX_WRAP_CLAMP_TO_BORDER || - wr == PIPE_TEX_WRAP_CLAMP_TO_BORDER)) { - if (i915->strict_conformance) { - assert(0); - /* sampler->fallback = true; */ - /* TODO */ - } - } - } -#endif - - state[1] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT); -} - - -void i915_update_samplers( struct i915_context *i915 ) -{ - uint unit; - - i915->current.sampler_enable_nr = 0; - i915->current.sampler_enable_flags = 0x0; - - for (unit = 0; unit < I915_TEX_UNITS; unit++) { - /* determine unit enable/disable by looking for a bound texture */ - /* could also examine the fragment program? */ - if (i915->texture[unit]) { - update_sampler( i915, - unit, - i915->sampler[unit], /* sampler state */ - i915->texture[unit], /* texture */ - i915->current.sampler[unit] /* the result */ - ); - - i915->current.sampler_enable_nr++; - i915->current.sampler_enable_flags |= (1 << unit); - } - } - - i915->hardware_dirty |= I915_HW_SAMPLER; -} - - -static uint -translate_texture_format(enum pipe_format pipeFormat) -{ - switch (pipeFormat) { - case PIPE_FORMAT_U_L8: - return MAPSURF_8BIT | MT_8BIT_L8; - case PIPE_FORMAT_U_I8: - return MAPSURF_8BIT | MT_8BIT_I8; - case PIPE_FORMAT_U_A8: - return MAPSURF_8BIT | MT_8BIT_A8; - case PIPE_FORMAT_U_A8_L8: - return MAPSURF_16BIT | MT_16BIT_AY88; - case PIPE_FORMAT_R5G6B5_UNORM: - return MAPSURF_16BIT | MT_16BIT_RGB565; - case PIPE_FORMAT_A1R5G5B5_UNORM: - return MAPSURF_16BIT | MT_16BIT_ARGB1555; - case PIPE_FORMAT_A4R4G4B4_UNORM: - return MAPSURF_16BIT | MT_16BIT_ARGB4444; - case PIPE_FORMAT_A8R8G8B8_UNORM: - return MAPSURF_32BIT | MT_32BIT_ARGB8888; - case PIPE_FORMAT_YCBCR_REV: - return (MAPSURF_422 | MT_422_YCRCB_NORMAL); - case PIPE_FORMAT_YCBCR: - return (MAPSURF_422 | MT_422_YCRCB_SWAPY); -#if 0 - case PIPE_FORMAT_RGB_FXT1: - case PIPE_FORMAT_RGBA_FXT1: - return (MAPSURF_COMPRESSED | MT_COMPRESS_FXT1); -#endif - case PIPE_FORMAT_Z16_UNORM: - return (MAPSURF_16BIT | MT_16BIT_L16); -#if 0 - case PIPE_FORMAT_RGBA_DXT1: - case PIPE_FORMAT_RGB_DXT1: - return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT1); - case PIPE_FORMAT_RGBA_DXT3: - return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT2_3); - case PIPE_FORMAT_RGBA_DXT5: - return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); -#endif - case PIPE_FORMAT_S8Z24_UNORM: - return (MAPSURF_32BIT | MT_32BIT_xL824); - default: - debug_printf("i915: translate_texture_format() bad image format %x\n", - pipeFormat); - assert(0); - return 0; - } -} - - -static void -i915_update_texture(struct i915_context *i915, uint unit, - uint state[6]) -{ - const struct i915_texture *tex = i915->texture[unit]; - const struct pipe_texture *pt = &tex->base; - uint format, pitch; - const uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; - const uint num_levels = pt->last_level; - - assert(tex); - assert(width); - assert(height); - assert(depth); - - format = translate_texture_format(pt->format); - pitch = tex->pitch * pt->cpp; - - assert(format); - assert(pitch); - - /* MS3 state */ - state[0] = - (((height - 1) << MS3_HEIGHT_SHIFT) - | ((width - 1) << MS3_WIDTH_SHIFT) - | format - | MS3_USE_FENCE_REGS); - - /* MS4 state */ - state[1] = - ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) - | MS4_CUBE_FACE_ENA_MASK - | ((num_levels * 4) << MS4_MAX_LOD_SHIFT) - | ((depth - 1) << MS4_VOLUME_DEPTH_SHIFT)); -} - - -void -i915_update_textures(struct i915_context *i915) -{ - uint unit; - - for (unit = 0; unit < I915_TEX_UNITS; unit++) { - /* determine unit enable/disable by looking for a bound texture */ - /* could also examine the fragment program? */ - if (i915->texture[unit]) { - i915_update_texture(i915, unit, i915->current.texbuffer[unit]); - } - } - - i915->hardware_dirty |= I915_HW_MAP; -} diff --git a/src/mesa/pipe/i915simple/i915_strings.c b/src/mesa/pipe/i915simple/i915_strings.c deleted file mode 100644 index c713bf7208..0000000000 --- a/src/mesa/pipe/i915simple/i915_strings.c +++ /dev/null @@ -1,83 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915_context.h" -#include "i915_reg.h" - - -static const char *i915_get_vendor( struct pipe_context *pipe ) -{ - return "Tungsten Graphics, Inc."; -} - - -static const char *i915_get_name( struct pipe_context *pipe ) -{ - static char buffer[128]; - const char *chipset; - - switch (i915_context(pipe)->pci_id) { - case PCI_CHIP_I915_G: - chipset = "915G"; - break; - case PCI_CHIP_I915_GM: - chipset = "915GM"; - break; - case PCI_CHIP_I945_G: - chipset = "945G"; - break; - case PCI_CHIP_I945_GM: - chipset = "945GM"; - break; - case PCI_CHIP_I945_GME: - chipset = "945GME"; - break; - case PCI_CHIP_G33_G: - chipset = "G33"; - break; - case PCI_CHIP_Q35_G: - chipset = "Q35"; - break; - case PCI_CHIP_Q33_G: - chipset = "Q33"; - break; - default: - chipset = "unknown"; - break; - } - - sprintf(buffer, "pipe/i915 (chipset: %s)", chipset); - return buffer; -} - - -void -i915_init_string_functions(struct i915_context *i915) -{ - i915->pipe.get_name = i915_get_name; - i915->pipe.get_vendor = i915_get_vendor; -} diff --git a/src/mesa/pipe/i915simple/i915_surface.c b/src/mesa/pipe/i915simple/i915_surface.c deleted file mode 100644 index de0cc5fe06..0000000000 --- a/src/mesa/pipe/i915simple/i915_surface.c +++ /dev/null @@ -1,191 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915_context.h" -#include "i915_blit.h" -#include "i915_state.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" -#include "pipe/util/p_tile.h" - - -/* - * XXX note: same as code in sp_surface.c - */ -static struct pipe_surface * -i915_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - struct i915_texture *tex = (struct i915_texture *)pt; - struct pipe_surface *ps; - unsigned offset; /* in bytes */ - - offset = tex->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE) { - offset += tex->image_offset[level][face] * pt->cpp; - } - else if (pt->target == PIPE_TEXTURE_3D) { - offset += tex->image_offset[level][zslice] * pt->cpp; - } - else { - assert(face == 0); - assert(zslice == 0); - } - - ps = pipe->winsys->surface_alloc(pipe->winsys); - if (ps) { - assert(ps->refcount); - assert(ps->winsys); - pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); - ps->format = pt->format; - ps->cpp = pt->cpp; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->pitch = tex->pitch; - ps->offset = offset; - } - return ps; -} - - - -/* Assumes all values are within bounds -- no checking at this level - - * do it higher up if required. - */ -static void -i915_surface_copy(struct pipe_context *pipe, - unsigned do_flip, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - struct pipe_surface *src, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - assert( dst != src ); - assert( dst->cpp == src->cpp ); - - if (0) { - pipe_copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - pipe_surface_map(src), - do_flip ? -(int) src->pitch : src->pitch, - srcx, do_flip ? 1 - srcy - height : srcy); - - pipe_surface_unmap(src); - pipe_surface_unmap(dst); - } - else { - i915_copy_blit( i915_context(pipe), - do_flip, - dst->cpp, - (short) src->pitch, src->buffer, src->offset, - (short) dst->pitch, dst->buffer, dst->offset, - (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); - } -} - -/* Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static void * -get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) -{ - return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; -} - - -static void -i915_surface_fill(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - unsigned width, unsigned height, unsigned value) -{ - if (0) { - unsigned i, j; - void *dst_map = pipe_surface_map(dst); - - switch (dst->cpp) { - case 1: { - ubyte *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; - case 2: { - ushort *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; - case 4: { - unsigned *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - break; - default: - assert(0); - break; - } - - pipe_surface_unmap( dst ); - } - else { - i915_fill_blit( i915_context(pipe), - dst->cpp, - (short) dst->pitch, - dst->buffer, dst->offset, - (short) dstx, (short) dsty, - (short) width, (short) height, - value ); - } -} - - -void -i915_init_surface_functions(struct i915_context *i915) -{ - i915->pipe.get_tex_surface = i915_get_tex_surface; - - i915->pipe.surface_copy = i915_surface_copy; - i915->pipe.surface_fill = i915_surface_fill; -} diff --git a/src/mesa/pipe/i915simple/i915_texture.c b/src/mesa/pipe/i915simple/i915_texture.c deleted file mode 100644 index 6d37ae3d74..0000000000 --- a/src/mesa/pipe/i915simple/i915_texture.c +++ /dev/null @@ -1,536 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Michel Dänzer - */ - -#include "pipe/p_state.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "pipe/p_winsys.h" - -#include "i915_context.h" -#include "i915_texture.h" -#include "i915_debug.h" - - -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - - - -static void -i915_miptree_set_level_info(struct i915_texture *tex, - unsigned level, - unsigned nr_images, - unsigned x, unsigned y, unsigned w, unsigned h, unsigned d) -{ - struct pipe_texture *pt = &tex->base; - - assert(level < PIPE_MAX_TEXTURE_LEVELS); - - pt->width[level] = w; - pt->height[level] = h; - pt->depth[level] = d; - - tex->level_offset[level] = (x + y * tex->pitch) * pt->cpp; - tex->nr_images[level] = nr_images; - - /* - DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - level, w, h, d, x, y, tex->level_offset[level]); - */ - - /* Not sure when this would happen, but anyway: - */ - if (tex->image_offset[level]) { - FREE(tex->image_offset[level]); - tex->image_offset[level] = NULL; - } - - assert(nr_images); - assert(!tex->image_offset[level]); - - tex->image_offset[level] = (unsigned *) MALLOC(nr_images * sizeof(unsigned)); - tex->image_offset[level][0] = 0; -} - - -static void -i915_miptree_set_image_offset(struct i915_texture *tex, - unsigned level, unsigned img, unsigned x, unsigned y) -{ - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - - assert(img < tex->nr_images[level]); - - tex->image_offset[level][img] = (x + y * tex->pitch); - - /* - DBG("%s level %d img %d pos %d,%d image_offset %x\n", - __FUNCTION__, level, img, x, y, tex->image_offset[level][img]); - */ -} - - -static void -i945_miptree_layout_2d( struct i915_texture *tex ) -{ - struct pipe_texture *pt = &tex->base; - int align_h = 2, align_w = 4; - unsigned level; - unsigned x = 0; - unsigned y = 0; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - - tex->pitch = pt->width[0]; - - /* May need to adjust pitch to accomodate the placement of - * the 2nd mipmap level. This occurs when the alignment - * constraints of mipmap placement push the right edge of the - * 2nd mipmap level out past the width of its parent. - */ - if (pt->last_level > 0) { - unsigned mip1_width = align_int(minify(pt->width[0]), align_w) - + minify(minify(pt->width[0])); - - if (mip1_width > pt->width[0]) - tex->pitch = mip1_width; - } - - /* Pitch must be a whole number of dwords, even though we - * express it in texels. - */ - tex->pitch = align_int(tex->pitch * pt->cpp, 4) / pt->cpp; - tex->total_height = 0; - - for (level = 0; level <= pt->last_level; level++) { - unsigned img_height; - - i915_miptree_set_level_info(tex, level, 1, x, y, width, height, 1); - - if (pt->compressed) - img_height = MAX2(1, height/4); - else - img_height = align_int(height, align_h); - - - /* Because the images are packed better, the final offset - * might not be the maximal one: - */ - tex->total_height = MAX2(tex->total_height, y + img_height); - - /* Layout_below: step right after second mipmap level. - */ - if (level == 1) { - x += align_int(width, align_w); - } - else { - y += img_height; - } - - width = minify(width); - height = minify(height); - } -} - - -static const int initial_offsets[6][2] = { - {0, 0}, - {0, 2}, - {1, 0}, - {1, 2}, - {1, 1}, - {1, 3} -}; - -static const int step_offsets[6][2] = { - {0, 2}, - {0, 2}, - {-1, 2}, - {-1, 2}, - {-1, 1}, - {-1, 1} -}; - - -static boolean -i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) -{ - struct pipe_texture *pt = &tex->base; - unsigned level; - - switch (pt->target) { - case PIPE_TEXTURE_CUBE: { - const unsigned dim = pt->width[0]; - unsigned face; - unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* double pitch for cube layouts */ - tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; - tex->total_height = dim * 4; - - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_level_info(tex, level, 6, - 0, 0, - /*OLD: tex->pitch, tex->total_height,*/ - lvlWidth, lvlHeight, - 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * dim; - unsigned y = initial_offsets[face][1] * dim; - unsigned d = dim; - - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_image_offset(tex, level, face, x, y); - d >>= 1; - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - } - } - break; - } - case PIPE_TEXTURE_3D:{ - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - unsigned stack_height = 0; - - /* Calculate the size of a single slice. - */ - tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; - - /* XXX: hardware expects/requires 9 levels at minimum. - */ - for (level = 0; level <= MAX2(8, pt->last_level); - level++) { - i915_miptree_set_level_info(tex, level, depth, 0, tex->total_height, - width, height, depth); - - - stack_height += MAX2(2, height); - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - - /* Fixup depth image_offsets: - */ - depth = pt->depth[0]; - for (level = 0; level <= pt->last_level; level++) { - unsigned i; - for (i = 0; i < depth; i++) - i915_miptree_set_image_offset(tex, level, i, - 0, i * stack_height); - - depth = minify(depth); - } - - - /* Multiply slice size by texture depth for total size. It's - * remarkable how wasteful of memory the i915 texture layouts - * are. They are largely fixed in the i945. - */ - tex->total_height = stack_height * pt->depth[0]; - break; - } - - default:{ - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned img_height; - - tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; - tex->total_height = 0; - - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_level_info(tex, level, 1, - 0, tex->total_height, - width, height, 1); - - if (pt->compressed) - img_height = MAX2(1, height / 4); - else - img_height = (MAX2(2, height) + 1) & ~1; - - tex->total_height += img_height; - - width = minify(width); - height = minify(height); - } - break; - } - } - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - tex->pitch, - tex->total_height, pt->cpp, tex->pitch * tex->total_height * pt->cpp); - */ - - return TRUE; -} - - -static boolean -i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) -{ - struct pipe_texture *pt = &tex->base; - unsigned level; - - switch (pt->target) { - case PIPE_TEXTURE_CUBE:{ - const unsigned dim = pt->width[0]; - unsigned face; - unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* Depending on the size of the largest images, pitch can be - * determined either by the old-style packing of cubemap faces, - * or the final row of 4x4, 2x2 and 1x1 faces below this. - */ - if (dim > 32) - tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; - else - tex->pitch = 14 * 8; - - tex->total_height = dim * 4 + 4; - - /* Set all the levels to effectively occupy the whole rectangular region. - */ - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_level_info(tex, level, 6, - 0, 0, - lvlWidth, lvlHeight, 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - - for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * dim; - unsigned y = initial_offsets[face][1] * dim; - unsigned d = dim; - - if (dim == 4 && face >= 4) { - y = tex->total_height - 4; - x = (face - 4) * 8; - } - else if (dim < 4 && (face > 0)) { - y = tex->total_height - 4; - x = face * 8; - } - - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_image_offset(tex, level, face, x, y); - - d >>= 1; - - switch (d) { - case 4: - switch (face) { - case PIPE_TEX_FACE_POS_X: - case PIPE_TEX_FACE_NEG_X: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - case PIPE_TEX_FACE_POS_Y: - case PIPE_TEX_FACE_NEG_Y: - y += 12; - x -= 8; - break; - case PIPE_TEX_FACE_POS_Z: - case PIPE_TEX_FACE_NEG_Z: - y = tex->total_height - 4; - x = (face - 4) * 8; - break; - } - - case 2: - y = tex->total_height - 4; - x = 16 + face * 8; - break; - - case 1: - x += 48; - break; - - default: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - } - } - } - break; - } - case PIPE_TEXTURE_3D:{ - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - unsigned pack_x_pitch, pack_x_nr; - unsigned pack_y_pitch; - unsigned level; - - tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; - tex->total_height = 0; - - pack_y_pitch = MAX2(pt->height[0], 2); - pack_x_pitch = tex->pitch; - pack_x_nr = 1; - - for (level = 0; level <= pt->last_level; level++) { - unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; - int x = 0; - int y = 0; - unsigned q, j; - - i915_miptree_set_level_info(tex, level, nr_images, - 0, tex->total_height, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - i915_miptree_set_image_offset(tex, level, q, x, y); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - tex->total_height += y; - - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= tex->pitch); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - } - - width = minify(width); - height = minify(height); - depth = minify(depth); - } - break; - } - - case PIPE_TEXTURE_1D: - case PIPE_TEXTURE_2D: -// case PIPE_TEXTURE_RECTANGLE: - i945_miptree_layout_2d(tex); - break; - default: - assert(0); - return FALSE; - } - - /* - DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - tex->pitch, - tex->total_height, pt->cpp, tex->pitch * tex->total_height * pt->cpp); - */ - - return TRUE; -} - - -struct pipe_texture * -i915_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat) -{ - struct i915_texture *tex = CALLOC_STRUCT(i915_texture); - - if (tex) { - struct i915_context *i915 = i915_context(pipe); - - tex->base = *templat; - - if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : - i915_miptree_layout(pipe, tex)) - tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, - PIPE_BUFFER_USAGE_PIXEL, - tex->pitch * tex->base.cpp * - tex->total_height); - - if (!tex->buffer) { - FREE(tex); - return NULL; - } - } - - return &tex->base; -} - - -void -i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) -{ - if (!*pt) - return; - - /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); - */ - if (--(*pt)->refcount <= 0) { - struct i915_texture *tex = (struct i915_texture *)*pt; - uint i; - - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); - */ - - pipe_buffer_reference(pipe->winsys, &tex->buffer, NULL); - - for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) - if (tex->image_offset[i]) - FREE(tex->image_offset[i]); - - FREE(tex); - } - *pt = NULL; -} diff --git a/src/mesa/pipe/i915simple/i915_texture.h b/src/mesa/pipe/i915simple/i915_texture.h deleted file mode 100644 index 330d111dc7..0000000000 --- a/src/mesa/pipe/i915simple/i915_texture.h +++ /dev/null @@ -1,17 +0,0 @@ - -#ifndef I915_TEXTURE_H -#define I915_TEXTURE_H - -struct pipe_context; -struct pipe_texture; - - -struct pipe_texture * -i915_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat); - -extern void -i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); - - -#endif /* I915_TEXTURE_H */ diff --git a/src/mesa/pipe/i915simple/i915_winsys.h b/src/mesa/pipe/i915simple/i915_winsys.h deleted file mode 100644 index fe49710852..0000000000 --- a/src/mesa/pipe/i915simple/i915_winsys.h +++ /dev/null @@ -1,115 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * This is the interface that i915simple requires any window system - * hosting it to implement. This is the only include file in i915simple - * which is public. - * - */ - -#ifndef I915_WINSYS_H -#define I915_WINSYS_H - - -#include "pipe/p_defines.h" - - -/* Pipe drivers are (meant to be!) independent of both GL and the - * window system. The window system provides a buffer manager and a - * set of additional hooks for things like command buffer submission, - * etc. - * - * There clearly has to be some agreement between the window system - * driver and the hardware driver about the format of command buffers, - * etc. - */ - -struct pipe_buffer; -struct pipe_winsys; - - -/** - * Additional winsys interface for i915simple. - * - * It is an over-simple batchbuffer mechanism. Will want to improve the - * performance of this, perhaps based on the cmdstream stuff. It - * would be pretty impossible to implement swz on top of this - * interface. - * - * Will also need additions/changes to implement static/dynamic - * indirect state. - */ -struct i915_winsys { - - /** - * Reserve space on batch buffer. - * - * Returns a null pointer if there is insufficient space in the batch buffer - * to hold the requested number of dwords and relocations. - * - * The number of dwords should also include the number of relocations. - */ - unsigned *(*batch_start)( struct i915_winsys *sws, - unsigned dwords, - unsigned relocs ); - - void (*batch_dword)( struct i915_winsys *sws, - unsigned dword ); - - /** - * Emit a relocation to a buffer. - * - * Used not only when the buffer addresses are not pinned, but also to - * ensure refered buffers will not be destroyed until the current batch - * buffer execution is finished. - * - * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and - * I915_BUFFER_ACCESS_READ macros. - */ - void (*batch_reloc)( struct i915_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta ); - - void (*batch_flush)( struct i915_winsys *sws ); - void (*batch_finish)( struct i915_winsys *sws ); -}; - -#define I915_BUFFER_ACCESS_WRITE 0x1 -#define I915_BUFFER_ACCESS_READ 0x2 - -#define I915_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) - - -struct pipe_context *i915_create( struct pipe_winsys *, - struct i915_winsys *, - unsigned pci_id ); - - -#endif diff --git a/src/mesa/pipe/i965simple/Makefile b/src/mesa/pipe/i965simple/Makefile deleted file mode 100644 index 48c00ab50b..0000000000 --- a/src/mesa/pipe/i965simple/Makefile +++ /dev/null @@ -1,66 +0,0 @@ - -TOP = ../../../.. -include $(TOP)/configs/current - -LIBNAME = i965simple - -DRIVER_SOURCES = \ - brw_blit.c \ - brw_flush.c \ - brw_strings.c \ - brw_surface.c \ - brw_cc.c \ - brw_clip.c \ - brw_clip_line.c \ - brw_clip_point.c \ - brw_clip_state.c \ - brw_clip_tri.c \ - brw_clip_util.c \ - brw_context.c \ - brw_curbe.c \ - brw_draw.c \ - brw_draw_upload.c \ - brw_eu.c \ - brw_eu_debug.c \ - brw_eu_emit.c \ - brw_eu_util.c \ - brw_gs.c \ - brw_gs_emit.c \ - brw_gs_state.c \ - brw_misc_state.c \ - brw_sf.c \ - brw_sf_emit.c \ - brw_sf_state.c \ - brw_shader_info.c \ - brw_state.c \ - brw_state_batch.c \ - brw_state_cache.c \ - brw_state_pool.c \ - brw_state_upload.c \ - brw_tex_layout.c \ - brw_urb.c \ - brw_util.c \ - brw_vs.c \ - brw_vs_emit.c \ - brw_vs_state.c \ - brw_wm.c \ - brw_wm_iz.c \ - brw_wm_decl.c \ - brw_wm_glsl.c \ - brw_wm_sampler_state.c \ - brw_wm_state.c \ - brw_wm_surface_state.c - -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(COMMON_BM_SOURCES) \ - $(MINIGLX_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -DRIVER_DEFINES = -I. - -include ../Makefile.template - -symlinks: diff --git a/src/mesa/pipe/i965simple/SConscript b/src/mesa/pipe/i965simple/SConscript deleted file mode 100644 index 74621de84c..0000000000 --- a/src/mesa/pipe/i965simple/SConscript +++ /dev/null @@ -1,55 +0,0 @@ -Import('*') - -env = env.Clone() - -i965simple = env.ConvenienceLibrary( - target = 'i965simple', - source = [ - 'brw_blit.c', - 'brw_cc.c', - 'brw_clip.c', - 'brw_clip_line.c', - 'brw_clip_point.c', - 'brw_clip_state.c', - 'brw_clip_tri.c', - 'brw_clip_util.c', - 'brw_context.c', - 'brw_curbe.c', - 'brw_draw.c', - 'brw_draw_upload.c', - 'brw_eu.c', - 'brw_eu_debug.c', - 'brw_eu_emit.c', - 'brw_eu_util.c', - 'brw_flush.c', - 'brw_gs.c', - 'brw_gs_emit.c', - 'brw_gs_state.c', - 'brw_misc_state.c', - 'brw_sf.c', - 'brw_sf_emit.c', - 'brw_sf_state.c', - 'brw_shader_info.c', - 'brw_state.c', - 'brw_state_batch.c', - 'brw_state_cache.c', - 'brw_state_pool.c', - 'brw_state_upload.c', - 'brw_strings.c', - 'brw_surface.c', - 'brw_tex_layout.c', - 'brw_urb.c', - 'brw_util.c', - 'brw_vs.c', - 'brw_vs_emit.c', - 'brw_vs_state.c', - 'brw_wm.c', - 'brw_wm_decl.c', - 'brw_wm_glsl.c', - 'brw_wm_iz.c', - 'brw_wm_sampler_state.c', - 'brw_wm_state.c', - 'brw_wm_surface_state.c', - ]) - -Export('i965simple') diff --git a/src/mesa/pipe/i965simple/brw_batch.h b/src/mesa/pipe/i965simple/brw_batch.h deleted file mode 100644 index 5f5932a488..0000000000 --- a/src/mesa/pipe/i965simple/brw_batch.h +++ /dev/null @@ -1,59 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef BRW_BATCH_H -#define BRW_BATCH_H - -#include "brw_winsys.h" - -#define BATCH_LOCALS - -#define INTEL_BATCH_NO_CLIPRECTS 0x1 -#define INTEL_BATCH_CLIPRECTS 0x2 - -#define BEGIN_BATCH( dwords, relocs ) \ - brw->winsys->batch_start(brw->winsys, dwords, relocs) - -#define OUT_BATCH( dword ) \ - brw->winsys->batch_dword(brw->winsys, dword) - -#define OUT_RELOC( buf, flags, delta ) \ - brw->winsys->batch_reloc(brw->winsys, buf, flags, delta) - -#define ADVANCE_BATCH() \ - brw->winsys->batch_end( brw->winsys ) - -/* XXX: this is bogus - need proper handling for out-of-memory in batchbuffer. - */ -#define FLUSH_BATCH(fence) do { \ - brw->winsys->batch_flush(brw->winsys, fence); \ - brw->hardware_dirty = ~0; \ -} while (0) - -#define BRW_BATCH_STRUCT(brw, s) brw_batchbuffer_data( brw->winsys, (s), sizeof(*(s))) - -#endif diff --git a/src/mesa/pipe/i965simple/brw_blit.c b/src/mesa/pipe/i965simple/brw_blit.c deleted file mode 100644 index 8494f70493..0000000000 --- a/src/mesa/pipe/i965simple/brw_blit.c +++ /dev/null @@ -1,218 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include -#include - -#include "brw_batch.h" -#include "brw_blit.h" -#include "brw_context.h" -#include "brw_reg.h" - -#include "pipe/p_context.h" -#include "pipe/p_winsys.h" - -#define FILE_DEBUG_FLAG DEBUG_BLIT - -void brw_fill_blit(struct brw_context *brw, - unsigned cpp, - short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - boolean dst_tiled, - short x, short y, - short w, short h, - unsigned color) -{ - unsigned BR13, CMD; - BATCH_LOCALS; - - dst_pitch *= cpp; - - switch(cpp) { - case 1: - case 2: - case 3: - BR13 = (0xF0 << 16) | (1<<24); - CMD = XY_COLOR_BLT_CMD; - break; - case 4: - BR13 = (0xF0 << 16) | (1<<24) | (1<<25); - CMD = XY_COLOR_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; - break; - default: - return; - } - - if (dst_tiled) { - CMD |= XY_DST_TILED; - dst_pitch /= 4; - } - - BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH( CMD ); - OUT_BATCH( dst_pitch | BR13 ); - OUT_BATCH( (y << 16) | x ); - OUT_BATCH( ((y+h) << 16) | (x+w) ); - OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, dst_offset ); - OUT_BATCH( color ); - ADVANCE_BATCH(); -} - -static unsigned translate_raster_op(unsigned logicop) -{ - switch(logicop) { - case PIPE_LOGICOP_CLEAR: return 0x00; - case PIPE_LOGICOP_AND: return 0x88; - case PIPE_LOGICOP_AND_REVERSE: return 0x44; - case PIPE_LOGICOP_COPY: return 0xCC; - case PIPE_LOGICOP_AND_INVERTED: return 0x22; - case PIPE_LOGICOP_NOOP: return 0xAA; - case PIPE_LOGICOP_XOR: return 0x66; - case PIPE_LOGICOP_OR: return 0xEE; - case PIPE_LOGICOP_NOR: return 0x11; - case PIPE_LOGICOP_EQUIV: return 0x99; - case PIPE_LOGICOP_INVERT: return 0x55; - case PIPE_LOGICOP_OR_REVERSE: return 0xDD; - case PIPE_LOGICOP_COPY_INVERTED: return 0x33; - case PIPE_LOGICOP_OR_INVERTED: return 0xBB; - case PIPE_LOGICOP_NAND: return 0x77; - case PIPE_LOGICOP_SET: return 0xFF; - default: return 0; - } -} - - -/* Copy BitBlt - */ -void brw_copy_blit(struct brw_context *brw, - unsigned do_flip, - unsigned cpp, - short src_pitch, - struct pipe_buffer *src_buffer, - unsigned src_offset, - boolean src_tiled, - short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - boolean dst_tiled, - short src_x, short src_y, - short dst_x, short dst_y, - short w, short h, - unsigned logic_op) -{ - unsigned CMD, BR13; - int dst_y2 = dst_y + h; - int dst_x2 = dst_x + w; - BATCH_LOCALS; - - - DBG("%s src:buf(%d)/%d %d,%d dst:buf(%d)/%d %d,%d sz:%dx%d op:%d\n", - __FUNCTION__, - src_buffer, src_pitch, src_x, src_y, - dst_buffer, dst_pitch, dst_x, dst_y, - w,h,logic_op); - - assert( logic_op - PIPE_LOGICOP_CLEAR >= 0 ); - assert( logic_op - PIPE_LOGICOP_CLEAR < 0x10 ); - - src_pitch *= cpp; - dst_pitch *= cpp; - - switch(cpp) { - case 1: - case 2: - case 3: - BR13 = (translate_raster_op(logic_op) << 16) | (1<<24); - CMD = XY_SRC_COPY_BLT_CMD; - break; - case 4: - BR13 = (translate_raster_op(logic_op) << 16) | (1<<24) | - (1<<25); - CMD = XY_SRC_COPY_BLT_CMD | XY_BLT_WRITE_ALPHA | XY_BLT_WRITE_RGB; - break; - default: - return; - } - - if (src_tiled) { - CMD |= XY_SRC_TILED; - src_pitch /= 4; - } - - if (dst_tiled) { - CMD |= XY_DST_TILED; - dst_pitch /= 4; - } - - if (dst_y2 < dst_y || - dst_x2 < dst_x) { - return; - } - - dst_pitch &= 0xffff; - src_pitch &= 0xffff; - - /* Initial y values don't seem to work with negative pitches. If - * we adjust the offsets manually (below), it seems to work fine. - * - * On the other hand, if we always adjust, the hardware doesn't - * know which blit directions to use, so overlapping copypixels get - * the wrong result. - */ - if (dst_pitch > 0 && src_pitch > 0) { - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH( CMD ); - OUT_BATCH( dst_pitch | BR13 ); - OUT_BATCH( (dst_y << 16) | dst_x ); - OUT_BATCH( (dst_y2 << 16) | dst_x2 ); - OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, - dst_offset ); - OUT_BATCH( (src_y << 16) | src_x ); - OUT_BATCH( src_pitch ); - OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ, - src_offset ); - ADVANCE_BATCH(); - } - else { - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH( CMD ); - OUT_BATCH( (dst_pitch & 0xffff) | BR13 ); - OUT_BATCH( (0 << 16) | dst_x ); - OUT_BATCH( (h << 16) | dst_x2 ); - OUT_RELOC( dst_buffer, BRW_BUFFER_ACCESS_WRITE, - dst_offset + dst_y * dst_pitch ); - OUT_BATCH( (src_pitch & 0xffff) ); - OUT_RELOC( src_buffer, BRW_BUFFER_ACCESS_READ, - src_offset + src_y * src_pitch ); - ADVANCE_BATCH(); - } -} - - - diff --git a/src/mesa/pipe/i965simple/brw_blit.h b/src/mesa/pipe/i965simple/brw_blit.h deleted file mode 100644 index 111c5d91d3..0000000000 --- a/src/mesa/pipe/i965simple/brw_blit.h +++ /dev/null @@ -1,33 +0,0 @@ -#ifndef BRW_BLIT_H -#define BRW_BLIT_H - -#include "pipe/p_compiler.h" - -struct pipe_buffer; -struct brw_context; - -void brw_fill_blit(struct brw_context *intel, - unsigned cpp, - short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - boolean dst_tiled, - short x, short y, - short w, short h, - unsigned color); -void brw_copy_blit(struct brw_context *intel, - unsigned do_flip, - unsigned cpp, - short src_pitch, - struct pipe_buffer *src_buffer, - unsigned src_offset, - boolean src_tiled, - short dst_pitch, - struct pipe_buffer *dst_buffer, - unsigned dst_offset, - boolean dst_tiled, - short src_x, short src_y, - short dst_x, short dst_y, - short w, short h, - unsigned logic_op); -#endif diff --git a/src/mesa/pipe/i965simple/brw_cc.c b/src/mesa/pipe/i965simple/brw_cc.c deleted file mode 100644 index 337e4f95f6..0000000000 --- a/src/mesa/pipe/i965simple/brw_cc.c +++ /dev/null @@ -1,269 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_util.h" - -#include "brw_context.h" -#include "brw_state.h" -#include "brw_defines.h" -#include "brw_util.h" - - -static int brw_translate_compare_func(int func) -{ - switch(func) { - case PIPE_FUNC_NEVER: - return BRW_COMPAREFUNCTION_NEVER; - case PIPE_FUNC_LESS: - return BRW_COMPAREFUNCTION_LESS; - case PIPE_FUNC_LEQUAL: - return BRW_COMPAREFUNCTION_LEQUAL; - case PIPE_FUNC_GREATER: - return BRW_COMPAREFUNCTION_GREATER; - case PIPE_FUNC_GEQUAL: - return BRW_COMPAREFUNCTION_GEQUAL; - case PIPE_FUNC_NOTEQUAL: - return BRW_COMPAREFUNCTION_NOTEQUAL; - case PIPE_FUNC_EQUAL: - return BRW_COMPAREFUNCTION_EQUAL; - case PIPE_FUNC_ALWAYS: - return BRW_COMPAREFUNCTION_ALWAYS; - } - - debug_printf("Unknown value in %s: %x\n", __FUNCTION__, func); - return BRW_COMPAREFUNCTION_ALWAYS; -} - -static int brw_translate_stencil_op(int op) -{ - switch(op) { - case PIPE_STENCIL_OP_KEEP: - return BRW_STENCILOP_KEEP; - case PIPE_STENCIL_OP_ZERO: - return BRW_STENCILOP_ZERO; - case PIPE_STENCIL_OP_REPLACE: - return BRW_STENCILOP_REPLACE; - case PIPE_STENCIL_OP_INCR: - return BRW_STENCILOP_INCRSAT; - case PIPE_STENCIL_OP_DECR: - return BRW_STENCILOP_DECRSAT; - case PIPE_STENCIL_OP_INCR_WRAP: - return BRW_STENCILOP_INCR; - case PIPE_STENCIL_OP_DECR_WRAP: - return BRW_STENCILOP_DECR; - case PIPE_STENCIL_OP_INVERT: - return BRW_STENCILOP_INVERT; - default: - return BRW_STENCILOP_ZERO; - } -} - - -static int brw_translate_logic_op(int opcode) -{ - switch(opcode) { - case PIPE_LOGICOP_CLEAR: - return BRW_LOGICOPFUNCTION_CLEAR; - case PIPE_LOGICOP_AND: - return BRW_LOGICOPFUNCTION_AND; - case PIPE_LOGICOP_AND_REVERSE: - return BRW_LOGICOPFUNCTION_AND_REVERSE; - case PIPE_LOGICOP_COPY: - return BRW_LOGICOPFUNCTION_COPY; - case PIPE_LOGICOP_COPY_INVERTED: - return BRW_LOGICOPFUNCTION_COPY_INVERTED; - case PIPE_LOGICOP_AND_INVERTED: - return BRW_LOGICOPFUNCTION_AND_INVERTED; - case PIPE_LOGICOP_NOOP: - return BRW_LOGICOPFUNCTION_NOOP; - case PIPE_LOGICOP_XOR: - return BRW_LOGICOPFUNCTION_XOR; - case PIPE_LOGICOP_OR: - return BRW_LOGICOPFUNCTION_OR; - case PIPE_LOGICOP_OR_INVERTED: - return BRW_LOGICOPFUNCTION_OR_INVERTED; - case PIPE_LOGICOP_NOR: - return BRW_LOGICOPFUNCTION_NOR; - case PIPE_LOGICOP_EQUIV: - return BRW_LOGICOPFUNCTION_EQUIV; - case PIPE_LOGICOP_INVERT: - return BRW_LOGICOPFUNCTION_INVERT; - case PIPE_LOGICOP_OR_REVERSE: - return BRW_LOGICOPFUNCTION_OR_REVERSE; - case PIPE_LOGICOP_NAND: - return BRW_LOGICOPFUNCTION_NAND; - case PIPE_LOGICOP_SET: - return BRW_LOGICOPFUNCTION_SET; - default: - return BRW_LOGICOPFUNCTION_SET; - } -} - - -static void upload_cc_vp( struct brw_context *brw ) -{ - struct brw_cc_viewport ccv; - - memset(&ccv, 0, sizeof(ccv)); - - ccv.min_depth = 0.0; - ccv.max_depth = 1.0; - - brw->cc.vp_gs_offset = brw_cache_data( &brw->cache[BRW_CC_VP], &ccv ); -} - -const struct brw_tracked_state brw_cc_vp = { - .dirty = { - .brw = BRW_NEW_SCENE, - .cache = 0 - }, - .update = upload_cc_vp -}; - - -static void upload_cc_unit( struct brw_context *brw ) -{ - struct brw_cc_unit_state cc; - - memset(&cc, 0, sizeof(cc)); - - /* BRW_NEW_DEPTH_STENCIL */ - if (brw->attribs.DepthStencil->stencil[0].enabled) { - cc.cc0.stencil_enable = brw->attribs.DepthStencil->stencil[0].enabled; - cc.cc0.stencil_func = brw_translate_compare_func(brw->attribs.DepthStencil->stencil[0].func); - cc.cc0.stencil_fail_op = brw_translate_stencil_op(brw->attribs.DepthStencil->stencil[0].fail_op); - cc.cc0.stencil_pass_depth_fail_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil[0].zfail_op); - cc.cc0.stencil_pass_depth_pass_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil[0].zpass_op); - cc.cc1.stencil_ref = brw->attribs.DepthStencil->stencil[0].ref_value; - cc.cc1.stencil_write_mask = brw->attribs.DepthStencil->stencil[0].write_mask; - cc.cc1.stencil_test_mask = brw->attribs.DepthStencil->stencil[0].value_mask; - - if (brw->attribs.DepthStencil->stencil[1].enabled) { - cc.cc0.bf_stencil_enable = brw->attribs.DepthStencil->stencil[1].enabled; - cc.cc0.bf_stencil_func = brw_translate_compare_func( - brw->attribs.DepthStencil->stencil[1].func); - cc.cc0.bf_stencil_fail_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil[1].fail_op); - cc.cc0.bf_stencil_pass_depth_fail_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil[1].zfail_op); - cc.cc0.bf_stencil_pass_depth_pass_op = brw_translate_stencil_op( - brw->attribs.DepthStencil->stencil[1].zpass_op); - cc.cc1.bf_stencil_ref = brw->attribs.DepthStencil->stencil[1].ref_value; - cc.cc2.bf_stencil_write_mask = brw->attribs.DepthStencil->stencil[1].write_mask; - cc.cc2.bf_stencil_test_mask = brw->attribs.DepthStencil->stencil[1].value_mask; - } - - /* Not really sure about this: - */ - if (brw->attribs.DepthStencil->stencil[0].write_mask || - brw->attribs.DepthStencil->stencil[1].write_mask) - cc.cc0.stencil_write_enable = 1; - } - - /* BRW_NEW_BLEND */ - if (brw->attribs.Blend->logicop_enable) { - cc.cc2.logicop_enable = 1; - cc.cc5.logicop_func = brw_translate_logic_op( brw->attribs.Blend->logicop_func ); - } - else if (brw->attribs.Blend->blend_enable) { - int eqRGB = brw->attribs.Blend->rgb_func; - int eqA = brw->attribs.Blend->alpha_func; - int srcRGB = brw->attribs.Blend->rgb_src_factor; - int dstRGB = brw->attribs.Blend->rgb_dst_factor; - int srcA = brw->attribs.Blend->alpha_src_factor; - int dstA = brw->attribs.Blend->alpha_dst_factor; - - if (eqRGB == PIPE_BLEND_MIN || eqRGB == PIPE_BLEND_MAX) { - srcRGB = dstRGB = PIPE_BLENDFACTOR_ONE; - } - - if (eqA == PIPE_BLEND_MIN || eqA == PIPE_BLEND_MAX) { - srcA = dstA = PIPE_BLENDFACTOR_ONE; - } - - cc.cc6.dest_blend_factor = brw_translate_blend_factor(dstRGB); - cc.cc6.src_blend_factor = brw_translate_blend_factor(srcRGB); - cc.cc6.blend_function = brw_translate_blend_equation( eqRGB ); - - cc.cc5.ia_dest_blend_factor = brw_translate_blend_factor(dstA); - cc.cc5.ia_src_blend_factor = brw_translate_blend_factor(srcA); - cc.cc5.ia_blend_function = brw_translate_blend_equation( eqA ); - - cc.cc3.blend_enable = 1; - cc.cc3.ia_blend_enable = (srcA != srcRGB || - dstA != dstRGB || - eqA != eqRGB); - } - - /* BRW_NEW_ALPHATEST - */ - if (brw->attribs.DepthStencil->alpha.enabled) { - cc.cc3.alpha_test = 1; - cc.cc3.alpha_test_func = - brw_translate_compare_func(brw->attribs.DepthStencil->alpha.func); - - UNCLAMPED_FLOAT_TO_UBYTE(cc.cc7.alpha_ref.ub[0], - brw->attribs.DepthStencil->alpha.ref); - - cc.cc3.alpha_test_format = BRW_ALPHATEST_FORMAT_UNORM8; - } - - if (brw->attribs.Blend->dither) { - cc.cc5.dither_enable = 1; - cc.cc6.y_dither_offset = 0; - cc.cc6.x_dither_offset = 0; - } - - if (brw->attribs.DepthStencil->depth.enabled) { - cc.cc2.depth_test = brw->attribs.DepthStencil->depth.enabled; - cc.cc2.depth_test_function = brw_translate_compare_func(brw->attribs.DepthStencil->depth.func); - cc.cc2.depth_write_enable = brw->attribs.DepthStencil->depth.writemask; - } - - /* CACHE_NEW_CC_VP */ - cc.cc4.cc_viewport_state_offset = brw->cc.vp_gs_offset >> 5; - - if (BRW_DEBUG & DEBUG_STATS) - cc.cc5.statistics_enable = 1; - - brw->cc.state_gs_offset = brw_cache_data( &brw->cache[BRW_CC_UNIT], &cc ); -} - -const struct brw_tracked_state brw_cc_unit = { - .dirty = { - .brw = BRW_NEW_DEPTH_STENCIL | BRW_NEW_BLEND | BRW_NEW_ALPHA_TEST, - .cache = CACHE_NEW_CC_VP - }, - .update = upload_cc_unit -}; - diff --git a/src/mesa/pipe/i965simple/brw_clip.c b/src/mesa/pipe/i965simple/brw_clip.c deleted file mode 100644 index 268124cc53..0000000000 --- a/src/mesa/pipe/i965simple/brw_clip.c +++ /dev/null @@ -1,206 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_state.h" -#include "brw_clip.h" - -#define FRONT_UNFILLED_BIT 0x1 -#define BACK_UNFILLED_BIT 0x2 - - -static void compile_clip_prog( struct brw_context *brw, - struct brw_clip_prog_key *key ) -{ - struct brw_clip_compile c; - const unsigned *program; - unsigned program_size; - unsigned delta; - unsigned i; - - memset(&c, 0, sizeof(c)); - - /* Begin the compilation: - */ - brw_init_compile(&c.func); - - c.func.single_program_flow = 1; - - c.key = *key; - - - /* Need to locate the two positions present in vertex + header. - * These are currently hardcoded: - */ - c.header_position_offset = ATTR_SIZE; - - for (i = 0, delta = REG_SIZE; i < PIPE_MAX_SHADER_OUTPUTS; i++) - if (c.key.attrs & (1<primitive) { - case PIPE_PRIM_TRIANGLES: -#if 0 - if (key->do_unfilled) - brw_emit_unfilled_clip( &c ); - else -#endif - brw_emit_tri_clip( &c ); - break; - case PIPE_PRIM_LINES: - brw_emit_line_clip( &c ); - break; - case PIPE_PRIM_POINTS: - brw_emit_point_clip( &c ); - break; - default: - assert(0); - return; - } - - - - /* get the program - */ - program = brw_get_program(&c.func, &program_size); - - /* Upload - */ - brw->clip.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_CLIP_PROG], - &c.key, - sizeof(c.key), - program, - program_size, - &c.prog_data, - &brw->clip.prog_data ); -} - - -static boolean search_cache( struct brw_context *brw, - struct brw_clip_prog_key *key ) -{ - return brw_search_cache(&brw->cache[BRW_CLIP_PROG], - key, sizeof(*key), - &brw->clip.prog_data, - &brw->clip.prog_gs_offset); -} - - - - -/* Calculate interpolants for triangle and line rasterization. - */ -static void upload_clip_prog(struct brw_context *brw) -{ - struct brw_clip_prog_key key; - - memset(&key, 0, sizeof(key)); - - /* Populate the key: - */ - /* BRW_NEW_REDUCED_PRIMITIVE */ - key.primitive = brw->reduced_primitive; - /* CACHE_NEW_VS_PROG */ - key.attrs = brw->vs.prog_data->outputs_written; - /* BRW_NEW_RASTER */ - key.do_flat_shading = (brw->attribs.Raster->flatshade); - /* BRW_NEW_CLIP */ - key.nr_userclip = brw->attribs.Clip.nr; /* XXX */ - -#if 0 - key.clip_mode = BRW_CLIPMODE_NORMAL; - - if (key.primitive == PIPE_PRIM_TRIANGLES) { - if (brw->attribs.Raster->cull_mode == PIPE_WINDING_BOTH) - key.clip_mode = BRW_CLIPMODE_REJECT_ALL; - else { - if (brw->attribs.Raster->fill_cw != PIPE_POLYGON_MODE_FILL || - brw->attribs.Raster->fill_ccw != PIPE_POLYGON_MODE_FILL) - key.do_unfilled = 1; - - /* Most cases the fixed function units will handle. Cases where - * one or more polygon faces are unfilled will require help: - */ - if (key.do_unfilled) { - key.clip_mode = BRW_CLIPMODE_CLIP_NON_REJECTED; - - if (brw->attribs.Raster->offset_cw || - brw->attribs.Raster->offset_ccw) { - key.offset_units = brw->attribs.Raster->offset_units; - key.offset_factor = brw->attribs.Raster->offset_scale; - } - key.fill_ccw = brw->attribs.Raster->fill_ccw; - key.fill_cw = brw->attribs.Raster->fill_cw; - key.offset_ccw = brw->attribs.Raster->offset_ccw; - key.offset_cw = brw->attribs.Raster->offset_cw; - if (brw->attribs.Raster->light_twoside && - key.fill_cw != CLIP_CULL) - key.copy_bfc_cw = 1; - } - } - } -#else - key.clip_mode = BRW_CLIPMODE_ACCEPT_ALL; -#endif - - if (!search_cache(brw, &key)) - compile_clip_prog( brw, &key ); -} - -const struct brw_tracked_state brw_clip_prog = { - .dirty = { - .brw = (BRW_NEW_RASTERIZER | - BRW_NEW_CLIP | - BRW_NEW_REDUCED_PRIMITIVE), - .cache = CACHE_NEW_VS_PROG - }, - .update = upload_clip_prog -}; diff --git a/src/mesa/pipe/i965simple/brw_clip.h b/src/mesa/pipe/i965simple/brw_clip.h deleted file mode 100644 index a89d08b791..0000000000 --- a/src/mesa/pipe/i965simple/brw_clip.h +++ /dev/null @@ -1,170 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#ifndef BRW_CLIP_H -#define BRW_CLIP_H - - -#include "brw_context.h" -#include "brw_eu.h" - -#define MAX_VERTS (3+6+6) - -/* Note that if unfilled primitives are being emitted, we have to fix - * up polygon offset and flatshading at this point: - */ -struct brw_clip_prog_key { - unsigned attrs:32; - unsigned primitive:4; - unsigned nr_userclip:3; - unsigned do_flat_shading:1; - unsigned do_unfilled:1; - unsigned fill_cw:2; /* includes cull information */ - unsigned fill_ccw:2; /* includes cull information */ - unsigned offset_cw:1; - unsigned offset_ccw:1; - unsigned pad0:17; - - unsigned copy_bfc_cw:1; - unsigned copy_bfc_ccw:1; - unsigned clip_mode:3; - unsigned pad1:27; - - float offset_factor; - float offset_units; -}; - - -#define CLIP_LINE 0 -#define CLIP_POINT 1 -#define CLIP_FILL 2 -#define CLIP_CULL 3 - - -#define PRIM_MASK (0x1f) - -struct brw_clip_compile { - struct brw_compile func; - struct brw_clip_prog_key key; - struct brw_clip_prog_data prog_data; - - struct { - struct brw_reg R0; - struct brw_reg vertex[MAX_VERTS]; - - struct brw_reg t; - struct brw_reg t0, t1; - struct brw_reg dp0, dp1; - - struct brw_reg dpPrev; - struct brw_reg dp; - struct brw_reg loopcount; - struct brw_reg nr_verts; - struct brw_reg planemask; - - struct brw_reg inlist; - struct brw_reg outlist; - struct brw_reg freelist; - - struct brw_reg dir; - struct brw_reg tmp0, tmp1; - struct brw_reg offset; - - struct brw_reg fixed_planes; - struct brw_reg plane_equation; - } reg; - - /* 3 different ways of expressing vertex size: - */ - unsigned nr_attrs; - unsigned nr_regs; - unsigned nr_bytes; - - unsigned first_tmp; - unsigned last_tmp; - - boolean need_direction; - - unsigned last_mrf; - - unsigned header_position_offset; - unsigned offset[PIPE_ATTRIB_MAX]; -}; - -#define ATTR_SIZE (4*4) - -/* Points are only culled, so no need for a clip routine, however it - * works out easier to have a dummy one. - */ -void brw_emit_unfilled_clip( struct brw_clip_compile *c ); -void brw_emit_tri_clip( struct brw_clip_compile *c ); -void brw_emit_line_clip( struct brw_clip_compile *c ); -void brw_emit_point_clip( struct brw_clip_compile *c ); - -/* brw_clip_tri.c, for use by the unfilled clip routine: - */ -void brw_clip_tri_init_vertices( struct brw_clip_compile *c ); -void brw_clip_tri_flat_shade( struct brw_clip_compile *c ); -void brw_clip_tri( struct brw_clip_compile *c ); -void brw_clip_tri_emit_polygon( struct brw_clip_compile *c ); -void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, - unsigned nr_verts ); - - -/* Utils: - */ - -void brw_clip_interp_vertex( struct brw_clip_compile *c, - struct brw_indirect dest_ptr, - struct brw_indirect v0_ptr, /* from */ - struct brw_indirect v1_ptr, /* to */ - struct brw_reg t0, - boolean force_edgeflag ); - -void brw_clip_init_planes( struct brw_clip_compile *c ); - -void brw_clip_emit_vue(struct brw_clip_compile *c, - struct brw_indirect vert, - boolean allocate, - boolean eot, - unsigned header); - -void brw_clip_kill_thread(struct brw_clip_compile *c); - -struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c ); -struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c ); - -void brw_clip_copy_colors( struct brw_clip_compile *c, - unsigned to, unsigned from ); - -void brw_clip_init_clipmask( struct brw_clip_compile *c ); - -#endif diff --git a/src/mesa/pipe/i965simple/brw_clip_line.c b/src/mesa/pipe/i965simple/brw_clip_line.c deleted file mode 100644 index 75d9e5fcda..0000000000 --- a/src/mesa/pipe/i965simple/brw_clip_line.c +++ /dev/null @@ -1,245 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_clip.h" - - - -static void brw_clip_line_alloc_regs( struct brw_clip_compile *c ) -{ - unsigned i = 0,j; - - /* Register usage is static, precompute here: - */ - c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; - - if (c->key.nr_userclip) { - c->reg.fixed_planes = brw_vec4_grf(i, 0); - i += (6 + c->key.nr_userclip + 1) / 2; - - c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2; - } - else - c->prog_data.curb_read_length = 0; - - - /* Payload vertices plus space for more generated vertices: - */ - for (j = 0; j < 4; j++) { - c->reg.vertex[j] = brw_vec4_grf(i, 0); - i += c->nr_regs; - } - - c->reg.t = brw_vec1_grf(i, 0); - c->reg.t0 = brw_vec1_grf(i, 1); - c->reg.t1 = brw_vec1_grf(i, 2); - c->reg.planemask = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD); - c->reg.plane_equation = brw_vec4_grf(i, 4); - i++; - - c->reg.dp0 = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */ - c->reg.dp1 = brw_vec1_grf(i, 4); - i++; - - if (!c->key.nr_userclip) { - c->reg.fixed_planes = brw_vec8_grf(i, 0); - i++; - } - - - c->first_tmp = i; - c->last_tmp = i; - - c->prog_data.urb_read_length = c->nr_regs; /* ? */ - c->prog_data.total_grf = i; -} - - - -/* Line clipping, more or less following the following algorithm: - * - * for (p=0;p t1) t1 = t; - * } else { - * float t = dp0 / (dp0 - dp1); - * if (t > t0) t0 = t; - * } - * - * if (t0 + t1 >= 1.0) - * return; - * } - * } - * - * interp( ctx, newvtx0, vtx0, vtx1, t0 ); - * interp( ctx, newvtx1, vtx1, vtx0, t1 ); - * - */ -static void clip_and_emit_line( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_indirect vtx0 = brw_indirect(0, 0); - struct brw_indirect vtx1 = brw_indirect(1, 0); - struct brw_indirect newvtx0 = brw_indirect(2, 0); - struct brw_indirect newvtx1 = brw_indirect(3, 0); - struct brw_indirect plane_ptr = brw_indirect(4, 0); - struct brw_instruction *plane_loop; - struct brw_instruction *plane_active; - struct brw_instruction *is_negative; - struct brw_instruction *is_neg2; - struct brw_instruction *not_culled; - struct brw_reg v1_null_ud = retype(vec1(brw_null_reg()), BRW_REGISTER_TYPE_UD); - - brw_MOV(p, get_addr_reg(vtx0), brw_address(c->reg.vertex[0])); - brw_MOV(p, get_addr_reg(vtx1), brw_address(c->reg.vertex[1])); - brw_MOV(p, get_addr_reg(newvtx0), brw_address(c->reg.vertex[2])); - brw_MOV(p, get_addr_reg(newvtx1), brw_address(c->reg.vertex[3])); - brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c)); - - /* Note: init t0, t1 together: - */ - brw_MOV(p, vec2(c->reg.t0), brw_imm_f(0)); - - brw_clip_init_planes(c); - brw_clip_init_clipmask(c); - - /* -ve rhw workaround */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), - brw_imm_ud(1<<20)); - brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(0x3f)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - - plane_loop = brw_DO(p, BRW_EXECUTE_1); - { - /* if (planemask & 1) - */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_AND(p, v1_null_ud, c->reg.planemask, brw_imm_ud(1)); - - plane_active = brw_IF(p, BRW_EXECUTE_1); - { - if (c->key.nr_userclip) - brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0)); - else - brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0)); - -#if 0 - /* dp = DP4(vtx->position, plane) - */ - brw_DP4(p, vec4(c->reg.dp0), deref_4f(vtx0, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); - - /* if (IS_NEGATIVE(dp1)) - */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_L); - brw_DP4(p, vec4(c->reg.dp1), deref_4f(vtx1, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); -#else - #warning "disabled" -#endif - is_negative = brw_IF(p, BRW_EXECUTE_1); - { - brw_ADD(p, c->reg.t, c->reg.dp1, negate(c->reg.dp0)); - brw_math_invert(p, c->reg.t, c->reg.t); - brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp1); - - brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t1 ); - brw_MOV(p, c->reg.t1, c->reg.t); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - } - is_negative = brw_ELSE(p, is_negative); - { - /* Coming back in. We know that both cannot be negative - * because the line would have been culled in that case. - */ - - /* If both are positive, do nothing */ - brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.dp0, brw_imm_f(0.0)); - is_neg2 = brw_IF(p, BRW_EXECUTE_1); - { - brw_ADD(p, c->reg.t, c->reg.dp0, negate(c->reg.dp1)); - brw_math_invert(p, c->reg.t, c->reg.t); - brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp0); - - brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_G, c->reg.t, c->reg.t0 ); - brw_MOV(p, c->reg.t0, c->reg.t); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - } - brw_ENDIF(p, is_neg2); - } - brw_ENDIF(p, is_negative); - } - brw_ENDIF(p, plane_active); - - /* plane_ptr++; - */ - brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c)); - - /* while (planemask>>=1) != 0 - */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1)); - } - brw_WHILE(p, plane_loop); - - brw_ADD(p, c->reg.t, c->reg.t0, c->reg.t1); - brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.t, brw_imm_f(1.0)); - not_culled = brw_IF(p, BRW_EXECUTE_1); - { - brw_clip_interp_vertex(c, newvtx0, vtx0, vtx1, c->reg.t0, FALSE); - brw_clip_interp_vertex(c, newvtx1, vtx1, vtx0, c->reg.t1, FALSE); - - brw_clip_emit_vue(c, newvtx0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START); - brw_clip_emit_vue(c, newvtx1, 0, 1, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END); - } - brw_ENDIF(p, not_culled); - brw_clip_kill_thread(c); -} - - - -void brw_emit_line_clip( struct brw_clip_compile *c ) -{ - brw_clip_line_alloc_regs(c); - - if (c->key.do_flat_shading) - brw_clip_copy_colors(c, 0, 1); - - clip_and_emit_line(c); -} diff --git a/src/mesa/pipe/i965simple/brw_clip_point.c b/src/mesa/pipe/i965simple/brw_clip_point.c deleted file mode 100644 index 6fce7210d1..0000000000 --- a/src/mesa/pipe/i965simple/brw_clip_point.c +++ /dev/null @@ -1,47 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_clip.h" - - -/* Point clipping, nothing to do? - */ -void brw_emit_point_clip( struct brw_clip_compile *c ) -{ - /* Send an empty message to kill the thread: - */ - brw_clip_tri_alloc_regs(c, 0); - brw_clip_kill_thread(c); -} diff --git a/src/mesa/pipe/i965simple/brw_clip_state.c b/src/mesa/pipe/i965simple/brw_clip_state.c deleted file mode 100644 index ea5c05a279..0000000000 --- a/src/mesa/pipe/i965simple/brw_clip_state.c +++ /dev/null @@ -1,92 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_context.h" -#include "brw_state.h" -#include "brw_defines.h" -#include "pipe/p_util.h" - - -static void upload_clip_unit( struct brw_context *brw ) -{ - struct brw_clip_unit_state clip; - - memset(&clip, 0, sizeof(clip)); - - /* CACHE_NEW_CLIP_PROG */ - clip.thread0.grf_reg_count = - align(brw->clip.prog_data->total_grf, 16) / 16 - 1; - clip.thread0.kernel_start_pointer = brw->clip.prog_gs_offset >> 6; - clip.thread3.urb_entry_read_length = brw->clip.prog_data->urb_read_length; - clip.thread3.const_urb_entry_read_length = brw->clip.prog_data->curb_read_length; - clip.clip5.clip_mode = brw->clip.prog_data->clip_mode; - - /* BRW_NEW_CURBE_OFFSETS */ - clip.thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2; - - /* BRW_NEW_URB_FENCE */ - clip.thread4.nr_urb_entries = brw->urb.nr_clip_entries; - clip.thread4.urb_entry_allocation_size = brw->urb.vsize - 1; - clip.thread4.max_threads = 1; /* 2 threads */ - - if (BRW_DEBUG & DEBUG_STATS) - clip.thread4.stats_enable = 1; - - /* CONSTANT */ - clip.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; - clip.thread1.single_program_flow = 1; - clip.thread3.dispatch_grf_start_reg = 1; - clip.thread3.urb_entry_read_offset = 0; - clip.clip5.userclip_enable_flags = 0x7f; - clip.clip5.userclip_must_clip = 1; - clip.clip5.guard_band_enable = 0; - clip.clip5.viewport_z_clip_enable = 1; - clip.clip5.viewport_xy_clip_enable = 1; - clip.clip5.vertex_position_space = BRW_CLIP_NDCSPACE; - clip.clip5.api_mode = BRW_CLIP_API_OGL; - clip.clip6.clipper_viewport_state_ptr = 0; - clip.viewport_xmin = -1; - clip.viewport_xmax = 1; - clip.viewport_ymin = -1; - clip.viewport_ymax = 1; - - brw->clip.state_gs_offset = brw_cache_data( &brw->cache[BRW_CLIP_UNIT], &clip ); -} - - -const struct brw_tracked_state brw_clip_unit = { - .dirty = { - .brw = (BRW_NEW_CURBE_OFFSETS | - BRW_NEW_URB_FENCE), - .cache = CACHE_NEW_CLIP_PROG - }, - .update = upload_clip_unit -}; diff --git a/src/mesa/pipe/i965simple/brw_clip_tri.c b/src/mesa/pipe/i965simple/brw_clip_tri.c deleted file mode 100644 index c5da7b825e..0000000000 --- a/src/mesa/pipe/i965simple/brw_clip_tri.c +++ /dev/null @@ -1,566 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_clip.h" - -static struct brw_reg get_tmp( struct brw_clip_compile *c ) -{ - struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0); - - if (++c->last_tmp > c->prog_data.total_grf) - c->prog_data.total_grf = c->last_tmp; - - return tmp; -} - -static void release_tmps( struct brw_clip_compile *c ) -{ - c->last_tmp = c->first_tmp; -} - - -void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, - unsigned nr_verts ) -{ - unsigned i = 0,j; - - /* Register usage is static, precompute here: - */ - c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; - - if (c->key.nr_userclip) { - c->reg.fixed_planes = brw_vec4_grf(i, 0); - i += (6 + c->key.nr_userclip + 1) / 2; - - c->prog_data.curb_read_length = (6 + c->key.nr_userclip + 1) / 2; - } - else - c->prog_data.curb_read_length = 0; - - - /* Payload vertices plus space for more generated vertices: - */ - for (j = 0; j < nr_verts; j++) { - c->reg.vertex[j] = brw_vec4_grf(i, 0); - i += c->nr_regs; - } - - if (c->nr_attrs & 1) { - for (j = 0; j < 3; j++) { - unsigned delta = c->nr_attrs*16 + 32; - brw_MOV(&c->func, byte_offset(c->reg.vertex[j], delta), brw_imm_f(0)); - } - } - - c->reg.t = brw_vec1_grf(i, 0); - c->reg.loopcount = retype(brw_vec1_grf(i, 1), BRW_REGISTER_TYPE_UD); - c->reg.nr_verts = retype(brw_vec1_grf(i, 2), BRW_REGISTER_TYPE_UD); - c->reg.planemask = retype(brw_vec1_grf(i, 3), BRW_REGISTER_TYPE_UD); - c->reg.plane_equation = brw_vec4_grf(i, 4); - i++; - - c->reg.dpPrev = brw_vec1_grf(i, 0); /* fixme - dp4 will clobber r.1,2,3 */ - c->reg.dp = brw_vec1_grf(i, 4); - i++; - - c->reg.inlist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); - i++; - - c->reg.outlist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); - i++; - - c->reg.freelist = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, i, 0); - i++; - - if (!c->key.nr_userclip) { - c->reg.fixed_planes = brw_vec8_grf(i, 0); - i++; - } - - if (c->key.do_unfilled) { - c->reg.dir = brw_vec4_grf(i, 0); - c->reg.offset = brw_vec4_grf(i, 4); - i++; - c->reg.tmp0 = brw_vec4_grf(i, 0); - c->reg.tmp1 = brw_vec4_grf(i, 4); - i++; - } - - c->first_tmp = i; - c->last_tmp = i; - - c->prog_data.urb_read_length = c->nr_regs; /* ? */ - c->prog_data.total_grf = i; -} - - - -void brw_clip_tri_init_vertices( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */ - struct brw_instruction *is_rev; - - /* Initial list of indices for incoming vertexes: - */ - brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); - brw_CMP(p, - vec1(brw_null_reg()), - BRW_CONDITIONAL_EQ, - tmp0, - brw_imm_ud(_3DPRIM_TRISTRIP_REVERSE)); - - /* XXX: Is there an easier way to do this? Need to reverse every - * second tristrip element: Can ignore sometimes? - */ - is_rev = brw_IF(p, BRW_EXECUTE_1); - { - brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[1]) ); - brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[0]) ); - if (c->need_direction) - brw_MOV(p, c->reg.dir, brw_imm_f(-1)); - } - is_rev = brw_ELSE(p, is_rev); - { - brw_MOV(p, get_element(c->reg.inlist, 0), brw_address(c->reg.vertex[0]) ); - brw_MOV(p, get_element(c->reg.inlist, 1), brw_address(c->reg.vertex[1]) ); - if (c->need_direction) - brw_MOV(p, c->reg.dir, brw_imm_f(1)); - } - brw_ENDIF(p, is_rev); - - brw_MOV(p, get_element(c->reg.inlist, 2), brw_address(c->reg.vertex[2]) ); - brw_MOV(p, brw_vec8_grf(c->reg.outlist.nr, 0), brw_imm_f(0)); - brw_MOV(p, c->reg.nr_verts, brw_imm_ud(3)); -} - - - -void brw_clip_tri_flat_shade( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *is_poly; - struct brw_reg tmp0 = c->reg.loopcount; /* handy temporary */ - - brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); - brw_CMP(p, - vec1(brw_null_reg()), - BRW_CONDITIONAL_EQ, - tmp0, - brw_imm_ud(_3DPRIM_POLYGON)); - - is_poly = brw_IF(p, BRW_EXECUTE_1); - { - brw_clip_copy_colors(c, 1, 0); - brw_clip_copy_colors(c, 2, 0); - } - is_poly = brw_ELSE(p, is_poly); - { - brw_clip_copy_colors(c, 0, 2); - brw_clip_copy_colors(c, 1, 2); - } - brw_ENDIF(p, is_poly); -} - - - -/* Use mesa's clipping algorithms, translated to GEN4 assembly. - */ -void brw_clip_tri( struct brw_clip_compile *c ) -{ -#if 0 - struct brw_compile *p = &c->func; - struct brw_indirect vtx = brw_indirect(0, 0); - struct brw_indirect vtxPrev = brw_indirect(1, 0); - struct brw_indirect vtxOut = brw_indirect(2, 0); - struct brw_indirect plane_ptr = brw_indirect(3, 0); - struct brw_indirect inlist_ptr = brw_indirect(4, 0); - struct brw_indirect outlist_ptr = brw_indirect(5, 0); - struct brw_indirect freelist_ptr = brw_indirect(6, 0); - struct brw_instruction *plane_loop; - struct brw_instruction *plane_active; - struct brw_instruction *vertex_loop; - struct brw_instruction *next_test; - struct brw_instruction *prev_test; - - brw_MOV(p, get_addr_reg(vtxPrev), brw_address(c->reg.vertex[2]) ); - brw_MOV(p, get_addr_reg(plane_ptr), brw_clip_plane0_address(c)); - brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist)); - brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist)); - - brw_MOV(p, get_addr_reg(freelist_ptr), brw_address(c->reg.vertex[3]) ); - - plane_loop = brw_DO(p, BRW_EXECUTE_1); - { - /* if (planemask & 1) - */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_AND(p, vec1(brw_null_reg()), c->reg.planemask, brw_imm_ud(1)); - - plane_active = brw_IF(p, BRW_EXECUTE_1); - { - /* vtxOut = freelist_ptr++ - */ - brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(freelist_ptr) ); - brw_ADD(p, get_addr_reg(freelist_ptr), get_addr_reg(freelist_ptr), brw_imm_uw(c->nr_regs * REG_SIZE)); - - if (c->key.nr_userclip) - brw_MOV(p, c->reg.plane_equation, deref_4f(plane_ptr, 0)); - else - brw_MOV(p, c->reg.plane_equation, deref_4b(plane_ptr, 0)); - - brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); - brw_MOV(p, c->reg.nr_verts, brw_imm_ud(0)); - - vertex_loop = brw_DO(p, BRW_EXECUTE_1); - { - /* vtx = *input_ptr; - */ - brw_MOV(p, get_addr_reg(vtx), deref_1uw(inlist_ptr, 0)); - - /* IS_NEGATIVE(prev) */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_L); - brw_DP4(p, vec4(c->reg.dpPrev), deref_4f(vtxPrev, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); - prev_test = brw_IF(p, BRW_EXECUTE_1); - { - /* IS_POSITIVE(next) - */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_GE); - brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); - next_test = brw_IF(p, BRW_EXECUTE_1); - { - - /* Coming back in. - */ - brw_ADD(p, c->reg.t, c->reg.dpPrev, negate(c->reg.dp)); - brw_math_invert(p, c->reg.t, c->reg.t); - brw_MUL(p, c->reg.t, c->reg.t, c->reg.dpPrev); - - /* If (vtxOut == 0) vtxOut = vtxPrev - */ - brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) ); - brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtxPrev) ); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - - brw_clip_interp_vertex(c, vtxOut, vtxPrev, vtx, c->reg.t, FALSE); - - /* *outlist_ptr++ = vtxOut; - * nr_verts++; - * vtxOut = 0; - */ - brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut)); - brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); - brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); - brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) ); - } - brw_ENDIF(p, next_test); - - } - prev_test = brw_ELSE(p, prev_test); - { - /* *outlist_ptr++ = vtxPrev; - * nr_verts++; - */ - brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxPrev)); - brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); - brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); - - /* IS_NEGATIVE(next) - */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_L); - brw_DP4(p, vec4(c->reg.dp), deref_4f(vtx, c->offset[VERT_RESULT_HPOS]), c->reg.plane_equation); - next_test = brw_IF(p, BRW_EXECUTE_1); - { - /* Going out of bounds. Avoid division by zero as we - * know dp != dpPrev from DIFFERENT_SIGNS, above. - */ - brw_ADD(p, c->reg.t, c->reg.dp, negate(c->reg.dpPrev)); - brw_math_invert(p, c->reg.t, c->reg.t); - brw_MUL(p, c->reg.t, c->reg.t, c->reg.dp); - - /* If (vtxOut == 0) vtxOut = vtx - */ - brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_EQ, get_addr_reg(vtxOut), brw_imm_uw(0) ); - brw_MOV(p, get_addr_reg(vtxOut), get_addr_reg(vtx) ); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - - brw_clip_interp_vertex(c, vtxOut, vtx, vtxPrev, c->reg.t, TRUE); - - /* *outlist_ptr++ = vtxOut; - * nr_verts++; - * vtxOut = 0; - */ - brw_MOV(p, deref_1uw(outlist_ptr, 0), get_addr_reg(vtxOut)); - brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_uw(sizeof(short))); - brw_ADD(p, c->reg.nr_verts, c->reg.nr_verts, brw_imm_ud(1)); - brw_MOV(p, get_addr_reg(vtxOut), brw_imm_uw(0) ); - } - brw_ENDIF(p, next_test); - } - brw_ENDIF(p, prev_test); - - /* vtxPrev = vtx; - * inlist_ptr++; - */ - brw_MOV(p, get_addr_reg(vtxPrev), get_addr_reg(vtx)); - brw_ADD(p, get_addr_reg(inlist_ptr), get_addr_reg(inlist_ptr), brw_imm_uw(sizeof(short))); - - /* while (--loopcount != 0) - */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); - } - brw_WHILE(p, vertex_loop); - - /* vtxPrev = *(outlist_ptr-1) OR: outlist[nr_verts-1] - * inlist = outlist - * inlist_ptr = &inlist[0] - * outlist_ptr = &outlist[0] - */ - brw_ADD(p, get_addr_reg(outlist_ptr), get_addr_reg(outlist_ptr), brw_imm_w(-2)); - brw_MOV(p, get_addr_reg(vtxPrev), deref_1uw(outlist_ptr, 0)); - brw_MOV(p, brw_vec8_grf(c->reg.inlist.nr, 0), brw_vec8_grf(c->reg.outlist.nr, 0)); - brw_MOV(p, get_addr_reg(inlist_ptr), brw_address(c->reg.inlist)); - brw_MOV(p, get_addr_reg(outlist_ptr), brw_address(c->reg.outlist)); - } - brw_ENDIF(p, plane_active); - - /* plane_ptr++; - */ - brw_ADD(p, get_addr_reg(plane_ptr), get_addr_reg(plane_ptr), brw_clip_plane_stride(c)); - - /* nr_verts >= 3 - */ - brw_CMP(p, - vec1(brw_null_reg()), - BRW_CONDITIONAL_GE, - c->reg.nr_verts, - brw_imm_ud(3)); - - /* && (planemask>>=1) != 0 - */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_SHR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud(1)); - } - brw_WHILE(p, plane_loop); -#else - #warning "disabled" -#endif -} - - - -void brw_clip_tri_emit_polygon(struct brw_clip_compile *c) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *loop, *if_insn; - - /* for (loopcount = nr_verts-2; loopcount > 0; loopcount--) - */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_G); - brw_ADD(p, - c->reg.loopcount, - c->reg.nr_verts, - brw_imm_d(-2)); - - if_insn = brw_IF(p, BRW_EXECUTE_1); - { - struct brw_indirect v0 = brw_indirect(0, 0); - struct brw_indirect vptr = brw_indirect(1, 0); - - brw_MOV(p, get_addr_reg(vptr), brw_address(c->reg.inlist)); - brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); - - brw_clip_emit_vue(c, v0, 1, 0, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_START)); - - brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2)); - brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); - - loop = brw_DO(p, BRW_EXECUTE_1); - { - brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_TRIFAN << 2)); - - brw_ADD(p, get_addr_reg(vptr), get_addr_reg(vptr), brw_imm_uw(2)); - brw_MOV(p, get_addr_reg(v0), deref_1uw(vptr, 0)); - - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); - } - brw_WHILE(p, loop); - - brw_clip_emit_vue(c, v0, 0, 1, ((_3DPRIM_TRIFAN << 2) | R02_PRIM_END)); - } - brw_ENDIF(p, if_insn); -} - -static void do_clip_tri( struct brw_clip_compile *c ) -{ - brw_clip_init_planes(c); - - brw_clip_tri(c); -} - - -static void maybe_do_clip_tri( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *do_clip; - - brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0)); - do_clip = brw_IF(p, BRW_EXECUTE_1); - { - do_clip_tri(c); - } - brw_ENDIF(p, do_clip); -} - -static void brw_clip_test( struct brw_clip_compile *c ) -{ -#if 0 - struct brw_reg t = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); - struct brw_reg t1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); - struct brw_reg t2 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); - struct brw_reg t3 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); - - struct brw_reg v0 = get_tmp(c); - struct brw_reg v1 = get_tmp(c); - struct brw_reg v2 = get_tmp(c); - - struct brw_indirect vt0 = brw_indirect(0, 0); - struct brw_indirect vt1 = brw_indirect(1, 0); - struct brw_indirect vt2 = brw_indirect(2, 0); - - struct brw_compile *p = &c->func; - - brw_MOV(p, get_addr_reg(vt0), brw_address(c->reg.vertex[0])); - brw_MOV(p, get_addr_reg(vt1), brw_address(c->reg.vertex[1])); - brw_MOV(p, get_addr_reg(vt2), brw_address(c->reg.vertex[2])); - brw_MOV(p, v0, deref_4f(vt0, c->offset[VERT_RESULT_HPOS])); - brw_MOV(p, v1, deref_4f(vt1, c->offset[VERT_RESULT_HPOS])); - brw_MOV(p, v2, deref_4f(vt2, c->offset[VERT_RESULT_HPOS])); - - /* test nearz, xmin, ymin plane */ - brw_CMP(p, t1, BRW_CONDITIONAL_LE, negate(v0), get_element(v0, 3)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_CMP(p, t2, BRW_CONDITIONAL_LE, negate(v1), get_element(v1, 3)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_CMP(p, t3, BRW_CONDITIONAL_LE, negate(v2), get_element(v2, 3)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_XOR(p, t, t1, t2); - brw_XOR(p, t1, t2, t3); - brw_OR(p, t, t, t1); - - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, - get_element(t, 0), brw_imm_ud(0)); - brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<5))); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, - get_element(t, 1), brw_imm_ud(0)); - brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<3))); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, - get_element(t, 2), brw_imm_ud(0)); - brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<1))); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - - /* test farz, xmax, ymax plane */ - brw_CMP(p, t1, BRW_CONDITIONAL_L, v0, get_element(v0, 3)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_CMP(p, t2, BRW_CONDITIONAL_L, v1, get_element(v1, 3)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_CMP(p, t3, BRW_CONDITIONAL_L, v2, get_element(v2, 3)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - - brw_XOR(p, t, t1, t2); - brw_XOR(p, t1, t2, t3); - brw_OR(p, t, t, t1); - - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, - get_element(t, 0), brw_imm_ud(0)); - brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<4))); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, - get_element(t, 1), brw_imm_ud(0)); - brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<2))); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_NZ, - get_element(t, 2), brw_imm_ud(0)); - brw_OR(p, c->reg.planemask, c->reg.planemask, brw_imm_ud((1<<0))); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - - release_tmps(c); -#else - #warning "disabled" -#endif -} - - -void brw_emit_tri_clip( struct brw_clip_compile *c ) -{ - struct brw_instruction *neg_rhw; - struct brw_compile *p = &c->func; - brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6); - brw_clip_tri_init_vertices(c); - brw_clip_init_clipmask(c); - - /* if -ve rhw workaround bit is set, - do cliptest */ - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_AND(p, brw_null_reg(), get_element_ud(c->reg.R0, 2), - brw_imm_ud(1<<20)); - neg_rhw = brw_IF(p, BRW_EXECUTE_1); - { - brw_clip_test(c); - } - brw_ENDIF(p, neg_rhw); - - /* Can't push into do_clip_tri because with polygon (or quad) - * flatshading, need to apply the flatshade here because we don't - * respect the PV when converting to trifan for emit: - */ - if (c->key.do_flat_shading) - brw_clip_tri_flat_shade(c); - - if (c->key.clip_mode == BRW_CLIPMODE_NORMAL) - do_clip_tri(c); - else - maybe_do_clip_tri(c); - - brw_clip_tri_emit_polygon(c); - - /* Send an empty message to kill the thread: - */ - brw_clip_kill_thread(c); -} diff --git a/src/mesa/pipe/i965simple/brw_clip_unfilled.c b/src/mesa/pipe/i965simple/brw_clip_unfilled.c deleted file mode 100644 index b774a76dd6..0000000000 --- a/src/mesa/pipe/i965simple/brw_clip_unfilled.c +++ /dev/null @@ -1,477 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_clip.h" - - - -/* This is performed against the original triangles, so no indirection - * required: -BZZZT! - */ -static void compute_tri_direction( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_reg e = c->reg.tmp0; - struct brw_reg f = c->reg.tmp1; - struct brw_reg v0 = byte_offset(c->reg.vertex[0], c->offset[VERT_RESULT_HPOS]); - struct brw_reg v1 = byte_offset(c->reg.vertex[1], c->offset[VERT_RESULT_HPOS]); - struct brw_reg v2 = byte_offset(c->reg.vertex[2], c->offset[VERT_RESULT_HPOS]); - - - /* Calculate the vectors of two edges of the triangle: - */ - brw_ADD(p, e, v0, negate(v2)); - brw_ADD(p, f, v1, negate(v2)); - - /* Take their crossproduct: - */ - brw_set_access_mode(p, BRW_ALIGN_16); - brw_MUL(p, vec4(brw_null_reg()), brw_swizzle(e, 1,2,0,3), brw_swizzle(f,2,0,1,3)); - brw_MAC(p, vec4(e), negate(brw_swizzle(e, 2,0,1,3)), brw_swizzle(f,1,2,0,3)); - brw_set_access_mode(p, BRW_ALIGN_1); - - brw_MUL(p, c->reg.dir, c->reg.dir, vec4(e)); -} - - -static void cull_direction( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *ccw; - unsigned conditional; - - assert (!(c->key.fill_ccw == CLIP_CULL && - c->key.fill_cw == CLIP_CULL)); - - if (c->key.fill_ccw == CLIP_CULL) - conditional = BRW_CONDITIONAL_GE; - else - conditional = BRW_CONDITIONAL_L; - - brw_CMP(p, - vec1(brw_null_reg()), - conditional, - get_element(c->reg.dir, 2), - brw_imm_f(0)); - - ccw = brw_IF(p, BRW_EXECUTE_1); - { - brw_clip_kill_thread(c); - } - brw_ENDIF(p, ccw); -} - - - -static void copy_bfc( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *ccw; - unsigned conditional; - - /* Do we have any colors to copy? - */ - if (!(c->offset[VERT_RESULT_COL0] && c->offset[VERT_RESULT_BFC0]) && - !(c->offset[VERT_RESULT_COL1] && c->offset[VERT_RESULT_BFC1])) - return; - - /* In some wierd degnerate cases we can end up testing the - * direction twice, once for culling and once for bfc copying. Oh - * well, that's what you get for setting wierd GL state. - */ - if (c->key.copy_bfc_ccw) - conditional = BRW_CONDITIONAL_GE; - else - conditional = BRW_CONDITIONAL_L; - - brw_CMP(p, - vec1(brw_null_reg()), - conditional, - get_element(c->reg.dir, 2), - brw_imm_f(0)); - - ccw = brw_IF(p, BRW_EXECUTE_1); - { - unsigned i; - - for (i = 0; i < 3; i++) { - if (c->offset[VERT_RESULT_COL0] && c->offset[VERT_RESULT_BFC0]) - brw_MOV(p, - byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_COL0]), - byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_BFC0])); - - if (c->offset[VERT_RESULT_COL1] && c->offset[VERT_RESULT_BFC1]) - brw_MOV(p, - byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_COL1]), - byte_offset(c->reg.vertex[i], c->offset[VERT_RESULT_BFC1])); - } - } - brw_ENDIF(p, ccw); -} - - - - -/* - float iz = 1.0 / dir.z; - float ac = dir.x * iz; - float bc = dir.y * iz; - offset = ctx->Polygon.OffsetUnits * DEPTH_SCALE; - offset += MAX2( abs(ac), abs(bc) ) * ctx->Polygon.OffsetFactor; - offset *= MRD; -*/ -static void compute_offset( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_reg off = c->reg.offset; - struct brw_reg dir = c->reg.dir; - - brw_math_invert(p, get_element(off, 2), get_element(dir, 2)); - brw_MUL(p, vec2(off), dir, get_element(off, 2)); - - brw_CMP(p, - vec1(brw_null_reg()), - BRW_CONDITIONAL_GE, - brw_abs(get_element(off, 0)), - brw_abs(get_element(off, 1))); - - brw_SEL(p, vec1(off), brw_abs(get_element(off, 0)), brw_abs(get_element(off, 1))); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - - brw_MUL(p, vec1(off), off, brw_imm_f(c->key.offset_factor)); - brw_ADD(p, vec1(off), off, brw_imm_f(c->key.offset_units)); -} - - -static void merge_edgeflags( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *is_poly; - struct brw_reg tmp0 = get_element_ud(c->reg.tmp0, 0); - - brw_AND(p, tmp0, get_element_ud(c->reg.R0, 2), brw_imm_ud(PRIM_MASK)); - brw_CMP(p, - vec1(brw_null_reg()), - BRW_CONDITIONAL_EQ, - tmp0, - brw_imm_ud(_3DPRIM_POLYGON)); - - /* Get away with using reg.vertex because we know that this is not - * a _3DPRIM_TRISTRIP_REVERSE: - */ - is_poly = brw_IF(p, BRW_EXECUTE_1); - { - brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ); - brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<8)); - brw_MOV(p, byte_offset(c->reg.vertex[0], c->offset[VERT_RESULT_EDGE]), brw_imm_f(0)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - - brw_set_conditionalmod(p, BRW_CONDITIONAL_EQ); - brw_AND(p, vec1(brw_null_reg()), get_element_ud(c->reg.R0, 2), brw_imm_ud(1<<9)); - brw_MOV(p, byte_offset(c->reg.vertex[2], c->offset[VERT_RESULT_EDGE]), brw_imm_f(0)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - } - brw_ENDIF(p, is_poly); -} - - - -static void apply_one_offset( struct brw_clip_compile *c, - struct brw_indirect vert ) -{ - struct brw_compile *p = &c->func; - struct brw_reg pos = deref_4f(vert, c->offset[VERT_RESULT_HPOS]); - struct brw_reg z = get_element(pos, 2); - - brw_ADD(p, z, z, vec1(c->reg.offset)); -} - - - -/*********************************************************************** - * Output clipped polygon as an unfilled primitive: - */ -static void emit_lines(struct brw_clip_compile *c, - boolean do_offset) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *loop; - struct brw_instruction *draw_edge; - struct brw_indirect v0 = brw_indirect(0, 0); - struct brw_indirect v1 = brw_indirect(1, 0); - struct brw_indirect v0ptr = brw_indirect(2, 0); - struct brw_indirect v1ptr = brw_indirect(3, 0); - - /* Need a seperate loop for offset: - */ - if (do_offset) { - brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); - brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); - - loop = brw_DO(p, BRW_EXECUTE_1); - { - brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); - brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); - - apply_one_offset(c, v0); - - brw_set_conditionalmod(p, BRW_CONDITIONAL_G); - brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); - } - brw_WHILE(p, loop); - } - - /* v1ptr = &inlist[nr_verts] - * *v1ptr = v0 - */ - brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); - brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); - brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v0ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW)); - brw_ADD(p, get_addr_reg(v1ptr), get_addr_reg(v1ptr), retype(c->reg.nr_verts, BRW_REGISTER_TYPE_UW)); - brw_MOV(p, deref_1uw(v1ptr, 0), deref_1uw(v0ptr, 0)); - - loop = brw_DO(p, BRW_EXECUTE_1); - { - brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); - brw_MOV(p, get_addr_reg(v1), deref_1uw(v0ptr, 2)); - brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); - - /* draw edge if edgeflag != 0 */ - brw_CMP(p, - vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, - deref_1f(v0, c->offset[VERT_RESULT_EDGE]), - brw_imm_f(0)); - draw_edge = brw_IF(p, BRW_EXECUTE_1); - { - brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_START); - brw_clip_emit_vue(c, v1, 1, 0, (_3DPRIM_LINESTRIP << 2) | R02_PRIM_END); - } - brw_ENDIF(p, draw_edge); - - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); - } - brw_WHILE(p, loop); -} - - - -static void emit_points(struct brw_clip_compile *c, - boolean do_offset ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *loop; - struct brw_instruction *draw_point; - - struct brw_indirect v0 = brw_indirect(0, 0); - struct brw_indirect v0ptr = brw_indirect(2, 0); - - brw_MOV(p, c->reg.loopcount, c->reg.nr_verts); - brw_MOV(p, get_addr_reg(v0ptr), brw_address(c->reg.inlist)); - - loop = brw_DO(p, BRW_EXECUTE_1); - { - brw_MOV(p, get_addr_reg(v0), deref_1uw(v0ptr, 0)); - brw_ADD(p, get_addr_reg(v0ptr), get_addr_reg(v0ptr), brw_imm_uw(2)); - - /* draw if edgeflag != 0 - */ - brw_CMP(p, - vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, - deref_1f(v0, c->offset[VERT_RESULT_EDGE]), - brw_imm_f(0)); - draw_point = brw_IF(p, BRW_EXECUTE_1); - { - if (do_offset) - apply_one_offset(c, v0); - - brw_clip_emit_vue(c, v0, 1, 0, (_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END); - } - brw_ENDIF(p, draw_point); - - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - brw_ADD(p, c->reg.loopcount, c->reg.loopcount, brw_imm_d(-1)); - } - brw_WHILE(p, loop); -} - - - - - - - -static void emit_primitives( struct brw_clip_compile *c, - unsigned mode, - boolean do_offset ) -{ - switch (mode) { - case CLIP_FILL: - brw_clip_tri_emit_polygon(c); - break; - - case CLIP_LINE: - emit_lines(c, do_offset); - break; - - case CLIP_POINT: - emit_points(c, do_offset); - break; - - case CLIP_CULL: - assert(0); - break; - } -} - - - -static void emit_unfilled_primitives( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *ccw; - - /* Direction culling has already been done. - */ - if (c->key.fill_ccw != c->key.fill_cw && - c->key.fill_ccw != CLIP_CULL && - c->key.fill_cw != CLIP_CULL) - { - brw_CMP(p, - vec1(brw_null_reg()), - BRW_CONDITIONAL_GE, - get_element(c->reg.dir, 2), - brw_imm_f(0)); - - ccw = brw_IF(p, BRW_EXECUTE_1); - { - emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw); - } - ccw = brw_ELSE(p, ccw); - { - emit_primitives(c, c->key.fill_cw, c->key.offset_cw); - } - brw_ENDIF(p, ccw); - } - else if (c->key.fill_cw != CLIP_CULL) { - emit_primitives(c, c->key.fill_cw, c->key.offset_cw); - } - else if (c->key.fill_ccw != CLIP_CULL) { - emit_primitives(c, c->key.fill_ccw, c->key.offset_ccw); - } -} - - - - -static void check_nr_verts( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *if_insn; - - brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_L, c->reg.nr_verts, brw_imm_d(3)); - if_insn = brw_IF(p, BRW_EXECUTE_1); - { - brw_clip_kill_thread(c); - } - brw_ENDIF(p, if_insn); -} - - -void brw_emit_unfilled_clip( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *do_clip; - - - c->need_direction = ((c->key.offset_ccw || c->key.offset_cw) || - (c->key.fill_ccw != c->key.fill_cw) || - c->key.fill_ccw == CLIP_CULL || - c->key.fill_cw == CLIP_CULL || - c->key.copy_bfc_cw || - c->key.copy_bfc_ccw); - - brw_clip_tri_alloc_regs(c, 3 + c->key.nr_userclip + 6); - brw_clip_tri_init_vertices(c); - - assert(c->offset[VERT_RESULT_EDGE]); - - if (c->key.fill_ccw == CLIP_CULL && - c->key.fill_cw == CLIP_CULL) { - brw_clip_kill_thread(c); - return; - } - - merge_edgeflags(c); - - /* Need to use the inlist indirection here: - */ - if (c->need_direction) - compute_tri_direction(c); - - if (c->key.fill_ccw == CLIP_CULL || - c->key.fill_cw == CLIP_CULL) - cull_direction(c); - - if (c->key.offset_ccw || - c->key.offset_cw) - compute_offset(c); - - if (c->key.copy_bfc_ccw || - c->key.copy_bfc_cw) - copy_bfc(c); - - /* Need to do this whether we clip or not: - */ - if (c->key.do_flat_shading) - brw_clip_tri_flat_shade(c); - - brw_clip_init_clipmask(c); - brw_CMP(p, vec1(brw_null_reg()), BRW_CONDITIONAL_NZ, c->reg.planemask, brw_imm_ud(0)); - do_clip = brw_IF(p, BRW_EXECUTE_1); - { - brw_clip_init_planes(c); - brw_clip_tri(c); - check_nr_verts(c); - } - brw_ENDIF(p, do_clip); - - emit_unfilled_primitives(c); - brw_clip_kill_thread(c); -} - - - diff --git a/src/mesa/pipe/i965simple/brw_clip_util.c b/src/mesa/pipe/i965simple/brw_clip_util.c deleted file mode 100644 index 6d58ceafff..0000000000 --- a/src/mesa/pipe/i965simple/brw_clip_util.c +++ /dev/null @@ -1,351 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_clip.h" - - - - - -static struct brw_reg get_tmp( struct brw_clip_compile *c ) -{ - struct brw_reg tmp = brw_vec4_grf(c->last_tmp, 0); - - if (++c->last_tmp > c->prog_data.total_grf) - c->prog_data.total_grf = c->last_tmp; - - return tmp; -} - -static void release_tmp( struct brw_clip_compile *c, struct brw_reg tmp ) -{ - if (tmp.nr == c->last_tmp-1) - c->last_tmp--; -} - - -static struct brw_reg make_plane_ud(unsigned x, unsigned y, unsigned z, unsigned w) -{ - return brw_imm_ud((w<<24) | (z<<16) | (y<<8) | x); -} - - -void brw_clip_init_planes( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - - if (!c->key.nr_userclip) { - brw_MOV(p, get_element_ud(c->reg.fixed_planes, 0), make_plane_ud( 0, 0, 0xff, 1)); - brw_MOV(p, get_element_ud(c->reg.fixed_planes, 1), make_plane_ud( 0, 0, 1, 1)); - brw_MOV(p, get_element_ud(c->reg.fixed_planes, 2), make_plane_ud( 0, 0xff, 0, 1)); - brw_MOV(p, get_element_ud(c->reg.fixed_planes, 3), make_plane_ud( 0, 1, 0, 1)); - brw_MOV(p, get_element_ud(c->reg.fixed_planes, 4), make_plane_ud(0xff, 0, 0, 1)); - brw_MOV(p, get_element_ud(c->reg.fixed_planes, 5), make_plane_ud( 1, 0, 0, 1)); - } -} - - - -#define W 3 - -/* Project 'pos' to screen space (or back again), overwrite with results: - */ -static void brw_clip_project_position(struct brw_clip_compile *c, struct brw_reg pos ) -{ - struct brw_compile *p = &c->func; - - /* calc rhw - */ - brw_math_invert(p, get_element(pos, W), get_element(pos, W)); - - /* value.xyz *= value.rhw - */ - brw_set_access_mode(p, BRW_ALIGN_16); - brw_MUL(p, brw_writemask(pos, TGSI_WRITEMASK_XYZ), pos, brw_swizzle1(pos, W)); - brw_set_access_mode(p, BRW_ALIGN_1); -} - - -static void brw_clip_project_vertex( struct brw_clip_compile *c, - struct brw_indirect vert_addr ) -{ -#if 0 - struct brw_compile *p = &c->func; - struct brw_reg tmp = get_tmp(c); - - /* Fixup position. Extract from the original vertex and re-project - * to screen space: - */ - brw_MOV(p, tmp, deref_4f(vert_addr, c->offset[VERT_RESULT_HPOS])); - brw_clip_project_position(c, tmp); - brw_MOV(p, deref_4f(vert_addr, c->header_position_offset), tmp); - - release_tmp(c, tmp); -#else - #warning "disabled" -#endif -} - - - - -/* Interpolate between two vertices and put the result into a0.0. - * Increment a0.0 accordingly. - */ -void brw_clip_interp_vertex( struct brw_clip_compile *c, - struct brw_indirect dest_ptr, - struct brw_indirect v0_ptr, /* from */ - struct brw_indirect v1_ptr, /* to */ - struct brw_reg t0, - boolean force_edgeflag) -{ -#if 0 - struct brw_compile *p = &c->func; - struct brw_reg tmp = get_tmp(c); - unsigned i; - - /* Just copy the vertex header: - */ - brw_copy_indirect_to_indirect(p, dest_ptr, v0_ptr, 1); - - /* Iterate over each attribute (could be done in pairs?) - */ - for (i = 0; i < c->nr_attrs; i++) { - unsigned delta = i*16 + 32; - - if (delta == c->offset[VERT_RESULT_EDGE]) { - if (force_edgeflag) - brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(1)); - else - brw_MOV(p, deref_4f(dest_ptr, delta), deref_4f(v0_ptr, delta)); - } - else { - /* Interpolate: - * - * New = attr0 + t*attr1 - t*attr0 - */ - brw_MUL(p, - vec4(brw_null_reg()), - deref_4f(v1_ptr, delta), - t0); - - brw_MAC(p, - tmp, - negate(deref_4f(v0_ptr, delta)), - t0); - - brw_ADD(p, - deref_4f(dest_ptr, delta), - deref_4f(v0_ptr, delta), - tmp); - } - } - - if (i & 1) { - unsigned delta = i*16 + 32; - brw_MOV(p, deref_4f(dest_ptr, delta), brw_imm_f(0)); - } - - release_tmp(c, tmp); - - /* Recreate the projected (NDC) coordinate in the new vertex - * header: - */ - brw_clip_project_vertex(c, dest_ptr ); -#else - #warning "disabled" -#endif -} - - - - -#define MAX_MRF 16 - -void brw_clip_emit_vue(struct brw_clip_compile *c, - struct brw_indirect vert, - boolean allocate, - boolean eot, - unsigned header) -{ - struct brw_compile *p = &c->func; - unsigned start = c->last_mrf; - - assert(!(allocate && eot)); - - /* Cycle through mrf regs - probably futile as we have to wait for - * the allocation response anyway. Also, the order this function - * is invoked doesn't correspond to the order the instructions will - * be executed, so it won't have any effect in many cases. - */ -#if 0 - if (start + c->nr_regs + 1 >= MAX_MRF) - start = 0; - - c->last_mrf = start + c->nr_regs + 1; -#endif - - /* Copy the vertex from vertn into m1..mN+1: - */ - brw_copy_from_indirect(p, brw_message_reg(start+1), vert, c->nr_regs); - - /* Overwrite PrimType and PrimStart in the message header, for - * each vertex in turn: - */ - brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header)); - - - /* Send each vertex as a seperate write to the urb. This - * is different to the concept in brw_sf_emit.c, where - * subsequent writes are used to build up a single urb - * entry. Each of these writes instantiates a seperate - * urb entry - (I think... what about 'allocate'?) - */ - brw_urb_WRITE(p, - allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), - start, - c->reg.R0, - allocate, - 1, /* used */ - c->nr_regs + 1, /* msg length */ - allocate ? 1 : 0, /* response_length */ - eot, /* eot */ - 1, /* writes_complete */ - 0, /* urb offset */ - BRW_URB_SWIZZLE_NONE); -} - - - -void brw_clip_kill_thread(struct brw_clip_compile *c) -{ - struct brw_compile *p = &c->func; - - /* Send an empty message to kill the thread and release any - * allocated urb entry: - */ - brw_urb_WRITE(p, - retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), - 0, - c->reg.R0, - 0, /* allocate */ - 0, /* used */ - 0, /* msg len */ - 0, /* response len */ - 1, /* eot */ - 1, /* writes complete */ - 0, - BRW_URB_SWIZZLE_NONE); -} - - - - -struct brw_reg brw_clip_plane0_address( struct brw_clip_compile *c ) -{ - return brw_address(c->reg.fixed_planes); -} - - -struct brw_reg brw_clip_plane_stride( struct brw_clip_compile *c ) -{ - if (c->key.nr_userclip) { - return brw_imm_uw(16); - } - else { - return brw_imm_uw(4); - } -} - - -/* If flatshading, distribute color from provoking vertex prior to - * clipping. - */ -void brw_clip_copy_colors( struct brw_clip_compile *c, - unsigned to, unsigned from ) -{ -#if 0 - struct brw_compile *p = &c->func; - - if (c->offset[VERT_RESULT_COL0]) - brw_MOV(p, - byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL0]), - byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL0])); - - if (c->offset[VERT_RESULT_COL1]) - brw_MOV(p, - byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_COL1]), - byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_COL1])); - - if (c->offset[VERT_RESULT_BFC0]) - brw_MOV(p, - byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC0]), - byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC0])); - - if (c->offset[VERT_RESULT_BFC1]) - brw_MOV(p, - byte_offset(c->reg.vertex[to], c->offset[VERT_RESULT_BFC1]), - byte_offset(c->reg.vertex[from], c->offset[VERT_RESULT_BFC1])); -#else - #warning "disabled" -#endif -} - - - -void brw_clip_init_clipmask( struct brw_clip_compile *c ) -{ - struct brw_compile *p = &c->func; - struct brw_reg incoming = get_element_ud(c->reg.R0, 2); - - /* Shift so that lowest outcode bit is rightmost: - */ - brw_SHR(p, c->reg.planemask, incoming, brw_imm_ud(26)); - - if (c->key.nr_userclip) { - struct brw_reg tmp = retype(vec1(get_tmp(c)), BRW_REGISTER_TYPE_UD); - - /* Rearrange userclip outcodes so that they come directly after - * the fixed plane bits. - */ - brw_AND(p, tmp, incoming, brw_imm_ud(0x3f<<14)); - brw_SHR(p, tmp, tmp, brw_imm_ud(8)); - brw_OR(p, c->reg.planemask, c->reg.planemask, tmp); - - release_tmp(c, tmp); - } -} - diff --git a/src/mesa/pipe/i965simple/brw_context.c b/src/mesa/pipe/i965simple/brw_context.c deleted file mode 100644 index 5e58701e91..0000000000 --- a/src/mesa/pipe/i965simple/brw_context.c +++ /dev/null @@ -1,245 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_draw.h" -#include "brw_vs.h" -#include "brw_tex_layout.h" -#include "brw_winsys.h" - -#include "pipe/p_winsys.h" -#include "pipe/p_context.h" -#include "pipe/p_util.h" - -/*************************************** - * Mesa's Driver Functions - ***************************************/ - -#ifndef BRW_DEBUG -int BRW_DEBUG = (0); -#endif - -static void brw_destroy(struct pipe_context *pipe) -{ - struct brw_context *brw = brw_context(pipe); - - FREE(brw); -} - -static void brw_clear(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue) -{ - int x, y, w, h; - /* FIXME: corny... */ - - x = 0; - y = 0; - w = ps->width; - h = ps->height; - - pipe->surface_fill(pipe, ps, x, y, w, h, clearValue); -} - - -static int -brw_get_param(struct pipe_context *pipe, int param) -{ - switch (param) { - case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; - case PIPE_CAP_NPOT_TEXTURES: - return 1; - case PIPE_CAP_TWO_SIDED_STENCIL: - return 1; - case PIPE_CAP_GLSL: - return 0; - case PIPE_CAP_S3TC: - return 0; - case PIPE_CAP_ANISOTROPIC_FILTER: - return 0; - case PIPE_CAP_POINT_SPRITE: - return 0; - case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; - case PIPE_CAP_OCCLUSION_QUERY: - return 0; - case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 1; - case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 11; /* max 1024x1024 */ - case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: - return 8; /* max 128x128x128 */ - case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 11; /* max 1024x1024 */ - default: - return 0; - } -} - - -static float -brw_get_paramf(struct pipe_context *pipe, int param) -{ - switch (param) { - case PIPE_CAP_MAX_LINE_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_LINE_WIDTH_AA: - return 7.5; - - case PIPE_CAP_MAX_POINT_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_POINT_WIDTH_AA: - return 255.0; - - case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 4.0; - - case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: - return 16.0; - - default: - return 0; - } -} - -static boolean -brw_is_format_supported( struct pipe_context *pipe, - enum pipe_format format, uint type ) -{ -#if 0 - /* XXX: This is broken -- rewrite if still needed. */ - static const unsigned tex_supported[] = { - PIPE_FORMAT_U_R8_G8_B8_A8, - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_U_L8, - PIPE_FORMAT_U_A8, - PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_L8_A8, - PIPE_FORMAT_YCBCR, - PIPE_FORMAT_YCBCR_REV, - PIPE_FORMAT_S8_Z24, - }; - - - /* Actually a lot more than this - add later: - */ - static const unsigned render_supported[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - }; - - /* - */ - static const unsigned z_stencil_supported[] = { - PIPE_FORMAT_U_Z16, - PIPE_FORMAT_U_Z32, - PIPE_FORMAT_S8_Z24, - }; - - switch (type) { - case PIPE_RENDER_FORMAT: - *numFormats = Elements(render_supported); - return render_supported; - - case PIPE_TEX_FORMAT: - *numFormats = Elements(tex_supported); - return render_supported; - - case PIPE_Z_STENCIL_FORMAT: - *numFormats = Elements(render_supported); - return render_supported; - - default: - *numFormats = 0; - return NULL; - } -#else - switch (format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - return TRUE; - default: - return FALSE; - }; - return FALSE; -#endif -} - - - - -struct pipe_context *brw_create(struct pipe_winsys *pipe_winsys, - struct brw_winsys *brw_winsys, - unsigned pci_id) -{ - struct brw_context *brw; - - pipe_winsys->printf(pipe_winsys, - "%s: creating brw_context with pci id 0x%x\n", - __FUNCTION__, pci_id); - - brw = CALLOC_STRUCT(brw_context); - if (brw == NULL) - return NULL; - - brw->winsys = brw_winsys; - brw->pipe.winsys = pipe_winsys; - - brw->pipe.destroy = brw_destroy; - brw->pipe.is_format_supported = brw_is_format_supported; - brw->pipe.get_param = brw_get_param; - brw->pipe.get_paramf = brw_get_paramf; - brw->pipe.clear = brw_clear; - brw->pipe.texture_create = brw_texture_create; - brw->pipe.texture_release = brw_texture_release; - - brw_init_surface_functions(brw); - brw_init_state_functions(brw); - brw_init_flush_functions(brw); - brw_init_string_functions(brw); - brw_init_draw_functions( brw ); - - - brw_init_state( brw ); - - brw->pci_id = pci_id; - brw->dirty = ~0; - brw->hardware_dirty = ~0; - - memset(&brw->wm.bind, ~0, sizeof(brw->wm.bind)); - - return &brw->pipe; -} - diff --git a/src/mesa/pipe/i965simple/brw_context.h b/src/mesa/pipe/i965simple/brw_context.h deleted file mode 100644 index 65664d853d..0000000000 --- a/src/mesa/pipe/i965simple/brw_context.h +++ /dev/null @@ -1,690 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRWCONTEXT_INC -#define BRWCONTEXT_INC - - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" - -#include "brw_structs.h" -#include "brw_winsys.h" - - -/* Glossary: - * - * URB - uniform resource buffer. A mid-sized buffer which is - * partitioned between the fixed function units and used for passing - * values (vertices, primitives, constants) between them. - * - * CURBE - constant URB entry. An urb region (entry) used to hold - * constant values which the fixed function units can be instructed to - * preload into the GRF when spawining a thread. - * - * VUE - vertex URB entry. An urb entry holding a vertex and usually - * a vertex header. The header contains control information and - * things like primitive type, Begin/end flags and clip codes. - * - * PUE - primitive URB entry. An urb entry produced by the setup (SF) - * unit holding rasterization and interpolation parameters. - * - * GRF - general register file. One of several register files - * addressable by programmed threads. The inputs (r0, payload, curbe, - * urb) of the thread are preloaded to this area before the thread is - * spawned. The registers are individually 8 dwords wide and suitable - * for general usage. Registers holding thread input values are not - * special and may be overwritten. - * - * MRF - message register file. Threads communicate (and terminate) - * by sending messages. Message parameters are placed in contigous - * MRF registers. All program output is via these messages. URB - * entries are populated by sending a message to the shared URB - * function containing the new data, together with a control word, - * often an unmodified copy of R0. - * - * R0 - GRF register 0. Typically holds control information used when - * sending messages to other threads. - * - * EU or GEN4 EU: The name of the programmable subsystem of the - * i965 hardware. Threads are executed by the EU, the registers - * described above are part of the EU architecture. - * - * Fixed function units: - * - * CS - Command streamer. Notional first unit, little software - * interaction. Holds the URB entries used for constant data, ie the - * CURBEs. - * - * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of - * this unit is responsible for pulling vertices out of vertex buffers - * in vram and injecting them into the processing pipe as VUEs. If - * enabled, it first passes them to a VS thread which is a good place - * for the driver to implement any active vertex shader. - * - * GS - Geometry Shader. This corresponds to a new DX10 concept. If - * enabled, incoming strips etc are passed to GS threads in individual - * line/triangle/point units. The GS thread may perform arbitary - * computation and emit whatever primtives with whatever vertices it - * chooses. This makes GS an excellent place to implement GL's - * unfilled polygon modes, though of course it is capable of much - * more. Additionally, GS is used to translate away primitives not - * handled by latter units, including Quads and Lineloops. - * - * CS - Clipper. Mesa's clipping algorithms are imported to run on - * this unit. The fixed function part performs cliptesting against - * the 6 fixed clipplanes and makes descisions on whether or not the - * incoming primitive needs to be passed to a thread for clipping. - * User clip planes are handled via cooperation with the VS thread. - * - * SF - Strips Fans or Setup: Triangles are prepared for - * rasterization. Interpolation coefficients are calculated. - * Flatshading and two-side lighting usually performed here. - * - * WM - Windower. Interpolation of vertex attributes performed here. - * Fragment shader implemented here. SIMD aspects of EU taken full - * advantage of, as pixels are processed in blocks of 16. - * - * CC - Color Calculator. No EU threads associated with this unit. - * Handles blending and (presumably) depth and stencil testing. - */ - -#define BRW_MAX_CURBE (32*16) - -struct brw_context; -struct brw_winsys; - - -/* Raised when we receive new state across the pipe interface: - */ -#define BRW_NEW_VIEWPORT 0x1 -#define BRW_NEW_RASTERIZER 0x2 -#define BRW_NEW_FS 0x4 -#define BRW_NEW_BLEND 0x8 -#define BRW_NEW_CLIP 0x10 -#define BRW_NEW_SCISSOR 0x20 -#define BRW_NEW_STIPPLE 0x40 -#define BRW_NEW_FRAMEBUFFER 0x80 -#define BRW_NEW_ALPHA_TEST 0x100 -#define BRW_NEW_DEPTH_STENCIL 0x200 -#define BRW_NEW_SAMPLER 0x400 -#define BRW_NEW_TEXTURE 0x800 -#define BRW_NEW_CONSTANTS 0x1000 -#define BRW_NEW_VBO 0x2000 -#define BRW_NEW_VS 0x4000 - -/* Raised for other internal events: - */ -#define BRW_NEW_URB_FENCE 0x10000 -#define BRW_NEW_PSP 0x20000 -#define BRW_NEW_CURBE_OFFSETS 0x40000 -#define BRW_NEW_REDUCED_PRIMITIVE 0x80000 -#define BRW_NEW_PRIMITIVE 0x100000 -#define BRW_NEW_SCENE 0x200000 -#define BRW_NEW_SF_LINKAGE 0x400000 - -extern int BRW_DEBUG; - -#define DEBUG_TEXTURE 0x1 -#define DEBUG_STATE 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_PRIMS 0x8 -#define DEBUG_VERTS 0x10 -#define DEBUG_FALLBACKS 0x20 -#define DEBUG_VERBOSE 0x40 -#define DEBUG_DRI 0x80 -#define DEBUG_DMA 0x100 -#define DEBUG_SANITY 0x200 -#define DEBUG_SYNC 0x400 -#define DEBUG_SLEEP 0x800 -#define DEBUG_PIXEL 0x1000 -#define DEBUG_STATS 0x2000 -#define DEBUG_TILE 0x4000 -#define DEBUG_SINGLE_THREAD 0x8000 -#define DEBUG_WM 0x10000 -#define DEBUG_URB 0x20000 -#define DEBUG_VS 0x40000 -#define DEBUG_BATCH 0x80000 -#define DEBUG_BUFMGR 0x100000 -#define DEBUG_BLIT 0x200000 -#define DEBUG_REGION 0x400000 -#define DEBUG_MIPTREE 0x800000 - -#define DBG(...) do { \ - if (BRW_DEBUG & FILE_DEBUG_FLAG) \ - brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ -} while(0) - -#define PRINT(...) do { \ - brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ -} while(0) - -struct brw_state_flags { - unsigned cache; - unsigned brw; -}; - - -struct brw_shader_info { - int nr_regs[8]; /* TGSI_FILE_* */ -}; - - - -struct brw_vertex_program { - struct pipe_shader_state program; - struct brw_shader_info info; - int id; -}; - - - -struct brw_fragment_program { - struct pipe_shader_state program; - struct brw_shader_info info; - - boolean UsesDepth; - boolean UsesKill; - boolean ComputesDepth; - int id; -}; - - - - -struct pipe_setup_linkage { - struct { - unsigned vp_output:5; - unsigned interp_mode:4; - unsigned bf_vp_output:5; - } fp_input[PIPE_MAX_SHADER_INPUTS]; - - unsigned fp_input_count:5; - unsigned max_vp_output:5; -}; - - - -struct brw_texture { - struct pipe_texture base; - - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ - unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; -}; - -/* Data about a particular attempt to compile a program. Note that - * there can be many of these, each in a different GL state - * corresponding to a different brw_wm_prog_key struct, with different - * compiled programs: - */ -/* Data about a particular attempt to compile a program. Note that - * there can be many of these, each in a different GL state - * corresponding to a different brw_wm_prog_key struct, with different - * compiled programs: - */ - -struct brw_wm_prog_data { - unsigned curb_read_length; - unsigned urb_read_length; - - unsigned first_curbe_grf; - unsigned total_grf; - unsigned total_scratch; - - /* Internally generated constants for the CURBE. These are loaded - * ahead of the data from the constant buffer. - */ - const float internal_const[8]; - unsigned nr_internal_consts; - unsigned max_const; - - boolean error; -}; - -struct brw_sf_prog_data { - unsigned urb_read_length; - unsigned total_grf; - - /* Each vertex may have upto 12 attributes, 4 components each, - * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 - * rows. - * - * Actually we use 4 for each, so call it 12 rows. - */ - unsigned urb_entry_size; -}; - -struct brw_clip_prog_data { - unsigned curb_read_length; /* user planes? */ - unsigned clip_mode; - unsigned urb_read_length; - unsigned total_grf; -}; - -struct brw_gs_prog_data { - unsigned urb_read_length; - unsigned total_grf; -}; - -struct brw_vs_prog_data { - unsigned curb_read_length; - unsigned urb_read_length; - unsigned total_grf; - unsigned outputs_written; - - unsigned inputs_read; - - unsigned max_const; - - float imm_buf[PIPE_MAX_CONSTANT][4]; - unsigned num_imm; - unsigned num_consts; - - /* Used for calculating urb partitions: - */ - unsigned urb_entry_size; -}; - - -#define BRW_MAX_TEX_UNIT 8 -#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 - -/* Create a fixed sized struct for caching binding tables: - */ -struct brw_surface_binding_table { - unsigned surf_ss_offset[BRW_WM_MAX_SURF]; -}; - - -struct brw_cache; - -struct brw_mem_pool { - struct pipe_buffer *buffer; - - unsigned size; - unsigned offset; /* offset of first free byte */ - - struct brw_context *brw; -}; - -struct brw_cache_item { - unsigned hash; - unsigned key_size; /* for variable-sized keys */ - const void *key; - - unsigned offset; /* offset within pool's buffer */ - unsigned data_size; - - struct brw_cache_item *next; -}; - - - -struct brw_cache { - unsigned id; - - const char *name; - - struct brw_context *brw; - struct brw_mem_pool *pool; - - struct brw_cache_item **items; - unsigned size, n_items; - - unsigned key_size; /* for fixed-size keys */ - unsigned aux_size; - - unsigned last_addr; /* offset of active item */ -}; - - - - -/* Considered adding a member to this struct to document which flags - * an update might raise so that ordering of the state atoms can be - * checked or derived at runtime. Dropped the idea in favor of having - * a debug mode where the state is monitored for flags which are - * raised that have already been tested against. - */ -struct brw_tracked_state { - struct brw_state_flags dirty; - void (*update)( struct brw_context *brw ); -}; - - -/* Flags for brw->state.cache. - */ -#define CACHE_NEW_CC_VP (1< 32. Wouldn't life - * be easier if C allowed arrays of packed elements? - */ -#define ATTRIB_BIT_DWORDS ((PIPE_ATTRIB_MAX+31)/32) - - - - -struct brw_vertex_info { - unsigned varying; /* varying:1[PIPE_ATTRIB_MAX] */ - unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_ATTRIB_MAX] */ -}; - - - - - -struct brw_context -{ - struct pipe_context pipe; - struct brw_winsys *winsys; - - unsigned primitive; - unsigned reduced_primitive; - - boolean emit_state_always; - - struct { - struct brw_state_flags dirty; - } state; - - - struct { - const struct pipe_blend_state *Blend; - const struct pipe_depth_stencil_alpha_state *DepthStencil; - const struct pipe_poly_stipple *PolygonStipple; - const struct pipe_rasterizer_state *Raster; - const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; - const struct brw_vertex_program *VertexProgram; - const struct brw_fragment_program *FragmentProgram; - - struct pipe_clip_state Clip; - struct pipe_blend_color BlendColor; - struct pipe_scissor_state Scissor; - struct pipe_viewport_state Viewport; - struct pipe_framebuffer_state FrameBuffer; - - const struct pipe_constant_buffer *Constants[2]; - const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; - } attribs; - - struct brw_mem_pool pool[BRW_MAX_POOL]; - struct brw_cache cache[BRW_MAX_CACHE]; - struct brw_cached_batch_item *cached_batch_items; - - struct { - - /* Arrays with buffer objects to copy non-bufferobj arrays into - * for upload: - */ - struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX]; - - struct brw_vertex_element_state inputs[PIPE_ATTRIB_MAX]; - -#define BRW_NR_UPLOAD_BUFS 17 -#define BRW_UPLOAD_INIT_SIZE (128*1024) - - /* Summary of size and varying of active arrays, so we can check - * for changes to this state: - */ - struct brw_vertex_info info; - } vb; - - - unsigned hardware_dirty; - unsigned dirty; - unsigned pci_id; - /* BRW_NEW_URB_ALLOCATIONS: - */ - struct { - unsigned vsize; /* vertex size plus header in urb registers */ - unsigned csize; /* constant buffer size in urb registers */ - unsigned sfsize; /* setup data size in urb registers */ - - boolean constrained; - - unsigned nr_vs_entries; - unsigned nr_gs_entries; - unsigned nr_clip_entries; - unsigned nr_sf_entries; - unsigned nr_cs_entries; - -/* unsigned vs_size; */ -/* unsigned gs_size; */ -/* unsigned clip_size; */ -/* unsigned sf_size; */ -/* unsigned cs_size; */ - - unsigned vs_start; - unsigned gs_start; - unsigned clip_start; - unsigned sf_start; - unsigned cs_start; - } urb; - - - /* BRW_NEW_CURBE_OFFSETS: - */ - struct { - unsigned wm_start; - unsigned wm_size; - unsigned clip_start; - unsigned clip_size; - unsigned vs_start; - unsigned vs_size; - unsigned total_size; - - unsigned gs_offset; - - float *last_buf; - unsigned last_bufsz; - } curbe; - - struct { - struct brw_vs_prog_data *prog_data; - - unsigned prog_gs_offset; - unsigned state_gs_offset; - } vs; - - struct { - struct brw_gs_prog_data *prog_data; - - boolean prog_active; - unsigned prog_gs_offset; - unsigned state_gs_offset; - } gs; - - struct { - struct brw_clip_prog_data *prog_data; - - unsigned prog_gs_offset; - unsigned vp_gs_offset; - unsigned state_gs_offset; - } clip; - - - struct { - struct brw_sf_prog_data *prog_data; - - struct pipe_setup_linkage linkage; - - unsigned prog_gs_offset; - unsigned vp_gs_offset; - unsigned state_gs_offset; - } sf; - - struct { - struct brw_wm_prog_data *prog_data; - -// struct brw_wm_compiler *compile_data; - - - /** - * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER - * cache - */ - struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; - - unsigned render_surf; - unsigned nr_surfaces; - - unsigned max_threads; - struct pipe_buffer *scratch_buffer; - unsigned scratch_buffer_size; - - unsigned sampler_count; - unsigned sampler_gs_offset; - - struct brw_surface_binding_table bind; - unsigned bind_ss_offset; - - unsigned prog_gs_offset; - unsigned state_gs_offset; - } wm; - - - struct { - unsigned vp_gs_offset; - unsigned state_gs_offset; - } cc; - - - /* Used to give every program string a unique id - */ - unsigned program_id; -}; - - -#define BRW_PACKCOLOR8888(r,g,b,a) ((r<<24) | (g<<16) | (b<<8) | a) - - -/*====================================================================== - * brw_vtbl.c - */ -void brw_do_flush( struct brw_context *brw, - unsigned flags ); - - -/*====================================================================== - * brw_state.c - */ -void brw_validate_state(struct brw_context *brw); -void brw_init_state(struct brw_context *brw); -void brw_destroy_state(struct brw_context *brw); - - -/*====================================================================== - * brw_tex.c - */ -void brwUpdateTextureState( struct brw_context *brw ); - - -/* brw_urb.c - */ -void brw_upload_urb_fence(struct brw_context *brw); - -void brw_upload_constant_buffer_state(struct brw_context *brw); - -void brw_init_surface_functions(struct brw_context *brw); -void brw_init_state_functions(struct brw_context *brw); -void brw_init_flush_functions(struct brw_context *brw); -void brw_init_string_functions(struct brw_context *brw); - -/*====================================================================== - * Inline conversion functions. These are better-typed than the - * macros used previously: - */ -static inline struct brw_context * -brw_context( struct pipe_context *ctx ) -{ - return (struct brw_context *)ctx; -} - -#endif - diff --git a/src/mesa/pipe/i965simple/brw_curbe.c b/src/mesa/pipe/i965simple/brw_curbe.c deleted file mode 100644 index 52bbd525c1..0000000000 --- a/src/mesa/pipe/i965simple/brw_curbe.c +++ /dev/null @@ -1,368 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_state.h" -#include "brw_batch.h" -#include "brw_util.h" -#include "brw_wm.h" -#include "pipe/p_state.h" -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" - -#define FILE_DEBUG_FLAG DEBUG_FALLBACKS - -/* Partition the CURBE between the various users of constant values: - */ -static void calculate_curbe_offsets( struct brw_context *brw ) -{ - /* CACHE_NEW_WM_PROG */ - unsigned nr_fp_regs = align(brw->wm.prog_data->max_const, 16); - - /* BRW_NEW_VERTEX_PROGRAM */ - unsigned nr_vp_regs = align(brw->vs.prog_data->max_const, 16); - unsigned nr_clip_regs = 0; - unsigned total_regs; - -#if 0 - /* BRW_NEW_CLIP ? */ - if (brw->attribs.Transform->ClipPlanesEnabled) { - unsigned nr_planes = 6 + brw_count_bits(brw->attribs.Transform->ClipPlanesEnabled); - nr_clip_regs = align(nr_planes * 4, 16); - } -#endif - - - total_regs = nr_fp_regs + nr_vp_regs + nr_clip_regs; - - /* This can happen - what to do? Probably rather than falling - * back, the best thing to do is emit programs which code the - * constants as immediate values. Could do this either as a static - * cap on WM and VS, or adaptively. - * - * Unfortunately, this is currently dependent on the results of the - * program generation process (in the case of wm), so this would - * introduce the need to re-generate programs in the event of a - * curbe allocation failure. - */ - /* Max size is 32 - just large enough to - * hold the 128 parameters allowed by - * the fragment and vertex program - * api's. It's not clear what happens - * when both VP and FP want to use 128 - * parameters, though. - */ - assert(total_regs <= 32); - - /* Lazy resize: - */ - if (nr_fp_regs > brw->curbe.wm_size || - nr_vp_regs > brw->curbe.vs_size || - nr_clip_regs != brw->curbe.clip_size || - (total_regs < brw->curbe.total_size / 4 && - brw->curbe.total_size > 16)) { - - unsigned reg = 0; - - /* Calculate a new layout: - */ - reg = 0; - brw->curbe.wm_start = reg; - brw->curbe.wm_size = nr_fp_regs; reg += nr_fp_regs; - brw->curbe.clip_start = reg; - brw->curbe.clip_size = nr_clip_regs; reg += nr_clip_regs; - brw->curbe.vs_start = reg; - brw->curbe.vs_size = nr_vp_regs; reg += nr_vp_regs; - brw->curbe.total_size = reg; - -#if 0 - if (0) - DBG("curbe wm %d+%d clip %d+%d vs %d+%d\n", - brw->curbe.wm_start, - brw->curbe.wm_size, - brw->curbe.clip_start, - brw->curbe.clip_size, - brw->curbe.vs_start, - brw->curbe.vs_size ); -#endif - - brw->state.dirty.brw |= BRW_NEW_CURBE_OFFSETS; - } -} - - -const struct brw_tracked_state brw_curbe_offsets = { - .dirty = { - .brw = (BRW_NEW_CLIP | - BRW_NEW_VS), - .cache = CACHE_NEW_WM_PROG - }, - .update = calculate_curbe_offsets -}; - - - -/* Define the number of curbes within CS's urb allocation. Multiple - * urb entries -> multiple curbes. These will be used by - * fixed-function hardware in a double-buffering scheme to avoid a - * pipeline stall each time the contents of the curbe is changed. - */ -void brw_upload_constant_buffer_state(struct brw_context *brw) -{ - struct brw_constant_buffer_state cbs; - memset(&cbs, 0, sizeof(cbs)); - - /* It appears that this is the state packet for the CS unit, ie. the - * urb entries detailed here are housed in the CS range from the - * URB_FENCE command. - */ - cbs.header.opcode = CMD_CONST_BUFFER_STATE; - cbs.header.length = sizeof(cbs)/4 - 2; - - /* BRW_NEW_URB_FENCE */ - cbs.bits0.nr_urb_entries = brw->urb.nr_cs_entries; - cbs.bits0.urb_entry_size = brw->urb.csize - 1; - - assert(brw->urb.nr_cs_entries); - BRW_CACHED_BATCH_STRUCT(brw, &cbs); -} - - -static float fixed_plane[6][4] = { - { 0, 0, -1, 1 }, - { 0, 0, 1, 1 }, - { 0, -1, 0, 1 }, - { 0, 1, 0, 1 }, - {-1, 0, 0, 1 }, - { 1, 0, 0, 1 } -}; - -/* Upload a new set of constants. Too much variability to go into the - * cache mechanism, but maybe would benefit from a comparison against - * the current uploaded set of constants. - */ -static void upload_constant_buffer(struct brw_context *brw) -{ - struct brw_mem_pool *pool = &brw->pool[BRW_GS_POOL]; - unsigned sz = brw->curbe.total_size; - unsigned bufsz = sz * sizeof(float); - float *buf; - unsigned i; - - - if (sz == 0) { - struct brw_constant_buffer cb; - cb.header.opcode = CMD_CONST_BUFFER; - cb.header.length = sizeof(cb)/4 - 2; - cb.header.valid = 0; - cb.bits0.buffer_length = 0; - cb.bits0.buffer_address = 0; - BRW_BATCH_STRUCT(brw, &cb); - - if (brw->curbe.last_buf) { - free(brw->curbe.last_buf); - brw->curbe.last_buf = NULL; - brw->curbe.last_bufsz = 0; - } - - return; - } - - buf = (float *)malloc(bufsz); - - memset(buf, 0, bufsz); - - if (brw->curbe.wm_size) { - unsigned offset = brw->curbe.wm_start * 16; - - /* First the constant buffer constants: - */ - - /* Then any internally generated constants: - */ - for (i = 0; i < brw->wm.prog_data->nr_internal_consts; i++) - buf[offset + i] = brw->wm.prog_data->internal_const[i]; - - assert(brw->wm.prog_data->max_const == - brw->wm.prog_data->nr_internal_consts); - } - - - /* The clipplanes are actually delivered to both CLIP and VS units. - * VS uses them to calculate the outcode bitmasks. - */ - if (brw->curbe.clip_size) { - unsigned offset = brw->curbe.clip_start * 16; - unsigned j; - - /* If any planes are going this way, send them all this way: - */ - for (i = 0; i < 6; i++) { - buf[offset + i * 4 + 0] = fixed_plane[i][0]; - buf[offset + i * 4 + 1] = fixed_plane[i][1]; - buf[offset + i * 4 + 2] = fixed_plane[i][2]; - buf[offset + i * 4 + 3] = fixed_plane[i][3]; - } - - /* Clip planes: BRW_NEW_CLIP: - */ - for (j = 0; j < brw->attribs.Clip.nr; j++) { - buf[offset + i * 4 + 0] = brw->attribs.Clip.ucp[j][0]; - buf[offset + i * 4 + 1] = brw->attribs.Clip.ucp[j][1]; - buf[offset + i * 4 + 2] = brw->attribs.Clip.ucp[j][2]; - buf[offset + i * 4 + 3] = brw->attribs.Clip.ucp[j][3]; - i++; - } - } - - - if (brw->curbe.vs_size) { - unsigned offset = brw->curbe.vs_start * 16; - /*unsigned nr = vp->max_const;*/ - const struct pipe_constant_buffer *cbuffer = brw->attribs.Constants[0]; - struct pipe_winsys *ws = brw->pipe.winsys; - /* FIXME: buffer size is num_consts + num_immediates */ - if (brw->vs.prog_data->num_consts) { - /* map the vertex constant buffer and copy to curbe: */ - void *data = ws->buffer_map(ws, cbuffer->buffer, 0); - /* FIXME: this is wrong. the cbuffer->size currently - * represents size of consts + immediates. so if we'll - * have both we'll copy over the end of the buffer - * with the subsequent memcpy */ - memcpy(&buf[offset], data, cbuffer->size); - ws->buffer_unmap(ws, cbuffer->buffer); - offset += cbuffer->size; - } - /*immediates*/ - if (brw->vs.prog_data->num_imm) { - memcpy(&buf[offset], brw->vs.prog_data->imm_buf, - brw->vs.prog_data->num_imm * 4 * sizeof(float)); - } - } - - if (1) { - for (i = 0; i < sz; i+=4) - debug_printf("curbe %d.%d: %f %f %f %f\n", i/8, i&4, - buf[i+0], buf[i+1], buf[i+2], buf[i+3]); - - debug_printf("last_buf %p buf %p sz %d/%d cmp %d\n", - brw->curbe.last_buf, buf, - bufsz, brw->curbe.last_bufsz, - brw->curbe.last_buf ? memcmp(buf, brw->curbe.last_buf, bufsz) : -1); - } - - if (brw->curbe.last_buf && - bufsz == brw->curbe.last_bufsz && - memcmp(buf, brw->curbe.last_buf, bufsz) == 0) { - free(buf); -/* return; */ - } - else { - if (brw->curbe.last_buf) - free(brw->curbe.last_buf); - brw->curbe.last_buf = buf; - brw->curbe.last_bufsz = bufsz; - - - if (!brw_pool_alloc(pool, - bufsz, - 1 << 6, - &brw->curbe.gs_offset)) { - debug_printf("out of GS memory for curbe\n"); - assert(0); - return; - } - - - /* Copy data to the buffer: - */ - brw->winsys->buffer_subdata_typed(brw->winsys, - pool->buffer, - brw->curbe.gs_offset, - bufsz, - buf, - BRW_CONSTANT_BUFFER ); - } - - /* TODO: only emit the constant_buffer packet when necessary, ie: - - contents have changed - - offset has changed - - hw requirements due to other packets emitted. - */ - { - struct brw_constant_buffer cb; - - memset(&cb, 0, sizeof(cb)); - - cb.header.opcode = CMD_CONST_BUFFER; - cb.header.length = sizeof(cb)/4 - 2; - cb.header.valid = 1; - cb.bits0.buffer_length = sz - 1; - cb.bits0.buffer_address = brw->curbe.gs_offset >> 6; - - /* Because this provokes an action (ie copy the constants into the - * URB), it shouldn't be shortcircuited if identical to the - * previous time - because eg. the urb destination may have - * changed, or the urb contents different to last time. - * - * Note that the data referred to is actually copied internally, - * not just used in place according to passed pointer. - * - * It appears that the CS unit takes care of using each available - * URB entry (Const URB Entry == CURBE) in turn, and issuing - * flushes as necessary when doublebuffering of CURBEs isn't - * possible. - */ - BRW_BATCH_STRUCT(brw, &cb); - } -} - -/* This tracked state is unique in that the state it monitors varies - * dynamically depending on the parameters tracked by the fragment and - * vertex programs. This is the template used as a starting point, - * each context will maintain a copy of this internally and update as - * required. - */ -const struct brw_tracked_state brw_constant_buffer = { - .dirty = { - .brw = (BRW_NEW_CLIP | - BRW_NEW_CONSTANTS | - BRW_NEW_URB_FENCE | /* Implicit - hardware requires this, not used above */ - BRW_NEW_PSP | /* Implicit - hardware requires this, not used above */ - BRW_NEW_CURBE_OFFSETS), - .cache = (CACHE_NEW_WM_PROG) - }, - .update = upload_constant_buffer -}; - diff --git a/src/mesa/pipe/i965simple/brw_defines.h b/src/mesa/pipe/i965simple/brw_defines.h deleted file mode 100644 index 9379a397f6..0000000000 --- a/src/mesa/pipe/i965simple/brw_defines.h +++ /dev/null @@ -1,852 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRW_DEFINES_H -#define BRW_DEFINES_H - -/* - */ -#define MI_NOOP 0x00 -#define MI_USER_INTERRUPT 0x02 -#define MI_WAIT_FOR_EVENT 0x03 -#define MI_FLUSH 0x04 -#define MI_REPORT_HEAD 0x07 -#define MI_ARB_ON_OFF 0x08 -#define MI_BATCH_BUFFER_END 0x0A -#define MI_OVERLAY_FLIP 0x11 -#define MI_LOAD_SCAN_LINES_INCL 0x12 -#define MI_LOAD_SCAN_LINES_EXCL 0x13 -#define MI_DISPLAY_BUFFER_INFO 0x14 -#define MI_SET_CONTEXT 0x18 -#define MI_STORE_DATA_IMM 0x20 -#define MI_STORE_DATA_INDEX 0x21 -#define MI_LOAD_REGISTER_IMM 0x22 -#define MI_STORE_REGISTER_MEM 0x24 -#define MI_BATCH_BUFFER_START 0x31 - -#define MI_SYNCHRONOUS_FLIP 0x0 -#define MI_ASYNCHRONOUS_FLIP 0x1 - -#define MI_BUFFER_SECURE 0x0 -#define MI_BUFFER_NONSECURE 0x1 - -#define MI_ARBITRATE_AT_CHAIN_POINTS 0x0 -#define MI_ARBITRATE_BETWEEN_INSTS 0x1 -#define MI_NO_ARBITRATION 0x3 - -#define MI_CONDITION_CODE_WAIT_DISABLED 0x0 -#define MI_CONDITION_CODE_WAIT_0 0x1 -#define MI_CONDITION_CODE_WAIT_1 0x2 -#define MI_CONDITION_CODE_WAIT_2 0x3 -#define MI_CONDITION_CODE_WAIT_3 0x4 -#define MI_CONDITION_CODE_WAIT_4 0x5 - -#define MI_DISPLAY_PIPE_A 0x0 -#define MI_DISPLAY_PIPE_B 0x1 - -#define MI_DISPLAY_PLANE_A 0x0 -#define MI_DISPLAY_PLANE_B 0x1 -#define MI_DISPLAY_PLANE_C 0x2 - -#define MI_STANDARD_FLIP 0x0 -#define MI_ENQUEUE_FLIP_PERFORM_BASE_FRAME_NUMBER_LOAD 0x1 -#define MI_ENQUEUE_FLIP_TARGET_FRAME_NUMBER_RELATIVE 0x2 -#define MI_ENQUEUE_FLIP_ABSOLUTE_TARGET_FRAME_NUMBER 0x3 - -#define MI_PHYSICAL_ADDRESS 0x0 -#define MI_VIRTUAL_ADDRESS 0x1 - -#define MI_BUFFER_MEMORY_MAIN 0x0 -#define MI_BUFFER_MEMORY_GTT 0x2 -#define MI_BUFFER_MEMORY_PER_PROCESS_GTT 0x3 - -#define MI_FLIP_CONTINUE 0x0 -#define MI_FLIP_ON 0x1 -#define MI_FLIP_OFF 0x2 - -#define MI_UNTRUSTED_REGISTER_SPACE 0x0 -#define MI_TRUSTED_REGISTER_SPACE 0x1 - -/* 3D state: - */ -#define _3DOP_3DSTATE_PIPELINED 0x0 -#define _3DOP_3DSTATE_NONPIPELINED 0x1 -#define _3DOP_3DCONTROL 0x2 -#define _3DOP_3DPRIMITIVE 0x3 - -#define _3DSTATE_PIPELINED_POINTERS 0x00 -#define _3DSTATE_BINDING_TABLE_POINTERS 0x01 -#define _3DSTATE_VERTEX_BUFFERS 0x08 -#define _3DSTATE_VERTEX_ELEMENTS 0x09 -#define _3DSTATE_INDEX_BUFFER 0x0A -#define _3DSTATE_VF_STATISTICS 0x0B -#define _3DSTATE_DRAWING_RECTANGLE 0x00 -#define _3DSTATE_CONSTANT_COLOR 0x01 -#define _3DSTATE_SAMPLER_PALETTE_LOAD 0x02 -#define _3DSTATE_CHROMA_KEY 0x04 -#define _3DSTATE_DEPTH_BUFFER 0x05 -#define _3DSTATE_POLY_STIPPLE_OFFSET 0x06 -#define _3DSTATE_POLY_STIPPLE_PATTERN 0x07 -#define _3DSTATE_LINE_STIPPLE 0x08 -#define _3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP 0x09 -#define _3DCONTROL 0x00 -#define _3DPRIMITIVE 0x00 - -#define PIPE_CONTROL_NOWRITE 0x00 -#define PIPE_CONTROL_WRITEIMMEDIATE 0x01 -#define PIPE_CONTROL_WRITEDEPTH 0x02 -#define PIPE_CONTROL_WRITETIMESTAMP 0x03 - -#define PIPE_CONTROL_GTTWRITE_PROCESS_LOCAL 0x00 -#define PIPE_CONTROL_GTTWRITE_GLOBAL 0x01 - -#define _3DPRIM_POINTLIST 0x01 -#define _3DPRIM_LINELIST 0x02 -#define _3DPRIM_LINESTRIP 0x03 -#define _3DPRIM_TRILIST 0x04 -#define _3DPRIM_TRISTRIP 0x05 -#define _3DPRIM_TRIFAN 0x06 -#define _3DPRIM_QUADLIST 0x07 -#define _3DPRIM_QUADSTRIP 0x08 -#define _3DPRIM_LINELIST_ADJ 0x09 -#define _3DPRIM_LINESTRIP_ADJ 0x0A -#define _3DPRIM_TRILIST_ADJ 0x0B -#define _3DPRIM_TRISTRIP_ADJ 0x0C -#define _3DPRIM_TRISTRIP_REVERSE 0x0D -#define _3DPRIM_POLYGON 0x0E -#define _3DPRIM_RECTLIST 0x0F -#define _3DPRIM_LINELOOP 0x10 -#define _3DPRIM_POINTLIST_BF 0x11 -#define _3DPRIM_LINESTRIP_CONT 0x12 -#define _3DPRIM_LINESTRIP_BF 0x13 -#define _3DPRIM_LINESTRIP_CONT_BF 0x14 -#define _3DPRIM_TRIFAN_NOSTIPPLE 0x15 - -#define _3DPRIM_VERTEXBUFFER_ACCESS_SEQUENTIAL 0 -#define _3DPRIM_VERTEXBUFFER_ACCESS_RANDOM 1 - -#define BRW_ANISORATIO_2 0 -#define BRW_ANISORATIO_4 1 -#define BRW_ANISORATIO_6 2 -#define BRW_ANISORATIO_8 3 -#define BRW_ANISORATIO_10 4 -#define BRW_ANISORATIO_12 5 -#define BRW_ANISORATIO_14 6 -#define BRW_ANISORATIO_16 7 - -#define BRW_BLENDFACTOR_ONE 0x1 -#define BRW_BLENDFACTOR_SRC_COLOR 0x2 -#define BRW_BLENDFACTOR_SRC_ALPHA 0x3 -#define BRW_BLENDFACTOR_DST_ALPHA 0x4 -#define BRW_BLENDFACTOR_DST_COLOR 0x5 -#define BRW_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 -#define BRW_BLENDFACTOR_CONST_COLOR 0x7 -#define BRW_BLENDFACTOR_CONST_ALPHA 0x8 -#define BRW_BLENDFACTOR_SRC1_COLOR 0x9 -#define BRW_BLENDFACTOR_SRC1_ALPHA 0x0A -#define BRW_BLENDFACTOR_ZERO 0x11 -#define BRW_BLENDFACTOR_INV_SRC_COLOR 0x12 -#define BRW_BLENDFACTOR_INV_SRC_ALPHA 0x13 -#define BRW_BLENDFACTOR_INV_DST_ALPHA 0x14 -#define BRW_BLENDFACTOR_INV_DST_COLOR 0x15 -#define BRW_BLENDFACTOR_INV_CONST_COLOR 0x17 -#define BRW_BLENDFACTOR_INV_CONST_ALPHA 0x18 -#define BRW_BLENDFACTOR_INV_SRC1_COLOR 0x19 -#define BRW_BLENDFACTOR_INV_SRC1_ALPHA 0x1A - -#define BRW_BLENDFUNCTION_ADD 0 -#define BRW_BLENDFUNCTION_SUBTRACT 1 -#define BRW_BLENDFUNCTION_REVERSE_SUBTRACT 2 -#define BRW_BLENDFUNCTION_MIN 3 -#define BRW_BLENDFUNCTION_MAX 4 - -#define BRW_ALPHATEST_FORMAT_UNORM8 0 -#define BRW_ALPHATEST_FORMAT_FLOAT32 1 - -#define BRW_CHROMAKEY_KILL_ON_ANY_MATCH 0 -#define BRW_CHROMAKEY_REPLACE_BLACK 1 - -#define BRW_CLIP_API_OGL 0 -#define BRW_CLIP_API_DX 1 - -#define BRW_CLIPMODE_NORMAL 0 -#define BRW_CLIPMODE_CLIP_ALL 1 -#define BRW_CLIPMODE_CLIP_NON_REJECTED 2 -#define BRW_CLIPMODE_REJECT_ALL 3 -#define BRW_CLIPMODE_ACCEPT_ALL 4 - -#define BRW_CLIP_NDCSPACE 0 -#define BRW_CLIP_SCREENSPACE 1 - -#define BRW_COMPAREFUNCTION_ALWAYS 0 -#define BRW_COMPAREFUNCTION_NEVER 1 -#define BRW_COMPAREFUNCTION_LESS 2 -#define BRW_COMPAREFUNCTION_EQUAL 3 -#define BRW_COMPAREFUNCTION_LEQUAL 4 -#define BRW_COMPAREFUNCTION_GREATER 5 -#define BRW_COMPAREFUNCTION_NOTEQUAL 6 -#define BRW_COMPAREFUNCTION_GEQUAL 7 - -#define BRW_COVERAGE_PIXELS_HALF 0 -#define BRW_COVERAGE_PIXELS_1 1 -#define BRW_COVERAGE_PIXELS_2 2 -#define BRW_COVERAGE_PIXELS_4 3 - -#define BRW_CULLMODE_BOTH 0 -#define BRW_CULLMODE_NONE 1 -#define BRW_CULLMODE_FRONT 2 -#define BRW_CULLMODE_BACK 3 - -#define BRW_DEFAULTCOLOR_R8G8B8A8_UNORM 0 -#define BRW_DEFAULTCOLOR_R32G32B32A32_FLOAT 1 - -#define BRW_DEPTHFORMAT_D32_FLOAT_S8X24_UINT 0 -#define BRW_DEPTHFORMAT_D32_FLOAT 1 -#define BRW_DEPTHFORMAT_D24_UNORM_S8_UINT 2 -#define BRW_DEPTHFORMAT_D16_UNORM 5 - -#define BRW_FLOATING_POINT_IEEE_754 0 -#define BRW_FLOATING_POINT_NON_IEEE_754 1 - -#define BRW_FRONTWINDING_CW 0 -#define BRW_FRONTWINDING_CCW 1 - -#define BRW_SPRITE_POINT_ENABLE 16 - -#define BRW_INDEX_BYTE 0 -#define BRW_INDEX_WORD 1 -#define BRW_INDEX_DWORD 2 - -#define BRW_LOGICOPFUNCTION_CLEAR 0 -#define BRW_LOGICOPFUNCTION_NOR 1 -#define BRW_LOGICOPFUNCTION_AND_INVERTED 2 -#define BRW_LOGICOPFUNCTION_COPY_INVERTED 3 -#define BRW_LOGICOPFUNCTION_AND_REVERSE 4 -#define BRW_LOGICOPFUNCTION_INVERT 5 -#define BRW_LOGICOPFUNCTION_XOR 6 -#define BRW_LOGICOPFUNCTION_NAND 7 -#define BRW_LOGICOPFUNCTION_AND 8 -#define BRW_LOGICOPFUNCTION_EQUIV 9 -#define BRW_LOGICOPFUNCTION_NOOP 10 -#define BRW_LOGICOPFUNCTION_OR_INVERTED 11 -#define BRW_LOGICOPFUNCTION_COPY 12 -#define BRW_LOGICOPFUNCTION_OR_REVERSE 13 -#define BRW_LOGICOPFUNCTION_OR 14 -#define BRW_LOGICOPFUNCTION_SET 15 - -#define BRW_MAPFILTER_NEAREST 0x0 -#define BRW_MAPFILTER_LINEAR 0x1 -#define BRW_MAPFILTER_ANISOTROPIC 0x2 - -#define BRW_MIPFILTER_NONE 0 -#define BRW_MIPFILTER_NEAREST 1 -#define BRW_MIPFILTER_LINEAR 3 - -#define BRW_POLYGON_FRONT_FACING 0 -#define BRW_POLYGON_BACK_FACING 1 - -#define BRW_PREFILTER_ALWAYS 0x0 -#define BRW_PREFILTER_NEVER 0x1 -#define BRW_PREFILTER_LESS 0x2 -#define BRW_PREFILTER_EQUAL 0x3 -#define BRW_PREFILTER_LEQUAL 0x4 -#define BRW_PREFILTER_GREATER 0x5 -#define BRW_PREFILTER_NOTEQUAL 0x6 -#define BRW_PREFILTER_GEQUAL 0x7 - -#define BRW_PROVOKING_VERTEX_0 0 -#define BRW_PROVOKING_VERTEX_1 1 -#define BRW_PROVOKING_VERTEX_2 2 - -#define BRW_RASTRULE_UPPER_LEFT 0 -#define BRW_RASTRULE_UPPER_RIGHT 1 - -#define BRW_RENDERTARGET_CLAMPRANGE_UNORM 0 -#define BRW_RENDERTARGET_CLAMPRANGE_SNORM 1 -#define BRW_RENDERTARGET_CLAMPRANGE_FORMAT 2 - -#define BRW_STENCILOP_KEEP 0 -#define BRW_STENCILOP_ZERO 1 -#define BRW_STENCILOP_REPLACE 2 -#define BRW_STENCILOP_INCRSAT 3 -#define BRW_STENCILOP_DECRSAT 4 -#define BRW_STENCILOP_INCR 5 -#define BRW_STENCILOP_DECR 6 -#define BRW_STENCILOP_INVERT 7 - -#define BRW_SURFACE_MIPMAPLAYOUT_BELOW 0 -#define BRW_SURFACE_MIPMAPLAYOUT_RIGHT 1 - -#define BRW_SURFACEFORMAT_R32G32B32A32_FLOAT 0x000 -#define BRW_SURFACEFORMAT_R32G32B32A32_SINT 0x001 -#define BRW_SURFACEFORMAT_R32G32B32A32_UINT 0x002 -#define BRW_SURFACEFORMAT_R32G32B32A32_UNORM 0x003 -#define BRW_SURFACEFORMAT_R32G32B32A32_SNORM 0x004 -#define BRW_SURFACEFORMAT_R64G64_FLOAT 0x005 -#define BRW_SURFACEFORMAT_R32G32B32X32_FLOAT 0x006 -#define BRW_SURFACEFORMAT_R32G32B32A32_SSCALED 0x007 -#define BRW_SURFACEFORMAT_R32G32B32A32_USCALED 0x008 -#define BRW_SURFACEFORMAT_R32G32B32_FLOAT 0x040 -#define BRW_SURFACEFORMAT_R32G32B32_SINT 0x041 -#define BRW_SURFACEFORMAT_R32G32B32_UINT 0x042 -#define BRW_SURFACEFORMAT_R32G32B32_UNORM 0x043 -#define BRW_SURFACEFORMAT_R32G32B32_SNORM 0x044 -#define BRW_SURFACEFORMAT_R32G32B32_SSCALED 0x045 -#define BRW_SURFACEFORMAT_R32G32B32_USCALED 0x046 -#define BRW_SURFACEFORMAT_R16G16B16A16_UNORM 0x080 -#define BRW_SURFACEFORMAT_R16G16B16A16_SNORM 0x081 -#define BRW_SURFACEFORMAT_R16G16B16A16_SINT 0x082 -#define BRW_SURFACEFORMAT_R16G16B16A16_UINT 0x083 -#define BRW_SURFACEFORMAT_R16G16B16A16_FLOAT 0x084 -#define BRW_SURFACEFORMAT_R32G32_FLOAT 0x085 -#define BRW_SURFACEFORMAT_R32G32_SINT 0x086 -#define BRW_SURFACEFORMAT_R32G32_UINT 0x087 -#define BRW_SURFACEFORMAT_R32_FLOAT_X8X24_TYPELESS 0x088 -#define BRW_SURFACEFORMAT_X32_TYPELESS_G8X24_UINT 0x089 -#define BRW_SURFACEFORMAT_L32A32_FLOAT 0x08A -#define BRW_SURFACEFORMAT_R32G32_UNORM 0x08B -#define BRW_SURFACEFORMAT_R32G32_SNORM 0x08C -#define BRW_SURFACEFORMAT_R64_FLOAT 0x08D -#define BRW_SURFACEFORMAT_R16G16B16X16_UNORM 0x08E -#define BRW_SURFACEFORMAT_R16G16B16X16_FLOAT 0x08F -#define BRW_SURFACEFORMAT_A32X32_FLOAT 0x090 -#define BRW_SURFACEFORMAT_L32X32_FLOAT 0x091 -#define BRW_SURFACEFORMAT_I32X32_FLOAT 0x092 -#define BRW_SURFACEFORMAT_R16G16B16A16_SSCALED 0x093 -#define BRW_SURFACEFORMAT_R16G16B16A16_USCALED 0x094 -#define BRW_SURFACEFORMAT_R32G32_SSCALED 0x095 -#define BRW_SURFACEFORMAT_R32G32_USCALED 0x096 -#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM 0x0C0 -#define BRW_SURFACEFORMAT_B8G8R8A8_UNORM_SRGB 0x0C1 -#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM 0x0C2 -#define BRW_SURFACEFORMAT_R10G10B10A2_UNORM_SRGB 0x0C3 -#define BRW_SURFACEFORMAT_R10G10B10A2_UINT 0x0C4 -#define BRW_SURFACEFORMAT_R10G10B10_SNORM_A2_UNORM 0x0C5 -#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM 0x0C7 -#define BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB 0x0C8 -#define BRW_SURFACEFORMAT_R8G8B8A8_SNORM 0x0C9 -#define BRW_SURFACEFORMAT_R8G8B8A8_SINT 0x0CA -#define BRW_SURFACEFORMAT_R8G8B8A8_UINT 0x0CB -#define BRW_SURFACEFORMAT_R16G16_UNORM 0x0CC -#define BRW_SURFACEFORMAT_R16G16_SNORM 0x0CD -#define BRW_SURFACEFORMAT_R16G16_SINT 0x0CE -#define BRW_SURFACEFORMAT_R16G16_UINT 0x0CF -#define BRW_SURFACEFORMAT_R16G16_FLOAT 0x0D0 -#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM 0x0D1 -#define BRW_SURFACEFORMAT_B10G10R10A2_UNORM_SRGB 0x0D2 -#define BRW_SURFACEFORMAT_R11G11B10_FLOAT 0x0D3 -#define BRW_SURFACEFORMAT_R32_SINT 0x0D6 -#define BRW_SURFACEFORMAT_R32_UINT 0x0D7 -#define BRW_SURFACEFORMAT_R32_FLOAT 0x0D8 -#define BRW_SURFACEFORMAT_R24_UNORM_X8_TYPELESS 0x0D9 -#define BRW_SURFACEFORMAT_X24_TYPELESS_G8_UINT 0x0DA -#define BRW_SURFACEFORMAT_L16A16_UNORM 0x0DF -#define BRW_SURFACEFORMAT_I24X8_UNORM 0x0E0 -#define BRW_SURFACEFORMAT_L24X8_UNORM 0x0E1 -#define BRW_SURFACEFORMAT_A24X8_UNORM 0x0E2 -#define BRW_SURFACEFORMAT_I32_FLOAT 0x0E3 -#define BRW_SURFACEFORMAT_L32_FLOAT 0x0E4 -#define BRW_SURFACEFORMAT_A32_FLOAT 0x0E5 -#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM 0x0E9 -#define BRW_SURFACEFORMAT_B8G8R8X8_UNORM_SRGB 0x0EA -#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM 0x0EB -#define BRW_SURFACEFORMAT_R8G8B8X8_UNORM_SRGB 0x0EC -#define BRW_SURFACEFORMAT_R9G9B9E5_SHAREDEXP 0x0ED -#define BRW_SURFACEFORMAT_B10G10R10X2_UNORM 0x0EE -#define BRW_SURFACEFORMAT_L16A16_FLOAT 0x0F0 -#define BRW_SURFACEFORMAT_R32_UNORM 0x0F1 -#define BRW_SURFACEFORMAT_R32_SNORM 0x0F2 -#define BRW_SURFACEFORMAT_R10G10B10X2_USCALED 0x0F3 -#define BRW_SURFACEFORMAT_R8G8B8A8_SSCALED 0x0F4 -#define BRW_SURFACEFORMAT_R8G8B8A8_USCALED 0x0F5 -#define BRW_SURFACEFORMAT_R16G16_SSCALED 0x0F6 -#define BRW_SURFACEFORMAT_R16G16_USCALED 0x0F7 -#define BRW_SURFACEFORMAT_R32_SSCALED 0x0F8 -#define BRW_SURFACEFORMAT_R32_USCALED 0x0F9 -#define BRW_SURFACEFORMAT_B5G6R5_UNORM 0x100 -#define BRW_SURFACEFORMAT_B5G6R5_UNORM_SRGB 0x101 -#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM 0x102 -#define BRW_SURFACEFORMAT_B5G5R5A1_UNORM_SRGB 0x103 -#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM 0x104 -#define BRW_SURFACEFORMAT_B4G4R4A4_UNORM_SRGB 0x105 -#define BRW_SURFACEFORMAT_R8G8_UNORM 0x106 -#define BRW_SURFACEFORMAT_R8G8_SNORM 0x107 -#define BRW_SURFACEFORMAT_R8G8_SINT 0x108 -#define BRW_SURFACEFORMAT_R8G8_UINT 0x109 -#define BRW_SURFACEFORMAT_R16_UNORM 0x10A -#define BRW_SURFACEFORMAT_R16_SNORM 0x10B -#define BRW_SURFACEFORMAT_R16_SINT 0x10C -#define BRW_SURFACEFORMAT_R16_UINT 0x10D -#define BRW_SURFACEFORMAT_R16_FLOAT 0x10E -#define BRW_SURFACEFORMAT_I16_UNORM 0x111 -#define BRW_SURFACEFORMAT_L16_UNORM 0x112 -#define BRW_SURFACEFORMAT_A16_UNORM 0x113 -#define BRW_SURFACEFORMAT_L8A8_UNORM 0x114 -#define BRW_SURFACEFORMAT_I16_FLOAT 0x115 -#define BRW_SURFACEFORMAT_L16_FLOAT 0x116 -#define BRW_SURFACEFORMAT_A16_FLOAT 0x117 -#define BRW_SURFACEFORMAT_R5G5_SNORM_B6_UNORM 0x119 -#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM 0x11A -#define BRW_SURFACEFORMAT_B5G5R5X1_UNORM_SRGB 0x11B -#define BRW_SURFACEFORMAT_R8G8_SSCALED 0x11C -#define BRW_SURFACEFORMAT_R8G8_USCALED 0x11D -#define BRW_SURFACEFORMAT_R16_SSCALED 0x11E -#define BRW_SURFACEFORMAT_R16_USCALED 0x11F -#define BRW_SURFACEFORMAT_R8_UNORM 0x140 -#define BRW_SURFACEFORMAT_R8_SNORM 0x141 -#define BRW_SURFACEFORMAT_R8_SINT 0x142 -#define BRW_SURFACEFORMAT_R8_UINT 0x143 -#define BRW_SURFACEFORMAT_A8_UNORM 0x144 -#define BRW_SURFACEFORMAT_I8_UNORM 0x145 -#define BRW_SURFACEFORMAT_L8_UNORM 0x146 -#define BRW_SURFACEFORMAT_P4A4_UNORM 0x147 -#define BRW_SURFACEFORMAT_A4P4_UNORM 0x148 -#define BRW_SURFACEFORMAT_R8_SSCALED 0x149 -#define BRW_SURFACEFORMAT_R8_USCALED 0x14A -#define BRW_SURFACEFORMAT_R1_UINT 0x181 -#define BRW_SURFACEFORMAT_YCRCB_NORMAL 0x182 -#define BRW_SURFACEFORMAT_YCRCB_SWAPUVY 0x183 -#define BRW_SURFACEFORMAT_BC1_UNORM 0x186 -#define BRW_SURFACEFORMAT_BC2_UNORM 0x187 -#define BRW_SURFACEFORMAT_BC3_UNORM 0x188 -#define BRW_SURFACEFORMAT_BC4_UNORM 0x189 -#define BRW_SURFACEFORMAT_BC5_UNORM 0x18A -#define BRW_SURFACEFORMAT_BC1_UNORM_SRGB 0x18B -#define BRW_SURFACEFORMAT_BC2_UNORM_SRGB 0x18C -#define BRW_SURFACEFORMAT_BC3_UNORM_SRGB 0x18D -#define BRW_SURFACEFORMAT_MONO8 0x18E -#define BRW_SURFACEFORMAT_YCRCB_SWAPUV 0x18F -#define BRW_SURFACEFORMAT_YCRCB_SWAPY 0x190 -#define BRW_SURFACEFORMAT_DXT1_RGB 0x191 -#define BRW_SURFACEFORMAT_FXT1 0x192 -#define BRW_SURFACEFORMAT_R8G8B8_UNORM 0x193 -#define BRW_SURFACEFORMAT_R8G8B8_SNORM 0x194 -#define BRW_SURFACEFORMAT_R8G8B8_SSCALED 0x195 -#define BRW_SURFACEFORMAT_R8G8B8_USCALED 0x196 -#define BRW_SURFACEFORMAT_R64G64B64A64_FLOAT 0x197 -#define BRW_SURFACEFORMAT_R64G64B64_FLOAT 0x198 -#define BRW_SURFACEFORMAT_BC4_SNORM 0x199 -#define BRW_SURFACEFORMAT_BC5_SNORM 0x19A -#define BRW_SURFACEFORMAT_R16G16B16_UNORM 0x19C -#define BRW_SURFACEFORMAT_R16G16B16_SNORM 0x19D -#define BRW_SURFACEFORMAT_R16G16B16_SSCALED 0x19E -#define BRW_SURFACEFORMAT_R16G16B16_USCALED 0x19F - -#define BRW_SURFACERETURNFORMAT_FLOAT32 0 -#define BRW_SURFACERETURNFORMAT_S1 1 - -#define BRW_SURFACE_1D 0 -#define BRW_SURFACE_2D 1 -#define BRW_SURFACE_3D 2 -#define BRW_SURFACE_CUBE 3 -#define BRW_SURFACE_BUFFER 4 -#define BRW_SURFACE_NULL 7 - -#define BRW_TEXCOORDMODE_WRAP 0 -#define BRW_TEXCOORDMODE_MIRROR 1 -#define BRW_TEXCOORDMODE_CLAMP 2 -#define BRW_TEXCOORDMODE_CUBE 3 -#define BRW_TEXCOORDMODE_CLAMP_BORDER 4 -#define BRW_TEXCOORDMODE_MIRROR_ONCE 5 - -#define BRW_THREAD_PRIORITY_NORMAL 0 -#define BRW_THREAD_PRIORITY_HIGH 1 - -#define BRW_TILEWALK_XMAJOR 0 -#define BRW_TILEWALK_YMAJOR 1 - -#define BRW_VERTEX_SUBPIXEL_PRECISION_8BITS 0 -#define BRW_VERTEX_SUBPIXEL_PRECISION_4BITS 1 - -#define BRW_VERTEXBUFFER_ACCESS_VERTEXDATA 0 -#define BRW_VERTEXBUFFER_ACCESS_INSTANCEDATA 1 - -#define BRW_VFCOMPONENT_NOSTORE 0 -#define BRW_VFCOMPONENT_STORE_SRC 1 -#define BRW_VFCOMPONENT_STORE_0 2 -#define BRW_VFCOMPONENT_STORE_1_FLT 3 -#define BRW_VFCOMPONENT_STORE_1_INT 4 -#define BRW_VFCOMPONENT_STORE_VID 5 -#define BRW_VFCOMPONENT_STORE_IID 6 -#define BRW_VFCOMPONENT_STORE_PID 7 - - - -/* Execution Unit (EU) defines - */ - -#define BRW_ALIGN_1 0 -#define BRW_ALIGN_16 1 - -#define BRW_ADDRESS_DIRECT 0 -#define BRW_ADDRESS_REGISTER_INDIRECT_REGISTER 1 - -#define BRW_CHANNEL_X 0 -#define BRW_CHANNEL_Y 1 -#define BRW_CHANNEL_Z 2 -#define BRW_CHANNEL_W 3 - -#define BRW_COMPRESSION_NONE 0 -#define BRW_COMPRESSION_2NDHALF 1 -#define BRW_COMPRESSION_COMPRESSED 2 - -#define BRW_CONDITIONAL_NONE 0 -#define BRW_CONDITIONAL_Z 1 -#define BRW_CONDITIONAL_NZ 2 -#define BRW_CONDITIONAL_EQ 1 /* Z */ -#define BRW_CONDITIONAL_NEQ 2 /* NZ */ -#define BRW_CONDITIONAL_G 3 -#define BRW_CONDITIONAL_GE 4 -#define BRW_CONDITIONAL_L 5 -#define BRW_CONDITIONAL_LE 6 -#define BRW_CONDITIONAL_C 7 -#define BRW_CONDITIONAL_O 8 - -#define BRW_DEBUG_NONE 0 -#define BRW_DEBUG_BREAKPOINT 1 - -#define BRW_DEPENDENCY_NORMAL 0 -#define BRW_DEPENDENCY_NOTCLEARED 1 -#define BRW_DEPENDENCY_NOTCHECKED 2 -#define BRW_DEPENDENCY_DISABLE 3 - -#define BRW_EXECUTE_1 0 -#define BRW_EXECUTE_2 1 -#define BRW_EXECUTE_4 2 -#define BRW_EXECUTE_8 3 -#define BRW_EXECUTE_16 4 -#define BRW_EXECUTE_32 5 - -#define BRW_HORIZONTAL_STRIDE_0 0 -#define BRW_HORIZONTAL_STRIDE_1 1 -#define BRW_HORIZONTAL_STRIDE_2 2 -#define BRW_HORIZONTAL_STRIDE_4 3 - -#define BRW_INSTRUCTION_NORMAL 0 -#define BRW_INSTRUCTION_SATURATE 1 - -#define BRW_MASK_ENABLE 0 -#define BRW_MASK_DISABLE 1 - -#define BRW_OPCODE_MOV 1 -#define BRW_OPCODE_SEL 2 -#define BRW_OPCODE_NOT 4 -#define BRW_OPCODE_AND 5 -#define BRW_OPCODE_OR 6 -#define BRW_OPCODE_XOR 7 -#define BRW_OPCODE_SHR 8 -#define BRW_OPCODE_SHL 9 -#define BRW_OPCODE_RSR 10 -#define BRW_OPCODE_RSL 11 -#define BRW_OPCODE_ASR 12 -#define BRW_OPCODE_CMP 16 -#define BRW_OPCODE_JMPI 32 -#define BRW_OPCODE_IF 34 -#define BRW_OPCODE_IFF 35 -#define BRW_OPCODE_ELSE 36 -#define BRW_OPCODE_ENDIF 37 -#define BRW_OPCODE_DO 38 -#define BRW_OPCODE_WHILE 39 -#define BRW_OPCODE_BREAK 40 -#define BRW_OPCODE_CONTINUE 41 -#define BRW_OPCODE_HALT 42 -#define BRW_OPCODE_MSAVE 44 -#define BRW_OPCODE_MRESTORE 45 -#define BRW_OPCODE_PUSH 46 -#define BRW_OPCODE_POP 47 -#define BRW_OPCODE_WAIT 48 -#define BRW_OPCODE_SEND 49 -#define BRW_OPCODE_ADD 64 -#define BRW_OPCODE_MUL 65 -#define BRW_OPCODE_AVG 66 -#define BRW_OPCODE_FRC 67 -#define BRW_OPCODE_RNDU 68 -#define BRW_OPCODE_RNDD 69 -#define BRW_OPCODE_RNDE 70 -#define BRW_OPCODE_RNDZ 71 -#define BRW_OPCODE_MAC 72 -#define BRW_OPCODE_MACH 73 -#define BRW_OPCODE_LZD 74 -#define BRW_OPCODE_SAD2 80 -#define BRW_OPCODE_SADA2 81 -#define BRW_OPCODE_DP4 84 -#define BRW_OPCODE_DPH 85 -#define BRW_OPCODE_DP3 86 -#define BRW_OPCODE_DP2 87 -#define BRW_OPCODE_DPA2 88 -#define BRW_OPCODE_LINE 89 -#define BRW_OPCODE_NOP 126 - -#define BRW_PREDICATE_NONE 0 -#define BRW_PREDICATE_NORMAL 1 -#define BRW_PREDICATE_ALIGN1_ANYV 2 -#define BRW_PREDICATE_ALIGN1_ALLV 3 -#define BRW_PREDICATE_ALIGN1_ANY2H 4 -#define BRW_PREDICATE_ALIGN1_ALL2H 5 -#define BRW_PREDICATE_ALIGN1_ANY4H 6 -#define BRW_PREDICATE_ALIGN1_ALL4H 7 -#define BRW_PREDICATE_ALIGN1_ANY8H 8 -#define BRW_PREDICATE_ALIGN1_ALL8H 9 -#define BRW_PREDICATE_ALIGN1_ANY16H 10 -#define BRW_PREDICATE_ALIGN1_ALL16H 11 -#define BRW_PREDICATE_ALIGN16_REPLICATE_X 2 -#define BRW_PREDICATE_ALIGN16_REPLICATE_Y 3 -#define BRW_PREDICATE_ALIGN16_REPLICATE_Z 4 -#define BRW_PREDICATE_ALIGN16_REPLICATE_W 5 -#define BRW_PREDICATE_ALIGN16_ANY4H 6 -#define BRW_PREDICATE_ALIGN16_ALL4H 7 - -#define BRW_ARCHITECTURE_REGISTER_FILE 0 -#define BRW_GENERAL_REGISTER_FILE 1 -#define BRW_MESSAGE_REGISTER_FILE 2 -#define BRW_IMMEDIATE_VALUE 3 - -#define BRW_REGISTER_TYPE_UD 0 -#define BRW_REGISTER_TYPE_D 1 -#define BRW_REGISTER_TYPE_UW 2 -#define BRW_REGISTER_TYPE_W 3 -#define BRW_REGISTER_TYPE_UB 4 -#define BRW_REGISTER_TYPE_B 5 -#define BRW_REGISTER_TYPE_VF 5 /* packed float vector, immediates only? */ -#define BRW_REGISTER_TYPE_HF 6 -#define BRW_REGISTER_TYPE_V 6 /* packed int vector, immediates only, uword dest only */ -#define BRW_REGISTER_TYPE_F 7 - -#define BRW_ARF_NULL 0x00 -#define BRW_ARF_ADDRESS 0x10 -#define BRW_ARF_ACCUMULATOR 0x20 -#define BRW_ARF_FLAG 0x30 -#define BRW_ARF_MASK 0x40 -#define BRW_ARF_MASK_STACK 0x50 -#define BRW_ARF_MASK_STACK_DEPTH 0x60 -#define BRW_ARF_STATE 0x70 -#define BRW_ARF_CONTROL 0x80 -#define BRW_ARF_NOTIFICATION_COUNT 0x90 -#define BRW_ARF_IP 0xA0 - -#define BRW_AMASK 0 -#define BRW_IMASK 1 -#define BRW_LMASK 2 -#define BRW_CMASK 3 - - - -#define BRW_THREAD_NORMAL 0 -#define BRW_THREAD_ATOMIC 1 -#define BRW_THREAD_SWITCH 2 - -#define BRW_VERTICAL_STRIDE_0 0 -#define BRW_VERTICAL_STRIDE_1 1 -#define BRW_VERTICAL_STRIDE_2 2 -#define BRW_VERTICAL_STRIDE_4 3 -#define BRW_VERTICAL_STRIDE_8 4 -#define BRW_VERTICAL_STRIDE_16 5 -#define BRW_VERTICAL_STRIDE_32 6 -#define BRW_VERTICAL_STRIDE_64 7 -#define BRW_VERTICAL_STRIDE_128 8 -#define BRW_VERTICAL_STRIDE_256 9 -#define BRW_VERTICAL_STRIDE_ONE_DIMENSIONAL 0xF - -#define BRW_WIDTH_1 0 -#define BRW_WIDTH_2 1 -#define BRW_WIDTH_4 2 -#define BRW_WIDTH_8 3 -#define BRW_WIDTH_16 4 - -#define BRW_STATELESS_BUFFER_BOUNDARY_1K 0 -#define BRW_STATELESS_BUFFER_BOUNDARY_2K 1 -#define BRW_STATELESS_BUFFER_BOUNDARY_4K 2 -#define BRW_STATELESS_BUFFER_BOUNDARY_8K 3 -#define BRW_STATELESS_BUFFER_BOUNDARY_16K 4 -#define BRW_STATELESS_BUFFER_BOUNDARY_32K 5 -#define BRW_STATELESS_BUFFER_BOUNDARY_64K 6 -#define BRW_STATELESS_BUFFER_BOUNDARY_128K 7 -#define BRW_STATELESS_BUFFER_BOUNDARY_256K 8 -#define BRW_STATELESS_BUFFER_BOUNDARY_512K 9 -#define BRW_STATELESS_BUFFER_BOUNDARY_1M 10 -#define BRW_STATELESS_BUFFER_BOUNDARY_2M 11 - -#define BRW_POLYGON_FACING_FRONT 0 -#define BRW_POLYGON_FACING_BACK 1 - -#define BRW_MESSAGE_TARGET_NULL 0 -#define BRW_MESSAGE_TARGET_MATH 1 -#define BRW_MESSAGE_TARGET_SAMPLER 2 -#define BRW_MESSAGE_TARGET_GATEWAY 3 -#define BRW_MESSAGE_TARGET_DATAPORT_READ 4 -#define BRW_MESSAGE_TARGET_DATAPORT_WRITE 5 -#define BRW_MESSAGE_TARGET_URB 6 -#define BRW_MESSAGE_TARGET_THREAD_SPAWNER 7 - -#define BRW_SAMPLER_RETURN_FORMAT_FLOAT32 0 -#define BRW_SAMPLER_RETURN_FORMAT_UINT32 2 -#define BRW_SAMPLER_RETURN_FORMAT_SINT32 3 - -#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE 0 -#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE 0 -#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS 0 -#define BRW_SAMPLER_MESSAGE_SIMD8_KILLPIX 1 -#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_LOD 1 -#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_LOD 1 -#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_GRADIENTS 2 -#define BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE_GRADIENTS 2 -#define BRW_SAMPLER_MESSAGE_SIMD4X2_SAMPLE_COMPARE 0 -#define BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_COMPARE 2 -#define BRW_SAMPLER_MESSAGE_SIMD4X2_RESINFO 2 -#define BRW_SAMPLER_MESSAGE_SIMD8_RESINFO 2 -#define BRW_SAMPLER_MESSAGE_SIMD16_RESINFO 2 -#define BRW_SAMPLER_MESSAGE_SIMD4X2_LD 3 -#define BRW_SAMPLER_MESSAGE_SIMD8_LD 3 -#define BRW_SAMPLER_MESSAGE_SIMD16_LD 3 - -#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDLOW 0 -#define BRW_DATAPORT_OWORD_BLOCK_1_OWORDHIGH 1 -#define BRW_DATAPORT_OWORD_BLOCK_2_OWORDS 2 -#define BRW_DATAPORT_OWORD_BLOCK_4_OWORDS 3 -#define BRW_DATAPORT_OWORD_BLOCK_8_OWORDS 4 - -#define BRW_DATAPORT_OWORD_DUAL_BLOCK_1OWORD 0 -#define BRW_DATAPORT_OWORD_DUAL_BLOCK_4OWORDS 2 - -#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_8DWORDS 2 -#define BRW_DATAPORT_DWORD_SCATTERED_BLOCK_16DWORDS 3 - -#define BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ 0 -#define BRW_DATAPORT_READ_MESSAGE_OWORD_DUAL_BLOCK_READ 1 -#define BRW_DATAPORT_READ_MESSAGE_DWORD_BLOCK_READ 2 -#define BRW_DATAPORT_READ_MESSAGE_DWORD_SCATTERED_READ 3 - -#define BRW_DATAPORT_READ_TARGET_DATA_CACHE 0 -#define BRW_DATAPORT_READ_TARGET_RENDER_CACHE 1 -#define BRW_DATAPORT_READ_TARGET_SAMPLER_CACHE 2 - -#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE 0 -#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE_REPLICATED 1 -#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN01 2 -#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_DUAL_SOURCE_SUBSPAN23 3 -#define BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD8_SINGLE_SOURCE_SUBSPAN01 4 - -#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE 0 -#define BRW_DATAPORT_WRITE_MESSAGE_OWORD_DUAL_BLOCK_WRITE 1 -#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_BLOCK_WRITE 2 -#define BRW_DATAPORT_WRITE_MESSAGE_DWORD_SCATTERED_WRITE 3 -#define BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE 4 -#define BRW_DATAPORT_WRITE_MESSAGE_STREAMED_VERTEX_BUFFER_WRITE 5 -#define BRW_DATAPORT_WRITE_MESSAGE_FLUSH_RENDER_CACHE 7 - -#define BRW_MATH_FUNCTION_INV 1 -#define BRW_MATH_FUNCTION_LOG 2 -#define BRW_MATH_FUNCTION_EXP 3 -#define BRW_MATH_FUNCTION_SQRT 4 -#define BRW_MATH_FUNCTION_RSQ 5 -#define BRW_MATH_FUNCTION_SIN 6 /* was 7 */ -#define BRW_MATH_FUNCTION_COS 7 /* was 8 */ -#define BRW_MATH_FUNCTION_SINCOS 8 /* was 6 */ -#define BRW_MATH_FUNCTION_TAN 9 -#define BRW_MATH_FUNCTION_POW 10 -#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT_AND_REMAINDER 11 -#define BRW_MATH_FUNCTION_INT_DIV_QUOTIENT 12 -#define BRW_MATH_FUNCTION_INT_DIV_REMAINDER 13 - -#define BRW_MATH_INTEGER_UNSIGNED 0 -#define BRW_MATH_INTEGER_SIGNED 1 - -#define BRW_MATH_PRECISION_FULL 0 -#define BRW_MATH_PRECISION_PARTIAL 1 - -#define BRW_MATH_SATURATE_NONE 0 -#define BRW_MATH_SATURATE_SATURATE 1 - -#define BRW_MATH_DATA_VECTOR 0 -#define BRW_MATH_DATA_SCALAR 1 - -#define BRW_URB_OPCODE_WRITE 0 - -#define BRW_URB_SWIZZLE_NONE 0 -#define BRW_URB_SWIZZLE_INTERLEAVE 1 -#define BRW_URB_SWIZZLE_TRANSPOSE 2 - -#define BRW_SCRATCH_SPACE_SIZE_1K 0 -#define BRW_SCRATCH_SPACE_SIZE_2K 1 -#define BRW_SCRATCH_SPACE_SIZE_4K 2 -#define BRW_SCRATCH_SPACE_SIZE_8K 3 -#define BRW_SCRATCH_SPACE_SIZE_16K 4 -#define BRW_SCRATCH_SPACE_SIZE_32K 5 -#define BRW_SCRATCH_SPACE_SIZE_64K 6 -#define BRW_SCRATCH_SPACE_SIZE_128K 7 -#define BRW_SCRATCH_SPACE_SIZE_256K 8 -#define BRW_SCRATCH_SPACE_SIZE_512K 9 -#define BRW_SCRATCH_SPACE_SIZE_1M 10 -#define BRW_SCRATCH_SPACE_SIZE_2M 11 - - - - -#define CMD_URB_FENCE 0x6000 -#define CMD_CONST_BUFFER_STATE 0x6001 -#define CMD_CONST_BUFFER 0x6002 - -#define CMD_STATE_BASE_ADDRESS 0x6101 -#define CMD_STATE_INSN_POINTER 0x6102 -#define CMD_PIPELINE_SELECT 0x6104 - -#define CMD_PIPELINED_STATE_POINTERS 0x7800 -#define CMD_BINDING_TABLE_PTRS 0x7801 -#define CMD_VERTEX_BUFFER 0x7808 -#define CMD_VERTEX_ELEMENT 0x7809 -#define CMD_INDEX_BUFFER 0x780a -#define CMD_VF_STATISTICS 0x780b - -#define CMD_DRAW_RECT 0x7900 -#define CMD_BLEND_CONSTANT_COLOR 0x7901 -#define CMD_CHROMA_KEY 0x7904 -#define CMD_DEPTH_BUFFER 0x7905 -#define CMD_POLY_STIPPLE_OFFSET 0x7906 -#define CMD_POLY_STIPPLE_PATTERN 0x7907 -#define CMD_LINE_STIPPLE_PATTERN 0x7908 -#define CMD_GLOBAL_DEPTH_OFFSET_CLAMP 0x7909 - -#define CMD_PIPE_CONTROL 0x7a00 - -#define CMD_3D_PRIM 0x7b00 - -#define CMD_MI_FLUSH 0x0200 - - -/* Various values from the R0 vertex header: - */ -#define R02_PRIM_END 0x1 -#define R02_PRIM_START 0x2 - - - -#endif diff --git a/src/mesa/pipe/i965simple/brw_draw.c b/src/mesa/pipe/i965simple/brw_draw.c deleted file mode 100644 index 7598e3dc8a..0000000000 --- a/src/mesa/pipe/i965simple/brw_draw.c +++ /dev/null @@ -1,239 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include - -#include "brw_batch.h" -#include "brw_draw.h" -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_state.h" - -#include "pipe/p_context.h" -#include "pipe/p_winsys.h" - -static unsigned hw_prim[PIPE_PRIM_POLYGON+1] = { - _3DPRIM_POINTLIST, - _3DPRIM_LINELIST, - _3DPRIM_LINELOOP, - _3DPRIM_LINESTRIP, - _3DPRIM_TRILIST, - _3DPRIM_TRISTRIP, - _3DPRIM_TRIFAN, - _3DPRIM_QUADLIST, - _3DPRIM_QUADSTRIP, - _3DPRIM_POLYGON -}; - - -static const int reduced_prim[PIPE_PRIM_POLYGON+1] = { - PIPE_PRIM_POINTS, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES -}; - - -/* When the primitive changes, set a state bit and re-validate. Not - * the nicest and would rather deal with this by having all the - * programs be immune to the active primitive (ie. cope with all - * possibilities). That may not be realistic however. - */ -static void brw_set_prim(struct brw_context *brw, int prim) -{ - PRINT("PRIM: %d\n", prim); - - /* Slight optimization to avoid the GS program when not needed: - */ - if (prim == PIPE_PRIM_QUAD_STRIP && - brw->attribs.Raster->flatshade && - brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_FILL && - brw->attribs.Raster->fill_ccw == PIPE_POLYGON_MODE_FILL) - prim = PIPE_PRIM_TRIANGLE_STRIP; - - if (prim != brw->primitive) { - brw->primitive = prim; - brw->state.dirty.brw |= BRW_NEW_PRIMITIVE; - - if (reduced_prim[prim] != brw->reduced_primitive) { - brw->reduced_primitive = reduced_prim[prim]; - brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE; - } - - brw_validate_state(brw); - } - -} - - -static unsigned trim(int prim, unsigned length) -{ - if (prim == PIPE_PRIM_QUAD_STRIP) - return length > 3 ? (length - length % 2) : 0; - else if (prim == PIPE_PRIM_QUADS) - return length - length % 4; - else - return length; -} - - - -static boolean brw_emit_prim( struct brw_context *brw, - boolean indexed, - unsigned start, - unsigned count ) - -{ - struct brw_3d_primitive prim_packet; - - if (BRW_DEBUG & DEBUG_PRIMS) - PRINT("PRIM: %d %d %d\n", brw->primitive, start, count); - - prim_packet.header.opcode = CMD_3D_PRIM; - prim_packet.header.length = sizeof(prim_packet)/4 - 2; - prim_packet.header.pad = 0; - prim_packet.header.topology = hw_prim[brw->primitive]; - prim_packet.header.indexed = indexed; - - prim_packet.verts_per_instance = trim(brw->primitive, count); - prim_packet.start_vert_location = start; - prim_packet.instance_count = 1; - prim_packet.start_instance_location = 0; - prim_packet.base_vert_location = 0; - - if (prim_packet.verts_per_instance == 0) - return TRUE; - - return brw_batchbuffer_data( brw->winsys, - &prim_packet, - sizeof(prim_packet) ); -} - - -/* May fail if out of video memory for texture or vbo upload, or on - * fallback conditions. - */ -static boolean brw_try_draw_elements( struct pipe_context *pipe, - struct pipe_buffer *index_buffer, - unsigned index_size, - unsigned mode, - unsigned start, - unsigned count ) -{ - struct brw_context *brw = brw_context(pipe); - - /* Set the first primitive ahead of validate_state: - */ - brw_set_prim(brw, mode); - - /* Upload index, vertex data: - */ - if (index_buffer && - !brw_upload_indices( brw, index_buffer, index_size, start, count )) - return FALSE; - - if (!brw_upload_vertex_buffers(brw)) - return FALSE; - - if (!brw_upload_vertex_elements( brw )) - return FALSE; - - /* XXX: Need to separate validate and upload of state. - */ - if (brw->state.dirty.brw) - brw_validate_state( brw ); - - if (!brw_emit_prim(brw, index_buffer != NULL, - start, count)) - return FALSE; - - return TRUE; -} - - - -static boolean brw_draw_elements( struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, - unsigned start, - unsigned count ) -{ - if (!brw_try_draw_elements( pipe, - indexBuffer, - indexSize, - mode, start, count )) - { - /* flush ? */ - - if (!brw_try_draw_elements( pipe, - indexBuffer, - indexSize, - mode, start, - count )) { - assert(0); - return FALSE; - } - } - - return TRUE; -} - - - -static boolean brw_draw_arrays( struct pipe_context *pipe, - unsigned mode, - unsigned start, - unsigned count ) -{ - if (!brw_try_draw_elements( pipe, NULL, 0, mode, start, count )) { - /* flush ? */ - - if (!brw_try_draw_elements( pipe, NULL, 0, mode, start, count )) { - assert(0); - return FALSE; - } - } - - return TRUE; -} - - - -void brw_init_draw_functions( struct brw_context *brw ) -{ - brw->pipe.draw_arrays = brw_draw_arrays; - brw->pipe.draw_elements = brw_draw_elements; -} - - diff --git a/src/mesa/pipe/i965simple/brw_draw.h b/src/mesa/pipe/i965simple/brw_draw.h deleted file mode 100644 index 62fe0d5d0e..0000000000 --- a/src/mesa/pipe/i965simple/brw_draw.h +++ /dev/null @@ -1,55 +0,0 @@ - /************************************************************************** - * - * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef BRW_DRAW_H -#define BRW_DRAW_H - -#include "pipe/p_context.h" - -struct brw_context; - - - -void brw_init_draw_functions( struct brw_context *brw ); - - -boolean brw_upload_vertices( struct brw_context *brw, - unsigned min_index, - unsigned max_index ); - -boolean brw_upload_indices(struct brw_context *brw, - const struct pipe_buffer *index_buffer, - int ib_size, int start, int count); - -boolean brw_upload_vertex_buffers( struct brw_context *brw ); -boolean brw_upload_vertex_elements( struct brw_context *brw ); - -unsigned brw_translate_surface_format( unsigned id ); - - - -#endif diff --git a/src/mesa/pipe/i965simple/brw_draw_upload.c b/src/mesa/pipe/i965simple/brw_draw_upload.c deleted file mode 100644 index aa85d93866..0000000000 --- a/src/mesa/pipe/i965simple/brw_draw_upload.c +++ /dev/null @@ -1,299 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include - -#include "brw_batch.h" -#include "brw_draw.h" -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_state.h" - -struct brw_array_state { - union header_union header; - - struct { - union { - struct { - unsigned pitch:11; - unsigned pad:15; - unsigned access_type:1; - unsigned vb_index:5; - } bits; - unsigned dword; - } vb0; - - struct pipe_buffer *buffer; - unsigned offset; - - unsigned max_index; - unsigned instance_data_step_rate; - - } vb[BRW_VBP_MAX]; -}; - - - -unsigned brw_translate_surface_format( unsigned id ) -{ - switch (id) { - case PIPE_FORMAT_R64_FLOAT: - return BRW_SURFACEFORMAT_R64_FLOAT; - case PIPE_FORMAT_R64G64_FLOAT: - return BRW_SURFACEFORMAT_R64G64_FLOAT; - case PIPE_FORMAT_R64G64B64_FLOAT: - return BRW_SURFACEFORMAT_R64G64B64_FLOAT; - case PIPE_FORMAT_R64G64B64A64_FLOAT: - return BRW_SURFACEFORMAT_R64G64B64A64_FLOAT; - - case PIPE_FORMAT_R32_FLOAT: - return BRW_SURFACEFORMAT_R32_FLOAT; - case PIPE_FORMAT_R32G32_FLOAT: - return BRW_SURFACEFORMAT_R32G32_FLOAT; - case PIPE_FORMAT_R32G32B32_FLOAT: - return BRW_SURFACEFORMAT_R32G32B32_FLOAT; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return BRW_SURFACEFORMAT_R32G32B32A32_FLOAT; - - case PIPE_FORMAT_R32_UNORM: - return BRW_SURFACEFORMAT_R32_UNORM; - case PIPE_FORMAT_R32G32_UNORM: - return BRW_SURFACEFORMAT_R32G32_UNORM; - case PIPE_FORMAT_R32G32B32_UNORM: - return BRW_SURFACEFORMAT_R32G32B32_UNORM; - case PIPE_FORMAT_R32G32B32A32_UNORM: - return BRW_SURFACEFORMAT_R32G32B32A32_UNORM; - - case PIPE_FORMAT_R32_USCALED: - return BRW_SURFACEFORMAT_R32_USCALED; - case PIPE_FORMAT_R32G32_USCALED: - return BRW_SURFACEFORMAT_R32G32_USCALED; - case PIPE_FORMAT_R32G32B32_USCALED: - return BRW_SURFACEFORMAT_R32G32B32_USCALED; - case PIPE_FORMAT_R32G32B32A32_USCALED: - return BRW_SURFACEFORMAT_R32G32B32A32_USCALED; - - case PIPE_FORMAT_R32_SNORM: - return BRW_SURFACEFORMAT_R32_SNORM; - case PIPE_FORMAT_R32G32_SNORM: - return BRW_SURFACEFORMAT_R32G32_SNORM; - case PIPE_FORMAT_R32G32B32_SNORM: - return BRW_SURFACEFORMAT_R32G32B32_SNORM; - case PIPE_FORMAT_R32G32B32A32_SNORM: - return BRW_SURFACEFORMAT_R32G32B32A32_SNORM; - - case PIPE_FORMAT_R32_SSCALED: - return BRW_SURFACEFORMAT_R32_SSCALED; - case PIPE_FORMAT_R32G32_SSCALED: - return BRW_SURFACEFORMAT_R32G32_SSCALED; - case PIPE_FORMAT_R32G32B32_SSCALED: - return BRW_SURFACEFORMAT_R32G32B32_SSCALED; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - return BRW_SURFACEFORMAT_R32G32B32A32_SSCALED; - - case PIPE_FORMAT_R16_UNORM: - return BRW_SURFACEFORMAT_R16_UNORM; - case PIPE_FORMAT_R16G16_UNORM: - return BRW_SURFACEFORMAT_R16G16_UNORM; - case PIPE_FORMAT_R16G16B16_UNORM: - return BRW_SURFACEFORMAT_R16G16B16_UNORM; - case PIPE_FORMAT_R16G16B16A16_UNORM: - return BRW_SURFACEFORMAT_R16G16B16A16_UNORM; - - case PIPE_FORMAT_R16_USCALED: - return BRW_SURFACEFORMAT_R16_USCALED; - case PIPE_FORMAT_R16G16_USCALED: - return BRW_SURFACEFORMAT_R16G16_USCALED; - case PIPE_FORMAT_R16G16B16_USCALED: - return BRW_SURFACEFORMAT_R16G16B16_USCALED; - case PIPE_FORMAT_R16G16B16A16_USCALED: - return BRW_SURFACEFORMAT_R16G16B16A16_USCALED; - - case PIPE_FORMAT_R16_SNORM: - return BRW_SURFACEFORMAT_R16_SNORM; - case PIPE_FORMAT_R16G16_SNORM: - return BRW_SURFACEFORMAT_R16G16_SNORM; - case PIPE_FORMAT_R16G16B16_SNORM: - return BRW_SURFACEFORMAT_R16G16B16_SNORM; - case PIPE_FORMAT_R16G16B16A16_SNORM: - return BRW_SURFACEFORMAT_R16G16B16A16_SNORM; - - case PIPE_FORMAT_R16_SSCALED: - return BRW_SURFACEFORMAT_R16_SSCALED; - case PIPE_FORMAT_R16G16_SSCALED: - return BRW_SURFACEFORMAT_R16G16_SSCALED; - case PIPE_FORMAT_R16G16B16_SSCALED: - return BRW_SURFACEFORMAT_R16G16B16_SSCALED; - case PIPE_FORMAT_R16G16B16A16_SSCALED: - return BRW_SURFACEFORMAT_R16G16B16A16_SSCALED; - - case PIPE_FORMAT_R8_UNORM: - return BRW_SURFACEFORMAT_R8_UNORM; - case PIPE_FORMAT_R8G8_UNORM: - return BRW_SURFACEFORMAT_R8G8_UNORM; - case PIPE_FORMAT_R8G8B8_UNORM: - return BRW_SURFACEFORMAT_R8G8B8_UNORM; - case PIPE_FORMAT_R8G8B8A8_UNORM: - return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; - - case PIPE_FORMAT_R8_USCALED: - return BRW_SURFACEFORMAT_R8_USCALED; - case PIPE_FORMAT_R8G8_USCALED: - return BRW_SURFACEFORMAT_R8G8_USCALED; - case PIPE_FORMAT_R8G8B8_USCALED: - return BRW_SURFACEFORMAT_R8G8B8_USCALED; - case PIPE_FORMAT_R8G8B8A8_USCALED: - return BRW_SURFACEFORMAT_R8G8B8A8_USCALED; - - case PIPE_FORMAT_R8_SNORM: - return BRW_SURFACEFORMAT_R8_SNORM; - case PIPE_FORMAT_R8G8_SNORM: - return BRW_SURFACEFORMAT_R8G8_SNORM; - case PIPE_FORMAT_R8G8B8_SNORM: - return BRW_SURFACEFORMAT_R8G8B8_SNORM; - case PIPE_FORMAT_R8G8B8A8_SNORM: - return BRW_SURFACEFORMAT_R8G8B8A8_SNORM; - - case PIPE_FORMAT_R8_SSCALED: - return BRW_SURFACEFORMAT_R8_SSCALED; - case PIPE_FORMAT_R8G8_SSCALED: - return BRW_SURFACEFORMAT_R8G8_SSCALED; - case PIPE_FORMAT_R8G8B8_SSCALED: - return BRW_SURFACEFORMAT_R8G8B8_SSCALED; - case PIPE_FORMAT_R8G8B8A8_SSCALED: - return BRW_SURFACEFORMAT_R8G8B8A8_SSCALED; - - default: - assert(0); - return 0; - } -} - -static unsigned get_index_type(int type) -{ - switch (type) { - case 1: return BRW_INDEX_BYTE; - case 2: return BRW_INDEX_WORD; - case 4: return BRW_INDEX_DWORD; - default: assert(0); return 0; - } -} - - -boolean brw_upload_vertex_buffers( struct brw_context *brw ) -{ - struct brw_array_state vbp; - unsigned nr_enabled = 0; - unsigned i; - - memset(&vbp, 0, sizeof(vbp)); - - /* This is a hardware limit: - */ - - for (i = 0; i < BRW_VEP_MAX; i++) - { - if (brw->vb.vbo_array[i] == NULL) { - nr_enabled = i; - break; - } - - vbp.vb[i].vb0.bits.pitch = brw->vb.vbo_array[i]->pitch; - vbp.vb[i].vb0.bits.pad = 0; - vbp.vb[i].vb0.bits.access_type = BRW_VERTEXBUFFER_ACCESS_VERTEXDATA; - vbp.vb[i].vb0.bits.vb_index = i; - vbp.vb[i].offset = brw->vb.vbo_array[i]->buffer_offset; - vbp.vb[i].buffer = brw->vb.vbo_array[i]->buffer; - vbp.vb[i].max_index = brw->vb.vbo_array[i]->max_index; - } - - - vbp.header.bits.length = (1 + nr_enabled * 4) - 2; - vbp.header.bits.opcode = CMD_VERTEX_BUFFER; - - BEGIN_BATCH(vbp.header.bits.length+2, 0); - OUT_BATCH( vbp.header.dword ); - - for (i = 0; i < nr_enabled; i++) { - OUT_BATCH( vbp.vb[i].vb0.dword ); - OUT_RELOC( vbp.vb[i].buffer, PIPE_BUFFER_USAGE_GPU_READ, - vbp.vb[i].offset); - OUT_BATCH( vbp.vb[i].max_index ); - OUT_BATCH( vbp.vb[i].instance_data_step_rate ); - } - ADVANCE_BATCH(); - return TRUE; -} - - - -boolean brw_upload_vertex_elements( struct brw_context *brw ) -{ - struct brw_vertex_element_packet vep; - - unsigned i; - unsigned nr_enabled = brw->attribs.VertexProgram->program.num_inputs; - - memset(&vep, 0, sizeof(vep)); - - for (i = 0; i < nr_enabled; i++) - vep.ve[i] = brw->vb.inputs[i]; - - - vep.header.length = (1 + nr_enabled * sizeof(vep.ve[0])/4) - 2; - vep.header.opcode = CMD_VERTEX_ELEMENT; - brw_cached_batch_struct(brw, &vep, 4 + nr_enabled * sizeof(vep.ve[0])); - - return TRUE; -} - -boolean brw_upload_indices( struct brw_context *brw, - const struct pipe_buffer *index_buffer, - int ib_size, int start, int count) -{ - /* Emit the indexbuffer packet: - */ - { - struct brw_indexbuffer ib; - - memset(&ib, 0, sizeof(ib)); - - ib.header.bits.opcode = CMD_INDEX_BUFFER; - ib.header.bits.length = sizeof(ib)/4 - 2; - ib.header.bits.index_format = get_index_type(ib_size); - ib.header.bits.cut_index_enable = 0; - - - BEGIN_BATCH(4, 0); - OUT_BATCH( ib.header.dword ); - OUT_RELOC( index_buffer, PIPE_BUFFER_USAGE_GPU_READ, start); - OUT_RELOC( index_buffer, PIPE_BUFFER_USAGE_GPU_READ, start + count); - OUT_BATCH( 0 ); - ADVANCE_BATCH(); - } - return TRUE; -} diff --git a/src/mesa/pipe/i965simple/brw_eu.c b/src/mesa/pipe/i965simple/brw_eu.c deleted file mode 100644 index e2002d1821..0000000000 --- a/src/mesa/pipe/i965simple/brw_eu.c +++ /dev/null @@ -1,130 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_eu.h" - - - -/* How does predicate control work when execution_size != 8? Do I - * need to test/set for 0xffff when execution_size is 16? - */ -void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value ) -{ - p->current->header.predicate_control = BRW_PREDICATE_NONE; - - if (value != 0xff) { - if (value != p->flag_value) { - brw_push_insn_state(p); - brw_MOV(p, brw_flag_reg(), brw_imm_uw(value)); - p->flag_value = value; - brw_pop_insn_state(p); - } - - p->current->header.predicate_control = BRW_PREDICATE_NORMAL; - } -} - -void brw_set_predicate_control( struct brw_compile *p, unsigned pc ) -{ - p->current->header.predicate_control = pc; -} - -void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional ) -{ - p->current->header.destreg__conditonalmod = conditional; -} - -void brw_set_access_mode( struct brw_compile *p, unsigned access_mode ) -{ - p->current->header.access_mode = access_mode; -} - -void brw_set_compression_control( struct brw_compile *p, boolean compression_control ) -{ - p->current->header.compression_control = compression_control; -} - -void brw_set_mask_control( struct brw_compile *p, unsigned value ) -{ - p->current->header.mask_control = value; -} - -void brw_set_saturate( struct brw_compile *p, unsigned value ) -{ - p->current->header.saturate = value; -} - -void brw_push_insn_state( struct brw_compile *p ) -{ - assert(p->current != &p->stack[BRW_EU_MAX_INSN_STACK-1]); - memcpy(p->current+1, p->current, sizeof(struct brw_instruction)); - p->current++; -} - -void brw_pop_insn_state( struct brw_compile *p ) -{ - assert(p->current != p->stack); - p->current--; -} - - -/*********************************************************************** - */ -void brw_init_compile( struct brw_compile *p ) -{ - p->nr_insn = 0; - p->current = p->stack; - memset(p->current, 0, sizeof(p->current[0])); - - /* Some defaults? - */ - brw_set_mask_control(p, BRW_MASK_ENABLE); /* what does this do? */ - brw_set_saturate(p, 0); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - brw_set_predicate_control_flag_value(p, 0xff); -} - - -const unsigned *brw_get_program( struct brw_compile *p, - unsigned *sz ) -{ - unsigned i; - - for (i = 0; i < 8; i++) - brw_NOP(p); - - *sz = p->nr_insn * sizeof(struct brw_instruction); - return (const unsigned *)p->store; -} - diff --git a/src/mesa/pipe/i965simple/brw_eu.h b/src/mesa/pipe/i965simple/brw_eu.h deleted file mode 100644 index 23151ae9ed..0000000000 --- a/src/mesa/pipe/i965simple/brw_eu.h +++ /dev/null @@ -1,888 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRW_EU_H -#define BRW_EU_H - -#include "brw_structs.h" -#include "brw_defines.h" - -#include "pipe/p_compiler.h" -#include "pipe/p_shader_tokens.h" - -#define BRW_SWIZZLE4(a,b,c,d) (((a)<<0) | ((b)<<2) | ((c)<<4) | ((d)<<6)) -#define BRW_GET_SWZ(swz, idx) (((swz) >> ((idx)*2)) & 0x3) - -#define BRW_SWIZZLE_NOOP BRW_SWIZZLE4(0,1,2,3) -#define BRW_SWIZZLE_XYZW BRW_SWIZZLE4(0,1,2,3) -#define BRW_SWIZZLE_XXXX BRW_SWIZZLE4(0,0,0,0) -#define BRW_SWIZZLE_XYXY BRW_SWIZZLE4(0,1,0,1) - - -#define REG_SIZE (8*4) - - -/* These aren't hardware structs, just something useful for us to pass around: - * - * Align1 operation has a lot of control over input ranges. Used in - * WM programs to implement shaders decomposed into "channel serial" - * or "structure of array" form: - */ -struct brw_reg -{ - unsigned type:4; - unsigned file:2; - unsigned nr:8; - unsigned subnr:5; /* :1 in align16 */ - unsigned negate:1; /* source only */ - unsigned abs:1; /* source only */ - unsigned vstride:4; /* source only */ - unsigned width:3; /* src only, align1 only */ - unsigned hstride:2; /* src only, align1 only */ - unsigned address_mode:1; /* relative addressing, hopefully! */ - unsigned pad0:1; - - union { - struct { - unsigned swizzle:8; /* src only, align16 only */ - unsigned writemask:4; /* dest only, align16 only */ - int indirect_offset:10; /* relative addressing offset */ - unsigned pad1:10; /* two dwords total */ - } bits; - - float f; - int d; - unsigned ud; - } dw1; -}; - - -struct brw_indirect { - unsigned addr_subnr:4; - int addr_offset:10; - unsigned pad:18; -}; - - -#define BRW_EU_MAX_INSN_STACK 5 -#define BRW_EU_MAX_INSN 1200 - -struct brw_compile { - struct brw_instruction store[BRW_EU_MAX_INSN]; - unsigned nr_insn; - - /* Allow clients to push/pop instruction state: - */ - struct brw_instruction stack[BRW_EU_MAX_INSN_STACK]; - struct brw_instruction *current; - - unsigned flag_value; - boolean single_program_flow; -}; - - - -static __inline int type_sz( unsigned type ) -{ - switch( type ) { - case BRW_REGISTER_TYPE_UD: - case BRW_REGISTER_TYPE_D: - case BRW_REGISTER_TYPE_F: - return 4; - case BRW_REGISTER_TYPE_HF: - case BRW_REGISTER_TYPE_UW: - case BRW_REGISTER_TYPE_W: - return 2; - case BRW_REGISTER_TYPE_UB: - case BRW_REGISTER_TYPE_B: - return 1; - default: - return 0; - } -} - -static __inline struct brw_reg brw_reg( unsigned file, - unsigned nr, - unsigned subnr, - unsigned type, - unsigned vstride, - unsigned width, - unsigned hstride, - unsigned swizzle, - unsigned writemask) -{ - - struct brw_reg reg; - reg.type = type; - reg.file = file; - reg.nr = nr; - reg.subnr = subnr * type_sz(type); - reg.negate = 0; - reg.abs = 0; - reg.vstride = vstride; - reg.width = width; - reg.hstride = hstride; - reg.address_mode = BRW_ADDRESS_DIRECT; - reg.pad0 = 0; - - /* Could do better: If the reg is r5.3<0;1,0>, we probably want to - * set swizzle and writemask to W, as the lower bits of subnr will - * be lost when converted to align16. This is probably too much to - * keep track of as you'd want it adjusted by suboffset(), etc. - * Perhaps fix up when converting to align16? - */ - reg.dw1.bits.swizzle = swizzle; - reg.dw1.bits.writemask = writemask; - reg.dw1.bits.indirect_offset = 0; - reg.dw1.bits.pad1 = 0; - return reg; -} - -static __inline struct brw_reg brw_vec16_reg( unsigned file, - unsigned nr, - unsigned subnr ) -{ - return brw_reg(file, - nr, - subnr, - BRW_REGISTER_TYPE_F, - BRW_VERTICAL_STRIDE_16, - BRW_WIDTH_16, - BRW_HORIZONTAL_STRIDE_1, - BRW_SWIZZLE_XYZW, - TGSI_WRITEMASK_XYZW); -} - -static __inline struct brw_reg brw_vec8_reg( unsigned file, - unsigned nr, - unsigned subnr ) -{ - return brw_reg(file, - nr, - subnr, - BRW_REGISTER_TYPE_F, - BRW_VERTICAL_STRIDE_8, - BRW_WIDTH_8, - BRW_HORIZONTAL_STRIDE_1, - BRW_SWIZZLE_XYZW, - TGSI_WRITEMASK_XYZW); -} - - -static __inline struct brw_reg brw_vec4_reg( unsigned file, - unsigned nr, - unsigned subnr ) -{ - return brw_reg(file, - nr, - subnr, - BRW_REGISTER_TYPE_F, - BRW_VERTICAL_STRIDE_4, - BRW_WIDTH_4, - BRW_HORIZONTAL_STRIDE_1, - BRW_SWIZZLE_XYZW, - TGSI_WRITEMASK_XYZW); -} - - -static __inline struct brw_reg brw_vec2_reg( unsigned file, - unsigned nr, - unsigned subnr ) -{ - return brw_reg(file, - nr, - subnr, - BRW_REGISTER_TYPE_F, - BRW_VERTICAL_STRIDE_2, - BRW_WIDTH_2, - BRW_HORIZONTAL_STRIDE_1, - BRW_SWIZZLE_XYXY, - TGSI_WRITEMASK_XY); -} - -static __inline struct brw_reg brw_vec1_reg( unsigned file, - unsigned nr, - unsigned subnr ) -{ - return brw_reg(file, - nr, - subnr, - BRW_REGISTER_TYPE_F, - BRW_VERTICAL_STRIDE_0, - BRW_WIDTH_1, - BRW_HORIZONTAL_STRIDE_0, - BRW_SWIZZLE_XXXX, - TGSI_WRITEMASK_X); -} - - -static __inline struct brw_reg retype( struct brw_reg reg, - unsigned type ) -{ - reg.type = type; - return reg; -} - -static __inline struct brw_reg suboffset( struct brw_reg reg, - unsigned delta ) -{ - reg.subnr += delta * type_sz(reg.type); - return reg; -} - - -static __inline struct brw_reg offset( struct brw_reg reg, - unsigned delta ) -{ - reg.nr += delta; - return reg; -} - - -static __inline struct brw_reg byte_offset( struct brw_reg reg, - unsigned bytes ) -{ - unsigned newoffset = reg.nr * REG_SIZE + reg.subnr + bytes; - reg.nr = newoffset / REG_SIZE; - reg.subnr = newoffset % REG_SIZE; - return reg; -} - - -static __inline struct brw_reg brw_uw16_reg( unsigned file, - unsigned nr, - unsigned subnr ) -{ - return suboffset(retype(brw_vec16_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); -} - -static __inline struct brw_reg brw_uw8_reg( unsigned file, - unsigned nr, - unsigned subnr ) -{ - return suboffset(retype(brw_vec8_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); -} - -static __inline struct brw_reg brw_uw1_reg( unsigned file, - unsigned nr, - unsigned subnr ) -{ - return suboffset(retype(brw_vec1_reg(file, nr, 0), BRW_REGISTER_TYPE_UW), subnr); -} - -static __inline struct brw_reg brw_imm_reg( unsigned type ) -{ - return brw_reg( BRW_IMMEDIATE_VALUE, - 0, - 0, - type, - BRW_VERTICAL_STRIDE_0, - BRW_WIDTH_1, - BRW_HORIZONTAL_STRIDE_0, - 0, - 0); -} - -static __inline struct brw_reg brw_imm_f( float f ) -{ - struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_F); - imm.dw1.f = f; - return imm; -} - -static __inline struct brw_reg brw_imm_d( int d ) -{ - struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_D); - imm.dw1.d = d; - return imm; -} - -static __inline struct brw_reg brw_imm_ud( unsigned ud ) -{ - struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UD); - imm.dw1.ud = ud; - return imm; -} - -static __inline struct brw_reg brw_imm_uw( ushort uw ) -{ - struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_UW); - imm.dw1.ud = uw; - return imm; -} - -static __inline struct brw_reg brw_imm_w( short w ) -{ - struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_W); - imm.dw1.d = w; - return imm; -} - -/* brw_imm_b and brw_imm_ub aren't supported by hardware - the type - * numbers alias with _V and _VF below: - */ - -/* Vector of eight signed half-byte values: - */ -static __inline struct brw_reg brw_imm_v( unsigned v ) -{ - struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_V); - imm.vstride = BRW_VERTICAL_STRIDE_0; - imm.width = BRW_WIDTH_8; - imm.hstride = BRW_HORIZONTAL_STRIDE_1; - imm.dw1.ud = v; - return imm; -} - -/* Vector of four 8-bit float values: - */ -static __inline struct brw_reg brw_imm_vf( unsigned v ) -{ - struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); - imm.vstride = BRW_VERTICAL_STRIDE_0; - imm.width = BRW_WIDTH_4; - imm.hstride = BRW_HORIZONTAL_STRIDE_1; - imm.dw1.ud = v; - return imm; -} - -#define VF_ZERO 0x0 -#define VF_ONE 0x30 -#define VF_NEG (1<<7) - -static __inline struct brw_reg brw_imm_vf4( unsigned v0, - unsigned v1, - unsigned v2, - unsigned v3) -{ - struct brw_reg imm = brw_imm_reg(BRW_REGISTER_TYPE_VF); - imm.vstride = BRW_VERTICAL_STRIDE_0; - imm.width = BRW_WIDTH_4; - imm.hstride = BRW_HORIZONTAL_STRIDE_1; - imm.dw1.ud = ((v0 << 0) | - (v1 << 8) | - (v2 << 16) | - (v3 << 24)); - return imm; -} - - -static __inline struct brw_reg brw_address( struct brw_reg reg ) -{ - return brw_imm_uw(reg.nr * REG_SIZE + reg.subnr); -} - - -static __inline struct brw_reg brw_vec1_grf( unsigned nr, - unsigned subnr ) -{ - return brw_vec1_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); -} - -static __inline struct brw_reg brw_vec8_grf( unsigned nr, - unsigned subnr ) -{ - return brw_vec8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); -} - -static __inline struct brw_reg brw_vec4_grf( unsigned nr, - unsigned subnr ) -{ - return brw_vec4_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); -} - - -static __inline struct brw_reg brw_vec2_grf( unsigned nr, - unsigned subnr ) -{ - return brw_vec2_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); -} - -static __inline struct brw_reg brw_uw8_grf( unsigned nr, - unsigned subnr ) -{ - return brw_uw8_reg(BRW_GENERAL_REGISTER_FILE, nr, subnr); -} - -static __inline struct brw_reg brw_null_reg( void ) -{ - return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, - BRW_ARF_NULL, - 0); -} - -static __inline struct brw_reg brw_address_reg( unsigned subnr ) -{ - return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, - BRW_ARF_ADDRESS, - subnr); -} - -/* If/else instructions break in align16 mode if writemask & swizzle - * aren't xyzw. This goes against the convention for other scalar - * regs: - */ -static __inline struct brw_reg brw_ip_reg( void ) -{ - return brw_reg(BRW_ARCHITECTURE_REGISTER_FILE, - BRW_ARF_IP, - 0, - BRW_REGISTER_TYPE_UD, - BRW_VERTICAL_STRIDE_4, /* ? */ - BRW_WIDTH_1, - BRW_HORIZONTAL_STRIDE_0, - BRW_SWIZZLE_XYZW, /* NOTE! */ - TGSI_WRITEMASK_XYZW); /* NOTE! */ -} - -static __inline struct brw_reg brw_acc_reg( void ) -{ - return brw_vec8_reg(BRW_ARCHITECTURE_REGISTER_FILE, - BRW_ARF_ACCUMULATOR, - 0); -} - - -static __inline struct brw_reg brw_flag_reg( void ) -{ - return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, - BRW_ARF_FLAG, - 0); -} - - -static __inline struct brw_reg brw_mask_reg( unsigned subnr ) -{ - return brw_uw1_reg(BRW_ARCHITECTURE_REGISTER_FILE, - BRW_ARF_MASK, - subnr); -} - -static __inline struct brw_reg brw_message_reg( unsigned nr ) -{ - return brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, - nr, - 0); -} - - - - -/* This is almost always called with a numeric constant argument, so - * make things easy to evaluate at compile time: - */ -static __inline unsigned cvt( unsigned val ) -{ - switch (val) { - case 0: return 0; - case 1: return 1; - case 2: return 2; - case 4: return 3; - case 8: return 4; - case 16: return 5; - case 32: return 6; - } - return 0; -} - -static __inline struct brw_reg stride( struct brw_reg reg, - unsigned vstride, - unsigned width, - unsigned hstride ) -{ - - reg.vstride = cvt(vstride); - reg.width = cvt(width) - 1; - reg.hstride = cvt(hstride); - return reg; -} - -static __inline struct brw_reg vec16( struct brw_reg reg ) -{ - return stride(reg, 16,16,1); -} - -static __inline struct brw_reg vec8( struct brw_reg reg ) -{ - return stride(reg, 8,8,1); -} - -static __inline struct brw_reg vec4( struct brw_reg reg ) -{ - return stride(reg, 4,4,1); -} - -static __inline struct brw_reg vec2( struct brw_reg reg ) -{ - return stride(reg, 2,2,1); -} - -static __inline struct brw_reg vec1( struct brw_reg reg ) -{ - return stride(reg, 0,1,0); -} - -static __inline struct brw_reg get_element( struct brw_reg reg, unsigned elt ) -{ - return vec1(suboffset(reg, elt)); -} - -static __inline struct brw_reg get_element_ud( struct brw_reg reg, unsigned elt ) -{ - return vec1(suboffset(retype(reg, BRW_REGISTER_TYPE_UD), elt)); -} - - -static __inline struct brw_reg brw_swizzle( struct brw_reg reg, - unsigned x, - unsigned y, - unsigned z, - unsigned w) -{ - reg.dw1.bits.swizzle = BRW_SWIZZLE4(BRW_GET_SWZ(reg.dw1.bits.swizzle, x), - BRW_GET_SWZ(reg.dw1.bits.swizzle, y), - BRW_GET_SWZ(reg.dw1.bits.swizzle, z), - BRW_GET_SWZ(reg.dw1.bits.swizzle, w)); - return reg; -} - - -static __inline struct brw_reg brw_swizzle1( struct brw_reg reg, - unsigned x ) -{ - return brw_swizzle(reg, x, x, x, x); -} - -static __inline struct brw_reg brw_writemask( struct brw_reg reg, - unsigned mask ) -{ - reg.dw1.bits.writemask &= mask; - return reg; -} - -static __inline struct brw_reg brw_set_writemask( struct brw_reg reg, - unsigned mask ) -{ - reg.dw1.bits.writemask = mask; - return reg; -} - -static __inline struct brw_reg negate( struct brw_reg reg ) -{ - reg.negate ^= 1; - return reg; -} - -static __inline struct brw_reg brw_abs( struct brw_reg reg ) -{ - reg.abs = 1; - return reg; -} - -/*********************************************************************** - */ -static __inline struct brw_reg brw_vec4_indirect( unsigned subnr, - int offset ) -{ - struct brw_reg reg = brw_vec4_grf(0, 0); - reg.subnr = subnr; - reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; - reg.dw1.bits.indirect_offset = offset; - return reg; -} - -static __inline struct brw_reg brw_vec1_indirect( unsigned subnr, - int offset ) -{ - struct brw_reg reg = brw_vec1_grf(0, 0); - reg.subnr = subnr; - reg.address_mode = BRW_ADDRESS_REGISTER_INDIRECT_REGISTER; - reg.dw1.bits.indirect_offset = offset; - return reg; -} - -static __inline struct brw_reg deref_4f(struct brw_indirect ptr, int offset) -{ - return brw_vec4_indirect(ptr.addr_subnr, ptr.addr_offset + offset); -} - -static __inline struct brw_reg deref_1f(struct brw_indirect ptr, int offset) -{ - return brw_vec1_indirect(ptr.addr_subnr, ptr.addr_offset + offset); -} - -static __inline struct brw_reg deref_4b(struct brw_indirect ptr, int offset) -{ - return retype(deref_4f(ptr, offset), BRW_REGISTER_TYPE_B); -} - -static __inline struct brw_reg deref_1uw(struct brw_indirect ptr, int offset) -{ - return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UW); -} - -static __inline struct brw_reg deref_1ud(struct brw_indirect ptr, int offset) -{ - return retype(deref_1f(ptr, offset), BRW_REGISTER_TYPE_UD); -} - -static __inline struct brw_reg get_addr_reg(struct brw_indirect ptr) -{ - return brw_address_reg(ptr.addr_subnr); -} - -static __inline struct brw_indirect brw_indirect_offset( struct brw_indirect ptr, int offset ) -{ - ptr.addr_offset += offset; - return ptr; -} - -static __inline struct brw_indirect brw_indirect( unsigned addr_subnr, int offset ) -{ - struct brw_indirect ptr; - ptr.addr_subnr = addr_subnr; - ptr.addr_offset = offset; - ptr.pad = 0; - return ptr; -} - -static __inline struct brw_instruction *current_insn( struct brw_compile *p) -{ - return &p->store[p->nr_insn]; -} - -void brw_pop_insn_state( struct brw_compile *p ); -void brw_push_insn_state( struct brw_compile *p ); -void brw_set_mask_control( struct brw_compile *p, unsigned value ); -void brw_set_saturate( struct brw_compile *p, unsigned value ); -void brw_set_access_mode( struct brw_compile *p, unsigned access_mode ); -void brw_set_compression_control( struct brw_compile *p, boolean control ); -void brw_set_predicate_control_flag_value( struct brw_compile *p, unsigned value ); -void brw_set_predicate_control( struct brw_compile *p, unsigned pc ); -void brw_set_conditionalmod( struct brw_compile *p, unsigned conditional ); - -void brw_init_compile( struct brw_compile *p ); -const unsigned *brw_get_program( struct brw_compile *p, unsigned *sz ); - - -struct brw_instruction *brw_alu1( struct brw_compile *p, - unsigned opcode, - struct brw_reg dest, - struct brw_reg src ); - -struct brw_instruction *brw_alu2(struct brw_compile *p, - unsigned opcode, - struct brw_reg dest, - struct brw_reg src0, - struct brw_reg src1 ); - -/* Helpers for regular instructions: - */ -#define ALU1(OP) \ -struct brw_instruction *brw_##OP(struct brw_compile *p, \ - struct brw_reg dest, \ - struct brw_reg src0); - -#define ALU2(OP) \ -struct brw_instruction *brw_##OP(struct brw_compile *p, \ - struct brw_reg dest, \ - struct brw_reg src0, \ - struct brw_reg src1); - -ALU1(MOV) -ALU2(SEL) -ALU1(NOT) -ALU2(AND) -ALU2(OR) -ALU2(XOR) -ALU2(SHR) -ALU2(SHL) -ALU2(RSR) -ALU2(RSL) -ALU2(ASR) -ALU2(JMPI) -ALU2(ADD) -ALU2(MUL) -ALU1(FRC) -ALU1(RNDD) -ALU2(MAC) -ALU2(MACH) -ALU1(LZD) -ALU2(DP4) -ALU2(DPH) -ALU2(DP3) -ALU2(DP2) -ALU2(LINE) - -#undef ALU1 -#undef ALU2 - - - -/* Helpers for SEND instruction: - */ -void brw_urb_WRITE(struct brw_compile *p, - struct brw_reg dest, - unsigned msg_reg_nr, - struct brw_reg src0, - boolean allocate, - boolean used, - unsigned msg_length, - unsigned response_length, - boolean eot, - boolean writes_complete, - unsigned offset, - unsigned swizzle); - -void brw_fb_WRITE(struct brw_compile *p, - struct brw_reg dest, - unsigned msg_reg_nr, - struct brw_reg src0, - unsigned binding_table_index, - unsigned msg_length, - unsigned response_length, - boolean eot); - -void brw_SAMPLE(struct brw_compile *p, - struct brw_reg dest, - unsigned msg_reg_nr, - struct brw_reg src0, - unsigned binding_table_index, - unsigned sampler, - unsigned writemask, - unsigned msg_type, - unsigned response_length, - unsigned msg_length, - boolean eot); - -void brw_math_16( struct brw_compile *p, - struct brw_reg dest, - unsigned function, - unsigned saturate, - unsigned msg_reg_nr, - struct brw_reg src, - unsigned precision ); - -void brw_math( struct brw_compile *p, - struct brw_reg dest, - unsigned function, - unsigned saturate, - unsigned msg_reg_nr, - struct brw_reg src, - unsigned data_type, - unsigned precision ); - -void brw_dp_READ_16( struct brw_compile *p, - struct brw_reg dest, - unsigned msg_reg_nr, - unsigned scratch_offset ); - -void brw_dp_WRITE_16( struct brw_compile *p, - struct brw_reg src, - unsigned msg_reg_nr, - unsigned scratch_offset ); - -/* If/else/endif. Works by manipulating the execution flags on each - * channel. - */ -struct brw_instruction *brw_IF(struct brw_compile *p, - unsigned execute_size); - -struct brw_instruction *brw_ELSE(struct brw_compile *p, - struct brw_instruction *if_insn); - -void brw_ENDIF(struct brw_compile *p, - struct brw_instruction *if_or_else_insn); - - -/* DO/WHILE loops: - */ -struct brw_instruction *brw_DO(struct brw_compile *p, - unsigned execute_size); - -struct brw_instruction *brw_WHILE(struct brw_compile *p, - struct brw_instruction *patch_insn); - -struct brw_instruction *brw_BREAK(struct brw_compile *p); -struct brw_instruction *brw_CONT(struct brw_compile *p); -/* Forward jumps: - */ -void brw_land_fwd_jump(struct brw_compile *p, - struct brw_instruction *jmp_insn); - - - -void brw_NOP(struct brw_compile *p); - -/* Special case: there is never a destination, execution size will be - * taken from src0: - */ -void brw_CMP(struct brw_compile *p, - struct brw_reg dest, - unsigned conditional, - struct brw_reg src0, - struct brw_reg src1); - -void brw_print_reg( struct brw_reg reg ); - - -/*********************************************************************** - * brw_eu_util.c: - */ - -void brw_copy_indirect_to_indirect(struct brw_compile *p, - struct brw_indirect dst_ptr, - struct brw_indirect src_ptr, - unsigned count); - -void brw_copy_from_indirect(struct brw_compile *p, - struct brw_reg dst, - struct brw_indirect ptr, - unsigned count); - -void brw_copy4(struct brw_compile *p, - struct brw_reg dst, - struct brw_reg src, - unsigned count); - -void brw_copy8(struct brw_compile *p, - struct brw_reg dst, - struct brw_reg src, - unsigned count); - -void brw_math_invert( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg src); - -void brw_set_src1( struct brw_instruction *insn, - struct brw_reg reg ); -#endif diff --git a/src/mesa/pipe/i965simple/brw_eu_debug.c b/src/mesa/pipe/i965simple/brw_eu_debug.c deleted file mode 100644 index 4a94ddefa6..0000000000 --- a/src/mesa/pipe/i965simple/brw_eu_debug.c +++ /dev/null @@ -1,90 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "pipe/p_debug.h" - -#include "brw_eu.h" - -void brw_print_reg( struct brw_reg hwreg ) -{ - static const char *file[] = { - "arf", - "grf", - "msg", - "imm" - }; - - static const char *type[] = { - "ud", - "d", - "uw", - "w", - "ub", - "vf", - "hf", - "f" - }; - - debug_printf("%s%s", - hwreg.abs ? "abs/" : "", - hwreg.negate ? "-" : ""); - - if (hwreg.file == BRW_GENERAL_REGISTER_FILE && - hwreg.nr % 2 == 0 && - hwreg.subnr == 0 && - hwreg.vstride == BRW_VERTICAL_STRIDE_8 && - hwreg.width == BRW_WIDTH_8 && - hwreg.hstride == BRW_HORIZONTAL_STRIDE_1 && - hwreg.type == BRW_REGISTER_TYPE_F) { - debug_printf("vec%d", hwreg.nr); - } - else if (hwreg.file == BRW_GENERAL_REGISTER_FILE && - hwreg.vstride == BRW_VERTICAL_STRIDE_0 && - hwreg.width == BRW_WIDTH_1 && - hwreg.hstride == BRW_HORIZONTAL_STRIDE_0 && - hwreg.type == BRW_REGISTER_TYPE_F) { - debug_printf("scl%d.%d", hwreg.nr, hwreg.subnr / 4); - } - else { - debug_printf("%s%d.%d<%d;%d,%d>:%s", - file[hwreg.file], - hwreg.nr, - hwreg.subnr / type_sz(hwreg.type), - hwreg.vstride ? (1<<(hwreg.vstride-1)) : 0, - 1< - */ - - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_eu.h" - - - - -/*********************************************************************** - * Internal helper for constructing instructions - */ - -static void guess_execution_size( struct brw_instruction *insn, - struct brw_reg reg ) -{ - if (reg.width == BRW_WIDTH_8 && - insn->header.compression_control == BRW_COMPRESSION_COMPRESSED) - insn->header.execution_size = BRW_EXECUTE_16; - else - insn->header.execution_size = reg.width; /* note - definitions are compatible */ -} - - -static void brw_set_dest( struct brw_instruction *insn, - struct brw_reg dest ) -{ - insn->bits1.da1.dest_reg_file = dest.file; - insn->bits1.da1.dest_reg_type = dest.type; - insn->bits1.da1.dest_address_mode = dest.address_mode; - - if (dest.address_mode == BRW_ADDRESS_DIRECT) { - insn->bits1.da1.dest_reg_nr = dest.nr; - - if (insn->header.access_mode == BRW_ALIGN_1) { - insn->bits1.da1.dest_subreg_nr = dest.subnr; - insn->bits1.da1.dest_horiz_stride = BRW_HORIZONTAL_STRIDE_1; - } - else { - insn->bits1.da16.dest_subreg_nr = dest.subnr / 16; - insn->bits1.da16.dest_writemask = dest.dw1.bits.writemask; - } - } - else { - insn->bits1.ia1.dest_subreg_nr = dest.subnr; - - /* These are different sizes in align1 vs align16: - */ - if (insn->header.access_mode == BRW_ALIGN_1) { - insn->bits1.ia1.dest_indirect_offset = dest.dw1.bits.indirect_offset; - insn->bits1.ia1.dest_horiz_stride = BRW_HORIZONTAL_STRIDE_1; - } - else { - insn->bits1.ia16.dest_indirect_offset = dest.dw1.bits.indirect_offset; - } - } - - /* NEW: Set the execution size based on dest.width and - * insn->compression_control: - */ - guess_execution_size(insn, dest); -} - -static void brw_set_src0( struct brw_instruction *insn, - struct brw_reg reg ) -{ - assert(reg.file != BRW_MESSAGE_REGISTER_FILE); - - insn->bits1.da1.src0_reg_file = reg.file; - insn->bits1.da1.src0_reg_type = reg.type; - insn->bits2.da1.src0_abs = reg.abs; - insn->bits2.da1.src0_negate = reg.negate; - insn->bits2.da1.src0_address_mode = reg.address_mode; - - if (reg.file == BRW_IMMEDIATE_VALUE) { - insn->bits3.ud = reg.dw1.ud; - - /* Required to set some fields in src1 as well: - */ - insn->bits1.da1.src1_reg_file = 0; /* arf */ - insn->bits1.da1.src1_reg_type = reg.type; - } - else - { - if (reg.address_mode == BRW_ADDRESS_DIRECT) { - if (insn->header.access_mode == BRW_ALIGN_1) { - insn->bits2.da1.src0_subreg_nr = reg.subnr; - insn->bits2.da1.src0_reg_nr = reg.nr; - } - else { - insn->bits2.da16.src0_subreg_nr = reg.subnr / 16; - insn->bits2.da16.src0_reg_nr = reg.nr; - } - } - else { - insn->bits2.ia1.src0_subreg_nr = reg.subnr; - - if (insn->header.access_mode == BRW_ALIGN_1) { - insn->bits2.ia1.src0_indirect_offset = reg.dw1.bits.indirect_offset; - } - else { - insn->bits2.ia16.src0_subreg_nr = reg.dw1.bits.indirect_offset; - } - } - - if (insn->header.access_mode == BRW_ALIGN_1) { - if (reg.width == BRW_WIDTH_1 && - insn->header.execution_size == BRW_EXECUTE_1) { - insn->bits2.da1.src0_horiz_stride = BRW_HORIZONTAL_STRIDE_0; - insn->bits2.da1.src0_width = BRW_WIDTH_1; - insn->bits2.da1.src0_vert_stride = BRW_VERTICAL_STRIDE_0; - } - else { - insn->bits2.da1.src0_horiz_stride = reg.hstride; - insn->bits2.da1.src0_width = reg.width; - insn->bits2.da1.src0_vert_stride = reg.vstride; - } - } - else { - insn->bits2.da16.src0_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); - insn->bits2.da16.src0_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); - insn->bits2.da16.src0_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); - insn->bits2.da16.src0_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); - - /* This is an oddity of the fact we're using the same - * descriptions for registers in align_16 as align_1: - */ - if (reg.vstride == BRW_VERTICAL_STRIDE_8) - insn->bits2.da16.src0_vert_stride = BRW_VERTICAL_STRIDE_4; - else - insn->bits2.da16.src0_vert_stride = reg.vstride; - } - } -} - - -void brw_set_src1( struct brw_instruction *insn, - struct brw_reg reg ) -{ - assert(reg.file != BRW_MESSAGE_REGISTER_FILE); - - insn->bits1.da1.src1_reg_file = reg.file; - insn->bits1.da1.src1_reg_type = reg.type; - insn->bits3.da1.src1_abs = reg.abs; - insn->bits3.da1.src1_negate = reg.negate; - - /* Only src1 can be immediate in two-argument instructions. - */ - assert(insn->bits1.da1.src0_reg_file != BRW_IMMEDIATE_VALUE); - - if (reg.file == BRW_IMMEDIATE_VALUE) { - insn->bits3.ud = reg.dw1.ud; - } - else { - /* This is a hardware restriction, which may or may not be lifted - * in the future: - */ - assert (reg.address_mode == BRW_ADDRESS_DIRECT); - //assert (reg.file == BRW_GENERAL_REGISTER_FILE); - - if (insn->header.access_mode == BRW_ALIGN_1) { - insn->bits3.da1.src1_subreg_nr = reg.subnr; - insn->bits3.da1.src1_reg_nr = reg.nr; - } - else { - insn->bits3.da16.src1_subreg_nr = reg.subnr / 16; - insn->bits3.da16.src1_reg_nr = reg.nr; - } - - if (insn->header.access_mode == BRW_ALIGN_1) { - if (reg.width == BRW_WIDTH_1 && - insn->header.execution_size == BRW_EXECUTE_1) { - insn->bits3.da1.src1_horiz_stride = BRW_HORIZONTAL_STRIDE_0; - insn->bits3.da1.src1_width = BRW_WIDTH_1; - insn->bits3.da1.src1_vert_stride = BRW_VERTICAL_STRIDE_0; - } - else { - insn->bits3.da1.src1_horiz_stride = reg.hstride; - insn->bits3.da1.src1_width = reg.width; - insn->bits3.da1.src1_vert_stride = reg.vstride; - } - } - else { - insn->bits3.da16.src1_swz_x = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_X); - insn->bits3.da16.src1_swz_y = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Y); - insn->bits3.da16.src1_swz_z = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_Z); - insn->bits3.da16.src1_swz_w = BRW_GET_SWZ(reg.dw1.bits.swizzle, BRW_CHANNEL_W); - - /* This is an oddity of the fact we're using the same - * descriptions for registers in align_16 as align_1: - */ - if (reg.vstride == BRW_VERTICAL_STRIDE_8) - insn->bits3.da16.src1_vert_stride = BRW_VERTICAL_STRIDE_4; - else - insn->bits3.da16.src1_vert_stride = reg.vstride; - } - } -} - - - -static void brw_set_math_message( struct brw_instruction *insn, - unsigned msg_length, - unsigned response_length, - unsigned function, - unsigned integer_type, - boolean low_precision, - boolean saturate, - unsigned dataType ) -{ - brw_set_src1(insn, brw_imm_d(0)); - - insn->bits3.math.function = function; - insn->bits3.math.int_type = integer_type; - insn->bits3.math.precision = low_precision; - insn->bits3.math.saturate = saturate; - insn->bits3.math.data_type = dataType; - insn->bits3.math.response_length = response_length; - insn->bits3.math.msg_length = msg_length; - insn->bits3.math.msg_target = BRW_MESSAGE_TARGET_MATH; - insn->bits3.math.end_of_thread = 0; -} - -static void brw_set_urb_message( struct brw_instruction *insn, - boolean allocate, - boolean used, - unsigned msg_length, - unsigned response_length, - boolean end_of_thread, - boolean complete, - unsigned offset, - unsigned swizzle_control ) -{ - brw_set_src1(insn, brw_imm_d(0)); - - insn->bits3.urb.opcode = 0; /* ? */ - insn->bits3.urb.offset = offset; - insn->bits3.urb.swizzle_control = swizzle_control; - insn->bits3.urb.allocate = allocate; - insn->bits3.urb.used = used; /* ? */ - insn->bits3.urb.complete = complete; - insn->bits3.urb.response_length = response_length; - insn->bits3.urb.msg_length = msg_length; - insn->bits3.urb.msg_target = BRW_MESSAGE_TARGET_URB; - insn->bits3.urb.end_of_thread = end_of_thread; -} - -static void brw_set_dp_write_message( struct brw_instruction *insn, - unsigned binding_table_index, - unsigned msg_control, - unsigned msg_type, - unsigned msg_length, - unsigned pixel_scoreboard_clear, - unsigned response_length, - unsigned end_of_thread ) -{ - brw_set_src1(insn, brw_imm_d(0)); - - insn->bits3.dp_write.binding_table_index = binding_table_index; - insn->bits3.dp_write.msg_control = msg_control; - insn->bits3.dp_write.pixel_scoreboard_clear = pixel_scoreboard_clear; - insn->bits3.dp_write.msg_type = msg_type; - insn->bits3.dp_write.send_commit_msg = 0; - insn->bits3.dp_write.response_length = response_length; - insn->bits3.dp_write.msg_length = msg_length; - insn->bits3.dp_write.msg_target = BRW_MESSAGE_TARGET_DATAPORT_WRITE; - insn->bits3.urb.end_of_thread = end_of_thread; -} - -static void brw_set_dp_read_message( struct brw_instruction *insn, - unsigned binding_table_index, - unsigned msg_control, - unsigned msg_type, - unsigned target_cache, - unsigned msg_length, - unsigned response_length, - unsigned end_of_thread ) -{ - brw_set_src1(insn, brw_imm_d(0)); - - insn->bits3.dp_read.binding_table_index = binding_table_index; - insn->bits3.dp_read.msg_control = msg_control; - insn->bits3.dp_read.msg_type = msg_type; - insn->bits3.dp_read.target_cache = target_cache; - insn->bits3.dp_read.response_length = response_length; - insn->bits3.dp_read.msg_length = msg_length; - insn->bits3.dp_read.msg_target = BRW_MESSAGE_TARGET_DATAPORT_READ; - insn->bits3.dp_read.end_of_thread = end_of_thread; -} - -static void brw_set_sampler_message( struct brw_instruction *insn, - unsigned binding_table_index, - unsigned sampler, - unsigned msg_type, - unsigned response_length, - unsigned msg_length, - boolean eot) -{ - brw_set_src1(insn, brw_imm_d(0)); - - insn->bits3.sampler.binding_table_index = binding_table_index; - insn->bits3.sampler.sampler = sampler; - insn->bits3.sampler.msg_type = msg_type; - insn->bits3.sampler.return_format = BRW_SAMPLER_RETURN_FORMAT_FLOAT32; - insn->bits3.sampler.response_length = response_length; - insn->bits3.sampler.msg_length = msg_length; - insn->bits3.sampler.end_of_thread = eot; - insn->bits3.sampler.msg_target = BRW_MESSAGE_TARGET_SAMPLER; -} - - - -static struct brw_instruction *next_insn( struct brw_compile *p, - unsigned opcode ) -{ - struct brw_instruction *insn; - - assert(p->nr_insn + 1 < BRW_EU_MAX_INSN); - - insn = &p->store[p->nr_insn++]; - memcpy(insn, p->current, sizeof(*insn)); - - /* Reset this one-shot flag: - */ - - if (p->current->header.destreg__conditonalmod) { - p->current->header.destreg__conditonalmod = 0; - p->current->header.predicate_control = BRW_PREDICATE_NORMAL; - } - - insn->header.opcode = opcode; - return insn; -} - - -struct brw_instruction *brw_alu1( struct brw_compile *p, - unsigned opcode, - struct brw_reg dest, - struct brw_reg src ) -{ - struct brw_instruction *insn = next_insn(p, opcode); - brw_set_dest(insn, dest); - brw_set_src0(insn, src); - return insn; -} - -struct brw_instruction *brw_alu2(struct brw_compile *p, - unsigned opcode, - struct brw_reg dest, - struct brw_reg src0, - struct brw_reg src1 ) -{ - struct brw_instruction *insn = next_insn(p, opcode); - brw_set_dest(insn, dest); - brw_set_src0(insn, src0); - brw_set_src1(insn, src1); - return insn; -} - - -/*********************************************************************** - * Convenience routines. - */ -#define ALU1(OP) \ -struct brw_instruction *brw_##OP(struct brw_compile *p, \ - struct brw_reg dest, \ - struct brw_reg src0) \ -{ \ - return brw_alu1(p, BRW_OPCODE_##OP, dest, src0); \ -} - -#define ALU2(OP) \ -struct brw_instruction *brw_##OP(struct brw_compile *p, \ - struct brw_reg dest, \ - struct brw_reg src0, \ - struct brw_reg src1) \ -{ \ - return brw_alu2(p, BRW_OPCODE_##OP, dest, src0, src1); \ -} - - -ALU1(MOV) -ALU2(SEL) -ALU1(NOT) -ALU2(AND) -ALU2(OR) -ALU2(XOR) -ALU2(SHR) -ALU2(SHL) -ALU2(RSR) -ALU2(RSL) -ALU2(ASR) -ALU2(ADD) -ALU2(MUL) -ALU1(FRC) -ALU1(RNDD) -ALU2(MAC) -ALU2(MACH) -ALU1(LZD) -ALU2(DP4) -ALU2(DPH) -ALU2(DP3) -ALU2(DP2) -ALU2(LINE) - - - - -void brw_NOP(struct brw_compile *p) -{ - struct brw_instruction *insn = next_insn(p, BRW_OPCODE_NOP); - brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); - brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); - brw_set_src1(insn, brw_imm_ud(0x0)); -} - - - - - -/*********************************************************************** - * Comparisons, if/else/endif - */ - -struct brw_instruction *brw_JMPI(struct brw_compile *p, - struct brw_reg dest, - struct brw_reg src0, - struct brw_reg src1) -{ - struct brw_instruction *insn = brw_alu2(p, BRW_OPCODE_JMPI, dest, src0, src1); - - p->current->header.predicate_control = BRW_PREDICATE_NONE; - - return insn; -} - -/* EU takes the value from the flag register and pushes it onto some - * sort of a stack (presumably merging with any flag value already on - * the stack). Within an if block, the flags at the top of the stack - * control execution on each channel of the unit, eg. on each of the - * 16 pixel values in our wm programs. - * - * When the matching 'else' instruction is reached (presumably by - * countdown of the instruction count patched in by our ELSE/ENDIF - * functions), the relevent flags are inverted. - * - * When the matching 'endif' instruction is reached, the flags are - * popped off. If the stack is now empty, normal execution resumes. - * - * No attempt is made to deal with stack overflow (14 elements?). - */ -struct brw_instruction *brw_IF(struct brw_compile *p, unsigned execute_size) -{ - struct brw_instruction *insn; - - if (p->single_program_flow) { - assert(execute_size == BRW_EXECUTE_1); - - insn = next_insn(p, BRW_OPCODE_ADD); - insn->header.predicate_inverse = 1; - } else { - insn = next_insn(p, BRW_OPCODE_IF); - } - - /* Override the defaults for this instruction: - */ - brw_set_dest(insn, brw_ip_reg()); - brw_set_src0(insn, brw_ip_reg()); - brw_set_src1(insn, brw_imm_d(0x0)); - - insn->header.execution_size = execute_size; - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.predicate_control = BRW_PREDICATE_NORMAL; - insn->header.mask_control = BRW_MASK_ENABLE; - - p->current->header.predicate_control = BRW_PREDICATE_NONE; - - return insn; -} - - -struct brw_instruction *brw_ELSE(struct brw_compile *p, - struct brw_instruction *if_insn) -{ - struct brw_instruction *insn; - - if (p->single_program_flow) { - insn = next_insn(p, BRW_OPCODE_ADD); - } else { - insn = next_insn(p, BRW_OPCODE_ELSE); - } - - brw_set_dest(insn, brw_ip_reg()); - brw_set_src0(insn, brw_ip_reg()); - brw_set_src1(insn, brw_imm_d(0x0)); - - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.execution_size = if_insn->header.execution_size; - insn->header.mask_control = BRW_MASK_ENABLE; - - /* Patch the if instruction to point at this instruction. - */ - if (p->single_program_flow) { - assert(if_insn->header.opcode == BRW_OPCODE_ADD); - - if_insn->bits3.ud = (insn - if_insn + 1) * 16; - } else { - assert(if_insn->header.opcode == BRW_OPCODE_IF); - - if_insn->bits3.if_else.jump_count = insn - if_insn; - if_insn->bits3.if_else.pop_count = 1; - if_insn->bits3.if_else.pad0 = 0; - } - - return insn; -} - -void brw_ENDIF(struct brw_compile *p, - struct brw_instruction *patch_insn) -{ - if (p->single_program_flow) { - /* In single program flow mode, there's no need to execute an ENDIF, - * since we don't need to do any stack operations, and if we're executing - * currently, we want to just continue executing. - */ - struct brw_instruction *next = &p->store[p->nr_insn]; - - assert(patch_insn->header.opcode == BRW_OPCODE_ADD); - - patch_insn->bits3.ud = (next - patch_insn) * 16; - } else { - struct brw_instruction *insn = next_insn(p, BRW_OPCODE_ENDIF); - - brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); - brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD)); - brw_set_src1(insn, brw_imm_d(0x0)); - - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.execution_size = patch_insn->header.execution_size; - insn->header.mask_control = BRW_MASK_ENABLE; - - assert(patch_insn->bits3.if_else.jump_count == 0); - - /* Patch the if or else instructions to point at this or the next - * instruction respectively. - */ - if (patch_insn->header.opcode == BRW_OPCODE_IF) { - /* Automagically turn it into an IFF: - */ - patch_insn->header.opcode = BRW_OPCODE_IFF; - patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1; - patch_insn->bits3.if_else.pop_count = 0; - patch_insn->bits3.if_else.pad0 = 0; - } else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) { - patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1; - patch_insn->bits3.if_else.pop_count = 1; - patch_insn->bits3.if_else.pad0 = 0; - } else { - assert(0); - } - - /* Also pop item off the stack in the endif instruction: - */ - insn->bits3.if_else.jump_count = 0; - insn->bits3.if_else.pop_count = 1; - insn->bits3.if_else.pad0 = 0; - } -} - -struct brw_instruction *brw_BREAK(struct brw_compile *p) -{ - struct brw_instruction *insn; - insn = next_insn(p, BRW_OPCODE_BREAK); - brw_set_dest(insn, brw_ip_reg()); - brw_set_src0(insn, brw_ip_reg()); - brw_set_src1(insn, brw_imm_d(0x0)); - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.execution_size = BRW_EXECUTE_8; - insn->header.mask_control = BRW_MASK_DISABLE; - insn->bits3.if_else.pad0 = 0; - return insn; -} - -struct brw_instruction *brw_CONT(struct brw_compile *p) -{ - struct brw_instruction *insn; - insn = next_insn(p, BRW_OPCODE_CONTINUE); - brw_set_dest(insn, brw_ip_reg()); - brw_set_src0(insn, brw_ip_reg()); - brw_set_src1(insn, brw_imm_d(0x0)); - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.execution_size = BRW_EXECUTE_8; - insn->header.mask_control = BRW_MASK_DISABLE; - insn->bits3.if_else.pad0 = 0; - return insn; -} - -/* DO/WHILE loop: - */ -struct brw_instruction *brw_DO(struct brw_compile *p, unsigned execute_size) -{ - if (p->single_program_flow) { - return &p->store[p->nr_insn]; - } else { - struct brw_instruction *insn = next_insn(p, BRW_OPCODE_DO); - - /* Override the defaults for this instruction: - */ - brw_set_dest(insn, brw_null_reg()); - brw_set_src0(insn, brw_null_reg()); - brw_set_src1(insn, brw_null_reg()); - - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.execution_size = execute_size; - insn->header.predicate_control = BRW_PREDICATE_NONE; - /* insn->header.mask_control = BRW_MASK_ENABLE; */ - insn->header.mask_control = BRW_MASK_DISABLE; - - return insn; - } -} - - - -struct brw_instruction *brw_WHILE(struct brw_compile *p, - struct brw_instruction *do_insn) -{ - struct brw_instruction *insn; - - if (p->single_program_flow) - insn = next_insn(p, BRW_OPCODE_ADD); - else - insn = next_insn(p, BRW_OPCODE_WHILE); - - brw_set_dest(insn, brw_ip_reg()); - brw_set_src0(insn, brw_ip_reg()); - brw_set_src1(insn, brw_imm_d(0x0)); - - insn->header.compression_control = BRW_COMPRESSION_NONE; - - if (p->single_program_flow) { - insn->header.execution_size = BRW_EXECUTE_1; - - insn->bits3.d = (do_insn - insn) * 16; - } else { - insn->header.execution_size = do_insn->header.execution_size; - - assert(do_insn->header.opcode == BRW_OPCODE_DO); - insn->bits3.if_else.jump_count = do_insn - insn; - insn->bits3.if_else.pop_count = 0; - insn->bits3.if_else.pad0 = 0; - } - -/* insn->header.mask_control = BRW_MASK_ENABLE; */ - - insn->header.mask_control = BRW_MASK_DISABLE; - p->current->header.predicate_control = BRW_PREDICATE_NONE; - return insn; -} - - -/* FORWARD JUMPS: - */ -void brw_land_fwd_jump(struct brw_compile *p, - struct brw_instruction *jmp_insn) -{ - struct brw_instruction *landing = &p->store[p->nr_insn]; - - assert(jmp_insn->header.opcode == BRW_OPCODE_JMPI); - assert(jmp_insn->bits1.da1.src1_reg_file = BRW_IMMEDIATE_VALUE); - - jmp_insn->bits3.ud = (landing - jmp_insn) - 1; -} - - - -/* To integrate with the above, it makes sense that the comparison - * instruction should populate the flag register. It might be simpler - * just to use the flag reg for most WM tasks? - */ -void brw_CMP(struct brw_compile *p, - struct brw_reg dest, - unsigned conditional, - struct brw_reg src0, - struct brw_reg src1) -{ - struct brw_instruction *insn = next_insn(p, BRW_OPCODE_CMP); - - insn->header.destreg__conditonalmod = conditional; - brw_set_dest(insn, dest); - brw_set_src0(insn, src0); - brw_set_src1(insn, src1); - -/* guess_execution_size(insn, src0); */ - - - /* Make it so that future instructions will use the computed flag - * value until brw_set_predicate_control_flag_value() is called - * again. - */ - if (dest.file == BRW_ARCHITECTURE_REGISTER_FILE && - dest.nr == 0) { - p->current->header.predicate_control = BRW_PREDICATE_NORMAL; - p->flag_value = 0xff; - } -} - - - -/*********************************************************************** - * Helpers for the various SEND message types: - */ - -/* Invert 8 values - */ -void brw_math( struct brw_compile *p, - struct brw_reg dest, - unsigned function, - unsigned saturate, - unsigned msg_reg_nr, - struct brw_reg src, - unsigned data_type, - unsigned precision ) -{ - struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); - unsigned msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1; - unsigned response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1; - - /* Example code doesn't set predicate_control for send - * instructions. - */ - insn->header.predicate_control = 0; - insn->header.destreg__conditonalmod = msg_reg_nr; - - brw_set_dest(insn, dest); - brw_set_src0(insn, src); - brw_set_math_message(insn, - msg_length, response_length, - function, - BRW_MATH_INTEGER_UNSIGNED, - precision, - saturate, - data_type); -} - -/* Use 2 send instructions to invert 16 elements - */ -void brw_math_16( struct brw_compile *p, - struct brw_reg dest, - unsigned function, - unsigned saturate, - unsigned msg_reg_nr, - struct brw_reg src, - unsigned precision ) -{ - struct brw_instruction *insn; - unsigned msg_length = (function == BRW_MATH_FUNCTION_POW) ? 2 : 1; - unsigned response_length = (function == BRW_MATH_FUNCTION_SINCOS) ? 2 : 1; - - /* First instruction: - */ - brw_push_insn_state(p); - brw_set_predicate_control_flag_value(p, 0xff); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - - insn = next_insn(p, BRW_OPCODE_SEND); - insn->header.destreg__conditonalmod = msg_reg_nr; - - brw_set_dest(insn, dest); - brw_set_src0(insn, src); - brw_set_math_message(insn, - msg_length, response_length, - function, - BRW_MATH_INTEGER_UNSIGNED, - precision, - saturate, - BRW_MATH_DATA_VECTOR); - - /* Second instruction: - */ - insn = next_insn(p, BRW_OPCODE_SEND); - insn->header.compression_control = BRW_COMPRESSION_2NDHALF; - insn->header.destreg__conditonalmod = msg_reg_nr+1; - - brw_set_dest(insn, offset(dest,1)); - brw_set_src0(insn, src); - brw_set_math_message(insn, - msg_length, response_length, - function, - BRW_MATH_INTEGER_UNSIGNED, - precision, - saturate, - BRW_MATH_DATA_VECTOR); - - brw_pop_insn_state(p); -} - - - - -void brw_dp_WRITE_16( struct brw_compile *p, - struct brw_reg src, - unsigned msg_reg_nr, - unsigned scratch_offset ) -{ - { - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - - brw_MOV(p, - retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D), - brw_imm_d(scratch_offset)); - - brw_pop_insn_state(p); - } - - { - unsigned msg_length = 3; - struct brw_reg dest = retype(brw_null_reg(), BRW_REGISTER_TYPE_UW); - struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); - - insn->header.predicate_control = 0; /* XXX */ - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.destreg__conditonalmod = msg_reg_nr; - - brw_set_dest(insn, dest); - brw_set_src0(insn, src); - - brw_set_dp_write_message(insn, - 255, /* bti */ - BRW_DATAPORT_OWORD_BLOCK_4_OWORDS, /* msg_control */ - BRW_DATAPORT_WRITE_MESSAGE_OWORD_BLOCK_WRITE, /* msg_type */ - msg_length, - 0, /* pixel scoreboard */ - 0, /* response_length */ - 0); /* eot */ - } - -} - - -void brw_dp_READ_16( struct brw_compile *p, - struct brw_reg dest, - unsigned msg_reg_nr, - unsigned scratch_offset ) -{ - { - brw_push_insn_state(p); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - brw_set_mask_control(p, BRW_MASK_DISABLE); - - brw_MOV(p, - retype(brw_vec1_grf(0, 2), BRW_REGISTER_TYPE_D), - brw_imm_d(scratch_offset)); - - brw_pop_insn_state(p); - } - - { - struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); - - insn->header.predicate_control = 0; /* XXX */ - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.destreg__conditonalmod = msg_reg_nr; - - brw_set_dest(insn, dest); /* UW? */ - brw_set_src0(insn, retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW)); - - brw_set_dp_read_message(insn, - 255, /* bti */ - 3, /* msg_control */ - BRW_DATAPORT_READ_MESSAGE_OWORD_BLOCK_READ, /* msg_type */ - 1, /* target cache */ - 1, /* msg_length */ - 2, /* response_length */ - 0); /* eot */ - } -} - - -void brw_fb_WRITE(struct brw_compile *p, - struct brw_reg dest, - unsigned msg_reg_nr, - struct brw_reg src0, - unsigned binding_table_index, - unsigned msg_length, - unsigned response_length, - boolean eot) -{ - struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); - - insn->header.predicate_control = 0; /* XXX */ - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.destreg__conditonalmod = msg_reg_nr; - - brw_set_dest(insn, dest); - brw_set_src0(insn, src0); - brw_set_dp_write_message(insn, - binding_table_index, - BRW_DATAPORT_RENDER_TARGET_WRITE_SIMD16_SINGLE_SOURCE, /* msg_control */ - BRW_DATAPORT_WRITE_MESSAGE_RENDER_TARGET_WRITE, /* msg_type */ - msg_length, - 1, /* pixel scoreboard */ - response_length, - eot); -} - - - -void brw_SAMPLE(struct brw_compile *p, - struct brw_reg dest, - unsigned msg_reg_nr, - struct brw_reg src0, - unsigned binding_table_index, - unsigned sampler, - unsigned writemask, - unsigned msg_type, - unsigned response_length, - unsigned msg_length, - boolean eot) -{ - boolean need_stall = 0; - - if(writemask == 0) { -/* debug_printf("%s: zero writemask??\n", __FUNCTION__); */ - return; - } - - /* Hardware doesn't do destination dependency checking on send - * instructions properly. Add a workaround which generates the - * dependency by other means. In practice it seems like this bug - * only crops up for texture samples, and only where registers are - * written by the send and then written again later without being - * read in between. Luckily for us, we already track that - * information and use it to modify the writemask for the - * instruction, so that is a guide for whether a workaround is - * needed. - */ - if (writemask != TGSI_WRITEMASK_XYZW) { - unsigned dst_offset = 0; - unsigned i, newmask = 0, len = 0; - - for (i = 0; i < 4; i++) { - if (writemask & (1<header.predicate_control = 0; /* XXX */ - insn->header.compression_control = BRW_COMPRESSION_NONE; - insn->header.destreg__conditonalmod = msg_reg_nr; - - brw_set_dest(insn, dest); - brw_set_src0(insn, src0); - brw_set_sampler_message(insn, - binding_table_index, - sampler, - msg_type, - response_length, - msg_length, - eot); - } - - if (need_stall) - { - struct brw_reg reg = vec8(offset(dest, response_length-1)); - - /* mov (8) r9.0<1>:f r9.0<8;8,1>:f { Align1 } - */ - brw_push_insn_state(p); - brw_set_compression_control(p, FALSE); - brw_MOV(p, reg, reg); - brw_pop_insn_state(p); - } - -} - -/* All these variables are pretty confusing - we might be better off - * using bitmasks and macros for this, in the old style. Or perhaps - * just having the caller instantiate the fields in dword3 itself. - */ -void brw_urb_WRITE(struct brw_compile *p, - struct brw_reg dest, - unsigned msg_reg_nr, - struct brw_reg src0, - boolean allocate, - boolean used, - unsigned msg_length, - unsigned response_length, - boolean eot, - boolean writes_complete, - unsigned offset, - unsigned swizzle) -{ - struct brw_instruction *insn = next_insn(p, BRW_OPCODE_SEND); - - assert(msg_length < 16); - - brw_set_dest(insn, dest); - brw_set_src0(insn, src0); - brw_set_src1(insn, brw_imm_d(0)); - - insn->header.destreg__conditonalmod = msg_reg_nr; - - brw_set_urb_message(insn, - allocate, - used, - msg_length, - response_length, - eot, - writes_complete, - offset, - swizzle); -} - diff --git a/src/mesa/pipe/i965simple/brw_eu_util.c b/src/mesa/pipe/i965simple/brw_eu_util.c deleted file mode 100644 index 3a65b141f0..0000000000 --- a/src/mesa/pipe/i965simple/brw_eu_util.c +++ /dev/null @@ -1,126 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_eu.h" - - -void brw_math_invert( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg src) -{ - brw_math( p, - dst, - BRW_MATH_FUNCTION_INV, - BRW_MATH_SATURATE_NONE, - 0, - src, - BRW_MATH_PRECISION_FULL, - BRW_MATH_DATA_VECTOR ); -} - - - -void brw_copy4(struct brw_compile *p, - struct brw_reg dst, - struct brw_reg src, - unsigned count) -{ - unsigned i; - - dst = vec4(dst); - src = vec4(src); - - for (i = 0; i < count; i++) - { - unsigned delta = i*32; - brw_MOV(p, byte_offset(dst, delta), byte_offset(src, delta)); - brw_MOV(p, byte_offset(dst, delta+16), byte_offset(src, delta+16)); - } -} - - -void brw_copy8(struct brw_compile *p, - struct brw_reg dst, - struct brw_reg src, - unsigned count) -{ - unsigned i; - - dst = vec8(dst); - src = vec8(src); - - for (i = 0; i < count; i++) - { - unsigned delta = i*32; - brw_MOV(p, byte_offset(dst, delta), byte_offset(src, delta)); - } -} - - -void brw_copy_indirect_to_indirect(struct brw_compile *p, - struct brw_indirect dst_ptr, - struct brw_indirect src_ptr, - unsigned count) -{ - unsigned i; - - for (i = 0; i < count; i++) - { - unsigned delta = i*32; - brw_MOV(p, deref_4f(dst_ptr, delta), deref_4f(src_ptr, delta)); - brw_MOV(p, deref_4f(dst_ptr, delta+16), deref_4f(src_ptr, delta+16)); - } -} - - -void brw_copy_from_indirect(struct brw_compile *p, - struct brw_reg dst, - struct brw_indirect ptr, - unsigned count) -{ - unsigned i; - - dst = vec4(dst); - - for (i = 0; i < count; i++) - { - unsigned delta = i*32; - brw_MOV(p, byte_offset(dst, delta), deref_4f(ptr, delta)); - brw_MOV(p, byte_offset(dst, delta+16), deref_4f(ptr, delta+16)); - } -} - - - - diff --git a/src/mesa/pipe/i965simple/brw_flush.c b/src/mesa/pipe/i965simple/brw_flush.c deleted file mode 100644 index 5216c680cf..0000000000 --- a/src/mesa/pipe/i965simple/brw_flush.c +++ /dev/null @@ -1,80 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - - -#include "pipe/p_defines.h" -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_batch.h" - - -/** - * In future we may want a fence-like interface instead of finish. - */ -static void brw_flush( struct pipe_context *pipe, - unsigned flags ) -{ - struct brw_context *brw = brw_context(pipe); - struct pipe_fence_handle *fence; - - /* Do we need to emit an MI_FLUSH command to flush the hardware - * caches? - */ - if (flags & (PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_TEXTURE_CACHE)) { - struct brw_mi_flush flush; - - memset(&flush, 0, sizeof(flush)); - flush.opcode = CMD_MI_FLUSH; - - if (!(flags & PIPE_FLUSH_RENDER_CACHE)) - flush.flags |= BRW_INHIBIT_FLUSH_RENDER_CACHE; - - if (flags & PIPE_FLUSH_TEXTURE_CACHE) - flush.flags |= BRW_FLUSH_READ_CACHE; - - BRW_BATCH_STRUCT(brw, &flush); - } - - /* If there are no flags, just flush pending commands to hardware: - */ - FLUSH_BATCH( &fence ); - - if (flags & PIPE_FLUSH_WAIT) { -// brw->winsys->wait_fence(brw->winsys, fence); - } -} - - - -void brw_init_flush_functions( struct brw_context *brw ) -{ - brw->pipe.flush = brw_flush; -} diff --git a/src/mesa/pipe/i965simple/brw_gs.c b/src/mesa/pipe/i965simple/brw_gs.c deleted file mode 100644 index de60868ccc..0000000000 --- a/src/mesa/pipe/i965simple/brw_gs.c +++ /dev/null @@ -1,196 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_state.h" -#include "brw_gs.h" - - - -static void compile_gs_prog( struct brw_context *brw, - struct brw_gs_prog_key *key ) -{ - struct brw_gs_compile c; - const unsigned *program; - unsigned program_size; - - memset(&c, 0, sizeof(c)); - - c.key = *key; - - /* Need to locate the two positions present in vertex + header. - * These are currently hardcoded: - */ - c.nr_attrs = brw_count_bits(c.key.attrs); - c.nr_regs = (c.nr_attrs + 1) / 2 + 1; /* are vertices packed, or reg-aligned? */ - c.nr_bytes = c.nr_regs * REG_SIZE; - - - /* Begin the compilation: - */ - brw_init_compile(&c.func); - - c.func.single_program_flow = 1; - - /* For some reason the thread is spawned with only 4 channels - * unmasked. - */ - brw_set_mask_control(&c.func, BRW_MASK_DISABLE); - - - /* Note that primitives which don't require a GS program have - * already been weeded out by this stage: - */ - switch (key->primitive) { - case PIPE_PRIM_QUADS: - brw_gs_quads( &c ); - break; - case PIPE_PRIM_QUAD_STRIP: - brw_gs_quad_strip( &c ); - break; - case PIPE_PRIM_LINE_LOOP: - brw_gs_lines( &c ); - break; - case PIPE_PRIM_LINES: - if (key->hint_gs_always) - brw_gs_lines( &c ); - else { - return; - } - break; - case PIPE_PRIM_TRIANGLES: - if (key->hint_gs_always) - brw_gs_tris( &c ); - else { - return; - } - break; - case PIPE_PRIM_POINTS: - if (key->hint_gs_always) - brw_gs_points( &c ); - else { - return; - } - break; - default: - return; - } - - /* get the program - */ - program = brw_get_program(&c.func, &program_size); - - /* Upload - */ - brw->gs.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_GS_PROG], - &c.key, - sizeof(c.key), - program, - program_size, - &c.prog_data, - &brw->gs.prog_data ); -} - - -static boolean search_cache( struct brw_context *brw, - struct brw_gs_prog_key *key ) -{ - return brw_search_cache(&brw->cache[BRW_GS_PROG], - key, sizeof(*key), - &brw->gs.prog_data, - &brw->gs.prog_gs_offset); -} - - -static const int gs_prim[PIPE_PRIM_POLYGON+1] = { - PIPE_PRIM_POINTS, - PIPE_PRIM_LINES, - PIPE_PRIM_LINE_LOOP, - PIPE_PRIM_LINES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_QUADS, - PIPE_PRIM_QUAD_STRIP, - PIPE_PRIM_TRIANGLES -}; - -static void populate_key( struct brw_context *brw, - struct brw_gs_prog_key *key ) -{ - memset(key, 0, sizeof(*key)); - - /* CACHE_NEW_VS_PROG */ - key->attrs = brw->vs.prog_data->outputs_written; - - /* BRW_NEW_PRIMITIVE */ - key->primitive = gs_prim[brw->primitive]; - - key->hint_gs_always = 0; /* debug code? */ - - key->need_gs_prog = (key->hint_gs_always || - brw->primitive == PIPE_PRIM_QUADS || - brw->primitive == PIPE_PRIM_QUAD_STRIP || - brw->primitive == PIPE_PRIM_LINE_LOOP); -} - -/* Calculate interpolants for triangle and line rasterization. - */ -static void upload_gs_prog( struct brw_context *brw ) -{ - struct brw_gs_prog_key key; - - /* Populate the key: - */ - populate_key(brw, &key); - - if (brw->gs.prog_active != key.need_gs_prog) { - brw->state.dirty.cache |= CACHE_NEW_GS_PROG; - brw->gs.prog_active = key.need_gs_prog; - } - - if (brw->gs.prog_active) { - if (!search_cache(brw, &key)) - compile_gs_prog( brw, &key ); - } -} - - -const struct brw_tracked_state brw_gs_prog = { - .dirty = { - .brw = BRW_NEW_PRIMITIVE, - .cache = CACHE_NEW_VS_PROG - }, - .update = upload_gs_prog -}; diff --git a/src/mesa/pipe/i965simple/brw_gs.h b/src/mesa/pipe/i965simple/brw_gs.h deleted file mode 100644 index f09141c6aa..0000000000 --- a/src/mesa/pipe/i965simple/brw_gs.h +++ /dev/null @@ -1,75 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRW_GS_H -#define BRW_GS_H - - -#include "brw_context.h" -#include "brw_eu.h" - -#define MAX_GS_VERTS (4) - -struct brw_gs_prog_key { - unsigned attrs:32; - unsigned primitive:4; - unsigned hint_gs_always:1; - unsigned need_gs_prog:1; - unsigned pad:26; -}; - -struct brw_gs_compile { - struct brw_compile func; - struct brw_gs_prog_key key; - struct brw_gs_prog_data prog_data; - - struct { - struct brw_reg R0; - struct brw_reg vertex[MAX_GS_VERTS]; - } reg; - - /* 3 different ways of expressing vertex size: - */ - unsigned nr_attrs; - unsigned nr_regs; - unsigned nr_bytes; -}; - -#define ATTR_SIZE (4*4) - -void brw_gs_quads( struct brw_gs_compile *c ); -void brw_gs_quad_strip( struct brw_gs_compile *c ); -void brw_gs_tris( struct brw_gs_compile *c ); -void brw_gs_lines( struct brw_gs_compile *c ); -void brw_gs_points( struct brw_gs_compile *c ); - -#endif diff --git a/src/mesa/pipe/i965simple/brw_gs_emit.c b/src/mesa/pipe/i965simple/brw_gs_emit.c deleted file mode 100644 index c3cc90b10f..0000000000 --- a/src/mesa/pipe/i965simple/brw_gs_emit.c +++ /dev/null @@ -1,148 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_gs.h" - -static void brw_gs_alloc_regs( struct brw_gs_compile *c, - unsigned nr_verts ) -{ - unsigned i = 0,j; - - /* Register usage is static, precompute here: - */ - c->reg.R0 = retype(brw_vec8_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; - - /* Payload vertices plus space for more generated vertices: - */ - for (j = 0; j < nr_verts; j++) { - c->reg.vertex[j] = brw_vec4_grf(i, 0); - i += c->nr_regs; - } - - c->prog_data.urb_read_length = c->nr_regs; - c->prog_data.total_grf = i; -} - - -static void brw_gs_emit_vue(struct brw_gs_compile *c, - struct brw_reg vert, - boolean last, - unsigned header) -{ - struct brw_compile *p = &c->func; - boolean allocate = !last; - - /* Overwrite PrimType and PrimStart in the message header, for - * each vertex in turn: - */ - brw_MOV(p, get_element_ud(c->reg.R0, 2), brw_imm_ud(header)); - - /* Copy the vertex from vertn into m1..mN+1: - */ - brw_copy8(p, brw_message_reg(1), vert, c->nr_regs); - - /* Send each vertex as a seperate write to the urb. This is - * different to the concept in brw_sf_emit.c, where subsequent - * writes are used to build up a single urb entry. Each of these - * writes instantiates a seperate urb entry, and a new one must be - * allocated each time. - */ - brw_urb_WRITE(p, - allocate ? c->reg.R0 : retype(brw_null_reg(), BRW_REGISTER_TYPE_UD), - 0, - c->reg.R0, - allocate, - 1, /* used */ - c->nr_regs + 1, /* msg length */ - allocate ? 1 : 0, /* response length */ - allocate ? 0 : 1, /* eot */ - 1, /* writes_complete */ - 0, /* urb offset */ - BRW_URB_SWIZZLE_NONE); -} - - - -void brw_gs_quads( struct brw_gs_compile *c ) -{ - brw_gs_alloc_regs(c, 4); - - /* Use polygons for correct edgeflag behaviour. Note that vertex 3 - * is the PV for quads, but vertex 0 for polygons: - */ - brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); - brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); -} - -void brw_gs_quad_strip( struct brw_gs_compile *c ) -{ - brw_gs_alloc_regs(c, 4); - - brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); - brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); -} - -void brw_gs_tris( struct brw_gs_compile *c ) -{ - brw_gs_alloc_regs(c, 3); - brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_TRILIST << 2) | R02_PRIM_START)); - brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_TRILIST << 2)); - brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_TRILIST << 2) | R02_PRIM_END)); -} - -void brw_gs_lines( struct brw_gs_compile *c ) -{ - brw_gs_alloc_regs(c, 2); - brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_START)); - brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_END)); -} - -void brw_gs_points( struct brw_gs_compile *c ) -{ - brw_gs_alloc_regs(c, 1); - brw_gs_emit_vue(c, c->reg.vertex[0], 1, ((_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END)); -} - - - - - - - - diff --git a/src/mesa/pipe/i965simple/brw_gs_state.c b/src/mesa/pipe/i965simple/brw_gs_state.c deleted file mode 100644 index 3932e9e939..0000000000 --- a/src/mesa/pipe/i965simple/brw_gs_state.c +++ /dev/null @@ -1,89 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - - -#include "brw_context.h" -#include "brw_state.h" -#include "brw_defines.h" -#include "pipe/p_util.h" - - - -static void upload_gs_unit( struct brw_context *brw ) -{ - struct brw_gs_unit_state gs; - - memset(&gs, 0, sizeof(gs)); - - /* CACHE_NEW_GS_PROG */ - if (brw->gs.prog_active) { - gs.thread0.grf_reg_count = - align(brw->gs.prog_data->total_grf, 16) / 16 - 1; - gs.thread0.kernel_start_pointer = brw->gs.prog_gs_offset >> 6; - gs.thread3.urb_entry_read_length = brw->gs.prog_data->urb_read_length; - } - else { - gs.thread0.grf_reg_count = 0; - gs.thread0.kernel_start_pointer = 0; - gs.thread3.urb_entry_read_length = 1; - } - - /* BRW_NEW_URB_FENCE */ - gs.thread4.nr_urb_entries = brw->urb.nr_gs_entries; - gs.thread4.urb_entry_allocation_size = brw->urb.vsize - 1; - - gs.thread4.max_threads = 0; /* Hardware requirement */ - - if (BRW_DEBUG & DEBUG_STATS) - gs.thread4.stats_enable = 1; - - /* CONSTANT */ - gs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; - gs.thread1.single_program_flow = 1; - gs.thread3.dispatch_grf_start_reg = 1; - gs.thread3.const_urb_entry_read_offset = 0; - gs.thread3.const_urb_entry_read_length = 0; - gs.thread3.urb_entry_read_offset = 0; - - - brw->gs.state_gs_offset = brw_cache_data( &brw->cache[BRW_GS_UNIT], &gs ); -} - - -const struct brw_tracked_state brw_gs_unit = { - .dirty = { - .brw = (BRW_NEW_CURBE_OFFSETS | - BRW_NEW_URB_FENCE), - .cache = CACHE_NEW_GS_PROG - }, - .update = upload_gs_unit -}; diff --git a/src/mesa/pipe/i965simple/brw_misc_state.c b/src/mesa/pipe/i965simple/brw_misc_state.c deleted file mode 100644 index 925049ecc1..0000000000 --- a/src/mesa/pipe/i965simple/brw_misc_state.c +++ /dev/null @@ -1,486 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_batch.h" -#include "brw_context.h" -#include "brw_state.h" -#include "brw_defines.h" - - - - - -/*********************************************************************** - * Blend color - */ - -static void upload_blend_constant_color(struct brw_context *brw) -{ - struct brw_blend_constant_color bcc; - - memset(&bcc, 0, sizeof(bcc)); - bcc.header.opcode = CMD_BLEND_CONSTANT_COLOR; - bcc.header.length = sizeof(bcc)/4-2; - bcc.blend_constant_color[0] = brw->attribs.BlendColor.color[0]; - bcc.blend_constant_color[1] = brw->attribs.BlendColor.color[1]; - bcc.blend_constant_color[2] = brw->attribs.BlendColor.color[2]; - bcc.blend_constant_color[3] = brw->attribs.BlendColor.color[3]; - - BRW_CACHED_BATCH_STRUCT(brw, &bcc); -} - - -const struct brw_tracked_state brw_blend_constant_color = { - .dirty = { - .brw = BRW_NEW_BLEND, - .cache = 0 - }, - .update = upload_blend_constant_color -}; - - -/*********************************************************************** - * Drawing rectangle - */ -static void upload_drawing_rect(struct brw_context *brw) -{ - struct brw_drawrect bdr; - - memset(&bdr, 0, sizeof(bdr)); - bdr.header.opcode = CMD_DRAW_RECT; - bdr.header.length = sizeof(bdr)/4 - 2; - bdr.xmin = 0; - bdr.ymin = 0; - bdr.xmax = brw->attribs.FrameBuffer.cbufs[0]->width; - bdr.ymax = brw->attribs.FrameBuffer.cbufs[0]->height; - bdr.xorg = 0; - bdr.yorg = 0; - - /* Can't use BRW_CACHED_BATCH_STRUCT because this is also emitted - * uncached in brw_draw.c: - */ - BRW_BATCH_STRUCT(brw, &bdr); -} - -const struct brw_tracked_state brw_drawing_rect = { - .dirty = { - .brw = BRW_NEW_SCENE, - .cache = 0 - }, - .update = upload_drawing_rect -}; - -/** - * Upload the binding table pointers, which point each stage's array of surface - * state pointers. - * - * The binding table pointers are relative to the surface state base address, - * which is the BRW_SS_POOL cache buffer. - */ -static void upload_binding_table_pointers(struct brw_context *brw) -{ - struct brw_binding_table_pointers btp; - memset(&btp, 0, sizeof(btp)); - - btp.header.opcode = CMD_BINDING_TABLE_PTRS; - btp.header.length = sizeof(btp)/4 - 2; - btp.vs = 0; - btp.gs = 0; - btp.clp = 0; - btp.sf = 0; - btp.wm = brw->wm.bind_ss_offset; - - BRW_CACHED_BATCH_STRUCT(brw, &btp); -} - -const struct brw_tracked_state brw_binding_table_pointers = { - .dirty = { - .brw = 0, - .cache = CACHE_NEW_SURF_BIND - }, - .update = upload_binding_table_pointers, -}; - - -/** - * Upload pointers to the per-stage state. - * - * The state pointers in this packet are all relative to the general state - * base address set by CMD_STATE_BASE_ADDRESS, which is the BRW_GS_POOL buffer. - */ -static void upload_pipelined_state_pointers(struct brw_context *brw ) -{ - struct brw_pipelined_state_pointers psp; - memset(&psp, 0, sizeof(psp)); - - psp.header.opcode = CMD_PIPELINED_STATE_POINTERS; - psp.header.length = sizeof(psp)/4 - 2; - - psp.vs.offset = brw->vs.state_gs_offset >> 5; - psp.sf.offset = brw->sf.state_gs_offset >> 5; - psp.wm.offset = brw->wm.state_gs_offset >> 5; - psp.cc.offset = brw->cc.state_gs_offset >> 5; - - /* GS gets turned on and off regularly. Need to re-emit URB fence - * after this occurs. - */ - if (brw->gs.prog_active) { - psp.gs.offset = brw->gs.state_gs_offset >> 5; - psp.gs.enable = 1; - } - - if (0) { - psp.clp.offset = brw->clip.state_gs_offset >> 5; - psp.clp.enable = 1; - } - - - if (BRW_CACHED_BATCH_STRUCT(brw, &psp)) - brw->state.dirty.brw |= BRW_NEW_PSP; -} - -const struct brw_tracked_state brw_pipelined_state_pointers = { - .dirty = { - .brw = 0, - .cache = (CACHE_NEW_VS_UNIT | - CACHE_NEW_GS_UNIT | - CACHE_NEW_GS_PROG | - CACHE_NEW_CLIP_UNIT | - CACHE_NEW_SF_UNIT | - CACHE_NEW_WM_UNIT | - CACHE_NEW_CC_UNIT) - }, - .update = upload_pipelined_state_pointers -}; - -static void upload_psp_urb_cbs(struct brw_context *brw ) -{ - upload_pipelined_state_pointers(brw); - brw_upload_urb_fence(brw); - brw_upload_constant_buffer_state(brw); -} - - -const struct brw_tracked_state brw_psp_urb_cbs = { - .dirty = { - .brw = BRW_NEW_URB_FENCE, - .cache = (CACHE_NEW_VS_UNIT | - CACHE_NEW_GS_UNIT | - CACHE_NEW_GS_PROG | - CACHE_NEW_CLIP_UNIT | - CACHE_NEW_SF_UNIT | - CACHE_NEW_WM_UNIT | - CACHE_NEW_CC_UNIT) - }, - .update = upload_psp_urb_cbs -}; - -/** - * Upload the depthbuffer offset and format. - * - * We have to do this per state validation as we need to emit the relocation - * in the batch buffer. - */ -static void upload_depthbuffer(struct brw_context *brw) -{ - struct pipe_surface *depth_surface = brw->attribs.FrameBuffer.zsbuf; - - BEGIN_BATCH(5, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (5 - 2)); - if (depth_surface == NULL) { - OUT_BATCH((BRW_DEPTHFORMAT_D32_FLOAT << 18) | - (BRW_SURFACE_NULL << 29)); - OUT_BATCH(0); - OUT_BATCH(0); - OUT_BATCH(0); - } else { - unsigned int format; - - switch (depth_surface->cpp) { - case 2: - format = BRW_DEPTHFORMAT_D16_UNORM; - break; - case 4: - if (depth_surface->format == PIPE_FORMAT_Z32_FLOAT) - format = BRW_DEPTHFORMAT_D32_FLOAT; - else - format = BRW_DEPTHFORMAT_D24_UNORM_S8_UINT; - break; - default: - assert(0); - return; - } - - OUT_BATCH(((depth_surface->pitch * depth_surface->cpp) - 1) | - (format << 18) | - (BRW_TILEWALK_YMAJOR << 26) | -// (depth_surface->region->tiled << 27) | - (BRW_SURFACE_2D << 29)); - OUT_RELOC(depth_surface->buffer, - PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE, 0); - OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) | - ((depth_surface->pitch - 1) << 6) | - ((depth_surface->height - 1) << 19)); - OUT_BATCH(0); - } - ADVANCE_BATCH(); -} - -const struct brw_tracked_state brw_depthbuffer = { - .dirty = { - .brw = BRW_NEW_SCENE, - .cache = 0 - }, - .update = upload_depthbuffer, -}; - - - - -/*********************************************************************** - * Polygon stipple packet - */ - -static void upload_polygon_stipple(struct brw_context *brw) -{ - struct brw_polygon_stipple bps; - unsigned i; - - memset(&bps, 0, sizeof(bps)); - bps.header.opcode = CMD_POLY_STIPPLE_PATTERN; - bps.header.length = sizeof(bps)/4-2; - - /* XXX: state tracker should send *all* state down initially! - */ - if (brw->attribs.PolygonStipple) - for (i = 0; i < 32; i++) - bps.stipple[i] = brw->attribs.PolygonStipple->stipple[31 - i]; /* invert */ - - BRW_CACHED_BATCH_STRUCT(brw, &bps); -} - -const struct brw_tracked_state brw_polygon_stipple = { - .dirty = { - .brw = BRW_NEW_STIPPLE, - .cache = 0 - }, - .update = upload_polygon_stipple -}; - - -/*********************************************************************** - * Line stipple packet - */ - -static void upload_line_stipple(struct brw_context *brw) -{ - struct brw_line_stipple bls; - float tmp; - int tmpi; - - memset(&bls, 0, sizeof(bls)); - bls.header.opcode = CMD_LINE_STIPPLE_PATTERN; - bls.header.length = sizeof(bls)/4 - 2; - - bls.bits0.pattern = brw->attribs.Raster->line_stipple_pattern; - bls.bits1.repeat_count = brw->attribs.Raster->line_stipple_factor; - - tmp = 1.0 / (float) brw->attribs.Raster->line_stipple_factor; - tmpi = tmp * (1<<13); - - - bls.bits1.inverse_repeat_count = tmpi; - - BRW_CACHED_BATCH_STRUCT(brw, &bls); -} - -const struct brw_tracked_state brw_line_stipple = { - .dirty = { - .brw = BRW_NEW_STIPPLE, - .cache = 0 - }, - .update = upload_line_stipple -}; - - -/*********************************************************************** - * Misc constant state packets - */ - -static void upload_pipe_control(struct brw_context *brw) -{ - struct brw_pipe_control pc; - - return; - - memset(&pc, 0, sizeof(pc)); - - pc.header.opcode = CMD_PIPE_CONTROL; - pc.header.length = sizeof(pc)/4 - 2; - pc.header.post_sync_operation = PIPE_CONTROL_NOWRITE; - - pc.header.instruction_state_cache_flush_enable = 1; - - pc.bits1.dest_addr_type = PIPE_CONTROL_GTTWRITE_GLOBAL; - - BRW_BATCH_STRUCT(brw, &pc); -} - -const struct brw_tracked_state brw_pipe_control = { - .dirty = { - .brw = BRW_NEW_SCENE, - .cache = 0 - }, - .update = upload_pipe_control -}; - - -/*********************************************************************** - * Misc invarient state packets - */ - -static void upload_invarient_state( struct brw_context *brw ) -{ - { - struct brw_mi_flush flush; - - memset(&flush, 0, sizeof(flush)); - flush.opcode = CMD_MI_FLUSH; - flush.flags = BRW_FLUSH_STATE_CACHE | BRW_FLUSH_READ_CACHE; - BRW_BATCH_STRUCT(brw, &flush); - } - - { - /* 0x61040000 Pipeline Select */ - /* PipelineSelect : 0 */ - struct brw_pipeline_select ps; - - memset(&ps, 0, sizeof(ps)); - ps.header.opcode = CMD_PIPELINE_SELECT; - ps.header.pipeline_select = 0; - BRW_BATCH_STRUCT(brw, &ps); - } - - { - struct brw_global_depth_offset_clamp gdo; - memset(&gdo, 0, sizeof(gdo)); - - /* Disable depth offset clamping. - */ - gdo.header.opcode = CMD_GLOBAL_DEPTH_OFFSET_CLAMP; - gdo.header.length = sizeof(gdo)/4 - 2; - gdo.depth_offset_clamp = 0.0; - - BRW_BATCH_STRUCT(brw, &gdo); - } - - - /* 0x61020000 State Instruction Pointer */ - { - struct brw_system_instruction_pointer sip; - memset(&sip, 0, sizeof(sip)); - - sip.header.opcode = CMD_STATE_INSN_POINTER; - sip.header.length = 0; - sip.bits0.pad = 0; - sip.bits0.system_instruction_pointer = 0; - BRW_BATCH_STRUCT(brw, &sip); - } - - - { - struct brw_vf_statistics vfs; - memset(&vfs, 0, sizeof(vfs)); - - vfs.opcode = CMD_VF_STATISTICS; - if (BRW_DEBUG & DEBUG_STATS) - vfs.statistics_enable = 1; - - BRW_BATCH_STRUCT(brw, &vfs); - } - - - { - struct brw_polygon_stipple_offset bpso; - - memset(&bpso, 0, sizeof(bpso)); - bpso.header.opcode = CMD_POLY_STIPPLE_OFFSET; - bpso.header.length = sizeof(bpso)/4-2; - bpso.bits0.x_offset = 0; - bpso.bits0.y_offset = 0; - - BRW_BATCH_STRUCT(brw, &bpso); - } -} - -const struct brw_tracked_state brw_invarient_state = { - .dirty = { - .brw = BRW_NEW_SCENE, - .cache = 0 - }, - .update = upload_invarient_state -}; - -/** - * Define the base addresses which some state is referenced from. - * - * This allows us to avoid having to emit relocations in many places for - * cached state, and instead emit pointers inside of large, mostly-static - * state pools. This comes at the expense of memory, and more expensive cache - * misses. - */ -static void upload_state_base_address( struct brw_context *brw ) -{ - /* Output the structure (brw_state_base_address) directly to the - * batchbuffer, so we can emit relocations inline. - */ - BEGIN_BATCH(6, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (6 - 2)); - OUT_RELOC(brw->pool[BRW_GS_POOL].buffer, - PIPE_BUFFER_USAGE_GPU_READ, - 1); /* General state base address */ - OUT_RELOC(brw->pool[BRW_SS_POOL].buffer, - PIPE_BUFFER_USAGE_GPU_READ, - 1); /* Surface state base address */ - OUT_BATCH(1); /* Indirect object base address */ - OUT_BATCH(1); /* General state upper bound */ - OUT_BATCH(1); /* Indirect object upper bound */ - ADVANCE_BATCH(); -} - - -const struct brw_tracked_state brw_state_base_address = { - .dirty = { - .brw = BRW_NEW_SCENE, - .cache = 0 - }, - .update = upload_state_base_address -}; diff --git a/src/mesa/pipe/i965simple/brw_reg.h b/src/mesa/pipe/i965simple/brw_reg.h deleted file mode 100644 index 9e885c3b3b..0000000000 --- a/src/mesa/pipe/i965simple/brw_reg.h +++ /dev/null @@ -1,76 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#define CMD_MI (0x0 << 29) -#define CMD_2D (0x2 << 29) -#define CMD_3D (0x3 << 29) - -#define MI_BATCH_BUFFER_END (CMD_MI | 0xA << 23) - -/* Stalls command execution waiting for the given events to have occurred. */ -#define MI_WAIT_FOR_EVENT (CMD_MI | (0x3 << 23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) - -/* Primitive dispatch on 830-945 */ -#define _3DPRIMITIVE (CMD_3D | (0x1f << 24)) -#define PRIM_INDIRECT (1<<23) -#define PRIM_INLINE (0<<23) -#define PRIM_INDIRECT_SEQUENTIAL (0<<17) -#define PRIM_INDIRECT_ELTS (1<<17) - -#define PRIM3D_TRILIST (0x0<<18) -#define PRIM3D_TRISTRIP (0x1<<18) -#define PRIM3D_TRISTRIP_RVRSE (0x2<<18) -#define PRIM3D_TRIFAN (0x3<<18) -#define PRIM3D_POLY (0x4<<18) -#define PRIM3D_LINELIST (0x5<<18) -#define PRIM3D_LINESTRIP (0x6<<18) -#define PRIM3D_RECTLIST (0x7<<18) -#define PRIM3D_POINTLIST (0x8<<18) -#define PRIM3D_DIB (0x9<<18) -#define PRIM3D_MASK (0x1f<<18) - -#define XY_SETUP_BLT_CMD (CMD_2D | (0x01 << 22) | 6) - -#define XY_COLOR_BLT_CMD (CMD_2D | (0x50 << 22) | 4) - -#define XY_SRC_COPY_BLT_CMD (CMD_2D | (0x53 << 22) | 6) - -/* BR00 */ -#define XY_BLT_WRITE_ALPHA (1 << 21) -#define XY_BLT_WRITE_RGB (1 << 20) -#define XY_SRC_TILED (1 << 15) -#define XY_DST_TILED (1 << 11) - -/* BR13 */ -#define BR13_565 (0x1 << 24) -#define BR13_8888 (0x3 << 24) - -#define FENCE_LINEAR 0 -#define FENCE_XMAJOR 1 -#define FENCE_YMAJOR 2 diff --git a/src/mesa/pipe/i965simple/brw_sf.c b/src/mesa/pipe/i965simple/brw_sf.c deleted file mode 100644 index 7c83b81c85..0000000000 --- a/src/mesa/pipe/i965simple/brw_sf.c +++ /dev/null @@ -1,351 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_sf.h" -#include "brw_state.h" -#include "tgsi/util/tgsi_parse.h" - - -static void compile_sf_prog( struct brw_context *brw, - struct brw_sf_prog_key *key ) -{ - struct brw_sf_compile c; - const unsigned *program; - unsigned program_size; - - memset(&c, 0, sizeof(c)); - - /* Begin the compilation: - */ - brw_init_compile(&c.func); - - c.key = *key; - - - c.nr_attrs = c.key.vp_output_count; - c.nr_attr_regs = (c.nr_attrs+1)/2; - - c.nr_setup_attrs = c.key.fp_input_count + 1; /* +1 for position */ - c.nr_setup_regs = (c.nr_setup_attrs+1)/2; - - c.prog_data.urb_read_length = c.nr_attr_regs; - c.prog_data.urb_entry_size = c.nr_setup_regs * 2; - - - /* Which primitive? Or all three? - */ - switch (key->primitive) { - case SF_TRIANGLES: - c.nr_verts = 3; - brw_emit_tri_setup( &c ); - break; - case SF_LINES: - c.nr_verts = 2; - brw_emit_line_setup( &c ); - break; - case SF_POINTS: - c.nr_verts = 1; - brw_emit_point_setup( &c ); - break; - - case SF_UNFILLED_TRIS: - default: - assert(0); - return; - } - - - - /* get the program - */ - program = brw_get_program(&c.func, &program_size); - - /* Upload - */ - brw->sf.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_SF_PROG], - &c.key, - sizeof(c.key), - program, - program_size, - &c.prog_data, - &brw->sf.prog_data ); -} - - -static boolean search_cache( struct brw_context *brw, - struct brw_sf_prog_key *key ) -{ - return brw_search_cache(&brw->cache[BRW_SF_PROG], - key, sizeof(*key), - &brw->sf.prog_data, - &brw->sf.prog_gs_offset); -} - - -/* Calculate interpolants for triangle and line rasterization. - */ -static void upload_sf_prog( struct brw_context *brw ) -{ - const struct brw_fragment_program *fs = brw->attribs.FragmentProgram; - struct brw_sf_prog_key key; - struct tgsi_parse_context parse; - int i, done = 0; - - - memset(&key, 0, sizeof(key)); - - /* Populate the key, noting state dependencies: - */ - /* CACHE_NEW_VS_PROG */ - key.vp_output_count = brw->vs.prog_data->outputs_written; - - /* BRW_NEW_FS */ - key.fp_input_count = brw->attribs.FragmentProgram->info.nr_regs[TGSI_FILE_INPUT]; - - - /* BRW_NEW_REDUCED_PRIMITIVE */ - switch (brw->reduced_primitive) { - case PIPE_PRIM_TRIANGLES: -// if (key.attrs & (1<program.tokens ); - while( !done && - !tgsi_parse_end_of_tokens( &parse ) ) - { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) - { - int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; - int interp_mode = parse.FullToken.FullDeclaration.Interpolation.Interpolate; - //int semantic = parse.FullToken.FullDeclaration.Semantic.SemanticName; - //int semantic_index = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - - debug_printf("fs input %d..%d interp mode %d\n", first, last, interp_mode); - - switch (interp_mode) { - case TGSI_INTERPOLATE_CONSTANT: - for (i = first; i <= last; i++) - key.const_mask |= (1 << i); - break; - case TGSI_INTERPOLATE_LINEAR: - for (i = first; i <= last; i++) - key.linear_mask |= (1 << i); - break; - case TGSI_INTERPOLATE_PERSPECTIVE: - for (i = first; i <= last; i++) - key.persp_mask |= (1 << i); - break; - default: - break; - } - - /* Also need stuff for flat shading, twosided color. - */ - - } - break; - default: - done = 1; - break; - } - } - - /* Hack: Adjust for position. Optimize away when not required (ie - * for perspective interpolation). - */ - key.persp_mask <<= 1; - key.linear_mask <<= 1; - key.linear_mask |= 1; - key.const_mask <<= 1; - - debug_printf("key.persp_mask: %x\n", key.persp_mask); - debug_printf("key.linear_mask: %x\n", key.linear_mask); - debug_printf("key.const_mask: %x\n", key.const_mask); - - -// key.do_point_sprite = brw->attribs.Point->PointSprite; -// key.SpriteOrigin = brw->attribs.Point->SpriteOrigin; - -// key.do_flat_shading = (brw->attribs.Raster->flatshade); -// key.do_twoside_color = (brw->attribs.Light->Enabled && brw->attribs.Light->Model.TwoSide); - -// if (key.do_twoside_color) -// key.frontface_ccw = (brw->attribs.Polygon->FrontFace == GL_CCW); - - - if (!search_cache(brw, &key)) - compile_sf_prog( brw, &key ); -} - - -const struct brw_tracked_state brw_sf_prog = { - .dirty = { - .brw = (BRW_NEW_RASTERIZER | - BRW_NEW_REDUCED_PRIMITIVE | - BRW_NEW_VS | - BRW_NEW_FS), - .cache = 0, - }, - .update = upload_sf_prog -}; - - - -#if 0 -/* Build a struct like the one we'd like the state tracker to pass to - * us. - */ -static void update_sf_linkage( struct brw_context *brw ) -{ - const struct brw_vertex_program *vs = brw->attribs.VertexProgram; - const struct brw_fragment_program *fs = brw->attribs.FragmentProgram; - struct pipe_setup_linkage state; - struct tgsi_parse_context parse; - - int i, j; - int nr_vp_outputs = 0; - int done = 0; - - struct { - unsigned semantic:8; - unsigned semantic_index:16; - } fp_semantic[32], vp_semantic[32]; - - memset(&state, 0, sizeof(state)); - - state.fp_input_count = 0; - - - - - - - assert(state.fp_input_count == fs->program.num_inputs); - - - /* Then scan vp outputs - */ - done = 0; - tgsi_parse_init( &parse, vs->program.tokens ); - while( !done && - !tgsi_parse_end_of_tokens( &parse ) ) - { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) - { - int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; - - for (i = first; i < last; i++) { - vp_semantic[i].semantic = - parse.FullToken.FullDeclaration.Semantic.SemanticName; - vp_semantic[i].semantic_index = - parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - } - - assert(last > nr_vp_outputs); - nr_vp_outputs = last; - } - break; - default: - done = 1; - break; - } - } - - - /* Now match based on semantic information. - */ - for (i = 0; i< state.fp_input_count; i++) { - for (j = 0; j < nr_vp_outputs; j++) { - if (fp_semantic[i].semantic == vp_semantic[j].semantic && - fp_semantic[i].semantic_index == vp_semantic[j].semantic_index) { - state.fp_input[i].vp_output = j; - } - } - if (fp_semantic[i].semantic == TGSI_SEMANTIC_COLOR) { - for (j = 0; j < nr_vp_outputs; j++) { - if (TGSI_SEMANTIC_BCOLOR == vp_semantic[j].semantic && - fp_semantic[i].semantic_index == vp_semantic[j].semantic_index) { - state.fp_input[i].bf_vp_output = j; - } - } - } - } - - if (memcmp(&brw->sf.linkage, &state, sizeof(state)) != 0) { - brw->sf.linkage = state; - brw->state.dirty.brw |= BRW_NEW_SF_LINKAGE; - } -} - - -const struct brw_tracked_state brw_sf_linkage = { - .dirty = { - .brw = (BRW_NEW_VS | - BRW_NEW_FS), - .cache = 0, - }, - .update = update_sf_linkage -}; - - -#endif diff --git a/src/mesa/pipe/i965simple/brw_sf.h b/src/mesa/pipe/i965simple/brw_sf.h deleted file mode 100644 index b7ada47560..0000000000 --- a/src/mesa/pipe/i965simple/brw_sf.h +++ /dev/null @@ -1,122 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRW_SF_H -#define BRW_SF_H - -#include "brw_context.h" -#include "brw_eu.h" - - -#define SF_POINTS 0 -#define SF_LINES 1 -#define SF_TRIANGLES 2 -#define SF_UNFILLED_TRIS 3 - - - -struct brw_sf_prog_key { - unsigned vp_output_count:5; - unsigned fp_input_count:5; - - unsigned primitive:2; - unsigned do_twoside_color:1; - unsigned do_flat_shading:1; - unsigned frontface_ccw:1; - unsigned do_point_sprite:1; - - /* Interpolation masks; - */ - unsigned linear_mask; - unsigned persp_mask; - unsigned const_mask; - - -// int SpriteOrigin; -}; - -struct brw_sf_point_tex { - boolean CoordReplace; -}; - -struct brw_sf_compile { - struct brw_compile func; - struct brw_sf_prog_key key; - struct brw_sf_prog_data prog_data; - - struct brw_reg pv; - struct brw_reg det; - struct brw_reg dx0; - struct brw_reg dx2; - struct brw_reg dy0; - struct brw_reg dy2; - - /* z and 1/w passed in seperately: - */ - struct brw_reg z[3]; - struct brw_reg inv_w[3]; - - /* The vertices: - */ - struct brw_reg vert[3]; - - /* Temporaries, allocated after last vertex reg. - */ - struct brw_reg inv_det; - struct brw_reg a1_sub_a0; - struct brw_reg a2_sub_a0; - struct brw_reg tmp; - - struct brw_reg m1Cx; - struct brw_reg m2Cy; - struct brw_reg m3C0; - - unsigned nr_verts; - unsigned nr_attrs; - unsigned nr_attr_regs; - unsigned nr_setup_attrs; - unsigned nr_setup_regs; -#if 0 - ubyte attr_to_idx[VERT_RESULT_MAX]; - ubyte idx_to_attr[VERT_RESULT_MAX]; - struct brw_sf_point_tex point_attrs[VERT_RESULT_MAX]; -#endif -}; - - -void brw_emit_tri_setup( struct brw_sf_compile *c ); -void brw_emit_line_setup( struct brw_sf_compile *c ); -void brw_emit_point_setup( struct brw_sf_compile *c ); -void brw_emit_point_sprite_setup( struct brw_sf_compile *c ); -void brw_emit_anyprim_setup( struct brw_sf_compile *c ); - -#endif diff --git a/src/mesa/pipe/i965simple/brw_sf_emit.c b/src/mesa/pipe/i965simple/brw_sf_emit.c deleted file mode 100644 index 78d6fa5e9e..0000000000 --- a/src/mesa/pipe/i965simple/brw_sf_emit.c +++ /dev/null @@ -1,382 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_defines.h" -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_util.h" -#include "brw_sf.h" - - - -/*********************************************************************** - * Triangle setup. - */ - - -static void alloc_regs( struct brw_sf_compile *c ) -{ - unsigned reg, i; - - /* Values computed by fixed function unit: - */ - c->pv = retype(brw_vec1_grf(1, 1), BRW_REGISTER_TYPE_UD); - c->det = brw_vec1_grf(1, 2); - c->dx0 = brw_vec1_grf(1, 3); - c->dx2 = brw_vec1_grf(1, 4); - c->dy0 = brw_vec1_grf(1, 5); - c->dy2 = brw_vec1_grf(1, 6); - - /* z and 1/w passed in seperately: - */ - c->z[0] = brw_vec1_grf(2, 0); - c->inv_w[0] = brw_vec1_grf(2, 1); - c->z[1] = brw_vec1_grf(2, 2); - c->inv_w[1] = brw_vec1_grf(2, 3); - c->z[2] = brw_vec1_grf(2, 4); - c->inv_w[2] = brw_vec1_grf(2, 5); - - /* The vertices: - */ - reg = 3; - for (i = 0; i < c->nr_verts; i++) { - c->vert[i] = brw_vec8_grf(reg, 0); - reg += c->nr_attr_regs; - } - - /* Temporaries, allocated after last vertex reg. - */ - c->inv_det = brw_vec1_grf(reg, 0); reg++; - c->a1_sub_a0 = brw_vec8_grf(reg, 0); reg++; - c->a2_sub_a0 = brw_vec8_grf(reg, 0); reg++; - c->tmp = brw_vec8_grf(reg, 0); reg++; - - /* Note grf allocation: - */ - c->prog_data.total_grf = reg; - - - /* Outputs of this program - interpolation coefficients for - * rasterization: - */ - c->m1Cx = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 1, 0); - c->m2Cy = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 2, 0); - c->m3C0 = brw_vec8_reg(BRW_MESSAGE_REGISTER_FILE, 3, 0); -} - - -static void copy_z_inv_w( struct brw_sf_compile *c ) -{ - struct brw_compile *p = &c->func; - unsigned i; - - brw_push_insn_state(p); - - /* Copy both scalars with a single MOV: - */ - for (i = 0; i < c->nr_verts; i++) - brw_MOV(p, vec2(suboffset(c->vert[i], 2)), vec2(c->z[i])); - - brw_pop_insn_state(p); -} - - -static void invert_det( struct brw_sf_compile *c) -{ - brw_math(&c->func, - c->inv_det, - BRW_MATH_FUNCTION_INV, - BRW_MATH_SATURATE_NONE, - 0, - c->det, - BRW_MATH_DATA_SCALAR, - BRW_MATH_PRECISION_FULL); - -} - -#define NON_PERPECTIVE_ATTRS (FRAG_BIT_WPOS | \ - FRAG_BIT_COL0 | \ - FRAG_BIT_COL1) - -static boolean calculate_masks( struct brw_sf_compile *c, - unsigned reg, - ushort *pc, - ushort *pc_persp, - ushort *pc_linear) -{ - boolean is_last_attr = (reg == c->nr_setup_regs - 1); - unsigned persp_mask = c->key.persp_mask; - unsigned linear_mask = c->key.linear_mask; - - debug_printf("persp_mask: %x\n", persp_mask); - debug_printf("linear_mask: %x\n", linear_mask); - - *pc_persp = 0; - *pc_linear = 0; - *pc = 0xf; - - if (persp_mask & (1 << (reg*2))) - *pc_persp = 0xf; - - if (linear_mask & (1 << (reg*2))) - *pc_linear = 0xf; - - /* Maybe only processs one attribute on the final round: - */ - if (reg*2+1 < c->nr_setup_attrs) { - *pc |= 0xf0; - - if (persp_mask & (1 << (reg*2+1))) - *pc_persp |= 0xf0; - - if (linear_mask & (1 << (reg*2+1))) - *pc_linear |= 0xf0; - } - - debug_printf("pc: %x\n", *pc); - debug_printf("pc_persp: %x\n", *pc_persp); - debug_printf("pc_linear: %x\n", *pc_linear); - - - return is_last_attr; -} - - - -void brw_emit_tri_setup( struct brw_sf_compile *c ) -{ - struct brw_compile *p = &c->func; - unsigned i; - - debug_printf("%s START ==============\n", __FUNCTION__); - - c->nr_verts = 3; - alloc_regs(c); - invert_det(c); - copy_z_inv_w(c); - - - for (i = 0; i < c->nr_setup_regs; i++) - { - /* Pair of incoming attributes: - */ - struct brw_reg a0 = offset(c->vert[0], i); - struct brw_reg a1 = offset(c->vert[1], i); - struct brw_reg a2 = offset(c->vert[2], i); - ushort pc = 0, pc_persp = 0, pc_linear = 0; - boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); - - if (pc_persp) - { - brw_set_predicate_control_flag_value(p, pc_persp); - brw_MUL(p, a0, a0, c->inv_w[0]); - brw_MUL(p, a1, a1, c->inv_w[1]); - brw_MUL(p, a2, a2, c->inv_w[2]); - } - - - /* Calculate coefficients for interpolated values: - */ - if (pc_linear) - { - brw_set_predicate_control_flag_value(p, pc_linear); - - brw_ADD(p, c->a1_sub_a0, a1, negate(a0)); - brw_ADD(p, c->a2_sub_a0, a2, negate(a0)); - - /* calculate dA/dx - */ - brw_MUL(p, brw_null_reg(), c->a1_sub_a0, c->dy2); - brw_MAC(p, c->tmp, c->a2_sub_a0, negate(c->dy0)); - brw_MUL(p, c->m1Cx, c->tmp, c->inv_det); - - /* calculate dA/dy - */ - brw_MUL(p, brw_null_reg(), c->a2_sub_a0, c->dx0); - brw_MAC(p, c->tmp, c->a1_sub_a0, negate(c->dx2)); - brw_MUL(p, c->m2Cy, c->tmp, c->inv_det); - } - - { - brw_set_predicate_control_flag_value(p, pc); - /* start point for interpolation - */ - brw_MOV(p, c->m3C0, a0); - - /* Copy m0..m3 to URB. m0 is implicitly copied from r0 in - * the send instruction: - */ - brw_urb_WRITE(p, - brw_null_reg(), - 0, - brw_vec8_grf(0, 0), /* r0, will be copied to m0 */ - 0, /* allocate */ - 1, /* used */ - 4, /* msg len */ - 0, /* response len */ - last, /* eot */ - last, /* writes complete */ - i*4, /* offset */ - BRW_URB_SWIZZLE_TRANSPOSE); /* XXX: Swizzle control "SF to windower" */ - } - } - - debug_printf("%s DONE ==============\n", __FUNCTION__); - -} - - - -void brw_emit_line_setup( struct brw_sf_compile *c ) -{ - struct brw_compile *p = &c->func; - unsigned i; - - - c->nr_verts = 2; - alloc_regs(c); - invert_det(c); - copy_z_inv_w(c); - - for (i = 0; i < c->nr_setup_regs; i++) - { - /* Pair of incoming attributes: - */ - struct brw_reg a0 = offset(c->vert[0], i); - struct brw_reg a1 = offset(c->vert[1], i); - ushort pc, pc_persp, pc_linear; - boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); - - if (pc_persp) - { - brw_set_predicate_control_flag_value(p, pc_persp); - brw_MUL(p, a0, a0, c->inv_w[0]); - brw_MUL(p, a1, a1, c->inv_w[1]); - } - - /* Calculate coefficients for position, color: - */ - if (pc_linear) { - brw_set_predicate_control_flag_value(p, pc_linear); - - brw_ADD(p, c->a1_sub_a0, a1, negate(a0)); - - brw_MUL(p, c->tmp, c->a1_sub_a0, c->dx0); - brw_MUL(p, c->m1Cx, c->tmp, c->inv_det); - - brw_MUL(p, c->tmp, c->a1_sub_a0, c->dy0); - brw_MUL(p, c->m2Cy, c->tmp, c->inv_det); - } - - { - brw_set_predicate_control_flag_value(p, pc); - - /* start point for interpolation - */ - brw_MOV(p, c->m3C0, a0); - - /* Copy m0..m3 to URB. - */ - brw_urb_WRITE(p, - brw_null_reg(), - 0, - brw_vec8_grf(0, 0), - 0, /* allocate */ - 1, /* used */ - 4, /* msg len */ - 0, /* response len */ - last, /* eot */ - last, /* writes complete */ - i*4, /* urb destination offset */ - BRW_URB_SWIZZLE_TRANSPOSE); - } - } -} - - -/* Points setup - several simplifications as all attributes are - * constant across the face of the point (point sprites excluded!) - */ -void brw_emit_point_setup( struct brw_sf_compile *c ) -{ - struct brw_compile *p = &c->func; - unsigned i; - - c->nr_verts = 1; - alloc_regs(c); - copy_z_inv_w(c); - - brw_MOV(p, c->m1Cx, brw_imm_ud(0)); /* zero - move out of loop */ - brw_MOV(p, c->m2Cy, brw_imm_ud(0)); /* zero - move out of loop */ - - for (i = 0; i < c->nr_setup_regs; i++) - { - struct brw_reg a0 = offset(c->vert[0], i); - ushort pc, pc_persp, pc_linear; - boolean last = calculate_masks(c, i, &pc, &pc_persp, &pc_linear); - - if (pc_persp) - { - /* This seems odd as the values are all constant, but the - * fragment shader will be expecting it: - */ - brw_set_predicate_control_flag_value(p, pc_persp); - brw_MUL(p, a0, a0, c->inv_w[0]); - } - - - /* The delta values are always zero, just send the starting - * coordinate. Again, this is to fit in with the interpolation - * code in the fragment shader. - */ - { - brw_set_predicate_control_flag_value(p, pc); - - brw_MOV(p, c->m3C0, a0); /* constant value */ - - /* Copy m0..m3 to URB. - */ - brw_urb_WRITE(p, - brw_null_reg(), - 0, - brw_vec8_grf(0, 0), - 0, /* allocate */ - 1, /* used */ - 4, /* msg len */ - 0, /* response len */ - last, /* eot */ - last, /* writes complete */ - i*4, /* urb destination offset */ - BRW_URB_SWIZZLE_TRANSPOSE); - } - } -} diff --git a/src/mesa/pipe/i965simple/brw_sf_state.c b/src/mesa/pipe/i965simple/brw_sf_state.c deleted file mode 100644 index 9acd3ea61b..0000000000 --- a/src/mesa/pipe/i965simple/brw_sf_state.c +++ /dev/null @@ -1,180 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - - -#include "brw_context.h" -#include "brw_state.h" -#include "brw_defines.h" -#include "pipe/p_util.h" - -static void upload_sf_vp(struct brw_context *brw) -{ - struct brw_sf_viewport sfv; - - memset(&sfv, 0, sizeof(sfv)); - - - /* BRW_NEW_VIEWPORT */ - { - const float *scale = brw->attribs.Viewport.scale; - const float *trans = brw->attribs.Viewport.translate; - - sfv.viewport.m00 = scale[0]; - sfv.viewport.m11 = scale[1]; - sfv.viewport.m22 = scale[2]; - sfv.viewport.m30 = trans[0]; - sfv.viewport.m31 = trans[1]; - sfv.viewport.m32 = trans[2]; - } - - /* _NEW_SCISSOR */ - sfv.scissor.xmin = brw->attribs.Scissor.minx; - sfv.scissor.xmax = brw->attribs.Scissor.maxx - 1; - sfv.scissor.ymin = brw->attribs.Scissor.miny; - sfv.scissor.ymax = brw->attribs.Scissor.maxy - 1; - - brw->sf.vp_gs_offset = brw_cache_data( &brw->cache[BRW_SF_VP], &sfv ); -} - -const struct brw_tracked_state brw_sf_vp = { - .dirty = { - .brw = (BRW_NEW_SCISSOR | - BRW_NEW_VIEWPORT), - .cache = 0 - }, - .update = upload_sf_vp -}; - -static void upload_sf_unit( struct brw_context *brw ) -{ - struct brw_sf_unit_state sf; - memset(&sf, 0, sizeof(sf)); - - /* CACHE_NEW_SF_PROG */ - sf.thread0.grf_reg_count = align(brw->sf.prog_data->total_grf, 16) / 16 - 1; - sf.thread0.kernel_start_pointer = brw->sf.prog_gs_offset >> 6; - sf.thread3.urb_entry_read_length = brw->sf.prog_data->urb_read_length; - - sf.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; - sf.thread3.dispatch_grf_start_reg = 3; - sf.thread3.urb_entry_read_offset = 1; - - /* BRW_NEW_URB_FENCE */ - sf.thread4.nr_urb_entries = brw->urb.nr_sf_entries; - sf.thread4.urb_entry_allocation_size = brw->urb.sfsize - 1; - sf.thread4.max_threads = MIN2(12, brw->urb.nr_sf_entries / 2) - 1; - - if (BRW_DEBUG & DEBUG_SINGLE_THREAD) - sf.thread4.max_threads = 0; - - if (BRW_DEBUG & DEBUG_STATS) - sf.thread4.stats_enable = 1; - - /* CACHE_NEW_SF_VP */ - sf.sf5.sf_viewport_state_offset = brw->sf.vp_gs_offset >> 5; - sf.sf5.viewport_transform = 1; - - /* BRW_NEW_RASTER */ - if (brw->attribs.Raster->scissor) - sf.sf6.scissor = 1; - -#if 0 - if (brw->attribs.Polygon->FrontFace == GL_CCW) - sf.sf5.front_winding = BRW_FRONTWINDING_CCW; - else - sf.sf5.front_winding = BRW_FRONTWINDING_CW; - - - if (brw->attribs.Polygon->CullFlag) { - switch (brw->attribs.Polygon->CullFaceMode) { - case GL_FRONT: - sf.sf6.cull_mode = BRW_CULLMODE_FRONT; - break; - case GL_BACK: - sf.sf6.cull_mode = BRW_CULLMODE_BACK; - break; - case GL_FRONT_AND_BACK: - sf.sf6.cull_mode = BRW_CULLMODE_BOTH; - break; - default: - assert(0); - break; - } - } - else - sf.sf6.cull_mode = BRW_CULLMODE_NONE; -#else - sf.sf5.front_winding = BRW_FRONTWINDING_CCW; - sf.sf6.cull_mode = BRW_CULLMODE_NONE; -#endif - - sf.sf6.line_width = CLAMP(brw->attribs.Raster->line_width, 1.0, 5.0) * (1<<1); - - sf.sf6.line_endcap_aa_region_width = 1; - if (brw->attribs.Raster->line_smooth) - sf.sf6.aa_enable = 1; - else if (sf.sf6.line_width <= 0x2) - sf.sf6.line_width = 0; - - sf.sf6.point_rast_rule = 1; /* opengl conventions */ - - sf.sf7.sprite_point = brw->attribs.Raster->point_sprite; - sf.sf7.point_size = CLAMP(brw->attribs.Raster->line_width, 1.0, 255.0) * (1<<3); - sf.sf7.use_point_size_state = !brw->attribs.Raster->point_size_per_vertex; - - /* might be BRW_NEW_PRIMITIVE if we have to adjust pv for polygons: - */ - sf.sf7.trifan_pv = 2; - sf.sf7.linestrip_pv = 1; - sf.sf7.tristrip_pv = 2; - sf.sf7.line_last_pixel_enable = 0; - - /* Set bias for OpenGL rasterization rules: - */ - sf.sf6.dest_org_vbias = 0x8; - sf.sf6.dest_org_hbias = 0x8; - - brw->sf.state_gs_offset = brw_cache_data( &brw->cache[BRW_SF_UNIT], &sf ); -} - - -const struct brw_tracked_state brw_sf_unit = { - .dirty = { - .brw = (BRW_NEW_RASTERIZER | - BRW_NEW_URB_FENCE), - .cache = (CACHE_NEW_SF_VP | - CACHE_NEW_SF_PROG) - }, - .update = upload_sf_unit -}; - - diff --git a/src/mesa/pipe/i965simple/brw_shader_info.c b/src/mesa/pipe/i965simple/brw_shader_info.c deleted file mode 100644 index 431b45466a..0000000000 --- a/src/mesa/pipe/i965simple/brw_shader_info.c +++ /dev/null @@ -1,49 +0,0 @@ - -#include "brw_context.h" -#include "brw_state.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" - - - - -void brw_shader_info(const struct tgsi_token *tokens, - struct brw_shader_info *info ) -{ - struct tgsi_parse_context parse; - int done = 0; - - tgsi_parse_init( &parse, tokens ); - - while( !done && - !tgsi_parse_end_of_tokens( &parse ) ) - { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - { - const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; - unsigned last = decl->u.DeclarationRange.Last; - - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - // Broken by crazy wpos init: - //assert( info->nr_regs[decl->Declaration.File] <= last); - - info->nr_regs[decl->Declaration.File] = MAX2(info->nr_regs[decl->Declaration.File], - last+1); - break; - } - case TGSI_TOKEN_TYPE_IMMEDIATE: - case TGSI_TOKEN_TYPE_INSTRUCTION: - default: - done = 1; - break; - } - } - - tgsi_parse_free (&parse); - -} diff --git a/src/mesa/pipe/i965simple/brw_state.c b/src/mesa/pipe/i965simple/brw_state.c deleted file mode 100644 index 95dfce88e4..0000000000 --- a/src/mesa/pipe/i965simple/brw_state.c +++ /dev/null @@ -1,424 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Zack Rusin - * Keith Whitwell - */ - - -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_dump.h" - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_state.h" -#include "brw_draw.h" - - -#define DUP( TYPE, VAL ) \ -do { \ - struct TYPE *x = malloc(sizeof(*x)); \ - memcpy(x, VAL, sizeof(*x) ); \ - return x; \ -} while (0) - -/************************************************************************ - * Blend - */ -static void * -brw_create_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend) -{ - DUP( pipe_blend_state, blend ); -} - -static void brw_bind_blend_state(struct pipe_context *pipe, - void *blend) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Blend = (struct pipe_blend_state*)blend; - brw->state.dirty.brw |= BRW_NEW_BLEND; -} - - -static void brw_delete_blend_state(struct pipe_context *pipe, void *blend) -{ - free(blend); -} - -static void brw_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.BlendColor = *blend_color; - - brw->state.dirty.brw |= BRW_NEW_BLEND; -} - -/************************************************************************ - * Sampler - */ - -static void * -brw_create_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) -{ - DUP( pipe_sampler_state, sampler ); -} - -static void brw_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Samplers[unit] = sampler; - brw->state.dirty.brw |= BRW_NEW_SAMPLER; -} - -static void brw_delete_sampler_state(struct pipe_context *pipe, - void *sampler) -{ - free(sampler); -} - - -/************************************************************************ - * Depth stencil - */ - -static void * -brw_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) -{ - DUP( pipe_depth_stencil_alpha_state, depth_stencil ); -} - -static void brw_bind_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; - - brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; -} - -static void brw_delete_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - free(depth_stencil); -} - -/************************************************************************ - * Scissor - */ -static void brw_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct brw_context *brw = brw_context(pipe); - - memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) ); - brw->state.dirty.brw |= BRW_NEW_SCISSOR; -} - - -/************************************************************************ - * Stipple - */ - -static void brw_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ -} - - -/************************************************************************ - * Fragment shader - */ - -static void * brw_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *shader) -{ - struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); - - /* XXX: Do I have to duplicate the tokens as well?? - */ - brw_fp->program = *shader; - brw_fp->id = brw_context(pipe)->program_id++; - - brw_shader_info(shader->tokens, - &brw_fp->info); - - tgsi_dump(shader->tokens, 0); - - - return (void *)brw_fp; -} - -static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader; - brw->state.dirty.brw |= BRW_NEW_FS; -} - -static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) -{ - FREE(shader); -} - - -/************************************************************************ - * Vertex shader and other TNL state - */ - -static void *brw_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *shader) -{ - struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); - - /* XXX: Do I have to duplicate the tokens as well?? - */ - brw_vp->program = *shader; - brw_vp->id = brw_context(pipe)->program_id++; - brw_shader_info(shader->tokens, - &brw_vp->info); - - tgsi_dump(shader->tokens, 0); - - return (void *)brw_vp; -} - -static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; - brw->state.dirty.brw |= BRW_NEW_VS; - - debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); -} - -static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) -{ - FREE(shader); -} - - -static void brw_set_clip_state( struct pipe_context *pipe, - const struct pipe_clip_state *clip ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Clip = *clip; -} - - -static void brw_set_viewport_state( struct pipe_context *pipe, - const struct pipe_viewport_state *viewport ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Viewport = *viewport; /* struct copy */ - brw->state.dirty.brw |= BRW_NEW_VIEWPORT; - - /* pass the viewport info to the draw module */ - //draw_set_viewport_state(brw->draw, viewport); -} - - -static void brw_set_vertex_buffer( struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer ) -{ - struct brw_context *brw = brw_context(pipe); - brw->vb.vbo_array[index] = buffer; -} - -static void brw_set_vertex_element(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_element *element) -{ - /* flush ? */ - struct brw_context *brw = brw_context(pipe); - - assert(index < PIPE_ATTRIB_MAX); - struct brw_vertex_element_state el; - memset(&el, 0, sizeof(el)); - - el.ve0.src_offset = element->src_offset; - el.ve0.src_format = brw_translate_surface_format(element->src_format); - el.ve0.valid = 1; - el.ve0.vertex_buffer_index = element->vertex_buffer_index; - - el.ve1.dst_offset = index * 4; - - el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; - - switch (element->nr_components) { - case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; - case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; - case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; - break; - } - - brw->vb.inputs[index] = el; -} - - - -/************************************************************************ - * Constant buffers - */ - -static void brw_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) -{ - struct brw_context *brw = brw_context(pipe); - - assert(buf == 0 || index == 0); - - brw->attribs.Constants[shader] = buf; - brw->state.dirty.brw |= BRW_NEW_CONSTANTS; -} - - -/************************************************************************ - * Texture surfaces - */ - - -static void brw_set_sampler_texture(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Texture[unit] = (struct brw_texture*)texture; /* ptr, not struct */ - - brw->state.dirty.brw |= BRW_NEW_TEXTURE; -} - - -/************************************************************************ - * Render targets, etc - */ - -static void brw_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.FrameBuffer = *fb; /* struct copy */ - - brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER; -} - - - -/************************************************************************ - * Rasterizer state - */ - -static void * -brw_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *rasterizer) -{ - DUP(pipe_rasterizer_state, rasterizer); -} - -static void brw_bind_rasterizer_state( struct pipe_context *pipe, - void *setup ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Raster = (struct pipe_rasterizer_state *)setup; - - /* Also pass-through to draw module: - */ - //draw_set_rasterizer_state(brw->draw, setup); - - brw->state.dirty.brw |= BRW_NEW_RASTERIZER; -} - -static void brw_delete_rasterizer_state(struct pipe_context *pipe, - void *setup) -{ - free(setup); -} - - - -void -brw_init_state_functions( struct brw_context *brw ) -{ - brw->pipe.create_blend_state = brw_create_blend_state; - brw->pipe.bind_blend_state = brw_bind_blend_state; - brw->pipe.delete_blend_state = brw_delete_blend_state; - - brw->pipe.create_sampler_state = brw_create_sampler_state; - brw->pipe.bind_sampler_state = brw_bind_sampler_state; - brw->pipe.delete_sampler_state = brw_delete_sampler_state; - - brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; - brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; - brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; - - brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; - brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; - brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state; - brw->pipe.create_fs_state = brw_create_fs_state; - brw->pipe.bind_fs_state = brw_bind_fs_state; - brw->pipe.delete_fs_state = brw_delete_fs_state; - brw->pipe.create_vs_state = brw_create_vs_state; - brw->pipe.bind_vs_state = brw_bind_vs_state; - brw->pipe.delete_vs_state = brw_delete_vs_state; - - brw->pipe.set_blend_color = brw_set_blend_color; - brw->pipe.set_clip_state = brw_set_clip_state; - brw->pipe.set_constant_buffer = brw_set_constant_buffer; - brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; - -// brw->pipe.set_feedback_state = brw_set_feedback_state; -// brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; - - brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; - brw->pipe.set_scissor_state = brw_set_scissor_state; - brw->pipe.set_sampler_texture = brw_set_sampler_texture; - brw->pipe.set_viewport_state = brw_set_viewport_state; - brw->pipe.set_vertex_buffer = brw_set_vertex_buffer; - brw->pipe.set_vertex_element = brw_set_vertex_element; -} diff --git a/src/mesa/pipe/i965simple/brw_state.h b/src/mesa/pipe/i965simple/brw_state.h deleted file mode 100644 index 258e9a556e..0000000000 --- a/src/mesa/pipe/i965simple/brw_state.h +++ /dev/null @@ -1,158 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRW_STATE_H -#define BRW_STATE_H - -#include "brw_context.h" -#include "brw_winsys.h" - - -const struct brw_tracked_state brw_blend_constant_color; -const struct brw_tracked_state brw_cc_unit; -const struct brw_tracked_state brw_cc_vp; -const struct brw_tracked_state brw_clip_prog; -const struct brw_tracked_state brw_clip_unit; -const struct brw_tracked_state brw_constant_buffer_state; -const struct brw_tracked_state brw_constant_buffer; -const struct brw_tracked_state brw_curbe_offsets; -const struct brw_tracked_state brw_invarient_state; -const struct brw_tracked_state brw_gs_prog; -const struct brw_tracked_state brw_gs_unit; -const struct brw_tracked_state brw_drawing_rect; -const struct brw_tracked_state brw_line_stipple; -const struct brw_tracked_state brw_pipelined_state_pointers; -const struct brw_tracked_state brw_binding_table_pointers; -const struct brw_tracked_state brw_depthbuffer; -const struct brw_tracked_state brw_polygon_stipple_offset; -const struct brw_tracked_state brw_polygon_stipple; -const struct brw_tracked_state brw_program_parameters; -const struct brw_tracked_state brw_recalculate_urb_fence; -const struct brw_tracked_state brw_sf_prog; -const struct brw_tracked_state brw_sf_unit; -const struct brw_tracked_state brw_sf_vp; -const struct brw_tracked_state brw_state_base_address; -const struct brw_tracked_state brw_urb_fence; -const struct brw_tracked_state brw_vertex_state; -const struct brw_tracked_state brw_vs_prog; -const struct brw_tracked_state brw_vs_unit; -const struct brw_tracked_state brw_wm_prog; -const struct brw_tracked_state brw_wm_samplers; -const struct brw_tracked_state brw_wm_surfaces; -const struct brw_tracked_state brw_wm_unit; - -const struct brw_tracked_state brw_psp_urb_cbs; - -const struct brw_tracked_state brw_active_vertprog; -const struct brw_tracked_state brw_tnl_vertprog; -const struct brw_tracked_state brw_pipe_control; - -const struct brw_tracked_state brw_clear_surface_cache; -const struct brw_tracked_state brw_clear_batch_cache; - -/*********************************************************************** - * brw_state_cache.c - */ -unsigned brw_cache_data(struct brw_cache *cache, - const void *data ); - -unsigned brw_cache_data_sz(struct brw_cache *cache, - const void *data, - unsigned data_sz); - -unsigned brw_upload_cache( struct brw_cache *cache, - const void *key, - unsigned key_sz, - const void *data, - unsigned data_sz, - const void *aux, - void *aux_return ); - -boolean brw_search_cache( struct brw_cache *cache, - const void *key, - unsigned key_size, - void *aux_return, - unsigned *offset_return); - -void brw_init_caches( struct brw_context *brw ); -void brw_destroy_caches( struct brw_context *brw ); - -static inline struct pipe_buffer *brw_cache_buffer(struct brw_context *brw, - enum brw_cache_id id) -{ - return brw->cache[id].pool->buffer; -} - -/*********************************************************************** - * brw_state_batch.c - */ -#define BRW_CACHED_BATCH_STRUCT(brw, s) brw_cached_batch_struct( brw, (s), sizeof(*(s)) ) - -boolean brw_cached_batch_struct( struct brw_context *brw, - const void *data, - unsigned sz ); - -void brw_destroy_batch_cache( struct brw_context *brw ); - - -/*********************************************************************** - * brw_state_pool.c - */ -void brw_init_pools( struct brw_context *brw ); -void brw_destroy_pools( struct brw_context *brw ); - -boolean brw_pool_alloc( struct brw_mem_pool *pool, - unsigned size, - unsigned alignment, - unsigned *offset_return); - -void brw_pool_fence( struct brw_context *brw, - struct brw_mem_pool *pool, - unsigned fence ); - - -void brw_pool_check_wrap( struct brw_context *brw, - struct brw_mem_pool *pool ); - -void brw_clear_all_caches( struct brw_context *brw ); -void brw_invalidate_pools( struct brw_context *brw ); -void brw_clear_batch_cache_flush( struct brw_context *brw ); - - -/* brw_shader_info.c - */ - -void brw_shader_info(const struct tgsi_token *tokens, - struct brw_shader_info *info ); - - -#endif diff --git a/src/mesa/pipe/i965simple/brw_state_batch.c b/src/mesa/pipe/i965simple/brw_state_batch.c deleted file mode 100644 index 35db76b594..0000000000 --- a/src/mesa/pipe/i965simple/brw_state_batch.c +++ /dev/null @@ -1,113 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_state.h" -#include "brw_winsys.h" - -#include "pipe/p_util.h" - -/* A facility similar to the data caching code above, which aims to - * prevent identical commands being issued repeatedly. - */ -boolean brw_cached_batch_struct( struct brw_context *brw, - const void *data, - unsigned sz ) -{ - struct brw_cached_batch_item *item = brw->cached_batch_items; - struct header *newheader = (struct header *)data; - - if (brw->emit_state_always) { - brw_batchbuffer_data(brw->winsys, data, sz); - return TRUE; - } - - while (item) { - if (item->header->opcode == newheader->opcode) { - if (item->sz == sz && memcmp(item->header, newheader, sz) == 0) - return FALSE; - if (item->sz != sz) { - FREE(item->header); - item->header = MALLOC(sz); - item->sz = sz; - } - goto emit; - } - item = item->next; - } - - assert(!item); - item = CALLOC_STRUCT(brw_cached_batch_item); - item->header = MALLOC(sz); - item->sz = sz; - item->next = brw->cached_batch_items; - brw->cached_batch_items = item; - -emit: - memcpy(item->header, newheader, sz); - brw_batchbuffer_data(brw->winsys, data, sz); - return TRUE; -} - -static void clear_batch_cache( struct brw_context *brw ) -{ - struct brw_cached_batch_item *item = brw->cached_batch_items; - - while (item) { - struct brw_cached_batch_item *next = item->next; - free((void *)item->header); - free(item); - item = next; - } - - brw->cached_batch_items = NULL; - - - brw_clear_all_caches(brw); - - brw_invalidate_pools(brw); -} - -void brw_clear_batch_cache_flush( struct brw_context *brw ) -{ - clear_batch_cache(brw); - -/* brw_do_flush(brw, BRW_FLUSH_STATE_CACHE|BRW_FLUSH_READ_CACHE); */ - - brw->state.dirty.brw |= ~0; - brw->state.dirty.cache |= ~0; -} - - - -void brw_destroy_batch_cache( struct brw_context *brw ) -{ - clear_batch_cache(brw); -} diff --git a/src/mesa/pipe/i965simple/brw_state_cache.c b/src/mesa/pipe/i965simple/brw_state_cache.c deleted file mode 100644 index b3a5124461..0000000000 --- a/src/mesa/pipe/i965simple/brw_state_cache.c +++ /dev/null @@ -1,443 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_state.h" - -#include "brw_wm.h" -#include "brw_vs.h" -#include "brw_clip.h" -#include "brw_sf.h" -#include "brw_gs.h" - -#include "pipe/p_util.h" - - - -/*********************************************************************** - * Check cache for uploaded version of struct, else upload new one. - * Fail when memory is exhausted. - * - * XXX: FIXME: Currently search is so slow it would be quicker to - * regenerate the data every time... - */ - -static unsigned hash_key( const void *key, unsigned key_size ) -{ - unsigned *ikey = (unsigned *)key; - unsigned hash = 0, i; - - assert(key_size % 4 == 0); - - /* I'm sure this can be improved on: - */ - for (i = 0; i < key_size/4; i++) - hash ^= ikey[i]; - - return hash; -} - -static struct brw_cache_item *search_cache( struct brw_cache *cache, - unsigned hash, - const void *key, - unsigned key_size) -{ - struct brw_cache_item *c; - - for (c = cache->items[hash % cache->size]; c; c = c->next) { - if (c->hash == hash && - c->key_size == key_size && - memcmp(c->key, key, key_size) == 0) - return c; - } - - return NULL; -} - - -static void rehash( struct brw_cache *cache ) -{ - struct brw_cache_item **items; - struct brw_cache_item *c, *next; - unsigned size, i; - - size = cache->size * 3; - items = (struct brw_cache_item**) MALLOC(size * sizeof(*items)); - memset(items, 0, size * sizeof(*items)); - - for (i = 0; i < cache->size; i++) - for (c = cache->items[i]; c; c = next) { - next = c->next; - c->next = items[c->hash % size]; - items[c->hash % size] = c; - } - - FREE(cache->items); - cache->items = items; - cache->size = size; -} - - -boolean brw_search_cache( struct brw_cache *cache, - const void *key, - unsigned key_size, - void *aux_return, - unsigned *offset_return) -{ - struct brw_cache_item *item; - unsigned addr = 0; - unsigned hash = hash_key(key, key_size); - - item = search_cache(cache, hash, key, key_size); - - if (item) { - if (aux_return) - *(void **)aux_return = (void *)((char *)item->key + item->key_size); - - *offset_return = addr = item->offset; - } - - if (item == NULL || addr != cache->last_addr) { - cache->brw->state.dirty.cache |= 1<id; - cache->last_addr = addr; - } - - return item != NULL; -} - -unsigned brw_upload_cache( struct brw_cache *cache, - const void *key, - unsigned key_size, - const void *data, - unsigned data_size, - const void *aux, - void *aux_return ) -{ - unsigned offset; - struct brw_cache_item *item = CALLOC_STRUCT(brw_cache_item); - unsigned hash = hash_key(key, key_size); - void *tmp = MALLOC(key_size + cache->aux_size); - - if (!brw_pool_alloc(cache->pool, data_size, 1 << 6, &offset)) { - /* Should not be possible: - */ - debug_printf("brw_pool_alloc failed\n"); - exit(1); - } - - memcpy(tmp, key, key_size); - - if (cache->aux_size) - memcpy(tmp+key_size, aux, cache->aux_size); - - item->key = tmp; - item->hash = hash; - item->key_size = key_size; - item->offset = offset; - item->data_size = data_size; - - if (++cache->n_items > cache->size * 1.5) - rehash(cache); - - hash %= cache->size; - item->next = cache->items[hash]; - cache->items[hash] = item; - - if (aux_return) { - assert(cache->aux_size); - *(void **)aux_return = (void *)((char *)item->key + item->key_size); - } - - if (BRW_DEBUG & DEBUG_STATE) - debug_printf("upload %s: %d bytes to pool buffer %p offset %x\n", - cache->name, - data_size, - (void*)cache->pool->buffer, - offset); - - /* Copy data to the buffer: - */ - cache->brw->winsys->buffer_subdata_typed(cache->brw->winsys, - cache->pool->buffer, - offset, - data_size, - data, - cache->id); - - cache->brw->state.dirty.cache |= 1<id; - cache->last_addr = offset; - - return offset; -} - -/* This doesn't really work with aux data. Use search/upload instead - */ -unsigned brw_cache_data_sz(struct brw_cache *cache, - const void *data, - unsigned data_size) -{ - unsigned addr; - - if (!brw_search_cache(cache, data, data_size, NULL, &addr)) { - addr = brw_upload_cache(cache, - data, data_size, - data, data_size, - NULL, NULL); - } - - return addr; -} - -unsigned brw_cache_data(struct brw_cache *cache, - const void *data) -{ - return brw_cache_data_sz(cache, data, cache->key_size); -} - -enum pool_type { - DW_SURFACE_STATE, - DW_GENERAL_STATE -}; - -static void brw_init_cache( struct brw_context *brw, - const char *name, - unsigned id, - unsigned key_size, - unsigned aux_size, - enum pool_type pool_type) -{ - struct brw_cache *cache = &brw->cache[id]; - cache->brw = brw; - cache->id = id; - cache->name = name; - cache->items = NULL; - - cache->size = 7; - cache->n_items = 0; - cache->items = (struct brw_cache_item **) - CALLOC(cache->size, sizeof(struct brw_cache_item)); - - - cache->key_size = key_size; - cache->aux_size = aux_size; - switch (pool_type) { - case DW_GENERAL_STATE: cache->pool = &brw->pool[BRW_GS_POOL]; break; - case DW_SURFACE_STATE: cache->pool = &brw->pool[BRW_SS_POOL]; break; - default: assert(0); break; - } -} - -void brw_init_caches( struct brw_context *brw ) -{ - - brw_init_cache(brw, - "CC_VP", - BRW_CC_VP, - sizeof(struct brw_cc_viewport), - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "CC_UNIT", - BRW_CC_UNIT, - sizeof(struct brw_cc_unit_state), - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "WM_PROG", - BRW_WM_PROG, - sizeof(struct brw_wm_prog_key), - sizeof(struct brw_wm_prog_data), - DW_GENERAL_STATE); - - brw_init_cache(brw, - "SAMPLER_DEFAULT_COLOR", - BRW_SAMPLER_DEFAULT_COLOR, - sizeof(struct brw_sampler_default_color), - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "SAMPLER", - BRW_SAMPLER, - 0, /* variable key/data size */ - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "WM_UNIT", - BRW_WM_UNIT, - sizeof(struct brw_wm_unit_state), - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "SF_PROG", - BRW_SF_PROG, - sizeof(struct brw_sf_prog_key), - sizeof(struct brw_sf_prog_data), - DW_GENERAL_STATE); - - brw_init_cache(brw, - "SF_VP", - BRW_SF_VP, - sizeof(struct brw_sf_viewport), - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "SF_UNIT", - BRW_SF_UNIT, - sizeof(struct brw_sf_unit_state), - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "VS_UNIT", - BRW_VS_UNIT, - sizeof(struct brw_vs_unit_state), - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "VS_PROG", - BRW_VS_PROG, - sizeof(struct brw_vs_prog_key), - sizeof(struct brw_vs_prog_data), - DW_GENERAL_STATE); - - brw_init_cache(brw, - "CLIP_UNIT", - BRW_CLIP_UNIT, - sizeof(struct brw_clip_unit_state), - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "CLIP_PROG", - BRW_CLIP_PROG, - sizeof(struct brw_clip_prog_key), - sizeof(struct brw_clip_prog_data), - DW_GENERAL_STATE); - - brw_init_cache(brw, - "GS_UNIT", - BRW_GS_UNIT, - sizeof(struct brw_gs_unit_state), - 0, - DW_GENERAL_STATE); - - brw_init_cache(brw, - "GS_PROG", - BRW_GS_PROG, - sizeof(struct brw_gs_prog_key), - sizeof(struct brw_gs_prog_data), - DW_GENERAL_STATE); - - brw_init_cache(brw, - "SS_SURFACE", - BRW_SS_SURFACE, - sizeof(struct brw_surface_state), - 0, - DW_SURFACE_STATE); - - brw_init_cache(brw, - "SS_SURF_BIND", - BRW_SS_SURF_BIND, - sizeof(struct brw_surface_binding_table), - 0, - DW_SURFACE_STATE); -} - - -/* When we lose hardware context, need to invalidate the surface cache - * as these structs must be explicitly re-uploaded. They are subject - * to fixup by the memory manager as they contain absolute agp - * offsets, so we need to ensure there is a fresh version of the - * struct available to receive the fixup. - * - * XXX: Need to ensure that there aren't two versions of a surface or - * bufferobj with different backing data active in the same buffer at - * once? Otherwise the cache could confuse them. Maybe better not to - * cache at all? - * - * --> Isn't this the same as saying need to ensure batch is flushed - * before new data is uploaded to an existing buffer? We - * already try to make sure of that. - */ -static void clear_cache( struct brw_cache *cache ) -{ - struct brw_cache_item *c, *next; - unsigned i; - - for (i = 0; i < cache->size; i++) { - for (c = cache->items[i]; c; c = next) { - next = c->next; - free((void *)c->key); - free(c); - } - cache->items[i] = NULL; - } - - cache->n_items = 0; -} - -void brw_clear_all_caches( struct brw_context *brw ) -{ - int i; - - if (BRW_DEBUG & DEBUG_STATE) - debug_printf("%s\n", __FUNCTION__); - - for (i = 0; i < BRW_MAX_CACHE; i++) - clear_cache(&brw->cache[i]); - - if (brw->curbe.last_buf) { - FREE(brw->curbe.last_buf); - brw->curbe.last_buf = NULL; - } - - brw->state.dirty.brw |= ~0; - brw->state.dirty.cache |= ~0; -} - - - - - -void brw_destroy_caches( struct brw_context *brw ) -{ - unsigned i; - - for (i = 0; i < BRW_MAX_CACHE; i++) - clear_cache(&brw->cache[i]); -} diff --git a/src/mesa/pipe/i965simple/brw_state_pool.c b/src/mesa/pipe/i965simple/brw_state_pool.c deleted file mode 100644 index f3174bfe0a..0000000000 --- a/src/mesa/pipe/i965simple/brw_state_pool.c +++ /dev/null @@ -1,137 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -/** @file brw_state_pool.c - * Implements the state pool allocator. - * - * For the 965, we create two state pools for state cache entries. Objects - * will be allocated into the pools depending on which state base address - * their pointer is relative to in other 965 state. - * - * The state pools are relatively simple: As objects are allocated, increment - * the offset to allocate space. When the pool is "full" (rather, close to - * full), we reset the pool and reset the state cache entries that point into - * the pool. - */ - -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "brw_context.h" -#include "brw_state.h" - -boolean brw_pool_alloc( struct brw_mem_pool *pool, - unsigned size, - unsigned alignment, - unsigned *offset_return) -{ - unsigned fixup = align(pool->offset, alignment) - pool->offset; - - size = align(size, 4); - - if (pool->offset + fixup + size >= pool->size) { - debug_printf("%s failed\n", __FUNCTION__); - assert(0); - exit(0); - } - - pool->offset += fixup; - *offset_return = pool->offset; - pool->offset += size; - - return TRUE; -} - -static -void brw_invalidate_pool( struct brw_mem_pool *pool ) -{ - if (BRW_DEBUG & DEBUG_STATE) - debug_printf("\n\n\n %s \n\n\n", __FUNCTION__); - - pool->offset = 0; - - brw_clear_all_caches(pool->brw); -} - - -static void brw_init_pool( struct brw_context *brw, - unsigned pool_id, - unsigned size ) -{ - struct brw_mem_pool *pool = &brw->pool[pool_id]; - - pool->size = size; - pool->brw = brw; - - pool->buffer = brw->pipe.winsys->buffer_create(brw->pipe.winsys, - 4096, - 0 /* DRM_BO_FLAG_MEM_TT */, - size); -} - -static void brw_destroy_pool( struct brw_context *brw, - unsigned pool_id ) -{ - struct brw_mem_pool *pool = &brw->pool[pool_id]; - - pipe_buffer_reference( pool->brw->pipe.winsys, - &pool->buffer, - NULL ); -} - - -void brw_pool_check_wrap( struct brw_context *brw, - struct brw_mem_pool *pool ) -{ - if (pool->offset > (pool->size * 3) / 4) { - brw->state.dirty.brw |= BRW_NEW_SCENE; - } - -} - -void brw_init_pools( struct brw_context *brw ) -{ - brw_init_pool(brw, BRW_GS_POOL, 0x80000); - brw_init_pool(brw, BRW_SS_POOL, 0x80000); -} - -void brw_destroy_pools( struct brw_context *brw ) -{ - brw_destroy_pool(brw, BRW_GS_POOL); - brw_destroy_pool(brw, BRW_SS_POOL); -} - - -void brw_invalidate_pools( struct brw_context *brw ) -{ - brw_invalidate_pool(&brw->pool[BRW_GS_POOL]); - brw_invalidate_pool(&brw->pool[BRW_SS_POOL]); -} diff --git a/src/mesa/pipe/i965simple/brw_state_upload.c b/src/mesa/pipe/i965simple/brw_state_upload.c deleted file mode 100644 index e727601e1e..0000000000 --- a/src/mesa/pipe/i965simple/brw_state_upload.c +++ /dev/null @@ -1,202 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_state.h" - -#include "pipe/p_util.h" - -/* This is used to initialize brw->state.atoms[]. We could use this - * list directly except for a single atom, brw_constant_buffer, which - * has a .dirty value which changes according to the parameters of the - * current fragment and vertex programs, and so cannot be a static - * value. - */ -const struct brw_tracked_state *atoms[] = -{ - &brw_vs_prog, - &brw_gs_prog, - &brw_clip_prog, - &brw_sf_prog, - &brw_wm_prog, - - /* Once all the programs are done, we know how large urb entry - * sizes need to be and can decide if we need to change the urb - * layout. - */ - &brw_curbe_offsets, - &brw_recalculate_urb_fence, - - - &brw_cc_vp, - &brw_cc_unit, - - &brw_wm_surfaces, /* must do before samplers */ - &brw_wm_samplers, - - &brw_wm_unit, - &brw_sf_vp, - &brw_sf_unit, - &brw_vs_unit, /* always required, enabled or not */ - &brw_clip_unit, - &brw_gs_unit, - - /* Command packets: - */ - &brw_invarient_state, - &brw_state_base_address, - &brw_pipe_control, - - &brw_binding_table_pointers, - &brw_blend_constant_color, - - &brw_drawing_rect, - &brw_depthbuffer, - - &brw_polygon_stipple, - &brw_line_stipple, - - &brw_psp_urb_cbs, - - &brw_constant_buffer -}; - - -void brw_init_state( struct brw_context *brw ) -{ - brw_init_pools(brw); - brw_init_caches(brw); - - brw->state.dirty.brw = ~0; - brw->emit_state_always = 0; -} - - -void brw_destroy_state( struct brw_context *brw ) -{ - brw_destroy_caches(brw); - brw_destroy_batch_cache(brw); - brw_destroy_pools(brw); -} - -/*********************************************************************** - */ - -static boolean check_state( const struct brw_state_flags *a, - const struct brw_state_flags *b ) -{ - return ((a->brw & b->brw) || - (a->cache & b->cache)); -} - -static void accumulate_state( struct brw_state_flags *a, - const struct brw_state_flags *b ) -{ - a->brw |= b->brw; - a->cache |= b->cache; -} - - -static void xor_states( struct brw_state_flags *result, - const struct brw_state_flags *a, - const struct brw_state_flags *b ) -{ - result->brw = a->brw ^ b->brw; - result->cache = a->cache ^ b->cache; -} - - -/*********************************************************************** - * Emit all state: - */ -void brw_validate_state( struct brw_context *brw ) -{ - struct brw_state_flags *state = &brw->state.dirty; - unsigned i; - - if (brw->emit_state_always) - state->brw |= ~0; - - if (state->cache == 0 && - state->brw == 0) - return; - - if (brw->state.dirty.brw & BRW_NEW_SCENE) - brw_clear_batch_cache_flush(brw); - - if (BRW_DEBUG) { - /* Debug version which enforces various sanity checks on the - * state flags which are generated and checked to help ensure - * state atoms are ordered correctly in the list. - */ - struct brw_state_flags examined, prev; - memset(&examined, 0, sizeof(examined)); - prev = *state; - - for (i = 0; i < Elements(atoms); i++) { - const struct brw_tracked_state *atom = atoms[i]; - struct brw_state_flags generated; - - assert(atom->dirty.brw || - atom->dirty.cache); - assert(atom->update); - - if (check_state(state, &atom->dirty)) { - atom->update( brw ); - } - - accumulate_state(&examined, &atom->dirty); - - /* generated = (prev ^ state) - * if (examined & generated) - * fail; - */ - xor_states(&generated, &prev, state); - assert(!check_state(&examined, &generated)); - prev = *state; - } - } - else { - for (i = 0; i < Elements(atoms); i++) { - const struct brw_tracked_state *atom = atoms[i]; - - assert(atom->dirty.brw || - atom->dirty.cache); - assert(atom->update); - - if (check_state(state, &atom->dirty)) - atom->update( brw ); - } - } - - memset(state, 0, sizeof(*state)); -} diff --git a/src/mesa/pipe/i965simple/brw_strings.c b/src/mesa/pipe/i965simple/brw_strings.c deleted file mode 100644 index 29a41ed1e9..0000000000 --- a/src/mesa/pipe/i965simple/brw_strings.c +++ /dev/null @@ -1,72 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "brw_context.h" -#include "brw_reg.h" - - -static const char *brw_get_vendor( struct pipe_context *pipe ) -{ - return "Tungsten Graphics, Inc."; -} - - -static const char *brw_get_name( struct pipe_context *pipe ) -{ - static char buffer[128]; - const char *chipset; - - switch (brw_context(pipe)->pci_id) { - case PCI_CHIP_I965_Q: - chipset = "Intel(R) 965Q"; - break; - case PCI_CHIP_I965_G: - case PCI_CHIP_I965_G_1: - chipset = "Intel(R) 965G"; - break; - case PCI_CHIP_I965_GM: - chipset = "Intel(R) 965GM"; - break; - case PCI_CHIP_I965_GME: - chipset = "Intel(R) 965GME/GLE"; - break; - default: - chipset = "unknown"; - break; - } - - sprintf(buffer, "pipe/i965 (chipset: %s)", chipset); - return buffer; -} - - -void -brw_init_string_functions(struct brw_context *brw) -{ - brw->pipe.get_name = brw_get_name; - brw->pipe.get_vendor = brw_get_vendor; -} diff --git a/src/mesa/pipe/i965simple/brw_structs.h b/src/mesa/pipe/i965simple/brw_structs.h deleted file mode 100644 index bbb087e95d..0000000000 --- a/src/mesa/pipe/i965simple/brw_structs.h +++ /dev/null @@ -1,1348 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRW_STRUCTS_H -#define BRW_STRUCTS_H - -#include "pipe/p_compiler.h" - -/* Command packets: - */ -struct header -{ - unsigned length:16; - unsigned opcode:16; -}; - - -union header_union -{ - struct header bits; - unsigned dword; -}; - -struct brw_3d_control -{ - struct - { - unsigned length:8; - unsigned notify_enable:1; - unsigned pad:3; - unsigned wc_flush_enable:1; - unsigned depth_stall_enable:1; - unsigned operation:2; - unsigned opcode:16; - } header; - - struct - { - unsigned pad:2; - unsigned dest_addr_type:1; - unsigned dest_addr:29; - } dest; - - unsigned dword2; - unsigned dword3; -}; - - -struct brw_3d_primitive -{ - struct - { - unsigned length:8; - unsigned pad:2; - unsigned topology:5; - unsigned indexed:1; - unsigned opcode:16; - } header; - - unsigned verts_per_instance; - unsigned start_vert_location; - unsigned instance_count; - unsigned start_instance_location; - unsigned base_vert_location; -}; - -/* These seem to be passed around as function args, so it works out - * better to keep them as #defines: - */ -#define BRW_FLUSH_READ_CACHE 0x1 -#define BRW_FLUSH_STATE_CACHE 0x2 -#define BRW_INHIBIT_FLUSH_RENDER_CACHE 0x4 -#define BRW_FLUSH_SNAPSHOT_COUNTERS 0x8 - -struct brw_mi_flush -{ - unsigned flags:4; - unsigned pad:12; - unsigned opcode:16; -}; - -struct brw_vf_statistics -{ - unsigned statistics_enable:1; - unsigned pad:15; - unsigned opcode:16; -}; - - - -struct brw_binding_table_pointers -{ - struct header header; - unsigned vs; - unsigned gs; - unsigned clp; - unsigned sf; - unsigned wm; -}; - - -struct brw_blend_constant_color -{ - struct header header; - float blend_constant_color[4]; -}; - - -struct brw_depthbuffer -{ - union header_union header; - - union { - struct { - unsigned pitch:18; - unsigned format:3; - unsigned pad:4; - unsigned depth_offset_disable:1; - unsigned tile_walk:1; - unsigned tiled_surface:1; - unsigned pad2:1; - unsigned surface_type:3; - } bits; - unsigned dword; - } dword1; - - unsigned dword2_base_addr; - - union { - struct { - unsigned pad:1; - unsigned mipmap_layout:1; - unsigned lod:4; - unsigned width:13; - unsigned height:13; - } bits; - unsigned dword; - } dword3; - - union { - struct { - unsigned pad:12; - unsigned min_array_element:9; - unsigned depth:11; - } bits; - unsigned dword; - } dword4; -}; - -struct brw_drawrect -{ - struct header header; - unsigned xmin:16; - unsigned ymin:16; - unsigned xmax:16; - unsigned ymax:16; - unsigned xorg:16; - unsigned yorg:16; -}; - - - - -struct brw_global_depth_offset_clamp -{ - struct header header; - float depth_offset_clamp; -}; - -struct brw_indexbuffer -{ - union { - struct - { - unsigned length:8; - unsigned index_format:2; - unsigned cut_index_enable:1; - unsigned pad:5; - unsigned opcode:16; - } bits; - unsigned dword; - - } header; - - unsigned buffer_start; - unsigned buffer_end; -}; - - -struct brw_line_stipple -{ - struct header header; - - struct - { - unsigned pattern:16; - unsigned pad:16; - } bits0; - - struct - { - unsigned repeat_count:9; - unsigned pad:7; - unsigned inverse_repeat_count:16; - } bits1; -}; - - -struct brw_pipelined_state_pointers -{ - struct header header; - - struct { - unsigned pad:5; - unsigned offset:27; - } vs; - - struct - { - unsigned enable:1; - unsigned pad:4; - unsigned offset:27; - } gs; - - struct - { - unsigned enable:1; - unsigned pad:4; - unsigned offset:27; - } clp; - - struct - { - unsigned pad:5; - unsigned offset:27; - } sf; - - struct - { - unsigned pad:5; - unsigned offset:27; - } wm; - - struct - { - unsigned pad:5; - unsigned offset:27; /* KW: check me! */ - } cc; -}; - - -struct brw_polygon_stipple_offset -{ - struct header header; - - struct { - unsigned y_offset:5; - unsigned pad:3; - unsigned x_offset:5; - unsigned pad0:19; - } bits0; -}; - - - -struct brw_polygon_stipple -{ - struct header header; - unsigned stipple[32]; -}; - - - -struct brw_pipeline_select -{ - struct - { - unsigned pipeline_select:1; - unsigned pad:15; - unsigned opcode:16; - } header; -}; - - -struct brw_pipe_control -{ - struct - { - unsigned length:8; - unsigned notify_enable:1; - unsigned pad:2; - unsigned instruction_state_cache_flush_enable:1; - unsigned write_cache_flush_enable:1; - unsigned depth_stall_enable:1; - unsigned post_sync_operation:2; - - unsigned opcode:16; - } header; - - struct - { - unsigned pad:2; - unsigned dest_addr_type:1; - unsigned dest_addr:29; - } bits1; - - unsigned data0; - unsigned data1; -}; - - -struct brw_urb_fence -{ - struct - { - unsigned length:8; - unsigned vs_realloc:1; - unsigned gs_realloc:1; - unsigned clp_realloc:1; - unsigned sf_realloc:1; - unsigned vfe_realloc:1; - unsigned cs_realloc:1; - unsigned pad:2; - unsigned opcode:16; - } header; - - struct - { - unsigned vs_fence:10; - unsigned gs_fence:10; - unsigned clp_fence:10; - unsigned pad:2; - } bits0; - - struct - { - unsigned sf_fence:10; - unsigned vf_fence:10; - unsigned cs_fence:10; - unsigned pad:2; - } bits1; -}; - -struct brw_constant_buffer_state /* previously brw_command_streamer */ -{ - struct header header; - - struct - { - unsigned nr_urb_entries:3; - unsigned pad:1; - unsigned urb_entry_size:5; - unsigned pad0:23; - } bits0; -}; - -struct brw_constant_buffer -{ - struct - { - unsigned length:8; - unsigned valid:1; - unsigned pad:7; - unsigned opcode:16; - } header; - - struct - { - unsigned buffer_length:6; - unsigned buffer_address:26; - } bits0; -}; - -struct brw_state_base_address -{ - struct header header; - - struct - { - unsigned modify_enable:1; - unsigned pad:4; - unsigned general_state_address:27; - } bits0; - - struct - { - unsigned modify_enable:1; - unsigned pad:4; - unsigned surface_state_address:27; - } bits1; - - struct - { - unsigned modify_enable:1; - unsigned pad:4; - unsigned indirect_object_state_address:27; - } bits2; - - struct - { - unsigned modify_enable:1; - unsigned pad:11; - unsigned general_state_upper_bound:20; - } bits3; - - struct - { - unsigned modify_enable:1; - unsigned pad:11; - unsigned indirect_object_state_upper_bound:20; - } bits4; -}; - -struct brw_state_prefetch -{ - struct header header; - - struct - { - unsigned prefetch_count:3; - unsigned pad:3; - unsigned prefetch_pointer:26; - } bits0; -}; - -struct brw_system_instruction_pointer -{ - struct header header; - - struct - { - unsigned pad:4; - unsigned system_instruction_pointer:28; - } bits0; -}; - - - - -/* State structs for the various fixed function units: - */ - - -struct thread0 -{ - unsigned pad0:1; - unsigned grf_reg_count:3; - unsigned pad1:2; - unsigned kernel_start_pointer:26; -}; - -struct thread1 -{ - unsigned ext_halt_exception_enable:1; - unsigned sw_exception_enable:1; - unsigned mask_stack_exception_enable:1; - unsigned timeout_exception_enable:1; - unsigned illegal_op_exception_enable:1; - unsigned pad0:3; - unsigned depth_coef_urb_read_offset:6; /* WM only */ - unsigned pad1:2; - unsigned floating_point_mode:1; - unsigned thread_priority:1; - unsigned binding_table_entry_count:8; - unsigned pad3:5; - unsigned single_program_flow:1; -}; - -struct thread2 -{ - unsigned per_thread_scratch_space:4; - unsigned pad0:6; - unsigned scratch_space_base_pointer:22; -}; - - -struct thread3 -{ - unsigned dispatch_grf_start_reg:4; - unsigned urb_entry_read_offset:6; - unsigned pad0:1; - unsigned urb_entry_read_length:6; - unsigned pad1:1; - unsigned const_urb_entry_read_offset:6; - unsigned pad2:1; - unsigned const_urb_entry_read_length:6; - unsigned pad3:1; -}; - - - -struct brw_clip_unit_state -{ - struct thread0 thread0; - struct - { - unsigned pad0:7; - unsigned sw_exception_enable:1; - unsigned pad1:3; - unsigned mask_stack_exception_enable:1; - unsigned pad2:1; - unsigned illegal_op_exception_enable:1; - unsigned pad3:2; - unsigned floating_point_mode:1; - unsigned thread_priority:1; - unsigned binding_table_entry_count:8; - unsigned pad4:5; - unsigned single_program_flow:1; - } thread1; - - struct thread2 thread2; - struct thread3 thread3; - - struct - { - unsigned pad0:9; - unsigned gs_output_stats:1; /* not always */ - unsigned stats_enable:1; - unsigned nr_urb_entries:7; - unsigned pad1:1; - unsigned urb_entry_allocation_size:5; - unsigned pad2:1; - unsigned max_threads:1; /* may be less */ - unsigned pad3:6; - } thread4; - - struct - { - unsigned pad0:13; - unsigned clip_mode:3; - unsigned userclip_enable_flags:8; - unsigned userclip_must_clip:1; - unsigned pad1:1; - unsigned guard_band_enable:1; - unsigned viewport_z_clip_enable:1; - unsigned viewport_xy_clip_enable:1; - unsigned vertex_position_space:1; - unsigned api_mode:1; - unsigned pad2:1; - } clip5; - - struct - { - unsigned pad0:5; - unsigned clipper_viewport_state_ptr:27; - } clip6; - - - float viewport_xmin; - float viewport_xmax; - float viewport_ymin; - float viewport_ymax; -}; - - - -struct brw_cc_unit_state -{ - struct - { - unsigned pad0:3; - unsigned bf_stencil_pass_depth_pass_op:3; - unsigned bf_stencil_pass_depth_fail_op:3; - unsigned bf_stencil_fail_op:3; - unsigned bf_stencil_func:3; - unsigned bf_stencil_enable:1; - unsigned pad1:2; - unsigned stencil_write_enable:1; - unsigned stencil_pass_depth_pass_op:3; - unsigned stencil_pass_depth_fail_op:3; - unsigned stencil_fail_op:3; - unsigned stencil_func:3; - unsigned stencil_enable:1; - } cc0; - - - struct - { - unsigned bf_stencil_ref:8; - unsigned stencil_write_mask:8; - unsigned stencil_test_mask:8; - unsigned stencil_ref:8; - } cc1; - - - struct - { - unsigned logicop_enable:1; - unsigned pad0:10; - unsigned depth_write_enable:1; - unsigned depth_test_function:3; - unsigned depth_test:1; - unsigned bf_stencil_write_mask:8; - unsigned bf_stencil_test_mask:8; - } cc2; - - - struct - { - unsigned pad0:8; - unsigned alpha_test_func:3; - unsigned alpha_test:1; - unsigned blend_enable:1; - unsigned ia_blend_enable:1; - unsigned pad1:1; - unsigned alpha_test_format:1; - unsigned pad2:16; - } cc3; - - struct - { - unsigned pad0:5; - unsigned cc_viewport_state_offset:27; - } cc4; - - struct - { - unsigned pad0:2; - unsigned ia_dest_blend_factor:5; - unsigned ia_src_blend_factor:5; - unsigned ia_blend_function:3; - unsigned statistics_enable:1; - unsigned logicop_func:4; - unsigned pad1:11; - unsigned dither_enable:1; - } cc5; - - struct - { - unsigned clamp_post_alpha_blend:1; - unsigned clamp_pre_alpha_blend:1; - unsigned clamp_range:2; - unsigned pad0:11; - unsigned y_dither_offset:2; - unsigned x_dither_offset:2; - unsigned dest_blend_factor:5; - unsigned src_blend_factor:5; - unsigned blend_function:3; - } cc6; - - struct { - union { - float f; - ubyte ub[4]; - } alpha_ref; - } cc7; -}; - - - -struct brw_sf_unit_state -{ - struct thread0 thread0; - struct thread1 thread1; - struct thread2 thread2; - struct thread3 thread3; - - struct - { - unsigned pad0:10; - unsigned stats_enable:1; - unsigned nr_urb_entries:7; - unsigned pad1:1; - unsigned urb_entry_allocation_size:5; - unsigned pad2:1; - unsigned max_threads:6; - unsigned pad3:1; - } thread4; - - struct - { - unsigned front_winding:1; - unsigned viewport_transform:1; - unsigned pad0:3; - unsigned sf_viewport_state_offset:27; - } sf5; - - struct - { - unsigned pad0:9; - unsigned dest_org_vbias:4; - unsigned dest_org_hbias:4; - unsigned scissor:1; - unsigned disable_2x2_trifilter:1; - unsigned disable_zero_pix_trifilter:1; - unsigned point_rast_rule:2; - unsigned line_endcap_aa_region_width:2; - unsigned line_width:4; - unsigned fast_scissor_disable:1; - unsigned cull_mode:2; - unsigned aa_enable:1; - } sf6; - - struct - { - unsigned point_size:11; - unsigned use_point_size_state:1; - unsigned subpixel_precision:1; - unsigned sprite_point:1; - unsigned pad0:11; - unsigned trifan_pv:2; - unsigned linestrip_pv:2; - unsigned tristrip_pv:2; - unsigned line_last_pixel_enable:1; - } sf7; - -}; - - -struct brw_gs_unit_state -{ - struct thread0 thread0; - struct thread1 thread1; - struct thread2 thread2; - struct thread3 thread3; - - struct - { - unsigned pad0:10; - unsigned stats_enable:1; - unsigned nr_urb_entries:7; - unsigned pad1:1; - unsigned urb_entry_allocation_size:5; - unsigned pad2:1; - unsigned max_threads:1; - unsigned pad3:6; - } thread4; - - struct - { - unsigned sampler_count:3; - unsigned pad0:2; - unsigned sampler_state_pointer:27; - } gs5; - - - struct - { - unsigned max_vp_index:4; - unsigned pad0:26; - unsigned reorder_enable:1; - unsigned pad1:1; - } gs6; -}; - - -struct brw_vs_unit_state -{ - struct thread0 thread0; - struct thread1 thread1; - struct thread2 thread2; - struct thread3 thread3; - - struct - { - unsigned pad0:10; - unsigned stats_enable:1; - unsigned nr_urb_entries:7; - unsigned pad1:1; - unsigned urb_entry_allocation_size:5; - unsigned pad2:1; - unsigned max_threads:4; - unsigned pad3:3; - } thread4; - - struct - { - unsigned sampler_count:3; - unsigned pad0:2; - unsigned sampler_state_pointer:27; - } vs5; - - struct - { - unsigned vs_enable:1; - unsigned vert_cache_disable:1; - unsigned pad0:30; - } vs6; -}; - - -struct brw_wm_unit_state -{ - struct thread0 thread0; - struct thread1 thread1; - struct thread2 thread2; - struct thread3 thread3; - - struct { - unsigned stats_enable:1; - unsigned pad0:1; - unsigned sampler_count:3; - unsigned sampler_state_pointer:27; - } wm4; - - struct - { - unsigned enable_8_pix:1; - unsigned enable_16_pix:1; - unsigned enable_32_pix:1; - unsigned pad0:7; - unsigned legacy_global_depth_bias:1; - unsigned line_stipple:1; - unsigned depth_offset:1; - unsigned polygon_stipple:1; - unsigned line_aa_region_width:2; - unsigned line_endcap_aa_region_width:2; - unsigned early_depth_test:1; - unsigned thread_dispatch_enable:1; - unsigned program_uses_depth:1; - unsigned program_computes_depth:1; - unsigned program_uses_killpixel:1; - unsigned legacy_line_rast: 1; - unsigned pad1:1; - unsigned max_threads:6; - unsigned pad2:1; - } wm5; - - float global_depth_offset_constant; - float global_depth_offset_scale; -}; - -struct brw_sampler_default_color { - float color[4]; -}; - -struct brw_sampler_state -{ - - struct - { - unsigned shadow_function:3; - unsigned lod_bias:11; - unsigned min_filter:3; - unsigned mag_filter:3; - unsigned mip_filter:2; - unsigned base_level:5; - unsigned pad:1; - unsigned lod_preclamp:1; - unsigned default_color_mode:1; - unsigned pad0:1; - unsigned disable:1; - } ss0; - - struct - { - unsigned r_wrap_mode:3; - unsigned t_wrap_mode:3; - unsigned s_wrap_mode:3; - unsigned pad:3; - unsigned max_lod:10; - unsigned min_lod:10; - } ss1; - - - struct - { - unsigned pad:5; - unsigned default_color_pointer:27; - } ss2; - - struct - { - unsigned pad:19; - unsigned max_aniso:3; - unsigned chroma_key_mode:1; - unsigned chroma_key_index:2; - unsigned chroma_key_enable:1; - unsigned monochrome_filter_width:3; - unsigned monochrome_filter_height:3; - } ss3; -}; - - -struct brw_clipper_viewport -{ - float xmin; - float xmax; - float ymin; - float ymax; -}; - -struct brw_cc_viewport -{ - float min_depth; - float max_depth; -}; - -struct brw_sf_viewport -{ - struct { - float m00; - float m11; - float m22; - float m30; - float m31; - float m32; - } viewport; - - struct { - short xmin; - short ymin; - short xmax; - short ymax; - } scissor; -}; - -/* Documented in the subsystem/shared-functions/sampler chapter... - */ -struct brw_surface_state -{ - struct { - unsigned cube_pos_z:1; - unsigned cube_neg_z:1; - unsigned cube_pos_y:1; - unsigned cube_neg_y:1; - unsigned cube_pos_x:1; - unsigned cube_neg_x:1; - unsigned pad:4; - unsigned mipmap_layout_mode:1; - unsigned vert_line_stride_ofs:1; - unsigned vert_line_stride:1; - unsigned color_blend:1; - unsigned writedisable_blue:1; - unsigned writedisable_green:1; - unsigned writedisable_red:1; - unsigned writedisable_alpha:1; - unsigned surface_format:9; - unsigned data_return_format:1; - unsigned pad0:1; - unsigned surface_type:3; - } ss0; - - struct { - unsigned base_addr; - } ss1; - - struct { - unsigned pad:2; - unsigned mip_count:4; - unsigned width:13; - unsigned height:13; - } ss2; - - struct { - unsigned tile_walk:1; - unsigned tiled_surface:1; - unsigned pad:1; - unsigned pitch:18; - unsigned depth:11; - } ss3; - - struct { - unsigned pad:19; - unsigned min_array_elt:9; - unsigned min_lod:4; - } ss4; -}; - - - -struct brw_vertex_buffer_state -{ - struct { - unsigned pitch:11; - unsigned pad:15; - unsigned access_type:1; - unsigned vb_index:5; - } vb0; - - unsigned start_addr; - unsigned max_index; -#if 1 - unsigned instance_data_step_rate; /* not included for sequential/random vertices? */ -#endif -}; - -#define BRW_VBP_MAX 17 - -struct brw_vb_array_state { - struct header header; - struct brw_vertex_buffer_state vb[BRW_VBP_MAX]; -}; - - -struct brw_vertex_element_state -{ - struct - { - unsigned src_offset:11; - unsigned pad:5; - unsigned src_format:9; - unsigned pad0:1; - unsigned valid:1; - unsigned vertex_buffer_index:5; - } ve0; - - struct - { - unsigned dst_offset:8; - unsigned pad:8; - unsigned vfcomponent3:4; - unsigned vfcomponent2:4; - unsigned vfcomponent1:4; - unsigned vfcomponent0:4; - } ve1; -}; - -#define BRW_VEP_MAX 18 - -struct brw_vertex_element_packet { - struct header header; - struct brw_vertex_element_state ve[BRW_VEP_MAX]; /* note: less than _TNL_ATTRIB_MAX */ -}; - - -struct brw_urb_immediate { - unsigned opcode:4; - unsigned offset:6; - unsigned swizzle_control:2; - unsigned pad:1; - unsigned allocate:1; - unsigned used:1; - unsigned complete:1; - unsigned response_length:4; - unsigned msg_length:4; - unsigned msg_target:4; - unsigned pad1:3; - unsigned end_of_thread:1; -}; - -/* Instruction format for the execution units: - */ - -struct brw_instruction -{ - struct - { - unsigned opcode:7; - unsigned pad:1; - unsigned access_mode:1; - unsigned mask_control:1; - unsigned dependency_control:2; - unsigned compression_control:2; - unsigned thread_control:2; - unsigned predicate_control:4; - unsigned predicate_inverse:1; - unsigned execution_size:3; - unsigned destreg__conditonalmod:4; /* destreg - send, conditionalmod - others */ - unsigned pad0:2; - unsigned debug_control:1; - unsigned saturate:1; - } header; - - union { - struct - { - unsigned dest_reg_file:2; - unsigned dest_reg_type:3; - unsigned src0_reg_file:2; - unsigned src0_reg_type:3; - unsigned src1_reg_file:2; - unsigned src1_reg_type:3; - unsigned pad:1; - unsigned dest_subreg_nr:5; - unsigned dest_reg_nr:8; - unsigned dest_horiz_stride:2; - unsigned dest_address_mode:1; - } da1; - - struct - { - unsigned dest_reg_file:2; - unsigned dest_reg_type:3; - unsigned src0_reg_file:2; - unsigned src0_reg_type:3; - unsigned pad:6; - int dest_indirect_offset:10; /* offset against the deref'd address reg */ - unsigned dest_subreg_nr:3; /* subnr for the address reg a0.x */ - unsigned dest_horiz_stride:2; - unsigned dest_address_mode:1; - } ia1; - - struct - { - unsigned dest_reg_file:2; - unsigned dest_reg_type:3; - unsigned src0_reg_file:2; - unsigned src0_reg_type:3; - unsigned src1_reg_file:2; - unsigned src1_reg_type:3; - unsigned pad0:1; - unsigned dest_writemask:4; - unsigned dest_subreg_nr:1; - unsigned dest_reg_nr:8; - unsigned pad1:2; - unsigned dest_address_mode:1; - } da16; - - struct - { - unsigned dest_reg_file:2; - unsigned dest_reg_type:3; - unsigned src0_reg_file:2; - unsigned src0_reg_type:3; - unsigned pad0:6; - unsigned dest_writemask:4; - int dest_indirect_offset:6; - unsigned dest_subreg_nr:3; - unsigned pad1:2; - unsigned dest_address_mode:1; - } ia16; - } bits1; - - - union { - struct - { - unsigned src0_subreg_nr:5; - unsigned src0_reg_nr:8; - unsigned src0_abs:1; - unsigned src0_negate:1; - unsigned src0_address_mode:1; - unsigned src0_horiz_stride:2; - unsigned src0_width:3; - unsigned src0_vert_stride:4; - unsigned flag_reg_nr:1; - unsigned pad:6; - } da1; - - struct - { - int src0_indirect_offset:10; - unsigned src0_subreg_nr:3; - unsigned src0_abs:1; - unsigned src0_negate:1; - unsigned src0_address_mode:1; - unsigned src0_horiz_stride:2; - unsigned src0_width:3; - unsigned src0_vert_stride:4; - unsigned flag_reg_nr:1; - unsigned pad:6; - } ia1; - - struct - { - unsigned src0_swz_x:2; - unsigned src0_swz_y:2; - unsigned src0_subreg_nr:1; - unsigned src0_reg_nr:8; - unsigned src0_abs:1; - unsigned src0_negate:1; - unsigned src0_address_mode:1; - unsigned src0_swz_z:2; - unsigned src0_swz_w:2; - unsigned pad0:1; - unsigned src0_vert_stride:4; - unsigned flag_reg_nr:1; - unsigned pad1:6; - } da16; - - struct - { - unsigned src0_swz_x:2; - unsigned src0_swz_y:2; - int src0_indirect_offset:6; - unsigned src0_subreg_nr:3; - unsigned src0_abs:1; - unsigned src0_negate:1; - unsigned src0_address_mode:1; - unsigned src0_swz_z:2; - unsigned src0_swz_w:2; - unsigned pad0:1; - unsigned src0_vert_stride:4; - unsigned flag_reg_nr:1; - unsigned pad1:6; - } ia16; - - } bits2; - - union - { - struct - { - unsigned src1_subreg_nr:5; - unsigned src1_reg_nr:8; - unsigned src1_abs:1; - unsigned src1_negate:1; - unsigned pad:1; - unsigned src1_horiz_stride:2; - unsigned src1_width:3; - unsigned src1_vert_stride:4; - unsigned pad0:7; - } da1; - - struct - { - unsigned src1_swz_x:2; - unsigned src1_swz_y:2; - unsigned src1_subreg_nr:1; - unsigned src1_reg_nr:8; - unsigned src1_abs:1; - unsigned src1_negate:1; - unsigned pad0:1; - unsigned src1_swz_z:2; - unsigned src1_swz_w:2; - unsigned pad1:1; - unsigned src1_vert_stride:4; - unsigned pad2:7; - } da16; - - struct - { - int src1_indirect_offset:10; - unsigned src1_subreg_nr:3; - unsigned src1_abs:1; - unsigned src1_negate:1; - unsigned pad0:1; - unsigned src1_horiz_stride:2; - unsigned src1_width:3; - unsigned src1_vert_stride:4; - unsigned flag_reg_nr:1; - unsigned pad1:6; - } ia1; - - struct - { - unsigned src1_swz_x:2; - unsigned src1_swz_y:2; - int src1_indirect_offset:6; - unsigned src1_subreg_nr:3; - unsigned src1_abs:1; - unsigned src1_negate:1; - unsigned pad0:1; - unsigned src1_swz_z:2; - unsigned src1_swz_w:2; - unsigned pad1:1; - unsigned src1_vert_stride:4; - unsigned flag_reg_nr:1; - unsigned pad2:6; - } ia16; - - - struct - { - int jump_count:16; /* note: signed */ - unsigned pop_count:4; - unsigned pad0:12; - } if_else; - - struct { - unsigned function:4; - unsigned int_type:1; - unsigned precision:1; - unsigned saturate:1; - unsigned data_type:1; - unsigned pad0:8; - unsigned response_length:4; - unsigned msg_length:4; - unsigned msg_target:4; - unsigned pad1:3; - unsigned end_of_thread:1; - } math; - - struct { - unsigned binding_table_index:8; - unsigned sampler:4; - unsigned return_format:2; - unsigned msg_type:2; - unsigned response_length:4; - unsigned msg_length:4; - unsigned msg_target:4; - unsigned pad1:3; - unsigned end_of_thread:1; - } sampler; - - struct brw_urb_immediate urb; - - struct { - unsigned binding_table_index:8; - unsigned msg_control:4; - unsigned msg_type:2; - unsigned target_cache:2; - unsigned response_length:4; - unsigned msg_length:4; - unsigned msg_target:4; - unsigned pad1:3; - unsigned end_of_thread:1; - } dp_read; - - struct { - unsigned binding_table_index:8; - unsigned msg_control:3; - unsigned pixel_scoreboard_clear:1; - unsigned msg_type:3; - unsigned send_commit_msg:1; - unsigned response_length:4; - unsigned msg_length:4; - unsigned msg_target:4; - unsigned pad1:3; - unsigned end_of_thread:1; - } dp_write; - - struct { - unsigned pad:16; - unsigned response_length:4; - unsigned msg_length:4; - unsigned msg_target:4; - unsigned pad1:3; - unsigned end_of_thread:1; - } generic; - - int d; - unsigned ud; - } bits3; -}; - - -#endif diff --git a/src/mesa/pipe/i965simple/brw_surface.c b/src/mesa/pipe/i965simple/brw_surface.c deleted file mode 100644 index 518845e4b2..0000000000 --- a/src/mesa/pipe/i965simple/brw_surface.c +++ /dev/null @@ -1,210 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "brw_blit.h" -#include "brw_context.h" -#include "brw_state.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" -#include "pipe/util/p_tile.h" - - -/* - * XXX note: same as code in sp_surface.c - */ -static struct pipe_surface * -brw_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - struct brw_texture *tex = (struct brw_texture *)pt; - struct pipe_surface *ps; - unsigned offset; /* in bytes */ - - offset = tex->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE) { - offset += tex->image_offset[level][face] * pt->cpp; - } - else if (pt->target == PIPE_TEXTURE_3D) { - offset += tex->image_offset[level][zslice] * pt->cpp; - } - else { - assert(face == 0); - assert(zslice == 0); - } - - ps = pipe->winsys->surface_alloc(pipe->winsys); - if (ps) { - assert(ps->format); - assert(ps->refcount); - pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); - ps->format = pt->format; - ps->cpp = pt->cpp; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->pitch = tex->pitch; - ps->offset = offset; - } - return ps; -} - - -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -static void -brw_surface_data(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - const void *src, unsigned src_pitch, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - pipe_copy_rect(pipe_surface_map(dst) + dst->offset, - dst->cpp, dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - pipe_surface_unmap(dst); -} - - -/* Assumes all values are within bounds -- no checking at this level - - * do it higher up if required. - */ -static void -brw_surface_copy(struct pipe_context *pipe, - unsigned do_flip, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - struct pipe_surface *src, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - assert(dst != src); - assert(dst->cpp == src->cpp); - - if (0) { - pipe_copy_rect(pipe_surface_map(dst) + dst->offset, - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - pipe_surface_map(src) + src->offset, - do_flip ? -src->pitch : src->pitch, - srcx, do_flip ? 1 - srcy - height : srcy); - - pipe_surface_unmap(src); - pipe_surface_unmap(dst); - } - else { - brw_copy_blit(brw_context(pipe), - do_flip, - dst->cpp, - (short) src->pitch, src->buffer, src->offset, FALSE, - (short) dst->pitch, dst->buffer, dst->offset, FALSE, - (short) srcx, (short) srcy, (short) dstx, (short) dsty, - (short) width, (short) height, PIPE_LOGICOP_COPY); - } -} - -/* Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static void * -get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) -{ - return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; -} - - -static void -brw_surface_fill(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - unsigned width, unsigned height, unsigned value) -{ - if (0) { - unsigned i, j; - void *dst_map = pipe_surface_map(dst); - - switch (dst->cpp) { - case 1: { - ubyte *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; - case 2: { - ushort *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; - case 4: { - unsigned *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - break; - default: - assert(0); - break; - } - - pipe_surface_unmap( dst ); - } - else { - brw_fill_blit(brw_context(pipe), - dst->cpp, - (short) dst->pitch, - dst->buffer, dst->offset, FALSE, - (short) dstx, (short) dsty, - (short) width, (short) height, - value); - } -} - -void -brw_init_surface_functions(struct brw_context *brw) -{ - brw->pipe.get_tex_surface = brw_get_tex_surface; - brw->pipe.surface_copy = brw_surface_copy; - brw->pipe.surface_fill = brw_surface_fill; -} diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.c b/src/mesa/pipe/i965simple/brw_tex_layout.c deleted file mode 100644 index 90561f1307..0000000000 --- a/src/mesa/pipe/i965simple/brw_tex_layout.c +++ /dev/null @@ -1,353 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -/* Code to layout images in a mipmap tree for i965. - */ - -#include "brw_tex_layout.h" - -#include "pipe/p_state.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" - -#include "brw_context.h" - -#define FILE_DEBUG_FLAG DEBUG_TEXTURE - -#if 0 -unsigned intel_compressed_alignment(unsigned internalFormat) -{ - unsigned alignment = 4; - - switch (internalFormat) { - case GL_COMPRESSED_RGB_FXT1_3DFX: - case GL_COMPRESSED_RGBA_FXT1_3DFX: - alignment = 8; - break; - - default: - break; - } - - return alignment; -} -#endif - -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - - -static boolean brw_miptree_layout(struct pipe_context *, struct brw_texture *); - -static void intel_miptree_set_image_offset(struct brw_texture *tex, - unsigned level, - unsigned img, - unsigned x, unsigned y) -{ - struct pipe_texture *pt = &tex->base; - if (img == 0 && level == 0) - assert(x == 0 && y == 0); - assert(img < tex->nr_images[level]); - - tex->image_offset[level][img] = (x + y * tex->pitch) * pt->cpp; -} - -static void intel_miptree_set_level_info(struct brw_texture *tex, - unsigned level, - unsigned nr_images, - unsigned x, unsigned y, - unsigned w, unsigned h, unsigned d) -{ - struct pipe_texture *pt = &tex->base; - - assert(level < PIPE_MAX_TEXTURE_LEVELS); - - pt->width[level] = w; - pt->height[level] = h; - pt->depth[level] = d; - - tex->level_offset[level] = (x + y * tex->pitch) * pt->cpp; - tex->nr_images[level] = nr_images; - - /* - DBG("%s level %d size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - level, w, h, d, x, y, tex->level_offset[level]); - */ - - /* Not sure when this would happen, but anyway: - */ - if (tex->image_offset[level]) { - FREE(tex->image_offset[level]); - tex->image_offset[level] = NULL; - } - - assert(nr_images); - assert(!tex->image_offset[level]); - - tex->image_offset[level] = (unsigned *) MALLOC(nr_images * sizeof(unsigned)); - tex->image_offset[level][0] = 0; -} - -static void i945_miptree_layout_2d(struct brw_texture *tex) -{ - struct pipe_texture *pt = &tex->base; - unsigned align_h = 2, align_w = 4; - unsigned level; - unsigned x = 0; - unsigned y = 0; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - - tex->pitch = pt->width[0]; - -#if 0 - if (pt->compressed) { - align_w = intel_compressed_alignment(pt->internal_format); - tex->pitch = ALIGN(pt->width[0], align_w); - } -#endif - - /* May need to adjust pitch to accomodate the placement of - * the 2nd mipmap. This occurs when the alignment - * constraints of mipmap placement push the right edge of the - * 2nd mipmap out past the width of its parent. - */ - if (pt->last_level > 0) { - unsigned mip1_width; - - if (pt->compressed) { - mip1_width = align(minify(pt->width[0]), align_w) - + align(minify(minify(pt->width[0])), align_w); - } else { - mip1_width = align(minify(pt->width[0]), align_w) - + minify(minify(pt->width[0])); - } - - if (mip1_width > tex->pitch) { - tex->pitch = mip1_width; - } - } - - /* Pitch must be a whole number of dwords, even though we - * express it in texels. - */ - tex->pitch = align(tex->pitch * pt->cpp, 4) / pt->cpp; - tex->total_height = 0; - - for (level = 0; level <= pt->last_level; level++) { - unsigned img_height; - - intel_miptree_set_level_info(tex, level, 1, x, y, width, - height, 1); - - if (pt->compressed) - img_height = MAX2(1, height/4); - else - img_height = align(height, align_h); - - - /* Because the images are packed better, the final offset - * might not be the maximal one: - */ - tex->total_height = MAX2(tex->total_height, y + img_height); - - /* Layout_below: step right after second mipmap. - */ - if (level == 1) { - x += align(width, align_w); - } - else { - y += img_height; - } - - width = minify(width); - height = minify(height); - } -} - -static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture *tex) -{ - struct pipe_texture *pt = &tex->base; - /* XXX: these vary depending on image format: - */ -/* int align_w = 4; */ - - switch (pt->target) { - case PIPE_TEXTURE_CUBE: - case PIPE_TEXTURE_3D: { - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - unsigned pack_x_pitch, pack_x_nr; - unsigned pack_y_pitch; - unsigned level; - unsigned align_h = 2; - unsigned align_w = 4; - - tex->total_height = 0; -#if 0 - if (pt->compressed) { - align_w = intel_compressed_alignment(pt->internal_format); - pt->pitch = align(width, align_w); - pack_y_pitch = (height + 3) / 4; - } else -#endif - { - tex->pitch = align(pt->width[0] * pt->cpp, 4) / pt->cpp; - pack_y_pitch = align(pt->height[0], align_h); - } - - pack_x_pitch = tex->pitch; - pack_x_nr = 1; - - for (level = 0; level <= pt->last_level; level++) { - unsigned nr_images = pt->target == PIPE_TEXTURE_3D ? depth : 6; - int x = 0; - int y = 0; - uint q, j; - - intel_miptree_set_level_info(tex, level, nr_images, - 0, tex->total_height, - width, height, depth); - - for (q = 0; q < nr_images;) { - for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - intel_miptree_set_image_offset(tex, level, q, x, y); - x += pack_x_pitch; - } - - x = 0; - y += pack_y_pitch; - } - - - tex->total_height += y; - width = minify(width); - height = minify(height); - depth = minify(depth); - - if (pt->compressed) { - pack_y_pitch = (height + 3) / 4; - - if (pack_x_pitch > align(width, align_w)) { - pack_x_pitch = align(width, align_w); - pack_x_nr <<= 1; - } - } else { - if (pack_x_pitch > 4) { - pack_x_pitch >>= 1; - pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= tex->pitch); - } - - if (pack_y_pitch > 2) { - pack_y_pitch >>= 1; - pack_y_pitch = align(pack_y_pitch, align_h); - } - } - - } - break; - } - - default: - i945_miptree_layout_2d(tex); - break; - } -#if 0 - PRINT("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, - pt->pitch, - pt->total_height, - pt->cpp, - pt->pitch * pt->total_height * pt->cpp ); -#endif - - return TRUE; -} - - -struct pipe_texture * -brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) -{ - struct brw_texture *tex = CALLOC_STRUCT(brw_texture); - - if (tex) { - tex->base = *templat; - - if (brw_miptree_layout(pipe, tex)) - tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, - PIPE_BUFFER_USAGE_PIXEL, - tex->pitch * tex->base.cpp * - tex->total_height); - - if (!tex->buffer) { - FREE(tex); - return NULL; - } - } - - return &tex->base; -} - -void -brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) -{ - if (!*pt) - return; - - /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); - */ - if (--(*pt)->refcount <= 0) { - struct brw_texture *tex = (struct brw_texture *)*pt; - uint i; - - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); - */ - - pipe_buffer_reference(pipe->winsys, &tex->buffer, NULL); - - for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) - if (tex->image_offset[i]) - free(tex->image_offset[i]); - - free(tex); - } - *pt = NULL; -} diff --git a/src/mesa/pipe/i965simple/brw_tex_layout.h b/src/mesa/pipe/i965simple/brw_tex_layout.h deleted file mode 100644 index cfd6b1ef3a..0000000000 --- a/src/mesa/pipe/i965simple/brw_tex_layout.h +++ /dev/null @@ -1,15 +0,0 @@ -#ifndef BRW_TEX_LAYOUT_H -#define BRW_TEX_LAYOUT_H - -#include "pipe/p_compiler.h" - -struct pipe_context; -struct pipe_texture; - -extern struct pipe_texture * -brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat); - -extern void -brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); - -#endif diff --git a/src/mesa/pipe/i965simple/brw_urb.c b/src/mesa/pipe/i965simple/brw_urb.c deleted file mode 100644 index 101a4367b9..0000000000 --- a/src/mesa/pipe/i965simple/brw_urb.c +++ /dev/null @@ -1,186 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -//#include "brw_state.h" -#include "brw_batch.h" -#include "brw_defines.h" - -#define VS 0 -#define GS 1 -#define CLP 2 -#define SF 3 -#define CS 4 - -/* XXX: Are the min_entry_size numbers useful? - * XXX: Verify min_nr_entries, esp for VS. - * XXX: Verify SF min_entry_size. - */ -static const struct { - unsigned min_nr_entries; - unsigned preferred_nr_entries; - unsigned min_entry_size; - unsigned max_entry_size; -} limits[CS+1] = { - { 8, 32, 1, 5 }, /* vs */ - { 4, 8, 1, 5 }, /* gs */ - { 6, 8, 1, 5 }, /* clp */ - { 1, 8, 1, 12 }, /* sf */ - { 1, 4, 1, 32 } /* cs */ -}; - - -static boolean check_urb_layout( struct brw_context *brw ) -{ - brw->urb.vs_start = 0; - brw->urb.gs_start = brw->urb.nr_vs_entries * brw->urb.vsize; - brw->urb.clip_start = brw->urb.gs_start + brw->urb.nr_gs_entries * brw->urb.vsize; - brw->urb.sf_start = brw->urb.clip_start + brw->urb.nr_clip_entries * brw->urb.vsize; - brw->urb.cs_start = brw->urb.sf_start + brw->urb.nr_sf_entries * brw->urb.sfsize; - - return brw->urb.cs_start + brw->urb.nr_cs_entries * brw->urb.csize <= 256; -} - -/* Most minimal update, forces re-emit of URB fence packet after GS - * unit turned on/off. - */ -static void recalculate_urb_fence( struct brw_context *brw ) -{ - unsigned csize = brw->curbe.total_size; - unsigned vsize = brw->vs.prog_data->urb_entry_size; - unsigned sfsize = brw->sf.prog_data->urb_entry_size; - - if (csize < limits[CS].min_entry_size) - csize = limits[CS].min_entry_size; - - if (vsize < limits[VS].min_entry_size) - vsize = limits[VS].min_entry_size; - - if (sfsize < limits[SF].min_entry_size) - sfsize = limits[SF].min_entry_size; - - if (brw->urb.vsize < vsize || - brw->urb.sfsize < sfsize || - brw->urb.csize < csize || - (brw->urb.constrained && (brw->urb.vsize > brw->urb.vsize || - brw->urb.sfsize > brw->urb.sfsize || - brw->urb.csize > brw->urb.csize))) { - - - brw->urb.csize = csize; - brw->urb.sfsize = sfsize; - brw->urb.vsize = vsize; - - brw->urb.nr_vs_entries = limits[VS].preferred_nr_entries; - brw->urb.nr_gs_entries = limits[GS].preferred_nr_entries; - brw->urb.nr_clip_entries = limits[CLP].preferred_nr_entries; - brw->urb.nr_sf_entries = limits[SF].preferred_nr_entries; - brw->urb.nr_cs_entries = limits[CS].preferred_nr_entries; - - if (!check_urb_layout(brw)) { - brw->urb.nr_vs_entries = limits[VS].min_nr_entries; - brw->urb.nr_gs_entries = limits[GS].min_nr_entries; - brw->urb.nr_clip_entries = limits[CLP].min_nr_entries; - brw->urb.nr_sf_entries = limits[SF].min_nr_entries; - brw->urb.nr_cs_entries = limits[CS].min_nr_entries; - - brw->urb.constrained = 1; - - if (!check_urb_layout(brw)) { - /* This is impossible, given the maximal sizes of urb - * entries and the values for minimum nr of entries - * provided above. - */ - debug_printf("couldn't calculate URB layout!\n"); - exit(1); - } - - if (BRW_DEBUG & (DEBUG_URB|DEBUG_FALLBACKS)) - debug_printf("URB CONSTRAINED\n"); - } - else - brw->urb.constrained = 0; - - if (BRW_DEBUG & DEBUG_URB) - debug_printf("URB fence: %d ..VS.. %d ..GS.. %d ..CLP.. %d ..SF.. %d ..CS.. %d\n", - brw->urb.vs_start, - brw->urb.gs_start, - brw->urb.clip_start, - brw->urb.sf_start, - brw->urb.cs_start, - 256); - - brw->state.dirty.brw |= BRW_NEW_URB_FENCE; - } -} - - -const struct brw_tracked_state brw_recalculate_urb_fence = { - .dirty = { - .brw = BRW_NEW_CURBE_OFFSETS, - .cache = (CACHE_NEW_VS_PROG | - CACHE_NEW_SF_PROG) - }, - .update = recalculate_urb_fence -}; - - - - - -void brw_upload_urb_fence(struct brw_context *brw) -{ - struct brw_urb_fence uf; - memset(&uf, 0, sizeof(uf)); - - uf.header.opcode = CMD_URB_FENCE; - uf.header.length = sizeof(uf)/4-2; - uf.header.vs_realloc = 1; - uf.header.gs_realloc = 1; - uf.header.clp_realloc = 1; - uf.header.sf_realloc = 1; - uf.header.vfe_realloc = 1; - uf.header.cs_realloc = 1; - - /* The ordering below is correct, not the layout in the - * instruction. - * - * There are 256 urb reg pairs in total. - */ - uf.bits0.vs_fence = brw->urb.gs_start; - uf.bits0.gs_fence = brw->urb.clip_start; - uf.bits0.clp_fence = brw->urb.sf_start; - uf.bits1.sf_fence = brw->urb.cs_start; - uf.bits1.cs_fence = 256; - - BRW_BATCH_STRUCT(brw, &uf); -} diff --git a/src/mesa/pipe/i965simple/brw_util.c b/src/mesa/pipe/i965simple/brw_util.c deleted file mode 100644 index 42391d7c8c..0000000000 --- a/src/mesa/pipe/i965simple/brw_util.c +++ /dev/null @@ -1,104 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_util.h" -#include "brw_defines.h" - -#include "pipe/p_defines.h" - -unsigned brw_count_bits( unsigned val ) -{ - unsigned i; - for (i = 0; val ; val >>= 1) - if (val & 1) - i++; - return i; -} - - -unsigned brw_translate_blend_equation( int mode ) -{ - switch (mode) { - case PIPE_BLEND_ADD: - return BRW_BLENDFUNCTION_ADD; - case PIPE_BLEND_MIN: - return BRW_BLENDFUNCTION_MIN; - case PIPE_BLEND_MAX: - return BRW_BLENDFUNCTION_MAX; - case PIPE_BLEND_SUBTRACT: - return BRW_BLENDFUNCTION_SUBTRACT; - case PIPE_BLEND_REVERSE_SUBTRACT: - return BRW_BLENDFUNCTION_REVERSE_SUBTRACT; - default: - assert(0); - return BRW_BLENDFUNCTION_ADD; - } -} - -unsigned brw_translate_blend_factor( int factor ) -{ - switch(factor) { - case PIPE_BLENDFACTOR_ZERO: - return BRW_BLENDFACTOR_ZERO; - case PIPE_BLENDFACTOR_SRC_ALPHA: - return BRW_BLENDFACTOR_SRC_ALPHA; - case PIPE_BLENDFACTOR_ONE: - return BRW_BLENDFACTOR_ONE; - case PIPE_BLENDFACTOR_SRC_COLOR: - return BRW_BLENDFACTOR_SRC_COLOR; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - return BRW_BLENDFACTOR_INV_SRC_COLOR; - case PIPE_BLENDFACTOR_DST_COLOR: - return BRW_BLENDFACTOR_DST_COLOR; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - return BRW_BLENDFACTOR_INV_DST_COLOR; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - return BRW_BLENDFACTOR_INV_SRC_ALPHA; - case PIPE_BLENDFACTOR_DST_ALPHA: - return BRW_BLENDFACTOR_DST_ALPHA; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - return BRW_BLENDFACTOR_INV_DST_ALPHA; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - return BRW_BLENDFACTOR_SRC_ALPHA_SATURATE; - case PIPE_BLENDFACTOR_CONST_COLOR: - return BRW_BLENDFACTOR_CONST_COLOR; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - return BRW_BLENDFACTOR_INV_CONST_COLOR; - case PIPE_BLENDFACTOR_CONST_ALPHA: - return BRW_BLENDFACTOR_CONST_ALPHA; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - return BRW_BLENDFACTOR_INV_CONST_ALPHA; - default: - assert(0); - return BRW_BLENDFACTOR_ZERO; - } -} diff --git a/src/mesa/pipe/i965simple/brw_util.h b/src/mesa/pipe/i965simple/brw_util.h deleted file mode 100644 index d60e5934db..0000000000 --- a/src/mesa/pipe/i965simple/brw_util.h +++ /dev/null @@ -1,43 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRW_UTIL_H -#define BRW_UTIL_H - -#include "pipe/p_state.h" - -extern unsigned brw_count_bits( unsigned val ); -extern unsigned brw_translate_blend_factor( int factor ); -extern unsigned brw_translate_blend_equation( int mode ); - - -#endif diff --git a/src/mesa/pipe/i965simple/brw_vs.c b/src/mesa/pipe/i965simple/brw_vs.c deleted file mode 100644 index 738c6346d5..0000000000 --- a/src/mesa/pipe/i965simple/brw_vs.c +++ /dev/null @@ -1,120 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_vs.h" -#include "brw_util.h" -#include "brw_state.h" - - -static void do_vs_prog( struct brw_context *brw, - const struct brw_vertex_program *vp, - struct brw_vs_prog_key *key ) -{ - unsigned program_size; - const unsigned *program; - struct brw_vs_compile c; - - memset(&c, 0, sizeof(c)); - memcpy(&c.key, key, sizeof(*key)); - - brw_init_compile(&c.func); - c.vp = vp; - - c.prog_data.outputs_written = vp->program.num_outputs; - c.prog_data.inputs_read = vp->program.num_inputs; - -#if 0 - if (c.key.copy_edgeflag) { - c.prog_data.outputs_written |= 1<vs.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_VS_PROG], - &c.key, - sizeof(c.key), - program, - program_size, - &c.prog_data, - &brw->vs.prog_data); -} - - -static void brw_upload_vs_prog( struct brw_context *brw ) -{ - struct brw_vs_prog_key key; - const struct brw_vertex_program *vp = brw->attribs.VertexProgram; - - assert(vp); - - memset(&key, 0, sizeof(key)); - - /* Just upload the program verbatim for now. Always send it all - * the inputs it asks for, whether they are varying or not. - */ - key.program_string_id = vp->id; - key.nr_userclip = brw->attribs.Clip.nr; - key.copy_edgeflag = (brw->attribs.Raster->fill_cw != PIPE_POLYGON_MODE_FILL || - brw->attribs.Raster->fill_ccw != PIPE_POLYGON_MODE_FILL); - - /* Make an early check for the key. - */ - if (brw_search_cache(&brw->cache[BRW_VS_PROG], - &key, sizeof(key), - &brw->vs.prog_data, - &brw->vs.prog_gs_offset)) - return; - - do_vs_prog(brw, vp, &key); -} - - -/* See brw_vs.c: - */ -const struct brw_tracked_state brw_vs_prog = { - .dirty = { - .brw = BRW_NEW_VS, - .cache = 0 - }, - .update = brw_upload_vs_prog -}; diff --git a/src/mesa/pipe/i965simple/brw_vs.h b/src/mesa/pipe/i965simple/brw_vs.h deleted file mode 100644 index 0e58f043b0..0000000000 --- a/src/mesa/pipe/i965simple/brw_vs.h +++ /dev/null @@ -1,82 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRW_VS_H -#define BRW_VS_H - - -#include "brw_context.h" -#include "brw_eu.h" - - -struct brw_vs_prog_key { - unsigned program_string_id; - unsigned nr_userclip:4; - unsigned copy_edgeflag:1; - unsigned know_w_is_one:1; - unsigned pad:26; -}; - - -struct brw_vs_compile { - struct brw_compile func; - struct brw_vs_prog_key key; - struct brw_vs_prog_data prog_data; - - struct brw_vertex_program *vp; - - unsigned nr_inputs; - - unsigned first_output; - unsigned nr_outputs; - - unsigned first_tmp; - unsigned last_tmp; - - struct brw_reg r0; - struct brw_reg r1; - struct brw_reg regs[12][128]; - struct brw_reg tmp; - struct brw_reg stack; - - struct { - boolean used_in_src; - struct brw_reg reg; - } output_regs[128]; - - struct brw_reg userplane[6]; - -}; - -void brw_vs_emit( struct brw_vs_compile *c ); - -#endif diff --git a/src/mesa/pipe/i965simple/brw_vs_emit.c b/src/mesa/pipe/i965simple/brw_vs_emit.c deleted file mode 100644 index 98915ba101..0000000000 --- a/src/mesa/pipe/i965simple/brw_vs_emit.c +++ /dev/null @@ -1,1332 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_context.h" -#include "brw_vs.h" - -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" - -struct brw_prog_info { - unsigned num_temps; - unsigned num_addrs; - unsigned num_consts; - - unsigned writes_psize; - - unsigned pos_idx; - unsigned result_edge_idx; - unsigned edge_flag_idx; - unsigned psize_idx; -}; - -/* Do things as simply as possible. Allocate and populate all regs - * ahead of time. - */ -static void brw_vs_alloc_regs( struct brw_vs_compile *c, - struct brw_prog_info *info ) -{ - unsigned i, reg = 0, mrf; - unsigned nr_params; - - /* r0 -- reserved as usual - */ - c->r0 = brw_vec8_grf(reg, 0); reg++; - - /* User clip planes from curbe: - */ - if (c->key.nr_userclip) { - for (i = 0; i < c->key.nr_userclip; i++) { - c->userplane[i] = stride( brw_vec4_grf(reg+3+i/2, (i%2) * 4), 0, 4, 1); - } - - /* Deal with curbe alignment: - */ - reg += ((6+c->key.nr_userclip+3)/4)*2; - } - - /* Vertex program parameters from curbe: - */ - nr_params = c->prog_data.max_const; - for (i = 0; i < nr_params; i++) { - c->regs[TGSI_FILE_CONSTANT][i] = stride(brw_vec4_grf(reg+i/2, (i%2) * 4), 0, 4, 1); - } - reg += (nr_params+1)/2; - c->prog_data.curb_read_length = reg - 1; - - - - /* Allocate input regs: - */ - c->nr_inputs = c->vp->program.num_inputs; - for (i = 0; i < c->nr_inputs; i++) { - c->regs[TGSI_FILE_INPUT][i] = brw_vec8_grf(reg, 0); - reg++; - } - - - /* Allocate outputs: TODO: could organize the non-position outputs - * to go straight into message regs. - */ - c->nr_outputs = 0; - c->first_output = reg; - mrf = 4; - for (i = 0; i < c->vp->program.num_outputs; i++) { - c->nr_outputs++; -#if 0 - if (i == VERT_RESULT_HPOS) { - c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); - reg++; - } - else if (i == VERT_RESULT_PSIZ) { - c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); - reg++; - mrf++; /* just a placeholder? XXX fix later stages & remove this */ - } - else { - c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf); - mrf++; - } -#else - /*treat pos differently for now */ - if (i == info->pos_idx) { - c->regs[TGSI_FILE_OUTPUT][i] = brw_vec8_grf(reg, 0); - reg++; - } else { - c->regs[TGSI_FILE_OUTPUT][i] = brw_message_reg(mrf); - mrf++; - } -#endif - } - - /* Allocate program temporaries: - */ - for (i = 0; i < info->num_temps; i++) { - c->regs[TGSI_FILE_TEMPORARY][i] = brw_vec8_grf(reg, 0); - reg++; - } - - /* Address reg(s). Don't try to use the internal address reg until - * deref time. - */ - for (i = 0; i < info->num_addrs; i++) { - c->regs[TGSI_FILE_ADDRESS][i] = brw_reg(BRW_GENERAL_REGISTER_FILE, - reg, - 0, - BRW_REGISTER_TYPE_D, - BRW_VERTICAL_STRIDE_8, - BRW_WIDTH_8, - BRW_HORIZONTAL_STRIDE_1, - BRW_SWIZZLE_XXXX, - TGSI_WRITEMASK_X); - reg++; - } - - for (i = 0; i < 128; i++) { - if (c->output_regs[i].used_in_src) { - c->output_regs[i].reg = brw_vec8_grf(reg, 0); - reg++; - } - } - - c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, reg, 0); - reg += 2; - - - /* Some opcodes need an internal temporary: - */ - c->first_tmp = reg; - c->last_tmp = reg; /* for allocation purposes */ - - /* Each input reg holds data from two vertices. The - * urb_read_length is the number of registers read from *each* - * vertex urb, so is half the amount: - */ - c->prog_data.urb_read_length = (c->nr_inputs+1)/2; - - c->prog_data.urb_entry_size = (c->nr_outputs+2+3)/4; - c->prog_data.total_grf = reg; -} - - -static struct brw_reg get_tmp( struct brw_vs_compile *c ) -{ - struct brw_reg tmp = brw_vec8_grf(c->last_tmp, 0); - - if (++c->last_tmp > c->prog_data.total_grf) - c->prog_data.total_grf = c->last_tmp; - - return tmp; -} - -static void release_tmp( struct brw_vs_compile *c, struct brw_reg tmp ) -{ - if (tmp.nr == c->last_tmp-1) - c->last_tmp--; -} - -static void release_tmps( struct brw_vs_compile *c ) -{ - c->last_tmp = c->first_tmp; -} - - -static void unalias1( struct brw_vs_compile *c, - struct brw_reg dst, - struct brw_reg arg0, - void (*func)( struct brw_vs_compile *, - struct brw_reg, - struct brw_reg )) -{ - if (dst.file == arg0.file && dst.nr == arg0.nr) { - struct brw_compile *p = &c->func; - struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask); - func(c, tmp, arg0); - brw_MOV(p, dst, tmp); - } - else { - func(c, dst, arg0); - } -} - -static void unalias2( struct brw_vs_compile *c, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1, - void (*func)( struct brw_vs_compile *, - struct brw_reg, - struct brw_reg, - struct brw_reg )) -{ - if ((dst.file == arg0.file && dst.nr == arg0.nr) || - (dst.file == arg1.file && dst.nr == arg1.nr)) { - struct brw_compile *p = &c->func; - struct brw_reg tmp = brw_writemask(get_tmp(c), dst.dw1.bits.writemask); - func(c, tmp, arg0, arg1); - brw_MOV(p, dst, tmp); - } - else { - func(c, dst, arg0, arg1); - } -} - -static void emit_sop( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1, - unsigned cond) -{ - brw_push_insn_state(p); - brw_CMP(p, brw_null_reg(), cond, arg0, arg1); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_MOV(p, dst, brw_imm_f(1.0f)); - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_MOV(p, dst, brw_imm_f(0.0f)); - brw_pop_insn_state(p); -} - -static void emit_seq( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1 ) -{ - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_EQ); -} - -static void emit_sne( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1 ) -{ - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_NEQ); -} -static void emit_slt( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1 ) -{ - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_L); -} - -static void emit_sle( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1 ) -{ - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_LE); -} - -static void emit_sgt( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1 ) -{ - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_G); -} - -static void emit_sge( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1 ) -{ - emit_sop(p, dst, arg0, arg1, BRW_CONDITIONAL_GE); -} - -static void emit_max( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1 ) -{ - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1); - brw_SEL(p, dst, arg1, arg0); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); -} - -static void emit_min( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1 ) -{ - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, arg0, arg1); - brw_SEL(p, dst, arg0, arg1); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); -} - - -static void emit_math1( struct brw_vs_compile *c, - unsigned function, - struct brw_reg dst, - struct brw_reg arg0, - unsigned precision) -{ - /* There are various odd behaviours with SEND on the simulator. In - * addition there are documented issues with the fact that the GEN4 - * processor doesn't do dependency control properly on SEND - * results. So, on balance, this kludge to get around failures - * with writemasked math results looks like it might be necessary - * whether that turns out to be a simulator bug or not: - */ - struct brw_compile *p = &c->func; - struct brw_reg tmp = dst; - boolean need_tmp = (dst.dw1.bits.writemask != 0xf || - dst.file != BRW_GENERAL_REGISTER_FILE); - - if (need_tmp) - tmp = get_tmp(c); - - brw_math(p, - tmp, - function, - BRW_MATH_SATURATE_NONE, - 2, - arg0, - BRW_MATH_DATA_SCALAR, - precision); - - if (need_tmp) { - brw_MOV(p, dst, tmp); - release_tmp(c, tmp); - } -} - -static void emit_math2( struct brw_vs_compile *c, - unsigned function, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1, - unsigned precision) -{ - struct brw_compile *p = &c->func; - struct brw_reg tmp = dst; - boolean need_tmp = (dst.dw1.bits.writemask != 0xf || - dst.file != BRW_GENERAL_REGISTER_FILE); - - if (need_tmp) - tmp = get_tmp(c); - - brw_MOV(p, brw_message_reg(3), arg1); - - brw_math(p, - tmp, - function, - BRW_MATH_SATURATE_NONE, - 2, - arg0, - BRW_MATH_DATA_SCALAR, - precision); - - if (need_tmp) { - brw_MOV(p, dst, tmp); - release_tmp(c, tmp); - } -} - - - -static void emit_exp_noalias( struct brw_vs_compile *c, - struct brw_reg dst, - struct brw_reg arg0 ) -{ - struct brw_compile *p = &c->func; - - - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_X) { - struct brw_reg tmp = get_tmp(c); - struct brw_reg tmp_d = retype(tmp, BRW_REGISTER_TYPE_D); - - /* tmp_d = floor(arg0.x) */ - brw_RNDD(p, tmp_d, brw_swizzle1(arg0, 0)); - - /* result[0] = 2.0 ^ tmp */ - - /* Adjust exponent for floating point: - * exp += 127 - */ - brw_ADD(p, brw_writemask(tmp_d, TGSI_WRITEMASK_X), tmp_d, brw_imm_d(127)); - - /* Install exponent and sign. - * Excess drops off the edge: - */ - brw_SHL(p, brw_writemask(retype(dst, BRW_REGISTER_TYPE_D), TGSI_WRITEMASK_X), - tmp_d, brw_imm_d(23)); - - release_tmp(c, tmp); - } - - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Y) { - /* result[1] = arg0.x - floor(arg0.x) */ - brw_FRC(p, brw_writemask(dst, TGSI_WRITEMASK_Y), brw_swizzle1(arg0, 0)); - } - - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) { - /* As with the LOG instruction, we might be better off just - * doing a taylor expansion here, seeing as we have to do all - * the prep work. - * - * If mathbox partial precision is too low, consider also: - * result[3] = result[0] * EXP(result[1]) - */ - emit_math1(c, - BRW_MATH_FUNCTION_EXP, - brw_writemask(dst, TGSI_WRITEMASK_Z), - brw_swizzle1(arg0, 0), - BRW_MATH_PRECISION_PARTIAL); - } - - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) { - /* result[3] = 1.0; */ - brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_W), brw_imm_f(1)); - } -} - - -static void emit_log_noalias( struct brw_vs_compile *c, - struct brw_reg dst, - struct brw_reg arg0 ) -{ - struct brw_compile *p = &c->func; - struct brw_reg tmp = dst; - struct brw_reg tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD); - struct brw_reg arg0_ud = retype(arg0, BRW_REGISTER_TYPE_UD); - boolean need_tmp = (dst.dw1.bits.writemask != 0xf || - dst.file != BRW_GENERAL_REGISTER_FILE); - - if (need_tmp) { - tmp = get_tmp(c); - tmp_ud = retype(tmp, BRW_REGISTER_TYPE_UD); - } - - /* Perform mant = frexpf(fabsf(x), &exp), adjust exp and mnt - * according to spec: - * - * These almost look likey they could be joined up, but not really - * practical: - * - * result[0].f = (x.i & ((1<<31)-1) >> 23) - 127 - * result[1].i = (x.i & ((1<<23)-1) + (127<<23) - */ - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_XZ) { - brw_AND(p, - brw_writemask(tmp_ud, TGSI_WRITEMASK_X), - brw_swizzle1(arg0_ud, 0), - brw_imm_ud((1U<<31)-1)); - - brw_SHR(p, - brw_writemask(tmp_ud, TGSI_WRITEMASK_X), - tmp_ud, - brw_imm_ud(23)); - - brw_ADD(p, - brw_writemask(tmp, TGSI_WRITEMASK_X), - retype(tmp_ud, BRW_REGISTER_TYPE_D), /* does it matter? */ - brw_imm_d(-127)); - } - - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_YZ) { - brw_AND(p, - brw_writemask(tmp_ud, TGSI_WRITEMASK_Y), - brw_swizzle1(arg0_ud, 0), - brw_imm_ud((1<<23)-1)); - - brw_OR(p, - brw_writemask(tmp_ud, TGSI_WRITEMASK_Y), - tmp_ud, - brw_imm_ud(127<<23)); - } - - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) { - /* result[2] = result[0] + LOG2(result[1]); */ - - /* Why bother? The above is just a hint how to do this with a - * taylor series. Maybe we *should* use a taylor series as by - * the time all the above has been done it's almost certainly - * quicker than calling the mathbox, even with low precision. - * - * Options are: - * - result[0] + mathbox.LOG2(result[1]) - * - mathbox.LOG2(arg0.x) - * - result[0] + inline_taylor_approx(result[1]) - */ - emit_math1(c, - BRW_MATH_FUNCTION_LOG, - brw_writemask(tmp, TGSI_WRITEMASK_Z), - brw_swizzle1(tmp, 1), - BRW_MATH_PRECISION_FULL); - - brw_ADD(p, - brw_writemask(tmp, TGSI_WRITEMASK_Z), - brw_swizzle1(tmp, 2), - brw_swizzle1(tmp, 0)); - } - - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) { - /* result[3] = 1.0; */ - brw_MOV(p, brw_writemask(tmp, TGSI_WRITEMASK_W), brw_imm_f(1)); - } - - if (need_tmp) { - brw_MOV(p, dst, tmp); - release_tmp(c, tmp); - } -} - - - - -/* Need to unalias - consider swizzles: r0 = DST r0.xxxx r1 - */ -static void emit_dst_noalias( struct brw_vs_compile *c, - struct brw_reg dst, - struct brw_reg arg0, - struct brw_reg arg1) -{ - struct brw_compile *p = &c->func; - - /* There must be a better way to do this: - */ - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_X) - brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_X), brw_imm_f(1.0)); - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Y) - brw_MUL(p, brw_writemask(dst, TGSI_WRITEMASK_Y), arg0, arg1); - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_Z) - brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_Z), arg0); - if (dst.dw1.bits.writemask & TGSI_WRITEMASK_W) - brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_W), arg1); -} - -static void emit_xpd( struct brw_compile *p, - struct brw_reg dst, - struct brw_reg t, - struct brw_reg u) -{ - brw_MUL(p, brw_null_reg(), brw_swizzle(t, 1,2,0,3), brw_swizzle(u,2,0,1,3)); - brw_MAC(p, dst, negate(brw_swizzle(t, 2,0,1,3)), brw_swizzle(u,1,2,0,3)); -} - - - -static void emit_lit_noalias( struct brw_vs_compile *c, - struct brw_reg dst, - struct brw_reg arg0 ) -{ - struct brw_compile *p = &c->func; - struct brw_instruction *if_insn; - struct brw_reg tmp = dst; - boolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE); - - if (need_tmp) - tmp = get_tmp(c); - - brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_YZ), brw_imm_f(0)); - brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_XW), brw_imm_f(1)); - - /* Need to use BRW_EXECUTE_8 and also do an 8-wide compare in order - * to get all channels active inside the IF. In the clipping code - * we run with NoMask, so it's not an option and we can use - * BRW_EXECUTE_1 for all comparisions. - */ - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,0), brw_imm_f(0)); - if_insn = brw_IF(p, BRW_EXECUTE_8); - { - brw_MOV(p, brw_writemask(dst, TGSI_WRITEMASK_Y), brw_swizzle1(arg0,0)); - - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_G, brw_swizzle1(arg0,1), brw_imm_f(0)); - brw_MOV(p, brw_writemask(tmp, TGSI_WRITEMASK_Z), brw_swizzle1(arg0,1)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - - emit_math2(c, - BRW_MATH_FUNCTION_POW, - brw_writemask(dst, TGSI_WRITEMASK_Z), - brw_swizzle1(tmp, 2), - brw_swizzle1(arg0, 3), - BRW_MATH_PRECISION_PARTIAL); - } - - brw_ENDIF(p, if_insn); -} - - - - - -/* TODO: relative addressing! - */ -static struct brw_reg get_reg( struct brw_vs_compile *c, - unsigned file, - unsigned index ) -{ - switch (file) { - case TGSI_FILE_TEMPORARY: - case TGSI_FILE_INPUT: - case TGSI_FILE_OUTPUT: - assert(c->regs[file][index].nr != 0); - return c->regs[file][index]; - case TGSI_FILE_CONSTANT: - assert(c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_imm].nr != 0); - return c->regs[TGSI_FILE_CONSTANT][index + c->prog_data.num_imm]; - case TGSI_FILE_IMMEDIATE: - assert(c->regs[TGSI_FILE_CONSTANT][index].nr != 0); - return c->regs[TGSI_FILE_CONSTANT][index]; - case TGSI_FILE_ADDRESS: - assert(index == 0); - return c->regs[file][index]; - - case TGSI_FILE_NULL: /* undef values */ - return brw_null_reg(); - - default: - assert(0); - return brw_null_reg(); - } -} - - - -static struct brw_reg deref( struct brw_vs_compile *c, - struct brw_reg arg, - int offset) -{ - struct brw_compile *p = &c->func; - struct brw_reg tmp = vec4(get_tmp(c)); - struct brw_reg vp_address = retype(vec1(get_reg(c, TGSI_FILE_ADDRESS, 0)), BRW_REGISTER_TYPE_UW); - unsigned byte_offset = arg.nr * 32 + arg.subnr + offset * 16; - struct brw_reg indirect = brw_vec4_indirect(0,0); - - { - brw_push_insn_state(p); - brw_set_access_mode(p, BRW_ALIGN_1); - - /* This is pretty clunky - load the address register twice and - * fetch each 4-dword value in turn. There must be a way to do - * this in a single pass, but I couldn't get it to work. - */ - brw_ADD(p, brw_address_reg(0), vp_address, brw_imm_d(byte_offset)); - brw_MOV(p, tmp, indirect); - - brw_ADD(p, brw_address_reg(0), suboffset(vp_address, 8), brw_imm_d(byte_offset)); - brw_MOV(p, suboffset(tmp, 4), indirect); - - brw_pop_insn_state(p); - } - - return vec8(tmp); -} - - -static void emit_arl( struct brw_vs_compile *c, - struct brw_reg dst, - struct brw_reg arg0 ) -{ - struct brw_compile *p = &c->func; - struct brw_reg tmp = dst; - boolean need_tmp = (dst.file != BRW_GENERAL_REGISTER_FILE); - - if (need_tmp) - tmp = get_tmp(c); - - brw_RNDD(p, tmp, arg0); - brw_MUL(p, dst, tmp, brw_imm_d(16)); - - if (need_tmp) - release_tmp(c, tmp); -} - - -/* Will return mangled results for SWZ op. The emit_swz() function - * ignores this result and recalculates taking extended swizzles into - * account. - */ -static struct brw_reg get_arg( struct brw_vs_compile *c, - struct tgsi_src_register *src ) -{ - struct brw_reg reg; - - if (src->File == TGSI_FILE_NULL) - return brw_null_reg(); - -#if 0 - if (src->RelAddr) - reg = deref(c, c->regs[PROGRAM_STATE_VAR][0], src->Index); - else -#endif - reg = get_reg(c, src->File, src->Index); - - /* Convert 3-bit swizzle to 2-bit. - */ - reg.dw1.bits.swizzle = BRW_SWIZZLE4(src->SwizzleX, - src->SwizzleY, - src->SwizzleZ, - src->SwizzleW); - - /* Note this is ok for non-swizzle instructions: - */ - reg.negate = src->Negate ? 1 : 0; - - return reg; -} - - -static struct brw_reg get_dst( struct brw_vs_compile *c, - const struct tgsi_dst_register *dst ) -{ - struct brw_reg reg = get_reg(c, dst->File, dst->Index); - - reg.dw1.bits.writemask = dst->WriteMask; - - return reg; -} - - - - -static void emit_swz( struct brw_vs_compile *c, - struct brw_reg dst, - struct tgsi_src_register src ) -{ - struct brw_compile *p = &c->func; - unsigned zeros_mask = 0; - unsigned ones_mask = 0; - unsigned src_mask = 0; - ubyte src_swz[4]; - boolean need_tmp = (src.Negate && - dst.file != BRW_GENERAL_REGISTER_FILE); - struct brw_reg tmp = dst; - unsigned i; - - if (need_tmp) - tmp = get_tmp(c); - - for (i = 0; i < 4; i++) { - if (dst.dw1.bits.writemask & (1<regs[PROGRAM_STATE_VAR][0], src.Index); - else -#endif - arg0 = get_reg(c, src.File, src.Index); - - arg0 = brw_swizzle(arg0, - src_swz[0], src_swz[1], - src_swz[2], src_swz[3]); - - brw_MOV(p, brw_writemask(tmp, src_mask), arg0); - } - - if (zeros_mask) - brw_MOV(p, brw_writemask(tmp, zeros_mask), brw_imm_f(0)); - - if (ones_mask) - brw_MOV(p, brw_writemask(tmp, ones_mask), brw_imm_f(1)); - - if (src.Negate) - brw_MOV(p, brw_writemask(tmp, src.Negate), negate(tmp)); - - if (need_tmp) { - brw_MOV(p, dst, tmp); - release_tmp(c, tmp); - } -} - - - -/* Post-vertex-program processing. Send the results to the URB. - */ -static void emit_vertex_write( struct brw_vs_compile *c, struct brw_prog_info *info) -{ - struct brw_compile *p = &c->func; - struct brw_reg m0 = brw_message_reg(0); - struct brw_reg pos = c->regs[TGSI_FILE_OUTPUT][info->pos_idx]; - struct brw_reg ndc; - - if (c->key.copy_edgeflag) { - brw_MOV(p, - get_reg(c, TGSI_FILE_OUTPUT, info->result_edge_idx), - get_reg(c, TGSI_FILE_INPUT, info->edge_flag_idx)); - } - - - /* Build ndc coords? TODO: Shortcircuit when w is known to be one. - */ - if (!c->key.know_w_is_one) { - ndc = get_tmp(c); - emit_math1(c, BRW_MATH_FUNCTION_INV, ndc, brw_swizzle1(pos, 3), BRW_MATH_PRECISION_FULL); - brw_MUL(p, brw_writemask(ndc, TGSI_WRITEMASK_XYZ), pos, ndc); - } - else { - ndc = pos; - } - - /* This includes the workaround for -ve rhw, so is no longer an - * optional step: - */ - if (info->writes_psize || - c->key.nr_userclip || - !c->key.know_w_is_one) - { - struct brw_reg header1 = retype(get_tmp(c), BRW_REGISTER_TYPE_UD); - unsigned i; - - brw_MOV(p, header1, brw_imm_ud(0)); - - brw_set_access_mode(p, BRW_ALIGN_16); - - if (info->writes_psize) { - struct brw_reg psiz = c->regs[TGSI_FILE_OUTPUT][info->psize_idx]; - brw_MUL(p, brw_writemask(header1, TGSI_WRITEMASK_W), - brw_swizzle1(psiz, 0), brw_imm_f(1<<11)); - brw_AND(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, - brw_imm_ud(0x7ff<<8)); - } - - - for (i = 0; i < c->key.nr_userclip; i++) { - brw_set_conditionalmod(p, BRW_CONDITIONAL_L); - brw_DP4(p, brw_null_reg(), pos, c->userplane[i]); - brw_OR(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, brw_imm_ud(1<key.know_w_is_one) { - brw_CMP(p, - vec8(brw_null_reg()), - BRW_CONDITIONAL_L, - brw_swizzle1(ndc, 3), - brw_imm_f(0)); - - brw_OR(p, brw_writemask(header1, TGSI_WRITEMASK_W), header1, brw_imm_ud(1<<6)); - brw_MOV(p, ndc, brw_imm_f(0)); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - } - - brw_set_access_mode(p, BRW_ALIGN_1); /* why? */ - brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), header1); - brw_set_access_mode(p, BRW_ALIGN_16); - - release_tmp(c, header1); - } - else { - brw_MOV(p, retype(brw_message_reg(1), BRW_REGISTER_TYPE_UD), brw_imm_ud(0)); - } - - - /* Emit the (interleaved) headers for the two vertices - an 8-reg - * of zeros followed by two sets of NDC coordinates: - */ - brw_set_access_mode(p, BRW_ALIGN_1); - brw_MOV(p, offset(m0, 2), ndc); - brw_MOV(p, offset(m0, 3), pos); - - - brw_urb_WRITE(p, - brw_null_reg(), /* dest */ - 0, /* starting mrf reg nr */ - c->r0, /* src */ - 0, /* allocate */ - 1, /* used */ - c->nr_outputs + 3, /* msg len */ - 0, /* response len */ - 1, /* eot */ - 1, /* writes complete */ - 0, /* urb destination offset */ - BRW_URB_SWIZZLE_INTERLEAVE); - -} - -static void -post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst ) -{ - struct tgsi_parse_context parse; - const struct tgsi_token *tokens = c->vp->program.tokens; - tgsi_parse_init(&parse, tokens); - while (!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - if (parse.FullToken.Token.Type == TGSI_TOKEN_TYPE_INSTRUCTION) { -#if 0 - struct brw_instruction *brw_inst1, *brw_inst2; - const struct tgsi_full_instruction *inst1, *inst2; - int offset; - inst1 = &parse.FullToken.FullInstruction; - brw_inst1 = inst1->Data; - switch (inst1->Opcode) { - case TGSI_OPCODE_CAL: - case TGSI_OPCODE_BRA: - target_insn = inst1->BranchTarget; - inst2 = &c->vp->program.Base.Instructions[target_insn]; - brw_inst2 = inst2->Data; - offset = brw_inst2 - brw_inst1; - brw_set_src1(brw_inst1, brw_imm_d(offset*16)); - break; - case TGSI_OPCODE_END: - offset = end_inst - brw_inst1; - brw_set_src1(brw_inst1, brw_imm_d(offset*16)); - break; - default: - break; - } -#endif - } - } - tgsi_parse_free(&parse); -} - -static void process_declaration(const struct tgsi_full_declaration *decl, - struct brw_prog_info *info) -{ - int first = decl->u.DeclarationRange.First; - int last = decl->u.DeclarationRange.Last; - - assert (decl->Declaration.Declare != TGSI_DECLARE_MASK); - - switch(decl->Declaration.File) { - case TGSI_FILE_CONSTANT: - info->num_consts += last - first + 1; - break; - case TGSI_FILE_INPUT: { - } - break; - case TGSI_FILE_OUTPUT: { - assert(last == first); /* for now */ - if (decl->Declaration.Semantic) { - switch (decl->Semantic.SemanticName) { - case TGSI_SEMANTIC_POSITION: { - info->pos_idx = first; - } - break; - case TGSI_SEMANTIC_COLOR: - break; - case TGSI_SEMANTIC_BCOLOR: - break; - case TGSI_SEMANTIC_FOG: - break; - case TGSI_SEMANTIC_PSIZE: { - info->writes_psize = TRUE; - info->psize_idx = first; - } - break; - case TGSI_SEMANTIC_GENERIC: - break; - } - } - } - break; - case TGSI_FILE_TEMPORARY: { - info->num_temps += (last - first) + 1; - } - break; - case TGSI_FILE_SAMPLER: { - } - break; - case TGSI_FILE_ADDRESS: { - info->num_addrs += (last - first) + 1; - } - break; - case TGSI_FILE_IMMEDIATE: { - } - break; - case TGSI_FILE_NULL: { - } - break; - } -} - -static void process_instruction(struct brw_vs_compile *c, - struct tgsi_full_instruction *inst, - struct brw_prog_info *info) -{ - struct brw_reg args[3], dst; - struct brw_compile *p = &c->func; - struct brw_indirect stack_index = brw_indirect(0, 0); - unsigned i; - unsigned index; - unsigned file; - /*FIXME: might not be the only one*/ - const struct tgsi_dst_register *dst_reg = &inst->FullDstRegisters[0].DstRegister; - /* - struct brw_instruction *if_inst[MAX_IFSN]; - unsigned insn, if_insn = 0; - */ - - for (i = 0; i < 3; i++) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - index = src->SrcRegister.Index; - file = src->SrcRegister.File; - if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src) - args[i] = c->output_regs[index].reg; - else - args[i] = get_arg(c, &src->SrcRegister); - } - - /* Get dest regs. Note that it is possible for a reg to be both - * dst and arg, given the static allocation of registers. So - * care needs to be taken emitting multi-operation instructions. - */ - index = dst_reg->Index; - file = dst_reg->File; - if (file == TGSI_FILE_OUTPUT && c->output_regs[index].used_in_src) - dst = c->output_regs[index].reg; - else - dst = get_dst(c, dst_reg); - - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ABS: - brw_MOV(p, dst, brw_abs(args[0])); - break; - case TGSI_OPCODE_ADD: - brw_ADD(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_DP3: - brw_DP3(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_DP4: - brw_DP4(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_DPH: - brw_DPH(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_DST: - unalias2(c, dst, args[0], args[1], emit_dst_noalias); - break; - case TGSI_OPCODE_EXP: - unalias1(c, dst, args[0], emit_exp_noalias); - break; - case TGSI_OPCODE_EX2: - emit_math1(c, BRW_MATH_FUNCTION_EXP, dst, args[0], BRW_MATH_PRECISION_FULL); - break; - case TGSI_OPCODE_ARL: - emit_arl(c, dst, args[0]); - break; - case TGSI_OPCODE_FLR: - brw_RNDD(p, dst, args[0]); - break; - case TGSI_OPCODE_FRC: - brw_FRC(p, dst, args[0]); - break; - case TGSI_OPCODE_LOG: - unalias1(c, dst, args[0], emit_log_noalias); - break; - case TGSI_OPCODE_LG2: - emit_math1(c, BRW_MATH_FUNCTION_LOG, dst, args[0], BRW_MATH_PRECISION_FULL); - break; - case TGSI_OPCODE_LIT: - unalias1(c, dst, args[0], emit_lit_noalias); - break; - case TGSI_OPCODE_MAD: - brw_MOV(p, brw_acc_reg(), args[2]); - brw_MAC(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_MAX: - emit_max(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_MIN: - emit_min(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_MOV: -#if 0 - case TGSI_OPCODE_SWZ: - /* The args[0] value can't be used here as it won't have - * correctly encoded the full swizzle: - */ - emit_swz(c, dst, inst->SrcReg[0] ); -#endif - brw_MOV(p, dst, args[0]); - break; - case TGSI_OPCODE_MUL: - brw_MUL(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_POW: - emit_math2(c, BRW_MATH_FUNCTION_POW, dst, args[0], args[1], BRW_MATH_PRECISION_FULL); - break; - case TGSI_OPCODE_RCP: - emit_math1(c, BRW_MATH_FUNCTION_INV, dst, args[0], BRW_MATH_PRECISION_FULL); - break; - case TGSI_OPCODE_RSQ: - emit_math1(c, BRW_MATH_FUNCTION_RSQ, dst, args[0], BRW_MATH_PRECISION_FULL); - break; - - case TGSI_OPCODE_SEQ: - emit_seq(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SNE: - emit_sne(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SGE: - emit_sge(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SGT: - emit_sgt(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SLT: - emit_slt(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SLE: - emit_sle(p, dst, args[0], args[1]); - break; - case TGSI_OPCODE_SUB: - brw_ADD(p, dst, args[0], negate(args[1])); - break; - case TGSI_OPCODE_XPD: - emit_xpd(p, dst, args[0], args[1]); - break; -#if 0 - case TGSI_OPCODE_IF: - assert(if_insn < MAX_IFSN); - if_inst[if_insn++] = brw_IF(p, BRW_EXECUTE_8); - break; - case TGSI_OPCODE_ELSE: - if_inst[if_insn-1] = brw_ELSE(p, if_inst[if_insn-1]); - break; - case TGSI_OPCODE_ENDIF: - assert(if_insn > 0); - brw_ENDIF(p, if_inst[--if_insn]); - break; - case TGSI_OPCODE_BRA: - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); - brw_set_predicate_control_flag_value(p, 0xff); - break; - case TGSI_OPCODE_CAL: - brw_set_access_mode(p, BRW_ALIGN_1); - brw_ADD(p, deref_1uw(stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); - brw_set_access_mode(p, BRW_ALIGN_16); - brw_ADD(p, get_addr_reg(stack_index), - get_addr_reg(stack_index), brw_imm_d(4)); - inst->Data = &p->store[p->nr_insn]; - brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); - break; -#endif - case TGSI_OPCODE_RET: -#if 0 - brw_ADD(p, get_addr_reg(stack_index), - get_addr_reg(stack_index), brw_imm_d(-4)); - brw_set_access_mode(p, BRW_ALIGN_1); - brw_MOV(p, brw_ip_reg(), deref_1uw(stack_index, 0)); - brw_set_access_mode(p, BRW_ALIGN_16); -#else - /*brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16));*/ -#endif - break; - case TGSI_OPCODE_END: - brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); - break; - case TGSI_OPCODE_BGNSUB: - case TGSI_OPCODE_ENDSUB: - break; - default: - debug_printf("Unsupport opcode %d in vertex shader\n", inst->Instruction.Opcode); - break; - } - - if (dst_reg->File == TGSI_FILE_OUTPUT - && dst_reg->Index != info->pos_idx - && c->output_regs[dst_reg->Index].used_in_src) - brw_MOV(p, get_dst(c, dst_reg), dst); - - release_tmps(c); -} - -/* Emit the fragment program instructions here. - */ -void brw_vs_emit(struct brw_vs_compile *c) -{ -#define MAX_IFSN 32 - struct brw_compile *p = &c->func; - struct brw_instruction *end_inst; - struct tgsi_parse_context parse; - struct brw_indirect stack_index = brw_indirect(0, 0); - const struct tgsi_token *tokens = c->vp->program.tokens; - struct brw_prog_info prog_info; - unsigned allocated_registers = 0; - memset(&prog_info, 0, sizeof(struct brw_prog_info)); - - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - brw_set_access_mode(p, BRW_ALIGN_16); - - tgsi_parse_init(&parse, tokens); - /* Message registers can't be read, so copy the output into GRF register - if they are used in source registers */ - while (!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - unsigned i; - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_INSTRUCTION: { - const struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction; - for (i = 0; i < 3; ++i) { - const struct tgsi_src_register *src = &inst->FullSrcRegisters[i].SrcRegister; - unsigned index = src->Index; - unsigned file = src->File; - if (file == TGSI_FILE_OUTPUT) - c->output_regs[index].used_in_src = TRUE; - } - } - break; - default: - /* nothing */ - break; - } - } - tgsi_parse_free(&parse); - - tgsi_parse_init(&parse, tokens); - - while (!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: { - struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; - process_declaration(decl, &prog_info); - } - break; - case TGSI_TOKEN_TYPE_IMMEDIATE: { - struct tgsi_full_immediate *imm = &parse.FullToken.FullImmediate; - /*assert(imm->Immediate.Size == 4);*/ - c->prog_data.imm_buf[c->prog_data.num_imm][0] = imm->u.ImmediateFloat32[0].Float; - c->prog_data.imm_buf[c->prog_data.num_imm][1] = imm->u.ImmediateFloat32[1].Float; - c->prog_data.imm_buf[c->prog_data.num_imm][2] = imm->u.ImmediateFloat32[2].Float; - c->prog_data.imm_buf[c->prog_data.num_imm][3] = imm->u.ImmediateFloat32[3].Float; - c->prog_data.num_imm++; - } - break; - case TGSI_TOKEN_TYPE_INSTRUCTION: { - struct tgsi_full_instruction *inst = &parse.FullToken.FullInstruction; - if (!allocated_registers) { - /* first instruction (declerations finished). - * now that we know what vars are being used allocate - * registers for them.*/ - c->prog_data.num_consts = prog_info.num_consts; - c->prog_data.max_const = prog_info.num_consts + c->prog_data.num_imm; - brw_vs_alloc_regs(c, &prog_info); - - brw_set_access_mode(p, BRW_ALIGN_1); - brw_MOV(p, get_addr_reg(stack_index), brw_address(c->stack)); - brw_set_access_mode(p, BRW_ALIGN_16); - allocated_registers = 1; - } - process_instruction(c, inst, &prog_info); - } - break; - } - } - - end_inst = &p->store[p->nr_insn]; - emit_vertex_write(c, &prog_info); - post_vs_emit(c, end_inst); - tgsi_parse_free(&parse); - -} diff --git a/src/mesa/pipe/i965simple/brw_vs_state.c b/src/mesa/pipe/i965simple/brw_vs_state.c deleted file mode 100644 index c73469929c..0000000000 --- a/src/mesa/pipe/i965simple/brw_vs_state.c +++ /dev/null @@ -1,102 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_state.h" -#include "brw_defines.h" - -#include "pipe/p_util.h" - -static void upload_vs_unit( struct brw_context *brw ) -{ - struct brw_vs_unit_state vs; - - memset(&vs, 0, sizeof(vs)); - - /* CACHE_NEW_VS_PROG */ - vs.thread0.kernel_start_pointer = brw->vs.prog_gs_offset >> 6; - vs.thread0.grf_reg_count = align(brw->vs.prog_data->total_grf, 16) / 16 - 1; - vs.thread3.urb_entry_read_length = brw->vs.prog_data->urb_read_length; - vs.thread3.const_urb_entry_read_length = brw->vs.prog_data->curb_read_length; - vs.thread3.dispatch_grf_start_reg = 1; - - - /* BRW_NEW_URB_FENCE */ - vs.thread4.nr_urb_entries = brw->urb.nr_vs_entries; - vs.thread4.urb_entry_allocation_size = brw->urb.vsize - 1; - vs.thread4.max_threads = MIN2( - MAX2(0, (brw->urb.nr_vs_entries - 6) / 2 - 1), - 15); - - - - if (BRW_DEBUG & DEBUG_SINGLE_THREAD) - vs.thread4.max_threads = 0; - - /* BRW_NEW_CURBE_OFFSETS, _NEW_TRANSFORM */ - if (0 /*brw->attribs.Clip->ClipPlanesEnabled*/) { - /* Note that we read in the userclip planes as well, hence - * clip_start: - */ - vs.thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2; - } - else { - vs.thread3.const_urb_entry_read_offset = brw->curbe.vs_start * 2; - } - - vs.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; - vs.thread3.urb_entry_read_offset = 0; - - /* No samplers for ARB_vp programs: - */ - vs.vs5.sampler_count = 0; - - if (BRW_DEBUG & DEBUG_STATS) - vs.thread4.stats_enable = 1; - - /* Vertex program always enabled: - */ - vs.vs6.vs_enable = 1; - - brw->vs.state_gs_offset = brw_cache_data( &brw->cache[BRW_VS_UNIT], &vs ); -} - - -const struct brw_tracked_state brw_vs_unit = { - .dirty = { - .brw = (BRW_NEW_CLIP | - BRW_NEW_CURBE_OFFSETS | - BRW_NEW_URB_FENCE), - .cache = CACHE_NEW_VS_PROG - }, - .update = upload_vs_unit -}; diff --git a/src/mesa/pipe/i965simple/brw_winsys.h b/src/mesa/pipe/i965simple/brw_winsys.h deleted file mode 100644 index 3523a58614..0000000000 --- a/src/mesa/pipe/i965simple/brw_winsys.h +++ /dev/null @@ -1,205 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * This is the interface that i965simple requires any window system - * hosting it to implement. This is the only include file in i965simple - * which is public. - * - */ - -#ifndef BRW_WINSYS_H -#define BRW_WINSYS_H - - -#include "pipe/p_defines.h" - - -/* Pipe drivers are (meant to be!) independent of both GL and the - * window system. The window system provides a buffer manager and a - * set of additional hooks for things like command buffer submission, - * etc. - * - * There clearly has to be some agreement between the window system - * driver and the hardware driver about the format of command buffers, - * etc. - */ - -struct pipe_buffer; -struct pipe_fence_handle; -struct pipe_winsys; - -/* The pipe driver currently understands the following chipsets: - */ -#define PCI_CHIP_I965_G 0x29A2 -#define PCI_CHIP_I965_Q 0x2992 -#define PCI_CHIP_I965_G_1 0x2982 -#define PCI_CHIP_I965_GM 0x2A02 -#define PCI_CHIP_I965_GME 0x2A12 - - -/* These are the names of all the state caches managed by the driver. - * - * When data is uploaded to a buffer with buffer_subdata, we use the - * special version of that function below so that information about - * what type of data this is can be passed to the winsys backend. - * That in turn allows the correct flags to be set in the aub file - * dump to allow human-readable file dumps later on. - */ - -enum brw_cache_id { - BRW_CC_VP, - BRW_CC_UNIT, - BRW_WM_PROG, - BRW_SAMPLER_DEFAULT_COLOR, - BRW_SAMPLER, - BRW_WM_UNIT, - BRW_SF_PROG, - BRW_SF_VP, - BRW_SF_UNIT, - BRW_VS_UNIT, - BRW_VS_PROG, - BRW_GS_UNIT, - BRW_GS_PROG, - BRW_CLIP_VP, - BRW_CLIP_UNIT, - BRW_CLIP_PROG, - BRW_SS_SURFACE, - BRW_SS_SURF_BIND, - - BRW_MAX_CACHE -}; - -#define BRW_CONSTANT_BUFFER BRW_MAX_CACHE - -/** - * Additional winsys interface for i965simple. - * - * It is an over-simple batchbuffer mechanism. Will want to improve the - * performance of this, perhaps based on the cmdstream stuff. It - * would be pretty impossible to implement swz on top of this - * interface. - * - * Will also need additions/changes to implement static/dynamic - * indirect state. - */ -struct brw_winsys { - - /** - * Reserve space on batch buffer. - * - * Returns a null pointer if there is insufficient space in the batch buffer - * to hold the requested number of dwords and relocations. - * - * The number of dwords should also include the number of relocations. - */ - unsigned *(*batch_start)(struct brw_winsys *sws, - unsigned dwords, - unsigned relocs); - - void (*batch_dword)(struct brw_winsys *sws, - unsigned dword); - - /** - * Emit a relocation to a buffer. - * - * Used not only when the buffer addresses are not pinned, but also to - * ensure refered buffers will not be destroyed until the current batch - * buffer execution is finished. - * - * The access flags is a combination of I915_BUFFER_ACCESS_WRITE and - * I915_BUFFER_ACCESS_READ macros. - */ - void (*batch_reloc)(struct brw_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta); - - - /* Not used yet, but really want this: - */ - void (*batch_end)( struct brw_winsys *sws ); - - /** - * Flush the batch buffer. - * - * Fence argument must point to NULL or to a previous fence, and the caller - * must call fence_reference when done with the fence. - */ - void (*batch_flush)(struct brw_winsys *sws, - struct pipe_fence_handle **fence); - - - /* A version of buffer_subdata that includes information for the - * simulator: - */ - void (*buffer_subdata_typed)(struct brw_winsys *sws, - struct pipe_buffer *buf, - unsigned long offset, - unsigned long size, - const void *data, - unsigned data_type); - - - /* A cheat so we don't have to think about relocations in a couple - * of places yet: - */ - unsigned (*get_buffer_offset)( struct brw_winsys *sws, - struct pipe_buffer *buf, - unsigned flags ); - -}; - -#define BRW_BUFFER_ACCESS_WRITE 0x1 -#define BRW_BUFFER_ACCESS_READ 0x2 - -#define BRW_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) - - -struct pipe_context *brw_create(struct pipe_winsys *, - struct brw_winsys *, - unsigned pci_id); - -static inline boolean brw_batchbuffer_data(struct brw_winsys *winsys, - const void *data, - unsigned bytes) -{ - static const unsigned incr = sizeof(unsigned); - uint i; - const unsigned *udata = (const unsigned*)(data); - unsigned size = bytes/incr; - - winsys->batch_start(winsys, size, 0); - for (i = 0; i < size; ++i) { - winsys->batch_dword(winsys, udata[i]); - } - winsys->batch_end(winsys); - - return (i == size); -} -#endif diff --git a/src/mesa/pipe/i965simple/brw_wm.c b/src/mesa/pipe/i965simple/brw_wm.c deleted file mode 100644 index 539b170744..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm.c +++ /dev/null @@ -1,210 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_util.h" -#include "brw_wm.h" -#include "brw_eu.h" -#include "brw_state.h" -#include "pipe/p_util.h" - - - -static void do_wm_prog( struct brw_context *brw, - struct brw_fragment_program *fp, - struct brw_wm_prog_key *key) -{ - struct brw_wm_compile *c = CALLOC_STRUCT(brw_wm_compile); - const unsigned *program; - unsigned program_size; - - c->key = *key; - c->fp = fp; - - c->delta_xy[0] = brw_null_reg(); - c->delta_xy[1] = brw_null_reg(); - c->pixel_xy[0] = brw_null_reg(); - c->pixel_xy[1] = brw_null_reg(); - c->pixel_w = brw_null_reg(); - - - debug_printf("XXXXXXXX FP\n"); - - brw_wm_glsl_emit(c); - - /* get the program - */ - program = brw_get_program(&c->func, &program_size); - - /* - */ - brw->wm.prog_gs_offset = brw_upload_cache( &brw->cache[BRW_WM_PROG], - &c->key, - sizeof(c->key), - program, - program_size, - &c->prog_data, - &brw->wm.prog_data ); - - FREE(c); -} - - - -static void brw_wm_populate_key( struct brw_context *brw, - struct brw_wm_prog_key *key ) -{ - /* BRW_NEW_FRAGMENT_PROGRAM */ - struct brw_fragment_program *fp = - (struct brw_fragment_program *)brw->attribs.FragmentProgram; - unsigned lookup = 0; - unsigned line_aa; - - memset(key, 0, sizeof(*key)); - - /* Build the index for table lookup - */ - /* BRW_NEW_DEPTH_STENCIL */ - if (fp->UsesKill || - brw->attribs.DepthStencil->alpha.enabled) - lookup |= IZ_PS_KILL_ALPHATEST_BIT; - - if (fp->ComputesDepth) - lookup |= IZ_PS_COMPUTES_DEPTH_BIT; - - if (brw->attribs.DepthStencil->depth.enabled) - lookup |= IZ_DEPTH_TEST_ENABLE_BIT; - - if (brw->attribs.DepthStencil->depth.enabled && - brw->attribs.DepthStencil->depth.writemask) /* ?? */ - lookup |= IZ_DEPTH_WRITE_ENABLE_BIT; - - if (brw->attribs.DepthStencil->stencil[0].enabled) { - lookup |= IZ_STENCIL_TEST_ENABLE_BIT; - - if (brw->attribs.DepthStencil->stencil[0].write_mask || - brw->attribs.DepthStencil->stencil[1].write_mask) - lookup |= IZ_STENCIL_WRITE_ENABLE_BIT; - } - - /* XXX: when should this be disabled? - */ - if (1) - lookup |= IZ_EARLY_DEPTH_TEST_BIT; - - - line_aa = AA_NEVER; - - /* _NEW_LINE, _NEW_POLYGON, BRW_NEW_REDUCED_PRIMITIVE */ - if (brw->attribs.Raster->line_smooth) { - if (brw->reduced_primitive == PIPE_PRIM_LINES) { - line_aa = AA_ALWAYS; - } - else if (brw->reduced_primitive == PIPE_PRIM_TRIANGLES) { - if (brw->attribs.Raster->fill_ccw == PIPE_POLYGON_MODE_LINE) { - line_aa = AA_SOMETIMES; - - if (brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_LINE || - (brw->attribs.Raster->cull_mode == PIPE_WINDING_CW)) - line_aa = AA_ALWAYS; - } - else if (brw->attribs.Raster->fill_cw == PIPE_POLYGON_MODE_LINE) { - line_aa = AA_SOMETIMES; - - if (brw->attribs.Raster->cull_mode == PIPE_WINDING_CCW) - line_aa = AA_ALWAYS; - } - } - } - - brw_wm_lookup_iz(line_aa, - lookup, - key); - - -#if 0 - /* BRW_NEW_SAMPLER - * - * Not doing any of this at the moment: - */ - for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { - const struct pipe_sampler_state *unit = brw->attribs.Samplers[i]; - - if (unit) { - - if (unit->compare && - unit->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - key->shadowtex_mask |= 1<Image[0][t->BaseLevel]->InternalFormat == GL_YCBCR_MESA) - key->yuvtex_mask |= 1<program_string_id = fp->id; - -} - - -static void brw_upload_wm_prog( struct brw_context *brw ) -{ - struct brw_wm_prog_key key; - struct brw_fragment_program *fp = (struct brw_fragment_program *) - brw->attribs.FragmentProgram; - - brw_wm_populate_key(brw, &key); - - /* Make an early check for the key. - */ - if (brw_search_cache(&brw->cache[BRW_WM_PROG], - &key, sizeof(key), - &brw->wm.prog_data, - &brw->wm.prog_gs_offset)) - return; - - do_wm_prog(brw, fp, &key); -} - - -const struct brw_tracked_state brw_wm_prog = { - .dirty = { - .brw = (BRW_NEW_FS | - BRW_NEW_REDUCED_PRIMITIVE), - .cache = 0 - }, - .update = brw_upload_wm_prog -}; - diff --git a/src/mesa/pipe/i965simple/brw_wm.h b/src/mesa/pipe/i965simple/brw_wm.h deleted file mode 100644 index a1ac0f504a..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm.h +++ /dev/null @@ -1,142 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRW_WM_H -#define BRW_WM_H - - -#include "brw_context.h" -#include "brw_eu.h" - -/* A big lookup table is used to figure out which and how many - * additional regs will inserted before the main payload in the WM - * program execution. These mainly relate to depth and stencil - * processing and the early-depth-test optimization. - */ -#define IZ_PS_KILL_ALPHATEST_BIT 0x1 -#define IZ_PS_COMPUTES_DEPTH_BIT 0x2 -#define IZ_DEPTH_WRITE_ENABLE_BIT 0x4 -#define IZ_DEPTH_TEST_ENABLE_BIT 0x8 -#define IZ_STENCIL_WRITE_ENABLE_BIT 0x10 -#define IZ_STENCIL_TEST_ENABLE_BIT 0x20 -#define IZ_EARLY_DEPTH_TEST_BIT 0x40 -#define IZ_BIT_MAX 0x80 - -#define AA_NEVER 0 -#define AA_SOMETIMES 1 -#define AA_ALWAYS 2 - -struct brw_wm_prog_key { - unsigned source_depth_reg:3; - unsigned aa_dest_stencil_reg:3; - unsigned dest_depth_reg:3; - unsigned nr_depth_regs:3; - unsigned shadowtex_mask:8; - unsigned computes_depth:1; /* could be derived from program string */ - unsigned source_depth_to_render_target:1; - unsigned runtime_check_aads_emit:1; - - unsigned yuvtex_mask:8; - - unsigned program_string_id; -}; - - - - - -#define PROGRAM_INTERNAL_PARAM -#define MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS 1024 /* 72 for GL_ARB_f_p */ -#define BRW_WM_MAX_INSN (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + PIPE_ATTRIB_MAX + 3) -#define BRW_WM_MAX_GRF 128 /* hardware limit */ -#define BRW_WM_MAX_VREG (BRW_WM_MAX_INSN * 4) -#define BRW_WM_MAX_REF (BRW_WM_MAX_INSN * 12) -#define BRW_WM_MAX_PARAM 256 -#define BRW_WM_MAX_CONST 256 -#define BRW_WM_MAX_KILLS MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS - -#define PAYLOAD_DEPTH (PIPE_ATTRIB_MAX) - -#define MAX_IFSN 32 -#define MAX_LOOP_DEPTH 32 - -struct brw_wm_compile { - struct brw_compile func; - struct brw_wm_prog_key key; - struct brw_wm_prog_data prog_data; /* result */ - - struct brw_fragment_program *fp; - - unsigned grf_limit; - unsigned max_wm_grf; - - - struct brw_reg pixel_xy[2]; - struct brw_reg delta_xy[2]; - struct brw_reg pixel_w; - - - struct brw_reg wm_regs[8][32][4]; - - struct brw_reg payload_depth[4]; - struct brw_reg payload_coef[16]; - - struct brw_reg emit_mask_reg; - - struct brw_instruction *if_inst[MAX_IFSN]; - int if_insn; - - struct brw_instruction *loop_inst[MAX_LOOP_DEPTH]; - int loop_insn; - - struct brw_instruction *inst0; - struct brw_instruction *inst1; - - struct brw_reg stack; - struct brw_indirect stack_index; - - unsigned reg_index; - - unsigned tmp_start; - unsigned tmp_index; -}; - - - -void brw_wm_lookup_iz( unsigned line_aa, - unsigned lookup, - struct brw_wm_prog_key *key ); - -void brw_wm_glsl_emit(struct brw_wm_compile *c); -void brw_wm_emit_decls(struct brw_wm_compile *c); - -#endif diff --git a/src/mesa/pipe/i965simple/brw_wm_decl.c b/src/mesa/pipe/i965simple/brw_wm_decl.c deleted file mode 100644 index b45a333a2e..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_decl.c +++ /dev/null @@ -1,383 +0,0 @@ - -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_wm.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" - -static struct brw_reg alloc_tmp(struct brw_wm_compile *c) -{ - c->tmp_index++; - c->reg_index = MAX2(c->reg_index, c->tmp_start + c->tmp_index); - return brw_vec8_grf(c->tmp_start + c->tmp_index, 0); -} - -static void release_tmps(struct brw_wm_compile *c) -{ - c->tmp_index = 0; -} - - - -static int is_null( struct brw_reg reg ) -{ - return (reg.file == BRW_ARCHITECTURE_REGISTER_FILE && - reg.nr == BRW_ARF_NULL); -} - -static void emit_pixel_xy( struct brw_wm_compile *c ) -{ - if (is_null(c->pixel_xy[0])) { - - struct brw_compile *p = &c->func; - struct brw_reg r1_uw = retype(brw_vec1_grf(1, 0), BRW_REGISTER_TYPE_UW); - - c->pixel_xy[0] = vec8(retype(alloc_tmp(c), BRW_REGISTER_TYPE_UW)); - c->pixel_xy[1] = vec8(retype(alloc_tmp(c), BRW_REGISTER_TYPE_UW)); - - /* Calculate pixel centers by adding 1 or 0 to each of the - * micro-tile coordinates passed in r1. - */ - brw_ADD(p, - c->pixel_xy[0], - stride(suboffset(r1_uw, 4), 2, 4, 0), - brw_imm_v(0x10101010)); - - brw_ADD(p, - c->pixel_xy[1], - stride(suboffset(r1_uw, 5), 2, 4, 0), - brw_imm_v(0x11001100)); - } -} - - - - - - -static void emit_delta_xy( struct brw_wm_compile *c ) -{ - if (is_null(c->delta_xy[0])) { - struct brw_compile *p = &c->func; - struct brw_reg r1 = brw_vec1_grf(1, 0); - - emit_pixel_xy(c); - - c->delta_xy[0] = alloc_tmp(c); - c->delta_xy[1] = alloc_tmp(c); - - /* Calc delta X,Y by subtracting origin in r1 from the pixel - * centers. - */ - brw_ADD(p, - c->delta_xy[0], - retype(c->pixel_xy[0], BRW_REGISTER_TYPE_UW), - negate(r1)); - - brw_ADD(p, - c->delta_xy[1], - retype(c->pixel_xy[1], BRW_REGISTER_TYPE_UW), - negate(suboffset(r1,1))); - } -} - - - -#if 0 -static void emit_pixel_w( struct brw_wm_compile *c ) -{ - if (is_null(c->pixel_w)) { - struct brw_compile *p = &c->func; - - struct brw_reg interp_wpos = c->coef_wpos; - - c->pixel_w = alloc_tmp(c); - - emit_delta_xy(c); - - /* Calc 1/w - just linterp wpos[3] optimized by putting the - * result straight into a message reg. - */ - struct brw_reg interp3 = brw_vec1_grf(interp_wpos.nr+1, 4); - brw_LINE(p, brw_null_reg(), interp3, c->delta_xy[0]); - brw_MAC(p, brw_message_reg(2), suboffset(interp3, 1), c->delta_xy[1]); - - /* Calc w */ - brw_math_16( p, - c->pixel_w, - BRW_MATH_FUNCTION_INV, - BRW_MATH_SATURATE_NONE, - 2, - brw_null_reg(), - BRW_MATH_PRECISION_FULL); - } -} -#endif - - -static void emit_cinterp(struct brw_wm_compile *c, - int idx, - int mask ) -{ - struct brw_compile *p = &c->func; - struct brw_reg interp[4]; - struct brw_reg coef = c->payload_coef[idx]; - int i; - - interp[0] = brw_vec1_grf(coef.nr, 0); - interp[1] = brw_vec1_grf(coef.nr, 4); - interp[2] = brw_vec1_grf(coef.nr+1, 0); - interp[3] = brw_vec1_grf(coef.nr+1, 4); - - for(i = 0; i < 4; i++ ) { - if (mask & (1<wm_regs[TGSI_FILE_INPUT][idx][i]; - brw_MOV(p, dst, suboffset(interp[i],3)); - } - } -} - -static void emit_linterp(struct brw_wm_compile *c, - int idx, - int mask ) -{ - struct brw_compile *p = &c->func; - struct brw_reg interp[4]; - struct brw_reg coef = c->payload_coef[idx]; - int i; - - emit_delta_xy(c); - - interp[0] = brw_vec1_grf(coef.nr, 0); - interp[1] = brw_vec1_grf(coef.nr, 4); - interp[2] = brw_vec1_grf(coef.nr+1, 0); - interp[3] = brw_vec1_grf(coef.nr+1, 4); - - for(i = 0; i < 4; i++ ) { - if (mask & (1<wm_regs[TGSI_FILE_INPUT][idx][i]; - brw_LINE(p, brw_null_reg(), interp[i], c->delta_xy[0]); - brw_MAC(p, dst, suboffset(interp[i],1), c->delta_xy[1]); - } - } -} - -#if 0 -static void emit_pinterp(struct brw_wm_compile *c, - int idx, - int mask ) -{ - struct brw_compile *p = &c->func; - struct brw_reg interp[4]; - struct brw_reg coef = c->payload_coef[idx]; - int i; - - get_delta_xy(c); - get_pixel_w(c); - - interp[0] = brw_vec1_grf(coef.nr, 0); - interp[1] = brw_vec1_grf(coef.nr, 4); - interp[2] = brw_vec1_grf(coef.nr+1, 0); - interp[3] = brw_vec1_grf(coef.nr+1, 4); - - for(i = 0; i < 4; i++ ) { - if (mask & (1<delta_xy[0]); - brw_MAC(p, dst, suboffset(interp[i],1), c->delta_xy[1]); - brw_MUL(p, dst, dst, c->pixel_w); - } - } -} -#endif - - - -#if 0 -static void emit_wpos( ) -{ - struct prog_dst_register dst = dst_reg(PROGRAM_INPUT, idx); - struct tgsi_full_src_register interp = src_reg(PROGRAM_PAYLOAD, idx); - struct tgsi_full_src_register deltas = get_delta_xy(c); - struct tgsi_full_src_register arg2; - unsigned opcode; - - opcode = WM_LINTERP; - arg2 = src_undef(); - - /* Have to treat wpos.xy specially: - */ - emit_op(c, - WM_WPOSXY, - dst_mask(dst, WRITEMASK_XY), - 0, 0, 0, - get_pixel_xy(c), - src_undef(), - src_undef()); - - dst = dst_mask(dst, WRITEMASK_ZW); - - /* PROGRAM_INPUT.attr.xyzw = INTERP payload.interp[attr].x, deltas.xyw - */ - emit_op(c, - WM_LINTERP, - dst, - 0, 0, 0, - interp, - deltas, - arg2); -} -#endif - - - - -/* Perform register allocation: - * - * -- r0??? - * -- passthrough depth regs (and stencil/aa??) - * -- curbe ?? - * -- inputs (coefficients) - * - * Use a totally static register allocation. This will perform poorly - * but is an easy way to get started (again). - */ -static void prealloc_reg(struct brw_wm_compile *c) -{ - int i, j; - int nr_curbe_regs = 0; - - /* R0, then some depth related regs: - */ - for (i = 0; i < c->key.nr_depth_regs; i++) { - c->payload_depth[i] = brw_vec8_grf(i*2, 0); - c->reg_index += 2; - } - - - /* Then a copy of our part of the CURBE entry: - */ - { - int nr_constants = c->fp->info.nr_regs[TGSI_FILE_CONSTANT]; - int index = 0; - - c->prog_data.max_const = 4*nr_constants; - for (i = 0; i < nr_constants; i++) { - for (j = 0; j < 4; j++, index++) - c->wm_regs[TGSI_FILE_CONSTANT][i][j] = brw_vec1_grf(c->reg_index + index/8, - index%8); - } - - nr_curbe_regs = 2*((4*nr_constants+15)/16); - c->reg_index += nr_curbe_regs; - } - - /* Adjust for parameter coefficients for position, which are - * currently always provided. - */ -// c->position_coef[i] = brw_vec8_grf(c->reg_index, 0); - c->reg_index += 2; - - /* Next we receive the plane coefficients for parameter - * interpolation: - */ - for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_INPUT]; i++) { - c->payload_coef[i] = brw_vec8_grf(c->reg_index, 0); - c->reg_index += 2; - } - - c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; - c->prog_data.urb_read_length = (c->fp->program.num_inputs + 1) * 2; - c->prog_data.curb_read_length = nr_curbe_regs; - - /* That's the end of the payload, now we can start allocating registers. - */ - c->emit_mask_reg = brw_uw1_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); - c->reg_index++; - - c->stack = brw_uw16_reg(BRW_GENERAL_REGISTER_FILE, c->reg_index, 0); - c->reg_index += 2; - - /* Now allocate room for the interpolated inputs and staging - * registers for the outputs: - */ - for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_INPUT]; i++) - for (j = 0; j < 4; j++) - c->wm_regs[TGSI_FILE_INPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); - - for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_OUTPUT]; i++) - for (j = 0; j < 4; j++) - c->wm_regs[TGSI_FILE_OUTPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); - - /* Beyond this we should only need registers for internal temporaries: - */ - c->tmp_start = c->reg_index; -} - - - - - -/* Need to interpolate fragment program inputs in as a preamble to the - * shader. A more sophisticated compiler would do this on demand, but - * we'll do it up front: - */ -void brw_wm_emit_decls(struct brw_wm_compile *c) -{ - struct tgsi_parse_context parse; - int done = 0; - - prealloc_reg(c); - - tgsi_parse_init( &parse, c->fp->program.tokens ); - - while( !done && - !tgsi_parse_end_of_tokens( &parse ) ) - { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - { - const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; - unsigned first = decl->u.DeclarationRange.First; - unsigned last = decl->u.DeclarationRange.Last; - unsigned mask = decl->Declaration.UsageMask; /* ? */ - unsigned i; - - if (decl->Declaration.File != TGSI_FILE_INPUT) - break; - - assert(decl->Declaration.Interpolate); - - for( i = first; i <= last; i++ ) { - switch (decl->Interpolation.Interpolate) { - case TGSI_INTERPOLATE_CONSTANT: - emit_cinterp(c, i, mask); - break; - - case TGSI_INTERPOLATE_LINEAR: - emit_linterp(c, i, mask); - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - //emit_pinterp(c, i, mask); - emit_linterp(c, i, mask); - break; - } - } - break; - } - case TGSI_TOKEN_TYPE_IMMEDIATE: - case TGSI_TOKEN_TYPE_INSTRUCTION: - default: - done = 1; - break; - } - } - - tgsi_parse_free (&parse); - - release_tmps(c); -} diff --git a/src/mesa/pipe/i965simple/brw_wm_glsl.c b/src/mesa/pipe/i965simple/brw_wm_glsl.c deleted file mode 100644 index d95645d108..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_glsl.c +++ /dev/null @@ -1,1079 +0,0 @@ - -#include "brw_context.h" -#include "brw_eu.h" -#include "brw_wm.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" - - - -static int get_scalar_dst_index(struct tgsi_full_instruction *inst) -{ - struct tgsi_dst_register dst = inst->FullDstRegisters[0].DstRegister; - int i; - for (i = 0; i < 4; i++) - if (dst.WriteMask & (1<tmp_index++; - c->reg_index = MAX2(c->reg_index, c->tmp_index); - return brw_vec8_grf(c->tmp_start + c->tmp_index, 0); -} - -static void release_tmps(struct brw_wm_compile *c) -{ - c->tmp_index = 0; -} - - -static struct brw_reg -get_reg(struct brw_wm_compile *c, int file, int index, int component ) -{ - switch (file) { - case TGSI_FILE_NULL: - return brw_null_reg(); - - case TGSI_FILE_SAMPLER: - /* Should never get here: - */ - assert (0); - return brw_null_reg(); - - case TGSI_FILE_IMMEDIATE: - /* These need a different path: - */ - assert(0); - return brw_null_reg(); - - - case TGSI_FILE_CONSTANT: - case TGSI_FILE_INPUT: - case TGSI_FILE_OUTPUT: - case TGSI_FILE_TEMPORARY: - case TGSI_FILE_ADDRESS: - return c->wm_regs[file][index][component]; - - default: - assert(0); - return brw_null_reg(); - } -} - - -static struct brw_reg get_dst_reg(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst, - int component) -{ - return get_reg(c, - inst->FullDstRegisters[0].DstRegister.File, - inst->FullDstRegisters[0].DstRegister.Index, - component); -} - -static int get_swz( struct tgsi_src_register src, int index ) -{ - switch (index & 3) { - case 0: return src.SwizzleX; - case 1: return src.SwizzleY; - case 2: return src.SwizzleZ; - case 3: return src.SwizzleW; - default: return 0; - } -} - -static int get_ext_swz( struct tgsi_src_register_ext_swz src, int index ) -{ - switch (index & 3) { - case 0: return src.ExtSwizzleX; - case 1: return src.ExtSwizzleY; - case 2: return src.ExtSwizzleZ; - case 3: return src.ExtSwizzleW; - default: return 0; - } -} - -static struct brw_reg get_src_reg(struct brw_wm_compile *c, - struct tgsi_full_src_register *src, - int index) -{ - struct brw_reg reg; - int component = index; - int neg = 0; - int abs = 0; - - if (src->SrcRegister.Negate) - neg = 1; - - component = get_swz(src->SrcRegister, component); - - /* Yes, there are multiple negates: - */ - switch (component & 3) { - case 0: neg ^= src->SrcRegisterExtSwz.NegateX; break; - case 1: neg ^= src->SrcRegisterExtSwz.NegateY; break; - case 2: neg ^= src->SrcRegisterExtSwz.NegateZ; break; - case 3: neg ^= src->SrcRegisterExtSwz.NegateW; break; - } - - /* And multiple swizzles, fun isn't it: - */ - component = get_ext_swz(src->SrcRegisterExtSwz, component); - - /* Can't handle this, don't know if we need to: - */ - assert(src->SrcRegisterExtSwz.ExtDivide == TGSI_EXTSWIZZLE_ONE); - - /* Not handling indirect lookups yet: - */ - assert(src->SrcRegister.Indirect == 0); - - /* Don't know what dimension means: - */ - assert(src->SrcRegister.Dimension == 0); - - /* Will never handle any of this stuff: - */ - assert(src->SrcRegisterExtMod.Complement == 0); - assert(src->SrcRegisterExtMod.Bias == 0); - assert(src->SrcRegisterExtMod.Scale2X == 0); - - if (src->SrcRegisterExtMod.Absolute) - abs = 1; - - /* Another negate! This is a post-absolute negate, which we - * can't do. Need to clean the crap out of tgsi somehow. - */ - assert(src->SrcRegisterExtMod.Negate == 0); - - switch( component ) { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - reg = get_reg(c, - src->SrcRegister.File, - src->SrcRegister.Index, - component ); - - if (neg) - reg = negate(reg); - - if (abs) - reg = brw_abs(reg); - - break; - - /* XXX: this won't really work in the general case, but we know - * that the extended swizzle is only allowed in the SWZ - * instruction (right??), in which case using an immediate - * directly will work. - */ - case TGSI_EXTSWIZZLE_ZERO: - reg = brw_imm_f(0); - break; - - case TGSI_EXTSWIZZLE_ONE: - if (neg && !abs) - reg = brw_imm_f(-1.0); - else - reg = brw_imm_f(1.0); - break; - - default: - assert(0); - break; - } - - - return reg; -} - -static void emit_abs( struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - - int i; - struct brw_compile *p = &c->func; - brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); - for (i = 0; i < 4; i++) { - if (mask & (1<FullSrcRegisters[0], i); - brw_MOV(p, dst, brw_abs(src)); /* NOTE */ - } - } - brw_set_saturate(p, 0); -} - - -static void emit_xpd(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - int i; - struct brw_compile *p = &c->func; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - for (i = 0; i < 4; i++) { - unsigned i2 = (i+2)%3; - unsigned i1 = (i+1)%3; - if (mask & (1<FullSrcRegisters[0], i2)); - src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i1); - brw_MUL(p, brw_null_reg(), src0, src1); - src0 = get_src_reg(c, &inst->FullSrcRegisters[0], i1); - src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i2); - brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); - brw_MAC(p, dst, src0, src1); - brw_set_saturate(p, 0); - } - } - brw_set_saturate(p, 0); -} - -static void emit_dp3(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_reg src0[3], src1[3], dst; - int i; - struct brw_compile *p = &c->func; - for (i = 0; i < 3; i++) { - src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); - src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i); - } - - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); - brw_MUL(p, brw_null_reg(), src0[0], src1[0]); - brw_MAC(p, brw_null_reg(), src0[1], src1[1]); - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - brw_MAC(p, dst, src0[2], src1[2]); - brw_set_saturate(p, 0); -} - -static void emit_dp4(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_reg src0[4], src1[4], dst; - int i; - struct brw_compile *p = &c->func; - for (i = 0; i < 4; i++) { - src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); - src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i); - } - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); - brw_MUL(p, brw_null_reg(), src0[0], src1[0]); - brw_MAC(p, brw_null_reg(), src0[1], src1[1]); - brw_MAC(p, brw_null_reg(), src0[2], src1[2]); - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - brw_MAC(p, dst, src0[3], src1[3]); - brw_set_saturate(p, 0); -} - -static void emit_dph(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_reg src0[4], src1[4], dst; - int i; - struct brw_compile *p = &c->func; - for (i = 0; i < 4; i++) { - src0[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); - src1[i] = get_src_reg(c, &inst->FullSrcRegisters[1], i); - } - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); - brw_MUL(p, brw_null_reg(), src0[0], src1[0]); - brw_MAC(p, brw_null_reg(), src0[1], src1[1]); - brw_MAC(p, dst, src0[2], src1[2]); - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - brw_ADD(p, dst, src0[3], src1[3]); - brw_set_saturate(p, 0); -} - -static void emit_math1(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst, unsigned func) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, dst; - - src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); - brw_MOV(p, brw_message_reg(2), src0); - brw_math(p, - dst, - func, - ((inst->Instruction.Saturate != TGSI_SAT_NONE) - ? BRW_MATH_SATURATE_SATURATE - : BRW_MATH_SATURATE_NONE), - 2, - brw_null_reg(), - BRW_MATH_DATA_VECTOR, - BRW_MATH_PRECISION_FULL); -} - - -static void emit_alu2(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst, - unsigned opcode) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, src1, dst; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - int i; - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<FullSrcRegisters[0], i); - src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); - brw_alu2(p, opcode, dst, src0, src1); - } - } - brw_set_saturate(p, 0); -} - - -static void emit_alu1(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst, - unsigned opcode) -{ - struct brw_compile *p = &c->func; - struct brw_reg src0, dst; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - int i; - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - for (i = 0 ; i < 4; i++) { - if (mask & (1<FullSrcRegisters[0], i); - brw_alu1(p, opcode, dst, src0); - } - } - if (inst->Instruction.Saturate != TGSI_SAT_NONE) - brw_set_saturate(p, 0); -} - - -static void emit_max(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - struct brw_reg src0, src1, dst; - int i; - brw_push_insn_state(p); - for (i = 0; i < 4; i++) { - if (mask & (1<FullSrcRegisters[0], i); - src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - brw_MOV(p, dst, src0); - brw_set_saturate(p, 0); - - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src0, src1); - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_MOV(p, dst, src1); - brw_set_saturate(p, 0); - brw_set_predicate_control_flag_value(p, 0xff); - } - } - brw_pop_insn_state(p); -} - -static void emit_min(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - struct brw_reg src0, src1, dst; - int i; - brw_push_insn_state(p); - for (i = 0; i < 4; i++) { - if (mask & (1<FullSrcRegisters[0], i); - src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - brw_MOV(p, dst, src0); - brw_set_saturate(p, 0); - - brw_CMP(p, brw_null_reg(), BRW_CONDITIONAL_L, src1, src0); - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_MOV(p, dst, src1); - brw_set_saturate(p, 0); - brw_set_predicate_control_flag_value(p, 0xff); - } - } - brw_pop_insn_state(p); -} - -static void emit_pow(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_compile *p = &c->func; - struct brw_reg dst, src0, src1; - dst = get_dst_reg(c, inst, get_scalar_dst_index(inst)); - src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); - src1 = get_src_reg(c, &inst->FullSrcRegisters[1], 0); - - brw_MOV(p, brw_message_reg(2), src0); - brw_MOV(p, brw_message_reg(3), src1); - - brw_math(p, - dst, - BRW_MATH_FUNCTION_POW, - (inst->Instruction.Saturate != TGSI_SAT_NONE - ? BRW_MATH_SATURATE_SATURATE - : BRW_MATH_SATURATE_NONE), - 2, - brw_null_reg(), - BRW_MATH_DATA_VECTOR, - BRW_MATH_PRECISION_FULL); -} - -static void emit_lrp(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - struct brw_reg dst, tmp1, tmp2, src0, src1, src2; - int i; - for (i = 0; i < 4; i++) { - if (mask & (1<FullSrcRegisters[0], i); - - src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); - - if (src1.nr == dst.nr) { - tmp1 = alloc_tmp(c); - brw_MOV(p, tmp1, src1); - } else - tmp1 = src1; - - src2 = get_src_reg(c, &inst->FullSrcRegisters[2], i); - if (src2.nr == dst.nr) { - tmp2 = alloc_tmp(c); - brw_MOV(p, tmp2, src2); - } else - tmp2 = src2; - - brw_ADD(p, dst, negate(src0), brw_imm_f(1.0)); - brw_MUL(p, brw_null_reg(), dst, tmp2); - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - brw_MAC(p, dst, src0, tmp1); - brw_set_saturate(p, 0); - } - release_tmps(c); - } -} - -static void emit_kil(struct brw_wm_compile *c) -{ - struct brw_compile *p = &c->func; - struct brw_reg depth = retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_UW); - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); - brw_NOT(p, c->emit_mask_reg, brw_mask_reg(1)); //IMASK - brw_AND(p, depth, c->emit_mask_reg, depth); - brw_pop_insn_state(p); -} - -static void emit_mad(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - struct brw_reg dst, src0, src1, src2; - int i; - - for (i = 0; i < 4; i++) { - if (mask & (1<FullSrcRegisters[0], i); - src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); - src2 = get_src_reg(c, &inst->FullSrcRegisters[2], i); - brw_MUL(p, dst, src0, src1); - - brw_set_saturate(p, (inst->Instruction.Saturate != TGSI_SAT_NONE) ? 1 : 0); - brw_ADD(p, dst, dst, src2); - brw_set_saturate(p, 0); - } - } -} - -static void emit_sop(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst, unsigned cond) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - struct brw_reg dst, src0, src1; - int i; - - brw_push_insn_state(p); - for (i = 0; i < 4; i++) { - if (mask & (1<FullSrcRegisters[0], i); - src1 = get_src_reg(c, &inst->FullSrcRegisters[1], i); - brw_CMP(p, brw_null_reg(), cond, src0, src1); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - brw_MOV(p, dst, brw_imm_f(0.0)); - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - brw_MOV(p, dst, brw_imm_f(1.0)); - } - } - brw_pop_insn_state(p); -} - - -static void emit_ddx(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_compile *p = &c->func; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - struct brw_reg interp[4]; - struct brw_reg dst; - struct brw_reg src0, w; - unsigned nr, i; - src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); - w = get_src_reg(c, &inst->FullSrcRegisters[1], 3); - nr = src0.nr; - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); - for(i = 0; i < 4; i++ ) { - if (mask & (1<func; - unsigned mask = inst->FullDstRegisters[0].DstRegister.WriteMask; - struct brw_reg interp[4]; - struct brw_reg dst; - struct brw_reg src0, w; - unsigned nr, i; - - src0 = get_src_reg(c, &inst->FullSrcRegisters[0], 0); - nr = src0.nr; - w = get_src_reg(c, &inst->FullSrcRegisters[1], 3); - interp[0] = brw_vec1_grf(nr, 0); - interp[1] = brw_vec1_grf(nr, 4); - interp[2] = brw_vec1_grf(nr+1, 0); - interp[3] = brw_vec1_grf(nr+1, 4); - brw_set_saturate(p, inst->Instruction.Saturate != TGSI_SAT_NONE); - for(i = 0; i < 4; i++ ) { - if (mask & (1<func; - struct brw_reg payload_reg = c->payload_depth[0]; - struct brw_reg dst[4], src[4]; - unsigned i; - for (i = 0; i < 4; i++) - dst[i] = get_dst_reg(c, inst, i); - for (i = 0; i < 4; i++) - src[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); - -#if 0 - switch (inst->TexSrcTarget) { - case TEXTURE_1D_INDEX: - brw_MOV(p, brw_message_reg(2), src[0]); - brw_MOV(p, brw_message_reg(3), brw_imm_f(0)); - brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); - break; - case TEXTURE_2D_INDEX: - case TEXTURE_RECT_INDEX: - brw_MOV(p, brw_message_reg(2), src[0]); - brw_MOV(p, brw_message_reg(3), src[1]); - brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); - break; - default: - brw_MOV(p, brw_message_reg(2), src[0]); - brw_MOV(p, brw_message_reg(3), src[1]); - brw_MOV(p, brw_message_reg(4), src[2]); - break; - } -#else - brw_MOV(p, brw_message_reg(2), src[0]); - brw_MOV(p, brw_message_reg(3), src[1]); - brw_MOV(p, brw_message_reg(4), brw_imm_f(0)); -#endif - - brw_MOV(p, brw_message_reg(5), src[3]); - brw_MOV(p, brw_message_reg(6), brw_imm_f(0)); - brw_SAMPLE(p, - retype(vec8(dst[0]), BRW_REGISTER_TYPE_UW), - 1, - retype(payload_reg, BRW_REGISTER_TYPE_UW), - inst->TexSrcUnit + 1, /* surface */ - inst->TexSrcUnit, /* sampler */ - inst->FullDstRegisters[0].DstRegister.WriteMask, - BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS, - 4, - 4, - 0); -#endif -} - -static void emit_tex(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ -#if 0 - struct brw_compile *p = &c->func; - struct brw_reg payload_reg = c->payload_depth[0]; - struct brw_reg dst[4], src[4]; - unsigned msg_len; - unsigned i, nr; - unsigned emit; - boolean shadow = (c->key.shadowtex_mask & (1<TexSrcUnit)) ? 1 : 0; - - for (i = 0; i < 4; i++) - dst[i] = get_dst_reg(c, inst, i); - for (i = 0; i < 4; i++) - src[i] = get_src_reg(c, &inst->FullSrcRegisters[0], i); - -#if 0 - switch (inst->TexSrcTarget) { - case TEXTURE_1D_INDEX: - emit = WRITEMASK_X; - nr = 1; - break; - case TEXTURE_2D_INDEX: - case TEXTURE_RECT_INDEX: - emit = WRITEMASK_XY; - nr = 2; - break; - default: - emit = WRITEMASK_XYZ; - nr = 3; - break; - } -#else - emit = WRITEMASK_XY; - nr = 2; -#endif - - msg_len = 1; - - for (i = 0; i < nr; i++) { - static const unsigned swz[4] = {0,1,2,2}; - if (emit & (1<TexSrcUnit + 1, /* surface */ - inst->TexSrcUnit, /* sampler */ - inst->FullDstRegisters[0].DstRegister.WriteMask, - BRW_SAMPLER_MESSAGE_SIMD8_SAMPLE, - 4, - shadow ? 6 : 4, - 0); - - if (shadow) - brw_MOV(p, dst[3], brw_imm_f(1.0)); -#endif -} - - - - - - - - -static void emit_fb_write(struct brw_wm_compile *c, - struct tgsi_full_instruction *inst) -{ - struct brw_compile *p = &c->func; - int nr = 2; - int channel; - int base_reg = 0; - - // src0 = output color - // src1 = payload_depth[0] - // src2 = output depth - // dst = ??? - - - - /* Reserve a space for AA - may not be needed: - */ - if (c->key.aa_dest_stencil_reg) - nr += 1; - - { - brw_push_insn_state(p); - for (channel = 0; channel < 4; channel++) { - struct brw_reg src0 = c->wm_regs[TGSI_FILE_OUTPUT][0][channel]; - - /* mov (8) m2.0<1>:ud r28.0<8;8,1>:ud { Align1 } */ - /* mov (8) m6.0<1>:ud r29.0<8;8,1>:ud { Align1 SecHalf } */ - brw_MOV(p, brw_message_reg(nr + channel), src0); - } - /* skip over the regs populated above: */ - nr += 8; - brw_pop_insn_state(p); - } - - - /* Pass through control information: - */ - /* mov (8) m1.0<1>:ud r1.0<8;8,1>:ud { Align1 NoMask } */ - { - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); /* ? */ - brw_MOV(p, - brw_message_reg(base_reg + 1), - brw_vec8_grf(1, 0)); - brw_pop_insn_state(p); - } - - /* Send framebuffer write message: */ - brw_fb_WRITE(p, - retype(vec8(brw_null_reg()), BRW_REGISTER_TYPE_UW), - base_reg, - retype(brw_vec8_grf(0, 0), BRW_REGISTER_TYPE_UW), - 0, /* render surface always 0 */ - nr, - 0, - 1); - -} - - -static void brw_wm_emit_instruction( struct brw_wm_compile *c, - struct tgsi_full_instruction *inst ) -{ - struct brw_compile *p = &c->func; - -#if 0 - if (inst->CondUpdate) - brw_set_conditionalmod(p, BRW_CONDITIONAL_NZ); - else - brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); -#else - brw_set_conditionalmod(p, BRW_CONDITIONAL_NONE); -#endif - - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ABS: - emit_abs(c, inst); - break; - case TGSI_OPCODE_ADD: - emit_alu2(c, inst, BRW_OPCODE_ADD); - break; - case TGSI_OPCODE_SUB: - assert(0); -// emit_alu2(c, inst, BRW_OPCODE_SUB); - break; - case TGSI_OPCODE_FRC: - emit_alu1(c, inst, BRW_OPCODE_FRC); - break; - case TGSI_OPCODE_FLR: - assert(0); -// emit_alu1(c, inst, BRW_OPCODE_FLR); - break; - case TGSI_OPCODE_LRP: - emit_lrp(c, inst); - break; - case TGSI_OPCODE_INT: - emit_alu1(c, inst, BRW_OPCODE_RNDD); - break; - case TGSI_OPCODE_MOV: - emit_alu1(c, inst, BRW_OPCODE_MOV); - break; - case TGSI_OPCODE_DP3: - emit_dp3(c, inst); - break; - case TGSI_OPCODE_DP4: - emit_dp4(c, inst); - break; - case TGSI_OPCODE_XPD: - emit_xpd(c, inst); - break; - case TGSI_OPCODE_DPH: - emit_dph(c, inst); - break; - case TGSI_OPCODE_RCP: - emit_math1(c, inst, BRW_MATH_FUNCTION_INV); - break; - case TGSI_OPCODE_RSQ: - emit_math1(c, inst, BRW_MATH_FUNCTION_RSQ); - break; - case TGSI_OPCODE_SIN: - emit_math1(c, inst, BRW_MATH_FUNCTION_SIN); - break; - case TGSI_OPCODE_COS: - emit_math1(c, inst, BRW_MATH_FUNCTION_COS); - break; - case TGSI_OPCODE_EX2: - emit_math1(c, inst, BRW_MATH_FUNCTION_EXP); - break; - case TGSI_OPCODE_LG2: - emit_math1(c, inst, BRW_MATH_FUNCTION_LOG); - break; - case TGSI_OPCODE_MAX: - emit_max(c, inst); - break; - case TGSI_OPCODE_MIN: - emit_min(c, inst); - break; - case TGSI_OPCODE_DDX: - emit_ddx(c, inst); - break; - case TGSI_OPCODE_DDY: - emit_ddy(c, inst); - break; - case TGSI_OPCODE_SLT: - emit_sop(c, inst, BRW_CONDITIONAL_L); - break; - case TGSI_OPCODE_SLE: - emit_sop(c, inst, BRW_CONDITIONAL_LE); - break; - case TGSI_OPCODE_SGT: - emit_sop(c, inst, BRW_CONDITIONAL_G); - break; - case TGSI_OPCODE_SGE: - emit_sop(c, inst, BRW_CONDITIONAL_GE); - break; - case TGSI_OPCODE_SEQ: - emit_sop(c, inst, BRW_CONDITIONAL_EQ); - break; - case TGSI_OPCODE_SNE: - emit_sop(c, inst, BRW_CONDITIONAL_NEQ); - break; - case TGSI_OPCODE_MUL: - emit_alu2(c, inst, BRW_OPCODE_MUL); - break; - case TGSI_OPCODE_POW: - emit_pow(c, inst); - break; - case TGSI_OPCODE_MAD: - emit_mad(c, inst); - break; - case TGSI_OPCODE_TEX: - emit_tex(c, inst); - break; - case TGSI_OPCODE_TXB: - emit_txb(c, inst); - break; - case TGSI_OPCODE_TEXKILL: - emit_kil(c); - break; - case TGSI_OPCODE_IF: - assert(c->if_insn < MAX_IFSN); - c->if_inst[c->if_insn++] = brw_IF(p, BRW_EXECUTE_8); - break; - case TGSI_OPCODE_ELSE: - c->if_inst[c->if_insn-1] = brw_ELSE(p, c->if_inst[c->if_insn-1]); - break; - case TGSI_OPCODE_ENDIF: - assert(c->if_insn > 0); - brw_ENDIF(p, c->if_inst[--c->if_insn]); - break; - case TGSI_OPCODE_BGNSUB: - case TGSI_OPCODE_ENDSUB: - break; - case TGSI_OPCODE_CAL: - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); - brw_set_access_mode(p, BRW_ALIGN_1); - brw_ADD(p, deref_1ud(c->stack_index, 0), brw_ip_reg(), brw_imm_d(3*16)); - brw_set_access_mode(p, BRW_ALIGN_16); - brw_ADD(p, - get_addr_reg(c->stack_index), - get_addr_reg(c->stack_index), brw_imm_d(4)); -// orig_inst = inst->Data; -// orig_inst->Data = &p->store[p->nr_insn]; - assert(0); - brw_ADD(p, brw_ip_reg(), brw_ip_reg(), brw_imm_d(1*16)); - brw_pop_insn_state(p); - break; - - case TGSI_OPCODE_RET: -#if 0 - brw_push_insn_state(p); - brw_set_mask_control(p, BRW_MASK_DISABLE); - brw_ADD(p, - get_addr_reg(c->stack_index), - get_addr_reg(c->stack_index), brw_imm_d(-4)); - brw_set_access_mode(p, BRW_ALIGN_1); - brw_MOV(p, brw_ip_reg(), deref_1ud(c->stack_index, 0)); - brw_set_access_mode(p, BRW_ALIGN_16); - brw_pop_insn_state(p); -#else - emit_fb_write(c, inst); -#endif - - break; - case TGSI_OPCODE_LOOP: - c->loop_inst[c->loop_insn++] = brw_DO(p, BRW_EXECUTE_8); - break; - case TGSI_OPCODE_BRK: - brw_BREAK(p); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - break; - case TGSI_OPCODE_CONT: - brw_CONT(p); - brw_set_predicate_control(p, BRW_PREDICATE_NONE); - break; - case TGSI_OPCODE_ENDLOOP: - c->loop_insn--; - c->inst0 = c->inst1 = brw_WHILE(p, c->loop_inst[c->loop_insn]); - /* patch all the BREAK instructions from - last BEGINLOOP */ - while (c->inst0 > c->loop_inst[c->loop_insn]) { - c->inst0--; - if (c->inst0->header.opcode == BRW_OPCODE_BREAK) { - c->inst0->bits3.if_else.jump_count = c->inst1 - c->inst0 + 1; - c->inst0->bits3.if_else.pop_count = 0; - } else if (c->inst0->header.opcode == BRW_OPCODE_CONTINUE) { - c->inst0->bits3.if_else.jump_count = c->inst1 - c->inst0; - c->inst0->bits3.if_else.pop_count = 0; - } - } - break; - case TGSI_OPCODE_END: - emit_fb_write(c, inst); - break; - - default: - debug_printf("unsupported IR in fragment shader %d\n", - inst->Instruction.Opcode); - } -#if 0 - if (inst->CondUpdate) - brw_set_predicate_control(p, BRW_PREDICATE_NORMAL); - else - brw_set_predicate_control(p, BRW_PREDICATE_NONE); -#endif -} - - - - - - -void brw_wm_glsl_emit(struct brw_wm_compile *c) -{ - struct tgsi_parse_context parse; - struct brw_compile *p = &c->func; - - brw_init_compile(&c->func); - brw_set_compression_control(p, BRW_COMPRESSION_NONE); - - c->reg_index = 0; - c->if_insn = 0; - c->loop_insn = 0; - c->stack_index = brw_indirect(0,0); - - /* Do static register allocation and parameter interpolation: - */ - brw_wm_emit_decls( c ); - - /* Emit the actual program. All done with very direct translation, - * hopefully we can improve on this shortly... - */ - brw_MOV(p, get_addr_reg(c->stack_index), brw_address(c->stack)); - - tgsi_parse_init( &parse, c->fp->program.tokens ); - - while( !tgsi_parse_end_of_tokens( &parse ) ) - { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - /* already done */ - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - /* not handled yet */ - assert(0); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - brw_wm_emit_instruction(c, &parse.FullToken.FullInstruction); - break; - - default: - assert( 0 ); - } - } - - tgsi_parse_free (&parse); - - /* Fix up call targets: - */ -#if 0 - { - unsigned nr_insns = c->fp->program.Base.NumInstructions; - unsigned insn, target_insn; - struct tgsi_full_instruction *inst1, *inst2; - struct brw_instruction *brw_inst1, *brw_inst2; - int offset; - for (insn = 0; insn < nr_insns; insn++) { - inst1 = &c->fp->program.Base.Instructions[insn]; - brw_inst1 = inst1->Data; - switch (inst1->Opcode) { - case TGSI_OPCODE_CAL: - target_insn = inst1->BranchTarget; - inst2 = &c->fp->program.Base.Instructions[target_insn]; - brw_inst2 = inst2->Data; - offset = brw_inst2 - brw_inst1; - brw_set_src1(brw_inst1, brw_imm_d(offset*16)); - break; - default: - break; - } - } - } -#endif - - c->prog_data.total_grf = c->reg_index; - c->prog_data.total_scratch = 0; -} diff --git a/src/mesa/pipe/i965simple/brw_wm_iz.c b/src/mesa/pipe/i965simple/brw_wm_iz.c deleted file mode 100644 index 6c5f25bf39..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_iz.c +++ /dev/null @@ -1,214 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_wm.h" - - -#undef P /* prompted depth */ -#undef C /* computed */ -#undef N /* non-promoted? */ - -#define P 0 -#define C 1 -#define N 2 - -const struct { - unsigned mode:2; - unsigned sd_present:1; - unsigned sd_to_rt:1; - unsigned dd_present:1; - unsigned ds_present:1; -} wm_iz_table[IZ_BIT_MAX] = -{ - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { C, 0, 1, 0, 0 }, - { C, 0, 1, 0, 0 }, - { C, 1, 1, 0, 0 }, - { C, 1, 1, 0, 0 }, - { C, 0, 1, 0, 0 }, - { C, 0, 1, 0, 0 }, - { C, 1, 1, 1, 0 }, - { C, 1, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { C, 1, 1, 1, 0 }, - { C, 1, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { C, 0, 1, 0, 0 }, - { C, 0, 1, 0, 0 }, - { C, 1, 1, 0, 0 }, - { C, 1, 1, 0, 0 }, - { C, 0, 1, 0, 0 }, - { C, 0, 1, 0, 0 }, - { C, 1, 1, 1, 0 }, - { C, 1, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { C, 1, 1, 1, 0 }, - { C, 1, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { C, 0, 0, 0, 1 }, - { C, 0, 0, 0, 1 }, - { C, 0, 1, 0, 1 }, - { C, 0, 1, 0, 1 }, - { C, 1, 1, 0, 1 }, - { C, 1, 1, 0, 1 }, - { C, 0, 1, 0, 1 }, - { C, 0, 1, 0, 1 }, - { C, 1, 1, 1, 1 }, - { C, 1, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { C, 1, 1, 1, 1 }, - { C, 1, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { C, 0, 0, 0, 1 }, - { C, 0, 0, 0, 1 }, - { C, 0, 1, 0, 1 }, - { C, 0, 1, 0, 1 }, - { C, 1, 1, 0, 1 }, - { C, 1, 1, 0, 1 }, - { C, 0, 1, 0, 1 }, - { C, 0, 1, 0, 1 }, - { C, 1, 1, 1, 1 }, - { C, 1, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { C, 1, 1, 1, 1 }, - { C, 1, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { N, 1, 1, 0, 0 }, - { N, 0, 1, 0, 0 }, - { N, 0, 1, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { C, 0, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { P, 0, 0, 0, 0 }, - { N, 1, 1, 0, 0 }, - { C, 0, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { N, 1, 1, 0, 0 }, - { N, 0, 1, 0, 0 }, - { N, 0, 1, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { C, 0, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { P, 0, 0, 0, 0 }, - { N, 1, 1, 0, 0 }, - { C, 0, 1, 1, 0 }, - { C, 0, 1, 1, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { N, 1, 1, 0, 1 }, - { N, 0, 1, 0, 1 }, - { N, 0, 1, 0, 1 }, - { P, 0, 0, 0, 0 }, - { P, 0, 0, 0, 0 }, - { C, 0, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { P, 0, 0, 0, 0 }, - { N, 1, 1, 0, 1 }, - { C, 0, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { P, 0, 0, 0, 0 }, - { C, 0, 0, 0, 1 }, - { P, 0, 0, 0, 0 }, - { C, 0, 1, 0, 1 }, - { P, 0, 0, 0, 0 }, - { C, 1, 1, 0, 1 }, - { C, 0, 1, 0, 1 }, - { C, 0, 1, 0, 1 }, - { P, 0, 0, 0, 0 }, - { C, 1, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { P, 0, 0, 0, 0 }, - { C, 1, 1, 1, 1 }, - { C, 0, 1, 1, 1 }, - { C, 0, 1, 1, 1 } -}; - -void brw_wm_lookup_iz( unsigned line_aa, - unsigned lookup, - struct brw_wm_prog_key *key ) -{ - unsigned reg = 2; - - assert (lookup < IZ_BIT_MAX); - - if (lookup & IZ_PS_COMPUTES_DEPTH_BIT) - key->computes_depth = 1; - - if (wm_iz_table[lookup].sd_present) { - key->source_depth_reg = reg; - reg += 2; - } - - if (wm_iz_table[lookup].sd_to_rt) - key->source_depth_to_render_target = 1; - - if (wm_iz_table[lookup].ds_present || line_aa != AA_NEVER) { - key->aa_dest_stencil_reg = reg; - key->runtime_check_aads_emit = (!wm_iz_table[lookup].ds_present && - line_aa == AA_SOMETIMES); - reg++; - } - - if (wm_iz_table[lookup].dd_present) { - key->dest_depth_reg = reg; - reg+=2; - } - - key->nr_depth_regs = (reg+1)/2; -} - diff --git a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c b/src/mesa/pipe/i965simple/brw_wm_sampler_state.c deleted file mode 100644 index de42ffc5b1..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_sampler_state.c +++ /dev/null @@ -1,273 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_state.h" -#include "brw_defines.h" - -#include "pipe/p_util.h" - - -#define COMPAREFUNC_ALWAYS 0 -#define COMPAREFUNC_NEVER 0x1 -#define COMPAREFUNC_LESS 0x2 -#define COMPAREFUNC_EQUAL 0x3 -#define COMPAREFUNC_LEQUAL 0x4 -#define COMPAREFUNC_GREATER 0x5 -#define COMPAREFUNC_NOTEQUAL 0x6 -#define COMPAREFUNC_GEQUAL 0x7 - -/* Samplers aren't strictly wm state from the hardware's perspective, - * but that is the only situation in which we use them in this driver. - */ - -static int intel_translate_shadow_compare_func(unsigned func) -{ - switch(func) { - case PIPE_FUNC_NEVER: - return COMPAREFUNC_ALWAYS; - case PIPE_FUNC_LESS: - return COMPAREFUNC_LEQUAL; - case PIPE_FUNC_LEQUAL: - return COMPAREFUNC_LESS; - case PIPE_FUNC_GREATER: - return COMPAREFUNC_GEQUAL; - case PIPE_FUNC_GEQUAL: - return COMPAREFUNC_GREATER; - case PIPE_FUNC_NOTEQUAL: - return COMPAREFUNC_EQUAL; - case PIPE_FUNC_EQUAL: - return COMPAREFUNC_NOTEQUAL; - case PIPE_FUNC_ALWAYS: - return COMPAREFUNC_NEVER; - } - - debug_printf("Unknown value in %s: %x\n", __FUNCTION__, func); - return COMPAREFUNC_NEVER; -} - -/* The brw (and related graphics cores) do not support GL_CLAMP. The - * Intel drivers for "other operating systems" implement GL_CLAMP as - * GL_CLAMP_TO_EDGE, so the same is done here. - */ -static unsigned translate_wrap_mode( int wrap ) -{ - switch( wrap ) { - case PIPE_TEX_WRAP_REPEAT: - return BRW_TEXCOORDMODE_WRAP; - case PIPE_TEX_WRAP_CLAMP: - return BRW_TEXCOORDMODE_CLAMP; - case PIPE_TEX_WRAP_CLAMP_TO_EDGE: - return BRW_TEXCOORDMODE_CLAMP; /* conform likes it this way */ - case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - return BRW_TEXCOORDMODE_CLAMP_BORDER; - case PIPE_TEX_WRAP_MIRROR_REPEAT: - return BRW_TEXCOORDMODE_MIRROR; - default: - return BRW_TEXCOORDMODE_WRAP; - } -} - - -static unsigned U_FIXED(float value, unsigned frac_bits) -{ - value *= (1<cache[BRW_SAMPLER_DEFAULT_COLOR], &sdc ); -} - - -/* - */ -static void brw_update_sampler_state( const struct pipe_sampler_state *pipe_sampler, - unsigned sdc_gs_offset, - struct brw_sampler_state *sampler) -{ - memset(sampler, 0, sizeof(*sampler)); - - switch (pipe_sampler->min_mip_filter) { - case PIPE_TEX_FILTER_NEAREST: - sampler->ss0.min_filter = BRW_MAPFILTER_NEAREST; - break; - case PIPE_TEX_FILTER_LINEAR: - sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR; - break; - default: - break; - } - - switch (pipe_sampler->min_mip_filter) { - case PIPE_TEX_MIPFILTER_NEAREST: - sampler->ss0.mip_filter = BRW_MIPFILTER_NEAREST; - break; - case PIPE_TEX_MIPFILTER_LINEAR: - sampler->ss0.mip_filter = BRW_MIPFILTER_LINEAR; - break; - case PIPE_TEX_MIPFILTER_NONE: - sampler->ss0.mip_filter = BRW_MIPFILTER_NONE; - break; - default: - break; - } - /* Set Anisotropy: - */ - if (pipe_sampler->max_anisotropy > 1.0) { - sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC; - sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC; - - if (pipe_sampler->max_anisotropy > 2.0) { - sampler->ss3.max_aniso = MAX2((pipe_sampler->max_anisotropy - 2) / 2, - BRW_ANISORATIO_16); - } - } - else { - switch (pipe_sampler->mag_img_filter) { - case PIPE_TEX_FILTER_NEAREST: - sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST; - break; - case PIPE_TEX_FILTER_LINEAR: - sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR; - break; - default: - break; - } - } - - sampler->ss1.s_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_s); - sampler->ss1.r_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_r); - sampler->ss1.t_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_t); - - /* Fulsim complains if I don't do this. Hardware doesn't mind: - */ -#if 0 - if (texObj->Target == GL_TEXTURE_CUBE_MAP_ARB) { - sampler->ss1.r_wrap_mode = BRW_TEXCOORDMODE_CUBE; - sampler->ss1.s_wrap_mode = BRW_TEXCOORDMODE_CUBE; - sampler->ss1.t_wrap_mode = BRW_TEXCOORDMODE_CUBE; - } -#endif - - /* Set shadow function: - */ - if (pipe_sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - /* Shadowing is "enabled" by emitting a particular sampler - * message (sample_c). So need to recompile WM program when - * shadow comparison is enabled on each/any texture unit. - */ - sampler->ss0.shadow_function = intel_translate_shadow_compare_func(pipe_sampler->compare_func); - } - - /* Set LOD bias: - */ - sampler->ss0.lod_bias = S_FIXED(CLAMP(pipe_sampler->lod_bias, -16, 15), 6); - - sampler->ss0.lod_preclamp = 1; /* OpenGL mode */ - sampler->ss0.default_color_mode = 0; /* OpenGL/DX10 mode */ - - /* Set BaseMipLevel, MaxLOD, MinLOD: - * - * XXX: I don't think that using firstLevel, lastLevel works, - * because we always setup the surface state as if firstLevel == - * level zero. Probably have to subtract firstLevel from each of - * these: - */ - sampler->ss0.base_level = U_FIXED(0, 1); - - sampler->ss1.max_lod = U_FIXED(MIN2(MAX2(pipe_sampler->max_lod, 0), 13), 6); - sampler->ss1.min_lod = U_FIXED(MIN2(MAX2(pipe_sampler->min_lod, 0), 13), 6); - - sampler->ss2.default_color_pointer = sdc_gs_offset >> 5; -} - - - -/* All samplers must be uploaded in a single contiguous array, which - * complicates various things. However, this is still too confusing - - * FIXME: simplify all the different new texture state flags. - */ -static void upload_wm_samplers(struct brw_context *brw) -{ - unsigned unit; - unsigned sampler_count = 0; - - /* BRW_NEW_SAMPLER */ - for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { - /* determine unit enable/disable by looking for a bound texture */ - if (brw->attribs.Texture[unit]) { - const struct pipe_sampler_state *sampler = brw->attribs.Samplers[unit]; - unsigned sdc_gs_offset = upload_default_color(brw, sampler->border_color); - - brw_update_sampler_state(sampler, - sdc_gs_offset, - &brw->wm.sampler[unit]); - - sampler_count = unit + 1; - } - } - - if (brw->wm.sampler_count != sampler_count) { - brw->wm.sampler_count = sampler_count; - brw->state.dirty.cache |= CACHE_NEW_SAMPLER; - } - - brw->wm.sampler_gs_offset = 0; - - if (brw->wm.sampler_count) - brw->wm.sampler_gs_offset = - brw_cache_data_sz(&brw->cache[BRW_SAMPLER], - brw->wm.sampler, - sizeof(struct brw_sampler_state) * brw->wm.sampler_count); -} - -const struct brw_tracked_state brw_wm_samplers = { - .dirty = { - .brw = BRW_NEW_SAMPLER, - .cache = 0 - }, - .update = upload_wm_samplers -}; - diff --git a/src/mesa/pipe/i965simple/brw_wm_state.c b/src/mesa/pipe/i965simple/brw_wm_state.c deleted file mode 100644 index 5ccd488842..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_state.c +++ /dev/null @@ -1,194 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#include "brw_context.h" -#include "brw_state.h" -#include "brw_defines.h" -#include "brw_wm.h" -#include "pipe/p_util.h" - -/*********************************************************************** - * WM unit - fragment programs and rasterization - */ -static void upload_wm_unit(struct brw_context *brw ) -{ - struct brw_wm_unit_state wm; - unsigned max_threads; - unsigned per_thread; - - if (BRW_DEBUG & DEBUG_SINGLE_THREAD) - max_threads = 0; - else - max_threads = 31; - - - memset(&wm, 0, sizeof(wm)); - - /* CACHE_NEW_WM_PROG */ - wm.thread0.grf_reg_count = align(brw->wm.prog_data->total_grf, 16) / 16 - 1; - wm.thread0.kernel_start_pointer = brw->wm.prog_gs_offset >> 6; - wm.thread3.dispatch_grf_start_reg = brw->wm.prog_data->first_curbe_grf; - wm.thread3.urb_entry_read_length = brw->wm.prog_data->urb_read_length; - wm.thread3.const_urb_entry_read_length = brw->wm.prog_data->curb_read_length; - - wm.wm5.max_threads = max_threads; - - per_thread = align(brw->wm.prog_data->total_scratch, 1024); - assert(per_thread <= 12 * 1024); - -#if 0 - if (brw->wm.prog_data->total_scratch) { - unsigned total = per_thread * (max_threads + 1); - - /* Scratch space -- just have to make sure there is sufficient - * allocated for the active program and current number of threads. - */ - brw->wm.scratch_buffer_size = total; - if (brw->wm.scratch_buffer && - brw->wm.scratch_buffer_size > brw->wm.scratch_buffer->size) { - dri_bo_unreference(brw->wm.scratch_buffer); - brw->wm.scratch_buffer = NULL; - } - if (!brw->wm.scratch_buffer) { - brw->wm.scratch_buffer = dri_bo_alloc(intel->intelScreen->bufmgr, - "wm scratch", - brw->wm.scratch_buffer_size, - 4096, DRM_BO_FLAG_MEM_TT); - } - } - /* XXX: Scratch buffers are not implemented correectly. - * - * The scratch offset to be programmed into wm is relative to the general - * state base address. However, using dri_bo_alloc/dri_bo_emit_reloc (or - * the previous bmGenBuffers scheme), we get an offset relative to the - * start of framebuffer. Even before then, it was broken in other ways, - * so just fail for now if we hit that path. - */ - assert(brw->wm.prog_data->total_scratch == 0); -#endif - - /* CACHE_NEW_SURFACE */ - wm.thread1.binding_table_entry_count = brw->wm.nr_surfaces; - - /* BRW_NEW_CURBE_OFFSETS */ - wm.thread3.const_urb_entry_read_offset = brw->curbe.wm_start * 2; - - wm.thread3.urb_entry_read_offset = 0; - wm.thread1.depth_coef_urb_read_offset = 1; - wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; - - /* CACHE_NEW_SAMPLER */ - wm.wm4.sampler_count = (brw->wm.sampler_count + 1) / 4; - wm.wm4.sampler_state_pointer = brw->wm.sampler_gs_offset >> 5; - - /* BRW_NEW_FRAGMENT_PROGRAM */ - { - const struct brw_fragment_program *fp = brw->attribs.FragmentProgram; - - if (fp->UsesDepth) - wm.wm5.program_uses_depth = 1; /* as far as we can tell */ - - if (fp->ComputesDepth) - wm.wm5.program_computes_depth = 1; - - /* BRW_NEW_ALPHA_TEST */ - if (fp->UsesKill || - brw->attribs.DepthStencil->alpha.enabled) - wm.wm5.program_uses_killpixel = 1; - - wm.wm5.enable_8_pix = 1; - } - - wm.wm5.thread_dispatch_enable = 1; /* AKA: color_write */ - wm.wm5.legacy_line_rast = 0; - wm.wm5.legacy_global_depth_bias = 0; - wm.wm5.early_depth_test = 1; /* never need to disable */ - wm.wm5.line_aa_region_width = 0; - wm.wm5.line_endcap_aa_region_width = 1; - - /* BRW_NEW_RASTERIZER */ - if (brw->attribs.Raster->poly_stipple_enable) - wm.wm5.polygon_stipple = 1; - -#if 0 - if (brw->attribs.Polygon->OffsetFill) { - wm.wm5.depth_offset = 1; - /* Something wierd going on with legacy_global_depth_bias, - * offset_constant, scaling and MRD. This value passes glean - * but gives some odd results elsewere (eg. the - * quad-offset-units test). - */ - wm.global_depth_offset_constant = brw->attribs.Polygon->OffsetUnits * 2; - - /* This is the only value that passes glean: - */ - wm.global_depth_offset_scale = brw->attribs.Polygon->OffsetFactor; - } -#endif - - if (brw->attribs.Raster->line_stipple_enable) { - wm.wm5.line_stipple = 1; - } - - if (BRW_DEBUG & DEBUG_STATS) - wm.wm4.stats_enable = 1; - - brw->wm.state_gs_offset = brw_cache_data( &brw->cache[BRW_WM_UNIT], &wm ); - - if (brw->wm.prog_data->total_scratch) { - /* - dri_emit_reloc(brw->cache[BRW_WM_UNIT].pool->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - (per_thread / 1024) - 1, - brw->wm.state_gs_offset + - ((char *)&wm.thread2 - (char *)&wm), - brw->wm.scratch_buffer); - */ - } else { - wm.thread2.scratch_space_base_pointer = 0; - } -} - -const struct brw_tracked_state brw_wm_unit = { - .dirty = { - .brw = (BRW_NEW_RASTERIZER | - BRW_NEW_ALPHA_TEST | - BRW_NEW_FS | - BRW_NEW_CURBE_OFFSETS), - - .cache = (CACHE_NEW_SURFACE | - CACHE_NEW_WM_PROG | - CACHE_NEW_SAMPLER) - }, - .update = upload_wm_unit -}; - diff --git a/src/mesa/pipe/i965simple/brw_wm_surface_state.c b/src/mesa/pipe/i965simple/brw_wm_surface_state.c deleted file mode 100644 index d16d919bce..0000000000 --- a/src/mesa/pipe/i965simple/brw_wm_surface_state.c +++ /dev/null @@ -1,304 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_context.h" -#include "brw_state.h" -#include "brw_defines.h" - -static unsigned translate_tex_target( enum pipe_texture_target target ) -{ - switch (target) { - case PIPE_TEXTURE_1D: - return BRW_SURFACE_1D; - - case PIPE_TEXTURE_2D: - return BRW_SURFACE_2D; - - case PIPE_TEXTURE_3D: - return BRW_SURFACE_3D; - - case PIPE_TEXTURE_CUBE: - return BRW_SURFACE_CUBE; - - default: - assert(0); - return 0; - } -} - -static unsigned translate_tex_format( enum pipe_format pipe_format ) -{ - switch( pipe_format ) { - case PIPE_FORMAT_U_L8: - return BRW_SURFACEFORMAT_L8_UNORM; - - case PIPE_FORMAT_U_I8: - return BRW_SURFACEFORMAT_I8_UNORM; - - case PIPE_FORMAT_U_A8: - return BRW_SURFACEFORMAT_A8_UNORM; - - case PIPE_FORMAT_U_A8_L8: - return BRW_SURFACEFORMAT_L8A8_UNORM; - - case PIPE_FORMAT_R8G8B8_UNORM: - assert(0); /* not supported for sampling */ - return BRW_SURFACEFORMAT_R8G8B8_UNORM; - - case PIPE_FORMAT_B8G8R8A8_UNORM: - return BRW_SURFACEFORMAT_B8G8R8A8_UNORM; - - case PIPE_FORMAT_R8G8B8A8_UNORM: - return BRW_SURFACEFORMAT_R8G8B8A8_UNORM; - - case PIPE_FORMAT_R5G6B5_UNORM: - return BRW_SURFACEFORMAT_B5G6R5_UNORM; - - case PIPE_FORMAT_A1R5G5B5_UNORM: - return BRW_SURFACEFORMAT_B5G5R5A1_UNORM; - - case PIPE_FORMAT_A4R4G4B4_UNORM: - return BRW_SURFACEFORMAT_B4G4R4A4_UNORM; - - case PIPE_FORMAT_YCBCR_REV: - return BRW_SURFACEFORMAT_YCRCB_NORMAL; - - case PIPE_FORMAT_YCBCR: - return BRW_SURFACEFORMAT_YCRCB_SWAPUVY; -#if 0 - case PIPE_FORMAT_RGB_FXT1: - case PIPE_FORMAT_RGBA_FXT1: - return BRW_SURFACEFORMAT_FXT1; -#endif - - case PIPE_FORMAT_Z16_UNORM: - return BRW_SURFACEFORMAT_I16_UNORM; -#if 0 - case PIPE_FORMAT_RGB_DXT1: - return BRW_SURFACEFORMAT_DXT1_RGB; - - case PIPE_FORMAT_RGBA_DXT1: - return BRW_SURFACEFORMAT_BC1_UNORM; - - case PIPE_FORMAT_RGBA_DXT3: - return BRW_SURFACEFORMAT_BC2_UNORM; - - case PIPE_FORMAT_RGBA_DXT5: - return BRW_SURFACEFORMAT_BC3_UNORM; - - case PIPE_FORMAT_SRGBA8: - return BRW_SURFACEFORMAT_R8G8B8A8_UNORM_SRGB; - case PIPE_FORMAT_SRGB_DXT1: - return BRW_SURFACEFORMAT_BC1_UNORM_SRGB; -#endif - - default: - assert(0); - return 0; - } -} - -static unsigned brw_buffer_offset(struct brw_context *brw, - struct pipe_buffer *buffer) -{ - return brw->winsys->get_buffer_offset(brw->winsys, - buffer, - 0); -} - -static -void brw_update_texture_surface( struct brw_context *brw, - unsigned unit ) -{ - const struct brw_texture *tObj = brw->attribs.Texture[unit]; - struct brw_surface_state surf; - - memset(&surf, 0, sizeof(surf)); - - surf.ss0.mipmap_layout_mode = BRW_SURFACE_MIPMAPLAYOUT_BELOW; - surf.ss0.surface_type = translate_tex_target(tObj->base.target); - surf.ss0.surface_format = translate_tex_format(tObj->base.format); - - /* This is ok for all textures with channel width 8bit or less: - */ -/* surf.ss0.data_return_format = BRW_SURFACERETURNFORMAT_S1; */ - - /* Updated in emit_reloc */ - surf.ss1.base_addr = brw_buffer_offset( brw, tObj->buffer ); - - surf.ss2.mip_count = tObj->base.last_level; - surf.ss2.width = tObj->base.width[0] - 1; - surf.ss2.height = tObj->base.height[0] - 1; - - surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; - surf.ss3.tiled_surface = 0; /* always zero */ - surf.ss3.pitch = tObj->pitch - 1; - surf.ss3.depth = tObj->base.depth[0] - 1; - - surf.ss4.min_lod = 0; - - if (tObj->base.target == PIPE_TEXTURE_CUBE) { - surf.ss0.cube_pos_x = 1; - surf.ss0.cube_pos_y = 1; - surf.ss0.cube_pos_z = 1; - surf.ss0.cube_neg_x = 1; - surf.ss0.cube_neg_y = 1; - surf.ss0.cube_neg_z = 1; - } - - brw->wm.bind.surf_ss_offset[unit + 1] = - brw_cache_data( &brw->cache[BRW_SS_SURFACE], &surf ); -} - - - -#define OFFSET(TYPE, FIELD) ( (unsigned)&(((TYPE *)0)->FIELD) ) - - -static void upload_wm_surfaces(struct brw_context *brw ) -{ - unsigned i; - - { - struct brw_surface_state surf; - - /* BRW_NEW_FRAMEBUFFER - */ - struct pipe_surface *pipe_surface = brw->attribs.FrameBuffer.cbufs[0];/*fixme*/ - - memset(&surf, 0, sizeof(surf)); - - if (pipe_surface != NULL) { - if (pipe_surface->cpp == 4) - surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; - else - surf.ss0.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM; - - surf.ss0.surface_type = BRW_SURFACE_2D; - - surf.ss1.base_addr = brw_buffer_offset( brw, pipe_surface->buffer ); - - surf.ss2.width = pipe_surface->width - 1; - surf.ss2.height = pipe_surface->height - 1; - surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; - surf.ss3.tiled_surface = 0; - surf.ss3.pitch = (pipe_surface->pitch * pipe_surface->cpp) - 1; - } else { - surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; - surf.ss0.surface_type = BRW_SURFACE_NULL; - } - - /* BRW_NEW_BLEND */ - surf.ss0.color_blend = (!brw->attribs.Blend->logicop_enable && - brw->attribs.Blend->blend_enable); - - - surf.ss0.writedisable_red = !(brw->attribs.Blend->colormask & PIPE_MASK_R); - surf.ss0.writedisable_green = !(brw->attribs.Blend->colormask & PIPE_MASK_G); - surf.ss0.writedisable_blue = !(brw->attribs.Blend->colormask & PIPE_MASK_B); - surf.ss0.writedisable_alpha = !(brw->attribs.Blend->colormask & PIPE_MASK_A); - - - - - brw->wm.bind.surf_ss_offset[0] = brw_cache_data( &brw->cache[BRW_SS_SURFACE], &surf ); - - brw->wm.nr_surfaces = 1; - } - - - /* BRW_NEW_TEXTURE - */ - for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { - const struct brw_texture *texUnit = brw->attribs.Texture[i]; - - if (texUnit && - texUnit->base.refcount/*(texUnit->refcount > 0) == really used */) { - - brw_update_texture_surface(brw, i); - - brw->wm.nr_surfaces = i+2; - } - else { - brw->wm.bind.surf_ss_offset[i+1] = 0; - } - } - - brw->wm.bind_ss_offset = brw_cache_data( &brw->cache[BRW_SS_SURF_BIND], - &brw->wm.bind ); -} - - -/* KW: Will find a different way to acheive this, see for example the - * state caches with relocs in the i915 swz driver. - */ -#if 0 -static void emit_reloc_wm_surfaces(struct brw_context *brw) -{ - int unit; - - if (brw->state.draw_region != NULL) { - /* Emit framebuffer relocation */ - dri_emit_reloc(brw_cache_buffer(brw, BRW_SS_SURFACE), - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0, - brw->wm.bind.surf_ss_offset[0] + - offsetof(struct brw_surface_state, ss1), - brw->state.draw_region->buffer); - } - - /* Emit relocations for texture buffers */ - for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { - struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[unit]; - struct gl_texture_object *tObj = texUnit->_Current; - struct intel_texture_object *intelObj = intel_texture_object(tObj); - - if (texUnit->_ReallyEnabled && intelObj->mt != NULL) { - dri_emit_reloc(brw_cache_buffer(brw, BRW_SS_SURFACE), - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - 0, - brw->wm.bind.surf_ss_offset[unit + 1] + - offsetof(struct brw_surface_state, ss1), - intelObj->mt->region->buffer); - } - } -} -#endif - -const struct brw_tracked_state brw_wm_surfaces = { - .dirty = { - .brw = (BRW_NEW_FRAMEBUFFER | - BRW_NEW_BLEND | - BRW_NEW_TEXTURE), - .cache = 0 - }, - .update = upload_wm_surfaces, -}; diff --git a/src/mesa/pipe/llvm/Makefile b/src/mesa/pipe/llvm/Makefile deleted file mode 100644 index 9c6e16d86b..0000000000 --- a/src/mesa/pipe/llvm/Makefile +++ /dev/null @@ -1,83 +0,0 @@ -# -*-makefile-*- -TOP = ../../../.. -include $(TOP)/configs/current - -LIBNAME = gallivm - - -GALLIVM_SOURCES = \ - gallivm.cpp \ - gallivm_cpu.cpp \ - instructions.cpp \ - loweringpass.cpp \ - tgsitollvm.cpp \ - storage.cpp \ - storagesoa.cpp \ - instructionssoa.cpp - -INC_SOURCES = gallivm_builtins.cpp - -CPP_SOURCES = \ - $(GALLIVM_SOURCES) - -C_SOURCES = -ASM_SOURCES = - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(CPP_SOURCES:.cpp=.o) \ - $(ASM_SOURCES:.S=.o) - -### Include directories -INCLUDES = \ - -I. \ - -I$(TOP)/src/mesa/pipe \ - -I$(TOP)/src/mesa \ - -I$(TOP)/include - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.cpp.o: - $(CXX) -c $(INCLUDES) $(LLVM_CXXFLAGS) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -##### TARGETS ##### - -default:: depend symlinks $(LIBNAME) - - -$(LIBNAME): $(OBJECTS) Makefile - $(TOP)/bin/mklib -o $@ -static $(OBJECTS) - - -depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ - $(ASM_SOURCES) $(INC_SOURCES) 2> /dev/null - - -gallivm_builtins.cpp: llvm_builtins.c - clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean: - -rm -f *.o */*.o *~ *.so *~ server/*.o - -rm -f depend depend.bak - -rm -f gallivm_builtins.cpp - -symlinks: - - -include depend diff --git a/src/mesa/pipe/llvm/gallivm.cpp b/src/mesa/pipe/llvm/gallivm.cpp deleted file mode 100644 index da0105c2c9..0000000000 --- a/src/mesa/pipe/llvm/gallivm.cpp +++ /dev/null @@ -1,327 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "gallivm.h" -#include "gallivm_p.h" - -#include "instructions.h" -#include "loweringpass.h" -#include "storage.h" -#include "tgsitollvm.h" - -#include "pipe/p_context.h" -#include "pipe/p_shader_tokens.h" - -#include "pipe/tgsi/exec/tgsi_exec.h" -#include "pipe/tgsi/util/tgsi_dump.h" - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -static int GLOBAL_ID = 0; - -using namespace llvm; - -static inline -void AddStandardCompilePasses(PassManager &PM) -{ - PM.add(new LoweringPass()); - PM.add(createVerifierPass()); // Verify that input is correct - - PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp - - //PM.add(createStripSymbolsPass(true)); - - PM.add(createRaiseAllocationsPass()); // call %malloc -> malloc inst - PM.add(createCFGSimplificationPass()); // Clean up disgusting code - PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas - PM.add(createGlobalOptimizerPass()); // Optimize out global vars - PM.add(createGlobalDCEPass()); // Remove unused fns and globs - PM.add(createIPConstantPropagationPass());// IP Constant Propagation - PM.add(createDeadArgEliminationPass()); // Dead argument elimination - PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE - PM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE - - PM.add(createPruneEHPass()); // Remove dead EH info - - PM.add(createFunctionInliningPass()); // Inline small functions - PM.add(createArgumentPromotionPass()); // Scalarize uninlined fn args - - PM.add(createTailDuplicationPass()); // Simplify cfg by copying code - PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl. - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas - PM.add(createInstructionCombiningPass()); // Combine silly seq's - PM.add(createCondPropagationPass()); // Propagate conditionals - - PM.add(createTailCallEliminationPass()); // Eliminate tail calls - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createReassociatePass()); // Reassociate expressions - PM.add(createLoopRotatePass()); - PM.add(createLICMPass()); // Hoist loop invariants - PM.add(createLoopUnswitchPass()); // Unswitch loops. - PM.add(createLoopIndexSplitPass()); // Index split loops. - PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc - PM.add(createIndVarSimplifyPass()); // Canonicalize indvars - PM.add(createLoopUnrollPass()); // Unroll small loops - PM.add(createInstructionCombiningPass()); // Clean up after the unroller - PM.add(createGVNPass()); // Remove redundancies - PM.add(createSCCPPass()); // Constant prop with SCCP - - // Run instcombine after redundancy elimination to exploit opportunities - // opened up by them. - PM.add(createInstructionCombiningPass()); - PM.add(createCondPropagationPass()); // Propagate conditionals - - PM.add(createDeadStoreEliminationPass()); // Delete dead stores - PM.add(createAggressiveDCEPass()); // SSA based 'Aggressive DCE' - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createSimplifyLibCallsPass()); // Library Call Optimizations - PM.add(createDeadTypeEliminationPass()); // Eliminate dead types - PM.add(createConstantMergePass()); // Merge dup global constants -} - -void gallivm_prog_delete(struct gallivm_prog *prog) -{ - delete prog->module; - prog->module = 0; - prog->function = 0; - free(prog); -} - -static inline void -constant_interpolation(float (*inputs)[16][4], - const struct tgsi_interp_coef *coefs, - unsigned attrib, - unsigned chan) -{ - unsigned i; - - for (i = 0; i < QUAD_SIZE; ++i) { - inputs[i][attrib][chan] = coefs[attrib].a0[chan]; - } -} - -static inline void -linear_interpolation(float (*inputs)[16][4], - const struct tgsi_interp_coef *coefs, - unsigned attrib, - unsigned chan) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = inputs[i][0][0]; - const float y = inputs[i][0][1]; - - inputs[i][attrib][chan] = - coefs[attrib].a0[chan] + - coefs[attrib].dadx[chan] * x + - coefs[attrib].dady[chan] * y; - } -} - -static inline void -perspective_interpolation(float (*inputs)[16][4], - const struct tgsi_interp_coef *coefs, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = inputs[i][0][0]; - const float y = inputs[i][0][1]; - /* WPOS.w here is really 1/w */ - const float w = 1.0f / inputs[i][0][3]; - assert(inputs[i][0][3] != 0.0); - - inputs[i][attrib][chan] = - (coefs[attrib].a0[chan] + - coefs[attrib].dadx[chan] * x + - coefs[attrib].dady[chan] * y) * w; - } -} - -void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix) -{ - if (!ir || !ir->module) - return; - - if (file_prefix) { - std::ostringstream stream; - stream << file_prefix; - stream << ir->id; - stream << ".ll"; - std::string name = stream.str(); - std::ofstream out(name.c_str()); - if (!out) { - std::cerr<<"Can't open file : "<module); - out.close(); - } else { - const llvm::Module::FunctionListType &funcs = ir->module->getFunctionList(); - llvm::Module::FunctionListType::const_iterator itr; - std::cout<<"; ---------- Start shader "<id<id<num_interp; ++i) { - const gallivm_interpolate &interp = prog->interpolators[i]; - switch (interp.type) { - case TGSI_INTERPOLATE_CONSTANT: - constant_interpolation(inputs, coef, interp.attrib, interp.chan); - break; - - case TGSI_INTERPOLATE_LINEAR: - linear_interpolation(inputs, coef, interp.attrib, interp.chan); - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - perspective_interpolation(inputs, coef, interp.attrib, interp.chan); - break; - - default: - assert( 0 ); - } - } -} - - -struct gallivm_ir * gallivm_ir_new(enum gallivm_shader_type type) -{ - struct gallivm_ir *ir = - (struct gallivm_ir *)calloc(1, sizeof(struct gallivm_ir)); - ++GLOBAL_ID; - ir->id = GLOBAL_ID; - ir->type = type; - - return ir; -} - -void gallivm_ir_set_layout(struct gallivm_ir *ir, - enum gallivm_vector_layout layout) -{ - ir->layout = layout; -} - -void gallivm_ir_set_components(struct gallivm_ir *ir, int num) -{ - ir->num_components = num; -} - -void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, - const struct tgsi_token *tokens) -{ - std::cout << "Creating llvm from: " <module = mod; - gallivm_ir_dump(ir, 0); -} - -void gallivm_ir_delete(struct gallivm_ir *ir) -{ - delete ir->module; - free(ir); -} - -struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) -{ - struct gallivm_prog *prog = - (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); - llvm::Module *mod = llvm::CloneModule(ir->module); - prog->num_consts = ir->num_consts; - memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators)); - prog->num_interp = ir->num_interp; - - /* Run optimization passes over it */ - PassManager passes; - passes.add(new TargetData(mod)); - AddStandardCompilePasses(passes); - passes.run(*mod); - prog->module = mod; - - std::cout << "After optimizations:"<dump(); - - return prog; -} - -#endif /* MESA_LLVM */ diff --git a/src/mesa/pipe/llvm/gallivm.h b/src/mesa/pipe/llvm/gallivm.h deleted file mode 100644 index 92da4bca7f..0000000000 --- a/src/mesa/pipe/llvm/gallivm.h +++ /dev/null @@ -1,103 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef GALLIVM_H -#define GALLIVM_H - -#if defined __cplusplus -extern "C" { -#endif - -#include "pipe/p_state.h" - -#ifdef MESA_LLVM - -struct tgsi_token; - -struct gallivm_ir; -struct gallivm_prog; -struct gallivm_cpu_engine; -struct tgsi_interp_coef; -struct tgsi_sampler; -struct tgsi_exec_vector; - -enum gallivm_shader_type { - GALLIVM_VS, - GALLIVM_FS -}; - -enum gallivm_vector_layout { - GALLIVM_AOS, - GALLIVM_SOA -}; - -struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type); -void gallivm_ir_set_layout(struct gallivm_ir *ir, - enum gallivm_vector_layout layout); -void gallivm_ir_set_components(struct gallivm_ir *ir, int num); -void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, - const struct tgsi_token *tokens); -void gallivm_ir_delete(struct gallivm_ir *ir); - - -struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir); - -void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - const struct tgsi_interp_coef *coefs); -void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); - - -struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); -struct gallivm_cpu_engine *gallivm_global_cpu_engine(); -int gallivm_cpu_vs_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - struct tgsi_exec_vector *temps); -int gallivm_cpu_fs_exec(struct gallivm_prog *prog, - float x, float y, - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*consts)[4], - struct tgsi_sampler *samplers); -void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); -void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); - - -#endif /* MESA_LLVM */ - -#if defined __cplusplus -} // extern "C" -#endif - -#endif diff --git a/src/mesa/pipe/llvm/gallivm_builtins.cpp b/src/mesa/pipe/llvm/gallivm_builtins.cpp deleted file mode 100644 index 1796f0a177..0000000000 --- a/src/mesa/pipe/llvm/gallivm_builtins.cpp +++ /dev/null @@ -1,567 +0,0 @@ -// Generated by llvm2cpp - DO NOT MODIFY! - - -Module* createGallivmBuiltins(Module *mod) { - -mod->setModuleIdentifier("shader"); - -// Type Definitions -ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 25); - -PointerType* PointerTy_1 = PointerType::get(ArrayTy_0, 0); - -std::vectorFuncTy_2_args; -FuncTy_2_args.push_back(Type::FloatTy); -FuncTy_2_args.push_back(Type::FloatTy); -FunctionType* FuncTy_2 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_2_args, - /*isVarArg=*/false); - -PointerType* PointerTy_3 = PointerType::get(FuncTy_2, 0); - -VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); - -std::vectorFuncTy_5_args; -FuncTy_5_args.push_back(VectorTy_4); -FunctionType* FuncTy_5 = FunctionType::get( - /*Result=*/VectorTy_4, - /*Params=*/FuncTy_5_args, - /*isVarArg=*/false); - -std::vectorFuncTy_6_args; -FuncTy_6_args.push_back(VectorTy_4); -FuncTy_6_args.push_back(VectorTy_4); -FuncTy_6_args.push_back(VectorTy_4); -FunctionType* FuncTy_6 = FunctionType::get( - /*Result=*/VectorTy_4, - /*Params=*/FuncTy_6_args, - /*isVarArg=*/false); - -VectorType* VectorTy_7 = VectorType::get(IntegerType::get(32), 4); - -std::vectorFuncTy_9_args; -FunctionType* FuncTy_9 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_9_args, - /*isVarArg=*/true); - -PointerType* PointerTy_8 = PointerType::get(FuncTy_9, 0); - -PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8), 0); - -std::vectorFuncTy_12_args; -FuncTy_12_args.push_back(Type::FloatTy); -FunctionType* FuncTy_12 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_12_args, - /*isVarArg=*/false); - -PointerType* PointerTy_11 = PointerType::get(FuncTy_12, 0); - -std::vectorFuncTy_13_args; -FuncTy_13_args.push_back(VectorTy_4); -FunctionType* FuncTy_13 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_13_args, - /*isVarArg=*/false); - - -// Function Declarations - -Function* func_approx = new Function( - /*Type=*/FuncTy_2, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"approx", mod); -func_approx->setCallingConv(CallingConv::C); -const ParamAttrsList *func_approx_PAL = 0; -func_approx->setParamAttrs(func_approx_PAL); - -Function* func_powf = new Function( - /*Type=*/FuncTy_2, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"powf", mod); // (external, no body) -func_powf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_powf_PAL = 0; -func_powf->setParamAttrs(func_powf_PAL); - -Function* func_lit = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"lit", mod); -func_lit->setCallingConv(CallingConv::C); -const ParamAttrsList *func_lit_PAL = 0; -func_lit->setParamAttrs(func_lit_PAL); - -Function* func_cmp = new Function( - /*Type=*/FuncTy_6, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"cmp", mod); -func_cmp->setCallingConv(CallingConv::C); -const ParamAttrsList *func_cmp_PAL = 0; -{ - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_cmp_PAL = ParamAttrsList::get(Attrs); - -} -func_cmp->setParamAttrs(func_cmp_PAL); - -Function* func_vcos = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"vcos", mod); -func_vcos->setCallingConv(CallingConv::C); -const ParamAttrsList *func_vcos_PAL = 0; -func_vcos->setParamAttrs(func_vcos_PAL); - -Function* func_printf = new Function( - /*Type=*/FuncTy_9, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", mod); // (external, no body) -func_printf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_printf_PAL = 0; -func_printf->setParamAttrs(func_printf_PAL); - -Function* func_cosf = new Function( - /*Type=*/FuncTy_12, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"cosf", mod); // (external, no body) -func_cosf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_cosf_PAL = 0; -func_cosf->setParamAttrs(func_cosf_PAL); - -Function* func_scs = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"scs", mod); -func_scs->setCallingConv(CallingConv::C); -const ParamAttrsList *func_scs_PAL = 0; -func_scs->setParamAttrs(func_scs_PAL); - -Function* func_sinf = new Function( - /*Type=*/FuncTy_12, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"sinf", mod); // (external, no body) -func_sinf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_sinf_PAL = 0; -func_sinf->setParamAttrs(func_sinf_PAL); - -Function* func_vsin = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"vsin", mod); -func_vsin->setCallingConv(CallingConv::C); -const ParamAttrsList *func_vsin_PAL = 0; -func_vsin->setParamAttrs(func_vsin_PAL); - -Function* func_kilp = new Function( - /*Type=*/FuncTy_13, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"kilp", mod); -func_kilp->setCallingConv(CallingConv::C); -const ParamAttrsList *func_kilp_PAL = 0; -{ - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_kilp_PAL = ParamAttrsList::get(Attrs); - -} -func_kilp->setParamAttrs(func_kilp_PAL); - -// Global Variable Declarations - - -GlobalVariable* gvar_array__str = new GlobalVariable( -/*Type=*/ArrayTy_0, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str", -mod); - -GlobalVariable* gvar_array__str1 = new GlobalVariable( -/*Type=*/ArrayTy_0, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str1", -mod); - -// Constant Definitions -Constant* const_array_14 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); -Constant* const_array_15 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); -ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); -ConstantFP* const_float_17 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); -Constant* const_float_18 = Constant::getNullValue(Type::FloatTy); -Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); -std::vector const_packed_20_elems; -ConstantFP* const_float_21 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); -const_packed_20_elems.push_back(const_float_21); -UndefValue* const_float_22 = UndefValue::get(Type::FloatTy); -const_packed_20_elems.push_back(const_float_22); -const_packed_20_elems.push_back(const_float_22); -const_packed_20_elems.push_back(const_float_21); -Constant* const_packed_20 = ConstantVector::get(VectorTy_4, const_packed_20_elems); -ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); -ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "3", 10)); -ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "2", 10)); -std::vector const_packed_26_elems; -const_packed_26_elems.push_back(const_float_21); -const_packed_26_elems.push_back(const_float_18); -const_packed_26_elems.push_back(const_float_18); -const_packed_26_elems.push_back(const_float_21); -Constant* const_packed_26 = ConstantVector::get(VectorTy_4, const_packed_26_elems); -Constant* const_double_27 = Constant::getNullValue(Type::DoubleTy); -std::vector const_packed_28_elems; -const_packed_28_elems.push_back(const_int32_19); -ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "5", 10)); -const_packed_28_elems.push_back(const_int32_29); -const_packed_28_elems.push_back(const_int32_25); -const_packed_28_elems.push_back(const_int32_24); -Constant* const_packed_28 = ConstantVector::get(VectorTy_7, const_packed_28_elems); -std::vector const_packed_30_elems; -const_packed_30_elems.push_back(const_int32_19); -const_packed_30_elems.push_back(const_int32_23); -ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "6", 10)); -const_packed_30_elems.push_back(const_int32_31); -const_packed_30_elems.push_back(const_int32_24); -Constant* const_packed_30 = ConstantVector::get(VectorTy_7, const_packed_30_elems); -std::vector const_packed_32_elems; -const_packed_32_elems.push_back(const_int32_19); -const_packed_32_elems.push_back(const_int32_23); -const_packed_32_elems.push_back(const_int32_25); -ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "7", 10)); -const_packed_32_elems.push_back(const_int32_33); -Constant* const_packed_32 = ConstantVector::get(VectorTy_7, const_packed_32_elems); -std::vector const_ptr_34_indices; -const_ptr_34_indices.push_back(const_int32_19); -const_ptr_34_indices.push_back(const_int32_19); -Constant* const_ptr_34 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_34_indices[0], const_ptr_34_indices.size() ); -UndefValue* const_packed_35 = UndefValue::get(VectorTy_4); -std::vector const_ptr_36_indices; -const_ptr_36_indices.push_back(const_int32_19); -const_ptr_36_indices.push_back(const_int32_19); -Constant* const_ptr_36 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_36_indices[0], const_ptr_36_indices.size() ); - -// Global Variable Definitions -gvar_array__str->setInitializer(const_array_14); -gvar_array__str1->setInitializer(const_array_15); - -// Function Definitions - -// Function: approx (func_approx) -{ - Function::arg_iterator args = func_approx->arg_begin(); - Value* float_a = args++; - float_a->setName("a"); - Value* float_b = args++; - float_b->setName("b"); - - BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); - - // Block entry (label_entry) - FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_16, "cmp", label_entry); - SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_16, float_b, "b.addr.0", label_entry); - FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_17, "cmp3", label_entry); - SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_17, float_b_addr_0, "b.addr.1", label_entry); - FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_18, "cmp7", label_entry); - SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_18, float_a, "a.addr.0", label_entry); - std::vector float_call_params; - float_call_params.push_back(float_a_addr_0); - float_call_params.push_back(float_b_addr_1); - CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); - float_call->setCallingConv(CallingConv::C); - float_call->setTailCall(true);const ParamAttrsList *float_call_PAL = 0; - float_call->setParamAttrs(float_call_PAL); - - new ReturnInst(float_call, label_entry); - -} - -// Function: lit (func_lit) -{ - Function::arg_iterator args = func_lit->arg_begin(); - Value* packed_tmp = args++; - packed_tmp->setName("tmp"); - - BasicBlock* label_entry_38 = new BasicBlock("entry",func_lit,0); - BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); - BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); - - // Block entry (label_entry_38) - ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp6", label_entry_38); - FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp6, const_float_18, "cmp", label_entry_38); - new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_39, label_entry_38); - - // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp10 = new InsertElementInst(const_packed_20, float_tmp6, const_int32_23, "tmp10", label_ifthen); - ExtractElementInst* float_tmp12 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp14", label_ifthen); - std::vector float_call_41_params; - float_call_41_params.push_back(float_tmp12); - float_call_41_params.push_back(float_tmp14); - CallInst* float_call_41 = new CallInst(func_approx, float_call_41_params.begin(), float_call_41_params.end(), "call", label_ifthen); - float_call_41->setCallingConv(CallingConv::C); - float_call_41->setTailCall(true);const ParamAttrsList *float_call_41_PAL = 0; - float_call_41->setParamAttrs(float_call_41_PAL); - - InsertElementInst* packed_tmp16 = new InsertElementInst(packed_tmp10, float_call_41, const_int32_25, "tmp16", label_ifthen); - new ReturnInst(packed_tmp16, label_ifthen); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock) - new ReturnInst(const_packed_26, label_UnifiedReturnBlock); - -} - -// Function: cmp (func_cmp) -{ - Function::arg_iterator args = func_cmp->arg_begin(); - Value* packed_tmp0 = args++; - packed_tmp0->setName("tmp0"); - Value* packed_tmp1 = args++; - packed_tmp1->setName("tmp1"); - Value* packed_tmp2 = args++; - packed_tmp2->setName("tmp2"); - - BasicBlock* label_entry_44 = new BasicBlock("entry",func_cmp,0); - BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); - BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); - BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); - BasicBlock* label_cond_cont34 = new BasicBlock("cond.cont34",func_cmp,0); - BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); - BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); - - // Block entry (label_entry_44) - ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_44); - CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_44); - FCmpInst* int1_cmp_45 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_44); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_44); - CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_44); - FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_44); - SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_45, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_44); - new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_44); - - // Block cond.?14 (label_cond__14) - ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_28, "tmp233", label_cond__14); - ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp254", label_cond__14); - CastInst* double_conv265 = new FPExtInst(float_tmp254, Type::DoubleTy, "conv265", label_cond__14); - FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_27, "cmp276", label_cond__14); - new BranchInst(label_cond__28, label_cond_cont34, int1_cmp276, label_cond__14); - - // Block cond.cont20 (label_cond_cont20) - ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_28, "tmp23", label_cond_cont20); - ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp25", label_cond_cont20); - CastInst* double_conv26 = new FPExtInst(float_tmp25, Type::DoubleTy, "conv26", label_cond_cont20); - FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_27, "cmp27", label_cond_cont20); - new BranchInst(label_cond__28, label_cond_cont34, int1_cmp27, label_cond_cont20); - - // Block cond.?28 (label_cond__28) - PHINode* packed_tmp23_reg2mem_0 = new PHINode(VectorTy_4, "tmp23.reg2mem.0", label_cond__28); - packed_tmp23_reg2mem_0->reserveOperandSpace(2); - packed_tmp23_reg2mem_0->addIncoming(packed_tmp233, label_cond__14); - packed_tmp23_reg2mem_0->addIncoming(packed_tmp23, label_cond_cont20); - - ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_30, "tmp378", label_cond__28); - ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp399", label_cond__28); - CastInst* double_conv4010 = new FPExtInst(float_tmp399, Type::DoubleTy, "conv4010", label_cond__28); - FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_27, "cmp4111", label_cond__28); - new BranchInst(label_cond__42, label_cond_cont48, int1_cmp4111, label_cond__28); - - // Block cond.cont34 (label_cond_cont34) - PHINode* packed_tmp23_reg2mem_1 = new PHINode(VectorTy_4, "tmp23.reg2mem.1", label_cond_cont34); - packed_tmp23_reg2mem_1->reserveOperandSpace(2); - packed_tmp23_reg2mem_1->addIncoming(packed_tmp233, label_cond__14); - packed_tmp23_reg2mem_1->addIncoming(packed_tmp23, label_cond_cont20); - - ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_30, "tmp37", label_cond_cont34); - ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp39", label_cond_cont34); - CastInst* double_conv40 = new FPExtInst(float_tmp39, Type::DoubleTy, "conv40", label_cond_cont34); - FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_27, "cmp41", label_cond_cont34); - new BranchInst(label_cond__42, label_cond_cont48, int1_cmp41, label_cond_cont34); - - // Block cond.?42 (label_cond__42) - PHINode* packed_tmp37_reg2mem_0 = new PHINode(VectorTy_4, "tmp37.reg2mem.0", label_cond__42); - packed_tmp37_reg2mem_0->reserveOperandSpace(2); - packed_tmp37_reg2mem_0->addIncoming(packed_tmp378, label_cond__28); - packed_tmp37_reg2mem_0->addIncoming(packed_tmp37, label_cond_cont34); - - ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_32, "tmp5113", label_cond__42); - new ReturnInst(packed_tmp5113, label_cond__42); - - // Block cond.cont48 (label_cond_cont48) - PHINode* packed_tmp37_reg2mem_1 = new PHINode(VectorTy_4, "tmp37.reg2mem.1", label_cond_cont48); - packed_tmp37_reg2mem_1->reserveOperandSpace(2); - packed_tmp37_reg2mem_1->addIncoming(packed_tmp378, label_cond__28); - packed_tmp37_reg2mem_1->addIncoming(packed_tmp37, label_cond_cont34); - - ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_32, "tmp51", label_cond_cont48); - new ReturnInst(packed_tmp51, label_cond_cont48); - -} - -// Function: vcos (func_vcos) -{ - Function::arg_iterator args = func_vcos->arg_begin(); - Value* packed_val = args++; - packed_val->setName("val"); - - BasicBlock* label_entry_53 = new BasicBlock("entry",func_vcos,0); - - // Block entry (label_entry_53) - ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_53); - CastInst* double_conv_54 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_53); - ExtractElementInst* float_tmp3_55 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_53); - CastInst* double_conv4 = new FPExtInst(float_tmp3_55, Type::DoubleTy, "conv4", label_entry_53); - ExtractElementInst* float_tmp6_56 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_53); - CastInst* double_conv7 = new FPExtInst(float_tmp6_56, Type::DoubleTy, "conv7", label_entry_53); - ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_53); - CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_53); - std::vector int32_call_params; - int32_call_params.push_back(const_ptr_34); - int32_call_params.push_back(double_conv_54); - int32_call_params.push_back(double_conv4); - int32_call_params.push_back(double_conv7); - int32_call_params.push_back(double_conv10); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_53); - int32_call->setCallingConv(CallingConv::C); - int32_call->setTailCall(true);const ParamAttrsList *int32_call_PAL = 0; - int32_call->setParamAttrs(int32_call_PAL); - - CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_53); - float_call13->setCallingConv(CallingConv::C); - float_call13->setTailCall(true);const ParamAttrsList *float_call13_PAL = 0; - float_call13->setParamAttrs(float_call13_PAL); - - InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_53); - CallInst* float_call18 = new CallInst(func_cosf, float_tmp1, "call18", label_entry_53); - float_call18->setCallingConv(CallingConv::C); - float_call18->setTailCall(true);const ParamAttrsList *float_call18_PAL = 0; - float_call18->setParamAttrs(float_call18_PAL); - - InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call18, const_int32_23, "tmp20", label_entry_53); - CallInst* float_call23 = new CallInst(func_cosf, float_tmp1, "call23", label_entry_53); - float_call23->setCallingConv(CallingConv::C); - float_call23->setTailCall(true);const ParamAttrsList *float_call23_PAL = 0; - float_call23->setParamAttrs(float_call23_PAL); - - InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call23, const_int32_25, "tmp25", label_entry_53); - CallInst* float_call28 = new CallInst(func_cosf, float_tmp1, "call28", label_entry_53); - float_call28->setCallingConv(CallingConv::C); - float_call28->setTailCall(true);const ParamAttrsList *float_call28_PAL = 0; - float_call28->setParamAttrs(float_call28_PAL); - - InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call28, const_int32_24, "tmp30", label_entry_53); - CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_53); - CastInst* double_conv36 = new FPExtInst(float_call18, Type::DoubleTy, "conv36", label_entry_53); - CastInst* double_conv39 = new FPExtInst(float_call23, Type::DoubleTy, "conv39", label_entry_53); - CastInst* double_conv42 = new FPExtInst(float_call28, Type::DoubleTy, "conv42", label_entry_53); - std::vector int32_call43_params; - int32_call43_params.push_back(const_ptr_36); - int32_call43_params.push_back(double_conv33); - int32_call43_params.push_back(double_conv36); - int32_call43_params.push_back(double_conv39); - int32_call43_params.push_back(double_conv42); - CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_53); - int32_call43->setCallingConv(CallingConv::C); - int32_call43->setTailCall(true);const ParamAttrsList *int32_call43_PAL = 0; - int32_call43->setParamAttrs(int32_call43_PAL); - - new ReturnInst(packed_tmp30, label_entry_53); - -} - -// Function: scs (func_scs) -{ - Function::arg_iterator args = func_scs->arg_begin(); - Value* packed_val_58 = args++; - packed_val_58->setName("val"); - - BasicBlock* label_entry_59 = new BasicBlock("entry",func_scs,0); - - // Block entry (label_entry_59) - ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_58, const_int32_19, "tmp2", label_entry_59); - CallInst* float_call_60 = new CallInst(func_cosf, float_tmp2, "call", label_entry_59); - float_call_60->setCallingConv(CallingConv::C); - float_call_60->setTailCall(true);const ParamAttrsList *float_call_60_PAL = 0; - float_call_60->setParamAttrs(float_call_60_PAL); - - InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_60, const_int32_19, "tmp5", label_entry_59); - CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_59); - float_call7->setCallingConv(CallingConv::C); - float_call7->setTailCall(true);const ParamAttrsList *float_call7_PAL = 0; - float_call7->setParamAttrs(float_call7_PAL); - - InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_59); - new ReturnInst(packed_tmp9, label_entry_59); - -} - -// Function: vsin (func_vsin) -{ - Function::arg_iterator args = func_vsin->arg_begin(); - Value* packed_val_62 = args++; - packed_val_62->setName("val"); - - BasicBlock* label_entry_63 = new BasicBlock("entry",func_vsin,0); - - // Block entry (label_entry_63) - ExtractElementInst* float_tmp2_64 = new ExtractElementInst(packed_val_62, const_int32_19, "tmp2", label_entry_63); - CallInst* float_call_65 = new CallInst(func_sinf, float_tmp2_64, "call", label_entry_63); - float_call_65->setCallingConv(CallingConv::C); - float_call_65->setTailCall(true);const ParamAttrsList *float_call_65_PAL = 0; - float_call_65->setParamAttrs(float_call_65_PAL); - - InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_65, const_int32_19, "tmp6", label_entry_63); - InsertElementInst* packed_tmp9_66 = new InsertElementInst(packed_tmp6, float_call_65, const_int32_23, "tmp9", label_entry_63); - InsertElementInst* packed_tmp12 = new InsertElementInst(packed_tmp9_66, float_call_65, const_int32_25, "tmp12", label_entry_63); - InsertElementInst* packed_tmp15_67 = new InsertElementInst(packed_tmp12, float_call_65, const_int32_24, "tmp15", label_entry_63); - new ReturnInst(packed_tmp15_67, label_entry_63); - -} - -// Function: kilp (func_kilp) -{ - Function::arg_iterator args = func_kilp->arg_begin(); - Value* packed_val_69 = args++; - packed_val_69->setName("val"); - - BasicBlock* label_entry_70 = new BasicBlock("entry",func_kilp,0); - BasicBlock* label_lor_rhs = new BasicBlock("lor_rhs",func_kilp,0); - BasicBlock* label_lor_rhs5 = new BasicBlock("lor_rhs5",func_kilp,0); - BasicBlock* label_lor_rhs11 = new BasicBlock("lor_rhs11",func_kilp,0); - BasicBlock* label_UnifiedReturnBlock_71 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); - - // Block entry (label_entry_70) - ExtractElementInst* float_tmp1_72 = new ExtractElementInst(packed_val_69, const_int32_19, "tmp1", label_entry_70); - FCmpInst* int1_cmp_73 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_72, const_float_18, "cmp", label_entry_70); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs, int1_cmp_73, label_entry_70); - - // Block lor_rhs (label_lor_rhs) - ExtractElementInst* float_tmp3_75 = new ExtractElementInst(packed_val_69, const_int32_23, "tmp3", label_lor_rhs); - FCmpInst* int1_cmp4 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_75, const_float_18, "cmp4", label_lor_rhs); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs5, int1_cmp4, label_lor_rhs); - - // Block lor_rhs5 (label_lor_rhs5) - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_val_69, const_int32_25, "tmp7", label_lor_rhs5); - FCmpInst* int1_cmp8 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp7, const_float_18, "cmp8", label_lor_rhs5); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs11, int1_cmp8, label_lor_rhs5); - - // Block lor_rhs11 (label_lor_rhs11) - ExtractElementInst* float_tmp13 = new ExtractElementInst(packed_val_69, const_int32_24, "tmp13", label_lor_rhs11); - FCmpInst* int1_cmp14 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp13, const_float_18, "cmp14", label_lor_rhs11); - CastInst* int32_retval = new ZExtInst(int1_cmp14, IntegerType::get(32), "retval", label_lor_rhs11); - new ReturnInst(int32_retval, label_lor_rhs11); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock_71) - new ReturnInst(const_int32_23, label_UnifiedReturnBlock_71); - -} - -return mod; - -} diff --git a/src/mesa/pipe/llvm/gallivm_cpu.cpp b/src/mesa/pipe/llvm/gallivm_cpu.cpp deleted file mode 100644 index dc4d92a72a..0000000000 --- a/src/mesa/pipe/llvm/gallivm_cpu.cpp +++ /dev/null @@ -1,202 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "gallivm.h" -#include "gallivm_p.h" - -#include "instructions.h" -#include "loweringpass.h" -#include "storage.h" -#include "tgsitollvm.h" - -#include "pipe/p_context.h" -#include "pipe/p_shader_tokens.h" - -#include "pipe/tgsi/exec/tgsi_exec.h" -#include "pipe/tgsi/util/tgsi_dump.h" - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -struct gallivm_cpu_engine { - llvm::ExecutionEngine *engine; -}; - -static struct gallivm_cpu_engine *CPU = 0; - -typedef int (*fragment_shader_runner)(float x, float y, - float (*dests)[16][4], - float (*inputs)[16][4], - int num_attribs, - float (*consts)[4], int num_consts, - struct tgsi_sampler *samplers); - -int gallivm_cpu_fs_exec(struct gallivm_prog *prog, - float fx, float fy, - float (*dests)[16][4], - float (*inputs)[16][4], - float (*consts)[4], - struct tgsi_sampler *samplers) -{ - fragment_shader_runner runner = reinterpret_cast(prog->function); - assert(runner); - - return runner(fx, fy, dests, inputs, prog->num_interp, - consts, prog->num_consts, - samplers); -} - -static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) -{ - llvm::Module *mod = prog->module; - llvm::Function *func = 0; - - switch (prog->type) { - case GALLIVM_VS: - func = mod->getFunction("vs_shader"); - break; - case GALLIVM_FS: - func = mod->getFunction("fs_shader"); - break; - default: - assert(!"Unknown shader type!"); - break; - } - return func; -} - -/*! - This function creates a CPU based execution engine for the given gallivm_prog. - gallivm_cpu_engine should be used as a singleton throughout the library. Before - executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile. - The gallivm_prog instance which is being passed to the constructor is being - automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile - with it again. - */ -struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) -{ - struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *) - calloc(1, sizeof(struct gallivm_cpu_engine)); - llvm::Module *mod = static_cast(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); - ee->DisableLazyCompilation(); - cpu->engine = ee; - - llvm::Function *func = func_for_shader(prog); - - prog->function = ee->getPointerToFunction(func); - CPU = cpu; - return cpu; -} - - -/*! - This function JIT compiles the given gallivm_prog with the given cpu based execution engine. - The reference to the generated machine code entry point will be stored - in the gallivm_prog program. After executing this function one can call gallivm_prog_exec - in order to execute the gallivm_prog on the CPU. - */ -void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) -{ - llvm::Module *mod = static_cast(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = cpu->engine; - assert(ee); - /*FIXME : remove */ - ee->DisableLazyCompilation(); - ee->addModuleProvider(mp); - - llvm::Function *func = func_for_shader(prog); - prog->function = ee->getPointerToFunction(func); -} - -void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) -{ - free(cpu); -} - -struct gallivm_cpu_engine * gallivm_global_cpu_engine() -{ - return CPU; -} - - -typedef void (*vertex_shader_runner)(void *ainputs, - void *dests, - float (*aconsts)[4], - void *temps); - - -/*! - This function is used to execute the gallivm_prog in software. Before calling - this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile - function. - */ -int gallivm_cpu_vs_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - struct tgsi_exec_vector *temps) -{ - vertex_shader_runner runner = reinterpret_cast(prog->function); - assert(runner); - /*FIXME*/ - runner(inputs, dests, consts, temps); - - return 0; -} - -#endif diff --git a/src/mesa/pipe/llvm/gallivm_p.h b/src/mesa/pipe/llvm/gallivm_p.h deleted file mode 100644 index cfe7b1901b..0000000000 --- a/src/mesa/pipe/llvm/gallivm_p.h +++ /dev/null @@ -1,110 +0,0 @@ -#ifndef GALLIVM_P_H -#define GALLIVM_P_H - -#ifdef MESA_LLVM - -#include "gallivm.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/p_compiler.h" - -namespace llvm { - class Module; -} - -#if defined __cplusplus -extern "C" { -#endif - -enum gallivm_shader_type; -enum gallivm_vector_layout; - -struct gallivm_interpolate { - int attrib; - int chan; - int type; -}; - -struct gallivm_ir { - llvm::Module *module; - int id; - enum gallivm_shader_type type; - enum gallivm_vector_layout layout; - int num_components; - int num_consts; - - //FIXME: this might not be enough for some shaders - struct gallivm_interpolate interpolators[32*4]; - int num_interp; -}; - -struct gallivm_prog { - llvm::Module *module; - void *function; - - int id; - enum gallivm_shader_type type; - - int num_consts; - - //FIXME: this might not be enough for some shaders - struct gallivm_interpolate interpolators[32*4]; - int num_interp; -}; - -static INLINE void gallivm_swizzle_components(int swizzle, - int *xc, int *yc, - int *zc, int *wc) -{ - int x = swizzle / 1000; swizzle -= x * 1000; - int y = swizzle / 100; swizzle -= y * 100; - int z = swizzle / 10; swizzle -= z * 10; - int w = swizzle; - - if (xc) *xc = x; - if (yc) *yc = y; - if (zc) *zc = z; - if (wc) *wc = w; -} - -static INLINE boolean gallivm_is_swizzle(int swizzle) -{ - const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + - TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; - return swizzle != NO_SWIZZLE; -} - -static INLINE int gallivm_x_swizzle(int swizzle) -{ - int x; - gallivm_swizzle_components(swizzle, &x, 0, 0, 0); - return x; -} - -static INLINE int gallivm_y_swizzle(int swizzle) -{ - int y; - gallivm_swizzle_components(swizzle, 0, &y, 0, 0); - return y; -} - -static INLINE int gallivm_z_swizzle(int swizzle) -{ - int z; - gallivm_swizzle_components(swizzle, 0, 0, &z, 0); - return z; -} - -static INLINE int gallivm_w_swizzle(int swizzle) -{ - int w; - gallivm_swizzle_components(swizzle, 0, 0, 0, &w); - return w; -} - -#endif /* MESA_LLVM */ - -#if defined __cplusplus -} // extern "C" -#endif - -#endif diff --git a/src/mesa/pipe/llvm/instructions.cpp b/src/mesa/pipe/llvm/instructions.cpp deleted file mode 100644 index 55d39fa5f1..0000000000 --- a/src/mesa/pipe/llvm/instructions.cpp +++ /dev/null @@ -1,889 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "instructions.h" - -#include "storage.h" - -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -using namespace llvm; - -#include "gallivm_builtins.cpp" - -static inline std::string createFuncName(int label) -{ - std::ostringstream stream; - stream << "function"; - stream << label; - return stream.str(); -} - -Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, - Storage *storage) - : m_mod(mod), m_func(func), m_builder(block), m_idx(0), - m_storage(storage) -{ - m_floatVecType = VectorType::get(Type::FloatTy, 4); - - m_llvmFSqrt = 0; - m_llvmFAbs = 0; - m_llvmPow = 0; - m_llvmFloor = 0; - m_llvmFlog = 0; - m_llvmLit = 0; - m_fmtPtr = 0; - - createGallivmBuiltins(m_mod); -} - -llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) -{ - return m_builder.CreateAdd(in1, in2, name("add")); -} - -llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in3) -{ - Value *mulRes = mul(in1, in2); - return add(mulRes, in3); -} - -llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2) -{ - return m_builder.CreateMul(in1, in2, name("mul")); -} - -const char * Instructions::name(const char *prefix) -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2) -{ - Value *mulRes = mul(in1, in2); - Value *x = m_builder.CreateExtractElement(mulRes, - m_storage->constantInt(0), - name("extractx")); - Value *y = m_builder.CreateExtractElement(mulRes, - m_storage->constantInt(1), - name("extracty")); - Value *z = m_builder.CreateExtractElement(mulRes, - m_storage->constantInt(2), - name("extractz")); - Value *xy = m_builder.CreateAdd(x, y,name("xy")); - Value *dot3 = m_builder.CreateAdd(xy, z, name("dot3")); - return vectorFromVals(dot3, dot3, dot3, dot3); -} - -llvm::Value *Instructions::callFSqrt(llvm::Value *val) -{ - if (!m_llvmFSqrt) { - // predeclare the intrinsic - std::vector fsqrtArgs; - fsqrtArgs.push_back(Type::FloatTy); - ParamAttrsList *fsqrtPal = 0; - FunctionType* fsqrtType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/fsqrtArgs, - /*isVarArg=*/false); - m_llvmFSqrt = new Function( - /*Type=*/fsqrtType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"llvm.sqrt.f32", m_mod); - m_llvmFSqrt->setCallingConv(CallingConv::C); - m_llvmFSqrt->setParamAttrs(fsqrtPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFSqrt, val, - name("sqrt")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::rsq(llvm::Value *in1) -{ - Value *x = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("extractx")); - Value *abs = callFAbs(x); - Value *sqrt = callFSqrt(abs); - - Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - sqrt, - name("rsqrt")); - return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt); -} - -llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w) -{ - Constant *const_vec = Constant::getNullValue(m_floatVecType); - Value *res = m_builder.CreateInsertElement(const_vec, x, - m_storage->constantInt(0), - name("vecx")); - res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), - name("vecxy")); - res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), - name("vecxyz")); - if (w) - res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), - name("vecxyzw")); - return res; -} - -llvm::Value *Instructions::callFAbs(llvm::Value *val) -{ - if (!m_llvmFAbs) { - // predeclare the intrinsic - std::vector fabsArgs; - fabsArgs.push_back(Type::FloatTy); - ParamAttrsList *fabsPal = 0; - FunctionType* fabsType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/fabsArgs, - /*isVarArg=*/false); - m_llvmFAbs = new Function( - /*Type=*/fabsType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"fabs", m_mod); - m_llvmFAbs->setCallingConv(CallingConv::C); - m_llvmFAbs->setParamAttrs(fabsPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFAbs, val, - name("fabs")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::lit(llvm::Value *in) -{ - if (!m_llvmLit) { - m_llvmLit = m_mod->getFunction("lit"); - } - CallInst *call = m_builder.CreateCall(m_llvmLit, in, name("litres")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) -{ - Value *res = m_builder.CreateSub(in1, in2, name("sub")); - return res; -} - -llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) -{ - if (!m_llvmPow) { - // predeclare the intrinsic - std::vector powArgs; - powArgs.push_back(Type::FloatTy); - powArgs.push_back(Type::FloatTy); - ParamAttrsList *powPal = 0; - FunctionType* powType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/powArgs, - /*isVarArg=*/false); - m_llvmPow = new Function( - /*Type=*/powType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"llvm.pow.f32", m_mod); - m_llvmPow->setCallingConv(CallingConv::C); - m_llvmPow->setParamAttrs(powPal); - } - std::vector params; - params.push_back(val1); - params.push_back(val2); - CallInst *call = m_builder.CreateCall(m_llvmPow, params.begin(), params.end(), - name("pow")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2) -{ - Value *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - Value *x2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(0), - name("x2")); - llvm::Value *val = callPow(x1, x2); - return vectorFromVals(val, val, val, val); -} - -llvm::Value * Instructions::rcp(llvm::Value *in1) -{ - Value *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - Value *res = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - x1, name("rcp")); - return vectorFromVals(res, res, res, res); -} - -llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) -{ - Value *mulRes = mul(in1, in2); - std::vector vec = extractVector(mulRes); - Value *xy = m_builder.CreateAdd(vec[0], vec[1], name("xy")); - Value *xyz = m_builder.CreateAdd(xy, vec[2], name("xyz")); - Value *dot4 = m_builder.CreateAdd(xyz, vec[3], name("dot4")); - return vectorFromVals(dot4, dot4, dot4, dot4); -} - -llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) -{ - Value *mulRes = mul(in1, in2); - std::vector vec1 = extractVector(mulRes); - Value *xy = m_builder.CreateAdd(vec1[0], vec1[1], name("xy")); - Value *xyz = m_builder.CreateAdd(xy, vec1[2], name("xyz")); - Value *dph = m_builder.CreateAdd(xyz, vec1[3], name("dph")); - return vectorFromVals(dph, dph, dph, dph); -} - -llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) -{ - Value *y1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(1), - name("y1")); - Value *z = m_builder.CreateExtractElement(in1, - m_storage->constantInt(2), - name("z")); - Value *y2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(1), - name("y2")); - Value *w = m_builder.CreateExtractElement(in2, - m_storage->constantInt(3), - name("w")); - Value *ry = m_builder.CreateMul(y1, y2, name("tyuy")); - return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), - ry, z, w); -} - -llvm::Value * Instructions::ex2(llvm::Value *in) -{ - llvm::Value *val = callPow(ConstantFP::get(Type::FloatTy, APFloat(2.f)), - m_builder.CreateExtractElement( - in, m_storage->constantInt(0), - name("x1"))); - return vectorFromVals(val, val, val, val); -} - -llvm::Value * Instructions::callFloor(llvm::Value *val) -{ - if (!m_llvmFloor) { - // predeclare the intrinsic - std::vector floorArgs; - floorArgs.push_back(Type::FloatTy); - ParamAttrsList *floorPal = 0; - FunctionType* floorType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/floorArgs, - /*isVarArg=*/false); - m_llvmFloor = new Function( - /*Type=*/floorType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"floorf", m_mod); - m_llvmFloor->setCallingConv(CallingConv::C); - m_llvmFloor->setParamAttrs(floorPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFloor, val, - name("floorf")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::floor(llvm::Value *in) -{ - std::vector vec = extractVector(in); - return vectorFromVals(callFloor(vec[0]), callFloor(vec[1]), - callFloor(vec[2]), callFloor(vec[3])); -} - -llvm::Value * Instructions::arl(llvm::Value *in) -{ - return floor(in); -} - -llvm::Value * Instructions::frc(llvm::Value *in) -{ - llvm::Value *flr = floor(in); - return sub(in, flr); -} - -llvm::Value * Instructions::callFLog(llvm::Value *val) -{ - if (!m_llvmFlog) { - // predeclare the intrinsic - std::vector flogArgs; - flogArgs.push_back(Type::FloatTy); - ParamAttrsList *flogPal = 0; - FunctionType* flogType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/flogArgs, - /*isVarArg=*/false); - m_llvmFlog = new Function( - /*Type=*/flogType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"logf", m_mod); - m_llvmFlog->setCallingConv(CallingConv::C); - m_llvmFlog->setParamAttrs(flogPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFlog, val, - name("logf")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::lg2(llvm::Value *in) -{ - std::vector vec = extractVector(in); - llvm::Value *const_vec = constVector(1.442695f, 1.442695f, - 1.442695f, 1.442695f); - return mul(vectorFromVals(callFLog(vec[0]), callFLog(vec[1]), - callFLog(vec[2]), callFLog(vec[3])), const_vec); -} - -llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) -{ - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); - Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], - name("selx")); - - Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); - Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], - name("sely")); - - Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); - Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], - name("selz")); - - Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); - Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], - name("selw")); - - return vectorFromVals(selx, sely, selz, selw); -} - -llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) -{ - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], - name("xcmp")); - Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], - name("selx")); - - Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], - name("ycmp")); - Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], - name("sely")); - - Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], - name("zcmp")); - Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], - name("selz")); - - Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], - name("wcmp")); - Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], - name("selw")); - - return vectorFromVals(selx, sely, selz, selw); -} - -void Instructions::printVector(llvm::Value *val) -{ - static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A"; - - if (!m_fmtPtr) { - Constant *format = ConstantArray::get(frmt, true); - ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1); - GlobalVariable* globalFormat = new GlobalVariable( - /*Type=*/arrayTy, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name(".str"), - m_mod); - globalFormat->setInitializer(format); - - Constant* const_int0 = Constant::getNullValue(IntegerType::get(32)); - std::vector const_ptr_21_indices; - const_ptr_21_indices.push_back(const_int0); - const_ptr_21_indices.push_back(const_int0); - m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat, - &const_ptr_21_indices[0], const_ptr_21_indices.size()); - } - - Function *func_printf = m_mod->getFunction("printf"); - if (!func_printf) - func_printf = declarePrintf(); - assert(func_printf); - std::vector vec = extractVector(val); - Value *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx")); - Value *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy")); - Value *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz")); - Value *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw")); - std::vector params; - params.push_back(m_fmtPtr); - params.push_back(dx); - params.push_back(dy); - params.push_back(dz); - params.push_back(dw); - CallInst *call = m_builder.CreateCall(func_printf, params.begin(), params.end(), - name("printf")); - call->setCallingConv(CallingConv::C); - call->setTailCall(true); -} - -llvm::Function * Instructions::declarePrintf() -{ - std::vector args; - ParamAttrsList *params = 0; - FunctionType* funcTy = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/args, - /*isVarArg=*/true); - Function* func_printf = new Function( - /*Type=*/funcTy, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", m_mod); - func_printf->setCallingConv(CallingConv::C); - func_printf->setParamAttrs(params); - return func_printf; -} - - -llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) -{ - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - Constant *const0f = Constant::getNullValue(Type::FloatTy); - - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], name("xcmp")); - Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - - Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], name("ycmp")); - Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - - Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], name("zcmp")); - Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - - Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], name("wcmp")); - Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); - - return vectorFromVals(x, y, z, w); -} -llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) -{ - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - Constant *const0f = Constant::getNullValue(Type::FloatTy); - - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOGE(vec1[0], vec2[0], name("xcmp")); - Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - - Value *ycmp = m_builder.CreateFCmpOGE(vec1[1], vec2[1], name("ycmp")); - Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - - Value *zcmp = m_builder.CreateFCmpOGE(vec1[2], vec2[2], name("zcmp")); - Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - - Value *wcmp = m_builder.CreateFCmpOGE(vec1[3], vec2[3], name("wcmp")); - Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); - - return vectorFromVals(x, y, z, w); -} - - -llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) -{ - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - Constant *const0f = Constant::getNullValue(Type::FloatTy); - - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); - Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - - Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); - Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - - Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); - Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - - Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); - Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); - - return vectorFromVals(x, y, z, w); -} - -llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) -{ - Value *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - Value *y1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(1), - name("y1")); - Value *z1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(2), - name("z1")); - - Value *x2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(0), - name("x2")); - Value *y2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(1), - name("y2")); - Value *z2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(2), - name("z2")); - Value *y1z2 = mul(y1, z2); - Value *z1y2 = mul(z1, y2); - - Value *z1x2 = mul(z1, x2); - Value *x1z2 = mul(x1, z2); - - Value *x1y2 = mul(x1, y2); - Value *y1x2 = mul(y1, x2); - - return vectorFromVals(sub(y1z2, z1y2), sub(z1x2, x1z2), sub(x1y2, y1x2)); -} - - -llvm::Value * Instructions::abs(llvm::Value *in) -{ - std::vector vec = extractVector(in); - Value *xabs = callFAbs(vec[0]); - Value *yabs = callFAbs(vec[1]); - Value *zabs = callFAbs(vec[2]); - Value *wabs = callFAbs(vec[3]); - return vectorFromVals(xabs, yabs, zabs, wabs); -} - -void Instructions::ifop(llvm::Value *in) -{ - BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); - BasicBlock *ifend = new BasicBlock(name("ifthenend"), m_func,0); - - //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0); - //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0); - //BasicBlock *wblock = new BasicBlock(name("wblock"), m_func,0); - - Constant *float0 = Constant::getNullValue(Type::FloatTy); - - Value *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0), - name("extractx")); - Value *xcmp = m_builder.CreateFCmpUNE(x, float0, name("xcmp")); - m_builder.CreateCondBr(xcmp, ifthen, ifend); - //m_builder.SetInsertPoint(yblock); - - m_builder.SetInsertPoint(ifthen); - m_ifStack.push(ifend); -} - -llvm::BasicBlock * Instructions::currentBlock() const -{ - return m_builder.GetInsertBlock(); -} - -void Instructions::elseop() -{ - assert(!m_ifStack.empty()); - BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); - m_builder.CreateBr(ifend); - m_builder.SetInsertPoint(m_ifStack.top()); - currentBlock()->setName(name("ifelse")); - m_ifStack.pop(); - m_ifStack.push(ifend); -} - -void Instructions::endif() -{ - assert(!m_ifStack.empty()); - m_builder.CreateBr(m_ifStack.top()); - m_builder.SetInsertPoint(m_ifStack.top()); - m_ifStack.pop(); -} - -llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in3) -{ - llvm::Value *m = mul(in1, in2); - llvm::Value *vec1 = constVector(1.f, 1.f, 1.f, 1.f); - llvm::Value *s = sub(vec1, in1); - return add(m, mul(s, in3)); -} - -void Instructions::beginLoop() -{ - BasicBlock *begin = new BasicBlock(name("loop"), m_func,0); - BasicBlock *end = new BasicBlock(name("endloop"), m_func,0); - - m_builder.CreateBr(begin); - Loop loop; - loop.begin = begin; - loop.end = end; - m_builder.SetInsertPoint(begin); - m_loopStack.push(loop); -} - -void Instructions::endLoop() -{ - assert(!m_loopStack.empty()); - Loop loop = m_loopStack.top(); - m_builder.CreateBr(loop.begin); - loop.end->moveAfter(currentBlock()); - m_builder.SetInsertPoint(loop.end); - m_loopStack.pop(); -} - -void Instructions::brk() -{ - assert(!m_loopStack.empty()); - BasicBlock *unr = new BasicBlock(name("unreachable"), m_func,0); - m_builder.CreateBr(m_loopStack.top().end); - m_builder.SetInsertPoint(unr); -} - -llvm::Value * Instructions::trunc(llvm::Value *in) -{ - std::vector vec = extractVector(in); - Value *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32), - name("ftoix")); - Value *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32), - name("ftoiy")); - Value *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32), - name("ftoiz")); - Value *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32), - name("ftoiw")); - Value *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy, - name("fx")); - Value *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy, - name("fy")); - Value *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy, - name("fz")); - Value *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy, - name("fw")); - return vectorFromVals(fx, fy, fz, fw); -} - -void Instructions::end() -{ - m_builder.CreateRetVoid(); -} - -void Instructions::cal(int label, llvm::Value *input) -{ - std::vector params; - params.push_back(input); - llvm::Function *func = findFunction(label); - - m_builder.CreateCall(func, params.begin(), params.end()); -} - -llvm::Function * Instructions::declareFunc(int label) -{ - PointerType *vecPtr = PointerType::getUnqual(m_floatVecType); - std::vector args; - args.push_back(vecPtr); - args.push_back(vecPtr); - args.push_back(vecPtr); - args.push_back(vecPtr); - ParamAttrsList *params = 0; - FunctionType *funcType = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/args, - /*isVarArg=*/false); - std::string name = createFuncName(label); - Function *func = new Function( - /*Type=*/funcType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/name.c_str(), m_mod); - func->setCallingConv(CallingConv::C); - func->setParamAttrs(params); - return func; -} - -void Instructions::bgnSub(unsigned label) -{ - llvm::Function *func = findFunction(label); - - Function::arg_iterator args = func->arg_begin(); - Value *ptr_INPUT = args++; - ptr_INPUT->setName("INPUT"); - m_storage->pushArguments(ptr_INPUT); - - llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); - - m_func = func; - m_builder.SetInsertPoint(entry); -} - -void Instructions::endSub() -{ - m_func = 0; - m_builder.SetInsertPoint(0); -} - -llvm::Function * Instructions::findFunction(int label) -{ - llvm::Function *func = m_functions[label]; - if (!func) { - func = declareFunc(label); - m_functions[label] = func; - } - return func; -} - -llvm::Value * Instructions::constVector(float x, float y, float z, float w) -{ - std::vector vec(4); - vec[0] = ConstantFP::get(Type::FloatTy, APFloat(x)); - vec[1] = ConstantFP::get(Type::FloatTy, APFloat(y)); - vec[2] = ConstantFP::get(Type::FloatTy, APFloat(z)); - vec[3] = ConstantFP::get(Type::FloatTy, APFloat(w)); - return ConstantVector::get(m_floatVecType, vec); -} - - -std::vector Instructions::extractVector(llvm::Value *vec) -{ - std::vector elems(4); - elems[0] = m_builder.CreateExtractElement(vec, m_storage->constantInt(0), - name("x")); - elems[1] = m_builder.CreateExtractElement(vec, m_storage->constantInt(1), - name("y")); - elems[2] = m_builder.CreateExtractElement(vec, m_storage->constantInt(2), - name("z")); - elems[3] = m_builder.CreateExtractElement(vec, m_storage->constantInt(3), - name("w")); - return elems; -} - -llvm::Value * Instructions::cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3) -{ - llvm::Function *func = m_mod->getFunction("cmp"); - assert(func); - - std::vector params; - params.push_back(in1); - params.push_back(in2); - params.push_back(in3); - CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), name("cmpres")); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::cos(llvm::Value *in) -{ -#if 0 - llvm::Function *func = m_mod->getFunction("vcos"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("cosres")); - call->setTailCall(false); - return call; -#else - std::vector elems = extractVector(in); - Function *func = m_mod->getFunction("cosf"); - assert(func); - CallInst *cos = m_builder.CreateCall(func, elems[0], name("cosres")); - cos->setCallingConv(CallingConv::C); - cos->setTailCall(true); - return vectorFromVals(cos, cos, cos, cos); -#endif -} - -llvm::Value * Instructions::scs(llvm::Value *in) -{ - llvm::Function *func = m_mod->getFunction("scs"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("scsres")); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::kilp(llvm::Value *in) -{ - llvm::Function *func = m_mod->getFunction("kilp"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("kilpres")); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::sin(llvm::Value *in) -{ - llvm::Function *func = m_mod->getFunction("vsin"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("sinres")); - call->setTailCall(false); - return call; -} -#endif //MESA_LLVM - - diff --git a/src/mesa/pipe/llvm/instructions.h b/src/mesa/pipe/llvm/instructions.h deleted file mode 100644 index 9ebc17dd8e..0000000000 --- a/src/mesa/pipe/llvm/instructions.h +++ /dev/null @@ -1,152 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef INSTRUCTIONS_H -#define INSTRUCTIONS_H - -#include -#include -#include -#include - -#include -#include - -namespace llvm { - class VectorType; - class Function; -} - -class Storage; - -class Instructions -{ -public: - Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, - Storage *storage); - - llvm::BasicBlock *currentBlock() const; - - llvm::Value *abs(llvm::Value *in1); - llvm::Value *arl(llvm::Value *in1); - llvm::Value *add(llvm::Value *in1, llvm::Value *in2); - void beginLoop(); - void bgnSub(unsigned); - void brk(); - void cal(int label, llvm::Value *input); - llvm::Value *cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); - llvm::Value *cos(llvm::Value *in); - llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); - void elseop(); - void endif(); - void endLoop(); - void end(); - void endSub(); - llvm::Value *ex2(llvm::Value *in); - llvm::Value *floor(llvm::Value *in); - llvm::Value *frc(llvm::Value *in); - void ifop(llvm::Value *in); - llvm::Value *kilp(llvm::Value *in); - llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in3); - llvm::Value *lit(llvm::Value *in); - llvm::Value *lg2(llvm::Value *in); - llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in2); - llvm::Value *min(llvm::Value *in1, llvm::Value *in2); - llvm::Value *max(llvm::Value *in1, llvm::Value *in2); - llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); - llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); - llvm::Value *rcp(llvm::Value *in); - llvm::Value *rsq(llvm::Value *in); - llvm::Value *scs(llvm::Value *in); - llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); - llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); - llvm::Value *sin(llvm::Value *in); - llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); - llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); - llvm::Value *trunc(llvm::Value *in); - - void printVector(llvm::Value *val); -private: - const char *name(const char *prefix); - - llvm::Value *callFAbs(llvm::Value *val); - llvm::Value *callFloor(llvm::Value *val); - llvm::Value *callFSqrt(llvm::Value *val); - llvm::Value *callFLog(llvm::Value *val); - llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2); - - llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w=0); - - llvm::Value *constVector(float x, float y, float z, float w); - - llvm::Function *declarePrintf(); - llvm::Function *declareFunc(int label); - - llvm::Function *findFunction(int label); - - std::vector extractVector(llvm::Value *vec); -private: - llvm::Module *m_mod; - llvm::Function *m_func; - char m_name[32]; - llvm::LLVMFoldingBuilder m_builder; - int m_idx; - - llvm::VectorType *m_floatVecType; - - llvm::Function *m_llvmFSqrt; - llvm::Function *m_llvmFAbs; - llvm::Function *m_llvmPow; - llvm::Function *m_llvmFloor; - llvm::Function *m_llvmFlog; - llvm::Function *m_llvmLit; - - llvm::Constant *m_fmtPtr; - - std::stack m_ifStack; - struct Loop { - llvm::BasicBlock *begin; - llvm::BasicBlock *end; - }; - std::stack m_loopStack; - std::map m_functions; - Storage *m_storage; -}; - -#endif diff --git a/src/mesa/pipe/llvm/instructionssoa.cpp b/src/mesa/pipe/llvm/instructionssoa.cpp deleted file mode 100644 index a4d5046637..0000000000 --- a/src/mesa/pipe/llvm/instructionssoa.cpp +++ /dev/null @@ -1,121 +0,0 @@ -#include "instructionssoa.h" - -#include "storagesoa.h" - -#include - -using namespace llvm; - -InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, - llvm::BasicBlock *block, StorageSoa *storage) - : m_builder(block), - m_storage(storage), - m_idx(0) -{ -} - -const char * InstructionsSoa::name(const char *prefix) const -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -llvm::Value * InstructionsSoa::vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w) -{ - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - Constant *constVector = Constant::getNullValue(vectorType); - Value *res = m_builder.CreateInsertElement(constVector, x, - m_storage->constantInt(0), - name("vecx")); - res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), - name("vecxy")); - res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), - name("vecxyz")); - if (w) - res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), - name("vecxyzw")); - return res; -} - -std::vector InstructionsSoa::arl(const std::vector in) -{ - std::vector res(4); - - //Extract x's - llvm::Value *x1 = m_builder.CreateExtractElement(in[0], - m_storage->constantInt(0), - name("extractX")); - //cast it to an unsigned int - x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast")); - - res[0] = x1;//vectorFromVals(x1, x2, x3, x4); - //only x is valid. the others shouldn't be necessary - /* - res[1] = Constant::getNullValue(m_floatVecType); - res[2] = Constant::getNullValue(m_floatVecType); - res[3] = Constant::getNullValue(m_floatVecType); - */ - - return res; -} - - -std::vector InstructionsSoa::add(const std::vector in1, - const std::vector in2) -{ - std::vector res(4); - - res[0] = m_builder.CreateAdd(in1[0], in2[0], name("addx")); - res[1] = m_builder.CreateAdd(in1[1], in2[1], name("addy")); - res[2] = m_builder.CreateAdd(in1[2], in2[2], name("addz")); - res[3] = m_builder.CreateAdd(in1[3], in2[3], name("addw")); - - return res; -} - -std::vector InstructionsSoa::mul(const std::vector in1, - const std::vector in2) -{ - std::vector res(4); - - res[0] = m_builder.CreateMul(in1[0], in2[0], name("mulx")); - res[1] = m_builder.CreateMul(in1[1], in2[1], name("muly")); - res[2] = m_builder.CreateMul(in1[2], in2[2], name("mulz")); - res[3] = m_builder.CreateMul(in1[3], in2[3], name("mulw")); - - return res; -} - -void InstructionsSoa::end() -{ - m_builder.CreateRetVoid(); -} - -std::vector InstructionsSoa::madd(const std::vector in1, - const std::vector in2, - const std::vector in3) -{ - std::vector res = mul(in1, in2); - return add(res, in3); -} - -std::vector InstructionsSoa::extractVector(llvm::Value *vector) -{ - std::vector res(4); - res[0] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(0), - name("extract1X")); - res[1] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(1), - name("extract2X")); - res[2] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(2), - name("extract3X")); - res[3] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(3), - name("extract4X")); - - return res; -} diff --git a/src/mesa/pipe/llvm/instructionssoa.h b/src/mesa/pipe/llvm/instructionssoa.h deleted file mode 100644 index 4169dcbb2e..0000000000 --- a/src/mesa/pipe/llvm/instructionssoa.h +++ /dev/null @@ -1,74 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INSTRUCTIONSSOA_H -#define INSTRUCTIONSSOA_H - -#include - -#include - -namespace llvm { - class Module; - class Function; - class BasicBlock; - class Value; -} -class StorageSoa; - -class InstructionsSoa -{ -public: - InstructionsSoa(llvm::Module *mod, llvm::Function *func, - llvm::BasicBlock *block, StorageSoa *storage); - - std::vector arl(const std::vector in); - - std::vector add(const std::vector in1, - const std::vector in2); - std::vector madd(const std::vector in1, - const std::vector in2, - const std::vector in3); - std::vector mul(const std::vector in1, - const std::vector in2); - void end(); - - std::vector extractVector(llvm::Value *vector); -private: - const char * name(const char *prefix) const; - llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w); -private: - llvm::LLVMFoldingBuilder m_builder; - StorageSoa *m_storage; -private: - mutable int m_idx; - mutable char m_name[32]; -}; - - -#endif diff --git a/src/mesa/pipe/llvm/llvm_builtins.c b/src/mesa/pipe/llvm/llvm_builtins.c deleted file mode 100644 index 4f98d754ba..0000000000 --- a/src/mesa/pipe/llvm/llvm_builtins.c +++ /dev/null @@ -1,115 +0,0 @@ -/*clang --emit-llvm llvm_builtins.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=gallivm_builtins.cpp -f -for=shader -funcname=createGallivmBuiltins*/ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -typedef __attribute__(( ocu_vector_type(4) )) float float4; - -extern float powf(float a, float b); - -inline float approx(float a, float b) -{ - if (b < -128.0f) b = -128.0f; - if (b > 128.0f) b = 128.0f; - if (a < 0) a = 0; - return powf(a, b); -} - -inline float4 lit(float4 tmp) -{ - float4 result; - result.x = 1.0; - result.w = 1.0; - if (tmp.x > 0) { - result.y = tmp.x; - result.z = approx(tmp.y, tmp.w); - } else { - result.y = 0; - result.z = 0; - } - return result; -} - -inline float4 cmp(float4 tmp0, float4 tmp1, float4 tmp2) -{ - float4 result; - - result.x = (tmp0.x < 0.0) ? tmp1.x : tmp2.x; - result.y = (tmp0.y < 0.0) ? tmp1.y : tmp2.y; - result.z = (tmp0.z < 0.0) ? tmp1.z : tmp2.z; - result.w = (tmp0.w < 0.0) ? tmp1.w : tmp2.w; - - return result; -} - -extern float cosf(float val); -extern float sinf(float val); - -inline float4 vcos(float4 val) -{ - float4 result; - printf("VEC IN is %f %f %f %f\n", val.x, val.y, val.z, val.w); - result.x = cosf(val.x); - result.y = cosf(val.x); - result.z = cosf(val.x); - result.w = cosf(val.x); - printf("VEC OUT is %f %f %f %f\n", result.x, result.y, result.z, result.w); - return result; -} - -inline float4 scs(float4 val) -{ - float4 result; - float tmp = val.x; - result.x = cosf(tmp); - result.y = sinf(tmp); - return result; -} - - -inline float4 vsin(float4 val) -{ - float4 result; - float tmp = val.x; - float res = sinf(tmp); - result.x = res; - result.y = res; - result.z = res; - result.w = res; - return result; -} - -inline int kilp(float4 val) -{ - if (val.x < 0 || val.y < 0 || val.z < 0 || val.w < 0) - return 1; - else - return 0; -} diff --git a/src/mesa/pipe/llvm/loweringpass.cpp b/src/mesa/pipe/llvm/loweringpass.cpp deleted file mode 100644 index 556dbec366..0000000000 --- a/src/mesa/pipe/llvm/loweringpass.cpp +++ /dev/null @@ -1,17 +0,0 @@ -#include "loweringpass.h" - -using namespace llvm; - -char LoweringPass::ID = 0; -RegisterPass X("lowering", "Lowering Pass"); - -LoweringPass::LoweringPass() - : ModulePass((intptr_t)&ID) -{ -} - -bool LoweringPass::runOnModule(Module &m) -{ - llvm::cerr << "Hello: " << m.getModuleIdentifier() << "\n"; - return false; -} diff --git a/src/mesa/pipe/llvm/loweringpass.h b/src/mesa/pipe/llvm/loweringpass.h deleted file mode 100644 index f62dcf6ba7..0000000000 --- a/src/mesa/pipe/llvm/loweringpass.h +++ /dev/null @@ -1,15 +0,0 @@ -#ifndef LOWERINGPASS_H -#define LOWERINGPASS_H - -#include "llvm/Pass.h" -#include "llvm/Module.h" - -struct LoweringPass : public llvm::ModulePass -{ - static char ID; - LoweringPass(); - - virtual bool runOnModule(llvm::Module &m); -}; - -#endif diff --git a/src/mesa/pipe/llvm/storage.cpp b/src/mesa/pipe/llvm/storage.cpp deleted file mode 100644 index c4326de8c5..0000000000 --- a/src/mesa/pipe/llvm/storage.cpp +++ /dev/null @@ -1,364 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "storage.h" - -#include "gallivm_p.h" - -#include "pipe/p_shader_tokens.h" -#include -#include -#include - -#include -#include -#include -#include -#include - -using namespace llvm; - -Storage::Storage(llvm::BasicBlock *block, llvm::Value *input) - : m_block(block), - m_INPUT(input), - m_addrs(32), - m_idx(0) -{ - m_floatVecType = VectorType::get(Type::FloatTy, 4); - m_intVecType = VectorType::get(IntegerType::get(32), 4); - - m_undefFloatVec = UndefValue::get(m_floatVecType); - m_undefIntVec = UndefValue::get(m_intVecType); - m_extSwizzleVec = 0; - - m_numConsts = 0; -} - -//can only build vectors with all members in the [0, 9] range -llvm::Constant *Storage::shuffleMask(int vec) -{ - if (!m_extSwizzleVec) { - std::vector elems; - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); - m_extSwizzleVec = ConstantVector::get(m_floatVecType, elems); - } - - if (m_intVecs.find(vec) != m_intVecs.end()) { - return m_intVecs[vec]; - } - int origVec = vec; - Constant* const_vec = 0; - if (origVec == 0) { - const_vec = Constant::getNullValue(m_intVecType); - } else { - int x = gallivm_x_swizzle(vec); - int y = gallivm_y_swizzle(vec); - int z = gallivm_z_swizzle(vec); - int w = gallivm_w_swizzle(vec); - std::vector elems; - elems.push_back(constantInt(x)); - elems.push_back(constantInt(y)); - elems.push_back(constantInt(z)); - elems.push_back(constantInt(w)); - const_vec = ConstantVector::get(m_intVecType, elems); - } - - m_intVecs[origVec] = const_vec; - return const_vec; -} - -llvm::ConstantInt *Storage::constantInt(int idx) -{ - if (m_constInts.find(idx) != m_constInts.end()) { - return m_constInts[idx]; - } - ConstantInt *const_int = ConstantInt::get(APInt(32, idx)); - m_constInts[idx] = const_int; - return const_int; -} - -llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) -{ - Value *val = element(InputsArg, idx, indIdx); - LoadInst *load = new LoadInst(val, name("input"), false, m_block); - load->setAlignment(8); - - return load; -} - -llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) -{ - m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; - - Value *elem = element(ConstsArg, idx, indIdx); - LoadInst *load = new LoadInst(elem, name("const"), false, m_block); - load->setAlignment(8); - return load; -} - -llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) -{ - Constant *mask = shuffleMask(shuffle); - ShuffleVectorInst *res = - new ShuffleVectorInst(vec, m_extSwizzleVec, mask, - name("shuffle"), m_block); - return res; -} - - -llvm::Value *Storage::tempElement(int idx, llvm::Value *indIdx) -{ - Value *elem = element(TempsArg, idx, indIdx); - - LoadInst *load = new LoadInst(elem, name("temp"), false, m_block); - load->setAlignment(8); - - return load; -} - -void Storage::setTempElement(int idx, llvm::Value *val, int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = 0; - if (m_tempWriteMap[idx]) - templ = tempElement(idx); - val = maskWrite(val, mask, templ); - } - Value *elem = element(TempsArg, idx); - StoreInst *st = new StoreInst(val, elem, false, m_block); - st->setAlignment(8); - m_tempWriteMap[idx] = true; -} - -void Storage::setOutputElement(int dstIdx, llvm::Value *val, int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = 0; - if (m_destWriteMap[dstIdx]) - templ = outputElement(dstIdx); - val = maskWrite(val, mask, templ); - } - - Value *elem = element(DestsArg, dstIdx); - StoreInst *st = new StoreInst(val, elem, false, m_block); - st->setAlignment(8); - m_destWriteMap[dstIdx] = true; -} - -llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ) -{ - llvm::Value *dst = templ; - if (!dst) - dst = Constant::getNullValue(m_floatVecType); - if ((mask & TGSI_WRITEMASK_X)) { - llvm::Value *x = new ExtractElementInst(src, unsigned(0), - name("x"), m_block); - dst = new InsertElementInst(dst, x, unsigned(0), - name("dstx"), m_block); - } - if ((mask & TGSI_WRITEMASK_Y)) { - llvm::Value *y = new ExtractElementInst(src, unsigned(1), - name("y"), m_block); - dst = new InsertElementInst(dst, y, unsigned(1), - name("dsty"), m_block); - } - if ((mask & TGSI_WRITEMASK_Z)) { - llvm::Value *z = new ExtractElementInst(src, unsigned(2), - name("z"), m_block); - dst = new InsertElementInst(dst, z, unsigned(2), - name("dstz"), m_block); - } - if ((mask & TGSI_WRITEMASK_W)) { - llvm::Value *w = new ExtractElementInst(src, unsigned(3), - name("w"), m_block); - dst = new InsertElementInst(dst, w, unsigned(3), - name("dstw"), m_block); - } - return dst; -} - -const char * Storage::name(const char *prefix) -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -int Storage::numConsts() const -{ - return m_numConsts; -} - -llvm::Value * Storage::addrElement(int idx) const -{ - Value *ret = m_addrs[idx]; - if (!ret) - return m_undefFloatVec; - return ret; -} - -void Storage::setAddrElement(int idx, llvm::Value *val, int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = m_addrs[idx]; - val = maskWrite(val, mask, templ); - } - m_addrs[idx] = val; -} - -llvm::Value * Storage::extractIndex(llvm::Value *vec) -{ - llvm::Value *x = new ExtractElementInst(vec, unsigned(0), - name("x"), m_block); - return new FPToSIInst(x, IntegerType::get(32), name("intidx"), m_block); -} - -void Storage::setCurrentBlock(llvm::BasicBlock *block) -{ - m_block = block; -} - -llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx) -{ - Value *elem = element(DestsArg, idx, indIdx); - LoadInst *load = new LoadInst(elem, name("output"), false, m_block); - load->setAlignment(8); - - return load; -} - -llvm::Value * Storage::inputPtr() const -{ - return m_INPUT; -} - -void Storage::pushArguments(llvm::Value *input) -{ - m_argStack.push(m_INPUT); - - m_INPUT = input; -} - -void Storage::popArguments() -{ - m_INPUT = m_argStack.top(); - m_argStack.pop(); -} - -void Storage::pushTemps() -{ - m_extSwizzleVec = 0; -} - -void Storage::popTemps() -{ -} - -llvm::Value * Storage::immediateElement(int idx) -{ - return m_immediates[idx]; -} - -void Storage::addImmediate(float *val) -{ - std::vector vec(4); - vec[0] = ConstantFP::get(Type::FloatTy, APFloat(val[0])); - vec[1] = ConstantFP::get(Type::FloatTy, APFloat(val[1])); - vec[2] = ConstantFP::get(Type::FloatTy, APFloat(val[2])); - vec[3] = ConstantFP::get(Type::FloatTy, APFloat(val[3])); - m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); -} - - -llvm::Value * Storage::elemPtr(Args arg) -{ - std::vector indices; - indices.push_back(constantInt(0)); - indices.push_back(constantInt(static_cast(arg))); - GetElementPtrInst *getElem = new GetElementPtrInst(m_INPUT, - indices.begin(), - indices.end(), - name("input_ptr"), - m_block); - return new LoadInst(getElem, name("input_field"), false, m_block); -} - -llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx, - llvm::Value *indIdx ) -{ - GetElementPtrInst *getElem = 0; - - if (indIdx) { - getElem = new GetElementPtrInst(ptr, - BinaryOperator::create(Instruction::Add, - indIdx, - constantInt(idx), - name("add"), - m_block), - name("field"), - m_block); - } else { - getElem = new GetElementPtrInst(ptr, - constantInt(idx), - name("field"), - m_block); - } - return getElem; -} - -llvm::Value * Storage::element(Args arg, int idx, llvm::Value *indIdx ) -{ - Value *val = elemPtr(arg); - return elemIdx(val, idx, indIdx); -} - -void Storage::setKilElement(llvm::Value *val) -{ - std::vector indices; - indices.push_back(constantInt(0)); - indices.push_back(constantInt(static_cast(KilArg))); - GetElementPtrInst *elem = new GetElementPtrInst(m_INPUT, - indices.begin(), - indices.end(), - name("kil_ptr"), - m_block); - StoreInst *st = new StoreInst(val, elem, false, m_block); - st->setAlignment(8); -} - -#endif //MESA_LLVM - - diff --git a/src/mesa/pipe/llvm/storage.h b/src/mesa/pipe/llvm/storage.h deleted file mode 100644 index 8574f7554e..0000000000 --- a/src/mesa/pipe/llvm/storage.h +++ /dev/null @@ -1,133 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef STORAGE_H -#define STORAGE_H - -#include -#include -#include -#include - -namespace llvm { - class BasicBlock; - class Constant; - class ConstantInt; - class LoadInst; - class Value; - class VectorType; -} - -class Storage -{ -public: - Storage(llvm::BasicBlock *block, - llvm::Value *input); - - llvm::Value *inputPtr() const; - - void setCurrentBlock(llvm::BasicBlock *block); - - llvm::ConstantInt *constantInt(int); - llvm::Constant *shuffleMask(int vec); - llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); - llvm::Value *constElement(int idx, llvm::Value *indIdx =0); - llvm::Value *outputElement(int idx, llvm::Value *indIdx =0); - llvm::Value *tempElement(int idx, llvm::Value *indIdx =0); - llvm::Value *immediateElement(int idx); - - void setOutputElement(int dstIdx, llvm::Value *val, int mask); - void setTempElement(int idx, llvm::Value *val, int mask); - - llvm::Value *addrElement(int idx) const; - void setAddrElement(int idx, llvm::Value *val, int mask); - - void setKilElement(llvm::Value *val); - - llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); - - llvm::Value *extractIndex(llvm::Value *vec); - - int numConsts() const; - - void pushArguments(llvm::Value *input); - void popArguments(); - void pushTemps(); - void popTemps(); - - void addImmediate(float *val); - -private: - llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); - const char *name(const char *prefix); - - enum Args { - DestsArg = 0, - InputsArg = 1, - TempsArg = 2, - ConstsArg = 3, - KilArg = 4 - }; - llvm::Value *elemPtr(Args arg); - llvm::Value *elemIdx(llvm::Value *ptr, int idx, - llvm::Value *indIdx = 0); - llvm::Value *element(Args arg, int idx, llvm::Value *indIdx = 0); - -private: - llvm::BasicBlock *m_block; - llvm::Value *m_INPUT; - - std::map m_constInts; - std::map m_intVecs; - std::vector m_addrs; - std::vector m_immediates; - - llvm::VectorType *m_floatVecType; - llvm::VectorType *m_intVecType; - - char m_name[32]; - int m_idx; - - int m_numConsts; - - std::map m_destWriteMap; - std::map m_tempWriteMap; - - llvm::Value *m_undefFloatVec; - llvm::Value *m_undefIntVec; - llvm::Value *m_extSwizzleVec; - - std::stack m_argStack; - std::stack > m_tempStack; -}; - -#endif diff --git a/src/mesa/pipe/llvm/storagesoa.cpp b/src/mesa/pipe/llvm/storagesoa.cpp deleted file mode 100644 index ed0674a96f..0000000000 --- a/src/mesa/pipe/llvm/storagesoa.cpp +++ /dev/null @@ -1,389 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "storagesoa.h" - -#include "gallivm_p.h" - -#include "pipe/p_shader_tokens.h" -#include "pipe/p_debug.h" - -#include -#include -#include - -#include -#include -#include -#include -#include - -using namespace llvm; - - -StorageSoa::StorageSoa(llvm::BasicBlock *block, - llvm::Value *input, - llvm::Value *output, - llvm::Value *consts, - llvm::Value *temps) - : m_block(block), - m_input(input), - m_output(output), - m_consts(consts), - m_temps(temps), - m_immediates(0), - m_idx(0) -{ -} - -void StorageSoa::addImmediate(float *vec) -{ - std::vector vals(4); - vals[0] = vec[0]; - vals[1] = vec[1]; - vals[2] = vec[2]; - vals[3] = vec[3]; - m_immediatesToFlush.push_back(vals); -} - -void StorageSoa::declareImmediates() -{ - if (m_immediatesToFlush.empty()) - return; - - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - ArrayType *vectorChannels = ArrayType::get(vectorType, 4); - ArrayType *arrayType = ArrayType::get(vectorChannels, m_immediatesToFlush.size()); - - m_immediates = new GlobalVariable( - /*Type=*/arrayType, - /*isConstant=*/false, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name("immediates"), - currentModule()); - - std::vector arrayVals; - for (unsigned int i = 0; i < m_immediatesToFlush.size(); ++i) { - std::vector vec = m_immediatesToFlush[i]; - std::vector vals(4); - std::vector channelArray; - - vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0]; - llvm::Constant *xChannel = createConstGlobalVector(vals); - - vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1]; - llvm::Constant *yChannel = createConstGlobalVector(vals); - - vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2]; - llvm::Constant *zChannel = createConstGlobalVector(vals); - - vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3]; - llvm::Constant *wChannel = createConstGlobalVector(vals); - channelArray.push_back(xChannel); - channelArray.push_back(yChannel); - channelArray.push_back(zChannel); - channelArray.push_back(wChannel); - Constant *constChannels = ConstantArray::get(vectorChannels, - channelArray); - arrayVals.push_back(constChannels); - } - Constant *constArray = ConstantArray::get(arrayType, arrayVals); - m_immediates->setInitializer(constArray); - - m_immediatesToFlush.clear(); -} - -llvm::Value *StorageSoa::addrElement(int idx) const -{ - std::map::const_iterator itr = m_addresses.find(idx); - if (itr == m_addresses.end()) { - debug_printf("Trying to access invalid shader 'address'\n"); - return 0; - } - llvm::Value * res = (*itr).second; - - res = new LoadInst(res, name("addr"), false, m_block); - - return res; -} - -std::vector StorageSoa::inputElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_input, idx, 0); - res[1] = element(m_input, idx, 1); - res[2] = element(m_input, idx, 2); - res[3] = element(m_input, idx, 3); - - return res; -} - -std::vector StorageSoa::constElement(llvm::Value *idx) -{ - std::vector res(4); - llvm::Value *xChannel, *yChannel, *zChannel, *wChannel; - - xChannel = elementPointer(m_consts, idx, 0); - yChannel = elementPointer(m_consts, idx, 1); - zChannel = elementPointer(m_consts, idx, 2); - wChannel = elementPointer(m_consts, idx, 3); - - res[0] = alignedArrayLoad(xChannel); - res[1] = alignedArrayLoad(yChannel); - res[2] = alignedArrayLoad(zChannel); - res[3] = alignedArrayLoad(wChannel); - - return res; -} - -std::vector StorageSoa::outputElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_output, idx, 0); - res[1] = element(m_output, idx, 1); - res[2] = element(m_output, idx, 2); - res[3] = element(m_output, idx, 3); - - return res; -} - -std::vector StorageSoa::tempElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_temps, idx, 0); - res[1] = element(m_temps, idx, 1); - res[2] = element(m_temps, idx, 2); - res[3] = element(m_temps, idx, 3); - - return res; -} - -std::vector StorageSoa::immediateElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_immediates, idx, 0); - res[1] = element(m_immediates, idx, 1); - res[2] = element(m_immediates, idx, 2); - res[3] = element(m_immediates, idx, 3); - - return res; -} - -llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index, - int channel) const -{ - std::vector indices; - if (m_immediates == ptr) - indices.push_back(constantInt(0)); - indices.push_back(index); - indices.push_back(constantInt(channel)); - - GetElementPtrInst *getElem = new GetElementPtrInst(ptr, - indices.begin(), - indices.end(), - name("ptr"), - m_block); - return getElem; -} - -llvm::Value * StorageSoa::element(llvm::Value *ptr, llvm::Value *index, - int channel) const -{ - llvm::Value *res = elementPointer(ptr, index, channel); - LoadInst *load = new LoadInst(res, name("element"), false, m_block); - //load->setAlignment(8); - return load; -} - -const char * StorageSoa::name(const char *prefix) const -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -llvm::ConstantInt * StorageSoa::constantInt(int idx) const -{ - if (m_constInts.find(idx) != m_constInts.end()) { - return m_constInts[idx]; - } - ConstantInt *constInt = ConstantInt::get(APInt(32, idx)); - m_constInts[idx] = constInt; - return constInt; -} - -llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val) -{ - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - PointerType *vectorPtr = PointerType::get(vectorType, 0); - - CastInst *cast = new BitCastInst(val, vectorPtr, name("toVector"), m_block); - LoadInst *load = new LoadInst(cast, name("alignLoad"), false, m_block); - load->setAlignment(8); - return load; -} - -llvm::Module * StorageSoa::currentModule() const -{ - if (!m_block || !m_block->getParent()) - return 0; - - return m_block->getParent()->getParent(); -} - -llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector &vec) -{ - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - std::vector immValues; - ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0])); - ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1])); - ConstantFP *constz = ConstantFP::get(Type::FloatTy, APFloat(vec[2])); - ConstantFP *constw = ConstantFP::get(Type::FloatTy, APFloat(vec[3])); - immValues.push_back(constx); - immValues.push_back(consty); - immValues.push_back(constz); - immValues.push_back(constw); - Constant *constVector = ConstantVector::get(vectorType, immValues); - - return constVector; -} - -std::vector StorageSoa::load(Argument type, int idx, int swizzle, - llvm::Value *indIdx) -{ - std::vector val(4); - - //if we have an indirect index, always use that - // if not use the integer offset to create one - llvm::Value *realIndex = 0; - if (indIdx) - realIndex = indIdx; - else - realIndex = constantInt(idx); - debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx); - - switch(type) { - case Input: - val = inputElement(realIndex); - break; - case Output: - val = outputElement(realIndex); - break; - case Temp: - val = tempElement(realIndex); - break; - case Const: - val = constElement(realIndex); - break; - case Immediate: - val = immediateElement(realIndex); - break; - case Address: - debug_printf("Address not handled in the load phase!\n"); - assert(0); - break; - } - if (!gallivm_is_swizzle(swizzle)) - return val; - - std::vector res(4); - - res[0] = val[gallivm_x_swizzle(swizzle)]; - res[1] = val[gallivm_y_swizzle(swizzle)]; - res[2] = val[gallivm_z_swizzle(swizzle)]; - res[3] = val[gallivm_w_swizzle(swizzle)]; - return res; -} - -void StorageSoa::store(Argument type, int idx, const std::vector &val, - int mask) -{ - llvm::Value *out = 0; - switch(type) { - case Output: - out = m_output; - break; - case Temp: - out = m_temps; - break; - case Input: - out = m_input; - break; - case Address: { - llvm::Value *addr = m_addresses[idx]; - if (!addr) { - addAddress(idx); - addr = m_addresses[idx]; - assert(addr); - } - new StoreInst(val[0], addr, false, m_block); - return; - break; - } - default: - debug_printf("Can't save output of this type: %d !\n", type); - assert(0); - break; - } - llvm::Value *realIndex = constantInt(idx); - if ((mask & TGSI_WRITEMASK_X)) { - llvm::Value *xChannel = elementPointer(out, realIndex, 0); - new StoreInst(val[0], xChannel, false, m_block); - } - if ((mask & TGSI_WRITEMASK_Y)) { - llvm::Value *yChannel = elementPointer(out, realIndex, 1); - new StoreInst(val[1], yChannel, false, m_block); - } - if ((mask & TGSI_WRITEMASK_Z)) { - llvm::Value *zChannel = elementPointer(out, realIndex, 2); - new StoreInst(val[2], zChannel, false, m_block); - } - if ((mask & TGSI_WRITEMASK_W)) { - llvm::Value *wChannel = elementPointer(out, realIndex, 3); - new StoreInst(val[3], wChannel, false, m_block); - } -} - -void StorageSoa::addAddress(int idx) -{ - GlobalVariable *val = new GlobalVariable( - /*Type=*/IntegerType::get(32), - /*isConstant=*/false, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name("address"), - currentModule()); - val->setInitializer(Constant::getNullValue(IntegerType::get(32))); - - debug_printf("adding to %d\n", idx); - m_addresses[idx] = val; -} diff --git a/src/mesa/pipe/llvm/storagesoa.h b/src/mesa/pipe/llvm/storagesoa.h deleted file mode 100644 index 6443351f27..0000000000 --- a/src/mesa/pipe/llvm/storagesoa.h +++ /dev/null @@ -1,111 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef STORAGESOA_H -#define STORAGESOA_H - -#include -#include -#include - -namespace llvm { - class BasicBlock; - class Constant; - class ConstantInt; - class GlobalVariable; - class LoadInst; - class Value; - class VectorType; - class Module; -} - -class StorageSoa -{ -public: - enum Argument { - Input, - Output, - Temp, - Const, - Immediate, - Address - }; -public: - StorageSoa(llvm::BasicBlock *block, - llvm::Value *input, - llvm::Value *output, - llvm::Value *consts, - llvm::Value *temps); - - - std::vector load(Argument type, int idx, int swizzle, - llvm::Value *indIdx =0); - void store(Argument type, int idx, const std::vector &val, - int mask); - - void addImmediate(float *vec); - void declareImmediates(); - - void addAddress(int idx); - - llvm::Value * addrElement(int idx) const; - - llvm::ConstantInt *constantInt(int) const; -private: - llvm::Value *elementPointer(llvm::Value *ptr, llvm::Value *indIdx, - int channel) const; - llvm::Value *element(llvm::Value *ptr, llvm::Value *idx, - int channel) const; - const char *name(const char *prefix) const; - llvm::Value *alignedArrayLoad(llvm::Value *val); - llvm::Module *currentModule() const; - llvm::Constant *createConstGlobalVector(const std::vector &vec); - - std::vector inputElement(llvm::Value *indIdx); - std::vector constElement(llvm::Value *indIdx); - std::vector outputElement(llvm::Value *indIdx); - std::vector tempElement(llvm::Value *indIdx); - std::vector immediateElement(llvm::Value *indIdx); -private: - llvm::BasicBlock *m_block; - - llvm::Value *m_input; - llvm::Value *m_output; - llvm::Value *m_consts; - llvm::Value *m_temps; - llvm::GlobalVariable *m_immediates; - - std::map m_addresses; - - std::vector > m_immediatesToFlush; - - mutable std::map m_constInts; - mutable char m_name[32]; - mutable int m_idx; -}; - -#endif diff --git a/src/mesa/pipe/llvm/tgsitollvm.cpp b/src/mesa/pipe/llvm/tgsitollvm.cpp deleted file mode 100644 index 0de595e678..0000000000 --- a/src/mesa/pipe/llvm/tgsitollvm.cpp +++ /dev/null @@ -1,1221 +0,0 @@ -#include "tgsitollvm.h" - -#include "gallivm.h" -#include "gallivm_p.h" - -#include "storage.h" -#include "instructions.h" -#include "storagesoa.h" -#include "instructionssoa.h" - -#include "pipe/p_shader_tokens.h" - -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/exec/tgsi_exec.h" -#include "pipe/tgsi/util/tgsi_util.h" -#include "pipe/tgsi/util/tgsi_build.h" -#include "pipe/tgsi/util/tgsi_dump.h" - - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - - -#include -#include -#include - -using namespace llvm; - -static inline FunctionType *vertexShaderFunctionType() -{ - //Function takes three arguments, - // the calling code has to make sure the types it will - // pass are castable to the following: - // [4 x <4 x float>] inputs, - // [4 x <4 x float>] output, - // [4 x [4 x float]] consts, - // [4 x <4 x float>] temps - - std::vector funcArgs; - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - ArrayType *vectorArray = ArrayType::get(vectorType, 4); - PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0); - - ArrayType *floatArray = ArrayType::get(Type::FloatTy, 4); - ArrayType *constsArray = ArrayType::get(floatArray, 4); - PointerType *constsArrayPtr = PointerType::get(constsArray, 0); - - funcArgs.push_back(vectorArrayPtr);//inputs - funcArgs.push_back(vectorArrayPtr);//output - funcArgs.push_back(constsArrayPtr);//consts - funcArgs.push_back(vectorArrayPtr);//temps - - FunctionType *functionType = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/funcArgs, - /*isVarArg=*/false); - - return functionType; -} - -static inline void -add_interpolator(struct gallivm_ir *ir, - struct gallivm_interpolate *interp) -{ - ir->interpolators[ir->num_interp] = *interp; - ++ir->num_interp; -} - -static void -translate_declaration(struct gallivm_ir *prog, - llvm::Module *module, - Storage *storage, - struct tgsi_full_declaration *decl, - struct tgsi_full_declaration *fd) -{ - if (decl->Declaration.File == TGSI_FILE_INPUT) { - unsigned first, last, mask; - uint interp_method; - - assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - /* Do not touch WPOS.xy */ - if (first == 0) { - mask &= ~TGSI_WRITEMASK_XY; - if (mask == TGSI_WRITEMASK_NONE) { - first++; - if (first > last) { - return; - } - } - } - - interp_method = decl->Interpolation.Interpolate; - - if (mask == TGSI_WRITEMASK_XYZW) { - unsigned i, j; - - for (i = first; i <= last; i++) { - for (j = 0; j < NUM_CHANNELS; j++) { - //interp( mach, i, j ); - struct gallivm_interpolate interp; - interp.type = interp_method; - interp.attrib = i; - interp.chan = j; - add_interpolator(prog, &interp); - } - } - } else { - unsigned i, j; - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - struct gallivm_interpolate interp; - interp.type = interp_method; - interp.attrib = i; - interp.chan = j; - add_interpolator(prog, &interp); - } - } - } - } - } -} - -static void -translate_declarationir(struct gallivm_ir *, - llvm::Module *, - StorageSoa *storage, - struct tgsi_full_declaration *decl, - struct tgsi_full_declaration *) -{ - if (decl->Declaration.File == TGSI_FILE_ADDRESS) { - int idx = decl->u.DeclarationRange.First; - storage->addAddress(idx); - } -} - -static void -translate_immediate(Storage *storage, - struct tgsi_full_immediate *imm) -{ - float vec[4]; - int i; - for (i = 0; i < imm->Immediate.Size - 1; ++i) { - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; - break; - default: - assert(0); - } - } - storage->addImmediate(vec); -} - - -static void -translate_immediateir(StorageSoa *storage, - struct tgsi_full_immediate *imm) -{ - float vec[4]; - int i; - for (i = 0; i < imm->Immediate.Size - 1; ++i) { - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; - break; - default: - assert(0); - } - } - storage->addImmediate(vec); -} - -static inline int -swizzleInt(struct tgsi_full_src_register *src) -{ - int swizzle = 0; - int start = 1000; - - for (int k = 0; k < 4; ++k) { - swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; - start /= 10; - } - return swizzle; -} - -static inline llvm::Value * -swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, - Storage *storage) -{ - int swizzle = swizzleInt(src); - - if (gallivm_is_swizzle(swizzle)) { - /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ - val = storage->shuffleVector(val, swizzle); - } - return val; -} - -static void -translate_instruction(llvm::Module *module, - Storage *storage, - Instructions *instr, - struct tgsi_full_instruction *inst, - struct tgsi_full_instruction *fi, - unsigned instno) -{ - llvm::Value *inputs[4]; - inputs[0] = 0; - inputs[1] = 0; - inputs[2] = 0; - inputs[3] = 0; - - for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - llvm::Value *val = 0; - llvm::Value *indIdx = 0; - - if (src->SrcRegister.Indirect) { - indIdx = storage->addrElement(src->SrcRegisterInd.Index); - indIdx = storage->extractIndex(indIdx); - } - if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->constElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->inputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->tempElement(src->SrcRegister.Index); - } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->outputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->immediateElement(src->SrcRegister.Index); - } else { - fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); - return; - } - - inputs[i] = swizzleVector(val, src, storage); - } - - /*if (inputs[0]) - instr->printVector(inputs[0]); - if (inputs[1]) - instr->printVector(inputs[1]);*/ - llvm::Value *out = 0; - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: { - out = instr->arl(inputs[0]); - } - break; - case TGSI_OPCODE_MOV: { - out = inputs[0]; - } - break; - case TGSI_OPCODE_LIT: { - out = instr->lit(inputs[0]); - } - break; - case TGSI_OPCODE_RCP: { - out = instr->rcp(inputs[0]); - } - break; - case TGSI_OPCODE_RSQ: { - out = instr->rsq(inputs[0]); - } - break; - case TGSI_OPCODE_EXP: - break; - case TGSI_OPCODE_LOG: - break; - case TGSI_OPCODE_MUL: { - out = instr->mul(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_ADD: { - out = instr->add(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP3: { - out = instr->dp3(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP4: { - out = instr->dp4(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DST: { - out = instr->dst(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MIN: { - out = instr->min(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAX: { - out = instr->max(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SLT: { - out = instr->slt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SGE: { - out = instr->sge(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAD: { - out = instr->madd(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SUB: { - out = instr->sub(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_LERP: { - out = instr->lerp(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_CND: - break; - case TGSI_OPCODE_CND0: - break; - case TGSI_OPCODE_DOT2ADD: - break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { - out = instr->frc(inputs[0]); - } - break; - case TGSI_OPCODE_CLAMP: - break; - case TGSI_OPCODE_FLOOR: { - out = instr->floor(inputs[0]); - } - break; - case TGSI_OPCODE_ROUND: - break; - case TGSI_OPCODE_EXPBASE2: { - out = instr->ex2(inputs[0]); - } - break; - case TGSI_OPCODE_LOGBASE2: { - out = instr->lg2(inputs[0]); - } - break; - case TGSI_OPCODE_POWER: { - out = instr->pow(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_CROSSPRODUCT: { - out = instr->cross(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; - case TGSI_OPCODE_ABS: { - out = instr->abs(inputs[0]); - } - break; - case TGSI_OPCODE_RCC: - break; - case TGSI_OPCODE_DPH: { - out = instr->dph(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_COS: { - out = instr->cos(inputs[0]); - } - break; - case TGSI_OPCODE_DDX: - break; - case TGSI_OPCODE_DDY: - break; - case TGSI_OPCODE_KILP: { - out = instr->kilp(inputs[0]); - storage->setKilElement(out); - return; - } - break; - case TGSI_OPCODE_PK2H: - break; - case TGSI_OPCODE_PK2US: - break; - case TGSI_OPCODE_PK4B: - break; - case TGSI_OPCODE_PK4UB: - break; - case TGSI_OPCODE_RFL: - break; - case TGSI_OPCODE_SEQ: - break; - case TGSI_OPCODE_SFL: - break; - case TGSI_OPCODE_SGT: { - out = instr->sgt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SIN: { - out = instr->sin(inputs[0]); - } - break; - case TGSI_OPCODE_SLE: - break; - case TGSI_OPCODE_SNE: - break; - case TGSI_OPCODE_STR: - break; - case TGSI_OPCODE_TEX: - break; - case TGSI_OPCODE_TXD: - break; - case TGSI_OPCODE_UP2H: - break; - case TGSI_OPCODE_UP2US: - break; - case TGSI_OPCODE_UP4B: - break; - case TGSI_OPCODE_UP4UB: - break; - case TGSI_OPCODE_X2D: - break; - case TGSI_OPCODE_ARA: - break; - case TGSI_OPCODE_ARR: - break; - case TGSI_OPCODE_BRA: - break; - case TGSI_OPCODE_CAL: { - instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr()); - return; - } - break; - case TGSI_OPCODE_RET: { - instr->end(); - return; - } - break; - case TGSI_OPCODE_SSG: - break; - case TGSI_OPCODE_CMP: { - out = instr->cmp(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SCS: { - out = instr->scs(inputs[0]); - } - break; - case TGSI_OPCODE_TXB: - break; - case TGSI_OPCODE_NRM: - break; - case TGSI_OPCODE_DIV: - break; - case TGSI_OPCODE_DP2: - break; - case TGSI_OPCODE_TXL: - break; - case TGSI_OPCODE_BRK: { - instr->brk(); - return; - } - break; - case TGSI_OPCODE_IF: { - instr->ifop(inputs[0]); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_LOOP: - break; - case TGSI_OPCODE_REP: - break; - case TGSI_OPCODE_ELSE: { - instr->elseop(); - storage->setCurrentBlock(instr->currentBlock()); - return; //only state update - } - break; - case TGSI_OPCODE_ENDIF: { - instr->endif(); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_ENDLOOP: - break; - case TGSI_OPCODE_ENDREP: - break; - case TGSI_OPCODE_PUSHA: - break; - case TGSI_OPCODE_POPA: - break; - case TGSI_OPCODE_CEIL: - break; - case TGSI_OPCODE_I2F: - break; - case TGSI_OPCODE_NOT: - break; - case TGSI_OPCODE_TRUNC: { - out = instr->trunc(inputs[0]); - } - break; - case TGSI_OPCODE_SHL: - break; - case TGSI_OPCODE_SHR: - break; - case TGSI_OPCODE_AND: - break; - case TGSI_OPCODE_OR: - break; - case TGSI_OPCODE_MOD: - break; - case TGSI_OPCODE_XOR: - break; - case TGSI_OPCODE_SAD: - break; - case TGSI_OPCODE_TXF: - break; - case TGSI_OPCODE_TXQ: - break; - case TGSI_OPCODE_CONT: - break; - case TGSI_OPCODE_EMIT: - break; - case TGSI_OPCODE_ENDPRIM: - break; - case TGSI_OPCODE_BGNLOOP2: { - instr->beginLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_BGNSUB: { - instr->bgnSub(instno); - storage->setCurrentBlock(instr->currentBlock()); - storage->pushTemps(); - return; - } - break; - case TGSI_OPCODE_ENDLOOP2: { - instr->endLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_ENDSUB: { - instr->endSub(); - storage->setCurrentBlock(instr->currentBlock()); - storage->popArguments(); - storage->popTemps(); - return; - } - break; - case TGSI_OPCODE_NOISE1: - break; - case TGSI_OPCODE_NOISE2: - break; - case TGSI_OPCODE_NOISE3: - break; - case TGSI_OPCODE_NOISE4: - break; - case TGSI_OPCODE_NOP: - break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; - case TGSI_OPCODE_NRM4: - break; - case TGSI_OPCODE_CALLNZ: - break; - case TGSI_OPCODE_IFC: - break; - case TGSI_OPCODE_BREAKC: - break; - case TGSI_OPCODE_KIL: - break; - case TGSI_OPCODE_END: - instr->end(); - return; - break; - default: - fprintf(stderr, "ERROR: Unknown opcode %d\n", - inst->Instruction.Opcode); - assert(0); - break; - } - - if (!out) { - fprintf(stderr, "ERROR: unsupported opcode %d\n", - inst->Instruction.Opcode); - assert(!"Unsupported opcode"); - } - - /* # not sure if we need this */ - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - /*TXT( "_SAT" );*/ - break; - case TGSI_SAT_MINUS_PLUS_ONE: - /*TXT( "_SAT[-1,1]" );*/ - break; - default: - assert( 0 ); - } - - /* store results */ - for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else { - fprintf(stderr, "ERROR: unsupported LLVM destination!"); - assert(!"wrong destination"); - } - } -} - - -static void -translate_instructionir(llvm::Module *module, - StorageSoa *storage, - InstructionsSoa *instr, - struct tgsi_full_instruction *inst, - struct tgsi_full_instruction *fi, - unsigned instno) -{ - std::vector< std::vector > inputs(inst->Instruction.NumSrcRegs); - - for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - std::vector val; - llvm::Value *indIdx = 0; - int swizzle = swizzleInt(src); - - if (src->SrcRegister.Indirect) { - indIdx = storage->addrElement(src->SrcRegisterInd.Index); - } - if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->load(StorageSoa::Const, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->load(StorageSoa::Input, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->load(StorageSoa::Temp, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->load(StorageSoa::Output, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->load(StorageSoa::Immediate, - src->SrcRegister.Index, swizzle, indIdx); - } else { - fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); - return; - } - - inputs[i] = val; - } - - std::vector out(4); - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: { - out = instr->arl(inputs[0]); - } - break; - case TGSI_OPCODE_MOV: { - out = inputs[0]; - } - break; - case TGSI_OPCODE_LIT: { - } - break; - case TGSI_OPCODE_RCP: { - } - break; - case TGSI_OPCODE_RSQ: { - } - break; - case TGSI_OPCODE_EXP: - break; - case TGSI_OPCODE_LOG: - break; - case TGSI_OPCODE_MUL: { - out = instr->mul(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_ADD: { - out = instr->add(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP3: { - } - break; - case TGSI_OPCODE_DP4: { - } - break; - case TGSI_OPCODE_DST: { - } - break; - case TGSI_OPCODE_MIN: { - } - break; - case TGSI_OPCODE_MAX: { - } - break; - case TGSI_OPCODE_SLT: { - } - break; - case TGSI_OPCODE_SGE: { - } - break; - case TGSI_OPCODE_MAD: { - out = instr->madd(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SUB: { - } - break; - case TGSI_OPCODE_LERP: { - } - break; - case TGSI_OPCODE_CND: - break; - case TGSI_OPCODE_CND0: - break; - case TGSI_OPCODE_DOT2ADD: - break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { - } - break; - case TGSI_OPCODE_CLAMP: - break; - case TGSI_OPCODE_FLOOR: { - } - break; - case TGSI_OPCODE_ROUND: - break; - case TGSI_OPCODE_EXPBASE2: { - } - break; - case TGSI_OPCODE_LOGBASE2: { - } - break; - case TGSI_OPCODE_POWER: { - } - break; - case TGSI_OPCODE_CROSSPRODUCT: { - } - break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; - case TGSI_OPCODE_ABS: { - } - break; - case TGSI_OPCODE_RCC: - break; - case TGSI_OPCODE_DPH: { - } - break; - case TGSI_OPCODE_COS: { - } - break; - case TGSI_OPCODE_DDX: - break; - case TGSI_OPCODE_DDY: - break; - case TGSI_OPCODE_KILP: { - } - break; - case TGSI_OPCODE_PK2H: - break; - case TGSI_OPCODE_PK2US: - break; - case TGSI_OPCODE_PK4B: - break; - case TGSI_OPCODE_PK4UB: - break; - case TGSI_OPCODE_RFL: - break; - case TGSI_OPCODE_SEQ: - break; - case TGSI_OPCODE_SFL: - break; - case TGSI_OPCODE_SGT: { - } - break; - case TGSI_OPCODE_SIN: { - } - break; - case TGSI_OPCODE_SLE: - break; - case TGSI_OPCODE_SNE: - break; - case TGSI_OPCODE_STR: - break; - case TGSI_OPCODE_TEX: - break; - case TGSI_OPCODE_TXD: - break; - case TGSI_OPCODE_UP2H: - break; - case TGSI_OPCODE_UP2US: - break; - case TGSI_OPCODE_UP4B: - break; - case TGSI_OPCODE_UP4UB: - break; - case TGSI_OPCODE_X2D: - break; - case TGSI_OPCODE_ARA: - break; - case TGSI_OPCODE_ARR: - break; - case TGSI_OPCODE_BRA: - break; - case TGSI_OPCODE_CAL: { - } - break; - case TGSI_OPCODE_RET: { - } - break; - case TGSI_OPCODE_SSG: - break; - case TGSI_OPCODE_CMP: { - } - break; - case TGSI_OPCODE_SCS: { - } - break; - case TGSI_OPCODE_TXB: - break; - case TGSI_OPCODE_NRM: - break; - case TGSI_OPCODE_DIV: - break; - case TGSI_OPCODE_DP2: - break; - case TGSI_OPCODE_TXL: - break; - case TGSI_OPCODE_BRK: { - } - break; - case TGSI_OPCODE_IF: { - } - break; - case TGSI_OPCODE_LOOP: - break; - case TGSI_OPCODE_REP: - break; - case TGSI_OPCODE_ELSE: { - } - break; - case TGSI_OPCODE_ENDIF: { - } - break; - case TGSI_OPCODE_ENDLOOP: - break; - case TGSI_OPCODE_ENDREP: - break; - case TGSI_OPCODE_PUSHA: - break; - case TGSI_OPCODE_POPA: - break; - case TGSI_OPCODE_CEIL: - break; - case TGSI_OPCODE_I2F: - break; - case TGSI_OPCODE_NOT: - break; - case TGSI_OPCODE_TRUNC: { - } - break; - case TGSI_OPCODE_SHL: - break; - case TGSI_OPCODE_SHR: - break; - case TGSI_OPCODE_AND: - break; - case TGSI_OPCODE_OR: - break; - case TGSI_OPCODE_MOD: - break; - case TGSI_OPCODE_XOR: - break; - case TGSI_OPCODE_SAD: - break; - case TGSI_OPCODE_TXF: - break; - case TGSI_OPCODE_TXQ: - break; - case TGSI_OPCODE_CONT: - break; - case TGSI_OPCODE_EMIT: - break; - case TGSI_OPCODE_ENDPRIM: - break; - case TGSI_OPCODE_BGNLOOP2: { - } - break; - case TGSI_OPCODE_BGNSUB: { - } - break; - case TGSI_OPCODE_ENDLOOP2: { - } - break; - case TGSI_OPCODE_ENDSUB: { - } - break; - case TGSI_OPCODE_NOISE1: - break; - case TGSI_OPCODE_NOISE2: - break; - case TGSI_OPCODE_NOISE3: - break; - case TGSI_OPCODE_NOISE4: - break; - case TGSI_OPCODE_NOP: - break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; - case TGSI_OPCODE_NRM4: - break; - case TGSI_OPCODE_CALLNZ: - break; - case TGSI_OPCODE_IFC: - break; - case TGSI_OPCODE_BREAKC: - break; - case TGSI_OPCODE_KIL: - break; - case TGSI_OPCODE_END: - instr->end(); - return; - break; - default: - fprintf(stderr, "ERROR: Unknown opcode %d\n", - inst->Instruction.Opcode); - assert(0); - break; - } - - if (!out[0]) { - fprintf(stderr, "ERROR: unsupported opcode %d\n", - inst->Instruction.Opcode); - assert(!"Unsupported opcode"); - } - - /* store results */ - for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->store(StorageSoa::Output, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->store(StorageSoa::Temp, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->store(StorageSoa::Address, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else { - fprintf(stderr, "ERROR: unsupported LLVM destination!"); - assert(!"wrong destination"); - } - } -} - -llvm::Module * -tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) -{ - llvm::Module *mod = new Module("shader"); - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned instno = 0; - Function* shader = mod->getFunction("execute_shader"); - std::ostringstream stream; - if (ir->type == GALLIVM_VS) { - stream << "vs_shader"; - } else { - stream << "fs_shader"; - } - stream << ir->id; - std::string func_name = stream.str(); - shader->setName(func_name.c_str()); - - Function::arg_iterator args = shader->arg_begin(); - Value *ptr_INPUT = args++; - ptr_INPUT->setName("input"); - - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); - - tgsi_parse_init(&parse, tokens); - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - Storage storage(label_entry, ptr_INPUT); - Instructions instr(mod, shader, label_entry, &storage); - while(!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - translate_declaration(ir, mod, &storage, - &parse.FullToken.FullDeclaration, - &fd); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - translate_immediate(&storage, - &parse.FullToken.FullImmediate); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - translate_instruction(mod, &storage, &instr, - &parse.FullToken.FullInstruction, - &fi, instno); - ++instno; - break; - - default: - assert(0); - } - } - - tgsi_parse_free(&parse); - - ir->num_consts = storage.numConsts(); - return mod; -} - -llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, - const struct tgsi_token *tokens) -{ - llvm::Module *mod = new Module("shader"); - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned instno = 0; - std::ostringstream stream; - if (ir->type == GALLIVM_VS) { - stream << "vs_shader"; - } else { - stream << "fs_shader"; - } - //stream << ir->id; - std::string func_name = stream.str(); - Function *shader = llvm::cast(mod->getOrInsertFunction( - func_name.c_str(), - vertexShaderFunctionType())); - - Function::arg_iterator args = shader->arg_begin(); - Value *input = args++; - input->setName("inputs"); - Value *output = args++; - output->setName("outputs"); - Value *consts = args++; - consts->setName("consts"); - Value *temps = args++; - temps->setName("temps"); - - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); - - tgsi_parse_init(&parse, tokens); - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - - StorageSoa storage(label_entry, input, output, consts, temps); - InstructionsSoa instr(mod, shader, label_entry, &storage); - - while(!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - translate_declarationir(ir, mod, &storage, - &parse.FullToken.FullDeclaration, - &fd); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - translate_immediateir(&storage, - &parse.FullToken.FullImmediate); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - storage.declareImmediates(); - translate_instructionir(mod, &storage, &instr, - &parse.FullToken.FullInstruction, - &fi, instno); - ++instno; - break; - - default: - assert(0); - } - } - - tgsi_parse_free(&parse); - - return mod; -} diff --git a/src/mesa/pipe/llvm/tgsitollvm.h b/src/mesa/pipe/llvm/tgsitollvm.h deleted file mode 100644 index 7ada04d629..0000000000 --- a/src/mesa/pipe/llvm/tgsitollvm.h +++ /dev/null @@ -1,20 +0,0 @@ -#ifndef TGSITOLLVM_H -#define TGSITOLLVM_H - - -namespace llvm { - class Module; -} - -struct gallivm_ir; -struct tgsi_token; - - -llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, - const struct tgsi_token *tokens); - - -llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, - const struct tgsi_token *tokens); - -#endif diff --git a/src/mesa/pipe/p_compiler.h b/src/mesa/pipe/p_compiler.h deleted file mode 100644 index 30cd729c56..0000000000 --- a/src/mesa/pipe/p_compiler.h +++ /dev/null @@ -1,116 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef P_COMPILER_H -#define P_COMPILER_H - - -#include -#include - - -#if defined(_WIN32) && !defined(__WIN32__) -#define __WIN32__ -#endif - -#if defined(_MSC_VER) && !defined(__MSC__) -#define __MSC__ -#endif - - -typedef unsigned int uint; -typedef unsigned char ubyte; -typedef unsigned char boolean; -typedef unsigned short ushort; -typedef unsigned long long uint64; - - -#if defined(__MSC__) - -typedef char int8_t; -typedef unsigned char uint8_t; -typedef short int16_t; -typedef unsigned short uint16_t; -typedef long int32_t; -typedef unsigned long uint32_t; -typedef long long int64_t; -typedef unsigned long long uint64_t; - -#if defined(_WIN64) -typedef unsigned __int64 uintptr_t; -#else -typedef unsigned int uintptr_t; -#endif - -#else -#include -#endif - - -#define TRUE 1 -#define FALSE 0 - - -/* Function inlining */ -#if defined(__GNUC__) -# define INLINE __inline__ -#elif defined(__MSC__) -# define INLINE __inline -#elif defined(__ICL) -# define INLINE __inline -#elif defined(__INTEL_COMPILER) -# define INLINE inline -#elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100) -# define INLINE __inline -#else -# define INLINE -#endif - - -#if defined __GNUC__ -#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___aligned[SIZE] __attribute__(( aligned( 16 ) )) -#define ALIGN16_ASSIGN(NAME) NAME##___aligned -#define ALIGN16_ATTRIB __attribute__(( aligned( 16 ) )) -#else -#define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___unaligned[SIZE + 1] -#define ALIGN16_ASSIGN(NAME) align16(NAME##___unaligned) -#define ALIGN16_ATTRIB -#endif - - - -/** For calling code-gen'd functions */ -#if !defined(XSTDCALL) -#if defined(WIN32) -#define XSTDCALL __stdcall -#else -#define XSTDCALL -#endif -#endif - - -#endif /* P_COMPILER_H */ diff --git a/src/mesa/pipe/p_context.h b/src/mesa/pipe/p_context.h deleted file mode 100644 index 39f95695fb..0000000000 --- a/src/mesa/pipe/p_context.h +++ /dev/null @@ -1,221 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef PIPE_CONTEXT_H -#define PIPE_CONTEXT_H - -#include "p_state.h" - - -struct pipe_state_cache; - -/* Opaque driver handles: - */ -struct pipe_query; - -/** - * Gallium rendering context. Basically: - * - state setting functions - * - VBO drawing functions - * - surface functions - * - device queries - */ -struct pipe_context { - struct pipe_winsys *winsys; - - void *priv; /** context private data (for DRI for example) */ - - void (*destroy)( struct pipe_context * ); - - /* - * Queries - */ - /** type is one of PIPE_SURFACE, PIPE_TEXTURE, etc. */ - boolean (*is_format_supported)( struct pipe_context *pipe, - enum pipe_format format, uint type ); - - const char *(*get_name)( struct pipe_context *pipe ); - - const char *(*get_vendor)( struct pipe_context *pipe ); - - int (*get_param)( struct pipe_context *pipe, int param ); - float (*get_paramf)( struct pipe_context *pipe, int param ); - - - /* - * Drawing. - * Return false on fallbacks (temporary??) - */ - boolean (*draw_arrays)( struct pipe_context *pipe, - unsigned mode, unsigned start, unsigned count); - - boolean (*draw_elements)( struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count); - - - /** - * Query objects - */ - struct pipe_query *(*create_query)( struct pipe_context *pipe, - unsigned query_type ); - - void (*destroy_query)(struct pipe_context *pipe, - struct pipe_query *q); - - void (*begin_query)(struct pipe_context *pipe, struct pipe_query *q); - void (*end_query)(struct pipe_context *pipe, struct pipe_query *q); - - boolean (*get_query_result)(struct pipe_context *pipe, - struct pipe_query *q, - boolean wait, - uint64 *result); - - /* - * State functions - */ - void * (*create_blend_state)(struct pipe_context *, - const struct pipe_blend_state *); - void (*bind_blend_state)(struct pipe_context *, void *); - void (*delete_blend_state)(struct pipe_context *, void *); - - void * (*create_sampler_state)(struct pipe_context *, - const struct pipe_sampler_state *); - void (*bind_sampler_state)(struct pipe_context *, unsigned unit, void *); - void (*delete_sampler_state)(struct pipe_context *, void *); - - void * (*create_rasterizer_state)(struct pipe_context *, - const struct pipe_rasterizer_state *); - void (*bind_rasterizer_state)(struct pipe_context *, void *); - void (*delete_rasterizer_state)(struct pipe_context *, void *); - - void * (*create_depth_stencil_alpha_state)(struct pipe_context *, - const struct pipe_depth_stencil_alpha_state *); - void (*bind_depth_stencil_alpha_state)(struct pipe_context *, void *); - void (*delete_depth_stencil_alpha_state)(struct pipe_context *, void *); - - void * (*create_fs_state)(struct pipe_context *, - const struct pipe_shader_state *); - void (*bind_fs_state)(struct pipe_context *, void *); - void (*delete_fs_state)(struct pipe_context *, void *); - - void * (*create_vs_state)(struct pipe_context *, - const struct pipe_shader_state *); - void (*bind_vs_state)(struct pipe_context *, void *); - void (*delete_vs_state)(struct pipe_context *, void *); - - /* The following look more properties than states. - * maybe combine a few of them into states or pass them - * in the bind calls to the state */ - void (*set_blend_color)( struct pipe_context *, - const struct pipe_blend_color * ); - - void (*set_clip_state)( struct pipe_context *, - const struct pipe_clip_state * ); - - void (*set_constant_buffer)( struct pipe_context *, - uint shader, uint index, - const struct pipe_constant_buffer *buf ); - - void (*set_framebuffer_state)( struct pipe_context *, - const struct pipe_framebuffer_state * ); - - void (*set_polygon_stipple)( struct pipe_context *, - const struct pipe_poly_stipple * ); - - void (*set_scissor_state)( struct pipe_context *, - const struct pipe_scissor_state * ); - - - /* Currently a sampler is constrained to sample from a single texture: - */ - void (*set_sampler_texture)( struct pipe_context *, - unsigned sampler, - struct pipe_texture * ); - - void (*set_viewport_state)( struct pipe_context *, - const struct pipe_viewport_state * ); - - /* - * Vertex arrays - */ - void (*set_vertex_buffer)( struct pipe_context *, - unsigned index, - const struct pipe_vertex_buffer * ); - - void (*set_vertex_element)( struct pipe_context *, - unsigned index, - const struct pipe_vertex_element * ); - - - /* - * Surface functions - */ - - void (*surface_copy)(struct pipe_context *pipe, - unsigned do_flip, /*<< flip surface contents vertically */ - struct pipe_surface *dest, - unsigned destx, unsigned desty, - struct pipe_surface *src, /* don't make this const - - need to map/unmap */ - unsigned srcx, unsigned srcy, - unsigned width, unsigned height); - - void (*surface_fill)(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - unsigned width, unsigned height, - unsigned value); - - void (*clear)(struct pipe_context *pipe, - struct pipe_surface *ps, - unsigned clearValue); - - - /* - * Texture functions - */ - struct pipe_texture * (*texture_create)(struct pipe_context *pipe, - const struct pipe_texture *templat); - - void (*texture_release)(struct pipe_context *pipe, - struct pipe_texture **pt); - - /** Get a surface which is a "view" into a texture */ - struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, - struct pipe_texture *texture, - unsigned face, unsigned level, - unsigned zslice); - - /* Flush rendering: - */ - void (*flush)( struct pipe_context *pipe, - unsigned flags ); -}; - -#endif /* PIPE_CONTEXT_H */ diff --git a/src/mesa/pipe/p_debug.h b/src/mesa/pipe/p_debug.h deleted file mode 100644 index 2a11627b36..0000000000 --- a/src/mesa/pipe/p_debug.h +++ /dev/null @@ -1,86 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * @file - * Cross-platform debugging helpers. - * - * For now it just has assert and printf replacements, but it might be extended - * with stack trace reports and more advanced logging in the near future. - * - * @author Jose Fonseca - */ - -#ifndef P_DEBUG_H_ -#define P_DEBUG_H_ - - -#include - - -#ifdef __cplusplus -extern "C" { -#endif - - -#ifdef DBG -#ifndef DEBUG -#define DEBUG 1 -#endif -#else -#ifndef NDEBUG -#define NDEBUG 1 -#endif -#endif - - -void debug_printf(const char *format, ...); - -void debug_vprintf(const char *format, va_list ap); - -void debug_assert_fail(const char *expr, const char *file, unsigned line); - - -/** Assert macro */ -#ifdef DEBUG -#define debug_assert(expr) ((expr) ? (void)0 : debug_assert_fail(#expr, __FILE__, __LINE__)) -#else -#define debug_assert(expr) ((void)0) -#endif - - -#ifdef assert -#undef assert -#endif -#define assert(expr) debug_assert(expr) - - -#ifdef __cplusplus -} -#endif - -#endif /* P_DEBUG_H_ */ diff --git a/src/mesa/pipe/p_defines.h b/src/mesa/pipe/p_defines.h deleted file mode 100644 index 0bf53ecb79..0000000000 --- a/src/mesa/pipe/p_defines.h +++ /dev/null @@ -1,270 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef PIPE_DEFINES_H -#define PIPE_DEFINES_H - -#include "p_format.h" - -#define PIPE_BLENDFACTOR_ONE 0x1 -#define PIPE_BLENDFACTOR_SRC_COLOR 0x2 -#define PIPE_BLENDFACTOR_SRC_ALPHA 0x3 -#define PIPE_BLENDFACTOR_DST_ALPHA 0x4 -#define PIPE_BLENDFACTOR_DST_COLOR 0x5 -#define PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE 0x6 -#define PIPE_BLENDFACTOR_CONST_COLOR 0x7 -#define PIPE_BLENDFACTOR_CONST_ALPHA 0x8 -#define PIPE_BLENDFACTOR_SRC1_COLOR 0x9 -#define PIPE_BLENDFACTOR_SRC1_ALPHA 0x0A -#define PIPE_BLENDFACTOR_ZERO 0x11 -#define PIPE_BLENDFACTOR_INV_SRC_COLOR 0x12 -#define PIPE_BLENDFACTOR_INV_SRC_ALPHA 0x13 -#define PIPE_BLENDFACTOR_INV_DST_ALPHA 0x14 -#define PIPE_BLENDFACTOR_INV_DST_COLOR 0x15 -#define PIPE_BLENDFACTOR_INV_CONST_COLOR 0x17 -#define PIPE_BLENDFACTOR_INV_CONST_ALPHA 0x18 -#define PIPE_BLENDFACTOR_INV_SRC1_COLOR 0x19 -#define PIPE_BLENDFACTOR_INV_SRC1_ALPHA 0x1A - -#define PIPE_BLEND_ADD 0 -#define PIPE_BLEND_SUBTRACT 1 -#define PIPE_BLEND_REVERSE_SUBTRACT 2 -#define PIPE_BLEND_MIN 3 -#define PIPE_BLEND_MAX 4 - -#define PIPE_LOGICOP_CLEAR 0 -#define PIPE_LOGICOP_NOR 1 -#define PIPE_LOGICOP_AND_INVERTED 2 -#define PIPE_LOGICOP_COPY_INVERTED 3 -#define PIPE_LOGICOP_AND_REVERSE 4 -#define PIPE_LOGICOP_INVERT 5 -#define PIPE_LOGICOP_XOR 6 -#define PIPE_LOGICOP_NAND 7 -#define PIPE_LOGICOP_AND 8 -#define PIPE_LOGICOP_EQUIV 9 -#define PIPE_LOGICOP_NOOP 10 -#define PIPE_LOGICOP_OR_INVERTED 11 -#define PIPE_LOGICOP_COPY 12 -#define PIPE_LOGICOP_OR_REVERSE 13 -#define PIPE_LOGICOP_OR 14 -#define PIPE_LOGICOP_SET 15 - -#define PIPE_MASK_R 0x1 -#define PIPE_MASK_G 0x2 -#define PIPE_MASK_B 0x4 -#define PIPE_MASK_A 0x8 -#define PIPE_MASK_RGBA 0xf - - -/** - * Inequality functions. Used for depth test, stencil compare, alpha - * test, shadow compare, etc. - */ -#define PIPE_FUNC_NEVER 0 -#define PIPE_FUNC_LESS 1 -#define PIPE_FUNC_EQUAL 2 -#define PIPE_FUNC_LEQUAL 3 -#define PIPE_FUNC_GREATER 4 -#define PIPE_FUNC_NOTEQUAL 5 -#define PIPE_FUNC_GEQUAL 6 -#define PIPE_FUNC_ALWAYS 7 - -/** Polygon fill mode */ -#define PIPE_POLYGON_MODE_FILL 0 -#define PIPE_POLYGON_MODE_LINE 1 -#define PIPE_POLYGON_MODE_POINT 2 - -/** Polygon front/back window, also for culling */ -#define PIPE_WINDING_NONE 0 -#define PIPE_WINDING_CW 1 -#define PIPE_WINDING_CCW 2 -#define PIPE_WINDING_BOTH (PIPE_WINDING_CW | PIPE_WINDING_CCW) - -/** Stencil ops */ -#define PIPE_STENCIL_OP_KEEP 0 -#define PIPE_STENCIL_OP_ZERO 1 -#define PIPE_STENCIL_OP_REPLACE 2 -#define PIPE_STENCIL_OP_INCR 3 -#define PIPE_STENCIL_OP_DECR 4 -#define PIPE_STENCIL_OP_INCR_WRAP 5 -#define PIPE_STENCIL_OP_DECR_WRAP 6 -#define PIPE_STENCIL_OP_INVERT 7 - -/** Texture types */ -enum pipe_texture_target { - PIPE_TEXTURE_1D = 0, - PIPE_TEXTURE_2D = 1, - PIPE_TEXTURE_3D = 2, - PIPE_TEXTURE_CUBE = 3 -}; - -#define PIPE_TEX_FACE_POS_X 0 -#define PIPE_TEX_FACE_NEG_X 1 -#define PIPE_TEX_FACE_POS_Y 2 -#define PIPE_TEX_FACE_NEG_Y 3 -#define PIPE_TEX_FACE_POS_Z 4 -#define PIPE_TEX_FACE_NEG_Z 5 - -#define PIPE_TEX_WRAP_REPEAT 0 -#define PIPE_TEX_WRAP_CLAMP 1 -#define PIPE_TEX_WRAP_CLAMP_TO_EDGE 2 -#define PIPE_TEX_WRAP_CLAMP_TO_BORDER 3 -#define PIPE_TEX_WRAP_MIRROR_REPEAT 4 -#define PIPE_TEX_WRAP_MIRROR_CLAMP 5 -#define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE 6 -#define PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER 7 - -/* Between mipmaps, ie mipfilter - */ -#define PIPE_TEX_MIPFILTER_NEAREST 0 -#define PIPE_TEX_MIPFILTER_LINEAR 1 -#define PIPE_TEX_MIPFILTER_NONE 2 - -/* Within a mipmap, ie min/mag filter - */ -#define PIPE_TEX_FILTER_NEAREST 0 -#define PIPE_TEX_FILTER_LINEAR 1 -//#define PIPE_TEX_FILTER_ANISO 2 - - -#define PIPE_TEX_COMPARE_NONE 0 -#define PIPE_TEX_COMPARE_R_TO_TEXTURE 1 - -#define PIPE_TEX_FACE_POS_X 0 -#define PIPE_TEX_FACE_NEG_X 1 -#define PIPE_TEX_FACE_POS_Y 2 -#define PIPE_TEX_FACE_NEG_Y 3 -#define PIPE_TEX_FACE_POS_Z 4 -#define PIPE_TEX_FACE_NEG_Z 5 -#define PIPE_TEX_FACE_MAX 6 - -/** - * Surfaces, textures, etc. (others may be added) - */ -#define PIPE_TEXTURE 1 -#define PIPE_SURFACE 2 /**< user-created surfaces */ - - -/** - * Surface status - */ -#define PIPE_SURFACE_STATUS_UNDEFINED 0 -#define PIPE_SURFACE_STATUS_DEFINED 1 -#define PIPE_SURFACE_STATUS_CLEAR 2 - - -/** - * Buffer usage flags - */ -#define PIPE_BUFFER_USAGE_CPU_READ (1 << 0) -#define PIPE_BUFFER_USAGE_CPU_WRITE (1 << 1) -#define PIPE_BUFFER_USAGE_GPU_READ (1 << 2) -#define PIPE_BUFFER_USAGE_GPU_WRITE (1 << 3) -#define PIPE_BUFFER_USAGE_PIXEL (1 << 4) -#define PIPE_BUFFER_USAGE_VERTEX (1 << 5) -#define PIPE_BUFFER_USAGE_INDEX (1 << 6) -#define PIPE_BUFFER_USAGE_CONSTANT (1 << 7) -/** Pipe driver custam usage flags should be greater or equal to this value */ -#define PIPE_BUFFER_USAGE_CUSTOM (1 << 16) - - -/** - * Flush types: - */ -#define PIPE_FLUSH_RENDER_CACHE 0x1 -#define PIPE_FLUSH_TEXTURE_CACHE 0x2 -#define PIPE_FLUSH_WAIT 0x4 -#define PIPE_FLUSH_SWAPBUFFERS 0x8 - - -/** - * Shaders - */ -#define PIPE_SHADER_VERTEX 0 -#define PIPE_SHADER_FRAGMENT 1 -#define PIPE_SHADER_TYPES 2 - - -/** - * Primitive types: - */ -#define PIPE_PRIM_POINTS 0 -#define PIPE_PRIM_LINES 1 -#define PIPE_PRIM_LINE_LOOP 2 -#define PIPE_PRIM_LINE_STRIP 3 -#define PIPE_PRIM_TRIANGLES 4 -#define PIPE_PRIM_TRIANGLE_STRIP 5 -#define PIPE_PRIM_TRIANGLE_FAN 6 -#define PIPE_PRIM_QUADS 7 -#define PIPE_PRIM_QUAD_STRIP 8 -#define PIPE_PRIM_POLYGON 9 - - -/** - * Query object types - */ -#define PIPE_QUERY_OCCLUSION_COUNTER 0 -#define PIPE_QUERY_PRIMITIVES_GENERATED 1 -#define PIPE_QUERY_PRIMITIVES_EMITTED 2 -#define PIPE_QUERY_TYPES 3 - - -/** - * Point sprite coord modes - */ -#define PIPE_SPRITE_COORD_NONE 0 -#define PIPE_SPRITE_COORD_UPPER_LEFT 1 -#define PIPE_SPRITE_COORD_LOWER_LEFT 2 - - -/** - * Implementation capabilities/limits - * Passed to pipe->get_param() - * XXX this will need some fine tuning... - */ -#define PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS 1 -#define PIPE_CAP_NPOT_TEXTURES 2 -#define PIPE_CAP_TWO_SIDED_STENCIL 3 -#define PIPE_CAP_GLSL 4 /* XXX need something better */ -#define PIPE_CAP_S3TC 5 -#define PIPE_CAP_ANISOTROPIC_FILTER 6 -#define PIPE_CAP_POINT_SPRITE 7 -#define PIPE_CAP_MAX_RENDER_TARGETS 8 -#define PIPE_CAP_OCCLUSION_QUERY 9 -#define PIPE_CAP_TEXTURE_SHADOW_MAP 10 -#define PIPE_CAP_MAX_TEXTURE_2D_LEVELS 11 -#define PIPE_CAP_MAX_TEXTURE_3D_LEVELS 12 -#define PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS 13 -#define PIPE_CAP_MAX_LINE_WIDTH 14 -#define PIPE_CAP_MAX_LINE_WIDTH_AA 15 -#define PIPE_CAP_MAX_POINT_WIDTH 16 -#define PIPE_CAP_MAX_POINT_WIDTH_AA 17 -#define PIPE_CAP_MAX_TEXTURE_ANISOTROPY 18 -#define PIPE_CAP_MAX_TEXTURE_LOD_BIAS 19 -#define PIPE_CAP_BITMAP_TEXCOORD_BIAS 20 - -#endif diff --git a/src/mesa/pipe/p_format.h b/src/mesa/pipe/p_format.h deleted file mode 100644 index c9ad324315..0000000000 --- a/src/mesa/pipe/p_format.h +++ /dev/null @@ -1,421 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef PIPE_FORMAT_H -#define PIPE_FORMAT_H - -#include // for sprintf - -#include "p_compiler.h" -#include "p_debug.h" - -/** - * The PIPE_FORMAT is a 32-bit wide bitfield that encodes all the information - * needed to uniquely describe a pixel format. - */ - -/** - * Possible format layouts -- occupy first 2 bits. The interpretation of - * the remaining 30 bits depends on a particual format layout. - */ -#define PIPE_FORMAT_LAYOUT_RGBAZS 0 -#define PIPE_FORMAT_LAYOUT_YCBCR 1 - -static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ -{ - return f & 0x3; -} - -/** - * RGBAZS Format Layout. - */ - -/** - * Format component selectors for RGBAZS layout. - */ -#define PIPE_FORMAT_COMP_R 0 -#define PIPE_FORMAT_COMP_G 1 -#define PIPE_FORMAT_COMP_B 2 -#define PIPE_FORMAT_COMP_A 3 -#define PIPE_FORMAT_COMP_0 4 -#define PIPE_FORMAT_COMP_1 5 -#define PIPE_FORMAT_COMP_Z 6 -#define PIPE_FORMAT_COMP_S 7 - -/** - * Format types for RGBAZS layout. - */ -#define PIPE_FORMAT_TYPE_UNKNOWN 0 -#define PIPE_FORMAT_TYPE_FLOAT 1 -#define PIPE_FORMAT_TYPE_UNORM 2 -#define PIPE_FORMAT_TYPE_SNORM 3 -#define PIPE_FORMAT_TYPE_USCALED 4 -#define PIPE_FORMAT_TYPE_SSCALED 5 - -/** - * Because the destination vector is assumed to be RGBA FLOAT, we - * need to know how to swizzle and expand components from the source - * vector. - * Let's take U_A1_R5_G5_B5 as an example. X swizzle is A, X size - * is 1 bit and type is UNORM. So we take the most significant bit - * from source vector, convert 0 to 0.0 and 1 to 1.0 and save it - * in the last component of the destination RGBA component. - * Next, Y swizzle is R, Y size is 5 and type is UNORM. We normalize - * those 5 bits into [0.0; 1.0] range and put it into second - * component of the destination vector. Rinse and repeat for - * components Z and W. - * If any of size fields is zero, it means the source format contains - * less than four components. - * If any swizzle is 0 or 1, the corresponding destination component - * should be filled with 0.0 and 1.0, respectively. - */ -typedef uint pipe_format_rgbazs_t; - -static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) -{ - return (f >> shift) & mask; -} - -/* XXX: The bit layout needs to be revised, can't currently encode 10-bit components. */ - -#define pf_swizzle_x(f) pf_get(f, 2, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_y(f) pf_get(f, 5, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_z(f) pf_get(f, 8, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_w(f) pf_get(f, 11, 0x7) /**< PIPE_FORMAT_COMP_ */ -#define pf_swizzle_xyzw(f,i) pf_get(f, 2+((i)*3), 0x7) -#define pf_size_x(f) pf_get(f, 14, 0x7) /**< Size of X */ -#define pf_size_y(f) pf_get(f, 17, 0x7) /**< Size of Y */ -#define pf_size_z(f) pf_get(f, 20, 0x7) /**< Size of Z */ -#define pf_size_w(f) pf_get(f, 23, 0x7) /**< Size of W */ -#define pf_size_xyzw(f,i) pf_get(f, 14+((i)*3), 0x7) -#define pf_exp8(f) pf_get(f, 26, 0x3) /**< Scale size by 8 ^ exp8 */ -#define pf_type(f) pf_get(f, 28, 0xf) /**< PIPE_FORMAT_TYPE_ */ - -/** - * Helper macro to encode the above structure into a 32-bit value. - */ -#define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP8, TYPE ) (\ - (PIPE_FORMAT_LAYOUT_RGBAZS << 0) |\ - ((SWZ) << 2) |\ - ((SIZEX) << 14) |\ - ((SIZEY) << 17) |\ - ((SIZEZ) << 20) |\ - ((SIZEW) << 23) |\ - ((EXP8) << 26) |\ - ((TYPE) << 28) ) - -/** - * Helper macro to encode the swizzle part of the structure above. - */ -#define _PIPE_FORMAT_SWZ( SWZX, SWZY, SWZZ, SWZW ) (((SWZX) << 0) | ((SWZY) << 3) | ((SWZZ) << 6) | ((SWZW) << 9)) - -/** - * Shorthand macro for RGBAZS layout with component sizes in 1-bit units. - */ -#define _PIPE_FORMAT_RGBAZS_1( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE ) - -/** - * Shorthand macro for RGBAZS layout with component sizes in 8-bit units. - */ -#define _PIPE_FORMAT_RGBAZS_8( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) - -/** - * Shorthand macro for RGBAZS layout with component sizes in 64-bit units. - */ -#define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 2, TYPE ) - -/** - * Shorthand macro for common format swizzles. - */ -#define _PIPE_FORMAT_R000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_RG00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) -#define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) -#define _PIPE_FORMAT_BGRA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_A ) -#define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R ) -#define _PIPE_FORMAT_RRR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) -#define _PIPE_FORMAT_RRRR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R ) -#define _PIPE_FORMAT_RRRG _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G ) -#define _PIPE_FORMAT_Z000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_ZS00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) - -/** - * YCBCR Format Layout. - */ - -/** - * This only contains a flag that indicates whether the format is reversed or - * not. - */ -typedef uint pipe_format_ycbcr_t; - -/** - * Helper macro to encode the above structure into a 32-bit value. - */ -#define _PIPE_FORMAT_YCBCR( REV ) (\ - (PIPE_FORMAT_LAYOUT_YCBCR << 0) |\ - ((REV) << 2) ) - -static INLINE uint pf_rev(pipe_format_ycbcr_t f) -{ - return (f >> 2) & 0x1; -} - -/** - * Texture/surface image formats (preliminary) - */ - -/* KW: Added lots of surface formats to support vertex element layout - * definitions, and eventually render-to-vertex-buffer. Could - * consider making float/int/uint/scaled/normalized a separate - * parameter, but on the other hand there are special cases like - * z24s8, compressed textures, ycbcr, etc that won't fit that model. - */ - -enum pipe_format { - PIPE_FORMAT_NONE = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_0000, 0, 0, 0, 0, PIPE_FORMAT_TYPE_UNKNOWN ), - PIPE_FORMAT_A8R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_B8G8R8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_A1R5G5B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_A4R4G4B4_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_U_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ - PIPE_FORMAT_U_A8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ - PIPE_FORMAT_U_I8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ - PIPE_FORMAT_U_A8_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha, luminance */ - PIPE_FORMAT_YCBCR = _PIPE_FORMAT_YCBCR( 0 ), - PIPE_FORMAT_YCBCR_REV = _PIPE_FORMAT_YCBCR( 1 ), - PIPE_FORMAT_Z16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_Z32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_Z32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_S8Z24_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_Z24S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte stencil */ - PIPE_FORMAT_R64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R64G64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R64G64B64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R64G64B64A64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32G32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32G32B32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32G32B32A32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32G32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32G32B32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32G32B32A32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32G32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32G32B32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32G32B32A32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32G32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32G32B32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32G32B32A32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R32G32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R32G32B32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R32G32B32A32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16G16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16G16B16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16G16B16A16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16G16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16G16B16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16G16B16A16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16G16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16G16B16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16G16B16A16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16G16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16G16B16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16G16B16A16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8G8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8G8B8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8G8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8G8B8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8G8B8A8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8G8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8G8B8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8G8B8A8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8B8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8B8A8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) -}; - - -/** - * XXX should remove this, but S8_UNORM is a poor name - */ -#define PIPE_FORMAT_U_S8 PIPE_FORMAT_S8_UNORM - - -/** - * Builds pipe format name from format token. - */ -static INLINE char *pf_sprint_name( char *str, uint format ) -{ - strcpy( str, "PIPE_FORMAT_" ); - switch (pf_layout( format )) { - case PIPE_FORMAT_LAYOUT_RGBAZS: { - pipe_format_rgbazs_t rgbazs = (pipe_format_rgbazs_t) format; - uint i; - uint scale = 1 << (pf_exp8( rgbazs ) * 3); - - for (i = 0; i < 4; i++) { - uint size = pf_size_xyzw( rgbazs, i ); - - if (size == 0) { - break; - } - switch (pf_swizzle_xyzw( rgbazs, i )) { - case PIPE_FORMAT_COMP_R: - strcat( str, "R" ); - break; - case PIPE_FORMAT_COMP_G: - strcat( str, "G" ); - break; - case PIPE_FORMAT_COMP_B: - strcat( str, "B" ); - break; - case PIPE_FORMAT_COMP_A: - strcat( str, "A" ); - break; - case PIPE_FORMAT_COMP_0: - strcat( str, "0" ); - break; - case PIPE_FORMAT_COMP_1: - strcat( str, "1" ); - break; - case PIPE_FORMAT_COMP_Z: - strcat( str, "Z" ); - break; - case PIPE_FORMAT_COMP_S: - strcat( str, "S" ); - break; - } - sprintf( &str[strlen( str )], "%u", size * scale ); - } - if (i != 0) { - strcat( str, "_" ); - } - switch (pf_type( rgbazs )) { - case PIPE_FORMAT_TYPE_UNKNOWN: - strcat( str, "NONE" ); - break; - case PIPE_FORMAT_TYPE_FLOAT: - strcat( str, "FLOAT" ); - break; - case PIPE_FORMAT_TYPE_UNORM: - strcat( str, "UNORM" ); - break; - case PIPE_FORMAT_TYPE_SNORM: - strcat( str, "SNORM" ); - break; - case PIPE_FORMAT_TYPE_USCALED: - strcat( str, "USCALED" ); - break; - case PIPE_FORMAT_TYPE_SSCALED: - strcat( str, "SSCALED" ); - break; - } - } - break; - case PIPE_FORMAT_LAYOUT_YCBCR: { - pipe_format_ycbcr_t ycbcr = (pipe_format_ycbcr_t) format; - - strcat( str, "YCBCR" ); - if (pf_rev( ycbcr )) { - strcat( str, "_REV" ); - } - } - break; - } - return str; -} - -static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) -{ - uint size; - - if (pf_swizzle_x(format) == comp) { - size = pf_size_x(format); - } - else if (pf_swizzle_y(format) == comp) { - size = pf_size_y(format); - } - else if (pf_swizzle_z(format) == comp) { - size = pf_size_z(format); - } - else if (pf_swizzle_w(format) == comp) { - size = pf_size_w(format); - } - else { - size = 0; - } - return size << (pf_exp8(format) * 3); -} - -static INLINE uint pf_get_bits( enum pipe_format format ) -{ - if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { - return - pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_S ); - } - else { - assert( pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR ); - - /* TODO */ - assert( 0 ); - return 0; - } -} - -static INLINE uint pf_get_size( enum pipe_format format ) { - assert(pf_get_bits(format) % 8 == 0); - return pf_get_bits(format) / 8; -} - -#endif diff --git a/src/mesa/pipe/p_inlines.h b/src/mesa/pipe/p_inlines.h deleted file mode 100644 index ebf6ed86bc..0000000000 --- a/src/mesa/pipe/p_inlines.h +++ /dev/null @@ -1,112 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef P_INLINES_H -#define P_INLINES_H - -#include "p_context.h" -#include "p_defines.h" -#include "p_winsys.h" - - -static INLINE void * -pipe_surface_map(struct pipe_surface *surface) -{ - return (char *)surface->winsys->buffer_map( surface->winsys, surface->buffer, - PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_CPU_READ ) - + surface->offset; -} - -static INLINE void -pipe_surface_unmap(struct pipe_surface *surface) -{ - surface->winsys->buffer_unmap( surface->winsys, surface->buffer ); -} - -/** - * Set 'ptr' to point to 'surf' and update reference counting. - * The old thing pointed to, if any, will be unreferenced first. - * 'surf' may be NULL. - */ -static INLINE void -pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) -{ - /* bump the refcount first */ - if (surf) - surf->refcount++; - - if (*ptr /* && --(*ptr)->refcount == 0 */) { - struct pipe_winsys *winsys = (*ptr)->winsys; - winsys->surface_release(winsys, ptr); - assert(!*ptr); - } - - *ptr = surf; -} - - -/* XXX: thread safety issues! - */ -static INLINE void -pipe_buffer_reference(struct pipe_winsys *winsys, - struct pipe_buffer **ptr, - struct pipe_buffer *buf) -{ - if (buf) - buf->refcount++; - - if (*ptr && --(*ptr)->refcount == 0) - winsys->buffer_destroy( winsys, *ptr ); - - *ptr = buf; -} - - - -/** - * \sa pipe_surface_reference - */ -static INLINE void -pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, - struct pipe_texture *pt) -{ - assert(ptr); - - if (pt) - pt->refcount++; - - if (*ptr) { - pipe->texture_release(pipe, ptr); - assert(!*ptr); - } - - *ptr = pt; -} - - -#endif /* P_INLINES_H */ diff --git a/src/mesa/pipe/p_shader_tokens.h b/src/mesa/pipe/p_shader_tokens.h deleted file mode 100644 index 3ce35310f6..0000000000 --- a/src/mesa/pipe/p_shader_tokens.h +++ /dev/null @@ -1,806 +0,0 @@ -#if !defined TGSI_TOKEN_H -#define TGSI_TOKEN_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_version -{ - unsigned MajorVersion : 8; - unsigned MinorVersion : 8; - unsigned Padding : 16; -}; - -struct tgsi_header -{ - unsigned HeaderSize : 8; - unsigned BodySize : 24; -}; - -#define TGSI_PROCESSOR_FRAGMENT 0 -#define TGSI_PROCESSOR_VERTEX 1 -#define TGSI_PROCESSOR_GEOMETRY 2 - -struct tgsi_processor -{ - unsigned Processor : 4; /* TGSI_PROCESSOR_ */ - unsigned Padding : 28; -}; - -#define TGSI_TOKEN_TYPE_DECLARATION 0 -#define TGSI_TOKEN_TYPE_IMMEDIATE 1 -#define TGSI_TOKEN_TYPE_INSTRUCTION 2 - -struct tgsi_token -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_ */ - unsigned Size : 8; /* UINT */ - unsigned Padding : 19; - unsigned Extended : 1; /* BOOL */ -}; - -#define TGSI_FILE_NULL 0 -#define TGSI_FILE_CONSTANT 1 -#define TGSI_FILE_INPUT 2 -#define TGSI_FILE_OUTPUT 3 -#define TGSI_FILE_TEMPORARY 4 -#define TGSI_FILE_SAMPLER 5 -#define TGSI_FILE_ADDRESS 6 -#define TGSI_FILE_IMMEDIATE 7 - -#define TGSI_DECLARE_RANGE 0 -#define TGSI_DECLARE_MASK 1 - -#define TGSI_WRITEMASK_NONE 0x00 -#define TGSI_WRITEMASK_X 0x01 -#define TGSI_WRITEMASK_Y 0x02 -#define TGSI_WRITEMASK_XY 0x03 -#define TGSI_WRITEMASK_Z 0x04 -#define TGSI_WRITEMASK_XZ 0x05 -#define TGSI_WRITEMASK_YZ 0x06 -#define TGSI_WRITEMASK_XYZ 0x07 -#define TGSI_WRITEMASK_W 0x08 -#define TGSI_WRITEMASK_XW 0x09 -#define TGSI_WRITEMASK_YW 0x0A -#define TGSI_WRITEMASK_XYW 0x0B -#define TGSI_WRITEMASK_ZW 0x0C -#define TGSI_WRITEMASK_XZW 0x0D -#define TGSI_WRITEMASK_YZW 0x0E -#define TGSI_WRITEMASK_XYZW 0x0F - -struct tgsi_declaration -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ - unsigned Size : 8; /* UINT */ - unsigned File : 4; /* one of TGSI_FILE_x */ - unsigned Declare : 4; /* one of TGSI_DECLARE_x */ - unsigned UsageMask : 4; /* bitmask of TGSI_WRITEMASK_x flags */ - unsigned Interpolate : 1; /* BOOL, any interpolation info? */ - unsigned Semantic : 1; /* BOOL, any semantic info? */ - unsigned Padding : 5; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_declaration_range -{ - unsigned First : 16; /* UINT */ - unsigned Last : 16; /* UINT */ -}; - -struct tgsi_declaration_mask -{ - unsigned Mask : 32; /* UINT */ -}; - -#define TGSI_INTERPOLATE_CONSTANT 0 -#define TGSI_INTERPOLATE_LINEAR 1 -#define TGSI_INTERPOLATE_PERSPECTIVE 2 - -struct tgsi_declaration_interpolation -{ - unsigned Interpolate : 4; /* TGSI_INTERPOLATE_ */ - unsigned Padding : 28; -}; - -#define TGSI_SEMANTIC_POSITION 0 -#define TGSI_SEMANTIC_COLOR 1 -#define TGSI_SEMANTIC_BCOLOR 2 /**< back-face color */ -#define TGSI_SEMANTIC_FOG 3 -#define TGSI_SEMANTIC_PSIZE 4 -#define TGSI_SEMANTIC_GENERIC 5 -#define TGSI_SEMANTIC_NORMAL 6 -#define TGSI_SEMANTIC_COUNT 7 /**< number of semantic values */ - -struct tgsi_declaration_semantic -{ - unsigned SemanticName : 8; /* one of TGSI_SEMANTIC_ */ - unsigned SemanticIndex : 16; /* UINT */ - unsigned Padding : 8; -}; - -#define TGSI_IMM_FLOAT32 0 - -struct tgsi_immediate -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_IMMEDIATE */ - unsigned Size : 8; /* UINT */ - unsigned DataType : 4; /* TGSI_IMM_ */ - unsigned Padding : 15; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_immediate_float32 -{ - float Float; -}; - -/* - * GL_NV_vertex_program - */ -#define TGSI_OPCODE_ARL 0 -#define TGSI_OPCODE_MOV 1 -#define TGSI_OPCODE_LIT 2 -#define TGSI_OPCODE_RCP 3 -#define TGSI_OPCODE_RSQ 4 -#define TGSI_OPCODE_EXP 5 -#define TGSI_OPCODE_LOG 6 -#define TGSI_OPCODE_MUL 7 -#define TGSI_OPCODE_ADD 8 -#define TGSI_OPCODE_DP3 9 -#define TGSI_OPCODE_DP4 10 -#define TGSI_OPCODE_DST 11 -#define TGSI_OPCODE_MIN 12 -#define TGSI_OPCODE_MAX 13 -#define TGSI_OPCODE_SLT 14 -#define TGSI_OPCODE_SGE 15 -#define TGSI_OPCODE_MAD 16 - -/* - * GL_ATI_fragment_shader - */ -#define TGSI_OPCODE_SUB 17 -#define TGSI_OPCODE_DOT3 TGSI_OPCODE_DP3 -#define TGSI_OPCODE_DOT4 TGSI_OPCODE_DP4 -#define TGSI_OPCODE_LERP 18 -#define TGSI_OPCODE_CND 19 -#define TGSI_OPCODE_CND0 20 -#define TGSI_OPCODE_DOT2ADD 21 - -/* - * GL_EXT_vertex_shader - */ -#define TGSI_OPCODE_INDEX 22 -#define TGSI_OPCODE_NEGATE 23 -#define TGSI_OPCODE_MADD TGSI_OPCODE_MAD -#define TGSI_OPCODE_FRAC 24 -#define TGSI_OPCODE_SETGE TGSI_OPCODE_SGE -#define TGSI_OPCODE_SETLT TGSI_OPCODE_SLT -#define TGSI_OPCODE_CLAMP 25 -#define TGSI_OPCODE_FLOOR 26 -#define TGSI_OPCODE_ROUND 27 -#define TGSI_OPCODE_EXPBASE2 28 -#define TGSI_OPCODE_LOGBASE2 29 -#define TGSI_OPCODE_POWER 30 -#define TGSI_OPCODE_RECIP TGSI_OPCODE_RCP -#define TGSI_OPCODE_RECIPSQRT TGSI_OPCODE_RSQ -#define TGSI_OPCODE_CROSSPRODUCT 31 -#define TGSI_OPCODE_MULTIPLYMATRIX 32 - -/* - * GL_NV_vertex_program1_1 - */ -#define TGSI_OPCODE_ABS 33 -#define TGSI_OPCODE_RCC 34 -#define TGSI_OPCODE_DPH 35 - -/* - * GL_NV_fragment_program - */ -#define TGSI_OPCODE_COS 36 -#define TGSI_OPCODE_DDX 37 -#define TGSI_OPCODE_DDY 38 -#define TGSI_OPCODE_EX2 TGSI_OPCODE_EXPBASE2 -#define TGSI_OPCODE_FLR TGSI_OPCODE_FLOOR -#define TGSI_OPCODE_FRC TGSI_OPCODE_FRAC -#define TGSI_OPCODE_KILP 39 /* predicated kill */ -#define TGSI_OPCODE_LG2 TGSI_OPCODE_LOGBASE2 -#define TGSI_OPCODE_LRP TGSI_OPCODE_LERP -#define TGSI_OPCODE_PK2H 40 -#define TGSI_OPCODE_PK2US 41 -#define TGSI_OPCODE_PK4B 42 -#define TGSI_OPCODE_PK4UB 43 -#define TGSI_OPCODE_POW TGSI_OPCODE_POWER -#define TGSI_OPCODE_RFL 44 -#define TGSI_OPCODE_SEQ 45 -#define TGSI_OPCODE_SFL 46 -#define TGSI_OPCODE_SGT 47 -#define TGSI_OPCODE_SIN 48 -#define TGSI_OPCODE_SLE 49 -#define TGSI_OPCODE_SNE 50 -#define TGSI_OPCODE_STR 51 -#define TGSI_OPCODE_TEX 52 -#define TGSI_OPCODE_TXD 53 -#define TGSI_OPCODE_UP2H 54 -#define TGSI_OPCODE_UP2US 55 -#define TGSI_OPCODE_UP4B 56 -#define TGSI_OPCODE_UP4UB 57 -#define TGSI_OPCODE_X2D 58 - -/* - * GL_NV_vertex_program2 - */ -#define TGSI_OPCODE_ARA 59 -#define TGSI_OPCODE_ARR 60 -#define TGSI_OPCODE_BRA 61 -#define TGSI_OPCODE_CAL 62 -#define TGSI_OPCODE_RET 63 -#define TGSI_OPCODE_SSG 64 - -/* - * GL_ARB_vertex_program - */ -#define TGSI_OPCODE_SWZ TGSI_OPCODE_MOV -#define TGSI_OPCODE_XPD TGSI_OPCODE_CROSSPRODUCT - -/* - * GL_ARB_fragment_program - */ -#define TGSI_OPCODE_CMP 65 -#define TGSI_OPCODE_SCS 66 -#define TGSI_OPCODE_TXB 67 - -/* - * GL_NV_fragment_program_option - */ -/* No new opcode */ - -/* - * GL_NV_fragment_program2 - */ -#define TGSI_OPCODE_NRM 68 -#define TGSI_OPCODE_DIV 69 -#define TGSI_OPCODE_DP2 70 -#define TGSI_OPCODE_DP2A TGSI_OPCODE_DOT2ADD -#define TGSI_OPCODE_TXL 71 -#define TGSI_OPCODE_BRK 72 -#define TGSI_OPCODE_IF 73 -#define TGSI_OPCODE_LOOP 74 -#define TGSI_OPCODE_REP 75 -#define TGSI_OPCODE_ELSE 76 -#define TGSI_OPCODE_ENDIF 77 -#define TGSI_OPCODE_ENDLOOP 78 -#define TGSI_OPCODE_ENDREP 79 - -/* - * GL_NV_vertex_program2_option - */ - -/* - * GL_NV_vertex_program3 - */ -#define TGSI_OPCODE_PUSHA 80 -#define TGSI_OPCODE_POPA 81 - -/* - * GL_NV_gpu_program4 - */ -#define TGSI_OPCODE_CEIL 82 -#define TGSI_OPCODE_I2F 83 -#define TGSI_OPCODE_NOT 84 -#define TGSI_OPCODE_TRUNC 85 -#define TGSI_OPCODE_SHL 86 -#define TGSI_OPCODE_SHR 87 -#define TGSI_OPCODE_AND 88 -#define TGSI_OPCODE_OR 89 -#define TGSI_OPCODE_MOD 90 -#define TGSI_OPCODE_XOR 91 -#define TGSI_OPCODE_SAD 92 -#define TGSI_OPCODE_TXF 93 -#define TGSI_OPCODE_TXQ 94 -#define TGSI_OPCODE_CONT 95 - -/* - * GL_NV_vertex_program4 - */ -/* Same as GL_NV_gpu_program4 */ - -/* - * GL_NV_fragment_program4 - */ -/* Same as GL_NV_gpu_program4 */ - -/* - * GL_NV_geometry_program4 - */ -/* Same as GL_NV_gpu_program4 */ -#define TGSI_OPCODE_EMIT 96 -#define TGSI_OPCODE_ENDPRIM 97 - -/* - * GLSL - */ -#define TGSI_OPCODE_BGNLOOP2 98 -#define TGSI_OPCODE_BGNSUB 99 -#define TGSI_OPCODE_ENDLOOP2 100 -#define TGSI_OPCODE_ENDSUB 101 -#define TGSI_OPCODE_INT TGSI_OPCODE_TRUNC -#define TGSI_OPCODE_NOISE1 102 -#define TGSI_OPCODE_NOISE2 103 -#define TGSI_OPCODE_NOISE3 104 -#define TGSI_OPCODE_NOISE4 105 -#define TGSI_OPCODE_NOP 106 - -/* - * ps_1_1 - */ -#define TGSI_OPCODE_TEXCOORD TGSI_OPCODE_NOP -#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL -#define TGSI_OPCODE_TEXBEM 107 -#define TGSI_OPCODE_TEXBEML 108 -#define TGSI_OPCODE_TEXREG2AR 109 -#define TGSI_OPCODE_TEXM3X2PAD 110 -#define TGSI_OPCODE_TEXM3X2TEX 111 -#define TGSI_OPCODE_TEXM3X3PAD 112 -#define TGSI_OPCODE_TEXM3X3TEX 113 -#define TGSI_OPCODE_TEXM3X3SPEC 114 -#define TGSI_OPCODE_TEXM3X3VSPEC 115 - -/* - * ps_1_2 - */ -#define TGSI_OPCODE_TEXREG2GB 116 -#define TGSI_OPCODE_TEXREG2RGB 117 -#define TGSI_OPCODE_TEXDP3TEX 118 -#define TGSI_OPCODE_TEXDP3 119 -#define TGSI_OPCODE_TEXM3X3 120 -/* CMP - use TGSI_OPCODE_CND0 */ - -/* - * ps_1_3 - */ -#define TGSI_OPCODE_TEXM3X2DEPTH 121 -/* CMP - use TGSI_OPCODE_CND0 */ - -/* - * ps_1_4 - */ -#define TGSI_OPCODE_TEXCRD TGSI_OPCODE_TEXCOORD -#define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX -#define TGSI_OPCODE_TEXDEPTH 122 -#define TGSI_OPCODE_BEM 123 - -/* - * ps_2_0 - */ -#define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX -#define TGSI_OPCODE_M4X3 124 -#define TGSI_OPCODE_M3X4 125 -#define TGSI_OPCODE_M3X3 126 -#define TGSI_OPCODE_M3X2 127 -#define TGSI_OPCODE_CRS TGSI_OPCODE_XPD -#define TGSI_OPCODE_NRM4 128 -#define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS -#define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB -#define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A - -/* - * ps_2_x - */ -#define TGSI_OPCODE_CALL TGSI_OPCODE_CAL -#define TGSI_OPCODE_CALLNZ 129 -#define TGSI_OPCODE_IFC 130 -#define TGSI_OPCODE_BREAK TGSI_OPCODE_BRK -#define TGSI_OPCODE_BREAKC 131 -#define TGSI_OPCODE_DSX TGSI_OPCODE_DDX -#define TGSI_OPCODE_DSY TGSI_OPCODE_DDY -#define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD - -/* - * vs_1_1 - */ -#define TGSI_OPCODE_EXPP TGSI_OPCODE_EXP -#define TGSI_OPCODE_LOGP TGSI_OPCODE_LG2 - -/* - * vs_2_0 - */ -#define TGSI_OPCODE_SGN TGSI_OPCODE_SSG -#define TGSI_OPCODE_MOVA TGSI_OPCODE_ARR - -/* - * vs_2_x - */ - -#define TGSI_OPCODE_KIL 132 /* unpredicated kill */ -#define TGSI_OPCODE_END 133 /* aka HALT */ - -#define TGSI_OPCODE_LAST 134 - -#define TGSI_SAT_NONE 0 /* do not saturate */ -#define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ -#define TGSI_SAT_MINUS_PLUS_ONE 2 /* clamp to [-1,1] */ - -/* - * Opcode is the operation code to execute. A given operation defines the - * semantics how the source registers (if any) are interpreted and what is - * written to the destination registers (if any) as a result of execution. - * - * NumDstRegs and NumSrcRegs is the number of destination and source registers, - * respectively. For a given operation code, those numbers are fixed and are - * present here only for convenience. - * - * If Extended is TRUE, it is now executed. - * - * Saturate controls how are final results in destination registers modified. - */ - -struct tgsi_instruction -{ - unsigned Type : 4; /* TGSI_TOKEN_TYPE_INSTRUCTION */ - unsigned Size : 8; /* UINT */ - unsigned Opcode : 8; /* TGSI_OPCODE_ */ - unsigned Saturate : 2; /* TGSI_SAT_ */ - unsigned NumDstRegs : 2; /* UINT */ - unsigned NumSrcRegs : 4; /* UINT */ - unsigned Padding : 3; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_instruction::Extended is TRUE, tgsi_instruction_ext follows. - * - * Then, tgsi_instruction::NumDstRegs of tgsi_dst_register follow. - * - * Then, tgsi_instruction::NumSrcRegs of tgsi_src_register follow. - * - * tgsi_instruction::Size contains the total number of words that make the - * instruction, including the instruction word. - */ - -#define TGSI_INSTRUCTION_EXT_TYPE_NV 0 -#define TGSI_INSTRUCTION_EXT_TYPE_LABEL 1 -#define TGSI_INSTRUCTION_EXT_TYPE_TEXTURE 2 -#define TGSI_INSTRUCTION_EXT_TYPE_PREDICATE 3 - -struct tgsi_instruction_ext -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_ */ - unsigned Padding : 27; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_NV, it should - * be cast to tgsi_instruction_ext_nv. - * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_LABEL, it - * should be cast to tgsi_instruction_ext_label. - * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_TEXTURE, it - * should be cast to tgsi_instruction_ext_texture. - * - * If tgsi_instruction_ext::Type is TGSI_INSTRUCTION_EXT_TYPE_PREDICATE, it - * should be cast to tgsi_instruction_ext_predicate. - * - * If tgsi_instruction_ext::Extended is TRUE, another tgsi_instruction_ext - * follows. - */ - -#define TGSI_PRECISION_DEFAULT 0 -#define TGSI_PRECISION_FLOAT32 1 -#define TGSI_PRECISION_FLOAT16 2 -#define TGSI_PRECISION_FIXED12 3 - -#define TGSI_CC_GT 0 -#define TGSI_CC_EQ 1 -#define TGSI_CC_LT 2 -#define TGSI_CC_UN 3 -#define TGSI_CC_GE 4 -#define TGSI_CC_LE 5 -#define TGSI_CC_NE 6 -#define TGSI_CC_TR 7 -#define TGSI_CC_FL 8 - -#define TGSI_SWIZZLE_X 0 -#define TGSI_SWIZZLE_Y 1 -#define TGSI_SWIZZLE_Z 2 -#define TGSI_SWIZZLE_W 3 - -/* - * Precision controls the precision at which the operation should be executed. - * - * CondDstUpdate enables condition code register writes. When this field is - * TRUE, CondDstIndex specifies the index of the condition code register to - * update. - * - * CondFlowEnable enables conditional execution of the operation. When this - * field is TRUE, CondFlowIndex specifies the index of the condition code - * register to test against CondMask with component swizzle controled by - * CondSwizzleX, CondSwizzleY, CondSwizzleZ and CondSwizzleW. If the test fails, - * the operation is not executed. - */ - -struct tgsi_instruction_ext_nv -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_NV */ - unsigned Precision : 4; /* TGSI_PRECISION_ */ - unsigned CondDstIndex : 4; /* UINT */ - unsigned CondFlowIndex : 4; /* UINT */ - unsigned CondMask : 4; /* TGSI_CC_ */ - unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned CondDstUpdate : 1; /* BOOL */ - unsigned CondFlowEnable : 1; /* BOOL */ - unsigned Padding : 1; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_instruction_ext_label -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_LABEL */ - unsigned Label : 24; /* UINT */ - unsigned Padding : 3; - unsigned Extended : 1; /* BOOL */ -}; - -#define TGSI_TEXTURE_UNKNOWN 0 -#define TGSI_TEXTURE_1D 1 -#define TGSI_TEXTURE_2D 2 -#define TGSI_TEXTURE_3D 3 -#define TGSI_TEXTURE_CUBE 4 -#define TGSI_TEXTURE_RECT 5 -#define TGSI_TEXTURE_SHADOW1D 6 -#define TGSI_TEXTURE_SHADOW2D 7 -#define TGSI_TEXTURE_SHADOWRECT 8 - -struct tgsi_instruction_ext_texture -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_TEXTURE */ - unsigned Texture : 8; /* TGSI_TEXTURE_ */ - unsigned Padding : 19; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_instruction_ext_predicate -{ - unsigned Type : 4; /* TGSI_INSTRUCTION_EXT_TYPE_PREDICATE */ - unsigned PredDstIndex : 4; /* UINT */ - unsigned PredWriteMask : 4; /* TGSI_WRITEMASK_ */ - unsigned Padding : 19; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * File specifies the register array to access. - * - * Index specifies the element number of a register in the register file. - * - * If Indirect is TRUE, Index should be offset by the X component of a source - * register that follows. The register can be now fetched into local storage - * for further processing. - * - * If Negate is TRUE, all components of the fetched register are negated. - * - * The fetched register components are swizzled according to SwizzleX, SwizzleY, - * SwizzleZ and SwizzleW. - * - * If Extended is TRUE, any further modifications to the source register are - * made to this temporary storage. - */ - -struct tgsi_src_register -{ - unsigned File : 4; /* TGSI_FILE_ */ - unsigned SwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned SwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned Negate : 1; /* BOOL */ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - int Index : 16; /* SINT */ - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_src_register::Extended is TRUE, tgsi_src_register_ext follows. - * - * Then, if tgsi_src_register::Indirect is TRUE, another tgsi_src_register - * follows. - * - * Then, if tgsi_src_register::Dimension is TRUE, tgsi_dimension follows. - */ - -#define TGSI_SRC_REGISTER_EXT_TYPE_SWZ 0 -#define TGSI_SRC_REGISTER_EXT_TYPE_MOD 1 - -struct tgsi_src_register_ext -{ - unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_ */ - unsigned Padding : 27; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_SWZ, - * it should be cast to tgsi_src_register_ext_swz. - * - * If tgsi_src_register_ext::Type is TGSI_SRC_REGISTER_EXT_TYPE_MOD, - * it should be cast to tgsi_src_register_ext_mod. - * - * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext - * follows. - */ - -#define TGSI_EXTSWIZZLE_X TGSI_SWIZZLE_X -#define TGSI_EXTSWIZZLE_Y TGSI_SWIZZLE_Y -#define TGSI_EXTSWIZZLE_Z TGSI_SWIZZLE_Z -#define TGSI_EXTSWIZZLE_W TGSI_SWIZZLE_W -#define TGSI_EXTSWIZZLE_ZERO 4 -#define TGSI_EXTSWIZZLE_ONE 5 - -/* - * ExtSwizzleX, ExtSwizzleY, ExtSwizzleZ and ExtSwizzleW swizzle the source - * register in an extended manner. - * - * NegateX, NegateY, NegateZ and NegateW negate individual components of the - * source register. - * - * ExtDivide specifies which component is used to divide all components of the - * source register. - */ - -struct tgsi_src_register_ext_swz -{ - unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_SWZ */ - unsigned ExtSwizzleX : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned ExtSwizzleY : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned ExtSwizzleZ : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned ExtSwizzleW : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned NegateX : 1; /* BOOL */ - unsigned NegateY : 1; /* BOOL */ - unsigned NegateZ : 1; /* BOOL */ - unsigned NegateW : 1; /* BOOL */ - unsigned ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ - unsigned Padding : 3; - unsigned Extended : 1; /* BOOL */ -}; - -/** - * Extra src register modifiers - * - * If Complement is TRUE, the source register is modified by subtracting it - * from 1.0. - * - * If Bias is TRUE, the source register is modified by subtracting 0.5 from it. - * - * If Scale2X is TRUE, the source register is modified by multiplying it by 2.0. - * - * If Absolute is TRUE, the source register is modified by removing the sign. - * - * If Negate is TRUE, the source register is modified by negating it. - */ - -struct tgsi_src_register_ext_mod -{ - unsigned Type : 4; /* TGSI_SRC_REGISTER_EXT_TYPE_MOD */ - unsigned Complement : 1; /* BOOL */ - unsigned Bias : 1; /* BOOL */ - unsigned Scale2X : 1; /* BOOL */ - unsigned Absolute : 1; /* BOOL */ - unsigned Negate : 1; /* BOOL */ - unsigned Padding : 22; - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_dimension -{ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - unsigned Padding : 13; - int Index : 16; /* SINT */ - unsigned Extended : 1; /* BOOL */ -}; - -struct tgsi_dst_register -{ - unsigned File : 4; /* TGSI_FILE_ */ - unsigned WriteMask : 4; /* TGSI_WRITEMASK_ */ - unsigned Indirect : 1; /* BOOL */ - unsigned Dimension : 1; /* BOOL */ - int Index : 16; /* SINT */ - unsigned Padding : 5; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * If tgsi_dst_register::Extended is TRUE, tgsi_dst_register_ext follows. - * - * Then, if tgsi_dst_register::Indirect is TRUE, tgsi_src_register follows. - */ - -#define TGSI_DST_REGISTER_EXT_TYPE_CONDCODE 0 -#define TGSI_DST_REGISTER_EXT_TYPE_MODULATE 1 -#define TGSI_DST_REGISTER_EXT_TYPE_PREDICATE 2 - -struct tgsi_dst_register_ext -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_ */ - unsigned Padding : 27; - unsigned Extended : 1; /* BOOL */ -}; - -/** - * Extra destination register modifiers - * - * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_CONDCODE, - * it should be cast to tgsi_dst_register_ext_condcode. - * - * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_MODULATE, - * it should be cast to tgsi_dst_register_ext_modulate. - * - * If tgsi_dst_register_ext::Type is TGSI_DST_REGISTER_EXT_TYPE_PREDICATE, - * it should be cast to tgsi_dst_register_ext_predicate. - * - * If tgsi_dst_register_ext::Extended is TRUE, another tgsi_dst_register_ext - * follows. - */ -struct tgsi_dst_register_ext_concode -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_CONDCODE */ - unsigned CondMask : 4; /* TGSI_CC_ */ - unsigned CondSwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned CondSrcIndex : 4; /* UINT */ - unsigned Padding : 11; - unsigned Extended : 1; /* BOOL */ -}; - -#define TGSI_MODULATE_1X 0 -#define TGSI_MODULATE_2X 1 -#define TGSI_MODULATE_4X 2 -#define TGSI_MODULATE_8X 3 -#define TGSI_MODULATE_HALF 4 -#define TGSI_MODULATE_QUARTER 5 -#define TGSI_MODULATE_EIGHTH 6 - -struct tgsi_dst_register_ext_modulate -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_MODULATE */ - unsigned Modulate : 4; /* TGSI_MODULATE_ */ - unsigned Padding : 23; - unsigned Extended : 1; /* BOOL */ -}; - -/* - * Currently, the following constraints apply. - * - * - PredSwizzleXYZW is either set to identity or replicate. - * - PredSrcIndex is 0. - */ - -struct tgsi_dst_register_ext_predicate -{ - unsigned Type : 4; /* TGSI_DST_REGISTER_EXT_TYPE_PREDICATE */ - unsigned PredSwizzleX : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSwizzleY : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSwizzleZ : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSwizzleW : 2; /* TGSI_SWIZZLE_ */ - unsigned PredSrcIndex : 4; /* UINT */ - unsigned Negate : 1; /* BOOL */ - unsigned Padding : 14; - unsigned Extended : 1; /* BOOL */ -}; - - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_TOKEN_H - diff --git a/src/mesa/pipe/p_state.h b/src/mesa/pipe/p_state.h deleted file mode 100644 index 4d3a6b2f41..0000000000 --- a/src/mesa/pipe/p_state.h +++ /dev/null @@ -1,322 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * Abstract graphics pipe state objects. - * - * Basic notes: - * 1. Want compact representations, so we use bitfields. - * 2. Put bitfields before other (GLfloat) fields. - */ - - -#ifndef PIPE_STATE_H -#define PIPE_STATE_H - -#include "p_compiler.h" -#include "p_defines.h" -#include "p_format.h" - -/** - * Implementation limits - */ -#define PIPE_MAX_SAMPLERS 8 -#define PIPE_MAX_CLIP_PLANES 6 -#define PIPE_MAX_CONSTANT 32 -#define PIPE_ATTRIB_MAX 32 -#define PIPE_MAX_COLOR_BUFS 8 -#define PIPE_MAX_TEXTURE_LEVELS 16 -#define PIPE_MAX_FEEDBACK_ATTRIBS 16 -#define PIPE_MAX_SHADER_INPUTS 16 -#define PIPE_MAX_SHADER_OUTPUTS 16 - - -/* fwd decls */ -struct pipe_surface; -struct pipe_winsys; - - - -/** - * The driver will certainly subclass this to include actual memory - * management information. - */ -struct pipe_buffer { - unsigned alignment; - unsigned usage; - unsigned size; - - /** Reference count */ - unsigned refcount; -}; - - - - -/** - * Primitive (point/line/tri) rasterization info - */ -struct pipe_rasterizer_state -{ - unsigned flatshade:1; - unsigned light_twoside:1; - unsigned front_winding:2; /**< PIPE_WINDING_x */ - unsigned cull_mode:2; /**< PIPE_WINDING_x */ - unsigned fill_cw:2; /**< PIPE_POLYGON_MODE_x */ - unsigned fill_ccw:2; /**< PIPE_POLYGON_MODE_x */ - unsigned offset_cw:1; - unsigned offset_ccw:1; - unsigned scissor:1; - unsigned poly_smooth:1; - unsigned poly_stipple_enable:1; - unsigned point_smooth:1; - unsigned point_sprite:1; - unsigned point_size_per_vertex:1; /**< size computed in vertex shader */ - unsigned multisample:1; /* XXX maybe more ms state in future */ - unsigned line_smooth:1; - unsigned line_stipple_enable:1; - unsigned line_stipple_factor:8; /**< [1..256] actually */ - unsigned line_stipple_pattern:16; - unsigned bypass_clipping:1; - unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ - - float line_width; - float point_size; /**< used when no per-vertex size */ - float offset_units; - float offset_scale; - ubyte sprite_coord_mode[PIPE_MAX_SHADER_OUTPUTS]; /**< PIPE_SPRITE_COORD_ */ -}; - - -struct pipe_poly_stipple { - unsigned stipple[32]; -}; - - -struct pipe_viewport_state { - float scale[4]; - float translate[4]; -}; - - -struct pipe_scissor_state { - unsigned minx:16; - unsigned miny:16; - unsigned maxx:16; - unsigned maxy:16; -}; - - -struct pipe_clip_state { - float ucp[PIPE_MAX_CLIP_PLANES][4]; - unsigned nr; -}; - - -/** - * Constants for vertex/fragment shaders - */ -struct pipe_constant_buffer { - struct pipe_buffer *buffer; - unsigned size; /** in bytes */ -}; - - -struct pipe_shader_state { - const struct tgsi_token *tokens; - ubyte num_inputs; - ubyte num_outputs; - ubyte input_map[PIPE_MAX_SHADER_INPUTS]; /* XXX this may be temporary */ - ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ - ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; - ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ - ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; -}; - - -struct pipe_depth_stencil_alpha_state -{ - struct { - unsigned enabled:1; /**< depth test enabled? */ - unsigned writemask:1; /**< allow depth buffer writes? */ - unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ - unsigned occlusion_count:1; /**< do occlusion counting? */ - } depth; - struct { - unsigned enabled:1; - unsigned func:3; /**< PIPE_FUNC_x */ - unsigned fail_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned zpass_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned zfail_op:3; /**< PIPE_STENCIL_OP_x */ - ubyte ref_value; - ubyte value_mask; - ubyte write_mask; - } stencil[2]; /**< [0] = front, [1] = back */ - struct { - unsigned enabled:1; - unsigned func:3; /**< PIPE_FUNC_x */ - float ref; /**< reference value */ - } alpha; -}; - - -struct pipe_blend_state { - unsigned blend_enable:1; - - unsigned rgb_func:3; /**< PIPE_BLEND_x */ - unsigned rgb_src_factor:5; /**< PIPE_BLENDFACTOR_x */ - unsigned rgb_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ - - unsigned alpha_func:3; /**< PIPE_BLEND_x */ - unsigned alpha_src_factor:5; /**< PIPE_BLENDFACTOR_x */ - unsigned alpha_dst_factor:5; /**< PIPE_BLENDFACTOR_x */ - - unsigned logicop_enable:1; - unsigned logicop_func:4; /**< PIPE_LOGICOP_x */ - - unsigned colormask:4; /**< bitmask of PIPE_MASK_R/G/B/A */ - unsigned dither:1; -}; - - -struct pipe_blend_color { - float color[4]; -}; - - -struct pipe_framebuffer_state -{ - /** multiple colorbuffers for multiple render targets */ - unsigned num_cbufs; - struct pipe_surface *cbufs[PIPE_MAX_COLOR_BUFS]; - - struct pipe_surface *zsbuf; /**< Z/stencil buffer */ -}; - - -/** - * Texture sampler state. - */ -struct pipe_sampler_state -{ - unsigned wrap_s:3; /**< PIPE_TEX_WRAP_x */ - unsigned wrap_t:3; /**< PIPE_TEX_WRAP_x */ - unsigned wrap_r:3; /**< PIPE_TEX_WRAP_x */ - unsigned min_img_filter:2; /**< PIPE_TEX_FILTER_x */ - unsigned min_mip_filter:2; /**< PIPE_TEX_MIPFILTER_x */ - unsigned mag_img_filter:2; /**< PIPE_TEX_FILTER_x */ - unsigned compare:1; /**< shadow/depth compare enabled? */ - unsigned compare_mode:1; /**< PIPE_TEX_COMPARE_x */ - unsigned compare_func:3; /**< PIPE_FUNC_x */ - unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */ - float shadow_ambient; /**< shadow test fail color/intensity */ - float lod_bias; /**< LOD/lambda bias */ - float min_lod, max_lod; /**< LOD clamp range, after bias */ - float border_color[4]; - float max_anisotropy; -}; - - -/** - * 2D surface. This is basically a view into a memory buffer. - * May be a renderbuffer, texture mipmap level, etc. - */ -struct pipe_surface -{ - struct pipe_buffer *buffer; /**< driver private buffer handle */ - enum pipe_format format; /**< PIPE_FORMAT_x */ - unsigned status; /**< PIPE_SURFACE_STATUS_x */ - unsigned clear_value; /**< may be temporary */ - unsigned cpp; /**< bytes per pixel */ - unsigned width, height; - unsigned pitch; /**< in pixels */ - unsigned offset; /**< offset from start of buffer, in bytes */ - unsigned refcount; - struct pipe_winsys *winsys; /**< winsys which owns/created the surface */ -}; - - -/** - * Texture. Represents one or several texture images on one or several mipmap - * levels. - */ -struct pipe_texture -{ - /* Effectively the key: - */ - enum pipe_texture_target target; /**< PIPE_TEXTURE_x */ - enum pipe_format format; /**< PIPE_FORMAT_x */ - - unsigned last_level; /**< Index of last mipmap level present/defined */ - - unsigned width[PIPE_MAX_TEXTURE_LEVELS]; - unsigned height[PIPE_MAX_TEXTURE_LEVELS]; - unsigned depth[PIPE_MAX_TEXTURE_LEVELS]; - unsigned cpp; - - unsigned compressed:1; - - /* These are also refcounted: - */ - unsigned refcount; -}; - - -/** - * A vertex buffer. Typically, all the vertex data/attributes for - * drawing something will be in one buffer. But it's also possible, for - * example, to put colors in one buffer and texcoords in another. - */ -struct pipe_vertex_buffer -{ - unsigned pitch:11; /**< stride to same attrib in next vertex, in bytes */ - unsigned max_index; /**< number of vertices in this buffer */ - unsigned buffer_offset; /**< offset to start of data in buffer, in bytes */ - struct pipe_buffer *buffer; /**< the actual buffer */ -}; - - -/** - * Information to describe a vertex attribute (position, color, etc) - */ -struct pipe_vertex_element -{ - /** Offset of this attribute, in bytes, from the start of the vertex */ - unsigned src_offset:11; - - /** Which vertex_buffer (as given to pipe->set_vertex_buffer()) does - * this attribute live in? - */ - unsigned vertex_buffer_index:5; - unsigned nr_components:3; - - enum pipe_format src_format; /**< PIPE_FORMAT_* */ -}; - - -#endif diff --git a/src/mesa/pipe/p_thread.h b/src/mesa/pipe/p_thread.h deleted file mode 100644 index cd432c547c..0000000000 --- a/src/mesa/pipe/p_thread.h +++ /dev/null @@ -1,54 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef P_THREAD_H -#define P_THREAD_H - -#include "p_compiler.h" - -/* - * XXX: We should come up with a system-independent thread definitions. - * XXX: Patching glthread defs for now. - */ - -#ifndef __MSC__ - -#include "glapi/glthread.h" - -#else /* __MSC__ */ - -typedef int _glthread_Mutex; - -#define _glthread_INIT_MUTEX( M ) ((void) (M)) -#define _glthread_LOCK_MUTEX( M ) ((void) (M)) -#define _glthread_UNLOCK_MUTEX( M ) ((void) (M)) - -#define sched_yield() ((void) 0) - -#endif /* __MSC__ */ - -#endif /* P_THREAD_H */ diff --git a/src/mesa/pipe/p_util.h b/src/mesa/pipe/p_util.h deleted file mode 100644 index d7da2801c9..0000000000 --- a/src/mesa/pipe/p_util.h +++ /dev/null @@ -1,408 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef P_UTIL_H -#define P_UTIL_H - -#include "p_compiler.h" -#include "p_debug.h" -#include - - -#ifdef WIN32 - -#ifdef __cplusplus -extern "C" -{ -#endif - -void * __stdcall -EngAllocMem( - unsigned long Flags, - unsigned long MemSize, - unsigned long Tag ); - -void __stdcall -EngFreeMem( - void *Mem ); - -#ifdef __cplusplus -} -#endif - -static INLINE void * -MALLOC( unsigned size ) -{ - return EngAllocMem( 0, size, 'D3AG' ); -} - -static INLINE void * -CALLOC( unsigned count, unsigned size ) -{ - void *ptr = MALLOC( count * size ); - if( ptr ) { - memset( ptr, 0, count * size ); - } - return ptr; -} - -static INLINE void -FREE( void *ptr ) -{ - if( ptr ) { - EngFreeMem( ptr ); - } -} - -static INLINE void * -REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) -{ - void *new_ptr; - if( new_size <= old_size ) { - return old_ptr; - } - new_ptr = MALLOC( new_size ); - if( new_ptr ) { - memcpy( new_ptr, old_ptr, old_size ); - } - FREE( old_ptr ); - return new_ptr; -} - -#define GETENV( X ) NULL - -#else /* WIN32 */ - -#define MALLOC( SIZE ) malloc( SIZE ) - -#define CALLOC( COUNT, SIZE ) calloc( COUNT, SIZE ) - -#define FREE( PTR ) free( PTR ) - -#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) - -#define GETENV( X ) getenv( X ) - -#endif /* WIN32 */ - -#define MALLOC_STRUCT(T) (struct T *) MALLOC(sizeof(struct T)) - -#define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) - - -/** - * Return a pointer aligned to next multiple of N bytes. - */ -static INLINE void * -align_pointer( void *unaligned, uint alignment ) -{ - if (sizeof(void *) == 64) { - union { - void *p; - uint64 u; - } pu; - pu.p = unaligned; - pu.u = (pu.u + alignment - 1) & ~(uint64) (alignment - 1); - return pu.p; - } - else { - /* 32-bit pointers */ - union { - void *p; - uint u; - } pu; - pu.p = unaligned; - pu.u = (pu.u + alignment - 1) & ~(alignment - 1); - return pu.p; - } -} - -/** - * Return memory on given byte alignment - */ -static INLINE void * -align_malloc(size_t bytes, uint alignment) -{ -#if defined(HAVE_POSIX_MEMALIGN) - void *mem; - (void) posix_memalign(& mem, alignment, bytes); - return mem; -#else - char *ptr, *buf; - - assert( alignment > 0 ); - - ptr = (char *) MALLOC(bytes + alignment + sizeof(void *)); - if (!ptr) - return NULL; - - buf = (char *) align_pointer( ptr + sizeof(void *), alignment ); - *(char **)(buf - sizeof(void *)) = ptr; - - return buf; -#endif /* defined(HAVE_POSIX_MEMALIGN) */ -} - -/** - * Free memory returned by align_malloc(). - */ -static INLINE void -align_free(void *ptr) -{ -#if defined(HAVE_POSIX_MEMALIGN) - FREE(ptr); -#else - void **cubbyHole = (void **) ((char *) ptr - sizeof(void *)); - void *realAddr = *cubbyHole; - FREE(realAddr); -#endif /* defined(HAVE_POSIX_MEMALIGN) */ -} - - - -/** - * Duplicate a block of memory. - */ -static INLINE void * -mem_dup(const void *src, uint size) -{ - void *dup = MALLOC(size); - if (dup) - memcpy(dup, src, size); - return dup; -} - - - -#define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) -#define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) -#define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) - -#define Elements(x) sizeof(x)/sizeof(*(x)) -#define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER)) - -/** - * Return a pointer aligned to next multiple of 16 bytes. - */ -static INLINE void * -align16( void *unaligned ) -{ - return align_pointer( unaligned, 16 ); -} - - -static INLINE int align_int(int x, int align) -{ - return (x + align - 1) & ~(align - 1); -} - - - -#if defined(__MSC__) && defined(__WIN32__) -static INLINE unsigned ffs( unsigned u ) -{ - unsigned i; - - if( u == 0 ) { - return 0; - } - - __asm bsf eax, [u] - __asm inc eax - __asm mov [i], eax - - return i; -} -#endif - -union fi { - float f; - int i; - unsigned ui; -}; - -#define UBYTE_TO_FLOAT( ub ) ((float)(ub) / 255.0F) - -#define IEEE_0996 0x3f7f0000 /* 0.996 or so */ - -/* This function/macro is sensitive to precision. Test very carefully - * if you change it! - */ -#define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \ - do { \ - union fi __tmp; \ - __tmp.f = (F); \ - if (__tmp.i < 0) \ - UB = (ubyte) 0; \ - else if (__tmp.i >= IEEE_0996) \ - UB = (ubyte) 255; \ - else { \ - __tmp.f = __tmp.f * (255.0f/256.0f) + 32768.0f; \ - UB = (ubyte) __tmp.i; \ - } \ - } while (0) - - - -static INLINE unsigned pack_ub4( unsigned char b0, - unsigned char b1, - unsigned char b2, - unsigned char b3 ) -{ - return ((((unsigned int)b0) << 0) | - (((unsigned int)b1) << 8) | - (((unsigned int)b2) << 16) | - (((unsigned int)b3) << 24)); -} - -static INLINE unsigned fui( float f ) -{ - union fi fi; - fi.f = f; - return fi.ui; -} - -static INLINE unsigned char float_to_ubyte( float f ) -{ - unsigned char ub; - UNCLAMPED_FLOAT_TO_UBYTE(ub, f); - return ub; -} - -static INLINE unsigned pack_ui32_float4( float a, - float b, - float c, - float d ) -{ - return pack_ub4( float_to_ubyte(a), - float_to_ubyte(b), - float_to_ubyte(c), - float_to_ubyte(d) ); -} - -#define COPY_4V( DST, SRC ) \ -do { \ - (DST)[0] = (SRC)[0]; \ - (DST)[1] = (SRC)[1]; \ - (DST)[2] = (SRC)[2]; \ - (DST)[3] = (SRC)[3]; \ -} while (0) - - -#define COPY_4FV( DST, SRC ) COPY_4V(DST, SRC) - - -#define ASSIGN_4V( DST, V0, V1, V2, V3 ) \ -do { \ - (DST)[0] = (V0); \ - (DST)[1] = (V1); \ - (DST)[2] = (V2); \ - (DST)[3] = (V3); \ -} while (0) - - -static INLINE int ifloor(float f) -{ - int ai, bi; - double af, bf; - union fi u; - - af = (3 << 22) + 0.5 + (double)f; - bf = (3 << 22) + 0.5 - (double)f; - u.f = (float) af; ai = u.i; - u.f = (float) bf; bi = u.i; - return (ai - bi) >> 1; -} - - -#if defined(__GNUC__) && defined(__i386__) -static INLINE int iround(float f) -{ - int r; - __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st"); - return r; -} -#elif defined(__MSC__) && defined(__WIN32__) -static INLINE int iround(float f) -{ - int r; - _asm { - fld f - fistp r - } - return r; -} -#else -#define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F))) -#endif - - -/* Could maybe have an inline version of this? - */ -#if defined(__GNUC__) -#define FABSF(x) fabsf(x) -#else -#define FABSF(x) ((float) fabs(x)) -#endif - -/* Pretty fast, and accurate. - * Based on code from http://www.flipcode.com/totd/ - */ -static INLINE float LOG2(float val) -{ - union fi num; - int log_2; - - num.f = val; - log_2 = ((num.i >> 23) & 255) - 128; - num.i &= ~(255 << 23); - num.i += 127 << 23; - num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3; - return num.f + log_2; -} - -#if defined(__GNUC__) -#define CEILF(x) ceilf(x) -#else -#define CEILF(x) ((float) ceil(x)) -#endif - -static INLINE int align(int value, int alignment) -{ - return (value + alignment - 1) & ~(alignment - 1); -} - - -/* util/p_util.c - */ -extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, - unsigned dst_x, unsigned dst_y, unsigned width, - unsigned height, const ubyte * src, - int src_pitch, unsigned src_x, int src_y); - - -#endif diff --git a/src/mesa/pipe/p_winsys.h b/src/mesa/pipe/p_winsys.h deleted file mode 100644 index 1e81eebd78..0000000000 --- a/src/mesa/pipe/p_winsys.h +++ /dev/null @@ -1,160 +0,0 @@ - /************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef P_WINSYS_H -#define P_WINSYS_H - - -#include "p_format.h" - -/** - * \file - * This is the interface that Gallium3D requires any window system - * hosting it to implement. This is the only include file in Gallium3D - * which is public. - */ - - -/** Opaque type */ -struct pipe_fence_handle; - -struct pipe_surface; - - -/** - * Gallium3D drivers are (meant to be!) independent of both GL and the - * window system. The window system provides a buffer manager and a - * set of additional hooks for things like command buffer submission, - * etc. - * - * There clearly has to be some agreement between the window system - * driver and the hardware driver about the format of command buffers, - * etc. - */ -struct pipe_winsys -{ - /** Returns name of this winsys interface */ - const char *(*get_name)( struct pipe_winsys *sws ); - - /** - * Do any special operations to ensure frontbuffer contents are - * displayed, eg copy fake frontbuffer. - */ - void (*flush_frontbuffer)( struct pipe_winsys *sws, - struct pipe_surface *surf, - void *context_private ); - - /** Debug output */ - void (*printf)( struct pipe_winsys *sws, - const char *, ... ); - - - /** allocate a new surface (no context dependency) */ - struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws); - - /** - * Allocate storage for a pipe_surface. - * Returns 0 if succeeds. - */ - int (*surface_alloc_storage)(struct pipe_winsys *ws, - struct pipe_surface *surf, - unsigned width, unsigned height, - enum pipe_format format, - unsigned flags); - - void (*surface_release)(struct pipe_winsys *ws, struct pipe_surface **s); - - - /** - * Buffer management. Buffer attributes are mostly fixed over its lifetime. - * - * Remember that gallium gets to choose the interface it needs, and the - * window systems must then implement that interface (rather than the - * other way around...). - * - * usage is a bitmask of PIPE_BUFFER_USAGE_PIXEL/VERTEX/INDEX/CONSTANT. This - * usage argument is only an optimization hint, not a guarantee, therefore - * proper behavior must be observed in all circumstances. - */ - struct pipe_buffer *(*buffer_create)( struct pipe_winsys *sws, - unsigned alignment, - unsigned usage, - unsigned size ); - - /** Create a buffer that wraps user-space data */ - struct pipe_buffer *(*user_buffer_create)(struct pipe_winsys *sws, - void *ptr, - unsigned bytes); - - /** - * Map the entire data store of a buffer object into the client's address. - * flags is bitmask of PIPE_BUFFER_USAGE_CPU_READ/WRITE flags. - */ - void *(*buffer_map)( struct pipe_winsys *sws, - struct pipe_buffer *buf, - unsigned usage ); - - void (*buffer_unmap)( struct pipe_winsys *sws, - struct pipe_buffer *buf ); - - void (*buffer_destroy)( struct pipe_winsys *sws, - struct pipe_buffer *buf ); - - - /** Set ptr = fence, with reference counting */ - void (*fence_reference)( struct pipe_winsys *sws, - struct pipe_fence_handle **ptr, - struct pipe_fence_handle *fence ); - - /** - * Checks whether the fence has been signalled. - * - * The meaning of flag is pipe-driver specific. - * - * Returns zero if it has. - */ - int (*fence_signalled)( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ); - - /** - * Wait for the fence to finish. - * - * The meaning of flag is pipe-driver specific. - * - * Returns zero on success. - */ - int (*fence_finish)( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ); - - -}; - - - -#endif /* P_WINSYS_H */ diff --git a/src/mesa/pipe/pipebuffer/Makefile b/src/mesa/pipe/pipebuffer/Makefile deleted file mode 100644 index 75764a9a18..0000000000 --- a/src/mesa/pipe/pipebuffer/Makefile +++ /dev/null @@ -1,23 +0,0 @@ - -TOP = ../../../.. -include $(TOP)/configs/current - -LIBNAME = pipebuffer - -DRIVER_SOURCES = \ - pb_buffer_fenced.c \ - pb_buffer_malloc.c \ - pb_bufmgr_fenced.c \ - pb_bufmgr_mm.c \ - pb_bufmgr_pool.c \ - pb_winsys.c - -C_SOURCES = \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -include ../Makefile.template - -symlinks: - diff --git a/src/mesa/pipe/pipebuffer/linked_list.h b/src/mesa/pipe/pipebuffer/linked_list.h deleted file mode 100644 index e99817fb13..0000000000 --- a/src/mesa/pipe/pipebuffer/linked_list.h +++ /dev/null @@ -1,91 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND. USA. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - **************************************************************************/ - -/** - * \file - * List macros heavily inspired by the Linux kernel - * list handling. No list looping yet. - * - * Is not threadsafe, so common operations need to - * be protected using an external mutex. - */ - -#ifndef LINKED_LIST_H_ -#define LINKED_LIST_H_ - - -#include - - -struct list_head -{ - struct list_head *prev; - struct list_head *next; -}; - - -#define LIST_INITHEAD(__item) \ - do { \ - (__item)->prev = (__item); \ - (__item)->next = (__item); \ - } while (0) - -#define LIST_ADD(__item, __list) \ - do { \ - (__item)->prev = (__list); \ - (__item)->next = (__list)->next; \ - (__list)->next->prev = (__item); \ - (__list)->next = (__item); \ - } while (0) - -#define LIST_ADDTAIL(__item, __list) \ - do { \ - (__item)->next = (__list); \ - (__item)->prev = (__list)->prev; \ - (__list)->prev->next = (__item); \ - (__list)->prev = (__item); \ - } while(0) - -#define LIST_DEL(__item) \ - do { \ - (__item)->prev->next = (__item)->next; \ - (__item)->next->prev = (__item)->prev; \ - } while(0) - -#define LIST_DELINIT(__item) \ - do { \ - (__item)->prev->next = (__item)->next; \ - (__item)->next->prev = (__item)->prev; \ - (__item)->next = (__item); \ - (__item)->prev = (__item); \ - } while(0) - -#define LIST_ENTRY(__type, __item, __field) \ - ((__type *)(((char *)(__item)) - offsetof(__type, __field))) - - -#endif /*LINKED_LIST_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer.h b/src/mesa/pipe/pipebuffer/pb_buffer.h deleted file mode 100644 index 97beb5f72a..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_buffer.h +++ /dev/null @@ -1,202 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Generic code for buffers. - * - * Behind a pipe buffle handle there can be DMA buffers, client (or user) - * buffers, regular malloced buffers, etc. This file provides an abstract base - * buffer handle that allows the driver to cope with all those kinds of buffers - * in a more flexible way. - * - * There is no obligation of a winsys driver to use this library. And a pipe - * driver should be completly agnostic about it. - * - * \author José Fonseca - */ - -#ifndef PB_BUFFER_H_ -#define PB_BUFFER_H_ - - -#include "pipe/p_compiler.h" -#include "pipe/p_debug.h" -#include "pipe/p_state.h" -#include "pipe/p_inlines.h" - - -struct pb_vtbl; - -/** - * Buffer description. - * - * Used when allocating the buffer. - */ -struct pb_desc -{ - unsigned alignment; - unsigned usage; -}; - - -/** - * Base class for all pb_* buffers. - */ -struct pb_buffer -{ - struct pipe_buffer base; - - /** - * Pointer to the virtual function table. - * - * Avoid accessing this table directly. Use the inline functions below - * instead to avoid mistakes. - */ - const struct pb_vtbl *vtbl; -}; - - -/** - * Virtual function table for the buffer storage operations. - * - * Note that creation is not done through this table. - */ -struct pb_vtbl -{ - void (*destroy)( struct pb_buffer *buf ); - - /** - * Map the entire data store of a buffer object into the client's address. - * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. - */ - void *(*map)( struct pb_buffer *buf, - unsigned flags ); - - void (*unmap)( struct pb_buffer *buf ); - - /** - * Get the base buffer and the offset. - * - * A buffer can be subdivided in smaller buffers. This method should return - * the underlaying buffer, and the relative offset. - * - * Buffers without an underlaying base buffer should return themselves, with - * a zero offset. - * - * Note that this will increase the reference count of the base buffer. - */ - void (*get_base_buffer)( struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset ); -}; - - -static INLINE struct pipe_buffer * -pb_pipe_buffer( struct pb_buffer *pbuf ) -{ - assert(pbuf); - return &pbuf->base; -} - - -static INLINE struct pb_buffer * -pb_buffer( struct pipe_buffer *buf ) -{ - assert(buf); - /* Could add a magic cookie check on debug builds. - */ - return (struct pb_buffer *)buf; -} - - -/* Accessor functions for pb->vtbl: - */ -static INLINE void * -pb_map(struct pb_buffer *buf, - unsigned flags) -{ - assert(buf); - return buf->vtbl->map(buf, flags); -} - - -static INLINE void -pb_unmap(struct pb_buffer *buf) -{ - assert(buf); - buf->vtbl->unmap(buf); -} - - -static INLINE void -pb_get_base_buffer( struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset ) -{ - buf->vtbl->get_base_buffer(buf, base_buf, offset); -} - - -static INLINE void -pb_destroy(struct pb_buffer *buf) -{ - assert(buf); - buf->vtbl->destroy(buf); -} - - -/* XXX: thread safety issues! - */ -static INLINE void -pb_reference(struct pb_buffer **dst, - struct pb_buffer *src) -{ - if (src) - src->base.refcount++; - - if (*dst && --(*dst)->base.refcount == 0) - pb_destroy( *dst ); - - *dst = src; -} - - -/** - * Malloc-based buffer to store data that can't be used by the graphics - * hardware. - */ -struct pb_buffer * -pb_malloc_buffer_create(size_t size, - const struct pb_desc *desc); - - -void -pb_init_winsys(struct pipe_winsys *winsys); - - -#endif /*PB_BUFFER_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c deleted file mode 100644 index f4fc3f6d71..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.c +++ /dev/null @@ -1,299 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Implementation of fenced buffers. - * - * \author José Fonseca - * \author Thomas Hellström - */ - - -#include "linked_list.h" - -#include "p_compiler.h" -#include "p_debug.h" -#include "p_winsys.h" -#include "p_thread.h" -#include "p_util.h" - -#include "pb_buffer.h" -#include "pb_buffer_fenced.h" - -#ifndef __MSC__ -#include -#endif - - -/** - * Convenience macro (type safe). - */ -#define SUPER(__derived) (&(__derived)->base) - - -struct fenced_buffer_list -{ - _glthread_Mutex mutex; - - struct pipe_winsys *winsys; - - size_t numDelayed; - size_t checkDelayed; - - struct list_head delayed; -}; - - -/** - * Wrapper around a pipe buffer which adds fencing and reference counting. - */ -struct fenced_buffer -{ - struct pb_buffer base; - - struct pb_buffer *buffer; - - struct pipe_fence_handle *fence; - - struct list_head head; - struct fenced_buffer_list *list; -}; - - -static INLINE struct fenced_buffer * -fenced_buffer(struct pb_buffer *buf) -{ - assert(buf); - assert(buf->vtbl == &fenced_buffer_vtbl); - return (struct fenced_buffer *)buf; -} - - - - -static void -_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, - int wait) -{ - struct pipe_winsys *winsys = fenced_list->winsys; - struct fenced_buffer *fenced_buf; - struct list_head *list, *prev; - int signaled = -1; - - list = fenced_list->delayed.next; - - if (fenced_list->numDelayed > 3) { - unsigned i; - - for (i = 0; i < fenced_list->numDelayed; i += 3) { - list = list->next; - } - } - - prev = list->prev; - for (; list != &fenced_list->delayed; list = prev, prev = list->prev) { - - fenced_buf = LIST_ENTRY(struct fenced_buffer, list, head); - - if (signaled != 0) { - if (wait) { - signaled = winsys->fence_finish(winsys, fenced_buf->fence, 0); - } - else { - signaled = winsys->fence_signalled(winsys, fenced_buf->fence, 0); - } - } - - if (signaled != 0) - /* XXX: we are assuming that buffers are freed in the same order they - * are fenced which may not always be true... - */ - break; - - winsys->fence_reference(winsys, &fenced_buf->fence, NULL); - - LIST_DEL(list); - fenced_list->numDelayed--; - - /* Do the delayed destroy: - */ - pb_reference(&fenced_buf->buffer, NULL); - FREE(fenced_buf); - } -} - - -static void -fenced_buffer_destroy(struct pb_buffer *buf) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - struct fenced_buffer_list *fenced_list = fenced_buf->list; - - if (fenced_buf->fence) { - LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); - fenced_list->numDelayed++; - } - else { - pb_reference(&fenced_buf->buffer, NULL); - FREE(fenced_buf); - } - - if ((fenced_list->numDelayed % fenced_list->checkDelayed) == 0) - _fenced_buffer_list_check_free(fenced_list, 0); -} - - -static void * -fenced_buffer_map(struct pb_buffer *buf, - unsigned flags) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - return pb_map(fenced_buf->buffer, flags); -} - - -static void -fenced_buffer_unmap(struct pb_buffer *buf) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - pb_unmap(fenced_buf->buffer); -} - - -static void -fenced_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - pb_get_base_buffer(fenced_buf->buffer, base_buf, offset); -} - - -const struct pb_vtbl -fenced_buffer_vtbl = { - fenced_buffer_destroy, - fenced_buffer_map, - fenced_buffer_unmap, - fenced_buffer_get_base_buffer -}; - - -struct pb_buffer * -fenced_buffer_create(struct fenced_buffer_list *fenced_list, - struct pb_buffer *buffer) -{ - struct fenced_buffer *buf; - - if(!buffer) - return NULL; - - buf = CALLOC_STRUCT(fenced_buffer); - if(!buf) - return NULL; - - buf->base.base.refcount = 1; - buf->base.base.alignment = buffer->base.alignment; - buf->base.base.usage = buffer->base.usage; - buf->base.base.size = buffer->base.size; - - buf->base.vtbl = &fenced_buffer_vtbl; - buf->buffer = buffer; - buf->list = fenced_list; - - return &buf->base; -} - - -void -buffer_fence(struct pb_buffer *buf, - struct pipe_fence_handle *fence) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - struct fenced_buffer_list *fenced_list = fenced_buf->list; - struct pipe_winsys *winsys = fenced_list->winsys; - - _glthread_LOCK_MUTEX(fenced_list->mutex); - winsys->fence_reference(winsys, &fenced_buf->fence, fence); - _glthread_UNLOCK_MUTEX(fenced_list->mutex); -} - - -struct fenced_buffer_list * -fenced_buffer_list_create(struct pipe_winsys *winsys) -{ - struct fenced_buffer_list *fenced_list; - - fenced_list = (struct fenced_buffer_list *)CALLOC(1, sizeof(*fenced_list)); - if (!fenced_list) - return NULL; - - fenced_list->winsys = winsys; - - LIST_INITHEAD(&fenced_list->delayed); - - fenced_list->numDelayed = 0; - - /* TODO: don't hard code this */ - fenced_list->checkDelayed = 5; - - _glthread_INIT_MUTEX(fenced_list->mutex); - - return fenced_list; -} - - -void -fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, - int wait) -{ - _glthread_LOCK_MUTEX(fenced_list->mutex); - _fenced_buffer_list_check_free(fenced_list, wait); - _glthread_UNLOCK_MUTEX(fenced_list->mutex); -} - - -void -fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) -{ - _glthread_LOCK_MUTEX(fenced_list->mutex); - - /* Wait on outstanding fences */ - while (fenced_list->numDelayed) { - _glthread_UNLOCK_MUTEX(fenced_list->mutex); - sched_yield(); - _fenced_buffer_list_check_free(fenced_list, 1); - _glthread_LOCK_MUTEX(fenced_list->mutex); - } - - _glthread_UNLOCK_MUTEX(fenced_list->mutex); - - FREE(fenced_list); -} - - diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h b/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h deleted file mode 100644 index c40b9c75e1..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_buffer_fenced.h +++ /dev/null @@ -1,117 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Buffer fencing. - * - * "Fenced buffers" is actually a misnomer. They should be referred as - * "fenceable buffers", i.e, buffers that can be fenced, but I couldn't find - * the word "fenceable" in the dictionary. - * - * A "fenced buffer" is a decorator around a normal buffer, which adds two - * special properties: - * - the ability for the destruction to be delayed by a fence; - * - reference counting. - * - * Usually DMA buffers have a life-time that will extend the life-time of its - * handle. The end-of-life is dictated by the fence signalling. - * - * Between the handle's destruction, and the fence signalling, the buffer is - * stored in a fenced buffer list. - * - * \author José Fonseca - */ - -#ifndef PB_BUFFER_FENCED_H_ -#define PB_BUFFER_FENCED_H_ - - -#include "pipe/p_debug.h" - - -struct pipe_winsys; -struct pipe_buffer; -struct pipe_fence_handle; - - -/** - * List of buffers which are awaiting fence signalling. - */ -struct fenced_buffer_list; - - -/** - * The fenced buffer's virtual function table. - * - * NOTE: Made public for debugging purposes. - */ -extern const struct pb_vtbl fenced_buffer_vtbl; - - -/** - * Create a fenced buffer list. - * - * See also fenced_bufmgr_create for a more convenient way to use this. - */ -struct fenced_buffer_list * -fenced_buffer_list_create(struct pipe_winsys *winsys); - - -/** - * Walk the fenced buffer list to check and free signalled buffers. - */ -void -fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, - int wait); - -void -fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list); - - -/** - * Wrap a buffer in a fenced buffer. - * - * NOTE: this will not increase the buffer reference count. - */ -struct pb_buffer * -fenced_buffer_create(struct fenced_buffer_list *fenced, - struct pb_buffer *buffer); - - -/** - * Set a buffer's fence. - * - * NOTE: Although it takes a generic pb_buffer argument, it will fail - * on everything but buffers returned by fenced_buffer_create. - */ -void -buffer_fence(struct pb_buffer *buf, - struct pipe_fence_handle *fence); - - -#endif /*PB_BUFFER_FENCED_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c b/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c deleted file mode 100644 index 9e8244f909..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_buffer_malloc.c +++ /dev/null @@ -1,127 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Implementation of malloc-based buffers to store data that can't be processed - * by the hardware. - * - * \author José Fonseca - */ - - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pb_buffer.h" - - -struct malloc_buffer -{ - struct pb_buffer base; - void *data; -}; - - -extern const struct pb_vtbl malloc_buffer_vtbl; - -static INLINE struct malloc_buffer * -malloc_buffer(struct pb_buffer *buf) -{ - assert(buf); - assert(buf->vtbl == &malloc_buffer_vtbl); - return (struct malloc_buffer *)buf; -} - - -static void -malloc_buffer_destroy(struct pb_buffer *buf) -{ - align_free(malloc_buffer(buf)->data); - FREE(buf); -} - - -static void * -malloc_buffer_map(struct pb_buffer *buf, - unsigned flags) -{ - return malloc_buffer(buf)->data; -} - - -static void -malloc_buffer_unmap(struct pb_buffer *buf) -{ - /* No-op */ -} - - -static void -malloc_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - *base_buf = buf; - *offset = 0; -} - - -const struct pb_vtbl -malloc_buffer_vtbl = { - malloc_buffer_destroy, - malloc_buffer_map, - malloc_buffer_unmap, - malloc_buffer_get_base_buffer -}; - - -struct pb_buffer * -pb_malloc_buffer_create(size_t size, - const struct pb_desc *desc) -{ - struct malloc_buffer *buf; - - /* TODO: do a single allocation */ - - buf = CALLOC_STRUCT(malloc_buffer); - if(!buf) - return NULL; - - buf->base.base.refcount = 1; - buf->base.base.alignment = desc->alignment; - buf->base.base.usage = desc->usage; - buf->base.base.size = size; - buf->base.vtbl = &malloc_buffer_vtbl; - - buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment); - if(!buf->data) { - align_free(buf); - return NULL; - } - - return &buf->base; -} diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr.h b/src/mesa/pipe/pipebuffer/pb_bufmgr.h deleted file mode 100644 index 1ddf784c97..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr.h +++ /dev/null @@ -1,126 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Buffer management. - * - * A buffer manager does only one basic thing: it creates buffers. Actually, - * "buffer factory" would probably a more accurate description. - * - * You can chain buffer managers so that you can have a finer grained memory - * management and pooling. - * - * For example, for a simple batch buffer manager you would chain: - * - the native buffer manager, which provides DMA memory from the graphics - * memory space; - * - the pool buffer manager, which keep around a pool of equally sized buffers - * to avoid latency associated with the native buffer manager; - * - the fenced buffer manager, which will delay buffer destruction until the - * the moment the card finishing processing it. - * - * \author José Fonseca - */ - -#ifndef PB_BUFMGR_H_ -#define PB_BUFMGR_H_ - - -#include - - -struct pb_desc; -struct pipe_buffer; -struct pipe_winsys; - - -/** - * Abstract base class for all buffer managers. - */ -struct pb_manager -{ - /* XXX: we will likely need more allocation flags */ - struct pb_buffer * - (*create_buffer)( struct pb_manager *mgr, - size_t size, - const struct pb_desc *desc); - - void - (*destroy)( struct pb_manager *mgr ); -}; - - -/** - * Static buffer pool manager. - * - * Manages the allocation of equally sized buffers. It does so by allocating - * a single big buffer and divide it equally sized buffers. - * - * It is meant to manage the allocation of batch buffer pools. - */ -struct pb_manager * -pool_bufmgr_create(struct pb_manager *provider, - size_t n, size_t size, - const struct pb_desc *desc); - - -/** - * Wraper around the old memory manager. - * - * It managers buffers of different sizes. It does so by allocating a buffer - * with the size of the heap, and then using the old mm memory manager to manage - * that heap. - */ -struct pb_manager * -mm_bufmgr_create(struct pb_manager *provider, - size_t size, size_t align2); - -/** - * Same as mm_bufmgr_create. - * - * Buffer will be release when the manager is destroyed. - */ -struct pb_manager * -mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, - size_t size, size_t align2); - - -/** - * Fenced buffer manager. - * - * This manager is just meant for convenience. It wraps the buffers returned - * by another manager in fenced buffers, so that - * - * NOTE: the buffer manager that provides the buffers will be destroyed - * at the same time. - */ -struct pb_manager * -fenced_bufmgr_create(struct pb_manager *provider, - struct pipe_winsys *winsys); - - -#endif /*PB_BUFMGR_H_*/ diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c deleted file mode 100644 index c535d3276c..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_fenced.c +++ /dev/null @@ -1,131 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ - -/** - * \file - * A buffer manager that wraps buffers in fenced buffers. - * - * \author José Fonseca - */ - - -#include "p_debug.h" -#include "p_util.h" - -#include "pb_buffer.h" -#include "pb_buffer_fenced.h" -#include "pb_bufmgr.h" - - -struct fenced_pb_manager -{ - struct pb_manager base; - - struct pb_manager *provider; - - struct fenced_buffer_list *fenced_list; -}; - - -static INLINE struct fenced_pb_manager * -fenced_pb_manager(struct pb_manager *mgr) -{ - assert(mgr); - return (struct fenced_pb_manager *)mgr; -} - - -static struct pb_buffer * -fenced_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, - const struct pb_desc *desc) -{ - struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); - struct pb_buffer *buf; - struct pb_buffer *fenced_buf; - - /* check for free buffers before allocating new ones */ - fenced_buffer_list_check_free(fenced_mgr->fenced_list, 0); - - buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); - if(!buf) { - /* try harder to get a buffer */ - fenced_buffer_list_check_free(fenced_mgr->fenced_list, 1); - - buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); - if(!buf) { - /* give up */ - return NULL; - } - } - - fenced_buf = fenced_buffer_create(fenced_mgr->fenced_list, buf); - if(!fenced_buf) { - assert(buf->base.refcount == 1); - pb_destroy(buf); - } - - return fenced_buf; -} - - -static void -fenced_bufmgr_destroy(struct pb_manager *mgr) -{ - struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); - - fenced_buffer_list_destroy(fenced_mgr->fenced_list); - - fenced_mgr->provider->destroy(fenced_mgr->provider); - - FREE(fenced_mgr); -} - - -struct pb_manager * -fenced_bufmgr_create(struct pb_manager *provider, - struct pipe_winsys *winsys) -{ - struct fenced_pb_manager *fenced_mgr; - - fenced_mgr = (struct fenced_pb_manager *)CALLOC(1, sizeof(*fenced_mgr)); - if (!fenced_mgr) - return NULL; - - fenced_mgr->base.destroy = fenced_bufmgr_destroy; - fenced_mgr->base.create_buffer = fenced_bufmgr_create_buffer; - - fenced_mgr->provider = provider; - fenced_mgr->fenced_list = fenced_buffer_list_create(winsys); - if(!fenced_mgr->fenced_list) { - FREE(fenced_mgr); - return NULL; - } - - return &fenced_mgr->base; -} diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c deleted file mode 100644 index 8b1b51c0e2..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_mm.c +++ /dev/null @@ -1,593 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * Copyright 1999 Wittawat Yamwong - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Buffer manager using the old texture memory manager. - * - * \author José Fonseca - */ - - -#include "linked_list.h" - -#include "p_defines.h" -#include "p_debug.h" -#include "p_thread.h" -#include "p_util.h" -#include "pb_buffer.h" -#include "pb_bufmgr.h" - - -/** - * Convenience macro (type safe). - */ -#define SUPER(__derived) (&(__derived)->base) - - -struct mem_block -{ - struct mem_block *next, *prev; - struct mem_block *next_free, *prev_free; - struct mem_block *heap; - int ofs, size; - unsigned int free:1; - unsigned int reserved:1; -}; - - -#ifdef DEBUG -/** - * For debugging purposes. - */ -static void -mmDumpMemInfo(const struct mem_block *heap) -{ - debug_printf("Memory heap %p:\n", (void *)heap); - if (heap == 0) { - debug_printf(" heap == 0\n"); - } else { - const struct mem_block *p; - - for(p = heap->next; p != heap; p = p->next) { - debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); - } - - debug_printf("\nFree list:\n"); - - for(p = heap->next_free; p != heap; p = p->next_free) { - debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); - } - - } - debug_printf("End of memory blocks\n"); -} -#endif - - -/** - * input: total size in bytes - * return: a heap pointer if OK, NULL if error - */ -static struct mem_block * -mmInit(int ofs, int size) -{ - struct mem_block *heap, *block; - - if (size <= 0) - return NULL; - - heap = CALLOC_STRUCT(mem_block); - if (!heap) - return NULL; - - block = CALLOC_STRUCT(mem_block); - if (!block) { - FREE(heap); - return NULL; - } - - heap->next = block; - heap->prev = block; - heap->next_free = block; - heap->prev_free = block; - - block->heap = heap; - block->next = heap; - block->prev = heap; - block->next_free = heap; - block->prev_free = heap; - - block->ofs = ofs; - block->size = size; - block->free = 1; - - return heap; -} - - -static struct mem_block * -SliceBlock(struct mem_block *p, - int startofs, int size, - int reserved, int alignment) -{ - struct mem_block *newblock; - - /* break left [p, newblock, p->next], then p = newblock */ - if (startofs > p->ofs) { - newblock = CALLOC_STRUCT(mem_block); - if (!newblock) - return NULL; - newblock->ofs = startofs; - newblock->size = p->size - (startofs - p->ofs); - newblock->free = 1; - newblock->heap = p->heap; - - newblock->next = p->next; - newblock->prev = p; - p->next->prev = newblock; - p->next = newblock; - - newblock->next_free = p->next_free; - newblock->prev_free = p; - p->next_free->prev_free = newblock; - p->next_free = newblock; - - p->size -= newblock->size; - p = newblock; - } - - /* break right, also [p, newblock, p->next] */ - if (size < p->size) { - newblock = CALLOC_STRUCT(mem_block); - if (!newblock) - return NULL; - newblock->ofs = startofs + size; - newblock->size = p->size - size; - newblock->free = 1; - newblock->heap = p->heap; - - newblock->next = p->next; - newblock->prev = p; - p->next->prev = newblock; - p->next = newblock; - - newblock->next_free = p->next_free; - newblock->prev_free = p; - p->next_free->prev_free = newblock; - p->next_free = newblock; - - p->size = size; - } - - /* p = middle block */ - p->free = 0; - - /* Remove p from the free list: - */ - p->next_free->prev_free = p->prev_free; - p->prev_free->next_free = p->next_free; - - p->next_free = 0; - p->prev_free = 0; - - p->reserved = reserved; - return p; -} - - -/** - * Allocate 'size' bytes with 2^align2 bytes alignment, - * restrict the search to free memory after 'startSearch' - * depth and back buffers should be in different 4mb banks - * to get better page hits if possible - * input: size = size of block - * align2 = 2^align2 bytes alignment - * startSearch = linear offset from start of heap to begin search - * return: pointer to the allocated block, 0 if error - */ -static struct mem_block * -mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) -{ - struct mem_block *p; - const int mask = (1 << align2)-1; - int startofs = 0; - int endofs; - - if (!heap || align2 < 0 || size <= 0) - return NULL; - - for (p = heap->next_free; p != heap; p = p->next_free) { - assert(p->free); - - startofs = (p->ofs + mask) & ~mask; - if ( startofs < startSearch ) { - startofs = startSearch; - } - endofs = startofs+size; - if (endofs <= (p->ofs+p->size)) - break; - } - - if (p == heap) - return NULL; - - assert(p->free); - p = SliceBlock(p,startofs,size,0,mask+1); - - return p; -} - - -#if 0 -/** - * Free block starts at offset - * input: pointer to a heap, start offset - * return: pointer to a block - */ -static struct mem_block * -mmFindBlock(struct mem_block *heap, int start) -{ - struct mem_block *p; - - for (p = heap->next; p != heap; p = p->next) { - if (p->ofs == start) - return p; - } - - return NULL; -} -#endif - - -static INLINE int -Join2Blocks(struct mem_block *p) -{ - /* XXX there should be some assertions here */ - - /* NOTE: heap->free == 0 */ - - if (p->free && p->next->free) { - struct mem_block *q = p->next; - - assert(p->ofs + p->size == q->ofs); - p->size += q->size; - - p->next = q->next; - q->next->prev = p; - - q->next_free->prev_free = q->prev_free; - q->prev_free->next_free = q->next_free; - - FREE(q); - return 1; - } - return 0; -} - - -/** - * Free block starts at offset - * input: pointer to a block - * return: 0 if OK, -1 if error - */ -static int -mmFreeMem(struct mem_block *b) -{ - if (!b) - return 0; - - if (b->free) { - debug_printf("block already free\n"); - return -1; - } - if (b->reserved) { - debug_printf("block is reserved\n"); - return -1; - } - - b->free = 1; - b->next_free = b->heap->next_free; - b->prev_free = b->heap; - b->next_free->prev_free = b; - b->prev_free->next_free = b; - - Join2Blocks(b); - if (b->prev != b->heap) - Join2Blocks(b->prev); - - return 0; -} - - -/** - * destroy MM - */ -static void -mmDestroy(struct mem_block *heap) -{ - struct mem_block *p; - - if (!heap) - return; - - for (p = heap->next; p != heap; ) { - struct mem_block *next = p->next; - FREE(p); - p = next; - } - - FREE(heap); -} - - -struct mm_pb_manager -{ - struct pb_manager base; - - _glthread_Mutex mutex; - - size_t size; - struct mem_block *heap; - - size_t align2; - - struct pb_buffer *buffer; - void *map; -}; - - -static INLINE struct mm_pb_manager * -mm_pb_manager(struct pb_manager *mgr) -{ - assert(mgr); - return (struct mm_pb_manager *)mgr; -} - - -struct mm_buffer -{ - struct pb_buffer base; - - struct mm_pb_manager *mgr; - - struct mem_block *block; -}; - - -static INLINE struct mm_buffer * -mm_buffer(struct pb_buffer *buf) -{ - assert(buf); - return (struct mm_buffer *)buf; -} - - -static void -mm_buffer_destroy(struct pb_buffer *buf) -{ - struct mm_buffer *mm_buf = mm_buffer(buf); - struct mm_pb_manager *mm = mm_buf->mgr; - - assert(buf->base.refcount == 0); - - _glthread_LOCK_MUTEX(mm->mutex); - mmFreeMem(mm_buf->block); - FREE(buf); - _glthread_UNLOCK_MUTEX(mm->mutex); -} - - -static void * -mm_buffer_map(struct pb_buffer *buf, - unsigned flags) -{ - struct mm_buffer *mm_buf = mm_buffer(buf); - struct mm_pb_manager *mm = mm_buf->mgr; - - return (unsigned char *) mm->map + mm_buf->block->ofs; -} - - -static void -mm_buffer_unmap(struct pb_buffer *buf) -{ - /* No-op */ -} - - -static void -mm_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - struct mm_buffer *mm_buf = mm_buffer(buf); - struct mm_pb_manager *mm = mm_buf->mgr; - pb_get_base_buffer(mm->buffer, base_buf, offset); - *offset += mm_buf->block->ofs; -} - - -static const struct pb_vtbl -mm_buffer_vtbl = { - mm_buffer_destroy, - mm_buffer_map, - mm_buffer_unmap, - mm_buffer_get_base_buffer -}; - - -static struct pb_buffer * -mm_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, - const struct pb_desc *desc) -{ - struct mm_pb_manager *mm = mm_pb_manager(mgr); - struct mm_buffer *mm_buf; - - /* We don't handle alignments larger then the one initially setup */ - assert(desc->alignment % (1 << mm->align2) == 0); - if(desc->alignment % (1 << mm->align2)) - return NULL; - - _glthread_LOCK_MUTEX(mm->mutex); - - mm_buf = CALLOC_STRUCT(mm_buffer); - if (!mm_buf) { - _glthread_UNLOCK_MUTEX(mm->mutex); - return NULL; - } - - mm_buf->base.base.refcount = 1; - mm_buf->base.base.alignment = desc->alignment; - mm_buf->base.base.usage = desc->usage; - mm_buf->base.base.size = size; - - mm_buf->base.vtbl = &mm_buffer_vtbl; - - mm_buf->mgr = mm; - - mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); - if(!mm_buf->block) { - debug_printf("warning: heap full\n"); -#if 0 - mmDumpMemInfo(mm->heap); -#endif - - mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); - if(!mm_buf->block) { - assert(0); - FREE(mm_buf); - _glthread_UNLOCK_MUTEX(mm->mutex); - return NULL; - } - } - - /* Some sanity checks */ - assert(0 <= mm_buf->block->ofs && mm_buf->block->ofs < mm->size); - assert(size <= mm_buf->block->size && mm_buf->block->ofs + mm_buf->block->size <= mm->size); - - _glthread_UNLOCK_MUTEX(mm->mutex); - return SUPER(mm_buf); -} - - -static void -mm_bufmgr_destroy(struct pb_manager *mgr) -{ - struct mm_pb_manager *mm = mm_pb_manager(mgr); - - _glthread_LOCK_MUTEX(mm->mutex); - - mmDestroy(mm->heap); - - pb_unmap(mm->buffer); - pb_reference(&mm->buffer, NULL); - - _glthread_UNLOCK_MUTEX(mm->mutex); - - FREE(mgr); -} - - -struct pb_manager * -mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, - size_t size, size_t align2) -{ - struct mm_pb_manager *mm; - - if(!buffer) - return NULL; - - mm = CALLOC_STRUCT(mm_pb_manager); - if (!mm) - return NULL; - - mm->base.create_buffer = mm_bufmgr_create_buffer; - mm->base.destroy = mm_bufmgr_destroy; - - mm->size = size; - mm->align2 = align2; /* 64-byte alignment */ - - _glthread_INIT_MUTEX(mm->mutex); - - mm->buffer = buffer; - - mm->map = pb_map(mm->buffer, - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE); - if(!mm->map) - goto failure; - - mm->heap = mmInit(0, size); - if (!mm->heap) - goto failure; - - return SUPER(mm); - -failure: -if(mm->heap) - mmDestroy(mm->heap); - if(mm->map) - pb_unmap(mm->buffer); - if(mm) - FREE(mm); - return NULL; -} - - -struct pb_manager * -mm_bufmgr_create(struct pb_manager *provider, - size_t size, size_t align2) -{ - struct pb_buffer *buffer; - struct pb_manager *mgr; - struct pb_desc desc; - - assert(provider); - assert(provider->create_buffer); - - memset(&desc, 0, sizeof(desc)); - desc.alignment = 1 << align2; - - buffer = provider->create_buffer(provider, size, &desc); - if (!buffer) - return NULL; - - mgr = mm_bufmgr_create_from_buffer(buffer, size, align2); - if (!mgr) { - pb_reference(&buffer, NULL); - return NULL; - } - - return mgr; -} diff --git a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c b/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c deleted file mode 100644 index 04477a865a..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_bufmgr_pool.c +++ /dev/null @@ -1,288 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ - -/** - * \file - * Batch buffer pool management. - * - * \author José Fonseca - * \author Thomas Hellström - */ - - -#include "linked_list.h" - -#include "p_compiler.h" -#include "p_debug.h" -#include "p_thread.h" -#include "p_defines.h" -#include "p_util.h" - -#include "pb_buffer.h" -#include "pb_bufmgr.h" - - -/** - * Convenience macro (type safe). - */ -#define SUPER(__derived) (&(__derived)->base) - - -struct pool_pb_manager -{ - struct pb_manager base; - - _glthread_Mutex mutex; - - size_t bufSize; - size_t bufAlign; - - size_t numFree; - size_t numTot; - - struct list_head free; - - struct pb_buffer *buffer; - void *map; - - struct pool_buffer *bufs; -}; - - -static INLINE struct pool_pb_manager * -pool_pb_manager(struct pb_manager *mgr) -{ - assert(mgr); - return (struct pool_pb_manager *)mgr; -} - - -struct pool_buffer -{ - struct pb_buffer base; - - struct pool_pb_manager *mgr; - - struct list_head head; - - size_t start; -}; - - -static INLINE struct pool_buffer * -pool_buffer(struct pb_buffer *buf) -{ - assert(buf); - return (struct pool_buffer *)buf; -} - - - -static void -pool_buffer_destroy(struct pb_buffer *buf) -{ - struct pool_buffer *pool_buf = pool_buffer(buf); - struct pool_pb_manager *pool = pool_buf->mgr; - - assert(pool_buf->base.base.refcount == 0); - - _glthread_LOCK_MUTEX(pool->mutex); - LIST_ADD(&pool_buf->head, &pool->free); - pool->numFree++; - _glthread_UNLOCK_MUTEX(pool->mutex); -} - - -static void * -pool_buffer_map(struct pb_buffer *buf, unsigned flags) -{ - struct pool_buffer *pool_buf = pool_buffer(buf); - struct pool_pb_manager *pool = pool_buf->mgr; - void *map; - - _glthread_LOCK_MUTEX(pool->mutex); - map = (unsigned char *) pool->map + pool_buf->start; - _glthread_UNLOCK_MUTEX(pool->mutex); - return map; -} - - -static void -pool_buffer_unmap(struct pb_buffer *buf) -{ - /* No-op */ -} - - -static void -pool_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - struct pool_buffer *pool_buf = pool_buffer(buf); - struct pool_pb_manager *pool = pool_buf->mgr; - pb_get_base_buffer(pool->buffer, base_buf, offset); - *offset += pool_buf->start; -} - - -static const struct pb_vtbl -pool_buffer_vtbl = { - pool_buffer_destroy, - pool_buffer_map, - pool_buffer_unmap, - pool_buffer_get_base_buffer -}; - - -static struct pb_buffer * -pool_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, - const struct pb_desc *desc) -{ - struct pool_pb_manager *pool = pool_pb_manager(mgr); - struct pool_buffer *pool_buf; - struct list_head *item; - - assert(size == pool->bufSize); - assert(pool->bufAlign % desc->alignment == 0); - - _glthread_LOCK_MUTEX(pool->mutex); - - if (pool->numFree == 0) { - _glthread_UNLOCK_MUTEX(pool->mutex); - debug_printf("warning: out of fixed size buffer objects\n"); - return NULL; - } - - item = pool->free.next; - - if (item == &pool->free) { - _glthread_UNLOCK_MUTEX(pool->mutex); - debug_printf("error: fixed size buffer pool corruption\n"); - return NULL; - } - - LIST_DEL(item); - --pool->numFree; - - _glthread_UNLOCK_MUTEX(pool->mutex); - - pool_buf = LIST_ENTRY(struct pool_buffer, item, head); - assert(pool_buf->base.base.refcount == 0); - pool_buf->base.base.refcount = 1; - pool_buf->base.base.alignment = desc->alignment; - pool_buf->base.base.usage = desc->usage; - - return SUPER(pool_buf); -} - - -static void -pool_bufmgr_destroy(struct pb_manager *mgr) -{ - struct pool_pb_manager *pool = pool_pb_manager(mgr); - _glthread_LOCK_MUTEX(pool->mutex); - - FREE(pool->bufs); - - pb_unmap(pool->buffer); - pb_reference(&pool->buffer, NULL); - - _glthread_UNLOCK_MUTEX(pool->mutex); - - FREE(mgr); -} - - -struct pb_manager * -pool_bufmgr_create(struct pb_manager *provider, - size_t numBufs, - size_t bufSize, - const struct pb_desc *desc) -{ - struct pool_pb_manager *pool; - struct pool_buffer *pool_buf; - size_t i; - - pool = (struct pool_pb_manager *)CALLOC(1, sizeof(*pool)); - if (!pool) - return NULL; - - pool->base.destroy = pool_bufmgr_destroy; - pool->base.create_buffer = pool_bufmgr_create_buffer; - - LIST_INITHEAD(&pool->free); - - pool->numTot = numBufs; - pool->numFree = numBufs; - pool->bufSize = bufSize; - pool->bufAlign = desc->alignment; - - _glthread_INIT_MUTEX(pool->mutex); - - pool->buffer = provider->create_buffer(provider, numBufs*bufSize, desc); - if (!pool->buffer) - goto failure; - - pool->map = pb_map(pool->buffer, - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE); - if(!pool->map) - goto failure; - - pool->bufs = (struct pool_buffer *)CALLOC(numBufs, sizeof(*pool->bufs)); - if (!pool->bufs) - goto failure; - - pool_buf = pool->bufs; - for (i = 0; i < numBufs; ++i) { - pool_buf->base.base.refcount = 0; - pool_buf->base.base.alignment = 0; - pool_buf->base.base.usage = 0; - pool_buf->base.base.size = bufSize; - pool_buf->base.vtbl = &pool_buffer_vtbl; - pool_buf->mgr = pool; - pool_buf->start = i * bufSize; - LIST_ADDTAIL(&pool_buf->head, &pool->free); - pool_buf++; - } - - return SUPER(pool); - -failure: - if(pool->bufs) - FREE(pool->bufs); - if(pool->map) - pb_unmap(pool->buffer); - if(pool->buffer) - pb_reference(&pool->buffer, NULL); - if(pool) - FREE(pool); - return NULL; -} diff --git a/src/mesa/pipe/pipebuffer/pb_winsys.c b/src/mesa/pipe/pipebuffer/pb_winsys.c deleted file mode 100644 index 978944091f..0000000000 --- a/src/mesa/pipe/pipebuffer/pb_winsys.c +++ /dev/null @@ -1,170 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Implementation of client buffer (also designated as "user buffers"), which - * are just state-tracker owned data masqueraded as buffers. - * - * \author José Fonseca - */ - - -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" - -#include "pb_buffer.h" - - -/** - * User buffers are special buffers that initially reference memory - * held by the user but which may if necessary copy that memory into - * device memory behind the scenes, for submission to hardware. - * - * These are particularly useful when the referenced data is never - * submitted to hardware at all, in the particular case of software - * vertex processing. - */ -struct pb_user_buffer -{ - struct pb_buffer base; - void *data; -}; - - -extern const struct pb_vtbl pb_user_buffer_vtbl; - - -static INLINE struct pb_user_buffer * -pb_user_buffer(struct pb_buffer *buf) -{ - assert(buf); - assert(buf->vtbl == &pb_user_buffer_vtbl); - return (struct pb_user_buffer *)buf; -} - - -static void -pb_user_buffer_destroy(struct pb_buffer *buf) -{ - assert(buf); - FREE(buf); -} - - -static void * -pb_user_buffer_map(struct pb_buffer *buf, - unsigned flags) -{ - return pb_user_buffer(buf)->data; -} - - -static void -pb_user_buffer_unmap(struct pb_buffer *buf) -{ - /* No-op */ -} - - -static void -pb_user_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - *base_buf = buf; - *offset = 0; -} - - -const struct pb_vtbl -pb_user_buffer_vtbl = { - pb_user_buffer_destroy, - pb_user_buffer_map, - pb_user_buffer_unmap, - pb_user_buffer_get_base_buffer -}; - - -static struct pipe_buffer * -pb_winsys_user_buffer_create(struct pipe_winsys *winsys, - void *data, - unsigned bytes) -{ - struct pb_user_buffer *buf = CALLOC_STRUCT(pb_user_buffer); - - if(!buf) - return NULL; - - buf->base.base.refcount = 1; - buf->base.base.size = bytes; - buf->base.base.alignment = 0; - buf->base.base.usage = 0; - - buf->base.vtbl = &pb_user_buffer_vtbl; - buf->data = data; - - return &buf->base.base; -} - - -static void * -pb_winsys_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned flags) -{ - (void)winsys; - return pb_map(pb_buffer(buf), flags); -} - - -static void -pb_winsys_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - (void)winsys; - pb_unmap(pb_buffer(buf)); -} - - -static void -pb_winsys_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - (void)winsys; - pb_destroy(pb_buffer(buf)); -} - - -void -pb_init_winsys(struct pipe_winsys *winsys) -{ - winsys->user_buffer_create = pb_winsys_user_buffer_create; - winsys->buffer_map = pb_winsys_buffer_map; - winsys->buffer_unmap = pb_winsys_buffer_unmap; - winsys->buffer_destroy = pb_winsys_buffer_destroy; -} diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile deleted file mode 100644 index 31438a882e..0000000000 --- a/src/mesa/pipe/softpipe/Makefile +++ /dev/null @@ -1,50 +0,0 @@ - -TOP = ../../../.. -include $(TOP)/configs/current - -LIBNAME = softpipe - -DRIVER_SOURCES = \ - sp_clear.c \ - sp_flush.c \ - sp_query.c \ - sp_context.c \ - sp_draw_arrays.c \ - sp_prim_setup.c \ - sp_prim_vbuf.c \ - sp_quad.c \ - sp_quad_alpha_test.c \ - sp_quad_blend.c \ - sp_quad_bufloop.c \ - sp_quad_colormask.c \ - sp_quad_coverage.c \ - sp_quad_depth_test.c \ - sp_quad_earlyz.c \ - sp_quad_fs.c \ - sp_quad_occlusion.c \ - sp_quad_output.c \ - sp_quad_stencil.c \ - sp_quad_stipple.c \ - sp_state_blend.c \ - sp_state_clip.c \ - sp_state_derived.c \ - sp_state_fs.c \ - sp_state_sampler.c \ - sp_state_rasterizer.c \ - sp_state_surface.c \ - sp_state_vertex.c \ - sp_texture.c \ - sp_tex_sample.c \ - sp_tile_cache.c \ - sp_surface.c - -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -include ../Makefile.template - -symlinks: - diff --git a/src/mesa/pipe/softpipe/SConscript b/src/mesa/pipe/softpipe/SConscript deleted file mode 100644 index d581ee8d3c..0000000000 --- a/src/mesa/pipe/softpipe/SConscript +++ /dev/null @@ -1,42 +0,0 @@ -Import('*') - -env = env.Clone() - -softpipe = env.ConvenienceLibrary( - target = 'softpipe', - source = [ - 'sp_clear.c', - 'sp_context.c', - 'sp_draw_arrays.c', - 'sp_flush.c', - 'sp_prim_setup.c', - 'sp_prim_vbuf.c', - 'sp_quad_alpha_test.c', - 'sp_quad_blend.c', - 'sp_quad_bufloop.c', - 'sp_quad.c', - 'sp_quad_colormask.c', - 'sp_quad_coverage.c', - 'sp_quad_depth_test.c', - 'sp_quad_earlyz.c', - 'sp_quad_fs.c', - 'sp_quad_occlusion.c', - 'sp_quad_output.c', - 'sp_quad_stencil.c', - 'sp_quad_stipple.c', - 'sp_query.c', - 'sp_state_blend.c', - 'sp_state_clip.c', - 'sp_state_derived.c', - 'sp_state_fs.c', - 'sp_state_rasterizer.c', - 'sp_state_sampler.c', - 'sp_state_surface.c', - 'sp_state_vertex.c', - 'sp_surface.c', - 'sp_tex_sample.c', - 'sp_texture.c', - 'sp_tile_cache.c', - ]) - -Export('softpipe') \ No newline at end of file diff --git a/src/mesa/pipe/softpipe/sp_clear.c b/src/mesa/pipe/softpipe/sp_clear.c deleted file mode 100644 index 8d295a30ca..0000000000 --- a/src/mesa/pipe/softpipe/sp_clear.c +++ /dev/null @@ -1,73 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Brian Paul - */ - - -#include "pipe/p_defines.h" -#include "sp_clear.h" -#include "sp_context.h" -#include "sp_surface.h" -#include "sp_state.h" -#include "sp_tile_cache.h" - - -/** - * Clear the given surface to the specified value. - * No masking, no scissor (clear entire buffer). - */ -void -softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - uint i; - -#if 0 - softpipe_update_derived(softpipe); /* not needed?? */ -#endif - - if (ps == sp_tile_cache_get_surface(softpipe->zsbuf_cache)) { - sp_tile_cache_clear(softpipe->zsbuf_cache, clearValue); -#if TILE_CLEAR_OPTIMIZATION - return; -#endif - } - - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[i])) { - sp_tile_cache_clear(softpipe->cbuf_cache[i], clearValue); - } - } - -#if !TILE_CLEAR_OPTIMIZATION - /* non-cached surface */ - pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); -#endif -} diff --git a/src/mesa/pipe/softpipe/sp_clear.h b/src/mesa/pipe/softpipe/sp_clear.h deleted file mode 100644 index a8ed1c4ecc..0000000000 --- a/src/mesa/pipe/softpipe/sp_clear.h +++ /dev/null @@ -1,43 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Brian Paul - */ - -#ifndef SP_CLEAR_H -#define SP_CLEAR_H - -#include "pipe/p_state.h" -struct pipe_context; - -extern void -softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, - unsigned clearValue); - - -#endif /* SP_CLEAR_H */ diff --git a/src/mesa/pipe/softpipe/sp_context.c b/src/mesa/pipe/softpipe/sp_context.c deleted file mode 100644 index cea6b90104..0000000000 --- a/src/mesa/pipe/softpipe/sp_context.c +++ /dev/null @@ -1,333 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - -#include "pipe/draw/draw_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "sp_clear.h" -#include "sp_context.h" -#include "sp_flush.h" -#include "sp_prim_setup.h" -#include "sp_prim_vbuf.h" -#include "sp_state.h" -#include "sp_surface.h" -#include "sp_tile_cache.h" -#include "sp_texture.h" -#include "sp_winsys.h" -#include "sp_query.h" - - - -/** - * Query format support for creating a texture, drawing surface, etc. - * \param format the format to test - * \param type one of PIPE_TEXTURE, PIPE_SURFACE - */ -static boolean -softpipe_is_format_supported( struct pipe_context *pipe, - enum pipe_format format, uint type ) -{ - switch (type) { - case PIPE_TEXTURE: - /* softpipe supports all texture formats */ - return TRUE; - case PIPE_SURFACE: - /* softpipe supports all (off-screen) surface formats */ - return TRUE; - default: - assert(0); - return FALSE; - } -} - - -/** - * Map any drawing surfaces which aren't already mapped - */ -void -softpipe_map_surfaces(struct softpipe_context *sp) -{ - unsigned i; - - for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); - } - - sp_tile_cache_map_surfaces(sp->zsbuf_cache); -} - - -/** - * Unmap any mapped drawing surfaces - */ -void -softpipe_unmap_surfaces(struct softpipe_context *sp) -{ - uint i; - - for (i = 0; i < sp->framebuffer.num_cbufs; i++) - sp_flush_tile_cache(sp, sp->cbuf_cache[i]); - sp_flush_tile_cache(sp, sp->zsbuf_cache); - - for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); - } - sp_tile_cache_unmap_surfaces(sp->zsbuf_cache); -} - - -static void softpipe_destroy( struct pipe_context *pipe ) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - struct pipe_winsys *ws = pipe->winsys; - uint i; - - draw_destroy( softpipe->draw ); - - softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); - softpipe->quad.earlyz->destroy( softpipe->quad.earlyz ); - softpipe->quad.shade->destroy( softpipe->quad.shade ); - softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); - softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); - softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); - softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); - softpipe->quad.coverage->destroy( softpipe->quad.coverage ); - softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); - softpipe->quad.blend->destroy( softpipe->quad.blend ); - softpipe->quad.colormask->destroy( softpipe->quad.colormask ); - softpipe->quad.output->destroy( softpipe->quad.output ); - - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) - sp_destroy_tile_cache(softpipe->cbuf_cache[i]); - sp_destroy_tile_cache(softpipe->zsbuf_cache); - - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) - sp_destroy_tile_cache(softpipe->tex_cache[i]); - - for (i = 0; i < Elements(softpipe->constants); i++) { - if (softpipe->constants[i].buffer) { - pipe_buffer_reference(ws, &softpipe->constants[i].buffer, NULL); - } - } - - FREE( softpipe ); -} - - -static const char *softpipe_get_name( struct pipe_context *pipe ) -{ - return "softpipe"; -} - -static const char *softpipe_get_vendor( struct pipe_context *pipe ) -{ - return "Tungsten Graphics, Inc."; -} - -static int softpipe_get_param(struct pipe_context *pipe, int param) -{ - switch (param) { - case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; - case PIPE_CAP_NPOT_TEXTURES: - return 1; - case PIPE_CAP_TWO_SIDED_STENCIL: - return 1; - case PIPE_CAP_GLSL: - return 1; - case PIPE_CAP_S3TC: - return 0; - case PIPE_CAP_ANISOTROPIC_FILTER: - return 0; - case PIPE_CAP_POINT_SPRITE: - return 1; - case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; - case PIPE_CAP_OCCLUSION_QUERY: - return 1; - case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 1; - case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 12; /* max 2Kx2K */ - case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: - return 8; /* max 128x128x128 */ - case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 12; /* max 2Kx2K */ - default: - return 0; - } -} - -static float softpipe_get_paramf(struct pipe_context *pipe, int param) -{ - switch (param) { - case PIPE_CAP_MAX_LINE_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_LINE_WIDTH_AA: - return 255.0; /* arbitrary */ - - case PIPE_CAP_MAX_POINT_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_POINT_WIDTH_AA: - return 255.0; /* arbitrary */ - - case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 0.0; - - case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: - return 16.0; /* arbitrary */ - - default: - return 0; - } -} - -struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, - struct softpipe_winsys *softpipe_winsys ) -{ - struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); - uint i; - -#if defined(__i386__) || defined(__386__) - softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; -#else - softpipe->use_sse = FALSE; -#endif - - softpipe->dump_fs = GETENV( "GALLIUM_DUMP_FS" ) != NULL; - - softpipe->pipe.winsys = pipe_winsys; - softpipe->pipe.destroy = softpipe_destroy; - - /* queries */ - softpipe->pipe.is_format_supported = softpipe_is_format_supported; - softpipe->pipe.get_name = softpipe_get_name; - softpipe->pipe.get_vendor = softpipe_get_vendor; - softpipe->pipe.get_param = softpipe_get_param; - softpipe->pipe.get_paramf = softpipe_get_paramf; - - /* state setters */ - softpipe->pipe.create_blend_state = softpipe_create_blend_state; - softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; - softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; - - softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; - softpipe->pipe.bind_sampler_state = softpipe_bind_sampler_state; - softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; - - softpipe->pipe.create_depth_stencil_alpha_state = softpipe_create_depth_stencil_state; - softpipe->pipe.bind_depth_stencil_alpha_state = softpipe_bind_depth_stencil_state; - softpipe->pipe.delete_depth_stencil_alpha_state = softpipe_delete_depth_stencil_state; - - softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; - softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; - softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; - - softpipe->pipe.create_fs_state = softpipe_create_fs_state; - softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; - softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; - - softpipe->pipe.create_vs_state = softpipe_create_vs_state; - softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; - softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; - - softpipe->pipe.set_blend_color = softpipe_set_blend_color; - softpipe->pipe.set_clip_state = softpipe_set_clip_state; - softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; - softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; - softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; - softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; - softpipe->pipe.set_sampler_texture = softpipe_set_sampler_texture; - softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; - - softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; - softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; - - softpipe->pipe.draw_arrays = softpipe_draw_arrays; - softpipe->pipe.draw_elements = softpipe_draw_elements; - - softpipe->pipe.clear = softpipe_clear; - softpipe->pipe.flush = softpipe_flush; - - softpipe_init_query_funcs( softpipe ); - - /* textures */ - softpipe->pipe.texture_create = softpipe_texture_create; - softpipe->pipe.texture_release = softpipe_texture_release; - softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; - - /* - * Alloc caches for accessing drawing surfaces and textures. - * Must be before quad stage setup! - */ - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) - softpipe->cbuf_cache[i] = sp_create_tile_cache(); - softpipe->zsbuf_cache = sp_create_tile_cache(); - - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) - softpipe->tex_cache[i] = sp_create_tile_cache(); - - - /* setup quad rendering stages */ - softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); - softpipe->quad.earlyz = sp_quad_earlyz_stage(softpipe); - softpipe->quad.shade = sp_quad_shade_stage(softpipe); - softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); - softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); - softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); - softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); - softpipe->quad.coverage = sp_quad_coverage_stage(softpipe); - softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); - softpipe->quad.blend = sp_quad_blend_stage(softpipe); - softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); - softpipe->quad.output = sp_quad_output_stage(softpipe); - - softpipe->winsys = softpipe_winsys; - - /* - * Create drawing context and plug our rendering stage into it. - */ - softpipe->draw = draw_create(); - assert(softpipe->draw); - softpipe->setup = sp_draw_render_stage(softpipe); - - if (GETENV( "SP_VBUF" ) != NULL) { - sp_init_vbuf(softpipe); - } - else { - draw_set_rasterize_stage(softpipe->draw, softpipe->setup); - } - - sp_init_surface_functions(softpipe); - - return &softpipe->pipe; -} diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h deleted file mode 100644 index aff8c2cc5d..0000000000 --- a/src/mesa/pipe/softpipe/sp_context.h +++ /dev/null @@ -1,152 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef SP_CONTEXT_H -#define SP_CONTEXT_H - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" - -#include "pipe/draw/draw_vertex.h" - -#include "sp_quad.h" - - -struct softpipe_winsys; -struct softpipe_vbuf_render; -struct draw_context; -struct draw_stage; -struct softpipe_tile_cache; -struct sp_fragment_shader_state; -struct sp_vertex_shader_state; - - -struct softpipe_context { - struct pipe_context pipe; /**< base class */ - struct softpipe_winsys *winsys; /**< window system interface */ - - - /* The most recent drawing state as set by the driver: - */ - const struct pipe_blend_state *blend; - const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct pipe_depth_stencil_alpha_state *depth_stencil; - const struct pipe_rasterizer_state *rasterizer; - const struct sp_fragment_shader_state *fs; - const struct sp_vertex_shader_state *vs; - - struct pipe_blend_color blend_color; - struct pipe_clip_state clip; - struct pipe_constant_buffer constants[2]; - struct pipe_framebuffer_state framebuffer; - struct pipe_poly_stipple poly_stipple; - struct pipe_scissor_state scissor; - struct softpipe_texture *texture[PIPE_MAX_SAMPLERS]; - struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - unsigned dirty; - - /* Counter for occlusion queries. Note this supports overlapping - * queries. - */ - uint64 occlusion_count; - - /* - * Mapped vertex buffers - */ - ubyte *mapped_vbuffer[PIPE_ATTRIB_MAX]; - - /** Mapped constant buffers */ - void *mapped_constants[PIPE_SHADER_TYPES]; - - /** Vertex format */ - struct vertex_info vertex_info; - struct vertex_info vertex_info_vbuf; - - int psize_slot; - -#if 0 - /* Stipple derived state: - */ - ubyte stipple_masks[16][16]; -#endif - - /** Derived from scissor and surface bounds: */ - struct pipe_scissor_state cliprect; - - unsigned line_stipple_counter; - - /** Software quad rendering pipeline */ - struct { - struct quad_stage *polygon_stipple; - struct quad_stage *earlyz; - struct quad_stage *shade; - struct quad_stage *alpha_test; - struct quad_stage *stencil_test; - struct quad_stage *depth_test; - struct quad_stage *occlusion; - struct quad_stage *coverage; - struct quad_stage *bufloop; - struct quad_stage *blend; - struct quad_stage *colormask; - struct quad_stage *output; - - struct quad_stage *first; /**< points to one of the above stages */ - } quad; - - /** The primitive drawing context */ - struct draw_context *draw; - struct draw_stage *setup; - struct draw_stage *vbuf; - struct softpipe_vbuf_render *vbuf_render; - - uint current_cbuf; /**< current color buffer being written to */ - - struct softpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS]; - struct softpipe_tile_cache *zsbuf_cache; - - struct softpipe_tile_cache *tex_cache[PIPE_MAX_SAMPLERS]; - - int use_sse : 1; - int dump_fs : 1; -}; - - - - -static INLINE struct softpipe_context * -softpipe_context( struct pipe_context *pipe ) -{ - return (struct softpipe_context *)pipe; -} - - -#endif /* SP_CONTEXT_H */ diff --git a/src/mesa/pipe/softpipe/sp_draw_arrays.c b/src/mesa/pipe/softpipe/sp_draw_arrays.c deleted file mode 100644 index 71a303a8b5..0000000000 --- a/src/mesa/pipe/softpipe/sp_draw_arrays.c +++ /dev/null @@ -1,164 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Brian Paul - * Keith Whitwell - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_context.h" -#include "pipe/p_winsys.h" - -#include "sp_context.h" -#include "sp_state.h" - -#include "pipe/draw/draw_context.h" - - - -static void -softpipe_map_constant_buffers(struct softpipe_context *sp) -{ - struct pipe_winsys *ws = sp->pipe.winsys; - uint i; - for (i = 0; i < 2; i++) { - if (sp->constants[i].size) - sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - } - - draw_set_mapped_constant_buffer(sp->draw, - sp->mapped_constants[PIPE_SHADER_VERTEX]); -} - -static void -softpipe_unmap_constant_buffers(struct softpipe_context *sp) -{ - struct pipe_winsys *ws = sp->pipe.winsys; - uint i; - for (i = 0; i < 2; i++) { - if (sp->constants[i].size) - ws->buffer_unmap(ws, sp->constants[i].buffer); - sp->mapped_constants[i] = NULL; - } -} - - -boolean -softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, - unsigned start, unsigned count) -{ - return softpipe_draw_elements(pipe, NULL, 0, mode, start, count); -} - - - -/** - * Draw vertex arrays, with optional indexing. - * Basically, map the vertex buffers (and drawing surfaces), then hand off - * the drawing to the 'draw' module. - * - * XXX should the element buffer be specified/bound with a separate function? - */ -boolean -softpipe_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count) -{ - struct softpipe_context *sp = softpipe_context(pipe); - struct draw_context *draw = sp->draw; - unsigned i; - - /* first, check that the primitive is not malformed. It is the - * state tracker's responsibility to do send only correctly formed - * primitives down. It currently isn't doing that though... - */ -#if 1 - count = draw_trim_prim( mode, count ); -#else - if (!draw_validate_prim( mode, count )) - assert(0); -#endif - - - if (sp->dirty) - softpipe_update_derived( sp ); - - softpipe_map_surfaces(sp); - softpipe_map_constant_buffers(sp); - - /* - * Map vertex buffers - */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { - if (sp->vertex_buffer[i].buffer) { - void *buf - = pipe->winsys->buffer_map(pipe->winsys, - sp->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_vertex_buffer(draw, i, buf); - } - } - /* Map index buffer, if present */ - if (indexBuffer) { - void *mapped_indexes - = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); - } - else { - /* no index/element buffer */ - draw_set_mapped_element_buffer(draw, 0, NULL); - } - - - /* draw! */ - draw_arrays(draw, mode, start, count); - - /* - * unmap vertex/index buffers - will cause draw module to flush - */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { - if (sp->vertex_buffer[i].buffer) { - draw_set_mapped_vertex_buffer(draw, i, NULL); - pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); - } - } - if (indexBuffer) { - draw_set_mapped_element_buffer(draw, 0, NULL); - pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); - } - - - /* Note: leave drawing surfaces mapped */ - softpipe_unmap_constant_buffers(sp); - - return TRUE; -} diff --git a/src/mesa/pipe/softpipe/sp_flush.c b/src/mesa/pipe/softpipe/sp_flush.c deleted file mode 100644 index ced0d5d098..0000000000 --- a/src/mesa/pipe/softpipe/sp_flush.c +++ /dev/null @@ -1,76 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - - -#include "pipe/p_defines.h" -#include "pipe/draw/draw_context.h" -#include "sp_flush.h" -#include "sp_context.h" -#include "sp_surface.h" -#include "sp_state.h" -#include "sp_tile_cache.h" -#include "sp_winsys.h" - - -/* There will be actual work to do here. In future we may want a - * fence-like interface instead of finish, and perhaps flush will take - * flags to indicate what type of flush is required. - */ -void -softpipe_flush( struct pipe_context *pipe, - unsigned flags ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - uint i; - - draw_flush(softpipe->draw); - - /* - flush the quad pipeline - * - flush the texture cache - * - flush the render cache - */ - - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) - if (softpipe->cbuf_cache[i]) - sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]); - - if (softpipe->zsbuf_cache) - sp_flush_tile_cache(softpipe, softpipe->zsbuf_cache); - - /* Need this call for hardware buffers before swapbuffers. - * - * there should probably be another/different flush-type function - * that's called before swapbuffers because we don't always want - * to unmap surfaces when flushing. - */ - softpipe_unmap_surfaces(softpipe); -} - diff --git a/src/mesa/pipe/softpipe/sp_flush.h b/src/mesa/pipe/softpipe/sp_flush.h deleted file mode 100644 index 34ec617866..0000000000 --- a/src/mesa/pipe/softpipe/sp_flush.h +++ /dev/null @@ -1,35 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SP_FLUSH_H -#define SP_FLUSH_H - -struct pipe_context; - -void softpipe_flush(struct pipe_context *pipe, unsigned flags ); - -#endif diff --git a/src/mesa/pipe/softpipe/sp_headers.h b/src/mesa/pipe/softpipe/sp_headers.h deleted file mode 100644 index 0ae31d8796..0000000000 --- a/src/mesa/pipe/softpipe/sp_headers.h +++ /dev/null @@ -1,82 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef SP_HEADERS_H -#define SP_HEADERS_H - -#include "pipe/tgsi/exec/tgsi_exec.h" - -#define PRIM_POINT 1 -#define PRIM_LINE 2 -#define PRIM_TRI 3 - - -/* The rasterizer generates 2x2 quads of fragment and feeds them to - * the current fp_machine (see below). - * Remember that Y=0=top with Y increasing down the window. - */ -#define QUAD_TOP_LEFT 0 -#define QUAD_TOP_RIGHT 1 -#define QUAD_BOTTOM_LEFT 2 -#define QUAD_BOTTOM_RIGHT 3 - -#define MASK_TOP_LEFT (1 << QUAD_TOP_LEFT) -#define MASK_TOP_RIGHT (1 << QUAD_TOP_RIGHT) -#define MASK_BOTTOM_LEFT (1 << QUAD_BOTTOM_LEFT) -#define MASK_BOTTOM_RIGHT (1 << QUAD_BOTTOM_RIGHT) -#define MASK_ALL 0xf - - -/** - * Encodes everything we need to know about a 2x2 pixel block. Uses - * "Channel-Serial" or "SoA" layout. - */ -struct quad_header { - int x0; - int y0; - unsigned mask:4; - unsigned facing:1; /**< Front (0) or back (1) facing? */ - unsigned prim:2; /**< PRIM_POINT, LINE, TRI */ - - struct { - float color[NUM_CHANNELS][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ - float depth[QUAD_SIZE]; - } outputs; - - float coverage[QUAD_SIZE]; /** fragment coverage for antialiasing */ - - const struct tgsi_interp_coef *coef; - const struct tgsi_interp_coef *posCoef; - - unsigned nr_attrs; -}; - - -#endif /* SP_HEADERS_H */ diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.c b/src/mesa/pipe/softpipe/sp_prim_setup.c deleted file mode 100644 index 2772048661..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_setup.c +++ /dev/null @@ -1,1247 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Primitive rasterization/rendering (points, lines, triangles) - * - * \author Keith Whitwell - * \author Brian Paul - */ - - -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_quad.h" -#include "sp_state.h" -#include "sp_prim_setup.h" -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vertex.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" - -#define DEBUG_VERTS 0 - -/** - * Triangle edge info - */ -struct edge { - float dx; /**< X(v1) - X(v0), used only during setup */ - float dy; /**< Y(v1) - Y(v0), used only during setup */ - float dxdy; /**< dx/dy */ - float sx, sy; /**< first sample point coord */ - int lines; /**< number of lines on this edge */ -}; - - -/** - * Triangle setup info (derived from draw_stage). - * Also used for line drawing (taking some liberties). - */ -struct setup_stage { - struct draw_stage stage; /**< This must be first (base class) */ - - struct softpipe_context *softpipe; - - /* Vertices are just an array of floats making up each attribute in - * turn. Currently fixed at 4 floats, but should change in time. - * Codegen will help cope with this. - */ - const struct vertex_header *vmax; - const struct vertex_header *vmid; - const struct vertex_header *vmin; - const struct vertex_header *vprovoke; - - struct edge ebot; - struct edge etop; - struct edge emaj; - - float oneoverarea; - - struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; - struct tgsi_interp_coef posCoef; /* For Z, W */ - struct quad_header quad; - - struct { - int left[2]; /**< [0] = row0, [1] = row1 */ - int right[2]; - int y; - unsigned y_flags; - unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ - } span; -}; - - - -/** - * Basically a cast wrapper. - */ -static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) -{ - return (struct setup_stage *)stage; -} - - -/** - * Clip setup->quad against the scissor/surface bounds. - */ -static INLINE void -quad_clip(struct setup_stage *setup) -{ - const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; - - if (setup->quad.x0 >= maxx || - setup->quad.y0 >= maxy || - setup->quad.x0 + 1 < minx || - setup->quad.y0 + 1 < miny) { - /* totally clipped */ - setup->quad.mask = 0x0; - return; - } - if (setup->quad.x0 < minx) - setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup->quad.y0 < miny) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - if (setup->quad.x0 == maxx - 1) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup->quad.y0 == maxy - 1) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); -} - - -/** - * Emit a quad (pass to next stage) with clipping. - */ -static INLINE void -clip_emit_quad(struct setup_stage *setup) -{ - quad_clip(setup); - if (setup->quad.mask) { - struct softpipe_context *sp = setup->softpipe; - sp->quad.first->run(sp->quad.first, &setup->quad); - } -} - - -/** - * Emit a quad (pass to next stage). No clipping is done. - */ -static INLINE void -emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) -{ - struct softpipe_context *sp = setup->softpipe; - setup->quad.x0 = x; - setup->quad.y0 = y; - setup->quad.mask = mask; - sp->quad.first->run(sp->quad.first, &setup->quad); -} - - -/** - * Given an X or Y coordinate, return the block/quad coordinate that it - * belongs to. - */ -static INLINE int block( int x ) -{ - return x & ~1; -} - - -/** - * Compute mask which indicates which pixels in the 2x2 quad are actually inside - * the triangle's bounds. - * - * this is pretty nasty... may need to rework flush_spans again to - * fix it, if possible. - */ -static unsigned calculate_mask( struct setup_stage *setup, int x ) -{ - unsigned mask = 0x0; - - if (x >= setup->span.left[0] && x < setup->span.right[0]) - mask |= MASK_TOP_LEFT; - - if (x >= setup->span.left[1] && x < setup->span.right[1]) - mask |= MASK_BOTTOM_LEFT; - - if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) - mask |= MASK_TOP_RIGHT; - - if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) - mask |= MASK_BOTTOM_RIGHT; - - return mask; -} - - -/** - * Render a horizontal span of quads - */ -static void flush_spans( struct setup_stage *setup ) -{ - int minleft, maxright; - int x; - - switch (setup->span.y_flags) { - case 0x3: - /* both odd and even lines written (both quad rows) */ - minleft = MIN2(setup->span.left[0], setup->span.left[1]); - maxright = MAX2(setup->span.right[0], setup->span.right[1]); - break; - - case 0x1: - /* only even line written (quad top row) */ - minleft = setup->span.left[0]; - maxright = setup->span.right[0]; - break; - - case 0x2: - /* only odd line written (quad bottom row) */ - minleft = setup->span.left[1]; - maxright = setup->span.right[1]; - break; - - default: - return; - } - - /* XXX this loop could be moved into the above switch cases and - * calculate_mask() could be simplified a bit... - */ - for (x = block(minleft); x <= block(maxright); x += 2) { - emit_quad( setup, x, setup->span.y, - calculate_mask( setup, x ) ); - } - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; -} - -#if DEBUG_VERTS -static void print_vertex(const struct setup_stage *setup, - const struct vertex_header *v) -{ - int i; - debug_printf("Vertex: (%p)\n", v); - for (i = 0; i < setup->quad.nr_attrs; i++) { - debug_printf(" %d: %f %f %f %f\n", i, - v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); - } -} -#endif - -static boolean setup_sort_vertices( struct setup_stage *setup, - const struct prim_header *prim ) -{ - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - const struct vertex_header *v2 = prim->v[2]; - -#if DEBUG_VERTS - debug_printf("Triangle:\n"); - print_vertex(setup, v0); - print_vertex(setup, v1); - print_vertex(setup, v2); -#endif - - setup->vprovoke = v2; - - /* determine bottom to top order of vertices */ - { - float y0 = v0->data[0][1]; - float y1 = v1->data[0][1]; - float y2 = v2->data[0][1]; - if (y0 <= y1) { - if (y1 <= y2) { - /* y0<=y1<=y2 */ - setup->vmin = v0; - setup->vmid = v1; - setup->vmax = v2; - } - else if (y2 <= y0) { - /* y2<=y0<=y1 */ - setup->vmin = v2; - setup->vmid = v0; - setup->vmax = v1; - } - else { - /* y0<=y2<=y1 */ - setup->vmin = v0; - setup->vmid = v2; - setup->vmax = v1; - } - } - else { - if (y0 <= y2) { - /* y1<=y0<=y2 */ - setup->vmin = v1; - setup->vmid = v0; - setup->vmax = v2; - } - else if (y2 <= y1) { - /* y2<=y1<=y0 */ - setup->vmin = v2; - setup->vmid = v1; - setup->vmax = v0; - } - else { - /* y1<=y2<=y0 */ - setup->vmin = v1; - setup->vmid = v2; - setup->vmax = v0; - } - } - } - - setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; - setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; - setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; - - /* - * Compute triangle's area. Use 1/area to compute partial - * derivatives of attributes later. - * - * The area will be the same as prim->det, but the sign may be - * different depending on how the vertices get sorted above. - * - * To determine whether the primitive is front or back facing we - * use the prim->det value because its sign is correct. - */ - { - const float area = (setup->emaj.dx * setup->ebot.dy - - setup->ebot.dx * setup->emaj.dy); - - setup->oneoverarea = 1.0f / area; - /* - debug_printf("%s one-over-area %f area %f det %f\n", - __FUNCTION__, setup->oneoverarea, area, prim->det ); - */ - } - - /* We need to know if this is a front or back-facing triangle for: - * - the GLSL gl_FrontFacing fragment attribute (bool) - * - two-sided stencil test - */ - setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); - - return TRUE; -} - - -/** - * Compute a0 for a constant-valued coefficient (GL_FLAT shading). - * The value value comes from vertex->data[slot][i]. - * The result will be put into setup->coef[slot].a0[i]. - * \param slot which attribute slot - * \param i which component of the slot (0..3) - */ -static void const_coeff( struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - assert(i <= 3); - - coef->dadx[i] = 0; - coef->dady[i] = 0; - - /* need provoking vertex info! - */ - coef->a0[i] = setup->vprovoke->data[vertSlot][i]; -} - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a triangle. - */ -static void tri_linear_coeff( struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - float botda = setup->vmid->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; - float majda = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - float dadx = a * setup->oneoverarea; - float dady = b * setup->oneoverarea; - - assert(i <= 3); - - coef->dadx[i] = dadx; - coef->dady[i] = dady; - - /* calculate a0 as the value which would be sampled for the - * fragment at (0,0), taking into account that we want to sample at - * pixel centers, in other words (0.5, 0.5). - * - * this is neat but unfortunately not a good way to do things for - * triangles with very large values of dadx or dady as it will - * result in the subtraction and re-addition from a0 of a very - * large number, which means we'll end up loosing a lot of the - * fractional bits and precision from a0. the way to fix this is - * to define a0 as the sample at a pixel center somewhere near vmin - * instead - i'll switch to this later. - */ - coef->a0[i] = (setup->vmin->data[vertSlot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); - - /* - debug_printf("attr[%d].%c: %f dx:%f dy:%f\n", - slot, "xyzw"[i], - setup->coef[slot].a0[i], - setup->coef[slot].dadx[i], - setup->coef[slot].dady[i]); - */ -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a triangle. - * We basically multiply the vertex value by 1/w before computing - * the plane coefficients (a0, dadx, dady). - * Later, when we compute the value at a particular fragment position we'll - * divide the interpolated value by the interpolated W at that fragment. - */ -static void tri_persp_coeff( struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - /* premultiply by 1/w (v->data[0][3] is always W): - */ - float mina = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; - float mida = setup->vmid->data[vertSlot][i] * setup->vmid->data[0][3]; - float maxa = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; - float botda = mida - mina; - float majda = maxa - mina; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - float dadx = a * setup->oneoverarea; - float dady = b * setup->oneoverarea; - - /* - debug_printf("tri persp %d,%d: %f %f %f\n", vertSlot, i, - setup->vmin->data[vertSlot][i], - setup->vmid->data[vertSlot][i], - setup->vmax->data[vertSlot][i] - ); - */ - assert(i <= 3); - - coef->dadx[i] = dadx; - coef->dady[i] = dady; - coef->a0[i] = (mina - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); -} - - -/** - * Special coefficient setup for gl_FragCoord. - * X and Y are trivial, though Y has to be inverted for OpenGL. - * Z and W are copied from posCoef which should have already been computed. - * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask. - */ -static void -setup_fragcoord_coeff(struct setup_stage *setup) -{ - /*X*/ - setup->coef[0].a0[0] = 0; - setup->coef[0].dadx[0] = 1.0; - setup->coef[0].dady[0] = 0.0; - /*Y*/ - if (setup->softpipe->rasterizer->origin_lower_left) { - /* y=0=bottom */ - const int winHeight = setup->softpipe->framebuffer.cbufs[0]->height; - setup->coef[0].a0[1] = (float) (winHeight - 1); - setup->coef[0].dady[1] = -1.0; - } - else { - /* y=0=top */ - setup->coef[0].a0[1] = 0.0; - setup->coef[0].dady[1] = 1.0; - } - setup->coef[0].dadx[1] = 0.0; - /*Z*/ - setup->coef[0].a0[2] = setup->posCoef.a0[2]; - setup->coef[0].dadx[2] = setup->posCoef.dadx[2]; - setup->coef[0].dady[2] = setup->posCoef.dady[2]; - /*W*/ - setup->coef[0].a0[3] = setup->posCoef.a0[3]; - setup->coef[0].dadx[3] = setup->posCoef.dadx[3]; - setup->coef[0].dady[3] = setup->posCoef.dady[3]; -} - - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. - */ -static void setup_tri_coefficients( struct setup_stage *setup ) -{ - struct softpipe_context *softpipe = setup->softpipe; - const struct pipe_shader_state *fs = &softpipe->fs->shader; - const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); - uint fragSlot; - - /* z and w are done by linear interpolation: - */ - tri_linear_coeff(setup, &setup->posCoef, 0, 2); - tri_linear_coeff(setup, &setup->posCoef, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { - const uint vertSlot = vinfo->src_index[fragSlot]; - uint j; - - switch (vinfo->interp_mode[fragSlot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_POS: - assert(fragSlot == 0); - setup_fragcoord_coeff(setup); - break; - default: - assert(0); - } - - if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - } -} - - - -static void setup_tri_edges( struct setup_stage *setup ) -{ - float vmin_x = setup->vmin->data[0][0] + 0.5f; - float vmid_x = setup->vmid->data[0][0] + 0.5f; - - float vmin_y = setup->vmin->data[0][1] - 0.5f; - float vmid_y = setup->vmid->data[0][1] - 0.5f; - float vmax_y = setup->vmax->data[0][1] - 0.5f; - - setup->emaj.sy = CEILF(vmin_y); - setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); - setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; - setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; - - setup->etop.sy = CEILF(vmid_y); - setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); - setup->etop.dxdy = setup->etop.dx / setup->etop.dy; - setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; - - setup->ebot.sy = CEILF(vmin_y); - setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); - setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; - setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; -} - - -/** - * Render the upper or lower half of a triangle. - * Scissoring/cliprect is applied here too. - */ -static void subtriangle( struct setup_stage *setup, - struct edge *eleft, - struct edge *eright, - unsigned lines ) -{ - const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; - int y, start_y, finish_y; - int sy = (int)eleft->sy; - - assert((int)eleft->sy == (int) eright->sy); - - /* clip top/bottom */ - start_y = sy; - finish_y = sy + lines; - - if (start_y < miny) - start_y = miny; - - if (finish_y > maxy) - finish_y = maxy; - - start_y -= sy; - finish_y -= sy; - - /* - debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); - */ - - for (y = start_y; y < finish_y; y++) { - - /* avoid accumulating adds as floats don't have the precision to - * accurately iterate large triangle edges that way. luckily we - * can just multiply these days. - * - * this is all drowned out by the attribute interpolation anyway. - */ - int left = (int)(eleft->sx + y * eleft->dxdy); - int right = (int)(eright->sx + y * eright->dxdy); - - /* clip left/right */ - if (left < minx) - left = minx; - if (right > maxx) - right = maxx; - - if (left < right) { - int _y = sy + y; - if (block(_y) != setup->span.y) { - flush_spans(setup); - setup->span.y = block(_y); - } - - setup->span.left[_y&1] = left; - setup->span.right[_y&1] = right; - setup->span.y_flags |= 1<<(_y&1); - } - } - - - /* save the values so that emaj can be restarted: - */ - eleft->sx += lines * eleft->dxdy; - eright->sx += lines * eright->dxdy; - eleft->sy += lines; - eright->sy += lines; -} - - -/** - * Do setup for triangle rasterization, then render the triangle. - */ -static void setup_tri( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct setup_stage *setup = setup_stage( stage ); - - /* - debug_printf("%s\n", __FUNCTION__ ); - */ - - setup_sort_vertices( setup, prim ); - setup_tri_coefficients( setup ); - setup_tri_edges( setup ); - - setup->quad.prim = PRIM_TRI; - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; - /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ - - /* init_constant_attribs( setup ); */ - - if (setup->oneoverarea < 0.0) { - /* emaj on left: - */ - subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); - subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); - } - else { - /* emaj on right: - */ - subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); - subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); - } - - flush_spans( setup ); -} - - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a line. - */ -static void -line_linear_coeff(struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - const float da = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; - const float dadx = da * setup->emaj.dx * setup->oneoverarea; - const float dady = da * setup->emaj.dy * setup->oneoverarea; - coef->dadx[i] = dadx; - coef->dady[i] = dady; - coef->a0[i] = (setup->vmin->data[vertSlot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a line. - */ -static void -line_persp_coeff(struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - /* XXX double-check/verify this arithmetic */ - const float a0 = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; - const float a1 = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; - const float da = a1 - a0; - const float dadx = da * setup->emaj.dx * setup->oneoverarea; - const float dady = da * setup->emaj.dy * setup->oneoverarea; - coef->dadx[i] = dadx; - coef->dady[i] = dady; - coef->a0[i] = (setup->vmin->data[vertSlot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); -} - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmax are initialized. - */ -static INLINE void -setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) -{ - struct softpipe_context *softpipe = setup->softpipe; - const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; - const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); - uint fragSlot; - - /* use setup->vmin, vmax to point to vertices */ - setup->vprovoke = prim->v[1]; - setup->vmin = prim->v[0]; - setup->vmax = prim->v[1]; - - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - /* NOTE: this is not really 1/area */ - setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx + - setup->emaj.dy * setup->emaj.dy); - - /* z and w are done by linear interpolation: - */ - line_linear_coeff(setup, &setup->posCoef, 0, 2); - line_linear_coeff(setup, &setup->posCoef, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { - const uint vertSlot = vinfo->src_index[fragSlot]; - uint j; - - switch (vinfo->interp_mode[fragSlot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_POS: - assert(fragSlot == 0); - assert(0); /* XXX fix this: */ - setup_fragcoord_coeff(setup); - break; - default: - assert(0); - } - - if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - } -} - - -/** - * Plot a pixel in a line segment. - */ -static INLINE void -plot(struct setup_stage *setup, int x, int y) -{ - const int iy = y & 1; - const int ix = x & 1; - const int quadX = x - ix; - const int quadY = y - iy; - const int mask = (1 << ix) << (2 * iy); - - if (quadX != setup->quad.x0 || - quadY != setup->quad.y0) - { - /* flush prev quad, start new quad */ - - if (setup->quad.x0 != -1) - clip_emit_quad(setup); - - setup->quad.x0 = quadX; - setup->quad.y0 = quadY; - setup->quad.mask = 0x0; - } - - setup->quad.mask |= mask; -} - - -/** - * Do setup for line rasterization, then render the line. - * Single-pixel width, no stipple, etc. We rely on the 'draw' module - * to handle stippling and wide lines. - */ -static void -setup_line(struct draw_stage *stage, struct prim_header *prim) -{ - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - struct setup_stage *setup = setup_stage( stage ); - int x0 = (int) v0->data[0][0]; - int x1 = (int) v1->data[0][0]; - int y0 = (int) v0->data[0][1]; - int y1 = (int) v1->data[0][1]; - int dx = x1 - x0; - int dy = y1 - y0; - int xstep, ystep; - - if (dx == 0 && dy == 0) - return; - - setup_line_coefficients(setup, prim); - - if (dx < 0) { - dx = -dx; /* make positive */ - xstep = -1; - } - else { - xstep = 1; - } - - if (dy < 0) { - dy = -dy; /* make positive */ - ystep = -1; - } - else { - ystep = 1; - } - - assert(dx >= 0); - assert(dy >= 0); - - setup->quad.x0 = setup->quad.y0 = -1; - setup->quad.mask = 0x0; - setup->quad.prim = PRIM_LINE; - /* XXX temporary: set coverage to 1.0 so the line appears - * if AA mode happens to be enabled. - */ - setup->quad.coverage[0] = - setup->quad.coverage[1] = - setup->quad.coverage[2] = - setup->quad.coverage[3] = 1.0; - - if (dx > dy) { - /*** X-major line ***/ - int i; - const int errorInc = dy + dy; - int error = errorInc - dx; - const int errorDec = error - dx; - - for (i = 0; i < dx; i++) { - plot(setup, x0, y0); - - x0 += xstep; - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - y0 += ystep; - } - } - } - else { - /*** Y-major line ***/ - int i; - const int errorInc = dx + dx; - int error = errorInc - dy; - const int errorDec = error - dy; - - for (i = 0; i < dy; i++) { - plot(setup, x0, y0); - - y0 += ystep; - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - x0 += xstep; - } - } - } - - /* draw final quad */ - if (setup->quad.mask) { - clip_emit_quad(setup); - } -} - - -static void -point_persp_coeff(struct setup_stage *setup, - const struct vertex_header *vert, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - assert(i <= 3); - coef->dadx[i] = 0.0F; - coef->dady[i] = 0.0F; - coef->a0[i] = vert->data[vertSlot][i] * vert->data[0][3]; -} - - -/** - * Do setup for point rasterization, then render the point. - * Round or square points... - * XXX could optimize a lot for 1-pixel points. - */ -static void -setup_point(struct draw_stage *stage, struct prim_header *prim) -{ - struct setup_stage *setup = setup_stage( stage ); - struct softpipe_context *softpipe = setup->softpipe; - const struct pipe_shader_state *fs = &softpipe->fs->shader; - const struct vertex_header *v0 = prim->v[0]; - const int sizeAttr = setup->softpipe->psize_slot; - const float size - = sizeAttr > 0 ? v0->data[sizeAttr][0] - : setup->softpipe->rasterizer->point_size; - const float halfSize = 0.5F * size; - const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth; - const float x = v0->data[0][0]; /* Note: data[0] is always position */ - const float y = v0->data[0][1]; - const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); - uint fragSlot; - - /* For points, all interpolants are constant-valued. - * However, for point sprites, we'll need to setup texcoords appropriately. - * XXX: which coefficients are the texcoords??? - * We may do point sprites as textured quads... - * - * KW: We don't know which coefficients are texcoords - ultimately - * the choice of what interpolation mode to use for each attribute - * should be determined by the fragment program, using - * per-attribute declaration statements that include interpolation - * mode as a parameter. So either the fragment program will have - * to be adjusted for pointsprite vs normal point behaviour, or - * otherwise a special interpolation mode will have to be defined - * which matches the required behaviour for point sprites. But - - * the latter is not a feature of normal hardware, and as such - * probably should be ruled out on that basis. - */ - setup->vprovoke = prim->v[0]; - - /* setup Z, W */ - const_coeff(setup, &setup->posCoef, 0, 2); - const_coeff(setup, &setup->posCoef, 0, 3); - - for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { - const uint vertSlot = vinfo->src_index[fragSlot]; - uint j; - - switch (vinfo->interp_mode[fragSlot]) { - case INTERP_CONSTANT: - /* fall-through */ - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - point_persp_coeff(setup, setup->vprovoke, - &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_POS: - assert(fragSlot == 0); - assert(0); /* XXX fix this: */ - setup_fragcoord_coeff(setup); - break; - default: - assert(0); - } - - if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - } - - setup->quad.prim = PRIM_POINT; - - if (halfSize <= 0.5 && !round) { - /* special case for 1-pixel points */ - const int ix = ((int) x) & 1; - const int iy = ((int) y) & 1; - setup->quad.x0 = (int) x - ix; - setup->quad.y0 = (int) y - iy; - setup->quad.mask = (1 << ix) << (2 * iy); - clip_emit_quad(setup); - } - else { - if (round) { - /* rounded points */ - const int ixmin = block((int) (x - halfSize)); - const int ixmax = block((int) (x + halfSize)); - const int iymin = block((int) (y - halfSize)); - const int iymax = block((int) (y + halfSize)); - const float rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ - const float rmax = halfSize + 0.7071F; - const float rmin2 = MAX2(0.0F, rmin * rmin); - const float rmax2 = rmax * rmax; - const float cscale = 1.0F / (rmax2 - rmin2); - int ix, iy; - - for (iy = iymin; iy <= iymax; iy += 2) { - for (ix = ixmin; ix <= ixmax; ix += 2) { - float dx, dy, dist2, cover; - - setup->quad.mask = 0x0; - - dx = (ix + 0.5f) - x; - dy = (iy + 0.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_TOP_LEFT; - } - - dx = (ix + 1.5f) - x; - dy = (iy + 0.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_TOP_RIGHT; - } - - dx = (ix + 0.5f) - x; - dy = (iy + 1.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_BOTTOM_LEFT; - } - - dx = (ix + 1.5f) - x; - dy = (iy + 1.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_BOTTOM_RIGHT; - } - - if (setup->quad.mask) { - setup->quad.x0 = ix; - setup->quad.y0 = iy; - clip_emit_quad(setup); - } - } - } - } - else { - /* square points */ - const int xmin = (int) (x + 0.75 - halfSize); - const int ymin = (int) (y + 0.25 - halfSize); - const int xmax = xmin + (int) size; - const int ymax = ymin + (int) size; - /* XXX could apply scissor to xmin,ymin,xmax,ymax now */ - const int ixmin = block(xmin); - const int ixmax = block(xmax - 1); - const int iymin = block(ymin); - const int iymax = block(ymax - 1); - int ix, iy; - - /* - debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax); - */ - for (iy = iymin; iy <= iymax; iy += 2) { - uint rowMask = 0xf; - if (iy < ymin) { - /* above the top edge */ - rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - } - if (iy + 1 >= ymax) { - /* below the bottom edge */ - rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); - } - - for (ix = ixmin; ix <= ixmax; ix += 2) { - uint mask = rowMask; - - if (ix < xmin) { - /* fragment is past left edge of point, turn off left bits */ - mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - } - if (ix + 1 >= xmax) { - /* past the right edge */ - mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - } - - setup->quad.mask = mask; - setup->quad.x0 = ix; - setup->quad.y0 = iy; - clip_emit_quad(setup); - } - } - } - } -} - - - -static void setup_begin( struct draw_stage *stage ) -{ - struct setup_stage *setup = setup_stage(stage); - struct softpipe_context *sp = setup->softpipe; - const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; - - setup->quad.nr_attrs = fs->num_inputs; - - sp->quad.first->begin(sp->quad.first); - - stage->point = setup_point; - stage->line = setup_line; - stage->tri = setup_tri; -} - - -static void setup_first_point( struct draw_stage *stage, - struct prim_header *header ) -{ - setup_begin(stage); - stage->point( stage, header ); -} - -static void setup_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - setup_begin(stage); - stage->line( stage, header ); -} - - -static void setup_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - setup_begin(stage); - stage->tri( stage, header ); -} - - - -static void setup_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->point = setup_first_point; - stage->line = setup_first_line; - stage->tri = setup_first_tri; -} - - -static void reset_stipple_counter( struct draw_stage *stage ) -{ -} - - -static void render_destroy( struct draw_stage *stage ) -{ - FREE( stage ); -} - - -/** - * Create a new primitive setup/render stage. - */ -struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) -{ - struct setup_stage *setup = CALLOC_STRUCT(setup_stage); - - setup->softpipe = softpipe; - setup->stage.draw = softpipe->draw; - setup->stage.point = setup_first_point; - setup->stage.line = setup_first_line; - setup->stage.tri = setup_first_tri; - setup->stage.flush = setup_flush; - setup->stage.reset_stipple_counter = reset_stipple_counter; - setup->stage.destroy = render_destroy; - - setup->quad.coef = setup->coef; - setup->quad.posCoef = &setup->posCoef; - - return &setup->stage; -} diff --git a/src/mesa/pipe/softpipe/sp_prim_setup.h b/src/mesa/pipe/softpipe/sp_prim_setup.h deleted file mode 100644 index f3e8a79dd9..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_setup.h +++ /dev/null @@ -1,79 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef SP_PRIM_SETUP_H -#define SP_PRIM_SETUP_H - - -/** - * vbuf is a special stage to gather the stream of triangles, lines, points - * together and reconstruct vertex buffers for hardware upload. - * - * First attempt, work in progress. - * - * TODO: - * - separate out vertex buffer building and primitive emit, ie >1 draw per vb. - * - tell vbuf stage how to build hw vertices directly - * - pass vbuf stage a buffer pointer for direct emit to agp/vram. - * - * - * - * Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - - -struct draw_stage; -struct softpipe_context; - - -typedef void (*vbuf_draw_func)( struct pipe_context *pipe, - unsigned prim, - const ushort *elements, - unsigned nr_elements, - const void *vertex_buffer, - unsigned nr_vertices ); - - -extern struct draw_stage * -sp_draw_render_stage( struct softpipe_context *softpipe ); - - -extern struct draw_stage * -sp_draw_vbuf_stage( struct draw_context *draw_context, - struct pipe_context *pipe, - vbuf_draw_func draw ); - - -#endif /* SP_PRIM_SETUP_H */ diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.c b/src/mesa/pipe/softpipe/sp_prim_vbuf.c deleted file mode 100644 index 7f71fdb6a9..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.c +++ /dev/null @@ -1,221 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Post-transform vertex buffering. This is an optional part of the - * softpipe rendering pipeline. - * Probably not desired in general, but useful for testing/debuggin. - * Enabled/Disabled with SP_VBUF env var. - * - * Authors - * Brian Paul - */ - - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_prim_vbuf.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vbuf.h" - - -#define SP_MAX_VBUF_INDEXES 1024 -#define SP_MAX_VBUF_SIZE 4096 - - -/** - * Subclass of vbuf_render. - */ -struct softpipe_vbuf_render -{ - struct vbuf_render base; - struct softpipe_context *softpipe; - uint prim; - uint vertex_size; - void *vertex_buffer; -}; - - -/** cast wrapper */ -static struct softpipe_vbuf_render * -softpipe_vbuf_render(struct vbuf_render *vbr) -{ - return (struct softpipe_vbuf_render *) vbr; -} - - - -static const struct vertex_info * -sp_vbuf_get_vertex_info(struct vbuf_render *vbr) -{ - struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - return softpipe_get_vbuf_vertex_info(cvbr->softpipe); -} - - -static void * -sp_vbuf_allocate_vertices(struct vbuf_render *vbr, - ushort vertex_size, ushort nr_vertices) -{ - struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - assert(!cvbr->vertex_buffer); - cvbr->vertex_buffer = align_malloc(vertex_size * nr_vertices, 16); - cvbr->vertex_size = vertex_size; - return cvbr->vertex_buffer; -} - - -static void -sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, - unsigned vertex_size, unsigned vertices_used) -{ - struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - align_free(vertices); - assert(vertices == cvbr->vertex_buffer); - cvbr->vertex_buffer = NULL; -} - - -static void -sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) -{ - struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - cvbr->prim = prim; -} - - -/** - * Recalculate prim's determinant. - * XXX is this needed? - */ -static void -calc_det(struct prim_header *header) -{ - /* Window coords: */ - const float *v0 = header->v[0]->data[0]; - const float *v1 = header->v[1]->data[0]; - const float *v2 = header->v[2]->data[0]; - - /* edge vectors e = v0 - v2, f = v1 - v2 */ - const float ex = v0[0] - v2[0]; - const float ey = v0[1] - v2[1]; - const float fx = v1[0] - v2[0]; - const float fy = v1[1] - v2[1]; - - /* det = cross(e,f).z */ - header->det = ex * fy - ey * fx; -} - - -static void -sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) -{ - struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - struct softpipe_context *softpipe = cvbr->softpipe; - struct draw_stage *setup = softpipe->setup; - struct prim_header prim; - unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); - unsigned i, j; - void *vertex_buffer = cvbr->vertex_buffer; - - prim.det = 0; - prim.reset_line_stipple = 0; - prim.edgeflags = 0; - prim.pad = 0; - - switch (cvbr->prim) { - case PIPE_PRIM_TRIANGLES: - for (i = 0; i < nr_indices; i += 3) { - for (j = 0; j < 3; j++) - prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + - indices[i+j] * vertex_size); - - calc_det(&prim); - setup->tri( setup, &prim ); - } - break; - - case PIPE_PRIM_LINES: - for (i = 0; i < nr_indices; i += 2) { - for (j = 0; j < 2; j++) - prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + - indices[i+j] * vertex_size); - - setup->line( setup, &prim ); - } - break; - - case PIPE_PRIM_POINTS: - for (i = 0; i < nr_indices; i++) { - prim.v[0] = (struct vertex_header *)((char *)vertex_buffer + - indices[i] * vertex_size); - setup->point( setup, &prim ); - } - break; - } - - setup->flush( setup, 0 ); -} - - -static void -sp_vbuf_destroy(struct vbuf_render *vbr) -{ - struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - cvbr->softpipe->vbuf_render = NULL; - FREE(cvbr); -} - - -/** - * Initialize the post-transform vertex buffer information for the given - * context. - */ -void -sp_init_vbuf(struct softpipe_context *sp) -{ - assert(sp->draw); - - sp->vbuf_render = CALLOC_STRUCT(softpipe_vbuf_render); - - sp->vbuf_render->base.max_indices = SP_MAX_VBUF_INDEXES; - sp->vbuf_render->base.max_vertex_buffer_bytes = SP_MAX_VBUF_SIZE; - - sp->vbuf_render->base.get_vertex_info = sp_vbuf_get_vertex_info; - sp->vbuf_render->base.allocate_vertices = sp_vbuf_allocate_vertices; - sp->vbuf_render->base.set_primitive = sp_vbuf_set_primitive; - sp->vbuf_render->base.draw = sp_vbuf_draw; - sp->vbuf_render->base.release_vertices = sp_vbuf_release_vertices; - sp->vbuf_render->base.destroy = sp_vbuf_destroy; - - sp->vbuf_render->softpipe = sp; - - sp->vbuf = draw_vbuf_stage(sp->draw, &sp->vbuf_render->base); - - draw_set_rasterize_stage(sp->draw, sp->vbuf); -} diff --git a/src/mesa/pipe/softpipe/sp_prim_vbuf.h b/src/mesa/pipe/softpipe/sp_prim_vbuf.h deleted file mode 100644 index 1de9cc2a89..0000000000 --- a/src/mesa/pipe/softpipe/sp_prim_vbuf.h +++ /dev/null @@ -1,38 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SP_VBUF_H -#define SP_VBUF_H - - -struct softpipe_context; - -extern void -sp_init_vbuf(struct softpipe_context *softpipe); - - -#endif /* SP_VBUF_H */ diff --git a/src/mesa/pipe/softpipe/sp_quad.c b/src/mesa/pipe/softpipe/sp_quad.c deleted file mode 100644 index 6bd468a51c..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad.c +++ /dev/null @@ -1,118 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "sp_context.h" -#include "sp_state.h" -#include "pipe/p_shader_tokens.h" - -static void -sp_push_quad_first( - struct softpipe_context *sp, - struct quad_stage *quad ) -{ - quad->next = sp->quad.first; - sp->quad.first = quad; -} - -static void -sp_build_depth_stencil( - struct softpipe_context *sp ) -{ - if (sp->depth_stencil->stencil[0].enabled || - sp->depth_stencil->stencil[1].enabled) { - sp_push_quad_first( sp, sp->quad.stencil_test ); - } - else if (sp->depth_stencil->depth.enabled && - sp->framebuffer.zsbuf) { - sp_push_quad_first( sp, sp->quad.depth_test ); - } -} - -void -sp_build_quad_pipeline(struct softpipe_context *sp) -{ - boolean early_depth_test = - sp->depth_stencil->depth.enabled && - sp->framebuffer.zsbuf && - !sp->depth_stencil->alpha.enabled && - sp->fs->shader.output_semantic_name[0] != TGSI_SEMANTIC_POSITION; - - /* build up the pipeline in reverse order... */ - - sp->quad.first = sp->quad.output; - - if (sp->blend->colormask != 0xf) { - sp_push_quad_first( sp, sp->quad.colormask ); - } - - if (sp->blend->blend_enable || - sp->blend->logicop_enable) { - sp_push_quad_first( sp, sp->quad.blend ); - } - - if (sp->framebuffer.num_cbufs == 1) { - /* the usual case: write to exactly one colorbuf */ - sp->current_cbuf = 0; - } - else { - /* insert bufloop stage */ - sp_push_quad_first( sp, sp->quad.bufloop ); - } - - if (sp->depth_stencil->depth.occlusion_count) { - sp_push_quad_first( sp, sp->quad.occlusion ); - } - - if (sp->rasterizer->poly_smooth || - sp->rasterizer->line_smooth || - sp->rasterizer->point_smooth) { - sp_push_quad_first( sp, sp->quad.coverage ); - } - - if (!early_depth_test) { - sp_build_depth_stencil( sp ); - } - - if (sp->depth_stencil->alpha.enabled) { - sp_push_quad_first( sp, sp->quad.alpha_test ); - } - - /* XXX always enable shader? */ - if (1) { - sp_push_quad_first( sp, sp->quad.shade ); - } - - if (early_depth_test) { - sp_build_depth_stencil( sp ); - sp_push_quad_first( sp, sp->quad.earlyz ); - } - - if (sp->rasterizer->poly_stipple_enable) { - sp_push_quad_first( sp, sp->quad.polygon_stipple ); - } -} diff --git a/src/mesa/pipe/softpipe/sp_quad.h b/src/mesa/pipe/softpipe/sp_quad.h deleted file mode 100644 index f1e0281764..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad.h +++ /dev/null @@ -1,70 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef SP_QUAD_H -#define SP_QUAD_H - - -struct softpipe_context; -struct quad_header; - - -struct quad_stage { - struct softpipe_context *softpipe; - - struct quad_stage *next; - - void (*begin)(struct quad_stage *qs); - - /** the stage action */ - void (*run)(struct quad_stage *qs, struct quad_header *quad); - - void (*destroy)(struct quad_stage *qs); -}; - - -struct quad_stage *sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_earlyz_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_alpha_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); - -void sp_build_quad_pipeline(struct softpipe_context *sp); - -void sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad); - -#endif /* SP_QUAD_H */ diff --git a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c b/src/mesa/pipe/softpipe/sp_quad_alpha_test.c deleted file mode 100644 index 4ffeac35e1..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_alpha_test.c +++ /dev/null @@ -1,108 +0,0 @@ - -/** - * quad alpha test - */ - -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_quad.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" - - -static void -alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - const float ref = softpipe->depth_stencil->alpha.ref; - unsigned passMask = 0x0, j; - const float *aaaa = quad->outputs.color[3]; - - switch (softpipe->depth_stencil->alpha.func) { - case PIPE_FUNC_NEVER: - quad->mask = 0x0; - break; - case PIPE_FUNC_LESS: - /* - * If mask were an array [4] we could do this SIMD-style: - * passMask = (quad->outputs.color[3] <= vec4(ref)); - */ - for (j = 0; j < QUAD_SIZE; j++) { - if (aaaa[j] < ref) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_EQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (aaaa[j] == ref) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_LEQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (aaaa[j] <= ref) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_GREATER: - for (j = 0; j < QUAD_SIZE; j++) { - if (aaaa[j] > ref) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_NOTEQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (aaaa[j] != ref) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_GEQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (aaaa[j] >= ref) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_ALWAYS: - passMask = MASK_ALL; - break; - default: - abort(); - } - - quad->mask &= passMask; - - if (quad->mask) - qs->next->run(qs->next, quad); -} - - -static void alpha_test_begin(struct quad_stage *qs) -{ - qs->next->begin(qs->next); -} - - -static void alpha_test_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -struct quad_stage * -sp_quad_alpha_test_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = alpha_test_begin; - stage->run = alpha_test_quad; - stage->destroy = alpha_test_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_blend.c b/src/mesa/pipe/softpipe/sp_quad_blend.c deleted file mode 100644 index 17f3ecd0b8..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_blend.c +++ /dev/null @@ -1,749 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * quad blending - * \author Brian Paul - */ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_tile_cache.h" -#include "sp_quad.h" - - -#define VEC4_COPY(DST, SRC) \ -do { \ - DST[0] = SRC[0]; \ - DST[1] = SRC[1]; \ - DST[2] = SRC[2]; \ - DST[3] = SRC[3]; \ -} while(0) - -#define VEC4_SCALAR(DST, SRC) \ -do { \ - DST[0] = SRC; \ - DST[1] = SRC; \ - DST[2] = SRC; \ - DST[3] = SRC; \ -} while(0) - -#define VEC4_ADD(R, A, B) \ -do { \ - R[0] = A[0] + B[0]; \ - R[1] = A[1] + B[1]; \ - R[2] = A[2] + B[2]; \ - R[3] = A[3] + B[3]; \ -} while (0) - -#define VEC4_SUB(R, A, B) \ -do { \ - R[0] = A[0] - B[0]; \ - R[1] = A[1] - B[1]; \ - R[2] = A[2] - B[2]; \ - R[3] = A[3] - B[3]; \ -} while (0) - -#define VEC4_MUL(R, A, B) \ -do { \ - R[0] = A[0] * B[0]; \ - R[1] = A[1] * B[1]; \ - R[2] = A[2] * B[2]; \ - R[3] = A[3] * B[3]; \ -} while (0) - -#define VEC4_MIN(R, A, B) \ -do { \ - R[0] = (A[0] < B[0]) ? A[0] : B[0]; \ - R[1] = (A[1] < B[1]) ? A[1] : B[1]; \ - R[2] = (A[2] < B[2]) ? A[2] : B[2]; \ - R[3] = (A[3] < B[3]) ? A[3] : B[3]; \ -} while (0) - -#define VEC4_MAX(R, A, B) \ -do { \ - R[0] = (A[0] > B[0]) ? A[0] : B[0]; \ - R[1] = (A[1] > B[1]) ? A[1] : B[1]; \ - R[2] = (A[2] > B[2]) ? A[2] : B[2]; \ - R[3] = (A[3] > B[3]) ? A[3] : B[3]; \ -} while (0) - - - -static void -logicop_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - float dest[4][QUAD_SIZE]; - ubyte src[4][4], dst[4][4], res[4][4]; - uint *src4 = (uint *) src; - uint *dst4 = (uint *) dst; - uint *res4 = (uint *) res; - struct softpipe_cached_tile * - tile = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - float (*quadColor)[4] = quad->outputs.color; - uint i, j; - - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); - int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); - for (i = 0; i < 4; i++) { - dest[i][j] = tile->data.color[y][x][i]; - } - } - - /* convert to ubyte */ - for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][0], dest[j][0]); /* P0 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][1], dest[j][1]); /* P1 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][2], dest[j][2]); /* P2 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][3], dest[j][3]); /* P3 */ - - UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quadColor[j][0]); /* P0 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quadColor[j][1]); /* P1 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quadColor[j][2]); /* P2 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quadColor[j][3]); /* P3 */ - } - - switch (softpipe->blend->logicop_func) { - case PIPE_LOGICOP_CLEAR: - for (j = 0; j < 4; j++) - res4[j] = 0; - break; - case PIPE_LOGICOP_NOR: - for (j = 0; j < 4; j++) - res4[j] = ~(src4[j] | dst4[j]); - break; - case PIPE_LOGICOP_AND_INVERTED: - for (j = 0; j < 4; j++) - res4[j] = ~src4[j] & dst4[j]; - break; - case PIPE_LOGICOP_COPY_INVERTED: - for (j = 0; j < 4; j++) - res4[j] = ~src4[j]; - break; - case PIPE_LOGICOP_AND_REVERSE: - for (j = 0; j < 4; j++) - res4[j] = src4[j] & ~dst4[j]; - break; - case PIPE_LOGICOP_INVERT: - for (j = 0; j < 4; j++) - res4[j] = ~dst4[j]; - break; - case PIPE_LOGICOP_XOR: - for (j = 0; j < 4; j++) - res4[j] = dst4[j] ^ src4[j]; - break; - case PIPE_LOGICOP_NAND: - for (j = 0; j < 4; j++) - res4[j] = ~(src4[j] & dst4[j]); - break; - case PIPE_LOGICOP_AND: - for (j = 0; j < 4; j++) - res4[j] = src4[j] & dst4[j]; - break; - case PIPE_LOGICOP_EQUIV: - for (j = 0; j < 4; j++) - res4[j] = ~(src4[j] ^ dst4[j]); - break; - case PIPE_LOGICOP_NOOP: - for (j = 0; j < 4; j++) - res4[j] = dst4[j]; - break; - case PIPE_LOGICOP_OR_INVERTED: - for (j = 0; j < 4; j++) - res4[j] = ~src4[j] | dst4[j]; - break; - case PIPE_LOGICOP_COPY: - for (j = 0; j < 4; j++) - res4[j] = src4[j]; - break; - case PIPE_LOGICOP_OR_REVERSE: - for (j = 0; j < 4; j++) - res4[j] = src4[j] | ~dst4[j]; - break; - case PIPE_LOGICOP_OR: - for (j = 0; j < 4; j++) - res4[j] = src4[j] | dst4[j]; - break; - case PIPE_LOGICOP_SET: - for (j = 0; j < 4; j++) - res4[j] = ~0; - break; - default: - assert(0); - } - - for (j = 0; j < 4; j++) { - quadColor[j][0] = UBYTE_TO_FLOAT(res[j][0]); - quadColor[j][1] = UBYTE_TO_FLOAT(res[j][1]); - quadColor[j][2] = UBYTE_TO_FLOAT(res[j][2]); - quadColor[j][3] = UBYTE_TO_FLOAT(res[j][3]); - } - - /* pass quad to next stage */ - qs->next->run(qs->next, quad); -} - - - - -static void -blend_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - static const float zero[4] = { 0, 0, 0, 0 }; - static const float one[4] = { 1, 1, 1, 1 }; - float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - float (*quadColor)[4] = quad->outputs.color; - uint i, j; - - if (softpipe->blend->logicop_enable) { - logicop_quad(qs, quad); - return; - } - - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); - int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); - for (i = 0; i < 4; i++) { - dest[i][j] = tile->data.color[y][x][i]; - } - } - - /* - * Compute src/first term RGB - */ - switch (softpipe->blend->rgb_src_factor) { - case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[0], quadColor[0]); /* R */ - VEC4_COPY(source[1], quadColor[1]); /* G */ - VEC4_COPY(source[2], quadColor[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - VEC4_MUL(source[0], quadColor[0], quadColor[0]); /* R */ - VEC4_MUL(source[1], quadColor[1], quadColor[1]); /* G */ - VEC4_MUL(source[2], quadColor[2], quadColor[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - { - const float *alpha = quadColor[3]; - VEC4_MUL(source[0], quadColor[0], alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_DST_COLOR: - VEC4_MUL(source[0], quadColor[0], dest[0]); /* R */ - VEC4_MUL(source[1], quadColor[1], dest[1]); /* G */ - VEC4_MUL(source[2], quadColor[2], dest[2]); /* B */ - break; - case PIPE_BLENDFACTOR_DST_ALPHA: - { - const float *alpha = dest[3]; - VEC4_MUL(source[0], quadColor[0], alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - { - const float *alpha = quadColor[3]; - float diff[4]; - VEC4_SUB(diff, one, dest[3]); - VEC4_MIN(source[0], alpha, diff); /* R */ - VEC4_MIN(source[1], alpha, diff); /* G */ - VEC4_MIN(source[2], alpha, diff); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ - VEC4_MUL(source[0], quadColor[0], comp); /* R */ - VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ - VEC4_MUL(source[1], quadColor[1], comp); /* G */ - VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ - VEC4_MUL(source[2], quadColor[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float alpha[4]; - VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quadColor[0], alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_SRC1_COLOR: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_SRC1_ALPHA: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(source[0], zero); /* R */ - VEC4_COPY(source[1], zero); /* G */ - VEC4_COPY(source[2], zero); /* B */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ - VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ - VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ - VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, quadColor[3]); - VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, dest[3]); - VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[0]); /* R */ - VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, dest[1]); /* G */ - VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, dest[2]); /* B */ - VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - { - float inv_comp[4]; - /* R */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); - VEC4_MUL(source[0], quadColor[0], inv_comp); - /* G */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); - VEC4_MUL(source[1], quadColor[1], inv_comp); - /* B */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); - VEC4_MUL(source[2], quadColor[2], inv_comp); - } - break; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_alpha[4]; - VEC4_SCALAR(inv_alpha, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: - assert(0); /* to do */ - break; - default: - abort(); - } - - /* - * Compute src/first term A - */ - switch (softpipe->blend->alpha_src_factor) { - case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[3], quadColor[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_SRC_ALPHA: - { - const float *alpha = quadColor[3]; - VEC4_MUL(source[3], quadColor[3], alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - { - const float *alpha = quadColor[3]; - float diff[4]; - VEC4_SUB(diff, one, dest[3]); - VEC4_MIN(source[3], alpha, diff); /* A */ - } - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(source[3], quadColor[3], comp); /* A */ - } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(source[3], zero); /* A */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, quadColor[3]); - VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, dest[3]); - VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_comp[4]; - /* A */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(source[3], quadColor[3], inv_comp); - } - break; - default: - abort(); - } - - - /* - * Compute dest/second term RGB - */ - switch (softpipe->blend->rgb_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - /* dest = dest * 1 NO-OP, leave dest as-is */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - VEC4_MUL(dest[0], dest[0], quadColor[0]); /* R */ - VEC4_MUL(dest[1], dest[1], quadColor[1]); /* G */ - VEC4_MUL(dest[2], dest[2], quadColor[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - VEC4_MUL(dest[0], dest[0], quadColor[3]); /* R * A */ - VEC4_MUL(dest[1], dest[1], quadColor[3]); /* G * A */ - VEC4_MUL(dest[2], dest[2], quadColor[3]); /* B * A */ - break; - case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */ - VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */ - VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */ - break; - case PIPE_BLENDFACTOR_DST_COLOR: - VEC4_MUL(dest[0], dest[0], dest[0]); /* R */ - VEC4_MUL(dest[1], dest[1], dest[1]); /* G */ - VEC4_MUL(dest[2], dest[2], dest[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - assert(0); /* illegal */ - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ - VEC4_MUL(dest[0], dest[0], comp); /* R */ - VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ - VEC4_MUL(dest[1], dest[1], comp); /* G */ - VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ - VEC4_MUL(dest[2], dest[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(dest[0], dest[0], comp); /* R */ - VEC4_MUL(dest[1], dest[1], comp); /* G */ - VEC4_MUL(dest[2], dest[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[0], zero); /* R */ - VEC4_COPY(dest[1], zero); /* G */ - VEC4_COPY(dest[2], zero); /* B */ - break; - case PIPE_BLENDFACTOR_SRC1_COLOR: - case PIPE_BLENDFACTOR_SRC1_ALPHA: - /* XXX what are these? */ - assert(0); - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ - VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ - VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ - VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ - VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ - VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quadColor[3]); - VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ - VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ - VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[3]); /* A */ - VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ - VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ - VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[0]); /* R */ - VEC4_MUL(dest[0], dest[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, dest[1]); /* G */ - VEC4_MUL(dest[1], dest[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, dest[2]); /* B */ - VEC4_MUL(dest[2], dest[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - { - float inv_comp[4]; - /* R */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); - VEC4_MUL(dest[0], dest[0], inv_comp); - /* G */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); - VEC4_MUL(dest[1], dest[1], inv_comp); - /* B */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); - VEC4_MUL(dest[2], dest[2], inv_comp); - } - break; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_comp[4]; - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(dest[0], dest[0], inv_comp); - VEC4_MUL(dest[1], dest[1], inv_comp); - VEC4_MUL(dest[2], dest[2], inv_comp); - } - break; - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: - /* XXX what are these? */ - assert(0); - break; - default: - assert(0); - } - - /* - * Compute dest/second term A - */ - switch (softpipe->blend->alpha_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - /* dest = dest * 1 NO-OP, leave dest as-is */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_SRC_ALPHA: - VEC4_MUL(dest[3], dest[3], quadColor[3]); /* A * A */ - break; - case PIPE_BLENDFACTOR_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(dest[3], dest[3], dest[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - assert(0); /* illegal */ - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(dest[3], dest[3], comp); /* A */ - } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[3], zero); /* A */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quadColor[3]); - VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[3]); /* A */ - VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_comp[4]; - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(dest[3], dest[3], inv_comp); - } - break; - default: - assert(0); - } - - /* - * Combine RGB terms - */ - switch (softpipe->blend->rgb_func) { - case PIPE_BLEND_ADD: - VEC4_ADD(quadColor[0], source[0], dest[0]); /* R */ - VEC4_ADD(quadColor[1], source[1], dest[1]); /* G */ - VEC4_ADD(quadColor[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quadColor[0], source[0], dest[0]); /* R */ - VEC4_SUB(quadColor[1], source[1], dest[1]); /* G */ - VEC4_SUB(quadColor[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quadColor[0], dest[0], source[0]); /* R */ - VEC4_SUB(quadColor[1], dest[1], source[1]); /* G */ - VEC4_SUB(quadColor[2], dest[2], source[2]); /* B */ - break; - case PIPE_BLEND_MIN: - VEC4_MIN(quadColor[0], source[0], dest[0]); /* R */ - VEC4_MIN(quadColor[1], source[1], dest[1]); /* G */ - VEC4_MIN(quadColor[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_MAX: - VEC4_MAX(quadColor[0], source[0], dest[0]); /* R */ - VEC4_MAX(quadColor[1], source[1], dest[1]); /* G */ - VEC4_MAX(quadColor[2], source[2], dest[2]); /* B */ - break; - default: - assert(0); - } - - /* - * Combine A terms - */ - switch (softpipe->blend->alpha_func) { - case PIPE_BLEND_ADD: - VEC4_ADD(quadColor[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quadColor[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quadColor[3], dest[3], source[3]); /* A */ - break; - case PIPE_BLEND_MIN: - VEC4_MIN(quadColor[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_MAX: - VEC4_MAX(quadColor[3], source[3], dest[3]); /* A */ - break; - default: - abort(); - } - - /* pass blended quad to next stage */ - qs->next->run(qs->next, quad); -} - - -static void blend_begin(struct quad_stage *qs) -{ - qs->next->begin(qs->next); -} - - -static void blend_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = blend_begin; - stage->run = blend_quad; - stage->destroy = blend_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_bufloop.c b/src/mesa/pipe/softpipe/sp_quad_bufloop.c deleted file mode 100644 index 2ae4e22a7d..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_bufloop.c +++ /dev/null @@ -1,72 +0,0 @@ - -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_quad.h" - - -/** - * Loop over colorbuffers, passing quad to next stage each time. - */ -static void -cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - float tmp[4][QUAD_SIZE]; - unsigned i; - - assert(sizeof(quad->outputs.color) == sizeof(tmp)); - assert(softpipe->framebuffer.num_cbufs <= PIPE_MAX_COLOR_BUFS); - - /* make copy of original colors since they can get modified - * by blending and masking. - * XXX we won't have to do this if the fragment program actually emits - * N separate colors and we're drawing to N color buffers (MRT). - * But if we emitted one color and glDrawBuffer(GL_FRONT_AND_BACK) is - * in effect, we need to save/restore colors like this. - */ - memcpy(tmp, quad->outputs.color, sizeof(tmp)); - - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { - /* set current cbuffer */ - softpipe->current_cbuf = i; - - /* pass blended quad to next stage */ - qs->next->run(qs->next, quad); - - /* restore quad's colors for next buffer */ - memcpy(quad->outputs.color, tmp, sizeof(tmp)); - } -} - - -static void cbuf_loop_begin(struct quad_stage *qs) -{ - qs->next->begin(qs->next); -} - - -static void cbuf_loop_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -/** - * Create the colorbuffer loop stage. - * This is used to implement multiple render targets and GL_FRONT_AND_BACK - * rendering. - */ -struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = cbuf_loop_begin; - stage->run = cbuf_loop_quad; - stage->destroy = cbuf_loop_destroy; - - return stage; -} - diff --git a/src/mesa/pipe/softpipe/sp_quad_colormask.c b/src/mesa/pipe/softpipe/sp_quad_colormask.c deleted file mode 100644 index 1f09d900ca..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_colormask.c +++ /dev/null @@ -1,110 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief quad colormask stage - * \author Brian Paul - */ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_quad.h" -#include "sp_tile_cache.h" - - - -/** - * XXX colormask could be rolled into blending... - */ -static void -colormask_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - float dest[4][QUAD_SIZE]; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - float (*quadColor)[4] = quad->outputs.color; - uint i, j; - - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); - int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); - for (i = 0; i < 4; i++) { - dest[i][j] = tile->data.color[y][x][i]; - } - } - - /* R */ - if (!(softpipe->blend->colormask & PIPE_MASK_R)) - COPY_4V(quadColor[0], dest[0]); - - /* G */ - if (!(softpipe->blend->colormask & PIPE_MASK_G)) - COPY_4V(quadColor[1], dest[1]); - - /* B */ - if (!(softpipe->blend->colormask & PIPE_MASK_B)) - COPY_4V(quadColor[2], dest[2]); - - /* A */ - if (!(softpipe->blend->colormask & PIPE_MASK_A)) - COPY_4V(quadColor[3], dest[3]); - - /* pass quad to next stage */ - qs->next->run(qs->next, quad); -} - - -static void colormask_begin(struct quad_stage *qs) -{ - qs->next->begin(qs->next); -} - - -static void colormask_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = colormask_begin; - stage->run = colormask_quad; - stage->destroy = colormask_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_coverage.c b/src/mesa/pipe/softpipe/sp_quad_coverage.c deleted file mode 100644 index b3d3fae22f..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_coverage.c +++ /dev/null @@ -1,88 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * \brief Apply AA coverage to quad alpha valus - * \author Brian Paul - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_quad.h" - - -/** - * Multiply quad's alpha values by the fragment coverage. - */ -static void -coverage_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - - if ((softpipe->rasterizer->poly_smooth && quad->prim == PRIM_TRI) || - (softpipe->rasterizer->line_smooth && quad->prim == PRIM_LINE) || - (softpipe->rasterizer->point_smooth && quad->prim == PRIM_POINT)) { - float (*quadColor)[4] = quad->outputs.color; - unsigned j; - for (j = 0; j < QUAD_SIZE; j++) { - assert(quad->coverage[j] >= 0.0); - assert(quad->coverage[j] <= 1.0); - quadColor[3][j] *= quad->coverage[j]; - } - } - - qs->next->run(qs->next, quad); -} - - -static void coverage_begin(struct quad_stage *qs) -{ - qs->next->begin(qs->next); -} - - -static void coverage_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = coverage_begin; - stage->run = coverage_quad; - stage->destroy = coverage_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_depth_test.c b/src/mesa/pipe/softpipe/sp_quad_depth_test.c deleted file mode 100644 index a9a0754f27..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_depth_test.c +++ /dev/null @@ -1,276 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Quad depth testing - */ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_quad.h" -#include "sp_tile_cache.h" - - -/** - * Do depth testing for a quad. - * Not static since it's used by the stencil code. - */ - -/* - * To increase efficiency, we should probably have multiple versions - * of this function that are specifically for Z16, Z32 and FP Z buffers. - * Try to effectively do that with codegen... - */ - -void -sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - struct pipe_surface *ps = softpipe->framebuffer.zsbuf; - const enum pipe_format format = ps->format; - unsigned bzzzz[QUAD_SIZE]; /**< Z values fetched from depth buffer */ - unsigned qzzzz[QUAD_SIZE]; /**< Z values from the quad */ - unsigned zmask = 0; - unsigned j; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->x0, quad->y0); - - assert(ps); /* shouldn't get here if there's no zbuffer */ - - /* - * Convert quad's float depth values to int depth values (qzzzz). - * If the Z buffer stores integer values, we _have_ to do the depth - * compares with integers (not floats). Otherwise, the float->int->float - * conversion of Z values (which isn't an identity function) will cause - * Z-fighting errors. - * - * Also, get the zbuffer values (bzzzz) from the cached tile. - */ - switch (format) { - case PIPE_FORMAT_Z16_UNORM: - { - float scale = 65535.0; - - for (j = 0; j < QUAD_SIZE; j++) { - qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); - } - - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - bzzzz[j] = tile->data.depth16[y][x]; - } - } - break; - case PIPE_FORMAT_Z32_UNORM: - { - double scale = (double) (uint) ~0UL; - - for (j = 0; j < QUAD_SIZE; j++) { - qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); - } - - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - bzzzz[j] = tile->data.depth32[y][x]; - } - } - break; - case PIPE_FORMAT_S8Z24_UNORM: - { - float scale = (float) ((1 << 24) - 1); - - for (j = 0; j < QUAD_SIZE; j++) { - qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); - } - - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - bzzzz[j] = tile->data.depth32[y][x] & 0xffffff; - } - } - break; - case PIPE_FORMAT_Z24S8_UNORM: - { - float scale = (float) ((1 << 24) - 1); - - for (j = 0; j < QUAD_SIZE; j++) { - qzzzz[j] = (unsigned) (quad->outputs.depth[j] * scale); - } - - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - bzzzz[j] = tile->data.depth32[y][x] >> 8; - } - } - break; - default: - assert(0); - } - - switch (softpipe->depth_stencil->depth.func) { - case PIPE_FUNC_NEVER: - /* zmask = 0 */ - break; - case PIPE_FUNC_LESS: - /* Note this is pretty much a single sse or cell instruction. - * Like this: quad->mask &= (quad->outputs.depth < zzzz); - */ - for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] < bzzzz[j]) - zmask |= 1 << j; - } - break; - case PIPE_FUNC_EQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] == bzzzz[j]) - zmask |= 1 << j; - } - break; - case PIPE_FUNC_LEQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] <= bzzzz[j]) - zmask |= (1 << j); - } - break; - case PIPE_FUNC_GREATER: - for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] > bzzzz[j]) - zmask |= (1 << j); - } - break; - case PIPE_FUNC_NOTEQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] != bzzzz[j]) - zmask |= (1 << j); - } - break; - case PIPE_FUNC_GEQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (qzzzz[j] >= bzzzz[j]) - zmask |= (1 << j); - } - break; - case PIPE_FUNC_ALWAYS: - zmask = MASK_ALL; - break; - default: - abort(); - } - - quad->mask &= zmask; - - if (softpipe->depth_stencil->depth.writemask) { - - /* This is also efficient with sse / spe instructions: - */ - for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - bzzzz[j] = qzzzz[j]; - } - } - - /* put updated Z values back into cached tile */ - switch (format) { - case PIPE_FORMAT_Z16_UNORM: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - tile->data.depth16[y][x] = (ushort) bzzzz[j]; - } - break; - case PIPE_FORMAT_Z32_UNORM: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - tile->data.depth32[y][x] = bzzzz[j]; - } - break; - case PIPE_FORMAT_S8Z24_UNORM: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - uint s8z24 = tile->data.depth32[y][x]; - s8z24 = (s8z24 & 0xff000000) | bzzzz[j]; - tile->data.depth32[y][x] = s8z24; - } - break; - case PIPE_FORMAT_Z24S8_UNORM: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - uint z24s8 = tile->data.depth32[y][x]; - z24s8 = (z24s8 & 0xff) | (bzzzz[j] << 24); - tile->data.depth32[y][x] = z24s8; - } - break; - default: - assert(0); - } - } -} - - -static void -depth_test_quad(struct quad_stage *qs, struct quad_header *quad) -{ - sp_depth_test_quad(qs, quad); - - if (quad->mask) - qs->next->run(qs->next, quad); -} - - -static void depth_test_begin(struct quad_stage *qs) -{ - qs->next->begin(qs->next); -} - - -static void depth_test_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = depth_test_begin; - stage->run = depth_test_quad; - stage->destroy = depth_test_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_earlyz.c b/src/mesa/pipe/softpipe/sp_quad_earlyz.c deleted file mode 100644 index 22ea99049f..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_earlyz.c +++ /dev/null @@ -1,88 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Quad early-z testing - */ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "sp_headers.h" -#include "sp_quad.h" - - -/** - * All this stage does is compute the quad's Z values (which is normally - * done by the shading stage). - * The next stage will do the actual depth test. - */ -static void -earlyz_quad( - struct quad_stage *qs, - struct quad_header *quad ) -{ - const float fx = (float) quad->x0; - const float fy = (float) quad->y0; - const float dzdx = quad->posCoef->dadx[2]; - const float dzdy = quad->posCoef->dady[2]; - const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy; - - quad->outputs.depth[0] = z0; - quad->outputs.depth[1] = z0 + dzdx; - quad->outputs.depth[2] = z0 + dzdy; - quad->outputs.depth[3] = z0 + dzdx + dzdy; - - qs->next->run( qs->next, quad ); -} - -static void -earlyz_begin( - struct quad_stage *qs ) -{ - qs->next->begin( qs->next ); -} - -static void -earlyz_destroy( - struct quad_stage *qs ) -{ - FREE( qs ); -} - -struct quad_stage * -sp_quad_earlyz_stage( - struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT( quad_stage ); - - stage->softpipe = softpipe; - stage->begin = earlyz_begin; - stage->run = earlyz_quad; - stage->destroy = earlyz_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c deleted file mode 100644 index 3316858413..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ /dev/null @@ -1,390 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" - -#include "x86/rtasm/x86sse.h" - -#ifdef MESA_LLVM -#include "pipe/llvm/gallivm.h" -#endif - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_headers.h" -#include "sp_quad.h" -#include "sp_texture.h" -#include "sp_tex_sample.h" - - -struct quad_shade_stage -{ - struct quad_stage stage; - struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; - struct tgsi_exec_machine machine; - struct tgsi_exec_vector *inputs, *outputs; - int colorOutSlot, depthOutSlot; -#ifdef MESA_LLVM - struct gallivm_prog *llvm_prog; -#endif -}; - - -/** cast wrapper */ -static INLINE struct quad_shade_stage * -quad_shade_stage(struct quad_stage *qs) -{ - return (struct quad_shade_stage *) qs; -} - - -/** - * Compute quad X,Y,Z,W for the four fragments in a quad. - * Note that we only need to "compute" X and Y for the upper-left fragment. - * We could do less work if we're not depth testing, or there's no - * perspective-corrected attributes, but that's seldom. - */ -static void -setup_pos_vector(const struct tgsi_interp_coef *coef, - float x, float y, - struct tgsi_exec_vector *quadpos) -{ - uint chan; - /* do X */ - quadpos->xyzw[0].f[0] = x; - /* do Y */ - quadpos->xyzw[1].f[0] = y; - /* do Z and W for all fragments in the quad */ - for (chan = 2; chan < 4; chan++) { - const float dadx = coef->dadx[chan]; - const float dady = coef->dady[chan]; - const float a0 = coef->a0[chan] + dadx * x + dady * y; - quadpos->xyzw[chan].f[0] = a0; - quadpos->xyzw[chan].f[1] = a0 + dadx; - quadpos->xyzw[chan].f[2] = a0 + dady; - quadpos->xyzw[chan].f[3] = a0 + dadx + dady; - } -} - - -typedef void (XSTDCALL *codegen_function)( - const struct tgsi_exec_vector *input, - struct tgsi_exec_vector *output, - float (*constant)[4], - struct tgsi_exec_vector *temporary, - const struct tgsi_interp_coef *coef -#if 0 - ,const struct tgsi_exec_vector *quadPos -#endif - ); - - -/** - * Execute fragment shader for the four fragments in the quad. - */ -static void -shade_quad( - struct quad_stage *qs, - struct quad_header *quad ) -{ - struct quad_shade_stage *qss = quad_shade_stage( qs ); - struct softpipe_context *softpipe = qs->softpipe; - struct tgsi_exec_machine *machine = &qss->machine; - - /* Consts do not require 16 byte alignment. */ - machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; - - machine->InterpCoefs = quad->coef; - - /* Compute X, Y, Z, W vals for this quad */ - setup_pos_vector(quad->posCoef, (float) quad->x0, (float) quad->y0, &machine->QuadPos); - - /* run shader */ -#if defined(__i386__) || defined(__386__) - if( softpipe->use_sse ) { - codegen_function func = (codegen_function) x86_get_func( &softpipe->fs->sse2_program ); - func( - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps, - machine->InterpCoefs -#if 0 - ,machine->QuadPos -#endif - ); - quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]); - } - else -#endif - { - quad->mask &= tgsi_exec_machine_run( machine ); - } - - /* store result color */ - if (qss->colorOutSlot >= 0) { - /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] - == TGSI_SEMANTIC_COLOR); - memcpy( - quad->outputs.color, - &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); - } - - /* - * XXX the following code for updating quad->outputs.depth - * isn't really needed if we did early z testing. - */ - - /* store result Z */ - if (qss->depthOutSlot >= 0) { - /* output[slot] is new Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; - } - } - else { - /* copy input Z (which was interpolated by the executor) to output Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; - /* XXX not sure the above line is always correct. The following - * might be better: - quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i]; - */ - } - } - - /* shader may cull fragments */ - if( quad->mask ) { - qs->next->run( qs->next, quad ); - } -} - -#if 0 -#ifdef MESA_LLVM -#define DLLVM 0 -static void -shade_quad_llvm(struct quad_stage *qs, - struct quad_header *quad) -{ - struct quad_shade_stage *qss = quad_shade_stage(qs); - struct softpipe_context *softpipe = qs->softpipe; - float dests[4][16][4] ALIGN16_ATTRIB; - float inputs[4][16][4] ALIGN16_ATTRIB; - const float fx = (float) quad->x0; - const float fy = (float) quad->y0; - struct gallivm_prog *llvm = qss->llvm_prog; - - inputs[0][0][0] = fx; - inputs[1][0][0] = fx + 1.0f; - inputs[2][0][0] = fx; - inputs[3][0][0] = fx + 1.0f; - - inputs[0][0][1] = fy; - inputs[1][0][1] = fy; - inputs[2][0][1] = fy + 1.0f; - inputs[3][0][1] = fy + 1.0f; -#if DLLVM - debug_printf("MASK = %d\n", quad->mask); -#endif - gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef); -#if DLLVM - for (int i = 0; i < 4; ++i) { - for (int j = 0; j < 2; ++j) { - debug_printf("IN(%d,%d) [%f %f %f %f]\n", i, j, - inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]); - } - } -#endif - - quad->mask &= - gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs, - softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], - qss->samplers); -#if DLLVM - debug_printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n", - dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], - dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]); -#endif - - /* store result color */ - if (qss->colorOutSlot >= 0) { - unsigned i; - /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] - == TGSI_SEMANTIC_COLOR); - for (i = 0; i < QUAD_SIZE; ++i) { - quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0]; - quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1]; - quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2]; - quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3]; - } - } -#if DLLVM - for (int i = 0; i < QUAD_SIZE; ++i) { - debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, - quad->outputs.color[0][i], - quad->outputs.color[1][i], - quad->outputs.color[2][i], - quad->outputs.color[3][i]); - } -#endif - - /* store result Z */ - if (qss->depthOutSlot >= 0) { - /* output[slot] is new Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = dests[i][0][2]; - } - } - else { - /* copy input Z (which was interpolated by the executor) to output Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = inputs[i][0][2]; - } - } -#if DLLVM - debug_printf("D [%f, %f, %f, %f] mask = %d\n", - quad->outputs.depth[0], - quad->outputs.depth[1], - quad->outputs.depth[2], - quad->outputs.depth[3], quad->mask); -#endif - - /* shader may cull fragments */ - if( quad->mask ) { - qs->next->run( qs->next, quad ); - } -} -#endif /*MESA_LLVM*/ -#endif - -/** - * Per-primitive (or per-begin?) setup - */ -static void shade_begin(struct quad_stage *qs) -{ - struct quad_shade_stage *qss = quad_shade_stage(qs); - struct softpipe_context *softpipe = qs->softpipe; - unsigned i; - - /* set TGSI sampler state that varies */ - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - qss->samplers[i].state = softpipe->sampler[i]; - qss->samplers[i].texture = &softpipe->texture[i]->base; - } - -#ifdef MESA_LLVM - qss->llvm_prog = softpipe->fs->llvm_prog; -#endif - /* XXX only do this if the fragment shader changes... */ - tgsi_exec_machine_init(&qss->machine, - softpipe->fs->shader.tokens, - PIPE_MAX_SAMPLERS, - qss->samplers ); - - /* find output slots for depth, color */ - qss->colorOutSlot = -1; - qss->depthOutSlot = -1; - for (i = 0; i < qss->stage.softpipe->fs->shader.num_outputs; i++) { - switch (qss->stage.softpipe->fs->shader.output_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - qss->depthOutSlot = i; - break; - case TGSI_SEMANTIC_COLOR: - qss->colorOutSlot = i; - break; - } - } - - qs->next->begin(qs->next); -} - - -static void shade_destroy(struct quad_stage *qs) -{ - struct quad_shade_stage *qss = (struct quad_shade_stage *) qs; - - tgsi_exec_machine_free_data(&qss->machine); - FREE( qss->inputs ); - FREE( qss->outputs ); - FREE( qs ); -} - - -struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) -{ - struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); - uint i; - - /* allocate storage for program inputs/outputs, aligned to 16 bytes */ - qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); - qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); - qss->machine.Inputs = align16(qss->inputs); - qss->machine.Outputs = align16(qss->outputs); - - qss->stage.softpipe = softpipe; - qss->stage.begin = shade_begin; -#ifdef MESA_LLVM - /* disable until ported to accept - * x/y and soa layout - qss->stage.run = shade_quad_llvm; - */ - softpipe->use_sse = FALSE; - qss->stage.run = shade_quad; -#else - qss->stage.run = shade_quad; -#endif - qss->stage.destroy = shade_destroy; - - /* set TGSI sampler state that's constant */ - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - assert(softpipe->tex_cache[i]); - qss->samplers[i].get_samples = sp_get_samples; - qss->samplers[i].pipe = &softpipe->pipe; - qss->samplers[i].cache = softpipe->tex_cache[i]; - } - - return &qss->stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_occlusion.c b/src/mesa/pipe/softpipe/sp_quad_occlusion.c deleted file mode 100644 index 54254df1f1..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_occlusion.c +++ /dev/null @@ -1,85 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * \brief Quad occlusion counter stage - * \author Brian Paul - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_quad.h" - -static unsigned count_bits( unsigned val ) -{ - unsigned i; - - for (i = 0; val ; val >>= 1) - i += (val & 1); - - return i; -} - -static void -occlusion_count_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - - softpipe->occlusion_count += count_bits(quad->mask); - - qs->next->run(qs->next, quad); -} - - -static void occlusion_begin(struct quad_stage *qs) -{ - qs->next->begin(qs->next); -} - - -static void occlusion_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = occlusion_begin; - stage->run = occlusion_count_quad; - stage->destroy = occlusion_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_output.c b/src/mesa/pipe/softpipe/sp_quad_output.c deleted file mode 100644 index cfe8f11808..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_output.c +++ /dev/null @@ -1,90 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_quad.h" -#include "sp_tile_cache.h" - - -/** - * Write quad to framebuffer, taking mask into account. - * - * Note that surfaces support only full quad reads and writes. - */ -static void -output_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - /* in-tile pos: */ - const int itx = quad->x0 % TILE_SIZE; - const int ity = quad->y0 % TILE_SIZE; - float (*quadColor)[4] = quad->outputs.color; - int i, j; - - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - int x = itx + (j & 1); - int y = ity + (j >> 1); - for (i = 0; i < 4; i++) { /* loop over color chans */ - tile->data.color[y][x][i] = quadColor[i][j]; - } - } - } -} - - -static void output_begin(struct quad_stage *qs) -{ - assert(qs->next == NULL); -} - - -static void output_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = output_begin; - stage->run = output_quad; - stage->destroy = output_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_stencil.c b/src/mesa/pipe/softpipe/sp_quad_stencil.c deleted file mode 100644 index 92a0da0083..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_stencil.c +++ /dev/null @@ -1,352 +0,0 @@ - -/** - * \brief Quad stencil testing - */ - - -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_tile_cache.h" -#include "sp_quad.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" - - -/** Only 8-bit stencil supported */ -#define STENCIL_MAX 0xff - - -/** - * Do the basic stencil test (compare stencil buffer values against the - * reference value. - * - * \param stencilVals the stencil values from the stencil buffer - * \param func the stencil func (PIPE_FUNC_x) - * \param ref the stencil reference value - * \param valMask the stencil value mask indicating which bits of the stencil - * values and ref value are to be used. - * \return mask indicating which pixels passed the stencil test - */ -static unsigned -do_stencil_test(const ubyte stencilVals[QUAD_SIZE], unsigned func, - unsigned ref, unsigned valMask) -{ - unsigned passMask = 0x0; - unsigned j; - - ref &= valMask; - - switch (func) { - case PIPE_FUNC_NEVER: - /* passMask = 0x0 */ - break; - case PIPE_FUNC_LESS: - for (j = 0; j < QUAD_SIZE; j++) { - if (ref < (stencilVals[j] & valMask)) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_EQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (ref == (stencilVals[j] & valMask)) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_LEQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (ref <= (stencilVals[j] & valMask)) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_GREATER: - for (j = 0; j < QUAD_SIZE; j++) { - if (ref > (stencilVals[j] & valMask)) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_NOTEQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (ref != (stencilVals[j] & valMask)) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_GEQUAL: - for (j = 0; j < QUAD_SIZE; j++) { - if (ref >= (stencilVals[j] & valMask)) { - passMask |= (1 << j); - } - } - break; - case PIPE_FUNC_ALWAYS: - passMask = MASK_ALL; - break; - default: - assert(0); - } - - return passMask; -} - - -/** - * Apply the stencil operator to stencil values. - * - * \param stencilVals the stencil buffer values (read and written) - * \param mask indicates which pixels to update - * \param op the stencil operator (PIPE_STENCIL_OP_x) - * \param ref the stencil reference value - * \param wrtMask writemask controlling which bits are changed in the - * stencil values - */ -static void -apply_stencil_op(ubyte stencilVals[QUAD_SIZE], - unsigned mask, unsigned op, ubyte ref, ubyte wrtMask) -{ - unsigned j; - ubyte newstencil[QUAD_SIZE]; - - for (j = 0; j < QUAD_SIZE; j++) { - newstencil[j] = stencilVals[j]; - } - - switch (op) { - case PIPE_STENCIL_OP_KEEP: - /* no-op */ - break; - case PIPE_STENCIL_OP_ZERO: - for (j = 0; j < QUAD_SIZE; j++) { - if (mask & (1 << j)) { - newstencil[j] = 0; - } - } - break; - case PIPE_STENCIL_OP_REPLACE: - for (j = 0; j < QUAD_SIZE; j++) { - if (mask & (1 << j)) { - newstencil[j] = ref; - } - } - break; - case PIPE_STENCIL_OP_INCR: - for (j = 0; j < QUAD_SIZE; j++) { - if (mask & (1 << j)) { - if (stencilVals[j] < STENCIL_MAX) { - newstencil[j] = stencilVals[j] + 1; - } - } - } - break; - case PIPE_STENCIL_OP_DECR: - for (j = 0; j < QUAD_SIZE; j++) { - if (mask & (1 << j)) { - if (stencilVals[j] > 0) { - newstencil[j] = stencilVals[j] - 1; - } - } - } - break; - case PIPE_STENCIL_OP_INCR_WRAP: - for (j = 0; j < QUAD_SIZE; j++) { - if (mask & (1 << j)) { - newstencil[j] = stencilVals[j] + 1; - } - } - break; - case PIPE_STENCIL_OP_DECR_WRAP: - for (j = 0; j < QUAD_SIZE; j++) { - if (mask & (1 << j)) { - newstencil[j] = stencilVals[j] - 1; - } - } - break; - case PIPE_STENCIL_OP_INVERT: - for (j = 0; j < QUAD_SIZE; j++) { - if (mask & (1 << j)) { - newstencil[j] = ~stencilVals[j]; - } - } - break; - default: - assert(0); - } - - /* - * update the stencil values - */ - if (wrtMask != STENCIL_MAX) { - /* apply bit-wise stencil buffer writemask */ - for (j = 0; j < QUAD_SIZE; j++) { - stencilVals[j] = (wrtMask & newstencil[j]) | (~wrtMask & stencilVals[j]); - } - } - else { - for (j = 0; j < QUAD_SIZE; j++) { - stencilVals[j] = newstencil[j]; - } - } -} - - -/** - * Do stencil (and depth) testing. Stenciling depends on the outcome of - * depth testing. - */ -static void -stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) -{ - struct softpipe_context *softpipe = qs->softpipe; - struct pipe_surface *ps = softpipe->framebuffer.zsbuf; - unsigned func, zFailOp, zPassOp, failOp; - ubyte ref, wrtMask, valMask; - ubyte stencilVals[QUAD_SIZE]; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, softpipe->zsbuf_cache, quad->x0, quad->y0); - uint j; - uint face = quad->facing; - - if (!softpipe->depth_stencil->stencil[1].enabled) { - /* single-sided stencil test, use front (face=0) state */ - face = 0; - } - - /* choose front or back face function, operator, etc */ - /* XXX we could do these initializations once per primitive */ - func = softpipe->depth_stencil->stencil[face].func; - failOp = softpipe->depth_stencil->stencil[face].fail_op; - zFailOp = softpipe->depth_stencil->stencil[face].zfail_op; - zPassOp = softpipe->depth_stencil->stencil[face].zpass_op; - ref = softpipe->depth_stencil->stencil[face].ref_value; - wrtMask = softpipe->depth_stencil->stencil[face].write_mask; - valMask = softpipe->depth_stencil->stencil[face].value_mask; - - assert(ps); /* shouldn't get here if there's no stencil buffer */ - - /* get stencil values from cached tile */ - switch (ps->format) { - case PIPE_FORMAT_S8Z24_UNORM: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - stencilVals[j] = tile->data.depth32[y][x] >> 24; - } - break; - case PIPE_FORMAT_Z24S8_UNORM: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - stencilVals[j] = tile->data.depth32[y][x] & 0xff; - } - break; - case PIPE_FORMAT_U_S8: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - stencilVals[j] = tile->data.stencil8[y][x]; - } - break; - default: - assert(0); - } - - /* do the stencil test first */ - { - unsigned passMask, failMask; - passMask = do_stencil_test(stencilVals, func, ref, valMask); - failMask = quad->mask & ~passMask; - quad->mask &= passMask; - - if (failOp != PIPE_STENCIL_OP_KEEP) { - apply_stencil_op(stencilVals, failMask, failOp, ref, wrtMask); - } - } - - if (quad->mask) { - /* now the pixels that passed the stencil test are depth tested */ - if (softpipe->depth_stencil->depth.enabled) { - const unsigned origMask = quad->mask; - - sp_depth_test_quad(qs, quad); /* quad->mask is updated */ - - /* update stencil buffer values according to z pass/fail result */ - if (zFailOp != PIPE_STENCIL_OP_KEEP) { - const unsigned failMask = origMask & ~quad->mask; - apply_stencil_op(stencilVals, failMask, zFailOp, ref, wrtMask); - } - - if (zPassOp != PIPE_STENCIL_OP_KEEP) { - const unsigned passMask = origMask & quad->mask; - apply_stencil_op(stencilVals, passMask, zPassOp, ref, wrtMask); - } - } - else { - /* no depth test, apply Zpass operator to stencil buffer values */ - apply_stencil_op(stencilVals, quad->mask, zPassOp, ref, wrtMask); - } - - } - - /* put new stencil values into cached tile */ - switch (ps->format) { - case PIPE_FORMAT_S8Z24_UNORM: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - uint s8z24 = tile->data.depth32[y][x]; - s8z24 = (stencilVals[j] << 24) | (s8z24 & 0xffffff); - tile->data.depth32[y][x] = s8z24; - } - break; - case PIPE_FORMAT_Z24S8_UNORM: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - uint z24s8 = tile->data.depth32[y][x]; - z24s8 = (z24s8 & 0xffffff00) | stencilVals[j]; - tile->data.depth32[y][x] = z24s8; - } - break; - case PIPE_FORMAT_U_S8: - for (j = 0; j < QUAD_SIZE; j++) { - int x = quad->x0 % TILE_SIZE + (j & 1); - int y = quad->y0 % TILE_SIZE + (j >> 1); - tile->data.stencil8[y][x] = stencilVals[j]; - } - break; - default: - assert(0); - } - - if (quad->mask) - qs->next->run(qs->next, quad); -} - - -static void stencil_begin(struct quad_stage *qs) -{ - qs->next->begin(qs->next); -} - - -static void stencil_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = stencil_begin; - stage->run = stencil_test_quad; - stage->destroy = stencil_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_quad_stipple.c b/src/mesa/pipe/softpipe/sp_quad_stipple.c deleted file mode 100644 index 8660432259..0000000000 --- a/src/mesa/pipe/softpipe/sp_quad_stipple.c +++ /dev/null @@ -1,94 +0,0 @@ - -/** - * quad polygon stipple stage - */ - -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_quad.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" - - -/** - * Apply polygon stipple to quads produced by triangle rasterization - */ -static void -stipple_quad(struct quad_stage *qs, struct quad_header *quad) -{ - static const uint bit31 = 1 << 31; - static const uint bit30 = 1 << 30; - - if (quad->prim == PRIM_TRI) { - struct softpipe_context *softpipe = qs->softpipe; - /* need to invert Y to index into OpenGL's stipple pattern */ - int y0, y1; - uint stipple0, stipple1; - if (softpipe->rasterizer->origin_lower_left) { - y0 = softpipe->framebuffer.cbufs[0]->height - 1 - quad->y0; - y1 = y0 - 1; - } - else { - y0 = quad->y0; - y1 = y0 + 1; - } - stipple0 = softpipe->poly_stipple.stipple[y0 % 32]; - stipple1 = softpipe->poly_stipple.stipple[y1 % 32]; - -#if 1 - { - const int col0 = quad->x0 % 32; - if ((stipple0 & (bit31 >> col0)) == 0) - quad->mask &= ~MASK_TOP_LEFT; - - if ((stipple0 & (bit30 >> col0)) == 0) - quad->mask &= ~MASK_TOP_RIGHT; - - if ((stipple1 & (bit31 >> col0)) == 0) - quad->mask &= ~MASK_BOTTOM_LEFT; - - if ((stipple1 & (bit30 >> col0)) == 0) - quad->mask &= ~MASK_BOTTOM_RIGHT; - } -#else - /* We'd like to use this code, but we'd need to redefine - * MASK_TOP_LEFT to be (1 << 1) and MASK_TOP_RIGHT to be (1 << 0), - * and similarly for the BOTTOM bits. But that may have undesirable - * side effects elsewhere. - */ - const int col0 = 30 - (quad->x0 % 32); - quad->mask &= (((stipple0 >> col0) & 0x3) | - (((stipple1 >> col0) & 0x3) << 2)); -#endif - if (!quad->mask) - return; - } - - qs->next->run(qs->next, quad); -} - - -static void stipple_begin(struct quad_stage *qs) -{ - qs->next->begin(qs->next); -} - - -static void stipple_destroy(struct quad_stage *qs) -{ - FREE( qs ); -} - - -struct quad_stage * -sp_quad_polygon_stipple_stage( struct softpipe_context *softpipe ) -{ - struct quad_stage *stage = CALLOC_STRUCT(quad_stage); - - stage->softpipe = softpipe; - stage->begin = stipple_begin; - stage->run = stipple_quad; - stage->destroy = stipple_destroy; - - return stage; -} diff --git a/src/mesa/pipe/softpipe/sp_query.c b/src/mesa/pipe/softpipe/sp_query.c deleted file mode 100644 index 6a8a43aeda..0000000000 --- a/src/mesa/pipe/softpipe/sp_query.c +++ /dev/null @@ -1,107 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - -#include "pipe/draw/draw_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_query.h" - -struct softpipe_query { - uint64 start; - uint64 end; -}; - - -static struct softpipe_query *softpipe_query( struct pipe_query *p ) -{ - return (struct softpipe_query *)p; -} - -static struct pipe_query * -softpipe_create_query(struct pipe_context *pipe, - unsigned type) -{ - assert(type == PIPE_QUERY_OCCLUSION_COUNTER); - return (struct pipe_query *)CALLOC_STRUCT( softpipe_query ); -} - - -static void -softpipe_destroy_query(struct pipe_context *pipe, struct pipe_query *q) -{ - FREE(q); -} - - -static void -softpipe_begin_query(struct pipe_context *pipe, struct pipe_query *q) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - struct softpipe_query *sq = softpipe_query(q); - - sq->start = softpipe->occlusion_count; -} - - -static void -softpipe_end_query(struct pipe_context *pipe, struct pipe_query *q) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - struct softpipe_query *sq = softpipe_query(q); - - sq->end = softpipe->occlusion_count; -} - - -static boolean -softpipe_get_query_result(struct pipe_context *pipe, - struct pipe_query *q, - boolean wait, - uint64 *result ) -{ - struct softpipe_query *sq = softpipe_query(q); - *result = sq->end - sq->start; - return TRUE; -} - - -void softpipe_init_query_funcs(struct softpipe_context *softpipe ) -{ - softpipe->pipe.create_query = softpipe_create_query; - softpipe->pipe.destroy_query = softpipe_destroy_query; - softpipe->pipe.begin_query = softpipe_begin_query; - softpipe->pipe.end_query = softpipe_end_query; - softpipe->pipe.get_query_result = softpipe_get_query_result; -} - - diff --git a/src/mesa/pipe/softpipe/sp_query.h b/src/mesa/pipe/softpipe/sp_query.h deleted file mode 100644 index 05060a4575..0000000000 --- a/src/mesa/pipe/softpipe/sp_query.h +++ /dev/null @@ -1,39 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - -#ifndef SP_QUERY_H -#define SP_QUERY_H - -struct softpipe_context; -extern void softpipe_init_query_funcs(struct softpipe_context * ); - - -#endif /* SP_QUERY_H */ diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h deleted file mode 100644 index b79db0d1f1..0000000000 --- a/src/mesa/pipe/softpipe/sp_state.h +++ /dev/null @@ -1,187 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef SP_STATE_H -#define SP_STATE_H - -#include "pipe/p_state.h" - -#include "x86/rtasm/x86sse.h" - - -#define SP_NEW_VIEWPORT 0x1 -#define SP_NEW_RASTERIZER 0x2 -#define SP_NEW_FS 0x4 -#define SP_NEW_BLEND 0x8 -#define SP_NEW_CLIP 0x10 -#define SP_NEW_SCISSOR 0x20 -#define SP_NEW_STIPPLE 0x40 -#define SP_NEW_FRAMEBUFFER 0x80 -#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100 -#define SP_NEW_CONSTANTS 0x200 -#define SP_NEW_SAMPLER 0x400 -#define SP_NEW_TEXTURE 0x800 -#define SP_NEW_VERTEX 0x1000 -#define SP_NEW_VS 0x2000 -#define SP_NEW_QUERY 0x4000 - - - -#ifdef MESA_LLVM -struct gallivm_prog; -#endif - - -struct vertex_info; - - -/** Subclass of pipe_shader_state */ -struct sp_fragment_shader_state { - struct pipe_shader_state shader; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif -#ifdef MESA_LLVM - struct gallivm_prog *llvm_prog; -#endif -}; - - -/** Subclass of pipe_shader_state */ -struct sp_vertex_shader_state { - struct pipe_shader_state shader; - struct draw_vertex_shader *draw_data; -}; - - - -void * -softpipe_create_blend_state(struct pipe_context *, - const struct pipe_blend_state *); -void softpipe_bind_blend_state(struct pipe_context *, - void *); -void softpipe_delete_blend_state(struct pipe_context *, - void *); - -void * -softpipe_create_sampler_state(struct pipe_context *, - const struct pipe_sampler_state *); -void softpipe_bind_sampler_state(struct pipe_context *, unsigned, void *); -void softpipe_delete_sampler_state(struct pipe_context *, void *); - -void * -softpipe_create_depth_stencil_state(struct pipe_context *, - const struct pipe_depth_stencil_alpha_state *); -void softpipe_bind_depth_stencil_state(struct pipe_context *, void *); -void softpipe_delete_depth_stencil_state(struct pipe_context *, void *); - -void * -softpipe_create_rasterizer_state(struct pipe_context *, - const struct pipe_rasterizer_state *); -void softpipe_bind_rasterizer_state(struct pipe_context *, void *); -void softpipe_delete_rasterizer_state(struct pipe_context *, void *); - -void softpipe_set_framebuffer_state( struct pipe_context *, - const struct pipe_framebuffer_state * ); - -void softpipe_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ); - -void softpipe_set_clip_state( struct pipe_context *, - const struct pipe_clip_state * ); - -void softpipe_set_constant_buffer(struct pipe_context *, - uint shader, uint index, - const struct pipe_constant_buffer *buf); - -void *softpipe_create_fs_state(struct pipe_context *, - const struct pipe_shader_state *); -void softpipe_bind_fs_state(struct pipe_context *, void *); -void softpipe_delete_fs_state(struct pipe_context *, void *); -void *softpipe_create_vs_state(struct pipe_context *, - const struct pipe_shader_state *); -void softpipe_bind_vs_state(struct pipe_context *, void *); -void softpipe_delete_vs_state(struct pipe_context *, void *); - -void softpipe_set_polygon_stipple( struct pipe_context *, - const struct pipe_poly_stipple * ); - -void softpipe_set_scissor_state( struct pipe_context *, - const struct pipe_scissor_state * ); - -void softpipe_set_sampler_texture( struct pipe_context *, - unsigned unit, - struct pipe_texture * ); - -void softpipe_set_viewport_state( struct pipe_context *, - const struct pipe_viewport_state * ); - -void softpipe_set_vertex_element(struct pipe_context *, - unsigned index, - const struct pipe_vertex_element *); - -void softpipe_set_vertex_buffer(struct pipe_context *, - unsigned index, - const struct pipe_vertex_buffer *); - - -void softpipe_update_derived( struct softpipe_context *softpipe ); - - -boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, - unsigned start, unsigned count); - -boolean softpipe_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count); - - -void -softpipe_map_surfaces(struct softpipe_context *sp); - -void -softpipe_unmap_surfaces(struct softpipe_context *sp); - -void -softpipe_map_texture_surfaces(struct softpipe_context *sp); - -void -softpipe_unmap_texture_surfaces(struct softpipe_context *sp); - - -struct vertex_info * -softpipe_get_vertex_info(struct softpipe_context *softpipe); - -struct vertex_info * -softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe); - - -#endif diff --git a/src/mesa/pipe/softpipe/sp_state_blend.c b/src/mesa/pipe/softpipe/sp_state_blend.c deleted file mode 100644 index 2d40d6bd8f..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_blend.c +++ /dev/null @@ -1,98 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_state.h" - - -void * -softpipe_create_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend) -{ - return mem_dup(blend, sizeof(*blend)); -} - -void softpipe_bind_blend_state( struct pipe_context *pipe, - void *blend ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->blend = (const struct pipe_blend_state *)blend; - - softpipe->dirty |= SP_NEW_BLEND; -} - -void softpipe_delete_blend_state(struct pipe_context *pipe, - void *blend) -{ - FREE( blend ); -} - - -void softpipe_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->blend_color = *blend_color; - - softpipe->dirty |= SP_NEW_BLEND; -} - - -/** XXX move someday? Or consolidate all these simple state setters - * into one file. - */ - - -void * -softpipe_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) -{ - return mem_dup(depth_stencil, sizeof(*depth_stencil)); -} - -void -softpipe_bind_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->depth_stencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; - - softpipe->dirty |= SP_NEW_DEPTH_STENCIL_ALPHA; -} - -void -softpipe_delete_depth_stencil_state(struct pipe_context *pipe, void *depth) -{ - FREE( depth ); -} diff --git a/src/mesa/pipe/softpipe/sp_state_clip.c b/src/mesa/pipe/softpipe/sp_state_clip.c deleted file mode 100644 index 08c5f06d05..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_clip.c +++ /dev/null @@ -1,83 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "sp_context.h" -#include "sp_state.h" -#include "pipe/draw/draw_context.h" - - -void softpipe_set_clip_state( struct pipe_context *pipe, - const struct pipe_clip_state *clip ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - /* pass the clip state to the draw module */ - draw_set_clip_state(softpipe->draw, clip); -} - - - -/* Called when driver state tracker notices changes to the viewport - * matrix: - */ -void softpipe_set_viewport_state( struct pipe_context *pipe, - const struct pipe_viewport_state *viewport ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->viewport = *viewport; /* struct copy */ - softpipe->dirty |= SP_NEW_VIEWPORT; - - /* pass the viewport info to the draw module */ - draw_set_viewport_state(softpipe->draw, viewport); - - /* Using tnl/ and vf/ modules is temporary while getting started. - * Full pipe will have vertex shader, vertex fetch of its own. - */ -} - - -void softpipe_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - memcpy( &softpipe->scissor, scissor, sizeof(*scissor) ); - softpipe->dirty |= SP_NEW_SCISSOR; -} - - -void softpipe_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - memcpy( &softpipe->poly_stipple, stipple, sizeof(*stipple) ); - softpipe->dirty |= SP_NEW_STIPPLE; -} diff --git a/src/mesa/pipe/softpipe/sp_state_derived.c b/src/mesa/pipe/softpipe/sp_state_derived.c deleted file mode 100644 index 372597869f..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_derived.c +++ /dev/null @@ -1,235 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_vertex.h" -#include "pipe/draw/draw_private.h" -#include "sp_context.h" -#include "sp_state.h" - - -/** - * Search vertex program's outputs to find a match for the given - * semantic name/index. Return the index of the output slot. - * - * Return 0 if not found. This will cause the fragment program to use - * vertex attrib 0 (position) in the cases where the fragment program - * attempts to use a missing vertex program output. This is an undefined - * condition that users shouldn't hit anyway. - */ -static int -find_vs_output(const struct pipe_shader_state *vs, - uint semantic_name, - uint semantic_index) -{ - uint i; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == semantic_name && - vs->output_semantic_index[i] == semantic_index) - return i; - } - return 0; -} - - -/** - * Mark the current vertex layout as "invalid". - * We'll validate the vertex layout later, when we start to actually - * render a point or line or tri. - */ -static void -invalidate_vertex_layout(struct softpipe_context *softpipe) -{ - softpipe->vertex_info.num_attribs = 0; -} - - -/** - * The vertex info describes how to convert the post-transformed vertices - * (simple float[][4]) used by the 'draw' module into vertices for - * rasterization. - * - * This function validates the vertex layout and returns a pointer to a - * vertex_info object. - */ -struct vertex_info * -softpipe_get_vertex_info(struct softpipe_context *softpipe) -{ - struct vertex_info *vinfo = &softpipe->vertex_info; - - if (vinfo->num_attribs == 0) { - /* compute vertex layout now */ - const struct pipe_shader_state *vs = &softpipe->vs->shader; - const struct pipe_shader_state *fs = &softpipe->fs->shader; - const enum interp_mode colorInterp - = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; - uint i; - - if (softpipe->vbuf) { - /* if using the post-transform vertex buffer, tell draw_vbuf to - * simply emit the whole post-xform vertex as-is: - */ - struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; - vinfo_vbuf->num_attribs = 0; - draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); - vinfo_vbuf->size = 4 * vs->num_outputs - + sizeof(struct vertex_header) / 4; - } - - /* - * Loop over fragment shader inputs, searching for the matching output - * from the vertex shader. - */ - vinfo->num_attribs = 0; - for (i = 0; i < fs->num_inputs; i++) { - int src; - switch (fs->input_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); - break; - - case TGSI_SEMANTIC_COLOR: - src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, - fs->input_semantic_index[i]); - draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); - break; - - case TGSI_SEMANTIC_FOG: - src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); - break; - - case TGSI_SEMANTIC_GENERIC: - /* this includes texcoords and varying vars */ - src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, - fs->input_semantic_index[i]); - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); - break; - - default: - assert(0); - } - } - - softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); - if (softpipe->psize_slot > 0) { - draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, - softpipe->psize_slot); - } - - draw_compute_vertex_size(vinfo); - } - - return vinfo; -} - - -/** - * Called from vbuf module. - * - * Note that there's actually two different vertex layouts in softpipe. - * - * The normal one is computed in softpipe_get_vertex_info() above and is - * used by the point/line/tri "setup" code. - * - * The other one (this one) is only used by the vbuf module (which is - * not normally used by default but used in testing). For the vbuf module, - * we basically want to pass-through the draw module's vertex layout as-is. - * When the softpipe vbuf code begins drawing, the normal vertex layout - * will come into play again. - */ -struct vertex_info * -softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe) -{ - (void) softpipe_get_vertex_info(softpipe); - return &softpipe->vertex_info_vbuf; -} - - -/** - * Recompute cliprect from scissor bounds, scissor enable and surface size. - */ -static void -compute_cliprect(struct softpipe_context *sp) -{ - unsigned surfWidth, surfHeight; - - if (sp->framebuffer.num_cbufs > 0) { - surfWidth = sp->framebuffer.cbufs[0]->width; - surfHeight = sp->framebuffer.cbufs[0]->height; - } - else { - /* no surface? */ - surfWidth = sp->scissor.maxx; - surfHeight = sp->scissor.maxy; - } - - if (sp->rasterizer->scissor) { - /* clip to scissor rect */ - sp->cliprect.minx = MAX2(sp->scissor.minx, 0); - sp->cliprect.miny = MAX2(sp->scissor.miny, 0); - sp->cliprect.maxx = MIN2(sp->scissor.maxx, surfWidth); - sp->cliprect.maxy = MIN2(sp->scissor.maxy, surfHeight); - } - else { - /* clip to surface bounds */ - sp->cliprect.minx = 0; - sp->cliprect.miny = 0; - sp->cliprect.maxx = surfWidth; - sp->cliprect.maxy = surfHeight; - } -} - - -/* Hopefully this will remain quite simple, otherwise need to pull in - * something like the state tracker mechanism. - */ -void softpipe_update_derived( struct softpipe_context *softpipe ) -{ - if (softpipe->dirty & (SP_NEW_RASTERIZER | - SP_NEW_FS | - SP_NEW_VS)) - invalidate_vertex_layout( softpipe ); - - if (softpipe->dirty & (SP_NEW_SCISSOR | - SP_NEW_DEPTH_STENCIL_ALPHA | - SP_NEW_FRAMEBUFFER)) - compute_cliprect(softpipe); - - if (softpipe->dirty & (SP_NEW_BLEND | - SP_NEW_DEPTH_STENCIL_ALPHA | - SP_NEW_FRAMEBUFFER | - SP_NEW_RASTERIZER | - SP_NEW_FS | - SP_NEW_QUERY)) - sp_build_quad_pipeline(softpipe); - - softpipe->dirty = 0; -} diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c deleted file mode 100644 index 0b814fc284..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ /dev/null @@ -1,179 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "sp_context.h" -#include "sp_state.h" - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" -#include "pipe/draw/draw_context.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/llvm/gallivm.h" -#include "pipe/tgsi/util/tgsi_dump.h" -#include "pipe/tgsi/exec/tgsi_sse2.h" - - -void * -softpipe_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - struct sp_fragment_shader_state *state; - - /* Decide whether we'll be codegenerating this shader and if so do - * that now. - */ - - state = CALLOC_STRUCT(sp_fragment_shader_state); - if (!state) - return NULL; - - state->shader = *templ; - - if (softpipe->dump_fs) { - tgsi_dump(state->shader.tokens, 0); - } - -#ifdef MESA_LLVM - state->llvm_prog = 0; - -#if 0 - if (!gallivm_global_cpu_engine()) { - gallivm_cpu_engine_create(state->llvm_prog); - } - else - gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); -#endif - -#elif defined(__i386__) || defined(__386__) - if (softpipe->use_sse) { - x86_init_func( &state->sse2_program ); - tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); - } -#endif - - return state; -} - - -void -softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->fs = (struct sp_fragment_shader_state *) fs; - - softpipe->dirty |= SP_NEW_FS; -} - - -void -softpipe_delete_fs_state(struct pipe_context *pipe, void *fs) -{ - struct sp_fragment_shader_state *state = fs; - -#if defined(__i386__) || defined(__386__) - x86_release_func( &state->sse2_program ); -#endif - - FREE( state ); -} - - -void * -softpipe_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - struct sp_vertex_shader_state *state; - - state = CALLOC_STRUCT(sp_vertex_shader_state); - if (state == NULL ) { - return NULL; - } - - state->shader = *templ; - - state->draw_data = draw_create_vertex_shader(softpipe->draw, - &state->shader); - if (state->draw_data == NULL) { - FREE( state ); - return NULL; - } - - return state; -} - - -void -softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - softpipe->vs = (const struct sp_vertex_shader_state *)vs; - - draw_bind_vertex_shader(softpipe->draw, softpipe->vs->draw_data); - - softpipe->dirty |= SP_NEW_VS; -} - - -void -softpipe_delete_vs_state(struct pipe_context *pipe, void *vs) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - struct sp_vertex_shader_state *state = - (struct sp_vertex_shader_state *)vs; - - draw_delete_vertex_shader(softpipe->draw, state->draw_data); - FREE( state ); -} - - - -void -softpipe_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - struct pipe_winsys *ws = pipe->winsys; - - assert(shader < PIPE_SHADER_TYPES); - assert(index == 0); - - /* note: reference counting */ - pipe_buffer_reference(ws, - &softpipe->constants[shader].buffer, - buf->buffer); - softpipe->constants[shader].size = buf->size; - - softpipe->dirty |= SP_NEW_CONSTANTS; -} diff --git a/src/mesa/pipe/softpipe/sp_state_rasterizer.c b/src/mesa/pipe/softpipe/sp_state_rasterizer.c deleted file mode 100644 index 53755099dd..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_rasterizer.c +++ /dev/null @@ -1,62 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "sp_context.h" -#include "sp_state.h" -#include "pipe/draw/draw_context.h" - - - -void * -softpipe_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *rast) -{ - return mem_dup(rast, sizeof(*rast)); -} - -void softpipe_bind_rasterizer_state(struct pipe_context *pipe, - void *setup) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - /* pass-through to draw module */ - draw_set_rasterizer_state(softpipe->draw, setup); - - softpipe->rasterizer = (struct pipe_rasterizer_state *)setup; - - softpipe->dirty |= SP_NEW_RASTERIZER; -} - -void softpipe_delete_rasterizer_state(struct pipe_context *pipe, - void *rasterizer) -{ - FREE( rasterizer ); -} - - diff --git a/src/mesa/pipe/softpipe/sp_state_sampler.c b/src/mesa/pipe/softpipe/sp_state_sampler.c deleted file mode 100644 index ea348c7e95..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_sampler.c +++ /dev/null @@ -1,93 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: - * Brian Paul - */ - -#include "pipe/p_util.h" - -#include "pipe/draw/draw_context.h" - -#include "sp_context.h" -#include "sp_context.h" -#include "sp_state.h" -#include "sp_texture.h" -#include "sp_tile_cache.h" -#include "pipe/draw/draw_context.h" - - - -void * -softpipe_create_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) -{ - return mem_dup(sampler, sizeof(*sampler)); -} - -void -softpipe_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - draw_flush(softpipe->draw); - - assert(unit < PIPE_MAX_SAMPLERS); - softpipe->sampler[unit] = (struct pipe_sampler_state *)sampler; - - softpipe->dirty |= SP_NEW_SAMPLER; -} - - -void -softpipe_delete_sampler_state(struct pipe_context *pipe, - void *sampler) -{ - FREE( sampler ); -} - - -void -softpipe_set_sampler_texture(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - - draw_flush(softpipe->draw); - - assert(unit < PIPE_MAX_SAMPLERS); - softpipe->texture[unit] = softpipe_texture(texture); /* ptr, not struct */ - - sp_tile_cache_set_texture(softpipe->tex_cache[unit], texture); - - softpipe->dirty |= SP_NEW_TEXTURE; -} - - - diff --git a/src/mesa/pipe/softpipe/sp_state_surface.c b/src/mesa/pipe/softpipe/sp_state_surface.c deleted file mode 100644 index e2c6893e9f..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_surface.c +++ /dev/null @@ -1,109 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ -#include "p_inlines.h" - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_surface.h" -#include "sp_tile_cache.h" - - -/** - * XXX this might get moved someday - * Set the framebuffer surface info: color buffers, zbuffer, stencil buffer. - * Here, we flush the old surfaces and update the tile cache to point to the new - * surfaces. - */ -void -softpipe_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb) -{ - struct softpipe_context *sp = softpipe_context(pipe); - uint i; - - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { - /* check if changing cbuf */ - if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) { - /* flush old */ - sp_flush_tile_cache(sp, sp->cbuf_cache[i]); - - /* assign new */ - sp->framebuffer.cbufs[i] = fb->cbufs[i]; - - /* update cache */ - sp_tile_cache_set_surface(sp->cbuf_cache[i], fb->cbufs[i]); - } - } - - sp->framebuffer.num_cbufs = fb->num_cbufs; - - /* zbuf changing? */ - if (sp->framebuffer.zsbuf != fb->zsbuf) { - /* flush old */ - sp_flush_tile_cache(sp, sp->zsbuf_cache); - - /* assign new */ - sp->framebuffer.zsbuf = fb->zsbuf; - - /* update cache */ - sp_tile_cache_set_surface(sp->zsbuf_cache, fb->zsbuf); - } - -#if 0 - /* XXX combined depth/stencil here */ - - /* sbuf changing? */ - if (sp->framebuffer.sbuf != fb->sbuf) { - /* flush old */ - sp_flush_tile_cache(sp, sp->sbuf_cache_sep); - - /* assign new */ - sp->framebuffer.sbuf = fb->sbuf; - - /* update cache */ - if (fb->sbuf != fb->zbuf) { - /* separate stencil buf */ - sp->sbuf_cache = sp->sbuf_cache_sep; - sp_tile_cache_set_surface(sp->sbuf_cache, fb->sbuf); - } - else { - /* combined depth/stencil */ - sp->sbuf_cache = sp->zbuf_cache; - sp_tile_cache_set_surface(sp->sbuf_cache, fb->sbuf); - } - } -#endif - - sp->dirty |= SP_NEW_FRAMEBUFFER; -} - - - - diff --git a/src/mesa/pipe/softpipe/sp_state_vertex.c b/src/mesa/pipe/softpipe/sp_state_vertex.c deleted file mode 100644 index 09ff540ccf..0000000000 --- a/src/mesa/pipe/softpipe/sp_state_vertex.c +++ /dev/null @@ -1,64 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_surface.h" - -#include "pipe/draw/draw_context.h" - - -void -softpipe_set_vertex_element(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_element *attrib) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - assert(index < PIPE_ATTRIB_MAX); - softpipe->vertex_element[index] = *attrib; /* struct copy */ - softpipe->dirty |= SP_NEW_VERTEX; - - draw_set_vertex_element(softpipe->draw, index, attrib); -} - - -void -softpipe_set_vertex_buffer(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - assert(index < PIPE_ATTRIB_MAX); - softpipe->vertex_buffer[index] = *buffer; /* struct copy */ - softpipe->dirty |= SP_NEW_VERTEX; - - draw_set_vertex_buffer(softpipe->draw, index, buffer); -} diff --git a/src/mesa/pipe/softpipe/sp_surface.c b/src/mesa/pipe/softpipe/sp_surface.c deleted file mode 100644 index 8802ced187..0000000000 --- a/src/mesa/pipe/softpipe/sp_surface.c +++ /dev/null @@ -1,159 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" -#include "pipe/util/p_tile.h" -#include "sp_context.h" -#include "sp_surface.h" - - - -/* Assumes all values are within bounds -- no checking at this level - - * do it higher up if required. - */ -static void -sp_surface_copy(struct pipe_context *pipe, - unsigned do_flip, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - struct pipe_surface *src, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - assert( dst->cpp == src->cpp ); - - pipe_copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, - width, height, - pipe_surface_map(src), - do_flip ? -(int) src->pitch : src->pitch, - srcx, do_flip ? 1 - srcy - height : srcy); - - pipe_surface_unmap(src); - pipe_surface_unmap(dst); -} - - -static void * -get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) -{ - return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; -} - - -#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8)) - - -/** - * Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static void -sp_surface_fill(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - unsigned width, unsigned height, unsigned value) -{ - unsigned i, j; - void *dst_map = pipe_surface_map(dst); - - assert(dst->pitch > 0); - assert(width <= dst->pitch); - - - switch (dst->cpp) { - case 1: - { - ubyte *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; - case 2: - { - ushort *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; - case 4: - { - unsigned *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - break; - case 8: - { - /* expand the 4-byte clear value to an 8-byte value */ - ushort *row = (ushort *) get_pointer(dst, dst_map, dstx, dsty); - ushort val0 = UBYTE_TO_USHORT((value >> 0) & 0xff); - ushort val1 = UBYTE_TO_USHORT((value >> 8) & 0xff); - ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff); - ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff); - val0 = (val0 << 8) | val0; - val1 = (val1 << 8) | val1; - val2 = (val2 << 8) | val2; - val3 = (val3 << 8) | val3; - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) { - row[j*4+0] = val0; - row[j*4+1] = val1; - row[j*4+2] = val2; - row[j*4+3] = val3; - } - row += dst->pitch * 4; - } - } - break; - default: - assert(0); - break; - } - - pipe_surface_unmap( dst ); -} - - -void -sp_init_surface_functions(struct softpipe_context *sp) -{ - sp->pipe.surface_copy = sp_surface_copy; - sp->pipe.surface_fill = sp_surface_fill; -} diff --git a/src/mesa/pipe/softpipe/sp_surface.h b/src/mesa/pipe/softpipe/sp_surface.h deleted file mode 100644 index 22de3ba43f..0000000000 --- a/src/mesa/pipe/softpipe/sp_surface.h +++ /dev/null @@ -1,42 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef SP_SURFACE_H -#define SP_SURFACE_H - - -struct softpipe_context; - - -extern void -sp_init_surface_functions(struct softpipe_context *sp); - - -#endif /* SP_SURFACE_H */ diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.c b/src/mesa/pipe/softpipe/sp_tex_sample.c deleted file mode 100644 index 325bdb86da..0000000000 --- a/src/mesa/pipe/softpipe/sp_tex_sample.c +++ /dev/null @@ -1,916 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Texture sampling - * - * Authors: - * Brian Paul - */ - -#include "sp_context.h" -#include "sp_headers.h" -#include "sp_surface.h" -#include "sp_tex_sample.h" -#include "sp_tile_cache.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/tgsi/exec/tgsi_exec.h" - - -/* - * Note, the FRAC macro has to work perfectly. Otherwise you'll sometimes - * see 1-pixel bands of improperly weighted linear-filtered textures. - * The tests/texwrap.c demo is a good test. - * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0. - * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x). - */ -#define FRAC(f) ((f) - ifloor(f)) - - -/** - * Linear interpolation macro - */ -#define LERP(T, A, B) ( (A) + (T) * ((B) - (A)) ) - - -/** - * Do 2D/biliner interpolation of float values. - * v00, v10, v01 and v11 are typically four texture samples in a square/box. - * a and b are the horizontal and vertical interpolants. - * It's important that this function is inlined when compiled with - * optimization! If we find that's not true on some systems, convert - * to a macro. - */ -static INLINE float -lerp_2d(float a, float b, - float v00, float v10, float v01, float v11) -{ - const float temp0 = LERP(a, v00, v10); - const float temp1 = LERP(a, v01, v11); - return LERP(b, temp0, temp1); -} - - -/** - * If A is a signed integer, A % B doesn't give the right value for A < 0 - * (in terms of texture repeat). Just casting to unsigned fixes that. - */ -#define REMAINDER(A, B) ((unsigned) (A) % (unsigned) (B)) - - -/** - * Apply texture coord wrapping mode and return integer texture index. - * \param wrapMode PIPE_TEX_WRAP_x - * \param s the texcoord - * \param size the texture image size - * \return integer texture index - */ -static INLINE int -nearest_texcoord(unsigned wrapMode, float s, unsigned size) -{ - int i; - switch (wrapMode) { - case PIPE_TEX_WRAP_REPEAT: - /* s limited to [0,1) */ - /* i limited to [0,size-1] */ - i = ifloor(s * size); - i = REMAINDER(i, size); - return i; - case PIPE_TEX_WRAP_CLAMP: - /* s limited to [0,1] */ - /* i limited to [0,size-1] */ - if (s <= 0.0F) - i = 0; - else if (s >= 1.0F) - i = size - 1; - else - i = ifloor(s * size); - return i; - case PIPE_TEX_WRAP_CLAMP_TO_EDGE: - { - /* s limited to [min,max] */ - /* i limited to [0, size-1] */ - const float min = 1.0F / (2.0F * size); - const float max = 1.0F - min; - if (s < min) - i = 0; - else if (s > max) - i = size - 1; - else - i = ifloor(s * size); - } - return i; - case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - { - /* s limited to [min,max] */ - /* i limited to [-1, size] */ - const float min = -1.0F / (2.0F * size); - const float max = 1.0F - min; - if (s <= min) - i = -1; - else if (s >= max) - i = size; - else - i = ifloor(s * size); - } - return i; - case PIPE_TEX_WRAP_MIRROR_REPEAT: - { - const float min = 1.0F / (2.0F * size); - const float max = 1.0F - min; - const int flr = ifloor(s); - float u; - if (flr & 1) - u = 1.0F - (s - (float) flr); - else - u = s - (float) flr; - if (u < min) - i = 0; - else if (u > max) - i = size - 1; - else - i = ifloor(u * size); - } - return i; - case PIPE_TEX_WRAP_MIRROR_CLAMP: - { - /* s limited to [0,1] */ - /* i limited to [0,size-1] */ - const float u = FABSF(s); - if (u <= 0.0F) - i = 0; - else if (u >= 1.0F) - i = size - 1; - else - i = ifloor(u * size); - } - return i; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: - { - /* s limited to [min,max] */ - /* i limited to [0, size-1] */ - const float min = 1.0F / (2.0F * size); - const float max = 1.0F - min; - const float u = FABSF(s); - if (u < min) - i = 0; - else if (u > max) - i = size - 1; - else - i = ifloor(u * size); - } - return i; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: - { - /* s limited to [min,max] */ - /* i limited to [0, size-1] */ - const float min = -1.0F / (2.0F * size); - const float max = 1.0F - min; - const float u = FABSF(s); - if (u < min) - i = -1; - else if (u > max) - i = size; - else - i = ifloor(u * size); - } - return i; - default: - assert(0); - return 0; - } -} - - -/** - * Used to compute texel locations for linear sampling. - * \param wrapMode PIPE_TEX_WRAP_x - * \param s the texcoord - * \param size the texture image size - * \param i0 returns first texture index - * \param i1 returns second texture index (usually *i0 + 1) - * \param a returns blend factor/weight between texture indexes - */ -static INLINE void -linear_texcoord(unsigned wrapMode, float s, unsigned size, - int *i0, int *i1, float *a) -{ - float u; - switch (wrapMode) { - case PIPE_TEX_WRAP_REPEAT: - u = s * size - 0.5F; - *i0 = REMAINDER(ifloor(u), size); - *i1 = REMAINDER(*i0 + 1, size); - break; - case PIPE_TEX_WRAP_CLAMP: - if (s <= 0.0F) - u = 0.0F; - else if (s >= 1.0F) - u = (float) size; - else - u = s * size; - u -= 0.5F; - *i0 = ifloor(u); - *i1 = *i0 + 1; - break; - case PIPE_TEX_WRAP_CLAMP_TO_EDGE: - if (s <= 0.0F) - u = 0.0F; - else if (s >= 1.0F) - u = (float) size; - else - u = s * size; - u -= 0.5F; - *i0 = ifloor(u); - *i1 = *i0 + 1; - if (*i0 < 0) - *i0 = 0; - if (*i1 >= (int) size) - *i1 = size - 1; - break; - case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - { - const float min = -1.0F / (2.0F * size); - const float max = 1.0F - min; - if (s <= min) - u = min * size; - else if (s >= max) - u = max * size; - else - u = s * size; - u -= 0.5F; - *i0 = ifloor(u); - *i1 = *i0 + 1; - } - break; - case PIPE_TEX_WRAP_MIRROR_REPEAT: - { - const int flr = ifloor(s); - if (flr & 1) - u = 1.0F - (s - (float) flr); - else - u = s - (float) flr; - u = (u * size) - 0.5F; - *i0 = ifloor(u); - *i1 = *i0 + 1; - if (*i0 < 0) - *i0 = 0; - if (*i1 >= (int) size) - *i1 = size - 1; - } - break; - case PIPE_TEX_WRAP_MIRROR_CLAMP: - u = FABSF(s); - if (u >= 1.0F) - u = (float) size; - else - u *= size; - u -= 0.5F; - *i0 = ifloor(u); - *i1 = *i0 + 1; - break; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: - u = FABSF(s); - if (u >= 1.0F) - u = (float) size; - else - u *= size; - u -= 0.5F; - *i0 = ifloor(u); - *i1 = *i0 + 1; - if (*i0 < 0) - *i0 = 0; - if (*i1 >= (int) size) - *i1 = size - 1; - break; - case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: - { - const float min = -1.0F / (2.0F * size); - const float max = 1.0F - min; - u = FABSF(s); - if (u <= min) - u = min * size; - else if (u >= max) - u = max * size; - else - u *= size; - u -= 0.5F; - *i0 = ifloor(u); - *i1 = *i0 + 1; - } - break; - default: - assert(0); - } - *a = FRAC(u); -} - - -static unsigned -choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) -{ - /* - major axis - direction target sc tc ma - ---------- ------------------------------- --- --- --- - +rx TEXTURE_CUBE_MAP_POSITIVE_X_EXT -rz -ry rx - -rx TEXTURE_CUBE_MAP_NEGATIVE_X_EXT +rz -ry rx - +ry TEXTURE_CUBE_MAP_POSITIVE_Y_EXT +rx +rz ry - -ry TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT +rx -rz ry - +rz TEXTURE_CUBE_MAP_POSITIVE_Z_EXT +rx -ry rz - -rz TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT -rx -ry rz - */ - const float arx = FABSF(rx), ary = FABSF(ry), arz = FABSF(rz); - unsigned face; - float sc, tc, ma; - - if (arx > ary && arx > arz) { - if (rx >= 0.0F) { - face = PIPE_TEX_FACE_POS_X; - sc = -rz; - tc = -ry; - ma = arx; - } - else { - face = PIPE_TEX_FACE_NEG_X; - sc = rz; - tc = -ry; - ma = arx; - } - } - else if (ary > arx && ary > arz) { - if (ry >= 0.0F) { - face = PIPE_TEX_FACE_POS_Y; - sc = rx; - tc = rz; - ma = ary; - } - else { - face = PIPE_TEX_FACE_NEG_Y; - sc = rx; - tc = -rz; - ma = ary; - } - } - else { - if (rz > 0.0F) { - face = PIPE_TEX_FACE_POS_Z; - sc = rx; - tc = -ry; - ma = arz; - } - else { - face = PIPE_TEX_FACE_NEG_Z; - sc = -rx; - tc = -ry; - ma = arz; - } - } - - *newS = ( sc / ma + 1.0F ) * 0.5F; - *newT = ( tc / ma + 1.0F ) * 0.5F; - - return face; -} - - -/** - * Examine the quad's texture coordinates to compute the partial - * derivatives w.r.t X and Y, then compute lambda (level of detail). - * - * This is only done for fragment shaders, not vertex shaders. - */ -static float -compute_lambda(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias) -{ - float rho, lambda; - - assert(s); - { - float dsdx = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]; - float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; - dsdx = FABSF(dsdx); - dsdy = FABSF(dsdy); - rho = MAX2(dsdx, dsdy); - if (sampler->state->normalized_coords) - rho *= sampler->texture->width[0]; - } - if (t) { - float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; - float dtdy = t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]; - float max; - dtdx = FABSF(dtdx); - dtdy = FABSF(dtdy); - max = MAX2(dtdx, dtdy); - if (sampler->state->normalized_coords) - max *= sampler->texture->height[0]; - rho = MAX2(rho, max); - } - if (p) { - float dpdx = p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT]; - float dpdy = p[QUAD_TOP_LEFT] - p[QUAD_BOTTOM_LEFT]; - float max; - dpdx = FABSF(dpdx); - dpdy = FABSF(dpdy); - max = MAX2(dpdx, dpdy); - if (sampler->state->normalized_coords) - max *= sampler->texture->depth[0]; - rho = MAX2(rho, max); - } - - lambda = LOG2(rho); - lambda += lodbias + sampler->state->lod_bias; - lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); - - return lambda; -} - - -/** - * Do several things here: - * 1. Compute lambda from the texcoords, if needed - * 2. Determine if we're minifying or magnifying - * 3. If minifying, choose mipmap levels - * 4. Return image filter to use within mipmap images - */ -static void -choose_mipmap_levels(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - unsigned *level0, unsigned *level1, float *levelBlend, - unsigned *imgFilter) -{ - if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { - /* no mipmap selection needed */ - *imgFilter = sampler->state->mag_img_filter; - *level0 = *level1 = (int) sampler->state->min_lod; - } - else { - float lambda; - - if (1) - /* fragment shader */ - lambda = compute_lambda(sampler, s, t, p, lodbias); - else - /* vertex shader */ - lambda = lodbias; /* not really a bias, but absolute LOD */ - - if (lambda < 0.0) { /* XXX threshold depends on the filter */ - /* magnifying */ - *imgFilter = sampler->state->mag_img_filter; - *level0 = *level1 = 0; - } - else { - /* minifying */ - *imgFilter = sampler->state->min_img_filter; - - /* choose mipmap level(s) and compute the blend factor between them */ - if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NEAREST) { - /* Nearest mipmap level */ - const int lvl = (int) (lambda + 0.5); - *level0 = - *level1 = CLAMP(lvl, 0, (int) sampler->texture->last_level); - } - else { - /* Linear interpolation between mipmap levels */ - const int lvl = (int) lambda; - *level0 = CLAMP(lvl, 0, (int) sampler->texture->last_level); - *level1 = CLAMP(lvl + 1, 0, (int) sampler->texture->last_level); - *levelBlend = FRAC(lambda); /* blending weight between levels */ - } - } - } -} - - -/** - * Get a texel from a texture, using the texture tile cache. - * - * \param face the cube face in 0..5 - * \param level the mipmap level - * \param x the x coord of texel within 2D image - * \param y the y coord of texel within 2D image - * \param z which slice of a 3D texture - * \param rgba the quad to put the texel/color into - * \param j which element of the rgba quad to write to - * - * XXX maybe move this into sp_tile_cache.c and merge with the - * sp_get_cached_tile_tex() function. Also, get 4 texels instead of 1... - */ -static void -get_texel(struct tgsi_sampler *sampler, - unsigned face, unsigned level, int x, int y, int z, - float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j) -{ - const int tx = x % TILE_SIZE; - const int ty = y % TILE_SIZE; - const struct softpipe_cached_tile *tile - = sp_get_cached_tile_tex(sampler->pipe, sampler->cache, - x, y, z, face, level); - rgba[0][j] = tile->data.color[ty][tx][0]; - rgba[1][j] = tile->data.color[ty][tx][1]; - rgba[2][j] = tile->data.color[ty][tx][2]; - rgba[3][j] = tile->data.color[ty][tx][3]; -} - - -/** - * Compare texcoord 'p' (aka R) against texture value 'rgba[0]' - * When we sampled the depth texture, the depth value was put into all - * RGBA channels. We look at the red channel here. - */ -static INLINE void -shadow_compare(uint compare_func, - float rgba[NUM_CHANNELS][QUAD_SIZE], - const float p[QUAD_SIZE], - uint j) -{ - int k; - switch (compare_func) { - case PIPE_FUNC_LESS: - k = p[j] < rgba[0][j]; - break; - case PIPE_FUNC_LEQUAL: - k = p[j] <= rgba[0][j]; - break; - case PIPE_FUNC_GREATER: - k = p[j] > rgba[0][j]; - break; - case PIPE_FUNC_GEQUAL: - k = p[j] >= rgba[0][j]; - break; - case PIPE_FUNC_EQUAL: - k = p[j] == rgba[0][j]; - break; - case PIPE_FUNC_NOTEQUAL: - k = p[j] != rgba[0][j]; - break; - case PIPE_FUNC_ALWAYS: - k = 1; - break; - case PIPE_FUNC_NEVER: - k = 0; - break; - default: - assert(0); - } - - rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k; -} - - -/** - * Common code for sampling 1D/2D/cube textures. - * Could probably extend for 3D... - */ -static void -sp_get_samples_2d_common(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE], - const unsigned faces[4]) -{ - const uint compare_func = sampler->state->compare_func; - unsigned level0, level1, j, imgFilter; - int width, height; - float levelBlend; - - choose_mipmap_levels(sampler, s, t, p, lodbias, - &level0, &level1, &levelBlend, &imgFilter); - - if (sampler->state->normalized_coords) { - width = sampler->texture->width[level0]; - height = sampler->texture->height[level0]; - } - else { - width = height = 1; - } - - assert(width > 0); - - switch (imgFilter) { - case PIPE_TEX_FILTER_NEAREST: - for (j = 0; j < QUAD_SIZE; j++) { - int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); - int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); - get_texel(sampler, faces[j], level0, x, y, 0, rgba, j); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, rgba, p, j); - } - - if (level0 != level1) { - /* get texels from second mipmap level and blend */ - float rgba2[4][4]; - unsigned c; - x = x / 2; - y = y / 2; - get_texel(sampler, faces[j], level1, x, y, 0, rgba2, j); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ - shadow_compare(compare_func, rgba2, p, j); - } - - for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); - } - } - } - break; - case PIPE_TEX_FILTER_LINEAR: - for (j = 0; j < QUAD_SIZE; j++) { - float tx[4][4], a, b; - int x0, y0, x1, y1, c; - linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &a); - linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &b); - get_texel(sampler, faces[j], level0, x0, y0, 0, tx, 0); - get_texel(sampler, faces[j], level0, x1, y0, 0, tx, 1); - get_texel(sampler, faces[j], level0, x0, y1, 0, tx, 2); - get_texel(sampler, faces[j], level0, x1, y1, 0, tx, 3); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { - shadow_compare(compare_func, tx, p, 0); - shadow_compare(compare_func, tx, p, 1); - shadow_compare(compare_func, tx, p, 2); - shadow_compare(compare_func, tx, p, 3); - } - - for (c = 0; c < 4; c++) { - rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); - } - - if (level0 != level1) { - /* get texels from second mipmap level and blend */ - float rgba2[4][4]; - x0 = x0 / 2; - y0 = y0 / 2; - x1 = x1 / 2; - y1 = y1 / 2; - get_texel(sampler, faces[j], level1, x0, y0, 0, tx, 0); - get_texel(sampler, faces[j], level1, x1, y0, 0, tx, 1); - get_texel(sampler, faces[j], level1, x0, y1, 0, tx, 2); - get_texel(sampler, faces[j], level1, x1, y1, 0, tx, 3); - if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){ - shadow_compare(compare_func, tx, p, 0); - shadow_compare(compare_func, tx, p, 1); - shadow_compare(compare_func, tx, p, 2); - shadow_compare(compare_func, tx, p, 3); - } - - for (c = 0; c < 4; c++) { - rgba2[c][j] = lerp_2d(a, b, - tx[c][0], tx[c][1], tx[c][2], tx[c][3]); - } - - for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); - } - } - } - break; - default: - assert(0); - } -} - - -static void -sp_get_samples_1d(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]) -{ - static const unsigned faces[4] = {0, 0, 0, 0}; - static const float tzero[4] = {0, 0, 0, 0}; - sp_get_samples_2d_common(sampler, s, tzero, NULL, lodbias, rgba, faces); -} - - -static void -sp_get_samples_2d(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]) -{ - static const unsigned faces[4] = {0, 0, 0, 0}; - sp_get_samples_2d_common(sampler, s, t, p, lodbias, rgba, faces); -} - - -static void -sp_get_samples_3d(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]) -{ - /* get/map pipe_surfaces corresponding to 3D tex slices */ - unsigned level0, level1, j, imgFilter; - int width, height, depth; - float levelBlend; - const uint face = 0; - - choose_mipmap_levels(sampler, s, t, p, lodbias, - &level0, &level1, &levelBlend, &imgFilter); - - if (sampler->state->normalized_coords) { - width = sampler->texture->width[level0]; - height = sampler->texture->height[level0]; - depth = sampler->texture->depth[level0]; - } - else { - width = height = depth = 1; - } - - assert(width > 0); - assert(height > 0); - assert(depth > 0); - - switch (imgFilter) { - case PIPE_TEX_FILTER_NEAREST: - for (j = 0; j < QUAD_SIZE; j++) { - int x = nearest_texcoord(sampler->state->wrap_s, s[j], width); - int y = nearest_texcoord(sampler->state->wrap_t, t[j], height); - int z = nearest_texcoord(sampler->state->wrap_r, p[j], depth); - get_texel(sampler, face, level0, x, y, z, rgba, j); - - if (level0 != level1) { - /* get texels from second mipmap level and blend */ - float rgba2[4][4]; - unsigned c; - x /= 2; - y /= 2; - z /= 2; - get_texel(sampler, face, level1, x, y, z, rgba2, j); - for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(levelBlend, rgba2[c][j], rgba[c][j]); - } - } - } - break; - case PIPE_TEX_FILTER_LINEAR: - for (j = 0; j < QUAD_SIZE; j++) { - float texel0[4][4], texel1[4][4]; - float xw, yw, zw; /* interpolation weights */ - int x0, x1, y0, y1, z0, z1, c; - linear_texcoord(sampler->state->wrap_s, s[j], width, &x0, &x1, &xw); - linear_texcoord(sampler->state->wrap_t, t[j], height, &y0, &y1, &yw); - linear_texcoord(sampler->state->wrap_r, p[j], depth, &z0, &z1, &zw); - get_texel(sampler, face, level0, x0, y0, z0, texel0, 0); - get_texel(sampler, face, level0, x1, y0, z0, texel0, 1); - get_texel(sampler, face, level0, x0, y1, z0, texel0, 2); - get_texel(sampler, face, level0, x1, y1, z0, texel0, 3); - get_texel(sampler, face, level0, x0, y0, z1, texel1, 0); - get_texel(sampler, face, level0, x1, y0, z1, texel1, 1); - get_texel(sampler, face, level0, x0, y1, z1, texel1, 2); - get_texel(sampler, face, level0, x1, y1, z1, texel1, 3); - - /* 3D lerp */ - for (c = 0; c < 4; c++) { - float ctemp0[4][4], ctemp1[4][4]; - ctemp0[c][j] = lerp_2d(xw, yw, - texel0[c][0], texel0[c][1], - texel0[c][2], texel0[c][3]); - ctemp1[c][j] = lerp_2d(xw, yw, - texel1[c][0], texel1[c][1], - texel1[c][2], texel1[c][3]); - rgba[c][j] = LERP(zw, ctemp0[c][j], ctemp1[c][j]); - } - - if (level0 != level1) { - /* get texels from second mipmap level and blend */ - float rgba2[4][4]; - x0 /= 2; - y0 /= 2; - z0 /= 2; - x1 /= 2; - y1 /= 2; - z1 /= 2; - get_texel(sampler, face, level1, x0, y0, z0, texel0, 0); - get_texel(sampler, face, level1, x1, y0, z0, texel0, 1); - get_texel(sampler, face, level1, x0, y1, z0, texel0, 2); - get_texel(sampler, face, level1, x1, y1, z0, texel0, 3); - get_texel(sampler, face, level1, x0, y0, z1, texel1, 0); - get_texel(sampler, face, level1, x1, y0, z1, texel1, 1); - get_texel(sampler, face, level1, x0, y1, z1, texel1, 2); - get_texel(sampler, face, level1, x1, y1, z1, texel1, 3); - - /* 3D lerp */ - for (c = 0; c < 4; c++) { - float ctemp0[4][4], ctemp1[4][4]; - ctemp0[c][j] = lerp_2d(xw, yw, - texel0[c][0], texel0[c][1], - texel0[c][2], texel0[c][3]); - ctemp1[c][j] = lerp_2d(xw, yw, - texel1[c][0], texel1[c][1], - texel1[c][2], texel1[c][3]); - rgba2[c][j] = LERP(zw, ctemp0[c][j], ctemp1[c][j]); - } - - /* blend mipmap levels */ - for (c = 0; c < NUM_CHANNELS; c++) { - rgba[c][j] = LERP(levelBlend, rgba[c][j], rgba2[c][j]); - } - } - } - break; - default: - assert(0); - } -} - - -static void -sp_get_samples_cube(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]) -{ - unsigned faces[QUAD_SIZE], j; - float ssss[4], tttt[4]; - for (j = 0; j < QUAD_SIZE; j++) { - faces[j] = choose_cube_face(s[j], t[j], p[j], ssss + j, tttt + j); - } - sp_get_samples_2d_common(sampler, ssss, tttt, NULL, lodbias, rgba, faces); -} - - -/** - * Called via tgsi_sampler::get_samples() - * Use the sampler's state setting to get a filtered RGBA value - * from the sampler's texture. - * - * XXX we can implement many versions of this function, each - * tightly coded for a specific combination of sampler state - * (nearest + repeat), (bilinear mipmap + clamp), etc. - * - * The update_samplers() function in st_atom_sampler.c could create - * a new tgsi_sampler object for each state combo it finds.... - */ -void -sp_get_samples(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]) -{ - if (!sampler->texture) - return; - - switch (sampler->texture->target) { - case PIPE_TEXTURE_1D: - sp_get_samples_1d(sampler, s, t, p, lodbias, rgba); - break; - case PIPE_TEXTURE_2D: - sp_get_samples_2d(sampler, s, t, p, lodbias, rgba); - break; - case PIPE_TEXTURE_3D: - sp_get_samples_3d(sampler, s, t, p, lodbias, rgba); - break; - case PIPE_TEXTURE_CUBE: - sp_get_samples_cube(sampler, s, t, p, lodbias, rgba); - break; - default: - assert(0); - } -} - diff --git a/src/mesa/pipe/softpipe/sp_tex_sample.h b/src/mesa/pipe/softpipe/sp_tex_sample.h deleted file mode 100644 index 404bfd0c36..0000000000 --- a/src/mesa/pipe/softpipe/sp_tex_sample.h +++ /dev/null @@ -1,17 +0,0 @@ -#ifndef SP_TEX_SAMPLE_H -#define SP_TEX_SAMPLE_H - - -struct tgsi_sampler; - - -extern void -sp_get_samples(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]); - - -#endif /* SP_TEX_SAMPLE_H */ diff --git a/src/mesa/pipe/softpipe/sp_texture.c b/src/mesa/pipe/softpipe/sp_texture.c deleted file mode 100644 index 6de7a9b543..0000000000 --- a/src/mesa/pipe/softpipe/sp_texture.c +++ /dev/null @@ -1,166 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Michel Dänzer - */ - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "pipe/p_winsys.h" - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_texture.h" - - -/* Simple, maximally packed layout. - */ - -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - - -static void -softpipe_texture_layout(struct softpipe_texture * spt) -{ - struct pipe_texture *pt = &spt->base; - unsigned level; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - - spt->buffer_size = 0; - - for (level = 0; level <= pt->last_level; level++) { - pt->width[level] = width; - pt->height[level] = height; - pt->depth[level] = depth; - - spt->level_offset[level] = spt->buffer_size; - - spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * - ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * - width * pt->cpp; - - width = minify(width); - height = minify(height); - depth = minify(depth); - } -} - - -struct pipe_texture * -softpipe_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat) -{ - struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); - if (!spt) - return NULL; - - spt->base = *templat; - - softpipe_texture_layout(spt); - - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); - if (!spt->buffer) { - FREE(spt); - return NULL; - } - - return &spt->base; -} - - -void -softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) -{ - if (!*pt) - return; - - /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); - */ - if (--(*pt)->refcount <= 0) { - struct softpipe_texture *spt = softpipe_texture(*pt); - - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); - */ - - pipe_buffer_reference(pipe->winsys, &spt->buffer, NULL); - - FREE(spt); - } - *pt = NULL; -} - - -/** - * Called via pipe->get_tex_surface() - */ -struct pipe_surface * -softpipe_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - struct softpipe_texture *spt = softpipe_texture(pt); - struct pipe_surface *ps; - - assert(level <= pt->last_level); - - ps = pipe->winsys->surface_alloc(pipe->winsys); - if (ps) { - assert(ps->refcount); - assert(ps->winsys); - pipe_buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); - ps->format = pt->format; - ps->cpp = pt->cpp; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->pitch = ps->width; - ps->offset = spt->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { - ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - (pt->compressed ? ps->height/4 : ps->height) * - ps->width * ps->cpp; - } else { - assert(face == 0); - assert(zslice == 0); - } - } - return ps; -} diff --git a/src/mesa/pipe/softpipe/sp_texture.h b/src/mesa/pipe/softpipe/sp_texture.h deleted file mode 100644 index fa646c0de9..0000000000 --- a/src/mesa/pipe/softpipe/sp_texture.h +++ /dev/null @@ -1,71 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SP_TEXTURE_H -#define SP_TEXTURE_H - - -struct pipe_context; -struct pipe_texture; - - -struct softpipe_texture -{ - struct pipe_texture base; - - unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; - unsigned long buffer_size; -}; - - -/** cast wrapper */ -static INLINE struct softpipe_texture * -softpipe_texture(struct pipe_texture *pt) -{ - return (struct softpipe_texture *) pt; -} - - - -extern struct pipe_texture * -softpipe_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat); - -extern void -softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); - -extern struct pipe_surface * -softpipe_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice); - - -#endif /* SP_TEXTURE */ diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.c b/src/mesa/pipe/softpipe/sp_tile_cache.c deleted file mode 100644 index 1597361b82..0000000000 --- a/src/mesa/pipe/softpipe/sp_tile_cache.c +++ /dev/null @@ -1,585 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Framebuffer/surface tile caching. - * - * Author: - * Brian Paul - */ - -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/util/p_tile.h" -#include "sp_context.h" -#include "sp_surface.h" -#include "sp_tile_cache.h" - -#define NUM_ENTRIES 30 - - -/** XXX move these */ -#define MAX_WIDTH 2048 -#define MAX_HEIGHT 2048 - - -struct softpipe_tile_cache -{ - struct pipe_surface *surface; /**< the surface we're caching */ - void *surface_map; - struct pipe_texture *texture; /**< if caching a texture */ - struct softpipe_cached_tile entries[NUM_ENTRIES]; - uint clear_flags[(MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32]; - float clear_color[4]; - uint clear_val; - boolean depth_stencil; /** Is the surface a depth/stencil format? */ - - struct pipe_surface *tex_surf; - void *tex_surf_map; - int tex_face, tex_level, tex_z; - - struct softpipe_cached_tile tile; /**< scratch tile for clears */ -}; - - -/** - * Return the position in the cache for the tile that contains win pos (x,y). - * We currently use a direct mapped cache so this is like a hack key. - * At some point we should investige something more sophisticated, like - * a LRU replacement policy. - */ -#define CACHE_POS(x, y) \ - (((x) / TILE_SIZE + ((y) / TILE_SIZE) * 5) % NUM_ENTRIES) - - - -/** - * Is the tile at (x,y) in cleared state? - */ -static INLINE uint -is_clear_flag_set(const uint *bitvec, int x, int y) -{ - int pos, bit; - x /= TILE_SIZE; - y /= TILE_SIZE; - pos = y * (MAX_WIDTH / TILE_SIZE) + x; - assert(pos / 32 < (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32); - bit = bitvec[pos / 32] & (1 << (pos & 31)); - return bit; -} - - -/** - * Mark the tile at (x,y) as not cleared. - */ -static INLINE void -clear_clear_flag(uint *bitvec, int x, int y) -{ - int pos; - x /= TILE_SIZE; - y /= TILE_SIZE; - pos = y * (MAX_WIDTH / TILE_SIZE) + x; - assert(pos / 32 < (MAX_WIDTH / TILE_SIZE) * (MAX_HEIGHT / TILE_SIZE) / 32); - bitvec[pos / 32] &= ~(1 << (pos & 31)); -} - - -struct softpipe_tile_cache * -sp_create_tile_cache(void) -{ - struct softpipe_tile_cache *tc; - uint pos; - - tc = CALLOC_STRUCT( softpipe_tile_cache ); - if (tc) { - for (pos = 0; pos < NUM_ENTRIES; pos++) { - tc->entries[pos].x = - tc->entries[pos].y = -1; - } - } - return tc; -} - - -void -sp_destroy_tile_cache(struct softpipe_tile_cache *tc) -{ - uint pos; - - for (pos = 0; pos < NUM_ENTRIES; pos++) { - //assert(tc->entries[pos].x < 0); - } - if (tc->surface) { - pipe_surface_reference(&tc->surface, NULL); - } - if (tc->tex_surf) { - pipe_surface_reference(&tc->tex_surf, NULL); - } - - FREE( tc ); -} - - -/** - * Specify the surface to cache. - */ -void -sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, - struct pipe_surface *ps) -{ - assert(!tc->texture); - - if (tc->surface_map) { - /*assert(tc->surface != ps);*/ - pipe_surface_unmap(tc->surface); - } - - pipe_surface_reference(&tc->surface, ps); - - if (ps) { - if (tc->surface_map) - tc->surface_map = pipe_surface_map(ps); - - tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || - ps->format == PIPE_FORMAT_Z16_UNORM || - ps->format == PIPE_FORMAT_Z32_UNORM || - ps->format == PIPE_FORMAT_U_S8); - } -} - - -/** - * Return the surface being cached. - */ -struct pipe_surface * -sp_tile_cache_get_surface(struct softpipe_tile_cache *tc) -{ - return tc->surface; -} - - -void -sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc) -{ - if (tc->surface && !tc->surface_map) - tc->surface_map = pipe_surface_map(tc->surface); - - if (tc->tex_surf && !tc->tex_surf_map) - tc->tex_surf_map = pipe_surface_map(tc->tex_surf); -} - - -void -sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc) -{ - if (tc->surface_map) { - pipe_surface_unmap(tc->surface); - tc->surface_map = NULL; - } - - if (tc->tex_surf_map) { - pipe_surface_unmap(tc->tex_surf); - tc->tex_surf_map = NULL; - } -} - - -/** - * Specify the texture to cache. - */ -void -sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, - struct pipe_texture *texture) -{ - uint i; - - assert(!tc->surface); - - tc->texture = texture; - - if (tc->tex_surf_map) { - pipe_surface_unmap(tc->tex_surf); - tc->tex_surf_map = NULL; - } - pipe_surface_reference(&tc->tex_surf, NULL); - - /* mark as entries as invalid/empty */ - /* XXX we should try to avoid this when the teximage hasn't changed */ - for (i = 0; i < NUM_ENTRIES; i++) { - tc->entries[i].x = -1; - } - - tc->tex_face = -1; /* any invalid value here */ -} - - -/** - * Set pixels in a tile to the given clear color/value, float. - */ -static void -clear_tile_rgba(struct softpipe_cached_tile *tile, - enum pipe_format format, - const float clear_value[4]) -{ - if (clear_value[0] == 0.0 && - clear_value[1] == 0.0 && - clear_value[2] == 0.0 && - clear_value[3] == 0.0) { - memset(tile->data.color, 0, sizeof(tile->data.color)); - } - else { - uint i, j; - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.color[i][j][0] = clear_value[0]; - tile->data.color[i][j][1] = clear_value[1]; - tile->data.color[i][j][2] = clear_value[2]; - tile->data.color[i][j][3] = clear_value[3]; - } - } - } -} - - -/** - * Set a tile to a solid value/color. - */ -static void -clear_tile(struct softpipe_cached_tile *tile, - enum pipe_format format, - uint clear_value) -{ - uint i, j; - - switch (pf_get_size(format)) { - case 1: - memset(tile->data.any, 0, TILE_SIZE * TILE_SIZE); - break; - case 2: - if (clear_value == 0) { - memset(tile->data.any, 0, 2 * TILE_SIZE * TILE_SIZE); - } - else { - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.depth16[i][j] = (ushort) clear_value; - } - } - } - break; - case 4: - if (clear_value == 0) { - memset(tile->data.any, 0, 4 * TILE_SIZE * TILE_SIZE); - } - else { - for (i = 0; i < TILE_SIZE; i++) { - for (j = 0; j < TILE_SIZE; j++) { - tile->data.color32[i][j] = clear_value; - } - } - } - break; - default: - assert(0); - } -} - - -/** - * Actually clear the tiles which were flagged as being in a clear state. - */ -static void -sp_tile_cache_flush_clear(struct pipe_context *pipe, - struct softpipe_tile_cache *tc) -{ - struct pipe_surface *ps = tc->surface; - const uint w = tc->surface->width; - const uint h = tc->surface->height; - uint x, y; - uint numCleared = 0; - - /* clear the scratch tile to the clear value */ - clear_tile(&tc->tile, ps->format, tc->clear_val); - - /* push the tile to all positions marked as clear */ - for (y = 0; y < h; y += TILE_SIZE) { - for (x = 0; x < w; x += TILE_SIZE) { - if (is_clear_flag_set(tc->clear_flags, x, y)) { - pipe_put_tile_raw(pipe, ps, - x, y, TILE_SIZE, TILE_SIZE, - tc->tile.data.color32, 0/*STRIDE*/); - - /* do this? */ - clear_clear_flag(tc->clear_flags, x, y); - - numCleared++; - } - } - } -#if 0 - debug_printf("num cleared: %u\n", numCleared); -#endif -} - - -/** - * Flush the tile cache: write all dirty tiles back to the surface. - * any tiles "flagged" as cleared will be "really" cleared. - */ -void -sp_flush_tile_cache(struct softpipe_context *softpipe, - struct softpipe_tile_cache *tc) -{ - struct pipe_context *pipe = &softpipe->pipe; - struct pipe_surface *ps = tc->surface; - int inuse = 0, pos; - - if (!ps || !ps->buffer) - return; - - for (pos = 0; pos < NUM_ENTRIES; pos++) { - struct softpipe_cached_tile *tile = tc->entries + pos; - if (tile->x >= 0) { - if (tc->depth_stencil) { - pipe_put_tile_raw(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); - } - else { - pipe_put_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - } - tile->x = tile->y = -1; /* mark as empty */ - inuse++; - } - } - -#if TILE_CLEAR_OPTIMIZATION - sp_tile_cache_flush_clear(&softpipe->pipe, tc); -#endif - -#if 0 - debug_printf("flushed tiles in use: %d\n", inuse); -#endif -} - - -/** - * Get a tile from the cache. - * \param x, y position of tile, in pixels - */ -struct softpipe_cached_tile * -sp_get_cached_tile(struct softpipe_context *softpipe, - struct softpipe_tile_cache *tc, int x, int y) -{ - struct pipe_context *pipe = &softpipe->pipe; - struct pipe_surface *ps = tc->surface; - - /* tile pos in framebuffer: */ - const int tile_x = x & ~(TILE_SIZE - 1); - const int tile_y = y & ~(TILE_SIZE - 1); - - /* cache pos/entry: */ - const int pos = CACHE_POS(x, y); - struct softpipe_cached_tile *tile = tc->entries + pos; - - if (tile_x != tile->x || - tile_y != tile->y) { - - if (tile->x != -1) { - /* put dirty tile back in framebuffer */ - if (tc->depth_stencil) { - pipe_put_tile_raw(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); - } - else { - pipe_put_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - } - } - - tile->x = tile_x; - tile->y = tile_y; - - if (is_clear_flag_set(tc->clear_flags, x, y)) { - /* don't get tile from framebuffer, just clear it */ - if (tc->depth_stencil) { - clear_tile(tile, ps->format, tc->clear_val); - } - else { - clear_tile_rgba(tile, ps->format, tc->clear_color); - } - clear_clear_flag(tc->clear_flags, x, y); - } - else { - /* get new tile data from surface */ - if (tc->depth_stencil) { - pipe_get_tile_raw(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); - } - else { - pipe_get_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - } - } - } - - return tile; -} - - -/** - * Given the texture face, level, zslice, x and y values, compute - * the cache entry position/index where we'd hope to find the - * cached texture tile. - * This is basically a direct-map cache. - * XXX There's probably lots of ways in which we can improve this. - */ -static INLINE uint -tex_cache_pos(int x, int y, int z, int face, int level) -{ - uint entry = x + y * 5 + z * 4 + face + level; - return entry % NUM_ENTRIES; -} - - -/** - * Similar to sp_get_cached_tile() but for textures. - * Tiles are read-only and indexed with more params. - */ -const struct softpipe_cached_tile * -sp_get_cached_tile_tex(struct pipe_context *pipe, - struct softpipe_tile_cache *tc, int x, int y, int z, - int face, int level) -{ - /* tile pos in framebuffer: */ - const int tile_x = x & ~(TILE_SIZE - 1); - const int tile_y = y & ~(TILE_SIZE - 1); - /* cache pos/entry: */ - const uint pos = tex_cache_pos(x / TILE_SIZE, y / TILE_SIZE, z, - face, level); - struct softpipe_cached_tile *tile = tc->entries + pos; - - if (tile_x != tile->x || - tile_y != tile->y || - z != tile->z || - face != tile->face || - level != tile->level) { - /* cache miss */ - - /* check if we need to get a new surface */ - if (!tc->tex_surf || - tc->tex_face != face || - tc->tex_level != level || - tc->tex_z != z) { - /* get new surface (view into texture) */ - - if (tc->tex_surf_map) - pipe_surface_unmap(tc->tex_surf); - - tc->tex_surf = pipe->get_tex_surface(pipe, tc->texture, face, level, z); - tc->tex_surf_map = pipe_surface_map(tc->tex_surf); - - tc->tex_face = face; - tc->tex_level = level; - tc->tex_z = z; - } - - /* get tile from the surface (view into texture) */ - pipe_get_tile_rgba(pipe, tc->tex_surf, - tile_x, tile_y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); - tile->x = tile_x; - tile->y = tile_y; - tile->z = z; - tile->face = face; - tile->level = level; - } - - return tile; -} - - -/** - * When a whole surface is being cleared to a value we can avoid - * fetching tiles above. - * Save the color and set a 'clearflag' for each tile of the screen. - */ -void -sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue) -{ - uint r, g, b, a; - uint pos; - - tc->clear_val = clearValue; - - switch (tc->surface->format) { - case PIPE_FORMAT_R8G8B8A8_UNORM: - r = (clearValue >> 24) & 0xff; - g = (clearValue >> 16) & 0xff; - b = (clearValue >> 8) & 0xff; - a = (clearValue ) & 0xff; - break; - case PIPE_FORMAT_A8R8G8B8_UNORM: - r = (clearValue >> 16) & 0xff; - g = (clearValue >> 8) & 0xff; - b = (clearValue ) & 0xff; - a = (clearValue >> 24) & 0xff; - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - r = (clearValue >> 8) & 0xff; - g = (clearValue >> 16) & 0xff; - b = (clearValue >> 24) & 0xff; - a = (clearValue ) & 0xff; - break; - default: - r = g = b = a = 0; - } - - tc->clear_color[0] = r / 255.0f; - tc->clear_color[1] = g / 255.0f; - tc->clear_color[2] = b / 255.0f; - tc->clear_color[3] = a / 255.0f; - -#if TILE_CLEAR_OPTIMIZATION - /* set flags to indicate all the tiles are cleared */ - memset(tc->clear_flags, 255, sizeof(tc->clear_flags)); -#else - /* disable the optimization */ - memset(tc->clear_flags, 0, sizeof(tc->clear_flags)); -#endif - - for (pos = 0; pos < NUM_ENTRIES; pos++) { - struct softpipe_cached_tile *tile = tc->entries + pos; - tile->x = tile->y = -1; - } -} diff --git a/src/mesa/pipe/softpipe/sp_tile_cache.h b/src/mesa/pipe/softpipe/sp_tile_cache.h deleted file mode 100644 index 7fd1081286..0000000000 --- a/src/mesa/pipe/softpipe/sp_tile_cache.h +++ /dev/null @@ -1,104 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef SP_TILE_CACHE_H -#define SP_TILE_CACHE_H - -#define TILE_CLEAR_OPTIMIZATION 1 - - -#include "pipe/p_compiler.h" - - -struct softpipe_context; -struct softpipe_tile_cache; - - -/** - * Cache tile size (width and height). This needs to be a power of two. - */ -#define TILE_SIZE 64 - - - -struct softpipe_cached_tile -{ - int x, y; /**< pos of tile in window coords */ - int z, face, level; /**< Extra texture indexes */ - union { - float color[TILE_SIZE][TILE_SIZE][4]; - uint color32[TILE_SIZE][TILE_SIZE]; - uint depth32[TILE_SIZE][TILE_SIZE]; - ushort depth16[TILE_SIZE][TILE_SIZE]; - ubyte stencil8[TILE_SIZE][TILE_SIZE]; - ubyte any[1]; - } data; -}; - - -extern struct softpipe_tile_cache * -sp_create_tile_cache(void); - -extern void -sp_destroy_tile_cache(struct softpipe_tile_cache *tc); - -extern void -sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, - struct pipe_surface *sps); - -extern struct pipe_surface * -sp_tile_cache_get_surface(struct softpipe_tile_cache *tc); - -extern void -sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc); - -extern void -sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc); - -extern void -sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, - struct pipe_texture *texture); - -extern void -sp_flush_tile_cache(struct softpipe_context *softpipe, - struct softpipe_tile_cache *tc); - -extern void -sp_tile_cache_clear(struct softpipe_tile_cache *tc, uint clearValue); - -extern struct softpipe_cached_tile * -sp_get_cached_tile(struct softpipe_context *softpipe, - struct softpipe_tile_cache *tc, int x, int y); - -extern const struct softpipe_cached_tile * -sp_get_cached_tile_tex(struct pipe_context *pipe, - struct softpipe_tile_cache *tc, int x, int y, int z, - int face, int level); - - -#endif /* SP_TILE_CACHE_H */ - diff --git a/src/mesa/pipe/softpipe/sp_winsys.h b/src/mesa/pipe/softpipe/sp_winsys.h deleted file mode 100644 index d6b379f58c..0000000000 --- a/src/mesa/pipe/softpipe/sp_winsys.h +++ /dev/null @@ -1,57 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* This is the interface that softpipe requires any window system - * hosting it to implement. This is the only include file in softpipe - * which is public. - */ - - -#ifndef SP_WINSYS_H -#define SP_WINSYS_H - - -#include "pipe/p_compiler.h" /* for boolean */ - -enum pipe_format; - -struct softpipe_winsys { - /** test if the given format is supported for front/back color bufs */ - boolean (*is_format_supported)( struct softpipe_winsys *sws, - enum pipe_format format ); - -}; - -struct pipe_winsys; -struct pipe_context; - - -struct pipe_context *softpipe_create( struct pipe_winsys *, - struct softpipe_winsys * ); - - -#endif /* SP_WINSYS_H */ diff --git a/src/mesa/pipe/tgsi/Makefile b/src/mesa/pipe/tgsi/Makefile deleted file mode 100644 index 12a8bd0409..0000000000 --- a/src/mesa/pipe/tgsi/Makefile +++ /dev/null @@ -1,3 +0,0 @@ -default: - cd ../.. ; make - diff --git a/src/mesa/pipe/tgsi/exec/Makefile b/src/mesa/pipe/tgsi/exec/Makefile deleted file mode 100644 index eb8b14e0e8..0000000000 --- a/src/mesa/pipe/tgsi/exec/Makefile +++ /dev/null @@ -1,3 +0,0 @@ -default: - cd ../../.. ; make - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c deleted file mode 100644 index 37e6007068..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ /dev/null @@ -1,2485 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * TGSI interpretor/executor. - * - * Flow control information: - * - * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) - * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special - * care since a condition may be true for some quad components but false - * for other components. - * - * We basically execute all statements (even if they're in the part of - * an IF/ELSE clause that's "not taken") and use a special mask to - * control writing to destination registers. This is the ExecMask. - * See store_dest(). - * - * The ExecMask is computed from three other masks (CondMask, LoopMask and - * ContMask) which are controlled by the flow control instructions (namely: - * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). - * - * - * Authors: - * Michal Krol - * Brian Paul - */ - -#include "pipe/p_compiler.h" -#include "pipe/p_state.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/util/tgsi_util.h" -#include "tgsi_exec.h" - -#define TILE_TOP_LEFT 0 -#define TILE_TOP_RIGHT 1 -#define TILE_BOTTOM_LEFT 2 -#define TILE_BOTTOM_RIGHT 3 - -/* - * Shorthand locations of various utility registers (_I = Index, _C = Channel) - */ -#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I -#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C -#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I -#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C -#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I -#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C -#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I -#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C -#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I -#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C -#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I -#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C -#define TEMP_128_I TGSI_EXEC_TEMP_128_I -#define TEMP_128_C TGSI_EXEC_TEMP_128_C -#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I -#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C -#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I -#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C -#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I -#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C -#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I -#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C -#define TEMP_R0 TGSI_EXEC_TEMP_R0 - -#define FOR_EACH_CHANNEL(CHAN)\ - for (CHAN = 0; CHAN < 4; CHAN++) - -#define IS_CHANNEL_ENABLED(INST, CHAN)\ - ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) - -#define IS_CHANNEL_ENABLED2(INST, CHAN)\ - ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) - -#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED( INST, CHAN )) - -#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED2( INST, CHAN )) - - -/** The execution mask depends on the conditional mask and the loop mask */ -#define UPDATE_EXEC_MASK(MACH) \ - MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask - - -#define CHAN_X 0 -#define CHAN_Y 1 -#define CHAN_Z 2 -#define CHAN_W 3 - - - -static void -tgsi_exec_prepare( struct tgsi_exec_machine *mach ) -{ - struct tgsi_exec_labels *labels = &mach->Labels; - struct tgsi_parse_context parse; - struct tgsi_full_instruction *instructions; - struct tgsi_full_declaration *declarations; - uint maxInstructions = 10, numInstructions = 0; - uint maxDeclarations = 10, numDeclarations = 0; - uint k; - uint instno = 0; - - mach->ImmLimit = 0; - labels->count = 0; - - declarations = (struct tgsi_full_declaration *) - MALLOC( maxDeclarations * sizeof(struct tgsi_full_declaration) ); - - instructions = (struct tgsi_full_instruction *) - MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) ); - - k = tgsi_parse_init( &parse, mach->Tokens ); - if (k != TGSI_PARSE_OK) { - debug_printf("Problem parsing!\n"); - return; - } - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - uint pointer = parse.Position; - uint i; - - tgsi_parse_token( &parse ); - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - /* save expanded declaration */ - if (numDeclarations == maxDeclarations) { - declarations = REALLOC(declarations, - maxDeclarations - * sizeof(struct tgsi_full_declaration), - (maxDeclarations + 10) - * sizeof(struct tgsi_full_declaration)); - maxDeclarations += 10; - } - memcpy(declarations + numDeclarations, - &parse.FullToken.FullDeclaration, - sizeof(declarations[0])); - numDeclarations++; - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; - assert( size % 4 == 0 ); - assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES ); - - for( i = 0; i < size; i++ ) { - mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; - } - mach->ImmLimit += size / 4; - } - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - assert( labels->count < 128 ); - - labels->labels[labels->count][0] = instno; - labels->labels[labels->count][1] = pointer; - labels->count++; - - /* save expanded instruction */ - if (numInstructions == maxInstructions) { - instructions = REALLOC(instructions, - maxInstructions - * sizeof(struct tgsi_full_instruction), - (maxInstructions + 10) - * sizeof(struct tgsi_full_instruction)); - maxInstructions += 10; - } - memcpy(instructions + numInstructions, - &parse.FullToken.FullInstruction, - sizeof(instructions[0])); - numInstructions++; - break; - - default: - assert( 0 ); - } - } - tgsi_parse_free (&parse); - - if (mach->Declarations) { - FREE( mach->Declarations ); - } - mach->Declarations = declarations; - mach->NumDeclarations = numDeclarations; - - if (mach->Instructions) { - FREE( mach->Instructions ); - } - mach->Instructions = instructions; - mach->NumInstructions = numInstructions; -} - - -/** - * Initialize machine state by expanding tokens to full instructions, - * allocating temporary storage, setting up constants, etc. - * After this, we can call tgsi_exec_machine_run() many times. - */ -void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach, - const struct tgsi_token *tokens, - uint numSamplers, - struct tgsi_sampler *samplers) -{ - uint i, k; - struct tgsi_parse_context parse; - -#if 0 - tgsi_dump(tokens, 0); -#endif - - mach->Tokens = tokens; - - mach->Samplers = samplers; - - k = tgsi_parse_init (&parse, mach->Tokens); - if (k != TGSI_PARSE_OK) { - debug_printf( "Problem parsing!\n" ); - return; - } - - mach->Processor = parse.FullHeader.Processor.Processor; - tgsi_parse_free (&parse); - - mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); - mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; - - /* Setup constants. */ - for( i = 0; i < 4; i++ ) { - mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; - mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; - mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; - mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; - mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; - mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; - mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; - mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; - } - - tgsi_exec_prepare( mach ); -} - - -void -tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach) -{ - if (mach->Instructions) { - FREE(mach->Instructions); - mach->Instructions = NULL; - mach->NumInstructions = 0; - } - if (mach->Declarations) { - FREE(mach->Declarations); - mach->Declarations = NULL; - mach->NumDeclarations = 0; - } -} - - -static void -micro_abs( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) fabs( (double) src->f[0] ); - dst->f[1] = (float) fabs( (double) src->f[1] ); - dst->f[2] = (float) fabs( (double) src->f[2] ); - dst->f[3] = (float) fabs( (double) src->f[3] ); -} - -static void -micro_add( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] + src1->f[0]; - dst->f[1] = src0->f[1] + src1->f[1]; - dst->f[2] = src0->f[2] + src1->f[2]; - dst->f[3] = src0->f[3] + src1->f[3]; -} - -static void -micro_iadd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] + src1->i[0]; - dst->i[1] = src0->i[1] + src1->i[1]; - dst->i[2] = src0->i[2] + src1->i[2]; - dst->i[3] = src0->i[3] + src1->i[3]; -} - -static void -micro_and( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] & src1->u[0]; - dst->u[1] = src0->u[1] & src1->u[1]; - dst->u[2] = src0->u[2] & src1->u[2]; - dst->u[3] = src0->u[3] & src1->u[3]; -} - -static void -micro_ceil( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) ceil( (double) src->f[0] ); - dst->f[1] = (float) ceil( (double) src->f[1] ); - dst->f[2] = (float) ceil( (double) src->f[2] ); - dst->f[3] = (float) ceil( (double) src->f[3] ); -} - -static void -micro_cos( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) cos( (double) src->f[0] ); - dst->f[1] = (float) cos( (double) src->f[1] ); - dst->f[2] = (float) cos( (double) src->f[2] ); - dst->f[3] = (float) cos( (double) src->f[3] ); -} - -static void -micro_ddx( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_ddy( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_div( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] / src1->f[0]; - dst->f[1] = src0->f[1] / src1->f[1]; - dst->f[2] = src0->f[2] / src1->f[2]; - dst->f[3] = src0->f[3] / src1->f[3]; -} - -static void -micro_udiv( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] / src1->u[0]; - dst->u[1] = src0->u[1] / src1->u[1]; - dst->u[2] = src0->u[2] / src1->u[2]; - dst->u[3] = src0->u[3] / src1->u[3]; -} - -static void -micro_eq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_ieq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; -} - -static void -micro_exp2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src) -{ - dst->f[0] = (float) pow( 2.0, (double) src->f[0] ); - dst->f[1] = (float) pow( 2.0, (double) src->f[1] ); - dst->f[2] = (float) pow( 2.0, (double) src->f[2] ); - dst->f[3] = (float) pow( 2.0, (double) src->f[3] ); -} - -static void -micro_f2it( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->i[0] = (int) src->f[0]; - dst->i[1] = (int) src->f[1]; - dst->i[2] = (int) src->f[2]; - dst->i[3] = (int) src->f[3]; -} - -static void -micro_f2ut( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = (uint) src->f[0]; - dst->u[1] = (uint) src->f[1]; - dst->u[2] = (uint) src->f[2]; - dst->u[3] = (uint) src->f[3]; -} - -static void -micro_flr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) floor( (double) src->f[0] ); - dst->f[1] = (float) floor( (double) src->f[1] ); - dst->f[2] = (float) floor( (double) src->f[2] ); - dst->f[3] = (float) floor( (double) src->f[3] ); -} - -static void -micro_frc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = src->f[0] - (float) floor( (double) src->f[0] ); - dst->f[1] = src->f[1] - (float) floor( (double) src->f[1] ); - dst->f[2] = src->f[2] - (float) floor( (double) src->f[2] ); - dst->f[3] = src->f[3] - (float) floor( (double) src->f[3] ); -} - -static void -micro_ge( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_i2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) src->i[0]; - dst->f[1] = (float) src->i[1]; - dst->f[2] = (float) src->i[2]; - dst->f[3] = (float) src->i[3]; -} - -static void -micro_lg2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) log( (double) src->f[0] ) * 1.442695f; - dst->f[1] = (float) log( (double) src->f[1] ) * 1.442695f; - dst->f[2] = (float) log( (double) src->f[2] ) * 1.442695f; - dst->f[3] = (float) log( (double) src->f[3] ) * 1.442695f; -} - -static void -micro_lt( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_ilt( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; -} - -static void -micro_ult( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; -} - -static void -micro_max( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; -} - -static void -micro_imax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; -} - -static void -micro_umax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; -} - -static void -micro_min( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; -} - -static void -micro_imin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; -} - -static void -micro_umin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; -} - -static void -micro_umod( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] % src1->u[0]; - dst->u[1] = src0->u[1] % src1->u[1]; - dst->u[2] = src0->u[2] % src1->u[2]; - dst->u[3] = src0->u[3] % src1->u[3]; -} - -static void -micro_mul( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] * src1->f[0]; - dst->f[1] = src0->f[1] * src1->f[1]; - dst->f[2] = src0->f[2] * src1->f[2]; - dst->f[3] = src0->f[3] * src1->f[3]; -} - -static void -micro_imul( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] * src1->i[0]; - dst->i[1] = src0->i[1] * src1->i[1]; - dst->i[2] = src0->i[2] * src1->i[2]; - dst->i[3] = src0->i[3] * src1->i[3]; -} - -static void -micro_imul64( - union tgsi_exec_channel *dst0, - union tgsi_exec_channel *dst1, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst1->i[0] = src0->i[0] * src1->i[0]; - dst1->i[1] = src0->i[1] * src1->i[1]; - dst1->i[2] = src0->i[2] * src1->i[2]; - dst1->i[3] = src0->i[3] * src1->i[3]; - dst0->i[0] = 0; - dst0->i[1] = 0; - dst0->i[2] = 0; - dst0->i[3] = 0; -} - -static void -micro_umul64( - union tgsi_exec_channel *dst0, - union tgsi_exec_channel *dst1, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst1->u[0] = src0->u[0] * src1->u[0]; - dst1->u[1] = src0->u[1] * src1->u[1]; - dst1->u[2] = src0->u[2] * src1->u[2]; - dst1->u[3] = src0->u[3] * src1->u[3]; - dst0->u[0] = 0; - dst0->u[1] = 0; - dst0->u[2] = 0; - dst0->u[3] = 0; -} - -static void -micro_movc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2 ) -{ - dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; - dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; - dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; - dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; -} - -static void -micro_neg( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = -src->f[0]; - dst->f[1] = -src->f[1]; - dst->f[2] = -src->f[2]; - dst->f[3] = -src->f[3]; -} - -static void -micro_ineg( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->i[0] = -src->i[0]; - dst->i[1] = -src->i[1]; - dst->i[2] = -src->i[2]; - dst->i[3] = -src->i[3]; -} - -static void -micro_not( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = ~src->u[0]; - dst->u[1] = ~src->u[1]; - dst->u[2] = ~src->u[2]; - dst->u[3] = ~src->u[3]; -} - -static void -micro_or( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] | src1->u[0]; - dst->u[1] = src0->u[1] | src1->u[1]; - dst->u[2] = src0->u[2] | src1->u[2]; - dst->u[3] = src0->u[3] | src1->u[3]; -} - -static void -micro_pow( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = (float) pow( (double) src0->f[0], (double) src1->f[0] ); - dst->f[1] = (float) pow( (double) src0->f[1], (double) src1->f[1] ); - dst->f[2] = (float) pow( (double) src0->f[2], (double) src1->f[2] ); - dst->f[3] = (float) pow( (double) src0->f[3], (double) src1->f[3] ); -} - -static void -micro_rnd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) floor( (double) (src->f[0] + 0.5f) ); - dst->f[1] = (float) floor( (double) (src->f[1] + 0.5f) ); - dst->f[2] = (float) floor( (double) (src->f[2] + 0.5f) ); - dst->f[3] = (float) floor( (double) (src->f[3] + 0.5f) ); -} - -static void -micro_shl( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] << src1->i[0]; - dst->i[1] = src0->i[1] << src1->i[1]; - dst->i[2] = src0->i[2] << src1->i[2]; - dst->i[3] = src0->i[3] << src1->i[3]; -} - -static void -micro_ishr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] >> src1->i[0]; - dst->i[1] = src0->i[1] >> src1->i[1]; - dst->i[2] = src0->i[2] >> src1->i[2]; - dst->i[3] = src0->i[3] >> src1->i[3]; -} - -static void -micro_trunc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0 ) -{ - dst->f[0] = (float) (int) src0->f[0]; - dst->f[1] = (float) (int) src0->f[1]; - dst->f[2] = (float) (int) src0->f[2]; - dst->f[3] = (float) (int) src0->f[3]; -} - -static void -micro_ushr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] >> src1->u[0]; - dst->u[1] = src0->u[1] >> src1->u[1]; - dst->u[2] = src0->u[2] >> src1->u[2]; - dst->u[3] = src0->u[3] >> src1->u[3]; -} - -static void -micro_sin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) sin( (double) src->f[0] ); - dst->f[1] = (float) sin( (double) src->f[1] ); - dst->f[2] = (float) sin( (double) src->f[2] ); - dst->f[3] = (float) sin( (double) src->f[3] ); -} - -static void -micro_sqrt( union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) sqrt( (double) src->f[0] ); - dst->f[1] = (float) sqrt( (double) src->f[1] ); - dst->f[2] = (float) sqrt( (double) src->f[2] ); - dst->f[3] = (float) sqrt( (double) src->f[3] ); -} - -static void -micro_sub( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] - src1->f[0]; - dst->f[1] = src0->f[1] - src1->f[1]; - dst->f[2] = src0->f[2] - src1->f[2]; - dst->f[3] = src0->f[3] - src1->f[3]; -} - -static void -micro_u2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) src->u[0]; - dst->f[1] = (float) src->u[1]; - dst->f[2] = (float) src->u[2]; - dst->f[3] = (float) src->u[3]; -} - -static void -micro_xor( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] ^ src1->u[0]; - dst->u[1] = src0->u[1] ^ src1->u[1]; - dst->u[2] = src0->u[2] ^ src1->u[2]; - dst->u[3] = src0->u[3] ^ src1->u[3]; -} - -static void -fetch_src_file_channel( - const struct tgsi_exec_machine *mach, - const uint file, - const uint swizzle, - const union tgsi_exec_channel *index, - union tgsi_exec_channel *chan ) -{ - switch( swizzle ) { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - switch( file ) { - case TGSI_FILE_CONSTANT: - chan->f[0] = mach->Consts[index->i[0]][swizzle]; - chan->f[1] = mach->Consts[index->i[1]][swizzle]; - chan->f[2] = mach->Consts[index->i[2]][swizzle]; - chan->f[3] = mach->Consts[index->i[3]][swizzle]; - break; - - case TGSI_FILE_INPUT: - chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_TEMPORARY: - chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_IMMEDIATE: - assert( index->i[0] < (int) mach->ImmLimit ); - chan->f[0] = mach->Imms[index->i[0]][swizzle]; - assert( index->i[1] < (int) mach->ImmLimit ); - chan->f[1] = mach->Imms[index->i[1]][swizzle]; - assert( index->i[2] < (int) mach->ImmLimit ); - chan->f[2] = mach->Imms[index->i[2]][swizzle]; - assert( index->i[3] < (int) mach->ImmLimit ); - chan->f[3] = mach->Imms[index->i[3]][swizzle]; - break; - - case TGSI_FILE_ADDRESS: - chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_OUTPUT: - /* vertex/fragment output vars can be read too */ - chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; - break; - - default: - assert( 0 ); - } - break; - - case TGSI_EXTSWIZZLE_ZERO: - *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; - break; - - case TGSI_EXTSWIZZLE_ONE: - *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; - break; - - default: - assert( 0 ); - } -} - -static void -fetch_source( - const struct tgsi_exec_machine *mach, - union tgsi_exec_channel *chan, - const struct tgsi_full_src_register *reg, - const uint chan_index ) -{ - union tgsi_exec_channel index; - uint swizzle; - - index.i[0] = - index.i[1] = - index.i[2] = - index.i[3] = reg->SrcRegister.Index; - - if (reg->SrcRegister.Indirect) { - union tgsi_exec_channel index2; - union tgsi_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterInd, CHAN_X ); - fetch_src_file_channel( - mach, - reg->SrcRegisterInd.File, - swizzle, - &index2, - &indir_index ); - - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; - } - - if( reg->SrcRegister.Dimension ) { - switch( reg->SrcRegister.File ) { - case TGSI_FILE_INPUT: - index.i[0] *= 17; - index.i[1] *= 17; - index.i[2] *= 17; - index.i[3] *= 17; - break; - case TGSI_FILE_CONSTANT: - index.i[0] *= 4096; - index.i[1] *= 4096; - index.i[2] *= 4096; - index.i[3] *= 4096; - break; - default: - assert( 0 ); - } - - index.i[0] += reg->SrcRegisterDim.Index; - index.i[1] += reg->SrcRegisterDim.Index; - index.i[2] += reg->SrcRegisterDim.Index; - index.i[3] += reg->SrcRegisterDim.Index; - - if (reg->SrcRegisterDim.Indirect) { - union tgsi_exec_channel index2; - union tgsi_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterDimInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); - fetch_src_file_channel( - mach, - reg->SrcRegisterDimInd.File, - swizzle, - &index2, - &indir_index ); - - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; - } - } - - swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); - fetch_src_file_channel( - mach, - reg->SrcRegister.File, - swizzle, - &index, - chan ); - - switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { - case TGSI_UTIL_SIGN_CLEAR: - micro_abs( chan, chan ); - break; - - case TGSI_UTIL_SIGN_SET: - micro_abs( chan, chan ); - micro_neg( chan, chan ); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - micro_neg( chan, chan ); - break; - - case TGSI_UTIL_SIGN_KEEP: - break; - } - - if (reg->SrcRegisterExtMod.Complement) { - micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan ); - } -} - -static void -store_dest( - struct tgsi_exec_machine *mach, - const union tgsi_exec_channel *chan, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - uint chan_index ) -{ - union tgsi_exec_channel *dst; - - switch( reg->DstRegister.File ) { - case TGSI_FILE_NULL: - return; - - case TGSI_FILE_OUTPUT: - dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] - + reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_TEMPORARY: - dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_ADDRESS: - dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; - break; - - default: - assert( 0 ); - return; - } - - switch (inst->Instruction.Saturate) - { - case TGSI_SAT_NONE: - if (mach->ExecMask & 0x1) - dst->i[0] = chan->i[0]; - if (mach->ExecMask & 0x2) - dst->i[1] = chan->i[1]; - if (mach->ExecMask & 0x4) - dst->i[2] = chan->i[2]; - if (mach->ExecMask & 0x8) - dst->i[3] = chan->i[3]; - break; - - case TGSI_SAT_ZERO_ONE: - /* XXX need to obey ExecMask here */ - micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); - micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); - break; - - case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); - break; - - default: - assert( 0 ); - } -} - -#define FETCH(VAL,INDEX,CHAN)\ - fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) - -#define STORE(VAL,INDEX,CHAN)\ - store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) - - -/** - * Execute ARB-style KIL which is predicated by a src register. - * Kill fragment if any of the four values is less than zero. - */ -static void -exec_kilp(struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst) -{ - uint uniquemask; - uint chan_index; - uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ - union tgsi_exec_channel r[1]; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - for (chan_index = 0; chan_index < 4; chan_index++) - { - uint swizzle; - uint i; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle ( - &inst->FullSrcRegisters[0], - chan_index); - - /* check if the component has not been already tested */ - if (uniquemask & (1 << swizzle)) - continue; - uniquemask |= 1 << swizzle; - - FETCH(&r[0], 0, chan_index); - for (i = 0; i < 4; i++) - if (r[0].f[i] < 0.0f) - kilmask |= 1 << i; - } - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; -} - - -/* - * Fetch a texel using STR texture coordinates. - */ -static void -fetch_texel( struct tgsi_sampler *sampler, - const union tgsi_exec_channel *s, - const union tgsi_exec_channel *t, - const union tgsi_exec_channel *p, - float lodbias, /* XXX should be float[4] */ - union tgsi_exec_channel *r, - union tgsi_exec_channel *g, - union tgsi_exec_channel *b, - union tgsi_exec_channel *a ) -{ - uint j; - float rgba[NUM_CHANNELS][QUAD_SIZE]; - - sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); - - for (j = 0; j < 4; j++) { - r->f[j] = rgba[0][j]; - g->f[j] = rgba[1][j]; - b->f[j] = rgba[2][j]; - a->f[j] = rgba[3][j]; - } -} - - -static void -exec_tex(struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst, - boolean biasLod) -{ - const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; - union tgsi_exec_channel r[8]; - uint chan_index; - float lodBias; - - /* debug_printf("Sampler %u unit %u\n", sampler, unit); */ - - switch (inst->InstructionExtTexture.Texture) { - case TGSI_TEXTURE_1D: - - FETCH(&r[0], 0, CHAN_X); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[1], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[1] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - if (biasLod) { - FETCH(&r[1], 0, CHAN_W); - lodBias = r[2].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ - &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ - break; - - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, /* inputs */ - &r[0], &r[1], &r[2], &r[3]); /* outputs */ - break; - - case TGSI_TEXTURE_3D: - case TGSI_TEXTURE_CUBE: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, - &r[0], &r[1], &r[2], &r[3]); - break; - - default: - assert (0); - } - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[chan_index], 0, chan_index ); - } -} - - -/** - * Evaluate a constant-valued coefficient at the position of the - * current quad. - */ -static void -eval_constant_coef( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; - } -} - -/** - * Evaluate a linear-valued coefficient at the position of the - * current quad. - */ -static void -eval_linear_coef( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - const float x = mach->QuadPos.xyzw[0].f[0]; - const float y = mach->QuadPos.xyzw[1].f[0]; - const float dadx = mach->InterpCoefs[attrib].dadx[chan]; - const float dady = mach->InterpCoefs[attrib].dady[chan]; - const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; - mach->Inputs[attrib].xyzw[chan].f[0] = a0; - mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; - mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; - mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; -} - -/** - * Evaluate a perspective-valued coefficient at the position of the - * current quad. - */ -static void -eval_perspective_coef( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - const float x = mach->QuadPos.xyzw[0].f[0]; - const float y = mach->QuadPos.xyzw[1].f[0]; - const float dadx = mach->InterpCoefs[attrib].dadx[chan]; - const float dady = mach->InterpCoefs[attrib].dady[chan]; - const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; - const float *w = mach->QuadPos.xyzw[3].f; - /* divide by W here */ - mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; - mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; - mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; - mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; -} - - -typedef void (* eval_coef_func)( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ); - -static void -exec_declaration( - struct tgsi_exec_machine *mach, - const struct tgsi_full_declaration *decl ) -{ - if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { - if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask; - eval_coef_func eval; - - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - switch( decl->Interpolation.Interpolate ) { - case TGSI_INTERPOLATE_CONSTANT: - eval = eval_constant_coef; - break; - - case TGSI_INTERPOLATE_LINEAR: - eval = eval_linear_coef; - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - eval = eval_perspective_coef; - break; - - default: - assert( 0 ); - } - - if( mask == TGSI_WRITEMASK_XYZW ) { - unsigned i, j; - - for( i = first; i <= last; i++ ) { - for( j = 0; j < NUM_CHANNELS; j++ ) { - eval( mach, i, j ); - } - } - } - else { - unsigned i, j; - - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - eval( mach, i, j ); - } - } - } - } - } - } -} - -static void -exec_instruction( - struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst, - int *pc ) -{ - uint chan_index; - union tgsi_exec_channel r[8]; - - (*pc)++; - - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_f2it( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LIT: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_X ); - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[1], 0, CHAN_Y ); - micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - - FETCH( &r[2], 0, CHAN_W ); - micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); - micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); - micro_pow( &r[1], &r[1], &r[2] ); - micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Z ); - } - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH( &r[0], 0, CHAN_X ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH( &r[0], 0, CHAN_X ); - micro_sqrt( &r[0], &r[0] ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXP: - assert (0); - break; - - case TGSI_OPCODE_LOG: - assert (0); - break; - - case TGSI_OPCODE_MUL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_mul( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_ADD: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Z ); - FETCH( &r[2], 1, CHAN_Z ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_W); - FETCH(&r[2], 1, CHAN_W); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DST: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - FETCH( &r[0], 0, CHAN_Y ); - FETCH( &r[1], 1, CHAN_Y); - micro_mul( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_Z ); - STORE( &r[0], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - FETCH( &r[0], 1, CHAN_W ); - STORE( &r[0], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MIN: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - /* XXX use micro_min()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_MAX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - /* XXX use micro_max()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); - - STORE(&r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_mul( &r[0], &r[0], &r[1] ); - FETCH( &r[1], 2, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SUB: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_sub( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - micro_sub( &r[1], &r[1], &r[2] ); - micro_mul( &r[0], &r[0], &r[1] ); - micro_add( &r[0], &r[0], &r[2] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_CND: - assert (0); - break; - - case TGSI_OPCODE_CND0: - assert (0); - break; - - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - assert (0); - break; - - case TGSI_OPCODE_INDEX: - assert (0); - break; - - case TGSI_OPCODE_NEGATE: - assert (0); - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_frc( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CLAMP: - assert (0); - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_flr( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_ROUND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_rnd( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ - FETCH(&r[0], 0, CHAN_X); - - micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ - FETCH( &r[0], 0, CHAN_X ); - micro_lg2( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_pow( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - FETCH(&r[0], 0, CHAN_Y); - FETCH(&r[1], 1, CHAN_Z); - - micro_mul( &r[2], &r[0], &r[1] ); - - FETCH(&r[3], 0, CHAN_Z); - FETCH(&r[4], 1, CHAN_Y); - - micro_mul( &r[5], &r[3], &r[4] ); - micro_sub( &r[2], &r[2], &r[5] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &r[2], 0, CHAN_X ); - } - - FETCH(&r[2], 1, CHAN_X); - - micro_mul( &r[3], &r[3], &r[2] ); - - FETCH(&r[5], 0, CHAN_X); - - micro_mul( &r[1], &r[1], &r[5] ); - micro_sub( &r[3], &r[3], &r[1] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - STORE( &r[3], 0, CHAN_Y ); - } - - micro_mul( &r[5], &r[5], &r[4] ); - micro_mul( &r[0], &r[0], &r[2] ); - micro_sub( &r[5], &r[5], &r[0] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - STORE( &r[5], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MULTIPLYMATRIX: - assert (0); - break; - - case TGSI_OPCODE_ABS: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - - micro_abs( &r[0], &r[0] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_RCC: - assert (0); - break; - - case TGSI_OPCODE_DPH: - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 1, CHAN_W); - - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_COS: - FETCH(&r[0], 0, CHAN_X); - - micro_cos( &r[0], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddx( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDY: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddy( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_KILP: - exec_kilp (mach, inst); - break; - - case TGSI_OPCODE_KIL: - /* for enabled ExecMask bits, set the killed bit */ - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; - break; - - case TGSI_OPCODE_PK2H: - assert (0); - break; - - case TGSI_OPCODE_PK2US: - assert (0); - break; - - case TGSI_OPCODE_PK4B: - assert (0); - break; - - case TGSI_OPCODE_PK4UB: - assert (0); - break; - - case TGSI_OPCODE_RFL: - assert (0); - break; - - case TGSI_OPCODE_SEQ: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], - &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], - &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SFL: - assert (0); - break; - - case TGSI_OPCODE_SGT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SIN: - FETCH( &r[0], 0, CHAN_X ); - micro_sin( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SNE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_STR: - assert (0); - break; - - case TGSI_OPCODE_TEX: - /* simple texture lookup */ - /* src[0] = texcoord */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, FALSE); - break; - - case TGSI_OPCODE_TXB: - /* Texture lookup with lod bias */ - /* src[0] = texcoord (src[0].w = LOD bias) */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); - break; - - case TGSI_OPCODE_TXD: - /* Texture lookup with explict partial derivatives */ - /* src[0] = texcoord */ - /* src[1] = d[strq]/dx */ - /* src[2] = d[strq]/dy */ - /* src[3] = sampler unit */ - assert (0); - break; - - case TGSI_OPCODE_TXL: - /* Texture lookup with explit LOD */ - /* src[0] = texcoord (src[0].w = LOD) */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); - break; - - case TGSI_OPCODE_UP2H: - assert (0); - break; - - case TGSI_OPCODE_UP2US: - assert (0); - break; - - case TGSI_OPCODE_UP4B: - assert (0); - break; - - case TGSI_OPCODE_UP4UB: - assert (0); - break; - - case TGSI_OPCODE_X2D: - assert (0); - break; - - case TGSI_OPCODE_ARA: - assert (0); - break; - - case TGSI_OPCODE_ARR: - assert (0); - break; - - case TGSI_OPCODE_BRA: - assert (0); - break; - - case TGSI_OPCODE_CAL: - /* skip the call if no execution channels are enabled */ - if (mach->ExecMask) { - /* do the call */ - - /* push the Cond, Loop, Cont stacks */ - assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); - mach->CondStack[mach->CondStackTop++] = mach->CondMask; - assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; - assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; - - assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); - mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; - - /* note that PC was already incremented above */ - mach->CallStack[mach->CallStackTop++] = *pc; - *pc = inst->InstructionExtLabel.Label; - } - break; - - case TGSI_OPCODE_RET: - mach->FuncMask &= ~mach->ExecMask; - UPDATE_EXEC_MASK(mach); - - if (mach->ExecMask == 0x0) { - /* really return now (otherwise, keep executing */ - - if (mach->CallStackTop == 0) { - /* returning from main() */ - *pc = -1; - return; - } - *pc = mach->CallStack[--mach->CallStackTop]; - - /* pop the Cond, Loop, Cont stacks */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; - assert(mach->FuncStackTop > 0); - mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; - - UPDATE_EXEC_MASK(mach); - } - break; - - case TGSI_OPCODE_SSG: - assert (0); - break; - - case TGSI_OPCODE_CMP: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_SCS: - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( &r[0], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - micro_cos( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - micro_sin( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_Y ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_NRM: - assert (0); - break; - - case TGSI_OPCODE_DIV: - assert( 0 ); - break; - - case TGSI_OPCODE_DP2: - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_IF: - /* push CondMask */ - assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); - mach->CondStack[mach->CondStackTop++] = mach->CondMask; - FETCH( &r[0], 0, CHAN_X ); - /* update CondMask */ - if( ! r[0].u[0] ) { - mach->CondMask &= ~0x1; - } - if( ! r[0].u[1] ) { - mach->CondMask &= ~0x2; - } - if( ! r[0].u[2] ) { - mach->CondMask &= ~0x4; - } - if( ! r[0].u[3] ) { - mach->CondMask &= ~0x8; - } - UPDATE_EXEC_MASK(mach); - /* Todo: If CondMask==0, jump to ELSE */ - break; - - case TGSI_OPCODE_ELSE: - /* invert CondMask wrt previous mask */ - { - uint prevMask; - assert(mach->CondStackTop > 0); - prevMask = mach->CondStack[mach->CondStackTop - 1]; - mach->CondMask = ~mach->CondMask & prevMask; - UPDATE_EXEC_MASK(mach); - /* Todo: If CondMask==0, jump to ENDIF */ - } - break; - - case TGSI_OPCODE_ENDIF: - /* pop CondMask */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_END: - /* halt execution */ - *pc = -1; - break; - - case TGSI_OPCODE_REP: - assert (0); - break; - - case TGSI_OPCODE_ENDREP: - assert (0); - break; - - case TGSI_OPCODE_PUSHA: - assert (0); - break; - - case TGSI_OPCODE_POPA: - assert (0); - break; - - case TGSI_OPCODE_CEIL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ceil( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_I2F: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_i2f( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_NOT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_not( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_TRUNC: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_trunc( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SHL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_shl( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SHR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ishr( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_AND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_and( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_OR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_or( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOD: - assert (0); - break; - - case TGSI_OPCODE_XOR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_xor( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SAD: - assert (0); - break; - - case TGSI_OPCODE_TXF: - assert (0); - break; - - case TGSI_OPCODE_TXQ: - assert (0); - break; - - case TGSI_OPCODE_EMIT: - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; - break; - - case TGSI_OPCODE_ENDPRIM: - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; - break; - - case TGSI_OPCODE_LOOP: - /* fall-through (for now) */ - case TGSI_OPCODE_BGNLOOP2: - /* push LoopMask and ContMasks */ - assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; - assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; - break; - - case TGSI_OPCODE_ENDLOOP: - /* fall-through (for now at least) */ - case TGSI_OPCODE_ENDLOOP2: - /* Restore ContMask, but don't pop */ - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; - if (mach->LoopMask) { - /* repeat loop: jump to instruction just past BGNLOOP */ - *pc = inst->InstructionExtLabel.Label + 1; - } - else { - /* exit loop: pop LoopMask */ - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - /* pop ContMask */ - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; - } - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_BRK: - /* turn off loop channels for each enabled exec channel */ - mach->LoopMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_CONT: - /* turn off cont channels for each enabled exec channel */ - mach->ContMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_BGNSUB: - /* no-op */ - break; - - case TGSI_OPCODE_ENDSUB: - /* no-op */ - break; - - case TGSI_OPCODE_NOISE1: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE2: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE3: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE4: - assert( 0 ); - break; - - case TGSI_OPCODE_NOP: - break; - - default: - assert( 0 ); - } -} - - -/** - * Run TGSI interpreter. - * \return bitmask of "alive" quad components - */ -uint -tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) -{ - uint i; - int pc = 0; - - mach->CondMask = 0xf; - mach->LoopMask = 0xf; - mach->ContMask = 0xf; - mach->FuncMask = 0xf; - mach->ExecMask = 0xf; - - mach->CondStackTop = 0; /* temporarily subvert this assertion */ - assert(mach->CondStackTop == 0); - assert(mach->LoopStackTop == 0); - assert(mach->ContStackTop == 0); - assert(mach->CallStackTop == 0); - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; - - if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; - mach->Primitives[0] = 0; - } - - - /* execute declarations (interpolants) */ - for (i = 0; i < mach->NumDeclarations; i++) { - exec_declaration( mach, mach->Declarations+i ); - } - - /* execute instructions, until pc is set to -1 */ - while (pc != -1) { - assert(pc < mach->NumInstructions); - exec_instruction( mach, mach->Instructions + pc, &pc ); - } - -#if 0 - /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ - if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { - /* - * Scale back depth component. - */ - for (i = 0; i < 4; i++) - mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; - } -#endif - - return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; -} - - diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h deleted file mode 100644 index 1fb66ee960..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ /dev/null @@ -1,239 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#if !defined TGSI_EXEC_H -#define TGSI_EXEC_H - -#include "pipe/p_compiler.h" - -#if defined __cplusplus -extern "C" { -#endif - -#define NUM_CHANNELS 4 /* R,G,B,A */ -#define QUAD_SIZE 4 /* 4 pixel/quad */ - -/** - * Registers may be treated as float, signed int or unsigned int. - */ -union tgsi_exec_channel -{ - float f[QUAD_SIZE]; - int i[QUAD_SIZE]; - unsigned u[QUAD_SIZE]; -}; - -/** - * A vector[RGBA] of channels[4 pixels] - */ -struct tgsi_exec_vector -{ - union tgsi_exec_channel xyzw[NUM_CHANNELS]; -}; - -/** - * For fragment programs, information for computing fragment input - * values from plane equation of the triangle/line. - */ -struct tgsi_interp_coef -{ - float a0[NUM_CHANNELS]; /* in an xyzw layout */ - float dadx[NUM_CHANNELS]; - float dady[NUM_CHANNELS]; -}; - - -struct softpipe_tile_cache; /**< Opaque to TGSI */ - -/** - * Information for sampling textures, which must be implemented - * by code outside the TGSI executor. - */ -struct tgsi_sampler -{ - const struct pipe_sampler_state *state; - struct pipe_texture *texture; - /** Get samples for four fragments in a quad */ - void (*get_samples)(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]); - void *pipe; /*XXX temporary*/ - struct softpipe_tile_cache *cache; -}; - -/** - * For branching/calling subroutines. - */ -struct tgsi_exec_labels -{ - unsigned labels[128][2]; - unsigned count; -}; - -/* - * Locations of various utility registers (_I = Index, _C = Channel) - */ -#define TGSI_EXEC_TEMP_00000000_I 32 -#define TGSI_EXEC_TEMP_00000000_C 0 - -#define TGSI_EXEC_TEMP_7FFFFFFF_I 32 -#define TGSI_EXEC_TEMP_7FFFFFFF_C 1 - -#define TGSI_EXEC_TEMP_80000000_I 32 -#define TGSI_EXEC_TEMP_80000000_C 2 - -#define TGSI_EXEC_TEMP_FFFFFFFF_I 32 -#define TGSI_EXEC_TEMP_FFFFFFFF_C 3 - -#define TGSI_EXEC_TEMP_ONE_I 33 -#define TGSI_EXEC_TEMP_ONE_C 0 - -#define TGSI_EXEC_TEMP_TWO_I 33 -#define TGSI_EXEC_TEMP_TWO_C 1 - -#define TGSI_EXEC_TEMP_128_I 33 -#define TGSI_EXEC_TEMP_128_C 2 - -#define TGSI_EXEC_TEMP_MINUS_128_I 33 -#define TGSI_EXEC_TEMP_MINUS_128_C 3 - -#define TGSI_EXEC_TEMP_KILMASK_I 34 -#define TGSI_EXEC_TEMP_KILMASK_C 0 - -#define TGSI_EXEC_TEMP_OUTPUT_I 34 -#define TGSI_EXEC_TEMP_OUTPUT_C 1 - -#define TGSI_EXEC_TEMP_PRIMITIVE_I 34 -#define TGSI_EXEC_TEMP_PRIMITIVE_C 2 - -#define TGSI_EXEC_TEMP_R0 35 - -#define TGSI_EXEC_NUM_TEMPS (32 + 4) -#define TGSI_EXEC_NUM_ADDRS 1 -#define TGSI_EXEC_NUM_IMMEDIATES 256 - -#define TGSI_EXEC_MAX_COND_NESTING 10 -#define TGSI_EXEC_MAX_LOOP_NESTING 10 -#define TGSI_EXEC_MAX_CALL_NESTING 10 - -/** - * Run-time virtual machine state for executing TGSI shader. - */ -struct tgsi_exec_machine -{ - /* - * 32 program temporaries - * 4 internal temporaries - * 1 address - * 1 temporary of padding to align to 16 bytes - */ - struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS + 1]; - - /* - * This will point to _Temps after aligning to 16B boundary. - */ - struct tgsi_exec_vector *Temps; - struct tgsi_exec_vector *Addrs; - - struct tgsi_sampler *Samplers; - - float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; - unsigned ImmLimit; - float (*Consts)[4]; - struct tgsi_exec_vector *Inputs; - struct tgsi_exec_vector *Outputs; - const struct tgsi_token *Tokens; - unsigned Processor; - - /* GEOMETRY processor only. */ - unsigned *Primitives; - - /* FRAGMENT processor only. */ - const struct tgsi_interp_coef *InterpCoefs; - struct tgsi_exec_vector QuadPos; - - /* Conditional execution masks */ - uint CondMask; /**< For IF/ELSE/ENDIF */ - uint LoopMask; /**< For BGNLOOP/ENDLOOP */ - uint ContMask; /**< For loop CONT statements */ - uint FuncMask; /**< For function calls */ - uint ExecMask; /**< = CondMask & LoopMask */ - - /** Condition mask stack (for nested conditionals) */ - uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; - int CondStackTop; - - /** Loop mask stack (for nested loops) */ - uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; - int LoopStackTop; - - /** Loop continue mask stack (see comments in tgsi_exec.c) */ - uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; - int ContStackTop; - - /** Function execution mask stack (for executing subroutine code) */ - uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; - int FuncStackTop; - - /** Function call stack for saving/restoring the program counter */ - uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; - int CallStackTop; - - struct tgsi_full_instruction *Instructions; - uint NumInstructions; - - struct tgsi_full_declaration *Declarations; - uint NumDeclarations; - - struct tgsi_exec_labels Labels; -}; - - -void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach, - const struct tgsi_token *tokens, - unsigned numSamplers, - struct tgsi_sampler *samplers); - -uint -tgsi_exec_machine_run( - struct tgsi_exec_machine *mach ); - - -void -tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach); - - -#if defined __cplusplus -} /* extern "C" */ -#endif - -#endif /* TGSI_EXEC_H */ diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c deleted file mode 100755 index 1e56e4afb6..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ /dev/null @@ -1,2378 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/util/tgsi_util.h" -#include "tgsi_exec.h" -#include "tgsi_sse2.h" - -#include "x86/rtasm/x86sse.h" - -#if defined(__i386__) || defined(__386__) - -#define DUMP_SSE 0 - -#if DUMP_SSE - -static void -_print_reg( - struct x86_reg reg ) -{ - if (reg.mod != mod_REG) - debug_printf( "[" ); - - switch( reg.file ) { - case file_REG32: - switch( reg.idx ) { - case reg_AX: - debug_printf( "EAX" ); - break; - case reg_CX: - debug_printf( "ECX" ); - break; - case reg_DX: - debug_printf( "EDX" ); - break; - case reg_BX: - debug_printf( "EBX" ); - break; - case reg_SP: - debug_printf( "ESP" ); - break; - case reg_BP: - debug_printf( "EBP" ); - break; - case reg_SI: - debug_printf( "ESI" ); - break; - case reg_DI: - debug_printf( "EDI" ); - break; - } - break; - case file_MMX: - assert( 0 ); - break; - case file_XMM: - debug_printf( "XMM%u", reg.idx ); - break; - case file_x87: - assert( 0 ); - break; - } - - if (reg.mod == mod_DISP8 || - reg.mod == mod_DISP32) - debug_printf("+%d", reg.disp); - - if (reg.mod != mod_REG) - debug_printf( "]" ); -} - -static void -_fill( - const char *op ) -{ - unsigned count = 10 - strlen( op ); - - while( count-- ) { - debug_printf( " " ); - } -} - -#define DUMP_START() debug_printf( "\nsse-dump start ----------------" ) -#define DUMP_END() debug_printf( "\nsse-dump end ----------------\n" ) -#define DUMP( OP ) debug_printf( "\n%s", OP ) -#define DUMP_I( OP, I ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - debug_printf( "%u", I ); } while( 0 ) -#define DUMP_R( OP, R0 ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - _print_reg( R0 ); } while( 0 ) -#define DUMP_RR( OP, R0, R1 ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - _print_reg( R0 );\ - debug_printf( ", " );\ - _print_reg( R1 ); } while( 0 ) -#define DUMP_RRI( OP, R0, R1, I ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - _print_reg( R0 );\ - debug_printf( ", " );\ - _print_reg( R1 );\ - debug_printf( ", " );\ - debug_printf( "%u", I ); } while( 0 ) - -#else - -#define DUMP_START() -#define DUMP_END() -#define DUMP( OP ) -#define DUMP_I( OP, I ) -#define DUMP_R( OP, R0 ) -#define DUMP_RR( OP, R0, R1 ) -#define DUMP_RRI( OP, R0, R1, I ) - -#endif - -#define FOR_EACH_CHANNEL( CHAN )\ - for( CHAN = 0; CHAN < 4; CHAN++ ) - -#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ - ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) - -#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ - if( IS_DST0_CHANNEL_ENABLED( INST, CHAN )) - -#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\ - FOR_EACH_CHANNEL( CHAN )\ - IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN ) - -#define CHAN_X 0 -#define CHAN_Y 1 -#define CHAN_Z 2 -#define CHAN_W 3 - -#define TEMP_R0 TGSI_EXEC_TEMP_R0 - -/** - * X86 utility functions. - */ - -static struct x86_reg -make_xmm( - unsigned xmm ) -{ - return x86_make_reg( - file_XMM, - (enum x86_reg_name) xmm ); -} - -/** - * X86 register mapping helpers. - */ - -static struct x86_reg -get_const_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_CX ); -} - -static struct x86_reg -get_input_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_AX ); -} - -static struct x86_reg -get_output_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_DX ); -} - -static struct x86_reg -get_temp_base( void ) -{ -#ifdef WIN32 - return x86_make_reg( - file_REG32, - reg_BX ); -#else - return x86_make_reg( - file_REG32, - reg_SI ); -#endif -} - -static struct x86_reg -get_coef_base( void ) -{ - return get_output_base(); -} - -/** - * Data access helpers. - */ - -static struct x86_reg -get_argument( - unsigned index ) -{ - return x86_make_disp( - x86_make_reg( file_REG32, reg_SP ), - (index + 1) * 4 ); -} - -static struct x86_reg -get_const( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_const_base(), - (vec * 4 + chan) * 4 ); -} - -static struct x86_reg -get_input( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_input_base(), - (vec * 4 + chan) * 16 ); -} - -static struct x86_reg -get_output( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_output_base(), - (vec * 4 + chan) * 16 ); -} - -static struct x86_reg -get_temp( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_temp_base(), - (vec * 4 + chan) * 16 ); -} - -static struct x86_reg -get_coef( - unsigned vec, - unsigned chan, - unsigned member ) -{ - return x86_make_disp( - get_coef_base(), - ((vec * 3 + member) * 4 + chan) * 4 ); -} - -/** - * X86 rtasm wrappers. - */ - -static void -emit_addps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ADDPS", dst, src ); - sse_addps( func, dst, src ); -} - -static void -emit_andnps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ANDNPS", dst, src ); - sse_andnps( func, dst, src ); -} - -static void -emit_andps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ANDPS", dst, src ); - sse_andps( func, dst, src ); -} - -static void -emit_call( - struct x86_function *func, - void (* addr)() ) -{ - struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); - - DUMP_I( "CALL", addr ); - x86_mov_reg_imm( func, ecx, (unsigned long) addr ); - x86_call( func, ecx ); -} - -static void -emit_cmpps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src, - enum sse_cc cc ) -{ - DUMP_RRI( "CMPPS", dst, src, cc ); - sse_cmpps( func, dst, src, cc ); -} - -static void -emit_cvttps2dq( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "CVTTPS2DQ", dst, src ); - sse2_cvttps2dq( func, dst, src ); -} - -static void -emit_maxps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MAXPS", dst, src ); - sse_maxps( func, dst, src ); -} - -static void -emit_minps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MINPS", dst, src ); - sse_minps( func, dst, src ); -} - -static void -emit_mov( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOV", dst, src ); - x86_mov( func, dst, src ); -} - -static void -emit_movaps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOVAPS", dst, src ); - sse_movaps( func, dst, src ); -} - -static void -emit_movss( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOVSS", dst, src ); - sse_movss( func, dst, src ); -} - -static void -emit_movups( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOVUPS", dst, src ); - sse_movups( func, dst, src ); -} - -static void -emit_mulps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MULPS", dst, src ); - sse_mulps( func, dst, src ); -} - -static void -emit_or( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "OR", dst, src ); - x86_or( func, dst, src ); -} - -static void -emit_orps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ORPS", dst, src ); - sse_orps( func, dst, src ); -} - -static void -emit_pmovmskb( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "PMOVMSKB", dst, src ); - sse_pmovmskb( func, dst, src ); -} - -static void -emit_pop( - struct x86_function *func, - struct x86_reg dst ) -{ - DUMP_R( "POP", dst ); - x86_pop( func, dst ); -} - -static void -emit_push( - struct x86_function *func, - struct x86_reg dst ) -{ - DUMP_R( "PUSH", dst ); - x86_push( func, dst ); -} - -static void -emit_rcpps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "RCPPS", dst, src ); - sse2_rcpps( func, dst, src ); -} - -#ifdef WIN32 -static void -emit_retw( - struct x86_function *func, - unsigned size ) -{ - DUMP_I( "RET", size ); - x86_retw( func, size ); -} -#else -static void -emit_ret( - struct x86_function *func ) -{ - DUMP( "RET" ); - x86_ret( func ); -} -#endif - -static void -emit_rsqrtps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "RSQRTPS", dst, src ); - sse_rsqrtps( func, dst, src ); -} - -static void -emit_shufps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src, - unsigned char shuf ) -{ - DUMP_RRI( "SHUFPS", dst, src, shuf ); - sse_shufps( func, dst, src, shuf ); -} - -static void -emit_subps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "SUBPS", dst, src ); - sse_subps( func, dst, src ); -} - -static void -emit_xorps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "XORPS", dst, src ); - sse_xorps( func, dst, src ); -} - -/** - * Data fetch helpers. - */ - -static void -emit_const( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movss( - func, - make_xmm( xmm ), - get_const( vec, chan ) ); - emit_shufps( - func, - make_xmm( xmm ), - make_xmm( xmm ), - SHUF( 0, 0, 0, 0 ) ); -} - -static void -emit_inputf( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movups( - func, - make_xmm( xmm ), - get_input( vec, chan ) ); -} - -static void -emit_output( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movups( - func, - get_output( vec, chan ), - make_xmm( xmm ) ); -} - -static void -emit_tempf( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movaps( - func, - make_xmm( xmm ), - get_temp( vec, chan ) ); -} - -static void -emit_coef( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan, - unsigned member ) -{ - emit_movss( - func, - make_xmm( xmm ), - get_coef( vec, chan, member ) ); - emit_shufps( - func, - make_xmm( xmm ), - make_xmm( xmm ), - SHUF( 0, 0, 0, 0 ) ); -} - -/** - * Data store helpers. - */ - -static void -emit_inputs( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movups( - func, - get_input( vec, chan ), - make_xmm( xmm ) ); -} - -static void -emit_temps( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movaps( - func, - get_temp( vec, chan ), - make_xmm( xmm ) ); -} - -static void -emit_addrs( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_temps( - func, - xmm, - vec + TGSI_EXEC_NUM_TEMPS, - chan ); -} - -/** - * Coefficent fetch helpers. - */ - -static void -emit_coef_a0( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 0 ); -} - -static void -emit_coef_dadx( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 1 ); -} - -static void -emit_coef_dady( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 2 ); -} - -/** - * Function call helpers. - */ - -static void -emit_push_gp( - struct x86_function *func ) -{ - emit_push( - func, - get_const_base() ); - emit_push( - func, - get_input_base() ); - emit_push( - func, - get_output_base() ); - - /* It is important on non-win32 platforms that temp base is pushed last. - */ - emit_push( - func, - get_temp_base() ); -} - -static void -emit_pop_gp( - struct x86_function *func ) -{ - /* Restore GP registers in a reverse order. - */ - emit_pop( - func, - get_temp_base() ); - emit_pop( - func, - get_output_base() ); - emit_pop( - func, - get_input_base() ); - emit_pop( - func, - get_const_base() ); -} - -static void -emit_func_call_dst( - struct x86_function *func, - unsigned xmm_dst, - void (*code)() ) -{ - emit_movaps( - func, - get_temp( TEMP_R0, 0 ), - make_xmm( xmm_dst ) ); - - emit_push_gp( - func ); - -#ifdef WIN32 - emit_push( - func, - get_temp( TEMP_R0, 0 ) ); -#endif - - emit_call( - func, - code ); - - emit_pop_gp( - func ); - - emit_movaps( - func, - make_xmm( xmm_dst ), - get_temp( TEMP_R0, 0 ) ); -} - -static void -emit_func_call_dst_src( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src, - void (*code)() ) -{ - emit_movaps( - func, - get_temp( TEMP_R0, 1 ), - make_xmm( xmm_src ) ); - - emit_func_call_dst( - func, - xmm_dst, - code ); -} - -/** - * Low-level instruction translators. - */ - -static void -emit_abs( - struct x86_function *func, - unsigned xmm ) -{ - emit_andps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_7FFFFFFF_I, - TGSI_EXEC_TEMP_7FFFFFFF_C ) ); -} - -static void -emit_add( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_addps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void XSTDCALL -cos4f( - float *store ) -{ -#ifdef WIN32 - store[0] = (float) cos( (double) store[0] ); - store[1] = (float) cos( (double) store[1] ); - store[2] = (float) cos( (double) store[2] ); - store[3] = (float) cos( (double) store[3] ); -#else - const unsigned X = TEMP_R0 * 16; - store[X + 0] = cosf( store[X + 0] ); - store[X + 1] = cosf( store[X + 1] ); - store[X + 2] = cosf( store[X + 2] ); - store[X + 3] = cosf( store[X + 3] ); -#endif -} - -static void -emit_cos( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - cos4f ); -} - -static void XSTDCALL -ex24f( - float *store ) -{ -#ifdef WIN32 - store[0] = (float) pow( 2.0, (double) store[0] ); - store[1] = (float) pow( 2.0, (double) store[1] ); - store[2] = (float) pow( 2.0, (double) store[2] ); - store[3] = (float) pow( 2.0, (double) store[3] ); -#else - const unsigned X = TEMP_R0 * 16; - store[X + 0] = powf( 2.0f, store[X + 0] ); - store[X + 1] = powf( 2.0f, store[X + 1] ); - store[X + 2] = powf( 2.0f, store[X + 2] ); - store[X + 3] = powf( 2.0f, store[X + 3] ); -#endif -} - -static void -emit_ex2( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - ex24f ); -} - -static void -emit_f2it( - struct x86_function *func, - unsigned xmm ) -{ - emit_cvttps2dq( - func, - make_xmm( xmm ), - make_xmm( xmm ) ); -} - -static void XSTDCALL -flr4f( - float *store ) -{ -#ifdef WIN32 - const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif - store[X + 0] = (float) floor( (double) store[X + 0] ); - store[X + 1] = (float) floor( (double) store[X + 1] ); - store[X + 2] = (float) floor( (double) store[X + 2] ); - store[X + 3] = (float) floor( (double) store[X + 3] ); -} - -static void -emit_flr( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - flr4f ); -} - -static void XSTDCALL -frc4f( - float *store ) -{ -#ifdef WIN32 - const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif - store[X + 0] -= (float) floor( (double) store[X + 0] ); - store[X + 1] -= (float) floor( (double) store[X + 1] ); - store[X + 2] -= (float) floor( (double) store[X + 2] ); - store[X + 3] -= (float) floor( (double) store[X + 3] ); -} - -static void -emit_frc( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - frc4f ); -} - -static void XSTDCALL -lg24f( - float *store ) -{ -#ifdef WIN32 - const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif - store[X + 0] = LOG2( store[X + 0] ); - store[X + 1] = LOG2( store[X + 1] ); - store[X + 2] = LOG2( store[X + 2] ); - store[X + 3] = LOG2( store[X + 3] ); -} - -static void -emit_lg2( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - lg24f ); -} - -static void -emit_MOV( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_movups( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_mul (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) -{ - emit_mulps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_neg( - struct x86_function *func, - unsigned xmm ) -{ - emit_xorps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C ) ); -} - -static void XSTDCALL -pow4f( - float *store ) -{ -#ifdef WIN32 - store[0] = (float) pow( (double) store[0], (double) store[4] ); - store[1] = (float) pow( (double) store[1], (double) store[5] ); - store[2] = (float) pow( (double) store[2], (double) store[6] ); - store[3] = (float) pow( (double) store[3], (double) store[7] ); -#else - const unsigned X = TEMP_R0 * 16; - store[X + 0] = powf( store[X + 0], store[X + 4] ); - store[X + 1] = powf( store[X + 1], store[X + 5] ); - store[X + 2] = powf( store[X + 2], store[X + 6] ); - store[X + 3] = powf( store[X + 3], store[X + 7] ); -#endif -} - -static void -emit_pow( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_func_call_dst_src( - func, - xmm_dst, - xmm_src, - pow4f ); -} - -static void -emit_rcp ( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_rcpps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_rsqrt( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_rsqrtps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_setsign( - struct x86_function *func, - unsigned xmm ) -{ - emit_orps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C ) ); -} - -static void XSTDCALL -sin4f( - float *store ) -{ -#ifdef WIN32 - store[0] = (float) sin( (double) store[0] ); - store[1] = (float) sin( (double) store[1] ); - store[2] = (float) sin( (double) store[2] ); - store[3] = (float) sin( (double) store[3] ); -#else - const unsigned X = TEMP_R0 * 16; - store[X + 0] = sinf( store[X + 0] ); - store[X + 1] = sinf( store[X + 1] ); - store[X + 2] = sinf( store[X + 2] ); - store[X + 3] = sinf( store[X + 3] ); -#endif -} - -static void -emit_sin (struct x86_function *func, - unsigned xmm_dst) -{ - emit_func_call_dst( - func, - xmm_dst, - sin4f ); -} - -static void -emit_sub( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_subps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -/** - * Register fetch. - */ - -static void -emit_fetch( - struct x86_function *func, - unsigned xmm, - const struct tgsi_full_src_register *reg, - const unsigned chan_index ) -{ - unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); - - switch( swizzle ) { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - switch( reg->SrcRegister.File ) { - case TGSI_FILE_CONSTANT: - emit_const( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - case TGSI_FILE_INPUT: - emit_inputf( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - case TGSI_FILE_TEMPORARY: - emit_tempf( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - default: - assert( 0 ); - } - break; - - case TGSI_EXTSWIZZLE_ZERO: - emit_tempf( - func, - xmm, - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ); - break; - - case TGSI_EXTSWIZZLE_ONE: - emit_tempf( - func, - xmm, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - break; - - default: - assert( 0 ); - } - - switch( tgsi_util_get_full_src_register_sign_mode( reg, chan_index ) ) { - case TGSI_UTIL_SIGN_CLEAR: - emit_abs( func, xmm ); - break; - - case TGSI_UTIL_SIGN_SET: - emit_setsign( func, xmm ); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - emit_neg( func, xmm ); - break; - - case TGSI_UTIL_SIGN_KEEP: - break; - } -} - -#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\ - emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN ) - -/** - * Register store. - */ - -static void -emit_store( - struct x86_function *func, - unsigned xmm, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - unsigned chan_index ) -{ - switch( reg->DstRegister.File ) { - case TGSI_FILE_OUTPUT: - emit_output( - func, - xmm, - reg->DstRegister.Index, - chan_index ); - break; - - case TGSI_FILE_TEMPORARY: - emit_temps( - func, - xmm, - reg->DstRegister.Index, - chan_index ); - break; - - case TGSI_FILE_ADDRESS: - emit_addrs( - func, - xmm, - reg->DstRegister.Index, - chan_index ); - break; - - default: - assert( 0 ); - } - - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - - case TGSI_SAT_ZERO_ONE: -// assert( 0 ); - break; - - case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); - break; - } -} - -#define STORE( FUNC, INST, XMM, INDEX, CHAN )\ - emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) - -/** - * High-level instruction translators. - */ - -static void -emit_kil( - struct x86_function *func, - const struct tgsi_full_src_register *reg ) -{ - unsigned uniquemask; - unsigned registers[4]; - unsigned nextregister = 0; - unsigned firstchan = ~0; - unsigned chan_index; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - FOR_EACH_CHANNEL( chan_index ) { - unsigned swizzle; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle( - reg, - chan_index ); - - /* check if the component has not been already tested */ - if( !(uniquemask & (1 << swizzle)) ) { - uniquemask |= 1 << swizzle; - - /* allocate register */ - registers[chan_index] = nextregister; - emit_fetch( - func, - nextregister, - reg, - chan_index ); - nextregister++; - - /* mark the first channel used */ - if( firstchan == ~0 ) { - firstchan = chan_index; - } - } - } - - emit_push( - func, - x86_make_reg( file_REG32, reg_AX ) ); - emit_push( - func, - x86_make_reg( file_REG32, reg_DX ) ); - - FOR_EACH_CHANNEL( chan_index ) { - if( uniquemask & (1 << chan_index) ) { - emit_cmpps( - func, - make_xmm( registers[chan_index] ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ), - cc_LessThan ); - - if( chan_index == firstchan ) { - emit_pmovmskb( - func, - x86_make_reg( file_REG32, reg_AX ), - make_xmm( registers[chan_index] ) ); - } - else { - emit_pmovmskb( - func, - x86_make_reg( file_REG32, reg_DX ), - make_xmm( registers[chan_index] ) ); - emit_or( - func, - x86_make_reg( file_REG32, reg_AX ), - x86_make_reg( file_REG32, reg_DX ) ); - } - } - } - - emit_or( - func, - get_temp( - TGSI_EXEC_TEMP_KILMASK_I, - TGSI_EXEC_TEMP_KILMASK_C ), - x86_make_reg( file_REG32, reg_AX ) ); - - emit_pop( - func, - x86_make_reg( file_REG32, reg_DX ) ); - emit_pop( - func, - x86_make_reg( file_REG32, reg_AX ) ); -} - -static void -emit_setcc( - struct x86_function *func, - struct tgsi_full_instruction *inst, - enum sse_cc cc ) -{ - unsigned chan_index; - - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_cmpps( - func, - make_xmm( 0 ), - make_xmm( 1 ), - cc ); - emit_andps( - func, - make_xmm( 0 ), - get_temp( - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ) ); - STORE( func, *inst, 0, 0, chan_index ); - } -} - -static void -emit_cmp( - struct x86_function *func, - struct tgsi_full_instruction *inst ) -{ - unsigned chan_index; - - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - FETCH( func, *inst, 2, 2, chan_index ); - emit_cmpps( - func, - make_xmm( 0 ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ), - cc_LessThan ); - emit_andps( - func, - make_xmm( 1 ), - make_xmm( 0 ) ); - emit_andnps( - func, - make_xmm( 0 ), - make_xmm( 2 ) ); - emit_orps( - func, - make_xmm( 0 ), - make_xmm( 1 ) ); - STORE( func, *inst, 0, 0, chan_index ); - } -} - -static int -emit_instruction( - struct x86_function *func, - struct tgsi_full_instruction *inst ) -{ - unsigned chan_index; - - switch( inst->Instruction.Opcode ) { - case TGSI_OPCODE_ARL: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_f2it( func, 0 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_LIT: - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C); - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - STORE( func, *inst, 0, 0, CHAN_X ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - STORE( func, *inst, 0, 0, CHAN_W ); - } - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_maxps( - func, - make_xmm( 0 ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ) ); - STORE( func, *inst, 0, 0, CHAN_Y ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 1, 0, CHAN_Y ); - emit_maxps( - func, - make_xmm( 1 ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ) ); - FETCH( func, *inst, 2, 0, CHAN_W ); - emit_minps( - func, - make_xmm( 2 ), - get_temp( - TGSI_EXEC_TEMP_128_I, - TGSI_EXEC_TEMP_128_C ) ); - emit_maxps( - func, - make_xmm( 2 ), - get_temp( - TGSI_EXEC_TEMP_MINUS_128_I, - TGSI_EXEC_TEMP_MINUS_128_C ) ); - emit_pow( func, 1, 2 ); - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_xorps( - func, - make_xmm( 2 ), - make_xmm( 2 ) ); - emit_cmpps( - func, - make_xmm( 2 ), - make_xmm( 0 ), - cc_LessThanEqual ); - emit_andps( - func, - make_xmm( 2 ), - make_xmm( 1 ) ); - STORE( func, *inst, 2, 0, CHAN_Z ); - } - } - break; - - case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_rcp( func, 0, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_rsqrt( func, 0, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXP: - return 0; - break; - - case TGSI_OPCODE_LOG: - return 0; - break; - - case TGSI_OPCODE_MUL: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_mul( func, 0, 1 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_ADD: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_add( func, 0, 1 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_mul( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Y ); - FETCH( func, *inst, 2, 1, CHAN_Y ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Z ); - FETCH( func, *inst, 2, 1, CHAN_Z ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_mul( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Y ); - FETCH( func, *inst, 2, 1, CHAN_Y ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Z ); - FETCH( func, *inst, 2, 1, CHAN_Z ); - emit_mul(func, 1, 2 ); - emit_add(func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_W ); - FETCH( func, *inst, 2, 1, CHAN_W ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DST: - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - STORE( func, *inst, 0, 0, CHAN_X ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - FETCH( func, *inst, 1, 1, CHAN_Y ); - emit_mul( func, 0, 1 ); - STORE( func, *inst, 0, 0, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - FETCH( func, *inst, 0, 0, CHAN_Z ); - STORE( func, *inst, 0, 0, CHAN_Z ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - FETCH( func, *inst, 0, 1, CHAN_W ); - STORE( func, *inst, 0, 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MIN: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_minps( - func, - make_xmm( 0 ), - make_xmm( 1 ) ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_MAX: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_maxps( - func, - make_xmm( 0 ), - make_xmm( 1 ) ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - emit_setcc( func, inst, cc_LessThan ); - break; - - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - emit_setcc( func, inst, cc_NotLessThan ); - break; - - case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - FETCH( func, *inst, 2, 2, chan_index ); - emit_mul( func, 0, 1 ); - emit_add( func, 0, 2 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_SUB: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_sub( func, 0, 1 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - FETCH( func, *inst, 2, 2, chan_index ); - emit_sub( func, 1, 2 ); - emit_mul( func, 0, 1 ); - emit_add( func, 0, 2 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_CND: - return 0; - break; - - case TGSI_OPCODE_CND0: - return 0; - break; - - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - return 0; - break; - - case TGSI_OPCODE_INDEX: - return 0; - break; - - case TGSI_OPCODE_NEGATE: - return 0; - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_frc( func, 0 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_CLAMP: - return 0; - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_flr( func, 0 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_ROUND: - return 0; - break; - - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_ex2( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_lg2( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_pow( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( func, *inst, 1, 1, CHAN_Z ); - FETCH( func, *inst, 3, 0, CHAN_Z ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - FETCH( func, *inst, 4, 1, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - emit_MOV( func, 2, 0 ); - emit_mul( func, 2, 1 ); - emit_MOV( func, 5, 3 ); - emit_mul( func, 5, 4 ); - emit_sub( func, 2, 5 ); - STORE( func, *inst, 2, 0, CHAN_X ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 2, 1, CHAN_X ); - FETCH( func, *inst, 5, 0, CHAN_X ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - emit_mul( func, 3, 2 ); - emit_mul( func, 1, 5 ); - emit_sub( func, 3, 1 ); - STORE( func, *inst, 3, 0, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - emit_mul( func, 5, 4 ); - emit_mul( func, 0, 2 ); - emit_sub( func, 5, 0 ); - STORE( func, *inst, 5, 0, CHAN_Z ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); - STORE( func, *inst, 0, 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MULTIPLYMATRIX: - return 0; - break; - - case TGSI_OPCODE_ABS: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_abs( func, 0) ; - - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_RCC: - return 0; - break; - - case TGSI_OPCODE_DPH: - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_mul( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Y ); - FETCH( func, *inst, 2, 1, CHAN_Y ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Z ); - FETCH( func, *inst, 2, 1, CHAN_Z ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 1, CHAN_W ); - emit_add( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_COS: - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_cos( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDX: - return 0; - break; - - case TGSI_OPCODE_DDY: - return 0; - break; - - case TGSI_OPCODE_KIL: - emit_kil( func, &inst->FullSrcRegisters[0] ); - break; - - case TGSI_OPCODE_PK2H: - return 0; - break; - - case TGSI_OPCODE_PK2US: - return 0; - break; - - case TGSI_OPCODE_PK4B: - return 0; - break; - - case TGSI_OPCODE_PK4UB: - return 0; - break; - - case TGSI_OPCODE_RFL: - return 0; - break; - - case TGSI_OPCODE_SEQ: - return 0; - break; - - case TGSI_OPCODE_SFL: - return 0; - break; - - case TGSI_OPCODE_SGT: - return 0; - break; - - case TGSI_OPCODE_SIN: - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_sin( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLE: - return 0; - break; - - case TGSI_OPCODE_SNE: - return 0; - break; - - case TGSI_OPCODE_STR: - return 0; - break; - - case TGSI_OPCODE_TEX: - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_TXD: - return 0; - break; - - case TGSI_OPCODE_UP2H: - return 0; - break; - - case TGSI_OPCODE_UP2US: - return 0; - break; - - case TGSI_OPCODE_UP4B: - return 0; - break; - - case TGSI_OPCODE_UP4UB: - return 0; - break; - - case TGSI_OPCODE_X2D: - return 0; - break; - - case TGSI_OPCODE_ARA: - return 0; - break; - - case TGSI_OPCODE_ARR: - return 0; - break; - - case TGSI_OPCODE_BRA: - return 0; - break; - - case TGSI_OPCODE_CAL: - return 0; - break; - - case TGSI_OPCODE_RET: - case TGSI_OPCODE_END: -#ifdef WIN32 - emit_retw( func, 16 ); -#else - emit_ret( func ); -#endif - break; - - case TGSI_OPCODE_SSG: - return 0; - break; - - case TGSI_OPCODE_CMP: - emit_cmp (func, inst); - break; - - case TGSI_OPCODE_SCS: - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_cos( func, 0 ); - STORE( func, *inst, 0, 0, CHAN_X ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - emit_sin( func, 0 ); - STORE( func, *inst, 0, 0, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - FETCH( func, *inst, 0, TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ); - STORE( func, *inst, 0, 0, CHAN_Z ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); - STORE( func, *inst, 0, 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_TXB: - return 0; - break; - - case TGSI_OPCODE_NRM: - return 0; - break; - - case TGSI_OPCODE_DIV: - return 0; - break; - - case TGSI_OPCODE_DP2: - return 0; - break; - - case TGSI_OPCODE_TXL: - return 0; - break; - - case TGSI_OPCODE_BRK: - return 0; - break; - - case TGSI_OPCODE_IF: - return 0; - break; - - case TGSI_OPCODE_LOOP: - return 0; - break; - - case TGSI_OPCODE_REP: - return 0; - break; - - case TGSI_OPCODE_ELSE: - return 0; - break; - - case TGSI_OPCODE_ENDIF: - return 0; - break; - - case TGSI_OPCODE_ENDLOOP: - return 0; - break; - - case TGSI_OPCODE_ENDREP: - return 0; - break; - - case TGSI_OPCODE_PUSHA: - return 0; - break; - - case TGSI_OPCODE_POPA: - return 0; - break; - - case TGSI_OPCODE_CEIL: - return 0; - break; - - case TGSI_OPCODE_I2F: - return 0; - break; - - case TGSI_OPCODE_NOT: - return 0; - break; - - case TGSI_OPCODE_TRUNC: - return 0; - break; - - case TGSI_OPCODE_SHL: - return 0; - break; - - case TGSI_OPCODE_SHR: - return 0; - break; - - case TGSI_OPCODE_AND: - return 0; - break; - - case TGSI_OPCODE_OR: - return 0; - break; - - case TGSI_OPCODE_MOD: - return 0; - break; - - case TGSI_OPCODE_XOR: - return 0; - break; - - case TGSI_OPCODE_SAD: - return 0; - break; - - case TGSI_OPCODE_TXF: - return 0; - break; - - case TGSI_OPCODE_TXQ: - return 0; - break; - - case TGSI_OPCODE_CONT: - return 0; - break; - - case TGSI_OPCODE_EMIT: - return 0; - break; - - case TGSI_OPCODE_ENDPRIM: - return 0; - break; - - default: - return 0; - } - - return 1; -} - -static void -emit_declaration( - struct x86_function *func, - struct tgsi_full_declaration *decl ) -{ - if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask; - unsigned i, j; - - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - /* Do not touch WPOS.xy */ - if( first == 0 ) { - mask &= ~TGSI_WRITEMASK_XY; - if( mask == TGSI_WRITEMASK_NONE ) { - first++; - } - } - - for( i = first; i <= last; i++ ) { - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - switch( decl->Interpolation.Interpolate ) { - case TGSI_INTERPOLATE_CONSTANT: - emit_coef_a0( func, 0, i, j ); - emit_inputs( func, 0, i, j ); - break; - - case TGSI_INTERPOLATE_LINEAR: - emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); - emit_coef_dadx( func, 1, i, j ); - emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); - emit_coef_dady( func, 3, i, j ); - emit_mul( func, 0, 1 ); /* x * dadx */ - emit_coef_a0( func, 4, i, j ); - emit_mul( func, 2, 3 ); /* y * dady */ - emit_add( func, 0, 4 ); /* x * dadx + a0 */ - emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ - emit_inputs( func, 0, i, j ); - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); - emit_coef_dadx( func, 1, i, j ); - emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); - emit_coef_dady( func, 3, i, j ); - emit_mul( func, 0, 1 ); /* x * dadx */ - emit_inputf( func, 4, 0, TGSI_SWIZZLE_W ); - emit_coef_a0( func, 5, i, j ); - emit_rcp( func, 4, 4 ); /* 1.0 / w */ - emit_mul( func, 2, 3 ); /* y * dady */ - emit_add( func, 0, 5 ); /* x * dadx + a0 */ - emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ - emit_mul( func, 0, 4 ); /* (x * dadx + y * dady + a0) / w */ - emit_inputs( func, 0, i, j ); - break; - - default: - assert( 0 ); - break; - } - } - } - } - } -} - -unsigned -tgsi_emit_sse2( - struct tgsi_token *tokens, - struct x86_function *func ) -{ - struct tgsi_parse_context parse; - unsigned ok = 1; - - DUMP_START(); - - func->csr = func->store; - - emit_mov( - func, - get_input_base(), - get_argument( 0 ) ); - emit_mov( - func, - get_output_base(), - get_argument( 1 ) ); - emit_mov( - func, - get_const_base(), - get_argument( 2 ) ); - emit_mov( - func, - get_temp_base(), - get_argument( 3 ) ); - - tgsi_parse_init( &parse, tokens ); - - while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - ok = emit_instruction( - func, - &parse.FullToken.FullInstruction ); - - if (!ok) { - debug_printf("failed to translate tgsi opcode %d\n", - parse.FullToken.FullInstruction.Instruction.Opcode ); - } - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - /* XXX implement this */ - ok = 0; - debug_printf("failed to emit immediate value\n"); - break; - - default: - assert( 0 ); - ok = 0; - break; - } - } - - tgsi_parse_free( &parse ); - - DUMP_END(); - - return ok; -} - -/** - * Fragment shaders are responsible for interpolating shader inputs. Because on - * x86 we have only 4 GP registers, and here we have 5 shader arguments (input, - * output, const, temp and coef), the code is split into two phases -- - * DECLARATION and INSTRUCTION phase. - * GP register holding the output argument is aliased with the coeff argument, - * as outputs are not needed in the DECLARATION phase. - */ -unsigned -tgsi_emit_sse2_fs( - struct tgsi_token *tokens, - struct x86_function *func ) -{ - struct tgsi_parse_context parse; - boolean instruction_phase = FALSE; - - DUMP_START(); - - func->csr = func->store; - - /* DECLARATION phase, do not load output argument. */ - emit_mov( - func, - get_input_base(), - get_argument( 0 ) ); - emit_mov( - func, - get_const_base(), - get_argument( 2 ) ); - emit_mov( - func, - get_temp_base(), - get_argument( 3 ) ); - emit_mov( - func, - get_coef_base(), - get_argument( 4 ) ); - - tgsi_parse_init( &parse, tokens ); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - emit_declaration( - func, - &parse.FullToken.FullDeclaration ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - if( !instruction_phase ) { - /* INSTRUCTION phase, overwrite coeff with output. */ - instruction_phase = TRUE; - emit_mov( - func, - get_output_base(), - get_argument( 1 ) ); - } - emit_instruction( - func, - &parse.FullToken.FullInstruction ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - /* XXX implement this */ - assert(0); - break; - - default: - assert( 0 ); - } - } - - tgsi_parse_free( &parse ); - - DUMP_END(); - - return 1; -} - -#endif /* i386 */ diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h b/src/mesa/pipe/tgsi/exec/tgsi_sse2.h deleted file mode 100755 index 9bee371766..0000000000 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.h +++ /dev/null @@ -1,26 +0,0 @@ -#if !defined TGSI_SSE2_H -#define TGSI_SSE2_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_token; -struct x86_function; - -unsigned -tgsi_emit_sse2( - struct tgsi_token *tokens, - struct x86_function *function ); - -unsigned -tgsi_emit_sse2_fs( - struct tgsi_token *tokens, - struct x86_function *function ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_SSE2_H - diff --git a/src/mesa/pipe/tgsi/util/tgsi_build.c b/src/mesa/pipe/tgsi/util/tgsi_build.c deleted file mode 100644 index a00ff1c2a5..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_build.c +++ /dev/null @@ -1,1371 +0,0 @@ -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_build.h" -#include "tgsi_parse.h" - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ) -{ - struct tgsi_version version; - - version.MajorVersion = 1; - version.MinorVersion = 1; - version.Padding = 0; - - return version; -} - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ) -{ - struct tgsi_header header; - - header.HeaderSize = 1; - header.BodySize = 0; - - return header; -} - -static void -header_headersize_grow( struct tgsi_header *header ) -{ - assert( header->HeaderSize < 0xFF ); - assert( header->BodySize == 0 ); - - header->HeaderSize++; -} - -static void -header_bodysize_grow( struct tgsi_header *header ) -{ - assert( header->BodySize < 0xFFFFFF ); - - header->BodySize++; -} - -struct tgsi_processor -tgsi_default_processor( void ) -{ - struct tgsi_processor processor; - - processor.Processor = TGSI_PROCESSOR_FRAGMENT; - processor.Padding = 0; - - return processor; -} - -struct tgsi_processor -tgsi_build_processor( - unsigned type, - struct tgsi_header *header ) -{ - struct tgsi_processor processor; - - processor = tgsi_default_processor(); - processor.Processor = type; - - header_headersize_grow( header ); - - return processor; -} - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ) -{ - struct tgsi_declaration declaration; - - declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; - declaration.Size = 1; - declaration.File = TGSI_FILE_NULL; - declaration.Declare = TGSI_DECLARE_RANGE; - declaration.UsageMask = TGSI_WRITEMASK_XYZW; - declaration.Interpolate = 0; - declaration.Semantic = 0; - declaration.Padding = 0; - declaration.Extended = 0; - - return declaration; -} - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned declare, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ) -{ - struct tgsi_declaration declaration; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( declare <= TGSI_DECLARE_MASK ); - - declaration = tgsi_default_declaration(); - declaration.File = file; - declaration.Declare = declare; - declaration.UsageMask = usage_mask; - declaration.Interpolate = interpolate; - declaration.Semantic = semantic; - - header_bodysize_grow( header ); - - return declaration; -} - -static void -declaration_grow( - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - assert( declaration->Size < 0xFF ); - - declaration->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ) -{ - struct tgsi_full_declaration full_declaration; - - full_declaration.Declaration = tgsi_default_declaration(); - full_declaration.Interpolation = tgsi_default_declaration_interpolation(); - full_declaration.Semantic = tgsi_default_declaration_semantic(); - - return full_declaration; -} - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0; - struct tgsi_declaration *declaration; - - if( maxsize <= size ) - return 0; - declaration = (struct tgsi_declaration *) &tokens[size]; - size++; - - *declaration = tgsi_build_declaration( - full_decl->Declaration.File, - full_decl->Declaration.Declare, - full_decl->Declaration.UsageMask, - full_decl->Declaration.Interpolate, - full_decl->Declaration.Semantic, - header ); - - switch( full_decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - { - struct tgsi_declaration_range *dr; - - if( maxsize <= size ) - return 0; - dr = (struct tgsi_declaration_range *) &tokens[size]; - size++; - - *dr = tgsi_build_declaration_range( - full_decl->u.DeclarationRange.First, - full_decl->u.DeclarationRange.Last, - declaration, - header ); - break; - } - - case TGSI_DECLARE_MASK: - { - struct tgsi_declaration_mask *dm; - - if( maxsize <= size ) - return 0; - dm = (struct tgsi_declaration_mask *) &tokens[size]; - size++; - - *dm = tgsi_build_declaration_mask( - full_decl->u.DeclarationMask.Mask, - declaration, - header ); - break; - } - - default: - assert( 0 ); - } - - if( full_decl->Declaration.Interpolate ) { - struct tgsi_declaration_interpolation *di; - - if( maxsize <= size ) - return 0; - di = (struct tgsi_declaration_interpolation *) &tokens[size]; - size++; - - *di = tgsi_build_declaration_interpolation( - full_decl->Interpolation.Interpolate, - declaration, - header ); - } - - if( full_decl->Declaration.Semantic ) { - struct tgsi_declaration_semantic *ds; - - if( maxsize <= size ) - return 0; - ds = (struct tgsi_declaration_semantic *) &tokens[size]; - size++; - - *ds = tgsi_build_declaration_semantic( - full_decl->Semantic.SemanticName, - full_decl->Semantic.SemanticIndex, - declaration, - header ); - } - - return size; -} - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_range declaration_range; - - assert( last >= first ); - assert( last <= 0xFFFF ); - - declaration_range.First = first; - declaration_range.Last = last; - - declaration_grow( declaration, header ); - - return declaration_range; -} - -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_mask declaration_mask; - - declaration_mask.Mask = mask; - - declaration_grow( declaration, header ); - - return declaration_mask; -} - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ) -{ - struct tgsi_declaration_interpolation di; - - di.Interpolate = TGSI_INTERPOLATE_CONSTANT; - di.Padding = 0; - - return di; -} - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_interpolation di; - - assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); - - di = tgsi_default_declaration_interpolation(); - di.Interpolate = interpolate; - - declaration_grow( declaration, header ); - - return di; -} - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ) -{ - struct tgsi_declaration_semantic ds; - - ds.SemanticName = TGSI_SEMANTIC_POSITION; - ds.SemanticIndex = 0; - ds.Padding = 0; - - return ds; -} - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_semantic ds; - - assert( semantic_name <= TGSI_SEMANTIC_COUNT ); - assert( semantic_index <= 0xFFFF ); - - ds = tgsi_default_declaration_semantic(); - ds.SemanticName = semantic_name; - ds.SemanticIndex = semantic_index; - - declaration_grow( declaration, header ); - - return ds; -} - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ) -{ - struct tgsi_immediate immediate; - - immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; - immediate.Size = 1; - immediate.DataType = TGSI_IMM_FLOAT32; - immediate.Padding = 0; - immediate.Extended = 0; - - return immediate; -} - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ) -{ - struct tgsi_immediate immediate; - - immediate = tgsi_default_immediate(); - - header_bodysize_grow( header ); - - return immediate; -} - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ) -{ - struct tgsi_full_immediate fullimm; - - fullimm.Immediate = tgsi_default_immediate(); - fullimm.u.Pointer = (void *) 0; - - return fullimm; -} - -static void -immediate_grow( - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - assert( immediate->Size < 0xFF ); - - immediate->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - float value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - struct tgsi_immediate_float32 immediate_float32; - - immediate_float32.Float = value; - - immediate_grow( immediate, header ); - - return immediate_float32; -} - -unsigned -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0, i; - struct tgsi_immediate *immediate; - - if( maxsize <= size ) - return 0; - immediate = (struct tgsi_immediate *) &tokens[size]; - size++; - - *immediate = tgsi_build_immediate( header ); - - for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { - struct tgsi_immediate_float32 *if32; - - if( maxsize <= size ) - return 0; - if32 = (struct tgsi_immediate_float32 *) &tokens[size]; - size++; - - *if32 = tgsi_build_immediate_float32( - full_imm->u.ImmediateFloat32[i].Float, - immediate, - header ); - } - - return size; -} - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ) -{ - struct tgsi_instruction instruction; - - instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; - instruction.Size = 1; - instruction.Opcode = TGSI_OPCODE_MOV; - instruction.Saturate = TGSI_SAT_NONE; - instruction.NumDstRegs = 1; - instruction.NumSrcRegs = 1; - instruction.Padding = 0; - instruction.Extended = 0; - - return instruction; -} - -struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ) -{ - struct tgsi_instruction instruction; - - assert (opcode <= TGSI_OPCODE_LAST); - assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); - assert (num_dst_regs <= 3); - assert (num_src_regs <= 15); - - instruction = tgsi_default_instruction(); - instruction.Opcode = opcode; - instruction.Saturate = saturate; - instruction.NumDstRegs = num_dst_regs; - instruction.NumSrcRegs = num_src_regs; - - header_bodysize_grow( header ); - - return instruction; -} - -static void -instruction_grow( - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - assert (instruction->Size < 0xFF); - - instruction->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ) -{ - struct tgsi_full_instruction full_instruction; - unsigned i; - - full_instruction.Instruction = tgsi_default_instruction(); - full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); - full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); - full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); - for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { - full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); - } - for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { - full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); - } - - return full_instruction; -} - -unsigned -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0; - unsigned i; - struct tgsi_instruction *instruction; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - instruction = (struct tgsi_instruction *) &tokens[size]; - size++; - - *instruction = tgsi_build_instruction( - full_inst->Instruction.Opcode, - full_inst->Instruction.Saturate, - full_inst->Instruction.NumDstRegs, - full_inst->Instruction.NumSrcRegs, - header ); - prev_token = (struct tgsi_token *) instruction; - - if( tgsi_compare_instruction_ext_nv( - full_inst->InstructionExtNv, - tgsi_default_instruction_ext_nv() ) ) { - struct tgsi_instruction_ext_nv *instruction_ext_nv; - - if( maxsize <= size ) - return 0; - instruction_ext_nv = - (struct tgsi_instruction_ext_nv *) &tokens[size]; - size++; - - *instruction_ext_nv = tgsi_build_instruction_ext_nv( - full_inst->InstructionExtNv.Precision, - full_inst->InstructionExtNv.CondDstIndex, - full_inst->InstructionExtNv.CondFlowIndex, - full_inst->InstructionExtNv.CondMask, - full_inst->InstructionExtNv.CondSwizzleX, - full_inst->InstructionExtNv.CondSwizzleY, - full_inst->InstructionExtNv.CondSwizzleZ, - full_inst->InstructionExtNv.CondSwizzleW, - full_inst->InstructionExtNv.CondDstUpdate, - full_inst->InstructionExtNv.CondFlowEnable, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_nv; - } - - if( tgsi_compare_instruction_ext_label( - full_inst->InstructionExtLabel, - tgsi_default_instruction_ext_label() ) ) { - struct tgsi_instruction_ext_label *instruction_ext_label; - - if( maxsize <= size ) - return 0; - instruction_ext_label = - (struct tgsi_instruction_ext_label *) &tokens[size]; - size++; - - *instruction_ext_label = tgsi_build_instruction_ext_label( - full_inst->InstructionExtLabel.Label, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_label; - } - - if( tgsi_compare_instruction_ext_texture( - full_inst->InstructionExtTexture, - tgsi_default_instruction_ext_texture() ) ) { - struct tgsi_instruction_ext_texture *instruction_ext_texture; - - if( maxsize <= size ) - return 0; - instruction_ext_texture = - (struct tgsi_instruction_ext_texture *) &tokens[size]; - size++; - - *instruction_ext_texture = tgsi_build_instruction_ext_texture( - full_inst->InstructionExtTexture.Texture, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_texture; - } - - for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { - const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; - struct tgsi_dst_register *dst_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - dst_register = (struct tgsi_dst_register *) &tokens[size]; - size++; - - *dst_register = tgsi_build_dst_register( - reg->DstRegister.File, - reg->DstRegister.WriteMask, - reg->DstRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register; - - if( tgsi_compare_dst_register_ext_concode( - reg->DstRegisterExtConcode, - tgsi_default_dst_register_ext_concode() ) ) { - struct tgsi_dst_register_ext_concode *dst_register_ext_concode; - - if( maxsize <= size ) - return 0; - dst_register_ext_concode = - (struct tgsi_dst_register_ext_concode *) &tokens[size]; - size++; - - *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( - reg->DstRegisterExtConcode.CondMask, - reg->DstRegisterExtConcode.CondSwizzleX, - reg->DstRegisterExtConcode.CondSwizzleY, - reg->DstRegisterExtConcode.CondSwizzleZ, - reg->DstRegisterExtConcode.CondSwizzleW, - reg->DstRegisterExtConcode.CondSrcIndex, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_concode; - } - - if( tgsi_compare_dst_register_ext_modulate( - reg->DstRegisterExtModulate, - tgsi_default_dst_register_ext_modulate() ) ) { - struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; - - if( maxsize <= size ) - return 0; - dst_register_ext_modulate = - (struct tgsi_dst_register_ext_modulate *) &tokens[size]; - size++; - - *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( - reg->DstRegisterExtModulate.Modulate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_modulate; - } - } - - for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { - const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; - struct tgsi_src_register *src_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - src_register = (struct tgsi_src_register *) &tokens[size]; - size++; - - *src_register = tgsi_build_src_register( - reg->SrcRegister.File, - reg->SrcRegister.SwizzleX, - reg->SrcRegister.SwizzleY, - reg->SrcRegister.SwizzleZ, - reg->SrcRegister.SwizzleW, - reg->SrcRegister.Negate, - reg->SrcRegister.Indirect, - reg->SrcRegister.Dimension, - reg->SrcRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register; - - if( tgsi_compare_src_register_ext_swz( - reg->SrcRegisterExtSwz, - tgsi_default_src_register_ext_swz() ) ) { - struct tgsi_src_register_ext_swz *src_register_ext_swz; - - if( maxsize <= size ) - return 0; - src_register_ext_swz = - (struct tgsi_src_register_ext_swz *) &tokens[size]; - size++; - - *src_register_ext_swz = tgsi_build_src_register_ext_swz( - reg->SrcRegisterExtSwz.ExtSwizzleX, - reg->SrcRegisterExtSwz.ExtSwizzleY, - reg->SrcRegisterExtSwz.ExtSwizzleZ, - reg->SrcRegisterExtSwz.ExtSwizzleW, - reg->SrcRegisterExtSwz.NegateX, - reg->SrcRegisterExtSwz.NegateY, - reg->SrcRegisterExtSwz.NegateZ, - reg->SrcRegisterExtSwz.NegateW, - reg->SrcRegisterExtSwz.ExtDivide, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_swz; - } - - if( tgsi_compare_src_register_ext_mod( - reg->SrcRegisterExtMod, - tgsi_default_src_register_ext_mod() ) ) { - struct tgsi_src_register_ext_mod *src_register_ext_mod; - - if( maxsize <= size ) - return 0; - src_register_ext_mod = - (struct tgsi_src_register_ext_mod *) &tokens[size]; - size++; - - *src_register_ext_mod = tgsi_build_src_register_ext_mod( - reg->SrcRegisterExtMod.Complement, - reg->SrcRegisterExtMod.Bias, - reg->SrcRegisterExtMod.Scale2X, - reg->SrcRegisterExtMod.Absolute, - reg->SrcRegisterExtMod.Negate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_mod; - } - - if( reg->SrcRegister.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterInd.File, - reg->SrcRegisterInd.SwizzleX, - reg->SrcRegisterInd.SwizzleY, - reg->SrcRegisterInd.SwizzleZ, - reg->SrcRegisterInd.SwizzleW, - reg->SrcRegisterInd.Negate, - reg->SrcRegisterInd.Indirect, - reg->SrcRegisterInd.Dimension, - reg->SrcRegisterInd.Index, - instruction, - header ); - } - - if( reg->SrcRegister.Dimension ) { - struct tgsi_dimension *dim; - - assert( !reg->SrcRegisterDim.Dimension ); - - if( maxsize <= size ) - return 0; - dim = (struct tgsi_dimension *) &tokens[size]; - size++; - - *dim = tgsi_build_dimension( - reg->SrcRegisterDim.Indirect, - reg->SrcRegisterDim.Index, - instruction, - header ); - - if( reg->SrcRegisterDim.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterDimInd.File, - reg->SrcRegisterDimInd.SwizzleX, - reg->SrcRegisterDimInd.SwizzleY, - reg->SrcRegisterDimInd.SwizzleZ, - reg->SrcRegisterDimInd.SwizzleW, - reg->SrcRegisterDimInd.Negate, - reg->SrcRegisterDimInd.Indirect, - reg->SrcRegisterDimInd.Dimension, - reg->SrcRegisterDimInd.Index, - instruction, - header ); - } - } - } - - return size; -} - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; - instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; - instruction_ext_nv.CondDstIndex = 0; - instruction_ext_nv.CondFlowIndex = 0; - instruction_ext_nv.CondMask = TGSI_CC_TR; - instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; - instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; - instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; - instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; - instruction_ext_nv.CondDstUpdate = 0; - instruction_ext_nv.CondFlowEnable = 0; - instruction_ext_nv.Padding = 0; - instruction_ext_nv.Extended = 0; - - return instruction_ext_nv; -} - -union token_u32 -{ - unsigned u32; -}; - -unsigned -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - unsigned precision, - unsigned cond_dst_index, - unsigned cond_flow_index, - unsigned cond_mask, - unsigned cond_swizzle_x, - unsigned cond_swizzle_y, - unsigned cond_swizzle_z, - unsigned cond_swizzle_w, - unsigned cond_dst_update, - unsigned cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv = tgsi_default_instruction_ext_nv(); - instruction_ext_nv.Precision = precision; - instruction_ext_nv.CondDstIndex = cond_dst_index; - instruction_ext_nv.CondFlowIndex = cond_flow_index; - instruction_ext_nv.CondMask = cond_mask; - instruction_ext_nv.CondSwizzleX = cond_swizzle_x; - instruction_ext_nv.CondSwizzleY = cond_swizzle_y; - instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; - instruction_ext_nv.CondSwizzleW = cond_swizzle_w; - instruction_ext_nv.CondDstUpdate = cond_dst_update; - instruction_ext_nv.CondFlowEnable = cond_flow_update; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_nv; -} - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; - instruction_ext_label.Label = 0; - instruction_ext_label.Padding = 0; - instruction_ext_label.Extended = 0; - - return instruction_ext_label; -} - -unsigned -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - unsigned label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label = tgsi_default_instruction_ext_label(); - instruction_ext_label.Label = label; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_label; -} - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; - instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; - instruction_ext_texture.Padding = 0; - instruction_ext_texture.Extended = 0; - - return instruction_ext_texture; -} - -unsigned -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - unsigned texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture = tgsi_default_instruction_ext_texture(); - instruction_ext_texture.Texture = texture; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_texture; -} - -struct tgsi_src_register -tgsi_default_src_register( void ) -{ - struct tgsi_src_register src_register; - - src_register.File = TGSI_FILE_NULL; - src_register.SwizzleX = TGSI_SWIZZLE_X; - src_register.SwizzleY = TGSI_SWIZZLE_Y; - src_register.SwizzleZ = TGSI_SWIZZLE_Z; - src_register.SwizzleW = TGSI_SWIZZLE_W; - src_register.Negate = 0; - src_register.Indirect = 0; - src_register.Dimension = 0; - src_register.Index = 0; - src_register.Extended = 0; - - return src_register; -} - -struct tgsi_src_register -tgsi_build_src_register( - unsigned file, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - unsigned negate, - unsigned indirect, - unsigned dimension, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register src_register; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( swizzle_x <= TGSI_SWIZZLE_W ); - assert( swizzle_y <= TGSI_SWIZZLE_W ); - assert( swizzle_z <= TGSI_SWIZZLE_W ); - assert( swizzle_w <= TGSI_SWIZZLE_W ); - assert( negate <= 1 ); - assert( index >= -0x8000 && index <= 0x7FFF ); - - src_register = tgsi_default_src_register(); - src_register.File = file; - src_register.SwizzleX = swizzle_x; - src_register.SwizzleY = swizzle_y; - src_register.SwizzleZ = swizzle_z; - src_register.SwizzleW = swizzle_w; - src_register.Negate = negate; - src_register.Indirect = indirect; - src_register.Dimension = dimension; - src_register.Index = index; - - instruction_grow( instruction, header ); - - return src_register; -} - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ) -{ - struct tgsi_full_src_register full_src_register; - - full_src_register.SrcRegister = tgsi_default_src_register(); - full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); - full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); - full_src_register.SrcRegisterInd = tgsi_default_src_register(); - full_src_register.SrcRegisterDim = tgsi_default_dimension(); - full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); - - return full_src_register; -} - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; - src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; - src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; - src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; - src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; - src_register_ext_swz.NegateX = 0; - src_register_ext_swz.NegateY = 0; - src_register_ext_swz.NegateZ = 0; - src_register_ext_swz.NegateW = 0; - src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; - src_register_ext_swz.Padding = 0; - src_register_ext_swz.Extended = 0; - - return src_register_ext_swz; -} - -unsigned -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - unsigned ext_swizzle_x, - unsigned ext_swizzle_y, - unsigned ext_swizzle_z, - unsigned ext_swizzle_w, - unsigned negate_x, - unsigned negate_y, - unsigned negate_z, - unsigned negate_w, - unsigned ext_divide, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE ); - assert( negate_x <= 1 ); - assert( negate_y <= 1 ); - assert( negate_z <= 1 ); - assert( negate_w <= 1 ); - assert( ext_divide <= TGSI_EXTSWIZZLE_ONE ); - - src_register_ext_swz = tgsi_default_src_register_ext_swz(); - src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; - src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; - src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; - src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; - src_register_ext_swz.NegateX = negate_x; - src_register_ext_swz.NegateY = negate_y; - src_register_ext_swz.NegateZ = negate_z; - src_register_ext_swz.NegateW = negate_w; - src_register_ext_swz.ExtDivide = ext_divide; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_swz; -} - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; - src_register_ext_mod.Complement = 0; - src_register_ext_mod.Bias = 0; - src_register_ext_mod.Scale2X = 0; - src_register_ext_mod.Absolute = 0; - src_register_ext_mod.Negate = 0; - src_register_ext_mod.Padding = 0; - src_register_ext_mod.Extended = 0; - - return src_register_ext_mod; -} - -unsigned -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - unsigned complement, - unsigned bias, - unsigned scale_2x, - unsigned absolute, - unsigned negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - assert( complement <= 1 ); - assert( bias <= 1 ); - assert( scale_2x <= 1 ); - assert( absolute <= 1 ); - assert( negate <= 1 ); - - src_register_ext_mod = tgsi_default_src_register_ext_mod(); - src_register_ext_mod.Complement = complement; - src_register_ext_mod.Bias = bias; - src_register_ext_mod.Scale2X = scale_2x; - src_register_ext_mod.Absolute = absolute; - src_register_ext_mod.Negate = negate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_mod; -} - -struct tgsi_dimension -tgsi_default_dimension( void ) -{ - struct tgsi_dimension dimension; - - dimension.Indirect = 0; - dimension.Dimension = 0; - dimension.Padding = 0; - dimension.Index = 0; - dimension.Extended = 0; - - return dimension; -} - -struct tgsi_dimension -tgsi_build_dimension( - unsigned indirect, - unsigned index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dimension dimension; - - dimension = tgsi_default_dimension(); - dimension.Indirect = indirect; - dimension.Index = index; - - instruction_grow( instruction, header ); - - return dimension; -} - -struct tgsi_dst_register -tgsi_default_dst_register( void ) -{ - struct tgsi_dst_register dst_register; - - dst_register.File = TGSI_FILE_NULL; - dst_register.WriteMask = TGSI_WRITEMASK_XYZW; - dst_register.Indirect = 0; - dst_register.Dimension = 0; - dst_register.Index = 0; - dst_register.Padding = 0; - dst_register.Extended = 0; - - return dst_register; -} - -struct tgsi_dst_register -tgsi_build_dst_register( - unsigned file, - unsigned mask, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register dst_register; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( mask <= TGSI_WRITEMASK_XYZW ); - assert( index >= -32768 && index <= 32767 ); - - dst_register = tgsi_default_dst_register(); - dst_register.File = file; - dst_register.WriteMask = mask; - dst_register.Index = index; - - instruction_grow( instruction, header ); - - return dst_register; -} - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ) -{ - struct tgsi_full_dst_register full_dst_register; - - full_dst_register.DstRegister = tgsi_default_dst_register(); - full_dst_register.DstRegisterExtConcode = - tgsi_default_dst_register_ext_concode(); - full_dst_register.DstRegisterExtModulate = - tgsi_default_dst_register_ext_modulate(); - - return full_dst_register; -} - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; - dst_register_ext_concode.CondMask = TGSI_CC_TR; - dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; - dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; - dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; - dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; - dst_register_ext_concode.CondSrcIndex = 0; - dst_register_ext_concode.Padding = 0; - dst_register_ext_concode.Extended = 0; - - return dst_register_ext_concode; -} - -unsigned -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - unsigned cc, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - int index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - assert( cc <= TGSI_CC_FL ); - assert( swizzle_x <= TGSI_SWIZZLE_W ); - assert( swizzle_y <= TGSI_SWIZZLE_W ); - assert( swizzle_z <= TGSI_SWIZZLE_W ); - assert( swizzle_w <= TGSI_SWIZZLE_W ); - assert( index >= -32768 && index <= 32767 ); - - dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); - dst_register_ext_concode.CondMask = cc; - dst_register_ext_concode.CondSwizzleX = swizzle_x; - dst_register_ext_concode.CondSwizzleY = swizzle_y; - dst_register_ext_concode.CondSwizzleZ = swizzle_z; - dst_register_ext_concode.CondSwizzleW = swizzle_w; - dst_register_ext_concode.CondSrcIndex = index; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_concode; -} - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; - dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; - dst_register_ext_modulate.Padding = 0; - dst_register_ext_modulate.Extended = 0; - - return dst_register_ext_modulate; -} - -unsigned -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - unsigned modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - assert( modulate <= TGSI_MODULATE_EIGHTH ); - - dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); - dst_register_ext_modulate.Modulate = modulate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_modulate; -} - diff --git a/src/mesa/pipe/tgsi/util/tgsi_build.h b/src/mesa/pipe/tgsi/util/tgsi_build.h deleted file mode 100644 index 116c78abf3..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_build.h +++ /dev/null @@ -1,320 +0,0 @@ -#if !defined TGSI_BUILD_H -#define TGSI_BUILD_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ); - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ); - -struct tgsi_processor -tgsi_default_processor( void ); - -struct tgsi_processor -tgsi_build_processor( - unsigned processor, - struct tgsi_header *header ); - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ); - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned declare, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ); - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ); - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ); - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ); - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ); - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ); - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ); - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - float value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ); - -unsigned -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ); - -struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ); - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ); - -unsigned -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ); - -unsigned -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ); - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - unsigned precision, - unsigned cond_dst_index, - unsigned cond_flow_index, - unsigned cond_mask, - unsigned cond_swizzle_x, - unsigned cond_swizzle_y, - unsigned cond_swizzle_z, - unsigned cond_swizzle_w, - unsigned cond_dst_update, - unsigned cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ); - -unsigned -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ); - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - unsigned label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ); - -unsigned -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ); - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - unsigned texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register -tgsi_default_src_register( void ); - -struct tgsi_src_register -tgsi_build_src_register( - unsigned file, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - unsigned negate, - unsigned indirect, - unsigned dimension, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ); - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ); - -unsigned -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ); - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - unsigned ext_swizzle_x, - unsigned ext_swizzle_y, - unsigned ext_swizzle_z, - unsigned ext_swizzle_w, - unsigned negate_x, - unsigned negate_y, - unsigned negate_z, - unsigned negate_w, - unsigned ext_divide, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ); - -unsigned -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ); - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - unsigned complement, - unsigned bias, - unsigned scale_2x, - unsigned absolute, - unsigned negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dimension -tgsi_default_dimension( void ); - -struct tgsi_dimension -tgsi_build_dimension( - unsigned indirect, - unsigned index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register -tgsi_default_dst_register( void ); - -struct tgsi_dst_register -tgsi_build_dst_register( - unsigned file, - unsigned mask, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ); - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ); - -unsigned -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ); - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - unsigned cc, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - int index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ); - -unsigned -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ); - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - unsigned modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_BUILD_H - diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.c b/src/mesa/pipe/tgsi/util/tgsi_dump.c deleted file mode 100644 index b5c54847e0..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_dump.c +++ /dev/null @@ -1,1581 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_dump.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" - -struct gen_dump -{ - unsigned tabs; - void (* write)( - struct gen_dump *dump, - const void *data, - unsigned size ); -}; - -struct text_dump -{ - struct gen_dump base; - char *text; - unsigned length; - unsigned capacity; -}; - -static void -_text_dump_write( - struct gen_dump *dump, - const void *data, - unsigned size ) -{ - struct text_dump *td = (struct text_dump *) dump; - unsigned new_length = td->length + size; - - if( new_length >= td->capacity ) { - unsigned new_capacity = td->capacity; - - do { - if( new_capacity == 0 ) { - new_capacity = 256; - } - else { - new_capacity *= 2; - } - } while( new_length >= new_capacity ); - td->text = (char *) REALLOC( - td->text, - td->capacity, - new_capacity ); - td->capacity = new_capacity; - } - memcpy( - &td->text[td->length], - data, - size ); - td->length = new_length; - td->text[td->length] = '\0'; -} - -struct file_dump -{ - struct gen_dump base; - FILE *file; -}; - -static void -_file_dump_write( - struct gen_dump *dump, - const void *data, - unsigned size ) -{ - struct file_dump *fd = (struct file_dump *) dump; - -#if 0 - fwrite( data, 1, size, fd->file ); -#else - { - unsigned i; - - for (i = 0; i < size; i++ ) { - fprintf( fd->file, "%c", ((const char *) data)[i] ); - } - } -#endif -} - -static void -gen_dump_str( - struct gen_dump *dump, - const char *str ) -{ - unsigned i; - size_t len = strlen( str ); - - for (i = 0; i < len; i++) { - dump->write( dump, &str[i], 1 ); - if (str[i] == '\n') { - unsigned i; - - for (i = 0; i < dump->tabs; i++) { - dump->write( dump, " ", 4 ); - } - } - } -} - -static void -gen_dump_chr( - struct gen_dump *dump, - const char chr ) -{ - dump->write( dump, &chr, 1 ); -} - -static void -gen_dump_uix( - struct gen_dump *dump, - const unsigned ui ) -{ - char str[36]; - - sprintf( str, "0x%x", ui ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_uid( - struct gen_dump *dump, - const unsigned ui ) -{ - char str[16]; - - sprintf( str, "%u", ui ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_sid( - struct gen_dump *dump, - const int si ) -{ - char str[16]; - - sprintf( str, "%d", si ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_flt( - struct gen_dump *dump, - const float flt ) -{ - char str[48]; - - sprintf( str, "%10.4f", flt ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_enum( - struct gen_dump *dump, - const unsigned e, - const char **enums, - const unsigned enums_count ) -{ - if (e >= enums_count) { - gen_dump_uid( dump, e ); - } - else { - gen_dump_str( dump, enums[e] ); - } -} - -static void -gen_dump_tab( - struct gen_dump *dump ) -{ - ++dump->tabs; -} - -static void -gen_dump_untab( - struct gen_dump *dump ) -{ - assert( dump->tabs > 0 ); - - --dump->tabs; -} - -#define TXT(S) gen_dump_str( dump, S ) -#define CHR(C) gen_dump_chr( dump, C ) -#define UIX(I) gen_dump_uix( dump, I ) -#define UID(I) gen_dump_uid( dump, I ) -#define SID(I) gen_dump_sid( dump, I ) -#define FLT(F) gen_dump_flt( dump, F ) -#define TAB() gen_dump_tab( dump ) -#define UNT() gen_dump_untab( dump ) -#define ENM(E,ENUMS) gen_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) - -static const char *TGSI_PROCESSOR_TYPES[] = -{ - "PROCESSOR_FRAGMENT", - "PROCESSOR_VERTEX", - "PROCESSOR_GEOMETRY" -}; - -static const char *TGSI_PROCESSOR_TYPES_SHORT[] = -{ - "FRAG", - "VERT", - "GEOM" -}; - -static const char *TGSI_TOKEN_TYPES[] = -{ - "TOKEN_TYPE_DECLARATION", - "TOKEN_TYPE_IMMEDIATE", - "TOKEN_TYPE_INSTRUCTION" -}; - -static const char *TGSI_FILES[] = -{ - "FILE_NULL", - "FILE_CONSTANT", - "FILE_INPUT", - "FILE_OUTPUT", - "FILE_TEMPORARY", - "FILE_SAMPLER", - "FILE_ADDRESS", - "FILE_IMMEDIATE" -}; - -static const char *TGSI_FILES_SHORT[] = -{ - "NULL", - "CONST", - "IN", - "OUT", - "TEMP", - "SAMP", - "ADDR", - "IMM" -}; - -static const char *TGSI_DECLARES[] = -{ - "DECLARE_RANGE", - "DECLARE_MASK" -}; - -static const char *TGSI_INTERPOLATES[] = -{ - "INTERPOLATE_CONSTANT", - "INTERPOLATE_LINEAR", - "INTERPOLATE_PERSPECTIVE", - "INTERPOLATE_ATTRIB" -}; - -static const char *TGSI_INTERPOLATES_SHORT[] = -{ - "CONSTANT", - "LINEAR", - "PERSPECTIVE", - "ATTRIB" -}; - -static const char *TGSI_SEMANTICS[] = -{ - "SEMANTIC_POSITION", - "SEMANTIC_COLOR", - "SEMANTIC_BCOLOR", - "SEMANTIC_FOG", - "SEMANTIC_PSIZE", - "SEMANTIC_GENERIC," -}; - -static const char *TGSI_SEMANTICS_SHORT[] = -{ - "POSITION", - "COLOR", - "BCOLOR", - "FOG", - "PSIZE", - "GENERIC", -}; - -static const char *TGSI_IMMS[] = -{ - "IMM_FLOAT32" -}; - -static const char *TGSI_IMMS_SHORT[] = -{ - "FLT32" -}; - -static const char *TGSI_OPCODES[] = -{ - "OPCODE_ARL", - "OPCODE_MOV", - "OPCODE_LIT", - "OPCODE_RCP", - "OPCODE_RSQ", - "OPCODE_EXP", - "OPCODE_LOG", - "OPCODE_MUL", - "OPCODE_ADD", - "OPCODE_DP3", - "OPCODE_DP4", - "OPCODE_DST", - "OPCODE_MIN", - "OPCODE_MAX", - "OPCODE_SLT", - "OPCODE_SGE", - "OPCODE_MAD", - "OPCODE_SUB", - "OPCODE_LERP", - "OPCODE_CND", - "OPCODE_CND0", - "OPCODE_DOT2ADD", - "OPCODE_INDEX", - "OPCODE_NEGATE", - "OPCODE_FRAC", - "OPCODE_CLAMP", - "OPCODE_FLOOR", - "OPCODE_ROUND", - "OPCODE_EXPBASE2", - "OPCODE_LOGBASE2", - "OPCODE_POWER", - "OPCODE_CROSSPRODUCT", - "OPCODE_MULTIPLYMATRIX", - "OPCODE_ABS", - "OPCODE_RCC", - "OPCODE_DPH", - "OPCODE_COS", - "OPCODE_DDX", - "OPCODE_DDY", - "OPCODE_KILP", - "OPCODE_PK2H", - "OPCODE_PK2US", - "OPCODE_PK4B", - "OPCODE_PK4UB", - "OPCODE_RFL", - "OPCODE_SEQ", - "OPCODE_SFL", - "OPCODE_SGT", - "OPCODE_SIN", - "OPCODE_SLE", - "OPCODE_SNE", - "OPCODE_STR", - "OPCODE_TEX", - "OPCODE_TXD", - "OPCODE_UP2H", - "OPCODE_UP2US", - "OPCODE_UP4B", - "OPCODE_UP4UB", - "OPCODE_X2D", - "OPCODE_ARA", - "OPCODE_ARR", - "OPCODE_BRA", - "OPCODE_CAL", - "OPCODE_RET", - "OPCODE_SSG", - "OPCODE_CMP", - "OPCODE_SCS", - "OPCODE_TXB", - "OPCODE_NRM", - "OPCODE_DIV", - "OPCODE_DP2", - "OPCODE_TXL", - "OPCODE_BRK", - "OPCODE_IF", - "OPCODE_LOOP", - "OPCODE_REP", - "OPCODE_ELSE", - "OPCODE_ENDIF", - "OPCODE_ENDLOOP", - "OPCODE_ENDREP", - "OPCODE_PUSHA", - "OPCODE_POPA", - "OPCODE_CEIL", - "OPCODE_I2F", - "OPCODE_NOT", - "OPCODE_TRUNC", - "OPCODE_SHL", - "OPCODE_SHR", - "OPCODE_AND", - "OPCODE_OR", - "OPCODE_MOD", - "OPCODE_XOR", - "OPCODE_SAD", - "OPCODE_TXF", - "OPCODE_TXQ", - "OPCODE_CONT", - "OPCODE_EMIT", - "OPCODE_ENDPRIM", - "OPCODE_BGNLOOP2", - "OPCODE_BGNSUB", - "OPCODE_ENDLOOP2", - "OPCODE_ENDSUB", - "OPCODE_NOISE1", - "OPCODE_NOISE2", - "OPCODE_NOISE3", - "OPCODE_NOISE4", - "OPCODE_NOP", - "OPCODE_TEXBEM", - "OPCODE_TEXBEML", - "OPCODE_TEXREG2AR", - "OPCODE_TEXM3X2PAD", - "OPCODE_TEXM3X2TEX", - "OPCODE_TEXM3X3PAD", - "OPCODE_TEXM3X3TEX", - "OPCODE_TEXM3X3SPEC", - "OPCODE_TEXM3X3VSPEC", - "OPCODE_TEXREG2GB", - "OPCODE_TEXREG2RGB", - "OPCODE_TEXDP3TEX", - "OPCODE_TEXDP3", - "OPCODE_TEXM3X3", - "OPCODE_TEXM3X2DEPTH", - "OPCODE_TEXDEPTH", - "OPCODE_BEM", - "OPCODE_M4X3", - "OPCODE_M3X4", - "OPCODE_M3X3", - "OPCODE_M3X2", - "OPCODE_NRM4", - "OPCODE_CALLNZ", - "OPCODE_IFC", - "OPCODE_BREAKC", - "OPCODE_KIL", - "OPCODE_END" -}; - -static const char *TGSI_OPCODES_SHORT[] = -{ - "ARL", - "MOV", - "LIT", - "RCP", - "RSQ", - "EXP", - "LOG", - "MUL", - "ADD", - "DP3", - "DP4", - "DST", - "MIN", - "MAX", - "SLT", - "SGE", - "MAD", - "SUB", - "LERP", - "CND", - "CND0", - "DOT2ADD", - "INDEX", - "NEGATE", - "FRAC", - "CLAMP", - "FLOOR", - "ROUND", - "EXPBASE2", - "LOGBASE2", - "POWER", - "CROSSPRODUCT", - "MULTIPLYMATRIX", - "ABS", - "RCC", - "DPH", - "COS", - "DDX", - "DDY", - "KILP", - "PK2H", - "PK2US", - "PK4B", - "PK4UB", - "RFL", - "SEQ", - "SFL", - "SGT", - "SIN", - "SLE", - "SNE", - "STR", - "TEX", - "TXD", - "UP2H", - "UP2US", - "UP4B", - "UP4UB", - "X2D", - "ARA", - "ARR", - "BRA", - "CAL", - "RET", - "SSG", - "CMP", - "SCS", - "TXB", - "NRM", - "DIV", - "DP2", - "TXL", - "BRK", - "IF", - "LOOP", - "REP", - "ELSE", - "ENDIF", - "ENDLOOP", - "ENDREP", - "PUSHA", - "POPA", - "CEIL", - "I2F", - "NOT", - "TRUNC", - "SHL", - "SHR", - "AND", - "OR", - "MOD", - "XOR", - "SAD", - "TXF", - "TXQ", - "CONT", - "EMIT", - "ENDPRIM", - "BGNLOOP2", - "BGNSUB", - "ENDLOOP2", - "ENDSUB", - "NOISE1", - "NOISE2", - "NOISE3", - "NOISE4", - "NOP", - "TEXBEM", - "TEXBEML", - "TEXREG2AR", - "TEXM3X2PAD", - "TEXM3X2TEX", - "TEXM3X3PAD", - "TEXM3X3TEX", - "TEXM3X3SPEC", - "TEXM3X3VSPEC", - "TEXREG2GB", - "TEXREG2RGB", - "TEXDP3TEX", - "TEXDP3", - "TEXM3X3", - "TEXM3X2DEPTH", - "TEXDEPTH", - "BEM", - "M4X3", - "M3X4", - "M3X3", - "M3X2", - "NRM4", - "CALLNZ", - "IFC", - "BREAKC", - "KIL", - "END" -}; - -static const char *TGSI_SATS[] = -{ - "SAT_NONE", - "SAT_ZERO_ONE", - "SAT_MINUS_PLUS_ONE" -}; - -static const char *TGSI_INSTRUCTION_EXTS[] = -{ - "INSTRUCTION_EXT_TYPE_NV", - "INSTRUCTION_EXT_TYPE_LABEL", - "INSTRUCTION_EXT_TYPE_TEXTURE" -}; - -static const char *TGSI_PRECISIONS[] = -{ - "PRECISION_DEFAULT", - "TGSI_PRECISION_FLOAT32", - "TGSI_PRECISION_FLOAT16", - "TGSI_PRECISION_FIXED12" -}; - -static const char *TGSI_CCS[] = -{ - "CC_GT", - "CC_EQ", - "CC_LT", - "CC_UN", - "CC_GE", - "CC_LE", - "CC_NE", - "CC_TR", - "CC_FL" -}; - -static const char *TGSI_SWIZZLES[] = -{ - "SWIZZLE_X", - "SWIZZLE_Y", - "SWIZZLE_Z", - "SWIZZLE_W" -}; - -static const char *TGSI_SWIZZLES_SHORT[] = -{ - "x", - "y", - "z", - "w" -}; - -static const char *TGSI_TEXTURES[] = -{ - "TEXTURE_UNKNOWN", - "TEXTURE_1D", - "TEXTURE_2D", - "TEXTURE_3D", - "TEXTURE_CUBE", - "TEXTURE_RECT", - "TEXTURE_SHADOW1D", - "TEXTURE_SHADOW2D", - "TEXTURE_SHADOWRECT" -}; - -static const char *TGSI_SRC_REGISTER_EXTS[] = -{ - "SRC_REGISTER_EXT_TYPE_SWZ", - "SRC_REGISTER_EXT_TYPE_MOD" -}; - -static const char *TGSI_EXTSWIZZLES[] = -{ - "EXTSWIZZLE_X", - "EXTSWIZZLE_Y", - "EXTSWIZZLE_Z", - "EXTSWIZZLE_W", - "EXTSWIZZLE_ZERO", - "EXTSWIZZLE_ONE" -}; - -static const char *TGSI_EXTSWIZZLES_SHORT[] = -{ - "x", - "y", - "z", - "w", - "0", - "1" -}; - -static const char *TGSI_WRITEMASKS[] = -{ - "0", - "WRITEMASK_X", - "WRITEMASK_Y", - "WRITEMASK_XY", - "WRITEMASK_Z", - "WRITEMASK_XZ", - "WRITEMASK_YZ", - "WRITEMASK_XYZ", - "WRITEMASK_W", - "WRITEMASK_XW", - "WRITEMASK_YW", - "WRITEMASK_XYW", - "WRITEMASK_ZW", - "WRITEMASK_XZW", - "WRITEMASK_YZW", - "WRITEMASK_XYZW" -}; - -static const char *TGSI_DST_REGISTER_EXTS[] = -{ - "DST_REGISTER_EXT_TYPE_CONDCODE", - "DST_REGISTER_EXT_TYPE_MODULATE" -}; - -static const char *TGSI_MODULATES[] = -{ - "MODULATE_1X", - "MODULATE_2X", - "MODULATE_4X", - "MODULATE_8X", - "MODULATE_HALF", - "MODULATE_QUARTER", - "MODULATE_EIGHTH" -}; - -static void -dump_declaration_short( - struct gen_dump *dump, - struct tgsi_full_declaration *decl ) -{ - TXT( "\nDCL " ); - ENM( decl->Declaration.File, TGSI_FILES_SHORT ); - - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - CHR( '[' ); - UID( decl->u.DeclarationRange.First ); - if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { - TXT( ".." ); - UID( decl->u.DeclarationRange.Last ); - } - CHR( ']' ); - break; - default: - assert( 0 ); - } - - if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { - CHR( '.' ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'x' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'w' ); - } - } - - if( decl->Declaration.Interpolate ) { - TXT( ", " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); - } - - if( decl->Declaration.Semantic ) { - TXT( ", " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); - CHR( '[' ); - UID( decl->Semantic.SemanticIndex ); - CHR( ']' ); - } -} - -static void -dump_declaration_verbose( - struct gen_dump *dump, - struct tgsi_full_declaration *decl, - unsigned ignored, - unsigned deflt, - struct tgsi_full_declaration *fd ) -{ - TXT( "\nFile : " ); - ENM( decl->Declaration.File, TGSI_FILES ); - TXT( "\nDeclare : " ); - ENM( decl->Declaration.Declare, TGSI_DECLARES ); - if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { - TXT( "\nUsageMask : " ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'X' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'Y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'Z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'W' ); - } - } - if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { - TXT( "\nInterpolate: " ); - UID( decl->Declaration.Interpolate ); - } - if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { - TXT( "\nSemantic : " ); - UID( decl->Declaration.Semantic ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Declaration.Padding ); - } - - CHR( '\n' ); - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - TXT( "\nFirst: " ); - UID( decl->u.DeclarationRange.First ); - TXT( "\nLast : " ); - UID( decl->u.DeclarationRange.Last ); - break; - - case TGSI_DECLARE_MASK: - TXT( "\nMask: " ); - UIX( decl->u.DeclarationMask.Mask ); - break; - - default: - assert( 0 ); - } - - if( decl->Declaration.Interpolate ) { - CHR( '\n' ); - TXT( "\nInterpolate: " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Interpolation.Padding ); - } - } - - if( decl->Declaration.Semantic ) { - CHR( '\n' ); - TXT( "\nSemanticName : " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); - TXT( "\nSemanticIndex: " ); - UID( decl->Semantic.SemanticIndex ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Semantic.Padding ); - } - } -} - -static void -dump_immediate_short( - struct gen_dump *dump, - struct tgsi_full_immediate *imm ) -{ - unsigned i; - - TXT( "\nIMM " ); - ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); - - TXT( " { " ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - - if( i < imm->Immediate.Size - 2 ) { - TXT( ", " ); - } - } - TXT( " }" ); -} - -static void -dump_immediate_verbose( - struct gen_dump *dump, - struct tgsi_full_immediate *imm, - unsigned ignored ) -{ - unsigned i; - - TXT( "\nDataType : " ); - ENM( imm->Immediate.DataType, TGSI_IMMS ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( imm->Immediate.Padding ); - } - - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - CHR( '\n' ); - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - TXT( "\nFloat: " ); - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - } -} - -static void -dump_instruction_short( - struct gen_dump *dump, - struct tgsi_full_instruction *inst, - unsigned instno ) -{ - unsigned i; - boolean first_reg = TRUE; - - CHR( '\n' ); - UID( instno ); - CHR( ':' ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); - - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - TXT( "_SAT" ); - break; - case TGSI_SAT_MINUS_PLUS_ONE: - TXT( "_SAT[-1,1]" ); - break; - default: - assert( 0 ); - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if( !first_reg ) { - CHR( ',' ); - } - CHR( ' ' ); - - ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); - SID( dst->DstRegister.Index ); - CHR( ']' ); - - if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { - CHR( '.' ); - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { - CHR( 'x' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { - CHR( 'y' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { - CHR( 'z' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { - CHR( 'w' ); - } - } - - first_reg = FALSE; - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - - if( !first_reg ) { - CHR( ',' ); - } - CHR( ' ' ); - - if( src->SrcRegisterExtMod.Complement ) { - TXT( "(1 - " ); - } - if( src->SrcRegisterExtMod.Negate ) { - CHR( '-' ); - } - if( src->SrcRegisterExtMod.Absolute ) { - CHR( '|' ); - } - if( src->SrcRegister.Negate ) { - CHR( '-' ); - } - - ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); - SID( src->SrcRegister.Index ); - CHR( ']' ); - - if (src->SrcRegister.Extended) { - if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || - src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || - src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || - src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { - CHR( '.' ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); - } - } - else if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || - src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || - src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || - src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { - CHR( '.' ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); - } - - if( src->SrcRegisterExtMod.Absolute ) { - CHR( '|' ); - } - if( src->SrcRegisterExtMod.Complement ) { - CHR( ')' ); - } - - first_reg = FALSE; - } - - switch( inst->Instruction.Opcode ) { - case TGSI_OPCODE_IF: - case TGSI_OPCODE_ELSE: - case TGSI_OPCODE_BGNLOOP2: - case TGSI_OPCODE_ENDLOOP2: - case TGSI_OPCODE_CAL: - TXT( " :" ); - UID( inst->InstructionExtLabel.Label ); - break; - } -} - -static void -dump_instruction_verbose( - struct gen_dump *dump, - struct tgsi_full_instruction *inst, - unsigned ignored, - unsigned deflt, - struct tgsi_full_instruction *fi ) -{ - unsigned i; - - TXT( "\nOpcode : " ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES ); - if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { - TXT( "\nSaturate : " ); - ENM( inst->Instruction.Saturate, TGSI_SATS ); - } - if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { - TXT( "\nNumDstRegs : " ); - UID( inst->Instruction.NumDstRegs ); - } - if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { - TXT( "\nNumSrcRegs : " ); - UID( inst->Instruction.NumSrcRegs ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->Instruction.Padding ); - } - - if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { - TXT( "\nPrecision : " ); - ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); - } - if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { - TXT( "\nCondDstIndex : " ); - UID( inst->InstructionExtNv.CondDstIndex ); - } - if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { - TXT( "\nCondFlowIndex : " ); - UID( inst->InstructionExtNv.CondFlowIndex ); - } - if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { - TXT( "\nCondMask : " ); - ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { - TXT( "\nCondSwizzleX : " ); - ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { - TXT( "\nCondSwizzleY : " ); - ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ : " ); - ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { - TXT( "\nCondSwizzleW : " ); - ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { - TXT( "\nCondDstUpdate : " ); - UID( inst->InstructionExtNv.CondDstUpdate ); - } - if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { - TXT( "\nCondFlowEnable: " ); - UID( inst->InstructionExtNv.CondFlowEnable ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtNv.Padding ); - if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { - TXT( "\nExtended : " ); - UID( inst->InstructionExtNv.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { - TXT( "\nLabel : " ); - UID( inst->InstructionExtLabel.Label ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtLabel.Padding ); - if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtLabel.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { - TXT( "\nTexture : " ); - ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtTexture.Padding ); - if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtTexture.Extended ); - } - } - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : " ); - ENM( dst->DstRegister.File, TGSI_FILES ); - if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { - TXT( "\nWriteMask: " ); - ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); - } - if( ignored ) { - if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( dst->DstRegister.Indirect ); - } - if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( dst->DstRegister.Dimension ); - } - } - if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { - TXT( "\nIndex : " ); - SID( dst->DstRegister.Index ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegister.Padding ); - if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegister.Extended ); - } - } - - if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { - TXT( "\nCondMask : " ); - ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { - TXT( "\nCondSwizzleX: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { - TXT( "\nCondSwizzleY: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { - TXT( "\nCondSwizzleW: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { - TXT( "\nCondSrcIndex: " ); - UID( dst->DstRegisterExtConcode.CondSrcIndex ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtConcode.Padding ); - if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegisterExtConcode.Extended ); - } - } - } - - if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { - TXT( "\nModulate: " ); - ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtModulate.Padding ); - if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { - TXT( "\nExtended: " ); - UID( dst->DstRegisterExtModulate.Extended ); - } - } - } - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : "); - ENM( src->SrcRegister.File, TGSI_FILES ); - if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { - TXT( "\nSwizzleX : " ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { - TXT( "\nSwizzleY : " ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { - TXT( "\nSwizzleZ : " ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { - TXT( "\nSwizzleW : " ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegister.Negate ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( src->SrcRegister.Indirect ); - } - if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( src->SrcRegister.Dimension ); - } - } - if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { - TXT( "\nIndex : " ); - SID( src->SrcRegister.Index ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegister.Extended ); - } - } - - if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { - TXT( "\nExtSwizzleX: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { - TXT( "\nExtSwizzleY: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { - TXT( "\nExtSwizzleZ: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { - TXT( "\nExtSwizzleW: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { - TXT( "\nNegateX : " ); - UID( src->SrcRegisterExtSwz.NegateX ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { - TXT( "\nNegateY : " ); - UID( src->SrcRegisterExtSwz.NegateY ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { - TXT( "\nNegateZ : " ); - UID( src->SrcRegisterExtSwz.NegateZ ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { - TXT( "\nNegateW : " ); - UID( src->SrcRegisterExtSwz.NegateW ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { - TXT( "\nExtDivide : " ); - ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtSwz.Padding ); - if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtSwz.Extended ); - } - } - } - - if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { - TXT( "\nComplement: " ); - UID( src->SrcRegisterExtMod.Complement ); - } - if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { - TXT( "\nBias : " ); - UID( src->SrcRegisterExtMod.Bias ); - } - if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { - TXT( "\nScale2X : " ); - UID( src->SrcRegisterExtMod.Scale2X ); - } - if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { - TXT( "\nAbsolute : " ); - UID( src->SrcRegisterExtMod.Absolute ); - } - if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegisterExtMod.Negate ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtMod.Padding ); - if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtMod.Extended ); - } - } - } - } -} - -static void -dump_gen( - struct gen_dump *dump, - const struct tgsi_token *tokens, - unsigned flags ) -{ - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned verbose = flags & TGSI_DUMP_VERBOSE; - unsigned ignored = !(flags & TGSI_DUMP_NO_IGNORED); - unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); - unsigned instno = 0; - - dump->tabs = 0; - - /* sanity check */ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); - - tgsi_parse_init( &parse, tokens ); - - TXT( "tgsi-dump begin -----------------" ); - - CHR( '\n' ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); - CHR( ' ' ); - UID( parse.FullVersion.Version.MajorVersion ); - CHR( '.' ); - UID( parse.FullVersion.Version.MinorVersion ); - - if( verbose ) { - TXT( "\nMajorVersion: " ); - UID( parse.FullVersion.Version.MajorVersion ); - TXT( "\nMinorVersion: " ); - UID( parse.FullVersion.Version.MinorVersion ); - CHR( '\n' ); - - TXT( "\nHeaderSize: " ); - UID( parse.FullHeader.Header.HeaderSize ); - TXT( "\nBodySize : " ); - UID( parse.FullHeader.Header.BodySize ); - TXT( "\nProcessor : " ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); - CHR( '\n' ); - } - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_short( - dump, - &parse.FullToken.FullDeclaration ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_short( - dump, - &parse.FullToken.FullImmediate ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_short( - dump, - &parse.FullToken.FullInstruction, - instno ); - instno++; - break; - - default: - assert( 0 ); - } - - if( verbose ) { - TXT( "\nType : " ); - ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); - if( ignored ) { - TXT( "\nSize : " ); - UID( parse.FullToken.Token.Size ); - if( deflt || parse.FullToken.Token.Extended ) { - TXT( "\nExtended : " ); - UID( parse.FullToken.Token.Extended ); - } - } - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_verbose( - dump, - &parse.FullToken.FullDeclaration, - ignored, - deflt, - &fd ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_verbose( - dump, - &parse.FullToken.FullImmediate, - ignored ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_verbose( - dump, - &parse.FullToken.FullInstruction, - ignored, - deflt, - &fi ); - break; - - default: - assert( 0 ); - } - - CHR( '\n' ); - } - } - - TXT( "\ntgsi-dump end -------------------\n" ); - - tgsi_parse_free( &parse ); -} - - -static void -sanity_checks(void) -{ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); - assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "END") == 0); -} - - -void -tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ) -{ - struct file_dump dump; - - sanity_checks(); - - dump.base.write = _file_dump_write; -#if 0 - { - static unsigned counter = 0; - char buffer[64]; - sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); - dump.file = fopen( buffer, "wt" ); - } -#else - dump.file = stderr; -#endif - - dump_gen( - &dump.base, - tokens, - flags ); - -#if 0 - fclose( dump.file ); -#endif -} - -void -tgsi_dump_str( - char **str, - const struct tgsi_token *tokens, - unsigned flags ) -{ - struct text_dump dump; - - dump.base.write = _text_dump_write; - dump.text = NULL; - dump.length = 0; - dump.capacity = 0; - - dump_gen( - &dump.base, - tokens, - flags ); - - *str = dump.text; -} diff --git a/src/mesa/pipe/tgsi/util/tgsi_dump.h b/src/mesa/pipe/tgsi/util/tgsi_dump.h deleted file mode 100644 index 1adc9db251..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_dump.h +++ /dev/null @@ -1,28 +0,0 @@ -#if !defined TGSI_DUMP_H -#define TGSI_DUMP_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -#define TGSI_DUMP_VERBOSE 1 -#define TGSI_DUMP_NO_IGNORED 2 -#define TGSI_DUMP_NO_DEFAULT 4 - -void -tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ); - -void -tgsi_dump_str( - char **str, - const struct tgsi_token *tokens, - unsigned flags ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_DUMP_H - diff --git a/src/mesa/pipe/tgsi/util/tgsi_parse.c b/src/mesa/pipe/tgsi/util/tgsi_parse.c deleted file mode 100644 index bf6b89ce56..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_parse.c +++ /dev/null @@ -1,319 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ) -{ - full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; -} - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ) -{ - if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { - FREE( full_token->FullImmediate.u.Pointer ); - } -} - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ) -{ - ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; - if( ctx->FullVersion.Version.MajorVersion > 1 ) { - return TGSI_PARSE_ERROR; - } - - ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; - if( ctx->FullHeader.Header.HeaderSize >= 2 ) { - ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; - } - else { - ctx->FullHeader.Processor = tgsi_default_processor(); - } - - ctx->Tokens = tokens; - ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; - - tgsi_full_token_init( &ctx->FullToken ); - - return TGSI_PARSE_OK; -} - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ) -{ - tgsi_full_token_free( &ctx->FullToken ); -} - -boolean -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ) -{ - return ctx->Position >= - 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; -} - -static void -next_token( - struct tgsi_parse_context *ctx, - void *token ) -{ - assert( !tgsi_parse_end_of_tokens( ctx ) ); - - *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; -} - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ) -{ - struct tgsi_token token; - unsigned i; - - tgsi_full_token_free( &ctx->FullToken ); - tgsi_full_token_init( &ctx->FullToken ); - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; - - *decl = tgsi_default_full_declaration(); - decl->Declaration = *(struct tgsi_declaration *) &token; - - switch( decl->Declaration.Type ) { - case TGSI_DECLARE_RANGE: - next_token( ctx, &decl->u.DeclarationRange ); - break; - - case TGSI_DECLARE_MASK: - next_token( ctx, &decl->u.DeclarationMask ); - break; - - default: - assert (0); - } - - if( decl->Declaration.Interpolate ) { - next_token( ctx, &decl->Interpolation ); - } - - if( decl->Declaration.Semantic ) { - next_token( ctx, &decl->Semantic ); - } - - break; - } - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; - - *imm = tgsi_default_full_immediate(); - imm->Immediate = *(struct tgsi_immediate *) &token; - - assert( !imm->Immediate.Extended ); - - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - imm->u.Pointer = MALLOC( - sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - next_token( ctx, &imm->u.ImmediateFloat32[i] ); - } - break; - - default: - assert( 0 ); - } - - break; - } - - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; - unsigned extended; - - *inst = tgsi_default_full_instruction(); - inst->Instruction = *(struct tgsi_instruction *) &token; - - extended = inst->Instruction.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_INSTRUCTION_EXT_TYPE_NV: - inst->InstructionExtNv = - *(struct tgsi_instruction_ext_nv *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_LABEL: - inst->InstructionExtLabel = - *(struct tgsi_instruction_ext_label *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: - inst->InstructionExtTexture = - *(struct tgsi_instruction_ext_texture *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); - assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); - - extended = inst->FullDstRegisters[i].DstRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: - inst->FullDstRegisters[i].DstRegisterExtConcode = - *(struct tgsi_dst_register_ext_concode *) &token; - break; - - case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: - inst->FullDstRegisters[i].DstRegisterExtModulate = - *(struct tgsi_dst_register_ext_modulate *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - } - - assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); - - extended = inst->FullSrcRegisters[i].SrcRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: - inst->FullSrcRegisters[i].SrcRegisterExtSwz = - *(struct tgsi_src_register_ext_swz *) &token; - break; - - case TGSI_SRC_REGISTER_EXT_TYPE_MOD: - inst->FullSrcRegisters[i].SrcRegisterExtMod = - *(struct tgsi_src_register_ext_mod *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - - if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); - - /* - * No support for multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); - - if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - } - } - - break; - } - - default: - assert( 0 ); - } -} - diff --git a/src/mesa/pipe/tgsi/util/tgsi_parse.h b/src/mesa/pipe/tgsi/util/tgsi_parse.h deleted file mode 100644 index 9372da8d5d..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_parse.h +++ /dev/null @@ -1,121 +0,0 @@ -#if !defined TGSI_PARSE_H -#define TGSI_PARSE_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_full_version -{ - struct tgsi_version Version; -}; - -struct tgsi_full_header -{ - struct tgsi_header Header; - struct tgsi_processor Processor; -}; - -struct tgsi_full_dst_register -{ - struct tgsi_dst_register DstRegister; - struct tgsi_dst_register_ext_concode DstRegisterExtConcode; - struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; -}; - -struct tgsi_full_src_register -{ - struct tgsi_src_register SrcRegister; - struct tgsi_src_register_ext_swz SrcRegisterExtSwz; - struct tgsi_src_register_ext_mod SrcRegisterExtMod; - struct tgsi_src_register SrcRegisterInd; - struct tgsi_dimension SrcRegisterDim; - struct tgsi_src_register SrcRegisterDimInd; -}; - -struct tgsi_full_declaration -{ - struct tgsi_declaration Declaration; - union - { - struct tgsi_declaration_range DeclarationRange; - struct tgsi_declaration_mask DeclarationMask; - } u; - struct tgsi_declaration_interpolation Interpolation; - struct tgsi_declaration_semantic Semantic; -}; - -struct tgsi_full_immediate -{ - struct tgsi_immediate Immediate; - union - { - void *Pointer; - struct tgsi_immediate_float32 *ImmediateFloat32; - } u; -}; - -#define TGSI_FULL_MAX_DST_REGISTERS 2 -#define TGSI_FULL_MAX_SRC_REGISTERS 3 - -struct tgsi_full_instruction -{ - struct tgsi_instruction Instruction; - struct tgsi_instruction_ext_nv InstructionExtNv; - struct tgsi_instruction_ext_label InstructionExtLabel; - struct tgsi_instruction_ext_texture InstructionExtTexture; - struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; - struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; -}; - -union tgsi_full_token -{ - struct tgsi_token Token; - struct tgsi_full_declaration FullDeclaration; - struct tgsi_full_immediate FullImmediate; - struct tgsi_full_instruction FullInstruction; -}; - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ); - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ); - -struct tgsi_parse_context -{ - const struct tgsi_token *Tokens; - unsigned Position; - struct tgsi_full_version FullVersion; - struct tgsi_full_header FullHeader; - union tgsi_full_token FullToken; -}; - -#define TGSI_PARSE_OK 0 -#define TGSI_PARSE_ERROR 1 - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ); - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ); - -boolean -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ); - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_PARSE_H - diff --git a/src/mesa/pipe/tgsi/util/tgsi_transform.c b/src/mesa/pipe/tgsi/util/tgsi_transform.c deleted file mode 100644 index 357f77b05a..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_transform.c +++ /dev/null @@ -1,199 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * TGSI program transformation utility. - * - * Authors: Brian Paul - */ - - -#include "tgsi_transform.h" - - - -static void -emit_instruction(struct tgsi_transform_context *ctx, - const struct tgsi_full_instruction *inst) -{ - uint ti = ctx->ti; - - ti += tgsi_build_full_instruction(inst, - ctx->tokens_out + ti, - ctx->header, - ctx->max_tokens_out - ti); - ctx->ti = ti; -} - - -static void -emit_declaration(struct tgsi_transform_context *ctx, - const struct tgsi_full_declaration *decl) -{ - uint ti = ctx->ti; - - ti += tgsi_build_full_declaration(decl, - ctx->tokens_out + ti, - ctx->header, - ctx->max_tokens_out - ti); - ctx->ti = ti; -} - - -static void -emit_immediate(struct tgsi_transform_context *ctx, - const struct tgsi_full_immediate *imm) -{ - uint ti = ctx->ti; - - ti += tgsi_build_full_immediate(imm, - ctx->tokens_out + ti, - ctx->header, - ctx->max_tokens_out - ti); - ctx->ti = ti; -} - - - -/** - * Apply user-defined transformations to the input shader to produce - * the output shader. - * For example, a register search-and-replace operation could be applied - * by defining a transform_instruction() callback that examined and changed - * the instruction src/dest regs. - * - * \return number of tokens emitted - */ -int -tgsi_transform_shader(const struct tgsi_token *tokens_in, - struct tgsi_token *tokens_out, - uint max_tokens_out, - struct tgsi_transform_context *ctx) -{ - uint procType; - - /* input shader */ - struct tgsi_parse_context parse; - - /* output shader */ - struct tgsi_processor *processor; - - - /** - ** callback context init - **/ - ctx->emit_instruction = emit_instruction; - ctx->emit_declaration = emit_declaration; - ctx->emit_immediate = emit_immediate; - ctx->tokens_out = tokens_out; - ctx->max_tokens_out = max_tokens_out; - - - /** - ** Setup to begin parsing input shader - **/ - if (tgsi_parse_init( &parse, tokens_in ) != TGSI_PARSE_OK) { - debug_printf("tgsi_parse_init() failed in tgsi_transform_shader()!\n"); - return -1; - } - procType = parse.FullHeader.Processor.Processor; - assert(procType == TGSI_PROCESSOR_FRAGMENT || - procType == TGSI_PROCESSOR_VERTEX || - procType == TGSI_PROCESSOR_GEOMETRY); - - - /** - ** Setup output shader - **/ - *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version(); - - ctx->header = (struct tgsi_header *) (tokens_out + 1); - *ctx->header = tgsi_build_header(); - - processor = (struct tgsi_processor *) (tokens_out + 2); - *processor = tgsi_build_processor( procType, ctx->header ); - - ctx->ti = 3; - - - /** - ** Loop over incoming program tokens/instructions - */ - while( !tgsi_parse_end_of_tokens( &parse ) ) { - - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *fullinst - = &parse.FullToken.FullInstruction; - - if (ctx->transform_instruction) - ctx->transform_instruction(ctx, fullinst); - else - ctx->emit_instruction(ctx, fullinst); - } - break; - - case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *fulldecl - = &parse.FullToken.FullDeclaration; - - if (ctx->transform_declaration) - ctx->transform_declaration(ctx, fulldecl); - else - ctx->emit_declaration(ctx, fulldecl); - } - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - struct tgsi_full_immediate *fullimm - = &parse.FullToken.FullImmediate; - - if (ctx->transform_immediate) - ctx->transform_immediate(ctx, fullimm); - else - ctx->emit_immediate(ctx, fullimm); - } - break; - - default: - assert( 0 ); - } - } - - if (ctx->epilog) { - ctx->epilog(ctx); - } - - tgsi_parse_free (&parse); - - return ctx->ti; -} diff --git a/src/mesa/pipe/tgsi/util/tgsi_transform.h b/src/mesa/pipe/tgsi/util/tgsi_transform.h deleted file mode 100644 index 365d8c298c..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_transform.h +++ /dev/null @@ -1,93 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_TRANSFORM_H -#define TGSI_TRANSFORM_H - - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/util/tgsi_build.h" - - - -/** - * Subclass this to add caller-specific data - */ -struct tgsi_transform_context -{ -/**** PUBLIC ***/ - - /** - * User-defined callbacks invoked per instruction. - */ - void (*transform_instruction)(struct tgsi_transform_context *ctx, - struct tgsi_full_instruction *inst); - - void (*transform_declaration)(struct tgsi_transform_context *ctx, - struct tgsi_full_declaration *decl); - - void (*transform_immediate)(struct tgsi_transform_context *ctx, - struct tgsi_full_immediate *imm); - - /** - * Called at end of input program to allow caller to append extra - * instructions. Return number of tokens emitted. - */ - void (*epilog)(struct tgsi_transform_context *ctx); - - -/*** PRIVATE ***/ - - /** - * These are setup by tgsi_transform_shader() and cannot be overridden. - * Meant to be called from in the above user callback functions. - */ - void (*emit_instruction)(struct tgsi_transform_context *ctx, - const struct tgsi_full_instruction *inst); - void (*emit_declaration)(struct tgsi_transform_context *ctx, - const struct tgsi_full_declaration *decl); - void (*emit_immediate)(struct tgsi_transform_context *ctx, - const struct tgsi_full_immediate *imm); - - struct tgsi_header *header; - uint max_tokens_out; - struct tgsi_token *tokens_out; - uint ti; -}; - - - -extern int -tgsi_transform_shader(const struct tgsi_token *tokens_in, - struct tgsi_token *tokens_out, - uint max_tokens_out, - struct tgsi_transform_context *ctx); - - -#endif /* TGSI_TRANSFORM_H */ diff --git a/src/mesa/pipe/tgsi/util/tgsi_util.c b/src/mesa/pipe/tgsi/util/tgsi_util.c deleted file mode 100644 index 4cdd89182a..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_util.c +++ /dev/null @@ -1,274 +0,0 @@ -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" -#include "tgsi_util.h" - -union pointer_hack -{ - void *pointer; - unsigned long long uint64; -}; - -void * -tgsi_align_128bit( - void *unaligned ) -{ - union pointer_hack ph; - - ph.uint64 = 0; - ph.pointer = unaligned; - ph.uint64 = (ph.uint64 + 15) & ~15; - return ph.pointer; -} - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->SwizzleX; - case 1: - return reg->SwizzleY; - case 2: - return reg->SwizzleZ; - case 3: - return reg->SwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->ExtSwizzleX; - case 1: - return reg->ExtSwizzleY; - case 2: - return reg->ExtSwizzleZ; - case 3: - return reg->ExtSwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned swizzle; - - /* - * First, calculate the extended swizzle for a given channel. This will give - * us either a channel index into the simple swizzle or a constant 1 or 0. - */ - swizzle = tgsi_util_get_src_register_extswizzle( - ®->SrcRegisterExtSwz, - component ); - - assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); - assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); - assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); - assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); - - /* - * Second, calculate the simple swizzle for the unswizzled channel index. - * Leave the constants intact, they are not affected by the simple swizzle. - */ - if( swizzle <= TGSI_SWIZZLE_W ) { - swizzle = tgsi_util_get_src_register_swizzle( - ®->SrcRegister, - component ); - } - - return swizzle; -} - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - unsigned swizzle, - unsigned component ) -{ - switch( component ) { - case 0: - reg->SwizzleX = swizzle; - break; - case 1: - reg->SwizzleY = swizzle; - break; - case 2: - reg->SwizzleZ = swizzle; - break; - case 3: - reg->SwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - unsigned swizzle, - unsigned component ) -{ - switch( component ) { - case 0: - reg->ExtSwizzleX = swizzle; - break; - case 1: - reg->ExtSwizzleY = swizzle; - break; - case 2: - reg->ExtSwizzleZ = swizzle; - break; - case 3: - reg->ExtSwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->NegateX; - case 1: - return reg->NegateY; - case 2: - return reg->NegateZ; - case 3: - return reg->NegateW; - default: - assert( 0 ); - } - return 0; -} - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ) -{ - switch( component ) { - case 0: - reg->NegateX = negate; - break; - case 1: - reg->NegateY = negate; - break; - case 2: - reg->NegateZ = negate; - break; - case 3: - reg->NegateW = negate; - break; - default: - assert( 0 ); - } -} - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned sign_mode; - - if( reg->SrcRegisterExtMod.Absolute ) { - /* Consider only the post-abs negation. */ - - if( reg->SrcRegisterExtMod.Negate ) { - sign_mode = TGSI_UTIL_SIGN_SET; - } - else { - sign_mode = TGSI_UTIL_SIGN_CLEAR; - } - } - else { - /* Accumulate the three negations. */ - - unsigned negate; - - negate = reg->SrcRegister.Negate; - if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { - negate = !negate; - } - if( reg->SrcRegisterExtMod.Negate ) { - negate = !negate; - } - - if( negate ) { - sign_mode = TGSI_UTIL_SIGN_TOGGLE; - } - else { - sign_mode = TGSI_UTIL_SIGN_KEEP; - } - } - - return sign_mode; -} - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - unsigned sign_mode ) -{ - reg->SrcRegisterExtSwz.NegateX = 0; - reg->SrcRegisterExtSwz.NegateY = 0; - reg->SrcRegisterExtSwz.NegateZ = 0; - reg->SrcRegisterExtSwz.NegateW = 0; - - switch (sign_mode) - { - case TGSI_UTIL_SIGN_CLEAR: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_SET: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 1; - break; - - case TGSI_UTIL_SIGN_TOGGLE: - reg->SrcRegister.Negate = 1; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_KEEP: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - default: - assert( 0 ); - } -} - diff --git a/src/mesa/pipe/tgsi/util/tgsi_util.h b/src/mesa/pipe/tgsi/util/tgsi_util.h deleted file mode 100644 index ef14446f0e..0000000000 --- a/src/mesa/pipe/tgsi/util/tgsi_util.h +++ /dev/null @@ -1,70 +0,0 @@ -#if !defined TGSI_UTIL_H -#define TGSI_UTIL_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -void * -tgsi_align_128bit( - void *unaligned ); - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component); - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - unsigned swizzle, - unsigned component ); - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - unsigned swizzle, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ); - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ); - -#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ -#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ -#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ -#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - unsigned sign_mode ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_UTIL_H - diff --git a/src/mesa/pipe/util/p_debug.c b/src/mesa/pipe/util/p_debug.c deleted file mode 100644 index b9607a6ba7..0000000000 --- a/src/mesa/pipe/util/p_debug.c +++ /dev/null @@ -1,76 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include - -#ifdef WIN32 -#include -#include -#else -#include -#include -#endif - -#include "pipe/p_debug.h" -#include "pipe/p_compiler.h" - - -void debug_vprintf(const char *format, va_list ap) -{ -#ifdef WIN32 - EngDebugPrint("Gallium3D: ", (PCHAR)format, ap); -#else - vfprintf(stderr, format, ap); -#endif -} - - -void debug_printf(const char *format, ...) -{ - va_list ap; - va_start(ap, format); - debug_vprintf(format, ap); - va_end(ap); -} - - -static INLINE void debug_abort(void) -{ -#ifdef WIN32 - EngDebugBreak(); -#else - abort(); -#endif -} - - -void debug_assert_fail(const char *expr, const char *file, unsigned line) -{ - debug_printf("%s:%i: Assertion `%s' failed.\n", file, line, expr); - debug_abort(); -} diff --git a/src/mesa/pipe/util/p_tile.c b/src/mesa/pipe/util/p_tile.c deleted file mode 100644 index 3f795a3898..0000000000 --- a/src/mesa/pipe/util/p_tile.c +++ /dev/null @@ -1,699 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * RGBA/float tile get/put functions. - * Usable both by drivers and state trackers. - * Surfaces should already be in a mapped state. - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" - -#include "p_tile.h" - - - -/** - * Move raw block of pixels from surface to user memory. - * This should be usable by any hw driver that has mappable surfaces. - */ -void -pipe_get_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride) -{ - const uint cpp = ps->cpp; - const ubyte *pSrc; - const uint src_stride = ps->pitch * cpp; - ubyte *pDest; - uint i; - - if (dst_stride == 0) { - dst_stride = w * cpp; - } - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - pSrc = (const ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; - pDest = (ubyte *) p; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w * cpp); - pDest += dst_stride; - pSrc += src_stride; - } - - pipe_surface_unmap(ps); -} - - -/** - * Move raw block of pixels from user memory to surface. - * This should be usable by any hw driver that has mappable surfaces. - */ -void -pipe_put_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride) -{ - const uint cpp = ps->cpp; - const ubyte *pSrc; - const uint dst_stride = ps->pitch * cpp; - ubyte *pDest; - uint i; - - if (src_stride == 0) { - src_stride = w * cpp; - } - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - pSrc = (const ubyte *) p; - pDest = (ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w * cpp); - pDest += dst_stride; - pSrc += src_stride; - } - - pipe_surface_unmap(ps); -} - - - - -/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */ -#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) - -#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ - us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) - - - -/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ - -static void -a8r8g8b8_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const unsigned pixel = *src++; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - } - p += dst_stride; - } -} - - -static void -a8r8g8b8_put_tile_rgba(unsigned *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - *dst++ = (a << 24) | (r << 16) | (g << 8) | b; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ - -static void -b8g8r8a8_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const unsigned pixel = *src++; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - } - p += dst_stride; - } -} - - -static void -b8g8r8a8_put_tile_rgba(unsigned *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - *dst++ = (b << 24) | (g << 16) | (r << 8) | a; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ - -static void -a1r5g5b5_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const ushort pixel = *src++; - pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); - pRow[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); - pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - pRow[3] = ((pixel >> 15) ) * 1.0f; - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ - -static void -a4r4g4b4_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const ushort pixel = *src++; - pRow[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); - pRow[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); - pRow[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); - pRow[3] = ((pixel >> 12) ) * (1.0f / 15.0f); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_R5G6B5_UNORM ***/ - -static void -r5g6b5_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const ushort pixel = *src++; - pRow[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); - pRow[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); - pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - pRow[3] = 1.0f; - } - p += dst_stride; - } -} - - -static void -r5g5b5_put_tile_rgba(ushort *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); - uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); - uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); - *dst++ = (r << 11) | (g << 5) | (b); - } - p += src_stride; - } -} - - - -/*** PIPE_FORMAT_Z16_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z16_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const float scale = 1.0f / 65535.0f; - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = *src++ * scale; - } - p += dst_stride; - } -} - - - - -/*** PIPE_FORMAT_U_L8 ***/ - -static void -l8_get_tile_rgba(ubyte *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(*src); - pRow[3] = 1.0; - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_U_A8 ***/ - -static void -a8_get_tile_rgba(ubyte *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = 0.0; - pRow[3] = UBYTE_TO_FLOAT(*src); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ - -static void -r16g16b16a16_get_tile_rgba(short *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src += 4, pRow += 4) { - pRow[0] = SHORT_TO_FLOAT(src[0]); - pRow[1] = SHORT_TO_FLOAT(src[1]); - pRow[2] = SHORT_TO_FLOAT(src[2]); - pRow[3] = SHORT_TO_FLOAT(src[3]); - } - p += dst_stride; - } -} - - -static void -r16g16b16a16_put_tile_rgba(short *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, dst += 4, pRow += 4) { - UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); - UNCLAMPED_FLOAT_TO_SHORT(dst[1], pRow[1]); - UNCLAMPED_FLOAT_TO_SHORT(dst[2], pRow[2]); - UNCLAMPED_FLOAT_TO_SHORT(dst[3], pRow[3]); - } - p += src_stride; - } -} - - - -/*** PIPE_FORMAT_U_I8 ***/ - -static void -i8_get_tile_rgba(ubyte *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = UBYTE_TO_FLOAT(*src); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_U_A8_L8 ***/ - -static void -a8_l8_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - ushort p = *src++; - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(p & 0xff); - pRow[3] = UBYTE_TO_FLOAT(p >> 8); - } - p += dst_stride; - } -} - - - - -/*** PIPE_FORMAT_Z32_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z32_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const double scale = 1.0 / (double) 0xffffffff; - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (float) (*src++ * scale); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_S8Z24_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -s8z24_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (float) (scale * (*src++ & 0xffffff)); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_Z24S8_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -z24s8_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (float) (scale * (*src++ >> 8)); - } - p += dst_stride; - } -} - - -void -pipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p) -{ - unsigned dst_stride = w * 4; - void *packed; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - packed = MALLOC(h * w * ps->cpp); - - if (!packed) - return; - - pipe_get_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); - - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - a1r5g5b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_A4R4G4B4_UNORM: - a4r4g4b4_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_R5G6B5_UNORM: - r5g6b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_U_L8: - l8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_U_A8: - a8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_U_I8: - i8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_U_A8_L8: - a8_l8_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_get_tile_rgba((short *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_Z16_UNORM: - z16_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_Z32_UNORM: - z32_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_S8Z24_UNORM: - s8z24_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_Z24S8_UNORM: - z24s8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - default: - assert(0); - } - - FREE(packed); -} - - -void -pipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p) -{ - unsigned src_stride = w * 4; - void *packed; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - packed = MALLOC(h * w * ps->cpp); - - if (!packed) - return; - - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - /*a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_R5G6B5_UNORM: - r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_R8G8B8A8_UNORM: - break; - case PIPE_FORMAT_U_L8: - /*l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_U_A8: - /*a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_U_I8: - /*i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_U_A8_L8: - /*a8_l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_Z16_UNORM: - /*z16_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_Z32_UNORM: - /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_S8Z24_UNORM: - /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_Z24S8_UNORM: - /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ - break; - default: - assert(0); - } - - pipe_put_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); - - FREE(packed); -} diff --git a/src/mesa/pipe/util/p_tile.h b/src/mesa/pipe/util/p_tile.h deleted file mode 100644 index cd8124bf11..0000000000 --- a/src/mesa/pipe/util/p_tile.h +++ /dev/null @@ -1,81 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef P_TILE_H -#define P_TILE_H - -#include "pipe/p_compiler.h" - -struct pipe_context; -struct pipe_surface; - - -/** - * Clip tile against surface dims. - * \return TRUE if tile is totally clipped, FALSE otherwise - */ -static INLINE boolean -pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps) -{ - if (x >= ps->width) - return TRUE; - if (y >= ps->height) - return TRUE; - if (x + *w > ps->width) - *w = ps->width - x; - if (y + *h > ps->height) - *h = ps->height - y; - return FALSE; -} - - -extern void -pipe_get_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride); - -extern void -pipe_put_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride); - - -extern void -pipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p); - -extern void -pipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p); - -#endif diff --git a/src/mesa/pipe/util/p_util.c b/src/mesa/pipe/util/p_util.c deleted file mode 100644 index 2a92f8e408..0000000000 --- a/src/mesa/pipe/util/p_util.c +++ /dev/null @@ -1,73 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Miscellaneous utility functions. - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" - - -/** - * Copy 2D rect from one place to another. - * Position and sizes are in pixels. - */ -void -pipe_copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte * src, - int src_pitch, - unsigned src_x, - int src_y) -{ - unsigned i; - - dst_pitch *= cpp; - src_pitch *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * src_pitch; - width *= cpp; - - if (width == dst_pitch && width == src_pitch) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; - } - } -} diff --git a/src/mesa/pipe/xlib/brw_aub.c b/src/mesa/pipe/xlib/brw_aub.c deleted file mode 100644 index 541d50c6e4..0000000000 --- a/src/mesa/pipe/xlib/brw_aub.c +++ /dev/null @@ -1,392 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#include "brw_aub.h" -#include "pipe/p_context.h" -#include "pipe/p_state.h" -#include "imports.h" -//#include "intel_winsys.h" - - -struct brw_aubfile { - FILE *file; - unsigned next_free_page; -}; - - -extern char *__progname; - - -struct aub_file_header { - unsigned int instruction_type; - unsigned int pad0:16; - unsigned int minor:8; - unsigned int major:8; - unsigned char application[8*4]; - unsigned int day:8; - unsigned int month:8; - unsigned int year:16; - unsigned int timezone:8; - unsigned int second:8; - unsigned int minute:8; - unsigned int hour:8; - unsigned int comment_length:16; - unsigned int pad1:16; -}; - -struct aub_block_header { - unsigned int instruction_type; - unsigned int operation:8; - unsigned int type:8; - unsigned int address_space:8; - unsigned int pad0:8; - unsigned int general_state_type:8; - unsigned int surface_state_type:8; - unsigned int pad1:16; - unsigned int address; - unsigned int length; -}; - -struct aub_dump_bmp { - unsigned int instruction_type; - unsigned int xmin:16; - unsigned int ymin:16; - unsigned int pitch:16; - unsigned int bpp:8; - unsigned int format:8; - unsigned int xsize:16; - unsigned int ysize:16; - unsigned int addr; - unsigned int unknown; -}; - -enum bh_operation { - BH_COMMENT, - BH_DATA_WRITE, - BH_COMMAND_WRITE, - BH_MMI0_WRITE32, - BH_END_SCENE, - BH_CONFIG_MEMORY_MAP, - BH_MAX_OPERATION -}; - -enum command_write_type { - CW_HWB_RING = 1, - CW_PRIMARY_RING_A, - CW_PRIMARY_RING_B, /* XXX - disagreement with listaub! */ - CW_PRIMARY_RING_C, - CW_MAX_TYPE -}; - -enum memory_map_type { - MM_DEFAULT, - MM_DYNAMIC, - MM_MAX_TYPE -}; - -enum address_space { - ADDR_GTT, - ADDR_LOCAL, - ADDR_MAIN, - ADDR_MAX -}; - - -#define AUB_FILE_HEADER 0xe085000b -#define AUB_BLOCK_HEADER 0xe0c10003 -#define AUB_DUMP_BMP 0xe09e0004 - -/* Registers to control page table - */ -#define PGETBL_CTL 0x2020 -#define PGETBL_ENABLED 0x1 - -#define NR_GTT_ENTRIES 65536 /* 256 mb */ - -#define FAIL \ -do { \ - fprintf(stderr, "failed to write aub data at %s/%d\n", __FUNCTION__, __LINE__); \ - exit(1); \ -} while (0) - - -/* Emit the headers at the top of each aubfile. Initialize the GTT. - */ -static void init_aubfile( FILE *aub_file ) -{ - struct aub_file_header fh; - struct aub_block_header bh; - unsigned int data; - - static int nr; - - nr++; - - /* Emit the aub header: - */ - memset(&fh, 0, sizeof(fh)); - - fh.instruction_type = AUB_FILE_HEADER; - fh.minor = 0x0; - fh.major = 0x7; - memcpy(fh.application, __progname, sizeof(fh.application)); - fh.day = (nr>>24) & 0xff; - fh.month = 0x0; - fh.year = 0x0; - fh.timezone = 0x0; - fh.second = nr & 0xff; - fh.minute = (nr>>8) & 0xff; - fh.hour = (nr>>16) & 0xff; - fh.comment_length = 0x0; - - if (fwrite(&fh, sizeof(fh), 1, aub_file) < 0) - FAIL; - - /* Setup the GTT starting at main memory address zero (!): - */ - memset(&bh, 0, sizeof(bh)); - - bh.instruction_type = AUB_BLOCK_HEADER; - bh.operation = BH_MMI0_WRITE32; - bh.type = 0x0; - bh.address_space = ADDR_GTT; /* ??? */ - bh.general_state_type = 0x0; - bh.surface_state_type = 0x0; - bh.address = PGETBL_CTL; - bh.length = 0x4; - - if (fwrite(&bh, sizeof(bh), 1, aub_file) < 0) - FAIL; - - data = 0x0 | PGETBL_ENABLED; - - if (fwrite(&data, sizeof(data), 1, aub_file) < 0) - FAIL; -} - - -static void init_aub_gtt( struct brw_aubfile *aubfile, - unsigned start_offset, - unsigned size ) -{ - FILE *aub_file = aubfile->file; - struct aub_block_header bh; - unsigned int i; - - assert(start_offset + size < NR_GTT_ENTRIES * 4096); - - - memset(&bh, 0, sizeof(bh)); - - bh.instruction_type = AUB_BLOCK_HEADER; - bh.operation = BH_DATA_WRITE; - bh.type = 0x0; - bh.address_space = ADDR_MAIN; - bh.general_state_type = 0x0; - bh.surface_state_type = 0x0; - bh.address = start_offset / 4096 * 4; - bh.length = size / 4096 * 4; - - if (fwrite(&bh, sizeof(bh), 1, aub_file) < 0) - FAIL; - - for (i = 0; i < size / 4096; i++) { - unsigned data = aubfile->next_free_page | 1; - - aubfile->next_free_page += 4096; - - if (fwrite(&data, sizeof(data), 1, aub_file) < 0) - FAIL; - } - -} - -static void write_block_header( FILE *aub_file, - struct aub_block_header *bh, - const unsigned *data, - unsigned sz ) -{ - sz = (sz + 3) & ~3; - - if (fwrite(bh, sizeof(*bh), 1, aub_file) < 0) - FAIL; - - if (fwrite(data, sz, 1, aub_file) < 0) - FAIL; - - fflush(aub_file); -} - - -static void write_dump_bmp( FILE *aub_file, - struct aub_dump_bmp *db ) -{ - if (fwrite(db, sizeof(*db), 1, aub_file) < 0) - FAIL; - - fflush(aub_file); -} - - - -void brw_aub_gtt_data( struct brw_aubfile *aubfile, - unsigned offset, - const void *data, - unsigned sz, - unsigned type, - unsigned state_type ) -{ - struct aub_block_header bh; - - bh.instruction_type = AUB_BLOCK_HEADER; - bh.operation = BH_DATA_WRITE; - bh.type = type; - bh.address_space = ADDR_GTT; - bh.pad0 = 0; - - if (type == DW_GENERAL_STATE) { - bh.general_state_type = state_type; - bh.surface_state_type = 0; - } - else { - bh.general_state_type = 0; - bh.surface_state_type = state_type; - } - - bh.pad1 = 0; - bh.address = offset; - bh.length = sz; - - write_block_header(aubfile->file, &bh, data, sz); -} - - - -void brw_aub_gtt_cmds( struct brw_aubfile *aubfile, - unsigned offset, - const void *data, - unsigned sz ) -{ - struct aub_block_header bh; - unsigned type = CW_PRIMARY_RING_A; - - - bh.instruction_type = AUB_BLOCK_HEADER; - bh.operation = BH_COMMAND_WRITE; - bh.type = type; - bh.address_space = ADDR_GTT; - bh.pad0 = 0; - bh.general_state_type = 0; - bh.surface_state_type = 0; - bh.pad1 = 0; - bh.address = offset; - bh.length = sz; - - write_block_header(aubfile->file, &bh, data, sz); -} - -void brw_aub_dump_bmp( struct brw_aubfile *aubfile, - struct pipe_surface *surface, - unsigned gtt_offset ) -{ - struct aub_dump_bmp db; - unsigned format; - - if (surface->cpp == 4) - format = 0x7; - else - format = 0x3; - - db.instruction_type = AUB_DUMP_BMP; - db.xmin = 0; - db.ymin = 0; - db.format = format; - db.bpp = surface->cpp * 8; - db.pitch = surface->pitch; - db.xsize = surface->width; - db.ysize = surface->height; - db.addr = gtt_offset; - db.unknown = /* surface->tiled ? 0x4 : */ 0x0; - - write_dump_bmp(aubfile->file, &db); -} - - - -struct brw_aubfile *brw_aubfile_create( void ) -{ - struct brw_aubfile *aubfile = CALLOC_STRUCT(brw_aubfile); - char filename[80]; - int val; - static int i = 0; - - i++; - - if (_mesa_getenv("INTEL_AUBFILE")) { - val = snprintf(filename, sizeof(filename), "%s%d.aub", _mesa_getenv("INTEL_AUBFILE"), i%4); - _mesa_printf("--> Aub file: %s\n", filename); - aubfile->file = fopen(filename, "w"); - } - else { - val = snprintf(filename, sizeof(filename), "%s.aub", __progname); - if (val < 0 || val > sizeof(filename)) - strcpy(filename, "default.aub"); - - _mesa_printf("--> Aub file: %s\n", filename); - aubfile->file = fopen(filename, "w"); - } - - if (!aubfile->file) { - _mesa_printf("couldn't open aubfile\n"); - exit(1); - } - - init_aubfile(aubfile->file); - - /* The GTT is located starting address zero in main memory. Pages - * to populate the gtt start after this point. - */ - aubfile->next_free_page = (NR_GTT_ENTRIES * 4 + 4095) & ~4095; - - /* More or less correspond with all the agp regions mapped by the - * driver: - */ - init_aub_gtt(aubfile, 0, 4096*4); - init_aub_gtt(aubfile, AUB_BUF_START, AUB_BUF_SIZE); - - return aubfile; -} - -void brw_aub_destroy( struct brw_aubfile *aubfile ) -{ - fclose(aubfile->file); - FREE(aubfile); -} diff --git a/src/mesa/pipe/xlib/brw_aub.h b/src/mesa/pipe/xlib/brw_aub.h deleted file mode 100644 index f5c60c7be2..0000000000 --- a/src/mesa/pipe/xlib/brw_aub.h +++ /dev/null @@ -1,114 +0,0 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - -#ifndef BRW_AUB_H -#define BRW_AUB_H - -/* We set up this region, buffers may be allocated here: - */ -#define AUB_BUF_START (4096*4) -#define AUB_BUF_SIZE (8*1024*1024) - -struct intel_context; -struct pipe_surface; - -struct brw_aubfile *brw_aubfile_create( void ); - -void brw_aub_destroy( struct brw_aubfile *aubfile ); - -void brw_aub_gtt_data( struct brw_aubfile *aubfile, - unsigned offset, - const void *data, - unsigned sz, - unsigned type, - unsigned state_type ); - -void brw_aub_gtt_cmds( struct brw_aubfile *aubfile, - unsigned offset, - const void *data, - unsigned sz ); - -void brw_aub_dump_bmp( struct brw_aubfile *aubfile, - struct pipe_surface *surface, - unsigned gtt_offset ); - - -enum data_write_type { - DW_NOTYPE, - DW_BATCH_BUFFER, - DW_BIN_BUFFER, - DW_BIN_POINTER_LIST, - DW_SLOW_STATE_BUFFER, - DW_VERTEX_BUFFER, - DW_2D_MAP, - DW_CUBE_MAP, - DW_INDIRECT_STATE_BUFFER, - DW_VOLUME_MAP, - DW_1D_MAP, - DW_CONSTANT_BUFFER, - DW_CONSTANT_URB_ENTRY, - DW_INDEX_BUFFER, - DW_GENERAL_STATE, - DW_SURFACE_STATE, - DW_MEDIA_OBJECT_INDIRECT_DATA, - DW_MAX_TYPE -}; - -enum data_write_general_state_type { - DWGS_NOTYPE, - DWGS_VERTEX_SHADER_STATE, - DWGS_GEOMETRY_SHADER_STATE , - DWGS_CLIPPER_STATE, - DWGS_STRIPS_FANS_STATE, - DWGS_WINDOWER_IZ_STATE, - DWGS_COLOR_CALC_STATE, - DWGS_CLIPPER_VIEWPORT_STATE, /* was 0x7 */ - DWGS_STRIPS_FANS_VIEWPORT_STATE, - DWGS_COLOR_CALC_VIEWPORT_STATE, /* was 0x9 */ - DWGS_SAMPLER_STATE, - DWGS_KERNEL_INSTRUCTIONS, - DWGS_SCRATCH_SPACE, - DWGS_SAMPLER_DEFAULT_COLOR, - DWGS_INTERFACE_DESCRIPTOR, - DWGS_VLD_STATE, - DWGS_VFE_STATE, - DWGS_MAX_TYPE -}; - -enum data_write_surface_state_type { - DWSS_NOTYPE, - DWSS_BINDING_TABLE_STATE, - DWSS_SURFACE_STATE, - DWSS_MAX_TYPE -}; - - -#endif diff --git a/src/mesa/pipe/xlib/fakeglx.c b/src/mesa/pipe/xlib/fakeglx.c deleted file mode 100644 index 902a755075..0000000000 --- a/src/mesa/pipe/xlib/fakeglx.c +++ /dev/null @@ -1,3188 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 7.1 - * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/* - * This is an emulation of the GLX API which allows Mesa/GLX-based programs - * to run on X servers which do not have the real GLX extension. - * - * Thanks to the contributors: - * - * Initial version: Philip Brown (phil@bolthole.com) - * Better glXGetConfig() support: Armin Liebchen (liebchen@asylum.cs.utah.edu) - * Further visual-handling refinements: Wolfram Gloger - * (wmglo@Dent.MED.Uni-Muenchen.DE). - * - * Notes: - * Don't be fooled, stereo isn't supported yet. - */ - - - -#include "glxheader.h" -#include "glxapi.h" -#include "GL/xmesa.h" -#include "context.h" -#include "config.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" -#include "version.h" -#include "xfonts.h" -#include "xmesaP.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_public.h" - - -#ifdef __VMS -#define _mesa_sprintf sprintf -#endif - -/* This indicates the client-side GLX API and GLX encoder version. */ -#define CLIENT_MAJOR_VERSION 1 -#define CLIENT_MINOR_VERSION 4 /* but don't have 1.3's pbuffers, etc yet */ - -/* This indicates the server-side GLX decoder version. - * GLX 1.4 indicates OpenGL 1.3 support - */ -#define SERVER_MAJOR_VERSION 1 -#define SERVER_MINOR_VERSION 4 - -/* This is appended onto the glXGetClient/ServerString version strings. */ -#define MESA_GLX_VERSION "Mesa " MESA_VERSION_STRING - -/* Who implemented this GLX? */ -#define VENDOR "Brian Paul" - -#define EXTENSIONS \ - "GLX_MESA_set_3dfx_mode " \ - "GLX_MESA_copy_sub_buffer " \ - "GLX_MESA_pixmap_colormap " \ - "GLX_MESA_release_buffers " \ - "GLX_ARB_get_proc_address " \ - "GLX_EXT_texture_from_pixmap " \ - "GLX_EXT_visual_info " \ - "GLX_EXT_visual_rating " \ - /*"GLX_SGI_video_sync "*/ \ - "GLX_SGIX_fbconfig " \ - "GLX_SGIX_pbuffer " - -/* - * Our fake GLX context will contain a "real" GLX context and an XMesa context. - * - * Note that a pointer to a __GLXcontext is a pointer to a fake_glx_context, - * and vice versa. - * - * We really just need this structure in order to make the libGL functions - * glXGetCurrentContext(), glXGetCurrentDrawable() and glXGetCurrentDisplay() - * work correctly. - */ -struct fake_glx_context { - __GLXcontext glxContext; /* this MUST be first! */ - XMesaContext xmesaContext; -}; - - - -/**********************************************************************/ -/*** GLX Visual Code ***/ -/**********************************************************************/ - -#define DONT_CARE -1 - - -static XMesaVisual *VisualTable = NULL; -static int NumVisuals = 0; - - -/* - * This struct and some code fragments borrowed - * from Mark Kilgard's GLUT library. - */ -typedef struct _OverlayInfo { - /* Avoid 64-bit portability problems by being careful to use - longs due to the way XGetWindowProperty is specified. Note - that these parameters are passed as CARD32s over X - protocol. */ - unsigned long overlay_visual; - long transparent_type; - long value; - long layer; -} OverlayInfo; - - - -/* Macro to handle c_class vs class field name in XVisualInfo struct */ -#if defined(__cplusplus) || defined(c_plusplus) -#define CLASS c_class -#else -#define CLASS class -#endif - - - -/* - * Test if the given XVisualInfo is usable for Mesa rendering. - */ -static GLboolean -is_usable_visual( XVisualInfo *vinfo ) -{ - switch (vinfo->CLASS) { - case StaticGray: - case GrayScale: - /* Any StaticGray/GrayScale visual works in RGB or CI mode */ - return GL_TRUE; - case StaticColor: - case PseudoColor: - /* Any StaticColor/PseudoColor visual of at least 4 bits */ - if (vinfo->depth>=4) { - return GL_TRUE; - } - else { - return GL_FALSE; - } - case TrueColor: - case DirectColor: - /* Any depth of TrueColor or DirectColor works in RGB mode */ - return GL_TRUE; - default: - /* This should never happen */ - return GL_FALSE; - } -} - - - -/** - * Get an array OverlayInfo records for specified screen. - * \param dpy the display - * \param screen screen number - * \param numOverlays returns numver of OverlayInfo records - * \return pointer to OverlayInfo array, free with XFree() - */ -static OverlayInfo * -GetOverlayInfo(Display *dpy, int screen, int *numOverlays) -{ - Atom overlayVisualsAtom; - Atom actualType; - Status status; - unsigned char *ovInfo; - unsigned long sizeData, bytesLeft; - int actualFormat; - - /* - * The SERVER_OVERLAY_VISUALS property on the root window contains - * a list of overlay visuals. Get that list now. - */ - overlayVisualsAtom = XInternAtom(dpy,"SERVER_OVERLAY_VISUALS", True); - if (overlayVisualsAtom == None) { - return 0; - } - - status = XGetWindowProperty(dpy, RootWindow(dpy, screen), - overlayVisualsAtom, 0L, (long) 10000, False, - overlayVisualsAtom, &actualType, &actualFormat, - &sizeData, &bytesLeft, - &ovInfo); - - if (status != Success || actualType != overlayVisualsAtom || - actualFormat != 32 || sizeData < 4) { - /* something went wrong */ - XFree((void *) ovInfo); - *numOverlays = 0; - return NULL; - } - - *numOverlays = sizeData / 4; - return (OverlayInfo *) ovInfo; -} - - - -/** - * Return the level (overlay, normal, underlay) of a given XVisualInfo. - * Input: dpy - the X display - * vinfo - the XVisualInfo to test - * Return: level of the visual: - * 0 = normal planes - * >0 = overlay planes - * <0 = underlay planes - */ -static int -level_of_visual( Display *dpy, XVisualInfo *vinfo ) -{ - OverlayInfo *overlay_info; - int numOverlaysPerScreen, i; - - overlay_info = GetOverlayInfo(dpy, vinfo->screen, &numOverlaysPerScreen); - if (!overlay_info) { - return 0; - } - - /* search the overlay visual list for the visual ID of interest */ - for (i = 0; i < numOverlaysPerScreen; i++) { - const OverlayInfo *ov = overlay_info + i; - if (ov->overlay_visual == vinfo->visualid) { - /* found the visual */ - if (/*ov->transparent_type==1 &&*/ ov->layer!=0) { - int level = ov->layer; - XFree((void *) overlay_info); - return level; - } - else { - XFree((void *) overlay_info); - return 0; - } - } - } - - /* The visual ID was not found in the overlay list. */ - XFree((void *) overlay_info); - return 0; -} - - - - -/* - * Given an XVisualInfo and RGB, Double, and Depth buffer flags, save the - * configuration in our list of GLX visuals. - */ -static XMesaVisual -save_glx_visual( Display *dpy, XVisualInfo *vinfo, - GLboolean rgbFlag, GLboolean alphaFlag, GLboolean dbFlag, - GLboolean stereoFlag, - GLint depth_size, GLint stencil_size, - GLint accumRedSize, GLint accumGreenSize, - GLint accumBlueSize, GLint accumAlphaSize, - GLint level, GLint numAuxBuffers ) -{ - GLboolean ximageFlag = GL_TRUE; - XMesaVisual xmvis; - GLint i; - GLboolean comparePointers; - - if (dbFlag) { - /* Check if the MESA_BACK_BUFFER env var is set */ - char *backbuffer = _mesa_getenv("MESA_BACK_BUFFER"); - if (backbuffer) { - if (backbuffer[0]=='p' || backbuffer[0]=='P') { - ximageFlag = GL_FALSE; - } - else if (backbuffer[0]=='x' || backbuffer[0]=='X') { - ximageFlag = GL_TRUE; - } - else { - _mesa_warning(NULL, "Mesa: invalid value for MESA_BACK_BUFFER environment variable, using an XImage."); - } - } - } - - if (stereoFlag) { - /* stereo not supported */ - return NULL; - } - - /* Comparing IDs uses less memory but sometimes fails. */ - /* XXX revisit this after 3.0 is finished. */ - if (_mesa_getenv("MESA_GLX_VISUAL_HACK")) - comparePointers = GL_TRUE; - else - comparePointers = GL_FALSE; - - /* Force the visual to have an alpha channel */ - if (rgbFlag && _mesa_getenv("MESA_GLX_FORCE_ALPHA")) - alphaFlag = GL_TRUE; - - /* First check if a matching visual is already in the list */ - for (i=0; idisplay == dpy - && v->mesa_visual.level == level - && v->mesa_visual.numAuxBuffers == numAuxBuffers - && v->ximage_flag == ximageFlag - && v->mesa_visual.rgbMode == rgbFlag - && v->mesa_visual.doubleBufferMode == dbFlag - && v->mesa_visual.stereoMode == stereoFlag - && (v->mesa_visual.alphaBits > 0) == alphaFlag - && (v->mesa_visual.depthBits >= depth_size || depth_size == 0) - && (v->mesa_visual.stencilBits >= stencil_size || stencil_size == 0) - && (v->mesa_visual.accumRedBits >= accumRedSize || accumRedSize == 0) - && (v->mesa_visual.accumGreenBits >= accumGreenSize || accumGreenSize == 0) - && (v->mesa_visual.accumBlueBits >= accumBlueSize || accumBlueSize == 0) - && (v->mesa_visual.accumAlphaBits >= accumAlphaSize || accumAlphaSize == 0)) { - /* now either compare XVisualInfo pointers or visual IDs */ - if ((!comparePointers && v->visinfo->visualid == vinfo->visualid) - || (comparePointers && v->vishandle == vinfo)) { - return v; - } - } - } - - /* Create a new visual and add it to the list. */ - - xmvis = XMesaCreateVisual( dpy, vinfo, rgbFlag, alphaFlag, dbFlag, - stereoFlag, ximageFlag, - depth_size, stencil_size, - accumRedSize, accumBlueSize, - accumBlueSize, accumAlphaSize, 0, level, - GLX_NONE_EXT ); - if (xmvis) { - /* Save a copy of the pointer now so we can find this visual again - * if we need to search for it in find_glx_visual(). - */ - xmvis->vishandle = vinfo; - /* Allocate more space for additional visual */ - VisualTable = (XMesaVisual *) _mesa_realloc( VisualTable, - sizeof(XMesaVisual) * NumVisuals, - sizeof(XMesaVisual) * (NumVisuals + 1)); - /* add xmvis to the list */ - VisualTable[NumVisuals] = xmvis; - NumVisuals++; - /* XXX minor hack, because XMesaCreateVisual doesn't support an - * aux buffers parameter. - */ - xmvis->mesa_visual.numAuxBuffers = numAuxBuffers; - } - return xmvis; -} - - -/** - * Return the default number of bits for the Z buffer. - * If defined, use the MESA_GLX_DEPTH_BITS env var value. - * Otherwise, use the DEFAULT_SOFTWARE_DEPTH_BITS constant. - * XXX probably do the same thing for stencil, accum, etc. - */ -static GLint -default_depth_bits(void) -{ - int zBits; - const char *zEnv = _mesa_getenv("MESA_GLX_DEPTH_BITS"); - if (zEnv) - zBits = _mesa_atoi(zEnv); - else - zBits = DEFAULT_SOFTWARE_DEPTH_BITS; - return zBits; -} - -static GLint -default_alpha_bits(void) -{ - int aBits; - const char *aEnv = _mesa_getenv("MESA_GLX_ALPHA_BITS"); - if (aEnv) - aBits = _mesa_atoi(aEnv); - else - aBits = 0; - return aBits; -} - -static GLint -default_accum_bits(void) -{ - return 16; -} - - - -/* - * Create a GLX visual from a regular XVisualInfo. - * This is called when Fake GLX is given an XVisualInfo which wasn't - * returned by glXChooseVisual. Since this is the first time we're - * considering this visual we'll take a guess at reasonable values - * for depth buffer size, stencil size, accum size, etc. - * This is the best we can do with a client-side emulation of GLX. - */ -static XMesaVisual -create_glx_visual( Display *dpy, XVisualInfo *visinfo ) -{ - int vislevel; - GLint zBits = 24; /*default_depth_bits();*/ - GLint accBits = default_accum_bits(); - GLboolean alphaFlag = default_alpha_bits() > 0; - - vislevel = level_of_visual( dpy, visinfo ); - if (vislevel) { - /* Configure this visual as a CI, single-buffered overlay */ - return save_glx_visual( dpy, visinfo, - GL_FALSE, /* rgb */ - GL_FALSE, /* alpha */ - GL_FALSE, /* double */ - GL_FALSE, /* stereo */ - 0, /* depth bits */ - 0, /* stencil bits */ - 0,0,0,0, /* accum bits */ - vislevel, /* level */ - 0 /* numAux */ - ); - } - else if (is_usable_visual( visinfo )) { - if (_mesa_getenv("MESA_GLX_FORCE_CI")) { - /* Configure this visual as a COLOR INDEX visual. */ - return save_glx_visual( dpy, visinfo, - GL_FALSE, /* rgb */ - GL_FALSE, /* alpha */ - GL_TRUE, /* double */ - GL_FALSE, /* stereo */ - zBits, - STENCIL_BITS, - 0, 0, 0, 0, /* accum bits */ - 0, /* level */ - 0 /* numAux */ - ); - } - else { - /* Configure this visual as RGB, double-buffered, depth-buffered. */ - /* This is surely wrong for some people's needs but what else */ - /* can be done? They should use glXChooseVisual(). */ - return save_glx_visual( dpy, visinfo, - GL_TRUE, /* rgb */ - alphaFlag, /* alpha */ - GL_TRUE, /* double */ - GL_FALSE, /* stereo */ - zBits, - STENCIL_BITS, - accBits, /* r */ - accBits, /* g */ - accBits, /* b */ - accBits, /* a */ - 0, /* level */ - 0 /* numAux */ - ); - } - } - else { - _mesa_warning(NULL, "Mesa: error in glXCreateContext: bad visual\n"); - return NULL; - } -} - - - -/* - * Find the GLX visual associated with an XVisualInfo. - */ -static XMesaVisual -find_glx_visual( Display *dpy, XVisualInfo *vinfo ) -{ - int i; - - /* try to match visual id */ - for (i=0;idisplay==dpy - && VisualTable[i]->visinfo->visualid == vinfo->visualid) { - return VisualTable[i]; - } - } - - /* if that fails, try to match pointers */ - for (i=0;idisplay==dpy && VisualTable[i]->vishandle==vinfo) { - return VisualTable[i]; - } - } - - return NULL; -} - - - -/** - * Return the transparent pixel value for a GLX visual. - * Input: glxvis - the glx_visual - * Return: a pixel value or -1 if no transparent pixel - */ -static int -transparent_pixel( XMesaVisual glxvis ) -{ - Display *dpy = glxvis->display; - XVisualInfo *vinfo = glxvis->visinfo; - OverlayInfo *overlay_info; - int numOverlaysPerScreen, i; - - overlay_info = GetOverlayInfo(dpy, vinfo->screen, &numOverlaysPerScreen); - if (!overlay_info) { - return -1; - } - - for (i = 0; i < numOverlaysPerScreen; i++) { - const OverlayInfo *ov = overlay_info + i; - if (ov->overlay_visual == vinfo->visualid) { - /* found it! */ - if (ov->transparent_type == 0) { - /* type 0 indicates no transparency */ - XFree((void *) overlay_info); - return -1; - } - else { - /* ov->value is the transparent pixel */ - XFree((void *) overlay_info); - return ov->value; - } - } - } - - /* The visual ID was not found in the overlay list. */ - XFree((void *) overlay_info); - return -1; -} - - - -/** - * Try to get an X visual which matches the given arguments. - */ -static XVisualInfo * -get_visual( Display *dpy, int scr, unsigned int depth, int xclass ) -{ - XVisualInfo temp, *vis; - long mask; - int n; - unsigned int default_depth; - int default_class; - - mask = VisualScreenMask | VisualDepthMask | VisualClassMask; - temp.screen = scr; - temp.depth = depth; - temp.CLASS = xclass; - - default_depth = DefaultDepth(dpy,scr); - default_class = DefaultVisual(dpy,scr)->CLASS; - - if (depth==default_depth && xclass==default_class) { - /* try to get root window's visual */ - temp.visualid = DefaultVisual(dpy,scr)->visualid; - mask |= VisualIDMask; - } - - vis = XGetVisualInfo( dpy, mask, &temp, &n ); - - /* In case bits/pixel > 24, make sure color channels are still <=8 bits. - * An SGI Infinite Reality system, for example, can have 30bpp pixels: - * 10 bits per color channel. Mesa's limited to a max of 8 bits/channel. - */ - if (vis && depth > 24 && (xclass==TrueColor || xclass==DirectColor)) { - if (_mesa_bitcount((GLuint) vis->red_mask ) <= 8 && - _mesa_bitcount((GLuint) vis->green_mask) <= 8 && - _mesa_bitcount((GLuint) vis->blue_mask ) <= 8) { - return vis; - } - else { - XFree((void *) vis); - return NULL; - } - } - - return vis; -} - - - -/* - * Retrieve the value of the given environment variable and find - * the X visual which matches it. - * Input: dpy - the display - * screen - the screen number - * varname - the name of the environment variable - * Return: an XVisualInfo pointer to NULL if error. - */ -static XVisualInfo * -get_env_visual(Display *dpy, int scr, const char *varname) -{ - char value[100], type[100]; - int depth, xclass = -1; - XVisualInfo *vis; - - if (!_mesa_getenv( varname )) { - return NULL; - } - - _mesa_strncpy( value, _mesa_getenv(varname), 100 ); - value[99] = 0; - - sscanf( value, "%s %d", type, &depth ); - - if (_mesa_strcmp(type,"TrueColor")==0) xclass = TrueColor; - else if (_mesa_strcmp(type,"DirectColor")==0) xclass = DirectColor; - else if (_mesa_strcmp(type,"PseudoColor")==0) xclass = PseudoColor; - else if (_mesa_strcmp(type,"StaticColor")==0) xclass = StaticColor; - else if (_mesa_strcmp(type,"GrayScale")==0) xclass = GrayScale; - else if (_mesa_strcmp(type,"StaticGray")==0) xclass = StaticGray; - - if (xclass>-1 && depth>0) { - vis = get_visual( dpy, scr, depth, xclass ); - if (vis) { - return vis; - } - } - - _mesa_warning(NULL, "GLX unable to find visual class=%s, depth=%d.", - type, depth); - - return NULL; -} - - - -/* - * Select an X visual which satisfies the RGBA/CI flag and minimum depth. - * Input: dpy, screen - X display and screen number - * rgba - GL_TRUE = RGBA mode, GL_FALSE = CI mode - * min_depth - minimum visual depth - * preferred_class - preferred GLX visual class or DONT_CARE - * Return: pointer to an XVisualInfo or NULL. - */ -static XVisualInfo * -choose_x_visual( Display *dpy, int screen, GLboolean rgba, int min_depth, - int preferred_class ) -{ - XVisualInfo *vis; - int xclass, visclass = 0; - int depth; - - if (rgba) { - Atom hp_cr_maps = XInternAtom(dpy, "_HP_RGB_SMOOTH_MAP_LIST", True); - /* First see if the MESA_RGB_VISUAL env var is defined */ - vis = get_env_visual( dpy, screen, "MESA_RGB_VISUAL" ); - if (vis) { - return vis; - } - /* Otherwise, search for a suitable visual */ - if (preferred_class==DONT_CARE) { - for (xclass=0;xclass<6;xclass++) { - switch (xclass) { - case 0: visclass = TrueColor; break; - case 1: visclass = DirectColor; break; - case 2: visclass = PseudoColor; break; - case 3: visclass = StaticColor; break; - case 4: visclass = GrayScale; break; - case 5: visclass = StaticGray; break; - } - if (min_depth==0) { - /* start with shallowest */ - for (depth=0;depth<=32;depth++) { - if (visclass==TrueColor && depth==8 && !hp_cr_maps) { - /* Special case: try to get 8-bit PseudoColor before */ - /* 8-bit TrueColor */ - vis = get_visual( dpy, screen, 8, PseudoColor ); - if (vis) { - return vis; - } - } - vis = get_visual( dpy, screen, depth, visclass ); - if (vis) { - return vis; - } - } - } - else { - /* start with deepest */ - for (depth=32;depth>=min_depth;depth--) { - if (visclass==TrueColor && depth==8 && !hp_cr_maps) { - /* Special case: try to get 8-bit PseudoColor before */ - /* 8-bit TrueColor */ - vis = get_visual( dpy, screen, 8, PseudoColor ); - if (vis) { - return vis; - } - } - vis = get_visual( dpy, screen, depth, visclass ); - if (vis) { - return vis; - } - } - } - } - } - else { - /* search for a specific visual class */ - switch (preferred_class) { - case GLX_TRUE_COLOR_EXT: visclass = TrueColor; break; - case GLX_DIRECT_COLOR_EXT: visclass = DirectColor; break; - case GLX_PSEUDO_COLOR_EXT: visclass = PseudoColor; break; - case GLX_STATIC_COLOR_EXT: visclass = StaticColor; break; - case GLX_GRAY_SCALE_EXT: visclass = GrayScale; break; - case GLX_STATIC_GRAY_EXT: visclass = StaticGray; break; - default: return NULL; - } - if (min_depth==0) { - /* start with shallowest */ - for (depth=0;depth<=32;depth++) { - vis = get_visual( dpy, screen, depth, visclass ); - if (vis) { - return vis; - } - } - } - else { - /* start with deepest */ - for (depth=32;depth>=min_depth;depth--) { - vis = get_visual( dpy, screen, depth, visclass ); - if (vis) { - return vis; - } - } - } - } - } - else { - /* First see if the MESA_CI_VISUAL env var is defined */ - vis = get_env_visual( dpy, screen, "MESA_CI_VISUAL" ); - if (vis) { - return vis; - } - /* Otherwise, search for a suitable visual, starting with shallowest */ - if (preferred_class==DONT_CARE) { - for (xclass=0;xclass<4;xclass++) { - switch (xclass) { - case 0: visclass = PseudoColor; break; - case 1: visclass = StaticColor; break; - case 2: visclass = GrayScale; break; - case 3: visclass = StaticGray; break; - } - /* try 8-bit up through 16-bit */ - for (depth=8;depth<=16;depth++) { - vis = get_visual( dpy, screen, depth, visclass ); - if (vis) { - return vis; - } - } - /* try min_depth up to 8-bit */ - for (depth=min_depth;depth<8;depth++) { - vis = get_visual( dpy, screen, depth, visclass ); - if (vis) { - return vis; - } - } - } - } - else { - /* search for a specific visual class */ - switch (preferred_class) { - case GLX_TRUE_COLOR_EXT: visclass = TrueColor; break; - case GLX_DIRECT_COLOR_EXT: visclass = DirectColor; break; - case GLX_PSEUDO_COLOR_EXT: visclass = PseudoColor; break; - case GLX_STATIC_COLOR_EXT: visclass = StaticColor; break; - case GLX_GRAY_SCALE_EXT: visclass = GrayScale; break; - case GLX_STATIC_GRAY_EXT: visclass = StaticGray; break; - default: return NULL; - } - /* try 8-bit up through 16-bit */ - for (depth=8;depth<=16;depth++) { - vis = get_visual( dpy, screen, depth, visclass ); - if (vis) { - return vis; - } - } - /* try min_depth up to 8-bit */ - for (depth=min_depth;depth<8;depth++) { - vis = get_visual( dpy, screen, depth, visclass ); - if (vis) { - return vis; - } - } - } - } - - /* didn't find a visual */ - return NULL; -} - - - -/* - * Find the deepest X over/underlay visual of at least min_depth. - * Input: dpy, screen - X display and screen number - * level - the over/underlay level - * trans_type - transparent pixel type: GLX_NONE_EXT, - * GLX_TRANSPARENT_RGB_EXT, GLX_TRANSPARENT_INDEX_EXT, - * or DONT_CARE - * trans_value - transparent pixel value or DONT_CARE - * min_depth - minimum visual depth - * preferred_class - preferred GLX visual class or DONT_CARE - * Return: pointer to an XVisualInfo or NULL. - */ -static XVisualInfo * -choose_x_overlay_visual( Display *dpy, int scr, GLboolean rgbFlag, - int level, int trans_type, int trans_value, - int min_depth, int preferred_class ) -{ - OverlayInfo *overlay_info; - int numOverlaysPerScreen; - int i; - XVisualInfo *deepvis; - int deepest; - - /*DEBUG int tt, tv; */ - - switch (preferred_class) { - case GLX_TRUE_COLOR_EXT: preferred_class = TrueColor; break; - case GLX_DIRECT_COLOR_EXT: preferred_class = DirectColor; break; - case GLX_PSEUDO_COLOR_EXT: preferred_class = PseudoColor; break; - case GLX_STATIC_COLOR_EXT: preferred_class = StaticColor; break; - case GLX_GRAY_SCALE_EXT: preferred_class = GrayScale; break; - case GLX_STATIC_GRAY_EXT: preferred_class = StaticGray; break; - default: preferred_class = DONT_CARE; - } - - overlay_info = GetOverlayInfo(dpy, scr, &numOverlaysPerScreen); - if (!overlay_info) { - return NULL; - } - - /* Search for the deepest overlay which satisifies all criteria. */ - deepest = min_depth; - deepvis = NULL; - - for (i = 0; i < numOverlaysPerScreen; i++) { - const OverlayInfo *ov = overlay_info + i; - XVisualInfo *vislist, vistemplate; - int count; - - if (ov->layer!=level) { - /* failed overlay level criteria */ - continue; - } - if (!(trans_type==DONT_CARE - || (trans_type==GLX_TRANSPARENT_INDEX_EXT - && ov->transparent_type>0) - || (trans_type==GLX_NONE_EXT && ov->transparent_type==0))) { - /* failed transparent pixel type criteria */ - continue; - } - if (trans_value!=DONT_CARE && trans_value!=ov->value) { - /* failed transparent pixel value criteria */ - continue; - } - - /* get XVisualInfo and check the depth */ - vistemplate.visualid = ov->overlay_visual; - vistemplate.screen = scr; - vislist = XGetVisualInfo( dpy, VisualIDMask | VisualScreenMask, - &vistemplate, &count ); - - if (count!=1) { - /* something went wrong */ - continue; - } - if (preferred_class!=DONT_CARE && preferred_class!=vislist->CLASS) { - /* wrong visual class */ - continue; - } - - /* if RGB was requested, make sure we have True/DirectColor */ - if (rgbFlag && vislist->CLASS != TrueColor - && vislist->CLASS != DirectColor) - continue; - - /* if CI was requested, make sure we have a color indexed visual */ - if (!rgbFlag - && (vislist->CLASS == TrueColor || vislist->CLASS == DirectColor)) - continue; - - if (deepvis==NULL || vislist->depth > deepest) { - /* YES! found a satisfactory visual */ - if (deepvis) { - XFree( deepvis ); - } - deepest = vislist->depth; - deepvis = vislist; - /* DEBUG tt = ov->transparent_type;*/ - /* DEBUG tv = ov->value; */ - } - } - -/*DEBUG - if (deepvis) { - printf("chose 0x%x: layer=%d depth=%d trans_type=%d trans_value=%d\n", - deepvis->visualid, level, deepvis->depth, tt, tv ); - } -*/ - return deepvis; -} - - -/**********************************************************************/ -/*** Display-related functions ***/ -/**********************************************************************/ - - -/** - * Free all XMesaVisuals which are associated with the given display. - */ -static void -destroy_visuals_on_display(Display *dpy) -{ - int i; - for (i = 0; i < NumVisuals; i++) { - if (VisualTable[i]->display == dpy) { - /* remove this visual */ - int j; - free(VisualTable[i]); - for (j = i; j < NumVisuals - 1; j++) - VisualTable[j] = VisualTable[j + 1]; - NumVisuals--; - } - } -} - - -/** - * Called from XCloseDisplay() to let us free our display-related data. - */ -static int -close_display_callback(Display *dpy, XExtCodes *codes) -{ - destroy_visuals_on_display(dpy); - xmesa_destroy_buffers_on_display(dpy); - return 0; -} - - -/** - * Look for the named extension on given display and return a pointer - * to the _XExtension data, or NULL if extension not found. - */ -static _XExtension * -lookup_extension(Display *dpy, const char *extName) -{ - _XExtension *ext; - for (ext = dpy->ext_procs; ext; ext = ext->next) { - if (ext->name && strcmp(ext->name, extName) == 0) { - return ext; - } - } - return NULL; -} - - -/** - * Whenever we're given a new Display pointer, call this function to - * register our close_display_callback function. - */ -static void -register_with_display(Display *dpy) -{ - const char *extName = "MesaGLX"; - _XExtension *ext; - - ext = lookup_extension(dpy, extName); - if (!ext) { - XExtCodes *c = XAddExtension(dpy); - ext = dpy->ext_procs; /* new extension is at head of list */ - assert(c->extension == ext->codes.extension); - ext->name = _mesa_strdup(extName); - ext->close_display = close_display_callback; - } -} - - -/**********************************************************************/ -/*** Begin Fake GLX API Functions ***/ -/**********************************************************************/ - - -/** - * Helper used by glXChooseVisual and glXChooseFBConfig. - * The fbConfig parameter must be GL_FALSE for the former and GL_TRUE for - * the later. - * In either case, the attribute list is terminated with the value 'None'. - */ -static XMesaVisual -choose_visual( Display *dpy, int screen, const int *list, GLboolean fbConfig ) -{ - const GLboolean rgbModeDefault = fbConfig; - const int *parselist; - XVisualInfo *vis; - int min_ci = 0; - int min_red=0, min_green=0, min_blue=0; - GLboolean rgb_flag = rgbModeDefault; - GLboolean alpha_flag = GL_FALSE; - GLboolean double_flag = GL_FALSE; - GLboolean stereo_flag = GL_FALSE; - GLint depth_size = 0; - GLint stencil_size = 0; - GLint accumRedSize = 0; - GLint accumGreenSize = 0; - GLint accumBlueSize = 0; - GLint accumAlphaSize = 0; - int level = 0; - int visual_type = DONT_CARE; - int trans_type = DONT_CARE; - int trans_value = DONT_CARE; - GLint caveat = DONT_CARE; - XMesaVisual xmvis = NULL; - int desiredVisualID = -1; - int numAux = 0; - - parselist = list; - - while (*parselist) { - - switch (*parselist) { - case GLX_USE_GL: - if (fbConfig) { - /* invalid token */ - return NULL; - } - else { - /* skip */ - parselist++; - } - break; - case GLX_BUFFER_SIZE: - parselist++; - min_ci = *parselist++; - break; - case GLX_LEVEL: - parselist++; - level = *parselist++; - break; - case GLX_RGBA: - if (fbConfig) { - /* invalid token */ - return NULL; - } - else { - rgb_flag = GL_TRUE; - parselist++; - } - break; - case GLX_DOUBLEBUFFER: - parselist++; - if (fbConfig) { - double_flag = *parselist++; - } - else { - double_flag = GL_TRUE; - } - break; - case GLX_STEREO: - parselist++; - if (fbConfig) { - stereo_flag = *parselist++; - } - else { - stereo_flag = GL_TRUE; - } - break; - case GLX_AUX_BUFFERS: - parselist++; - numAux = *parselist++; - if (numAux > MAX_AUX_BUFFERS) - return NULL; - break; - case GLX_RED_SIZE: - parselist++; - min_red = *parselist++; - break; - case GLX_GREEN_SIZE: - parselist++; - min_green = *parselist++; - break; - case GLX_BLUE_SIZE: - parselist++; - min_blue = *parselist++; - break; - case GLX_ALPHA_SIZE: - parselist++; - { - GLint size = *parselist++; - alpha_flag = size ? GL_TRUE : GL_FALSE; - } - break; - case GLX_DEPTH_SIZE: - parselist++; - depth_size = *parselist++; - break; - case GLX_STENCIL_SIZE: - parselist++; - stencil_size = *parselist++; - break; - case GLX_ACCUM_RED_SIZE: - parselist++; - { - GLint size = *parselist++; - accumRedSize = MAX2( accumRedSize, size ); - } - break; - case GLX_ACCUM_GREEN_SIZE: - parselist++; - { - GLint size = *parselist++; - accumGreenSize = MAX2( accumGreenSize, size ); - } - break; - case GLX_ACCUM_BLUE_SIZE: - parselist++; - { - GLint size = *parselist++; - accumBlueSize = MAX2( accumBlueSize, size ); - } - break; - case GLX_ACCUM_ALPHA_SIZE: - parselist++; - { - GLint size = *parselist++; - accumAlphaSize = MAX2( accumAlphaSize, size ); - } - break; - - /* - * GLX_EXT_visual_info extension - */ - case GLX_X_VISUAL_TYPE_EXT: - parselist++; - visual_type = *parselist++; - break; - case GLX_TRANSPARENT_TYPE_EXT: - parselist++; - trans_type = *parselist++; - break; - case GLX_TRANSPARENT_INDEX_VALUE_EXT: - parselist++; - trans_value = *parselist++; - break; - case GLX_TRANSPARENT_RED_VALUE_EXT: - case GLX_TRANSPARENT_GREEN_VALUE_EXT: - case GLX_TRANSPARENT_BLUE_VALUE_EXT: - case GLX_TRANSPARENT_ALPHA_VALUE_EXT: - /* ignore */ - parselist++; - parselist++; - break; - - /* - * GLX_EXT_visual_info extension - */ - case GLX_VISUAL_CAVEAT_EXT: - parselist++; - caveat = *parselist++; /* ignored for now */ - break; - - /* - * GLX_ARB_multisample - */ - case GLX_SAMPLE_BUFFERS_ARB: - /* ms not supported */ - return NULL; - case GLX_SAMPLES_ARB: - /* ms not supported */ - return NULL; - - /* - * FBConfig attribs. - */ - case GLX_RENDER_TYPE: - if (!fbConfig) - return NULL; - parselist++; - if (*parselist == GLX_RGBA_BIT) { - rgb_flag = GL_TRUE; - } - else if (*parselist == GLX_COLOR_INDEX_BIT) { - rgb_flag = GL_FALSE; - } - else if (*parselist == 0) { - rgb_flag = GL_TRUE; - } - parselist++; - break; - case GLX_DRAWABLE_TYPE: - if (!fbConfig) - return NULL; - parselist++; - if (*parselist & ~(GLX_WINDOW_BIT | GLX_PIXMAP_BIT | GLX_PBUFFER_BIT)) { - return NULL; /* bad bit */ - } - parselist++; - break; - case GLX_FBCONFIG_ID: - if (!fbConfig) - return NULL; - parselist++; - desiredVisualID = *parselist++; - break; - case GLX_X_RENDERABLE: - if (!fbConfig) - return NULL; - parselist += 2; - /* ignore */ - break; - -#ifdef GLX_EXT_texture_from_pixmap - case GLX_BIND_TO_TEXTURE_RGB_EXT: - parselist++; /*skip*/ - break; - case GLX_BIND_TO_TEXTURE_RGBA_EXT: - parselist++; /*skip*/ - break; - case GLX_BIND_TO_MIPMAP_TEXTURE_EXT: - parselist++; /*skip*/ - break; - case GLX_BIND_TO_TEXTURE_TARGETS_EXT: - parselist++; - if (*parselist & ~(GLX_TEXTURE_1D_BIT_EXT | - GLX_TEXTURE_2D_BIT_EXT | - GLX_TEXTURE_RECTANGLE_BIT_EXT)) { - /* invalid bit */ - return NULL; - } - break; - case GLX_Y_INVERTED_EXT: - parselist++; /*skip*/ - break; -#endif - - case None: - /* end of list */ - break; - - default: - /* undefined attribute */ - _mesa_warning(NULL, "unexpected attrib 0x%x in choose_visual()", - *parselist); - return NULL; - } - } - - (void) caveat; - - /* - * Since we're only simulating the GLX extension this function will never - * find any real GL visuals. Instead, all we can do is try to find an RGB - * or CI visual of appropriate depth. Other requested attributes such as - * double buffering, depth buffer, etc. will be associated with the X - * visual and stored in the VisualTable[]. - */ - if (desiredVisualID != -1) { - /* try to get a specific visual, by visualID */ - XVisualInfo temp; - int n; - temp.visualid = desiredVisualID; - temp.screen = screen; - vis = XGetVisualInfo(dpy, VisualIDMask | VisualScreenMask, &temp, &n); - if (vis) { - /* give the visual some useful GLX attributes */ - double_flag = GL_TRUE; - if (vis->depth > 8) - rgb_flag = GL_TRUE; - depth_size = 24; /*default_depth_bits();*/ - stencil_size = STENCIL_BITS; - /* XXX accum??? */ - } - } - else if (level==0) { - /* normal color planes */ - if (rgb_flag) { - /* Get an RGB visual */ - int min_rgb = min_red + min_green + min_blue; - if (min_rgb>1 && min_rgb<8) { - /* a special case to be sure we can get a monochrome visual */ - min_rgb = 1; - } - vis = choose_x_visual( dpy, screen, rgb_flag, min_rgb, visual_type ); - } - else { - /* Get a color index visual */ - vis = choose_x_visual( dpy, screen, rgb_flag, min_ci, visual_type ); - accumRedSize = accumGreenSize = accumBlueSize = accumAlphaSize = 0; - } - } - else { - /* over/underlay planes */ - if (rgb_flag) { - /* rgba overlay */ - int min_rgb = min_red + min_green + min_blue; - if (min_rgb>1 && min_rgb<8) { - /* a special case to be sure we can get a monochrome visual */ - min_rgb = 1; - } - vis = choose_x_overlay_visual( dpy, screen, rgb_flag, level, - trans_type, trans_value, min_rgb, visual_type ); - } - else { - /* color index overlay */ - vis = choose_x_overlay_visual( dpy, screen, rgb_flag, level, - trans_type, trans_value, min_ci, visual_type ); - } - } - - if (vis) { - /* Note: we're not exactly obeying the glXChooseVisual rules here. - * When GLX_DEPTH_SIZE = 1 is specified we're supposed to choose the - * largest depth buffer size, which is 32bits/value. Instead, we - * return 16 to maintain performance with earlier versions of Mesa. - */ - if (stencil_size > 0) - depth_size = 24; /* if Z and stencil, always use 24+8 format */ - else if (depth_size > 24) - depth_size = 32; - else if (depth_size > 16) - depth_size = 24; - else if (depth_size > 0) { - depth_size = default_depth_bits(); - } - - if (!alpha_flag) { - alpha_flag = default_alpha_bits() > 0; - } - - /* we only support one size of stencil and accum buffers. */ - if (stencil_size > 0) - stencil_size = STENCIL_BITS; - if (accumRedSize > 0 || accumGreenSize > 0 || accumBlueSize > 0 || - accumAlphaSize > 0) { - accumRedSize = - accumGreenSize = - accumBlueSize = default_accum_bits(); - accumAlphaSize = alpha_flag ? accumRedSize : 0; - } - - xmvis = save_glx_visual( dpy, vis, rgb_flag, alpha_flag, double_flag, - stereo_flag, depth_size, stencil_size, - accumRedSize, accumGreenSize, - accumBlueSize, accumAlphaSize, level, numAux ); - } - - return xmvis; -} - - -static XVisualInfo * -Fake_glXChooseVisual( Display *dpy, int screen, int *list ) -{ - XMesaVisual xmvis; - - /* register ourselves as an extension on this display */ - register_with_display(dpy); - - xmvis = choose_visual(dpy, screen, list, GL_FALSE); - if (xmvis) { -#if 0 - return xmvis->vishandle; -#else - /* create a new vishandle - the cached one may be stale */ - xmvis->vishandle = (XVisualInfo *) _mesa_malloc(sizeof(XVisualInfo)); - if (xmvis->vishandle) { - _mesa_memcpy(xmvis->vishandle, xmvis->visinfo, sizeof(XVisualInfo)); - } - return xmvis->vishandle; -#endif - } - else - return NULL; -} - - -static GLXContext -Fake_glXCreateContext( Display *dpy, XVisualInfo *visinfo, - GLXContext share_list, Bool direct ) -{ - XMesaVisual xmvis; - struct fake_glx_context *glxCtx; - struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list; - - if (!dpy || !visinfo) - return 0; - - glxCtx = CALLOC_STRUCT(fake_glx_context); - if (!glxCtx) - return 0; - - /* deallocate unused windows/buffers */ -#if 0 - XMesaGarbageCollect(); -#endif - - xmvis = find_glx_visual( dpy, visinfo ); - if (!xmvis) { - /* This visual wasn't found with glXChooseVisual() */ - xmvis = create_glx_visual( dpy, visinfo ); - if (!xmvis) { - /* unusable visual */ - _mesa_free(glxCtx); - return NULL; - } - } - - glxCtx->xmesaContext = XMesaCreateContext(xmvis, - shareCtx ? shareCtx->xmesaContext : NULL); - if (!glxCtx->xmesaContext) { - _mesa_free(glxCtx); - return NULL; - } - - glxCtx->glxContext.isDirect = GL_FALSE; - glxCtx->glxContext.currentDpy = dpy; - glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ - - assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); - - return (GLXContext) glxCtx; -} - - -/* XXX these may have to be removed due to thread-safety issues. */ -static GLXContext MakeCurrent_PrevContext = 0; -static GLXDrawable MakeCurrent_PrevDrawable = 0; -static GLXDrawable MakeCurrent_PrevReadable = 0; -static XMesaBuffer MakeCurrent_PrevDrawBuffer = 0; -static XMesaBuffer MakeCurrent_PrevReadBuffer = 0; - - -/* GLX 1.3 and later */ -static Bool -Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, - GLXDrawable read, GLXContext ctx ) -{ - struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; - - if (ctx && draw && read) { - XMesaBuffer drawBuffer, readBuffer; - XMesaContext xmctx = glxCtx->xmesaContext; - - /* Find the XMesaBuffer which corresponds to the GLXDrawable 'draw' */ - if (ctx == MakeCurrent_PrevContext - && draw == MakeCurrent_PrevDrawable) { - drawBuffer = MakeCurrent_PrevDrawBuffer; - } - else { - drawBuffer = XMesaFindBuffer( dpy, draw ); - } - if (!drawBuffer) { - /* drawable must be a new window! */ - drawBuffer = XMesaCreateWindowBuffer( xmctx->xm_visual, draw ); - if (!drawBuffer) { - /* Out of memory, or context/drawable depth mismatch */ - return False; - } -#ifdef FX - FXcreateContext( xmctx->xm_visual, draw, xmctx, drawBuffer ); -#endif - } - - /* Find the XMesaBuffer which corresponds to the GLXDrawable 'read' */ - if (ctx == MakeCurrent_PrevContext - && read == MakeCurrent_PrevReadable) { - readBuffer = MakeCurrent_PrevReadBuffer; - } - else { - readBuffer = XMesaFindBuffer( dpy, read ); - } - if (!readBuffer) { - /* drawable must be a new window! */ - readBuffer = XMesaCreateWindowBuffer( xmctx->xm_visual, read ); - if (!readBuffer) { - /* Out of memory, or context/drawable depth mismatch */ - return False; - } -#ifdef FX - FXcreateContext( xmctx->xm_visual, read, xmctx, readBuffer ); -#endif - } - - MakeCurrent_PrevContext = ctx; - MakeCurrent_PrevDrawable = draw; - MakeCurrent_PrevReadable = read; - MakeCurrent_PrevDrawBuffer = drawBuffer; - MakeCurrent_PrevReadBuffer = readBuffer; - - /* Now make current! */ - if (XMesaMakeCurrent2(xmctx, drawBuffer, readBuffer)) { - ((__GLXcontext *) ctx)->currentDpy = dpy; - ((__GLXcontext *) ctx)->currentDrawable = draw; - ((__GLXcontext *) ctx)->currentReadable = read; - return True; - } - else { - return False; - } - } - else if (!ctx && !draw && !read) { - /* release current context w/out assigning new one. */ - XMesaMakeCurrent( NULL, NULL ); - MakeCurrent_PrevContext = 0; - MakeCurrent_PrevDrawable = 0; - MakeCurrent_PrevReadable = 0; - MakeCurrent_PrevDrawBuffer = 0; - MakeCurrent_PrevReadBuffer = 0; - return True; - } - else { - /* The args must either all be non-zero or all zero. - * This is an error. - */ - return False; - } -} - - -static Bool -Fake_glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx ) -{ - return Fake_glXMakeContextCurrent( dpy, drawable, drawable, ctx ); -} - - -static GLXPixmap -Fake_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ) -{ - XMesaVisual v; - XMesaBuffer b; - - v = find_glx_visual( dpy, visinfo ); - if (!v) { - v = create_glx_visual( dpy, visinfo ); - if (!v) { - /* unusable visual */ - return 0; - } - } - - b = XMesaCreatePixmapBuffer( v, pixmap, 0 ); - if (!b) { - return 0; - } - return b->drawable; -} - - -/*** GLX_MESA_pixmap_colormap ***/ - -static GLXPixmap -Fake_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, - Pixmap pixmap, Colormap cmap ) -{ - XMesaVisual v; - XMesaBuffer b; - - v = find_glx_visual( dpy, visinfo ); - if (!v) { - v = create_glx_visual( dpy, visinfo ); - if (!v) { - /* unusable visual */ - return 0; - } - } - - b = XMesaCreatePixmapBuffer( v, pixmap, cmap ); - if (!b) { - return 0; - } - return b->drawable; -} - - -static void -Fake_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap ) -{ - XMesaBuffer b = XMesaFindBuffer(dpy, pixmap); - if (b) { - XMesaDestroyBuffer(b); - } - else if (_mesa_getenv("MESA_DEBUG")) { - _mesa_warning(NULL, "Mesa: glXDestroyGLXPixmap: invalid pixmap\n"); - } -} - - -static void -Fake_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, - unsigned long mask ) -{ - struct fake_glx_context *fakeSrc = (struct fake_glx_context *) src; - struct fake_glx_context *fakeDst = (struct fake_glx_context *) dst; - XMesaContext xm_src = fakeSrc->xmesaContext; - XMesaContext xm_dst = fakeDst->xmesaContext; - (void) dpy; - if (MakeCurrent_PrevContext == src) { - _mesa_Flush(); - } - st_copy_context_state( xm_src->st, xm_dst->st, (GLuint) mask ); -} - - -static Bool -Fake_glXQueryExtension( Display *dpy, int *errorb, int *event ) -{ - /* Mesa's GLX isn't really an X extension but we try to act like one. */ - (void) dpy; - (void) errorb; - (void) event; - return True; -} - - -extern void _kw_ungrab_all( Display *dpy ); -void _kw_ungrab_all( Display *dpy ) -{ - XUngrabPointer( dpy, CurrentTime ); - XUngrabKeyboard( dpy, CurrentTime ); -} - - -static void -Fake_glXDestroyContext( Display *dpy, GLXContext ctx ) -{ - struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; - (void) dpy; - MakeCurrent_PrevContext = 0; - MakeCurrent_PrevDrawable = 0; - MakeCurrent_PrevReadable = 0; - MakeCurrent_PrevDrawBuffer = 0; - MakeCurrent_PrevReadBuffer = 0; - XMesaDestroyContext( glxCtx->xmesaContext ); - XMesaGarbageCollect(); - _mesa_free(glxCtx); -} - - -static Bool -Fake_glXIsDirect( Display *dpy, GLXContext ctx ) -{ - (void) dpy; - (void) ctx; - return False; -} - - - -static void -Fake_glXSwapBuffers( Display *dpy, GLXDrawable drawable ) -{ - XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable ); - - if (buffer) { - XMesaSwapBuffers(buffer); - } - else if (_mesa_getenv("MESA_DEBUG")) { - _mesa_warning(NULL, "glXSwapBuffers: invalid drawable 0x%x\n", - (int) drawable); - } -} - - - -/*** GLX_MESA_copy_sub_buffer ***/ - -static void -Fake_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable, - int x, int y, int width, int height ) -{ - XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable ); - if (buffer) { - XMesaCopySubBuffer(buffer, x, y, width, height); - } - else if (_mesa_getenv("MESA_DEBUG")) { - _mesa_warning(NULL, "Mesa: glXCopySubBufferMESA: invalid drawable\n"); - } -} - - -static Bool -Fake_glXQueryVersion( Display *dpy, int *maj, int *min ) -{ - (void) dpy; - /* Return GLX version, not Mesa version */ - assert(CLIENT_MAJOR_VERSION == SERVER_MAJOR_VERSION); - *maj = CLIENT_MAJOR_VERSION; - *min = MIN2( CLIENT_MINOR_VERSION, SERVER_MINOR_VERSION ); - return True; -} - - -/* - * Query the GLX attributes of the given XVisualInfo. - */ -static int -get_config( XMesaVisual xmvis, int attrib, int *value, GLboolean fbconfig ) -{ - ASSERT(xmvis); - switch(attrib) { - case GLX_USE_GL: - if (fbconfig) - return GLX_BAD_ATTRIBUTE; - *value = (int) True; - return 0; - case GLX_BUFFER_SIZE: - *value = xmvis->visinfo->depth; - return 0; - case GLX_LEVEL: - *value = xmvis->mesa_visual.level; - return 0; - case GLX_RGBA: - if (fbconfig) - return GLX_BAD_ATTRIBUTE; - if (xmvis->mesa_visual.rgbMode) { - *value = True; - } - else { - *value = False; - } - return 0; - case GLX_DOUBLEBUFFER: - *value = (int) xmvis->mesa_visual.doubleBufferMode; - return 0; - case GLX_STEREO: - *value = (int) xmvis->mesa_visual.stereoMode; - return 0; - case GLX_AUX_BUFFERS: - *value = xmvis->mesa_visual.numAuxBuffers; - return 0; - case GLX_RED_SIZE: - *value = xmvis->mesa_visual.redBits; - return 0; - case GLX_GREEN_SIZE: - *value = xmvis->mesa_visual.greenBits; - return 0; - case GLX_BLUE_SIZE: - *value = xmvis->mesa_visual.blueBits; - return 0; - case GLX_ALPHA_SIZE: - *value = xmvis->mesa_visual.alphaBits; - return 0; - case GLX_DEPTH_SIZE: - *value = xmvis->mesa_visual.depthBits; - return 0; - case GLX_STENCIL_SIZE: - *value = xmvis->mesa_visual.stencilBits; - return 0; - case GLX_ACCUM_RED_SIZE: - *value = xmvis->mesa_visual.accumRedBits; - return 0; - case GLX_ACCUM_GREEN_SIZE: - *value = xmvis->mesa_visual.accumGreenBits; - return 0; - case GLX_ACCUM_BLUE_SIZE: - *value = xmvis->mesa_visual.accumBlueBits; - return 0; - case GLX_ACCUM_ALPHA_SIZE: - *value = xmvis->mesa_visual.accumAlphaBits; - return 0; - - /* - * GLX_EXT_visual_info extension - */ - case GLX_X_VISUAL_TYPE_EXT: - switch (xmvis->visinfo->CLASS) { - case StaticGray: *value = GLX_STATIC_GRAY_EXT; return 0; - case GrayScale: *value = GLX_GRAY_SCALE_EXT; return 0; - case StaticColor: *value = GLX_STATIC_GRAY_EXT; return 0; - case PseudoColor: *value = GLX_PSEUDO_COLOR_EXT; return 0; - case TrueColor: *value = GLX_TRUE_COLOR_EXT; return 0; - case DirectColor: *value = GLX_DIRECT_COLOR_EXT; return 0; - } - return 0; - case GLX_TRANSPARENT_TYPE_EXT: - if (xmvis->mesa_visual.level==0) { - /* normal planes */ - *value = GLX_NONE_EXT; - } - else if (xmvis->mesa_visual.level>0) { - /* overlay */ - if (xmvis->mesa_visual.rgbMode) { - *value = GLX_TRANSPARENT_RGB_EXT; - } - else { - *value = GLX_TRANSPARENT_INDEX_EXT; - } - } - else if (xmvis->mesa_visual.level<0) { - /* underlay */ - *value = GLX_NONE_EXT; - } - return 0; - case GLX_TRANSPARENT_INDEX_VALUE_EXT: - { - int pixel = transparent_pixel( xmvis ); - if (pixel>=0) { - *value = pixel; - } - /* else undefined */ - } - return 0; - case GLX_TRANSPARENT_RED_VALUE_EXT: - /* undefined */ - return 0; - case GLX_TRANSPARENT_GREEN_VALUE_EXT: - /* undefined */ - return 0; - case GLX_TRANSPARENT_BLUE_VALUE_EXT: - /* undefined */ - return 0; - case GLX_TRANSPARENT_ALPHA_VALUE_EXT: - /* undefined */ - return 0; - - /* - * GLX_EXT_visual_info extension - */ - case GLX_VISUAL_CAVEAT_EXT: - /* test for zero, just in case */ - if (xmvis->mesa_visual.visualRating > 0) - *value = xmvis->mesa_visual.visualRating; - else - *value = GLX_NONE_EXT; - return 0; - - /* - * GLX_ARB_multisample - */ - case GLX_SAMPLE_BUFFERS_ARB: - *value = 0; - return 0; - case GLX_SAMPLES_ARB: - *value = 0; - return 0; - - /* - * For FBConfigs: - */ - case GLX_SCREEN_EXT: - if (!fbconfig) - return GLX_BAD_ATTRIBUTE; - *value = xmvis->visinfo->screen; - break; - case GLX_DRAWABLE_TYPE: /*SGIX too */ - if (!fbconfig) - return GLX_BAD_ATTRIBUTE; - *value = GLX_WINDOW_BIT | GLX_PIXMAP_BIT | GLX_PBUFFER_BIT; - break; - case GLX_RENDER_TYPE_SGIX: - if (!fbconfig) - return GLX_BAD_ATTRIBUTE; - if (xmvis->mesa_visual.rgbMode) - *value = GLX_RGBA_BIT; - else - *value = GLX_COLOR_INDEX_BIT; - break; - case GLX_X_RENDERABLE_SGIX: - if (!fbconfig) - return GLX_BAD_ATTRIBUTE; - *value = True; /* XXX really? */ - break; - case GLX_FBCONFIG_ID_SGIX: - if (!fbconfig) - return GLX_BAD_ATTRIBUTE; - *value = xmvis->visinfo->visualid; - break; - case GLX_MAX_PBUFFER_WIDTH: - if (!fbconfig) - return GLX_BAD_ATTRIBUTE; - /* XXX or MAX_WIDTH? */ - *value = DisplayWidth(xmvis->display, xmvis->visinfo->screen); - break; - case GLX_MAX_PBUFFER_HEIGHT: - if (!fbconfig) - return GLX_BAD_ATTRIBUTE; - *value = DisplayHeight(xmvis->display, xmvis->visinfo->screen); - break; - case GLX_MAX_PBUFFER_PIXELS: - if (!fbconfig) - return GLX_BAD_ATTRIBUTE; - *value = DisplayWidth(xmvis->display, xmvis->visinfo->screen) * - DisplayHeight(xmvis->display, xmvis->visinfo->screen); - break; - case GLX_VISUAL_ID: - if (!fbconfig) - return GLX_BAD_ATTRIBUTE; - *value = xmvis->visinfo->visualid; - break; - -#ifdef GLX_EXT_texture_from_pixmap - case GLX_BIND_TO_TEXTURE_RGB_EXT: - *value = True; /*XXX*/ - break; - case GLX_BIND_TO_TEXTURE_RGBA_EXT: - /* XXX review */ - *value = xmvis->mesa_visual.alphaBits > 0 ? True : False; - break; - case GLX_BIND_TO_MIPMAP_TEXTURE_EXT: - *value = True; /*XXX*/ - break; - case GLX_BIND_TO_TEXTURE_TARGETS_EXT: - *value = (GLX_TEXTURE_1D_BIT_EXT | - GLX_TEXTURE_2D_BIT_EXT | - GLX_TEXTURE_RECTANGLE_BIT_EXT); /*XXX*/ - break; - case GLX_Y_INVERTED_EXT: - *value = True; /*XXX*/ - break; -#endif - - default: - return GLX_BAD_ATTRIBUTE; - } - return Success; -} - - -static int -Fake_glXGetConfig( Display *dpy, XVisualInfo *visinfo, - int attrib, int *value ) -{ - XMesaVisual xmvis; - int k; - if (!dpy || !visinfo) - return GLX_BAD_ATTRIBUTE; - - xmvis = find_glx_visual( dpy, visinfo ); - if (!xmvis) { - /* this visual wasn't obtained with glXChooseVisual */ - xmvis = create_glx_visual( dpy, visinfo ); - if (!xmvis) { - /* this visual can't be used for GL rendering */ - if (attrib==GLX_USE_GL) { - *value = (int) False; - return 0; - } - else { - return GLX_BAD_VISUAL; - } - } - } - - k = get_config(xmvis, attrib, value, GL_FALSE); - return k; -} - - -static void -Fake_glXWaitGL( void ) -{ - XMesaContext xmesa = XMesaGetCurrentContext(); - XMesaFlush( xmesa ); -} - - - -static void -Fake_glXWaitX( void ) -{ - XMesaContext xmesa = XMesaGetCurrentContext(); - XMesaFlush( xmesa ); -} - - -static const char * -get_extensions( void ) -{ -#ifdef FX - const char *fx = _mesa_getenv("MESA_GLX_FX"); - if (fx && fx[0] != 'd') { - return EXTENSIONS; - } -#endif - return EXTENSIONS + 23; /* skip "GLX_MESA_set_3dfx_mode" */ -} - - - -/* GLX 1.1 and later */ -static const char * -Fake_glXQueryExtensionsString( Display *dpy, int screen ) -{ - (void) dpy; - (void) screen; - return get_extensions(); -} - - - -/* GLX 1.1 and later */ -static const char * -Fake_glXQueryServerString( Display *dpy, int screen, int name ) -{ - static char version[1000]; - _mesa_sprintf(version, "%d.%d %s", - SERVER_MAJOR_VERSION, SERVER_MINOR_VERSION, MESA_GLX_VERSION); - - (void) dpy; - (void) screen; - - switch (name) { - case GLX_EXTENSIONS: - return get_extensions(); - case GLX_VENDOR: - return VENDOR; - case GLX_VERSION: - return version; - default: - return NULL; - } -} - - - -/* GLX 1.1 and later */ -static const char * -Fake_glXGetClientString( Display *dpy, int name ) -{ - static char version[1000]; - _mesa_sprintf(version, "%d.%d %s", CLIENT_MAJOR_VERSION, - CLIENT_MINOR_VERSION, MESA_GLX_VERSION); - - (void) dpy; - - switch (name) { - case GLX_EXTENSIONS: - return get_extensions(); - case GLX_VENDOR: - return VENDOR; - case GLX_VERSION: - return version; - default: - return NULL; - } -} - - - -/* - * GLX 1.3 and later - */ - - -static int -Fake_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config, - int attribute, int *value ) -{ - XMesaVisual v = (XMesaVisual) config; - (void) dpy; - (void) config; - - if (!dpy || !config || !value) - return -1; - - return get_config(v, attribute, value, GL_TRUE); -} - - -static GLXFBConfig * -Fake_glXGetFBConfigs( Display *dpy, int screen, int *nelements ) -{ - XVisualInfo *visuals, visTemplate; - const long visMask = VisualScreenMask; - int i; - - /* Get list of all X visuals */ - visTemplate.screen = screen; - visuals = XGetVisualInfo(dpy, visMask, &visTemplate, nelements); - if (*nelements > 0) { - XMesaVisual *results; - results = (XMesaVisual *) _mesa_malloc(*nelements * sizeof(XMesaVisual)); - if (!results) { - *nelements = 0; - return NULL; - } - for (i = 0; i < *nelements; i++) { - results[i] = create_glx_visual(dpy, visuals + i); - } - return (GLXFBConfig *) results; - } - return NULL; -} - - -static GLXFBConfig * -Fake_glXChooseFBConfig( Display *dpy, int screen, - const int *attribList, int *nitems ) -{ - XMesaVisual xmvis; - - if (!attribList || !attribList[0]) { - /* return list of all configs (per GLX_SGIX_fbconfig spec) */ - return Fake_glXGetFBConfigs(dpy, screen, nitems); - } - - xmvis = choose_visual(dpy, screen, attribList, GL_TRUE); - if (xmvis) { - GLXFBConfig *config = (GLXFBConfig *) _mesa_malloc(sizeof(XMesaVisual)); - if (!config) { - *nitems = 0; - return NULL; - } - *nitems = 1; - config[0] = (GLXFBConfig) xmvis; - return (GLXFBConfig *) config; - } - else { - *nitems = 0; - return NULL; - } -} - - -static XVisualInfo * -Fake_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config ) -{ - if (dpy && config) { - XMesaVisual xmvis = (XMesaVisual) config; -#if 0 - return xmvis->vishandle; -#else - /* create a new vishandle - the cached one may be stale */ - xmvis->vishandle = (XVisualInfo *) _mesa_malloc(sizeof(XVisualInfo)); - if (xmvis->vishandle) { - _mesa_memcpy(xmvis->vishandle, xmvis->visinfo, sizeof(XVisualInfo)); - } - return xmvis->vishandle; -#endif - } - else { - return NULL; - } -} - - -static GLXWindow -Fake_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win, - const int *attribList ) -{ - XMesaVisual xmvis = (XMesaVisual) config; - XMesaBuffer xmbuf; - if (!xmvis) - return 0; - - xmbuf = XMesaCreateWindowBuffer(xmvis, win); - if (!xmbuf) - return 0; - -#ifdef FX - /* XXX this will segfault if actually called */ - FXcreateContext(xmvis, win, NULL, xmbuf); -#endif - - (void) dpy; - (void) attribList; /* Ignored in GLX 1.3 */ - - return win; /* A hack for now */ -} - - -static void -Fake_glXDestroyWindow( Display *dpy, GLXWindow window ) -{ - XMesaBuffer b = XMesaFindBuffer(dpy, (XMesaDrawable) window); - if (b) - XMesaDestroyBuffer(b); - /* don't destroy X window */ -} - - -/* XXX untested */ -static GLXPixmap -Fake_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, - const int *attribList ) -{ - XMesaVisual v = (XMesaVisual) config; - XMesaBuffer b; - const int *attr; - int target = 0, format = 0, mipmap = 0; - int value; - - if (!dpy || !config || !pixmap) - return 0; - - for (attr = attribList; *attr; attr++) { - switch (*attr) { - case GLX_TEXTURE_FORMAT_EXT: - attr++; - switch (*attr) { - case GLX_TEXTURE_FORMAT_NONE_EXT: - case GLX_TEXTURE_FORMAT_RGB_EXT: - case GLX_TEXTURE_FORMAT_RGBA_EXT: - format = *attr; - break; - default: - /* error */ - return 0; - } - break; - case GLX_TEXTURE_TARGET_EXT: - attr++; - switch (*attr) { - case GLX_TEXTURE_1D_EXT: - case GLX_TEXTURE_2D_EXT: - case GLX_TEXTURE_RECTANGLE_EXT: - target = *attr; - break; - default: - /* error */ - return 0; - } - break; - case GLX_MIPMAP_TEXTURE_EXT: - attr++; - if (*attr) - mipmap = 1; - break; - default: - /* error */ - return 0; - } - } - - if (format == GLX_TEXTURE_FORMAT_RGB_EXT) { - if (get_config(v, GLX_BIND_TO_TEXTURE_RGB_EXT, - &value, GL_TRUE) != Success - || !value) { - return 0; /* error! */ - } - } - else if (format == GLX_TEXTURE_FORMAT_RGBA_EXT) { - if (get_config(v, GLX_BIND_TO_TEXTURE_RGBA_EXT, - &value, GL_TRUE) != Success - || !value) { - return 0; /* error! */ - } - } - if (mipmap) { - if (get_config(v, GLX_BIND_TO_MIPMAP_TEXTURE_EXT, - &value, GL_TRUE) != Success - || !value) { - return 0; /* error! */ - } - } - if (target == GLX_TEXTURE_1D_EXT) { - if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT, - &value, GL_TRUE) != Success - || (value & GLX_TEXTURE_1D_BIT_EXT) == 0) { - return 0; /* error! */ - } - } - else if (target == GLX_TEXTURE_2D_EXT) { - if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT, - &value, GL_TRUE) != Success - || (value & GLX_TEXTURE_2D_BIT_EXT) == 0) { - return 0; /* error! */ - } - } - if (target == GLX_TEXTURE_RECTANGLE_EXT) { - if (get_config(v, GLX_BIND_TO_TEXTURE_TARGETS_EXT, - &value, GL_TRUE) != Success - || (value & GLX_TEXTURE_RECTANGLE_BIT_EXT) == 0) { - return 0; /* error! */ - } - } - - if (format || target || mipmap) { - /* texture from pixmap */ - b = XMesaCreatePixmapTextureBuffer(v, pixmap, 0, format, target, mipmap); - } - else { - b = XMesaCreatePixmapBuffer( v, pixmap, 0 ); - } - if (!b) { - return 0; - } - - return pixmap; -} - - -static void -Fake_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap ) -{ - XMesaBuffer b = XMesaFindBuffer(dpy, (XMesaDrawable)pixmap); - if (b) - XMesaDestroyBuffer(b); - /* don't destroy X pixmap */ -} - - -static GLXPbuffer -Fake_glXCreatePbuffer( Display *dpy, GLXFBConfig config, - const int *attribList ) -{ - XMesaVisual xmvis = (XMesaVisual) config; - XMesaBuffer xmbuf; - const int *attrib; - int width = 0, height = 0; - GLboolean useLargest = GL_FALSE, preserveContents = GL_FALSE; - - (void) dpy; - - for (attrib = attribList; *attrib; attrib++) { - switch (*attrib) { - case GLX_PBUFFER_WIDTH: - attrib++; - width = *attrib; - break; - case GLX_PBUFFER_HEIGHT: - attrib++; - height = *attrib; - break; - case GLX_PRESERVED_CONTENTS: - attrib++; - preserveContents = *attrib; /* ignored */ - break; - case GLX_LARGEST_PBUFFER: - attrib++; - useLargest = *attrib; /* ignored */ - break; - default: - return 0; - } - } - - /* not used at this time */ - (void) useLargest; - (void) preserveContents; - - if (width == 0 || height == 0) - return 0; - - xmbuf = XMesaCreatePBuffer( xmvis, 0, width, height); - /* A GLXPbuffer handle must be an X Drawable because that's what - * glXMakeCurrent takes. - */ - if (xmbuf) - return (GLXPbuffer) xmbuf->drawable; - else - return 0; -} - - -static void -Fake_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf ) -{ - XMesaBuffer b = XMesaFindBuffer(dpy, pbuf); - if (b) { - XMesaDestroyBuffer(b); - } -} - - -static void -Fake_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, - unsigned int *value ) -{ - XMesaBuffer xmbuf = XMesaFindBuffer(dpy, draw); - if (!xmbuf) - return; - - switch (attribute) { - case GLX_WIDTH: - *value = xmesa_buffer_width(xmbuf); - break; - case GLX_HEIGHT: - *value = xmesa_buffer_width(xmbuf); - break; - case GLX_PRESERVED_CONTENTS: - *value = True; - break; - case GLX_LARGEST_PBUFFER: - *value = xmesa_buffer_width(xmbuf) * xmesa_buffer_height(xmbuf); - break; - case GLX_FBCONFIG_ID: - *value = xmbuf->xm_visual->visinfo->visualid; - return; -#ifdef GLX_EXT_texture_from_pixmap - case GLX_TEXTURE_FORMAT_EXT: - *value = xmbuf->TextureFormat; - break; - case GLX_TEXTURE_TARGET_EXT: - *value = xmbuf->TextureTarget; - break; - case GLX_MIPMAP_TEXTURE_EXT: - *value = xmbuf->TextureMipmap; - break; -#endif - - default: - return; /* raise BadValue error */ - } -} - - -static GLXContext -Fake_glXCreateNewContext( Display *dpy, GLXFBConfig config, - int renderType, GLXContext shareList, Bool direct ) -{ - struct fake_glx_context *glxCtx; - struct fake_glx_context *shareCtx = (struct fake_glx_context *) shareList; - XMesaVisual xmvis = (XMesaVisual) config; - - if (!dpy || !config || - (renderType != GLX_RGBA_TYPE && renderType != GLX_COLOR_INDEX_TYPE)) - return 0; - - glxCtx = CALLOC_STRUCT(fake_glx_context); - if (!glxCtx) - return 0; - - /* deallocate unused windows/buffers */ - XMesaGarbageCollect(); - - glxCtx->xmesaContext = XMesaCreateContext(xmvis, - shareCtx ? shareCtx->xmesaContext : NULL); - if (!glxCtx->xmesaContext) { - _mesa_free(glxCtx); - return NULL; - } - - glxCtx->glxContext.isDirect = GL_FALSE; - glxCtx->glxContext.currentDpy = dpy; - glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ - - assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); - - return (GLXContext) glxCtx; -} - - -static int -Fake_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value ) -{ - struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; - XMesaContext xmctx = glxCtx->xmesaContext; - - (void) dpy; - (void) ctx; - - switch (attribute) { - case GLX_FBCONFIG_ID: - *value = xmctx->xm_visual->visinfo->visualid; - break; - case GLX_RENDER_TYPE: - if (xmctx->xm_visual->mesa_visual.rgbMode) - *value = GLX_RGBA_BIT; - else - *value = GLX_COLOR_INDEX_BIT; - break; - case GLX_SCREEN: - *value = 0; - return Success; - default: - return GLX_BAD_ATTRIBUTE; - } - return 0; -} - - -static void -Fake_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask ) -{ - XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); - if (xmbuf) - xmbuf->selectedEvents = mask; -} - - -static void -Fake_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable, - unsigned long *mask ) -{ - XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); - if (xmbuf) - *mask = xmbuf->selectedEvents; - else - *mask = 0; -} - - - -/*** GLX_SGI_swap_control ***/ - -static int -Fake_glXSwapIntervalSGI(int interval) -{ - (void) interval; - return 0; -} - - - -/*** GLX_SGI_video_sync ***/ - -static unsigned int FrameCounter = 0; - -static int -Fake_glXGetVideoSyncSGI(unsigned int *count) -{ - /* this is a bogus implementation */ - *count = FrameCounter++; - return 0; -} - -static int -Fake_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) -{ - if (divisor <= 0 || remainder < 0) - return GLX_BAD_VALUE; - /* this is a bogus implementation */ - FrameCounter++; - while (FrameCounter % divisor != remainder) - FrameCounter++; - *count = FrameCounter; - return 0; -} - - - -/*** GLX_SGI_make_current_read ***/ - -static Bool -Fake_glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) -{ - return Fake_glXMakeContextCurrent( dpy, draw, read, ctx ); -} - -/* not used -static GLXDrawable -Fake_glXGetCurrentReadDrawableSGI(void) -{ - return 0; -} -*/ - - -/*** GLX_SGIX_video_source ***/ -#if defined(_VL_H) - -static GLXVideoSourceSGIX -Fake_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode) -{ - (void) dpy; - (void) screen; - (void) server; - (void) path; - (void) nodeClass; - (void) drainNode; - return 0; -} - -static void -Fake_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) -{ - (void) dpy; - (void) src; -} - -#endif - - -/*** GLX_EXT_import_context ***/ - -static void -Fake_glXFreeContextEXT(Display *dpy, GLXContext context) -{ - (void) dpy; - (void) context; -} - -static GLXContextID -Fake_glXGetContextIDEXT(const GLXContext context) -{ - (void) context; - return 0; -} - -static GLXContext -Fake_glXImportContextEXT(Display *dpy, GLXContextID contextID) -{ - (void) dpy; - (void) contextID; - return 0; -} - -static int -Fake_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value) -{ - (void) dpy; - (void) context; - (void) attribute; - (void) value; - return 0; -} - - - -/*** GLX_SGIX_fbconfig ***/ - -static int -Fake_glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value) -{ - return Fake_glXGetFBConfigAttrib(dpy, config, attribute, value); -} - -static GLXFBConfigSGIX * -Fake_glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements) -{ - return (GLXFBConfig *) Fake_glXChooseFBConfig(dpy, screen, attrib_list, nelements); -} - - -static GLXPixmap -Fake_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap) -{ - XMesaVisual xmvis = (XMesaVisual) config; - XMesaBuffer xmbuf = XMesaCreatePixmapBuffer(xmvis, pixmap, 0); - return xmbuf->drawable; /* need to return an X ID */ -} - - -static GLXContext -Fake_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct) -{ - XMesaVisual xmvis = (XMesaVisual) config; - struct fake_glx_context *glxCtx; - struct fake_glx_context *shareCtx = (struct fake_glx_context *) share_list; - - glxCtx = CALLOC_STRUCT(fake_glx_context); - if (!glxCtx) - return 0; - - /* deallocate unused windows/buffers */ - XMesaGarbageCollect(); - - glxCtx->xmesaContext = XMesaCreateContext(xmvis, - shareCtx ? shareCtx->xmesaContext : NULL); - if (!glxCtx->xmesaContext) { - _mesa_free(glxCtx); - return NULL; - } - - glxCtx->glxContext.isDirect = GL_FALSE; - glxCtx->glxContext.currentDpy = dpy; - glxCtx->glxContext.xid = (XID) glxCtx; /* self pointer */ - - assert((void *) glxCtx == (void *) &(glxCtx->glxContext)); - - return (GLXContext) glxCtx; -} - - -static XVisualInfo * -Fake_glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config) -{ - return Fake_glXGetVisualFromFBConfig(dpy, config); -} - - -static GLXFBConfigSGIX -Fake_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) -{ - XMesaVisual xmvis = find_glx_visual(dpy, vis); - if (!xmvis) { - /* This visual wasn't found with glXChooseVisual() */ - xmvis = create_glx_visual(dpy, vis); - } - - return (GLXFBConfigSGIX) xmvis; -} - - - -/*** GLX_SGIX_pbuffer ***/ - -static GLXPbufferSGIX -Fake_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, - unsigned int width, unsigned int height, - int *attribList) -{ - XMesaVisual xmvis = (XMesaVisual) config; - XMesaBuffer xmbuf; - const int *attrib; - GLboolean useLargest = GL_FALSE, preserveContents = GL_FALSE; - - (void) dpy; - - for (attrib = attribList; attrib && *attrib; attrib++) { - switch (*attrib) { - case GLX_PRESERVED_CONTENTS_SGIX: - attrib++; - preserveContents = *attrib; /* ignored */ - break; - case GLX_LARGEST_PBUFFER_SGIX: - attrib++; - useLargest = *attrib; /* ignored */ - break; - default: - return 0; - } - } - - /* not used at this time */ - (void) useLargest; - (void) preserveContents; - - xmbuf = XMesaCreatePBuffer( xmvis, 0, width, height); - /* A GLXPbuffer handle must be an X Drawable because that's what - * glXMakeCurrent takes. - */ - return (GLXPbuffer) xmbuf->drawable; -} - - -static void -Fake_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) -{ - XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf); - if (xmbuf) { - XMesaDestroyBuffer(xmbuf); - } -} - - -static int -Fake_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value) -{ - const XMesaBuffer xmbuf = XMesaFindBuffer(dpy, pbuf); - - if (!xmbuf) { - /* Generate GLXBadPbufferSGIX for bad pbuffer */ - return 0; - } - - switch (attribute) { - case GLX_PRESERVED_CONTENTS_SGIX: - *value = True; - break; - case GLX_LARGEST_PBUFFER_SGIX: - *value = xmesa_buffer_width(xmbuf) * xmesa_buffer_height(xmbuf); - break; - case GLX_WIDTH_SGIX: - *value = xmesa_buffer_width(xmbuf); - break; - case GLX_HEIGHT_SGIX: - *value = xmesa_buffer_height(xmbuf); - break; - case GLX_EVENT_MASK_SGIX: - *value = 0; /* XXX might be wrong */ - break; - default: - *value = 0; - } - return 0; -} - - -static void -Fake_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) -{ - XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); - if (xmbuf) { - /* Note: we'll never generate clobber events */ - xmbuf->selectedEvents = mask; - } -} - - -static void -Fake_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask) -{ - XMesaBuffer xmbuf = XMesaFindBuffer(dpy, drawable); - if (xmbuf) { - *mask = xmbuf->selectedEvents; - } - else { - *mask = 0; - } -} - - - -/*** GLX_SGI_cushion ***/ - -static void -Fake_glXCushionSGI(Display *dpy, Window win, float cushion) -{ - (void) dpy; - (void) win; - (void) cushion; -} - - - -/*** GLX_SGIX_video_resize ***/ - -static int -Fake_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window) -{ - (void) dpy; - (void) screen; - (void) channel; - (void) window; - return 0; -} - -static int -Fake_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h) -{ - (void) dpy; - (void) screen; - (void) channel; - (void) x; - (void) y; - (void) w; - (void) h; - return 0; -} - -static int -Fake_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h) -{ - (void) dpy; - (void) screen; - (void) channel; - (void) x; - (void) y; - (void) w; - (void) h; - return 0; -} - -static int -Fake_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh) -{ - (void) dpy; - (void) screen; - (void) channel; - (void) dx; - (void) dy; - (void) dw; - (void) dh; - return 0; -} - -static int -Fake_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype) -{ - (void) dpy; - (void) screen; - (void) channel; - (void) synctype; - return 0; -} - - - -/*** GLX_SGIX_dmbuffer **/ - -#if defined(_DM_BUFFER_H_) -static Bool -Fake_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer) -{ - (void) dpy; - (void) pbuffer; - (void) params; - (void) dmbuffer; - return False; -} -#endif - - -/*** GLX_SGIX_swap_group ***/ - -static void -Fake_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member) -{ - (void) dpy; - (void) drawable; - (void) member; -} - - - -/*** GLX_SGIX_swap_barrier ***/ - -static void -Fake_glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier) -{ - (void) dpy; - (void) drawable; - (void) barrier; -} - -static Bool -Fake_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) -{ - (void) dpy; - (void) screen; - (void) max; - return False; -} - - - -/*** GLX_SUN_get_transparent_index ***/ - -static Status -Fake_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent) -{ - (void) dpy; - (void) overlay; - (void) underlay; - (void) pTransparent; - return 0; -} - - - -/*** GLX_MESA_release_buffers ***/ - -/* - * Release the depth, stencil, accum buffers attached to a GLXDrawable - * (a window or pixmap) prior to destroying the GLXDrawable. - */ -static Bool -Fake_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d ) -{ - XMesaBuffer b = XMesaFindBuffer(dpy, d); - if (b) { - XMesaDestroyBuffer(b); - return True; - } - return False; -} - - - -/*** GLX_MESA_set_3dfx_mode ***/ - -static Bool -Fake_glXSet3DfxModeMESA( int mode ) -{ - return XMesaSetFXmode( mode ); -} - - - -/*** GLX_NV_vertex_array range ***/ -static void * -Fake_glXAllocateMemoryNV( GLsizei size, - GLfloat readFrequency, - GLfloat writeFrequency, - GLfloat priority ) -{ - (void) size; - (void) readFrequency; - (void) writeFrequency; - (void) priority; - return NULL; -} - - -static void -Fake_glXFreeMemoryNV( GLvoid *pointer ) -{ - (void) pointer; -} - - -/*** GLX_MESA_agp_offset ***/ - -static GLuint -Fake_glXGetAGPOffsetMESA( const GLvoid *pointer ) -{ - (void) pointer; - return ~0; -} - - -/*** GLX_EXT_texture_from_pixmap ***/ - -static void -Fake_glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, - const int *attrib_list) -{ - XMesaBuffer b = XMesaFindBuffer(dpy, drawable); - if (b) - XMesaBindTexImage(dpy, b, buffer, attrib_list); -} - -static void -Fake_glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer) -{ - XMesaBuffer b = XMesaFindBuffer(dpy, drawable); - if (b) - XMesaReleaseTexImage(dpy, b, buffer); -} - - -/* silence warning */ -extern struct _glxapi_table *_mesa_GetGLXDispatchTable(void); - - -/** - * Create a new GLX API dispatch table with its function pointers - * initialized to point to Mesa's "fake" GLX API functions. - * Note: there's a similar function (_real_GetGLXDispatchTable) that - * returns a new dispatch table with all pointers initalized to point - * to "real" GLX functions (which understand GLX wire protocol, etc). - */ -struct _glxapi_table * -_mesa_GetGLXDispatchTable(void) -{ - static struct _glxapi_table glx; - - /* be sure our dispatch table size <= libGL's table */ - { - GLuint size = sizeof(struct _glxapi_table) / sizeof(void *); - (void) size; - assert(_glxapi_get_dispatch_table_size() >= size); - } - - /* initialize the whole table to no-ops */ - _glxapi_set_no_op_table(&glx); - - /* now initialize the table with the functions I implement */ - glx.ChooseVisual = Fake_glXChooseVisual; - glx.CopyContext = Fake_glXCopyContext; - glx.CreateContext = Fake_glXCreateContext; - glx.CreateGLXPixmap = Fake_glXCreateGLXPixmap; - glx.DestroyContext = Fake_glXDestroyContext; - glx.DestroyGLXPixmap = Fake_glXDestroyGLXPixmap; - glx.GetConfig = Fake_glXGetConfig; - /*glx.GetCurrentContext = Fake_glXGetCurrentContext;*/ - /*glx.GetCurrentDrawable = Fake_glXGetCurrentDrawable;*/ - glx.IsDirect = Fake_glXIsDirect; - glx.MakeCurrent = Fake_glXMakeCurrent; - glx.QueryExtension = Fake_glXQueryExtension; - glx.QueryVersion = Fake_glXQueryVersion; - glx.SwapBuffers = Fake_glXSwapBuffers; - glx.UseXFont = Fake_glXUseXFont; - glx.WaitGL = Fake_glXWaitGL; - glx.WaitX = Fake_glXWaitX; - - /*** GLX_VERSION_1_1 ***/ - glx.GetClientString = Fake_glXGetClientString; - glx.QueryExtensionsString = Fake_glXQueryExtensionsString; - glx.QueryServerString = Fake_glXQueryServerString; - - /*** GLX_VERSION_1_2 ***/ - /*glx.GetCurrentDisplay = Fake_glXGetCurrentDisplay;*/ - - /*** GLX_VERSION_1_3 ***/ - glx.ChooseFBConfig = Fake_glXChooseFBConfig; - glx.CreateNewContext = Fake_glXCreateNewContext; - glx.CreatePbuffer = Fake_glXCreatePbuffer; - glx.CreatePixmap = Fake_glXCreatePixmap; - glx.CreateWindow = Fake_glXCreateWindow; - glx.DestroyPbuffer = Fake_glXDestroyPbuffer; - glx.DestroyPixmap = Fake_glXDestroyPixmap; - glx.DestroyWindow = Fake_glXDestroyWindow; - /*glx.GetCurrentReadDrawable = Fake_glXGetCurrentReadDrawable;*/ - glx.GetFBConfigAttrib = Fake_glXGetFBConfigAttrib; - glx.GetFBConfigs = Fake_glXGetFBConfigs; - glx.GetSelectedEvent = Fake_glXGetSelectedEvent; - glx.GetVisualFromFBConfig = Fake_glXGetVisualFromFBConfig; - glx.MakeContextCurrent = Fake_glXMakeContextCurrent; - glx.QueryContext = Fake_glXQueryContext; - glx.QueryDrawable = Fake_glXQueryDrawable; - glx.SelectEvent = Fake_glXSelectEvent; - - /*** GLX_SGI_swap_control ***/ - glx.SwapIntervalSGI = Fake_glXSwapIntervalSGI; - - /*** GLX_SGI_video_sync ***/ - glx.GetVideoSyncSGI = Fake_glXGetVideoSyncSGI; - glx.WaitVideoSyncSGI = Fake_glXWaitVideoSyncSGI; - - /*** GLX_SGI_make_current_read ***/ - glx.MakeCurrentReadSGI = Fake_glXMakeCurrentReadSGI; - /*glx.GetCurrentReadDrawableSGI = Fake_glXGetCurrentReadDrawableSGI;*/ - -/*** GLX_SGIX_video_source ***/ -#if defined(_VL_H) - glx.CreateGLXVideoSourceSGIX = Fake_glXCreateGLXVideoSourceSGIX; - glx.DestroyGLXVideoSourceSGIX = Fake_glXDestroyGLXVideoSourceSGIX; -#endif - - /*** GLX_EXT_import_context ***/ - glx.FreeContextEXT = Fake_glXFreeContextEXT; - glx.GetContextIDEXT = Fake_glXGetContextIDEXT; - /*glx.GetCurrentDisplayEXT = Fake_glXGetCurrentDisplayEXT;*/ - glx.ImportContextEXT = Fake_glXImportContextEXT; - glx.QueryContextInfoEXT = Fake_glXQueryContextInfoEXT; - - /*** GLX_SGIX_fbconfig ***/ - glx.GetFBConfigAttribSGIX = Fake_glXGetFBConfigAttribSGIX; - glx.ChooseFBConfigSGIX = Fake_glXChooseFBConfigSGIX; - glx.CreateGLXPixmapWithConfigSGIX = Fake_glXCreateGLXPixmapWithConfigSGIX; - glx.CreateContextWithConfigSGIX = Fake_glXCreateContextWithConfigSGIX; - glx.GetVisualFromFBConfigSGIX = Fake_glXGetVisualFromFBConfigSGIX; - glx.GetFBConfigFromVisualSGIX = Fake_glXGetFBConfigFromVisualSGIX; - - /*** GLX_SGIX_pbuffer ***/ - glx.CreateGLXPbufferSGIX = Fake_glXCreateGLXPbufferSGIX; - glx.DestroyGLXPbufferSGIX = Fake_glXDestroyGLXPbufferSGIX; - glx.QueryGLXPbufferSGIX = Fake_glXQueryGLXPbufferSGIX; - glx.SelectEventSGIX = Fake_glXSelectEventSGIX; - glx.GetSelectedEventSGIX = Fake_glXGetSelectedEventSGIX; - - /*** GLX_SGI_cushion ***/ - glx.CushionSGI = Fake_glXCushionSGI; - - /*** GLX_SGIX_video_resize ***/ - glx.BindChannelToWindowSGIX = Fake_glXBindChannelToWindowSGIX; - glx.ChannelRectSGIX = Fake_glXChannelRectSGIX; - glx.QueryChannelRectSGIX = Fake_glXQueryChannelRectSGIX; - glx.QueryChannelDeltasSGIX = Fake_glXQueryChannelDeltasSGIX; - glx.ChannelRectSyncSGIX = Fake_glXChannelRectSyncSGIX; - - /*** GLX_SGIX_dmbuffer **/ -#if defined(_DM_BUFFER_H_) - glx.AssociateDMPbufferSGIX = NULL; -#endif - - /*** GLX_SGIX_swap_group ***/ - glx.JoinSwapGroupSGIX = Fake_glXJoinSwapGroupSGIX; - - /*** GLX_SGIX_swap_barrier ***/ - glx.BindSwapBarrierSGIX = Fake_glXBindSwapBarrierSGIX; - glx.QueryMaxSwapBarriersSGIX = Fake_glXQueryMaxSwapBarriersSGIX; - - /*** GLX_SUN_get_transparent_index ***/ - glx.GetTransparentIndexSUN = Fake_glXGetTransparentIndexSUN; - - /*** GLX_MESA_copy_sub_buffer ***/ - glx.CopySubBufferMESA = Fake_glXCopySubBufferMESA; - - /*** GLX_MESA_release_buffers ***/ - glx.ReleaseBuffersMESA = Fake_glXReleaseBuffersMESA; - - /*** GLX_MESA_pixmap_colormap ***/ - glx.CreateGLXPixmapMESA = Fake_glXCreateGLXPixmapMESA; - - /*** GLX_MESA_set_3dfx_mode ***/ - glx.Set3DfxModeMESA = Fake_glXSet3DfxModeMESA; - - /*** GLX_NV_vertex_array_range ***/ - glx.AllocateMemoryNV = Fake_glXAllocateMemoryNV; - glx.FreeMemoryNV = Fake_glXFreeMemoryNV; - - /*** GLX_MESA_agp_offset ***/ - glx.GetAGPOffsetMESA = Fake_glXGetAGPOffsetMESA; - - /*** GLX_EXT_texture_from_pixmap ***/ - glx.BindTexImageEXT = Fake_glXBindTexImageEXT; - glx.ReleaseTexImageEXT = Fake_glXReleaseTexImageEXT; - - return &glx; -} diff --git a/src/mesa/pipe/xlib/glxapi.c b/src/mesa/pipe/xlib/glxapi.c deleted file mode 100644 index c2ccce6f52..0000000000 --- a/src/mesa/pipe/xlib/glxapi.c +++ /dev/null @@ -1,1408 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 7.1 - * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/* - * This is the GLX API dispatcher. Calls to the glX* functions are - * either routed to the real GLX encoders or to Mesa's pseudo-GLX functions. - * See the glxapi.h file for more details. - */ - - -#include -#include -#include -#include -#include "main/glheader.h" -#include "glapi/glapi.h" -#include "glxapi.h" - - -extern struct _glxapi_table *_real_GetGLXDispatchTable(void); -extern struct _glxapi_table *_mesa_GetGLXDispatchTable(void); - - -struct display_dispatch { - Display *Dpy; - struct _glxapi_table *Table; - struct display_dispatch *Next; -}; - -static struct display_dispatch *DispatchList = NULL; - - -/* Display -> Dispatch caching */ -static Display *prevDisplay = NULL; -static struct _glxapi_table *prevTable = NULL; - - -static struct _glxapi_table * -get_dispatch(Display *dpy) -{ - if (!dpy) - return NULL; - - /* search list of display/dispatch pairs for this display */ - { - const struct display_dispatch *d = DispatchList; - while (d) { - if (d->Dpy == dpy) { - prevDisplay = dpy; - prevTable = d->Table; - return d->Table; /* done! */ - } - d = d->Next; - } - } - - /* A new display, determine if we should use real GLX - * or Mesa's pseudo-GLX. - */ - { - struct _glxapi_table *t = _mesa_GetGLXDispatchTable(); - - if (t) { - struct display_dispatch *d; - d = (struct display_dispatch *) malloc(sizeof(struct display_dispatch)); - if (d) { - d->Dpy = dpy; - d->Table = t; - /* insert at head of list */ - d->Next = DispatchList; - DispatchList = d; - /* update cache */ - prevDisplay = dpy; - prevTable = t; - return t; - } - } - } - - /* If we get here that means we can't use real GLX on this display - * and the Mesa pseudo-GLX software renderer wasn't compiled in. - * Or, we ran out of memory! - */ - return NULL; -} - - -/* Don't use the GET_DISPATCH defined in glthread.h */ -#undef GET_DISPATCH - -#define GET_DISPATCH(DPY, TABLE) \ - if (DPY == prevDisplay) { \ - TABLE = prevTable; \ - } \ - else if (!DPY) { \ - TABLE = NULL; \ - } \ - else { \ - TABLE = get_dispatch(DPY); \ - } - - - - -/** - * GLX API current context. - */ -#if defined(GLX_USE_TLS) -PUBLIC __thread void * CurrentContext - __attribute__((tls_model("initial-exec"))); -#elif defined(THREADS) -static _glthread_TSD ContextTSD; /**< Per-thread context pointer */ -#else -static GLXContext CurrentContext = 0; -#endif - - -static void -SetCurrentContext(GLXContext c) -{ -#if defined(GLX_USE_TLS) - CurrentContext = c; -#elif defined(THREADS) - _glthread_SetTSD(&ContextTSD, c); -#else - CurrentContext = c; -#endif -} - - -/* - * GLX API entrypoints - */ - -/*** GLX_VERSION_1_0 ***/ - -XVisualInfo PUBLIC * -glXChooseVisual(Display *dpy, int screen, int *list) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->ChooseVisual)(dpy, screen, list); -} - - -void PUBLIC -glXCopyContext(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->CopyContext)(dpy, src, dst, mask); -} - - -GLXContext PUBLIC -glXCreateContext(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateContext)(dpy, visinfo, shareList, direct); -} - - -GLXPixmap PUBLIC -glXCreateGLXPixmap(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXPixmap)(dpy, visinfo, pixmap); -} - - -void PUBLIC -glXDestroyContext(Display *dpy, GLXContext ctx) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - if (glXGetCurrentContext() == ctx) - SetCurrentContext(NULL); - (t->DestroyContext)(dpy, ctx); -} - - -void PUBLIC -glXDestroyGLXPixmap(Display *dpy, GLXPixmap pixmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyGLXPixmap)(dpy, pixmap); -} - - -int PUBLIC -glXGetConfig(Display *dpy, XVisualInfo *visinfo, int attrib, int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return GLX_NO_EXTENSION; - return (t->GetConfig)(dpy, visinfo, attrib, value); -} - - -GLXContext PUBLIC -glXGetCurrentContext(void) -{ -#if defined(GLX_USE_TLS) - return CurrentContext; -#elif defined(THREADS) - return (GLXContext) _glthread_GetTSD(&ContextTSD); -#else - return CurrentContext; -#endif -} - - -GLXDrawable PUBLIC -glXGetCurrentDrawable(void) -{ - __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); - return gc ? gc->currentDrawable : 0; -} - - -Bool PUBLIC -glXIsDirect(Display *dpy, GLXContext ctx) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->IsDirect)(dpy, ctx); -} - - -Bool PUBLIC -glXMakeCurrent(Display *dpy, GLXDrawable drawable, GLXContext ctx) -{ - Bool b; - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) { - return False; - } - b = (*t->MakeCurrent)(dpy, drawable, ctx); - if (b) { - SetCurrentContext(ctx); - } - return b; -} - - -Bool PUBLIC -glXQueryExtension(Display *dpy, int *errorb, int *event) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->QueryExtension)(dpy, errorb, event); -} - - -Bool PUBLIC -glXQueryVersion(Display *dpy, int *maj, int *min) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->QueryVersion)(dpy, maj, min); -} - - -void PUBLIC -glXSwapBuffers(Display *dpy, GLXDrawable drawable) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->SwapBuffers)(dpy, drawable); -} - - -void PUBLIC -glXUseXFont(Font font, int first, int count, int listBase) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->UseXFont)(font, first, count, listBase); -} - - -void PUBLIC -glXWaitGL(void) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->WaitGL)(); -} - - -void PUBLIC -glXWaitX(void) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->WaitX)(); -} - - - -/*** GLX_VERSION_1_1 ***/ - -const char PUBLIC * -glXGetClientString(Display *dpy, int name) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->GetClientString)(dpy, name); -} - - -const char PUBLIC * -glXQueryExtensionsString(Display *dpy, int screen) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->QueryExtensionsString)(dpy, screen); -} - - -const char PUBLIC * -glXQueryServerString(Display *dpy, int screen, int name) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->QueryServerString)(dpy, screen, name); -} - - -/*** GLX_VERSION_1_2 ***/ - -Display PUBLIC * -glXGetCurrentDisplay(void) -{ - /* Same code as in libGL's glxext.c */ - __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); - if (NULL == gc) return NULL; - return gc->currentDpy; -} - - - -/*** GLX_VERSION_1_3 ***/ - -GLXFBConfig PUBLIC * -glXChooseFBConfig(Display *dpy, int screen, const int *attribList, int *nitems) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ChooseFBConfig)(dpy, screen, attribList, nitems); -} - - -GLXContext PUBLIC -glXCreateNewContext(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateNewContext)(dpy, config, renderType, shareList, direct); -} - - -GLXPbuffer PUBLIC -glXCreatePbuffer(Display *dpy, GLXFBConfig config, const int *attribList) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreatePbuffer)(dpy, config, attribList); -} - - -GLXPixmap PUBLIC -glXCreatePixmap(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreatePixmap)(dpy, config, pixmap, attribList); -} - - -GLXWindow PUBLIC -glXCreateWindow(Display *dpy, GLXFBConfig config, Window win, const int *attribList) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateWindow)(dpy, config, win, attribList); -} - - -void PUBLIC -glXDestroyPbuffer(Display *dpy, GLXPbuffer pbuf) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyPbuffer)(dpy, pbuf); -} - - -void PUBLIC -glXDestroyPixmap(Display *dpy, GLXPixmap pixmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyPixmap)(dpy, pixmap); -} - - -void PUBLIC -glXDestroyWindow(Display *dpy, GLXWindow window) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyWindow)(dpy, window); -} - - -GLXDrawable PUBLIC -glXGetCurrentReadDrawable(void) -{ - __GLXcontext *gc = (__GLXcontext *) glXGetCurrentContext(); - return gc ? gc->currentReadable : 0; -} - - -int PUBLIC -glXGetFBConfigAttrib(Display *dpy, GLXFBConfig config, int attribute, int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return GLX_NO_EXTENSION; - return (t->GetFBConfigAttrib)(dpy, config, attribute, value); -} - - -GLXFBConfig PUBLIC * -glXGetFBConfigs(Display *dpy, int screen, int *nelements) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->GetFBConfigs)(dpy, screen, nelements); -} - -void PUBLIC -glXGetSelectedEvent(Display *dpy, GLXDrawable drawable, unsigned long *mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->GetSelectedEvent)(dpy, drawable, mask); -} - - -XVisualInfo PUBLIC * -glXGetVisualFromFBConfig(Display *dpy, GLXFBConfig config) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->GetVisualFromFBConfig)(dpy, config); -} - - -Bool PUBLIC -glXMakeContextCurrent(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) -{ - Bool b; - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - b = (t->MakeContextCurrent)(dpy, draw, read, ctx); - if (b) { - SetCurrentContext(ctx); - } - return b; -} - - -int PUBLIC -glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - assert(t); - if (!t) - return 0; /* XXX correct? */ - return (t->QueryContext)(dpy, ctx, attribute, value); -} - - -void PUBLIC -glXQueryDrawable(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->QueryDrawable)(dpy, draw, attribute, value); -} - - -void PUBLIC -glXSelectEvent(Display *dpy, GLXDrawable drawable, unsigned long mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->SelectEvent)(dpy, drawable, mask); -} - - - -/*** GLX_SGI_swap_control ***/ - -int PUBLIC -glXSwapIntervalSGI(int interval) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->SwapIntervalSGI)(interval); -} - - - -/*** GLX_SGI_video_sync ***/ - -int PUBLIC -glXGetVideoSyncSGI(unsigned int *count) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t || !glXGetCurrentContext()) - return GLX_BAD_CONTEXT; - return (t->GetVideoSyncSGI)(count); -} - -int PUBLIC -glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t || !glXGetCurrentContext()) - return GLX_BAD_CONTEXT; - return (t->WaitVideoSyncSGI)(divisor, remainder, count); -} - - - -/*** GLX_SGI_make_current_read ***/ - -Bool PUBLIC -glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->MakeCurrentReadSGI)(dpy, draw, read, ctx); -} - -GLXDrawable PUBLIC -glXGetCurrentReadDrawableSGI(void) -{ - return glXGetCurrentReadDrawable(); -} - - -#if defined(_VL_H) - -GLXVideoSourceSGIX PUBLIC -glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXVideoSourceSGIX)(dpy, screen, server, path, nodeClass, drainNode); -} - -void PUBLIC -glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->DestroyGLXVideoSourceSGIX)(dpy, src); -} - -#endif - - -/*** GLX_EXT_import_context ***/ - -void PUBLIC -glXFreeContextEXT(Display *dpy, GLXContext context) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->FreeContextEXT)(dpy, context); -} - -GLXContextID PUBLIC -glXGetContextIDEXT(const GLXContext context) -{ - return ((__GLXcontext *) context)->xid; -} - -Display PUBLIC * -glXGetCurrentDisplayEXT(void) -{ - return glXGetCurrentDisplay(); -} - -GLXContext PUBLIC -glXImportContextEXT(Display *dpy, GLXContextID contextID) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ImportContextEXT)(dpy, contextID); -} - -int PUBLIC -glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute,int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; /* XXX ok? */ - return (t->QueryContextInfoEXT)(dpy, context, attribute, value); -} - - - -/*** GLX_SGIX_fbconfig ***/ - -int PUBLIC -glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->GetFBConfigAttribSGIX)(dpy, config, attribute, value); -} - -GLXFBConfigSGIX PUBLIC * -glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ChooseFBConfigSGIX)(dpy, screen, attrib_list, nelements); -} - -GLXPixmap PUBLIC -glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXPixmapWithConfigSGIX)(dpy, config, pixmap); -} - -GLXContext PUBLIC -glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateContextWithConfigSGIX)(dpy, config, render_type, share_list, direct); -} - -XVisualInfo PUBLIC * -glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->GetVisualFromFBConfigSGIX)(dpy, config); -} - -GLXFBConfigSGIX PUBLIC -glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->GetFBConfigFromVisualSGIX)(dpy, vis); -} - - - -/*** GLX_SGIX_pbuffer ***/ - -GLXPbufferSGIX PUBLIC -glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attrib_list) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXPbufferSGIX)(dpy, config, width, height, attrib_list); -} - -void PUBLIC -glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->DestroyGLXPbufferSGIX)(dpy, pbuf); -} - -int PUBLIC -glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->QueryGLXPbufferSGIX)(dpy, pbuf, attribute, value); -} - -void PUBLIC -glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->SelectEventSGIX)(dpy, drawable, mask); -} - -void PUBLIC -glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->GetSelectedEventSGIX)(dpy, drawable, mask); -} - - - -/*** GLX_SGI_cushion ***/ - -void PUBLIC -glXCushionSGI(Display *dpy, Window win, float cushion) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->CushionSGI)(dpy, win, cushion); -} - - - -/*** GLX_SGIX_video_resize ***/ - -int PUBLIC -glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->BindChannelToWindowSGIX)(dpy, screen, channel, window); -} - -int PUBLIC -glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ChannelRectSGIX)(dpy, screen, channel, x, y, w, h); -} - -int PUBLIC -glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->QueryChannelRectSGIX)(dpy, screen, channel, x, y, w, h); -} - -int PUBLIC -glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->QueryChannelDeltasSGIX)(dpy, screen, channel, dx, dy, dw, dh); -} - -int PUBLIC -glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->ChannelRectSyncSGIX)(dpy, screen, channel, synctype); -} - - - -#if defined(_DM_BUFFER_H_) - -Bool PUBLIC -glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->AssociateDMPbufferSGIX)(dpy, pbuffer, params, dmbuffer); -} - -#endif - - -/*** GLX_SGIX_swap_group ***/ - -void PUBLIC -glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (*t->JoinSwapGroupSGIX)(dpy, drawable, member); -} - - -/*** GLX_SGIX_swap_barrier ***/ - -void PUBLIC -glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (*t->BindSwapBarrierSGIX)(dpy, drawable, barrier); -} - -Bool PUBLIC -glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (*t->QueryMaxSwapBarriersSGIX)(dpy, screen, max); -} - - - -/*** GLX_SUN_get_transparent_index ***/ - -Status PUBLIC -glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (*t->GetTransparentIndexSUN)(dpy, overlay, underlay, pTransparent); -} - - - -/*** GLX_MESA_copy_sub_buffer ***/ - -void PUBLIC -glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->CopySubBufferMESA)(dpy, drawable, x, y, width, height); -} - - - -/*** GLX_MESA_release_buffers ***/ - -Bool PUBLIC -glXReleaseBuffersMESA(Display *dpy, Window w) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->ReleaseBuffersMESA)(dpy, w); -} - - - -/*** GLX_MESA_pixmap_colormap ***/ - -GLXPixmap PUBLIC -glXCreateGLXPixmapMESA(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (!t) - return 0; - return (t->CreateGLXPixmapMESA)(dpy, visinfo, pixmap, cmap); -} - - - -/*** GLX_MESA_set_3dfx_mode ***/ - -Bool PUBLIC -glXSet3DfxModeMESA(int mode) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return False; - return (t->Set3DfxModeMESA)(mode); -} - - - -/*** GLX_NV_vertex_array_range ***/ - -void PUBLIC * -glXAllocateMemoryNV( GLsizei size, - GLfloat readFrequency, - GLfloat writeFrequency, - GLfloat priority ) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return NULL; - return (t->AllocateMemoryNV)(size, readFrequency, writeFrequency, priority); -} - - -void PUBLIC -glXFreeMemoryNV( GLvoid *pointer ) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return; - (t->FreeMemoryNV)(pointer); -} - - - - -/*** GLX_MESA_agp_offset */ - -GLuint PUBLIC -glXGetAGPOffsetMESA( const GLvoid *pointer ) -{ - struct _glxapi_table *t; - Display *dpy = glXGetCurrentDisplay(); - GET_DISPATCH(dpy, t); - if (!t) - return ~0; - return (t->GetAGPOffsetMESA)(pointer); -} - - -/*** GLX_MESA_allocate_memory */ - -void * -glXAllocateMemoryMESA(Display *dpy, int scrn, size_t size, - float readfreq, float writefreq, float priority) -{ - /* dummy */ - return NULL; -} - -void -glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer) -{ - /* dummy */ -} - - -GLuint -glXGetMemoryOffsetMESA(Display *dpy, int scrn, const void *pointer) -{ - /* dummy */ - return 0; -} - - -/*** GLX_EXT_texture_from_pixmap */ - -void -glXBindTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer, - const int *attrib_list) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (t) - t->BindTexImageEXT(dpy, drawable, buffer, attrib_list); -} - -void -glXReleaseTexImageEXT(Display *dpy, GLXDrawable drawable, int buffer) -{ - struct _glxapi_table *t; - GET_DISPATCH(dpy, t); - if (t) - t->ReleaseTexImageEXT(dpy, drawable, buffer); -} - - -/**********************************************************************/ -/* GLX API management functions */ -/**********************************************************************/ - - -const char * -_glxapi_get_version(void) -{ - return "1.3"; -} - - -/* - * Return array of extension strings. - */ -const char ** -_glxapi_get_extensions(void) -{ - static const char *extensions[] = { -#ifdef GLX_EXT_import_context - "GLX_EXT_import_context", -#endif -#ifdef GLX_SGI_video_sync - "GLX_SGI_video_sync", -#endif -#ifdef GLX_MESA_copy_sub_buffer - "GLX_MESA_copy_sub_buffer", -#endif -#ifdef GLX_MESA_release_buffers - "GLX_MESA_release_buffers", -#endif -#ifdef GLX_MESA_pixmap_colormap - "GLX_MESA_pixmap_colormap", -#endif -#ifdef GLX_MESA_set_3dfx_mode - "GLX_MESA_set_3dfx_mode", -#endif -#ifdef GLX_SGIX_fbconfig - "GLX_SGIX_fbconfig", -#endif -#ifdef GLX_SGIX_pbuffer - "GLX_SGIX_pbuffer", -#endif -#ifdef GLX_EXT_texture_from_pixmap - "GLX_EXT_texture_from_pixmap", -#endif - NULL - }; - return extensions; -} - - -/* - * Return size of the GLX dispatch table, in entries, not bytes. - */ -GLuint -_glxapi_get_dispatch_table_size(void) -{ - return sizeof(struct _glxapi_table) / sizeof(void *); -} - - -static int -generic_no_op_func(void) -{ - return 0; -} - - -/* - * Initialize all functions in given dispatch table to be no-ops - */ -void -_glxapi_set_no_op_table(struct _glxapi_table *t) -{ - typedef int (*nop_func)(void); - nop_func *dispatch = (nop_func *) t; - GLuint n = _glxapi_get_dispatch_table_size(); - GLuint i; - for (i = 0; i < n; i++) { - dispatch[i] = generic_no_op_func; - } -} - - -struct name_address_pair { - const char *Name; - __GLXextFuncPtr Address; -}; - -static struct name_address_pair GLX_functions[] = { - /*** GLX_VERSION_1_0 ***/ - { "glXChooseVisual", (__GLXextFuncPtr) glXChooseVisual }, - { "glXCopyContext", (__GLXextFuncPtr) glXCopyContext }, - { "glXCreateContext", (__GLXextFuncPtr) glXCreateContext }, - { "glXCreateGLXPixmap", (__GLXextFuncPtr) glXCreateGLXPixmap }, - { "glXDestroyContext", (__GLXextFuncPtr) glXDestroyContext }, - { "glXDestroyGLXPixmap", (__GLXextFuncPtr) glXDestroyGLXPixmap }, - { "glXGetConfig", (__GLXextFuncPtr) glXGetConfig }, - { "glXGetCurrentContext", (__GLXextFuncPtr) glXGetCurrentContext }, - { "glXGetCurrentDrawable", (__GLXextFuncPtr) glXGetCurrentDrawable }, - { "glXIsDirect", (__GLXextFuncPtr) glXIsDirect }, - { "glXMakeCurrent", (__GLXextFuncPtr) glXMakeCurrent }, - { "glXQueryExtension", (__GLXextFuncPtr) glXQueryExtension }, - { "glXQueryVersion", (__GLXextFuncPtr) glXQueryVersion }, - { "glXSwapBuffers", (__GLXextFuncPtr) glXSwapBuffers }, - { "glXUseXFont", (__GLXextFuncPtr) glXUseXFont }, - { "glXWaitGL", (__GLXextFuncPtr) glXWaitGL }, - { "glXWaitX", (__GLXextFuncPtr) glXWaitX }, - - /*** GLX_VERSION_1_1 ***/ - { "glXGetClientString", (__GLXextFuncPtr) glXGetClientString }, - { "glXQueryExtensionsString", (__GLXextFuncPtr) glXQueryExtensionsString }, - { "glXQueryServerString", (__GLXextFuncPtr) glXQueryServerString }, - - /*** GLX_VERSION_1_2 ***/ - { "glXGetCurrentDisplay", (__GLXextFuncPtr) glXGetCurrentDisplay }, - - /*** GLX_VERSION_1_3 ***/ - { "glXChooseFBConfig", (__GLXextFuncPtr) glXChooseFBConfig }, - { "glXCreateNewContext", (__GLXextFuncPtr) glXCreateNewContext }, - { "glXCreatePbuffer", (__GLXextFuncPtr) glXCreatePbuffer }, - { "glXCreatePixmap", (__GLXextFuncPtr) glXCreatePixmap }, - { "glXCreateWindow", (__GLXextFuncPtr) glXCreateWindow }, - { "glXDestroyPbuffer", (__GLXextFuncPtr) glXDestroyPbuffer }, - { "glXDestroyPixmap", (__GLXextFuncPtr) glXDestroyPixmap }, - { "glXDestroyWindow", (__GLXextFuncPtr) glXDestroyWindow }, - { "glXGetCurrentReadDrawable", (__GLXextFuncPtr) glXGetCurrentReadDrawable }, - { "glXGetFBConfigAttrib", (__GLXextFuncPtr) glXGetFBConfigAttrib }, - { "glXGetFBConfigs", (__GLXextFuncPtr) glXGetFBConfigs }, - { "glXGetSelectedEvent", (__GLXextFuncPtr) glXGetSelectedEvent }, - { "glXGetVisualFromFBConfig", (__GLXextFuncPtr) glXGetVisualFromFBConfig }, - { "glXMakeContextCurrent", (__GLXextFuncPtr) glXMakeContextCurrent }, - { "glXQueryContext", (__GLXextFuncPtr) glXQueryContext }, - { "glXQueryDrawable", (__GLXextFuncPtr) glXQueryDrawable }, - { "glXSelectEvent", (__GLXextFuncPtr) glXSelectEvent }, - - /*** GLX_VERSION_1_4 ***/ - { "glXGetProcAddress", (__GLXextFuncPtr) glXGetProcAddress }, - - /*** GLX_SGI_swap_control ***/ - { "glXSwapIntervalSGI", (__GLXextFuncPtr) glXSwapIntervalSGI }, - - /*** GLX_SGI_video_sync ***/ - { "glXGetVideoSyncSGI", (__GLXextFuncPtr) glXGetVideoSyncSGI }, - { "glXWaitVideoSyncSGI", (__GLXextFuncPtr) glXWaitVideoSyncSGI }, - - /*** GLX_SGI_make_current_read ***/ - { "glXMakeCurrentReadSGI", (__GLXextFuncPtr) glXMakeCurrentReadSGI }, - { "glXGetCurrentReadDrawableSGI", (__GLXextFuncPtr) glXGetCurrentReadDrawableSGI }, - - /*** GLX_SGIX_video_source ***/ -#if defined(_VL_H) - { "glXCreateGLXVideoSourceSGIX", (__GLXextFuncPtr) glXCreateGLXVideoSourceSGIX }, - { "glXDestroyGLXVideoSourceSGIX", (__GLXextFuncPtr) glXDestroyGLXVideoSourceSGIX }, -#endif - - /*** GLX_EXT_import_context ***/ - { "glXFreeContextEXT", (__GLXextFuncPtr) glXFreeContextEXT }, - { "glXGetContextIDEXT", (__GLXextFuncPtr) glXGetContextIDEXT }, - { "glXGetCurrentDisplayEXT", (__GLXextFuncPtr) glXGetCurrentDisplayEXT }, - { "glXImportContextEXT", (__GLXextFuncPtr) glXImportContextEXT }, - { "glXQueryContextInfoEXT", (__GLXextFuncPtr) glXQueryContextInfoEXT }, - - /*** GLX_SGIX_fbconfig ***/ - { "glXGetFBConfigAttribSGIX", (__GLXextFuncPtr) glXGetFBConfigAttribSGIX }, - { "glXChooseFBConfigSGIX", (__GLXextFuncPtr) glXChooseFBConfigSGIX }, - { "glXCreateGLXPixmapWithConfigSGIX", (__GLXextFuncPtr) glXCreateGLXPixmapWithConfigSGIX }, - { "glXCreateContextWithConfigSGIX", (__GLXextFuncPtr) glXCreateContextWithConfigSGIX }, - { "glXGetVisualFromFBConfigSGIX", (__GLXextFuncPtr) glXGetVisualFromFBConfigSGIX }, - { "glXGetFBConfigFromVisualSGIX", (__GLXextFuncPtr) glXGetFBConfigFromVisualSGIX }, - - /*** GLX_SGIX_pbuffer ***/ - { "glXCreateGLXPbufferSGIX", (__GLXextFuncPtr) glXCreateGLXPbufferSGIX }, - { "glXDestroyGLXPbufferSGIX", (__GLXextFuncPtr) glXDestroyGLXPbufferSGIX }, - { "glXQueryGLXPbufferSGIX", (__GLXextFuncPtr) glXQueryGLXPbufferSGIX }, - { "glXSelectEventSGIX", (__GLXextFuncPtr) glXSelectEventSGIX }, - { "glXGetSelectedEventSGIX", (__GLXextFuncPtr) glXGetSelectedEventSGIX }, - - /*** GLX_SGI_cushion ***/ - { "glXCushionSGI", (__GLXextFuncPtr) glXCushionSGI }, - - /*** GLX_SGIX_video_resize ***/ - { "glXBindChannelToWindowSGIX", (__GLXextFuncPtr) glXBindChannelToWindowSGIX }, - { "glXChannelRectSGIX", (__GLXextFuncPtr) glXChannelRectSGIX }, - { "glXQueryChannelRectSGIX", (__GLXextFuncPtr) glXQueryChannelRectSGIX }, - { "glXQueryChannelDeltasSGIX", (__GLXextFuncPtr) glXQueryChannelDeltasSGIX }, - { "glXChannelRectSyncSGIX", (__GLXextFuncPtr) glXChannelRectSyncSGIX }, - - /*** GLX_SGIX_dmbuffer **/ -#if defined(_DM_BUFFER_H_) - { "glXAssociateDMPbufferSGIX", (__GLXextFuncPtr) glXAssociateDMPbufferSGIX }, -#endif - - /*** GLX_SGIX_swap_group ***/ - { "glXJoinSwapGroupSGIX", (__GLXextFuncPtr) glXJoinSwapGroupSGIX }, - - /*** GLX_SGIX_swap_barrier ***/ - { "glXBindSwapBarrierSGIX", (__GLXextFuncPtr) glXBindSwapBarrierSGIX }, - { "glXQueryMaxSwapBarriersSGIX", (__GLXextFuncPtr) glXQueryMaxSwapBarriersSGIX }, - - /*** GLX_SUN_get_transparent_index ***/ - { "glXGetTransparentIndexSUN", (__GLXextFuncPtr) glXGetTransparentIndexSUN }, - - /*** GLX_MESA_copy_sub_buffer ***/ - { "glXCopySubBufferMESA", (__GLXextFuncPtr) glXCopySubBufferMESA }, - - /*** GLX_MESA_pixmap_colormap ***/ - { "glXCreateGLXPixmapMESA", (__GLXextFuncPtr) glXCreateGLXPixmapMESA }, - - /*** GLX_MESA_release_buffers ***/ - { "glXReleaseBuffersMESA", (__GLXextFuncPtr) glXReleaseBuffersMESA }, - - /*** GLX_MESA_set_3dfx_mode ***/ - { "glXSet3DfxModeMESA", (__GLXextFuncPtr) glXSet3DfxModeMESA }, - - /*** GLX_ARB_get_proc_address ***/ - { "glXGetProcAddressARB", (__GLXextFuncPtr) glXGetProcAddressARB }, - - /*** GLX_NV_vertex_array_range ***/ - { "glXAllocateMemoryNV", (__GLXextFuncPtr) glXAllocateMemoryNV }, - { "glXFreeMemoryNV", (__GLXextFuncPtr) glXFreeMemoryNV }, - - /*** GLX_MESA_agp_offset ***/ - { "glXGetAGPOffsetMESA", (__GLXextFuncPtr) glXGetAGPOffsetMESA }, - - /*** GLX_MESA_allocate_memory ***/ - { "glXAllocateMemoryMESA", (__GLXextFuncPtr) glXAllocateMemoryMESA }, - { "glXFreeMemoryMESA", (__GLXextFuncPtr) glXFreeMemoryMESA }, - { "glXGetMemoryOffsetMESA", (__GLXextFuncPtr) glXGetMemoryOffsetMESA }, - - /*** GLX_EXT_texture_from_pixmap ***/ - { "glXBindTexImageEXT", (__GLXextFuncPtr) glXBindTexImageEXT }, - { "glXReleaseTexImageEXT", (__GLXextFuncPtr) glXReleaseTexImageEXT }, - - { NULL, NULL } /* end of list */ -}; - - - -/* - * Return address of named glX function, or NULL if not found. - */ -__GLXextFuncPtr -_glxapi_get_proc_address(const char *funcName) -{ - GLuint i; - for (i = 0; GLX_functions[i].Name; i++) { - if (strcmp(GLX_functions[i].Name, funcName) == 0) - return GLX_functions[i].Address; - } - return NULL; -} - - - -/* - * This function does not get dispatched through the dispatch table - * since it's really a "meta" function. - */ -__GLXextFuncPtr -glXGetProcAddressARB(const GLubyte *procName) -{ - __GLXextFuncPtr f; - - f = _glxapi_get_proc_address((const char *) procName); - if (f) { - return f; - } - - f = (__GLXextFuncPtr) _glapi_get_proc_address((const char *) procName); - return f; -} - - -/* GLX 1.4 */ -void (*glXGetProcAddress(const GLubyte *procName))() -{ - return glXGetProcAddressARB(procName); -} diff --git a/src/mesa/pipe/xlib/glxapi.h b/src/mesa/pipe/xlib/glxapi.h deleted file mode 100644 index 37de81e55a..0000000000 --- a/src/mesa/pipe/xlib/glxapi.h +++ /dev/null @@ -1,228 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.3 - * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -#ifndef _glxapi_h_ -#define _glxapi_h_ - - -#define GLX_GLXEXT_PROTOTYPES -#include "GL/glx.h" - - -/* The GLX API dispatcher (i.e. this code) is being built into stand-alone - * Mesa. We don't know anything about XFree86 or real GLX so we define a - * minimal __GLXContextRec here so some of the functions in this file can - * work properly. - */ -typedef struct __GLXcontextRec { - Display *currentDpy; - GLboolean isDirect; - GLXDrawable currentDrawable; - GLXDrawable currentReadable; - XID xid; -} __GLXcontext; - - -/* - * Almost all the GLX API functions get routed through this dispatch table. - * The exceptions are the glXGetCurrentXXX() functions. - * - * This dispatch table allows multiple GLX client-side modules to coexist. - * Specifically, a real GLX library (like SGI's or the Utah GLX) and Mesa's - * pseudo-GLX can be present at the same time. The former being used on - * GLX-enabled X servers and the later on non-GLX X servers. - * - * Red Hat has been using this since Red Hat Linux 7.0 (I think). - * This'll be a standard feature in XFree86 4.3. It basically allows one - * libGL to do both DRI-rendering and "fake GLX" rendering to X displays - * that lack the GLX extension. - */ -struct _glxapi_table { - /*** GLX_VERSION_1_0 ***/ - XVisualInfo *(*ChooseVisual)(Display *dpy, int screen, int *list); - void (*CopyContext)(Display *dpy, GLXContext src, GLXContext dst, unsigned long mask); - GLXContext (*CreateContext)(Display *dpy, XVisualInfo *visinfo, GLXContext shareList, Bool direct); - GLXPixmap (*CreateGLXPixmap)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap); - void (*DestroyContext)(Display *dpy, GLXContext ctx); - void (*DestroyGLXPixmap)(Display *dpy, GLXPixmap pixmap); - int (*GetConfig)(Display *dpy, XVisualInfo *visinfo, int attrib, int *value); - /*GLXContext (*GetCurrentContext)(void);*/ - /*GLXDrawable (*GetCurrentDrawable)(void);*/ - Bool (*IsDirect)(Display *dpy, GLXContext ctx); - Bool (*MakeCurrent)(Display *dpy, GLXDrawable drawable, GLXContext ctx); - Bool (*QueryExtension)(Display *dpy, int *errorb, int *event); - Bool (*QueryVersion)(Display *dpy, int *maj, int *min); - void (*SwapBuffers)(Display *dpy, GLXDrawable drawable); - void (*UseXFont)(Font font, int first, int count, int listBase); - void (*WaitGL)(void); - void (*WaitX)(void); - - /*** GLX_VERSION_1_1 ***/ - const char *(*GetClientString)(Display *dpy, int name); - const char *(*QueryExtensionsString)(Display *dpy, int screen); - const char *(*QueryServerString)(Display *dpy, int screen, int name); - - /*** GLX_VERSION_1_2 ***/ - /*Display *(*GetCurrentDisplay)(void);*/ - - /*** GLX_VERSION_1_3 ***/ - GLXFBConfig *(*ChooseFBConfig)(Display *dpy, int screen, const int *attribList, int *nitems); - GLXContext (*CreateNewContext)(Display *dpy, GLXFBConfig config, int renderType, GLXContext shareList, Bool direct); - GLXPbuffer (*CreatePbuffer)(Display *dpy, GLXFBConfig config, const int *attribList); - GLXPixmap (*CreatePixmap)(Display *dpy, GLXFBConfig config, Pixmap pixmap, const int *attribList); - GLXWindow (*CreateWindow)(Display *dpy, GLXFBConfig config, Window win, const int *attribList); - void (*DestroyPbuffer)(Display *dpy, GLXPbuffer pbuf); - void (*DestroyPixmap)(Display *dpy, GLXPixmap pixmap); - void (*DestroyWindow)(Display *dpy, GLXWindow window); - /*GLXDrawable (*GetCurrentReadDrawable)(void);*/ - int (*GetFBConfigAttrib)(Display *dpy, GLXFBConfig config, int attribute, int *value); - GLXFBConfig *(*GetFBConfigs)(Display *dpy, int screen, int *nelements); - void (*GetSelectedEvent)(Display *dpy, GLXDrawable drawable, unsigned long *mask); - XVisualInfo *(*GetVisualFromFBConfig)(Display *dpy, GLXFBConfig config); - Bool (*MakeContextCurrent)(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx); - int (*QueryContext)(Display *dpy, GLXContext ctx, int attribute, int *value); - void (*QueryDrawable)(Display *dpy, GLXDrawable draw, int attribute, unsigned int *value); - void (*SelectEvent)(Display *dpy, GLXDrawable drawable, unsigned long mask); - - /*** GLX_SGI_swap_control ***/ - int (*SwapIntervalSGI)(int); - - /*** GLX_SGI_video_sync ***/ - int (*GetVideoSyncSGI)(unsigned int *count); - int (*WaitVideoSyncSGI)(int divisor, int remainder, unsigned int *count); - - /*** GLX_SGI_make_current_read ***/ - Bool (*MakeCurrentReadSGI)(Display *, GLXDrawable, GLXDrawable, GLXContext); - /*GLXDrawable (*GetCurrentReadDrawableSGI)(void);*/ - - /*** GLX_SGIX_video_source (needs video library) ***/ -#if defined(_VL_H_) - GLXVideoSourceSGIX (*CreateGLXVideoSourceSGIX)(Display *, int, VLServer, VLPath, int, VLNode); - void (*DestroyGLXVideoSourceSGIX)(Display *, GLXVideoSourceSGIX); -#else - void *CreateGLXVideoSourceSGIX; - void *DestroyGLXVideoSourceSGIX; -#endif - - /*** GLX_EXT_import_context ***/ - void (*FreeContextEXT)(Display *dpy, GLXContext context); - GLXContextID (*GetContextIDEXT)(const GLXContext context); - /*Display *(*GetCurrentDisplayEXT)(void);*/ - GLXContext (*ImportContextEXT)(Display *dpy, GLXContextID contextID); - int (*QueryContextInfoEXT)(Display *dpy, GLXContext context, int attribute,int *value); - - /*** GLX_SGIX_fbconfig ***/ - int (*GetFBConfigAttribSGIX)(Display *, GLXFBConfigSGIX, int, int *); - GLXFBConfigSGIX * (*ChooseFBConfigSGIX)(Display *, int, int *, int *); - GLXPixmap (*CreateGLXPixmapWithConfigSGIX)(Display *, GLXFBConfigSGIX, Pixmap); - GLXContext (*CreateContextWithConfigSGIX)(Display *, GLXFBConfigSGIX, int, GLXContext, Bool); - XVisualInfo * (*GetVisualFromFBConfigSGIX)(Display *, GLXFBConfigSGIX); - GLXFBConfigSGIX (*GetFBConfigFromVisualSGIX)(Display *, XVisualInfo *); - - /*** GLX_SGIX_pbuffer ***/ - GLXPbufferSGIX (*CreateGLXPbufferSGIX)(Display *, GLXFBConfigSGIX, unsigned int, unsigned int, int *); - void (*DestroyGLXPbufferSGIX)(Display *, GLXPbufferSGIX); - int (*QueryGLXPbufferSGIX)(Display *, GLXPbufferSGIX, int, unsigned int *); - void (*SelectEventSGIX)(Display *, GLXDrawable, unsigned long); - void (*GetSelectedEventSGIX)(Display *, GLXDrawable, unsigned long *); - - /*** GLX_SGI_cushion ***/ - void (*CushionSGI)(Display *, Window, float); - - /*** GLX_SGIX_video_resize ***/ - int (*BindChannelToWindowSGIX)(Display *, int, int, Window); - int (*ChannelRectSGIX)(Display *, int, int, int, int, int, int); - int (*QueryChannelRectSGIX)(Display *, int, int, int *, int *, int *, int *); - int (*QueryChannelDeltasSGIX)(Display *, int, int, int *, int *, int *, int *); - int (*ChannelRectSyncSGIX)(Display *, int, int, GLenum); - - /*** GLX_SGIX_dmbuffer (needs dmedia library) ***/ -#if defined (_DM_BUFFER_H_) - Bool (*AssociateDMPbufferSGIX)(Display *, GLXPbufferSGIX, DMparams *, DMbuffer); -#else - void *AssociciateDMPbufferSGIX; -#endif - - /*** GLX_SGIX_swap_group ***/ - void (*JoinSwapGroupSGIX)(Display *, GLXDrawable, GLXDrawable); - - /*** GLX_SGIX_swap_barrier ***/ - void (*BindSwapBarrierSGIX)(Display *, GLXDrawable, int); - Bool (*QueryMaxSwapBarriersSGIX)(Display *, int, int *); - - /*** GLX_SUN_get_transparent_index ***/ - Status (*GetTransparentIndexSUN)(Display *, Window, Window, long *); - - /*** GLX_MESA_copy_sub_buffer ***/ - void (*CopySubBufferMESA)(Display *dpy, GLXDrawable drawable, int x, int y, int width, int height); - - /*** GLX_MESA_release_buffers ***/ - Bool (*ReleaseBuffersMESA)(Display *dpy, Window w); - - /*** GLX_MESA_pixmap_colormap ***/ - GLXPixmap (*CreateGLXPixmapMESA)(Display *dpy, XVisualInfo *visinfo, Pixmap pixmap, Colormap cmap); - - /*** GLX_MESA_set_3dfx_mode ***/ - Bool (*Set3DfxModeMESA)(int mode); - - /*** GLX_NV_vertex_array_range ***/ - void * (*AllocateMemoryNV)( GLsizei size, - GLfloat readFrequency, - GLfloat writeFrequency, - GLfloat priority ); - void (*FreeMemoryNV)( GLvoid *pointer ); - - /*** GLX_MESA_agp_offset ***/ - GLuint (*GetAGPOffsetMESA)( const GLvoid *pointer ); - - /*** GLX_EXT_texture_from_pixmap ***/ - void (*BindTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer, - const int *attrib_list); - void (*ReleaseTexImageEXT)(Display *dpy, GLXDrawable drawable, int buffer); -}; - - - -extern const char * -_glxapi_get_version(void); - - -extern const char ** -_glxapi_get_extensions(void); - - -extern GLuint -_glxapi_get_dispatch_table_size(void); - - -extern void -_glxapi_set_no_op_table(struct _glxapi_table *t); - - -extern __GLXextFuncPtr -_glxapi_get_proc_address(const char *funcName); - - -#endif diff --git a/src/mesa/pipe/xlib/glxheader.h b/src/mesa/pipe/xlib/glxheader.h deleted file mode 100644 index a402191f13..0000000000 --- a/src/mesa/pipe/xlib/glxheader.h +++ /dev/null @@ -1,62 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5.1 - * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -#ifndef GLX_HEADER_H -#define GLX_HEADER_H - -#ifdef __VMS -#include -#endif - -#include "glheader.h" - -#ifdef XFree86Server - -# include "resource.h" -# include "windowstr.h" - -#else - -# include -# include -# include -# ifdef USE_XSHM /* was SHM */ -# include -# include -# include -# endif -# include -# include - -#endif - - - -/* this silences a compiler warning on several systems */ -struct timespec; -struct itimerspec; - - -#endif /*GLX_HEADER*/ diff --git a/src/mesa/pipe/xlib/realglx.c b/src/mesa/pipe/xlib/realglx.c deleted file mode 100644 index 30adb7465b..0000000000 --- a/src/mesa/pipe/xlib/realglx.c +++ /dev/null @@ -1,180 +0,0 @@ - -/* - * Mesa 3-D graphics library - * Version: 5.1 - * - * Copyright (C) 1999-2002 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -#include -#include -#include "realglx.h" -#include "glxapi.h" - - -struct _glxapi_table * -_real_GetGLXDispatchTable(void) -{ - static struct _glxapi_table glx; - - /* be sure our dispatch table size <= libGL's table */ - { - GLuint size = sizeof(struct _glxapi_table) / sizeof(void *); - (void) size; - assert(_glxapi_get_dispatch_table_size() >= size); - } - - /* initialize the whole table to no-ops */ - _glxapi_set_no_op_table(&glx); - - /* now initialize the table with the functions I implement */ - - /*** GLX_VERSION_1_0 ***/ - glx.ChooseVisual = _real_glXChooseVisual; - glx.CopyContext = _real_glXCopyContext; - glx.CreateContext = _real_glXCreateContext; - glx.CreateGLXPixmap = _real_glXCreateGLXPixmap; - glx.DestroyContext = _real_glXDestroyContext; - glx.DestroyGLXPixmap = _real_glXDestroyGLXPixmap; - glx.GetConfig = _real_glXGetConfig; - /*glx.GetCurrentContext = _real_glXGetCurrentContext;*/ - /*glx.GetCurrentDrawable = _real_glXGetCurrentDrawable;*/ - glx.IsDirect = _real_glXIsDirect; - glx.MakeCurrent = _real_glXMakeCurrent; - glx.QueryExtension = _real_glXQueryExtension; - glx.QueryVersion = _real_glXQueryVersion; - glx.SwapBuffers = _real_glXSwapBuffers; - glx.UseXFont = _real_glXUseXFont; - glx.WaitGL = _real_glXWaitGL; - glx.WaitX = _real_glXWaitX; - - /*** GLX_VERSION_1_1 ***/ - glx.GetClientString = _real_glXGetClientString; - glx.QueryExtensionsString = _real_glXQueryExtensionsString; - glx.QueryServerString = _real_glXQueryServerString; - - /*** GLX_VERSION_1_2 ***/ - /*glx.GetCurrentDisplay = _real_glXGetCurrentDisplay;*/ - - /*** GLX_VERSION_1_3 ***/ - glx.ChooseFBConfig = _real_glXChooseFBConfig; - glx.CreateNewContext = _real_glXCreateNewContext; - glx.CreatePbuffer = _real_glXCreatePbuffer; - glx.CreatePixmap = _real_glXCreatePixmap; - glx.CreateWindow = _real_glXCreateWindow; - glx.DestroyPbuffer = _real_glXDestroyPbuffer; - glx.DestroyPixmap = _real_glXDestroyPixmap; - glx.DestroyWindow = _real_glXDestroyWindow; - /*glx.GetCurrentReadDrawable = _real_glXGetCurrentReadDrawable;*/ - glx.GetFBConfigAttrib = _real_glXGetFBConfigAttrib; - glx.GetFBConfigs = _real_glXGetFBConfigs; - glx.GetSelectedEvent = _real_glXGetSelectedEvent; - glx.GetVisualFromFBConfig = _real_glXGetVisualFromFBConfig; - glx.MakeContextCurrent = _real_glXMakeContextCurrent; - glx.QueryContext = _real_glXQueryContext; - glx.QueryDrawable = _real_glXQueryDrawable; - glx.SelectEvent = _real_glXSelectEvent; - - /*** GLX_SGI_swap_control ***/ - glx.SwapIntervalSGI = _real_glXSwapIntervalSGI; - - /*** GLX_SGI_video_sync ***/ - glx.GetVideoSyncSGI = _real_glXGetVideoSyncSGI; - glx.WaitVideoSyncSGI = _real_glXWaitVideoSyncSGI; - - /*** GLX_SGI_make_current_read ***/ - glx.MakeCurrentReadSGI = _real_glXMakeCurrentReadSGI; - /*glx.GetCurrentReadDrawableSGI = _real_glXGetCurrentReadDrawableSGI;*/ - -#if defined(_VL_H) - /*** GLX_SGIX_video_source ***/ - glx.CreateGLXVideoSourceSGIX = _real_glXCreateGLXVideoSourceSGIX; - glx.DestroyGLXVideoSourceSGIX = _real_glXDestroyGLXVideoSourceSGIX; -#endif - - /*** GLX_EXT_import_context ***/ - glx.FreeContextEXT = _real_glXFreeContextEXT; - /*glx.GetContextIDEXT = _real_glXGetContextIDEXT;*/ - /*glx.GetCurrentDisplayEXT = _real_glXGetCurrentDisplayEXT;*/ - glx.ImportContextEXT = _real_glXImportContextEXT; - glx.QueryContextInfoEXT = _real_glXQueryContextInfoEXT; - - /*** GLX_SGIX_fbconfig ***/ - glx.GetFBConfigAttribSGIX = _real_glXGetFBConfigAttribSGIX; - glx.ChooseFBConfigSGIX = _real_glXChooseFBConfigSGIX; - glx.CreateGLXPixmapWithConfigSGIX = _real_glXCreateGLXPixmapWithConfigSGIX; - glx.CreateContextWithConfigSGIX = _real_glXCreateContextWithConfigSGIX; - glx.GetVisualFromFBConfigSGIX = _real_glXGetVisualFromFBConfigSGIX; - glx.GetFBConfigFromVisualSGIX = _real_glXGetFBConfigFromVisualSGIX; - - /*** GLX_SGIX_pbuffer ***/ - glx.CreateGLXPbufferSGIX = _real_glXCreateGLXPbufferSGIX; - glx.DestroyGLXPbufferSGIX = _real_glXDestroyGLXPbufferSGIX; - glx.QueryGLXPbufferSGIX = _real_glXQueryGLXPbufferSGIX; - glx.SelectEventSGIX = _real_glXSelectEventSGIX; - glx.GetSelectedEventSGIX = _real_glXGetSelectedEventSGIX; - - /*** GLX_SGI_cushion ***/ - glx.CushionSGI = _real_glXCushionSGI; - - /*** GLX_SGIX_video_resize ***/ - glx.BindChannelToWindowSGIX = _real_glXBindChannelToWindowSGIX; - glx.ChannelRectSGIX = _real_glXChannelRectSGIX; - glx.QueryChannelRectSGIX = _real_glXQueryChannelRectSGIX; - glx.QueryChannelDeltasSGIX = _real_glXQueryChannelDeltasSGIX; - glx.ChannelRectSyncSGIX = _real_glXChannelRectSyncSGIX; - -#if defined(_DM_BUFFER_H_) - /*** (GLX_SGIX_dmbuffer ***/ - glx.AssociateDMPbufferSGIX = NULL; -#endif - - /*** GLX_SGIX_swap_group ***/ - glx.JoinSwapGroupSGIX = _real_glXJoinSwapGroupSGIX; - - /*** GLX_SGIX_swap_barrier ***/ - glx.BindSwapBarrierSGIX = _real_glXBindSwapBarrierSGIX; - glx.QueryMaxSwapBarriersSGIX = _real_glXQueryMaxSwapBarriersSGIX; - - /*** GLX_SUN_get_transparent_index ***/ - glx.GetTransparentIndexSUN = _real_glXGetTransparentIndexSUN; - - /*** GLX_MESA_copy_sub_buffer ***/ - glx.CopySubBufferMESA = _real_glXCopySubBufferMESA; - - /*** GLX_MESA_release_buffers ***/ - glx.ReleaseBuffersMESA = _real_glXReleaseBuffersMESA; - - /*** GLX_MESA_pixmap_colormap ***/ - glx.CreateGLXPixmapMESA = _real_glXCreateGLXPixmapMESA; - - /*** GLX_MESA_set_3dfx_mode ***/ - glx.Set3DfxModeMESA = _real_glXSet3DfxModeMESA; - - /*** GLX_NV_vertex_array_range ***/ - glx.AllocateMemoryNV = _real_glXAllocateMemoryNV; - glx.FreeMemoryNV = _real_glXFreeMemoryNV; - - /*** GLX_MESA_agp_offset ***/ - glx.GetAGPOffsetMESA = _real_glXGetAGPOffsetMESA; - - return &glx; -} diff --git a/src/mesa/pipe/xlib/realglx.h b/src/mesa/pipe/xlib/realglx.h deleted file mode 100644 index 150129db68..0000000000 --- a/src/mesa/pipe/xlib/realglx.h +++ /dev/null @@ -1,326 +0,0 @@ - -/* - * Mesa 3-D graphics library - * Version: 3.5 - * - * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -#ifndef REALGLX_H -#define REALGLX_H - - -extern struct _glxapi_table * -_real_GetGLXDispatchTable(void); - - -/* - * Basically just need these to prevent compiler warnings. - */ - - -extern XVisualInfo * -_real_glXChooseVisual( Display *dpy, int screen, int *list ); - -extern GLXContext -_real_glXCreateContext( Display *dpy, XVisualInfo *visinfo, - GLXContext share_list, Bool direct ); - -extern GLXPixmap -_real_glXCreateGLXPixmap( Display *dpy, XVisualInfo *visinfo, Pixmap pixmap ); - -extern GLXPixmap -_real_glXCreateGLXPixmapMESA( Display *dpy, XVisualInfo *visinfo, - Pixmap pixmap, Colormap cmap ); - -extern void -_real_glXDestroyGLXPixmap( Display *dpy, GLXPixmap pixmap ); - -extern void -_real_glXCopyContext( Display *dpy, GLXContext src, GLXContext dst, - unsigned long mask ); - -extern Bool -_real_glXMakeCurrent( Display *dpy, GLXDrawable drawable, GLXContext ctx ); - -extern Bool -_real_glXQueryExtension( Display *dpy, int *errorb, int *event ); - -extern void -_real_glXDestroyContext( Display *dpy, GLXContext ctx ); - -extern Bool -_real_glXIsDirect( Display *dpy, GLXContext ctx ); - -extern void -_real_glXSwapBuffers( Display *dpy, GLXDrawable drawable ); - -extern void -_real_glXUseXFont( Font font, int first, int count, int listbase ); - -extern Bool -_real_glXQueryVersion( Display *dpy, int *maj, int *min ); - -extern int -_real_glXGetConfig( Display *dpy, XVisualInfo *visinfo, - int attrib, int *value ); - -extern void -_real_glXWaitGL( void ); - - -extern void -_real_glXWaitX( void ); - -/* GLX 1.1 and later */ -extern const char * -_real_glXQueryExtensionsString( Display *dpy, int screen ); - -/* GLX 1.1 and later */ -extern const char * -_real_glXQueryServerString( Display *dpy, int screen, int name ); - -/* GLX 1.1 and later */ -extern const char * -_real_glXGetClientString( Display *dpy, int name ); - - -/* - * GLX 1.3 and later - */ - -extern GLXFBConfig * -_real_glXChooseFBConfig( Display *dpy, int screen, - const int *attribList, int *nitems ); - -extern int -_real_glXGetFBConfigAttrib( Display *dpy, GLXFBConfig config, - int attribute, int *value ); - -extern GLXFBConfig * -_real_glXGetFBConfigs( Display *dpy, int screen, int *nelements ); - -extern XVisualInfo * -_real_glXGetVisualFromFBConfig( Display *dpy, GLXFBConfig config ); - -extern GLXWindow -_real_glXCreateWindow( Display *dpy, GLXFBConfig config, Window win, - const int *attribList ); - -extern void -_real_glXDestroyWindow( Display *dpy, GLXWindow window ); - -extern GLXPixmap -_real_glXCreatePixmap( Display *dpy, GLXFBConfig config, Pixmap pixmap, - const int *attribList ); - -extern void -_real_glXDestroyPixmap( Display *dpy, GLXPixmap pixmap ); - -extern GLXPbuffer -_real_glXCreatePbuffer( Display *dpy, GLXFBConfig config, - const int *attribList ); - -extern void -_real_glXDestroyPbuffer( Display *dpy, GLXPbuffer pbuf ); - -extern void -_real_glXQueryDrawable( Display *dpy, GLXDrawable draw, int attribute, - unsigned int *value ); - -extern GLXContext -_real_glXCreateNewContext( Display *dpy, GLXFBConfig config, - int renderType, GLXContext shareList, Bool direct ); - - -extern Bool -_real_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, - GLXDrawable read, GLXContext ctx ); - -extern int -_real_glXQueryContext( Display *dpy, GLXContext ctx, int attribute, int *value ); - -extern void -_real_glXSelectEvent( Display *dpy, GLXDrawable drawable, unsigned long mask ); - -extern void -_real_glXGetSelectedEvent( Display *dpy, GLXDrawable drawable, - unsigned long *mask ); - -#ifdef GLX_SGI_swap_control -extern int -_real_glXSwapIntervalSGI(int interval); -#endif - - -#ifdef GLX_SGI_video_sync -extern int -_real_glXGetVideoSyncSGI(unsigned int *count); - -extern int -_real_glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count); -#endif - - -#ifdef GLX_SGI_make_current_read -extern Bool -_real_glXMakeCurrentReadSGI(Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx); - -extern GLXDrawable -_real_glXGetCurrentReadDrawableSGI(void); -#endif - -#if defined(_VL_H) && defined(GLX_SGIX_video_source) -extern GLXVideoSourceSGIX -_real_glXCreateGLXVideoSourceSGIX(Display *dpy, int screen, VLServer server, VLPath path, int nodeClass, VLNode drainNode); - -extern void -_real_glXDestroyGLXVideoSourceSGIX(Display *dpy, GLXVideoSourceSGIX src); -#endif - -#ifdef GLX_EXT_import_context -extern void -_real_glXFreeContextEXT(Display *dpy, GLXContext context); - -extern GLXContextID -_real_glXGetContextIDEXT(const GLXContext context); - -extern Display * -_real_glXGetCurrentDisplayEXT(void); - -extern GLXContext -_real_glXImportContextEXT(Display *dpy, GLXContextID contextID); - -extern int -_real_glXQueryContextInfoEXT(Display *dpy, GLXContext context, int attribute, int *value); -#endif - -#ifdef GLX_SGIX_fbconfig -extern int -_real_glXGetFBConfigAttribSGIX(Display *dpy, GLXFBConfigSGIX config, int attribute, int *value); - -extern GLXFBConfigSGIX * -_real_glXChooseFBConfigSGIX(Display *dpy, int screen, int *attrib_list, int *nelements); - -extern GLXPixmap -_real_glXCreateGLXPixmapWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, Pixmap pixmap); - -extern GLXContext -_real_glXCreateContextWithConfigSGIX(Display *dpy, GLXFBConfigSGIX config, int render_type, GLXContext share_list, Bool direct); - -extern XVisualInfo * -_real_glXGetVisualFromFBConfigSGIX(Display *dpy, GLXFBConfigSGIX config); - -extern GLXFBConfigSGIX -_real_glXGetFBConfigFromVisualSGIX(Display *dpy, XVisualInfo *vis); -#endif - -#ifdef GLX_SGIX_pbuffer -extern GLXPbufferSGIX -_real_glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, unsigned int width, unsigned int height, int *attrib_list); - -extern void -_real_glXDestroyGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf); - -extern int -_real_glXQueryGLXPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuf, int attribute, unsigned int *value); - -extern void -_real_glXSelectEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long mask); - -extern void -_real_glXGetSelectedEventSGIX(Display *dpy, GLXDrawable drawable, unsigned long *mask); -#endif - -#ifdef GLX_SGI_cushion -extern void -_real_glXCushionSGI(Display *dpy, Window win, float cushion); -#endif - -#ifdef GLX_SGIX_video_resize -extern int -_real_glXBindChannelToWindowSGIX(Display *dpy, int screen, int channel , Window window); - -extern int -_real_glXChannelRectSGIX(Display *dpy, int screen, int channel, int x, int y, int w, int h); - -extern int -_real_glXQueryChannelRectSGIX(Display *dpy, int screen, int channel, int *x, int *y, int *w, int *h); - -extern int -_real_glXQueryChannelDeltasSGIX(Display *dpy, int screen, int channel, int *dx, int *dy, int *dw, int *dh); - -extern int -_real_glXChannelRectSyncSGIX(Display *dpy, int screen, int channel, GLenum synctype); -#endif - -#if defined(_DM_BUFFER_H_) && defined(GLX_SGIX_dmbuffer) -extern Bool -_real_glXAssociateDMPbufferSGIX(Display *dpy, GLXPbufferSGIX pbuffer, DMparams *params, DMbuffer dmbuffer); -#endif - -#ifdef GLX_SGIX_swap_group -extern void -_real_glXJoinSwapGroupSGIX(Display *dpy, GLXDrawable drawable, GLXDrawable member); -#endif - -#ifdef GLX_SGIX_swap_barrier -extern void -_real_glXBindSwapBarrierSGIX(Display *dpy, GLXDrawable drawable, int barrier); - -extern Bool -_real_glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max); -#endif - -#ifdef GLX_SUN_get_transparent_index -extern Status -_real_glXGetTransparentIndexSUN(Display *dpy, Window overlay, Window underlay, long *pTransparent); -#endif - -#ifdef GLX_MESA_release_buffers -extern Bool -_real_glXReleaseBuffersMESA( Display *dpy, GLXDrawable d ); -#endif - -#ifdef GLX_MESA_set_3dfx_mode -extern Bool -_real_glXSet3DfxModeMESA( int mode ); -#endif - -#ifdef GLX_NV_vertex_array_range -extern void * -_real_glXAllocateMemoryNV(GLsizei size, GLfloat readfreq, GLfloat writefreq, GLfloat priority); -extern void -_real_glXFreeMemoryNV(GLvoid *pointer); -#endif - -#ifdef GLX_MESA_agp_offset -extern GLuint -_real_glXGetAGPOffsetMESA(const GLvoid *pointer); -#endif - -#ifdef GLX_MESA_copy_sub_buffer -extern void -_real_glXCopySubBufferMESA( Display *dpy, GLXDrawable drawable, - int x, int y, int width, int height ); -#endif - -#endif /* REALGLX_H */ diff --git a/src/mesa/pipe/xlib/xfonts.c b/src/mesa/pipe/xlib/xfonts.c deleted file mode 100644 index d72c600bd1..0000000000 --- a/src/mesa/pipe/xlib/xfonts.c +++ /dev/null @@ -1,377 +0,0 @@ - -/* - * Mesa 3-D graphics library - * Version: 3.5 - * - * Copyright (C) 1999-2000 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/* xfonts.c -- glXUseXFont() for Mesa written by - * Copyright (C) 1995 Thorsten.Ohl @ Physik.TH-Darmstadt.de - */ - -#ifdef __VMS -#include -#endif - -#include "glxheader.h" -#include "context.h" -#include "imports.h" -#include "xfonts.h" - - -/* Some debugging info. */ - -#ifdef DEBUG -#undef _R -#undef _G -#undef _B -#include - -int debug_xfonts = 0; - -static void -dump_char_struct(XCharStruct * ch, char *prefix) -{ - printf("%slbearing = %d, rbearing = %d, width = %d\n", - prefix, ch->lbearing, ch->rbearing, ch->width); - printf("%sascent = %d, descent = %d, attributes = %u\n", - prefix, ch->ascent, ch->descent, (unsigned int) ch->attributes); -} - -static void -dump_font_struct(XFontStruct * font) -{ - printf("ascent = %d, descent = %d\n", font->ascent, font->descent); - printf("char_or_byte2 = (%u,%u)\n", - font->min_char_or_byte2, font->max_char_or_byte2); - printf("byte1 = (%u,%u)\n", font->min_byte1, font->max_byte1); - printf("all_chars_exist = %s\n", font->all_chars_exist ? "True" : "False"); - printf("default_char = %c (\\%03o)\n", - (char) (isprint(font->default_char) ? font->default_char : ' '), - font->default_char); - dump_char_struct(&font->min_bounds, "min> "); - dump_char_struct(&font->max_bounds, "max> "); -#if 0 - for (c = font->min_char_or_byte2; c <= font->max_char_or_byte2; c++) { - char prefix[8]; - sprintf(prefix, "%d> ", c); - dump_char_struct(&font->per_char[c], prefix); - } -#endif -} - -static void -dump_bitmap(unsigned int width, unsigned int height, GLubyte * bitmap) -{ - unsigned int x, y; - - printf(" "); - for (x = 0; x < 8 * width; x++) - printf("%o", 7 - (x % 8)); - putchar('\n'); - for (y = 0; y < height; y++) { - printf("%3o:", y); - for (x = 0; x < 8 * width; x++) - putchar((bitmap[width * (height - y - 1) + x / 8] & (1 << (7 - (x % - 8)))) - ? '*' : '.'); - printf(" "); - for (x = 0; x < width; x++) - printf("0x%02x, ", bitmap[width * (height - y - 1) + x]); - putchar('\n'); - } -} -#endif /* DEBUG */ - - -/* Implementation. */ - -/* Fill a BITMAP with a character C from thew current font - in the graphics context GC. WIDTH is the width in bytes - and HEIGHT is the height in bits. - - Note that the generated bitmaps must be used with - - glPixelStorei (GL_UNPACK_SWAP_BYTES, GL_FALSE); - glPixelStorei (GL_UNPACK_LSB_FIRST, GL_FALSE); - glPixelStorei (GL_UNPACK_ROW_LENGTH, 0); - glPixelStorei (GL_UNPACK_SKIP_ROWS, 0); - glPixelStorei (GL_UNPACK_SKIP_PIXELS, 0); - glPixelStorei (GL_UNPACK_ALIGNMENT, 1); - - Possible optimizations: - - * use only one reusable pixmap with the maximum dimensions. - * draw the entire font into a single pixmap (careful with - proportional fonts!). -*/ - - -/* - * Generate OpenGL-compatible bitmap. - */ -static void -fill_bitmap(Display * dpy, Window win, GC gc, - unsigned int width, unsigned int height, - int x0, int y0, unsigned int c, GLubyte * bitmap) -{ - XImage *image; - unsigned int x, y; - Pixmap pixmap; - XChar2b char2b; - - pixmap = XCreatePixmap(dpy, win, 8 * width, height, 1); - XSetForeground(dpy, gc, 0); - XFillRectangle(dpy, pixmap, gc, 0, 0, 8 * width, height); - XSetForeground(dpy, gc, 1); - - char2b.byte1 = (c >> 8) & 0xff; - char2b.byte2 = (c & 0xff); - - XDrawString16(dpy, pixmap, gc, x0, y0, &char2b, 1); - - image = XGetImage(dpy, pixmap, 0, 0, 8 * width, height, 1, XYPixmap); - if (image) { - /* Fill the bitmap (X11 and OpenGL are upside down wrt each other). */ - for (y = 0; y < height; y++) - for (x = 0; x < 8 * width; x++) - if (XGetPixel(image, x, y)) - bitmap[width * (height - y - 1) + x / 8] |= - (1 << (7 - (x % 8))); - XDestroyImage(image); - } - - XFreePixmap(dpy, pixmap); -} - -/* - * determine if a given glyph is valid and return the - * corresponding XCharStruct. - */ -static XCharStruct * -isvalid(XFontStruct * fs, unsigned int which) -{ - unsigned int rows, pages; - unsigned int byte1 = 0, byte2 = 0; - int i, valid = 1; - - rows = fs->max_byte1 - fs->min_byte1 + 1; - pages = fs->max_char_or_byte2 - fs->min_char_or_byte2 + 1; - - if (rows == 1) { - /* "linear" fonts */ - if ((fs->min_char_or_byte2 > which) || (fs->max_char_or_byte2 < which)) - valid = 0; - } - else { - /* "matrix" fonts */ - byte2 = which & 0xff; - byte1 = which >> 8; - if ((fs->min_char_or_byte2 > byte2) || - (fs->max_char_or_byte2 < byte2) || - (fs->min_byte1 > byte1) || (fs->max_byte1 < byte1)) - valid = 0; - } - - if (valid) { - if (fs->per_char) { - if (rows == 1) { - /* "linear" fonts */ - return (fs->per_char + (which - fs->min_char_or_byte2)); - } - else { - /* "matrix" fonts */ - i = ((byte1 - fs->min_byte1) * pages) + - (byte2 - fs->min_char_or_byte2); - return (fs->per_char + i); - } - } - else { - return (&fs->min_bounds); - } - } - return (NULL); -} - - -void -Fake_glXUseXFont(Font font, int first, int count, int listbase) -{ - Display *dpy; - Window win; - Pixmap pixmap; - GC gc; - XGCValues values; - unsigned long valuemask; - XFontStruct *fs; - GLint swapbytes, lsbfirst, rowlength; - GLint skiprows, skippixels, alignment; - unsigned int max_width, max_height, max_bm_width, max_bm_height; - GLubyte *bm; - int i; - - dpy = glXGetCurrentDisplay(); - if (!dpy) - return; /* I guess glXMakeCurrent wasn't called */ - win = RootWindow(dpy, DefaultScreen(dpy)); - - fs = XQueryFont(dpy, font); - if (!fs) { - _mesa_error(NULL, GL_INVALID_VALUE, - "Couldn't get font structure information"); - return; - } - - /* Allocate a bitmap that can fit all characters. */ - max_width = fs->max_bounds.rbearing - fs->min_bounds.lbearing; - max_height = fs->max_bounds.ascent + fs->max_bounds.descent; - max_bm_width = (max_width + 7) / 8; - max_bm_height = max_height; - - bm = (GLubyte *) MALLOC((max_bm_width * max_bm_height) * sizeof(GLubyte)); - if (!bm) { - XFreeFontInfo(NULL, fs, 1); - _mesa_error(NULL, GL_OUT_OF_MEMORY, - "Couldn't allocate bitmap in glXUseXFont()"); - return; - } - -#if 0 - /* get the page info */ - pages = fs->max_char_or_byte2 - fs->min_char_or_byte2 + 1; - firstchar = (fs->min_byte1 << 8) + fs->min_char_or_byte2; - lastchar = (fs->max_byte1 << 8) + fs->max_char_or_byte2; - rows = fs->max_byte1 - fs->min_byte1 + 1; - unsigned int first_char, last_char, pages, rows; -#endif - - /* Save the current packing mode for bitmaps. */ - glGetIntegerv(GL_UNPACK_SWAP_BYTES, &swapbytes); - glGetIntegerv(GL_UNPACK_LSB_FIRST, &lsbfirst); - glGetIntegerv(GL_UNPACK_ROW_LENGTH, &rowlength); - glGetIntegerv(GL_UNPACK_SKIP_ROWS, &skiprows); - glGetIntegerv(GL_UNPACK_SKIP_PIXELS, &skippixels); - glGetIntegerv(GL_UNPACK_ALIGNMENT, &alignment); - - /* Enforce a standard packing mode which is compatible with - fill_bitmap() from above. This is actually the default mode, - except for the (non)alignment. */ - glPixelStorei(GL_UNPACK_SWAP_BYTES, GL_FALSE); - glPixelStorei(GL_UNPACK_LSB_FIRST, GL_FALSE); - glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); - glPixelStorei(GL_UNPACK_SKIP_ROWS, 0); - glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0); - glPixelStorei(GL_UNPACK_ALIGNMENT, 1); - - pixmap = XCreatePixmap(dpy, win, 10, 10, 1); - values.foreground = BlackPixel(dpy, DefaultScreen(dpy)); - values.background = WhitePixel(dpy, DefaultScreen(dpy)); - values.font = fs->fid; - valuemask = GCForeground | GCBackground | GCFont; - gc = XCreateGC(dpy, pixmap, valuemask, &values); - XFreePixmap(dpy, pixmap); - -#ifdef DEBUG - if (debug_xfonts) - dump_font_struct(fs); -#endif - - for (i = 0; i < count; i++) { - unsigned int width, height, bm_width, bm_height; - GLfloat x0, y0, dx, dy; - XCharStruct *ch; - int x, y; - unsigned int c = first + i; - int list = listbase + i; - int valid; - - /* check on index validity and get the bounds */ - ch = isvalid(fs, c); - if (!ch) { - ch = &fs->max_bounds; - valid = 0; - } - else { - valid = 1; - } - -#ifdef DEBUG - if (debug_xfonts) { - char s[7]; - sprintf(s, isprint(c) ? "%c> " : "\\%03o> ", c); - dump_char_struct(ch, s); - } -#endif - - /* glBitmap()' parameters: - straight from the glXUseXFont(3) manpage. */ - width = ch->rbearing - ch->lbearing; - height = ch->ascent + ch->descent; - x0 = -ch->lbearing; - y0 = ch->descent - 0; /* XXX used to subtract 1 here */ - /* but that caused a conformace failure */ - dx = ch->width; - dy = 0; - - /* X11's starting point. */ - x = -ch->lbearing; - y = ch->ascent; - - /* Round the width to a multiple of eight. We will use this also - for the pixmap for capturing the X11 font. This is slightly - inefficient, but it makes the OpenGL part real easy. */ - bm_width = (width + 7) / 8; - bm_height = height; - - glNewList(list, GL_COMPILE); - if (valid && (bm_width > 0) && (bm_height > 0)) { - - MEMSET(bm, '\0', bm_width * bm_height); - fill_bitmap(dpy, win, gc, bm_width, bm_height, x, y, c, bm); - - glBitmap(width, height, x0, y0, dx, dy, bm); -#ifdef DEBUG - if (debug_xfonts) { - printf("width/height = %u/%u\n", width, height); - printf("bm_width/bm_height = %u/%u\n", bm_width, bm_height); - dump_bitmap(bm_width, bm_height, bm); - } -#endif - } - else { - glBitmap(0, 0, 0.0, 0.0, dx, dy, NULL); - } - glEndList(); - } - - FREE(bm); - XFreeFontInfo(NULL, fs, 1); - XFreeGC(dpy, gc); - - /* Restore saved packing modes. */ - glPixelStorei(GL_UNPACK_SWAP_BYTES, swapbytes); - glPixelStorei(GL_UNPACK_LSB_FIRST, lsbfirst); - glPixelStorei(GL_UNPACK_ROW_LENGTH, rowlength); - glPixelStorei(GL_UNPACK_SKIP_ROWS, skiprows); - glPixelStorei(GL_UNPACK_SKIP_PIXELS, skippixels); - glPixelStorei(GL_UNPACK_ALIGNMENT, alignment); -} diff --git a/src/mesa/pipe/xlib/xfonts.h b/src/mesa/pipe/xlib/xfonts.h deleted file mode 100644 index e36f42f817..0000000000 --- a/src/mesa/pipe/xlib/xfonts.h +++ /dev/null @@ -1,41 +0,0 @@ - -/* - * Mesa 3-D graphics library - * Version: 3.5 - * - * Copyright (C) 1999-2000 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -#ifndef XFONTS_H -#define XFONTS_H - -#ifdef __VMS -#include -#endif - -#include - - -extern void Fake_glXUseXFont( Font font, int first, int count, int listbase ); - - -#endif - diff --git a/src/mesa/pipe/xlib/xm_api.c b/src/mesa/pipe/xlib/xm_api.c deleted file mode 100644 index e5fef1d7a8..0000000000 --- a/src/mesa/pipe/xlib/xm_api.c +++ /dev/null @@ -1,1380 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 7.1 - * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file xm_api.c - * - * All the XMesa* API functions. - * - * - * NOTES: - * - * The window coordinate system origin (0,0) is in the lower-left corner - * of the window. X11's window coordinate origin is in the upper-left - * corner of the window. Therefore, most drawing functions in this - * file have to flip Y coordinates. - * - * Define USE_XSHM in the Makefile with -DUSE_XSHM if you want to compile - * in support for the MIT Shared Memory extension. If enabled, when you - * use an Ximage for the back buffer in double buffered mode, the "swap" - * operation will be faster. You must also link with -lXext. - * - * Byte swapping: If the Mesa host and the X display use a different - * byte order then there's some trickiness to be aware of when using - * XImages. The byte ordering used for the XImage is that of the X - * display, not the Mesa host. - * The color-to-pixel encoding for True/DirectColor must be done - * according to the display's visual red_mask, green_mask, and blue_mask. - * If XPutPixel is used to put a pixel into an XImage then XPutPixel will - * do byte swapping if needed. If one wants to directly "poke" the pixel - * into the XImage's buffer then the pixel must be byte swapped first. - * - */ - -#ifdef __CYGWIN__ -#undef WIN32 -#undef __WIN32__ -#endif - -#include "glxheader.h" -#include "GL/xmesa.h" -#include "xmesaP.h" -#include "main/context.h" -#include "main/framebuffer.h" -#include "glapi/glthread.h" - -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_context.h" - -#include "xm_winsys_aub.h" - -/** - * Global X driver lock - */ -_glthread_Mutex _xmesa_lock; - - -int xmesa_mode; - - -/**********************************************************************/ -/***** X Utility Functions *****/ -/**********************************************************************/ - - -/** - * Return the host's byte order as LSBFirst or MSBFirst ala X. - */ -#ifndef XFree86Server -static int host_byte_order( void ) -{ - int i = 1; - char *cptr = (char *) &i; - return (*cptr==1) ? LSBFirst : MSBFirst; -} -#endif - - -/** - * Check if the X Shared Memory extension is available. - * Return: 0 = not available - * 1 = shared XImage support available - * 2 = shared Pixmap support available also - */ -static int check_for_xshm( XMesaDisplay *display ) -{ -#if defined(USE_XSHM) && !defined(XFree86Server) - int major, minor, ignore; - Bool pixmaps; - - if (XQueryExtension( display, "MIT-SHM", &ignore, &ignore, &ignore )) { - if (XShmQueryVersion( display, &major, &minor, &pixmaps )==True) { - return (pixmaps==True) ? 2 : 1; - } - else { - return 0; - } - } - else { - return 0; - } -#else - /* No XSHM support */ - return 0; -#endif -} - - -/** - * Return the true number of bits per pixel for XImages. - * For example, if we request a 24-bit deep visual we may actually need/get - * 32bpp XImages. This function returns the appropriate bpp. - * Input: dpy - the X display - * visinfo - desribes the visual to be used for XImages - * Return: true number of bits per pixel for XImages - */ -static int -bits_per_pixel( XMesaVisual xmv ) -{ -#ifdef XFree86Server - const int depth = xmv->nplanes; - int i; - assert(depth > 0); - for (i = 0; i < screenInfo.numPixmapFormats; i++) { - if (screenInfo.formats[i].depth == depth) - return screenInfo.formats[i].bitsPerPixel; - } - return depth; /* should never get here, but this should be safe */ -#else - XMesaDisplay *dpy = xmv->display; - XMesaVisualInfo visinfo = xmv->visinfo; - XMesaImage *img; - int bitsPerPixel; - /* Create a temporary XImage */ - img = XCreateImage( dpy, visinfo->visual, visinfo->depth, - ZPixmap, 0, /*format, offset*/ - (char*) MALLOC(8), /*data*/ - 1, 1, /*width, height*/ - 32, /*bitmap_pad*/ - 0 /*bytes_per_line*/ - ); - assert(img); - /* grab the bits/pixel value */ - bitsPerPixel = img->bits_per_pixel; - /* free the XImage */ - _mesa_free( img->data ); - img->data = NULL; - XMesaDestroyImage( img ); - return bitsPerPixel; -#endif -} - - - -/* - * Determine if a given X window ID is valid (window exists). - * Do this by calling XGetWindowAttributes() for the window and - * checking if we catch an X error. - * Input: dpy - the display - * win - the window to check for existance - * Return: GL_TRUE - window exists - * GL_FALSE - window doesn't exist - */ -#ifndef XFree86Server -static GLboolean WindowExistsFlag; - -static int window_exists_err_handler( XMesaDisplay* dpy, XErrorEvent* xerr ) -{ - (void) dpy; - if (xerr->error_code == BadWindow) { - WindowExistsFlag = GL_FALSE; - } - return 0; -} - -static GLboolean window_exists( XMesaDisplay *dpy, Window win ) -{ - XWindowAttributes wa; - int (*old_handler)( XMesaDisplay*, XErrorEvent* ); - WindowExistsFlag = GL_TRUE; - old_handler = XSetErrorHandler(window_exists_err_handler); - XGetWindowAttributes( dpy, win, &wa ); /* dummy request */ - XSetErrorHandler(old_handler); - return WindowExistsFlag; -} - -static Status -get_drawable_size( XMesaDisplay *dpy, Drawable d, uint *width, uint *height ) -{ - Window root; - Status stat; - int xpos, ypos; - unsigned int w, h, bw, depth; - stat = XGetGeometry(dpy, d, &root, &xpos, &ypos, &w, &h, &bw, &depth); - *width = w; - *height = h; - return stat; -} -#endif - - -/** - * Return the size of the window (or pixmap) that corresponds to the - * given XMesaBuffer. - * \param width returns width in pixels - * \param height returns height in pixels - */ -static void -xmesa_get_window_size(XMesaDisplay *dpy, XMesaBuffer b, - GLuint *width, GLuint *height) -{ -#ifdef XFree86Server - *width = MIN2(b->drawable->width, MAX_WIDTH); - *height = MIN2(b->drawable->height, MAX_HEIGHT); -#else - Status stat; - - _glthread_LOCK_MUTEX(_xmesa_lock); - XSync(b->xm_visual->display, 0); /* added for Chromium */ - stat = get_drawable_size(dpy, b->drawable, width, height); - _glthread_UNLOCK_MUTEX(_xmesa_lock); - - if (!stat) { - /* probably querying a window that's recently been destroyed */ - _mesa_warning(NULL, "XGetGeometry failed!\n"); - *width = *height = 1; - } -#endif -} - - -/** - * Choose the pixel format for the given visual. - * This will tell the gallium driver how to pack pixel data into - * drawing surfaces. - */ -static GLuint -choose_pixel_format(XMesaVisual v) -{ - if ( GET_REDMASK(v) == 0x0000ff - && GET_GREENMASK(v) == 0x00ff00 - && GET_BLUEMASK(v) == 0xff0000 - && v->BitsPerPixel == 32) { - if (CHECK_BYTE_ORDER(v)) { - /* no byteswapping needed */ - return 0 /* PIXEL_FORMAT_U_A8_B8_G8_R8 */; - } - else { - return PIPE_FORMAT_R8G8B8A8_UNORM; - } - } - else if ( GET_REDMASK(v) == 0xff0000 - && GET_GREENMASK(v) == 0x00ff00 - && GET_BLUEMASK(v) == 0x0000ff - && v->BitsPerPixel == 32) { - if (CHECK_BYTE_ORDER(v)) { - /* no byteswapping needed */ - return PIPE_FORMAT_A8R8G8B8_UNORM; - } - else { - return PIPE_FORMAT_B8G8R8A8_UNORM; - } - } - else if ( GET_REDMASK(v) == 0xf800 - && GET_GREENMASK(v) == 0x07e0 - && GET_BLUEMASK(v) == 0x001f - && CHECK_BYTE_ORDER(v) - && v->BitsPerPixel == 16) { - /* 5-6-5 RGB */ - return PIPE_FORMAT_R5G6B5_UNORM; - } - - assert(0); - return 0; -} - - - -/**********************************************************************/ -/***** Linked list of XMesaBuffers *****/ -/**********************************************************************/ - -XMesaBuffer XMesaBufferList = NULL; - - -/** - * Allocate a new XMesaBuffer object which corresponds to the given drawable. - * Note that XMesaBuffer is derived from GLframebuffer. - * The new XMesaBuffer will not have any size (Width=Height=0). - * - * \param d the corresponding X drawable (window or pixmap) - * \param type either WINDOW, PIXMAP or PBUFFER, describing d - * \param vis the buffer's visual - * \param cmap the window's colormap, if known. - * \return new XMesaBuffer or NULL if any problem - */ -static XMesaBuffer -create_xmesa_buffer(XMesaDrawable d, BufferType type, - XMesaVisual vis, XMesaColormap cmap) -{ - XMesaBuffer b; - GLframebuffer *fb; - enum pipe_format colorFormat, depthFormat, stencilFormat; - uint width, height; - - ASSERT(type == WINDOW || type == PIXMAP || type == PBUFFER); - - b = (XMesaBuffer) CALLOC_STRUCT(xmesa_buffer); - if (!b) - return NULL; - - b->drawable = d; - - b->xm_visual = vis; - b->type = type; - b->cmap = cmap; - - /* determine PIPE_FORMATs for buffers */ - colorFormat = choose_pixel_format(vis); - - if (vis->mesa_visual.depthBits == 0) - depthFormat = PIPE_FORMAT_NONE; - else if (vis->mesa_visual.depthBits <= 16) - depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (vis->mesa_visual.depthBits <= 24) - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - else - depthFormat = PIPE_FORMAT_Z32_UNORM; - - if (vis->mesa_visual.stencilBits == 8) { - if (depthFormat == PIPE_FORMAT_S8Z24_UNORM) - stencilFormat = depthFormat; - else - stencilFormat = PIPE_FORMAT_S8_UNORM; - } - else { - stencilFormat = PIPE_FORMAT_NONE; - } - - - get_drawable_size(vis->display, d, &width, &height); - - /* - * Create framebuffer, but we'll plug in our own renderbuffers below. - */ - b->stfb = st_create_framebuffer(&vis->mesa_visual, - colorFormat, depthFormat, stencilFormat, - width, height, - (void *) b); - fb = &b->stfb->Base; - - /* - * Create scratch XImage for xmesa_display_surface() - */ - b->tempImage = XCreateImage(vis->display, - vis->visinfo->visual, - vis->visinfo->depth, - ZPixmap, 0, /* format, offset */ - NULL, /* data */ - 0, 0, /* size */ - 32, /* bitmap_pad */ - 0); /* bytes_per_line */ - - /* GLX_EXT_texture_from_pixmap */ - b->TextureTarget = 0; - b->TextureFormat = GLX_TEXTURE_FORMAT_NONE_EXT; - b->TextureMipmap = 0; - - /* insert buffer into linked list */ - b->Next = XMesaBufferList; - XMesaBufferList = b; - - return b; -} - - -/** - * Find an XMesaBuffer by matching X display and colormap but NOT matching - * the notThis buffer. - */ -XMesaBuffer -xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis) -{ - XMesaBuffer b; - for (b = XMesaBufferList; b; b = b->Next) { - if (b->xm_visual->display == dpy && - b->cmap == cmap && - b != notThis) { - return b; - } - } - return NULL; -} - - -/** - * Remove buffer from linked list, delete if no longer referenced. - */ -static void -xmesa_free_buffer(XMesaBuffer buffer) -{ - XMesaBuffer prev = NULL, b; - - for (b = XMesaBufferList; b; b = b->Next) { - if (b == buffer) { - struct gl_framebuffer *fb = &buffer->stfb->Base; - - /* unlink buffer from list */ - if (prev) - prev->Next = buffer->Next; - else - XMesaBufferList = buffer->Next; - - /* mark as delete pending */ - fb->DeletePending = GL_TRUE; - - /* Since the X window for the XMesaBuffer is going away, we don't - * want to dereference this pointer in the future. - */ - b->drawable = 0; - - buffer->tempImage->data = NULL; - XDestroyImage(buffer->tempImage); - - /* Unreference. If count = zero we'll really delete the buffer */ - _mesa_unreference_framebuffer(&fb); - - XFreeGC(b->xm_visual->display, b->gc); - - free(buffer); - - return; - } - /* continue search */ - prev = b; - } - /* buffer not found in XMesaBufferList */ - _mesa_problem(NULL,"xmesa_free_buffer() - buffer not found\n"); -} - - - -/**********************************************************************/ -/***** Misc Private Functions *****/ -/**********************************************************************/ - - -/** - * When a context is bound for the first time, we can finally finish - * initializing the context's visual and buffer information. - * \param v the XMesaVisual to initialize - * \param b the XMesaBuffer to initialize (may be NULL) - * \param rgb_flag TRUE = RGBA mode, FALSE = color index mode - * \param window the window/pixmap we're rendering into - * \param cmap the colormap associated with the window/pixmap - * \return GL_TRUE=success, GL_FALSE=failure - */ -static GLboolean -initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, - GLboolean rgb_flag, XMesaDrawable window, - XMesaColormap cmap) -{ -#ifdef XFree86Server - int client = (window) ? CLIENT_ID(window->id) : 0; -#endif - - ASSERT(!b || b->xm_visual == v); - - /* Save true bits/pixel */ - v->BitsPerPixel = bits_per_pixel(v); - assert(v->BitsPerPixel > 0); - - if (rgb_flag == GL_FALSE) { - /* COLOR-INDEXED WINDOW: not supported*/ - return GL_FALSE; - } - else { - /* RGB WINDOW: - * We support RGB rendering into almost any kind of visual. - */ - const int xclass = v->mesa_visual.visualType; - if (xclass != GLX_TRUE_COLOR && xclass == !GLX_DIRECT_COLOR) { - _mesa_warning(NULL, - "XMesa: RGB mode rendering not supported in given visual.\n"); - return GL_FALSE; - } - v->mesa_visual.indexBits = 0; - } - - /* - * If MESA_INFO env var is set print out some debugging info - * which can help Brian figure out what's going on when a user - * reports bugs. - */ - if (_mesa_getenv("MESA_INFO")) { - _mesa_printf("X/Mesa visual = %p\n", (void *) v); - _mesa_printf("X/Mesa level = %d\n", v->mesa_visual.level); - _mesa_printf("X/Mesa depth = %d\n", GET_VISUAL_DEPTH(v)); - _mesa_printf("X/Mesa bits per pixel = %d\n", v->BitsPerPixel); - } - - if (b && window) { - /* these should have been set in create_xmesa_buffer */ - ASSERT(b->drawable == window); - - /* Setup for single/double buffering */ - if (v->mesa_visual.doubleBufferMode) { - /* Double buffered */ - b->shm = check_for_xshm( v->display ); - } - - /* X11 graphics context */ -#ifdef XFree86Server - b->gc = CreateScratchGC(v->display, window->depth); -#else - b->gc = XCreateGC( v->display, window, 0, NULL ); -#endif - XMesaSetFunction( v->display, b->gc, GXcopy ); - } - - return GL_TRUE; -} - - - -#define NUM_VISUAL_TYPES 6 - -/** - * Convert an X visual type to a GLX visual type. - * - * \param visualType X visual type (i.e., \c TrueColor, \c StaticGray, etc.) - * to be converted. - * \return If \c visualType is a valid X visual type, a GLX visual type will - * be returned. Otherwise \c GLX_NONE will be returned. - * - * \note - * This code was lifted directly from lib/GL/glx/glcontextmodes.c in the - * DRI CVS tree. - */ -static GLint -xmesa_convert_from_x_visual_type( int visualType ) -{ - static const int glx_visual_types[ NUM_VISUAL_TYPES ] = { - GLX_STATIC_GRAY, GLX_GRAY_SCALE, - GLX_STATIC_COLOR, GLX_PSEUDO_COLOR, - GLX_TRUE_COLOR, GLX_DIRECT_COLOR - }; - - return ( (unsigned) visualType < NUM_VISUAL_TYPES ) - ? glx_visual_types[ visualType ] : GLX_NONE; -} - - -/**********************************************************************/ -/***** Public Functions *****/ -/**********************************************************************/ - - -/* - * Create a new X/Mesa visual. - * Input: display - X11 display - * visinfo - an XVisualInfo pointer - * rgb_flag - GL_TRUE = RGB mode, - * GL_FALSE = color index mode - * alpha_flag - alpha buffer requested? - * db_flag - GL_TRUE = double-buffered, - * GL_FALSE = single buffered - * stereo_flag - stereo visual? - * ximage_flag - GL_TRUE = use an XImage for back buffer, - * GL_FALSE = use an off-screen pixmap for back buffer - * depth_size - requested bits/depth values, or zero - * stencil_size - requested bits/stencil values, or zero - * accum_red_size - requested bits/red accum values, or zero - * accum_green_size - requested bits/green accum values, or zero - * accum_blue_size - requested bits/blue accum values, or zero - * accum_alpha_size - requested bits/alpha accum values, or zero - * num_samples - number of samples/pixel if multisampling, or zero - * level - visual level, usually 0 - * visualCaveat - ala the GLX extension, usually GLX_NONE - * Return; a new XMesaVisual or 0 if error. - */ -PUBLIC -XMesaVisual XMesaCreateVisual( XMesaDisplay *display, - XMesaVisualInfo visinfo, - GLboolean rgb_flag, - GLboolean alpha_flag, - GLboolean db_flag, - GLboolean stereo_flag, - GLboolean ximage_flag, - GLint depth_size, - GLint stencil_size, - GLint accum_red_size, - GLint accum_green_size, - GLint accum_blue_size, - GLint accum_alpha_size, - GLint num_samples, - GLint level, - GLint visualCaveat ) -{ - XMesaVisual v; - GLint red_bits, green_bits, blue_bits, alpha_bits; - -#ifndef XFree86Server - /* For debugging only */ - if (_mesa_getenv("MESA_XSYNC")) { - /* This makes debugging X easier. - * In your debugger, set a breakpoint on _XError to stop when an - * X protocol error is generated. - */ - XSynchronize( display, 1 ); - } -#endif - - v = (XMesaVisual) CALLOC_STRUCT(xmesa_visual); - if (!v) { - return NULL; - } - - v->display = display; - - /* Save a copy of the XVisualInfo struct because the user may X_mesa_free() - * the struct but we may need some of the information contained in it - * at a later time. - */ -#ifndef XFree86Server - v->visinfo = (XVisualInfo *) MALLOC(sizeof(*visinfo)); - if(!v->visinfo) { - _mesa_free(v); - return NULL; - } - MEMCPY(v->visinfo, visinfo, sizeof(*visinfo)); -#endif - - v->ximage_flag = ximage_flag; - -#ifdef XFree86Server - /* We could calculate these values by ourselves. nplanes is either the sum - * of the red, green, and blue bits or the number index bits. - * ColormapEntries is either (1U << index_bits) or - * (1U << max(redBits, greenBits, blueBits)). - */ - assert(visinfo->nplanes > 0); - v->nplanes = visinfo->nplanes; - v->ColormapEntries = visinfo->ColormapEntries; - - v->mesa_visual.redMask = visinfo->redMask; - v->mesa_visual.greenMask = visinfo->greenMask; - v->mesa_visual.blueMask = visinfo->blueMask; - v->mesa_visual.visualID = visinfo->vid; - v->mesa_visual.screen = 0; /* FIXME: What should be done here? */ -#else - v->mesa_visual.redMask = visinfo->red_mask; - v->mesa_visual.greenMask = visinfo->green_mask; - v->mesa_visual.blueMask = visinfo->blue_mask; - v->mesa_visual.visualID = visinfo->visualid; - v->mesa_visual.screen = visinfo->screen; -#endif - -#if defined(XFree86Server) || !(defined(__cplusplus) || defined(c_plusplus)) - v->mesa_visual.visualType = xmesa_convert_from_x_visual_type(visinfo->class); -#else - v->mesa_visual.visualType = xmesa_convert_from_x_visual_type(visinfo->c_class); -#endif - - v->mesa_visual.visualRating = visualCaveat; - - if (alpha_flag) - v->mesa_visual.alphaBits = 8; - - (void) initialize_visual_and_buffer( v, NULL, rgb_flag, 0, 0 ); - - { - const int xclass = v->mesa_visual.visualType; - if (xclass == GLX_TRUE_COLOR || xclass == GLX_DIRECT_COLOR) { - red_bits = _mesa_bitcount(GET_REDMASK(v)); - green_bits = _mesa_bitcount(GET_GREENMASK(v)); - blue_bits = _mesa_bitcount(GET_BLUEMASK(v)); - } - else { - /* this is an approximation */ - int depth; - depth = GET_VISUAL_DEPTH(v); - red_bits = depth / 3; - depth -= red_bits; - green_bits = depth / 2; - depth -= green_bits; - blue_bits = depth; - alpha_bits = 0; - assert( red_bits + green_bits + blue_bits == GET_VISUAL_DEPTH(v) ); - } - alpha_bits = v->mesa_visual.alphaBits; - } - - _mesa_initialize_visual( &v->mesa_visual, - rgb_flag, db_flag, stereo_flag, - red_bits, green_bits, - blue_bits, alpha_bits, - v->mesa_visual.indexBits, - depth_size, - stencil_size, - accum_red_size, accum_green_size, - accum_blue_size, accum_alpha_size, - 0 ); - - /* XXX minor hack */ - v->mesa_visual.level = level; - return v; -} - - -PUBLIC -void XMesaDestroyVisual( XMesaVisual v ) -{ -#ifndef XFree86Server - _mesa_free(v->visinfo); -#endif - _mesa_free(v); -} - - - -/** - * Create a new XMesaContext. - * \param v the XMesaVisual - * \param share_list another XMesaContext with which to share display - * lists or NULL if no sharing is wanted. - * \return an XMesaContext or NULL if error. - */ -PUBLIC -XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) -{ - static GLboolean firstTime = GL_TRUE; - struct pipe_context *pipe; - XMesaContext c; - GLcontext *mesaCtx; - uint pf; - - if (firstTime) { - _glthread_INIT_MUTEX(_xmesa_lock); - firstTime = GL_FALSE; - } - - /* Note: the XMesaContext contains a Mesa GLcontext struct (inheritance) */ - c = (XMesaContext) CALLOC_STRUCT(xmesa_context); - if (!c) - return NULL; - - pf = choose_pixel_format(v); - assert(pf); - - if (!getenv("XM_AUB")) { - xmesa_mode = XMESA_SOFTPIPE; - pipe = xmesa_create_pipe_context( c, pf ); - } - else { - xmesa_mode = XMESA_AUB; - pipe = xmesa_create_i965simple( xmesa_get_pipe_winsys_aub() ); - } - - c->st = st_create_context(pipe, &v->mesa_visual, - share_list ? share_list->st : NULL); - mesaCtx = c->st->ctx; - c->st->ctx->DriverCtx = c; - -#if 00 - _mesa_enable_sw_extensions(mesaCtx); - _mesa_enable_1_3_extensions(mesaCtx); - _mesa_enable_1_4_extensions(mesaCtx); - _mesa_enable_1_5_extensions(mesaCtx); - _mesa_enable_2_0_extensions(mesaCtx); -#endif - -#ifdef XFree86Server - /* If we're running in the X server, do bounds checking to prevent - * segfaults and server crashes! - */ - mesaCtx->Const.CheckArrayBounds = GL_TRUE; -#endif - - /* finish up xmesa context initializations */ - c->xm_visual = v; - c->xm_buffer = NULL; /* set later by XMesaMakeCurrent */ - - c->st->haveFramebufferSurfaces = GL_TRUE; - - return c; -} - - - -PUBLIC -void XMesaDestroyContext( XMesaContext c ) -{ - st_destroy_context(c->st); - _mesa_free(c); -} - - - -/** - * Private function for creating an XMesaBuffer which corresponds to an - * X window or pixmap. - * \param v the window's XMesaVisual - * \param w the window we're wrapping - * \return new XMesaBuffer or NULL if error - */ -PUBLIC XMesaBuffer -XMesaCreateWindowBuffer(XMesaVisual v, XMesaWindow w) -{ -#ifndef XFree86Server - XWindowAttributes attr; -#endif - XMesaBuffer b; - XMesaColormap cmap; - int depth; - - assert(v); - assert(w); - - /* Check that window depth matches visual depth */ -#ifdef XFree86Server - depth = ((XMesaDrawable)w)->depth; -#else - XGetWindowAttributes( v->display, w, &attr ); - depth = attr.depth; -#endif - if (GET_VISUAL_DEPTH(v) != depth) { - _mesa_warning(NULL, "XMesaCreateWindowBuffer: depth mismatch between visual (%d) and window (%d)!\n", - GET_VISUAL_DEPTH(v), depth); - return NULL; - } - - /* Find colormap */ -#ifdef XFree86Server - cmap = (ColormapPtr)LookupIDByType(wColormap(w), RT_COLORMAP); -#else - if (attr.colormap) { - cmap = attr.colormap; - } - else { - _mesa_warning(NULL, "Window %u has no colormap!\n", (unsigned int) w); - /* this is weird, a window w/out a colormap!? */ - /* OK, let's just allocate a new one and hope for the best */ - cmap = XCreateColormap(v->display, w, attr.visual, AllocNone); - } -#endif - - b = create_xmesa_buffer((XMesaDrawable) w, WINDOW, v, cmap); - if (!b) - return NULL; - - if (!initialize_visual_and_buffer( v, b, v->mesa_visual.rgbMode, - (XMesaDrawable) w, cmap )) { - xmesa_free_buffer(b); - return NULL; - } - - return b; -} - - - -/** - * Create a new XMesaBuffer from an X pixmap. - * - * \param v the XMesaVisual - * \param p the pixmap - * \param cmap the colormap, may be 0 if using a \c GLX_TRUE_COLOR or - * \c GLX_DIRECT_COLOR visual for the pixmap - * \returns new XMesaBuffer or NULL if error - */ -PUBLIC XMesaBuffer -XMesaCreatePixmapBuffer(XMesaVisual v, XMesaPixmap p, XMesaColormap cmap) -{ - XMesaBuffer b; - - assert(v); - - b = create_xmesa_buffer((XMesaDrawable) p, PIXMAP, v, cmap); - if (!b) - return NULL; - - if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode, - (XMesaDrawable) p, cmap)) { - xmesa_free_buffer(b); - return NULL; - } - - return b; -} - - -/** - * For GLX_EXT_texture_from_pixmap - */ -XMesaBuffer -XMesaCreatePixmapTextureBuffer(XMesaVisual v, XMesaPixmap p, - XMesaColormap cmap, - int format, int target, int mipmap) -{ - GET_CURRENT_CONTEXT(ctx); - XMesaBuffer b; - GLuint width, height; - - assert(v); - - b = create_xmesa_buffer((XMesaDrawable) p, PIXMAP, v, cmap); - if (!b) - return NULL; - - /* get pixmap size, update framebuffer/renderbuffer dims */ - xmesa_get_window_size(v->display, b, &width, &height); - _mesa_resize_framebuffer(NULL, &(b->stfb->Base), width, height); - - if (target == 0) { - /* examine dims */ - if (ctx->Extensions.ARB_texture_non_power_of_two) { - target = GLX_TEXTURE_2D_EXT; - } - else if ( _mesa_bitcount(width) == 1 - && _mesa_bitcount(height) == 1) { - /* power of two size */ - if (height == 1) { - target = GLX_TEXTURE_1D_EXT; - } - else { - target = GLX_TEXTURE_2D_EXT; - } - } - else if (ctx->Extensions.NV_texture_rectangle) { - target = GLX_TEXTURE_RECTANGLE_EXT; - } - else { - /* non power of two textures not supported */ - XMesaDestroyBuffer(b); - return 0; - } - } - - b->TextureTarget = target; - b->TextureFormat = format; - b->TextureMipmap = mipmap; - - if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode, - (XMesaDrawable) p, cmap)) { - xmesa_free_buffer(b); - return NULL; - } - - return b; -} - - - -XMesaBuffer -XMesaCreatePBuffer(XMesaVisual v, XMesaColormap cmap, - unsigned int width, unsigned int height) -{ -#ifndef XFree86Server - XMesaWindow root; - XMesaDrawable drawable; /* X Pixmap Drawable */ - XMesaBuffer b; - - /* allocate pixmap for front buffer */ - root = RootWindow( v->display, v->visinfo->screen ); - drawable = XCreatePixmap(v->display, root, width, height, - v->visinfo->depth); - if (!drawable) - return NULL; - - b = create_xmesa_buffer(drawable, PBUFFER, v, cmap); - if (!b) - return NULL; - - if (!initialize_visual_and_buffer(v, b, v->mesa_visual.rgbMode, - drawable, cmap)) { - xmesa_free_buffer(b); - return NULL; - } - - return b; -#else - return 0; -#endif -} - - - -/* - * Deallocate an XMesaBuffer structure and all related info. - */ -PUBLIC void -XMesaDestroyBuffer(XMesaBuffer b) -{ - xmesa_free_buffer(b); -} - - -/** - * Query the current window size and update the corresponding GLframebuffer - * and all attached renderbuffers. - * Called when: - * 1. the first time a buffer is bound to a context. - * 2. from the XMesaResizeBuffers() API function. - * 3. SwapBuffers. XXX probabaly from xm_flush_frontbuffer() too... - * Note: it's possible (and legal) for xmctx to be NULL. That can happen - * when resizing a buffer when no rendering context is bound. - */ -void -xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer) -{ - GLuint width, height; - xmesa_get_window_size(drawBuffer->xm_visual->display, drawBuffer, &width, &height); - st_resize_framebuffer(drawBuffer->stfb, width, height); -} - - -/* - * Bind buffer b to context c and make c the current rendering context. - */ -GLboolean XMesaMakeCurrent( XMesaContext c, XMesaBuffer b ) -{ - return XMesaMakeCurrent2( c, b, b ); -} - - -/* - * Bind buffer b to context c and make c the current rendering context. - */ -PUBLIC -GLboolean XMesaMakeCurrent2( XMesaContext c, XMesaBuffer drawBuffer, - XMesaBuffer readBuffer ) -{ - if (c) { - if (!drawBuffer || !readBuffer) - return GL_FALSE; /* must specify buffers! */ - -#if 0 - /* XXX restore this optimization */ - if (&(c->mesa) == _mesa_get_current_context() - && c->mesa.DrawBuffer == &drawBuffer->mesa_buffer - && c->mesa.ReadBuffer == &readBuffer->mesa_buffer - && xmesa_buffer(c->mesa.DrawBuffer)->wasCurrent) { - /* same context and buffer, do nothing */ - return GL_TRUE; - } -#endif - - c->xm_buffer = drawBuffer; - - /* Call this periodically to detect when the user has begun using - * GL rendering from multiple threads. - */ - _glapi_check_multithread(); - - st_make_current(c->st, drawBuffer->stfb, readBuffer->stfb); - - xmesa_check_and_update_buffer_size(c, drawBuffer); - if (readBuffer != drawBuffer) - xmesa_check_and_update_buffer_size(c, readBuffer); - - /* Solution to Stephane Rehel's problem with glXReleaseBuffersMESA(): */ - drawBuffer->wasCurrent = GL_TRUE; - } - else { - /* Detach */ - st_make_current( NULL, NULL, NULL ); - } - return GL_TRUE; -} - - -/* - * Unbind the context c from its buffer. - */ -GLboolean XMesaUnbindContext( XMesaContext c ) -{ - /* A no-op for XFree86 integration purposes */ - return GL_TRUE; -} - - -XMesaContext XMesaGetCurrentContext( void ) -{ - GET_CURRENT_CONTEXT(ctx); - if (ctx) { - XMesaContext xmesa = xmesa_context(ctx); - return xmesa; - } - else { - return 0; - } -} - - -XMesaBuffer XMesaGetCurrentBuffer( void ) -{ - GET_CURRENT_CONTEXT(ctx); - if (ctx) { - XMesaBuffer xmbuf = xmesa_buffer(ctx->DrawBuffer); - return xmbuf; - } - else { - return 0; - } -} - - -/* New in Mesa 3.1 */ -XMesaBuffer XMesaGetCurrentReadBuffer( void ) -{ - GET_CURRENT_CONTEXT(ctx); - if (ctx) { - return xmesa_buffer(ctx->ReadBuffer); - } - else { - return 0; - } -} - - -#ifdef XFree86Server -PUBLIC -GLboolean XMesaForceCurrent(XMesaContext c) -{ - if (c) { - _glapi_set_dispatch(c->mesa.CurrentDispatch); - - if (&(c->mesa) != _mesa_get_current_context()) { - _mesa_make_current(&c->mesa, c->mesa.DrawBuffer, c->mesa.ReadBuffer); - } - } - else { - _mesa_make_current(NULL, NULL, NULL); - } - return GL_TRUE; -} - - -PUBLIC -GLboolean XMesaLoseCurrent(XMesaContext c) -{ - (void) c; - _mesa_make_current(NULL, NULL, NULL); - return GL_TRUE; -} - - -PUBLIC -GLboolean XMesaCopyContext( XMesaContext xm_src, XMesaContext xm_dst, GLuint mask ) -{ - _mesa_copy_context(&xm_src->mesa, &xm_dst->mesa, mask); - return GL_TRUE; -} -#endif /* XFree86Server */ - - -#ifndef FX -GLboolean XMesaSetFXmode( GLint mode ) -{ - (void) mode; - return GL_FALSE; -} -#endif - - - -/* - * Copy the back buffer to the front buffer. If there's no back buffer - * this is a no-op. - */ -PUBLIC -void XMesaSwapBuffers( XMesaBuffer b ) -{ - struct pipe_surface *surf; - - /* If we're swapping the buffer associated with the current context - * we have to flush any pending rendering commands first. - */ - st_notify_swapbuffers(b->stfb); - - surf = st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT); - if (surf) { - if (xmesa_mode == XMESA_AUB) - xmesa_display_aub( surf ); - else - xmesa_display_surface(b, surf); - } - - xmesa_check_and_update_buffer_size(NULL, b); -} - - - -/* - * Copy sub-region of back buffer to front buffer - */ -void XMesaCopySubBuffer( XMesaBuffer b, int x, int y, int width, int height ) -{ - struct pipe_surface *surf_front - = st_get_framebuffer_surface(b->stfb, ST_SURFACE_FRONT_LEFT); - struct pipe_surface *surf_back - = st_get_framebuffer_surface(b->stfb, ST_SURFACE_BACK_LEFT); - struct pipe_context *pipe = NULL; /* XXX fix */ - - if (!surf_front || !surf_back) - return; - - pipe->surface_copy(pipe, - FALSE, - surf_front, x, y, /* dest */ - surf_back, x, y, /* src */ - width, height); -} - - - -/* - * Return the depth buffer associated with an XMesaBuffer. - * Input: b - the XMesa buffer handle - * Output: width, height - size of buffer in pixels - * bytesPerValue - bytes per depth value (2 or 4) - * buffer - pointer to depth buffer values - * Return: GL_TRUE or GL_FALSE to indicate success or failure. - */ -GLboolean XMesaGetDepthBuffer( XMesaBuffer b, GLint *width, GLint *height, - GLint *bytesPerValue, void **buffer ) -{ - *width = 0; - *height = 0; - *bytesPerValue = 0; - *buffer = 0; - return GL_FALSE; -} - - -void XMesaFlush( XMesaContext c ) -{ - if (c && c->xm_visual->display) { -#ifdef XFree86Server - /* NOT_NEEDED */ -#else - XSync( c->xm_visual->display, False ); -#endif - } -} - - - -const char *XMesaGetString( XMesaContext c, int name ) -{ - (void) c; - if (name==XMESA_VERSION) { - return "5.0"; - } - else if (name==XMESA_EXTENSIONS) { - return ""; - } - else { - return NULL; - } -} - - - -XMesaBuffer XMesaFindBuffer( XMesaDisplay *dpy, XMesaDrawable d ) -{ - XMesaBuffer b; - for (b=XMesaBufferList; b; b=b->Next) { - if (b->drawable == d && b->xm_visual->display == dpy) { - return b; - } - } - return NULL; -} - - -/** - * Free/destroy all XMesaBuffers associated with given display. - */ -void xmesa_destroy_buffers_on_display(XMesaDisplay *dpy) -{ - XMesaBuffer b, next; - for (b = XMesaBufferList; b; b = next) { - next = b->Next; - if (b->xm_visual->display == dpy) { - xmesa_free_buffer(b); - } - } -} - - -/* - * Look for XMesaBuffers whose X window has been destroyed. - * Deallocate any such XMesaBuffers. - */ -void XMesaGarbageCollect( void ) -{ - XMesaBuffer b, next; - for (b=XMesaBufferList; b; b=next) { - next = b->Next; - if (b->xm_visual && - b->xm_visual->display && - b->drawable && - b->type == WINDOW) { -#ifdef XFree86Server - /* NOT_NEEDED */ -#else - XSync(b->xm_visual->display, False); - if (!window_exists( b->xm_visual->display, b->drawable )) { - /* found a dead window, free the ancillary info */ - XMesaDestroyBuffer( b ); - } -#endif - } - } -} - - -unsigned long XMesaDitherColor( XMesaContext xmesa, GLint x, GLint y, - GLfloat red, GLfloat green, - GLfloat blue, GLfloat alpha ) -{ - /* no longer supported */ - return 0; -} - - -/* - * This is typically called when the window size changes and we need - * to reallocate the buffer's back/depth/stencil/accum buffers. - */ -PUBLIC void -XMesaResizeBuffers( XMesaBuffer b ) -{ - GET_CURRENT_CONTEXT(ctx); - XMesaContext xmctx = xmesa_context(ctx); - if (!xmctx) - return; - xmesa_check_and_update_buffer_size(xmctx, b); -} - - - - -PUBLIC void -XMesaBindTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer, - const int *attrib_list) -{ -} - - - -PUBLIC void -XMesaReleaseTexImage(XMesaDisplay *dpy, XMesaBuffer drawable, int buffer) -{ -} - diff --git a/src/mesa/pipe/xlib/xm_image.c b/src/mesa/pipe/xlib/xm_image.c deleted file mode 100644 index 087b4e4c3a..0000000000 --- a/src/mesa/pipe/xlib/xm_image.c +++ /dev/null @@ -1,133 +0,0 @@ -/************************************************************************** - -Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. -All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining a -copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sub license, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial portions -of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. -IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR -ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, -TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE -SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Kevin E. Martin - * Brian Paul - */ - -#include -#include - -#include "glxheader.h" -#include "xmesaP.h" - -#ifdef XFree86Server - -#ifdef ROUNDUP -#undef ROUNDUP -#endif - -#define ROUNDUP(nbytes, pad) ((((nbytes) + ((pad)-1)) / (pad)) * ((pad)>>3)) - -XMesaImage *XMesaCreateImage(int bitsPerPixel, int width, int height, char *data) -{ - XMesaImage *image; - - image = (XMesaImage *)xalloc(sizeof(XMesaImage)); - - if (image) { - image->width = width; - image->height = height; - image->data = data; - /* Always pad to 32 bits */ - image->bytes_per_line = ROUNDUP((bitsPerPixel * width), 32); - image->bits_per_pixel = bitsPerPixel; - } - - return image; -} - -void XMesaDestroyImage(XMesaImage *image) -{ - if (image->data) - free(image->data); - xfree(image); -} - -unsigned long XMesaGetPixel(XMesaImage *image, int x, int y) -{ - CARD8 *row = (CARD8 *)(image->data + y*image->bytes_per_line); - CARD8 *i8; - CARD16 *i16; - CARD32 *i32; - switch (image->bits_per_pixel) { - case 8: - i8 = (CARD8 *)row; - return i8[x]; - break; - case 15: - case 16: - i16 = (CARD16 *)row; - return i16[x]; - break; - case 24: /* WARNING: architecture specific code */ - i8 = (CARD8 *)row; - return (((CARD32)i8[x*3]) | - (((CARD32)i8[x*3+1])<<8) | - (((CARD32)i8[x*3+2])<<16)); - break; - case 32: - i32 = (CARD32 *)row; - return i32[x]; - break; - } - return 0; -} - -#ifndef XMESA_USE_PUTPIXEL_MACRO -void XMesaPutPixel(XMesaImage *image, int x, int y, unsigned long pixel) -{ - CARD8 *row = (CARD8 *)(image->data + y*image->bytes_per_line); - CARD8 *i8; - CARD16 *i16; - CARD32 *i32; - switch (image->bits_per_pixel) { - case 8: - i8 = (CARD8 *)row; - i8[x] = (CARD8)pixel; - break; - case 15: - case 16: - i16 = (CARD16 *)row; - i16[x] = (CARD16)pixel; - break; - case 24: /* WARNING: architecture specific code */ - i8 = (CARD8 *)__row; - i8[x*3] = (CARD8)(p); - i8[x*3+1] = (CARD8)(p>>8); - i8[x*3+2] = (CARD8)(p>>16); - case 32: - i32 = (CARD32 *)row; - i32[x] = (CARD32)pixel; - break; - } -} -#endif - -#endif /* XFree86Server */ diff --git a/src/mesa/pipe/xlib/xm_image.h b/src/mesa/pipe/xlib/xm_image.h deleted file mode 100644 index 2a5e0f3777..0000000000 --- a/src/mesa/pipe/xlib/xm_image.h +++ /dev/null @@ -1,77 +0,0 @@ -/************************************************************************** - -Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. -All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining a -copy of this software and associated documentation files (the -"Software"), to deal in the Software without restriction, including -without limitation the rights to use, copy, modify, merge, publish, -distribute, sub license, and/or sell copies of the Software, and to -permit persons to whom the Software is furnished to do so, subject to -the following conditions: - -The above copyright notice and this permission notice (including the -next paragraph) shall be included in all copies or substantial portions -of the Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. -IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR -ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, -TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE -SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - -/* - * Authors: - * Kevin E. Martin - * Brian Paul - */ - -#ifndef _XM_IMAGE_H_ -#define _XM_IMAGE_H_ - -#define XMESA_USE_PUTPIXEL_MACRO - -extern XMesaImage *XMesaCreateImage(int bitsPerPixel, int width, int height, - char *data); -extern void XMesaDestroyImage(XMesaImage *image); -extern unsigned long XMesaGetPixel(XMesaImage *image, int x, int y); -#ifdef XMESA_USE_PUTPIXEL_MACRO -#define XMesaPutPixel(__i,__x,__y,__p) \ -{ \ - CARD8 *__row = (CARD8 *)(__i->data + __y*__i->bytes_per_line); \ - CARD8 *__i8; \ - CARD16 *__i16; \ - CARD32 *__i32; \ - switch (__i->bits_per_pixel) { \ - case 8: \ - __i8 = (CARD8 *)__row; \ - __i8[__x] = (CARD8)__p; \ - break; \ - case 15: \ - case 16: \ - __i16 = (CARD16 *)__row; \ - __i16[__x] = (CARD16)__p; \ - break; \ - case 24: /* WARNING: architecture specific code */ \ - __i8 = (CARD8 *)__row; \ - __i8[__x*3] = (CARD8)(__p); \ - __i8[__x*3+1] = (CARD8)(__p>>8); \ - __i8[__x*3+2] = (CARD8)(__p>>16); \ - break; \ - case 32: \ - __i32 = (CARD32 *)__row; \ - __i32[__x] = (CARD32)__p; \ - break; \ - } \ -} -#else -extern void XMesaPutPixel(XMesaImage *image, int x, int y, - unsigned long pixel); -#endif - -#endif /* _XM_IMAGE_H_ */ diff --git a/src/mesa/pipe/xlib/xm_winsys.c b/src/mesa/pipe/xlib/xm_winsys.c deleted file mode 100644 index c3cd22eea3..0000000000 --- a/src/mesa/pipe/xlib/xm_winsys.c +++ /dev/null @@ -1,466 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ - -/* - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#include "glxheader.h" -#include "xmesaP.h" - -#include "pipe/p_winsys.h" -#include "pipe/p_format.h" -#include "pipe/p_context.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/softpipe/sp_winsys.h" - -#ifdef GALLIUM_CELL -#include "pipe/cell/ppu/cell_context.h" -#include "pipe/cell/ppu/cell_winsys.h" -#else -#define TILE_SIZE 32 /* avoid compilation errors */ -#endif - -#include "xm_winsys_aub.h" - - -/** - * Low-level OS/window system memory buffer - */ -struct xm_buffer -{ - struct pipe_buffer base; - boolean userBuffer; /** Is this a user-space buffer? */ - void *data; - void *mapped; -}; - - -struct xmesa_surface -{ - struct pipe_surface surface; - - int tileSize; -}; - - -/** - * Derived from softpipe_winsys. - * We just need one extra field which indicates the pixel format to use for - * drawing surfaces so that we're compatible with the XVisual/window format. - */ -struct xmesa_softpipe_winsys -{ - struct softpipe_winsys spws; - enum pipe_format pixelformat; -}; - - - -/** Cast wrapper */ -static INLINE struct xmesa_surface * -xmesa_surface(struct pipe_surface *ps) -{ -// assert(0); - return (struct xmesa_surface *) ps; -} - -/** cast wrapper */ -static INLINE struct xmesa_softpipe_winsys * -xmesa_softpipe_winsys(struct softpipe_winsys *spws) -{ - return (struct xmesa_softpipe_winsys *) spws; -} - -/** - * Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque - * buffer pointer... - */ -static INLINE struct xm_buffer * -xm_buffer( struct pipe_buffer *buf ) -{ - return (struct xm_buffer *)buf; -} - - - -/* Most callbacks map direcly onto dri_bufmgr operations: - */ -static void * -xm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buf, - unsigned flags) -{ - struct xm_buffer *xm_buf = xm_buffer(buf); - xm_buf->mapped = xm_buf->data; - return xm_buf->mapped; -} - -static void -xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) -{ - struct xm_buffer *xm_buf = xm_buffer(buf); - xm_buf->mapped = NULL; -} - -static void -xm_buffer_destroy(struct pipe_winsys *pws, - struct pipe_buffer *buf) -{ - struct xm_buffer *oldBuf = xm_buffer(buf); - - if (oldBuf->data) { - if (!oldBuf->userBuffer) - align_free(oldBuf->data); - oldBuf->data = NULL; - } - - free(oldBuf); -} - - -/** - * Display a surface that's in a tiled configuration. That is, all the - * pixels for a TILE_SIZExTILE_SIZE block are contiguous in memory. - */ -static void -xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) -{ - XImage *ximage = b->tempImage; - struct xm_buffer *xm_buf = xm_buffer(surf->buffer); - const uint tilesPerRow = (surf->width + TILE_SIZE - 1) / TILE_SIZE; - uint x, y; - - /* check that the XImage has been previously initialized */ - assert(ximage->format); - assert(ximage->bitmap_unit); - - /* update XImage's fields */ - ximage->width = TILE_SIZE; - ximage->height = TILE_SIZE; - ximage->bytes_per_line = TILE_SIZE * 4; - - for (y = 0; y < surf->height; y += TILE_SIZE) { - for (x = 0; x < surf->width; x += TILE_SIZE) { - int dx = x; - int dy = y; - int tx = x / TILE_SIZE; - int ty = y / TILE_SIZE; - int offset = ty * tilesPerRow + tx; - - offset *= 4 * TILE_SIZE * TILE_SIZE; - - ximage->data = (char *) xm_buf->data + offset; - - XPutImage(b->xm_visual->display, b->drawable, b->gc, - ximage, 0, 0, dx, dy, TILE_SIZE, TILE_SIZE); - } - } -} - - -/** - * Display/copy the image in the surface into the X window specified - * by the XMesaBuffer. - */ -void -xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) -{ - XImage *ximage = b->tempImage; - struct xm_buffer *xm_buf = xm_buffer(surf->buffer); - const struct xmesa_surface *xm_surf - = xmesa_surface((struct pipe_surface *) surf); - - if (xm_surf->tileSize) { - xmesa_display_surface_tiled(b, surf); - return; - } - - /* check that the XImage has been previously initialized */ - assert(ximage->format); - assert(ximage->bitmap_unit); - - /* update XImage's fields */ - ximage->width = surf->width; - ximage->height = surf->height; - ximage->bytes_per_line = surf->pitch * (ximage->bits_per_pixel / 8); - ximage->data = xm_buf->data; - - /* display image in Window */ - XPutImage(b->xm_visual->display, b->drawable, b->gc, - ximage, 0, 0, 0, 0, surf->width, surf->height); -} - - -static void -xm_flush_frontbuffer(struct pipe_winsys *pws, - struct pipe_surface *surf, - void *context_private) -{ - /* The Xlib driver's front color surfaces are actually X Windows so - * this flush is a no-op. - * If we instead did front buffer rendering to a temporary XImage, - * this would be the place to copy the Ximage to the on-screen Window. - */ - XMesaContext xmctx = (XMesaContext) context_private; - xmesa_display_surface(xmctx->xm_buffer, surf); -} - - - -static void -xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) -{ - va_list args; - va_start( args, fmtString ); - vfprintf(stderr, fmtString, args); - va_end( args ); -} - - -static const char * -xm_get_name(struct pipe_winsys *pws) -{ - return "Xlib"; -} - - -static struct pipe_buffer * -xm_buffer_create(struct pipe_winsys *pws, - unsigned alignment, - unsigned usage, - unsigned size) -{ - struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); - buffer->base.refcount = 1; - buffer->base.alignment = alignment; - buffer->base.usage = usage; - buffer->base.size = size; - - /* align to 16-byte multiple for Cell */ - buffer->data = align_malloc(size, max(alignment, 16)); - - return &buffer->base; -} - - -/** - * Create buffer which wraps user-space data. - */ -static struct pipe_buffer * -xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) -{ - struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); - buffer->base.refcount = 1; - buffer->base.size = bytes; - buffer->userBuffer = TRUE; - buffer->data = ptr; - - return &buffer->base; -} - - - -/** - * Round n up to next multiple. - */ -static INLINE unsigned -round_up(unsigned n, unsigned multiple) -{ - return (n + multiple - 1) & ~(multiple - 1); -} - -static int -xm_surface_alloc_storage(struct pipe_winsys *winsys, - struct pipe_surface *surf, - unsigned width, unsigned height, - enum pipe_format format, - unsigned flags) -{ - const unsigned alignment = 64; - - surf->width = width; - surf->height = height; - surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = round_up(width, alignment / surf->cpp); - -#ifdef GALLIUM_CELL /* XXX a bit of a hack */ - height = round_up(height, TILE_SIZE); -#endif - - assert(!surf->buffer); - surf->buffer = winsys->buffer_create(winsys, alignment, - PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * height); - if(!surf->buffer) - return -1; - - return 0; -} - - -/** - * Called via pipe->surface_alloc() to create new surfaces (textures, - * renderbuffers, etc. - */ -static struct pipe_surface * -xm_surface_alloc(struct pipe_winsys *ws) -{ - struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); - - assert(ws); - - xms->surface.refcount = 1; - xms->surface.winsys = ws; - -#ifdef GALLIUM_CELL - if (!getenv("GALLIUM_NOCELL")) { - xms->tileSize = 32; /** probably temporary */ - } -#endif - - return &xms->surface; -} - - - -static void -xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) -{ - struct pipe_surface *surf = *s; - surf->refcount--; - if (surf->refcount == 0) { - if (surf->buffer) - pipe_buffer_reference(winsys, &surf->buffer, NULL); - free(surf); - } - *s = NULL; -} - - - -/** - * Return pointer to a pipe_winsys object. - * For Xlib, this is a singleton object. - * Nothing special for the Xlib driver so no subclassing or anything. - */ -struct pipe_winsys * -xmesa_get_pipe_winsys_aub(void) -{ - static struct pipe_winsys *ws = NULL; - - if (!ws && getenv("XM_AUB")) { - ws = xmesa_create_pipe_winsys_aub(); - } - else if (!ws) { - ws = CALLOC_STRUCT(pipe_winsys); - - /* Fill in this struct with callbacks that pipe will need to - * communicate with the window system, buffer manager, etc. - */ - ws->buffer_create = xm_buffer_create; - ws->user_buffer_create = xm_user_buffer_create; - ws->buffer_map = xm_buffer_map; - ws->buffer_unmap = xm_buffer_unmap; - ws->buffer_destroy = xm_buffer_destroy; - - ws->surface_alloc = xm_surface_alloc; - ws->surface_alloc_storage = xm_surface_alloc_storage; - ws->surface_release = xm_surface_release; - - ws->flush_frontbuffer = xm_flush_frontbuffer; - ws->printf = xm_printf; - ws->get_name = xm_get_name; - } - - return ws; -} - - -/** - * Called via softpipe_winsys->is_format_supported(). - * This function is only called to test formats for front/back color surfaces. - * The winsys being queried will have been created at glXCreateContext - * time, with a pixel format corresponding to the context's visual. - */ -static boolean -xmesa_is_format_supported(struct softpipe_winsys *sws, - enum pipe_format format) -{ - struct xmesa_softpipe_winsys *xmws = xmesa_softpipe_winsys(sws); - return (format == xmws->pixelformat); -} - - -/** - * Return pointer to a softpipe_winsys object. - */ -static struct softpipe_winsys * -xmesa_get_softpipe_winsys(uint pixelformat) -{ - struct xmesa_softpipe_winsys *xmws - = CALLOC_STRUCT(xmesa_softpipe_winsys); - if (!xmws) - return NULL; - - xmws->spws.is_format_supported = xmesa_is_format_supported; - xmws->pixelformat = pixelformat; - - return &xmws->spws; -} - - -struct pipe_context * -xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) -{ - struct pipe_winsys *pws = xmesa_get_pipe_winsys_aub(); - struct pipe_context *pipe; - -#ifdef GALLIUM_CELL - if (!getenv("GALLIUM_NOCELL")) { - struct cell_winsys *cws = cell_get_winsys(pixelformat); - pipe = cell_create_context(pws, cws); - if (pipe) - pipe->priv = xmesa; - return pipe; - } - else -#endif - { - struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(pixelformat); - pipe = softpipe_create( pws, spws ); - if (pipe) - pipe->priv = xmesa; - - return pipe; - } -} diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.c b/src/mesa/pipe/xlib/xm_winsys_aub.c deleted file mode 100644 index bf41570257..0000000000 --- a/src/mesa/pipe/xlib/xm_winsys_aub.c +++ /dev/null @@ -1,589 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ - -/* - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#include "glxheader.h" -#include "xmesaP.h" - -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/i965simple/brw_winsys.h" -#include "brw_aub.h" -#include "xm_winsys_aub.h" - - - -struct aub_buffer { - char *data; - unsigned offset; - unsigned size; - unsigned refcount; - unsigned map_count; - boolean dump_on_unmap; -}; - - - -struct aub_pipe_winsys { - struct pipe_winsys winsys; - - struct brw_aubfile *aubfile; - - /* This is simple, isn't it: - */ - char *pool; - unsigned size; - unsigned used; -}; - - -/* Turn a pipe winsys into an aub/pipe winsys: - */ -static inline struct aub_pipe_winsys * -aub_pipe_winsys( struct pipe_winsys *winsys ) -{ - return (struct aub_pipe_winsys *)winsys; -} - - - -static INLINE struct aub_buffer * -aub_bo( struct pipe_buffer *bo ) -{ - return (struct aub_buffer *)bo; -} - -static INLINE struct pipe_buffer * -pipe_bo( struct aub_buffer *bo ) -{ - return (struct pipe_buffer *)bo; -} - - - - -static void *aub_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned flags ) -{ - struct aub_buffer *sbo = aub_bo(buf); - - assert(sbo->data); - - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - sbo->dump_on_unmap = 1; - - sbo->map_count++; - return sbo->data; -} - -static void aub_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); - struct aub_buffer *sbo = aub_bo(buf); - - sbo->map_count--; - - if (sbo->map_count == 0 && - sbo->dump_on_unmap) { - - sbo->dump_on_unmap = 0; - - brw_aub_gtt_data( iws->aubfile, - sbo->offset, - sbo->data, - sbo->size, - 0, - 0); - } -} - - -static void -aub_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - free(buf); -} - - -void xmesa_buffer_subdata_aub(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned long offset, - unsigned long size, - const void *data, - unsigned aub_type, - unsigned aub_sub_type) -{ - struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); - struct aub_buffer *sbo = aub_bo(buf); - - assert(sbo->size > offset + size); - memcpy(sbo->data + offset, data, size); - - brw_aub_gtt_data( iws->aubfile, - sbo->offset + offset, - sbo->data + offset, - size, - aub_type, - aub_sub_type ); -} - -void xmesa_commands_aub(struct pipe_winsys *winsys, - unsigned *cmds, - unsigned nr_dwords) -{ - struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); - unsigned size = nr_dwords * 4; - - assert(iws->used + size < iws->size); - - brw_aub_gtt_cmds( iws->aubfile, - AUB_BUF_START + iws->used, - cmds, - nr_dwords * sizeof(int) ); - - iws->used += align(size, 4096); -} - - -static struct aub_pipe_winsys *global_winsys = NULL; - -void xmesa_display_aub( /* struct pipe_winsys *winsys, */ - struct pipe_surface *surface ) -{ -// struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); - brw_aub_dump_bmp( global_winsys->aubfile, - surface, - aub_bo(surface->buffer)->offset ); -} - - - -/* Pipe has no concept of pools. We choose the tex/region pool - * for all buffers. - */ -static struct pipe_buffer * -aub_buffer_create(struct pipe_winsys *winsys, - unsigned alignment, - unsigned usage, - unsigned size) -{ - struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); - struct aub_buffer *sbo = CALLOC_STRUCT(aub_buffer); - - sbo->refcount = 1; - - /* Could reuse buffers that are not referenced in current - * batchbuffer. Can't do that atm, so always reallocate: - */ - assert(iws->used + size < iws->size); - sbo->data = iws->pool + iws->used; - sbo->offset = AUB_BUF_START + iws->used; - iws->used += align(size, 4096); - - sbo->size = size; - - return pipe_bo(sbo); -} - - -static struct pipe_buffer * -aub_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) -{ - struct aub_buffer *sbo; - - /* Lets hope this is meant for upload, not as a result! - */ - sbo = aub_bo(aub_buffer_create( winsys, 0, 0, 0 )); - - sbo->data = ptr; - sbo->size = bytes; - - return pipe_bo(sbo); -} - - -/* The state tracker (should!) keep track of whether the fake - * frontbuffer has been touched by any rendering since the last time - * we copied its contents to the real frontbuffer. Our task is easy: - */ -static void -aub_flush_frontbuffer( struct pipe_winsys *winsys, - struct pipe_surface *surf, - void *context_private) -{ - xmesa_display_aub( surf ); -} - -static struct pipe_surface * -aub_i915_surface_alloc(struct pipe_winsys *winsys) -{ - struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); - if (surf) { - surf->refcount = 1; - surf->winsys = winsys; - } - return surf; -} - - -/** - * Round n up to next multiple. - */ -static INLINE unsigned -round_up(unsigned n, unsigned multiple) -{ - return (n + multiple - 1) & ~(multiple - 1); -} - -static int -aub_i915_surface_alloc_storage(struct pipe_winsys *winsys, - struct pipe_surface *surf, - unsigned width, unsigned height, - enum pipe_format format, - unsigned flags) -{ - const unsigned alignment = 64; - - surf->width = width; - surf->height = height; - surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = round_up(width, alignment / surf->cpp); - - assert(!surf->buffer); - surf->buffer = winsys->buffer_create(winsys, alignment, - PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * height); - if(!surf->buffer) - return -1; - - return 0; -} - -static void -aub_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) -{ - struct pipe_surface *surf = *s; - surf->refcount--; - if (surf->refcount == 0) { - if (surf->buffer) - pipe_buffer_reference(winsys, &surf->buffer, NULL); - free(surf); - } - *s = NULL; -} - - - -static void -aub_printf( struct pipe_winsys *winsys, const char *fmtString, ... ) -{ - va_list args; - va_start( args, fmtString ); - vfprintf(stderr, fmtString, args); - va_end( args ); -} - -static const char * -aub_get_name( struct pipe_winsys *winsys ) -{ - return "Aub/xlib"; -} - -struct pipe_winsys * -xmesa_create_pipe_winsys_aub( void ) -{ - struct aub_pipe_winsys *iws = CALLOC_STRUCT( aub_pipe_winsys ); - - /* Fill in this struct with callbacks that pipe will need to - * communicate with the window system, buffer manager, etc. - * - * Pipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate _DriBufferObject. - */ - iws->winsys.buffer_create = aub_buffer_create; - iws->winsys.user_buffer_create = aub_user_buffer_create; - iws->winsys.buffer_map = aub_buffer_map; - iws->winsys.buffer_unmap = aub_buffer_unmap; - iws->winsys.buffer_destroy = aub_buffer_destroy; - iws->winsys.flush_frontbuffer = aub_flush_frontbuffer; - iws->winsys.printf = aub_printf; - iws->winsys.get_name = aub_get_name; - - iws->winsys.surface_alloc = aub_i915_surface_alloc; - iws->winsys.surface_alloc_storage = aub_i915_surface_alloc_storage; - iws->winsys.surface_release = aub_i915_surface_release; - - iws->aubfile = brw_aubfile_create(); - iws->size = AUB_BUF_SIZE; - iws->pool = malloc(AUB_BUF_SIZE); - - /* HACK: static copy of this pointer: - */ - assert(global_winsys == NULL); - global_winsys = iws; - - return &iws->winsys; -} - - -void -xmesa_destroy_pipe_winsys_aub( struct pipe_winsys *winsys ) - -{ - struct aub_pipe_winsys *iws = aub_pipe_winsys(winsys); - brw_aub_destroy(iws->aubfile); - free(iws->pool); - free(iws); -} - - - - - - - -#define IWS_BATCHBUFFER_SIZE 1024 - -struct aub_brw_winsys { - struct brw_winsys winsys; /**< batch buffer funcs */ - struct aub_context *aub; - - struct pipe_winsys *pipe_winsys; - - unsigned batch_data[IWS_BATCHBUFFER_SIZE]; - unsigned batch_nr; - unsigned batch_size; - unsigned batch_alloc; -}; - - -/* Turn a i965simple winsys into an aub/i965simple winsys: - */ -static inline struct aub_brw_winsys * -aub_brw_winsys( struct brw_winsys *sws ) -{ - return (struct aub_brw_winsys *)sws; -} - - -/* Simple batchbuffer interface: - */ - -static unsigned *aub_i965_batch_start( struct brw_winsys *sws, - unsigned dwords, - unsigned relocs ) -{ - struct aub_brw_winsys *iws = aub_brw_winsys(sws); - - if (iws->batch_size < iws->batch_nr + dwords) - return NULL; - - iws->batch_alloc = iws->batch_nr + dwords; - return (void *)1; /* not a valid pointer! */ -} - -static void aub_i965_batch_dword( struct brw_winsys *sws, - unsigned dword ) -{ - struct aub_brw_winsys *iws = aub_brw_winsys(sws); - - assert(iws->batch_nr < iws->batch_alloc); - iws->batch_data[iws->batch_nr++] = dword; -} - -static void aub_i965_batch_reloc( struct brw_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta ) -{ - struct aub_brw_winsys *iws = aub_brw_winsys(sws); - - assert(iws->batch_nr < iws->batch_alloc); - iws->batch_data[iws->batch_nr++] = aub_bo(buf)->offset + delta; -} - -static unsigned aub_i965_get_buffer_offset( struct brw_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags ) -{ - return aub_bo(buf)->offset; -} - -static void aub_i965_batch_end( struct brw_winsys *sws ) -{ - struct aub_brw_winsys *iws = aub_brw_winsys(sws); - - assert(iws->batch_nr <= iws->batch_alloc); - iws->batch_alloc = 0; -} - -static void aub_i965_batch_flush( struct brw_winsys *sws, - struct pipe_fence_handle **fence ) -{ - struct aub_brw_winsys *iws = aub_brw_winsys(sws); - assert(iws->batch_nr <= iws->batch_size); - - if (iws->batch_nr) { - xmesa_commands_aub( iws->pipe_winsys, - iws->batch_data, - iws->batch_nr ); - } - - iws->batch_nr = 0; -} - - - -static void aub_i965_buffer_subdata_typed(struct brw_winsys *winsys, - struct pipe_buffer *buf, - unsigned long offset, - unsigned long size, - const void *data, - unsigned data_type) -{ - struct aub_brw_winsys *iws = aub_brw_winsys(winsys); - unsigned aub_type = DW_GENERAL_STATE; - unsigned aub_sub_type; - - switch (data_type) { - case BRW_CC_VP: - aub_sub_type = DWGS_COLOR_CALC_VIEWPORT_STATE; - break; - case BRW_CC_UNIT: - aub_sub_type = DWGS_COLOR_CALC_STATE; - break; - case BRW_WM_PROG: - aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; - break; - case BRW_SAMPLER_DEFAULT_COLOR: - aub_sub_type = DWGS_SAMPLER_DEFAULT_COLOR; - break; - case BRW_SAMPLER: - aub_sub_type = DWGS_SAMPLER_STATE; - break; - case BRW_WM_UNIT: - aub_sub_type = DWGS_WINDOWER_IZ_STATE; - break; - case BRW_SF_PROG: - aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; - break; - case BRW_SF_VP: - aub_sub_type = DWGS_STRIPS_FANS_VIEWPORT_STATE; - break; - case BRW_SF_UNIT: - aub_sub_type = DWGS_STRIPS_FANS_STATE; - break; - case BRW_VS_UNIT: - aub_sub_type = DWGS_VERTEX_SHADER_STATE; - break; - case BRW_VS_PROG: - aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; - break; - case BRW_GS_UNIT: - aub_sub_type = DWGS_GEOMETRY_SHADER_STATE; - break; - case BRW_GS_PROG: - aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; - break; - case BRW_CLIP_VP: - aub_sub_type = DWGS_CLIPPER_VIEWPORT_STATE; - break; - case BRW_CLIP_UNIT: - aub_sub_type = DWGS_CLIPPER_STATE; - break; - case BRW_CLIP_PROG: - aub_sub_type = DWGS_KERNEL_INSTRUCTIONS; - break; - case BRW_SS_SURFACE: - aub_type = DW_SURFACE_STATE; - aub_sub_type = DWSS_SURFACE_STATE; - break; - case BRW_SS_SURF_BIND: - aub_type = DW_SURFACE_STATE; - aub_sub_type = DWSS_BINDING_TABLE_STATE; - break; - case BRW_CONSTANT_BUFFER: - aub_type = DW_CONSTANT_URB_ENTRY; - aub_sub_type = 0; - break; - - default: - assert(0); - break; - } - - xmesa_buffer_subdata_aub( iws->pipe_winsys, - buf, - offset, - size, - data, - aub_type, - aub_sub_type ); -} - -/** - * Create i965 hardware rendering context. - */ -struct pipe_context * -xmesa_create_i965simple( struct pipe_winsys *winsys ) -{ - struct aub_brw_winsys *iws = CALLOC_STRUCT( aub_brw_winsys ); - - /* Fill in this struct with callbacks that i965simple will need to - * communicate with the window system, buffer manager, etc. - */ - iws->winsys.batch_start = aub_i965_batch_start; - iws->winsys.batch_dword = aub_i965_batch_dword; - iws->winsys.batch_reloc = aub_i965_batch_reloc; - iws->winsys.batch_end = aub_i965_batch_end; - iws->winsys.batch_flush = aub_i965_batch_flush; - iws->winsys.buffer_subdata_typed = aub_i965_buffer_subdata_typed; - iws->winsys.get_buffer_offset = aub_i965_get_buffer_offset; - - iws->pipe_winsys = winsys; - - iws->batch_size = IWS_BATCHBUFFER_SIZE; - - /* Create the i965simple context: - */ - return brw_create( winsys, - &iws->winsys, - 0 ); -} diff --git a/src/mesa/pipe/xlib/xm_winsys_aub.h b/src/mesa/pipe/xlib/xm_winsys_aub.h deleted file mode 100644 index 7bee199116..0000000000 --- a/src/mesa/pipe/xlib/xm_winsys_aub.h +++ /dev/null @@ -1,67 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef AUB_WINSYS_H -#define AUB_WINSYS_H - -struct pipe_context; -struct pipe_winsys; -struct pipe_buffer; -struct pipe_surface; - -struct pipe_winsys * -xmesa_create_pipe_winsys_aub( void ); - -void -xmesa_destroy_pipe_winsys_aub( struct pipe_winsys *winsys ); - - - -struct pipe_context * -xmesa_create_i965simple( struct pipe_winsys *winsys ); - - - -void xmesa_buffer_subdata_aub(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned long offset, - unsigned long size, - const void *data, - unsigned aub_type, - unsigned aub_sub_type); - -void xmesa_commands_aub(struct pipe_winsys *winsys, - unsigned *cmds, - unsigned nr_dwords); - - -void xmesa_display_aub( /* struct pipe_winsys *winsys, */ - struct pipe_surface *surface ); - -struct pipe_winsys *xmesa_get_pipe_winsys_aub(void); - -#endif diff --git a/src/mesa/pipe/xlib/xmesaP.h b/src/mesa/pipe/xlib/xmesaP.h deleted file mode 100644 index fa8d1f14b9..0000000000 --- a/src/mesa/pipe/xlib/xmesaP.h +++ /dev/null @@ -1,176 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 7.1 - * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -#ifndef XMESAP_H -#define XMESAP_H - - -#include "GL/xmesa.h" -#include "mtypes.h" -#ifdef XFree86Server -#include "xm_image.h" -#endif - -#include "state_tracker/st_context.h" -#include "state_tracker/st_public.h" - - -extern _glthread_Mutex _xmesa_lock; - -extern XMesaBuffer XMesaBufferList; - -/* - */ -#define XMESA_SOFTPIPE 1 -#define XMESA_AUB 2 -extern int xmesa_mode; - - -/** - * Visual inforation, derived from GLvisual. - * Basically corresponds to an XVisualInfo. - */ -struct xmesa_visual { - GLvisual mesa_visual; /* Device independent visual parameters */ - XMesaDisplay *display; /* The X11 display */ -#ifdef XFree86Server - GLint ColormapEntries; - GLint nplanes; -#else - XMesaVisualInfo visinfo; /* X's visual info (pointer to private copy) */ - XVisualInfo *vishandle; /* Only used in fakeglx.c */ -#endif - GLint BitsPerPixel; /* True bits per pixel for XImages */ - - GLboolean ximage_flag; /* Use XImage for back buffer (not pixmap)? */ -}; - - -/** - * Context info, derived from st_context. - * Basically corresponds to a GLXContext. - */ -struct xmesa_context { - struct st_context *st; - XMesaVisual xm_visual; /** pixel format info */ - XMesaBuffer xm_buffer; /** current drawbuffer */ -}; - - -/** - * Types of X/GLX drawables we might render into. - */ -typedef enum { - WINDOW, /* An X window */ - GLXWINDOW, /* GLX window */ - PIXMAP, /* GLX pixmap */ - PBUFFER /* GLX Pbuffer */ -} BufferType; - - -/** - * Framebuffer information, derived from. - * Basically corresponds to a GLXDrawable. - */ -struct xmesa_buffer { - struct st_framebuffer *stfb; - - GLboolean wasCurrent; /* was ever the current buffer? */ - XMesaVisual xm_visual; /* the X/Mesa visual */ - XMesaDrawable drawable; /* Usually the X window ID */ - XMesaColormap cmap; /* the X colormap */ - BufferType type; /* window, pixmap, pbuffer or glxwindow */ - - XMesaImage *tempImage; - unsigned long selectedEvents;/* for pbuffers only */ - - GLuint shm; /* X Shared Memory extension status: */ - /* 0 = not available */ - /* 1 = XImage support available */ - /* 2 = Pixmap support available too */ -#if defined(USE_XSHM) && !defined(XFree86Server) - XShmSegmentInfo shminfo; -#endif - - XMesaGC gc; /* scratch GC for span, line, tri drawing */ - - /* GLX_EXT_texture_from_pixmap */ - GLint TextureTarget; /** GLX_TEXTURE_1D_EXT, for example */ - GLint TextureFormat; /** GLX_TEXTURE_FORMAT_RGB_EXT, for example */ - GLint TextureMipmap; /** 0 or 1 */ - - struct xmesa_buffer *Next; /* Linked list pointer: */ -}; - - - -/** cast wrapper */ -static INLINE XMesaContext -xmesa_context(GLcontext *ctx) -{ - return (XMesaContext) ctx->DriverCtx; -} - - -/** cast wrapper */ -static INLINE XMesaBuffer -xmesa_buffer(GLframebuffer *fb) -{ - struct st_framebuffer *stfb = (struct st_framebuffer *) fb; - return (XMesaBuffer) st_framebuffer_private(stfb); -} - - -extern void -xmesa_delete_framebuffer(struct gl_framebuffer *fb); - -extern XMesaBuffer -xmesa_find_buffer(XMesaDisplay *dpy, XMesaColormap cmap, XMesaBuffer notThis); - -extern void -xmesa_check_and_update_buffer_size(XMesaContext xmctx, XMesaBuffer drawBuffer); - -extern void -xmesa_destroy_buffers_on_display(XMesaDisplay *dpy); - -extern struct pipe_context * -xmesa_create_pipe_context(XMesaContext xm, uint pixelformat); - -static INLINE GLuint -xmesa_buffer_width(XMesaBuffer b) -{ - return b->stfb->Base.Width; -} - -static INLINE GLuint -xmesa_buffer_height(XMesaBuffer b) -{ - return b->stfb->Base.Height; -} - -extern void -xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf); - -#endif -- cgit v1.2.3 From 6acd63a4980951727939c0dd545a0324965b3834 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Feb 2008 17:50:12 +0900 Subject: Code reorganization: update build. Update the Makefiles and includes for the new paths. Note that there hasn't been no separation of the Makefiles yet, and make is jumping all over the place. That will be taken care shortly. But for now, make should work. It was tested with linux and linux-dri. Linux-cell and linux-llvm might require some minor tweaks. --- src/gallium/Makefile | 12 +-- src/gallium/Makefile.template | 7 +- src/gallium/aux/Makefile | 24 +++++ src/gallium/aux/draw/draw_private.h | 2 +- src/gallium/aux/draw/draw_vertex.c | 4 +- src/gallium/aux/draw/draw_vertex_shader.c | 4 +- src/gallium/aux/llvm/Makefile | 4 +- src/gallium/aux/llvm/gallivm.cpp | 4 +- src/gallium/aux/llvm/gallivm_cpu.cpp | 4 +- src/gallium/aux/llvm/tgsitollvm.cpp | 10 +- src/gallium/aux/pipebuffer/Makefile | 2 +- src/gallium/aux/tgsi/exec/tgsi_exec.c | 4 +- src/gallium/aux/tgsi/exec/tgsi_sse2.c | 4 +- src/gallium/aux/tgsi/util/tgsi_transform.h | 4 +- src/gallium/drivers/Makefile | 24 +++++ src/gallium/drivers/cell/ppu/Makefile | 7 +- src/gallium/drivers/cell/ppu/cell_clear.c | 2 +- src/gallium/drivers/cell/ppu/cell_context.c | 6 +- src/gallium/drivers/cell/ppu/cell_context.h | 6 +- src/gallium/drivers/cell/ppu/cell_draw_arrays.c | 2 +- src/gallium/drivers/cell/ppu/cell_flush.c | 2 +- src/gallium/drivers/cell/ppu/cell_render.c | 2 +- src/gallium/drivers/cell/ppu/cell_spu.c | 2 +- src/gallium/drivers/cell/ppu/cell_spu.h | 2 +- src/gallium/drivers/cell/ppu/cell_state_blend.c | 2 +- src/gallium/drivers/cell/ppu/cell_state_clip.c | 2 +- src/gallium/drivers/cell/ppu/cell_state_derived.c | 4 +- src/gallium/drivers/cell/ppu/cell_state_fs.c | 8 +- .../drivers/cell/ppu/cell_state_rasterizer.c | 2 +- src/gallium/drivers/cell/ppu/cell_state_sampler.c | 2 +- src/gallium/drivers/cell/ppu/cell_state_vertex.c | 2 +- src/gallium/drivers/cell/ppu/cell_surface.c | 2 +- src/gallium/drivers/cell/ppu/cell_vbuf.c | 2 +- src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 6 +- src/gallium/drivers/cell/spu/Makefile | 6 +- src/gallium/drivers/cell/spu/spu_exec.c | 4 +- src/gallium/drivers/cell/spu/spu_exec.h | 2 +- src/gallium/drivers/cell/spu/spu_main.c | 2 +- src/gallium/drivers/cell/spu/spu_main.h | 4 +- src/gallium/drivers/cell/spu/spu_render.c | 2 +- src/gallium/drivers/cell/spu/spu_render.h | 2 +- src/gallium/drivers/cell/spu/spu_tile.h | 2 +- src/gallium/drivers/cell/spu/spu_util.c | 4 +- src/gallium/drivers/cell/spu/spu_vertex_shader.c | 6 +- src/gallium/drivers/failover/Makefile | 2 +- src/gallium/drivers/i915simple/Makefile | 2 +- src/gallium/drivers/i915simple/i915_context.c | 2 +- src/gallium/drivers/i915simple/i915_context.h | 2 +- .../drivers/i915simple/i915_fpc_translate.c | 4 +- src/gallium/drivers/i915simple/i915_prim_emit.c | 2 +- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 2 +- src/gallium/drivers/i915simple/i915_state.c | 2 +- .../drivers/i915simple/i915_state_derived.c | 4 +- src/gallium/drivers/i915simple/i915_strings.c | 2 +- src/gallium/drivers/i915simple/i915_surface.c | 2 +- src/gallium/drivers/i965simple/Makefile | 2 +- src/gallium/drivers/i965simple/brw_shader_info.c | 2 +- src/gallium/drivers/i965simple/brw_state.c | 2 +- src/gallium/drivers/i965simple/brw_strings.c | 2 +- src/gallium/drivers/i965simple/brw_surface.c | 2 +- src/gallium/drivers/i965simple/brw_vs_emit.c | 2 +- src/gallium/drivers/i965simple/brw_wm_decl.c | 2 +- src/gallium/drivers/i965simple/brw_wm_glsl.c | 2 +- src/gallium/drivers/softpipe/Makefile | 2 +- src/gallium/drivers/softpipe/sp_context.c | 2 +- src/gallium/drivers/softpipe/sp_context.h | 2 +- src/gallium/drivers/softpipe/sp_draw_arrays.c | 2 +- src/gallium/drivers/softpipe/sp_flush.c | 2 +- src/gallium/drivers/softpipe/sp_headers.h | 2 +- src/gallium/drivers/softpipe/sp_prim_setup.c | 4 +- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 6 +- src/gallium/drivers/softpipe/sp_quad_fs.c | 2 +- src/gallium/drivers/softpipe/sp_query.c | 2 +- src/gallium/drivers/softpipe/sp_state_clip.c | 2 +- src/gallium/drivers/softpipe/sp_state_derived.c | 6 +- src/gallium/drivers/softpipe/sp_state_fs.c | 8 +- src/gallium/drivers/softpipe/sp_state_rasterizer.c | 2 +- src/gallium/drivers/softpipe/sp_state_sampler.c | 4 +- src/gallium/drivers/softpipe/sp_state_vertex.c | 2 +- src/gallium/drivers/softpipe/sp_surface.c | 2 +- src/gallium/drivers/softpipe/sp_tex_sample.c | 2 +- src/gallium/drivers/softpipe/sp_tile_cache.c | 2 +- src/gallium/winsys/dri/Makefile | 38 +++++++ src/gallium/winsys/dri/Makefile.template | 113 +++++++++++++++++++++ src/gallium/winsys/dri/intel/Makefile | 8 +- src/gallium/winsys/dri/intel/intel_winsys_i915.c | 2 +- .../winsys/dri/intel/intel_winsys_softpipe.c | 2 +- src/gallium/winsys/xlib/xm_winsys.c | 6 +- src/gallium/winsys/xlib/xm_winsys_aub.c | 2 +- src/mesa/Makefile | 16 ++- src/mesa/drivers/x11/xm_api.c | 2 +- src/mesa/drivers/x11/xm_dd.c | 2 +- src/mesa/drivers/x11/xm_surface.c | 8 +- src/mesa/drivers/x11/xm_winsys.c | 2 +- src/mesa/drivers/x11/xmesaP.h | 4 +- src/mesa/sources | 83 +++++++-------- src/mesa/state_tracker/st_atom_shader.c | 2 +- src/mesa/state_tracker/st_cache.c | 4 +- src/mesa/state_tracker/st_cache.h | 2 +- src/mesa/state_tracker/st_cb_accum.c | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_cb_feedback.c | 6 +- src/mesa/state_tracker/st_cb_program.c | 4 +- src/mesa/state_tracker/st_cb_rasterpos.c | 4 +- src/mesa/state_tracker/st_cb_readpixels.c | 2 +- src/mesa/state_tracker/st_cb_texture.c | 2 +- src/mesa/state_tracker/st_context.c | 4 +- src/mesa/state_tracker/st_debug.c | 4 +- src/mesa/state_tracker/st_draw.c | 4 +- src/mesa/state_tracker/st_gen_mipmap.c | 2 +- src/mesa/state_tracker/st_mesa_to_tgsi.c | 6 +- src/mesa/state_tracker/st_program.c | 4 +- 112 files changed, 427 insertions(+), 223 deletions(-) create mode 100644 src/gallium/aux/Makefile create mode 100644 src/gallium/drivers/Makefile create mode 100644 src/gallium/winsys/dri/Makefile create mode 100644 src/gallium/winsys/dri/Makefile.template (limited to 'src') diff --git a/src/gallium/Makefile b/src/gallium/Makefile index d880d090c1..a13b9a52d3 100644 --- a/src/gallium/Makefile +++ b/src/gallium/Makefile @@ -1,16 +1,8 @@ -TOP = ../../.. +TOP = ../.. include $(TOP)/configs/current -ifeq ($(CONFIG_NAME), linux-cell) -CELL_DIR = cell -endif - -ifeq ($(CONFIG_NAME), linux-llvm) -LLVM_DIR = llvm -endif - -SUBDIRS = softpipe i915simple i965simple failover pipebuffer $(CELL_DIR) $(LLVM_DIR) +SUBDIRS = aux drivers default: subdirs diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template index 8e84f8eb2d..0717ed8dd2 100644 --- a/src/gallium/Makefile.template +++ b/src/gallium/Makefile.template @@ -15,7 +15,10 @@ OBJECTS = $(C_SOURCES:.c=.o) \ ### Include directories INCLUDES = \ -I. \ - -I$(TOP)/src/mesa/pipe \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/include/pipe \ + -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/drivers \ -I$(TOP)/src/mesa \ -I$(TOP)/include \ $(DRIVER_INCLUDES) @@ -38,7 +41,7 @@ INCLUDES = \ default: depend symlinks $(LIBNAME) -$(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/mesa/pipe/Makefile.template +$(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/gallium/Makefile.template $(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS) diff --git a/src/gallium/aux/Makefile b/src/gallium/aux/Makefile new file mode 100644 index 0000000000..da68498aa1 --- /dev/null +++ b/src/gallium/aux/Makefile @@ -0,0 +1,24 @@ +TOP = ../../.. +include $(TOP)/configs/current + + +ifeq ($(CONFIG_NAME), linux-llvm) +LLVM_DIR = llvm +endif + +SUBDIRS = pipebuffer $(LLVM_DIR) + + +default: subdirs + + +subdirs: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +clean: + rm -f `find . -name \*.[oa]` diff --git a/src/gallium/aux/draw/draw_private.h b/src/gallium/aux/draw/draw_private.h index b17eaaed65..3d09aef87c 100644 --- a/src/gallium/aux/draw/draw_private.h +++ b/src/gallium/aux/draw/draw_private.h @@ -45,7 +45,7 @@ #include "pipe/p_defines.h" #include "x86/rtasm/x86sse.h" -#include "pipe/tgsi/exec/tgsi_exec.h" +#include "tgsi/exec/tgsi_exec.h" struct gallivm_prog; diff --git a/src/gallium/aux/draw/draw_vertex.c b/src/gallium/aux/draw/draw_vertex.c index 2d6592150f..daf1ef4b80 100644 --- a/src/gallium/aux/draw/draw_vertex.c +++ b/src/gallium/aux/draw/draw_vertex.c @@ -34,8 +34,8 @@ */ -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vertex.h" +#include "draw/draw_private.h" +#include "draw/draw_vertex.h" /** diff --git a/src/gallium/aux/draw/draw_vertex_shader.c b/src/gallium/aux/draw/draw_vertex_shader.c index c824c1407e..377ecbb931 100644 --- a/src/gallium/aux/draw/draw_vertex_shader.c +++ b/src/gallium/aux/draw/draw_vertex_shader.c @@ -34,13 +34,13 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #if defined(__i386__) || defined(__386__) -#include "pipe/tgsi/exec/tgsi_sse2.h" +#include "tgsi/exec/tgsi_sse2.h" #endif #include "draw_private.h" #include "draw_context.h" #include "x86/rtasm/x86sse.h" -#include "pipe/llvm/gallivm.h" +#include "llvm/gallivm.h" #define DBG_VS 0 diff --git a/src/gallium/aux/llvm/Makefile b/src/gallium/aux/llvm/Makefile index 9c6e16d86b..e6ac399d08 100644 --- a/src/gallium/aux/llvm/Makefile +++ b/src/gallium/aux/llvm/Makefile @@ -30,7 +30,9 @@ OBJECTS = $(C_SOURCES:.c=.o) \ ### Include directories INCLUDES = \ -I. \ - -I$(TOP)/src/mesa/pipe \ + -I$(TOP)/src/gallium/drivers + -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/include \ -I$(TOP)/src/mesa \ -I$(TOP)/include diff --git a/src/gallium/aux/llvm/gallivm.cpp b/src/gallium/aux/llvm/gallivm.cpp index da0105c2c9..d14bb3b99a 100644 --- a/src/gallium/aux/llvm/gallivm.cpp +++ b/src/gallium/aux/llvm/gallivm.cpp @@ -42,8 +42,8 @@ #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/exec/tgsi_exec.h" -#include "pipe/tgsi/util/tgsi_dump.h" +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_dump.h" #include #include diff --git a/src/gallium/aux/llvm/gallivm_cpu.cpp b/src/gallium/aux/llvm/gallivm_cpu.cpp index dc4d92a72a..8f9830d0b1 100644 --- a/src/gallium/aux/llvm/gallivm_cpu.cpp +++ b/src/gallium/aux/llvm/gallivm_cpu.cpp @@ -42,8 +42,8 @@ #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/exec/tgsi_exec.h" -#include "pipe/tgsi/util/tgsi_dump.h" +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_dump.h" #include #include diff --git a/src/gallium/aux/llvm/tgsitollvm.cpp b/src/gallium/aux/llvm/tgsitollvm.cpp index 0de595e678..2cb4acce32 100644 --- a/src/gallium/aux/llvm/tgsitollvm.cpp +++ b/src/gallium/aux/llvm/tgsitollvm.cpp @@ -10,11 +10,11 @@ #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/exec/tgsi_exec.h" -#include "pipe/tgsi/util/tgsi_util.h" -#include "pipe/tgsi/util/tgsi_build.h" -#include "pipe/tgsi/util/tgsi_dump.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_util.h" +#include "tgsi/util/tgsi_build.h" +#include "tgsi/util/tgsi_dump.h" #include diff --git a/src/gallium/aux/pipebuffer/Makefile b/src/gallium/aux/pipebuffer/Makefile index 75764a9a18..588629e870 100644 --- a/src/gallium/aux/pipebuffer/Makefile +++ b/src/gallium/aux/pipebuffer/Makefile @@ -17,7 +17,7 @@ C_SOURCES = \ ASM_SOURCES = -include ../Makefile.template +include ../../Makefile.template symlinks: diff --git a/src/gallium/aux/tgsi/exec/tgsi_exec.c b/src/gallium/aux/tgsi/exec/tgsi_exec.c index 37e6007068..a8f64c2287 100644 --- a/src/gallium/aux/tgsi/exec/tgsi_exec.c +++ b/src/gallium/aux/tgsi/exec/tgsi_exec.c @@ -54,8 +54,8 @@ #include "pipe/p_state.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/util/tgsi_util.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_util.h" #include "tgsi_exec.h" #define TILE_TOP_LEFT 0 diff --git a/src/gallium/aux/tgsi/exec/tgsi_sse2.c b/src/gallium/aux/tgsi/exec/tgsi_sse2.c index 1e56e4afb6..593464db3e 100755 --- a/src/gallium/aux/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/aux/tgsi/exec/tgsi_sse2.c @@ -27,8 +27,8 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/util/tgsi_util.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_util.h" #include "tgsi_exec.h" #include "tgsi_sse2.h" diff --git a/src/gallium/aux/tgsi/util/tgsi_transform.h b/src/gallium/aux/tgsi/util/tgsi_transform.h index 365d8c298c..fcf85d603b 100644 --- a/src/gallium/aux/tgsi/util/tgsi_transform.h +++ b/src/gallium/aux/tgsi/util/tgsi_transform.h @@ -31,8 +31,8 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/util/tgsi_build.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_build.h" diff --git a/src/gallium/drivers/Makefile b/src/gallium/drivers/Makefile new file mode 100644 index 0000000000..c0345a9cb5 --- /dev/null +++ b/src/gallium/drivers/Makefile @@ -0,0 +1,24 @@ +TOP = ../../.. +include $(TOP)/configs/current + + +ifeq ($(CONFIG_NAME), linux-cell) +CELL_DIR = cell +endif + +SUBDIRS = softpipe i915simple i965simple failover pipebuffer $(CELL_DIR) + + +default: subdirs + + +subdirs: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +clean: + rm -f `find . -name \*.[oa]` diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile index 50060f5cd3..011863c11e 100644 --- a/src/gallium/drivers/cell/ppu/Makefile +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -40,8 +40,11 @@ SOURCES = \ OBJECTS = $(SOURCES:.c=.o) \ -INCLUDE_DIRS = -I$(TOP)/src/mesa - +INCLUDE_DIRS = \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/drivers .c.o: $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ diff --git a/src/gallium/drivers/cell/ppu/cell_clear.c b/src/gallium/drivers/cell/ppu/cell_clear.c index 07b908eec5..e588a30d5b 100644 --- a/src/gallium/drivers/cell/ppu/cell_clear.c +++ b/src/gallium/drivers/cell/ppu/cell_clear.c @@ -35,7 +35,7 @@ #include #include "pipe/p_inlines.h" #include "pipe/p_util.h" -#include "pipe/cell/common.h" +#include "cell/common.h" #include "cell_clear.h" #include "cell_context.h" #include "cell_batch.h" diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index bbe1fd7a11..e1eb22f468 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -37,9 +37,9 @@ #include "pipe/p_format.h" #include "pipe/p_util.h" #include "pipe/p_winsys.h" -#include "pipe/cell/common.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" +#include "cell/common.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" #include "cell_clear.h" #include "cell_context.h" #include "cell_draw_arrays.h" diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 3b63419b5e..6196c0c72f 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -32,10 +32,10 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/draw/draw_vertex.h" -#include "pipe/draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_vbuf.h" #include "cell_winsys.h" -#include "pipe/cell/common.h" +#include "cell/common.h" struct cell_vbuf_render; diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c index 717cd8370f..f12613649b 100644 --- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c @@ -39,7 +39,7 @@ #include "cell_draw_arrays.h" #include "cell_state.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" diff --git a/src/gallium/drivers/cell/ppu/cell_flush.c b/src/gallium/drivers/cell/ppu/cell_flush.c index f62bc4650c..20f27531fc 100644 --- a/src/gallium/drivers/cell/ppu/cell_flush.c +++ b/src/gallium/drivers/cell/ppu/cell_flush.c @@ -31,7 +31,7 @@ #include "cell_flush.h" #include "cell_spu.h" #include "cell_render.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" void diff --git a/src/gallium/drivers/cell/ppu/cell_render.c b/src/gallium/drivers/cell/ppu/cell_render.c index 4ab277a4b2..b663b37622 100644 --- a/src/gallium/drivers/cell/ppu/cell_render.c +++ b/src/gallium/drivers/cell/ppu/cell_render.c @@ -34,7 +34,7 @@ #include "cell_render.h" #include "cell_spu.h" #include "pipe/p_util.h" -#include "pipe/draw/draw_private.h" +#include "draw/draw_private.h" struct render_stage { diff --git a/src/gallium/drivers/cell/ppu/cell_spu.c b/src/gallium/drivers/cell/ppu/cell_spu.c index 7c83a47e57..419e74dc40 100644 --- a/src/gallium/drivers/cell/ppu/cell_spu.c +++ b/src/gallium/drivers/cell/ppu/cell_spu.c @@ -31,7 +31,7 @@ #include "cell_spu.h" #include "pipe/p_format.h" #include "pipe/p_state.h" -#include "pipe/cell/common.h" +#include "cell/common.h" /* diff --git a/src/gallium/drivers/cell/ppu/cell_spu.h b/src/gallium/drivers/cell/ppu/cell_spu.h index 19eff94f96..137f26612e 100644 --- a/src/gallium/drivers/cell/ppu/cell_spu.h +++ b/src/gallium/drivers/cell/ppu/cell_spu.h @@ -31,7 +31,7 @@ #include #include -#include "pipe/cell/common.h" +#include "cell/common.h" #include "cell_context.h" diff --git a/src/gallium/drivers/cell/ppu/cell_state_blend.c b/src/gallium/drivers/cell/ppu/cell_state_blend.c index 4fc60548c8..b6d6d71f0c 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_blend.c +++ b/src/gallium/drivers/cell/ppu/cell_state_blend.c @@ -29,7 +29,7 @@ */ #include "pipe/p_util.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "cell_context.h" #include "cell_state.h" diff --git a/src/gallium/drivers/cell/ppu/cell_state_clip.c b/src/gallium/drivers/cell/ppu/cell_state_clip.c index 4f43665941..0482f87e88 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_clip.c +++ b/src/gallium/drivers/cell/ppu/cell_state_clip.c @@ -30,7 +30,7 @@ #include "cell_context.h" #include "cell_state.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" void cell_set_clip_state( struct pipe_context *pipe, diff --git a/src/gallium/drivers/cell/ppu/cell_state_derived.c b/src/gallium/drivers/cell/ppu/cell_state_derived.c index 56daf5dfde..0c46829258 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_derived.c +++ b/src/gallium/drivers/cell/ppu/cell_state_derived.c @@ -27,8 +27,8 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_vertex.h" +#include "draw/draw_context.h" +#include "draw/draw_vertex.h" #include "cell_context.h" #include "cell_batch.h" #include "cell_state.h" diff --git a/src/gallium/drivers/cell/ppu/cell_state_fs.c b/src/gallium/drivers/cell/ppu/cell_state_fs.c index 3f46a87d18..b2ed699a5b 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_fs.c +++ b/src/gallium/drivers/cell/ppu/cell_state_fs.c @@ -29,12 +29,12 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #if 0 #include "pipe/p_shader_tokens.h" -#include "pipe/llvm/gallivm.h" -#include "pipe/tgsi/util/tgsi_dump.h" -#include "pipe/tgsi/exec/tgsi_sse2.h" +#include "llvm/gallivm.h" +#include "tgsi/util/tgsi_dump.h" +#include "tgsi/exec/tgsi_sse2.h" #endif #include "cell_context.h" diff --git a/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c b/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c index d8128ece54..7eca5b5765 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c +++ b/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c @@ -27,7 +27,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "cell_context.h" #include "cell_state.h" diff --git a/src/gallium/drivers/cell/ppu/cell_state_sampler.c b/src/gallium/drivers/cell/ppu/cell_state_sampler.c index ade6cc8338..a33421a4ad 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_sampler.c +++ b/src/gallium/drivers/cell/ppu/cell_state_sampler.c @@ -30,7 +30,7 @@ */ #include "pipe/p_util.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "cell_context.h" #include "cell_state.h" #include "cell_texture.h" diff --git a/src/gallium/drivers/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c index 0f01e920f9..563831b62d 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_vertex.c +++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c @@ -32,7 +32,7 @@ #include "cell_context.h" #include "cell_state.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" void diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c index fca93e4742..a35db0ef99 100644 --- a/src/gallium/drivers/cell/ppu/cell_surface.c +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -29,7 +29,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "pipe/util/p_tile.h" +#include "util/p_tile.h" #include "cell_context.h" #include "cell_surface.h" diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.c b/src/gallium/drivers/cell/ppu/cell_vbuf.c index e9fafe492e..cc727ff4ed 100644 --- a/src/gallium/drivers/cell/ppu/cell_vbuf.c +++ b/src/gallium/drivers/cell/ppu/cell_vbuf.c @@ -36,7 +36,7 @@ #include "cell_flush.h" #include "cell_spu.h" #include "cell_vbuf.h" -#include "pipe/draw/draw_vbuf.h" +#include "draw/draw_vbuf.h" /** Allow vertex data to be inlined after RENDER command */ diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index 80dd500b34..0ba4506505 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -38,9 +38,9 @@ #include "cell_spu.h" #include "cell_batch.h" -#include "pipe/cell/common.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" +#include "cell/common.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" /** * Run the vertex shader on all vertices in the vertex queue. diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile index f202971d73..7aa947299e 100644 --- a/src/gallium/drivers/cell/spu/Makefile +++ b/src/gallium/drivers/cell/spu/Makefile @@ -31,7 +31,11 @@ SPU_OBJECTS = $(SOURCES:.c=.o) \ SPU_ASM_OUT = $(SOURCES:.c=.s) \ -INCLUDE_DIRS = -I$(TOP)/src/mesa +INCLUDE_DIRS = \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/drivers .c.o: diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index e51008b9b3..109540b1f7 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -67,8 +67,8 @@ #include "pipe/p_state.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/util/tgsi_util.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_util.h" #include "spu_exec.h" #include "spu_main.h" #include "spu_vertex_shader.h" diff --git a/src/gallium/drivers/cell/spu/spu_exec.h b/src/gallium/drivers/cell/spu/spu_exec.h index b4c7661ef6..3e17c490d2 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.h +++ b/src/gallium/drivers/cell/spu/spu_exec.h @@ -29,7 +29,7 @@ #define SPU_EXEC_H #include "pipe/p_compiler.h" -#include "pipe/tgsi/exec/tgsi_exec.h" +#include "tgsi/exec/tgsi_exec.h" #if defined __cplusplus extern "C" { diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index e375197fe6..1e7243b863 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -38,7 +38,7 @@ #include "spu_tile.h" //#include "spu_test.h" #include "spu_vertex_shader.h" -#include "pipe/cell/common.h" +#include "cell/common.h" #include "pipe/p_defines.h" diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index 1710a17512..5c95d112ac 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -31,8 +31,8 @@ #include -#include "pipe/cell/common.h" -#include "pipe/draw/draw_vertex.h" +#include "cell/common.h" +#include "draw/draw_vertex.h" #include "pipe/p_state.h" diff --git a/src/gallium/drivers/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c index 932fb500b3..20e77aa2e6 100644 --- a/src/gallium/drivers/cell/spu/spu_render.c +++ b/src/gallium/drivers/cell/spu/spu_render.c @@ -34,7 +34,7 @@ #include "spu_render.h" #include "spu_tri.h" #include "spu_tile.h" -#include "pipe/cell/common.h" +#include "cell/common.h" diff --git a/src/gallium/drivers/cell/spu/spu_render.h b/src/gallium/drivers/cell/spu/spu_render.h index fbcdc5ec31..493434f087 100644 --- a/src/gallium/drivers/cell/spu/spu_render.h +++ b/src/gallium/drivers/cell/spu/spu_render.h @@ -29,7 +29,7 @@ #ifndef SPU_RENDER_H #define SPU_RENDER_H -#include "pipe/cell/common.h" +#include "cell/common.h" extern void cmd_render(const struct cell_command_render *render, uint *pos_incr); diff --git a/src/gallium/drivers/cell/spu/spu_tile.h b/src/gallium/drivers/cell/spu/spu_tile.h index e53340a55a..3105b848fd 100644 --- a/src/gallium/drivers/cell/spu/spu_tile.h +++ b/src/gallium/drivers/cell/spu/spu_tile.h @@ -32,7 +32,7 @@ #include #include #include "spu_main.h" -#include "pipe/cell/common.h" +#include "cell/common.h" diff --git a/src/gallium/drivers/cell/spu/spu_util.c b/src/gallium/drivers/cell/spu/spu_util.c index ac373240c1..ea4274a0a7 100644 --- a/src/gallium/drivers/cell/spu/spu_util.c +++ b/src/gallium/drivers/cell/spu/spu_util.c @@ -1,8 +1,8 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_parse.h" //#include "tgsi_build.h" -#include "pipe/tgsi/util/tgsi_util.h" +#include "tgsi/util/tgsi_util.h" unsigned tgsi_util_get_src_register_swizzle( diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.c b/src/gallium/drivers/cell/spu/spu_vertex_shader.c index c1cbbb6d1e..3f5bf41aa2 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_shader.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.c @@ -39,9 +39,9 @@ #include "pipe/p_shader_tokens.h" #include "spu_vertex_shader.h" #include "spu_exec.h" -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_context.h" -#include "pipe/cell/common.h" +#include "draw/draw_private.h" +#include "draw/draw_context.h" +#include "cell/common.h" #include "spu_main.h" static INLINE unsigned diff --git a/src/gallium/drivers/failover/Makefile b/src/gallium/drivers/failover/Makefile index 72d0895c74..14389bd055 100644 --- a/src/gallium/drivers/failover/Makefile +++ b/src/gallium/drivers/failover/Makefile @@ -15,7 +15,7 @@ C_SOURCES = \ ASM_SOURCES = -include ../Makefile.template +include ../../Makefile.template symlinks: diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile index 2f91de3afc..ee22ba86f9 100644 --- a/src/gallium/drivers/i915simple/Makefile +++ b/src/gallium/drivers/i915simple/Makefile @@ -32,7 +32,7 @@ C_SOURCES = \ ASM_SOURCES = -include ../Makefile.template +include ../../Makefile.template symlinks: diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 497623a700..7f71f8fd4f 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -32,7 +32,7 @@ #include "i915_texture.h" #include "i915_reg.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "pipe/p_util.h" diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index b4ea63c3e7..2d876925b2 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -33,7 +33,7 @@ #include "pipe/p_defines.h" #include "pipe/p_state.h" -#include "pipe/draw/draw_vertex.h" +#include "draw/draw_vertex.h" #define I915_TEX_UNITS 8 diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 868f0c7e04..6c1524c768 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -33,9 +33,9 @@ #include "i915_fpc.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_parse.h" -#include "pipe/draw/draw_vertex.h" +#include "draw/draw_vertex.h" /** diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c index c4a706c37d..44c4325936 100644 --- a/src/gallium/drivers/i915simple/i915_prim_emit.c +++ b/src/gallium/drivers/i915simple/i915_prim_emit.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/draw/draw_private.h" +#include "draw/draw_private.h" #include "pipe/p_util.h" #include "i915_context.h" diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index e069773fd4..c5bf6174f6 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -38,7 +38,7 @@ */ -#include "pipe/draw/draw_vbuf.h" +#include "draw/draw_vbuf.h" #include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index abd5571b88..294e6fad03 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -29,7 +29,7 @@ */ -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "pipe/p_winsys.h" #include "pipe/p_util.h" diff --git a/src/gallium/drivers/i915simple/i915_state_derived.c b/src/gallium/drivers/i915simple/i915_state_derived.c index 653983e4a9..4767584fc6 100644 --- a/src/gallium/drivers/i915simple/i915_state_derived.c +++ b/src/gallium/drivers/i915simple/i915_state_derived.c @@ -27,8 +27,8 @@ #include "pipe/p_util.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_vertex.h" +#include "draw/draw_context.h" +#include "draw/draw_vertex.h" #include "i915_context.h" #include "i915_state.h" #include "i915_reg.h" diff --git a/src/gallium/drivers/i915simple/i915_strings.c b/src/gallium/drivers/i915simple/i915_strings.c index c713bf7208..301fedea19 100644 --- a/src/gallium/drivers/i915simple/i915_strings.c +++ b/src/gallium/drivers/i915simple/i915_strings.c @@ -70,7 +70,7 @@ static const char *i915_get_name( struct pipe_context *pipe ) break; } - sprintf(buffer, "pipe/i915 (chipset: %s)", chipset); + sprintf(buffer, "i915 (chipset: %s)", chipset); return buffer; } diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index de0cc5fe06..17fd27895a 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -33,7 +33,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "pipe/util/p_tile.h" +#include "util/p_tile.h" /* diff --git a/src/gallium/drivers/i965simple/Makefile b/src/gallium/drivers/i965simple/Makefile index 48c00ab50b..1dec1f9749 100644 --- a/src/gallium/drivers/i965simple/Makefile +++ b/src/gallium/drivers/i965simple/Makefile @@ -61,6 +61,6 @@ ASM_SOURCES = DRIVER_DEFINES = -I. -include ../Makefile.template +include ../../Makefile.template symlinks: diff --git a/src/gallium/drivers/i965simple/brw_shader_info.c b/src/gallium/drivers/i965simple/brw_shader_info.c index 431b45466a..a762a870fe 100644 --- a/src/gallium/drivers/i965simple/brw_shader_info.c +++ b/src/gallium/drivers/i965simple/brw_shader_info.c @@ -3,7 +3,7 @@ #include "brw_state.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_parse.h" diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 95dfce88e4..f746d1cc57 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -33,7 +33,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_dump.h" +#include "tgsi/util/tgsi_dump.h" #include "brw_context.h" #include "brw_defines.h" diff --git a/src/gallium/drivers/i965simple/brw_strings.c b/src/gallium/drivers/i965simple/brw_strings.c index 29a41ed1e9..3d9c50961f 100644 --- a/src/gallium/drivers/i965simple/brw_strings.c +++ b/src/gallium/drivers/i965simple/brw_strings.c @@ -59,7 +59,7 @@ static const char *brw_get_name( struct pipe_context *pipe ) break; } - sprintf(buffer, "pipe/i965 (chipset: %s)", chipset); + sprintf(buffer, "i965 (chipset: %s)", chipset); return buffer; } diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index 518845e4b2..376a42b1a6 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -32,7 +32,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "pipe/util/p_tile.h" +#include "util/p_tile.h" /* diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c index 98915ba101..05df4860ed 100644 --- a/src/gallium/drivers/i965simple/brw_vs_emit.c +++ b/src/gallium/drivers/i965simple/brw_vs_emit.c @@ -33,7 +33,7 @@ #include "brw_vs.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_parse.h" struct brw_prog_info { unsigned num_temps; diff --git a/src/gallium/drivers/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c index b45a333a2e..97418a52e7 100644 --- a/src/gallium/drivers/i965simple/brw_wm_decl.c +++ b/src/gallium/drivers/i965simple/brw_wm_decl.c @@ -4,7 +4,7 @@ #include "brw_wm.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_parse.h" static struct brw_reg alloc_tmp(struct brw_wm_compile *c) { diff --git a/src/gallium/drivers/i965simple/brw_wm_glsl.c b/src/gallium/drivers/i965simple/brw_wm_glsl.c index d95645d108..44f946ea74 100644 --- a/src/gallium/drivers/i965simple/brw_wm_glsl.c +++ b/src/gallium/drivers/i965simple/brw_wm_glsl.c @@ -4,7 +4,7 @@ #include "brw_wm.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_parse.h" diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile index 31438a882e..2304ea4246 100644 --- a/src/gallium/drivers/softpipe/Makefile +++ b/src/gallium/drivers/softpipe/Makefile @@ -44,7 +44,7 @@ C_SOURCES = \ ASM_SOURCES = -include ../Makefile.template +include ../../Makefile.template symlinks: diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index cea6b90104..5e98f190bb 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -29,7 +29,7 @@ * Keith Whitwell */ -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_util.h" diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index aff8c2cc5d..8c79cb3ce4 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -34,7 +34,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/draw/draw_vertex.h" +#include "draw/draw_vertex.h" #include "sp_quad.h" diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 71a303a8b5..2049afda34 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -38,7 +38,7 @@ #include "sp_context.h" #include "sp_state.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c index ced0d5d098..2cbd0d7cab 100644 --- a/src/gallium/drivers/softpipe/sp_flush.c +++ b/src/gallium/drivers/softpipe/sp_flush.c @@ -31,7 +31,7 @@ #include "pipe/p_defines.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "sp_flush.h" #include "sp_context.h" #include "sp_surface.h" diff --git a/src/gallium/drivers/softpipe/sp_headers.h b/src/gallium/drivers/softpipe/sp_headers.h index 0ae31d8796..9cf8222133 100644 --- a/src/gallium/drivers/softpipe/sp_headers.h +++ b/src/gallium/drivers/softpipe/sp_headers.h @@ -31,7 +31,7 @@ #ifndef SP_HEADERS_H #define SP_HEADERS_H -#include "pipe/tgsi/exec/tgsi_exec.h" +#include "tgsi/exec/tgsi_exec.h" #define PRIM_POINT 1 #define PRIM_LINE 2 diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 2772048661..d73521ccbe 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -38,8 +38,8 @@ #include "sp_quad.h" #include "sp_state.h" #include "sp_prim_setup.h" -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vertex.h" +#include "draw/draw_private.h" +#include "draw/draw_vertex.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 7f71fdb6a9..69bea8a8f5 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -39,9 +39,9 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_prim_vbuf.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_vbuf.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" #define SP_MAX_VBUF_INDEXES 1024 diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 3316858413..b4c01a7ea8 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -42,7 +42,7 @@ #include "x86/rtasm/x86sse.h" #ifdef MESA_LLVM -#include "pipe/llvm/gallivm.h" +#include "llvm/gallivm.h" #endif #include "sp_context.h" diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c index 6a8a43aeda..adf9ccf64c 100644 --- a/src/gallium/drivers/softpipe/sp_query.c +++ b/src/gallium/drivers/softpipe/sp_query.c @@ -29,7 +29,7 @@ * Keith Whitwell */ -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_util.h" diff --git a/src/gallium/drivers/softpipe/sp_state_clip.c b/src/gallium/drivers/softpipe/sp_state_clip.c index 08c5f06d05..c797c0dd3b 100644 --- a/src/gallium/drivers/softpipe/sp_state_clip.c +++ b/src/gallium/drivers/softpipe/sp_state_clip.c @@ -29,7 +29,7 @@ */ #include "sp_context.h" #include "sp_state.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" void softpipe_set_clip_state( struct pipe_context *pipe, diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 372597869f..9d8fd8b750 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -27,9 +27,9 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_vertex.h" -#include "pipe/draw/draw_private.h" +#include "draw/draw_context.h" +#include "draw/draw_vertex.h" +#include "draw/draw_private.h" #include "sp_context.h" #include "sp_state.h" diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 0b814fc284..1e3cadd43d 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -32,11 +32,11 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "pipe/p_shader_tokens.h" -#include "pipe/llvm/gallivm.h" -#include "pipe/tgsi/util/tgsi_dump.h" -#include "pipe/tgsi/exec/tgsi_sse2.h" +#include "llvm/gallivm.h" +#include "tgsi/util/tgsi_dump.h" +#include "tgsi/exec/tgsi_sse2.h" void * diff --git a/src/gallium/drivers/softpipe/sp_state_rasterizer.c b/src/gallium/drivers/softpipe/sp_state_rasterizer.c index 53755099dd..98e04352db 100644 --- a/src/gallium/drivers/softpipe/sp_state_rasterizer.c +++ b/src/gallium/drivers/softpipe/sp_state_rasterizer.c @@ -29,7 +29,7 @@ #include "pipe/p_util.h" #include "sp_context.h" #include "sp_state.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c index ea348c7e95..460adccec4 100644 --- a/src/gallium/drivers/softpipe/sp_state_sampler.c +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -31,14 +31,14 @@ #include "pipe/p_util.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" #include "sp_context.h" #include "sp_context.h" #include "sp_state.h" #include "sp_texture.h" #include "sp_tile_cache.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c index 09ff540ccf..f01a10de3b 100644 --- a/src/gallium/drivers/softpipe/sp_state_vertex.c +++ b/src/gallium/drivers/softpipe/sp_state_vertex.c @@ -33,7 +33,7 @@ #include "sp_state.h" #include "sp_surface.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_context.h" void diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index 8802ced187..653449c4f1 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -29,7 +29,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "pipe/util/p_tile.h" +#include "util/p_tile.h" #include "sp_context.h" #include "sp_surface.h" diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 325bdb86da..2f82fd6abe 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -40,7 +40,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" -#include "pipe/tgsi/exec/tgsi_exec.h" +#include "tgsi/exec/tgsi_exec.h" /* diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 1597361b82..dde3fabc81 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -34,7 +34,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "pipe/util/p_tile.h" +#include "util/p_tile.h" #include "sp_context.h" #include "sp_surface.h" #include "sp_tile_cache.h" diff --git a/src/gallium/winsys/dri/Makefile b/src/gallium/winsys/dri/Makefile new file mode 100644 index 0000000000..f466ce6c3c --- /dev/null +++ b/src/gallium/winsys/dri/Makefile @@ -0,0 +1,38 @@ +# src/mesa/drivers/dri/Makefile + +TOP = ../../../.. + +include $(TOP)/configs/current + + + +default: $(TOP)/$(LIB_DIR) subdirs + + +$(TOP)/$(LIB_DIR): + -mkdir $(TOP)/$(LIB_DIR) + + +subdirs: + @for dir in $(DRI_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +install: + @for dir in $(DRI_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) install) || exit 1 ; \ + fi \ + done + + +clean: + @for dir in $(DRI_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) clean) ; \ + fi \ + done + -rm -f common/*.o diff --git a/src/gallium/winsys/dri/Makefile.template b/src/gallium/winsys/dri/Makefile.template new file mode 100644 index 0000000000..b96305c094 --- /dev/null +++ b/src/gallium/winsys/dri/Makefile.template @@ -0,0 +1,113 @@ +# -*-makefile-*- + +MESA_MODULES = $(TOP)/src/mesa/libmesa.a + +COMMON_GALLIUM_SOURCES = \ + $(TOP)/src/mesa/drivers/dri/common/utils.c \ + $(TOP)/src/mesa/drivers/dri/common/vblank.c \ + $(TOP)/src/mesa/drivers/dri/common/dri_util.c \ + $(TOP)/src/mesa/drivers/dri/common/xmlconfig.c + +COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \ + $(TOP)/src/mesa/drivers/common/driverfuncs.c \ + $(TOP)/src/mesa/drivers/dri/common/texmem.c \ + $(TOP)/src/mesa/drivers/dri/common/drirenderbuffer.c + +COMMON_BM_SOURCES = \ + $(TOP)/src/mesa/drivers/dri/common/dri_bufmgr.c \ + $(TOP)/src/mesa/drivers/dri/common/dri_drmpool.c + + +ifeq ($(WINDOW_SYSTEM),dri) +WINOBJ= +WINLIB= +INCLUDES = $(SHARED_INCLUDES) $(EXPAT_INCLUDES) + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(ASM_SOURCES:.S=.o) + +else +# miniglx +WINOBJ= +WINLIB=-L$(MESA)/src/glx/mini +MINIGLX_INCLUDES = -I$(TOP)/src/glx/mini +INCLUDES = $(MINIGLX_INCLUDES) \ + $(SHARED_INCLUDES) \ + $(PCIACCESS_CFLAGS) + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(MINIGLX_SOURCES:.c=.o) \ + $(ASM_SOURCES:.S=.o) +endif + + +### Include directories +SHARED_INCLUDES = \ + -I. \ + -I$(TOP)/src/mesa/drivers/dri/common \ + -Iserver \ + -I$(TOP)/include \ + -I$(TOP)/include/GL/internal \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main \ + -I$(TOP)/src/mesa/glapi \ + -I$(TOP)/src/mesa/math \ + -I$(TOP)/src/mesa/transform \ + -I$(TOP)/src/mesa/shader \ + -I$(TOP)/src/mesa/swrast \ + -I$(TOP)/src/mesa/swrast_setup \ + -I$(TOP)/src/egl/main \ + -I$(TOP)/src/egl/drivers/dri \ + $(LIBDRM_CFLAGS) + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + + +##### TARGETS ##### + +default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) + + +$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template + $(TOP)/bin/mklib -noprefix -o $@ \ + $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) + + +$(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) + $(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR) + + +depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ + $(ASM_SOURCES) 2> /dev/null + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean: + -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) + -rm -f depend depend.bak + + +install: $(LIBNAME) + $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) + $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) + + +include depend diff --git a/src/gallium/winsys/dri/intel/Makefile b/src/gallium/winsys/dri/intel/Makefile index 9ae0f01325..40654bb2ac 100644 --- a/src/gallium/winsys/dri/intel/Makefile +++ b/src/gallium/winsys/dri/intel/Makefile @@ -7,8 +7,8 @@ LIBNAME = i915tex_dri.so MINIGLX_SOURCES = server/intel_dri.c PIPE_DRIVERS = \ - $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a \ - $(TOP)/src/mesa/pipe/i915simple/libi915simple.a + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a DRIVER_SOURCES = \ intel_winsys_pipe.c \ @@ -28,11 +28,11 @@ C_SOURCES = \ ASM_SOURCES = -DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ +DRIVER_DEFINES = -I$(TOP)/src/mesa/drivers/dri/intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") include ../Makefile.template -intel_tex_layout.o: ../intel/intel_tex_layout.c +intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c symlinks: diff --git a/src/gallium/winsys/dri/intel/intel_winsys_i915.c b/src/gallium/winsys/dri/intel/intel_winsys_i915.c index 1ba6a9e1b2..0ed3890e93 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_i915.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_i915.c @@ -39,7 +39,7 @@ #include "intel_winsys.h" #include "pipe/p_util.h" -#include "pipe/i915simple/i915_winsys.h" +#include "i915simple/i915_winsys.h" struct intel_i915_winsys { diff --git a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c index cec3437831..9e483bdc9f 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c @@ -34,7 +34,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "pipe/p_format.h" -#include "pipe/softpipe/sp_winsys.h" +#include "softpipe/sp_winsys.h" struct intel_softpipe_winsys { diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index c3cd22eea3..8da596d419 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -41,11 +41,11 @@ #include "pipe/p_context.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "pipe/softpipe/sp_winsys.h" +#include "softpipe/sp_winsys.h" #ifdef GALLIUM_CELL -#include "pipe/cell/ppu/cell_context.h" -#include "pipe/cell/ppu/cell_winsys.h" +#include "cell/ppu/cell_context.h" +#include "cell/ppu/cell_winsys.h" #else #define TILE_SIZE 32 /* avoid compilation errors */ #endif diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c index bf41570257..dbfd37bda2 100644 --- a/src/gallium/winsys/xlib/xm_winsys_aub.c +++ b/src/gallium/winsys/xlib/xm_winsys_aub.c @@ -39,7 +39,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "pipe/i965simple/brw_winsys.h" +#include "i965simple/brw_winsys.h" #include "brw_aub.h" #include "xm_winsys_aub.h" diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 720f1b2e02..561608fedd 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -12,16 +12,16 @@ GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) PIPE_LIB = \ - $(TOP)/src/mesa/pipe/softpipe/libsoftpipe.a \ - $(TOP)/src/mesa/pipe/i965simple/libi965simple.a + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/i965simple/libi965simple.a ifeq ($(CONFIG_NAME), linux-cell) -CELL_LIB = $(TOP)/src/mesa/pipe/cell/ppu/libcell.a -CELL_LIB_SPU = $(TOP)/src/mesa/pipe/cell/spu/g3d_spu.a +CELL_LIB = $(TOP)/src/gallium/drivers/cell/ppu/libcell.a +CELL_LIB_SPU = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a endif ifeq ($(CONFIG_NAME), linux-llvm) -LLVM_LIB = $(TOP)/src/mesa/pipe/llvm/libgallivm.a +LLVM_LIB = $(TOP)/src/gallium/aux/llvm/libgallivm.a endif .SUFFIXES : .cpp @@ -71,7 +71,7 @@ libmesa.a: $(SOLO_OBJECTS) fi linux-solo: depend subdirs libmesa.a - cd drivers/dri ; $(MAKE) + cd $(TOP)/src/gallium/winsys/dri ; $(MAKE) ##################################################################### @@ -165,7 +165,6 @@ depend: $(ALL_SOURCES) subdirs: @ (cd x86 ; $(MAKE)) @ (cd x86-64 ; $(MAKE)) - (cd pipe ; $(MAKE)) install: default $(INSTALL) -d $(INSTALL_DIR)/include/GL @@ -178,7 +177,7 @@ install: default $(INSTALL) $(TOP)/$(LIB_DIR)/libOSMesa* $(INSTALL_DIR)/$(LIB_DIR); \ fi @if [ "${DRIVER_DIRS}" = "dri" ] ; then \ - cd drivers/dri ; $(MAKE) install ; \ + cd $(TOP)/gallium/winsys/dri ; $(MAKE) install ; \ fi ## NOT INSTALLED YET: @@ -198,7 +197,6 @@ clean: (cd drivers/dri && $(MAKE) clean) (cd x86 && $(MAKE) clean) (cd x86-64 && $(MAKE) clean) - (cd pipe ; $(MAKE) clean ) include depend diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c index 08c98eab48..18b033666f 100644 --- a/src/mesa/drivers/x11/xm_api.c +++ b/src/mesa/drivers/x11/xm_api.c @@ -85,7 +85,7 @@ #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" -#include "pipe/softpipe/sp_context.h" +#include "softpipe/sp_context.h" #include "pipe/p_defines.h" /** diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c index 8ae243ae66..34287effe1 100644 --- a/src/mesa/drivers/x11/xm_dd.c +++ b/src/mesa/drivers/x11/xm_dd.c @@ -53,7 +53,7 @@ #include "tnl/tnl.h" #include "tnl/t_context.h" -#include "pipe/softpipe/sp_context.h" +#include "softpipe/sp_context.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" #include "state_tracker/st_draw.h" diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 5533158ece..81616b92d9 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -45,10 +45,10 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/softpipe/sp_context.h" -#include "pipe/softpipe/sp_clear.h" -#include "pipe/softpipe/sp_tile_cache.h" -#include "pipe/softpipe/sp_surface.h" +#include "softpipe/sp_context.h" +#include "softpipe/sp_clear.h" +#include "softpipe/sp_tile_cache.h" +#include "softpipe/sp_surface.h" #include "state_tracker/st_context.h" diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index a690df2772..2edc697693 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -38,7 +38,7 @@ #include "main/macros.h" #include "pipe/p_winsys.h" -#include "pipe/softpipe/sp_winsys.h" +#include "softpipe/sp_winsys.h" /** diff --git a/src/mesa/drivers/x11/xmesaP.h b/src/mesa/drivers/x11/xmesaP.h index 4709d63394..fd2dfcd79a 100644 --- a/src/mesa/drivers/x11/xmesaP.h +++ b/src/mesa/drivers/x11/xmesaP.h @@ -37,8 +37,8 @@ #include "xm_image.h" #endif #include "state_tracker/st_cb_fbo.h" -#include "pipe/softpipe/sp_context.h" -#include "pipe/softpipe/sp_surface.h" +#include "softpipe/sp_context.h" +#include "softpipe/sp_surface.h" extern _glthread_Mutex _xmesa_lock; diff --git a/src/mesa/sources b/src/mesa/sources index 1165425183..2d07738210 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -158,45 +158,45 @@ VF_SOURCES = \ DRAW_SOURCES = \ - pipe/draw/draw_clip.c \ - pipe/draw/draw_context.c\ - pipe/draw/draw_cull.c \ - pipe/draw/draw_debug.c \ - pipe/draw/draw_flatshade.c \ - pipe/draw/draw_offset.c \ - pipe/draw/draw_prim.c \ - pipe/draw/draw_stipple.c \ - pipe/draw/draw_twoside.c \ - pipe/draw/draw_unfilled.c \ - pipe/draw/draw_validate.c \ - pipe/draw/draw_vbuf.c \ - pipe/draw/draw_vertex.c \ - pipe/draw/draw_vertex_cache.c \ - pipe/draw/draw_vertex_fetch.c \ - pipe/draw/draw_vertex_shader.c \ - pipe/draw/draw_vf.c \ - pipe/draw/draw_vf_generic.c \ - pipe/draw/draw_vf_sse.c \ - pipe/draw/draw_wide_prims.c + $(TOP)/src/gallium/aux/draw/draw_clip.c \ + $(TOP)/src/gallium/aux/draw/draw_context.c\ + $(TOP)/src/gallium/aux/draw/draw_cull.c \ + $(TOP)/src/gallium/aux/draw/draw_debug.c \ + $(TOP)/src/gallium/aux/draw/draw_flatshade.c \ + $(TOP)/src/gallium/aux/draw/draw_offset.c \ + $(TOP)/src/gallium/aux/draw/draw_prim.c \ + $(TOP)/src/gallium/aux/draw/draw_stipple.c \ + $(TOP)/src/gallium/aux/draw/draw_twoside.c \ + $(TOP)/src/gallium/aux/draw/draw_unfilled.c \ + $(TOP)/src/gallium/aux/draw/draw_validate.c \ + $(TOP)/src/gallium/aux/draw/draw_vbuf.c \ + $(TOP)/src/gallium/aux/draw/draw_vertex.c \ + $(TOP)/src/gallium/aux/draw/draw_vertex_cache.c \ + $(TOP)/src/gallium/aux/draw/draw_vertex_fetch.c \ + $(TOP)/src/gallium/aux/draw/draw_vertex_shader.c \ + $(TOP)/src/gallium/aux/draw/draw_vf.c \ + $(TOP)/src/gallium/aux/draw/draw_vf_generic.c \ + $(TOP)/src/gallium/aux/draw/draw_vf_sse.c \ + $(TOP)/src/gallium/aux/draw/draw_wide_prims.c TGSIEXEC_SOURCES = \ - pipe/tgsi/exec/tgsi_exec.c \ - pipe/tgsi/exec/tgsi_sse2.c + $(TOP)/src/gallium/aux/tgsi/exec/tgsi_exec.c \ + $(TOP)/src/gallium/aux/tgsi/exec/tgsi_sse2.c TGSIUTIL_SOURCES = \ - pipe/tgsi/util/tgsi_build.c \ - pipe/tgsi/util/tgsi_dump.c \ - pipe/tgsi/util/tgsi_parse.c \ - pipe/tgsi/util/tgsi_util.c + $(TOP)/src/gallium/aux/tgsi/util/tgsi_build.c \ + $(TOP)/src/gallium/aux/tgsi/util/tgsi_dump.c \ + $(TOP)/src/gallium/aux/tgsi/util/tgsi_parse.c \ + $(TOP)/src/gallium/aux/tgsi/util/tgsi_util.c STATECACHE_SOURCES = \ - pipe/cso_cache/cso_hash.c \ - pipe/cso_cache/cso_cache.c + $(TOP)/src/gallium/aux/cso_cache/cso_hash.c \ + $(TOP)/src/gallium/aux/cso_cache/cso_cache.c PIPEUTIL_SOURCES = \ - pipe/util/p_debug.c \ - pipe/util/p_tile.c \ - pipe/util/p_util.c + $(TOP)/src/gallium/aux/util/p_debug.c \ + $(TOP)/src/gallium/aux/util/p_tile.c \ + $(TOP)/src/gallium/aux/util/p_util.c STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ @@ -331,13 +331,13 @@ __COMMON_DRIVER_SOURCES = \ drivers/common/driverfuncs.c X11_DRIVER_SOURCES = \ - pipe/xlib/glxapi.c \ - pipe/xlib/fakeglx.c \ - pipe/xlib/xfonts.c \ - pipe/xlib/xm_api.c \ - pipe/xlib/xm_winsys.c \ - pipe/xlib/xm_winsys_aub.c \ - pipe/xlib/brw_aub.c + $(TOP)/src/gallium/winsys/xlib/glxapi.c \ + $(TOP)/src/gallium/winsys/xlib/fakeglx.c \ + $(TOP)/src/gallium/winsys/xlib/xfonts.c \ + $(TOP)/src/gallium/winsys/xlib/xm_api.c \ + $(TOP)/src/gallium/winsys/xlib/xm_winsys.c \ + $(TOP)/src/gallium/winsys/xlib/xm_winsys_aub.c \ + $(TOP)/src/gallium/winsys/xlib/brw_aub.c OSMESA_DRIVER_SOURCES = \ drivers/osmesa/osmesa.c @@ -425,7 +425,10 @@ FBDEV_DRIVER_OBJECTS = $(FBDEV_DRIVER_SOURCES:.c=.o) INCLUDE_DIRS = \ -I$(TOP)/include \ -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/main + -I$(TOP)/src/mesa/main \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/aux OLD_INCLUDE_DIRS = \ -I$(TOP)/src/mesa/tnl \ @@ -435,4 +438,4 @@ OLD_INCLUDE_DIRS = \ -I$(TOP)/src/mesa/shader \ -I$(TOP)/src/mesa/shader/grammar \ -I$(TOP)/src/mesa/shader/slang \ - -I$(TOP)/src/mesa/pipe/tgsi + -I$(TOP)/s$(TOP)/src/gallium/aux/tgsi diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 2c6ec8421b..b67b620eaa 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -43,7 +43,7 @@ #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" -#include "pipe/cso_cache/cso_cache.h" +#include "cso_cache/cso_cache.h" #include "st_context.h" #include "st_cache.h" diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index e0965b217a..2979e7fae5 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -36,8 +36,8 @@ #include "pipe/p_state.h" -#include "pipe/cso_cache/cso_cache.h" -#include "pipe/cso_cache/cso_hash.h" +#include "cso_cache/cso_cache.h" +#include "cso_cache/cso_hash.h" /* Those function will either find the state of the given template diff --git a/src/mesa/state_tracker/st_cache.h b/src/mesa/state_tracker/st_cache.h index e0c176b0ff..b81de316ec 100644 --- a/src/mesa/state_tracker/st_cache.h +++ b/src/mesa/state_tracker/st_cache.h @@ -33,7 +33,7 @@ #ifndef ST_CACHE_H #define ST_CACHE_H -#include "pipe/cso_cache/cso_cache.h" +#include "cso_cache/cso_cache.h" struct pipe_blend_state; struct pipe_sampler_state; diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 3a3bf9016d..663c4f205d 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -43,7 +43,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/util/p_tile.h" +#include "util/p_tile.h" #define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index f13199a3c0..e2d4e06da1 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -56,7 +56,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "pipe/util/p_tile.h" +#include "util/p_tile.h" #include "shader/prog_instruction.h" diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 31744151f1..5315294c07 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -53,10 +53,10 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/cso_cache/cso_cache.h" +#include "cso_cache/cso_cache.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" /** diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index af3ee65504..61d4f4c41c 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -39,8 +39,8 @@ #include "shader/programopt.h" #include "shader/shader_api.h" -#include "pipe/cso_cache/cso_cache.h" -#include "pipe/draw/draw_context.h" +#include "cso_cache/cso_cache.h" +#include "draw/draw_context.h" #include "st_context.h" #include "st_program.h" diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 7e347c4893..5b0eb6022b 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -44,8 +44,8 @@ #include "st_draw.h" #include "st_cb_rasterpos.h" #include "st_draw.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" #include "shader/prog_instruction.h" #include "vbo/vbo.h" diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 868c5f3c5f..c89c74229e 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -40,7 +40,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/util/p_tile.h" +#include "util/p_tile.h" #include "st_context.h" #include "st_cb_readpixels.h" #include "st_cb_fbo.h" diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 91a40288cc..03dbb30b0f 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -47,7 +47,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/util/p_tile.h" +#include "util/p_tile.h" #define DBG if (0) printf diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index bf4618bed8..09e389f9dc 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -54,8 +54,8 @@ #include "pipe/p_context.h" #include "pipe/p_winsys.h" #include "pipe/p_inlines.h" -#include "pipe/draw/draw_context.h" -#include "pipe/cso_cache/cso_cache.h" +#include "draw/draw_context.h" +#include "cso_cache/cso_cache.h" /** diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index 57450e52bf..5888bcb98a 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -31,9 +31,9 @@ #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_dump.h" +#include "tgsi/util/tgsi_dump.h" -#include "pipe/cso_cache/cso_cache.h" +#include "cso_cache/cso_cache.h" #include "st_context.h" #include "st_debug.h" diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index ae9f5c8b11..1c0fa8c6aa 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -47,8 +47,8 @@ #include "pipe/p_winsys.h" #include "pipe/p_inlines.h" -#include "pipe/draw/draw_private.h" -#include "pipe/draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_context.h" static GLuint double_types[4] = { diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 459941cca8..6c09b86033 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -37,7 +37,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "pipe/cso_cache/cso_cache.h" +#include "cso_cache/cso_cache.h" #include "st_context.h" #include "st_draw.h" diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 325aa20173..97206752af 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -32,9 +32,9 @@ #include "pipe/p_compiler.h" #include "pipe/p_shader_tokens.h" -#include "pipe/tgsi/util/tgsi_parse.h" -#include "pipe/tgsi/util/tgsi_build.h" -#include "pipe/tgsi/util/tgsi_util.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_build.h" +#include "tgsi/util/tgsi_util.h" #include "st_mesa_to_tgsi.h" #include "shader/prog_instruction.h" #include "shader/prog_parameter.h" diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index c8297baded..dc992ee9c2 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -38,8 +38,8 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" -#include "pipe/draw/draw_context.h" -#include "pipe/tgsi/util/tgsi_dump.h" +#include "draw/draw_context.h" +#include "tgsi/util/tgsi_dump.h" #include "st_context.h" #include "st_cache.h" -- cgit v1.2.3 From 6d3831b11d9f5aaba61cc2fb8ade61437ad7c335 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Feb 2008 17:52:14 +0900 Subject: Code reorganization: placeholder for state-trackers. This is meant for temporarily holding state-trackers, until they eventually find their way out of gallium tree. --- src/gallium/state_trackers/README | 2 ++ 1 file changed, 2 insertions(+) create mode 100644 src/gallium/state_trackers/README (limited to 'src') diff --git a/src/gallium/state_trackers/README b/src/gallium/state_trackers/README new file mode 100644 index 0000000000..28dd27bbd5 --- /dev/null +++ b/src/gallium/state_trackers/README @@ -0,0 +1,2 @@ +This directory is a placeholder for incubating state-trackers. Mesa's +state-tracker is in src/mesa. -- cgit v1.2.3 From fa9c160389ffc6d7a20773b77c937193f30339d8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 15 Feb 2008 08:56:04 +0000 Subject: tgsi: disable dummy sse2 texture code --- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 21 ++++++++++++++------- 1 file changed, 14 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 1d9e9a14ce..6df7588c92 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -1935,13 +1935,20 @@ emit_instruction( break; case TGSI_OPCODE_TEX: - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); + if (0) { + /* Disable dummy texture code: + */ + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + } + else { + return 0; } break; -- cgit v1.2.3 From c04a7f8929d674971a472ffa4d3a31200c22aa5a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 15 Feb 2008 09:31:22 +0000 Subject: gallium: reorganize fragment shader execution, unbreak sse This is probably going to get further reworked in the near future. Right now there's a new interface wrapped around each shader execution mode - exec, sse2, llvm. The llvm code was disabled already and has just been moved as-is to a new file, whereas the sse2 and exec code is actually enabled. The way the interfaces has turned out suggests to me that the correct approach is to actually have each shader include a pointer to a quad stage which will do a better job of encapsulating the execution environment than what I have here -- that's a second step however. --- src/mesa/pipe/draw/draw_vertex_shader.c | 10 +- src/mesa/pipe/softpipe/Makefile | 3 + src/mesa/pipe/softpipe/sp_context.h | 8 +- src/mesa/pipe/softpipe/sp_fs_exec.c | 112 ++++++++++++++++++ src/mesa/pipe/softpipe/sp_fs_llvm.c | 200 ++++++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_fs_sse.c | 192 ++++++++++++++++++++++++++++++ src/mesa/pipe/softpipe/sp_quad_fs.c | 200 ++------------------------------ src/mesa/pipe/softpipe/sp_state.h | 37 +++--- src/mesa/pipe/softpipe/sp_state_fs.c | 69 ++++------- src/mesa/pipe/tgsi/exec/tgsi_exec.c | 71 +++++------- src/mesa/pipe/tgsi/exec/tgsi_exec.h | 8 +- src/mesa/pipe/tgsi/exec/tgsi_sse2.c | 10 +- 12 files changed, 613 insertions(+), 307 deletions(-) create mode 100644 src/mesa/pipe/softpipe/sp_fs_exec.c create mode 100644 src/mesa/pipe/softpipe/sp_fs_llvm.c create mode 100644 src/mesa/pipe/softpipe/sp_fs_sse.c (limited to 'src') diff --git a/src/mesa/pipe/draw/draw_vertex_shader.c b/src/mesa/pipe/draw/draw_vertex_shader.c index c824c1407e..bf340ef9b0 100644 --- a/src/mesa/pipe/draw/draw_vertex_shader.c +++ b/src/mesa/pipe/draw/draw_vertex_shader.c @@ -305,11 +305,13 @@ draw_bind_vertex_shader(struct draw_context *draw, draw->vertex_shader = dvs; draw->num_vs_outputs = dvs->state->num_outputs; + tgsi_exec_machine_init(&draw->machine); + /* specify the vertex program to interpret/execute */ - tgsi_exec_machine_init(&draw->machine, - draw->vertex_shader->state->tokens, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/ ); + tgsi_exec_machine_bind_shader(&draw->machine, + draw->vertex_shader->state->tokens, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/ ); } diff --git a/src/mesa/pipe/softpipe/Makefile b/src/mesa/pipe/softpipe/Makefile index 31438a882e..304632abb2 100644 --- a/src/mesa/pipe/softpipe/Makefile +++ b/src/mesa/pipe/softpipe/Makefile @@ -5,6 +5,9 @@ include $(TOP)/configs/current LIBNAME = softpipe DRIVER_SOURCES = \ + sp_fs_exec.c \ + sp_fs_sse.c \ + sp_fs_llvm.c \ sp_clear.c \ sp_flush.c \ sp_query.c \ diff --git a/src/mesa/pipe/softpipe/sp_context.h b/src/mesa/pipe/softpipe/sp_context.h index aff8c2cc5d..5e24baf938 100644 --- a/src/mesa/pipe/softpipe/sp_context.h +++ b/src/mesa/pipe/softpipe/sp_context.h @@ -44,8 +44,8 @@ struct softpipe_vbuf_render; struct draw_context; struct draw_stage; struct softpipe_tile_cache; -struct sp_fragment_shader_state; -struct sp_vertex_shader_state; +struct sp_fragment_shader; +struct sp_vertex_shader; struct softpipe_context { @@ -59,8 +59,8 @@ struct softpipe_context { const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct pipe_depth_stencil_alpha_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; - const struct sp_fragment_shader_state *fs; - const struct sp_vertex_shader_state *vs; + const struct sp_fragment_shader *fs; + const struct sp_vertex_shader *vs; struct pipe_blend_color blend_color; struct pipe_clip_state clip; diff --git a/src/mesa/pipe/softpipe/sp_fs_exec.c b/src/mesa/pipe/softpipe/sp_fs_exec.c new file mode 100644 index 0000000000..b949492e09 --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_fs_exec.c @@ -0,0 +1,112 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_fs.h" +#include "sp_headers.h" + + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/tgsi/exec/tgsi_exec.h" + +struct sp_exec_fragment_shader { + struct sp_fragment_shader base; +}; + + + + +static void +exec_prepare( struct sp_fragment_shader *base, + struct tgsi_exec_machine *machine, + struct tgsi_sampler *samplers ) +{ + tgsi_exec_machine_bind_shader( machine, + base->shader.tokens, + PIPE_MAX_SAMPLERS, + samplers ); +} + + + + +/* TODO: hide the machine struct in here somewhere, remove from this + * interface: + */ +static unsigned +exec_run( struct sp_fragment_shader *base, + struct tgsi_exec_machine *machine, + struct quad_header *quad ) +{ + + /* Compute X, Y, Z, W vals for this quad */ + sp_setup_pos_vector(quad->posCoef, + (float)quad->x0, (float)quad->y0, + &machine->QuadPos); + + return tgsi_exec_machine_run( machine ); +} + + + +static void +exec_delete( struct sp_fragment_shader *base ) +{ + FREE(base); +} + + + + + +struct sp_fragment_shader * +softpipe_create_fs_exec(struct softpipe_context *softpipe, + const struct pipe_shader_state *templ) +{ + struct sp_exec_fragment_shader *shader; + + /* Decide whether we'll be codegenerating this shader and if so do + * that now. + */ + + shader = CALLOC_STRUCT(sp_exec_fragment_shader); + if (!shader) + return NULL; + + shader->base.shader = *templ; + shader->base.prepare = exec_prepare; + shader->base.run = exec_run; + shader->base.delete = exec_delete; + + return &shader->base; +} + diff --git a/src/mesa/pipe/softpipe/sp_fs_llvm.c b/src/mesa/pipe/softpipe/sp_fs_llvm.c new file mode 100644 index 0000000000..9237c1fbda --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_fs_llvm.c @@ -0,0 +1,200 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: + * Zack Rusin + */ + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_fs.h" + + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/tgsi/exec/tgsi_sse2.h" + +#if 0 + +struct sp_llvm_fragment_shader { + struct sp_fragment_shader base; + struct gallivm_prog *llvm_prog; +}; + +static void +shade_quad_llvm(struct quad_stage *qs, + struct quad_header *quad) +{ + struct quad_shade_stage *qss = quad_shade_stage(qs); + struct softpipe_context *softpipe = qs->softpipe; + float dests[4][16][4] ALIGN16_ATTRIB; + float inputs[4][16][4] ALIGN16_ATTRIB; + const float fx = (float) quad->x0; + const float fy = (float) quad->y0; + struct gallivm_prog *llvm = qss->llvm_prog; + + inputs[0][0][0] = fx; + inputs[1][0][0] = fx + 1.0f; + inputs[2][0][0] = fx; + inputs[3][0][0] = fx + 1.0f; + + inputs[0][0][1] = fy; + inputs[1][0][1] = fy; + inputs[2][0][1] = fy + 1.0f; + inputs[3][0][1] = fy + 1.0f; + + + gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef); + +#if DLLVM + debug_printf("MASK = %d\n", quad->mask); + for (int i = 0; i < 4; ++i) { + for (int j = 0; j < 2; ++j) { + debug_printf("IN(%d,%d) [%f %f %f %f]\n", i, j, + inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]); + } + } +#endif + + quad->mask &= + gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs, + softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], + qss->samplers); +#if DLLVM + debug_printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n", + dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], + dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]); +#endif + + /* store result color */ + if (qss->colorOutSlot >= 0) { + unsigned i; + /* XXX need to handle multiple color outputs someday */ + allvmrt(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] + == TGSI_SEMANTIC_COLOR); + for (i = 0; i < QUAD_SIZE; ++i) { + quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0]; + quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1]; + quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2]; + quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3]; + } + } +#if DLLVM + for (int i = 0; i < QUAD_SIZE; ++i) { + debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, + quad->outputs.color[0][i], + quad->outputs.color[1][i], + quad->outputs.color[2][i], + quad->outputs.color[3][i]); + } +#endif + + /* store result Z */ + if (qss->depthOutSlot >= 0) { + /* output[slot] is new Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = dests[i][0][2]; + } + } + else { + /* copy input Z (which was interpolated by the executor) to output Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = inputs[i][0][2]; + } + } +#if DLLVM + debug_printf("D [%f, %f, %f, %f] mask = %d\n", + quad->outputs.depth[0], + quad->outputs.depth[1], + quad->outputs.depth[2], + quad->outputs.depth[3], quad->mask); +#endif + + /* shader may cull fragments */ + if( quad->mask ) { + qs->next->run( qs->next, quad ); + } +} + + +unsigned +run_llvm_fs( struct sp_fragment_shader *base, + struct foo *machine ) +{ +} + + +void +delete_llvm_fs( struct sp_fragment_shader *base ) +{ + FREE(base); +} + + +struct sp_fragment_shader * +softpipe_create_fs_llvm(struct softpipe_context *softpipe, + const struct pipe_shader_state *templ) +{ + struct sp_llvm_fragment_shader *shader = NULL; + + /* LLVM fragment shaders currently disabled: + */ + state = CALLOC_STRUCT(sp_llvm_shader_state); + if (!state) + return NULL; + + state->llvm_prog = 0; + + if (!gallivm_global_cpu_engine()) { + gallivm_cpu_engine_create(state->llvm_prog); + } + else + gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); + + if (shader) { + shader->base.run = run_llvm_fs; + shader->base.delete = delete_llvm_fs; + } + + return shader; +} + + +#else + +struct sp_fragment_shader * +softpipe_create_fs_llvm(struct softpipe_context *softpipe, + const struct pipe_shader_state *templ) +{ + return NULL; +} + +#endif diff --git a/src/mesa/pipe/softpipe/sp_fs_sse.c b/src/mesa/pipe/softpipe/sp_fs_sse.c new file mode 100644 index 0000000000..713ece369e --- /dev/null +++ b/src/mesa/pipe/softpipe/sp_fs_sse.c @@ -0,0 +1,192 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_fs.h" +#include "sp_headers.h" + + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/tgsi/exec/tgsi_exec.h" +#include "pipe/tgsi/exec/tgsi_sse2.h" + + +#if defined(__i386__) || defined(__386__) + +#include "x86/rtasm/x86sse.h" + +/* Surely this should be defined somewhere in a tgsi header: + */ +typedef void (XSTDCALL *codegen_function)( + const struct tgsi_exec_vector *input, + struct tgsi_exec_vector *output, + float (*constant)[4], + struct tgsi_exec_vector *temporary, + const struct tgsi_interp_coef *coef + //, const struct tgsi_exec_vector *quadPos + ); + + +struct sp_sse_fragment_shader { + struct sp_fragment_shader base; + struct x86_function sse2_program; + codegen_function func; +}; + + +/** + * Compute quad X,Y,Z,W for the four fragments in a quad. + * + * This should really be part of the compiled shader. + */ +void +sp_setup_pos_vector(const struct tgsi_interp_coef *coef, + float x, float y, + struct tgsi_exec_vector *quadpos) +{ + uint chan; + /* do X */ + quadpos->xyzw[0].f[0] = x; + quadpos->xyzw[0].f[1] = x + 1; + quadpos->xyzw[0].f[2] = x; + quadpos->xyzw[0].f[3] = x + 1; + + /* do Y */ + quadpos->xyzw[1].f[0] = y; + quadpos->xyzw[1].f[1] = y; + quadpos->xyzw[1].f[2] = y + 1; + quadpos->xyzw[1].f[3] = y + 1; + + /* do Z and W for all fragments in the quad */ + for (chan = 2; chan < 4; chan++) { + const float dadx = coef->dadx[chan]; + const float dady = coef->dady[chan]; + const float a0 = coef->a0[chan] + dadx * x + dady * y; + quadpos->xyzw[chan].f[0] = a0; + quadpos->xyzw[chan].f[1] = a0 + dadx; + quadpos->xyzw[chan].f[2] = a0 + dady; + quadpos->xyzw[chan].f[3] = a0 + dadx + dady; + } +} + + +static void +sse_prepare( struct sp_fragment_shader *base, + struct tgsi_exec_machine *machine, + struct tgsi_sampler *samplers ) +{ +} + + +/* TODO: codegenerate the whole run function, skip this wrapper. + * TODO: break dependency on tgsi_exec_machine struct + * TODO: push Position calculation into the generated shader + * TODO: process >1 quad at a time + */ +static unsigned +sse_run( struct sp_fragment_shader *base, + struct tgsi_exec_machine *machine, + struct quad_header *quad ) +{ + struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base; + + /* Compute X, Y, Z, W vals for this quad -- place in temp[0] for now */ + sp_setup_pos_vector(quad->posCoef, + (float)quad->x0, (float)quad->y0, + machine->Temps); + + shader->func( machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps, + machine->InterpCoefs + // , &machine->QuadPos + ); + + return ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]); +} + + +static void +sse_delete( struct sp_fragment_shader *base ) +{ + struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base; + + x86_release_func( &shader->sse2_program ); + FREE(shader); +} + + +struct sp_fragment_shader * +softpipe_create_fs_sse(struct softpipe_context *softpipe, + const struct pipe_shader_state *templ) +{ + struct sp_sse_fragment_shader *shader; + + if (!softpipe->use_sse) + return NULL; + + shader = CALLOC_STRUCT(sp_sse_fragment_shader); + if (!shader) + return NULL; + + x86_init_func( &shader->sse2_program ); + + if (!tgsi_emit_sse2_fs( templ->tokens, &shader->sse2_program )) { + FREE(shader); + return NULL; + } + + shader->func = (codegen_function) x86_get_func( &shader->sse2_program ); + assert(shader->func); + + shader->base.shader = *templ; + shader->base.prepare = sse_prepare; + shader->base.run = sse_run; + shader->base.delete = sse_delete; + + return &shader->base; +} + + +#else + +/* Maybe put this varient in the header file. + */ +struct sp_fragment_shader * +softpipe_create_fs_sse(struct softpipe_context *softpipe, + const struct pipe_shader_state *templ) +{ + return NULL; +} + +#endif diff --git a/src/mesa/pipe/softpipe/sp_quad_fs.c b/src/mesa/pipe/softpipe/sp_quad_fs.c index 3316858413..cf1b1eff75 100644 --- a/src/mesa/pipe/softpipe/sp_quad_fs.c +++ b/src/mesa/pipe/softpipe/sp_quad_fs.c @@ -39,12 +39,6 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" -#include "x86/rtasm/x86sse.h" - -#ifdef MESA_LLVM -#include "pipe/llvm/gallivm.h" -#endif - #include "sp_context.h" #include "sp_state.h" #include "sp_headers.h" @@ -60,9 +54,6 @@ struct quad_shade_stage struct tgsi_exec_machine machine; struct tgsi_exec_vector *inputs, *outputs; int colorOutSlot, depthOutSlot; -#ifdef MESA_LLVM - struct gallivm_prog *llvm_prog; -#endif }; @@ -74,46 +65,6 @@ quad_shade_stage(struct quad_stage *qs) } -/** - * Compute quad X,Y,Z,W for the four fragments in a quad. - * Note that we only need to "compute" X and Y for the upper-left fragment. - * We could do less work if we're not depth testing, or there's no - * perspective-corrected attributes, but that's seldom. - */ -static void -setup_pos_vector(const struct tgsi_interp_coef *coef, - float x, float y, - struct tgsi_exec_vector *quadpos) -{ - uint chan; - /* do X */ - quadpos->xyzw[0].f[0] = x; - /* do Y */ - quadpos->xyzw[1].f[0] = y; - /* do Z and W for all fragments in the quad */ - for (chan = 2; chan < 4; chan++) { - const float dadx = coef->dadx[chan]; - const float dady = coef->dady[chan]; - const float a0 = coef->a0[chan] + dadx * x + dady * y; - quadpos->xyzw[chan].f[0] = a0; - quadpos->xyzw[chan].f[1] = a0 + dadx; - quadpos->xyzw[chan].f[2] = a0 + dady; - quadpos->xyzw[chan].f[3] = a0 + dadx + dady; - } -} - - -typedef void (XSTDCALL *codegen_function)( - const struct tgsi_exec_vector *input, - struct tgsi_exec_vector *output, - float (*constant)[4], - struct tgsi_exec_vector *temporary, - const struct tgsi_interp_coef *coef -#if 0 - ,const struct tgsi_exec_vector *quadPos -#endif - ); - /** * Execute fragment shader for the four fragments in the quad. @@ -132,30 +83,10 @@ shade_quad( machine->InterpCoefs = quad->coef; - /* Compute X, Y, Z, W vals for this quad */ - setup_pos_vector(quad->posCoef, (float) quad->x0, (float) quad->y0, &machine->QuadPos); - /* run shader */ -#if defined(__i386__) || defined(__386__) - if( softpipe->use_sse ) { - codegen_function func = (codegen_function) x86_get_func( &softpipe->fs->sse2_program ); - func( - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps, - machine->InterpCoefs -#if 0 - ,machine->QuadPos -#endif - ); - quad->mask &= ~(machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0]); - } - else -#endif - { - quad->mask &= tgsi_exec_machine_run( machine ); - } + quad->mask &= softpipe->fs->run( softpipe->fs, + &qss->machine, + quad ); /* store result color */ if (qss->colorOutSlot >= 0) { @@ -199,107 +130,6 @@ shade_quad( } } -#if 0 -#ifdef MESA_LLVM -#define DLLVM 0 -static void -shade_quad_llvm(struct quad_stage *qs, - struct quad_header *quad) -{ - struct quad_shade_stage *qss = quad_shade_stage(qs); - struct softpipe_context *softpipe = qs->softpipe; - float dests[4][16][4] ALIGN16_ATTRIB; - float inputs[4][16][4] ALIGN16_ATTRIB; - const float fx = (float) quad->x0; - const float fy = (float) quad->y0; - struct gallivm_prog *llvm = qss->llvm_prog; - - inputs[0][0][0] = fx; - inputs[1][0][0] = fx + 1.0f; - inputs[2][0][0] = fx; - inputs[3][0][0] = fx + 1.0f; - - inputs[0][0][1] = fy; - inputs[1][0][1] = fy; - inputs[2][0][1] = fy + 1.0f; - inputs[3][0][1] = fy + 1.0f; -#if DLLVM - debug_printf("MASK = %d\n", quad->mask); -#endif - gallivm_prog_inputs_interpolate(llvm, inputs, quad->coef); -#if DLLVM - for (int i = 0; i < 4; ++i) { - for (int j = 0; j < 2; ++j) { - debug_printf("IN(%d,%d) [%f %f %f %f]\n", i, j, - inputs[i][j][0], inputs[i][j][1], inputs[i][j][2], inputs[i][j][3]); - } - } -#endif - - quad->mask &= - gallivm_fragment_shader_exec(llvm, fx, fy, dests, inputs, - softpipe->mapped_constants[PIPE_SHADER_FRAGMENT], - qss->samplers); -#if DLLVM - debug_printf("OUT LLVM = 1[%f %f %f %f], 2[%f %f %f %f]\n", - dests[0][0][0], dests[0][0][1], dests[0][0][2], dests[0][0][3], - dests[0][1][0], dests[0][1][1], dests[0][1][2], dests[0][1][3]); -#endif - - /* store result color */ - if (qss->colorOutSlot >= 0) { - unsigned i; - /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] - == TGSI_SEMANTIC_COLOR); - for (i = 0; i < QUAD_SIZE; ++i) { - quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0]; - quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1]; - quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2]; - quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3]; - } - } -#if DLLVM - for (int i = 0; i < QUAD_SIZE; ++i) { - debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, - quad->outputs.color[0][i], - quad->outputs.color[1][i], - quad->outputs.color[2][i], - quad->outputs.color[3][i]); - } -#endif - - /* store result Z */ - if (qss->depthOutSlot >= 0) { - /* output[slot] is new Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = dests[i][0][2]; - } - } - else { - /* copy input Z (which was interpolated by the executor) to output Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = inputs[i][0][2]; - } - } -#if DLLVM - debug_printf("D [%f, %f, %f, %f] mask = %d\n", - quad->outputs.depth[0], - quad->outputs.depth[1], - quad->outputs.depth[2], - quad->outputs.depth[3], quad->mask); -#endif - - /* shader may cull fragments */ - if( quad->mask ) { - qs->next->run( qs->next, quad ); - } -} -#endif /*MESA_LLVM*/ -#endif - /** * Per-primitive (or per-begin?) setup */ @@ -315,15 +145,6 @@ static void shade_begin(struct quad_stage *qs) qss->samplers[i].texture = &softpipe->texture[i]->base; } -#ifdef MESA_LLVM - qss->llvm_prog = softpipe->fs->llvm_prog; -#endif - /* XXX only do this if the fragment shader changes... */ - tgsi_exec_machine_init(&qss->machine, - softpipe->fs->shader.tokens, - PIPE_MAX_SAMPLERS, - qss->samplers ); - /* find output slots for depth, color */ qss->colorOutSlot = -1; qss->depthOutSlot = -1; @@ -337,6 +158,10 @@ static void shade_begin(struct quad_stage *qs) break; } } + + softpipe->fs->prepare( softpipe->fs, + &qss->machine, + qss->samplers ); qs->next->begin(qs->next); } @@ -366,16 +191,7 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) qss->stage.softpipe = softpipe; qss->stage.begin = shade_begin; -#ifdef MESA_LLVM - /* disable until ported to accept - * x/y and soa layout - qss->stage.run = shade_quad_llvm; - */ - softpipe->use_sse = FALSE; - qss->stage.run = shade_quad; -#else qss->stage.run = shade_quad; -#endif qss->stage.destroy = shade_destroy; /* set TGSI sampler state that's constant */ @@ -386,5 +202,7 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) qss->samplers[i].cache = softpipe->tex_cache[i]; } + tgsi_exec_machine_init( &qss->machine ); + return &qss->stage; } diff --git a/src/mesa/pipe/softpipe/sp_state.h b/src/mesa/pipe/softpipe/sp_state.h index af955c1e17..431952f1aa 100644 --- a/src/mesa/pipe/softpipe/sp_state.h +++ b/src/mesa/pipe/softpipe/sp_state.h @@ -33,8 +33,6 @@ #include "pipe/p_state.h" -#include "x86/rtasm/x86sse.h" - #define SP_NEW_VIEWPORT 0x1 #define SP_NEW_RASTERIZER 0x2 @@ -53,27 +51,36 @@ #define SP_NEW_QUERY 0x4000 +struct tgsi_sampler; +struct tgsi_interp_coef; +struct tgsi_exec_machine; -#ifdef MESA_LLVM -struct gallivm_prog; -#endif +/** Subclass of pipe_shader_state (though it doesn't really need to be). + * + * This is starting to look an awful lot like a quad pipeline stage... + */ +struct sp_fragment_shader { + struct pipe_shader_state shader; + void (*prepare)( struct sp_fragment_shader *shader, + struct tgsi_exec_machine *machine, + struct tgsi_sampler *samplers); -/** Subclass of pipe_shader_state */ -struct sp_fragment_shader_state { - struct pipe_shader_state shader; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif -#ifdef MESA_LLVM - struct gallivm_prog *llvm_prog; -#endif + /* Run the shader - this interface will get cleaned up in the + * future: + */ + unsigned (*run)( struct sp_fragment_shader *shader, + struct tgsi_exec_machine *machine, + struct quad_header *quad ); + + + void (*delete)( struct sp_fragment_shader * ); }; /** Subclass of pipe_shader_state */ -struct sp_vertex_shader_state { +struct sp_vertex_shader { struct pipe_shader_state shader; struct draw_vertex_shader *draw_data; }; diff --git a/src/mesa/pipe/softpipe/sp_state_fs.c b/src/mesa/pipe/softpipe/sp_state_fs.c index 0b814fc284..c43f28583e 100644 --- a/src/mesa/pipe/softpipe/sp_state_fs.c +++ b/src/mesa/pipe/softpipe/sp_state_fs.c @@ -27,16 +27,15 @@ #include "sp_context.h" #include "sp_state.h" +#include "sp_fs.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "pipe/draw/draw_context.h" #include "pipe/p_shader_tokens.h" -#include "pipe/llvm/gallivm.h" +#include "pipe/draw/draw_context.h" #include "pipe/tgsi/util/tgsi_dump.h" -#include "pipe/tgsi/exec/tgsi_sse2.h" void * @@ -44,40 +43,22 @@ softpipe_create_fs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { struct softpipe_context *softpipe = softpipe_context(pipe); - struct sp_fragment_shader_state *state; + struct sp_fragment_shader *state; - /* Decide whether we'll be codegenerating this shader and if so do - * that now. - */ - - state = CALLOC_STRUCT(sp_fragment_shader_state); - if (!state) - return NULL; + if (softpipe->dump_fs) + tgsi_dump(templ->tokens, 0); - state->shader = *templ; - - if (softpipe->dump_fs) { - tgsi_dump(state->shader.tokens, 0); - } + state = softpipe_create_fs_llvm( softpipe, templ ); + if (state) + return state; + + state = softpipe_create_fs_sse( softpipe, templ ); + if (state) + return state; -#ifdef MESA_LLVM - state->llvm_prog = 0; - -#if 0 - if (!gallivm_global_cpu_engine()) { - gallivm_cpu_engine_create(state->llvm_prog); - } - else - gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); -#endif - -#elif defined(__i386__) || defined(__386__) - if (softpipe->use_sse) { - x86_init_func( &state->sse2_program ); - tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); - } -#endif + state = softpipe_create_fs_exec( softpipe, templ ); + assert(state); return state; } @@ -87,7 +68,7 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->fs = (struct sp_fragment_shader_state *) fs; + softpipe->fs = (struct sp_fragment_shader *) fs; softpipe->dirty |= SP_NEW_FS; } @@ -96,13 +77,9 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) void softpipe_delete_fs_state(struct pipe_context *pipe, void *fs) { - struct sp_fragment_shader_state *state = fs; - -#if defined(__i386__) || defined(__386__) - x86_release_func( &state->sse2_program ); -#endif - - FREE( state ); + struct sp_fragment_shader *state = fs; + + state->delete( state ); } @@ -111,9 +88,9 @@ softpipe_create_vs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { struct softpipe_context *softpipe = softpipe_context(pipe); - struct sp_vertex_shader_state *state; + struct sp_vertex_shader *state; - state = CALLOC_STRUCT(sp_vertex_shader_state); + state = CALLOC_STRUCT(sp_vertex_shader); if (state == NULL ) { return NULL; } @@ -136,7 +113,7 @@ softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->vs = (const struct sp_vertex_shader_state *)vs; + softpipe->vs = (const struct sp_vertex_shader *)vs; draw_bind_vertex_shader(softpipe->draw, softpipe->vs->draw_data); @@ -149,8 +126,8 @@ softpipe_delete_vs_state(struct pipe_context *pipe, void *vs) { struct softpipe_context *softpipe = softpipe_context(pipe); - struct sp_vertex_shader_state *state = - (struct sp_vertex_shader_state *)vs; + struct sp_vertex_shader *state = + (struct sp_vertex_shader *)vs; draw_delete_vertex_shader(softpipe->draw, state->draw_data); FREE( state ); diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.c b/src/mesa/pipe/tgsi/exec/tgsi_exec.c index 336ae1c8b6..92a6fd4d62 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.c @@ -120,18 +120,41 @@ -static void -tgsi_exec_prepare( struct tgsi_exec_machine *mach ) +/** + * Initialize machine state by expanding tokens to full instructions, + * allocating temporary storage, setting up constants, etc. + * After this, we can call tgsi_exec_machine_run() many times. + */ +void +tgsi_exec_machine_bind_shader( + struct tgsi_exec_machine *mach, + const struct tgsi_token *tokens, + uint numSamplers, + struct tgsi_sampler *samplers) { - struct tgsi_exec_labels *labels = &mach->Labels; + uint k; struct tgsi_parse_context parse; + struct tgsi_exec_labels *labels = &mach->Labels; struct tgsi_full_instruction *instructions; struct tgsi_full_declaration *declarations; uint maxInstructions = 10, numInstructions = 0; uint maxDeclarations = 10, numDeclarations = 0; - uint k; uint instno = 0; +#if 0 + tgsi_dump(tokens, 0); +#endif + + mach->Tokens = tokens; + mach->Samplers = samplers; + + k = tgsi_parse_init (&parse, mach->Tokens); + if (k != TGSI_PARSE_OK) { + debug_printf( "Problem parsing!\n" ); + return; + } + + mach->Processor = parse.FullHeader.Processor.Processor; mach->ImmLimit = 0; labels->count = 0; @@ -141,11 +164,6 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) instructions = (struct tgsi_full_instruction *) MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) ); - k = tgsi_parse_init( &parse, mach->Tokens ); - if (k != TGSI_PARSE_OK) { - debug_printf("Problem parsing!\n"); - return; - } while( !tgsi_parse_end_of_tokens( &parse ) ) { uint pointer = parse.Position; @@ -176,7 +194,8 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES ); for( i = 0; i < size; i++ ) { - mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + mach->Imms[mach->ImmLimit + i / 4][i % 4] = + parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; } mach->ImmLimit += size / 4; } @@ -224,37 +243,11 @@ tgsi_exec_prepare( struct tgsi_exec_machine *mach ) } -/** - * Initialize machine state by expanding tokens to full instructions, - * allocating temporary storage, setting up constants, etc. - * After this, we can call tgsi_exec_machine_run() many times. - */ void tgsi_exec_machine_init( - struct tgsi_exec_machine *mach, - const struct tgsi_token *tokens, - uint numSamplers, - struct tgsi_sampler *samplers) + struct tgsi_exec_machine *mach ) { - uint i, k; - struct tgsi_parse_context parse; - -#if 0 - tgsi_dump(tokens, 0); -#endif - - mach->Tokens = tokens; - - mach->Samplers = samplers; - - k = tgsi_parse_init (&parse, mach->Tokens); - if (k != TGSI_PARSE_OK) { - debug_printf( "Problem parsing!\n" ); - return; - } - - mach->Processor = parse.FullHeader.Processor.Processor; - tgsi_parse_free (&parse); + uint i; mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; @@ -270,8 +263,6 @@ tgsi_exec_machine_init( mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; } - - tgsi_exec_prepare( mach ); } diff --git a/src/mesa/pipe/tgsi/exec/tgsi_exec.h b/src/mesa/pipe/tgsi/exec/tgsi_exec.h index 1fb66ee960..45c49dd007 100644 --- a/src/mesa/pipe/tgsi/exec/tgsi_exec.h +++ b/src/mesa/pipe/tgsi/exec/tgsi_exec.h @@ -215,12 +215,16 @@ struct tgsi_exec_machine struct tgsi_exec_labels Labels; }; - void tgsi_exec_machine_init( + struct tgsi_exec_machine *mach ); + + +void +tgsi_exec_machine_bind_shader( struct tgsi_exec_machine *mach, const struct tgsi_token *tokens, - unsigned numSamplers, + uint numSamplers, struct tgsi_sampler *samplers); uint diff --git a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c index 6df7588c92..9bc9483e3b 100755 --- a/src/mesa/pipe/tgsi/exec/tgsi_sse2.c +++ b/src/mesa/pipe/tgsi/exec/tgsi_sse2.c @@ -2186,9 +2186,9 @@ emit_declaration( break; case TGSI_INTERPOLATE_LINEAR: - emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_tempf( func, 0, 0, TGSI_SWIZZLE_X ); emit_coef_dadx( func, 1, i, j ); - emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y ); emit_coef_dady( func, 3, i, j ); emit_mul( func, 0, 1 ); /* x * dadx */ emit_coef_a0( func, 4, i, j ); @@ -2199,12 +2199,12 @@ emit_declaration( break; case TGSI_INTERPOLATE_PERSPECTIVE: - emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_tempf( func, 0, 0, TGSI_SWIZZLE_X ); emit_coef_dadx( func, 1, i, j ); - emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y ); emit_coef_dady( func, 3, i, j ); emit_mul( func, 0, 1 ); /* x * dadx */ - emit_inputf( func, 4, 0, TGSI_SWIZZLE_W ); + emit_tempf( func, 4, 0, TGSI_SWIZZLE_W ); emit_coef_a0( func, 5, i, j ); emit_rcp( func, 4, 4 ); /* 1.0 / w */ emit_mul( func, 2, 3 ); /* y * dady */ -- cgit v1.2.3 From 92fcbf6e7bc622dcace226bb70ff6d5cdbdbaecb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Feb 2008 20:07:18 +0900 Subject: Code reorganization: s/aux/auxiliary/. "aux" is a reserved name on Windows (X_X) --- src/gallium/aux/Makefile | 24 - src/gallium/aux/cso_cache/cso_cache.c | 181 -- src/gallium/aux/cso_cache/cso_cache.h | 107 - src/gallium/aux/cso_cache/cso_hash.c | 388 --- src/gallium/aux/cso_cache/cso_hash.h | 62 - src/gallium/aux/draw/Makefile | 2 - src/gallium/aux/draw/draw_clip.c | 488 ---- src/gallium/aux/draw/draw_context.c | 293 --- src/gallium/aux/draw/draw_context.h | 142 -- src/gallium/aux/draw/draw_cull.c | 150 -- src/gallium/aux/draw/draw_debug.c | 113 - src/gallium/aux/draw/draw_flatshade.c | 205 -- src/gallium/aux/draw/draw_offset.c | 186 -- src/gallium/aux/draw/draw_prim.c | 482 ---- src/gallium/aux/draw/draw_private.h | 346 --- src/gallium/aux/draw/draw_stipple.c | 239 -- src/gallium/aux/draw/draw_twoside.c | 203 -- src/gallium/aux/draw/draw_unfilled.c | 206 -- src/gallium/aux/draw/draw_validate.c | 185 -- src/gallium/aux/draw/draw_vbuf.c | 570 ----- src/gallium/aux/draw/draw_vbuf.h | 106 - src/gallium/aux/draw/draw_vertex.c | 79 - src/gallium/aux/draw/draw_vertex.h | 111 - src/gallium/aux/draw/draw_vertex_cache.c | 196 -- src/gallium/aux/draw/draw_vertex_fetch.c | 510 ---- src/gallium/aux/draw/draw_vertex_shader.c | 325 --- src/gallium/aux/draw/draw_vf.c | 428 ---- src/gallium/aux/draw/draw_vf.h | 236 -- src/gallium/aux/draw/draw_vf_generic.c | 585 ----- src/gallium/aux/draw/draw_vf_sse.c | 614 ----- src/gallium/aux/draw/draw_wide_prims.c | 432 ---- src/gallium/aux/llvm/Makefile | 85 - src/gallium/aux/llvm/gallivm.cpp | 327 --- src/gallium/aux/llvm/gallivm.h | 103 - src/gallium/aux/llvm/gallivm_builtins.cpp | 567 ----- src/gallium/aux/llvm/gallivm_cpu.cpp | 202 -- src/gallium/aux/llvm/gallivm_p.h | 110 - src/gallium/aux/llvm/instructions.cpp | 889 ------- src/gallium/aux/llvm/instructions.h | 152 -- src/gallium/aux/llvm/instructionssoa.cpp | 121 - src/gallium/aux/llvm/instructionssoa.h | 74 - src/gallium/aux/llvm/llvm_builtins.c | 115 - src/gallium/aux/llvm/loweringpass.cpp | 17 - src/gallium/aux/llvm/loweringpass.h | 15 - src/gallium/aux/llvm/storage.cpp | 364 --- src/gallium/aux/llvm/storage.h | 133 -- src/gallium/aux/llvm/storagesoa.cpp | 389 --- src/gallium/aux/llvm/storagesoa.h | 111 - src/gallium/aux/llvm/tgsitollvm.cpp | 1221 ---------- src/gallium/aux/llvm/tgsitollvm.h | 20 - src/gallium/aux/pipebuffer/Makefile | 23 - src/gallium/aux/pipebuffer/linked_list.h | 91 - src/gallium/aux/pipebuffer/pb_buffer.h | 202 -- src/gallium/aux/pipebuffer/pb_buffer_fenced.c | 299 --- src/gallium/aux/pipebuffer/pb_buffer_fenced.h | 117 - src/gallium/aux/pipebuffer/pb_buffer_malloc.c | 127 - src/gallium/aux/pipebuffer/pb_bufmgr.h | 126 - src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c | 131 -- src/gallium/aux/pipebuffer/pb_bufmgr_mm.c | 593 ----- src/gallium/aux/pipebuffer/pb_bufmgr_pool.c | 288 --- src/gallium/aux/pipebuffer/pb_winsys.c | 170 -- src/gallium/aux/tgsi/Makefile | 3 - src/gallium/aux/tgsi/exec/Makefile | 3 - src/gallium/aux/tgsi/exec/tgsi_exec.c | 2485 -------------------- src/gallium/aux/tgsi/exec/tgsi_exec.h | 239 -- src/gallium/aux/tgsi/exec/tgsi_sse2.c | 2378 ------------------- src/gallium/aux/tgsi/exec/tgsi_sse2.h | 26 - src/gallium/aux/tgsi/util/tgsi_build.c | 1371 ----------- src/gallium/aux/tgsi/util/tgsi_build.h | 320 --- src/gallium/aux/tgsi/util/tgsi_dump.c | 1581 ------------- src/gallium/aux/tgsi/util/tgsi_dump.h | 28 - src/gallium/aux/tgsi/util/tgsi_parse.c | 319 --- src/gallium/aux/tgsi/util/tgsi_parse.h | 121 - src/gallium/aux/tgsi/util/tgsi_transform.c | 199 -- src/gallium/aux/tgsi/util/tgsi_transform.h | 93 - src/gallium/aux/tgsi/util/tgsi_util.c | 274 --- src/gallium/aux/tgsi/util/tgsi_util.h | 70 - src/gallium/aux/util/p_debug.c | 76 - src/gallium/aux/util/p_tile.c | 699 ------ src/gallium/aux/util/p_tile.h | 81 - src/gallium/aux/util/p_util.c | 73 - src/gallium/auxiliary/Makefile | 24 + src/gallium/auxiliary/cso_cache/cso_cache.c | 181 ++ src/gallium/auxiliary/cso_cache/cso_cache.h | 107 + src/gallium/auxiliary/cso_cache/cso_hash.c | 388 +++ src/gallium/auxiliary/cso_cache/cso_hash.h | 62 + src/gallium/auxiliary/draw/Makefile | 2 + src/gallium/auxiliary/draw/draw_clip.c | 488 ++++ src/gallium/auxiliary/draw/draw_context.c | 293 +++ src/gallium/auxiliary/draw/draw_context.h | 142 ++ src/gallium/auxiliary/draw/draw_cull.c | 150 ++ src/gallium/auxiliary/draw/draw_debug.c | 113 + src/gallium/auxiliary/draw/draw_flatshade.c | 205 ++ src/gallium/auxiliary/draw/draw_offset.c | 186 ++ src/gallium/auxiliary/draw/draw_prim.c | 482 ++++ src/gallium/auxiliary/draw/draw_private.h | 346 +++ src/gallium/auxiliary/draw/draw_stipple.c | 239 ++ src/gallium/auxiliary/draw/draw_twoside.c | 203 ++ src/gallium/auxiliary/draw/draw_unfilled.c | 206 ++ src/gallium/auxiliary/draw/draw_validate.c | 185 ++ src/gallium/auxiliary/draw/draw_vbuf.c | 570 +++++ src/gallium/auxiliary/draw/draw_vbuf.h | 106 + src/gallium/auxiliary/draw/draw_vertex.c | 79 + src/gallium/auxiliary/draw/draw_vertex.h | 111 + src/gallium/auxiliary/draw/draw_vertex_cache.c | 196 ++ src/gallium/auxiliary/draw/draw_vertex_fetch.c | 510 ++++ src/gallium/auxiliary/draw/draw_vertex_shader.c | 325 +++ src/gallium/auxiliary/draw/draw_vf.c | 428 ++++ src/gallium/auxiliary/draw/draw_vf.h | 236 ++ src/gallium/auxiliary/draw/draw_vf_generic.c | 585 +++++ src/gallium/auxiliary/draw/draw_vf_sse.c | 614 +++++ src/gallium/auxiliary/draw/draw_wide_prims.c | 432 ++++ src/gallium/auxiliary/llvm/Makefile | 85 + src/gallium/auxiliary/llvm/gallivm.cpp | 327 +++ src/gallium/auxiliary/llvm/gallivm.h | 103 + src/gallium/auxiliary/llvm/gallivm_builtins.cpp | 567 +++++ src/gallium/auxiliary/llvm/gallivm_cpu.cpp | 202 ++ src/gallium/auxiliary/llvm/gallivm_p.h | 110 + src/gallium/auxiliary/llvm/instructions.cpp | 889 +++++++ src/gallium/auxiliary/llvm/instructions.h | 152 ++ src/gallium/auxiliary/llvm/instructionssoa.cpp | 121 + src/gallium/auxiliary/llvm/instructionssoa.h | 74 + src/gallium/auxiliary/llvm/llvm_builtins.c | 115 + src/gallium/auxiliary/llvm/loweringpass.cpp | 17 + src/gallium/auxiliary/llvm/loweringpass.h | 15 + src/gallium/auxiliary/llvm/storage.cpp | 364 +++ src/gallium/auxiliary/llvm/storage.h | 133 ++ src/gallium/auxiliary/llvm/storagesoa.cpp | 389 +++ src/gallium/auxiliary/llvm/storagesoa.h | 111 + src/gallium/auxiliary/llvm/tgsitollvm.cpp | 1221 ++++++++++ src/gallium/auxiliary/llvm/tgsitollvm.h | 20 + src/gallium/auxiliary/pipebuffer/Makefile | 23 + src/gallium/auxiliary/pipebuffer/linked_list.h | 91 + src/gallium/auxiliary/pipebuffer/pb_buffer.h | 202 ++ .../auxiliary/pipebuffer/pb_buffer_fenced.c | 299 +++ .../auxiliary/pipebuffer/pb_buffer_fenced.h | 117 + .../auxiliary/pipebuffer/pb_buffer_malloc.c | 127 + src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 126 + .../auxiliary/pipebuffer/pb_bufmgr_fenced.c | 131 ++ src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c | 593 +++++ src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c | 288 +++ src/gallium/auxiliary/pipebuffer/pb_winsys.c | 170 ++ src/gallium/auxiliary/tgsi/Makefile | 3 + src/gallium/auxiliary/tgsi/exec/Makefile | 3 + src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 2485 ++++++++++++++++++++ src/gallium/auxiliary/tgsi/exec/tgsi_exec.h | 239 ++ src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2378 +++++++++++++++++++ src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h | 26 + src/gallium/auxiliary/tgsi/util/tgsi_build.c | 1371 +++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_build.h | 320 +++ src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 1581 +++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_dump.h | 28 + src/gallium/auxiliary/tgsi/util/tgsi_parse.c | 319 +++ src/gallium/auxiliary/tgsi/util/tgsi_parse.h | 121 + src/gallium/auxiliary/tgsi/util/tgsi_transform.c | 199 ++ src/gallium/auxiliary/tgsi/util/tgsi_transform.h | 93 + src/gallium/auxiliary/tgsi/util/tgsi_util.c | 274 +++ src/gallium/auxiliary/tgsi/util/tgsi_util.h | 70 + src/gallium/auxiliary/util/p_debug.c | 76 + src/gallium/auxiliary/util/p_tile.c | 699 ++++++ src/gallium/auxiliary/util/p_tile.h | 81 + src/gallium/auxiliary/util/p_util.c | 73 + 162 files changed, 25815 insertions(+), 25815 deletions(-) delete mode 100644 src/gallium/aux/Makefile delete mode 100644 src/gallium/aux/cso_cache/cso_cache.c delete mode 100644 src/gallium/aux/cso_cache/cso_cache.h delete mode 100644 src/gallium/aux/cso_cache/cso_hash.c delete mode 100644 src/gallium/aux/cso_cache/cso_hash.h delete mode 100644 src/gallium/aux/draw/Makefile delete mode 100644 src/gallium/aux/draw/draw_clip.c delete mode 100644 src/gallium/aux/draw/draw_context.c delete mode 100644 src/gallium/aux/draw/draw_context.h delete mode 100644 src/gallium/aux/draw/draw_cull.c delete mode 100644 src/gallium/aux/draw/draw_debug.c delete mode 100644 src/gallium/aux/draw/draw_flatshade.c delete mode 100644 src/gallium/aux/draw/draw_offset.c delete mode 100644 src/gallium/aux/draw/draw_prim.c delete mode 100644 src/gallium/aux/draw/draw_private.h delete mode 100644 src/gallium/aux/draw/draw_stipple.c delete mode 100644 src/gallium/aux/draw/draw_twoside.c delete mode 100644 src/gallium/aux/draw/draw_unfilled.c delete mode 100644 src/gallium/aux/draw/draw_validate.c delete mode 100644 src/gallium/aux/draw/draw_vbuf.c delete mode 100644 src/gallium/aux/draw/draw_vbuf.h delete mode 100644 src/gallium/aux/draw/draw_vertex.c delete mode 100644 src/gallium/aux/draw/draw_vertex.h delete mode 100644 src/gallium/aux/draw/draw_vertex_cache.c delete mode 100644 src/gallium/aux/draw/draw_vertex_fetch.c delete mode 100644 src/gallium/aux/draw/draw_vertex_shader.c delete mode 100644 src/gallium/aux/draw/draw_vf.c delete mode 100644 src/gallium/aux/draw/draw_vf.h delete mode 100644 src/gallium/aux/draw/draw_vf_generic.c delete mode 100644 src/gallium/aux/draw/draw_vf_sse.c delete mode 100644 src/gallium/aux/draw/draw_wide_prims.c delete mode 100644 src/gallium/aux/llvm/Makefile delete mode 100644 src/gallium/aux/llvm/gallivm.cpp delete mode 100644 src/gallium/aux/llvm/gallivm.h delete mode 100644 src/gallium/aux/llvm/gallivm_builtins.cpp delete mode 100644 src/gallium/aux/llvm/gallivm_cpu.cpp delete mode 100644 src/gallium/aux/llvm/gallivm_p.h delete mode 100644 src/gallium/aux/llvm/instructions.cpp delete mode 100644 src/gallium/aux/llvm/instructions.h delete mode 100644 src/gallium/aux/llvm/instructionssoa.cpp delete mode 100644 src/gallium/aux/llvm/instructionssoa.h delete mode 100644 src/gallium/aux/llvm/llvm_builtins.c delete mode 100644 src/gallium/aux/llvm/loweringpass.cpp delete mode 100644 src/gallium/aux/llvm/loweringpass.h delete mode 100644 src/gallium/aux/llvm/storage.cpp delete mode 100644 src/gallium/aux/llvm/storage.h delete mode 100644 src/gallium/aux/llvm/storagesoa.cpp delete mode 100644 src/gallium/aux/llvm/storagesoa.h delete mode 100644 src/gallium/aux/llvm/tgsitollvm.cpp delete mode 100644 src/gallium/aux/llvm/tgsitollvm.h delete mode 100644 src/gallium/aux/pipebuffer/Makefile delete mode 100644 src/gallium/aux/pipebuffer/linked_list.h delete mode 100644 src/gallium/aux/pipebuffer/pb_buffer.h delete mode 100644 src/gallium/aux/pipebuffer/pb_buffer_fenced.c delete mode 100644 src/gallium/aux/pipebuffer/pb_buffer_fenced.h delete mode 100644 src/gallium/aux/pipebuffer/pb_buffer_malloc.c delete mode 100644 src/gallium/aux/pipebuffer/pb_bufmgr.h delete mode 100644 src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c delete mode 100644 src/gallium/aux/pipebuffer/pb_bufmgr_mm.c delete mode 100644 src/gallium/aux/pipebuffer/pb_bufmgr_pool.c delete mode 100644 src/gallium/aux/pipebuffer/pb_winsys.c delete mode 100644 src/gallium/aux/tgsi/Makefile delete mode 100644 src/gallium/aux/tgsi/exec/Makefile delete mode 100644 src/gallium/aux/tgsi/exec/tgsi_exec.c delete mode 100644 src/gallium/aux/tgsi/exec/tgsi_exec.h delete mode 100755 src/gallium/aux/tgsi/exec/tgsi_sse2.c delete mode 100755 src/gallium/aux/tgsi/exec/tgsi_sse2.h delete mode 100644 src/gallium/aux/tgsi/util/tgsi_build.c delete mode 100644 src/gallium/aux/tgsi/util/tgsi_build.h delete mode 100644 src/gallium/aux/tgsi/util/tgsi_dump.c delete mode 100644 src/gallium/aux/tgsi/util/tgsi_dump.h delete mode 100644 src/gallium/aux/tgsi/util/tgsi_parse.c delete mode 100644 src/gallium/aux/tgsi/util/tgsi_parse.h delete mode 100644 src/gallium/aux/tgsi/util/tgsi_transform.c delete mode 100644 src/gallium/aux/tgsi/util/tgsi_transform.h delete mode 100644 src/gallium/aux/tgsi/util/tgsi_util.c delete mode 100644 src/gallium/aux/tgsi/util/tgsi_util.h delete mode 100644 src/gallium/aux/util/p_debug.c delete mode 100644 src/gallium/aux/util/p_tile.c delete mode 100644 src/gallium/aux/util/p_tile.h delete mode 100644 src/gallium/aux/util/p_util.c create mode 100644 src/gallium/auxiliary/Makefile create mode 100644 src/gallium/auxiliary/cso_cache/cso_cache.c create mode 100644 src/gallium/auxiliary/cso_cache/cso_cache.h create mode 100644 src/gallium/auxiliary/cso_cache/cso_hash.c create mode 100644 src/gallium/auxiliary/cso_cache/cso_hash.h create mode 100644 src/gallium/auxiliary/draw/Makefile create mode 100644 src/gallium/auxiliary/draw/draw_clip.c create mode 100644 src/gallium/auxiliary/draw/draw_context.c create mode 100644 src/gallium/auxiliary/draw/draw_context.h create mode 100644 src/gallium/auxiliary/draw/draw_cull.c create mode 100644 src/gallium/auxiliary/draw/draw_debug.c create mode 100644 src/gallium/auxiliary/draw/draw_flatshade.c create mode 100644 src/gallium/auxiliary/draw/draw_offset.c create mode 100644 src/gallium/auxiliary/draw/draw_prim.c create mode 100644 src/gallium/auxiliary/draw/draw_private.h create mode 100644 src/gallium/auxiliary/draw/draw_stipple.c create mode 100644 src/gallium/auxiliary/draw/draw_twoside.c create mode 100644 src/gallium/auxiliary/draw/draw_unfilled.c create mode 100644 src/gallium/auxiliary/draw/draw_validate.c create mode 100644 src/gallium/auxiliary/draw/draw_vbuf.c create mode 100644 src/gallium/auxiliary/draw/draw_vbuf.h create mode 100644 src/gallium/auxiliary/draw/draw_vertex.c create mode 100644 src/gallium/auxiliary/draw/draw_vertex.h create mode 100644 src/gallium/auxiliary/draw/draw_vertex_cache.c create mode 100644 src/gallium/auxiliary/draw/draw_vertex_fetch.c create mode 100644 src/gallium/auxiliary/draw/draw_vertex_shader.c create mode 100644 src/gallium/auxiliary/draw/draw_vf.c create mode 100644 src/gallium/auxiliary/draw/draw_vf.h create mode 100644 src/gallium/auxiliary/draw/draw_vf_generic.c create mode 100644 src/gallium/auxiliary/draw/draw_vf_sse.c create mode 100644 src/gallium/auxiliary/draw/draw_wide_prims.c create mode 100644 src/gallium/auxiliary/llvm/Makefile create mode 100644 src/gallium/auxiliary/llvm/gallivm.cpp create mode 100644 src/gallium/auxiliary/llvm/gallivm.h create mode 100644 src/gallium/auxiliary/llvm/gallivm_builtins.cpp create mode 100644 src/gallium/auxiliary/llvm/gallivm_cpu.cpp create mode 100644 src/gallium/auxiliary/llvm/gallivm_p.h create mode 100644 src/gallium/auxiliary/llvm/instructions.cpp create mode 100644 src/gallium/auxiliary/llvm/instructions.h create mode 100644 src/gallium/auxiliary/llvm/instructionssoa.cpp create mode 100644 src/gallium/auxiliary/llvm/instructionssoa.h create mode 100644 src/gallium/auxiliary/llvm/llvm_builtins.c create mode 100644 src/gallium/auxiliary/llvm/loweringpass.cpp create mode 100644 src/gallium/auxiliary/llvm/loweringpass.h create mode 100644 src/gallium/auxiliary/llvm/storage.cpp create mode 100644 src/gallium/auxiliary/llvm/storage.h create mode 100644 src/gallium/auxiliary/llvm/storagesoa.cpp create mode 100644 src/gallium/auxiliary/llvm/storagesoa.h create mode 100644 src/gallium/auxiliary/llvm/tgsitollvm.cpp create mode 100644 src/gallium/auxiliary/llvm/tgsitollvm.h create mode 100644 src/gallium/auxiliary/pipebuffer/Makefile create mode 100644 src/gallium/auxiliary/pipebuffer/linked_list.h create mode 100644 src/gallium/auxiliary/pipebuffer/pb_buffer.h create mode 100644 src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c create mode 100644 src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h create mode 100644 src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c create mode 100644 src/gallium/auxiliary/pipebuffer/pb_bufmgr.h create mode 100644 src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c create mode 100644 src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c create mode 100644 src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c create mode 100644 src/gallium/auxiliary/pipebuffer/pb_winsys.c create mode 100644 src/gallium/auxiliary/tgsi/Makefile create mode 100644 src/gallium/auxiliary/tgsi/exec/Makefile create mode 100644 src/gallium/auxiliary/tgsi/exec/tgsi_exec.c create mode 100644 src/gallium/auxiliary/tgsi/exec/tgsi_exec.h create mode 100755 src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c create mode 100755 src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_build.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_build.h create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_dump.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_dump.h create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_parse.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_parse.h create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_transform.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_transform.h create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_util.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_util.h create mode 100644 src/gallium/auxiliary/util/p_debug.c create mode 100644 src/gallium/auxiliary/util/p_tile.c create mode 100644 src/gallium/auxiliary/util/p_tile.h create mode 100644 src/gallium/auxiliary/util/p_util.c (limited to 'src') diff --git a/src/gallium/aux/Makefile b/src/gallium/aux/Makefile deleted file mode 100644 index da68498aa1..0000000000 --- a/src/gallium/aux/Makefile +++ /dev/null @@ -1,24 +0,0 @@ -TOP = ../../.. -include $(TOP)/configs/current - - -ifeq ($(CONFIG_NAME), linux-llvm) -LLVM_DIR = llvm -endif - -SUBDIRS = pipebuffer $(LLVM_DIR) - - -default: subdirs - - -subdirs: - @for dir in $(SUBDIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ - fi \ - done - - -clean: - rm -f `find . -name \*.[oa]` diff --git a/src/gallium/aux/cso_cache/cso_cache.c b/src/gallium/aux/cso_cache/cso_cache.c deleted file mode 100644 index 9e77e0774d..0000000000 --- a/src/gallium/aux/cso_cache/cso_cache.c +++ /dev/null @@ -1,181 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Zack Rusin - */ - -#include "cso_cache.h" -#include "cso_hash.h" - -#if 1 -static unsigned hash_key(const void *key, unsigned key_size) -{ - unsigned *ikey = (unsigned *)key; - unsigned hash = 0, i; - - assert(key_size % 4 == 0); - - /* I'm sure this can be improved on: - */ - for (i = 0; i < key_size/4; i++) - hash ^= ikey[i]; - - return hash; -} -#else -static unsigned hash_key(const unsigned char *p, int n) -{ - unsigned h = 0; - unsigned g; - - while (n--) { - h = (h << 4) + *p++; - if ((g = (h & 0xf0000000)) != 0) - h ^= g >> 23; - h &= ~g; - } - return h; -} -#endif - -unsigned cso_construct_key(void *item, int item_size) -{ - return hash_key((item), item_size); -} - -static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_type type) -{ - struct cso_hash *hash = 0; - - switch(type) { - case CSO_BLEND: - hash = sc->blend_hash; - break; - case CSO_SAMPLER: - hash = sc->sampler_hash; - break; - case CSO_DEPTH_STENCIL_ALPHA: - hash = sc->depth_stencil_hash; - break; - case CSO_RASTERIZER: - hash = sc->rasterizer_hash; - break; - case CSO_FRAGMENT_SHADER: - hash = sc->fs_hash; - break; - case CSO_VERTEX_SHADER: - hash = sc->vs_hash; - break; - } - - return hash; -} - -static int _cso_size_for_type(enum cso_cache_type type) -{ - switch(type) { - case CSO_BLEND: - return sizeof(struct pipe_blend_state); - case CSO_SAMPLER: - return sizeof(struct pipe_sampler_state); - case CSO_DEPTH_STENCIL_ALPHA: - return sizeof(struct pipe_depth_stencil_alpha_state); - case CSO_RASTERIZER: - return sizeof(struct pipe_rasterizer_state); - case CSO_FRAGMENT_SHADER: - return sizeof(struct pipe_shader_state); - case CSO_VERTEX_SHADER: - return sizeof(struct pipe_shader_state); - } - return 0; -} - -struct cso_hash_iter -cso_insert_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *state) -{ - struct cso_hash *hash = _cso_hash_for_type(sc, type); - return cso_hash_insert(hash, hash_key, state); -} - -struct cso_hash_iter -cso_find_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type) -{ - struct cso_hash *hash = _cso_hash_for_type(sc, type); - - return cso_hash_find(hash, hash_key); -} - -struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *templ) -{ - struct cso_hash_iter iter = cso_find_state(sc, hash_key, type); - int size = _cso_size_for_type(type); - while (!cso_hash_iter_is_null(iter)) { - void *iter_data = cso_hash_iter_data(iter); - if (!memcmp(iter_data, templ, size)) - return iter; - iter = cso_hash_iter_next(iter); - } - return iter; -} - -void * cso_take_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type) -{ - struct cso_hash *hash = _cso_hash_for_type(sc, type); - return cso_hash_take(hash, hash_key); -} - -struct cso_cache *cso_cache_create(void) -{ - struct cso_cache *sc = malloc(sizeof(struct cso_cache)); - - sc->blend_hash = cso_hash_create(); - sc->sampler_hash = cso_hash_create(); - sc->depth_stencil_hash = cso_hash_create(); - sc->rasterizer_hash = cso_hash_create(); - sc->fs_hash = cso_hash_create(); - sc->vs_hash = cso_hash_create(); - - return sc; -} - -void cso_cache_delete(struct cso_cache *sc) -{ - assert(sc); - cso_hash_delete(sc->blend_hash); - cso_hash_delete(sc->sampler_hash); - cso_hash_delete(sc->depth_stencil_hash); - cso_hash_delete(sc->rasterizer_hash); - cso_hash_delete(sc->fs_hash); - cso_hash_delete(sc->vs_hash); - free(sc); -} diff --git a/src/gallium/aux/cso_cache/cso_cache.h b/src/gallium/aux/cso_cache/cso_cache.h deleted file mode 100644 index 116e2eaa2c..0000000000 --- a/src/gallium/aux/cso_cache/cso_cache.h +++ /dev/null @@ -1,107 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#ifndef CSO_CACHE_H -#define CSO_CACHE_H - -#include "pipe/p_context.h" -#include "pipe/p_state.h" - - -struct cso_hash; - -struct cso_cache { - struct cso_hash *blend_hash; - struct cso_hash *depth_stencil_hash; - struct cso_hash *fs_hash; - struct cso_hash *vs_hash; - struct cso_hash *rasterizer_hash; - struct cso_hash *sampler_hash; -}; - -struct cso_blend { - struct pipe_blend_state state; - void *data; -}; - -struct cso_depth_stencil_alpha { - struct pipe_depth_stencil_alpha_state state; - void *data; -}; - -struct cso_rasterizer { - struct pipe_rasterizer_state state; - void *data; -}; - -struct cso_fragment_shader { - struct pipe_shader_state state; - void *data; -}; - -struct cso_vertex_shader { - struct pipe_shader_state state; - void *data; -}; - -struct cso_sampler { - struct pipe_sampler_state state; - void *data; -}; - - -enum cso_cache_type { - CSO_BLEND, - CSO_SAMPLER, - CSO_DEPTH_STENCIL_ALPHA, - CSO_RASTERIZER, - CSO_FRAGMENT_SHADER, - CSO_VERTEX_SHADER -}; - -unsigned cso_construct_key(void *item, int item_size); - -struct cso_cache *cso_cache_create(void); -void cso_cache_delete(struct cso_cache *sc); - -struct cso_hash_iter cso_insert_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *state); -struct cso_hash_iter cso_find_state(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type); -struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, - unsigned hash_key, enum cso_cache_type type, - void *templ); -void * cso_take_state(struct cso_cache *sc, unsigned hash_key, - enum cso_cache_type type); - -#endif diff --git a/src/gallium/aux/cso_cache/cso_hash.c b/src/gallium/aux/cso_cache/cso_hash.c deleted file mode 100644 index 0338cb3b47..0000000000 --- a/src/gallium/aux/cso_cache/cso_hash.c +++ /dev/null @@ -1,388 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#include "cso_hash.h" - -#include -#include -#include -#include - -#define MAX(a, b) ((a > b) ? (a) : (b)) - -static const int MinNumBits = 4; - -static const unsigned char prime_deltas[] = { - 0, 0, 1, 3, 1, 5, 3, 3, 1, 9, 7, 5, 3, 9, 25, 3, - 1, 21, 3, 21, 7, 15, 9, 5, 3, 29, 15, 0, 0, 0, 0, 0 -}; - -static int primeForNumBits(int numBits) -{ - return (1 << numBits) + prime_deltas[numBits]; -} - -/* - Returns the smallest integer n such that - primeForNumBits(n) >= hint. -*/ -static int countBits(int hint) -{ - int numBits = 0; - int bits = hint; - - while (bits > 1) { - bits >>= 1; - numBits++; - } - - if (numBits >= (int)sizeof(prime_deltas)) { - numBits = sizeof(prime_deltas) - 1; - } else if (primeForNumBits(numBits) < hint) { - ++numBits; - } - return numBits; -} - -struct cso_node { - struct cso_node *next; - unsigned key; - void *value; -}; - -struct cso_hash_data { - struct cso_node *fakeNext; - struct cso_node **buckets; - int size; - int nodeSize; - short userNumBits; - short numBits; - int numBuckets; -}; - -struct cso_hash { - union { - struct cso_hash_data *d; - struct cso_node *e; - } data; -}; - -static void *cso_data_allocate_node(struct cso_hash_data *hash) -{ - return malloc(hash->nodeSize); -} - -static void cso_data_free_node(struct cso_node *node) -{ - /* XXX still a leak here. - * Need to cast value ptr to original cso type, then free the - * driver-specific data hanging off of it. For example: - struct cso_sampler *csamp = (struct cso_sampler *) node->value; - free(csamp->data); - */ - free(node->value); - free(node); -} - -static struct cso_node * -cso_hash_create_node(struct cso_hash *hash, - unsigned akey, void *avalue, - struct cso_node **anextNode) -{ - struct cso_node *node = cso_data_allocate_node(hash->data.d); - node->key = akey; - node->value = avalue; - - node->next = (struct cso_node*)(*anextNode); - *anextNode = node; - ++hash->data.d->size; - return node; -} - -static void cso_data_rehash(struct cso_hash_data *hash, int hint) -{ - if (hint < 0) { - hint = countBits(-hint); - if (hint < MinNumBits) - hint = MinNumBits; - hash->userNumBits = hint; - while (primeForNumBits(hint) < (hash->size >> 1)) - ++hint; - } else if (hint < MinNumBits) { - hint = MinNumBits; - } - - if (hash->numBits != hint) { - struct cso_node *e = (struct cso_node *)(hash); - struct cso_node **oldBuckets = hash->buckets; - int oldNumBuckets = hash->numBuckets; - int i = 0; - - hash->numBits = hint; - hash->numBuckets = primeForNumBits(hint); - hash->buckets = malloc(sizeof(struct cso_node*) * hash->numBuckets); - for (i = 0; i < hash->numBuckets; ++i) - hash->buckets[i] = e; - - for (i = 0; i < oldNumBuckets; ++i) { - struct cso_node *firstNode = oldBuckets[i]; - while (firstNode != e) { - unsigned h = firstNode->key; - struct cso_node *lastNode = firstNode; - while (lastNode->next != e && lastNode->next->key == h) - lastNode = lastNode->next; - - struct cso_node *afterLastNode = lastNode->next; - struct cso_node **beforeFirstNode = &hash->buckets[h % hash->numBuckets]; - while (*beforeFirstNode != e) - beforeFirstNode = &(*beforeFirstNode)->next; - lastNode->next = *beforeFirstNode; - *beforeFirstNode = firstNode; - firstNode = afterLastNode; - } - } - free(oldBuckets); - } -} - -static void cso_data_might_grow(struct cso_hash_data *hash) -{ - if (hash->size >= hash->numBuckets) - cso_data_rehash(hash, hash->numBits + 1); -} - -static void cso_data_has_shrunk(struct cso_hash_data *hash) -{ - if (hash->size <= (hash->numBuckets >> 3) && - hash->numBits > hash->userNumBits) { - int max = MAX(hash->numBits-2, hash->userNumBits); - cso_data_rehash(hash, max); - } -} - -static struct cso_node *cso_data_first_node(struct cso_hash_data *hash) -{ - struct cso_node *e = (struct cso_node *)(hash); - struct cso_node **bucket = hash->buckets; - int n = hash->numBuckets; - while (n--) { - if (*bucket != e) - return *bucket; - ++bucket; - } - return e; -} - -static struct cso_node **cso_hash_find_node(struct cso_hash *hash, unsigned akey) -{ - struct cso_node **node; - - if (hash->data.d->numBuckets) { - node = (struct cso_node **)(&hash->data.d->buckets[akey % hash->data.d->numBuckets]); - assert(*node == hash->data.e || (*node)->next); - while (*node != hash->data.e && (*node)->key != akey) - node = &(*node)->next; - } else { - node = (struct cso_node **)((const struct cso_node * const *)(&hash->data.e)); - } - return node; -} - -struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, - unsigned key, void *data) -{ - cso_data_might_grow(hash->data.d); - - struct cso_node **nextNode = cso_hash_find_node(hash, key); - struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); - struct cso_hash_iter iter = {hash, node}; - return iter; -} - -struct cso_hash * cso_hash_create(void) -{ - struct cso_hash *hash = malloc(sizeof(struct cso_hash)); - hash->data.d = malloc(sizeof(struct cso_hash_data)); - hash->data.d->fakeNext = 0; - hash->data.d->buckets = 0; - hash->data.d->size = 0; - hash->data.d->nodeSize = sizeof(struct cso_node); - hash->data.d->userNumBits = MinNumBits; - hash->data.d->numBits = 0; - hash->data.d->numBuckets = 0; - - return hash; -} - -void cso_hash_delete(struct cso_hash *hash) -{ - struct cso_node *e_for_x = (struct cso_node *)(hash->data.d); - struct cso_node **bucket = (struct cso_node **)(hash->data.d->buckets); - int n = hash->data.d->numBuckets; - while (n--) { - struct cso_node *cur = *bucket++; - while (cur != e_for_x) { - struct cso_node *next = cur->next; - cso_data_free_node(cur); - cur = next; - } - } - free(hash->data.d->buckets); - free(hash->data.d); - free(hash); -} - -struct cso_hash_iter cso_hash_find(struct cso_hash *hash, - unsigned key) -{ - struct cso_node **nextNode = cso_hash_find_node(hash, key); - struct cso_hash_iter iter = {hash, *nextNode}; - return iter; -} - -unsigned cso_hash_iter_key(struct cso_hash_iter iter) -{ - if (!iter.node || iter.hash->data.e == iter.node) - return 0; - return iter.node->key; -} - -void * cso_hash_iter_data(struct cso_hash_iter iter) -{ - if (!iter.node || iter.hash->data.e == iter.node) - return 0; - return iter.node->value; -} - -static struct cso_node *cso_hash_data_next(struct cso_node *node) -{ - union { - struct cso_node *next; - struct cso_node *e; - struct cso_hash_data *d; - } a; - a.next = node->next; - if (!a.next) { - fprintf(stderr, "iterating beyond the last element\n"); - return 0; - } - if (a.next->next) - return a.next; - - int start = (node->key % a.d->numBuckets) + 1; - struct cso_node **bucket = a.d->buckets + start; - int n = a.d->numBuckets - start; - while (n--) { - if (*bucket != a.e) - return *bucket; - ++bucket; - } - return a.e; -} - - -static struct cso_node *cso_hash_data_prev(struct cso_node *node) -{ - union { - struct cso_node *e; - struct cso_hash_data *d; - } a; - - a.e = node; - while (a.e->next) - a.e = a.e->next; - - int start; - if (node == a.e) - start = a.d->numBuckets - 1; - else - start = node->key % a.d->numBuckets; - - struct cso_node *sentinel = node; - struct cso_node **bucket = a.d->buckets + start; - while (start >= 0) { - if (*bucket != sentinel) { - struct cso_node *prev = *bucket; - while (prev->next != sentinel) - prev = prev->next; - return prev; - } - - sentinel = a.e; - --bucket; - --start; - } - fprintf(stderr, "iterating backward beyond first element\n"); - return a.e; -} - -struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter) -{ - struct cso_hash_iter next = {iter.hash, cso_hash_data_next(iter.node)}; - return next; -} - -int cso_hash_iter_is_null(struct cso_hash_iter iter) -{ - if (!iter.node || iter.node == iter.hash->data.e) - return 1; - return 0; -} - -void * cso_hash_take(struct cso_hash *hash, - unsigned akey) -{ - struct cso_node **node = cso_hash_find_node(hash, akey); - if (*node != hash->data.e) { - void *t = (*node)->value; - struct cso_node *next = (*node)->next; - cso_data_free_node(*node); - *node = next; - --hash->data.d->size; - cso_data_has_shrunk(hash->data.d); - return t; - } - return 0; -} - -struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter) -{ - struct cso_hash_iter prev = {iter.hash, - cso_hash_data_prev(iter.node)}; - return prev; -} - -struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash) -{ - struct cso_hash_iter iter = {hash, cso_data_first_node(hash->data.d)}; - return iter; -} diff --git a/src/gallium/aux/cso_cache/cso_hash.h b/src/gallium/aux/cso_cache/cso_hash.h deleted file mode 100644 index b4aa111860..0000000000 --- a/src/gallium/aux/cso_cache/cso_hash.h +++ /dev/null @@ -1,62 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#ifndef CSO_HASH_H -#define CSO_HASH_H - -struct cso_hash; -struct cso_node; - -struct cso_hash_iter { - struct cso_hash *hash; - struct cso_node *node; -}; - -struct cso_hash *cso_hash_create(void); -void cso_hash_delete(struct cso_hash *hash); - -struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, - void *data); -void *cso_hash_take(struct cso_hash *hash, unsigned key); - -struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash); -struct cso_hash_iter cso_hash_find(struct cso_hash *hash, unsigned key); - - -int cso_hash_iter_is_null(struct cso_hash_iter iter); -unsigned cso_hash_iter_key(struct cso_hash_iter iter); -void *cso_hash_iter_data(struct cso_hash_iter iter); - -struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter); -struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); - -#endif diff --git a/src/gallium/aux/draw/Makefile b/src/gallium/aux/draw/Makefile deleted file mode 100644 index 451911a354..0000000000 --- a/src/gallium/aux/draw/Makefile +++ /dev/null @@ -1,2 +0,0 @@ -default: - cd .. ; make diff --git a/src/gallium/aux/draw/draw_clip.c b/src/gallium/aux/draw/draw_clip.c deleted file mode 100644 index e3051507ea..0000000000 --- a/src/gallium/aux/draw/draw_clip.c +++ /dev/null @@ -1,488 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Clipping stage - * - * \author Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" - -#include "draw_context.h" -#include "draw_private.h" - - -#ifndef IS_NEGATIVE -#define IS_NEGATIVE(X) ((X) < 0.0) -#endif - -#ifndef DIFFERENT_SIGNS -#define DIFFERENT_SIGNS(x, y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F) -#endif - -#ifndef MAX_CLIPPED_VERTICES -#define MAX_CLIPPED_VERTICES ((2 * (6 + PIPE_MAX_CLIP_PLANES))+1) -#endif - - - -struct clipper { - struct draw_stage stage; /**< base class */ - - /* Basically duplicate some of the flatshading logic here: - */ - boolean flat; - uint num_color_attribs; - uint color_attribs[4]; /* front/back primary/secondary colors */ - - float (*plane)[4]; -}; - - -/* This is a bit confusing: - */ -static INLINE struct clipper *clipper_stage( struct draw_stage *stage ) -{ - return (struct clipper *)stage; -} - - -#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) - - -/* All attributes are float[4], so this is easy: - */ -static void interp_attr( float *fdst, - float t, - const float *fin, - const float *fout ) -{ - fdst[0] = LINTERP( t, fout[0], fin[0] ); - fdst[1] = LINTERP( t, fout[1], fin[1] ); - fdst[2] = LINTERP( t, fout[2], fin[2] ); - fdst[3] = LINTERP( t, fout[3], fin[3] ); -} - -static void copy_colors( struct draw_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - const struct clipper *clipper = clipper_stage(stage); - uint i; - for (i = 0; i < clipper->num_color_attribs; i++) { - const uint attr = clipper->color_attribs[i]; - COPY_4FV(dst->data[attr], src->data[attr]); - } -} - - - -/* Interpolate between two vertices to produce a third. - */ -static void interp( const struct clipper *clip, - struct vertex_header *dst, - float t, - const struct vertex_header *out, - const struct vertex_header *in ) -{ - const unsigned nr_attrs = clip->stage.draw->num_vs_outputs; - unsigned j; - - /* Vertex header. - */ - { - dst->clipmask = 0; - dst->edgeflag = 0; - dst->pad = 0; - dst->vertex_id = UNDEFINED_VERTEX_ID; - } - - /* Clip coordinates: interpolate normally - */ - { - interp_attr(dst->clip, t, in->clip, out->clip); - } - - /* Do the projective divide and insert window coordinates: - */ - { - const float *pos = dst->clip; - const float *scale = clip->stage.draw->viewport.scale; - const float *trans = clip->stage.draw->viewport.translate; - const float oow = 1.0f / pos[3]; - - dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; - dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; - dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; - dst->data[0][3] = oow; - } - - /* Other attributes - * Note: start at 1 to skip winpos (data[0]) since we just computed - * it above. - */ - for (j = 1; j < nr_attrs; j++) { - interp_attr(dst->data[j], t, in->data[j], out->data[j]); - } -} - - -static void emit_poly( struct draw_stage *stage, - struct vertex_header **inlist, - unsigned n, - const struct prim_header *origPrim) -{ - struct prim_header header; - unsigned i; - - /* later stages may need the determinant, but only the sign matters */ - header.det = origPrim->det; - - for (i = 2; i < n; i++) { - header.v[0] = inlist[i-1]; - header.v[1] = inlist[i]; - header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ - - { - unsigned tmp1 = header.v[1]->edgeflag; - unsigned tmp2 = header.v[2]->edgeflag; - - if (i != n-1) header.v[1]->edgeflag = 0; - if (i != 2) header.v[2]->edgeflag = 0; - - header.edgeflags = ((header.v[0]->edgeflag << 0) | - (header.v[1]->edgeflag << 1) | - (header.v[2]->edgeflag << 2)); - - stage->next->tri( stage->next, &header ); - - header.v[1]->edgeflag = tmp1; - header.v[2]->edgeflag = tmp2; - } - } -} - - - - -/* Clip a triangle against the viewport and user clip planes. - */ -static void -do_clip_tri( struct draw_stage *stage, - struct prim_header *header, - unsigned clipmask ) -{ - struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *a[MAX_CLIPPED_VERTICES]; - struct vertex_header *b[MAX_CLIPPED_VERTICES]; - struct vertex_header **inlist = a; - struct vertex_header **outlist = b; - unsigned tmpnr = 0; - unsigned n = 3; - unsigned i; - - inlist[0] = header->v[0]; - inlist[1] = header->v[1]; - inlist[2] = header->v[2]; - - while (clipmask && n >= 3) { - const unsigned plane_idx = ffs(clipmask)-1; - const float *plane = clipper->plane[plane_idx]; - struct vertex_header *vert_prev = inlist[0]; - float dp_prev = dot4( vert_prev->clip, plane ); - unsigned outcount = 0; - - clipmask &= ~(1<clip, plane ); - - if (!IS_NEGATIVE(dp_prev)) { - outlist[outcount++] = vert_prev; - } - - if (DIFFERENT_SIGNS(dp, dp_prev)) { - struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; - outlist[outcount++] = new_vert; - - if (IS_NEGATIVE(dp)) { - /* Going out of bounds. Avoid division by zero as we - * know dp != dp_prev from DIFFERENT_SIGNS, above. - */ - float t = dp / (dp - dp_prev); - interp( clipper, new_vert, t, vert, vert_prev ); - - /* Force edgeflag true in this case: - */ - new_vert->edgeflag = 1; - } else { - /* Coming back in. - */ - float t = dp_prev / (dp_prev - dp); - interp( clipper, new_vert, t, vert_prev, vert ); - - /* Copy starting vert's edgeflag: - */ - new_vert->edgeflag = vert_prev->edgeflag; - } - } - - vert_prev = vert; - dp_prev = dp; - } - - { - struct vertex_header **tmp = inlist; - inlist = outlist; - outlist = tmp; - n = outcount; - } - } - - /* If flat-shading, copy color to new provoking vertex. - */ - if (clipper->flat && inlist[0] != header->v[2]) { - if (1) { - inlist[0] = dup_vert(stage, inlist[0], tmpnr++); - } - - copy_colors(stage, inlist[0], header->v[2]); - } - - - - /* Emit the polygon as triangles to the setup stage: - */ - if (n >= 3) - emit_poly( stage, inlist, n, header ); -} - - -/* Clip a line against the viewport and user clip planes. - */ -static void -do_clip_line( struct draw_stage *stage, - struct prim_header *header, - unsigned clipmask ) -{ - const struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - const float *pos0 = v0->clip; - const float *pos1 = v1->clip; - float t0 = 0.0F; - float t1 = 0.0F; - struct prim_header newprim; - - while (clipmask) { - const unsigned plane_idx = ffs(clipmask)-1; - const float *plane = clipper->plane[plane_idx]; - const float dp0 = dot4( pos0, plane ); - const float dp1 = dot4( pos1, plane ); - - if (dp1 < 0.0F) { - float t = dp1 / (dp1 - dp0); - t1 = MAX2(t1, t); - } - - if (dp0 < 0.0F) { - float t = dp0 / (dp0 - dp1); - t0 = MAX2(t0, t); - } - - if (t0 + t1 >= 1.0F) - return; /* discard */ - - clipmask &= ~(1 << plane_idx); /* turn off this plane's bit */ - } - - if (v0->clipmask) { - interp( clipper, stage->tmp[0], t0, v0, v1 ); - - if (clipper->flat) - copy_colors(stage, stage->tmp[0], v0); - - newprim.v[0] = stage->tmp[0]; - } - else { - newprim.v[0] = v0; - } - - if (v1->clipmask) { - interp( clipper, stage->tmp[1], t1, v1, v0 ); - newprim.v[1] = stage->tmp[1]; - } - else { - newprim.v[1] = v1; - } - - stage->next->line( stage->next, &newprim ); -} - - -static void -clip_point( struct draw_stage *stage, - struct prim_header *header ) -{ - if (header->v[0]->clipmask == 0) - stage->next->point( stage->next, header ); -} - - -static void -clip_line( struct draw_stage *stage, - struct prim_header *header ) -{ - unsigned clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask); - - if (clipmask == 0) { - /* no clipping needed */ - stage->next->line( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask) == 0) { - do_clip_line(stage, header, clipmask); - } - /* else, totally clipped */ -} - - -static void -clip_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - unsigned clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask | - header->v[2]->clipmask); - - if (clipmask == 0) { - /* no clipping needed */ - stage->next->tri( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask & - header->v[2]->clipmask) == 0) { - do_clip_tri(stage, header, clipmask); - } -} - -/* Update state. Could further delay this until we hit the first - * primitive that really requires clipping. - */ -static void -clip_init_state( struct draw_stage *stage ) -{ - struct clipper *clipper = clipper_stage( stage ); - - clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; - - if (clipper->flat) { - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; - uint i; - - clipper->num_color_attribs = 0; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || - vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - clipper->color_attribs[clipper->num_color_attribs++] = i; - } - } - } - - stage->tri = clip_tri; - stage->line = clip_line; -} - - - -static void clip_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - clip_init_state( stage ); - stage->tri( stage, header ); -} - -static void clip_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - clip_init_state( stage ); - stage->line( stage, header ); -} - - -static void clip_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->tri = clip_first_tri; - stage->line = clip_first_line; - stage->next->flush( stage->next, flags ); -} - - -static void clip_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void clip_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Allocate a new clipper stage. - * \return pointer to new stage object - */ -struct draw_stage *draw_clip_stage( struct draw_context *draw ) -{ - struct clipper *clipper = CALLOC_STRUCT(clipper); - - draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); - - clipper->stage.draw = draw; - clipper->stage.point = clip_point; - clipper->stage.line = clip_first_line; - clipper->stage.tri = clip_first_tri; - clipper->stage.flush = clip_flush; - clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; - clipper->stage.destroy = clip_destroy; - - clipper->plane = draw->plane; - - return &clipper->stage; -} diff --git a/src/gallium/aux/draw/draw_context.c b/src/gallium/aux/draw/draw_context.c deleted file mode 100644 index 4be3830316..0000000000 --- a/src/gallium/aux/draw/draw_context.c +++ /dev/null @@ -1,293 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "draw_context.h" -#include "draw_private.h" - - - -struct draw_context *draw_create( void ) -{ - struct draw_context *draw = CALLOC_STRUCT( draw_context ); - -#if defined(__i386__) || defined(__386__) - draw->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; -#else - draw->use_sse = FALSE; -#endif - - /* create pipeline stages */ - draw->pipeline.wide = draw_wide_stage( draw ); - draw->pipeline.stipple = draw_stipple_stage( draw ); - draw->pipeline.unfilled = draw_unfilled_stage( draw ); - draw->pipeline.twoside = draw_twoside_stage( draw ); - draw->pipeline.offset = draw_offset_stage( draw ); - draw->pipeline.clip = draw_clip_stage( draw ); - draw->pipeline.flatshade = draw_flatshade_stage( draw ); - draw->pipeline.cull = draw_cull_stage( draw ); - draw->pipeline.validate = draw_validate_stage( draw ); - draw->pipeline.first = draw->pipeline.validate; - - ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); - ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); - ASSIGN_4V( draw->plane[2], 0, -1, 0, 1 ); - ASSIGN_4V( draw->plane[3], 0, 1, 0, 1 ); - ASSIGN_4V( draw->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ - ASSIGN_4V( draw->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ - draw->nr_planes = 6; - - /* Statically allocate maximum sized vertices for the cache - could be cleverer... - */ - { - uint i; - const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; - char *tmp = align_malloc(Elements(draw->vcache.vertex) * size, 16); - - for (i = 0; i < Elements(draw->vcache.vertex); i++) - draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * size); - } - - draw->shader_queue_flush = draw_vertex_shader_queue_flush; - - draw->convert_wide_points = TRUE; - draw->convert_wide_lines = TRUE; - - draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ - - draw_vertex_cache_invalidate( draw ); - draw_set_mapped_element_buffer( draw, 0, NULL ); - - return draw; -} - - -void draw_destroy( struct draw_context *draw ) -{ - draw->pipeline.wide->destroy( draw->pipeline.wide ); - draw->pipeline.stipple->destroy( draw->pipeline.stipple ); - draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); - draw->pipeline.twoside->destroy( draw->pipeline.twoside ); - draw->pipeline.offset->destroy( draw->pipeline.offset ); - draw->pipeline.clip->destroy( draw->pipeline.clip ); - draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); - draw->pipeline.cull->destroy( draw->pipeline.cull ); - draw->pipeline.validate->destroy( draw->pipeline.validate ); - if (draw->pipeline.rasterize) - draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); - tgsi_exec_machine_free_data(&draw->machine); - align_free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ - FREE( draw ); -} - - - -void draw_flush( struct draw_context *draw ) -{ - draw_do_flush( draw, DRAW_FLUSH_BACKEND ); -} - - - -/** - * Register new primitive rasterization/rendering state. - * This causes the drawing pipeline to be rebuilt. - */ -void draw_set_rasterizer_state( struct draw_context *draw, - const struct pipe_rasterizer_state *raster ) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - - draw->rasterizer = raster; -} - - -/** - * Plug in the primitive rendering/rasterization stage (which is the last - * stage in the drawing pipeline). - * This is provided by the device driver. - */ -void draw_set_rasterize_stage( struct draw_context *draw, - struct draw_stage *stage ) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - - draw->pipeline.rasterize = stage; -} - - -/** - * Set the draw module's clipping state. - */ -void draw_set_clip_state( struct draw_context *draw, - const struct pipe_clip_state *clip ) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - - assert(clip->nr <= PIPE_MAX_CLIP_PLANES); - memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); - draw->nr_planes = 6 + clip->nr; -} - - -/** - * Set the draw module's viewport state. - */ -void draw_set_viewport_state( struct draw_context *draw, - const struct pipe_viewport_state *viewport ) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->viewport = *viewport; /* struct copy */ -} - - - -void -draw_set_vertex_buffer(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_buffer *buffer) -{ - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - assert(attr < PIPE_ATTRIB_MAX); - draw->vertex_buffer[attr] = *buffer; -} - - -void -draw_set_vertex_element(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_element *element) -{ - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - assert(attr < PIPE_ATTRIB_MAX); - draw->vertex_element[attr] = *element; -} - - -/** - * Tell drawing context where to find mapped vertex buffers. - */ -void -draw_set_mapped_vertex_buffer(struct draw_context *draw, - unsigned attr, const void *buffer) -{ - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - draw->user.vbuffer[attr] = buffer; -} - - -void -draw_set_mapped_constant_buffer(struct draw_context *draw, - const void *buffer) -{ - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - draw->user.constants = buffer; -} - - -/** - * Tells the draw module whether to convert wide points (size != 1) - * into triangles. - */ -void -draw_convert_wide_points(struct draw_context *draw, boolean enable) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->convert_wide_points = enable; -} - - -/** - * Tells the draw module whether to convert wide lines (width != 1) - * into triangles. - */ -void -draw_convert_wide_lines(struct draw_context *draw, boolean enable) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->convert_wide_lines = enable; -} - - -/** - * Allocate space for temporary post-transform vertices, such as for clipping. - */ -void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) -{ - assert(!stage->tmp); - - stage->nr_tmps = nr; - - if (nr) { - ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); - unsigned i; - - stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); - } -} - - -void draw_free_temp_verts( struct draw_stage *stage ) -{ - if (stage->tmp) { - FREE( stage->tmp[0] ); - FREE( stage->tmp ); - stage->tmp = NULL; - } -} - - -boolean draw_use_sse(struct draw_context *draw) -{ - return (boolean) draw->use_sse; -} - - -void draw_reset_vertex_ids(struct draw_context *draw) -{ - struct draw_stage *stage = draw->pipeline.first; - - while (stage) { - unsigned i; - - for (i = 0; i < stage->nr_tmps; i++) - stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; - - stage = stage->next; - } - - draw_vertex_cache_reset_vertex_ids(draw); -} diff --git a/src/gallium/aux/draw/draw_context.h b/src/gallium/aux/draw/draw_context.h deleted file mode 100644 index ddeb184497..0000000000 --- a/src/gallium/aux/draw/draw_context.h +++ /dev/null @@ -1,142 +0,0 @@ - -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Public interface into the drawing module. - */ - -/* Authors: Keith Whitwell - */ - - -#ifndef DRAW_CONTEXT_H -#define DRAW_CONTEXT_H - - -#include "pipe/p_state.h" - - -struct vertex_buffer; -struct vertex_info; -struct draw_context; -struct draw_stage; -struct draw_vertex_shader; - - -/** - * Clipmask flags - */ -/*@{*/ -#define CLIP_RIGHT_BIT 0x01 -#define CLIP_LEFT_BIT 0x02 -#define CLIP_TOP_BIT 0x04 -#define CLIP_BOTTOM_BIT 0x08 -#define CLIP_NEAR_BIT 0x10 -#define CLIP_FAR_BIT 0x20 -/*@}*/ - -/** - * Bitshift for each clip flag - */ -/*@{*/ -#define CLIP_RIGHT_SHIFT 0 -#define CLIP_LEFT_SHIFT 1 -#define CLIP_TOP_SHIFT 2 -#define CLIP_BOTTOM_SHIFT 3 -#define CLIP_NEAR_SHIFT 4 -#define CLIP_FAR_SHIFT 5 -/*@}*/ - - -struct draw_context *draw_create( void ); - -void draw_destroy( struct draw_context *draw ); - -void draw_set_viewport_state( struct draw_context *draw, - const struct pipe_viewport_state *viewport ); - -void draw_set_clip_state( struct draw_context *pipe, - const struct pipe_clip_state *clip ); - -void draw_set_rasterizer_state( struct draw_context *draw, - const struct pipe_rasterizer_state *raster ); - -void draw_set_rasterize_stage( struct draw_context *draw, - struct draw_stage *stage ); - -void draw_convert_wide_points(struct draw_context *draw, boolean enable); - -void draw_convert_wide_lines(struct draw_context *draw, boolean enable); - - -struct draw_vertex_shader * -draw_create_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader); -void draw_bind_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs); -void draw_delete_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs); - -boolean draw_use_sse(struct draw_context *draw); - -void draw_set_vertex_buffer(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_buffer *buffer); - -void draw_set_vertex_element(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_element *element); - -void draw_set_mapped_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ); - -void draw_set_mapped_vertex_buffer(struct draw_context *draw, - unsigned attr, const void *buffer); - -void draw_set_mapped_constant_buffer(struct draw_context *draw, - const void *buffer); - - -/*********************************************************************** - * draw_prim.c - */ - -void draw_arrays(struct draw_context *draw, unsigned prim, - unsigned start, unsigned count); - -void draw_flush(struct draw_context *draw); - -/*********************************************************************** - * draw_debug.c - */ -boolean draw_validate_prim( unsigned prim, unsigned length ); -unsigned draw_trim_prim( unsigned mode, unsigned count ); - - - -#endif /* DRAW_CONTEXT_H */ diff --git a/src/gallium/aux/draw/draw_cull.c b/src/gallium/aux/draw/draw_cull.c deleted file mode 100644 index 8177b0ac86..0000000000 --- a/src/gallium/aux/draw/draw_cull.c +++ /dev/null @@ -1,150 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Drawing stage for polygon culling - */ - -/* Authors: Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "draw_private.h" - - -struct cull_stage { - struct draw_stage stage; - unsigned winding; /**< which winding(s) to cull (one of PIPE_WINDING_x) */ -}; - - -static INLINE struct cull_stage *cull_stage( struct draw_stage *stage ) -{ - return (struct cull_stage *)stage; -} - - - - -static void cull_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - /* Window coords: */ - const float *v0 = header->v[0]->data[0]; - const float *v1 = header->v[1]->data[0]; - const float *v2 = header->v[2]->data[0]; - - /* edge vectors e = v0 - v2, f = v1 - v2 */ - const float ex = v0[0] - v2[0]; - const float ey = v0[1] - v2[1]; - const float fx = v1[0] - v2[0]; - const float fy = v1[1] - v2[1]; - - /* det = cross(e,f).z */ - header->det = ex * fy - ey * fx; - - if (header->det != 0) { - /* if (det < 0 then Z points toward camera and triangle is - * counter-clockwise winding. - */ - unsigned winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; - - if ((winding & cull_stage(stage)->winding) == 0) { - /* triangle is not culled, pass to next stage */ - stage->next->tri( stage->next, header ); - } - } -} - -static void cull_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct cull_stage *cull = cull_stage(stage); - - cull->winding = stage->draw->rasterizer->cull_mode; - - stage->tri = cull_tri; - stage->tri( stage, header ); -} - - - -static void cull_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void cull_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void cull_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->tri = cull_first_tri; - stage->next->flush( stage->next, flags ); -} - -static void cull_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void cull_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create a new polygon culling stage. - */ -struct draw_stage *draw_cull_stage( struct draw_context *draw ) -{ - struct cull_stage *cull = CALLOC_STRUCT(cull_stage); - - draw_alloc_temp_verts( &cull->stage, 0 ); - - cull->stage.draw = draw; - cull->stage.next = NULL; - cull->stage.point = cull_point; - cull->stage.line = cull_line; - cull->stage.tri = cull_first_tri; - cull->stage.flush = cull_flush; - cull->stage.reset_stipple_counter = cull_reset_stipple_counter; - cull->stage.destroy = cull_destroy; - - return &cull->stage; -} diff --git a/src/gallium/aux/draw/draw_debug.c b/src/gallium/aux/draw/draw_debug.c deleted file mode 100644 index d6220b5f62..0000000000 --- a/src/gallium/aux/draw/draw_debug.c +++ /dev/null @@ -1,113 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "draw_private.h" -#include "draw_context.h" - - - -static void -draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) -{ - assert(prim >= PIPE_PRIM_POINTS); - assert(prim <= PIPE_PRIM_POLYGON); - - switch (prim) { - case PIPE_PRIM_POINTS: - *first = 1; - *incr = 1; - break; - case PIPE_PRIM_LINES: - *first = 2; - *incr = 2; - break; - case PIPE_PRIM_LINE_STRIP: - *first = 2; - *incr = 1; - break; - case PIPE_PRIM_LINE_LOOP: - *first = 2; - *incr = 1; - break; - case PIPE_PRIM_TRIANGLES: - *first = 3; - *incr = 3; - break; - case PIPE_PRIM_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - break; - case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: - *first = 3; - *incr = 1; - break; - case PIPE_PRIM_QUADS: - *first = 4; - *incr = 4; - break; - case PIPE_PRIM_QUAD_STRIP: - *first = 4; - *incr = 2; - break; - default: - assert(0); - *first = 1; - *incr = 1; - break; - } -} - - -unsigned -draw_trim_prim( unsigned mode, unsigned count ) -{ - unsigned length, first, incr; - - draw_prim_info( mode, &first, &incr ); - - if (count < first) - length = 0; - else - length = count - (count - first) % incr; - - return length; -} - - -boolean -draw_validate_prim( unsigned mode, unsigned count ) -{ - return (count > 0 && - count == draw_trim_prim( mode, count )); -} - diff --git a/src/gallium/aux/draw/draw_flatshade.c b/src/gallium/aux/draw/draw_flatshade.c deleted file mode 100644 index 4398abbc60..0000000000 --- a/src/gallium/aux/draw/draw_flatshade.c +++ /dev/null @@ -1,205 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -/** subclass of draw_stage */ -struct flat_stage -{ - struct draw_stage stage; - - uint num_color_attribs; - uint color_attribs[4]; /* front/back primary/secondary colors */ -}; - - -static INLINE struct flat_stage * -flat_stage(struct draw_stage *stage) -{ - return (struct flat_stage *) stage; -} - - -/** Copy all the color attributes from 'src' vertex to 'dst' vertex */ -static INLINE void copy_colors( struct draw_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - const struct flat_stage *flat = flat_stage(stage); - uint i; - for (i = 0; i < flat->num_color_attribs; i++) { - const uint attr = flat->color_attribs[i]; - COPY_4FV(dst->data[attr], src->data[attr]); - } -} - - -/** Copy all the color attributes from src vertex to dst0 & dst1 vertices */ -static INLINE void copy_colors2( struct draw_stage *stage, - struct vertex_header *dst0, - struct vertex_header *dst1, - const struct vertex_header *src ) -{ - const struct flat_stage *flat = flat_stage(stage); - uint i; - for (i = 0; i < flat->num_color_attribs; i++) { - const uint attr = flat->color_attribs[i]; - COPY_4FV(dst0->data[attr], src->data[attr]); - COPY_4FV(dst1->data[attr], src->data[attr]); - } -} - - -/** - * Flatshade tri. Required for clipping and when unfilled tris are - * active, otherwise handled by hardware. - */ -static void flatshade_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = header->v[2]; - - copy_colors2(stage, tmp.v[0], tmp.v[1], tmp.v[2]); - - stage->next->tri( stage->next, &tmp ); -} - - -/** - * Flatshade line. Required for clipping. - */ -static void flatshade_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = header->v[1]; - - copy_colors(stage, tmp.v[0], tmp.v[1]); - - stage->next->line( stage->next, &tmp ); -} - - -static void flatshade_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void flatshade_init_state( struct draw_stage *stage ) -{ - struct flat_stage *flat = flat_stage(stage); - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; - uint i; - - /* Find which vertex shader outputs are colors, make a list */ - flat->num_color_attribs = 0; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || - vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - flat->color_attribs[flat->num_color_attribs++] = i; - } - } - - stage->line = flatshade_line; - stage->tri = flatshade_tri; -} - -static void flatshade_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - flatshade_init_state( stage ); - stage->tri( stage, header ); -} - -static void flatshade_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - flatshade_init_state( stage ); - stage->line( stage, header ); -} - - -static void flatshade_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->tri = flatshade_first_tri; - stage->line = flatshade_first_line; - stage->next->flush( stage->next, flags ); -} - - -static void flatshade_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void flatshade_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create flatshading drawing stage. - */ -struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) -{ - struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage); - - draw_alloc_temp_verts( &flatshade->stage, 2 ); - - flatshade->stage.draw = draw; - flatshade->stage.next = NULL; - flatshade->stage.point = flatshade_point; - flatshade->stage.line = flatshade_first_line; - flatshade->stage.tri = flatshade_first_tri; - flatshade->stage.flush = flatshade_flush; - flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; - flatshade->stage.destroy = flatshade_destroy; - - return &flatshade->stage; -} - - diff --git a/src/gallium/aux/draw/draw_offset.c b/src/gallium/aux/draw/draw_offset.c deleted file mode 100644 index dbc676deae..0000000000 --- a/src/gallium/aux/draw/draw_offset.c +++ /dev/null @@ -1,186 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief polygon offset state - * - * \author Keith Whitwell - * \author Brian Paul - */ - -#include "pipe/p_util.h" -#include "draw_private.h" - - - -struct offset_stage { - struct draw_stage stage; - - float scale; - float units; -}; - - - -static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) -{ - return (struct offset_stage *) stage; -} - - - - - -/** - * Offset tri Z. Some hardware can handle this, but not usually when - * doing unfilled rendering. - */ -static void do_offset_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct offset_stage *offset = offset_stage(stage); - float inv_det = 1.0f / header->det; - - /* Window coords: - */ - float *v0 = header->v[0]->data[0]; - float *v1 = header->v[1]->data[0]; - float *v2 = header->v[2]->data[0]; - - /* edge vectors e = v0 - v2, f = v1 - v2 */ - float ex = v0[0] - v2[0]; - float ey = v0[1] - v2[1]; - float ez = v0[2] - v2[2]; - float fx = v1[0] - v2[0]; - float fy = v1[1] - v2[1]; - float fz = v1[2] - v2[2]; - - /* (a,b) = cross(e,f).xy */ - float a = ey*fz - ez*fy; - float b = ez*fx - ex*fz; - - float dzdx = FABSF(a * inv_det); - float dzdy = FABSF(b * inv_det); - - float zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; - - /* - * Note: we're applying the offset and clamping per-vertex. - * Ideally, the offset is applied per-fragment prior to fragment shading. - */ - v0[2] = CLAMP(v0[2] + zoffset, 0.0f, 1.0f); - v1[2] = CLAMP(v1[2] + zoffset, 0.0f, 1.0f); - v2[2] = CLAMP(v2[2] + zoffset, 0.0f, 1.0f); - - stage->next->tri( stage->next, header ); -} - - -static void offset_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = dup_vert(stage, header->v[2], 2); - - do_offset_tri( stage, &tmp ); -} - - -static void offset_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct offset_stage *offset = offset_stage(stage); - float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ - - offset->units = stage->draw->rasterizer->offset_units * mrd; - offset->scale = stage->draw->rasterizer->offset_scale; - - stage->tri = offset_tri; - stage->tri( stage, header ); -} - - -static void offset_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void offset_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void offset_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->tri = offset_first_tri; - stage->next->flush( stage->next, flags ); -} - - -static void offset_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void offset_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create polygon offset drawing stage. - */ -struct draw_stage *draw_offset_stage( struct draw_context *draw ) -{ - struct offset_stage *offset = CALLOC_STRUCT(offset_stage); - - draw_alloc_temp_verts( &offset->stage, 3 ); - - offset->stage.draw = draw; - offset->stage.next = NULL; - offset->stage.point = offset_point; - offset->stage.line = offset_line; - offset->stage.tri = offset_first_tri; - offset->stage.flush = offset_flush; - offset->stage.reset_stipple_counter = offset_reset_stipple_counter; - offset->stage.destroy = offset_destroy; - - return &offset->stage; -} diff --git a/src/gallium/aux/draw/draw_prim.c b/src/gallium/aux/draw/draw_prim.c deleted file mode 100644 index 51e2242719..0000000000 --- a/src/gallium/aux/draw/draw_prim.c +++ /dev/null @@ -1,482 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_debug.h" - -#include "draw_private.h" -#include "draw_context.h" - - - -#define RP_NONE 0 -#define RP_POINT 1 -#define RP_LINE 2 -#define RP_TRI 3 - - -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { - RP_POINT, - RP_LINE, - RP_LINE, - RP_LINE, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI -}; - - -static void draw_prim_queue_flush( struct draw_context *draw ) -{ - unsigned i; - - if (0) - debug_printf("Flushing with %d prims, %d verts\n", - draw->pq.queue_nr, draw->vs.queue_nr); - - assert (draw->pq.queue_nr != 0); - - /* NOTE: we cannot save draw->pipeline->first in a local var because - * draw->pipeline->first is often changed by the first call to tri(), - * line(), etc. - */ - if (draw->rasterizer->line_stipple_enable) { - switch (draw->reduced_prim) { - case RP_TRI: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - - draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); - } - break; - case RP_LINE: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - - draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); - } - break; - case RP_POINT: - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); - break; - } - } - else { - switch (draw->reduced_prim) { - case RP_TRI: - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); - break; - case RP_LINE: - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); - break; - case RP_POINT: - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); - break; - } - } - - draw->pq.queue_nr = 0; - draw_vertex_cache_unreference( draw ); -} - - - -void draw_do_flush( struct draw_context *draw, unsigned flags ) -{ - if (0) - debug_printf("Flushing with %d verts, %d prims\n", - draw->vs.queue_nr, - draw->pq.queue_nr ); - - - if (flags >= DRAW_FLUSH_SHADER_QUEUE) { - if (draw->vs.queue_nr) - (*draw->shader_queue_flush)(draw); - - if (flags >= DRAW_FLUSH_PRIM_QUEUE) { - if (draw->pq.queue_nr) - draw_prim_queue_flush(draw); - - if (flags >= DRAW_FLUSH_VERTEX_CACHE) { - draw_vertex_cache_invalidate(draw); - - if (flags >= DRAW_FLUSH_STATE_CHANGE) { - draw->pipeline.first->flush( draw->pipeline.first, flags ); - draw->pipeline.first = draw->pipeline.validate; - draw->reduced_prim = ~0; - } - } - } - } -} - - - -/* Return a pointer to a freshly queued primitive header. Ensure that - * there is room in the vertex cache for a maximum of "nr_verts" new - * vertices. Flush primitive and/or vertex queues if necessary to - * make space. - */ -static struct prim_header *get_queued_prim( struct draw_context *draw, - unsigned nr_verts ) -{ - if (!draw_vertex_cache_check_space( draw, nr_verts )) { -// debug_printf("v"); - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE ); - } - else if (draw->pq.queue_nr == PRIM_QUEUE_LENGTH) { -// debug_printf("p"); - draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); - } - - assert(draw->pq.queue_nr < PRIM_QUEUE_LENGTH); - - return &draw->pq.queue[draw->pq.queue_nr++]; -} - - - -/** - * Add a point to the primitive queue. - * \param i0 index into user's vertex arrays - */ -static void do_point( struct draw_context *draw, - unsigned i0 ) -{ - struct prim_header *prim = get_queued_prim( draw, 1 ); - - prim->reset_line_stipple = 0; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->vcache.get_vertex( draw, i0 ); -} - - -/** - * Add a line to the primitive queue. - * \param i0 index into user's vertex arrays - * \param i1 index into user's vertex arrays - */ -static void do_line( struct draw_context *draw, - boolean reset_stipple, - unsigned i0, - unsigned i1 ) -{ - struct prim_header *prim = get_queued_prim( draw, 2 ); - - prim->reset_line_stipple = reset_stipple; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->vcache.get_vertex( draw, i0 ); - prim->v[1] = draw->vcache.get_vertex( draw, i1 ); -} - -/** - * Add a triangle to the primitive queue. - */ -static void do_triangle( struct draw_context *draw, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - - prim->reset_line_stipple = 1; - prim->edgeflags = ~0; - prim->pad = 0; - prim->v[0] = draw->vcache.get_vertex( draw, i0 ); - prim->v[1] = draw->vcache.get_vertex( draw, i1 ); - prim->v[2] = draw->vcache.get_vertex( draw, i2 ); -} - -static void do_ef_triangle( struct draw_context *draw, - boolean reset_stipple, - unsigned ef_mask, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - struct vertex_header *v0 = draw->vcache.get_vertex( draw, i0 ); - struct vertex_header *v1 = draw->vcache.get_vertex( draw, i1 ); - struct vertex_header *v2 = draw->vcache.get_vertex( draw, i2 ); - - prim->reset_line_stipple = reset_stipple; - - prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | - (v1->edgeflag << 1) | - (v2->edgeflag << 2)); - prim->pad = 0; - prim->v[0] = v0; - prim->v[1] = v1; - prim->v[2] = v2; -} - - -static void do_ef_quad( struct draw_context *draw, - unsigned v0, - unsigned v1, - unsigned v2, - unsigned v3 ) -{ - const unsigned omitEdge2 = ~(1 << 1); - const unsigned omitEdge3 = ~(1 << 2); - do_ef_triangle( draw, 1, omitEdge2, v0, v1, v3 ); - do_ef_triangle( draw, 0, omitEdge3, v1, v2, v3 ); -} - -static void do_quad( struct draw_context *draw, - unsigned v0, - unsigned v1, - unsigned v2, - unsigned v3 ) -{ - do_triangle( draw, v0, v1, v3 ); - do_triangle( draw, v1, v2, v3 ); -} - - -/** - * Main entrypoint to draw some number of points/lines/triangles - */ -static void -draw_prim( struct draw_context *draw, - unsigned prim, unsigned start, unsigned count ) -{ - unsigned i; - boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); - -// debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - - switch (prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i ++) { - do_point( draw, - start + i ); - } - break; - - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) { - do_line( draw, - TRUE, - start + i + 0, - start + i + 1); - } - break; - - case PIPE_PRIM_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, /* XXX: only if vb not split */ - start + i - 1, - start + i ); - } - - do_line( draw, - 0, - start + count - 1, - start + 0 ); - } - break; - - case PIPE_PRIM_LINE_STRIP: - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, - start + i - 1, - start + i ); - } - break; - - case PIPE_PRIM_TRIANGLES: - if (unfilled) { - for (i = 0; i+2 < count; i += 3) { - do_ef_triangle( draw, - 1, - ~0, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - else { - for (i = 0; i+2 < count; i += 3) { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - do_triangle( draw, - start + i + 1, - start + i + 0, - start + i + 2 ); - } - else { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - case PIPE_PRIM_TRIANGLE_FAN: - if (count >= 3) { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - - case PIPE_PRIM_QUADS: - if (unfilled) { - for (i = 0; i+3 < count; i += 4) { - do_ef_quad( draw, - start + i + 0, - start + i + 1, - start + i + 2, - start + i + 3); - } - } - else { - for (i = 0; i+3 < count; i += 4) { - do_quad( draw, - start + i + 0, - start + i + 1, - start + i + 2, - start + i + 3); - } - } - break; - - case PIPE_PRIM_QUAD_STRIP: - if (unfilled) { - for (i = 0; i+3 < count; i += 2) { - do_ef_quad( draw, - start + i + 2, - start + i + 0, - start + i + 1, - start + i + 3); - } - } - else { - for (i = 0; i+3 < count; i += 2) { - do_quad( draw, - start + i + 2, - start + i + 0, - start + i + 1, - start + i + 3); - } - } - break; - - case PIPE_PRIM_POLYGON: - if (unfilled) { - unsigned ef_mask = (1<<2) | (1<<0); - - for (i = 0; i+2 < count; i++) { - - if (i + 3 >= count) - ef_mask |= (1<<1); - - do_ef_triangle( draw, - i == 0, - ef_mask, - start + i + 1, - start + i + 2, - start + 0); - - ef_mask &= ~(1<<2); - } - } - else { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + i + 1, - start + i + 2, - start + 0); - } - } - break; - - default: - assert(0); - break; - } -} - - - - -/** - * Draw vertex arrays - * This is the main entrypoint into the drawing module. - * \param prim one of PIPE_PRIM_x - * \param start index of first vertex to draw - * \param count number of vertices to draw - */ -void -draw_arrays(struct draw_context *draw, unsigned prim, - unsigned start, unsigned count) -{ - if (reduced_prim[prim] != draw->reduced_prim) { - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->reduced_prim = reduced_prim[prim]; - } - - /* drawing done here: */ - draw_prim(draw, prim, start, count); -} - - diff --git a/src/gallium/aux/draw/draw_private.h b/src/gallium/aux/draw/draw_private.h deleted file mode 100644 index 3d09aef87c..0000000000 --- a/src/gallium/aux/draw/draw_private.h +++ /dev/null @@ -1,346 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Private data structures, etc for the draw module. - */ - - -/** - * Authors: - * Keith Whitwell - * Brian Paul - */ - - -#ifndef DRAW_PRIVATE_H -#define DRAW_PRIVATE_H - - -#include "pipe/p_state.h" -#include "pipe/p_defines.h" - -#include "x86/rtasm/x86sse.h" -#include "tgsi/exec/tgsi_exec.h" - - -struct gallivm_prog; -struct gallivm_cpu_engine; - -/** - * Basic vertex info. - * Carry some useful information around with the vertices in the prim pipe. - */ -struct vertex_header { - unsigned clipmask:12; - unsigned edgeflag:1; - unsigned pad:3; - unsigned vertex_id:16; - - float clip[4]; - - float data[][4]; /* Note variable size */ -}; - -/* NOTE: It should match vertex_id size above */ -#define UNDEFINED_VERTEX_ID 0xffff - -/* XXX This is too large */ -#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) - - - -/** - * Basic info for a point/line/triangle primitive. - */ -struct prim_header { - float det; /**< front/back face determinant */ - unsigned reset_line_stipple:1; - unsigned edgeflags:3; - unsigned pad:28; - struct vertex_header *v[3]; /**< 1 to 3 vertex pointers */ -}; - - - -struct draw_context; - -/** - * Base class for all primitive drawing stages. - */ -struct draw_stage -{ - struct draw_context *draw; /**< parent context */ - - struct draw_stage *next; /**< next stage in pipeline */ - - struct vertex_header **tmp; /**< temp vert storage, such as for clipping */ - unsigned nr_tmps; - - void (*point)( struct draw_stage *, - struct prim_header * ); - - void (*line)( struct draw_stage *, - struct prim_header * ); - - void (*tri)( struct draw_stage *, - struct prim_header * ); - - void (*flush)( struct draw_stage *, - unsigned flags ); - - void (*reset_stipple_counter)( struct draw_stage * ); - - void (*destroy)( struct draw_stage * ); -}; - - -#define PRIM_QUEUE_LENGTH 16 -#define VCACHE_SIZE 32 -#define VCACHE_OVERFLOW 4 -#define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ - -/** - * Private version of the compiled vertex_shader - */ -struct draw_vertex_shader { - const struct pipe_shader_state *state; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif -#ifdef MESA_LLVM - struct gallivm_prog *llvm_prog; -#endif -}; - - -/* Internal function for vertex fetch. - */ -typedef void (*fetch_func)(const void *ptr, float *attrib); -typedef void (*full_fetch_func)( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ); - - - -/** - * Private context for the drawing module. - */ -struct draw_context -{ - /** Drawing/primitive pipeline stages */ - struct { - struct draw_stage *first; /**< one of the following */ - - struct draw_stage *validate; - - /* stages (in logical order) */ - struct draw_stage *flatshade; - struct draw_stage *clip; - struct draw_stage *cull; - struct draw_stage *twoside; - struct draw_stage *offset; - struct draw_stage *unfilled; - struct draw_stage *stipple; - struct draw_stage *wide; - struct draw_stage *rasterize; - } pipeline; - - /* pipe state that we need: */ - const struct pipe_rasterizer_state *rasterizer; - struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - const struct draw_vertex_shader *vertex_shader; - - uint num_vs_outputs; /**< convenience, from vertex_shader */ - - /* user-space vertex data, buffers */ - struct { - /** vertex element/index buffer (ex: glDrawElements) */ - const void *elts; - /** bytes per index (0, 1, 2 or 4) */ - unsigned eltSize; - - /** vertex arrays */ - const void *vbuffer[PIPE_ATTRIB_MAX]; - - /** constant buffer (for vertex shader) */ - const void *constants; - } user; - - /* Clip derived state: - */ - float plane[12][4]; - unsigned nr_planes; - - boolean convert_wide_points; /**< convert wide points to tris? */ - boolean convert_wide_lines; /**< convert side lines to tris? */ - - unsigned reduced_prim; - - /** TGSI program interpreter runtime state */ - struct tgsi_exec_machine machine; - - /* Vertex fetch internal state - */ - struct { - const ubyte *src_ptr[PIPE_ATTRIB_MAX]; - unsigned pitch[PIPE_ATTRIB_MAX]; - fetch_func fetch[PIPE_ATTRIB_MAX]; - unsigned nr_attrs; - full_fetch_func fetch_func; - } vertex_fetch; - - /* Post-tnl vertex cache: - */ - struct { - unsigned referenced; /**< bitfield */ - unsigned idx[VCACHE_SIZE + VCACHE_OVERFLOW]; - struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; - unsigned overflow; - - /** To find space in the vertex cache: */ - struct vertex_header *(*get_vertex)( struct draw_context *draw, - unsigned i ); - } vcache; - - /* Vertex shader queue: - */ - struct { - struct { - unsigned elt; /**< index into the user's vertex arrays */ - struct vertex_header *dest; /**< points into vcache.vertex[] array */ - } queue[VS_QUEUE_LENGTH]; - unsigned queue_nr; - } vs; - - /** - * Run the vertex shader on all vertices in the vertex queue. - */ - void (*shader_queue_flush)(struct draw_context *draw); - - /* Prim pipeline queue: - */ - struct { - /* Need to queue up primitives until their vertices have been - * transformed by a vs queue flush. - */ - struct prim_header queue[PRIM_QUEUE_LENGTH]; - unsigned queue_nr; - } pq; - - int use_sse : 1; -#ifdef MESA_LLVM - struct gallivm_cpu_engine *engine; -#endif - - void *driver_private; -}; - - - -extern struct draw_stage *draw_unfilled_stage( struct draw_context *context ); -extern struct draw_stage *draw_twoside_stage( struct draw_context *context ); -extern struct draw_stage *draw_offset_stage( struct draw_context *context ); -extern struct draw_stage *draw_clip_stage( struct draw_context *context ); -extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); -extern struct draw_stage *draw_cull_stage( struct draw_context *context ); -extern struct draw_stage *draw_stipple_stage( struct draw_context *context ); -extern struct draw_stage *draw_wide_stage( struct draw_context *context ); -extern struct draw_stage *draw_validate_stage( struct draw_context *context ); - - -extern void draw_free_temp_verts( struct draw_stage *stage ); - -extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); - -extern void draw_reset_vertex_ids( struct draw_context *draw ); - - -extern int draw_vertex_cache_check_space( struct draw_context *draw, - unsigned nr_verts ); - -extern void draw_vertex_cache_invalidate( struct draw_context *draw ); -extern void draw_vertex_cache_unreference( struct draw_context *draw ); -extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); - - -extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); -#ifdef MESA_LLVM -extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw ); -#endif - -struct tgsi_exec_machine; - -extern void draw_update_vertex_fetch( struct draw_context *draw ); - - -#define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ -#define DRAW_FLUSH_PRIM_QUEUE 0x2 -#define DRAW_FLUSH_VERTEX_CACHE 0x4 -#define DRAW_FLUSH_STATE_CHANGE 0x8 -#define DRAW_FLUSH_BACKEND 0x10 - - -void draw_do_flush( struct draw_context *draw, unsigned flags ); - - - -/** - * Get a writeable copy of a vertex. - * \param stage drawing stage info - * \param vert the vertex to copy (source) - * \param idx index into stage's tmp[] array to put the copy (dest) - * \return pointer to the copied vertex - */ -static INLINE struct vertex_header * -dup_vert( struct draw_stage *stage, - const struct vertex_header *vert, - unsigned idx ) -{ - struct vertex_header *tmp = stage->tmp[idx]; - const uint vsize = sizeof(struct vertex_header) - + stage->draw->num_vs_outputs * 4 * sizeof(float); - memcpy(tmp, vert, vsize); - tmp->vertex_id = UNDEFINED_VERTEX_ID; - return tmp; -} - -static INLINE float -dot4(const float *a, const float *b) -{ - float result = (a[0]*b[0] + - a[1]*b[1] + - a[2]*b[2] + - a[3]*b[3]); - - return result; -} - -#endif /* DRAW_PRIVATE_H */ diff --git a/src/gallium/aux/draw/draw_stipple.c b/src/gallium/aux/draw/draw_stipple.c deleted file mode 100644 index 506f33512c..0000000000 --- a/src/gallium/aux/draw/draw_stipple.c +++ /dev/null @@ -1,239 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -/* Implement line stipple by cutting lines up into smaller lines. - * There are hundreds of ways to implement line stipple, this is one - * choice that should work in all situations, requires no state - * manipulations, but with a penalty in terms of large amounts of - * generated geometry. - */ - - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -/** Subclass of draw_stage */ -struct stipple_stage { - struct draw_stage stage; - float counter; - uint pattern; - uint factor; -}; - - -static INLINE struct stipple_stage * -stipple_stage(struct draw_stage *stage) -{ - return (struct stipple_stage *) stage; -} - - -/** - * Compute interpolated vertex attributes for 'dst' at position 't' - * between 'v0' and 'v1'. - * XXX using linear interpolation for all attribs at this time. - */ -static void -screen_interp( struct draw_context *draw, - struct vertex_header *dst, - float t, - const struct vertex_header *v0, - const struct vertex_header *v1 ) -{ - uint attr; - for (attr = 0; attr < draw->num_vs_outputs; attr++) { - const float *val0 = v0->data[attr]; - const float *val1 = v1->data[attr]; - float *newv = dst->data[attr]; - uint i; - for (i = 0; i < 4; i++) { - newv[i] = val0[i] + t * (val1[i] - val0[i]); - } - } -} - - -static void -emit_segment(struct draw_stage *stage, struct prim_header *header, - float t0, float t1) -{ - struct vertex_header *v0new = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1new = dup_vert(stage, header->v[1], 1); - struct prim_header newprim = *header; - - if (t0 > 0.0) { - screen_interp( stage->draw, v0new, t0, header->v[0], header->v[1] ); - newprim.v[0] = v0new; - } - - if (t1 < 1.0) { - screen_interp( stage->draw, v1new, t1, header->v[0], header->v[1] ); - newprim.v[1] = v1new; - } - - stage->next->line( stage->next, &newprim ); -} - - -static INLINE unsigned -stipple_test(int counter, ushort pattern, int factor) -{ - int b = (counter / factor) & 0xf; - return (1 << b) & pattern; -} - - -static void -stipple_line(struct draw_stage *stage, struct prim_header *header) -{ - struct stipple_stage *stipple = stipple_stage(stage); - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - const float *pos0 = v0->data[0]; - const float *pos1 = v1->data[0]; - float start = 0; - int state = 0; - - float x0 = pos0[0]; - float x1 = pos1[0]; - float y0 = pos0[1]; - float y1 = pos1[1]; - - float dx = x0 > x1 ? x0 - x1 : x1 - x0; - float dy = y0 > y1 ? y0 - y1 : y1 - y0; - - float length = MAX2(dx, dy); - int i; - - /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table. - */ - for (i = 0; i < length; i++) { - int result = stipple_test( (int) stipple->counter+i, - (ushort) stipple->pattern, stipple->factor ); - if (result != state) { - /* changing from "off" to "on" or vice versa */ - if (state) { - if (start != i) { - /* finishing an "on" segment */ - emit_segment( stage, header, start / length, i / length ); - } - } - else { - /* starting an "on" segment */ - start = (float) i; - } - state = result; - } - } - - if (state && start < length) - emit_segment( stage, header, start / length, 1.0 ); - - stipple->counter += length; -} - - -static void -reset_stipple_counter(struct draw_stage *stage) -{ - struct stipple_stage *stipple = stipple_stage(stage); - stipple->counter = 0; - stage->next->reset_stipple_counter( stage->next ); -} - - -static void -stipple_first_line(struct draw_stage *stage, - struct prim_header *header) -{ - struct stipple_stage *stipple = stipple_stage(stage); - struct draw_context *draw = stage->draw; - - stipple->pattern = draw->rasterizer->line_stipple_pattern; - stipple->factor = draw->rasterizer->line_stipple_factor + 1; - - stage->line = stipple_line; - stage->line( stage, header ); -} - - -static void -stipple_flush(struct draw_stage *stage, unsigned flags) -{ - stage->line = stipple_first_line; - stage->next->flush( stage->next, flags ); -} - - -static void -passthrough_point(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->point( stage->next, header ); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} - - -static void -stipple_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create line stippler stage - */ -struct draw_stage *draw_stipple_stage( struct draw_context *draw ) -{ - struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); - - draw_alloc_temp_verts( &stipple->stage, 2 ); - - stipple->stage.draw = draw; - stipple->stage.next = NULL; - stipple->stage.point = passthrough_point; - stipple->stage.line = stipple_first_line; - stipple->stage.tri = passthrough_tri; - stipple->stage.reset_stipple_counter = reset_stipple_counter; - stipple->stage.flush = stipple_flush; - stipple->stage.destroy = stipple_destroy; - - return &stipple->stage; -} diff --git a/src/gallium/aux/draw/draw_twoside.c b/src/gallium/aux/draw/draw_twoside.c deleted file mode 100644 index 1c38957987..0000000000 --- a/src/gallium/aux/draw/draw_twoside.c +++ /dev/null @@ -1,203 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -struct twoside_stage { - struct draw_stage stage; - float sign; /**< +1 or -1 */ - uint attrib_front0, attrib_back0; - uint attrib_front1, attrib_back1; -}; - - -static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage ) -{ - return (struct twoside_stage *)stage; -} - - - - -/** - * Copy back color(s) to front color(s). - */ -static INLINE struct vertex_header * -copy_bfc( struct twoside_stage *twoside, - const struct vertex_header *v, - unsigned idx ) -{ - struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); - - if (twoside->attrib_back0) { - COPY_4FV(tmp->data[twoside->attrib_front0], - tmp->data[twoside->attrib_back0]); - } - if (twoside->attrib_back1) { - COPY_4FV(tmp->data[twoside->attrib_front1], - tmp->data[twoside->attrib_back1]); - } - - return tmp; -} - - -/* Twoside tri: - */ -static void twoside_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - - if (header->det * twoside->sign < 0.0) { - /* this is a back-facing triangle */ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - /* copy back attribs to front attribs */ - tmp.v[0] = copy_bfc(twoside, header->v[0], 0); - tmp.v[1] = copy_bfc(twoside, header->v[1], 1); - tmp.v[2] = copy_bfc(twoside, header->v[2], 2); - - stage->next->tri( stage->next, &tmp ); - } - else { - stage->next->tri( stage->next, header ); - } -} - - -static void twoside_line( struct draw_stage *stage, - struct prim_header *header ) -{ - /* pass-through */ - stage->next->line( stage->next, header ); -} - - -static void twoside_point( struct draw_stage *stage, - struct prim_header *header ) -{ - /* pass-through */ - stage->next->point( stage->next, header ); -} - - -static void twoside_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; - uint i; - - twoside->attrib_front0 = 0; - twoside->attrib_front1 = 0; - twoside->attrib_back0 = 0; - twoside->attrib_back1 = 0; - - /* Find which vertex shader outputs are front/back colors */ - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { - if (vs->output_semantic_index[i] == 0) - twoside->attrib_front0 = i; - else - twoside->attrib_front1 = i; - } - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - if (vs->output_semantic_index[i] == 0) - twoside->attrib_back0 = i; - else - twoside->attrib_back1 = i; - } - } - - if (!twoside->attrib_back0) - twoside->attrib_front0 = 0; - - if (!twoside->attrib_back1) - twoside->attrib_front1 = 0; - - /* - * We'll multiply the primitive's determinant by this sign to determine - * if the triangle is back-facing (negative). - * sign = -1 for CCW, +1 for CW - */ - twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; - - stage->tri = twoside_tri; - stage->tri( stage, header ); -} - - -static void twoside_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->tri = twoside_first_tri; - stage->next->flush( stage->next, flags ); -} - - -static void twoside_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void twoside_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create twoside pipeline stage. - */ -struct draw_stage *draw_twoside_stage( struct draw_context *draw ) -{ - struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); - - draw_alloc_temp_verts( &twoside->stage, 3 ); - - twoside->stage.draw = draw; - twoside->stage.next = NULL; - twoside->stage.point = twoside_point; - twoside->stage.line = twoside_line; - twoside->stage.tri = twoside_first_tri; - twoside->stage.flush = twoside_flush; - twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; - twoside->stage.destroy = twoside_destroy; - - return &twoside->stage; -} diff --git a/src/gallium/aux/draw/draw_unfilled.c b/src/gallium/aux/draw/draw_unfilled.c deleted file mode 100644 index 8777cfdfc8..0000000000 --- a/src/gallium/aux/draw/draw_unfilled.c +++ /dev/null @@ -1,206 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Drawing stage for handling glPolygonMode(line/point). - * Convert triangles to points or lines as needed. - */ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "draw_private.h" - - -struct unfilled_stage { - struct draw_stage stage; - - /** [0] = front face, [1] = back face. - * legal values: PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE, - * and PIPE_POLYGON_MODE_POINT, - */ - unsigned mode[2]; -}; - - -static INLINE struct unfilled_stage *unfilled_stage( struct draw_stage *stage ) -{ - return (struct unfilled_stage *)stage; -} - - - -static void point( struct draw_stage *stage, - struct vertex_header *v0 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - stage->next->point( stage->next, &tmp ); -} - -static void line( struct draw_stage *stage, - struct vertex_header *v0, - struct vertex_header *v1 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - tmp.v[1] = v1; - stage->next->line( stage->next, &tmp ); -} - - -static void points( struct draw_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - - if (header->edgeflags & 0x1) point( stage, v0 ); - if (header->edgeflags & 0x2) point( stage, v1 ); - if (header->edgeflags & 0x4) point( stage, v2 ); -} - - -static void lines( struct draw_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - -#if 0 - assert(((header->edgeflags & 0x1) >> 0) == header->v[0]->edgeflag); - assert(((header->edgeflags & 0x2) >> 1) == header->v[1]->edgeflag); - assert(((header->edgeflags & 0x4) >> 2) == header->v[2]->edgeflag); -#endif - - if (header->edgeflags & 0x1) line( stage, v0, v1 ); - if (header->edgeflags & 0x2) line( stage, v1, v2 ); - if (header->edgeflags & 0x4) line( stage, v2, v0 ); -} - - -/* Unfilled tri: - * - * Note edgeflags in the vertex struct is not sufficient as we will - * need to manipulate them when decomposing primitives??? - */ -static void unfilled_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - unsigned mode = unfilled->mode[header->det >= 0.0]; - - switch (mode) { - case PIPE_POLYGON_MODE_FILL: - stage->next->tri( stage->next, header ); - break; - case PIPE_POLYGON_MODE_LINE: - lines( stage, header ); - break; - case PIPE_POLYGON_MODE_POINT: - points( stage, header ); - break; - default: - abort(); - } -} - - -static void unfilled_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - - unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */ - unfilled->mode[1] = stage->draw->rasterizer->fill_cw; /* back */ - - stage->tri = unfilled_tri; - stage->tri( stage, header ); -} - - -static void unfilled_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void unfilled_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void unfilled_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->next->flush( stage->next, flags ); - - stage->tri = unfilled_first_tri; -} - - -static void unfilled_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void unfilled_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create unfilled triangle stage. - */ -struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) -{ - struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); - - draw_alloc_temp_verts( &unfilled->stage, 0 ); - - unfilled->stage.draw = draw; - unfilled->stage.next = NULL; - unfilled->stage.tmp = NULL; - unfilled->stage.point = unfilled_point; - unfilled->stage.line = unfilled_line; - unfilled->stage.tri = unfilled_first_tri; - unfilled->stage.flush = unfilled_flush; - unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; - unfilled->stage.destroy = unfilled_destroy; - - return &unfilled->stage; -} diff --git a/src/gallium/aux/draw/draw_validate.c b/src/gallium/aux/draw/draw_validate.c deleted file mode 100644 index 4375ebabbc..0000000000 --- a/src/gallium/aux/draw/draw_validate.c +++ /dev/null @@ -1,185 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "draw_private.h" - - - - - -/** - * Rebuild the rendering pipeline. - */ -static struct draw_stage *validate_pipeline( struct draw_stage *stage ) -{ - struct draw_context *draw = stage->draw; - struct draw_stage *next = draw->pipeline.rasterize; - int need_det = 0; - int precalc_flat = 0; - - /* Set the validate's next stage to the rasterize stage, so that it - * can be found later if needed for flushing. - */ - stage->next = next; - - /* - * NOTE: we build up the pipeline in end-to-start order. - * - * TODO: make the current primitive part of the state and build - * shorter pipelines for lines & points. - */ - - if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines) || - (draw->rasterizer->point_size != 1.0 && draw->convert_wide_points) || - draw->rasterizer->point_sprite) { - draw->pipeline.wide->next = next; - next = draw->pipeline.wide; - } - - if (draw->rasterizer->line_stipple_enable) { - draw->pipeline.stipple->next = next; - next = draw->pipeline.stipple; - precalc_flat = 1; /* only needed for lines really */ - } - - if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { - draw->pipeline.unfilled->next = next; - next = draw->pipeline.unfilled; - precalc_flat = 1; /* only needed for triangles really */ - need_det = 1; - } - - if (draw->rasterizer->flatshade && precalc_flat) { - draw->pipeline.flatshade->next = next; - next = draw->pipeline.flatshade; - } - - if (draw->rasterizer->offset_cw || - draw->rasterizer->offset_ccw) { - draw->pipeline.offset->next = next; - next = draw->pipeline.offset; - need_det = 1; - } - - if (draw->rasterizer->light_twoside) { - draw->pipeline.twoside->next = next; - next = draw->pipeline.twoside; - need_det = 1; - } - - /* Always run the cull stage as we calculate determinant there - * also. - * - * This can actually be a win as culling out the triangles can lead - * to less work emitting vertices, smaller vertex buffers, etc. - * It's difficult to say whether this will be true in general. - */ - if (need_det || draw->rasterizer->cull_mode) { - draw->pipeline.cull->next = next; - next = draw->pipeline.cull; - } - - /* Clip stage - */ - if (!draw->rasterizer->bypass_clipping) - { - draw->pipeline.clip->next = next; - next = draw->pipeline.clip; - } - - - draw->pipeline.first = next; - return next; -} - -static void validate_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->tri( pipeline, header ); -} - -static void validate_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->line( pipeline, header ); -} - -static void validate_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->point( pipeline, header ); -} - -static void validate_reset_stipple_counter( struct draw_stage *stage ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->reset_stipple_counter( pipeline ); -} - -static void validate_flush( struct draw_stage *stage, - unsigned flags ) -{ - /* May need to pass a backend flush on to the rasterize stage. - */ - if (stage->next) - stage->next->flush( stage->next, flags ); -} - - -static void validate_destroy( struct draw_stage *stage ) -{ - FREE( stage ); -} - - -/** - * Create validate pipeline stage. - */ -struct draw_stage *draw_validate_stage( struct draw_context *draw ) -{ - struct draw_stage *stage = CALLOC_STRUCT(draw_stage); - - stage->draw = draw; - stage->next = NULL; - stage->point = validate_point; - stage->line = validate_line; - stage->tri = validate_tri; - stage->flush = validate_flush; - stage->reset_stipple_counter = validate_reset_stipple_counter; - stage->destroy = validate_destroy; - - return stage; -} diff --git a/src/gallium/aux/draw/draw_vbuf.c b/src/gallium/aux/draw/draw_vbuf.c deleted file mode 100644 index 71ac73912b..0000000000 --- a/src/gallium/aux/draw/draw_vbuf.c +++ /dev/null @@ -1,570 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Vertex buffer drawing stage. - * - * \author José Fonseca - * \author Keith Whitwell - */ - - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" - -#include "draw_vbuf.h" -#include "draw_private.h" -#include "draw_vertex.h" -#include "draw_vf.h" - - -/** - * Vertex buffer emit stage. - */ -struct vbuf_stage { - struct draw_stage stage; /**< This must be first (base class) */ - - struct vbuf_render *render; - - const struct vertex_info *vinfo; - - /** Vertex size in bytes */ - unsigned vertex_size; - - struct draw_vertex_fetch *vf; - - /* FIXME: we have no guarantee that 'unsigned' is 32bit */ - - /** Vertices in hardware format */ - unsigned *vertices; - unsigned *vertex_ptr; - unsigned max_vertices; - unsigned nr_vertices; - - /** Indices */ - ushort *indices; - unsigned max_indices; - unsigned nr_indices; - - /** Pipe primitive */ - unsigned prim; -}; - - -/** - * Basically a cast wrapper. - */ -static INLINE struct vbuf_stage * -vbuf_stage( struct draw_stage *stage ) -{ - assert(stage); - return (struct vbuf_stage *)stage; -} - - -static void vbuf_flush_indices( struct vbuf_stage *vbuf ); -static void vbuf_flush_vertices( struct vbuf_stage *vbuf ); -static void vbuf_alloc_vertices( struct vbuf_stage *vbuf ); - - -static INLINE boolean -overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) -{ - unsigned long used = (unsigned long) ((char *)ptr - (char *)map); - return (used + bytes) > bufsz; -} - - -static INLINE void -check_space( struct vbuf_stage *vbuf, unsigned nr ) -{ - if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { - vbuf_flush_vertices(vbuf); - vbuf_alloc_vertices(vbuf); - } - - if (vbuf->nr_indices + nr > vbuf->max_indices ) - vbuf_flush_indices(vbuf); -} - - -#if 0 -static INLINE void -dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) -{ - assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); - unsigned i, j, k; - - for (i = 0; i < vinfo->num_attribs; i++) { - j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - debug_printf("EMIT_OMIT:"); - break; - case EMIT_ALL: - assert(i == 0); - assert(j == 0); - debug_printf("EMIT_ALL:\t"); - for(k = 0; k < vinfo->size*4; ++k) - debug_printf("%02x ", *data++); - break; - case EMIT_1F: - debug_printf("EMIT_1F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_1F_PSIZE: - debug_printf("EMIT_1F_PSIZE:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_2F: - debug_printf("EMIT_2F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_3F: - debug_printf("EMIT_3F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - data += sizeof(float); - break; - case EMIT_4F: - debug_printf("EMIT_4F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_4UB: - debug_printf("EMIT_4UB:\t"); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - break; - default: - assert(0); - } - debug_printf("\n"); - } - debug_printf("\n"); -} -#endif - - -/** - * Extract the needed fields from post-transformed vertex and emit - * a hardware(driver) vertex. - * Recall that the vertices are constructed by the 'draw' module and - * have a couple of slots at the beginning (1-dword header, 4-dword - * clip pos) that we ignore here. We only use the vertex->data[] fields. - */ -static INLINE void -emit_vertex( struct vbuf_stage *vbuf, - struct vertex_header *vertex ) -{ -#if 0 - debug_printf("emit vertex %d to %p\n", - vbuf->nr_vertices, vbuf->vertex_ptr); -#endif - - if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { - if(vertex->vertex_id < vbuf->nr_vertices) - return; - else - debug_printf("Bad vertex id 0x%04x (>= 0x%04x)\n", - vertex->vertex_id, vbuf->nr_vertices); - return; - } - - vertex->vertex_id = vbuf->nr_vertices++; - - if(!vbuf->vf) { - const struct vertex_info *vinfo = vbuf->vinfo; - uint i; - uint count = 0; /* for debug/sanity */ - - assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); - - for (i = 0; i < vinfo->num_attribs; i++) { - uint j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; - case EMIT_ALL: - /* just copy the whole vertex as-is to the vbuf */ - assert(i == 0); - assert(j == 0); - memcpy(vbuf->vertex_ptr, vertex, vinfo->size * 4); - vbuf->vertex_ptr += vinfo->size; - count += vinfo->size; - break; - case EMIT_1F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - count++; - break; - case EMIT_1F_PSIZE: - *vbuf->vertex_ptr++ = fui(vbuf->stage.draw->rasterizer->point_size); - count++; - break; - case EMIT_2F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - count += 2; - break; - case EMIT_3F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); - count += 3; - break; - case EMIT_4F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][3]); - count += 4; - break; - case EMIT_4UB: - *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), - float_to_ubyte( vertex->data[j][1] ), - float_to_ubyte( vertex->data[j][0] ), - float_to_ubyte( vertex->data[j][3] )); - count += 1; - break; - default: - assert(0); - } - } - assert(count == vinfo->size); -#if 0 - { - static float data[256]; - draw_vf_emit_vertex(vbuf->vf, vertex, data); - if(memcmp((uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size, data, vbuf->vertex_size)) { - debug_printf("With VF:\n"); - dump_emitted_vertex(vbuf->vinfo, (uint8_t *)data); - debug_printf("Without VF:\n"); - dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size); - assert(0); - } - } -#endif - } - else { - draw_vf_emit_vertex(vbuf->vf, vertex, vbuf->vertex_ptr); - - vbuf->vertex_ptr += vbuf->vertex_size/4; - } -} - - -static void -vbuf_tri( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - check_space( vbuf, 3 ); - - for (i = 0; i < 3; i++) { - emit_vertex( vbuf, prim->v[i] ); - - vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; - } -} - - -static void -vbuf_line( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - check_space( vbuf, 2 ); - - for (i = 0; i < 2; i++) { - emit_vertex( vbuf, prim->v[i] ); - - vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; - } -} - - -static void -vbuf_point( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - check_space( vbuf, 1 ); - - emit_vertex( vbuf, prim->v[0] ); - - vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[0]->vertex_id; -} - - -/** - * Set the prim type for subsequent vertices. - * This may result in a new vertex size. The existing vbuffer (if any) - * will be flushed if needed and a new one allocated. - */ -static void -vbuf_set_prim( struct vbuf_stage *vbuf, uint newprim ) -{ - const struct vertex_info *vinfo; - unsigned vertex_size; - - assert(newprim == PIPE_PRIM_POINTS || - newprim == PIPE_PRIM_LINES || - newprim == PIPE_PRIM_TRIANGLES); - - vbuf->prim = newprim; - vbuf->render->set_primitive(vbuf->render, newprim); - - vinfo = vbuf->render->get_vertex_info(vbuf->render); - vertex_size = vinfo->size * sizeof(float); - - if (vertex_size != vbuf->vertex_size) - vbuf_flush_vertices(vbuf); - - vbuf->vinfo = vinfo; - vbuf->vertex_size = vertex_size; - if(vbuf->vf) - draw_vf_set_vertex_info(vbuf->vf, - vbuf->vinfo, - vbuf->stage.draw->rasterizer->point_size); - - if (!vbuf->vertices) - vbuf_alloc_vertices(vbuf); -} - - -static void -vbuf_first_tri( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - stage->tri = vbuf_tri; - vbuf_set_prim(vbuf, PIPE_PRIM_TRIANGLES); - stage->tri( stage, prim ); -} - - -static void -vbuf_first_line( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - stage->line = vbuf_line; - vbuf_set_prim(vbuf, PIPE_PRIM_LINES); - stage->line( stage, prim ); -} - - -static void -vbuf_first_point( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - stage->point = vbuf_point; - vbuf_set_prim(vbuf, PIPE_PRIM_POINTS); - stage->point( stage, prim ); -} - - -static void -vbuf_flush_indices( struct vbuf_stage *vbuf ) -{ - if(!vbuf->nr_indices) - return; - - assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == - vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); - - switch(vbuf->prim) { - case PIPE_PRIM_POINTS: - break; - case PIPE_PRIM_LINES: - assert(vbuf->nr_indices % 2 == 0); - break; - case PIPE_PRIM_TRIANGLES: - assert(vbuf->nr_indices % 3 == 0); - break; - default: - assert(0); - } - - vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); - - vbuf->nr_indices = 0; - - /* don't need to reset point/line/tri functions */ -#if 0 - stage->point = vbuf_first_point; - stage->line = vbuf_first_line; - stage->tri = vbuf_first_tri; -#endif -} - - -/** - * Flush existing vertex buffer and allocate a new one. - * - * XXX: We separate flush-on-index-full and flush-on-vb-full, but may - * raise issues uploading vertices if the hardware wants to flush when - * we flush. - */ -static void -vbuf_flush_vertices( struct vbuf_stage *vbuf ) -{ - if(vbuf->vertices) { - vbuf_flush_indices(vbuf); - - /* Reset temporary vertices ids */ - if(vbuf->nr_vertices) - draw_reset_vertex_ids( vbuf->stage.draw ); - - /* Free the vertex buffer */ - vbuf->render->release_vertices(vbuf->render, - vbuf->vertices, - vbuf->vertex_size, - vbuf->nr_vertices); - vbuf->max_vertices = vbuf->nr_vertices = 0; - vbuf->vertex_ptr = vbuf->vertices = NULL; - - } -} - - -static void -vbuf_alloc_vertices( struct vbuf_stage *vbuf ) -{ - assert(!vbuf->nr_indices); - assert(!vbuf->vertices); - - /* Allocate a new vertex buffer */ - vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; - vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, - (ushort) vbuf->vertex_size, - (ushort) vbuf->max_vertices); - vbuf->vertex_ptr = vbuf->vertices; -} - - - -static void -vbuf_flush( struct draw_stage *stage, unsigned flags ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - - stage->point = vbuf_first_point; - stage->line = vbuf_first_line; - stage->tri = vbuf_first_tri; - - if (flags & DRAW_FLUSH_BACKEND) - vbuf_flush_vertices( vbuf ); -} - - -static void -vbuf_reset_stipple_counter( struct draw_stage *stage ) -{ - /* XXX: Need to do something here for hardware with linestipple. - */ - (void) stage; -} - - -static void vbuf_destroy( struct draw_stage *stage ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - if(vbuf->indices) - align_free( vbuf->indices ); - - if(vbuf->vf) - draw_vf_destroy( vbuf->vf ); - - if (vbuf->render) - vbuf->render->destroy( vbuf->render ); - - FREE( stage ); -} - - -/** - * Create a new primitive vbuf/render stage. - */ -struct draw_stage *draw_vbuf_stage( struct draw_context *draw, - struct vbuf_render *render ) -{ - struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); - - if(!vbuf) - return NULL; - - vbuf->stage.draw = draw; - vbuf->stage.point = vbuf_first_point; - vbuf->stage.line = vbuf_first_line; - vbuf->stage.tri = vbuf_first_tri; - vbuf->stage.flush = vbuf_flush; - vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter; - vbuf->stage.destroy = vbuf_destroy; - - vbuf->render = render; - - assert(render->max_indices < UNDEFINED_VERTEX_ID); - vbuf->max_indices = render->max_indices; - vbuf->indices = (ushort *) - align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); - if(!vbuf->indices) - vbuf_destroy(&vbuf->stage); - - vbuf->vertices = NULL; - vbuf->vertex_ptr = vbuf->vertices; - - vbuf->prim = ~0; - - if(!GETENV("GALLIUM_NOVF")) - vbuf->vf = draw_vf_create(); - - return &vbuf->stage; -} diff --git a/src/gallium/aux/draw/draw_vbuf.h b/src/gallium/aux/draw/draw_vbuf.h deleted file mode 100644 index cfd2b9820c..0000000000 --- a/src/gallium/aux/draw/draw_vbuf.h +++ /dev/null @@ -1,106 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Vertex buffer drawing stage. - * - * \author Keith Whitwell - * \author José Fonseca - */ - -#ifndef DRAW_VBUF_H_ -#define DRAW_VBUF_H_ - - -#include "pipe/p_util.h" - - -struct draw_context; -struct vertex_info; - - -/** - * Interface for hardware vertex buffer rendering. - */ -struct vbuf_render { - - /** - * Driver limits. May be tuned lower to improve cache hits on - * index list. - */ - unsigned max_indices; - unsigned max_vertex_buffer_bytes; - - /** - * Get the hardware vertex format. - * - * XXX: have this in draw_context instead? - */ - const struct vertex_info *(*get_vertex_info)( struct vbuf_render * ); - - /** - * Request a destination for vertices. - * Hardware renderers will use ttm memory, others will just malloc - * something. - */ - void *(*allocate_vertices)( struct vbuf_render *, - ushort vertex_size, - ushort nr_vertices ); - - /** - * Notify the renderer of the current primitive when it changes. - * Prim is restricted to TRIANGLES, LINES and POINTS. - */ - void (*set_primitive)( struct vbuf_render *, unsigned prim ); - - /** - * DrawElements, note indices are ushort: - */ - void (*draw)( struct vbuf_render *, - const ushort *indices, - uint nr_indices ); - - /** - * Called when vbuf is done with this set of vertices: - */ - void (*release_vertices)( struct vbuf_render *, - void *vertices, - unsigned vertex_size, - unsigned vertices_used ); - - void (*destroy)( struct vbuf_render * ); -}; - - - -struct draw_stage * -draw_vbuf_stage( struct draw_context *draw, - struct vbuf_render *render ); - - -#endif /*DRAW_VBUF_H_*/ diff --git a/src/gallium/aux/draw/draw_vertex.c b/src/gallium/aux/draw/draw_vertex.c deleted file mode 100644 index daf1ef4b80..0000000000 --- a/src/gallium/aux/draw/draw_vertex.c +++ /dev/null @@ -1,79 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* - * Functions for specifying the post-transformation vertex layout. - * - * Author: - * Brian Paul - * Keith Whitwell - */ - - -#include "draw/draw_private.h" -#include "draw/draw_vertex.h" - - -/** - * Compute the size of a vertex, in dwords/floats, to update the - * vinfo->size field. - */ -void -draw_compute_vertex_size(struct vertex_info *vinfo) -{ - uint i; - - vinfo->size = 0; - for (i = 0; i < vinfo->num_attribs; i++) { - switch (vinfo->emit[i]) { - case EMIT_OMIT: - break; - case EMIT_4UB: - /* fall-through */ - case EMIT_1F_PSIZE: - /* fall-through */ - case EMIT_1F: - vinfo->size += 1; - break; - case EMIT_2F: - vinfo->size += 2; - break; - case EMIT_3F: - vinfo->size += 3; - break; - case EMIT_4F: - vinfo->size += 4; - break; - case EMIT_ALL: - /* fall-through */ - default: - assert(0); - } - } - - assert(vinfo->size * 4 <= MAX_VERTEX_SIZE); -} diff --git a/src/gallium/aux/draw/draw_vertex.h b/src/gallium/aux/draw/draw_vertex.h deleted file mode 100644 index 267c74203b..0000000000 --- a/src/gallium/aux/draw/draw_vertex.h +++ /dev/null @@ -1,111 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Post-transform vertex format info. The vertex_info struct is used by - * the draw_vbuf code to emit hardware-specific vertex layouts into hw - * vertex buffers. - * - * Author: - * Brian Paul - */ - - -#ifndef DRAW_VERTEX_H -#define DRAW_VERTEX_H - - -#include "pipe/p_state.h" - - -/** - * Vertex attribute emit modes - */ -enum attrib_emit { - EMIT_OMIT, /**< don't emit the attribute */ - EMIT_ALL, /**< emit whole post-xform vertex, w/ header */ - EMIT_1F, - EMIT_1F_PSIZE, /**< insert constant point size */ - EMIT_2F, - EMIT_3F, - EMIT_4F, - EMIT_4UB /**< XXX may need variations for RGBA vs BGRA, etc */ -}; - - -/** - * Attribute interpolation mode - */ -enum interp_mode { - INTERP_NONE, /**< never interpolate vertex header info */ - INTERP_POS, /**< special case for frag position */ - INTERP_CONSTANT, - INTERP_LINEAR, - INTERP_PERSPECTIVE -}; - - -/** - * Information about hardware/rasterization vertex layout. - */ -struct vertex_info -{ - uint num_attribs; - uint hwfmt[4]; /**< hardware format info for this format */ - enum interp_mode interp_mode[PIPE_MAX_SHADER_INPUTS]; - enum attrib_emit emit[PIPE_MAX_SHADER_INPUTS]; /**< EMIT_x */ - uint src_index[PIPE_MAX_SHADER_INPUTS]; /**< map to post-xform attribs */ - uint size; /**< total vertex size in dwords */ -}; - - - -/** - * Add another attribute to the given vertex_info object. - * \param src_index indicates which post-transformed vertex attrib slot - * corresponds to this attribute. - * \return slot in which the attribute was added - */ -static INLINE uint -draw_emit_vertex_attr(struct vertex_info *vinfo, - enum attrib_emit emit, enum interp_mode interp, - uint src_index) -{ - const uint n = vinfo->num_attribs; - assert(n < PIPE_MAX_SHADER_INPUTS); - vinfo->emit[n] = emit; - vinfo->interp_mode[n] = interp; - vinfo->src_index[n] = src_index; - vinfo->num_attribs++; - return n; -} - - -extern void draw_compute_vertex_size(struct vertex_info *vinfo); - - -#endif /* DRAW_VERTEX_H */ diff --git a/src/gallium/aux/draw/draw_vertex_cache.c b/src/gallium/aux/draw/draw_vertex_cache.c deleted file mode 100644 index 44427999cc..0000000000 --- a/src/gallium/aux/draw/draw_vertex_cache.c +++ /dev/null @@ -1,196 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "draw_private.h" -#include "draw_context.h" - - -void draw_vertex_cache_invalidate( struct draw_context *draw ) -{ - assert(draw->pq.queue_nr == 0); - assert(draw->vs.queue_nr == 0); - assert(draw->vcache.referenced == 0); - - memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); -} - - -/** - * Check if vertex is in cache, otherwise add it. It won't go through - * VS yet, not until there is a flush operation or the VS queue fills up. - * - * Note that cache entries are basically just two pointers: the first - * an index into the user's vertex arrays, the second a location in - * the vertex shader cache for the post-transformed vertex. - * - * \return pointer to location of (post-transformed) vertex header in the cache - */ -static struct vertex_header *get_vertex( struct draw_context *draw, - unsigned i ) -{ - unsigned slot = (i + (i>>5)) % VCACHE_SIZE; - - assert(slot < 32); /* so we don't exceed the bitfield size below */ - - /* Cache miss? - */ - if (draw->vcache.idx[slot] != i) { - - /* If slot is in use, use the overflow area: - */ - if (draw->vcache.referenced & (1 << slot)) { - slot = VCACHE_SIZE + draw->vcache.overflow++; - } - - assert(slot < Elements(draw->vcache.idx)); - - draw->vcache.idx[slot] = i; - - /* Add to vertex shader queue: - */ - assert(draw->vs.queue_nr < VS_QUEUE_LENGTH); - draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; - draw->vs.queue[draw->vs.queue_nr].elt = i; - draw->vs.queue_nr++; - - /* Need to set the vertex's edge flag here. If we're being called - * by do_ef_triangle(), that function needs edge flag info! - */ - draw->vcache.vertex[slot]->clipmask = 0; - draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ - draw->vcache.vertex[slot]->pad = 0; - draw->vcache.vertex[slot]->vertex_id = UNDEFINED_VERTEX_ID; - } - - - /* primitive flushing may have cleared the bitfield but did not - * clear the idx[] array values. Set the bit now. This fixes a - * bug found when drawing long triangle fans. - */ - draw->vcache.referenced |= (1 << slot); - return draw->vcache.vertex[slot]; -} - - -static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const unsigned *elts = (const unsigned *) draw->user.elts; - return get_vertex( draw, elts[i] ); -} - - -static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const ushort *elts = (const ushort *) draw->user.elts; - return get_vertex( draw, elts[i] ); -} - - -static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const ubyte *elts = (const ubyte *) draw->user.elts; - return get_vertex( draw, elts[i] ); -} - - -void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) -{ - unsigned i; - - for (i = 0; i < Elements(draw->vcache.vertex); i++) - draw->vcache.vertex[i]->vertex_id = UNDEFINED_VERTEX_ID; -} - - -void draw_vertex_cache_unreference( struct draw_context *draw ) -{ - draw->vcache.referenced = 0; - draw->vcache.overflow = 0; -} - - -int draw_vertex_cache_check_space( struct draw_context *draw, - unsigned nr_verts ) -{ - if (draw->vcache.overflow + nr_verts < VCACHE_OVERFLOW) { - /* The vs queue is sized so that this can never happen: - */ - assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); - return TRUE; - } - else - return FALSE; -} - - - -/** - * Tell the drawing context about the index/element buffer to use - * (ala glDrawElements) - * If no element buffer is to be used (i.e. glDrawArrays) then this - * should be called with eltSize=0 and elements=NULL. - * - * \param draw the drawing context - * \param eltSize size of each element (1, 2 or 4 bytes) - * \param elements the element buffer ptr - */ -void -draw_set_mapped_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ) -{ -// draw_statechange( draw ); - - /* choose the get_vertex() function to use */ - switch (eltSize) { - case 0: - draw->vcache.get_vertex = get_vertex; - break; - case 1: - draw->vcache.get_vertex = get_ubyte_elt_vertex; - break; - case 2: - draw->vcache.get_vertex = get_ushort_elt_vertex; - break; - case 4: - draw->vcache.get_vertex = get_uint_elt_vertex; - break; - default: - assert(0); - } - draw->user.elts = elements; - draw->user.eltSize = eltSize; -} - diff --git a/src/gallium/aux/draw/draw_vertex_fetch.c b/src/gallium/aux/draw/draw_vertex_fetch.c deleted file mode 100644 index e13df04605..0000000000 --- a/src/gallium/aux/draw/draw_vertex_fetch.c +++ /dev/null @@ -1,510 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" -#include "draw_context.h" - - -#define DRAW_DBG 0 - - -/** - * Fetch a float[4] vertex attribute from memory, doing format/type - * conversion as needed. - * - * This is probably needed/dupliocated elsewhere, eg format - * conversion, texture sampling etc. - */ -#define FETCH_ATTRIB( NAME, SZ, CVT ) \ -static void \ -fetch_##NAME(const void *ptr, float *attrib) \ -{ \ - static const float defaults[4] = { 0,0,0,1 }; \ - int i; \ - \ - for (i = 0; i < SZ; i++) { \ - attrib[i] = CVT; \ - } \ - \ - for (; i < 4; i++) { \ - attrib[i] = defaults[i]; \ - } \ -} - -#define CVT_64_FLOAT (float) ((double *) ptr)[i] -#define CVT_32_FLOAT ((float *) ptr)[i] - -#define CVT_8_USCALED (float) ((unsigned char *) ptr)[i] -#define CVT_16_USCALED (float) ((unsigned short *) ptr)[i] -#define CVT_32_USCALED (float) ((unsigned int *) ptr)[i] - -#define CVT_8_SSCALED (float) ((char *) ptr)[i] -#define CVT_16_SSCALED (float) ((short *) ptr)[i] -#define CVT_32_SSCALED (float) ((int *) ptr)[i] - -#define CVT_8_UNORM (float) ((unsigned char *) ptr)[i] / 255.0f -#define CVT_16_UNORM (float) ((unsigned short *) ptr)[i] / 65535.0f -#define CVT_32_UNORM (float) ((unsigned int *) ptr)[i] / 4294967295.0f - -#define CVT_8_SNORM (float) ((char *) ptr)[i] / 127.0f -#define CVT_16_SNORM (float) ((short *) ptr)[i] / 32767.0f -#define CVT_32_SNORM (float) ((int *) ptr)[i] / 2147483647.0f - -FETCH_ATTRIB( R64G64B64A64_FLOAT, 4, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) -FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) - -FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) -FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) - -FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) -FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) - -FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) -FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) - -FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) -FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) - -FETCH_ATTRIB( R32G32B32A32_SNORM, 4, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32B32_SNORM, 3, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32_SNORM, 2, CVT_32_SNORM ) -FETCH_ATTRIB( R32_SNORM, 1, CVT_32_SNORM ) - -FETCH_ATTRIB( R16G16B16A16_USCALED, 4, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16B16_USCALED, 3, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16_USCALED, 2, CVT_16_USCALED ) -FETCH_ATTRIB( R16_USCALED, 1, CVT_16_USCALED ) - -FETCH_ATTRIB( R16G16B16A16_SSCALED, 4, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16B16_SSCALED, 3, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16_SSCALED, 2, CVT_16_SSCALED ) -FETCH_ATTRIB( R16_SSCALED, 1, CVT_16_SSCALED ) - -FETCH_ATTRIB( R16G16B16A16_UNORM, 4, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16B16_UNORM, 3, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16_UNORM, 2, CVT_16_UNORM ) -FETCH_ATTRIB( R16_UNORM, 1, CVT_16_UNORM ) - -FETCH_ATTRIB( R16G16B16A16_SNORM, 4, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16B16_SNORM, 3, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16_SNORM, 2, CVT_16_SNORM ) -FETCH_ATTRIB( R16_SNORM, 1, CVT_16_SNORM ) - -FETCH_ATTRIB( R8G8B8A8_USCALED, 4, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8B8_USCALED, 3, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8_USCALED, 2, CVT_8_USCALED ) -FETCH_ATTRIB( R8_USCALED, 1, CVT_8_USCALED ) - -FETCH_ATTRIB( R8G8B8A8_SSCALED, 4, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8B8_SSCALED, 3, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8_SSCALED, 2, CVT_8_SSCALED ) -FETCH_ATTRIB( R8_SSCALED, 1, CVT_8_SSCALED ) - -FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8B8_UNORM, 3, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8_UNORM, 2, CVT_8_UNORM ) -FETCH_ATTRIB( R8_UNORM, 1, CVT_8_UNORM ) - -FETCH_ATTRIB( R8G8B8A8_SNORM, 4, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8B8_SNORM, 3, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8_SNORM, 2, CVT_8_SNORM ) -FETCH_ATTRIB( R8_SNORM, 1, CVT_8_SNORM ) - -FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) -//FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) - - - -static fetch_func get_fetch_func( enum pipe_format format ) -{ -#if 0 - { - char tmp[80]; - pf_sprint_name(tmp, format); - debug_printf("%s: %s\n", __FUNCTION__, tmp); - } -#endif - - switch (format) { - case PIPE_FORMAT_R64_FLOAT: - return fetch_R64_FLOAT; - case PIPE_FORMAT_R64G64_FLOAT: - return fetch_R64G64_FLOAT; - case PIPE_FORMAT_R64G64B64_FLOAT: - return fetch_R64G64B64_FLOAT; - case PIPE_FORMAT_R64G64B64A64_FLOAT: - return fetch_R64G64B64A64_FLOAT; - - case PIPE_FORMAT_R32_FLOAT: - return fetch_R32_FLOAT; - case PIPE_FORMAT_R32G32_FLOAT: - return fetch_R32G32_FLOAT; - case PIPE_FORMAT_R32G32B32_FLOAT: - return fetch_R32G32B32_FLOAT; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return fetch_R32G32B32A32_FLOAT; - - case PIPE_FORMAT_R32_UNORM: - return fetch_R32_UNORM; - case PIPE_FORMAT_R32G32_UNORM: - return fetch_R32G32_UNORM; - case PIPE_FORMAT_R32G32B32_UNORM: - return fetch_R32G32B32_UNORM; - case PIPE_FORMAT_R32G32B32A32_UNORM: - return fetch_R32G32B32A32_UNORM; - - case PIPE_FORMAT_R32_USCALED: - return fetch_R32_USCALED; - case PIPE_FORMAT_R32G32_USCALED: - return fetch_R32G32_USCALED; - case PIPE_FORMAT_R32G32B32_USCALED: - return fetch_R32G32B32_USCALED; - case PIPE_FORMAT_R32G32B32A32_USCALED: - return fetch_R32G32B32A32_USCALED; - - case PIPE_FORMAT_R32_SNORM: - return fetch_R32_SNORM; - case PIPE_FORMAT_R32G32_SNORM: - return fetch_R32G32_SNORM; - case PIPE_FORMAT_R32G32B32_SNORM: - return fetch_R32G32B32_SNORM; - case PIPE_FORMAT_R32G32B32A32_SNORM: - return fetch_R32G32B32A32_SNORM; - - case PIPE_FORMAT_R32_SSCALED: - return fetch_R32_SSCALED; - case PIPE_FORMAT_R32G32_SSCALED: - return fetch_R32G32_SSCALED; - case PIPE_FORMAT_R32G32B32_SSCALED: - return fetch_R32G32B32_SSCALED; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - return fetch_R32G32B32A32_SSCALED; - - case PIPE_FORMAT_R16_UNORM: - return fetch_R16_UNORM; - case PIPE_FORMAT_R16G16_UNORM: - return fetch_R16G16_UNORM; - case PIPE_FORMAT_R16G16B16_UNORM: - return fetch_R16G16B16_UNORM; - case PIPE_FORMAT_R16G16B16A16_UNORM: - return fetch_R16G16B16A16_UNORM; - - case PIPE_FORMAT_R16_USCALED: - return fetch_R16_USCALED; - case PIPE_FORMAT_R16G16_USCALED: - return fetch_R16G16_USCALED; - case PIPE_FORMAT_R16G16B16_USCALED: - return fetch_R16G16B16_USCALED; - case PIPE_FORMAT_R16G16B16A16_USCALED: - return fetch_R16G16B16A16_USCALED; - - case PIPE_FORMAT_R16_SNORM: - return fetch_R16_SNORM; - case PIPE_FORMAT_R16G16_SNORM: - return fetch_R16G16_SNORM; - case PIPE_FORMAT_R16G16B16_SNORM: - return fetch_R16G16B16_SNORM; - case PIPE_FORMAT_R16G16B16A16_SNORM: - return fetch_R16G16B16A16_SNORM; - - case PIPE_FORMAT_R16_SSCALED: - return fetch_R16_SSCALED; - case PIPE_FORMAT_R16G16_SSCALED: - return fetch_R16G16_SSCALED; - case PIPE_FORMAT_R16G16B16_SSCALED: - return fetch_R16G16B16_SSCALED; - case PIPE_FORMAT_R16G16B16A16_SSCALED: - return fetch_R16G16B16A16_SSCALED; - - case PIPE_FORMAT_R8_UNORM: - return fetch_R8_UNORM; - case PIPE_FORMAT_R8G8_UNORM: - return fetch_R8G8_UNORM; - case PIPE_FORMAT_R8G8B8_UNORM: - return fetch_R8G8B8_UNORM; - case PIPE_FORMAT_R8G8B8A8_UNORM: - return fetch_R8G8B8A8_UNORM; - - case PIPE_FORMAT_R8_USCALED: - return fetch_R8_USCALED; - case PIPE_FORMAT_R8G8_USCALED: - return fetch_R8G8_USCALED; - case PIPE_FORMAT_R8G8B8_USCALED: - return fetch_R8G8B8_USCALED; - case PIPE_FORMAT_R8G8B8A8_USCALED: - return fetch_R8G8B8A8_USCALED; - - case PIPE_FORMAT_R8_SNORM: - return fetch_R8_SNORM; - case PIPE_FORMAT_R8G8_SNORM: - return fetch_R8G8_SNORM; - case PIPE_FORMAT_R8G8B8_SNORM: - return fetch_R8G8B8_SNORM; - case PIPE_FORMAT_R8G8B8A8_SNORM: - return fetch_R8G8B8A8_SNORM; - - case PIPE_FORMAT_R8_SSCALED: - return fetch_R8_SSCALED; - case PIPE_FORMAT_R8G8_SSCALED: - return fetch_R8G8_SSCALED; - case PIPE_FORMAT_R8G8B8_SSCALED: - return fetch_R8G8B8_SSCALED; - case PIPE_FORMAT_R8G8B8A8_SSCALED: - return fetch_R8G8B8A8_SSCALED; - - case PIPE_FORMAT_A8R8G8B8_UNORM: - return fetch_A8R8G8B8_UNORM; - - case 0: - return NULL; /* not sure why this is needed */ - - default: - assert(0); - return NULL; - } -} - - -static void -transpose_4x4( float *out, const float *in ) -{ - /* This can be achieved in 12 sse instructions, plus the final - * stores I guess. This is probably a bit more than that - maybe - * 32 or so? - */ - out[0] = in[0]; out[1] = in[4]; out[2] = in[8]; out[3] = in[12]; - out[4] = in[1]; out[5] = in[5]; out[6] = in[9]; out[7] = in[13]; - out[8] = in[2]; out[9] = in[6]; out[10] = in[10]; out[11] = in[14]; - out[12] = in[3]; out[13] = in[7]; out[14] = in[11]; out[15] = in[15]; -} - - - -static void fetch_xyz_rgb( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - - assert(count <= 4); - -// debug_printf("%s\n", __FUNCTION__); - - /* loop over vertex attributes (vertex shader inputs) - */ - - for (i = 0; i < 4; i++) { - { - const float *in = (const float *)(src[0] + elts[i] * pitch[0]); - float *out = &machine->Inputs[0].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - - { - const float *in = (const float *)(src[1] + elts[i] * pitch[1]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - } -} - - - - -static void fetch_xyz_rgb_st( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - - assert(count <= 4); - - /* loop over vertex attributes (vertex shader inputs) - */ - - for (i = 0; i < 4; i++) { - { - const float *in = (const float *)(src[0] + elts[i] * pitch[0]); - float *out = &machine->Inputs[0].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - - { - const float *in = (const float *)(src[1] + elts[i] * pitch[1]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - - { - const float *in = (const float *)(src[2] + elts[i] * pitch[2]); - float *out = &machine->Inputs[2].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = 0.0f; - out[12] = 1.0f; - } - } -} - - - - -/** - * Fetch vertex attributes for 'count' vertices. - */ -static void generic_vertex_fetch( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - unsigned nr_attrs = draw->vertex_fetch.nr_attrs; - unsigned attr; - - assert(count <= 4); - -// debug_printf("%s %d\n", __FUNCTION__, count); - - /* loop over vertex attributes (vertex shader inputs) - */ - for (attr = 0; attr < nr_attrs; attr++) { - - const unsigned pitch = draw->vertex_fetch.pitch[attr]; - const ubyte *src = draw->vertex_fetch.src_ptr[attr]; - const fetch_func fetch = draw->vertex_fetch.fetch[attr]; - unsigned i; - float p[4][4]; - - - /* Fetch four attributes for four vertices. - * - * Could fetch directly into AOS format, but this is meant to be - * a prototype for an sse implementation, which would have - * difficulties doing that. - */ - for (i = 0; i < count; i++) - fetch( src + elts[i] * pitch, p[i] ); - - /* Be nice and zero out any missing vertices: - */ - for ( ; i < 4; i++) - p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; - - /* Transpose/swizzle into sse-friendly format. Currently - * assuming that all vertex shader inputs are float[4], but this - * isn't true -- if the vertex shader only wants tex0.xy, we - * could optimize for that. - * - * To do so fully without codegen would probably require an - * excessive number of fetch functions, but we could at least - * minimize the transpose step: - */ - transpose_4x4( (float *)&machine->Inputs[attr].xyzw[0].f[0], (float *)p ); - } -} - - - -void draw_update_vertex_fetch( struct draw_context *draw ) -{ - unsigned nr_attrs, i; - -// debug_printf("%s\n", __FUNCTION__); - - /* this may happend during context init */ - if (!draw->vertex_shader) - return; - - nr_attrs = draw->vertex_shader->state->num_inputs; - - for (i = 0; i < nr_attrs; i++) { - unsigned buf = draw->vertex_element[i].vertex_buffer_index; - enum pipe_format format = draw->vertex_element[i].src_format; - - draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + - draw->vertex_buffer[buf].buffer_offset + - draw->vertex_element[i].src_offset; - - draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; - draw->vertex_fetch.fetch[i] = get_fetch_func( format ); - } - - draw->vertex_fetch.nr_attrs = nr_attrs; - - draw->vertex_fetch.fetch_func = generic_vertex_fetch; - - switch (nr_attrs) { - case 2: - if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT) - draw->vertex_fetch.fetch_func = fetch_xyz_rgb; - break; - case 3: - if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[2].src_format == PIPE_FORMAT_R32G32_FLOAT) - draw->vertex_fetch.fetch_func = fetch_xyz_rgb_st; - break; - default: - break; - } - -} diff --git a/src/gallium/aux/draw/draw_vertex_shader.c b/src/gallium/aux/draw/draw_vertex_shader.c deleted file mode 100644 index 377ecbb931..0000000000 --- a/src/gallium/aux/draw/draw_vertex_shader.c +++ /dev/null @@ -1,325 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#if defined(__i386__) || defined(__386__) -#include "tgsi/exec/tgsi_sse2.h" -#endif -#include "draw_private.h" -#include "draw_context.h" - -#include "x86/rtasm/x86sse.h" -#include "llvm/gallivm.h" - - -#define DBG_VS 0 - - -static INLINE unsigned -compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) -{ - unsigned mask = 0; - unsigned i; - - /* Do the hardwired planes first: - */ - if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; - if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; - if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; - if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; - if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; - if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; - - /* Followed by any remaining ones: - */ - for (i = 6; i < nr; i++) { - if (dot4(clip, plane[i]) < 0) - mask |= (1<machine; - unsigned int j; - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - - assert(count <= 4); - assert(draw->vertex_shader->state->output_semantic_name[0] - == TGSI_SEMANTIC_POSITION); - - /* Consts does not require 16 byte alignment. */ - machine->Consts = (float (*)[4]) draw->user.constants; - - machine->Inputs = ALIGN16_ASSIGN(inputs); - machine->Outputs = ALIGN16_ASSIGN(outputs); - - draw->vertex_fetch.fetch_func( draw, machine, elts, count ); - - /* run shader */ -#ifdef MESA_LLVM - if (1) { - struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; - gallivm_cpu_vs_exec(prog, - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps); - } else -#elif defined(__i386__) || defined(__386__) - if (draw->use_sse) { - /* SSE */ - /* cast away const */ - struct draw_vertex_shader *shader - = (struct draw_vertex_shader *)draw->vertex_shader; - codegen_function func - = (codegen_function) x86_get_func( &shader->sse2_program ); - - if (func) - func( - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps ); - else - /* interpreter */ - tgsi_exec_machine_run( machine ); - } - else -#endif - { - /* interpreter */ - tgsi_exec_machine_run( machine ); - } - - /* store machine results */ - for (j = 0; j < count; j++) { - unsigned slot; - float x, y, z, w; - - /* Handle attr[0] (position) specially: - * - * XXX: Computing the clipmask should be done in the vertex - * program as a set of DP4 instructions appended to the - * user-provided code. - */ - x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; - -#if DBG_VS - debug_printf("output[%d]win: %f %f %f %f\n", j, - vOut[j]->data[0][0], - vOut[j]->data[0][1], - vOut[j]->data[0][2], - vOut[j]->data[0][3]); -#endif - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; -#if DBG_VS - debug_printf("output[%d][%d]: %f %f %f %f\n", j, slot, - vOut[j]->data[slot][0], - vOut[j]->data[slot][1], - vOut[j]->data[slot][2], - vOut[j]->data[slot][3]); -#endif - } - } /* loop over vertices */ -} - - -/** - * Run the vertex shader on all vertices in the vertex queue. - * Called by the draw module when the vertx cache needs to be flushed. - */ -void -draw_vertex_shader_queue_flush(struct draw_context *draw) -{ - unsigned i; - - assert(draw->vs.queue_nr != 0); - - /* XXX: do this on statechange: - */ - draw_update_vertex_fetch( draw ); - -// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); - - /* run vertex shader on vertex cache entries, four per invokation */ - for (i = 0; i < draw->vs.queue_nr; i += 4) { - struct vertex_header *dests[4]; - unsigned elts[4]; - int j, n = MIN2(4, draw->vs.queue_nr - i); - - for (j = 0; j < n; j++) { - elts[j] = draw->vs.queue[i + j].elt; - dests[j] = draw->vs.queue[i + j].dest; - } - - for ( ; j < 4; j++) { - elts[j] = elts[0]; - dests[j] = dests[0]; - } - - assert(n > 0); - assert(n <= 4); - - run_vertex_program(draw, elts, n, dests); - } - - draw->vs.queue_nr = 0; -} - - -struct draw_vertex_shader * -draw_create_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader) -{ - struct draw_vertex_shader *vs; - - vs = CALLOC_STRUCT( draw_vertex_shader ); - if (vs == NULL) { - return NULL; - } - - vs->state = shader; - -#ifdef MESA_LLVM - struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); - gallivm_ir_set_layout(ir, GALLIVM_SOA); - gallivm_ir_set_components(ir, 4); - gallivm_ir_fill_from_tgsi(ir, shader->tokens); - vs->llvm_prog = gallivm_ir_compile(ir); - gallivm_ir_delete(ir); - - draw->engine = gallivm_global_cpu_engine(); - if (!draw->engine) { - draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); - } - else { - gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); - } -#elif defined(__i386__) || defined(__386__) - if (draw->use_sse) { - /* cast-away const */ - struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; - - x86_init_func( &vs->sse2_program ); - if (!tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, - &vs->sse2_program )) { - x86_release_func( (struct x86_function *) &vs->sse2_program ); - fprintf(stdout /*err*/, - "tgsi_emit_sse2() failed, falling back to interpreter\n"); - } - } -#endif - - return vs; -} - - -void -draw_bind_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - - draw->vertex_shader = dvs; - draw->num_vs_outputs = dvs->state->num_outputs; - - /* specify the vertex program to interpret/execute */ - tgsi_exec_machine_init(&draw->machine, - draw->vertex_shader->state->tokens, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/ ); -} - - -void -draw_delete_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs) -{ -#if defined(__i386__) || defined(__386__) - x86_release_func( (struct x86_function *) &dvs->sse2_program ); -#endif - - FREE( dvs ); -} diff --git a/src/gallium/aux/draw/draw_vf.c b/src/gallium/aux/draw/draw_vf.c deleted file mode 100644 index dc3a5ecd21..0000000000 --- a/src/gallium/aux/draw/draw_vf.c +++ /dev/null @@ -1,428 +0,0 @@ -/* - * Copyright 2003 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: - * Keith Whitwell - */ - - -#include - -#include "pipe/p_compiler.h" -#include "pipe/p_util.h" - -#include "draw_vf.h" - - -#define DRAW_VF_DBG 0 - - -/* TODO: remove this */ -extern void -_mesa_exec_free( void *addr ); - - -static boolean match_fastpath( struct draw_vertex_fetch *vf, - const struct draw_vf_fastpath *fp) -{ - unsigned j; - - if (vf->attr_count != fp->attr_count) - return FALSE; - - for (j = 0; j < vf->attr_count; j++) - if (vf->attr[j].format != fp->attr[j].format || - vf->attr[j].inputsize != fp->attr[j].size || - vf->attr[j].vertoffset != fp->attr[j].offset) - return FALSE; - - if (fp->match_strides) { - if (vf->vertex_stride != fp->vertex_stride) - return FALSE; - - for (j = 0; j < vf->attr_count; j++) - if (vf->attr[j].inputstride != fp->attr[j].stride) - return FALSE; - } - - return TRUE; -} - -static boolean search_fastpath_emit( struct draw_vertex_fetch *vf ) -{ - struct draw_vf_fastpath *fp = vf->fastpath; - - for ( ; fp ; fp = fp->next) { - if (match_fastpath(vf, fp)) { - vf->emit = fp->func; - return TRUE; - } - } - - return FALSE; -} - -void draw_vf_register_fastpath( struct draw_vertex_fetch *vf, - boolean match_strides ) -{ - struct draw_vf_fastpath *fastpath = CALLOC_STRUCT(draw_vf_fastpath); - unsigned i; - - fastpath->vertex_stride = vf->vertex_stride; - fastpath->attr_count = vf->attr_count; - fastpath->match_strides = match_strides; - fastpath->func = vf->emit; - fastpath->attr = (struct draw_vf_attr_type *) - MALLOC(vf->attr_count * sizeof(fastpath->attr[0])); - - for (i = 0; i < vf->attr_count; i++) { - fastpath->attr[i].format = vf->attr[i].format; - fastpath->attr[i].stride = vf->attr[i].inputstride; - fastpath->attr[i].size = vf->attr[i].inputsize; - fastpath->attr[i].offset = vf->attr[i].vertoffset; - } - - fastpath->next = vf->fastpath; - vf->fastpath = fastpath; -} - - - - -/*********************************************************************** - * Build codegen functions or return generic ones: - */ -static void choose_emit_func( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *dest) -{ - vf->emit = NULL; - - /* Does this match an existing (hardwired, codegen or known-bad) - * fastpath? - */ - if (search_fastpath_emit(vf)) { - /* Use this result. If it is null, then it is already known - * that the current state will fail for codegen and there is no - * point trying again. - */ - } - else if (vf->codegen_emit) { - vf->codegen_emit( vf ); - } - - if (!vf->emit) { - draw_vf_generate_hardwired_emit(vf); - } - - /* Otherwise use the generic version: - */ - if (!vf->emit) - vf->emit = draw_vf_generic_emit; - - vf->emit( vf, count, dest ); -} - - - - - -/*********************************************************************** - * Public entrypoints, mostly dispatch to the above: - */ - - - -static unsigned -draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, - const struct draw_vf_attr_map *map, - unsigned nr, - unsigned vertex_stride ) -{ - unsigned offset = 0; - unsigned i, j; - - assert(nr < PIPE_ATTRIB_MAX); - - for (j = 0, i = 0; i < nr; i++) { - const unsigned format = map[i].format; - if (format == DRAW_EMIT_PAD) { -#if (DRAW_VF_DBG) - debug_printf("%d: pad %d, offset %d\n", i, - map[i].offset, offset); -#endif - - offset += map[i].offset; - - } - else { - vf->attr[j].attrib = map[i].attrib; - vf->attr[j].format = format; - vf->attr[j].insert = draw_vf_format_info[format].insert; - vf->attr[j].vertattrsize = draw_vf_format_info[format].attrsize; - vf->attr[j].vertoffset = offset; - vf->attr[j].isconst = draw_vf_format_info[format].isconst; - if(vf->attr[j].isconst) - memcpy(vf->attr[j].data, &map[i].data, vf->attr[j].vertattrsize); - -#if (DRAW_VF_DBG) - debug_printf("%d: %s, offset %d\n", i, - draw_vf_format_info[format].name, - vf->attr[j].vertoffset); -#endif - - offset += draw_vf_format_info[format].attrsize; - j++; - } - } - - vf->attr_count = j; - vf->vertex_stride = vertex_stride ? vertex_stride : offset; - vf->emit = choose_emit_func; - - assert(vf->vertex_stride >= offset); - return vf->vertex_stride; -} - - -void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, - const struct vertex_info *vinfo, - float point_size ) -{ - unsigned i, j, k; - struct draw_vf_attr *a = vf->attr; - struct draw_vf_attr_map attrs[PIPE_MAX_SHADER_INPUTS]; - unsigned count = 0; /* for debug/sanity */ - unsigned nr_attrs = 0; - - for (i = 0; i < vinfo->num_attribs; i++) { - j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; - case EMIT_ALL: { - /* just copy the whole vertex as-is to the vbuf */ - unsigned s = vinfo->size; - assert(i == 0); - assert(j == 0); - /* copy the vertex header */ - /* XXX: we actually don't copy the header, just pad it */ - attrs[nr_attrs].attrib = 0; - attrs[nr_attrs].format = DRAW_EMIT_PAD; - attrs[nr_attrs].offset = offsetof(struct vertex_header, data); - s -= offsetof(struct vertex_header, data)/4; - count += offsetof(struct vertex_header, data)/4; - nr_attrs++; - /* copy the vertex data */ - for(k = 0; k < (s & ~0x3); k += 4) { - attrs[nr_attrs].attrib = k/4; - attrs[nr_attrs].format = DRAW_EMIT_4F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 4; - } - /* tail */ - /* XXX: actually, this shouldn't be needed */ - attrs[nr_attrs].attrib = k/4; - attrs[nr_attrs].offset = 0; - switch(s & 0x3) { - case 0: - break; - case 1: - attrs[nr_attrs].format = DRAW_EMIT_1F; - nr_attrs++; - count += 1; - break; - case 2: - attrs[nr_attrs].format = DRAW_EMIT_2F; - nr_attrs++; - count += 2; - break; - case 3: - attrs[nr_attrs].format = DRAW_EMIT_3F; - nr_attrs++; - count += 3; - break; - } - break; - } - case EMIT_1F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_1F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count++; - break; - case EMIT_1F_PSIZE: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_1F_CONST; - attrs[nr_attrs].offset = 0; - attrs[nr_attrs].data.f[0] = point_size; - nr_attrs++; - count++; - break; - case EMIT_2F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_2F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 2; - break; - case EMIT_3F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_3F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 3; - break; - case EMIT_4F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_4F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 4; - break; - case EMIT_4UB: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_4UB_4F_BGRA; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 1; - break; - default: - assert(0); - } - } - - assert(count == vinfo->size); - - draw_vf_set_vertex_attributes(vf, - attrs, - nr_attrs, - vinfo->size * sizeof(float) ); - - for (j = 0; j < vf->attr_count; j++) { - a[j].inputsize = 4; - a[j].do_insert = a[j].insert[4 - 1]; - if(a[j].isconst) { - a[j].inputptr = a[j].data; - a[j].inputstride = 0; - } - } -} - - -#if 0 -/* Set attribute pointers, adjusted for start position: - */ -void draw_vf_set_sources( struct draw_vertex_fetch *vf, - GLvector4f * const sources[], - unsigned start ) -{ - struct draw_vf_attr *a = vf->attr; - unsigned j; - - for (j = 0; j < vf->attr_count; j++) { - const GLvector4f *vptr = sources[a[j].attrib]; - - if ((a[j].inputstride != vptr->stride) || - (a[j].inputsize != vptr->size)) - vf->emit = choose_emit_func; - - a[j].inputstride = vptr->stride; - a[j].inputsize = vptr->size; - a[j].do_insert = a[j].insert[vptr->size - 1]; - a[j].inputptr = ((uint8_t *)vptr->data) + start * vptr->stride; - } -} -#endif - - -/** - * Emit a vertex to dest. - */ -void draw_vf_emit_vertex( struct draw_vertex_fetch *vf, - struct vertex_header *vertex, - void *dest ) -{ - struct draw_vf_attr *a = vf->attr; - unsigned j; - - for (j = 0; j < vf->attr_count; j++) { - if (!a[j].isconst) { - a[j].inputptr = (uint8_t *)&vertex->data[a[j].attrib][0]; - a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ - } - } - - vf->emit( vf, 1, (uint8_t*) dest ); -} - - - -struct draw_vertex_fetch *draw_vf_create( void ) -{ - struct draw_vertex_fetch *vf = CALLOC_STRUCT(draw_vertex_fetch); - unsigned i; - - for (i = 0; i < PIPE_ATTRIB_MAX; i++) - vf->attr[i].vf = vf; - - vf->identity[0] = 0.0; - vf->identity[1] = 0.0; - vf->identity[2] = 0.0; - vf->identity[3] = 1.0; - - vf->codegen_emit = NULL; - -#ifdef USE_SSE_ASM - if (!GETENV("GALLIUM_NO_CODEGEN")) - vf->codegen_emit = draw_vf_generate_sse_emit; -#endif - - return vf; -} - - -void draw_vf_destroy( struct draw_vertex_fetch *vf ) -{ - struct draw_vf_fastpath *fp, *tmp; - - for (fp = vf->fastpath ; fp ; fp = tmp) { - tmp = fp->next; - FREE(fp->attr); - - /* KW: At the moment, fp->func is constrained to be allocated by - * _mesa_exec_alloc(), as the hardwired fastpaths in - * t_vertex_generic.c are handled specially. It would be nice - * to unify them, but this probably won't change until this - * module gets another overhaul. - */ - //_mesa_exec_free((void *) fp->func); - FREE(fp); - } - - vf->fastpath = NULL; - FREE(vf); -} diff --git a/src/gallium/aux/draw/draw_vf.h b/src/gallium/aux/draw/draw_vf.h deleted file mode 100644 index 011c8f0ff1..0000000000 --- a/src/gallium/aux/draw/draw_vf.h +++ /dev/null @@ -1,236 +0,0 @@ -/* - * Copyright 2008 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/** - * Vertex fetch/store/convert code. This functionality is used in two places: - * 1. Vertex fetch/convert - to grab vertex data from incoming vertex - * arrays and convert to format needed by vertex shaders. - * 2. Vertex store/emit - to convert simple float[][4] vertex attributes - * (which is the organization used throughout the draw/prim pipeline) to - * hardware-specific formats and emit into hardware vertex buffers. - * - * - * Authors: - * Keith Whitwell - */ - -#ifndef DRAW_VF_H -#define DRAW_VF_H - - -#include "pipe/p_compiler.h" -#include "pipe/p_state.h" - -#include "draw_vertex.h" -#include "draw_private.h" /* for vertex_header */ - - -enum draw_vf_attr_format { - DRAW_EMIT_1F, - DRAW_EMIT_2F, - DRAW_EMIT_3F, - DRAW_EMIT_4F, - DRAW_EMIT_3F_XYW, /**< for projective texture */ - DRAW_EMIT_1UB_1F, /**< for fog coordinate */ - DRAW_EMIT_3UB_3F_RGB, /**< for specular color */ - DRAW_EMIT_3UB_3F_BGR, /**< for specular color */ - DRAW_EMIT_4UB_4F_RGBA, /**< for color */ - DRAW_EMIT_4UB_4F_BGRA, /**< for color */ - DRAW_EMIT_4UB_4F_ARGB, /**< for color */ - DRAW_EMIT_4UB_4F_ABGR, /**< for color */ - DRAW_EMIT_1F_CONST, - DRAW_EMIT_2F_CONST, - DRAW_EMIT_3F_CONST, - DRAW_EMIT_4F_CONST, - DRAW_EMIT_PAD, /**< leave a hole of 'offset' bytes */ - DRAW_EMIT_MAX -}; - -struct draw_vf_attr_map -{ - /** Input attribute number */ - unsigned attrib; - - enum draw_vf_attr_format format; - - unsigned offset; - - /** - * Constant data for DRAW_EMIT_*_CONST - */ - union { - uint8_t ub[4]; - float f[4]; - } data; -}; - -struct draw_vertex_fetch; - - - -#if 0 -unsigned -draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, - const struct draw_vf_attr_map *map, - unsigned nr, - unsigned vertex_stride ); -#endif - -void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, - const struct vertex_info *vinfo, - float point_size ); - -#if 0 -void -draw_vf_set_sources( struct draw_vertex_fetch *vf, - GLvector4f * const attrib[], - unsigned start ); -#endif - -void -draw_vf_emit_vertex( struct draw_vertex_fetch *vf, - struct vertex_header *vertex, - void *dest ); - -struct draw_vertex_fetch * -draw_vf_create( void ); - -void -draw_vf_destroy( struct draw_vertex_fetch *vf ); - - - -/*********************************************************************** - * Internal functions and structs: - */ - -struct draw_vf_attr; - -typedef void (*draw_vf_extract_func)( const struct draw_vf_attr *a, - float *out, - const uint8_t *v ); - -typedef void (*draw_vf_insert_func)( const struct draw_vf_attr *a, - uint8_t *v, - const float *in ); - -typedef void (*draw_vf_emit_func)( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *dest ); - - - -/** - * Describes how to convert/move a vertex attribute from a vertex - * array to a vertex structure. - */ -struct draw_vf_attr -{ - struct draw_vertex_fetch *vf; - - unsigned format; - unsigned inputsize; - unsigned inputstride; - unsigned vertoffset; /**< position of the attrib in the vertex struct */ - - boolean isconst; /**< read from const data below */ - uint8_t data[16]; - - unsigned attrib; /**< which vertex attrib (0=position, etc) */ - unsigned vertattrsize; /**< size of the attribute in bytes */ - - uint8_t *inputptr; - const draw_vf_insert_func *insert; - draw_vf_insert_func do_insert; - draw_vf_extract_func extract; -}; - -struct draw_vertex_fetch -{ - struct draw_vf_attr attr[PIPE_ATTRIB_MAX]; - unsigned attr_count; - unsigned vertex_stride; - - draw_vf_emit_func emit; - - /* Parameters and constants for codegen: - */ - float identity[4]; - - struct draw_vf_fastpath *fastpath; - - void (*codegen_emit)( struct draw_vertex_fetch *vf ); -}; - - -struct draw_vf_attr_type { - unsigned format; - unsigned size; - unsigned stride; - unsigned offset; -}; - -/** XXX this could be moved into draw_vf.c */ -struct draw_vf_fastpath { - unsigned vertex_stride; - unsigned attr_count; - boolean match_strides; - - struct draw_vf_attr_type *attr; - - draw_vf_emit_func func; - struct draw_vf_fastpath *next; -}; - - -void -draw_vf_register_fastpath( struct draw_vertex_fetch *vtx, - boolean match_strides ); - -void -draw_vf_generic_emit( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *v ); - -void -draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ); - -void -draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ); - - -/** XXX this type and function could probably be moved into draw_vf.c */ -struct draw_vf_format_info { - const char *name; - draw_vf_insert_func insert[4]; - const unsigned attrsize; - const boolean isconst; -}; - -extern const struct draw_vf_format_info -draw_vf_format_info[DRAW_EMIT_MAX]; - - -#endif diff --git a/src/gallium/aux/draw/draw_vf_generic.c b/src/gallium/aux/draw/draw_vf_generic.c deleted file mode 100644 index 7a60a9db9c..0000000000 --- a/src/gallium/aux/draw/draw_vf_generic.c +++ /dev/null @@ -1,585 +0,0 @@ - -/* - * Copyright 2003 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: - * Keith Whitwell - */ - - -#include "pipe/p_compiler.h" -#include "pipe/p_debug.h" -#include "pipe/p_util.h" - -#include "draw_vf.h" - - - -static INLINE void insert_4f_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; - out[3] = in[3]; -} - -static INLINE void insert_4f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; - out[3] = 1; -} - -static INLINE void insert_4f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = 0; - out[3] = 1; -} - -static INLINE void insert_4f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = 0; - out[2] = 0; - out[3] = 1; -} - -static INLINE void insert_3f_xyw_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[3]; -} - -static INLINE void insert_3f_xyw_err( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - (void) a; (void) v; (void) in; - assert(0); -} - -static INLINE void insert_3f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; -} - -static INLINE void insert_3f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = 0; -} - -static INLINE void insert_3f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = 0; - out[2] = 0; -} - - -static INLINE void insert_2f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; -} - -static INLINE void insert_2f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = 0; -} - -static INLINE void insert_1f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; -} - -static INLINE void insert_null( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - (void) a; (void) v; (void) in; -} - -static INLINE void insert_4ub_4f_rgba_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); -} - -static INLINE void insert_4ub_4f_rgba_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_rgba_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[2] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_rgba_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - v[1] = 0; - v[2] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_bgra_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); -} - -static INLINE void insert_4ub_4f_bgra_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_bgra_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[0] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_bgra_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - v[1] = 0; - v[0] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_argb_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); -} - -static INLINE void insert_4ub_4f_argb_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_argb_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - v[3] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_argb_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - v[2] = 0x00; - v[3] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_abgr_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); -} - -static INLINE void insert_4ub_4f_abgr_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_abgr_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - v[1] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_abgr_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - v[2] = 0x00; - v[1] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_3ub_3f_rgb_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); -} - -static INLINE void insert_3ub_3f_rgb_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[2] = 0; -} - -static INLINE void insert_3ub_3f_rgb_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - v[1] = 0; - v[2] = 0; -} - -static INLINE void insert_3ub_3f_bgr_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); -} - -static INLINE void insert_3ub_3f_bgr_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[0] = 0; -} - -static INLINE void insert_3ub_3f_bgr_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - v[1] = 0; - v[0] = 0; -} - - -static INLINE void insert_1ub_1f_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); -} - - -const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX] = -{ - { "1f", - { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, - sizeof(float), FALSE }, - - { "2f", - { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, - 2 * sizeof(float), FALSE }, - - { "3f", - { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, - 3 * sizeof(float), FALSE }, - - { "4f", - { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, - 4 * sizeof(float), FALSE }, - - { "3f_xyw", - { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, - insert_3f_xyw_4 }, - 3 * sizeof(float), FALSE }, - - { "1ub_1f", - { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 }, - sizeof(uint8_t), FALSE }, - - { "3ub_3f_rgb", - { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3, - insert_3ub_3f_rgb_3 }, - 3 * sizeof(uint8_t), FALSE }, - - { "3ub_3f_bgr", - { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3, - insert_3ub_3f_bgr_3 }, - 3 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_rgba", - { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, - insert_4ub_4f_rgba_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_bgra", - { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3, - insert_4ub_4f_bgra_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_argb", - { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3, - insert_4ub_4f_argb_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_abgr", - { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3, - insert_4ub_4f_abgr_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "1f_const", - { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, - sizeof(float), TRUE }, - - { "2f_const", - { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, - 2 * sizeof(float), TRUE }, - - { "3f_const", - { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, - 3 * sizeof(float), TRUE }, - - { "4f_const", - { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, - 4 * sizeof(float), TRUE }, - - { "pad", - { NULL, NULL, NULL, NULL }, - 0, FALSE }, - -}; - - - - -/*********************************************************************** - * Hardwired fastpaths for emitting whole vertices or groups of - * vertices - */ -#define EMIT5(NR, F0, F1, F2, F3, F4, NAME) \ -static void NAME( struct draw_vertex_fetch *vf, \ - unsigned count, \ - uint8_t *v ) \ -{ \ - struct draw_vf_attr *a = vf->attr; \ - unsigned i; \ - \ - for (i = 0 ; i < count ; i++, v += vf->vertex_stride) { \ - if (NR > 0) { \ - F0( &a[0], v + a[0].vertoffset, (float *)a[0].inputptr ); \ - a[0].inputptr += a[0].inputstride; \ - } \ - \ - if (NR > 1) { \ - F1( &a[1], v + a[1].vertoffset, (float *)a[1].inputptr ); \ - a[1].inputptr += a[1].inputstride; \ - } \ - \ - if (NR > 2) { \ - F2( &a[2], v + a[2].vertoffset, (float *)a[2].inputptr ); \ - a[2].inputptr += a[2].inputstride; \ - } \ - \ - if (NR > 3) { \ - F3( &a[3], v + a[3].vertoffset, (float *)a[3].inputptr ); \ - a[3].inputptr += a[3].inputstride; \ - } \ - \ - if (NR > 4) { \ - F4( &a[4], v + a[4].vertoffset, (float *)a[4].inputptr ); \ - a[4].inputptr += a[4].inputstride; \ - } \ - } \ -} - - -#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \ - insert_null, insert_null, NAME) - -#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \ - insert_null, NAME) - -#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \ - insert_null, NAME) - - -EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4) - -EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2) - -EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2) - - -/* Use the codegen paths to select one of a number of hardwired - * fastpaths. - */ -void draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ) -{ - draw_vf_emit_func func = NULL; - - /* Does it fit a hardwired fastpath? Help! this is growing out of - * control! - */ - switch (vf->attr_count) { - case 2: - if (vf->attr[0].do_insert == insert_3f_3 && - vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - func = emit_xyz3_rgba4; - } - break; - case 3: - if (vf->attr[2].do_insert == insert_2f_2) { - if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - if (vf->attr[0].do_insert == insert_4f_4) - func = emit_xyzw4_rgba4_st2; - } - } - break; - case 4: - if (vf->attr[2].do_insert == insert_2f_2 && - vf->attr[3].do_insert == insert_2f_2) { - if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - if (vf->attr[0].do_insert == insert_4f_4) - func = emit_xyzw4_rgba4_st2_st2; - } - } - break; - } - - vf->emit = func; -} - -/*********************************************************************** - * Generic (non-codegen) functions for whole vertices or groups of - * vertices - */ - -void draw_vf_generic_emit( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *v ) -{ - struct draw_vf_attr *a = vf->attr; - const unsigned attr_count = vf->attr_count; - const unsigned stride = vf->vertex_stride; - unsigned i, j; - - for (i = 0 ; i < count ; i++, v += stride) { - for (j = 0; j < attr_count; j++) { - float *in = (float *)a[j].inputptr; - a[j].inputptr += a[j].inputstride; - a[j].do_insert( &a[j], v + a[j].vertoffset, in ); - } - } -} - - diff --git a/src/gallium/aux/draw/draw_vf_sse.c b/src/gallium/aux/draw/draw_vf_sse.c deleted file mode 100644 index 1ad2ae756d..0000000000 --- a/src/gallium/aux/draw/draw_vf_sse.c +++ /dev/null @@ -1,614 +0,0 @@ -/* - * Copyright 2003 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: - * Keith Whitwell - */ - - -#include "simple_list.h" - -#include "pipe/p_compiler.h" - -#include "draw_vf.h" - - -#if defined(USE_SSE_ASM) - -#include "x86/rtasm/x86sse.h" -#include "x86/common_x86_asm.h" - - -#define X 0 -#define Y 1 -#define Z 2 -#define W 3 - - -struct x86_program { - struct x86_function func; - - struct draw_vertex_fetch *vf; - boolean inputs_safe; - boolean outputs_safe; - boolean have_sse2; - - struct x86_reg identity; - struct x86_reg chan0; -}; - - -static struct x86_reg get_identity( struct x86_program *p ) -{ - return p->identity; -} - -static void emit_load4f_4( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movups(&p->func, dest, arg0); -} - -static void emit_load4f_3( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Have to jump through some hoops: - * - * c 0 0 0 - * c 0 0 1 - * 0 0 c 1 - * a b c 1 - */ - sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); - sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); - sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) ); - sse_movlps(&p->func, dest, arg0); -} - -static void emit_load4f_2( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Initialize from identity, then pull in low two words: - */ - sse_movups(&p->func, dest, get_identity(p)); - sse_movlps(&p->func, dest, arg0); -} - -static void emit_load4f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Pull in low word, then swizzle in identity */ - sse_movss(&p->func, dest, arg0); - sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); -} - - - -static void emit_load3f_3( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Over-reads by 1 dword - potential SEGV if input is a vertex - * array. - */ - if (p->inputs_safe) { - sse_movups(&p->func, dest, arg0); - } - else { - /* c 0 0 0 - * c c c c - * a b c c - */ - sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); - sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X)); - sse_movlps(&p->func, dest, arg0); - } -} - -static void emit_load3f_2( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - emit_load4f_2(p, dest, arg0); -} - -static void emit_load3f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - emit_load4f_1(p, dest, arg0); -} - -static void emit_load2f_2( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movlps(&p->func, dest, arg0); -} - -static void emit_load2f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - emit_load4f_1(p, dest, arg0); -} - -static void emit_load1f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movss(&p->func, dest, arg0); -} - -static void (*load[4][4])( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) = { - { emit_load1f_1, - emit_load1f_1, - emit_load1f_1, - emit_load1f_1 }, - - { emit_load2f_1, - emit_load2f_2, - emit_load2f_2, - emit_load2f_2 }, - - { emit_load3f_1, - emit_load3f_2, - emit_load3f_3, - emit_load3f_3 }, - - { emit_load4f_1, - emit_load4f_2, - emit_load4f_3, - emit_load4f_4 } -}; - -static void emit_load( struct x86_program *p, - struct x86_reg dest, - unsigned sz, - struct x86_reg src, - unsigned src_sz) -{ - load[sz-1][src_sz-1](p, dest, src); -} - -static void emit_store4f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movups(&p->func, dest, arg0); -} - -static void emit_store3f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - if (p->outputs_safe) { - /* Emit the extra dword anyway. This may hurt writecombining, - * may cause other problems. - */ - sse_movups(&p->func, dest, arg0); - } - else { - /* Alternate strategy - emit two, shuffle, emit one. - */ - sse_movlps(&p->func, dest, arg0); - sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */ - sse_movss(&p->func, x86_make_disp(dest,8), arg0); - } -} - -static void emit_store2f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movlps(&p->func, dest, arg0); -} - -static void emit_store1f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movss(&p->func, dest, arg0); -} - - -static void (*store[4])( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) = -{ - emit_store1f, - emit_store2f, - emit_store3f, - emit_store4f -}; - -static void emit_store( struct x86_program *p, - struct x86_reg dest, - unsigned sz, - struct x86_reg temp ) - -{ - store[sz-1](p, dest, temp); -} - -static void emit_pack_store_4ub( struct x86_program *p, - struct x86_reg dest, - struct x86_reg temp ) -{ - /* Scale by 255.0 - */ - sse_mulps(&p->func, temp, p->chan0); - - if (p->have_sse2) { - sse2_cvtps2dq(&p->func, temp, temp); - sse2_packssdw(&p->func, temp, temp); - sse2_packuswb(&p->func, temp, temp); - sse_movss(&p->func, dest, temp); - } - else { - struct x86_reg mmx0 = x86_make_reg(file_MMX, 0); - struct x86_reg mmx1 = x86_make_reg(file_MMX, 1); - sse_cvtps2pi(&p->func, mmx0, temp); - sse_movhlps(&p->func, temp, temp); - sse_cvtps2pi(&p->func, mmx1, temp); - mmx_packssdw(&p->func, mmx0, mmx1); - mmx_packuswb(&p->func, mmx0, mmx0); - mmx_movd(&p->func, dest, mmx0); - } -} - -static int get_offset( const void *a, const void *b ) -{ - return (const char *)b - (const char *)a; -} - -/* Not much happens here. Eventually use this function to try and - * avoid saving/reloading the source pointers each vertex (if some of - * them can fit in registers). - */ -static void get_src_ptr( struct x86_program *p, - struct x86_reg srcREG, - struct x86_reg vfREG, - struct draw_vf_attr *a ) -{ - struct draw_vertex_fetch *vf = p->vf; - struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); - - /* Load current a[j].inputptr - */ - x86_mov(&p->func, srcREG, ptr_to_src); -} - -static void update_src_ptr( struct x86_program *p, - struct x86_reg srcREG, - struct x86_reg vfREG, - struct draw_vf_attr *a ) -{ - if (a->inputstride) { - struct draw_vertex_fetch *vf = p->vf; - struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); - - /* add a[j].inputstride (hardcoded value - could just as easily - * pull the stride value from memory each time). - */ - x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride)); - - /* save new value of a[j].inputptr - */ - x86_mov(&p->func, ptr_to_src, srcREG); - } -} - - -/* Lots of hardcoding - * - * EAX -- pointer to current output vertex - * ECX -- pointer to current attribute - * - */ -static boolean build_vertex_emit( struct x86_program *p ) -{ - struct draw_vertex_fetch *vf = p->vf; - unsigned j = 0; - - struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX); - struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX); - struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); - struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI); - struct x86_reg temp = x86_make_reg(file_XMM, 0); - uint8_t *fixup, *label; - - /* Push a few regs? - */ - x86_push(&p->func, countEBP); - x86_push(&p->func, vfESI); - - - /* Get vertex count, compare to zero - */ - x86_xor(&p->func, srcECX, srcECX); - x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2)); - x86_cmp(&p->func, countEBP, srcECX); - fixup = x86_jcc_forward(&p->func, cc_E); - - /* Initialize destination register. - */ - x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3)); - - /* Move argument 1 (vf) into a reg: - */ - x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1)); - - - /* always load, needed or not: - */ - sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0]))); - - /* Note address for loop jump */ - label = x86_get_label(&p->func); - - /* Emit code for each of the attributes. Currently routes - * everything through SSE registers, even when it might be more - * efficient to stick with regular old x86. No optimization or - * other tricks - enough new ground to cover here just getting - * things working. - */ - while (j < vf->attr_count) { - struct draw_vf_attr *a = &vf->attr[j]; - struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset); - - /* Now, load an XMM reg from src, perhaps transform, then save. - * Could be shortcircuited in specific cases: - */ - switch (a->format) { - case DRAW_EMIT_1F: - case DRAW_EMIT_1F_CONST: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 1, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_2F: - case DRAW_EMIT_2F_CONST: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 2, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_3F: - case DRAW_EMIT_3F_CONST: - /* Potentially the worst case - hardcode 2+1 copying: - */ - if (0) { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 3, temp); - update_src_ptr(p, srcECX, vfESI, a); - } - else { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 2, temp); - if (a->inputsize > 2) { - emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1); - emit_store(p, x86_make_disp(dest,8), 1, temp); - } - else { - sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p)); - } - update_src_ptr(p, srcECX, vfESI, a); - } - break; - case DRAW_EMIT_4F: - case DRAW_EMIT_4F_CONST: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 4, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_3F_XYW: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z)); - emit_store(p, dest, 3, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - - case DRAW_EMIT_1UB_1F: - /* Test for PAD3 + 1UB: - */ - if (j > 0 && - a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3) - { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X)); - emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */ - update_src_ptr(p, srcECX, vfESI, a); - } - else { - debug_printf("Can't emit 1ub %x %x %d\n", - a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); - return FALSE; - } - break; - case DRAW_EMIT_3UB_3F_RGB: - case DRAW_EMIT_3UB_3F_BGR: - /* Test for 3UB + PAD1: - */ - if (j == vf->attr_count - 1 || - a[1].vertoffset >= a->vertoffset + 4) { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - if (a->format == DRAW_EMIT_3UB_3F_BGR) - sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - } - /* Test for 3UB + 1UB: - */ - else if (j < vf->attr_count - 1 && - a[1].format == DRAW_EMIT_1UB_1F && - a[1].vertoffset == a->vertoffset + 3) { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - update_src_ptr(p, srcECX, vfESI, a); - - /* Make room for incoming value: - */ - sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); - - get_src_ptr(p, srcECX, vfESI, &a[1]); - emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize); - update_src_ptr(p, srcECX, vfESI, &a[1]); - - /* Rearrange and possibly do BGR conversion: - */ - if (a->format == DRAW_EMIT_3UB_3F_BGR) - sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); - else - sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X)); - - emit_pack_store_4ub(p, dest, temp); - j++; /* NOTE: two attrs consumed */ - } - else { - debug_printf("Can't emit 3ub\n"); - } - return FALSE; /* add this later */ - break; - - case DRAW_EMIT_4UB_4F_RGBA: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4UB_4F_BGRA: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4UB_4F_ARGB: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4UB_4F_ABGR: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - default: - debug_printf("unknown a[%d].format %d\n", j, a->format); - return FALSE; /* catch any new opcodes */ - } - - /* Increment j by at least 1 - may have been incremented above also: - */ - j++; - } - - /* Next vertex: - */ - x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vf->vertex_stride)); - - /* decr count, loop if not zero - */ - x86_dec(&p->func, countEBP); - x86_test(&p->func, countEBP, countEBP); - x86_jcc(&p->func, cc_NZ, label); - - /* Exit mmx state? - */ - if (p->func.need_emms) - mmx_emms(&p->func); - - /* Land forward jump here: - */ - x86_fixup_fwd_jump(&p->func, fixup); - - /* Pop regs and return - */ - x86_pop(&p->func, x86_get_base_reg(vfESI)); - x86_pop(&p->func, countEBP); - x86_ret(&p->func); - - vf->emit = (draw_vf_emit_func)x86_get_func(&p->func); - return TRUE; -} - - - -void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) -{ - struct x86_program p; - - if (!cpu_has_xmm) { - vf->codegen_emit = NULL; - return; - } - - memset(&p, 0, sizeof(p)); - - p.vf = vf; - p.inputs_safe = 0; /* for now */ - p.outputs_safe = 1; /* for now */ - p.have_sse2 = cpu_has_xmm2; - p.identity = x86_make_reg(file_XMM, 6); - p.chan0 = x86_make_reg(file_XMM, 7); - - x86_init_func(&p.func); - - if (build_vertex_emit(&p)) { - draw_vf_register_fastpath( vf, TRUE ); - } - else { - /* Note the failure so that we don't keep trying to codegen an - * impossible state: - */ - draw_vf_register_fastpath( vf, FALSE ); - x86_release_func(&p.func); - } -} - -#else - -void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) -{ - /* Dummy version for when USE_SSE_ASM not defined */ -} - -#endif diff --git a/src/gallium/aux/draw/draw_wide_prims.c b/src/gallium/aux/draw/draw_wide_prims.c deleted file mode 100644 index 655774b155..0000000000 --- a/src/gallium/aux/draw/draw_wide_prims.c +++ /dev/null @@ -1,432 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -struct wide_stage { - struct draw_stage stage; - - float half_line_width; - float half_point_size; - - uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; - uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; - uint num_texcoords; - - int psize_slot; -}; - - - -static INLINE struct wide_stage *wide_stage( struct draw_stage *stage ) -{ - return (struct wide_stage *)stage; -} - - -static void passthrough_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void passthrough_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line(stage->next, header); -} - -static void passthrough_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->tri(stage->next, header); -} - - -/** - * Draw a wide line by drawing a quad (two triangles). - * XXX need to disable polygon stipple. - */ -static void wide_line( struct draw_stage *stage, - struct prim_header *header ) -{ - const struct wide_stage *wide = wide_stage(stage); - const float half_width = wide->half_line_width; - - struct prim_header tri; - - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - const float dx = FABSF(pos0[0] - pos2[0]); - const float dy = FABSF(pos0[1] - pos2[1]); - - /* - * Draw wide line as a quad (two tris) by "stretching" the line along - * X or Y. - * We need to tweak coords in several ways to be conformant here. - */ - - if (dx > dy) { - /* x-major line */ - pos0[1] = pos0[1] - half_width - 0.25f; - pos1[1] = pos1[1] + half_width - 0.25f; - pos2[1] = pos2[1] - half_width - 0.25f; - pos3[1] = pos3[1] + half_width - 0.25f; - if (pos0[0] < pos2[0]) { - /* left to right line */ - pos0[0] -= 0.5f; - pos1[0] -= 0.5f; - pos2[0] -= 0.5f; - pos3[0] -= 0.5f; - } - else { - /* right to left line */ - pos0[0] += 0.5f; - pos1[0] += 0.5f; - pos2[0] += 0.5f; - pos3[0] += 0.5f; - } - } - else { - /* y-major line */ - pos0[0] = pos0[0] - half_width + 0.25f; - pos1[0] = pos1[0] + half_width + 0.25f; - pos2[0] = pos2[0] - half_width + 0.25f; - pos3[0] = pos3[0] + half_width + 0.25f; - if (pos0[1] < pos2[1]) { - /* top to bottom line */ - pos0[1] -= 0.5f; - pos1[1] -= 0.5f; - pos2[1] -= 0.5f; - pos3[1] -= 0.5f; - } - else { - /* bottom to top line */ - pos0[1] += 0.5f; - pos1[1] += 0.5f; - pos2[1] += 0.5f; - pos3[1] += 0.5f; - } - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -/** - * Draw a wide line by drawing a quad, using geometry which will - * fullfill GL's antialiased line requirements. - */ -static void wide_line_aa(struct draw_stage *stage, - struct prim_header *header) -{ - const struct wide_stage *wide = wide_stage(stage); - const float half_width = wide->half_line_width; - struct prim_header tri; - struct vertex_header *v[4]; - float *pos; - float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; - float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; - const float len = (float) sqrt(dx * dx + dy * dy); - uint i; - - dx = dx * half_width / len; - dy = dy * half_width / len; - - /* allocate/dup new verts */ - for (i = 0; i < 4; i++) { - v[i] = dup_vert(stage, header->v[i/2], i); - } - - /* - * Quad for line from v0 to v1: - * - * 1 3 - * +-------------------------+ - * | | - * *v0 v1* - * | | - * +-------------------------+ - * 0 2 - */ - - pos = v[0]->data[0]; - pos[0] += dy; - pos[1] -= dx; - - pos = v[1]->data[0]; - pos[0] -= dy; - pos[1] += dx; - - pos = v[2]->data[0]; - pos[0] += dy; - pos[1] -= dx; - - pos = v[3]->data[0]; - pos[0] -= dy; - pos[1] += dx; - - tri.det = header->det; /* only the sign matters */ - - tri.v[0] = v[2]; tri.v[1] = v[1]; tri.v[2] = v[0]; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v[3]; tri.v[1] = v[1]; tri.v[2] = v[2]; - stage->next->tri( stage->next, &tri ); - -} - - -/** - * Set the vertex texcoords for sprite mode. - * Coords may be left untouched or set to a right-side-up or upside-down - * orientation. - */ -static void set_texcoords(const struct wide_stage *wide, - struct vertex_header *v, const float tc[4]) -{ - uint i; - for (i = 0; i < wide->num_texcoords; i++) { - if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { - uint j = wide->texcoord_slot[i]; - v->data[j][0] = tc[0]; - if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) - v->data[j][1] = 1.0f - tc[1]; - else - v->data[j][1] = tc[1]; - v->data[j][2] = tc[2]; - v->data[j][3] = tc[3]; - } - } -} - - -/* If there are lots of sprite points (and why wouldn't there be?) it - * would probably be more sensible to change hardware setup to - * optimize this rather than doing the whole thing in software like - * this. - */ -static void wide_point( struct draw_stage *stage, - struct prim_header *header ) -{ - const struct wide_stage *wide = wide_stage(stage); - const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; - float half_size; - float left_adj, right_adj; - - struct prim_header tri; - - /* four dups of original vertex */ - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - /* point size is either per-vertex or fixed size */ - if (wide->psize_slot >= 0) { - half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; - } - else { - half_size = wide->half_point_size; - } - - left_adj = -half_size + 0.25f; - right_adj = half_size + 0.25f; - - pos0[0] += left_adj; - pos0[1] -= half_size; - - pos1[0] += left_adj; - pos1[1] += half_size; - - pos2[0] += right_adj; - pos2[1] -= half_size; - - pos3[0] += right_adj; - pos3[1] += half_size; - - if (sprite) { - static const float tex00[4] = { 0, 0, 0, 1 }; - static const float tex01[4] = { 0, 1, 0, 1 }; - static const float tex11[4] = { 1, 1, 0, 1 }; - static const float tex10[4] = { 1, 0, 0, 1 }; - set_texcoords( wide, v0, tex00 ); - set_texcoords( wide, v1, tex01 ); - set_texcoords( wide, v2, tex10 ); - set_texcoords( wide, v3, tex11 ); - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -static void wide_first_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct wide_stage *wide = wide_stage(stage); - struct draw_context *draw = stage->draw; - - wide->half_point_size = 0.5f * draw->rasterizer->point_size; - - if (draw->rasterizer->point_size != 1.0) { - stage->point = wide_point; - } - else { - stage->point = passthrough_point; - } - - if (draw->rasterizer->point_sprite) { - /* find vertex shader texcoord outputs */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i, j = 0; - for (i = 0; i < vs->state->num_outputs; i++) { - if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { - wide->texcoord_slot[j] = i; - wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; - j++; - } - } - wide->num_texcoords = j; - } - - wide->psize_slot = -1; - - if (draw->rasterizer->point_size_per_vertex) { - /* find PSIZ vertex output */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i; - for (i = 0; i < vs->state->num_outputs; i++) { - if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { - wide->psize_slot = i; - break; - } - } - } - - stage->point( stage, header ); -} - - - -static void wide_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct wide_stage *wide = wide_stage(stage); - struct draw_context *draw = stage->draw; - - wide->half_line_width = 0.5f * draw->rasterizer->line_width; - - if (draw->rasterizer->line_width != 1.0) { - if (draw->rasterizer->line_smooth) - wide->stage.line = wide_line_aa; - else - wide->stage.line = wide_line; - } - else { - wide->stage.line = passthrough_line; - } - - stage->line( stage, header ); -} - - -static void wide_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->line = wide_first_line; - stage->point = wide_first_point; - stage->next->flush( stage->next, flags ); -} - - -static void wide_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void wide_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -struct draw_stage *draw_wide_stage( struct draw_context *draw ) -{ - struct wide_stage *wide = CALLOC_STRUCT(wide_stage); - - draw_alloc_temp_verts( &wide->stage, 4 ); - - wide->stage.draw = draw; - wide->stage.next = NULL; - wide->stage.point = wide_first_point; - wide->stage.line = wide_first_line; - wide->stage.tri = passthrough_tri; - wide->stage.flush = wide_flush; - wide->stage.reset_stipple_counter = wide_reset_stipple_counter; - wide->stage.destroy = wide_destroy; - - return &wide->stage; -} diff --git a/src/gallium/aux/llvm/Makefile b/src/gallium/aux/llvm/Makefile deleted file mode 100644 index e6ac399d08..0000000000 --- a/src/gallium/aux/llvm/Makefile +++ /dev/null @@ -1,85 +0,0 @@ -# -*-makefile-*- -TOP = ../../../.. -include $(TOP)/configs/current - -LIBNAME = gallivm - - -GALLIVM_SOURCES = \ - gallivm.cpp \ - gallivm_cpu.cpp \ - instructions.cpp \ - loweringpass.cpp \ - tgsitollvm.cpp \ - storage.cpp \ - storagesoa.cpp \ - instructionssoa.cpp - -INC_SOURCES = gallivm_builtins.cpp - -CPP_SOURCES = \ - $(GALLIVM_SOURCES) - -C_SOURCES = -ASM_SOURCES = - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(CPP_SOURCES:.cpp=.o) \ - $(ASM_SOURCES:.S=.o) - -### Include directories -INCLUDES = \ - -I. \ - -I$(TOP)/src/gallium/drivers - -I$(TOP)/src/gallium/aux \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/mesa \ - -I$(TOP)/include - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.cpp.o: - $(CXX) -c $(INCLUDES) $(LLVM_CXXFLAGS) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -##### TARGETS ##### - -default:: depend symlinks $(LIBNAME) - - -$(LIBNAME): $(OBJECTS) Makefile - $(TOP)/bin/mklib -o $@ -static $(OBJECTS) - - -depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ - $(ASM_SOURCES) $(INC_SOURCES) 2> /dev/null - - -gallivm_builtins.cpp: llvm_builtins.c - clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean: - -rm -f *.o */*.o *~ *.so *~ server/*.o - -rm -f depend depend.bak - -rm -f gallivm_builtins.cpp - -symlinks: - - -include depend diff --git a/src/gallium/aux/llvm/gallivm.cpp b/src/gallium/aux/llvm/gallivm.cpp deleted file mode 100644 index d14bb3b99a..0000000000 --- a/src/gallium/aux/llvm/gallivm.cpp +++ /dev/null @@ -1,327 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "gallivm.h" -#include "gallivm_p.h" - -#include "instructions.h" -#include "loweringpass.h" -#include "storage.h" -#include "tgsitollvm.h" - -#include "pipe/p_context.h" -#include "pipe/p_shader_tokens.h" - -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_dump.h" - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -static int GLOBAL_ID = 0; - -using namespace llvm; - -static inline -void AddStandardCompilePasses(PassManager &PM) -{ - PM.add(new LoweringPass()); - PM.add(createVerifierPass()); // Verify that input is correct - - PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp - - //PM.add(createStripSymbolsPass(true)); - - PM.add(createRaiseAllocationsPass()); // call %malloc -> malloc inst - PM.add(createCFGSimplificationPass()); // Clean up disgusting code - PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas - PM.add(createGlobalOptimizerPass()); // Optimize out global vars - PM.add(createGlobalDCEPass()); // Remove unused fns and globs - PM.add(createIPConstantPropagationPass());// IP Constant Propagation - PM.add(createDeadArgEliminationPass()); // Dead argument elimination - PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE - PM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE - - PM.add(createPruneEHPass()); // Remove dead EH info - - PM.add(createFunctionInliningPass()); // Inline small functions - PM.add(createArgumentPromotionPass()); // Scalarize uninlined fn args - - PM.add(createTailDuplicationPass()); // Simplify cfg by copying code - PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl. - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas - PM.add(createInstructionCombiningPass()); // Combine silly seq's - PM.add(createCondPropagationPass()); // Propagate conditionals - - PM.add(createTailCallEliminationPass()); // Eliminate tail calls - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createReassociatePass()); // Reassociate expressions - PM.add(createLoopRotatePass()); - PM.add(createLICMPass()); // Hoist loop invariants - PM.add(createLoopUnswitchPass()); // Unswitch loops. - PM.add(createLoopIndexSplitPass()); // Index split loops. - PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc - PM.add(createIndVarSimplifyPass()); // Canonicalize indvars - PM.add(createLoopUnrollPass()); // Unroll small loops - PM.add(createInstructionCombiningPass()); // Clean up after the unroller - PM.add(createGVNPass()); // Remove redundancies - PM.add(createSCCPPass()); // Constant prop with SCCP - - // Run instcombine after redundancy elimination to exploit opportunities - // opened up by them. - PM.add(createInstructionCombiningPass()); - PM.add(createCondPropagationPass()); // Propagate conditionals - - PM.add(createDeadStoreEliminationPass()); // Delete dead stores - PM.add(createAggressiveDCEPass()); // SSA based 'Aggressive DCE' - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createSimplifyLibCallsPass()); // Library Call Optimizations - PM.add(createDeadTypeEliminationPass()); // Eliminate dead types - PM.add(createConstantMergePass()); // Merge dup global constants -} - -void gallivm_prog_delete(struct gallivm_prog *prog) -{ - delete prog->module; - prog->module = 0; - prog->function = 0; - free(prog); -} - -static inline void -constant_interpolation(float (*inputs)[16][4], - const struct tgsi_interp_coef *coefs, - unsigned attrib, - unsigned chan) -{ - unsigned i; - - for (i = 0; i < QUAD_SIZE; ++i) { - inputs[i][attrib][chan] = coefs[attrib].a0[chan]; - } -} - -static inline void -linear_interpolation(float (*inputs)[16][4], - const struct tgsi_interp_coef *coefs, - unsigned attrib, - unsigned chan) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = inputs[i][0][0]; - const float y = inputs[i][0][1]; - - inputs[i][attrib][chan] = - coefs[attrib].a0[chan] + - coefs[attrib].dadx[chan] * x + - coefs[attrib].dady[chan] * y; - } -} - -static inline void -perspective_interpolation(float (*inputs)[16][4], - const struct tgsi_interp_coef *coefs, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = inputs[i][0][0]; - const float y = inputs[i][0][1]; - /* WPOS.w here is really 1/w */ - const float w = 1.0f / inputs[i][0][3]; - assert(inputs[i][0][3] != 0.0); - - inputs[i][attrib][chan] = - (coefs[attrib].a0[chan] + - coefs[attrib].dadx[chan] * x + - coefs[attrib].dady[chan] * y) * w; - } -} - -void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix) -{ - if (!ir || !ir->module) - return; - - if (file_prefix) { - std::ostringstream stream; - stream << file_prefix; - stream << ir->id; - stream << ".ll"; - std::string name = stream.str(); - std::ofstream out(name.c_str()); - if (!out) { - std::cerr<<"Can't open file : "<module); - out.close(); - } else { - const llvm::Module::FunctionListType &funcs = ir->module->getFunctionList(); - llvm::Module::FunctionListType::const_iterator itr; - std::cout<<"; ---------- Start shader "<id<id<num_interp; ++i) { - const gallivm_interpolate &interp = prog->interpolators[i]; - switch (interp.type) { - case TGSI_INTERPOLATE_CONSTANT: - constant_interpolation(inputs, coef, interp.attrib, interp.chan); - break; - - case TGSI_INTERPOLATE_LINEAR: - linear_interpolation(inputs, coef, interp.attrib, interp.chan); - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - perspective_interpolation(inputs, coef, interp.attrib, interp.chan); - break; - - default: - assert( 0 ); - } - } -} - - -struct gallivm_ir * gallivm_ir_new(enum gallivm_shader_type type) -{ - struct gallivm_ir *ir = - (struct gallivm_ir *)calloc(1, sizeof(struct gallivm_ir)); - ++GLOBAL_ID; - ir->id = GLOBAL_ID; - ir->type = type; - - return ir; -} - -void gallivm_ir_set_layout(struct gallivm_ir *ir, - enum gallivm_vector_layout layout) -{ - ir->layout = layout; -} - -void gallivm_ir_set_components(struct gallivm_ir *ir, int num) -{ - ir->num_components = num; -} - -void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, - const struct tgsi_token *tokens) -{ - std::cout << "Creating llvm from: " <module = mod; - gallivm_ir_dump(ir, 0); -} - -void gallivm_ir_delete(struct gallivm_ir *ir) -{ - delete ir->module; - free(ir); -} - -struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) -{ - struct gallivm_prog *prog = - (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); - llvm::Module *mod = llvm::CloneModule(ir->module); - prog->num_consts = ir->num_consts; - memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators)); - prog->num_interp = ir->num_interp; - - /* Run optimization passes over it */ - PassManager passes; - passes.add(new TargetData(mod)); - AddStandardCompilePasses(passes); - passes.run(*mod); - prog->module = mod; - - std::cout << "After optimizations:"<dump(); - - return prog; -} - -#endif /* MESA_LLVM */ diff --git a/src/gallium/aux/llvm/gallivm.h b/src/gallium/aux/llvm/gallivm.h deleted file mode 100644 index 92da4bca7f..0000000000 --- a/src/gallium/aux/llvm/gallivm.h +++ /dev/null @@ -1,103 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef GALLIVM_H -#define GALLIVM_H - -#if defined __cplusplus -extern "C" { -#endif - -#include "pipe/p_state.h" - -#ifdef MESA_LLVM - -struct tgsi_token; - -struct gallivm_ir; -struct gallivm_prog; -struct gallivm_cpu_engine; -struct tgsi_interp_coef; -struct tgsi_sampler; -struct tgsi_exec_vector; - -enum gallivm_shader_type { - GALLIVM_VS, - GALLIVM_FS -}; - -enum gallivm_vector_layout { - GALLIVM_AOS, - GALLIVM_SOA -}; - -struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type); -void gallivm_ir_set_layout(struct gallivm_ir *ir, - enum gallivm_vector_layout layout); -void gallivm_ir_set_components(struct gallivm_ir *ir, int num); -void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, - const struct tgsi_token *tokens); -void gallivm_ir_delete(struct gallivm_ir *ir); - - -struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir); - -void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - const struct tgsi_interp_coef *coefs); -void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); - - -struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); -struct gallivm_cpu_engine *gallivm_global_cpu_engine(); -int gallivm_cpu_vs_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - struct tgsi_exec_vector *temps); -int gallivm_cpu_fs_exec(struct gallivm_prog *prog, - float x, float y, - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*consts)[4], - struct tgsi_sampler *samplers); -void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); -void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); - - -#endif /* MESA_LLVM */ - -#if defined __cplusplus -} // extern "C" -#endif - -#endif diff --git a/src/gallium/aux/llvm/gallivm_builtins.cpp b/src/gallium/aux/llvm/gallivm_builtins.cpp deleted file mode 100644 index 1796f0a177..0000000000 --- a/src/gallium/aux/llvm/gallivm_builtins.cpp +++ /dev/null @@ -1,567 +0,0 @@ -// Generated by llvm2cpp - DO NOT MODIFY! - - -Module* createGallivmBuiltins(Module *mod) { - -mod->setModuleIdentifier("shader"); - -// Type Definitions -ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 25); - -PointerType* PointerTy_1 = PointerType::get(ArrayTy_0, 0); - -std::vectorFuncTy_2_args; -FuncTy_2_args.push_back(Type::FloatTy); -FuncTy_2_args.push_back(Type::FloatTy); -FunctionType* FuncTy_2 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_2_args, - /*isVarArg=*/false); - -PointerType* PointerTy_3 = PointerType::get(FuncTy_2, 0); - -VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); - -std::vectorFuncTy_5_args; -FuncTy_5_args.push_back(VectorTy_4); -FunctionType* FuncTy_5 = FunctionType::get( - /*Result=*/VectorTy_4, - /*Params=*/FuncTy_5_args, - /*isVarArg=*/false); - -std::vectorFuncTy_6_args; -FuncTy_6_args.push_back(VectorTy_4); -FuncTy_6_args.push_back(VectorTy_4); -FuncTy_6_args.push_back(VectorTy_4); -FunctionType* FuncTy_6 = FunctionType::get( - /*Result=*/VectorTy_4, - /*Params=*/FuncTy_6_args, - /*isVarArg=*/false); - -VectorType* VectorTy_7 = VectorType::get(IntegerType::get(32), 4); - -std::vectorFuncTy_9_args; -FunctionType* FuncTy_9 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_9_args, - /*isVarArg=*/true); - -PointerType* PointerTy_8 = PointerType::get(FuncTy_9, 0); - -PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8), 0); - -std::vectorFuncTy_12_args; -FuncTy_12_args.push_back(Type::FloatTy); -FunctionType* FuncTy_12 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_12_args, - /*isVarArg=*/false); - -PointerType* PointerTy_11 = PointerType::get(FuncTy_12, 0); - -std::vectorFuncTy_13_args; -FuncTy_13_args.push_back(VectorTy_4); -FunctionType* FuncTy_13 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_13_args, - /*isVarArg=*/false); - - -// Function Declarations - -Function* func_approx = new Function( - /*Type=*/FuncTy_2, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"approx", mod); -func_approx->setCallingConv(CallingConv::C); -const ParamAttrsList *func_approx_PAL = 0; -func_approx->setParamAttrs(func_approx_PAL); - -Function* func_powf = new Function( - /*Type=*/FuncTy_2, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"powf", mod); // (external, no body) -func_powf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_powf_PAL = 0; -func_powf->setParamAttrs(func_powf_PAL); - -Function* func_lit = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"lit", mod); -func_lit->setCallingConv(CallingConv::C); -const ParamAttrsList *func_lit_PAL = 0; -func_lit->setParamAttrs(func_lit_PAL); - -Function* func_cmp = new Function( - /*Type=*/FuncTy_6, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"cmp", mod); -func_cmp->setCallingConv(CallingConv::C); -const ParamAttrsList *func_cmp_PAL = 0; -{ - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_cmp_PAL = ParamAttrsList::get(Attrs); - -} -func_cmp->setParamAttrs(func_cmp_PAL); - -Function* func_vcos = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"vcos", mod); -func_vcos->setCallingConv(CallingConv::C); -const ParamAttrsList *func_vcos_PAL = 0; -func_vcos->setParamAttrs(func_vcos_PAL); - -Function* func_printf = new Function( - /*Type=*/FuncTy_9, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", mod); // (external, no body) -func_printf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_printf_PAL = 0; -func_printf->setParamAttrs(func_printf_PAL); - -Function* func_cosf = new Function( - /*Type=*/FuncTy_12, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"cosf", mod); // (external, no body) -func_cosf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_cosf_PAL = 0; -func_cosf->setParamAttrs(func_cosf_PAL); - -Function* func_scs = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"scs", mod); -func_scs->setCallingConv(CallingConv::C); -const ParamAttrsList *func_scs_PAL = 0; -func_scs->setParamAttrs(func_scs_PAL); - -Function* func_sinf = new Function( - /*Type=*/FuncTy_12, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"sinf", mod); // (external, no body) -func_sinf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_sinf_PAL = 0; -func_sinf->setParamAttrs(func_sinf_PAL); - -Function* func_vsin = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"vsin", mod); -func_vsin->setCallingConv(CallingConv::C); -const ParamAttrsList *func_vsin_PAL = 0; -func_vsin->setParamAttrs(func_vsin_PAL); - -Function* func_kilp = new Function( - /*Type=*/FuncTy_13, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"kilp", mod); -func_kilp->setCallingConv(CallingConv::C); -const ParamAttrsList *func_kilp_PAL = 0; -{ - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_kilp_PAL = ParamAttrsList::get(Attrs); - -} -func_kilp->setParamAttrs(func_kilp_PAL); - -// Global Variable Declarations - - -GlobalVariable* gvar_array__str = new GlobalVariable( -/*Type=*/ArrayTy_0, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str", -mod); - -GlobalVariable* gvar_array__str1 = new GlobalVariable( -/*Type=*/ArrayTy_0, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str1", -mod); - -// Constant Definitions -Constant* const_array_14 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); -Constant* const_array_15 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); -ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); -ConstantFP* const_float_17 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); -Constant* const_float_18 = Constant::getNullValue(Type::FloatTy); -Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); -std::vector const_packed_20_elems; -ConstantFP* const_float_21 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); -const_packed_20_elems.push_back(const_float_21); -UndefValue* const_float_22 = UndefValue::get(Type::FloatTy); -const_packed_20_elems.push_back(const_float_22); -const_packed_20_elems.push_back(const_float_22); -const_packed_20_elems.push_back(const_float_21); -Constant* const_packed_20 = ConstantVector::get(VectorTy_4, const_packed_20_elems); -ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); -ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "3", 10)); -ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "2", 10)); -std::vector const_packed_26_elems; -const_packed_26_elems.push_back(const_float_21); -const_packed_26_elems.push_back(const_float_18); -const_packed_26_elems.push_back(const_float_18); -const_packed_26_elems.push_back(const_float_21); -Constant* const_packed_26 = ConstantVector::get(VectorTy_4, const_packed_26_elems); -Constant* const_double_27 = Constant::getNullValue(Type::DoubleTy); -std::vector const_packed_28_elems; -const_packed_28_elems.push_back(const_int32_19); -ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "5", 10)); -const_packed_28_elems.push_back(const_int32_29); -const_packed_28_elems.push_back(const_int32_25); -const_packed_28_elems.push_back(const_int32_24); -Constant* const_packed_28 = ConstantVector::get(VectorTy_7, const_packed_28_elems); -std::vector const_packed_30_elems; -const_packed_30_elems.push_back(const_int32_19); -const_packed_30_elems.push_back(const_int32_23); -ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "6", 10)); -const_packed_30_elems.push_back(const_int32_31); -const_packed_30_elems.push_back(const_int32_24); -Constant* const_packed_30 = ConstantVector::get(VectorTy_7, const_packed_30_elems); -std::vector const_packed_32_elems; -const_packed_32_elems.push_back(const_int32_19); -const_packed_32_elems.push_back(const_int32_23); -const_packed_32_elems.push_back(const_int32_25); -ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "7", 10)); -const_packed_32_elems.push_back(const_int32_33); -Constant* const_packed_32 = ConstantVector::get(VectorTy_7, const_packed_32_elems); -std::vector const_ptr_34_indices; -const_ptr_34_indices.push_back(const_int32_19); -const_ptr_34_indices.push_back(const_int32_19); -Constant* const_ptr_34 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_34_indices[0], const_ptr_34_indices.size() ); -UndefValue* const_packed_35 = UndefValue::get(VectorTy_4); -std::vector const_ptr_36_indices; -const_ptr_36_indices.push_back(const_int32_19); -const_ptr_36_indices.push_back(const_int32_19); -Constant* const_ptr_36 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_36_indices[0], const_ptr_36_indices.size() ); - -// Global Variable Definitions -gvar_array__str->setInitializer(const_array_14); -gvar_array__str1->setInitializer(const_array_15); - -// Function Definitions - -// Function: approx (func_approx) -{ - Function::arg_iterator args = func_approx->arg_begin(); - Value* float_a = args++; - float_a->setName("a"); - Value* float_b = args++; - float_b->setName("b"); - - BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); - - // Block entry (label_entry) - FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_16, "cmp", label_entry); - SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_16, float_b, "b.addr.0", label_entry); - FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_17, "cmp3", label_entry); - SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_17, float_b_addr_0, "b.addr.1", label_entry); - FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_18, "cmp7", label_entry); - SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_18, float_a, "a.addr.0", label_entry); - std::vector float_call_params; - float_call_params.push_back(float_a_addr_0); - float_call_params.push_back(float_b_addr_1); - CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); - float_call->setCallingConv(CallingConv::C); - float_call->setTailCall(true);const ParamAttrsList *float_call_PAL = 0; - float_call->setParamAttrs(float_call_PAL); - - new ReturnInst(float_call, label_entry); - -} - -// Function: lit (func_lit) -{ - Function::arg_iterator args = func_lit->arg_begin(); - Value* packed_tmp = args++; - packed_tmp->setName("tmp"); - - BasicBlock* label_entry_38 = new BasicBlock("entry",func_lit,0); - BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); - BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); - - // Block entry (label_entry_38) - ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp6", label_entry_38); - FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp6, const_float_18, "cmp", label_entry_38); - new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_39, label_entry_38); - - // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp10 = new InsertElementInst(const_packed_20, float_tmp6, const_int32_23, "tmp10", label_ifthen); - ExtractElementInst* float_tmp12 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp14", label_ifthen); - std::vector float_call_41_params; - float_call_41_params.push_back(float_tmp12); - float_call_41_params.push_back(float_tmp14); - CallInst* float_call_41 = new CallInst(func_approx, float_call_41_params.begin(), float_call_41_params.end(), "call", label_ifthen); - float_call_41->setCallingConv(CallingConv::C); - float_call_41->setTailCall(true);const ParamAttrsList *float_call_41_PAL = 0; - float_call_41->setParamAttrs(float_call_41_PAL); - - InsertElementInst* packed_tmp16 = new InsertElementInst(packed_tmp10, float_call_41, const_int32_25, "tmp16", label_ifthen); - new ReturnInst(packed_tmp16, label_ifthen); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock) - new ReturnInst(const_packed_26, label_UnifiedReturnBlock); - -} - -// Function: cmp (func_cmp) -{ - Function::arg_iterator args = func_cmp->arg_begin(); - Value* packed_tmp0 = args++; - packed_tmp0->setName("tmp0"); - Value* packed_tmp1 = args++; - packed_tmp1->setName("tmp1"); - Value* packed_tmp2 = args++; - packed_tmp2->setName("tmp2"); - - BasicBlock* label_entry_44 = new BasicBlock("entry",func_cmp,0); - BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); - BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); - BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); - BasicBlock* label_cond_cont34 = new BasicBlock("cond.cont34",func_cmp,0); - BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); - BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); - - // Block entry (label_entry_44) - ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_44); - CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_44); - FCmpInst* int1_cmp_45 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_44); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_44); - CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_44); - FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_44); - SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_45, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_44); - new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_44); - - // Block cond.?14 (label_cond__14) - ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_28, "tmp233", label_cond__14); - ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp254", label_cond__14); - CastInst* double_conv265 = new FPExtInst(float_tmp254, Type::DoubleTy, "conv265", label_cond__14); - FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_27, "cmp276", label_cond__14); - new BranchInst(label_cond__28, label_cond_cont34, int1_cmp276, label_cond__14); - - // Block cond.cont20 (label_cond_cont20) - ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_28, "tmp23", label_cond_cont20); - ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp25", label_cond_cont20); - CastInst* double_conv26 = new FPExtInst(float_tmp25, Type::DoubleTy, "conv26", label_cond_cont20); - FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_27, "cmp27", label_cond_cont20); - new BranchInst(label_cond__28, label_cond_cont34, int1_cmp27, label_cond_cont20); - - // Block cond.?28 (label_cond__28) - PHINode* packed_tmp23_reg2mem_0 = new PHINode(VectorTy_4, "tmp23.reg2mem.0", label_cond__28); - packed_tmp23_reg2mem_0->reserveOperandSpace(2); - packed_tmp23_reg2mem_0->addIncoming(packed_tmp233, label_cond__14); - packed_tmp23_reg2mem_0->addIncoming(packed_tmp23, label_cond_cont20); - - ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_30, "tmp378", label_cond__28); - ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp399", label_cond__28); - CastInst* double_conv4010 = new FPExtInst(float_tmp399, Type::DoubleTy, "conv4010", label_cond__28); - FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_27, "cmp4111", label_cond__28); - new BranchInst(label_cond__42, label_cond_cont48, int1_cmp4111, label_cond__28); - - // Block cond.cont34 (label_cond_cont34) - PHINode* packed_tmp23_reg2mem_1 = new PHINode(VectorTy_4, "tmp23.reg2mem.1", label_cond_cont34); - packed_tmp23_reg2mem_1->reserveOperandSpace(2); - packed_tmp23_reg2mem_1->addIncoming(packed_tmp233, label_cond__14); - packed_tmp23_reg2mem_1->addIncoming(packed_tmp23, label_cond_cont20); - - ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_30, "tmp37", label_cond_cont34); - ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp39", label_cond_cont34); - CastInst* double_conv40 = new FPExtInst(float_tmp39, Type::DoubleTy, "conv40", label_cond_cont34); - FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_27, "cmp41", label_cond_cont34); - new BranchInst(label_cond__42, label_cond_cont48, int1_cmp41, label_cond_cont34); - - // Block cond.?42 (label_cond__42) - PHINode* packed_tmp37_reg2mem_0 = new PHINode(VectorTy_4, "tmp37.reg2mem.0", label_cond__42); - packed_tmp37_reg2mem_0->reserveOperandSpace(2); - packed_tmp37_reg2mem_0->addIncoming(packed_tmp378, label_cond__28); - packed_tmp37_reg2mem_0->addIncoming(packed_tmp37, label_cond_cont34); - - ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_32, "tmp5113", label_cond__42); - new ReturnInst(packed_tmp5113, label_cond__42); - - // Block cond.cont48 (label_cond_cont48) - PHINode* packed_tmp37_reg2mem_1 = new PHINode(VectorTy_4, "tmp37.reg2mem.1", label_cond_cont48); - packed_tmp37_reg2mem_1->reserveOperandSpace(2); - packed_tmp37_reg2mem_1->addIncoming(packed_tmp378, label_cond__28); - packed_tmp37_reg2mem_1->addIncoming(packed_tmp37, label_cond_cont34); - - ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_32, "tmp51", label_cond_cont48); - new ReturnInst(packed_tmp51, label_cond_cont48); - -} - -// Function: vcos (func_vcos) -{ - Function::arg_iterator args = func_vcos->arg_begin(); - Value* packed_val = args++; - packed_val->setName("val"); - - BasicBlock* label_entry_53 = new BasicBlock("entry",func_vcos,0); - - // Block entry (label_entry_53) - ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_53); - CastInst* double_conv_54 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_53); - ExtractElementInst* float_tmp3_55 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_53); - CastInst* double_conv4 = new FPExtInst(float_tmp3_55, Type::DoubleTy, "conv4", label_entry_53); - ExtractElementInst* float_tmp6_56 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_53); - CastInst* double_conv7 = new FPExtInst(float_tmp6_56, Type::DoubleTy, "conv7", label_entry_53); - ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_53); - CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_53); - std::vector int32_call_params; - int32_call_params.push_back(const_ptr_34); - int32_call_params.push_back(double_conv_54); - int32_call_params.push_back(double_conv4); - int32_call_params.push_back(double_conv7); - int32_call_params.push_back(double_conv10); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_53); - int32_call->setCallingConv(CallingConv::C); - int32_call->setTailCall(true);const ParamAttrsList *int32_call_PAL = 0; - int32_call->setParamAttrs(int32_call_PAL); - - CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_53); - float_call13->setCallingConv(CallingConv::C); - float_call13->setTailCall(true);const ParamAttrsList *float_call13_PAL = 0; - float_call13->setParamAttrs(float_call13_PAL); - - InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_53); - CallInst* float_call18 = new CallInst(func_cosf, float_tmp1, "call18", label_entry_53); - float_call18->setCallingConv(CallingConv::C); - float_call18->setTailCall(true);const ParamAttrsList *float_call18_PAL = 0; - float_call18->setParamAttrs(float_call18_PAL); - - InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call18, const_int32_23, "tmp20", label_entry_53); - CallInst* float_call23 = new CallInst(func_cosf, float_tmp1, "call23", label_entry_53); - float_call23->setCallingConv(CallingConv::C); - float_call23->setTailCall(true);const ParamAttrsList *float_call23_PAL = 0; - float_call23->setParamAttrs(float_call23_PAL); - - InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call23, const_int32_25, "tmp25", label_entry_53); - CallInst* float_call28 = new CallInst(func_cosf, float_tmp1, "call28", label_entry_53); - float_call28->setCallingConv(CallingConv::C); - float_call28->setTailCall(true);const ParamAttrsList *float_call28_PAL = 0; - float_call28->setParamAttrs(float_call28_PAL); - - InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call28, const_int32_24, "tmp30", label_entry_53); - CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_53); - CastInst* double_conv36 = new FPExtInst(float_call18, Type::DoubleTy, "conv36", label_entry_53); - CastInst* double_conv39 = new FPExtInst(float_call23, Type::DoubleTy, "conv39", label_entry_53); - CastInst* double_conv42 = new FPExtInst(float_call28, Type::DoubleTy, "conv42", label_entry_53); - std::vector int32_call43_params; - int32_call43_params.push_back(const_ptr_36); - int32_call43_params.push_back(double_conv33); - int32_call43_params.push_back(double_conv36); - int32_call43_params.push_back(double_conv39); - int32_call43_params.push_back(double_conv42); - CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_53); - int32_call43->setCallingConv(CallingConv::C); - int32_call43->setTailCall(true);const ParamAttrsList *int32_call43_PAL = 0; - int32_call43->setParamAttrs(int32_call43_PAL); - - new ReturnInst(packed_tmp30, label_entry_53); - -} - -// Function: scs (func_scs) -{ - Function::arg_iterator args = func_scs->arg_begin(); - Value* packed_val_58 = args++; - packed_val_58->setName("val"); - - BasicBlock* label_entry_59 = new BasicBlock("entry",func_scs,0); - - // Block entry (label_entry_59) - ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_58, const_int32_19, "tmp2", label_entry_59); - CallInst* float_call_60 = new CallInst(func_cosf, float_tmp2, "call", label_entry_59); - float_call_60->setCallingConv(CallingConv::C); - float_call_60->setTailCall(true);const ParamAttrsList *float_call_60_PAL = 0; - float_call_60->setParamAttrs(float_call_60_PAL); - - InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_60, const_int32_19, "tmp5", label_entry_59); - CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_59); - float_call7->setCallingConv(CallingConv::C); - float_call7->setTailCall(true);const ParamAttrsList *float_call7_PAL = 0; - float_call7->setParamAttrs(float_call7_PAL); - - InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_59); - new ReturnInst(packed_tmp9, label_entry_59); - -} - -// Function: vsin (func_vsin) -{ - Function::arg_iterator args = func_vsin->arg_begin(); - Value* packed_val_62 = args++; - packed_val_62->setName("val"); - - BasicBlock* label_entry_63 = new BasicBlock("entry",func_vsin,0); - - // Block entry (label_entry_63) - ExtractElementInst* float_tmp2_64 = new ExtractElementInst(packed_val_62, const_int32_19, "tmp2", label_entry_63); - CallInst* float_call_65 = new CallInst(func_sinf, float_tmp2_64, "call", label_entry_63); - float_call_65->setCallingConv(CallingConv::C); - float_call_65->setTailCall(true);const ParamAttrsList *float_call_65_PAL = 0; - float_call_65->setParamAttrs(float_call_65_PAL); - - InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_65, const_int32_19, "tmp6", label_entry_63); - InsertElementInst* packed_tmp9_66 = new InsertElementInst(packed_tmp6, float_call_65, const_int32_23, "tmp9", label_entry_63); - InsertElementInst* packed_tmp12 = new InsertElementInst(packed_tmp9_66, float_call_65, const_int32_25, "tmp12", label_entry_63); - InsertElementInst* packed_tmp15_67 = new InsertElementInst(packed_tmp12, float_call_65, const_int32_24, "tmp15", label_entry_63); - new ReturnInst(packed_tmp15_67, label_entry_63); - -} - -// Function: kilp (func_kilp) -{ - Function::arg_iterator args = func_kilp->arg_begin(); - Value* packed_val_69 = args++; - packed_val_69->setName("val"); - - BasicBlock* label_entry_70 = new BasicBlock("entry",func_kilp,0); - BasicBlock* label_lor_rhs = new BasicBlock("lor_rhs",func_kilp,0); - BasicBlock* label_lor_rhs5 = new BasicBlock("lor_rhs5",func_kilp,0); - BasicBlock* label_lor_rhs11 = new BasicBlock("lor_rhs11",func_kilp,0); - BasicBlock* label_UnifiedReturnBlock_71 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); - - // Block entry (label_entry_70) - ExtractElementInst* float_tmp1_72 = new ExtractElementInst(packed_val_69, const_int32_19, "tmp1", label_entry_70); - FCmpInst* int1_cmp_73 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_72, const_float_18, "cmp", label_entry_70); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs, int1_cmp_73, label_entry_70); - - // Block lor_rhs (label_lor_rhs) - ExtractElementInst* float_tmp3_75 = new ExtractElementInst(packed_val_69, const_int32_23, "tmp3", label_lor_rhs); - FCmpInst* int1_cmp4 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_75, const_float_18, "cmp4", label_lor_rhs); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs5, int1_cmp4, label_lor_rhs); - - // Block lor_rhs5 (label_lor_rhs5) - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_val_69, const_int32_25, "tmp7", label_lor_rhs5); - FCmpInst* int1_cmp8 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp7, const_float_18, "cmp8", label_lor_rhs5); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs11, int1_cmp8, label_lor_rhs5); - - // Block lor_rhs11 (label_lor_rhs11) - ExtractElementInst* float_tmp13 = new ExtractElementInst(packed_val_69, const_int32_24, "tmp13", label_lor_rhs11); - FCmpInst* int1_cmp14 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp13, const_float_18, "cmp14", label_lor_rhs11); - CastInst* int32_retval = new ZExtInst(int1_cmp14, IntegerType::get(32), "retval", label_lor_rhs11); - new ReturnInst(int32_retval, label_lor_rhs11); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock_71) - new ReturnInst(const_int32_23, label_UnifiedReturnBlock_71); - -} - -return mod; - -} diff --git a/src/gallium/aux/llvm/gallivm_cpu.cpp b/src/gallium/aux/llvm/gallivm_cpu.cpp deleted file mode 100644 index 8f9830d0b1..0000000000 --- a/src/gallium/aux/llvm/gallivm_cpu.cpp +++ /dev/null @@ -1,202 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "gallivm.h" -#include "gallivm_p.h" - -#include "instructions.h" -#include "loweringpass.h" -#include "storage.h" -#include "tgsitollvm.h" - -#include "pipe/p_context.h" -#include "pipe/p_shader_tokens.h" - -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_dump.h" - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -struct gallivm_cpu_engine { - llvm::ExecutionEngine *engine; -}; - -static struct gallivm_cpu_engine *CPU = 0; - -typedef int (*fragment_shader_runner)(float x, float y, - float (*dests)[16][4], - float (*inputs)[16][4], - int num_attribs, - float (*consts)[4], int num_consts, - struct tgsi_sampler *samplers); - -int gallivm_cpu_fs_exec(struct gallivm_prog *prog, - float fx, float fy, - float (*dests)[16][4], - float (*inputs)[16][4], - float (*consts)[4], - struct tgsi_sampler *samplers) -{ - fragment_shader_runner runner = reinterpret_cast(prog->function); - assert(runner); - - return runner(fx, fy, dests, inputs, prog->num_interp, - consts, prog->num_consts, - samplers); -} - -static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) -{ - llvm::Module *mod = prog->module; - llvm::Function *func = 0; - - switch (prog->type) { - case GALLIVM_VS: - func = mod->getFunction("vs_shader"); - break; - case GALLIVM_FS: - func = mod->getFunction("fs_shader"); - break; - default: - assert(!"Unknown shader type!"); - break; - } - return func; -} - -/*! - This function creates a CPU based execution engine for the given gallivm_prog. - gallivm_cpu_engine should be used as a singleton throughout the library. Before - executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile. - The gallivm_prog instance which is being passed to the constructor is being - automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile - with it again. - */ -struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) -{ - struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *) - calloc(1, sizeof(struct gallivm_cpu_engine)); - llvm::Module *mod = static_cast(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); - ee->DisableLazyCompilation(); - cpu->engine = ee; - - llvm::Function *func = func_for_shader(prog); - - prog->function = ee->getPointerToFunction(func); - CPU = cpu; - return cpu; -} - - -/*! - This function JIT compiles the given gallivm_prog with the given cpu based execution engine. - The reference to the generated machine code entry point will be stored - in the gallivm_prog program. After executing this function one can call gallivm_prog_exec - in order to execute the gallivm_prog on the CPU. - */ -void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) -{ - llvm::Module *mod = static_cast(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = cpu->engine; - assert(ee); - /*FIXME : remove */ - ee->DisableLazyCompilation(); - ee->addModuleProvider(mp); - - llvm::Function *func = func_for_shader(prog); - prog->function = ee->getPointerToFunction(func); -} - -void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) -{ - free(cpu); -} - -struct gallivm_cpu_engine * gallivm_global_cpu_engine() -{ - return CPU; -} - - -typedef void (*vertex_shader_runner)(void *ainputs, - void *dests, - float (*aconsts)[4], - void *temps); - - -/*! - This function is used to execute the gallivm_prog in software. Before calling - this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile - function. - */ -int gallivm_cpu_vs_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - struct tgsi_exec_vector *temps) -{ - vertex_shader_runner runner = reinterpret_cast(prog->function); - assert(runner); - /*FIXME*/ - runner(inputs, dests, consts, temps); - - return 0; -} - -#endif diff --git a/src/gallium/aux/llvm/gallivm_p.h b/src/gallium/aux/llvm/gallivm_p.h deleted file mode 100644 index cfe7b1901b..0000000000 --- a/src/gallium/aux/llvm/gallivm_p.h +++ /dev/null @@ -1,110 +0,0 @@ -#ifndef GALLIVM_P_H -#define GALLIVM_P_H - -#ifdef MESA_LLVM - -#include "gallivm.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/p_compiler.h" - -namespace llvm { - class Module; -} - -#if defined __cplusplus -extern "C" { -#endif - -enum gallivm_shader_type; -enum gallivm_vector_layout; - -struct gallivm_interpolate { - int attrib; - int chan; - int type; -}; - -struct gallivm_ir { - llvm::Module *module; - int id; - enum gallivm_shader_type type; - enum gallivm_vector_layout layout; - int num_components; - int num_consts; - - //FIXME: this might not be enough for some shaders - struct gallivm_interpolate interpolators[32*4]; - int num_interp; -}; - -struct gallivm_prog { - llvm::Module *module; - void *function; - - int id; - enum gallivm_shader_type type; - - int num_consts; - - //FIXME: this might not be enough for some shaders - struct gallivm_interpolate interpolators[32*4]; - int num_interp; -}; - -static INLINE void gallivm_swizzle_components(int swizzle, - int *xc, int *yc, - int *zc, int *wc) -{ - int x = swizzle / 1000; swizzle -= x * 1000; - int y = swizzle / 100; swizzle -= y * 100; - int z = swizzle / 10; swizzle -= z * 10; - int w = swizzle; - - if (xc) *xc = x; - if (yc) *yc = y; - if (zc) *zc = z; - if (wc) *wc = w; -} - -static INLINE boolean gallivm_is_swizzle(int swizzle) -{ - const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + - TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; - return swizzle != NO_SWIZZLE; -} - -static INLINE int gallivm_x_swizzle(int swizzle) -{ - int x; - gallivm_swizzle_components(swizzle, &x, 0, 0, 0); - return x; -} - -static INLINE int gallivm_y_swizzle(int swizzle) -{ - int y; - gallivm_swizzle_components(swizzle, 0, &y, 0, 0); - return y; -} - -static INLINE int gallivm_z_swizzle(int swizzle) -{ - int z; - gallivm_swizzle_components(swizzle, 0, 0, &z, 0); - return z; -} - -static INLINE int gallivm_w_swizzle(int swizzle) -{ - int w; - gallivm_swizzle_components(swizzle, 0, 0, 0, &w); - return w; -} - -#endif /* MESA_LLVM */ - -#if defined __cplusplus -} // extern "C" -#endif - -#endif diff --git a/src/gallium/aux/llvm/instructions.cpp b/src/gallium/aux/llvm/instructions.cpp deleted file mode 100644 index 55d39fa5f1..0000000000 --- a/src/gallium/aux/llvm/instructions.cpp +++ /dev/null @@ -1,889 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "instructions.h" - -#include "storage.h" - -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -using namespace llvm; - -#include "gallivm_builtins.cpp" - -static inline std::string createFuncName(int label) -{ - std::ostringstream stream; - stream << "function"; - stream << label; - return stream.str(); -} - -Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, - Storage *storage) - : m_mod(mod), m_func(func), m_builder(block), m_idx(0), - m_storage(storage) -{ - m_floatVecType = VectorType::get(Type::FloatTy, 4); - - m_llvmFSqrt = 0; - m_llvmFAbs = 0; - m_llvmPow = 0; - m_llvmFloor = 0; - m_llvmFlog = 0; - m_llvmLit = 0; - m_fmtPtr = 0; - - createGallivmBuiltins(m_mod); -} - -llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) -{ - return m_builder.CreateAdd(in1, in2, name("add")); -} - -llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in3) -{ - Value *mulRes = mul(in1, in2); - return add(mulRes, in3); -} - -llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2) -{ - return m_builder.CreateMul(in1, in2, name("mul")); -} - -const char * Instructions::name(const char *prefix) -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2) -{ - Value *mulRes = mul(in1, in2); - Value *x = m_builder.CreateExtractElement(mulRes, - m_storage->constantInt(0), - name("extractx")); - Value *y = m_builder.CreateExtractElement(mulRes, - m_storage->constantInt(1), - name("extracty")); - Value *z = m_builder.CreateExtractElement(mulRes, - m_storage->constantInt(2), - name("extractz")); - Value *xy = m_builder.CreateAdd(x, y,name("xy")); - Value *dot3 = m_builder.CreateAdd(xy, z, name("dot3")); - return vectorFromVals(dot3, dot3, dot3, dot3); -} - -llvm::Value *Instructions::callFSqrt(llvm::Value *val) -{ - if (!m_llvmFSqrt) { - // predeclare the intrinsic - std::vector fsqrtArgs; - fsqrtArgs.push_back(Type::FloatTy); - ParamAttrsList *fsqrtPal = 0; - FunctionType* fsqrtType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/fsqrtArgs, - /*isVarArg=*/false); - m_llvmFSqrt = new Function( - /*Type=*/fsqrtType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"llvm.sqrt.f32", m_mod); - m_llvmFSqrt->setCallingConv(CallingConv::C); - m_llvmFSqrt->setParamAttrs(fsqrtPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFSqrt, val, - name("sqrt")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::rsq(llvm::Value *in1) -{ - Value *x = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("extractx")); - Value *abs = callFAbs(x); - Value *sqrt = callFSqrt(abs); - - Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - sqrt, - name("rsqrt")); - return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt); -} - -llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w) -{ - Constant *const_vec = Constant::getNullValue(m_floatVecType); - Value *res = m_builder.CreateInsertElement(const_vec, x, - m_storage->constantInt(0), - name("vecx")); - res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), - name("vecxy")); - res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), - name("vecxyz")); - if (w) - res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), - name("vecxyzw")); - return res; -} - -llvm::Value *Instructions::callFAbs(llvm::Value *val) -{ - if (!m_llvmFAbs) { - // predeclare the intrinsic - std::vector fabsArgs; - fabsArgs.push_back(Type::FloatTy); - ParamAttrsList *fabsPal = 0; - FunctionType* fabsType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/fabsArgs, - /*isVarArg=*/false); - m_llvmFAbs = new Function( - /*Type=*/fabsType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"fabs", m_mod); - m_llvmFAbs->setCallingConv(CallingConv::C); - m_llvmFAbs->setParamAttrs(fabsPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFAbs, val, - name("fabs")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::lit(llvm::Value *in) -{ - if (!m_llvmLit) { - m_llvmLit = m_mod->getFunction("lit"); - } - CallInst *call = m_builder.CreateCall(m_llvmLit, in, name("litres")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) -{ - Value *res = m_builder.CreateSub(in1, in2, name("sub")); - return res; -} - -llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) -{ - if (!m_llvmPow) { - // predeclare the intrinsic - std::vector powArgs; - powArgs.push_back(Type::FloatTy); - powArgs.push_back(Type::FloatTy); - ParamAttrsList *powPal = 0; - FunctionType* powType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/powArgs, - /*isVarArg=*/false); - m_llvmPow = new Function( - /*Type=*/powType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"llvm.pow.f32", m_mod); - m_llvmPow->setCallingConv(CallingConv::C); - m_llvmPow->setParamAttrs(powPal); - } - std::vector params; - params.push_back(val1); - params.push_back(val2); - CallInst *call = m_builder.CreateCall(m_llvmPow, params.begin(), params.end(), - name("pow")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2) -{ - Value *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - Value *x2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(0), - name("x2")); - llvm::Value *val = callPow(x1, x2); - return vectorFromVals(val, val, val, val); -} - -llvm::Value * Instructions::rcp(llvm::Value *in1) -{ - Value *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - Value *res = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - x1, name("rcp")); - return vectorFromVals(res, res, res, res); -} - -llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) -{ - Value *mulRes = mul(in1, in2); - std::vector vec = extractVector(mulRes); - Value *xy = m_builder.CreateAdd(vec[0], vec[1], name("xy")); - Value *xyz = m_builder.CreateAdd(xy, vec[2], name("xyz")); - Value *dot4 = m_builder.CreateAdd(xyz, vec[3], name("dot4")); - return vectorFromVals(dot4, dot4, dot4, dot4); -} - -llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) -{ - Value *mulRes = mul(in1, in2); - std::vector vec1 = extractVector(mulRes); - Value *xy = m_builder.CreateAdd(vec1[0], vec1[1], name("xy")); - Value *xyz = m_builder.CreateAdd(xy, vec1[2], name("xyz")); - Value *dph = m_builder.CreateAdd(xyz, vec1[3], name("dph")); - return vectorFromVals(dph, dph, dph, dph); -} - -llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) -{ - Value *y1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(1), - name("y1")); - Value *z = m_builder.CreateExtractElement(in1, - m_storage->constantInt(2), - name("z")); - Value *y2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(1), - name("y2")); - Value *w = m_builder.CreateExtractElement(in2, - m_storage->constantInt(3), - name("w")); - Value *ry = m_builder.CreateMul(y1, y2, name("tyuy")); - return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), - ry, z, w); -} - -llvm::Value * Instructions::ex2(llvm::Value *in) -{ - llvm::Value *val = callPow(ConstantFP::get(Type::FloatTy, APFloat(2.f)), - m_builder.CreateExtractElement( - in, m_storage->constantInt(0), - name("x1"))); - return vectorFromVals(val, val, val, val); -} - -llvm::Value * Instructions::callFloor(llvm::Value *val) -{ - if (!m_llvmFloor) { - // predeclare the intrinsic - std::vector floorArgs; - floorArgs.push_back(Type::FloatTy); - ParamAttrsList *floorPal = 0; - FunctionType* floorType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/floorArgs, - /*isVarArg=*/false); - m_llvmFloor = new Function( - /*Type=*/floorType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"floorf", m_mod); - m_llvmFloor->setCallingConv(CallingConv::C); - m_llvmFloor->setParamAttrs(floorPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFloor, val, - name("floorf")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::floor(llvm::Value *in) -{ - std::vector vec = extractVector(in); - return vectorFromVals(callFloor(vec[0]), callFloor(vec[1]), - callFloor(vec[2]), callFloor(vec[3])); -} - -llvm::Value * Instructions::arl(llvm::Value *in) -{ - return floor(in); -} - -llvm::Value * Instructions::frc(llvm::Value *in) -{ - llvm::Value *flr = floor(in); - return sub(in, flr); -} - -llvm::Value * Instructions::callFLog(llvm::Value *val) -{ - if (!m_llvmFlog) { - // predeclare the intrinsic - std::vector flogArgs; - flogArgs.push_back(Type::FloatTy); - ParamAttrsList *flogPal = 0; - FunctionType* flogType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/flogArgs, - /*isVarArg=*/false); - m_llvmFlog = new Function( - /*Type=*/flogType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"logf", m_mod); - m_llvmFlog->setCallingConv(CallingConv::C); - m_llvmFlog->setParamAttrs(flogPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFlog, val, - name("logf")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::lg2(llvm::Value *in) -{ - std::vector vec = extractVector(in); - llvm::Value *const_vec = constVector(1.442695f, 1.442695f, - 1.442695f, 1.442695f); - return mul(vectorFromVals(callFLog(vec[0]), callFLog(vec[1]), - callFLog(vec[2]), callFLog(vec[3])), const_vec); -} - -llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) -{ - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); - Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], - name("selx")); - - Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); - Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], - name("sely")); - - Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); - Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], - name("selz")); - - Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); - Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], - name("selw")); - - return vectorFromVals(selx, sely, selz, selw); -} - -llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) -{ - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], - name("xcmp")); - Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], - name("selx")); - - Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], - name("ycmp")); - Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], - name("sely")); - - Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], - name("zcmp")); - Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], - name("selz")); - - Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], - name("wcmp")); - Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], - name("selw")); - - return vectorFromVals(selx, sely, selz, selw); -} - -void Instructions::printVector(llvm::Value *val) -{ - static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A"; - - if (!m_fmtPtr) { - Constant *format = ConstantArray::get(frmt, true); - ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1); - GlobalVariable* globalFormat = new GlobalVariable( - /*Type=*/arrayTy, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name(".str"), - m_mod); - globalFormat->setInitializer(format); - - Constant* const_int0 = Constant::getNullValue(IntegerType::get(32)); - std::vector const_ptr_21_indices; - const_ptr_21_indices.push_back(const_int0); - const_ptr_21_indices.push_back(const_int0); - m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat, - &const_ptr_21_indices[0], const_ptr_21_indices.size()); - } - - Function *func_printf = m_mod->getFunction("printf"); - if (!func_printf) - func_printf = declarePrintf(); - assert(func_printf); - std::vector vec = extractVector(val); - Value *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx")); - Value *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy")); - Value *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz")); - Value *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw")); - std::vector params; - params.push_back(m_fmtPtr); - params.push_back(dx); - params.push_back(dy); - params.push_back(dz); - params.push_back(dw); - CallInst *call = m_builder.CreateCall(func_printf, params.begin(), params.end(), - name("printf")); - call->setCallingConv(CallingConv::C); - call->setTailCall(true); -} - -llvm::Function * Instructions::declarePrintf() -{ - std::vector args; - ParamAttrsList *params = 0; - FunctionType* funcTy = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/args, - /*isVarArg=*/true); - Function* func_printf = new Function( - /*Type=*/funcTy, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", m_mod); - func_printf->setCallingConv(CallingConv::C); - func_printf->setParamAttrs(params); - return func_printf; -} - - -llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) -{ - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - Constant *const0f = Constant::getNullValue(Type::FloatTy); - - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], name("xcmp")); - Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - - Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], name("ycmp")); - Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - - Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], name("zcmp")); - Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - - Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], name("wcmp")); - Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); - - return vectorFromVals(x, y, z, w); -} -llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) -{ - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - Constant *const0f = Constant::getNullValue(Type::FloatTy); - - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOGE(vec1[0], vec2[0], name("xcmp")); - Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - - Value *ycmp = m_builder.CreateFCmpOGE(vec1[1], vec2[1], name("ycmp")); - Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - - Value *zcmp = m_builder.CreateFCmpOGE(vec1[2], vec2[2], name("zcmp")); - Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - - Value *wcmp = m_builder.CreateFCmpOGE(vec1[3], vec2[3], name("wcmp")); - Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); - - return vectorFromVals(x, y, z, w); -} - - -llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) -{ - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - Constant *const0f = Constant::getNullValue(Type::FloatTy); - - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); - Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - - Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); - Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - - Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); - Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - - Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); - Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); - - return vectorFromVals(x, y, z, w); -} - -llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) -{ - Value *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - Value *y1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(1), - name("y1")); - Value *z1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(2), - name("z1")); - - Value *x2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(0), - name("x2")); - Value *y2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(1), - name("y2")); - Value *z2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(2), - name("z2")); - Value *y1z2 = mul(y1, z2); - Value *z1y2 = mul(z1, y2); - - Value *z1x2 = mul(z1, x2); - Value *x1z2 = mul(x1, z2); - - Value *x1y2 = mul(x1, y2); - Value *y1x2 = mul(y1, x2); - - return vectorFromVals(sub(y1z2, z1y2), sub(z1x2, x1z2), sub(x1y2, y1x2)); -} - - -llvm::Value * Instructions::abs(llvm::Value *in) -{ - std::vector vec = extractVector(in); - Value *xabs = callFAbs(vec[0]); - Value *yabs = callFAbs(vec[1]); - Value *zabs = callFAbs(vec[2]); - Value *wabs = callFAbs(vec[3]); - return vectorFromVals(xabs, yabs, zabs, wabs); -} - -void Instructions::ifop(llvm::Value *in) -{ - BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); - BasicBlock *ifend = new BasicBlock(name("ifthenend"), m_func,0); - - //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0); - //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0); - //BasicBlock *wblock = new BasicBlock(name("wblock"), m_func,0); - - Constant *float0 = Constant::getNullValue(Type::FloatTy); - - Value *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0), - name("extractx")); - Value *xcmp = m_builder.CreateFCmpUNE(x, float0, name("xcmp")); - m_builder.CreateCondBr(xcmp, ifthen, ifend); - //m_builder.SetInsertPoint(yblock); - - m_builder.SetInsertPoint(ifthen); - m_ifStack.push(ifend); -} - -llvm::BasicBlock * Instructions::currentBlock() const -{ - return m_builder.GetInsertBlock(); -} - -void Instructions::elseop() -{ - assert(!m_ifStack.empty()); - BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); - m_builder.CreateBr(ifend); - m_builder.SetInsertPoint(m_ifStack.top()); - currentBlock()->setName(name("ifelse")); - m_ifStack.pop(); - m_ifStack.push(ifend); -} - -void Instructions::endif() -{ - assert(!m_ifStack.empty()); - m_builder.CreateBr(m_ifStack.top()); - m_builder.SetInsertPoint(m_ifStack.top()); - m_ifStack.pop(); -} - -llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in3) -{ - llvm::Value *m = mul(in1, in2); - llvm::Value *vec1 = constVector(1.f, 1.f, 1.f, 1.f); - llvm::Value *s = sub(vec1, in1); - return add(m, mul(s, in3)); -} - -void Instructions::beginLoop() -{ - BasicBlock *begin = new BasicBlock(name("loop"), m_func,0); - BasicBlock *end = new BasicBlock(name("endloop"), m_func,0); - - m_builder.CreateBr(begin); - Loop loop; - loop.begin = begin; - loop.end = end; - m_builder.SetInsertPoint(begin); - m_loopStack.push(loop); -} - -void Instructions::endLoop() -{ - assert(!m_loopStack.empty()); - Loop loop = m_loopStack.top(); - m_builder.CreateBr(loop.begin); - loop.end->moveAfter(currentBlock()); - m_builder.SetInsertPoint(loop.end); - m_loopStack.pop(); -} - -void Instructions::brk() -{ - assert(!m_loopStack.empty()); - BasicBlock *unr = new BasicBlock(name("unreachable"), m_func,0); - m_builder.CreateBr(m_loopStack.top().end); - m_builder.SetInsertPoint(unr); -} - -llvm::Value * Instructions::trunc(llvm::Value *in) -{ - std::vector vec = extractVector(in); - Value *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32), - name("ftoix")); - Value *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32), - name("ftoiy")); - Value *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32), - name("ftoiz")); - Value *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32), - name("ftoiw")); - Value *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy, - name("fx")); - Value *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy, - name("fy")); - Value *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy, - name("fz")); - Value *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy, - name("fw")); - return vectorFromVals(fx, fy, fz, fw); -} - -void Instructions::end() -{ - m_builder.CreateRetVoid(); -} - -void Instructions::cal(int label, llvm::Value *input) -{ - std::vector params; - params.push_back(input); - llvm::Function *func = findFunction(label); - - m_builder.CreateCall(func, params.begin(), params.end()); -} - -llvm::Function * Instructions::declareFunc(int label) -{ - PointerType *vecPtr = PointerType::getUnqual(m_floatVecType); - std::vector args; - args.push_back(vecPtr); - args.push_back(vecPtr); - args.push_back(vecPtr); - args.push_back(vecPtr); - ParamAttrsList *params = 0; - FunctionType *funcType = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/args, - /*isVarArg=*/false); - std::string name = createFuncName(label); - Function *func = new Function( - /*Type=*/funcType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/name.c_str(), m_mod); - func->setCallingConv(CallingConv::C); - func->setParamAttrs(params); - return func; -} - -void Instructions::bgnSub(unsigned label) -{ - llvm::Function *func = findFunction(label); - - Function::arg_iterator args = func->arg_begin(); - Value *ptr_INPUT = args++; - ptr_INPUT->setName("INPUT"); - m_storage->pushArguments(ptr_INPUT); - - llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); - - m_func = func; - m_builder.SetInsertPoint(entry); -} - -void Instructions::endSub() -{ - m_func = 0; - m_builder.SetInsertPoint(0); -} - -llvm::Function * Instructions::findFunction(int label) -{ - llvm::Function *func = m_functions[label]; - if (!func) { - func = declareFunc(label); - m_functions[label] = func; - } - return func; -} - -llvm::Value * Instructions::constVector(float x, float y, float z, float w) -{ - std::vector vec(4); - vec[0] = ConstantFP::get(Type::FloatTy, APFloat(x)); - vec[1] = ConstantFP::get(Type::FloatTy, APFloat(y)); - vec[2] = ConstantFP::get(Type::FloatTy, APFloat(z)); - vec[3] = ConstantFP::get(Type::FloatTy, APFloat(w)); - return ConstantVector::get(m_floatVecType, vec); -} - - -std::vector Instructions::extractVector(llvm::Value *vec) -{ - std::vector elems(4); - elems[0] = m_builder.CreateExtractElement(vec, m_storage->constantInt(0), - name("x")); - elems[1] = m_builder.CreateExtractElement(vec, m_storage->constantInt(1), - name("y")); - elems[2] = m_builder.CreateExtractElement(vec, m_storage->constantInt(2), - name("z")); - elems[3] = m_builder.CreateExtractElement(vec, m_storage->constantInt(3), - name("w")); - return elems; -} - -llvm::Value * Instructions::cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3) -{ - llvm::Function *func = m_mod->getFunction("cmp"); - assert(func); - - std::vector params; - params.push_back(in1); - params.push_back(in2); - params.push_back(in3); - CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), name("cmpres")); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::cos(llvm::Value *in) -{ -#if 0 - llvm::Function *func = m_mod->getFunction("vcos"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("cosres")); - call->setTailCall(false); - return call; -#else - std::vector elems = extractVector(in); - Function *func = m_mod->getFunction("cosf"); - assert(func); - CallInst *cos = m_builder.CreateCall(func, elems[0], name("cosres")); - cos->setCallingConv(CallingConv::C); - cos->setTailCall(true); - return vectorFromVals(cos, cos, cos, cos); -#endif -} - -llvm::Value * Instructions::scs(llvm::Value *in) -{ - llvm::Function *func = m_mod->getFunction("scs"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("scsres")); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::kilp(llvm::Value *in) -{ - llvm::Function *func = m_mod->getFunction("kilp"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("kilpres")); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::sin(llvm::Value *in) -{ - llvm::Function *func = m_mod->getFunction("vsin"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("sinres")); - call->setTailCall(false); - return call; -} -#endif //MESA_LLVM - - diff --git a/src/gallium/aux/llvm/instructions.h b/src/gallium/aux/llvm/instructions.h deleted file mode 100644 index 9ebc17dd8e..0000000000 --- a/src/gallium/aux/llvm/instructions.h +++ /dev/null @@ -1,152 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef INSTRUCTIONS_H -#define INSTRUCTIONS_H - -#include -#include -#include -#include - -#include -#include - -namespace llvm { - class VectorType; - class Function; -} - -class Storage; - -class Instructions -{ -public: - Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, - Storage *storage); - - llvm::BasicBlock *currentBlock() const; - - llvm::Value *abs(llvm::Value *in1); - llvm::Value *arl(llvm::Value *in1); - llvm::Value *add(llvm::Value *in1, llvm::Value *in2); - void beginLoop(); - void bgnSub(unsigned); - void brk(); - void cal(int label, llvm::Value *input); - llvm::Value *cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); - llvm::Value *cos(llvm::Value *in); - llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); - void elseop(); - void endif(); - void endLoop(); - void end(); - void endSub(); - llvm::Value *ex2(llvm::Value *in); - llvm::Value *floor(llvm::Value *in); - llvm::Value *frc(llvm::Value *in); - void ifop(llvm::Value *in); - llvm::Value *kilp(llvm::Value *in); - llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in3); - llvm::Value *lit(llvm::Value *in); - llvm::Value *lg2(llvm::Value *in); - llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in2); - llvm::Value *min(llvm::Value *in1, llvm::Value *in2); - llvm::Value *max(llvm::Value *in1, llvm::Value *in2); - llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); - llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); - llvm::Value *rcp(llvm::Value *in); - llvm::Value *rsq(llvm::Value *in); - llvm::Value *scs(llvm::Value *in); - llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); - llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); - llvm::Value *sin(llvm::Value *in); - llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); - llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); - llvm::Value *trunc(llvm::Value *in); - - void printVector(llvm::Value *val); -private: - const char *name(const char *prefix); - - llvm::Value *callFAbs(llvm::Value *val); - llvm::Value *callFloor(llvm::Value *val); - llvm::Value *callFSqrt(llvm::Value *val); - llvm::Value *callFLog(llvm::Value *val); - llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2); - - llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w=0); - - llvm::Value *constVector(float x, float y, float z, float w); - - llvm::Function *declarePrintf(); - llvm::Function *declareFunc(int label); - - llvm::Function *findFunction(int label); - - std::vector extractVector(llvm::Value *vec); -private: - llvm::Module *m_mod; - llvm::Function *m_func; - char m_name[32]; - llvm::LLVMFoldingBuilder m_builder; - int m_idx; - - llvm::VectorType *m_floatVecType; - - llvm::Function *m_llvmFSqrt; - llvm::Function *m_llvmFAbs; - llvm::Function *m_llvmPow; - llvm::Function *m_llvmFloor; - llvm::Function *m_llvmFlog; - llvm::Function *m_llvmLit; - - llvm::Constant *m_fmtPtr; - - std::stack m_ifStack; - struct Loop { - llvm::BasicBlock *begin; - llvm::BasicBlock *end; - }; - std::stack m_loopStack; - std::map m_functions; - Storage *m_storage; -}; - -#endif diff --git a/src/gallium/aux/llvm/instructionssoa.cpp b/src/gallium/aux/llvm/instructionssoa.cpp deleted file mode 100644 index a4d5046637..0000000000 --- a/src/gallium/aux/llvm/instructionssoa.cpp +++ /dev/null @@ -1,121 +0,0 @@ -#include "instructionssoa.h" - -#include "storagesoa.h" - -#include - -using namespace llvm; - -InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, - llvm::BasicBlock *block, StorageSoa *storage) - : m_builder(block), - m_storage(storage), - m_idx(0) -{ -} - -const char * InstructionsSoa::name(const char *prefix) const -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -llvm::Value * InstructionsSoa::vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w) -{ - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - Constant *constVector = Constant::getNullValue(vectorType); - Value *res = m_builder.CreateInsertElement(constVector, x, - m_storage->constantInt(0), - name("vecx")); - res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), - name("vecxy")); - res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), - name("vecxyz")); - if (w) - res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), - name("vecxyzw")); - return res; -} - -std::vector InstructionsSoa::arl(const std::vector in) -{ - std::vector res(4); - - //Extract x's - llvm::Value *x1 = m_builder.CreateExtractElement(in[0], - m_storage->constantInt(0), - name("extractX")); - //cast it to an unsigned int - x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast")); - - res[0] = x1;//vectorFromVals(x1, x2, x3, x4); - //only x is valid. the others shouldn't be necessary - /* - res[1] = Constant::getNullValue(m_floatVecType); - res[2] = Constant::getNullValue(m_floatVecType); - res[3] = Constant::getNullValue(m_floatVecType); - */ - - return res; -} - - -std::vector InstructionsSoa::add(const std::vector in1, - const std::vector in2) -{ - std::vector res(4); - - res[0] = m_builder.CreateAdd(in1[0], in2[0], name("addx")); - res[1] = m_builder.CreateAdd(in1[1], in2[1], name("addy")); - res[2] = m_builder.CreateAdd(in1[2], in2[2], name("addz")); - res[3] = m_builder.CreateAdd(in1[3], in2[3], name("addw")); - - return res; -} - -std::vector InstructionsSoa::mul(const std::vector in1, - const std::vector in2) -{ - std::vector res(4); - - res[0] = m_builder.CreateMul(in1[0], in2[0], name("mulx")); - res[1] = m_builder.CreateMul(in1[1], in2[1], name("muly")); - res[2] = m_builder.CreateMul(in1[2], in2[2], name("mulz")); - res[3] = m_builder.CreateMul(in1[3], in2[3], name("mulw")); - - return res; -} - -void InstructionsSoa::end() -{ - m_builder.CreateRetVoid(); -} - -std::vector InstructionsSoa::madd(const std::vector in1, - const std::vector in2, - const std::vector in3) -{ - std::vector res = mul(in1, in2); - return add(res, in3); -} - -std::vector InstructionsSoa::extractVector(llvm::Value *vector) -{ - std::vector res(4); - res[0] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(0), - name("extract1X")); - res[1] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(1), - name("extract2X")); - res[2] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(2), - name("extract3X")); - res[3] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(3), - name("extract4X")); - - return res; -} diff --git a/src/gallium/aux/llvm/instructionssoa.h b/src/gallium/aux/llvm/instructionssoa.h deleted file mode 100644 index 4169dcbb2e..0000000000 --- a/src/gallium/aux/llvm/instructionssoa.h +++ /dev/null @@ -1,74 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INSTRUCTIONSSOA_H -#define INSTRUCTIONSSOA_H - -#include - -#include - -namespace llvm { - class Module; - class Function; - class BasicBlock; - class Value; -} -class StorageSoa; - -class InstructionsSoa -{ -public: - InstructionsSoa(llvm::Module *mod, llvm::Function *func, - llvm::BasicBlock *block, StorageSoa *storage); - - std::vector arl(const std::vector in); - - std::vector add(const std::vector in1, - const std::vector in2); - std::vector madd(const std::vector in1, - const std::vector in2, - const std::vector in3); - std::vector mul(const std::vector in1, - const std::vector in2); - void end(); - - std::vector extractVector(llvm::Value *vector); -private: - const char * name(const char *prefix) const; - llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w); -private: - llvm::LLVMFoldingBuilder m_builder; - StorageSoa *m_storage; -private: - mutable int m_idx; - mutable char m_name[32]; -}; - - -#endif diff --git a/src/gallium/aux/llvm/llvm_builtins.c b/src/gallium/aux/llvm/llvm_builtins.c deleted file mode 100644 index 4f98d754ba..0000000000 --- a/src/gallium/aux/llvm/llvm_builtins.c +++ /dev/null @@ -1,115 +0,0 @@ -/*clang --emit-llvm llvm_builtins.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=gallivm_builtins.cpp -f -for=shader -funcname=createGallivmBuiltins*/ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -typedef __attribute__(( ocu_vector_type(4) )) float float4; - -extern float powf(float a, float b); - -inline float approx(float a, float b) -{ - if (b < -128.0f) b = -128.0f; - if (b > 128.0f) b = 128.0f; - if (a < 0) a = 0; - return powf(a, b); -} - -inline float4 lit(float4 tmp) -{ - float4 result; - result.x = 1.0; - result.w = 1.0; - if (tmp.x > 0) { - result.y = tmp.x; - result.z = approx(tmp.y, tmp.w); - } else { - result.y = 0; - result.z = 0; - } - return result; -} - -inline float4 cmp(float4 tmp0, float4 tmp1, float4 tmp2) -{ - float4 result; - - result.x = (tmp0.x < 0.0) ? tmp1.x : tmp2.x; - result.y = (tmp0.y < 0.0) ? tmp1.y : tmp2.y; - result.z = (tmp0.z < 0.0) ? tmp1.z : tmp2.z; - result.w = (tmp0.w < 0.0) ? tmp1.w : tmp2.w; - - return result; -} - -extern float cosf(float val); -extern float sinf(float val); - -inline float4 vcos(float4 val) -{ - float4 result; - printf("VEC IN is %f %f %f %f\n", val.x, val.y, val.z, val.w); - result.x = cosf(val.x); - result.y = cosf(val.x); - result.z = cosf(val.x); - result.w = cosf(val.x); - printf("VEC OUT is %f %f %f %f\n", result.x, result.y, result.z, result.w); - return result; -} - -inline float4 scs(float4 val) -{ - float4 result; - float tmp = val.x; - result.x = cosf(tmp); - result.y = sinf(tmp); - return result; -} - - -inline float4 vsin(float4 val) -{ - float4 result; - float tmp = val.x; - float res = sinf(tmp); - result.x = res; - result.y = res; - result.z = res; - result.w = res; - return result; -} - -inline int kilp(float4 val) -{ - if (val.x < 0 || val.y < 0 || val.z < 0 || val.w < 0) - return 1; - else - return 0; -} diff --git a/src/gallium/aux/llvm/loweringpass.cpp b/src/gallium/aux/llvm/loweringpass.cpp deleted file mode 100644 index 556dbec366..0000000000 --- a/src/gallium/aux/llvm/loweringpass.cpp +++ /dev/null @@ -1,17 +0,0 @@ -#include "loweringpass.h" - -using namespace llvm; - -char LoweringPass::ID = 0; -RegisterPass X("lowering", "Lowering Pass"); - -LoweringPass::LoweringPass() - : ModulePass((intptr_t)&ID) -{ -} - -bool LoweringPass::runOnModule(Module &m) -{ - llvm::cerr << "Hello: " << m.getModuleIdentifier() << "\n"; - return false; -} diff --git a/src/gallium/aux/llvm/loweringpass.h b/src/gallium/aux/llvm/loweringpass.h deleted file mode 100644 index f62dcf6ba7..0000000000 --- a/src/gallium/aux/llvm/loweringpass.h +++ /dev/null @@ -1,15 +0,0 @@ -#ifndef LOWERINGPASS_H -#define LOWERINGPASS_H - -#include "llvm/Pass.h" -#include "llvm/Module.h" - -struct LoweringPass : public llvm::ModulePass -{ - static char ID; - LoweringPass(); - - virtual bool runOnModule(llvm::Module &m); -}; - -#endif diff --git a/src/gallium/aux/llvm/storage.cpp b/src/gallium/aux/llvm/storage.cpp deleted file mode 100644 index c4326de8c5..0000000000 --- a/src/gallium/aux/llvm/storage.cpp +++ /dev/null @@ -1,364 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "storage.h" - -#include "gallivm_p.h" - -#include "pipe/p_shader_tokens.h" -#include -#include -#include - -#include -#include -#include -#include -#include - -using namespace llvm; - -Storage::Storage(llvm::BasicBlock *block, llvm::Value *input) - : m_block(block), - m_INPUT(input), - m_addrs(32), - m_idx(0) -{ - m_floatVecType = VectorType::get(Type::FloatTy, 4); - m_intVecType = VectorType::get(IntegerType::get(32), 4); - - m_undefFloatVec = UndefValue::get(m_floatVecType); - m_undefIntVec = UndefValue::get(m_intVecType); - m_extSwizzleVec = 0; - - m_numConsts = 0; -} - -//can only build vectors with all members in the [0, 9] range -llvm::Constant *Storage::shuffleMask(int vec) -{ - if (!m_extSwizzleVec) { - std::vector elems; - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); - m_extSwizzleVec = ConstantVector::get(m_floatVecType, elems); - } - - if (m_intVecs.find(vec) != m_intVecs.end()) { - return m_intVecs[vec]; - } - int origVec = vec; - Constant* const_vec = 0; - if (origVec == 0) { - const_vec = Constant::getNullValue(m_intVecType); - } else { - int x = gallivm_x_swizzle(vec); - int y = gallivm_y_swizzle(vec); - int z = gallivm_z_swizzle(vec); - int w = gallivm_w_swizzle(vec); - std::vector elems; - elems.push_back(constantInt(x)); - elems.push_back(constantInt(y)); - elems.push_back(constantInt(z)); - elems.push_back(constantInt(w)); - const_vec = ConstantVector::get(m_intVecType, elems); - } - - m_intVecs[origVec] = const_vec; - return const_vec; -} - -llvm::ConstantInt *Storage::constantInt(int idx) -{ - if (m_constInts.find(idx) != m_constInts.end()) { - return m_constInts[idx]; - } - ConstantInt *const_int = ConstantInt::get(APInt(32, idx)); - m_constInts[idx] = const_int; - return const_int; -} - -llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) -{ - Value *val = element(InputsArg, idx, indIdx); - LoadInst *load = new LoadInst(val, name("input"), false, m_block); - load->setAlignment(8); - - return load; -} - -llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) -{ - m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; - - Value *elem = element(ConstsArg, idx, indIdx); - LoadInst *load = new LoadInst(elem, name("const"), false, m_block); - load->setAlignment(8); - return load; -} - -llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) -{ - Constant *mask = shuffleMask(shuffle); - ShuffleVectorInst *res = - new ShuffleVectorInst(vec, m_extSwizzleVec, mask, - name("shuffle"), m_block); - return res; -} - - -llvm::Value *Storage::tempElement(int idx, llvm::Value *indIdx) -{ - Value *elem = element(TempsArg, idx, indIdx); - - LoadInst *load = new LoadInst(elem, name("temp"), false, m_block); - load->setAlignment(8); - - return load; -} - -void Storage::setTempElement(int idx, llvm::Value *val, int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = 0; - if (m_tempWriteMap[idx]) - templ = tempElement(idx); - val = maskWrite(val, mask, templ); - } - Value *elem = element(TempsArg, idx); - StoreInst *st = new StoreInst(val, elem, false, m_block); - st->setAlignment(8); - m_tempWriteMap[idx] = true; -} - -void Storage::setOutputElement(int dstIdx, llvm::Value *val, int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = 0; - if (m_destWriteMap[dstIdx]) - templ = outputElement(dstIdx); - val = maskWrite(val, mask, templ); - } - - Value *elem = element(DestsArg, dstIdx); - StoreInst *st = new StoreInst(val, elem, false, m_block); - st->setAlignment(8); - m_destWriteMap[dstIdx] = true; -} - -llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ) -{ - llvm::Value *dst = templ; - if (!dst) - dst = Constant::getNullValue(m_floatVecType); - if ((mask & TGSI_WRITEMASK_X)) { - llvm::Value *x = new ExtractElementInst(src, unsigned(0), - name("x"), m_block); - dst = new InsertElementInst(dst, x, unsigned(0), - name("dstx"), m_block); - } - if ((mask & TGSI_WRITEMASK_Y)) { - llvm::Value *y = new ExtractElementInst(src, unsigned(1), - name("y"), m_block); - dst = new InsertElementInst(dst, y, unsigned(1), - name("dsty"), m_block); - } - if ((mask & TGSI_WRITEMASK_Z)) { - llvm::Value *z = new ExtractElementInst(src, unsigned(2), - name("z"), m_block); - dst = new InsertElementInst(dst, z, unsigned(2), - name("dstz"), m_block); - } - if ((mask & TGSI_WRITEMASK_W)) { - llvm::Value *w = new ExtractElementInst(src, unsigned(3), - name("w"), m_block); - dst = new InsertElementInst(dst, w, unsigned(3), - name("dstw"), m_block); - } - return dst; -} - -const char * Storage::name(const char *prefix) -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -int Storage::numConsts() const -{ - return m_numConsts; -} - -llvm::Value * Storage::addrElement(int idx) const -{ - Value *ret = m_addrs[idx]; - if (!ret) - return m_undefFloatVec; - return ret; -} - -void Storage::setAddrElement(int idx, llvm::Value *val, int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = m_addrs[idx]; - val = maskWrite(val, mask, templ); - } - m_addrs[idx] = val; -} - -llvm::Value * Storage::extractIndex(llvm::Value *vec) -{ - llvm::Value *x = new ExtractElementInst(vec, unsigned(0), - name("x"), m_block); - return new FPToSIInst(x, IntegerType::get(32), name("intidx"), m_block); -} - -void Storage::setCurrentBlock(llvm::BasicBlock *block) -{ - m_block = block; -} - -llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx) -{ - Value *elem = element(DestsArg, idx, indIdx); - LoadInst *load = new LoadInst(elem, name("output"), false, m_block); - load->setAlignment(8); - - return load; -} - -llvm::Value * Storage::inputPtr() const -{ - return m_INPUT; -} - -void Storage::pushArguments(llvm::Value *input) -{ - m_argStack.push(m_INPUT); - - m_INPUT = input; -} - -void Storage::popArguments() -{ - m_INPUT = m_argStack.top(); - m_argStack.pop(); -} - -void Storage::pushTemps() -{ - m_extSwizzleVec = 0; -} - -void Storage::popTemps() -{ -} - -llvm::Value * Storage::immediateElement(int idx) -{ - return m_immediates[idx]; -} - -void Storage::addImmediate(float *val) -{ - std::vector vec(4); - vec[0] = ConstantFP::get(Type::FloatTy, APFloat(val[0])); - vec[1] = ConstantFP::get(Type::FloatTy, APFloat(val[1])); - vec[2] = ConstantFP::get(Type::FloatTy, APFloat(val[2])); - vec[3] = ConstantFP::get(Type::FloatTy, APFloat(val[3])); - m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); -} - - -llvm::Value * Storage::elemPtr(Args arg) -{ - std::vector indices; - indices.push_back(constantInt(0)); - indices.push_back(constantInt(static_cast(arg))); - GetElementPtrInst *getElem = new GetElementPtrInst(m_INPUT, - indices.begin(), - indices.end(), - name("input_ptr"), - m_block); - return new LoadInst(getElem, name("input_field"), false, m_block); -} - -llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx, - llvm::Value *indIdx ) -{ - GetElementPtrInst *getElem = 0; - - if (indIdx) { - getElem = new GetElementPtrInst(ptr, - BinaryOperator::create(Instruction::Add, - indIdx, - constantInt(idx), - name("add"), - m_block), - name("field"), - m_block); - } else { - getElem = new GetElementPtrInst(ptr, - constantInt(idx), - name("field"), - m_block); - } - return getElem; -} - -llvm::Value * Storage::element(Args arg, int idx, llvm::Value *indIdx ) -{ - Value *val = elemPtr(arg); - return elemIdx(val, idx, indIdx); -} - -void Storage::setKilElement(llvm::Value *val) -{ - std::vector indices; - indices.push_back(constantInt(0)); - indices.push_back(constantInt(static_cast(KilArg))); - GetElementPtrInst *elem = new GetElementPtrInst(m_INPUT, - indices.begin(), - indices.end(), - name("kil_ptr"), - m_block); - StoreInst *st = new StoreInst(val, elem, false, m_block); - st->setAlignment(8); -} - -#endif //MESA_LLVM - - diff --git a/src/gallium/aux/llvm/storage.h b/src/gallium/aux/llvm/storage.h deleted file mode 100644 index 8574f7554e..0000000000 --- a/src/gallium/aux/llvm/storage.h +++ /dev/null @@ -1,133 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef STORAGE_H -#define STORAGE_H - -#include -#include -#include -#include - -namespace llvm { - class BasicBlock; - class Constant; - class ConstantInt; - class LoadInst; - class Value; - class VectorType; -} - -class Storage -{ -public: - Storage(llvm::BasicBlock *block, - llvm::Value *input); - - llvm::Value *inputPtr() const; - - void setCurrentBlock(llvm::BasicBlock *block); - - llvm::ConstantInt *constantInt(int); - llvm::Constant *shuffleMask(int vec); - llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); - llvm::Value *constElement(int idx, llvm::Value *indIdx =0); - llvm::Value *outputElement(int idx, llvm::Value *indIdx =0); - llvm::Value *tempElement(int idx, llvm::Value *indIdx =0); - llvm::Value *immediateElement(int idx); - - void setOutputElement(int dstIdx, llvm::Value *val, int mask); - void setTempElement(int idx, llvm::Value *val, int mask); - - llvm::Value *addrElement(int idx) const; - void setAddrElement(int idx, llvm::Value *val, int mask); - - void setKilElement(llvm::Value *val); - - llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); - - llvm::Value *extractIndex(llvm::Value *vec); - - int numConsts() const; - - void pushArguments(llvm::Value *input); - void popArguments(); - void pushTemps(); - void popTemps(); - - void addImmediate(float *val); - -private: - llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); - const char *name(const char *prefix); - - enum Args { - DestsArg = 0, - InputsArg = 1, - TempsArg = 2, - ConstsArg = 3, - KilArg = 4 - }; - llvm::Value *elemPtr(Args arg); - llvm::Value *elemIdx(llvm::Value *ptr, int idx, - llvm::Value *indIdx = 0); - llvm::Value *element(Args arg, int idx, llvm::Value *indIdx = 0); - -private: - llvm::BasicBlock *m_block; - llvm::Value *m_INPUT; - - std::map m_constInts; - std::map m_intVecs; - std::vector m_addrs; - std::vector m_immediates; - - llvm::VectorType *m_floatVecType; - llvm::VectorType *m_intVecType; - - char m_name[32]; - int m_idx; - - int m_numConsts; - - std::map m_destWriteMap; - std::map m_tempWriteMap; - - llvm::Value *m_undefFloatVec; - llvm::Value *m_undefIntVec; - llvm::Value *m_extSwizzleVec; - - std::stack m_argStack; - std::stack > m_tempStack; -}; - -#endif diff --git a/src/gallium/aux/llvm/storagesoa.cpp b/src/gallium/aux/llvm/storagesoa.cpp deleted file mode 100644 index ed0674a96f..0000000000 --- a/src/gallium/aux/llvm/storagesoa.cpp +++ /dev/null @@ -1,389 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "storagesoa.h" - -#include "gallivm_p.h" - -#include "pipe/p_shader_tokens.h" -#include "pipe/p_debug.h" - -#include -#include -#include - -#include -#include -#include -#include -#include - -using namespace llvm; - - -StorageSoa::StorageSoa(llvm::BasicBlock *block, - llvm::Value *input, - llvm::Value *output, - llvm::Value *consts, - llvm::Value *temps) - : m_block(block), - m_input(input), - m_output(output), - m_consts(consts), - m_temps(temps), - m_immediates(0), - m_idx(0) -{ -} - -void StorageSoa::addImmediate(float *vec) -{ - std::vector vals(4); - vals[0] = vec[0]; - vals[1] = vec[1]; - vals[2] = vec[2]; - vals[3] = vec[3]; - m_immediatesToFlush.push_back(vals); -} - -void StorageSoa::declareImmediates() -{ - if (m_immediatesToFlush.empty()) - return; - - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - ArrayType *vectorChannels = ArrayType::get(vectorType, 4); - ArrayType *arrayType = ArrayType::get(vectorChannels, m_immediatesToFlush.size()); - - m_immediates = new GlobalVariable( - /*Type=*/arrayType, - /*isConstant=*/false, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name("immediates"), - currentModule()); - - std::vector arrayVals; - for (unsigned int i = 0; i < m_immediatesToFlush.size(); ++i) { - std::vector vec = m_immediatesToFlush[i]; - std::vector vals(4); - std::vector channelArray; - - vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0]; - llvm::Constant *xChannel = createConstGlobalVector(vals); - - vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1]; - llvm::Constant *yChannel = createConstGlobalVector(vals); - - vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2]; - llvm::Constant *zChannel = createConstGlobalVector(vals); - - vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3]; - llvm::Constant *wChannel = createConstGlobalVector(vals); - channelArray.push_back(xChannel); - channelArray.push_back(yChannel); - channelArray.push_back(zChannel); - channelArray.push_back(wChannel); - Constant *constChannels = ConstantArray::get(vectorChannels, - channelArray); - arrayVals.push_back(constChannels); - } - Constant *constArray = ConstantArray::get(arrayType, arrayVals); - m_immediates->setInitializer(constArray); - - m_immediatesToFlush.clear(); -} - -llvm::Value *StorageSoa::addrElement(int idx) const -{ - std::map::const_iterator itr = m_addresses.find(idx); - if (itr == m_addresses.end()) { - debug_printf("Trying to access invalid shader 'address'\n"); - return 0; - } - llvm::Value * res = (*itr).second; - - res = new LoadInst(res, name("addr"), false, m_block); - - return res; -} - -std::vector StorageSoa::inputElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_input, idx, 0); - res[1] = element(m_input, idx, 1); - res[2] = element(m_input, idx, 2); - res[3] = element(m_input, idx, 3); - - return res; -} - -std::vector StorageSoa::constElement(llvm::Value *idx) -{ - std::vector res(4); - llvm::Value *xChannel, *yChannel, *zChannel, *wChannel; - - xChannel = elementPointer(m_consts, idx, 0); - yChannel = elementPointer(m_consts, idx, 1); - zChannel = elementPointer(m_consts, idx, 2); - wChannel = elementPointer(m_consts, idx, 3); - - res[0] = alignedArrayLoad(xChannel); - res[1] = alignedArrayLoad(yChannel); - res[2] = alignedArrayLoad(zChannel); - res[3] = alignedArrayLoad(wChannel); - - return res; -} - -std::vector StorageSoa::outputElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_output, idx, 0); - res[1] = element(m_output, idx, 1); - res[2] = element(m_output, idx, 2); - res[3] = element(m_output, idx, 3); - - return res; -} - -std::vector StorageSoa::tempElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_temps, idx, 0); - res[1] = element(m_temps, idx, 1); - res[2] = element(m_temps, idx, 2); - res[3] = element(m_temps, idx, 3); - - return res; -} - -std::vector StorageSoa::immediateElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_immediates, idx, 0); - res[1] = element(m_immediates, idx, 1); - res[2] = element(m_immediates, idx, 2); - res[3] = element(m_immediates, idx, 3); - - return res; -} - -llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index, - int channel) const -{ - std::vector indices; - if (m_immediates == ptr) - indices.push_back(constantInt(0)); - indices.push_back(index); - indices.push_back(constantInt(channel)); - - GetElementPtrInst *getElem = new GetElementPtrInst(ptr, - indices.begin(), - indices.end(), - name("ptr"), - m_block); - return getElem; -} - -llvm::Value * StorageSoa::element(llvm::Value *ptr, llvm::Value *index, - int channel) const -{ - llvm::Value *res = elementPointer(ptr, index, channel); - LoadInst *load = new LoadInst(res, name("element"), false, m_block); - //load->setAlignment(8); - return load; -} - -const char * StorageSoa::name(const char *prefix) const -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -llvm::ConstantInt * StorageSoa::constantInt(int idx) const -{ - if (m_constInts.find(idx) != m_constInts.end()) { - return m_constInts[idx]; - } - ConstantInt *constInt = ConstantInt::get(APInt(32, idx)); - m_constInts[idx] = constInt; - return constInt; -} - -llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val) -{ - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - PointerType *vectorPtr = PointerType::get(vectorType, 0); - - CastInst *cast = new BitCastInst(val, vectorPtr, name("toVector"), m_block); - LoadInst *load = new LoadInst(cast, name("alignLoad"), false, m_block); - load->setAlignment(8); - return load; -} - -llvm::Module * StorageSoa::currentModule() const -{ - if (!m_block || !m_block->getParent()) - return 0; - - return m_block->getParent()->getParent(); -} - -llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector &vec) -{ - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - std::vector immValues; - ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0])); - ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1])); - ConstantFP *constz = ConstantFP::get(Type::FloatTy, APFloat(vec[2])); - ConstantFP *constw = ConstantFP::get(Type::FloatTy, APFloat(vec[3])); - immValues.push_back(constx); - immValues.push_back(consty); - immValues.push_back(constz); - immValues.push_back(constw); - Constant *constVector = ConstantVector::get(vectorType, immValues); - - return constVector; -} - -std::vector StorageSoa::load(Argument type, int idx, int swizzle, - llvm::Value *indIdx) -{ - std::vector val(4); - - //if we have an indirect index, always use that - // if not use the integer offset to create one - llvm::Value *realIndex = 0; - if (indIdx) - realIndex = indIdx; - else - realIndex = constantInt(idx); - debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx); - - switch(type) { - case Input: - val = inputElement(realIndex); - break; - case Output: - val = outputElement(realIndex); - break; - case Temp: - val = tempElement(realIndex); - break; - case Const: - val = constElement(realIndex); - break; - case Immediate: - val = immediateElement(realIndex); - break; - case Address: - debug_printf("Address not handled in the load phase!\n"); - assert(0); - break; - } - if (!gallivm_is_swizzle(swizzle)) - return val; - - std::vector res(4); - - res[0] = val[gallivm_x_swizzle(swizzle)]; - res[1] = val[gallivm_y_swizzle(swizzle)]; - res[2] = val[gallivm_z_swizzle(swizzle)]; - res[3] = val[gallivm_w_swizzle(swizzle)]; - return res; -} - -void StorageSoa::store(Argument type, int idx, const std::vector &val, - int mask) -{ - llvm::Value *out = 0; - switch(type) { - case Output: - out = m_output; - break; - case Temp: - out = m_temps; - break; - case Input: - out = m_input; - break; - case Address: { - llvm::Value *addr = m_addresses[idx]; - if (!addr) { - addAddress(idx); - addr = m_addresses[idx]; - assert(addr); - } - new StoreInst(val[0], addr, false, m_block); - return; - break; - } - default: - debug_printf("Can't save output of this type: %d !\n", type); - assert(0); - break; - } - llvm::Value *realIndex = constantInt(idx); - if ((mask & TGSI_WRITEMASK_X)) { - llvm::Value *xChannel = elementPointer(out, realIndex, 0); - new StoreInst(val[0], xChannel, false, m_block); - } - if ((mask & TGSI_WRITEMASK_Y)) { - llvm::Value *yChannel = elementPointer(out, realIndex, 1); - new StoreInst(val[1], yChannel, false, m_block); - } - if ((mask & TGSI_WRITEMASK_Z)) { - llvm::Value *zChannel = elementPointer(out, realIndex, 2); - new StoreInst(val[2], zChannel, false, m_block); - } - if ((mask & TGSI_WRITEMASK_W)) { - llvm::Value *wChannel = elementPointer(out, realIndex, 3); - new StoreInst(val[3], wChannel, false, m_block); - } -} - -void StorageSoa::addAddress(int idx) -{ - GlobalVariable *val = new GlobalVariable( - /*Type=*/IntegerType::get(32), - /*isConstant=*/false, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name("address"), - currentModule()); - val->setInitializer(Constant::getNullValue(IntegerType::get(32))); - - debug_printf("adding to %d\n", idx); - m_addresses[idx] = val; -} diff --git a/src/gallium/aux/llvm/storagesoa.h b/src/gallium/aux/llvm/storagesoa.h deleted file mode 100644 index 6443351f27..0000000000 --- a/src/gallium/aux/llvm/storagesoa.h +++ /dev/null @@ -1,111 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef STORAGESOA_H -#define STORAGESOA_H - -#include -#include -#include - -namespace llvm { - class BasicBlock; - class Constant; - class ConstantInt; - class GlobalVariable; - class LoadInst; - class Value; - class VectorType; - class Module; -} - -class StorageSoa -{ -public: - enum Argument { - Input, - Output, - Temp, - Const, - Immediate, - Address - }; -public: - StorageSoa(llvm::BasicBlock *block, - llvm::Value *input, - llvm::Value *output, - llvm::Value *consts, - llvm::Value *temps); - - - std::vector load(Argument type, int idx, int swizzle, - llvm::Value *indIdx =0); - void store(Argument type, int idx, const std::vector &val, - int mask); - - void addImmediate(float *vec); - void declareImmediates(); - - void addAddress(int idx); - - llvm::Value * addrElement(int idx) const; - - llvm::ConstantInt *constantInt(int) const; -private: - llvm::Value *elementPointer(llvm::Value *ptr, llvm::Value *indIdx, - int channel) const; - llvm::Value *element(llvm::Value *ptr, llvm::Value *idx, - int channel) const; - const char *name(const char *prefix) const; - llvm::Value *alignedArrayLoad(llvm::Value *val); - llvm::Module *currentModule() const; - llvm::Constant *createConstGlobalVector(const std::vector &vec); - - std::vector inputElement(llvm::Value *indIdx); - std::vector constElement(llvm::Value *indIdx); - std::vector outputElement(llvm::Value *indIdx); - std::vector tempElement(llvm::Value *indIdx); - std::vector immediateElement(llvm::Value *indIdx); -private: - llvm::BasicBlock *m_block; - - llvm::Value *m_input; - llvm::Value *m_output; - llvm::Value *m_consts; - llvm::Value *m_temps; - llvm::GlobalVariable *m_immediates; - - std::map m_addresses; - - std::vector > m_immediatesToFlush; - - mutable std::map m_constInts; - mutable char m_name[32]; - mutable int m_idx; -}; - -#endif diff --git a/src/gallium/aux/llvm/tgsitollvm.cpp b/src/gallium/aux/llvm/tgsitollvm.cpp deleted file mode 100644 index 2cb4acce32..0000000000 --- a/src/gallium/aux/llvm/tgsitollvm.cpp +++ /dev/null @@ -1,1221 +0,0 @@ -#include "tgsitollvm.h" - -#include "gallivm.h" -#include "gallivm_p.h" - -#include "storage.h" -#include "instructions.h" -#include "storagesoa.h" -#include "instructionssoa.h" - -#include "pipe/p_shader_tokens.h" - -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi/util/tgsi_build.h" -#include "tgsi/util/tgsi_dump.h" - - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - - -#include -#include -#include - -using namespace llvm; - -static inline FunctionType *vertexShaderFunctionType() -{ - //Function takes three arguments, - // the calling code has to make sure the types it will - // pass are castable to the following: - // [4 x <4 x float>] inputs, - // [4 x <4 x float>] output, - // [4 x [4 x float]] consts, - // [4 x <4 x float>] temps - - std::vector funcArgs; - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - ArrayType *vectorArray = ArrayType::get(vectorType, 4); - PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0); - - ArrayType *floatArray = ArrayType::get(Type::FloatTy, 4); - ArrayType *constsArray = ArrayType::get(floatArray, 4); - PointerType *constsArrayPtr = PointerType::get(constsArray, 0); - - funcArgs.push_back(vectorArrayPtr);//inputs - funcArgs.push_back(vectorArrayPtr);//output - funcArgs.push_back(constsArrayPtr);//consts - funcArgs.push_back(vectorArrayPtr);//temps - - FunctionType *functionType = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/funcArgs, - /*isVarArg=*/false); - - return functionType; -} - -static inline void -add_interpolator(struct gallivm_ir *ir, - struct gallivm_interpolate *interp) -{ - ir->interpolators[ir->num_interp] = *interp; - ++ir->num_interp; -} - -static void -translate_declaration(struct gallivm_ir *prog, - llvm::Module *module, - Storage *storage, - struct tgsi_full_declaration *decl, - struct tgsi_full_declaration *fd) -{ - if (decl->Declaration.File == TGSI_FILE_INPUT) { - unsigned first, last, mask; - uint interp_method; - - assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - /* Do not touch WPOS.xy */ - if (first == 0) { - mask &= ~TGSI_WRITEMASK_XY; - if (mask == TGSI_WRITEMASK_NONE) { - first++; - if (first > last) { - return; - } - } - } - - interp_method = decl->Interpolation.Interpolate; - - if (mask == TGSI_WRITEMASK_XYZW) { - unsigned i, j; - - for (i = first; i <= last; i++) { - for (j = 0; j < NUM_CHANNELS; j++) { - //interp( mach, i, j ); - struct gallivm_interpolate interp; - interp.type = interp_method; - interp.attrib = i; - interp.chan = j; - add_interpolator(prog, &interp); - } - } - } else { - unsigned i, j; - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - struct gallivm_interpolate interp; - interp.type = interp_method; - interp.attrib = i; - interp.chan = j; - add_interpolator(prog, &interp); - } - } - } - } - } -} - -static void -translate_declarationir(struct gallivm_ir *, - llvm::Module *, - StorageSoa *storage, - struct tgsi_full_declaration *decl, - struct tgsi_full_declaration *) -{ - if (decl->Declaration.File == TGSI_FILE_ADDRESS) { - int idx = decl->u.DeclarationRange.First; - storage->addAddress(idx); - } -} - -static void -translate_immediate(Storage *storage, - struct tgsi_full_immediate *imm) -{ - float vec[4]; - int i; - for (i = 0; i < imm->Immediate.Size - 1; ++i) { - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; - break; - default: - assert(0); - } - } - storage->addImmediate(vec); -} - - -static void -translate_immediateir(StorageSoa *storage, - struct tgsi_full_immediate *imm) -{ - float vec[4]; - int i; - for (i = 0; i < imm->Immediate.Size - 1; ++i) { - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; - break; - default: - assert(0); - } - } - storage->addImmediate(vec); -} - -static inline int -swizzleInt(struct tgsi_full_src_register *src) -{ - int swizzle = 0; - int start = 1000; - - for (int k = 0; k < 4; ++k) { - swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; - start /= 10; - } - return swizzle; -} - -static inline llvm::Value * -swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, - Storage *storage) -{ - int swizzle = swizzleInt(src); - - if (gallivm_is_swizzle(swizzle)) { - /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ - val = storage->shuffleVector(val, swizzle); - } - return val; -} - -static void -translate_instruction(llvm::Module *module, - Storage *storage, - Instructions *instr, - struct tgsi_full_instruction *inst, - struct tgsi_full_instruction *fi, - unsigned instno) -{ - llvm::Value *inputs[4]; - inputs[0] = 0; - inputs[1] = 0; - inputs[2] = 0; - inputs[3] = 0; - - for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - llvm::Value *val = 0; - llvm::Value *indIdx = 0; - - if (src->SrcRegister.Indirect) { - indIdx = storage->addrElement(src->SrcRegisterInd.Index); - indIdx = storage->extractIndex(indIdx); - } - if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->constElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->inputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->tempElement(src->SrcRegister.Index); - } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->outputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->immediateElement(src->SrcRegister.Index); - } else { - fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); - return; - } - - inputs[i] = swizzleVector(val, src, storage); - } - - /*if (inputs[0]) - instr->printVector(inputs[0]); - if (inputs[1]) - instr->printVector(inputs[1]);*/ - llvm::Value *out = 0; - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: { - out = instr->arl(inputs[0]); - } - break; - case TGSI_OPCODE_MOV: { - out = inputs[0]; - } - break; - case TGSI_OPCODE_LIT: { - out = instr->lit(inputs[0]); - } - break; - case TGSI_OPCODE_RCP: { - out = instr->rcp(inputs[0]); - } - break; - case TGSI_OPCODE_RSQ: { - out = instr->rsq(inputs[0]); - } - break; - case TGSI_OPCODE_EXP: - break; - case TGSI_OPCODE_LOG: - break; - case TGSI_OPCODE_MUL: { - out = instr->mul(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_ADD: { - out = instr->add(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP3: { - out = instr->dp3(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP4: { - out = instr->dp4(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DST: { - out = instr->dst(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MIN: { - out = instr->min(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAX: { - out = instr->max(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SLT: { - out = instr->slt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SGE: { - out = instr->sge(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAD: { - out = instr->madd(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SUB: { - out = instr->sub(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_LERP: { - out = instr->lerp(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_CND: - break; - case TGSI_OPCODE_CND0: - break; - case TGSI_OPCODE_DOT2ADD: - break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { - out = instr->frc(inputs[0]); - } - break; - case TGSI_OPCODE_CLAMP: - break; - case TGSI_OPCODE_FLOOR: { - out = instr->floor(inputs[0]); - } - break; - case TGSI_OPCODE_ROUND: - break; - case TGSI_OPCODE_EXPBASE2: { - out = instr->ex2(inputs[0]); - } - break; - case TGSI_OPCODE_LOGBASE2: { - out = instr->lg2(inputs[0]); - } - break; - case TGSI_OPCODE_POWER: { - out = instr->pow(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_CROSSPRODUCT: { - out = instr->cross(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; - case TGSI_OPCODE_ABS: { - out = instr->abs(inputs[0]); - } - break; - case TGSI_OPCODE_RCC: - break; - case TGSI_OPCODE_DPH: { - out = instr->dph(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_COS: { - out = instr->cos(inputs[0]); - } - break; - case TGSI_OPCODE_DDX: - break; - case TGSI_OPCODE_DDY: - break; - case TGSI_OPCODE_KILP: { - out = instr->kilp(inputs[0]); - storage->setKilElement(out); - return; - } - break; - case TGSI_OPCODE_PK2H: - break; - case TGSI_OPCODE_PK2US: - break; - case TGSI_OPCODE_PK4B: - break; - case TGSI_OPCODE_PK4UB: - break; - case TGSI_OPCODE_RFL: - break; - case TGSI_OPCODE_SEQ: - break; - case TGSI_OPCODE_SFL: - break; - case TGSI_OPCODE_SGT: { - out = instr->sgt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SIN: { - out = instr->sin(inputs[0]); - } - break; - case TGSI_OPCODE_SLE: - break; - case TGSI_OPCODE_SNE: - break; - case TGSI_OPCODE_STR: - break; - case TGSI_OPCODE_TEX: - break; - case TGSI_OPCODE_TXD: - break; - case TGSI_OPCODE_UP2H: - break; - case TGSI_OPCODE_UP2US: - break; - case TGSI_OPCODE_UP4B: - break; - case TGSI_OPCODE_UP4UB: - break; - case TGSI_OPCODE_X2D: - break; - case TGSI_OPCODE_ARA: - break; - case TGSI_OPCODE_ARR: - break; - case TGSI_OPCODE_BRA: - break; - case TGSI_OPCODE_CAL: { - instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr()); - return; - } - break; - case TGSI_OPCODE_RET: { - instr->end(); - return; - } - break; - case TGSI_OPCODE_SSG: - break; - case TGSI_OPCODE_CMP: { - out = instr->cmp(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SCS: { - out = instr->scs(inputs[0]); - } - break; - case TGSI_OPCODE_TXB: - break; - case TGSI_OPCODE_NRM: - break; - case TGSI_OPCODE_DIV: - break; - case TGSI_OPCODE_DP2: - break; - case TGSI_OPCODE_TXL: - break; - case TGSI_OPCODE_BRK: { - instr->brk(); - return; - } - break; - case TGSI_OPCODE_IF: { - instr->ifop(inputs[0]); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_LOOP: - break; - case TGSI_OPCODE_REP: - break; - case TGSI_OPCODE_ELSE: { - instr->elseop(); - storage->setCurrentBlock(instr->currentBlock()); - return; //only state update - } - break; - case TGSI_OPCODE_ENDIF: { - instr->endif(); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_ENDLOOP: - break; - case TGSI_OPCODE_ENDREP: - break; - case TGSI_OPCODE_PUSHA: - break; - case TGSI_OPCODE_POPA: - break; - case TGSI_OPCODE_CEIL: - break; - case TGSI_OPCODE_I2F: - break; - case TGSI_OPCODE_NOT: - break; - case TGSI_OPCODE_TRUNC: { - out = instr->trunc(inputs[0]); - } - break; - case TGSI_OPCODE_SHL: - break; - case TGSI_OPCODE_SHR: - break; - case TGSI_OPCODE_AND: - break; - case TGSI_OPCODE_OR: - break; - case TGSI_OPCODE_MOD: - break; - case TGSI_OPCODE_XOR: - break; - case TGSI_OPCODE_SAD: - break; - case TGSI_OPCODE_TXF: - break; - case TGSI_OPCODE_TXQ: - break; - case TGSI_OPCODE_CONT: - break; - case TGSI_OPCODE_EMIT: - break; - case TGSI_OPCODE_ENDPRIM: - break; - case TGSI_OPCODE_BGNLOOP2: { - instr->beginLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_BGNSUB: { - instr->bgnSub(instno); - storage->setCurrentBlock(instr->currentBlock()); - storage->pushTemps(); - return; - } - break; - case TGSI_OPCODE_ENDLOOP2: { - instr->endLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_ENDSUB: { - instr->endSub(); - storage->setCurrentBlock(instr->currentBlock()); - storage->popArguments(); - storage->popTemps(); - return; - } - break; - case TGSI_OPCODE_NOISE1: - break; - case TGSI_OPCODE_NOISE2: - break; - case TGSI_OPCODE_NOISE3: - break; - case TGSI_OPCODE_NOISE4: - break; - case TGSI_OPCODE_NOP: - break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; - case TGSI_OPCODE_NRM4: - break; - case TGSI_OPCODE_CALLNZ: - break; - case TGSI_OPCODE_IFC: - break; - case TGSI_OPCODE_BREAKC: - break; - case TGSI_OPCODE_KIL: - break; - case TGSI_OPCODE_END: - instr->end(); - return; - break; - default: - fprintf(stderr, "ERROR: Unknown opcode %d\n", - inst->Instruction.Opcode); - assert(0); - break; - } - - if (!out) { - fprintf(stderr, "ERROR: unsupported opcode %d\n", - inst->Instruction.Opcode); - assert(!"Unsupported opcode"); - } - - /* # not sure if we need this */ - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - /*TXT( "_SAT" );*/ - break; - case TGSI_SAT_MINUS_PLUS_ONE: - /*TXT( "_SAT[-1,1]" );*/ - break; - default: - assert( 0 ); - } - - /* store results */ - for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else { - fprintf(stderr, "ERROR: unsupported LLVM destination!"); - assert(!"wrong destination"); - } - } -} - - -static void -translate_instructionir(llvm::Module *module, - StorageSoa *storage, - InstructionsSoa *instr, - struct tgsi_full_instruction *inst, - struct tgsi_full_instruction *fi, - unsigned instno) -{ - std::vector< std::vector > inputs(inst->Instruction.NumSrcRegs); - - for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - std::vector val; - llvm::Value *indIdx = 0; - int swizzle = swizzleInt(src); - - if (src->SrcRegister.Indirect) { - indIdx = storage->addrElement(src->SrcRegisterInd.Index); - } - if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->load(StorageSoa::Const, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->load(StorageSoa::Input, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->load(StorageSoa::Temp, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->load(StorageSoa::Output, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->load(StorageSoa::Immediate, - src->SrcRegister.Index, swizzle, indIdx); - } else { - fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); - return; - } - - inputs[i] = val; - } - - std::vector out(4); - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: { - out = instr->arl(inputs[0]); - } - break; - case TGSI_OPCODE_MOV: { - out = inputs[0]; - } - break; - case TGSI_OPCODE_LIT: { - } - break; - case TGSI_OPCODE_RCP: { - } - break; - case TGSI_OPCODE_RSQ: { - } - break; - case TGSI_OPCODE_EXP: - break; - case TGSI_OPCODE_LOG: - break; - case TGSI_OPCODE_MUL: { - out = instr->mul(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_ADD: { - out = instr->add(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP3: { - } - break; - case TGSI_OPCODE_DP4: { - } - break; - case TGSI_OPCODE_DST: { - } - break; - case TGSI_OPCODE_MIN: { - } - break; - case TGSI_OPCODE_MAX: { - } - break; - case TGSI_OPCODE_SLT: { - } - break; - case TGSI_OPCODE_SGE: { - } - break; - case TGSI_OPCODE_MAD: { - out = instr->madd(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SUB: { - } - break; - case TGSI_OPCODE_LERP: { - } - break; - case TGSI_OPCODE_CND: - break; - case TGSI_OPCODE_CND0: - break; - case TGSI_OPCODE_DOT2ADD: - break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { - } - break; - case TGSI_OPCODE_CLAMP: - break; - case TGSI_OPCODE_FLOOR: { - } - break; - case TGSI_OPCODE_ROUND: - break; - case TGSI_OPCODE_EXPBASE2: { - } - break; - case TGSI_OPCODE_LOGBASE2: { - } - break; - case TGSI_OPCODE_POWER: { - } - break; - case TGSI_OPCODE_CROSSPRODUCT: { - } - break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; - case TGSI_OPCODE_ABS: { - } - break; - case TGSI_OPCODE_RCC: - break; - case TGSI_OPCODE_DPH: { - } - break; - case TGSI_OPCODE_COS: { - } - break; - case TGSI_OPCODE_DDX: - break; - case TGSI_OPCODE_DDY: - break; - case TGSI_OPCODE_KILP: { - } - break; - case TGSI_OPCODE_PK2H: - break; - case TGSI_OPCODE_PK2US: - break; - case TGSI_OPCODE_PK4B: - break; - case TGSI_OPCODE_PK4UB: - break; - case TGSI_OPCODE_RFL: - break; - case TGSI_OPCODE_SEQ: - break; - case TGSI_OPCODE_SFL: - break; - case TGSI_OPCODE_SGT: { - } - break; - case TGSI_OPCODE_SIN: { - } - break; - case TGSI_OPCODE_SLE: - break; - case TGSI_OPCODE_SNE: - break; - case TGSI_OPCODE_STR: - break; - case TGSI_OPCODE_TEX: - break; - case TGSI_OPCODE_TXD: - break; - case TGSI_OPCODE_UP2H: - break; - case TGSI_OPCODE_UP2US: - break; - case TGSI_OPCODE_UP4B: - break; - case TGSI_OPCODE_UP4UB: - break; - case TGSI_OPCODE_X2D: - break; - case TGSI_OPCODE_ARA: - break; - case TGSI_OPCODE_ARR: - break; - case TGSI_OPCODE_BRA: - break; - case TGSI_OPCODE_CAL: { - } - break; - case TGSI_OPCODE_RET: { - } - break; - case TGSI_OPCODE_SSG: - break; - case TGSI_OPCODE_CMP: { - } - break; - case TGSI_OPCODE_SCS: { - } - break; - case TGSI_OPCODE_TXB: - break; - case TGSI_OPCODE_NRM: - break; - case TGSI_OPCODE_DIV: - break; - case TGSI_OPCODE_DP2: - break; - case TGSI_OPCODE_TXL: - break; - case TGSI_OPCODE_BRK: { - } - break; - case TGSI_OPCODE_IF: { - } - break; - case TGSI_OPCODE_LOOP: - break; - case TGSI_OPCODE_REP: - break; - case TGSI_OPCODE_ELSE: { - } - break; - case TGSI_OPCODE_ENDIF: { - } - break; - case TGSI_OPCODE_ENDLOOP: - break; - case TGSI_OPCODE_ENDREP: - break; - case TGSI_OPCODE_PUSHA: - break; - case TGSI_OPCODE_POPA: - break; - case TGSI_OPCODE_CEIL: - break; - case TGSI_OPCODE_I2F: - break; - case TGSI_OPCODE_NOT: - break; - case TGSI_OPCODE_TRUNC: { - } - break; - case TGSI_OPCODE_SHL: - break; - case TGSI_OPCODE_SHR: - break; - case TGSI_OPCODE_AND: - break; - case TGSI_OPCODE_OR: - break; - case TGSI_OPCODE_MOD: - break; - case TGSI_OPCODE_XOR: - break; - case TGSI_OPCODE_SAD: - break; - case TGSI_OPCODE_TXF: - break; - case TGSI_OPCODE_TXQ: - break; - case TGSI_OPCODE_CONT: - break; - case TGSI_OPCODE_EMIT: - break; - case TGSI_OPCODE_ENDPRIM: - break; - case TGSI_OPCODE_BGNLOOP2: { - } - break; - case TGSI_OPCODE_BGNSUB: { - } - break; - case TGSI_OPCODE_ENDLOOP2: { - } - break; - case TGSI_OPCODE_ENDSUB: { - } - break; - case TGSI_OPCODE_NOISE1: - break; - case TGSI_OPCODE_NOISE2: - break; - case TGSI_OPCODE_NOISE3: - break; - case TGSI_OPCODE_NOISE4: - break; - case TGSI_OPCODE_NOP: - break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; - case TGSI_OPCODE_NRM4: - break; - case TGSI_OPCODE_CALLNZ: - break; - case TGSI_OPCODE_IFC: - break; - case TGSI_OPCODE_BREAKC: - break; - case TGSI_OPCODE_KIL: - break; - case TGSI_OPCODE_END: - instr->end(); - return; - break; - default: - fprintf(stderr, "ERROR: Unknown opcode %d\n", - inst->Instruction.Opcode); - assert(0); - break; - } - - if (!out[0]) { - fprintf(stderr, "ERROR: unsupported opcode %d\n", - inst->Instruction.Opcode); - assert(!"Unsupported opcode"); - } - - /* store results */ - for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->store(StorageSoa::Output, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->store(StorageSoa::Temp, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->store(StorageSoa::Address, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else { - fprintf(stderr, "ERROR: unsupported LLVM destination!"); - assert(!"wrong destination"); - } - } -} - -llvm::Module * -tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) -{ - llvm::Module *mod = new Module("shader"); - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned instno = 0; - Function* shader = mod->getFunction("execute_shader"); - std::ostringstream stream; - if (ir->type == GALLIVM_VS) { - stream << "vs_shader"; - } else { - stream << "fs_shader"; - } - stream << ir->id; - std::string func_name = stream.str(); - shader->setName(func_name.c_str()); - - Function::arg_iterator args = shader->arg_begin(); - Value *ptr_INPUT = args++; - ptr_INPUT->setName("input"); - - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); - - tgsi_parse_init(&parse, tokens); - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - Storage storage(label_entry, ptr_INPUT); - Instructions instr(mod, shader, label_entry, &storage); - while(!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - translate_declaration(ir, mod, &storage, - &parse.FullToken.FullDeclaration, - &fd); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - translate_immediate(&storage, - &parse.FullToken.FullImmediate); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - translate_instruction(mod, &storage, &instr, - &parse.FullToken.FullInstruction, - &fi, instno); - ++instno; - break; - - default: - assert(0); - } - } - - tgsi_parse_free(&parse); - - ir->num_consts = storage.numConsts(); - return mod; -} - -llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, - const struct tgsi_token *tokens) -{ - llvm::Module *mod = new Module("shader"); - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned instno = 0; - std::ostringstream stream; - if (ir->type == GALLIVM_VS) { - stream << "vs_shader"; - } else { - stream << "fs_shader"; - } - //stream << ir->id; - std::string func_name = stream.str(); - Function *shader = llvm::cast(mod->getOrInsertFunction( - func_name.c_str(), - vertexShaderFunctionType())); - - Function::arg_iterator args = shader->arg_begin(); - Value *input = args++; - input->setName("inputs"); - Value *output = args++; - output->setName("outputs"); - Value *consts = args++; - consts->setName("consts"); - Value *temps = args++; - temps->setName("temps"); - - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); - - tgsi_parse_init(&parse, tokens); - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - - StorageSoa storage(label_entry, input, output, consts, temps); - InstructionsSoa instr(mod, shader, label_entry, &storage); - - while(!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - translate_declarationir(ir, mod, &storage, - &parse.FullToken.FullDeclaration, - &fd); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - translate_immediateir(&storage, - &parse.FullToken.FullImmediate); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - storage.declareImmediates(); - translate_instructionir(mod, &storage, &instr, - &parse.FullToken.FullInstruction, - &fi, instno); - ++instno; - break; - - default: - assert(0); - } - } - - tgsi_parse_free(&parse); - - return mod; -} diff --git a/src/gallium/aux/llvm/tgsitollvm.h b/src/gallium/aux/llvm/tgsitollvm.h deleted file mode 100644 index 7ada04d629..0000000000 --- a/src/gallium/aux/llvm/tgsitollvm.h +++ /dev/null @@ -1,20 +0,0 @@ -#ifndef TGSITOLLVM_H -#define TGSITOLLVM_H - - -namespace llvm { - class Module; -} - -struct gallivm_ir; -struct tgsi_token; - - -llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, - const struct tgsi_token *tokens); - - -llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, - const struct tgsi_token *tokens); - -#endif diff --git a/src/gallium/aux/pipebuffer/Makefile b/src/gallium/aux/pipebuffer/Makefile deleted file mode 100644 index 588629e870..0000000000 --- a/src/gallium/aux/pipebuffer/Makefile +++ /dev/null @@ -1,23 +0,0 @@ - -TOP = ../../../.. -include $(TOP)/configs/current - -LIBNAME = pipebuffer - -DRIVER_SOURCES = \ - pb_buffer_fenced.c \ - pb_buffer_malloc.c \ - pb_bufmgr_fenced.c \ - pb_bufmgr_mm.c \ - pb_bufmgr_pool.c \ - pb_winsys.c - -C_SOURCES = \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -include ../../Makefile.template - -symlinks: - diff --git a/src/gallium/aux/pipebuffer/linked_list.h b/src/gallium/aux/pipebuffer/linked_list.h deleted file mode 100644 index e99817fb13..0000000000 --- a/src/gallium/aux/pipebuffer/linked_list.h +++ /dev/null @@ -1,91 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND. USA. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - **************************************************************************/ - -/** - * \file - * List macros heavily inspired by the Linux kernel - * list handling. No list looping yet. - * - * Is not threadsafe, so common operations need to - * be protected using an external mutex. - */ - -#ifndef LINKED_LIST_H_ -#define LINKED_LIST_H_ - - -#include - - -struct list_head -{ - struct list_head *prev; - struct list_head *next; -}; - - -#define LIST_INITHEAD(__item) \ - do { \ - (__item)->prev = (__item); \ - (__item)->next = (__item); \ - } while (0) - -#define LIST_ADD(__item, __list) \ - do { \ - (__item)->prev = (__list); \ - (__item)->next = (__list)->next; \ - (__list)->next->prev = (__item); \ - (__list)->next = (__item); \ - } while (0) - -#define LIST_ADDTAIL(__item, __list) \ - do { \ - (__item)->next = (__list); \ - (__item)->prev = (__list)->prev; \ - (__list)->prev->next = (__item); \ - (__list)->prev = (__item); \ - } while(0) - -#define LIST_DEL(__item) \ - do { \ - (__item)->prev->next = (__item)->next; \ - (__item)->next->prev = (__item)->prev; \ - } while(0) - -#define LIST_DELINIT(__item) \ - do { \ - (__item)->prev->next = (__item)->next; \ - (__item)->next->prev = (__item)->prev; \ - (__item)->next = (__item); \ - (__item)->prev = (__item); \ - } while(0) - -#define LIST_ENTRY(__type, __item, __field) \ - ((__type *)(((char *)(__item)) - offsetof(__type, __field))) - - -#endif /*LINKED_LIST_H_*/ diff --git a/src/gallium/aux/pipebuffer/pb_buffer.h b/src/gallium/aux/pipebuffer/pb_buffer.h deleted file mode 100644 index 97beb5f72a..0000000000 --- a/src/gallium/aux/pipebuffer/pb_buffer.h +++ /dev/null @@ -1,202 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Generic code for buffers. - * - * Behind a pipe buffle handle there can be DMA buffers, client (or user) - * buffers, regular malloced buffers, etc. This file provides an abstract base - * buffer handle that allows the driver to cope with all those kinds of buffers - * in a more flexible way. - * - * There is no obligation of a winsys driver to use this library. And a pipe - * driver should be completly agnostic about it. - * - * \author José Fonseca - */ - -#ifndef PB_BUFFER_H_ -#define PB_BUFFER_H_ - - -#include "pipe/p_compiler.h" -#include "pipe/p_debug.h" -#include "pipe/p_state.h" -#include "pipe/p_inlines.h" - - -struct pb_vtbl; - -/** - * Buffer description. - * - * Used when allocating the buffer. - */ -struct pb_desc -{ - unsigned alignment; - unsigned usage; -}; - - -/** - * Base class for all pb_* buffers. - */ -struct pb_buffer -{ - struct pipe_buffer base; - - /** - * Pointer to the virtual function table. - * - * Avoid accessing this table directly. Use the inline functions below - * instead to avoid mistakes. - */ - const struct pb_vtbl *vtbl; -}; - - -/** - * Virtual function table for the buffer storage operations. - * - * Note that creation is not done through this table. - */ -struct pb_vtbl -{ - void (*destroy)( struct pb_buffer *buf ); - - /** - * Map the entire data store of a buffer object into the client's address. - * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. - */ - void *(*map)( struct pb_buffer *buf, - unsigned flags ); - - void (*unmap)( struct pb_buffer *buf ); - - /** - * Get the base buffer and the offset. - * - * A buffer can be subdivided in smaller buffers. This method should return - * the underlaying buffer, and the relative offset. - * - * Buffers without an underlaying base buffer should return themselves, with - * a zero offset. - * - * Note that this will increase the reference count of the base buffer. - */ - void (*get_base_buffer)( struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset ); -}; - - -static INLINE struct pipe_buffer * -pb_pipe_buffer( struct pb_buffer *pbuf ) -{ - assert(pbuf); - return &pbuf->base; -} - - -static INLINE struct pb_buffer * -pb_buffer( struct pipe_buffer *buf ) -{ - assert(buf); - /* Could add a magic cookie check on debug builds. - */ - return (struct pb_buffer *)buf; -} - - -/* Accessor functions for pb->vtbl: - */ -static INLINE void * -pb_map(struct pb_buffer *buf, - unsigned flags) -{ - assert(buf); - return buf->vtbl->map(buf, flags); -} - - -static INLINE void -pb_unmap(struct pb_buffer *buf) -{ - assert(buf); - buf->vtbl->unmap(buf); -} - - -static INLINE void -pb_get_base_buffer( struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset ) -{ - buf->vtbl->get_base_buffer(buf, base_buf, offset); -} - - -static INLINE void -pb_destroy(struct pb_buffer *buf) -{ - assert(buf); - buf->vtbl->destroy(buf); -} - - -/* XXX: thread safety issues! - */ -static INLINE void -pb_reference(struct pb_buffer **dst, - struct pb_buffer *src) -{ - if (src) - src->base.refcount++; - - if (*dst && --(*dst)->base.refcount == 0) - pb_destroy( *dst ); - - *dst = src; -} - - -/** - * Malloc-based buffer to store data that can't be used by the graphics - * hardware. - */ -struct pb_buffer * -pb_malloc_buffer_create(size_t size, - const struct pb_desc *desc); - - -void -pb_init_winsys(struct pipe_winsys *winsys); - - -#endif /*PB_BUFFER_H_*/ diff --git a/src/gallium/aux/pipebuffer/pb_buffer_fenced.c b/src/gallium/aux/pipebuffer/pb_buffer_fenced.c deleted file mode 100644 index f4fc3f6d71..0000000000 --- a/src/gallium/aux/pipebuffer/pb_buffer_fenced.c +++ /dev/null @@ -1,299 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Implementation of fenced buffers. - * - * \author José Fonseca - * \author Thomas Hellström - */ - - -#include "linked_list.h" - -#include "p_compiler.h" -#include "p_debug.h" -#include "p_winsys.h" -#include "p_thread.h" -#include "p_util.h" - -#include "pb_buffer.h" -#include "pb_buffer_fenced.h" - -#ifndef __MSC__ -#include -#endif - - -/** - * Convenience macro (type safe). - */ -#define SUPER(__derived) (&(__derived)->base) - - -struct fenced_buffer_list -{ - _glthread_Mutex mutex; - - struct pipe_winsys *winsys; - - size_t numDelayed; - size_t checkDelayed; - - struct list_head delayed; -}; - - -/** - * Wrapper around a pipe buffer which adds fencing and reference counting. - */ -struct fenced_buffer -{ - struct pb_buffer base; - - struct pb_buffer *buffer; - - struct pipe_fence_handle *fence; - - struct list_head head; - struct fenced_buffer_list *list; -}; - - -static INLINE struct fenced_buffer * -fenced_buffer(struct pb_buffer *buf) -{ - assert(buf); - assert(buf->vtbl == &fenced_buffer_vtbl); - return (struct fenced_buffer *)buf; -} - - - - -static void -_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, - int wait) -{ - struct pipe_winsys *winsys = fenced_list->winsys; - struct fenced_buffer *fenced_buf; - struct list_head *list, *prev; - int signaled = -1; - - list = fenced_list->delayed.next; - - if (fenced_list->numDelayed > 3) { - unsigned i; - - for (i = 0; i < fenced_list->numDelayed; i += 3) { - list = list->next; - } - } - - prev = list->prev; - for (; list != &fenced_list->delayed; list = prev, prev = list->prev) { - - fenced_buf = LIST_ENTRY(struct fenced_buffer, list, head); - - if (signaled != 0) { - if (wait) { - signaled = winsys->fence_finish(winsys, fenced_buf->fence, 0); - } - else { - signaled = winsys->fence_signalled(winsys, fenced_buf->fence, 0); - } - } - - if (signaled != 0) - /* XXX: we are assuming that buffers are freed in the same order they - * are fenced which may not always be true... - */ - break; - - winsys->fence_reference(winsys, &fenced_buf->fence, NULL); - - LIST_DEL(list); - fenced_list->numDelayed--; - - /* Do the delayed destroy: - */ - pb_reference(&fenced_buf->buffer, NULL); - FREE(fenced_buf); - } -} - - -static void -fenced_buffer_destroy(struct pb_buffer *buf) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - struct fenced_buffer_list *fenced_list = fenced_buf->list; - - if (fenced_buf->fence) { - LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); - fenced_list->numDelayed++; - } - else { - pb_reference(&fenced_buf->buffer, NULL); - FREE(fenced_buf); - } - - if ((fenced_list->numDelayed % fenced_list->checkDelayed) == 0) - _fenced_buffer_list_check_free(fenced_list, 0); -} - - -static void * -fenced_buffer_map(struct pb_buffer *buf, - unsigned flags) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - return pb_map(fenced_buf->buffer, flags); -} - - -static void -fenced_buffer_unmap(struct pb_buffer *buf) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - pb_unmap(fenced_buf->buffer); -} - - -static void -fenced_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - pb_get_base_buffer(fenced_buf->buffer, base_buf, offset); -} - - -const struct pb_vtbl -fenced_buffer_vtbl = { - fenced_buffer_destroy, - fenced_buffer_map, - fenced_buffer_unmap, - fenced_buffer_get_base_buffer -}; - - -struct pb_buffer * -fenced_buffer_create(struct fenced_buffer_list *fenced_list, - struct pb_buffer *buffer) -{ - struct fenced_buffer *buf; - - if(!buffer) - return NULL; - - buf = CALLOC_STRUCT(fenced_buffer); - if(!buf) - return NULL; - - buf->base.base.refcount = 1; - buf->base.base.alignment = buffer->base.alignment; - buf->base.base.usage = buffer->base.usage; - buf->base.base.size = buffer->base.size; - - buf->base.vtbl = &fenced_buffer_vtbl; - buf->buffer = buffer; - buf->list = fenced_list; - - return &buf->base; -} - - -void -buffer_fence(struct pb_buffer *buf, - struct pipe_fence_handle *fence) -{ - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - struct fenced_buffer_list *fenced_list = fenced_buf->list; - struct pipe_winsys *winsys = fenced_list->winsys; - - _glthread_LOCK_MUTEX(fenced_list->mutex); - winsys->fence_reference(winsys, &fenced_buf->fence, fence); - _glthread_UNLOCK_MUTEX(fenced_list->mutex); -} - - -struct fenced_buffer_list * -fenced_buffer_list_create(struct pipe_winsys *winsys) -{ - struct fenced_buffer_list *fenced_list; - - fenced_list = (struct fenced_buffer_list *)CALLOC(1, sizeof(*fenced_list)); - if (!fenced_list) - return NULL; - - fenced_list->winsys = winsys; - - LIST_INITHEAD(&fenced_list->delayed); - - fenced_list->numDelayed = 0; - - /* TODO: don't hard code this */ - fenced_list->checkDelayed = 5; - - _glthread_INIT_MUTEX(fenced_list->mutex); - - return fenced_list; -} - - -void -fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, - int wait) -{ - _glthread_LOCK_MUTEX(fenced_list->mutex); - _fenced_buffer_list_check_free(fenced_list, wait); - _glthread_UNLOCK_MUTEX(fenced_list->mutex); -} - - -void -fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) -{ - _glthread_LOCK_MUTEX(fenced_list->mutex); - - /* Wait on outstanding fences */ - while (fenced_list->numDelayed) { - _glthread_UNLOCK_MUTEX(fenced_list->mutex); - sched_yield(); - _fenced_buffer_list_check_free(fenced_list, 1); - _glthread_LOCK_MUTEX(fenced_list->mutex); - } - - _glthread_UNLOCK_MUTEX(fenced_list->mutex); - - FREE(fenced_list); -} - - diff --git a/src/gallium/aux/pipebuffer/pb_buffer_fenced.h b/src/gallium/aux/pipebuffer/pb_buffer_fenced.h deleted file mode 100644 index c40b9c75e1..0000000000 --- a/src/gallium/aux/pipebuffer/pb_buffer_fenced.h +++ /dev/null @@ -1,117 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Buffer fencing. - * - * "Fenced buffers" is actually a misnomer. They should be referred as - * "fenceable buffers", i.e, buffers that can be fenced, but I couldn't find - * the word "fenceable" in the dictionary. - * - * A "fenced buffer" is a decorator around a normal buffer, which adds two - * special properties: - * - the ability for the destruction to be delayed by a fence; - * - reference counting. - * - * Usually DMA buffers have a life-time that will extend the life-time of its - * handle. The end-of-life is dictated by the fence signalling. - * - * Between the handle's destruction, and the fence signalling, the buffer is - * stored in a fenced buffer list. - * - * \author José Fonseca - */ - -#ifndef PB_BUFFER_FENCED_H_ -#define PB_BUFFER_FENCED_H_ - - -#include "pipe/p_debug.h" - - -struct pipe_winsys; -struct pipe_buffer; -struct pipe_fence_handle; - - -/** - * List of buffers which are awaiting fence signalling. - */ -struct fenced_buffer_list; - - -/** - * The fenced buffer's virtual function table. - * - * NOTE: Made public for debugging purposes. - */ -extern const struct pb_vtbl fenced_buffer_vtbl; - - -/** - * Create a fenced buffer list. - * - * See also fenced_bufmgr_create for a more convenient way to use this. - */ -struct fenced_buffer_list * -fenced_buffer_list_create(struct pipe_winsys *winsys); - - -/** - * Walk the fenced buffer list to check and free signalled buffers. - */ -void -fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, - int wait); - -void -fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list); - - -/** - * Wrap a buffer in a fenced buffer. - * - * NOTE: this will not increase the buffer reference count. - */ -struct pb_buffer * -fenced_buffer_create(struct fenced_buffer_list *fenced, - struct pb_buffer *buffer); - - -/** - * Set a buffer's fence. - * - * NOTE: Although it takes a generic pb_buffer argument, it will fail - * on everything but buffers returned by fenced_buffer_create. - */ -void -buffer_fence(struct pb_buffer *buf, - struct pipe_fence_handle *fence); - - -#endif /*PB_BUFFER_FENCED_H_*/ diff --git a/src/gallium/aux/pipebuffer/pb_buffer_malloc.c b/src/gallium/aux/pipebuffer/pb_buffer_malloc.c deleted file mode 100644 index 9e8244f909..0000000000 --- a/src/gallium/aux/pipebuffer/pb_buffer_malloc.c +++ /dev/null @@ -1,127 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Implementation of malloc-based buffers to store data that can't be processed - * by the hardware. - * - * \author José Fonseca - */ - - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pb_buffer.h" - - -struct malloc_buffer -{ - struct pb_buffer base; - void *data; -}; - - -extern const struct pb_vtbl malloc_buffer_vtbl; - -static INLINE struct malloc_buffer * -malloc_buffer(struct pb_buffer *buf) -{ - assert(buf); - assert(buf->vtbl == &malloc_buffer_vtbl); - return (struct malloc_buffer *)buf; -} - - -static void -malloc_buffer_destroy(struct pb_buffer *buf) -{ - align_free(malloc_buffer(buf)->data); - FREE(buf); -} - - -static void * -malloc_buffer_map(struct pb_buffer *buf, - unsigned flags) -{ - return malloc_buffer(buf)->data; -} - - -static void -malloc_buffer_unmap(struct pb_buffer *buf) -{ - /* No-op */ -} - - -static void -malloc_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - *base_buf = buf; - *offset = 0; -} - - -const struct pb_vtbl -malloc_buffer_vtbl = { - malloc_buffer_destroy, - malloc_buffer_map, - malloc_buffer_unmap, - malloc_buffer_get_base_buffer -}; - - -struct pb_buffer * -pb_malloc_buffer_create(size_t size, - const struct pb_desc *desc) -{ - struct malloc_buffer *buf; - - /* TODO: do a single allocation */ - - buf = CALLOC_STRUCT(malloc_buffer); - if(!buf) - return NULL; - - buf->base.base.refcount = 1; - buf->base.base.alignment = desc->alignment; - buf->base.base.usage = desc->usage; - buf->base.base.size = size; - buf->base.vtbl = &malloc_buffer_vtbl; - - buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment); - if(!buf->data) { - align_free(buf); - return NULL; - } - - return &buf->base; -} diff --git a/src/gallium/aux/pipebuffer/pb_bufmgr.h b/src/gallium/aux/pipebuffer/pb_bufmgr.h deleted file mode 100644 index 1ddf784c97..0000000000 --- a/src/gallium/aux/pipebuffer/pb_bufmgr.h +++ /dev/null @@ -1,126 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Buffer management. - * - * A buffer manager does only one basic thing: it creates buffers. Actually, - * "buffer factory" would probably a more accurate description. - * - * You can chain buffer managers so that you can have a finer grained memory - * management and pooling. - * - * For example, for a simple batch buffer manager you would chain: - * - the native buffer manager, which provides DMA memory from the graphics - * memory space; - * - the pool buffer manager, which keep around a pool of equally sized buffers - * to avoid latency associated with the native buffer manager; - * - the fenced buffer manager, which will delay buffer destruction until the - * the moment the card finishing processing it. - * - * \author José Fonseca - */ - -#ifndef PB_BUFMGR_H_ -#define PB_BUFMGR_H_ - - -#include - - -struct pb_desc; -struct pipe_buffer; -struct pipe_winsys; - - -/** - * Abstract base class for all buffer managers. - */ -struct pb_manager -{ - /* XXX: we will likely need more allocation flags */ - struct pb_buffer * - (*create_buffer)( struct pb_manager *mgr, - size_t size, - const struct pb_desc *desc); - - void - (*destroy)( struct pb_manager *mgr ); -}; - - -/** - * Static buffer pool manager. - * - * Manages the allocation of equally sized buffers. It does so by allocating - * a single big buffer and divide it equally sized buffers. - * - * It is meant to manage the allocation of batch buffer pools. - */ -struct pb_manager * -pool_bufmgr_create(struct pb_manager *provider, - size_t n, size_t size, - const struct pb_desc *desc); - - -/** - * Wraper around the old memory manager. - * - * It managers buffers of different sizes. It does so by allocating a buffer - * with the size of the heap, and then using the old mm memory manager to manage - * that heap. - */ -struct pb_manager * -mm_bufmgr_create(struct pb_manager *provider, - size_t size, size_t align2); - -/** - * Same as mm_bufmgr_create. - * - * Buffer will be release when the manager is destroyed. - */ -struct pb_manager * -mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, - size_t size, size_t align2); - - -/** - * Fenced buffer manager. - * - * This manager is just meant for convenience. It wraps the buffers returned - * by another manager in fenced buffers, so that - * - * NOTE: the buffer manager that provides the buffers will be destroyed - * at the same time. - */ -struct pb_manager * -fenced_bufmgr_create(struct pb_manager *provider, - struct pipe_winsys *winsys); - - -#endif /*PB_BUFMGR_H_*/ diff --git a/src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c deleted file mode 100644 index c535d3276c..0000000000 --- a/src/gallium/aux/pipebuffer/pb_bufmgr_fenced.c +++ /dev/null @@ -1,131 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ - -/** - * \file - * A buffer manager that wraps buffers in fenced buffers. - * - * \author José Fonseca - */ - - -#include "p_debug.h" -#include "p_util.h" - -#include "pb_buffer.h" -#include "pb_buffer_fenced.h" -#include "pb_bufmgr.h" - - -struct fenced_pb_manager -{ - struct pb_manager base; - - struct pb_manager *provider; - - struct fenced_buffer_list *fenced_list; -}; - - -static INLINE struct fenced_pb_manager * -fenced_pb_manager(struct pb_manager *mgr) -{ - assert(mgr); - return (struct fenced_pb_manager *)mgr; -} - - -static struct pb_buffer * -fenced_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, - const struct pb_desc *desc) -{ - struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); - struct pb_buffer *buf; - struct pb_buffer *fenced_buf; - - /* check for free buffers before allocating new ones */ - fenced_buffer_list_check_free(fenced_mgr->fenced_list, 0); - - buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); - if(!buf) { - /* try harder to get a buffer */ - fenced_buffer_list_check_free(fenced_mgr->fenced_list, 1); - - buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); - if(!buf) { - /* give up */ - return NULL; - } - } - - fenced_buf = fenced_buffer_create(fenced_mgr->fenced_list, buf); - if(!fenced_buf) { - assert(buf->base.refcount == 1); - pb_destroy(buf); - } - - return fenced_buf; -} - - -static void -fenced_bufmgr_destroy(struct pb_manager *mgr) -{ - struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); - - fenced_buffer_list_destroy(fenced_mgr->fenced_list); - - fenced_mgr->provider->destroy(fenced_mgr->provider); - - FREE(fenced_mgr); -} - - -struct pb_manager * -fenced_bufmgr_create(struct pb_manager *provider, - struct pipe_winsys *winsys) -{ - struct fenced_pb_manager *fenced_mgr; - - fenced_mgr = (struct fenced_pb_manager *)CALLOC(1, sizeof(*fenced_mgr)); - if (!fenced_mgr) - return NULL; - - fenced_mgr->base.destroy = fenced_bufmgr_destroy; - fenced_mgr->base.create_buffer = fenced_bufmgr_create_buffer; - - fenced_mgr->provider = provider; - fenced_mgr->fenced_list = fenced_buffer_list_create(winsys); - if(!fenced_mgr->fenced_list) { - FREE(fenced_mgr); - return NULL; - } - - return &fenced_mgr->base; -} diff --git a/src/gallium/aux/pipebuffer/pb_bufmgr_mm.c b/src/gallium/aux/pipebuffer/pb_bufmgr_mm.c deleted file mode 100644 index 8b1b51c0e2..0000000000 --- a/src/gallium/aux/pipebuffer/pb_bufmgr_mm.c +++ /dev/null @@ -1,593 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * Copyright 1999 Wittawat Yamwong - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Buffer manager using the old texture memory manager. - * - * \author José Fonseca - */ - - -#include "linked_list.h" - -#include "p_defines.h" -#include "p_debug.h" -#include "p_thread.h" -#include "p_util.h" -#include "pb_buffer.h" -#include "pb_bufmgr.h" - - -/** - * Convenience macro (type safe). - */ -#define SUPER(__derived) (&(__derived)->base) - - -struct mem_block -{ - struct mem_block *next, *prev; - struct mem_block *next_free, *prev_free; - struct mem_block *heap; - int ofs, size; - unsigned int free:1; - unsigned int reserved:1; -}; - - -#ifdef DEBUG -/** - * For debugging purposes. - */ -static void -mmDumpMemInfo(const struct mem_block *heap) -{ - debug_printf("Memory heap %p:\n", (void *)heap); - if (heap == 0) { - debug_printf(" heap == 0\n"); - } else { - const struct mem_block *p; - - for(p = heap->next; p != heap; p = p->next) { - debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); - } - - debug_printf("\nFree list:\n"); - - for(p = heap->next_free; p != heap; p = p->next_free) { - debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); - } - - } - debug_printf("End of memory blocks\n"); -} -#endif - - -/** - * input: total size in bytes - * return: a heap pointer if OK, NULL if error - */ -static struct mem_block * -mmInit(int ofs, int size) -{ - struct mem_block *heap, *block; - - if (size <= 0) - return NULL; - - heap = CALLOC_STRUCT(mem_block); - if (!heap) - return NULL; - - block = CALLOC_STRUCT(mem_block); - if (!block) { - FREE(heap); - return NULL; - } - - heap->next = block; - heap->prev = block; - heap->next_free = block; - heap->prev_free = block; - - block->heap = heap; - block->next = heap; - block->prev = heap; - block->next_free = heap; - block->prev_free = heap; - - block->ofs = ofs; - block->size = size; - block->free = 1; - - return heap; -} - - -static struct mem_block * -SliceBlock(struct mem_block *p, - int startofs, int size, - int reserved, int alignment) -{ - struct mem_block *newblock; - - /* break left [p, newblock, p->next], then p = newblock */ - if (startofs > p->ofs) { - newblock = CALLOC_STRUCT(mem_block); - if (!newblock) - return NULL; - newblock->ofs = startofs; - newblock->size = p->size - (startofs - p->ofs); - newblock->free = 1; - newblock->heap = p->heap; - - newblock->next = p->next; - newblock->prev = p; - p->next->prev = newblock; - p->next = newblock; - - newblock->next_free = p->next_free; - newblock->prev_free = p; - p->next_free->prev_free = newblock; - p->next_free = newblock; - - p->size -= newblock->size; - p = newblock; - } - - /* break right, also [p, newblock, p->next] */ - if (size < p->size) { - newblock = CALLOC_STRUCT(mem_block); - if (!newblock) - return NULL; - newblock->ofs = startofs + size; - newblock->size = p->size - size; - newblock->free = 1; - newblock->heap = p->heap; - - newblock->next = p->next; - newblock->prev = p; - p->next->prev = newblock; - p->next = newblock; - - newblock->next_free = p->next_free; - newblock->prev_free = p; - p->next_free->prev_free = newblock; - p->next_free = newblock; - - p->size = size; - } - - /* p = middle block */ - p->free = 0; - - /* Remove p from the free list: - */ - p->next_free->prev_free = p->prev_free; - p->prev_free->next_free = p->next_free; - - p->next_free = 0; - p->prev_free = 0; - - p->reserved = reserved; - return p; -} - - -/** - * Allocate 'size' bytes with 2^align2 bytes alignment, - * restrict the search to free memory after 'startSearch' - * depth and back buffers should be in different 4mb banks - * to get better page hits if possible - * input: size = size of block - * align2 = 2^align2 bytes alignment - * startSearch = linear offset from start of heap to begin search - * return: pointer to the allocated block, 0 if error - */ -static struct mem_block * -mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) -{ - struct mem_block *p; - const int mask = (1 << align2)-1; - int startofs = 0; - int endofs; - - if (!heap || align2 < 0 || size <= 0) - return NULL; - - for (p = heap->next_free; p != heap; p = p->next_free) { - assert(p->free); - - startofs = (p->ofs + mask) & ~mask; - if ( startofs < startSearch ) { - startofs = startSearch; - } - endofs = startofs+size; - if (endofs <= (p->ofs+p->size)) - break; - } - - if (p == heap) - return NULL; - - assert(p->free); - p = SliceBlock(p,startofs,size,0,mask+1); - - return p; -} - - -#if 0 -/** - * Free block starts at offset - * input: pointer to a heap, start offset - * return: pointer to a block - */ -static struct mem_block * -mmFindBlock(struct mem_block *heap, int start) -{ - struct mem_block *p; - - for (p = heap->next; p != heap; p = p->next) { - if (p->ofs == start) - return p; - } - - return NULL; -} -#endif - - -static INLINE int -Join2Blocks(struct mem_block *p) -{ - /* XXX there should be some assertions here */ - - /* NOTE: heap->free == 0 */ - - if (p->free && p->next->free) { - struct mem_block *q = p->next; - - assert(p->ofs + p->size == q->ofs); - p->size += q->size; - - p->next = q->next; - q->next->prev = p; - - q->next_free->prev_free = q->prev_free; - q->prev_free->next_free = q->next_free; - - FREE(q); - return 1; - } - return 0; -} - - -/** - * Free block starts at offset - * input: pointer to a block - * return: 0 if OK, -1 if error - */ -static int -mmFreeMem(struct mem_block *b) -{ - if (!b) - return 0; - - if (b->free) { - debug_printf("block already free\n"); - return -1; - } - if (b->reserved) { - debug_printf("block is reserved\n"); - return -1; - } - - b->free = 1; - b->next_free = b->heap->next_free; - b->prev_free = b->heap; - b->next_free->prev_free = b; - b->prev_free->next_free = b; - - Join2Blocks(b); - if (b->prev != b->heap) - Join2Blocks(b->prev); - - return 0; -} - - -/** - * destroy MM - */ -static void -mmDestroy(struct mem_block *heap) -{ - struct mem_block *p; - - if (!heap) - return; - - for (p = heap->next; p != heap; ) { - struct mem_block *next = p->next; - FREE(p); - p = next; - } - - FREE(heap); -} - - -struct mm_pb_manager -{ - struct pb_manager base; - - _glthread_Mutex mutex; - - size_t size; - struct mem_block *heap; - - size_t align2; - - struct pb_buffer *buffer; - void *map; -}; - - -static INLINE struct mm_pb_manager * -mm_pb_manager(struct pb_manager *mgr) -{ - assert(mgr); - return (struct mm_pb_manager *)mgr; -} - - -struct mm_buffer -{ - struct pb_buffer base; - - struct mm_pb_manager *mgr; - - struct mem_block *block; -}; - - -static INLINE struct mm_buffer * -mm_buffer(struct pb_buffer *buf) -{ - assert(buf); - return (struct mm_buffer *)buf; -} - - -static void -mm_buffer_destroy(struct pb_buffer *buf) -{ - struct mm_buffer *mm_buf = mm_buffer(buf); - struct mm_pb_manager *mm = mm_buf->mgr; - - assert(buf->base.refcount == 0); - - _glthread_LOCK_MUTEX(mm->mutex); - mmFreeMem(mm_buf->block); - FREE(buf); - _glthread_UNLOCK_MUTEX(mm->mutex); -} - - -static void * -mm_buffer_map(struct pb_buffer *buf, - unsigned flags) -{ - struct mm_buffer *mm_buf = mm_buffer(buf); - struct mm_pb_manager *mm = mm_buf->mgr; - - return (unsigned char *) mm->map + mm_buf->block->ofs; -} - - -static void -mm_buffer_unmap(struct pb_buffer *buf) -{ - /* No-op */ -} - - -static void -mm_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - struct mm_buffer *mm_buf = mm_buffer(buf); - struct mm_pb_manager *mm = mm_buf->mgr; - pb_get_base_buffer(mm->buffer, base_buf, offset); - *offset += mm_buf->block->ofs; -} - - -static const struct pb_vtbl -mm_buffer_vtbl = { - mm_buffer_destroy, - mm_buffer_map, - mm_buffer_unmap, - mm_buffer_get_base_buffer -}; - - -static struct pb_buffer * -mm_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, - const struct pb_desc *desc) -{ - struct mm_pb_manager *mm = mm_pb_manager(mgr); - struct mm_buffer *mm_buf; - - /* We don't handle alignments larger then the one initially setup */ - assert(desc->alignment % (1 << mm->align2) == 0); - if(desc->alignment % (1 << mm->align2)) - return NULL; - - _glthread_LOCK_MUTEX(mm->mutex); - - mm_buf = CALLOC_STRUCT(mm_buffer); - if (!mm_buf) { - _glthread_UNLOCK_MUTEX(mm->mutex); - return NULL; - } - - mm_buf->base.base.refcount = 1; - mm_buf->base.base.alignment = desc->alignment; - mm_buf->base.base.usage = desc->usage; - mm_buf->base.base.size = size; - - mm_buf->base.vtbl = &mm_buffer_vtbl; - - mm_buf->mgr = mm; - - mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); - if(!mm_buf->block) { - debug_printf("warning: heap full\n"); -#if 0 - mmDumpMemInfo(mm->heap); -#endif - - mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); - if(!mm_buf->block) { - assert(0); - FREE(mm_buf); - _glthread_UNLOCK_MUTEX(mm->mutex); - return NULL; - } - } - - /* Some sanity checks */ - assert(0 <= mm_buf->block->ofs && mm_buf->block->ofs < mm->size); - assert(size <= mm_buf->block->size && mm_buf->block->ofs + mm_buf->block->size <= mm->size); - - _glthread_UNLOCK_MUTEX(mm->mutex); - return SUPER(mm_buf); -} - - -static void -mm_bufmgr_destroy(struct pb_manager *mgr) -{ - struct mm_pb_manager *mm = mm_pb_manager(mgr); - - _glthread_LOCK_MUTEX(mm->mutex); - - mmDestroy(mm->heap); - - pb_unmap(mm->buffer); - pb_reference(&mm->buffer, NULL); - - _glthread_UNLOCK_MUTEX(mm->mutex); - - FREE(mgr); -} - - -struct pb_manager * -mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, - size_t size, size_t align2) -{ - struct mm_pb_manager *mm; - - if(!buffer) - return NULL; - - mm = CALLOC_STRUCT(mm_pb_manager); - if (!mm) - return NULL; - - mm->base.create_buffer = mm_bufmgr_create_buffer; - mm->base.destroy = mm_bufmgr_destroy; - - mm->size = size; - mm->align2 = align2; /* 64-byte alignment */ - - _glthread_INIT_MUTEX(mm->mutex); - - mm->buffer = buffer; - - mm->map = pb_map(mm->buffer, - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE); - if(!mm->map) - goto failure; - - mm->heap = mmInit(0, size); - if (!mm->heap) - goto failure; - - return SUPER(mm); - -failure: -if(mm->heap) - mmDestroy(mm->heap); - if(mm->map) - pb_unmap(mm->buffer); - if(mm) - FREE(mm); - return NULL; -} - - -struct pb_manager * -mm_bufmgr_create(struct pb_manager *provider, - size_t size, size_t align2) -{ - struct pb_buffer *buffer; - struct pb_manager *mgr; - struct pb_desc desc; - - assert(provider); - assert(provider->create_buffer); - - memset(&desc, 0, sizeof(desc)); - desc.alignment = 1 << align2; - - buffer = provider->create_buffer(provider, size, &desc); - if (!buffer) - return NULL; - - mgr = mm_bufmgr_create_from_buffer(buffer, size, align2); - if (!mgr) { - pb_reference(&buffer, NULL); - return NULL; - } - - return mgr; -} diff --git a/src/gallium/aux/pipebuffer/pb_bufmgr_pool.c b/src/gallium/aux/pipebuffer/pb_bufmgr_pool.c deleted file mode 100644 index 04477a865a..0000000000 --- a/src/gallium/aux/pipebuffer/pb_bufmgr_pool.c +++ /dev/null @@ -1,288 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ - -/** - * \file - * Batch buffer pool management. - * - * \author José Fonseca - * \author Thomas Hellström - */ - - -#include "linked_list.h" - -#include "p_compiler.h" -#include "p_debug.h" -#include "p_thread.h" -#include "p_defines.h" -#include "p_util.h" - -#include "pb_buffer.h" -#include "pb_bufmgr.h" - - -/** - * Convenience macro (type safe). - */ -#define SUPER(__derived) (&(__derived)->base) - - -struct pool_pb_manager -{ - struct pb_manager base; - - _glthread_Mutex mutex; - - size_t bufSize; - size_t bufAlign; - - size_t numFree; - size_t numTot; - - struct list_head free; - - struct pb_buffer *buffer; - void *map; - - struct pool_buffer *bufs; -}; - - -static INLINE struct pool_pb_manager * -pool_pb_manager(struct pb_manager *mgr) -{ - assert(mgr); - return (struct pool_pb_manager *)mgr; -} - - -struct pool_buffer -{ - struct pb_buffer base; - - struct pool_pb_manager *mgr; - - struct list_head head; - - size_t start; -}; - - -static INLINE struct pool_buffer * -pool_buffer(struct pb_buffer *buf) -{ - assert(buf); - return (struct pool_buffer *)buf; -} - - - -static void -pool_buffer_destroy(struct pb_buffer *buf) -{ - struct pool_buffer *pool_buf = pool_buffer(buf); - struct pool_pb_manager *pool = pool_buf->mgr; - - assert(pool_buf->base.base.refcount == 0); - - _glthread_LOCK_MUTEX(pool->mutex); - LIST_ADD(&pool_buf->head, &pool->free); - pool->numFree++; - _glthread_UNLOCK_MUTEX(pool->mutex); -} - - -static void * -pool_buffer_map(struct pb_buffer *buf, unsigned flags) -{ - struct pool_buffer *pool_buf = pool_buffer(buf); - struct pool_pb_manager *pool = pool_buf->mgr; - void *map; - - _glthread_LOCK_MUTEX(pool->mutex); - map = (unsigned char *) pool->map + pool_buf->start; - _glthread_UNLOCK_MUTEX(pool->mutex); - return map; -} - - -static void -pool_buffer_unmap(struct pb_buffer *buf) -{ - /* No-op */ -} - - -static void -pool_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - struct pool_buffer *pool_buf = pool_buffer(buf); - struct pool_pb_manager *pool = pool_buf->mgr; - pb_get_base_buffer(pool->buffer, base_buf, offset); - *offset += pool_buf->start; -} - - -static const struct pb_vtbl -pool_buffer_vtbl = { - pool_buffer_destroy, - pool_buffer_map, - pool_buffer_unmap, - pool_buffer_get_base_buffer -}; - - -static struct pb_buffer * -pool_bufmgr_create_buffer(struct pb_manager *mgr, - size_t size, - const struct pb_desc *desc) -{ - struct pool_pb_manager *pool = pool_pb_manager(mgr); - struct pool_buffer *pool_buf; - struct list_head *item; - - assert(size == pool->bufSize); - assert(pool->bufAlign % desc->alignment == 0); - - _glthread_LOCK_MUTEX(pool->mutex); - - if (pool->numFree == 0) { - _glthread_UNLOCK_MUTEX(pool->mutex); - debug_printf("warning: out of fixed size buffer objects\n"); - return NULL; - } - - item = pool->free.next; - - if (item == &pool->free) { - _glthread_UNLOCK_MUTEX(pool->mutex); - debug_printf("error: fixed size buffer pool corruption\n"); - return NULL; - } - - LIST_DEL(item); - --pool->numFree; - - _glthread_UNLOCK_MUTEX(pool->mutex); - - pool_buf = LIST_ENTRY(struct pool_buffer, item, head); - assert(pool_buf->base.base.refcount == 0); - pool_buf->base.base.refcount = 1; - pool_buf->base.base.alignment = desc->alignment; - pool_buf->base.base.usage = desc->usage; - - return SUPER(pool_buf); -} - - -static void -pool_bufmgr_destroy(struct pb_manager *mgr) -{ - struct pool_pb_manager *pool = pool_pb_manager(mgr); - _glthread_LOCK_MUTEX(pool->mutex); - - FREE(pool->bufs); - - pb_unmap(pool->buffer); - pb_reference(&pool->buffer, NULL); - - _glthread_UNLOCK_MUTEX(pool->mutex); - - FREE(mgr); -} - - -struct pb_manager * -pool_bufmgr_create(struct pb_manager *provider, - size_t numBufs, - size_t bufSize, - const struct pb_desc *desc) -{ - struct pool_pb_manager *pool; - struct pool_buffer *pool_buf; - size_t i; - - pool = (struct pool_pb_manager *)CALLOC(1, sizeof(*pool)); - if (!pool) - return NULL; - - pool->base.destroy = pool_bufmgr_destroy; - pool->base.create_buffer = pool_bufmgr_create_buffer; - - LIST_INITHEAD(&pool->free); - - pool->numTot = numBufs; - pool->numFree = numBufs; - pool->bufSize = bufSize; - pool->bufAlign = desc->alignment; - - _glthread_INIT_MUTEX(pool->mutex); - - pool->buffer = provider->create_buffer(provider, numBufs*bufSize, desc); - if (!pool->buffer) - goto failure; - - pool->map = pb_map(pool->buffer, - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE); - if(!pool->map) - goto failure; - - pool->bufs = (struct pool_buffer *)CALLOC(numBufs, sizeof(*pool->bufs)); - if (!pool->bufs) - goto failure; - - pool_buf = pool->bufs; - for (i = 0; i < numBufs; ++i) { - pool_buf->base.base.refcount = 0; - pool_buf->base.base.alignment = 0; - pool_buf->base.base.usage = 0; - pool_buf->base.base.size = bufSize; - pool_buf->base.vtbl = &pool_buffer_vtbl; - pool_buf->mgr = pool; - pool_buf->start = i * bufSize; - LIST_ADDTAIL(&pool_buf->head, &pool->free); - pool_buf++; - } - - return SUPER(pool); - -failure: - if(pool->bufs) - FREE(pool->bufs); - if(pool->map) - pb_unmap(pool->buffer); - if(pool->buffer) - pb_reference(&pool->buffer, NULL); - if(pool) - FREE(pool); - return NULL; -} diff --git a/src/gallium/aux/pipebuffer/pb_winsys.c b/src/gallium/aux/pipebuffer/pb_winsys.c deleted file mode 100644 index 978944091f..0000000000 --- a/src/gallium/aux/pipebuffer/pb_winsys.c +++ /dev/null @@ -1,170 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Implementation of client buffer (also designated as "user buffers"), which - * are just state-tracker owned data masqueraded as buffers. - * - * \author José Fonseca - */ - - -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" - -#include "pb_buffer.h" - - -/** - * User buffers are special buffers that initially reference memory - * held by the user but which may if necessary copy that memory into - * device memory behind the scenes, for submission to hardware. - * - * These are particularly useful when the referenced data is never - * submitted to hardware at all, in the particular case of software - * vertex processing. - */ -struct pb_user_buffer -{ - struct pb_buffer base; - void *data; -}; - - -extern const struct pb_vtbl pb_user_buffer_vtbl; - - -static INLINE struct pb_user_buffer * -pb_user_buffer(struct pb_buffer *buf) -{ - assert(buf); - assert(buf->vtbl == &pb_user_buffer_vtbl); - return (struct pb_user_buffer *)buf; -} - - -static void -pb_user_buffer_destroy(struct pb_buffer *buf) -{ - assert(buf); - FREE(buf); -} - - -static void * -pb_user_buffer_map(struct pb_buffer *buf, - unsigned flags) -{ - return pb_user_buffer(buf)->data; -} - - -static void -pb_user_buffer_unmap(struct pb_buffer *buf) -{ - /* No-op */ -} - - -static void -pb_user_buffer_get_base_buffer(struct pb_buffer *buf, - struct pb_buffer **base_buf, - unsigned *offset) -{ - *base_buf = buf; - *offset = 0; -} - - -const struct pb_vtbl -pb_user_buffer_vtbl = { - pb_user_buffer_destroy, - pb_user_buffer_map, - pb_user_buffer_unmap, - pb_user_buffer_get_base_buffer -}; - - -static struct pipe_buffer * -pb_winsys_user_buffer_create(struct pipe_winsys *winsys, - void *data, - unsigned bytes) -{ - struct pb_user_buffer *buf = CALLOC_STRUCT(pb_user_buffer); - - if(!buf) - return NULL; - - buf->base.base.refcount = 1; - buf->base.base.size = bytes; - buf->base.base.alignment = 0; - buf->base.base.usage = 0; - - buf->base.vtbl = &pb_user_buffer_vtbl; - buf->data = data; - - return &buf->base.base; -} - - -static void * -pb_winsys_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned flags) -{ - (void)winsys; - return pb_map(pb_buffer(buf), flags); -} - - -static void -pb_winsys_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - (void)winsys; - pb_unmap(pb_buffer(buf)); -} - - -static void -pb_winsys_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - (void)winsys; - pb_destroy(pb_buffer(buf)); -} - - -void -pb_init_winsys(struct pipe_winsys *winsys) -{ - winsys->user_buffer_create = pb_winsys_user_buffer_create; - winsys->buffer_map = pb_winsys_buffer_map; - winsys->buffer_unmap = pb_winsys_buffer_unmap; - winsys->buffer_destroy = pb_winsys_buffer_destroy; -} diff --git a/src/gallium/aux/tgsi/Makefile b/src/gallium/aux/tgsi/Makefile deleted file mode 100644 index 12a8bd0409..0000000000 --- a/src/gallium/aux/tgsi/Makefile +++ /dev/null @@ -1,3 +0,0 @@ -default: - cd ../.. ; make - diff --git a/src/gallium/aux/tgsi/exec/Makefile b/src/gallium/aux/tgsi/exec/Makefile deleted file mode 100644 index eb8b14e0e8..0000000000 --- a/src/gallium/aux/tgsi/exec/Makefile +++ /dev/null @@ -1,3 +0,0 @@ -default: - cd ../../.. ; make - diff --git a/src/gallium/aux/tgsi/exec/tgsi_exec.c b/src/gallium/aux/tgsi/exec/tgsi_exec.c deleted file mode 100644 index a8f64c2287..0000000000 --- a/src/gallium/aux/tgsi/exec/tgsi_exec.c +++ /dev/null @@ -1,2485 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * TGSI interpretor/executor. - * - * Flow control information: - * - * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) - * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special - * care since a condition may be true for some quad components but false - * for other components. - * - * We basically execute all statements (even if they're in the part of - * an IF/ELSE clause that's "not taken") and use a special mask to - * control writing to destination registers. This is the ExecMask. - * See store_dest(). - * - * The ExecMask is computed from three other masks (CondMask, LoopMask and - * ContMask) which are controlled by the flow control instructions (namely: - * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). - * - * - * Authors: - * Michal Krol - * Brian Paul - */ - -#include "pipe/p_compiler.h" -#include "pipe/p_state.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi_exec.h" - -#define TILE_TOP_LEFT 0 -#define TILE_TOP_RIGHT 1 -#define TILE_BOTTOM_LEFT 2 -#define TILE_BOTTOM_RIGHT 3 - -/* - * Shorthand locations of various utility registers (_I = Index, _C = Channel) - */ -#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I -#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C -#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I -#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C -#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I -#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C -#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I -#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C -#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I -#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C -#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I -#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C -#define TEMP_128_I TGSI_EXEC_TEMP_128_I -#define TEMP_128_C TGSI_EXEC_TEMP_128_C -#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I -#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C -#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I -#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C -#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I -#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C -#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I -#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C -#define TEMP_R0 TGSI_EXEC_TEMP_R0 - -#define FOR_EACH_CHANNEL(CHAN)\ - for (CHAN = 0; CHAN < 4; CHAN++) - -#define IS_CHANNEL_ENABLED(INST, CHAN)\ - ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) - -#define IS_CHANNEL_ENABLED2(INST, CHAN)\ - ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) - -#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED( INST, CHAN )) - -#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED2( INST, CHAN )) - - -/** The execution mask depends on the conditional mask and the loop mask */ -#define UPDATE_EXEC_MASK(MACH) \ - MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask - - -#define CHAN_X 0 -#define CHAN_Y 1 -#define CHAN_Z 2 -#define CHAN_W 3 - - - -static void -tgsi_exec_prepare( struct tgsi_exec_machine *mach ) -{ - struct tgsi_exec_labels *labels = &mach->Labels; - struct tgsi_parse_context parse; - struct tgsi_full_instruction *instructions; - struct tgsi_full_declaration *declarations; - uint maxInstructions = 10, numInstructions = 0; - uint maxDeclarations = 10, numDeclarations = 0; - uint k; - uint instno = 0; - - mach->ImmLimit = 0; - labels->count = 0; - - declarations = (struct tgsi_full_declaration *) - MALLOC( maxDeclarations * sizeof(struct tgsi_full_declaration) ); - - instructions = (struct tgsi_full_instruction *) - MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) ); - - k = tgsi_parse_init( &parse, mach->Tokens ); - if (k != TGSI_PARSE_OK) { - debug_printf("Problem parsing!\n"); - return; - } - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - uint pointer = parse.Position; - uint i; - - tgsi_parse_token( &parse ); - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - /* save expanded declaration */ - if (numDeclarations == maxDeclarations) { - declarations = REALLOC(declarations, - maxDeclarations - * sizeof(struct tgsi_full_declaration), - (maxDeclarations + 10) - * sizeof(struct tgsi_full_declaration)); - maxDeclarations += 10; - } - memcpy(declarations + numDeclarations, - &parse.FullToken.FullDeclaration, - sizeof(declarations[0])); - numDeclarations++; - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; - assert( size % 4 == 0 ); - assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES ); - - for( i = 0; i < size; i++ ) { - mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; - } - mach->ImmLimit += size / 4; - } - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - assert( labels->count < 128 ); - - labels->labels[labels->count][0] = instno; - labels->labels[labels->count][1] = pointer; - labels->count++; - - /* save expanded instruction */ - if (numInstructions == maxInstructions) { - instructions = REALLOC(instructions, - maxInstructions - * sizeof(struct tgsi_full_instruction), - (maxInstructions + 10) - * sizeof(struct tgsi_full_instruction)); - maxInstructions += 10; - } - memcpy(instructions + numInstructions, - &parse.FullToken.FullInstruction, - sizeof(instructions[0])); - numInstructions++; - break; - - default: - assert( 0 ); - } - } - tgsi_parse_free (&parse); - - if (mach->Declarations) { - FREE( mach->Declarations ); - } - mach->Declarations = declarations; - mach->NumDeclarations = numDeclarations; - - if (mach->Instructions) { - FREE( mach->Instructions ); - } - mach->Instructions = instructions; - mach->NumInstructions = numInstructions; -} - - -/** - * Initialize machine state by expanding tokens to full instructions, - * allocating temporary storage, setting up constants, etc. - * After this, we can call tgsi_exec_machine_run() many times. - */ -void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach, - const struct tgsi_token *tokens, - uint numSamplers, - struct tgsi_sampler *samplers) -{ - uint i, k; - struct tgsi_parse_context parse; - -#if 0 - tgsi_dump(tokens, 0); -#endif - - mach->Tokens = tokens; - - mach->Samplers = samplers; - - k = tgsi_parse_init (&parse, mach->Tokens); - if (k != TGSI_PARSE_OK) { - debug_printf( "Problem parsing!\n" ); - return; - } - - mach->Processor = parse.FullHeader.Processor.Processor; - tgsi_parse_free (&parse); - - mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); - mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; - - /* Setup constants. */ - for( i = 0; i < 4; i++ ) { - mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; - mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; - mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; - mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; - mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; - mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; - mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; - mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; - } - - tgsi_exec_prepare( mach ); -} - - -void -tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach) -{ - if (mach->Instructions) { - FREE(mach->Instructions); - mach->Instructions = NULL; - mach->NumInstructions = 0; - } - if (mach->Declarations) { - FREE(mach->Declarations); - mach->Declarations = NULL; - mach->NumDeclarations = 0; - } -} - - -static void -micro_abs( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) fabs( (double) src->f[0] ); - dst->f[1] = (float) fabs( (double) src->f[1] ); - dst->f[2] = (float) fabs( (double) src->f[2] ); - dst->f[3] = (float) fabs( (double) src->f[3] ); -} - -static void -micro_add( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] + src1->f[0]; - dst->f[1] = src0->f[1] + src1->f[1]; - dst->f[2] = src0->f[2] + src1->f[2]; - dst->f[3] = src0->f[3] + src1->f[3]; -} - -static void -micro_iadd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] + src1->i[0]; - dst->i[1] = src0->i[1] + src1->i[1]; - dst->i[2] = src0->i[2] + src1->i[2]; - dst->i[3] = src0->i[3] + src1->i[3]; -} - -static void -micro_and( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] & src1->u[0]; - dst->u[1] = src0->u[1] & src1->u[1]; - dst->u[2] = src0->u[2] & src1->u[2]; - dst->u[3] = src0->u[3] & src1->u[3]; -} - -static void -micro_ceil( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) ceil( (double) src->f[0] ); - dst->f[1] = (float) ceil( (double) src->f[1] ); - dst->f[2] = (float) ceil( (double) src->f[2] ); - dst->f[3] = (float) ceil( (double) src->f[3] ); -} - -static void -micro_cos( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) cos( (double) src->f[0] ); - dst->f[1] = (float) cos( (double) src->f[1] ); - dst->f[2] = (float) cos( (double) src->f[2] ); - dst->f[3] = (float) cos( (double) src->f[3] ); -} - -static void -micro_ddx( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_ddy( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_div( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] / src1->f[0]; - dst->f[1] = src0->f[1] / src1->f[1]; - dst->f[2] = src0->f[2] / src1->f[2]; - dst->f[3] = src0->f[3] / src1->f[3]; -} - -static void -micro_udiv( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] / src1->u[0]; - dst->u[1] = src0->u[1] / src1->u[1]; - dst->u[2] = src0->u[2] / src1->u[2]; - dst->u[3] = src0->u[3] / src1->u[3]; -} - -static void -micro_eq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_ieq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; -} - -static void -micro_exp2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src) -{ - dst->f[0] = (float) pow( 2.0, (double) src->f[0] ); - dst->f[1] = (float) pow( 2.0, (double) src->f[1] ); - dst->f[2] = (float) pow( 2.0, (double) src->f[2] ); - dst->f[3] = (float) pow( 2.0, (double) src->f[3] ); -} - -static void -micro_f2it( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->i[0] = (int) src->f[0]; - dst->i[1] = (int) src->f[1]; - dst->i[2] = (int) src->f[2]; - dst->i[3] = (int) src->f[3]; -} - -static void -micro_f2ut( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = (uint) src->f[0]; - dst->u[1] = (uint) src->f[1]; - dst->u[2] = (uint) src->f[2]; - dst->u[3] = (uint) src->f[3]; -} - -static void -micro_flr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) floor( (double) src->f[0] ); - dst->f[1] = (float) floor( (double) src->f[1] ); - dst->f[2] = (float) floor( (double) src->f[2] ); - dst->f[3] = (float) floor( (double) src->f[3] ); -} - -static void -micro_frc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = src->f[0] - (float) floor( (double) src->f[0] ); - dst->f[1] = src->f[1] - (float) floor( (double) src->f[1] ); - dst->f[2] = src->f[2] - (float) floor( (double) src->f[2] ); - dst->f[3] = src->f[3] - (float) floor( (double) src->f[3] ); -} - -static void -micro_ge( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_i2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) src->i[0]; - dst->f[1] = (float) src->i[1]; - dst->f[2] = (float) src->i[2]; - dst->f[3] = (float) src->i[3]; -} - -static void -micro_lg2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) log( (double) src->f[0] ) * 1.442695f; - dst->f[1] = (float) log( (double) src->f[1] ) * 1.442695f; - dst->f[2] = (float) log( (double) src->f[2] ) * 1.442695f; - dst->f[3] = (float) log( (double) src->f[3] ) * 1.442695f; -} - -static void -micro_lt( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_ilt( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; -} - -static void -micro_ult( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; -} - -static void -micro_max( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; -} - -static void -micro_imax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; -} - -static void -micro_umax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; -} - -static void -micro_min( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; -} - -static void -micro_imin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; -} - -static void -micro_umin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; -} - -static void -micro_umod( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] % src1->u[0]; - dst->u[1] = src0->u[1] % src1->u[1]; - dst->u[2] = src0->u[2] % src1->u[2]; - dst->u[3] = src0->u[3] % src1->u[3]; -} - -static void -micro_mul( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] * src1->f[0]; - dst->f[1] = src0->f[1] * src1->f[1]; - dst->f[2] = src0->f[2] * src1->f[2]; - dst->f[3] = src0->f[3] * src1->f[3]; -} - -static void -micro_imul( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] * src1->i[0]; - dst->i[1] = src0->i[1] * src1->i[1]; - dst->i[2] = src0->i[2] * src1->i[2]; - dst->i[3] = src0->i[3] * src1->i[3]; -} - -static void -micro_imul64( - union tgsi_exec_channel *dst0, - union tgsi_exec_channel *dst1, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst1->i[0] = src0->i[0] * src1->i[0]; - dst1->i[1] = src0->i[1] * src1->i[1]; - dst1->i[2] = src0->i[2] * src1->i[2]; - dst1->i[3] = src0->i[3] * src1->i[3]; - dst0->i[0] = 0; - dst0->i[1] = 0; - dst0->i[2] = 0; - dst0->i[3] = 0; -} - -static void -micro_umul64( - union tgsi_exec_channel *dst0, - union tgsi_exec_channel *dst1, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst1->u[0] = src0->u[0] * src1->u[0]; - dst1->u[1] = src0->u[1] * src1->u[1]; - dst1->u[2] = src0->u[2] * src1->u[2]; - dst1->u[3] = src0->u[3] * src1->u[3]; - dst0->u[0] = 0; - dst0->u[1] = 0; - dst0->u[2] = 0; - dst0->u[3] = 0; -} - -static void -micro_movc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2 ) -{ - dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; - dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; - dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; - dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; -} - -static void -micro_neg( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = -src->f[0]; - dst->f[1] = -src->f[1]; - dst->f[2] = -src->f[2]; - dst->f[3] = -src->f[3]; -} - -static void -micro_ineg( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->i[0] = -src->i[0]; - dst->i[1] = -src->i[1]; - dst->i[2] = -src->i[2]; - dst->i[3] = -src->i[3]; -} - -static void -micro_not( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = ~src->u[0]; - dst->u[1] = ~src->u[1]; - dst->u[2] = ~src->u[2]; - dst->u[3] = ~src->u[3]; -} - -static void -micro_or( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] | src1->u[0]; - dst->u[1] = src0->u[1] | src1->u[1]; - dst->u[2] = src0->u[2] | src1->u[2]; - dst->u[3] = src0->u[3] | src1->u[3]; -} - -static void -micro_pow( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = (float) pow( (double) src0->f[0], (double) src1->f[0] ); - dst->f[1] = (float) pow( (double) src0->f[1], (double) src1->f[1] ); - dst->f[2] = (float) pow( (double) src0->f[2], (double) src1->f[2] ); - dst->f[3] = (float) pow( (double) src0->f[3], (double) src1->f[3] ); -} - -static void -micro_rnd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) floor( (double) (src->f[0] + 0.5f) ); - dst->f[1] = (float) floor( (double) (src->f[1] + 0.5f) ); - dst->f[2] = (float) floor( (double) (src->f[2] + 0.5f) ); - dst->f[3] = (float) floor( (double) (src->f[3] + 0.5f) ); -} - -static void -micro_shl( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] << src1->i[0]; - dst->i[1] = src0->i[1] << src1->i[1]; - dst->i[2] = src0->i[2] << src1->i[2]; - dst->i[3] = src0->i[3] << src1->i[3]; -} - -static void -micro_ishr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] >> src1->i[0]; - dst->i[1] = src0->i[1] >> src1->i[1]; - dst->i[2] = src0->i[2] >> src1->i[2]; - dst->i[3] = src0->i[3] >> src1->i[3]; -} - -static void -micro_trunc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0 ) -{ - dst->f[0] = (float) (int) src0->f[0]; - dst->f[1] = (float) (int) src0->f[1]; - dst->f[2] = (float) (int) src0->f[2]; - dst->f[3] = (float) (int) src0->f[3]; -} - -static void -micro_ushr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] >> src1->u[0]; - dst->u[1] = src0->u[1] >> src1->u[1]; - dst->u[2] = src0->u[2] >> src1->u[2]; - dst->u[3] = src0->u[3] >> src1->u[3]; -} - -static void -micro_sin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) sin( (double) src->f[0] ); - dst->f[1] = (float) sin( (double) src->f[1] ); - dst->f[2] = (float) sin( (double) src->f[2] ); - dst->f[3] = (float) sin( (double) src->f[3] ); -} - -static void -micro_sqrt( union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) sqrt( (double) src->f[0] ); - dst->f[1] = (float) sqrt( (double) src->f[1] ); - dst->f[2] = (float) sqrt( (double) src->f[2] ); - dst->f[3] = (float) sqrt( (double) src->f[3] ); -} - -static void -micro_sub( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] - src1->f[0]; - dst->f[1] = src0->f[1] - src1->f[1]; - dst->f[2] = src0->f[2] - src1->f[2]; - dst->f[3] = src0->f[3] - src1->f[3]; -} - -static void -micro_u2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) src->u[0]; - dst->f[1] = (float) src->u[1]; - dst->f[2] = (float) src->u[2]; - dst->f[3] = (float) src->u[3]; -} - -static void -micro_xor( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] ^ src1->u[0]; - dst->u[1] = src0->u[1] ^ src1->u[1]; - dst->u[2] = src0->u[2] ^ src1->u[2]; - dst->u[3] = src0->u[3] ^ src1->u[3]; -} - -static void -fetch_src_file_channel( - const struct tgsi_exec_machine *mach, - const uint file, - const uint swizzle, - const union tgsi_exec_channel *index, - union tgsi_exec_channel *chan ) -{ - switch( swizzle ) { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - switch( file ) { - case TGSI_FILE_CONSTANT: - chan->f[0] = mach->Consts[index->i[0]][swizzle]; - chan->f[1] = mach->Consts[index->i[1]][swizzle]; - chan->f[2] = mach->Consts[index->i[2]][swizzle]; - chan->f[3] = mach->Consts[index->i[3]][swizzle]; - break; - - case TGSI_FILE_INPUT: - chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_TEMPORARY: - chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_IMMEDIATE: - assert( index->i[0] < (int) mach->ImmLimit ); - chan->f[0] = mach->Imms[index->i[0]][swizzle]; - assert( index->i[1] < (int) mach->ImmLimit ); - chan->f[1] = mach->Imms[index->i[1]][swizzle]; - assert( index->i[2] < (int) mach->ImmLimit ); - chan->f[2] = mach->Imms[index->i[2]][swizzle]; - assert( index->i[3] < (int) mach->ImmLimit ); - chan->f[3] = mach->Imms[index->i[3]][swizzle]; - break; - - case TGSI_FILE_ADDRESS: - chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_OUTPUT: - /* vertex/fragment output vars can be read too */ - chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; - break; - - default: - assert( 0 ); - } - break; - - case TGSI_EXTSWIZZLE_ZERO: - *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; - break; - - case TGSI_EXTSWIZZLE_ONE: - *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; - break; - - default: - assert( 0 ); - } -} - -static void -fetch_source( - const struct tgsi_exec_machine *mach, - union tgsi_exec_channel *chan, - const struct tgsi_full_src_register *reg, - const uint chan_index ) -{ - union tgsi_exec_channel index; - uint swizzle; - - index.i[0] = - index.i[1] = - index.i[2] = - index.i[3] = reg->SrcRegister.Index; - - if (reg->SrcRegister.Indirect) { - union tgsi_exec_channel index2; - union tgsi_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterInd, CHAN_X ); - fetch_src_file_channel( - mach, - reg->SrcRegisterInd.File, - swizzle, - &index2, - &indir_index ); - - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; - } - - if( reg->SrcRegister.Dimension ) { - switch( reg->SrcRegister.File ) { - case TGSI_FILE_INPUT: - index.i[0] *= 17; - index.i[1] *= 17; - index.i[2] *= 17; - index.i[3] *= 17; - break; - case TGSI_FILE_CONSTANT: - index.i[0] *= 4096; - index.i[1] *= 4096; - index.i[2] *= 4096; - index.i[3] *= 4096; - break; - default: - assert( 0 ); - } - - index.i[0] += reg->SrcRegisterDim.Index; - index.i[1] += reg->SrcRegisterDim.Index; - index.i[2] += reg->SrcRegisterDim.Index; - index.i[3] += reg->SrcRegisterDim.Index; - - if (reg->SrcRegisterDim.Indirect) { - union tgsi_exec_channel index2; - union tgsi_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterDimInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); - fetch_src_file_channel( - mach, - reg->SrcRegisterDimInd.File, - swizzle, - &index2, - &indir_index ); - - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; - } - } - - swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); - fetch_src_file_channel( - mach, - reg->SrcRegister.File, - swizzle, - &index, - chan ); - - switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { - case TGSI_UTIL_SIGN_CLEAR: - micro_abs( chan, chan ); - break; - - case TGSI_UTIL_SIGN_SET: - micro_abs( chan, chan ); - micro_neg( chan, chan ); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - micro_neg( chan, chan ); - break; - - case TGSI_UTIL_SIGN_KEEP: - break; - } - - if (reg->SrcRegisterExtMod.Complement) { - micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan ); - } -} - -static void -store_dest( - struct tgsi_exec_machine *mach, - const union tgsi_exec_channel *chan, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - uint chan_index ) -{ - union tgsi_exec_channel *dst; - - switch( reg->DstRegister.File ) { - case TGSI_FILE_NULL: - return; - - case TGSI_FILE_OUTPUT: - dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] - + reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_TEMPORARY: - dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_ADDRESS: - dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; - break; - - default: - assert( 0 ); - return; - } - - switch (inst->Instruction.Saturate) - { - case TGSI_SAT_NONE: - if (mach->ExecMask & 0x1) - dst->i[0] = chan->i[0]; - if (mach->ExecMask & 0x2) - dst->i[1] = chan->i[1]; - if (mach->ExecMask & 0x4) - dst->i[2] = chan->i[2]; - if (mach->ExecMask & 0x8) - dst->i[3] = chan->i[3]; - break; - - case TGSI_SAT_ZERO_ONE: - /* XXX need to obey ExecMask here */ - micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); - micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); - break; - - case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); - break; - - default: - assert( 0 ); - } -} - -#define FETCH(VAL,INDEX,CHAN)\ - fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) - -#define STORE(VAL,INDEX,CHAN)\ - store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) - - -/** - * Execute ARB-style KIL which is predicated by a src register. - * Kill fragment if any of the four values is less than zero. - */ -static void -exec_kilp(struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst) -{ - uint uniquemask; - uint chan_index; - uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ - union tgsi_exec_channel r[1]; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - for (chan_index = 0; chan_index < 4; chan_index++) - { - uint swizzle; - uint i; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle ( - &inst->FullSrcRegisters[0], - chan_index); - - /* check if the component has not been already tested */ - if (uniquemask & (1 << swizzle)) - continue; - uniquemask |= 1 << swizzle; - - FETCH(&r[0], 0, chan_index); - for (i = 0; i < 4; i++) - if (r[0].f[i] < 0.0f) - kilmask |= 1 << i; - } - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; -} - - -/* - * Fetch a texel using STR texture coordinates. - */ -static void -fetch_texel( struct tgsi_sampler *sampler, - const union tgsi_exec_channel *s, - const union tgsi_exec_channel *t, - const union tgsi_exec_channel *p, - float lodbias, /* XXX should be float[4] */ - union tgsi_exec_channel *r, - union tgsi_exec_channel *g, - union tgsi_exec_channel *b, - union tgsi_exec_channel *a ) -{ - uint j; - float rgba[NUM_CHANNELS][QUAD_SIZE]; - - sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); - - for (j = 0; j < 4; j++) { - r->f[j] = rgba[0][j]; - g->f[j] = rgba[1][j]; - b->f[j] = rgba[2][j]; - a->f[j] = rgba[3][j]; - } -} - - -static void -exec_tex(struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst, - boolean biasLod) -{ - const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; - union tgsi_exec_channel r[8]; - uint chan_index; - float lodBias; - - /* debug_printf("Sampler %u unit %u\n", sampler, unit); */ - - switch (inst->InstructionExtTexture.Texture) { - case TGSI_TEXTURE_1D: - - FETCH(&r[0], 0, CHAN_X); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[1], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[1] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - if (biasLod) { - FETCH(&r[1], 0, CHAN_W); - lodBias = r[2].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ - &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ - break; - - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, /* inputs */ - &r[0], &r[1], &r[2], &r[3]); /* outputs */ - break; - - case TGSI_TEXTURE_3D: - case TGSI_TEXTURE_CUBE: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, - &r[0], &r[1], &r[2], &r[3]); - break; - - default: - assert (0); - } - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[chan_index], 0, chan_index ); - } -} - - -/** - * Evaluate a constant-valued coefficient at the position of the - * current quad. - */ -static void -eval_constant_coef( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; - } -} - -/** - * Evaluate a linear-valued coefficient at the position of the - * current quad. - */ -static void -eval_linear_coef( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - const float x = mach->QuadPos.xyzw[0].f[0]; - const float y = mach->QuadPos.xyzw[1].f[0]; - const float dadx = mach->InterpCoefs[attrib].dadx[chan]; - const float dady = mach->InterpCoefs[attrib].dady[chan]; - const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; - mach->Inputs[attrib].xyzw[chan].f[0] = a0; - mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; - mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; - mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; -} - -/** - * Evaluate a perspective-valued coefficient at the position of the - * current quad. - */ -static void -eval_perspective_coef( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - const float x = mach->QuadPos.xyzw[0].f[0]; - const float y = mach->QuadPos.xyzw[1].f[0]; - const float dadx = mach->InterpCoefs[attrib].dadx[chan]; - const float dady = mach->InterpCoefs[attrib].dady[chan]; - const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; - const float *w = mach->QuadPos.xyzw[3].f; - /* divide by W here */ - mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; - mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; - mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; - mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; -} - - -typedef void (* eval_coef_func)( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ); - -static void -exec_declaration( - struct tgsi_exec_machine *mach, - const struct tgsi_full_declaration *decl ) -{ - if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { - if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask; - eval_coef_func eval; - - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - switch( decl->Interpolation.Interpolate ) { - case TGSI_INTERPOLATE_CONSTANT: - eval = eval_constant_coef; - break; - - case TGSI_INTERPOLATE_LINEAR: - eval = eval_linear_coef; - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - eval = eval_perspective_coef; - break; - - default: - assert( 0 ); - } - - if( mask == TGSI_WRITEMASK_XYZW ) { - unsigned i, j; - - for( i = first; i <= last; i++ ) { - for( j = 0; j < NUM_CHANNELS; j++ ) { - eval( mach, i, j ); - } - } - } - else { - unsigned i, j; - - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - eval( mach, i, j ); - } - } - } - } - } - } -} - -static void -exec_instruction( - struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst, - int *pc ) -{ - uint chan_index; - union tgsi_exec_channel r[8]; - - (*pc)++; - - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_f2it( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LIT: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_X ); - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[1], 0, CHAN_Y ); - micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - - FETCH( &r[2], 0, CHAN_W ); - micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); - micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); - micro_pow( &r[1], &r[1], &r[2] ); - micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Z ); - } - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH( &r[0], 0, CHAN_X ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH( &r[0], 0, CHAN_X ); - micro_sqrt( &r[0], &r[0] ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXP: - assert (0); - break; - - case TGSI_OPCODE_LOG: - assert (0); - break; - - case TGSI_OPCODE_MUL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_mul( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_ADD: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Z ); - FETCH( &r[2], 1, CHAN_Z ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_W); - FETCH(&r[2], 1, CHAN_W); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DST: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - FETCH( &r[0], 0, CHAN_Y ); - FETCH( &r[1], 1, CHAN_Y); - micro_mul( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_Z ); - STORE( &r[0], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - FETCH( &r[0], 1, CHAN_W ); - STORE( &r[0], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MIN: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - /* XXX use micro_min()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_MAX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - /* XXX use micro_max()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); - - STORE(&r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_mul( &r[0], &r[0], &r[1] ); - FETCH( &r[1], 2, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SUB: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_sub( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - micro_sub( &r[1], &r[1], &r[2] ); - micro_mul( &r[0], &r[0], &r[1] ); - micro_add( &r[0], &r[0], &r[2] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_CND: - assert (0); - break; - - case TGSI_OPCODE_CND0: - assert (0); - break; - - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - assert (0); - break; - - case TGSI_OPCODE_INDEX: - assert (0); - break; - - case TGSI_OPCODE_NEGATE: - assert (0); - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_frc( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CLAMP: - assert (0); - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_flr( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_ROUND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_rnd( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ - FETCH(&r[0], 0, CHAN_X); - - micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ - FETCH( &r[0], 0, CHAN_X ); - micro_lg2( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_pow( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - FETCH(&r[0], 0, CHAN_Y); - FETCH(&r[1], 1, CHAN_Z); - - micro_mul( &r[2], &r[0], &r[1] ); - - FETCH(&r[3], 0, CHAN_Z); - FETCH(&r[4], 1, CHAN_Y); - - micro_mul( &r[5], &r[3], &r[4] ); - micro_sub( &r[2], &r[2], &r[5] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &r[2], 0, CHAN_X ); - } - - FETCH(&r[2], 1, CHAN_X); - - micro_mul( &r[3], &r[3], &r[2] ); - - FETCH(&r[5], 0, CHAN_X); - - micro_mul( &r[1], &r[1], &r[5] ); - micro_sub( &r[3], &r[3], &r[1] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - STORE( &r[3], 0, CHAN_Y ); - } - - micro_mul( &r[5], &r[5], &r[4] ); - micro_mul( &r[0], &r[0], &r[2] ); - micro_sub( &r[5], &r[5], &r[0] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - STORE( &r[5], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MULTIPLYMATRIX: - assert (0); - break; - - case TGSI_OPCODE_ABS: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - - micro_abs( &r[0], &r[0] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_RCC: - assert (0); - break; - - case TGSI_OPCODE_DPH: - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 1, CHAN_W); - - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_COS: - FETCH(&r[0], 0, CHAN_X); - - micro_cos( &r[0], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddx( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDY: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddy( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_KILP: - exec_kilp (mach, inst); - break; - - case TGSI_OPCODE_KIL: - /* for enabled ExecMask bits, set the killed bit */ - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; - break; - - case TGSI_OPCODE_PK2H: - assert (0); - break; - - case TGSI_OPCODE_PK2US: - assert (0); - break; - - case TGSI_OPCODE_PK4B: - assert (0); - break; - - case TGSI_OPCODE_PK4UB: - assert (0); - break; - - case TGSI_OPCODE_RFL: - assert (0); - break; - - case TGSI_OPCODE_SEQ: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], - &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], - &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SFL: - assert (0); - break; - - case TGSI_OPCODE_SGT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SIN: - FETCH( &r[0], 0, CHAN_X ); - micro_sin( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SNE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_STR: - assert (0); - break; - - case TGSI_OPCODE_TEX: - /* simple texture lookup */ - /* src[0] = texcoord */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, FALSE); - break; - - case TGSI_OPCODE_TXB: - /* Texture lookup with lod bias */ - /* src[0] = texcoord (src[0].w = LOD bias) */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); - break; - - case TGSI_OPCODE_TXD: - /* Texture lookup with explict partial derivatives */ - /* src[0] = texcoord */ - /* src[1] = d[strq]/dx */ - /* src[2] = d[strq]/dy */ - /* src[3] = sampler unit */ - assert (0); - break; - - case TGSI_OPCODE_TXL: - /* Texture lookup with explit LOD */ - /* src[0] = texcoord (src[0].w = LOD) */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); - break; - - case TGSI_OPCODE_UP2H: - assert (0); - break; - - case TGSI_OPCODE_UP2US: - assert (0); - break; - - case TGSI_OPCODE_UP4B: - assert (0); - break; - - case TGSI_OPCODE_UP4UB: - assert (0); - break; - - case TGSI_OPCODE_X2D: - assert (0); - break; - - case TGSI_OPCODE_ARA: - assert (0); - break; - - case TGSI_OPCODE_ARR: - assert (0); - break; - - case TGSI_OPCODE_BRA: - assert (0); - break; - - case TGSI_OPCODE_CAL: - /* skip the call if no execution channels are enabled */ - if (mach->ExecMask) { - /* do the call */ - - /* push the Cond, Loop, Cont stacks */ - assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); - mach->CondStack[mach->CondStackTop++] = mach->CondMask; - assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; - assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; - - assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); - mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; - - /* note that PC was already incremented above */ - mach->CallStack[mach->CallStackTop++] = *pc; - *pc = inst->InstructionExtLabel.Label; - } - break; - - case TGSI_OPCODE_RET: - mach->FuncMask &= ~mach->ExecMask; - UPDATE_EXEC_MASK(mach); - - if (mach->ExecMask == 0x0) { - /* really return now (otherwise, keep executing */ - - if (mach->CallStackTop == 0) { - /* returning from main() */ - *pc = -1; - return; - } - *pc = mach->CallStack[--mach->CallStackTop]; - - /* pop the Cond, Loop, Cont stacks */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; - assert(mach->FuncStackTop > 0); - mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; - - UPDATE_EXEC_MASK(mach); - } - break; - - case TGSI_OPCODE_SSG: - assert (0); - break; - - case TGSI_OPCODE_CMP: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_SCS: - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( &r[0], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - micro_cos( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - micro_sin( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_Y ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_NRM: - assert (0); - break; - - case TGSI_OPCODE_DIV: - assert( 0 ); - break; - - case TGSI_OPCODE_DP2: - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_IF: - /* push CondMask */ - assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); - mach->CondStack[mach->CondStackTop++] = mach->CondMask; - FETCH( &r[0], 0, CHAN_X ); - /* update CondMask */ - if( ! r[0].u[0] ) { - mach->CondMask &= ~0x1; - } - if( ! r[0].u[1] ) { - mach->CondMask &= ~0x2; - } - if( ! r[0].u[2] ) { - mach->CondMask &= ~0x4; - } - if( ! r[0].u[3] ) { - mach->CondMask &= ~0x8; - } - UPDATE_EXEC_MASK(mach); - /* Todo: If CondMask==0, jump to ELSE */ - break; - - case TGSI_OPCODE_ELSE: - /* invert CondMask wrt previous mask */ - { - uint prevMask; - assert(mach->CondStackTop > 0); - prevMask = mach->CondStack[mach->CondStackTop - 1]; - mach->CondMask = ~mach->CondMask & prevMask; - UPDATE_EXEC_MASK(mach); - /* Todo: If CondMask==0, jump to ENDIF */ - } - break; - - case TGSI_OPCODE_ENDIF: - /* pop CondMask */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_END: - /* halt execution */ - *pc = -1; - break; - - case TGSI_OPCODE_REP: - assert (0); - break; - - case TGSI_OPCODE_ENDREP: - assert (0); - break; - - case TGSI_OPCODE_PUSHA: - assert (0); - break; - - case TGSI_OPCODE_POPA: - assert (0); - break; - - case TGSI_OPCODE_CEIL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ceil( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_I2F: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_i2f( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_NOT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_not( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_TRUNC: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_trunc( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SHL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_shl( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SHR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ishr( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_AND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_and( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_OR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_or( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOD: - assert (0); - break; - - case TGSI_OPCODE_XOR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_xor( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SAD: - assert (0); - break; - - case TGSI_OPCODE_TXF: - assert (0); - break; - - case TGSI_OPCODE_TXQ: - assert (0); - break; - - case TGSI_OPCODE_EMIT: - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; - break; - - case TGSI_OPCODE_ENDPRIM: - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; - break; - - case TGSI_OPCODE_LOOP: - /* fall-through (for now) */ - case TGSI_OPCODE_BGNLOOP2: - /* push LoopMask and ContMasks */ - assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; - assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; - break; - - case TGSI_OPCODE_ENDLOOP: - /* fall-through (for now at least) */ - case TGSI_OPCODE_ENDLOOP2: - /* Restore ContMask, but don't pop */ - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; - if (mach->LoopMask) { - /* repeat loop: jump to instruction just past BGNLOOP */ - *pc = inst->InstructionExtLabel.Label + 1; - } - else { - /* exit loop: pop LoopMask */ - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - /* pop ContMask */ - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; - } - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_BRK: - /* turn off loop channels for each enabled exec channel */ - mach->LoopMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_CONT: - /* turn off cont channels for each enabled exec channel */ - mach->ContMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_BGNSUB: - /* no-op */ - break; - - case TGSI_OPCODE_ENDSUB: - /* no-op */ - break; - - case TGSI_OPCODE_NOISE1: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE2: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE3: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE4: - assert( 0 ); - break; - - case TGSI_OPCODE_NOP: - break; - - default: - assert( 0 ); - } -} - - -/** - * Run TGSI interpreter. - * \return bitmask of "alive" quad components - */ -uint -tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) -{ - uint i; - int pc = 0; - - mach->CondMask = 0xf; - mach->LoopMask = 0xf; - mach->ContMask = 0xf; - mach->FuncMask = 0xf; - mach->ExecMask = 0xf; - - mach->CondStackTop = 0; /* temporarily subvert this assertion */ - assert(mach->CondStackTop == 0); - assert(mach->LoopStackTop == 0); - assert(mach->ContStackTop == 0); - assert(mach->CallStackTop == 0); - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; - - if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; - mach->Primitives[0] = 0; - } - - - /* execute declarations (interpolants) */ - for (i = 0; i < mach->NumDeclarations; i++) { - exec_declaration( mach, mach->Declarations+i ); - } - - /* execute instructions, until pc is set to -1 */ - while (pc != -1) { - assert(pc < mach->NumInstructions); - exec_instruction( mach, mach->Instructions + pc, &pc ); - } - -#if 0 - /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ - if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { - /* - * Scale back depth component. - */ - for (i = 0; i < 4; i++) - mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; - } -#endif - - return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; -} - - diff --git a/src/gallium/aux/tgsi/exec/tgsi_exec.h b/src/gallium/aux/tgsi/exec/tgsi_exec.h deleted file mode 100644 index 1fb66ee960..0000000000 --- a/src/gallium/aux/tgsi/exec/tgsi_exec.h +++ /dev/null @@ -1,239 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#if !defined TGSI_EXEC_H -#define TGSI_EXEC_H - -#include "pipe/p_compiler.h" - -#if defined __cplusplus -extern "C" { -#endif - -#define NUM_CHANNELS 4 /* R,G,B,A */ -#define QUAD_SIZE 4 /* 4 pixel/quad */ - -/** - * Registers may be treated as float, signed int or unsigned int. - */ -union tgsi_exec_channel -{ - float f[QUAD_SIZE]; - int i[QUAD_SIZE]; - unsigned u[QUAD_SIZE]; -}; - -/** - * A vector[RGBA] of channels[4 pixels] - */ -struct tgsi_exec_vector -{ - union tgsi_exec_channel xyzw[NUM_CHANNELS]; -}; - -/** - * For fragment programs, information for computing fragment input - * values from plane equation of the triangle/line. - */ -struct tgsi_interp_coef -{ - float a0[NUM_CHANNELS]; /* in an xyzw layout */ - float dadx[NUM_CHANNELS]; - float dady[NUM_CHANNELS]; -}; - - -struct softpipe_tile_cache; /**< Opaque to TGSI */ - -/** - * Information for sampling textures, which must be implemented - * by code outside the TGSI executor. - */ -struct tgsi_sampler -{ - const struct pipe_sampler_state *state; - struct pipe_texture *texture; - /** Get samples for four fragments in a quad */ - void (*get_samples)(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]); - void *pipe; /*XXX temporary*/ - struct softpipe_tile_cache *cache; -}; - -/** - * For branching/calling subroutines. - */ -struct tgsi_exec_labels -{ - unsigned labels[128][2]; - unsigned count; -}; - -/* - * Locations of various utility registers (_I = Index, _C = Channel) - */ -#define TGSI_EXEC_TEMP_00000000_I 32 -#define TGSI_EXEC_TEMP_00000000_C 0 - -#define TGSI_EXEC_TEMP_7FFFFFFF_I 32 -#define TGSI_EXEC_TEMP_7FFFFFFF_C 1 - -#define TGSI_EXEC_TEMP_80000000_I 32 -#define TGSI_EXEC_TEMP_80000000_C 2 - -#define TGSI_EXEC_TEMP_FFFFFFFF_I 32 -#define TGSI_EXEC_TEMP_FFFFFFFF_C 3 - -#define TGSI_EXEC_TEMP_ONE_I 33 -#define TGSI_EXEC_TEMP_ONE_C 0 - -#define TGSI_EXEC_TEMP_TWO_I 33 -#define TGSI_EXEC_TEMP_TWO_C 1 - -#define TGSI_EXEC_TEMP_128_I 33 -#define TGSI_EXEC_TEMP_128_C 2 - -#define TGSI_EXEC_TEMP_MINUS_128_I 33 -#define TGSI_EXEC_TEMP_MINUS_128_C 3 - -#define TGSI_EXEC_TEMP_KILMASK_I 34 -#define TGSI_EXEC_TEMP_KILMASK_C 0 - -#define TGSI_EXEC_TEMP_OUTPUT_I 34 -#define TGSI_EXEC_TEMP_OUTPUT_C 1 - -#define TGSI_EXEC_TEMP_PRIMITIVE_I 34 -#define TGSI_EXEC_TEMP_PRIMITIVE_C 2 - -#define TGSI_EXEC_TEMP_R0 35 - -#define TGSI_EXEC_NUM_TEMPS (32 + 4) -#define TGSI_EXEC_NUM_ADDRS 1 -#define TGSI_EXEC_NUM_IMMEDIATES 256 - -#define TGSI_EXEC_MAX_COND_NESTING 10 -#define TGSI_EXEC_MAX_LOOP_NESTING 10 -#define TGSI_EXEC_MAX_CALL_NESTING 10 - -/** - * Run-time virtual machine state for executing TGSI shader. - */ -struct tgsi_exec_machine -{ - /* - * 32 program temporaries - * 4 internal temporaries - * 1 address - * 1 temporary of padding to align to 16 bytes - */ - struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS + 1]; - - /* - * This will point to _Temps after aligning to 16B boundary. - */ - struct tgsi_exec_vector *Temps; - struct tgsi_exec_vector *Addrs; - - struct tgsi_sampler *Samplers; - - float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; - unsigned ImmLimit; - float (*Consts)[4]; - struct tgsi_exec_vector *Inputs; - struct tgsi_exec_vector *Outputs; - const struct tgsi_token *Tokens; - unsigned Processor; - - /* GEOMETRY processor only. */ - unsigned *Primitives; - - /* FRAGMENT processor only. */ - const struct tgsi_interp_coef *InterpCoefs; - struct tgsi_exec_vector QuadPos; - - /* Conditional execution masks */ - uint CondMask; /**< For IF/ELSE/ENDIF */ - uint LoopMask; /**< For BGNLOOP/ENDLOOP */ - uint ContMask; /**< For loop CONT statements */ - uint FuncMask; /**< For function calls */ - uint ExecMask; /**< = CondMask & LoopMask */ - - /** Condition mask stack (for nested conditionals) */ - uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; - int CondStackTop; - - /** Loop mask stack (for nested loops) */ - uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; - int LoopStackTop; - - /** Loop continue mask stack (see comments in tgsi_exec.c) */ - uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; - int ContStackTop; - - /** Function execution mask stack (for executing subroutine code) */ - uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; - int FuncStackTop; - - /** Function call stack for saving/restoring the program counter */ - uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; - int CallStackTop; - - struct tgsi_full_instruction *Instructions; - uint NumInstructions; - - struct tgsi_full_declaration *Declarations; - uint NumDeclarations; - - struct tgsi_exec_labels Labels; -}; - - -void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach, - const struct tgsi_token *tokens, - unsigned numSamplers, - struct tgsi_sampler *samplers); - -uint -tgsi_exec_machine_run( - struct tgsi_exec_machine *mach ); - - -void -tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach); - - -#if defined __cplusplus -} /* extern "C" */ -#endif - -#endif /* TGSI_EXEC_H */ diff --git a/src/gallium/aux/tgsi/exec/tgsi_sse2.c b/src/gallium/aux/tgsi/exec/tgsi_sse2.c deleted file mode 100755 index 593464db3e..0000000000 --- a/src/gallium/aux/tgsi/exec/tgsi_sse2.c +++ /dev/null @@ -1,2378 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi_exec.h" -#include "tgsi_sse2.h" - -#include "x86/rtasm/x86sse.h" - -#if defined(__i386__) || defined(__386__) - -#define DUMP_SSE 0 - -#if DUMP_SSE - -static void -_print_reg( - struct x86_reg reg ) -{ - if (reg.mod != mod_REG) - debug_printf( "[" ); - - switch( reg.file ) { - case file_REG32: - switch( reg.idx ) { - case reg_AX: - debug_printf( "EAX" ); - break; - case reg_CX: - debug_printf( "ECX" ); - break; - case reg_DX: - debug_printf( "EDX" ); - break; - case reg_BX: - debug_printf( "EBX" ); - break; - case reg_SP: - debug_printf( "ESP" ); - break; - case reg_BP: - debug_printf( "EBP" ); - break; - case reg_SI: - debug_printf( "ESI" ); - break; - case reg_DI: - debug_printf( "EDI" ); - break; - } - break; - case file_MMX: - assert( 0 ); - break; - case file_XMM: - debug_printf( "XMM%u", reg.idx ); - break; - case file_x87: - assert( 0 ); - break; - } - - if (reg.mod == mod_DISP8 || - reg.mod == mod_DISP32) - debug_printf("+%d", reg.disp); - - if (reg.mod != mod_REG) - debug_printf( "]" ); -} - -static void -_fill( - const char *op ) -{ - unsigned count = 10 - strlen( op ); - - while( count-- ) { - debug_printf( " " ); - } -} - -#define DUMP_START() debug_printf( "\nsse-dump start ----------------" ) -#define DUMP_END() debug_printf( "\nsse-dump end ----------------\n" ) -#define DUMP( OP ) debug_printf( "\n%s", OP ) -#define DUMP_I( OP, I ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - debug_printf( "%u", I ); } while( 0 ) -#define DUMP_R( OP, R0 ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - _print_reg( R0 ); } while( 0 ) -#define DUMP_RR( OP, R0, R1 ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - _print_reg( R0 );\ - debug_printf( ", " );\ - _print_reg( R1 ); } while( 0 ) -#define DUMP_RRI( OP, R0, R1, I ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - _print_reg( R0 );\ - debug_printf( ", " );\ - _print_reg( R1 );\ - debug_printf( ", " );\ - debug_printf( "%u", I ); } while( 0 ) - -#else - -#define DUMP_START() -#define DUMP_END() -#define DUMP( OP ) -#define DUMP_I( OP, I ) -#define DUMP_R( OP, R0 ) -#define DUMP_RR( OP, R0, R1 ) -#define DUMP_RRI( OP, R0, R1, I ) - -#endif - -#define FOR_EACH_CHANNEL( CHAN )\ - for( CHAN = 0; CHAN < 4; CHAN++ ) - -#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ - ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) - -#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ - if( IS_DST0_CHANNEL_ENABLED( INST, CHAN )) - -#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\ - FOR_EACH_CHANNEL( CHAN )\ - IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN ) - -#define CHAN_X 0 -#define CHAN_Y 1 -#define CHAN_Z 2 -#define CHAN_W 3 - -#define TEMP_R0 TGSI_EXEC_TEMP_R0 - -/** - * X86 utility functions. - */ - -static struct x86_reg -make_xmm( - unsigned xmm ) -{ - return x86_make_reg( - file_XMM, - (enum x86_reg_name) xmm ); -} - -/** - * X86 register mapping helpers. - */ - -static struct x86_reg -get_const_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_CX ); -} - -static struct x86_reg -get_input_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_AX ); -} - -static struct x86_reg -get_output_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_DX ); -} - -static struct x86_reg -get_temp_base( void ) -{ -#ifdef WIN32 - return x86_make_reg( - file_REG32, - reg_BX ); -#else - return x86_make_reg( - file_REG32, - reg_SI ); -#endif -} - -static struct x86_reg -get_coef_base( void ) -{ - return get_output_base(); -} - -/** - * Data access helpers. - */ - -static struct x86_reg -get_argument( - unsigned index ) -{ - return x86_make_disp( - x86_make_reg( file_REG32, reg_SP ), - (index + 1) * 4 ); -} - -static struct x86_reg -get_const( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_const_base(), - (vec * 4 + chan) * 4 ); -} - -static struct x86_reg -get_input( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_input_base(), - (vec * 4 + chan) * 16 ); -} - -static struct x86_reg -get_output( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_output_base(), - (vec * 4 + chan) * 16 ); -} - -static struct x86_reg -get_temp( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_temp_base(), - (vec * 4 + chan) * 16 ); -} - -static struct x86_reg -get_coef( - unsigned vec, - unsigned chan, - unsigned member ) -{ - return x86_make_disp( - get_coef_base(), - ((vec * 3 + member) * 4 + chan) * 4 ); -} - -/** - * X86 rtasm wrappers. - */ - -static void -emit_addps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ADDPS", dst, src ); - sse_addps( func, dst, src ); -} - -static void -emit_andnps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ANDNPS", dst, src ); - sse_andnps( func, dst, src ); -} - -static void -emit_andps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ANDPS", dst, src ); - sse_andps( func, dst, src ); -} - -static void -emit_call( - struct x86_function *func, - void (* addr)() ) -{ - struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); - - DUMP_I( "CALL", addr ); - x86_mov_reg_imm( func, ecx, (unsigned long) addr ); - x86_call( func, ecx ); -} - -static void -emit_cmpps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src, - enum sse_cc cc ) -{ - DUMP_RRI( "CMPPS", dst, src, cc ); - sse_cmpps( func, dst, src, cc ); -} - -static void -emit_cvttps2dq( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "CVTTPS2DQ", dst, src ); - sse2_cvttps2dq( func, dst, src ); -} - -static void -emit_maxps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MAXPS", dst, src ); - sse_maxps( func, dst, src ); -} - -static void -emit_minps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MINPS", dst, src ); - sse_minps( func, dst, src ); -} - -static void -emit_mov( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOV", dst, src ); - x86_mov( func, dst, src ); -} - -static void -emit_movaps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOVAPS", dst, src ); - sse_movaps( func, dst, src ); -} - -static void -emit_movss( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOVSS", dst, src ); - sse_movss( func, dst, src ); -} - -static void -emit_movups( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOVUPS", dst, src ); - sse_movups( func, dst, src ); -} - -static void -emit_mulps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MULPS", dst, src ); - sse_mulps( func, dst, src ); -} - -static void -emit_or( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "OR", dst, src ); - x86_or( func, dst, src ); -} - -static void -emit_orps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ORPS", dst, src ); - sse_orps( func, dst, src ); -} - -static void -emit_pmovmskb( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "PMOVMSKB", dst, src ); - sse_pmovmskb( func, dst, src ); -} - -static void -emit_pop( - struct x86_function *func, - struct x86_reg dst ) -{ - DUMP_R( "POP", dst ); - x86_pop( func, dst ); -} - -static void -emit_push( - struct x86_function *func, - struct x86_reg dst ) -{ - DUMP_R( "PUSH", dst ); - x86_push( func, dst ); -} - -static void -emit_rcpps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "RCPPS", dst, src ); - sse2_rcpps( func, dst, src ); -} - -#ifdef WIN32 -static void -emit_retw( - struct x86_function *func, - unsigned size ) -{ - DUMP_I( "RET", size ); - x86_retw( func, size ); -} -#else -static void -emit_ret( - struct x86_function *func ) -{ - DUMP( "RET" ); - x86_ret( func ); -} -#endif - -static void -emit_rsqrtps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "RSQRTPS", dst, src ); - sse_rsqrtps( func, dst, src ); -} - -static void -emit_shufps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src, - unsigned char shuf ) -{ - DUMP_RRI( "SHUFPS", dst, src, shuf ); - sse_shufps( func, dst, src, shuf ); -} - -static void -emit_subps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "SUBPS", dst, src ); - sse_subps( func, dst, src ); -} - -static void -emit_xorps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "XORPS", dst, src ); - sse_xorps( func, dst, src ); -} - -/** - * Data fetch helpers. - */ - -static void -emit_const( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movss( - func, - make_xmm( xmm ), - get_const( vec, chan ) ); - emit_shufps( - func, - make_xmm( xmm ), - make_xmm( xmm ), - SHUF( 0, 0, 0, 0 ) ); -} - -static void -emit_inputf( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movups( - func, - make_xmm( xmm ), - get_input( vec, chan ) ); -} - -static void -emit_output( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movups( - func, - get_output( vec, chan ), - make_xmm( xmm ) ); -} - -static void -emit_tempf( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movaps( - func, - make_xmm( xmm ), - get_temp( vec, chan ) ); -} - -static void -emit_coef( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan, - unsigned member ) -{ - emit_movss( - func, - make_xmm( xmm ), - get_coef( vec, chan, member ) ); - emit_shufps( - func, - make_xmm( xmm ), - make_xmm( xmm ), - SHUF( 0, 0, 0, 0 ) ); -} - -/** - * Data store helpers. - */ - -static void -emit_inputs( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movups( - func, - get_input( vec, chan ), - make_xmm( xmm ) ); -} - -static void -emit_temps( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_movaps( - func, - get_temp( vec, chan ), - make_xmm( xmm ) ); -} - -static void -emit_addrs( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_temps( - func, - xmm, - vec + TGSI_EXEC_NUM_TEMPS, - chan ); -} - -/** - * Coefficent fetch helpers. - */ - -static void -emit_coef_a0( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 0 ); -} - -static void -emit_coef_dadx( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 1 ); -} - -static void -emit_coef_dady( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 2 ); -} - -/** - * Function call helpers. - */ - -static void -emit_push_gp( - struct x86_function *func ) -{ - emit_push( - func, - get_const_base() ); - emit_push( - func, - get_input_base() ); - emit_push( - func, - get_output_base() ); - - /* It is important on non-win32 platforms that temp base is pushed last. - */ - emit_push( - func, - get_temp_base() ); -} - -static void -emit_pop_gp( - struct x86_function *func ) -{ - /* Restore GP registers in a reverse order. - */ - emit_pop( - func, - get_temp_base() ); - emit_pop( - func, - get_output_base() ); - emit_pop( - func, - get_input_base() ); - emit_pop( - func, - get_const_base() ); -} - -static void -emit_func_call_dst( - struct x86_function *func, - unsigned xmm_dst, - void (*code)() ) -{ - emit_movaps( - func, - get_temp( TEMP_R0, 0 ), - make_xmm( xmm_dst ) ); - - emit_push_gp( - func ); - -#ifdef WIN32 - emit_push( - func, - get_temp( TEMP_R0, 0 ) ); -#endif - - emit_call( - func, - code ); - - emit_pop_gp( - func ); - - emit_movaps( - func, - make_xmm( xmm_dst ), - get_temp( TEMP_R0, 0 ) ); -} - -static void -emit_func_call_dst_src( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src, - void (*code)() ) -{ - emit_movaps( - func, - get_temp( TEMP_R0, 1 ), - make_xmm( xmm_src ) ); - - emit_func_call_dst( - func, - xmm_dst, - code ); -} - -/** - * Low-level instruction translators. - */ - -static void -emit_abs( - struct x86_function *func, - unsigned xmm ) -{ - emit_andps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_7FFFFFFF_I, - TGSI_EXEC_TEMP_7FFFFFFF_C ) ); -} - -static void -emit_add( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_addps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void XSTDCALL -cos4f( - float *store ) -{ -#ifdef WIN32 - store[0] = (float) cos( (double) store[0] ); - store[1] = (float) cos( (double) store[1] ); - store[2] = (float) cos( (double) store[2] ); - store[3] = (float) cos( (double) store[3] ); -#else - const unsigned X = TEMP_R0 * 16; - store[X + 0] = cosf( store[X + 0] ); - store[X + 1] = cosf( store[X + 1] ); - store[X + 2] = cosf( store[X + 2] ); - store[X + 3] = cosf( store[X + 3] ); -#endif -} - -static void -emit_cos( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - cos4f ); -} - -static void XSTDCALL -ex24f( - float *store ) -{ -#ifdef WIN32 - store[0] = (float) pow( 2.0, (double) store[0] ); - store[1] = (float) pow( 2.0, (double) store[1] ); - store[2] = (float) pow( 2.0, (double) store[2] ); - store[3] = (float) pow( 2.0, (double) store[3] ); -#else - const unsigned X = TEMP_R0 * 16; - store[X + 0] = powf( 2.0f, store[X + 0] ); - store[X + 1] = powf( 2.0f, store[X + 1] ); - store[X + 2] = powf( 2.0f, store[X + 2] ); - store[X + 3] = powf( 2.0f, store[X + 3] ); -#endif -} - -static void -emit_ex2( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - ex24f ); -} - -static void -emit_f2it( - struct x86_function *func, - unsigned xmm ) -{ - emit_cvttps2dq( - func, - make_xmm( xmm ), - make_xmm( xmm ) ); -} - -static void XSTDCALL -flr4f( - float *store ) -{ -#ifdef WIN32 - const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif - store[X + 0] = (float) floor( (double) store[X + 0] ); - store[X + 1] = (float) floor( (double) store[X + 1] ); - store[X + 2] = (float) floor( (double) store[X + 2] ); - store[X + 3] = (float) floor( (double) store[X + 3] ); -} - -static void -emit_flr( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - flr4f ); -} - -static void XSTDCALL -frc4f( - float *store ) -{ -#ifdef WIN32 - const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif - store[X + 0] -= (float) floor( (double) store[X + 0] ); - store[X + 1] -= (float) floor( (double) store[X + 1] ); - store[X + 2] -= (float) floor( (double) store[X + 2] ); - store[X + 3] -= (float) floor( (double) store[X + 3] ); -} - -static void -emit_frc( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - frc4f ); -} - -static void XSTDCALL -lg24f( - float *store ) -{ -#ifdef WIN32 - const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif - store[X + 0] = LOG2( store[X + 0] ); - store[X + 1] = LOG2( store[X + 1] ); - store[X + 2] = LOG2( store[X + 2] ); - store[X + 3] = LOG2( store[X + 3] ); -} - -static void -emit_lg2( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - lg24f ); -} - -static void -emit_MOV( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_movups( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_mul (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) -{ - emit_mulps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_neg( - struct x86_function *func, - unsigned xmm ) -{ - emit_xorps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C ) ); -} - -static void XSTDCALL -pow4f( - float *store ) -{ -#ifdef WIN32 - store[0] = (float) pow( (double) store[0], (double) store[4] ); - store[1] = (float) pow( (double) store[1], (double) store[5] ); - store[2] = (float) pow( (double) store[2], (double) store[6] ); - store[3] = (float) pow( (double) store[3], (double) store[7] ); -#else - const unsigned X = TEMP_R0 * 16; - store[X + 0] = powf( store[X + 0], store[X + 4] ); - store[X + 1] = powf( store[X + 1], store[X + 5] ); - store[X + 2] = powf( store[X + 2], store[X + 6] ); - store[X + 3] = powf( store[X + 3], store[X + 7] ); -#endif -} - -static void -emit_pow( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_func_call_dst_src( - func, - xmm_dst, - xmm_src, - pow4f ); -} - -static void -emit_rcp ( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_rcpps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_rsqrt( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_rsqrtps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_setsign( - struct x86_function *func, - unsigned xmm ) -{ - emit_orps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C ) ); -} - -static void XSTDCALL -sin4f( - float *store ) -{ -#ifdef WIN32 - store[0] = (float) sin( (double) store[0] ); - store[1] = (float) sin( (double) store[1] ); - store[2] = (float) sin( (double) store[2] ); - store[3] = (float) sin( (double) store[3] ); -#else - const unsigned X = TEMP_R0 * 16; - store[X + 0] = sinf( store[X + 0] ); - store[X + 1] = sinf( store[X + 1] ); - store[X + 2] = sinf( store[X + 2] ); - store[X + 3] = sinf( store[X + 3] ); -#endif -} - -static void -emit_sin (struct x86_function *func, - unsigned xmm_dst) -{ - emit_func_call_dst( - func, - xmm_dst, - sin4f ); -} - -static void -emit_sub( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_subps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -/** - * Register fetch. - */ - -static void -emit_fetch( - struct x86_function *func, - unsigned xmm, - const struct tgsi_full_src_register *reg, - const unsigned chan_index ) -{ - unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); - - switch( swizzle ) { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - switch( reg->SrcRegister.File ) { - case TGSI_FILE_CONSTANT: - emit_const( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - case TGSI_FILE_INPUT: - emit_inputf( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - case TGSI_FILE_TEMPORARY: - emit_tempf( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - default: - assert( 0 ); - } - break; - - case TGSI_EXTSWIZZLE_ZERO: - emit_tempf( - func, - xmm, - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ); - break; - - case TGSI_EXTSWIZZLE_ONE: - emit_tempf( - func, - xmm, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - break; - - default: - assert( 0 ); - } - - switch( tgsi_util_get_full_src_register_sign_mode( reg, chan_index ) ) { - case TGSI_UTIL_SIGN_CLEAR: - emit_abs( func, xmm ); - break; - - case TGSI_UTIL_SIGN_SET: - emit_setsign( func, xmm ); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - emit_neg( func, xmm ); - break; - - case TGSI_UTIL_SIGN_KEEP: - break; - } -} - -#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\ - emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN ) - -/** - * Register store. - */ - -static void -emit_store( - struct x86_function *func, - unsigned xmm, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - unsigned chan_index ) -{ - switch( reg->DstRegister.File ) { - case TGSI_FILE_OUTPUT: - emit_output( - func, - xmm, - reg->DstRegister.Index, - chan_index ); - break; - - case TGSI_FILE_TEMPORARY: - emit_temps( - func, - xmm, - reg->DstRegister.Index, - chan_index ); - break; - - case TGSI_FILE_ADDRESS: - emit_addrs( - func, - xmm, - reg->DstRegister.Index, - chan_index ); - break; - - default: - assert( 0 ); - } - - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - - case TGSI_SAT_ZERO_ONE: -// assert( 0 ); - break; - - case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); - break; - } -} - -#define STORE( FUNC, INST, XMM, INDEX, CHAN )\ - emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) - -/** - * High-level instruction translators. - */ - -static void -emit_kil( - struct x86_function *func, - const struct tgsi_full_src_register *reg ) -{ - unsigned uniquemask; - unsigned registers[4]; - unsigned nextregister = 0; - unsigned firstchan = ~0; - unsigned chan_index; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - FOR_EACH_CHANNEL( chan_index ) { - unsigned swizzle; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle( - reg, - chan_index ); - - /* check if the component has not been already tested */ - if( !(uniquemask & (1 << swizzle)) ) { - uniquemask |= 1 << swizzle; - - /* allocate register */ - registers[chan_index] = nextregister; - emit_fetch( - func, - nextregister, - reg, - chan_index ); - nextregister++; - - /* mark the first channel used */ - if( firstchan == ~0 ) { - firstchan = chan_index; - } - } - } - - emit_push( - func, - x86_make_reg( file_REG32, reg_AX ) ); - emit_push( - func, - x86_make_reg( file_REG32, reg_DX ) ); - - FOR_EACH_CHANNEL( chan_index ) { - if( uniquemask & (1 << chan_index) ) { - emit_cmpps( - func, - make_xmm( registers[chan_index] ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ), - cc_LessThan ); - - if( chan_index == firstchan ) { - emit_pmovmskb( - func, - x86_make_reg( file_REG32, reg_AX ), - make_xmm( registers[chan_index] ) ); - } - else { - emit_pmovmskb( - func, - x86_make_reg( file_REG32, reg_DX ), - make_xmm( registers[chan_index] ) ); - emit_or( - func, - x86_make_reg( file_REG32, reg_AX ), - x86_make_reg( file_REG32, reg_DX ) ); - } - } - } - - emit_or( - func, - get_temp( - TGSI_EXEC_TEMP_KILMASK_I, - TGSI_EXEC_TEMP_KILMASK_C ), - x86_make_reg( file_REG32, reg_AX ) ); - - emit_pop( - func, - x86_make_reg( file_REG32, reg_DX ) ); - emit_pop( - func, - x86_make_reg( file_REG32, reg_AX ) ); -} - -static void -emit_setcc( - struct x86_function *func, - struct tgsi_full_instruction *inst, - enum sse_cc cc ) -{ - unsigned chan_index; - - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_cmpps( - func, - make_xmm( 0 ), - make_xmm( 1 ), - cc ); - emit_andps( - func, - make_xmm( 0 ), - get_temp( - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ) ); - STORE( func, *inst, 0, 0, chan_index ); - } -} - -static void -emit_cmp( - struct x86_function *func, - struct tgsi_full_instruction *inst ) -{ - unsigned chan_index; - - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - FETCH( func, *inst, 2, 2, chan_index ); - emit_cmpps( - func, - make_xmm( 0 ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ), - cc_LessThan ); - emit_andps( - func, - make_xmm( 1 ), - make_xmm( 0 ) ); - emit_andnps( - func, - make_xmm( 0 ), - make_xmm( 2 ) ); - emit_orps( - func, - make_xmm( 0 ), - make_xmm( 1 ) ); - STORE( func, *inst, 0, 0, chan_index ); - } -} - -static int -emit_instruction( - struct x86_function *func, - struct tgsi_full_instruction *inst ) -{ - unsigned chan_index; - - switch( inst->Instruction.Opcode ) { - case TGSI_OPCODE_ARL: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_f2it( func, 0 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_LIT: - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C); - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - STORE( func, *inst, 0, 0, CHAN_X ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - STORE( func, *inst, 0, 0, CHAN_W ); - } - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_maxps( - func, - make_xmm( 0 ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ) ); - STORE( func, *inst, 0, 0, CHAN_Y ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 1, 0, CHAN_Y ); - emit_maxps( - func, - make_xmm( 1 ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ) ); - FETCH( func, *inst, 2, 0, CHAN_W ); - emit_minps( - func, - make_xmm( 2 ), - get_temp( - TGSI_EXEC_TEMP_128_I, - TGSI_EXEC_TEMP_128_C ) ); - emit_maxps( - func, - make_xmm( 2 ), - get_temp( - TGSI_EXEC_TEMP_MINUS_128_I, - TGSI_EXEC_TEMP_MINUS_128_C ) ); - emit_pow( func, 1, 2 ); - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_xorps( - func, - make_xmm( 2 ), - make_xmm( 2 ) ); - emit_cmpps( - func, - make_xmm( 2 ), - make_xmm( 0 ), - cc_LessThanEqual ); - emit_andps( - func, - make_xmm( 2 ), - make_xmm( 1 ) ); - STORE( func, *inst, 2, 0, CHAN_Z ); - } - } - break; - - case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_rcp( func, 0, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_rsqrt( func, 0, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXP: - return 0; - break; - - case TGSI_OPCODE_LOG: - return 0; - break; - - case TGSI_OPCODE_MUL: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_mul( func, 0, 1 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_ADD: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_add( func, 0, 1 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_mul( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Y ); - FETCH( func, *inst, 2, 1, CHAN_Y ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Z ); - FETCH( func, *inst, 2, 1, CHAN_Z ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_mul( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Y ); - FETCH( func, *inst, 2, 1, CHAN_Y ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Z ); - FETCH( func, *inst, 2, 1, CHAN_Z ); - emit_mul(func, 1, 2 ); - emit_add(func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_W ); - FETCH( func, *inst, 2, 1, CHAN_W ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DST: - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - STORE( func, *inst, 0, 0, CHAN_X ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - FETCH( func, *inst, 1, 1, CHAN_Y ); - emit_mul( func, 0, 1 ); - STORE( func, *inst, 0, 0, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - FETCH( func, *inst, 0, 0, CHAN_Z ); - STORE( func, *inst, 0, 0, CHAN_Z ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - FETCH( func, *inst, 0, 1, CHAN_W ); - STORE( func, *inst, 0, 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MIN: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_minps( - func, - make_xmm( 0 ), - make_xmm( 1 ) ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_MAX: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_maxps( - func, - make_xmm( 0 ), - make_xmm( 1 ) ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - emit_setcc( func, inst, cc_LessThan ); - break; - - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - emit_setcc( func, inst, cc_NotLessThan ); - break; - - case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - FETCH( func, *inst, 2, 2, chan_index ); - emit_mul( func, 0, 1 ); - emit_add( func, 0, 2 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_SUB: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_sub( func, 0, 1 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - FETCH( func, *inst, 2, 2, chan_index ); - emit_sub( func, 1, 2 ); - emit_mul( func, 0, 1 ); - emit_add( func, 0, 2 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_CND: - return 0; - break; - - case TGSI_OPCODE_CND0: - return 0; - break; - - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - return 0; - break; - - case TGSI_OPCODE_INDEX: - return 0; - break; - - case TGSI_OPCODE_NEGATE: - return 0; - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_frc( func, 0 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_CLAMP: - return 0; - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_flr( func, 0 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_ROUND: - return 0; - break; - - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_ex2( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_lg2( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_pow( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( func, *inst, 1, 1, CHAN_Z ); - FETCH( func, *inst, 3, 0, CHAN_Z ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - FETCH( func, *inst, 4, 1, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - emit_MOV( func, 2, 0 ); - emit_mul( func, 2, 1 ); - emit_MOV( func, 5, 3 ); - emit_mul( func, 5, 4 ); - emit_sub( func, 2, 5 ); - STORE( func, *inst, 2, 0, CHAN_X ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 2, 1, CHAN_X ); - FETCH( func, *inst, 5, 0, CHAN_X ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - emit_mul( func, 3, 2 ); - emit_mul( func, 1, 5 ); - emit_sub( func, 3, 1 ); - STORE( func, *inst, 3, 0, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - emit_mul( func, 5, 4 ); - emit_mul( func, 0, 2 ); - emit_sub( func, 5, 0 ); - STORE( func, *inst, 5, 0, CHAN_Z ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); - STORE( func, *inst, 0, 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MULTIPLYMATRIX: - return 0; - break; - - case TGSI_OPCODE_ABS: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_abs( func, 0) ; - - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_RCC: - return 0; - break; - - case TGSI_OPCODE_DPH: - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_mul( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Y ); - FETCH( func, *inst, 2, 1, CHAN_Y ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Z ); - FETCH( func, *inst, 2, 1, CHAN_Z ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 1, CHAN_W ); - emit_add( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_COS: - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_cos( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDX: - return 0; - break; - - case TGSI_OPCODE_DDY: - return 0; - break; - - case TGSI_OPCODE_KIL: - emit_kil( func, &inst->FullSrcRegisters[0] ); - break; - - case TGSI_OPCODE_PK2H: - return 0; - break; - - case TGSI_OPCODE_PK2US: - return 0; - break; - - case TGSI_OPCODE_PK4B: - return 0; - break; - - case TGSI_OPCODE_PK4UB: - return 0; - break; - - case TGSI_OPCODE_RFL: - return 0; - break; - - case TGSI_OPCODE_SEQ: - return 0; - break; - - case TGSI_OPCODE_SFL: - return 0; - break; - - case TGSI_OPCODE_SGT: - return 0; - break; - - case TGSI_OPCODE_SIN: - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_sin( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLE: - return 0; - break; - - case TGSI_OPCODE_SNE: - return 0; - break; - - case TGSI_OPCODE_STR: - return 0; - break; - - case TGSI_OPCODE_TEX: - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_TXD: - return 0; - break; - - case TGSI_OPCODE_UP2H: - return 0; - break; - - case TGSI_OPCODE_UP2US: - return 0; - break; - - case TGSI_OPCODE_UP4B: - return 0; - break; - - case TGSI_OPCODE_UP4UB: - return 0; - break; - - case TGSI_OPCODE_X2D: - return 0; - break; - - case TGSI_OPCODE_ARA: - return 0; - break; - - case TGSI_OPCODE_ARR: - return 0; - break; - - case TGSI_OPCODE_BRA: - return 0; - break; - - case TGSI_OPCODE_CAL: - return 0; - break; - - case TGSI_OPCODE_RET: - case TGSI_OPCODE_END: -#ifdef WIN32 - emit_retw( func, 16 ); -#else - emit_ret( func ); -#endif - break; - - case TGSI_OPCODE_SSG: - return 0; - break; - - case TGSI_OPCODE_CMP: - emit_cmp (func, inst); - break; - - case TGSI_OPCODE_SCS: - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_cos( func, 0 ); - STORE( func, *inst, 0, 0, CHAN_X ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - emit_sin( func, 0 ); - STORE( func, *inst, 0, 0, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - FETCH( func, *inst, 0, TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ); - STORE( func, *inst, 0, 0, CHAN_Z ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); - STORE( func, *inst, 0, 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_TXB: - return 0; - break; - - case TGSI_OPCODE_NRM: - return 0; - break; - - case TGSI_OPCODE_DIV: - return 0; - break; - - case TGSI_OPCODE_DP2: - return 0; - break; - - case TGSI_OPCODE_TXL: - return 0; - break; - - case TGSI_OPCODE_BRK: - return 0; - break; - - case TGSI_OPCODE_IF: - return 0; - break; - - case TGSI_OPCODE_LOOP: - return 0; - break; - - case TGSI_OPCODE_REP: - return 0; - break; - - case TGSI_OPCODE_ELSE: - return 0; - break; - - case TGSI_OPCODE_ENDIF: - return 0; - break; - - case TGSI_OPCODE_ENDLOOP: - return 0; - break; - - case TGSI_OPCODE_ENDREP: - return 0; - break; - - case TGSI_OPCODE_PUSHA: - return 0; - break; - - case TGSI_OPCODE_POPA: - return 0; - break; - - case TGSI_OPCODE_CEIL: - return 0; - break; - - case TGSI_OPCODE_I2F: - return 0; - break; - - case TGSI_OPCODE_NOT: - return 0; - break; - - case TGSI_OPCODE_TRUNC: - return 0; - break; - - case TGSI_OPCODE_SHL: - return 0; - break; - - case TGSI_OPCODE_SHR: - return 0; - break; - - case TGSI_OPCODE_AND: - return 0; - break; - - case TGSI_OPCODE_OR: - return 0; - break; - - case TGSI_OPCODE_MOD: - return 0; - break; - - case TGSI_OPCODE_XOR: - return 0; - break; - - case TGSI_OPCODE_SAD: - return 0; - break; - - case TGSI_OPCODE_TXF: - return 0; - break; - - case TGSI_OPCODE_TXQ: - return 0; - break; - - case TGSI_OPCODE_CONT: - return 0; - break; - - case TGSI_OPCODE_EMIT: - return 0; - break; - - case TGSI_OPCODE_ENDPRIM: - return 0; - break; - - default: - return 0; - } - - return 1; -} - -static void -emit_declaration( - struct x86_function *func, - struct tgsi_full_declaration *decl ) -{ - if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask; - unsigned i, j; - - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - /* Do not touch WPOS.xy */ - if( first == 0 ) { - mask &= ~TGSI_WRITEMASK_XY; - if( mask == TGSI_WRITEMASK_NONE ) { - first++; - } - } - - for( i = first; i <= last; i++ ) { - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - switch( decl->Interpolation.Interpolate ) { - case TGSI_INTERPOLATE_CONSTANT: - emit_coef_a0( func, 0, i, j ); - emit_inputs( func, 0, i, j ); - break; - - case TGSI_INTERPOLATE_LINEAR: - emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); - emit_coef_dadx( func, 1, i, j ); - emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); - emit_coef_dady( func, 3, i, j ); - emit_mul( func, 0, 1 ); /* x * dadx */ - emit_coef_a0( func, 4, i, j ); - emit_mul( func, 2, 3 ); /* y * dady */ - emit_add( func, 0, 4 ); /* x * dadx + a0 */ - emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ - emit_inputs( func, 0, i, j ); - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); - emit_coef_dadx( func, 1, i, j ); - emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); - emit_coef_dady( func, 3, i, j ); - emit_mul( func, 0, 1 ); /* x * dadx */ - emit_inputf( func, 4, 0, TGSI_SWIZZLE_W ); - emit_coef_a0( func, 5, i, j ); - emit_rcp( func, 4, 4 ); /* 1.0 / w */ - emit_mul( func, 2, 3 ); /* y * dady */ - emit_add( func, 0, 5 ); /* x * dadx + a0 */ - emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ - emit_mul( func, 0, 4 ); /* (x * dadx + y * dady + a0) / w */ - emit_inputs( func, 0, i, j ); - break; - - default: - assert( 0 ); - break; - } - } - } - } - } -} - -unsigned -tgsi_emit_sse2( - struct tgsi_token *tokens, - struct x86_function *func ) -{ - struct tgsi_parse_context parse; - unsigned ok = 1; - - DUMP_START(); - - func->csr = func->store; - - emit_mov( - func, - get_input_base(), - get_argument( 0 ) ); - emit_mov( - func, - get_output_base(), - get_argument( 1 ) ); - emit_mov( - func, - get_const_base(), - get_argument( 2 ) ); - emit_mov( - func, - get_temp_base(), - get_argument( 3 ) ); - - tgsi_parse_init( &parse, tokens ); - - while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - ok = emit_instruction( - func, - &parse.FullToken.FullInstruction ); - - if (!ok) { - debug_printf("failed to translate tgsi opcode %d\n", - parse.FullToken.FullInstruction.Instruction.Opcode ); - } - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - /* XXX implement this */ - ok = 0; - debug_printf("failed to emit immediate value\n"); - break; - - default: - assert( 0 ); - ok = 0; - break; - } - } - - tgsi_parse_free( &parse ); - - DUMP_END(); - - return ok; -} - -/** - * Fragment shaders are responsible for interpolating shader inputs. Because on - * x86 we have only 4 GP registers, and here we have 5 shader arguments (input, - * output, const, temp and coef), the code is split into two phases -- - * DECLARATION and INSTRUCTION phase. - * GP register holding the output argument is aliased with the coeff argument, - * as outputs are not needed in the DECLARATION phase. - */ -unsigned -tgsi_emit_sse2_fs( - struct tgsi_token *tokens, - struct x86_function *func ) -{ - struct tgsi_parse_context parse; - boolean instruction_phase = FALSE; - - DUMP_START(); - - func->csr = func->store; - - /* DECLARATION phase, do not load output argument. */ - emit_mov( - func, - get_input_base(), - get_argument( 0 ) ); - emit_mov( - func, - get_const_base(), - get_argument( 2 ) ); - emit_mov( - func, - get_temp_base(), - get_argument( 3 ) ); - emit_mov( - func, - get_coef_base(), - get_argument( 4 ) ); - - tgsi_parse_init( &parse, tokens ); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - emit_declaration( - func, - &parse.FullToken.FullDeclaration ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - if( !instruction_phase ) { - /* INSTRUCTION phase, overwrite coeff with output. */ - instruction_phase = TRUE; - emit_mov( - func, - get_output_base(), - get_argument( 1 ) ); - } - emit_instruction( - func, - &parse.FullToken.FullInstruction ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - /* XXX implement this */ - assert(0); - break; - - default: - assert( 0 ); - } - } - - tgsi_parse_free( &parse ); - - DUMP_END(); - - return 1; -} - -#endif /* i386 */ diff --git a/src/gallium/aux/tgsi/exec/tgsi_sse2.h b/src/gallium/aux/tgsi/exec/tgsi_sse2.h deleted file mode 100755 index 9bee371766..0000000000 --- a/src/gallium/aux/tgsi/exec/tgsi_sse2.h +++ /dev/null @@ -1,26 +0,0 @@ -#if !defined TGSI_SSE2_H -#define TGSI_SSE2_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_token; -struct x86_function; - -unsigned -tgsi_emit_sse2( - struct tgsi_token *tokens, - struct x86_function *function ); - -unsigned -tgsi_emit_sse2_fs( - struct tgsi_token *tokens, - struct x86_function *function ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_SSE2_H - diff --git a/src/gallium/aux/tgsi/util/tgsi_build.c b/src/gallium/aux/tgsi/util/tgsi_build.c deleted file mode 100644 index a00ff1c2a5..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_build.c +++ /dev/null @@ -1,1371 +0,0 @@ -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_build.h" -#include "tgsi_parse.h" - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ) -{ - struct tgsi_version version; - - version.MajorVersion = 1; - version.MinorVersion = 1; - version.Padding = 0; - - return version; -} - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ) -{ - struct tgsi_header header; - - header.HeaderSize = 1; - header.BodySize = 0; - - return header; -} - -static void -header_headersize_grow( struct tgsi_header *header ) -{ - assert( header->HeaderSize < 0xFF ); - assert( header->BodySize == 0 ); - - header->HeaderSize++; -} - -static void -header_bodysize_grow( struct tgsi_header *header ) -{ - assert( header->BodySize < 0xFFFFFF ); - - header->BodySize++; -} - -struct tgsi_processor -tgsi_default_processor( void ) -{ - struct tgsi_processor processor; - - processor.Processor = TGSI_PROCESSOR_FRAGMENT; - processor.Padding = 0; - - return processor; -} - -struct tgsi_processor -tgsi_build_processor( - unsigned type, - struct tgsi_header *header ) -{ - struct tgsi_processor processor; - - processor = tgsi_default_processor(); - processor.Processor = type; - - header_headersize_grow( header ); - - return processor; -} - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ) -{ - struct tgsi_declaration declaration; - - declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; - declaration.Size = 1; - declaration.File = TGSI_FILE_NULL; - declaration.Declare = TGSI_DECLARE_RANGE; - declaration.UsageMask = TGSI_WRITEMASK_XYZW; - declaration.Interpolate = 0; - declaration.Semantic = 0; - declaration.Padding = 0; - declaration.Extended = 0; - - return declaration; -} - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned declare, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ) -{ - struct tgsi_declaration declaration; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( declare <= TGSI_DECLARE_MASK ); - - declaration = tgsi_default_declaration(); - declaration.File = file; - declaration.Declare = declare; - declaration.UsageMask = usage_mask; - declaration.Interpolate = interpolate; - declaration.Semantic = semantic; - - header_bodysize_grow( header ); - - return declaration; -} - -static void -declaration_grow( - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - assert( declaration->Size < 0xFF ); - - declaration->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ) -{ - struct tgsi_full_declaration full_declaration; - - full_declaration.Declaration = tgsi_default_declaration(); - full_declaration.Interpolation = tgsi_default_declaration_interpolation(); - full_declaration.Semantic = tgsi_default_declaration_semantic(); - - return full_declaration; -} - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0; - struct tgsi_declaration *declaration; - - if( maxsize <= size ) - return 0; - declaration = (struct tgsi_declaration *) &tokens[size]; - size++; - - *declaration = tgsi_build_declaration( - full_decl->Declaration.File, - full_decl->Declaration.Declare, - full_decl->Declaration.UsageMask, - full_decl->Declaration.Interpolate, - full_decl->Declaration.Semantic, - header ); - - switch( full_decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - { - struct tgsi_declaration_range *dr; - - if( maxsize <= size ) - return 0; - dr = (struct tgsi_declaration_range *) &tokens[size]; - size++; - - *dr = tgsi_build_declaration_range( - full_decl->u.DeclarationRange.First, - full_decl->u.DeclarationRange.Last, - declaration, - header ); - break; - } - - case TGSI_DECLARE_MASK: - { - struct tgsi_declaration_mask *dm; - - if( maxsize <= size ) - return 0; - dm = (struct tgsi_declaration_mask *) &tokens[size]; - size++; - - *dm = tgsi_build_declaration_mask( - full_decl->u.DeclarationMask.Mask, - declaration, - header ); - break; - } - - default: - assert( 0 ); - } - - if( full_decl->Declaration.Interpolate ) { - struct tgsi_declaration_interpolation *di; - - if( maxsize <= size ) - return 0; - di = (struct tgsi_declaration_interpolation *) &tokens[size]; - size++; - - *di = tgsi_build_declaration_interpolation( - full_decl->Interpolation.Interpolate, - declaration, - header ); - } - - if( full_decl->Declaration.Semantic ) { - struct tgsi_declaration_semantic *ds; - - if( maxsize <= size ) - return 0; - ds = (struct tgsi_declaration_semantic *) &tokens[size]; - size++; - - *ds = tgsi_build_declaration_semantic( - full_decl->Semantic.SemanticName, - full_decl->Semantic.SemanticIndex, - declaration, - header ); - } - - return size; -} - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_range declaration_range; - - assert( last >= first ); - assert( last <= 0xFFFF ); - - declaration_range.First = first; - declaration_range.Last = last; - - declaration_grow( declaration, header ); - - return declaration_range; -} - -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_mask declaration_mask; - - declaration_mask.Mask = mask; - - declaration_grow( declaration, header ); - - return declaration_mask; -} - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ) -{ - struct tgsi_declaration_interpolation di; - - di.Interpolate = TGSI_INTERPOLATE_CONSTANT; - di.Padding = 0; - - return di; -} - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_interpolation di; - - assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); - - di = tgsi_default_declaration_interpolation(); - di.Interpolate = interpolate; - - declaration_grow( declaration, header ); - - return di; -} - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ) -{ - struct tgsi_declaration_semantic ds; - - ds.SemanticName = TGSI_SEMANTIC_POSITION; - ds.SemanticIndex = 0; - ds.Padding = 0; - - return ds; -} - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_semantic ds; - - assert( semantic_name <= TGSI_SEMANTIC_COUNT ); - assert( semantic_index <= 0xFFFF ); - - ds = tgsi_default_declaration_semantic(); - ds.SemanticName = semantic_name; - ds.SemanticIndex = semantic_index; - - declaration_grow( declaration, header ); - - return ds; -} - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ) -{ - struct tgsi_immediate immediate; - - immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; - immediate.Size = 1; - immediate.DataType = TGSI_IMM_FLOAT32; - immediate.Padding = 0; - immediate.Extended = 0; - - return immediate; -} - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ) -{ - struct tgsi_immediate immediate; - - immediate = tgsi_default_immediate(); - - header_bodysize_grow( header ); - - return immediate; -} - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ) -{ - struct tgsi_full_immediate fullimm; - - fullimm.Immediate = tgsi_default_immediate(); - fullimm.u.Pointer = (void *) 0; - - return fullimm; -} - -static void -immediate_grow( - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - assert( immediate->Size < 0xFF ); - - immediate->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - float value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - struct tgsi_immediate_float32 immediate_float32; - - immediate_float32.Float = value; - - immediate_grow( immediate, header ); - - return immediate_float32; -} - -unsigned -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0, i; - struct tgsi_immediate *immediate; - - if( maxsize <= size ) - return 0; - immediate = (struct tgsi_immediate *) &tokens[size]; - size++; - - *immediate = tgsi_build_immediate( header ); - - for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { - struct tgsi_immediate_float32 *if32; - - if( maxsize <= size ) - return 0; - if32 = (struct tgsi_immediate_float32 *) &tokens[size]; - size++; - - *if32 = tgsi_build_immediate_float32( - full_imm->u.ImmediateFloat32[i].Float, - immediate, - header ); - } - - return size; -} - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ) -{ - struct tgsi_instruction instruction; - - instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; - instruction.Size = 1; - instruction.Opcode = TGSI_OPCODE_MOV; - instruction.Saturate = TGSI_SAT_NONE; - instruction.NumDstRegs = 1; - instruction.NumSrcRegs = 1; - instruction.Padding = 0; - instruction.Extended = 0; - - return instruction; -} - -struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ) -{ - struct tgsi_instruction instruction; - - assert (opcode <= TGSI_OPCODE_LAST); - assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); - assert (num_dst_regs <= 3); - assert (num_src_regs <= 15); - - instruction = tgsi_default_instruction(); - instruction.Opcode = opcode; - instruction.Saturate = saturate; - instruction.NumDstRegs = num_dst_regs; - instruction.NumSrcRegs = num_src_regs; - - header_bodysize_grow( header ); - - return instruction; -} - -static void -instruction_grow( - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - assert (instruction->Size < 0xFF); - - instruction->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ) -{ - struct tgsi_full_instruction full_instruction; - unsigned i; - - full_instruction.Instruction = tgsi_default_instruction(); - full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); - full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); - full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); - for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { - full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); - } - for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { - full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); - } - - return full_instruction; -} - -unsigned -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0; - unsigned i; - struct tgsi_instruction *instruction; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - instruction = (struct tgsi_instruction *) &tokens[size]; - size++; - - *instruction = tgsi_build_instruction( - full_inst->Instruction.Opcode, - full_inst->Instruction.Saturate, - full_inst->Instruction.NumDstRegs, - full_inst->Instruction.NumSrcRegs, - header ); - prev_token = (struct tgsi_token *) instruction; - - if( tgsi_compare_instruction_ext_nv( - full_inst->InstructionExtNv, - tgsi_default_instruction_ext_nv() ) ) { - struct tgsi_instruction_ext_nv *instruction_ext_nv; - - if( maxsize <= size ) - return 0; - instruction_ext_nv = - (struct tgsi_instruction_ext_nv *) &tokens[size]; - size++; - - *instruction_ext_nv = tgsi_build_instruction_ext_nv( - full_inst->InstructionExtNv.Precision, - full_inst->InstructionExtNv.CondDstIndex, - full_inst->InstructionExtNv.CondFlowIndex, - full_inst->InstructionExtNv.CondMask, - full_inst->InstructionExtNv.CondSwizzleX, - full_inst->InstructionExtNv.CondSwizzleY, - full_inst->InstructionExtNv.CondSwizzleZ, - full_inst->InstructionExtNv.CondSwizzleW, - full_inst->InstructionExtNv.CondDstUpdate, - full_inst->InstructionExtNv.CondFlowEnable, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_nv; - } - - if( tgsi_compare_instruction_ext_label( - full_inst->InstructionExtLabel, - tgsi_default_instruction_ext_label() ) ) { - struct tgsi_instruction_ext_label *instruction_ext_label; - - if( maxsize <= size ) - return 0; - instruction_ext_label = - (struct tgsi_instruction_ext_label *) &tokens[size]; - size++; - - *instruction_ext_label = tgsi_build_instruction_ext_label( - full_inst->InstructionExtLabel.Label, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_label; - } - - if( tgsi_compare_instruction_ext_texture( - full_inst->InstructionExtTexture, - tgsi_default_instruction_ext_texture() ) ) { - struct tgsi_instruction_ext_texture *instruction_ext_texture; - - if( maxsize <= size ) - return 0; - instruction_ext_texture = - (struct tgsi_instruction_ext_texture *) &tokens[size]; - size++; - - *instruction_ext_texture = tgsi_build_instruction_ext_texture( - full_inst->InstructionExtTexture.Texture, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_texture; - } - - for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { - const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; - struct tgsi_dst_register *dst_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - dst_register = (struct tgsi_dst_register *) &tokens[size]; - size++; - - *dst_register = tgsi_build_dst_register( - reg->DstRegister.File, - reg->DstRegister.WriteMask, - reg->DstRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register; - - if( tgsi_compare_dst_register_ext_concode( - reg->DstRegisterExtConcode, - tgsi_default_dst_register_ext_concode() ) ) { - struct tgsi_dst_register_ext_concode *dst_register_ext_concode; - - if( maxsize <= size ) - return 0; - dst_register_ext_concode = - (struct tgsi_dst_register_ext_concode *) &tokens[size]; - size++; - - *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( - reg->DstRegisterExtConcode.CondMask, - reg->DstRegisterExtConcode.CondSwizzleX, - reg->DstRegisterExtConcode.CondSwizzleY, - reg->DstRegisterExtConcode.CondSwizzleZ, - reg->DstRegisterExtConcode.CondSwizzleW, - reg->DstRegisterExtConcode.CondSrcIndex, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_concode; - } - - if( tgsi_compare_dst_register_ext_modulate( - reg->DstRegisterExtModulate, - tgsi_default_dst_register_ext_modulate() ) ) { - struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; - - if( maxsize <= size ) - return 0; - dst_register_ext_modulate = - (struct tgsi_dst_register_ext_modulate *) &tokens[size]; - size++; - - *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( - reg->DstRegisterExtModulate.Modulate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_modulate; - } - } - - for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { - const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; - struct tgsi_src_register *src_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - src_register = (struct tgsi_src_register *) &tokens[size]; - size++; - - *src_register = tgsi_build_src_register( - reg->SrcRegister.File, - reg->SrcRegister.SwizzleX, - reg->SrcRegister.SwizzleY, - reg->SrcRegister.SwizzleZ, - reg->SrcRegister.SwizzleW, - reg->SrcRegister.Negate, - reg->SrcRegister.Indirect, - reg->SrcRegister.Dimension, - reg->SrcRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register; - - if( tgsi_compare_src_register_ext_swz( - reg->SrcRegisterExtSwz, - tgsi_default_src_register_ext_swz() ) ) { - struct tgsi_src_register_ext_swz *src_register_ext_swz; - - if( maxsize <= size ) - return 0; - src_register_ext_swz = - (struct tgsi_src_register_ext_swz *) &tokens[size]; - size++; - - *src_register_ext_swz = tgsi_build_src_register_ext_swz( - reg->SrcRegisterExtSwz.ExtSwizzleX, - reg->SrcRegisterExtSwz.ExtSwizzleY, - reg->SrcRegisterExtSwz.ExtSwizzleZ, - reg->SrcRegisterExtSwz.ExtSwizzleW, - reg->SrcRegisterExtSwz.NegateX, - reg->SrcRegisterExtSwz.NegateY, - reg->SrcRegisterExtSwz.NegateZ, - reg->SrcRegisterExtSwz.NegateW, - reg->SrcRegisterExtSwz.ExtDivide, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_swz; - } - - if( tgsi_compare_src_register_ext_mod( - reg->SrcRegisterExtMod, - tgsi_default_src_register_ext_mod() ) ) { - struct tgsi_src_register_ext_mod *src_register_ext_mod; - - if( maxsize <= size ) - return 0; - src_register_ext_mod = - (struct tgsi_src_register_ext_mod *) &tokens[size]; - size++; - - *src_register_ext_mod = tgsi_build_src_register_ext_mod( - reg->SrcRegisterExtMod.Complement, - reg->SrcRegisterExtMod.Bias, - reg->SrcRegisterExtMod.Scale2X, - reg->SrcRegisterExtMod.Absolute, - reg->SrcRegisterExtMod.Negate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_mod; - } - - if( reg->SrcRegister.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterInd.File, - reg->SrcRegisterInd.SwizzleX, - reg->SrcRegisterInd.SwizzleY, - reg->SrcRegisterInd.SwizzleZ, - reg->SrcRegisterInd.SwizzleW, - reg->SrcRegisterInd.Negate, - reg->SrcRegisterInd.Indirect, - reg->SrcRegisterInd.Dimension, - reg->SrcRegisterInd.Index, - instruction, - header ); - } - - if( reg->SrcRegister.Dimension ) { - struct tgsi_dimension *dim; - - assert( !reg->SrcRegisterDim.Dimension ); - - if( maxsize <= size ) - return 0; - dim = (struct tgsi_dimension *) &tokens[size]; - size++; - - *dim = tgsi_build_dimension( - reg->SrcRegisterDim.Indirect, - reg->SrcRegisterDim.Index, - instruction, - header ); - - if( reg->SrcRegisterDim.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterDimInd.File, - reg->SrcRegisterDimInd.SwizzleX, - reg->SrcRegisterDimInd.SwizzleY, - reg->SrcRegisterDimInd.SwizzleZ, - reg->SrcRegisterDimInd.SwizzleW, - reg->SrcRegisterDimInd.Negate, - reg->SrcRegisterDimInd.Indirect, - reg->SrcRegisterDimInd.Dimension, - reg->SrcRegisterDimInd.Index, - instruction, - header ); - } - } - } - - return size; -} - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; - instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; - instruction_ext_nv.CondDstIndex = 0; - instruction_ext_nv.CondFlowIndex = 0; - instruction_ext_nv.CondMask = TGSI_CC_TR; - instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; - instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; - instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; - instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; - instruction_ext_nv.CondDstUpdate = 0; - instruction_ext_nv.CondFlowEnable = 0; - instruction_ext_nv.Padding = 0; - instruction_ext_nv.Extended = 0; - - return instruction_ext_nv; -} - -union token_u32 -{ - unsigned u32; -}; - -unsigned -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - unsigned precision, - unsigned cond_dst_index, - unsigned cond_flow_index, - unsigned cond_mask, - unsigned cond_swizzle_x, - unsigned cond_swizzle_y, - unsigned cond_swizzle_z, - unsigned cond_swizzle_w, - unsigned cond_dst_update, - unsigned cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv = tgsi_default_instruction_ext_nv(); - instruction_ext_nv.Precision = precision; - instruction_ext_nv.CondDstIndex = cond_dst_index; - instruction_ext_nv.CondFlowIndex = cond_flow_index; - instruction_ext_nv.CondMask = cond_mask; - instruction_ext_nv.CondSwizzleX = cond_swizzle_x; - instruction_ext_nv.CondSwizzleY = cond_swizzle_y; - instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; - instruction_ext_nv.CondSwizzleW = cond_swizzle_w; - instruction_ext_nv.CondDstUpdate = cond_dst_update; - instruction_ext_nv.CondFlowEnable = cond_flow_update; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_nv; -} - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; - instruction_ext_label.Label = 0; - instruction_ext_label.Padding = 0; - instruction_ext_label.Extended = 0; - - return instruction_ext_label; -} - -unsigned -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - unsigned label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label = tgsi_default_instruction_ext_label(); - instruction_ext_label.Label = label; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_label; -} - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; - instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; - instruction_ext_texture.Padding = 0; - instruction_ext_texture.Extended = 0; - - return instruction_ext_texture; -} - -unsigned -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - unsigned texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture = tgsi_default_instruction_ext_texture(); - instruction_ext_texture.Texture = texture; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_texture; -} - -struct tgsi_src_register -tgsi_default_src_register( void ) -{ - struct tgsi_src_register src_register; - - src_register.File = TGSI_FILE_NULL; - src_register.SwizzleX = TGSI_SWIZZLE_X; - src_register.SwizzleY = TGSI_SWIZZLE_Y; - src_register.SwizzleZ = TGSI_SWIZZLE_Z; - src_register.SwizzleW = TGSI_SWIZZLE_W; - src_register.Negate = 0; - src_register.Indirect = 0; - src_register.Dimension = 0; - src_register.Index = 0; - src_register.Extended = 0; - - return src_register; -} - -struct tgsi_src_register -tgsi_build_src_register( - unsigned file, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - unsigned negate, - unsigned indirect, - unsigned dimension, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register src_register; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( swizzle_x <= TGSI_SWIZZLE_W ); - assert( swizzle_y <= TGSI_SWIZZLE_W ); - assert( swizzle_z <= TGSI_SWIZZLE_W ); - assert( swizzle_w <= TGSI_SWIZZLE_W ); - assert( negate <= 1 ); - assert( index >= -0x8000 && index <= 0x7FFF ); - - src_register = tgsi_default_src_register(); - src_register.File = file; - src_register.SwizzleX = swizzle_x; - src_register.SwizzleY = swizzle_y; - src_register.SwizzleZ = swizzle_z; - src_register.SwizzleW = swizzle_w; - src_register.Negate = negate; - src_register.Indirect = indirect; - src_register.Dimension = dimension; - src_register.Index = index; - - instruction_grow( instruction, header ); - - return src_register; -} - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ) -{ - struct tgsi_full_src_register full_src_register; - - full_src_register.SrcRegister = tgsi_default_src_register(); - full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); - full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); - full_src_register.SrcRegisterInd = tgsi_default_src_register(); - full_src_register.SrcRegisterDim = tgsi_default_dimension(); - full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); - - return full_src_register; -} - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; - src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; - src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; - src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; - src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; - src_register_ext_swz.NegateX = 0; - src_register_ext_swz.NegateY = 0; - src_register_ext_swz.NegateZ = 0; - src_register_ext_swz.NegateW = 0; - src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; - src_register_ext_swz.Padding = 0; - src_register_ext_swz.Extended = 0; - - return src_register_ext_swz; -} - -unsigned -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - unsigned ext_swizzle_x, - unsigned ext_swizzle_y, - unsigned ext_swizzle_z, - unsigned ext_swizzle_w, - unsigned negate_x, - unsigned negate_y, - unsigned negate_z, - unsigned negate_w, - unsigned ext_divide, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE ); - assert( negate_x <= 1 ); - assert( negate_y <= 1 ); - assert( negate_z <= 1 ); - assert( negate_w <= 1 ); - assert( ext_divide <= TGSI_EXTSWIZZLE_ONE ); - - src_register_ext_swz = tgsi_default_src_register_ext_swz(); - src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; - src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; - src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; - src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; - src_register_ext_swz.NegateX = negate_x; - src_register_ext_swz.NegateY = negate_y; - src_register_ext_swz.NegateZ = negate_z; - src_register_ext_swz.NegateW = negate_w; - src_register_ext_swz.ExtDivide = ext_divide; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_swz; -} - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; - src_register_ext_mod.Complement = 0; - src_register_ext_mod.Bias = 0; - src_register_ext_mod.Scale2X = 0; - src_register_ext_mod.Absolute = 0; - src_register_ext_mod.Negate = 0; - src_register_ext_mod.Padding = 0; - src_register_ext_mod.Extended = 0; - - return src_register_ext_mod; -} - -unsigned -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - unsigned complement, - unsigned bias, - unsigned scale_2x, - unsigned absolute, - unsigned negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - assert( complement <= 1 ); - assert( bias <= 1 ); - assert( scale_2x <= 1 ); - assert( absolute <= 1 ); - assert( negate <= 1 ); - - src_register_ext_mod = tgsi_default_src_register_ext_mod(); - src_register_ext_mod.Complement = complement; - src_register_ext_mod.Bias = bias; - src_register_ext_mod.Scale2X = scale_2x; - src_register_ext_mod.Absolute = absolute; - src_register_ext_mod.Negate = negate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_mod; -} - -struct tgsi_dimension -tgsi_default_dimension( void ) -{ - struct tgsi_dimension dimension; - - dimension.Indirect = 0; - dimension.Dimension = 0; - dimension.Padding = 0; - dimension.Index = 0; - dimension.Extended = 0; - - return dimension; -} - -struct tgsi_dimension -tgsi_build_dimension( - unsigned indirect, - unsigned index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dimension dimension; - - dimension = tgsi_default_dimension(); - dimension.Indirect = indirect; - dimension.Index = index; - - instruction_grow( instruction, header ); - - return dimension; -} - -struct tgsi_dst_register -tgsi_default_dst_register( void ) -{ - struct tgsi_dst_register dst_register; - - dst_register.File = TGSI_FILE_NULL; - dst_register.WriteMask = TGSI_WRITEMASK_XYZW; - dst_register.Indirect = 0; - dst_register.Dimension = 0; - dst_register.Index = 0; - dst_register.Padding = 0; - dst_register.Extended = 0; - - return dst_register; -} - -struct tgsi_dst_register -tgsi_build_dst_register( - unsigned file, - unsigned mask, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register dst_register; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( mask <= TGSI_WRITEMASK_XYZW ); - assert( index >= -32768 && index <= 32767 ); - - dst_register = tgsi_default_dst_register(); - dst_register.File = file; - dst_register.WriteMask = mask; - dst_register.Index = index; - - instruction_grow( instruction, header ); - - return dst_register; -} - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ) -{ - struct tgsi_full_dst_register full_dst_register; - - full_dst_register.DstRegister = tgsi_default_dst_register(); - full_dst_register.DstRegisterExtConcode = - tgsi_default_dst_register_ext_concode(); - full_dst_register.DstRegisterExtModulate = - tgsi_default_dst_register_ext_modulate(); - - return full_dst_register; -} - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; - dst_register_ext_concode.CondMask = TGSI_CC_TR; - dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; - dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; - dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; - dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; - dst_register_ext_concode.CondSrcIndex = 0; - dst_register_ext_concode.Padding = 0; - dst_register_ext_concode.Extended = 0; - - return dst_register_ext_concode; -} - -unsigned -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - unsigned cc, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - int index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - assert( cc <= TGSI_CC_FL ); - assert( swizzle_x <= TGSI_SWIZZLE_W ); - assert( swizzle_y <= TGSI_SWIZZLE_W ); - assert( swizzle_z <= TGSI_SWIZZLE_W ); - assert( swizzle_w <= TGSI_SWIZZLE_W ); - assert( index >= -32768 && index <= 32767 ); - - dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); - dst_register_ext_concode.CondMask = cc; - dst_register_ext_concode.CondSwizzleX = swizzle_x; - dst_register_ext_concode.CondSwizzleY = swizzle_y; - dst_register_ext_concode.CondSwizzleZ = swizzle_z; - dst_register_ext_concode.CondSwizzleW = swizzle_w; - dst_register_ext_concode.CondSrcIndex = index; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_concode; -} - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; - dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; - dst_register_ext_modulate.Padding = 0; - dst_register_ext_modulate.Extended = 0; - - return dst_register_ext_modulate; -} - -unsigned -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - unsigned modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - assert( modulate <= TGSI_MODULATE_EIGHTH ); - - dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); - dst_register_ext_modulate.Modulate = modulate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_modulate; -} - diff --git a/src/gallium/aux/tgsi/util/tgsi_build.h b/src/gallium/aux/tgsi/util/tgsi_build.h deleted file mode 100644 index 116c78abf3..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_build.h +++ /dev/null @@ -1,320 +0,0 @@ -#if !defined TGSI_BUILD_H -#define TGSI_BUILD_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ); - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ); - -struct tgsi_processor -tgsi_default_processor( void ); - -struct tgsi_processor -tgsi_build_processor( - unsigned processor, - struct tgsi_header *header ); - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ); - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned declare, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ); - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ); - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ); - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ); - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ); - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ); - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ); - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - float value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ); - -unsigned -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ); - -struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ); - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ); - -unsigned -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ); - -unsigned -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ); - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - unsigned precision, - unsigned cond_dst_index, - unsigned cond_flow_index, - unsigned cond_mask, - unsigned cond_swizzle_x, - unsigned cond_swizzle_y, - unsigned cond_swizzle_z, - unsigned cond_swizzle_w, - unsigned cond_dst_update, - unsigned cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ); - -unsigned -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ); - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - unsigned label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ); - -unsigned -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ); - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - unsigned texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register -tgsi_default_src_register( void ); - -struct tgsi_src_register -tgsi_build_src_register( - unsigned file, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - unsigned negate, - unsigned indirect, - unsigned dimension, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ); - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ); - -unsigned -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ); - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - unsigned ext_swizzle_x, - unsigned ext_swizzle_y, - unsigned ext_swizzle_z, - unsigned ext_swizzle_w, - unsigned negate_x, - unsigned negate_y, - unsigned negate_z, - unsigned negate_w, - unsigned ext_divide, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ); - -unsigned -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ); - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - unsigned complement, - unsigned bias, - unsigned scale_2x, - unsigned absolute, - unsigned negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dimension -tgsi_default_dimension( void ); - -struct tgsi_dimension -tgsi_build_dimension( - unsigned indirect, - unsigned index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register -tgsi_default_dst_register( void ); - -struct tgsi_dst_register -tgsi_build_dst_register( - unsigned file, - unsigned mask, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ); - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ); - -unsigned -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ); - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - unsigned cc, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - int index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ); - -unsigned -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ); - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - unsigned modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_BUILD_H - diff --git a/src/gallium/aux/tgsi/util/tgsi_dump.c b/src/gallium/aux/tgsi/util/tgsi_dump.c deleted file mode 100644 index b5c54847e0..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_dump.c +++ /dev/null @@ -1,1581 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_dump.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" - -struct gen_dump -{ - unsigned tabs; - void (* write)( - struct gen_dump *dump, - const void *data, - unsigned size ); -}; - -struct text_dump -{ - struct gen_dump base; - char *text; - unsigned length; - unsigned capacity; -}; - -static void -_text_dump_write( - struct gen_dump *dump, - const void *data, - unsigned size ) -{ - struct text_dump *td = (struct text_dump *) dump; - unsigned new_length = td->length + size; - - if( new_length >= td->capacity ) { - unsigned new_capacity = td->capacity; - - do { - if( new_capacity == 0 ) { - new_capacity = 256; - } - else { - new_capacity *= 2; - } - } while( new_length >= new_capacity ); - td->text = (char *) REALLOC( - td->text, - td->capacity, - new_capacity ); - td->capacity = new_capacity; - } - memcpy( - &td->text[td->length], - data, - size ); - td->length = new_length; - td->text[td->length] = '\0'; -} - -struct file_dump -{ - struct gen_dump base; - FILE *file; -}; - -static void -_file_dump_write( - struct gen_dump *dump, - const void *data, - unsigned size ) -{ - struct file_dump *fd = (struct file_dump *) dump; - -#if 0 - fwrite( data, 1, size, fd->file ); -#else - { - unsigned i; - - for (i = 0; i < size; i++ ) { - fprintf( fd->file, "%c", ((const char *) data)[i] ); - } - } -#endif -} - -static void -gen_dump_str( - struct gen_dump *dump, - const char *str ) -{ - unsigned i; - size_t len = strlen( str ); - - for (i = 0; i < len; i++) { - dump->write( dump, &str[i], 1 ); - if (str[i] == '\n') { - unsigned i; - - for (i = 0; i < dump->tabs; i++) { - dump->write( dump, " ", 4 ); - } - } - } -} - -static void -gen_dump_chr( - struct gen_dump *dump, - const char chr ) -{ - dump->write( dump, &chr, 1 ); -} - -static void -gen_dump_uix( - struct gen_dump *dump, - const unsigned ui ) -{ - char str[36]; - - sprintf( str, "0x%x", ui ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_uid( - struct gen_dump *dump, - const unsigned ui ) -{ - char str[16]; - - sprintf( str, "%u", ui ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_sid( - struct gen_dump *dump, - const int si ) -{ - char str[16]; - - sprintf( str, "%d", si ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_flt( - struct gen_dump *dump, - const float flt ) -{ - char str[48]; - - sprintf( str, "%10.4f", flt ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_enum( - struct gen_dump *dump, - const unsigned e, - const char **enums, - const unsigned enums_count ) -{ - if (e >= enums_count) { - gen_dump_uid( dump, e ); - } - else { - gen_dump_str( dump, enums[e] ); - } -} - -static void -gen_dump_tab( - struct gen_dump *dump ) -{ - ++dump->tabs; -} - -static void -gen_dump_untab( - struct gen_dump *dump ) -{ - assert( dump->tabs > 0 ); - - --dump->tabs; -} - -#define TXT(S) gen_dump_str( dump, S ) -#define CHR(C) gen_dump_chr( dump, C ) -#define UIX(I) gen_dump_uix( dump, I ) -#define UID(I) gen_dump_uid( dump, I ) -#define SID(I) gen_dump_sid( dump, I ) -#define FLT(F) gen_dump_flt( dump, F ) -#define TAB() gen_dump_tab( dump ) -#define UNT() gen_dump_untab( dump ) -#define ENM(E,ENUMS) gen_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) - -static const char *TGSI_PROCESSOR_TYPES[] = -{ - "PROCESSOR_FRAGMENT", - "PROCESSOR_VERTEX", - "PROCESSOR_GEOMETRY" -}; - -static const char *TGSI_PROCESSOR_TYPES_SHORT[] = -{ - "FRAG", - "VERT", - "GEOM" -}; - -static const char *TGSI_TOKEN_TYPES[] = -{ - "TOKEN_TYPE_DECLARATION", - "TOKEN_TYPE_IMMEDIATE", - "TOKEN_TYPE_INSTRUCTION" -}; - -static const char *TGSI_FILES[] = -{ - "FILE_NULL", - "FILE_CONSTANT", - "FILE_INPUT", - "FILE_OUTPUT", - "FILE_TEMPORARY", - "FILE_SAMPLER", - "FILE_ADDRESS", - "FILE_IMMEDIATE" -}; - -static const char *TGSI_FILES_SHORT[] = -{ - "NULL", - "CONST", - "IN", - "OUT", - "TEMP", - "SAMP", - "ADDR", - "IMM" -}; - -static const char *TGSI_DECLARES[] = -{ - "DECLARE_RANGE", - "DECLARE_MASK" -}; - -static const char *TGSI_INTERPOLATES[] = -{ - "INTERPOLATE_CONSTANT", - "INTERPOLATE_LINEAR", - "INTERPOLATE_PERSPECTIVE", - "INTERPOLATE_ATTRIB" -}; - -static const char *TGSI_INTERPOLATES_SHORT[] = -{ - "CONSTANT", - "LINEAR", - "PERSPECTIVE", - "ATTRIB" -}; - -static const char *TGSI_SEMANTICS[] = -{ - "SEMANTIC_POSITION", - "SEMANTIC_COLOR", - "SEMANTIC_BCOLOR", - "SEMANTIC_FOG", - "SEMANTIC_PSIZE", - "SEMANTIC_GENERIC," -}; - -static const char *TGSI_SEMANTICS_SHORT[] = -{ - "POSITION", - "COLOR", - "BCOLOR", - "FOG", - "PSIZE", - "GENERIC", -}; - -static const char *TGSI_IMMS[] = -{ - "IMM_FLOAT32" -}; - -static const char *TGSI_IMMS_SHORT[] = -{ - "FLT32" -}; - -static const char *TGSI_OPCODES[] = -{ - "OPCODE_ARL", - "OPCODE_MOV", - "OPCODE_LIT", - "OPCODE_RCP", - "OPCODE_RSQ", - "OPCODE_EXP", - "OPCODE_LOG", - "OPCODE_MUL", - "OPCODE_ADD", - "OPCODE_DP3", - "OPCODE_DP4", - "OPCODE_DST", - "OPCODE_MIN", - "OPCODE_MAX", - "OPCODE_SLT", - "OPCODE_SGE", - "OPCODE_MAD", - "OPCODE_SUB", - "OPCODE_LERP", - "OPCODE_CND", - "OPCODE_CND0", - "OPCODE_DOT2ADD", - "OPCODE_INDEX", - "OPCODE_NEGATE", - "OPCODE_FRAC", - "OPCODE_CLAMP", - "OPCODE_FLOOR", - "OPCODE_ROUND", - "OPCODE_EXPBASE2", - "OPCODE_LOGBASE2", - "OPCODE_POWER", - "OPCODE_CROSSPRODUCT", - "OPCODE_MULTIPLYMATRIX", - "OPCODE_ABS", - "OPCODE_RCC", - "OPCODE_DPH", - "OPCODE_COS", - "OPCODE_DDX", - "OPCODE_DDY", - "OPCODE_KILP", - "OPCODE_PK2H", - "OPCODE_PK2US", - "OPCODE_PK4B", - "OPCODE_PK4UB", - "OPCODE_RFL", - "OPCODE_SEQ", - "OPCODE_SFL", - "OPCODE_SGT", - "OPCODE_SIN", - "OPCODE_SLE", - "OPCODE_SNE", - "OPCODE_STR", - "OPCODE_TEX", - "OPCODE_TXD", - "OPCODE_UP2H", - "OPCODE_UP2US", - "OPCODE_UP4B", - "OPCODE_UP4UB", - "OPCODE_X2D", - "OPCODE_ARA", - "OPCODE_ARR", - "OPCODE_BRA", - "OPCODE_CAL", - "OPCODE_RET", - "OPCODE_SSG", - "OPCODE_CMP", - "OPCODE_SCS", - "OPCODE_TXB", - "OPCODE_NRM", - "OPCODE_DIV", - "OPCODE_DP2", - "OPCODE_TXL", - "OPCODE_BRK", - "OPCODE_IF", - "OPCODE_LOOP", - "OPCODE_REP", - "OPCODE_ELSE", - "OPCODE_ENDIF", - "OPCODE_ENDLOOP", - "OPCODE_ENDREP", - "OPCODE_PUSHA", - "OPCODE_POPA", - "OPCODE_CEIL", - "OPCODE_I2F", - "OPCODE_NOT", - "OPCODE_TRUNC", - "OPCODE_SHL", - "OPCODE_SHR", - "OPCODE_AND", - "OPCODE_OR", - "OPCODE_MOD", - "OPCODE_XOR", - "OPCODE_SAD", - "OPCODE_TXF", - "OPCODE_TXQ", - "OPCODE_CONT", - "OPCODE_EMIT", - "OPCODE_ENDPRIM", - "OPCODE_BGNLOOP2", - "OPCODE_BGNSUB", - "OPCODE_ENDLOOP2", - "OPCODE_ENDSUB", - "OPCODE_NOISE1", - "OPCODE_NOISE2", - "OPCODE_NOISE3", - "OPCODE_NOISE4", - "OPCODE_NOP", - "OPCODE_TEXBEM", - "OPCODE_TEXBEML", - "OPCODE_TEXREG2AR", - "OPCODE_TEXM3X2PAD", - "OPCODE_TEXM3X2TEX", - "OPCODE_TEXM3X3PAD", - "OPCODE_TEXM3X3TEX", - "OPCODE_TEXM3X3SPEC", - "OPCODE_TEXM3X3VSPEC", - "OPCODE_TEXREG2GB", - "OPCODE_TEXREG2RGB", - "OPCODE_TEXDP3TEX", - "OPCODE_TEXDP3", - "OPCODE_TEXM3X3", - "OPCODE_TEXM3X2DEPTH", - "OPCODE_TEXDEPTH", - "OPCODE_BEM", - "OPCODE_M4X3", - "OPCODE_M3X4", - "OPCODE_M3X3", - "OPCODE_M3X2", - "OPCODE_NRM4", - "OPCODE_CALLNZ", - "OPCODE_IFC", - "OPCODE_BREAKC", - "OPCODE_KIL", - "OPCODE_END" -}; - -static const char *TGSI_OPCODES_SHORT[] = -{ - "ARL", - "MOV", - "LIT", - "RCP", - "RSQ", - "EXP", - "LOG", - "MUL", - "ADD", - "DP3", - "DP4", - "DST", - "MIN", - "MAX", - "SLT", - "SGE", - "MAD", - "SUB", - "LERP", - "CND", - "CND0", - "DOT2ADD", - "INDEX", - "NEGATE", - "FRAC", - "CLAMP", - "FLOOR", - "ROUND", - "EXPBASE2", - "LOGBASE2", - "POWER", - "CROSSPRODUCT", - "MULTIPLYMATRIX", - "ABS", - "RCC", - "DPH", - "COS", - "DDX", - "DDY", - "KILP", - "PK2H", - "PK2US", - "PK4B", - "PK4UB", - "RFL", - "SEQ", - "SFL", - "SGT", - "SIN", - "SLE", - "SNE", - "STR", - "TEX", - "TXD", - "UP2H", - "UP2US", - "UP4B", - "UP4UB", - "X2D", - "ARA", - "ARR", - "BRA", - "CAL", - "RET", - "SSG", - "CMP", - "SCS", - "TXB", - "NRM", - "DIV", - "DP2", - "TXL", - "BRK", - "IF", - "LOOP", - "REP", - "ELSE", - "ENDIF", - "ENDLOOP", - "ENDREP", - "PUSHA", - "POPA", - "CEIL", - "I2F", - "NOT", - "TRUNC", - "SHL", - "SHR", - "AND", - "OR", - "MOD", - "XOR", - "SAD", - "TXF", - "TXQ", - "CONT", - "EMIT", - "ENDPRIM", - "BGNLOOP2", - "BGNSUB", - "ENDLOOP2", - "ENDSUB", - "NOISE1", - "NOISE2", - "NOISE3", - "NOISE4", - "NOP", - "TEXBEM", - "TEXBEML", - "TEXREG2AR", - "TEXM3X2PAD", - "TEXM3X2TEX", - "TEXM3X3PAD", - "TEXM3X3TEX", - "TEXM3X3SPEC", - "TEXM3X3VSPEC", - "TEXREG2GB", - "TEXREG2RGB", - "TEXDP3TEX", - "TEXDP3", - "TEXM3X3", - "TEXM3X2DEPTH", - "TEXDEPTH", - "BEM", - "M4X3", - "M3X4", - "M3X3", - "M3X2", - "NRM4", - "CALLNZ", - "IFC", - "BREAKC", - "KIL", - "END" -}; - -static const char *TGSI_SATS[] = -{ - "SAT_NONE", - "SAT_ZERO_ONE", - "SAT_MINUS_PLUS_ONE" -}; - -static const char *TGSI_INSTRUCTION_EXTS[] = -{ - "INSTRUCTION_EXT_TYPE_NV", - "INSTRUCTION_EXT_TYPE_LABEL", - "INSTRUCTION_EXT_TYPE_TEXTURE" -}; - -static const char *TGSI_PRECISIONS[] = -{ - "PRECISION_DEFAULT", - "TGSI_PRECISION_FLOAT32", - "TGSI_PRECISION_FLOAT16", - "TGSI_PRECISION_FIXED12" -}; - -static const char *TGSI_CCS[] = -{ - "CC_GT", - "CC_EQ", - "CC_LT", - "CC_UN", - "CC_GE", - "CC_LE", - "CC_NE", - "CC_TR", - "CC_FL" -}; - -static const char *TGSI_SWIZZLES[] = -{ - "SWIZZLE_X", - "SWIZZLE_Y", - "SWIZZLE_Z", - "SWIZZLE_W" -}; - -static const char *TGSI_SWIZZLES_SHORT[] = -{ - "x", - "y", - "z", - "w" -}; - -static const char *TGSI_TEXTURES[] = -{ - "TEXTURE_UNKNOWN", - "TEXTURE_1D", - "TEXTURE_2D", - "TEXTURE_3D", - "TEXTURE_CUBE", - "TEXTURE_RECT", - "TEXTURE_SHADOW1D", - "TEXTURE_SHADOW2D", - "TEXTURE_SHADOWRECT" -}; - -static const char *TGSI_SRC_REGISTER_EXTS[] = -{ - "SRC_REGISTER_EXT_TYPE_SWZ", - "SRC_REGISTER_EXT_TYPE_MOD" -}; - -static const char *TGSI_EXTSWIZZLES[] = -{ - "EXTSWIZZLE_X", - "EXTSWIZZLE_Y", - "EXTSWIZZLE_Z", - "EXTSWIZZLE_W", - "EXTSWIZZLE_ZERO", - "EXTSWIZZLE_ONE" -}; - -static const char *TGSI_EXTSWIZZLES_SHORT[] = -{ - "x", - "y", - "z", - "w", - "0", - "1" -}; - -static const char *TGSI_WRITEMASKS[] = -{ - "0", - "WRITEMASK_X", - "WRITEMASK_Y", - "WRITEMASK_XY", - "WRITEMASK_Z", - "WRITEMASK_XZ", - "WRITEMASK_YZ", - "WRITEMASK_XYZ", - "WRITEMASK_W", - "WRITEMASK_XW", - "WRITEMASK_YW", - "WRITEMASK_XYW", - "WRITEMASK_ZW", - "WRITEMASK_XZW", - "WRITEMASK_YZW", - "WRITEMASK_XYZW" -}; - -static const char *TGSI_DST_REGISTER_EXTS[] = -{ - "DST_REGISTER_EXT_TYPE_CONDCODE", - "DST_REGISTER_EXT_TYPE_MODULATE" -}; - -static const char *TGSI_MODULATES[] = -{ - "MODULATE_1X", - "MODULATE_2X", - "MODULATE_4X", - "MODULATE_8X", - "MODULATE_HALF", - "MODULATE_QUARTER", - "MODULATE_EIGHTH" -}; - -static void -dump_declaration_short( - struct gen_dump *dump, - struct tgsi_full_declaration *decl ) -{ - TXT( "\nDCL " ); - ENM( decl->Declaration.File, TGSI_FILES_SHORT ); - - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - CHR( '[' ); - UID( decl->u.DeclarationRange.First ); - if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { - TXT( ".." ); - UID( decl->u.DeclarationRange.Last ); - } - CHR( ']' ); - break; - default: - assert( 0 ); - } - - if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { - CHR( '.' ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'x' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'w' ); - } - } - - if( decl->Declaration.Interpolate ) { - TXT( ", " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); - } - - if( decl->Declaration.Semantic ) { - TXT( ", " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); - CHR( '[' ); - UID( decl->Semantic.SemanticIndex ); - CHR( ']' ); - } -} - -static void -dump_declaration_verbose( - struct gen_dump *dump, - struct tgsi_full_declaration *decl, - unsigned ignored, - unsigned deflt, - struct tgsi_full_declaration *fd ) -{ - TXT( "\nFile : " ); - ENM( decl->Declaration.File, TGSI_FILES ); - TXT( "\nDeclare : " ); - ENM( decl->Declaration.Declare, TGSI_DECLARES ); - if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { - TXT( "\nUsageMask : " ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'X' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'Y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'Z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'W' ); - } - } - if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { - TXT( "\nInterpolate: " ); - UID( decl->Declaration.Interpolate ); - } - if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { - TXT( "\nSemantic : " ); - UID( decl->Declaration.Semantic ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Declaration.Padding ); - } - - CHR( '\n' ); - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - TXT( "\nFirst: " ); - UID( decl->u.DeclarationRange.First ); - TXT( "\nLast : " ); - UID( decl->u.DeclarationRange.Last ); - break; - - case TGSI_DECLARE_MASK: - TXT( "\nMask: " ); - UIX( decl->u.DeclarationMask.Mask ); - break; - - default: - assert( 0 ); - } - - if( decl->Declaration.Interpolate ) { - CHR( '\n' ); - TXT( "\nInterpolate: " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Interpolation.Padding ); - } - } - - if( decl->Declaration.Semantic ) { - CHR( '\n' ); - TXT( "\nSemanticName : " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); - TXT( "\nSemanticIndex: " ); - UID( decl->Semantic.SemanticIndex ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Semantic.Padding ); - } - } -} - -static void -dump_immediate_short( - struct gen_dump *dump, - struct tgsi_full_immediate *imm ) -{ - unsigned i; - - TXT( "\nIMM " ); - ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); - - TXT( " { " ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - - if( i < imm->Immediate.Size - 2 ) { - TXT( ", " ); - } - } - TXT( " }" ); -} - -static void -dump_immediate_verbose( - struct gen_dump *dump, - struct tgsi_full_immediate *imm, - unsigned ignored ) -{ - unsigned i; - - TXT( "\nDataType : " ); - ENM( imm->Immediate.DataType, TGSI_IMMS ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( imm->Immediate.Padding ); - } - - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - CHR( '\n' ); - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - TXT( "\nFloat: " ); - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - } -} - -static void -dump_instruction_short( - struct gen_dump *dump, - struct tgsi_full_instruction *inst, - unsigned instno ) -{ - unsigned i; - boolean first_reg = TRUE; - - CHR( '\n' ); - UID( instno ); - CHR( ':' ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); - - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - TXT( "_SAT" ); - break; - case TGSI_SAT_MINUS_PLUS_ONE: - TXT( "_SAT[-1,1]" ); - break; - default: - assert( 0 ); - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if( !first_reg ) { - CHR( ',' ); - } - CHR( ' ' ); - - ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); - SID( dst->DstRegister.Index ); - CHR( ']' ); - - if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { - CHR( '.' ); - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { - CHR( 'x' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { - CHR( 'y' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { - CHR( 'z' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { - CHR( 'w' ); - } - } - - first_reg = FALSE; - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - - if( !first_reg ) { - CHR( ',' ); - } - CHR( ' ' ); - - if( src->SrcRegisterExtMod.Complement ) { - TXT( "(1 - " ); - } - if( src->SrcRegisterExtMod.Negate ) { - CHR( '-' ); - } - if( src->SrcRegisterExtMod.Absolute ) { - CHR( '|' ); - } - if( src->SrcRegister.Negate ) { - CHR( '-' ); - } - - ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); - SID( src->SrcRegister.Index ); - CHR( ']' ); - - if (src->SrcRegister.Extended) { - if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || - src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || - src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || - src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { - CHR( '.' ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); - } - } - else if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || - src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || - src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || - src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { - CHR( '.' ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); - } - - if( src->SrcRegisterExtMod.Absolute ) { - CHR( '|' ); - } - if( src->SrcRegisterExtMod.Complement ) { - CHR( ')' ); - } - - first_reg = FALSE; - } - - switch( inst->Instruction.Opcode ) { - case TGSI_OPCODE_IF: - case TGSI_OPCODE_ELSE: - case TGSI_OPCODE_BGNLOOP2: - case TGSI_OPCODE_ENDLOOP2: - case TGSI_OPCODE_CAL: - TXT( " :" ); - UID( inst->InstructionExtLabel.Label ); - break; - } -} - -static void -dump_instruction_verbose( - struct gen_dump *dump, - struct tgsi_full_instruction *inst, - unsigned ignored, - unsigned deflt, - struct tgsi_full_instruction *fi ) -{ - unsigned i; - - TXT( "\nOpcode : " ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES ); - if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { - TXT( "\nSaturate : " ); - ENM( inst->Instruction.Saturate, TGSI_SATS ); - } - if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { - TXT( "\nNumDstRegs : " ); - UID( inst->Instruction.NumDstRegs ); - } - if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { - TXT( "\nNumSrcRegs : " ); - UID( inst->Instruction.NumSrcRegs ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->Instruction.Padding ); - } - - if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { - TXT( "\nPrecision : " ); - ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); - } - if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { - TXT( "\nCondDstIndex : " ); - UID( inst->InstructionExtNv.CondDstIndex ); - } - if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { - TXT( "\nCondFlowIndex : " ); - UID( inst->InstructionExtNv.CondFlowIndex ); - } - if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { - TXT( "\nCondMask : " ); - ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { - TXT( "\nCondSwizzleX : " ); - ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { - TXT( "\nCondSwizzleY : " ); - ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ : " ); - ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { - TXT( "\nCondSwizzleW : " ); - ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { - TXT( "\nCondDstUpdate : " ); - UID( inst->InstructionExtNv.CondDstUpdate ); - } - if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { - TXT( "\nCondFlowEnable: " ); - UID( inst->InstructionExtNv.CondFlowEnable ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtNv.Padding ); - if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { - TXT( "\nExtended : " ); - UID( inst->InstructionExtNv.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { - TXT( "\nLabel : " ); - UID( inst->InstructionExtLabel.Label ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtLabel.Padding ); - if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtLabel.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { - TXT( "\nTexture : " ); - ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtTexture.Padding ); - if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtTexture.Extended ); - } - } - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : " ); - ENM( dst->DstRegister.File, TGSI_FILES ); - if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { - TXT( "\nWriteMask: " ); - ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); - } - if( ignored ) { - if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( dst->DstRegister.Indirect ); - } - if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( dst->DstRegister.Dimension ); - } - } - if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { - TXT( "\nIndex : " ); - SID( dst->DstRegister.Index ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegister.Padding ); - if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegister.Extended ); - } - } - - if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { - TXT( "\nCondMask : " ); - ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { - TXT( "\nCondSwizzleX: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { - TXT( "\nCondSwizzleY: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { - TXT( "\nCondSwizzleW: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { - TXT( "\nCondSrcIndex: " ); - UID( dst->DstRegisterExtConcode.CondSrcIndex ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtConcode.Padding ); - if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegisterExtConcode.Extended ); - } - } - } - - if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { - TXT( "\nModulate: " ); - ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtModulate.Padding ); - if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { - TXT( "\nExtended: " ); - UID( dst->DstRegisterExtModulate.Extended ); - } - } - } - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; - - CHR( '\n' ); - TXT( "\nFile : "); - ENM( src->SrcRegister.File, TGSI_FILES ); - if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { - TXT( "\nSwizzleX : " ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { - TXT( "\nSwizzleY : " ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { - TXT( "\nSwizzleZ : " ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { - TXT( "\nSwizzleW : " ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegister.Negate ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( src->SrcRegister.Indirect ); - } - if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( src->SrcRegister.Dimension ); - } - } - if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { - TXT( "\nIndex : " ); - SID( src->SrcRegister.Index ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegister.Extended ); - } - } - - if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { - TXT( "\nExtSwizzleX: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { - TXT( "\nExtSwizzleY: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { - TXT( "\nExtSwizzleZ: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { - TXT( "\nExtSwizzleW: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { - TXT( "\nNegateX : " ); - UID( src->SrcRegisterExtSwz.NegateX ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { - TXT( "\nNegateY : " ); - UID( src->SrcRegisterExtSwz.NegateY ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { - TXT( "\nNegateZ : " ); - UID( src->SrcRegisterExtSwz.NegateZ ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { - TXT( "\nNegateW : " ); - UID( src->SrcRegisterExtSwz.NegateW ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { - TXT( "\nExtDivide : " ); - ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtSwz.Padding ); - if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtSwz.Extended ); - } - } - } - - if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { - CHR( '\n' ); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { - TXT( "\nComplement: " ); - UID( src->SrcRegisterExtMod.Complement ); - } - if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { - TXT( "\nBias : " ); - UID( src->SrcRegisterExtMod.Bias ); - } - if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { - TXT( "\nScale2X : " ); - UID( src->SrcRegisterExtMod.Scale2X ); - } - if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { - TXT( "\nAbsolute : " ); - UID( src->SrcRegisterExtMod.Absolute ); - } - if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegisterExtMod.Negate ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtMod.Padding ); - if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtMod.Extended ); - } - } - } - } -} - -static void -dump_gen( - struct gen_dump *dump, - const struct tgsi_token *tokens, - unsigned flags ) -{ - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned verbose = flags & TGSI_DUMP_VERBOSE; - unsigned ignored = !(flags & TGSI_DUMP_NO_IGNORED); - unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); - unsigned instno = 0; - - dump->tabs = 0; - - /* sanity check */ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); - - tgsi_parse_init( &parse, tokens ); - - TXT( "tgsi-dump begin -----------------" ); - - CHR( '\n' ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); - CHR( ' ' ); - UID( parse.FullVersion.Version.MajorVersion ); - CHR( '.' ); - UID( parse.FullVersion.Version.MinorVersion ); - - if( verbose ) { - TXT( "\nMajorVersion: " ); - UID( parse.FullVersion.Version.MajorVersion ); - TXT( "\nMinorVersion: " ); - UID( parse.FullVersion.Version.MinorVersion ); - CHR( '\n' ); - - TXT( "\nHeaderSize: " ); - UID( parse.FullHeader.Header.HeaderSize ); - TXT( "\nBodySize : " ); - UID( parse.FullHeader.Header.BodySize ); - TXT( "\nProcessor : " ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); - CHR( '\n' ); - } - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_short( - dump, - &parse.FullToken.FullDeclaration ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_short( - dump, - &parse.FullToken.FullImmediate ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_short( - dump, - &parse.FullToken.FullInstruction, - instno ); - instno++; - break; - - default: - assert( 0 ); - } - - if( verbose ) { - TXT( "\nType : " ); - ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); - if( ignored ) { - TXT( "\nSize : " ); - UID( parse.FullToken.Token.Size ); - if( deflt || parse.FullToken.Token.Extended ) { - TXT( "\nExtended : " ); - UID( parse.FullToken.Token.Extended ); - } - } - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_verbose( - dump, - &parse.FullToken.FullDeclaration, - ignored, - deflt, - &fd ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_verbose( - dump, - &parse.FullToken.FullImmediate, - ignored ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_verbose( - dump, - &parse.FullToken.FullInstruction, - ignored, - deflt, - &fi ); - break; - - default: - assert( 0 ); - } - - CHR( '\n' ); - } - } - - TXT( "\ntgsi-dump end -------------------\n" ); - - tgsi_parse_free( &parse ); -} - - -static void -sanity_checks(void) -{ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); - assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "END") == 0); -} - - -void -tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ) -{ - struct file_dump dump; - - sanity_checks(); - - dump.base.write = _file_dump_write; -#if 0 - { - static unsigned counter = 0; - char buffer[64]; - sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); - dump.file = fopen( buffer, "wt" ); - } -#else - dump.file = stderr; -#endif - - dump_gen( - &dump.base, - tokens, - flags ); - -#if 0 - fclose( dump.file ); -#endif -} - -void -tgsi_dump_str( - char **str, - const struct tgsi_token *tokens, - unsigned flags ) -{ - struct text_dump dump; - - dump.base.write = _text_dump_write; - dump.text = NULL; - dump.length = 0; - dump.capacity = 0; - - dump_gen( - &dump.base, - tokens, - flags ); - - *str = dump.text; -} diff --git a/src/gallium/aux/tgsi/util/tgsi_dump.h b/src/gallium/aux/tgsi/util/tgsi_dump.h deleted file mode 100644 index 1adc9db251..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_dump.h +++ /dev/null @@ -1,28 +0,0 @@ -#if !defined TGSI_DUMP_H -#define TGSI_DUMP_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -#define TGSI_DUMP_VERBOSE 1 -#define TGSI_DUMP_NO_IGNORED 2 -#define TGSI_DUMP_NO_DEFAULT 4 - -void -tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ); - -void -tgsi_dump_str( - char **str, - const struct tgsi_token *tokens, - unsigned flags ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_DUMP_H - diff --git a/src/gallium/aux/tgsi/util/tgsi_parse.c b/src/gallium/aux/tgsi/util/tgsi_parse.c deleted file mode 100644 index bf6b89ce56..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_parse.c +++ /dev/null @@ -1,319 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ) -{ - full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; -} - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ) -{ - if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { - FREE( full_token->FullImmediate.u.Pointer ); - } -} - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ) -{ - ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; - if( ctx->FullVersion.Version.MajorVersion > 1 ) { - return TGSI_PARSE_ERROR; - } - - ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; - if( ctx->FullHeader.Header.HeaderSize >= 2 ) { - ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; - } - else { - ctx->FullHeader.Processor = tgsi_default_processor(); - } - - ctx->Tokens = tokens; - ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; - - tgsi_full_token_init( &ctx->FullToken ); - - return TGSI_PARSE_OK; -} - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ) -{ - tgsi_full_token_free( &ctx->FullToken ); -} - -boolean -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ) -{ - return ctx->Position >= - 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; -} - -static void -next_token( - struct tgsi_parse_context *ctx, - void *token ) -{ - assert( !tgsi_parse_end_of_tokens( ctx ) ); - - *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; -} - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ) -{ - struct tgsi_token token; - unsigned i; - - tgsi_full_token_free( &ctx->FullToken ); - tgsi_full_token_init( &ctx->FullToken ); - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; - - *decl = tgsi_default_full_declaration(); - decl->Declaration = *(struct tgsi_declaration *) &token; - - switch( decl->Declaration.Type ) { - case TGSI_DECLARE_RANGE: - next_token( ctx, &decl->u.DeclarationRange ); - break; - - case TGSI_DECLARE_MASK: - next_token( ctx, &decl->u.DeclarationMask ); - break; - - default: - assert (0); - } - - if( decl->Declaration.Interpolate ) { - next_token( ctx, &decl->Interpolation ); - } - - if( decl->Declaration.Semantic ) { - next_token( ctx, &decl->Semantic ); - } - - break; - } - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; - - *imm = tgsi_default_full_immediate(); - imm->Immediate = *(struct tgsi_immediate *) &token; - - assert( !imm->Immediate.Extended ); - - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - imm->u.Pointer = MALLOC( - sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - next_token( ctx, &imm->u.ImmediateFloat32[i] ); - } - break; - - default: - assert( 0 ); - } - - break; - } - - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; - unsigned extended; - - *inst = tgsi_default_full_instruction(); - inst->Instruction = *(struct tgsi_instruction *) &token; - - extended = inst->Instruction.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_INSTRUCTION_EXT_TYPE_NV: - inst->InstructionExtNv = - *(struct tgsi_instruction_ext_nv *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_LABEL: - inst->InstructionExtLabel = - *(struct tgsi_instruction_ext_label *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: - inst->InstructionExtTexture = - *(struct tgsi_instruction_ext_texture *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); - assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); - - extended = inst->FullDstRegisters[i].DstRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: - inst->FullDstRegisters[i].DstRegisterExtConcode = - *(struct tgsi_dst_register_ext_concode *) &token; - break; - - case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: - inst->FullDstRegisters[i].DstRegisterExtModulate = - *(struct tgsi_dst_register_ext_modulate *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - } - - assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); - - extended = inst->FullSrcRegisters[i].SrcRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: - inst->FullSrcRegisters[i].SrcRegisterExtSwz = - *(struct tgsi_src_register_ext_swz *) &token; - break; - - case TGSI_SRC_REGISTER_EXT_TYPE_MOD: - inst->FullSrcRegisters[i].SrcRegisterExtMod = - *(struct tgsi_src_register_ext_mod *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - - if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); - - /* - * No support for multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); - - if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - } - } - - break; - } - - default: - assert( 0 ); - } -} - diff --git a/src/gallium/aux/tgsi/util/tgsi_parse.h b/src/gallium/aux/tgsi/util/tgsi_parse.h deleted file mode 100644 index 9372da8d5d..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_parse.h +++ /dev/null @@ -1,121 +0,0 @@ -#if !defined TGSI_PARSE_H -#define TGSI_PARSE_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -struct tgsi_full_version -{ - struct tgsi_version Version; -}; - -struct tgsi_full_header -{ - struct tgsi_header Header; - struct tgsi_processor Processor; -}; - -struct tgsi_full_dst_register -{ - struct tgsi_dst_register DstRegister; - struct tgsi_dst_register_ext_concode DstRegisterExtConcode; - struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; -}; - -struct tgsi_full_src_register -{ - struct tgsi_src_register SrcRegister; - struct tgsi_src_register_ext_swz SrcRegisterExtSwz; - struct tgsi_src_register_ext_mod SrcRegisterExtMod; - struct tgsi_src_register SrcRegisterInd; - struct tgsi_dimension SrcRegisterDim; - struct tgsi_src_register SrcRegisterDimInd; -}; - -struct tgsi_full_declaration -{ - struct tgsi_declaration Declaration; - union - { - struct tgsi_declaration_range DeclarationRange; - struct tgsi_declaration_mask DeclarationMask; - } u; - struct tgsi_declaration_interpolation Interpolation; - struct tgsi_declaration_semantic Semantic; -}; - -struct tgsi_full_immediate -{ - struct tgsi_immediate Immediate; - union - { - void *Pointer; - struct tgsi_immediate_float32 *ImmediateFloat32; - } u; -}; - -#define TGSI_FULL_MAX_DST_REGISTERS 2 -#define TGSI_FULL_MAX_SRC_REGISTERS 3 - -struct tgsi_full_instruction -{ - struct tgsi_instruction Instruction; - struct tgsi_instruction_ext_nv InstructionExtNv; - struct tgsi_instruction_ext_label InstructionExtLabel; - struct tgsi_instruction_ext_texture InstructionExtTexture; - struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; - struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; -}; - -union tgsi_full_token -{ - struct tgsi_token Token; - struct tgsi_full_declaration FullDeclaration; - struct tgsi_full_immediate FullImmediate; - struct tgsi_full_instruction FullInstruction; -}; - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ); - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ); - -struct tgsi_parse_context -{ - const struct tgsi_token *Tokens; - unsigned Position; - struct tgsi_full_version FullVersion; - struct tgsi_full_header FullHeader; - union tgsi_full_token FullToken; -}; - -#define TGSI_PARSE_OK 0 -#define TGSI_PARSE_ERROR 1 - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ); - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ); - -boolean -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ); - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_PARSE_H - diff --git a/src/gallium/aux/tgsi/util/tgsi_transform.c b/src/gallium/aux/tgsi/util/tgsi_transform.c deleted file mode 100644 index 357f77b05a..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_transform.c +++ /dev/null @@ -1,199 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * TGSI program transformation utility. - * - * Authors: Brian Paul - */ - - -#include "tgsi_transform.h" - - - -static void -emit_instruction(struct tgsi_transform_context *ctx, - const struct tgsi_full_instruction *inst) -{ - uint ti = ctx->ti; - - ti += tgsi_build_full_instruction(inst, - ctx->tokens_out + ti, - ctx->header, - ctx->max_tokens_out - ti); - ctx->ti = ti; -} - - -static void -emit_declaration(struct tgsi_transform_context *ctx, - const struct tgsi_full_declaration *decl) -{ - uint ti = ctx->ti; - - ti += tgsi_build_full_declaration(decl, - ctx->tokens_out + ti, - ctx->header, - ctx->max_tokens_out - ti); - ctx->ti = ti; -} - - -static void -emit_immediate(struct tgsi_transform_context *ctx, - const struct tgsi_full_immediate *imm) -{ - uint ti = ctx->ti; - - ti += tgsi_build_full_immediate(imm, - ctx->tokens_out + ti, - ctx->header, - ctx->max_tokens_out - ti); - ctx->ti = ti; -} - - - -/** - * Apply user-defined transformations to the input shader to produce - * the output shader. - * For example, a register search-and-replace operation could be applied - * by defining a transform_instruction() callback that examined and changed - * the instruction src/dest regs. - * - * \return number of tokens emitted - */ -int -tgsi_transform_shader(const struct tgsi_token *tokens_in, - struct tgsi_token *tokens_out, - uint max_tokens_out, - struct tgsi_transform_context *ctx) -{ - uint procType; - - /* input shader */ - struct tgsi_parse_context parse; - - /* output shader */ - struct tgsi_processor *processor; - - - /** - ** callback context init - **/ - ctx->emit_instruction = emit_instruction; - ctx->emit_declaration = emit_declaration; - ctx->emit_immediate = emit_immediate; - ctx->tokens_out = tokens_out; - ctx->max_tokens_out = max_tokens_out; - - - /** - ** Setup to begin parsing input shader - **/ - if (tgsi_parse_init( &parse, tokens_in ) != TGSI_PARSE_OK) { - debug_printf("tgsi_parse_init() failed in tgsi_transform_shader()!\n"); - return -1; - } - procType = parse.FullHeader.Processor.Processor; - assert(procType == TGSI_PROCESSOR_FRAGMENT || - procType == TGSI_PROCESSOR_VERTEX || - procType == TGSI_PROCESSOR_GEOMETRY); - - - /** - ** Setup output shader - **/ - *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version(); - - ctx->header = (struct tgsi_header *) (tokens_out + 1); - *ctx->header = tgsi_build_header(); - - processor = (struct tgsi_processor *) (tokens_out + 2); - *processor = tgsi_build_processor( procType, ctx->header ); - - ctx->ti = 3; - - - /** - ** Loop over incoming program tokens/instructions - */ - while( !tgsi_parse_end_of_tokens( &parse ) ) { - - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *fullinst - = &parse.FullToken.FullInstruction; - - if (ctx->transform_instruction) - ctx->transform_instruction(ctx, fullinst); - else - ctx->emit_instruction(ctx, fullinst); - } - break; - - case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *fulldecl - = &parse.FullToken.FullDeclaration; - - if (ctx->transform_declaration) - ctx->transform_declaration(ctx, fulldecl); - else - ctx->emit_declaration(ctx, fulldecl); - } - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - struct tgsi_full_immediate *fullimm - = &parse.FullToken.FullImmediate; - - if (ctx->transform_immediate) - ctx->transform_immediate(ctx, fullimm); - else - ctx->emit_immediate(ctx, fullimm); - } - break; - - default: - assert( 0 ); - } - } - - if (ctx->epilog) { - ctx->epilog(ctx); - } - - tgsi_parse_free (&parse); - - return ctx->ti; -} diff --git a/src/gallium/aux/tgsi/util/tgsi_transform.h b/src/gallium/aux/tgsi/util/tgsi_transform.h deleted file mode 100644 index fcf85d603b..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_transform.h +++ /dev/null @@ -1,93 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_TRANSFORM_H -#define TGSI_TRANSFORM_H - - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_build.h" - - - -/** - * Subclass this to add caller-specific data - */ -struct tgsi_transform_context -{ -/**** PUBLIC ***/ - - /** - * User-defined callbacks invoked per instruction. - */ - void (*transform_instruction)(struct tgsi_transform_context *ctx, - struct tgsi_full_instruction *inst); - - void (*transform_declaration)(struct tgsi_transform_context *ctx, - struct tgsi_full_declaration *decl); - - void (*transform_immediate)(struct tgsi_transform_context *ctx, - struct tgsi_full_immediate *imm); - - /** - * Called at end of input program to allow caller to append extra - * instructions. Return number of tokens emitted. - */ - void (*epilog)(struct tgsi_transform_context *ctx); - - -/*** PRIVATE ***/ - - /** - * These are setup by tgsi_transform_shader() and cannot be overridden. - * Meant to be called from in the above user callback functions. - */ - void (*emit_instruction)(struct tgsi_transform_context *ctx, - const struct tgsi_full_instruction *inst); - void (*emit_declaration)(struct tgsi_transform_context *ctx, - const struct tgsi_full_declaration *decl); - void (*emit_immediate)(struct tgsi_transform_context *ctx, - const struct tgsi_full_immediate *imm); - - struct tgsi_header *header; - uint max_tokens_out; - struct tgsi_token *tokens_out; - uint ti; -}; - - - -extern int -tgsi_transform_shader(const struct tgsi_token *tokens_in, - struct tgsi_token *tokens_out, - uint max_tokens_out, - struct tgsi_transform_context *ctx); - - -#endif /* TGSI_TRANSFORM_H */ diff --git a/src/gallium/aux/tgsi/util/tgsi_util.c b/src/gallium/aux/tgsi/util/tgsi_util.c deleted file mode 100644 index 4cdd89182a..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_util.c +++ /dev/null @@ -1,274 +0,0 @@ -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" -#include "tgsi_util.h" - -union pointer_hack -{ - void *pointer; - unsigned long long uint64; -}; - -void * -tgsi_align_128bit( - void *unaligned ) -{ - union pointer_hack ph; - - ph.uint64 = 0; - ph.pointer = unaligned; - ph.uint64 = (ph.uint64 + 15) & ~15; - return ph.pointer; -} - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->SwizzleX; - case 1: - return reg->SwizzleY; - case 2: - return reg->SwizzleZ; - case 3: - return reg->SwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->ExtSwizzleX; - case 1: - return reg->ExtSwizzleY; - case 2: - return reg->ExtSwizzleZ; - case 3: - return reg->ExtSwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned swizzle; - - /* - * First, calculate the extended swizzle for a given channel. This will give - * us either a channel index into the simple swizzle or a constant 1 or 0. - */ - swizzle = tgsi_util_get_src_register_extswizzle( - ®->SrcRegisterExtSwz, - component ); - - assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); - assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); - assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); - assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); - - /* - * Second, calculate the simple swizzle for the unswizzled channel index. - * Leave the constants intact, they are not affected by the simple swizzle. - */ - if( swizzle <= TGSI_SWIZZLE_W ) { - swizzle = tgsi_util_get_src_register_swizzle( - ®->SrcRegister, - component ); - } - - return swizzle; -} - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - unsigned swizzle, - unsigned component ) -{ - switch( component ) { - case 0: - reg->SwizzleX = swizzle; - break; - case 1: - reg->SwizzleY = swizzle; - break; - case 2: - reg->SwizzleZ = swizzle; - break; - case 3: - reg->SwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - unsigned swizzle, - unsigned component ) -{ - switch( component ) { - case 0: - reg->ExtSwizzleX = swizzle; - break; - case 1: - reg->ExtSwizzleY = swizzle; - break; - case 2: - reg->ExtSwizzleZ = swizzle; - break; - case 3: - reg->ExtSwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->NegateX; - case 1: - return reg->NegateY; - case 2: - return reg->NegateZ; - case 3: - return reg->NegateW; - default: - assert( 0 ); - } - return 0; -} - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ) -{ - switch( component ) { - case 0: - reg->NegateX = negate; - break; - case 1: - reg->NegateY = negate; - break; - case 2: - reg->NegateZ = negate; - break; - case 3: - reg->NegateW = negate; - break; - default: - assert( 0 ); - } -} - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned sign_mode; - - if( reg->SrcRegisterExtMod.Absolute ) { - /* Consider only the post-abs negation. */ - - if( reg->SrcRegisterExtMod.Negate ) { - sign_mode = TGSI_UTIL_SIGN_SET; - } - else { - sign_mode = TGSI_UTIL_SIGN_CLEAR; - } - } - else { - /* Accumulate the three negations. */ - - unsigned negate; - - negate = reg->SrcRegister.Negate; - if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { - negate = !negate; - } - if( reg->SrcRegisterExtMod.Negate ) { - negate = !negate; - } - - if( negate ) { - sign_mode = TGSI_UTIL_SIGN_TOGGLE; - } - else { - sign_mode = TGSI_UTIL_SIGN_KEEP; - } - } - - return sign_mode; -} - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - unsigned sign_mode ) -{ - reg->SrcRegisterExtSwz.NegateX = 0; - reg->SrcRegisterExtSwz.NegateY = 0; - reg->SrcRegisterExtSwz.NegateZ = 0; - reg->SrcRegisterExtSwz.NegateW = 0; - - switch (sign_mode) - { - case TGSI_UTIL_SIGN_CLEAR: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_SET: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 1; - break; - - case TGSI_UTIL_SIGN_TOGGLE: - reg->SrcRegister.Negate = 1; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_KEEP: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - default: - assert( 0 ); - } -} - diff --git a/src/gallium/aux/tgsi/util/tgsi_util.h b/src/gallium/aux/tgsi/util/tgsi_util.h deleted file mode 100644 index ef14446f0e..0000000000 --- a/src/gallium/aux/tgsi/util/tgsi_util.h +++ /dev/null @@ -1,70 +0,0 @@ -#if !defined TGSI_UTIL_H -#define TGSI_UTIL_H - -#if defined __cplusplus -extern "C" { -#endif // defined __cplusplus - -void * -tgsi_align_128bit( - void *unaligned ); - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component); - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - unsigned swizzle, - unsigned component ); - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - unsigned swizzle, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ); - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ); - -#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ -#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ -#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ -#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - unsigned sign_mode ); - -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus - -#endif // !defined TGSI_UTIL_H - diff --git a/src/gallium/aux/util/p_debug.c b/src/gallium/aux/util/p_debug.c deleted file mode 100644 index b9607a6ba7..0000000000 --- a/src/gallium/aux/util/p_debug.c +++ /dev/null @@ -1,76 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include - -#ifdef WIN32 -#include -#include -#else -#include -#include -#endif - -#include "pipe/p_debug.h" -#include "pipe/p_compiler.h" - - -void debug_vprintf(const char *format, va_list ap) -{ -#ifdef WIN32 - EngDebugPrint("Gallium3D: ", (PCHAR)format, ap); -#else - vfprintf(stderr, format, ap); -#endif -} - - -void debug_printf(const char *format, ...) -{ - va_list ap; - va_start(ap, format); - debug_vprintf(format, ap); - va_end(ap); -} - - -static INLINE void debug_abort(void) -{ -#ifdef WIN32 - EngDebugBreak(); -#else - abort(); -#endif -} - - -void debug_assert_fail(const char *expr, const char *file, unsigned line) -{ - debug_printf("%s:%i: Assertion `%s' failed.\n", file, line, expr); - debug_abort(); -} diff --git a/src/gallium/aux/util/p_tile.c b/src/gallium/aux/util/p_tile.c deleted file mode 100644 index 3f795a3898..0000000000 --- a/src/gallium/aux/util/p_tile.c +++ /dev/null @@ -1,699 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * RGBA/float tile get/put functions. - * Usable both by drivers and state trackers. - * Surfaces should already be in a mapped state. - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" - -#include "p_tile.h" - - - -/** - * Move raw block of pixels from surface to user memory. - * This should be usable by any hw driver that has mappable surfaces. - */ -void -pipe_get_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride) -{ - const uint cpp = ps->cpp; - const ubyte *pSrc; - const uint src_stride = ps->pitch * cpp; - ubyte *pDest; - uint i; - - if (dst_stride == 0) { - dst_stride = w * cpp; - } - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - pSrc = (const ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; - pDest = (ubyte *) p; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w * cpp); - pDest += dst_stride; - pSrc += src_stride; - } - - pipe_surface_unmap(ps); -} - - -/** - * Move raw block of pixels from user memory to surface. - * This should be usable by any hw driver that has mappable surfaces. - */ -void -pipe_put_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride) -{ - const uint cpp = ps->cpp; - const ubyte *pSrc; - const uint dst_stride = ps->pitch * cpp; - ubyte *pDest; - uint i; - - if (src_stride == 0) { - src_stride = w * cpp; - } - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - pSrc = (const ubyte *) p; - pDest = (ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; - - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w * cpp); - pDest += dst_stride; - pSrc += src_stride; - } - - pipe_surface_unmap(ps); -} - - - - -/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */ -#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) - -#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ - us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) - - - -/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ - -static void -a8r8g8b8_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const unsigned pixel = *src++; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - } - p += dst_stride; - } -} - - -static void -a8r8g8b8_put_tile_rgba(unsigned *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - *dst++ = (a << 24) | (r << 16) | (g << 8) | b; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ - -static void -b8g8r8a8_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const unsigned pixel = *src++; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - } - p += dst_stride; - } -} - - -static void -b8g8r8a8_put_tile_rgba(unsigned *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - *dst++ = (b << 24) | (g << 16) | (r << 8) | a; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ - -static void -a1r5g5b5_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const ushort pixel = *src++; - pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); - pRow[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); - pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - pRow[3] = ((pixel >> 15) ) * 1.0f; - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ - -static void -a4r4g4b4_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const ushort pixel = *src++; - pRow[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); - pRow[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); - pRow[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); - pRow[3] = ((pixel >> 12) ) * (1.0f / 15.0f); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_R5G6B5_UNORM ***/ - -static void -r5g6b5_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const ushort pixel = *src++; - pRow[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); - pRow[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); - pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - pRow[3] = 1.0f; - } - p += dst_stride; - } -} - - -static void -r5g5b5_put_tile_rgba(ushort *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); - uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); - uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); - *dst++ = (r << 11) | (g << 5) | (b); - } - p += src_stride; - } -} - - - -/*** PIPE_FORMAT_Z16_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z16_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const float scale = 1.0f / 65535.0f; - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = *src++ * scale; - } - p += dst_stride; - } -} - - - - -/*** PIPE_FORMAT_U_L8 ***/ - -static void -l8_get_tile_rgba(ubyte *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(*src); - pRow[3] = 1.0; - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_U_A8 ***/ - -static void -a8_get_tile_rgba(ubyte *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = 0.0; - pRow[3] = UBYTE_TO_FLOAT(*src); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ - -static void -r16g16b16a16_get_tile_rgba(short *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src += 4, pRow += 4) { - pRow[0] = SHORT_TO_FLOAT(src[0]); - pRow[1] = SHORT_TO_FLOAT(src[1]); - pRow[2] = SHORT_TO_FLOAT(src[2]); - pRow[3] = SHORT_TO_FLOAT(src[3]); - } - p += dst_stride; - } -} - - -static void -r16g16b16a16_put_tile_rgba(short *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, dst += 4, pRow += 4) { - UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); - UNCLAMPED_FLOAT_TO_SHORT(dst[1], pRow[1]); - UNCLAMPED_FLOAT_TO_SHORT(dst[2], pRow[2]); - UNCLAMPED_FLOAT_TO_SHORT(dst[3], pRow[3]); - } - p += src_stride; - } -} - - - -/*** PIPE_FORMAT_U_I8 ***/ - -static void -i8_get_tile_rgba(ubyte *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = UBYTE_TO_FLOAT(*src); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_U_A8_L8 ***/ - -static void -a8_l8_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - ushort p = *src++; - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(p & 0xff); - pRow[3] = UBYTE_TO_FLOAT(p >> 8); - } - p += dst_stride; - } -} - - - - -/*** PIPE_FORMAT_Z32_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z32_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const double scale = 1.0 / (double) 0xffffffff; - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (float) (*src++ * scale); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_S8Z24_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -s8z24_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (float) (scale * (*src++ & 0xffffff)); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_Z24S8_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -z24s8_get_tile_rgba(unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (float) (scale * (*src++ >> 8)); - } - p += dst_stride; - } -} - - -void -pipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p) -{ - unsigned dst_stride = w * 4; - void *packed; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - packed = MALLOC(h * w * ps->cpp); - - if (!packed) - return; - - pipe_get_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); - - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - a1r5g5b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_A4R4G4B4_UNORM: - a4r4g4b4_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_R5G6B5_UNORM: - r5g6b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_U_L8: - l8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_U_A8: - a8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_U_I8: - i8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_U_A8_L8: - a8_l8_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_get_tile_rgba((short *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_Z16_UNORM: - z16_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_Z32_UNORM: - z32_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_S8Z24_UNORM: - s8z24_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - case PIPE_FORMAT_Z24S8_UNORM: - z24s8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); - break; - default: - assert(0); - } - - FREE(packed); -} - - -void -pipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p) -{ - unsigned src_stride = w * 4; - void *packed; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - packed = MALLOC(h * w * ps->cpp); - - if (!packed) - return; - - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - /*a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_R5G6B5_UNORM: - r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_R8G8B8A8_UNORM: - break; - case PIPE_FORMAT_U_L8: - /*l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_U_A8: - /*a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_U_I8: - /*i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_U_A8_L8: - /*a8_l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_Z16_UNORM: - /*z16_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_Z32_UNORM: - /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_S8Z24_UNORM: - /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_Z24S8_UNORM: - /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ - break; - default: - assert(0); - } - - pipe_put_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); - - FREE(packed); -} diff --git a/src/gallium/aux/util/p_tile.h b/src/gallium/aux/util/p_tile.h deleted file mode 100644 index cd8124bf11..0000000000 --- a/src/gallium/aux/util/p_tile.h +++ /dev/null @@ -1,81 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef P_TILE_H -#define P_TILE_H - -#include "pipe/p_compiler.h" - -struct pipe_context; -struct pipe_surface; - - -/** - * Clip tile against surface dims. - * \return TRUE if tile is totally clipped, FALSE otherwise - */ -static INLINE boolean -pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps) -{ - if (x >= ps->width) - return TRUE; - if (y >= ps->height) - return TRUE; - if (x + *w > ps->width) - *w = ps->width - x; - if (y + *h > ps->height) - *h = ps->height - y; - return FALSE; -} - - -extern void -pipe_get_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride); - -extern void -pipe_put_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride); - - -extern void -pipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p); - -extern void -pipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p); - -#endif diff --git a/src/gallium/aux/util/p_util.c b/src/gallium/aux/util/p_util.c deleted file mode 100644 index 2a92f8e408..0000000000 --- a/src/gallium/aux/util/p_util.c +++ /dev/null @@ -1,73 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Miscellaneous utility functions. - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" - - -/** - * Copy 2D rect from one place to another. - * Position and sizes are in pixels. - */ -void -pipe_copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte * src, - int src_pitch, - unsigned src_x, - int src_y) -{ - unsigned i; - - dst_pitch *= cpp; - src_pitch *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * src_pitch; - width *= cpp; - - if (width == dst_pitch && width == src_pitch) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; - } - } -} diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile new file mode 100644 index 0000000000..da68498aa1 --- /dev/null +++ b/src/gallium/auxiliary/Makefile @@ -0,0 +1,24 @@ +TOP = ../../.. +include $(TOP)/configs/current + + +ifeq ($(CONFIG_NAME), linux-llvm) +LLVM_DIR = llvm +endif + +SUBDIRS = pipebuffer $(LLVM_DIR) + + +default: subdirs + + +subdirs: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +clean: + rm -f `find . -name \*.[oa]` diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c new file mode 100644 index 0000000000..9e77e0774d --- /dev/null +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -0,0 +1,181 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Zack Rusin + */ + +#include "cso_cache.h" +#include "cso_hash.h" + +#if 1 +static unsigned hash_key(const void *key, unsigned key_size) +{ + unsigned *ikey = (unsigned *)key; + unsigned hash = 0, i; + + assert(key_size % 4 == 0); + + /* I'm sure this can be improved on: + */ + for (i = 0; i < key_size/4; i++) + hash ^= ikey[i]; + + return hash; +} +#else +static unsigned hash_key(const unsigned char *p, int n) +{ + unsigned h = 0; + unsigned g; + + while (n--) { + h = (h << 4) + *p++; + if ((g = (h & 0xf0000000)) != 0) + h ^= g >> 23; + h &= ~g; + } + return h; +} +#endif + +unsigned cso_construct_key(void *item, int item_size) +{ + return hash_key((item), item_size); +} + +static struct cso_hash *_cso_hash_for_type(struct cso_cache *sc, enum cso_cache_type type) +{ + struct cso_hash *hash = 0; + + switch(type) { + case CSO_BLEND: + hash = sc->blend_hash; + break; + case CSO_SAMPLER: + hash = sc->sampler_hash; + break; + case CSO_DEPTH_STENCIL_ALPHA: + hash = sc->depth_stencil_hash; + break; + case CSO_RASTERIZER: + hash = sc->rasterizer_hash; + break; + case CSO_FRAGMENT_SHADER: + hash = sc->fs_hash; + break; + case CSO_VERTEX_SHADER: + hash = sc->vs_hash; + break; + } + + return hash; +} + +static int _cso_size_for_type(enum cso_cache_type type) +{ + switch(type) { + case CSO_BLEND: + return sizeof(struct pipe_blend_state); + case CSO_SAMPLER: + return sizeof(struct pipe_sampler_state); + case CSO_DEPTH_STENCIL_ALPHA: + return sizeof(struct pipe_depth_stencil_alpha_state); + case CSO_RASTERIZER: + return sizeof(struct pipe_rasterizer_state); + case CSO_FRAGMENT_SHADER: + return sizeof(struct pipe_shader_state); + case CSO_VERTEX_SHADER: + return sizeof(struct pipe_shader_state); + } + return 0; +} + +struct cso_hash_iter +cso_insert_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *state) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + return cso_hash_insert(hash, hash_key, state); +} + +struct cso_hash_iter +cso_find_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + + return cso_hash_find(hash, hash_key); +} + +struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *templ) +{ + struct cso_hash_iter iter = cso_find_state(sc, hash_key, type); + int size = _cso_size_for_type(type); + while (!cso_hash_iter_is_null(iter)) { + void *iter_data = cso_hash_iter_data(iter); + if (!memcmp(iter_data, templ, size)) + return iter; + iter = cso_hash_iter_next(iter); + } + return iter; +} + +void * cso_take_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type) +{ + struct cso_hash *hash = _cso_hash_for_type(sc, type); + return cso_hash_take(hash, hash_key); +} + +struct cso_cache *cso_cache_create(void) +{ + struct cso_cache *sc = malloc(sizeof(struct cso_cache)); + + sc->blend_hash = cso_hash_create(); + sc->sampler_hash = cso_hash_create(); + sc->depth_stencil_hash = cso_hash_create(); + sc->rasterizer_hash = cso_hash_create(); + sc->fs_hash = cso_hash_create(); + sc->vs_hash = cso_hash_create(); + + return sc; +} + +void cso_cache_delete(struct cso_cache *sc) +{ + assert(sc); + cso_hash_delete(sc->blend_hash); + cso_hash_delete(sc->sampler_hash); + cso_hash_delete(sc->depth_stencil_hash); + cso_hash_delete(sc->rasterizer_hash); + cso_hash_delete(sc->fs_hash); + cso_hash_delete(sc->vs_hash); + free(sc); +} diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h new file mode 100644 index 0000000000..116e2eaa2c --- /dev/null +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -0,0 +1,107 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#ifndef CSO_CACHE_H +#define CSO_CACHE_H + +#include "pipe/p_context.h" +#include "pipe/p_state.h" + + +struct cso_hash; + +struct cso_cache { + struct cso_hash *blend_hash; + struct cso_hash *depth_stencil_hash; + struct cso_hash *fs_hash; + struct cso_hash *vs_hash; + struct cso_hash *rasterizer_hash; + struct cso_hash *sampler_hash; +}; + +struct cso_blend { + struct pipe_blend_state state; + void *data; +}; + +struct cso_depth_stencil_alpha { + struct pipe_depth_stencil_alpha_state state; + void *data; +}; + +struct cso_rasterizer { + struct pipe_rasterizer_state state; + void *data; +}; + +struct cso_fragment_shader { + struct pipe_shader_state state; + void *data; +}; + +struct cso_vertex_shader { + struct pipe_shader_state state; + void *data; +}; + +struct cso_sampler { + struct pipe_sampler_state state; + void *data; +}; + + +enum cso_cache_type { + CSO_BLEND, + CSO_SAMPLER, + CSO_DEPTH_STENCIL_ALPHA, + CSO_RASTERIZER, + CSO_FRAGMENT_SHADER, + CSO_VERTEX_SHADER +}; + +unsigned cso_construct_key(void *item, int item_size); + +struct cso_cache *cso_cache_create(void); +void cso_cache_delete(struct cso_cache *sc); + +struct cso_hash_iter cso_insert_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *state); +struct cso_hash_iter cso_find_state(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type); +struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, + unsigned hash_key, enum cso_cache_type type, + void *templ); +void * cso_take_state(struct cso_cache *sc, unsigned hash_key, + enum cso_cache_type type); + +#endif diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c new file mode 100644 index 0000000000..0338cb3b47 --- /dev/null +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -0,0 +1,388 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#include "cso_hash.h" + +#include +#include +#include +#include + +#define MAX(a, b) ((a > b) ? (a) : (b)) + +static const int MinNumBits = 4; + +static const unsigned char prime_deltas[] = { + 0, 0, 1, 3, 1, 5, 3, 3, 1, 9, 7, 5, 3, 9, 25, 3, + 1, 21, 3, 21, 7, 15, 9, 5, 3, 29, 15, 0, 0, 0, 0, 0 +}; + +static int primeForNumBits(int numBits) +{ + return (1 << numBits) + prime_deltas[numBits]; +} + +/* + Returns the smallest integer n such that + primeForNumBits(n) >= hint. +*/ +static int countBits(int hint) +{ + int numBits = 0; + int bits = hint; + + while (bits > 1) { + bits >>= 1; + numBits++; + } + + if (numBits >= (int)sizeof(prime_deltas)) { + numBits = sizeof(prime_deltas) - 1; + } else if (primeForNumBits(numBits) < hint) { + ++numBits; + } + return numBits; +} + +struct cso_node { + struct cso_node *next; + unsigned key; + void *value; +}; + +struct cso_hash_data { + struct cso_node *fakeNext; + struct cso_node **buckets; + int size; + int nodeSize; + short userNumBits; + short numBits; + int numBuckets; +}; + +struct cso_hash { + union { + struct cso_hash_data *d; + struct cso_node *e; + } data; +}; + +static void *cso_data_allocate_node(struct cso_hash_data *hash) +{ + return malloc(hash->nodeSize); +} + +static void cso_data_free_node(struct cso_node *node) +{ + /* XXX still a leak here. + * Need to cast value ptr to original cso type, then free the + * driver-specific data hanging off of it. For example: + struct cso_sampler *csamp = (struct cso_sampler *) node->value; + free(csamp->data); + */ + free(node->value); + free(node); +} + +static struct cso_node * +cso_hash_create_node(struct cso_hash *hash, + unsigned akey, void *avalue, + struct cso_node **anextNode) +{ + struct cso_node *node = cso_data_allocate_node(hash->data.d); + node->key = akey; + node->value = avalue; + + node->next = (struct cso_node*)(*anextNode); + *anextNode = node; + ++hash->data.d->size; + return node; +} + +static void cso_data_rehash(struct cso_hash_data *hash, int hint) +{ + if (hint < 0) { + hint = countBits(-hint); + if (hint < MinNumBits) + hint = MinNumBits; + hash->userNumBits = hint; + while (primeForNumBits(hint) < (hash->size >> 1)) + ++hint; + } else if (hint < MinNumBits) { + hint = MinNumBits; + } + + if (hash->numBits != hint) { + struct cso_node *e = (struct cso_node *)(hash); + struct cso_node **oldBuckets = hash->buckets; + int oldNumBuckets = hash->numBuckets; + int i = 0; + + hash->numBits = hint; + hash->numBuckets = primeForNumBits(hint); + hash->buckets = malloc(sizeof(struct cso_node*) * hash->numBuckets); + for (i = 0; i < hash->numBuckets; ++i) + hash->buckets[i] = e; + + for (i = 0; i < oldNumBuckets; ++i) { + struct cso_node *firstNode = oldBuckets[i]; + while (firstNode != e) { + unsigned h = firstNode->key; + struct cso_node *lastNode = firstNode; + while (lastNode->next != e && lastNode->next->key == h) + lastNode = lastNode->next; + + struct cso_node *afterLastNode = lastNode->next; + struct cso_node **beforeFirstNode = &hash->buckets[h % hash->numBuckets]; + while (*beforeFirstNode != e) + beforeFirstNode = &(*beforeFirstNode)->next; + lastNode->next = *beforeFirstNode; + *beforeFirstNode = firstNode; + firstNode = afterLastNode; + } + } + free(oldBuckets); + } +} + +static void cso_data_might_grow(struct cso_hash_data *hash) +{ + if (hash->size >= hash->numBuckets) + cso_data_rehash(hash, hash->numBits + 1); +} + +static void cso_data_has_shrunk(struct cso_hash_data *hash) +{ + if (hash->size <= (hash->numBuckets >> 3) && + hash->numBits > hash->userNumBits) { + int max = MAX(hash->numBits-2, hash->userNumBits); + cso_data_rehash(hash, max); + } +} + +static struct cso_node *cso_data_first_node(struct cso_hash_data *hash) +{ + struct cso_node *e = (struct cso_node *)(hash); + struct cso_node **bucket = hash->buckets; + int n = hash->numBuckets; + while (n--) { + if (*bucket != e) + return *bucket; + ++bucket; + } + return e; +} + +static struct cso_node **cso_hash_find_node(struct cso_hash *hash, unsigned akey) +{ + struct cso_node **node; + + if (hash->data.d->numBuckets) { + node = (struct cso_node **)(&hash->data.d->buckets[akey % hash->data.d->numBuckets]); + assert(*node == hash->data.e || (*node)->next); + while (*node != hash->data.e && (*node)->key != akey) + node = &(*node)->next; + } else { + node = (struct cso_node **)((const struct cso_node * const *)(&hash->data.e)); + } + return node; +} + +struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, + unsigned key, void *data) +{ + cso_data_might_grow(hash->data.d); + + struct cso_node **nextNode = cso_hash_find_node(hash, key); + struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); + struct cso_hash_iter iter = {hash, node}; + return iter; +} + +struct cso_hash * cso_hash_create(void) +{ + struct cso_hash *hash = malloc(sizeof(struct cso_hash)); + hash->data.d = malloc(sizeof(struct cso_hash_data)); + hash->data.d->fakeNext = 0; + hash->data.d->buckets = 0; + hash->data.d->size = 0; + hash->data.d->nodeSize = sizeof(struct cso_node); + hash->data.d->userNumBits = MinNumBits; + hash->data.d->numBits = 0; + hash->data.d->numBuckets = 0; + + return hash; +} + +void cso_hash_delete(struct cso_hash *hash) +{ + struct cso_node *e_for_x = (struct cso_node *)(hash->data.d); + struct cso_node **bucket = (struct cso_node **)(hash->data.d->buckets); + int n = hash->data.d->numBuckets; + while (n--) { + struct cso_node *cur = *bucket++; + while (cur != e_for_x) { + struct cso_node *next = cur->next; + cso_data_free_node(cur); + cur = next; + } + } + free(hash->data.d->buckets); + free(hash->data.d); + free(hash); +} + +struct cso_hash_iter cso_hash_find(struct cso_hash *hash, + unsigned key) +{ + struct cso_node **nextNode = cso_hash_find_node(hash, key); + struct cso_hash_iter iter = {hash, *nextNode}; + return iter; +} + +unsigned cso_hash_iter_key(struct cso_hash_iter iter) +{ + if (!iter.node || iter.hash->data.e == iter.node) + return 0; + return iter.node->key; +} + +void * cso_hash_iter_data(struct cso_hash_iter iter) +{ + if (!iter.node || iter.hash->data.e == iter.node) + return 0; + return iter.node->value; +} + +static struct cso_node *cso_hash_data_next(struct cso_node *node) +{ + union { + struct cso_node *next; + struct cso_node *e; + struct cso_hash_data *d; + } a; + a.next = node->next; + if (!a.next) { + fprintf(stderr, "iterating beyond the last element\n"); + return 0; + } + if (a.next->next) + return a.next; + + int start = (node->key % a.d->numBuckets) + 1; + struct cso_node **bucket = a.d->buckets + start; + int n = a.d->numBuckets - start; + while (n--) { + if (*bucket != a.e) + return *bucket; + ++bucket; + } + return a.e; +} + + +static struct cso_node *cso_hash_data_prev(struct cso_node *node) +{ + union { + struct cso_node *e; + struct cso_hash_data *d; + } a; + + a.e = node; + while (a.e->next) + a.e = a.e->next; + + int start; + if (node == a.e) + start = a.d->numBuckets - 1; + else + start = node->key % a.d->numBuckets; + + struct cso_node *sentinel = node; + struct cso_node **bucket = a.d->buckets + start; + while (start >= 0) { + if (*bucket != sentinel) { + struct cso_node *prev = *bucket; + while (prev->next != sentinel) + prev = prev->next; + return prev; + } + + sentinel = a.e; + --bucket; + --start; + } + fprintf(stderr, "iterating backward beyond first element\n"); + return a.e; +} + +struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter) +{ + struct cso_hash_iter next = {iter.hash, cso_hash_data_next(iter.node)}; + return next; +} + +int cso_hash_iter_is_null(struct cso_hash_iter iter) +{ + if (!iter.node || iter.node == iter.hash->data.e) + return 1; + return 0; +} + +void * cso_hash_take(struct cso_hash *hash, + unsigned akey) +{ + struct cso_node **node = cso_hash_find_node(hash, akey); + if (*node != hash->data.e) { + void *t = (*node)->value; + struct cso_node *next = (*node)->next; + cso_data_free_node(*node); + *node = next; + --hash->data.d->size; + cso_data_has_shrunk(hash->data.d); + return t; + } + return 0; +} + +struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter) +{ + struct cso_hash_iter prev = {iter.hash, + cso_hash_data_prev(iter.node)}; + return prev; +} + +struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash) +{ + struct cso_hash_iter iter = {hash, cso_data_first_node(hash->data.d)}; + return iter; +} diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h new file mode 100644 index 0000000000..b4aa111860 --- /dev/null +++ b/src/gallium/auxiliary/cso_cache/cso_hash.h @@ -0,0 +1,62 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + */ + +#ifndef CSO_HASH_H +#define CSO_HASH_H + +struct cso_hash; +struct cso_node; + +struct cso_hash_iter { + struct cso_hash *hash; + struct cso_node *node; +}; + +struct cso_hash *cso_hash_create(void); +void cso_hash_delete(struct cso_hash *hash); + +struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, + void *data); +void *cso_hash_take(struct cso_hash *hash, unsigned key); + +struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash); +struct cso_hash_iter cso_hash_find(struct cso_hash *hash, unsigned key); + + +int cso_hash_iter_is_null(struct cso_hash_iter iter); +unsigned cso_hash_iter_key(struct cso_hash_iter iter); +void *cso_hash_iter_data(struct cso_hash_iter iter); + +struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter); +struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); + +#endif diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile new file mode 100644 index 0000000000..451911a354 --- /dev/null +++ b/src/gallium/auxiliary/draw/Makefile @@ -0,0 +1,2 @@ +default: + cd .. ; make diff --git a/src/gallium/auxiliary/draw/draw_clip.c b/src/gallium/auxiliary/draw/draw_clip.c new file mode 100644 index 0000000000..e3051507ea --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_clip.c @@ -0,0 +1,488 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Clipping stage + * + * \author Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + +#include "draw_context.h" +#include "draw_private.h" + + +#ifndef IS_NEGATIVE +#define IS_NEGATIVE(X) ((X) < 0.0) +#endif + +#ifndef DIFFERENT_SIGNS +#define DIFFERENT_SIGNS(x, y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F) +#endif + +#ifndef MAX_CLIPPED_VERTICES +#define MAX_CLIPPED_VERTICES ((2 * (6 + PIPE_MAX_CLIP_PLANES))+1) +#endif + + + +struct clipper { + struct draw_stage stage; /**< base class */ + + /* Basically duplicate some of the flatshading logic here: + */ + boolean flat; + uint num_color_attribs; + uint color_attribs[4]; /* front/back primary/secondary colors */ + + float (*plane)[4]; +}; + + +/* This is a bit confusing: + */ +static INLINE struct clipper *clipper_stage( struct draw_stage *stage ) +{ + return (struct clipper *)stage; +} + + +#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) + + +/* All attributes are float[4], so this is easy: + */ +static void interp_attr( float *fdst, + float t, + const float *fin, + const float *fout ) +{ + fdst[0] = LINTERP( t, fout[0], fin[0] ); + fdst[1] = LINTERP( t, fout[1], fin[1] ); + fdst[2] = LINTERP( t, fout[2], fin[2] ); + fdst[3] = LINTERP( t, fout[3], fin[3] ); +} + +static void copy_colors( struct draw_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + const struct clipper *clipper = clipper_stage(stage); + uint i; + for (i = 0; i < clipper->num_color_attribs; i++) { + const uint attr = clipper->color_attribs[i]; + COPY_4FV(dst->data[attr], src->data[attr]); + } +} + + + +/* Interpolate between two vertices to produce a third. + */ +static void interp( const struct clipper *clip, + struct vertex_header *dst, + float t, + const struct vertex_header *out, + const struct vertex_header *in ) +{ + const unsigned nr_attrs = clip->stage.draw->num_vs_outputs; + unsigned j; + + /* Vertex header. + */ + { + dst->clipmask = 0; + dst->edgeflag = 0; + dst->pad = 0; + dst->vertex_id = UNDEFINED_VERTEX_ID; + } + + /* Clip coordinates: interpolate normally + */ + { + interp_attr(dst->clip, t, in->clip, out->clip); + } + + /* Do the projective divide and insert window coordinates: + */ + { + const float *pos = dst->clip; + const float *scale = clip->stage.draw->viewport.scale; + const float *trans = clip->stage.draw->viewport.translate; + const float oow = 1.0f / pos[3]; + + dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; + dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; + dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; + dst->data[0][3] = oow; + } + + /* Other attributes + * Note: start at 1 to skip winpos (data[0]) since we just computed + * it above. + */ + for (j = 1; j < nr_attrs; j++) { + interp_attr(dst->data[j], t, in->data[j], out->data[j]); + } +} + + +static void emit_poly( struct draw_stage *stage, + struct vertex_header **inlist, + unsigned n, + const struct prim_header *origPrim) +{ + struct prim_header header; + unsigned i; + + /* later stages may need the determinant, but only the sign matters */ + header.det = origPrim->det; + + for (i = 2; i < n; i++) { + header.v[0] = inlist[i-1]; + header.v[1] = inlist[i]; + header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ + + { + unsigned tmp1 = header.v[1]->edgeflag; + unsigned tmp2 = header.v[2]->edgeflag; + + if (i != n-1) header.v[1]->edgeflag = 0; + if (i != 2) header.v[2]->edgeflag = 0; + + header.edgeflags = ((header.v[0]->edgeflag << 0) | + (header.v[1]->edgeflag << 1) | + (header.v[2]->edgeflag << 2)); + + stage->next->tri( stage->next, &header ); + + header.v[1]->edgeflag = tmp1; + header.v[2]->edgeflag = tmp2; + } + } +} + + + + +/* Clip a triangle against the viewport and user clip planes. + */ +static void +do_clip_tri( struct draw_stage *stage, + struct prim_header *header, + unsigned clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *a[MAX_CLIPPED_VERTICES]; + struct vertex_header *b[MAX_CLIPPED_VERTICES]; + struct vertex_header **inlist = a; + struct vertex_header **outlist = b; + unsigned tmpnr = 0; + unsigned n = 3; + unsigned i; + + inlist[0] = header->v[0]; + inlist[1] = header->v[1]; + inlist[2] = header->v[2]; + + while (clipmask && n >= 3) { + const unsigned plane_idx = ffs(clipmask)-1; + const float *plane = clipper->plane[plane_idx]; + struct vertex_header *vert_prev = inlist[0]; + float dp_prev = dot4( vert_prev->clip, plane ); + unsigned outcount = 0; + + clipmask &= ~(1<clip, plane ); + + if (!IS_NEGATIVE(dp_prev)) { + outlist[outcount++] = vert_prev; + } + + if (DIFFERENT_SIGNS(dp, dp_prev)) { + struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; + outlist[outcount++] = new_vert; + + if (IS_NEGATIVE(dp)) { + /* Going out of bounds. Avoid division by zero as we + * know dp != dp_prev from DIFFERENT_SIGNS, above. + */ + float t = dp / (dp - dp_prev); + interp( clipper, new_vert, t, vert, vert_prev ); + + /* Force edgeflag true in this case: + */ + new_vert->edgeflag = 1; + } else { + /* Coming back in. + */ + float t = dp_prev / (dp_prev - dp); + interp( clipper, new_vert, t, vert_prev, vert ); + + /* Copy starting vert's edgeflag: + */ + new_vert->edgeflag = vert_prev->edgeflag; + } + } + + vert_prev = vert; + dp_prev = dp; + } + + { + struct vertex_header **tmp = inlist; + inlist = outlist; + outlist = tmp; + n = outcount; + } + } + + /* If flat-shading, copy color to new provoking vertex. + */ + if (clipper->flat && inlist[0] != header->v[2]) { + if (1) { + inlist[0] = dup_vert(stage, inlist[0], tmpnr++); + } + + copy_colors(stage, inlist[0], header->v[2]); + } + + + + /* Emit the polygon as triangles to the setup stage: + */ + if (n >= 3) + emit_poly( stage, inlist, n, header ); +} + + +/* Clip a line against the viewport and user clip planes. + */ +static void +do_clip_line( struct draw_stage *stage, + struct prim_header *header, + unsigned clipmask ) +{ + const struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const float *pos0 = v0->clip; + const float *pos1 = v1->clip; + float t0 = 0.0F; + float t1 = 0.0F; + struct prim_header newprim; + + while (clipmask) { + const unsigned plane_idx = ffs(clipmask)-1; + const float *plane = clipper->plane[plane_idx]; + const float dp0 = dot4( pos0, plane ); + const float dp1 = dot4( pos1, plane ); + + if (dp1 < 0.0F) { + float t = dp1 / (dp1 - dp0); + t1 = MAX2(t1, t); + } + + if (dp0 < 0.0F) { + float t = dp0 / (dp0 - dp1); + t0 = MAX2(t0, t); + } + + if (t0 + t1 >= 1.0F) + return; /* discard */ + + clipmask &= ~(1 << plane_idx); /* turn off this plane's bit */ + } + + if (v0->clipmask) { + interp( clipper, stage->tmp[0], t0, v0, v1 ); + + if (clipper->flat) + copy_colors(stage, stage->tmp[0], v0); + + newprim.v[0] = stage->tmp[0]; + } + else { + newprim.v[0] = v0; + } + + if (v1->clipmask) { + interp( clipper, stage->tmp[1], t1, v1, v0 ); + newprim.v[1] = stage->tmp[1]; + } + else { + newprim.v[1] = v1; + } + + stage->next->line( stage->next, &newprim ); +} + + +static void +clip_point( struct draw_stage *stage, + struct prim_header *header ) +{ + if (header->v[0]->clipmask == 0) + stage->next->point( stage->next, header ); +} + + +static void +clip_line( struct draw_stage *stage, + struct prim_header *header ) +{ + unsigned clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask); + + if (clipmask == 0) { + /* no clipping needed */ + stage->next->line( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask) == 0) { + do_clip_line(stage, header, clipmask); + } + /* else, totally clipped */ +} + + +static void +clip_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + unsigned clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask | + header->v[2]->clipmask); + + if (clipmask == 0) { + /* no clipping needed */ + stage->next->tri( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask & + header->v[2]->clipmask) == 0) { + do_clip_tri(stage, header, clipmask); + } +} + +/* Update state. Could further delay this until we hit the first + * primitive that really requires clipping. + */ +static void +clip_init_state( struct draw_stage *stage ) +{ + struct clipper *clipper = clipper_stage( stage ); + + clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; + + if (clipper->flat) { + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + clipper->num_color_attribs = 0; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + clipper->color_attribs[clipper->num_color_attribs++] = i; + } + } + } + + stage->tri = clip_tri; + stage->line = clip_line; +} + + + +static void clip_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + clip_init_state( stage ); + stage->tri( stage, header ); +} + +static void clip_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + clip_init_state( stage ); + stage->line( stage, header ); +} + + +static void clip_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->tri = clip_first_tri; + stage->line = clip_first_line; + stage->next->flush( stage->next, flags ); +} + + +static void clip_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void clip_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Allocate a new clipper stage. + * \return pointer to new stage object + */ +struct draw_stage *draw_clip_stage( struct draw_context *draw ) +{ + struct clipper *clipper = CALLOC_STRUCT(clipper); + + draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); + + clipper->stage.draw = draw; + clipper->stage.point = clip_point; + clipper->stage.line = clip_first_line; + clipper->stage.tri = clip_first_tri; + clipper->stage.flush = clip_flush; + clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; + clipper->stage.destroy = clip_destroy; + + clipper->plane = draw->plane; + + return &clipper->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c new file mode 100644 index 0000000000..4be3830316 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -0,0 +1,293 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "draw_context.h" +#include "draw_private.h" + + + +struct draw_context *draw_create( void ) +{ + struct draw_context *draw = CALLOC_STRUCT( draw_context ); + +#if defined(__i386__) || defined(__386__) + draw->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; +#else + draw->use_sse = FALSE; +#endif + + /* create pipeline stages */ + draw->pipeline.wide = draw_wide_stage( draw ); + draw->pipeline.stipple = draw_stipple_stage( draw ); + draw->pipeline.unfilled = draw_unfilled_stage( draw ); + draw->pipeline.twoside = draw_twoside_stage( draw ); + draw->pipeline.offset = draw_offset_stage( draw ); + draw->pipeline.clip = draw_clip_stage( draw ); + draw->pipeline.flatshade = draw_flatshade_stage( draw ); + draw->pipeline.cull = draw_cull_stage( draw ); + draw->pipeline.validate = draw_validate_stage( draw ); + draw->pipeline.first = draw->pipeline.validate; + + ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); + ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); + ASSIGN_4V( draw->plane[2], 0, -1, 0, 1 ); + ASSIGN_4V( draw->plane[3], 0, 1, 0, 1 ); + ASSIGN_4V( draw->plane[4], 0, 0, 1, 1 ); /* yes these are correct */ + ASSIGN_4V( draw->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ + draw->nr_planes = 6; + + /* Statically allocate maximum sized vertices for the cache - could be cleverer... + */ + { + uint i; + const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; + char *tmp = align_malloc(Elements(draw->vcache.vertex) * size, 16); + + for (i = 0; i < Elements(draw->vcache.vertex); i++) + draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * size); + } + + draw->shader_queue_flush = draw_vertex_shader_queue_flush; + + draw->convert_wide_points = TRUE; + draw->convert_wide_lines = TRUE; + + draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ + + draw_vertex_cache_invalidate( draw ); + draw_set_mapped_element_buffer( draw, 0, NULL ); + + return draw; +} + + +void draw_destroy( struct draw_context *draw ) +{ + draw->pipeline.wide->destroy( draw->pipeline.wide ); + draw->pipeline.stipple->destroy( draw->pipeline.stipple ); + draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); + draw->pipeline.twoside->destroy( draw->pipeline.twoside ); + draw->pipeline.offset->destroy( draw->pipeline.offset ); + draw->pipeline.clip->destroy( draw->pipeline.clip ); + draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); + draw->pipeline.cull->destroy( draw->pipeline.cull ); + draw->pipeline.validate->destroy( draw->pipeline.validate ); + if (draw->pipeline.rasterize) + draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); + tgsi_exec_machine_free_data(&draw->machine); + align_free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ + FREE( draw ); +} + + + +void draw_flush( struct draw_context *draw ) +{ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); +} + + + +/** + * Register new primitive rasterization/rendering state. + * This causes the drawing pipeline to be rebuilt. + */ +void draw_set_rasterizer_state( struct draw_context *draw, + const struct pipe_rasterizer_state *raster ) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + draw->rasterizer = raster; +} + + +/** + * Plug in the primitive rendering/rasterization stage (which is the last + * stage in the drawing pipeline). + * This is provided by the device driver. + */ +void draw_set_rasterize_stage( struct draw_context *draw, + struct draw_stage *stage ) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + draw->pipeline.rasterize = stage; +} + + +/** + * Set the draw module's clipping state. + */ +void draw_set_clip_state( struct draw_context *draw, + const struct pipe_clip_state *clip ) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + assert(clip->nr <= PIPE_MAX_CLIP_PLANES); + memcpy(&draw->plane[6], clip->ucp, clip->nr * sizeof(clip->ucp[0])); + draw->nr_planes = 6 + clip->nr; +} + + +/** + * Set the draw module's viewport state. + */ +void draw_set_viewport_state( struct draw_context *draw, + const struct pipe_viewport_state *viewport ) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->viewport = *viewport; /* struct copy */ +} + + + +void +draw_set_vertex_buffer(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_buffer *buffer) +{ + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); + assert(attr < PIPE_ATTRIB_MAX); + draw->vertex_buffer[attr] = *buffer; +} + + +void +draw_set_vertex_element(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_element *element) +{ + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); + assert(attr < PIPE_ATTRIB_MAX); + draw->vertex_element[attr] = *element; +} + + +/** + * Tell drawing context where to find mapped vertex buffers. + */ +void +draw_set_mapped_vertex_buffer(struct draw_context *draw, + unsigned attr, const void *buffer) +{ + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); + draw->user.vbuffer[attr] = buffer; +} + + +void +draw_set_mapped_constant_buffer(struct draw_context *draw, + const void *buffer) +{ + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); + draw->user.constants = buffer; +} + + +/** + * Tells the draw module whether to convert wide points (size != 1) + * into triangles. + */ +void +draw_convert_wide_points(struct draw_context *draw, boolean enable) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->convert_wide_points = enable; +} + + +/** + * Tells the draw module whether to convert wide lines (width != 1) + * into triangles. + */ +void +draw_convert_wide_lines(struct draw_context *draw, boolean enable) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->convert_wide_lines = enable; +} + + +/** + * Allocate space for temporary post-transform vertices, such as for clipping. + */ +void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) +{ + assert(!stage->tmp); + + stage->nr_tmps = nr; + + if (nr) { + ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); + unsigned i; + + stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } +} + + +void draw_free_temp_verts( struct draw_stage *stage ) +{ + if (stage->tmp) { + FREE( stage->tmp[0] ); + FREE( stage->tmp ); + stage->tmp = NULL; + } +} + + +boolean draw_use_sse(struct draw_context *draw) +{ + return (boolean) draw->use_sse; +} + + +void draw_reset_vertex_ids(struct draw_context *draw) +{ + struct draw_stage *stage = draw->pipeline.first; + + while (stage) { + unsigned i; + + for (i = 0; i < stage->nr_tmps; i++) + stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; + + stage = stage->next; + } + + draw_vertex_cache_reset_vertex_ids(draw); +} diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h new file mode 100644 index 0000000000..ddeb184497 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -0,0 +1,142 @@ + +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Public interface into the drawing module. + */ + +/* Authors: Keith Whitwell + */ + + +#ifndef DRAW_CONTEXT_H +#define DRAW_CONTEXT_H + + +#include "pipe/p_state.h" + + +struct vertex_buffer; +struct vertex_info; +struct draw_context; +struct draw_stage; +struct draw_vertex_shader; + + +/** + * Clipmask flags + */ +/*@{*/ +#define CLIP_RIGHT_BIT 0x01 +#define CLIP_LEFT_BIT 0x02 +#define CLIP_TOP_BIT 0x04 +#define CLIP_BOTTOM_BIT 0x08 +#define CLIP_NEAR_BIT 0x10 +#define CLIP_FAR_BIT 0x20 +/*@}*/ + +/** + * Bitshift for each clip flag + */ +/*@{*/ +#define CLIP_RIGHT_SHIFT 0 +#define CLIP_LEFT_SHIFT 1 +#define CLIP_TOP_SHIFT 2 +#define CLIP_BOTTOM_SHIFT 3 +#define CLIP_NEAR_SHIFT 4 +#define CLIP_FAR_SHIFT 5 +/*@}*/ + + +struct draw_context *draw_create( void ); + +void draw_destroy( struct draw_context *draw ); + +void draw_set_viewport_state( struct draw_context *draw, + const struct pipe_viewport_state *viewport ); + +void draw_set_clip_state( struct draw_context *pipe, + const struct pipe_clip_state *clip ); + +void draw_set_rasterizer_state( struct draw_context *draw, + const struct pipe_rasterizer_state *raster ); + +void draw_set_rasterize_stage( struct draw_context *draw, + struct draw_stage *stage ); + +void draw_convert_wide_points(struct draw_context *draw, boolean enable); + +void draw_convert_wide_lines(struct draw_context *draw, boolean enable); + + +struct draw_vertex_shader * +draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader); +void draw_bind_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs); +void draw_delete_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs); + +boolean draw_use_sse(struct draw_context *draw); + +void draw_set_vertex_buffer(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_buffer *buffer); + +void draw_set_vertex_element(struct draw_context *draw, + unsigned attr, + const struct pipe_vertex_element *element); + +void draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ); + +void draw_set_mapped_vertex_buffer(struct draw_context *draw, + unsigned attr, const void *buffer); + +void draw_set_mapped_constant_buffer(struct draw_context *draw, + const void *buffer); + + +/*********************************************************************** + * draw_prim.c + */ + +void draw_arrays(struct draw_context *draw, unsigned prim, + unsigned start, unsigned count); + +void draw_flush(struct draw_context *draw); + +/*********************************************************************** + * draw_debug.c + */ +boolean draw_validate_prim( unsigned prim, unsigned length ); +unsigned draw_trim_prim( unsigned mode, unsigned count ); + + + +#endif /* DRAW_CONTEXT_H */ diff --git a/src/gallium/auxiliary/draw/draw_cull.c b/src/gallium/auxiliary/draw/draw_cull.c new file mode 100644 index 0000000000..8177b0ac86 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_cull.c @@ -0,0 +1,150 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Drawing stage for polygon culling + */ + +/* Authors: Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + +struct cull_stage { + struct draw_stage stage; + unsigned winding; /**< which winding(s) to cull (one of PIPE_WINDING_x) */ +}; + + +static INLINE struct cull_stage *cull_stage( struct draw_stage *stage ) +{ + return (struct cull_stage *)stage; +} + + + + +static void cull_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + /* Window coords: */ + const float *v0 = header->v[0]->data[0]; + const float *v1 = header->v[1]->data[0]; + const float *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0] - v2[0]; + const float ey = v0[1] - v2[1]; + const float fx = v1[0] - v2[0]; + const float fy = v1[1] - v2[1]; + + /* det = cross(e,f).z */ + header->det = ex * fy - ey * fx; + + if (header->det != 0) { + /* if (det < 0 then Z points toward camera and triangle is + * counter-clockwise winding. + */ + unsigned winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; + + if ((winding & cull_stage(stage)->winding) == 0) { + /* triangle is not culled, pass to next stage */ + stage->next->tri( stage->next, header ); + } + } +} + +static void cull_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct cull_stage *cull = cull_stage(stage); + + cull->winding = stage->draw->rasterizer->cull_mode; + + stage->tri = cull_tri; + stage->tri( stage, header ); +} + + + +static void cull_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void cull_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void cull_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->tri = cull_first_tri; + stage->next->flush( stage->next, flags ); +} + +static void cull_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void cull_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create a new polygon culling stage. + */ +struct draw_stage *draw_cull_stage( struct draw_context *draw ) +{ + struct cull_stage *cull = CALLOC_STRUCT(cull_stage); + + draw_alloc_temp_verts( &cull->stage, 0 ); + + cull->stage.draw = draw; + cull->stage.next = NULL; + cull->stage.point = cull_point; + cull->stage.line = cull_line; + cull->stage.tri = cull_first_tri; + cull->stage.flush = cull_flush; + cull->stage.reset_stipple_counter = cull_reset_stipple_counter; + cull->stage.destroy = cull_destroy; + + return &cull->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_debug.c b/src/gallium/auxiliary/draw/draw_debug.c new file mode 100644 index 0000000000..d6220b5f62 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_debug.c @@ -0,0 +1,113 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "draw_private.h" +#include "draw_context.h" + + + +static void +draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) +{ + assert(prim >= PIPE_PRIM_POINTS); + assert(prim <= PIPE_PRIM_POLYGON); + + switch (prim) { + case PIPE_PRIM_POINTS: + *first = 1; + *incr = 1; + break; + case PIPE_PRIM_LINES: + *first = 2; + *incr = 2; + break; + case PIPE_PRIM_LINE_STRIP: + *first = 2; + *incr = 1; + break; + case PIPE_PRIM_LINE_LOOP: + *first = 2; + *incr = 1; + break; + case PIPE_PRIM_TRIANGLES: + *first = 3; + *incr = 3; + break; + case PIPE_PRIM_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + break; + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + *first = 3; + *incr = 1; + break; + case PIPE_PRIM_QUADS: + *first = 4; + *incr = 4; + break; + case PIPE_PRIM_QUAD_STRIP: + *first = 4; + *incr = 2; + break; + default: + assert(0); + *first = 1; + *incr = 1; + break; + } +} + + +unsigned +draw_trim_prim( unsigned mode, unsigned count ) +{ + unsigned length, first, incr; + + draw_prim_info( mode, &first, &incr ); + + if (count < first) + length = 0; + else + length = count - (count - first) % incr; + + return length; +} + + +boolean +draw_validate_prim( unsigned mode, unsigned count ) +{ + return (count > 0 && + count == draw_trim_prim( mode, count )); +} + diff --git a/src/gallium/auxiliary/draw/draw_flatshade.c b/src/gallium/auxiliary/draw/draw_flatshade.c new file mode 100644 index 0000000000..4398abbc60 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_flatshade.c @@ -0,0 +1,205 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +/** subclass of draw_stage */ +struct flat_stage +{ + struct draw_stage stage; + + uint num_color_attribs; + uint color_attribs[4]; /* front/back primary/secondary colors */ +}; + + +static INLINE struct flat_stage * +flat_stage(struct draw_stage *stage) +{ + return (struct flat_stage *) stage; +} + + +/** Copy all the color attributes from 'src' vertex to 'dst' vertex */ +static INLINE void copy_colors( struct draw_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + const struct flat_stage *flat = flat_stage(stage); + uint i; + for (i = 0; i < flat->num_color_attribs; i++) { + const uint attr = flat->color_attribs[i]; + COPY_4FV(dst->data[attr], src->data[attr]); + } +} + + +/** Copy all the color attributes from src vertex to dst0 & dst1 vertices */ +static INLINE void copy_colors2( struct draw_stage *stage, + struct vertex_header *dst0, + struct vertex_header *dst1, + const struct vertex_header *src ) +{ + const struct flat_stage *flat = flat_stage(stage); + uint i; + for (i = 0; i < flat->num_color_attribs; i++) { + const uint attr = flat->color_attribs[i]; + COPY_4FV(dst0->data[attr], src->data[attr]); + COPY_4FV(dst1->data[attr], src->data[attr]); + } +} + + +/** + * Flatshade tri. Required for clipping and when unfilled tris are + * active, otherwise handled by hardware. + */ +static void flatshade_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = header->v[2]; + + copy_colors2(stage, tmp.v[0], tmp.v[1], tmp.v[2]); + + stage->next->tri( stage->next, &tmp ); +} + + +/** + * Flatshade line. Required for clipping. + */ +static void flatshade_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = header->v[1]; + + copy_colors(stage, tmp.v[0], tmp.v[1]); + + stage->next->line( stage->next, &tmp ); +} + + +static void flatshade_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void flatshade_init_state( struct draw_stage *stage ) +{ + struct flat_stage *flat = flat_stage(stage); + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + /* Find which vertex shader outputs are colors, make a list */ + flat->num_color_attribs = 0; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + flat->color_attribs[flat->num_color_attribs++] = i; + } + } + + stage->line = flatshade_line; + stage->tri = flatshade_tri; +} + +static void flatshade_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + flatshade_init_state( stage ); + stage->tri( stage, header ); +} + +static void flatshade_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + flatshade_init_state( stage ); + stage->line( stage, header ); +} + + +static void flatshade_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->tri = flatshade_first_tri; + stage->line = flatshade_first_line; + stage->next->flush( stage->next, flags ); +} + + +static void flatshade_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void flatshade_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create flatshading drawing stage. + */ +struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) +{ + struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage); + + draw_alloc_temp_verts( &flatshade->stage, 2 ); + + flatshade->stage.draw = draw; + flatshade->stage.next = NULL; + flatshade->stage.point = flatshade_point; + flatshade->stage.line = flatshade_first_line; + flatshade->stage.tri = flatshade_first_tri; + flatshade->stage.flush = flatshade_flush; + flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; + flatshade->stage.destroy = flatshade_destroy; + + return &flatshade->stage; +} + + diff --git a/src/gallium/auxiliary/draw/draw_offset.c b/src/gallium/auxiliary/draw/draw_offset.c new file mode 100644 index 0000000000..dbc676deae --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_offset.c @@ -0,0 +1,186 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief polygon offset state + * + * \author Keith Whitwell + * \author Brian Paul + */ + +#include "pipe/p_util.h" +#include "draw_private.h" + + + +struct offset_stage { + struct draw_stage stage; + + float scale; + float units; +}; + + + +static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) +{ + return (struct offset_stage *) stage; +} + + + + + +/** + * Offset tri Z. Some hardware can handle this, but not usually when + * doing unfilled rendering. + */ +static void do_offset_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + float inv_det = 1.0f / header->det; + + /* Window coords: + */ + float *v0 = header->v[0]->data[0]; + float *v1 = header->v[1]->data[0]; + float *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ + float ex = v0[0] - v2[0]; + float ey = v0[1] - v2[1]; + float ez = v0[2] - v2[2]; + float fx = v1[0] - v2[0]; + float fy = v1[1] - v2[1]; + float fz = v1[2] - v2[2]; + + /* (a,b) = cross(e,f).xy */ + float a = ey*fz - ez*fy; + float b = ez*fx - ex*fz; + + float dzdx = FABSF(a * inv_det); + float dzdy = FABSF(b * inv_det); + + float zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; + + /* + * Note: we're applying the offset and clamping per-vertex. + * Ideally, the offset is applied per-fragment prior to fragment shading. + */ + v0[2] = CLAMP(v0[2] + zoffset, 0.0f, 1.0f); + v1[2] = CLAMP(v1[2] + zoffset, 0.0f, 1.0f); + v2[2] = CLAMP(v2[2] + zoffset, 0.0f, 1.0f); + + stage->next->tri( stage->next, header ); +} + + +static void offset_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = dup_vert(stage, header->v[2], 2); + + do_offset_tri( stage, &tmp ); +} + + +static void offset_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ + + offset->units = stage->draw->rasterizer->offset_units * mrd; + offset->scale = stage->draw->rasterizer->offset_scale; + + stage->tri = offset_tri; + stage->tri( stage, header ); +} + + +static void offset_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void offset_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void offset_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->tri = offset_first_tri; + stage->next->flush( stage->next, flags ); +} + + +static void offset_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void offset_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create polygon offset drawing stage. + */ +struct draw_stage *draw_offset_stage( struct draw_context *draw ) +{ + struct offset_stage *offset = CALLOC_STRUCT(offset_stage); + + draw_alloc_temp_verts( &offset->stage, 3 ); + + offset->stage.draw = draw; + offset->stage.next = NULL; + offset->stage.point = offset_point; + offset->stage.line = offset_line; + offset->stage.tri = offset_first_tri; + offset->stage.flush = offset_flush; + offset->stage.reset_stipple_counter = offset_reset_stipple_counter; + offset->stage.destroy = offset_destroy; + + return &offset->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c new file mode 100644 index 0000000000..51e2242719 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -0,0 +1,482 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_debug.h" + +#include "draw_private.h" +#include "draw_context.h" + + + +#define RP_NONE 0 +#define RP_POINT 1 +#define RP_LINE 2 +#define RP_TRI 3 + + +static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { + RP_POINT, + RP_LINE, + RP_LINE, + RP_LINE, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI, + RP_TRI +}; + + +static void draw_prim_queue_flush( struct draw_context *draw ) +{ + unsigned i; + + if (0) + debug_printf("Flushing with %d prims, %d verts\n", + draw->pq.queue_nr, draw->vs.queue_nr); + + assert (draw->pq.queue_nr != 0); + + /* NOTE: we cannot save draw->pipeline->first in a local var because + * draw->pipeline->first is often changed by the first call to tri(), + * line(), etc. + */ + if (draw->rasterizer->line_stipple_enable) { + switch (draw->reduced_prim) { + case RP_TRI: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); + + draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); + } + break; + case RP_LINE: + for (i = 0; i < draw->pq.queue_nr; i++) { + if (draw->pq.queue[i].reset_line_stipple) + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); + + draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); + } + break; + case RP_POINT: + draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); + break; + } + } + else { + switch (draw->reduced_prim) { + case RP_TRI: + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); + break; + case RP_LINE: + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); + break; + case RP_POINT: + for (i = 0; i < draw->pq.queue_nr; i++) + draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); + break; + } + } + + draw->pq.queue_nr = 0; + draw_vertex_cache_unreference( draw ); +} + + + +void draw_do_flush( struct draw_context *draw, unsigned flags ) +{ + if (0) + debug_printf("Flushing with %d verts, %d prims\n", + draw->vs.queue_nr, + draw->pq.queue_nr ); + + + if (flags >= DRAW_FLUSH_SHADER_QUEUE) { + if (draw->vs.queue_nr) + (*draw->shader_queue_flush)(draw); + + if (flags >= DRAW_FLUSH_PRIM_QUEUE) { + if (draw->pq.queue_nr) + draw_prim_queue_flush(draw); + + if (flags >= DRAW_FLUSH_VERTEX_CACHE) { + draw_vertex_cache_invalidate(draw); + + if (flags >= DRAW_FLUSH_STATE_CHANGE) { + draw->pipeline.first->flush( draw->pipeline.first, flags ); + draw->pipeline.first = draw->pipeline.validate; + draw->reduced_prim = ~0; + } + } + } + } +} + + + +/* Return a pointer to a freshly queued primitive header. Ensure that + * there is room in the vertex cache for a maximum of "nr_verts" new + * vertices. Flush primitive and/or vertex queues if necessary to + * make space. + */ +static struct prim_header *get_queued_prim( struct draw_context *draw, + unsigned nr_verts ) +{ + if (!draw_vertex_cache_check_space( draw, nr_verts )) { +// debug_printf("v"); + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE ); + } + else if (draw->pq.queue_nr == PRIM_QUEUE_LENGTH) { +// debug_printf("p"); + draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); + } + + assert(draw->pq.queue_nr < PRIM_QUEUE_LENGTH); + + return &draw->pq.queue[draw->pq.queue_nr++]; +} + + + +/** + * Add a point to the primitive queue. + * \param i0 index into user's vertex arrays + */ +static void do_point( struct draw_context *draw, + unsigned i0 ) +{ + struct prim_header *prim = get_queued_prim( draw, 1 ); + + prim->reset_line_stipple = 0; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->vcache.get_vertex( draw, i0 ); +} + + +/** + * Add a line to the primitive queue. + * \param i0 index into user's vertex arrays + * \param i1 index into user's vertex arrays + */ +static void do_line( struct draw_context *draw, + boolean reset_stipple, + unsigned i0, + unsigned i1 ) +{ + struct prim_header *prim = get_queued_prim( draw, 2 ); + + prim->reset_line_stipple = reset_stipple; + prim->edgeflags = 1; + prim->pad = 0; + prim->v[0] = draw->vcache.get_vertex( draw, i0 ); + prim->v[1] = draw->vcache.get_vertex( draw, i1 ); +} + +/** + * Add a triangle to the primitive queue. + */ +static void do_triangle( struct draw_context *draw, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + + prim->reset_line_stipple = 1; + prim->edgeflags = ~0; + prim->pad = 0; + prim->v[0] = draw->vcache.get_vertex( draw, i0 ); + prim->v[1] = draw->vcache.get_vertex( draw, i1 ); + prim->v[2] = draw->vcache.get_vertex( draw, i2 ); +} + +static void do_ef_triangle( struct draw_context *draw, + boolean reset_stipple, + unsigned ef_mask, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + struct prim_header *prim = get_queued_prim( draw, 3 ); + struct vertex_header *v0 = draw->vcache.get_vertex( draw, i0 ); + struct vertex_header *v1 = draw->vcache.get_vertex( draw, i1 ); + struct vertex_header *v2 = draw->vcache.get_vertex( draw, i2 ); + + prim->reset_line_stipple = reset_stipple; + + prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | + (v1->edgeflag << 1) | + (v2->edgeflag << 2)); + prim->pad = 0; + prim->v[0] = v0; + prim->v[1] = v1; + prim->v[2] = v2; +} + + +static void do_ef_quad( struct draw_context *draw, + unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3 ) +{ + const unsigned omitEdge2 = ~(1 << 1); + const unsigned omitEdge3 = ~(1 << 2); + do_ef_triangle( draw, 1, omitEdge2, v0, v1, v3 ); + do_ef_triangle( draw, 0, omitEdge3, v1, v2, v3 ); +} + +static void do_quad( struct draw_context *draw, + unsigned v0, + unsigned v1, + unsigned v2, + unsigned v3 ) +{ + do_triangle( draw, v0, v1, v3 ); + do_triangle( draw, v1, v2, v3 ); +} + + +/** + * Main entrypoint to draw some number of points/lines/triangles + */ +static void +draw_prim( struct draw_context *draw, + unsigned prim, unsigned start, unsigned count ) +{ + unsigned i; + boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); + +// debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); + + switch (prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i ++) { + do_point( draw, + start + i ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) { + do_line( draw, + TRUE, + start + i + 0, + start + i + 1); + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, /* XXX: only if vb not split */ + start + i - 1, + start + i ); + } + + do_line( draw, + 0, + start + count - 1, + start + 0 ); + } + break; + + case PIPE_PRIM_LINE_STRIP: + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, + start + i - 1, + start + i ); + } + break; + + case PIPE_PRIM_TRIANGLES: + if (unfilled) { + for (i = 0; i+2 < count; i += 3) { + do_ef_triangle( draw, + 1, + ~0, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + else { + for (i = 0; i+2 < count; i += 3) { + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + do_triangle( draw, + start + i + 1, + start + i + 0, + start + i + 2 ); + } + else { + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + if (count >= 3) { + for (i = 0; i+2 < count; i++) { + do_triangle( draw, + start + 0, + start + i + 1, + start + i + 2 ); + } + } + break; + + + case PIPE_PRIM_QUADS: + if (unfilled) { + for (i = 0; i+3 < count; i += 4) { + do_ef_quad( draw, + start + i + 0, + start + i + 1, + start + i + 2, + start + i + 3); + } + } + else { + for (i = 0; i+3 < count; i += 4) { + do_quad( draw, + start + i + 0, + start + i + 1, + start + i + 2, + start + i + 3); + } + } + break; + + case PIPE_PRIM_QUAD_STRIP: + if (unfilled) { + for (i = 0; i+3 < count; i += 2) { + do_ef_quad( draw, + start + i + 2, + start + i + 0, + start + i + 1, + start + i + 3); + } + } + else { + for (i = 0; i+3 < count; i += 2) { + do_quad( draw, + start + i + 2, + start + i + 0, + start + i + 1, + start + i + 3); + } + } + break; + + case PIPE_PRIM_POLYGON: + if (unfilled) { + unsigned ef_mask = (1<<2) | (1<<0); + + for (i = 0; i+2 < count; i++) { + + if (i + 3 >= count) + ef_mask |= (1<<1); + + do_ef_triangle( draw, + i == 0, + ef_mask, + start + i + 1, + start + i + 2, + start + 0); + + ef_mask &= ~(1<<2); + } + } + else { + for (i = 0; i+2 < count; i++) { + do_triangle( draw, + start + i + 1, + start + i + 2, + start + 0); + } + } + break; + + default: + assert(0); + break; + } +} + + + + +/** + * Draw vertex arrays + * This is the main entrypoint into the drawing module. + * \param prim one of PIPE_PRIM_x + * \param start index of first vertex to draw + * \param count number of vertices to draw + */ +void +draw_arrays(struct draw_context *draw, unsigned prim, + unsigned start, unsigned count) +{ + if (reduced_prim[prim] != draw->reduced_prim) { + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->reduced_prim = reduced_prim[prim]; + } + + /* drawing done here: */ + draw_prim(draw, prim, start, count); +} + + diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h new file mode 100644 index 0000000000..3d09aef87c --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -0,0 +1,346 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Private data structures, etc for the draw module. + */ + + +/** + * Authors: + * Keith Whitwell + * Brian Paul + */ + + +#ifndef DRAW_PRIVATE_H +#define DRAW_PRIVATE_H + + +#include "pipe/p_state.h" +#include "pipe/p_defines.h" + +#include "x86/rtasm/x86sse.h" +#include "tgsi/exec/tgsi_exec.h" + + +struct gallivm_prog; +struct gallivm_cpu_engine; + +/** + * Basic vertex info. + * Carry some useful information around with the vertices in the prim pipe. + */ +struct vertex_header { + unsigned clipmask:12; + unsigned edgeflag:1; + unsigned pad:3; + unsigned vertex_id:16; + + float clip[4]; + + float data[][4]; /* Note variable size */ +}; + +/* NOTE: It should match vertex_id size above */ +#define UNDEFINED_VERTEX_ID 0xffff + +/* XXX This is too large */ +#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) + + + +/** + * Basic info for a point/line/triangle primitive. + */ +struct prim_header { + float det; /**< front/back face determinant */ + unsigned reset_line_stipple:1; + unsigned edgeflags:3; + unsigned pad:28; + struct vertex_header *v[3]; /**< 1 to 3 vertex pointers */ +}; + + + +struct draw_context; + +/** + * Base class for all primitive drawing stages. + */ +struct draw_stage +{ + struct draw_context *draw; /**< parent context */ + + struct draw_stage *next; /**< next stage in pipeline */ + + struct vertex_header **tmp; /**< temp vert storage, such as for clipping */ + unsigned nr_tmps; + + void (*point)( struct draw_stage *, + struct prim_header * ); + + void (*line)( struct draw_stage *, + struct prim_header * ); + + void (*tri)( struct draw_stage *, + struct prim_header * ); + + void (*flush)( struct draw_stage *, + unsigned flags ); + + void (*reset_stipple_counter)( struct draw_stage * ); + + void (*destroy)( struct draw_stage * ); +}; + + +#define PRIM_QUEUE_LENGTH 16 +#define VCACHE_SIZE 32 +#define VCACHE_OVERFLOW 4 +#define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ + +/** + * Private version of the compiled vertex_shader + */ +struct draw_vertex_shader { + const struct pipe_shader_state *state; +#if defined(__i386__) || defined(__386__) + struct x86_function sse2_program; +#endif +#ifdef MESA_LLVM + struct gallivm_prog *llvm_prog; +#endif +}; + + +/* Internal function for vertex fetch. + */ +typedef void (*fetch_func)(const void *ptr, float *attrib); +typedef void (*full_fetch_func)( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ); + + + +/** + * Private context for the drawing module. + */ +struct draw_context +{ + /** Drawing/primitive pipeline stages */ + struct { + struct draw_stage *first; /**< one of the following */ + + struct draw_stage *validate; + + /* stages (in logical order) */ + struct draw_stage *flatshade; + struct draw_stage *clip; + struct draw_stage *cull; + struct draw_stage *twoside; + struct draw_stage *offset; + struct draw_stage *unfilled; + struct draw_stage *stipple; + struct draw_stage *wide; + struct draw_stage *rasterize; + } pipeline; + + /* pipe state that we need: */ + const struct pipe_rasterizer_state *rasterizer; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + const struct draw_vertex_shader *vertex_shader; + + uint num_vs_outputs; /**< convenience, from vertex_shader */ + + /* user-space vertex data, buffers */ + struct { + /** vertex element/index buffer (ex: glDrawElements) */ + const void *elts; + /** bytes per index (0, 1, 2 or 4) */ + unsigned eltSize; + + /** vertex arrays */ + const void *vbuffer[PIPE_ATTRIB_MAX]; + + /** constant buffer (for vertex shader) */ + const void *constants; + } user; + + /* Clip derived state: + */ + float plane[12][4]; + unsigned nr_planes; + + boolean convert_wide_points; /**< convert wide points to tris? */ + boolean convert_wide_lines; /**< convert side lines to tris? */ + + unsigned reduced_prim; + + /** TGSI program interpreter runtime state */ + struct tgsi_exec_machine machine; + + /* Vertex fetch internal state + */ + struct { + const ubyte *src_ptr[PIPE_ATTRIB_MAX]; + unsigned pitch[PIPE_ATTRIB_MAX]; + fetch_func fetch[PIPE_ATTRIB_MAX]; + unsigned nr_attrs; + full_fetch_func fetch_func; + } vertex_fetch; + + /* Post-tnl vertex cache: + */ + struct { + unsigned referenced; /**< bitfield */ + unsigned idx[VCACHE_SIZE + VCACHE_OVERFLOW]; + struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; + unsigned overflow; + + /** To find space in the vertex cache: */ + struct vertex_header *(*get_vertex)( struct draw_context *draw, + unsigned i ); + } vcache; + + /* Vertex shader queue: + */ + struct { + struct { + unsigned elt; /**< index into the user's vertex arrays */ + struct vertex_header *dest; /**< points into vcache.vertex[] array */ + } queue[VS_QUEUE_LENGTH]; + unsigned queue_nr; + } vs; + + /** + * Run the vertex shader on all vertices in the vertex queue. + */ + void (*shader_queue_flush)(struct draw_context *draw); + + /* Prim pipeline queue: + */ + struct { + /* Need to queue up primitives until their vertices have been + * transformed by a vs queue flush. + */ + struct prim_header queue[PRIM_QUEUE_LENGTH]; + unsigned queue_nr; + } pq; + + int use_sse : 1; +#ifdef MESA_LLVM + struct gallivm_cpu_engine *engine; +#endif + + void *driver_private; +}; + + + +extern struct draw_stage *draw_unfilled_stage( struct draw_context *context ); +extern struct draw_stage *draw_twoside_stage( struct draw_context *context ); +extern struct draw_stage *draw_offset_stage( struct draw_context *context ); +extern struct draw_stage *draw_clip_stage( struct draw_context *context ); +extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); +extern struct draw_stage *draw_cull_stage( struct draw_context *context ); +extern struct draw_stage *draw_stipple_stage( struct draw_context *context ); +extern struct draw_stage *draw_wide_stage( struct draw_context *context ); +extern struct draw_stage *draw_validate_stage( struct draw_context *context ); + + +extern void draw_free_temp_verts( struct draw_stage *stage ); + +extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); + +extern void draw_reset_vertex_ids( struct draw_context *draw ); + + +extern int draw_vertex_cache_check_space( struct draw_context *draw, + unsigned nr_verts ); + +extern void draw_vertex_cache_invalidate( struct draw_context *draw ); +extern void draw_vertex_cache_unreference( struct draw_context *draw ); +extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); + + +extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); +#ifdef MESA_LLVM +extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw ); +#endif + +struct tgsi_exec_machine; + +extern void draw_update_vertex_fetch( struct draw_context *draw ); + + +#define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ +#define DRAW_FLUSH_PRIM_QUEUE 0x2 +#define DRAW_FLUSH_VERTEX_CACHE 0x4 +#define DRAW_FLUSH_STATE_CHANGE 0x8 +#define DRAW_FLUSH_BACKEND 0x10 + + +void draw_do_flush( struct draw_context *draw, unsigned flags ); + + + +/** + * Get a writeable copy of a vertex. + * \param stage drawing stage info + * \param vert the vertex to copy (source) + * \param idx index into stage's tmp[] array to put the copy (dest) + * \return pointer to the copied vertex + */ +static INLINE struct vertex_header * +dup_vert( struct draw_stage *stage, + const struct vertex_header *vert, + unsigned idx ) +{ + struct vertex_header *tmp = stage->tmp[idx]; + const uint vsize = sizeof(struct vertex_header) + + stage->draw->num_vs_outputs * 4 * sizeof(float); + memcpy(tmp, vert, vsize); + tmp->vertex_id = UNDEFINED_VERTEX_ID; + return tmp; +} + +static INLINE float +dot4(const float *a, const float *b) +{ + float result = (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); + + return result; +} + +#endif /* DRAW_PRIVATE_H */ diff --git a/src/gallium/auxiliary/draw/draw_stipple.c b/src/gallium/auxiliary/draw/draw_stipple.c new file mode 100644 index 0000000000..506f33512c --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_stipple.c @@ -0,0 +1,239 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +/* Implement line stipple by cutting lines up into smaller lines. + * There are hundreds of ways to implement line stipple, this is one + * choice that should work in all situations, requires no state + * manipulations, but with a penalty in terms of large amounts of + * generated geometry. + */ + + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +/** Subclass of draw_stage */ +struct stipple_stage { + struct draw_stage stage; + float counter; + uint pattern; + uint factor; +}; + + +static INLINE struct stipple_stage * +stipple_stage(struct draw_stage *stage) +{ + return (struct stipple_stage *) stage; +} + + +/** + * Compute interpolated vertex attributes for 'dst' at position 't' + * between 'v0' and 'v1'. + * XXX using linear interpolation for all attribs at this time. + */ +static void +screen_interp( struct draw_context *draw, + struct vertex_header *dst, + float t, + const struct vertex_header *v0, + const struct vertex_header *v1 ) +{ + uint attr; + for (attr = 0; attr < draw->num_vs_outputs; attr++) { + const float *val0 = v0->data[attr]; + const float *val1 = v1->data[attr]; + float *newv = dst->data[attr]; + uint i; + for (i = 0; i < 4; i++) { + newv[i] = val0[i] + t * (val1[i] - val0[i]); + } + } +} + + +static void +emit_segment(struct draw_stage *stage, struct prim_header *header, + float t0, float t1) +{ + struct vertex_header *v0new = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1new = dup_vert(stage, header->v[1], 1); + struct prim_header newprim = *header; + + if (t0 > 0.0) { + screen_interp( stage->draw, v0new, t0, header->v[0], header->v[1] ); + newprim.v[0] = v0new; + } + + if (t1 < 1.0) { + screen_interp( stage->draw, v1new, t1, header->v[0], header->v[1] ); + newprim.v[1] = v1new; + } + + stage->next->line( stage->next, &newprim ); +} + + +static INLINE unsigned +stipple_test(int counter, ushort pattern, int factor) +{ + int b = (counter / factor) & 0xf; + return (1 << b) & pattern; +} + + +static void +stipple_line(struct draw_stage *stage, struct prim_header *header) +{ + struct stipple_stage *stipple = stipple_stage(stage); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const float *pos0 = v0->data[0]; + const float *pos1 = v1->data[0]; + float start = 0; + int state = 0; + + float x0 = pos0[0]; + float x1 = pos1[0]; + float y0 = pos0[1]; + float y1 = pos1[1]; + + float dx = x0 > x1 ? x0 - x1 : x1 - x0; + float dy = y0 > y1 ? y0 - y1 : y1 - y0; + + float length = MAX2(dx, dy); + int i; + + /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table. + */ + for (i = 0; i < length; i++) { + int result = stipple_test( (int) stipple->counter+i, + (ushort) stipple->pattern, stipple->factor ); + if (result != state) { + /* changing from "off" to "on" or vice versa */ + if (state) { + if (start != i) { + /* finishing an "on" segment */ + emit_segment( stage, header, start / length, i / length ); + } + } + else { + /* starting an "on" segment */ + start = (float) i; + } + state = result; + } + } + + if (state && start < length) + emit_segment( stage, header, start / length, 1.0 ); + + stipple->counter += length; +} + + +static void +reset_stipple_counter(struct draw_stage *stage) +{ + struct stipple_stage *stipple = stipple_stage(stage); + stipple->counter = 0; + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +stipple_first_line(struct draw_stage *stage, + struct prim_header *header) +{ + struct stipple_stage *stipple = stipple_stage(stage); + struct draw_context *draw = stage->draw; + + stipple->pattern = draw->rasterizer->line_stipple_pattern; + stipple->factor = draw->rasterizer->line_stipple_factor + 1; + + stage->line = stipple_line; + stage->line( stage, header ); +} + + +static void +stipple_flush(struct draw_stage *stage, unsigned flags) +{ + stage->line = stipple_first_line; + stage->next->flush( stage->next, flags ); +} + + +static void +passthrough_point(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->point( stage->next, header ); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + +static void +stipple_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create line stippler stage + */ +struct draw_stage *draw_stipple_stage( struct draw_context *draw ) +{ + struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); + + draw_alloc_temp_verts( &stipple->stage, 2 ); + + stipple->stage.draw = draw; + stipple->stage.next = NULL; + stipple->stage.point = passthrough_point; + stipple->stage.line = stipple_first_line; + stipple->stage.tri = passthrough_tri; + stipple->stage.reset_stipple_counter = reset_stipple_counter; + stipple->stage.flush = stipple_flush; + stipple->stage.destroy = stipple_destroy; + + return &stipple->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_twoside.c b/src/gallium/auxiliary/draw/draw_twoside.c new file mode 100644 index 0000000000..1c38957987 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_twoside.c @@ -0,0 +1,203 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +struct twoside_stage { + struct draw_stage stage; + float sign; /**< +1 or -1 */ + uint attrib_front0, attrib_back0; + uint attrib_front1, attrib_back1; +}; + + +static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage ) +{ + return (struct twoside_stage *)stage; +} + + + + +/** + * Copy back color(s) to front color(s). + */ +static INLINE struct vertex_header * +copy_bfc( struct twoside_stage *twoside, + const struct vertex_header *v, + unsigned idx ) +{ + struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); + + if (twoside->attrib_back0) { + COPY_4FV(tmp->data[twoside->attrib_front0], + tmp->data[twoside->attrib_back0]); + } + if (twoside->attrib_back1) { + COPY_4FV(tmp->data[twoside->attrib_front1], + tmp->data[twoside->attrib_back1]); + } + + return tmp; +} + + +/* Twoside tri: + */ +static void twoside_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + if (header->det * twoside->sign < 0.0) { + /* this is a back-facing triangle */ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + /* copy back attribs to front attribs */ + tmp.v[0] = copy_bfc(twoside, header->v[0], 0); + tmp.v[1] = copy_bfc(twoside, header->v[1], 1); + tmp.v[2] = copy_bfc(twoside, header->v[2], 2); + + stage->next->tri( stage->next, &tmp ); + } + else { + stage->next->tri( stage->next, header ); + } +} + + +static void twoside_line( struct draw_stage *stage, + struct prim_header *header ) +{ + /* pass-through */ + stage->next->line( stage->next, header ); +} + + +static void twoside_point( struct draw_stage *stage, + struct prim_header *header ) +{ + /* pass-through */ + stage->next->point( stage->next, header ); +} + + +static void twoside_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + uint i; + + twoside->attrib_front0 = 0; + twoside->attrib_front1 = 0; + twoside->attrib_back0 = 0; + twoside->attrib_back1 = 0; + + /* Find which vertex shader outputs are front/back colors */ + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { + if (vs->output_semantic_index[i] == 0) + twoside->attrib_front0 = i; + else + twoside->attrib_front1 = i; + } + if (vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + if (vs->output_semantic_index[i] == 0) + twoside->attrib_back0 = i; + else + twoside->attrib_back1 = i; + } + } + + if (!twoside->attrib_back0) + twoside->attrib_front0 = 0; + + if (!twoside->attrib_back1) + twoside->attrib_front1 = 0; + + /* + * We'll multiply the primitive's determinant by this sign to determine + * if the triangle is back-facing (negative). + * sign = -1 for CCW, +1 for CW + */ + twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; + + stage->tri = twoside_tri; + stage->tri( stage, header ); +} + + +static void twoside_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->tri = twoside_first_tri; + stage->next->flush( stage->next, flags ); +} + + +static void twoside_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void twoside_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create twoside pipeline stage. + */ +struct draw_stage *draw_twoside_stage( struct draw_context *draw ) +{ + struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); + + draw_alloc_temp_verts( &twoside->stage, 3 ); + + twoside->stage.draw = draw; + twoside->stage.next = NULL; + twoside->stage.point = twoside_point; + twoside->stage.line = twoside_line; + twoside->stage.tri = twoside_first_tri; + twoside->stage.flush = twoside_flush; + twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; + twoside->stage.destroy = twoside_destroy; + + return &twoside->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_unfilled.c b/src/gallium/auxiliary/draw/draw_unfilled.c new file mode 100644 index 0000000000..8777cfdfc8 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_unfilled.c @@ -0,0 +1,206 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Drawing stage for handling glPolygonMode(line/point). + * Convert triangles to points or lines as needed. + */ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + +struct unfilled_stage { + struct draw_stage stage; + + /** [0] = front face, [1] = back face. + * legal values: PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE, + * and PIPE_POLYGON_MODE_POINT, + */ + unsigned mode[2]; +}; + + +static INLINE struct unfilled_stage *unfilled_stage( struct draw_stage *stage ) +{ + return (struct unfilled_stage *)stage; +} + + + +static void point( struct draw_stage *stage, + struct vertex_header *v0 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + stage->next->point( stage->next, &tmp ); +} + +static void line( struct draw_stage *stage, + struct vertex_header *v0, + struct vertex_header *v1 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + tmp.v[1] = v1; + stage->next->line( stage->next, &tmp ); +} + + +static void points( struct draw_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (header->edgeflags & 0x1) point( stage, v0 ); + if (header->edgeflags & 0x2) point( stage, v1 ); + if (header->edgeflags & 0x4) point( stage, v2 ); +} + + +static void lines( struct draw_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + +#if 0 + assert(((header->edgeflags & 0x1) >> 0) == header->v[0]->edgeflag); + assert(((header->edgeflags & 0x2) >> 1) == header->v[1]->edgeflag); + assert(((header->edgeflags & 0x4) >> 2) == header->v[2]->edgeflag); +#endif + + if (header->edgeflags & 0x1) line( stage, v0, v1 ); + if (header->edgeflags & 0x2) line( stage, v1, v2 ); + if (header->edgeflags & 0x4) line( stage, v2, v0 ); +} + + +/* Unfilled tri: + * + * Note edgeflags in the vertex struct is not sufficient as we will + * need to manipulate them when decomposing primitives??? + */ +static void unfilled_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + unsigned mode = unfilled->mode[header->det >= 0.0]; + + switch (mode) { + case PIPE_POLYGON_MODE_FILL: + stage->next->tri( stage->next, header ); + break; + case PIPE_POLYGON_MODE_LINE: + lines( stage, header ); + break; + case PIPE_POLYGON_MODE_POINT: + points( stage, header ); + break; + default: + abort(); + } +} + + +static void unfilled_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + + unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */ + unfilled->mode[1] = stage->draw->rasterizer->fill_cw; /* back */ + + stage->tri = unfilled_tri; + stage->tri( stage, header ); +} + + +static void unfilled_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void unfilled_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void unfilled_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->next->flush( stage->next, flags ); + + stage->tri = unfilled_first_tri; +} + + +static void unfilled_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void unfilled_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create unfilled triangle stage. + */ +struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) +{ + struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); + + draw_alloc_temp_verts( &unfilled->stage, 0 ); + + unfilled->stage.draw = draw; + unfilled->stage.next = NULL; + unfilled->stage.tmp = NULL; + unfilled->stage.point = unfilled_point; + unfilled->stage.line = unfilled_line; + unfilled->stage.tri = unfilled_first_tri; + unfilled->stage.flush = unfilled_flush; + unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; + unfilled->stage.destroy = unfilled_destroy; + + return &unfilled->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c new file mode 100644 index 0000000000..4375ebabbc --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -0,0 +1,185 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + + + + +/** + * Rebuild the rendering pipeline. + */ +static struct draw_stage *validate_pipeline( struct draw_stage *stage ) +{ + struct draw_context *draw = stage->draw; + struct draw_stage *next = draw->pipeline.rasterize; + int need_det = 0; + int precalc_flat = 0; + + /* Set the validate's next stage to the rasterize stage, so that it + * can be found later if needed for flushing. + */ + stage->next = next; + + /* + * NOTE: we build up the pipeline in end-to-start order. + * + * TODO: make the current primitive part of the state and build + * shorter pipelines for lines & points. + */ + + if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines) || + (draw->rasterizer->point_size != 1.0 && draw->convert_wide_points) || + draw->rasterizer->point_sprite) { + draw->pipeline.wide->next = next; + next = draw->pipeline.wide; + } + + if (draw->rasterizer->line_stipple_enable) { + draw->pipeline.stipple->next = next; + next = draw->pipeline.stipple; + precalc_flat = 1; /* only needed for lines really */ + } + + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { + draw->pipeline.unfilled->next = next; + next = draw->pipeline.unfilled; + precalc_flat = 1; /* only needed for triangles really */ + need_det = 1; + } + + if (draw->rasterizer->flatshade && precalc_flat) { + draw->pipeline.flatshade->next = next; + next = draw->pipeline.flatshade; + } + + if (draw->rasterizer->offset_cw || + draw->rasterizer->offset_ccw) { + draw->pipeline.offset->next = next; + next = draw->pipeline.offset; + need_det = 1; + } + + if (draw->rasterizer->light_twoside) { + draw->pipeline.twoside->next = next; + next = draw->pipeline.twoside; + need_det = 1; + } + + /* Always run the cull stage as we calculate determinant there + * also. + * + * This can actually be a win as culling out the triangles can lead + * to less work emitting vertices, smaller vertex buffers, etc. + * It's difficult to say whether this will be true in general. + */ + if (need_det || draw->rasterizer->cull_mode) { + draw->pipeline.cull->next = next; + next = draw->pipeline.cull; + } + + /* Clip stage + */ + if (!draw->rasterizer->bypass_clipping) + { + draw->pipeline.clip->next = next; + next = draw->pipeline.clip; + } + + + draw->pipeline.first = next; + return next; +} + +static void validate_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->tri( pipeline, header ); +} + +static void validate_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->line( pipeline, header ); +} + +static void validate_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->point( pipeline, header ); +} + +static void validate_reset_stipple_counter( struct draw_stage *stage ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->reset_stipple_counter( pipeline ); +} + +static void validate_flush( struct draw_stage *stage, + unsigned flags ) +{ + /* May need to pass a backend flush on to the rasterize stage. + */ + if (stage->next) + stage->next->flush( stage->next, flags ); +} + + +static void validate_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + +/** + * Create validate pipeline stage. + */ +struct draw_stage *draw_validate_stage( struct draw_context *draw ) +{ + struct draw_stage *stage = CALLOC_STRUCT(draw_stage); + + stage->draw = draw; + stage->next = NULL; + stage->point = validate_point; + stage->line = validate_line; + stage->tri = validate_tri; + stage->flush = validate_flush; + stage->reset_stipple_counter = validate_reset_stipple_counter; + stage->destroy = validate_destroy; + + return stage; +} diff --git a/src/gallium/auxiliary/draw/draw_vbuf.c b/src/gallium/auxiliary/draw/draw_vbuf.c new file mode 100644 index 0000000000..71ac73912b --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vbuf.c @@ -0,0 +1,570 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Vertex buffer drawing stage. + * + * \author José Fonseca + * \author Keith Whitwell + */ + + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" + +#include "draw_vbuf.h" +#include "draw_private.h" +#include "draw_vertex.h" +#include "draw_vf.h" + + +/** + * Vertex buffer emit stage. + */ +struct vbuf_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct vbuf_render *render; + + const struct vertex_info *vinfo; + + /** Vertex size in bytes */ + unsigned vertex_size; + + struct draw_vertex_fetch *vf; + + /* FIXME: we have no guarantee that 'unsigned' is 32bit */ + + /** Vertices in hardware format */ + unsigned *vertices; + unsigned *vertex_ptr; + unsigned max_vertices; + unsigned nr_vertices; + + /** Indices */ + ushort *indices; + unsigned max_indices; + unsigned nr_indices; + + /** Pipe primitive */ + unsigned prim; +}; + + +/** + * Basically a cast wrapper. + */ +static INLINE struct vbuf_stage * +vbuf_stage( struct draw_stage *stage ) +{ + assert(stage); + return (struct vbuf_stage *)stage; +} + + +static void vbuf_flush_indices( struct vbuf_stage *vbuf ); +static void vbuf_flush_vertices( struct vbuf_stage *vbuf ); +static void vbuf_alloc_vertices( struct vbuf_stage *vbuf ); + + +static INLINE boolean +overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) +{ + unsigned long used = (unsigned long) ((char *)ptr - (char *)map); + return (used + bytes) > bufsz; +} + + +static INLINE void +check_space( struct vbuf_stage *vbuf, unsigned nr ) +{ + if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { + vbuf_flush_vertices(vbuf); + vbuf_alloc_vertices(vbuf); + } + + if (vbuf->nr_indices + nr > vbuf->max_indices ) + vbuf_flush_indices(vbuf); +} + + +#if 0 +static INLINE void +dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) +{ + assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); + unsigned i, j, k; + + for (i = 0; i < vinfo->num_attribs; i++) { + j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + debug_printf("EMIT_OMIT:"); + break; + case EMIT_ALL: + assert(i == 0); + assert(j == 0); + debug_printf("EMIT_ALL:\t"); + for(k = 0; k < vinfo->size*4; ++k) + debug_printf("%02x ", *data++); + break; + case EMIT_1F: + debug_printf("EMIT_1F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_1F_PSIZE: + debug_printf("EMIT_1F_PSIZE:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_2F: + debug_printf("EMIT_2F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_3F: + debug_printf("EMIT_3F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + data += sizeof(float); + break; + case EMIT_4F: + debug_printf("EMIT_4F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_4UB: + debug_printf("EMIT_4UB:\t"); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + break; + default: + assert(0); + } + debug_printf("\n"); + } + debug_printf("\n"); +} +#endif + + +/** + * Extract the needed fields from post-transformed vertex and emit + * a hardware(driver) vertex. + * Recall that the vertices are constructed by the 'draw' module and + * have a couple of slots at the beginning (1-dword header, 4-dword + * clip pos) that we ignore here. We only use the vertex->data[] fields. + */ +static INLINE void +emit_vertex( struct vbuf_stage *vbuf, + struct vertex_header *vertex ) +{ +#if 0 + debug_printf("emit vertex %d to %p\n", + vbuf->nr_vertices, vbuf->vertex_ptr); +#endif + + if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { + if(vertex->vertex_id < vbuf->nr_vertices) + return; + else + debug_printf("Bad vertex id 0x%04x (>= 0x%04x)\n", + vertex->vertex_id, vbuf->nr_vertices); + return; + } + + vertex->vertex_id = vbuf->nr_vertices++; + + if(!vbuf->vf) { + const struct vertex_info *vinfo = vbuf->vinfo; + uint i; + uint count = 0; /* for debug/sanity */ + + assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); + + for (i = 0; i < vinfo->num_attribs; i++) { + uint j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + /* no-op */ + break; + case EMIT_ALL: + /* just copy the whole vertex as-is to the vbuf */ + assert(i == 0); + assert(j == 0); + memcpy(vbuf->vertex_ptr, vertex, vinfo->size * 4); + vbuf->vertex_ptr += vinfo->size; + count += vinfo->size; + break; + case EMIT_1F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + count++; + break; + case EMIT_1F_PSIZE: + *vbuf->vertex_ptr++ = fui(vbuf->stage.draw->rasterizer->point_size); + count++; + break; + case EMIT_2F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + count += 2; + break; + case EMIT_3F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); + count += 3; + break; + case EMIT_4F: + *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); + *vbuf->vertex_ptr++ = fui(vertex->data[j][3]); + count += 4; + break; + case EMIT_4UB: + *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), + float_to_ubyte( vertex->data[j][1] ), + float_to_ubyte( vertex->data[j][0] ), + float_to_ubyte( vertex->data[j][3] )); + count += 1; + break; + default: + assert(0); + } + } + assert(count == vinfo->size); +#if 0 + { + static float data[256]; + draw_vf_emit_vertex(vbuf->vf, vertex, data); + if(memcmp((uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size, data, vbuf->vertex_size)) { + debug_printf("With VF:\n"); + dump_emitted_vertex(vbuf->vinfo, (uint8_t *)data); + debug_printf("Without VF:\n"); + dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size); + assert(0); + } + } +#endif + } + else { + draw_vf_emit_vertex(vbuf->vf, vertex, vbuf->vertex_ptr); + + vbuf->vertex_ptr += vbuf->vertex_size/4; + } +} + + +static void +vbuf_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + check_space( vbuf, 3 ); + + for (i = 0; i < 3; i++) { + emit_vertex( vbuf, prim->v[i] ); + + vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; + } +} + + +static void +vbuf_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + check_space( vbuf, 2 ); + + for (i = 0; i < 2; i++) { + emit_vertex( vbuf, prim->v[i] ); + + vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; + } +} + + +static void +vbuf_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + check_space( vbuf, 1 ); + + emit_vertex( vbuf, prim->v[0] ); + + vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[0]->vertex_id; +} + + +/** + * Set the prim type for subsequent vertices. + * This may result in a new vertex size. The existing vbuffer (if any) + * will be flushed if needed and a new one allocated. + */ +static void +vbuf_set_prim( struct vbuf_stage *vbuf, uint newprim ) +{ + const struct vertex_info *vinfo; + unsigned vertex_size; + + assert(newprim == PIPE_PRIM_POINTS || + newprim == PIPE_PRIM_LINES || + newprim == PIPE_PRIM_TRIANGLES); + + vbuf->prim = newprim; + vbuf->render->set_primitive(vbuf->render, newprim); + + vinfo = vbuf->render->get_vertex_info(vbuf->render); + vertex_size = vinfo->size * sizeof(float); + + if (vertex_size != vbuf->vertex_size) + vbuf_flush_vertices(vbuf); + + vbuf->vinfo = vinfo; + vbuf->vertex_size = vertex_size; + if(vbuf->vf) + draw_vf_set_vertex_info(vbuf->vf, + vbuf->vinfo, + vbuf->stage.draw->rasterizer->point_size); + + if (!vbuf->vertices) + vbuf_alloc_vertices(vbuf); +} + + +static void +vbuf_first_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + stage->tri = vbuf_tri; + vbuf_set_prim(vbuf, PIPE_PRIM_TRIANGLES); + stage->tri( stage, prim ); +} + + +static void +vbuf_first_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + stage->line = vbuf_line; + vbuf_set_prim(vbuf, PIPE_PRIM_LINES); + stage->line( stage, prim ); +} + + +static void +vbuf_first_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + stage->point = vbuf_point; + vbuf_set_prim(vbuf, PIPE_PRIM_POINTS); + stage->point( stage, prim ); +} + + +static void +vbuf_flush_indices( struct vbuf_stage *vbuf ) +{ + if(!vbuf->nr_indices) + return; + + assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == + vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); + + switch(vbuf->prim) { + case PIPE_PRIM_POINTS: + break; + case PIPE_PRIM_LINES: + assert(vbuf->nr_indices % 2 == 0); + break; + case PIPE_PRIM_TRIANGLES: + assert(vbuf->nr_indices % 3 == 0); + break; + default: + assert(0); + } + + vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); + + vbuf->nr_indices = 0; + + /* don't need to reset point/line/tri functions */ +#if 0 + stage->point = vbuf_first_point; + stage->line = vbuf_first_line; + stage->tri = vbuf_first_tri; +#endif +} + + +/** + * Flush existing vertex buffer and allocate a new one. + * + * XXX: We separate flush-on-index-full and flush-on-vb-full, but may + * raise issues uploading vertices if the hardware wants to flush when + * we flush. + */ +static void +vbuf_flush_vertices( struct vbuf_stage *vbuf ) +{ + if(vbuf->vertices) { + vbuf_flush_indices(vbuf); + + /* Reset temporary vertices ids */ + if(vbuf->nr_vertices) + draw_reset_vertex_ids( vbuf->stage.draw ); + + /* Free the vertex buffer */ + vbuf->render->release_vertices(vbuf->render, + vbuf->vertices, + vbuf->vertex_size, + vbuf->nr_vertices); + vbuf->max_vertices = vbuf->nr_vertices = 0; + vbuf->vertex_ptr = vbuf->vertices = NULL; + + } +} + + +static void +vbuf_alloc_vertices( struct vbuf_stage *vbuf ) +{ + assert(!vbuf->nr_indices); + assert(!vbuf->vertices); + + /* Allocate a new vertex buffer */ + vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; + vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, + (ushort) vbuf->vertex_size, + (ushort) vbuf->max_vertices); + vbuf->vertex_ptr = vbuf->vertices; +} + + + +static void +vbuf_flush( struct draw_stage *stage, unsigned flags ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + + stage->point = vbuf_first_point; + stage->line = vbuf_first_line; + stage->tri = vbuf_first_tri; + + if (flags & DRAW_FLUSH_BACKEND) + vbuf_flush_vertices( vbuf ); +} + + +static void +vbuf_reset_stipple_counter( struct draw_stage *stage ) +{ + /* XXX: Need to do something here for hardware with linestipple. + */ + (void) stage; +} + + +static void vbuf_destroy( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + if(vbuf->indices) + align_free( vbuf->indices ); + + if(vbuf->vf) + draw_vf_destroy( vbuf->vf ); + + if (vbuf->render) + vbuf->render->destroy( vbuf->render ); + + FREE( stage ); +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *draw_vbuf_stage( struct draw_context *draw, + struct vbuf_render *render ) +{ + struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); + + if(!vbuf) + return NULL; + + vbuf->stage.draw = draw; + vbuf->stage.point = vbuf_first_point; + vbuf->stage.line = vbuf_first_line; + vbuf->stage.tri = vbuf_first_tri; + vbuf->stage.flush = vbuf_flush; + vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter; + vbuf->stage.destroy = vbuf_destroy; + + vbuf->render = render; + + assert(render->max_indices < UNDEFINED_VERTEX_ID); + vbuf->max_indices = render->max_indices; + vbuf->indices = (ushort *) + align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); + if(!vbuf->indices) + vbuf_destroy(&vbuf->stage); + + vbuf->vertices = NULL; + vbuf->vertex_ptr = vbuf->vertices; + + vbuf->prim = ~0; + + if(!GETENV("GALLIUM_NOVF")) + vbuf->vf = draw_vf_create(); + + return &vbuf->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_vbuf.h b/src/gallium/auxiliary/draw/draw_vbuf.h new file mode 100644 index 0000000000..cfd2b9820c --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vbuf.h @@ -0,0 +1,106 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Vertex buffer drawing stage. + * + * \author Keith Whitwell + * \author José Fonseca + */ + +#ifndef DRAW_VBUF_H_ +#define DRAW_VBUF_H_ + + +#include "pipe/p_util.h" + + +struct draw_context; +struct vertex_info; + + +/** + * Interface for hardware vertex buffer rendering. + */ +struct vbuf_render { + + /** + * Driver limits. May be tuned lower to improve cache hits on + * index list. + */ + unsigned max_indices; + unsigned max_vertex_buffer_bytes; + + /** + * Get the hardware vertex format. + * + * XXX: have this in draw_context instead? + */ + const struct vertex_info *(*get_vertex_info)( struct vbuf_render * ); + + /** + * Request a destination for vertices. + * Hardware renderers will use ttm memory, others will just malloc + * something. + */ + void *(*allocate_vertices)( struct vbuf_render *, + ushort vertex_size, + ushort nr_vertices ); + + /** + * Notify the renderer of the current primitive when it changes. + * Prim is restricted to TRIANGLES, LINES and POINTS. + */ + void (*set_primitive)( struct vbuf_render *, unsigned prim ); + + /** + * DrawElements, note indices are ushort: + */ + void (*draw)( struct vbuf_render *, + const ushort *indices, + uint nr_indices ); + + /** + * Called when vbuf is done with this set of vertices: + */ + void (*release_vertices)( struct vbuf_render *, + void *vertices, + unsigned vertex_size, + unsigned vertices_used ); + + void (*destroy)( struct vbuf_render * ); +}; + + + +struct draw_stage * +draw_vbuf_stage( struct draw_context *draw, + struct vbuf_render *render ); + + +#endif /*DRAW_VBUF_H_*/ diff --git a/src/gallium/auxiliary/draw/draw_vertex.c b/src/gallium/auxiliary/draw/draw_vertex.c new file mode 100644 index 0000000000..daf1ef4b80 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vertex.c @@ -0,0 +1,79 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * Functions for specifying the post-transformation vertex layout. + * + * Author: + * Brian Paul + * Keith Whitwell + */ + + +#include "draw/draw_private.h" +#include "draw/draw_vertex.h" + + +/** + * Compute the size of a vertex, in dwords/floats, to update the + * vinfo->size field. + */ +void +draw_compute_vertex_size(struct vertex_info *vinfo) +{ + uint i; + + vinfo->size = 0; + for (i = 0; i < vinfo->num_attribs; i++) { + switch (vinfo->emit[i]) { + case EMIT_OMIT: + break; + case EMIT_4UB: + /* fall-through */ + case EMIT_1F_PSIZE: + /* fall-through */ + case EMIT_1F: + vinfo->size += 1; + break; + case EMIT_2F: + vinfo->size += 2; + break; + case EMIT_3F: + vinfo->size += 3; + break; + case EMIT_4F: + vinfo->size += 4; + break; + case EMIT_ALL: + /* fall-through */ + default: + assert(0); + } + } + + assert(vinfo->size * 4 <= MAX_VERTEX_SIZE); +} diff --git a/src/gallium/auxiliary/draw/draw_vertex.h b/src/gallium/auxiliary/draw/draw_vertex.h new file mode 100644 index 0000000000..267c74203b --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vertex.h @@ -0,0 +1,111 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Post-transform vertex format info. The vertex_info struct is used by + * the draw_vbuf code to emit hardware-specific vertex layouts into hw + * vertex buffers. + * + * Author: + * Brian Paul + */ + + +#ifndef DRAW_VERTEX_H +#define DRAW_VERTEX_H + + +#include "pipe/p_state.h" + + +/** + * Vertex attribute emit modes + */ +enum attrib_emit { + EMIT_OMIT, /**< don't emit the attribute */ + EMIT_ALL, /**< emit whole post-xform vertex, w/ header */ + EMIT_1F, + EMIT_1F_PSIZE, /**< insert constant point size */ + EMIT_2F, + EMIT_3F, + EMIT_4F, + EMIT_4UB /**< XXX may need variations for RGBA vs BGRA, etc */ +}; + + +/** + * Attribute interpolation mode + */ +enum interp_mode { + INTERP_NONE, /**< never interpolate vertex header info */ + INTERP_POS, /**< special case for frag position */ + INTERP_CONSTANT, + INTERP_LINEAR, + INTERP_PERSPECTIVE +}; + + +/** + * Information about hardware/rasterization vertex layout. + */ +struct vertex_info +{ + uint num_attribs; + uint hwfmt[4]; /**< hardware format info for this format */ + enum interp_mode interp_mode[PIPE_MAX_SHADER_INPUTS]; + enum attrib_emit emit[PIPE_MAX_SHADER_INPUTS]; /**< EMIT_x */ + uint src_index[PIPE_MAX_SHADER_INPUTS]; /**< map to post-xform attribs */ + uint size; /**< total vertex size in dwords */ +}; + + + +/** + * Add another attribute to the given vertex_info object. + * \param src_index indicates which post-transformed vertex attrib slot + * corresponds to this attribute. + * \return slot in which the attribute was added + */ +static INLINE uint +draw_emit_vertex_attr(struct vertex_info *vinfo, + enum attrib_emit emit, enum interp_mode interp, + uint src_index) +{ + const uint n = vinfo->num_attribs; + assert(n < PIPE_MAX_SHADER_INPUTS); + vinfo->emit[n] = emit; + vinfo->interp_mode[n] = interp; + vinfo->src_index[n] = src_index; + vinfo->num_attribs++; + return n; +} + + +extern void draw_compute_vertex_size(struct vertex_info *vinfo); + + +#endif /* DRAW_VERTEX_H */ diff --git a/src/gallium/auxiliary/draw/draw_vertex_cache.c b/src/gallium/auxiliary/draw/draw_vertex_cache.c new file mode 100644 index 0000000000..44427999cc --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vertex_cache.c @@ -0,0 +1,196 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw_private.h" +#include "draw_context.h" + + +void draw_vertex_cache_invalidate( struct draw_context *draw ) +{ + assert(draw->pq.queue_nr == 0); + assert(draw->vs.queue_nr == 0); + assert(draw->vcache.referenced == 0); + + memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); +} + + +/** + * Check if vertex is in cache, otherwise add it. It won't go through + * VS yet, not until there is a flush operation or the VS queue fills up. + * + * Note that cache entries are basically just two pointers: the first + * an index into the user's vertex arrays, the second a location in + * the vertex shader cache for the post-transformed vertex. + * + * \return pointer to location of (post-transformed) vertex header in the cache + */ +static struct vertex_header *get_vertex( struct draw_context *draw, + unsigned i ) +{ + unsigned slot = (i + (i>>5)) % VCACHE_SIZE; + + assert(slot < 32); /* so we don't exceed the bitfield size below */ + + /* Cache miss? + */ + if (draw->vcache.idx[slot] != i) { + + /* If slot is in use, use the overflow area: + */ + if (draw->vcache.referenced & (1 << slot)) { + slot = VCACHE_SIZE + draw->vcache.overflow++; + } + + assert(slot < Elements(draw->vcache.idx)); + + draw->vcache.idx[slot] = i; + + /* Add to vertex shader queue: + */ + assert(draw->vs.queue_nr < VS_QUEUE_LENGTH); + draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; + draw->vs.queue[draw->vs.queue_nr].elt = i; + draw->vs.queue_nr++; + + /* Need to set the vertex's edge flag here. If we're being called + * by do_ef_triangle(), that function needs edge flag info! + */ + draw->vcache.vertex[slot]->clipmask = 0; + draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ + draw->vcache.vertex[slot]->pad = 0; + draw->vcache.vertex[slot]->vertex_id = UNDEFINED_VERTEX_ID; + } + + + /* primitive flushing may have cleared the bitfield but did not + * clear the idx[] array values. Set the bit now. This fixes a + * bug found when drawing long triangle fans. + */ + draw->vcache.referenced |= (1 << slot); + return draw->vcache.vertex[slot]; +} + + +static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const unsigned *elts = (const unsigned *) draw->user.elts; + return get_vertex( draw, elts[i] ); +} + + +static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const ushort *elts = (const ushort *) draw->user.elts; + return get_vertex( draw, elts[i] ); +} + + +static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, + unsigned i ) +{ + const ubyte *elts = (const ubyte *) draw->user.elts; + return get_vertex( draw, elts[i] ); +} + + +void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) +{ + unsigned i; + + for (i = 0; i < Elements(draw->vcache.vertex); i++) + draw->vcache.vertex[i]->vertex_id = UNDEFINED_VERTEX_ID; +} + + +void draw_vertex_cache_unreference( struct draw_context *draw ) +{ + draw->vcache.referenced = 0; + draw->vcache.overflow = 0; +} + + +int draw_vertex_cache_check_space( struct draw_context *draw, + unsigned nr_verts ) +{ + if (draw->vcache.overflow + nr_verts < VCACHE_OVERFLOW) { + /* The vs queue is sized so that this can never happen: + */ + assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); + return TRUE; + } + else + return FALSE; +} + + + +/** + * Tell the drawing context about the index/element buffer to use + * (ala glDrawElements) + * If no element buffer is to be used (i.e. glDrawArrays) then this + * should be called with eltSize=0 and elements=NULL. + * + * \param draw the drawing context + * \param eltSize size of each element (1, 2 or 4 bytes) + * \param elements the element buffer ptr + */ +void +draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ) +{ +// draw_statechange( draw ); + + /* choose the get_vertex() function to use */ + switch (eltSize) { + case 0: + draw->vcache.get_vertex = get_vertex; + break; + case 1: + draw->vcache.get_vertex = get_ubyte_elt_vertex; + break; + case 2: + draw->vcache.get_vertex = get_ushort_elt_vertex; + break; + case 4: + draw->vcache.get_vertex = get_uint_elt_vertex; + break; + default: + assert(0); + } + draw->user.elts = elements; + draw->user.eltSize = eltSize; +} + diff --git a/src/gallium/auxiliary/draw/draw_vertex_fetch.c b/src/gallium/auxiliary/draw/draw_vertex_fetch.c new file mode 100644 index 0000000000..e13df04605 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vertex_fetch.c @@ -0,0 +1,510 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" +#include "draw_context.h" + + +#define DRAW_DBG 0 + + +/** + * Fetch a float[4] vertex attribute from memory, doing format/type + * conversion as needed. + * + * This is probably needed/dupliocated elsewhere, eg format + * conversion, texture sampling etc. + */ +#define FETCH_ATTRIB( NAME, SZ, CVT ) \ +static void \ +fetch_##NAME(const void *ptr, float *attrib) \ +{ \ + static const float defaults[4] = { 0,0,0,1 }; \ + int i; \ + \ + for (i = 0; i < SZ; i++) { \ + attrib[i] = CVT; \ + } \ + \ + for (; i < 4; i++) { \ + attrib[i] = defaults[i]; \ + } \ +} + +#define CVT_64_FLOAT (float) ((double *) ptr)[i] +#define CVT_32_FLOAT ((float *) ptr)[i] + +#define CVT_8_USCALED (float) ((unsigned char *) ptr)[i] +#define CVT_16_USCALED (float) ((unsigned short *) ptr)[i] +#define CVT_32_USCALED (float) ((unsigned int *) ptr)[i] + +#define CVT_8_SSCALED (float) ((char *) ptr)[i] +#define CVT_16_SSCALED (float) ((short *) ptr)[i] +#define CVT_32_SSCALED (float) ((int *) ptr)[i] + +#define CVT_8_UNORM (float) ((unsigned char *) ptr)[i] / 255.0f +#define CVT_16_UNORM (float) ((unsigned short *) ptr)[i] / 65535.0f +#define CVT_32_UNORM (float) ((unsigned int *) ptr)[i] / 4294967295.0f + +#define CVT_8_SNORM (float) ((char *) ptr)[i] / 127.0f +#define CVT_16_SNORM (float) ((short *) ptr)[i] / 32767.0f +#define CVT_32_SNORM (float) ((int *) ptr)[i] / 2147483647.0f + +FETCH_ATTRIB( R64G64B64A64_FLOAT, 4, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) +FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) +FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) + +FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) +FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) +FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) + +FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) +FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) +FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) + +FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) +FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) +FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) + +FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) +FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) +FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) + +FETCH_ATTRIB( R32G32B32A32_SNORM, 4, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32B32_SNORM, 3, CVT_32_SNORM ) +FETCH_ATTRIB( R32G32_SNORM, 2, CVT_32_SNORM ) +FETCH_ATTRIB( R32_SNORM, 1, CVT_32_SNORM ) + +FETCH_ATTRIB( R16G16B16A16_USCALED, 4, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16B16_USCALED, 3, CVT_16_USCALED ) +FETCH_ATTRIB( R16G16_USCALED, 2, CVT_16_USCALED ) +FETCH_ATTRIB( R16_USCALED, 1, CVT_16_USCALED ) + +FETCH_ATTRIB( R16G16B16A16_SSCALED, 4, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16B16_SSCALED, 3, CVT_16_SSCALED ) +FETCH_ATTRIB( R16G16_SSCALED, 2, CVT_16_SSCALED ) +FETCH_ATTRIB( R16_SSCALED, 1, CVT_16_SSCALED ) + +FETCH_ATTRIB( R16G16B16A16_UNORM, 4, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16B16_UNORM, 3, CVT_16_UNORM ) +FETCH_ATTRIB( R16G16_UNORM, 2, CVT_16_UNORM ) +FETCH_ATTRIB( R16_UNORM, 1, CVT_16_UNORM ) + +FETCH_ATTRIB( R16G16B16A16_SNORM, 4, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16B16_SNORM, 3, CVT_16_SNORM ) +FETCH_ATTRIB( R16G16_SNORM, 2, CVT_16_SNORM ) +FETCH_ATTRIB( R16_SNORM, 1, CVT_16_SNORM ) + +FETCH_ATTRIB( R8G8B8A8_USCALED, 4, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8B8_USCALED, 3, CVT_8_USCALED ) +FETCH_ATTRIB( R8G8_USCALED, 2, CVT_8_USCALED ) +FETCH_ATTRIB( R8_USCALED, 1, CVT_8_USCALED ) + +FETCH_ATTRIB( R8G8B8A8_SSCALED, 4, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8B8_SSCALED, 3, CVT_8_SSCALED ) +FETCH_ATTRIB( R8G8_SSCALED, 2, CVT_8_SSCALED ) +FETCH_ATTRIB( R8_SSCALED, 1, CVT_8_SSCALED ) + +FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8B8_UNORM, 3, CVT_8_UNORM ) +FETCH_ATTRIB( R8G8_UNORM, 2, CVT_8_UNORM ) +FETCH_ATTRIB( R8_UNORM, 1, CVT_8_UNORM ) + +FETCH_ATTRIB( R8G8B8A8_SNORM, 4, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8B8_SNORM, 3, CVT_8_SNORM ) +FETCH_ATTRIB( R8G8_SNORM, 2, CVT_8_SNORM ) +FETCH_ATTRIB( R8_SNORM, 1, CVT_8_SNORM ) + +FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) +//FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) + + + +static fetch_func get_fetch_func( enum pipe_format format ) +{ +#if 0 + { + char tmp[80]; + pf_sprint_name(tmp, format); + debug_printf("%s: %s\n", __FUNCTION__, tmp); + } +#endif + + switch (format) { + case PIPE_FORMAT_R64_FLOAT: + return fetch_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return fetch_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return fetch_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return fetch_R64G64B64A64_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + return fetch_R32_FLOAT; + case PIPE_FORMAT_R32G32_FLOAT: + return fetch_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return fetch_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return fetch_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return fetch_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return fetch_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return fetch_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return fetch_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return fetch_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return fetch_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return fetch_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return fetch_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return fetch_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return fetch_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return fetch_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return fetch_R32G32B32A32_SNORM; + + case PIPE_FORMAT_R32_SSCALED: + return fetch_R32_SSCALED; + case PIPE_FORMAT_R32G32_SSCALED: + return fetch_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return fetch_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return fetch_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return fetch_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return fetch_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return fetch_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return fetch_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return fetch_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return fetch_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return fetch_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return fetch_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return fetch_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return fetch_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return fetch_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return fetch_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return fetch_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return fetch_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return fetch_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return fetch_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return fetch_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return fetch_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return fetch_R8G8B8_UNORM; + case PIPE_FORMAT_R8G8B8A8_UNORM: + return fetch_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return fetch_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return fetch_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return fetch_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return fetch_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return fetch_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return fetch_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return fetch_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return fetch_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return fetch_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return fetch_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return fetch_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return fetch_R8G8B8A8_SSCALED; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + return fetch_A8R8G8B8_UNORM; + + case 0: + return NULL; /* not sure why this is needed */ + + default: + assert(0); + return NULL; + } +} + + +static void +transpose_4x4( float *out, const float *in ) +{ + /* This can be achieved in 12 sse instructions, plus the final + * stores I guess. This is probably a bit more than that - maybe + * 32 or so? + */ + out[0] = in[0]; out[1] = in[4]; out[2] = in[8]; out[3] = in[12]; + out[4] = in[1]; out[5] = in[5]; out[6] = in[9]; out[7] = in[13]; + out[8] = in[2]; out[9] = in[6]; out[10] = in[10]; out[11] = in[14]; + out[12] = in[3]; out[13] = in[7]; out[14] = in[11]; out[15] = in[15]; +} + + + +static void fetch_xyz_rgb( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + + assert(count <= 4); + +// debug_printf("%s\n", __FUNCTION__); + + /* loop over vertex attributes (vertex shader inputs) + */ + + for (i = 0; i < 4; i++) { + { + const float *in = (const float *)(src[0] + elts[i] * pitch[0]); + float *out = &machine->Inputs[0].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[1] + elts[i] * pitch[1]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + } +} + + + + +static void fetch_xyz_rgb_st( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + + assert(count <= 4); + + /* loop over vertex attributes (vertex shader inputs) + */ + + for (i = 0; i < 4; i++) { + { + const float *in = (const float *)(src[0] + elts[i] * pitch[0]); + float *out = &machine->Inputs[0].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[1] + elts[i] * pitch[1]); + float *out = &machine->Inputs[1].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = in[2]; + out[12] = 1.0f; + } + + { + const float *in = (const float *)(src[2] + elts[i] * pitch[2]); + float *out = &machine->Inputs[2].xyzw[0].f[i]; + out[0] = in[0]; + out[4] = in[1]; + out[8] = 0.0f; + out[12] = 1.0f; + } + } +} + + + + +/** + * Fetch vertex attributes for 'count' vertices. + */ +static void generic_vertex_fetch( struct draw_context *draw, + struct tgsi_exec_machine *machine, + const unsigned *elts, + unsigned count ) +{ + unsigned nr_attrs = draw->vertex_fetch.nr_attrs; + unsigned attr; + + assert(count <= 4); + +// debug_printf("%s %d\n", __FUNCTION__, count); + + /* loop over vertex attributes (vertex shader inputs) + */ + for (attr = 0; attr < nr_attrs; attr++) { + + const unsigned pitch = draw->vertex_fetch.pitch[attr]; + const ubyte *src = draw->vertex_fetch.src_ptr[attr]; + const fetch_func fetch = draw->vertex_fetch.fetch[attr]; + unsigned i; + float p[4][4]; + + + /* Fetch four attributes for four vertices. + * + * Could fetch directly into AOS format, but this is meant to be + * a prototype for an sse implementation, which would have + * difficulties doing that. + */ + for (i = 0; i < count; i++) + fetch( src + elts[i] * pitch, p[i] ); + + /* Be nice and zero out any missing vertices: + */ + for ( ; i < 4; i++) + p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; + + /* Transpose/swizzle into sse-friendly format. Currently + * assuming that all vertex shader inputs are float[4], but this + * isn't true -- if the vertex shader only wants tex0.xy, we + * could optimize for that. + * + * To do so fully without codegen would probably require an + * excessive number of fetch functions, but we could at least + * minimize the transpose step: + */ + transpose_4x4( (float *)&machine->Inputs[attr].xyzw[0].f[0], (float *)p ); + } +} + + + +void draw_update_vertex_fetch( struct draw_context *draw ) +{ + unsigned nr_attrs, i; + +// debug_printf("%s\n", __FUNCTION__); + + /* this may happend during context init */ + if (!draw->vertex_shader) + return; + + nr_attrs = draw->vertex_shader->state->num_inputs; + + for (i = 0; i < nr_attrs; i++) { + unsigned buf = draw->vertex_element[i].vertex_buffer_index; + enum pipe_format format = draw->vertex_element[i].src_format; + + draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[i].src_offset; + + draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; + draw->vertex_fetch.fetch[i] = get_fetch_func( format ); + } + + draw->vertex_fetch.nr_attrs = nr_attrs; + + draw->vertex_fetch.fetch_func = generic_vertex_fetch; + + switch (nr_attrs) { + case 2: + if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT) + draw->vertex_fetch.fetch_func = fetch_xyz_rgb; + break; + case 3: + if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT && + draw->vertex_element[2].src_format == PIPE_FORMAT_R32G32_FLOAT) + draw->vertex_fetch.fetch_func = fetch_xyz_rgb_st; + break; + default: + break; + } + +} diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c new file mode 100644 index 0000000000..377ecbb931 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -0,0 +1,325 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#if defined(__i386__) || defined(__386__) +#include "tgsi/exec/tgsi_sse2.h" +#endif +#include "draw_private.h" +#include "draw_context.h" + +#include "x86/rtasm/x86sse.h" +#include "llvm/gallivm.h" + + +#define DBG_VS 0 + + +static INLINE unsigned +compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<machine; + unsigned int j; + + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); + assert(draw->vertex_shader->state->output_semantic_name[0] + == TGSI_SEMANTIC_POSITION); + + /* Consts does not require 16 byte alignment. */ + machine->Consts = (float (*)[4]) draw->user.constants; + + machine->Inputs = ALIGN16_ASSIGN(inputs); + machine->Outputs = ALIGN16_ASSIGN(outputs); + + draw->vertex_fetch.fetch_func( draw, machine, elts, count ); + + /* run shader */ +#ifdef MESA_LLVM + if (1) { + struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; + gallivm_cpu_vs_exec(prog, + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps); + } else +#elif defined(__i386__) || defined(__386__) + if (draw->use_sse) { + /* SSE */ + /* cast away const */ + struct draw_vertex_shader *shader + = (struct draw_vertex_shader *)draw->vertex_shader; + codegen_function func + = (codegen_function) x86_get_func( &shader->sse2_program ); + + if (func) + func( + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps ); + else + /* interpreter */ + tgsi_exec_machine_run( machine ); + } + else +#endif + { + /* interpreter */ + tgsi_exec_machine_run( machine ); + } + + /* store machine results */ + for (j = 0; j < count; j++) { + unsigned slot; + float x, y, z, w; + + /* Handle attr[0] (position) specially: + * + * XXX: Computing the clipmask should be done in the vertex + * program as a set of DP4 instructions appended to the + * user-provided code. + */ + x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + +#if DBG_VS + debug_printf("output[%d]win: %f %f %f %f\n", j, + vOut[j]->data[0][0], + vOut[j]->data[0][1], + vOut[j]->data[0][2], + vOut[j]->data[0][3]); +#endif + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; +#if DBG_VS + debug_printf("output[%d][%d]: %f %f %f %f\n", j, slot, + vOut[j]->data[slot][0], + vOut[j]->data[slot][1], + vOut[j]->data[slot][2], + vOut[j]->data[slot][3]); +#endif + } + } /* loop over vertices */ +} + + +/** + * Run the vertex shader on all vertices in the vertex queue. + * Called by the draw module when the vertx cache needs to be flushed. + */ +void +draw_vertex_shader_queue_flush(struct draw_context *draw) +{ + unsigned i; + + assert(draw->vs.queue_nr != 0); + + /* XXX: do this on statechange: + */ + draw_update_vertex_fetch( draw ); + +// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); + + /* run vertex shader on vertex cache entries, four per invokation */ + for (i = 0; i < draw->vs.queue_nr; i += 4) { + struct vertex_header *dests[4]; + unsigned elts[4]; + int j, n = MIN2(4, draw->vs.queue_nr - i); + + for (j = 0; j < n; j++) { + elts[j] = draw->vs.queue[i + j].elt; + dests[j] = draw->vs.queue[i + j].dest; + } + + for ( ; j < 4; j++) { + elts[j] = elts[0]; + dests[j] = dests[0]; + } + + assert(n > 0); + assert(n <= 4); + + run_vertex_program(draw, elts, n, dests); + } + + draw->vs.queue_nr = 0; +} + + +struct draw_vertex_shader * +draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader) +{ + struct draw_vertex_shader *vs; + + vs = CALLOC_STRUCT( draw_vertex_shader ); + if (vs == NULL) { + return NULL; + } + + vs->state = shader; + +#ifdef MESA_LLVM + struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); + gallivm_ir_set_layout(ir, GALLIVM_SOA); + gallivm_ir_set_components(ir, 4); + gallivm_ir_fill_from_tgsi(ir, shader->tokens); + vs->llvm_prog = gallivm_ir_compile(ir); + gallivm_ir_delete(ir); + + draw->engine = gallivm_global_cpu_engine(); + if (!draw->engine) { + draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); + } + else { + gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); + } +#elif defined(__i386__) || defined(__386__) + if (draw->use_sse) { + /* cast-away const */ + struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; + + x86_init_func( &vs->sse2_program ); + if (!tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, + &vs->sse2_program )) { + x86_release_func( (struct x86_function *) &vs->sse2_program ); + fprintf(stdout /*err*/, + "tgsi_emit_sse2() failed, falling back to interpreter\n"); + } + } +#endif + + return vs; +} + + +void +draw_bind_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + draw->vertex_shader = dvs; + draw->num_vs_outputs = dvs->state->num_outputs; + + /* specify the vertex program to interpret/execute */ + tgsi_exec_machine_init(&draw->machine, + draw->vertex_shader->state->tokens, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/ ); +} + + +void +draw_delete_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs) +{ +#if defined(__i386__) || defined(__386__) + x86_release_func( (struct x86_function *) &dvs->sse2_program ); +#endif + + FREE( dvs ); +} diff --git a/src/gallium/auxiliary/draw/draw_vf.c b/src/gallium/auxiliary/draw/draw_vf.c new file mode 100644 index 0000000000..dc3a5ecd21 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vf.c @@ -0,0 +1,428 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + + +#include + +#include "pipe/p_compiler.h" +#include "pipe/p_util.h" + +#include "draw_vf.h" + + +#define DRAW_VF_DBG 0 + + +/* TODO: remove this */ +extern void +_mesa_exec_free( void *addr ); + + +static boolean match_fastpath( struct draw_vertex_fetch *vf, + const struct draw_vf_fastpath *fp) +{ + unsigned j; + + if (vf->attr_count != fp->attr_count) + return FALSE; + + for (j = 0; j < vf->attr_count; j++) + if (vf->attr[j].format != fp->attr[j].format || + vf->attr[j].inputsize != fp->attr[j].size || + vf->attr[j].vertoffset != fp->attr[j].offset) + return FALSE; + + if (fp->match_strides) { + if (vf->vertex_stride != fp->vertex_stride) + return FALSE; + + for (j = 0; j < vf->attr_count; j++) + if (vf->attr[j].inputstride != fp->attr[j].stride) + return FALSE; + } + + return TRUE; +} + +static boolean search_fastpath_emit( struct draw_vertex_fetch *vf ) +{ + struct draw_vf_fastpath *fp = vf->fastpath; + + for ( ; fp ; fp = fp->next) { + if (match_fastpath(vf, fp)) { + vf->emit = fp->func; + return TRUE; + } + } + + return FALSE; +} + +void draw_vf_register_fastpath( struct draw_vertex_fetch *vf, + boolean match_strides ) +{ + struct draw_vf_fastpath *fastpath = CALLOC_STRUCT(draw_vf_fastpath); + unsigned i; + + fastpath->vertex_stride = vf->vertex_stride; + fastpath->attr_count = vf->attr_count; + fastpath->match_strides = match_strides; + fastpath->func = vf->emit; + fastpath->attr = (struct draw_vf_attr_type *) + MALLOC(vf->attr_count * sizeof(fastpath->attr[0])); + + for (i = 0; i < vf->attr_count; i++) { + fastpath->attr[i].format = vf->attr[i].format; + fastpath->attr[i].stride = vf->attr[i].inputstride; + fastpath->attr[i].size = vf->attr[i].inputsize; + fastpath->attr[i].offset = vf->attr[i].vertoffset; + } + + fastpath->next = vf->fastpath; + vf->fastpath = fastpath; +} + + + + +/*********************************************************************** + * Build codegen functions or return generic ones: + */ +static void choose_emit_func( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *dest) +{ + vf->emit = NULL; + + /* Does this match an existing (hardwired, codegen or known-bad) + * fastpath? + */ + if (search_fastpath_emit(vf)) { + /* Use this result. If it is null, then it is already known + * that the current state will fail for codegen and there is no + * point trying again. + */ + } + else if (vf->codegen_emit) { + vf->codegen_emit( vf ); + } + + if (!vf->emit) { + draw_vf_generate_hardwired_emit(vf); + } + + /* Otherwise use the generic version: + */ + if (!vf->emit) + vf->emit = draw_vf_generic_emit; + + vf->emit( vf, count, dest ); +} + + + + + +/*********************************************************************** + * Public entrypoints, mostly dispatch to the above: + */ + + + +static unsigned +draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, + const struct draw_vf_attr_map *map, + unsigned nr, + unsigned vertex_stride ) +{ + unsigned offset = 0; + unsigned i, j; + + assert(nr < PIPE_ATTRIB_MAX); + + for (j = 0, i = 0; i < nr; i++) { + const unsigned format = map[i].format; + if (format == DRAW_EMIT_PAD) { +#if (DRAW_VF_DBG) + debug_printf("%d: pad %d, offset %d\n", i, + map[i].offset, offset); +#endif + + offset += map[i].offset; + + } + else { + vf->attr[j].attrib = map[i].attrib; + vf->attr[j].format = format; + vf->attr[j].insert = draw_vf_format_info[format].insert; + vf->attr[j].vertattrsize = draw_vf_format_info[format].attrsize; + vf->attr[j].vertoffset = offset; + vf->attr[j].isconst = draw_vf_format_info[format].isconst; + if(vf->attr[j].isconst) + memcpy(vf->attr[j].data, &map[i].data, vf->attr[j].vertattrsize); + +#if (DRAW_VF_DBG) + debug_printf("%d: %s, offset %d\n", i, + draw_vf_format_info[format].name, + vf->attr[j].vertoffset); +#endif + + offset += draw_vf_format_info[format].attrsize; + j++; + } + } + + vf->attr_count = j; + vf->vertex_stride = vertex_stride ? vertex_stride : offset; + vf->emit = choose_emit_func; + + assert(vf->vertex_stride >= offset); + return vf->vertex_stride; +} + + +void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, + const struct vertex_info *vinfo, + float point_size ) +{ + unsigned i, j, k; + struct draw_vf_attr *a = vf->attr; + struct draw_vf_attr_map attrs[PIPE_MAX_SHADER_INPUTS]; + unsigned count = 0; /* for debug/sanity */ + unsigned nr_attrs = 0; + + for (i = 0; i < vinfo->num_attribs; i++) { + j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + /* no-op */ + break; + case EMIT_ALL: { + /* just copy the whole vertex as-is to the vbuf */ + unsigned s = vinfo->size; + assert(i == 0); + assert(j == 0); + /* copy the vertex header */ + /* XXX: we actually don't copy the header, just pad it */ + attrs[nr_attrs].attrib = 0; + attrs[nr_attrs].format = DRAW_EMIT_PAD; + attrs[nr_attrs].offset = offsetof(struct vertex_header, data); + s -= offsetof(struct vertex_header, data)/4; + count += offsetof(struct vertex_header, data)/4; + nr_attrs++; + /* copy the vertex data */ + for(k = 0; k < (s & ~0x3); k += 4) { + attrs[nr_attrs].attrib = k/4; + attrs[nr_attrs].format = DRAW_EMIT_4F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 4; + } + /* tail */ + /* XXX: actually, this shouldn't be needed */ + attrs[nr_attrs].attrib = k/4; + attrs[nr_attrs].offset = 0; + switch(s & 0x3) { + case 0: + break; + case 1: + attrs[nr_attrs].format = DRAW_EMIT_1F; + nr_attrs++; + count += 1; + break; + case 2: + attrs[nr_attrs].format = DRAW_EMIT_2F; + nr_attrs++; + count += 2; + break; + case 3: + attrs[nr_attrs].format = DRAW_EMIT_3F; + nr_attrs++; + count += 3; + break; + } + break; + } + case EMIT_1F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_1F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count++; + break; + case EMIT_1F_PSIZE: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_1F_CONST; + attrs[nr_attrs].offset = 0; + attrs[nr_attrs].data.f[0] = point_size; + nr_attrs++; + count++; + break; + case EMIT_2F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_2F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 2; + break; + case EMIT_3F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_3F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 3; + break; + case EMIT_4F: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_4F; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 4; + break; + case EMIT_4UB: + attrs[nr_attrs].attrib = j; + attrs[nr_attrs].format = DRAW_EMIT_4UB_4F_BGRA; + attrs[nr_attrs].offset = 0; + nr_attrs++; + count += 1; + break; + default: + assert(0); + } + } + + assert(count == vinfo->size); + + draw_vf_set_vertex_attributes(vf, + attrs, + nr_attrs, + vinfo->size * sizeof(float) ); + + for (j = 0; j < vf->attr_count; j++) { + a[j].inputsize = 4; + a[j].do_insert = a[j].insert[4 - 1]; + if(a[j].isconst) { + a[j].inputptr = a[j].data; + a[j].inputstride = 0; + } + } +} + + +#if 0 +/* Set attribute pointers, adjusted for start position: + */ +void draw_vf_set_sources( struct draw_vertex_fetch *vf, + GLvector4f * const sources[], + unsigned start ) +{ + struct draw_vf_attr *a = vf->attr; + unsigned j; + + for (j = 0; j < vf->attr_count; j++) { + const GLvector4f *vptr = sources[a[j].attrib]; + + if ((a[j].inputstride != vptr->stride) || + (a[j].inputsize != vptr->size)) + vf->emit = choose_emit_func; + + a[j].inputstride = vptr->stride; + a[j].inputsize = vptr->size; + a[j].do_insert = a[j].insert[vptr->size - 1]; + a[j].inputptr = ((uint8_t *)vptr->data) + start * vptr->stride; + } +} +#endif + + +/** + * Emit a vertex to dest. + */ +void draw_vf_emit_vertex( struct draw_vertex_fetch *vf, + struct vertex_header *vertex, + void *dest ) +{ + struct draw_vf_attr *a = vf->attr; + unsigned j; + + for (j = 0; j < vf->attr_count; j++) { + if (!a[j].isconst) { + a[j].inputptr = (uint8_t *)&vertex->data[a[j].attrib][0]; + a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ + } + } + + vf->emit( vf, 1, (uint8_t*) dest ); +} + + + +struct draw_vertex_fetch *draw_vf_create( void ) +{ + struct draw_vertex_fetch *vf = CALLOC_STRUCT(draw_vertex_fetch); + unsigned i; + + for (i = 0; i < PIPE_ATTRIB_MAX; i++) + vf->attr[i].vf = vf; + + vf->identity[0] = 0.0; + vf->identity[1] = 0.0; + vf->identity[2] = 0.0; + vf->identity[3] = 1.0; + + vf->codegen_emit = NULL; + +#ifdef USE_SSE_ASM + if (!GETENV("GALLIUM_NO_CODEGEN")) + vf->codegen_emit = draw_vf_generate_sse_emit; +#endif + + return vf; +} + + +void draw_vf_destroy( struct draw_vertex_fetch *vf ) +{ + struct draw_vf_fastpath *fp, *tmp; + + for (fp = vf->fastpath ; fp ; fp = tmp) { + tmp = fp->next; + FREE(fp->attr); + + /* KW: At the moment, fp->func is constrained to be allocated by + * _mesa_exec_alloc(), as the hardwired fastpaths in + * t_vertex_generic.c are handled specially. It would be nice + * to unify them, but this probably won't change until this + * module gets another overhaul. + */ + //_mesa_exec_free((void *) fp->func); + FREE(fp); + } + + vf->fastpath = NULL; + FREE(vf); +} diff --git a/src/gallium/auxiliary/draw/draw_vf.h b/src/gallium/auxiliary/draw/draw_vf.h new file mode 100644 index 0000000000..011c8f0ff1 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vf.h @@ -0,0 +1,236 @@ +/* + * Copyright 2008 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * Vertex fetch/store/convert code. This functionality is used in two places: + * 1. Vertex fetch/convert - to grab vertex data from incoming vertex + * arrays and convert to format needed by vertex shaders. + * 2. Vertex store/emit - to convert simple float[][4] vertex attributes + * (which is the organization used throughout the draw/prim pipeline) to + * hardware-specific formats and emit into hardware vertex buffers. + * + * + * Authors: + * Keith Whitwell + */ + +#ifndef DRAW_VF_H +#define DRAW_VF_H + + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" + +#include "draw_vertex.h" +#include "draw_private.h" /* for vertex_header */ + + +enum draw_vf_attr_format { + DRAW_EMIT_1F, + DRAW_EMIT_2F, + DRAW_EMIT_3F, + DRAW_EMIT_4F, + DRAW_EMIT_3F_XYW, /**< for projective texture */ + DRAW_EMIT_1UB_1F, /**< for fog coordinate */ + DRAW_EMIT_3UB_3F_RGB, /**< for specular color */ + DRAW_EMIT_3UB_3F_BGR, /**< for specular color */ + DRAW_EMIT_4UB_4F_RGBA, /**< for color */ + DRAW_EMIT_4UB_4F_BGRA, /**< for color */ + DRAW_EMIT_4UB_4F_ARGB, /**< for color */ + DRAW_EMIT_4UB_4F_ABGR, /**< for color */ + DRAW_EMIT_1F_CONST, + DRAW_EMIT_2F_CONST, + DRAW_EMIT_3F_CONST, + DRAW_EMIT_4F_CONST, + DRAW_EMIT_PAD, /**< leave a hole of 'offset' bytes */ + DRAW_EMIT_MAX +}; + +struct draw_vf_attr_map +{ + /** Input attribute number */ + unsigned attrib; + + enum draw_vf_attr_format format; + + unsigned offset; + + /** + * Constant data for DRAW_EMIT_*_CONST + */ + union { + uint8_t ub[4]; + float f[4]; + } data; +}; + +struct draw_vertex_fetch; + + + +#if 0 +unsigned +draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, + const struct draw_vf_attr_map *map, + unsigned nr, + unsigned vertex_stride ); +#endif + +void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, + const struct vertex_info *vinfo, + float point_size ); + +#if 0 +void +draw_vf_set_sources( struct draw_vertex_fetch *vf, + GLvector4f * const attrib[], + unsigned start ); +#endif + +void +draw_vf_emit_vertex( struct draw_vertex_fetch *vf, + struct vertex_header *vertex, + void *dest ); + +struct draw_vertex_fetch * +draw_vf_create( void ); + +void +draw_vf_destroy( struct draw_vertex_fetch *vf ); + + + +/*********************************************************************** + * Internal functions and structs: + */ + +struct draw_vf_attr; + +typedef void (*draw_vf_extract_func)( const struct draw_vf_attr *a, + float *out, + const uint8_t *v ); + +typedef void (*draw_vf_insert_func)( const struct draw_vf_attr *a, + uint8_t *v, + const float *in ); + +typedef void (*draw_vf_emit_func)( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *dest ); + + + +/** + * Describes how to convert/move a vertex attribute from a vertex + * array to a vertex structure. + */ +struct draw_vf_attr +{ + struct draw_vertex_fetch *vf; + + unsigned format; + unsigned inputsize; + unsigned inputstride; + unsigned vertoffset; /**< position of the attrib in the vertex struct */ + + boolean isconst; /**< read from const data below */ + uint8_t data[16]; + + unsigned attrib; /**< which vertex attrib (0=position, etc) */ + unsigned vertattrsize; /**< size of the attribute in bytes */ + + uint8_t *inputptr; + const draw_vf_insert_func *insert; + draw_vf_insert_func do_insert; + draw_vf_extract_func extract; +}; + +struct draw_vertex_fetch +{ + struct draw_vf_attr attr[PIPE_ATTRIB_MAX]; + unsigned attr_count; + unsigned vertex_stride; + + draw_vf_emit_func emit; + + /* Parameters and constants for codegen: + */ + float identity[4]; + + struct draw_vf_fastpath *fastpath; + + void (*codegen_emit)( struct draw_vertex_fetch *vf ); +}; + + +struct draw_vf_attr_type { + unsigned format; + unsigned size; + unsigned stride; + unsigned offset; +}; + +/** XXX this could be moved into draw_vf.c */ +struct draw_vf_fastpath { + unsigned vertex_stride; + unsigned attr_count; + boolean match_strides; + + struct draw_vf_attr_type *attr; + + draw_vf_emit_func func; + struct draw_vf_fastpath *next; +}; + + +void +draw_vf_register_fastpath( struct draw_vertex_fetch *vtx, + boolean match_strides ); + +void +draw_vf_generic_emit( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *v ); + +void +draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ); + +void +draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ); + + +/** XXX this type and function could probably be moved into draw_vf.c */ +struct draw_vf_format_info { + const char *name; + draw_vf_insert_func insert[4]; + const unsigned attrsize; + const boolean isconst; +}; + +extern const struct draw_vf_format_info +draw_vf_format_info[DRAW_EMIT_MAX]; + + +#endif diff --git a/src/gallium/auxiliary/draw/draw_vf_generic.c b/src/gallium/auxiliary/draw/draw_vf_generic.c new file mode 100644 index 0000000000..7a60a9db9c --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vf_generic.c @@ -0,0 +1,585 @@ + +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_util.h" + +#include "draw_vf.h" + + + +static INLINE void insert_4f_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = in[3]; +} + +static INLINE void insert_4f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = 1; +} + +static INLINE void insert_4f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; + out[3] = 1; +} + +static INLINE void insert_4f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; + out[3] = 1; +} + +static INLINE void insert_3f_xyw_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[3]; +} + +static INLINE void insert_3f_xyw_err( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + (void) a; (void) v; (void) in; + assert(0); +} + +static INLINE void insert_3f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; +} + +static INLINE void insert_3f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; + out[2] = 0; +} + +static INLINE void insert_3f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; + out[2] = 0; +} + + +static INLINE void insert_2f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = in[1]; +} + +static INLINE void insert_2f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; + out[1] = 0; +} + +static INLINE void insert_1f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + float *out = (float *)(v); + (void) a; + + out[0] = in[0]; +} + +static INLINE void insert_null( const struct draw_vf_attr *a, uint8_t *v, const float *in ) +{ + (void) a; (void) v; (void) in; +} + +static INLINE void insert_4ub_4f_rgba_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); +} + +static INLINE void insert_4ub_4f_rgba_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_rgba_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[2] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_rgba_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + v[1] = 0; + v[2] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); +} + +static INLINE void insert_4ub_4f_bgra_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[0] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_bgra_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + v[1] = 0; + v[0] = 0; + v[3] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); +} + +static INLINE void insert_4ub_4f_argb_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + v[3] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_argb_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); + v[2] = 0x00; + v[3] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_4( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); +} + +static INLINE void insert_4ub_4f_abgr_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); + v[1] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_4ub_4f_abgr_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); + v[2] = 0x00; + v[1] = 0x00; + v[0] = 0xff; +} + +static INLINE void insert_3ub_3f_rgb_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); +} + +static INLINE void insert_3ub_3f_rgb_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[2] = 0; +} + +static INLINE void insert_3ub_3f_rgb_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); + v[1] = 0; + v[2] = 0; +} + +static INLINE void insert_3ub_3f_bgr_3( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); +} + +static INLINE void insert_3ub_3f_bgr_2( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); + v[0] = 0; +} + +static INLINE void insert_3ub_3f_bgr_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); + v[1] = 0; + v[0] = 0; +} + + +static INLINE void insert_1ub_1f_1( const struct draw_vf_attr *a, uint8_t *v, + const float *in ) +{ + (void) a; + UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); +} + + +const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX] = +{ + { "1f", + { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, + sizeof(float), FALSE }, + + { "2f", + { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, + 2 * sizeof(float), FALSE }, + + { "3f", + { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, + 3 * sizeof(float), FALSE }, + + { "4f", + { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, + 4 * sizeof(float), FALSE }, + + { "3f_xyw", + { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, + insert_3f_xyw_4 }, + 3 * sizeof(float), FALSE }, + + { "1ub_1f", + { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 }, + sizeof(uint8_t), FALSE }, + + { "3ub_3f_rgb", + { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3, + insert_3ub_3f_rgb_3 }, + 3 * sizeof(uint8_t), FALSE }, + + { "3ub_3f_bgr", + { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3, + insert_3ub_3f_bgr_3 }, + 3 * sizeof(uint8_t), FALSE }, + + { "4ub_4f_rgba", + { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, + insert_4ub_4f_rgba_4 }, + 4 * sizeof(uint8_t), FALSE }, + + { "4ub_4f_bgra", + { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3, + insert_4ub_4f_bgra_4 }, + 4 * sizeof(uint8_t), FALSE }, + + { "4ub_4f_argb", + { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3, + insert_4ub_4f_argb_4 }, + 4 * sizeof(uint8_t), FALSE }, + + { "4ub_4f_abgr", + { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3, + insert_4ub_4f_abgr_4 }, + 4 * sizeof(uint8_t), FALSE }, + + { "1f_const", + { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, + sizeof(float), TRUE }, + + { "2f_const", + { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, + 2 * sizeof(float), TRUE }, + + { "3f_const", + { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, + 3 * sizeof(float), TRUE }, + + { "4f_const", + { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, + 4 * sizeof(float), TRUE }, + + { "pad", + { NULL, NULL, NULL, NULL }, + 0, FALSE }, + +}; + + + + +/*********************************************************************** + * Hardwired fastpaths for emitting whole vertices or groups of + * vertices + */ +#define EMIT5(NR, F0, F1, F2, F3, F4, NAME) \ +static void NAME( struct draw_vertex_fetch *vf, \ + unsigned count, \ + uint8_t *v ) \ +{ \ + struct draw_vf_attr *a = vf->attr; \ + unsigned i; \ + \ + for (i = 0 ; i < count ; i++, v += vf->vertex_stride) { \ + if (NR > 0) { \ + F0( &a[0], v + a[0].vertoffset, (float *)a[0].inputptr ); \ + a[0].inputptr += a[0].inputstride; \ + } \ + \ + if (NR > 1) { \ + F1( &a[1], v + a[1].vertoffset, (float *)a[1].inputptr ); \ + a[1].inputptr += a[1].inputstride; \ + } \ + \ + if (NR > 2) { \ + F2( &a[2], v + a[2].vertoffset, (float *)a[2].inputptr ); \ + a[2].inputptr += a[2].inputstride; \ + } \ + \ + if (NR > 3) { \ + F3( &a[3], v + a[3].vertoffset, (float *)a[3].inputptr ); \ + a[3].inputptr += a[3].inputstride; \ + } \ + \ + if (NR > 4) { \ + F4( &a[4], v + a[4].vertoffset, (float *)a[4].inputptr ); \ + a[4].inputptr += a[4].inputstride; \ + } \ + } \ +} + + +#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \ + insert_null, insert_null, NAME) + +#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \ + insert_null, NAME) + +#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \ + insert_null, NAME) + + +EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4) + +EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2) + +EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2) + + +/* Use the codegen paths to select one of a number of hardwired + * fastpaths. + */ +void draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ) +{ + draw_vf_emit_func func = NULL; + + /* Does it fit a hardwired fastpath? Help! this is growing out of + * control! + */ + switch (vf->attr_count) { + case 2: + if (vf->attr[0].do_insert == insert_3f_3 && + vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + func = emit_xyz3_rgba4; + } + break; + case 3: + if (vf->attr[2].do_insert == insert_2f_2) { + if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + if (vf->attr[0].do_insert == insert_4f_4) + func = emit_xyzw4_rgba4_st2; + } + } + break; + case 4: + if (vf->attr[2].do_insert == insert_2f_2 && + vf->attr[3].do_insert == insert_2f_2) { + if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { + if (vf->attr[0].do_insert == insert_4f_4) + func = emit_xyzw4_rgba4_st2_st2; + } + } + break; + } + + vf->emit = func; +} + +/*********************************************************************** + * Generic (non-codegen) functions for whole vertices or groups of + * vertices + */ + +void draw_vf_generic_emit( struct draw_vertex_fetch *vf, + unsigned count, + uint8_t *v ) +{ + struct draw_vf_attr *a = vf->attr; + const unsigned attr_count = vf->attr_count; + const unsigned stride = vf->vertex_stride; + unsigned i, j; + + for (i = 0 ; i < count ; i++, v += stride) { + for (j = 0; j < attr_count; j++) { + float *in = (float *)a[j].inputptr; + a[j].inputptr += a[j].inputstride; + a[j].do_insert( &a[j], v + a[j].vertoffset, in ); + } + } +} + + diff --git a/src/gallium/auxiliary/draw/draw_vf_sse.c b/src/gallium/auxiliary/draw/draw_vf_sse.c new file mode 100644 index 0000000000..1ad2ae756d --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vf_sse.c @@ -0,0 +1,614 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + + +#include "simple_list.h" + +#include "pipe/p_compiler.h" + +#include "draw_vf.h" + + +#if defined(USE_SSE_ASM) + +#include "x86/rtasm/x86sse.h" +#include "x86/common_x86_asm.h" + + +#define X 0 +#define Y 1 +#define Z 2 +#define W 3 + + +struct x86_program { + struct x86_function func; + + struct draw_vertex_fetch *vf; + boolean inputs_safe; + boolean outputs_safe; + boolean have_sse2; + + struct x86_reg identity; + struct x86_reg chan0; +}; + + +static struct x86_reg get_identity( struct x86_program *p ) +{ + return p->identity; +} + +static void emit_load4f_4( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movups(&p->func, dest, arg0); +} + +static void emit_load4f_3( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Have to jump through some hoops: + * + * c 0 0 0 + * c 0 0 1 + * 0 0 c 1 + * a b c 1 + */ + sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); + sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); + sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) ); + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load4f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Initialize from identity, then pull in low two words: + */ + sse_movups(&p->func, dest, get_identity(p)); + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load4f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Pull in low word, then swizzle in identity */ + sse_movss(&p->func, dest, arg0); + sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); +} + + + +static void emit_load3f_3( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + /* Over-reads by 1 dword - potential SEGV if input is a vertex + * array. + */ + if (p->inputs_safe) { + sse_movups(&p->func, dest, arg0); + } + else { + /* c 0 0 0 + * c c c c + * a b c c + */ + sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); + sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X)); + sse_movlps(&p->func, dest, arg0); + } +} + +static void emit_load3f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_2(p, dest, arg0); +} + +static void emit_load3f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_1(p, dest, arg0); +} + +static void emit_load2f_2( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movlps(&p->func, dest, arg0); +} + +static void emit_load2f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + emit_load4f_1(p, dest, arg0); +} + +static void emit_load1f_1( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movss(&p->func, dest, arg0); +} + +static void (*load[4][4])( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) = { + { emit_load1f_1, + emit_load1f_1, + emit_load1f_1, + emit_load1f_1 }, + + { emit_load2f_1, + emit_load2f_2, + emit_load2f_2, + emit_load2f_2 }, + + { emit_load3f_1, + emit_load3f_2, + emit_load3f_3, + emit_load3f_3 }, + + { emit_load4f_1, + emit_load4f_2, + emit_load4f_3, + emit_load4f_4 } +}; + +static void emit_load( struct x86_program *p, + struct x86_reg dest, + unsigned sz, + struct x86_reg src, + unsigned src_sz) +{ + load[sz-1][src_sz-1](p, dest, src); +} + +static void emit_store4f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movups(&p->func, dest, arg0); +} + +static void emit_store3f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + if (p->outputs_safe) { + /* Emit the extra dword anyway. This may hurt writecombining, + * may cause other problems. + */ + sse_movups(&p->func, dest, arg0); + } + else { + /* Alternate strategy - emit two, shuffle, emit one. + */ + sse_movlps(&p->func, dest, arg0); + sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */ + sse_movss(&p->func, x86_make_disp(dest,8), arg0); + } +} + +static void emit_store2f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movlps(&p->func, dest, arg0); +} + +static void emit_store1f( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) +{ + sse_movss(&p->func, dest, arg0); +} + + +static void (*store[4])( struct x86_program *p, + struct x86_reg dest, + struct x86_reg arg0 ) = +{ + emit_store1f, + emit_store2f, + emit_store3f, + emit_store4f +}; + +static void emit_store( struct x86_program *p, + struct x86_reg dest, + unsigned sz, + struct x86_reg temp ) + +{ + store[sz-1](p, dest, temp); +} + +static void emit_pack_store_4ub( struct x86_program *p, + struct x86_reg dest, + struct x86_reg temp ) +{ + /* Scale by 255.0 + */ + sse_mulps(&p->func, temp, p->chan0); + + if (p->have_sse2) { + sse2_cvtps2dq(&p->func, temp, temp); + sse2_packssdw(&p->func, temp, temp); + sse2_packuswb(&p->func, temp, temp); + sse_movss(&p->func, dest, temp); + } + else { + struct x86_reg mmx0 = x86_make_reg(file_MMX, 0); + struct x86_reg mmx1 = x86_make_reg(file_MMX, 1); + sse_cvtps2pi(&p->func, mmx0, temp); + sse_movhlps(&p->func, temp, temp); + sse_cvtps2pi(&p->func, mmx1, temp); + mmx_packssdw(&p->func, mmx0, mmx1); + mmx_packuswb(&p->func, mmx0, mmx0); + mmx_movd(&p->func, dest, mmx0); + } +} + +static int get_offset( const void *a, const void *b ) +{ + return (const char *)b - (const char *)a; +} + +/* Not much happens here. Eventually use this function to try and + * avoid saving/reloading the source pointers each vertex (if some of + * them can fit in registers). + */ +static void get_src_ptr( struct x86_program *p, + struct x86_reg srcREG, + struct x86_reg vfREG, + struct draw_vf_attr *a ) +{ + struct draw_vertex_fetch *vf = p->vf; + struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); + + /* Load current a[j].inputptr + */ + x86_mov(&p->func, srcREG, ptr_to_src); +} + +static void update_src_ptr( struct x86_program *p, + struct x86_reg srcREG, + struct x86_reg vfREG, + struct draw_vf_attr *a ) +{ + if (a->inputstride) { + struct draw_vertex_fetch *vf = p->vf; + struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); + + /* add a[j].inputstride (hardcoded value - could just as easily + * pull the stride value from memory each time). + */ + x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride)); + + /* save new value of a[j].inputptr + */ + x86_mov(&p->func, ptr_to_src, srcREG); + } +} + + +/* Lots of hardcoding + * + * EAX -- pointer to current output vertex + * ECX -- pointer to current attribute + * + */ +static boolean build_vertex_emit( struct x86_program *p ) +{ + struct draw_vertex_fetch *vf = p->vf; + unsigned j = 0; + + struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX); + struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX); + struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); + struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI); + struct x86_reg temp = x86_make_reg(file_XMM, 0); + uint8_t *fixup, *label; + + /* Push a few regs? + */ + x86_push(&p->func, countEBP); + x86_push(&p->func, vfESI); + + + /* Get vertex count, compare to zero + */ + x86_xor(&p->func, srcECX, srcECX); + x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2)); + x86_cmp(&p->func, countEBP, srcECX); + fixup = x86_jcc_forward(&p->func, cc_E); + + /* Initialize destination register. + */ + x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3)); + + /* Move argument 1 (vf) into a reg: + */ + x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1)); + + + /* always load, needed or not: + */ + sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0]))); + + /* Note address for loop jump */ + label = x86_get_label(&p->func); + + /* Emit code for each of the attributes. Currently routes + * everything through SSE registers, even when it might be more + * efficient to stick with regular old x86. No optimization or + * other tricks - enough new ground to cover here just getting + * things working. + */ + while (j < vf->attr_count) { + struct draw_vf_attr *a = &vf->attr[j]; + struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset); + + /* Now, load an XMM reg from src, perhaps transform, then save. + * Could be shortcircuited in specific cases: + */ + switch (a->format) { + case DRAW_EMIT_1F: + case DRAW_EMIT_1F_CONST: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 1, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_2F: + case DRAW_EMIT_2F_CONST: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 2, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_3F: + case DRAW_EMIT_3F_CONST: + /* Potentially the worst case - hardcode 2+1 copying: + */ + if (0) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + } + else { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 2, temp); + if (a->inputsize > 2) { + emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1); + emit_store(p, x86_make_disp(dest,8), 1, temp); + } + else { + sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p)); + } + update_src_ptr(p, srcECX, vfESI, a); + } + break; + case DRAW_EMIT_4F: + case DRAW_EMIT_4F_CONST: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_store(p, dest, 4, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_3F_XYW: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z)); + emit_store(p, dest, 3, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + + case DRAW_EMIT_1UB_1F: + /* Test for PAD3 + 1UB: + */ + if (j > 0 && + a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3) + { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X)); + emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */ + update_src_ptr(p, srcECX, vfESI, a); + } + else { + debug_printf("Can't emit 1ub %x %x %d\n", + a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); + return FALSE; + } + break; + case DRAW_EMIT_3UB_3F_RGB: + case DRAW_EMIT_3UB_3F_BGR: + /* Test for 3UB + PAD1: + */ + if (j == vf->attr_count - 1 || + a[1].vertoffset >= a->vertoffset + 4) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + if (a->format == DRAW_EMIT_3UB_3F_BGR) + sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + } + /* Test for 3UB + 1UB: + */ + else if (j < vf->attr_count - 1 && + a[1].format == DRAW_EMIT_1UB_1F && + a[1].vertoffset == a->vertoffset + 3) { + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); + update_src_ptr(p, srcECX, vfESI, a); + + /* Make room for incoming value: + */ + sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); + + get_src_ptr(p, srcECX, vfESI, &a[1]); + emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize); + update_src_ptr(p, srcECX, vfESI, &a[1]); + + /* Rearrange and possibly do BGR conversion: + */ + if (a->format == DRAW_EMIT_3UB_3F_BGR) + sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); + else + sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X)); + + emit_pack_store_4ub(p, dest, temp); + j++; /* NOTE: two attrs consumed */ + } + else { + debug_printf("Can't emit 3ub\n"); + } + return FALSE; /* add this later */ + break; + + case DRAW_EMIT_4UB_4F_RGBA: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_4UB_4F_BGRA: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_4UB_4F_ARGB: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + case DRAW_EMIT_4UB_4F_ABGR: + get_src_ptr(p, srcECX, vfESI, a); + emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); + sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); + emit_pack_store_4ub(p, dest, temp); + update_src_ptr(p, srcECX, vfESI, a); + break; + default: + debug_printf("unknown a[%d].format %d\n", j, a->format); + return FALSE; /* catch any new opcodes */ + } + + /* Increment j by at least 1 - may have been incremented above also: + */ + j++; + } + + /* Next vertex: + */ + x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vf->vertex_stride)); + + /* decr count, loop if not zero + */ + x86_dec(&p->func, countEBP); + x86_test(&p->func, countEBP, countEBP); + x86_jcc(&p->func, cc_NZ, label); + + /* Exit mmx state? + */ + if (p->func.need_emms) + mmx_emms(&p->func); + + /* Land forward jump here: + */ + x86_fixup_fwd_jump(&p->func, fixup); + + /* Pop regs and return + */ + x86_pop(&p->func, x86_get_base_reg(vfESI)); + x86_pop(&p->func, countEBP); + x86_ret(&p->func); + + vf->emit = (draw_vf_emit_func)x86_get_func(&p->func); + return TRUE; +} + + + +void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) +{ + struct x86_program p; + + if (!cpu_has_xmm) { + vf->codegen_emit = NULL; + return; + } + + memset(&p, 0, sizeof(p)); + + p.vf = vf; + p.inputs_safe = 0; /* for now */ + p.outputs_safe = 1; /* for now */ + p.have_sse2 = cpu_has_xmm2; + p.identity = x86_make_reg(file_XMM, 6); + p.chan0 = x86_make_reg(file_XMM, 7); + + x86_init_func(&p.func); + + if (build_vertex_emit(&p)) { + draw_vf_register_fastpath( vf, TRUE ); + } + else { + /* Note the failure so that we don't keep trying to codegen an + * impossible state: + */ + draw_vf_register_fastpath( vf, FALSE ); + x86_release_func(&p.func); + } +} + +#else + +void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) +{ + /* Dummy version for when USE_SSE_ASM not defined */ +} + +#endif diff --git a/src/gallium/auxiliary/draw/draw_wide_prims.c b/src/gallium/auxiliary/draw/draw_wide_prims.c new file mode 100644 index 0000000000..655774b155 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_wide_prims.c @@ -0,0 +1,432 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +struct wide_stage { + struct draw_stage stage; + + float half_line_width; + float half_point_size; + + uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; + uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; + uint num_texcoords; + + int psize_slot; +}; + + + +static INLINE struct wide_stage *wide_stage( struct draw_stage *stage ) +{ + return (struct wide_stage *)stage; +} + + +static void passthrough_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void passthrough_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line(stage->next, header); +} + +static void passthrough_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Draw a wide line by drawing a quad (two triangles). + * XXX need to disable polygon stipple. + */ +static void wide_line( struct draw_stage *stage, + struct prim_header *header ) +{ + const struct wide_stage *wide = wide_stage(stage); + const float half_width = wide->half_line_width; + + struct prim_header tri; + + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + const float dx = FABSF(pos0[0] - pos2[0]); + const float dy = FABSF(pos0[1] - pos2[1]); + + /* + * Draw wide line as a quad (two tris) by "stretching" the line along + * X or Y. + * We need to tweak coords in several ways to be conformant here. + */ + + if (dx > dy) { + /* x-major line */ + pos0[1] = pos0[1] - half_width - 0.25f; + pos1[1] = pos1[1] + half_width - 0.25f; + pos2[1] = pos2[1] - half_width - 0.25f; + pos3[1] = pos3[1] + half_width - 0.25f; + if (pos0[0] < pos2[0]) { + /* left to right line */ + pos0[0] -= 0.5f; + pos1[0] -= 0.5f; + pos2[0] -= 0.5f; + pos3[0] -= 0.5f; + } + else { + /* right to left line */ + pos0[0] += 0.5f; + pos1[0] += 0.5f; + pos2[0] += 0.5f; + pos3[0] += 0.5f; + } + } + else { + /* y-major line */ + pos0[0] = pos0[0] - half_width + 0.25f; + pos1[0] = pos1[0] + half_width + 0.25f; + pos2[0] = pos2[0] - half_width + 0.25f; + pos3[0] = pos3[0] + half_width + 0.25f; + if (pos0[1] < pos2[1]) { + /* top to bottom line */ + pos0[1] -= 0.5f; + pos1[1] -= 0.5f; + pos2[1] -= 0.5f; + pos3[1] -= 0.5f; + } + else { + /* bottom to top line */ + pos0[1] += 0.5f; + pos1[1] += 0.5f; + pos2[1] += 0.5f; + pos3[1] += 0.5f; + } + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +/** + * Draw a wide line by drawing a quad, using geometry which will + * fullfill GL's antialiased line requirements. + */ +static void wide_line_aa(struct draw_stage *stage, + struct prim_header *header) +{ + const struct wide_stage *wide = wide_stage(stage); + const float half_width = wide->half_line_width; + struct prim_header tri; + struct vertex_header *v[4]; + float *pos; + float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; + float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; + const float len = (float) sqrt(dx * dx + dy * dy); + uint i; + + dx = dx * half_width / len; + dy = dy * half_width / len; + + /* allocate/dup new verts */ + for (i = 0; i < 4; i++) { + v[i] = dup_vert(stage, header->v[i/2], i); + } + + /* + * Quad for line from v0 to v1: + * + * 1 3 + * +-------------------------+ + * | | + * *v0 v1* + * | | + * +-------------------------+ + * 0 2 + */ + + pos = v[0]->data[0]; + pos[0] += dy; + pos[1] -= dx; + + pos = v[1]->data[0]; + pos[0] -= dy; + pos[1] += dx; + + pos = v[2]->data[0]; + pos[0] += dy; + pos[1] -= dx; + + pos = v[3]->data[0]; + pos[0] -= dy; + pos[1] += dx; + + tri.det = header->det; /* only the sign matters */ + + tri.v[0] = v[2]; tri.v[1] = v[1]; tri.v[2] = v[0]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[3]; tri.v[1] = v[1]; tri.v[2] = v[2]; + stage->next->tri( stage->next, &tri ); + +} + + +/** + * Set the vertex texcoords for sprite mode. + * Coords may be left untouched or set to a right-side-up or upside-down + * orientation. + */ +static void set_texcoords(const struct wide_stage *wide, + struct vertex_header *v, const float tc[4]) +{ + uint i; + for (i = 0; i < wide->num_texcoords; i++) { + if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { + uint j = wide->texcoord_slot[i]; + v->data[j][0] = tc[0]; + if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) + v->data[j][1] = 1.0f - tc[1]; + else + v->data[j][1] = tc[1]; + v->data[j][2] = tc[2]; + v->data[j][3] = tc[3]; + } + } +} + + +/* If there are lots of sprite points (and why wouldn't there be?) it + * would probably be more sensible to change hardware setup to + * optimize this rather than doing the whole thing in software like + * this. + */ +static void wide_point( struct draw_stage *stage, + struct prim_header *header ) +{ + const struct wide_stage *wide = wide_stage(stage); + const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; + float half_size; + float left_adj, right_adj; + + struct prim_header tri; + + /* four dups of original vertex */ + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + /* point size is either per-vertex or fixed size */ + if (wide->psize_slot >= 0) { + half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; + } + else { + half_size = wide->half_point_size; + } + + left_adj = -half_size + 0.25f; + right_adj = half_size + 0.25f; + + pos0[0] += left_adj; + pos0[1] -= half_size; + + pos1[0] += left_adj; + pos1[1] += half_size; + + pos2[0] += right_adj; + pos2[1] -= half_size; + + pos3[0] += right_adj; + pos3[1] += half_size; + + if (sprite) { + static const float tex00[4] = { 0, 0, 0, 1 }; + static const float tex01[4] = { 0, 1, 0, 1 }; + static const float tex11[4] = { 1, 1, 0, 1 }; + static const float tex10[4] = { 1, 0, 0, 1 }; + set_texcoords( wide, v0, tex00 ); + set_texcoords( wide, v1, tex01 ); + set_texcoords( wide, v2, tex10 ); + set_texcoords( wide, v3, tex11 ); + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +static void wide_first_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct wide_stage *wide = wide_stage(stage); + struct draw_context *draw = stage->draw; + + wide->half_point_size = 0.5f * draw->rasterizer->point_size; + + if (draw->rasterizer->point_size != 1.0) { + stage->point = wide_point; + } + else { + stage->point = passthrough_point; + } + + if (draw->rasterizer->point_sprite) { + /* find vertex shader texcoord outputs */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i, j = 0; + for (i = 0; i < vs->state->num_outputs; i++) { + if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { + wide->texcoord_slot[j] = i; + wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; + j++; + } + } + wide->num_texcoords = j; + } + + wide->psize_slot = -1; + + if (draw->rasterizer->point_size_per_vertex) { + /* find PSIZ vertex output */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i; + for (i = 0; i < vs->state->num_outputs; i++) { + if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + wide->psize_slot = i; + break; + } + } + } + + stage->point( stage, header ); +} + + + +static void wide_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct wide_stage *wide = wide_stage(stage); + struct draw_context *draw = stage->draw; + + wide->half_line_width = 0.5f * draw->rasterizer->line_width; + + if (draw->rasterizer->line_width != 1.0) { + if (draw->rasterizer->line_smooth) + wide->stage.line = wide_line_aa; + else + wide->stage.line = wide_line; + } + else { + wide->stage.line = passthrough_line; + } + + stage->line( stage, header ); +} + + +static void wide_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->line = wide_first_line; + stage->point = wide_first_point; + stage->next->flush( stage->next, flags ); +} + + +static void wide_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void wide_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +struct draw_stage *draw_wide_stage( struct draw_context *draw ) +{ + struct wide_stage *wide = CALLOC_STRUCT(wide_stage); + + draw_alloc_temp_verts( &wide->stage, 4 ); + + wide->stage.draw = draw; + wide->stage.next = NULL; + wide->stage.point = wide_first_point; + wide->stage.line = wide_first_line; + wide->stage.tri = passthrough_tri; + wide->stage.flush = wide_flush; + wide->stage.reset_stipple_counter = wide_reset_stipple_counter; + wide->stage.destroy = wide_destroy; + + return &wide->stage; +} diff --git a/src/gallium/auxiliary/llvm/Makefile b/src/gallium/auxiliary/llvm/Makefile new file mode 100644 index 0000000000..e6ac399d08 --- /dev/null +++ b/src/gallium/auxiliary/llvm/Makefile @@ -0,0 +1,85 @@ +# -*-makefile-*- +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = gallivm + + +GALLIVM_SOURCES = \ + gallivm.cpp \ + gallivm_cpu.cpp \ + instructions.cpp \ + loweringpass.cpp \ + tgsitollvm.cpp \ + storage.cpp \ + storagesoa.cpp \ + instructionssoa.cpp + +INC_SOURCES = gallivm_builtins.cpp + +CPP_SOURCES = \ + $(GALLIVM_SOURCES) + +C_SOURCES = +ASM_SOURCES = + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(CPP_SOURCES:.cpp=.o) \ + $(ASM_SOURCES:.S=.o) + +### Include directories +INCLUDES = \ + -I. \ + -I$(TOP)/src/gallium/drivers + -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDES) $(LLVM_CXXFLAGS) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +##### TARGETS ##### + +default:: depend symlinks $(LIBNAME) + + +$(LIBNAME): $(OBJECTS) Makefile + $(TOP)/bin/mklib -o $@ -static $(OBJECTS) + + +depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ + $(ASM_SOURCES) $(INC_SOURCES) 2> /dev/null + + +gallivm_builtins.cpp: llvm_builtins.c + clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean: + -rm -f *.o */*.o *~ *.so *~ server/*.o + -rm -f depend depend.bak + -rm -f gallivm_builtins.cpp + +symlinks: + + +include depend diff --git a/src/gallium/auxiliary/llvm/gallivm.cpp b/src/gallium/auxiliary/llvm/gallivm.cpp new file mode 100644 index 0000000000..d14bb3b99a --- /dev/null +++ b/src/gallium/auxiliary/llvm/gallivm.cpp @@ -0,0 +1,327 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "instructions.h" +#include "loweringpass.h" +#include "storage.h" +#include "tgsitollvm.h" + +#include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_dump.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +static int GLOBAL_ID = 0; + +using namespace llvm; + +static inline +void AddStandardCompilePasses(PassManager &PM) +{ + PM.add(new LoweringPass()); + PM.add(createVerifierPass()); // Verify that input is correct + + PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp + + //PM.add(createStripSymbolsPass(true)); + + PM.add(createRaiseAllocationsPass()); // call %malloc -> malloc inst + PM.add(createCFGSimplificationPass()); // Clean up disgusting code + PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas + PM.add(createGlobalOptimizerPass()); // Optimize out global vars + PM.add(createGlobalDCEPass()); // Remove unused fns and globs + PM.add(createIPConstantPropagationPass());// IP Constant Propagation + PM.add(createDeadArgEliminationPass()); // Dead argument elimination + PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE + PM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE + + PM.add(createPruneEHPass()); // Remove dead EH info + + PM.add(createFunctionInliningPass()); // Inline small functions + PM.add(createArgumentPromotionPass()); // Scalarize uninlined fn args + + PM.add(createTailDuplicationPass()); // Simplify cfg by copying code + PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl. + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas + PM.add(createInstructionCombiningPass()); // Combine silly seq's + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createTailCallEliminationPass()); // Eliminate tail calls + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createReassociatePass()); // Reassociate expressions + PM.add(createLoopRotatePass()); + PM.add(createLICMPass()); // Hoist loop invariants + PM.add(createLoopUnswitchPass()); // Unswitch loops. + PM.add(createLoopIndexSplitPass()); // Index split loops. + PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc + PM.add(createIndVarSimplifyPass()); // Canonicalize indvars + PM.add(createLoopUnrollPass()); // Unroll small loops + PM.add(createInstructionCombiningPass()); // Clean up after the unroller + PM.add(createGVNPass()); // Remove redundancies + PM.add(createSCCPPass()); // Constant prop with SCCP + + // Run instcombine after redundancy elimination to exploit opportunities + // opened up by them. + PM.add(createInstructionCombiningPass()); + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createDeadStoreEliminationPass()); // Delete dead stores + PM.add(createAggressiveDCEPass()); // SSA based 'Aggressive DCE' + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createSimplifyLibCallsPass()); // Library Call Optimizations + PM.add(createDeadTypeEliminationPass()); // Eliminate dead types + PM.add(createConstantMergePass()); // Merge dup global constants +} + +void gallivm_prog_delete(struct gallivm_prog *prog) +{ + delete prog->module; + prog->module = 0; + prog->function = 0; + free(prog); +} + +static inline void +constant_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan) +{ + unsigned i; + + for (i = 0; i < QUAD_SIZE; ++i) { + inputs[i][attrib][chan] = coefs[attrib].a0[chan]; + } +} + +static inline void +linear_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = inputs[i][0][0]; + const float y = inputs[i][0][1]; + + inputs[i][attrib][chan] = + coefs[attrib].a0[chan] + + coefs[attrib].dadx[chan] * x + + coefs[attrib].dady[chan] * y; + } +} + +static inline void +perspective_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = inputs[i][0][0]; + const float y = inputs[i][0][1]; + /* WPOS.w here is really 1/w */ + const float w = 1.0f / inputs[i][0][3]; + assert(inputs[i][0][3] != 0.0); + + inputs[i][attrib][chan] = + (coefs[attrib].a0[chan] + + coefs[attrib].dadx[chan] * x + + coefs[attrib].dady[chan] * y) * w; + } +} + +void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix) +{ + if (!ir || !ir->module) + return; + + if (file_prefix) { + std::ostringstream stream; + stream << file_prefix; + stream << ir->id; + stream << ".ll"; + std::string name = stream.str(); + std::ofstream out(name.c_str()); + if (!out) { + std::cerr<<"Can't open file : "<module); + out.close(); + } else { + const llvm::Module::FunctionListType &funcs = ir->module->getFunctionList(); + llvm::Module::FunctionListType::const_iterator itr; + std::cout<<"; ---------- Start shader "<id<id<num_interp; ++i) { + const gallivm_interpolate &interp = prog->interpolators[i]; + switch (interp.type) { + case TGSI_INTERPOLATE_CONSTANT: + constant_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + case TGSI_INTERPOLATE_LINEAR: + linear_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + perspective_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + default: + assert( 0 ); + } + } +} + + +struct gallivm_ir * gallivm_ir_new(enum gallivm_shader_type type) +{ + struct gallivm_ir *ir = + (struct gallivm_ir *)calloc(1, sizeof(struct gallivm_ir)); + ++GLOBAL_ID; + ir->id = GLOBAL_ID; + ir->type = type; + + return ir; +} + +void gallivm_ir_set_layout(struct gallivm_ir *ir, + enum gallivm_vector_layout layout) +{ + ir->layout = layout; +} + +void gallivm_ir_set_components(struct gallivm_ir *ir, int num) +{ + ir->num_components = num; +} + +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, + const struct tgsi_token *tokens) +{ + std::cout << "Creating llvm from: " <module = mod; + gallivm_ir_dump(ir, 0); +} + +void gallivm_ir_delete(struct gallivm_ir *ir) +{ + delete ir->module; + free(ir); +} + +struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) +{ + struct gallivm_prog *prog = + (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); + llvm::Module *mod = llvm::CloneModule(ir->module); + prog->num_consts = ir->num_consts; + memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators)); + prog->num_interp = ir->num_interp; + + /* Run optimization passes over it */ + PassManager passes; + passes.add(new TargetData(mod)); + AddStandardCompilePasses(passes); + passes.run(*mod); + prog->module = mod; + + std::cout << "After optimizations:"<dump(); + + return prog; +} + +#endif /* MESA_LLVM */ diff --git a/src/gallium/auxiliary/llvm/gallivm.h b/src/gallium/auxiliary/llvm/gallivm.h new file mode 100644 index 0000000000..92da4bca7f --- /dev/null +++ b/src/gallium/auxiliary/llvm/gallivm.h @@ -0,0 +1,103 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef GALLIVM_H +#define GALLIVM_H + +#if defined __cplusplus +extern "C" { +#endif + +#include "pipe/p_state.h" + +#ifdef MESA_LLVM + +struct tgsi_token; + +struct gallivm_ir; +struct gallivm_prog; +struct gallivm_cpu_engine; +struct tgsi_interp_coef; +struct tgsi_sampler; +struct tgsi_exec_vector; + +enum gallivm_shader_type { + GALLIVM_VS, + GALLIVM_FS +}; + +enum gallivm_vector_layout { + GALLIVM_AOS, + GALLIVM_SOA +}; + +struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type); +void gallivm_ir_set_layout(struct gallivm_ir *ir, + enum gallivm_vector_layout layout); +void gallivm_ir_set_components(struct gallivm_ir *ir, int num); +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, + const struct tgsi_token *tokens); +void gallivm_ir_delete(struct gallivm_ir *ir); + + +struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir); + +void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + const struct tgsi_interp_coef *coefs); +void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); + + +struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); +struct gallivm_cpu_engine *gallivm_global_cpu_engine(); +int gallivm_cpu_vs_exec(struct gallivm_prog *prog, + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, + float (*consts)[4], + struct tgsi_exec_vector *temps); +int gallivm_cpu_fs_exec(struct gallivm_prog *prog, + float x, float y, + float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + float (*consts)[4], + struct tgsi_sampler *samplers); +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); + + +#endif /* MESA_LLVM */ + +#if defined __cplusplus +} // extern "C" +#endif + +#endif diff --git a/src/gallium/auxiliary/llvm/gallivm_builtins.cpp b/src/gallium/auxiliary/llvm/gallivm_builtins.cpp new file mode 100644 index 0000000000..1796f0a177 --- /dev/null +++ b/src/gallium/auxiliary/llvm/gallivm_builtins.cpp @@ -0,0 +1,567 @@ +// Generated by llvm2cpp - DO NOT MODIFY! + + +Module* createGallivmBuiltins(Module *mod) { + +mod->setModuleIdentifier("shader"); + +// Type Definitions +ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 25); + +PointerType* PointerTy_1 = PointerType::get(ArrayTy_0, 0); + +std::vectorFuncTy_2_args; +FuncTy_2_args.push_back(Type::FloatTy); +FuncTy_2_args.push_back(Type::FloatTy); +FunctionType* FuncTy_2 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_2_args, + /*isVarArg=*/false); + +PointerType* PointerTy_3 = PointerType::get(FuncTy_2, 0); + +VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); + +std::vectorFuncTy_5_args; +FuncTy_5_args.push_back(VectorTy_4); +FunctionType* FuncTy_5 = FunctionType::get( + /*Result=*/VectorTy_4, + /*Params=*/FuncTy_5_args, + /*isVarArg=*/false); + +std::vectorFuncTy_6_args; +FuncTy_6_args.push_back(VectorTy_4); +FuncTy_6_args.push_back(VectorTy_4); +FuncTy_6_args.push_back(VectorTy_4); +FunctionType* FuncTy_6 = FunctionType::get( + /*Result=*/VectorTy_4, + /*Params=*/FuncTy_6_args, + /*isVarArg=*/false); + +VectorType* VectorTy_7 = VectorType::get(IntegerType::get(32), 4); + +std::vectorFuncTy_9_args; +FunctionType* FuncTy_9 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_9_args, + /*isVarArg=*/true); + +PointerType* PointerTy_8 = PointerType::get(FuncTy_9, 0); + +PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8), 0); + +std::vectorFuncTy_12_args; +FuncTy_12_args.push_back(Type::FloatTy); +FunctionType* FuncTy_12 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_12_args, + /*isVarArg=*/false); + +PointerType* PointerTy_11 = PointerType::get(FuncTy_12, 0); + +std::vectorFuncTy_13_args; +FuncTy_13_args.push_back(VectorTy_4); +FunctionType* FuncTy_13 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_13_args, + /*isVarArg=*/false); + + +// Function Declarations + +Function* func_approx = new Function( + /*Type=*/FuncTy_2, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"approx", mod); +func_approx->setCallingConv(CallingConv::C); +const ParamAttrsList *func_approx_PAL = 0; +func_approx->setParamAttrs(func_approx_PAL); + +Function* func_powf = new Function( + /*Type=*/FuncTy_2, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"powf", mod); // (external, no body) +func_powf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_powf_PAL = 0; +func_powf->setParamAttrs(func_powf_PAL); + +Function* func_lit = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"lit", mod); +func_lit->setCallingConv(CallingConv::C); +const ParamAttrsList *func_lit_PAL = 0; +func_lit->setParamAttrs(func_lit_PAL); + +Function* func_cmp = new Function( + /*Type=*/FuncTy_6, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"cmp", mod); +func_cmp->setCallingConv(CallingConv::C); +const ParamAttrsList *func_cmp_PAL = 0; +{ + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_cmp_PAL = ParamAttrsList::get(Attrs); + +} +func_cmp->setParamAttrs(func_cmp_PAL); + +Function* func_vcos = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"vcos", mod); +func_vcos->setCallingConv(CallingConv::C); +const ParamAttrsList *func_vcos_PAL = 0; +func_vcos->setParamAttrs(func_vcos_PAL); + +Function* func_printf = new Function( + /*Type=*/FuncTy_9, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", mod); // (external, no body) +func_printf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_printf_PAL = 0; +func_printf->setParamAttrs(func_printf_PAL); + +Function* func_cosf = new Function( + /*Type=*/FuncTy_12, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"cosf", mod); // (external, no body) +func_cosf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_cosf_PAL = 0; +func_cosf->setParamAttrs(func_cosf_PAL); + +Function* func_scs = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"scs", mod); +func_scs->setCallingConv(CallingConv::C); +const ParamAttrsList *func_scs_PAL = 0; +func_scs->setParamAttrs(func_scs_PAL); + +Function* func_sinf = new Function( + /*Type=*/FuncTy_12, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"sinf", mod); // (external, no body) +func_sinf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_sinf_PAL = 0; +func_sinf->setParamAttrs(func_sinf_PAL); + +Function* func_vsin = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"vsin", mod); +func_vsin->setCallingConv(CallingConv::C); +const ParamAttrsList *func_vsin_PAL = 0; +func_vsin->setParamAttrs(func_vsin_PAL); + +Function* func_kilp = new Function( + /*Type=*/FuncTy_13, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"kilp", mod); +func_kilp->setCallingConv(CallingConv::C); +const ParamAttrsList *func_kilp_PAL = 0; +{ + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_kilp_PAL = ParamAttrsList::get(Attrs); + +} +func_kilp->setParamAttrs(func_kilp_PAL); + +// Global Variable Declarations + + +GlobalVariable* gvar_array__str = new GlobalVariable( +/*Type=*/ArrayTy_0, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str", +mod); + +GlobalVariable* gvar_array__str1 = new GlobalVariable( +/*Type=*/ArrayTy_0, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str1", +mod); + +// Constant Definitions +Constant* const_array_14 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); +Constant* const_array_15 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); +ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); +ConstantFP* const_float_17 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); +Constant* const_float_18 = Constant::getNullValue(Type::FloatTy); +Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); +std::vector const_packed_20_elems; +ConstantFP* const_float_21 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); +const_packed_20_elems.push_back(const_float_21); +UndefValue* const_float_22 = UndefValue::get(Type::FloatTy); +const_packed_20_elems.push_back(const_float_22); +const_packed_20_elems.push_back(const_float_22); +const_packed_20_elems.push_back(const_float_21); +Constant* const_packed_20 = ConstantVector::get(VectorTy_4, const_packed_20_elems); +ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); +ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "3", 10)); +ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "2", 10)); +std::vector const_packed_26_elems; +const_packed_26_elems.push_back(const_float_21); +const_packed_26_elems.push_back(const_float_18); +const_packed_26_elems.push_back(const_float_18); +const_packed_26_elems.push_back(const_float_21); +Constant* const_packed_26 = ConstantVector::get(VectorTy_4, const_packed_26_elems); +Constant* const_double_27 = Constant::getNullValue(Type::DoubleTy); +std::vector const_packed_28_elems; +const_packed_28_elems.push_back(const_int32_19); +ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "5", 10)); +const_packed_28_elems.push_back(const_int32_29); +const_packed_28_elems.push_back(const_int32_25); +const_packed_28_elems.push_back(const_int32_24); +Constant* const_packed_28 = ConstantVector::get(VectorTy_7, const_packed_28_elems); +std::vector const_packed_30_elems; +const_packed_30_elems.push_back(const_int32_19); +const_packed_30_elems.push_back(const_int32_23); +ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "6", 10)); +const_packed_30_elems.push_back(const_int32_31); +const_packed_30_elems.push_back(const_int32_24); +Constant* const_packed_30 = ConstantVector::get(VectorTy_7, const_packed_30_elems); +std::vector const_packed_32_elems; +const_packed_32_elems.push_back(const_int32_19); +const_packed_32_elems.push_back(const_int32_23); +const_packed_32_elems.push_back(const_int32_25); +ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "7", 10)); +const_packed_32_elems.push_back(const_int32_33); +Constant* const_packed_32 = ConstantVector::get(VectorTy_7, const_packed_32_elems); +std::vector const_ptr_34_indices; +const_ptr_34_indices.push_back(const_int32_19); +const_ptr_34_indices.push_back(const_int32_19); +Constant* const_ptr_34 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_34_indices[0], const_ptr_34_indices.size() ); +UndefValue* const_packed_35 = UndefValue::get(VectorTy_4); +std::vector const_ptr_36_indices; +const_ptr_36_indices.push_back(const_int32_19); +const_ptr_36_indices.push_back(const_int32_19); +Constant* const_ptr_36 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_36_indices[0], const_ptr_36_indices.size() ); + +// Global Variable Definitions +gvar_array__str->setInitializer(const_array_14); +gvar_array__str1->setInitializer(const_array_15); + +// Function Definitions + +// Function: approx (func_approx) +{ + Function::arg_iterator args = func_approx->arg_begin(); + Value* float_a = args++; + float_a->setName("a"); + Value* float_b = args++; + float_b->setName("b"); + + BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); + + // Block entry (label_entry) + FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_16, "cmp", label_entry); + SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_16, float_b, "b.addr.0", label_entry); + FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_17, "cmp3", label_entry); + SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_17, float_b_addr_0, "b.addr.1", label_entry); + FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_18, "cmp7", label_entry); + SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_18, float_a, "a.addr.0", label_entry); + std::vector float_call_params; + float_call_params.push_back(float_a_addr_0); + float_call_params.push_back(float_b_addr_1); + CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); + float_call->setCallingConv(CallingConv::C); + float_call->setTailCall(true);const ParamAttrsList *float_call_PAL = 0; + float_call->setParamAttrs(float_call_PAL); + + new ReturnInst(float_call, label_entry); + +} + +// Function: lit (func_lit) +{ + Function::arg_iterator args = func_lit->arg_begin(); + Value* packed_tmp = args++; + packed_tmp->setName("tmp"); + + BasicBlock* label_entry_38 = new BasicBlock("entry",func_lit,0); + BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); + BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); + + // Block entry (label_entry_38) + ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp6", label_entry_38); + FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp6, const_float_18, "cmp", label_entry_38); + new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_39, label_entry_38); + + // Block ifthen (label_ifthen) + InsertElementInst* packed_tmp10 = new InsertElementInst(const_packed_20, float_tmp6, const_int32_23, "tmp10", label_ifthen); + ExtractElementInst* float_tmp12 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp14", label_ifthen); + std::vector float_call_41_params; + float_call_41_params.push_back(float_tmp12); + float_call_41_params.push_back(float_tmp14); + CallInst* float_call_41 = new CallInst(func_approx, float_call_41_params.begin(), float_call_41_params.end(), "call", label_ifthen); + float_call_41->setCallingConv(CallingConv::C); + float_call_41->setTailCall(true);const ParamAttrsList *float_call_41_PAL = 0; + float_call_41->setParamAttrs(float_call_41_PAL); + + InsertElementInst* packed_tmp16 = new InsertElementInst(packed_tmp10, float_call_41, const_int32_25, "tmp16", label_ifthen); + new ReturnInst(packed_tmp16, label_ifthen); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock) + new ReturnInst(const_packed_26, label_UnifiedReturnBlock); + +} + +// Function: cmp (func_cmp) +{ + Function::arg_iterator args = func_cmp->arg_begin(); + Value* packed_tmp0 = args++; + packed_tmp0->setName("tmp0"); + Value* packed_tmp1 = args++; + packed_tmp1->setName("tmp1"); + Value* packed_tmp2 = args++; + packed_tmp2->setName("tmp2"); + + BasicBlock* label_entry_44 = new BasicBlock("entry",func_cmp,0); + BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); + BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); + BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); + BasicBlock* label_cond_cont34 = new BasicBlock("cond.cont34",func_cmp,0); + BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); + BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); + + // Block entry (label_entry_44) + ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_44); + CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_44); + FCmpInst* int1_cmp_45 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_44); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_44); + CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_44); + FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_44); + SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_45, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_44); + new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_44); + + // Block cond.?14 (label_cond__14) + ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_28, "tmp233", label_cond__14); + ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp254", label_cond__14); + CastInst* double_conv265 = new FPExtInst(float_tmp254, Type::DoubleTy, "conv265", label_cond__14); + FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_27, "cmp276", label_cond__14); + new BranchInst(label_cond__28, label_cond_cont34, int1_cmp276, label_cond__14); + + // Block cond.cont20 (label_cond_cont20) + ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_28, "tmp23", label_cond_cont20); + ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp25", label_cond_cont20); + CastInst* double_conv26 = new FPExtInst(float_tmp25, Type::DoubleTy, "conv26", label_cond_cont20); + FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_27, "cmp27", label_cond_cont20); + new BranchInst(label_cond__28, label_cond_cont34, int1_cmp27, label_cond_cont20); + + // Block cond.?28 (label_cond__28) + PHINode* packed_tmp23_reg2mem_0 = new PHINode(VectorTy_4, "tmp23.reg2mem.0", label_cond__28); + packed_tmp23_reg2mem_0->reserveOperandSpace(2); + packed_tmp23_reg2mem_0->addIncoming(packed_tmp233, label_cond__14); + packed_tmp23_reg2mem_0->addIncoming(packed_tmp23, label_cond_cont20); + + ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_30, "tmp378", label_cond__28); + ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp399", label_cond__28); + CastInst* double_conv4010 = new FPExtInst(float_tmp399, Type::DoubleTy, "conv4010", label_cond__28); + FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_27, "cmp4111", label_cond__28); + new BranchInst(label_cond__42, label_cond_cont48, int1_cmp4111, label_cond__28); + + // Block cond.cont34 (label_cond_cont34) + PHINode* packed_tmp23_reg2mem_1 = new PHINode(VectorTy_4, "tmp23.reg2mem.1", label_cond_cont34); + packed_tmp23_reg2mem_1->reserveOperandSpace(2); + packed_tmp23_reg2mem_1->addIncoming(packed_tmp233, label_cond__14); + packed_tmp23_reg2mem_1->addIncoming(packed_tmp23, label_cond_cont20); + + ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_30, "tmp37", label_cond_cont34); + ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp39", label_cond_cont34); + CastInst* double_conv40 = new FPExtInst(float_tmp39, Type::DoubleTy, "conv40", label_cond_cont34); + FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_27, "cmp41", label_cond_cont34); + new BranchInst(label_cond__42, label_cond_cont48, int1_cmp41, label_cond_cont34); + + // Block cond.?42 (label_cond__42) + PHINode* packed_tmp37_reg2mem_0 = new PHINode(VectorTy_4, "tmp37.reg2mem.0", label_cond__42); + packed_tmp37_reg2mem_0->reserveOperandSpace(2); + packed_tmp37_reg2mem_0->addIncoming(packed_tmp378, label_cond__28); + packed_tmp37_reg2mem_0->addIncoming(packed_tmp37, label_cond_cont34); + + ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_32, "tmp5113", label_cond__42); + new ReturnInst(packed_tmp5113, label_cond__42); + + // Block cond.cont48 (label_cond_cont48) + PHINode* packed_tmp37_reg2mem_1 = new PHINode(VectorTy_4, "tmp37.reg2mem.1", label_cond_cont48); + packed_tmp37_reg2mem_1->reserveOperandSpace(2); + packed_tmp37_reg2mem_1->addIncoming(packed_tmp378, label_cond__28); + packed_tmp37_reg2mem_1->addIncoming(packed_tmp37, label_cond_cont34); + + ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_32, "tmp51", label_cond_cont48); + new ReturnInst(packed_tmp51, label_cond_cont48); + +} + +// Function: vcos (func_vcos) +{ + Function::arg_iterator args = func_vcos->arg_begin(); + Value* packed_val = args++; + packed_val->setName("val"); + + BasicBlock* label_entry_53 = new BasicBlock("entry",func_vcos,0); + + // Block entry (label_entry_53) + ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_53); + CastInst* double_conv_54 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_53); + ExtractElementInst* float_tmp3_55 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_53); + CastInst* double_conv4 = new FPExtInst(float_tmp3_55, Type::DoubleTy, "conv4", label_entry_53); + ExtractElementInst* float_tmp6_56 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_53); + CastInst* double_conv7 = new FPExtInst(float_tmp6_56, Type::DoubleTy, "conv7", label_entry_53); + ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_53); + CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_53); + std::vector int32_call_params; + int32_call_params.push_back(const_ptr_34); + int32_call_params.push_back(double_conv_54); + int32_call_params.push_back(double_conv4); + int32_call_params.push_back(double_conv7); + int32_call_params.push_back(double_conv10); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_53); + int32_call->setCallingConv(CallingConv::C); + int32_call->setTailCall(true);const ParamAttrsList *int32_call_PAL = 0; + int32_call->setParamAttrs(int32_call_PAL); + + CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_53); + float_call13->setCallingConv(CallingConv::C); + float_call13->setTailCall(true);const ParamAttrsList *float_call13_PAL = 0; + float_call13->setParamAttrs(float_call13_PAL); + + InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_53); + CallInst* float_call18 = new CallInst(func_cosf, float_tmp1, "call18", label_entry_53); + float_call18->setCallingConv(CallingConv::C); + float_call18->setTailCall(true);const ParamAttrsList *float_call18_PAL = 0; + float_call18->setParamAttrs(float_call18_PAL); + + InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call18, const_int32_23, "tmp20", label_entry_53); + CallInst* float_call23 = new CallInst(func_cosf, float_tmp1, "call23", label_entry_53); + float_call23->setCallingConv(CallingConv::C); + float_call23->setTailCall(true);const ParamAttrsList *float_call23_PAL = 0; + float_call23->setParamAttrs(float_call23_PAL); + + InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call23, const_int32_25, "tmp25", label_entry_53); + CallInst* float_call28 = new CallInst(func_cosf, float_tmp1, "call28", label_entry_53); + float_call28->setCallingConv(CallingConv::C); + float_call28->setTailCall(true);const ParamAttrsList *float_call28_PAL = 0; + float_call28->setParamAttrs(float_call28_PAL); + + InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call28, const_int32_24, "tmp30", label_entry_53); + CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_53); + CastInst* double_conv36 = new FPExtInst(float_call18, Type::DoubleTy, "conv36", label_entry_53); + CastInst* double_conv39 = new FPExtInst(float_call23, Type::DoubleTy, "conv39", label_entry_53); + CastInst* double_conv42 = new FPExtInst(float_call28, Type::DoubleTy, "conv42", label_entry_53); + std::vector int32_call43_params; + int32_call43_params.push_back(const_ptr_36); + int32_call43_params.push_back(double_conv33); + int32_call43_params.push_back(double_conv36); + int32_call43_params.push_back(double_conv39); + int32_call43_params.push_back(double_conv42); + CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_53); + int32_call43->setCallingConv(CallingConv::C); + int32_call43->setTailCall(true);const ParamAttrsList *int32_call43_PAL = 0; + int32_call43->setParamAttrs(int32_call43_PAL); + + new ReturnInst(packed_tmp30, label_entry_53); + +} + +// Function: scs (func_scs) +{ + Function::arg_iterator args = func_scs->arg_begin(); + Value* packed_val_58 = args++; + packed_val_58->setName("val"); + + BasicBlock* label_entry_59 = new BasicBlock("entry",func_scs,0); + + // Block entry (label_entry_59) + ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_58, const_int32_19, "tmp2", label_entry_59); + CallInst* float_call_60 = new CallInst(func_cosf, float_tmp2, "call", label_entry_59); + float_call_60->setCallingConv(CallingConv::C); + float_call_60->setTailCall(true);const ParamAttrsList *float_call_60_PAL = 0; + float_call_60->setParamAttrs(float_call_60_PAL); + + InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_60, const_int32_19, "tmp5", label_entry_59); + CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_59); + float_call7->setCallingConv(CallingConv::C); + float_call7->setTailCall(true);const ParamAttrsList *float_call7_PAL = 0; + float_call7->setParamAttrs(float_call7_PAL); + + InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_59); + new ReturnInst(packed_tmp9, label_entry_59); + +} + +// Function: vsin (func_vsin) +{ + Function::arg_iterator args = func_vsin->arg_begin(); + Value* packed_val_62 = args++; + packed_val_62->setName("val"); + + BasicBlock* label_entry_63 = new BasicBlock("entry",func_vsin,0); + + // Block entry (label_entry_63) + ExtractElementInst* float_tmp2_64 = new ExtractElementInst(packed_val_62, const_int32_19, "tmp2", label_entry_63); + CallInst* float_call_65 = new CallInst(func_sinf, float_tmp2_64, "call", label_entry_63); + float_call_65->setCallingConv(CallingConv::C); + float_call_65->setTailCall(true);const ParamAttrsList *float_call_65_PAL = 0; + float_call_65->setParamAttrs(float_call_65_PAL); + + InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_65, const_int32_19, "tmp6", label_entry_63); + InsertElementInst* packed_tmp9_66 = new InsertElementInst(packed_tmp6, float_call_65, const_int32_23, "tmp9", label_entry_63); + InsertElementInst* packed_tmp12 = new InsertElementInst(packed_tmp9_66, float_call_65, const_int32_25, "tmp12", label_entry_63); + InsertElementInst* packed_tmp15_67 = new InsertElementInst(packed_tmp12, float_call_65, const_int32_24, "tmp15", label_entry_63); + new ReturnInst(packed_tmp15_67, label_entry_63); + +} + +// Function: kilp (func_kilp) +{ + Function::arg_iterator args = func_kilp->arg_begin(); + Value* packed_val_69 = args++; + packed_val_69->setName("val"); + + BasicBlock* label_entry_70 = new BasicBlock("entry",func_kilp,0); + BasicBlock* label_lor_rhs = new BasicBlock("lor_rhs",func_kilp,0); + BasicBlock* label_lor_rhs5 = new BasicBlock("lor_rhs5",func_kilp,0); + BasicBlock* label_lor_rhs11 = new BasicBlock("lor_rhs11",func_kilp,0); + BasicBlock* label_UnifiedReturnBlock_71 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); + + // Block entry (label_entry_70) + ExtractElementInst* float_tmp1_72 = new ExtractElementInst(packed_val_69, const_int32_19, "tmp1", label_entry_70); + FCmpInst* int1_cmp_73 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_72, const_float_18, "cmp", label_entry_70); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs, int1_cmp_73, label_entry_70); + + // Block lor_rhs (label_lor_rhs) + ExtractElementInst* float_tmp3_75 = new ExtractElementInst(packed_val_69, const_int32_23, "tmp3", label_lor_rhs); + FCmpInst* int1_cmp4 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_75, const_float_18, "cmp4", label_lor_rhs); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs5, int1_cmp4, label_lor_rhs); + + // Block lor_rhs5 (label_lor_rhs5) + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_val_69, const_int32_25, "tmp7", label_lor_rhs5); + FCmpInst* int1_cmp8 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp7, const_float_18, "cmp8", label_lor_rhs5); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs11, int1_cmp8, label_lor_rhs5); + + // Block lor_rhs11 (label_lor_rhs11) + ExtractElementInst* float_tmp13 = new ExtractElementInst(packed_val_69, const_int32_24, "tmp13", label_lor_rhs11); + FCmpInst* int1_cmp14 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp13, const_float_18, "cmp14", label_lor_rhs11); + CastInst* int32_retval = new ZExtInst(int1_cmp14, IntegerType::get(32), "retval", label_lor_rhs11); + new ReturnInst(int32_retval, label_lor_rhs11); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock_71) + new ReturnInst(const_int32_23, label_UnifiedReturnBlock_71); + +} + +return mod; + +} diff --git a/src/gallium/auxiliary/llvm/gallivm_cpu.cpp b/src/gallium/auxiliary/llvm/gallivm_cpu.cpp new file mode 100644 index 0000000000..8f9830d0b1 --- /dev/null +++ b/src/gallium/auxiliary/llvm/gallivm_cpu.cpp @@ -0,0 +1,202 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "instructions.h" +#include "loweringpass.h" +#include "storage.h" +#include "tgsitollvm.h" + +#include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_dump.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +struct gallivm_cpu_engine { + llvm::ExecutionEngine *engine; +}; + +static struct gallivm_cpu_engine *CPU = 0; + +typedef int (*fragment_shader_runner)(float x, float y, + float (*dests)[16][4], + float (*inputs)[16][4], + int num_attribs, + float (*consts)[4], int num_consts, + struct tgsi_sampler *samplers); + +int gallivm_cpu_fs_exec(struct gallivm_prog *prog, + float fx, float fy, + float (*dests)[16][4], + float (*inputs)[16][4], + float (*consts)[4], + struct tgsi_sampler *samplers) +{ + fragment_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + + return runner(fx, fy, dests, inputs, prog->num_interp, + consts, prog->num_consts, + samplers); +} + +static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) +{ + llvm::Module *mod = prog->module; + llvm::Function *func = 0; + + switch (prog->type) { + case GALLIVM_VS: + func = mod->getFunction("vs_shader"); + break; + case GALLIVM_FS: + func = mod->getFunction("fs_shader"); + break; + default: + assert(!"Unknown shader type!"); + break; + } + return func; +} + +/*! + This function creates a CPU based execution engine for the given gallivm_prog. + gallivm_cpu_engine should be used as a singleton throughout the library. Before + executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile. + The gallivm_prog instance which is being passed to the constructor is being + automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile + with it again. + */ +struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) +{ + struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *) + calloc(1, sizeof(struct gallivm_cpu_engine)); + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); + ee->DisableLazyCompilation(); + cpu->engine = ee; + + llvm::Function *func = func_for_shader(prog); + + prog->function = ee->getPointerToFunction(func); + CPU = cpu; + return cpu; +} + + +/*! + This function JIT compiles the given gallivm_prog with the given cpu based execution engine. + The reference to the generated machine code entry point will be stored + in the gallivm_prog program. After executing this function one can call gallivm_prog_exec + in order to execute the gallivm_prog on the CPU. + */ +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) +{ + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = cpu->engine; + assert(ee); + /*FIXME : remove */ + ee->DisableLazyCompilation(); + ee->addModuleProvider(mp); + + llvm::Function *func = func_for_shader(prog); + prog->function = ee->getPointerToFunction(func); +} + +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) +{ + free(cpu); +} + +struct gallivm_cpu_engine * gallivm_global_cpu_engine() +{ + return CPU; +} + + +typedef void (*vertex_shader_runner)(void *ainputs, + void *dests, + float (*aconsts)[4], + void *temps); + + +/*! + This function is used to execute the gallivm_prog in software. Before calling + this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile + function. + */ +int gallivm_cpu_vs_exec(struct gallivm_prog *prog, + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, + float (*consts)[4], + struct tgsi_exec_vector *temps) +{ + vertex_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + /*FIXME*/ + runner(inputs, dests, consts, temps); + + return 0; +} + +#endif diff --git a/src/gallium/auxiliary/llvm/gallivm_p.h b/src/gallium/auxiliary/llvm/gallivm_p.h new file mode 100644 index 0000000000..cfe7b1901b --- /dev/null +++ b/src/gallium/auxiliary/llvm/gallivm_p.h @@ -0,0 +1,110 @@ +#ifndef GALLIVM_P_H +#define GALLIVM_P_H + +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/p_compiler.h" + +namespace llvm { + class Module; +} + +#if defined __cplusplus +extern "C" { +#endif + +enum gallivm_shader_type; +enum gallivm_vector_layout; + +struct gallivm_interpolate { + int attrib; + int chan; + int type; +}; + +struct gallivm_ir { + llvm::Module *module; + int id; + enum gallivm_shader_type type; + enum gallivm_vector_layout layout; + int num_components; + int num_consts; + + //FIXME: this might not be enough for some shaders + struct gallivm_interpolate interpolators[32*4]; + int num_interp; +}; + +struct gallivm_prog { + llvm::Module *module; + void *function; + + int id; + enum gallivm_shader_type type; + + int num_consts; + + //FIXME: this might not be enough for some shaders + struct gallivm_interpolate interpolators[32*4]; + int num_interp; +}; + +static INLINE void gallivm_swizzle_components(int swizzle, + int *xc, int *yc, + int *zc, int *wc) +{ + int x = swizzle / 1000; swizzle -= x * 1000; + int y = swizzle / 100; swizzle -= y * 100; + int z = swizzle / 10; swizzle -= z * 10; + int w = swizzle; + + if (xc) *xc = x; + if (yc) *yc = y; + if (zc) *zc = z; + if (wc) *wc = w; +} + +static INLINE boolean gallivm_is_swizzle(int swizzle) +{ + const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + + TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; + return swizzle != NO_SWIZZLE; +} + +static INLINE int gallivm_x_swizzle(int swizzle) +{ + int x; + gallivm_swizzle_components(swizzle, &x, 0, 0, 0); + return x; +} + +static INLINE int gallivm_y_swizzle(int swizzle) +{ + int y; + gallivm_swizzle_components(swizzle, 0, &y, 0, 0); + return y; +} + +static INLINE int gallivm_z_swizzle(int swizzle) +{ + int z; + gallivm_swizzle_components(swizzle, 0, 0, &z, 0); + return z; +} + +static INLINE int gallivm_w_swizzle(int swizzle) +{ + int w; + gallivm_swizzle_components(swizzle, 0, 0, 0, &w); + return w; +} + +#endif /* MESA_LLVM */ + +#if defined __cplusplus +} // extern "C" +#endif + +#endif diff --git a/src/gallium/auxiliary/llvm/instructions.cpp b/src/gallium/auxiliary/llvm/instructions.cpp new file mode 100644 index 0000000000..55d39fa5f1 --- /dev/null +++ b/src/gallium/auxiliary/llvm/instructions.cpp @@ -0,0 +1,889 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "instructions.h" + +#include "storage.h" + +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +using namespace llvm; + +#include "gallivm_builtins.cpp" + +static inline std::string createFuncName(int label) +{ + std::ostringstream stream; + stream << "function"; + stream << label; + return stream.str(); +} + +Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, + Storage *storage) + : m_mod(mod), m_func(func), m_builder(block), m_idx(0), + m_storage(storage) +{ + m_floatVecType = VectorType::get(Type::FloatTy, 4); + + m_llvmFSqrt = 0; + m_llvmFAbs = 0; + m_llvmPow = 0; + m_llvmFloor = 0; + m_llvmFlog = 0; + m_llvmLit = 0; + m_fmtPtr = 0; + + createGallivmBuiltins(m_mod); +} + +llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) +{ + return m_builder.CreateAdd(in1, in2, name("add")); +} + +llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3) +{ + Value *mulRes = mul(in1, in2); + return add(mulRes, in3); +} + +llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2) +{ + return m_builder.CreateMul(in1, in2, name("mul")); +} + +const char * Instructions::name(const char *prefix) +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + Value *x = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(0), + name("extractx")); + Value *y = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(1), + name("extracty")); + Value *z = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(2), + name("extractz")); + Value *xy = m_builder.CreateAdd(x, y,name("xy")); + Value *dot3 = m_builder.CreateAdd(xy, z, name("dot3")); + return vectorFromVals(dot3, dot3, dot3, dot3); +} + +llvm::Value *Instructions::callFSqrt(llvm::Value *val) +{ + if (!m_llvmFSqrt) { + // predeclare the intrinsic + std::vector fsqrtArgs; + fsqrtArgs.push_back(Type::FloatTy); + ParamAttrsList *fsqrtPal = 0; + FunctionType* fsqrtType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/fsqrtArgs, + /*isVarArg=*/false); + m_llvmFSqrt = new Function( + /*Type=*/fsqrtType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"llvm.sqrt.f32", m_mod); + m_llvmFSqrt->setCallingConv(CallingConv::C); + m_llvmFSqrt->setParamAttrs(fsqrtPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFSqrt, val, + name("sqrt")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::rsq(llvm::Value *in1) +{ + Value *x = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("extractx")); + Value *abs = callFAbs(x); + Value *sqrt = callFSqrt(abs); + + Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + sqrt, + name("rsqrt")); + return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt); +} + +llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w) +{ + Constant *const_vec = Constant::getNullValue(m_floatVecType); + Value *res = m_builder.CreateInsertElement(const_vec, x, + m_storage->constantInt(0), + name("vecx")); + res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), + name("vecxy")); + res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), + name("vecxyz")); + if (w) + res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), + name("vecxyzw")); + return res; +} + +llvm::Value *Instructions::callFAbs(llvm::Value *val) +{ + if (!m_llvmFAbs) { + // predeclare the intrinsic + std::vector fabsArgs; + fabsArgs.push_back(Type::FloatTy); + ParamAttrsList *fabsPal = 0; + FunctionType* fabsType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/fabsArgs, + /*isVarArg=*/false); + m_llvmFAbs = new Function( + /*Type=*/fabsType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"fabs", m_mod); + m_llvmFAbs->setCallingConv(CallingConv::C); + m_llvmFAbs->setParamAttrs(fabsPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFAbs, val, + name("fabs")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::lit(llvm::Value *in) +{ + if (!m_llvmLit) { + m_llvmLit = m_mod->getFunction("lit"); + } + CallInst *call = m_builder.CreateCall(m_llvmLit, in, name("litres")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) +{ + Value *res = m_builder.CreateSub(in1, in2, name("sub")); + return res; +} + +llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) +{ + if (!m_llvmPow) { + // predeclare the intrinsic + std::vector powArgs; + powArgs.push_back(Type::FloatTy); + powArgs.push_back(Type::FloatTy); + ParamAttrsList *powPal = 0; + FunctionType* powType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/powArgs, + /*isVarArg=*/false); + m_llvmPow = new Function( + /*Type=*/powType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"llvm.pow.f32", m_mod); + m_llvmPow->setCallingConv(CallingConv::C); + m_llvmPow->setParamAttrs(powPal); + } + std::vector params; + params.push_back(val1); + params.push_back(val2); + CallInst *call = m_builder.CreateCall(m_llvmPow, params.begin(), params.end(), + name("pow")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2) +{ + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); + llvm::Value *val = callPow(x1, x2); + return vectorFromVals(val, val, val, val); +} + +llvm::Value * Instructions::rcp(llvm::Value *in1) +{ + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *res = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + x1, name("rcp")); + return vectorFromVals(res, res, res, res); +} + +llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + std::vector vec = extractVector(mulRes); + Value *xy = m_builder.CreateAdd(vec[0], vec[1], name("xy")); + Value *xyz = m_builder.CreateAdd(xy, vec[2], name("xyz")); + Value *dot4 = m_builder.CreateAdd(xyz, vec[3], name("dot4")); + return vectorFromVals(dot4, dot4, dot4, dot4); +} + +llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + std::vector vec1 = extractVector(mulRes); + Value *xy = m_builder.CreateAdd(vec1[0], vec1[1], name("xy")); + Value *xyz = m_builder.CreateAdd(xy, vec1[2], name("xyz")); + Value *dph = m_builder.CreateAdd(xyz, vec1[3], name("dph")); + return vectorFromVals(dph, dph, dph, dph); +} + +llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) +{ + Value *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + Value *z = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z")); + Value *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + Value *w = m_builder.CreateExtractElement(in2, + m_storage->constantInt(3), + name("w")); + Value *ry = m_builder.CreateMul(y1, y2, name("tyuy")); + return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), + ry, z, w); +} + +llvm::Value * Instructions::ex2(llvm::Value *in) +{ + llvm::Value *val = callPow(ConstantFP::get(Type::FloatTy, APFloat(2.f)), + m_builder.CreateExtractElement( + in, m_storage->constantInt(0), + name("x1"))); + return vectorFromVals(val, val, val, val); +} + +llvm::Value * Instructions::callFloor(llvm::Value *val) +{ + if (!m_llvmFloor) { + // predeclare the intrinsic + std::vector floorArgs; + floorArgs.push_back(Type::FloatTy); + ParamAttrsList *floorPal = 0; + FunctionType* floorType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/floorArgs, + /*isVarArg=*/false); + m_llvmFloor = new Function( + /*Type=*/floorType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"floorf", m_mod); + m_llvmFloor->setCallingConv(CallingConv::C); + m_llvmFloor->setParamAttrs(floorPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFloor, val, + name("floorf")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::floor(llvm::Value *in) +{ + std::vector vec = extractVector(in); + return vectorFromVals(callFloor(vec[0]), callFloor(vec[1]), + callFloor(vec[2]), callFloor(vec[3])); +} + +llvm::Value * Instructions::arl(llvm::Value *in) +{ + return floor(in); +} + +llvm::Value * Instructions::frc(llvm::Value *in) +{ + llvm::Value *flr = floor(in); + return sub(in, flr); +} + +llvm::Value * Instructions::callFLog(llvm::Value *val) +{ + if (!m_llvmFlog) { + // predeclare the intrinsic + std::vector flogArgs; + flogArgs.push_back(Type::FloatTy); + ParamAttrsList *flogPal = 0; + FunctionType* flogType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/flogArgs, + /*isVarArg=*/false); + m_llvmFlog = new Function( + /*Type=*/flogType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"logf", m_mod); + m_llvmFlog->setCallingConv(CallingConv::C); + m_llvmFlog->setParamAttrs(flogPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFlog, val, + name("logf")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::lg2(llvm::Value *in) +{ + std::vector vec = extractVector(in); + llvm::Value *const_vec = constVector(1.442695f, 1.442695f, + 1.442695f, 1.442695f); + return mul(vectorFromVals(callFLog(vec[0]), callFLog(vec[1]), + callFLog(vec[2]), callFLog(vec[3])), const_vec); +} + +llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) +{ + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); + Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); + + Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); + Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); + + Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); + Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); + + Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); + Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); + + return vectorFromVals(selx, sely, selz, selw); +} + +llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) +{ + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], + name("xcmp")); + Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); + + Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], + name("ycmp")); + Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); + + Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], + name("zcmp")); + Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); + + Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], + name("wcmp")); + Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); + + return vectorFromVals(selx, sely, selz, selw); +} + +void Instructions::printVector(llvm::Value *val) +{ + static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A"; + + if (!m_fmtPtr) { + Constant *format = ConstantArray::get(frmt, true); + ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1); + GlobalVariable* globalFormat = new GlobalVariable( + /*Type=*/arrayTy, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name(".str"), + m_mod); + globalFormat->setInitializer(format); + + Constant* const_int0 = Constant::getNullValue(IntegerType::get(32)); + std::vector const_ptr_21_indices; + const_ptr_21_indices.push_back(const_int0); + const_ptr_21_indices.push_back(const_int0); + m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat, + &const_ptr_21_indices[0], const_ptr_21_indices.size()); + } + + Function *func_printf = m_mod->getFunction("printf"); + if (!func_printf) + func_printf = declarePrintf(); + assert(func_printf); + std::vector vec = extractVector(val); + Value *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx")); + Value *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy")); + Value *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz")); + Value *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw")); + std::vector params; + params.push_back(m_fmtPtr); + params.push_back(dx); + params.push_back(dy); + params.push_back(dz); + params.push_back(dw); + CallInst *call = m_builder.CreateCall(func_printf, params.begin(), params.end(), + name("printf")); + call->setCallingConv(CallingConv::C); + call->setTailCall(true); +} + +llvm::Function * Instructions::declarePrintf() +{ + std::vector args; + ParamAttrsList *params = 0; + FunctionType* funcTy = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/args, + /*isVarArg=*/true); + Function* func_printf = new Function( + /*Type=*/funcTy, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", m_mod); + func_printf->setCallingConv(CallingConv::C); + func_printf->setParamAttrs(params); + return func_printf; +} + + +llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], name("xcmp")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + + Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], name("ycmp")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + + Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], name("zcmp")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + + Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], name("wcmp")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + + return vectorFromVals(x, y, z, w); +} +llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOGE(vec1[0], vec2[0], name("xcmp")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + + Value *ycmp = m_builder.CreateFCmpOGE(vec1[1], vec2[1], name("ycmp")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + + Value *zcmp = m_builder.CreateFCmpOGE(vec1[2], vec2[2], name("zcmp")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + + Value *wcmp = m_builder.CreateFCmpOGE(vec1[3], vec2[3], name("wcmp")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + + return vectorFromVals(x, y, z, w); +} + + +llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + + Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + + Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + + Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + + return vectorFromVals(x, y, z, w); +} + +llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) +{ + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + Value *z1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z1")); + + Value *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); + Value *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + Value *z2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(2), + name("z2")); + Value *y1z2 = mul(y1, z2); + Value *z1y2 = mul(z1, y2); + + Value *z1x2 = mul(z1, x2); + Value *x1z2 = mul(x1, z2); + + Value *x1y2 = mul(x1, y2); + Value *y1x2 = mul(y1, x2); + + return vectorFromVals(sub(y1z2, z1y2), sub(z1x2, x1z2), sub(x1y2, y1x2)); +} + + +llvm::Value * Instructions::abs(llvm::Value *in) +{ + std::vector vec = extractVector(in); + Value *xabs = callFAbs(vec[0]); + Value *yabs = callFAbs(vec[1]); + Value *zabs = callFAbs(vec[2]); + Value *wabs = callFAbs(vec[3]); + return vectorFromVals(xabs, yabs, zabs, wabs); +} + +void Instructions::ifop(llvm::Value *in) +{ + BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); + BasicBlock *ifend = new BasicBlock(name("ifthenend"), m_func,0); + + //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0); + //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0); + //BasicBlock *wblock = new BasicBlock(name("wblock"), m_func,0); + + Constant *float0 = Constant::getNullValue(Type::FloatTy); + + Value *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0), + name("extractx")); + Value *xcmp = m_builder.CreateFCmpUNE(x, float0, name("xcmp")); + m_builder.CreateCondBr(xcmp, ifthen, ifend); + //m_builder.SetInsertPoint(yblock); + + m_builder.SetInsertPoint(ifthen); + m_ifStack.push(ifend); +} + +llvm::BasicBlock * Instructions::currentBlock() const +{ + return m_builder.GetInsertBlock(); +} + +void Instructions::elseop() +{ + assert(!m_ifStack.empty()); + BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); + m_builder.CreateBr(ifend); + m_builder.SetInsertPoint(m_ifStack.top()); + currentBlock()->setName(name("ifelse")); + m_ifStack.pop(); + m_ifStack.push(ifend); +} + +void Instructions::endif() +{ + assert(!m_ifStack.empty()); + m_builder.CreateBr(m_ifStack.top()); + m_builder.SetInsertPoint(m_ifStack.top()); + m_ifStack.pop(); +} + +llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3) +{ + llvm::Value *m = mul(in1, in2); + llvm::Value *vec1 = constVector(1.f, 1.f, 1.f, 1.f); + llvm::Value *s = sub(vec1, in1); + return add(m, mul(s, in3)); +} + +void Instructions::beginLoop() +{ + BasicBlock *begin = new BasicBlock(name("loop"), m_func,0); + BasicBlock *end = new BasicBlock(name("endloop"), m_func,0); + + m_builder.CreateBr(begin); + Loop loop; + loop.begin = begin; + loop.end = end; + m_builder.SetInsertPoint(begin); + m_loopStack.push(loop); +} + +void Instructions::endLoop() +{ + assert(!m_loopStack.empty()); + Loop loop = m_loopStack.top(); + m_builder.CreateBr(loop.begin); + loop.end->moveAfter(currentBlock()); + m_builder.SetInsertPoint(loop.end); + m_loopStack.pop(); +} + +void Instructions::brk() +{ + assert(!m_loopStack.empty()); + BasicBlock *unr = new BasicBlock(name("unreachable"), m_func,0); + m_builder.CreateBr(m_loopStack.top().end); + m_builder.SetInsertPoint(unr); +} + +llvm::Value * Instructions::trunc(llvm::Value *in) +{ + std::vector vec = extractVector(in); + Value *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32), + name("ftoix")); + Value *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32), + name("ftoiy")); + Value *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32), + name("ftoiz")); + Value *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32), + name("ftoiw")); + Value *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy, + name("fx")); + Value *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy, + name("fy")); + Value *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy, + name("fz")); + Value *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy, + name("fw")); + return vectorFromVals(fx, fy, fz, fw); +} + +void Instructions::end() +{ + m_builder.CreateRetVoid(); +} + +void Instructions::cal(int label, llvm::Value *input) +{ + std::vector params; + params.push_back(input); + llvm::Function *func = findFunction(label); + + m_builder.CreateCall(func, params.begin(), params.end()); +} + +llvm::Function * Instructions::declareFunc(int label) +{ + PointerType *vecPtr = PointerType::getUnqual(m_floatVecType); + std::vector args; + args.push_back(vecPtr); + args.push_back(vecPtr); + args.push_back(vecPtr); + args.push_back(vecPtr); + ParamAttrsList *params = 0; + FunctionType *funcType = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/args, + /*isVarArg=*/false); + std::string name = createFuncName(label); + Function *func = new Function( + /*Type=*/funcType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/name.c_str(), m_mod); + func->setCallingConv(CallingConv::C); + func->setParamAttrs(params); + return func; +} + +void Instructions::bgnSub(unsigned label) +{ + llvm::Function *func = findFunction(label); + + Function::arg_iterator args = func->arg_begin(); + Value *ptr_INPUT = args++; + ptr_INPUT->setName("INPUT"); + m_storage->pushArguments(ptr_INPUT); + + llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); + + m_func = func; + m_builder.SetInsertPoint(entry); +} + +void Instructions::endSub() +{ + m_func = 0; + m_builder.SetInsertPoint(0); +} + +llvm::Function * Instructions::findFunction(int label) +{ + llvm::Function *func = m_functions[label]; + if (!func) { + func = declareFunc(label); + m_functions[label] = func; + } + return func; +} + +llvm::Value * Instructions::constVector(float x, float y, float z, float w) +{ + std::vector vec(4); + vec[0] = ConstantFP::get(Type::FloatTy, APFloat(x)); + vec[1] = ConstantFP::get(Type::FloatTy, APFloat(y)); + vec[2] = ConstantFP::get(Type::FloatTy, APFloat(z)); + vec[3] = ConstantFP::get(Type::FloatTy, APFloat(w)); + return ConstantVector::get(m_floatVecType, vec); +} + + +std::vector Instructions::extractVector(llvm::Value *vec) +{ + std::vector elems(4); + elems[0] = m_builder.CreateExtractElement(vec, m_storage->constantInt(0), + name("x")); + elems[1] = m_builder.CreateExtractElement(vec, m_storage->constantInt(1), + name("y")); + elems[2] = m_builder.CreateExtractElement(vec, m_storage->constantInt(2), + name("z")); + elems[3] = m_builder.CreateExtractElement(vec, m_storage->constantInt(3), + name("w")); + return elems; +} + +llvm::Value * Instructions::cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3) +{ + llvm::Function *func = m_mod->getFunction("cmp"); + assert(func); + + std::vector params; + params.push_back(in1); + params.push_back(in2); + params.push_back(in3); + CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), name("cmpres")); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::cos(llvm::Value *in) +{ +#if 0 + llvm::Function *func = m_mod->getFunction("vcos"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("cosres")); + call->setTailCall(false); + return call; +#else + std::vector elems = extractVector(in); + Function *func = m_mod->getFunction("cosf"); + assert(func); + CallInst *cos = m_builder.CreateCall(func, elems[0], name("cosres")); + cos->setCallingConv(CallingConv::C); + cos->setTailCall(true); + return vectorFromVals(cos, cos, cos, cos); +#endif +} + +llvm::Value * Instructions::scs(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("scs"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("scsres")); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::kilp(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("kilp"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("kilpres")); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::sin(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("vsin"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("sinres")); + call->setTailCall(false); + return call; +} +#endif //MESA_LLVM + + diff --git a/src/gallium/auxiliary/llvm/instructions.h b/src/gallium/auxiliary/llvm/instructions.h new file mode 100644 index 0000000000..9ebc17dd8e --- /dev/null +++ b/src/gallium/auxiliary/llvm/instructions.h @@ -0,0 +1,152 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef INSTRUCTIONS_H +#define INSTRUCTIONS_H + +#include +#include +#include +#include + +#include +#include + +namespace llvm { + class VectorType; + class Function; +} + +class Storage; + +class Instructions +{ +public: + Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, + Storage *storage); + + llvm::BasicBlock *currentBlock() const; + + llvm::Value *abs(llvm::Value *in1); + llvm::Value *arl(llvm::Value *in1); + llvm::Value *add(llvm::Value *in1, llvm::Value *in2); + void beginLoop(); + void bgnSub(unsigned); + void brk(); + void cal(int label, llvm::Value *input); + llvm::Value *cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); + llvm::Value *cos(llvm::Value *in); + llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); + void elseop(); + void endif(); + void endLoop(); + void end(); + void endSub(); + llvm::Value *ex2(llvm::Value *in); + llvm::Value *floor(llvm::Value *in); + llvm::Value *frc(llvm::Value *in); + void ifop(llvm::Value *in); + llvm::Value *kilp(llvm::Value *in); + llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3); + llvm::Value *lit(llvm::Value *in); + llvm::Value *lg2(llvm::Value *in); + llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in2); + llvm::Value *min(llvm::Value *in1, llvm::Value *in2); + llvm::Value *max(llvm::Value *in1, llvm::Value *in2); + llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); + llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); + llvm::Value *rcp(llvm::Value *in); + llvm::Value *rsq(llvm::Value *in); + llvm::Value *scs(llvm::Value *in); + llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sin(llvm::Value *in); + llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); + llvm::Value *trunc(llvm::Value *in); + + void printVector(llvm::Value *val); +private: + const char *name(const char *prefix); + + llvm::Value *callFAbs(llvm::Value *val); + llvm::Value *callFloor(llvm::Value *val); + llvm::Value *callFSqrt(llvm::Value *val); + llvm::Value *callFLog(llvm::Value *val); + llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2); + + llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w=0); + + llvm::Value *constVector(float x, float y, float z, float w); + + llvm::Function *declarePrintf(); + llvm::Function *declareFunc(int label); + + llvm::Function *findFunction(int label); + + std::vector extractVector(llvm::Value *vec); +private: + llvm::Module *m_mod; + llvm::Function *m_func; + char m_name[32]; + llvm::LLVMFoldingBuilder m_builder; + int m_idx; + + llvm::VectorType *m_floatVecType; + + llvm::Function *m_llvmFSqrt; + llvm::Function *m_llvmFAbs; + llvm::Function *m_llvmPow; + llvm::Function *m_llvmFloor; + llvm::Function *m_llvmFlog; + llvm::Function *m_llvmLit; + + llvm::Constant *m_fmtPtr; + + std::stack m_ifStack; + struct Loop { + llvm::BasicBlock *begin; + llvm::BasicBlock *end; + }; + std::stack m_loopStack; + std::map m_functions; + Storage *m_storage; +}; + +#endif diff --git a/src/gallium/auxiliary/llvm/instructionssoa.cpp b/src/gallium/auxiliary/llvm/instructionssoa.cpp new file mode 100644 index 0000000000..a4d5046637 --- /dev/null +++ b/src/gallium/auxiliary/llvm/instructionssoa.cpp @@ -0,0 +1,121 @@ +#include "instructionssoa.h" + +#include "storagesoa.h" + +#include + +using namespace llvm; + +InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, + llvm::BasicBlock *block, StorageSoa *storage) + : m_builder(block), + m_storage(storage), + m_idx(0) +{ +} + +const char * InstructionsSoa::name(const char *prefix) const +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::Value * InstructionsSoa::vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + Constant *constVector = Constant::getNullValue(vectorType); + Value *res = m_builder.CreateInsertElement(constVector, x, + m_storage->constantInt(0), + name("vecx")); + res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), + name("vecxy")); + res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), + name("vecxyz")); + if (w) + res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), + name("vecxyzw")); + return res; +} + +std::vector InstructionsSoa::arl(const std::vector in) +{ + std::vector res(4); + + //Extract x's + llvm::Value *x1 = m_builder.CreateExtractElement(in[0], + m_storage->constantInt(0), + name("extractX")); + //cast it to an unsigned int + x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast")); + + res[0] = x1;//vectorFromVals(x1, x2, x3, x4); + //only x is valid. the others shouldn't be necessary + /* + res[1] = Constant::getNullValue(m_floatVecType); + res[2] = Constant::getNullValue(m_floatVecType); + res[3] = Constant::getNullValue(m_floatVecType); + */ + + return res; +} + + +std::vector InstructionsSoa::add(const std::vector in1, + const std::vector in2) +{ + std::vector res(4); + + res[0] = m_builder.CreateAdd(in1[0], in2[0], name("addx")); + res[1] = m_builder.CreateAdd(in1[1], in2[1], name("addy")); + res[2] = m_builder.CreateAdd(in1[2], in2[2], name("addz")); + res[3] = m_builder.CreateAdd(in1[3], in2[3], name("addw")); + + return res; +} + +std::vector InstructionsSoa::mul(const std::vector in1, + const std::vector in2) +{ + std::vector res(4); + + res[0] = m_builder.CreateMul(in1[0], in2[0], name("mulx")); + res[1] = m_builder.CreateMul(in1[1], in2[1], name("muly")); + res[2] = m_builder.CreateMul(in1[2], in2[2], name("mulz")); + res[3] = m_builder.CreateMul(in1[3], in2[3], name("mulw")); + + return res; +} + +void InstructionsSoa::end() +{ + m_builder.CreateRetVoid(); +} + +std::vector InstructionsSoa::madd(const std::vector in1, + const std::vector in2, + const std::vector in3) +{ + std::vector res = mul(in1, in2); + return add(res, in3); +} + +std::vector InstructionsSoa::extractVector(llvm::Value *vector) +{ + std::vector res(4); + res[0] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(0), + name("extract1X")); + res[1] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(1), + name("extract2X")); + res[2] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(2), + name("extract3X")); + res[3] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(3), + name("extract4X")); + + return res; +} diff --git a/src/gallium/auxiliary/llvm/instructionssoa.h b/src/gallium/auxiliary/llvm/instructionssoa.h new file mode 100644 index 0000000000..4169dcbb2e --- /dev/null +++ b/src/gallium/auxiliary/llvm/instructionssoa.h @@ -0,0 +1,74 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INSTRUCTIONSSOA_H +#define INSTRUCTIONSSOA_H + +#include + +#include + +namespace llvm { + class Module; + class Function; + class BasicBlock; + class Value; +} +class StorageSoa; + +class InstructionsSoa +{ +public: + InstructionsSoa(llvm::Module *mod, llvm::Function *func, + llvm::BasicBlock *block, StorageSoa *storage); + + std::vector arl(const std::vector in); + + std::vector add(const std::vector in1, + const std::vector in2); + std::vector madd(const std::vector in1, + const std::vector in2, + const std::vector in3); + std::vector mul(const std::vector in1, + const std::vector in2); + void end(); + + std::vector extractVector(llvm::Value *vector); +private: + const char * name(const char *prefix) const; + llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w); +private: + llvm::LLVMFoldingBuilder m_builder; + StorageSoa *m_storage; +private: + mutable int m_idx; + mutable char m_name[32]; +}; + + +#endif diff --git a/src/gallium/auxiliary/llvm/llvm_builtins.c b/src/gallium/auxiliary/llvm/llvm_builtins.c new file mode 100644 index 0000000000..4f98d754ba --- /dev/null +++ b/src/gallium/auxiliary/llvm/llvm_builtins.c @@ -0,0 +1,115 @@ +/*clang --emit-llvm llvm_builtins.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=gallivm_builtins.cpp -f -for=shader -funcname=createGallivmBuiltins*/ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +typedef __attribute__(( ocu_vector_type(4) )) float float4; + +extern float powf(float a, float b); + +inline float approx(float a, float b) +{ + if (b < -128.0f) b = -128.0f; + if (b > 128.0f) b = 128.0f; + if (a < 0) a = 0; + return powf(a, b); +} + +inline float4 lit(float4 tmp) +{ + float4 result; + result.x = 1.0; + result.w = 1.0; + if (tmp.x > 0) { + result.y = tmp.x; + result.z = approx(tmp.y, tmp.w); + } else { + result.y = 0; + result.z = 0; + } + return result; +} + +inline float4 cmp(float4 tmp0, float4 tmp1, float4 tmp2) +{ + float4 result; + + result.x = (tmp0.x < 0.0) ? tmp1.x : tmp2.x; + result.y = (tmp0.y < 0.0) ? tmp1.y : tmp2.y; + result.z = (tmp0.z < 0.0) ? tmp1.z : tmp2.z; + result.w = (tmp0.w < 0.0) ? tmp1.w : tmp2.w; + + return result; +} + +extern float cosf(float val); +extern float sinf(float val); + +inline float4 vcos(float4 val) +{ + float4 result; + printf("VEC IN is %f %f %f %f\n", val.x, val.y, val.z, val.w); + result.x = cosf(val.x); + result.y = cosf(val.x); + result.z = cosf(val.x); + result.w = cosf(val.x); + printf("VEC OUT is %f %f %f %f\n", result.x, result.y, result.z, result.w); + return result; +} + +inline float4 scs(float4 val) +{ + float4 result; + float tmp = val.x; + result.x = cosf(tmp); + result.y = sinf(tmp); + return result; +} + + +inline float4 vsin(float4 val) +{ + float4 result; + float tmp = val.x; + float res = sinf(tmp); + result.x = res; + result.y = res; + result.z = res; + result.w = res; + return result; +} + +inline int kilp(float4 val) +{ + if (val.x < 0 || val.y < 0 || val.z < 0 || val.w < 0) + return 1; + else + return 0; +} diff --git a/src/gallium/auxiliary/llvm/loweringpass.cpp b/src/gallium/auxiliary/llvm/loweringpass.cpp new file mode 100644 index 0000000000..556dbec366 --- /dev/null +++ b/src/gallium/auxiliary/llvm/loweringpass.cpp @@ -0,0 +1,17 @@ +#include "loweringpass.h" + +using namespace llvm; + +char LoweringPass::ID = 0; +RegisterPass X("lowering", "Lowering Pass"); + +LoweringPass::LoweringPass() + : ModulePass((intptr_t)&ID) +{ +} + +bool LoweringPass::runOnModule(Module &m) +{ + llvm::cerr << "Hello: " << m.getModuleIdentifier() << "\n"; + return false; +} diff --git a/src/gallium/auxiliary/llvm/loweringpass.h b/src/gallium/auxiliary/llvm/loweringpass.h new file mode 100644 index 0000000000..f62dcf6ba7 --- /dev/null +++ b/src/gallium/auxiliary/llvm/loweringpass.h @@ -0,0 +1,15 @@ +#ifndef LOWERINGPASS_H +#define LOWERINGPASS_H + +#include "llvm/Pass.h" +#include "llvm/Module.h" + +struct LoweringPass : public llvm::ModulePass +{ + static char ID; + LoweringPass(); + + virtual bool runOnModule(llvm::Module &m); +}; + +#endif diff --git a/src/gallium/auxiliary/llvm/storage.cpp b/src/gallium/auxiliary/llvm/storage.cpp new file mode 100644 index 0000000000..c4326de8c5 --- /dev/null +++ b/src/gallium/auxiliary/llvm/storage.cpp @@ -0,0 +1,364 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "storage.h" + +#include "gallivm_p.h" + +#include "pipe/p_shader_tokens.h" +#include +#include +#include + +#include +#include +#include +#include +#include + +using namespace llvm; + +Storage::Storage(llvm::BasicBlock *block, llvm::Value *input) + : m_block(block), + m_INPUT(input), + m_addrs(32), + m_idx(0) +{ + m_floatVecType = VectorType::get(Type::FloatTy, 4); + m_intVecType = VectorType::get(IntegerType::get(32), 4); + + m_undefFloatVec = UndefValue::get(m_floatVecType); + m_undefIntVec = UndefValue::get(m_intVecType); + m_extSwizzleVec = 0; + + m_numConsts = 0; +} + +//can only build vectors with all members in the [0, 9] range +llvm::Constant *Storage::shuffleMask(int vec) +{ + if (!m_extSwizzleVec) { + std::vector elems; + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); + m_extSwizzleVec = ConstantVector::get(m_floatVecType, elems); + } + + if (m_intVecs.find(vec) != m_intVecs.end()) { + return m_intVecs[vec]; + } + int origVec = vec; + Constant* const_vec = 0; + if (origVec == 0) { + const_vec = Constant::getNullValue(m_intVecType); + } else { + int x = gallivm_x_swizzle(vec); + int y = gallivm_y_swizzle(vec); + int z = gallivm_z_swizzle(vec); + int w = gallivm_w_swizzle(vec); + std::vector elems; + elems.push_back(constantInt(x)); + elems.push_back(constantInt(y)); + elems.push_back(constantInt(z)); + elems.push_back(constantInt(w)); + const_vec = ConstantVector::get(m_intVecType, elems); + } + + m_intVecs[origVec] = const_vec; + return const_vec; +} + +llvm::ConstantInt *Storage::constantInt(int idx) +{ + if (m_constInts.find(idx) != m_constInts.end()) { + return m_constInts[idx]; + } + ConstantInt *const_int = ConstantInt::get(APInt(32, idx)); + m_constInts[idx] = const_int; + return const_int; +} + +llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) +{ + Value *val = element(InputsArg, idx, indIdx); + LoadInst *load = new LoadInst(val, name("input"), false, m_block); + load->setAlignment(8); + + return load; +} + +llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) +{ + m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; + + Value *elem = element(ConstsArg, idx, indIdx); + LoadInst *load = new LoadInst(elem, name("const"), false, m_block); + load->setAlignment(8); + return load; +} + +llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) +{ + Constant *mask = shuffleMask(shuffle); + ShuffleVectorInst *res = + new ShuffleVectorInst(vec, m_extSwizzleVec, mask, + name("shuffle"), m_block); + return res; +} + + +llvm::Value *Storage::tempElement(int idx, llvm::Value *indIdx) +{ + Value *elem = element(TempsArg, idx, indIdx); + + LoadInst *load = new LoadInst(elem, name("temp"), false, m_block); + load->setAlignment(8); + + return load; +} + +void Storage::setTempElement(int idx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = 0; + if (m_tempWriteMap[idx]) + templ = tempElement(idx); + val = maskWrite(val, mask, templ); + } + Value *elem = element(TempsArg, idx); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); + m_tempWriteMap[idx] = true; +} + +void Storage::setOutputElement(int dstIdx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = 0; + if (m_destWriteMap[dstIdx]) + templ = outputElement(dstIdx); + val = maskWrite(val, mask, templ); + } + + Value *elem = element(DestsArg, dstIdx); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); + m_destWriteMap[dstIdx] = true; +} + +llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ) +{ + llvm::Value *dst = templ; + if (!dst) + dst = Constant::getNullValue(m_floatVecType); + if ((mask & TGSI_WRITEMASK_X)) { + llvm::Value *x = new ExtractElementInst(src, unsigned(0), + name("x"), m_block); + dst = new InsertElementInst(dst, x, unsigned(0), + name("dstx"), m_block); + } + if ((mask & TGSI_WRITEMASK_Y)) { + llvm::Value *y = new ExtractElementInst(src, unsigned(1), + name("y"), m_block); + dst = new InsertElementInst(dst, y, unsigned(1), + name("dsty"), m_block); + } + if ((mask & TGSI_WRITEMASK_Z)) { + llvm::Value *z = new ExtractElementInst(src, unsigned(2), + name("z"), m_block); + dst = new InsertElementInst(dst, z, unsigned(2), + name("dstz"), m_block); + } + if ((mask & TGSI_WRITEMASK_W)) { + llvm::Value *w = new ExtractElementInst(src, unsigned(3), + name("w"), m_block); + dst = new InsertElementInst(dst, w, unsigned(3), + name("dstw"), m_block); + } + return dst; +} + +const char * Storage::name(const char *prefix) +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +int Storage::numConsts() const +{ + return m_numConsts; +} + +llvm::Value * Storage::addrElement(int idx) const +{ + Value *ret = m_addrs[idx]; + if (!ret) + return m_undefFloatVec; + return ret; +} + +void Storage::setAddrElement(int idx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = m_addrs[idx]; + val = maskWrite(val, mask, templ); + } + m_addrs[idx] = val; +} + +llvm::Value * Storage::extractIndex(llvm::Value *vec) +{ + llvm::Value *x = new ExtractElementInst(vec, unsigned(0), + name("x"), m_block); + return new FPToSIInst(x, IntegerType::get(32), name("intidx"), m_block); +} + +void Storage::setCurrentBlock(llvm::BasicBlock *block) +{ + m_block = block; +} + +llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx) +{ + Value *elem = element(DestsArg, idx, indIdx); + LoadInst *load = new LoadInst(elem, name("output"), false, m_block); + load->setAlignment(8); + + return load; +} + +llvm::Value * Storage::inputPtr() const +{ + return m_INPUT; +} + +void Storage::pushArguments(llvm::Value *input) +{ + m_argStack.push(m_INPUT); + + m_INPUT = input; +} + +void Storage::popArguments() +{ + m_INPUT = m_argStack.top(); + m_argStack.pop(); +} + +void Storage::pushTemps() +{ + m_extSwizzleVec = 0; +} + +void Storage::popTemps() +{ +} + +llvm::Value * Storage::immediateElement(int idx) +{ + return m_immediates[idx]; +} + +void Storage::addImmediate(float *val) +{ + std::vector vec(4); + vec[0] = ConstantFP::get(Type::FloatTy, APFloat(val[0])); + vec[1] = ConstantFP::get(Type::FloatTy, APFloat(val[1])); + vec[2] = ConstantFP::get(Type::FloatTy, APFloat(val[2])); + vec[3] = ConstantFP::get(Type::FloatTy, APFloat(val[3])); + m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); +} + + +llvm::Value * Storage::elemPtr(Args arg) +{ + std::vector indices; + indices.push_back(constantInt(0)); + indices.push_back(constantInt(static_cast(arg))); + GetElementPtrInst *getElem = new GetElementPtrInst(m_INPUT, + indices.begin(), + indices.end(), + name("input_ptr"), + m_block); + return new LoadInst(getElem, name("input_field"), false, m_block); +} + +llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx, + llvm::Value *indIdx ) +{ + GetElementPtrInst *getElem = 0; + + if (indIdx) { + getElem = new GetElementPtrInst(ptr, + BinaryOperator::create(Instruction::Add, + indIdx, + constantInt(idx), + name("add"), + m_block), + name("field"), + m_block); + } else { + getElem = new GetElementPtrInst(ptr, + constantInt(idx), + name("field"), + m_block); + } + return getElem; +} + +llvm::Value * Storage::element(Args arg, int idx, llvm::Value *indIdx ) +{ + Value *val = elemPtr(arg); + return elemIdx(val, idx, indIdx); +} + +void Storage::setKilElement(llvm::Value *val) +{ + std::vector indices; + indices.push_back(constantInt(0)); + indices.push_back(constantInt(static_cast(KilArg))); + GetElementPtrInst *elem = new GetElementPtrInst(m_INPUT, + indices.begin(), + indices.end(), + name("kil_ptr"), + m_block); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); +} + +#endif //MESA_LLVM + + diff --git a/src/gallium/auxiliary/llvm/storage.h b/src/gallium/auxiliary/llvm/storage.h new file mode 100644 index 0000000000..8574f7554e --- /dev/null +++ b/src/gallium/auxiliary/llvm/storage.h @@ -0,0 +1,133 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef STORAGE_H +#define STORAGE_H + +#include +#include +#include +#include + +namespace llvm { + class BasicBlock; + class Constant; + class ConstantInt; + class LoadInst; + class Value; + class VectorType; +} + +class Storage +{ +public: + Storage(llvm::BasicBlock *block, + llvm::Value *input); + + llvm::Value *inputPtr() const; + + void setCurrentBlock(llvm::BasicBlock *block); + + llvm::ConstantInt *constantInt(int); + llvm::Constant *shuffleMask(int vec); + llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); + llvm::Value *constElement(int idx, llvm::Value *indIdx =0); + llvm::Value *outputElement(int idx, llvm::Value *indIdx =0); + llvm::Value *tempElement(int idx, llvm::Value *indIdx =0); + llvm::Value *immediateElement(int idx); + + void setOutputElement(int dstIdx, llvm::Value *val, int mask); + void setTempElement(int idx, llvm::Value *val, int mask); + + llvm::Value *addrElement(int idx) const; + void setAddrElement(int idx, llvm::Value *val, int mask); + + void setKilElement(llvm::Value *val); + + llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); + + llvm::Value *extractIndex(llvm::Value *vec); + + int numConsts() const; + + void pushArguments(llvm::Value *input); + void popArguments(); + void pushTemps(); + void popTemps(); + + void addImmediate(float *val); + +private: + llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); + const char *name(const char *prefix); + + enum Args { + DestsArg = 0, + InputsArg = 1, + TempsArg = 2, + ConstsArg = 3, + KilArg = 4 + }; + llvm::Value *elemPtr(Args arg); + llvm::Value *elemIdx(llvm::Value *ptr, int idx, + llvm::Value *indIdx = 0); + llvm::Value *element(Args arg, int idx, llvm::Value *indIdx = 0); + +private: + llvm::BasicBlock *m_block; + llvm::Value *m_INPUT; + + std::map m_constInts; + std::map m_intVecs; + std::vector m_addrs; + std::vector m_immediates; + + llvm::VectorType *m_floatVecType; + llvm::VectorType *m_intVecType; + + char m_name[32]; + int m_idx; + + int m_numConsts; + + std::map m_destWriteMap; + std::map m_tempWriteMap; + + llvm::Value *m_undefFloatVec; + llvm::Value *m_undefIntVec; + llvm::Value *m_extSwizzleVec; + + std::stack m_argStack; + std::stack > m_tempStack; +}; + +#endif diff --git a/src/gallium/auxiliary/llvm/storagesoa.cpp b/src/gallium/auxiliary/llvm/storagesoa.cpp new file mode 100644 index 0000000000..ed0674a96f --- /dev/null +++ b/src/gallium/auxiliary/llvm/storagesoa.cpp @@ -0,0 +1,389 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "storagesoa.h" + +#include "gallivm_p.h" + +#include "pipe/p_shader_tokens.h" +#include "pipe/p_debug.h" + +#include +#include +#include + +#include +#include +#include +#include +#include + +using namespace llvm; + + +StorageSoa::StorageSoa(llvm::BasicBlock *block, + llvm::Value *input, + llvm::Value *output, + llvm::Value *consts, + llvm::Value *temps) + : m_block(block), + m_input(input), + m_output(output), + m_consts(consts), + m_temps(temps), + m_immediates(0), + m_idx(0) +{ +} + +void StorageSoa::addImmediate(float *vec) +{ + std::vector vals(4); + vals[0] = vec[0]; + vals[1] = vec[1]; + vals[2] = vec[2]; + vals[3] = vec[3]; + m_immediatesToFlush.push_back(vals); +} + +void StorageSoa::declareImmediates() +{ + if (m_immediatesToFlush.empty()) + return; + + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorChannels = ArrayType::get(vectorType, 4); + ArrayType *arrayType = ArrayType::get(vectorChannels, m_immediatesToFlush.size()); + + m_immediates = new GlobalVariable( + /*Type=*/arrayType, + /*isConstant=*/false, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name("immediates"), + currentModule()); + + std::vector arrayVals; + for (unsigned int i = 0; i < m_immediatesToFlush.size(); ++i) { + std::vector vec = m_immediatesToFlush[i]; + std::vector vals(4); + std::vector channelArray; + + vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0]; + llvm::Constant *xChannel = createConstGlobalVector(vals); + + vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1]; + llvm::Constant *yChannel = createConstGlobalVector(vals); + + vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2]; + llvm::Constant *zChannel = createConstGlobalVector(vals); + + vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3]; + llvm::Constant *wChannel = createConstGlobalVector(vals); + channelArray.push_back(xChannel); + channelArray.push_back(yChannel); + channelArray.push_back(zChannel); + channelArray.push_back(wChannel); + Constant *constChannels = ConstantArray::get(vectorChannels, + channelArray); + arrayVals.push_back(constChannels); + } + Constant *constArray = ConstantArray::get(arrayType, arrayVals); + m_immediates->setInitializer(constArray); + + m_immediatesToFlush.clear(); +} + +llvm::Value *StorageSoa::addrElement(int idx) const +{ + std::map::const_iterator itr = m_addresses.find(idx); + if (itr == m_addresses.end()) { + debug_printf("Trying to access invalid shader 'address'\n"); + return 0; + } + llvm::Value * res = (*itr).second; + + res = new LoadInst(res, name("addr"), false, m_block); + + return res; +} + +std::vector StorageSoa::inputElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_input, idx, 0); + res[1] = element(m_input, idx, 1); + res[2] = element(m_input, idx, 2); + res[3] = element(m_input, idx, 3); + + return res; +} + +std::vector StorageSoa::constElement(llvm::Value *idx) +{ + std::vector res(4); + llvm::Value *xChannel, *yChannel, *zChannel, *wChannel; + + xChannel = elementPointer(m_consts, idx, 0); + yChannel = elementPointer(m_consts, idx, 1); + zChannel = elementPointer(m_consts, idx, 2); + wChannel = elementPointer(m_consts, idx, 3); + + res[0] = alignedArrayLoad(xChannel); + res[1] = alignedArrayLoad(yChannel); + res[2] = alignedArrayLoad(zChannel); + res[3] = alignedArrayLoad(wChannel); + + return res; +} + +std::vector StorageSoa::outputElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_output, idx, 0); + res[1] = element(m_output, idx, 1); + res[2] = element(m_output, idx, 2); + res[3] = element(m_output, idx, 3); + + return res; +} + +std::vector StorageSoa::tempElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_temps, idx, 0); + res[1] = element(m_temps, idx, 1); + res[2] = element(m_temps, idx, 2); + res[3] = element(m_temps, idx, 3); + + return res; +} + +std::vector StorageSoa::immediateElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_immediates, idx, 0); + res[1] = element(m_immediates, idx, 1); + res[2] = element(m_immediates, idx, 2); + res[3] = element(m_immediates, idx, 3); + + return res; +} + +llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index, + int channel) const +{ + std::vector indices; + if (m_immediates == ptr) + indices.push_back(constantInt(0)); + indices.push_back(index); + indices.push_back(constantInt(channel)); + + GetElementPtrInst *getElem = new GetElementPtrInst(ptr, + indices.begin(), + indices.end(), + name("ptr"), + m_block); + return getElem; +} + +llvm::Value * StorageSoa::element(llvm::Value *ptr, llvm::Value *index, + int channel) const +{ + llvm::Value *res = elementPointer(ptr, index, channel); + LoadInst *load = new LoadInst(res, name("element"), false, m_block); + //load->setAlignment(8); + return load; +} + +const char * StorageSoa::name(const char *prefix) const +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::ConstantInt * StorageSoa::constantInt(int idx) const +{ + if (m_constInts.find(idx) != m_constInts.end()) { + return m_constInts[idx]; + } + ConstantInt *constInt = ConstantInt::get(APInt(32, idx)); + m_constInts[idx] = constInt; + return constInt; +} + +llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + PointerType *vectorPtr = PointerType::get(vectorType, 0); + + CastInst *cast = new BitCastInst(val, vectorPtr, name("toVector"), m_block); + LoadInst *load = new LoadInst(cast, name("alignLoad"), false, m_block); + load->setAlignment(8); + return load; +} + +llvm::Module * StorageSoa::currentModule() const +{ + if (!m_block || !m_block->getParent()) + return 0; + + return m_block->getParent()->getParent(); +} + +llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector &vec) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + std::vector immValues; + ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0])); + ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1])); + ConstantFP *constz = ConstantFP::get(Type::FloatTy, APFloat(vec[2])); + ConstantFP *constw = ConstantFP::get(Type::FloatTy, APFloat(vec[3])); + immValues.push_back(constx); + immValues.push_back(consty); + immValues.push_back(constz); + immValues.push_back(constw); + Constant *constVector = ConstantVector::get(vectorType, immValues); + + return constVector; +} + +std::vector StorageSoa::load(Argument type, int idx, int swizzle, + llvm::Value *indIdx) +{ + std::vector val(4); + + //if we have an indirect index, always use that + // if not use the integer offset to create one + llvm::Value *realIndex = 0; + if (indIdx) + realIndex = indIdx; + else + realIndex = constantInt(idx); + debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx); + + switch(type) { + case Input: + val = inputElement(realIndex); + break; + case Output: + val = outputElement(realIndex); + break; + case Temp: + val = tempElement(realIndex); + break; + case Const: + val = constElement(realIndex); + break; + case Immediate: + val = immediateElement(realIndex); + break; + case Address: + debug_printf("Address not handled in the load phase!\n"); + assert(0); + break; + } + if (!gallivm_is_swizzle(swizzle)) + return val; + + std::vector res(4); + + res[0] = val[gallivm_x_swizzle(swizzle)]; + res[1] = val[gallivm_y_swizzle(swizzle)]; + res[2] = val[gallivm_z_swizzle(swizzle)]; + res[3] = val[gallivm_w_swizzle(swizzle)]; + return res; +} + +void StorageSoa::store(Argument type, int idx, const std::vector &val, + int mask) +{ + llvm::Value *out = 0; + switch(type) { + case Output: + out = m_output; + break; + case Temp: + out = m_temps; + break; + case Input: + out = m_input; + break; + case Address: { + llvm::Value *addr = m_addresses[idx]; + if (!addr) { + addAddress(idx); + addr = m_addresses[idx]; + assert(addr); + } + new StoreInst(val[0], addr, false, m_block); + return; + break; + } + default: + debug_printf("Can't save output of this type: %d !\n", type); + assert(0); + break; + } + llvm::Value *realIndex = constantInt(idx); + if ((mask & TGSI_WRITEMASK_X)) { + llvm::Value *xChannel = elementPointer(out, realIndex, 0); + new StoreInst(val[0], xChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_Y)) { + llvm::Value *yChannel = elementPointer(out, realIndex, 1); + new StoreInst(val[1], yChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_Z)) { + llvm::Value *zChannel = elementPointer(out, realIndex, 2); + new StoreInst(val[2], zChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_W)) { + llvm::Value *wChannel = elementPointer(out, realIndex, 3); + new StoreInst(val[3], wChannel, false, m_block); + } +} + +void StorageSoa::addAddress(int idx) +{ + GlobalVariable *val = new GlobalVariable( + /*Type=*/IntegerType::get(32), + /*isConstant=*/false, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name("address"), + currentModule()); + val->setInitializer(Constant::getNullValue(IntegerType::get(32))); + + debug_printf("adding to %d\n", idx); + m_addresses[idx] = val; +} diff --git a/src/gallium/auxiliary/llvm/storagesoa.h b/src/gallium/auxiliary/llvm/storagesoa.h new file mode 100644 index 0000000000..6443351f27 --- /dev/null +++ b/src/gallium/auxiliary/llvm/storagesoa.h @@ -0,0 +1,111 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef STORAGESOA_H +#define STORAGESOA_H + +#include +#include +#include + +namespace llvm { + class BasicBlock; + class Constant; + class ConstantInt; + class GlobalVariable; + class LoadInst; + class Value; + class VectorType; + class Module; +} + +class StorageSoa +{ +public: + enum Argument { + Input, + Output, + Temp, + Const, + Immediate, + Address + }; +public: + StorageSoa(llvm::BasicBlock *block, + llvm::Value *input, + llvm::Value *output, + llvm::Value *consts, + llvm::Value *temps); + + + std::vector load(Argument type, int idx, int swizzle, + llvm::Value *indIdx =0); + void store(Argument type, int idx, const std::vector &val, + int mask); + + void addImmediate(float *vec); + void declareImmediates(); + + void addAddress(int idx); + + llvm::Value * addrElement(int idx) const; + + llvm::ConstantInt *constantInt(int) const; +private: + llvm::Value *elementPointer(llvm::Value *ptr, llvm::Value *indIdx, + int channel) const; + llvm::Value *element(llvm::Value *ptr, llvm::Value *idx, + int channel) const; + const char *name(const char *prefix) const; + llvm::Value *alignedArrayLoad(llvm::Value *val); + llvm::Module *currentModule() const; + llvm::Constant *createConstGlobalVector(const std::vector &vec); + + std::vector inputElement(llvm::Value *indIdx); + std::vector constElement(llvm::Value *indIdx); + std::vector outputElement(llvm::Value *indIdx); + std::vector tempElement(llvm::Value *indIdx); + std::vector immediateElement(llvm::Value *indIdx); +private: + llvm::BasicBlock *m_block; + + llvm::Value *m_input; + llvm::Value *m_output; + llvm::Value *m_consts; + llvm::Value *m_temps; + llvm::GlobalVariable *m_immediates; + + std::map m_addresses; + + std::vector > m_immediatesToFlush; + + mutable std::map m_constInts; + mutable char m_name[32]; + mutable int m_idx; +}; + +#endif diff --git a/src/gallium/auxiliary/llvm/tgsitollvm.cpp b/src/gallium/auxiliary/llvm/tgsitollvm.cpp new file mode 100644 index 0000000000..2cb4acce32 --- /dev/null +++ b/src/gallium/auxiliary/llvm/tgsitollvm.cpp @@ -0,0 +1,1221 @@ +#include "tgsitollvm.h" + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "storage.h" +#include "instructions.h" +#include "storagesoa.h" +#include "instructionssoa.h" + +#include "pipe/p_shader_tokens.h" + +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_util.h" +#include "tgsi/util/tgsi_build.h" +#include "tgsi/util/tgsi_dump.h" + + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + + +#include +#include +#include + +using namespace llvm; + +static inline FunctionType *vertexShaderFunctionType() +{ + //Function takes three arguments, + // the calling code has to make sure the types it will + // pass are castable to the following: + // [4 x <4 x float>] inputs, + // [4 x <4 x float>] output, + // [4 x [4 x float]] consts, + // [4 x <4 x float>] temps + + std::vector funcArgs; + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorArray = ArrayType::get(vectorType, 4); + PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0); + + ArrayType *floatArray = ArrayType::get(Type::FloatTy, 4); + ArrayType *constsArray = ArrayType::get(floatArray, 4); + PointerType *constsArrayPtr = PointerType::get(constsArray, 0); + + funcArgs.push_back(vectorArrayPtr);//inputs + funcArgs.push_back(vectorArrayPtr);//output + funcArgs.push_back(constsArrayPtr);//consts + funcArgs.push_back(vectorArrayPtr);//temps + + FunctionType *functionType = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/funcArgs, + /*isVarArg=*/false); + + return functionType; +} + +static inline void +add_interpolator(struct gallivm_ir *ir, + struct gallivm_interpolate *interp) +{ + ir->interpolators[ir->num_interp] = *interp; + ++ir->num_interp; +} + +static void +translate_declaration(struct gallivm_ir *prog, + llvm::Module *module, + Storage *storage, + struct tgsi_full_declaration *decl, + struct tgsi_full_declaration *fd) +{ + if (decl->Declaration.File == TGSI_FILE_INPUT) { + unsigned first, last, mask; + uint interp_method; + + assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if (first == 0) { + mask &= ~TGSI_WRITEMASK_XY; + if (mask == TGSI_WRITEMASK_NONE) { + first++; + if (first > last) { + return; + } + } + } + + interp_method = decl->Interpolation.Interpolate; + + if (mask == TGSI_WRITEMASK_XYZW) { + unsigned i, j; + + for (i = first; i <= last; i++) { + for (j = 0; j < NUM_CHANNELS; j++) { + //interp( mach, i, j ); + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } else { + unsigned i, j; + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } + } + } +} + +static void +translate_declarationir(struct gallivm_ir *, + llvm::Module *, + StorageSoa *storage, + struct tgsi_full_declaration *decl, + struct tgsi_full_declaration *) +{ + if (decl->Declaration.File == TGSI_FILE_ADDRESS) { + int idx = decl->u.DeclarationRange.First; + storage->addAddress(idx); + } +} + +static void +translate_immediate(Storage *storage, + struct tgsi_full_immediate *imm) +{ + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert(0); + } + } + storage->addImmediate(vec); +} + + +static void +translate_immediateir(StorageSoa *storage, + struct tgsi_full_immediate *imm) +{ + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert(0); + } + } + storage->addImmediate(vec); +} + +static inline int +swizzleInt(struct tgsi_full_src_register *src) +{ + int swizzle = 0; + int start = 1000; + + for (int k = 0; k < 4; ++k) { + swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; + start /= 10; + } + return swizzle; +} + +static inline llvm::Value * +swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, + Storage *storage) +{ + int swizzle = swizzleInt(src); + + if (gallivm_is_swizzle(swizzle)) { + /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ + val = storage->shuffleVector(val, swizzle); + } + return val; +} + +static void +translate_instruction(llvm::Module *module, + Storage *storage, + Instructions *instr, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi, + unsigned instno) +{ + llvm::Value *inputs[4]; + inputs[0] = 0; + inputs[1] = 0; + inputs[2] = 0; + inputs[3] = 0; + + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + llvm::Value *val = 0; + llvm::Value *indIdx = 0; + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + indIdx = storage->extractIndex(indIdx); + } + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->constElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->inputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->tempElement(src->SrcRegister.Index); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->outputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->immediateElement(src->SrcRegister.Index); + } else { + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); + return; + } + + inputs[i] = swizzleVector(val, src, storage); + } + + /*if (inputs[0]) + instr->printVector(inputs[0]); + if (inputs[1]) + instr->printVector(inputs[1]);*/ + llvm::Value *out = 0; + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); + } + break; + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } + break; + case TGSI_OPCODE_LIT: { + out = instr->lit(inputs[0]); + } + break; + case TGSI_OPCODE_RCP: { + out = instr->rcp(inputs[0]); + } + break; + case TGSI_OPCODE_RSQ: { + out = instr->rsq(inputs[0]); + } + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP3: { + out = instr->dp3(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP4: { + out = instr->dp4(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DST: { + out = instr->dst(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MIN: { + out = instr->min(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAX: { + out = instr->max(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SLT: { + out = instr->slt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SGE: { + out = instr->sge(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SUB: { + out = instr->sub(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_LERP: { + out = instr->lerp(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: { + out = instr->frc(inputs[0]); + } + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: { + out = instr->floor(inputs[0]); + } + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: { + out = instr->ex2(inputs[0]); + } + break; + case TGSI_OPCODE_LOGBASE2: { + out = instr->lg2(inputs[0]); + } + break; + case TGSI_OPCODE_POWER: { + out = instr->pow(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_CROSSPRODUCT: { + out = instr->cross(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: { + out = instr->abs(inputs[0]); + } + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: { + out = instr->dph(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_COS: { + out = instr->cos(inputs[0]); + } + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: { + out = instr->kilp(inputs[0]); + storage->setKilElement(out); + return; + } + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: { + out = instr->sgt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SIN: { + out = instr->sin(inputs[0]); + } + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: { + instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr()); + return; + } + break; + case TGSI_OPCODE_RET: { + instr->end(); + return; + } + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: { + out = instr->cmp(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SCS: { + out = instr->scs(inputs[0]); + } + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: { + instr->brk(); + return; + } + break; + case TGSI_OPCODE_IF: { + instr->ifop(inputs[0]); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: { + instr->elseop(); + storage->setCurrentBlock(instr->currentBlock()); + return; //only state update + } + break; + case TGSI_OPCODE_ENDIF: { + instr->endif(); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: { + out = instr->trunc(inputs[0]); + } + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: { + instr->beginLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_BGNSUB: { + instr->bgnSub(instno); + storage->setCurrentBlock(instr->currentBlock()); + storage->pushTemps(); + return; + } + break; + case TGSI_OPCODE_ENDLOOP2: { + instr->endLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_ENDSUB: { + instr->endSub(); + storage->setCurrentBlock(instr->currentBlock()); + storage->popArguments(); + storage->popTemps(); + return; + } + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + instr->end(); + return; + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + if (!out) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } + + /* # not sure if we need this */ + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + /*TXT( "_SAT" );*/ + break; + case TGSI_SAT_MINUS_PLUS_ONE: + /*TXT( "_SAT[-1,1]" );*/ + break; + default: + assert( 0 ); + } + + /* store results */ + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); + } + } +} + + +static void +translate_instructionir(llvm::Module *module, + StorageSoa *storage, + InstructionsSoa *instr, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi, + unsigned instno) +{ + std::vector< std::vector > inputs(inst->Instruction.NumSrcRegs); + + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + std::vector val; + llvm::Value *indIdx = 0; + int swizzle = swizzleInt(src); + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + } + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->load(StorageSoa::Const, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->load(StorageSoa::Input, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->load(StorageSoa::Temp, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->load(StorageSoa::Output, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->load(StorageSoa::Immediate, + src->SrcRegister.Index, swizzle, indIdx); + } else { + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); + return; + } + + inputs[i] = val; + } + + std::vector out(4); + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); + } + break; + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } + break; + case TGSI_OPCODE_LIT: { + } + break; + case TGSI_OPCODE_RCP: { + } + break; + case TGSI_OPCODE_RSQ: { + } + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP3: { + } + break; + case TGSI_OPCODE_DP4: { + } + break; + case TGSI_OPCODE_DST: { + } + break; + case TGSI_OPCODE_MIN: { + } + break; + case TGSI_OPCODE_MAX: { + } + break; + case TGSI_OPCODE_SLT: { + } + break; + case TGSI_OPCODE_SGE: { + } + break; + case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SUB: { + } + break; + case TGSI_OPCODE_LERP: { + } + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: { + } + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: { + } + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: { + } + break; + case TGSI_OPCODE_LOGBASE2: { + } + break; + case TGSI_OPCODE_POWER: { + } + break; + case TGSI_OPCODE_CROSSPRODUCT: { + } + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: { + } + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: { + } + break; + case TGSI_OPCODE_COS: { + } + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: { + } + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: { + } + break; + case TGSI_OPCODE_SIN: { + } + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: { + } + break; + case TGSI_OPCODE_RET: { + } + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: { + } + break; + case TGSI_OPCODE_SCS: { + } + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: { + } + break; + case TGSI_OPCODE_IF: { + } + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: { + } + break; + case TGSI_OPCODE_ENDIF: { + } + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: { + } + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: { + } + break; + case TGSI_OPCODE_BGNSUB: { + } + break; + case TGSI_OPCODE_ENDLOOP2: { + } + break; + case TGSI_OPCODE_ENDSUB: { + } + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + instr->end(); + return; + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + if (!out[0]) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } + + /* store results */ + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->store(StorageSoa::Output, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->store(StorageSoa::Temp, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->store(StorageSoa::Address, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); + } + } +} + +llvm::Module * +tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) +{ + llvm::Module *mod = new Module("shader"); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned instno = 0; + Function* shader = mod->getFunction("execute_shader"); + std::ostringstream stream; + if (ir->type == GALLIVM_VS) { + stream << "vs_shader"; + } else { + stream << "fs_shader"; + } + stream << ir->id; + std::string func_name = stream.str(); + shader->setName(func_name.c_str()); + + Function::arg_iterator args = shader->arg_begin(); + Value *ptr_INPUT = args++; + ptr_INPUT->setName("input"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + + tgsi_parse_init(&parse, tokens); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + Storage storage(label_entry, ptr_INPUT); + Instructions instr(mod, shader, label_entry, &storage); + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declaration(ir, mod, &storage, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediate(&storage, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + translate_instruction(mod, &storage, &instr, + &parse.FullToken.FullInstruction, + &fi, instno); + ++instno; + break; + + default: + assert(0); + } + } + + tgsi_parse_free(&parse); + + ir->num_consts = storage.numConsts(); + return mod; +} + +llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, + const struct tgsi_token *tokens) +{ + llvm::Module *mod = new Module("shader"); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned instno = 0; + std::ostringstream stream; + if (ir->type == GALLIVM_VS) { + stream << "vs_shader"; + } else { + stream << "fs_shader"; + } + //stream << ir->id; + std::string func_name = stream.str(); + Function *shader = llvm::cast(mod->getOrInsertFunction( + func_name.c_str(), + vertexShaderFunctionType())); + + Function::arg_iterator args = shader->arg_begin(); + Value *input = args++; + input->setName("inputs"); + Value *output = args++; + output->setName("outputs"); + Value *consts = args++; + consts->setName("consts"); + Value *temps = args++; + temps->setName("temps"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + + tgsi_parse_init(&parse, tokens); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + StorageSoa storage(label_entry, input, output, consts, temps); + InstructionsSoa instr(mod, shader, label_entry, &storage); + + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declarationir(ir, mod, &storage, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediateir(&storage, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + storage.declareImmediates(); + translate_instructionir(mod, &storage, &instr, + &parse.FullToken.FullInstruction, + &fi, instno); + ++instno; + break; + + default: + assert(0); + } + } + + tgsi_parse_free(&parse); + + return mod; +} diff --git a/src/gallium/auxiliary/llvm/tgsitollvm.h b/src/gallium/auxiliary/llvm/tgsitollvm.h new file mode 100644 index 0000000000..7ada04d629 --- /dev/null +++ b/src/gallium/auxiliary/llvm/tgsitollvm.h @@ -0,0 +1,20 @@ +#ifndef TGSITOLLVM_H +#define TGSITOLLVM_H + + +namespace llvm { + class Module; +} + +struct gallivm_ir; +struct tgsi_token; + + +llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, + const struct tgsi_token *tokens); + + +llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, + const struct tgsi_token *tokens); + +#endif diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile new file mode 100644 index 0000000000..588629e870 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/Makefile @@ -0,0 +1,23 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = pipebuffer + +DRIVER_SOURCES = \ + pb_buffer_fenced.c \ + pb_buffer_malloc.c \ + pb_bufmgr_fenced.c \ + pb_bufmgr_mm.c \ + pb_bufmgr_pool.c \ + pb_winsys.c + +C_SOURCES = \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/auxiliary/pipebuffer/linked_list.h b/src/gallium/auxiliary/pipebuffer/linked_list.h new file mode 100644 index 0000000000..e99817fb13 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/linked_list.h @@ -0,0 +1,91 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND. USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + **************************************************************************/ + +/** + * \file + * List macros heavily inspired by the Linux kernel + * list handling. No list looping yet. + * + * Is not threadsafe, so common operations need to + * be protected using an external mutex. + */ + +#ifndef LINKED_LIST_H_ +#define LINKED_LIST_H_ + + +#include + + +struct list_head +{ + struct list_head *prev; + struct list_head *next; +}; + + +#define LIST_INITHEAD(__item) \ + do { \ + (__item)->prev = (__item); \ + (__item)->next = (__item); \ + } while (0) + +#define LIST_ADD(__item, __list) \ + do { \ + (__item)->prev = (__list); \ + (__item)->next = (__list)->next; \ + (__list)->next->prev = (__item); \ + (__list)->next = (__item); \ + } while (0) + +#define LIST_ADDTAIL(__item, __list) \ + do { \ + (__item)->next = (__list); \ + (__item)->prev = (__list)->prev; \ + (__list)->prev->next = (__item); \ + (__list)->prev = (__item); \ + } while(0) + +#define LIST_DEL(__item) \ + do { \ + (__item)->prev->next = (__item)->next; \ + (__item)->next->prev = (__item)->prev; \ + } while(0) + +#define LIST_DELINIT(__item) \ + do { \ + (__item)->prev->next = (__item)->next; \ + (__item)->next->prev = (__item)->prev; \ + (__item)->next = (__item); \ + (__item)->prev = (__item); \ + } while(0) + +#define LIST_ENTRY(__type, __item, __field) \ + ((__type *)(((char *)(__item)) - offsetof(__type, __field))) + + +#endif /*LINKED_LIST_H_*/ diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h new file mode 100644 index 0000000000..97beb5f72a --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h @@ -0,0 +1,202 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Generic code for buffers. + * + * Behind a pipe buffle handle there can be DMA buffers, client (or user) + * buffers, regular malloced buffers, etc. This file provides an abstract base + * buffer handle that allows the driver to cope with all those kinds of buffers + * in a more flexible way. + * + * There is no obligation of a winsys driver to use this library. And a pipe + * driver should be completly agnostic about it. + * + * \author José Fonseca + */ + +#ifndef PB_BUFFER_H_ +#define PB_BUFFER_H_ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" + + +struct pb_vtbl; + +/** + * Buffer description. + * + * Used when allocating the buffer. + */ +struct pb_desc +{ + unsigned alignment; + unsigned usage; +}; + + +/** + * Base class for all pb_* buffers. + */ +struct pb_buffer +{ + struct pipe_buffer base; + + /** + * Pointer to the virtual function table. + * + * Avoid accessing this table directly. Use the inline functions below + * instead to avoid mistakes. + */ + const struct pb_vtbl *vtbl; +}; + + +/** + * Virtual function table for the buffer storage operations. + * + * Note that creation is not done through this table. + */ +struct pb_vtbl +{ + void (*destroy)( struct pb_buffer *buf ); + + /** + * Map the entire data store of a buffer object into the client's address. + * flags is bitmask of PIPE_BUFFER_FLAG_READ/WRITE. + */ + void *(*map)( struct pb_buffer *buf, + unsigned flags ); + + void (*unmap)( struct pb_buffer *buf ); + + /** + * Get the base buffer and the offset. + * + * A buffer can be subdivided in smaller buffers. This method should return + * the underlaying buffer, and the relative offset. + * + * Buffers without an underlaying base buffer should return themselves, with + * a zero offset. + * + * Note that this will increase the reference count of the base buffer. + */ + void (*get_base_buffer)( struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset ); +}; + + +static INLINE struct pipe_buffer * +pb_pipe_buffer( struct pb_buffer *pbuf ) +{ + assert(pbuf); + return &pbuf->base; +} + + +static INLINE struct pb_buffer * +pb_buffer( struct pipe_buffer *buf ) +{ + assert(buf); + /* Could add a magic cookie check on debug builds. + */ + return (struct pb_buffer *)buf; +} + + +/* Accessor functions for pb->vtbl: + */ +static INLINE void * +pb_map(struct pb_buffer *buf, + unsigned flags) +{ + assert(buf); + return buf->vtbl->map(buf, flags); +} + + +static INLINE void +pb_unmap(struct pb_buffer *buf) +{ + assert(buf); + buf->vtbl->unmap(buf); +} + + +static INLINE void +pb_get_base_buffer( struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset ) +{ + buf->vtbl->get_base_buffer(buf, base_buf, offset); +} + + +static INLINE void +pb_destroy(struct pb_buffer *buf) +{ + assert(buf); + buf->vtbl->destroy(buf); +} + + +/* XXX: thread safety issues! + */ +static INLINE void +pb_reference(struct pb_buffer **dst, + struct pb_buffer *src) +{ + if (src) + src->base.refcount++; + + if (*dst && --(*dst)->base.refcount == 0) + pb_destroy( *dst ); + + *dst = src; +} + + +/** + * Malloc-based buffer to store data that can't be used by the graphics + * hardware. + */ +struct pb_buffer * +pb_malloc_buffer_create(size_t size, + const struct pb_desc *desc); + + +void +pb_init_winsys(struct pipe_winsys *winsys); + + +#endif /*PB_BUFFER_H_*/ diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c new file mode 100644 index 0000000000..f4fc3f6d71 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -0,0 +1,299 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of fenced buffers. + * + * \author José Fonseca + * \author Thomas Hellström + */ + + +#include "linked_list.h" + +#include "p_compiler.h" +#include "p_debug.h" +#include "p_winsys.h" +#include "p_thread.h" +#include "p_util.h" + +#include "pb_buffer.h" +#include "pb_buffer_fenced.h" + +#ifndef __MSC__ +#include +#endif + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct fenced_buffer_list +{ + _glthread_Mutex mutex; + + struct pipe_winsys *winsys; + + size_t numDelayed; + size_t checkDelayed; + + struct list_head delayed; +}; + + +/** + * Wrapper around a pipe buffer which adds fencing and reference counting. + */ +struct fenced_buffer +{ + struct pb_buffer base; + + struct pb_buffer *buffer; + + struct pipe_fence_handle *fence; + + struct list_head head; + struct fenced_buffer_list *list; +}; + + +static INLINE struct fenced_buffer * +fenced_buffer(struct pb_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &fenced_buffer_vtbl); + return (struct fenced_buffer *)buf; +} + + + + +static void +_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait) +{ + struct pipe_winsys *winsys = fenced_list->winsys; + struct fenced_buffer *fenced_buf; + struct list_head *list, *prev; + int signaled = -1; + + list = fenced_list->delayed.next; + + if (fenced_list->numDelayed > 3) { + unsigned i; + + for (i = 0; i < fenced_list->numDelayed; i += 3) { + list = list->next; + } + } + + prev = list->prev; + for (; list != &fenced_list->delayed; list = prev, prev = list->prev) { + + fenced_buf = LIST_ENTRY(struct fenced_buffer, list, head); + + if (signaled != 0) { + if (wait) { + signaled = winsys->fence_finish(winsys, fenced_buf->fence, 0); + } + else { + signaled = winsys->fence_signalled(winsys, fenced_buf->fence, 0); + } + } + + if (signaled != 0) + /* XXX: we are assuming that buffers are freed in the same order they + * are fenced which may not always be true... + */ + break; + + winsys->fence_reference(winsys, &fenced_buf->fence, NULL); + + LIST_DEL(list); + fenced_list->numDelayed--; + + /* Do the delayed destroy: + */ + pb_reference(&fenced_buf->buffer, NULL); + FREE(fenced_buf); + } +} + + +static void +fenced_buffer_destroy(struct pb_buffer *buf) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + + if (fenced_buf->fence) { + LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); + fenced_list->numDelayed++; + } + else { + pb_reference(&fenced_buf->buffer, NULL); + FREE(fenced_buf); + } + + if ((fenced_list->numDelayed % fenced_list->checkDelayed) == 0) + _fenced_buffer_list_check_free(fenced_list, 0); +} + + +static void * +fenced_buffer_map(struct pb_buffer *buf, + unsigned flags) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + return pb_map(fenced_buf->buffer, flags); +} + + +static void +fenced_buffer_unmap(struct pb_buffer *buf) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + pb_unmap(fenced_buf->buffer); +} + + +static void +fenced_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + pb_get_base_buffer(fenced_buf->buffer, base_buf, offset); +} + + +const struct pb_vtbl +fenced_buffer_vtbl = { + fenced_buffer_destroy, + fenced_buffer_map, + fenced_buffer_unmap, + fenced_buffer_get_base_buffer +}; + + +struct pb_buffer * +fenced_buffer_create(struct fenced_buffer_list *fenced_list, + struct pb_buffer *buffer) +{ + struct fenced_buffer *buf; + + if(!buffer) + return NULL; + + buf = CALLOC_STRUCT(fenced_buffer); + if(!buf) + return NULL; + + buf->base.base.refcount = 1; + buf->base.base.alignment = buffer->base.alignment; + buf->base.base.usage = buffer->base.usage; + buf->base.base.size = buffer->base.size; + + buf->base.vtbl = &fenced_buffer_vtbl; + buf->buffer = buffer; + buf->list = fenced_list; + + return &buf->base; +} + + +void +buffer_fence(struct pb_buffer *buf, + struct pipe_fence_handle *fence) +{ + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer_list *fenced_list = fenced_buf->list; + struct pipe_winsys *winsys = fenced_list->winsys; + + _glthread_LOCK_MUTEX(fenced_list->mutex); + winsys->fence_reference(winsys, &fenced_buf->fence, fence); + _glthread_UNLOCK_MUTEX(fenced_list->mutex); +} + + +struct fenced_buffer_list * +fenced_buffer_list_create(struct pipe_winsys *winsys) +{ + struct fenced_buffer_list *fenced_list; + + fenced_list = (struct fenced_buffer_list *)CALLOC(1, sizeof(*fenced_list)); + if (!fenced_list) + return NULL; + + fenced_list->winsys = winsys; + + LIST_INITHEAD(&fenced_list->delayed); + + fenced_list->numDelayed = 0; + + /* TODO: don't hard code this */ + fenced_list->checkDelayed = 5; + + _glthread_INIT_MUTEX(fenced_list->mutex); + + return fenced_list; +} + + +void +fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait) +{ + _glthread_LOCK_MUTEX(fenced_list->mutex); + _fenced_buffer_list_check_free(fenced_list, wait); + _glthread_UNLOCK_MUTEX(fenced_list->mutex); +} + + +void +fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) +{ + _glthread_LOCK_MUTEX(fenced_list->mutex); + + /* Wait on outstanding fences */ + while (fenced_list->numDelayed) { + _glthread_UNLOCK_MUTEX(fenced_list->mutex); + sched_yield(); + _fenced_buffer_list_check_free(fenced_list, 1); + _glthread_LOCK_MUTEX(fenced_list->mutex); + } + + _glthread_UNLOCK_MUTEX(fenced_list->mutex); + + FREE(fenced_list); +} + + diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h new file mode 100644 index 0000000000..c40b9c75e1 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h @@ -0,0 +1,117 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer fencing. + * + * "Fenced buffers" is actually a misnomer. They should be referred as + * "fenceable buffers", i.e, buffers that can be fenced, but I couldn't find + * the word "fenceable" in the dictionary. + * + * A "fenced buffer" is a decorator around a normal buffer, which adds two + * special properties: + * - the ability for the destruction to be delayed by a fence; + * - reference counting. + * + * Usually DMA buffers have a life-time that will extend the life-time of its + * handle. The end-of-life is dictated by the fence signalling. + * + * Between the handle's destruction, and the fence signalling, the buffer is + * stored in a fenced buffer list. + * + * \author José Fonseca + */ + +#ifndef PB_BUFFER_FENCED_H_ +#define PB_BUFFER_FENCED_H_ + + +#include "pipe/p_debug.h" + + +struct pipe_winsys; +struct pipe_buffer; +struct pipe_fence_handle; + + +/** + * List of buffers which are awaiting fence signalling. + */ +struct fenced_buffer_list; + + +/** + * The fenced buffer's virtual function table. + * + * NOTE: Made public for debugging purposes. + */ +extern const struct pb_vtbl fenced_buffer_vtbl; + + +/** + * Create a fenced buffer list. + * + * See also fenced_bufmgr_create for a more convenient way to use this. + */ +struct fenced_buffer_list * +fenced_buffer_list_create(struct pipe_winsys *winsys); + + +/** + * Walk the fenced buffer list to check and free signalled buffers. + */ +void +fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, + int wait); + +void +fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list); + + +/** + * Wrap a buffer in a fenced buffer. + * + * NOTE: this will not increase the buffer reference count. + */ +struct pb_buffer * +fenced_buffer_create(struct fenced_buffer_list *fenced, + struct pb_buffer *buffer); + + +/** + * Set a buffer's fence. + * + * NOTE: Although it takes a generic pb_buffer argument, it will fail + * on everything but buffers returned by fenced_buffer_create. + */ +void +buffer_fence(struct pb_buffer *buf, + struct pipe_fence_handle *fence); + + +#endif /*PB_BUFFER_FENCED_H_*/ diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c new file mode 100644 index 0000000000..9e8244f909 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c @@ -0,0 +1,127 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of malloc-based buffers to store data that can't be processed + * by the hardware. + * + * \author José Fonseca + */ + + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pb_buffer.h" + + +struct malloc_buffer +{ + struct pb_buffer base; + void *data; +}; + + +extern const struct pb_vtbl malloc_buffer_vtbl; + +static INLINE struct malloc_buffer * +malloc_buffer(struct pb_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &malloc_buffer_vtbl); + return (struct malloc_buffer *)buf; +} + + +static void +malloc_buffer_destroy(struct pb_buffer *buf) +{ + align_free(malloc_buffer(buf)->data); + FREE(buf); +} + + +static void * +malloc_buffer_map(struct pb_buffer *buf, + unsigned flags) +{ + return malloc_buffer(buf)->data; +} + + +static void +malloc_buffer_unmap(struct pb_buffer *buf) +{ + /* No-op */ +} + + +static void +malloc_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + *base_buf = buf; + *offset = 0; +} + + +const struct pb_vtbl +malloc_buffer_vtbl = { + malloc_buffer_destroy, + malloc_buffer_map, + malloc_buffer_unmap, + malloc_buffer_get_base_buffer +}; + + +struct pb_buffer * +pb_malloc_buffer_create(size_t size, + const struct pb_desc *desc) +{ + struct malloc_buffer *buf; + + /* TODO: do a single allocation */ + + buf = CALLOC_STRUCT(malloc_buffer); + if(!buf) + return NULL; + + buf->base.base.refcount = 1; + buf->base.base.alignment = desc->alignment; + buf->base.base.usage = desc->usage; + buf->base.base.size = size; + buf->base.vtbl = &malloc_buffer_vtbl; + + buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment); + if(!buf->data) { + align_free(buf); + return NULL; + } + + return &buf->base; +} diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h new file mode 100644 index 0000000000..1ddf784c97 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -0,0 +1,126 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer management. + * + * A buffer manager does only one basic thing: it creates buffers. Actually, + * "buffer factory" would probably a more accurate description. + * + * You can chain buffer managers so that you can have a finer grained memory + * management and pooling. + * + * For example, for a simple batch buffer manager you would chain: + * - the native buffer manager, which provides DMA memory from the graphics + * memory space; + * - the pool buffer manager, which keep around a pool of equally sized buffers + * to avoid latency associated with the native buffer manager; + * - the fenced buffer manager, which will delay buffer destruction until the + * the moment the card finishing processing it. + * + * \author José Fonseca + */ + +#ifndef PB_BUFMGR_H_ +#define PB_BUFMGR_H_ + + +#include + + +struct pb_desc; +struct pipe_buffer; +struct pipe_winsys; + + +/** + * Abstract base class for all buffer managers. + */ +struct pb_manager +{ + /* XXX: we will likely need more allocation flags */ + struct pb_buffer * + (*create_buffer)( struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc); + + void + (*destroy)( struct pb_manager *mgr ); +}; + + +/** + * Static buffer pool manager. + * + * Manages the allocation of equally sized buffers. It does so by allocating + * a single big buffer and divide it equally sized buffers. + * + * It is meant to manage the allocation of batch buffer pools. + */ +struct pb_manager * +pool_bufmgr_create(struct pb_manager *provider, + size_t n, size_t size, + const struct pb_desc *desc); + + +/** + * Wraper around the old memory manager. + * + * It managers buffers of different sizes. It does so by allocating a buffer + * with the size of the heap, and then using the old mm memory manager to manage + * that heap. + */ +struct pb_manager * +mm_bufmgr_create(struct pb_manager *provider, + size_t size, size_t align2); + +/** + * Same as mm_bufmgr_create. + * + * Buffer will be release when the manager is destroyed. + */ +struct pb_manager * +mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, + size_t size, size_t align2); + + +/** + * Fenced buffer manager. + * + * This manager is just meant for convenience. It wraps the buffers returned + * by another manager in fenced buffers, so that + * + * NOTE: the buffer manager that provides the buffers will be destroyed + * at the same time. + */ +struct pb_manager * +fenced_bufmgr_create(struct pb_manager *provider, + struct pipe_winsys *winsys); + + +#endif /*PB_BUFMGR_H_*/ diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c new file mode 100644 index 0000000000..c535d3276c --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c @@ -0,0 +1,131 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * \file + * A buffer manager that wraps buffers in fenced buffers. + * + * \author José Fonseca + */ + + +#include "p_debug.h" +#include "p_util.h" + +#include "pb_buffer.h" +#include "pb_buffer_fenced.h" +#include "pb_bufmgr.h" + + +struct fenced_pb_manager +{ + struct pb_manager base; + + struct pb_manager *provider; + + struct fenced_buffer_list *fenced_list; +}; + + +static INLINE struct fenced_pb_manager * +fenced_pb_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct fenced_pb_manager *)mgr; +} + + +static struct pb_buffer * +fenced_bufmgr_create_buffer(struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc) +{ + struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); + struct pb_buffer *buf; + struct pb_buffer *fenced_buf; + + /* check for free buffers before allocating new ones */ + fenced_buffer_list_check_free(fenced_mgr->fenced_list, 0); + + buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); + if(!buf) { + /* try harder to get a buffer */ + fenced_buffer_list_check_free(fenced_mgr->fenced_list, 1); + + buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc); + if(!buf) { + /* give up */ + return NULL; + } + } + + fenced_buf = fenced_buffer_create(fenced_mgr->fenced_list, buf); + if(!fenced_buf) { + assert(buf->base.refcount == 1); + pb_destroy(buf); + } + + return fenced_buf; +} + + +static void +fenced_bufmgr_destroy(struct pb_manager *mgr) +{ + struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr); + + fenced_buffer_list_destroy(fenced_mgr->fenced_list); + + fenced_mgr->provider->destroy(fenced_mgr->provider); + + FREE(fenced_mgr); +} + + +struct pb_manager * +fenced_bufmgr_create(struct pb_manager *provider, + struct pipe_winsys *winsys) +{ + struct fenced_pb_manager *fenced_mgr; + + fenced_mgr = (struct fenced_pb_manager *)CALLOC(1, sizeof(*fenced_mgr)); + if (!fenced_mgr) + return NULL; + + fenced_mgr->base.destroy = fenced_bufmgr_destroy; + fenced_mgr->base.create_buffer = fenced_bufmgr_create_buffer; + + fenced_mgr->provider = provider; + fenced_mgr->fenced_list = fenced_buffer_list_create(winsys); + if(!fenced_mgr->fenced_list) { + FREE(fenced_mgr); + return NULL; + } + + return &fenced_mgr->base; +} diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c new file mode 100644 index 0000000000..8b1b51c0e2 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -0,0 +1,593 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 1999 Wittawat Yamwong + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer manager using the old texture memory manager. + * + * \author José Fonseca + */ + + +#include "linked_list.h" + +#include "p_defines.h" +#include "p_debug.h" +#include "p_thread.h" +#include "p_util.h" +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct mem_block +{ + struct mem_block *next, *prev; + struct mem_block *next_free, *prev_free; + struct mem_block *heap; + int ofs, size; + unsigned int free:1; + unsigned int reserved:1; +}; + + +#ifdef DEBUG +/** + * For debugging purposes. + */ +static void +mmDumpMemInfo(const struct mem_block *heap) +{ + debug_printf("Memory heap %p:\n", (void *)heap); + if (heap == 0) { + debug_printf(" heap == 0\n"); + } else { + const struct mem_block *p; + + for(p = heap->next; p != heap; p = p->next) { + debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + debug_printf("\nFree list:\n"); + + for(p = heap->next_free; p != heap; p = p->next_free) { + debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + } + debug_printf("End of memory blocks\n"); +} +#endif + + +/** + * input: total size in bytes + * return: a heap pointer if OK, NULL if error + */ +static struct mem_block * +mmInit(int ofs, int size) +{ + struct mem_block *heap, *block; + + if (size <= 0) + return NULL; + + heap = CALLOC_STRUCT(mem_block); + if (!heap) + return NULL; + + block = CALLOC_STRUCT(mem_block); + if (!block) { + FREE(heap); + return NULL; + } + + heap->next = block; + heap->prev = block; + heap->next_free = block; + heap->prev_free = block; + + block->heap = heap; + block->next = heap; + block->prev = heap; + block->next_free = heap; + block->prev_free = heap; + + block->ofs = ofs; + block->size = size; + block->free = 1; + + return heap; +} + + +static struct mem_block * +SliceBlock(struct mem_block *p, + int startofs, int size, + int reserved, int alignment) +{ + struct mem_block *newblock; + + /* break left [p, newblock, p->next], then p = newblock */ + if (startofs > p->ofs) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs; + newblock->size = p->size - (startofs - p->ofs); + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size -= newblock->size; + p = newblock; + } + + /* break right, also [p, newblock, p->next] */ + if (size < p->size) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs + size; + newblock->size = p->size - size; + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size = size; + } + + /* p = middle block */ + p->free = 0; + + /* Remove p from the free list: + */ + p->next_free->prev_free = p->prev_free; + p->prev_free->next_free = p->next_free; + + p->next_free = 0; + p->prev_free = 0; + + p->reserved = reserved; + return p; +} + + +/** + * Allocate 'size' bytes with 2^align2 bytes alignment, + * restrict the search to free memory after 'startSearch' + * depth and back buffers should be in different 4mb banks + * to get better page hits if possible + * input: size = size of block + * align2 = 2^align2 bytes alignment + * startSearch = linear offset from start of heap to begin search + * return: pointer to the allocated block, 0 if error + */ +static struct mem_block * +mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) +{ + struct mem_block *p; + const int mask = (1 << align2)-1; + int startofs = 0; + int endofs; + + if (!heap || align2 < 0 || size <= 0) + return NULL; + + for (p = heap->next_free; p != heap; p = p->next_free) { + assert(p->free); + + startofs = (p->ofs + mask) & ~mask; + if ( startofs < startSearch ) { + startofs = startSearch; + } + endofs = startofs+size; + if (endofs <= (p->ofs+p->size)) + break; + } + + if (p == heap) + return NULL; + + assert(p->free); + p = SliceBlock(p,startofs,size,0,mask+1); + + return p; +} + + +#if 0 +/** + * Free block starts at offset + * input: pointer to a heap, start offset + * return: pointer to a block + */ +static struct mem_block * +mmFindBlock(struct mem_block *heap, int start) +{ + struct mem_block *p; + + for (p = heap->next; p != heap; p = p->next) { + if (p->ofs == start) + return p; + } + + return NULL; +} +#endif + + +static INLINE int +Join2Blocks(struct mem_block *p) +{ + /* XXX there should be some assertions here */ + + /* NOTE: heap->free == 0 */ + + if (p->free && p->next->free) { + struct mem_block *q = p->next; + + assert(p->ofs + p->size == q->ofs); + p->size += q->size; + + p->next = q->next; + q->next->prev = p; + + q->next_free->prev_free = q->prev_free; + q->prev_free->next_free = q->next_free; + + FREE(q); + return 1; + } + return 0; +} + + +/** + * Free block starts at offset + * input: pointer to a block + * return: 0 if OK, -1 if error + */ +static int +mmFreeMem(struct mem_block *b) +{ + if (!b) + return 0; + + if (b->free) { + debug_printf("block already free\n"); + return -1; + } + if (b->reserved) { + debug_printf("block is reserved\n"); + return -1; + } + + b->free = 1; + b->next_free = b->heap->next_free; + b->prev_free = b->heap; + b->next_free->prev_free = b; + b->prev_free->next_free = b; + + Join2Blocks(b); + if (b->prev != b->heap) + Join2Blocks(b->prev); + + return 0; +} + + +/** + * destroy MM + */ +static void +mmDestroy(struct mem_block *heap) +{ + struct mem_block *p; + + if (!heap) + return; + + for (p = heap->next; p != heap; ) { + struct mem_block *next = p->next; + FREE(p); + p = next; + } + + FREE(heap); +} + + +struct mm_pb_manager +{ + struct pb_manager base; + + _glthread_Mutex mutex; + + size_t size; + struct mem_block *heap; + + size_t align2; + + struct pb_buffer *buffer; + void *map; +}; + + +static INLINE struct mm_pb_manager * +mm_pb_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct mm_pb_manager *)mgr; +} + + +struct mm_buffer +{ + struct pb_buffer base; + + struct mm_pb_manager *mgr; + + struct mem_block *block; +}; + + +static INLINE struct mm_buffer * +mm_buffer(struct pb_buffer *buf) +{ + assert(buf); + return (struct mm_buffer *)buf; +} + + +static void +mm_buffer_destroy(struct pb_buffer *buf) +{ + struct mm_buffer *mm_buf = mm_buffer(buf); + struct mm_pb_manager *mm = mm_buf->mgr; + + assert(buf->base.refcount == 0); + + _glthread_LOCK_MUTEX(mm->mutex); + mmFreeMem(mm_buf->block); + FREE(buf); + _glthread_UNLOCK_MUTEX(mm->mutex); +} + + +static void * +mm_buffer_map(struct pb_buffer *buf, + unsigned flags) +{ + struct mm_buffer *mm_buf = mm_buffer(buf); + struct mm_pb_manager *mm = mm_buf->mgr; + + return (unsigned char *) mm->map + mm_buf->block->ofs; +} + + +static void +mm_buffer_unmap(struct pb_buffer *buf) +{ + /* No-op */ +} + + +static void +mm_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + struct mm_buffer *mm_buf = mm_buffer(buf); + struct mm_pb_manager *mm = mm_buf->mgr; + pb_get_base_buffer(mm->buffer, base_buf, offset); + *offset += mm_buf->block->ofs; +} + + +static const struct pb_vtbl +mm_buffer_vtbl = { + mm_buffer_destroy, + mm_buffer_map, + mm_buffer_unmap, + mm_buffer_get_base_buffer +}; + + +static struct pb_buffer * +mm_bufmgr_create_buffer(struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc) +{ + struct mm_pb_manager *mm = mm_pb_manager(mgr); + struct mm_buffer *mm_buf; + + /* We don't handle alignments larger then the one initially setup */ + assert(desc->alignment % (1 << mm->align2) == 0); + if(desc->alignment % (1 << mm->align2)) + return NULL; + + _glthread_LOCK_MUTEX(mm->mutex); + + mm_buf = CALLOC_STRUCT(mm_buffer); + if (!mm_buf) { + _glthread_UNLOCK_MUTEX(mm->mutex); + return NULL; + } + + mm_buf->base.base.refcount = 1; + mm_buf->base.base.alignment = desc->alignment; + mm_buf->base.base.usage = desc->usage; + mm_buf->base.base.size = size; + + mm_buf->base.vtbl = &mm_buffer_vtbl; + + mm_buf->mgr = mm; + + mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); + if(!mm_buf->block) { + debug_printf("warning: heap full\n"); +#if 0 + mmDumpMemInfo(mm->heap); +#endif + + mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); + if(!mm_buf->block) { + assert(0); + FREE(mm_buf); + _glthread_UNLOCK_MUTEX(mm->mutex); + return NULL; + } + } + + /* Some sanity checks */ + assert(0 <= mm_buf->block->ofs && mm_buf->block->ofs < mm->size); + assert(size <= mm_buf->block->size && mm_buf->block->ofs + mm_buf->block->size <= mm->size); + + _glthread_UNLOCK_MUTEX(mm->mutex); + return SUPER(mm_buf); +} + + +static void +mm_bufmgr_destroy(struct pb_manager *mgr) +{ + struct mm_pb_manager *mm = mm_pb_manager(mgr); + + _glthread_LOCK_MUTEX(mm->mutex); + + mmDestroy(mm->heap); + + pb_unmap(mm->buffer); + pb_reference(&mm->buffer, NULL); + + _glthread_UNLOCK_MUTEX(mm->mutex); + + FREE(mgr); +} + + +struct pb_manager * +mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, + size_t size, size_t align2) +{ + struct mm_pb_manager *mm; + + if(!buffer) + return NULL; + + mm = CALLOC_STRUCT(mm_pb_manager); + if (!mm) + return NULL; + + mm->base.create_buffer = mm_bufmgr_create_buffer; + mm->base.destroy = mm_bufmgr_destroy; + + mm->size = size; + mm->align2 = align2; /* 64-byte alignment */ + + _glthread_INIT_MUTEX(mm->mutex); + + mm->buffer = buffer; + + mm->map = pb_map(mm->buffer, + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); + if(!mm->map) + goto failure; + + mm->heap = mmInit(0, size); + if (!mm->heap) + goto failure; + + return SUPER(mm); + +failure: +if(mm->heap) + mmDestroy(mm->heap); + if(mm->map) + pb_unmap(mm->buffer); + if(mm) + FREE(mm); + return NULL; +} + + +struct pb_manager * +mm_bufmgr_create(struct pb_manager *provider, + size_t size, size_t align2) +{ + struct pb_buffer *buffer; + struct pb_manager *mgr; + struct pb_desc desc; + + assert(provider); + assert(provider->create_buffer); + + memset(&desc, 0, sizeof(desc)); + desc.alignment = 1 << align2; + + buffer = provider->create_buffer(provider, size, &desc); + if (!buffer) + return NULL; + + mgr = mm_bufmgr_create_from_buffer(buffer, size, align2); + if (!mgr) { + pb_reference(&buffer, NULL); + return NULL; + } + + return mgr; +} diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c new file mode 100644 index 0000000000..04477a865a --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c @@ -0,0 +1,288 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * \file + * Batch buffer pool management. + * + * \author José Fonseca + * \author Thomas Hellström + */ + + +#include "linked_list.h" + +#include "p_compiler.h" +#include "p_debug.h" +#include "p_thread.h" +#include "p_defines.h" +#include "p_util.h" + +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct pool_pb_manager +{ + struct pb_manager base; + + _glthread_Mutex mutex; + + size_t bufSize; + size_t bufAlign; + + size_t numFree; + size_t numTot; + + struct list_head free; + + struct pb_buffer *buffer; + void *map; + + struct pool_buffer *bufs; +}; + + +static INLINE struct pool_pb_manager * +pool_pb_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct pool_pb_manager *)mgr; +} + + +struct pool_buffer +{ + struct pb_buffer base; + + struct pool_pb_manager *mgr; + + struct list_head head; + + size_t start; +}; + + +static INLINE struct pool_buffer * +pool_buffer(struct pb_buffer *buf) +{ + assert(buf); + return (struct pool_buffer *)buf; +} + + + +static void +pool_buffer_destroy(struct pb_buffer *buf) +{ + struct pool_buffer *pool_buf = pool_buffer(buf); + struct pool_pb_manager *pool = pool_buf->mgr; + + assert(pool_buf->base.base.refcount == 0); + + _glthread_LOCK_MUTEX(pool->mutex); + LIST_ADD(&pool_buf->head, &pool->free); + pool->numFree++; + _glthread_UNLOCK_MUTEX(pool->mutex); +} + + +static void * +pool_buffer_map(struct pb_buffer *buf, unsigned flags) +{ + struct pool_buffer *pool_buf = pool_buffer(buf); + struct pool_pb_manager *pool = pool_buf->mgr; + void *map; + + _glthread_LOCK_MUTEX(pool->mutex); + map = (unsigned char *) pool->map + pool_buf->start; + _glthread_UNLOCK_MUTEX(pool->mutex); + return map; +} + + +static void +pool_buffer_unmap(struct pb_buffer *buf) +{ + /* No-op */ +} + + +static void +pool_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + struct pool_buffer *pool_buf = pool_buffer(buf); + struct pool_pb_manager *pool = pool_buf->mgr; + pb_get_base_buffer(pool->buffer, base_buf, offset); + *offset += pool_buf->start; +} + + +static const struct pb_vtbl +pool_buffer_vtbl = { + pool_buffer_destroy, + pool_buffer_map, + pool_buffer_unmap, + pool_buffer_get_base_buffer +}; + + +static struct pb_buffer * +pool_bufmgr_create_buffer(struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc) +{ + struct pool_pb_manager *pool = pool_pb_manager(mgr); + struct pool_buffer *pool_buf; + struct list_head *item; + + assert(size == pool->bufSize); + assert(pool->bufAlign % desc->alignment == 0); + + _glthread_LOCK_MUTEX(pool->mutex); + + if (pool->numFree == 0) { + _glthread_UNLOCK_MUTEX(pool->mutex); + debug_printf("warning: out of fixed size buffer objects\n"); + return NULL; + } + + item = pool->free.next; + + if (item == &pool->free) { + _glthread_UNLOCK_MUTEX(pool->mutex); + debug_printf("error: fixed size buffer pool corruption\n"); + return NULL; + } + + LIST_DEL(item); + --pool->numFree; + + _glthread_UNLOCK_MUTEX(pool->mutex); + + pool_buf = LIST_ENTRY(struct pool_buffer, item, head); + assert(pool_buf->base.base.refcount == 0); + pool_buf->base.base.refcount = 1; + pool_buf->base.base.alignment = desc->alignment; + pool_buf->base.base.usage = desc->usage; + + return SUPER(pool_buf); +} + + +static void +pool_bufmgr_destroy(struct pb_manager *mgr) +{ + struct pool_pb_manager *pool = pool_pb_manager(mgr); + _glthread_LOCK_MUTEX(pool->mutex); + + FREE(pool->bufs); + + pb_unmap(pool->buffer); + pb_reference(&pool->buffer, NULL); + + _glthread_UNLOCK_MUTEX(pool->mutex); + + FREE(mgr); +} + + +struct pb_manager * +pool_bufmgr_create(struct pb_manager *provider, + size_t numBufs, + size_t bufSize, + const struct pb_desc *desc) +{ + struct pool_pb_manager *pool; + struct pool_buffer *pool_buf; + size_t i; + + pool = (struct pool_pb_manager *)CALLOC(1, sizeof(*pool)); + if (!pool) + return NULL; + + pool->base.destroy = pool_bufmgr_destroy; + pool->base.create_buffer = pool_bufmgr_create_buffer; + + LIST_INITHEAD(&pool->free); + + pool->numTot = numBufs; + pool->numFree = numBufs; + pool->bufSize = bufSize; + pool->bufAlign = desc->alignment; + + _glthread_INIT_MUTEX(pool->mutex); + + pool->buffer = provider->create_buffer(provider, numBufs*bufSize, desc); + if (!pool->buffer) + goto failure; + + pool->map = pb_map(pool->buffer, + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); + if(!pool->map) + goto failure; + + pool->bufs = (struct pool_buffer *)CALLOC(numBufs, sizeof(*pool->bufs)); + if (!pool->bufs) + goto failure; + + pool_buf = pool->bufs; + for (i = 0; i < numBufs; ++i) { + pool_buf->base.base.refcount = 0; + pool_buf->base.base.alignment = 0; + pool_buf->base.base.usage = 0; + pool_buf->base.base.size = bufSize; + pool_buf->base.vtbl = &pool_buffer_vtbl; + pool_buf->mgr = pool; + pool_buf->start = i * bufSize; + LIST_ADDTAIL(&pool_buf->head, &pool->free); + pool_buf++; + } + + return SUPER(pool); + +failure: + if(pool->bufs) + FREE(pool->bufs); + if(pool->map) + pb_unmap(pool->buffer); + if(pool->buffer) + pb_reference(&pool->buffer, NULL); + if(pool) + FREE(pool); + return NULL; +} diff --git a/src/gallium/auxiliary/pipebuffer/pb_winsys.c b/src/gallium/auxiliary/pipebuffer/pb_winsys.c new file mode 100644 index 0000000000..978944091f --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_winsys.c @@ -0,0 +1,170 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Implementation of client buffer (also designated as "user buffers"), which + * are just state-tracker owned data masqueraded as buffers. + * + * \author José Fonseca + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" + +#include "pb_buffer.h" + + +/** + * User buffers are special buffers that initially reference memory + * held by the user but which may if necessary copy that memory into + * device memory behind the scenes, for submission to hardware. + * + * These are particularly useful when the referenced data is never + * submitted to hardware at all, in the particular case of software + * vertex processing. + */ +struct pb_user_buffer +{ + struct pb_buffer base; + void *data; +}; + + +extern const struct pb_vtbl pb_user_buffer_vtbl; + + +static INLINE struct pb_user_buffer * +pb_user_buffer(struct pb_buffer *buf) +{ + assert(buf); + assert(buf->vtbl == &pb_user_buffer_vtbl); + return (struct pb_user_buffer *)buf; +} + + +static void +pb_user_buffer_destroy(struct pb_buffer *buf) +{ + assert(buf); + FREE(buf); +} + + +static void * +pb_user_buffer_map(struct pb_buffer *buf, + unsigned flags) +{ + return pb_user_buffer(buf)->data; +} + + +static void +pb_user_buffer_unmap(struct pb_buffer *buf) +{ + /* No-op */ +} + + +static void +pb_user_buffer_get_base_buffer(struct pb_buffer *buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + *base_buf = buf; + *offset = 0; +} + + +const struct pb_vtbl +pb_user_buffer_vtbl = { + pb_user_buffer_destroy, + pb_user_buffer_map, + pb_user_buffer_unmap, + pb_user_buffer_get_base_buffer +}; + + +static struct pipe_buffer * +pb_winsys_user_buffer_create(struct pipe_winsys *winsys, + void *data, + unsigned bytes) +{ + struct pb_user_buffer *buf = CALLOC_STRUCT(pb_user_buffer); + + if(!buf) + return NULL; + + buf->base.base.refcount = 1; + buf->base.base.size = bytes; + buf->base.base.alignment = 0; + buf->base.base.usage = 0; + + buf->base.vtbl = &pb_user_buffer_vtbl; + buf->data = data; + + return &buf->base.base; +} + + +static void * +pb_winsys_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags) +{ + (void)winsys; + return pb_map(pb_buffer(buf), flags); +} + + +static void +pb_winsys_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + (void)winsys; + pb_unmap(pb_buffer(buf)); +} + + +static void +pb_winsys_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + (void)winsys; + pb_destroy(pb_buffer(buf)); +} + + +void +pb_init_winsys(struct pipe_winsys *winsys) +{ + winsys->user_buffer_create = pb_winsys_user_buffer_create; + winsys->buffer_map = pb_winsys_buffer_map; + winsys->buffer_unmap = pb_winsys_buffer_unmap; + winsys->buffer_destroy = pb_winsys_buffer_destroy; +} diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile new file mode 100644 index 0000000000..12a8bd0409 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -0,0 +1,3 @@ +default: + cd ../.. ; make + diff --git a/src/gallium/auxiliary/tgsi/exec/Makefile b/src/gallium/auxiliary/tgsi/exec/Makefile new file mode 100644 index 0000000000..eb8b14e0e8 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/exec/Makefile @@ -0,0 +1,3 @@ +default: + cd ../../.. ; make + diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c new file mode 100644 index 0000000000..a8f64c2287 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -0,0 +1,2485 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI interpretor/executor. + * + * Flow control information: + * + * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) + * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special + * care since a condition may be true for some quad components but false + * for other components. + * + * We basically execute all statements (even if they're in the part of + * an IF/ELSE clause that's "not taken") and use a special mask to + * control writing to destination registers. This is the ExecMask. + * See store_dest(). + * + * The ExecMask is computed from three other masks (CondMask, LoopMask and + * ContMask) which are controlled by the flow control instructions (namely: + * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). + * + * + * Authors: + * Michal Krol + * Brian Paul + */ + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_util.h" +#include "tgsi_exec.h" + +#define TILE_TOP_LEFT 0 +#define TILE_TOP_RIGHT 1 +#define TILE_BOTTOM_LEFT 2 +#define TILE_BOTTOM_RIGHT 3 + +/* + * Shorthand locations of various utility registers (_I = Index, _C = Channel) + */ +#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I +#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C +#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I +#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C +#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I +#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C +#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I +#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C +#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I +#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C +#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I +#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C +#define TEMP_128_I TGSI_EXEC_TEMP_128_I +#define TEMP_128_C TGSI_EXEC_TEMP_128_C +#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I +#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C +#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I +#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C +#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I +#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C +#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I +#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +#define FOR_EACH_CHANNEL(CHAN)\ + for (CHAN = 0; CHAN < 4; CHAN++) + +#define IS_CHANNEL_ENABLED(INST, CHAN)\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IS_CHANNEL_ENABLED2(INST, CHAN)\ + ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) + +#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED2( INST, CHAN )) + + +/** The execution mask depends on the conditional mask and the loop mask */ +#define UPDATE_EXEC_MASK(MACH) \ + MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask + + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + + + +static void +tgsi_exec_prepare( struct tgsi_exec_machine *mach ) +{ + struct tgsi_exec_labels *labels = &mach->Labels; + struct tgsi_parse_context parse; + struct tgsi_full_instruction *instructions; + struct tgsi_full_declaration *declarations; + uint maxInstructions = 10, numInstructions = 0; + uint maxDeclarations = 10, numDeclarations = 0; + uint k; + uint instno = 0; + + mach->ImmLimit = 0; + labels->count = 0; + + declarations = (struct tgsi_full_declaration *) + MALLOC( maxDeclarations * sizeof(struct tgsi_full_declaration) ); + + instructions = (struct tgsi_full_instruction *) + MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) ); + + k = tgsi_parse_init( &parse, mach->Tokens ); + if (k != TGSI_PARSE_OK) { + debug_printf("Problem parsing!\n"); + return; + } + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + uint pointer = parse.Position; + uint i; + + tgsi_parse_token( &parse ); + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + /* save expanded declaration */ + if (numDeclarations == maxDeclarations) { + declarations = REALLOC(declarations, + maxDeclarations + * sizeof(struct tgsi_full_declaration), + (maxDeclarations + 10) + * sizeof(struct tgsi_full_declaration)); + maxDeclarations += 10; + } + memcpy(declarations + numDeclarations, + &parse.FullToken.FullDeclaration, + sizeof(declarations[0])); + numDeclarations++; + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; + assert( size % 4 == 0 ); + assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES ); + + for( i = 0; i < size; i++ ) { + mach->Imms[mach->ImmLimit + i / 4][i % 4] = parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } + mach->ImmLimit += size / 4; + } + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + assert( labels->count < 128 ); + + labels->labels[labels->count][0] = instno; + labels->labels[labels->count][1] = pointer; + labels->count++; + + /* save expanded instruction */ + if (numInstructions == maxInstructions) { + instructions = REALLOC(instructions, + maxInstructions + * sizeof(struct tgsi_full_instruction), + (maxInstructions + 10) + * sizeof(struct tgsi_full_instruction)); + maxInstructions += 10; + } + memcpy(instructions + numInstructions, + &parse.FullToken.FullInstruction, + sizeof(instructions[0])); + numInstructions++; + break; + + default: + assert( 0 ); + } + } + tgsi_parse_free (&parse); + + if (mach->Declarations) { + FREE( mach->Declarations ); + } + mach->Declarations = declarations; + mach->NumDeclarations = numDeclarations; + + if (mach->Instructions) { + FREE( mach->Instructions ); + } + mach->Instructions = instructions; + mach->NumInstructions = numInstructions; +} + + +/** + * Initialize machine state by expanding tokens to full instructions, + * allocating temporary storage, setting up constants, etc. + * After this, we can call tgsi_exec_machine_run() many times. + */ +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach, + const struct tgsi_token *tokens, + uint numSamplers, + struct tgsi_sampler *samplers) +{ + uint i, k; + struct tgsi_parse_context parse; + +#if 0 + tgsi_dump(tokens, 0); +#endif + + mach->Tokens = tokens; + + mach->Samplers = samplers; + + k = tgsi_parse_init (&parse, mach->Tokens); + if (k != TGSI_PARSE_OK) { + debug_printf( "Problem parsing!\n" ); + return; + } + + mach->Processor = parse.FullHeader.Processor.Processor; + tgsi_parse_free (&parse); + + mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); + mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; + + /* Setup constants. */ + for( i = 0; i < 4; i++ ) { + mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; + mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; + mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; + mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; + mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; + mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; + mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; + mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; + } + + tgsi_exec_prepare( mach ); +} + + +void +tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach) +{ + if (mach->Instructions) { + FREE(mach->Instructions); + mach->Instructions = NULL; + mach->NumInstructions = 0; + } + if (mach->Declarations) { + FREE(mach->Declarations); + mach->Declarations = NULL; + mach->NumDeclarations = 0; + } +} + + +static void +micro_abs( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) fabs( (double) src->f[0] ); + dst->f[1] = (float) fabs( (double) src->f[1] ); + dst->f[2] = (float) fabs( (double) src->f[2] ); + dst->f[3] = (float) fabs( (double) src->f[3] ); +} + +static void +micro_add( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] + src1->f[0]; + dst->f[1] = src0->f[1] + src1->f[1]; + dst->f[2] = src0->f[2] + src1->f[2]; + dst->f[3] = src0->f[3] + src1->f[3]; +} + +static void +micro_iadd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] + src1->i[0]; + dst->i[1] = src0->i[1] + src1->i[1]; + dst->i[2] = src0->i[2] + src1->i[2]; + dst->i[3] = src0->i[3] + src1->i[3]; +} + +static void +micro_and( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] & src1->u[0]; + dst->u[1] = src0->u[1] & src1->u[1]; + dst->u[2] = src0->u[2] & src1->u[2]; + dst->u[3] = src0->u[3] & src1->u[3]; +} + +static void +micro_ceil( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) ceil( (double) src->f[0] ); + dst->f[1] = (float) ceil( (double) src->f[1] ); + dst->f[2] = (float) ceil( (double) src->f[2] ); + dst->f[3] = (float) ceil( (double) src->f[3] ); +} + +static void +micro_cos( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) cos( (double) src->f[0] ); + dst->f[1] = (float) cos( (double) src->f[1] ); + dst->f[2] = (float) cos( (double) src->f[2] ); + dst->f[3] = (float) cos( (double) src->f[3] ); +} + +static void +micro_ddx( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_ddy( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_div( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] / src1->f[0]; + dst->f[1] = src0->f[1] / src1->f[1]; + dst->f[2] = src0->f[2] / src1->f[2]; + dst->f[3] = src0->f[3] / src1->f[3]; +} + +static void +micro_udiv( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] / src1->u[0]; + dst->u[1] = src0->u[1] / src1->u[1]; + dst->u[2] = src0->u[2] / src1->u[2]; + dst->u[3] = src0->u[3] / src1->u[3]; +} + +static void +micro_eq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ieq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_exp2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = (float) pow( 2.0, (double) src->f[0] ); + dst->f[1] = (float) pow( 2.0, (double) src->f[1] ); + dst->f[2] = (float) pow( 2.0, (double) src->f[2] ); + dst->f[3] = (float) pow( 2.0, (double) src->f[3] ); +} + +static void +micro_f2it( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = (int) src->f[0]; + dst->i[1] = (int) src->f[1]; + dst->i[2] = (int) src->f[2]; + dst->i[3] = (int) src->f[3]; +} + +static void +micro_f2ut( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = (uint) src->f[0]; + dst->u[1] = (uint) src->f[1]; + dst->u[2] = (uint) src->f[2]; + dst->u[3] = (uint) src->f[3]; +} + +static void +micro_flr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) floor( (double) src->f[0] ); + dst->f[1] = (float) floor( (double) src->f[1] ); + dst->f[2] = (float) floor( (double) src->f[2] ); + dst->f[3] = (float) floor( (double) src->f[3] ); +} + +static void +micro_frc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = src->f[0] - (float) floor( (double) src->f[0] ); + dst->f[1] = src->f[1] - (float) floor( (double) src->f[1] ); + dst->f[2] = src->f[2] - (float) floor( (double) src->f[2] ); + dst->f[3] = src->f[3] - (float) floor( (double) src->f[3] ); +} + +static void +micro_ge( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_i2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) src->i[0]; + dst->f[1] = (float) src->i[1]; + dst->f[2] = (float) src->i[2]; + dst->f[3] = (float) src->i[3]; +} + +static void +micro_lg2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) log( (double) src->f[0] ) * 1.442695f; + dst->f[1] = (float) log( (double) src->f[1] ) * 1.442695f; + dst->f[2] = (float) log( (double) src->f[2] ) * 1.442695f; + dst->f[3] = (float) log( (double) src->f[3] ) * 1.442695f; +} + +static void +micro_lt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ilt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_ult( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; +} + +static void +micro_max( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_min( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_umod( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] % src1->u[0]; + dst->u[1] = src0->u[1] % src1->u[1]; + dst->u[2] = src0->u[2] % src1->u[2]; + dst->u[3] = src0->u[3] % src1->u[3]; +} + +static void +micro_mul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] * src1->f[0]; + dst->f[1] = src0->f[1] * src1->f[1]; + dst->f[2] = src0->f[2] * src1->f[2]; + dst->f[3] = src0->f[3] * src1->f[3]; +} + +static void +micro_imul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] * src1->i[0]; + dst->i[1] = src0->i[1] * src1->i[1]; + dst->i[2] = src0->i[2] * src1->i[2]; + dst->i[3] = src0->i[3] * src1->i[3]; +} + +static void +micro_imul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->i[0] = src0->i[0] * src1->i[0]; + dst1->i[1] = src0->i[1] * src1->i[1]; + dst1->i[2] = src0->i[2] * src1->i[2]; + dst1->i[3] = src0->i[3] * src1->i[3]; + dst0->i[0] = 0; + dst0->i[1] = 0; + dst0->i[2] = 0; + dst0->i[3] = 0; +} + +static void +micro_umul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->u[0] = src0->u[0] * src1->u[0]; + dst1->u[1] = src0->u[1] * src1->u[1]; + dst1->u[2] = src0->u[2] * src1->u[2]; + dst1->u[3] = src0->u[3] * src1->u[3]; + dst0->u[0] = 0; + dst0->u[1] = 0; + dst0->u[2] = 0; + dst0->u[3] = 0; +} + +static void +micro_movc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2 ) +{ + dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; + dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; + dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; + dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; +} + +static void +micro_neg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = -src->f[0]; + dst->f[1] = -src->f[1]; + dst->f[2] = -src->f[2]; + dst->f[3] = -src->f[3]; +} + +static void +micro_ineg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = -src->i[0]; + dst->i[1] = -src->i[1]; + dst->i[2] = -src->i[2]; + dst->i[3] = -src->i[3]; +} + +static void +micro_not( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = ~src->u[0]; + dst->u[1] = ~src->u[1]; + dst->u[2] = ~src->u[2]; + dst->u[3] = ~src->u[3]; +} + +static void +micro_or( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] | src1->u[0]; + dst->u[1] = src0->u[1] | src1->u[1]; + dst->u[2] = src0->u[2] | src1->u[2]; + dst->u[3] = src0->u[3] | src1->u[3]; +} + +static void +micro_pow( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = (float) pow( (double) src0->f[0], (double) src1->f[0] ); + dst->f[1] = (float) pow( (double) src0->f[1], (double) src1->f[1] ); + dst->f[2] = (float) pow( (double) src0->f[2], (double) src1->f[2] ); + dst->f[3] = (float) pow( (double) src0->f[3], (double) src1->f[3] ); +} + +static void +micro_rnd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) floor( (double) (src->f[0] + 0.5f) ); + dst->f[1] = (float) floor( (double) (src->f[1] + 0.5f) ); + dst->f[2] = (float) floor( (double) (src->f[2] + 0.5f) ); + dst->f[3] = (float) floor( (double) (src->f[3] + 0.5f) ); +} + +static void +micro_shl( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] << src1->i[0]; + dst->i[1] = src0->i[1] << src1->i[1]; + dst->i[2] = src0->i[2] << src1->i[2]; + dst->i[3] = src0->i[3] << src1->i[3]; +} + +static void +micro_ishr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] >> src1->i[0]; + dst->i[1] = src0->i[1] >> src1->i[1]; + dst->i[2] = src0->i[2] >> src1->i[2]; + dst->i[3] = src0->i[3] >> src1->i[3]; +} + +static void +micro_trunc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0 ) +{ + dst->f[0] = (float) (int) src0->f[0]; + dst->f[1] = (float) (int) src0->f[1]; + dst->f[2] = (float) (int) src0->f[2]; + dst->f[3] = (float) (int) src0->f[3]; +} + +static void +micro_ushr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] >> src1->u[0]; + dst->u[1] = src0->u[1] >> src1->u[1]; + dst->u[2] = src0->u[2] >> src1->u[2]; + dst->u[3] = src0->u[3] >> src1->u[3]; +} + +static void +micro_sin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) sin( (double) src->f[0] ); + dst->f[1] = (float) sin( (double) src->f[1] ); + dst->f[2] = (float) sin( (double) src->f[2] ); + dst->f[3] = (float) sin( (double) src->f[3] ); +} + +static void +micro_sqrt( union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) sqrt( (double) src->f[0] ); + dst->f[1] = (float) sqrt( (double) src->f[1] ); + dst->f[2] = (float) sqrt( (double) src->f[2] ); + dst->f[3] = (float) sqrt( (double) src->f[3] ); +} + +static void +micro_sub( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] - src1->f[0]; + dst->f[1] = src0->f[1] - src1->f[1]; + dst->f[2] = src0->f[2] - src1->f[2]; + dst->f[3] = src0->f[3] - src1->f[3]; +} + +static void +micro_u2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) src->u[0]; + dst->f[1] = (float) src->u[1]; + dst->f[2] = (float) src->u[2]; + dst->f[3] = (float) src->u[3]; +} + +static void +micro_xor( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] ^ src1->u[0]; + dst->u[1] = src0->u[1] ^ src1->u[1]; + dst->u[2] = src0->u[2] ^ src1->u[2]; + dst->u[3] = src0->u[3] ^ src1->u[3]; +} + +static void +fetch_src_file_channel( + const struct tgsi_exec_machine *mach, + const uint file, + const uint swizzle, + const union tgsi_exec_channel *index, + union tgsi_exec_channel *chan ) +{ + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( file ) { + case TGSI_FILE_CONSTANT: + chan->f[0] = mach->Consts[index->i[0]][swizzle]; + chan->f[1] = mach->Consts[index->i[1]][swizzle]; + chan->f[2] = mach->Consts[index->i[2]][swizzle]; + chan->f[3] = mach->Consts[index->i[3]][swizzle]; + break; + + case TGSI_FILE_INPUT: + chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_TEMPORARY: + chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_IMMEDIATE: + assert( index->i[0] < (int) mach->ImmLimit ); + chan->f[0] = mach->Imms[index->i[0]][swizzle]; + assert( index->i[1] < (int) mach->ImmLimit ); + chan->f[1] = mach->Imms[index->i[1]][swizzle]; + assert( index->i[2] < (int) mach->ImmLimit ); + chan->f[2] = mach->Imms[index->i[2]][swizzle]; + assert( index->i[3] < (int) mach->ImmLimit ); + chan->f[3] = mach->Imms[index->i[3]][swizzle]; + break; + + case TGSI_FILE_ADDRESS: + chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_OUTPUT: + /* vertex/fragment output vars can be read too */ + chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; + break; + + case TGSI_EXTSWIZZLE_ONE: + *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; + break; + + default: + assert( 0 ); + } +} + +static void +fetch_source( + const struct tgsi_exec_machine *mach, + union tgsi_exec_channel *chan, + const struct tgsi_full_src_register *reg, + const uint chan_index ) +{ + union tgsi_exec_channel index; + uint swizzle; + + index.i[0] = + index.i[1] = + index.i[2] = + index.i[3] = reg->SrcRegister.Index; + + if (reg->SrcRegister.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + + if( reg->SrcRegister.Dimension ) { + switch( reg->SrcRegister.File ) { + case TGSI_FILE_INPUT: + index.i[0] *= 17; + index.i[1] *= 17; + index.i[2] *= 17; + index.i[3] *= 17; + break; + case TGSI_FILE_CONSTANT: + index.i[0] *= 4096; + index.i[1] *= 4096; + index.i[2] *= 4096; + index.i[3] *= 4096; + break; + default: + assert( 0 ); + } + + index.i[0] += reg->SrcRegisterDim.Index; + index.i[1] += reg->SrcRegisterDim.Index; + index.i[2] += reg->SrcRegisterDim.Index; + index.i[3] += reg->SrcRegisterDim.Index; + + if (reg->SrcRegisterDim.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterDimInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterDimInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + } + + swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + fetch_src_file_channel( + mach, + reg->SrcRegister.File, + swizzle, + &index, + chan ); + + switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { + case TGSI_UTIL_SIGN_CLEAR: + micro_abs( chan, chan ); + break; + + case TGSI_UTIL_SIGN_SET: + micro_abs( chan, chan ); + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } + + if (reg->SrcRegisterExtMod.Complement) { + micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan ); + } +} + +static void +store_dest( + struct tgsi_exec_machine *mach, + const union tgsi_exec_channel *chan, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + uint chan_index ) +{ + union tgsi_exec_channel *dst; + + switch( reg->DstRegister.File ) { + case TGSI_FILE_NULL: + return; + + case TGSI_FILE_OUTPUT: + dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + + reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_TEMPORARY: + dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_ADDRESS: + dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; + break; + + default: + assert( 0 ); + return; + } + + switch (inst->Instruction.Saturate) + { + case TGSI_SAT_NONE: + if (mach->ExecMask & 0x1) + dst->i[0] = chan->i[0]; + if (mach->ExecMask & 0x2) + dst->i[1] = chan->i[1]; + if (mach->ExecMask & 0x4) + dst->i[2] = chan->i[2]; + if (mach->ExecMask & 0x8) + dst->i[3] = chan->i[3]; + break; + + case TGSI_SAT_ZERO_ONE: + /* XXX need to obey ExecMask here */ + micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); + micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + + default: + assert( 0 ); + } +} + +#define FETCH(VAL,INDEX,CHAN)\ + fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) + +#define STORE(VAL,INDEX,CHAN)\ + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) + + +/** + * Execute ARB-style KIL which is predicated by a src register. + * Kill fragment if any of the four values is less than zero. + */ +static void +exec_kilp(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + uint uniquemask; + uint chan_index; + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + union tgsi_exec_channel r[1]; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + uint swizzle; + uint i; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle ( + &inst->FullSrcRegisters[0], + chan_index); + + /* check if the component has not been already tested */ + if (uniquemask & (1 << swizzle)) + continue; + uniquemask |= 1 << swizzle; + + FETCH(&r[0], 0, chan_index); + for (i = 0; i < 4; i++) + if (r[0].f[i] < 0.0f) + kilmask |= 1 << i; + } + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; +} + + +/* + * Fetch a texel using STR texture coordinates. + */ +static void +fetch_texel( struct tgsi_sampler *sampler, + const union tgsi_exec_channel *s, + const union tgsi_exec_channel *t, + const union tgsi_exec_channel *p, + float lodbias, /* XXX should be float[4] */ + union tgsi_exec_channel *r, + union tgsi_exec_channel *g, + union tgsi_exec_channel *b, + union tgsi_exec_channel *a ) +{ + uint j; + float rgba[NUM_CHANNELS][QUAD_SIZE]; + + sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); + + for (j = 0; j < 4; j++) { + r->f[j] = rgba[0][j]; + g->f[j] = rgba[1][j]; + b->f[j] = rgba[2][j]; + a->f[j] = rgba[3][j]; + } +} + + +static void +exec_tex(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + boolean biasLod) +{ + const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + union tgsi_exec_channel r[8]; + uint chan_index; + float lodBias; + + /* debug_printf("Sampler %u unit %u\n", sampler, unit); */ + + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: + + FETCH(&r[0], 0, CHAN_X); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[1], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[1] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[1], 0, CHAN_W); + lodBias = r[2].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ + &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ + break; + + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, /* inputs */ + &r[0], &r[1], &r[2], &r[3]); /* outputs */ + break; + + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { + case TGSI_EXTSWIZZLE_W: + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + break; + + case TGSI_EXTSWIZZLE_ONE: + break; + + default: + assert (0); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, + &r[0], &r[1], &r[2], &r[3]); + break; + + default: + assert (0); + } + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } +} + + +/** + * Evaluate a constant-valued coefficient at the position of the + * current quad. + */ +static void +eval_constant_coef( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; + } +} + +/** + * Evaluate a linear-valued coefficient at the position of the + * current quad. + */ +static void +eval_linear_coef( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + mach->Inputs[attrib].xyzw[chan].f[0] = a0; + mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; + mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; + mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; +} + +/** + * Evaluate a perspective-valued coefficient at the position of the + * current quad. + */ +static void +eval_perspective_coef( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + const float *w = mach->QuadPos.xyzw[3].f; + /* divide by W here */ + mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; + mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; + mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; + mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; +} + + +typedef void (* eval_coef_func)( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ); + +static void +exec_declaration( + struct tgsi_exec_machine *mach, + const struct tgsi_full_declaration *decl ) +{ + if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + eval_coef_func eval; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + switch( decl->Interpolation.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + eval = eval_constant_coef; + break; + + case TGSI_INTERPOLATE_LINEAR: + eval = eval_linear_coef; + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + eval = eval_perspective_coef; + break; + + default: + assert( 0 ); + } + + if( mask == TGSI_WRITEMASK_XYZW ) { + unsigned i, j; + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + eval( mach, i, j ); + } + } + } + else { + unsigned i, j; + + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + eval( mach, i, j ); + } + } + } + } + } + } +} + +static void +exec_instruction( + struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + int *pc ) +{ + uint chan_index; + union tgsi_exec_channel r[8]; + + (*pc)++; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_f2it( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_X ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[1], 0, CHAN_Y ); + micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + + FETCH( &r[2], 0, CHAN_W ); + micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); + micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); + micro_pow( &r[1], &r[1], &r[2] ); + micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Z ); + } + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( &r[0], 0, CHAN_X ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( &r[0], 0, CHAN_X ); + micro_sqrt( &r[0], &r[0] ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + assert (0); + break; + + case TGSI_OPCODE_LOG: + assert (0); + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_mul( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Z ); + FETCH( &r[2], 1, CHAN_Z ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_W); + FETCH(&r[2], 1, CHAN_W); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + FETCH( &r[0], 0, CHAN_Y ); + FETCH( &r[1], 1, CHAN_Y); + micro_mul( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_Z ); + STORE( &r[0], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + FETCH( &r[0], 1, CHAN_W ); + STORE( &r[0], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + /* XXX use micro_min()?? */ + micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + /* XXX use micro_max()?? */ + micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); + + STORE(&r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_mul( &r[0], &r[0], &r[1] ); + FETCH( &r[1], 2, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_sub( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_sub( &r[1], &r[1], &r[2] ); + micro_mul( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_CND: + assert (0); + break; + + case TGSI_OPCODE_CND0: + assert (0); + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert (0); + break; + + case TGSI_OPCODE_INDEX: + assert (0); + break; + + case TGSI_OPCODE_NEGATE: + assert (0); + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_frc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + assert (0); + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_flr( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_rnd( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH(&r[0], 0, CHAN_X); + + micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( &r[0], 0, CHAN_X ); + micro_lg2( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_pow( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + FETCH(&r[0], 0, CHAN_Y); + FETCH(&r[1], 1, CHAN_Z); + + micro_mul( &r[2], &r[0], &r[1] ); + + FETCH(&r[3], 0, CHAN_Z); + FETCH(&r[4], 1, CHAN_Y); + + micro_mul( &r[5], &r[3], &r[4] ); + micro_sub( &r[2], &r[2], &r[5] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &r[2], 0, CHAN_X ); + } + + FETCH(&r[2], 1, CHAN_X); + + micro_mul( &r[3], &r[3], &r[2] ); + + FETCH(&r[5], 0, CHAN_X); + + micro_mul( &r[1], &r[1], &r[5] ); + micro_sub( &r[3], &r[3], &r[1] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + STORE( &r[3], 0, CHAN_Y ); + } + + micro_mul( &r[5], &r[5], &r[4] ); + micro_mul( &r[0], &r[0], &r[2] ); + micro_sub( &r[5], &r[5], &r[0] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( &r[5], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + assert (0); + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + + micro_abs( &r[0], &r[0] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_RCC: + assert (0); + break; + + case TGSI_OPCODE_DPH: + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 1, CHAN_W); + + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH(&r[0], 0, CHAN_X); + + micro_cos( &r[0], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddx( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDY: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddy( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_KILP: + exec_kilp (mach, inst); + break; + + case TGSI_OPCODE_KIL: + /* for enabled ExecMask bits, set the killed bit */ + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; + break; + + case TGSI_OPCODE_PK2H: + assert (0); + break; + + case TGSI_OPCODE_PK2US: + assert (0); + break; + + case TGSI_OPCODE_PK4B: + assert (0); + break; + + case TGSI_OPCODE_PK4UB: + assert (0); + break; + + case TGSI_OPCODE_RFL: + assert (0); + break; + + case TGSI_OPCODE_SEQ: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_eq( &r[0], &r[0], &r[1], + &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], + &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SFL: + assert (0); + break; + + case TGSI_OPCODE_SGT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SIN: + FETCH( &r[0], 0, CHAN_X ); + micro_sin( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SNE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_STR: + assert (0); + break; + + case TGSI_OPCODE_TEX: + /* simple texture lookup */ + /* src[0] = texcoord */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, FALSE); + break; + + case TGSI_OPCODE_TXB: + /* Texture lookup with lod bias */ + /* src[0] = texcoord (src[0].w = LOD bias) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE); + break; + + case TGSI_OPCODE_TXD: + /* Texture lookup with explict partial derivatives */ + /* src[0] = texcoord */ + /* src[1] = d[strq]/dx */ + /* src[2] = d[strq]/dy */ + /* src[3] = sampler unit */ + assert (0); + break; + + case TGSI_OPCODE_TXL: + /* Texture lookup with explit LOD */ + /* src[0] = texcoord (src[0].w = LOD) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE); + break; + + case TGSI_OPCODE_UP2H: + assert (0); + break; + + case TGSI_OPCODE_UP2US: + assert (0); + break; + + case TGSI_OPCODE_UP4B: + assert (0); + break; + + case TGSI_OPCODE_UP4UB: + assert (0); + break; + + case TGSI_OPCODE_X2D: + assert (0); + break; + + case TGSI_OPCODE_ARA: + assert (0); + break; + + case TGSI_OPCODE_ARR: + assert (0); + break; + + case TGSI_OPCODE_BRA: + assert (0); + break; + + case TGSI_OPCODE_CAL: + /* skip the call if no execution channels are enabled */ + if (mach->ExecMask) { + /* do the call */ + + /* push the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + + assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); + mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; + + /* note that PC was already incremented above */ + mach->CallStack[mach->CallStackTop++] = *pc; + *pc = inst->InstructionExtLabel.Label; + } + break; + + case TGSI_OPCODE_RET: + mach->FuncMask &= ~mach->ExecMask; + UPDATE_EXEC_MASK(mach); + + if (mach->ExecMask == 0x0) { + /* really return now (otherwise, keep executing */ + + if (mach->CallStackTop == 0) { + /* returning from main() */ + *pc = -1; + return; + } + *pc = mach->CallStack[--mach->CallStackTop]; + + /* pop the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->FuncStackTop > 0); + mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; + + UPDATE_EXEC_MASK(mach); + } + break; + + case TGSI_OPCODE_SSG: + assert (0); + break; + + case TGSI_OPCODE_CMP: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_SCS: + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( &r[0], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + micro_cos( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + micro_sin( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_Y ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_NRM: + assert (0); + break; + + case TGSI_OPCODE_DIV: + assert( 0 ); + break; + + case TGSI_OPCODE_DP2: + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_IF: + /* push CondMask */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + FETCH( &r[0], 0, CHAN_X ); + /* update CondMask */ + if( ! r[0].u[0] ) { + mach->CondMask &= ~0x1; + } + if( ! r[0].u[1] ) { + mach->CondMask &= ~0x2; + } + if( ! r[0].u[2] ) { + mach->CondMask &= ~0x4; + } + if( ! r[0].u[3] ) { + mach->CondMask &= ~0x8; + } + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ELSE */ + break; + + case TGSI_OPCODE_ELSE: + /* invert CondMask wrt previous mask */ + { + uint prevMask; + assert(mach->CondStackTop > 0); + prevMask = mach->CondStack[mach->CondStackTop - 1]; + mach->CondMask = ~mach->CondMask & prevMask; + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ENDIF */ + } + break; + + case TGSI_OPCODE_ENDIF: + /* pop CondMask */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_END: + /* halt execution */ + *pc = -1; + break; + + case TGSI_OPCODE_REP: + assert (0); + break; + + case TGSI_OPCODE_ENDREP: + assert (0); + break; + + case TGSI_OPCODE_PUSHA: + assert (0); + break; + + case TGSI_OPCODE_POPA: + assert (0); + break; + + case TGSI_OPCODE_CEIL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ceil( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_I2F: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_i2f( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_NOT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_not( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TRUNC: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_trunc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_shl( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ishr( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_AND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_and( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_OR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_or( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOD: + assert (0); + break; + + case TGSI_OPCODE_XOR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_xor( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SAD: + assert (0); + break; + + case TGSI_OPCODE_TXF: + assert (0); + break; + + case TGSI_OPCODE_TXQ: + assert (0); + break; + + case TGSI_OPCODE_EMIT: + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; + break; + + case TGSI_OPCODE_ENDPRIM: + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; + break; + + case TGSI_OPCODE_LOOP: + /* fall-through (for now) */ + case TGSI_OPCODE_BGNLOOP2: + /* push LoopMask and ContMasks */ + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + break; + + case TGSI_OPCODE_ENDLOOP: + /* fall-through (for now at least) */ + case TGSI_OPCODE_ENDLOOP2: + /* Restore ContMask, but don't pop */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; + if (mach->LoopMask) { + /* repeat loop: jump to instruction just past BGNLOOP */ + *pc = inst->InstructionExtLabel.Label + 1; + } + else { + /* exit loop: pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + /* pop ContMask */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + } + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BRK: + /* turn off loop channels for each enabled exec channel */ + mach->LoopMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_CONT: + /* turn off cont channels for each enabled exec channel */ + mach->ContMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BGNSUB: + /* no-op */ + break; + + case TGSI_OPCODE_ENDSUB: + /* no-op */ + break; + + case TGSI_OPCODE_NOISE1: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE2: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE3: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE4: + assert( 0 ); + break; + + case TGSI_OPCODE_NOP: + break; + + default: + assert( 0 ); + } +} + + +/** + * Run TGSI interpreter. + * \return bitmask of "alive" quad components + */ +uint +tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) +{ + uint i; + int pc = 0; + + mach->CondMask = 0xf; + mach->LoopMask = 0xf; + mach->ContMask = 0xf; + mach->FuncMask = 0xf; + mach->ExecMask = 0xf; + + mach->CondStackTop = 0; /* temporarily subvert this assertion */ + assert(mach->CondStackTop == 0); + assert(mach->LoopStackTop == 0); + assert(mach->ContStackTop == 0); + assert(mach->CallStackTop == 0); + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; + + if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; + mach->Primitives[0] = 0; + } + + + /* execute declarations (interpolants) */ + for (i = 0; i < mach->NumDeclarations; i++) { + exec_declaration( mach, mach->Declarations+i ); + } + + /* execute instructions, until pc is set to -1 */ + while (pc != -1) { + assert(pc < mach->NumInstructions); + exec_instruction( mach, mach->Instructions + pc, &pc ); + } + +#if 0 + /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ + if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { + /* + * Scale back depth component. + */ + for (i = 0; i < 4; i++) + mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; + } +#endif + + return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; +} + + diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h new file mode 100644 index 0000000000..1fb66ee960 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h @@ -0,0 +1,239 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#if !defined TGSI_EXEC_H +#define TGSI_EXEC_H + +#include "pipe/p_compiler.h" + +#if defined __cplusplus +extern "C" { +#endif + +#define NUM_CHANNELS 4 /* R,G,B,A */ +#define QUAD_SIZE 4 /* 4 pixel/quad */ + +/** + * Registers may be treated as float, signed int or unsigned int. + */ +union tgsi_exec_channel +{ + float f[QUAD_SIZE]; + int i[QUAD_SIZE]; + unsigned u[QUAD_SIZE]; +}; + +/** + * A vector[RGBA] of channels[4 pixels] + */ +struct tgsi_exec_vector +{ + union tgsi_exec_channel xyzw[NUM_CHANNELS]; +}; + +/** + * For fragment programs, information for computing fragment input + * values from plane equation of the triangle/line. + */ +struct tgsi_interp_coef +{ + float a0[NUM_CHANNELS]; /* in an xyzw layout */ + float dadx[NUM_CHANNELS]; + float dady[NUM_CHANNELS]; +}; + + +struct softpipe_tile_cache; /**< Opaque to TGSI */ + +/** + * Information for sampling textures, which must be implemented + * by code outside the TGSI executor. + */ +struct tgsi_sampler +{ + const struct pipe_sampler_state *state; + struct pipe_texture *texture; + /** Get samples for four fragments in a quad */ + void (*get_samples)(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); + void *pipe; /*XXX temporary*/ + struct softpipe_tile_cache *cache; +}; + +/** + * For branching/calling subroutines. + */ +struct tgsi_exec_labels +{ + unsigned labels[128][2]; + unsigned count; +}; + +/* + * Locations of various utility registers (_I = Index, _C = Channel) + */ +#define TGSI_EXEC_TEMP_00000000_I 32 +#define TGSI_EXEC_TEMP_00000000_C 0 + +#define TGSI_EXEC_TEMP_7FFFFFFF_I 32 +#define TGSI_EXEC_TEMP_7FFFFFFF_C 1 + +#define TGSI_EXEC_TEMP_80000000_I 32 +#define TGSI_EXEC_TEMP_80000000_C 2 + +#define TGSI_EXEC_TEMP_FFFFFFFF_I 32 +#define TGSI_EXEC_TEMP_FFFFFFFF_C 3 + +#define TGSI_EXEC_TEMP_ONE_I 33 +#define TGSI_EXEC_TEMP_ONE_C 0 + +#define TGSI_EXEC_TEMP_TWO_I 33 +#define TGSI_EXEC_TEMP_TWO_C 1 + +#define TGSI_EXEC_TEMP_128_I 33 +#define TGSI_EXEC_TEMP_128_C 2 + +#define TGSI_EXEC_TEMP_MINUS_128_I 33 +#define TGSI_EXEC_TEMP_MINUS_128_C 3 + +#define TGSI_EXEC_TEMP_KILMASK_I 34 +#define TGSI_EXEC_TEMP_KILMASK_C 0 + +#define TGSI_EXEC_TEMP_OUTPUT_I 34 +#define TGSI_EXEC_TEMP_OUTPUT_C 1 + +#define TGSI_EXEC_TEMP_PRIMITIVE_I 34 +#define TGSI_EXEC_TEMP_PRIMITIVE_C 2 + +#define TGSI_EXEC_TEMP_R0 35 + +#define TGSI_EXEC_NUM_TEMPS (32 + 4) +#define TGSI_EXEC_NUM_ADDRS 1 +#define TGSI_EXEC_NUM_IMMEDIATES 256 + +#define TGSI_EXEC_MAX_COND_NESTING 10 +#define TGSI_EXEC_MAX_LOOP_NESTING 10 +#define TGSI_EXEC_MAX_CALL_NESTING 10 + +/** + * Run-time virtual machine state for executing TGSI shader. + */ +struct tgsi_exec_machine +{ + /* + * 32 program temporaries + * 4 internal temporaries + * 1 address + * 1 temporary of padding to align to 16 bytes + */ + struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS + 1]; + + /* + * This will point to _Temps after aligning to 16B boundary. + */ + struct tgsi_exec_vector *Temps; + struct tgsi_exec_vector *Addrs; + + struct tgsi_sampler *Samplers; + + float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; + unsigned ImmLimit; + float (*Consts)[4]; + struct tgsi_exec_vector *Inputs; + struct tgsi_exec_vector *Outputs; + const struct tgsi_token *Tokens; + unsigned Processor; + + /* GEOMETRY processor only. */ + unsigned *Primitives; + + /* FRAGMENT processor only. */ + const struct tgsi_interp_coef *InterpCoefs; + struct tgsi_exec_vector QuadPos; + + /* Conditional execution masks */ + uint CondMask; /**< For IF/ELSE/ENDIF */ + uint LoopMask; /**< For BGNLOOP/ENDLOOP */ + uint ContMask; /**< For loop CONT statements */ + uint FuncMask; /**< For function calls */ + uint ExecMask; /**< = CondMask & LoopMask */ + + /** Condition mask stack (for nested conditionals) */ + uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; + int CondStackTop; + + /** Loop mask stack (for nested loops) */ + uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopStackTop; + + /** Loop continue mask stack (see comments in tgsi_exec.c) */ + uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int ContStackTop; + + /** Function execution mask stack (for executing subroutine code) */ + uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; + int FuncStackTop; + + /** Function call stack for saving/restoring the program counter */ + uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; + int CallStackTop; + + struct tgsi_full_instruction *Instructions; + uint NumInstructions; + + struct tgsi_full_declaration *Declarations; + uint NumDeclarations; + + struct tgsi_exec_labels Labels; +}; + + +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach, + const struct tgsi_token *tokens, + unsigned numSamplers, + struct tgsi_sampler *samplers); + +uint +tgsi_exec_machine_run( + struct tgsi_exec_machine *mach ); + + +void +tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach); + + +#if defined __cplusplus +} /* extern "C" */ +#endif + +#endif /* TGSI_EXEC_H */ diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c new file mode 100755 index 0000000000..593464db3e --- /dev/null +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -0,0 +1,2378 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_util.h" +#include "tgsi_exec.h" +#include "tgsi_sse2.h" + +#include "x86/rtasm/x86sse.h" + +#if defined(__i386__) || defined(__386__) + +#define DUMP_SSE 0 + +#if DUMP_SSE + +static void +_print_reg( + struct x86_reg reg ) +{ + if (reg.mod != mod_REG) + debug_printf( "[" ); + + switch( reg.file ) { + case file_REG32: + switch( reg.idx ) { + case reg_AX: + debug_printf( "EAX" ); + break; + case reg_CX: + debug_printf( "ECX" ); + break; + case reg_DX: + debug_printf( "EDX" ); + break; + case reg_BX: + debug_printf( "EBX" ); + break; + case reg_SP: + debug_printf( "ESP" ); + break; + case reg_BP: + debug_printf( "EBP" ); + break; + case reg_SI: + debug_printf( "ESI" ); + break; + case reg_DI: + debug_printf( "EDI" ); + break; + } + break; + case file_MMX: + assert( 0 ); + break; + case file_XMM: + debug_printf( "XMM%u", reg.idx ); + break; + case file_x87: + assert( 0 ); + break; + } + + if (reg.mod == mod_DISP8 || + reg.mod == mod_DISP32) + debug_printf("+%d", reg.disp); + + if (reg.mod != mod_REG) + debug_printf( "]" ); +} + +static void +_fill( + const char *op ) +{ + unsigned count = 10 - strlen( op ); + + while( count-- ) { + debug_printf( " " ); + } +} + +#define DUMP_START() debug_printf( "\nsse-dump start ----------------" ) +#define DUMP_END() debug_printf( "\nsse-dump end ----------------\n" ) +#define DUMP( OP ) debug_printf( "\n%s", OP ) +#define DUMP_I( OP, I ) do {\ + debug_printf( "\n%s", OP );\ + _fill( OP );\ + debug_printf( "%u", I ); } while( 0 ) +#define DUMP_R( OP, R0 ) do {\ + debug_printf( "\n%s", OP );\ + _fill( OP );\ + _print_reg( R0 ); } while( 0 ) +#define DUMP_RR( OP, R0, R1 ) do {\ + debug_printf( "\n%s", OP );\ + _fill( OP );\ + _print_reg( R0 );\ + debug_printf( ", " );\ + _print_reg( R1 ); } while( 0 ) +#define DUMP_RRI( OP, R0, R1, I ) do {\ + debug_printf( "\n%s", OP );\ + _fill( OP );\ + _print_reg( R0 );\ + debug_printf( ", " );\ + _print_reg( R1 );\ + debug_printf( ", " );\ + debug_printf( "%u", I ); } while( 0 ) + +#else + +#define DUMP_START() +#define DUMP_END() +#define DUMP( OP ) +#define DUMP_I( OP, I ) +#define DUMP_R( OP, R0 ) +#define DUMP_RR( OP, R0, R1 ) +#define DUMP_RRI( OP, R0, R1, I ) + +#endif + +#define FOR_EACH_CHANNEL( CHAN )\ + for( CHAN = 0; CHAN < 4; CHAN++ ) + +#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ + if( IS_DST0_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\ + FOR_EACH_CHANNEL( CHAN )\ + IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN ) + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +/** + * X86 utility functions. + */ + +static struct x86_reg +make_xmm( + unsigned xmm ) +{ + return x86_make_reg( + file_XMM, + (enum x86_reg_name) xmm ); +} + +/** + * X86 register mapping helpers. + */ + +static struct x86_reg +get_const_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_CX ); +} + +static struct x86_reg +get_input_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_AX ); +} + +static struct x86_reg +get_output_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_DX ); +} + +static struct x86_reg +get_temp_base( void ) +{ +#ifdef WIN32 + return x86_make_reg( + file_REG32, + reg_BX ); +#else + return x86_make_reg( + file_REG32, + reg_SI ); +#endif +} + +static struct x86_reg +get_coef_base( void ) +{ + return get_output_base(); +} + +/** + * Data access helpers. + */ + +static struct x86_reg +get_argument( + unsigned index ) +{ + return x86_make_disp( + x86_make_reg( file_REG32, reg_SP ), + (index + 1) * 4 ); +} + +static struct x86_reg +get_const( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_const_base(), + (vec * 4 + chan) * 4 ); +} + +static struct x86_reg +get_input( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_input_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_output( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_output_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_temp( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_temp_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_coef( + unsigned vec, + unsigned chan, + unsigned member ) +{ + return x86_make_disp( + get_coef_base(), + ((vec * 3 + member) * 4 + chan) * 4 ); +} + +/** + * X86 rtasm wrappers. + */ + +static void +emit_addps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ADDPS", dst, src ); + sse_addps( func, dst, src ); +} + +static void +emit_andnps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ANDNPS", dst, src ); + sse_andnps( func, dst, src ); +} + +static void +emit_andps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ANDPS", dst, src ); + sse_andps( func, dst, src ); +} + +static void +emit_call( + struct x86_function *func, + void (* addr)() ) +{ + struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); + + DUMP_I( "CALL", addr ); + x86_mov_reg_imm( func, ecx, (unsigned long) addr ); + x86_call( func, ecx ); +} + +static void +emit_cmpps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src, + enum sse_cc cc ) +{ + DUMP_RRI( "CMPPS", dst, src, cc ); + sse_cmpps( func, dst, src, cc ); +} + +static void +emit_cvttps2dq( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "CVTTPS2DQ", dst, src ); + sse2_cvttps2dq( func, dst, src ); +} + +static void +emit_maxps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MAXPS", dst, src ); + sse_maxps( func, dst, src ); +} + +static void +emit_minps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MINPS", dst, src ); + sse_minps( func, dst, src ); +} + +static void +emit_mov( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOV", dst, src ); + x86_mov( func, dst, src ); +} + +static void +emit_movaps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOVAPS", dst, src ); + sse_movaps( func, dst, src ); +} + +static void +emit_movss( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOVSS", dst, src ); + sse_movss( func, dst, src ); +} + +static void +emit_movups( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MOVUPS", dst, src ); + sse_movups( func, dst, src ); +} + +static void +emit_mulps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "MULPS", dst, src ); + sse_mulps( func, dst, src ); +} + +static void +emit_or( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "OR", dst, src ); + x86_or( func, dst, src ); +} + +static void +emit_orps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "ORPS", dst, src ); + sse_orps( func, dst, src ); +} + +static void +emit_pmovmskb( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "PMOVMSKB", dst, src ); + sse_pmovmskb( func, dst, src ); +} + +static void +emit_pop( + struct x86_function *func, + struct x86_reg dst ) +{ + DUMP_R( "POP", dst ); + x86_pop( func, dst ); +} + +static void +emit_push( + struct x86_function *func, + struct x86_reg dst ) +{ + DUMP_R( "PUSH", dst ); + x86_push( func, dst ); +} + +static void +emit_rcpps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "RCPPS", dst, src ); + sse2_rcpps( func, dst, src ); +} + +#ifdef WIN32 +static void +emit_retw( + struct x86_function *func, + unsigned size ) +{ + DUMP_I( "RET", size ); + x86_retw( func, size ); +} +#else +static void +emit_ret( + struct x86_function *func ) +{ + DUMP( "RET" ); + x86_ret( func ); +} +#endif + +static void +emit_rsqrtps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "RSQRTPS", dst, src ); + sse_rsqrtps( func, dst, src ); +} + +static void +emit_shufps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src, + unsigned char shuf ) +{ + DUMP_RRI( "SHUFPS", dst, src, shuf ); + sse_shufps( func, dst, src, shuf ); +} + +static void +emit_subps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "SUBPS", dst, src ); + sse_subps( func, dst, src ); +} + +static void +emit_xorps( + struct x86_function *func, + struct x86_reg dst, + struct x86_reg src ) +{ + DUMP_RR( "XORPS", dst, src ); + sse_xorps( func, dst, src ); +} + +/** + * Data fetch helpers. + */ + +static void +emit_const( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movss( + func, + make_xmm( xmm ), + get_const( vec, chan ) ); + emit_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + +static void +emit_inputf( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movups( + func, + make_xmm( xmm ), + get_input( vec, chan ) ); +} + +static void +emit_output( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movups( + func, + get_output( vec, chan ), + make_xmm( xmm ) ); +} + +static void +emit_tempf( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movaps( + func, + make_xmm( xmm ), + get_temp( vec, chan ) ); +} + +static void +emit_coef( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan, + unsigned member ) +{ + emit_movss( + func, + make_xmm( xmm ), + get_coef( vec, chan, member ) ); + emit_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + +/** + * Data store helpers. + */ + +static void +emit_inputs( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movups( + func, + get_input( vec, chan ), + make_xmm( xmm ) ); +} + +static void +emit_temps( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movaps( + func, + get_temp( vec, chan ), + make_xmm( xmm ) ); +} + +static void +emit_addrs( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_temps( + func, + xmm, + vec + TGSI_EXEC_NUM_TEMPS, + chan ); +} + +/** + * Coefficent fetch helpers. + */ + +static void +emit_coef_a0( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 0 ); +} + +static void +emit_coef_dadx( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 1 ); +} + +static void +emit_coef_dady( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 2 ); +} + +/** + * Function call helpers. + */ + +static void +emit_push_gp( + struct x86_function *func ) +{ + emit_push( + func, + get_const_base() ); + emit_push( + func, + get_input_base() ); + emit_push( + func, + get_output_base() ); + + /* It is important on non-win32 platforms that temp base is pushed last. + */ + emit_push( + func, + get_temp_base() ); +} + +static void +emit_pop_gp( + struct x86_function *func ) +{ + /* Restore GP registers in a reverse order. + */ + emit_pop( + func, + get_temp_base() ); + emit_pop( + func, + get_output_base() ); + emit_pop( + func, + get_input_base() ); + emit_pop( + func, + get_const_base() ); +} + +static void +emit_func_call_dst( + struct x86_function *func, + unsigned xmm_dst, + void (*code)() ) +{ + emit_movaps( + func, + get_temp( TEMP_R0, 0 ), + make_xmm( xmm_dst ) ); + + emit_push_gp( + func ); + +#ifdef WIN32 + emit_push( + func, + get_temp( TEMP_R0, 0 ) ); +#endif + + emit_call( + func, + code ); + + emit_pop_gp( + func ); + + emit_movaps( + func, + make_xmm( xmm_dst ), + get_temp( TEMP_R0, 0 ) ); +} + +static void +emit_func_call_dst_src( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src, + void (*code)() ) +{ + emit_movaps( + func, + get_temp( TEMP_R0, 1 ), + make_xmm( xmm_src ) ); + + emit_func_call_dst( + func, + xmm_dst, + code ); +} + +/** + * Low-level instruction translators. + */ + +static void +emit_abs( + struct x86_function *func, + unsigned xmm ) +{ + emit_andps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_7FFFFFFF_I, + TGSI_EXEC_TEMP_7FFFFFFF_C ) ); +} + +static void +emit_add( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_addps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void XSTDCALL +cos4f( + float *store ) +{ +#ifdef WIN32 + store[0] = (float) cos( (double) store[0] ); + store[1] = (float) cos( (double) store[1] ); + store[2] = (float) cos( (double) store[2] ); + store[3] = (float) cos( (double) store[3] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = cosf( store[X + 0] ); + store[X + 1] = cosf( store[X + 1] ); + store[X + 2] = cosf( store[X + 2] ); + store[X + 3] = cosf( store[X + 3] ); +#endif +} + +static void +emit_cos( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + cos4f ); +} + +static void XSTDCALL +ex24f( + float *store ) +{ +#ifdef WIN32 + store[0] = (float) pow( 2.0, (double) store[0] ); + store[1] = (float) pow( 2.0, (double) store[1] ); + store[2] = (float) pow( 2.0, (double) store[2] ); + store[3] = (float) pow( 2.0, (double) store[3] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = powf( 2.0f, store[X + 0] ); + store[X + 1] = powf( 2.0f, store[X + 1] ); + store[X + 2] = powf( 2.0f, store[X + 2] ); + store[X + 3] = powf( 2.0f, store[X + 3] ); +#endif +} + +static void +emit_ex2( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + ex24f ); +} + +static void +emit_f2it( + struct x86_function *func, + unsigned xmm ) +{ + emit_cvttps2dq( + func, + make_xmm( xmm ), + make_xmm( xmm ) ); +} + +static void XSTDCALL +flr4f( + float *store ) +{ +#ifdef WIN32 + const unsigned X = 0; +#else + const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] = (float) floor( (double) store[X + 0] ); + store[X + 1] = (float) floor( (double) store[X + 1] ); + store[X + 2] = (float) floor( (double) store[X + 2] ); + store[X + 3] = (float) floor( (double) store[X + 3] ); +} + +static void +emit_flr( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + flr4f ); +} + +static void XSTDCALL +frc4f( + float *store ) +{ +#ifdef WIN32 + const unsigned X = 0; +#else + const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] -= (float) floor( (double) store[X + 0] ); + store[X + 1] -= (float) floor( (double) store[X + 1] ); + store[X + 2] -= (float) floor( (double) store[X + 2] ); + store[X + 3] -= (float) floor( (double) store[X + 3] ); +} + +static void +emit_frc( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + frc4f ); +} + +static void XSTDCALL +lg24f( + float *store ) +{ +#ifdef WIN32 + const unsigned X = 0; +#else + const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] = LOG2( store[X + 0] ); + store[X + 1] = LOG2( store[X + 1] ); + store[X + 2] = LOG2( store[X + 2] ); + store[X + 3] = LOG2( store[X + 3] ); +} + +static void +emit_lg2( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + lg24f ); +} + +static void +emit_MOV( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_movups( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_mul (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + emit_mulps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_neg( + struct x86_function *func, + unsigned xmm ) +{ + emit_xorps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); +} + +static void XSTDCALL +pow4f( + float *store ) +{ +#ifdef WIN32 + store[0] = (float) pow( (double) store[0], (double) store[4] ); + store[1] = (float) pow( (double) store[1], (double) store[5] ); + store[2] = (float) pow( (double) store[2], (double) store[6] ); + store[3] = (float) pow( (double) store[3], (double) store[7] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = powf( store[X + 0], store[X + 4] ); + store[X + 1] = powf( store[X + 1], store[X + 5] ); + store[X + 2] = powf( store[X + 2], store[X + 6] ); + store[X + 3] = powf( store[X + 3], store[X + 7] ); +#endif +} + +static void +emit_pow( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_func_call_dst_src( + func, + xmm_dst, + xmm_src, + pow4f ); +} + +static void +emit_rcp ( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_rcpps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_rsqrt( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_rsqrtps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_setsign( + struct x86_function *func, + unsigned xmm ) +{ + emit_orps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); +} + +static void XSTDCALL +sin4f( + float *store ) +{ +#ifdef WIN32 + store[0] = (float) sin( (double) store[0] ); + store[1] = (float) sin( (double) store[1] ); + store[2] = (float) sin( (double) store[2] ); + store[3] = (float) sin( (double) store[3] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = sinf( store[X + 0] ); + store[X + 1] = sinf( store[X + 1] ); + store[X + 2] = sinf( store[X + 2] ); + store[X + 3] = sinf( store[X + 3] ); +#endif +} + +static void +emit_sin (struct x86_function *func, + unsigned xmm_dst) +{ + emit_func_call_dst( + func, + xmm_dst, + sin4f ); +} + +static void +emit_sub( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_subps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +/** + * Register fetch. + */ + +static void +emit_fetch( + struct x86_function *func, + unsigned xmm, + const struct tgsi_full_src_register *reg, + const unsigned chan_index ) +{ + unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( reg->SrcRegister.File ) { + case TGSI_FILE_CONSTANT: + emit_const( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + case TGSI_FILE_INPUT: + emit_inputf( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + case TGSI_FILE_TEMPORARY: + emit_tempf( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + emit_tempf( + func, + xmm, + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ); + break; + + case TGSI_EXTSWIZZLE_ONE: + emit_tempf( + func, + xmm, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + break; + + default: + assert( 0 ); + } + + switch( tgsi_util_get_full_src_register_sign_mode( reg, chan_index ) ) { + case TGSI_UTIL_SIGN_CLEAR: + emit_abs( func, xmm ); + break; + + case TGSI_UTIL_SIGN_SET: + emit_setsign( func, xmm ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + emit_neg( func, xmm ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } +} + +#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\ + emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN ) + +/** + * Register store. + */ + +static void +emit_store( + struct x86_function *func, + unsigned xmm, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + unsigned chan_index ) +{ + switch( reg->DstRegister.File ) { + case TGSI_FILE_OUTPUT: + emit_output( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; + + case TGSI_FILE_TEMPORARY: + emit_temps( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; + + case TGSI_FILE_ADDRESS: + emit_addrs( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; + + default: + assert( 0 ); + } + + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + + case TGSI_SAT_ZERO_ONE: +// assert( 0 ); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + } +} + +#define STORE( FUNC, INST, XMM, INDEX, CHAN )\ + emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) + +/** + * High-level instruction translators. + */ + +static void +emit_kil( + struct x86_function *func, + const struct tgsi_full_src_register *reg ) +{ + unsigned uniquemask; + unsigned registers[4]; + unsigned nextregister = 0; + unsigned firstchan = ~0; + unsigned chan_index; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + FOR_EACH_CHANNEL( chan_index ) { + unsigned swizzle; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle( + reg, + chan_index ); + + /* check if the component has not been already tested */ + if( !(uniquemask & (1 << swizzle)) ) { + uniquemask |= 1 << swizzle; + + /* allocate register */ + registers[chan_index] = nextregister; + emit_fetch( + func, + nextregister, + reg, + chan_index ); + nextregister++; + + /* mark the first channel used */ + if( firstchan == ~0 ) { + firstchan = chan_index; + } + } + } + + emit_push( + func, + x86_make_reg( file_REG32, reg_AX ) ); + emit_push( + func, + x86_make_reg( file_REG32, reg_DX ) ); + + FOR_EACH_CHANNEL( chan_index ) { + if( uniquemask & (1 << chan_index) ) { + emit_cmpps( + func, + make_xmm( registers[chan_index] ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ), + cc_LessThan ); + + if( chan_index == firstchan ) { + emit_pmovmskb( + func, + x86_make_reg( file_REG32, reg_AX ), + make_xmm( registers[chan_index] ) ); + } + else { + emit_pmovmskb( + func, + x86_make_reg( file_REG32, reg_DX ), + make_xmm( registers[chan_index] ) ); + emit_or( + func, + x86_make_reg( file_REG32, reg_AX ), + x86_make_reg( file_REG32, reg_DX ) ); + } + } + } + + emit_or( + func, + get_temp( + TGSI_EXEC_TEMP_KILMASK_I, + TGSI_EXEC_TEMP_KILMASK_C ), + x86_make_reg( file_REG32, reg_AX ) ); + + emit_pop( + func, + x86_make_reg( file_REG32, reg_DX ) ); + emit_pop( + func, + x86_make_reg( file_REG32, reg_AX ) ); +} + +static void +emit_setcc( + struct x86_function *func, + struct tgsi_full_instruction *inst, + enum sse_cc cc ) +{ + unsigned chan_index; + + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_cmpps( + func, + make_xmm( 0 ), + make_xmm( 1 ), + cc ); + emit_andps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ) ); + STORE( func, *inst, 0, 0, chan_index ); + } +} + +static void +emit_cmp( + struct x86_function *func, + struct tgsi_full_instruction *inst ) +{ + unsigned chan_index; + + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_cmpps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ), + cc_LessThan ); + emit_andps( + func, + make_xmm( 1 ), + make_xmm( 0 ) ); + emit_andnps( + func, + make_xmm( 0 ), + make_xmm( 2 ) ); + emit_orps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } +} + +static int +emit_instruction( + struct x86_function *func, + struct tgsi_full_instruction *inst ) +{ + unsigned chan_index; + + switch( inst->Instruction.Opcode ) { + case TGSI_OPCODE_ARL: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_f2it( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + /* TGSI_OPCODE_SWZ */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C); + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + STORE( func, *inst, 0, 0, CHAN_X ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( func, *inst, 0, 0, CHAN_W ); + } + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_maxps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ) ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 1, 0, CHAN_Y ); + emit_maxps( + func, + make_xmm( 1 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ) ); + FETCH( func, *inst, 2, 0, CHAN_W ); + emit_minps( + func, + make_xmm( 2 ), + get_temp( + TGSI_EXEC_TEMP_128_I, + TGSI_EXEC_TEMP_128_C ) ); + emit_maxps( + func, + make_xmm( 2 ), + get_temp( + TGSI_EXEC_TEMP_MINUS_128_I, + TGSI_EXEC_TEMP_MINUS_128_C ) ); + emit_pow( func, 1, 2 ); + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_xorps( + func, + make_xmm( 2 ), + make_xmm( 2 ) ); + emit_cmpps( + func, + make_xmm( 2 ), + make_xmm( 0 ), + cc_LessThanEqual ); + emit_andps( + func, + make_xmm( 2 ), + make_xmm( 1 ) ); + STORE( func, *inst, 2, 0, CHAN_Z ); + } + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_rcp( func, 0, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_rsqrt( func, 0, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + return 0; + break; + + case TGSI_OPCODE_LOG: + return 0; + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_mul( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_add( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul(func, 1, 2 ); + emit_add(func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_W ); + FETCH( func, *inst, 2, 1, CHAN_W ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + FETCH( func, *inst, 1, 1, CHAN_Y ); + emit_mul( func, 0, 1 ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + FETCH( func, *inst, 0, 0, CHAN_Z ); + STORE( func, *inst, 0, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, 1, CHAN_W ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_minps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_maxps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + emit_setcc( func, inst, cc_LessThan ); + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + emit_setcc( func, inst, cc_NotLessThan ); + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_mul( func, 0, 1 ); + emit_add( func, 0, 2 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_sub( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_sub( func, 1, 2 ); + emit_mul( func, 0, 1 ); + emit_add( func, 0, 2 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CND: + return 0; + break; + + case TGSI_OPCODE_CND0: + return 0; + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + return 0; + break; + + case TGSI_OPCODE_INDEX: + return 0; + break; + + case TGSI_OPCODE_NEGATE: + return 0; + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_frc( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + return 0; + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_flr( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + return 0; + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_ex2( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_lg2( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_pow( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( func, *inst, 1, 1, CHAN_Z ); + FETCH( func, *inst, 3, 0, CHAN_Z ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + FETCH( func, *inst, 4, 1, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + emit_MOV( func, 2, 0 ); + emit_mul( func, 2, 1 ); + emit_MOV( func, 5, 3 ); + emit_mul( func, 5, 4 ); + emit_sub( func, 2, 5 ); + STORE( func, *inst, 2, 0, CHAN_X ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 2, 1, CHAN_X ); + FETCH( func, *inst, 5, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + emit_mul( func, 3, 2 ); + emit_mul( func, 1, 5 ); + emit_sub( func, 3, 1 ); + STORE( func, *inst, 3, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + emit_mul( func, 5, 4 ); + emit_mul( func, 0, 2 ); + emit_sub( func, 5, 0 ); + STORE( func, *inst, 5, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + return 0; + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_abs( func, 0) ; + + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_RCC: + return 0; + break; + + case TGSI_OPCODE_DPH: + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 1, CHAN_W ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_cos( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + return 0; + break; + + case TGSI_OPCODE_DDY: + return 0; + break; + + case TGSI_OPCODE_KIL: + emit_kil( func, &inst->FullSrcRegisters[0] ); + break; + + case TGSI_OPCODE_PK2H: + return 0; + break; + + case TGSI_OPCODE_PK2US: + return 0; + break; + + case TGSI_OPCODE_PK4B: + return 0; + break; + + case TGSI_OPCODE_PK4UB: + return 0; + break; + + case TGSI_OPCODE_RFL: + return 0; + break; + + case TGSI_OPCODE_SEQ: + return 0; + break; + + case TGSI_OPCODE_SFL: + return 0; + break; + + case TGSI_OPCODE_SGT: + return 0; + break; + + case TGSI_OPCODE_SIN: + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_sin( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + return 0; + break; + + case TGSI_OPCODE_SNE: + return 0; + break; + + case TGSI_OPCODE_STR: + return 0; + break; + + case TGSI_OPCODE_TEX: + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_TXD: + return 0; + break; + + case TGSI_OPCODE_UP2H: + return 0; + break; + + case TGSI_OPCODE_UP2US: + return 0; + break; + + case TGSI_OPCODE_UP4B: + return 0; + break; + + case TGSI_OPCODE_UP4UB: + return 0; + break; + + case TGSI_OPCODE_X2D: + return 0; + break; + + case TGSI_OPCODE_ARA: + return 0; + break; + + case TGSI_OPCODE_ARR: + return 0; + break; + + case TGSI_OPCODE_BRA: + return 0; + break; + + case TGSI_OPCODE_CAL: + return 0; + break; + + case TGSI_OPCODE_RET: + case TGSI_OPCODE_END: +#ifdef WIN32 + emit_retw( func, 16 ); +#else + emit_ret( func ); +#endif + break; + + case TGSI_OPCODE_SSG: + return 0; + break; + + case TGSI_OPCODE_CMP: + emit_cmp (func, inst); + break; + + case TGSI_OPCODE_SCS: + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_cos( func, 0 ); + STORE( func, *inst, 0, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + emit_sin( func, 0 ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + FETCH( func, *inst, 0, TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ); + STORE( func, *inst, 0, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_TXB: + return 0; + break; + + case TGSI_OPCODE_NRM: + return 0; + break; + + case TGSI_OPCODE_DIV: + return 0; + break; + + case TGSI_OPCODE_DP2: + return 0; + break; + + case TGSI_OPCODE_TXL: + return 0; + break; + + case TGSI_OPCODE_BRK: + return 0; + break; + + case TGSI_OPCODE_IF: + return 0; + break; + + case TGSI_OPCODE_LOOP: + return 0; + break; + + case TGSI_OPCODE_REP: + return 0; + break; + + case TGSI_OPCODE_ELSE: + return 0; + break; + + case TGSI_OPCODE_ENDIF: + return 0; + break; + + case TGSI_OPCODE_ENDLOOP: + return 0; + break; + + case TGSI_OPCODE_ENDREP: + return 0; + break; + + case TGSI_OPCODE_PUSHA: + return 0; + break; + + case TGSI_OPCODE_POPA: + return 0; + break; + + case TGSI_OPCODE_CEIL: + return 0; + break; + + case TGSI_OPCODE_I2F: + return 0; + break; + + case TGSI_OPCODE_NOT: + return 0; + break; + + case TGSI_OPCODE_TRUNC: + return 0; + break; + + case TGSI_OPCODE_SHL: + return 0; + break; + + case TGSI_OPCODE_SHR: + return 0; + break; + + case TGSI_OPCODE_AND: + return 0; + break; + + case TGSI_OPCODE_OR: + return 0; + break; + + case TGSI_OPCODE_MOD: + return 0; + break; + + case TGSI_OPCODE_XOR: + return 0; + break; + + case TGSI_OPCODE_SAD: + return 0; + break; + + case TGSI_OPCODE_TXF: + return 0; + break; + + case TGSI_OPCODE_TXQ: + return 0; + break; + + case TGSI_OPCODE_CONT: + return 0; + break; + + case TGSI_OPCODE_EMIT: + return 0; + break; + + case TGSI_OPCODE_ENDPRIM: + return 0; + break; + + default: + return 0; + } + + return 1; +} + +static void +emit_declaration( + struct x86_function *func, + struct tgsi_full_declaration *decl ) +{ + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + unsigned i, j; + + assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if( first == 0 ) { + mask &= ~TGSI_WRITEMASK_XY; + if( mask == TGSI_WRITEMASK_NONE ) { + first++; + } + } + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + switch( decl->Interpolation.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + emit_coef_a0( func, 0, i, j ); + emit_inputs( func, 0, i, j ); + break; + + case TGSI_INTERPOLATE_LINEAR: + emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_coef_dadx( func, 1, i, j ); + emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_coef_dady( func, 3, i, j ); + emit_mul( func, 0, 1 ); /* x * dadx */ + emit_coef_a0( func, 4, i, j ); + emit_mul( func, 2, 3 ); /* y * dady */ + emit_add( func, 0, 4 ); /* x * dadx + a0 */ + emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ + emit_inputs( func, 0, i, j ); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + emit_inputf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_coef_dadx( func, 1, i, j ); + emit_inputf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_coef_dady( func, 3, i, j ); + emit_mul( func, 0, 1 ); /* x * dadx */ + emit_inputf( func, 4, 0, TGSI_SWIZZLE_W ); + emit_coef_a0( func, 5, i, j ); + emit_rcp( func, 4, 4 ); /* 1.0 / w */ + emit_mul( func, 2, 3 ); /* y * dady */ + emit_add( func, 0, 5 ); /* x * dadx + a0 */ + emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ + emit_mul( func, 0, 4 ); /* (x * dadx + y * dady + a0) / w */ + emit_inputs( func, 0, i, j ); + break; + + default: + assert( 0 ); + break; + } + } + } + } + } +} + +unsigned +tgsi_emit_sse2( + struct tgsi_token *tokens, + struct x86_function *func ) +{ + struct tgsi_parse_context parse; + unsigned ok = 1; + + DUMP_START(); + + func->csr = func->store; + + emit_mov( + func, + get_input_base(), + get_argument( 0 ) ); + emit_mov( + func, + get_output_base(), + get_argument( 1 ) ); + emit_mov( + func, + get_const_base(), + get_argument( 2 ) ); + emit_mov( + func, + get_temp_base(), + get_argument( 3 ) ); + + tgsi_parse_init( &parse, tokens ); + + while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + ok = emit_instruction( + func, + &parse.FullToken.FullInstruction ); + + if (!ok) { + debug_printf("failed to translate tgsi opcode %d\n", + parse.FullToken.FullInstruction.Instruction.Opcode ); + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* XXX implement this */ + ok = 0; + debug_printf("failed to emit immediate value\n"); + break; + + default: + assert( 0 ); + ok = 0; + break; + } + } + + tgsi_parse_free( &parse ); + + DUMP_END(); + + return ok; +} + +/** + * Fragment shaders are responsible for interpolating shader inputs. Because on + * x86 we have only 4 GP registers, and here we have 5 shader arguments (input, + * output, const, temp and coef), the code is split into two phases -- + * DECLARATION and INSTRUCTION phase. + * GP register holding the output argument is aliased with the coeff argument, + * as outputs are not needed in the DECLARATION phase. + */ +unsigned +tgsi_emit_sse2_fs( + struct tgsi_token *tokens, + struct x86_function *func ) +{ + struct tgsi_parse_context parse; + boolean instruction_phase = FALSE; + + DUMP_START(); + + func->csr = func->store; + + /* DECLARATION phase, do not load output argument. */ + emit_mov( + func, + get_input_base(), + get_argument( 0 ) ); + emit_mov( + func, + get_const_base(), + get_argument( 2 ) ); + emit_mov( + func, + get_temp_base(), + get_argument( 3 ) ); + emit_mov( + func, + get_coef_base(), + get_argument( 4 ) ); + + tgsi_parse_init( &parse, tokens ); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + emit_declaration( + func, + &parse.FullToken.FullDeclaration ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + if( !instruction_phase ) { + /* INSTRUCTION phase, overwrite coeff with output. */ + instruction_phase = TRUE; + emit_mov( + func, + get_output_base(), + get_argument( 1 ) ); + } + emit_instruction( + func, + &parse.FullToken.FullInstruction ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* XXX implement this */ + assert(0); + break; + + default: + assert( 0 ); + } + } + + tgsi_parse_free( &parse ); + + DUMP_END(); + + return 1; +} + +#endif /* i386 */ diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h new file mode 100755 index 0000000000..9bee371766 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h @@ -0,0 +1,26 @@ +#if !defined TGSI_SSE2_H +#define TGSI_SSE2_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_token; +struct x86_function; + +unsigned +tgsi_emit_sse2( + struct tgsi_token *tokens, + struct x86_function *function ); + +unsigned +tgsi_emit_sse2_fs( + struct tgsi_token *tokens, + struct x86_function *function ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_SSE2_H + diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.c b/src/gallium/auxiliary/tgsi/util/tgsi_build.c new file mode 100644 index 0000000000..a00ff1c2a5 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.c @@ -0,0 +1,1371 @@ +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_build.h" +#include "tgsi_parse.h" + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ) +{ + struct tgsi_version version; + + version.MajorVersion = 1; + version.MinorVersion = 1; + version.Padding = 0; + + return version; +} + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ) +{ + struct tgsi_header header; + + header.HeaderSize = 1; + header.BodySize = 0; + + return header; +} + +static void +header_headersize_grow( struct tgsi_header *header ) +{ + assert( header->HeaderSize < 0xFF ); + assert( header->BodySize == 0 ); + + header->HeaderSize++; +} + +static void +header_bodysize_grow( struct tgsi_header *header ) +{ + assert( header->BodySize < 0xFFFFFF ); + + header->BodySize++; +} + +struct tgsi_processor +tgsi_default_processor( void ) +{ + struct tgsi_processor processor; + + processor.Processor = TGSI_PROCESSOR_FRAGMENT; + processor.Padding = 0; + + return processor; +} + +struct tgsi_processor +tgsi_build_processor( + unsigned type, + struct tgsi_header *header ) +{ + struct tgsi_processor processor; + + processor = tgsi_default_processor(); + processor.Processor = type; + + header_headersize_grow( header ); + + return processor; +} + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ) +{ + struct tgsi_declaration declaration; + + declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; + declaration.Size = 1; + declaration.File = TGSI_FILE_NULL; + declaration.Declare = TGSI_DECLARE_RANGE; + declaration.UsageMask = TGSI_WRITEMASK_XYZW; + declaration.Interpolate = 0; + declaration.Semantic = 0; + declaration.Padding = 0; + declaration.Extended = 0; + + return declaration; +} + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned declare, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ) +{ + struct tgsi_declaration declaration; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( declare <= TGSI_DECLARE_MASK ); + + declaration = tgsi_default_declaration(); + declaration.File = file; + declaration.Declare = declare; + declaration.UsageMask = usage_mask; + declaration.Interpolate = interpolate; + declaration.Semantic = semantic; + + header_bodysize_grow( header ); + + return declaration; +} + +static void +declaration_grow( + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + assert( declaration->Size < 0xFF ); + + declaration->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ) +{ + struct tgsi_full_declaration full_declaration; + + full_declaration.Declaration = tgsi_default_declaration(); + full_declaration.Interpolation = tgsi_default_declaration_interpolation(); + full_declaration.Semantic = tgsi_default_declaration_semantic(); + + return full_declaration; +} + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0; + struct tgsi_declaration *declaration; + + if( maxsize <= size ) + return 0; + declaration = (struct tgsi_declaration *) &tokens[size]; + size++; + + *declaration = tgsi_build_declaration( + full_decl->Declaration.File, + full_decl->Declaration.Declare, + full_decl->Declaration.UsageMask, + full_decl->Declaration.Interpolate, + full_decl->Declaration.Semantic, + header ); + + switch( full_decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + { + struct tgsi_declaration_range *dr; + + if( maxsize <= size ) + return 0; + dr = (struct tgsi_declaration_range *) &tokens[size]; + size++; + + *dr = tgsi_build_declaration_range( + full_decl->u.DeclarationRange.First, + full_decl->u.DeclarationRange.Last, + declaration, + header ); + break; + } + + case TGSI_DECLARE_MASK: + { + struct tgsi_declaration_mask *dm; + + if( maxsize <= size ) + return 0; + dm = (struct tgsi_declaration_mask *) &tokens[size]; + size++; + + *dm = tgsi_build_declaration_mask( + full_decl->u.DeclarationMask.Mask, + declaration, + header ); + break; + } + + default: + assert( 0 ); + } + + if( full_decl->Declaration.Interpolate ) { + struct tgsi_declaration_interpolation *di; + + if( maxsize <= size ) + return 0; + di = (struct tgsi_declaration_interpolation *) &tokens[size]; + size++; + + *di = tgsi_build_declaration_interpolation( + full_decl->Interpolation.Interpolate, + declaration, + header ); + } + + if( full_decl->Declaration.Semantic ) { + struct tgsi_declaration_semantic *ds; + + if( maxsize <= size ) + return 0; + ds = (struct tgsi_declaration_semantic *) &tokens[size]; + size++; + + *ds = tgsi_build_declaration_semantic( + full_decl->Semantic.SemanticName, + full_decl->Semantic.SemanticIndex, + declaration, + header ); + } + + return size; +} + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_range declaration_range; + + assert( last >= first ); + assert( last <= 0xFFFF ); + + declaration_range.First = first; + declaration_range.Last = last; + + declaration_grow( declaration, header ); + + return declaration_range; +} + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + unsigned mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_mask declaration_mask; + + declaration_mask.Mask = mask; + + declaration_grow( declaration, header ); + + return declaration_mask; +} + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ) +{ + struct tgsi_declaration_interpolation di; + + di.Interpolate = TGSI_INTERPOLATE_CONSTANT; + di.Padding = 0; + + return di; +} + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + unsigned interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_interpolation di; + + assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); + + di = tgsi_default_declaration_interpolation(); + di.Interpolate = interpolate; + + declaration_grow( declaration, header ); + + return di; +} + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ) +{ + struct tgsi_declaration_semantic ds; + + ds.SemanticName = TGSI_SEMANTIC_POSITION; + ds.SemanticIndex = 0; + ds.Padding = 0; + + return ds; +} + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_semantic ds; + + assert( semantic_name <= TGSI_SEMANTIC_COUNT ); + assert( semantic_index <= 0xFFFF ); + + ds = tgsi_default_declaration_semantic(); + ds.SemanticName = semantic_name; + ds.SemanticIndex = semantic_index; + + declaration_grow( declaration, header ); + + return ds; +} + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ) +{ + struct tgsi_immediate immediate; + + immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + immediate.Size = 1; + immediate.DataType = TGSI_IMM_FLOAT32; + immediate.Padding = 0; + immediate.Extended = 0; + + return immediate; +} + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ) +{ + struct tgsi_immediate immediate; + + immediate = tgsi_default_immediate(); + + header_bodysize_grow( header ); + + return immediate; +} + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ) +{ + struct tgsi_full_immediate fullimm; + + fullimm.Immediate = tgsi_default_immediate(); + fullimm.u.Pointer = (void *) 0; + + return fullimm; +} + +static void +immediate_grow( + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + assert( immediate->Size < 0xFF ); + + immediate->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + float value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + struct tgsi_immediate_float32 immediate_float32; + + immediate_float32.Float = value; + + immediate_grow( immediate, header ); + + return immediate_float32; +} + +unsigned +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0, i; + struct tgsi_immediate *immediate; + + if( maxsize <= size ) + return 0; + immediate = (struct tgsi_immediate *) &tokens[size]; + size++; + + *immediate = tgsi_build_immediate( header ); + + for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { + struct tgsi_immediate_float32 *if32; + + if( maxsize <= size ) + return 0; + if32 = (struct tgsi_immediate_float32 *) &tokens[size]; + size++; + + *if32 = tgsi_build_immediate_float32( + full_imm->u.ImmediateFloat32[i].Float, + immediate, + header ); + } + + return size; +} + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ) +{ + struct tgsi_instruction instruction; + + instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; + instruction.Size = 1; + instruction.Opcode = TGSI_OPCODE_MOV; + instruction.Saturate = TGSI_SAT_NONE; + instruction.NumDstRegs = 1; + instruction.NumSrcRegs = 1; + instruction.Padding = 0; + instruction.Extended = 0; + + return instruction; +} + +struct tgsi_instruction +tgsi_build_instruction( + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header ) +{ + struct tgsi_instruction instruction; + + assert (opcode <= TGSI_OPCODE_LAST); + assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); + assert (num_dst_regs <= 3); + assert (num_src_regs <= 15); + + instruction = tgsi_default_instruction(); + instruction.Opcode = opcode; + instruction.Saturate = saturate; + instruction.NumDstRegs = num_dst_regs; + instruction.NumSrcRegs = num_src_regs; + + header_bodysize_grow( header ); + + return instruction; +} + +static void +instruction_grow( + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + assert (instruction->Size < 0xFF); + + instruction->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ) +{ + struct tgsi_full_instruction full_instruction; + unsigned i; + + full_instruction.Instruction = tgsi_default_instruction(); + full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); + full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); + full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); + for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { + full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); + } + for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { + full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); + } + + return full_instruction; +} + +unsigned +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0; + unsigned i; + struct tgsi_instruction *instruction; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + instruction = (struct tgsi_instruction *) &tokens[size]; + size++; + + *instruction = tgsi_build_instruction( + full_inst->Instruction.Opcode, + full_inst->Instruction.Saturate, + full_inst->Instruction.NumDstRegs, + full_inst->Instruction.NumSrcRegs, + header ); + prev_token = (struct tgsi_token *) instruction; + + if( tgsi_compare_instruction_ext_nv( + full_inst->InstructionExtNv, + tgsi_default_instruction_ext_nv() ) ) { + struct tgsi_instruction_ext_nv *instruction_ext_nv; + + if( maxsize <= size ) + return 0; + instruction_ext_nv = + (struct tgsi_instruction_ext_nv *) &tokens[size]; + size++; + + *instruction_ext_nv = tgsi_build_instruction_ext_nv( + full_inst->InstructionExtNv.Precision, + full_inst->InstructionExtNv.CondDstIndex, + full_inst->InstructionExtNv.CondFlowIndex, + full_inst->InstructionExtNv.CondMask, + full_inst->InstructionExtNv.CondSwizzleX, + full_inst->InstructionExtNv.CondSwizzleY, + full_inst->InstructionExtNv.CondSwizzleZ, + full_inst->InstructionExtNv.CondSwizzleW, + full_inst->InstructionExtNv.CondDstUpdate, + full_inst->InstructionExtNv.CondFlowEnable, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_nv; + } + + if( tgsi_compare_instruction_ext_label( + full_inst->InstructionExtLabel, + tgsi_default_instruction_ext_label() ) ) { + struct tgsi_instruction_ext_label *instruction_ext_label; + + if( maxsize <= size ) + return 0; + instruction_ext_label = + (struct tgsi_instruction_ext_label *) &tokens[size]; + size++; + + *instruction_ext_label = tgsi_build_instruction_ext_label( + full_inst->InstructionExtLabel.Label, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_label; + } + + if( tgsi_compare_instruction_ext_texture( + full_inst->InstructionExtTexture, + tgsi_default_instruction_ext_texture() ) ) { + struct tgsi_instruction_ext_texture *instruction_ext_texture; + + if( maxsize <= size ) + return 0; + instruction_ext_texture = + (struct tgsi_instruction_ext_texture *) &tokens[size]; + size++; + + *instruction_ext_texture = tgsi_build_instruction_ext_texture( + full_inst->InstructionExtTexture.Texture, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_texture; + } + + for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { + const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; + struct tgsi_dst_register *dst_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + dst_register = (struct tgsi_dst_register *) &tokens[size]; + size++; + + *dst_register = tgsi_build_dst_register( + reg->DstRegister.File, + reg->DstRegister.WriteMask, + reg->DstRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register; + + if( tgsi_compare_dst_register_ext_concode( + reg->DstRegisterExtConcode, + tgsi_default_dst_register_ext_concode() ) ) { + struct tgsi_dst_register_ext_concode *dst_register_ext_concode; + + if( maxsize <= size ) + return 0; + dst_register_ext_concode = + (struct tgsi_dst_register_ext_concode *) &tokens[size]; + size++; + + *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( + reg->DstRegisterExtConcode.CondMask, + reg->DstRegisterExtConcode.CondSwizzleX, + reg->DstRegisterExtConcode.CondSwizzleY, + reg->DstRegisterExtConcode.CondSwizzleZ, + reg->DstRegisterExtConcode.CondSwizzleW, + reg->DstRegisterExtConcode.CondSrcIndex, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_concode; + } + + if( tgsi_compare_dst_register_ext_modulate( + reg->DstRegisterExtModulate, + tgsi_default_dst_register_ext_modulate() ) ) { + struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; + + if( maxsize <= size ) + return 0; + dst_register_ext_modulate = + (struct tgsi_dst_register_ext_modulate *) &tokens[size]; + size++; + + *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( + reg->DstRegisterExtModulate.Modulate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_modulate; + } + } + + for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { + const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; + struct tgsi_src_register *src_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + src_register = (struct tgsi_src_register *) &tokens[size]; + size++; + + *src_register = tgsi_build_src_register( + reg->SrcRegister.File, + reg->SrcRegister.SwizzleX, + reg->SrcRegister.SwizzleY, + reg->SrcRegister.SwizzleZ, + reg->SrcRegister.SwizzleW, + reg->SrcRegister.Negate, + reg->SrcRegister.Indirect, + reg->SrcRegister.Dimension, + reg->SrcRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register; + + if( tgsi_compare_src_register_ext_swz( + reg->SrcRegisterExtSwz, + tgsi_default_src_register_ext_swz() ) ) { + struct tgsi_src_register_ext_swz *src_register_ext_swz; + + if( maxsize <= size ) + return 0; + src_register_ext_swz = + (struct tgsi_src_register_ext_swz *) &tokens[size]; + size++; + + *src_register_ext_swz = tgsi_build_src_register_ext_swz( + reg->SrcRegisterExtSwz.ExtSwizzleX, + reg->SrcRegisterExtSwz.ExtSwizzleY, + reg->SrcRegisterExtSwz.ExtSwizzleZ, + reg->SrcRegisterExtSwz.ExtSwizzleW, + reg->SrcRegisterExtSwz.NegateX, + reg->SrcRegisterExtSwz.NegateY, + reg->SrcRegisterExtSwz.NegateZ, + reg->SrcRegisterExtSwz.NegateW, + reg->SrcRegisterExtSwz.ExtDivide, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_swz; + } + + if( tgsi_compare_src_register_ext_mod( + reg->SrcRegisterExtMod, + tgsi_default_src_register_ext_mod() ) ) { + struct tgsi_src_register_ext_mod *src_register_ext_mod; + + if( maxsize <= size ) + return 0; + src_register_ext_mod = + (struct tgsi_src_register_ext_mod *) &tokens[size]; + size++; + + *src_register_ext_mod = tgsi_build_src_register_ext_mod( + reg->SrcRegisterExtMod.Complement, + reg->SrcRegisterExtMod.Bias, + reg->SrcRegisterExtMod.Scale2X, + reg->SrcRegisterExtMod.Absolute, + reg->SrcRegisterExtMod.Negate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_mod; + } + + if( reg->SrcRegister.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterInd.File, + reg->SrcRegisterInd.SwizzleX, + reg->SrcRegisterInd.SwizzleY, + reg->SrcRegisterInd.SwizzleZ, + reg->SrcRegisterInd.SwizzleW, + reg->SrcRegisterInd.Negate, + reg->SrcRegisterInd.Indirect, + reg->SrcRegisterInd.Dimension, + reg->SrcRegisterInd.Index, + instruction, + header ); + } + + if( reg->SrcRegister.Dimension ) { + struct tgsi_dimension *dim; + + assert( !reg->SrcRegisterDim.Dimension ); + + if( maxsize <= size ) + return 0; + dim = (struct tgsi_dimension *) &tokens[size]; + size++; + + *dim = tgsi_build_dimension( + reg->SrcRegisterDim.Indirect, + reg->SrcRegisterDim.Index, + instruction, + header ); + + if( reg->SrcRegisterDim.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterDimInd.File, + reg->SrcRegisterDimInd.SwizzleX, + reg->SrcRegisterDimInd.SwizzleY, + reg->SrcRegisterDimInd.SwizzleZ, + reg->SrcRegisterDimInd.SwizzleW, + reg->SrcRegisterDimInd.Negate, + reg->SrcRegisterDimInd.Indirect, + reg->SrcRegisterDimInd.Dimension, + reg->SrcRegisterDimInd.Index, + instruction, + header ); + } + } + } + + return size; +} + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; + instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; + instruction_ext_nv.CondDstIndex = 0; + instruction_ext_nv.CondFlowIndex = 0; + instruction_ext_nv.CondMask = TGSI_CC_TR; + instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; + instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; + instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; + instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; + instruction_ext_nv.CondDstUpdate = 0; + instruction_ext_nv.CondFlowEnable = 0; + instruction_ext_nv.Padding = 0; + instruction_ext_nv.Extended = 0; + + return instruction_ext_nv; +} + +union token_u32 +{ + unsigned u32; +}; + +unsigned +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv = tgsi_default_instruction_ext_nv(); + instruction_ext_nv.Precision = precision; + instruction_ext_nv.CondDstIndex = cond_dst_index; + instruction_ext_nv.CondFlowIndex = cond_flow_index; + instruction_ext_nv.CondMask = cond_mask; + instruction_ext_nv.CondSwizzleX = cond_swizzle_x; + instruction_ext_nv.CondSwizzleY = cond_swizzle_y; + instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; + instruction_ext_nv.CondSwizzleW = cond_swizzle_w; + instruction_ext_nv.CondDstUpdate = cond_dst_update; + instruction_ext_nv.CondFlowEnable = cond_flow_update; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_nv; +} + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; + instruction_ext_label.Label = 0; + instruction_ext_label.Padding = 0; + instruction_ext_label.Extended = 0; + + return instruction_ext_label; +} + +unsigned +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + unsigned label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label = tgsi_default_instruction_ext_label(); + instruction_ext_label.Label = label; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_label; +} + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; + instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; + instruction_ext_texture.Padding = 0; + instruction_ext_texture.Extended = 0; + + return instruction_ext_texture; +} + +unsigned +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + unsigned texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture = tgsi_default_instruction_ext_texture(); + instruction_ext_texture.Texture = texture; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_texture; +} + +struct tgsi_src_register +tgsi_default_src_register( void ) +{ + struct tgsi_src_register src_register; + + src_register.File = TGSI_FILE_NULL; + src_register.SwizzleX = TGSI_SWIZZLE_X; + src_register.SwizzleY = TGSI_SWIZZLE_Y; + src_register.SwizzleZ = TGSI_SWIZZLE_Z; + src_register.SwizzleW = TGSI_SWIZZLE_W; + src_register.Negate = 0; + src_register.Indirect = 0; + src_register.Dimension = 0; + src_register.Index = 0; + src_register.Extended = 0; + + return src_register; +} + +struct tgsi_src_register +tgsi_build_src_register( + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register src_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( negate <= 1 ); + assert( index >= -0x8000 && index <= 0x7FFF ); + + src_register = tgsi_default_src_register(); + src_register.File = file; + src_register.SwizzleX = swizzle_x; + src_register.SwizzleY = swizzle_y; + src_register.SwizzleZ = swizzle_z; + src_register.SwizzleW = swizzle_w; + src_register.Negate = negate; + src_register.Indirect = indirect; + src_register.Dimension = dimension; + src_register.Index = index; + + instruction_grow( instruction, header ); + + return src_register; +} + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ) +{ + struct tgsi_full_src_register full_src_register; + + full_src_register.SrcRegister = tgsi_default_src_register(); + full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); + full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); + full_src_register.SrcRegisterInd = tgsi_default_src_register(); + full_src_register.SrcRegisterDim = tgsi_default_dimension(); + full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); + + return full_src_register; +} + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; + src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; + src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; + src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; + src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; + src_register_ext_swz.NegateX = 0; + src_register_ext_swz.NegateY = 0; + src_register_ext_swz.NegateZ = 0; + src_register_ext_swz.NegateW = 0; + src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; + src_register_ext_swz.Padding = 0; + src_register_ext_swz.Extended = 0; + + return src_register_ext_swz; +} + +unsigned +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + unsigned ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE ); + assert( negate_x <= 1 ); + assert( negate_y <= 1 ); + assert( negate_z <= 1 ); + assert( negate_w <= 1 ); + assert( ext_divide <= TGSI_EXTSWIZZLE_ONE ); + + src_register_ext_swz = tgsi_default_src_register_ext_swz(); + src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; + src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; + src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; + src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; + src_register_ext_swz.NegateX = negate_x; + src_register_ext_swz.NegateY = negate_y; + src_register_ext_swz.NegateZ = negate_z; + src_register_ext_swz.NegateW = negate_w; + src_register_ext_swz.ExtDivide = ext_divide; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_swz; +} + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; + src_register_ext_mod.Complement = 0; + src_register_ext_mod.Bias = 0; + src_register_ext_mod.Scale2X = 0; + src_register_ext_mod.Absolute = 0; + src_register_ext_mod.Negate = 0; + src_register_ext_mod.Padding = 0; + src_register_ext_mod.Extended = 0; + + return src_register_ext_mod; +} + +unsigned +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + assert( complement <= 1 ); + assert( bias <= 1 ); + assert( scale_2x <= 1 ); + assert( absolute <= 1 ); + assert( negate <= 1 ); + + src_register_ext_mod = tgsi_default_src_register_ext_mod(); + src_register_ext_mod.Complement = complement; + src_register_ext_mod.Bias = bias; + src_register_ext_mod.Scale2X = scale_2x; + src_register_ext_mod.Absolute = absolute; + src_register_ext_mod.Negate = negate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_mod; +} + +struct tgsi_dimension +tgsi_default_dimension( void ) +{ + struct tgsi_dimension dimension; + + dimension.Indirect = 0; + dimension.Dimension = 0; + dimension.Padding = 0; + dimension.Index = 0; + dimension.Extended = 0; + + return dimension; +} + +struct tgsi_dimension +tgsi_build_dimension( + unsigned indirect, + unsigned index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dimension dimension; + + dimension = tgsi_default_dimension(); + dimension.Indirect = indirect; + dimension.Index = index; + + instruction_grow( instruction, header ); + + return dimension; +} + +struct tgsi_dst_register +tgsi_default_dst_register( void ) +{ + struct tgsi_dst_register dst_register; + + dst_register.File = TGSI_FILE_NULL; + dst_register.WriteMask = TGSI_WRITEMASK_XYZW; + dst_register.Indirect = 0; + dst_register.Dimension = 0; + dst_register.Index = 0; + dst_register.Padding = 0; + dst_register.Extended = 0; + + return dst_register; +} + +struct tgsi_dst_register +tgsi_build_dst_register( + unsigned file, + unsigned mask, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register dst_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( mask <= TGSI_WRITEMASK_XYZW ); + assert( index >= -32768 && index <= 32767 ); + + dst_register = tgsi_default_dst_register(); + dst_register.File = file; + dst_register.WriteMask = mask; + dst_register.Index = index; + + instruction_grow( instruction, header ); + + return dst_register; +} + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ) +{ + struct tgsi_full_dst_register full_dst_register; + + full_dst_register.DstRegister = tgsi_default_dst_register(); + full_dst_register.DstRegisterExtConcode = + tgsi_default_dst_register_ext_concode(); + full_dst_register.DstRegisterExtModulate = + tgsi_default_dst_register_ext_modulate(); + + return full_dst_register; +} + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; + dst_register_ext_concode.CondMask = TGSI_CC_TR; + dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; + dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; + dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; + dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; + dst_register_ext_concode.CondSrcIndex = 0; + dst_register_ext_concode.Padding = 0; + dst_register_ext_concode.Extended = 0; + + return dst_register_ext_concode; +} + +unsigned +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + assert( cc <= TGSI_CC_FL ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( index >= -32768 && index <= 32767 ); + + dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); + dst_register_ext_concode.CondMask = cc; + dst_register_ext_concode.CondSwizzleX = swizzle_x; + dst_register_ext_concode.CondSwizzleY = swizzle_y; + dst_register_ext_concode.CondSwizzleZ = swizzle_z; + dst_register_ext_concode.CondSwizzleW = swizzle_w; + dst_register_ext_concode.CondSrcIndex = index; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_concode; +} + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; + dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; + dst_register_ext_modulate.Padding = 0; + dst_register_ext_modulate.Extended = 0; + + return dst_register_ext_modulate; +} + +unsigned +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + unsigned modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + assert( modulate <= TGSI_MODULATE_EIGHTH ); + + dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); + dst_register_ext_modulate.Modulate = modulate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_modulate; +} + diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.h b/src/gallium/auxiliary/tgsi/util/tgsi_build.h new file mode 100644 index 0000000000..116c78abf3 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.h @@ -0,0 +1,320 @@ +#if !defined TGSI_BUILD_H +#define TGSI_BUILD_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ); + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ); + +struct tgsi_processor +tgsi_default_processor( void ); + +struct tgsi_processor +tgsi_build_processor( + unsigned processor, + struct tgsi_header *header ); + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ); + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned declare, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ); + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ); + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_mask +tgsi_build_declaration_mask( + unsigned mask, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_interpolation +tgsi_default_declaration_interpolation( void ); + +struct tgsi_declaration_interpolation +tgsi_build_declaration_interpolation( + unsigned interpolate, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ); + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ); + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ); + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ); + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + float value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ); + +unsigned +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ); + +struct tgsi_instruction +tgsi_build_instruction( + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header ); + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ); + +unsigned +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ); + +unsigned +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ); + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ); + +unsigned +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ); + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + unsigned label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ); + +unsigned +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ); + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + unsigned texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register +tgsi_default_src_register( void ); + +struct tgsi_src_register +tgsi_build_src_register( + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ); + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ); + +unsigned +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ); + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + unsigned ext_divide, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ); + +unsigned +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ); + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dimension +tgsi_default_dimension( void ); + +struct tgsi_dimension +tgsi_build_dimension( + unsigned indirect, + unsigned index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register +tgsi_default_dst_register( void ); + +struct tgsi_dst_register +tgsi_build_dst_register( + unsigned file, + unsigned mask, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ); + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ); + +unsigned +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ); + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ); + +unsigned +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ); + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + unsigned modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_BUILD_H + diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c new file mode 100644 index 0000000000..b5c54847e0 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -0,0 +1,1581 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_dump.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" + +struct gen_dump +{ + unsigned tabs; + void (* write)( + struct gen_dump *dump, + const void *data, + unsigned size ); +}; + +struct text_dump +{ + struct gen_dump base; + char *text; + unsigned length; + unsigned capacity; +}; + +static void +_text_dump_write( + struct gen_dump *dump, + const void *data, + unsigned size ) +{ + struct text_dump *td = (struct text_dump *) dump; + unsigned new_length = td->length + size; + + if( new_length >= td->capacity ) { + unsigned new_capacity = td->capacity; + + do { + if( new_capacity == 0 ) { + new_capacity = 256; + } + else { + new_capacity *= 2; + } + } while( new_length >= new_capacity ); + td->text = (char *) REALLOC( + td->text, + td->capacity, + new_capacity ); + td->capacity = new_capacity; + } + memcpy( + &td->text[td->length], + data, + size ); + td->length = new_length; + td->text[td->length] = '\0'; +} + +struct file_dump +{ + struct gen_dump base; + FILE *file; +}; + +static void +_file_dump_write( + struct gen_dump *dump, + const void *data, + unsigned size ) +{ + struct file_dump *fd = (struct file_dump *) dump; + +#if 0 + fwrite( data, 1, size, fd->file ); +#else + { + unsigned i; + + for (i = 0; i < size; i++ ) { + fprintf( fd->file, "%c", ((const char *) data)[i] ); + } + } +#endif +} + +static void +gen_dump_str( + struct gen_dump *dump, + const char *str ) +{ + unsigned i; + size_t len = strlen( str ); + + for (i = 0; i < len; i++) { + dump->write( dump, &str[i], 1 ); + if (str[i] == '\n') { + unsigned i; + + for (i = 0; i < dump->tabs; i++) { + dump->write( dump, " ", 4 ); + } + } + } +} + +static void +gen_dump_chr( + struct gen_dump *dump, + const char chr ) +{ + dump->write( dump, &chr, 1 ); +} + +static void +gen_dump_uix( + struct gen_dump *dump, + const unsigned ui ) +{ + char str[36]; + + sprintf( str, "0x%x", ui ); + gen_dump_str( dump, str ); +} + +static void +gen_dump_uid( + struct gen_dump *dump, + const unsigned ui ) +{ + char str[16]; + + sprintf( str, "%u", ui ); + gen_dump_str( dump, str ); +} + +static void +gen_dump_sid( + struct gen_dump *dump, + const int si ) +{ + char str[16]; + + sprintf( str, "%d", si ); + gen_dump_str( dump, str ); +} + +static void +gen_dump_flt( + struct gen_dump *dump, + const float flt ) +{ + char str[48]; + + sprintf( str, "%10.4f", flt ); + gen_dump_str( dump, str ); +} + +static void +gen_dump_enum( + struct gen_dump *dump, + const unsigned e, + const char **enums, + const unsigned enums_count ) +{ + if (e >= enums_count) { + gen_dump_uid( dump, e ); + } + else { + gen_dump_str( dump, enums[e] ); + } +} + +static void +gen_dump_tab( + struct gen_dump *dump ) +{ + ++dump->tabs; +} + +static void +gen_dump_untab( + struct gen_dump *dump ) +{ + assert( dump->tabs > 0 ); + + --dump->tabs; +} + +#define TXT(S) gen_dump_str( dump, S ) +#define CHR(C) gen_dump_chr( dump, C ) +#define UIX(I) gen_dump_uix( dump, I ) +#define UID(I) gen_dump_uid( dump, I ) +#define SID(I) gen_dump_sid( dump, I ) +#define FLT(F) gen_dump_flt( dump, F ) +#define TAB() gen_dump_tab( dump ) +#define UNT() gen_dump_untab( dump ) +#define ENM(E,ENUMS) gen_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) + +static const char *TGSI_PROCESSOR_TYPES[] = +{ + "PROCESSOR_FRAGMENT", + "PROCESSOR_VERTEX", + "PROCESSOR_GEOMETRY" +}; + +static const char *TGSI_PROCESSOR_TYPES_SHORT[] = +{ + "FRAG", + "VERT", + "GEOM" +}; + +static const char *TGSI_TOKEN_TYPES[] = +{ + "TOKEN_TYPE_DECLARATION", + "TOKEN_TYPE_IMMEDIATE", + "TOKEN_TYPE_INSTRUCTION" +}; + +static const char *TGSI_FILES[] = +{ + "FILE_NULL", + "FILE_CONSTANT", + "FILE_INPUT", + "FILE_OUTPUT", + "FILE_TEMPORARY", + "FILE_SAMPLER", + "FILE_ADDRESS", + "FILE_IMMEDIATE" +}; + +static const char *TGSI_FILES_SHORT[] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + +static const char *TGSI_DECLARES[] = +{ + "DECLARE_RANGE", + "DECLARE_MASK" +}; + +static const char *TGSI_INTERPOLATES[] = +{ + "INTERPOLATE_CONSTANT", + "INTERPOLATE_LINEAR", + "INTERPOLATE_PERSPECTIVE", + "INTERPOLATE_ATTRIB" +}; + +static const char *TGSI_INTERPOLATES_SHORT[] = +{ + "CONSTANT", + "LINEAR", + "PERSPECTIVE", + "ATTRIB" +}; + +static const char *TGSI_SEMANTICS[] = +{ + "SEMANTIC_POSITION", + "SEMANTIC_COLOR", + "SEMANTIC_BCOLOR", + "SEMANTIC_FOG", + "SEMANTIC_PSIZE", + "SEMANTIC_GENERIC," +}; + +static const char *TGSI_SEMANTICS_SHORT[] = +{ + "POSITION", + "COLOR", + "BCOLOR", + "FOG", + "PSIZE", + "GENERIC", +}; + +static const char *TGSI_IMMS[] = +{ + "IMM_FLOAT32" +}; + +static const char *TGSI_IMMS_SHORT[] = +{ + "FLT32" +}; + +static const char *TGSI_OPCODES[] = +{ + "OPCODE_ARL", + "OPCODE_MOV", + "OPCODE_LIT", + "OPCODE_RCP", + "OPCODE_RSQ", + "OPCODE_EXP", + "OPCODE_LOG", + "OPCODE_MUL", + "OPCODE_ADD", + "OPCODE_DP3", + "OPCODE_DP4", + "OPCODE_DST", + "OPCODE_MIN", + "OPCODE_MAX", + "OPCODE_SLT", + "OPCODE_SGE", + "OPCODE_MAD", + "OPCODE_SUB", + "OPCODE_LERP", + "OPCODE_CND", + "OPCODE_CND0", + "OPCODE_DOT2ADD", + "OPCODE_INDEX", + "OPCODE_NEGATE", + "OPCODE_FRAC", + "OPCODE_CLAMP", + "OPCODE_FLOOR", + "OPCODE_ROUND", + "OPCODE_EXPBASE2", + "OPCODE_LOGBASE2", + "OPCODE_POWER", + "OPCODE_CROSSPRODUCT", + "OPCODE_MULTIPLYMATRIX", + "OPCODE_ABS", + "OPCODE_RCC", + "OPCODE_DPH", + "OPCODE_COS", + "OPCODE_DDX", + "OPCODE_DDY", + "OPCODE_KILP", + "OPCODE_PK2H", + "OPCODE_PK2US", + "OPCODE_PK4B", + "OPCODE_PK4UB", + "OPCODE_RFL", + "OPCODE_SEQ", + "OPCODE_SFL", + "OPCODE_SGT", + "OPCODE_SIN", + "OPCODE_SLE", + "OPCODE_SNE", + "OPCODE_STR", + "OPCODE_TEX", + "OPCODE_TXD", + "OPCODE_UP2H", + "OPCODE_UP2US", + "OPCODE_UP4B", + "OPCODE_UP4UB", + "OPCODE_X2D", + "OPCODE_ARA", + "OPCODE_ARR", + "OPCODE_BRA", + "OPCODE_CAL", + "OPCODE_RET", + "OPCODE_SSG", + "OPCODE_CMP", + "OPCODE_SCS", + "OPCODE_TXB", + "OPCODE_NRM", + "OPCODE_DIV", + "OPCODE_DP2", + "OPCODE_TXL", + "OPCODE_BRK", + "OPCODE_IF", + "OPCODE_LOOP", + "OPCODE_REP", + "OPCODE_ELSE", + "OPCODE_ENDIF", + "OPCODE_ENDLOOP", + "OPCODE_ENDREP", + "OPCODE_PUSHA", + "OPCODE_POPA", + "OPCODE_CEIL", + "OPCODE_I2F", + "OPCODE_NOT", + "OPCODE_TRUNC", + "OPCODE_SHL", + "OPCODE_SHR", + "OPCODE_AND", + "OPCODE_OR", + "OPCODE_MOD", + "OPCODE_XOR", + "OPCODE_SAD", + "OPCODE_TXF", + "OPCODE_TXQ", + "OPCODE_CONT", + "OPCODE_EMIT", + "OPCODE_ENDPRIM", + "OPCODE_BGNLOOP2", + "OPCODE_BGNSUB", + "OPCODE_ENDLOOP2", + "OPCODE_ENDSUB", + "OPCODE_NOISE1", + "OPCODE_NOISE2", + "OPCODE_NOISE3", + "OPCODE_NOISE4", + "OPCODE_NOP", + "OPCODE_TEXBEM", + "OPCODE_TEXBEML", + "OPCODE_TEXREG2AR", + "OPCODE_TEXM3X2PAD", + "OPCODE_TEXM3X2TEX", + "OPCODE_TEXM3X3PAD", + "OPCODE_TEXM3X3TEX", + "OPCODE_TEXM3X3SPEC", + "OPCODE_TEXM3X3VSPEC", + "OPCODE_TEXREG2GB", + "OPCODE_TEXREG2RGB", + "OPCODE_TEXDP3TEX", + "OPCODE_TEXDP3", + "OPCODE_TEXM3X3", + "OPCODE_TEXM3X2DEPTH", + "OPCODE_TEXDEPTH", + "OPCODE_BEM", + "OPCODE_M4X3", + "OPCODE_M3X4", + "OPCODE_M3X3", + "OPCODE_M3X2", + "OPCODE_NRM4", + "OPCODE_CALLNZ", + "OPCODE_IFC", + "OPCODE_BREAKC", + "OPCODE_KIL", + "OPCODE_END" +}; + +static const char *TGSI_OPCODES_SHORT[] = +{ + "ARL", + "MOV", + "LIT", + "RCP", + "RSQ", + "EXP", + "LOG", + "MUL", + "ADD", + "DP3", + "DP4", + "DST", + "MIN", + "MAX", + "SLT", + "SGE", + "MAD", + "SUB", + "LERP", + "CND", + "CND0", + "DOT2ADD", + "INDEX", + "NEGATE", + "FRAC", + "CLAMP", + "FLOOR", + "ROUND", + "EXPBASE2", + "LOGBASE2", + "POWER", + "CROSSPRODUCT", + "MULTIPLYMATRIX", + "ABS", + "RCC", + "DPH", + "COS", + "DDX", + "DDY", + "KILP", + "PK2H", + "PK2US", + "PK4B", + "PK4UB", + "RFL", + "SEQ", + "SFL", + "SGT", + "SIN", + "SLE", + "SNE", + "STR", + "TEX", + "TXD", + "UP2H", + "UP2US", + "UP4B", + "UP4UB", + "X2D", + "ARA", + "ARR", + "BRA", + "CAL", + "RET", + "SSG", + "CMP", + "SCS", + "TXB", + "NRM", + "DIV", + "DP2", + "TXL", + "BRK", + "IF", + "LOOP", + "REP", + "ELSE", + "ENDIF", + "ENDLOOP", + "ENDREP", + "PUSHA", + "POPA", + "CEIL", + "I2F", + "NOT", + "TRUNC", + "SHL", + "SHR", + "AND", + "OR", + "MOD", + "XOR", + "SAD", + "TXF", + "TXQ", + "CONT", + "EMIT", + "ENDPRIM", + "BGNLOOP2", + "BGNSUB", + "ENDLOOP2", + "ENDSUB", + "NOISE1", + "NOISE2", + "NOISE3", + "NOISE4", + "NOP", + "TEXBEM", + "TEXBEML", + "TEXREG2AR", + "TEXM3X2PAD", + "TEXM3X2TEX", + "TEXM3X3PAD", + "TEXM3X3TEX", + "TEXM3X3SPEC", + "TEXM3X3VSPEC", + "TEXREG2GB", + "TEXREG2RGB", + "TEXDP3TEX", + "TEXDP3", + "TEXM3X3", + "TEXM3X2DEPTH", + "TEXDEPTH", + "BEM", + "M4X3", + "M3X4", + "M3X3", + "M3X2", + "NRM4", + "CALLNZ", + "IFC", + "BREAKC", + "KIL", + "END" +}; + +static const char *TGSI_SATS[] = +{ + "SAT_NONE", + "SAT_ZERO_ONE", + "SAT_MINUS_PLUS_ONE" +}; + +static const char *TGSI_INSTRUCTION_EXTS[] = +{ + "INSTRUCTION_EXT_TYPE_NV", + "INSTRUCTION_EXT_TYPE_LABEL", + "INSTRUCTION_EXT_TYPE_TEXTURE" +}; + +static const char *TGSI_PRECISIONS[] = +{ + "PRECISION_DEFAULT", + "TGSI_PRECISION_FLOAT32", + "TGSI_PRECISION_FLOAT16", + "TGSI_PRECISION_FIXED12" +}; + +static const char *TGSI_CCS[] = +{ + "CC_GT", + "CC_EQ", + "CC_LT", + "CC_UN", + "CC_GE", + "CC_LE", + "CC_NE", + "CC_TR", + "CC_FL" +}; + +static const char *TGSI_SWIZZLES[] = +{ + "SWIZZLE_X", + "SWIZZLE_Y", + "SWIZZLE_Z", + "SWIZZLE_W" +}; + +static const char *TGSI_SWIZZLES_SHORT[] = +{ + "x", + "y", + "z", + "w" +}; + +static const char *TGSI_TEXTURES[] = +{ + "TEXTURE_UNKNOWN", + "TEXTURE_1D", + "TEXTURE_2D", + "TEXTURE_3D", + "TEXTURE_CUBE", + "TEXTURE_RECT", + "TEXTURE_SHADOW1D", + "TEXTURE_SHADOW2D", + "TEXTURE_SHADOWRECT" +}; + +static const char *TGSI_SRC_REGISTER_EXTS[] = +{ + "SRC_REGISTER_EXT_TYPE_SWZ", + "SRC_REGISTER_EXT_TYPE_MOD" +}; + +static const char *TGSI_EXTSWIZZLES[] = +{ + "EXTSWIZZLE_X", + "EXTSWIZZLE_Y", + "EXTSWIZZLE_Z", + "EXTSWIZZLE_W", + "EXTSWIZZLE_ZERO", + "EXTSWIZZLE_ONE" +}; + +static const char *TGSI_EXTSWIZZLES_SHORT[] = +{ + "x", + "y", + "z", + "w", + "0", + "1" +}; + +static const char *TGSI_WRITEMASKS[] = +{ + "0", + "WRITEMASK_X", + "WRITEMASK_Y", + "WRITEMASK_XY", + "WRITEMASK_Z", + "WRITEMASK_XZ", + "WRITEMASK_YZ", + "WRITEMASK_XYZ", + "WRITEMASK_W", + "WRITEMASK_XW", + "WRITEMASK_YW", + "WRITEMASK_XYW", + "WRITEMASK_ZW", + "WRITEMASK_XZW", + "WRITEMASK_YZW", + "WRITEMASK_XYZW" +}; + +static const char *TGSI_DST_REGISTER_EXTS[] = +{ + "DST_REGISTER_EXT_TYPE_CONDCODE", + "DST_REGISTER_EXT_TYPE_MODULATE" +}; + +static const char *TGSI_MODULATES[] = +{ + "MODULATE_1X", + "MODULATE_2X", + "MODULATE_4X", + "MODULATE_8X", + "MODULATE_HALF", + "MODULATE_QUARTER", + "MODULATE_EIGHTH" +}; + +static void +dump_declaration_short( + struct gen_dump *dump, + struct tgsi_full_declaration *decl ) +{ + TXT( "\nDCL " ); + ENM( decl->Declaration.File, TGSI_FILES_SHORT ); + + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + CHR( '[' ); + UID( decl->u.DeclarationRange.First ); + if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { + TXT( ".." ); + UID( decl->u.DeclarationRange.Last ); + } + CHR( ']' ); + break; + default: + assert( 0 ); + } + + if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + + if( decl->Declaration.Interpolate ) { + TXT( ", " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); + } + + if( decl->Declaration.Semantic ) { + TXT( ", " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); + CHR( '[' ); + UID( decl->Semantic.SemanticIndex ); + CHR( ']' ); + } +} + +static void +dump_declaration_verbose( + struct gen_dump *dump, + struct tgsi_full_declaration *decl, + unsigned ignored, + unsigned deflt, + struct tgsi_full_declaration *fd ) +{ + TXT( "\nFile : " ); + ENM( decl->Declaration.File, TGSI_FILES ); + TXT( "\nDeclare : " ); + ENM( decl->Declaration.Declare, TGSI_DECLARES ); + if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { + TXT( "\nUsageMask : " ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'X' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'Y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'Z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'W' ); + } + } + if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { + TXT( "\nInterpolate: " ); + UID( decl->Declaration.Interpolate ); + } + if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { + TXT( "\nSemantic : " ); + UID( decl->Declaration.Semantic ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Declaration.Padding ); + } + + CHR( '\n' ); + switch( decl->Declaration.Declare ) { + case TGSI_DECLARE_RANGE: + TXT( "\nFirst: " ); + UID( decl->u.DeclarationRange.First ); + TXT( "\nLast : " ); + UID( decl->u.DeclarationRange.Last ); + break; + + case TGSI_DECLARE_MASK: + TXT( "\nMask: " ); + UIX( decl->u.DeclarationMask.Mask ); + break; + + default: + assert( 0 ); + } + + if( decl->Declaration.Interpolate ) { + CHR( '\n' ); + TXT( "\nInterpolate: " ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Interpolation.Padding ); + } + } + + if( decl->Declaration.Semantic ) { + CHR( '\n' ); + TXT( "\nSemanticName : " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); + TXT( "\nSemanticIndex: " ); + UID( decl->Semantic.SemanticIndex ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Semantic.Padding ); + } + } +} + +static void +dump_immediate_short( + struct gen_dump *dump, + struct tgsi_full_immediate *imm ) +{ + unsigned i; + + TXT( "\nIMM " ); + ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); + + TXT( " { " ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + + if( i < imm->Immediate.Size - 2 ) { + TXT( ", " ); + } + } + TXT( " }" ); +} + +static void +dump_immediate_verbose( + struct gen_dump *dump, + struct tgsi_full_immediate *imm, + unsigned ignored ) +{ + unsigned i; + + TXT( "\nDataType : " ); + ENM( imm->Immediate.DataType, TGSI_IMMS ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( imm->Immediate.Padding ); + } + + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + CHR( '\n' ); + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + TXT( "\nFloat: " ); + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + } +} + +static void +dump_instruction_short( + struct gen_dump *dump, + struct tgsi_full_instruction *inst, + unsigned instno ) +{ + unsigned i; + boolean first_reg = TRUE; + + CHR( '\n' ); + UID( instno ); + CHR( ':' ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); + + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + TXT( "_SAT" ); + break; + case TGSI_SAT_MINUS_PLUS_ONE: + TXT( "_SAT[-1,1]" ); + break; + default: + assert( 0 ); + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( dst->DstRegister.Index ); + CHR( ']' ); + + if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { + CHR( '.' ); + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { + CHR( 'x' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { + CHR( 'y' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { + CHR( 'z' ); + } + if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { + CHR( 'w' ); + } + } + + first_reg = FALSE; + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + + if( !first_reg ) { + CHR( ',' ); + } + CHR( ' ' ); + + if( src->SrcRegisterExtMod.Complement ) { + TXT( "(1 - " ); + } + if( src->SrcRegisterExtMod.Negate ) { + CHR( '-' ); + } + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + if( src->SrcRegister.Negate ) { + CHR( '-' ); + } + + ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); + + CHR( '[' ); + SID( src->SrcRegister.Index ); + CHR( ']' ); + + if (src->SrcRegister.Extended) { + if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { + CHR( '.' ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); + } + } + else if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { + CHR( '.' ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); + } + + if( src->SrcRegisterExtMod.Absolute ) { + CHR( '|' ); + } + if( src->SrcRegisterExtMod.Complement ) { + CHR( ')' ); + } + + first_reg = FALSE; + } + + switch( inst->Instruction.Opcode ) { + case TGSI_OPCODE_IF: + case TGSI_OPCODE_ELSE: + case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_CAL: + TXT( " :" ); + UID( inst->InstructionExtLabel.Label ); + break; + } +} + +static void +dump_instruction_verbose( + struct gen_dump *dump, + struct tgsi_full_instruction *inst, + unsigned ignored, + unsigned deflt, + struct tgsi_full_instruction *fi ) +{ + unsigned i; + + TXT( "\nOpcode : " ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES ); + if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { + TXT( "\nSaturate : " ); + ENM( inst->Instruction.Saturate, TGSI_SATS ); + } + if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { + TXT( "\nNumDstRegs : " ); + UID( inst->Instruction.NumDstRegs ); + } + if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { + TXT( "\nNumSrcRegs : " ); + UID( inst->Instruction.NumSrcRegs ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->Instruction.Padding ); + } + + if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { + TXT( "\nPrecision : " ); + ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); + } + if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { + TXT( "\nCondDstIndex : " ); + UID( inst->InstructionExtNv.CondDstIndex ); + } + if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { + TXT( "\nCondFlowIndex : " ); + UID( inst->InstructionExtNv.CondFlowIndex ); + } + if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { + TXT( "\nCondMask : " ); + ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { + TXT( "\nCondSwizzleX : " ); + ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { + TXT( "\nCondSwizzleY : " ); + ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ : " ); + ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { + TXT( "\nCondSwizzleW : " ); + ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { + TXT( "\nCondDstUpdate : " ); + UID( inst->InstructionExtNv.CondDstUpdate ); + } + if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { + TXT( "\nCondFlowEnable: " ); + UID( inst->InstructionExtNv.CondFlowEnable ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtNv.Padding ); + if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { + TXT( "\nExtended : " ); + UID( inst->InstructionExtNv.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { + TXT( "\nLabel : " ); + UID( inst->InstructionExtLabel.Label ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtLabel.Padding ); + if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtLabel.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { + TXT( "\nTexture : " ); + ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtTexture.Padding ); + if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtTexture.Extended ); + } + } + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : " ); + ENM( dst->DstRegister.File, TGSI_FILES ); + if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { + TXT( "\nWriteMask: " ); + ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); + } + if( ignored ) { + if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( dst->DstRegister.Indirect ); + } + if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( dst->DstRegister.Dimension ); + } + } + if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { + TXT( "\nIndex : " ); + SID( dst->DstRegister.Index ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegister.Padding ); + if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegister.Extended ); + } + } + + if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { + TXT( "\nCondMask : " ); + ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { + TXT( "\nCondSwizzleX: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { + TXT( "\nCondSwizzleY: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { + TXT( "\nCondSwizzleW: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { + TXT( "\nCondSrcIndex: " ); + UID( dst->DstRegisterExtConcode.CondSrcIndex ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtConcode.Padding ); + if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegisterExtConcode.Extended ); + } + } + } + + if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { + TXT( "\nModulate: " ); + ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtModulate.Padding ); + if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { + TXT( "\nExtended: " ); + UID( dst->DstRegisterExtModulate.Extended ); + } + } + } + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; + + CHR( '\n' ); + TXT( "\nFile : "); + ENM( src->SrcRegister.File, TGSI_FILES ); + if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { + TXT( "\nSwizzleX : " ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { + TXT( "\nSwizzleY : " ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { + TXT( "\nSwizzleZ : " ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { + TXT( "\nSwizzleW : " ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegister.Negate ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( src->SrcRegister.Indirect ); + } + if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( src->SrcRegister.Dimension ); + } + } + if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { + TXT( "\nIndex : " ); + SID( src->SrcRegister.Index ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegister.Extended ); + } + } + + if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { + TXT( "\nExtSwizzleX: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { + TXT( "\nExtSwizzleY: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { + TXT( "\nExtSwizzleZ: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { + TXT( "\nExtSwizzleW: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { + TXT( "\nNegateX : " ); + UID( src->SrcRegisterExtSwz.NegateX ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { + TXT( "\nNegateY : " ); + UID( src->SrcRegisterExtSwz.NegateY ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { + TXT( "\nNegateZ : " ); + UID( src->SrcRegisterExtSwz.NegateZ ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { + TXT( "\nNegateW : " ); + UID( src->SrcRegisterExtSwz.NegateW ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { + TXT( "\nExtDivide : " ); + ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtSwz.Padding ); + if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtSwz.Extended ); + } + } + } + + if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { + CHR( '\n' ); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { + TXT( "\nComplement: " ); + UID( src->SrcRegisterExtMod.Complement ); + } + if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { + TXT( "\nBias : " ); + UID( src->SrcRegisterExtMod.Bias ); + } + if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { + TXT( "\nScale2X : " ); + UID( src->SrcRegisterExtMod.Scale2X ); + } + if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { + TXT( "\nAbsolute : " ); + UID( src->SrcRegisterExtMod.Absolute ); + } + if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegisterExtMod.Negate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtMod.Padding ); + if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtMod.Extended ); + } + } + } + } +} + +static void +dump_gen( + struct gen_dump *dump, + const struct tgsi_token *tokens, + unsigned flags ) +{ + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned verbose = flags & TGSI_DUMP_VERBOSE; + unsigned ignored = !(flags & TGSI_DUMP_NO_IGNORED); + unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); + unsigned instno = 0; + + dump->tabs = 0; + + /* sanity check */ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); + + tgsi_parse_init( &parse, tokens ); + + TXT( "tgsi-dump begin -----------------" ); + + CHR( '\n' ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); + CHR( ' ' ); + UID( parse.FullVersion.Version.MajorVersion ); + CHR( '.' ); + UID( parse.FullVersion.Version.MinorVersion ); + + if( verbose ) { + TXT( "\nMajorVersion: " ); + UID( parse.FullVersion.Version.MajorVersion ); + TXT( "\nMinorVersion: " ); + UID( parse.FullVersion.Version.MinorVersion ); + CHR( '\n' ); + + TXT( "\nHeaderSize: " ); + UID( parse.FullHeader.Header.HeaderSize ); + TXT( "\nBodySize : " ); + UID( parse.FullHeader.Header.BodySize ); + TXT( "\nProcessor : " ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + CHR( '\n' ); + } + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_short( + dump, + &parse.FullToken.FullDeclaration ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_short( + dump, + &parse.FullToken.FullImmediate ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_short( + dump, + &parse.FullToken.FullInstruction, + instno ); + instno++; + break; + + default: + assert( 0 ); + } + + if( verbose ) { + TXT( "\nType : " ); + ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); + if( ignored ) { + TXT( "\nSize : " ); + UID( parse.FullToken.Token.Size ); + if( deflt || parse.FullToken.Token.Extended ) { + TXT( "\nExtended : " ); + UID( parse.FullToken.Token.Extended ); + } + } + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_verbose( + dump, + &parse.FullToken.FullDeclaration, + ignored, + deflt, + &fd ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_verbose( + dump, + &parse.FullToken.FullImmediate, + ignored ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_verbose( + dump, + &parse.FullToken.FullInstruction, + ignored, + deflt, + &fi ); + break; + + default: + assert( 0 ); + } + + CHR( '\n' ); + } + } + + TXT( "\ntgsi-dump end -------------------\n" ); + + tgsi_parse_free( &parse ); +} + + +static void +sanity_checks(void) +{ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); + assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "END") == 0); +} + + +void +tgsi_dump( + const struct tgsi_token *tokens, + unsigned flags ) +{ + struct file_dump dump; + + sanity_checks(); + + dump.base.write = _file_dump_write; +#if 0 + { + static unsigned counter = 0; + char buffer[64]; + sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); + dump.file = fopen( buffer, "wt" ); + } +#else + dump.file = stderr; +#endif + + dump_gen( + &dump.base, + tokens, + flags ); + +#if 0 + fclose( dump.file ); +#endif +} + +void +tgsi_dump_str( + char **str, + const struct tgsi_token *tokens, + unsigned flags ) +{ + struct text_dump dump; + + dump.base.write = _text_dump_write; + dump.text = NULL; + dump.length = 0; + dump.capacity = 0; + + dump_gen( + &dump.base, + tokens, + flags ); + + *str = dump.text; +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h new file mode 100644 index 0000000000..1adc9db251 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h @@ -0,0 +1,28 @@ +#if !defined TGSI_DUMP_H +#define TGSI_DUMP_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +#define TGSI_DUMP_VERBOSE 1 +#define TGSI_DUMP_NO_IGNORED 2 +#define TGSI_DUMP_NO_DEFAULT 4 + +void +tgsi_dump( + const struct tgsi_token *tokens, + unsigned flags ); + +void +tgsi_dump_str( + char **str, + const struct tgsi_token *tokens, + unsigned flags ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_DUMP_H + diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c new file mode 100644 index 0000000000..bf6b89ce56 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c @@ -0,0 +1,319 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ) +{ + full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; +} + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ) +{ + if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { + FREE( full_token->FullImmediate.u.Pointer ); + } +} + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ) +{ + ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; + if( ctx->FullVersion.Version.MajorVersion > 1 ) { + return TGSI_PARSE_ERROR; + } + + ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; + if( ctx->FullHeader.Header.HeaderSize >= 2 ) { + ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; + } + else { + ctx->FullHeader.Processor = tgsi_default_processor(); + } + + ctx->Tokens = tokens; + ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; + + tgsi_full_token_init( &ctx->FullToken ); + + return TGSI_PARSE_OK; +} + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ) +{ + tgsi_full_token_free( &ctx->FullToken ); +} + +boolean +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ) +{ + return ctx->Position >= + 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; +} + +static void +next_token( + struct tgsi_parse_context *ctx, + void *token ) +{ + assert( !tgsi_parse_end_of_tokens( ctx ) ); + + *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; +} + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ) +{ + struct tgsi_token token; + unsigned i; + + tgsi_full_token_free( &ctx->FullToken ); + tgsi_full_token_init( &ctx->FullToken ); + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; + + *decl = tgsi_default_full_declaration(); + decl->Declaration = *(struct tgsi_declaration *) &token; + + switch( decl->Declaration.Type ) { + case TGSI_DECLARE_RANGE: + next_token( ctx, &decl->u.DeclarationRange ); + break; + + case TGSI_DECLARE_MASK: + next_token( ctx, &decl->u.DeclarationMask ); + break; + + default: + assert (0); + } + + if( decl->Declaration.Interpolate ) { + next_token( ctx, &decl->Interpolation ); + } + + if( decl->Declaration.Semantic ) { + next_token( ctx, &decl->Semantic ); + } + + break; + } + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; + + *imm = tgsi_default_full_immediate(); + imm->Immediate = *(struct tgsi_immediate *) &token; + + assert( !imm->Immediate.Extended ); + + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + imm->u.Pointer = MALLOC( + sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + next_token( ctx, &imm->u.ImmediateFloat32[i] ); + } + break; + + default: + assert( 0 ); + } + + break; + } + + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; + unsigned extended; + + *inst = tgsi_default_full_instruction(); + inst->Instruction = *(struct tgsi_instruction *) &token; + + extended = inst->Instruction.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_INSTRUCTION_EXT_TYPE_NV: + inst->InstructionExtNv = + *(struct tgsi_instruction_ext_nv *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_LABEL: + inst->InstructionExtLabel = + *(struct tgsi_instruction_ext_label *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: + inst->InstructionExtTexture = + *(struct tgsi_instruction_ext_texture *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); + assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); + + extended = inst->FullDstRegisters[i].DstRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: + inst->FullDstRegisters[i].DstRegisterExtConcode = + *(struct tgsi_dst_register_ext_concode *) &token; + break; + + case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: + inst->FullDstRegisters[i].DstRegisterExtModulate = + *(struct tgsi_dst_register_ext_modulate *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + } + + assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); + + extended = inst->FullSrcRegisters[i].SrcRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: + inst->FullSrcRegisters[i].SrcRegisterExtSwz = + *(struct tgsi_src_register_ext_swz *) &token; + break; + + case TGSI_SRC_REGISTER_EXT_TYPE_MOD: + inst->FullSrcRegisters[i].SrcRegisterExtMod = + *(struct tgsi_src_register_ext_mod *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + + if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); + + /* + * No support for multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); + + if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + } + } + + break; + } + + default: + assert( 0 ); + } +} + diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h new file mode 100644 index 0000000000..9372da8d5d --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h @@ -0,0 +1,121 @@ +#if !defined TGSI_PARSE_H +#define TGSI_PARSE_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +struct tgsi_full_version +{ + struct tgsi_version Version; +}; + +struct tgsi_full_header +{ + struct tgsi_header Header; + struct tgsi_processor Processor; +}; + +struct tgsi_full_dst_register +{ + struct tgsi_dst_register DstRegister; + struct tgsi_dst_register_ext_concode DstRegisterExtConcode; + struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; +}; + +struct tgsi_full_src_register +{ + struct tgsi_src_register SrcRegister; + struct tgsi_src_register_ext_swz SrcRegisterExtSwz; + struct tgsi_src_register_ext_mod SrcRegisterExtMod; + struct tgsi_src_register SrcRegisterInd; + struct tgsi_dimension SrcRegisterDim; + struct tgsi_src_register SrcRegisterDimInd; +}; + +struct tgsi_full_declaration +{ + struct tgsi_declaration Declaration; + union + { + struct tgsi_declaration_range DeclarationRange; + struct tgsi_declaration_mask DeclarationMask; + } u; + struct tgsi_declaration_interpolation Interpolation; + struct tgsi_declaration_semantic Semantic; +}; + +struct tgsi_full_immediate +{ + struct tgsi_immediate Immediate; + union + { + void *Pointer; + struct tgsi_immediate_float32 *ImmediateFloat32; + } u; +}; + +#define TGSI_FULL_MAX_DST_REGISTERS 2 +#define TGSI_FULL_MAX_SRC_REGISTERS 3 + +struct tgsi_full_instruction +{ + struct tgsi_instruction Instruction; + struct tgsi_instruction_ext_nv InstructionExtNv; + struct tgsi_instruction_ext_label InstructionExtLabel; + struct tgsi_instruction_ext_texture InstructionExtTexture; + struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; + struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; +}; + +union tgsi_full_token +{ + struct tgsi_token Token; + struct tgsi_full_declaration FullDeclaration; + struct tgsi_full_immediate FullImmediate; + struct tgsi_full_instruction FullInstruction; +}; + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ); + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ); + +struct tgsi_parse_context +{ + const struct tgsi_token *Tokens; + unsigned Position; + struct tgsi_full_version FullVersion; + struct tgsi_full_header FullHeader; + union tgsi_full_token FullToken; +}; + +#define TGSI_PARSE_OK 0 +#define TGSI_PARSE_ERROR 1 + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ); + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ); + +boolean +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ); + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_PARSE_H + diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_transform.c b/src/gallium/auxiliary/tgsi/util/tgsi_transform.c new file mode 100644 index 0000000000..357f77b05a --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_transform.c @@ -0,0 +1,199 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI program transformation utility. + * + * Authors: Brian Paul + */ + + +#include "tgsi_transform.h" + + + +static void +emit_instruction(struct tgsi_transform_context *ctx, + const struct tgsi_full_instruction *inst) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_instruction(inst, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + +static void +emit_declaration(struct tgsi_transform_context *ctx, + const struct tgsi_full_declaration *decl) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_declaration(decl, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + +static void +emit_immediate(struct tgsi_transform_context *ctx, + const struct tgsi_full_immediate *imm) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_immediate(imm, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + + +/** + * Apply user-defined transformations to the input shader to produce + * the output shader. + * For example, a register search-and-replace operation could be applied + * by defining a transform_instruction() callback that examined and changed + * the instruction src/dest regs. + * + * \return number of tokens emitted + */ +int +tgsi_transform_shader(const struct tgsi_token *tokens_in, + struct tgsi_token *tokens_out, + uint max_tokens_out, + struct tgsi_transform_context *ctx) +{ + uint procType; + + /* input shader */ + struct tgsi_parse_context parse; + + /* output shader */ + struct tgsi_processor *processor; + + + /** + ** callback context init + **/ + ctx->emit_instruction = emit_instruction; + ctx->emit_declaration = emit_declaration; + ctx->emit_immediate = emit_immediate; + ctx->tokens_out = tokens_out; + ctx->max_tokens_out = max_tokens_out; + + + /** + ** Setup to begin parsing input shader + **/ + if (tgsi_parse_init( &parse, tokens_in ) != TGSI_PARSE_OK) { + debug_printf("tgsi_parse_init() failed in tgsi_transform_shader()!\n"); + return -1; + } + procType = parse.FullHeader.Processor.Processor; + assert(procType == TGSI_PROCESSOR_FRAGMENT || + procType == TGSI_PROCESSOR_VERTEX || + procType == TGSI_PROCESSOR_GEOMETRY); + + + /** + ** Setup output shader + **/ + *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version(); + + ctx->header = (struct tgsi_header *) (tokens_out + 1); + *ctx->header = tgsi_build_header(); + + processor = (struct tgsi_processor *) (tokens_out + 2); + *processor = tgsi_build_processor( procType, ctx->header ); + + ctx->ti = 3; + + + /** + ** Loop over incoming program tokens/instructions + */ + while( !tgsi_parse_end_of_tokens( &parse ) ) { + + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *fullinst + = &parse.FullToken.FullInstruction; + + if (ctx->transform_instruction) + ctx->transform_instruction(ctx, fullinst); + else + ctx->emit_instruction(ctx, fullinst); + } + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *fulldecl + = &parse.FullToken.FullDeclaration; + + if (ctx->transform_declaration) + ctx->transform_declaration(ctx, fulldecl); + else + ctx->emit_declaration(ctx, fulldecl); + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *fullimm + = &parse.FullToken.FullImmediate; + + if (ctx->transform_immediate) + ctx->transform_immediate(ctx, fullimm); + else + ctx->emit_immediate(ctx, fullimm); + } + break; + + default: + assert( 0 ); + } + } + + if (ctx->epilog) { + ctx->epilog(ctx); + } + + tgsi_parse_free (&parse); + + return ctx->ti; +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_transform.h b/src/gallium/auxiliary/tgsi/util/tgsi_transform.h new file mode 100644 index 0000000000..fcf85d603b --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_transform.h @@ -0,0 +1,93 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_TRANSFORM_H +#define TGSI_TRANSFORM_H + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_build.h" + + + +/** + * Subclass this to add caller-specific data + */ +struct tgsi_transform_context +{ +/**** PUBLIC ***/ + + /** + * User-defined callbacks invoked per instruction. + */ + void (*transform_instruction)(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst); + + void (*transform_declaration)(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl); + + void (*transform_immediate)(struct tgsi_transform_context *ctx, + struct tgsi_full_immediate *imm); + + /** + * Called at end of input program to allow caller to append extra + * instructions. Return number of tokens emitted. + */ + void (*epilog)(struct tgsi_transform_context *ctx); + + +/*** PRIVATE ***/ + + /** + * These are setup by tgsi_transform_shader() and cannot be overridden. + * Meant to be called from in the above user callback functions. + */ + void (*emit_instruction)(struct tgsi_transform_context *ctx, + const struct tgsi_full_instruction *inst); + void (*emit_declaration)(struct tgsi_transform_context *ctx, + const struct tgsi_full_declaration *decl); + void (*emit_immediate)(struct tgsi_transform_context *ctx, + const struct tgsi_full_immediate *imm); + + struct tgsi_header *header; + uint max_tokens_out; + struct tgsi_token *tokens_out; + uint ti; +}; + + + +extern int +tgsi_transform_shader(const struct tgsi_token *tokens_in, + struct tgsi_token *tokens_out, + uint max_tokens_out, + struct tgsi_transform_context *ctx); + + +#endif /* TGSI_TRANSFORM_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_util.c b/src/gallium/auxiliary/tgsi/util/tgsi_util.c new file mode 100644 index 0000000000..4cdd89182a --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_util.c @@ -0,0 +1,274 @@ +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" +#include "tgsi_util.h" + +union pointer_hack +{ + void *pointer; + unsigned long long uint64; +}; + +void * +tgsi_align_128bit( + void *unaligned ) +{ + union pointer_hack ph; + + ph.uint64 = 0; + ph.pointer = unaligned; + ph.uint64 = (ph.uint64 + 15) & ~15; + return ph.pointer; +} + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->SwizzleX; + case 1: + return reg->SwizzleY; + case 2: + return reg->SwizzleZ; + case 3: + return reg->SwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->ExtSwizzleX; + case 1: + return reg->ExtSwizzleY; + case 2: + return reg->ExtSwizzleZ; + case 3: + return reg->ExtSwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned swizzle; + + /* + * First, calculate the extended swizzle for a given channel. This will give + * us either a channel index into the simple swizzle or a constant 1 or 0. + */ + swizzle = tgsi_util_get_src_register_extswizzle( + ®->SrcRegisterExtSwz, + component ); + + assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); + assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); + assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); + assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); + + /* + * Second, calculate the simple swizzle for the unswizzled channel index. + * Leave the constants intact, they are not affected by the simple swizzle. + */ + if( swizzle <= TGSI_SWIZZLE_W ) { + swizzle = tgsi_util_get_src_register_swizzle( + ®->SrcRegister, + component ); + } + + return swizzle; +} + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + unsigned swizzle, + unsigned component ) +{ + switch( component ) { + case 0: + reg->SwizzleX = swizzle; + break; + case 1: + reg->SwizzleY = swizzle; + break; + case 2: + reg->SwizzleZ = swizzle; + break; + case 3: + reg->SwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + unsigned swizzle, + unsigned component ) +{ + switch( component ) { + case 0: + reg->ExtSwizzleX = swizzle; + break; + case 1: + reg->ExtSwizzleY = swizzle; + break; + case 2: + reg->ExtSwizzleZ = swizzle; + break; + case 3: + reg->ExtSwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->NegateX; + case 1: + return reg->NegateY; + case 2: + return reg->NegateZ; + case 3: + return reg->NegateW; + default: + assert( 0 ); + } + return 0; +} + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ) +{ + switch( component ) { + case 0: + reg->NegateX = negate; + break; + case 1: + reg->NegateY = negate; + break; + case 2: + reg->NegateZ = negate; + break; + case 3: + reg->NegateW = negate; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned sign_mode; + + if( reg->SrcRegisterExtMod.Absolute ) { + /* Consider only the post-abs negation. */ + + if( reg->SrcRegisterExtMod.Negate ) { + sign_mode = TGSI_UTIL_SIGN_SET; + } + else { + sign_mode = TGSI_UTIL_SIGN_CLEAR; + } + } + else { + /* Accumulate the three negations. */ + + unsigned negate; + + negate = reg->SrcRegister.Negate; + if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { + negate = !negate; + } + if( reg->SrcRegisterExtMod.Negate ) { + negate = !negate; + } + + if( negate ) { + sign_mode = TGSI_UTIL_SIGN_TOGGLE; + } + else { + sign_mode = TGSI_UTIL_SIGN_KEEP; + } + } + + return sign_mode; +} + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + unsigned sign_mode ) +{ + reg->SrcRegisterExtSwz.NegateX = 0; + reg->SrcRegisterExtSwz.NegateY = 0; + reg->SrcRegisterExtSwz.NegateZ = 0; + reg->SrcRegisterExtSwz.NegateW = 0; + + switch (sign_mode) + { + case TGSI_UTIL_SIGN_CLEAR: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_SET: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 1; + break; + + case TGSI_UTIL_SIGN_TOGGLE: + reg->SrcRegister.Negate = 1; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_KEEP: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + default: + assert( 0 ); + } +} + diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_util.h b/src/gallium/auxiliary/tgsi/util/tgsi_util.h new file mode 100644 index 0000000000..ef14446f0e --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_util.h @@ -0,0 +1,70 @@ +#if !defined TGSI_UTIL_H +#define TGSI_UTIL_H + +#if defined __cplusplus +extern "C" { +#endif // defined __cplusplus + +void * +tgsi_align_128bit( + void *unaligned ); + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component); + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + unsigned swizzle, + unsigned component ); + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + unsigned swizzle, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ); + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ); + +#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ +#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ +#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ +#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + unsigned sign_mode ); + +#if defined __cplusplus +} // extern "C" +#endif // defined __cplusplus + +#endif // !defined TGSI_UTIL_H + diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c new file mode 100644 index 0000000000..b9607a6ba7 --- /dev/null +++ b/src/gallium/auxiliary/util/p_debug.c @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include + +#ifdef WIN32 +#include +#include +#else +#include +#include +#endif + +#include "pipe/p_debug.h" +#include "pipe/p_compiler.h" + + +void debug_vprintf(const char *format, va_list ap) +{ +#ifdef WIN32 + EngDebugPrint("Gallium3D: ", (PCHAR)format, ap); +#else + vfprintf(stderr, format, ap); +#endif +} + + +void debug_printf(const char *format, ...) +{ + va_list ap; + va_start(ap, format); + debug_vprintf(format, ap); + va_end(ap); +} + + +static INLINE void debug_abort(void) +{ +#ifdef WIN32 + EngDebugBreak(); +#else + abort(); +#endif +} + + +void debug_assert_fail(const char *expr, const char *file, unsigned line) +{ + debug_printf("%s:%i: Assertion `%s' failed.\n", file, line, expr); + debug_abort(); +} diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c new file mode 100644 index 0000000000..3f795a3898 --- /dev/null +++ b/src/gallium/auxiliary/util/p_tile.c @@ -0,0 +1,699 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * RGBA/float tile get/put functions. + * Usable both by drivers and state trackers. + * Surfaces should already be in a mapped state. + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" + +#include "p_tile.h" + + + +/** + * Move raw block of pixels from surface to user memory. + * This should be usable by any hw driver that has mappable surfaces. + */ +void +pipe_get_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride) +{ + const uint cpp = ps->cpp; + const ubyte *pSrc; + const uint src_stride = ps->pitch * cpp; + ubyte *pDest; + uint i; + + if (dst_stride == 0) { + dst_stride = w * cpp; + } + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + pSrc = (const ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; + pDest = (ubyte *) p; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w * cpp); + pDest += dst_stride; + pSrc += src_stride; + } + + pipe_surface_unmap(ps); +} + + +/** + * Move raw block of pixels from user memory to surface. + * This should be usable by any hw driver that has mappable surfaces. + */ +void +pipe_put_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride) +{ + const uint cpp = ps->cpp; + const ubyte *pSrc; + const uint dst_stride = ps->pitch * cpp; + ubyte *pDest; + uint i; + + if (src_stride == 0) { + src_stride = w * cpp; + } + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + pSrc = (const ubyte *) p; + pDest = (ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; + + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, w * cpp); + pDest += dst_stride; + pSrc += src_stride; + } + + pipe_surface_unmap(ps); +} + + + + +/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */ +#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) + +#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ + us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) + + + +/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ + +static void +a8r8g8b8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + } + p += dst_stride; + } +} + + +static void +a8r8g8b8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + *dst++ = (a << 24) | (r << 16) | (g << 8) | b; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ + +static void +b8g8r8a8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); + pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + } + p += dst_stride; + } +} + + +static void +b8g8r8a8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + *dst++ = (b << 24) | (g << 16) | (r << 8) | a; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ + +static void +a1r5g5b5_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); + pRow[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); + pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + pRow[3] = ((pixel >> 15) ) * 1.0f; + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ + +static void +a4r4g4b4_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); + pRow[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); + pRow[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); + pRow[3] = ((pixel >> 12) ) * (1.0f / 15.0f); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_R5G6B5_UNORM ***/ + +static void +r5g6b5_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); + pRow[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); + pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + pRow[3] = 1.0f; + } + p += dst_stride; + } +} + + +static void +r5g5b5_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); + uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); + uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); + *dst++ = (r << 11) | (g << 5) | (b); + } + p += src_stride; + } +} + + + +/*** PIPE_FORMAT_Z16_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z16_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const float scale = 1.0f / 65535.0f; + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = *src++ * scale; + } + p += dst_stride; + } +} + + + + +/*** PIPE_FORMAT_U_L8 ***/ + +static void +l8_get_tile_rgba(ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(*src); + pRow[3] = 1.0; + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_U_A8 ***/ + +static void +a8_get_tile_rgba(ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = 0.0; + pRow[3] = UBYTE_TO_FLOAT(*src); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ + +static void +r16g16b16a16_get_tile_rgba(short *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src += 4, pRow += 4) { + pRow[0] = SHORT_TO_FLOAT(src[0]); + pRow[1] = SHORT_TO_FLOAT(src[1]); + pRow[2] = SHORT_TO_FLOAT(src[2]); + pRow[3] = SHORT_TO_FLOAT(src[3]); + } + p += dst_stride; + } +} + + +static void +r16g16b16a16_put_tile_rgba(short *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, dst += 4, pRow += 4) { + UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); + UNCLAMPED_FLOAT_TO_SHORT(dst[1], pRow[1]); + UNCLAMPED_FLOAT_TO_SHORT(dst[2], pRow[2]); + UNCLAMPED_FLOAT_TO_SHORT(dst[3], pRow[3]); + } + p += src_stride; + } +} + + + +/*** PIPE_FORMAT_U_I8 ***/ + +static void +i8_get_tile_rgba(ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = UBYTE_TO_FLOAT(*src); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_U_A8_L8 ***/ + +static void +a8_l8_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + ushort p = *src++; + pRow[0] = + pRow[1] = + pRow[2] = UBYTE_TO_FLOAT(p & 0xff); + pRow[3] = UBYTE_TO_FLOAT(p >> 8); + } + p += dst_stride; + } +} + + + + +/*** PIPE_FORMAT_Z32_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z32_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const double scale = 1.0 / (double) 0xffffffff; + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (*src++ * scale); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_S8Z24_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +s8z24_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (scale * (*src++ & 0xffffff)); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_Z24S8_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +z24s8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (scale * (*src++ >> 8)); + } + p += dst_stride; + } +} + + +void +pipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p) +{ + unsigned dst_stride = w * 4; + void *packed; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + packed = MALLOC(h * w * ps->cpp); + + if (!packed) + return; + + pipe_get_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); + + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + a1r5g5b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_A4R4G4B4_UNORM: + a4r4g4b4_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g6b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_U_L8: + l8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_U_A8: + a8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_U_I8: + i8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_U_A8_L8: + a8_l8_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_get_tile_rgba((short *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_Z16_UNORM: + z16_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_Z32_UNORM: + z32_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_S8Z24_UNORM: + s8z24_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + case PIPE_FORMAT_Z24S8_UNORM: + z24s8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; + default: + assert(0); + } + + FREE(packed); +} + + +void +pipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p) +{ + unsigned src_stride = w * 4; + void *packed; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + packed = MALLOC(h * w * ps->cpp); + + if (!packed) + return; + + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + /*a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + break; + case PIPE_FORMAT_U_L8: + /*l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_U_A8: + /*a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_U_I8: + /*i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_U_A8_L8: + /*a8_l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_Z16_UNORM: + /*z16_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_Z32_UNORM: + /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_S8Z24_UNORM: + /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_Z24S8_UNORM: + /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ + break; + default: + assert(0); + } + + pipe_put_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); + + FREE(packed); +} diff --git a/src/gallium/auxiliary/util/p_tile.h b/src/gallium/auxiliary/util/p_tile.h new file mode 100644 index 0000000000..cd8124bf11 --- /dev/null +++ b/src/gallium/auxiliary/util/p_tile.h @@ -0,0 +1,81 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_TILE_H +#define P_TILE_H + +#include "pipe/p_compiler.h" + +struct pipe_context; +struct pipe_surface; + + +/** + * Clip tile against surface dims. + * \return TRUE if tile is totally clipped, FALSE otherwise + */ +static INLINE boolean +pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps) +{ + if (x >= ps->width) + return TRUE; + if (y >= ps->height) + return TRUE; + if (x + *w > ps->width) + *w = ps->width - x; + if (y + *h > ps->height) + *h = ps->height - y; + return FALSE; +} + + +extern void +pipe_get_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride); + +extern void +pipe_put_tile_raw(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride); + + +extern void +pipe_get_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p); + +extern void +pipe_put_tile_rgba(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p); + +#endif diff --git a/src/gallium/auxiliary/util/p_util.c b/src/gallium/auxiliary/util/p_util.c new file mode 100644 index 0000000000..2a92f8e408 --- /dev/null +++ b/src/gallium/auxiliary/util/p_util.c @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Miscellaneous utility functions. + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" + + +/** + * Copy 2D rect from one place to another. + * Position and sizes are in pixels. + */ +void +pipe_copy_rect(ubyte * dst, + unsigned cpp, + unsigned dst_pitch, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + int src_pitch, + unsigned src_x, + int src_y) +{ + unsigned i; + + dst_pitch *= cpp; + src_pitch *= cpp; + dst += dst_x * cpp; + src += src_x * cpp; + dst += dst_y * dst_pitch; + src += src_y * src_pitch; + width *= cpp; + + if (width == dst_pitch && width == src_pitch) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_pitch; + src += src_pitch; + } + } +} -- cgit v1.2.3 From 66f22aa3bf7fa546e946b45156aa578e202982c9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Feb 2008 20:11:40 +0900 Subject: Code reorganization: s/aux/auxiliary/ -- update build. --- src/gallium/Makefile | 2 +- src/gallium/Makefile.template | 2 +- src/gallium/auxiliary/llvm/Makefile | 2 +- src/gallium/drivers/cell/ppu/Makefile | 2 +- src/gallium/drivers/cell/spu/Makefile | 2 +- src/gallium/winsys/dri/Makefile.template | 2 +- src/mesa/Makefile | 2 +- src/mesa/sources | 66 ++++++++++++++++---------------- 8 files changed, 40 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/gallium/Makefile b/src/gallium/Makefile index a13b9a52d3..89e068a449 100644 --- a/src/gallium/Makefile +++ b/src/gallium/Makefile @@ -2,7 +2,7 @@ TOP = ../.. include $(TOP)/configs/current -SUBDIRS = aux drivers +SUBDIRS = auxiliary drivers default: subdirs diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template index 0717ed8dd2..83b25f9b47 100644 --- a/src/gallium/Makefile.template +++ b/src/gallium/Makefile.template @@ -17,7 +17,7 @@ INCLUDES = \ -I. \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/gallium/include/pipe \ - -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers \ -I$(TOP)/src/mesa \ -I$(TOP)/include \ diff --git a/src/gallium/auxiliary/llvm/Makefile b/src/gallium/auxiliary/llvm/Makefile index e6ac399d08..e0abf860c1 100644 --- a/src/gallium/auxiliary/llvm/Makefile +++ b/src/gallium/auxiliary/llvm/Makefile @@ -31,7 +31,7 @@ OBJECTS = $(C_SOURCES:.c=.o) \ INCLUDES = \ -I. \ -I$(TOP)/src/gallium/drivers - -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/mesa \ -I$(TOP)/include diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile index 011863c11e..a4c3f29e8a 100644 --- a/src/gallium/drivers/cell/ppu/Makefile +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -43,7 +43,7 @@ OBJECTS = $(SOURCES:.c=.o) \ INCLUDE_DIRS = \ -I$(TOP)/src/mesa \ -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers .c.o: diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile index 7aa947299e..30ef2450ec 100644 --- a/src/gallium/drivers/cell/spu/Makefile +++ b/src/gallium/drivers/cell/spu/Makefile @@ -34,7 +34,7 @@ SPU_ASM_OUT = $(SOURCES:.c=.s) \ INCLUDE_DIRS = \ -I$(TOP)/src/mesa \ -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers diff --git a/src/gallium/winsys/dri/Makefile.template b/src/gallium/winsys/dri/Makefile.template index b96305c094..2a261ed669 100644 --- a/src/gallium/winsys/dri/Makefile.template +++ b/src/gallium/winsys/dri/Makefile.template @@ -49,7 +49,7 @@ SHARED_INCLUDES = \ -I$(TOP)/include \ -I$(TOP)/include/GL/internal \ -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/aux \ + -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers \ -I$(TOP)/src/mesa \ -I$(TOP)/src/mesa/main \ diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 561608fedd..c8cb2b592f 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -21,7 +21,7 @@ CELL_LIB_SPU = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a endif ifeq ($(CONFIG_NAME), linux-llvm) -LLVM_LIB = $(TOP)/src/gallium/aux/llvm/libgallivm.a +LLVM_LIB = $(TOP)/src/gallium/auxiliary/llvm/libgallivm.a endif .SUFFIXES : .cpp diff --git a/src/mesa/sources b/src/mesa/sources index 2d07738210..cecd8a830f 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -158,45 +158,45 @@ VF_SOURCES = \ DRAW_SOURCES = \ - $(TOP)/src/gallium/aux/draw/draw_clip.c \ - $(TOP)/src/gallium/aux/draw/draw_context.c\ - $(TOP)/src/gallium/aux/draw/draw_cull.c \ - $(TOP)/src/gallium/aux/draw/draw_debug.c \ - $(TOP)/src/gallium/aux/draw/draw_flatshade.c \ - $(TOP)/src/gallium/aux/draw/draw_offset.c \ - $(TOP)/src/gallium/aux/draw/draw_prim.c \ - $(TOP)/src/gallium/aux/draw/draw_stipple.c \ - $(TOP)/src/gallium/aux/draw/draw_twoside.c \ - $(TOP)/src/gallium/aux/draw/draw_unfilled.c \ - $(TOP)/src/gallium/aux/draw/draw_validate.c \ - $(TOP)/src/gallium/aux/draw/draw_vbuf.c \ - $(TOP)/src/gallium/aux/draw/draw_vertex.c \ - $(TOP)/src/gallium/aux/draw/draw_vertex_cache.c \ - $(TOP)/src/gallium/aux/draw/draw_vertex_fetch.c \ - $(TOP)/src/gallium/aux/draw/draw_vertex_shader.c \ - $(TOP)/src/gallium/aux/draw/draw_vf.c \ - $(TOP)/src/gallium/aux/draw/draw_vf_generic.c \ - $(TOP)/src/gallium/aux/draw/draw_vf_sse.c \ - $(TOP)/src/gallium/aux/draw/draw_wide_prims.c + $(TOP)/src/gallium/auxiliary/draw/draw_clip.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_context.c\ + $(TOP)/src/gallium/auxiliary/draw/draw_cull.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_debug.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_flatshade.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_offset.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_prim.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_stipple.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_twoside.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_unfilled.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_validate.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vbuf.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vertex.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vertex_cache.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vertex_fetch.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vertex_shader.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vf.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vf_generic.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vf_sse.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_wide_prims.c TGSIEXEC_SOURCES = \ - $(TOP)/src/gallium/aux/tgsi/exec/tgsi_exec.c \ - $(TOP)/src/gallium/aux/tgsi/exec/tgsi_sse2.c + $(TOP)/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c \ + $(TOP)/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c TGSIUTIL_SOURCES = \ - $(TOP)/src/gallium/aux/tgsi/util/tgsi_build.c \ - $(TOP)/src/gallium/aux/tgsi/util/tgsi_dump.c \ - $(TOP)/src/gallium/aux/tgsi/util/tgsi_parse.c \ - $(TOP)/src/gallium/aux/tgsi/util/tgsi_util.c + $(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_build.c \ + $(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_dump.c \ + $(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_parse.c \ + $(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_util.c STATECACHE_SOURCES = \ - $(TOP)/src/gallium/aux/cso_cache/cso_hash.c \ - $(TOP)/src/gallium/aux/cso_cache/cso_cache.c + $(TOP)/src/gallium/auxiliary/cso_cache/cso_hash.c \ + $(TOP)/src/gallium/auxiliary/cso_cache/cso_cache.c PIPEUTIL_SOURCES = \ - $(TOP)/src/gallium/aux/util/p_debug.c \ - $(TOP)/src/gallium/aux/util/p_tile.c \ - $(TOP)/src/gallium/aux/util/p_util.c + $(TOP)/src/gallium/auxiliary/util/p_debug.c \ + $(TOP)/src/gallium/auxiliary/util/p_tile.c \ + $(TOP)/src/gallium/auxiliary/util/p_util.c STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ @@ -428,7 +428,7 @@ INCLUDE_DIRS = \ -I$(TOP)/src/mesa/main \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/src/gallium/aux + -I$(TOP)/src/gallium/auxiliary OLD_INCLUDE_DIRS = \ -I$(TOP)/src/mesa/tnl \ @@ -438,4 +438,4 @@ OLD_INCLUDE_DIRS = \ -I$(TOP)/src/mesa/shader \ -I$(TOP)/src/mesa/shader/grammar \ -I$(TOP)/src/mesa/shader/slang \ - -I$(TOP)/s$(TOP)/src/gallium/aux/tgsi + -I$(TOP)/s$(TOP)/src/gallium/auxiliary/tgsi -- cgit v1.2.3 From e822e09b89407d6cb8cd4a79e1c5c1e0955caf64 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 15 Feb 2008 13:35:46 +0000 Subject: softpipe: rename some functions to disambiguate --- src/gallium/drivers/softpipe/sp_fs_sse.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 28c5d8c556..d90066e025 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -100,7 +100,7 @@ sp_setup_pos_vector(const struct tgsi_interp_coef *coef, static void -sse_prepare( struct sp_fragment_shader *base, +fs_sse_prepare( struct sp_fragment_shader *base, struct tgsi_exec_machine *machine, struct tgsi_sampler *samplers ) { @@ -113,7 +113,7 @@ sse_prepare( struct sp_fragment_shader *base, * TODO: process >1 quad at a time */ static unsigned -sse_run( struct sp_fragment_shader *base, +fs_sse_run( struct sp_fragment_shader *base, struct tgsi_exec_machine *machine, struct quad_header *quad ) { @@ -137,7 +137,7 @@ sse_run( struct sp_fragment_shader *base, static void -sse_delete( struct sp_fragment_shader *base ) +fs_sse_delete( struct sp_fragment_shader *base ) { struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base; @@ -170,9 +170,9 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe, assert(shader->func); shader->base.shader = *templ; - shader->base.prepare = sse_prepare; - shader->base.run = sse_run; - shader->base.delete = sse_delete; + shader->base.prepare = fs_sse_prepare; + shader->base.run = fs_sse_run; + shader->base.delete = fs_sse_delete; return &shader->base; } -- cgit v1.2.3 From b29d8d27292c2ad956d3f0a307603f00ee01af28 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 15 Feb 2008 13:37:01 +0000 Subject: draw: subclass vertex shaders according to execution method Create new files for shaders compiled/executed with llvm, sse, exec respectively --- src/gallium/auxiliary/draw/Makefile | 2 +- src/gallium/auxiliary/draw/draw_private.h | 40 ++-- src/gallium/auxiliary/draw/draw_vertex_shader.c | 229 ++------------------- src/gallium/auxiliary/draw/draw_vs.h | 50 +++++ src/gallium/auxiliary/draw/draw_vs_exec.c | 186 ++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs_llvm.c | 237 ++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs_sse.c | 251 ++++++++++++++++++++++++ src/mesa/sources | 3 + 8 files changed, 766 insertions(+), 232 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_vs.h create mode 100644 src/gallium/auxiliary/draw/draw_vs_exec.c create mode 100644 src/gallium/auxiliary/draw/draw_vs_llvm.c create mode 100644 src/gallium/auxiliary/draw/draw_vs_sse.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 451911a354..fe9b150f30 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -1,2 +1,2 @@ default: - cd .. ; make + cd ../../../mesa ; make diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 3d09aef87c..bc11259cb2 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -128,13 +128,25 @@ struct draw_stage * Private version of the compiled vertex_shader */ struct draw_vertex_shader { + + /* This member will disappear shortly: + */ const struct pipe_shader_state *state; -#if defined(__i386__) || defined(__386__) - struct x86_function sse2_program; -#endif -#ifdef MESA_LLVM - struct gallivm_prog *llvm_prog; -#endif + + void (*prepare)( struct draw_vertex_shader *shader, + struct draw_context *draw ); + + /* Run the shader - this interface will get cleaned up in the + * future: + */ + void (*run)( struct draw_vertex_shader *shader, + struct draw_context *draw, + const unsigned *elts, + unsigned count, + struct vertex_header *vOut[] ); + + + void (*delete)( struct draw_vertex_shader * ); }; @@ -176,7 +188,7 @@ struct draw_context struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; - const struct draw_vertex_shader *vertex_shader; + struct draw_vertex_shader *vertex_shader; uint num_vs_outputs; /**< convenience, from vertex_shader */ @@ -201,6 +213,7 @@ struct draw_context boolean convert_wide_points; /**< convert wide points to tris? */ boolean convert_wide_lines; /**< convert side lines to tris? */ + boolean use_sse; unsigned reduced_prim; @@ -255,11 +268,10 @@ struct draw_context unsigned queue_nr; } pq; - int use_sse : 1; -#ifdef MESA_LLVM - struct gallivm_cpu_engine *engine; -#endif - + + /* This (and the tgsi_exec_machine struct) probably need to be moved somewhere private. + */ + struct gallivm_cpu_engine *engine; void *driver_private; }; @@ -290,11 +302,7 @@ extern void draw_vertex_cache_invalidate( struct draw_context *draw ); extern void draw_vertex_cache_unreference( struct draw_context *draw ); extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); - extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); -#ifdef MESA_LLVM -extern void draw_vertex_shader_queue_flush_llvm( struct draw_context *draw ); -#endif struct tgsi_exec_machine; diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index 9413f8b43a..f68f6e3244 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -33,177 +33,10 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#if defined(__i386__) || defined(__386__) -#include "tgsi/exec/tgsi_sse2.h" -#endif #include "draw_private.h" #include "draw_context.h" +#include "draw_vs.h" -#include "x86/rtasm/x86sse.h" -#include "llvm/gallivm.h" - - -#define DBG_VS 0 - - -static INLINE unsigned -compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) -{ - unsigned mask = 0; - unsigned i; - - /* Do the hardwired planes first: - */ - if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; - if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; - if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; - if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; - if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; - if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; - - /* Followed by any remaining ones: - */ - for (i = 6; i < nr; i++) { - if (dot4(clip, plane[i]) < 0) - mask |= (1<machine; - unsigned int j; - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - - assert(count <= 4); - assert(draw->vertex_shader->state->output_semantic_name[0] - == TGSI_SEMANTIC_POSITION); - - /* Consts does not require 16 byte alignment. */ - machine->Consts = (float (*)[4]) draw->user.constants; - - machine->Inputs = ALIGN16_ASSIGN(inputs); - machine->Outputs = ALIGN16_ASSIGN(outputs); - - draw->vertex_fetch.fetch_func( draw, machine, elts, count ); - - /* run shader */ -#ifdef MESA_LLVM - if (1) { - struct gallivm_prog *prog = draw->vertex_shader->llvm_prog; - gallivm_cpu_vs_exec(prog, - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps); - } else -#elif defined(__i386__) || defined(__386__) - if (draw->use_sse) { - /* SSE */ - /* cast away const */ - struct draw_vertex_shader *shader - = (struct draw_vertex_shader *)draw->vertex_shader; - codegen_function func - = (codegen_function) x86_get_func( &shader->sse2_program ); - - if (func) - func( - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps ); - else - /* interpreter */ - tgsi_exec_machine_run( machine ); - } - else -#endif - { - /* interpreter */ - tgsi_exec_machine_run( machine ); - } - - /* store machine results */ - for (j = 0; j < count; j++) { - unsigned slot; - float x, y, z, w; - - /* Handle attr[0] (position) specially: - * - * XXX: Computing the clipmask should be done in the vertex - * program as a set of DP4 instructions appended to the - * user-provided code. - */ - x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; - -#if DBG_VS - debug_printf("output[%d]win: %f %f %f %f\n", j, - vOut[j]->data[0][0], - vOut[j]->data[0][1], - vOut[j]->data[0][2], - vOut[j]->data[0][3]); -#endif - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; -#if DBG_VS - debug_printf("output[%d][%d]: %f %f %f %f\n", j, slot, - vOut[j]->data[slot][0], - vOut[j]->data[slot][1], - vOut[j]->data[slot][2], - vOut[j]->data[slot][3]); -#endif - } - } /* loop over vertices */ -} /** @@ -213,13 +46,14 @@ run_vertex_program(struct draw_context *draw, void draw_vertex_shader_queue_flush(struct draw_context *draw) { + struct draw_vertex_shader *shader = draw->vertex_shader; unsigned i; assert(draw->vs.queue_nr != 0); /* XXX: do this on statechange: */ - draw_update_vertex_fetch( draw ); + shader->prepare( shader, draw ); // fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); @@ -242,7 +76,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) assert(n > 0); assert(n <= 4); - run_vertex_program(draw, elts, n, dests); + shader->run(shader, draw, elts, n, dests); } draw->vs.queue_nr = 0; @@ -255,43 +89,16 @@ draw_create_vertex_shader(struct draw_context *draw, { struct draw_vertex_shader *vs; - vs = CALLOC_STRUCT( draw_vertex_shader ); - if (vs == NULL) { - return NULL; - } - - vs->state = shader; + vs = draw_create_vs_llvm( draw, shader ); + if (vs) + return vs; -#ifdef MESA_LLVM - struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); - gallivm_ir_set_layout(ir, GALLIVM_SOA); - gallivm_ir_set_components(ir, 4); - gallivm_ir_fill_from_tgsi(ir, shader->tokens); - vs->llvm_prog = gallivm_ir_compile(ir); - gallivm_ir_delete(ir); - - draw->engine = gallivm_global_cpu_engine(); - if (!draw->engine) { - draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); - } - else { - gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); - } -#elif defined(__i386__) || defined(__386__) - if (draw->use_sse) { - /* cast-away const */ - struct pipe_shader_state *sh = (struct pipe_shader_state *) shader; - - x86_init_func( &vs->sse2_program ); - if (!tgsi_emit_sse2( (struct tgsi_token *) sh->tokens, - &vs->sse2_program )) { - x86_release_func( (struct x86_function *) &vs->sse2_program ); - fprintf(stdout /*err*/, - "tgsi_emit_sse2() failed, falling back to interpreter\n"); - } - } -#endif + vs = draw_create_vs_sse( draw, shader ); + if (vs) + return vs; + vs = draw_create_vs_exec( draw, shader ); + assert(vs); return vs; } @@ -307,11 +114,7 @@ draw_bind_vertex_shader(struct draw_context *draw, tgsi_exec_machine_init(&draw->machine); - /* specify the vertex program to interpret/execute */ - tgsi_exec_machine_bind_shader(&draw->machine, - draw->vertex_shader->state->tokens, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/ ); + dvs->prepare( dvs, draw ); } @@ -319,9 +122,5 @@ void draw_delete_vertex_shader(struct draw_context *draw, struct draw_vertex_shader *dvs) { -#if defined(__i386__) || defined(__386__) - x86_release_func( (struct x86_function *) &dvs->sse2_program ); -#endif - - FREE( dvs ); + dvs->delete( dvs ); } diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h new file mode 100644 index 0000000000..4ee7e705e9 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef DRAW_VS_H +#define DRAW_VS_H + +struct draw_vertex_shader; +struct draw_context; +struct pipe_shader_state; + +struct draw_vertex_shader * +draw_create_vs_exec(struct draw_context *draw, + const struct pipe_shader_state *templ); + +struct draw_vertex_shader * +draw_create_vs_sse(struct draw_context *draw, + const struct pipe_shader_state *templ); + +struct draw_vertex_shader * +draw_create_vs_llvm(struct draw_context *draw, + const struct pipe_shader_state *templ); + +#endif diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c new file mode 100644 index 0000000000..8588879400 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -0,0 +1,186 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + +#include "draw_private.h" +#include "draw_context.h" +#include "draw_vs.h" + + +static INLINE unsigned +compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<machine, + shader->state->tokens, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/ ); + + draw_update_vertex_fetch( draw ); +} + + +/** + * Transform vertices with the current vertex program/shader + * Up to four vertices can be shaded at a time. + * \param vbuffer the input vertex data + * \param elts indexes of four input vertices + * \param count number of vertices to shade [1..4] + * \param vOut array of pointers to four output vertices + */ +static void +vs_exec_run( struct draw_vertex_shader *shader, + struct draw_context *draw, + const unsigned *elts, + unsigned count, + struct vertex_header *vOut[] ) +{ + struct tgsi_exec_machine *machine = &draw->machine; + unsigned int j; + + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); + assert(draw->vertex_shader->state->output_semantic_name[0] + == TGSI_SEMANTIC_POSITION); + + machine->Consts = (float (*)[4]) draw->user.constants; + machine->Inputs = ALIGN16_ASSIGN(inputs); + machine->Outputs = ALIGN16_ASSIGN(outputs); + + draw->vertex_fetch.fetch_func( draw, machine, elts, count ); + + /* run interpreter */ + tgsi_exec_machine_run( machine ); + + + /* store machine results */ + for (j = 0; j < count; j++) { + unsigned slot; + float x, y, z, w; + + /* Handle attr[0] (position) specially: + * + * XXX: Computing the clipmask should be done in the vertex + * program as a set of DP4 instructions appended to the + * user-provided code. + */ + x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + } /* loop over vertices */ +} + + + +static void +vs_exec_delete( struct draw_vertex_shader *dvs ) +{ + FREE( dvs ); +} + + +struct draw_vertex_shader * +draw_create_vs_exec(struct draw_context *draw, + const struct pipe_shader_state *state) +{ + struct draw_vertex_shader *vs = CALLOC_STRUCT( draw_vertex_shader ); + + if (vs == NULL) + return NULL; + + vs->state = state; + vs->prepare = vs_exec_prepare; + vs->run = vs_exec_run; + vs->delete = vs_exec_delete; + + return vs; +} diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c new file mode 100644 index 0000000000..44022b6e07 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -0,0 +1,237 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin + * Keith Whitwell + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" +#include "draw_context.h" +#include "draw_vs.h" + +#ifdef MESA_LLVM + +#include "llvm/gallivm.h" + +struct draw_llvm_vertex_shader { + struct draw_vertex_shader base; + struct gallivm_prog *llvm_prog; +}; + + +static INLINE unsigned +compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<machine; + unsigned int j; + + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + + assert(count <= 4); + assert(draw->vertex_shader->state->output_semantic_name[0] + == TGSI_SEMANTIC_POSITION); + + /* Consts does not require 16 byte alignment. */ + machine->Consts = (float (*)[4]) draw->user.constants; + + machine->Inputs = ALIGN16_ASSIGN(inputs); + machine->Outputs = ALIGN16_ASSIGN(outputs); + + draw->vertex_fetch.fetch_func( draw, machine, elts, count ); + + /* run shader */ + gallivm_cpu_vs_exec(shader->llvm_prog, + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps); + + /* store machine results */ + for (j = 0; j < count; j++) { + unsigned slot; + float x, y, z, w; + + x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + } /* loop over vertices */ +} + +static void +vs_llvm_delete( struct draw_vertex_shader *base ) +{ + struct draw_llvm_vertex_shader *shader = + (struct draw_llvm_vertex_shader *)base; + + /* Do something to free compiled shader: + */ + + FREE( shader ); +} + + + + +struct draw_vertex_shader * +draw_create_vs_llvm(struct draw_context *draw, + const struct pipe_shader_state *templ) +{ + struct draw_llvm_vertex_shader *vs; + + vs = CALLOC_STRUCT( draw_llvm_vertex_shader ); + if (vs == NULL) + return NULL; + + vs->base.state = templ; + vs->base.prepare = vs_llvm_prepare; + vs->base.run = vs_llvm_run; + vs->base.delete = vs_llvm_delete; + + { + struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); + gallivm_ir_set_layout(ir, GALLIVM_SOA); + gallivm_ir_set_components(ir, 4); + gallivm_ir_fill_from_tgsi(ir, vs->base.state->tokens); + vs->llvm_prog = gallivm_ir_compile(ir); + gallivm_ir_delete(ir); + } + + draw->engine = gallivm_global_cpu_engine(); + + /* XXX: Why are there two versions of this? Shouldn't creating the + * engine be a separate operation to compiling a shader? + */ + if (!draw->engine) { + draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); + } + else { + gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); + } + + return &vs->base; +} + + + + + +#else + +struct draw_vertex_shader * +draw_create_vs_llvm(struct draw_context *draw, + const struct pipe_shader_state *shader) +{ + return NULL; +} + +#endif diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c new file mode 100644 index 0000000000..04349cb404 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -0,0 +1,251 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "draw_vs.h" + +#if defined(__i386__) || defined(__386__) + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + +#include "draw_private.h" +#include "draw_context.h" + +#include "x86/rtasm/x86sse.h" +#include "tgsi/exec/tgsi_sse2.h" + + +typedef void (XSTDCALL *codegen_function) ( + const struct tgsi_exec_vector *input, + struct tgsi_exec_vector *output, + float (*constant)[4], + struct tgsi_exec_vector *temporary ); + + +struct draw_sse_vertex_shader { + struct draw_vertex_shader base; + struct x86_function sse2_program; + codegen_function func; +}; + + +/* Should be part of the generated shader: + */ +static INLINE unsigned +compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) +{ + unsigned mask = 0; + unsigned i; + + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<machine; + unsigned int j; + + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + const float *scale = draw->viewport.scale; + const float *trans = draw->viewport.translate; + + assert(count <= 4); + assert(draw->vertex_shader->state->output_semantic_name[0] + == TGSI_SEMANTIC_POSITION); + + /* Consts does not require 16 byte alignment. */ + machine->Consts = (float (*)[4]) draw->user.constants; + machine->Inputs = ALIGN16_ASSIGN(inputs); + machine->Outputs = ALIGN16_ASSIGN(outputs); + + + /* Fetch vertices. This may at some point be integrated into the + * compiled shader -- that would require a reorganization where + * multiple versions of the compiled shader might exist, + * specialized for each fetch state. + */ + draw->vertex_fetch.fetch_func( draw, machine, elts, count ); + + + /* run compiled shader + */ + shader->func( + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps ); + + + /* XXX: Computing the clipmask and emitting results should be done + * in the vertex program as a set of instructions appended to + * the user-provided code. + */ + for (j = 0; j < count; j++) { + unsigned slot; + float x, y, z, w; + + x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + } +} + + + +static void +vs_sse_delete( struct draw_vertex_shader *base ) +{ + struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; + + x86_release_func( &shader->sse2_program ); + + FREE( shader ); +} + + +struct draw_vertex_shader * +draw_create_vs_sse(struct draw_context *draw, + const struct pipe_shader_state *templ) +{ + struct draw_sse_vertex_shader *vs; + + if (!draw->use_sse) + return NULL; + + vs = CALLOC_STRUCT( draw_sse_vertex_shader ); + if (vs == NULL) + return NULL; + + vs->base.state = templ; + vs->base.prepare = vs_sse_prepare; + vs->base.run = vs_sse_run; + vs->base.delete = vs_sse_delete; + + x86_init_func( &vs->sse2_program ); + + if (!tgsi_emit_sse2( (struct tgsi_token *) vs->base.state->tokens, + &vs->sse2_program )) + goto fail; + + vs->func = (codegen_function) x86_get_func( &vs->sse2_program ); + + return &vs->base; + +fail: + fprintf(stderr, "tgsi_emit_sse2() failed, falling back to interpreter\n"); + + x86_release_func( &vs->sse2_program ); + + FREE(vs); + return NULL; +} + + + +#else + +struct draw_vertex_shader * +draw_create_vs_sse( struct draw_context *draw, + const struct pipe_shader_state *templ ) +{ + return NULL; +} + + +#endif + diff --git a/src/mesa/sources b/src/mesa/sources index cecd8a830f..f83d247a1e 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -159,6 +159,9 @@ VF_SOURCES = \ DRAW_SOURCES = \ $(TOP)/src/gallium/auxiliary/draw/draw_clip.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vs_exec.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vs_sse.c \ + $(TOP)/src/gallium/auxiliary/draw/draw_vs_llvm.c \ $(TOP)/src/gallium/auxiliary/draw/draw_context.c\ $(TOP)/src/gallium/auxiliary/draw/draw_cull.c \ $(TOP)/src/gallium/auxiliary/draw/draw_debug.c \ -- cgit v1.2.3 From c179bc990108a8ea691ceab03fd68a12396ab538 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 15 Feb 2008 13:39:24 +0000 Subject: tgsi: pass through failure to sse-codegenerate for fragment programs too. In particular, will fallback to interpreted execution for shaders with TEX instructions. --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 79209575bc..62c6a69c63 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -2308,6 +2308,7 @@ tgsi_emit_sse2_fs( { struct tgsi_parse_context parse; boolean instruction_phase = FALSE; + unsigned ok = 1; DUMP_START(); @@ -2333,7 +2334,7 @@ tgsi_emit_sse2_fs( tgsi_parse_init( &parse, tokens ); - while( !tgsi_parse_end_of_tokens( &parse ) ) { + while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { @@ -2352,17 +2353,18 @@ tgsi_emit_sse2_fs( get_output_base(), get_argument( 1 ) ); } - emit_instruction( + ok = emit_instruction( func, &parse.FullToken.FullInstruction ); break; case TGSI_TOKEN_TYPE_IMMEDIATE: /* XXX implement this */ - assert(0); + ok = 0; break; default: + ok = 0; assert( 0 ); } } @@ -2371,7 +2373,7 @@ tgsi_emit_sse2_fs( DUMP_END(); - return 1; + return ok; } #endif /* i386 */ -- cgit v1.2.3 From 2cc0c3b99703bb10e0b320001183acb728e5488e Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 14 Feb 2008 18:29:51 -0800 Subject: Correct the convert to and from float instructions --- src/mesa/ppc/rtasm/spe_asm.c | 34 ++++++++++++++++++++++++++++++++++ src/mesa/ppc/rtasm/spe_asm.h | 12 ++++++++---- 2 files changed, 42 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/ppc/rtasm/spe_asm.c b/src/mesa/ppc/rtasm/spe_asm.c index f8aff9050b..1037637250 100644 --- a/src/mesa/ppc/rtasm/spe_asm.c +++ b/src/mesa/ppc/rtasm/spe_asm.c @@ -87,6 +87,20 @@ union spe_inst_RI7 { }; +/** + * Encode one output register with one input reg. and an 8-bit signed immed + */ +union spe_inst_RI8 { + uint32_t bits; + struct { + unsigned op:10; + unsigned i8:8; + unsigned rA:7; + unsigned rT:7; + } inst; +}; + + /** * Encode one output register with one input reg. and a 10-bit signed immed */ @@ -169,6 +183,20 @@ static void emit_RI7(struct spe_function *p, unsigned op, unsigned rT, +static void emit_RI8(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, int imm) +{ + union spe_inst_RI8 inst; + inst.inst.op = op; + inst.inst.i8 = imm; + inst.inst.rA = rA; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + + static void emit_RI10(struct spe_function *p, unsigned op, unsigned rT, unsigned rA, int imm) { @@ -238,6 +266,12 @@ void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ emit_RI7(p, _op, rT, rA, imm); \ } +#define EMIT_RI8(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ +{ \ + emit_RI8(p, _op, rT, rA, 155 - imm); \ +} + #define EMIT_RI10(_name, _op) \ void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ { \ diff --git a/src/mesa/ppc/rtasm/spe_asm.h b/src/mesa/ppc/rtasm/spe_asm.h index 9532669b8c..6d69ae655d 100644 --- a/src/mesa/ppc/rtasm/spe_asm.h +++ b/src/mesa/ppc/rtasm/spe_asm.h @@ -60,6 +60,9 @@ extern void spe_release_func(struct spe_function *p); #define EMIT_RI7(_name, _op) \ extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ int imm) +#define EMIT_RI8(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + int imm) #define EMIT_RI10(_name, _op) \ extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ int imm) @@ -270,10 +273,10 @@ EMIT_RR (spe_dfnma, 0x35f); EMIT_R (spe_frest, 0x1b8); EMIT_R (spe_frsqest, 0x1b9); EMIT_RR (spe_fi, 0x3d4); -EMIT_RI7 (spe_csflt, 0x3da); -EMIT_RI7 (spe_cflts, 0x3d8); -EMIT_RI7 (spe_cuflt, 0x3db); -EMIT_RI7 (spe_cfltu, 0x3d9); +EMIT_RI8 (spe_csflt, 0x1da); +EMIT_RI8 (spe_cflts, 0x1d8); +EMIT_RI8 (spe_cuflt, 0x1db); +EMIT_RI8 (spe_cfltu, 0x1d9); EMIT_R (spe_frds, 0x3b9); EMIT_R (spe_fesd, 0x3b8); EMIT_RR (spe_dfceq, 0x3c3); @@ -302,6 +305,7 @@ EMIT_R (spe_wrch, 0x10d); #undef EMIT_RR #undef EMIT_RRR #undef EMIT_RI7 +#undef EMIT_RI8 #undef EMIT_RI10 #undef EMIT_RI16 #undef EMIT_RI18 -- cgit v1.2.3 From 55e64b63adfca3cba5847601b1a68e885da725a4 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 14 Feb 2008 18:30:48 -0800 Subject: Initial version of code gen for attribute fetch --- src/mesa/pipe/cell/ppu/cell_vertex_fetch.c | 392 +++++++++++++++++++++++++++++ 1 file changed, 392 insertions(+) create mode 100644 src/mesa/pipe/cell/ppu/cell_vertex_fetch.c (limited to 'src') diff --git a/src/mesa/pipe/cell/ppu/cell_vertex_fetch.c b/src/mesa/pipe/cell/ppu/cell_vertex_fetch.c new file mode 100644 index 0000000000..f2432f4ff5 --- /dev/null +++ b/src/mesa/pipe/cell/ppu/cell_vertex_fetch.c @@ -0,0 +1,392 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_format.h" + +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_private.h" + +#include "pipe/cell/ppu/cell_context.h" +#include "ppc/rtasm/spe_asm.h" + +typedef uint64_t register_mask; + +int allocate_available_register(register_mask *m) +{ + unsigned i; + for (i = 0; i < 64; i++) { + const uint64_t mask = (1ULL << i); + + if ((m[0] & mask) != 0) { + m[0] &= ~mask; + return i; + } + } + + return -1; +} + + +int allocate_register(register_mask *m, unsigned reg) +{ + assert((m[0] & (1ULL << reg)) != 0); + + m[0] &= ~(1ULL << reg); + return reg; +} + + +void release_register(register_mask *m, unsigned reg) +{ + assert((m[0] & (1ULL << reg)) == 0); + + m[0] |= (1ULL << reg); +} + + +/** + * Emit a 4x4 matrix transpose operation + * + * \param p Function that the transpose operation is to be appended to + * \param m Live register mask + * \param row0 Register containing row 0 of the source matrix + * \param row1 Register containing row 1 of the source matrix + * \param row2 Register containing row 2 of the source matrix + * \param row3 Register containing row 3 of the source matrix + * \param dest_ptr Register containing the address of the destination matrix + * \param shuf_ptr Register containing the address of the shuffled data + * \param count Number of colums to actually be written to the destination + * + * \note + * This function assumes that the registers named by \c row0, \c row1, + * \c row2, and \c row3 are scratch and can be modified by the generated code. + * Furthermore, these registers will be released, via calls to + * \c release_register, by this function. + * + * \note + * This function requires that four temporary are available on entry. + */ +static void +emit_matrix_transpose(struct spe_function *p, register_mask *m, + unsigned row0, unsigned row1, unsigned row2, + unsigned row3, unsigned dest_ptr, + unsigned shuf_ptr, unsigned count) +{ + int shuf_hi = allocate_available_register(m); + int shuf_lo = allocate_available_register(m); + int t1 = allocate_available_register(m); + int t2 = allocate_available_register(m); + int t3; + int t4; + int col0; + int col1; + int col2; + int col3; + + + spe_lqd(p, shuf_hi, shuf_ptr, 3); + spe_lqd(p, shuf_lo, shuf_ptr, 4); + spe_shufb(p, t1, row0, row2, shuf_hi); + spe_shufb(p, t2, row0, row2, shuf_lo); + + + /* row0 and row2 are now no longer needed. Re-use those registers as + * temporaries. + */ + t3 = row0; + t4 = row2; + + spe_shufb(p, t3, row1, row3, shuf_hi); + spe_shufb(p, t4, row1, row3, shuf_lo); + + + /* row1 and row3 are now no longer needed. Re-use those registers as + * temporaries. + */ + col0 = row1; + col1 = row3; + + spe_shufb(p, col0, t1, t3, shuf_hi); + if (count > 1) { + spe_shufb(p, col1, t1, t3, shuf_lo); + } + + /* t1 and t3 are now no longer needed. Re-use those registers as + * temporaries. + */ + col2 = t1; + col3 = t3; + + if (count > 2) { + spe_shufb(p, col2, t2, t4, shuf_hi); + } + + if (count > 3) { + spe_shufb(p, col3, t2, t4, shuf_lo); + } + + + /* Store the results. Remember that the stqd instruction is encoded using + * the qword offset (stand-alone assemblers to the byte-offset to + * qword-offset conversion for you), so the byte-offset needs be divided by + * 16. + */ + switch (count) { + case 4: + spe_stqd(p, col3, dest_ptr, 3); + case 3: + spe_stqd(p, col2, dest_ptr, 2); + case 2: + spe_stqd(p, col1, dest_ptr, 1); + case 1: + spe_stqd(p, col0, dest_ptr, 0); + } + + + /* Release all of the temporary registers used. + */ + release_register(m, col0); + release_register(m, col1); + release_register(m, col2); + release_register(m, col3); + release_register(m, shuf_hi); + release_register(m, shuf_lo); + release_register(m, t2); + release_register(m, t4); +} + + +static void +emit_fetch(struct spe_function *p, register_mask *m, + unsigned in_ptr, unsigned *offset, + unsigned out_ptr, unsigned shuf_ptr, + enum pipe_format format) +{ + const unsigned count = (pf_size_x(format) != 0) + (pf_size_y(format) != 0) + + (pf_size_z(format) != 0) + (pf_size_w(format) != 0); + const unsigned type = pf_type(format); + const unsigned bytes = pf_size_x(format); + + int v0 = allocate_available_register(m); + int v1 = allocate_available_register(m); + int v2 = allocate_available_register(m); + int v3 = allocate_available_register(m); + int tmp = allocate_available_register(m); + int float_zero = -1; + int float_one = -1; + float scale_signed = 0.0; + float scale_unsigned = 0.0; + + spe_lqd(p, v0, in_ptr, 0 + offset[0]); + spe_lqd(p, v1, in_ptr, 1 + offset[0]); + spe_lqd(p, v2, in_ptr, 2 + offset[0]); + spe_lqd(p, v3, in_ptr, 3 + offset[0]); + offset[0] += 4; + + switch (bytes) { + case 1: + scale_signed = 1.0f / 127.0f; + scale_unsigned = 1.0f / 255.0f; + spe_lqd(p, tmp, shuf_ptr, 1); + spe_shufb(p, v0, v0, v0, tmp); + spe_shufb(p, v1, v1, v1, tmp); + spe_shufb(p, v2, v2, v2, tmp); + spe_shufb(p, v3, v3, v3, tmp); + break; + case 2: + scale_signed = 1.0f / 32767.0f; + scale_unsigned = 1.0f / 65535.0f; + spe_lqd(p, tmp, shuf_ptr, 2); + spe_shufb(p, v0, v0, v0, tmp); + spe_shufb(p, v1, v1, v1, tmp); + spe_shufb(p, v2, v2, v2, tmp); + spe_shufb(p, v3, v3, v3, tmp); + break; + case 4: + scale_signed = 1.0f / 2147483647.0f; + scale_unsigned = 1.0f / 4294967295.0f; + break; + default: + assert(0); + break; + } + + switch (type) { + case PIPE_FORMAT_TYPE_FLOAT: + break; + case PIPE_FORMAT_TYPE_UNORM: + spe_ilhu(p, tmp, ((unsigned) scale_unsigned) >> 16); + spe_iohl(p, tmp, ((unsigned) scale_unsigned) & 0x0ffff); + spe_cuflt(p, v0, v0, 0); + spe_fm(p, v0, v0, tmp); + break; + case PIPE_FORMAT_TYPE_SNORM: + spe_ilhu(p, tmp, ((unsigned) scale_signed) >> 16); + spe_iohl(p, tmp, ((unsigned) scale_signed) & 0x0ffff); + spe_csflt(p, v0, v0, 0); + spe_fm(p, v0, v0, tmp); + break; + case PIPE_FORMAT_TYPE_USCALED: + spe_cuflt(p, v0, v0, 0); + break; + case PIPE_FORMAT_TYPE_SSCALED: + spe_csflt(p, v0, v0, 0); + break; + } + + + if (count < 4) { + float_one = allocate_available_register(m); + spe_il(p, float_one, 1); + spe_cuflt(p, float_one, float_one, 0); + + if (count < 3) { + float_zero = allocate_available_register(m); + spe_il(p, float_zero, 0); + } + } + + release_register(m, tmp); + + emit_matrix_transpose(p, m, v0, v1, v2, v3, out_ptr, shuf_ptr, count); + + switch (count) { + case 1: + spe_stqd(p, float_zero, out_ptr, 1); + case 2: + spe_stqd(p, float_zero, out_ptr, 2); + case 3: + spe_stqd(p, float_one, out_ptr, 3); + } + + if (float_zero != -1) { + release_register(m, float_zero); + } + + if (float_one != -1) { + release_register(m, float_one); + } +} + + +void cell_update_vertex_fetch(struct draw_context *draw) +{ + struct cell_context *const cell = + (struct cell_context *) draw->driver_private; + register_mask m = ~0; + struct spe_function *p = &cell->attrib_fetch; + unsigned function_index[PIPE_ATTRIB_MAX]; + unsigned unique_attr_formats; + int out_ptr; + int in_ptr; + int shuf_ptr; + unsigned i; + unsigned j; + + + /* Determine how many unique input attribute formats there are. At the + * same time, store the index of the lowest numbered attribute that has + * the same format as any non-unique format. + */ + unique_attr_formats = 1; + function_index[0] = 0; + for (i = 1; i < draw->vertex_fetch.nr_attrs; i++) { + const enum pipe_format curr_fmt = draw->vertex_element[i].src_format; + + for (j = 0; j < i; j++) { + if (curr_fmt == draw->vertex_element[j].src_format) { + break; + } + } + + if (j == i) { + unique_attr_formats++; + } + + function_index[i] = j; + } + + + /* Each fetch function can be a maximum of 34 instructions (note: this is + * actually a slight over-estimate). That means (34 * 4) = 136 bytes + * each maximum. + */ + spe_init_func(p, 136 * unique_attr_formats); + + + /* Registers 0, 1, and 2 are reserved by the ABI. + */ + allocate_register(&m, 0); + allocate_register(&m, 1); + allocate_register(&m, 2); + + + /* Allocate registers for the function's input parameters. + */ + out_ptr = allocate_register(&m, 3); + in_ptr = allocate_register(&m, 4); + shuf_ptr = allocate_register(&m, 5); + + + /* Generate code for the individual attribute fetch functions. + */ + for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { + unsigned offset; + + if (function_index[i] == i) { + cell->attrib_fetch_offsets[i] = (unsigned) ((void *) p->csr + - (void *) p->store); + + offset = 0; + emit_fetch(p, & m, in_ptr, &offset, out_ptr, shuf_ptr, + draw->vertex_element[i].src_format); + spe_bi(p, 0, 0, 0); + + /* Round up to the next 16-byte boundary. + */ + if ((((unsigned) p->store) & 0x0f) != 0) { + const unsigned align = ((unsigned) p->store) & 0x0f; + p->store = (uint32_t *) (((void *) p->store) + align); + } + } else { + /* Use the same function entry-point as a previously seen attribute + * with the same format. + */ + cell->attrib_fetch_offsets[i] = + cell->attrib_fetch_offsets[function_index[i]]; + } + } + + static first_time = 1; + if (first_time) { + first_time = 0; + const unsigned instructions = p->csr - p->store; + for (i = 0; i < instructions; i++) { + printf("\t.long\t0x%08x\n", p->store[i]); + } + } +} -- cgit v1.2.3 From 397b81bd1c7984b1667af7ef954e053263a7a661 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 15 Feb 2008 09:43:13 -0800 Subject: Move cell_vertex_fetch.c for recent code reorg. --- src/gallium/drivers/cell/ppu/cell_vertex_fetch.c | 392 +++++++++++++++++++++++ src/mesa/pipe/cell/ppu/cell_vertex_fetch.c | 392 ----------------------- 2 files changed, 392 insertions(+), 392 deletions(-) create mode 100644 src/gallium/drivers/cell/ppu/cell_vertex_fetch.c delete mode 100644 src/mesa/pipe/cell/ppu/cell_vertex_fetch.c (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c new file mode 100644 index 0000000000..f2432f4ff5 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c @@ -0,0 +1,392 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "pipe/p_format.h" + +#include "pipe/draw/draw_context.h" +#include "pipe/draw/draw_private.h" + +#include "pipe/cell/ppu/cell_context.h" +#include "ppc/rtasm/spe_asm.h" + +typedef uint64_t register_mask; + +int allocate_available_register(register_mask *m) +{ + unsigned i; + for (i = 0; i < 64; i++) { + const uint64_t mask = (1ULL << i); + + if ((m[0] & mask) != 0) { + m[0] &= ~mask; + return i; + } + } + + return -1; +} + + +int allocate_register(register_mask *m, unsigned reg) +{ + assert((m[0] & (1ULL << reg)) != 0); + + m[0] &= ~(1ULL << reg); + return reg; +} + + +void release_register(register_mask *m, unsigned reg) +{ + assert((m[0] & (1ULL << reg)) == 0); + + m[0] |= (1ULL << reg); +} + + +/** + * Emit a 4x4 matrix transpose operation + * + * \param p Function that the transpose operation is to be appended to + * \param m Live register mask + * \param row0 Register containing row 0 of the source matrix + * \param row1 Register containing row 1 of the source matrix + * \param row2 Register containing row 2 of the source matrix + * \param row3 Register containing row 3 of the source matrix + * \param dest_ptr Register containing the address of the destination matrix + * \param shuf_ptr Register containing the address of the shuffled data + * \param count Number of colums to actually be written to the destination + * + * \note + * This function assumes that the registers named by \c row0, \c row1, + * \c row2, and \c row3 are scratch and can be modified by the generated code. + * Furthermore, these registers will be released, via calls to + * \c release_register, by this function. + * + * \note + * This function requires that four temporary are available on entry. + */ +static void +emit_matrix_transpose(struct spe_function *p, register_mask *m, + unsigned row0, unsigned row1, unsigned row2, + unsigned row3, unsigned dest_ptr, + unsigned shuf_ptr, unsigned count) +{ + int shuf_hi = allocate_available_register(m); + int shuf_lo = allocate_available_register(m); + int t1 = allocate_available_register(m); + int t2 = allocate_available_register(m); + int t3; + int t4; + int col0; + int col1; + int col2; + int col3; + + + spe_lqd(p, shuf_hi, shuf_ptr, 3); + spe_lqd(p, shuf_lo, shuf_ptr, 4); + spe_shufb(p, t1, row0, row2, shuf_hi); + spe_shufb(p, t2, row0, row2, shuf_lo); + + + /* row0 and row2 are now no longer needed. Re-use those registers as + * temporaries. + */ + t3 = row0; + t4 = row2; + + spe_shufb(p, t3, row1, row3, shuf_hi); + spe_shufb(p, t4, row1, row3, shuf_lo); + + + /* row1 and row3 are now no longer needed. Re-use those registers as + * temporaries. + */ + col0 = row1; + col1 = row3; + + spe_shufb(p, col0, t1, t3, shuf_hi); + if (count > 1) { + spe_shufb(p, col1, t1, t3, shuf_lo); + } + + /* t1 and t3 are now no longer needed. Re-use those registers as + * temporaries. + */ + col2 = t1; + col3 = t3; + + if (count > 2) { + spe_shufb(p, col2, t2, t4, shuf_hi); + } + + if (count > 3) { + spe_shufb(p, col3, t2, t4, shuf_lo); + } + + + /* Store the results. Remember that the stqd instruction is encoded using + * the qword offset (stand-alone assemblers to the byte-offset to + * qword-offset conversion for you), so the byte-offset needs be divided by + * 16. + */ + switch (count) { + case 4: + spe_stqd(p, col3, dest_ptr, 3); + case 3: + spe_stqd(p, col2, dest_ptr, 2); + case 2: + spe_stqd(p, col1, dest_ptr, 1); + case 1: + spe_stqd(p, col0, dest_ptr, 0); + } + + + /* Release all of the temporary registers used. + */ + release_register(m, col0); + release_register(m, col1); + release_register(m, col2); + release_register(m, col3); + release_register(m, shuf_hi); + release_register(m, shuf_lo); + release_register(m, t2); + release_register(m, t4); +} + + +static void +emit_fetch(struct spe_function *p, register_mask *m, + unsigned in_ptr, unsigned *offset, + unsigned out_ptr, unsigned shuf_ptr, + enum pipe_format format) +{ + const unsigned count = (pf_size_x(format) != 0) + (pf_size_y(format) != 0) + + (pf_size_z(format) != 0) + (pf_size_w(format) != 0); + const unsigned type = pf_type(format); + const unsigned bytes = pf_size_x(format); + + int v0 = allocate_available_register(m); + int v1 = allocate_available_register(m); + int v2 = allocate_available_register(m); + int v3 = allocate_available_register(m); + int tmp = allocate_available_register(m); + int float_zero = -1; + int float_one = -1; + float scale_signed = 0.0; + float scale_unsigned = 0.0; + + spe_lqd(p, v0, in_ptr, 0 + offset[0]); + spe_lqd(p, v1, in_ptr, 1 + offset[0]); + spe_lqd(p, v2, in_ptr, 2 + offset[0]); + spe_lqd(p, v3, in_ptr, 3 + offset[0]); + offset[0] += 4; + + switch (bytes) { + case 1: + scale_signed = 1.0f / 127.0f; + scale_unsigned = 1.0f / 255.0f; + spe_lqd(p, tmp, shuf_ptr, 1); + spe_shufb(p, v0, v0, v0, tmp); + spe_shufb(p, v1, v1, v1, tmp); + spe_shufb(p, v2, v2, v2, tmp); + spe_shufb(p, v3, v3, v3, tmp); + break; + case 2: + scale_signed = 1.0f / 32767.0f; + scale_unsigned = 1.0f / 65535.0f; + spe_lqd(p, tmp, shuf_ptr, 2); + spe_shufb(p, v0, v0, v0, tmp); + spe_shufb(p, v1, v1, v1, tmp); + spe_shufb(p, v2, v2, v2, tmp); + spe_shufb(p, v3, v3, v3, tmp); + break; + case 4: + scale_signed = 1.0f / 2147483647.0f; + scale_unsigned = 1.0f / 4294967295.0f; + break; + default: + assert(0); + break; + } + + switch (type) { + case PIPE_FORMAT_TYPE_FLOAT: + break; + case PIPE_FORMAT_TYPE_UNORM: + spe_ilhu(p, tmp, ((unsigned) scale_unsigned) >> 16); + spe_iohl(p, tmp, ((unsigned) scale_unsigned) & 0x0ffff); + spe_cuflt(p, v0, v0, 0); + spe_fm(p, v0, v0, tmp); + break; + case PIPE_FORMAT_TYPE_SNORM: + spe_ilhu(p, tmp, ((unsigned) scale_signed) >> 16); + spe_iohl(p, tmp, ((unsigned) scale_signed) & 0x0ffff); + spe_csflt(p, v0, v0, 0); + spe_fm(p, v0, v0, tmp); + break; + case PIPE_FORMAT_TYPE_USCALED: + spe_cuflt(p, v0, v0, 0); + break; + case PIPE_FORMAT_TYPE_SSCALED: + spe_csflt(p, v0, v0, 0); + break; + } + + + if (count < 4) { + float_one = allocate_available_register(m); + spe_il(p, float_one, 1); + spe_cuflt(p, float_one, float_one, 0); + + if (count < 3) { + float_zero = allocate_available_register(m); + spe_il(p, float_zero, 0); + } + } + + release_register(m, tmp); + + emit_matrix_transpose(p, m, v0, v1, v2, v3, out_ptr, shuf_ptr, count); + + switch (count) { + case 1: + spe_stqd(p, float_zero, out_ptr, 1); + case 2: + spe_stqd(p, float_zero, out_ptr, 2); + case 3: + spe_stqd(p, float_one, out_ptr, 3); + } + + if (float_zero != -1) { + release_register(m, float_zero); + } + + if (float_one != -1) { + release_register(m, float_one); + } +} + + +void cell_update_vertex_fetch(struct draw_context *draw) +{ + struct cell_context *const cell = + (struct cell_context *) draw->driver_private; + register_mask m = ~0; + struct spe_function *p = &cell->attrib_fetch; + unsigned function_index[PIPE_ATTRIB_MAX]; + unsigned unique_attr_formats; + int out_ptr; + int in_ptr; + int shuf_ptr; + unsigned i; + unsigned j; + + + /* Determine how many unique input attribute formats there are. At the + * same time, store the index of the lowest numbered attribute that has + * the same format as any non-unique format. + */ + unique_attr_formats = 1; + function_index[0] = 0; + for (i = 1; i < draw->vertex_fetch.nr_attrs; i++) { + const enum pipe_format curr_fmt = draw->vertex_element[i].src_format; + + for (j = 0; j < i; j++) { + if (curr_fmt == draw->vertex_element[j].src_format) { + break; + } + } + + if (j == i) { + unique_attr_formats++; + } + + function_index[i] = j; + } + + + /* Each fetch function can be a maximum of 34 instructions (note: this is + * actually a slight over-estimate). That means (34 * 4) = 136 bytes + * each maximum. + */ + spe_init_func(p, 136 * unique_attr_formats); + + + /* Registers 0, 1, and 2 are reserved by the ABI. + */ + allocate_register(&m, 0); + allocate_register(&m, 1); + allocate_register(&m, 2); + + + /* Allocate registers for the function's input parameters. + */ + out_ptr = allocate_register(&m, 3); + in_ptr = allocate_register(&m, 4); + shuf_ptr = allocate_register(&m, 5); + + + /* Generate code for the individual attribute fetch functions. + */ + for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { + unsigned offset; + + if (function_index[i] == i) { + cell->attrib_fetch_offsets[i] = (unsigned) ((void *) p->csr + - (void *) p->store); + + offset = 0; + emit_fetch(p, & m, in_ptr, &offset, out_ptr, shuf_ptr, + draw->vertex_element[i].src_format); + spe_bi(p, 0, 0, 0); + + /* Round up to the next 16-byte boundary. + */ + if ((((unsigned) p->store) & 0x0f) != 0) { + const unsigned align = ((unsigned) p->store) & 0x0f; + p->store = (uint32_t *) (((void *) p->store) + align); + } + } else { + /* Use the same function entry-point as a previously seen attribute + * with the same format. + */ + cell->attrib_fetch_offsets[i] = + cell->attrib_fetch_offsets[function_index[i]]; + } + } + + static first_time = 1; + if (first_time) { + first_time = 0; + const unsigned instructions = p->csr - p->store; + for (i = 0; i < instructions; i++) { + printf("\t.long\t0x%08x\n", p->store[i]); + } + } +} diff --git a/src/mesa/pipe/cell/ppu/cell_vertex_fetch.c b/src/mesa/pipe/cell/ppu/cell_vertex_fetch.c deleted file mode 100644 index f2432f4ff5..0000000000 --- a/src/mesa/pipe/cell/ppu/cell_vertex_fetch.c +++ /dev/null @@ -1,392 +0,0 @@ -/* - * (C) Copyright IBM Corporation 2008 - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -#include -#include "pipe/p_defines.h" -#include "pipe/p_context.h" -#include "pipe/p_format.h" - -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" - -#include "pipe/cell/ppu/cell_context.h" -#include "ppc/rtasm/spe_asm.h" - -typedef uint64_t register_mask; - -int allocate_available_register(register_mask *m) -{ - unsigned i; - for (i = 0; i < 64; i++) { - const uint64_t mask = (1ULL << i); - - if ((m[0] & mask) != 0) { - m[0] &= ~mask; - return i; - } - } - - return -1; -} - - -int allocate_register(register_mask *m, unsigned reg) -{ - assert((m[0] & (1ULL << reg)) != 0); - - m[0] &= ~(1ULL << reg); - return reg; -} - - -void release_register(register_mask *m, unsigned reg) -{ - assert((m[0] & (1ULL << reg)) == 0); - - m[0] |= (1ULL << reg); -} - - -/** - * Emit a 4x4 matrix transpose operation - * - * \param p Function that the transpose operation is to be appended to - * \param m Live register mask - * \param row0 Register containing row 0 of the source matrix - * \param row1 Register containing row 1 of the source matrix - * \param row2 Register containing row 2 of the source matrix - * \param row3 Register containing row 3 of the source matrix - * \param dest_ptr Register containing the address of the destination matrix - * \param shuf_ptr Register containing the address of the shuffled data - * \param count Number of colums to actually be written to the destination - * - * \note - * This function assumes that the registers named by \c row0, \c row1, - * \c row2, and \c row3 are scratch and can be modified by the generated code. - * Furthermore, these registers will be released, via calls to - * \c release_register, by this function. - * - * \note - * This function requires that four temporary are available on entry. - */ -static void -emit_matrix_transpose(struct spe_function *p, register_mask *m, - unsigned row0, unsigned row1, unsigned row2, - unsigned row3, unsigned dest_ptr, - unsigned shuf_ptr, unsigned count) -{ - int shuf_hi = allocate_available_register(m); - int shuf_lo = allocate_available_register(m); - int t1 = allocate_available_register(m); - int t2 = allocate_available_register(m); - int t3; - int t4; - int col0; - int col1; - int col2; - int col3; - - - spe_lqd(p, shuf_hi, shuf_ptr, 3); - spe_lqd(p, shuf_lo, shuf_ptr, 4); - spe_shufb(p, t1, row0, row2, shuf_hi); - spe_shufb(p, t2, row0, row2, shuf_lo); - - - /* row0 and row2 are now no longer needed. Re-use those registers as - * temporaries. - */ - t3 = row0; - t4 = row2; - - spe_shufb(p, t3, row1, row3, shuf_hi); - spe_shufb(p, t4, row1, row3, shuf_lo); - - - /* row1 and row3 are now no longer needed. Re-use those registers as - * temporaries. - */ - col0 = row1; - col1 = row3; - - spe_shufb(p, col0, t1, t3, shuf_hi); - if (count > 1) { - spe_shufb(p, col1, t1, t3, shuf_lo); - } - - /* t1 and t3 are now no longer needed. Re-use those registers as - * temporaries. - */ - col2 = t1; - col3 = t3; - - if (count > 2) { - spe_shufb(p, col2, t2, t4, shuf_hi); - } - - if (count > 3) { - spe_shufb(p, col3, t2, t4, shuf_lo); - } - - - /* Store the results. Remember that the stqd instruction is encoded using - * the qword offset (stand-alone assemblers to the byte-offset to - * qword-offset conversion for you), so the byte-offset needs be divided by - * 16. - */ - switch (count) { - case 4: - spe_stqd(p, col3, dest_ptr, 3); - case 3: - spe_stqd(p, col2, dest_ptr, 2); - case 2: - spe_stqd(p, col1, dest_ptr, 1); - case 1: - spe_stqd(p, col0, dest_ptr, 0); - } - - - /* Release all of the temporary registers used. - */ - release_register(m, col0); - release_register(m, col1); - release_register(m, col2); - release_register(m, col3); - release_register(m, shuf_hi); - release_register(m, shuf_lo); - release_register(m, t2); - release_register(m, t4); -} - - -static void -emit_fetch(struct spe_function *p, register_mask *m, - unsigned in_ptr, unsigned *offset, - unsigned out_ptr, unsigned shuf_ptr, - enum pipe_format format) -{ - const unsigned count = (pf_size_x(format) != 0) + (pf_size_y(format) != 0) - + (pf_size_z(format) != 0) + (pf_size_w(format) != 0); - const unsigned type = pf_type(format); - const unsigned bytes = pf_size_x(format); - - int v0 = allocate_available_register(m); - int v1 = allocate_available_register(m); - int v2 = allocate_available_register(m); - int v3 = allocate_available_register(m); - int tmp = allocate_available_register(m); - int float_zero = -1; - int float_one = -1; - float scale_signed = 0.0; - float scale_unsigned = 0.0; - - spe_lqd(p, v0, in_ptr, 0 + offset[0]); - spe_lqd(p, v1, in_ptr, 1 + offset[0]); - spe_lqd(p, v2, in_ptr, 2 + offset[0]); - spe_lqd(p, v3, in_ptr, 3 + offset[0]); - offset[0] += 4; - - switch (bytes) { - case 1: - scale_signed = 1.0f / 127.0f; - scale_unsigned = 1.0f / 255.0f; - spe_lqd(p, tmp, shuf_ptr, 1); - spe_shufb(p, v0, v0, v0, tmp); - spe_shufb(p, v1, v1, v1, tmp); - spe_shufb(p, v2, v2, v2, tmp); - spe_shufb(p, v3, v3, v3, tmp); - break; - case 2: - scale_signed = 1.0f / 32767.0f; - scale_unsigned = 1.0f / 65535.0f; - spe_lqd(p, tmp, shuf_ptr, 2); - spe_shufb(p, v0, v0, v0, tmp); - spe_shufb(p, v1, v1, v1, tmp); - spe_shufb(p, v2, v2, v2, tmp); - spe_shufb(p, v3, v3, v3, tmp); - break; - case 4: - scale_signed = 1.0f / 2147483647.0f; - scale_unsigned = 1.0f / 4294967295.0f; - break; - default: - assert(0); - break; - } - - switch (type) { - case PIPE_FORMAT_TYPE_FLOAT: - break; - case PIPE_FORMAT_TYPE_UNORM: - spe_ilhu(p, tmp, ((unsigned) scale_unsigned) >> 16); - spe_iohl(p, tmp, ((unsigned) scale_unsigned) & 0x0ffff); - spe_cuflt(p, v0, v0, 0); - spe_fm(p, v0, v0, tmp); - break; - case PIPE_FORMAT_TYPE_SNORM: - spe_ilhu(p, tmp, ((unsigned) scale_signed) >> 16); - spe_iohl(p, tmp, ((unsigned) scale_signed) & 0x0ffff); - spe_csflt(p, v0, v0, 0); - spe_fm(p, v0, v0, tmp); - break; - case PIPE_FORMAT_TYPE_USCALED: - spe_cuflt(p, v0, v0, 0); - break; - case PIPE_FORMAT_TYPE_SSCALED: - spe_csflt(p, v0, v0, 0); - break; - } - - - if (count < 4) { - float_one = allocate_available_register(m); - spe_il(p, float_one, 1); - spe_cuflt(p, float_one, float_one, 0); - - if (count < 3) { - float_zero = allocate_available_register(m); - spe_il(p, float_zero, 0); - } - } - - release_register(m, tmp); - - emit_matrix_transpose(p, m, v0, v1, v2, v3, out_ptr, shuf_ptr, count); - - switch (count) { - case 1: - spe_stqd(p, float_zero, out_ptr, 1); - case 2: - spe_stqd(p, float_zero, out_ptr, 2); - case 3: - spe_stqd(p, float_one, out_ptr, 3); - } - - if (float_zero != -1) { - release_register(m, float_zero); - } - - if (float_one != -1) { - release_register(m, float_one); - } -} - - -void cell_update_vertex_fetch(struct draw_context *draw) -{ - struct cell_context *const cell = - (struct cell_context *) draw->driver_private; - register_mask m = ~0; - struct spe_function *p = &cell->attrib_fetch; - unsigned function_index[PIPE_ATTRIB_MAX]; - unsigned unique_attr_formats; - int out_ptr; - int in_ptr; - int shuf_ptr; - unsigned i; - unsigned j; - - - /* Determine how many unique input attribute formats there are. At the - * same time, store the index of the lowest numbered attribute that has - * the same format as any non-unique format. - */ - unique_attr_formats = 1; - function_index[0] = 0; - for (i = 1; i < draw->vertex_fetch.nr_attrs; i++) { - const enum pipe_format curr_fmt = draw->vertex_element[i].src_format; - - for (j = 0; j < i; j++) { - if (curr_fmt == draw->vertex_element[j].src_format) { - break; - } - } - - if (j == i) { - unique_attr_formats++; - } - - function_index[i] = j; - } - - - /* Each fetch function can be a maximum of 34 instructions (note: this is - * actually a slight over-estimate). That means (34 * 4) = 136 bytes - * each maximum. - */ - spe_init_func(p, 136 * unique_attr_formats); - - - /* Registers 0, 1, and 2 are reserved by the ABI. - */ - allocate_register(&m, 0); - allocate_register(&m, 1); - allocate_register(&m, 2); - - - /* Allocate registers for the function's input parameters. - */ - out_ptr = allocate_register(&m, 3); - in_ptr = allocate_register(&m, 4); - shuf_ptr = allocate_register(&m, 5); - - - /* Generate code for the individual attribute fetch functions. - */ - for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { - unsigned offset; - - if (function_index[i] == i) { - cell->attrib_fetch_offsets[i] = (unsigned) ((void *) p->csr - - (void *) p->store); - - offset = 0; - emit_fetch(p, & m, in_ptr, &offset, out_ptr, shuf_ptr, - draw->vertex_element[i].src_format); - spe_bi(p, 0, 0, 0); - - /* Round up to the next 16-byte boundary. - */ - if ((((unsigned) p->store) & 0x0f) != 0) { - const unsigned align = ((unsigned) p->store) & 0x0f; - p->store = (uint32_t *) (((void *) p->store) + align); - } - } else { - /* Use the same function entry-point as a previously seen attribute - * with the same format. - */ - cell->attrib_fetch_offsets[i] = - cell->attrib_fetch_offsets[function_index[i]]; - } - } - - static first_time = 1; - if (first_time) { - first_time = 0; - const unsigned instructions = p->csr - p->store; - for (i = 0; i < instructions; i++) { - printf("\t.long\t0x%08x\n", p->store[i]); - } - } -} -- cgit v1.2.3 From b08d3fa249c830d274dca362b8f824b75fe26945 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 15 Feb 2008 10:00:31 -0800 Subject: Make this file build on non-SSE builds (e.g., Cell) --- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 04349cb404..27bc66812c 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -243,7 +243,7 @@ struct draw_vertex_shader * draw_create_vs_sse( struct draw_context *draw, const struct pipe_shader_state *templ ) { - return NULL; + return (void *) 0; } -- cgit v1.2.3 From 3320b1874e810583f95b93a89697b2955987b84f Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 15 Feb 2008 11:03:54 -0800 Subject: Cell: Enable code gen for SPE attribute fetch Doubles are still unsupported. --- src/gallium/drivers/cell/common.h | 15 +- src/gallium/drivers/cell/ppu/Makefile | 1 + src/gallium/drivers/cell/ppu/cell_context.h | 4 + src/gallium/drivers/cell/ppu/cell_vertex_fetch.c | 15 +- src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 21 +- src/gallium/drivers/cell/spu/spu_main.c | 22 +- src/gallium/drivers/cell/spu/spu_vertex_fetch.c | 477 +--------------------- src/gallium/drivers/cell/spu/spu_vertex_shader.h | 6 +- 8 files changed, 71 insertions(+), 490 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index 4de514c358..74b131fbef 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -90,6 +90,7 @@ #define CELL_CMD_STATE_VS_ARRAY_INFO 16 #define CELL_CMD_STATE_BLEND 17 #define CELL_CMD_VS_EXECUTE 18 +#define CELL_CMD_STATE_ATTRIB_FETCH 19 #define CELL_NUM_BUFFERS 4 @@ -128,13 +129,19 @@ struct cell_command_clear_surface */ struct cell_array_info { - uint64_t base; /**< Base address of the 0th element. */ - uint attr; /**< Attribute that this state is for. */ - uint pitch; /**< Byte pitch from one entry to the next. */ - uint format; /**< Pipe format of each entry. */ + uint64_t base; /**< Base address of the 0th element. */ + uint attr; /**< Attribute that this state is for. */ + uint pitch; /**< Byte pitch from one entry to the next. */ + uint size; + uint function_offset; } ALIGN16_ATTRIB; +struct cell_attribute_fetch_code { + uint64_t base; + uint size; +}; + struct cell_shader_info { unsigned num_outputs; diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile index a4c3f29e8a..196ab777f5 100644 --- a/src/gallium/drivers/cell/ppu/Makefile +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -34,6 +34,7 @@ SOURCES = \ cell_surface.c \ cell_texture.c \ cell_vbuf.c \ + cell_vertex_fetch.c \ cell_vertex_shader.c \ cell_winsys.c diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 6196c0c72f..91f8e542a2 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -36,6 +36,7 @@ #include "draw/draw_vbuf.h" #include "cell_winsys.h" #include "cell/common.h" +#include "ppc/rtasm/spe_asm.h" struct cell_vbuf_render; @@ -111,6 +112,9 @@ struct cell_context /** [4] to ensure 16-byte alignment for each status word */ uint buffer_status[CELL_MAX_SPUS][CELL_NUM_BUFFERS][4] ALIGN16_ATTRIB; + + struct spe_function attrib_fetch; + unsigned attrib_fetch_offsets[PIPE_ATTRIB_MAX]; }; diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c index f2432f4ff5..f10689a959 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c @@ -27,10 +27,10 @@ #include "pipe/p_context.h" #include "pipe/p_format.h" -#include "pipe/draw/draw_context.h" -#include "pipe/draw/draw_private.h" +#include "../auxiliary/draw/draw_context.h" +#include "../auxiliary/draw/draw_private.h" -#include "pipe/cell/ppu/cell_context.h" +#include "cell_context.h" #include "ppc/rtasm/spe_asm.h" typedef uint64_t register_mask; @@ -380,13 +380,4 @@ void cell_update_vertex_fetch(struct draw_context *draw) cell->attrib_fetch_offsets[function_index[i]]; } } - - static first_time = 1; - if (first_time) { - first_time = 0; - const unsigned instructions = p->csr - p->store; - for (i = 0; i < instructions; i++) { - printf("\t.long\t0x%08x\n", p->store[i]); - } - } } diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index 0ba4506505..6a1d3bc20a 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -55,14 +55,32 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) uint64_t *batch; struct cell_array_info *array_info; unsigned i, j; + struct cell_attribute_fetch_code *cf; assert(draw->vs.queue_nr != 0); /* XXX: do this on statechange: */ draw_update_vertex_fetch(draw); + cell_update_vertex_fetch(draw); + + + batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*cf)); + batch[0] = CELL_CMD_STATE_ATTRIB_FETCH; + cf = (struct cell_attribute_fetch_code *) (&batch[1]); + cf->base = cell->attrib_fetch.store; + cf->size = ROUNDUP16((unsigned)((void *) cell->attrib_fetch.csr + - (void *) cell->attrib_fetch.store)); + for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { + const enum pipe_format format = draw->vertex_element[i].src_format; + const unsigned count = ((pf_size_x(format) != 0) + + (pf_size_y(format) != 0) + + (pf_size_z(format) != 0) + + (pf_size_w(format) != 0)); + const unsigned size = pf_size_x(format) * count; + batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*array_info)); batch[0] = CELL_CMD_STATE_VS_ARRAY_INFO; @@ -72,7 +90,8 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) array_info->base = (uintptr_t) draw->vertex_fetch.src_ptr[i]; array_info->attr = i; array_info->pitch = draw->vertex_fetch.pitch[i]; - array_info->format = draw->vertex_element[i].src_format; + array_info->size = size; + array_info->function_offset = cell->attrib_fetch_offsets[i]; } batch = cell_batch_alloc(cell, sizeof(batch[0]) diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 1e7243b863..fcbf0f841e 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -54,6 +54,9 @@ struct spu_global spu; struct spu_vs_context draw; +static unsigned char attribute_fetch_code_buffer[136 * PIPE_ATTRIB_MAX] + ALIGN16_ATTRIB; + /** * Tell the PPU that this SPU has finished copying a buffer to * local store and that it may be reused by the PPU. @@ -306,7 +309,8 @@ cmd_state_vs_array_info(const struct cell_array_info *vs_info) ASSERT(attr < PIPE_ATTRIB_MAX); draw.vertex_fetch.src_ptr[attr] = vs_info->base; draw.vertex_fetch.pitch[attr] = vs_info->pitch; - draw.vertex_fetch.format[attr] = vs_info->format; + draw.vertex_fetch.size[attr] = vs_info->size; + draw.vertex_fetch.code_offset[attr] = vs_info->function_offset; draw.vertex_fetch.dirty = 1; } @@ -433,6 +437,22 @@ cmd_batch(uint opcode) cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); pos += (1 + ROUNDUP8(sizeof(struct cell_array_info)) / 8); break; + case CELL_CMD_STATE_ATTRIB_FETCH: { + struct cell_attribute_fetch_code *code = + (struct cell_attribute_fetch_code *) &buffer[pos+1]; + + mfc_get(attribute_fetch_code_buffer, + (unsigned int) code->base, /* src */ + code->size, + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); + + draw.vertex_fetch.code = attribute_fetch_code_buffer; + pos += (1 + ROUNDUP8(sizeof(struct cell_attribute_fetch_code)) / 8); + break; + } default: printf("SPU %u: bad opcode: 0x%llx\n", spu.init.id, buffer[pos]); ASSERT(0); diff --git a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c index 45e3c26c00..55c6c28717 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c @@ -1,6 +1,7 @@ /************************************************************************** * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * (C) Copyright IBM Corporation 2008 * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -28,10 +29,10 @@ /* * Authors: * Keith Whitwell + * Ian Romanick */ #include -#include #include "pipe/p_util.h" #include "pipe/p_state.h" @@ -59,6 +60,10 @@ #define DRAW_DBG 0 +typedef void (*spu_fetch_func)(qword *out, const qword *in, + const qword *shuffle_data); + + static const qword fetch_shuffle_data[] = { /* Shuffle used by CVT_64_FLOAT */ @@ -97,22 +102,6 @@ static const qword fetch_shuffle_data[] = { }; -static INLINE void -trans4x4(qword row0, qword row1, qword row2, qword row3, qword *out, - const qword *shuffle) -{ - qword t1 = si_shufb(row0, row2, shuffle[3]); - qword t2 = si_shufb(row0, row2, shuffle[4]); - qword t3 = si_shufb(row1, row3, shuffle[3]); - qword t4 = si_shufb(row1, row3, shuffle[4]); - - out[0] = si_shufb(t1, t3, shuffle[3]); - out[1] = si_shufb(t1, t3, shuffle[4]); - out[2] = si_shufb(t2, t4, shuffle[3]); - out[3] = si_shufb(t2, t4, shuffle[4]); -} - - /** * Fetch between 1 and 32 bytes from an unaligned address */ @@ -151,446 +140,6 @@ fetch_unaligned(qword *dst, unsigned ea, unsigned size) } -#define CVT_32_FLOAT(q, s) (*(q)) - -static INLINE qword -CVT_64_FLOAT(const qword *qw, const qword *shuffle) -{ - qword a = si_frds(qw[0]); - qword b = si_frds(si_rotqbyi(qw[0], 8)); - qword c = si_frds(qw[1]); - qword d = si_frds(si_rotqbyi(qw[1], 8)); - - qword ab = si_shufb(a, b, shuffle[0]); - qword cd = si_shufb(c, d, si_rotqbyi(shuffle[0], 8)); - - return si_or(ab, cd); -} - - -static INLINE qword -CVT_8_USCALED(const qword *qw, const qword *shuffle) -{ - return si_cuflt(si_shufb(*qw, *qw, shuffle[1]), 0); -} - - -static INLINE qword -CVT_16_USCALED(const qword *qw, const qword *shuffle) -{ - return si_cuflt(si_shufb(*qw, *qw, shuffle[2]), 0); -} - - -static INLINE qword -CVT_32_USCALED(const qword *qw, const qword *shuffle) -{ - (void) shuffle; - return si_cuflt(*qw, 0); -} - -static INLINE qword -CVT_8_SSCALED(const qword *qw, const qword *shuffle) -{ - return si_csflt(si_shufb(*qw, *qw, shuffle[1]), 0); -} - - -static INLINE qword -CVT_16_SSCALED(const qword *qw, const qword *shuffle) -{ - return si_csflt(si_shufb(*qw, *qw, shuffle[2]), 0); -} - - -static INLINE qword -CVT_32_SSCALED(const qword *qw, const qword *shuffle) -{ - (void) shuffle; - return si_csflt(*qw, 0); -} - - -static INLINE qword -CVT_8_UNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 255.0f); - return si_fm(CVT_8_USCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_16_UNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 65535.0f); - return si_fm(CVT_16_USCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_32_UNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 4294967295.0f); - return si_fm(CVT_32_USCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_8_SNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 127.0f); - return si_fm(CVT_8_SSCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_16_SNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 32767.0f); - return si_fm(CVT_16_SSCALED(qw, shuffle), scale); -} - - -static INLINE qword -CVT_32_SNORM(const qword *qw, const qword *shuffle) -{ - const qword scale = (qword) spu_splats(1.0f / 2147483647.0f); - return si_fm(CVT_32_SSCALED(qw, shuffle), scale); -} - -#define SZ_4 si_il(0U) -#define SZ_3 si_fsmbi(0x000f) -#define SZ_2 si_fsmbi(0x00ff) -#define SZ_1 si_fsmbi(0x0fff) - -/** - * Fetch a float[4] vertex attribute from memory, doing format/type - * conversion as needed. - * - * This is probably needed/dupliocated elsewhere, eg format - * conversion, texture sampling etc. - */ -#define FETCH_ATTRIB( NAME, SZ, CVT, N ) \ -static void \ -fetch_##NAME(qword *out, const qword *in, qword defaults, \ - const qword *shuffle) \ -{ \ - qword tmp[4]; \ - \ - tmp[0] = si_selb(CVT(in + (0 * N), shuffle), defaults, SZ); \ - tmp[1] = si_selb(CVT(in + (1 * N), shuffle), defaults, SZ); \ - tmp[2] = si_selb(CVT(in + (2 * N), shuffle), defaults, SZ); \ - tmp[3] = si_selb(CVT(in + (3 * N), shuffle), defaults, SZ); \ - trans4x4(tmp[0], tmp[1], tmp[2], tmp[3], out, shuffle); \ -} - - -FETCH_ATTRIB( R64G64B64A64_FLOAT, SZ_4, CVT_64_FLOAT, 2 ) -FETCH_ATTRIB( R64G64B64_FLOAT, SZ_3, CVT_64_FLOAT, 2 ) -FETCH_ATTRIB( R64G64_FLOAT, SZ_2, CVT_64_FLOAT, 2 ) -FETCH_ATTRIB( R64_FLOAT, SZ_1, CVT_64_FLOAT, 2 ) - -FETCH_ATTRIB( R32G32B32A32_FLOAT, SZ_4, CVT_32_FLOAT, 1 ) -FETCH_ATTRIB( R32G32B32_FLOAT, SZ_3, CVT_32_FLOAT, 1 ) -FETCH_ATTRIB( R32G32_FLOAT, SZ_2, CVT_32_FLOAT, 1 ) -FETCH_ATTRIB( R32_FLOAT, SZ_1, CVT_32_FLOAT, 1 ) - -FETCH_ATTRIB( R32G32B32A32_USCALED, SZ_4, CVT_32_USCALED, 1 ) -FETCH_ATTRIB( R32G32B32_USCALED, SZ_3, CVT_32_USCALED, 1 ) -FETCH_ATTRIB( R32G32_USCALED, SZ_2, CVT_32_USCALED, 1 ) -FETCH_ATTRIB( R32_USCALED, SZ_1, CVT_32_USCALED, 1 ) - -FETCH_ATTRIB( R32G32B32A32_SSCALED, SZ_4, CVT_32_SSCALED, 1 ) -FETCH_ATTRIB( R32G32B32_SSCALED, SZ_3, CVT_32_SSCALED, 1 ) -FETCH_ATTRIB( R32G32_SSCALED, SZ_2, CVT_32_SSCALED, 1 ) -FETCH_ATTRIB( R32_SSCALED, SZ_1, CVT_32_SSCALED, 1 ) - -FETCH_ATTRIB( R32G32B32A32_UNORM, SZ_4, CVT_32_UNORM, 1 ) -FETCH_ATTRIB( R32G32B32_UNORM, SZ_3, CVT_32_UNORM, 1 ) -FETCH_ATTRIB( R32G32_UNORM, SZ_2, CVT_32_UNORM, 1 ) -FETCH_ATTRIB( R32_UNORM, SZ_1, CVT_32_UNORM, 1 ) - -FETCH_ATTRIB( R32G32B32A32_SNORM, SZ_4, CVT_32_SNORM, 1 ) -FETCH_ATTRIB( R32G32B32_SNORM, SZ_3, CVT_32_SNORM, 1 ) -FETCH_ATTRIB( R32G32_SNORM, SZ_2, CVT_32_SNORM, 1 ) -FETCH_ATTRIB( R32_SNORM, SZ_1, CVT_32_SNORM, 1 ) - -FETCH_ATTRIB( R16G16B16A16_USCALED, SZ_4, CVT_16_USCALED, 1 ) -FETCH_ATTRIB( R16G16B16_USCALED, SZ_3, CVT_16_USCALED, 1 ) -FETCH_ATTRIB( R16G16_USCALED, SZ_2, CVT_16_USCALED, 1 ) -FETCH_ATTRIB( R16_USCALED, SZ_1, CVT_16_USCALED, 1 ) - -FETCH_ATTRIB( R16G16B16A16_SSCALED, SZ_4, CVT_16_SSCALED, 1 ) -FETCH_ATTRIB( R16G16B16_SSCALED, SZ_3, CVT_16_SSCALED, 1 ) -FETCH_ATTRIB( R16G16_SSCALED, SZ_2, CVT_16_SSCALED, 1 ) -FETCH_ATTRIB( R16_SSCALED, SZ_1, CVT_16_SSCALED, 1 ) - -FETCH_ATTRIB( R16G16B16A16_UNORM, SZ_4, CVT_16_UNORM, 1 ) -FETCH_ATTRIB( R16G16B16_UNORM, SZ_3, CVT_16_UNORM, 1 ) -FETCH_ATTRIB( R16G16_UNORM, SZ_2, CVT_16_UNORM, 1 ) -FETCH_ATTRIB( R16_UNORM, SZ_1, CVT_16_UNORM, 1 ) - -FETCH_ATTRIB( R16G16B16A16_SNORM, SZ_4, CVT_16_SNORM, 1 ) -FETCH_ATTRIB( R16G16B16_SNORM, SZ_3, CVT_16_SNORM, 1 ) -FETCH_ATTRIB( R16G16_SNORM, SZ_2, CVT_16_SNORM, 1 ) -FETCH_ATTRIB( R16_SNORM, SZ_1, CVT_16_SNORM, 1 ) - -FETCH_ATTRIB( R8G8B8A8_USCALED, SZ_4, CVT_8_USCALED, 1 ) -FETCH_ATTRIB( R8G8B8_USCALED, SZ_3, CVT_8_USCALED, 1 ) -FETCH_ATTRIB( R8G8_USCALED, SZ_2, CVT_8_USCALED, 1 ) -FETCH_ATTRIB( R8_USCALED, SZ_1, CVT_8_USCALED, 1 ) - -FETCH_ATTRIB( R8G8B8A8_SSCALED, SZ_4, CVT_8_SSCALED, 1 ) -FETCH_ATTRIB( R8G8B8_SSCALED, SZ_3, CVT_8_SSCALED, 1 ) -FETCH_ATTRIB( R8G8_SSCALED, SZ_2, CVT_8_SSCALED, 1 ) -FETCH_ATTRIB( R8_SSCALED, SZ_1, CVT_8_SSCALED, 1 ) - -FETCH_ATTRIB( R8G8B8A8_UNORM, SZ_4, CVT_8_UNORM, 1 ) -FETCH_ATTRIB( R8G8B8_UNORM, SZ_3, CVT_8_UNORM, 1 ) -FETCH_ATTRIB( R8G8_UNORM, SZ_2, CVT_8_UNORM, 1 ) -FETCH_ATTRIB( R8_UNORM, SZ_1, CVT_8_UNORM, 1 ) - -FETCH_ATTRIB( R8G8B8A8_SNORM, SZ_4, CVT_8_SNORM, 1 ) -FETCH_ATTRIB( R8G8B8_SNORM, SZ_3, CVT_8_SNORM, 1 ) -FETCH_ATTRIB( R8G8_SNORM, SZ_2, CVT_8_SNORM, 1 ) -FETCH_ATTRIB( R8_SNORM, SZ_1, CVT_8_SNORM, 1 ) - -FETCH_ATTRIB( A8R8G8B8_UNORM, SZ_4, CVT_8_UNORM, 1 ) - - - -static spu_fetch_func get_fetch_func( enum pipe_format format ) -{ - switch (format) { - case PIPE_FORMAT_R64_FLOAT: - return fetch_R64_FLOAT; - case PIPE_FORMAT_R64G64_FLOAT: - return fetch_R64G64_FLOAT; - case PIPE_FORMAT_R64G64B64_FLOAT: - return fetch_R64G64B64_FLOAT; - case PIPE_FORMAT_R64G64B64A64_FLOAT: - return fetch_R64G64B64A64_FLOAT; - - case PIPE_FORMAT_R32_FLOAT: - return fetch_R32_FLOAT; - case PIPE_FORMAT_R32G32_FLOAT: - return fetch_R32G32_FLOAT; - case PIPE_FORMAT_R32G32B32_FLOAT: - return fetch_R32G32B32_FLOAT; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return fetch_R32G32B32A32_FLOAT; - - case PIPE_FORMAT_R32_UNORM: - return fetch_R32_UNORM; - case PIPE_FORMAT_R32G32_UNORM: - return fetch_R32G32_UNORM; - case PIPE_FORMAT_R32G32B32_UNORM: - return fetch_R32G32B32_UNORM; - case PIPE_FORMAT_R32G32B32A32_UNORM: - return fetch_R32G32B32A32_UNORM; - - case PIPE_FORMAT_R32_USCALED: - return fetch_R32_USCALED; - case PIPE_FORMAT_R32G32_USCALED: - return fetch_R32G32_USCALED; - case PIPE_FORMAT_R32G32B32_USCALED: - return fetch_R32G32B32_USCALED; - case PIPE_FORMAT_R32G32B32A32_USCALED: - return fetch_R32G32B32A32_USCALED; - - case PIPE_FORMAT_R32_SNORM: - return fetch_R32_SNORM; - case PIPE_FORMAT_R32G32_SNORM: - return fetch_R32G32_SNORM; - case PIPE_FORMAT_R32G32B32_SNORM: - return fetch_R32G32B32_SNORM; - case PIPE_FORMAT_R32G32B32A32_SNORM: - return fetch_R32G32B32A32_SNORM; - - case PIPE_FORMAT_R32_SSCALED: - return fetch_R32_SSCALED; - case PIPE_FORMAT_R32G32_SSCALED: - return fetch_R32G32_SSCALED; - case PIPE_FORMAT_R32G32B32_SSCALED: - return fetch_R32G32B32_SSCALED; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - return fetch_R32G32B32A32_SSCALED; - - case PIPE_FORMAT_R16_UNORM: - return fetch_R16_UNORM; - case PIPE_FORMAT_R16G16_UNORM: - return fetch_R16G16_UNORM; - case PIPE_FORMAT_R16G16B16_UNORM: - return fetch_R16G16B16_UNORM; - case PIPE_FORMAT_R16G16B16A16_UNORM: - return fetch_R16G16B16A16_UNORM; - - case PIPE_FORMAT_R16_USCALED: - return fetch_R16_USCALED; - case PIPE_FORMAT_R16G16_USCALED: - return fetch_R16G16_USCALED; - case PIPE_FORMAT_R16G16B16_USCALED: - return fetch_R16G16B16_USCALED; - case PIPE_FORMAT_R16G16B16A16_USCALED: - return fetch_R16G16B16A16_USCALED; - - case PIPE_FORMAT_R16_SNORM: - return fetch_R16_SNORM; - case PIPE_FORMAT_R16G16_SNORM: - return fetch_R16G16_SNORM; - case PIPE_FORMAT_R16G16B16_SNORM: - return fetch_R16G16B16_SNORM; - case PIPE_FORMAT_R16G16B16A16_SNORM: - return fetch_R16G16B16A16_SNORM; - - case PIPE_FORMAT_R16_SSCALED: - return fetch_R16_SSCALED; - case PIPE_FORMAT_R16G16_SSCALED: - return fetch_R16G16_SSCALED; - case PIPE_FORMAT_R16G16B16_SSCALED: - return fetch_R16G16B16_SSCALED; - case PIPE_FORMAT_R16G16B16A16_SSCALED: - return fetch_R16G16B16A16_SSCALED; - - case PIPE_FORMAT_R8_UNORM: - return fetch_R8_UNORM; - case PIPE_FORMAT_R8G8_UNORM: - return fetch_R8G8_UNORM; - case PIPE_FORMAT_R8G8B8_UNORM: - return fetch_R8G8B8_UNORM; - case PIPE_FORMAT_R8G8B8A8_UNORM: - return fetch_R8G8B8A8_UNORM; - - case PIPE_FORMAT_R8_USCALED: - return fetch_R8_USCALED; - case PIPE_FORMAT_R8G8_USCALED: - return fetch_R8G8_USCALED; - case PIPE_FORMAT_R8G8B8_USCALED: - return fetch_R8G8B8_USCALED; - case PIPE_FORMAT_R8G8B8A8_USCALED: - return fetch_R8G8B8A8_USCALED; - - case PIPE_FORMAT_R8_SNORM: - return fetch_R8_SNORM; - case PIPE_FORMAT_R8G8_SNORM: - return fetch_R8G8_SNORM; - case PIPE_FORMAT_R8G8B8_SNORM: - return fetch_R8G8B8_SNORM; - case PIPE_FORMAT_R8G8B8A8_SNORM: - return fetch_R8G8B8A8_SNORM; - - case PIPE_FORMAT_R8_SSCALED: - return fetch_R8_SSCALED; - case PIPE_FORMAT_R8G8_SSCALED: - return fetch_R8G8_SSCALED; - case PIPE_FORMAT_R8G8B8_SSCALED: - return fetch_R8G8B8_SSCALED; - case PIPE_FORMAT_R8G8B8A8_SSCALED: - return fetch_R8G8B8A8_SSCALED; - - case PIPE_FORMAT_A8R8G8B8_UNORM: - return fetch_A8R8G8B8_UNORM; - - case 0: - return NULL; /* not sure why this is needed */ - - default: - assert(0); - return NULL; - } -} - - -static unsigned get_vertex_size( enum pipe_format format ) -{ - switch (format) { - case PIPE_FORMAT_R64_FLOAT: - return 8; - case PIPE_FORMAT_R64G64_FLOAT: - return 2 * 8; - case PIPE_FORMAT_R64G64B64_FLOAT: - return 3 * 8; - case PIPE_FORMAT_R64G64B64A64_FLOAT: - return 4 * 8; - - case PIPE_FORMAT_R32_SSCALED: - case PIPE_FORMAT_R32_SNORM: - case PIPE_FORMAT_R32_USCALED: - case PIPE_FORMAT_R32_UNORM: - case PIPE_FORMAT_R32_FLOAT: - return 4; - case PIPE_FORMAT_R32G32_SSCALED: - case PIPE_FORMAT_R32G32_SNORM: - case PIPE_FORMAT_R32G32_USCALED: - case PIPE_FORMAT_R32G32_UNORM: - case PIPE_FORMAT_R32G32_FLOAT: - return 2 * 4; - case PIPE_FORMAT_R32G32B32_SSCALED: - case PIPE_FORMAT_R32G32B32_SNORM: - case PIPE_FORMAT_R32G32B32_USCALED: - case PIPE_FORMAT_R32G32B32_UNORM: - case PIPE_FORMAT_R32G32B32_FLOAT: - return 3 * 4; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - case PIPE_FORMAT_R32G32B32A32_SNORM: - case PIPE_FORMAT_R32G32B32A32_USCALED: - case PIPE_FORMAT_R32G32B32A32_UNORM: - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return 4 * 4; - - case PIPE_FORMAT_R16_SSCALED: - case PIPE_FORMAT_R16_SNORM: - case PIPE_FORMAT_R16_UNORM: - case PIPE_FORMAT_R16_USCALED: - return 2; - case PIPE_FORMAT_R16G16_SSCALED: - case PIPE_FORMAT_R16G16_SNORM: - case PIPE_FORMAT_R16G16_USCALED: - case PIPE_FORMAT_R16G16_UNORM: - return 2 * 2; - case PIPE_FORMAT_R16G16B16_SSCALED: - case PIPE_FORMAT_R16G16B16_SNORM: - case PIPE_FORMAT_R16G16B16_USCALED: - case PIPE_FORMAT_R16G16B16_UNORM: - return 3 * 2; - case PIPE_FORMAT_R16G16B16A16_SSCALED: - case PIPE_FORMAT_R16G16B16A16_SNORM: - case PIPE_FORMAT_R16G16B16A16_USCALED: - case PIPE_FORMAT_R16G16B16A16_UNORM: - return 4 * 2; - - case PIPE_FORMAT_R8_SSCALED: - case PIPE_FORMAT_R8_SNORM: - case PIPE_FORMAT_R8_USCALED: - case PIPE_FORMAT_R8_UNORM: - return 1; - case PIPE_FORMAT_R8G8_SSCALED: - case PIPE_FORMAT_R8G8_SNORM: - case PIPE_FORMAT_R8G8_USCALED: - case PIPE_FORMAT_R8G8_UNORM: - return 2 * 1; - case PIPE_FORMAT_R8G8B8_SSCALED: - case PIPE_FORMAT_R8G8B8_SNORM: - case PIPE_FORMAT_R8G8B8_USCALED: - case PIPE_FORMAT_R8G8B8_UNORM: - return 3 * 1; - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R8G8B8A8_SSCALED: - case PIPE_FORMAT_R8G8B8A8_SNORM: - case PIPE_FORMAT_R8G8B8A8_USCALED: - case PIPE_FORMAT_R8G8B8A8_UNORM: - return 4 * 1; - - case 0: - return 0; /* not sure why this is needed */ - - default: - assert(0); - return 0; - } -} - - /** * Fetch vertex attributes for 'count' vertices. */ @@ -612,10 +161,10 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, /* loop over vertex attributes (vertex shader inputs) */ for (attr = 0; attr < nr_attrs; attr++) { - const qword default_values = (qword)(vec_float4){ 0.0, 0.0, 0.0, 1.0 }; const unsigned pitch = draw->vertex_fetch.pitch[attr]; const uint64_t src = draw->vertex_fetch.src_ptr[attr]; - const spu_fetch_func fetch = draw->vertex_fetch.fetch[attr]; + const spu_fetch_func fetch = (spu_fetch_func) + (draw->vertex_fetch.code + draw->vertex_fetch.code_offset[attr]); unsigned i; unsigned idx; const unsigned bytes_per_entry = draw->vertex_fetch.size[attr]; @@ -644,8 +193,7 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, /* Convert all 4 vertices to vectors of float. */ - (*fetch)(&machine->Inputs[attr].xyzw[0].q, in, default_values, - fetch_shuffle_data); + (*fetch)(&machine->Inputs[attr].xyzw[0].q, in, fetch_shuffle_data); } } @@ -662,12 +210,5 @@ void spu_update_vertex_fetch( struct spu_vs_context *draw ) } - for (i = 0; i < draw->vertex_fetch.nr_attrs; i++) { - draw->vertex_fetch.fetch[i] = - get_fetch_func(draw->vertex_fetch.format[i]); - draw->vertex_fetch.size[i] = - get_vertex_size(draw->vertex_fetch.format[i]); - } - draw->vertex_fetch.fetch_func = generic_vertex_fetch; } diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.h b/src/gallium/drivers/cell/spu/spu_vertex_shader.h index b5bf31e67d..0fb0bc28d0 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_shader.h +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.h @@ -6,8 +6,6 @@ struct spu_vs_context; -typedef void (*spu_fetch_func)(qword *out, const qword *in, qword defaults, - const qword *shuffle_data); typedef void (*spu_full_fetch_func)( struct spu_vs_context *draw, struct spu_exec_machine *machine, const unsigned *elts, @@ -20,12 +18,12 @@ struct spu_vs_context { uint64_t src_ptr[PIPE_ATTRIB_MAX]; unsigned pitch[PIPE_ATTRIB_MAX]; unsigned size[PIPE_ATTRIB_MAX]; - enum pipe_format format[PIPE_ATTRIB_MAX]; + unsigned code_offset[PIPE_ATTRIB_MAX]; unsigned nr_attrs; boolean dirty; - spu_fetch_func fetch[PIPE_ATTRIB_MAX]; spu_full_fetch_func fetch_func; + void *code; } vertex_fetch; /* Clip derived state: -- cgit v1.2.3 From 08c9534107fcaf06f9b801551524ed5dc724db13 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 16 Feb 2008 10:05:01 -0700 Subject: gallium: implement min vs. mag filter determination for non-mipmapped textures Fixes tests/minmag.c --- src/gallium/drivers/softpipe/sp_tex_sample.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 2f82fd6abe..c0128f81d7 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -476,6 +476,19 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, /* no mipmap selection needed */ *imgFilter = sampler->state->mag_img_filter; *level0 = *level1 = (int) sampler->state->min_lod; + + if (sampler->state->min_img_filter != sampler->state->mag_img_filter) { + /* non-mipmapped texture, but still need to determine if doing + * minification or magnification. + */ + float lambda = compute_lambda(sampler, s, t, p, lodbias); + if (lambda < 0.5) { /* XXX this may need tweaking... */ + *imgFilter = sampler->state->mag_img_filter; + } + else { + *imgFilter = sampler->state->min_img_filter; + } + } } else { float lambda; -- cgit v1.2.3 From 3b2a291888d8e62787de03f8529806fb562bd186 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 16 Feb 2008 13:50:31 -0700 Subject: gallium: tweak texture filter min/mag thresholds --- src/gallium/drivers/softpipe/sp_tex_sample.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index c0128f81d7..a15bd43166 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -474,7 +474,6 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, { if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { /* no mipmap selection needed */ - *imgFilter = sampler->state->mag_img_filter; *level0 = *level1 = (int) sampler->state->min_lod; if (sampler->state->min_img_filter != sampler->state->mag_img_filter) { @@ -482,13 +481,16 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, * minification or magnification. */ float lambda = compute_lambda(sampler, s, t, p, lodbias); - if (lambda < 0.5) { /* XXX this may need tweaking... */ + if (lambda <= 0.0) { *imgFilter = sampler->state->mag_img_filter; } else { *imgFilter = sampler->state->min_img_filter; } } + else { + *imgFilter = sampler->state->mag_img_filter; + } } else { float lambda; @@ -500,7 +502,7 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, /* vertex shader */ lambda = lodbias; /* not really a bias, but absolute LOD */ - if (lambda < 0.0) { /* XXX threshold depends on the filter */ + if (lambda <= 0.0) { /* XXX threshold depends on the filter */ /* magnifying */ *imgFilter = sampler->state->mag_img_filter; *level0 = *level1 = 0; -- cgit v1.2.3 From 0c6bbd41bd6dc1041eaca7c907d3768d107c1afa Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 16 Feb 2008 13:55:47 -0700 Subject: gallium: add missing mip level clamp --- src/gallium/drivers/softpipe/sp_tex_sample.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index a15bd43166..c54e9d385c 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -474,7 +474,8 @@ choose_mipmap_levels(struct tgsi_sampler *sampler, { if (sampler->state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) { /* no mipmap selection needed */ - *level0 = *level1 = (int) sampler->state->min_lod; + *level0 = *level1 = CLAMP((int) sampler->state->min_lod, + 0, (int) sampler->texture->last_level); if (sampler->state->min_img_filter != sampler->state->mag_img_filter) { /* non-mipmapped texture, but still need to determine if doing -- cgit v1.2.3 From 4a79156812d574249b51e1692f4615aa31bf0e50 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sun, 17 Feb 2008 09:42:26 -0500 Subject: fix the build --- src/gallium/auxiliary/llvm/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/llvm/Makefile b/src/gallium/auxiliary/llvm/Makefile index e0abf860c1..39fac6ea4a 100644 --- a/src/gallium/auxiliary/llvm/Makefile +++ b/src/gallium/auxiliary/llvm/Makefile @@ -30,7 +30,7 @@ OBJECTS = $(C_SOURCES:.c=.o) \ ### Include directories INCLUDES = \ -I. \ - -I$(TOP)/src/gallium/drivers + -I$(TOP)/src/gallium/drivers \ -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/mesa \ -- cgit v1.2.3 From bfd5916eafb9a97ad10f1d4a8738e7dcb02e04f4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 18 Feb 2008 14:25:04 +0900 Subject: Code reorganization: split gallium and mesa makefiles. In other words, don't build src/gallium source code from within src/mesa/Makefile. Also, allow to customize which gallium auxiliary dirs, driver driver, winsys dirs get built from the config/* files. --- src/gallium/Makefile.template | 1 + src/gallium/auxiliary/Makefile | 6 +- src/gallium/auxiliary/cso_cache/Makefile | 18 +++++ src/gallium/auxiliary/draw/Makefile | 41 ++++++++++- src/gallium/auxiliary/tgsi/Makefile | 24 ++++++- src/gallium/auxiliary/tgsi/exec/Makefile | 3 - src/gallium/auxiliary/util/Makefile | 20 ++++++ src/gallium/drivers/Makefile | 6 +- src/gallium/winsys/Makefile | 20 ++++++ src/gallium/winsys/dri/Makefile.template | 6 +- src/gallium/winsys/xlib/Makefile | 113 +++++++++++++++++++++++++++++++ src/mesa/Makefile | 74 ++------------------ src/mesa/sources | 56 +-------------- 13 files changed, 247 insertions(+), 141 deletions(-) create mode 100644 src/gallium/auxiliary/cso_cache/Makefile delete mode 100644 src/gallium/auxiliary/tgsi/exec/Makefile create mode 100644 src/gallium/auxiliary/util/Makefile create mode 100644 src/gallium/winsys/Makefile create mode 100644 src/gallium/winsys/xlib/Makefile (limited to 'src') diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template index 83b25f9b47..6ad58c205c 100644 --- a/src/gallium/Makefile.template +++ b/src/gallium/Makefile.template @@ -20,6 +20,7 @@ INCLUDES = \ -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers \ -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main \ -I$(TOP)/include \ $(DRIVER_INCLUDES) diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile index da68498aa1..eaa0f2fe4e 100644 --- a/src/gallium/auxiliary/Makefile +++ b/src/gallium/auxiliary/Makefile @@ -2,11 +2,7 @@ TOP = ../../.. include $(TOP)/configs/current -ifeq ($(CONFIG_NAME), linux-llvm) -LLVM_DIR = llvm -endif - -SUBDIRS = pipebuffer $(LLVM_DIR) +SUBDIRS = $(GALLIUM_AUXILIARY_DIRS) default: subdirs diff --git a/src/gallium/auxiliary/cso_cache/Makefile b/src/gallium/auxiliary/cso_cache/Makefile new file mode 100644 index 0000000000..8248b097fd --- /dev/null +++ b/src/gallium/auxiliary/cso_cache/Makefile @@ -0,0 +1,18 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = cso_cache + +DRIVER_SOURCES = \ + cso_cache.c \ + cso_hash.c + +C_SOURCES = \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index fe9b150f30..c56b63d85b 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -1,2 +1,39 @@ -default: - cd ../../../mesa ; make +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = draw + +DRIVER_SOURCES = \ + draw_clip.c \ + draw_vs_exec.c \ + draw_vs_sse.c \ + draw_vs_llvm.c \ + draw_context.c\ + draw_cull.c \ + draw_debug.c \ + draw_flatshade.c \ + draw_offset.c \ + draw_prim.c \ + draw_stipple.c \ + draw_twoside.c \ + draw_unfilled.c \ + draw_validate.c \ + draw_vbuf.c \ + draw_vertex.c \ + draw_vertex_cache.c \ + draw_vertex_fetch.c \ + draw_vertex_shader.c \ + draw_vf.c \ + draw_vf_generic.c \ + draw_vf_sse.c \ + draw_wide_prims.c + +C_SOURCES = \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index 12a8bd0409..c10ab396d8 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -1,3 +1,23 @@ -default: - cd ../.. ; make + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = tgsi + +DRIVER_SOURCES = \ + exec/tgsi_exec.c \ + exec/tgsi_sse2.c \ + util/tgsi_build.c \ + util/tgsi_dump.c \ + util/tgsi_parse.c \ + util/tgsi_util.c + +C_SOURCES = \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: diff --git a/src/gallium/auxiliary/tgsi/exec/Makefile b/src/gallium/auxiliary/tgsi/exec/Makefile deleted file mode 100644 index eb8b14e0e8..0000000000 --- a/src/gallium/auxiliary/tgsi/exec/Makefile +++ /dev/null @@ -1,3 +0,0 @@ -default: - cd ../../.. ; make - diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile new file mode 100644 index 0000000000..b8cb148c4f --- /dev/null +++ b/src/gallium/auxiliary/util/Makefile @@ -0,0 +1,20 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = util + +DRIVER_SOURCES = \ + p_debug.c \ + p_tile.c \ + p_util.c + +C_SOURCES = \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/drivers/Makefile b/src/gallium/drivers/Makefile index c0345a9cb5..8dcfb18a16 100644 --- a/src/gallium/drivers/Makefile +++ b/src/gallium/drivers/Makefile @@ -2,11 +2,7 @@ TOP = ../../.. include $(TOP)/configs/current -ifeq ($(CONFIG_NAME), linux-cell) -CELL_DIR = cell -endif - -SUBDIRS = softpipe i915simple i965simple failover pipebuffer $(CELL_DIR) +SUBDIRS = softpipe i915simple i965simple failover default: subdirs diff --git a/src/gallium/winsys/Makefile b/src/gallium/winsys/Makefile new file mode 100644 index 0000000000..3dc5ea5744 --- /dev/null +++ b/src/gallium/winsys/Makefile @@ -0,0 +1,20 @@ +TOP = ../../.. +include $(TOP)/configs/current + + +SUBDIRS = $(GALLIUM_WINSYS_DIRS) + + +default: subdirs + + +subdirs: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +clean: + rm -f `find . -name \*.[oa]` diff --git a/src/gallium/winsys/dri/Makefile.template b/src/gallium/winsys/dri/Makefile.template index 2a261ed669..65a93bd53e 100644 --- a/src/gallium/winsys/dri/Makefile.template +++ b/src/gallium/winsys/dri/Makefile.template @@ -1,7 +1,9 @@ # -*-makefile-*- -MESA_MODULES = $(TOP)/src/mesa/libmesa.a - +MESA_MODULES = \ + $(TOP)/src/mesa/libmesa.a \ + $(GALLIUM_AUXILIARIES) + COMMON_GALLIUM_SOURCES = \ $(TOP)/src/mesa/drivers/dri/common/utils.c \ $(TOP)/src/mesa/drivers/dri/common/vblank.c \ diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile new file mode 100644 index 0000000000..2664ac47ce --- /dev/null +++ b/src/gallium/winsys/xlib/Makefile @@ -0,0 +1,113 @@ +# src/mesa/Makefile + +TOP = ../../../.. +include $(TOP)/configs/current + + +INCLUDE_DIRS = \ + -I$(TOP)/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/auxiliary + +X11_DRIVER_SOURCES = \ + glxapi.c \ + fakeglx.c \ + xfonts.c \ + xm_api.c \ + xm_winsys.c \ + xm_winsys_aub.c \ + brw_aub.c + + +GL_MAJOR = 1 +GL_MINOR = 5 +GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) + + +PIPE_LIB = \ + $(GALLIUM_DRIVERS) \ + $(TOP)/src/mesa/libglapi.a \ + $(TOP)/src/mesa/libmesa.a \ + $(GALLIUM_AUXILIARIES) + +ifeq ($(CONFIG_NAME), linux-cell) +CELL_LIB = $(TOP)/src/gallium/drivers/cell/ppu/libcell.a +CELL_LIB_SPU = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a +endif + +ifeq ($(CONFIG_NAME), linux-llvm) +LLVM_LIB = $(TOP)/src/gallium/auxiliary/llvm/libgallivm.a +endif + + +.SUFFIXES : .cpp + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDE_DIRS) $(CXXFLAGS) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + + +default: $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) + + +###################################################################### +# Stand-alone Mesa libGL and libOSMesa +STAND_ALONE_DRIVER_SOURCES = \ + $(X11_DRIVER_SOURCES) + +STAND_ALONE_DRIVER_OBJECTS = $(STAND_ALONE_DRIVER_SOURCES:.c=.o) + +STAND_ALONE_OBJECTS = \ + $(STAND_ALONE_DRIVER_OBJECTS) + +# Make the GL library +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(LLVM_LIB) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) + $(TOP)/bin/mklib -o $(GL_LIB) \ + -linker "$(CC)" \ + -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ + -install $(TOP)/$(LIB_DIR) \ + $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ + --start-group $(PIPE_LIB) $(LLVM_LIB) --end-group $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) + + +###################################################################### +# Generic stuff + +depend: $(ALL_SOURCES) + @ echo "running $(MKDEP)" + @ rm -f depend # workaround oops on gutsy?!? + @ touch depend + @ $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \ + > /dev/null 2>/dev/null + + +install: default + $(INSTALL) -d $(INSTALL_DIR)/include/GL + $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) -m 644 $(TOP)/include/GL/*.h $(INSTALL_DIR)/include/GL + @if [ -e $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) ]; then \ + $(INSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \ + fi + +## NOT INSTALLED YET: +## $(INSTALL) -d $(INSTALL_DIR)/include/GLES +## $(INSTALL) -m 644 include/GLES/*.h $(INSTALL_DIR)/include/GLES + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` $(TOP)/include/GL/*.h + +clean: + -rm -f *.o + + +include depend diff --git a/src/mesa/Makefile b/src/mesa/Makefile index c8cb2b592f..86a9cf0b88 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -11,19 +11,6 @@ GL_MINOR = 5 GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) -PIPE_LIB = \ - $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ - $(TOP)/src/gallium/drivers/i965simple/libi965simple.a - -ifeq ($(CONFIG_NAME), linux-cell) -CELL_LIB = $(TOP)/src/gallium/drivers/cell/ppu/libcell.a -CELL_LIB_SPU = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a -endif - -ifeq ($(CONFIG_NAME), linux-llvm) -LLVM_LIB = $(TOP)/src/gallium/auxiliary/llvm/libgallivm.a -endif - .SUFFIXES : .cpp .c.o: @@ -36,33 +23,14 @@ endif $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ -# Figure out what to make here -default: - @if [ "${DRIVER_DIRS}" = "dri" ] ; then \ - $(MAKE) linux-solo ; \ - elif [ "${DRIVER_DIRS}" = "osmesa" ] ; then \ - $(MAKE) osmesa-only ; \ - elif [ "$(DRIVER_DIRS)" = "beos" ]; then \ - $(MAKE) beos ; \ - elif [ "$(DRIVER_DIRS)" = "directfb" ]; then \ - $(MAKE) directfb ; \ - elif [ "$(DRIVER_DIRS)" = "fbdev osmesa" ]; then \ - $(MAKE) fbdev ; $(MAKE) osmesa-only ; \ - else \ - $(MAKE) stand-alone ; \ - fi - +default: depend subdirs libmesa.a -###################################################################### -# BeOS driver target - -beos: depend subdirs libmesa.a - cd drivers/beos; $(MAKE) +ifneq ($(DRIVER_DIRS),dri) +default: libglapi.a +endif ###################################################################### -# Linux DRI drivers - # Make archive of core object files libmesa.a: $(SOLO_OBJECTS) @ $(TOP)/bin/mklib -o mesa -static $(SOLO_OBJECTS); @@ -70,32 +38,8 @@ libmesa.a: $(SOLO_OBJECTS) mimeset -f "$@" ; \ fi -linux-solo: depend subdirs libmesa.a - cd $(TOP)/src/gallium/winsys/dri ; $(MAKE) - - -##################################################################### -# Stand-alone Mesa libGL, no built-in drivers (DirectFB) - -libgl-core: $(CORE_OBJECTS) - @ $(TOP)/bin/mklib -o $(GL_LIB) \ - -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ - -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) $(CORE_OBJECTS) \ - $(GL_LIB_DEPS) - -directfb: depend subdirs libgl-core - cd drivers/directfb ; $(MAKE) - - -##################################################################### -# fbdev Mesa driver (libGL.so) - -fbdev: $(CORE_OBJECTS) $(FBDEV_DRIVER_OBJECTS) $(COMMON_DRIVER_OBJECTS) - @ $(TOP)/bin/mklib -o $(GL_LIB) \ - -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ - -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ - $(CORE_OBJECTS) $(FBDEV_DRIVER_OBJECTS) \ - $(COMMON_DRIVER_OBJECTS) $(GL_LIB_DEPS) +libglapi.a: $(GLAPI_OBJECTS) + @ $(TOP)/bin/mklib -o glapi -static $(GLAPI_OBJECTS) ###################################################################### @@ -176,9 +120,6 @@ install: default @if [ -e $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) ]; then \ $(INSTALL) $(TOP)/$(LIB_DIR)/libOSMesa* $(INSTALL_DIR)/$(LIB_DIR); \ fi - @if [ "${DRIVER_DIRS}" = "dri" ] ; then \ - cd $(TOP)/gallium/winsys/dri ; $(MAKE) install ; \ - fi ## NOT INSTALLED YET: ## $(INSTALL) -d $(INSTALL_DIR)/include/GLES @@ -192,9 +133,8 @@ tags: clean: -rm -f */*.o -rm -f */*/*.o - -rm -f depend depend.bak libmesa.a + -rm -f depend depend.bak libmesa.a libglapi.a -rm -f drivers/*/*.o - (cd drivers/dri && $(MAKE) clean) (cd x86 && $(MAKE) clean) (cd x86-64 && $(MAKE) clean) diff --git a/src/mesa/sources b/src/mesa/sources index f83d247a1e..0d185fd5f3 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -156,51 +156,6 @@ VF_SOURCES = \ vf/vf_generic.c \ vf/vf_sse.c - -DRAW_SOURCES = \ - $(TOP)/src/gallium/auxiliary/draw/draw_clip.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vs_exec.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vs_sse.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vs_llvm.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_context.c\ - $(TOP)/src/gallium/auxiliary/draw/draw_cull.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_debug.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_flatshade.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_offset.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_prim.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_stipple.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_twoside.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_unfilled.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_validate.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vbuf.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vertex.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vertex_cache.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vertex_fetch.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vertex_shader.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vf.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vf_generic.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_vf_sse.c \ - $(TOP)/src/gallium/auxiliary/draw/draw_wide_prims.c - -TGSIEXEC_SOURCES = \ - $(TOP)/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c \ - $(TOP)/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c - -TGSIUTIL_SOURCES = \ - $(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_build.c \ - $(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_dump.c \ - $(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_parse.c \ - $(TOP)/src/gallium/auxiliary/tgsi/util/tgsi_util.c - -STATECACHE_SOURCES = \ - $(TOP)/src/gallium/auxiliary/cso_cache/cso_hash.c \ - $(TOP)/src/gallium/auxiliary/cso_cache/cso_cache.c - -PIPEUTIL_SOURCES = \ - $(TOP)/src/gallium/auxiliary/util/p_debug.c \ - $(TOP)/src/gallium/auxiliary/util/p_tile.c \ - $(TOP)/src/gallium/auxiliary/util/p_util.c - STATETRACKER_SOURCES = \ state_tracker/st_atom.c \ state_tracker/st_atom_blend.c \ @@ -229,7 +184,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_readpixels.c \ state_tracker/st_cb_strings.c \ state_tracker/st_cb_texture.c \ - state_tracker/st_cache.c \ + state_tracker/st_cache.c \ state_tracker/st_context.c \ state_tracker/st_debug.c \ state_tracker/st_draw.c \ @@ -333,15 +288,6 @@ SPARC_API = \ __COMMON_DRIVER_SOURCES = \ drivers/common/driverfuncs.c -X11_DRIVER_SOURCES = \ - $(TOP)/src/gallium/winsys/xlib/glxapi.c \ - $(TOP)/src/gallium/winsys/xlib/fakeglx.c \ - $(TOP)/src/gallium/winsys/xlib/xfonts.c \ - $(TOP)/src/gallium/winsys/xlib/xm_api.c \ - $(TOP)/src/gallium/winsys/xlib/xm_winsys.c \ - $(TOP)/src/gallium/winsys/xlib/xm_winsys_aub.c \ - $(TOP)/src/gallium/winsys/xlib/brw_aub.c - OSMESA_DRIVER_SOURCES = \ drivers/osmesa/osmesa.c -- cgit v1.2.3 From 39fe5851a57a7218eafce7dab971738bdb780166 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 18 Feb 2008 15:07:17 +0900 Subject: Actually use GALLIUM_DRIVER_DIRS. --- src/gallium/drivers/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/Makefile b/src/gallium/drivers/Makefile index 8dcfb18a16..6161cb6ff8 100644 --- a/src/gallium/drivers/Makefile +++ b/src/gallium/drivers/Makefile @@ -2,7 +2,7 @@ TOP = ../../.. include $(TOP)/configs/current -SUBDIRS = softpipe i915simple i965simple failover +SUBDIRS = $(GALLIUM_DRIVER_DIRS) default: subdirs -- cgit v1.2.3 From 8450b14676a2f5c6423b16bc4bc7a1ec5a6a987e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 18 Feb 2008 15:37:52 +0900 Subject: Cleanup depend files. --- src/gallium/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/Makefile b/src/gallium/Makefile index 89e068a449..568747e157 100644 --- a/src/gallium/Makefile +++ b/src/gallium/Makefile @@ -18,3 +18,4 @@ subdirs: clean: rm -f `find . -name \*.[oa]` + rm -f `find . -name depend` -- cgit v1.2.3 From 33ceb6716a2166db75659fa66d85fb4cfb9633c7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 18 Feb 2008 10:52:44 +0000 Subject: Update scons build for new code layout. --- src/SConscript | 7 ++ src/gallium/SConscript | 23 +++- src/gallium/auxiliary/cso_cache/SConscript | 10 ++ src/gallium/auxiliary/draw/SConscript | 31 +++++ src/gallium/auxiliary/pipebuffer/SConscript | 14 +++ src/gallium/auxiliary/tgsi/SConscript | 14 +++ src/gallium/auxiliary/util/SConscript | 11 ++ src/gallium/drivers/failover/SConscript | 13 +++ src/gallium/winsys/SConscript | 10 ++ src/gallium/winsys/dri/SConscript | 51 +++++++++ src/gallium/winsys/dri/intel/SConscript | 14 +-- src/gallium/winsys/xlib/SConscript | 28 +++++ src/mesa/SConscript | 170 ++-------------------------- src/mesa/drivers/dri/SConscript | 48 -------- 14 files changed, 224 insertions(+), 220 deletions(-) create mode 100644 src/SConscript create mode 100644 src/gallium/auxiliary/cso_cache/SConscript create mode 100644 src/gallium/auxiliary/draw/SConscript create mode 100644 src/gallium/auxiliary/pipebuffer/SConscript create mode 100644 src/gallium/auxiliary/tgsi/SConscript create mode 100644 src/gallium/auxiliary/util/SConscript create mode 100644 src/gallium/drivers/failover/SConscript create mode 100644 src/gallium/winsys/SConscript create mode 100644 src/gallium/winsys/dri/SConscript create mode 100644 src/gallium/winsys/xlib/SConscript delete mode 100644 src/mesa/drivers/dri/SConscript (limited to 'src') diff --git a/src/SConscript b/src/SConscript new file mode 100644 index 0000000000..5b09943894 --- /dev/null +++ b/src/SConscript @@ -0,0 +1,7 @@ +Import('*') + +SConscript([ + 'gallium/SConscript', + 'mesa/SConscript', + 'gallium/winsys/SConscript', +]) diff --git a/src/gallium/SConscript b/src/gallium/SConscript index d9c20e0100..a835f6d661 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -1,9 +1,24 @@ Import('*') -#env = env.Clone() +env = env.Clone() + +auxiliaries = [] + +Export('auxiliaries') + SConscript([ - 'softpipe/SConscript', - 'i915simple/SConscript', - 'i965simple/SConscript', + # NOTE: order matters! + 'auxiliary/util/SConscript', + 'auxiliary/tgsi/SConscript', + 'auxiliary/cso_cache/SConscript', + 'auxiliary/draw/SConscript', + #'auxiliary/llvm/SConscript', + 'auxiliary/pipebuffer/SConscript', + + 'drivers/softpipe/SConscript', + 'drivers/i915simple/SConscript', + 'drivers/i965simple/SConscript', + 'drivers/failover/SConscript', + #'drivers/cell/SConscript', ]) diff --git a/src/gallium/auxiliary/cso_cache/SConscript b/src/gallium/auxiliary/cso_cache/SConscript new file mode 100644 index 0000000000..9751881613 --- /dev/null +++ b/src/gallium/auxiliary/cso_cache/SConscript @@ -0,0 +1,10 @@ +Import('*') + +cso_cache = env.ConvenienceLibrary( + target = 'cso_cache', + source = [ + 'cso_cache.c', + 'cso_hash.c', + ]) + +auxiliaries.insert(0, cso_cache) diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript new file mode 100644 index 0000000000..8e3a8caa74 --- /dev/null +++ b/src/gallium/auxiliary/draw/SConscript @@ -0,0 +1,31 @@ +Import('*') + +draw = env.ConvenienceLibrary( + target = 'draw', + source = [ + 'draw_clip.c', + 'draw_vs_exec.c', + 'draw_vs_sse.c', + 'draw_vs_llvm.c', + 'draw_context.c', + 'draw_cull.c', + 'draw_debug.c', + 'draw_flatshade.c', + 'draw_offset.c', + 'draw_prim.c', + 'draw_stipple.c', + 'draw_twoside.c', + 'draw_unfilled.c', + 'draw_validate.c', + 'draw_vbuf.c', + 'draw_vertex.c', + 'draw_vertex_cache.c', + 'draw_vertex_fetch.c', + 'draw_vertex_shader.c', + 'draw_vf.c', + 'draw_vf_generic.c', + 'draw_vf_sse.c', + 'draw_wide_prims.c', + ]) + +auxiliaries.insert(0, draw) diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript new file mode 100644 index 0000000000..3d41fd84a6 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/SConscript @@ -0,0 +1,14 @@ +Import('*') + +pipebuffer = env.ConvenienceLibrary( + target = 'pipebuffer', + source = [ + 'pb_buffer_fenced.c', + 'pb_buffer_malloc.c', + 'pb_bufmgr_fenced.c', + 'pb_bufmgr_mm.c', + 'pb_bufmgr_pool.c', + 'pb_winsys.c', + ]) + +auxiliaries.insert(0, pipebuffer) diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript new file mode 100644 index 0000000000..8464bfe944 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -0,0 +1,14 @@ +Import('*') + +tgsi = env.ConvenienceLibrary( + target = 'tgsi', + source = [ + 'exec/tgsi_exec.c', + 'exec/tgsi_sse2.c', + 'util/tgsi_build.c', + 'util/tgsi_dump.c', + 'util/tgsi_parse.c', + 'util/tgsi_util.c', + ]) + +auxiliaries.insert(0, tgsi) diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript new file mode 100644 index 0000000000..b126cf44d6 --- /dev/null +++ b/src/gallium/auxiliary/util/SConscript @@ -0,0 +1,11 @@ +Import('*') + +util = env.ConvenienceLibrary( + target = 'util', + source = [ + 'p_debug.c', + 'p_tile.c', + 'p_util.c', + ]) + +auxiliaries.insert(0, util) diff --git a/src/gallium/drivers/failover/SConscript b/src/gallium/drivers/failover/SConscript new file mode 100644 index 0000000000..f8e9b1b491 --- /dev/null +++ b/src/gallium/drivers/failover/SConscript @@ -0,0 +1,13 @@ +Import('*') + +env = env.Clone() + +failover = env.ConvenienceLibrary( + target = 'failover', + source = [ + 'fo_state.c', + 'fo_state_emit.c', + 'fo_context.c', + ]) + +Export('failover') diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript new file mode 100644 index 0000000000..32215d8d58 --- /dev/null +++ b/src/gallium/winsys/SConscript @@ -0,0 +1,10 @@ +Import('*') + +if dri: + SConscript([ + 'dri/SConscript', + ]) +else: + SConscript([ + 'xlib/SConscript', + ]) diff --git a/src/gallium/winsys/dri/SConscript b/src/gallium/winsys/dri/SConscript new file mode 100644 index 0000000000..8c56ce917c --- /dev/null +++ b/src/gallium/winsys/dri/SConscript @@ -0,0 +1,51 @@ +Import('*') + +drienv = env.Clone() + +drienv.Replace(CPPPATH = [ + '#src/mesa/drivers/dri/common', + '#include', + '#include/GL/internal', + '#src/gallium/include', + '#src/gallium/auxiliary', + '#src/gallium/drivers', + '#src/mesa', + '#src/mesa/main', + '#src/mesa/glapi', + '#src/mesa/math', + '#src/mesa/transform', + '#src/mesa/shader', + '#src/mesa/swrast', + '#src/mesa/swrast_setup', + '#src/egl/main', + '#src/egl/drivers/dri', +]) + +drienv.ParseConfig('pkg-config --cflags --libs libdrm') + +COMMON_GALLIUM_SOURCES = [ + '#src/mesa/drivers/dri/common/utils.c', + '#src/mesa/drivers/dri/common/vblank.c', + '#src/mesa/drivers/dri/common/dri_util.c', + '#src/mesa/drivers/dri/common/xmlconfig.c', +] + +COMMON_BM_SOURCES = [ + '#src/mesa/drivers/dri/common/dri_bufmgr.c', + '#src/mesa/drivers/dri/common/dri_drmpool.c', +] + +Export([ + 'drienv', + 'COMMON_GALLIUM_SOURCES', + 'COMMON_BM_SOURCES', +]) + +# TODO: Installation +#install: $(LIBNAME) +# $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) +# $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) + +SConscript([ + 'intel/SConscript', +]) diff --git a/src/gallium/winsys/dri/intel/SConscript b/src/gallium/winsys/dri/intel/SConscript index a7cc10450e..525ba580e8 100644 --- a/src/gallium/winsys/dri/intel/SConscript +++ b/src/gallium/winsys/dri/intel/SConscript @@ -9,11 +9,6 @@ env.Append(CPPPATH = [ #MINIGLX_SOURCES = server/intel_dri.c -pipe_drivers = [ - softpipe, - i915simple -] - DRIVER_SOURCES = [ 'intel_winsys_pipe.c', 'intel_winsys_softpipe.c', @@ -31,11 +26,14 @@ sources = \ COMMON_BM_SOURCES + \ DRIVER_SOURCES -# DRIVER_DEFINES = -I../intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ -# && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") +drivers = [ + softpipe, + i915simple +] +# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions env.SharedLibrary( target ='i915tex_dri.so', source = sources, - LIBS = pipe_drivers + env['LIBS'], + LIBS = mesa + drivers + auxiliaries + env['LIBS'], ) \ No newline at end of file diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript new file mode 100644 index 0000000000..f8aa5ef945 --- /dev/null +++ b/src/gallium/winsys/xlib/SConscript @@ -0,0 +1,28 @@ +####################################################################### +# SConscript for xlib winsys + +Import('*') + + +sources = [ + 'glxapi.c', + 'fakeglx.c', + 'xfonts.c', + 'xm_api.c', + 'xm_winsys.c', + 'xm_winsys_aub.c', + 'brw_aub.c', +] + +drivers = [ + softpipe, + i915simple, + i965simple, +] + +# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions +env.SharedLibrary( + target ='GL', + source = sources, + LIBS = glapi + mesa + drivers + auxiliaries + env['LIBS'], +) diff --git a/src/mesa/SConscript b/src/mesa/SConscript index faf8c84872..a828133580 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -1,7 +1,5 @@ ####################################################################### -# SConscript for mesa -# -# TODO: Split this into per-module SConscripts +# SConscript for Mesa Import('*') @@ -116,53 +114,6 @@ VF_SOURCES = [ 'vf/vf_sse.c', ] -DRAW_SOURCES = [ - 'pipe/draw/draw_clip.c', - 'pipe/draw/draw_context.c', - 'pipe/draw/draw_cull.c', - 'pipe/draw/draw_debug.c', - 'pipe/draw/draw_flatshade.c', - 'pipe/draw/draw_offset.c', - 'pipe/draw/draw_prim.c', - 'pipe/draw/draw_stipple.c', - 'pipe/draw/draw_twoside.c', - 'pipe/draw/draw_unfilled.c', - 'pipe/draw/draw_validate.c', - 'pipe/draw/draw_vbuf.c', - 'pipe/draw/draw_vertex.c', - 'pipe/draw/draw_vertex_cache.c', - 'pipe/draw/draw_vertex_fetch.c', - 'pipe/draw/draw_vertex_shader.c', - 'pipe/draw/draw_vertex_shader_llvm.c', - 'pipe/draw/draw_vf.c', - 'pipe/draw/draw_vf_generic.c', - 'pipe/draw/draw_vf_sse.c', - 'pipe/draw/draw_wide_prims.c', -] - -TGSIEXEC_SOURCES = [ - 'pipe/tgsi/exec/tgsi_exec.c', - 'pipe/tgsi/exec/tgsi_sse2.c', -] - -TGSIUTIL_SOURCES = [ - 'pipe/tgsi/util/tgsi_build.c', - 'pipe/tgsi/util/tgsi_dump.c', - 'pipe/tgsi/util/tgsi_parse.c', - 'pipe/tgsi/util/tgsi_util.c', -] - -STATECACHE_SOURCES = [ - 'pipe/cso_cache/cso_hash.c', - 'pipe/cso_cache/cso_cache.c', -] - -PIPEUTIL_SOURCES = [ - 'pipe/util/p_debug.c', - 'pipe/util/p_tile.c', - 'pipe/util/p_util.c', -] - STATETRACKER_SOURCES = [ 'state_tracker/st_atom.c', 'state_tracker/st_atom_blend.c', @@ -311,126 +262,25 @@ else: ASM_SOURCES = [] API_SOURCES = [] - -####################################################################### -# Driver sources - - -X11_DRIVER_SOURCES = [ - 'pipe/xlib/glxapi.c', - 'pipe/xlib/fakeglx.c', - 'pipe/xlib/xfonts.c', - 'pipe/xlib/xm_api.c', - 'pipe/xlib/xm_winsys.c', - 'pipe/xlib/xm_winsys_aub.c', - 'pipe/xlib/brw_aub.c', -] - -OSMESA_DRIVER_SOURCES = [ - 'drivers/osmesa/osmesa.c', -] - -GLIDE_DRIVER_SOURCES = [ - 'drivers/glide/fxapi.c', - 'drivers/glide/fxdd.c', - 'drivers/glide/fxddspan.c', - 'drivers/glide/fxddtex.c', - 'drivers/glide/fxsetup.c', - 'drivers/glide/fxtexman.c', - 'drivers/glide/fxtris.c', - 'drivers/glide/fxvb.c', - 'drivers/glide/fxglidew.c', - 'drivers/glide/fxg.c', -] - -SVGA_DRIVER_SOURCES = [ - 'drivers/svga/svgamesa.c', - 'drivers/svga/svgamesa8.c', - 'drivers/svga/svgamesa15.c', - 'drivers/svga/svgamesa16.c', - 'drivers/svga/svgamesa24.c', - 'drivers/svga/svgamesa32.c', -] - -FBDEV_DRIVER_SOURCES = [ - 'drivers/fbdev/glfbdev.c', -] - - -### All the core C sources - SOLO_SOURCES = \ MAIN_SOURCES + \ MATH_SOURCES + \ VBO_SOURCES + \ VF_SOURCES + \ - DRAW_SOURCES + \ - TGSIEXEC_SOURCES + \ - TGSIUTIL_SOURCES + \ - PIPEUTIL_SOURCES + \ - STATECACHE_SOURCES + \ STATETRACKER_SOURCES + \ SHADER_SOURCES + \ ASM_SOURCES + \ SLANG_SOURCES -CORE_SOURCES = \ - GLAPI_SOURCES + API_SOURCES + \ - SOLO_SOURCES - -ALL_SOURCES = \ - GLAPI_SOURCES + API_SOURCES + \ - SOLO_SOURCES + \ - ASM_SOURCES + \ - X11_DRIVER_SOURCES + \ - FBDEV_DRIVER_SOURCES + \ - OSMESA_DRIVER_SOURCES - - -###################################################################### -# Gallium sources - -SConscript([ - 'pipe/SConscript', -]) - - -###################################################################### -# libGL +mesa = env.ConvenienceLibrary( + target = 'mesa', + source = SOLO_SOURCES, +) +Export('mesa') if not dri: - STAND_ALONE_DRIVER_SOURCES = \ - CORE_SOURCES + \ - X11_DRIVER_SOURCES - - Import( - 'softpipe', - 'i915simple', - 'i965simple' + glapi = env.ConvenienceLibrary( + target = 'glapi', + source = GLAPI_SOURCES + API_SOURCES, ) - - pipe_drivers = [ - softpipe, - i965simple - ] - - env.SharedLibrary( - target ='GL', - source = STAND_ALONE_DRIVER_SOURCES, - LIBS = [softpipe, i965simple] + env['LIBS'], - ) - - -###################################################################### -# Driver sources - -if dri: - mesa = env.ConvenienceLibrary( - target = 'mesa', - source = SOLO_SOURCES, - ) - env.Prepend(LIBS = [mesa]) - - SConscript([ - 'drivers/dri/SConscript', - ]) + Export('glapi') diff --git a/src/mesa/drivers/dri/SConscript b/src/mesa/drivers/dri/SConscript deleted file mode 100644 index d32bd08669..0000000000 --- a/src/mesa/drivers/dri/SConscript +++ /dev/null @@ -1,48 +0,0 @@ -Import('*') - -drienv = env.Clone() - -drienv.Replace(CPPPATH = [ - '#src/mesa/drivers/dri/common', - '#include', - '#include/GL/internal', - '#src/mesa', - '#src/mesa/main', - '#src/mesa/glapi', - '#src/mesa/math', - '#src/mesa/transform', - '#src/mesa/shader', - '#src/mesa/swrast', - '#src/mesa/swrast_setup', - '#src/egl/main', - '#src/egl/drivers/dri', -]) - -drienv.ParseConfig('pkg-config --cflags --libs libdrm') - -COMMON_GALLIUM_SOURCES = [ - '../common/utils.c', - '../common/vblank.c', - '../common/dri_util.c', - '../common/xmlconfig.c', -] - -COMMON_BM_SOURCES = [ - '../common/dri_bufmgr.c', - '../common/dri_drmpool.c', -] - -Export([ - 'drienv', - 'COMMON_GALLIUM_SOURCES', - 'COMMON_BM_SOURCES', -]) - -# TODO: Installation -#install: $(LIBNAME) -# $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) -# $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) - -SConscript([ - 'intel_winsys/SConscript', -]) -- cgit v1.2.3 From 687a8b96ef13658bbe779d0011ce1144844f1972 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 18 Feb 2008 20:02:42 +0900 Subject: Standardize on using the pipe/ include prefix. --- src/gallium/Makefile.template | 1 - src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 10 +++++----- src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c | 4 ++-- src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c | 8 ++++---- src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c | 10 +++++----- src/gallium/drivers/i915simple/i915_state_immediate.c | 2 +- src/gallium/drivers/i915simple/i915_state_inlines.h | 4 ++-- src/gallium/drivers/softpipe/sp_state_surface.c | 2 +- 8 files changed, 20 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template index 6ad58c205c..6698212e77 100644 --- a/src/gallium/Makefile.template +++ b/src/gallium/Makefile.template @@ -16,7 +16,6 @@ OBJECTS = $(C_SOURCES:.c=.o) \ INCLUDES = \ -I. \ -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/include/pipe \ -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers \ -I$(TOP)/src/mesa \ diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index f4fc3f6d71..bc85c4b19f 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -36,11 +36,11 @@ #include "linked_list.h" -#include "p_compiler.h" -#include "p_debug.h" -#include "p_winsys.h" -#include "p_thread.h" -#include "p_util.h" +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_winsys.h" +#include "pipe/p_thread.h" +#include "pipe/p_util.h" #include "pb_buffer.h" #include "pb_buffer_fenced.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c index c535d3276c..bffca5b244 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c @@ -34,8 +34,8 @@ */ -#include "p_debug.h" -#include "p_util.h" +#include "pipe/p_debug.h" +#include "pipe/p_util.h" #include "pb_buffer.h" #include "pb_buffer_fenced.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index 8b1b51c0e2..969aab51b5 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -36,10 +36,10 @@ #include "linked_list.h" -#include "p_defines.h" -#include "p_debug.h" -#include "p_thread.h" -#include "p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_debug.h" +#include "pipe/p_thread.h" +#include "pipe/p_util.h" #include "pb_buffer.h" #include "pb_bufmgr.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c index 04477a865a..beb145b7cb 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c @@ -37,11 +37,11 @@ #include "linked_list.h" -#include "p_compiler.h" -#include "p_debug.h" -#include "p_thread.h" -#include "p_defines.h" -#include "p_util.h" +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_thread.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "pb_buffer.h" #include "pb_bufmgr.h" diff --git a/src/gallium/drivers/i915simple/i915_state_immediate.c b/src/gallium/drivers/i915simple/i915_state_immediate.c index 07031fc6c5..dfbbcab624 100644 --- a/src/gallium/drivers/i915simple/i915_state_immediate.c +++ b/src/gallium/drivers/i915simple/i915_state_immediate.c @@ -33,7 +33,7 @@ #include "i915_context.h" #include "i915_state.h" #include "i915_reg.h" -#include "p_util.h" +#include "pipe/p_util.h" /* All state expressable with the LOAD_STATE_IMMEDIATE_1 packet. diff --git a/src/gallium/drivers/i915simple/i915_state_inlines.h b/src/gallium/drivers/i915simple/i915_state_inlines.h index 0934ac79a4..378de8f9c4 100644 --- a/src/gallium/drivers/i915simple/i915_state_inlines.h +++ b/src/gallium/drivers/i915simple/i915_state_inlines.h @@ -28,8 +28,8 @@ #ifndef I915_STATE_INLINES_H #define I915_STATE_INLINES_H -#include "p_compiler.h" -#include "p_defines.h" +#include "pipe/p_compiler.h" +#include "pipe/p_defines.h" #include "i915_reg.h" diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c index e2c6893e9f..124b18b708 100644 --- a/src/gallium/drivers/softpipe/sp_state_surface.c +++ b/src/gallium/drivers/softpipe/sp_state_surface.c @@ -27,7 +27,7 @@ /* Authors: Keith Whitwell */ -#include "p_inlines.h" +#include "pipe/p_inlines.h" #include "sp_context.h" #include "sp_state.h" -- cgit v1.2.3 From 3f3b09d6d86cfd277c5837d15466ee703897aa3d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 18 Feb 2008 20:05:06 +0900 Subject: Rename llvm -> gallivm. Following the directory == library name policy simplifies the build system. --- src/gallium/auxiliary/gallivm/Makefile | 85 ++ src/gallium/auxiliary/gallivm/gallivm.cpp | 327 ++++++ src/gallium/auxiliary/gallivm/gallivm.h | 103 ++ src/gallium/auxiliary/gallivm/gallivm_builtins.cpp | 567 +++++++++ src/gallium/auxiliary/gallivm/gallivm_cpu.cpp | 202 ++++ src/gallium/auxiliary/gallivm/gallivm_p.h | 110 ++ src/gallium/auxiliary/gallivm/instructions.cpp | 889 ++++++++++++++ src/gallium/auxiliary/gallivm/instructions.h | 152 +++ src/gallium/auxiliary/gallivm/instructionssoa.cpp | 121 ++ src/gallium/auxiliary/gallivm/instructionssoa.h | 74 ++ src/gallium/auxiliary/gallivm/llvm_builtins.c | 115 ++ src/gallium/auxiliary/gallivm/loweringpass.cpp | 17 + src/gallium/auxiliary/gallivm/loweringpass.h | 15 + src/gallium/auxiliary/gallivm/storage.cpp | 364 ++++++ src/gallium/auxiliary/gallivm/storage.h | 133 +++ src/gallium/auxiliary/gallivm/storagesoa.cpp | 389 +++++++ src/gallium/auxiliary/gallivm/storagesoa.h | 111 ++ src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 1221 ++++++++++++++++++++ src/gallium/auxiliary/gallivm/tgsitollvm.h | 20 + src/gallium/auxiliary/llvm/Makefile | 85 -- src/gallium/auxiliary/llvm/gallivm.cpp | 327 ------ src/gallium/auxiliary/llvm/gallivm.h | 103 -- src/gallium/auxiliary/llvm/gallivm_builtins.cpp | 567 --------- src/gallium/auxiliary/llvm/gallivm_cpu.cpp | 202 ---- src/gallium/auxiliary/llvm/gallivm_p.h | 110 -- src/gallium/auxiliary/llvm/instructions.cpp | 889 -------------- src/gallium/auxiliary/llvm/instructions.h | 152 --- src/gallium/auxiliary/llvm/instructionssoa.cpp | 121 -- src/gallium/auxiliary/llvm/instructionssoa.h | 74 -- src/gallium/auxiliary/llvm/llvm_builtins.c | 115 -- src/gallium/auxiliary/llvm/loweringpass.cpp | 17 - src/gallium/auxiliary/llvm/loweringpass.h | 15 - src/gallium/auxiliary/llvm/storage.cpp | 364 ------ src/gallium/auxiliary/llvm/storage.h | 133 --- src/gallium/auxiliary/llvm/storagesoa.cpp | 389 ------- src/gallium/auxiliary/llvm/storagesoa.h | 111 -- src/gallium/auxiliary/llvm/tgsitollvm.cpp | 1221 -------------------- src/gallium/auxiliary/llvm/tgsitollvm.h | 20 - 38 files changed, 5015 insertions(+), 5015 deletions(-) create mode 100644 src/gallium/auxiliary/gallivm/Makefile create mode 100644 src/gallium/auxiliary/gallivm/gallivm.cpp create mode 100644 src/gallium/auxiliary/gallivm/gallivm.h create mode 100644 src/gallium/auxiliary/gallivm/gallivm_builtins.cpp create mode 100644 src/gallium/auxiliary/gallivm/gallivm_cpu.cpp create mode 100644 src/gallium/auxiliary/gallivm/gallivm_p.h create mode 100644 src/gallium/auxiliary/gallivm/instructions.cpp create mode 100644 src/gallium/auxiliary/gallivm/instructions.h create mode 100644 src/gallium/auxiliary/gallivm/instructionssoa.cpp create mode 100644 src/gallium/auxiliary/gallivm/instructionssoa.h create mode 100644 src/gallium/auxiliary/gallivm/llvm_builtins.c create mode 100644 src/gallium/auxiliary/gallivm/loweringpass.cpp create mode 100644 src/gallium/auxiliary/gallivm/loweringpass.h create mode 100644 src/gallium/auxiliary/gallivm/storage.cpp create mode 100644 src/gallium/auxiliary/gallivm/storage.h create mode 100644 src/gallium/auxiliary/gallivm/storagesoa.cpp create mode 100644 src/gallium/auxiliary/gallivm/storagesoa.h create mode 100644 src/gallium/auxiliary/gallivm/tgsitollvm.cpp create mode 100644 src/gallium/auxiliary/gallivm/tgsitollvm.h delete mode 100644 src/gallium/auxiliary/llvm/Makefile delete mode 100644 src/gallium/auxiliary/llvm/gallivm.cpp delete mode 100644 src/gallium/auxiliary/llvm/gallivm.h delete mode 100644 src/gallium/auxiliary/llvm/gallivm_builtins.cpp delete mode 100644 src/gallium/auxiliary/llvm/gallivm_cpu.cpp delete mode 100644 src/gallium/auxiliary/llvm/gallivm_p.h delete mode 100644 src/gallium/auxiliary/llvm/instructions.cpp delete mode 100644 src/gallium/auxiliary/llvm/instructions.h delete mode 100644 src/gallium/auxiliary/llvm/instructionssoa.cpp delete mode 100644 src/gallium/auxiliary/llvm/instructionssoa.h delete mode 100644 src/gallium/auxiliary/llvm/llvm_builtins.c delete mode 100644 src/gallium/auxiliary/llvm/loweringpass.cpp delete mode 100644 src/gallium/auxiliary/llvm/loweringpass.h delete mode 100644 src/gallium/auxiliary/llvm/storage.cpp delete mode 100644 src/gallium/auxiliary/llvm/storage.h delete mode 100644 src/gallium/auxiliary/llvm/storagesoa.cpp delete mode 100644 src/gallium/auxiliary/llvm/storagesoa.h delete mode 100644 src/gallium/auxiliary/llvm/tgsitollvm.cpp delete mode 100644 src/gallium/auxiliary/llvm/tgsitollvm.h (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/Makefile b/src/gallium/auxiliary/gallivm/Makefile new file mode 100644 index 0000000000..39fac6ea4a --- /dev/null +++ b/src/gallium/auxiliary/gallivm/Makefile @@ -0,0 +1,85 @@ +# -*-makefile-*- +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = gallivm + + +GALLIVM_SOURCES = \ + gallivm.cpp \ + gallivm_cpu.cpp \ + instructions.cpp \ + loweringpass.cpp \ + tgsitollvm.cpp \ + storage.cpp \ + storagesoa.cpp \ + instructionssoa.cpp + +INC_SOURCES = gallivm_builtins.cpp + +CPP_SOURCES = \ + $(GALLIVM_SOURCES) + +C_SOURCES = +ASM_SOURCES = + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(CPP_SOURCES:.cpp=.o) \ + $(ASM_SOURCES:.S=.o) + +### Include directories +INCLUDES = \ + -I. \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/auxiliary \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/include + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDES) $(LLVM_CXXFLAGS) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +##### TARGETS ##### + +default:: depend symlinks $(LIBNAME) + + +$(LIBNAME): $(OBJECTS) Makefile + $(TOP)/bin/mklib -o $@ -static $(OBJECTS) + + +depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ + $(ASM_SOURCES) $(INC_SOURCES) 2> /dev/null + + +gallivm_builtins.cpp: llvm_builtins.c + clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean: + -rm -f *.o */*.o *~ *.so *~ server/*.o + -rm -f depend depend.bak + -rm -f gallivm_builtins.cpp + +symlinks: + + +include depend diff --git a/src/gallium/auxiliary/gallivm/gallivm.cpp b/src/gallium/auxiliary/gallivm/gallivm.cpp new file mode 100644 index 0000000000..d14bb3b99a --- /dev/null +++ b/src/gallium/auxiliary/gallivm/gallivm.cpp @@ -0,0 +1,327 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "instructions.h" +#include "loweringpass.h" +#include "storage.h" +#include "tgsitollvm.h" + +#include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_dump.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +static int GLOBAL_ID = 0; + +using namespace llvm; + +static inline +void AddStandardCompilePasses(PassManager &PM) +{ + PM.add(new LoweringPass()); + PM.add(createVerifierPass()); // Verify that input is correct + + PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp + + //PM.add(createStripSymbolsPass(true)); + + PM.add(createRaiseAllocationsPass()); // call %malloc -> malloc inst + PM.add(createCFGSimplificationPass()); // Clean up disgusting code + PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas + PM.add(createGlobalOptimizerPass()); // Optimize out global vars + PM.add(createGlobalDCEPass()); // Remove unused fns and globs + PM.add(createIPConstantPropagationPass());// IP Constant Propagation + PM.add(createDeadArgEliminationPass()); // Dead argument elimination + PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE + PM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE + + PM.add(createPruneEHPass()); // Remove dead EH info + + PM.add(createFunctionInliningPass()); // Inline small functions + PM.add(createArgumentPromotionPass()); // Scalarize uninlined fn args + + PM.add(createTailDuplicationPass()); // Simplify cfg by copying code + PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl. + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas + PM.add(createInstructionCombiningPass()); // Combine silly seq's + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createTailCallEliminationPass()); // Eliminate tail calls + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createReassociatePass()); // Reassociate expressions + PM.add(createLoopRotatePass()); + PM.add(createLICMPass()); // Hoist loop invariants + PM.add(createLoopUnswitchPass()); // Unswitch loops. + PM.add(createLoopIndexSplitPass()); // Index split loops. + PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc + PM.add(createIndVarSimplifyPass()); // Canonicalize indvars + PM.add(createLoopUnrollPass()); // Unroll small loops + PM.add(createInstructionCombiningPass()); // Clean up after the unroller + PM.add(createGVNPass()); // Remove redundancies + PM.add(createSCCPPass()); // Constant prop with SCCP + + // Run instcombine after redundancy elimination to exploit opportunities + // opened up by them. + PM.add(createInstructionCombiningPass()); + PM.add(createCondPropagationPass()); // Propagate conditionals + + PM.add(createDeadStoreEliminationPass()); // Delete dead stores + PM.add(createAggressiveDCEPass()); // SSA based 'Aggressive DCE' + PM.add(createCFGSimplificationPass()); // Merge & remove BBs + PM.add(createSimplifyLibCallsPass()); // Library Call Optimizations + PM.add(createDeadTypeEliminationPass()); // Eliminate dead types + PM.add(createConstantMergePass()); // Merge dup global constants +} + +void gallivm_prog_delete(struct gallivm_prog *prog) +{ + delete prog->module; + prog->module = 0; + prog->function = 0; + free(prog); +} + +static inline void +constant_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan) +{ + unsigned i; + + for (i = 0; i < QUAD_SIZE; ++i) { + inputs[i][attrib][chan] = coefs[attrib].a0[chan]; + } +} + +static inline void +linear_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = inputs[i][0][0]; + const float y = inputs[i][0][1]; + + inputs[i][attrib][chan] = + coefs[attrib].a0[chan] + + coefs[attrib].dadx[chan] * x + + coefs[attrib].dady[chan] * y; + } +} + +static inline void +perspective_interpolation(float (*inputs)[16][4], + const struct tgsi_interp_coef *coefs, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + const float x = inputs[i][0][0]; + const float y = inputs[i][0][1]; + /* WPOS.w here is really 1/w */ + const float w = 1.0f / inputs[i][0][3]; + assert(inputs[i][0][3] != 0.0); + + inputs[i][attrib][chan] = + (coefs[attrib].a0[chan] + + coefs[attrib].dadx[chan] * x + + coefs[attrib].dady[chan] * y) * w; + } +} + +void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix) +{ + if (!ir || !ir->module) + return; + + if (file_prefix) { + std::ostringstream stream; + stream << file_prefix; + stream << ir->id; + stream << ".ll"; + std::string name = stream.str(); + std::ofstream out(name.c_str()); + if (!out) { + std::cerr<<"Can't open file : "<module); + out.close(); + } else { + const llvm::Module::FunctionListType &funcs = ir->module->getFunctionList(); + llvm::Module::FunctionListType::const_iterator itr; + std::cout<<"; ---------- Start shader "<id<id<num_interp; ++i) { + const gallivm_interpolate &interp = prog->interpolators[i]; + switch (interp.type) { + case TGSI_INTERPOLATE_CONSTANT: + constant_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + case TGSI_INTERPOLATE_LINEAR: + linear_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + perspective_interpolation(inputs, coef, interp.attrib, interp.chan); + break; + + default: + assert( 0 ); + } + } +} + + +struct gallivm_ir * gallivm_ir_new(enum gallivm_shader_type type) +{ + struct gallivm_ir *ir = + (struct gallivm_ir *)calloc(1, sizeof(struct gallivm_ir)); + ++GLOBAL_ID; + ir->id = GLOBAL_ID; + ir->type = type; + + return ir; +} + +void gallivm_ir_set_layout(struct gallivm_ir *ir, + enum gallivm_vector_layout layout) +{ + ir->layout = layout; +} + +void gallivm_ir_set_components(struct gallivm_ir *ir, int num) +{ + ir->num_components = num; +} + +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, + const struct tgsi_token *tokens) +{ + std::cout << "Creating llvm from: " <module = mod; + gallivm_ir_dump(ir, 0); +} + +void gallivm_ir_delete(struct gallivm_ir *ir) +{ + delete ir->module; + free(ir); +} + +struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) +{ + struct gallivm_prog *prog = + (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); + llvm::Module *mod = llvm::CloneModule(ir->module); + prog->num_consts = ir->num_consts; + memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators)); + prog->num_interp = ir->num_interp; + + /* Run optimization passes over it */ + PassManager passes; + passes.add(new TargetData(mod)); + AddStandardCompilePasses(passes); + passes.run(*mod); + prog->module = mod; + + std::cout << "After optimizations:"<dump(); + + return prog; +} + +#endif /* MESA_LLVM */ diff --git a/src/gallium/auxiliary/gallivm/gallivm.h b/src/gallium/auxiliary/gallivm/gallivm.h new file mode 100644 index 0000000000..92da4bca7f --- /dev/null +++ b/src/gallium/auxiliary/gallivm/gallivm.h @@ -0,0 +1,103 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef GALLIVM_H +#define GALLIVM_H + +#if defined __cplusplus +extern "C" { +#endif + +#include "pipe/p_state.h" + +#ifdef MESA_LLVM + +struct tgsi_token; + +struct gallivm_ir; +struct gallivm_prog; +struct gallivm_cpu_engine; +struct tgsi_interp_coef; +struct tgsi_sampler; +struct tgsi_exec_vector; + +enum gallivm_shader_type { + GALLIVM_VS, + GALLIVM_FS +}; + +enum gallivm_vector_layout { + GALLIVM_AOS, + GALLIVM_SOA +}; + +struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type); +void gallivm_ir_set_layout(struct gallivm_ir *ir, + enum gallivm_vector_layout layout); +void gallivm_ir_set_components(struct gallivm_ir *ir, int num); +void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, + const struct tgsi_token *tokens); +void gallivm_ir_delete(struct gallivm_ir *ir); + + +struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir); + +void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + const struct tgsi_interp_coef *coefs); +void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); + + +struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); +struct gallivm_cpu_engine *gallivm_global_cpu_engine(); +int gallivm_cpu_vs_exec(struct gallivm_prog *prog, + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, + float (*consts)[4], + struct tgsi_exec_vector *temps); +int gallivm_cpu_fs_exec(struct gallivm_prog *prog, + float x, float y, + float (*dests)[PIPE_MAX_SHADER_INPUTS][4], + float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], + float (*consts)[4], + struct tgsi_sampler *samplers); +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); + + +#endif /* MESA_LLVM */ + +#if defined __cplusplus +} // extern "C" +#endif + +#endif diff --git a/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp b/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp new file mode 100644 index 0000000000..1796f0a177 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp @@ -0,0 +1,567 @@ +// Generated by llvm2cpp - DO NOT MODIFY! + + +Module* createGallivmBuiltins(Module *mod) { + +mod->setModuleIdentifier("shader"); + +// Type Definitions +ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 25); + +PointerType* PointerTy_1 = PointerType::get(ArrayTy_0, 0); + +std::vectorFuncTy_2_args; +FuncTy_2_args.push_back(Type::FloatTy); +FuncTy_2_args.push_back(Type::FloatTy); +FunctionType* FuncTy_2 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_2_args, + /*isVarArg=*/false); + +PointerType* PointerTy_3 = PointerType::get(FuncTy_2, 0); + +VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); + +std::vectorFuncTy_5_args; +FuncTy_5_args.push_back(VectorTy_4); +FunctionType* FuncTy_5 = FunctionType::get( + /*Result=*/VectorTy_4, + /*Params=*/FuncTy_5_args, + /*isVarArg=*/false); + +std::vectorFuncTy_6_args; +FuncTy_6_args.push_back(VectorTy_4); +FuncTy_6_args.push_back(VectorTy_4); +FuncTy_6_args.push_back(VectorTy_4); +FunctionType* FuncTy_6 = FunctionType::get( + /*Result=*/VectorTy_4, + /*Params=*/FuncTy_6_args, + /*isVarArg=*/false); + +VectorType* VectorTy_7 = VectorType::get(IntegerType::get(32), 4); + +std::vectorFuncTy_9_args; +FunctionType* FuncTy_9 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_9_args, + /*isVarArg=*/true); + +PointerType* PointerTy_8 = PointerType::get(FuncTy_9, 0); + +PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8), 0); + +std::vectorFuncTy_12_args; +FuncTy_12_args.push_back(Type::FloatTy); +FunctionType* FuncTy_12 = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/FuncTy_12_args, + /*isVarArg=*/false); + +PointerType* PointerTy_11 = PointerType::get(FuncTy_12, 0); + +std::vectorFuncTy_13_args; +FuncTy_13_args.push_back(VectorTy_4); +FunctionType* FuncTy_13 = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/FuncTy_13_args, + /*isVarArg=*/false); + + +// Function Declarations + +Function* func_approx = new Function( + /*Type=*/FuncTy_2, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"approx", mod); +func_approx->setCallingConv(CallingConv::C); +const ParamAttrsList *func_approx_PAL = 0; +func_approx->setParamAttrs(func_approx_PAL); + +Function* func_powf = new Function( + /*Type=*/FuncTy_2, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"powf", mod); // (external, no body) +func_powf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_powf_PAL = 0; +func_powf->setParamAttrs(func_powf_PAL); + +Function* func_lit = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"lit", mod); +func_lit->setCallingConv(CallingConv::C); +const ParamAttrsList *func_lit_PAL = 0; +func_lit->setParamAttrs(func_lit_PAL); + +Function* func_cmp = new Function( + /*Type=*/FuncTy_6, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"cmp", mod); +func_cmp->setCallingConv(CallingConv::C); +const ParamAttrsList *func_cmp_PAL = 0; +{ + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_cmp_PAL = ParamAttrsList::get(Attrs); + +} +func_cmp->setParamAttrs(func_cmp_PAL); + +Function* func_vcos = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"vcos", mod); +func_vcos->setCallingConv(CallingConv::C); +const ParamAttrsList *func_vcos_PAL = 0; +func_vcos->setParamAttrs(func_vcos_PAL); + +Function* func_printf = new Function( + /*Type=*/FuncTy_9, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", mod); // (external, no body) +func_printf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_printf_PAL = 0; +func_printf->setParamAttrs(func_printf_PAL); + +Function* func_cosf = new Function( + /*Type=*/FuncTy_12, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"cosf", mod); // (external, no body) +func_cosf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_cosf_PAL = 0; +func_cosf->setParamAttrs(func_cosf_PAL); + +Function* func_scs = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"scs", mod); +func_scs->setCallingConv(CallingConv::C); +const ParamAttrsList *func_scs_PAL = 0; +func_scs->setParamAttrs(func_scs_PAL); + +Function* func_sinf = new Function( + /*Type=*/FuncTy_12, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"sinf", mod); // (external, no body) +func_sinf->setCallingConv(CallingConv::C); +const ParamAttrsList *func_sinf_PAL = 0; +func_sinf->setParamAttrs(func_sinf_PAL); + +Function* func_vsin = new Function( + /*Type=*/FuncTy_5, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"vsin", mod); +func_vsin->setCallingConv(CallingConv::C); +const ParamAttrsList *func_vsin_PAL = 0; +func_vsin->setParamAttrs(func_vsin_PAL); + +Function* func_kilp = new Function( + /*Type=*/FuncTy_13, + /*Linkage=*/GlobalValue::WeakLinkage, + /*Name=*/"kilp", mod); +func_kilp->setCallingConv(CallingConv::C); +const ParamAttrsList *func_kilp_PAL = 0; +{ + ParamAttrsVector Attrs; + ParamAttrsWithIndex PAWI; + PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; + Attrs.push_back(PAWI); + func_kilp_PAL = ParamAttrsList::get(Attrs); + +} +func_kilp->setParamAttrs(func_kilp_PAL); + +// Global Variable Declarations + + +GlobalVariable* gvar_array__str = new GlobalVariable( +/*Type=*/ArrayTy_0, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str", +mod); + +GlobalVariable* gvar_array__str1 = new GlobalVariable( +/*Type=*/ArrayTy_0, +/*isConstant=*/true, +/*Linkage=*/GlobalValue::InternalLinkage, +/*Initializer=*/0, // has initializer, specified below +/*Name=*/".str1", +mod); + +// Constant Definitions +Constant* const_array_14 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); +Constant* const_array_15 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); +ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); +ConstantFP* const_float_17 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); +Constant* const_float_18 = Constant::getNullValue(Type::FloatTy); +Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); +std::vector const_packed_20_elems; +ConstantFP* const_float_21 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); +const_packed_20_elems.push_back(const_float_21); +UndefValue* const_float_22 = UndefValue::get(Type::FloatTy); +const_packed_20_elems.push_back(const_float_22); +const_packed_20_elems.push_back(const_float_22); +const_packed_20_elems.push_back(const_float_21); +Constant* const_packed_20 = ConstantVector::get(VectorTy_4, const_packed_20_elems); +ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); +ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "3", 10)); +ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "2", 10)); +std::vector const_packed_26_elems; +const_packed_26_elems.push_back(const_float_21); +const_packed_26_elems.push_back(const_float_18); +const_packed_26_elems.push_back(const_float_18); +const_packed_26_elems.push_back(const_float_21); +Constant* const_packed_26 = ConstantVector::get(VectorTy_4, const_packed_26_elems); +Constant* const_double_27 = Constant::getNullValue(Type::DoubleTy); +std::vector const_packed_28_elems; +const_packed_28_elems.push_back(const_int32_19); +ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "5", 10)); +const_packed_28_elems.push_back(const_int32_29); +const_packed_28_elems.push_back(const_int32_25); +const_packed_28_elems.push_back(const_int32_24); +Constant* const_packed_28 = ConstantVector::get(VectorTy_7, const_packed_28_elems); +std::vector const_packed_30_elems; +const_packed_30_elems.push_back(const_int32_19); +const_packed_30_elems.push_back(const_int32_23); +ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "6", 10)); +const_packed_30_elems.push_back(const_int32_31); +const_packed_30_elems.push_back(const_int32_24); +Constant* const_packed_30 = ConstantVector::get(VectorTy_7, const_packed_30_elems); +std::vector const_packed_32_elems; +const_packed_32_elems.push_back(const_int32_19); +const_packed_32_elems.push_back(const_int32_23); +const_packed_32_elems.push_back(const_int32_25); +ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "7", 10)); +const_packed_32_elems.push_back(const_int32_33); +Constant* const_packed_32 = ConstantVector::get(VectorTy_7, const_packed_32_elems); +std::vector const_ptr_34_indices; +const_ptr_34_indices.push_back(const_int32_19); +const_ptr_34_indices.push_back(const_int32_19); +Constant* const_ptr_34 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_34_indices[0], const_ptr_34_indices.size() ); +UndefValue* const_packed_35 = UndefValue::get(VectorTy_4); +std::vector const_ptr_36_indices; +const_ptr_36_indices.push_back(const_int32_19); +const_ptr_36_indices.push_back(const_int32_19); +Constant* const_ptr_36 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_36_indices[0], const_ptr_36_indices.size() ); + +// Global Variable Definitions +gvar_array__str->setInitializer(const_array_14); +gvar_array__str1->setInitializer(const_array_15); + +// Function Definitions + +// Function: approx (func_approx) +{ + Function::arg_iterator args = func_approx->arg_begin(); + Value* float_a = args++; + float_a->setName("a"); + Value* float_b = args++; + float_b->setName("b"); + + BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); + + // Block entry (label_entry) + FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_16, "cmp", label_entry); + SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_16, float_b, "b.addr.0", label_entry); + FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_17, "cmp3", label_entry); + SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_17, float_b_addr_0, "b.addr.1", label_entry); + FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_18, "cmp7", label_entry); + SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_18, float_a, "a.addr.0", label_entry); + std::vector float_call_params; + float_call_params.push_back(float_a_addr_0); + float_call_params.push_back(float_b_addr_1); + CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); + float_call->setCallingConv(CallingConv::C); + float_call->setTailCall(true);const ParamAttrsList *float_call_PAL = 0; + float_call->setParamAttrs(float_call_PAL); + + new ReturnInst(float_call, label_entry); + +} + +// Function: lit (func_lit) +{ + Function::arg_iterator args = func_lit->arg_begin(); + Value* packed_tmp = args++; + packed_tmp->setName("tmp"); + + BasicBlock* label_entry_38 = new BasicBlock("entry",func_lit,0); + BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); + BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); + + // Block entry (label_entry_38) + ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp6", label_entry_38); + FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp6, const_float_18, "cmp", label_entry_38); + new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_39, label_entry_38); + + // Block ifthen (label_ifthen) + InsertElementInst* packed_tmp10 = new InsertElementInst(const_packed_20, float_tmp6, const_int32_23, "tmp10", label_ifthen); + ExtractElementInst* float_tmp12 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp12", label_ifthen); + ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp14", label_ifthen); + std::vector float_call_41_params; + float_call_41_params.push_back(float_tmp12); + float_call_41_params.push_back(float_tmp14); + CallInst* float_call_41 = new CallInst(func_approx, float_call_41_params.begin(), float_call_41_params.end(), "call", label_ifthen); + float_call_41->setCallingConv(CallingConv::C); + float_call_41->setTailCall(true);const ParamAttrsList *float_call_41_PAL = 0; + float_call_41->setParamAttrs(float_call_41_PAL); + + InsertElementInst* packed_tmp16 = new InsertElementInst(packed_tmp10, float_call_41, const_int32_25, "tmp16", label_ifthen); + new ReturnInst(packed_tmp16, label_ifthen); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock) + new ReturnInst(const_packed_26, label_UnifiedReturnBlock); + +} + +// Function: cmp (func_cmp) +{ + Function::arg_iterator args = func_cmp->arg_begin(); + Value* packed_tmp0 = args++; + packed_tmp0->setName("tmp0"); + Value* packed_tmp1 = args++; + packed_tmp1->setName("tmp1"); + Value* packed_tmp2 = args++; + packed_tmp2->setName("tmp2"); + + BasicBlock* label_entry_44 = new BasicBlock("entry",func_cmp,0); + BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); + BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); + BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); + BasicBlock* label_cond_cont34 = new BasicBlock("cond.cont34",func_cmp,0); + BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); + BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); + + // Block entry (label_entry_44) + ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_44); + CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_44); + FCmpInst* int1_cmp_45 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_44); + ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_44); + CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_44); + FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_44); + SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_45, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_44); + new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_44); + + // Block cond.?14 (label_cond__14) + ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_28, "tmp233", label_cond__14); + ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp254", label_cond__14); + CastInst* double_conv265 = new FPExtInst(float_tmp254, Type::DoubleTy, "conv265", label_cond__14); + FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_27, "cmp276", label_cond__14); + new BranchInst(label_cond__28, label_cond_cont34, int1_cmp276, label_cond__14); + + // Block cond.cont20 (label_cond_cont20) + ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_28, "tmp23", label_cond_cont20); + ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp25", label_cond_cont20); + CastInst* double_conv26 = new FPExtInst(float_tmp25, Type::DoubleTy, "conv26", label_cond_cont20); + FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_27, "cmp27", label_cond_cont20); + new BranchInst(label_cond__28, label_cond_cont34, int1_cmp27, label_cond_cont20); + + // Block cond.?28 (label_cond__28) + PHINode* packed_tmp23_reg2mem_0 = new PHINode(VectorTy_4, "tmp23.reg2mem.0", label_cond__28); + packed_tmp23_reg2mem_0->reserveOperandSpace(2); + packed_tmp23_reg2mem_0->addIncoming(packed_tmp233, label_cond__14); + packed_tmp23_reg2mem_0->addIncoming(packed_tmp23, label_cond_cont20); + + ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_30, "tmp378", label_cond__28); + ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp399", label_cond__28); + CastInst* double_conv4010 = new FPExtInst(float_tmp399, Type::DoubleTy, "conv4010", label_cond__28); + FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_27, "cmp4111", label_cond__28); + new BranchInst(label_cond__42, label_cond_cont48, int1_cmp4111, label_cond__28); + + // Block cond.cont34 (label_cond_cont34) + PHINode* packed_tmp23_reg2mem_1 = new PHINode(VectorTy_4, "tmp23.reg2mem.1", label_cond_cont34); + packed_tmp23_reg2mem_1->reserveOperandSpace(2); + packed_tmp23_reg2mem_1->addIncoming(packed_tmp233, label_cond__14); + packed_tmp23_reg2mem_1->addIncoming(packed_tmp23, label_cond_cont20); + + ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_30, "tmp37", label_cond_cont34); + ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp39", label_cond_cont34); + CastInst* double_conv40 = new FPExtInst(float_tmp39, Type::DoubleTy, "conv40", label_cond_cont34); + FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_27, "cmp41", label_cond_cont34); + new BranchInst(label_cond__42, label_cond_cont48, int1_cmp41, label_cond_cont34); + + // Block cond.?42 (label_cond__42) + PHINode* packed_tmp37_reg2mem_0 = new PHINode(VectorTy_4, "tmp37.reg2mem.0", label_cond__42); + packed_tmp37_reg2mem_0->reserveOperandSpace(2); + packed_tmp37_reg2mem_0->addIncoming(packed_tmp378, label_cond__28); + packed_tmp37_reg2mem_0->addIncoming(packed_tmp37, label_cond_cont34); + + ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_32, "tmp5113", label_cond__42); + new ReturnInst(packed_tmp5113, label_cond__42); + + // Block cond.cont48 (label_cond_cont48) + PHINode* packed_tmp37_reg2mem_1 = new PHINode(VectorTy_4, "tmp37.reg2mem.1", label_cond_cont48); + packed_tmp37_reg2mem_1->reserveOperandSpace(2); + packed_tmp37_reg2mem_1->addIncoming(packed_tmp378, label_cond__28); + packed_tmp37_reg2mem_1->addIncoming(packed_tmp37, label_cond_cont34); + + ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_32, "tmp51", label_cond_cont48); + new ReturnInst(packed_tmp51, label_cond_cont48); + +} + +// Function: vcos (func_vcos) +{ + Function::arg_iterator args = func_vcos->arg_begin(); + Value* packed_val = args++; + packed_val->setName("val"); + + BasicBlock* label_entry_53 = new BasicBlock("entry",func_vcos,0); + + // Block entry (label_entry_53) + ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_53); + CastInst* double_conv_54 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_53); + ExtractElementInst* float_tmp3_55 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_53); + CastInst* double_conv4 = new FPExtInst(float_tmp3_55, Type::DoubleTy, "conv4", label_entry_53); + ExtractElementInst* float_tmp6_56 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_53); + CastInst* double_conv7 = new FPExtInst(float_tmp6_56, Type::DoubleTy, "conv7", label_entry_53); + ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_53); + CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_53); + std::vector int32_call_params; + int32_call_params.push_back(const_ptr_34); + int32_call_params.push_back(double_conv_54); + int32_call_params.push_back(double_conv4); + int32_call_params.push_back(double_conv7); + int32_call_params.push_back(double_conv10); + CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_53); + int32_call->setCallingConv(CallingConv::C); + int32_call->setTailCall(true);const ParamAttrsList *int32_call_PAL = 0; + int32_call->setParamAttrs(int32_call_PAL); + + CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_53); + float_call13->setCallingConv(CallingConv::C); + float_call13->setTailCall(true);const ParamAttrsList *float_call13_PAL = 0; + float_call13->setParamAttrs(float_call13_PAL); + + InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_53); + CallInst* float_call18 = new CallInst(func_cosf, float_tmp1, "call18", label_entry_53); + float_call18->setCallingConv(CallingConv::C); + float_call18->setTailCall(true);const ParamAttrsList *float_call18_PAL = 0; + float_call18->setParamAttrs(float_call18_PAL); + + InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call18, const_int32_23, "tmp20", label_entry_53); + CallInst* float_call23 = new CallInst(func_cosf, float_tmp1, "call23", label_entry_53); + float_call23->setCallingConv(CallingConv::C); + float_call23->setTailCall(true);const ParamAttrsList *float_call23_PAL = 0; + float_call23->setParamAttrs(float_call23_PAL); + + InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call23, const_int32_25, "tmp25", label_entry_53); + CallInst* float_call28 = new CallInst(func_cosf, float_tmp1, "call28", label_entry_53); + float_call28->setCallingConv(CallingConv::C); + float_call28->setTailCall(true);const ParamAttrsList *float_call28_PAL = 0; + float_call28->setParamAttrs(float_call28_PAL); + + InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call28, const_int32_24, "tmp30", label_entry_53); + CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_53); + CastInst* double_conv36 = new FPExtInst(float_call18, Type::DoubleTy, "conv36", label_entry_53); + CastInst* double_conv39 = new FPExtInst(float_call23, Type::DoubleTy, "conv39", label_entry_53); + CastInst* double_conv42 = new FPExtInst(float_call28, Type::DoubleTy, "conv42", label_entry_53); + std::vector int32_call43_params; + int32_call43_params.push_back(const_ptr_36); + int32_call43_params.push_back(double_conv33); + int32_call43_params.push_back(double_conv36); + int32_call43_params.push_back(double_conv39); + int32_call43_params.push_back(double_conv42); + CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_53); + int32_call43->setCallingConv(CallingConv::C); + int32_call43->setTailCall(true);const ParamAttrsList *int32_call43_PAL = 0; + int32_call43->setParamAttrs(int32_call43_PAL); + + new ReturnInst(packed_tmp30, label_entry_53); + +} + +// Function: scs (func_scs) +{ + Function::arg_iterator args = func_scs->arg_begin(); + Value* packed_val_58 = args++; + packed_val_58->setName("val"); + + BasicBlock* label_entry_59 = new BasicBlock("entry",func_scs,0); + + // Block entry (label_entry_59) + ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_58, const_int32_19, "tmp2", label_entry_59); + CallInst* float_call_60 = new CallInst(func_cosf, float_tmp2, "call", label_entry_59); + float_call_60->setCallingConv(CallingConv::C); + float_call_60->setTailCall(true);const ParamAttrsList *float_call_60_PAL = 0; + float_call_60->setParamAttrs(float_call_60_PAL); + + InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_60, const_int32_19, "tmp5", label_entry_59); + CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_59); + float_call7->setCallingConv(CallingConv::C); + float_call7->setTailCall(true);const ParamAttrsList *float_call7_PAL = 0; + float_call7->setParamAttrs(float_call7_PAL); + + InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_59); + new ReturnInst(packed_tmp9, label_entry_59); + +} + +// Function: vsin (func_vsin) +{ + Function::arg_iterator args = func_vsin->arg_begin(); + Value* packed_val_62 = args++; + packed_val_62->setName("val"); + + BasicBlock* label_entry_63 = new BasicBlock("entry",func_vsin,0); + + // Block entry (label_entry_63) + ExtractElementInst* float_tmp2_64 = new ExtractElementInst(packed_val_62, const_int32_19, "tmp2", label_entry_63); + CallInst* float_call_65 = new CallInst(func_sinf, float_tmp2_64, "call", label_entry_63); + float_call_65->setCallingConv(CallingConv::C); + float_call_65->setTailCall(true);const ParamAttrsList *float_call_65_PAL = 0; + float_call_65->setParamAttrs(float_call_65_PAL); + + InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_65, const_int32_19, "tmp6", label_entry_63); + InsertElementInst* packed_tmp9_66 = new InsertElementInst(packed_tmp6, float_call_65, const_int32_23, "tmp9", label_entry_63); + InsertElementInst* packed_tmp12 = new InsertElementInst(packed_tmp9_66, float_call_65, const_int32_25, "tmp12", label_entry_63); + InsertElementInst* packed_tmp15_67 = new InsertElementInst(packed_tmp12, float_call_65, const_int32_24, "tmp15", label_entry_63); + new ReturnInst(packed_tmp15_67, label_entry_63); + +} + +// Function: kilp (func_kilp) +{ + Function::arg_iterator args = func_kilp->arg_begin(); + Value* packed_val_69 = args++; + packed_val_69->setName("val"); + + BasicBlock* label_entry_70 = new BasicBlock("entry",func_kilp,0); + BasicBlock* label_lor_rhs = new BasicBlock("lor_rhs",func_kilp,0); + BasicBlock* label_lor_rhs5 = new BasicBlock("lor_rhs5",func_kilp,0); + BasicBlock* label_lor_rhs11 = new BasicBlock("lor_rhs11",func_kilp,0); + BasicBlock* label_UnifiedReturnBlock_71 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); + + // Block entry (label_entry_70) + ExtractElementInst* float_tmp1_72 = new ExtractElementInst(packed_val_69, const_int32_19, "tmp1", label_entry_70); + FCmpInst* int1_cmp_73 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_72, const_float_18, "cmp", label_entry_70); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs, int1_cmp_73, label_entry_70); + + // Block lor_rhs (label_lor_rhs) + ExtractElementInst* float_tmp3_75 = new ExtractElementInst(packed_val_69, const_int32_23, "tmp3", label_lor_rhs); + FCmpInst* int1_cmp4 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_75, const_float_18, "cmp4", label_lor_rhs); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs5, int1_cmp4, label_lor_rhs); + + // Block lor_rhs5 (label_lor_rhs5) + ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_val_69, const_int32_25, "tmp7", label_lor_rhs5); + FCmpInst* int1_cmp8 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp7, const_float_18, "cmp8", label_lor_rhs5); + new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs11, int1_cmp8, label_lor_rhs5); + + // Block lor_rhs11 (label_lor_rhs11) + ExtractElementInst* float_tmp13 = new ExtractElementInst(packed_val_69, const_int32_24, "tmp13", label_lor_rhs11); + FCmpInst* int1_cmp14 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp13, const_float_18, "cmp14", label_lor_rhs11); + CastInst* int32_retval = new ZExtInst(int1_cmp14, IntegerType::get(32), "retval", label_lor_rhs11); + new ReturnInst(int32_retval, label_lor_rhs11); + + // Block UnifiedReturnBlock (label_UnifiedReturnBlock_71) + new ReturnInst(const_int32_23, label_UnifiedReturnBlock_71); + +} + +return mod; + +} diff --git a/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp new file mode 100644 index 0000000000..8f9830d0b1 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp @@ -0,0 +1,202 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "instructions.h" +#include "loweringpass.h" +#include "storage.h" +#include "tgsitollvm.h" + +#include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_dump.h" + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +struct gallivm_cpu_engine { + llvm::ExecutionEngine *engine; +}; + +static struct gallivm_cpu_engine *CPU = 0; + +typedef int (*fragment_shader_runner)(float x, float y, + float (*dests)[16][4], + float (*inputs)[16][4], + int num_attribs, + float (*consts)[4], int num_consts, + struct tgsi_sampler *samplers); + +int gallivm_cpu_fs_exec(struct gallivm_prog *prog, + float fx, float fy, + float (*dests)[16][4], + float (*inputs)[16][4], + float (*consts)[4], + struct tgsi_sampler *samplers) +{ + fragment_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + + return runner(fx, fy, dests, inputs, prog->num_interp, + consts, prog->num_consts, + samplers); +} + +static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) +{ + llvm::Module *mod = prog->module; + llvm::Function *func = 0; + + switch (prog->type) { + case GALLIVM_VS: + func = mod->getFunction("vs_shader"); + break; + case GALLIVM_FS: + func = mod->getFunction("fs_shader"); + break; + default: + assert(!"Unknown shader type!"); + break; + } + return func; +} + +/*! + This function creates a CPU based execution engine for the given gallivm_prog. + gallivm_cpu_engine should be used as a singleton throughout the library. Before + executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile. + The gallivm_prog instance which is being passed to the constructor is being + automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile + with it again. + */ +struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) +{ + struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *) + calloc(1, sizeof(struct gallivm_cpu_engine)); + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); + ee->DisableLazyCompilation(); + cpu->engine = ee; + + llvm::Function *func = func_for_shader(prog); + + prog->function = ee->getPointerToFunction(func); + CPU = cpu; + return cpu; +} + + +/*! + This function JIT compiles the given gallivm_prog with the given cpu based execution engine. + The reference to the generated machine code entry point will be stored + in the gallivm_prog program. After executing this function one can call gallivm_prog_exec + in order to execute the gallivm_prog on the CPU. + */ +void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) +{ + llvm::Module *mod = static_cast(prog->module); + llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); + llvm::ExecutionEngine *ee = cpu->engine; + assert(ee); + /*FIXME : remove */ + ee->DisableLazyCompilation(); + ee->addModuleProvider(mp); + + llvm::Function *func = func_for_shader(prog); + prog->function = ee->getPointerToFunction(func); +} + +void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) +{ + free(cpu); +} + +struct gallivm_cpu_engine * gallivm_global_cpu_engine() +{ + return CPU; +} + + +typedef void (*vertex_shader_runner)(void *ainputs, + void *dests, + float (*aconsts)[4], + void *temps); + + +/*! + This function is used to execute the gallivm_prog in software. Before calling + this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile + function. + */ +int gallivm_cpu_vs_exec(struct gallivm_prog *prog, + struct tgsi_exec_vector *inputs, + struct tgsi_exec_vector *dests, + float (*consts)[4], + struct tgsi_exec_vector *temps) +{ + vertex_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); + /*FIXME*/ + runner(inputs, dests, consts, temps); + + return 0; +} + +#endif diff --git a/src/gallium/auxiliary/gallivm/gallivm_p.h b/src/gallium/auxiliary/gallivm/gallivm_p.h new file mode 100644 index 0000000000..cfe7b1901b --- /dev/null +++ b/src/gallium/auxiliary/gallivm/gallivm_p.h @@ -0,0 +1,110 @@ +#ifndef GALLIVM_P_H +#define GALLIVM_P_H + +#ifdef MESA_LLVM + +#include "gallivm.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/p_compiler.h" + +namespace llvm { + class Module; +} + +#if defined __cplusplus +extern "C" { +#endif + +enum gallivm_shader_type; +enum gallivm_vector_layout; + +struct gallivm_interpolate { + int attrib; + int chan; + int type; +}; + +struct gallivm_ir { + llvm::Module *module; + int id; + enum gallivm_shader_type type; + enum gallivm_vector_layout layout; + int num_components; + int num_consts; + + //FIXME: this might not be enough for some shaders + struct gallivm_interpolate interpolators[32*4]; + int num_interp; +}; + +struct gallivm_prog { + llvm::Module *module; + void *function; + + int id; + enum gallivm_shader_type type; + + int num_consts; + + //FIXME: this might not be enough for some shaders + struct gallivm_interpolate interpolators[32*4]; + int num_interp; +}; + +static INLINE void gallivm_swizzle_components(int swizzle, + int *xc, int *yc, + int *zc, int *wc) +{ + int x = swizzle / 1000; swizzle -= x * 1000; + int y = swizzle / 100; swizzle -= y * 100; + int z = swizzle / 10; swizzle -= z * 10; + int w = swizzle; + + if (xc) *xc = x; + if (yc) *yc = y; + if (zc) *zc = z; + if (wc) *wc = w; +} + +static INLINE boolean gallivm_is_swizzle(int swizzle) +{ + const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + + TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; + return swizzle != NO_SWIZZLE; +} + +static INLINE int gallivm_x_swizzle(int swizzle) +{ + int x; + gallivm_swizzle_components(swizzle, &x, 0, 0, 0); + return x; +} + +static INLINE int gallivm_y_swizzle(int swizzle) +{ + int y; + gallivm_swizzle_components(swizzle, 0, &y, 0, 0); + return y; +} + +static INLINE int gallivm_z_swizzle(int swizzle) +{ + int z; + gallivm_swizzle_components(swizzle, 0, 0, &z, 0); + return z; +} + +static INLINE int gallivm_w_swizzle(int swizzle) +{ + int w; + gallivm_swizzle_components(swizzle, 0, 0, 0, &w); + return w; +} + +#endif /* MESA_LLVM */ + +#if defined __cplusplus +} // extern "C" +#endif + +#endif diff --git a/src/gallium/auxiliary/gallivm/instructions.cpp b/src/gallium/auxiliary/gallivm/instructions.cpp new file mode 100644 index 0000000000..55d39fa5f1 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/instructions.cpp @@ -0,0 +1,889 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "instructions.h" + +#include "storage.h" + +#include +#include +#include +#include +#include +#include +#include + +#include +#include +#include + +using namespace llvm; + +#include "gallivm_builtins.cpp" + +static inline std::string createFuncName(int label) +{ + std::ostringstream stream; + stream << "function"; + stream << label; + return stream.str(); +} + +Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, + Storage *storage) + : m_mod(mod), m_func(func), m_builder(block), m_idx(0), + m_storage(storage) +{ + m_floatVecType = VectorType::get(Type::FloatTy, 4); + + m_llvmFSqrt = 0; + m_llvmFAbs = 0; + m_llvmPow = 0; + m_llvmFloor = 0; + m_llvmFlog = 0; + m_llvmLit = 0; + m_fmtPtr = 0; + + createGallivmBuiltins(m_mod); +} + +llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) +{ + return m_builder.CreateAdd(in1, in2, name("add")); +} + +llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3) +{ + Value *mulRes = mul(in1, in2); + return add(mulRes, in3); +} + +llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2) +{ + return m_builder.CreateMul(in1, in2, name("mul")); +} + +const char * Instructions::name(const char *prefix) +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + Value *x = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(0), + name("extractx")); + Value *y = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(1), + name("extracty")); + Value *z = m_builder.CreateExtractElement(mulRes, + m_storage->constantInt(2), + name("extractz")); + Value *xy = m_builder.CreateAdd(x, y,name("xy")); + Value *dot3 = m_builder.CreateAdd(xy, z, name("dot3")); + return vectorFromVals(dot3, dot3, dot3, dot3); +} + +llvm::Value *Instructions::callFSqrt(llvm::Value *val) +{ + if (!m_llvmFSqrt) { + // predeclare the intrinsic + std::vector fsqrtArgs; + fsqrtArgs.push_back(Type::FloatTy); + ParamAttrsList *fsqrtPal = 0; + FunctionType* fsqrtType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/fsqrtArgs, + /*isVarArg=*/false); + m_llvmFSqrt = new Function( + /*Type=*/fsqrtType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"llvm.sqrt.f32", m_mod); + m_llvmFSqrt->setCallingConv(CallingConv::C); + m_llvmFSqrt->setParamAttrs(fsqrtPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFSqrt, val, + name("sqrt")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::rsq(llvm::Value *in1) +{ + Value *x = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("extractx")); + Value *abs = callFAbs(x); + Value *sqrt = callFSqrt(abs); + + Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + sqrt, + name("rsqrt")); + return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt); +} + +llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w) +{ + Constant *const_vec = Constant::getNullValue(m_floatVecType); + Value *res = m_builder.CreateInsertElement(const_vec, x, + m_storage->constantInt(0), + name("vecx")); + res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), + name("vecxy")); + res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), + name("vecxyz")); + if (w) + res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), + name("vecxyzw")); + return res; +} + +llvm::Value *Instructions::callFAbs(llvm::Value *val) +{ + if (!m_llvmFAbs) { + // predeclare the intrinsic + std::vector fabsArgs; + fabsArgs.push_back(Type::FloatTy); + ParamAttrsList *fabsPal = 0; + FunctionType* fabsType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/fabsArgs, + /*isVarArg=*/false); + m_llvmFAbs = new Function( + /*Type=*/fabsType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"fabs", m_mod); + m_llvmFAbs->setCallingConv(CallingConv::C); + m_llvmFAbs->setParamAttrs(fabsPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFAbs, val, + name("fabs")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::lit(llvm::Value *in) +{ + if (!m_llvmLit) { + m_llvmLit = m_mod->getFunction("lit"); + } + CallInst *call = m_builder.CreateCall(m_llvmLit, in, name("litres")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) +{ + Value *res = m_builder.CreateSub(in1, in2, name("sub")); + return res; +} + +llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) +{ + if (!m_llvmPow) { + // predeclare the intrinsic + std::vector powArgs; + powArgs.push_back(Type::FloatTy); + powArgs.push_back(Type::FloatTy); + ParamAttrsList *powPal = 0; + FunctionType* powType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/powArgs, + /*isVarArg=*/false); + m_llvmPow = new Function( + /*Type=*/powType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"llvm.pow.f32", m_mod); + m_llvmPow->setCallingConv(CallingConv::C); + m_llvmPow->setParamAttrs(powPal); + } + std::vector params; + params.push_back(val1); + params.push_back(val2); + CallInst *call = m_builder.CreateCall(m_llvmPow, params.begin(), params.end(), + name("pow")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2) +{ + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); + llvm::Value *val = callPow(x1, x2); + return vectorFromVals(val, val, val, val); +} + +llvm::Value * Instructions::rcp(llvm::Value *in1) +{ + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *res = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, + APFloat(1.f)), + x1, name("rcp")); + return vectorFromVals(res, res, res, res); +} + +llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + std::vector vec = extractVector(mulRes); + Value *xy = m_builder.CreateAdd(vec[0], vec[1], name("xy")); + Value *xyz = m_builder.CreateAdd(xy, vec[2], name("xyz")); + Value *dot4 = m_builder.CreateAdd(xyz, vec[3], name("dot4")); + return vectorFromVals(dot4, dot4, dot4, dot4); +} + +llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) +{ + Value *mulRes = mul(in1, in2); + std::vector vec1 = extractVector(mulRes); + Value *xy = m_builder.CreateAdd(vec1[0], vec1[1], name("xy")); + Value *xyz = m_builder.CreateAdd(xy, vec1[2], name("xyz")); + Value *dph = m_builder.CreateAdd(xyz, vec1[3], name("dph")); + return vectorFromVals(dph, dph, dph, dph); +} + +llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) +{ + Value *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + Value *z = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z")); + Value *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + Value *w = m_builder.CreateExtractElement(in2, + m_storage->constantInt(3), + name("w")); + Value *ry = m_builder.CreateMul(y1, y2, name("tyuy")); + return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), + ry, z, w); +} + +llvm::Value * Instructions::ex2(llvm::Value *in) +{ + llvm::Value *val = callPow(ConstantFP::get(Type::FloatTy, APFloat(2.f)), + m_builder.CreateExtractElement( + in, m_storage->constantInt(0), + name("x1"))); + return vectorFromVals(val, val, val, val); +} + +llvm::Value * Instructions::callFloor(llvm::Value *val) +{ + if (!m_llvmFloor) { + // predeclare the intrinsic + std::vector floorArgs; + floorArgs.push_back(Type::FloatTy); + ParamAttrsList *floorPal = 0; + FunctionType* floorType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/floorArgs, + /*isVarArg=*/false); + m_llvmFloor = new Function( + /*Type=*/floorType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"floorf", m_mod); + m_llvmFloor->setCallingConv(CallingConv::C); + m_llvmFloor->setParamAttrs(floorPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFloor, val, + name("floorf")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::floor(llvm::Value *in) +{ + std::vector vec = extractVector(in); + return vectorFromVals(callFloor(vec[0]), callFloor(vec[1]), + callFloor(vec[2]), callFloor(vec[3])); +} + +llvm::Value * Instructions::arl(llvm::Value *in) +{ + return floor(in); +} + +llvm::Value * Instructions::frc(llvm::Value *in) +{ + llvm::Value *flr = floor(in); + return sub(in, flr); +} + +llvm::Value * Instructions::callFLog(llvm::Value *val) +{ + if (!m_llvmFlog) { + // predeclare the intrinsic + std::vector flogArgs; + flogArgs.push_back(Type::FloatTy); + ParamAttrsList *flogPal = 0; + FunctionType* flogType = FunctionType::get( + /*Result=*/Type::FloatTy, + /*Params=*/flogArgs, + /*isVarArg=*/false); + m_llvmFlog = new Function( + /*Type=*/flogType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"logf", m_mod); + m_llvmFlog->setCallingConv(CallingConv::C); + m_llvmFlog->setParamAttrs(flogPal); + } + CallInst *call = m_builder.CreateCall(m_llvmFlog, val, + name("logf")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::lg2(llvm::Value *in) +{ + std::vector vec = extractVector(in); + llvm::Value *const_vec = constVector(1.442695f, 1.442695f, + 1.442695f, 1.442695f); + return mul(vectorFromVals(callFLog(vec[0]), callFLog(vec[1]), + callFLog(vec[2]), callFLog(vec[3])), const_vec); +} + +llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) +{ + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); + Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); + + Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); + Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); + + Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); + Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); + + Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); + Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); + + return vectorFromVals(selx, sely, selz, selw); +} + +llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) +{ + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], + name("xcmp")); + Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], + name("selx")); + + Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], + name("ycmp")); + Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], + name("sely")); + + Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], + name("zcmp")); + Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], + name("selz")); + + Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], + name("wcmp")); + Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], + name("selw")); + + return vectorFromVals(selx, sely, selz, selw); +} + +void Instructions::printVector(llvm::Value *val) +{ + static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A"; + + if (!m_fmtPtr) { + Constant *format = ConstantArray::get(frmt, true); + ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1); + GlobalVariable* globalFormat = new GlobalVariable( + /*Type=*/arrayTy, + /*isConstant=*/true, + /*Linkage=*/GlobalValue::InternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name(".str"), + m_mod); + globalFormat->setInitializer(format); + + Constant* const_int0 = Constant::getNullValue(IntegerType::get(32)); + std::vector const_ptr_21_indices; + const_ptr_21_indices.push_back(const_int0); + const_ptr_21_indices.push_back(const_int0); + m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat, + &const_ptr_21_indices[0], const_ptr_21_indices.size()); + } + + Function *func_printf = m_mod->getFunction("printf"); + if (!func_printf) + func_printf = declarePrintf(); + assert(func_printf); + std::vector vec = extractVector(val); + Value *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx")); + Value *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy")); + Value *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz")); + Value *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw")); + std::vector params; + params.push_back(m_fmtPtr); + params.push_back(dx); + params.push_back(dy); + params.push_back(dz); + params.push_back(dw); + CallInst *call = m_builder.CreateCall(func_printf, params.begin(), params.end(), + name("printf")); + call->setCallingConv(CallingConv::C); + call->setTailCall(true); +} + +llvm::Function * Instructions::declarePrintf() +{ + std::vector args; + ParamAttrsList *params = 0; + FunctionType* funcTy = FunctionType::get( + /*Result=*/IntegerType::get(32), + /*Params=*/args, + /*isVarArg=*/true); + Function* func_printf = new Function( + /*Type=*/funcTy, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/"printf", m_mod); + func_printf->setCallingConv(CallingConv::C); + func_printf->setParamAttrs(params); + return func_printf; +} + + +llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], name("xcmp")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + + Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], name("ycmp")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + + Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], name("zcmp")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + + Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], name("wcmp")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + + return vectorFromVals(x, y, z, w); +} +llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOGE(vec1[0], vec2[0], name("xcmp")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + + Value *ycmp = m_builder.CreateFCmpOGE(vec1[1], vec2[1], name("ycmp")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + + Value *zcmp = m_builder.CreateFCmpOGE(vec1[2], vec2[2], name("zcmp")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + + Value *wcmp = m_builder.CreateFCmpOGE(vec1[3], vec2[3], name("wcmp")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + + return vectorFromVals(x, y, z, w); +} + + +llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) +{ + Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const0f = Constant::getNullValue(Type::FloatTy); + + std::vector vec1 = extractVector(in1); + std::vector vec2 = extractVector(in2); + + Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); + Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); + + Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); + Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); + + Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); + Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); + + Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); + Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); + + return vectorFromVals(x, y, z, w); +} + +llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) +{ + Value *x1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(0), + name("x1")); + Value *y1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(1), + name("y1")); + Value *z1 = m_builder.CreateExtractElement(in1, + m_storage->constantInt(2), + name("z1")); + + Value *x2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(0), + name("x2")); + Value *y2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(1), + name("y2")); + Value *z2 = m_builder.CreateExtractElement(in2, + m_storage->constantInt(2), + name("z2")); + Value *y1z2 = mul(y1, z2); + Value *z1y2 = mul(z1, y2); + + Value *z1x2 = mul(z1, x2); + Value *x1z2 = mul(x1, z2); + + Value *x1y2 = mul(x1, y2); + Value *y1x2 = mul(y1, x2); + + return vectorFromVals(sub(y1z2, z1y2), sub(z1x2, x1z2), sub(x1y2, y1x2)); +} + + +llvm::Value * Instructions::abs(llvm::Value *in) +{ + std::vector vec = extractVector(in); + Value *xabs = callFAbs(vec[0]); + Value *yabs = callFAbs(vec[1]); + Value *zabs = callFAbs(vec[2]); + Value *wabs = callFAbs(vec[3]); + return vectorFromVals(xabs, yabs, zabs, wabs); +} + +void Instructions::ifop(llvm::Value *in) +{ + BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); + BasicBlock *ifend = new BasicBlock(name("ifthenend"), m_func,0); + + //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0); + //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0); + //BasicBlock *wblock = new BasicBlock(name("wblock"), m_func,0); + + Constant *float0 = Constant::getNullValue(Type::FloatTy); + + Value *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0), + name("extractx")); + Value *xcmp = m_builder.CreateFCmpUNE(x, float0, name("xcmp")); + m_builder.CreateCondBr(xcmp, ifthen, ifend); + //m_builder.SetInsertPoint(yblock); + + m_builder.SetInsertPoint(ifthen); + m_ifStack.push(ifend); +} + +llvm::BasicBlock * Instructions::currentBlock() const +{ + return m_builder.GetInsertBlock(); +} + +void Instructions::elseop() +{ + assert(!m_ifStack.empty()); + BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); + m_builder.CreateBr(ifend); + m_builder.SetInsertPoint(m_ifStack.top()); + currentBlock()->setName(name("ifelse")); + m_ifStack.pop(); + m_ifStack.push(ifend); +} + +void Instructions::endif() +{ + assert(!m_ifStack.empty()); + m_builder.CreateBr(m_ifStack.top()); + m_builder.SetInsertPoint(m_ifStack.top()); + m_ifStack.pop(); +} + +llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3) +{ + llvm::Value *m = mul(in1, in2); + llvm::Value *vec1 = constVector(1.f, 1.f, 1.f, 1.f); + llvm::Value *s = sub(vec1, in1); + return add(m, mul(s, in3)); +} + +void Instructions::beginLoop() +{ + BasicBlock *begin = new BasicBlock(name("loop"), m_func,0); + BasicBlock *end = new BasicBlock(name("endloop"), m_func,0); + + m_builder.CreateBr(begin); + Loop loop; + loop.begin = begin; + loop.end = end; + m_builder.SetInsertPoint(begin); + m_loopStack.push(loop); +} + +void Instructions::endLoop() +{ + assert(!m_loopStack.empty()); + Loop loop = m_loopStack.top(); + m_builder.CreateBr(loop.begin); + loop.end->moveAfter(currentBlock()); + m_builder.SetInsertPoint(loop.end); + m_loopStack.pop(); +} + +void Instructions::brk() +{ + assert(!m_loopStack.empty()); + BasicBlock *unr = new BasicBlock(name("unreachable"), m_func,0); + m_builder.CreateBr(m_loopStack.top().end); + m_builder.SetInsertPoint(unr); +} + +llvm::Value * Instructions::trunc(llvm::Value *in) +{ + std::vector vec = extractVector(in); + Value *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32), + name("ftoix")); + Value *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32), + name("ftoiy")); + Value *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32), + name("ftoiz")); + Value *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32), + name("ftoiw")); + Value *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy, + name("fx")); + Value *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy, + name("fy")); + Value *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy, + name("fz")); + Value *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy, + name("fw")); + return vectorFromVals(fx, fy, fz, fw); +} + +void Instructions::end() +{ + m_builder.CreateRetVoid(); +} + +void Instructions::cal(int label, llvm::Value *input) +{ + std::vector params; + params.push_back(input); + llvm::Function *func = findFunction(label); + + m_builder.CreateCall(func, params.begin(), params.end()); +} + +llvm::Function * Instructions::declareFunc(int label) +{ + PointerType *vecPtr = PointerType::getUnqual(m_floatVecType); + std::vector args; + args.push_back(vecPtr); + args.push_back(vecPtr); + args.push_back(vecPtr); + args.push_back(vecPtr); + ParamAttrsList *params = 0; + FunctionType *funcType = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/args, + /*isVarArg=*/false); + std::string name = createFuncName(label); + Function *func = new Function( + /*Type=*/funcType, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Name=*/name.c_str(), m_mod); + func->setCallingConv(CallingConv::C); + func->setParamAttrs(params); + return func; +} + +void Instructions::bgnSub(unsigned label) +{ + llvm::Function *func = findFunction(label); + + Function::arg_iterator args = func->arg_begin(); + Value *ptr_INPUT = args++; + ptr_INPUT->setName("INPUT"); + m_storage->pushArguments(ptr_INPUT); + + llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); + + m_func = func; + m_builder.SetInsertPoint(entry); +} + +void Instructions::endSub() +{ + m_func = 0; + m_builder.SetInsertPoint(0); +} + +llvm::Function * Instructions::findFunction(int label) +{ + llvm::Function *func = m_functions[label]; + if (!func) { + func = declareFunc(label); + m_functions[label] = func; + } + return func; +} + +llvm::Value * Instructions::constVector(float x, float y, float z, float w) +{ + std::vector vec(4); + vec[0] = ConstantFP::get(Type::FloatTy, APFloat(x)); + vec[1] = ConstantFP::get(Type::FloatTy, APFloat(y)); + vec[2] = ConstantFP::get(Type::FloatTy, APFloat(z)); + vec[3] = ConstantFP::get(Type::FloatTy, APFloat(w)); + return ConstantVector::get(m_floatVecType, vec); +} + + +std::vector Instructions::extractVector(llvm::Value *vec) +{ + std::vector elems(4); + elems[0] = m_builder.CreateExtractElement(vec, m_storage->constantInt(0), + name("x")); + elems[1] = m_builder.CreateExtractElement(vec, m_storage->constantInt(1), + name("y")); + elems[2] = m_builder.CreateExtractElement(vec, m_storage->constantInt(2), + name("z")); + elems[3] = m_builder.CreateExtractElement(vec, m_storage->constantInt(3), + name("w")); + return elems; +} + +llvm::Value * Instructions::cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3) +{ + llvm::Function *func = m_mod->getFunction("cmp"); + assert(func); + + std::vector params; + params.push_back(in1); + params.push_back(in2); + params.push_back(in3); + CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), name("cmpres")); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::cos(llvm::Value *in) +{ +#if 0 + llvm::Function *func = m_mod->getFunction("vcos"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("cosres")); + call->setTailCall(false); + return call; +#else + std::vector elems = extractVector(in); + Function *func = m_mod->getFunction("cosf"); + assert(func); + CallInst *cos = m_builder.CreateCall(func, elems[0], name("cosres")); + cos->setCallingConv(CallingConv::C); + cos->setTailCall(true); + return vectorFromVals(cos, cos, cos, cos); +#endif +} + +llvm::Value * Instructions::scs(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("scs"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("scsres")); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::kilp(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("kilp"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("kilpres")); + call->setTailCall(false); + return call; +} + +llvm::Value * Instructions::sin(llvm::Value *in) +{ + llvm::Function *func = m_mod->getFunction("vsin"); + assert(func); + + CallInst *call = m_builder.CreateCall(func, in, name("sinres")); + call->setTailCall(false); + return call; +} +#endif //MESA_LLVM + + diff --git a/src/gallium/auxiliary/gallivm/instructions.h b/src/gallium/auxiliary/gallivm/instructions.h new file mode 100644 index 0000000000..9ebc17dd8e --- /dev/null +++ b/src/gallium/auxiliary/gallivm/instructions.h @@ -0,0 +1,152 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef INSTRUCTIONS_H +#define INSTRUCTIONS_H + +#include +#include +#include +#include + +#include +#include + +namespace llvm { + class VectorType; + class Function; +} + +class Storage; + +class Instructions +{ +public: + Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, + Storage *storage); + + llvm::BasicBlock *currentBlock() const; + + llvm::Value *abs(llvm::Value *in1); + llvm::Value *arl(llvm::Value *in1); + llvm::Value *add(llvm::Value *in1, llvm::Value *in2); + void beginLoop(); + void bgnSub(unsigned); + void brk(); + void cal(int label, llvm::Value *input); + llvm::Value *cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); + llvm::Value *cos(llvm::Value *in); + llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); + llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); + void elseop(); + void endif(); + void endLoop(); + void end(); + void endSub(); + llvm::Value *ex2(llvm::Value *in); + llvm::Value *floor(llvm::Value *in); + llvm::Value *frc(llvm::Value *in); + void ifop(llvm::Value *in); + llvm::Value *kilp(llvm::Value *in); + llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in3); + llvm::Value *lit(llvm::Value *in); + llvm::Value *lg2(llvm::Value *in); + llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, + llvm::Value *in2); + llvm::Value *min(llvm::Value *in1, llvm::Value *in2); + llvm::Value *max(llvm::Value *in1, llvm::Value *in2); + llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); + llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); + llvm::Value *rcp(llvm::Value *in); + llvm::Value *rsq(llvm::Value *in); + llvm::Value *scs(llvm::Value *in); + llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sin(llvm::Value *in); + llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); + llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); + llvm::Value *trunc(llvm::Value *in); + + void printVector(llvm::Value *val); +private: + const char *name(const char *prefix); + + llvm::Value *callFAbs(llvm::Value *val); + llvm::Value *callFloor(llvm::Value *val); + llvm::Value *callFSqrt(llvm::Value *val); + llvm::Value *callFLog(llvm::Value *val); + llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2); + + llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w=0); + + llvm::Value *constVector(float x, float y, float z, float w); + + llvm::Function *declarePrintf(); + llvm::Function *declareFunc(int label); + + llvm::Function *findFunction(int label); + + std::vector extractVector(llvm::Value *vec); +private: + llvm::Module *m_mod; + llvm::Function *m_func; + char m_name[32]; + llvm::LLVMFoldingBuilder m_builder; + int m_idx; + + llvm::VectorType *m_floatVecType; + + llvm::Function *m_llvmFSqrt; + llvm::Function *m_llvmFAbs; + llvm::Function *m_llvmPow; + llvm::Function *m_llvmFloor; + llvm::Function *m_llvmFlog; + llvm::Function *m_llvmLit; + + llvm::Constant *m_fmtPtr; + + std::stack m_ifStack; + struct Loop { + llvm::BasicBlock *begin; + llvm::BasicBlock *end; + }; + std::stack m_loopStack; + std::map m_functions; + Storage *m_storage; +}; + +#endif diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp new file mode 100644 index 0000000000..a4d5046637 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -0,0 +1,121 @@ +#include "instructionssoa.h" + +#include "storagesoa.h" + +#include + +using namespace llvm; + +InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, + llvm::BasicBlock *block, StorageSoa *storage) + : m_builder(block), + m_storage(storage), + m_idx(0) +{ +} + +const char * InstructionsSoa::name(const char *prefix) const +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::Value * InstructionsSoa::vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + Constant *constVector = Constant::getNullValue(vectorType); + Value *res = m_builder.CreateInsertElement(constVector, x, + m_storage->constantInt(0), + name("vecx")); + res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), + name("vecxy")); + res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), + name("vecxyz")); + if (w) + res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), + name("vecxyzw")); + return res; +} + +std::vector InstructionsSoa::arl(const std::vector in) +{ + std::vector res(4); + + //Extract x's + llvm::Value *x1 = m_builder.CreateExtractElement(in[0], + m_storage->constantInt(0), + name("extractX")); + //cast it to an unsigned int + x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast")); + + res[0] = x1;//vectorFromVals(x1, x2, x3, x4); + //only x is valid. the others shouldn't be necessary + /* + res[1] = Constant::getNullValue(m_floatVecType); + res[2] = Constant::getNullValue(m_floatVecType); + res[3] = Constant::getNullValue(m_floatVecType); + */ + + return res; +} + + +std::vector InstructionsSoa::add(const std::vector in1, + const std::vector in2) +{ + std::vector res(4); + + res[0] = m_builder.CreateAdd(in1[0], in2[0], name("addx")); + res[1] = m_builder.CreateAdd(in1[1], in2[1], name("addy")); + res[2] = m_builder.CreateAdd(in1[2], in2[2], name("addz")); + res[3] = m_builder.CreateAdd(in1[3], in2[3], name("addw")); + + return res; +} + +std::vector InstructionsSoa::mul(const std::vector in1, + const std::vector in2) +{ + std::vector res(4); + + res[0] = m_builder.CreateMul(in1[0], in2[0], name("mulx")); + res[1] = m_builder.CreateMul(in1[1], in2[1], name("muly")); + res[2] = m_builder.CreateMul(in1[2], in2[2], name("mulz")); + res[3] = m_builder.CreateMul(in1[3], in2[3], name("mulw")); + + return res; +} + +void InstructionsSoa::end() +{ + m_builder.CreateRetVoid(); +} + +std::vector InstructionsSoa::madd(const std::vector in1, + const std::vector in2, + const std::vector in3) +{ + std::vector res = mul(in1, in2); + return add(res, in3); +} + +std::vector InstructionsSoa::extractVector(llvm::Value *vector) +{ + std::vector res(4); + res[0] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(0), + name("extract1X")); + res[1] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(1), + name("extract2X")); + res[2] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(2), + name("extract3X")); + res[3] = m_builder.CreateExtractElement(vector, + m_storage->constantInt(3), + name("extract4X")); + + return res; +} diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h new file mode 100644 index 0000000000..4169dcbb2e --- /dev/null +++ b/src/gallium/auxiliary/gallivm/instructionssoa.h @@ -0,0 +1,74 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INSTRUCTIONSSOA_H +#define INSTRUCTIONSSOA_H + +#include + +#include + +namespace llvm { + class Module; + class Function; + class BasicBlock; + class Value; +} +class StorageSoa; + +class InstructionsSoa +{ +public: + InstructionsSoa(llvm::Module *mod, llvm::Function *func, + llvm::BasicBlock *block, StorageSoa *storage); + + std::vector arl(const std::vector in); + + std::vector add(const std::vector in1, + const std::vector in2); + std::vector madd(const std::vector in1, + const std::vector in2, + const std::vector in3); + std::vector mul(const std::vector in1, + const std::vector in2); + void end(); + + std::vector extractVector(llvm::Value *vector); +private: + const char * name(const char *prefix) const; + llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, + llvm::Value *z, llvm::Value *w); +private: + llvm::LLVMFoldingBuilder m_builder; + StorageSoa *m_storage; +private: + mutable int m_idx; + mutable char m_name[32]; +}; + + +#endif diff --git a/src/gallium/auxiliary/gallivm/llvm_builtins.c b/src/gallium/auxiliary/gallivm/llvm_builtins.c new file mode 100644 index 0000000000..4f98d754ba --- /dev/null +++ b/src/gallium/auxiliary/gallivm/llvm_builtins.c @@ -0,0 +1,115 @@ +/*clang --emit-llvm llvm_builtins.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=gallivm_builtins.cpp -f -for=shader -funcname=createGallivmBuiltins*/ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +typedef __attribute__(( ocu_vector_type(4) )) float float4; + +extern float powf(float a, float b); + +inline float approx(float a, float b) +{ + if (b < -128.0f) b = -128.0f; + if (b > 128.0f) b = 128.0f; + if (a < 0) a = 0; + return powf(a, b); +} + +inline float4 lit(float4 tmp) +{ + float4 result; + result.x = 1.0; + result.w = 1.0; + if (tmp.x > 0) { + result.y = tmp.x; + result.z = approx(tmp.y, tmp.w); + } else { + result.y = 0; + result.z = 0; + } + return result; +} + +inline float4 cmp(float4 tmp0, float4 tmp1, float4 tmp2) +{ + float4 result; + + result.x = (tmp0.x < 0.0) ? tmp1.x : tmp2.x; + result.y = (tmp0.y < 0.0) ? tmp1.y : tmp2.y; + result.z = (tmp0.z < 0.0) ? tmp1.z : tmp2.z; + result.w = (tmp0.w < 0.0) ? tmp1.w : tmp2.w; + + return result; +} + +extern float cosf(float val); +extern float sinf(float val); + +inline float4 vcos(float4 val) +{ + float4 result; + printf("VEC IN is %f %f %f %f\n", val.x, val.y, val.z, val.w); + result.x = cosf(val.x); + result.y = cosf(val.x); + result.z = cosf(val.x); + result.w = cosf(val.x); + printf("VEC OUT is %f %f %f %f\n", result.x, result.y, result.z, result.w); + return result; +} + +inline float4 scs(float4 val) +{ + float4 result; + float tmp = val.x; + result.x = cosf(tmp); + result.y = sinf(tmp); + return result; +} + + +inline float4 vsin(float4 val) +{ + float4 result; + float tmp = val.x; + float res = sinf(tmp); + result.x = res; + result.y = res; + result.z = res; + result.w = res; + return result; +} + +inline int kilp(float4 val) +{ + if (val.x < 0 || val.y < 0 || val.z < 0 || val.w < 0) + return 1; + else + return 0; +} diff --git a/src/gallium/auxiliary/gallivm/loweringpass.cpp b/src/gallium/auxiliary/gallivm/loweringpass.cpp new file mode 100644 index 0000000000..556dbec366 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/loweringpass.cpp @@ -0,0 +1,17 @@ +#include "loweringpass.h" + +using namespace llvm; + +char LoweringPass::ID = 0; +RegisterPass X("lowering", "Lowering Pass"); + +LoweringPass::LoweringPass() + : ModulePass((intptr_t)&ID) +{ +} + +bool LoweringPass::runOnModule(Module &m) +{ + llvm::cerr << "Hello: " << m.getModuleIdentifier() << "\n"; + return false; +} diff --git a/src/gallium/auxiliary/gallivm/loweringpass.h b/src/gallium/auxiliary/gallivm/loweringpass.h new file mode 100644 index 0000000000..f62dcf6ba7 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/loweringpass.h @@ -0,0 +1,15 @@ +#ifndef LOWERINGPASS_H +#define LOWERINGPASS_H + +#include "llvm/Pass.h" +#include "llvm/Module.h" + +struct LoweringPass : public llvm::ModulePass +{ + static char ID; + LoweringPass(); + + virtual bool runOnModule(llvm::Module &m); +}; + +#endif diff --git a/src/gallium/auxiliary/gallivm/storage.cpp b/src/gallium/auxiliary/gallivm/storage.cpp new file mode 100644 index 0000000000..c4326de8c5 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/storage.cpp @@ -0,0 +1,364 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +#ifdef MESA_LLVM + +#include "storage.h" + +#include "gallivm_p.h" + +#include "pipe/p_shader_tokens.h" +#include +#include +#include + +#include +#include +#include +#include +#include + +using namespace llvm; + +Storage::Storage(llvm::BasicBlock *block, llvm::Value *input) + : m_block(block), + m_INPUT(input), + m_addrs(32), + m_idx(0) +{ + m_floatVecType = VectorType::get(Type::FloatTy, 4); + m_intVecType = VectorType::get(IntegerType::get(32), 4); + + m_undefFloatVec = UndefValue::get(m_floatVecType); + m_undefIntVec = UndefValue::get(m_intVecType); + m_extSwizzleVec = 0; + + m_numConsts = 0; +} + +//can only build vectors with all members in the [0, 9] range +llvm::Constant *Storage::shuffleMask(int vec) +{ + if (!m_extSwizzleVec) { + std::vector elems; + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); + elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); + m_extSwizzleVec = ConstantVector::get(m_floatVecType, elems); + } + + if (m_intVecs.find(vec) != m_intVecs.end()) { + return m_intVecs[vec]; + } + int origVec = vec; + Constant* const_vec = 0; + if (origVec == 0) { + const_vec = Constant::getNullValue(m_intVecType); + } else { + int x = gallivm_x_swizzle(vec); + int y = gallivm_y_swizzle(vec); + int z = gallivm_z_swizzle(vec); + int w = gallivm_w_swizzle(vec); + std::vector elems; + elems.push_back(constantInt(x)); + elems.push_back(constantInt(y)); + elems.push_back(constantInt(z)); + elems.push_back(constantInt(w)); + const_vec = ConstantVector::get(m_intVecType, elems); + } + + m_intVecs[origVec] = const_vec; + return const_vec; +} + +llvm::ConstantInt *Storage::constantInt(int idx) +{ + if (m_constInts.find(idx) != m_constInts.end()) { + return m_constInts[idx]; + } + ConstantInt *const_int = ConstantInt::get(APInt(32, idx)); + m_constInts[idx] = const_int; + return const_int; +} + +llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) +{ + Value *val = element(InputsArg, idx, indIdx); + LoadInst *load = new LoadInst(val, name("input"), false, m_block); + load->setAlignment(8); + + return load; +} + +llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) +{ + m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; + + Value *elem = element(ConstsArg, idx, indIdx); + LoadInst *load = new LoadInst(elem, name("const"), false, m_block); + load->setAlignment(8); + return load; +} + +llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) +{ + Constant *mask = shuffleMask(shuffle); + ShuffleVectorInst *res = + new ShuffleVectorInst(vec, m_extSwizzleVec, mask, + name("shuffle"), m_block); + return res; +} + + +llvm::Value *Storage::tempElement(int idx, llvm::Value *indIdx) +{ + Value *elem = element(TempsArg, idx, indIdx); + + LoadInst *load = new LoadInst(elem, name("temp"), false, m_block); + load->setAlignment(8); + + return load; +} + +void Storage::setTempElement(int idx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = 0; + if (m_tempWriteMap[idx]) + templ = tempElement(idx); + val = maskWrite(val, mask, templ); + } + Value *elem = element(TempsArg, idx); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); + m_tempWriteMap[idx] = true; +} + +void Storage::setOutputElement(int dstIdx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = 0; + if (m_destWriteMap[dstIdx]) + templ = outputElement(dstIdx); + val = maskWrite(val, mask, templ); + } + + Value *elem = element(DestsArg, dstIdx); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); + m_destWriteMap[dstIdx] = true; +} + +llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ) +{ + llvm::Value *dst = templ; + if (!dst) + dst = Constant::getNullValue(m_floatVecType); + if ((mask & TGSI_WRITEMASK_X)) { + llvm::Value *x = new ExtractElementInst(src, unsigned(0), + name("x"), m_block); + dst = new InsertElementInst(dst, x, unsigned(0), + name("dstx"), m_block); + } + if ((mask & TGSI_WRITEMASK_Y)) { + llvm::Value *y = new ExtractElementInst(src, unsigned(1), + name("y"), m_block); + dst = new InsertElementInst(dst, y, unsigned(1), + name("dsty"), m_block); + } + if ((mask & TGSI_WRITEMASK_Z)) { + llvm::Value *z = new ExtractElementInst(src, unsigned(2), + name("z"), m_block); + dst = new InsertElementInst(dst, z, unsigned(2), + name("dstz"), m_block); + } + if ((mask & TGSI_WRITEMASK_W)) { + llvm::Value *w = new ExtractElementInst(src, unsigned(3), + name("w"), m_block); + dst = new InsertElementInst(dst, w, unsigned(3), + name("dstw"), m_block); + } + return dst; +} + +const char * Storage::name(const char *prefix) +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +int Storage::numConsts() const +{ + return m_numConsts; +} + +llvm::Value * Storage::addrElement(int idx) const +{ + Value *ret = m_addrs[idx]; + if (!ret) + return m_undefFloatVec; + return ret; +} + +void Storage::setAddrElement(int idx, llvm::Value *val, int mask) +{ + if (mask != TGSI_WRITEMASK_XYZW) { + llvm::Value *templ = m_addrs[idx]; + val = maskWrite(val, mask, templ); + } + m_addrs[idx] = val; +} + +llvm::Value * Storage::extractIndex(llvm::Value *vec) +{ + llvm::Value *x = new ExtractElementInst(vec, unsigned(0), + name("x"), m_block); + return new FPToSIInst(x, IntegerType::get(32), name("intidx"), m_block); +} + +void Storage::setCurrentBlock(llvm::BasicBlock *block) +{ + m_block = block; +} + +llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx) +{ + Value *elem = element(DestsArg, idx, indIdx); + LoadInst *load = new LoadInst(elem, name("output"), false, m_block); + load->setAlignment(8); + + return load; +} + +llvm::Value * Storage::inputPtr() const +{ + return m_INPUT; +} + +void Storage::pushArguments(llvm::Value *input) +{ + m_argStack.push(m_INPUT); + + m_INPUT = input; +} + +void Storage::popArguments() +{ + m_INPUT = m_argStack.top(); + m_argStack.pop(); +} + +void Storage::pushTemps() +{ + m_extSwizzleVec = 0; +} + +void Storage::popTemps() +{ +} + +llvm::Value * Storage::immediateElement(int idx) +{ + return m_immediates[idx]; +} + +void Storage::addImmediate(float *val) +{ + std::vector vec(4); + vec[0] = ConstantFP::get(Type::FloatTy, APFloat(val[0])); + vec[1] = ConstantFP::get(Type::FloatTy, APFloat(val[1])); + vec[2] = ConstantFP::get(Type::FloatTy, APFloat(val[2])); + vec[3] = ConstantFP::get(Type::FloatTy, APFloat(val[3])); + m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); +} + + +llvm::Value * Storage::elemPtr(Args arg) +{ + std::vector indices; + indices.push_back(constantInt(0)); + indices.push_back(constantInt(static_cast(arg))); + GetElementPtrInst *getElem = new GetElementPtrInst(m_INPUT, + indices.begin(), + indices.end(), + name("input_ptr"), + m_block); + return new LoadInst(getElem, name("input_field"), false, m_block); +} + +llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx, + llvm::Value *indIdx ) +{ + GetElementPtrInst *getElem = 0; + + if (indIdx) { + getElem = new GetElementPtrInst(ptr, + BinaryOperator::create(Instruction::Add, + indIdx, + constantInt(idx), + name("add"), + m_block), + name("field"), + m_block); + } else { + getElem = new GetElementPtrInst(ptr, + constantInt(idx), + name("field"), + m_block); + } + return getElem; +} + +llvm::Value * Storage::element(Args arg, int idx, llvm::Value *indIdx ) +{ + Value *val = elemPtr(arg); + return elemIdx(val, idx, indIdx); +} + +void Storage::setKilElement(llvm::Value *val) +{ + std::vector indices; + indices.push_back(constantInt(0)); + indices.push_back(constantInt(static_cast(KilArg))); + GetElementPtrInst *elem = new GetElementPtrInst(m_INPUT, + indices.begin(), + indices.end(), + name("kil_ptr"), + m_block); + StoreInst *st = new StoreInst(val, elem, false, m_block); + st->setAlignment(8); +} + +#endif //MESA_LLVM + + diff --git a/src/gallium/auxiliary/gallivm/storage.h b/src/gallium/auxiliary/gallivm/storage.h new file mode 100644 index 0000000000..8574f7554e --- /dev/null +++ b/src/gallium/auxiliary/gallivm/storage.h @@ -0,0 +1,133 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ + +#ifndef STORAGE_H +#define STORAGE_H + +#include +#include +#include +#include + +namespace llvm { + class BasicBlock; + class Constant; + class ConstantInt; + class LoadInst; + class Value; + class VectorType; +} + +class Storage +{ +public: + Storage(llvm::BasicBlock *block, + llvm::Value *input); + + llvm::Value *inputPtr() const; + + void setCurrentBlock(llvm::BasicBlock *block); + + llvm::ConstantInt *constantInt(int); + llvm::Constant *shuffleMask(int vec); + llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); + llvm::Value *constElement(int idx, llvm::Value *indIdx =0); + llvm::Value *outputElement(int idx, llvm::Value *indIdx =0); + llvm::Value *tempElement(int idx, llvm::Value *indIdx =0); + llvm::Value *immediateElement(int idx); + + void setOutputElement(int dstIdx, llvm::Value *val, int mask); + void setTempElement(int idx, llvm::Value *val, int mask); + + llvm::Value *addrElement(int idx) const; + void setAddrElement(int idx, llvm::Value *val, int mask); + + void setKilElement(llvm::Value *val); + + llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); + + llvm::Value *extractIndex(llvm::Value *vec); + + int numConsts() const; + + void pushArguments(llvm::Value *input); + void popArguments(); + void pushTemps(); + void popTemps(); + + void addImmediate(float *val); + +private: + llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); + const char *name(const char *prefix); + + enum Args { + DestsArg = 0, + InputsArg = 1, + TempsArg = 2, + ConstsArg = 3, + KilArg = 4 + }; + llvm::Value *elemPtr(Args arg); + llvm::Value *elemIdx(llvm::Value *ptr, int idx, + llvm::Value *indIdx = 0); + llvm::Value *element(Args arg, int idx, llvm::Value *indIdx = 0); + +private: + llvm::BasicBlock *m_block; + llvm::Value *m_INPUT; + + std::map m_constInts; + std::map m_intVecs; + std::vector m_addrs; + std::vector m_immediates; + + llvm::VectorType *m_floatVecType; + llvm::VectorType *m_intVecType; + + char m_name[32]; + int m_idx; + + int m_numConsts; + + std::map m_destWriteMap; + std::map m_tempWriteMap; + + llvm::Value *m_undefFloatVec; + llvm::Value *m_undefIntVec; + llvm::Value *m_extSwizzleVec; + + std::stack m_argStack; + std::stack > m_tempStack; +}; + +#endif diff --git a/src/gallium/auxiliary/gallivm/storagesoa.cpp b/src/gallium/auxiliary/gallivm/storagesoa.cpp new file mode 100644 index 0000000000..ed0674a96f --- /dev/null +++ b/src/gallium/auxiliary/gallivm/storagesoa.cpp @@ -0,0 +1,389 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "storagesoa.h" + +#include "gallivm_p.h" + +#include "pipe/p_shader_tokens.h" +#include "pipe/p_debug.h" + +#include +#include +#include + +#include +#include +#include +#include +#include + +using namespace llvm; + + +StorageSoa::StorageSoa(llvm::BasicBlock *block, + llvm::Value *input, + llvm::Value *output, + llvm::Value *consts, + llvm::Value *temps) + : m_block(block), + m_input(input), + m_output(output), + m_consts(consts), + m_temps(temps), + m_immediates(0), + m_idx(0) +{ +} + +void StorageSoa::addImmediate(float *vec) +{ + std::vector vals(4); + vals[0] = vec[0]; + vals[1] = vec[1]; + vals[2] = vec[2]; + vals[3] = vec[3]; + m_immediatesToFlush.push_back(vals); +} + +void StorageSoa::declareImmediates() +{ + if (m_immediatesToFlush.empty()) + return; + + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorChannels = ArrayType::get(vectorType, 4); + ArrayType *arrayType = ArrayType::get(vectorChannels, m_immediatesToFlush.size()); + + m_immediates = new GlobalVariable( + /*Type=*/arrayType, + /*isConstant=*/false, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name("immediates"), + currentModule()); + + std::vector arrayVals; + for (unsigned int i = 0; i < m_immediatesToFlush.size(); ++i) { + std::vector vec = m_immediatesToFlush[i]; + std::vector vals(4); + std::vector channelArray; + + vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0]; + llvm::Constant *xChannel = createConstGlobalVector(vals); + + vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1]; + llvm::Constant *yChannel = createConstGlobalVector(vals); + + vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2]; + llvm::Constant *zChannel = createConstGlobalVector(vals); + + vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3]; + llvm::Constant *wChannel = createConstGlobalVector(vals); + channelArray.push_back(xChannel); + channelArray.push_back(yChannel); + channelArray.push_back(zChannel); + channelArray.push_back(wChannel); + Constant *constChannels = ConstantArray::get(vectorChannels, + channelArray); + arrayVals.push_back(constChannels); + } + Constant *constArray = ConstantArray::get(arrayType, arrayVals); + m_immediates->setInitializer(constArray); + + m_immediatesToFlush.clear(); +} + +llvm::Value *StorageSoa::addrElement(int idx) const +{ + std::map::const_iterator itr = m_addresses.find(idx); + if (itr == m_addresses.end()) { + debug_printf("Trying to access invalid shader 'address'\n"); + return 0; + } + llvm::Value * res = (*itr).second; + + res = new LoadInst(res, name("addr"), false, m_block); + + return res; +} + +std::vector StorageSoa::inputElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_input, idx, 0); + res[1] = element(m_input, idx, 1); + res[2] = element(m_input, idx, 2); + res[3] = element(m_input, idx, 3); + + return res; +} + +std::vector StorageSoa::constElement(llvm::Value *idx) +{ + std::vector res(4); + llvm::Value *xChannel, *yChannel, *zChannel, *wChannel; + + xChannel = elementPointer(m_consts, idx, 0); + yChannel = elementPointer(m_consts, idx, 1); + zChannel = elementPointer(m_consts, idx, 2); + wChannel = elementPointer(m_consts, idx, 3); + + res[0] = alignedArrayLoad(xChannel); + res[1] = alignedArrayLoad(yChannel); + res[2] = alignedArrayLoad(zChannel); + res[3] = alignedArrayLoad(wChannel); + + return res; +} + +std::vector StorageSoa::outputElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_output, idx, 0); + res[1] = element(m_output, idx, 1); + res[2] = element(m_output, idx, 2); + res[3] = element(m_output, idx, 3); + + return res; +} + +std::vector StorageSoa::tempElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_temps, idx, 0); + res[1] = element(m_temps, idx, 1); + res[2] = element(m_temps, idx, 2); + res[3] = element(m_temps, idx, 3); + + return res; +} + +std::vector StorageSoa::immediateElement(llvm::Value *idx) +{ + std::vector res(4); + + res[0] = element(m_immediates, idx, 0); + res[1] = element(m_immediates, idx, 1); + res[2] = element(m_immediates, idx, 2); + res[3] = element(m_immediates, idx, 3); + + return res; +} + +llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index, + int channel) const +{ + std::vector indices; + if (m_immediates == ptr) + indices.push_back(constantInt(0)); + indices.push_back(index); + indices.push_back(constantInt(channel)); + + GetElementPtrInst *getElem = new GetElementPtrInst(ptr, + indices.begin(), + indices.end(), + name("ptr"), + m_block); + return getElem; +} + +llvm::Value * StorageSoa::element(llvm::Value *ptr, llvm::Value *index, + int channel) const +{ + llvm::Value *res = elementPointer(ptr, index, channel); + LoadInst *load = new LoadInst(res, name("element"), false, m_block); + //load->setAlignment(8); + return load; +} + +const char * StorageSoa::name(const char *prefix) const +{ + ++m_idx; + snprintf(m_name, 32, "%s%d", prefix, m_idx); + return m_name; +} + +llvm::ConstantInt * StorageSoa::constantInt(int idx) const +{ + if (m_constInts.find(idx) != m_constInts.end()) { + return m_constInts[idx]; + } + ConstantInt *constInt = ConstantInt::get(APInt(32, idx)); + m_constInts[idx] = constInt; + return constInt; +} + +llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + PointerType *vectorPtr = PointerType::get(vectorType, 0); + + CastInst *cast = new BitCastInst(val, vectorPtr, name("toVector"), m_block); + LoadInst *load = new LoadInst(cast, name("alignLoad"), false, m_block); + load->setAlignment(8); + return load; +} + +llvm::Module * StorageSoa::currentModule() const +{ + if (!m_block || !m_block->getParent()) + return 0; + + return m_block->getParent()->getParent(); +} + +llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector &vec) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + std::vector immValues; + ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0])); + ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1])); + ConstantFP *constz = ConstantFP::get(Type::FloatTy, APFloat(vec[2])); + ConstantFP *constw = ConstantFP::get(Type::FloatTy, APFloat(vec[3])); + immValues.push_back(constx); + immValues.push_back(consty); + immValues.push_back(constz); + immValues.push_back(constw); + Constant *constVector = ConstantVector::get(vectorType, immValues); + + return constVector; +} + +std::vector StorageSoa::load(Argument type, int idx, int swizzle, + llvm::Value *indIdx) +{ + std::vector val(4); + + //if we have an indirect index, always use that + // if not use the integer offset to create one + llvm::Value *realIndex = 0; + if (indIdx) + realIndex = indIdx; + else + realIndex = constantInt(idx); + debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx); + + switch(type) { + case Input: + val = inputElement(realIndex); + break; + case Output: + val = outputElement(realIndex); + break; + case Temp: + val = tempElement(realIndex); + break; + case Const: + val = constElement(realIndex); + break; + case Immediate: + val = immediateElement(realIndex); + break; + case Address: + debug_printf("Address not handled in the load phase!\n"); + assert(0); + break; + } + if (!gallivm_is_swizzle(swizzle)) + return val; + + std::vector res(4); + + res[0] = val[gallivm_x_swizzle(swizzle)]; + res[1] = val[gallivm_y_swizzle(swizzle)]; + res[2] = val[gallivm_z_swizzle(swizzle)]; + res[3] = val[gallivm_w_swizzle(swizzle)]; + return res; +} + +void StorageSoa::store(Argument type, int idx, const std::vector &val, + int mask) +{ + llvm::Value *out = 0; + switch(type) { + case Output: + out = m_output; + break; + case Temp: + out = m_temps; + break; + case Input: + out = m_input; + break; + case Address: { + llvm::Value *addr = m_addresses[idx]; + if (!addr) { + addAddress(idx); + addr = m_addresses[idx]; + assert(addr); + } + new StoreInst(val[0], addr, false, m_block); + return; + break; + } + default: + debug_printf("Can't save output of this type: %d !\n", type); + assert(0); + break; + } + llvm::Value *realIndex = constantInt(idx); + if ((mask & TGSI_WRITEMASK_X)) { + llvm::Value *xChannel = elementPointer(out, realIndex, 0); + new StoreInst(val[0], xChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_Y)) { + llvm::Value *yChannel = elementPointer(out, realIndex, 1); + new StoreInst(val[1], yChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_Z)) { + llvm::Value *zChannel = elementPointer(out, realIndex, 2); + new StoreInst(val[2], zChannel, false, m_block); + } + if ((mask & TGSI_WRITEMASK_W)) { + llvm::Value *wChannel = elementPointer(out, realIndex, 3); + new StoreInst(val[3], wChannel, false, m_block); + } +} + +void StorageSoa::addAddress(int idx) +{ + GlobalVariable *val = new GlobalVariable( + /*Type=*/IntegerType::get(32), + /*isConstant=*/false, + /*Linkage=*/GlobalValue::ExternalLinkage, + /*Initializer=*/0, // has initializer, specified below + /*Name=*/name("address"), + currentModule()); + val->setInitializer(Constant::getNullValue(IntegerType::get(32))); + + debug_printf("adding to %d\n", idx); + m_addresses[idx] = val; +} diff --git a/src/gallium/auxiliary/gallivm/storagesoa.h b/src/gallium/auxiliary/gallivm/storagesoa.h new file mode 100644 index 0000000000..6443351f27 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/storagesoa.h @@ -0,0 +1,111 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef STORAGESOA_H +#define STORAGESOA_H + +#include +#include +#include + +namespace llvm { + class BasicBlock; + class Constant; + class ConstantInt; + class GlobalVariable; + class LoadInst; + class Value; + class VectorType; + class Module; +} + +class StorageSoa +{ +public: + enum Argument { + Input, + Output, + Temp, + Const, + Immediate, + Address + }; +public: + StorageSoa(llvm::BasicBlock *block, + llvm::Value *input, + llvm::Value *output, + llvm::Value *consts, + llvm::Value *temps); + + + std::vector load(Argument type, int idx, int swizzle, + llvm::Value *indIdx =0); + void store(Argument type, int idx, const std::vector &val, + int mask); + + void addImmediate(float *vec); + void declareImmediates(); + + void addAddress(int idx); + + llvm::Value * addrElement(int idx) const; + + llvm::ConstantInt *constantInt(int) const; +private: + llvm::Value *elementPointer(llvm::Value *ptr, llvm::Value *indIdx, + int channel) const; + llvm::Value *element(llvm::Value *ptr, llvm::Value *idx, + int channel) const; + const char *name(const char *prefix) const; + llvm::Value *alignedArrayLoad(llvm::Value *val); + llvm::Module *currentModule() const; + llvm::Constant *createConstGlobalVector(const std::vector &vec); + + std::vector inputElement(llvm::Value *indIdx); + std::vector constElement(llvm::Value *indIdx); + std::vector outputElement(llvm::Value *indIdx); + std::vector tempElement(llvm::Value *indIdx); + std::vector immediateElement(llvm::Value *indIdx); +private: + llvm::BasicBlock *m_block; + + llvm::Value *m_input; + llvm::Value *m_output; + llvm::Value *m_consts; + llvm::Value *m_temps; + llvm::GlobalVariable *m_immediates; + + std::map m_addresses; + + std::vector > m_immediatesToFlush; + + mutable std::map m_constInts; + mutable char m_name[32]; + mutable int m_idx; +}; + +#endif diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp new file mode 100644 index 0000000000..2cb4acce32 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -0,0 +1,1221 @@ +#include "tgsitollvm.h" + +#include "gallivm.h" +#include "gallivm_p.h" + +#include "storage.h" +#include "instructions.h" +#include "storagesoa.h" +#include "instructionssoa.h" + +#include "pipe/p_shader_tokens.h" + +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_util.h" +#include "tgsi/util/tgsi_build.h" +#include "tgsi/util/tgsi_dump.h" + + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + + +#include +#include +#include + +using namespace llvm; + +static inline FunctionType *vertexShaderFunctionType() +{ + //Function takes three arguments, + // the calling code has to make sure the types it will + // pass are castable to the following: + // [4 x <4 x float>] inputs, + // [4 x <4 x float>] output, + // [4 x [4 x float]] consts, + // [4 x <4 x float>] temps + + std::vector funcArgs; + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorArray = ArrayType::get(vectorType, 4); + PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0); + + ArrayType *floatArray = ArrayType::get(Type::FloatTy, 4); + ArrayType *constsArray = ArrayType::get(floatArray, 4); + PointerType *constsArrayPtr = PointerType::get(constsArray, 0); + + funcArgs.push_back(vectorArrayPtr);//inputs + funcArgs.push_back(vectorArrayPtr);//output + funcArgs.push_back(constsArrayPtr);//consts + funcArgs.push_back(vectorArrayPtr);//temps + + FunctionType *functionType = FunctionType::get( + /*Result=*/Type::VoidTy, + /*Params=*/funcArgs, + /*isVarArg=*/false); + + return functionType; +} + +static inline void +add_interpolator(struct gallivm_ir *ir, + struct gallivm_interpolate *interp) +{ + ir->interpolators[ir->num_interp] = *interp; + ++ir->num_interp; +} + +static void +translate_declaration(struct gallivm_ir *prog, + llvm::Module *module, + Storage *storage, + struct tgsi_full_declaration *decl, + struct tgsi_full_declaration *fd) +{ + if (decl->Declaration.File == TGSI_FILE_INPUT) { + unsigned first, last, mask; + uint interp_method; + + assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); + + first = decl->u.DeclarationRange.First; + last = decl->u.DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + /* Do not touch WPOS.xy */ + if (first == 0) { + mask &= ~TGSI_WRITEMASK_XY; + if (mask == TGSI_WRITEMASK_NONE) { + first++; + if (first > last) { + return; + } + } + } + + interp_method = decl->Interpolation.Interpolate; + + if (mask == TGSI_WRITEMASK_XYZW) { + unsigned i, j; + + for (i = first; i <= last; i++) { + for (j = 0; j < NUM_CHANNELS; j++) { + //interp( mach, i, j ); + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } else { + unsigned i, j; + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + struct gallivm_interpolate interp; + interp.type = interp_method; + interp.attrib = i; + interp.chan = j; + add_interpolator(prog, &interp); + } + } + } + } + } +} + +static void +translate_declarationir(struct gallivm_ir *, + llvm::Module *, + StorageSoa *storage, + struct tgsi_full_declaration *decl, + struct tgsi_full_declaration *) +{ + if (decl->Declaration.File == TGSI_FILE_ADDRESS) { + int idx = decl->u.DeclarationRange.First; + storage->addAddress(idx); + } +} + +static void +translate_immediate(Storage *storage, + struct tgsi_full_immediate *imm) +{ + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert(0); + } + } + storage->addImmediate(vec); +} + + +static void +translate_immediateir(StorageSoa *storage, + struct tgsi_full_immediate *imm) +{ + float vec[4]; + int i; + for (i = 0; i < imm->Immediate.Size - 1; ++i) { + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + vec[i] = imm->u.ImmediateFloat32[i].Float; + break; + default: + assert(0); + } + } + storage->addImmediate(vec); +} + +static inline int +swizzleInt(struct tgsi_full_src_register *src) +{ + int swizzle = 0; + int start = 1000; + + for (int k = 0; k < 4; ++k) { + swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; + start /= 10; + } + return swizzle; +} + +static inline llvm::Value * +swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, + Storage *storage) +{ + int swizzle = swizzleInt(src); + + if (gallivm_is_swizzle(swizzle)) { + /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ + val = storage->shuffleVector(val, swizzle); + } + return val; +} + +static void +translate_instruction(llvm::Module *module, + Storage *storage, + Instructions *instr, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi, + unsigned instno) +{ + llvm::Value *inputs[4]; + inputs[0] = 0; + inputs[1] = 0; + inputs[2] = 0; + inputs[3] = 0; + + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + llvm::Value *val = 0; + llvm::Value *indIdx = 0; + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + indIdx = storage->extractIndex(indIdx); + } + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->constElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->inputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->tempElement(src->SrcRegister.Index); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->outputElement(src->SrcRegister.Index, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->immediateElement(src->SrcRegister.Index); + } else { + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); + return; + } + + inputs[i] = swizzleVector(val, src, storage); + } + + /*if (inputs[0]) + instr->printVector(inputs[0]); + if (inputs[1]) + instr->printVector(inputs[1]);*/ + llvm::Value *out = 0; + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); + } + break; + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } + break; + case TGSI_OPCODE_LIT: { + out = instr->lit(inputs[0]); + } + break; + case TGSI_OPCODE_RCP: { + out = instr->rcp(inputs[0]); + } + break; + case TGSI_OPCODE_RSQ: { + out = instr->rsq(inputs[0]); + } + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP3: { + out = instr->dp3(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP4: { + out = instr->dp4(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DST: { + out = instr->dst(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MIN: { + out = instr->min(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAX: { + out = instr->max(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SLT: { + out = instr->slt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SGE: { + out = instr->sge(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SUB: { + out = instr->sub(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_LERP: { + out = instr->lerp(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: { + out = instr->frc(inputs[0]); + } + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: { + out = instr->floor(inputs[0]); + } + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: { + out = instr->ex2(inputs[0]); + } + break; + case TGSI_OPCODE_LOGBASE2: { + out = instr->lg2(inputs[0]); + } + break; + case TGSI_OPCODE_POWER: { + out = instr->pow(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_CROSSPRODUCT: { + out = instr->cross(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: { + out = instr->abs(inputs[0]); + } + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: { + out = instr->dph(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_COS: { + out = instr->cos(inputs[0]); + } + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: { + out = instr->kilp(inputs[0]); + storage->setKilElement(out); + return; + } + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: { + out = instr->sgt(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_SIN: { + out = instr->sin(inputs[0]); + } + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: { + instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr()); + return; + } + break; + case TGSI_OPCODE_RET: { + instr->end(); + return; + } + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: { + out = instr->cmp(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SCS: { + out = instr->scs(inputs[0]); + } + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: { + instr->brk(); + return; + } + break; + case TGSI_OPCODE_IF: { + instr->ifop(inputs[0]); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: { + instr->elseop(); + storage->setCurrentBlock(instr->currentBlock()); + return; //only state update + } + break; + case TGSI_OPCODE_ENDIF: { + instr->endif(); + storage->setCurrentBlock(instr->currentBlock()); + return; //just update the state + } + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: { + out = instr->trunc(inputs[0]); + } + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: { + instr->beginLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_BGNSUB: { + instr->bgnSub(instno); + storage->setCurrentBlock(instr->currentBlock()); + storage->pushTemps(); + return; + } + break; + case TGSI_OPCODE_ENDLOOP2: { + instr->endLoop(); + storage->setCurrentBlock(instr->currentBlock()); + return; + } + break; + case TGSI_OPCODE_ENDSUB: { + instr->endSub(); + storage->setCurrentBlock(instr->currentBlock()); + storage->popArguments(); + storage->popTemps(); + return; + } + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + instr->end(); + return; + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + if (!out) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } + + /* # not sure if we need this */ + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + /*TXT( "_SAT" );*/ + break; + case TGSI_SAT_MINUS_PLUS_ONE: + /*TXT( "_SAT[-1,1]" );*/ + break; + default: + assert( 0 ); + } + + /* store results */ + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); + } + } +} + + +static void +translate_instructionir(llvm::Module *module, + StorageSoa *storage, + InstructionsSoa *instr, + struct tgsi_full_instruction *inst, + struct tgsi_full_instruction *fi, + unsigned instno) +{ + std::vector< std::vector > inputs(inst->Instruction.NumSrcRegs); + + for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + std::vector val; + llvm::Value *indIdx = 0; + int swizzle = swizzleInt(src); + + if (src->SrcRegister.Indirect) { + indIdx = storage->addrElement(src->SrcRegisterInd.Index); + } + if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { + val = storage->load(StorageSoa::Const, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { + val = storage->load(StorageSoa::Input, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { + val = storage->load(StorageSoa::Temp, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { + val = storage->load(StorageSoa::Output, + src->SrcRegister.Index, swizzle, indIdx); + } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { + val = storage->load(StorageSoa::Immediate, + src->SrcRegister.Index, swizzle, indIdx); + } else { + fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); + return; + } + + inputs[i] = val; + } + + std::vector out(4); + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: { + out = instr->arl(inputs[0]); + } + break; + case TGSI_OPCODE_MOV: { + out = inputs[0]; + } + break; + case TGSI_OPCODE_LIT: { + } + break; + case TGSI_OPCODE_RCP: { + } + break; + case TGSI_OPCODE_RSQ: { + } + break; + case TGSI_OPCODE_EXP: + break; + case TGSI_OPCODE_LOG: + break; + case TGSI_OPCODE_MUL: { + out = instr->mul(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_ADD: { + out = instr->add(inputs[0], inputs[1]); + } + break; + case TGSI_OPCODE_DP3: { + } + break; + case TGSI_OPCODE_DP4: { + } + break; + case TGSI_OPCODE_DST: { + } + break; + case TGSI_OPCODE_MIN: { + } + break; + case TGSI_OPCODE_MAX: { + } + break; + case TGSI_OPCODE_SLT: { + } + break; + case TGSI_OPCODE_SGE: { + } + break; + case TGSI_OPCODE_MAD: { + out = instr->madd(inputs[0], inputs[1], inputs[2]); + } + break; + case TGSI_OPCODE_SUB: { + } + break; + case TGSI_OPCODE_LERP: { + } + break; + case TGSI_OPCODE_CND: + break; + case TGSI_OPCODE_CND0: + break; + case TGSI_OPCODE_DOT2ADD: + break; + case TGSI_OPCODE_INDEX: + break; + case TGSI_OPCODE_NEGATE: + break; + case TGSI_OPCODE_FRAC: { + } + break; + case TGSI_OPCODE_CLAMP: + break; + case TGSI_OPCODE_FLOOR: { + } + break; + case TGSI_OPCODE_ROUND: + break; + case TGSI_OPCODE_EXPBASE2: { + } + break; + case TGSI_OPCODE_LOGBASE2: { + } + break; + case TGSI_OPCODE_POWER: { + } + break; + case TGSI_OPCODE_CROSSPRODUCT: { + } + break; + case TGSI_OPCODE_MULTIPLYMATRIX: + break; + case TGSI_OPCODE_ABS: { + } + break; + case TGSI_OPCODE_RCC: + break; + case TGSI_OPCODE_DPH: { + } + break; + case TGSI_OPCODE_COS: { + } + break; + case TGSI_OPCODE_DDX: + break; + case TGSI_OPCODE_DDY: + break; + case TGSI_OPCODE_KILP: { + } + break; + case TGSI_OPCODE_PK2H: + break; + case TGSI_OPCODE_PK2US: + break; + case TGSI_OPCODE_PK4B: + break; + case TGSI_OPCODE_PK4UB: + break; + case TGSI_OPCODE_RFL: + break; + case TGSI_OPCODE_SEQ: + break; + case TGSI_OPCODE_SFL: + break; + case TGSI_OPCODE_SGT: { + } + break; + case TGSI_OPCODE_SIN: { + } + break; + case TGSI_OPCODE_SLE: + break; + case TGSI_OPCODE_SNE: + break; + case TGSI_OPCODE_STR: + break; + case TGSI_OPCODE_TEX: + break; + case TGSI_OPCODE_TXD: + break; + case TGSI_OPCODE_UP2H: + break; + case TGSI_OPCODE_UP2US: + break; + case TGSI_OPCODE_UP4B: + break; + case TGSI_OPCODE_UP4UB: + break; + case TGSI_OPCODE_X2D: + break; + case TGSI_OPCODE_ARA: + break; + case TGSI_OPCODE_ARR: + break; + case TGSI_OPCODE_BRA: + break; + case TGSI_OPCODE_CAL: { + } + break; + case TGSI_OPCODE_RET: { + } + break; + case TGSI_OPCODE_SSG: + break; + case TGSI_OPCODE_CMP: { + } + break; + case TGSI_OPCODE_SCS: { + } + break; + case TGSI_OPCODE_TXB: + break; + case TGSI_OPCODE_NRM: + break; + case TGSI_OPCODE_DIV: + break; + case TGSI_OPCODE_DP2: + break; + case TGSI_OPCODE_TXL: + break; + case TGSI_OPCODE_BRK: { + } + break; + case TGSI_OPCODE_IF: { + } + break; + case TGSI_OPCODE_LOOP: + break; + case TGSI_OPCODE_REP: + break; + case TGSI_OPCODE_ELSE: { + } + break; + case TGSI_OPCODE_ENDIF: { + } + break; + case TGSI_OPCODE_ENDLOOP: + break; + case TGSI_OPCODE_ENDREP: + break; + case TGSI_OPCODE_PUSHA: + break; + case TGSI_OPCODE_POPA: + break; + case TGSI_OPCODE_CEIL: + break; + case TGSI_OPCODE_I2F: + break; + case TGSI_OPCODE_NOT: + break; + case TGSI_OPCODE_TRUNC: { + } + break; + case TGSI_OPCODE_SHL: + break; + case TGSI_OPCODE_SHR: + break; + case TGSI_OPCODE_AND: + break; + case TGSI_OPCODE_OR: + break; + case TGSI_OPCODE_MOD: + break; + case TGSI_OPCODE_XOR: + break; + case TGSI_OPCODE_SAD: + break; + case TGSI_OPCODE_TXF: + break; + case TGSI_OPCODE_TXQ: + break; + case TGSI_OPCODE_CONT: + break; + case TGSI_OPCODE_EMIT: + break; + case TGSI_OPCODE_ENDPRIM: + break; + case TGSI_OPCODE_BGNLOOP2: { + } + break; + case TGSI_OPCODE_BGNSUB: { + } + break; + case TGSI_OPCODE_ENDLOOP2: { + } + break; + case TGSI_OPCODE_ENDSUB: { + } + break; + case TGSI_OPCODE_NOISE1: + break; + case TGSI_OPCODE_NOISE2: + break; + case TGSI_OPCODE_NOISE3: + break; + case TGSI_OPCODE_NOISE4: + break; + case TGSI_OPCODE_NOP: + break; + case TGSI_OPCODE_TEXBEM: + break; + case TGSI_OPCODE_TEXBEML: + break; + case TGSI_OPCODE_TEXREG2AR: + break; + case TGSI_OPCODE_TEXM3X2PAD: + break; + case TGSI_OPCODE_TEXM3X2TEX: + break; + case TGSI_OPCODE_TEXM3X3PAD: + break; + case TGSI_OPCODE_TEXM3X3TEX: + break; + case TGSI_OPCODE_TEXM3X3SPEC: + break; + case TGSI_OPCODE_TEXM3X3VSPEC: + break; + case TGSI_OPCODE_TEXREG2GB: + break; + case TGSI_OPCODE_TEXREG2RGB: + break; + case TGSI_OPCODE_TEXDP3TEX: + break; + case TGSI_OPCODE_TEXDP3: + break; + case TGSI_OPCODE_TEXM3X3: + break; + case TGSI_OPCODE_TEXM3X2DEPTH: + break; + case TGSI_OPCODE_TEXDEPTH: + break; + case TGSI_OPCODE_BEM: + break; + case TGSI_OPCODE_M4X3: + break; + case TGSI_OPCODE_M3X4: + break; + case TGSI_OPCODE_M3X3: + break; + case TGSI_OPCODE_M3X2: + break; + case TGSI_OPCODE_NRM4: + break; + case TGSI_OPCODE_CALLNZ: + break; + case TGSI_OPCODE_IFC: + break; + case TGSI_OPCODE_BREAKC: + break; + case TGSI_OPCODE_KIL: + break; + case TGSI_OPCODE_END: + instr->end(); + return; + break; + default: + fprintf(stderr, "ERROR: Unknown opcode %d\n", + inst->Instruction.Opcode); + assert(0); + break; + } + + if (!out[0]) { + fprintf(stderr, "ERROR: unsupported opcode %d\n", + inst->Instruction.Opcode); + assert(!"Unsupported opcode"); + } + + /* store results */ + for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { + storage->store(StorageSoa::Output, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { + storage->store(StorageSoa::Temp, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { + storage->store(StorageSoa::Address, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); + } else { + fprintf(stderr, "ERROR: unsupported LLVM destination!"); + assert(!"wrong destination"); + } + } +} + +llvm::Module * +tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) +{ + llvm::Module *mod = new Module("shader"); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned instno = 0; + Function* shader = mod->getFunction("execute_shader"); + std::ostringstream stream; + if (ir->type == GALLIVM_VS) { + stream << "vs_shader"; + } else { + stream << "fs_shader"; + } + stream << ir->id; + std::string func_name = stream.str(); + shader->setName(func_name.c_str()); + + Function::arg_iterator args = shader->arg_begin(); + Value *ptr_INPUT = args++; + ptr_INPUT->setName("input"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + + tgsi_parse_init(&parse, tokens); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + Storage storage(label_entry, ptr_INPUT); + Instructions instr(mod, shader, label_entry, &storage); + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declaration(ir, mod, &storage, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediate(&storage, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + translate_instruction(mod, &storage, &instr, + &parse.FullToken.FullInstruction, + &fi, instno); + ++instno; + break; + + default: + assert(0); + } + } + + tgsi_parse_free(&parse); + + ir->num_consts = storage.numConsts(); + return mod; +} + +llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, + const struct tgsi_token *tokens) +{ + llvm::Module *mod = new Module("shader"); + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + unsigned instno = 0; + std::ostringstream stream; + if (ir->type == GALLIVM_VS) { + stream << "vs_shader"; + } else { + stream << "fs_shader"; + } + //stream << ir->id; + std::string func_name = stream.str(); + Function *shader = llvm::cast(mod->getOrInsertFunction( + func_name.c_str(), + vertexShaderFunctionType())); + + Function::arg_iterator args = shader->arg_begin(); + Value *input = args++; + input->setName("inputs"); + Value *output = args++; + output->setName("outputs"); + Value *consts = args++; + consts->setName("consts"); + Value *temps = args++; + temps->setName("temps"); + + BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + + tgsi_parse_init(&parse, tokens); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + StorageSoa storage(label_entry, input, output, consts, temps); + InstructionsSoa instr(mod, shader, label_entry, &storage); + + while(!tgsi_parse_end_of_tokens(&parse)) { + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_DECLARATION: + translate_declarationir(ir, mod, &storage, + &parse.FullToken.FullDeclaration, + &fd); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + translate_immediateir(&storage, + &parse.FullToken.FullImmediate); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + storage.declareImmediates(); + translate_instructionir(mod, &storage, &instr, + &parse.FullToken.FullInstruction, + &fi, instno); + ++instno; + break; + + default: + assert(0); + } + } + + tgsi_parse_free(&parse); + + return mod; +} diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.h b/src/gallium/auxiliary/gallivm/tgsitollvm.h new file mode 100644 index 0000000000..7ada04d629 --- /dev/null +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.h @@ -0,0 +1,20 @@ +#ifndef TGSITOLLVM_H +#define TGSITOLLVM_H + + +namespace llvm { + class Module; +} + +struct gallivm_ir; +struct tgsi_token; + + +llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, + const struct tgsi_token *tokens); + + +llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, + const struct tgsi_token *tokens); + +#endif diff --git a/src/gallium/auxiliary/llvm/Makefile b/src/gallium/auxiliary/llvm/Makefile deleted file mode 100644 index 39fac6ea4a..0000000000 --- a/src/gallium/auxiliary/llvm/Makefile +++ /dev/null @@ -1,85 +0,0 @@ -# -*-makefile-*- -TOP = ../../../.. -include $(TOP)/configs/current - -LIBNAME = gallivm - - -GALLIVM_SOURCES = \ - gallivm.cpp \ - gallivm_cpu.cpp \ - instructions.cpp \ - loweringpass.cpp \ - tgsitollvm.cpp \ - storage.cpp \ - storagesoa.cpp \ - instructionssoa.cpp - -INC_SOURCES = gallivm_builtins.cpp - -CPP_SOURCES = \ - $(GALLIVM_SOURCES) - -C_SOURCES = -ASM_SOURCES = - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(CPP_SOURCES:.cpp=.o) \ - $(ASM_SOURCES:.S=.o) - -### Include directories -INCLUDES = \ - -I. \ - -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/src/gallium/auxiliary \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/mesa \ - -I$(TOP)/include - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(LLVM_CFLAGS) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.cpp.o: - $(CXX) -c $(INCLUDES) $(LLVM_CXXFLAGS) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -##### TARGETS ##### - -default:: depend symlinks $(LIBNAME) - - -$(LIBNAME): $(OBJECTS) Makefile - $(TOP)/bin/mklib -o $@ -static $(OBJECTS) - - -depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ - $(ASM_SOURCES) $(INC_SOURCES) 2> /dev/null - - -gallivm_builtins.cpp: llvm_builtins.c - clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean: - -rm -f *.o */*.o *~ *.so *~ server/*.o - -rm -f depend depend.bak - -rm -f gallivm_builtins.cpp - -symlinks: - - -include depend diff --git a/src/gallium/auxiliary/llvm/gallivm.cpp b/src/gallium/auxiliary/llvm/gallivm.cpp deleted file mode 100644 index d14bb3b99a..0000000000 --- a/src/gallium/auxiliary/llvm/gallivm.cpp +++ /dev/null @@ -1,327 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "gallivm.h" -#include "gallivm_p.h" - -#include "instructions.h" -#include "loweringpass.h" -#include "storage.h" -#include "tgsitollvm.h" - -#include "pipe/p_context.h" -#include "pipe/p_shader_tokens.h" - -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_dump.h" - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -static int GLOBAL_ID = 0; - -using namespace llvm; - -static inline -void AddStandardCompilePasses(PassManager &PM) -{ - PM.add(new LoweringPass()); - PM.add(createVerifierPass()); // Verify that input is correct - - PM.add(createLowerSetJmpPass()); // Lower llvm.setjmp/.longjmp - - //PM.add(createStripSymbolsPass(true)); - - PM.add(createRaiseAllocationsPass()); // call %malloc -> malloc inst - PM.add(createCFGSimplificationPass()); // Clean up disgusting code - PM.add(createPromoteMemoryToRegisterPass());// Kill useless allocas - PM.add(createGlobalOptimizerPass()); // Optimize out global vars - PM.add(createGlobalDCEPass()); // Remove unused fns and globs - PM.add(createIPConstantPropagationPass());// IP Constant Propagation - PM.add(createDeadArgEliminationPass()); // Dead argument elimination - PM.add(createInstructionCombiningPass()); // Clean up after IPCP & DAE - PM.add(createCFGSimplificationPass()); // Clean up after IPCP & DAE - - PM.add(createPruneEHPass()); // Remove dead EH info - - PM.add(createFunctionInliningPass()); // Inline small functions - PM.add(createArgumentPromotionPass()); // Scalarize uninlined fn args - - PM.add(createTailDuplicationPass()); // Simplify cfg by copying code - PM.add(createInstructionCombiningPass()); // Cleanup for scalarrepl. - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createScalarReplAggregatesPass()); // Break up aggregate allocas - PM.add(createInstructionCombiningPass()); // Combine silly seq's - PM.add(createCondPropagationPass()); // Propagate conditionals - - PM.add(createTailCallEliminationPass()); // Eliminate tail calls - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createReassociatePass()); // Reassociate expressions - PM.add(createLoopRotatePass()); - PM.add(createLICMPass()); // Hoist loop invariants - PM.add(createLoopUnswitchPass()); // Unswitch loops. - PM.add(createLoopIndexSplitPass()); // Index split loops. - PM.add(createInstructionCombiningPass()); // Clean up after LICM/reassoc - PM.add(createIndVarSimplifyPass()); // Canonicalize indvars - PM.add(createLoopUnrollPass()); // Unroll small loops - PM.add(createInstructionCombiningPass()); // Clean up after the unroller - PM.add(createGVNPass()); // Remove redundancies - PM.add(createSCCPPass()); // Constant prop with SCCP - - // Run instcombine after redundancy elimination to exploit opportunities - // opened up by them. - PM.add(createInstructionCombiningPass()); - PM.add(createCondPropagationPass()); // Propagate conditionals - - PM.add(createDeadStoreEliminationPass()); // Delete dead stores - PM.add(createAggressiveDCEPass()); // SSA based 'Aggressive DCE' - PM.add(createCFGSimplificationPass()); // Merge & remove BBs - PM.add(createSimplifyLibCallsPass()); // Library Call Optimizations - PM.add(createDeadTypeEliminationPass()); // Eliminate dead types - PM.add(createConstantMergePass()); // Merge dup global constants -} - -void gallivm_prog_delete(struct gallivm_prog *prog) -{ - delete prog->module; - prog->module = 0; - prog->function = 0; - free(prog); -} - -static inline void -constant_interpolation(float (*inputs)[16][4], - const struct tgsi_interp_coef *coefs, - unsigned attrib, - unsigned chan) -{ - unsigned i; - - for (i = 0; i < QUAD_SIZE; ++i) { - inputs[i][attrib][chan] = coefs[attrib].a0[chan]; - } -} - -static inline void -linear_interpolation(float (*inputs)[16][4], - const struct tgsi_interp_coef *coefs, - unsigned attrib, - unsigned chan) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = inputs[i][0][0]; - const float y = inputs[i][0][1]; - - inputs[i][attrib][chan] = - coefs[attrib].a0[chan] + - coefs[attrib].dadx[chan] * x + - coefs[attrib].dady[chan] * y; - } -} - -static inline void -perspective_interpolation(float (*inputs)[16][4], - const struct tgsi_interp_coef *coefs, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - const float x = inputs[i][0][0]; - const float y = inputs[i][0][1]; - /* WPOS.w here is really 1/w */ - const float w = 1.0f / inputs[i][0][3]; - assert(inputs[i][0][3] != 0.0); - - inputs[i][attrib][chan] = - (coefs[attrib].a0[chan] + - coefs[attrib].dadx[chan] * x + - coefs[attrib].dady[chan] * y) * w; - } -} - -void gallivm_ir_dump(struct gallivm_ir *ir, const char *file_prefix) -{ - if (!ir || !ir->module) - return; - - if (file_prefix) { - std::ostringstream stream; - stream << file_prefix; - stream << ir->id; - stream << ".ll"; - std::string name = stream.str(); - std::ofstream out(name.c_str()); - if (!out) { - std::cerr<<"Can't open file : "<module); - out.close(); - } else { - const llvm::Module::FunctionListType &funcs = ir->module->getFunctionList(); - llvm::Module::FunctionListType::const_iterator itr; - std::cout<<"; ---------- Start shader "<id<id<num_interp; ++i) { - const gallivm_interpolate &interp = prog->interpolators[i]; - switch (interp.type) { - case TGSI_INTERPOLATE_CONSTANT: - constant_interpolation(inputs, coef, interp.attrib, interp.chan); - break; - - case TGSI_INTERPOLATE_LINEAR: - linear_interpolation(inputs, coef, interp.attrib, interp.chan); - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - perspective_interpolation(inputs, coef, interp.attrib, interp.chan); - break; - - default: - assert( 0 ); - } - } -} - - -struct gallivm_ir * gallivm_ir_new(enum gallivm_shader_type type) -{ - struct gallivm_ir *ir = - (struct gallivm_ir *)calloc(1, sizeof(struct gallivm_ir)); - ++GLOBAL_ID; - ir->id = GLOBAL_ID; - ir->type = type; - - return ir; -} - -void gallivm_ir_set_layout(struct gallivm_ir *ir, - enum gallivm_vector_layout layout) -{ - ir->layout = layout; -} - -void gallivm_ir_set_components(struct gallivm_ir *ir, int num) -{ - ir->num_components = num; -} - -void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, - const struct tgsi_token *tokens) -{ - std::cout << "Creating llvm from: " <module = mod; - gallivm_ir_dump(ir, 0); -} - -void gallivm_ir_delete(struct gallivm_ir *ir) -{ - delete ir->module; - free(ir); -} - -struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) -{ - struct gallivm_prog *prog = - (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); - llvm::Module *mod = llvm::CloneModule(ir->module); - prog->num_consts = ir->num_consts; - memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators)); - prog->num_interp = ir->num_interp; - - /* Run optimization passes over it */ - PassManager passes; - passes.add(new TargetData(mod)); - AddStandardCompilePasses(passes); - passes.run(*mod); - prog->module = mod; - - std::cout << "After optimizations:"<dump(); - - return prog; -} - -#endif /* MESA_LLVM */ diff --git a/src/gallium/auxiliary/llvm/gallivm.h b/src/gallium/auxiliary/llvm/gallivm.h deleted file mode 100644 index 92da4bca7f..0000000000 --- a/src/gallium/auxiliary/llvm/gallivm.h +++ /dev/null @@ -1,103 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef GALLIVM_H -#define GALLIVM_H - -#if defined __cplusplus -extern "C" { -#endif - -#include "pipe/p_state.h" - -#ifdef MESA_LLVM - -struct tgsi_token; - -struct gallivm_ir; -struct gallivm_prog; -struct gallivm_cpu_engine; -struct tgsi_interp_coef; -struct tgsi_sampler; -struct tgsi_exec_vector; - -enum gallivm_shader_type { - GALLIVM_VS, - GALLIVM_FS -}; - -enum gallivm_vector_layout { - GALLIVM_AOS, - GALLIVM_SOA -}; - -struct gallivm_ir *gallivm_ir_new(enum gallivm_shader_type type); -void gallivm_ir_set_layout(struct gallivm_ir *ir, - enum gallivm_vector_layout layout); -void gallivm_ir_set_components(struct gallivm_ir *ir, int num); -void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, - const struct tgsi_token *tokens); -void gallivm_ir_delete(struct gallivm_ir *ir); - - -struct gallivm_prog *gallivm_ir_compile(struct gallivm_ir *ir); - -void gallivm_prog_inputs_interpolate(struct gallivm_prog *prog, - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - const struct tgsi_interp_coef *coefs); -void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); - - -struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); -struct gallivm_cpu_engine *gallivm_global_cpu_engine(); -int gallivm_cpu_vs_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - struct tgsi_exec_vector *temps); -int gallivm_cpu_fs_exec(struct gallivm_prog *prog, - float x, float y, - float (*dests)[PIPE_MAX_SHADER_INPUTS][4], - float (*inputs)[PIPE_MAX_SHADER_INPUTS][4], - float (*consts)[4], - struct tgsi_sampler *samplers); -void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *ee, struct gallivm_prog *prog); -void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); - - -#endif /* MESA_LLVM */ - -#if defined __cplusplus -} // extern "C" -#endif - -#endif diff --git a/src/gallium/auxiliary/llvm/gallivm_builtins.cpp b/src/gallium/auxiliary/llvm/gallivm_builtins.cpp deleted file mode 100644 index 1796f0a177..0000000000 --- a/src/gallium/auxiliary/llvm/gallivm_builtins.cpp +++ /dev/null @@ -1,567 +0,0 @@ -// Generated by llvm2cpp - DO NOT MODIFY! - - -Module* createGallivmBuiltins(Module *mod) { - -mod->setModuleIdentifier("shader"); - -// Type Definitions -ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 25); - -PointerType* PointerTy_1 = PointerType::get(ArrayTy_0, 0); - -std::vectorFuncTy_2_args; -FuncTy_2_args.push_back(Type::FloatTy); -FuncTy_2_args.push_back(Type::FloatTy); -FunctionType* FuncTy_2 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_2_args, - /*isVarArg=*/false); - -PointerType* PointerTy_3 = PointerType::get(FuncTy_2, 0); - -VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); - -std::vectorFuncTy_5_args; -FuncTy_5_args.push_back(VectorTy_4); -FunctionType* FuncTy_5 = FunctionType::get( - /*Result=*/VectorTy_4, - /*Params=*/FuncTy_5_args, - /*isVarArg=*/false); - -std::vectorFuncTy_6_args; -FuncTy_6_args.push_back(VectorTy_4); -FuncTy_6_args.push_back(VectorTy_4); -FuncTy_6_args.push_back(VectorTy_4); -FunctionType* FuncTy_6 = FunctionType::get( - /*Result=*/VectorTy_4, - /*Params=*/FuncTy_6_args, - /*isVarArg=*/false); - -VectorType* VectorTy_7 = VectorType::get(IntegerType::get(32), 4); - -std::vectorFuncTy_9_args; -FunctionType* FuncTy_9 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_9_args, - /*isVarArg=*/true); - -PointerType* PointerTy_8 = PointerType::get(FuncTy_9, 0); - -PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8), 0); - -std::vectorFuncTy_12_args; -FuncTy_12_args.push_back(Type::FloatTy); -FunctionType* FuncTy_12 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_12_args, - /*isVarArg=*/false); - -PointerType* PointerTy_11 = PointerType::get(FuncTy_12, 0); - -std::vectorFuncTy_13_args; -FuncTy_13_args.push_back(VectorTy_4); -FunctionType* FuncTy_13 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_13_args, - /*isVarArg=*/false); - - -// Function Declarations - -Function* func_approx = new Function( - /*Type=*/FuncTy_2, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"approx", mod); -func_approx->setCallingConv(CallingConv::C); -const ParamAttrsList *func_approx_PAL = 0; -func_approx->setParamAttrs(func_approx_PAL); - -Function* func_powf = new Function( - /*Type=*/FuncTy_2, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"powf", mod); // (external, no body) -func_powf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_powf_PAL = 0; -func_powf->setParamAttrs(func_powf_PAL); - -Function* func_lit = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"lit", mod); -func_lit->setCallingConv(CallingConv::C); -const ParamAttrsList *func_lit_PAL = 0; -func_lit->setParamAttrs(func_lit_PAL); - -Function* func_cmp = new Function( - /*Type=*/FuncTy_6, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"cmp", mod); -func_cmp->setCallingConv(CallingConv::C); -const ParamAttrsList *func_cmp_PAL = 0; -{ - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_cmp_PAL = ParamAttrsList::get(Attrs); - -} -func_cmp->setParamAttrs(func_cmp_PAL); - -Function* func_vcos = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"vcos", mod); -func_vcos->setCallingConv(CallingConv::C); -const ParamAttrsList *func_vcos_PAL = 0; -func_vcos->setParamAttrs(func_vcos_PAL); - -Function* func_printf = new Function( - /*Type=*/FuncTy_9, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", mod); // (external, no body) -func_printf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_printf_PAL = 0; -func_printf->setParamAttrs(func_printf_PAL); - -Function* func_cosf = new Function( - /*Type=*/FuncTy_12, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"cosf", mod); // (external, no body) -func_cosf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_cosf_PAL = 0; -func_cosf->setParamAttrs(func_cosf_PAL); - -Function* func_scs = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"scs", mod); -func_scs->setCallingConv(CallingConv::C); -const ParamAttrsList *func_scs_PAL = 0; -func_scs->setParamAttrs(func_scs_PAL); - -Function* func_sinf = new Function( - /*Type=*/FuncTy_12, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"sinf", mod); // (external, no body) -func_sinf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_sinf_PAL = 0; -func_sinf->setParamAttrs(func_sinf_PAL); - -Function* func_vsin = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"vsin", mod); -func_vsin->setCallingConv(CallingConv::C); -const ParamAttrsList *func_vsin_PAL = 0; -func_vsin->setParamAttrs(func_vsin_PAL); - -Function* func_kilp = new Function( - /*Type=*/FuncTy_13, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"kilp", mod); -func_kilp->setCallingConv(CallingConv::C); -const ParamAttrsList *func_kilp_PAL = 0; -{ - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_kilp_PAL = ParamAttrsList::get(Attrs); - -} -func_kilp->setParamAttrs(func_kilp_PAL); - -// Global Variable Declarations - - -GlobalVariable* gvar_array__str = new GlobalVariable( -/*Type=*/ArrayTy_0, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str", -mod); - -GlobalVariable* gvar_array__str1 = new GlobalVariable( -/*Type=*/ArrayTy_0, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str1", -mod); - -// Constant Definitions -Constant* const_array_14 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); -Constant* const_array_15 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); -ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); -ConstantFP* const_float_17 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); -Constant* const_float_18 = Constant::getNullValue(Type::FloatTy); -Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); -std::vector const_packed_20_elems; -ConstantFP* const_float_21 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); -const_packed_20_elems.push_back(const_float_21); -UndefValue* const_float_22 = UndefValue::get(Type::FloatTy); -const_packed_20_elems.push_back(const_float_22); -const_packed_20_elems.push_back(const_float_22); -const_packed_20_elems.push_back(const_float_21); -Constant* const_packed_20 = ConstantVector::get(VectorTy_4, const_packed_20_elems); -ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); -ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "3", 10)); -ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "2", 10)); -std::vector const_packed_26_elems; -const_packed_26_elems.push_back(const_float_21); -const_packed_26_elems.push_back(const_float_18); -const_packed_26_elems.push_back(const_float_18); -const_packed_26_elems.push_back(const_float_21); -Constant* const_packed_26 = ConstantVector::get(VectorTy_4, const_packed_26_elems); -Constant* const_double_27 = Constant::getNullValue(Type::DoubleTy); -std::vector const_packed_28_elems; -const_packed_28_elems.push_back(const_int32_19); -ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "5", 10)); -const_packed_28_elems.push_back(const_int32_29); -const_packed_28_elems.push_back(const_int32_25); -const_packed_28_elems.push_back(const_int32_24); -Constant* const_packed_28 = ConstantVector::get(VectorTy_7, const_packed_28_elems); -std::vector const_packed_30_elems; -const_packed_30_elems.push_back(const_int32_19); -const_packed_30_elems.push_back(const_int32_23); -ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "6", 10)); -const_packed_30_elems.push_back(const_int32_31); -const_packed_30_elems.push_back(const_int32_24); -Constant* const_packed_30 = ConstantVector::get(VectorTy_7, const_packed_30_elems); -std::vector const_packed_32_elems; -const_packed_32_elems.push_back(const_int32_19); -const_packed_32_elems.push_back(const_int32_23); -const_packed_32_elems.push_back(const_int32_25); -ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "7", 10)); -const_packed_32_elems.push_back(const_int32_33); -Constant* const_packed_32 = ConstantVector::get(VectorTy_7, const_packed_32_elems); -std::vector const_ptr_34_indices; -const_ptr_34_indices.push_back(const_int32_19); -const_ptr_34_indices.push_back(const_int32_19); -Constant* const_ptr_34 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_34_indices[0], const_ptr_34_indices.size() ); -UndefValue* const_packed_35 = UndefValue::get(VectorTy_4); -std::vector const_ptr_36_indices; -const_ptr_36_indices.push_back(const_int32_19); -const_ptr_36_indices.push_back(const_int32_19); -Constant* const_ptr_36 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_36_indices[0], const_ptr_36_indices.size() ); - -// Global Variable Definitions -gvar_array__str->setInitializer(const_array_14); -gvar_array__str1->setInitializer(const_array_15); - -// Function Definitions - -// Function: approx (func_approx) -{ - Function::arg_iterator args = func_approx->arg_begin(); - Value* float_a = args++; - float_a->setName("a"); - Value* float_b = args++; - float_b->setName("b"); - - BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); - - // Block entry (label_entry) - FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_16, "cmp", label_entry); - SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_16, float_b, "b.addr.0", label_entry); - FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_17, "cmp3", label_entry); - SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_17, float_b_addr_0, "b.addr.1", label_entry); - FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_18, "cmp7", label_entry); - SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_18, float_a, "a.addr.0", label_entry); - std::vector float_call_params; - float_call_params.push_back(float_a_addr_0); - float_call_params.push_back(float_b_addr_1); - CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); - float_call->setCallingConv(CallingConv::C); - float_call->setTailCall(true);const ParamAttrsList *float_call_PAL = 0; - float_call->setParamAttrs(float_call_PAL); - - new ReturnInst(float_call, label_entry); - -} - -// Function: lit (func_lit) -{ - Function::arg_iterator args = func_lit->arg_begin(); - Value* packed_tmp = args++; - packed_tmp->setName("tmp"); - - BasicBlock* label_entry_38 = new BasicBlock("entry",func_lit,0); - BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); - BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); - - // Block entry (label_entry_38) - ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp6", label_entry_38); - FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp6, const_float_18, "cmp", label_entry_38); - new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_39, label_entry_38); - - // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp10 = new InsertElementInst(const_packed_20, float_tmp6, const_int32_23, "tmp10", label_ifthen); - ExtractElementInst* float_tmp12 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp14", label_ifthen); - std::vector float_call_41_params; - float_call_41_params.push_back(float_tmp12); - float_call_41_params.push_back(float_tmp14); - CallInst* float_call_41 = new CallInst(func_approx, float_call_41_params.begin(), float_call_41_params.end(), "call", label_ifthen); - float_call_41->setCallingConv(CallingConv::C); - float_call_41->setTailCall(true);const ParamAttrsList *float_call_41_PAL = 0; - float_call_41->setParamAttrs(float_call_41_PAL); - - InsertElementInst* packed_tmp16 = new InsertElementInst(packed_tmp10, float_call_41, const_int32_25, "tmp16", label_ifthen); - new ReturnInst(packed_tmp16, label_ifthen); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock) - new ReturnInst(const_packed_26, label_UnifiedReturnBlock); - -} - -// Function: cmp (func_cmp) -{ - Function::arg_iterator args = func_cmp->arg_begin(); - Value* packed_tmp0 = args++; - packed_tmp0->setName("tmp0"); - Value* packed_tmp1 = args++; - packed_tmp1->setName("tmp1"); - Value* packed_tmp2 = args++; - packed_tmp2->setName("tmp2"); - - BasicBlock* label_entry_44 = new BasicBlock("entry",func_cmp,0); - BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); - BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); - BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); - BasicBlock* label_cond_cont34 = new BasicBlock("cond.cont34",func_cmp,0); - BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); - BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); - - // Block entry (label_entry_44) - ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_44); - CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_44); - FCmpInst* int1_cmp_45 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_44); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_44); - CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_44); - FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_44); - SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_45, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_44); - new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_44); - - // Block cond.?14 (label_cond__14) - ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_28, "tmp233", label_cond__14); - ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp254", label_cond__14); - CastInst* double_conv265 = new FPExtInst(float_tmp254, Type::DoubleTy, "conv265", label_cond__14); - FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_27, "cmp276", label_cond__14); - new BranchInst(label_cond__28, label_cond_cont34, int1_cmp276, label_cond__14); - - // Block cond.cont20 (label_cond_cont20) - ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_28, "tmp23", label_cond_cont20); - ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp25", label_cond_cont20); - CastInst* double_conv26 = new FPExtInst(float_tmp25, Type::DoubleTy, "conv26", label_cond_cont20); - FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_27, "cmp27", label_cond_cont20); - new BranchInst(label_cond__28, label_cond_cont34, int1_cmp27, label_cond_cont20); - - // Block cond.?28 (label_cond__28) - PHINode* packed_tmp23_reg2mem_0 = new PHINode(VectorTy_4, "tmp23.reg2mem.0", label_cond__28); - packed_tmp23_reg2mem_0->reserveOperandSpace(2); - packed_tmp23_reg2mem_0->addIncoming(packed_tmp233, label_cond__14); - packed_tmp23_reg2mem_0->addIncoming(packed_tmp23, label_cond_cont20); - - ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_30, "tmp378", label_cond__28); - ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp399", label_cond__28); - CastInst* double_conv4010 = new FPExtInst(float_tmp399, Type::DoubleTy, "conv4010", label_cond__28); - FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_27, "cmp4111", label_cond__28); - new BranchInst(label_cond__42, label_cond_cont48, int1_cmp4111, label_cond__28); - - // Block cond.cont34 (label_cond_cont34) - PHINode* packed_tmp23_reg2mem_1 = new PHINode(VectorTy_4, "tmp23.reg2mem.1", label_cond_cont34); - packed_tmp23_reg2mem_1->reserveOperandSpace(2); - packed_tmp23_reg2mem_1->addIncoming(packed_tmp233, label_cond__14); - packed_tmp23_reg2mem_1->addIncoming(packed_tmp23, label_cond_cont20); - - ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_30, "tmp37", label_cond_cont34); - ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp39", label_cond_cont34); - CastInst* double_conv40 = new FPExtInst(float_tmp39, Type::DoubleTy, "conv40", label_cond_cont34); - FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_27, "cmp41", label_cond_cont34); - new BranchInst(label_cond__42, label_cond_cont48, int1_cmp41, label_cond_cont34); - - // Block cond.?42 (label_cond__42) - PHINode* packed_tmp37_reg2mem_0 = new PHINode(VectorTy_4, "tmp37.reg2mem.0", label_cond__42); - packed_tmp37_reg2mem_0->reserveOperandSpace(2); - packed_tmp37_reg2mem_0->addIncoming(packed_tmp378, label_cond__28); - packed_tmp37_reg2mem_0->addIncoming(packed_tmp37, label_cond_cont34); - - ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_32, "tmp5113", label_cond__42); - new ReturnInst(packed_tmp5113, label_cond__42); - - // Block cond.cont48 (label_cond_cont48) - PHINode* packed_tmp37_reg2mem_1 = new PHINode(VectorTy_4, "tmp37.reg2mem.1", label_cond_cont48); - packed_tmp37_reg2mem_1->reserveOperandSpace(2); - packed_tmp37_reg2mem_1->addIncoming(packed_tmp378, label_cond__28); - packed_tmp37_reg2mem_1->addIncoming(packed_tmp37, label_cond_cont34); - - ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_32, "tmp51", label_cond_cont48); - new ReturnInst(packed_tmp51, label_cond_cont48); - -} - -// Function: vcos (func_vcos) -{ - Function::arg_iterator args = func_vcos->arg_begin(); - Value* packed_val = args++; - packed_val->setName("val"); - - BasicBlock* label_entry_53 = new BasicBlock("entry",func_vcos,0); - - // Block entry (label_entry_53) - ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_53); - CastInst* double_conv_54 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_53); - ExtractElementInst* float_tmp3_55 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_53); - CastInst* double_conv4 = new FPExtInst(float_tmp3_55, Type::DoubleTy, "conv4", label_entry_53); - ExtractElementInst* float_tmp6_56 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_53); - CastInst* double_conv7 = new FPExtInst(float_tmp6_56, Type::DoubleTy, "conv7", label_entry_53); - ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_53); - CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_53); - std::vector int32_call_params; - int32_call_params.push_back(const_ptr_34); - int32_call_params.push_back(double_conv_54); - int32_call_params.push_back(double_conv4); - int32_call_params.push_back(double_conv7); - int32_call_params.push_back(double_conv10); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_53); - int32_call->setCallingConv(CallingConv::C); - int32_call->setTailCall(true);const ParamAttrsList *int32_call_PAL = 0; - int32_call->setParamAttrs(int32_call_PAL); - - CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_53); - float_call13->setCallingConv(CallingConv::C); - float_call13->setTailCall(true);const ParamAttrsList *float_call13_PAL = 0; - float_call13->setParamAttrs(float_call13_PAL); - - InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_53); - CallInst* float_call18 = new CallInst(func_cosf, float_tmp1, "call18", label_entry_53); - float_call18->setCallingConv(CallingConv::C); - float_call18->setTailCall(true);const ParamAttrsList *float_call18_PAL = 0; - float_call18->setParamAttrs(float_call18_PAL); - - InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call18, const_int32_23, "tmp20", label_entry_53); - CallInst* float_call23 = new CallInst(func_cosf, float_tmp1, "call23", label_entry_53); - float_call23->setCallingConv(CallingConv::C); - float_call23->setTailCall(true);const ParamAttrsList *float_call23_PAL = 0; - float_call23->setParamAttrs(float_call23_PAL); - - InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call23, const_int32_25, "tmp25", label_entry_53); - CallInst* float_call28 = new CallInst(func_cosf, float_tmp1, "call28", label_entry_53); - float_call28->setCallingConv(CallingConv::C); - float_call28->setTailCall(true);const ParamAttrsList *float_call28_PAL = 0; - float_call28->setParamAttrs(float_call28_PAL); - - InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call28, const_int32_24, "tmp30", label_entry_53); - CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_53); - CastInst* double_conv36 = new FPExtInst(float_call18, Type::DoubleTy, "conv36", label_entry_53); - CastInst* double_conv39 = new FPExtInst(float_call23, Type::DoubleTy, "conv39", label_entry_53); - CastInst* double_conv42 = new FPExtInst(float_call28, Type::DoubleTy, "conv42", label_entry_53); - std::vector int32_call43_params; - int32_call43_params.push_back(const_ptr_36); - int32_call43_params.push_back(double_conv33); - int32_call43_params.push_back(double_conv36); - int32_call43_params.push_back(double_conv39); - int32_call43_params.push_back(double_conv42); - CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_53); - int32_call43->setCallingConv(CallingConv::C); - int32_call43->setTailCall(true);const ParamAttrsList *int32_call43_PAL = 0; - int32_call43->setParamAttrs(int32_call43_PAL); - - new ReturnInst(packed_tmp30, label_entry_53); - -} - -// Function: scs (func_scs) -{ - Function::arg_iterator args = func_scs->arg_begin(); - Value* packed_val_58 = args++; - packed_val_58->setName("val"); - - BasicBlock* label_entry_59 = new BasicBlock("entry",func_scs,0); - - // Block entry (label_entry_59) - ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_58, const_int32_19, "tmp2", label_entry_59); - CallInst* float_call_60 = new CallInst(func_cosf, float_tmp2, "call", label_entry_59); - float_call_60->setCallingConv(CallingConv::C); - float_call_60->setTailCall(true);const ParamAttrsList *float_call_60_PAL = 0; - float_call_60->setParamAttrs(float_call_60_PAL); - - InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_60, const_int32_19, "tmp5", label_entry_59); - CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_59); - float_call7->setCallingConv(CallingConv::C); - float_call7->setTailCall(true);const ParamAttrsList *float_call7_PAL = 0; - float_call7->setParamAttrs(float_call7_PAL); - - InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_59); - new ReturnInst(packed_tmp9, label_entry_59); - -} - -// Function: vsin (func_vsin) -{ - Function::arg_iterator args = func_vsin->arg_begin(); - Value* packed_val_62 = args++; - packed_val_62->setName("val"); - - BasicBlock* label_entry_63 = new BasicBlock("entry",func_vsin,0); - - // Block entry (label_entry_63) - ExtractElementInst* float_tmp2_64 = new ExtractElementInst(packed_val_62, const_int32_19, "tmp2", label_entry_63); - CallInst* float_call_65 = new CallInst(func_sinf, float_tmp2_64, "call", label_entry_63); - float_call_65->setCallingConv(CallingConv::C); - float_call_65->setTailCall(true);const ParamAttrsList *float_call_65_PAL = 0; - float_call_65->setParamAttrs(float_call_65_PAL); - - InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_65, const_int32_19, "tmp6", label_entry_63); - InsertElementInst* packed_tmp9_66 = new InsertElementInst(packed_tmp6, float_call_65, const_int32_23, "tmp9", label_entry_63); - InsertElementInst* packed_tmp12 = new InsertElementInst(packed_tmp9_66, float_call_65, const_int32_25, "tmp12", label_entry_63); - InsertElementInst* packed_tmp15_67 = new InsertElementInst(packed_tmp12, float_call_65, const_int32_24, "tmp15", label_entry_63); - new ReturnInst(packed_tmp15_67, label_entry_63); - -} - -// Function: kilp (func_kilp) -{ - Function::arg_iterator args = func_kilp->arg_begin(); - Value* packed_val_69 = args++; - packed_val_69->setName("val"); - - BasicBlock* label_entry_70 = new BasicBlock("entry",func_kilp,0); - BasicBlock* label_lor_rhs = new BasicBlock("lor_rhs",func_kilp,0); - BasicBlock* label_lor_rhs5 = new BasicBlock("lor_rhs5",func_kilp,0); - BasicBlock* label_lor_rhs11 = new BasicBlock("lor_rhs11",func_kilp,0); - BasicBlock* label_UnifiedReturnBlock_71 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); - - // Block entry (label_entry_70) - ExtractElementInst* float_tmp1_72 = new ExtractElementInst(packed_val_69, const_int32_19, "tmp1", label_entry_70); - FCmpInst* int1_cmp_73 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_72, const_float_18, "cmp", label_entry_70); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs, int1_cmp_73, label_entry_70); - - // Block lor_rhs (label_lor_rhs) - ExtractElementInst* float_tmp3_75 = new ExtractElementInst(packed_val_69, const_int32_23, "tmp3", label_lor_rhs); - FCmpInst* int1_cmp4 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_75, const_float_18, "cmp4", label_lor_rhs); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs5, int1_cmp4, label_lor_rhs); - - // Block lor_rhs5 (label_lor_rhs5) - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_val_69, const_int32_25, "tmp7", label_lor_rhs5); - FCmpInst* int1_cmp8 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp7, const_float_18, "cmp8", label_lor_rhs5); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs11, int1_cmp8, label_lor_rhs5); - - // Block lor_rhs11 (label_lor_rhs11) - ExtractElementInst* float_tmp13 = new ExtractElementInst(packed_val_69, const_int32_24, "tmp13", label_lor_rhs11); - FCmpInst* int1_cmp14 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp13, const_float_18, "cmp14", label_lor_rhs11); - CastInst* int32_retval = new ZExtInst(int1_cmp14, IntegerType::get(32), "retval", label_lor_rhs11); - new ReturnInst(int32_retval, label_lor_rhs11); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock_71) - new ReturnInst(const_int32_23, label_UnifiedReturnBlock_71); - -} - -return mod; - -} diff --git a/src/gallium/auxiliary/llvm/gallivm_cpu.cpp b/src/gallium/auxiliary/llvm/gallivm_cpu.cpp deleted file mode 100644 index 8f9830d0b1..0000000000 --- a/src/gallium/auxiliary/llvm/gallivm_cpu.cpp +++ /dev/null @@ -1,202 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "gallivm.h" -#include "gallivm_p.h" - -#include "instructions.h" -#include "loweringpass.h" -#include "storage.h" -#include "tgsitollvm.h" - -#include "pipe/p_context.h" -#include "pipe/p_shader_tokens.h" - -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_dump.h" - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -struct gallivm_cpu_engine { - llvm::ExecutionEngine *engine; -}; - -static struct gallivm_cpu_engine *CPU = 0; - -typedef int (*fragment_shader_runner)(float x, float y, - float (*dests)[16][4], - float (*inputs)[16][4], - int num_attribs, - float (*consts)[4], int num_consts, - struct tgsi_sampler *samplers); - -int gallivm_cpu_fs_exec(struct gallivm_prog *prog, - float fx, float fy, - float (*dests)[16][4], - float (*inputs)[16][4], - float (*consts)[4], - struct tgsi_sampler *samplers) -{ - fragment_shader_runner runner = reinterpret_cast(prog->function); - assert(runner); - - return runner(fx, fy, dests, inputs, prog->num_interp, - consts, prog->num_consts, - samplers); -} - -static inline llvm::Function *func_for_shader(struct gallivm_prog *prog) -{ - llvm::Module *mod = prog->module; - llvm::Function *func = 0; - - switch (prog->type) { - case GALLIVM_VS: - func = mod->getFunction("vs_shader"); - break; - case GALLIVM_FS: - func = mod->getFunction("fs_shader"); - break; - default: - assert(!"Unknown shader type!"); - break; - } - return func; -} - -/*! - This function creates a CPU based execution engine for the given gallivm_prog. - gallivm_cpu_engine should be used as a singleton throughout the library. Before - executing gallivm_prog_exec one needs to call gallivm_cpu_jit_compile. - The gallivm_prog instance which is being passed to the constructor is being - automatically JIT compiled so one shouldn't call gallivm_cpu_jit_compile - with it again. - */ -struct gallivm_cpu_engine * gallivm_cpu_engine_create(struct gallivm_prog *prog) -{ - struct gallivm_cpu_engine *cpu = (struct gallivm_cpu_engine *) - calloc(1, sizeof(struct gallivm_cpu_engine)); - llvm::Module *mod = static_cast(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = llvm::ExecutionEngine::create(mp, false); - ee->DisableLazyCompilation(); - cpu->engine = ee; - - llvm::Function *func = func_for_shader(prog); - - prog->function = ee->getPointerToFunction(func); - CPU = cpu; - return cpu; -} - - -/*! - This function JIT compiles the given gallivm_prog with the given cpu based execution engine. - The reference to the generated machine code entry point will be stored - in the gallivm_prog program. After executing this function one can call gallivm_prog_exec - in order to execute the gallivm_prog on the CPU. - */ -void gallivm_cpu_jit_compile(struct gallivm_cpu_engine *cpu, struct gallivm_prog *prog) -{ - llvm::Module *mod = static_cast(prog->module); - llvm::ExistingModuleProvider *mp = new llvm::ExistingModuleProvider(mod); - llvm::ExecutionEngine *ee = cpu->engine; - assert(ee); - /*FIXME : remove */ - ee->DisableLazyCompilation(); - ee->addModuleProvider(mp); - - llvm::Function *func = func_for_shader(prog); - prog->function = ee->getPointerToFunction(func); -} - -void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *cpu) -{ - free(cpu); -} - -struct gallivm_cpu_engine * gallivm_global_cpu_engine() -{ - return CPU; -} - - -typedef void (*vertex_shader_runner)(void *ainputs, - void *dests, - float (*aconsts)[4], - void *temps); - - -/*! - This function is used to execute the gallivm_prog in software. Before calling - this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile - function. - */ -int gallivm_cpu_vs_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - struct tgsi_exec_vector *temps) -{ - vertex_shader_runner runner = reinterpret_cast(prog->function); - assert(runner); - /*FIXME*/ - runner(inputs, dests, consts, temps); - - return 0; -} - -#endif diff --git a/src/gallium/auxiliary/llvm/gallivm_p.h b/src/gallium/auxiliary/llvm/gallivm_p.h deleted file mode 100644 index cfe7b1901b..0000000000 --- a/src/gallium/auxiliary/llvm/gallivm_p.h +++ /dev/null @@ -1,110 +0,0 @@ -#ifndef GALLIVM_P_H -#define GALLIVM_P_H - -#ifdef MESA_LLVM - -#include "gallivm.h" -#include "pipe/p_shader_tokens.h" -#include "pipe/p_compiler.h" - -namespace llvm { - class Module; -} - -#if defined __cplusplus -extern "C" { -#endif - -enum gallivm_shader_type; -enum gallivm_vector_layout; - -struct gallivm_interpolate { - int attrib; - int chan; - int type; -}; - -struct gallivm_ir { - llvm::Module *module; - int id; - enum gallivm_shader_type type; - enum gallivm_vector_layout layout; - int num_components; - int num_consts; - - //FIXME: this might not be enough for some shaders - struct gallivm_interpolate interpolators[32*4]; - int num_interp; -}; - -struct gallivm_prog { - llvm::Module *module; - void *function; - - int id; - enum gallivm_shader_type type; - - int num_consts; - - //FIXME: this might not be enough for some shaders - struct gallivm_interpolate interpolators[32*4]; - int num_interp; -}; - -static INLINE void gallivm_swizzle_components(int swizzle, - int *xc, int *yc, - int *zc, int *wc) -{ - int x = swizzle / 1000; swizzle -= x * 1000; - int y = swizzle / 100; swizzle -= y * 100; - int z = swizzle / 10; swizzle -= z * 10; - int w = swizzle; - - if (xc) *xc = x; - if (yc) *yc = y; - if (zc) *zc = z; - if (wc) *wc = w; -} - -static INLINE boolean gallivm_is_swizzle(int swizzle) -{ - const int NO_SWIZZLE = TGSI_SWIZZLE_X * 1000 + TGSI_SWIZZLE_Y * 100 + - TGSI_SWIZZLE_Z * 10 + TGSI_SWIZZLE_W; - return swizzle != NO_SWIZZLE; -} - -static INLINE int gallivm_x_swizzle(int swizzle) -{ - int x; - gallivm_swizzle_components(swizzle, &x, 0, 0, 0); - return x; -} - -static INLINE int gallivm_y_swizzle(int swizzle) -{ - int y; - gallivm_swizzle_components(swizzle, 0, &y, 0, 0); - return y; -} - -static INLINE int gallivm_z_swizzle(int swizzle) -{ - int z; - gallivm_swizzle_components(swizzle, 0, 0, &z, 0); - return z; -} - -static INLINE int gallivm_w_swizzle(int swizzle) -{ - int w; - gallivm_swizzle_components(swizzle, 0, 0, 0, &w); - return w; -} - -#endif /* MESA_LLVM */ - -#if defined __cplusplus -} // extern "C" -#endif - -#endif diff --git a/src/gallium/auxiliary/llvm/instructions.cpp b/src/gallium/auxiliary/llvm/instructions.cpp deleted file mode 100644 index 55d39fa5f1..0000000000 --- a/src/gallium/auxiliary/llvm/instructions.cpp +++ /dev/null @@ -1,889 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "instructions.h" - -#include "storage.h" - -#include -#include -#include -#include -#include -#include -#include - -#include -#include -#include - -using namespace llvm; - -#include "gallivm_builtins.cpp" - -static inline std::string createFuncName(int label) -{ - std::ostringstream stream; - stream << "function"; - stream << label; - return stream.str(); -} - -Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, - Storage *storage) - : m_mod(mod), m_func(func), m_builder(block), m_idx(0), - m_storage(storage) -{ - m_floatVecType = VectorType::get(Type::FloatTy, 4); - - m_llvmFSqrt = 0; - m_llvmFAbs = 0; - m_llvmPow = 0; - m_llvmFloor = 0; - m_llvmFlog = 0; - m_llvmLit = 0; - m_fmtPtr = 0; - - createGallivmBuiltins(m_mod); -} - -llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) -{ - return m_builder.CreateAdd(in1, in2, name("add")); -} - -llvm::Value * Instructions::madd(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in3) -{ - Value *mulRes = mul(in1, in2); - return add(mulRes, in3); -} - -llvm::Value * Instructions::mul(llvm::Value *in1, llvm::Value *in2) -{ - return m_builder.CreateMul(in1, in2, name("mul")); -} - -const char * Instructions::name(const char *prefix) -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -llvm::Value * Instructions::dp3(llvm::Value *in1, llvm::Value *in2) -{ - Value *mulRes = mul(in1, in2); - Value *x = m_builder.CreateExtractElement(mulRes, - m_storage->constantInt(0), - name("extractx")); - Value *y = m_builder.CreateExtractElement(mulRes, - m_storage->constantInt(1), - name("extracty")); - Value *z = m_builder.CreateExtractElement(mulRes, - m_storage->constantInt(2), - name("extractz")); - Value *xy = m_builder.CreateAdd(x, y,name("xy")); - Value *dot3 = m_builder.CreateAdd(xy, z, name("dot3")); - return vectorFromVals(dot3, dot3, dot3, dot3); -} - -llvm::Value *Instructions::callFSqrt(llvm::Value *val) -{ - if (!m_llvmFSqrt) { - // predeclare the intrinsic - std::vector fsqrtArgs; - fsqrtArgs.push_back(Type::FloatTy); - ParamAttrsList *fsqrtPal = 0; - FunctionType* fsqrtType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/fsqrtArgs, - /*isVarArg=*/false); - m_llvmFSqrt = new Function( - /*Type=*/fsqrtType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"llvm.sqrt.f32", m_mod); - m_llvmFSqrt->setCallingConv(CallingConv::C); - m_llvmFSqrt->setParamAttrs(fsqrtPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFSqrt, val, - name("sqrt")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::rsq(llvm::Value *in1) -{ - Value *x = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("extractx")); - Value *abs = callFAbs(x); - Value *sqrt = callFSqrt(abs); - - Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - sqrt, - name("rsqrt")); - return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt); -} - -llvm::Value * Instructions::vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w) -{ - Constant *const_vec = Constant::getNullValue(m_floatVecType); - Value *res = m_builder.CreateInsertElement(const_vec, x, - m_storage->constantInt(0), - name("vecx")); - res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), - name("vecxy")); - res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), - name("vecxyz")); - if (w) - res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), - name("vecxyzw")); - return res; -} - -llvm::Value *Instructions::callFAbs(llvm::Value *val) -{ - if (!m_llvmFAbs) { - // predeclare the intrinsic - std::vector fabsArgs; - fabsArgs.push_back(Type::FloatTy); - ParamAttrsList *fabsPal = 0; - FunctionType* fabsType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/fabsArgs, - /*isVarArg=*/false); - m_llvmFAbs = new Function( - /*Type=*/fabsType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"fabs", m_mod); - m_llvmFAbs->setCallingConv(CallingConv::C); - m_llvmFAbs->setParamAttrs(fabsPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFAbs, val, - name("fabs")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::lit(llvm::Value *in) -{ - if (!m_llvmLit) { - m_llvmLit = m_mod->getFunction("lit"); - } - CallInst *call = m_builder.CreateCall(m_llvmLit, in, name("litres")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::sub(llvm::Value *in1, llvm::Value *in2) -{ - Value *res = m_builder.CreateSub(in1, in2, name("sub")); - return res; -} - -llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) -{ - if (!m_llvmPow) { - // predeclare the intrinsic - std::vector powArgs; - powArgs.push_back(Type::FloatTy); - powArgs.push_back(Type::FloatTy); - ParamAttrsList *powPal = 0; - FunctionType* powType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/powArgs, - /*isVarArg=*/false); - m_llvmPow = new Function( - /*Type=*/powType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"llvm.pow.f32", m_mod); - m_llvmPow->setCallingConv(CallingConv::C); - m_llvmPow->setParamAttrs(powPal); - } - std::vector params; - params.push_back(val1); - params.push_back(val2); - CallInst *call = m_builder.CreateCall(m_llvmPow, params.begin(), params.end(), - name("pow")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::pow(llvm::Value *in1, llvm::Value *in2) -{ - Value *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - Value *x2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(0), - name("x2")); - llvm::Value *val = callPow(x1, x2); - return vectorFromVals(val, val, val, val); -} - -llvm::Value * Instructions::rcp(llvm::Value *in1) -{ - Value *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - Value *res = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - x1, name("rcp")); - return vectorFromVals(res, res, res, res); -} - -llvm::Value * Instructions::dp4(llvm::Value *in1, llvm::Value *in2) -{ - Value *mulRes = mul(in1, in2); - std::vector vec = extractVector(mulRes); - Value *xy = m_builder.CreateAdd(vec[0], vec[1], name("xy")); - Value *xyz = m_builder.CreateAdd(xy, vec[2], name("xyz")); - Value *dot4 = m_builder.CreateAdd(xyz, vec[3], name("dot4")); - return vectorFromVals(dot4, dot4, dot4, dot4); -} - -llvm::Value * Instructions::dph(llvm::Value *in1, llvm::Value *in2) -{ - Value *mulRes = mul(in1, in2); - std::vector vec1 = extractVector(mulRes); - Value *xy = m_builder.CreateAdd(vec1[0], vec1[1], name("xy")); - Value *xyz = m_builder.CreateAdd(xy, vec1[2], name("xyz")); - Value *dph = m_builder.CreateAdd(xyz, vec1[3], name("dph")); - return vectorFromVals(dph, dph, dph, dph); -} - -llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) -{ - Value *y1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(1), - name("y1")); - Value *z = m_builder.CreateExtractElement(in1, - m_storage->constantInt(2), - name("z")); - Value *y2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(1), - name("y2")); - Value *w = m_builder.CreateExtractElement(in2, - m_storage->constantInt(3), - name("w")); - Value *ry = m_builder.CreateMul(y1, y2, name("tyuy")); - return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), - ry, z, w); -} - -llvm::Value * Instructions::ex2(llvm::Value *in) -{ - llvm::Value *val = callPow(ConstantFP::get(Type::FloatTy, APFloat(2.f)), - m_builder.CreateExtractElement( - in, m_storage->constantInt(0), - name("x1"))); - return vectorFromVals(val, val, val, val); -} - -llvm::Value * Instructions::callFloor(llvm::Value *val) -{ - if (!m_llvmFloor) { - // predeclare the intrinsic - std::vector floorArgs; - floorArgs.push_back(Type::FloatTy); - ParamAttrsList *floorPal = 0; - FunctionType* floorType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/floorArgs, - /*isVarArg=*/false); - m_llvmFloor = new Function( - /*Type=*/floorType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"floorf", m_mod); - m_llvmFloor->setCallingConv(CallingConv::C); - m_llvmFloor->setParamAttrs(floorPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFloor, val, - name("floorf")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::floor(llvm::Value *in) -{ - std::vector vec = extractVector(in); - return vectorFromVals(callFloor(vec[0]), callFloor(vec[1]), - callFloor(vec[2]), callFloor(vec[3])); -} - -llvm::Value * Instructions::arl(llvm::Value *in) -{ - return floor(in); -} - -llvm::Value * Instructions::frc(llvm::Value *in) -{ - llvm::Value *flr = floor(in); - return sub(in, flr); -} - -llvm::Value * Instructions::callFLog(llvm::Value *val) -{ - if (!m_llvmFlog) { - // predeclare the intrinsic - std::vector flogArgs; - flogArgs.push_back(Type::FloatTy); - ParamAttrsList *flogPal = 0; - FunctionType* flogType = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/flogArgs, - /*isVarArg=*/false); - m_llvmFlog = new Function( - /*Type=*/flogType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"logf", m_mod); - m_llvmFlog->setCallingConv(CallingConv::C); - m_llvmFlog->setParamAttrs(flogPal); - } - CallInst *call = m_builder.CreateCall(m_llvmFlog, val, - name("logf")); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::lg2(llvm::Value *in) -{ - std::vector vec = extractVector(in); - llvm::Value *const_vec = constVector(1.442695f, 1.442695f, - 1.442695f, 1.442695f); - return mul(vectorFromVals(callFLog(vec[0]), callFLog(vec[1]), - callFLog(vec[2]), callFLog(vec[3])), const_vec); -} - -llvm::Value * Instructions::min(llvm::Value *in1, llvm::Value *in2) -{ - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); - Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], - name("selx")); - - Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); - Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], - name("sely")); - - Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); - Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], - name("selz")); - - Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); - Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], - name("selw")); - - return vectorFromVals(selx, sely, selz, selw); -} - -llvm::Value * Instructions::max(llvm::Value *in1, llvm::Value *in2) -{ - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], - name("xcmp")); - Value *selx = m_builder.CreateSelect(xcmp, vec1[0], vec2[0], - name("selx")); - - Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], - name("ycmp")); - Value *sely = m_builder.CreateSelect(ycmp, vec1[1], vec2[1], - name("sely")); - - Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], - name("zcmp")); - Value *selz = m_builder.CreateSelect(zcmp, vec1[2], vec2[2], - name("selz")); - - Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], - name("wcmp")); - Value *selw = m_builder.CreateSelect(wcmp, vec1[3], vec2[3], - name("selw")); - - return vectorFromVals(selx, sely, selz, selw); -} - -void Instructions::printVector(llvm::Value *val) -{ - static const char *frmt = "Vector is [%f, %f, %f, %f]\x0A"; - - if (!m_fmtPtr) { - Constant *format = ConstantArray::get(frmt, true); - ArrayType *arrayTy = ArrayType::get(IntegerType::get(8), strlen(frmt) + 1); - GlobalVariable* globalFormat = new GlobalVariable( - /*Type=*/arrayTy, - /*isConstant=*/true, - /*Linkage=*/GlobalValue::InternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name(".str"), - m_mod); - globalFormat->setInitializer(format); - - Constant* const_int0 = Constant::getNullValue(IntegerType::get(32)); - std::vector const_ptr_21_indices; - const_ptr_21_indices.push_back(const_int0); - const_ptr_21_indices.push_back(const_int0); - m_fmtPtr = ConstantExpr::getGetElementPtr(globalFormat, - &const_ptr_21_indices[0], const_ptr_21_indices.size()); - } - - Function *func_printf = m_mod->getFunction("printf"); - if (!func_printf) - func_printf = declarePrintf(); - assert(func_printf); - std::vector vec = extractVector(val); - Value *dx = m_builder.CreateFPExt(vec[0], Type::DoubleTy, name("dx")); - Value *dy = m_builder.CreateFPExt(vec[1], Type::DoubleTy, name("dy")); - Value *dz = m_builder.CreateFPExt(vec[2], Type::DoubleTy, name("dz")); - Value *dw = m_builder.CreateFPExt(vec[3], Type::DoubleTy, name("dw")); - std::vector params; - params.push_back(m_fmtPtr); - params.push_back(dx); - params.push_back(dy); - params.push_back(dz); - params.push_back(dw); - CallInst *call = m_builder.CreateCall(func_printf, params.begin(), params.end(), - name("printf")); - call->setCallingConv(CallingConv::C); - call->setTailCall(true); -} - -llvm::Function * Instructions::declarePrintf() -{ - std::vector args; - ParamAttrsList *params = 0; - FunctionType* funcTy = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/args, - /*isVarArg=*/true); - Function* func_printf = new Function( - /*Type=*/funcTy, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", m_mod); - func_printf->setCallingConv(CallingConv::C); - func_printf->setParamAttrs(params); - return func_printf; -} - - -llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) -{ - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - Constant *const0f = Constant::getNullValue(Type::FloatTy); - - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - Value *xcmp = m_builder.CreateFCmpOGT(vec1[0], vec2[0], name("xcmp")); - Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - - Value *ycmp = m_builder.CreateFCmpOGT(vec1[1], vec2[1], name("ycmp")); - Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - - Value *zcmp = m_builder.CreateFCmpOGT(vec1[2], vec2[2], name("zcmp")); - Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - - Value *wcmp = m_builder.CreateFCmpOGT(vec1[3], vec2[3], name("wcmp")); - Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); - - return vectorFromVals(x, y, z, w); -} -llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) -{ - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - Constant *const0f = Constant::getNullValue(Type::FloatTy); - - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOGE(vec1[0], vec2[0], name("xcmp")); - Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - - Value *ycmp = m_builder.CreateFCmpOGE(vec1[1], vec2[1], name("ycmp")); - Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - - Value *zcmp = m_builder.CreateFCmpOGE(vec1[2], vec2[2], name("zcmp")); - Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - - Value *wcmp = m_builder.CreateFCmpOGE(vec1[3], vec2[3], name("wcmp")); - Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); - - return vectorFromVals(x, y, z, w); -} - - -llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) -{ - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); - Constant *const0f = Constant::getNullValue(Type::FloatTy); - - std::vector vec1 = extractVector(in1); - std::vector vec2 = extractVector(in2); - - Value *xcmp = m_builder.CreateFCmpOLT(vec1[0], vec2[0], name("xcmp")); - Value *x = m_builder.CreateSelect(xcmp, const1f, const0f, name("xsel")); - - Value *ycmp = m_builder.CreateFCmpOLT(vec1[1], vec2[1], name("ycmp")); - Value *y = m_builder.CreateSelect(ycmp, const1f, const0f, name("ysel")); - - Value *zcmp = m_builder.CreateFCmpOLT(vec1[2], vec2[2], name("zcmp")); - Value *z = m_builder.CreateSelect(zcmp, const1f, const0f, name("zsel")); - - Value *wcmp = m_builder.CreateFCmpOLT(vec1[3], vec2[3], name("wcmp")); - Value *w = m_builder.CreateSelect(wcmp, const1f, const0f, name("wsel")); - - return vectorFromVals(x, y, z, w); -} - -llvm::Value * Instructions::cross(llvm::Value *in1, llvm::Value *in2) -{ - Value *x1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(0), - name("x1")); - Value *y1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(1), - name("y1")); - Value *z1 = m_builder.CreateExtractElement(in1, - m_storage->constantInt(2), - name("z1")); - - Value *x2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(0), - name("x2")); - Value *y2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(1), - name("y2")); - Value *z2 = m_builder.CreateExtractElement(in2, - m_storage->constantInt(2), - name("z2")); - Value *y1z2 = mul(y1, z2); - Value *z1y2 = mul(z1, y2); - - Value *z1x2 = mul(z1, x2); - Value *x1z2 = mul(x1, z2); - - Value *x1y2 = mul(x1, y2); - Value *y1x2 = mul(y1, x2); - - return vectorFromVals(sub(y1z2, z1y2), sub(z1x2, x1z2), sub(x1y2, y1x2)); -} - - -llvm::Value * Instructions::abs(llvm::Value *in) -{ - std::vector vec = extractVector(in); - Value *xabs = callFAbs(vec[0]); - Value *yabs = callFAbs(vec[1]); - Value *zabs = callFAbs(vec[2]); - Value *wabs = callFAbs(vec[3]); - return vectorFromVals(xabs, yabs, zabs, wabs); -} - -void Instructions::ifop(llvm::Value *in) -{ - BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); - BasicBlock *ifend = new BasicBlock(name("ifthenend"), m_func,0); - - //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0); - //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0); - //BasicBlock *wblock = new BasicBlock(name("wblock"), m_func,0); - - Constant *float0 = Constant::getNullValue(Type::FloatTy); - - Value *x = m_builder.CreateExtractElement(in, m_storage->constantInt(0), - name("extractx")); - Value *xcmp = m_builder.CreateFCmpUNE(x, float0, name("xcmp")); - m_builder.CreateCondBr(xcmp, ifthen, ifend); - //m_builder.SetInsertPoint(yblock); - - m_builder.SetInsertPoint(ifthen); - m_ifStack.push(ifend); -} - -llvm::BasicBlock * Instructions::currentBlock() const -{ - return m_builder.GetInsertBlock(); -} - -void Instructions::elseop() -{ - assert(!m_ifStack.empty()); - BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); - m_builder.CreateBr(ifend); - m_builder.SetInsertPoint(m_ifStack.top()); - currentBlock()->setName(name("ifelse")); - m_ifStack.pop(); - m_ifStack.push(ifend); -} - -void Instructions::endif() -{ - assert(!m_ifStack.empty()); - m_builder.CreateBr(m_ifStack.top()); - m_builder.SetInsertPoint(m_ifStack.top()); - m_ifStack.pop(); -} - -llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in3) -{ - llvm::Value *m = mul(in1, in2); - llvm::Value *vec1 = constVector(1.f, 1.f, 1.f, 1.f); - llvm::Value *s = sub(vec1, in1); - return add(m, mul(s, in3)); -} - -void Instructions::beginLoop() -{ - BasicBlock *begin = new BasicBlock(name("loop"), m_func,0); - BasicBlock *end = new BasicBlock(name("endloop"), m_func,0); - - m_builder.CreateBr(begin); - Loop loop; - loop.begin = begin; - loop.end = end; - m_builder.SetInsertPoint(begin); - m_loopStack.push(loop); -} - -void Instructions::endLoop() -{ - assert(!m_loopStack.empty()); - Loop loop = m_loopStack.top(); - m_builder.CreateBr(loop.begin); - loop.end->moveAfter(currentBlock()); - m_builder.SetInsertPoint(loop.end); - m_loopStack.pop(); -} - -void Instructions::brk() -{ - assert(!m_loopStack.empty()); - BasicBlock *unr = new BasicBlock(name("unreachable"), m_func,0); - m_builder.CreateBr(m_loopStack.top().end); - m_builder.SetInsertPoint(unr); -} - -llvm::Value * Instructions::trunc(llvm::Value *in) -{ - std::vector vec = extractVector(in); - Value *icastx = m_builder.CreateFPToSI(vec[0], IntegerType::get(32), - name("ftoix")); - Value *icasty = m_builder.CreateFPToSI(vec[1], IntegerType::get(32), - name("ftoiy")); - Value *icastz = m_builder.CreateFPToSI(vec[2], IntegerType::get(32), - name("ftoiz")); - Value *icastw = m_builder.CreateFPToSI(vec[3], IntegerType::get(32), - name("ftoiw")); - Value *fx = m_builder.CreateSIToFP(icastx, Type::FloatTy, - name("fx")); - Value *fy = m_builder.CreateSIToFP(icasty, Type::FloatTy, - name("fy")); - Value *fz = m_builder.CreateSIToFP(icastz, Type::FloatTy, - name("fz")); - Value *fw = m_builder.CreateSIToFP(icastw, Type::FloatTy, - name("fw")); - return vectorFromVals(fx, fy, fz, fw); -} - -void Instructions::end() -{ - m_builder.CreateRetVoid(); -} - -void Instructions::cal(int label, llvm::Value *input) -{ - std::vector params; - params.push_back(input); - llvm::Function *func = findFunction(label); - - m_builder.CreateCall(func, params.begin(), params.end()); -} - -llvm::Function * Instructions::declareFunc(int label) -{ - PointerType *vecPtr = PointerType::getUnqual(m_floatVecType); - std::vector args; - args.push_back(vecPtr); - args.push_back(vecPtr); - args.push_back(vecPtr); - args.push_back(vecPtr); - ParamAttrsList *params = 0; - FunctionType *funcType = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/args, - /*isVarArg=*/false); - std::string name = createFuncName(label); - Function *func = new Function( - /*Type=*/funcType, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/name.c_str(), m_mod); - func->setCallingConv(CallingConv::C); - func->setParamAttrs(params); - return func; -} - -void Instructions::bgnSub(unsigned label) -{ - llvm::Function *func = findFunction(label); - - Function::arg_iterator args = func->arg_begin(); - Value *ptr_INPUT = args++; - ptr_INPUT->setName("INPUT"); - m_storage->pushArguments(ptr_INPUT); - - llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); - - m_func = func; - m_builder.SetInsertPoint(entry); -} - -void Instructions::endSub() -{ - m_func = 0; - m_builder.SetInsertPoint(0); -} - -llvm::Function * Instructions::findFunction(int label) -{ - llvm::Function *func = m_functions[label]; - if (!func) { - func = declareFunc(label); - m_functions[label] = func; - } - return func; -} - -llvm::Value * Instructions::constVector(float x, float y, float z, float w) -{ - std::vector vec(4); - vec[0] = ConstantFP::get(Type::FloatTy, APFloat(x)); - vec[1] = ConstantFP::get(Type::FloatTy, APFloat(y)); - vec[2] = ConstantFP::get(Type::FloatTy, APFloat(z)); - vec[3] = ConstantFP::get(Type::FloatTy, APFloat(w)); - return ConstantVector::get(m_floatVecType, vec); -} - - -std::vector Instructions::extractVector(llvm::Value *vec) -{ - std::vector elems(4); - elems[0] = m_builder.CreateExtractElement(vec, m_storage->constantInt(0), - name("x")); - elems[1] = m_builder.CreateExtractElement(vec, m_storage->constantInt(1), - name("y")); - elems[2] = m_builder.CreateExtractElement(vec, m_storage->constantInt(2), - name("z")); - elems[3] = m_builder.CreateExtractElement(vec, m_storage->constantInt(3), - name("w")); - return elems; -} - -llvm::Value * Instructions::cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3) -{ - llvm::Function *func = m_mod->getFunction("cmp"); - assert(func); - - std::vector params; - params.push_back(in1); - params.push_back(in2); - params.push_back(in3); - CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), name("cmpres")); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::cos(llvm::Value *in) -{ -#if 0 - llvm::Function *func = m_mod->getFunction("vcos"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("cosres")); - call->setTailCall(false); - return call; -#else - std::vector elems = extractVector(in); - Function *func = m_mod->getFunction("cosf"); - assert(func); - CallInst *cos = m_builder.CreateCall(func, elems[0], name("cosres")); - cos->setCallingConv(CallingConv::C); - cos->setTailCall(true); - return vectorFromVals(cos, cos, cos, cos); -#endif -} - -llvm::Value * Instructions::scs(llvm::Value *in) -{ - llvm::Function *func = m_mod->getFunction("scs"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("scsres")); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::kilp(llvm::Value *in) -{ - llvm::Function *func = m_mod->getFunction("kilp"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("kilpres")); - call->setTailCall(false); - return call; -} - -llvm::Value * Instructions::sin(llvm::Value *in) -{ - llvm::Function *func = m_mod->getFunction("vsin"); - assert(func); - - CallInst *call = m_builder.CreateCall(func, in, name("sinres")); - call->setTailCall(false); - return call; -} -#endif //MESA_LLVM - - diff --git a/src/gallium/auxiliary/llvm/instructions.h b/src/gallium/auxiliary/llvm/instructions.h deleted file mode 100644 index 9ebc17dd8e..0000000000 --- a/src/gallium/auxiliary/llvm/instructions.h +++ /dev/null @@ -1,152 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef INSTRUCTIONS_H -#define INSTRUCTIONS_H - -#include -#include -#include -#include - -#include -#include - -namespace llvm { - class VectorType; - class Function; -} - -class Storage; - -class Instructions -{ -public: - Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, - Storage *storage); - - llvm::BasicBlock *currentBlock() const; - - llvm::Value *abs(llvm::Value *in1); - llvm::Value *arl(llvm::Value *in1); - llvm::Value *add(llvm::Value *in1, llvm::Value *in2); - void beginLoop(); - void bgnSub(unsigned); - void brk(); - void cal(int label, llvm::Value *input); - llvm::Value *cmp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); - llvm::Value *cos(llvm::Value *in); - llvm::Value *cross(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dp3(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dp4(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dph(llvm::Value *in1, llvm::Value *in2); - llvm::Value *dst(llvm::Value *in1, llvm::Value *in2); - void elseop(); - void endif(); - void endLoop(); - void end(); - void endSub(); - llvm::Value *ex2(llvm::Value *in); - llvm::Value *floor(llvm::Value *in); - llvm::Value *frc(llvm::Value *in); - void ifop(llvm::Value *in); - llvm::Value *kilp(llvm::Value *in); - llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in3); - llvm::Value *lit(llvm::Value *in); - llvm::Value *lg2(llvm::Value *in); - llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in2); - llvm::Value *min(llvm::Value *in1, llvm::Value *in2); - llvm::Value *max(llvm::Value *in1, llvm::Value *in2); - llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); - llvm::Value *pow(llvm::Value *in1, llvm::Value *in2); - llvm::Value *rcp(llvm::Value *in); - llvm::Value *rsq(llvm::Value *in); - llvm::Value *scs(llvm::Value *in); - llvm::Value *sge(llvm::Value *in1, llvm::Value *in2); - llvm::Value *sgt(llvm::Value *in1, llvm::Value *in2); - llvm::Value *sin(llvm::Value *in); - llvm::Value *slt(llvm::Value *in1, llvm::Value *in2); - llvm::Value *sub(llvm::Value *in1, llvm::Value *in2); - llvm::Value *trunc(llvm::Value *in); - - void printVector(llvm::Value *val); -private: - const char *name(const char *prefix); - - llvm::Value *callFAbs(llvm::Value *val); - llvm::Value *callFloor(llvm::Value *val); - llvm::Value *callFSqrt(llvm::Value *val); - llvm::Value *callFLog(llvm::Value *val); - llvm::Value *callPow(llvm::Value *val1, llvm::Value *val2); - - llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w=0); - - llvm::Value *constVector(float x, float y, float z, float w); - - llvm::Function *declarePrintf(); - llvm::Function *declareFunc(int label); - - llvm::Function *findFunction(int label); - - std::vector extractVector(llvm::Value *vec); -private: - llvm::Module *m_mod; - llvm::Function *m_func; - char m_name[32]; - llvm::LLVMFoldingBuilder m_builder; - int m_idx; - - llvm::VectorType *m_floatVecType; - - llvm::Function *m_llvmFSqrt; - llvm::Function *m_llvmFAbs; - llvm::Function *m_llvmPow; - llvm::Function *m_llvmFloor; - llvm::Function *m_llvmFlog; - llvm::Function *m_llvmLit; - - llvm::Constant *m_fmtPtr; - - std::stack m_ifStack; - struct Loop { - llvm::BasicBlock *begin; - llvm::BasicBlock *end; - }; - std::stack m_loopStack; - std::map m_functions; - Storage *m_storage; -}; - -#endif diff --git a/src/gallium/auxiliary/llvm/instructionssoa.cpp b/src/gallium/auxiliary/llvm/instructionssoa.cpp deleted file mode 100644 index a4d5046637..0000000000 --- a/src/gallium/auxiliary/llvm/instructionssoa.cpp +++ /dev/null @@ -1,121 +0,0 @@ -#include "instructionssoa.h" - -#include "storagesoa.h" - -#include - -using namespace llvm; - -InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, - llvm::BasicBlock *block, StorageSoa *storage) - : m_builder(block), - m_storage(storage), - m_idx(0) -{ -} - -const char * InstructionsSoa::name(const char *prefix) const -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -llvm::Value * InstructionsSoa::vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w) -{ - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - Constant *constVector = Constant::getNullValue(vectorType); - Value *res = m_builder.CreateInsertElement(constVector, x, - m_storage->constantInt(0), - name("vecx")); - res = m_builder.CreateInsertElement(res, y, m_storage->constantInt(1), - name("vecxy")); - res = m_builder.CreateInsertElement(res, z, m_storage->constantInt(2), - name("vecxyz")); - if (w) - res = m_builder.CreateInsertElement(res, w, m_storage->constantInt(3), - name("vecxyzw")); - return res; -} - -std::vector InstructionsSoa::arl(const std::vector in) -{ - std::vector res(4); - - //Extract x's - llvm::Value *x1 = m_builder.CreateExtractElement(in[0], - m_storage->constantInt(0), - name("extractX")); - //cast it to an unsigned int - x1 = m_builder.CreateFPToUI(x1, IntegerType::get(32), name("x1IntCast")); - - res[0] = x1;//vectorFromVals(x1, x2, x3, x4); - //only x is valid. the others shouldn't be necessary - /* - res[1] = Constant::getNullValue(m_floatVecType); - res[2] = Constant::getNullValue(m_floatVecType); - res[3] = Constant::getNullValue(m_floatVecType); - */ - - return res; -} - - -std::vector InstructionsSoa::add(const std::vector in1, - const std::vector in2) -{ - std::vector res(4); - - res[0] = m_builder.CreateAdd(in1[0], in2[0], name("addx")); - res[1] = m_builder.CreateAdd(in1[1], in2[1], name("addy")); - res[2] = m_builder.CreateAdd(in1[2], in2[2], name("addz")); - res[3] = m_builder.CreateAdd(in1[3], in2[3], name("addw")); - - return res; -} - -std::vector InstructionsSoa::mul(const std::vector in1, - const std::vector in2) -{ - std::vector res(4); - - res[0] = m_builder.CreateMul(in1[0], in2[0], name("mulx")); - res[1] = m_builder.CreateMul(in1[1], in2[1], name("muly")); - res[2] = m_builder.CreateMul(in1[2], in2[2], name("mulz")); - res[3] = m_builder.CreateMul(in1[3], in2[3], name("mulw")); - - return res; -} - -void InstructionsSoa::end() -{ - m_builder.CreateRetVoid(); -} - -std::vector InstructionsSoa::madd(const std::vector in1, - const std::vector in2, - const std::vector in3) -{ - std::vector res = mul(in1, in2); - return add(res, in3); -} - -std::vector InstructionsSoa::extractVector(llvm::Value *vector) -{ - std::vector res(4); - res[0] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(0), - name("extract1X")); - res[1] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(1), - name("extract2X")); - res[2] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(2), - name("extract3X")); - res[3] = m_builder.CreateExtractElement(vector, - m_storage->constantInt(3), - name("extract4X")); - - return res; -} diff --git a/src/gallium/auxiliary/llvm/instructionssoa.h b/src/gallium/auxiliary/llvm/instructionssoa.h deleted file mode 100644 index 4169dcbb2e..0000000000 --- a/src/gallium/auxiliary/llvm/instructionssoa.h +++ /dev/null @@ -1,74 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INSTRUCTIONSSOA_H -#define INSTRUCTIONSSOA_H - -#include - -#include - -namespace llvm { - class Module; - class Function; - class BasicBlock; - class Value; -} -class StorageSoa; - -class InstructionsSoa -{ -public: - InstructionsSoa(llvm::Module *mod, llvm::Function *func, - llvm::BasicBlock *block, StorageSoa *storage); - - std::vector arl(const std::vector in); - - std::vector add(const std::vector in1, - const std::vector in2); - std::vector madd(const std::vector in1, - const std::vector in2, - const std::vector in3); - std::vector mul(const std::vector in1, - const std::vector in2); - void end(); - - std::vector extractVector(llvm::Value *vector); -private: - const char * name(const char *prefix) const; - llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, - llvm::Value *z, llvm::Value *w); -private: - llvm::LLVMFoldingBuilder m_builder; - StorageSoa *m_storage; -private: - mutable int m_idx; - mutable char m_name[32]; -}; - - -#endif diff --git a/src/gallium/auxiliary/llvm/llvm_builtins.c b/src/gallium/auxiliary/llvm/llvm_builtins.c deleted file mode 100644 index 4f98d754ba..0000000000 --- a/src/gallium/auxiliary/llvm/llvm_builtins.c +++ /dev/null @@ -1,115 +0,0 @@ -/*clang --emit-llvm llvm_builtins.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=gallivm_builtins.cpp -f -for=shader -funcname=createGallivmBuiltins*/ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -typedef __attribute__(( ocu_vector_type(4) )) float float4; - -extern float powf(float a, float b); - -inline float approx(float a, float b) -{ - if (b < -128.0f) b = -128.0f; - if (b > 128.0f) b = 128.0f; - if (a < 0) a = 0; - return powf(a, b); -} - -inline float4 lit(float4 tmp) -{ - float4 result; - result.x = 1.0; - result.w = 1.0; - if (tmp.x > 0) { - result.y = tmp.x; - result.z = approx(tmp.y, tmp.w); - } else { - result.y = 0; - result.z = 0; - } - return result; -} - -inline float4 cmp(float4 tmp0, float4 tmp1, float4 tmp2) -{ - float4 result; - - result.x = (tmp0.x < 0.0) ? tmp1.x : tmp2.x; - result.y = (tmp0.y < 0.0) ? tmp1.y : tmp2.y; - result.z = (tmp0.z < 0.0) ? tmp1.z : tmp2.z; - result.w = (tmp0.w < 0.0) ? tmp1.w : tmp2.w; - - return result; -} - -extern float cosf(float val); -extern float sinf(float val); - -inline float4 vcos(float4 val) -{ - float4 result; - printf("VEC IN is %f %f %f %f\n", val.x, val.y, val.z, val.w); - result.x = cosf(val.x); - result.y = cosf(val.x); - result.z = cosf(val.x); - result.w = cosf(val.x); - printf("VEC OUT is %f %f %f %f\n", result.x, result.y, result.z, result.w); - return result; -} - -inline float4 scs(float4 val) -{ - float4 result; - float tmp = val.x; - result.x = cosf(tmp); - result.y = sinf(tmp); - return result; -} - - -inline float4 vsin(float4 val) -{ - float4 result; - float tmp = val.x; - float res = sinf(tmp); - result.x = res; - result.y = res; - result.z = res; - result.w = res; - return result; -} - -inline int kilp(float4 val) -{ - if (val.x < 0 || val.y < 0 || val.z < 0 || val.w < 0) - return 1; - else - return 0; -} diff --git a/src/gallium/auxiliary/llvm/loweringpass.cpp b/src/gallium/auxiliary/llvm/loweringpass.cpp deleted file mode 100644 index 556dbec366..0000000000 --- a/src/gallium/auxiliary/llvm/loweringpass.cpp +++ /dev/null @@ -1,17 +0,0 @@ -#include "loweringpass.h" - -using namespace llvm; - -char LoweringPass::ID = 0; -RegisterPass X("lowering", "Lowering Pass"); - -LoweringPass::LoweringPass() - : ModulePass((intptr_t)&ID) -{ -} - -bool LoweringPass::runOnModule(Module &m) -{ - llvm::cerr << "Hello: " << m.getModuleIdentifier() << "\n"; - return false; -} diff --git a/src/gallium/auxiliary/llvm/loweringpass.h b/src/gallium/auxiliary/llvm/loweringpass.h deleted file mode 100644 index f62dcf6ba7..0000000000 --- a/src/gallium/auxiliary/llvm/loweringpass.h +++ /dev/null @@ -1,15 +0,0 @@ -#ifndef LOWERINGPASS_H -#define LOWERINGPASS_H - -#include "llvm/Pass.h" -#include "llvm/Module.h" - -struct LoweringPass : public llvm::ModulePass -{ - static char ID; - LoweringPass(); - - virtual bool runOnModule(llvm::Module &m); -}; - -#endif diff --git a/src/gallium/auxiliary/llvm/storage.cpp b/src/gallium/auxiliary/llvm/storage.cpp deleted file mode 100644 index c4326de8c5..0000000000 --- a/src/gallium/auxiliary/llvm/storage.cpp +++ /dev/null @@ -1,364 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ -#ifdef MESA_LLVM - -#include "storage.h" - -#include "gallivm_p.h" - -#include "pipe/p_shader_tokens.h" -#include -#include -#include - -#include -#include -#include -#include -#include - -using namespace llvm; - -Storage::Storage(llvm::BasicBlock *block, llvm::Value *input) - : m_block(block), - m_INPUT(input), - m_addrs(32), - m_idx(0) -{ - m_floatVecType = VectorType::get(Type::FloatTy, 4); - m_intVecType = VectorType::get(IntegerType::get(32), 4); - - m_undefFloatVec = UndefValue::get(m_floatVecType); - m_undefIntVec = UndefValue::get(m_intVecType); - m_extSwizzleVec = 0; - - m_numConsts = 0; -} - -//can only build vectors with all members in the [0, 9] range -llvm::Constant *Storage::shuffleMask(int vec) -{ - if (!m_extSwizzleVec) { - std::vector elems; - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); - m_extSwizzleVec = ConstantVector::get(m_floatVecType, elems); - } - - if (m_intVecs.find(vec) != m_intVecs.end()) { - return m_intVecs[vec]; - } - int origVec = vec; - Constant* const_vec = 0; - if (origVec == 0) { - const_vec = Constant::getNullValue(m_intVecType); - } else { - int x = gallivm_x_swizzle(vec); - int y = gallivm_y_swizzle(vec); - int z = gallivm_z_swizzle(vec); - int w = gallivm_w_swizzle(vec); - std::vector elems; - elems.push_back(constantInt(x)); - elems.push_back(constantInt(y)); - elems.push_back(constantInt(z)); - elems.push_back(constantInt(w)); - const_vec = ConstantVector::get(m_intVecType, elems); - } - - m_intVecs[origVec] = const_vec; - return const_vec; -} - -llvm::ConstantInt *Storage::constantInt(int idx) -{ - if (m_constInts.find(idx) != m_constInts.end()) { - return m_constInts[idx]; - } - ConstantInt *const_int = ConstantInt::get(APInt(32, idx)); - m_constInts[idx] = const_int; - return const_int; -} - -llvm::Value *Storage::inputElement(int idx, llvm::Value *indIdx) -{ - Value *val = element(InputsArg, idx, indIdx); - LoadInst *load = new LoadInst(val, name("input"), false, m_block); - load->setAlignment(8); - - return load; -} - -llvm::Value *Storage::constElement(int idx, llvm::Value *indIdx) -{ - m_numConsts = ((idx + 1) > m_numConsts) ? (idx + 1) : m_numConsts; - - Value *elem = element(ConstsArg, idx, indIdx); - LoadInst *load = new LoadInst(elem, name("const"), false, m_block); - load->setAlignment(8); - return load; -} - -llvm::Value *Storage::shuffleVector(llvm::Value *vec, int shuffle) -{ - Constant *mask = shuffleMask(shuffle); - ShuffleVectorInst *res = - new ShuffleVectorInst(vec, m_extSwizzleVec, mask, - name("shuffle"), m_block); - return res; -} - - -llvm::Value *Storage::tempElement(int idx, llvm::Value *indIdx) -{ - Value *elem = element(TempsArg, idx, indIdx); - - LoadInst *load = new LoadInst(elem, name("temp"), false, m_block); - load->setAlignment(8); - - return load; -} - -void Storage::setTempElement(int idx, llvm::Value *val, int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = 0; - if (m_tempWriteMap[idx]) - templ = tempElement(idx); - val = maskWrite(val, mask, templ); - } - Value *elem = element(TempsArg, idx); - StoreInst *st = new StoreInst(val, elem, false, m_block); - st->setAlignment(8); - m_tempWriteMap[idx] = true; -} - -void Storage::setOutputElement(int dstIdx, llvm::Value *val, int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = 0; - if (m_destWriteMap[dstIdx]) - templ = outputElement(dstIdx); - val = maskWrite(val, mask, templ); - } - - Value *elem = element(DestsArg, dstIdx); - StoreInst *st = new StoreInst(val, elem, false, m_block); - st->setAlignment(8); - m_destWriteMap[dstIdx] = true; -} - -llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ) -{ - llvm::Value *dst = templ; - if (!dst) - dst = Constant::getNullValue(m_floatVecType); - if ((mask & TGSI_WRITEMASK_X)) { - llvm::Value *x = new ExtractElementInst(src, unsigned(0), - name("x"), m_block); - dst = new InsertElementInst(dst, x, unsigned(0), - name("dstx"), m_block); - } - if ((mask & TGSI_WRITEMASK_Y)) { - llvm::Value *y = new ExtractElementInst(src, unsigned(1), - name("y"), m_block); - dst = new InsertElementInst(dst, y, unsigned(1), - name("dsty"), m_block); - } - if ((mask & TGSI_WRITEMASK_Z)) { - llvm::Value *z = new ExtractElementInst(src, unsigned(2), - name("z"), m_block); - dst = new InsertElementInst(dst, z, unsigned(2), - name("dstz"), m_block); - } - if ((mask & TGSI_WRITEMASK_W)) { - llvm::Value *w = new ExtractElementInst(src, unsigned(3), - name("w"), m_block); - dst = new InsertElementInst(dst, w, unsigned(3), - name("dstw"), m_block); - } - return dst; -} - -const char * Storage::name(const char *prefix) -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -int Storage::numConsts() const -{ - return m_numConsts; -} - -llvm::Value * Storage::addrElement(int idx) const -{ - Value *ret = m_addrs[idx]; - if (!ret) - return m_undefFloatVec; - return ret; -} - -void Storage::setAddrElement(int idx, llvm::Value *val, int mask) -{ - if (mask != TGSI_WRITEMASK_XYZW) { - llvm::Value *templ = m_addrs[idx]; - val = maskWrite(val, mask, templ); - } - m_addrs[idx] = val; -} - -llvm::Value * Storage::extractIndex(llvm::Value *vec) -{ - llvm::Value *x = new ExtractElementInst(vec, unsigned(0), - name("x"), m_block); - return new FPToSIInst(x, IntegerType::get(32), name("intidx"), m_block); -} - -void Storage::setCurrentBlock(llvm::BasicBlock *block) -{ - m_block = block; -} - -llvm::Value * Storage::outputElement(int idx, llvm::Value *indIdx) -{ - Value *elem = element(DestsArg, idx, indIdx); - LoadInst *load = new LoadInst(elem, name("output"), false, m_block); - load->setAlignment(8); - - return load; -} - -llvm::Value * Storage::inputPtr() const -{ - return m_INPUT; -} - -void Storage::pushArguments(llvm::Value *input) -{ - m_argStack.push(m_INPUT); - - m_INPUT = input; -} - -void Storage::popArguments() -{ - m_INPUT = m_argStack.top(); - m_argStack.pop(); -} - -void Storage::pushTemps() -{ - m_extSwizzleVec = 0; -} - -void Storage::popTemps() -{ -} - -llvm::Value * Storage::immediateElement(int idx) -{ - return m_immediates[idx]; -} - -void Storage::addImmediate(float *val) -{ - std::vector vec(4); - vec[0] = ConstantFP::get(Type::FloatTy, APFloat(val[0])); - vec[1] = ConstantFP::get(Type::FloatTy, APFloat(val[1])); - vec[2] = ConstantFP::get(Type::FloatTy, APFloat(val[2])); - vec[3] = ConstantFP::get(Type::FloatTy, APFloat(val[3])); - m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); -} - - -llvm::Value * Storage::elemPtr(Args arg) -{ - std::vector indices; - indices.push_back(constantInt(0)); - indices.push_back(constantInt(static_cast(arg))); - GetElementPtrInst *getElem = new GetElementPtrInst(m_INPUT, - indices.begin(), - indices.end(), - name("input_ptr"), - m_block); - return new LoadInst(getElem, name("input_field"), false, m_block); -} - -llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx, - llvm::Value *indIdx ) -{ - GetElementPtrInst *getElem = 0; - - if (indIdx) { - getElem = new GetElementPtrInst(ptr, - BinaryOperator::create(Instruction::Add, - indIdx, - constantInt(idx), - name("add"), - m_block), - name("field"), - m_block); - } else { - getElem = new GetElementPtrInst(ptr, - constantInt(idx), - name("field"), - m_block); - } - return getElem; -} - -llvm::Value * Storage::element(Args arg, int idx, llvm::Value *indIdx ) -{ - Value *val = elemPtr(arg); - return elemIdx(val, idx, indIdx); -} - -void Storage::setKilElement(llvm::Value *val) -{ - std::vector indices; - indices.push_back(constantInt(0)); - indices.push_back(constantInt(static_cast(KilArg))); - GetElementPtrInst *elem = new GetElementPtrInst(m_INPUT, - indices.begin(), - indices.end(), - name("kil_ptr"), - m_block); - StoreInst *st = new StoreInst(val, elem, false, m_block); - st->setAlignment(8); -} - -#endif //MESA_LLVM - - diff --git a/src/gallium/auxiliary/llvm/storage.h b/src/gallium/auxiliary/llvm/storage.h deleted file mode 100644 index 8574f7554e..0000000000 --- a/src/gallium/auxiliary/llvm/storage.h +++ /dev/null @@ -1,133 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin zack@tungstengraphics.com - */ - -#ifndef STORAGE_H -#define STORAGE_H - -#include -#include -#include -#include - -namespace llvm { - class BasicBlock; - class Constant; - class ConstantInt; - class LoadInst; - class Value; - class VectorType; -} - -class Storage -{ -public: - Storage(llvm::BasicBlock *block, - llvm::Value *input); - - llvm::Value *inputPtr() const; - - void setCurrentBlock(llvm::BasicBlock *block); - - llvm::ConstantInt *constantInt(int); - llvm::Constant *shuffleMask(int vec); - llvm::Value *inputElement(int idx, llvm::Value *indIdx =0); - llvm::Value *constElement(int idx, llvm::Value *indIdx =0); - llvm::Value *outputElement(int idx, llvm::Value *indIdx =0); - llvm::Value *tempElement(int idx, llvm::Value *indIdx =0); - llvm::Value *immediateElement(int idx); - - void setOutputElement(int dstIdx, llvm::Value *val, int mask); - void setTempElement(int idx, llvm::Value *val, int mask); - - llvm::Value *addrElement(int idx) const; - void setAddrElement(int idx, llvm::Value *val, int mask); - - void setKilElement(llvm::Value *val); - - llvm::Value *shuffleVector(llvm::Value *vec, int shuffle); - - llvm::Value *extractIndex(llvm::Value *vec); - - int numConsts() const; - - void pushArguments(llvm::Value *input); - void popArguments(); - void pushTemps(); - void popTemps(); - - void addImmediate(float *val); - -private: - llvm::Value *maskWrite(llvm::Value *src, int mask, llvm::Value *templ); - const char *name(const char *prefix); - - enum Args { - DestsArg = 0, - InputsArg = 1, - TempsArg = 2, - ConstsArg = 3, - KilArg = 4 - }; - llvm::Value *elemPtr(Args arg); - llvm::Value *elemIdx(llvm::Value *ptr, int idx, - llvm::Value *indIdx = 0); - llvm::Value *element(Args arg, int idx, llvm::Value *indIdx = 0); - -private: - llvm::BasicBlock *m_block; - llvm::Value *m_INPUT; - - std::map m_constInts; - std::map m_intVecs; - std::vector m_addrs; - std::vector m_immediates; - - llvm::VectorType *m_floatVecType; - llvm::VectorType *m_intVecType; - - char m_name[32]; - int m_idx; - - int m_numConsts; - - std::map m_destWriteMap; - std::map m_tempWriteMap; - - llvm::Value *m_undefFloatVec; - llvm::Value *m_undefIntVec; - llvm::Value *m_extSwizzleVec; - - std::stack m_argStack; - std::stack > m_tempStack; -}; - -#endif diff --git a/src/gallium/auxiliary/llvm/storagesoa.cpp b/src/gallium/auxiliary/llvm/storagesoa.cpp deleted file mode 100644 index ed0674a96f..0000000000 --- a/src/gallium/auxiliary/llvm/storagesoa.cpp +++ /dev/null @@ -1,389 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "storagesoa.h" - -#include "gallivm_p.h" - -#include "pipe/p_shader_tokens.h" -#include "pipe/p_debug.h" - -#include -#include -#include - -#include -#include -#include -#include -#include - -using namespace llvm; - - -StorageSoa::StorageSoa(llvm::BasicBlock *block, - llvm::Value *input, - llvm::Value *output, - llvm::Value *consts, - llvm::Value *temps) - : m_block(block), - m_input(input), - m_output(output), - m_consts(consts), - m_temps(temps), - m_immediates(0), - m_idx(0) -{ -} - -void StorageSoa::addImmediate(float *vec) -{ - std::vector vals(4); - vals[0] = vec[0]; - vals[1] = vec[1]; - vals[2] = vec[2]; - vals[3] = vec[3]; - m_immediatesToFlush.push_back(vals); -} - -void StorageSoa::declareImmediates() -{ - if (m_immediatesToFlush.empty()) - return; - - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - ArrayType *vectorChannels = ArrayType::get(vectorType, 4); - ArrayType *arrayType = ArrayType::get(vectorChannels, m_immediatesToFlush.size()); - - m_immediates = new GlobalVariable( - /*Type=*/arrayType, - /*isConstant=*/false, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name("immediates"), - currentModule()); - - std::vector arrayVals; - for (unsigned int i = 0; i < m_immediatesToFlush.size(); ++i) { - std::vector vec = m_immediatesToFlush[i]; - std::vector vals(4); - std::vector channelArray; - - vals[0] = vec[0]; vals[1] = vec[0]; vals[2] = vec[0]; vals[3] = vec[0]; - llvm::Constant *xChannel = createConstGlobalVector(vals); - - vals[0] = vec[1]; vals[1] = vec[1]; vals[2] = vec[1]; vals[3] = vec[1]; - llvm::Constant *yChannel = createConstGlobalVector(vals); - - vals[0] = vec[2]; vals[1] = vec[2]; vals[2] = vec[2]; vals[3] = vec[2]; - llvm::Constant *zChannel = createConstGlobalVector(vals); - - vals[0] = vec[3]; vals[1] = vec[3]; vals[2] = vec[3]; vals[3] = vec[3]; - llvm::Constant *wChannel = createConstGlobalVector(vals); - channelArray.push_back(xChannel); - channelArray.push_back(yChannel); - channelArray.push_back(zChannel); - channelArray.push_back(wChannel); - Constant *constChannels = ConstantArray::get(vectorChannels, - channelArray); - arrayVals.push_back(constChannels); - } - Constant *constArray = ConstantArray::get(arrayType, arrayVals); - m_immediates->setInitializer(constArray); - - m_immediatesToFlush.clear(); -} - -llvm::Value *StorageSoa::addrElement(int idx) const -{ - std::map::const_iterator itr = m_addresses.find(idx); - if (itr == m_addresses.end()) { - debug_printf("Trying to access invalid shader 'address'\n"); - return 0; - } - llvm::Value * res = (*itr).second; - - res = new LoadInst(res, name("addr"), false, m_block); - - return res; -} - -std::vector StorageSoa::inputElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_input, idx, 0); - res[1] = element(m_input, idx, 1); - res[2] = element(m_input, idx, 2); - res[3] = element(m_input, idx, 3); - - return res; -} - -std::vector StorageSoa::constElement(llvm::Value *idx) -{ - std::vector res(4); - llvm::Value *xChannel, *yChannel, *zChannel, *wChannel; - - xChannel = elementPointer(m_consts, idx, 0); - yChannel = elementPointer(m_consts, idx, 1); - zChannel = elementPointer(m_consts, idx, 2); - wChannel = elementPointer(m_consts, idx, 3); - - res[0] = alignedArrayLoad(xChannel); - res[1] = alignedArrayLoad(yChannel); - res[2] = alignedArrayLoad(zChannel); - res[3] = alignedArrayLoad(wChannel); - - return res; -} - -std::vector StorageSoa::outputElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_output, idx, 0); - res[1] = element(m_output, idx, 1); - res[2] = element(m_output, idx, 2); - res[3] = element(m_output, idx, 3); - - return res; -} - -std::vector StorageSoa::tempElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_temps, idx, 0); - res[1] = element(m_temps, idx, 1); - res[2] = element(m_temps, idx, 2); - res[3] = element(m_temps, idx, 3); - - return res; -} - -std::vector StorageSoa::immediateElement(llvm::Value *idx) -{ - std::vector res(4); - - res[0] = element(m_immediates, idx, 0); - res[1] = element(m_immediates, idx, 1); - res[2] = element(m_immediates, idx, 2); - res[3] = element(m_immediates, idx, 3); - - return res; -} - -llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index, - int channel) const -{ - std::vector indices; - if (m_immediates == ptr) - indices.push_back(constantInt(0)); - indices.push_back(index); - indices.push_back(constantInt(channel)); - - GetElementPtrInst *getElem = new GetElementPtrInst(ptr, - indices.begin(), - indices.end(), - name("ptr"), - m_block); - return getElem; -} - -llvm::Value * StorageSoa::element(llvm::Value *ptr, llvm::Value *index, - int channel) const -{ - llvm::Value *res = elementPointer(ptr, index, channel); - LoadInst *load = new LoadInst(res, name("element"), false, m_block); - //load->setAlignment(8); - return load; -} - -const char * StorageSoa::name(const char *prefix) const -{ - ++m_idx; - snprintf(m_name, 32, "%s%d", prefix, m_idx); - return m_name; -} - -llvm::ConstantInt * StorageSoa::constantInt(int idx) const -{ - if (m_constInts.find(idx) != m_constInts.end()) { - return m_constInts[idx]; - } - ConstantInt *constInt = ConstantInt::get(APInt(32, idx)); - m_constInts[idx] = constInt; - return constInt; -} - -llvm::Value *StorageSoa::alignedArrayLoad(llvm::Value *val) -{ - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - PointerType *vectorPtr = PointerType::get(vectorType, 0); - - CastInst *cast = new BitCastInst(val, vectorPtr, name("toVector"), m_block); - LoadInst *load = new LoadInst(cast, name("alignLoad"), false, m_block); - load->setAlignment(8); - return load; -} - -llvm::Module * StorageSoa::currentModule() const -{ - if (!m_block || !m_block->getParent()) - return 0; - - return m_block->getParent()->getParent(); -} - -llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector &vec) -{ - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - std::vector immValues; - ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0])); - ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1])); - ConstantFP *constz = ConstantFP::get(Type::FloatTy, APFloat(vec[2])); - ConstantFP *constw = ConstantFP::get(Type::FloatTy, APFloat(vec[3])); - immValues.push_back(constx); - immValues.push_back(consty); - immValues.push_back(constz); - immValues.push_back(constw); - Constant *constVector = ConstantVector::get(vectorType, immValues); - - return constVector; -} - -std::vector StorageSoa::load(Argument type, int idx, int swizzle, - llvm::Value *indIdx) -{ - std::vector val(4); - - //if we have an indirect index, always use that - // if not use the integer offset to create one - llvm::Value *realIndex = 0; - if (indIdx) - realIndex = indIdx; - else - realIndex = constantInt(idx); - debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx); - - switch(type) { - case Input: - val = inputElement(realIndex); - break; - case Output: - val = outputElement(realIndex); - break; - case Temp: - val = tempElement(realIndex); - break; - case Const: - val = constElement(realIndex); - break; - case Immediate: - val = immediateElement(realIndex); - break; - case Address: - debug_printf("Address not handled in the load phase!\n"); - assert(0); - break; - } - if (!gallivm_is_swizzle(swizzle)) - return val; - - std::vector res(4); - - res[0] = val[gallivm_x_swizzle(swizzle)]; - res[1] = val[gallivm_y_swizzle(swizzle)]; - res[2] = val[gallivm_z_swizzle(swizzle)]; - res[3] = val[gallivm_w_swizzle(swizzle)]; - return res; -} - -void StorageSoa::store(Argument type, int idx, const std::vector &val, - int mask) -{ - llvm::Value *out = 0; - switch(type) { - case Output: - out = m_output; - break; - case Temp: - out = m_temps; - break; - case Input: - out = m_input; - break; - case Address: { - llvm::Value *addr = m_addresses[idx]; - if (!addr) { - addAddress(idx); - addr = m_addresses[idx]; - assert(addr); - } - new StoreInst(val[0], addr, false, m_block); - return; - break; - } - default: - debug_printf("Can't save output of this type: %d !\n", type); - assert(0); - break; - } - llvm::Value *realIndex = constantInt(idx); - if ((mask & TGSI_WRITEMASK_X)) { - llvm::Value *xChannel = elementPointer(out, realIndex, 0); - new StoreInst(val[0], xChannel, false, m_block); - } - if ((mask & TGSI_WRITEMASK_Y)) { - llvm::Value *yChannel = elementPointer(out, realIndex, 1); - new StoreInst(val[1], yChannel, false, m_block); - } - if ((mask & TGSI_WRITEMASK_Z)) { - llvm::Value *zChannel = elementPointer(out, realIndex, 2); - new StoreInst(val[2], zChannel, false, m_block); - } - if ((mask & TGSI_WRITEMASK_W)) { - llvm::Value *wChannel = elementPointer(out, realIndex, 3); - new StoreInst(val[3], wChannel, false, m_block); - } -} - -void StorageSoa::addAddress(int idx) -{ - GlobalVariable *val = new GlobalVariable( - /*Type=*/IntegerType::get(32), - /*isConstant=*/false, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Initializer=*/0, // has initializer, specified below - /*Name=*/name("address"), - currentModule()); - val->setInitializer(Constant::getNullValue(IntegerType::get(32))); - - debug_printf("adding to %d\n", idx); - m_addresses[idx] = val; -} diff --git a/src/gallium/auxiliary/llvm/storagesoa.h b/src/gallium/auxiliary/llvm/storagesoa.h deleted file mode 100644 index 6443351f27..0000000000 --- a/src/gallium/auxiliary/llvm/storagesoa.h +++ /dev/null @@ -1,111 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef STORAGESOA_H -#define STORAGESOA_H - -#include -#include -#include - -namespace llvm { - class BasicBlock; - class Constant; - class ConstantInt; - class GlobalVariable; - class LoadInst; - class Value; - class VectorType; - class Module; -} - -class StorageSoa -{ -public: - enum Argument { - Input, - Output, - Temp, - Const, - Immediate, - Address - }; -public: - StorageSoa(llvm::BasicBlock *block, - llvm::Value *input, - llvm::Value *output, - llvm::Value *consts, - llvm::Value *temps); - - - std::vector load(Argument type, int idx, int swizzle, - llvm::Value *indIdx =0); - void store(Argument type, int idx, const std::vector &val, - int mask); - - void addImmediate(float *vec); - void declareImmediates(); - - void addAddress(int idx); - - llvm::Value * addrElement(int idx) const; - - llvm::ConstantInt *constantInt(int) const; -private: - llvm::Value *elementPointer(llvm::Value *ptr, llvm::Value *indIdx, - int channel) const; - llvm::Value *element(llvm::Value *ptr, llvm::Value *idx, - int channel) const; - const char *name(const char *prefix) const; - llvm::Value *alignedArrayLoad(llvm::Value *val); - llvm::Module *currentModule() const; - llvm::Constant *createConstGlobalVector(const std::vector &vec); - - std::vector inputElement(llvm::Value *indIdx); - std::vector constElement(llvm::Value *indIdx); - std::vector outputElement(llvm::Value *indIdx); - std::vector tempElement(llvm::Value *indIdx); - std::vector immediateElement(llvm::Value *indIdx); -private: - llvm::BasicBlock *m_block; - - llvm::Value *m_input; - llvm::Value *m_output; - llvm::Value *m_consts; - llvm::Value *m_temps; - llvm::GlobalVariable *m_immediates; - - std::map m_addresses; - - std::vector > m_immediatesToFlush; - - mutable std::map m_constInts; - mutable char m_name[32]; - mutable int m_idx; -}; - -#endif diff --git a/src/gallium/auxiliary/llvm/tgsitollvm.cpp b/src/gallium/auxiliary/llvm/tgsitollvm.cpp deleted file mode 100644 index 2cb4acce32..0000000000 --- a/src/gallium/auxiliary/llvm/tgsitollvm.cpp +++ /dev/null @@ -1,1221 +0,0 @@ -#include "tgsitollvm.h" - -#include "gallivm.h" -#include "gallivm_p.h" - -#include "storage.h" -#include "instructions.h" -#include "storagesoa.h" -#include "instructionssoa.h" - -#include "pipe/p_shader_tokens.h" - -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi/util/tgsi_build.h" -#include "tgsi/util/tgsi_dump.h" - - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - - -#include -#include -#include - -using namespace llvm; - -static inline FunctionType *vertexShaderFunctionType() -{ - //Function takes three arguments, - // the calling code has to make sure the types it will - // pass are castable to the following: - // [4 x <4 x float>] inputs, - // [4 x <4 x float>] output, - // [4 x [4 x float]] consts, - // [4 x <4 x float>] temps - - std::vector funcArgs; - VectorType *vectorType = VectorType::get(Type::FloatTy, 4); - ArrayType *vectorArray = ArrayType::get(vectorType, 4); - PointerType *vectorArrayPtr = PointerType::get(vectorArray, 0); - - ArrayType *floatArray = ArrayType::get(Type::FloatTy, 4); - ArrayType *constsArray = ArrayType::get(floatArray, 4); - PointerType *constsArrayPtr = PointerType::get(constsArray, 0); - - funcArgs.push_back(vectorArrayPtr);//inputs - funcArgs.push_back(vectorArrayPtr);//output - funcArgs.push_back(constsArrayPtr);//consts - funcArgs.push_back(vectorArrayPtr);//temps - - FunctionType *functionType = FunctionType::get( - /*Result=*/Type::VoidTy, - /*Params=*/funcArgs, - /*isVarArg=*/false); - - return functionType; -} - -static inline void -add_interpolator(struct gallivm_ir *ir, - struct gallivm_interpolate *interp) -{ - ir->interpolators[ir->num_interp] = *interp; - ++ir->num_interp; -} - -static void -translate_declaration(struct gallivm_ir *prog, - llvm::Module *module, - Storage *storage, - struct tgsi_full_declaration *decl, - struct tgsi_full_declaration *fd) -{ - if (decl->Declaration.File == TGSI_FILE_INPUT) { - unsigned first, last, mask; - uint interp_method; - - assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - /* Do not touch WPOS.xy */ - if (first == 0) { - mask &= ~TGSI_WRITEMASK_XY; - if (mask == TGSI_WRITEMASK_NONE) { - first++; - if (first > last) { - return; - } - } - } - - interp_method = decl->Interpolation.Interpolate; - - if (mask == TGSI_WRITEMASK_XYZW) { - unsigned i, j; - - for (i = first; i <= last; i++) { - for (j = 0; j < NUM_CHANNELS; j++) { - //interp( mach, i, j ); - struct gallivm_interpolate interp; - interp.type = interp_method; - interp.attrib = i; - interp.chan = j; - add_interpolator(prog, &interp); - } - } - } else { - unsigned i, j; - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - struct gallivm_interpolate interp; - interp.type = interp_method; - interp.attrib = i; - interp.chan = j; - add_interpolator(prog, &interp); - } - } - } - } - } -} - -static void -translate_declarationir(struct gallivm_ir *, - llvm::Module *, - StorageSoa *storage, - struct tgsi_full_declaration *decl, - struct tgsi_full_declaration *) -{ - if (decl->Declaration.File == TGSI_FILE_ADDRESS) { - int idx = decl->u.DeclarationRange.First; - storage->addAddress(idx); - } -} - -static void -translate_immediate(Storage *storage, - struct tgsi_full_immediate *imm) -{ - float vec[4]; - int i; - for (i = 0; i < imm->Immediate.Size - 1; ++i) { - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; - break; - default: - assert(0); - } - } - storage->addImmediate(vec); -} - - -static void -translate_immediateir(StorageSoa *storage, - struct tgsi_full_immediate *imm) -{ - float vec[4]; - int i; - for (i = 0; i < imm->Immediate.Size - 1; ++i) { - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - vec[i] = imm->u.ImmediateFloat32[i].Float; - break; - default: - assert(0); - } - } - storage->addImmediate(vec); -} - -static inline int -swizzleInt(struct tgsi_full_src_register *src) -{ - int swizzle = 0; - int start = 1000; - - for (int k = 0; k < 4; ++k) { - swizzle += tgsi_util_get_full_src_register_extswizzle(src, k) * start; - start /= 10; - } - return swizzle; -} - -static inline llvm::Value * -swizzleVector(llvm::Value *val, struct tgsi_full_src_register *src, - Storage *storage) -{ - int swizzle = swizzleInt(src); - - if (gallivm_is_swizzle(swizzle)) { - /*fprintf(stderr, "XXXXXXXX swizzle = %d\n", swizzle);*/ - val = storage->shuffleVector(val, swizzle); - } - return val; -} - -static void -translate_instruction(llvm::Module *module, - Storage *storage, - Instructions *instr, - struct tgsi_full_instruction *inst, - struct tgsi_full_instruction *fi, - unsigned instno) -{ - llvm::Value *inputs[4]; - inputs[0] = 0; - inputs[1] = 0; - inputs[2] = 0; - inputs[3] = 0; - - for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - llvm::Value *val = 0; - llvm::Value *indIdx = 0; - - if (src->SrcRegister.Indirect) { - indIdx = storage->addrElement(src->SrcRegisterInd.Index); - indIdx = storage->extractIndex(indIdx); - } - if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->constElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->inputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->tempElement(src->SrcRegister.Index); - } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->outputElement(src->SrcRegister.Index, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->immediateElement(src->SrcRegister.Index); - } else { - fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); - return; - } - - inputs[i] = swizzleVector(val, src, storage); - } - - /*if (inputs[0]) - instr->printVector(inputs[0]); - if (inputs[1]) - instr->printVector(inputs[1]);*/ - llvm::Value *out = 0; - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: { - out = instr->arl(inputs[0]); - } - break; - case TGSI_OPCODE_MOV: { - out = inputs[0]; - } - break; - case TGSI_OPCODE_LIT: { - out = instr->lit(inputs[0]); - } - break; - case TGSI_OPCODE_RCP: { - out = instr->rcp(inputs[0]); - } - break; - case TGSI_OPCODE_RSQ: { - out = instr->rsq(inputs[0]); - } - break; - case TGSI_OPCODE_EXP: - break; - case TGSI_OPCODE_LOG: - break; - case TGSI_OPCODE_MUL: { - out = instr->mul(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_ADD: { - out = instr->add(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP3: { - out = instr->dp3(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP4: { - out = instr->dp4(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DST: { - out = instr->dst(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MIN: { - out = instr->min(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAX: { - out = instr->max(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SLT: { - out = instr->slt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SGE: { - out = instr->sge(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MAD: { - out = instr->madd(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SUB: { - out = instr->sub(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_LERP: { - out = instr->lerp(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_CND: - break; - case TGSI_OPCODE_CND0: - break; - case TGSI_OPCODE_DOT2ADD: - break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { - out = instr->frc(inputs[0]); - } - break; - case TGSI_OPCODE_CLAMP: - break; - case TGSI_OPCODE_FLOOR: { - out = instr->floor(inputs[0]); - } - break; - case TGSI_OPCODE_ROUND: - break; - case TGSI_OPCODE_EXPBASE2: { - out = instr->ex2(inputs[0]); - } - break; - case TGSI_OPCODE_LOGBASE2: { - out = instr->lg2(inputs[0]); - } - break; - case TGSI_OPCODE_POWER: { - out = instr->pow(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_CROSSPRODUCT: { - out = instr->cross(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; - case TGSI_OPCODE_ABS: { - out = instr->abs(inputs[0]); - } - break; - case TGSI_OPCODE_RCC: - break; - case TGSI_OPCODE_DPH: { - out = instr->dph(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_COS: { - out = instr->cos(inputs[0]); - } - break; - case TGSI_OPCODE_DDX: - break; - case TGSI_OPCODE_DDY: - break; - case TGSI_OPCODE_KILP: { - out = instr->kilp(inputs[0]); - storage->setKilElement(out); - return; - } - break; - case TGSI_OPCODE_PK2H: - break; - case TGSI_OPCODE_PK2US: - break; - case TGSI_OPCODE_PK4B: - break; - case TGSI_OPCODE_PK4UB: - break; - case TGSI_OPCODE_RFL: - break; - case TGSI_OPCODE_SEQ: - break; - case TGSI_OPCODE_SFL: - break; - case TGSI_OPCODE_SGT: { - out = instr->sgt(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_SIN: { - out = instr->sin(inputs[0]); - } - break; - case TGSI_OPCODE_SLE: - break; - case TGSI_OPCODE_SNE: - break; - case TGSI_OPCODE_STR: - break; - case TGSI_OPCODE_TEX: - break; - case TGSI_OPCODE_TXD: - break; - case TGSI_OPCODE_UP2H: - break; - case TGSI_OPCODE_UP2US: - break; - case TGSI_OPCODE_UP4B: - break; - case TGSI_OPCODE_UP4UB: - break; - case TGSI_OPCODE_X2D: - break; - case TGSI_OPCODE_ARA: - break; - case TGSI_OPCODE_ARR: - break; - case TGSI_OPCODE_BRA: - break; - case TGSI_OPCODE_CAL: { - instr->cal(inst->InstructionExtLabel.Label, storage->inputPtr()); - return; - } - break; - case TGSI_OPCODE_RET: { - instr->end(); - return; - } - break; - case TGSI_OPCODE_SSG: - break; - case TGSI_OPCODE_CMP: { - out = instr->cmp(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SCS: { - out = instr->scs(inputs[0]); - } - break; - case TGSI_OPCODE_TXB: - break; - case TGSI_OPCODE_NRM: - break; - case TGSI_OPCODE_DIV: - break; - case TGSI_OPCODE_DP2: - break; - case TGSI_OPCODE_TXL: - break; - case TGSI_OPCODE_BRK: { - instr->brk(); - return; - } - break; - case TGSI_OPCODE_IF: { - instr->ifop(inputs[0]); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_LOOP: - break; - case TGSI_OPCODE_REP: - break; - case TGSI_OPCODE_ELSE: { - instr->elseop(); - storage->setCurrentBlock(instr->currentBlock()); - return; //only state update - } - break; - case TGSI_OPCODE_ENDIF: { - instr->endif(); - storage->setCurrentBlock(instr->currentBlock()); - return; //just update the state - } - break; - case TGSI_OPCODE_ENDLOOP: - break; - case TGSI_OPCODE_ENDREP: - break; - case TGSI_OPCODE_PUSHA: - break; - case TGSI_OPCODE_POPA: - break; - case TGSI_OPCODE_CEIL: - break; - case TGSI_OPCODE_I2F: - break; - case TGSI_OPCODE_NOT: - break; - case TGSI_OPCODE_TRUNC: { - out = instr->trunc(inputs[0]); - } - break; - case TGSI_OPCODE_SHL: - break; - case TGSI_OPCODE_SHR: - break; - case TGSI_OPCODE_AND: - break; - case TGSI_OPCODE_OR: - break; - case TGSI_OPCODE_MOD: - break; - case TGSI_OPCODE_XOR: - break; - case TGSI_OPCODE_SAD: - break; - case TGSI_OPCODE_TXF: - break; - case TGSI_OPCODE_TXQ: - break; - case TGSI_OPCODE_CONT: - break; - case TGSI_OPCODE_EMIT: - break; - case TGSI_OPCODE_ENDPRIM: - break; - case TGSI_OPCODE_BGNLOOP2: { - instr->beginLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_BGNSUB: { - instr->bgnSub(instno); - storage->setCurrentBlock(instr->currentBlock()); - storage->pushTemps(); - return; - } - break; - case TGSI_OPCODE_ENDLOOP2: { - instr->endLoop(); - storage->setCurrentBlock(instr->currentBlock()); - return; - } - break; - case TGSI_OPCODE_ENDSUB: { - instr->endSub(); - storage->setCurrentBlock(instr->currentBlock()); - storage->popArguments(); - storage->popTemps(); - return; - } - break; - case TGSI_OPCODE_NOISE1: - break; - case TGSI_OPCODE_NOISE2: - break; - case TGSI_OPCODE_NOISE3: - break; - case TGSI_OPCODE_NOISE4: - break; - case TGSI_OPCODE_NOP: - break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; - case TGSI_OPCODE_NRM4: - break; - case TGSI_OPCODE_CALLNZ: - break; - case TGSI_OPCODE_IFC: - break; - case TGSI_OPCODE_BREAKC: - break; - case TGSI_OPCODE_KIL: - break; - case TGSI_OPCODE_END: - instr->end(); - return; - break; - default: - fprintf(stderr, "ERROR: Unknown opcode %d\n", - inst->Instruction.Opcode); - assert(0); - break; - } - - if (!out) { - fprintf(stderr, "ERROR: unsupported opcode %d\n", - inst->Instruction.Opcode); - assert(!"Unsupported opcode"); - } - - /* # not sure if we need this */ - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - /*TXT( "_SAT" );*/ - break; - case TGSI_SAT_MINUS_PLUS_ONE: - /*TXT( "_SAT[-1,1]" );*/ - break; - default: - assert( 0 ); - } - - /* store results */ - for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->setOutputElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->setTempElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->setAddrElement(dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else { - fprintf(stderr, "ERROR: unsupported LLVM destination!"); - assert(!"wrong destination"); - } - } -} - - -static void -translate_instructionir(llvm::Module *module, - StorageSoa *storage, - InstructionsSoa *instr, - struct tgsi_full_instruction *inst, - struct tgsi_full_instruction *fi, - unsigned instno) -{ - std::vector< std::vector > inputs(inst->Instruction.NumSrcRegs); - - for (int i = 0; i < inst->Instruction.NumSrcRegs; ++i) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - std::vector val; - llvm::Value *indIdx = 0; - int swizzle = swizzleInt(src); - - if (src->SrcRegister.Indirect) { - indIdx = storage->addrElement(src->SrcRegisterInd.Index); - } - if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->load(StorageSoa::Const, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->load(StorageSoa::Input, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->load(StorageSoa::Temp, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->load(StorageSoa::Output, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->load(StorageSoa::Immediate, - src->SrcRegister.Index, swizzle, indIdx); - } else { - fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); - return; - } - - inputs[i] = val; - } - - std::vector out(4); - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: { - out = instr->arl(inputs[0]); - } - break; - case TGSI_OPCODE_MOV: { - out = inputs[0]; - } - break; - case TGSI_OPCODE_LIT: { - } - break; - case TGSI_OPCODE_RCP: { - } - break; - case TGSI_OPCODE_RSQ: { - } - break; - case TGSI_OPCODE_EXP: - break; - case TGSI_OPCODE_LOG: - break; - case TGSI_OPCODE_MUL: { - out = instr->mul(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_ADD: { - out = instr->add(inputs[0], inputs[1]); - } - break; - case TGSI_OPCODE_DP3: { - } - break; - case TGSI_OPCODE_DP4: { - } - break; - case TGSI_OPCODE_DST: { - } - break; - case TGSI_OPCODE_MIN: { - } - break; - case TGSI_OPCODE_MAX: { - } - break; - case TGSI_OPCODE_SLT: { - } - break; - case TGSI_OPCODE_SGE: { - } - break; - case TGSI_OPCODE_MAD: { - out = instr->madd(inputs[0], inputs[1], inputs[2]); - } - break; - case TGSI_OPCODE_SUB: { - } - break; - case TGSI_OPCODE_LERP: { - } - break; - case TGSI_OPCODE_CND: - break; - case TGSI_OPCODE_CND0: - break; - case TGSI_OPCODE_DOT2ADD: - break; - case TGSI_OPCODE_INDEX: - break; - case TGSI_OPCODE_NEGATE: - break; - case TGSI_OPCODE_FRAC: { - } - break; - case TGSI_OPCODE_CLAMP: - break; - case TGSI_OPCODE_FLOOR: { - } - break; - case TGSI_OPCODE_ROUND: - break; - case TGSI_OPCODE_EXPBASE2: { - } - break; - case TGSI_OPCODE_LOGBASE2: { - } - break; - case TGSI_OPCODE_POWER: { - } - break; - case TGSI_OPCODE_CROSSPRODUCT: { - } - break; - case TGSI_OPCODE_MULTIPLYMATRIX: - break; - case TGSI_OPCODE_ABS: { - } - break; - case TGSI_OPCODE_RCC: - break; - case TGSI_OPCODE_DPH: { - } - break; - case TGSI_OPCODE_COS: { - } - break; - case TGSI_OPCODE_DDX: - break; - case TGSI_OPCODE_DDY: - break; - case TGSI_OPCODE_KILP: { - } - break; - case TGSI_OPCODE_PK2H: - break; - case TGSI_OPCODE_PK2US: - break; - case TGSI_OPCODE_PK4B: - break; - case TGSI_OPCODE_PK4UB: - break; - case TGSI_OPCODE_RFL: - break; - case TGSI_OPCODE_SEQ: - break; - case TGSI_OPCODE_SFL: - break; - case TGSI_OPCODE_SGT: { - } - break; - case TGSI_OPCODE_SIN: { - } - break; - case TGSI_OPCODE_SLE: - break; - case TGSI_OPCODE_SNE: - break; - case TGSI_OPCODE_STR: - break; - case TGSI_OPCODE_TEX: - break; - case TGSI_OPCODE_TXD: - break; - case TGSI_OPCODE_UP2H: - break; - case TGSI_OPCODE_UP2US: - break; - case TGSI_OPCODE_UP4B: - break; - case TGSI_OPCODE_UP4UB: - break; - case TGSI_OPCODE_X2D: - break; - case TGSI_OPCODE_ARA: - break; - case TGSI_OPCODE_ARR: - break; - case TGSI_OPCODE_BRA: - break; - case TGSI_OPCODE_CAL: { - } - break; - case TGSI_OPCODE_RET: { - } - break; - case TGSI_OPCODE_SSG: - break; - case TGSI_OPCODE_CMP: { - } - break; - case TGSI_OPCODE_SCS: { - } - break; - case TGSI_OPCODE_TXB: - break; - case TGSI_OPCODE_NRM: - break; - case TGSI_OPCODE_DIV: - break; - case TGSI_OPCODE_DP2: - break; - case TGSI_OPCODE_TXL: - break; - case TGSI_OPCODE_BRK: { - } - break; - case TGSI_OPCODE_IF: { - } - break; - case TGSI_OPCODE_LOOP: - break; - case TGSI_OPCODE_REP: - break; - case TGSI_OPCODE_ELSE: { - } - break; - case TGSI_OPCODE_ENDIF: { - } - break; - case TGSI_OPCODE_ENDLOOP: - break; - case TGSI_OPCODE_ENDREP: - break; - case TGSI_OPCODE_PUSHA: - break; - case TGSI_OPCODE_POPA: - break; - case TGSI_OPCODE_CEIL: - break; - case TGSI_OPCODE_I2F: - break; - case TGSI_OPCODE_NOT: - break; - case TGSI_OPCODE_TRUNC: { - } - break; - case TGSI_OPCODE_SHL: - break; - case TGSI_OPCODE_SHR: - break; - case TGSI_OPCODE_AND: - break; - case TGSI_OPCODE_OR: - break; - case TGSI_OPCODE_MOD: - break; - case TGSI_OPCODE_XOR: - break; - case TGSI_OPCODE_SAD: - break; - case TGSI_OPCODE_TXF: - break; - case TGSI_OPCODE_TXQ: - break; - case TGSI_OPCODE_CONT: - break; - case TGSI_OPCODE_EMIT: - break; - case TGSI_OPCODE_ENDPRIM: - break; - case TGSI_OPCODE_BGNLOOP2: { - } - break; - case TGSI_OPCODE_BGNSUB: { - } - break; - case TGSI_OPCODE_ENDLOOP2: { - } - break; - case TGSI_OPCODE_ENDSUB: { - } - break; - case TGSI_OPCODE_NOISE1: - break; - case TGSI_OPCODE_NOISE2: - break; - case TGSI_OPCODE_NOISE3: - break; - case TGSI_OPCODE_NOISE4: - break; - case TGSI_OPCODE_NOP: - break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; - case TGSI_OPCODE_M4X3: - break; - case TGSI_OPCODE_M3X4: - break; - case TGSI_OPCODE_M3X3: - break; - case TGSI_OPCODE_M3X2: - break; - case TGSI_OPCODE_NRM4: - break; - case TGSI_OPCODE_CALLNZ: - break; - case TGSI_OPCODE_IFC: - break; - case TGSI_OPCODE_BREAKC: - break; - case TGSI_OPCODE_KIL: - break; - case TGSI_OPCODE_END: - instr->end(); - return; - break; - default: - fprintf(stderr, "ERROR: Unknown opcode %d\n", - inst->Instruction.Opcode); - assert(0); - break; - } - - if (!out[0]) { - fprintf(stderr, "ERROR: unsupported opcode %d\n", - inst->Instruction.Opcode); - assert(!"Unsupported opcode"); - } - - /* store results */ - for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->store(StorageSoa::Output, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->store(StorageSoa::Temp, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->store(StorageSoa::Address, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else { - fprintf(stderr, "ERROR: unsupported LLVM destination!"); - assert(!"wrong destination"); - } - } -} - -llvm::Module * -tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) -{ - llvm::Module *mod = new Module("shader"); - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned instno = 0; - Function* shader = mod->getFunction("execute_shader"); - std::ostringstream stream; - if (ir->type == GALLIVM_VS) { - stream << "vs_shader"; - } else { - stream << "fs_shader"; - } - stream << ir->id; - std::string func_name = stream.str(); - shader->setName(func_name.c_str()); - - Function::arg_iterator args = shader->arg_begin(); - Value *ptr_INPUT = args++; - ptr_INPUT->setName("input"); - - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); - - tgsi_parse_init(&parse, tokens); - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - Storage storage(label_entry, ptr_INPUT); - Instructions instr(mod, shader, label_entry, &storage); - while(!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - translate_declaration(ir, mod, &storage, - &parse.FullToken.FullDeclaration, - &fd); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - translate_immediate(&storage, - &parse.FullToken.FullImmediate); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - translate_instruction(mod, &storage, &instr, - &parse.FullToken.FullInstruction, - &fi, instno); - ++instno; - break; - - default: - assert(0); - } - } - - tgsi_parse_free(&parse); - - ir->num_consts = storage.numConsts(); - return mod; -} - -llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, - const struct tgsi_token *tokens) -{ - llvm::Module *mod = new Module("shader"); - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned instno = 0; - std::ostringstream stream; - if (ir->type == GALLIVM_VS) { - stream << "vs_shader"; - } else { - stream << "fs_shader"; - } - //stream << ir->id; - std::string func_name = stream.str(); - Function *shader = llvm::cast(mod->getOrInsertFunction( - func_name.c_str(), - vertexShaderFunctionType())); - - Function::arg_iterator args = shader->arg_begin(); - Value *input = args++; - input->setName("inputs"); - Value *output = args++; - output->setName("outputs"); - Value *consts = args++; - consts->setName("consts"); - Value *temps = args++; - temps->setName("temps"); - - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); - - tgsi_parse_init(&parse, tokens); - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - - StorageSoa storage(label_entry, input, output, consts, temps); - InstructionsSoa instr(mod, shader, label_entry, &storage); - - while(!tgsi_parse_end_of_tokens(&parse)) { - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - translate_declarationir(ir, mod, &storage, - &parse.FullToken.FullDeclaration, - &fd); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - translate_immediateir(&storage, - &parse.FullToken.FullImmediate); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - storage.declareImmediates(); - translate_instructionir(mod, &storage, &instr, - &parse.FullToken.FullInstruction, - &fi, instno); - ++instno; - break; - - default: - assert(0); - } - } - - tgsi_parse_free(&parse); - - return mod; -} diff --git a/src/gallium/auxiliary/llvm/tgsitollvm.h b/src/gallium/auxiliary/llvm/tgsitollvm.h deleted file mode 100644 index 7ada04d629..0000000000 --- a/src/gallium/auxiliary/llvm/tgsitollvm.h +++ /dev/null @@ -1,20 +0,0 @@ -#ifndef TGSITOLLVM_H -#define TGSITOLLVM_H - - -namespace llvm { - class Module; -} - -struct gallivm_ir; -struct tgsi_token; - - -llvm::Module * tgsi_to_llvm(struct gallivm_ir *ir, - const struct tgsi_token *tokens); - - -llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, - const struct tgsi_token *tokens); - -#endif -- cgit v1.2.3 From 0448dbd64a2ef217349f4ada4777d432bc82e46d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 18 Feb 2008 12:33:34 +0000 Subject: Update for llvm -> gallivm rename. --- src/gallium/SConscript | 2 +- src/gallium/auxiliary/draw/draw_vs_llvm.c | 2 +- src/gallium/winsys/xlib/Makefile | 8 ++------ src/mesa/Makefile | 4 ++-- 4 files changed, 6 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/SConscript b/src/gallium/SConscript index a835f6d661..ea55ed2ed5 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -13,7 +13,7 @@ SConscript([ 'auxiliary/tgsi/SConscript', 'auxiliary/cso_cache/SConscript', 'auxiliary/draw/SConscript', - #'auxiliary/llvm/SConscript', + #'auxiliary/gallivm/SConscript', 'auxiliary/pipebuffer/SConscript', 'drivers/softpipe/SConscript', diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 44022b6e07..0fd557d667 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -40,7 +40,7 @@ #ifdef MESA_LLVM -#include "llvm/gallivm.h" +#include "gallivm/gallivm.h" struct draw_llvm_vertex_shader { struct draw_vertex_shader base; diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile index 2664ac47ce..17405388ee 100644 --- a/src/gallium/winsys/xlib/Makefile +++ b/src/gallium/winsys/xlib/Makefile @@ -38,10 +38,6 @@ CELL_LIB = $(TOP)/src/gallium/drivers/cell/ppu/libcell.a CELL_LIB_SPU = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a endif -ifeq ($(CONFIG_NAME), linux-llvm) -LLVM_LIB = $(TOP)/src/gallium/auxiliary/llvm/libgallivm.a -endif - .SUFFIXES : .cpp @@ -69,13 +65,13 @@ STAND_ALONE_OBJECTS = \ $(STAND_ALONE_DRIVER_OBJECTS) # Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(LLVM_LIB) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(TOP)/bin/mklib -o $(GL_LIB) \ -linker "$(CC)" \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ - --start-group $(PIPE_LIB) $(LLVM_LIB) --end-group $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) + --start-group $(PIPE_LIB) --end-group $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) ###################################################################### diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 86a9cf0b88..2403223db2 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -67,13 +67,13 @@ stand-alone: depend subdirs $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) $(TOP)/$(LIB_DIR)/$ osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(LLVM_LIB) +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) @ $(TOP)/bin/mklib -o $(GL_LIB) \ -linker "$(CC)" \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ - $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(LLVM_LIB) $(GL_LIB_DEPS) + $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) # Make the OSMesa library $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) \ -- cgit v1.2.3 From aceeb80d4f706980aaf71b8e098d4c6718d8ac90 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 18 Feb 2008 16:19:05 -0700 Subject: gallium: antialiased line drawing New draw/prim stage: draw_aaline. When installed, lines are replaced by textured quads to do antialiasing. The current user-defined fragment shader is modified to do a texture fetch and modulate fragment alpha. --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/draw_aaline.c | 832 ++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_context.c | 22 + src/gallium/auxiliary/draw/draw_context.h | 21 +- src/gallium/auxiliary/draw/draw_prim.c | 7 + src/gallium/auxiliary/draw/draw_private.h | 12 +- src/gallium/auxiliary/draw/draw_validate.c | 8 +- src/gallium/auxiliary/tgsi/Makefile | 1 + src/gallium/drivers/softpipe/sp_context.c | 3 + src/gallium/drivers/softpipe/sp_state_derived.c | 17 +- 10 files changed, 914 insertions(+), 10 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_aaline.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index c56b63d85b..c8000cbe9c 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -4,6 +4,7 @@ include $(TOP)/configs/current LIBNAME = draw DRIVER_SOURCES = \ + draw_aaline.c \ draw_clip.c \ draw_vs_exec.c \ draw_vs_sse.c \ diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c new file mode 100644 index 0000000000..f1fee4f8ba --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -0,0 +1,832 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * AA line stage: AA lines are converted to texture mapped triangles. + * + * Authors: Brian Paul + */ + + +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/util/tgsi_transform.h" +#include "tgsi/util/tgsi_dump.h" + +#include "draw_context.h" +#include "draw_private.h" + + +/** + * Max texture level for the alpha texture used for antialiasing + */ +#define MAX_TEXTURE_LEVEL 5 /* 32 x 32 */ + + +/** + * Subclass of pipe_shader_state to carry extra fragment shader info. + */ +struct aaline_fragment_shader +{ + struct pipe_shader_state state; + void *driver_fs; + void *aaline_fs; + void *aapoint_fs; /* not yet */ + void *sprite_fs; /* not yet */ +}; + + +/** + * Subclass of draw_stage + */ +struct aaline_stage +{ + struct draw_stage stage; + + float half_line_width; + + /** For AA lines, this is the vertex attrib slot for the new texcoords */ + uint tex_slot; + + void *sampler_cso; + struct pipe_texture *texture; + uint sampler_unit; + + + /* + * Currently bound state + */ + struct aaline_fragment_shader *fs; + struct { + void *sampler[PIPE_MAX_SAMPLERS]; + struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + } state; + + /* + * Driver interface/override functions + */ + void * (*driver_create_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*driver_bind_fs_state)(struct pipe_context *, void *); + void (*driver_delete_fs_state)(struct pipe_context *, void *); + + void (*driver_bind_sampler_state)(struct pipe_context *, unsigned, void *); + + void (*driver_set_sampler_texture)(struct pipe_context *, + unsigned sampler, + struct pipe_texture *); + + struct pipe_context *pipe; +}; + + + +/** + * Subclass of tgsi_transform_context, used for transforming the + * user's fragment shader to add the special AA instructions. + */ +struct aa_transform_context { + struct tgsi_transform_context base; + uint tempsUsed; /**< bitmask */ + int colorOutput; /**< which output is the primary color */ + int maxSampler; /**< max sampler index found */ + int maxInput, maxGeneric; /**< max input index found */ + int colorTemp, texTemp; /**< temp registers */ + boolean firstInstruction; +}; + + +/** + * TGSI declaration transform callback. + * Look for a free sampler, a free input attrib, and two free temp regs. + */ +static void +aa_transform_decl(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl) +{ + struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; + + if (decl->Declaration.File == TGSI_FILE_OUTPUT && + decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR && + decl->Semantic.SemanticIndex == 0) { + aactx->colorOutput = decl->u.DeclarationRange.First; + } + else if (decl->Declaration.File == TGSI_FILE_SAMPLER) { + if (decl->u.DeclarationRange.Last > aactx->maxSampler) + aactx->maxSampler = decl->u.DeclarationRange.Last + 1; + } + else if (decl->Declaration.File == TGSI_FILE_INPUT) { + if (decl->u.DeclarationRange.Last > aactx->maxInput) + aactx->maxInput = decl->u.DeclarationRange.Last; + if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && + decl->Semantic.SemanticIndex > aactx->maxGeneric) { + aactx->maxGeneric = decl->Semantic.SemanticIndex; + } + } + else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + aactx->tempsUsed |= (1 << i); + } + } + + ctx->emit_declaration(ctx, decl); +} + + +/** + * TGSI instruction transform callback. + * Replace writes to result.color w/ a temp reg. + * Upon END instruction, insert texture sampling code for antialiasing. + */ +static void +aa_transform_inst(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst) +{ + struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; + + if (aactx->firstInstruction) { + /* emit our new declarations before the first instruction */ + + struct tgsi_full_declaration decl; + uint i; + + /* find two free temp regs */ + for (i = 0; i < 32; i++) { + if ((aactx->tempsUsed & (1 << i)) == 0) { + /* found a free temp */ + if (aactx->colorTemp < 0) + aactx->colorTemp = i; + else if (aactx->texTemp < 0) + aactx->texTemp = i; + else + break; + } + } + assert(aactx->colorTemp >= 0); + assert(aactx->texTemp >= 0); + + /* declare new generic input/texcoord */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; + decl.Semantic.SemanticIndex = aactx->maxGeneric + 1; + decl.Declaration.Interpolate = 1; + /* XXX this could be linear... */ + decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->maxInput + 1; + ctx->emit_declaration(ctx, &decl); + + /* declare new sampler */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_SAMPLER; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->maxSampler + 1; + ctx->emit_declaration(ctx, &decl); + + /* declare new temp regs */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->texTemp; + ctx->emit_declaration(ctx, &decl); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->colorTemp; + ctx->emit_declaration(ctx, &decl); + + aactx->firstInstruction = FALSE; + } + + if (inst->Instruction.Opcode == TGSI_OPCODE_END && + aactx->colorOutput != -1) { + struct tgsi_full_instruction newInst; + + /* TEX */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_TEX; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->texTemp; + newInst.Instruction.NumSrcRegs = 2; + newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->maxInput + 1; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->maxSampler + 1; + + ctx->emit_instruction(ctx, &newInst); + + /* MOV rgb */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MOV; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; + ctx->emit_instruction(ctx, &newInst); + + /* MUL alpha */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->texTemp; + ctx->emit_instruction(ctx, &newInst); + + /* END */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_END; + newInst.Instruction.NumDstRegs = 0; + newInst.Instruction.NumSrcRegs = 0; + ctx->emit_instruction(ctx, &newInst); + } + else { + /* Not an END instruction. + * Look for writes to result.color and replace with colorTemp reg. + */ + uint i; + + for (i = 0; i < inst->Instruction.NumDstRegs; i++) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + if (dst->DstRegister.File == TGSI_FILE_OUTPUT && + dst->DstRegister.Index == aactx->colorOutput) { + dst->DstRegister.File = TGSI_FILE_TEMPORARY; + dst->DstRegister.Index = aactx->colorTemp; + } + } + + ctx->emit_instruction(ctx, inst); + } +} + + +/** + * Generate the frag shader we'll use for drawing AA lines. + * This will be the user's shader plus some texture/modulate instructions. + */ +static void +generate_aaline_fs(struct aaline_stage *aaline) +{ + const struct pipe_shader_state *orig_fs = &aaline->fs->state; + struct draw_context *draw = aaline->stage.draw; + struct pipe_shader_state aaline_fs; + struct aa_transform_context transform; + +#define MAX 1000 + + aaline_fs = *orig_fs; /* copy to init */ + aaline_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); + + memset(&transform, 0, sizeof(transform)); + transform.colorOutput = -1; + transform.maxSampler = -1; + transform.maxInput = -1; + transform.maxGeneric = -1; + transform.colorTemp = -1; + transform.texTemp = -1; + transform.firstInstruction = TRUE; + transform.base.transform_instruction = aa_transform_inst; + transform.base.transform_declaration = aa_transform_decl; + + tgsi_transform_shader(orig_fs->tokens, + (struct tgsi_token *) aaline_fs.tokens, + MAX, &transform.base); + +#if 0 /* DEBUG */ + tgsi_dump(orig_fs->tokens, 0); + tgsi_dump(aaline_fs.tokens, 0); +#endif + + aaline_fs.input_semantic_name[aaline_fs.num_inputs] = TGSI_SEMANTIC_GENERIC; + aaline_fs.input_semantic_index[aaline_fs.num_inputs] = transform.maxGeneric + 1; + aaline_fs.num_inputs++; + + aaline->fs->aaline_fs + = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); + + /* advertise the extra post-transform vertex attributes which will have + * the texcoords. + */ + draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; + draw->extra_vp_outputs.semantic_index = transform.maxGeneric + 1; +} + + +/** + * Create the texture map we'll use for antialiasing the lines. + */ +static void +aaline_create_texture(struct aaline_stage *aaline) +{ + struct pipe_context *pipe = aaline->pipe; + struct pipe_texture texTemp; + uint level; + + memset(&texTemp, 0, sizeof(texTemp)); + texTemp.target = PIPE_TEXTURE_2D; + texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */ + texTemp.last_level = MAX_TEXTURE_LEVEL; + texTemp.width[0] = 1 << MAX_TEXTURE_LEVEL; + texTemp.height[0] = 1 << MAX_TEXTURE_LEVEL; + texTemp.depth[0] = 1; + texTemp.cpp = 1; + + aaline->texture = pipe->texture_create(pipe, &texTemp); + + /* Fill in mipmap images. + * Basically each level is solid opaque, except for the outermost + * texels which are zero. Special case the 1x1 and 2x2 levels. + */ + for (level = 0; level <= MAX_TEXTURE_LEVEL; level++) { + struct pipe_surface *surface; + const uint size = aaline->texture->width[level]; + ubyte *data; + uint i, j; + + assert(aaline->texture->width[level] == aaline->texture->height[level]); + + surface = pipe->get_tex_surface(pipe, aaline->texture, 0, level, 0); + data = pipe_surface_map(surface); + + for (i = 0; i < size; i++) { + for (j = 0; j < size; j++) { + uint d; + if (size == 1) { + d = 255; + } + else if (size == 2) { + d = 200; /* tuneable */ + } + else if (i == 0 || j == 0 || i == size - 1 || j == size - 1) { + d = 0; + } + else { + d = 255; + } + data[i * surface->pitch + j] = d; + } + } + + /* unmap */ + pipe_surface_unmap(surface); + pipe_surface_reference(&surface, NULL); + } +} + + +/** + * Create the sampler CSO that'll be used for antialiasing. + * By using a mipmapped texture, we don't have to generate a different + * texture image for each line size. + */ +static void +aaline_create_sampler(struct aaline_stage *aaline) +{ + struct pipe_sampler_state sampler; + struct pipe_context *pipe = aaline->pipe; + + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_LINEAR; + sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.normalized_coords = 1; + sampler.min_lod = 0.0f; + sampler.max_lod = MAX_TEXTURE_LEVEL; + + aaline->sampler_cso = pipe->create_sampler_state(pipe, &sampler); +} + + +/** + * When we're about to draw our first AA line in a batch, this function is + * called to tell the driver to bind our modified fragment shader. + */ +static void +bind_aaline_fragment_shader(struct aaline_stage *aaline) +{ + if (!aaline->fs->aaline_fs) { + generate_aaline_fs(aaline); + } + aaline->driver_bind_fs_state(aaline->pipe, aaline->fs->aaline_fs); +} + + + +static INLINE struct aaline_stage * +aaline_stage( struct draw_stage *stage ) +{ + return (struct aaline_stage *) stage; +} + + +static void +passthrough_point(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->point(stage->next, header); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Draw a wide line by drawing a quad, using geometry which will + * fullfill GL's antialiased line requirements. + */ +static void +aaline_line(struct draw_stage *stage, struct prim_header *header) +{ + const struct aaline_stage *aaline = aaline_stage(stage); + const float half_width = aaline->half_line_width; + struct prim_header tri; + struct vertex_header *v[8]; + uint texPos = aaline->tex_slot; + float *pos, *tex; + float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; + float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; + float a = atan2(dy, dx); + float c_a = cos(a), s_a = sin(a); + uint i; + + /* XXX the ends of lines aren't quite perfect yet, but probably passable */ + dx = 0.5 * half_width; + dy = half_width; + + /* allocate/dup new verts */ + for (i = 0; i < 8; i++) { + v[i] = dup_vert(stage, header->v[i/4], i); + } + + /* + * Quad strip for line from v0 to v1 (*=endpoints): + * + * 1 3 5 7 + * +---+---------------------+---+ + * | | + * | *v0 v1* | + * | | + * +---+---------------------+---+ + * 0 2 4 6 + */ + + /* new verts */ + pos = v[0]->data[0]; + pos[0] += (-dx * c_a - dy * s_a); + pos[1] += (-dx * s_a + dy * c_a); + + pos = v[1]->data[0]; + pos[0] += (-dx * c_a - -dy * s_a); + pos[1] += (-dx * s_a + -dy * c_a); + + pos = v[2]->data[0]; + pos[0] += ( dx * c_a - dy * s_a); + pos[1] += ( dx * s_a + dy * c_a); + + pos = v[3]->data[0]; + pos[0] += ( dx * c_a - -dy * s_a); + pos[1] += ( dx * s_a + -dy * c_a); + + pos = v[4]->data[0]; + pos[0] += (-dx * c_a - dy * s_a); + pos[1] += (-dx * s_a + dy * c_a); + + pos = v[5]->data[0]; + pos[0] += (-dx * c_a - -dy * s_a); + pos[1] += (-dx * s_a + -dy * c_a); + + pos = v[6]->data[0]; + pos[0] += ( dx * c_a - dy * s_a); + pos[1] += ( dx * s_a + dy * c_a); + + pos = v[7]->data[0]; + pos[0] += ( dx * c_a - -dy * s_a); + pos[1] += ( dx * s_a + -dy * c_a); + + /* new texcoords */ + tex = v[0]->data[texPos]; + ASSIGN_4V(tex, 0, 0, 0, 1); + + tex = v[1]->data[texPos]; + ASSIGN_4V(tex, 0, 1, 0, 1); + + tex = v[2]->data[texPos]; + ASSIGN_4V(tex, .5, 0, 0, 1); + + tex = v[3]->data[texPos]; + ASSIGN_4V(tex, .5, 1, 0, 1); + + tex = v[4]->data[texPos]; + ASSIGN_4V(tex, .5, 0, 0, 1); + + tex = v[5]->data[texPos]; + ASSIGN_4V(tex, .5, 1, 0, 1); + + tex = v[6]->data[texPos]; + ASSIGN_4V(tex, 1, 0, 0, 1); + + tex = v[7]->data[texPos]; + ASSIGN_4V(tex, 1, 1, 0, 1); + + /* emit 6 tris for the quad strip */ + tri.v[0] = v[2]; tri.v[1] = v[1]; tri.v[2] = v[0]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[3]; tri.v[1] = v[1]; tri.v[2] = v[2]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[4]; tri.v[1] = v[3]; tri.v[2] = v[2]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[5]; tri.v[1] = v[3]; tri.v[2] = v[4]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[6]; tri.v[1] = v[5]; tri.v[2] = v[4]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[7]; tri.v[1] = v[5]; tri.v[2] = v[6]; + stage->next->tri( stage->next, &tri ); +} + + +static void +aaline_first_line(struct draw_stage *stage, struct prim_header *header) +{ + auto struct aaline_stage *aaline = aaline_stage(stage); + struct draw_context *draw = stage->draw; + struct pipe_context *pipe = aaline->pipe; + + assert(draw->rasterizer->line_smooth); + + if (draw->rasterizer->line_width <= 3.0) + aaline->half_line_width = 1.5f; + else + aaline->half_line_width = 0.5f * draw->rasterizer->line_width; + + aaline->tex_slot = draw->num_vs_outputs; + assert(aaline->tex_slot > 0); /* output[0] is vertex pos */ + draw->extra_vp_outputs.slot = aaline->tex_slot; + + /* + * Bind our fragprog, sampler and texture + */ + bind_aaline_fragment_shader(aaline); + + aaline->driver_bind_sampler_state(pipe, aaline->sampler_unit, aaline->sampler_cso); + aaline->driver_set_sampler_texture(pipe, aaline->sampler_unit, aaline->texture); + + /* now really draw first line */ + stage->line = aaline_line; + stage->line(stage, header); +} + + +static void +aaline_flush(struct draw_stage *stage, unsigned flags) +{ + struct draw_context *draw = stage->draw; + struct aaline_stage *aaline = aaline_stage(stage); + struct pipe_context *pipe = aaline->pipe; + + stage->line = aaline_first_line; + stage->next->flush( stage->next, flags ); + + /* restore original frag shader */ + aaline->driver_bind_fs_state(pipe, aaline->fs->driver_fs); + + /* XXX restore original texture, sampler state */ + aaline->driver_bind_sampler_state(pipe, aaline->sampler_unit, + aaline->state.sampler[aaline->sampler_unit]); + aaline->driver_set_sampler_texture(pipe, aaline->sampler_unit, + aaline->state.texture[aaline->sampler_unit]); + + draw->extra_vp_outputs.slot = 0; +} + + +static void +aaline_reset_stipple_counter(struct draw_stage *stage) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +aaline_destroy(struct draw_stage *stage) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +static struct aaline_stage * +draw_aaline_stage(struct draw_context *draw) +{ + struct aaline_stage *aaline = CALLOC_STRUCT(aaline_stage); + + draw_alloc_temp_verts( &aaline->stage, 8 ); + + aaline->stage.draw = draw; + aaline->stage.next = NULL; + aaline->stage.point = passthrough_point; + aaline->stage.line = aaline_first_line; + aaline->stage.tri = passthrough_tri; + aaline->stage.flush = aaline_flush; + aaline->stage.reset_stipple_counter = aaline_reset_stipple_counter; + aaline->stage.destroy = aaline_destroy; + + return aaline; +} + + +/* + * XXX temporary? solution to mapping a pipe_context to a aaline_stage. + */ + +#define MAX_CONTEXTS 10 + +static struct pipe_context *Pipe[MAX_CONTEXTS]; +static struct aaline_stage *Stage[MAX_CONTEXTS]; +static uint NumContexts; + +static void +add_aa_pipe_context(struct pipe_context *pipe, struct aaline_stage *aa) +{ + assert(NumContexts < MAX_CONTEXTS); + Pipe[NumContexts] = pipe; + Stage[NumContexts] = aa; + NumContexts++; +} + +static struct aaline_stage * +aaline_stage_from_pipe(struct pipe_context *pipe) +{ + uint i; + for (i = 0; i < NumContexts; i++) { + if (Pipe[i] == pipe) + return Stage[i]; + } + assert(0); + return NULL; +} + + +/** + * This function overrides the driver's create_fs_state() function and + * will typically be called by the state tracker. + */ +static void * +aaline_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *fs) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + struct aaline_fragment_shader *aafs = CALLOC_STRUCT(aaline_fragment_shader); + + if (aafs) { + aafs->state = *fs; + + /* pass-through */ + aafs->driver_fs = aaline->driver_create_fs_state(aaline->pipe, fs); + } + + return aafs; +} + + +static void +aaline_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs; + /* save current */ + aaline->fs = aafs; + /* pass-through */ + aaline->driver_bind_fs_state(aaline->pipe, aafs->driver_fs); +} + + +static void +aaline_delete_fs_state(struct pipe_context *pipe, void *fs) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs; + /* pass-through */ + aaline->driver_delete_fs_state(aaline->pipe, aafs->driver_fs); + FREE(aafs); +} + + +static void +aaline_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + /* save current */ + aaline->state.sampler[unit] = sampler; + /* pass-through */ + aaline->driver_bind_sampler_state(aaline->pipe, unit, sampler); +} + + +static void +aaline_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, struct pipe_texture *texture) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + /* save current */ + aaline->state.texture[sampler] = texture; + /* pass-through */ + aaline->driver_set_sampler_texture(aaline->pipe, sampler, texture); +} + + +/** + * Called by drivers that want to install this AA line prim stage + * into the draw module's pipeline. This will not be used if the + * hardware has native support for AA lines. + */ +void +draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) +{ + struct aaline_stage *aaline; + + /* + * Create / install AA line drawing / prim stage + */ + aaline = draw_aaline_stage( draw ); + assert(aaline); + draw->pipeline.aaline = &aaline->stage; + + aaline->pipe = pipe; + + /* create special texture, sampler state */ + aaline_create_texture(aaline); + aaline_create_sampler(aaline); + + /* save original driver functions */ + aaline->driver_create_fs_state = pipe->create_fs_state; + aaline->driver_bind_fs_state = pipe->bind_fs_state; + aaline->driver_delete_fs_state = pipe->delete_fs_state; + + aaline->driver_bind_sampler_state = pipe->bind_sampler_state; + aaline->driver_set_sampler_texture = pipe->set_sampler_texture; + + /* override the driver's functions */ + pipe->create_fs_state = aaline_create_fs_state; + pipe->bind_fs_state = aaline_bind_fs_state; + pipe->delete_fs_state = aaline_delete_fs_state; + + pipe->bind_sampler_state = aaline_bind_sampler_state; + pipe->set_sampler_texture = aaline_set_sampler_texture; + + add_aa_pipe_context(pipe, aaline); +} diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 4be3830316..a7f3b4aecb 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -103,6 +103,8 @@ void draw_destroy( struct draw_context *draw ) draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); draw->pipeline.cull->destroy( draw->pipeline.cull ); draw->pipeline.validate->destroy( draw->pipeline.validate ); + if (draw->pipeline.aaline) + draw->pipeline.aaline->destroy( draw->pipeline.aaline ); if (draw->pipeline.rasterize) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); tgsi_exec_machine_free_data(&draw->machine); @@ -239,6 +241,26 @@ draw_convert_wide_lines(struct draw_context *draw, boolean enable) } +/** + * The draw module may sometimes generate vertices with extra attributes + * (such as texcoords for AA lines). The driver can call this function + * to find those attributes. + */ +int +draw_find_vs_output(struct draw_context *draw, + uint semantic_name, uint semantic_index) +{ + /* XXX there may be more than one extra vertex attrib. + * For example, simulated gl_FragCoord and gl_PointCoord. + */ + if (draw->extra_vp_outputs.semantic_name == semantic_name && + draw->extra_vp_outputs.semantic_index == semantic_index) { + return draw->extra_vp_outputs.slot; + } + return 0; +} + + /** * Allocate space for temporary post-transform vertices, such as for clipping. */ diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index ddeb184497..82035aa8ad 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -41,6 +41,7 @@ #include "pipe/p_state.h" +struct pipe_context; struct vertex_buffer; struct vertex_info; struct draw_context; @@ -93,6 +94,20 @@ void draw_convert_wide_points(struct draw_context *draw, boolean enable); void draw_convert_wide_lines(struct draw_context *draw, boolean enable); +boolean draw_use_sse(struct draw_context *draw); + +void +draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe); + + +int +draw_find_vs_output(struct draw_context *draw, + uint semantic_name, uint semantic_index); + + +/* + * Vertex shader functions + */ struct draw_vertex_shader * draw_create_vertex_shader(struct draw_context *draw, @@ -102,7 +117,11 @@ void draw_bind_vertex_shader(struct draw_context *draw, void draw_delete_vertex_shader(struct draw_context *draw, struct draw_vertex_shader *dvs); -boolean draw_use_sse(struct draw_context *draw); + + +/* + * Vertex data functions + */ void draw_set_vertex_buffer(struct draw_context *draw, unsigned attr, diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index 51e2242719..dd9a848863 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -121,11 +121,15 @@ static void draw_prim_queue_flush( struct draw_context *draw ) void draw_do_flush( struct draw_context *draw, unsigned flags ) { + static boolean flushing = FALSE; + if (0) debug_printf("Flushing with %d verts, %d prims\n", draw->vs.queue_nr, draw->pq.queue_nr ); + if (!flushing) { + flushing = TRUE; if (flags >= DRAW_FLUSH_SHADER_QUEUE) { if (draw->vs.queue_nr) @@ -146,6 +150,9 @@ void draw_do_flush( struct draw_context *draw, unsigned flags ) } } } + + flushing = FALSE; + } } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index bc11259cb2..dd75f9aefc 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -48,6 +48,7 @@ #include "tgsi/exec/tgsi_exec.h" +struct pipe_context; struct gallivm_prog; struct gallivm_cpu_engine; @@ -179,6 +180,7 @@ struct draw_context struct draw_stage *offset; struct draw_stage *unfilled; struct draw_stage *stipple; + struct draw_stage *aaline; struct draw_stage *wide; struct draw_stage *rasterize; } pipeline; @@ -212,9 +214,17 @@ struct draw_context unsigned nr_planes; boolean convert_wide_points; /**< convert wide points to tris? */ - boolean convert_wide_lines; /**< convert side lines to tris? */ + boolean convert_wide_lines; /**< convert wide lines to tris? */ boolean use_sse; + /* If a prim stage introduces new vertex attributes, they'll be stored here + */ + struct { + uint semantic_name; + uint semantic_index; + int slot; + } extra_vp_outputs; + unsigned reduced_prim; /** TGSI program interpreter runtime state */ diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index 4375ebabbc..97e40e372b 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -58,7 +58,13 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) * shorter pipelines for lines & points. */ - if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines) || + if (draw->rasterizer->line_smooth && draw->pipeline.aaline) { + draw->pipeline.aaline->next = next; + next = draw->pipeline.aaline; + } + + if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines + && !draw->rasterizer->line_smooth) || (draw->rasterizer->point_size != 1.0 && draw->convert_wide_points) || draw->rasterizer->point_sprite) { draw->pipeline.wide->next = next; diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index c10ab396d8..8bb62b2a0a 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -10,6 +10,7 @@ DRIVER_SOURCES = \ util/tgsi_build.c \ util/tgsi_dump.c \ util/tgsi_parse.c \ + util/tgsi_transform.c \ util/tgsi_util.c C_SOURCES = \ diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 5e98f190bb..254c6adca4 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -327,6 +327,9 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, draw_set_rasterize_stage(softpipe->draw, softpipe->setup); } + /* enable aaline stage */ + draw_install_aaline_stage(softpipe->draw, &softpipe->pipe); + sp_init_surface_functions(softpipe); return &softpipe->pipe; diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 9d8fd8b750..f9f2c5eaa8 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -44,7 +44,8 @@ * condition that users shouldn't hit anyway. */ static int -find_vs_output(const struct pipe_shader_state *vs, +find_vs_output(struct softpipe_context *sp, + const struct pipe_shader_state *vs, uint semantic_name, uint semantic_index) { @@ -54,7 +55,9 @@ find_vs_output(const struct pipe_shader_state *vs, vs->output_semantic_index[i] == semantic_index) return i; } - return 0; + + /* See if the draw module is introducing a new attribute... */ + return draw_find_vs_output(sp->draw, semantic_name, semantic_index); } @@ -111,24 +114,24 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) int src; switch (fs->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: - src = find_vs_output(vs, TGSI_SEMANTIC_POSITION, 0); + src = find_vs_output(softpipe, vs, TGSI_SEMANTIC_POSITION, 0); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); break; case TGSI_SEMANTIC_COLOR: - src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, + src = find_vs_output(softpipe, vs, TGSI_SEMANTIC_COLOR, fs->input_semantic_index[i]); draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); break; case TGSI_SEMANTIC_FOG: - src = find_vs_output(vs, TGSI_SEMANTIC_FOG, 0); + src = find_vs_output(softpipe, vs, TGSI_SEMANTIC_FOG, 0); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ - src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, + src = find_vs_output(softpipe, vs, TGSI_SEMANTIC_GENERIC, fs->input_semantic_index[i]); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; @@ -138,7 +141,7 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) } } - softpipe->psize_slot = find_vs_output(vs, TGSI_SEMANTIC_PSIZE, 0); + softpipe->psize_slot = find_vs_output(softpipe, vs, TGSI_SEMANTIC_PSIZE, 0); if (softpipe->psize_slot > 0) { draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, softpipe->psize_slot); -- cgit v1.2.3 From 478c14453b33b20724bcdb70cf1f54f4addb71ab Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 18 Feb 2008 16:50:07 -0700 Subject: gallium: remove the prototype/unused wide_line_aa() function --- src/gallium/auxiliary/draw/draw_wide_prims.c | 69 +--------------------------- 1 file changed, 1 insertion(+), 68 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_wide_prims.c b/src/gallium/auxiliary/draw/draw_wide_prims.c index 655774b155..1f8069bdca 100644 --- a/src/gallium/auxiliary/draw/draw_wide_prims.c +++ b/src/gallium/auxiliary/draw/draw_wide_prims.c @@ -161,70 +161,6 @@ static void wide_line( struct draw_stage *stage, } -/** - * Draw a wide line by drawing a quad, using geometry which will - * fullfill GL's antialiased line requirements. - */ -static void wide_line_aa(struct draw_stage *stage, - struct prim_header *header) -{ - const struct wide_stage *wide = wide_stage(stage); - const float half_width = wide->half_line_width; - struct prim_header tri; - struct vertex_header *v[4]; - float *pos; - float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; - float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; - const float len = (float) sqrt(dx * dx + dy * dy); - uint i; - - dx = dx * half_width / len; - dy = dy * half_width / len; - - /* allocate/dup new verts */ - for (i = 0; i < 4; i++) { - v[i] = dup_vert(stage, header->v[i/2], i); - } - - /* - * Quad for line from v0 to v1: - * - * 1 3 - * +-------------------------+ - * | | - * *v0 v1* - * | | - * +-------------------------+ - * 0 2 - */ - - pos = v[0]->data[0]; - pos[0] += dy; - pos[1] -= dx; - - pos = v[1]->data[0]; - pos[0] -= dy; - pos[1] += dx; - - pos = v[2]->data[0]; - pos[0] += dy; - pos[1] -= dx; - - pos = v[3]->data[0]; - pos[0] -= dy; - pos[1] += dx; - - tri.det = header->det; /* only the sign matters */ - - tri.v[0] = v[2]; tri.v[1] = v[1]; tri.v[2] = v[0]; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v[3]; tri.v[1] = v[1]; tri.v[2] = v[2]; - stage->next->tri( stage->next, &tri ); - -} - - /** * Set the vertex texcoords for sprite mode. * Coords may be left untouched or set to a right-side-up or upside-down @@ -379,10 +315,7 @@ static void wide_first_line( struct draw_stage *stage, wide->half_line_width = 0.5f * draw->rasterizer->line_width; if (draw->rasterizer->line_width != 1.0) { - if (draw->rasterizer->line_smooth) - wide->stage.line = wide_line_aa; - else - wide->stage.line = wide_line; + wide->stage.line = wide_line; } else { wide->stage.line = passthrough_line; -- cgit v1.2.3 From ae9931dad24703d99530b2761c759cef1cbc0fb5 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 18 Feb 2008 18:36:20 -0700 Subject: gallium: call draw_flush() for scissor/stipple state changes --- src/gallium/drivers/softpipe/sp_state_clip.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_clip.c b/src/gallium/drivers/softpipe/sp_state_clip.c index c797c0dd3b..d8ea979957 100644 --- a/src/gallium/drivers/softpipe/sp_state_clip.c +++ b/src/gallium/drivers/softpipe/sp_state_clip.c @@ -68,6 +68,8 @@ void softpipe_set_scissor_state( struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); + draw_flush(softpipe->draw); + memcpy( &softpipe->scissor, scissor, sizeof(*scissor) ); softpipe->dirty |= SP_NEW_SCISSOR; } @@ -78,6 +80,8 @@ void softpipe_set_polygon_stipple( struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); + draw_flush(softpipe->draw); + memcpy( &softpipe->poly_stipple, stipple, sizeof(*stipple) ); softpipe->dirty |= SP_NEW_STIPPLE; } -- cgit v1.2.3 From 6c7f663cb96cac4873129f835614181405bd3f6e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 18 Feb 2008 18:39:55 -0700 Subject: gallium: move draw_set_viewport_state() call, plus code clean-up, remove obsolete comments --- src/gallium/drivers/softpipe/sp_state_clip.c | 16 ++++------------ 1 file changed, 4 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_clip.c b/src/gallium/drivers/softpipe/sp_state_clip.c index d8ea979957..4946c776e3 100644 --- a/src/gallium/drivers/softpipe/sp_state_clip.c +++ b/src/gallium/drivers/softpipe/sp_state_clip.c @@ -42,24 +42,16 @@ void softpipe_set_clip_state( struct pipe_context *pipe, } - -/* Called when driver state tracker notices changes to the viewport - * matrix: - */ void softpipe_set_viewport_state( struct pipe_context *pipe, const struct pipe_viewport_state *viewport ) { struct softpipe_context *softpipe = softpipe_context(pipe); - softpipe->viewport = *viewport; /* struct copy */ - softpipe->dirty |= SP_NEW_VIEWPORT; - /* pass the viewport info to the draw module */ draw_set_viewport_state(softpipe->draw, viewport); - /* Using tnl/ and vf/ modules is temporary while getting started. - * Full pipe will have vertex shader, vertex fetch of its own. - */ + softpipe->viewport = *viewport; /* struct copy */ + softpipe->dirty |= SP_NEW_VIEWPORT; } @@ -70,7 +62,7 @@ void softpipe_set_scissor_state( struct pipe_context *pipe, draw_flush(softpipe->draw); - memcpy( &softpipe->scissor, scissor, sizeof(*scissor) ); + softpipe->scissor = *scissor; /* struct copy */ softpipe->dirty |= SP_NEW_SCISSOR; } @@ -82,6 +74,6 @@ void softpipe_set_polygon_stipple( struct pipe_context *pipe, draw_flush(softpipe->draw); - memcpy( &softpipe->poly_stipple, stipple, sizeof(*stipple) ); + softpipe->poly_stipple = *stipple; /* struct copy */ softpipe->dirty |= SP_NEW_STIPPLE; } -- cgit v1.2.3 From e279b1c57ac8d17703d80f1b644fd1d4f115101b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 18 Feb 2008 14:35:19 +0000 Subject: More llvm -> gallivm. Forgot this one on the last commit. --- src/gallium/drivers/cell/ppu/cell_state_fs.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_fs.c b/src/gallium/drivers/cell/ppu/cell_state_fs.c index b2ed699a5b..f3958fa429 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_fs.c +++ b/src/gallium/drivers/cell/ppu/cell_state_fs.c @@ -32,7 +32,7 @@ #include "draw/draw_context.h" #if 0 #include "pipe/p_shader_tokens.h" -#include "llvm/gallivm.h" +#include "gallivm/gallivm.h" #include "tgsi/util/tgsi_dump.h" #include "tgsi/exec/tgsi_sse2.h" #endif -- cgit v1.2.3 From e773a813cf475e2a7ad79ea1ec698bf2530d0433 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 10:50:39 +0900 Subject: Initial scons support to build gallivm. Not yet complete. --- src/gallium/SConscript | 6 +++++- src/gallium/auxiliary/gallivm/SConscript | 16 ++++++++++++++++ 2 files changed, 21 insertions(+), 1 deletion(-) create mode 100644 src/gallium/auxiliary/gallivm/SConscript (limited to 'src') diff --git a/src/gallium/SConscript b/src/gallium/SConscript index ea55ed2ed5..c505eee792 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -13,9 +13,13 @@ SConscript([ 'auxiliary/tgsi/SConscript', 'auxiliary/cso_cache/SConscript', 'auxiliary/draw/SConscript', - #'auxiliary/gallivm/SConscript', 'auxiliary/pipebuffer/SConscript', +]) + +if llvm: + SConscript(['auxiliary/gallivm/SConscript']) +SConscript([ 'drivers/softpipe/SConscript', 'drivers/i915simple/SConscript', 'drivers/i965simple/SConscript', diff --git a/src/gallium/auxiliary/gallivm/SConscript b/src/gallium/auxiliary/gallivm/SConscript new file mode 100644 index 0000000000..c0aa51b90a --- /dev/null +++ b/src/gallium/auxiliary/gallivm/SConscript @@ -0,0 +1,16 @@ +Import('*') + +gallivm = env.ConvenienceLibrary( + target = 'gallivm', + source = [ + 'gallivm.cpp', + 'gallivm_cpu.cpp', + 'instructions.cpp', + 'loweringpass.cpp', + 'tgsitollvm.cpp', + 'storage.cpp', + 'storagesoa.cpp', + 'instructionssoa.cpp', + ]) + +auxiliaries.insert(0, gallivm) -- cgit v1.2.3 From df8ab3140ce05599e1dc983ac211a30fc845d9b5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 11:49:48 +0900 Subject: Bring rtasm from mesa to gallium. --- src/gallium/SConscript | 1 + src/gallium/auxiliary/rtasm/Makefile | 20 + src/gallium/auxiliary/rtasm/SConscript | 11 + src/gallium/auxiliary/rtasm/execmem.c | 133 ++++ src/gallium/auxiliary/rtasm/execmem.h | 45 ++ src/gallium/auxiliary/rtasm/mm.c | 283 ++++++++ src/gallium/auxiliary/rtasm/mm.h | 89 +++ src/gallium/auxiliary/rtasm/x86sse.c | 1195 ++++++++++++++++++++++++++++++++ src/gallium/auxiliary/rtasm/x86sse.h | 256 +++++++ 9 files changed, 2033 insertions(+) create mode 100644 src/gallium/auxiliary/rtasm/Makefile create mode 100644 src/gallium/auxiliary/rtasm/SConscript create mode 100644 src/gallium/auxiliary/rtasm/execmem.c create mode 100644 src/gallium/auxiliary/rtasm/execmem.h create mode 100644 src/gallium/auxiliary/rtasm/mm.c create mode 100644 src/gallium/auxiliary/rtasm/mm.h create mode 100644 src/gallium/auxiliary/rtasm/x86sse.c create mode 100644 src/gallium/auxiliary/rtasm/x86sse.h (limited to 'src') diff --git a/src/gallium/SConscript b/src/gallium/SConscript index c505eee792..a08b4b830e 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -10,6 +10,7 @@ Export('auxiliaries') SConscript([ # NOTE: order matters! 'auxiliary/util/SConscript', + 'auxiliary/rtasm/SConscript', 'auxiliary/tgsi/SConscript', 'auxiliary/cso_cache/SConscript', 'auxiliary/draw/SConscript', diff --git a/src/gallium/auxiliary/rtasm/Makefile b/src/gallium/auxiliary/rtasm/Makefile new file mode 100644 index 0000000000..b3b9934e10 --- /dev/null +++ b/src/gallium/auxiliary/rtasm/Makefile @@ -0,0 +1,20 @@ + +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = rtasm + +DRIVER_SOURCES = \ + x86sse.c \ + mm.c \ + execmem.c + +C_SOURCES = \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/auxiliary/rtasm/SConscript b/src/gallium/auxiliary/rtasm/SConscript new file mode 100644 index 0000000000..c5b1551786 --- /dev/null +++ b/src/gallium/auxiliary/rtasm/SConscript @@ -0,0 +1,11 @@ +Import('*') + +rtasm = env.ConvenienceLibrary( + target = 'rtasm', + source = [ + 'x86sse.c', + 'mm.c', + 'execmem.c', + ]) + +auxiliaries.insert(0, rtasm) diff --git a/src/gallium/auxiliary/rtasm/execmem.c b/src/gallium/auxiliary/rtasm/execmem.c new file mode 100644 index 0000000000..c7c35f7ef2 --- /dev/null +++ b/src/gallium/auxiliary/rtasm/execmem.c @@ -0,0 +1,133 @@ +/************************************************************************** + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * \file exemem.c + * Functions for allocating executable memory. + * + * \author Keith Whitwell + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_thread.h" + +#include "execmem.h" + + +#if defined(__linux__) + +/* + * Allocate a large block of memory which can hold code then dole it out + * in pieces by means of the generic memory manager code. +*/ + +#include +#include +#include "mm.h" + +#define EXEC_HEAP_SIZE (10*1024*1024) + +_glthread_DECLARE_STATIC_MUTEX(exec_mutex); + +static struct mem_block *exec_heap = NULL; +static unsigned char *exec_mem = NULL; + + +static void +init_heap(void) +{ + if (!exec_heap) + exec_heap = mmInit( 0, EXEC_HEAP_SIZE ); + + if (!exec_mem) + exec_mem = (unsigned char *) mmap(0, EXEC_HEAP_SIZE, + PROT_EXEC | PROT_READ | PROT_WRITE, + MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); +} + + +void * +_mesa_exec_malloc(size_t size) +{ + struct mem_block *block = NULL; + void *addr = NULL; + + _glthread_LOCK_MUTEX(exec_mutex); + + init_heap(); + + if (exec_heap) { + size = (size + 31) & ~31; + block = mmAllocMem( exec_heap, size, 32, 0 ); + } + + if (block) + addr = exec_mem + block->ofs; + else + debug_printf("_mesa_exec_malloc failed\n"); + + _glthread_UNLOCK_MUTEX(exec_mutex); + + return addr; +} + + +void +_mesa_exec_free(void *addr) +{ + _glthread_LOCK_MUTEX(exec_mutex); + + if (exec_heap) { + struct mem_block *block = mmFindBlock(exec_heap, (unsigned char *)addr - exec_mem); + + if (block) + mmFreeMem(block); + } + + _glthread_UNLOCK_MUTEX(exec_mutex); +} + + +#else + +/* + * Just use regular memory. + */ + +void * +_mesa_exec_malloc(GLuint size) +{ + return _mesa_malloc( size ); +} + + +void +_mesa_exec_free(void *addr) +{ + _mesa_free(addr); +} + + +#endif diff --git a/src/gallium/auxiliary/rtasm/execmem.h b/src/gallium/auxiliary/rtasm/execmem.h new file mode 100644 index 0000000000..9fd4569165 --- /dev/null +++ b/src/gallium/auxiliary/rtasm/execmem.h @@ -0,0 +1,45 @@ +/************************************************************************** + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file exemem.c + * Functions for allocating executable memory. + * + * \author Keith Whitwell + */ + +#ifndef _EXECMEM_H_ +#define _EXECMEM_H_ + +#include "pipe/p_compiler.h" + + +extern void * +_mesa_exec_malloc( size_t size ); + + +extern void +_mesa_exec_free( void *addr ); + + +#endif diff --git a/src/gallium/auxiliary/rtasm/mm.c b/src/gallium/auxiliary/rtasm/mm.c new file mode 100644 index 0000000000..15f50491da --- /dev/null +++ b/src/gallium/auxiliary/rtasm/mm.c @@ -0,0 +1,283 @@ +/* + * GLX Hardware Device Driver common code + * Copyright (C) 1999 Wittawat Yamwong + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * WITTAWAT YAMWONG, OR ANY OTHER CONTRIBUTORS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE + * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_util.h" +#include "pipe/p_debug.h" + +#include "mm.h" + + +void +mmDumpMemInfo(const struct mem_block *heap) +{ + debug_printf("Memory heap %p:\n", (void *)heap); + if (heap == 0) { + debug_printf(" heap == 0\n"); + } else { + const struct mem_block *p; + + for(p = heap->next; p != heap; p = p->next) { + debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + debug_printf("\nFree list:\n"); + + for(p = heap->next_free; p != heap; p = p->next_free) { + debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + } + debug_printf("End of memory blocks\n"); +} + +struct mem_block * +mmInit(int ofs, int size) +{ + struct mem_block *heap, *block; + + if (size <= 0) + return NULL; + + heap = CALLOC_STRUCT(mem_block); + if (!heap) + return NULL; + + block = CALLOC_STRUCT(mem_block); + if (!block) { + FREE(heap); + return NULL; + } + + heap->next = block; + heap->prev = block; + heap->next_free = block; + heap->prev_free = block; + + block->heap = heap; + block->next = heap; + block->prev = heap; + block->next_free = heap; + block->prev_free = heap; + + block->ofs = ofs; + block->size = size; + block->free = 1; + + return heap; +} + + +static struct mem_block * +SliceBlock(struct mem_block *p, + int startofs, int size, + int reserved, int alignment) +{ + struct mem_block *newblock; + + /* break left [p, newblock, p->next], then p = newblock */ + if (startofs > p->ofs) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs; + newblock->size = p->size - (startofs - p->ofs); + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size -= newblock->size; + p = newblock; + } + + /* break right, also [p, newblock, p->next] */ + if (size < p->size) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs + size; + newblock->size = p->size - size; + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size = size; + } + + /* p = middle block */ + p->free = 0; + + /* Remove p from the free list: + */ + p->next_free->prev_free = p->prev_free; + p->prev_free->next_free = p->next_free; + + p->next_free = 0; + p->prev_free = 0; + + p->reserved = reserved; + return p; +} + + +struct mem_block * +mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) +{ + struct mem_block *p; + const int mask = (1 << align2)-1; + int startofs = 0; + int endofs; + + if (!heap || align2 < 0 || size <= 0) + return NULL; + + for (p = heap->next_free; p != heap; p = p->next_free) { + assert(p->free); + + startofs = (p->ofs + mask) & ~mask; + if ( startofs < startSearch ) { + startofs = startSearch; + } + endofs = startofs+size; + if (endofs <= (p->ofs+p->size)) + break; + } + + if (p == heap) + return NULL; + + assert(p->free); + p = SliceBlock(p,startofs,size,0,mask+1); + + return p; +} + + +struct mem_block * +mmFindBlock(struct mem_block *heap, int start) +{ + struct mem_block *p; + + for (p = heap->next; p != heap; p = p->next) { + if (p->ofs == start) + return p; + } + + return NULL; +} + + +static INLINE int +Join2Blocks(struct mem_block *p) +{ + /* XXX there should be some assertions here */ + + /* NOTE: heap->free == 0 */ + + if (p->free && p->next->free) { + struct mem_block *q = p->next; + + assert(p->ofs + p->size == q->ofs); + p->size += q->size; + + p->next = q->next; + q->next->prev = p; + + q->next_free->prev_free = q->prev_free; + q->prev_free->next_free = q->next_free; + + FREE(q); + return 1; + } + return 0; +} + +int +mmFreeMem(struct mem_block *b) +{ + if (!b) + return 0; + + if (b->free) { + debug_printf("block already free\n"); + return -1; + } + if (b->reserved) { + debug_printf("block is reserved\n"); + return -1; + } + + b->free = 1; + b->next_free = b->heap->next_free; + b->prev_free = b->heap; + b->next_free->prev_free = b; + b->prev_free->next_free = b; + + Join2Blocks(b); + if (b->prev != b->heap) + Join2Blocks(b->prev); + + return 0; +} + + +void +mmDestroy(struct mem_block *heap) +{ + struct mem_block *p; + + if (!heap) + return; + + for (p = heap->next; p != heap; ) { + struct mem_block *next = p->next; + FREE(p); + p = next; + } + + FREE(heap); +} diff --git a/src/gallium/auxiliary/rtasm/mm.h b/src/gallium/auxiliary/rtasm/mm.h new file mode 100644 index 0000000000..f469b18d3e --- /dev/null +++ b/src/gallium/auxiliary/rtasm/mm.h @@ -0,0 +1,89 @@ +/* + * GLX Hardware Device Driver common code + * Copyright (C) 1999 Wittawat Yamwong + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * KEITH WHITWELL, OR ANY OTHER CONTRIBUTORS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE + * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * Memory manager code. Primarily used by device drivers to manage texture + * heaps, etc. + */ + + +#ifndef MM_H +#define MM_H + + +struct mem_block { + struct mem_block *next, *prev; + struct mem_block *next_free, *prev_free; + struct mem_block *heap; + int ofs,size; + unsigned int free:1; + unsigned int reserved:1; +}; + + + +/** + * input: total size in bytes + * return: a heap pointer if OK, NULL if error + */ +extern struct mem_block *mmInit(int ofs, int size); + +/** + * Allocate 'size' bytes with 2^align2 bytes alignment, + * restrict the search to free memory after 'startSearch' + * depth and back buffers should be in different 4mb banks + * to get better page hits if possible + * input: size = size of block + * align2 = 2^align2 bytes alignment + * startSearch = linear offset from start of heap to begin search + * return: pointer to the allocated block, 0 if error + */ +extern struct mem_block *mmAllocMem(struct mem_block *heap, int size, int align2, + int startSearch); + +/** + * Free block starts at offset + * input: pointer to a block + * return: 0 if OK, -1 if error + */ +extern int mmFreeMem(struct mem_block *b); + +/** + * Free block starts at offset + * input: pointer to a heap, start offset + * return: pointer to a block + */ +extern struct mem_block *mmFindBlock(struct mem_block *heap, int start); + +/** + * destroy MM + */ +extern void mmDestroy(struct mem_block *mmInit); + +/** + * For debuging purpose. + */ +extern void mmDumpMemInfo(const struct mem_block *mmInit); + +#endif diff --git a/src/gallium/auxiliary/rtasm/x86sse.c b/src/gallium/auxiliary/rtasm/x86sse.c new file mode 100644 index 0000000000..fff6f77a6b --- /dev/null +++ b/src/gallium/auxiliary/rtasm/x86sse.c @@ -0,0 +1,1195 @@ +#if defined(__i386__) || defined(__386__) + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" + +#include "x86sse.h" + +#define DISASSEM 0 +#define X86_TWOB 0x0f + +static unsigned char *cptr( void (*label)() ) +{ + return (unsigned char *)(unsigned long)label; +} + + +static void do_realloc( struct x86_function *p ) +{ + if (p->size == 0) { + p->size = 1024; + p->store = _mesa_exec_malloc(p->size); + p->csr = p->store; + } + else { + unsigned used = p->csr - p->store; + unsigned char *tmp = p->store; + p->size *= 2; + p->store = _mesa_exec_malloc(p->size); + memcpy(p->store, tmp, used); + p->csr = p->store + used; + _mesa_exec_free(tmp); + } +} + +/* Emit bytes to the instruction stream: + */ +static unsigned char *reserve( struct x86_function *p, int bytes ) +{ + if (p->csr + bytes - p->store > p->size) + do_realloc(p); + + { + unsigned char *csr = p->csr; + p->csr += bytes; + return csr; + } +} + + + +static void emit_1b( struct x86_function *p, char b0 ) +{ + char *csr = (char *)reserve(p, 1); + *csr = b0; +} + +static void emit_1i( struct x86_function *p, int i0 ) +{ + int *icsr = (int *)reserve(p, sizeof(i0)); + *icsr = i0; +} + +static void emit_1ub( struct x86_function *p, unsigned char b0 ) +{ + unsigned char *csr = reserve(p, 1); + *csr++ = b0; +} + +static void emit_2ub( struct x86_function *p, unsigned char b0, unsigned char b1 ) +{ + unsigned char *csr = reserve(p, 2); + *csr++ = b0; + *csr++ = b1; +} + +static void emit_3ub( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2 ) +{ + unsigned char *csr = reserve(p, 3); + *csr++ = b0; + *csr++ = b1; + *csr++ = b2; +} + + +/* Build a modRM byte + possible displacement. No treatment of SIB + * indexing. BZZT - no way to encode an absolute address. + */ +static void emit_modrm( struct x86_function *p, + struct x86_reg reg, + struct x86_reg regmem ) +{ + unsigned char val = 0; + + assert(reg.mod == mod_REG); + + val |= regmem.mod << 6; /* mod field */ + val |= reg.idx << 3; /* reg field */ + val |= regmem.idx; /* r/m field */ + + emit_1ub(p, val); + + /* Oh-oh we've stumbled into the SIB thing. + */ + if (regmem.file == file_REG32 && + regmem.idx == reg_SP) { + emit_1ub(p, 0x24); /* simplistic! */ + } + + switch (regmem.mod) { + case mod_REG: + case mod_INDIRECT: + break; + case mod_DISP8: + emit_1b(p, regmem.disp); + break; + case mod_DISP32: + emit_1i(p, regmem.disp); + break; + default: + assert(0); + break; + } +} + + +static void emit_modrm_noreg( struct x86_function *p, + unsigned op, + struct x86_reg regmem ) +{ + struct x86_reg dummy = x86_make_reg(file_REG32, op); + emit_modrm(p, dummy, regmem); +} + +/* Many x86 instructions have two opcodes to cope with the situations + * where the destination is a register or memory reference + * respectively. This function selects the correct opcode based on + * the arguments presented. + */ +static void emit_op_modrm( struct x86_function *p, + unsigned char op_dst_is_reg, + unsigned char op_dst_is_mem, + struct x86_reg dst, + struct x86_reg src ) +{ + switch (dst.mod) { + case mod_REG: + emit_1ub(p, op_dst_is_reg); + emit_modrm(p, dst, src); + break; + case mod_INDIRECT: + case mod_DISP32: + case mod_DISP8: + assert(src.mod == mod_REG); + emit_1ub(p, op_dst_is_mem); + emit_modrm(p, src, dst); + break; + default: + assert(0); + break; + } +} + + + + + + + +/* Create and manipulate registers and regmem values: + */ +struct x86_reg x86_make_reg( enum x86_reg_file file, + enum x86_reg_name idx ) +{ + struct x86_reg reg; + + reg.file = file; + reg.idx = idx; + reg.mod = mod_REG; + reg.disp = 0; + + return reg; +} + +struct x86_reg x86_make_disp( struct x86_reg reg, + int disp ) +{ + assert(reg.file == file_REG32); + + if (reg.mod == mod_REG) + reg.disp = disp; + else + reg.disp += disp; + + if (reg.disp == 0) + reg.mod = mod_INDIRECT; + else if (reg.disp <= 127 && reg.disp >= -128) + reg.mod = mod_DISP8; + else + reg.mod = mod_DISP32; + + return reg; +} + +struct x86_reg x86_deref( struct x86_reg reg ) +{ + return x86_make_disp(reg, 0); +} + +struct x86_reg x86_get_base_reg( struct x86_reg reg ) +{ + return x86_make_reg( reg.file, reg.idx ); +} + +unsigned char *x86_get_label( struct x86_function *p ) +{ + return p->csr; +} + + + +/*********************************************************************** + * x86 instructions + */ + + +void x86_jcc( struct x86_function *p, + enum x86_cc cc, + unsigned char *label ) +{ + int offset = label - (x86_get_label(p) + 2); + + if (offset <= 127 && offset >= -128) { + emit_1ub(p, 0x70 + cc); + emit_1b(p, (char) offset); + } + else { + offset = label - (x86_get_label(p) + 6); + emit_2ub(p, 0x0f, 0x80 + cc); + emit_1i(p, offset); + } +} + +/* Always use a 32bit offset for forward jumps: + */ +unsigned char *x86_jcc_forward( struct x86_function *p, + enum x86_cc cc ) +{ + emit_2ub(p, 0x0f, 0x80 + cc); + emit_1i(p, 0); + return x86_get_label(p); +} + +unsigned char *x86_jmp_forward( struct x86_function *p) +{ + emit_1ub(p, 0xe9); + emit_1i(p, 0); + return x86_get_label(p); +} + +unsigned char *x86_call_forward( struct x86_function *p) +{ + emit_1ub(p, 0xe8); + emit_1i(p, 0); + return x86_get_label(p); +} + +/* Fixup offset from forward jump: + */ +void x86_fixup_fwd_jump( struct x86_function *p, + unsigned char *fixup ) +{ + *(int *)(fixup - 4) = x86_get_label(p) - fixup; +} + +void x86_jmp( struct x86_function *p, unsigned char *label) +{ + emit_1ub(p, 0xe9); + emit_1i(p, label - x86_get_label(p) - 4); +} + +#if 0 +/* This doesn't work once we start reallocating & copying the + * generated code on buffer fills, because the call is relative to the + * current pc. + */ +void x86_call( struct x86_function *p, void (*label)()) +{ + emit_1ub(p, 0xe8); + emit_1i(p, cptr(label) - x86_get_label(p) - 4); +} +#else +void x86_call( struct x86_function *p, struct x86_reg reg) +{ + emit_1ub(p, 0xff); + emit_modrm(p, reg, reg); +} +#endif + + +/* michal: + * Temporary. As I need immediate operands, and dont want to mess with the codegen, + * I load the immediate into general purpose register and use it. + */ +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) +{ + assert(dst.mod == mod_REG); + emit_1ub(p, 0xb8 + dst.idx); + emit_1i(p, imm); +} + +void x86_push( struct x86_function *p, + struct x86_reg reg ) +{ + assert(reg.mod == mod_REG); + emit_1ub(p, 0x50 + reg.idx); + p->stack_offset += 4; +} + +void x86_pop( struct x86_function *p, + struct x86_reg reg ) +{ + assert(reg.mod == mod_REG); + emit_1ub(p, 0x58 + reg.idx); + p->stack_offset -= 4; +} + +void x86_inc( struct x86_function *p, + struct x86_reg reg ) +{ + assert(reg.mod == mod_REG); + emit_1ub(p, 0x40 + reg.idx); +} + +void x86_dec( struct x86_function *p, + struct x86_reg reg ) +{ + assert(reg.mod == mod_REG); + emit_1ub(p, 0x48 + reg.idx); +} + +void x86_ret( struct x86_function *p ) +{ + emit_1ub(p, 0xc3); +} + +void x86_sahf( struct x86_function *p ) +{ + emit_1ub(p, 0x9e); +} + +void x86_mov( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x8b, 0x89, dst, src ); +} + +void x86_xor( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x33, 0x31, dst, src ); +} + +void x86_cmp( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x3b, 0x39, dst, src ); +} + +void x86_lea( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_1ub(p, 0x8d); + emit_modrm( p, dst, src ); +} + +void x86_test( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_1ub(p, 0x85); + emit_modrm( p, dst, src ); +} + +void x86_add( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm(p, 0x03, 0x01, dst, src ); +} + +void x86_mul( struct x86_function *p, + struct x86_reg src ) +{ + assert (src.file == file_REG32 && src.mod == mod_REG); + emit_op_modrm(p, 0xf7, 0, x86_make_reg (file_REG32, reg_SP), src ); +} + +void x86_sub( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm(p, 0x2b, 0x29, dst, src ); +} + +void x86_or( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x0b, 0x09, dst, src ); +} + +void x86_and( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x23, 0x21, dst, src ); +} + + + +/*********************************************************************** + * SSE instructions + */ + + +void sse_movss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, 0xF3, X86_TWOB); + emit_op_modrm( p, 0x10, 0x11, dst, src ); +} + +void sse_movaps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x28, 0x29, dst, src ); +} + +void sse_movups( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x10, 0x11, dst, src ); +} + +void sse_movhps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.mod != mod_REG || src.mod != mod_REG); + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x16, 0x17, dst, src ); /* cf movlhps */ +} + +void sse_movlps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.mod != mod_REG || src.mod != mod_REG); + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x12, 0x13, dst, src ); /* cf movhlps */ +} + +void sse_maxps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x5F); + emit_modrm( p, dst, src ); +} + +void sse_maxss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x5F); + emit_modrm( p, dst, src ); +} + +void sse_divss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x5E); + emit_modrm( p, dst, src ); +} + +void sse_minps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x5D); + emit_modrm( p, dst, src ); +} + +void sse_subps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x5C); + emit_modrm( p, dst, src ); +} + +void sse_mulps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x59); + emit_modrm( p, dst, src ); +} + +void sse_mulss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x59); + emit_modrm( p, dst, src ); +} + +void sse_addps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x58); + emit_modrm( p, dst, src ); +} + +void sse_addss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x58); + emit_modrm( p, dst, src ); +} + +void sse_andnps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x55); + emit_modrm( p, dst, src ); +} + +void sse_andps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x54); + emit_modrm( p, dst, src ); +} + +void sse_rsqrtps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x52); + emit_modrm( p, dst, src ); +} + +void sse_rsqrtss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x52); + emit_modrm( p, dst, src ); + +} + +void sse_movhlps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.mod == mod_REG && src.mod == mod_REG); + emit_2ub(p, X86_TWOB, 0x12); + emit_modrm( p, dst, src ); +} + +void sse_movlhps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.mod == mod_REG && src.mod == mod_REG); + emit_2ub(p, X86_TWOB, 0x16); + emit_modrm( p, dst, src ); +} + +void sse_orps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x56); + emit_modrm( p, dst, src ); +} + +void sse_xorps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x57); + emit_modrm( p, dst, src ); +} + +void sse_cvtps2pi( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.file == file_MMX && + (src.file == file_XMM || src.mod != mod_REG)); + + p->need_emms = 1; + + emit_2ub(p, X86_TWOB, 0x2d); + emit_modrm( p, dst, src ); +} + + +/* Shufps can also be used to implement a reduced swizzle when dest == + * arg0. + */ +void sse_shufps( struct x86_function *p, + struct x86_reg dest, + struct x86_reg arg0, + unsigned char shuf) +{ + emit_2ub(p, X86_TWOB, 0xC6); + emit_modrm(p, dest, arg0); + emit_1ub(p, shuf); +} + +void sse_cmpps( struct x86_function *p, + struct x86_reg dest, + struct x86_reg arg0, + unsigned char cc) +{ + emit_2ub(p, X86_TWOB, 0xC2); + emit_modrm(p, dest, arg0); + emit_1ub(p, cc); +} + +void sse_pmovmskb( struct x86_function *p, + struct x86_reg dest, + struct x86_reg src) +{ + emit_3ub(p, 0x66, X86_TWOB, 0xD7); + emit_modrm(p, dest, src); +} + +/*********************************************************************** + * SSE2 instructions + */ + +/** + * Perform a reduced swizzle: + */ +void sse2_pshufd( struct x86_function *p, + struct x86_reg dest, + struct x86_reg arg0, + unsigned char shuf) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x70); + emit_modrm(p, dest, arg0); + emit_1ub(p, shuf); +} + +void sse2_cvttps2dq( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub( p, 0xF3, X86_TWOB, 0x5B ); + emit_modrm( p, dst, src ); +} + +void sse2_cvtps2dq( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x5B); + emit_modrm( p, dst, src ); +} + +void sse2_packssdw( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x6B); + emit_modrm( p, dst, src ); +} + +void sse2_packsswb( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x63); + emit_modrm( p, dst, src ); +} + +void sse2_packuswb( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x67); + emit_modrm( p, dst, src ); +} + +void sse2_rcpps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x53); + emit_modrm( p, dst, src ); +} + +void sse2_rcpss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x53); + emit_modrm( p, dst, src ); +} + +void sse2_movd( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, 0x66, X86_TWOB); + emit_op_modrm( p, 0x6e, 0x7e, dst, src ); +} + + + + +/*********************************************************************** + * x87 instructions + */ +void x87_fist( struct x86_function *p, struct x86_reg dst ) +{ + emit_1ub(p, 0xdb); + emit_modrm_noreg(p, 2, dst); +} + +void x87_fistp( struct x86_function *p, struct x86_reg dst ) +{ + emit_1ub(p, 0xdb); + emit_modrm_noreg(p, 3, dst); +} + +void x87_fild( struct x86_function *p, struct x86_reg arg ) +{ + emit_1ub(p, 0xdf); + emit_modrm_noreg(p, 0, arg); +} + +void x87_fldz( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xee); +} + + +void x87_fldcw( struct x86_function *p, struct x86_reg arg ) +{ + assert(arg.file == file_REG32); + assert(arg.mod != mod_REG); + emit_1ub(p, 0xd9); + emit_modrm_noreg(p, 5, arg); +} + +void x87_fld1( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xe8); +} + +void x87_fldl2e( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xea); +} + +void x87_fldln2( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xed); +} + +void x87_fwait( struct x86_function *p ) +{ + emit_1ub(p, 0x9b); +} + +void x87_fnclex( struct x86_function *p ) +{ + emit_2ub(p, 0xdb, 0xe2); +} + +void x87_fclex( struct x86_function *p ) +{ + x87_fwait(p); + x87_fnclex(p); +} + + +static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg, + unsigned char dst0ub0, + unsigned char dst0ub1, + unsigned char arg0ub0, + unsigned char arg0ub1, + unsigned char argmem_noreg) +{ + assert(dst.file == file_x87); + + if (arg.file == file_x87) { + if (dst.idx == 0) + emit_2ub(p, dst0ub0, dst0ub1+arg.idx); + else if (arg.idx == 0) + emit_2ub(p, arg0ub0, arg0ub1+arg.idx); + else + assert(0); + } + else if (dst.idx == 0) { + assert(arg.file == file_REG32); + emit_1ub(p, 0xd8); + emit_modrm_noreg(p, argmem_noreg, arg); + } + else + assert(0); +} + +void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xc8, + 0xdc, 0xc8, + 4); +} + +void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xe0, + 0xdc, 0xe8, + 4); +} + +void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xe8, + 0xdc, 0xe0, + 5); +} + +void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xc0, + 0xdc, 0xc0, + 0); +} + +void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xf0, + 0xdc, 0xf8, + 6); +} + +void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xf8, + 0xdc, 0xf0, + 7); +} + +void x87_fmulp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xc8+dst.idx); +} + +void x87_fsubp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xe8+dst.idx); +} + +void x87_fsubrp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xe0+dst.idx); +} + +void x87_faddp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xc0+dst.idx); +} + +void x87_fdivp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xf8+dst.idx); +} + +void x87_fdivrp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xf0+dst.idx); +} + +void x87_fucom( struct x86_function *p, struct x86_reg arg ) +{ + assert(arg.file == file_x87); + emit_2ub(p, 0xdd, 0xe0+arg.idx); +} + +void x87_fucomp( struct x86_function *p, struct x86_reg arg ) +{ + assert(arg.file == file_x87); + emit_2ub(p, 0xdd, 0xe8+arg.idx); +} + +void x87_fucompp( struct x86_function *p ) +{ + emit_2ub(p, 0xda, 0xe9); +} + +void x87_fxch( struct x86_function *p, struct x86_reg arg ) +{ + assert(arg.file == file_x87); + emit_2ub(p, 0xd9, 0xc8+arg.idx); +} + +void x87_fabs( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xe1); +} + +void x87_fchs( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xe0); +} + +void x87_fcos( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xff); +} + + +void x87_fprndint( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfc); +} + +void x87_fscale( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfd); +} + +void x87_fsin( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfe); +} + +void x87_fsincos( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfb); +} + +void x87_fsqrt( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfa); +} + +void x87_fxtract( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xf4); +} + +/* st0 = (2^st0)-1 + * + * Restrictions: -1.0 <= st0 <= 1.0 + */ +void x87_f2xm1( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xf0); +} + +/* st1 = st1 * log2(st0); + * pop_stack; + */ +void x87_fyl2x( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xf1); +} + +/* st1 = st1 * log2(st0 + 1.0); + * pop_stack; + * + * A fast operation, with restrictions: -.29 < st0 < .29 + */ +void x87_fyl2xp1( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xf9); +} + + +void x87_fld( struct x86_function *p, struct x86_reg arg ) +{ + if (arg.file == file_x87) + emit_2ub(p, 0xd9, 0xc0 + arg.idx); + else { + emit_1ub(p, 0xd9); + emit_modrm_noreg(p, 0, arg); + } +} + +void x87_fst( struct x86_function *p, struct x86_reg dst ) +{ + if (dst.file == file_x87) + emit_2ub(p, 0xdd, 0xd0 + dst.idx); + else { + emit_1ub(p, 0xd9); + emit_modrm_noreg(p, 2, dst); + } +} + +void x87_fstp( struct x86_function *p, struct x86_reg dst ) +{ + if (dst.file == file_x87) + emit_2ub(p, 0xdd, 0xd8 + dst.idx); + else { + emit_1ub(p, 0xd9); + emit_modrm_noreg(p, 3, dst); + } +} + +void x87_fcom( struct x86_function *p, struct x86_reg dst ) +{ + if (dst.file == file_x87) + emit_2ub(p, 0xd8, 0xd0 + dst.idx); + else { + emit_1ub(p, 0xd8); + emit_modrm_noreg(p, 2, dst); + } +} + +void x87_fcomp( struct x86_function *p, struct x86_reg dst ) +{ + if (dst.file == file_x87) + emit_2ub(p, 0xd8, 0xd8 + dst.idx); + else { + emit_1ub(p, 0xd8); + emit_modrm_noreg(p, 3, dst); + } +} + + +void x87_fnstsw( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_REG32); + + if (dst.idx == reg_AX && + dst.mod == mod_REG) + emit_2ub(p, 0xdf, 0xe0); + else { + emit_1ub(p, 0xdd); + emit_modrm_noreg(p, 7, dst); + } +} + + + + +/*********************************************************************** + * MMX instructions + */ + +void mmx_emms( struct x86_function *p ) +{ + assert(p->need_emms); + emit_2ub(p, 0x0f, 0x77); + p->need_emms = 0; +} + +void mmx_packssdw( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.file == file_MMX && + (src.file == file_MMX || src.mod != mod_REG)); + + p->need_emms = 1; + + emit_2ub(p, X86_TWOB, 0x6b); + emit_modrm( p, dst, src ); +} + +void mmx_packuswb( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.file == file_MMX && + (src.file == file_MMX || src.mod != mod_REG)); + + p->need_emms = 1; + + emit_2ub(p, X86_TWOB, 0x67); + emit_modrm( p, dst, src ); +} + +void mmx_movd( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + p->need_emms = 1; + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x6e, 0x7e, dst, src ); +} + +void mmx_movq( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + p->need_emms = 1; + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x6f, 0x7f, dst, src ); +} + + +/*********************************************************************** + * Helper functions + */ + + +/* Retreive a reference to one of the function arguments, taking into + * account any push/pop activity: + */ +struct x86_reg x86_fn_arg( struct x86_function *p, + unsigned arg ) +{ + return x86_make_disp(x86_make_reg(file_REG32, reg_SP), + p->stack_offset + arg * 4); /* ??? */ +} + + +void x86_init_func( struct x86_function *p ) +{ + p->size = 0; + p->store = NULL; + p->csr = p->store; +} + +void x86_init_func_size( struct x86_function *p, unsigned code_size ) +{ + p->size = code_size; + p->store = _mesa_exec_malloc(code_size); + p->csr = p->store; +} + +void x86_release_func( struct x86_function *p ) +{ + _mesa_exec_free(p->store); + p->store = NULL; + p->csr = NULL; + p->size = 0; +} + + +void (*x86_get_func( struct x86_function *p ))(void) +{ + if (DISASSEM && p->store) + _mesa_printf("disassemble %p %p\n", p->store, p->csr); + return (void (*)(void)) (unsigned long) p->store; +} + +#else + +void x86sse_dummy( void ) +{ +} + +#endif diff --git a/src/gallium/auxiliary/rtasm/x86sse.h b/src/gallium/auxiliary/rtasm/x86sse.h new file mode 100644 index 0000000000..c2aa416492 --- /dev/null +++ b/src/gallium/auxiliary/rtasm/x86sse.h @@ -0,0 +1,256 @@ + +#ifndef _X86SSE_H_ +#define _X86SSE_H_ + +#if defined(__i386__) || defined(__386__) + +/* It is up to the caller to ensure that instructions issued are + * suitable for the host cpu. There are no checks made in this module + * for mmx/sse/sse2 support on the cpu. + */ +struct x86_reg { + unsigned file:3; + unsigned idx:3; + unsigned mod:2; /* mod_REG if this is just a register */ + int disp:24; /* only +/- 23bits of offset - should be enough... */ +}; + +struct x86_function { + unsigned size; + unsigned char *store; + unsigned char *csr; + unsigned stack_offset; + int need_emms; + const char *fn; +}; + +enum x86_reg_file { + file_REG32, + file_MMX, + file_XMM, + file_x87 +}; + +/* Values for mod field of modr/m byte + */ +enum x86_reg_mod { + mod_INDIRECT, + mod_DISP8, + mod_DISP32, + mod_REG +}; + +enum x86_reg_name { + reg_AX, + reg_CX, + reg_DX, + reg_BX, + reg_SP, + reg_BP, + reg_SI, + reg_DI +}; + + +enum x86_cc { + cc_O, /* overflow */ + cc_NO, /* not overflow */ + cc_NAE, /* not above or equal / carry */ + cc_AE, /* above or equal / not carry */ + cc_E, /* equal / zero */ + cc_NE /* not equal / not zero */ +}; + +enum sse_cc { + cc_Equal, + cc_LessThan, + cc_LessThanEqual, + cc_Unordered, + cc_NotEqual, + cc_NotLessThan, + cc_NotLessThanEqual, + cc_Ordered +}; + +#define cc_Z cc_E +#define cc_NZ cc_NE + +/* Begin/end/retreive function creation: + */ + + +void x86_init_func( struct x86_function *p ); +void x86_init_func_size( struct x86_function *p, unsigned code_size ); +void x86_release_func( struct x86_function *p ); +void (*x86_get_func( struct x86_function *p ))( void ); + + + +/* Create and manipulate registers and regmem values: + */ +struct x86_reg x86_make_reg( enum x86_reg_file file, + enum x86_reg_name idx ); + +struct x86_reg x86_make_disp( struct x86_reg reg, + int disp ); + +struct x86_reg x86_deref( struct x86_reg reg ); + +struct x86_reg x86_get_base_reg( struct x86_reg reg ); + + +/* Labels, jumps and fixup: + */ +unsigned char *x86_get_label( struct x86_function *p ); + +void x86_jcc( struct x86_function *p, + enum x86_cc cc, + unsigned char *label ); + +unsigned char *x86_jcc_forward( struct x86_function *p, + enum x86_cc cc ); + +unsigned char *x86_jmp_forward( struct x86_function *p); + +unsigned char *x86_call_forward( struct x86_function *p); + +void x86_fixup_fwd_jump( struct x86_function *p, + unsigned char *fixup ); + +void x86_jmp( struct x86_function *p, unsigned char *label ); + +/* void x86_call( struct x86_function *p, void (*label)() ); */ +void x86_call( struct x86_function *p, struct x86_reg reg); + +/* michal: + * Temporary. As I need immediate operands, and dont want to mess with the codegen, + * I load the immediate into general purpose register and use it. + */ +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ); + + +/* Macro for sse_shufps() and sse2_pshufd(): + */ +#define SHUF(_x,_y,_z,_w) (((_x)<<0) | ((_y)<<2) | ((_z)<<4) | ((_w)<<6)) +#define SHUF_NOOP RSW(0,1,2,3) +#define GET_SHUF(swz, idx) (((swz) >> ((idx)*2)) & 0x3) + +void mmx_emms( struct x86_function *p ); +void mmx_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void mmx_movq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void mmx_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void mmx_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); + +void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_cvttps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, + unsigned char shuf ); +void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); + +void sse_addps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_addss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, + unsigned char cc ); +void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movaps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movhlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movlhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movups( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_mulps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_mulss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_orps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, + unsigned char shuf ); +void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); + +void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_cmp( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_dec( struct x86_function *p, struct x86_reg reg ); +void x86_inc( struct x86_function *p, struct x86_reg reg ); +void x86_lea( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_mov( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_mul( struct x86_function *p, struct x86_reg src ); +void x86_or( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_pop( struct x86_function *p, struct x86_reg reg ); +void x86_push( struct x86_function *p, struct x86_reg reg ); +void x86_ret( struct x86_function *p ); +void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_sahf( struct x86_function *p ); + +void x87_f2xm1( struct x86_function *p ); +void x87_fabs( struct x86_function *p ); +void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_faddp( struct x86_function *p, struct x86_reg dst ); +void x87_fchs( struct x86_function *p ); +void x87_fclex( struct x86_function *p ); +void x87_fcom( struct x86_function *p, struct x86_reg dst ); +void x87_fcomp( struct x86_function *p, struct x86_reg dst ); +void x87_fcos( struct x86_function *p ); +void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fdivp( struct x86_function *p, struct x86_reg dst ); +void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fdivrp( struct x86_function *p, struct x86_reg dst ); +void x87_fild( struct x86_function *p, struct x86_reg arg ); +void x87_fist( struct x86_function *p, struct x86_reg dst ); +void x87_fistp( struct x86_function *p, struct x86_reg dst ); +void x87_fld( struct x86_function *p, struct x86_reg arg ); +void x87_fld1( struct x86_function *p ); +void x87_fldcw( struct x86_function *p, struct x86_reg arg ); +void x87_fldl2e( struct x86_function *p ); +void x87_fldln2( struct x86_function *p ); +void x87_fldz( struct x86_function *p ); +void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fmulp( struct x86_function *p, struct x86_reg dst ); +void x87_fnclex( struct x86_function *p ); +void x87_fprndint( struct x86_function *p ); +void x87_fscale( struct x86_function *p ); +void x87_fsin( struct x86_function *p ); +void x87_fsincos( struct x86_function *p ); +void x87_fsqrt( struct x86_function *p ); +void x87_fst( struct x86_function *p, struct x86_reg dst ); +void x87_fstp( struct x86_function *p, struct x86_reg dst ); +void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fsubp( struct x86_function *p, struct x86_reg dst ); +void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fsubrp( struct x86_function *p, struct x86_reg dst ); +void x87_fxch( struct x86_function *p, struct x86_reg dst ); +void x87_fxtract( struct x86_function *p ); +void x87_fyl2x( struct x86_function *p ); +void x87_fyl2xp1( struct x86_function *p ); +void x87_fwait( struct x86_function *p ); +void x87_fnstsw( struct x86_function *p, struct x86_reg dst ); +void x87_fucompp( struct x86_function *p ); +void x87_fucomp( struct x86_function *p, struct x86_reg arg ); +void x87_fucom( struct x86_function *p, struct x86_reg arg ); + + + +/* Retreive a reference to one of the function arguments, taking into + * account any push/pop activity. Note - doesn't track explict + * manipulation of ESP by other instructions. + */ +struct x86_reg x86_fn_arg( struct x86_function *p, unsigned arg ); + +#endif +#endif -- cgit v1.2.3 From 39ea0308425ad04618061129c63c22ac0efb0692 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 12:00:48 +0900 Subject: Rename rtasm files. --- src/gallium/auxiliary/rtasm/Makefile | 4 +- src/gallium/auxiliary/rtasm/SConscript | 4 +- src/gallium/auxiliary/rtasm/execmem.c | 133 --- src/gallium/auxiliary/rtasm/execmem.h | 45 - src/gallium/auxiliary/rtasm/rtasm_execmem.c | 134 +++ src/gallium/auxiliary/rtasm/rtasm_execmem.h | 45 + src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 1196 +++++++++++++++++++++++++++ src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 256 ++++++ src/gallium/auxiliary/rtasm/x86sse.c | 1195 -------------------------- src/gallium/auxiliary/rtasm/x86sse.h | 256 ------ 10 files changed, 1635 insertions(+), 1633 deletions(-) delete mode 100644 src/gallium/auxiliary/rtasm/execmem.c delete mode 100644 src/gallium/auxiliary/rtasm/execmem.h create mode 100644 src/gallium/auxiliary/rtasm/rtasm_execmem.c create mode 100644 src/gallium/auxiliary/rtasm/rtasm_execmem.h create mode 100644 src/gallium/auxiliary/rtasm/rtasm_x86sse.c create mode 100644 src/gallium/auxiliary/rtasm/rtasm_x86sse.h delete mode 100644 src/gallium/auxiliary/rtasm/x86sse.c delete mode 100644 src/gallium/auxiliary/rtasm/x86sse.h (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/Makefile b/src/gallium/auxiliary/rtasm/Makefile index b3b9934e10..7c8ac60794 100644 --- a/src/gallium/auxiliary/rtasm/Makefile +++ b/src/gallium/auxiliary/rtasm/Makefile @@ -5,9 +5,9 @@ include $(TOP)/configs/current LIBNAME = rtasm DRIVER_SOURCES = \ + execmem.c \ x86sse.c \ - mm.c \ - execmem.c + mm.c C_SOURCES = \ $(DRIVER_SOURCES) diff --git a/src/gallium/auxiliary/rtasm/SConscript b/src/gallium/auxiliary/rtasm/SConscript index c5b1551786..de8456e0ca 100644 --- a/src/gallium/auxiliary/rtasm/SConscript +++ b/src/gallium/auxiliary/rtasm/SConscript @@ -3,9 +3,9 @@ Import('*') rtasm = env.ConvenienceLibrary( target = 'rtasm', source = [ - 'x86sse.c', + 'rtasm_execmem.c', + 'rtasm_x86sse.c', 'mm.c', - 'execmem.c', ]) auxiliaries.insert(0, rtasm) diff --git a/src/gallium/auxiliary/rtasm/execmem.c b/src/gallium/auxiliary/rtasm/execmem.c deleted file mode 100644 index c7c35f7ef2..0000000000 --- a/src/gallium/auxiliary/rtasm/execmem.c +++ /dev/null @@ -1,133 +0,0 @@ -/************************************************************************** - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * \file exemem.c - * Functions for allocating executable memory. - * - * \author Keith Whitwell - */ - - -#include "pipe/p_compiler.h" -#include "pipe/p_thread.h" - -#include "execmem.h" - - -#if defined(__linux__) - -/* - * Allocate a large block of memory which can hold code then dole it out - * in pieces by means of the generic memory manager code. -*/ - -#include -#include -#include "mm.h" - -#define EXEC_HEAP_SIZE (10*1024*1024) - -_glthread_DECLARE_STATIC_MUTEX(exec_mutex); - -static struct mem_block *exec_heap = NULL; -static unsigned char *exec_mem = NULL; - - -static void -init_heap(void) -{ - if (!exec_heap) - exec_heap = mmInit( 0, EXEC_HEAP_SIZE ); - - if (!exec_mem) - exec_mem = (unsigned char *) mmap(0, EXEC_HEAP_SIZE, - PROT_EXEC | PROT_READ | PROT_WRITE, - MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); -} - - -void * -_mesa_exec_malloc(size_t size) -{ - struct mem_block *block = NULL; - void *addr = NULL; - - _glthread_LOCK_MUTEX(exec_mutex); - - init_heap(); - - if (exec_heap) { - size = (size + 31) & ~31; - block = mmAllocMem( exec_heap, size, 32, 0 ); - } - - if (block) - addr = exec_mem + block->ofs; - else - debug_printf("_mesa_exec_malloc failed\n"); - - _glthread_UNLOCK_MUTEX(exec_mutex); - - return addr; -} - - -void -_mesa_exec_free(void *addr) -{ - _glthread_LOCK_MUTEX(exec_mutex); - - if (exec_heap) { - struct mem_block *block = mmFindBlock(exec_heap, (unsigned char *)addr - exec_mem); - - if (block) - mmFreeMem(block); - } - - _glthread_UNLOCK_MUTEX(exec_mutex); -} - - -#else - -/* - * Just use regular memory. - */ - -void * -_mesa_exec_malloc(GLuint size) -{ - return _mesa_malloc( size ); -} - - -void -_mesa_exec_free(void *addr) -{ - _mesa_free(addr); -} - - -#endif diff --git a/src/gallium/auxiliary/rtasm/execmem.h b/src/gallium/auxiliary/rtasm/execmem.h deleted file mode 100644 index 9fd4569165..0000000000 --- a/src/gallium/auxiliary/rtasm/execmem.h +++ /dev/null @@ -1,45 +0,0 @@ -/************************************************************************** - * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file exemem.c - * Functions for allocating executable memory. - * - * \author Keith Whitwell - */ - -#ifndef _EXECMEM_H_ -#define _EXECMEM_H_ - -#include "pipe/p_compiler.h" - - -extern void * -_mesa_exec_malloc( size_t size ); - - -extern void -_mesa_exec_free( void *addr ); - - -#endif diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.c b/src/gallium/auxiliary/rtasm/rtasm_execmem.c new file mode 100644 index 0000000000..cb13db2498 --- /dev/null +++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.c @@ -0,0 +1,134 @@ +/************************************************************************** + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * \file exemem.c + * Functions for allocating executable memory. + * + * \author Keith Whitwell + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_thread.h" + +#include "rtasm_execmem.h" + + +#if defined(__linux__) + +/* + * Allocate a large block of memory which can hold code then dole it out + * in pieces by means of the generic memory manager code. +*/ + +#include +#include +#include "mm.h" + +#define EXEC_HEAP_SIZE (10*1024*1024) + +_glthread_DECLARE_STATIC_MUTEX(exec_mutex); + +static struct mem_block *exec_heap = NULL; +static unsigned char *exec_mem = NULL; + + +static void +init_heap(void) +{ + if (!exec_heap) + exec_heap = mmInit( 0, EXEC_HEAP_SIZE ); + + if (!exec_mem) + exec_mem = (unsigned char *) mmap(0, EXEC_HEAP_SIZE, + PROT_EXEC | PROT_READ | PROT_WRITE, + MAP_PRIVATE | MAP_ANONYMOUS, -1, 0); +} + + +void * +rtasm_exec_malloc(size_t size) +{ + struct mem_block *block = NULL; + void *addr = NULL; + + _glthread_LOCK_MUTEX(exec_mutex); + + init_heap(); + + if (exec_heap) { + size = (size + 31) & ~31; + block = mmAllocMem( exec_heap, size, 32, 0 ); + } + + if (block) + addr = exec_mem + block->ofs; + else + debug_printf("rtasm_exec_malloc failed\n"); + + _glthread_UNLOCK_MUTEX(exec_mutex); + + return addr; +} + + +void +rtasm_exec_free(void *addr) +{ + _glthread_LOCK_MUTEX(exec_mutex); + + if (exec_heap) { + struct mem_block *block = mmFindBlock(exec_heap, (unsigned char *)addr - exec_mem); + + if (block) + mmFreeMem(block); + } + + _glthread_UNLOCK_MUTEX(exec_mutex); +} + + +#else + +/* + * Just use regular memory. + */ + +void * +rtasm_exec_malloc(GLuint size) +{ + return MALLOC( size ); +} + + +void +rtasm_exec_free(void *addr) +{ + FREE(addr); +} + + +#endif diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.h b/src/gallium/auxiliary/rtasm/rtasm_execmem.h new file mode 100644 index 0000000000..155c6d34e0 --- /dev/null +++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.h @@ -0,0 +1,45 @@ +/************************************************************************** + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file exemem.c + * Functions for allocating executable memory. + * + * \author Keith Whitwell + */ + +#ifndef _RTASM_EXECMEM_H_ +#define _RTASM_EXECMEM_H_ + +#include "pipe/p_compiler.h" + + +extern void * +rtasm_exec_malloc( size_t size ); + + +extern void +rtasm_exec_free( void *addr ); + + +#endif diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c new file mode 100644 index 0000000000..3c885a9fff --- /dev/null +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -0,0 +1,1196 @@ +#if defined(__i386__) || defined(__386__) + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" + +#include "rtasm_execmem.h" +#include "rtasm_x86sse.h" + +#define DISASSEM 0 +#define X86_TWOB 0x0f + +static unsigned char *cptr( void (*label)() ) +{ + return (unsigned char *)(unsigned long)label; +} + + +static void do_realloc( struct x86_function *p ) +{ + if (p->size == 0) { + p->size = 1024; + p->store = rtasm_exec_malloc(p->size); + p->csr = p->store; + } + else { + unsigned used = p->csr - p->store; + unsigned char *tmp = p->store; + p->size *= 2; + p->store = rtasm_exec_malloc(p->size); + memcpy(p->store, tmp, used); + p->csr = p->store + used; + rtasm_exec_free(tmp); + } +} + +/* Emit bytes to the instruction stream: + */ +static unsigned char *reserve( struct x86_function *p, int bytes ) +{ + if (p->csr + bytes - p->store > p->size) + do_realloc(p); + + { + unsigned char *csr = p->csr; + p->csr += bytes; + return csr; + } +} + + + +static void emit_1b( struct x86_function *p, char b0 ) +{ + char *csr = (char *)reserve(p, 1); + *csr = b0; +} + +static void emit_1i( struct x86_function *p, int i0 ) +{ + int *icsr = (int *)reserve(p, sizeof(i0)); + *icsr = i0; +} + +static void emit_1ub( struct x86_function *p, unsigned char b0 ) +{ + unsigned char *csr = reserve(p, 1); + *csr++ = b0; +} + +static void emit_2ub( struct x86_function *p, unsigned char b0, unsigned char b1 ) +{ + unsigned char *csr = reserve(p, 2); + *csr++ = b0; + *csr++ = b1; +} + +static void emit_3ub( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2 ) +{ + unsigned char *csr = reserve(p, 3); + *csr++ = b0; + *csr++ = b1; + *csr++ = b2; +} + + +/* Build a modRM byte + possible displacement. No treatment of SIB + * indexing. BZZT - no way to encode an absolute address. + */ +static void emit_modrm( struct x86_function *p, + struct x86_reg reg, + struct x86_reg regmem ) +{ + unsigned char val = 0; + + assert(reg.mod == mod_REG); + + val |= regmem.mod << 6; /* mod field */ + val |= reg.idx << 3; /* reg field */ + val |= regmem.idx; /* r/m field */ + + emit_1ub(p, val); + + /* Oh-oh we've stumbled into the SIB thing. + */ + if (regmem.file == file_REG32 && + regmem.idx == reg_SP) { + emit_1ub(p, 0x24); /* simplistic! */ + } + + switch (regmem.mod) { + case mod_REG: + case mod_INDIRECT: + break; + case mod_DISP8: + emit_1b(p, regmem.disp); + break; + case mod_DISP32: + emit_1i(p, regmem.disp); + break; + default: + assert(0); + break; + } +} + + +static void emit_modrm_noreg( struct x86_function *p, + unsigned op, + struct x86_reg regmem ) +{ + struct x86_reg dummy = x86_make_reg(file_REG32, op); + emit_modrm(p, dummy, regmem); +} + +/* Many x86 instructions have two opcodes to cope with the situations + * where the destination is a register or memory reference + * respectively. This function selects the correct opcode based on + * the arguments presented. + */ +static void emit_op_modrm( struct x86_function *p, + unsigned char op_dst_is_reg, + unsigned char op_dst_is_mem, + struct x86_reg dst, + struct x86_reg src ) +{ + switch (dst.mod) { + case mod_REG: + emit_1ub(p, op_dst_is_reg); + emit_modrm(p, dst, src); + break; + case mod_INDIRECT: + case mod_DISP32: + case mod_DISP8: + assert(src.mod == mod_REG); + emit_1ub(p, op_dst_is_mem); + emit_modrm(p, src, dst); + break; + default: + assert(0); + break; + } +} + + + + + + + +/* Create and manipulate registers and regmem values: + */ +struct x86_reg x86_make_reg( enum x86_reg_file file, + enum x86_reg_name idx ) +{ + struct x86_reg reg; + + reg.file = file; + reg.idx = idx; + reg.mod = mod_REG; + reg.disp = 0; + + return reg; +} + +struct x86_reg x86_make_disp( struct x86_reg reg, + int disp ) +{ + assert(reg.file == file_REG32); + + if (reg.mod == mod_REG) + reg.disp = disp; + else + reg.disp += disp; + + if (reg.disp == 0) + reg.mod = mod_INDIRECT; + else if (reg.disp <= 127 && reg.disp >= -128) + reg.mod = mod_DISP8; + else + reg.mod = mod_DISP32; + + return reg; +} + +struct x86_reg x86_deref( struct x86_reg reg ) +{ + return x86_make_disp(reg, 0); +} + +struct x86_reg x86_get_base_reg( struct x86_reg reg ) +{ + return x86_make_reg( reg.file, reg.idx ); +} + +unsigned char *x86_get_label( struct x86_function *p ) +{ + return p->csr; +} + + + +/*********************************************************************** + * x86 instructions + */ + + +void x86_jcc( struct x86_function *p, + enum x86_cc cc, + unsigned char *label ) +{ + int offset = label - (x86_get_label(p) + 2); + + if (offset <= 127 && offset >= -128) { + emit_1ub(p, 0x70 + cc); + emit_1b(p, (char) offset); + } + else { + offset = label - (x86_get_label(p) + 6); + emit_2ub(p, 0x0f, 0x80 + cc); + emit_1i(p, offset); + } +} + +/* Always use a 32bit offset for forward jumps: + */ +unsigned char *x86_jcc_forward( struct x86_function *p, + enum x86_cc cc ) +{ + emit_2ub(p, 0x0f, 0x80 + cc); + emit_1i(p, 0); + return x86_get_label(p); +} + +unsigned char *x86_jmp_forward( struct x86_function *p) +{ + emit_1ub(p, 0xe9); + emit_1i(p, 0); + return x86_get_label(p); +} + +unsigned char *x86_call_forward( struct x86_function *p) +{ + emit_1ub(p, 0xe8); + emit_1i(p, 0); + return x86_get_label(p); +} + +/* Fixup offset from forward jump: + */ +void x86_fixup_fwd_jump( struct x86_function *p, + unsigned char *fixup ) +{ + *(int *)(fixup - 4) = x86_get_label(p) - fixup; +} + +void x86_jmp( struct x86_function *p, unsigned char *label) +{ + emit_1ub(p, 0xe9); + emit_1i(p, label - x86_get_label(p) - 4); +} + +#if 0 +/* This doesn't work once we start reallocating & copying the + * generated code on buffer fills, because the call is relative to the + * current pc. + */ +void x86_call( struct x86_function *p, void (*label)()) +{ + emit_1ub(p, 0xe8); + emit_1i(p, cptr(label) - x86_get_label(p) - 4); +} +#else +void x86_call( struct x86_function *p, struct x86_reg reg) +{ + emit_1ub(p, 0xff); + emit_modrm(p, reg, reg); +} +#endif + + +/* michal: + * Temporary. As I need immediate operands, and dont want to mess with the codegen, + * I load the immediate into general purpose register and use it. + */ +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) +{ + assert(dst.mod == mod_REG); + emit_1ub(p, 0xb8 + dst.idx); + emit_1i(p, imm); +} + +void x86_push( struct x86_function *p, + struct x86_reg reg ) +{ + assert(reg.mod == mod_REG); + emit_1ub(p, 0x50 + reg.idx); + p->stack_offset += 4; +} + +void x86_pop( struct x86_function *p, + struct x86_reg reg ) +{ + assert(reg.mod == mod_REG); + emit_1ub(p, 0x58 + reg.idx); + p->stack_offset -= 4; +} + +void x86_inc( struct x86_function *p, + struct x86_reg reg ) +{ + assert(reg.mod == mod_REG); + emit_1ub(p, 0x40 + reg.idx); +} + +void x86_dec( struct x86_function *p, + struct x86_reg reg ) +{ + assert(reg.mod == mod_REG); + emit_1ub(p, 0x48 + reg.idx); +} + +void x86_ret( struct x86_function *p ) +{ + emit_1ub(p, 0xc3); +} + +void x86_sahf( struct x86_function *p ) +{ + emit_1ub(p, 0x9e); +} + +void x86_mov( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x8b, 0x89, dst, src ); +} + +void x86_xor( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x33, 0x31, dst, src ); +} + +void x86_cmp( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x3b, 0x39, dst, src ); +} + +void x86_lea( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_1ub(p, 0x8d); + emit_modrm( p, dst, src ); +} + +void x86_test( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_1ub(p, 0x85); + emit_modrm( p, dst, src ); +} + +void x86_add( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm(p, 0x03, 0x01, dst, src ); +} + +void x86_mul( struct x86_function *p, + struct x86_reg src ) +{ + assert (src.file == file_REG32 && src.mod == mod_REG); + emit_op_modrm(p, 0xf7, 0, x86_make_reg (file_REG32, reg_SP), src ); +} + +void x86_sub( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm(p, 0x2b, 0x29, dst, src ); +} + +void x86_or( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x0b, 0x09, dst, src ); +} + +void x86_and( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_op_modrm( p, 0x23, 0x21, dst, src ); +} + + + +/*********************************************************************** + * SSE instructions + */ + + +void sse_movss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, 0xF3, X86_TWOB); + emit_op_modrm( p, 0x10, 0x11, dst, src ); +} + +void sse_movaps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x28, 0x29, dst, src ); +} + +void sse_movups( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x10, 0x11, dst, src ); +} + +void sse_movhps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.mod != mod_REG || src.mod != mod_REG); + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x16, 0x17, dst, src ); /* cf movlhps */ +} + +void sse_movlps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.mod != mod_REG || src.mod != mod_REG); + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x12, 0x13, dst, src ); /* cf movhlps */ +} + +void sse_maxps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x5F); + emit_modrm( p, dst, src ); +} + +void sse_maxss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x5F); + emit_modrm( p, dst, src ); +} + +void sse_divss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x5E); + emit_modrm( p, dst, src ); +} + +void sse_minps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x5D); + emit_modrm( p, dst, src ); +} + +void sse_subps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x5C); + emit_modrm( p, dst, src ); +} + +void sse_mulps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x59); + emit_modrm( p, dst, src ); +} + +void sse_mulss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x59); + emit_modrm( p, dst, src ); +} + +void sse_addps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x58); + emit_modrm( p, dst, src ); +} + +void sse_addss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x58); + emit_modrm( p, dst, src ); +} + +void sse_andnps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x55); + emit_modrm( p, dst, src ); +} + +void sse_andps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x54); + emit_modrm( p, dst, src ); +} + +void sse_rsqrtps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x52); + emit_modrm( p, dst, src ); +} + +void sse_rsqrtss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x52); + emit_modrm( p, dst, src ); + +} + +void sse_movhlps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.mod == mod_REG && src.mod == mod_REG); + emit_2ub(p, X86_TWOB, 0x12); + emit_modrm( p, dst, src ); +} + +void sse_movlhps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.mod == mod_REG && src.mod == mod_REG); + emit_2ub(p, X86_TWOB, 0x16); + emit_modrm( p, dst, src ); +} + +void sse_orps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x56); + emit_modrm( p, dst, src ); +} + +void sse_xorps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x57); + emit_modrm( p, dst, src ); +} + +void sse_cvtps2pi( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.file == file_MMX && + (src.file == file_XMM || src.mod != mod_REG)); + + p->need_emms = 1; + + emit_2ub(p, X86_TWOB, 0x2d); + emit_modrm( p, dst, src ); +} + + +/* Shufps can also be used to implement a reduced swizzle when dest == + * arg0. + */ +void sse_shufps( struct x86_function *p, + struct x86_reg dest, + struct x86_reg arg0, + unsigned char shuf) +{ + emit_2ub(p, X86_TWOB, 0xC6); + emit_modrm(p, dest, arg0); + emit_1ub(p, shuf); +} + +void sse_cmpps( struct x86_function *p, + struct x86_reg dest, + struct x86_reg arg0, + unsigned char cc) +{ + emit_2ub(p, X86_TWOB, 0xC2); + emit_modrm(p, dest, arg0); + emit_1ub(p, cc); +} + +void sse_pmovmskb( struct x86_function *p, + struct x86_reg dest, + struct x86_reg src) +{ + emit_3ub(p, 0x66, X86_TWOB, 0xD7); + emit_modrm(p, dest, src); +} + +/*********************************************************************** + * SSE2 instructions + */ + +/** + * Perform a reduced swizzle: + */ +void sse2_pshufd( struct x86_function *p, + struct x86_reg dest, + struct x86_reg arg0, + unsigned char shuf) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x70); + emit_modrm(p, dest, arg0); + emit_1ub(p, shuf); +} + +void sse2_cvttps2dq( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub( p, 0xF3, X86_TWOB, 0x5B ); + emit_modrm( p, dst, src ); +} + +void sse2_cvtps2dq( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x5B); + emit_modrm( p, dst, src ); +} + +void sse2_packssdw( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x6B); + emit_modrm( p, dst, src ); +} + +void sse2_packsswb( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x63); + emit_modrm( p, dst, src ); +} + +void sse2_packuswb( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x67); + emit_modrm( p, dst, src ); +} + +void sse2_rcpps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x53); + emit_modrm( p, dst, src ); +} + +void sse2_rcpss( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0xF3, X86_TWOB, 0x53); + emit_modrm( p, dst, src ); +} + +void sse2_movd( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, 0x66, X86_TWOB); + emit_op_modrm( p, 0x6e, 0x7e, dst, src ); +} + + + + +/*********************************************************************** + * x87 instructions + */ +void x87_fist( struct x86_function *p, struct x86_reg dst ) +{ + emit_1ub(p, 0xdb); + emit_modrm_noreg(p, 2, dst); +} + +void x87_fistp( struct x86_function *p, struct x86_reg dst ) +{ + emit_1ub(p, 0xdb); + emit_modrm_noreg(p, 3, dst); +} + +void x87_fild( struct x86_function *p, struct x86_reg arg ) +{ + emit_1ub(p, 0xdf); + emit_modrm_noreg(p, 0, arg); +} + +void x87_fldz( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xee); +} + + +void x87_fldcw( struct x86_function *p, struct x86_reg arg ) +{ + assert(arg.file == file_REG32); + assert(arg.mod != mod_REG); + emit_1ub(p, 0xd9); + emit_modrm_noreg(p, 5, arg); +} + +void x87_fld1( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xe8); +} + +void x87_fldl2e( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xea); +} + +void x87_fldln2( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xed); +} + +void x87_fwait( struct x86_function *p ) +{ + emit_1ub(p, 0x9b); +} + +void x87_fnclex( struct x86_function *p ) +{ + emit_2ub(p, 0xdb, 0xe2); +} + +void x87_fclex( struct x86_function *p ) +{ + x87_fwait(p); + x87_fnclex(p); +} + + +static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg, + unsigned char dst0ub0, + unsigned char dst0ub1, + unsigned char arg0ub0, + unsigned char arg0ub1, + unsigned char argmem_noreg) +{ + assert(dst.file == file_x87); + + if (arg.file == file_x87) { + if (dst.idx == 0) + emit_2ub(p, dst0ub0, dst0ub1+arg.idx); + else if (arg.idx == 0) + emit_2ub(p, arg0ub0, arg0ub1+arg.idx); + else + assert(0); + } + else if (dst.idx == 0) { + assert(arg.file == file_REG32); + emit_1ub(p, 0xd8); + emit_modrm_noreg(p, argmem_noreg, arg); + } + else + assert(0); +} + +void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xc8, + 0xdc, 0xc8, + 4); +} + +void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xe0, + 0xdc, 0xe8, + 4); +} + +void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xe8, + 0xdc, 0xe0, + 5); +} + +void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xc0, + 0xdc, 0xc0, + 0); +} + +void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xf0, + 0xdc, 0xf8, + 6); +} + +void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +{ + x87_arith_op(p, dst, arg, + 0xd8, 0xf8, + 0xdc, 0xf0, + 7); +} + +void x87_fmulp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xc8+dst.idx); +} + +void x87_fsubp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xe8+dst.idx); +} + +void x87_fsubrp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xe0+dst.idx); +} + +void x87_faddp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xc0+dst.idx); +} + +void x87_fdivp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xf8+dst.idx); +} + +void x87_fdivrp( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_x87); + assert(dst.idx >= 1); + emit_2ub(p, 0xde, 0xf0+dst.idx); +} + +void x87_fucom( struct x86_function *p, struct x86_reg arg ) +{ + assert(arg.file == file_x87); + emit_2ub(p, 0xdd, 0xe0+arg.idx); +} + +void x87_fucomp( struct x86_function *p, struct x86_reg arg ) +{ + assert(arg.file == file_x87); + emit_2ub(p, 0xdd, 0xe8+arg.idx); +} + +void x87_fucompp( struct x86_function *p ) +{ + emit_2ub(p, 0xda, 0xe9); +} + +void x87_fxch( struct x86_function *p, struct x86_reg arg ) +{ + assert(arg.file == file_x87); + emit_2ub(p, 0xd9, 0xc8+arg.idx); +} + +void x87_fabs( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xe1); +} + +void x87_fchs( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xe0); +} + +void x87_fcos( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xff); +} + + +void x87_fprndint( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfc); +} + +void x87_fscale( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfd); +} + +void x87_fsin( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfe); +} + +void x87_fsincos( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfb); +} + +void x87_fsqrt( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xfa); +} + +void x87_fxtract( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xf4); +} + +/* st0 = (2^st0)-1 + * + * Restrictions: -1.0 <= st0 <= 1.0 + */ +void x87_f2xm1( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xf0); +} + +/* st1 = st1 * log2(st0); + * pop_stack; + */ +void x87_fyl2x( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xf1); +} + +/* st1 = st1 * log2(st0 + 1.0); + * pop_stack; + * + * A fast operation, with restrictions: -.29 < st0 < .29 + */ +void x87_fyl2xp1( struct x86_function *p ) +{ + emit_2ub(p, 0xd9, 0xf9); +} + + +void x87_fld( struct x86_function *p, struct x86_reg arg ) +{ + if (arg.file == file_x87) + emit_2ub(p, 0xd9, 0xc0 + arg.idx); + else { + emit_1ub(p, 0xd9); + emit_modrm_noreg(p, 0, arg); + } +} + +void x87_fst( struct x86_function *p, struct x86_reg dst ) +{ + if (dst.file == file_x87) + emit_2ub(p, 0xdd, 0xd0 + dst.idx); + else { + emit_1ub(p, 0xd9); + emit_modrm_noreg(p, 2, dst); + } +} + +void x87_fstp( struct x86_function *p, struct x86_reg dst ) +{ + if (dst.file == file_x87) + emit_2ub(p, 0xdd, 0xd8 + dst.idx); + else { + emit_1ub(p, 0xd9); + emit_modrm_noreg(p, 3, dst); + } +} + +void x87_fcom( struct x86_function *p, struct x86_reg dst ) +{ + if (dst.file == file_x87) + emit_2ub(p, 0xd8, 0xd0 + dst.idx); + else { + emit_1ub(p, 0xd8); + emit_modrm_noreg(p, 2, dst); + } +} + +void x87_fcomp( struct x86_function *p, struct x86_reg dst ) +{ + if (dst.file == file_x87) + emit_2ub(p, 0xd8, 0xd8 + dst.idx); + else { + emit_1ub(p, 0xd8); + emit_modrm_noreg(p, 3, dst); + } +} + + +void x87_fnstsw( struct x86_function *p, struct x86_reg dst ) +{ + assert(dst.file == file_REG32); + + if (dst.idx == reg_AX && + dst.mod == mod_REG) + emit_2ub(p, 0xdf, 0xe0); + else { + emit_1ub(p, 0xdd); + emit_modrm_noreg(p, 7, dst); + } +} + + + + +/*********************************************************************** + * MMX instructions + */ + +void mmx_emms( struct x86_function *p ) +{ + assert(p->need_emms); + emit_2ub(p, 0x0f, 0x77); + p->need_emms = 0; +} + +void mmx_packssdw( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.file == file_MMX && + (src.file == file_MMX || src.mod != mod_REG)); + + p->need_emms = 1; + + emit_2ub(p, X86_TWOB, 0x6b); + emit_modrm( p, dst, src ); +} + +void mmx_packuswb( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + assert(dst.file == file_MMX && + (src.file == file_MMX || src.mod != mod_REG)); + + p->need_emms = 1; + + emit_2ub(p, X86_TWOB, 0x67); + emit_modrm( p, dst, src ); +} + +void mmx_movd( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + p->need_emms = 1; + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x6e, 0x7e, dst, src ); +} + +void mmx_movq( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + p->need_emms = 1; + emit_1ub(p, X86_TWOB); + emit_op_modrm( p, 0x6f, 0x7f, dst, src ); +} + + +/*********************************************************************** + * Helper functions + */ + + +/* Retreive a reference to one of the function arguments, taking into + * account any push/pop activity: + */ +struct x86_reg x86_fn_arg( struct x86_function *p, + unsigned arg ) +{ + return x86_make_disp(x86_make_reg(file_REG32, reg_SP), + p->stack_offset + arg * 4); /* ??? */ +} + + +void x86_init_func( struct x86_function *p ) +{ + p->size = 0; + p->store = NULL; + p->csr = p->store; +} + +void x86_init_func_size( struct x86_function *p, unsigned code_size ) +{ + p->size = code_size; + p->store = rtasm_exec_malloc(code_size); + p->csr = p->store; +} + +void x86_release_func( struct x86_function *p ) +{ + rtasm_exec_free(p->store); + p->store = NULL; + p->csr = NULL; + p->size = 0; +} + + +void (*x86_get_func( struct x86_function *p ))(void) +{ + if (DISASSEM && p->store) + debug_printf("disassemble %p %p\n", p->store, p->csr); + return (void (*)(void)) (unsigned long) p->store; +} + +#else + +void x86sse_dummy( void ) +{ +} + +#endif diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h new file mode 100644 index 0000000000..c2aa416492 --- /dev/null +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -0,0 +1,256 @@ + +#ifndef _X86SSE_H_ +#define _X86SSE_H_ + +#if defined(__i386__) || defined(__386__) + +/* It is up to the caller to ensure that instructions issued are + * suitable for the host cpu. There are no checks made in this module + * for mmx/sse/sse2 support on the cpu. + */ +struct x86_reg { + unsigned file:3; + unsigned idx:3; + unsigned mod:2; /* mod_REG if this is just a register */ + int disp:24; /* only +/- 23bits of offset - should be enough... */ +}; + +struct x86_function { + unsigned size; + unsigned char *store; + unsigned char *csr; + unsigned stack_offset; + int need_emms; + const char *fn; +}; + +enum x86_reg_file { + file_REG32, + file_MMX, + file_XMM, + file_x87 +}; + +/* Values for mod field of modr/m byte + */ +enum x86_reg_mod { + mod_INDIRECT, + mod_DISP8, + mod_DISP32, + mod_REG +}; + +enum x86_reg_name { + reg_AX, + reg_CX, + reg_DX, + reg_BX, + reg_SP, + reg_BP, + reg_SI, + reg_DI +}; + + +enum x86_cc { + cc_O, /* overflow */ + cc_NO, /* not overflow */ + cc_NAE, /* not above or equal / carry */ + cc_AE, /* above or equal / not carry */ + cc_E, /* equal / zero */ + cc_NE /* not equal / not zero */ +}; + +enum sse_cc { + cc_Equal, + cc_LessThan, + cc_LessThanEqual, + cc_Unordered, + cc_NotEqual, + cc_NotLessThan, + cc_NotLessThanEqual, + cc_Ordered +}; + +#define cc_Z cc_E +#define cc_NZ cc_NE + +/* Begin/end/retreive function creation: + */ + + +void x86_init_func( struct x86_function *p ); +void x86_init_func_size( struct x86_function *p, unsigned code_size ); +void x86_release_func( struct x86_function *p ); +void (*x86_get_func( struct x86_function *p ))( void ); + + + +/* Create and manipulate registers and regmem values: + */ +struct x86_reg x86_make_reg( enum x86_reg_file file, + enum x86_reg_name idx ); + +struct x86_reg x86_make_disp( struct x86_reg reg, + int disp ); + +struct x86_reg x86_deref( struct x86_reg reg ); + +struct x86_reg x86_get_base_reg( struct x86_reg reg ); + + +/* Labels, jumps and fixup: + */ +unsigned char *x86_get_label( struct x86_function *p ); + +void x86_jcc( struct x86_function *p, + enum x86_cc cc, + unsigned char *label ); + +unsigned char *x86_jcc_forward( struct x86_function *p, + enum x86_cc cc ); + +unsigned char *x86_jmp_forward( struct x86_function *p); + +unsigned char *x86_call_forward( struct x86_function *p); + +void x86_fixup_fwd_jump( struct x86_function *p, + unsigned char *fixup ); + +void x86_jmp( struct x86_function *p, unsigned char *label ); + +/* void x86_call( struct x86_function *p, void (*label)() ); */ +void x86_call( struct x86_function *p, struct x86_reg reg); + +/* michal: + * Temporary. As I need immediate operands, and dont want to mess with the codegen, + * I load the immediate into general purpose register and use it. + */ +void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ); + + +/* Macro for sse_shufps() and sse2_pshufd(): + */ +#define SHUF(_x,_y,_z,_w) (((_x)<<0) | ((_y)<<2) | ((_z)<<4) | ((_w)<<6)) +#define SHUF_NOOP RSW(0,1,2,3) +#define GET_SHUF(swz, idx) (((swz) >> ((idx)*2)) & 0x3) + +void mmx_emms( struct x86_function *p ); +void mmx_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void mmx_movq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void mmx_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void mmx_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); + +void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_cvttps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, + unsigned char shuf ); +void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); + +void sse_addps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_addss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, + unsigned char cc ); +void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movaps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movhlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movlhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_movups( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_mulps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_mulss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_orps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, + unsigned char shuf ); +void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); + +void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_cmp( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_dec( struct x86_function *p, struct x86_reg reg ); +void x86_inc( struct x86_function *p, struct x86_reg reg ); +void x86_lea( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_mov( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_mul( struct x86_function *p, struct x86_reg src ); +void x86_or( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_pop( struct x86_function *p, struct x86_reg reg ); +void x86_push( struct x86_function *p, struct x86_reg reg ); +void x86_ret( struct x86_function *p ); +void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void x86_sahf( struct x86_function *p ); + +void x87_f2xm1( struct x86_function *p ); +void x87_fabs( struct x86_function *p ); +void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_faddp( struct x86_function *p, struct x86_reg dst ); +void x87_fchs( struct x86_function *p ); +void x87_fclex( struct x86_function *p ); +void x87_fcom( struct x86_function *p, struct x86_reg dst ); +void x87_fcomp( struct x86_function *p, struct x86_reg dst ); +void x87_fcos( struct x86_function *p ); +void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fdivp( struct x86_function *p, struct x86_reg dst ); +void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fdivrp( struct x86_function *p, struct x86_reg dst ); +void x87_fild( struct x86_function *p, struct x86_reg arg ); +void x87_fist( struct x86_function *p, struct x86_reg dst ); +void x87_fistp( struct x86_function *p, struct x86_reg dst ); +void x87_fld( struct x86_function *p, struct x86_reg arg ); +void x87_fld1( struct x86_function *p ); +void x87_fldcw( struct x86_function *p, struct x86_reg arg ); +void x87_fldl2e( struct x86_function *p ); +void x87_fldln2( struct x86_function *p ); +void x87_fldz( struct x86_function *p ); +void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fmulp( struct x86_function *p, struct x86_reg dst ); +void x87_fnclex( struct x86_function *p ); +void x87_fprndint( struct x86_function *p ); +void x87_fscale( struct x86_function *p ); +void x87_fsin( struct x86_function *p ); +void x87_fsincos( struct x86_function *p ); +void x87_fsqrt( struct x86_function *p ); +void x87_fst( struct x86_function *p, struct x86_reg dst ); +void x87_fstp( struct x86_function *p, struct x86_reg dst ); +void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fsubp( struct x86_function *p, struct x86_reg dst ); +void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); +void x87_fsubrp( struct x86_function *p, struct x86_reg dst ); +void x87_fxch( struct x86_function *p, struct x86_reg dst ); +void x87_fxtract( struct x86_function *p ); +void x87_fyl2x( struct x86_function *p ); +void x87_fyl2xp1( struct x86_function *p ); +void x87_fwait( struct x86_function *p ); +void x87_fnstsw( struct x86_function *p, struct x86_reg dst ); +void x87_fucompp( struct x86_function *p ); +void x87_fucomp( struct x86_function *p, struct x86_reg arg ); +void x87_fucom( struct x86_function *p, struct x86_reg arg ); + + + +/* Retreive a reference to one of the function arguments, taking into + * account any push/pop activity. Note - doesn't track explict + * manipulation of ESP by other instructions. + */ +struct x86_reg x86_fn_arg( struct x86_function *p, unsigned arg ); + +#endif +#endif diff --git a/src/gallium/auxiliary/rtasm/x86sse.c b/src/gallium/auxiliary/rtasm/x86sse.c deleted file mode 100644 index fff6f77a6b..0000000000 --- a/src/gallium/auxiliary/rtasm/x86sse.c +++ /dev/null @@ -1,1195 +0,0 @@ -#if defined(__i386__) || defined(__386__) - -#include "pipe/p_compiler.h" -#include "pipe/p_debug.h" - -#include "x86sse.h" - -#define DISASSEM 0 -#define X86_TWOB 0x0f - -static unsigned char *cptr( void (*label)() ) -{ - return (unsigned char *)(unsigned long)label; -} - - -static void do_realloc( struct x86_function *p ) -{ - if (p->size == 0) { - p->size = 1024; - p->store = _mesa_exec_malloc(p->size); - p->csr = p->store; - } - else { - unsigned used = p->csr - p->store; - unsigned char *tmp = p->store; - p->size *= 2; - p->store = _mesa_exec_malloc(p->size); - memcpy(p->store, tmp, used); - p->csr = p->store + used; - _mesa_exec_free(tmp); - } -} - -/* Emit bytes to the instruction stream: - */ -static unsigned char *reserve( struct x86_function *p, int bytes ) -{ - if (p->csr + bytes - p->store > p->size) - do_realloc(p); - - { - unsigned char *csr = p->csr; - p->csr += bytes; - return csr; - } -} - - - -static void emit_1b( struct x86_function *p, char b0 ) -{ - char *csr = (char *)reserve(p, 1); - *csr = b0; -} - -static void emit_1i( struct x86_function *p, int i0 ) -{ - int *icsr = (int *)reserve(p, sizeof(i0)); - *icsr = i0; -} - -static void emit_1ub( struct x86_function *p, unsigned char b0 ) -{ - unsigned char *csr = reserve(p, 1); - *csr++ = b0; -} - -static void emit_2ub( struct x86_function *p, unsigned char b0, unsigned char b1 ) -{ - unsigned char *csr = reserve(p, 2); - *csr++ = b0; - *csr++ = b1; -} - -static void emit_3ub( struct x86_function *p, unsigned char b0, unsigned char b1, unsigned char b2 ) -{ - unsigned char *csr = reserve(p, 3); - *csr++ = b0; - *csr++ = b1; - *csr++ = b2; -} - - -/* Build a modRM byte + possible displacement. No treatment of SIB - * indexing. BZZT - no way to encode an absolute address. - */ -static void emit_modrm( struct x86_function *p, - struct x86_reg reg, - struct x86_reg regmem ) -{ - unsigned char val = 0; - - assert(reg.mod == mod_REG); - - val |= regmem.mod << 6; /* mod field */ - val |= reg.idx << 3; /* reg field */ - val |= regmem.idx; /* r/m field */ - - emit_1ub(p, val); - - /* Oh-oh we've stumbled into the SIB thing. - */ - if (regmem.file == file_REG32 && - regmem.idx == reg_SP) { - emit_1ub(p, 0x24); /* simplistic! */ - } - - switch (regmem.mod) { - case mod_REG: - case mod_INDIRECT: - break; - case mod_DISP8: - emit_1b(p, regmem.disp); - break; - case mod_DISP32: - emit_1i(p, regmem.disp); - break; - default: - assert(0); - break; - } -} - - -static void emit_modrm_noreg( struct x86_function *p, - unsigned op, - struct x86_reg regmem ) -{ - struct x86_reg dummy = x86_make_reg(file_REG32, op); - emit_modrm(p, dummy, regmem); -} - -/* Many x86 instructions have two opcodes to cope with the situations - * where the destination is a register or memory reference - * respectively. This function selects the correct opcode based on - * the arguments presented. - */ -static void emit_op_modrm( struct x86_function *p, - unsigned char op_dst_is_reg, - unsigned char op_dst_is_mem, - struct x86_reg dst, - struct x86_reg src ) -{ - switch (dst.mod) { - case mod_REG: - emit_1ub(p, op_dst_is_reg); - emit_modrm(p, dst, src); - break; - case mod_INDIRECT: - case mod_DISP32: - case mod_DISP8: - assert(src.mod == mod_REG); - emit_1ub(p, op_dst_is_mem); - emit_modrm(p, src, dst); - break; - default: - assert(0); - break; - } -} - - - - - - - -/* Create and manipulate registers and regmem values: - */ -struct x86_reg x86_make_reg( enum x86_reg_file file, - enum x86_reg_name idx ) -{ - struct x86_reg reg; - - reg.file = file; - reg.idx = idx; - reg.mod = mod_REG; - reg.disp = 0; - - return reg; -} - -struct x86_reg x86_make_disp( struct x86_reg reg, - int disp ) -{ - assert(reg.file == file_REG32); - - if (reg.mod == mod_REG) - reg.disp = disp; - else - reg.disp += disp; - - if (reg.disp == 0) - reg.mod = mod_INDIRECT; - else if (reg.disp <= 127 && reg.disp >= -128) - reg.mod = mod_DISP8; - else - reg.mod = mod_DISP32; - - return reg; -} - -struct x86_reg x86_deref( struct x86_reg reg ) -{ - return x86_make_disp(reg, 0); -} - -struct x86_reg x86_get_base_reg( struct x86_reg reg ) -{ - return x86_make_reg( reg.file, reg.idx ); -} - -unsigned char *x86_get_label( struct x86_function *p ) -{ - return p->csr; -} - - - -/*********************************************************************** - * x86 instructions - */ - - -void x86_jcc( struct x86_function *p, - enum x86_cc cc, - unsigned char *label ) -{ - int offset = label - (x86_get_label(p) + 2); - - if (offset <= 127 && offset >= -128) { - emit_1ub(p, 0x70 + cc); - emit_1b(p, (char) offset); - } - else { - offset = label - (x86_get_label(p) + 6); - emit_2ub(p, 0x0f, 0x80 + cc); - emit_1i(p, offset); - } -} - -/* Always use a 32bit offset for forward jumps: - */ -unsigned char *x86_jcc_forward( struct x86_function *p, - enum x86_cc cc ) -{ - emit_2ub(p, 0x0f, 0x80 + cc); - emit_1i(p, 0); - return x86_get_label(p); -} - -unsigned char *x86_jmp_forward( struct x86_function *p) -{ - emit_1ub(p, 0xe9); - emit_1i(p, 0); - return x86_get_label(p); -} - -unsigned char *x86_call_forward( struct x86_function *p) -{ - emit_1ub(p, 0xe8); - emit_1i(p, 0); - return x86_get_label(p); -} - -/* Fixup offset from forward jump: - */ -void x86_fixup_fwd_jump( struct x86_function *p, - unsigned char *fixup ) -{ - *(int *)(fixup - 4) = x86_get_label(p) - fixup; -} - -void x86_jmp( struct x86_function *p, unsigned char *label) -{ - emit_1ub(p, 0xe9); - emit_1i(p, label - x86_get_label(p) - 4); -} - -#if 0 -/* This doesn't work once we start reallocating & copying the - * generated code on buffer fills, because the call is relative to the - * current pc. - */ -void x86_call( struct x86_function *p, void (*label)()) -{ - emit_1ub(p, 0xe8); - emit_1i(p, cptr(label) - x86_get_label(p) - 4); -} -#else -void x86_call( struct x86_function *p, struct x86_reg reg) -{ - emit_1ub(p, 0xff); - emit_modrm(p, reg, reg); -} -#endif - - -/* michal: - * Temporary. As I need immediate operands, and dont want to mess with the codegen, - * I load the immediate into general purpose register and use it. - */ -void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) -{ - assert(dst.mod == mod_REG); - emit_1ub(p, 0xb8 + dst.idx); - emit_1i(p, imm); -} - -void x86_push( struct x86_function *p, - struct x86_reg reg ) -{ - assert(reg.mod == mod_REG); - emit_1ub(p, 0x50 + reg.idx); - p->stack_offset += 4; -} - -void x86_pop( struct x86_function *p, - struct x86_reg reg ) -{ - assert(reg.mod == mod_REG); - emit_1ub(p, 0x58 + reg.idx); - p->stack_offset -= 4; -} - -void x86_inc( struct x86_function *p, - struct x86_reg reg ) -{ - assert(reg.mod == mod_REG); - emit_1ub(p, 0x40 + reg.idx); -} - -void x86_dec( struct x86_function *p, - struct x86_reg reg ) -{ - assert(reg.mod == mod_REG); - emit_1ub(p, 0x48 + reg.idx); -} - -void x86_ret( struct x86_function *p ) -{ - emit_1ub(p, 0xc3); -} - -void x86_sahf( struct x86_function *p ) -{ - emit_1ub(p, 0x9e); -} - -void x86_mov( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_op_modrm( p, 0x8b, 0x89, dst, src ); -} - -void x86_xor( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_op_modrm( p, 0x33, 0x31, dst, src ); -} - -void x86_cmp( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_op_modrm( p, 0x3b, 0x39, dst, src ); -} - -void x86_lea( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_1ub(p, 0x8d); - emit_modrm( p, dst, src ); -} - -void x86_test( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_1ub(p, 0x85); - emit_modrm( p, dst, src ); -} - -void x86_add( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_op_modrm(p, 0x03, 0x01, dst, src ); -} - -void x86_mul( struct x86_function *p, - struct x86_reg src ) -{ - assert (src.file == file_REG32 && src.mod == mod_REG); - emit_op_modrm(p, 0xf7, 0, x86_make_reg (file_REG32, reg_SP), src ); -} - -void x86_sub( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_op_modrm(p, 0x2b, 0x29, dst, src ); -} - -void x86_or( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_op_modrm( p, 0x0b, 0x09, dst, src ); -} - -void x86_and( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_op_modrm( p, 0x23, 0x21, dst, src ); -} - - - -/*********************************************************************** - * SSE instructions - */ - - -void sse_movss( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, 0xF3, X86_TWOB); - emit_op_modrm( p, 0x10, 0x11, dst, src ); -} - -void sse_movaps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_1ub(p, X86_TWOB); - emit_op_modrm( p, 0x28, 0x29, dst, src ); -} - -void sse_movups( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_1ub(p, X86_TWOB); - emit_op_modrm( p, 0x10, 0x11, dst, src ); -} - -void sse_movhps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - assert(dst.mod != mod_REG || src.mod != mod_REG); - emit_1ub(p, X86_TWOB); - emit_op_modrm( p, 0x16, 0x17, dst, src ); /* cf movlhps */ -} - -void sse_movlps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - assert(dst.mod != mod_REG || src.mod != mod_REG); - emit_1ub(p, X86_TWOB); - emit_op_modrm( p, 0x12, 0x13, dst, src ); /* cf movhlps */ -} - -void sse_maxps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x5F); - emit_modrm( p, dst, src ); -} - -void sse_maxss( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0xF3, X86_TWOB, 0x5F); - emit_modrm( p, dst, src ); -} - -void sse_divss( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0xF3, X86_TWOB, 0x5E); - emit_modrm( p, dst, src ); -} - -void sse_minps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x5D); - emit_modrm( p, dst, src ); -} - -void sse_subps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x5C); - emit_modrm( p, dst, src ); -} - -void sse_mulps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x59); - emit_modrm( p, dst, src ); -} - -void sse_mulss( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0xF3, X86_TWOB, 0x59); - emit_modrm( p, dst, src ); -} - -void sse_addps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x58); - emit_modrm( p, dst, src ); -} - -void sse_addss( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0xF3, X86_TWOB, 0x58); - emit_modrm( p, dst, src ); -} - -void sse_andnps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x55); - emit_modrm( p, dst, src ); -} - -void sse_andps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x54); - emit_modrm( p, dst, src ); -} - -void sse_rsqrtps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x52); - emit_modrm( p, dst, src ); -} - -void sse_rsqrtss( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0xF3, X86_TWOB, 0x52); - emit_modrm( p, dst, src ); - -} - -void sse_movhlps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - assert(dst.mod == mod_REG && src.mod == mod_REG); - emit_2ub(p, X86_TWOB, 0x12); - emit_modrm( p, dst, src ); -} - -void sse_movlhps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - assert(dst.mod == mod_REG && src.mod == mod_REG); - emit_2ub(p, X86_TWOB, 0x16); - emit_modrm( p, dst, src ); -} - -void sse_orps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x56); - emit_modrm( p, dst, src ); -} - -void sse_xorps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x57); - emit_modrm( p, dst, src ); -} - -void sse_cvtps2pi( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - assert(dst.file == file_MMX && - (src.file == file_XMM || src.mod != mod_REG)); - - p->need_emms = 1; - - emit_2ub(p, X86_TWOB, 0x2d); - emit_modrm( p, dst, src ); -} - - -/* Shufps can also be used to implement a reduced swizzle when dest == - * arg0. - */ -void sse_shufps( struct x86_function *p, - struct x86_reg dest, - struct x86_reg arg0, - unsigned char shuf) -{ - emit_2ub(p, X86_TWOB, 0xC6); - emit_modrm(p, dest, arg0); - emit_1ub(p, shuf); -} - -void sse_cmpps( struct x86_function *p, - struct x86_reg dest, - struct x86_reg arg0, - unsigned char cc) -{ - emit_2ub(p, X86_TWOB, 0xC2); - emit_modrm(p, dest, arg0); - emit_1ub(p, cc); -} - -void sse_pmovmskb( struct x86_function *p, - struct x86_reg dest, - struct x86_reg src) -{ - emit_3ub(p, 0x66, X86_TWOB, 0xD7); - emit_modrm(p, dest, src); -} - -/*********************************************************************** - * SSE2 instructions - */ - -/** - * Perform a reduced swizzle: - */ -void sse2_pshufd( struct x86_function *p, - struct x86_reg dest, - struct x86_reg arg0, - unsigned char shuf) -{ - emit_3ub(p, 0x66, X86_TWOB, 0x70); - emit_modrm(p, dest, arg0); - emit_1ub(p, shuf); -} - -void sse2_cvttps2dq( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub( p, 0xF3, X86_TWOB, 0x5B ); - emit_modrm( p, dst, src ); -} - -void sse2_cvtps2dq( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0x66, X86_TWOB, 0x5B); - emit_modrm( p, dst, src ); -} - -void sse2_packssdw( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0x66, X86_TWOB, 0x6B); - emit_modrm( p, dst, src ); -} - -void sse2_packsswb( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0x66, X86_TWOB, 0x63); - emit_modrm( p, dst, src ); -} - -void sse2_packuswb( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0x66, X86_TWOB, 0x67); - emit_modrm( p, dst, src ); -} - -void sse2_rcpps( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, X86_TWOB, 0x53); - emit_modrm( p, dst, src ); -} - -void sse2_rcpss( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_3ub(p, 0xF3, X86_TWOB, 0x53); - emit_modrm( p, dst, src ); -} - -void sse2_movd( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - emit_2ub(p, 0x66, X86_TWOB); - emit_op_modrm( p, 0x6e, 0x7e, dst, src ); -} - - - - -/*********************************************************************** - * x87 instructions - */ -void x87_fist( struct x86_function *p, struct x86_reg dst ) -{ - emit_1ub(p, 0xdb); - emit_modrm_noreg(p, 2, dst); -} - -void x87_fistp( struct x86_function *p, struct x86_reg dst ) -{ - emit_1ub(p, 0xdb); - emit_modrm_noreg(p, 3, dst); -} - -void x87_fild( struct x86_function *p, struct x86_reg arg ) -{ - emit_1ub(p, 0xdf); - emit_modrm_noreg(p, 0, arg); -} - -void x87_fldz( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xee); -} - - -void x87_fldcw( struct x86_function *p, struct x86_reg arg ) -{ - assert(arg.file == file_REG32); - assert(arg.mod != mod_REG); - emit_1ub(p, 0xd9); - emit_modrm_noreg(p, 5, arg); -} - -void x87_fld1( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xe8); -} - -void x87_fldl2e( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xea); -} - -void x87_fldln2( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xed); -} - -void x87_fwait( struct x86_function *p ) -{ - emit_1ub(p, 0x9b); -} - -void x87_fnclex( struct x86_function *p ) -{ - emit_2ub(p, 0xdb, 0xe2); -} - -void x87_fclex( struct x86_function *p ) -{ - x87_fwait(p); - x87_fnclex(p); -} - - -static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg, - unsigned char dst0ub0, - unsigned char dst0ub1, - unsigned char arg0ub0, - unsigned char arg0ub1, - unsigned char argmem_noreg) -{ - assert(dst.file == file_x87); - - if (arg.file == file_x87) { - if (dst.idx == 0) - emit_2ub(p, dst0ub0, dst0ub1+arg.idx); - else if (arg.idx == 0) - emit_2ub(p, arg0ub0, arg0ub1+arg.idx); - else - assert(0); - } - else if (dst.idx == 0) { - assert(arg.file == file_REG32); - emit_1ub(p, 0xd8); - emit_modrm_noreg(p, argmem_noreg, arg); - } - else - assert(0); -} - -void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) -{ - x87_arith_op(p, dst, arg, - 0xd8, 0xc8, - 0xdc, 0xc8, - 4); -} - -void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) -{ - x87_arith_op(p, dst, arg, - 0xd8, 0xe0, - 0xdc, 0xe8, - 4); -} - -void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) -{ - x87_arith_op(p, dst, arg, - 0xd8, 0xe8, - 0xdc, 0xe0, - 5); -} - -void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) -{ - x87_arith_op(p, dst, arg, - 0xd8, 0xc0, - 0xdc, 0xc0, - 0); -} - -void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) -{ - x87_arith_op(p, dst, arg, - 0xd8, 0xf0, - 0xdc, 0xf8, - 6); -} - -void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) -{ - x87_arith_op(p, dst, arg, - 0xd8, 0xf8, - 0xdc, 0xf0, - 7); -} - -void x87_fmulp( struct x86_function *p, struct x86_reg dst ) -{ - assert(dst.file == file_x87); - assert(dst.idx >= 1); - emit_2ub(p, 0xde, 0xc8+dst.idx); -} - -void x87_fsubp( struct x86_function *p, struct x86_reg dst ) -{ - assert(dst.file == file_x87); - assert(dst.idx >= 1); - emit_2ub(p, 0xde, 0xe8+dst.idx); -} - -void x87_fsubrp( struct x86_function *p, struct x86_reg dst ) -{ - assert(dst.file == file_x87); - assert(dst.idx >= 1); - emit_2ub(p, 0xde, 0xe0+dst.idx); -} - -void x87_faddp( struct x86_function *p, struct x86_reg dst ) -{ - assert(dst.file == file_x87); - assert(dst.idx >= 1); - emit_2ub(p, 0xde, 0xc0+dst.idx); -} - -void x87_fdivp( struct x86_function *p, struct x86_reg dst ) -{ - assert(dst.file == file_x87); - assert(dst.idx >= 1); - emit_2ub(p, 0xde, 0xf8+dst.idx); -} - -void x87_fdivrp( struct x86_function *p, struct x86_reg dst ) -{ - assert(dst.file == file_x87); - assert(dst.idx >= 1); - emit_2ub(p, 0xde, 0xf0+dst.idx); -} - -void x87_fucom( struct x86_function *p, struct x86_reg arg ) -{ - assert(arg.file == file_x87); - emit_2ub(p, 0xdd, 0xe0+arg.idx); -} - -void x87_fucomp( struct x86_function *p, struct x86_reg arg ) -{ - assert(arg.file == file_x87); - emit_2ub(p, 0xdd, 0xe8+arg.idx); -} - -void x87_fucompp( struct x86_function *p ) -{ - emit_2ub(p, 0xda, 0xe9); -} - -void x87_fxch( struct x86_function *p, struct x86_reg arg ) -{ - assert(arg.file == file_x87); - emit_2ub(p, 0xd9, 0xc8+arg.idx); -} - -void x87_fabs( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xe1); -} - -void x87_fchs( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xe0); -} - -void x87_fcos( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xff); -} - - -void x87_fprndint( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xfc); -} - -void x87_fscale( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xfd); -} - -void x87_fsin( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xfe); -} - -void x87_fsincos( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xfb); -} - -void x87_fsqrt( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xfa); -} - -void x87_fxtract( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xf4); -} - -/* st0 = (2^st0)-1 - * - * Restrictions: -1.0 <= st0 <= 1.0 - */ -void x87_f2xm1( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xf0); -} - -/* st1 = st1 * log2(st0); - * pop_stack; - */ -void x87_fyl2x( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xf1); -} - -/* st1 = st1 * log2(st0 + 1.0); - * pop_stack; - * - * A fast operation, with restrictions: -.29 < st0 < .29 - */ -void x87_fyl2xp1( struct x86_function *p ) -{ - emit_2ub(p, 0xd9, 0xf9); -} - - -void x87_fld( struct x86_function *p, struct x86_reg arg ) -{ - if (arg.file == file_x87) - emit_2ub(p, 0xd9, 0xc0 + arg.idx); - else { - emit_1ub(p, 0xd9); - emit_modrm_noreg(p, 0, arg); - } -} - -void x87_fst( struct x86_function *p, struct x86_reg dst ) -{ - if (dst.file == file_x87) - emit_2ub(p, 0xdd, 0xd0 + dst.idx); - else { - emit_1ub(p, 0xd9); - emit_modrm_noreg(p, 2, dst); - } -} - -void x87_fstp( struct x86_function *p, struct x86_reg dst ) -{ - if (dst.file == file_x87) - emit_2ub(p, 0xdd, 0xd8 + dst.idx); - else { - emit_1ub(p, 0xd9); - emit_modrm_noreg(p, 3, dst); - } -} - -void x87_fcom( struct x86_function *p, struct x86_reg dst ) -{ - if (dst.file == file_x87) - emit_2ub(p, 0xd8, 0xd0 + dst.idx); - else { - emit_1ub(p, 0xd8); - emit_modrm_noreg(p, 2, dst); - } -} - -void x87_fcomp( struct x86_function *p, struct x86_reg dst ) -{ - if (dst.file == file_x87) - emit_2ub(p, 0xd8, 0xd8 + dst.idx); - else { - emit_1ub(p, 0xd8); - emit_modrm_noreg(p, 3, dst); - } -} - - -void x87_fnstsw( struct x86_function *p, struct x86_reg dst ) -{ - assert(dst.file == file_REG32); - - if (dst.idx == reg_AX && - dst.mod == mod_REG) - emit_2ub(p, 0xdf, 0xe0); - else { - emit_1ub(p, 0xdd); - emit_modrm_noreg(p, 7, dst); - } -} - - - - -/*********************************************************************** - * MMX instructions - */ - -void mmx_emms( struct x86_function *p ) -{ - assert(p->need_emms); - emit_2ub(p, 0x0f, 0x77); - p->need_emms = 0; -} - -void mmx_packssdw( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - assert(dst.file == file_MMX && - (src.file == file_MMX || src.mod != mod_REG)); - - p->need_emms = 1; - - emit_2ub(p, X86_TWOB, 0x6b); - emit_modrm( p, dst, src ); -} - -void mmx_packuswb( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - assert(dst.file == file_MMX && - (src.file == file_MMX || src.mod != mod_REG)); - - p->need_emms = 1; - - emit_2ub(p, X86_TWOB, 0x67); - emit_modrm( p, dst, src ); -} - -void mmx_movd( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - p->need_emms = 1; - emit_1ub(p, X86_TWOB); - emit_op_modrm( p, 0x6e, 0x7e, dst, src ); -} - -void mmx_movq( struct x86_function *p, - struct x86_reg dst, - struct x86_reg src ) -{ - p->need_emms = 1; - emit_1ub(p, X86_TWOB); - emit_op_modrm( p, 0x6f, 0x7f, dst, src ); -} - - -/*********************************************************************** - * Helper functions - */ - - -/* Retreive a reference to one of the function arguments, taking into - * account any push/pop activity: - */ -struct x86_reg x86_fn_arg( struct x86_function *p, - unsigned arg ) -{ - return x86_make_disp(x86_make_reg(file_REG32, reg_SP), - p->stack_offset + arg * 4); /* ??? */ -} - - -void x86_init_func( struct x86_function *p ) -{ - p->size = 0; - p->store = NULL; - p->csr = p->store; -} - -void x86_init_func_size( struct x86_function *p, unsigned code_size ) -{ - p->size = code_size; - p->store = _mesa_exec_malloc(code_size); - p->csr = p->store; -} - -void x86_release_func( struct x86_function *p ) -{ - _mesa_exec_free(p->store); - p->store = NULL; - p->csr = NULL; - p->size = 0; -} - - -void (*x86_get_func( struct x86_function *p ))(void) -{ - if (DISASSEM && p->store) - _mesa_printf("disassemble %p %p\n", p->store, p->csr); - return (void (*)(void)) (unsigned long) p->store; -} - -#else - -void x86sse_dummy( void ) -{ -} - -#endif diff --git a/src/gallium/auxiliary/rtasm/x86sse.h b/src/gallium/auxiliary/rtasm/x86sse.h deleted file mode 100644 index c2aa416492..0000000000 --- a/src/gallium/auxiliary/rtasm/x86sse.h +++ /dev/null @@ -1,256 +0,0 @@ - -#ifndef _X86SSE_H_ -#define _X86SSE_H_ - -#if defined(__i386__) || defined(__386__) - -/* It is up to the caller to ensure that instructions issued are - * suitable for the host cpu. There are no checks made in this module - * for mmx/sse/sse2 support on the cpu. - */ -struct x86_reg { - unsigned file:3; - unsigned idx:3; - unsigned mod:2; /* mod_REG if this is just a register */ - int disp:24; /* only +/- 23bits of offset - should be enough... */ -}; - -struct x86_function { - unsigned size; - unsigned char *store; - unsigned char *csr; - unsigned stack_offset; - int need_emms; - const char *fn; -}; - -enum x86_reg_file { - file_REG32, - file_MMX, - file_XMM, - file_x87 -}; - -/* Values for mod field of modr/m byte - */ -enum x86_reg_mod { - mod_INDIRECT, - mod_DISP8, - mod_DISP32, - mod_REG -}; - -enum x86_reg_name { - reg_AX, - reg_CX, - reg_DX, - reg_BX, - reg_SP, - reg_BP, - reg_SI, - reg_DI -}; - - -enum x86_cc { - cc_O, /* overflow */ - cc_NO, /* not overflow */ - cc_NAE, /* not above or equal / carry */ - cc_AE, /* above or equal / not carry */ - cc_E, /* equal / zero */ - cc_NE /* not equal / not zero */ -}; - -enum sse_cc { - cc_Equal, - cc_LessThan, - cc_LessThanEqual, - cc_Unordered, - cc_NotEqual, - cc_NotLessThan, - cc_NotLessThanEqual, - cc_Ordered -}; - -#define cc_Z cc_E -#define cc_NZ cc_NE - -/* Begin/end/retreive function creation: - */ - - -void x86_init_func( struct x86_function *p ); -void x86_init_func_size( struct x86_function *p, unsigned code_size ); -void x86_release_func( struct x86_function *p ); -void (*x86_get_func( struct x86_function *p ))( void ); - - - -/* Create and manipulate registers and regmem values: - */ -struct x86_reg x86_make_reg( enum x86_reg_file file, - enum x86_reg_name idx ); - -struct x86_reg x86_make_disp( struct x86_reg reg, - int disp ); - -struct x86_reg x86_deref( struct x86_reg reg ); - -struct x86_reg x86_get_base_reg( struct x86_reg reg ); - - -/* Labels, jumps and fixup: - */ -unsigned char *x86_get_label( struct x86_function *p ); - -void x86_jcc( struct x86_function *p, - enum x86_cc cc, - unsigned char *label ); - -unsigned char *x86_jcc_forward( struct x86_function *p, - enum x86_cc cc ); - -unsigned char *x86_jmp_forward( struct x86_function *p); - -unsigned char *x86_call_forward( struct x86_function *p); - -void x86_fixup_fwd_jump( struct x86_function *p, - unsigned char *fixup ); - -void x86_jmp( struct x86_function *p, unsigned char *label ); - -/* void x86_call( struct x86_function *p, void (*label)() ); */ -void x86_call( struct x86_function *p, struct x86_reg reg); - -/* michal: - * Temporary. As I need immediate operands, and dont want to mess with the codegen, - * I load the immediate into general purpose register and use it. - */ -void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ); - - -/* Macro for sse_shufps() and sse2_pshufd(): - */ -#define SHUF(_x,_y,_z,_w) (((_x)<<0) | ((_y)<<2) | ((_z)<<4) | ((_w)<<6)) -#define SHUF_NOOP RSW(0,1,2,3) -#define GET_SHUF(swz, idx) (((swz) >> ((idx)*2)) & 0x3) - -void mmx_emms( struct x86_function *p ); -void mmx_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void mmx_movq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void mmx_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void mmx_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); - -void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_cvttps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_pshufd( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - unsigned char shuf ); -void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); - -void sse_addps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_addss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, - unsigned char cc ); -void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_movaps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_movhlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_movhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_movlhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_movlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_movss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_movups( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_mulps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_mulss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_orps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, - unsigned char shuf ); -void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); - -void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void x86_cmp( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void x86_dec( struct x86_function *p, struct x86_reg reg ); -void x86_inc( struct x86_function *p, struct x86_reg reg ); -void x86_lea( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void x86_mov( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void x86_mul( struct x86_function *p, struct x86_reg src ); -void x86_or( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void x86_pop( struct x86_function *p, struct x86_reg reg ); -void x86_push( struct x86_function *p, struct x86_reg reg ); -void x86_ret( struct x86_function *p ); -void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -void x86_sahf( struct x86_function *p ); - -void x87_f2xm1( struct x86_function *p ); -void x87_fabs( struct x86_function *p ); -void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); -void x87_faddp( struct x86_function *p, struct x86_reg dst ); -void x87_fchs( struct x86_function *p ); -void x87_fclex( struct x86_function *p ); -void x87_fcom( struct x86_function *p, struct x86_reg dst ); -void x87_fcomp( struct x86_function *p, struct x86_reg dst ); -void x87_fcos( struct x86_function *p ); -void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); -void x87_fdivp( struct x86_function *p, struct x86_reg dst ); -void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); -void x87_fdivrp( struct x86_function *p, struct x86_reg dst ); -void x87_fild( struct x86_function *p, struct x86_reg arg ); -void x87_fist( struct x86_function *p, struct x86_reg dst ); -void x87_fistp( struct x86_function *p, struct x86_reg dst ); -void x87_fld( struct x86_function *p, struct x86_reg arg ); -void x87_fld1( struct x86_function *p ); -void x87_fldcw( struct x86_function *p, struct x86_reg arg ); -void x87_fldl2e( struct x86_function *p ); -void x87_fldln2( struct x86_function *p ); -void x87_fldz( struct x86_function *p ); -void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); -void x87_fmulp( struct x86_function *p, struct x86_reg dst ); -void x87_fnclex( struct x86_function *p ); -void x87_fprndint( struct x86_function *p ); -void x87_fscale( struct x86_function *p ); -void x87_fsin( struct x86_function *p ); -void x87_fsincos( struct x86_function *p ); -void x87_fsqrt( struct x86_function *p ); -void x87_fst( struct x86_function *p, struct x86_reg dst ); -void x87_fstp( struct x86_function *p, struct x86_reg dst ); -void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); -void x87_fsubp( struct x86_function *p, struct x86_reg dst ); -void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); -void x87_fsubrp( struct x86_function *p, struct x86_reg dst ); -void x87_fxch( struct x86_function *p, struct x86_reg dst ); -void x87_fxtract( struct x86_function *p ); -void x87_fyl2x( struct x86_function *p ); -void x87_fyl2xp1( struct x86_function *p ); -void x87_fwait( struct x86_function *p ); -void x87_fnstsw( struct x86_function *p, struct x86_reg dst ); -void x87_fucompp( struct x86_function *p ); -void x87_fucomp( struct x86_function *p, struct x86_reg arg ); -void x87_fucom( struct x86_function *p, struct x86_reg arg ); - - - -/* Retreive a reference to one of the function arguments, taking into - * account any push/pop activity. Note - doesn't track explict - * manipulation of ESP by other instructions. - */ -struct x86_reg x86_fn_arg( struct x86_function *p, unsigned arg ); - -#endif -#endif -- cgit v1.2.3 From d2f6c9ab10656f6ecda131a6785a60565026d249 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 12:05:32 +0900 Subject: Add copyright headers to all rtasm source files. --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 23 +++++++++++++++++++++++ src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 28 +++++++++++++++++++++++++--- 2 files changed, 48 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 3c885a9fff..b332192a62 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -1,3 +1,26 @@ +/************************************************************************** + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #if defined(__i386__) || defined(__386__) #include "pipe/p_compiler.h" diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index c2aa416492..e4576001bf 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -1,6 +1,28 @@ - -#ifndef _X86SSE_H_ -#define _X86SSE_H_ +/************************************************************************** + * + * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _RTASM_X86SSE_H_ +#define _RTASM_X86SSE_H_ #if defined(__i386__) || defined(__386__) -- cgit v1.2.3 From 17158c2f00f5bee29ec8239367fd5498f22e4a91 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 12:24:42 +0900 Subject: Move mm.c code into util module. Using the u_ prefix to distingish the c source files that support gallium interfaces and those that have really no relation with gallium itself. --- src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c | 302 +----------------------- src/gallium/auxiliary/rtasm/Makefile | 5 +- src/gallium/auxiliary/rtasm/SConscript | 3 +- src/gallium/auxiliary/rtasm/mm.c | 283 ---------------------- src/gallium/auxiliary/rtasm/mm.h | 89 ------- src/gallium/auxiliary/rtasm/rtasm_execmem.c | 2 +- src/gallium/auxiliary/util/Makefile | 3 +- src/gallium/auxiliary/util/SConscript | 1 + src/gallium/auxiliary/util/u_mm.c | 283 ++++++++++++++++++++++ src/gallium/auxiliary/util/u_mm.h | 91 +++++++ 10 files changed, 382 insertions(+), 680 deletions(-) delete mode 100644 src/gallium/auxiliary/rtasm/mm.c delete mode 100644 src/gallium/auxiliary/rtasm/mm.h create mode 100644 src/gallium/auxiliary/util/u_mm.c create mode 100644 src/gallium/auxiliary/util/u_mm.h (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index 969aab51b5..983a105347 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -1,7 +1,6 @@ /************************************************************************** * * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * Copyright 1999 Wittawat Yamwong * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -40,6 +39,7 @@ #include "pipe/p_debug.h" #include "pipe/p_thread.h" #include "pipe/p_util.h" +#include "util/u_mm.h" #include "pb_buffer.h" #include "pb_bufmgr.h" @@ -50,306 +50,6 @@ #define SUPER(__derived) (&(__derived)->base) -struct mem_block -{ - struct mem_block *next, *prev; - struct mem_block *next_free, *prev_free; - struct mem_block *heap; - int ofs, size; - unsigned int free:1; - unsigned int reserved:1; -}; - - -#ifdef DEBUG -/** - * For debugging purposes. - */ -static void -mmDumpMemInfo(const struct mem_block *heap) -{ - debug_printf("Memory heap %p:\n", (void *)heap); - if (heap == 0) { - debug_printf(" heap == 0\n"); - } else { - const struct mem_block *p; - - for(p = heap->next; p != heap; p = p->next) { - debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); - } - - debug_printf("\nFree list:\n"); - - for(p = heap->next_free; p != heap; p = p->next_free) { - debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); - } - - } - debug_printf("End of memory blocks\n"); -} -#endif - - -/** - * input: total size in bytes - * return: a heap pointer if OK, NULL if error - */ -static struct mem_block * -mmInit(int ofs, int size) -{ - struct mem_block *heap, *block; - - if (size <= 0) - return NULL; - - heap = CALLOC_STRUCT(mem_block); - if (!heap) - return NULL; - - block = CALLOC_STRUCT(mem_block); - if (!block) { - FREE(heap); - return NULL; - } - - heap->next = block; - heap->prev = block; - heap->next_free = block; - heap->prev_free = block; - - block->heap = heap; - block->next = heap; - block->prev = heap; - block->next_free = heap; - block->prev_free = heap; - - block->ofs = ofs; - block->size = size; - block->free = 1; - - return heap; -} - - -static struct mem_block * -SliceBlock(struct mem_block *p, - int startofs, int size, - int reserved, int alignment) -{ - struct mem_block *newblock; - - /* break left [p, newblock, p->next], then p = newblock */ - if (startofs > p->ofs) { - newblock = CALLOC_STRUCT(mem_block); - if (!newblock) - return NULL; - newblock->ofs = startofs; - newblock->size = p->size - (startofs - p->ofs); - newblock->free = 1; - newblock->heap = p->heap; - - newblock->next = p->next; - newblock->prev = p; - p->next->prev = newblock; - p->next = newblock; - - newblock->next_free = p->next_free; - newblock->prev_free = p; - p->next_free->prev_free = newblock; - p->next_free = newblock; - - p->size -= newblock->size; - p = newblock; - } - - /* break right, also [p, newblock, p->next] */ - if (size < p->size) { - newblock = CALLOC_STRUCT(mem_block); - if (!newblock) - return NULL; - newblock->ofs = startofs + size; - newblock->size = p->size - size; - newblock->free = 1; - newblock->heap = p->heap; - - newblock->next = p->next; - newblock->prev = p; - p->next->prev = newblock; - p->next = newblock; - - newblock->next_free = p->next_free; - newblock->prev_free = p; - p->next_free->prev_free = newblock; - p->next_free = newblock; - - p->size = size; - } - - /* p = middle block */ - p->free = 0; - - /* Remove p from the free list: - */ - p->next_free->prev_free = p->prev_free; - p->prev_free->next_free = p->next_free; - - p->next_free = 0; - p->prev_free = 0; - - p->reserved = reserved; - return p; -} - - -/** - * Allocate 'size' bytes with 2^align2 bytes alignment, - * restrict the search to free memory after 'startSearch' - * depth and back buffers should be in different 4mb banks - * to get better page hits if possible - * input: size = size of block - * align2 = 2^align2 bytes alignment - * startSearch = linear offset from start of heap to begin search - * return: pointer to the allocated block, 0 if error - */ -static struct mem_block * -mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) -{ - struct mem_block *p; - const int mask = (1 << align2)-1; - int startofs = 0; - int endofs; - - if (!heap || align2 < 0 || size <= 0) - return NULL; - - for (p = heap->next_free; p != heap; p = p->next_free) { - assert(p->free); - - startofs = (p->ofs + mask) & ~mask; - if ( startofs < startSearch ) { - startofs = startSearch; - } - endofs = startofs+size; - if (endofs <= (p->ofs+p->size)) - break; - } - - if (p == heap) - return NULL; - - assert(p->free); - p = SliceBlock(p,startofs,size,0,mask+1); - - return p; -} - - -#if 0 -/** - * Free block starts at offset - * input: pointer to a heap, start offset - * return: pointer to a block - */ -static struct mem_block * -mmFindBlock(struct mem_block *heap, int start) -{ - struct mem_block *p; - - for (p = heap->next; p != heap; p = p->next) { - if (p->ofs == start) - return p; - } - - return NULL; -} -#endif - - -static INLINE int -Join2Blocks(struct mem_block *p) -{ - /* XXX there should be some assertions here */ - - /* NOTE: heap->free == 0 */ - - if (p->free && p->next->free) { - struct mem_block *q = p->next; - - assert(p->ofs + p->size == q->ofs); - p->size += q->size; - - p->next = q->next; - q->next->prev = p; - - q->next_free->prev_free = q->prev_free; - q->prev_free->next_free = q->next_free; - - FREE(q); - return 1; - } - return 0; -} - - -/** - * Free block starts at offset - * input: pointer to a block - * return: 0 if OK, -1 if error - */ -static int -mmFreeMem(struct mem_block *b) -{ - if (!b) - return 0; - - if (b->free) { - debug_printf("block already free\n"); - return -1; - } - if (b->reserved) { - debug_printf("block is reserved\n"); - return -1; - } - - b->free = 1; - b->next_free = b->heap->next_free; - b->prev_free = b->heap; - b->next_free->prev_free = b; - b->prev_free->next_free = b; - - Join2Blocks(b); - if (b->prev != b->heap) - Join2Blocks(b->prev); - - return 0; -} - - -/** - * destroy MM - */ -static void -mmDestroy(struct mem_block *heap) -{ - struct mem_block *p; - - if (!heap) - return; - - for (p = heap->next; p != heap; ) { - struct mem_block *next = p->next; - FREE(p); - p = next; - } - - FREE(heap); -} - - struct mm_pb_manager { struct pb_manager base; diff --git a/src/gallium/auxiliary/rtasm/Makefile b/src/gallium/auxiliary/rtasm/Makefile index 7c8ac60794..edfae2a204 100644 --- a/src/gallium/auxiliary/rtasm/Makefile +++ b/src/gallium/auxiliary/rtasm/Makefile @@ -5,9 +5,8 @@ include $(TOP)/configs/current LIBNAME = rtasm DRIVER_SOURCES = \ - execmem.c \ - x86sse.c \ - mm.c + rtasm_execmem.c \ + rtasm_x86sse.c C_SOURCES = \ $(DRIVER_SOURCES) diff --git a/src/gallium/auxiliary/rtasm/SConscript b/src/gallium/auxiliary/rtasm/SConscript index de8456e0ca..6eca1fe4c0 100644 --- a/src/gallium/auxiliary/rtasm/SConscript +++ b/src/gallium/auxiliary/rtasm/SConscript @@ -4,8 +4,7 @@ rtasm = env.ConvenienceLibrary( target = 'rtasm', source = [ 'rtasm_execmem.c', - 'rtasm_x86sse.c', - 'mm.c', + 'rtasm_x86sse.c' ]) auxiliaries.insert(0, rtasm) diff --git a/src/gallium/auxiliary/rtasm/mm.c b/src/gallium/auxiliary/rtasm/mm.c deleted file mode 100644 index 15f50491da..0000000000 --- a/src/gallium/auxiliary/rtasm/mm.c +++ /dev/null @@ -1,283 +0,0 @@ -/* - * GLX Hardware Device Driver common code - * Copyright (C) 1999 Wittawat Yamwong - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * WITTAWAT YAMWONG, OR ANY OTHER CONTRIBUTORS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE - * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - */ - - -#include "pipe/p_compiler.h" -#include "pipe/p_util.h" -#include "pipe/p_debug.h" - -#include "mm.h" - - -void -mmDumpMemInfo(const struct mem_block *heap) -{ - debug_printf("Memory heap %p:\n", (void *)heap); - if (heap == 0) { - debug_printf(" heap == 0\n"); - } else { - const struct mem_block *p; - - for(p = heap->next; p != heap; p = p->next) { - debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); - } - - debug_printf("\nFree list:\n"); - - for(p = heap->next_free; p != heap; p = p->next_free) { - debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, - p->free ? 'F':'.', - p->reserved ? 'R':'.'); - } - - } - debug_printf("End of memory blocks\n"); -} - -struct mem_block * -mmInit(int ofs, int size) -{ - struct mem_block *heap, *block; - - if (size <= 0) - return NULL; - - heap = CALLOC_STRUCT(mem_block); - if (!heap) - return NULL; - - block = CALLOC_STRUCT(mem_block); - if (!block) { - FREE(heap); - return NULL; - } - - heap->next = block; - heap->prev = block; - heap->next_free = block; - heap->prev_free = block; - - block->heap = heap; - block->next = heap; - block->prev = heap; - block->next_free = heap; - block->prev_free = heap; - - block->ofs = ofs; - block->size = size; - block->free = 1; - - return heap; -} - - -static struct mem_block * -SliceBlock(struct mem_block *p, - int startofs, int size, - int reserved, int alignment) -{ - struct mem_block *newblock; - - /* break left [p, newblock, p->next], then p = newblock */ - if (startofs > p->ofs) { - newblock = CALLOC_STRUCT(mem_block); - if (!newblock) - return NULL; - newblock->ofs = startofs; - newblock->size = p->size - (startofs - p->ofs); - newblock->free = 1; - newblock->heap = p->heap; - - newblock->next = p->next; - newblock->prev = p; - p->next->prev = newblock; - p->next = newblock; - - newblock->next_free = p->next_free; - newblock->prev_free = p; - p->next_free->prev_free = newblock; - p->next_free = newblock; - - p->size -= newblock->size; - p = newblock; - } - - /* break right, also [p, newblock, p->next] */ - if (size < p->size) { - newblock = CALLOC_STRUCT(mem_block); - if (!newblock) - return NULL; - newblock->ofs = startofs + size; - newblock->size = p->size - size; - newblock->free = 1; - newblock->heap = p->heap; - - newblock->next = p->next; - newblock->prev = p; - p->next->prev = newblock; - p->next = newblock; - - newblock->next_free = p->next_free; - newblock->prev_free = p; - p->next_free->prev_free = newblock; - p->next_free = newblock; - - p->size = size; - } - - /* p = middle block */ - p->free = 0; - - /* Remove p from the free list: - */ - p->next_free->prev_free = p->prev_free; - p->prev_free->next_free = p->next_free; - - p->next_free = 0; - p->prev_free = 0; - - p->reserved = reserved; - return p; -} - - -struct mem_block * -mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) -{ - struct mem_block *p; - const int mask = (1 << align2)-1; - int startofs = 0; - int endofs; - - if (!heap || align2 < 0 || size <= 0) - return NULL; - - for (p = heap->next_free; p != heap; p = p->next_free) { - assert(p->free); - - startofs = (p->ofs + mask) & ~mask; - if ( startofs < startSearch ) { - startofs = startSearch; - } - endofs = startofs+size; - if (endofs <= (p->ofs+p->size)) - break; - } - - if (p == heap) - return NULL; - - assert(p->free); - p = SliceBlock(p,startofs,size,0,mask+1); - - return p; -} - - -struct mem_block * -mmFindBlock(struct mem_block *heap, int start) -{ - struct mem_block *p; - - for (p = heap->next; p != heap; p = p->next) { - if (p->ofs == start) - return p; - } - - return NULL; -} - - -static INLINE int -Join2Blocks(struct mem_block *p) -{ - /* XXX there should be some assertions here */ - - /* NOTE: heap->free == 0 */ - - if (p->free && p->next->free) { - struct mem_block *q = p->next; - - assert(p->ofs + p->size == q->ofs); - p->size += q->size; - - p->next = q->next; - q->next->prev = p; - - q->next_free->prev_free = q->prev_free; - q->prev_free->next_free = q->next_free; - - FREE(q); - return 1; - } - return 0; -} - -int -mmFreeMem(struct mem_block *b) -{ - if (!b) - return 0; - - if (b->free) { - debug_printf("block already free\n"); - return -1; - } - if (b->reserved) { - debug_printf("block is reserved\n"); - return -1; - } - - b->free = 1; - b->next_free = b->heap->next_free; - b->prev_free = b->heap; - b->next_free->prev_free = b; - b->prev_free->next_free = b; - - Join2Blocks(b); - if (b->prev != b->heap) - Join2Blocks(b->prev); - - return 0; -} - - -void -mmDestroy(struct mem_block *heap) -{ - struct mem_block *p; - - if (!heap) - return; - - for (p = heap->next; p != heap; ) { - struct mem_block *next = p->next; - FREE(p); - p = next; - } - - FREE(heap); -} diff --git a/src/gallium/auxiliary/rtasm/mm.h b/src/gallium/auxiliary/rtasm/mm.h deleted file mode 100644 index f469b18d3e..0000000000 --- a/src/gallium/auxiliary/rtasm/mm.h +++ /dev/null @@ -1,89 +0,0 @@ -/* - * GLX Hardware Device Driver common code - * Copyright (C) 1999 Wittawat Yamwong - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * KEITH WHITWELL, OR ANY OTHER CONTRIBUTORS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE - * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/** - * Memory manager code. Primarily used by device drivers to manage texture - * heaps, etc. - */ - - -#ifndef MM_H -#define MM_H - - -struct mem_block { - struct mem_block *next, *prev; - struct mem_block *next_free, *prev_free; - struct mem_block *heap; - int ofs,size; - unsigned int free:1; - unsigned int reserved:1; -}; - - - -/** - * input: total size in bytes - * return: a heap pointer if OK, NULL if error - */ -extern struct mem_block *mmInit(int ofs, int size); - -/** - * Allocate 'size' bytes with 2^align2 bytes alignment, - * restrict the search to free memory after 'startSearch' - * depth and back buffers should be in different 4mb banks - * to get better page hits if possible - * input: size = size of block - * align2 = 2^align2 bytes alignment - * startSearch = linear offset from start of heap to begin search - * return: pointer to the allocated block, 0 if error - */ -extern struct mem_block *mmAllocMem(struct mem_block *heap, int size, int align2, - int startSearch); - -/** - * Free block starts at offset - * input: pointer to a block - * return: 0 if OK, -1 if error - */ -extern int mmFreeMem(struct mem_block *b); - -/** - * Free block starts at offset - * input: pointer to a heap, start offset - * return: pointer to a block - */ -extern struct mem_block *mmFindBlock(struct mem_block *heap, int start); - -/** - * destroy MM - */ -extern void mmDestroy(struct mem_block *mmInit); - -/** - * For debuging purpose. - */ -extern void mmDumpMemInfo(const struct mem_block *mmInit); - -#endif diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.c b/src/gallium/auxiliary/rtasm/rtasm_execmem.c index cb13db2498..9c78fa5626 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_execmem.c +++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.c @@ -46,7 +46,7 @@ #include #include -#include "mm.h" +#include "util/u_mm.h" #define EXEC_HEAP_SIZE (10*1024*1024) diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index b8cb148c4f..7cc2aa44f9 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -7,7 +7,8 @@ LIBNAME = util DRIVER_SOURCES = \ p_debug.c \ p_tile.c \ - p_util.c + p_util.c \ + u_mm.c C_SOURCES = \ $(DRIVER_SOURCES) diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index b126cf44d6..4717941434 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -6,6 +6,7 @@ util = env.ConvenienceLibrary( 'p_debug.c', 'p_tile.c', 'p_util.c', + 'u_mm.c', ]) auxiliaries.insert(0, util) diff --git a/src/gallium/auxiliary/util/u_mm.c b/src/gallium/auxiliary/util/u_mm.c new file mode 100644 index 0000000000..b49ae074e0 --- /dev/null +++ b/src/gallium/auxiliary/util/u_mm.c @@ -0,0 +1,283 @@ +/************************************************************************** + * + * Copyright (C) 1999 Wittawat Yamwong + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * WITTAWAT YAMWONG, OR ANY OTHER CONTRIBUTORS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE + * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_compiler.h" +#include "pipe/p_util.h" +#include "pipe/p_debug.h" + +#include "util/u_mm.h" + + +void +mmDumpMemInfo(const struct mem_block *heap) +{ + debug_printf("Memory heap %p:\n", (void *)heap); + if (heap == 0) { + debug_printf(" heap == 0\n"); + } else { + const struct mem_block *p; + + for(p = heap->next; p != heap; p = p->next) { + debug_printf(" Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + debug_printf("\nFree list:\n"); + + for(p = heap->next_free; p != heap; p = p->next_free) { + debug_printf(" FREE Offset:%08x, Size:%08x, %c%c\n",p->ofs,p->size, + p->free ? 'F':'.', + p->reserved ? 'R':'.'); + } + + } + debug_printf("End of memory blocks\n"); +} + +struct mem_block * +mmInit(int ofs, int size) +{ + struct mem_block *heap, *block; + + if (size <= 0) + return NULL; + + heap = CALLOC_STRUCT(mem_block); + if (!heap) + return NULL; + + block = CALLOC_STRUCT(mem_block); + if (!block) { + FREE(heap); + return NULL; + } + + heap->next = block; + heap->prev = block; + heap->next_free = block; + heap->prev_free = block; + + block->heap = heap; + block->next = heap; + block->prev = heap; + block->next_free = heap; + block->prev_free = heap; + + block->ofs = ofs; + block->size = size; + block->free = 1; + + return heap; +} + + +static struct mem_block * +SliceBlock(struct mem_block *p, + int startofs, int size, + int reserved, int alignment) +{ + struct mem_block *newblock; + + /* break left [p, newblock, p->next], then p = newblock */ + if (startofs > p->ofs) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs; + newblock->size = p->size - (startofs - p->ofs); + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size -= newblock->size; + p = newblock; + } + + /* break right, also [p, newblock, p->next] */ + if (size < p->size) { + newblock = CALLOC_STRUCT(mem_block); + if (!newblock) + return NULL; + newblock->ofs = startofs + size; + newblock->size = p->size - size; + newblock->free = 1; + newblock->heap = p->heap; + + newblock->next = p->next; + newblock->prev = p; + p->next->prev = newblock; + p->next = newblock; + + newblock->next_free = p->next_free; + newblock->prev_free = p; + p->next_free->prev_free = newblock; + p->next_free = newblock; + + p->size = size; + } + + /* p = middle block */ + p->free = 0; + + /* Remove p from the free list: + */ + p->next_free->prev_free = p->prev_free; + p->prev_free->next_free = p->next_free; + + p->next_free = 0; + p->prev_free = 0; + + p->reserved = reserved; + return p; +} + + +struct mem_block * +mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) +{ + struct mem_block *p; + const int mask = (1 << align2)-1; + int startofs = 0; + int endofs; + + if (!heap || align2 < 0 || size <= 0) + return NULL; + + for (p = heap->next_free; p != heap; p = p->next_free) { + assert(p->free); + + startofs = (p->ofs + mask) & ~mask; + if ( startofs < startSearch ) { + startofs = startSearch; + } + endofs = startofs+size; + if (endofs <= (p->ofs+p->size)) + break; + } + + if (p == heap) + return NULL; + + assert(p->free); + p = SliceBlock(p,startofs,size,0,mask+1); + + return p; +} + + +struct mem_block * +mmFindBlock(struct mem_block *heap, int start) +{ + struct mem_block *p; + + for (p = heap->next; p != heap; p = p->next) { + if (p->ofs == start) + return p; + } + + return NULL; +} + + +static INLINE int +Join2Blocks(struct mem_block *p) +{ + /* XXX there should be some assertions here */ + + /* NOTE: heap->free == 0 */ + + if (p->free && p->next->free) { + struct mem_block *q = p->next; + + assert(p->ofs + p->size == q->ofs); + p->size += q->size; + + p->next = q->next; + q->next->prev = p; + + q->next_free->prev_free = q->prev_free; + q->prev_free->next_free = q->next_free; + + FREE(q); + return 1; + } + return 0; +} + +int +mmFreeMem(struct mem_block *b) +{ + if (!b) + return 0; + + if (b->free) { + debug_printf("block already free\n"); + return -1; + } + if (b->reserved) { + debug_printf("block is reserved\n"); + return -1; + } + + b->free = 1; + b->next_free = b->heap->next_free; + b->prev_free = b->heap; + b->next_free->prev_free = b; + b->prev_free->next_free = b; + + Join2Blocks(b); + if (b->prev != b->heap) + Join2Blocks(b->prev); + + return 0; +} + + +void +mmDestroy(struct mem_block *heap) +{ + struct mem_block *p; + + if (!heap) + return; + + for (p = heap->next; p != heap; ) { + struct mem_block *next = p->next; + FREE(p); + p = next; + } + + FREE(heap); +} diff --git a/src/gallium/auxiliary/util/u_mm.h b/src/gallium/auxiliary/util/u_mm.h new file mode 100644 index 0000000000..b226b101cb --- /dev/null +++ b/src/gallium/auxiliary/util/u_mm.h @@ -0,0 +1,91 @@ +/************************************************************************** + * + * Copyright (C) 1999 Wittawat Yamwong + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * KEITH WHITWELL, OR ANY OTHER CONTRIBUTORS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE + * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * @file + * Memory manager code. Primarily used by device drivers to manage texture + * heaps, etc. + */ + + +#ifndef _U_MM_H_ +#define _U_MM_H_ + + +struct mem_block { + struct mem_block *next, *prev; + struct mem_block *next_free, *prev_free; + struct mem_block *heap; + int ofs,size; + unsigned int free:1; + unsigned int reserved:1; +}; + + + +/** + * input: total size in bytes + * return: a heap pointer if OK, NULL if error + */ +extern struct mem_block *mmInit(int ofs, int size); + +/** + * Allocate 'size' bytes with 2^align2 bytes alignment, + * restrict the search to free memory after 'startSearch' + * depth and back buffers should be in different 4mb banks + * to get better page hits if possible + * input: size = size of block + * align2 = 2^align2 bytes alignment + * startSearch = linear offset from start of heap to begin search + * return: pointer to the allocated block, 0 if error + */ +extern struct mem_block *mmAllocMem(struct mem_block *heap, int size, int align2, + int startSearch); + +/** + * Free block starts at offset + * input: pointer to a block + * return: 0 if OK, -1 if error + */ +extern int mmFreeMem(struct mem_block *b); + +/** + * Free block starts at offset + * input: pointer to a heap, start offset + * return: pointer to a block + */ +extern struct mem_block *mmFindBlock(struct mem_block *heap, int start); + +/** + * destroy MM + */ +extern void mmDestroy(struct mem_block *mmInit); + +/** + * For debuging purpose. + */ +extern void mmDumpMemInfo(const struct mem_block *mmInit); + +#endif -- cgit v1.2.3 From f430d95a36d55141cd9ef911aab70364ce4a4108 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 12:52:28 +0900 Subject: Use gallium's rtasm module. --- src/gallium/auxiliary/draw/draw_private.h | 2 +- src/gallium/auxiliary/draw/draw_vf.c | 10 +++------- src/gallium/auxiliary/draw/draw_vf_sse.c | 2 +- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2 +- src/gallium/drivers/softpipe/sp_fs_sse.c | 2 +- src/mesa/state_tracker/st_program.h | 1 - 7 files changed, 8 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index dd75f9aefc..6c7e860861 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -44,7 +44,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "x86/rtasm/x86sse.h" +#include "rtasm/rtasm_x86sse.h" #include "tgsi/exec/tgsi_exec.h" diff --git a/src/gallium/auxiliary/draw/draw_vf.c b/src/gallium/auxiliary/draw/draw_vf.c index dc3a5ecd21..901ff20a7e 100644 --- a/src/gallium/auxiliary/draw/draw_vf.c +++ b/src/gallium/auxiliary/draw/draw_vf.c @@ -30,6 +30,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_util.h" +#include "rtasm/rtasm_execmem.h" #include "draw_vf.h" @@ -37,11 +38,6 @@ #define DRAW_VF_DBG 0 -/* TODO: remove this */ -extern void -_mesa_exec_free( void *addr ); - - static boolean match_fastpath( struct draw_vertex_fetch *vf, const struct draw_vf_fastpath *fp) { @@ -414,12 +410,12 @@ void draw_vf_destroy( struct draw_vertex_fetch *vf ) FREE(fp->attr); /* KW: At the moment, fp->func is constrained to be allocated by - * _mesa_exec_alloc(), as the hardwired fastpaths in + * rtasm_exec_alloc(), as the hardwired fastpaths in * t_vertex_generic.c are handled specially. It would be nice * to unify them, but this probably won't change until this * module gets another overhaul. */ - //_mesa_exec_free((void *) fp->func); + //rtasm_exec_free((void *) fp->func); FREE(fp); } diff --git a/src/gallium/auxiliary/draw/draw_vf_sse.c b/src/gallium/auxiliary/draw/draw_vf_sse.c index 1ad2ae756d..1e889deeea 100644 --- a/src/gallium/auxiliary/draw/draw_vf_sse.c +++ b/src/gallium/auxiliary/draw/draw_vf_sse.c @@ -35,7 +35,7 @@ #if defined(USE_SSE_ASM) -#include "x86/rtasm/x86sse.h" +#include "rtasm/rtasm_x86sse.h" #include "x86/common_x86_asm.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 27bc66812c..11ef0c503d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -41,7 +41,7 @@ #include "draw_private.h" #include "draw_context.h" -#include "x86/rtasm/x86sse.h" +#include "rtasm/rtasm_x86sse.h" #include "tgsi/exec/tgsi_sse2.h" diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 62c6a69c63..29a7f842ed 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -32,7 +32,7 @@ #include "tgsi_exec.h" #include "tgsi_sse2.h" -#include "x86/rtasm/x86sse.h" +#include "rtasm/rtasm_x86sse.h" #if defined(__i386__) || defined(__386__) diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index d90066e025..b18772f4e6 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -42,7 +42,7 @@ #if defined(__i386__) || defined(__386__) -#include "x86/rtasm/x86sse.h" +#include "rtasm/rtasm_x86sse.h" /* Surely this should be defined somewhere in a tgsi header: */ diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index ea1dde4a7a..25cf3e94a8 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -36,7 +36,6 @@ #include "mtypes.h" #include "pipe/p_shader_tokens.h" -#include "x86/rtasm/x86sse.h" #define ST_MAX_SHADER_TOKENS 1024 -- cgit v1.2.3 From 90b2beb661f630966788a6e909dc759c99e38973 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 13:27:13 +0900 Subject: Simplify makefile boilerplate code. Don't define ASM_SOURCES variable globally -- reserve that variable to be defined locally by makefiles, together with C_SOURCES and CPP_SOURCES. --- src/gallium/auxiliary/cso_cache/Makefile | 7 +------ src/gallium/auxiliary/draw/Makefile | 7 +------ src/gallium/auxiliary/pipebuffer/Makefile | 8 +------- src/gallium/auxiliary/rtasm/Makefile | 8 +------- src/gallium/auxiliary/tgsi/Makefile | 8 +------- src/gallium/auxiliary/util/Makefile | 8 +------- src/gallium/drivers/failover/Makefile | 9 +-------- src/gallium/drivers/i915simple/Makefile | 9 +-------- src/gallium/drivers/i965simple/Makefile | 25 +++++++------------------ src/gallium/drivers/softpipe/Makefile | 9 +-------- src/gallium/winsys/dri/Makefile.template | 5 +++-- src/glx/x11/Makefile | 6 +++--- src/mesa/sources | 6 +++--- 13 files changed, 25 insertions(+), 90 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/Makefile b/src/gallium/auxiliary/cso_cache/Makefile index 8248b097fd..3e49266163 100644 --- a/src/gallium/auxiliary/cso_cache/Makefile +++ b/src/gallium/auxiliary/cso_cache/Makefile @@ -3,15 +3,10 @@ include $(TOP)/configs/current LIBNAME = cso_cache -DRIVER_SOURCES = \ +C_SOURCES = \ cso_cache.c \ cso_hash.c -C_SOURCES = \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template symlinks: diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index c8000cbe9c..1ee9eca0ca 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -3,7 +3,7 @@ include $(TOP)/configs/current LIBNAME = draw -DRIVER_SOURCES = \ +C_SOURCES = \ draw_aaline.c \ draw_clip.c \ draw_vs_exec.c \ @@ -29,11 +29,6 @@ DRIVER_SOURCES = \ draw_vf_sse.c \ draw_wide_prims.c -C_SOURCES = \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template symlinks: diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile index 588629e870..a9fa518c67 100644 --- a/src/gallium/auxiliary/pipebuffer/Makefile +++ b/src/gallium/auxiliary/pipebuffer/Makefile @@ -1,10 +1,9 @@ - TOP = ../../../.. include $(TOP)/configs/current LIBNAME = pipebuffer -DRIVER_SOURCES = \ +C_SOURCES = \ pb_buffer_fenced.c \ pb_buffer_malloc.c \ pb_bufmgr_fenced.c \ @@ -12,11 +11,6 @@ DRIVER_SOURCES = \ pb_bufmgr_pool.c \ pb_winsys.c -C_SOURCES = \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template symlinks: diff --git a/src/gallium/auxiliary/rtasm/Makefile b/src/gallium/auxiliary/rtasm/Makefile index edfae2a204..bc339d2aa6 100644 --- a/src/gallium/auxiliary/rtasm/Makefile +++ b/src/gallium/auxiliary/rtasm/Makefile @@ -1,18 +1,12 @@ - TOP = ../../../.. include $(TOP)/configs/current LIBNAME = rtasm -DRIVER_SOURCES = \ +C_SOURCES = \ rtasm_execmem.c \ rtasm_x86sse.c -C_SOURCES = \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template symlinks: diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index 8bb62b2a0a..71f64b747c 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -1,10 +1,9 @@ - TOP = ../../../.. include $(TOP)/configs/current LIBNAME = tgsi -DRIVER_SOURCES = \ +C_SOURCES = \ exec/tgsi_exec.c \ exec/tgsi_sse2.c \ util/tgsi_build.c \ @@ -13,11 +12,6 @@ DRIVER_SOURCES = \ util/tgsi_transform.c \ util/tgsi_util.c -C_SOURCES = \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template symlinks: diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 7cc2aa44f9..906a46d6b4 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -1,20 +1,14 @@ - TOP = ../../../.. include $(TOP)/configs/current LIBNAME = util -DRIVER_SOURCES = \ +C_SOURCES = \ p_debug.c \ p_tile.c \ p_util.c \ u_mm.c -C_SOURCES = \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template symlinks: diff --git a/src/gallium/drivers/failover/Makefile b/src/gallium/drivers/failover/Makefile index 14389bd055..f08b8df07a 100644 --- a/src/gallium/drivers/failover/Makefile +++ b/src/gallium/drivers/failover/Makefile @@ -1,20 +1,13 @@ - TOP = ../../../.. include $(TOP)/configs/current LIBNAME = failover -DRIVER_SOURCES = \ +C_SOURCES = \ fo_state.c \ fo_state_emit.c \ fo_context.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template symlinks: diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile index ee22ba86f9..2a75f5d57c 100644 --- a/src/gallium/drivers/i915simple/Makefile +++ b/src/gallium/drivers/i915simple/Makefile @@ -1,10 +1,9 @@ - TOP = ../../../.. include $(TOP)/configs/current LIBNAME = i915simple -DRIVER_SOURCES = \ +C_SOURCES = \ i915_blit.c \ i915_clear.c \ i915_flush.c \ @@ -26,12 +25,6 @@ DRIVER_SOURCES = \ i915_fpc_translate.c \ i915_surface.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template symlinks: diff --git a/src/gallium/drivers/i965simple/Makefile b/src/gallium/drivers/i965simple/Makefile index 1dec1f9749..cc8580836c 100644 --- a/src/gallium/drivers/i965simple/Makefile +++ b/src/gallium/drivers/i965simple/Makefile @@ -1,14 +1,13 @@ - TOP = ../../../.. include $(TOP)/configs/current LIBNAME = i965simple -DRIVER_SOURCES = \ - brw_blit.c \ - brw_flush.c \ - brw_strings.c \ - brw_surface.c \ +C_SOURCES = \ + brw_blit.c \ + brw_flush.c \ + brw_strings.c \ + brw_surface.c \ brw_cc.c \ brw_clip.c \ brw_clip_line.c \ @@ -31,8 +30,8 @@ DRIVER_SOURCES = \ brw_sf.c \ brw_sf_emit.c \ brw_sf_state.c \ - brw_shader_info.c \ - brw_state.c \ + brw_shader_info.c \ + brw_state.c \ brw_state_batch.c \ brw_state_cache.c \ brw_state_pool.c \ @@ -51,16 +50,6 @@ DRIVER_SOURCES = \ brw_wm_state.c \ brw_wm_surface_state.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(COMMON_BM_SOURCES) \ - $(MINIGLX_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -DRIVER_DEFINES = -I. - include ../../Makefile.template symlinks: diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile index 5479daf8ea..539ffb77f5 100644 --- a/src/gallium/drivers/softpipe/Makefile +++ b/src/gallium/drivers/softpipe/Makefile @@ -1,10 +1,9 @@ - TOP = ../../../.. include $(TOP)/configs/current LIBNAME = softpipe -DRIVER_SOURCES = \ +C_SOURCES = \ sp_fs_exec.c \ sp_fs_sse.c \ sp_fs_llvm.c \ @@ -41,12 +40,6 @@ DRIVER_SOURCES = \ sp_tile_cache.c \ sp_surface.c -C_SOURCES = \ - $(COMMON_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - include ../../Makefile.template symlinks: diff --git a/src/gallium/winsys/dri/Makefile.template b/src/gallium/winsys/dri/Makefile.template index 65a93bd53e..3bc1fdd4d4 100644 --- a/src/gallium/winsys/dri/Makefile.template +++ b/src/gallium/winsys/dri/Makefile.template @@ -25,8 +25,9 @@ WINOBJ= WINLIB= INCLUDES = $(SHARED_INCLUDES) $(EXPAT_INCLUDES) -OBJECTS = $(C_SOURCES:.c=.o) \ - $(ASM_SOURCES:.S=.o) +OBJECTS = \ + $(C_SOURCES:.c=.o) \ + $(ASM_SOURCES:.S=.o) else # miniglx diff --git a/src/glx/x11/Makefile b/src/glx/x11/Makefile index 5f74fcff06..b404727f08 100644 --- a/src/glx/x11/Makefile +++ b/src/glx/x11/Makefile @@ -35,7 +35,7 @@ SOURCES = \ include $(TOP)/src/mesa/sources -MESA_ASM_API = $(addprefix $(TOP)/src/mesa/, $(ASM_API)) +MESA_GLAPI_ASM_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_ASM_SOURCES)) MESA_GLAPI_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_SOURCES)) MESA_GLAPI_OBJECTS = $(addprefix $(TOP)/src/mesa/, $(GLAPI_OBJECTS)) @@ -70,11 +70,11 @@ $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(OBJECTS) Makefile -install $(TOP)/$(LIB_DIR) $(GL_LIB_DEPS) $(OBJECTS) -depend: $(SOURCES) $(MESA_GLAPI_SOURCES) $(MESA_ASM_API) Makefile +depend: $(SOURCES) $(MESA_GLAPI_SOURCES) $(MESA_GLAPI_ASM_SOURCES) Makefile rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) \ - $(MESA_GLAPI_SOURCES) $(MESA_ASM_API) + $(MESA_GLAPI_SOURCES) $(MESA_GLAPI_ASM_SOURCES) # Emacs tags diff --git a/src/mesa/sources b/src/mesa/sources index 0d185fd5f3..9e56694893 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -320,7 +320,7 @@ FBDEV_DRIVER_SOURCES = \ ALL_SOURCES = \ $(GLAPI_SOURCES) \ $(SOLO_SOURCES) \ - $(ASM_SOURCES) \ + $(MESA_ASM_SOURCES) \ $(COMMON_DRIVER_SOURCES)\ $(X11_DRIVER_SOURCES) \ $(FBDEV_DRIVER_SOURCES) \ @@ -353,11 +353,11 @@ CORE_SOURCES = \ SOLO_OBJECTS = \ $(SOLO_SOURCES:.c=.o) \ - $(ASM_SOURCES:.S=.o) + $(MESA_ASM_SOURCES:.S=.o) GLAPI_OBJECTS = \ $(GLAPI_SOURCES:.c=.o) \ - $(ASM_API:.S=.o) + $(GLAPI_ASM_SOURCES:.S=.o) CORE_OBJECTS = $(SOLO_OBJECTS) $(GLAPI_OBJECTS) -- cgit v1.2.3 From b0eef0dc2557febea7d425fee1f9c2da382898a6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 13:41:38 +0900 Subject: Add run-time cpu capabilities detection stubs. --- src/gallium/auxiliary/draw/draw_vf_sse.c | 6 ++-- src/gallium/auxiliary/rtasm/Makefile | 1 + src/gallium/auxiliary/rtasm/SConscript | 1 + src/gallium/auxiliary/rtasm/rtasm_cpu.c | 50 ++++++++++++++++++++++++++++++++ src/gallium/auxiliary/rtasm/rtasm_cpu.h | 42 +++++++++++++++++++++++++++ 5 files changed, 97 insertions(+), 3 deletions(-) create mode 100644 src/gallium/auxiliary/rtasm/rtasm_cpu.c create mode 100644 src/gallium/auxiliary/rtasm/rtasm_cpu.h (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vf_sse.c b/src/gallium/auxiliary/draw/draw_vf_sse.c index 1e889deeea..6076f9849d 100644 --- a/src/gallium/auxiliary/draw/draw_vf_sse.c +++ b/src/gallium/auxiliary/draw/draw_vf_sse.c @@ -35,8 +35,8 @@ #if defined(USE_SSE_ASM) +#include "rtasm/rtasm_cpu.h" #include "rtasm/rtasm_x86sse.h" -#include "x86/common_x86_asm.h" #define X 0 @@ -576,7 +576,7 @@ void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) { struct x86_program p; - if (!cpu_has_xmm) { + if (!rtasm_cpu_has_sse()) { vf->codegen_emit = NULL; return; } @@ -586,7 +586,7 @@ void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) p.vf = vf; p.inputs_safe = 0; /* for now */ p.outputs_safe = 1; /* for now */ - p.have_sse2 = cpu_has_xmm2; + p.have_sse2 = rtasm_cpu_has_sse2(); p.identity = x86_make_reg(file_XMM, 6); p.chan0 = x86_make_reg(file_XMM, 7); diff --git a/src/gallium/auxiliary/rtasm/Makefile b/src/gallium/auxiliary/rtasm/Makefile index bc339d2aa6..9b972f8f86 100644 --- a/src/gallium/auxiliary/rtasm/Makefile +++ b/src/gallium/auxiliary/rtasm/Makefile @@ -4,6 +4,7 @@ include $(TOP)/configs/current LIBNAME = rtasm C_SOURCES = \ + rtasm_cpu.c \ rtasm_execmem.c \ rtasm_x86sse.c diff --git a/src/gallium/auxiliary/rtasm/SConscript b/src/gallium/auxiliary/rtasm/SConscript index 6eca1fe4c0..ac41a4f212 100644 --- a/src/gallium/auxiliary/rtasm/SConscript +++ b/src/gallium/auxiliary/rtasm/SConscript @@ -3,6 +3,7 @@ Import('*') rtasm = env.ConvenienceLibrary( target = 'rtasm', source = [ + 'rtasm_cpu.c', 'rtasm_execmem.c', 'rtasm_x86sse.c' ]) diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.c b/src/gallium/auxiliary/rtasm/rtasm_cpu.c new file mode 100644 index 0000000000..eb3359750b --- /dev/null +++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.c @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "rtasm_cpu.h" + + +int rtasm_cpu_has_sse(void) +{ + /* FIXME: actually detect this at run-time */ +#if defined(__i386__) || defined(__386__) + return 1; +#else + return 0; +#endif +} + +int rtasm_cpu_has_sse2(void) +{ + /* FIXME: actually detect this at run-time */ +#if defined(__i386__) || defined(__386__) + return 1; +#else + return 0; +#endif +} diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.h b/src/gallium/auxiliary/rtasm/rtasm_cpu.h new file mode 100644 index 0000000000..ebc71634fd --- /dev/null +++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.h @@ -0,0 +1,42 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Runtime detection of CPU capabilities. + */ + +#ifndef _RTASM_CPU_H_ +#define _RTASM_CPU_H_ + + +int rtasm_cpu_has_sse(void); + +int rtasm_cpu_has_sse2(void); + + +#endif /* _RTASM_CPU_H_ */ -- cgit v1.2.3 From 5d78212d752e021555356bbb9cc5993ad6d9e847 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 14:00:16 +0900 Subject: Bring in ppc spe rtasm into gallium's rtasm module. Moving files since these are not being used outside gallium. --- src/gallium/auxiliary/rtasm/Makefile | 3 +- src/gallium/auxiliary/rtasm/SConscript | 3 +- src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c | 386 +++++++++++++++++++++++ src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h | 314 ++++++++++++++++++ src/gallium/drivers/cell/ppu/cell_context.h | 2 +- src/gallium/drivers/cell/ppu/cell_vertex_fetch.c | 2 +- src/mesa/ppc/rtasm/spe_asm.c | 385 ---------------------- src/mesa/ppc/rtasm/spe_asm.h | 314 ------------------ src/mesa/sources | 1 - 9 files changed, 706 insertions(+), 704 deletions(-) create mode 100644 src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c create mode 100644 src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h delete mode 100644 src/mesa/ppc/rtasm/spe_asm.c delete mode 100644 src/mesa/ppc/rtasm/spe_asm.h (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/Makefile b/src/gallium/auxiliary/rtasm/Makefile index 9b972f8f86..39b8a4dbd7 100644 --- a/src/gallium/auxiliary/rtasm/Makefile +++ b/src/gallium/auxiliary/rtasm/Makefile @@ -6,7 +6,8 @@ LIBNAME = rtasm C_SOURCES = \ rtasm_cpu.c \ rtasm_execmem.c \ - rtasm_x86sse.c + rtasm_x86sse.c \ + rtasm_ppc_spe.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/rtasm/SConscript b/src/gallium/auxiliary/rtasm/SConscript index ac41a4f212..8ea25922aa 100644 --- a/src/gallium/auxiliary/rtasm/SConscript +++ b/src/gallium/auxiliary/rtasm/SConscript @@ -5,7 +5,8 @@ rtasm = env.ConvenienceLibrary( source = [ 'rtasm_cpu.c', 'rtasm_execmem.c', - 'rtasm_x86sse.c' + 'rtasm_x86sse.c', + 'rtasm_ppc_spe.c', ]) auxiliaries.insert(0, rtasm) diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c new file mode 100644 index 0000000000..95a2d6fcbb --- /dev/null +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c @@ -0,0 +1,386 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file + * Real-time assembly generation interface for Cell B.E. SPEs. + * + * \author Ian Romanick + */ + +#include "pipe/p_compiler.h" +#include "pipe/p_util.h" +#include "rtasm_ppc_spe.h" + +#ifdef GALLIUM_CELL +/** + * SPE instruction types + * + * There are 6 primary instruction encodings used on the Cell's SPEs. Each of + * the following unions encodes one type. + * + * \bug + * If, at some point, we start generating SPE code from a little-endian host + * these unions will not work. + */ +/*@{*/ +/** + * Encode one output register with two input registers + */ +union spe_inst_RR { + uint32_t bits; + struct { + unsigned op:11; + unsigned rB:7; + unsigned rA:7; + unsigned rT:7; + } inst; +}; + + +/** + * Encode one output register with three input registers + */ +union spe_inst_RRR { + uint32_t bits; + struct { + unsigned op:4; + unsigned rT:7; + unsigned rB:7; + unsigned rA:7; + unsigned rC:7; + } inst; +}; + + +/** + * Encode one output register with one input reg. and a 7-bit signed immed + */ +union spe_inst_RI7 { + uint32_t bits; + struct { + unsigned op:11; + unsigned i7:7; + unsigned rA:7; + unsigned rT:7; + } inst; +}; + + +/** + * Encode one output register with one input reg. and an 8-bit signed immed + */ +union spe_inst_RI8 { + uint32_t bits; + struct { + unsigned op:10; + unsigned i8:8; + unsigned rA:7; + unsigned rT:7; + } inst; +}; + + +/** + * Encode one output register with one input reg. and a 10-bit signed immed + */ +union spe_inst_RI10 { + uint32_t bits; + struct { + unsigned op:8; + unsigned i10:10; + unsigned rA:7; + unsigned rT:7; + } inst; +}; + + +/** + * Encode one output register with a 16-bit signed immediate + */ +union spe_inst_RI16 { + uint32_t bits; + struct { + unsigned op:9; + unsigned i16:16; + unsigned rT:7; + } inst; +}; + + +/** + * Encode one output register with a 18-bit signed immediate + */ +union spe_inst_RI18 { + uint32_t bits; + struct { + unsigned op:7; + unsigned i18:18; + unsigned rT:7; + } inst; +}; +/*@}*/ + + +static void emit_RR(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, unsigned rB) +{ + union spe_inst_RR inst; + inst.inst.op = op; + inst.inst.rB = rB; + inst.inst.rA = rA; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + +static void emit_RRR(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, unsigned rB, unsigned rC) +{ + union spe_inst_RRR inst; + inst.inst.op = op; + inst.inst.rT = rT; + inst.inst.rB = rB; + inst.inst.rA = rA; + inst.inst.rC = rC; + *p->csr = inst.bits; + p->csr++; +} + + +static void emit_RI7(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, int imm) +{ + union spe_inst_RI7 inst; + inst.inst.op = op; + inst.inst.i7 = imm; + inst.inst.rA = rA; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + + +static void emit_RI8(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, int imm) +{ + union spe_inst_RI8 inst; + inst.inst.op = op; + inst.inst.i8 = imm; + inst.inst.rA = rA; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + + +static void emit_RI10(struct spe_function *p, unsigned op, unsigned rT, + unsigned rA, int imm) +{ + union spe_inst_RI10 inst; + inst.inst.op = op; + inst.inst.i10 = imm; + inst.inst.rA = rA; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + +static void emit_RI16(struct spe_function *p, unsigned op, unsigned rT, + int imm) +{ + union spe_inst_RI16 inst; + inst.inst.op = op; + inst.inst.i16 = imm; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + +static void emit_RI18(struct spe_function *p, unsigned op, unsigned rT, + int imm) +{ + union spe_inst_RI18 inst; + inst.inst.op = op; + inst.inst.i18 = imm; + inst.inst.rT = rT; + *p->csr = inst.bits; + p->csr++; +} + + + + +#define EMIT_(_name, _op) \ +void _name (struct spe_function *p, unsigned rT) \ +{ \ + emit_RR(p, _op, rT, 0, 0); \ +} + +#define EMIT_R(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA) \ +{ \ + emit_RR(p, _op, rT, rA, 0); \ +} + +#define EMIT_RR(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, unsigned rB) \ +{ \ + emit_RR(p, _op, rT, rA, rB); \ +} + +#define EMIT_RRR(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, unsigned rB, unsigned rC) \ +{ \ + emit_RRR(p, _op, rT, rA, rB, rC); \ +} + +#define EMIT_RI7(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ +{ \ + emit_RI7(p, _op, rT, rA, imm); \ +} + +#define EMIT_RI8(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ +{ \ + emit_RI8(p, _op, rT, rA, 155 - imm); \ +} + +#define EMIT_RI10(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ +{ \ + emit_RI10(p, _op, rT, rA, imm); \ +} + +#define EMIT_RI16(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, int imm) \ +{ \ + emit_RI16(p, _op, rT, imm); \ +} + +#define EMIT_RI18(_name, _op) \ +void _name (struct spe_function *p, unsigned rT, int imm) \ +{ \ + emit_RI18(p, _op, rT, imm); \ +} + +#define EMIT_I16(_name, _op) \ +void _name (struct spe_function *p, int imm) \ +{ \ + emit_RI16(p, _op, 0, imm); \ +} + +#include "rtasm_ppc_spe.h" + + +/* + */ +void spe_init_func(struct spe_function *p, unsigned code_size) +{ + p->store = align_malloc(code_size, 16); + p->csr = p->store; +} + + +void spe_release_func(struct spe_function *p) +{ + align_free(p->store); + p->store = NULL; + p->csr = NULL; +} + + +void spe_bi(struct spe_function *p, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x1a8, 0, rA, (d << 5) | (e << 4)); +} + +void spe_iret(struct spe_function *p, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x1aa, 0, rA, (d << 5) | (e << 4)); +} + +void spe_bisled(struct spe_function *p, unsigned rT, unsigned rA, int d, + int e) +{ + emit_RI7(p, 0x1ab, rT, rA, (d << 5) | (e << 4)); +} + +void spe_bisl(struct spe_function *p, unsigned rT, unsigned rA, int d, + int e) +{ + emit_RI7(p, 0x1a9, rT, rA, (d << 5) | (e << 4)); +} + +void spe_biz(struct spe_function *p, unsigned rT, unsigned rA, int d, + int e) +{ + emit_RI7(p, 0x128, rT, rA, (d << 5) | (e << 4)); +} + +void spe_binz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x129, rT, rA, (d << 5) | (e << 4)); +} + +void spe_bihz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x12a, rT, rA, (d << 5) | (e << 4)); +} + +void spe_bihnz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) +{ + emit_RI7(p, 0x12b, rT, rA, (d << 5) | (e << 4)); +} + + +/* Hint-for-branch instructions + */ +#if 0 +hbr; +hbra; +hbrr; +#endif + + +/* Control instructions + */ +#if 0 +stop; +EMIT_RR (spe_stopd, 0x140); +EMIT_ (spe_lnop, 0x001); +EMIT_ (spe_nop, 0x201); +sync; +EMIT_ (spe_dsync, 0x003); +EMIT_R (spe_mfspr, 0x00c); +EMIT_R (spe_mtspr, 0x10c); +#endif + +#endif /* GALLIUM_CELL */ diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h new file mode 100644 index 0000000000..10ce44b3a0 --- /dev/null +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h @@ -0,0 +1,314 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file + * Real-time assembly generation interface for Cell B.E. SPEs. + * + * \author Ian Romanick + */ + +#ifndef RTASM_PPC_SPE_H +#define RTASM_PPC_SPE_H + +struct spe_function { + /** + * + */ + uint32_t *store; + uint32_t *csr; + const char *fn; +}; + +extern void spe_init_func(struct spe_function *p, unsigned code_size); +extern void spe_release_func(struct spe_function *p); + +#endif /* RTASM_PPC_SPE_H */ + +#ifndef EMIT_ +#define EMIT_(name, _op) \ + extern void _name (struct spe_function *p, unsigned rT) +#define EMIT_R(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA) +#define EMIT_RR(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + unsigned rB) +#define EMIT_RRR(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + unsigned rB, unsigned rC) +#define EMIT_RI7(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + int imm) +#define EMIT_RI8(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + int imm) +#define EMIT_RI10(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ + int imm) +#define EMIT_RI16(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, int imm) +#define EMIT_RI18(_name, _op) \ + extern void _name (struct spe_function *p, unsigned rT, int imm) +#define EMIT_I16(_name, _op) \ + extern void _name (struct spe_function *p, int imm) +#define UNDEF_EMIT_MACROS +#endif /* EMIT_ */ + + +/* Memory load / store instructions + */ +EMIT_RI10(spe_lqd, 0x034); +EMIT_RR (spe_lqx, 0x1c4); +EMIT_RI16(spe_lqa, 0x061); +EMIT_RI16(spe_lqr, 0x067); +EMIT_RI10(spe_stqd, 0x024); +EMIT_RR (spe_stqx, 0x144); +EMIT_RI16(spe_stqa, 0x041); +EMIT_RI16(spe_stqr, 0x047); +EMIT_RI7 (spe_cbd, 0x1f4); +EMIT_RR (spe_cbx, 0x1d4); +EMIT_RI7 (spe_chd, 0x1f5); +EMIT_RI7 (spe_chx, 0x1d5); +EMIT_RI7 (spe_cwd, 0x1f6); +EMIT_RI7 (spe_cwx, 0x1d6); +EMIT_RI7 (spe_cdd, 0x1f7); +EMIT_RI7 (spe_cdx, 0x1d7); + + +/* Constant formation instructions + */ +EMIT_RI16(spe_ilh, 0x083); +EMIT_RI16(spe_ilhu, 0x082); +EMIT_RI16(spe_il, 0x081); +EMIT_RI18(spe_ila, 0x021); +EMIT_RI16(spe_iohl, 0x0c1); +EMIT_RI16(spe_fsmbi, 0x0c5); + + +/* Integer and logical instructions + */ +EMIT_RR (spe_ah, 0x0c8); +EMIT_RI10(spe_ahi, 0x01d); +EMIT_RR (spe_a, 0x0c0); +EMIT_RI10(spe_ai, 0x01c); +EMIT_RR (spe_sfh, 0x048); +EMIT_RI10(spe_sfhi, 0x00d); +EMIT_RR (spe_sf, 0x040); +EMIT_RI10(spe_sfi, 0x00c); +EMIT_RR (spe_addx, 0x340); +EMIT_RR (spe_cg, 0x0c2); +EMIT_RR (spe_cgx, 0x342); +EMIT_RR (spe_sfx, 0x341); +EMIT_RR (spe_bg, 0x042); +EMIT_RR (spe_bgx, 0x343); +EMIT_RR (spe_mpy, 0x3c4); +EMIT_RR (spe_mpyu, 0x3cc); +EMIT_RI10(spe_mpyi, 0x074); +EMIT_RI10(spe_mpyui, 0x075); +EMIT_RRR (spe_mpya, 0x00c); +EMIT_RR (spe_mpyh, 0x3c5); +EMIT_RR (spe_mpys, 0x3c7); +EMIT_RR (spe_mpyhh, 0x3c6); +EMIT_RR (spe_mpyhha, 0x346); +EMIT_RR (spe_mpyhhu, 0x3ce); +EMIT_RR (spe_mpyhhau, 0x34e); +EMIT_R (spe_clz, 0x2a5); +EMIT_R (spe_cntb, 0x2b4); +EMIT_R (spe_fsmb, 0x1b6); +EMIT_R (spe_fsmh, 0x1b5); +EMIT_R (spe_fsm, 0x1b4); +EMIT_R (spe_gbb, 0x1b2); +EMIT_R (spe_gbh, 0x1b1); +EMIT_R (spe_gb, 0x1b0); +EMIT_RR (spe_avgb, 0x0d3); +EMIT_RR (spe_absdb, 0x053); +EMIT_RR (spe_sumb, 0x253); +EMIT_R (spe_xsbh, 0x2b6); +EMIT_R (spe_xshw, 0x2ae); +EMIT_R (spe_xswd, 0x2a6); +EMIT_RR (spe_and, 0x0c1); +EMIT_RR (spe_andc, 0x2c1); +EMIT_RI10(spe_andbi, 0x016); +EMIT_RI10(spe_andhi, 0x015); +EMIT_RI10(spe_andi, 0x014); +EMIT_RR (spe_or, 0x041); +EMIT_RR (spe_orc, 0x2c9); +EMIT_RI10(spe_orbi, 0x006); +EMIT_RI10(spe_orhi, 0x005); +EMIT_RI10(spe_ori, 0x004); +EMIT_R (spe_orx, 0x1f0); +EMIT_RR (spe_xor, 0x241); +EMIT_RI10(spe_xorbi, 0x026); +EMIT_RI10(spe_xorhi, 0x025); +EMIT_RI10(spe_xori, 0x024); +EMIT_RR (spe_nand, 0x0c9); +EMIT_RR (spe_nor, 0x049); +EMIT_RR (spe_eqv, 0x249); +EMIT_RRR (spe_selb, 0x008); +EMIT_RRR (spe_shufb, 0x00b); + + +/* Shift and rotate instructions + */ +EMIT_RR (spe_shlh, 0x05f); +EMIT_RI7 (spe_shlhi, 0x07f); +EMIT_RR (spe_shl, 0x05b); +EMIT_RI7 (spe_shli, 0x07b); +EMIT_RR (spe_shlqbi, 0x1db); +EMIT_RI7 (spe_shlqbii, 0x1fb); +EMIT_RR (spe_shlqby, 0x1df); +EMIT_RI7 (spe_shlqbyi, 0x1ff); +EMIT_RR (spe_shlqbybi, 0x1cf); +EMIT_RR (spe_roth, 0x05c); +EMIT_RI7 (spe_rothi, 0x07c); +EMIT_RR (spe_rot, 0x058); +EMIT_RI7 (spe_roti, 0x078); +EMIT_RR (spe_rotqby, 0x1dc); +EMIT_RI7 (spe_rotqbyi, 0x1fc); +EMIT_RR (spe_rotqbybi, 0x1cc); +EMIT_RR (spe_rotqbi, 0x1d8); +EMIT_RI7 (spe_rotqbii, 0x1f8); +EMIT_RR (spe_rothm, 0x05d); +EMIT_RI7 (spe_rothmi, 0x07d); +EMIT_RR (spe_rotm, 0x059); +EMIT_RI7 (spe_rotmi, 0x079); +EMIT_RR (spe_rotqmby, 0x1dd); +EMIT_RI7 (spe_rotqmbyi, 0x1fd); +EMIT_RR (spe_rotqmbybi, 0x1cd); +EMIT_RR (spe_rotqmbi, 0x1c9); +EMIT_RI7 (spe_rotqmbii, 0x1f9); +EMIT_RR (spe_rotmah, 0x05e); +EMIT_RI7 (spe_rotmahi, 0x07e); +EMIT_RR (spe_rotma, 0x05a); +EMIT_RI7 (spe_rotmai, 0x07a); + + +/* Compare, branch, and halt instructions + */ +EMIT_RR (spe_heq, 0x3d8); +EMIT_RI10(spe_heqi, 0x07f); +EMIT_RR (spe_hgt, 0x258); +EMIT_RI10(spe_hgti, 0x04f); +EMIT_RR (spe_hlgt, 0x2d8); +EMIT_RI10(spe_hlgti, 0x05f); +EMIT_RR (spe_ceqb, 0x3d0); +EMIT_RI10(spe_ceqbi, 0x07e); +EMIT_RR (spe_ceqh, 0x3c8); +EMIT_RI10(spe_ceqhi, 0x07d); +EMIT_RR (spe_ceq, 0x3c0); +EMIT_RI10(spe_ceqi, 0x07c); +EMIT_RR (spe_cgtb, 0x250); +EMIT_RI10(spe_cgtbi, 0x04e); +EMIT_RR (spe_cgth, 0x248); +EMIT_RI10(spe_cgthi, 0x04d); +EMIT_RR (spe_cgt, 0x240); +EMIT_RI10(spe_cgti, 0x04c); +EMIT_RR (spe_clgtb, 0x2d0); +EMIT_RI10(spe_clgtbi, 0x05e); +EMIT_RR (spe_clgth, 0x2c8); +EMIT_RI10(spe_clgthi, 0x05d); +EMIT_RR (spe_clgt, 0x2c0); +EMIT_RI10(spe_clgti, 0x05c); +EMIT_I16 (spe_br, 0x064); +EMIT_I16 (spe_bra, 0x060); +EMIT_RI16(spe_brsl, 0x066); +EMIT_RI16(spe_brasl, 0x062); +EMIT_RI16(spe_brnz, 0x042); +EMIT_RI16(spe_brz, 0x040); +EMIT_RI16(spe_brhnz, 0x046); +EMIT_RI16(spe_brhz, 0x044); + +extern void spe_bi(struct spe_function *p, unsigned rA, int d, int e); +extern void spe_iret(struct spe_function *p, unsigned rA, int d, int e); +extern void spe_bisled(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spe_bisl(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spe_biz(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spe_binz(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spe_bihz(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); +extern void spe_bihnz(struct spe_function *p, unsigned rT, unsigned rA, + int d, int e); + + +/* Floating-point instructions + */ +EMIT_RR (spe_fa, 0x2c4); +EMIT_RR (spe_dfa, 0x2cc); +EMIT_RR (spe_fs, 0x2c5); +EMIT_RR (spe_dfs, 0x2cd); +EMIT_RR (spe_fm, 0x2c6); +EMIT_RR (spe_dfm, 0x2ce); +EMIT_RRR (spe_fma, 0x00e); +EMIT_RR (spe_dfma, 0x35c); +EMIT_RRR (spe_fnms, 0x00d); +EMIT_RR (spe_dfnms, 0x35e); +EMIT_RRR (spe_fms, 0x00f); +EMIT_RR (spe_dfms, 0x35d); +EMIT_RR (spe_dfnma, 0x35f); +EMIT_R (spe_frest, 0x1b8); +EMIT_R (spe_frsqest, 0x1b9); +EMIT_RR (spe_fi, 0x3d4); +EMIT_RI8 (spe_csflt, 0x1da); +EMIT_RI8 (spe_cflts, 0x1d8); +EMIT_RI8 (spe_cuflt, 0x1db); +EMIT_RI8 (spe_cfltu, 0x1d9); +EMIT_R (spe_frds, 0x3b9); +EMIT_R (spe_fesd, 0x3b8); +EMIT_RR (spe_dfceq, 0x3c3); +EMIT_RR (spe_dfcmeq, 0x3cb); +EMIT_RR (spe_dfcgt, 0x2c3); +EMIT_RR (spe_dfcmgt, 0x2cb); +EMIT_RI7 (spe_dftsv, 0x3bf); +EMIT_RR (spe_fceq, 0x3c2); +EMIT_RR (spe_fcmeq, 0x3ca); +EMIT_RR (spe_fcgt, 0x2c2); +EMIT_RR (spe_fcmgt, 0x2ca); +EMIT_R (spe_fscrwr, 0x3ba); +EMIT_ (spe_fscrrd, 0x398); + + +/* Channel instructions + */ +EMIT_R (spe_rdch, 0x00d); +EMIT_R (spe_rdchcnt, 0x00f); +EMIT_R (spe_wrch, 0x10d); + + +#ifdef UNDEF_EMIT_MACROS +#undef EMIT_ +#undef EMIT_R +#undef EMIT_RR +#undef EMIT_RRR +#undef EMIT_RI7 +#undef EMIT_RI8 +#undef EMIT_RI10 +#undef EMIT_RI16 +#undef EMIT_RI18 +#undef EMIT_I16 +#undef UNDEF_EMIT_MACROS +#endif /* EMIT_ */ diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 91f8e542a2..3b687bb868 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -36,7 +36,7 @@ #include "draw/draw_vbuf.h" #include "cell_winsys.h" #include "cell/common.h" -#include "ppc/rtasm/spe_asm.h" +#include "rtasm/rtasm_ppc_spe.h" struct cell_vbuf_render; diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c index f10689a959..9cf74bab47 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c @@ -31,7 +31,7 @@ #include "../auxiliary/draw/draw_private.h" #include "cell_context.h" -#include "ppc/rtasm/spe_asm.h" +#include "rtasm/rtasm_ppc_spe.h" typedef uint64_t register_mask; diff --git a/src/mesa/ppc/rtasm/spe_asm.c b/src/mesa/ppc/rtasm/spe_asm.c deleted file mode 100644 index 1037637250..0000000000 --- a/src/mesa/ppc/rtasm/spe_asm.c +++ /dev/null @@ -1,385 +0,0 @@ -/* - * (C) Copyright IBM Corporation 2008 - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file spe_asm.c - * Real-time assembly generation interface for Cell B.E. SPEs. - * - * \author Ian Romanick - */ -#ifdef GALLIUM_CELL -#include -#include -#include "spe_asm.h" - -/** - * SPE instruction types - * - * There are 6 primary instruction encodings used on the Cell's SPEs. Each of - * the following unions encodes one type. - * - * \bug - * If, at some point, we start generating SPE code from a little-endian host - * these unions will not work. - */ -/*@{*/ -/** - * Encode one output register with two input registers - */ -union spe_inst_RR { - uint32_t bits; - struct { - unsigned op:11; - unsigned rB:7; - unsigned rA:7; - unsigned rT:7; - } inst; -}; - - -/** - * Encode one output register with three input registers - */ -union spe_inst_RRR { - uint32_t bits; - struct { - unsigned op:4; - unsigned rT:7; - unsigned rB:7; - unsigned rA:7; - unsigned rC:7; - } inst; -}; - - -/** - * Encode one output register with one input reg. and a 7-bit signed immed - */ -union spe_inst_RI7 { - uint32_t bits; - struct { - unsigned op:11; - unsigned i7:7; - unsigned rA:7; - unsigned rT:7; - } inst; -}; - - -/** - * Encode one output register with one input reg. and an 8-bit signed immed - */ -union spe_inst_RI8 { - uint32_t bits; - struct { - unsigned op:10; - unsigned i8:8; - unsigned rA:7; - unsigned rT:7; - } inst; -}; - - -/** - * Encode one output register with one input reg. and a 10-bit signed immed - */ -union spe_inst_RI10 { - uint32_t bits; - struct { - unsigned op:8; - unsigned i10:10; - unsigned rA:7; - unsigned rT:7; - } inst; -}; - - -/** - * Encode one output register with a 16-bit signed immediate - */ -union spe_inst_RI16 { - uint32_t bits; - struct { - unsigned op:9; - unsigned i16:16; - unsigned rT:7; - } inst; -}; - - -/** - * Encode one output register with a 18-bit signed immediate - */ -union spe_inst_RI18 { - uint32_t bits; - struct { - unsigned op:7; - unsigned i18:18; - unsigned rT:7; - } inst; -}; -/*@}*/ - - -static void emit_RR(struct spe_function *p, unsigned op, unsigned rT, - unsigned rA, unsigned rB) -{ - union spe_inst_RR inst; - inst.inst.op = op; - inst.inst.rB = rB; - inst.inst.rA = rA; - inst.inst.rT = rT; - *p->csr = inst.bits; - p->csr++; -} - - -static void emit_RRR(struct spe_function *p, unsigned op, unsigned rT, - unsigned rA, unsigned rB, unsigned rC) -{ - union spe_inst_RRR inst; - inst.inst.op = op; - inst.inst.rT = rT; - inst.inst.rB = rB; - inst.inst.rA = rA; - inst.inst.rC = rC; - *p->csr = inst.bits; - p->csr++; -} - - -static void emit_RI7(struct spe_function *p, unsigned op, unsigned rT, - unsigned rA, int imm) -{ - union spe_inst_RI7 inst; - inst.inst.op = op; - inst.inst.i7 = imm; - inst.inst.rA = rA; - inst.inst.rT = rT; - *p->csr = inst.bits; - p->csr++; -} - - - -static void emit_RI8(struct spe_function *p, unsigned op, unsigned rT, - unsigned rA, int imm) -{ - union spe_inst_RI8 inst; - inst.inst.op = op; - inst.inst.i8 = imm; - inst.inst.rA = rA; - inst.inst.rT = rT; - *p->csr = inst.bits; - p->csr++; -} - - - -static void emit_RI10(struct spe_function *p, unsigned op, unsigned rT, - unsigned rA, int imm) -{ - union spe_inst_RI10 inst; - inst.inst.op = op; - inst.inst.i10 = imm; - inst.inst.rA = rA; - inst.inst.rT = rT; - *p->csr = inst.bits; - p->csr++; -} - - -static void emit_RI16(struct spe_function *p, unsigned op, unsigned rT, - int imm) -{ - union spe_inst_RI16 inst; - inst.inst.op = op; - inst.inst.i16 = imm; - inst.inst.rT = rT; - *p->csr = inst.bits; - p->csr++; -} - - -static void emit_RI18(struct spe_function *p, unsigned op, unsigned rT, - int imm) -{ - union spe_inst_RI18 inst; - inst.inst.op = op; - inst.inst.i18 = imm; - inst.inst.rT = rT; - *p->csr = inst.bits; - p->csr++; -} - - - - -#define EMIT_(_name, _op) \ -void _name (struct spe_function *p, unsigned rT) \ -{ \ - emit_RR(p, _op, rT, 0, 0); \ -} - -#define EMIT_R(_name, _op) \ -void _name (struct spe_function *p, unsigned rT, unsigned rA) \ -{ \ - emit_RR(p, _op, rT, rA, 0); \ -} - -#define EMIT_RR(_name, _op) \ -void _name (struct spe_function *p, unsigned rT, unsigned rA, unsigned rB) \ -{ \ - emit_RR(p, _op, rT, rA, rB); \ -} - -#define EMIT_RRR(_name, _op) \ -void _name (struct spe_function *p, unsigned rT, unsigned rA, unsigned rB, unsigned rC) \ -{ \ - emit_RRR(p, _op, rT, rA, rB, rC); \ -} - -#define EMIT_RI7(_name, _op) \ -void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ -{ \ - emit_RI7(p, _op, rT, rA, imm); \ -} - -#define EMIT_RI8(_name, _op) \ -void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ -{ \ - emit_RI8(p, _op, rT, rA, 155 - imm); \ -} - -#define EMIT_RI10(_name, _op) \ -void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ -{ \ - emit_RI10(p, _op, rT, rA, imm); \ -} - -#define EMIT_RI16(_name, _op) \ -void _name (struct spe_function *p, unsigned rT, int imm) \ -{ \ - emit_RI16(p, _op, rT, imm); \ -} - -#define EMIT_RI18(_name, _op) \ -void _name (struct spe_function *p, unsigned rT, int imm) \ -{ \ - emit_RI18(p, _op, rT, imm); \ -} - -#define EMIT_I16(_name, _op) \ -void _name (struct spe_function *p, int imm) \ -{ \ - emit_RI16(p, _op, 0, imm); \ -} - -#include "spe_asm.h" - - -/* - */ -void spe_init_func(struct spe_function *p, unsigned code_size) -{ - p->store = _mesa_align_malloc(code_size, 16); - p->csr = p->store; -} - - -void spe_release_func(struct spe_function *p) -{ - _mesa_align_free(p->store); - p->store = NULL; - p->csr = NULL; -} - - -void spe_bi(struct spe_function *p, unsigned rA, int d, int e) -{ - emit_RI7(p, 0x1a8, 0, rA, (d << 5) | (e << 4)); -} - -void spe_iret(struct spe_function *p, unsigned rA, int d, int e) -{ - emit_RI7(p, 0x1aa, 0, rA, (d << 5) | (e << 4)); -} - -void spe_bisled(struct spe_function *p, unsigned rT, unsigned rA, int d, - int e) -{ - emit_RI7(p, 0x1ab, rT, rA, (d << 5) | (e << 4)); -} - -void spe_bisl(struct spe_function *p, unsigned rT, unsigned rA, int d, - int e) -{ - emit_RI7(p, 0x1a9, rT, rA, (d << 5) | (e << 4)); -} - -void spe_biz(struct spe_function *p, unsigned rT, unsigned rA, int d, - int e) -{ - emit_RI7(p, 0x128, rT, rA, (d << 5) | (e << 4)); -} - -void spe_binz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) -{ - emit_RI7(p, 0x129, rT, rA, (d << 5) | (e << 4)); -} - -void spe_bihz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) -{ - emit_RI7(p, 0x12a, rT, rA, (d << 5) | (e << 4)); -} - -void spe_bihnz(struct spe_function *p, unsigned rT, unsigned rA, int d, int e) -{ - emit_RI7(p, 0x12b, rT, rA, (d << 5) | (e << 4)); -} - - -/* Hint-for-branch instructions - */ -#if 0 -hbr; -hbra; -hbrr; -#endif - - -/* Control instructions - */ -#if 0 -stop; -EMIT_RR (spe_stopd, 0x140); -EMIT_ (spe_lnop, 0x001); -EMIT_ (spe_nop, 0x201); -sync; -EMIT_ (spe_dsync, 0x003); -EMIT_R (spe_mfspr, 0x00c); -EMIT_R (spe_mtspr, 0x10c); -#endif - -#endif /* GALLIUM_CELL */ diff --git a/src/mesa/ppc/rtasm/spe_asm.h b/src/mesa/ppc/rtasm/spe_asm.h deleted file mode 100644 index 6d69ae655d..0000000000 --- a/src/mesa/ppc/rtasm/spe_asm.h +++ /dev/null @@ -1,314 +0,0 @@ -/* - * (C) Copyright IBM Corporation 2008 - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file spe_asm.h - * Real-time assembly generation interface for Cell B.E. SPEs. - * - * \author Ian Romanick - */ - -#ifndef SPE_ASM_H -#define SPE_ASM_H - -struct spe_function { - /** - * - */ - uint32_t *store; - uint32_t *csr; - const char *fn; -}; - -extern void spe_init_func(struct spe_function *p, unsigned code_size); -extern void spe_release_func(struct spe_function *p); - -#endif /* SPE_ASM_H */ - -#ifndef EMIT_ -#define EMIT_(name, _op) \ - extern void _name (struct spe_function *p, unsigned rT) -#define EMIT_R(_name, _op) \ - extern void _name (struct spe_function *p, unsigned rT, unsigned rA) -#define EMIT_RR(_name, _op) \ - extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ - unsigned rB) -#define EMIT_RRR(_name, _op) \ - extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ - unsigned rB, unsigned rC) -#define EMIT_RI7(_name, _op) \ - extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ - int imm) -#define EMIT_RI8(_name, _op) \ - extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ - int imm) -#define EMIT_RI10(_name, _op) \ - extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ - int imm) -#define EMIT_RI16(_name, _op) \ - extern void _name (struct spe_function *p, unsigned rT, int imm) -#define EMIT_RI18(_name, _op) \ - extern void _name (struct spe_function *p, unsigned rT, int imm) -#define EMIT_I16(_name, _op) \ - extern void _name (struct spe_function *p, int imm) -#define UNDEF_EMIT_MACROS -#endif /* EMIT_ */ - - -/* Memory load / store instructions - */ -EMIT_RI10(spe_lqd, 0x034); -EMIT_RR (spe_lqx, 0x1c4); -EMIT_RI16(spe_lqa, 0x061); -EMIT_RI16(spe_lqr, 0x067); -EMIT_RI10(spe_stqd, 0x024); -EMIT_RR (spe_stqx, 0x144); -EMIT_RI16(spe_stqa, 0x041); -EMIT_RI16(spe_stqr, 0x047); -EMIT_RI7 (spe_cbd, 0x1f4); -EMIT_RR (spe_cbx, 0x1d4); -EMIT_RI7 (spe_chd, 0x1f5); -EMIT_RI7 (spe_chx, 0x1d5); -EMIT_RI7 (spe_cwd, 0x1f6); -EMIT_RI7 (spe_cwx, 0x1d6); -EMIT_RI7 (spe_cdd, 0x1f7); -EMIT_RI7 (spe_cdx, 0x1d7); - - -/* Constant formation instructions - */ -EMIT_RI16(spe_ilh, 0x083); -EMIT_RI16(spe_ilhu, 0x082); -EMIT_RI16(spe_il, 0x081); -EMIT_RI18(spe_ila, 0x021); -EMIT_RI16(spe_iohl, 0x0c1); -EMIT_RI16(spe_fsmbi, 0x0c5); - - -/* Integer and logical instructions - */ -EMIT_RR (spe_ah, 0x0c8); -EMIT_RI10(spe_ahi, 0x01d); -EMIT_RR (spe_a, 0x0c0); -EMIT_RI10(spe_ai, 0x01c); -EMIT_RR (spe_sfh, 0x048); -EMIT_RI10(spe_sfhi, 0x00d); -EMIT_RR (spe_sf, 0x040); -EMIT_RI10(spe_sfi, 0x00c); -EMIT_RR (spe_addx, 0x340); -EMIT_RR (spe_cg, 0x0c2); -EMIT_RR (spe_cgx, 0x342); -EMIT_RR (spe_sfx, 0x341); -EMIT_RR (spe_bg, 0x042); -EMIT_RR (spe_bgx, 0x343); -EMIT_RR (spe_mpy, 0x3c4); -EMIT_RR (spe_mpyu, 0x3cc); -EMIT_RI10(spe_mpyi, 0x074); -EMIT_RI10(spe_mpyui, 0x075); -EMIT_RRR (spe_mpya, 0x00c); -EMIT_RR (spe_mpyh, 0x3c5); -EMIT_RR (spe_mpys, 0x3c7); -EMIT_RR (spe_mpyhh, 0x3c6); -EMIT_RR (spe_mpyhha, 0x346); -EMIT_RR (spe_mpyhhu, 0x3ce); -EMIT_RR (spe_mpyhhau, 0x34e); -EMIT_R (spe_clz, 0x2a5); -EMIT_R (spe_cntb, 0x2b4); -EMIT_R (spe_fsmb, 0x1b6); -EMIT_R (spe_fsmh, 0x1b5); -EMIT_R (spe_fsm, 0x1b4); -EMIT_R (spe_gbb, 0x1b2); -EMIT_R (spe_gbh, 0x1b1); -EMIT_R (spe_gb, 0x1b0); -EMIT_RR (spe_avgb, 0x0d3); -EMIT_RR (spe_absdb, 0x053); -EMIT_RR (spe_sumb, 0x253); -EMIT_R (spe_xsbh, 0x2b6); -EMIT_R (spe_xshw, 0x2ae); -EMIT_R (spe_xswd, 0x2a6); -EMIT_RR (spe_and, 0x0c1); -EMIT_RR (spe_andc, 0x2c1); -EMIT_RI10(spe_andbi, 0x016); -EMIT_RI10(spe_andhi, 0x015); -EMIT_RI10(spe_andi, 0x014); -EMIT_RR (spe_or, 0x041); -EMIT_RR (spe_orc, 0x2c9); -EMIT_RI10(spe_orbi, 0x006); -EMIT_RI10(spe_orhi, 0x005); -EMIT_RI10(spe_ori, 0x004); -EMIT_R (spe_orx, 0x1f0); -EMIT_RR (spe_xor, 0x241); -EMIT_RI10(spe_xorbi, 0x026); -EMIT_RI10(spe_xorhi, 0x025); -EMIT_RI10(spe_xori, 0x024); -EMIT_RR (spe_nand, 0x0c9); -EMIT_RR (spe_nor, 0x049); -EMIT_RR (spe_eqv, 0x249); -EMIT_RRR (spe_selb, 0x008); -EMIT_RRR (spe_shufb, 0x00b); - - -/* Shift and rotate instructions - */ -EMIT_RR (spe_shlh, 0x05f); -EMIT_RI7 (spe_shlhi, 0x07f); -EMIT_RR (spe_shl, 0x05b); -EMIT_RI7 (spe_shli, 0x07b); -EMIT_RR (spe_shlqbi, 0x1db); -EMIT_RI7 (spe_shlqbii, 0x1fb); -EMIT_RR (spe_shlqby, 0x1df); -EMIT_RI7 (spe_shlqbyi, 0x1ff); -EMIT_RR (spe_shlqbybi, 0x1cf); -EMIT_RR (spe_roth, 0x05c); -EMIT_RI7 (spe_rothi, 0x07c); -EMIT_RR (spe_rot, 0x058); -EMIT_RI7 (spe_roti, 0x078); -EMIT_RR (spe_rotqby, 0x1dc); -EMIT_RI7 (spe_rotqbyi, 0x1fc); -EMIT_RR (spe_rotqbybi, 0x1cc); -EMIT_RR (spe_rotqbi, 0x1d8); -EMIT_RI7 (spe_rotqbii, 0x1f8); -EMIT_RR (spe_rothm, 0x05d); -EMIT_RI7 (spe_rothmi, 0x07d); -EMIT_RR (spe_rotm, 0x059); -EMIT_RI7 (spe_rotmi, 0x079); -EMIT_RR (spe_rotqmby, 0x1dd); -EMIT_RI7 (spe_rotqmbyi, 0x1fd); -EMIT_RR (spe_rotqmbybi, 0x1cd); -EMIT_RR (spe_rotqmbi, 0x1c9); -EMIT_RI7 (spe_rotqmbii, 0x1f9); -EMIT_RR (spe_rotmah, 0x05e); -EMIT_RI7 (spe_rotmahi, 0x07e); -EMIT_RR (spe_rotma, 0x05a); -EMIT_RI7 (spe_rotmai, 0x07a); - - -/* Compare, branch, and halt instructions - */ -EMIT_RR (spe_heq, 0x3d8); -EMIT_RI10(spe_heqi, 0x07f); -EMIT_RR (spe_hgt, 0x258); -EMIT_RI10(spe_hgti, 0x04f); -EMIT_RR (spe_hlgt, 0x2d8); -EMIT_RI10(spe_hlgti, 0x05f); -EMIT_RR (spe_ceqb, 0x3d0); -EMIT_RI10(spe_ceqbi, 0x07e); -EMIT_RR (spe_ceqh, 0x3c8); -EMIT_RI10(spe_ceqhi, 0x07d); -EMIT_RR (spe_ceq, 0x3c0); -EMIT_RI10(spe_ceqi, 0x07c); -EMIT_RR (spe_cgtb, 0x250); -EMIT_RI10(spe_cgtbi, 0x04e); -EMIT_RR (spe_cgth, 0x248); -EMIT_RI10(spe_cgthi, 0x04d); -EMIT_RR (spe_cgt, 0x240); -EMIT_RI10(spe_cgti, 0x04c); -EMIT_RR (spe_clgtb, 0x2d0); -EMIT_RI10(spe_clgtbi, 0x05e); -EMIT_RR (spe_clgth, 0x2c8); -EMIT_RI10(spe_clgthi, 0x05d); -EMIT_RR (spe_clgt, 0x2c0); -EMIT_RI10(spe_clgti, 0x05c); -EMIT_I16 (spe_br, 0x064); -EMIT_I16 (spe_bra, 0x060); -EMIT_RI16(spe_brsl, 0x066); -EMIT_RI16(spe_brasl, 0x062); -EMIT_RI16(spe_brnz, 0x042); -EMIT_RI16(spe_brz, 0x040); -EMIT_RI16(spe_brhnz, 0x046); -EMIT_RI16(spe_brhz, 0x044); - -extern void spe_bi(struct spe_function *p, unsigned rA, int d, int e); -extern void spe_iret(struct spe_function *p, unsigned rA, int d, int e); -extern void spe_bisled(struct spe_function *p, unsigned rT, unsigned rA, - int d, int e); -extern void spe_bisl(struct spe_function *p, unsigned rT, unsigned rA, - int d, int e); -extern void spe_biz(struct spe_function *p, unsigned rT, unsigned rA, - int d, int e); -extern void spe_binz(struct spe_function *p, unsigned rT, unsigned rA, - int d, int e); -extern void spe_bihz(struct spe_function *p, unsigned rT, unsigned rA, - int d, int e); -extern void spe_bihnz(struct spe_function *p, unsigned rT, unsigned rA, - int d, int e); - - -/* Floating-point instructions - */ -EMIT_RR (spe_fa, 0x2c4); -EMIT_RR (spe_dfa, 0x2cc); -EMIT_RR (spe_fs, 0x2c5); -EMIT_RR (spe_dfs, 0x2cd); -EMIT_RR (spe_fm, 0x2c6); -EMIT_RR (spe_dfm, 0x2ce); -EMIT_RRR (spe_fma, 0x00e); -EMIT_RR (spe_dfma, 0x35c); -EMIT_RRR (spe_fnms, 0x00d); -EMIT_RR (spe_dfnms, 0x35e); -EMIT_RRR (spe_fms, 0x00f); -EMIT_RR (spe_dfms, 0x35d); -EMIT_RR (spe_dfnma, 0x35f); -EMIT_R (spe_frest, 0x1b8); -EMIT_R (spe_frsqest, 0x1b9); -EMIT_RR (spe_fi, 0x3d4); -EMIT_RI8 (spe_csflt, 0x1da); -EMIT_RI8 (spe_cflts, 0x1d8); -EMIT_RI8 (spe_cuflt, 0x1db); -EMIT_RI8 (spe_cfltu, 0x1d9); -EMIT_R (spe_frds, 0x3b9); -EMIT_R (spe_fesd, 0x3b8); -EMIT_RR (spe_dfceq, 0x3c3); -EMIT_RR (spe_dfcmeq, 0x3cb); -EMIT_RR (spe_dfcgt, 0x2c3); -EMIT_RR (spe_dfcmgt, 0x2cb); -EMIT_RI7 (spe_dftsv, 0x3bf); -EMIT_RR (spe_fceq, 0x3c2); -EMIT_RR (spe_fcmeq, 0x3ca); -EMIT_RR (spe_fcgt, 0x2c2); -EMIT_RR (spe_fcmgt, 0x2ca); -EMIT_R (spe_fscrwr, 0x3ba); -EMIT_ (spe_fscrrd, 0x398); - - -/* Channel instructions - */ -EMIT_R (spe_rdch, 0x00d); -EMIT_R (spe_rdchcnt, 0x00f); -EMIT_R (spe_wrch, 0x10d); - - -#ifdef UNDEF_EMIT_MACROS -#undef EMIT_ -#undef EMIT_R -#undef EMIT_RR -#undef EMIT_RRR -#undef EMIT_RI7 -#undef EMIT_RI8 -#undef EMIT_RI10 -#undef EMIT_RI16 -#undef EMIT_RI18 -#undef EMIT_I16 -#undef UNDEF_EMIT_MACROS -#endif /* EMIT_ */ diff --git a/src/mesa/sources b/src/mesa/sources index 9e56694893..f0bf7b31fb 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -246,7 +246,6 @@ ASM_C_SOURCES = \ x86/rtasm/x86sse.c \ sparc/sparc.c \ ppc/common_ppc.c \ - ppc/rtasm/spe_asm.c \ x86-64/x86-64.c X86_SOURCES = \ -- cgit v1.2.3 From b9da3791c934e05b82063a8c79c423a0a8e29a94 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 15:07:53 +0900 Subject: Remove src/mesa and src/mesa/main from gallium source include paths. --- src/gallium/Makefile.template | 2 - src/gallium/auxiliary/draw/draw_vf_sse.c | 3 +- src/gallium/auxiliary/pipebuffer/linked_list.h | 91 ------ .../auxiliary/pipebuffer/pb_buffer_fenced.c | 3 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c | 3 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c | 3 +- src/gallium/auxiliary/util/u_double_list.h | 91 ++++++ src/gallium/auxiliary/util/u_simple_list.h | 197 +++++++++++++ src/gallium/drivers/i915simple/i915_debug.c | 2 - src/gallium/include/pipe/p_thread.h | 327 +++++++++++++++++++-- src/gallium/winsys/dri/intel/SConscript | 2 +- src/gallium/winsys/dri/intel/intel_batchpool.c | 15 +- src/gallium/winsys/xlib/SConscript | 6 + src/gallium/winsys/xlib/brw_aub.c | 16 +- src/mesa/SConscript | 7 + src/mesa/x86/common_x86_asm.S | 2 +- 16 files changed, 620 insertions(+), 150 deletions(-) delete mode 100644 src/gallium/auxiliary/pipebuffer/linked_list.h create mode 100644 src/gallium/auxiliary/util/u_double_list.h create mode 100644 src/gallium/auxiliary/util/u_simple_list.h (limited to 'src') diff --git a/src/gallium/Makefile.template b/src/gallium/Makefile.template index 6698212e77..4e462b5c97 100644 --- a/src/gallium/Makefile.template +++ b/src/gallium/Makefile.template @@ -18,8 +18,6 @@ INCLUDES = \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/main \ -I$(TOP)/include \ $(DRIVER_INCLUDES) diff --git a/src/gallium/auxiliary/draw/draw_vf_sse.c b/src/gallium/auxiliary/draw/draw_vf_sse.c index 6076f9849d..aff4ffd985 100644 --- a/src/gallium/auxiliary/draw/draw_vf_sse.c +++ b/src/gallium/auxiliary/draw/draw_vf_sse.c @@ -26,9 +26,8 @@ */ -#include "simple_list.h" - #include "pipe/p_compiler.h" +#include "util/u_simple_list.h" #include "draw_vf.h" diff --git a/src/gallium/auxiliary/pipebuffer/linked_list.h b/src/gallium/auxiliary/pipebuffer/linked_list.h deleted file mode 100644 index e99817fb13..0000000000 --- a/src/gallium/auxiliary/pipebuffer/linked_list.h +++ /dev/null @@ -1,91 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND. USA. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - **************************************************************************/ - -/** - * \file - * List macros heavily inspired by the Linux kernel - * list handling. No list looping yet. - * - * Is not threadsafe, so common operations need to - * be protected using an external mutex. - */ - -#ifndef LINKED_LIST_H_ -#define LINKED_LIST_H_ - - -#include - - -struct list_head -{ - struct list_head *prev; - struct list_head *next; -}; - - -#define LIST_INITHEAD(__item) \ - do { \ - (__item)->prev = (__item); \ - (__item)->next = (__item); \ - } while (0) - -#define LIST_ADD(__item, __list) \ - do { \ - (__item)->prev = (__list); \ - (__item)->next = (__list)->next; \ - (__list)->next->prev = (__item); \ - (__list)->next = (__item); \ - } while (0) - -#define LIST_ADDTAIL(__item, __list) \ - do { \ - (__item)->next = (__list); \ - (__item)->prev = (__list)->prev; \ - (__list)->prev->next = (__item); \ - (__list)->prev = (__item); \ - } while(0) - -#define LIST_DEL(__item) \ - do { \ - (__item)->prev->next = (__item)->next; \ - (__item)->next->prev = (__item)->prev; \ - } while(0) - -#define LIST_DELINIT(__item) \ - do { \ - (__item)->prev->next = (__item)->next; \ - (__item)->next->prev = (__item)->prev; \ - (__item)->next = (__item); \ - (__item)->prev = (__item); \ - } while(0) - -#define LIST_ENTRY(__type, __item, __field) \ - ((__type *)(((char *)(__item)) - offsetof(__type, __field))) - - -#endif /*LINKED_LIST_H_*/ diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index bc85c4b19f..e2ee72ed1f 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -34,13 +34,12 @@ */ -#include "linked_list.h" - #include "pipe/p_compiler.h" #include "pipe/p_debug.h" #include "pipe/p_winsys.h" #include "pipe/p_thread.h" #include "pipe/p_util.h" +#include "util/u_double_list.h" #include "pb_buffer.h" #include "pb_buffer_fenced.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index 983a105347..ff4fd123f3 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -33,12 +33,11 @@ */ -#include "linked_list.h" - #include "pipe/p_defines.h" #include "pipe/p_debug.h" #include "pipe/p_thread.h" #include "pipe/p_util.h" +#include "util/u_double_list.h" #include "util/u_mm.h" #include "pb_buffer.h" #include "pb_bufmgr.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c index beb145b7cb..528e9528f6 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c @@ -35,13 +35,12 @@ */ -#include "linked_list.h" - #include "pipe/p_compiler.h" #include "pipe/p_debug.h" #include "pipe/p_thread.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "util/u_double_list.h" #include "pb_buffer.h" #include "pb_bufmgr.h" diff --git a/src/gallium/auxiliary/util/u_double_list.h b/src/gallium/auxiliary/util/u_double_list.h new file mode 100644 index 0000000000..8cb10c9820 --- /dev/null +++ b/src/gallium/auxiliary/util/u_double_list.h @@ -0,0 +1,91 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND. USA. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + **************************************************************************/ + +/** + * \file + * List macros heavily inspired by the Linux kernel + * list handling. No list looping yet. + * + * Is not threadsafe, so common operations need to + * be protected using an external mutex. + */ + +#ifndef _U_DOUBLE_LIST_H_ +#define _U_DOUBLE_LIST_H_ + + +#include + + +struct list_head +{ + struct list_head *prev; + struct list_head *next; +}; + + +#define LIST_INITHEAD(__item) \ + do { \ + (__item)->prev = (__item); \ + (__item)->next = (__item); \ + } while (0) + +#define LIST_ADD(__item, __list) \ + do { \ + (__item)->prev = (__list); \ + (__item)->next = (__list)->next; \ + (__list)->next->prev = (__item); \ + (__list)->next = (__item); \ + } while (0) + +#define LIST_ADDTAIL(__item, __list) \ + do { \ + (__item)->next = (__list); \ + (__item)->prev = (__list)->prev; \ + (__list)->prev->next = (__item); \ + (__list)->prev = (__item); \ + } while(0) + +#define LIST_DEL(__item) \ + do { \ + (__item)->prev->next = (__item)->next; \ + (__item)->next->prev = (__item)->prev; \ + } while(0) + +#define LIST_DELINIT(__item) \ + do { \ + (__item)->prev->next = (__item)->next; \ + (__item)->next->prev = (__item)->prev; \ + (__item)->next = (__item); \ + (__item)->prev = (__item); \ + } while(0) + +#define LIST_ENTRY(__type, __item, __field) \ + ((__type *)(((char *)(__item)) - offsetof(__type, __field))) + + +#endif /*_U_DOUBLE_LIST_H_*/ diff --git a/src/gallium/auxiliary/util/u_simple_list.h b/src/gallium/auxiliary/util/u_simple_list.h new file mode 100644 index 0000000000..f5f43b0faa --- /dev/null +++ b/src/gallium/auxiliary/util/u_simple_list.h @@ -0,0 +1,197 @@ +/** + * \file simple_list.h + * Simple macros for type-safe, intrusive lists. + * + * Intended to work with a list sentinal which is created as an empty + * list. Insert & delete are O(1). + * + * \author + * (C) 1997, Keith Whitwell + */ + +/* + * Mesa 3-D graphics library + * Version: 3.5 + * + * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef _U_SIMPLE_LIST_H_ +#define _U_SIMPLE_LIST_H_ + +/** + * Remove an element from list. + * + * \param elem element to remove. + */ +#define remove_from_list(elem) \ +do { \ + (elem)->next->prev = (elem)->prev; \ + (elem)->prev->next = (elem)->next; \ +} while (0) + +/** + * Insert an element to the list head. + * + * \param list list. + * \param elem element to insert. + */ +#define insert_at_head(list, elem) \ +do { \ + (elem)->prev = list; \ + (elem)->next = (list)->next; \ + (list)->next->prev = elem; \ + (list)->next = elem; \ +} while(0) + +/** + * Insert an element to the list tail. + * + * \param list list. + * \param elem element to insert. + */ +#define insert_at_tail(list, elem) \ +do { \ + (elem)->next = list; \ + (elem)->prev = (list)->prev; \ + (list)->prev->next = elem; \ + (list)->prev = elem; \ +} while(0) + +/** + * Move an element to the list head. + * + * \param list list. + * \param elem element to move. + */ +#define move_to_head(list, elem) \ +do { \ + remove_from_list(elem); \ + insert_at_head(list, elem); \ +} while (0) + +/** + * Move an element to the list tail. + * + * \param list list. + * \param elem element to move. + */ +#define move_to_tail(list, elem) \ +do { \ + remove_from_list(elem); \ + insert_at_tail(list, elem); \ +} while (0) + +/** + * Make a empty list empty. + * + * \param sentinal list (sentinal element). + */ +#define make_empty_list(sentinal) \ +do { \ + (sentinal)->next = sentinal; \ + (sentinal)->prev = sentinal; \ +} while (0) + +/** + * Get list first element. + * + * \param list list. + * + * \return pointer to first element. + */ +#define first_elem(list) ((list)->next) + +/** + * Get list last element. + * + * \param list list. + * + * \return pointer to last element. + */ +#define last_elem(list) ((list)->prev) + +/** + * Get next element. + * + * \param elem element. + * + * \return pointer to next element. + */ +#define next_elem(elem) ((elem)->next) + +/** + * Get previous element. + * + * \param elem element. + * + * \return pointer to previous element. + */ +#define prev_elem(elem) ((elem)->prev) + +/** + * Test whether element is at end of the list. + * + * \param list list. + * \param elem element. + * + * \return non-zero if element is at end of list, or zero otherwise. + */ +#define at_end(list, elem) ((elem) == (list)) + +/** + * Test if a list is empty. + * + * \param list list. + * + * \return non-zero if list empty, or zero otherwise. + */ +#define is_empty_list(list) ((list)->next == (list)) + +/** + * Walk through the elements of a list. + * + * \param ptr pointer to the current element. + * \param list list. + * + * \note It should be followed by a { } block or a single statement, as in a \c + * for loop. + */ +#define foreach(ptr, list) \ + for( ptr=(list)->next ; ptr!=list ; ptr=(ptr)->next ) + +/** + * Walk through the elements of a list. + * + * Same as #foreach but lets you unlink the current value during a list + * traversal. Useful for freeing a list, element by element. + * + * \param ptr pointer to the current element. + * \param t temporary pointer. + * \param list list. + * + * \note It should be followed by a { } block or a single statement, as in a \c + * for loop. + */ +#define foreach_s(ptr, t, list) \ + for(ptr=(list)->next,t=(ptr)->next; list != ptr; ptr=t, t=(t)->next) + +#endif /* _U_SIMPLE_LIST_H_ */ diff --git a/src/gallium/drivers/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c index 94db44e1aa..78102dbac2 100644 --- a/src/gallium/drivers/i915simple/i915_debug.c +++ b/src/gallium/drivers/i915simple/i915_debug.c @@ -25,8 +25,6 @@ * **************************************************************************/ -//#include "imports.h" - #include "i915_reg.h" #include "i915_context.h" #include "i915_winsys.h" diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h index cd432c547c..4325abc951 100644 --- a/src/gallium/include/pipe/p_thread.h +++ b/src/gallium/include/pipe/p_thread.h @@ -1,54 +1,317 @@ /************************************************************************** * + * Copyright 1999-2006 Brian Paul * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * **************************************************************************/ -#ifndef P_THREAD_H -#define P_THREAD_H +/** + * @file + * Thread + * + * Initial version by John Stone (j.stone@acm.org) (johns@cs.umr.edu) + * and Christoph Poliwoda (poliwoda@volumegraphics.com) + * Revised by Keith Whitwell + * Adapted for new gl dispatcher by Brian Paul + * + * + * + * DOCUMENTATION + * + * This thread module exports the following types: + * _glthread_TSD Thread-specific data area + * _glthread_Thread Thread datatype + * _glthread_Mutex Mutual exclusion lock + * + * Macros: + * _glthread_DECLARE_STATIC_MUTEX(name) Declare a non-local mutex + * _glthread_INIT_MUTEX(name) Initialize a mutex + * _glthread_LOCK_MUTEX(name) Lock a mutex + * _glthread_UNLOCK_MUTEX(name) Unlock a mutex + * + * Functions: + * _glthread_GetID(v) Get integer thread ID + * _glthread_InitTSD() Initialize thread-specific data + * _glthread_GetTSD() Get thread-specific data + * _glthread_SetTSD() Set thread-specific data + * + * If this file is accidentally included by a non-threaded build, + * it should not cause the build to fail, or otherwise cause problems. + * In general, it should only be included when needed however. + */ + +#ifndef _P_THREAD_H_ +#define _P_THREAD_H_ + + +#if defined(USE_MGL_NAMESPACE) +#define _glapi_Dispatch _mglapi_Dispatch +#endif + + + +#if (defined(PTHREADS) || defined(SOLARIS_THREADS) ||\ + defined(WIN32_THREADS) || defined(USE_XTHREADS) || defined(BEOS_THREADS)) \ + && !defined(THREADS) +# define THREADS +#endif + +#ifdef VMS +#include +#endif + +/* + * POSIX threads. This should be your choice in the Unix world + * whenever possible. When building with POSIX threads, be sure + * to enable any compiler flags which will cause the MT-safe + * libc (if one exists) to be used when linking, as well as any + * header macros for MT-safe errno, etc. For Solaris, this is the -mt + * compiler flag. On Solaris with gcc, use -D_REENTRANT to enable + * proper compiling for MT-safe libc etc. + */ +#if defined(PTHREADS) +#include /* POSIX threads headers */ + +typedef struct { + pthread_key_t key; + int initMagic; +} _glthread_TSD; + +typedef pthread_t _glthread_Thread; + +typedef pthread_mutex_t _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) \ + static _glthread_Mutex name = PTHREAD_MUTEX_INITIALIZER + +#define _glthread_INIT_MUTEX(name) \ + pthread_mutex_init(&(name), NULL) + +#define _glthread_DESTROY_MUTEX(name) \ + pthread_mutex_destroy(&(name)) + +#define _glthread_LOCK_MUTEX(name) \ + (void) pthread_mutex_lock(&(name)) + +#define _glthread_UNLOCK_MUTEX(name) \ + (void) pthread_mutex_unlock(&(name)) + +#endif /* PTHREADS */ + + + + +/* + * Solaris threads. Use only up to Solaris 2.4. + * Solaris 2.5 and higher provide POSIX threads. + * Be sure to compile with -mt on the Solaris compilers, or + * use -D_REENTRANT if using gcc. + */ +#ifdef SOLARIS_THREADS +#include + +typedef struct { + thread_key_t key; + mutex_t keylock; + int initMagic; +} _glthread_TSD; + +typedef thread_t _glthread_Thread; + +typedef mutex_t _glthread_Mutex; + +/* XXX need to really implement mutex-related macros */ +#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 +#define _glthread_INIT_MUTEX(name) (void) name +#define _glthread_DESTROY_MUTEX(name) (void) name +#define _glthread_LOCK_MUTEX(name) (void) name +#define _glthread_UNLOCK_MUTEX(name) (void) name + +#endif /* SOLARIS_THREADS */ + + + + +/* + * Windows threads. Should work with Windows NT and 95. + * IMPORTANT: Link with multithreaded runtime library when THREADS are + * used! + */ +#ifdef WIN32_THREADS +#include + +typedef struct { + DWORD key; + int initMagic; +} _glthread_TSD; + +typedef HANDLE _glthread_Thread; + +typedef CRITICAL_SECTION _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) /*static*/ _glthread_Mutex name = {0,0,0,0,0,0} +#define _glthread_INIT_MUTEX(name) InitializeCriticalSection(&name) +#define _glthread_DESTROY_MUTEX(name) DeleteCriticalSection(&name) +#define _glthread_LOCK_MUTEX(name) EnterCriticalSection(&name) +#define _glthread_UNLOCK_MUTEX(name) LeaveCriticalSection(&name) + +#endif /* WIN32_THREADS */ + + -#include "p_compiler.h" /* - * XXX: We should come up with a system-independent thread definitions. - * XXX: Patching glthread defs for now. + * XFree86 has its own thread wrapper, Xthreads.h + * We wrap it again for GL. */ +#ifdef USE_XTHREADS +#include + +typedef struct { + xthread_key_t key; + int initMagic; +} _glthread_TSD; + +typedef xthread_t _glthread_Thread; + +typedef xmutex_rec _glthread_Mutex; + +#ifdef XMUTEX_INITIALIZER +#define _glthread_DECLARE_STATIC_MUTEX(name) \ + static _glthread_Mutex name = XMUTEX_INITIALIZER +#else +#define _glthread_DECLARE_STATIC_MUTEX(name) \ + static _glthread_Mutex name +#endif + +#define _glthread_INIT_MUTEX(name) \ + xmutex_init(&(name)) + +#define _glthread_DESTROY_MUTEX(name) \ + xmutex_clear(&(name)) + +#define _glthread_LOCK_MUTEX(name) \ + (void) xmutex_lock(&(name)) + +#define _glthread_UNLOCK_MUTEX(name) \ + (void) xmutex_unlock(&(name)) + +#endif /* USE_XTHREADS */ + + + +/* + * BeOS threads. R5.x required. + */ +#ifdef BEOS_THREADS + +#include +#include + +typedef struct { + int32 key; + int initMagic; +} _glthread_TSD; + +typedef thread_id _glthread_Thread; + +/* Use Benaphore, aka speeder semaphore */ +typedef struct { + int32 lock; + sem_id sem; +} benaphore; +typedef benaphore _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = { 0, 0 } +#define _glthread_INIT_MUTEX(name) name.sem = create_sem(0, #name"_benaphore"), name.lock = 0 +#define _glthread_DESTROY_MUTEX(name) delete_sem(name.sem), name.lock = 0 +#define _glthread_LOCK_MUTEX(name) if (name.sem == 0) _glthread_INIT_MUTEX(name); \ + if (atomic_add(&(name.lock), 1) >= 1) acquire_sem(name.sem) +#define _glthread_UNLOCK_MUTEX(name) if (atomic_add(&(name.lock), -1) > 1) release_sem(name.sem) + +#endif /* BEOS_THREADS */ + + + +#ifndef THREADS + +/* + * THREADS not defined + */ + +typedef unsigned _glthread_TSD; + +typedef unsigned _glthread_Thread; + +typedef unsigned _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 + +#define _glthread_INIT_MUTEX(name) (void) name + +#define _glthread_DESTROY_MUTEX(name) (void) name + +#define _glthread_LOCK_MUTEX(name) (void) name + +#define _glthread_UNLOCK_MUTEX(name) (void) name + +#endif /* THREADS */ + + + +/* + * Platform independent thread specific data API. + */ + +extern unsigned long +_glthread_GetID(void); + + +extern void +_glthread_InitTSD(_glthread_TSD *); + + +extern void * +_glthread_GetTSD(_glthread_TSD *); -#ifndef __MSC__ -#include "glapi/glthread.h" +extern void +_glthread_SetTSD(_glthread_TSD *, void *); -#else /* __MSC__ */ +#if defined(GLX_USE_TLS) -typedef int _glthread_Mutex; +extern __thread struct _glapi_table * _glapi_tls_Dispatch + __attribute__((tls_model("initial-exec"))); -#define _glthread_INIT_MUTEX( M ) ((void) (M)) -#define _glthread_LOCK_MUTEX( M ) ((void) (M)) -#define _glthread_UNLOCK_MUTEX( M ) ((void) (M)) +#define GET_DISPATCH() _glapi_tls_Dispatch -#define sched_yield() ((void) 0) +#elif !defined(GL_CALL) +# if defined(THREADS) +# define GET_DISPATCH() \ + ((__builtin_expect( _glapi_Dispatch != NULL, 1 )) \ + ? _glapi_Dispatch : _glapi_get_dispatch()) +# else +# define GET_DISPATCH() _glapi_Dispatch +# endif /* defined(THREADS) */ +#endif /* ndef GL_CALL */ -#endif /* __MSC__ */ -#endif /* P_THREAD_H */ +#endif /* _P_THREAD_H_ */ diff --git a/src/gallium/winsys/dri/intel/SConscript b/src/gallium/winsys/dri/intel/SConscript index 525ba580e8..0ad19d42a8 100644 --- a/src/gallium/winsys/dri/intel/SConscript +++ b/src/gallium/winsys/dri/intel/SConscript @@ -35,5 +35,5 @@ drivers = [ env.SharedLibrary( target ='i915tex_dri.so', source = sources, - LIBS = mesa + drivers + auxiliaries + env['LIBS'], + LIBS = drivers + mesa + auxiliaries + env['LIBS'], ) \ No newline at end of file diff --git a/src/gallium/winsys/dri/intel/intel_batchpool.c b/src/gallium/winsys/dri/intel/intel_batchpool.c index 33b56817f6..ce154c7b88 100644 --- a/src/gallium/winsys/dri/intel/intel_batchpool.c +++ b/src/gallium/winsys/dri/intel/intel_batchpool.c @@ -36,9 +36,12 @@ #include #include +#include #include -#include "imports.h" -#include "glthread.h" + +#include "pipe/p_compiler.h" +#include "pipe/p_thread.h" + #include "dri_bufpool.h" #include "dri_bufmgr.h" #include "intel_batchpool.h" @@ -196,7 +199,7 @@ pool_create(struct _DriBufferPool *pool, _glthread_LOCK_MUTEX(p->mutex); if (p->numFree == 0) - pool_checkFree(p, GL_TRUE); + pool_checkFree(p, TRUE); if (p->numFree == 0) { fprintf(stderr, "Out of fixed size buffer objects\n"); @@ -278,7 +281,7 @@ pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, return -EBUSY; } - buf->mapped = GL_TRUE; + buf->mapped = TRUE; *virtual = (unsigned char *) p->virtual + buf->start; _glthread_UNLOCK_MUTEX(p->mutex); return 0; @@ -361,7 +364,7 @@ pool_validate(struct _DriBufferPool *pool, void *private) BBuf *buf = (BBuf *) private; BPool *p = buf->parent; _glthread_LOCK_MUTEX(p->mutex); - buf->unfenced = GL_TRUE; + buf->unfenced = TRUE; _glthread_UNLOCK_MUTEX(p->mutex); return 0; } @@ -379,7 +382,7 @@ pool_takedown(struct _DriBufferPool *pool) while ((p->numFree < p->numTot) && p->numDelayed) { _glthread_UNLOCK_MUTEX(p->mutex); sched_yield(); - pool_checkFree(p, GL_TRUE); + pool_checkFree(p, TRUE); _glthread_LOCK_MUTEX(p->mutex); } diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript index f8aa5ef945..c38b5be52c 100644 --- a/src/gallium/winsys/xlib/SConscript +++ b/src/gallium/winsys/xlib/SConscript @@ -3,6 +3,12 @@ Import('*') +env = env.Clone() + +env.Append(CPPPATH = [ + '#/src/mesa', + '#/src/mesa/main', +]) sources = [ 'glxapi.c', diff --git a/src/gallium/winsys/xlib/brw_aub.c b/src/gallium/winsys/xlib/brw_aub.c index 541d50c6e4..10eedd8402 100644 --- a/src/gallium/winsys/xlib/brw_aub.c +++ b/src/gallium/winsys/xlib/brw_aub.c @@ -29,11 +29,13 @@ * Keith Whitwell */ +#include +#include #include "brw_aub.h" #include "pipe/p_context.h" #include "pipe/p_state.h" -#include "imports.h" -//#include "intel_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_debug.h" struct brw_aubfile { @@ -350,9 +352,9 @@ struct brw_aubfile *brw_aubfile_create( void ) i++; - if (_mesa_getenv("INTEL_AUBFILE")) { - val = snprintf(filename, sizeof(filename), "%s%d.aub", _mesa_getenv("INTEL_AUBFILE"), i%4); - _mesa_printf("--> Aub file: %s\n", filename); + if (getenv("INTEL_AUBFILE")) { + val = snprintf(filename, sizeof(filename), "%s%d.aub", getenv("INTEL_AUBFILE"), i%4); + debug_printf("--> Aub file: %s\n", filename); aubfile->file = fopen(filename, "w"); } else { @@ -360,12 +362,12 @@ struct brw_aubfile *brw_aubfile_create( void ) if (val < 0 || val > sizeof(filename)) strcpy(filename, "default.aub"); - _mesa_printf("--> Aub file: %s\n", filename); + debug_printf("--> Aub file: %s\n", filename); aubfile->file = fopen(filename, "w"); } if (!aubfile->file) { - _mesa_printf("couldn't open aubfile\n"); + debug_printf("couldn't open aubfile\n"); exit(1); } diff --git a/src/mesa/SConscript b/src/mesa/SConscript index a828133580..db18c61fac 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -4,6 +4,13 @@ Import('*') +env = env.Clone() + +# Includes +env.Append(CPPPATH = [ + '#/src/mesa', + '#/src/mesa/main', +]) ####################################################################### # Core sources diff --git a/src/mesa/x86/common_x86_asm.S b/src/mesa/x86/common_x86_asm.S index ef3cc9eb59..09c86b05ba 100644 --- a/src/mesa/x86/common_x86_asm.S +++ b/src/mesa/x86/common_x86_asm.S @@ -39,7 +39,7 @@ * in there will break the build on some platforms. */ -#include "matypes.h" +#include "assyntax.h" #include "common_x86_features.h" SEG_TEXT -- cgit v1.2.3 From b62f0ddd09f8ef9c400feca321412b3f0bc77e63 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 18:56:55 +0900 Subject: Portability fixes. Eliminate C99/C++ constructs. (We should actually disable gcc C99 syntax options if we are serious about portability.) --- src/gallium/auxiliary/cso_cache/cso_hash.c | 35 ++++++++++++++++++++---------- 1 file changed, 23 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index 0338cb3b47..b40217c524 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -158,11 +158,14 @@ static void cso_data_rehash(struct cso_hash_data *hash, int hint) while (firstNode != e) { unsigned h = firstNode->key; struct cso_node *lastNode = firstNode; + struct cso_node *afterLastNode; + struct cso_node **beforeFirstNode; + while (lastNode->next != e && lastNode->next->key == h) lastNode = lastNode->next; - struct cso_node *afterLastNode = lastNode->next; - struct cso_node **beforeFirstNode = &hash->buckets[h % hash->numBuckets]; + afterLastNode = lastNode->next; + beforeFirstNode = &hash->buckets[h % hash->numBuckets]; while (*beforeFirstNode != e) beforeFirstNode = &(*beforeFirstNode)->next; lastNode->next = *beforeFirstNode; @@ -222,10 +225,12 @@ struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, { cso_data_might_grow(hash->data.d); - struct cso_node **nextNode = cso_hash_find_node(hash, key); - struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); - struct cso_hash_iter iter = {hash, node}; - return iter; + { + struct cso_node **nextNode = cso_hash_find_node(hash, key); + struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); + struct cso_hash_iter iter = {hash, node}; + return iter; + } } struct cso_hash * cso_hash_create(void) @@ -290,6 +295,10 @@ static struct cso_node *cso_hash_data_next(struct cso_node *node) struct cso_node *e; struct cso_hash_data *d; } a; + int start; + struct cso_node **bucket; + int n; + a.next = node->next; if (!a.next) { fprintf(stderr, "iterating beyond the last element\n"); @@ -298,9 +307,9 @@ static struct cso_node *cso_hash_data_next(struct cso_node *node) if (a.next->next) return a.next; - int start = (node->key % a.d->numBuckets) + 1; - struct cso_node **bucket = a.d->buckets + start; - int n = a.d->numBuckets - start; + start = (node->key % a.d->numBuckets) + 1; + bucket = a.d->buckets + start; + n = a.d->numBuckets - start; while (n--) { if (*bucket != a.e) return *bucket; @@ -316,19 +325,21 @@ static struct cso_node *cso_hash_data_prev(struct cso_node *node) struct cso_node *e; struct cso_hash_data *d; } a; + int start; + struct cso_node *sentinel; + struct cso_node **bucket; a.e = node; while (a.e->next) a.e = a.e->next; - int start; if (node == a.e) start = a.d->numBuckets - 1; else start = node->key % a.d->numBuckets; - struct cso_node *sentinel = node; - struct cso_node **bucket = a.d->buckets + start; + sentinel = node; + bucket = a.d->buckets + start; while (start >= 0) { if (*bucket != sentinel) { struct cso_node *prev = *bucket; -- cgit v1.2.3 From 5480a6bc13a555f99a89fc801cfe153182697dda Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Feb 2008 18:57:25 +0900 Subject: Fix windows build. --- src/gallium/auxiliary/rtasm/rtasm_execmem.c | 3 ++- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 2 +- src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 2 +- 3 files changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.c b/src/gallium/auxiliary/rtasm/rtasm_execmem.c index 9c78fa5626..300c1c2d9d 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_execmem.c +++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.c @@ -33,6 +33,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_debug.h" #include "pipe/p_thread.h" +#include "pipe/p_util.h" #include "rtasm_execmem.h" @@ -118,7 +119,7 @@ rtasm_exec_free(void *addr) */ void * -rtasm_exec_malloc(GLuint size) +rtasm_exec_malloc(size_t size) { return MALLOC( size ); } diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index b332192a62..dcbf76f600 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -21,7 +21,7 @@ * **************************************************************************/ -#if defined(__i386__) || defined(__386__) +#if defined(__i386__) || defined(__386__) || defined(i386) #include "pipe/p_compiler.h" #include "pipe/p_debug.h" diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index e4576001bf..606b41eb35 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -24,7 +24,7 @@ #ifndef _RTASM_X86SSE_H_ #define _RTASM_X86SSE_H_ -#if defined(__i386__) || defined(__386__) +#if defined(__i386__) || defined(__386__) || defined(i386) /* It is up to the caller to ensure that instructions issued are * suitable for the host cpu. There are no checks made in this module -- cgit v1.2.3 From 4362c6e59d575a039e654e1520bbff89b73fc8f2 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 18 Feb 2008 18:51:57 -0800 Subject: Cell: trivial clean-ups --- src/gallium/drivers/cell/spu/spu_exec.c | 8 ++------ 1 file changed, 2 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 109540b1f7..0eb5ea1a3f 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -146,17 +146,13 @@ spu_exec_machine_init(struct spu_exec_machine *mach, struct spu_sampler *samplers, unsigned processor) { - qword zero; - qword not_zero; - uint i; + const qword zero = si_il(0); + const qword not_zero = si_il(~0); mach->Samplers = samplers; mach->Processor = processor; mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; - zero = si_xor(zero, zero); - not_zero = si_xori(zero, 0xff); - /* Setup constants. */ mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].q = zero; mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].q = not_zero; -- cgit v1.2.3 From 66be2810c3be07dd1ee45a60cfc632725837f2cd Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 18 Feb 2008 18:55:39 -0800 Subject: Cell: emit vertex shaders and uniforms more intelligently --- src/gallium/drivers/cell/common.h | 22 ++++---- src/gallium/drivers/cell/ppu/cell_state_emit.c | 18 ++++++ src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 17 +++--- src/gallium/drivers/cell/spu/spu_main.c | 9 +++ src/gallium/drivers/cell/spu/spu_vertex_fetch.c | 6 +- src/gallium/drivers/cell/spu/spu_vertex_shader.c | 68 ++++++++++------------- src/gallium/drivers/cell/spu/spu_vertex_shader.h | 5 ++ 7 files changed, 85 insertions(+), 60 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index 74b131fbef..cf892206c6 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -87,10 +87,12 @@ #define CELL_CMD_STATE_TEXTURE 13 #define CELL_CMD_STATE_VERTEX_INFO 14 #define CELL_CMD_STATE_VIEWPORT 15 -#define CELL_CMD_STATE_VS_ARRAY_INFO 16 -#define CELL_CMD_STATE_BLEND 17 -#define CELL_CMD_VS_EXECUTE 18 -#define CELL_CMD_STATE_ATTRIB_FETCH 19 +#define CELL_CMD_STATE_UNIFORMS 16 +#define CELL_CMD_STATE_VS_ARRAY_INFO 17 +#define CELL_CMD_STATE_BIND_VS 18 +#define CELL_CMD_STATE_BLEND 19 +#define CELL_CMD_STATE_ATTRIB_FETCH 20 +#define CELL_CMD_VS_EXECUTE 21 #define CELL_NUM_BUFFERS 4 @@ -144,14 +146,13 @@ struct cell_attribute_fetch_code { struct cell_shader_info { - unsigned num_outputs; - uint64_t declarations; - unsigned num_declarations; uint64_t instructions; - unsigned num_instructions; - uint64_t uniforms; uint64_t immediates; + + unsigned num_outputs; + unsigned num_declarations; + unsigned num_instructions; unsigned num_immediates; } ALIGN16_ATTRIB; @@ -160,10 +161,9 @@ struct cell_shader_info struct cell_command_vs { uint64_t opcode; /**< CELL_CMD_VS_EXECUTE */ - struct cell_shader_info shader; + uint64_t vOut[SPU_VERTS_PER_BATCH]; unsigned num_elts; unsigned elts[SPU_VERTS_PER_BATCH]; - uint64_t vOut[SPU_VERTS_PER_BATCH]; float plane[12][4]; unsigned nr_planes; unsigned nr_attrs; diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 5d2a786449..49c0d130c5 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -31,6 +31,8 @@ #include "cell_state_emit.h" #include "cell_batch.h" #include "cell_texture.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" static void @@ -100,4 +102,20 @@ cell_emit_state(struct cell_context *cell) emit_state_cmd(cell, CELL_CMD_STATE_VERTEX_INFO, &cell->vertex_info, sizeof(struct vertex_info)); } + + if (cell->dirty & CELL_NEW_VS) { + const struct draw_context *const draw = cell->draw; + struct cell_shader_info info; + + info.num_outputs = draw->num_vs_outputs; + info.declarations = (uintptr_t) draw->machine.Declarations; + info.num_declarations = draw->machine.NumDeclarations; + info.instructions = (uintptr_t) draw->machine.Instructions; + info.num_instructions = draw->machine.NumInstructions; + info.immediates = (uintptr_t) draw->machine.Imms; + info.num_immediates = draw->machine.ImmLimit / 4; + + emit_state_cmd(cell, CELL_CMD_STATE_BIND_VS, + & info, sizeof(info)); + } } diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index 6a1d3bc20a..64c7821c19 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -54,6 +54,7 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) struct cell_command_vs *const vs = &cell_global.command[0].vs; uint64_t *batch; struct cell_array_info *array_info; + struct cell_shader_info *shader_info; unsigned i, j; struct cell_attribute_fetch_code *cf; @@ -100,17 +101,17 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) (void) memcpy(&batch[1], &draw->viewport, sizeof(struct pipe_viewport_state)); + { + uint64_t uniforms = (uintptr_t) draw->user.constants; + + batch = cell_batch_alloc(cell, 2 *sizeof(batch[0])); + batch[0] = CELL_CMD_STATE_UNIFORMS; + batch[1] = uniforms; + } + cell_batch_flush(cell); vs->opcode = CELL_CMD_VS_EXECUTE; - vs->shader.num_outputs = draw->num_vs_outputs; - vs->shader.declarations = (uintptr_t) draw->machine.Declarations; - vs->shader.num_declarations = draw->machine.NumDeclarations; - vs->shader.instructions = (uintptr_t) draw->machine.Instructions; - vs->shader.num_instructions = draw->machine.NumInstructions; - vs->shader.uniforms = (uintptr_t) draw->user.constants; - vs->shader.immediates = (uintptr_t) draw->machine.Imms; - vs->shader.num_immediates = draw->machine.ImmLimit / 4; vs->nr_attrs = draw->vertex_fetch.nr_attrs; (void) memcpy(vs->plane, draw->plane, sizeof(draw->plane)); diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index fcbf0f841e..dbc3705c24 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -433,10 +433,19 @@ cmd_batch(uint opcode) sizeof(struct pipe_viewport_state)); pos += (1 + ROUNDUP8(sizeof(struct pipe_viewport_state)) / 8); break; + case CELL_CMD_STATE_UNIFORMS: + draw.constants = (float (*)[4]) (uintptr_t) buffer[pos + 1]; + pos += 2; + break; case CELL_CMD_STATE_VS_ARRAY_INFO: cmd_state_vs_array_info((struct cell_array_info *) &buffer[pos+1]); pos += (1 + ROUNDUP8(sizeof(struct cell_array_info)) / 8); break; + case CELL_CMD_STATE_BIND_VS: + spu_bind_vertex_shader(&draw, + (struct cell_shader_info *) &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct cell_shader_info)) / 8); + break; case CELL_CMD_STATE_ATTRIB_FETCH: { struct cell_attribute_fetch_code *code = (struct cell_attribute_fetch_code *) &buffer[pos+1]; diff --git a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c index 55c6c28717..e5d9910ff3 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c @@ -64,7 +64,7 @@ typedef void (*spu_fetch_func)(qword *out, const qword *in, const qword *shuffle_data); -static const qword fetch_shuffle_data[] = { +static const qword fetch_shuffle_data[5] ALIGN16_ATTRIB = { /* Shuffle used by CVT_64_FLOAT */ { @@ -108,7 +108,7 @@ static const qword fetch_shuffle_data[] = { static INLINE void fetch_unaligned(qword *dst, unsigned ea, unsigned size) { - qword tmp[4]; + qword tmp[4] ALIGN16_ATTRIB; const int shift = ea & 0x0f; const unsigned aligned_start_ea = ea & ~0x0f; const unsigned aligned_end_ea = (ea + size) & ~0x0f; @@ -169,7 +169,7 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, unsigned idx; const unsigned bytes_per_entry = draw->vertex_fetch.size[attr]; const unsigned quads_per_entry = (bytes_per_entry + 15) / 16; - qword in[2 * 4]; + qword in[2 * 4] ALIGN16_ATTRIB; /* Fetch four attributes for four vertices. diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.c b/src/gallium/drivers/cell/spu/spu_vertex_shader.c index 3f5bf41aa2..8363efeeb6 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_shader.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.c @@ -165,63 +165,55 @@ run_vertex_program(struct spu_vs_context *draw, } -static void -spu_bind_vertex_shader(struct spu_vs_context *draw, - void *uniforms, - void *planes, - unsigned nr_planes, - unsigned num_outputs - ) -{ - draw->constants = (float (*)[4]) uniforms; - - (void) memcpy(draw->plane, planes, sizeof(float) * 4 * nr_planes); - draw->nr_planes = nr_planes; - draw->num_vs_outputs = num_outputs; - - /* specify the shader to interpret/execute */ - spu_exec_machine_init(&draw->machine, - PIPE_MAX_SAMPLERS, - NULL /*samplers*/, - PIPE_SHADER_VERTEX); -} - - unsigned char immediates[(sizeof(float) * 4 * TGSI_EXEC_NUM_IMMEDIATES) + 32] ALIGN16_ATTRIB; + void -spu_execute_vertex_shader(struct spu_vs_context *draw, - const struct cell_command_vs *vs) +spu_bind_vertex_shader(struct spu_vs_context *draw, + struct cell_shader_info *vs) { - unsigned i; - - const uint64_t immediate_addr = vs->shader.immediates; + const unsigned immediate_addr = vs->immediates; const unsigned immediate_size = - ROUNDUP16((sizeof(float) * 4 * vs->shader.num_immediates) - + (immediate_addr & 0x0f)); + ROUNDUP16((sizeof(float) * 4 * vs->num_immediates) + + (immediate_addr & 0x0f)); + mfc_get(immediates, immediate_addr & ~0x0f, immediate_size, TAG_VERTEX_BUFFER, 0, 0); draw->machine.Instructions = (struct tgsi_full_instruction *) - vs->shader.instructions; - draw->machine.NumInstructions = vs->shader.num_instructions; + vs->instructions; + draw->machine.NumInstructions = vs->num_instructions; draw->machine.Declarations = (struct tgsi_full_declaration *) - vs->shader.declarations; - draw->machine.NumDeclarations = vs->shader.num_declarations; + vs->declarations; + draw->machine.NumDeclarations = vs->num_declarations; - draw->vertex_fetch.nr_attrs = vs->nr_attrs; + draw->num_vs_outputs = vs->num_outputs; + + /* specify the shader to interpret/execute */ + spu_exec_machine_init(&draw->machine, + PIPE_MAX_SAMPLERS, + NULL /*samplers*/, + PIPE_SHADER_VERTEX); wait_on_mask(1 << TAG_VERTEX_BUFFER); (void) memcpy(& draw->machine.Imms, &immediates[immediate_addr & 0x0f], - sizeof(float) * 4 * vs->shader.num_immediates); + sizeof(float) * 4 * vs->num_immediates); +} - spu_bind_vertex_shader(draw, vs->shader.uniforms, - vs->plane, vs->nr_planes, - vs->shader.num_outputs); + +void +spu_execute_vertex_shader(struct spu_vs_context *draw, + const struct cell_command_vs *vs) +{ + unsigned i; + + (void) memcpy(draw->plane, vs->plane, sizeof(float) * 4 * vs->nr_planes); + draw->nr_planes = vs->nr_planes; + draw->vertex_fetch.nr_attrs = vs->nr_attrs; for (i = 0; i < vs->num_elts; i += 4) { const unsigned batch_size = MIN2(vs->num_elts - i, 4); diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.h b/src/gallium/drivers/cell/spu/spu_vertex_shader.h index 0fb0bc28d0..54a4b8d9b9 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_shader.h +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.h @@ -1,6 +1,7 @@ #ifndef SPU_VERTEX_SHADER_H #define SPU_VERTEX_SHADER_H +#include "cell/common.h" #include "pipe/p_format.h" #include "spu_exec.h" @@ -54,6 +55,10 @@ static INLINE void spu_vertex_fetch(struct spu_vs_context *draw, struct cell_command_vs; +extern void +spu_bind_vertex_shader(struct spu_vs_context *draw, + struct cell_shader_info *vs); + extern void spu_execute_vertex_shader(struct spu_vs_context *draw, const struct cell_command_vs *vs); -- cgit v1.2.3 From 3e329ea7e41e8d97de5b5f345ecab0833c8afe70 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Feb 2008 11:14:54 -0700 Subject: gallium: updated cell build Building on Ian's Cell build fix. Put libcell.a in the gallium/drivers/cell/ directory. General Makefile clean-up, simplification, updated comments. --- src/gallium/drivers/cell/ppu/Makefile | 12 ++++++++---- src/gallium/winsys/xlib/Makefile | 19 ++++++++++--------- 2 files changed, 18 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile index 196ab777f5..3c3f622a2f 100644 --- a/src/gallium/drivers/cell/ppu/Makefile +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -1,6 +1,6 @@ # Gallium3D Cell driver: PPU code -# This makefile builds the g3dcell.a library which gets pulled into +# This makefile builds the libcell.a library which gets pulled into # the main libGL.so library @@ -8,10 +8,14 @@ TOP = ../../../../.. include $(TOP)/configs/linux-cell -#PROG = gl4 +# This is the "top-level" cell PPU driver code, will get pulled into libGL.so +# by the winsys Makefile. +CELL_LIB = ../libcell.a -CELL_LIB = libcell.a +# This is the SPU code. We'd like to be able to put this into the libcell.a +# archive with the PPU code, but nesting .a libs doesn't seem to work. +# So, it's pulled into libGL.so in gallium/winsys/xlib/Makefile SPU_CODE_MODULE = ../spu/g3d_spu.a @@ -56,7 +60,7 @@ default: $(CELL_LIB) $(CELL_LIB): $(OBJECTS) $(SPU_CODE_MODULE) -# ar -ru $(CELL_LIB) $(OBJECTS) $(SPU_CODE_MODULE) +# ar -ru $(CELL_LIB) $(OBJECTS) $(SPU_CODE_MODULE) # doesn't work ar -ru $(CELL_LIB) $(OBJECTS) #$(PROG): $(PPU_OBJECTS) diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile index 17405388ee..09f10e5ea8 100644 --- a/src/gallium/winsys/xlib/Makefile +++ b/src/gallium/winsys/xlib/Makefile @@ -1,4 +1,4 @@ -# src/mesa/Makefile +# src/gallium/winsys/xlib/Makefile TOP = ../../../.. include $(TOP)/configs/current @@ -27,16 +27,17 @@ GL_MINOR = 5 GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) +ifeq ($(CONFIG_NAME), linux-cell) +# The SPU code is in a separate .a file, unfortunately +CELL_SPU_LIB = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a +endif + PIPE_LIB = \ $(GALLIUM_DRIVERS) \ $(TOP)/src/mesa/libglapi.a \ $(TOP)/src/mesa/libmesa.a \ - $(GALLIUM_AUXILIARIES) - -ifeq ($(CONFIG_NAME), linux-cell) -CELL_LIB = $(TOP)/src/gallium/drivers/cell/ppu/libcell.a -CELL_LIB_SPU = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a -endif + $(GALLIUM_AUXILIARIES) \ + $(CELL_SPU_LIB) \ .SUFFIXES : .cpp @@ -65,13 +66,13 @@ STAND_ALONE_OBJECTS = \ $(STAND_ALONE_DRIVER_OBJECTS) # Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(TOP)/bin/mklib -o $(GL_LIB) \ -linker "$(CC)" \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ - --start-group $(PIPE_LIB) --end-group $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) + --start-group $(PIPE_LIB) --end-group $(GL_LIB_DEPS) ###################################################################### -- cgit v1.2.3 From b1c8fa5b6002296d9abe21c06d5cb81a3f70828a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Feb 2008 14:55:18 -0700 Subject: gallium: implement correct sampling for RECT targets / unnormalized texcoords --- src/gallium/drivers/softpipe/sp_tex_sample.c | 168 +++++++++++++++++++++++---- 1 file changed, 143 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index c54e9d385c..ff872f360e 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -332,6 +332,61 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, } +/** + * For RECT textures / unnormalized texcoords + * Only a subset of wrap modes supported. + */ +static INLINE int +nearest_texcoord_unnorm(unsigned wrapMode, float s, unsigned size) +{ + int i; + switch (wrapMode) { + case PIPE_TEX_WRAP_CLAMP: + i = ifloor(s); + return CLAMP(i, 0, size-1); + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + /* fall-through */ + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + return ifloor( CLAMP(s, 0.5F, size - 0.5F) ); + default: + assert(0); + return 0; + } +} + + +/** + * For RECT textures / unnormalized texcoords. + * Only a subset of wrap modes supported. + */ +static INLINE void +linear_texcoord_unnorm(unsigned wrapMode, float s, unsigned size, + int *i0, int *i1, float *a) +{ + switch (wrapMode) { + case PIPE_TEX_WRAP_CLAMP: + /* Not exactly what the spec says, but it matches NVIDIA output */ + s = CLAMP(s - 0.5F, 0.0, size - 1.0); + *i0 = ifloor(s); + *i1 = *i0 + 1; + break; + case PIPE_TEX_WRAP_CLAMP_TO_EDGE: + /* fall-through */ + case PIPE_TEX_WRAP_CLAMP_TO_BORDER: + s = CLAMP(s, 0.5F, size - 0.5F); + s -= 0.5F; + *i0 = ifloor(s); + *i1 = *i0 + 1; + if (*i1 > size - 1) + *i1 = size - 1; + break; + default: + assert(0); + } + *a = FRAC(s); +} + + static unsigned choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) { @@ -415,15 +470,15 @@ compute_lambda(struct tgsi_sampler *sampler, { float rho, lambda; + assert(sampler->state->normalized_coords); + assert(s); { float dsdx = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]; float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; dsdx = FABSF(dsdx); dsdy = FABSF(dsdy); - rho = MAX2(dsdx, dsdy); - if (sampler->state->normalized_coords) - rho *= sampler->texture->width[0]; + rho = MAX2(dsdx, dsdy) * sampler->texture->width[0]; } if (t) { float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; @@ -431,9 +486,7 @@ compute_lambda(struct tgsi_sampler *sampler, float max; dtdx = FABSF(dtdx); dtdy = FABSF(dtdy); - max = MAX2(dtdx, dtdy); - if (sampler->state->normalized_coords) - max *= sampler->texture->height[0]; + max = MAX2(dtdx, dtdy) * sampler->texture->height[0]; rho = MAX2(rho, max); } if (p) { @@ -442,9 +495,7 @@ compute_lambda(struct tgsi_sampler *sampler, float max; dpdx = FABSF(dpdx); dpdy = FABSF(dpdy); - max = MAX2(dpdx, dpdy); - if (sampler->state->normalized_coords) - max *= sampler->texture->depth[0]; + max = MAX2(dpdx, dpdy) * sampler->texture->depth[0]; rho = MAX2(rho, max); } @@ -628,13 +679,10 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, choose_mipmap_levels(sampler, s, t, p, lodbias, &level0, &level1, &levelBlend, &imgFilter); - if (sampler->state->normalized_coords) { - width = sampler->texture->width[level0]; - height = sampler->texture->height[level0]; - } - else { - width = height = 1; - } + assert(sampler->state->normalized_coords); + + width = sampler->texture->width[level0]; + height = sampler->texture->height[level0]; assert(width > 0); @@ -765,14 +813,11 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, choose_mipmap_levels(sampler, s, t, p, lodbias, &level0, &level1, &levelBlend, &imgFilter); - if (sampler->state->normalized_coords) { - width = sampler->texture->width[level0]; - height = sampler->texture->height[level0]; - depth = sampler->texture->depth[level0]; - } - else { - width = height = depth = 1; - } + assert(sampler->state->normalized_coords); + + width = sampler->texture->width[level0]; + height = sampler->texture->height[level0]; + depth = sampler->texture->depth[level0]; assert(width > 0); assert(height > 0); @@ -889,6 +934,73 @@ sp_get_samples_cube(struct tgsi_sampler *sampler, } +static void +sp_get_samples_rect(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]) +{ + //sp_get_samples_2d_common(sampler, s, t, p, lodbias, rgba, faces); + static const uint face = 0; + const uint compare_func = sampler->state->compare_func; + unsigned level0, level1, j, imgFilter; + int width, height; + float levelBlend; + + choose_mipmap_levels(sampler, s, t, p, lodbias, + &level0, &level1, &levelBlend, &imgFilter); + + /* texture RECTS cannot be mipmapped */ + assert(level0 == level1); + + width = sampler->texture->width[level0]; + height = sampler->texture->height[level0]; + + assert(width > 0); + + switch (imgFilter) { + case PIPE_TEX_FILTER_NEAREST: + for (j = 0; j < QUAD_SIZE; j++) { + int x = nearest_texcoord_unnorm(sampler->state->wrap_s, s[j], width); + int y = nearest_texcoord_unnorm(sampler->state->wrap_t, t[j], height); + get_texel(sampler, face, level0, x, y, 0, rgba, j); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(compare_func, rgba, p, j); + } + } + break; + case PIPE_TEX_FILTER_LINEAR: + for (j = 0; j < QUAD_SIZE; j++) { + float tx[4][4], a, b; + int x0, y0, x1, y1, c; + linear_texcoord_unnorm(sampler->state->wrap_s, s[j], width, &x0, &x1, &a); + linear_texcoord_unnorm(sampler->state->wrap_t, t[j], height, &y0, &y1, &b); + get_texel(sampler, face, level0, x0, y0, 0, tx, 0); + get_texel(sampler, face, level0, x1, y0, 0, tx, 1); + get_texel(sampler, face, level0, x0, y1, 0, tx, 2); + get_texel(sampler, face, level0, x1, y1, 0, tx, 3); + if (sampler->state->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + shadow_compare(compare_func, tx, p, 0); + shadow_compare(compare_func, tx, p, 1); + shadow_compare(compare_func, tx, p, 2); + shadow_compare(compare_func, tx, p, 3); + } + + for (c = 0; c < 4; c++) { + rgba[c][j] = lerp_2d(a, b, tx[c][0], tx[c][1], tx[c][2], tx[c][3]); + } + } + break; + default: + assert(0); + } +} + + + + /** * Called via tgsi_sampler::get_samples() * Use the sampler's state setting to get a filtered RGBA value @@ -914,15 +1026,21 @@ sp_get_samples(struct tgsi_sampler *sampler, switch (sampler->texture->target) { case PIPE_TEXTURE_1D: + assert(sampler->state->normalized_coords); sp_get_samples_1d(sampler, s, t, p, lodbias, rgba); break; case PIPE_TEXTURE_2D: - sp_get_samples_2d(sampler, s, t, p, lodbias, rgba); + if (sampler->state->normalized_coords) + sp_get_samples_2d(sampler, s, t, p, lodbias, rgba); + else + sp_get_samples_rect(sampler, s, t, p, lodbias, rgba); break; case PIPE_TEXTURE_3D: + assert(sampler->state->normalized_coords); sp_get_samples_3d(sampler, s, t, p, lodbias, rgba); break; case PIPE_TEXTURE_CUBE: + assert(sampler->state->normalized_coords); sp_get_samples_cube(sampler, s, t, p, lodbias, rgba); break; default: -- cgit v1.2.3 From a2c06c5b5c8b3fb9f6d65bcd288d62e112e6a603 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Feb 2008 14:56:01 -0700 Subject: gallium: don't hard-code attrib slot=0 in setup_fragcoord_coeff() --- src/gallium/drivers/softpipe/sp_prim_setup.c | 41 ++++++++++++---------------- 1 file changed, 18 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index d73521ccbe..7b1e131ee1 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -476,33 +476,33 @@ static void tri_persp_coeff( struct setup_stage *setup, * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask. */ static void -setup_fragcoord_coeff(struct setup_stage *setup) +setup_fragcoord_coeff(struct setup_stage *setup, uint slot) { /*X*/ - setup->coef[0].a0[0] = 0; - setup->coef[0].dadx[0] = 1.0; - setup->coef[0].dady[0] = 0.0; + setup->coef[slot].a0[0] = 0; + setup->coef[slot].dadx[0] = 1.0; + setup->coef[slot].dady[0] = 0.0; /*Y*/ if (setup->softpipe->rasterizer->origin_lower_left) { /* y=0=bottom */ const int winHeight = setup->softpipe->framebuffer.cbufs[0]->height; - setup->coef[0].a0[1] = (float) (winHeight - 1); - setup->coef[0].dady[1] = -1.0; + setup->coef[slot].a0[1] = (float) (winHeight - 1); + setup->coef[slot].dady[1] = -1.0; } else { /* y=0=top */ - setup->coef[0].a0[1] = 0.0; - setup->coef[0].dady[1] = 1.0; + setup->coef[slot].a0[1] = 0.0; + setup->coef[slot].dady[1] = 1.0; } - setup->coef[0].dadx[1] = 0.0; + setup->coef[slot].dadx[1] = 0.0; /*Z*/ - setup->coef[0].a0[2] = setup->posCoef.a0[2]; - setup->coef[0].dadx[2] = setup->posCoef.dadx[2]; - setup->coef[0].dady[2] = setup->posCoef.dady[2]; + setup->coef[slot].a0[2] = setup->posCoef.a0[2]; + setup->coef[slot].dadx[2] = setup->posCoef.dadx[2]; + setup->coef[slot].dady[2] = setup->posCoef.dady[2]; /*W*/ - setup->coef[0].a0[3] = setup->posCoef.a0[3]; - setup->coef[0].dadx[3] = setup->posCoef.dadx[3]; - setup->coef[0].dady[3] = setup->posCoef.dady[3]; + setup->coef[slot].a0[3] = setup->posCoef.a0[3]; + setup->coef[slot].dadx[3] = setup->posCoef.dadx[3]; + setup->coef[slot].dady[3] = setup->posCoef.dady[3]; } @@ -543,8 +543,7 @@ static void setup_tri_coefficients( struct setup_stage *setup ) tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); break; case INTERP_POS: - assert(fragSlot == 0); - setup_fragcoord_coeff(setup); + setup_fragcoord_coeff(setup, fragSlot); break; default: assert(0); @@ -798,9 +797,7 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); break; case INTERP_POS: - assert(fragSlot == 0); - assert(0); /* XXX fix this: */ - setup_fragcoord_coeff(setup); + setup_fragcoord_coeff(setup, fragSlot); break; default: assert(0); @@ -1022,9 +1019,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) &setup->coef[fragSlot], vertSlot, j); break; case INTERP_POS: - assert(fragSlot == 0); - assert(0); /* XXX fix this: */ - setup_fragcoord_coeff(setup); + setup_fragcoord_coeff(setup, fragSlot); break; default: assert(0); -- cgit v1.2.3 From 4ec46e4869b60b60c7ddf43168604713b5c4c359 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Feb 2008 14:58:23 -0700 Subject: gallium: add some casts to prevent likely msvc warnings --- src/gallium/drivers/softpipe/sp_tex_sample.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index ff872f360e..43d5085895 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -347,7 +347,7 @@ nearest_texcoord_unnorm(unsigned wrapMode, float s, unsigned size) case PIPE_TEX_WRAP_CLAMP_TO_EDGE: /* fall-through */ case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - return ifloor( CLAMP(s, 0.5F, size - 0.5F) ); + return ifloor( CLAMP(s, 0.5F, (float) size - 0.5F) ); default: assert(0); return 0; @@ -366,14 +366,14 @@ linear_texcoord_unnorm(unsigned wrapMode, float s, unsigned size, switch (wrapMode) { case PIPE_TEX_WRAP_CLAMP: /* Not exactly what the spec says, but it matches NVIDIA output */ - s = CLAMP(s - 0.5F, 0.0, size - 1.0); + s = CLAMP(s - 0.5F, 0.0, (float) size - 1.0); *i0 = ifloor(s); *i1 = *i0 + 1; break; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: /* fall-through */ case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - s = CLAMP(s, 0.5F, size - 0.5F); + s = CLAMP(s, 0.5F, (float) size - 0.5F); s -= 0.5F; *i0 = ifloor(s); *i1 = *i0 + 1; -- cgit v1.2.3 From 75a4524f2c6444b27055e539da052827670b62cf Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Feb 2008 16:28:25 -0700 Subject: gallium: initialize the killmask register to zero before running shader This fixes mysterious missing fragments when running with SSE. --- src/gallium/drivers/softpipe/sp_fs_sse.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index b18772f4e6..53050b7823 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -124,6 +124,9 @@ fs_sse_run( struct sp_fragment_shader *base, (float)quad->x0, (float)quad->y0, machine->Temps); + /* init kill mask */ + machine->Temps[TGSI_EXEC_TEMP_KILMASK_I].xyzw[TGSI_EXEC_TEMP_KILMASK_C].u[0] = 0x0; + shader->func( machine->Inputs, machine->Outputs, machine->Consts, -- cgit v1.2.3 From 49c3f3b537cdad847eaa24f90d01c4b1f604f724 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 19 Feb 2008 16:51:19 -0700 Subject: gallium: general clean-up of xlib winsys Makefile --- src/gallium/winsys/xlib/Makefile | 41 ++++++++++++---------------------------- 1 file changed, 12 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile index 09f10e5ea8..c443330942 100644 --- a/src/gallium/winsys/xlib/Makefile +++ b/src/gallium/winsys/xlib/Makefile @@ -4,6 +4,11 @@ TOP = ../../../.. include $(TOP)/configs/current +GL_MAJOR = 1 +GL_MINOR = 5 +GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) + + INCLUDE_DIRS = \ -I$(TOP)/include \ -I$(TOP)/src/mesa \ @@ -12,7 +17,7 @@ INCLUDE_DIRS = \ -I$(TOP)/src/gallium/drivers \ -I$(TOP)/src/gallium/auxiliary -X11_DRIVER_SOURCES = \ +XLIB_WINSYS_SOURCES = \ glxapi.c \ fakeglx.c \ xfonts.c \ @@ -21,10 +26,7 @@ X11_DRIVER_SOURCES = \ xm_winsys_aub.c \ brw_aub.c - -GL_MAJOR = 1 -GL_MINOR = 5 -GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) +XLIB_WINSYS_OBJECTS = $(XLIB_WINSYS_SOURCES:.c=.o) ifeq ($(CONFIG_NAME), linux-cell) @@ -32,7 +34,7 @@ ifeq ($(CONFIG_NAME), linux-cell) CELL_SPU_LIB = $(TOP)/src/gallium/drivers/cell/spu/g3d_spu.a endif -PIPE_LIB = \ +LIBS = \ $(GALLIUM_DRIVERS) \ $(TOP)/src/mesa/libglapi.a \ $(TOP)/src/mesa/libmesa.a \ @@ -48,35 +50,20 @@ PIPE_LIB = \ .cpp.o: $(CXX) -c $(INCLUDE_DIRS) $(CXXFLAGS) $< -o $@ -.S.o: - $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ default: $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) -###################################################################### -# Stand-alone Mesa libGL and libOSMesa -STAND_ALONE_DRIVER_SOURCES = \ - $(X11_DRIVER_SOURCES) - -STAND_ALONE_DRIVER_OBJECTS = $(STAND_ALONE_DRIVER_SOURCES:.c=.o) - -STAND_ALONE_OBJECTS = \ - $(STAND_ALONE_DRIVER_OBJECTS) - -# Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) +# Make the libGL.so library +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(XLIB_WINSYS_OBJECTS) $(LIBS) $(TOP)/bin/mklib -o $(GL_LIB) \ -linker "$(CC)" \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ - $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ - --start-group $(PIPE_LIB) --end-group $(GL_LIB_DEPS) - + $(MKLIB_OPTIONS) $(XLIB_WINSYS_OBJECTS) \ + --start-group $(LIBS) --end-group $(GL_LIB_DEPS) -###################################################################### -# Generic stuff depend: $(ALL_SOURCES) @ echo "running $(MKDEP)" @@ -94,10 +81,6 @@ install: default $(INSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \ fi -## NOT INSTALLED YET: -## $(INSTALL) -d $(INSTALL_DIR)/include/GLES -## $(INSTALL) -m 644 include/GLES/*.h $(INSTALL_DIR)/include/GLES - # Emacs tags tags: -- cgit v1.2.3 From 22a0b85eaebf767f5b03bf899596e09f5cc03876 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 11:15:59 -0700 Subject: gallium: use pipe_texture_reference() in sp_tile_cache_set_texture() --- src/gallium/drivers/softpipe/sp_state_sampler.c | 2 +- src/gallium/drivers/softpipe/sp_tile_cache.c | 5 +++-- src/gallium/drivers/softpipe/sp_tile_cache.h | 3 ++- 3 files changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c index 460adccec4..9246915e19 100644 --- a/src/gallium/drivers/softpipe/sp_state_sampler.c +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -84,7 +84,7 @@ softpipe_set_sampler_texture(struct pipe_context *pipe, assert(unit < PIPE_MAX_SAMPLERS); softpipe->texture[unit] = softpipe_texture(texture); /* ptr, not struct */ - sp_tile_cache_set_texture(softpipe->tex_cache[unit], texture); + sp_tile_cache_set_texture(pipe, softpipe->tex_cache[unit], texture); softpipe->dirty |= SP_NEW_TEXTURE; } diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index dde3fabc81..9ed3c5072d 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -212,14 +212,15 @@ sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc) * Specify the texture to cache. */ void -sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, +sp_tile_cache_set_texture(struct pipe_context *pipe, + struct softpipe_tile_cache *tc, struct pipe_texture *texture) { uint i; assert(!tc->surface); - tc->texture = texture; + pipe_texture_reference(pipe, &tc->texture, texture); if (tc->tex_surf_map) { pipe_surface_unmap(tc->tex_surf); diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.h b/src/gallium/drivers/softpipe/sp_tile_cache.h index 7fd1081286..2631e29a3a 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.h +++ b/src/gallium/drivers/softpipe/sp_tile_cache.h @@ -80,7 +80,8 @@ extern void sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc); extern void -sp_tile_cache_set_texture(struct softpipe_tile_cache *tc, +sp_tile_cache_set_texture(struct pipe_context *pipe, + struct softpipe_tile_cache *tc, struct pipe_texture *texture); extern void -- cgit v1.2.3 From 58edb0683db45c449b219988a8715cf8fd69e42d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 11:20:25 -0700 Subject: gallium: state tracker didn't always notify drivers of texobj data changes Calling glTexSubImage() or glTexImage() to replace texture data didn't reliably cause pipe->set_sampler_texture() to get called so drivers didn't always get notified of new texture data. The st_texture_object->pt pointer doesn't always indicate changed data so added a dirtyData field. --- src/mesa/state_tracker/st_atom_texture.c | 18 ++++----- src/mesa/state_tracker/st_cb_drawpixels.c | 3 +- src/mesa/state_tracker/st_cb_fbo.c | 1 + src/mesa/state_tracker/st_cb_texture.c | 42 ++++---------------- src/mesa/state_tracker/st_cb_texture.h | 32 ++++++++++++++-- src/mesa/state_tracker/st_context.h | 22 +---------- src/mesa/state_tracker/st_gen_mipmap.c | 4 +- src/mesa/state_tracker/st_texture.h | 64 +++++++++++++++++++++++++++++++ 8 files changed, 115 insertions(+), 71 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 2a836d630b..9fead7e314 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -34,6 +34,7 @@ #include "st_context.h" #include "st_atom.h" +#include "st_texture.h" #include "st_cb_texture.h" #include "pipe/p_context.h" @@ -53,27 +54,26 @@ update_textures(struct st_context *st) for (unit = 0; unit < st->ctx->Const.MaxTextureCoordUnits; unit++) { const GLuint su = fprog->Base.SamplerUnits[unit]; struct gl_texture_object *texObj = st->ctx->Texture.Unit[su]._Current; - struct pipe_texture *pt; + struct st_texture_object *stObj = st_texture_object(texObj); if (texObj) { GLboolean flush, retval; retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); /* XXX retval indicates whether there's a texture border */ - - pt = st_get_texobj_texture(texObj); - } - else { - pt = NULL; } /* XXX: need to ensure that textures are unbound/removed from * this table before being deleted, otherwise the pointer * comparison below could fail. */ - if (st->state.sampler_texture[unit] != pt) { - st->state.sampler_texture[unit] = pt; - st->pipe->set_sampler_texture(st->pipe, unit, pt); + if (st->state.sampler_texture[unit] != stObj || + (stObj && stObj->dirtyData)) { + struct pipe_texture *pt = st_get_stobj_texture(stObj); + st->state.sampler_texture[unit] = stObj; + st->pipe->set_sampler_texture(st->pipe, unit, pt); + if (stObj) + stObj->dirtyData = GL_FALSE; } } } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index e2d4e06da1..585cae3743 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -726,7 +726,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); pipe->bind_fs_state(pipe, ctx->st->state.fs->data); pipe->bind_vs_state(pipe, ctx->st->state.vs->cso->data); - pipe->set_sampler_texture(pipe, unit, ctx->st->state.sampler_texture[unit]); + pipe->set_sampler_texture(pipe, unit, + st_get_stobj_texture(ctx->st->state.sampler_texture[unit])); pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]->data); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); } diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 4341623267..781425b546 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -48,6 +48,7 @@ #include "st_cb_texture.h" #include "st_format.h" #include "st_public.h" +#include "st_texture.h" diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 03dbb30b0f..7226b0dd98 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -53,33 +53,6 @@ #define DBG if (0) printf -struct st_texture_object -{ - struct gl_texture_object base; /* The "parent" object */ - - /* The texture must include at levels [0..lastLevel] once validated: - */ - GLuint lastLevel; - - /* On validation any active images held in main memory or in other - * textures will be copied to this texture and the old storage freed. - */ - struct pipe_texture *pt; - - GLboolean imageOverride; - GLint depthOverride; - GLuint pitchOverride; -}; - - - -static INLINE struct st_texture_object * -st_texture_object(struct gl_texture_object *obj) -{ - return (struct st_texture_object *) obj; -} - - static INLINE struct st_texture_image * st_texture_image(struct gl_texture_image *img) { @@ -87,14 +60,6 @@ st_texture_image(struct gl_texture_image *img) } -struct pipe_texture * -st_get_texobj_texture(struct gl_texture_object *texObj) -{ - struct st_texture_object *stObj = st_texture_object(texObj); - return stObj->pt; -} - - static enum pipe_texture_target gl_target_to_pipe(GLenum target) { @@ -725,6 +690,9 @@ st_TexImage(GLcontext * ctx, texImage->Data = NULL; } + /* flag data as dirty */ + stObj->dirtyData = GL_TRUE; + #if 01 if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); @@ -900,6 +868,7 @@ st_TexSubimage(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { + struct st_texture_object *stObj = st_texture_object(texObj); struct st_texture_image *stImage = st_texture_image(texImage); GLuint dstRowStride; GLuint srcImageStride = _mesa_image_image_stride(packing, width, height, @@ -961,6 +930,9 @@ st_TexSubimage(GLcontext * ctx, st_texture_image_unmap(stImage); texImage->Data = NULL; } + + /* flag data as dirty */ + stObj->dirtyData = GL_TRUE; } diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h index 878256ec26..843745fcd6 100644 --- a/src/mesa/state_tracker/st_cb_texture.h +++ b/src/mesa/state_tracker/st_cb_texture.h @@ -1,9 +1,33 @@ -#ifndef ST_CB_TEXTURE_H -#define ST_CB_TEXTURE_H +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ -extern struct pipe_texture * -st_get_texobj_texture(struct gl_texture_object *texObj); +#ifndef ST_CB_TEXTURE_H +#define ST_CB_TEXTURE_H extern GLboolean diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 59d1590f05..5be4769be4 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -59,26 +59,6 @@ struct st_tracked_state { -struct st_texture_image -{ - struct gl_texture_image base; - - /* These aren't stored in gl_texture_image - */ - GLuint level; - GLuint face; - - /* If stImage->pt != NULL, image data is stored here. - * Else if stImage->base.Data != NULL, image is stored there. - * Else there is no image data. - */ - struct pipe_texture *pt; - - struct pipe_surface *surface; -}; - - - struct st_context { GLcontext *ctx; @@ -106,7 +86,7 @@ struct st_context struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; struct pipe_framebuffer_state framebuffer; - struct pipe_texture *sampler_texture[PIPE_MAX_SAMPLERS]; + struct st_texture_object *sampler_texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_viewport_state viewport; diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 6c09b86033..a0b4b973aa 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -43,6 +43,7 @@ #include "st_draw.h" #include "st_gen_mipmap.h" #include "st_program.h" +#include "st_texture.h" #include "st_cb_drawpixels.h" #include "st_cb_texture.h" @@ -302,7 +303,8 @@ st_render_mipmap(struct st_context *st, pipe->bind_vs_state(pipe, st->state.vs->cso->data); if (st->state.sampler[0]) pipe->bind_sampler_state(pipe, 0, st->state.sampler[0]->data); - pipe->set_sampler_texture(pipe, 0, st->state.sampler_texture[0]); + pipe->set_sampler_texture(pipe, 0, + st_get_stobj_texture(st->state.sampler_texture[0])); pipe->set_viewport_state(pipe, &st->state.viewport); return TRUE; diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 72324cd9ab..78f5f451ed 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -35,6 +35,70 @@ struct pipe_context; struct pipe_texture; +struct st_texture_image +{ + struct gl_texture_image base; + + /* These aren't stored in gl_texture_image + */ + GLuint level; + GLuint face; + + /* If stImage->pt != NULL, image data is stored here. + * Else if stImage->base.Data != NULL, image is stored there. + * Else there is no image data. + */ + struct pipe_texture *pt; + + struct pipe_surface *surface; +}; + + + +struct st_texture_object +{ + struct gl_texture_object base; /* The "parent" object */ + + /* The texture must include at levels [0..lastLevel] once validated: + */ + GLuint lastLevel; + + /* On validation any active images held in main memory or in other + * textures will be copied to this texture and the old storage freed. + */ + struct pipe_texture *pt; + + GLboolean imageOverride; + GLint depthOverride; + GLuint pitchOverride; + + GLboolean dirtyData; +}; + + +static INLINE struct st_texture_object * +st_texture_object(struct gl_texture_object *obj) +{ + return (struct st_texture_object *) obj; +} + + +static INLINE struct pipe_texture * +st_get_texobj_texture(struct gl_texture_object *texObj) +{ + struct st_texture_object *stObj = st_texture_object(texObj); + return stObj ? stObj->pt : NULL; +} + + +static INLINE struct pipe_texture * +st_get_stobj_texture(struct st_texture_object *stObj) +{ + return stObj ? stObj->pt : NULL; +} + + + extern struct pipe_texture * st_texture_create(struct st_context *st, enum pipe_texture_target target, -- cgit v1.2.3 From 3ccbaa977f96eaa849093875dd0944f744ee1e21 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 13:11:51 -0700 Subject: gallium: re-enable GenerateMipmap calls --- src/mesa/state_tracker/st_cb_texture.c | 19 ++----------------- 1 file changed, 2 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 7226b0dd98..2e7d78e582 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -693,11 +693,9 @@ st_TexImage(GLcontext * ctx, /* flag data as dirty */ stObj->dirtyData = GL_TRUE; -#if 01 if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); } -#endif } @@ -915,14 +913,9 @@ st_TexSubimage(GLcontext * ctx, } } -#if 0 - /* GL_SGIS_generate_mipmap */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - _mesa_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } -#endif _mesa_unmap_teximage_pbo(ctx, packing); @@ -1184,17 +1177,9 @@ do_copy_texsubimage(GLcontext *ctx, pipe_surface_reference(&dest_surface, NULL); -#if 0 - /* GL_SGIS_generate_mipmap -- this can be accelerated now. - * XXX Add a ctx->Driver.GenerateMipmaps() function? - */ if (level == texObj->BaseLevel && texObj->GenerateMipmap) { - intel_generate_mipmap(ctx, target, - &ctx->Texture.Unit[ctx->Texture.CurrentUnit], - texObj); + ctx->Driver.GenerateMipmap(ctx, target, texObj); } -#endif - } -- cgit v1.2.3 From 1d45787d4a70c55e7fa899d13b9139430e2fa3e2 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 13:12:36 -0700 Subject: gallium: mark texture object data dirty in do_copy_texsubimage() --- src/mesa/state_tracker/st_cb_texture.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 2e7d78e582..778fb536bc 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1091,6 +1091,7 @@ do_copy_texsubimage(GLcontext *ctx, struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); struct st_texture_image *stImage = st_texture_image(texImage); + struct st_texture_object *stObj = st_texture_object(texObj); GLenum baseFormat = texImage->InternalFormat; struct gl_framebuffer *fb = ctx->ReadBuffer; struct st_renderbuffer *strb; @@ -1177,6 +1178,9 @@ do_copy_texsubimage(GLcontext *ctx, pipe_surface_reference(&dest_surface, NULL); + /* flag data as dirty */ + stObj->dirtyData = GL_TRUE; + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); } -- cgit v1.2.3 From d5640a2dbdc4454d0405f2cd5b18fc49b1ca7694 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 13:24:52 -0700 Subject: gallium: new pipe->texture_update() function Called whenever texture data is changed (glTexImage, glTexSubImage, glCopyTexSubImage, etc). --- src/gallium/drivers/cell/ppu/cell_context.c | 1 + src/gallium/drivers/cell/ppu/cell_texture.c | 8 +++++ src/gallium/drivers/cell/ppu/cell_texture.h | 3 ++ src/gallium/drivers/failover/fo_context.c | 9 +++-- src/gallium/drivers/i915simple/i915_context.c | 1 + src/gallium/drivers/i915simple/i915_texture.c | 7 ++++ src/gallium/drivers/i915simple/i915_texture.h | 4 +++ src/gallium/drivers/i965simple/brw_context.c | 1 + src/gallium/drivers/i965simple/brw_tex_layout.c | 8 +++++ src/gallium/drivers/i965simple/brw_tex_layout.h | 3 ++ src/gallium/drivers/softpipe/sp_context.c | 1 + src/gallium/drivers/softpipe/sp_texture.c | 15 +++++++++ src/gallium/drivers/softpipe/sp_texture.h | 4 +++ src/gallium/drivers/softpipe/sp_tile_cache.c | 45 ++++++++++++++----------- src/gallium/include/pipe/p_context.h | 8 +++++ src/mesa/state_tracker/st_atom_texture.c | 14 +++++--- 16 files changed, 104 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index e1eb22f468..b6ba14578c 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -244,6 +244,7 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) /* textures */ cell->pipe.texture_create = cell_texture_create; cell->pipe.texture_release = cell_texture_release; + cell->pipe.texture_update = cell_texture_update; cell->pipe.get_tex_surface = cell_get_tex_surface; cell->pipe.set_sampler_texture = cell_set_sampler_texture; diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index c8ef36002f..4629eb1320 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -128,6 +128,14 @@ cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } +void +cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) +{ + /* XXX TO DO: re-tile the texture data ... */ + +} + + /** * Called via pipe->get_tex_surface() */ diff --git a/src/gallium/drivers/cell/ppu/cell_texture.h b/src/gallium/drivers/cell/ppu/cell_texture.h index 0264fed88e..07e81582f4 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.h +++ b/src/gallium/drivers/cell/ppu/cell_texture.h @@ -67,6 +67,9 @@ cell_texture_create(struct pipe_context *pipe, extern void cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); +extern void +cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture); + extern struct pipe_surface * cell_get_tex_surface(struct pipe_context *pipe, struct pipe_texture *pt, diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c index 7ce4a7df17..156f7399b0 100644 --- a/src/gallium/drivers/failover/fo_context.c +++ b/src/gallium/drivers/failover/fo_context.c @@ -137,15 +137,14 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover_init_state_functions( failover ); -#if 0 - failover->pipe.surface_alloc = hw->surface_alloc; -#endif - failover->pipe.get_tex_surface = hw->get_tex_surface; - failover->pipe.surface_copy = hw->surface_copy; failover->pipe.surface_fill = hw->surface_fill; + failover->pipe.texture_create = hw->texture_create; failover->pipe.texture_release = hw->texture_release; + failover->pipe.texture_update = hw->texture_update; + failover->pipe.get_tex_surface = hw->get_tex_surface; + failover->pipe.flush = hw->flush; failover->dirty = 0; diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 7f71f8fd4f..97773f1256 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -302,6 +302,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->pipe.texture_create = i915_texture_create; i915->pipe.texture_release = i915_texture_release; + i915->pipe.texture_update = i915_texture_update; i915->dirty = ~0; i915->hardware_dirty = ~0; diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 6d37ae3d74..4ba76d19ad 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -534,3 +534,10 @@ i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } *pt = NULL; } + + +void +i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) +{ + /* no-op? */ +} diff --git a/src/gallium/drivers/i915simple/i915_texture.h b/src/gallium/drivers/i915simple/i915_texture.h index 330d111dc7..0312977552 100644 --- a/src/gallium/drivers/i915simple/i915_texture.h +++ b/src/gallium/drivers/i915simple/i915_texture.h @@ -14,4 +14,8 @@ extern void i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); +extern void +i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture); + + #endif /* I915_TEXTURE_H */ diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c index 5e58701e91..2e2380a8d6 100644 --- a/src/gallium/drivers/i965simple/brw_context.c +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -224,6 +224,7 @@ struct pipe_context *brw_create(struct pipe_winsys *pipe_winsys, brw->pipe.clear = brw_clear; brw->pipe.texture_create = brw_texture_create; brw->pipe.texture_release = brw_texture_release; + brw->pipe.texture_update = brw_texture_update; brw_init_surface_functions(brw); brw_init_state_functions(brw); diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 90561f1307..220591da9a 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -351,3 +351,11 @@ brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } *pt = NULL; } + + +void +brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) +{ + /* no-op? */ +} + diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.h b/src/gallium/drivers/i965simple/brw_tex_layout.h index cfd6b1ef3a..7d118d0fa8 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.h +++ b/src/gallium/drivers/i965simple/brw_tex_layout.h @@ -12,4 +12,7 @@ brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat extern void brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); +extern void +brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture); + #endif diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 254c6adca4..316020cba6 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -283,6 +283,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, /* textures */ softpipe->pipe.texture_create = softpipe_texture_create; softpipe->pipe.texture_release = softpipe_texture_release; + softpipe->pipe.texture_update = softpipe_texture_update; softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; /* diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 6de7a9b543..8f31f05e47 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -39,6 +39,7 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_texture.h" +#include "sp_tile_cache.h" /* Simple, maximally packed layout. @@ -128,6 +129,20 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } +void +softpipe_texture_update(struct pipe_context *pipe, + struct pipe_texture *texture) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + uint unit; + for (unit = 0; unit < PIPE_MAX_SAMPLERS; unit++) { + if (softpipe->texture[unit] == texture) { + sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); + } + } +} + + /** * Called via pipe->get_tex_surface() */ diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h index fa646c0de9..50fc100427 100644 --- a/src/gallium/drivers/softpipe/sp_texture.h +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -62,6 +62,10 @@ softpipe_texture_create(struct pipe_context *pipe, extern void softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); +extern void +softpipe_texture_update(struct pipe_context *pipe, + struct pipe_texture *texture); + extern struct pipe_surface * softpipe_get_tex_surface(struct pipe_context *pipe, struct pipe_texture *pt, diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 9ed3c5072d..da30dd6c48 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -359,30 +359,37 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, struct pipe_surface *ps = tc->surface; int inuse = 0, pos; - if (!ps || !ps->buffer) - return; - - for (pos = 0; pos < NUM_ENTRIES; pos++) { - struct softpipe_cached_tile *tile = tc->entries + pos; - if (tile->x >= 0) { - if (tc->depth_stencil) { - pipe_put_tile_raw(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); - } - else { - pipe_put_tile_rgba(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - (float *) tile->data.color); + if (ps && ps->buffer) { + /* caching a drawing surface */ + for (pos = 0; pos < NUM_ENTRIES; pos++) { + struct softpipe_cached_tile *tile = tc->entries + pos; + if (tile->x >= 0) { + if (tc->depth_stencil) { + pipe_put_tile_raw(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); + } + else { + pipe_put_tile_rgba(pipe, ps, + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + (float *) tile->data.color); + } + tile->x = tile->y = -1; /* mark as empty */ + inuse++; } - tile->x = tile->y = -1; /* mark as empty */ - inuse++; } - } #if TILE_CLEAR_OPTIMIZATION - sp_tile_cache_flush_clear(&softpipe->pipe, tc); + sp_tile_cache_flush_clear(&softpipe->pipe, tc); #endif + } + else if (tc->texture) { + /* caching a texture, mark all entries as embpy */ + for (pos = 0; pos < NUM_ENTRIES; pos++) { + tc->entries[pos].x = -1; + } + tc->tex_face = -1; + } #if 0 debug_printf("flushed tiles in use: %d\n", inuse); diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 39f95695fb..036c4c8964 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -206,6 +206,14 @@ struct pipe_context { void (*texture_release)(struct pipe_context *pipe, struct pipe_texture **pt); + /** + * Called when texture data is changed. + * Note: we could pass some hints about which mip levels or cube faces + * have changed... + */ + void (*texture_update)(struct pipe_context *pipe, + struct pipe_texture *texture); + /** Get a surface which is a "view" into a texture */ struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, struct pipe_texture *texture, diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 9fead7e314..a4ac726816 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -67,14 +67,20 @@ update_textures(struct st_context *st) * this table before being deleted, otherwise the pointer * comparison below could fail. */ - if (st->state.sampler_texture[unit] != stObj || - (stObj && stObj->dirtyData)) { + if (st->state.sampler_texture[unit] != stObj) { struct pipe_texture *pt = st_get_stobj_texture(stObj); st->state.sampler_texture[unit] = stObj; st->pipe->set_sampler_texture(st->pipe, unit, pt); - if (stObj) - stObj->dirtyData = GL_FALSE; } + + stObj = st->state.sampler_texture[unit]; + + if (stObj && stObj->dirtyData) { + struct pipe_texture *pt = st_get_stobj_texture(stObj); + st->pipe->texture_update(st->pipe, pt); + stObj->dirtyData = GL_FALSE; + } + } } -- cgit v1.2.3 From 882a4b505484a50f1ccc2cf3ae0c3a52d4ec1be3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 14:00:42 -0700 Subject: gallium: minor re-org of 915 surface/texture code --- src/gallium/drivers/i915simple/i915_context.c | 4 +- src/gallium/drivers/i915simple/i915_surface.c | 45 +-------------------- src/gallium/drivers/i915simple/i915_texture.c | 58 +++++++++++++++++++++++++-- src/gallium/drivers/i915simple/i915_texture.h | 37 ++++++++++++----- 4 files changed, 84 insertions(+), 60 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 97773f1256..acfa349439 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -296,13 +296,11 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915_init_state_functions(i915); i915_init_flush_functions(i915); i915_init_string_functions(i915); + i915_init_texture_functions(i915); i915->pci_id = pci_id; i915->flags.is_i945 = is_i945; - i915->pipe.texture_create = i915_texture_create; - i915->pipe.texture_release = i915_texture_release; - i915->pipe.texture_update = i915_texture_update; i915->dirty = ~0; i915->hardware_dirty = ~0; diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index 17fd27895a..f4fbedbe9b 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -36,48 +36,6 @@ #include "util/p_tile.h" -/* - * XXX note: same as code in sp_surface.c - */ -static struct pipe_surface * -i915_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - struct i915_texture *tex = (struct i915_texture *)pt; - struct pipe_surface *ps; - unsigned offset; /* in bytes */ - - offset = tex->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE) { - offset += tex->image_offset[level][face] * pt->cpp; - } - else if (pt->target == PIPE_TEXTURE_3D) { - offset += tex->image_offset[level][zslice] * pt->cpp; - } - else { - assert(face == 0); - assert(zslice == 0); - } - - ps = pipe->winsys->surface_alloc(pipe->winsys); - if (ps) { - assert(ps->refcount); - assert(ps->winsys); - pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); - ps->format = pt->format; - ps->cpp = pt->cpp; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->pitch = tex->pitch; - ps->offset = offset; - } - return ps; -} - - - /* Assumes all values are within bounds -- no checking at this level - * do it higher up if required. */ @@ -115,6 +73,7 @@ i915_surface_copy(struct pipe_context *pipe, } } + /* Fill a rectangular sub-region. Need better logic about when to * push buffers into AGP - will currently do so whenever possible. */ @@ -184,8 +143,6 @@ i915_surface_fill(struct pipe_context *pipe, void i915_init_surface_functions(struct i915_context *i915) { - i915->pipe.get_tex_surface = i915_get_tex_surface; - i915->pipe.surface_copy = i915_surface_copy; i915->pipe.surface_fill = i915_surface_fill; } diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 4ba76d19ad..b235fae96d 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -478,7 +478,7 @@ i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) } -struct pipe_texture * +static struct pipe_texture * i915_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) { @@ -506,7 +506,7 @@ i915_texture_create(struct pipe_context *pipe, } -void +static void i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) { if (!*pt) @@ -536,8 +536,60 @@ i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } -void +static void i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) { /* no-op? */ } + + +/* + * XXX note: same as code in sp_surface.c + */ +static struct pipe_surface * +i915_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct i915_texture *tex = (struct i915_texture *)pt; + struct pipe_surface *ps; + unsigned offset; /* in bytes */ + + offset = tex->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE) { + offset += tex->image_offset[level][face] * pt->cpp; + } + else if (pt->target == PIPE_TEXTURE_3D) { + offset += tex->image_offset[level][zslice] * pt->cpp; + } + else { + assert(face == 0); + assert(zslice == 0); + } + + ps = pipe->winsys->surface_alloc(pipe->winsys); + if (ps) { + assert(ps->refcount); + assert(ps->winsys); + pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = tex->pitch; + ps->offset = offset; + } + return ps; +} + + + +void +i915_init_texture_functions(struct i915_context *i915) +{ + i915->pipe.texture_create = i915_texture_create; + i915->pipe.texture_release = i915_texture_release; + i915->pipe.texture_update = i915_texture_update; + i915->pipe.get_tex_surface = i915_get_tex_surface; +} diff --git a/src/gallium/drivers/i915simple/i915_texture.h b/src/gallium/drivers/i915simple/i915_texture.h index 0312977552..6d8d41178f 100644 --- a/src/gallium/drivers/i915simple/i915_texture.h +++ b/src/gallium/drivers/i915simple/i915_texture.h @@ -1,21 +1,38 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ #ifndef I915_TEXTURE_H #define I915_TEXTURE_H struct pipe_context; -struct pipe_texture; - - -struct pipe_texture * -i915_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat); - -extern void -i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); extern void -i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture); +i915_init_texture_functions(struct i915_context *i915); #endif /* I915_TEXTURE_H */ -- cgit v1.2.3 From 4eae65c8e052976a130564560699e60e1a3a9cc3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 14:04:05 -0700 Subject: gallium: re-org of i965 texture/surface code, functions --- src/gallium/drivers/i965simple/brw_context.c | 4 +- src/gallium/drivers/i965simple/brw_surface.c | 43 +----------------- src/gallium/drivers/i965simple/brw_tex_layout.c | 60 +++++++++++++++++++++++-- src/gallium/drivers/i965simple/brw_tex_layout.h | 10 +---- 4 files changed, 60 insertions(+), 57 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c index 2e2380a8d6..6fb840708e 100644 --- a/src/gallium/drivers/i965simple/brw_context.c +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -222,11 +222,9 @@ struct pipe_context *brw_create(struct pipe_winsys *pipe_winsys, brw->pipe.get_param = brw_get_param; brw->pipe.get_paramf = brw_get_paramf; brw->pipe.clear = brw_clear; - brw->pipe.texture_create = brw_texture_create; - brw->pipe.texture_release = brw_texture_release; - brw->pipe.texture_update = brw_texture_update; brw_init_surface_functions(brw); + brw_init_texture_functions(brw); brw_init_state_functions(brw); brw_init_flush_functions(brw); brw_init_string_functions(brw); diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index 376a42b1a6..dc4846d39f 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -35,47 +35,6 @@ #include "util/p_tile.h" -/* - * XXX note: same as code in sp_surface.c - */ -static struct pipe_surface * -brw_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - struct brw_texture *tex = (struct brw_texture *)pt; - struct pipe_surface *ps; - unsigned offset; /* in bytes */ - - offset = tex->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE) { - offset += tex->image_offset[level][face] * pt->cpp; - } - else if (pt->target == PIPE_TEXTURE_3D) { - offset += tex->image_offset[level][zslice] * pt->cpp; - } - else { - assert(face == 0); - assert(zslice == 0); - } - - ps = pipe->winsys->surface_alloc(pipe->winsys); - if (ps) { - assert(ps->format); - assert(ps->refcount); - pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); - ps->format = pt->format; - ps->cpp = pt->cpp; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->pitch = tex->pitch; - ps->offset = offset; - } - return ps; -} - - /* Upload data to a rectangular sub-region. Lots of choices how to do this: * * - memcpy by span to current destination @@ -201,10 +160,10 @@ brw_surface_fill(struct pipe_context *pipe, } } + void brw_init_surface_functions(struct brw_context *brw) { - brw->pipe.get_tex_surface = brw_get_tex_surface; brw->pipe.surface_copy = brw_surface_copy; brw->pipe.surface_fill = brw_surface_fill; } diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 220591da9a..043a2ff9a4 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -300,8 +300,9 @@ static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture } -struct pipe_texture * -brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) +static struct pipe_texture * +brw_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat) { struct brw_texture *tex = CALLOC_STRUCT(brw_texture); @@ -323,7 +324,8 @@ brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat return &tex->base; } -void + +static void brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) { if (!*pt) @@ -353,9 +355,59 @@ brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } -void +static void brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) { /* no-op? */ } + +/* + * XXX note: same as code in sp_surface.c + */ +static struct pipe_surface * +brw_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct brw_texture *tex = (struct brw_texture *)pt; + struct pipe_surface *ps; + unsigned offset; /* in bytes */ + + offset = tex->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE) { + offset += tex->image_offset[level][face] * pt->cpp; + } + else if (pt->target == PIPE_TEXTURE_3D) { + offset += tex->image_offset[level][zslice] * pt->cpp; + } + else { + assert(face == 0); + assert(zslice == 0); + } + + ps = pipe->winsys->surface_alloc(pipe->winsys); + if (ps) { + assert(ps->format); + assert(ps->refcount); + pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = tex->pitch; + ps->offset = offset; + } + return ps; +} + + +void +brw_init_texture_functions(struct brw_context *brw) +{ + brw->pipe.texture_create = brw_texture_create; + brw->pipe.texture_release = brw_texture_release; + brw->pipe.texture_update = brw_texture_update; + brw->pipe.get_tex_surface = brw_get_tex_surface; +} diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.h b/src/gallium/drivers/i965simple/brw_tex_layout.h index 7d118d0fa8..ed49baeef8 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.h +++ b/src/gallium/drivers/i965simple/brw_tex_layout.h @@ -1,18 +1,12 @@ #ifndef BRW_TEX_LAYOUT_H #define BRW_TEX_LAYOUT_H -#include "pipe/p_compiler.h" -struct pipe_context; -struct pipe_texture; +struct brw_context; -extern struct pipe_texture * -brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat); extern void -brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); +brw_init_texture_functions(struct brw_context *brw); -extern void -brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture); #endif -- cgit v1.2.3 From 9171e63f414866aef155b17d3c85c9a236a872d6 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 14:21:45 -0700 Subject: cell: put most simple state-setter functions in new cell_pipe_state.c file Also, re-org of texture/surface functions. --- src/gallium/drivers/cell/ppu/Makefile | 6 +- src/gallium/drivers/cell/ppu/cell_context.c | 35 +-- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 323 +++++++++++++++++++++++++ src/gallium/drivers/cell/ppu/cell_pipe_state.h | 39 +++ src/gallium/drivers/cell/ppu/cell_state.h | 69 +----- src/gallium/drivers/cell/ppu/cell_texture.c | 21 +- src/gallium/drivers/cell/ppu/cell_texture.h | 20 +- 7 files changed, 389 insertions(+), 124 deletions(-) create mode 100644 src/gallium/drivers/cell/ppu/cell_pipe_state.c create mode 100644 src/gallium/drivers/cell/ppu/cell_pipe_state.h (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile index 3c3f622a2f..6b6dfca890 100644 --- a/src/gallium/drivers/cell/ppu/Makefile +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -25,14 +25,10 @@ SOURCES = \ cell_context.c \ cell_draw_arrays.c \ cell_flush.c \ - cell_state_blend.c \ - cell_state_clip.c \ cell_state_derived.c \ cell_state_emit.c \ cell_state_fs.c \ - cell_state_rasterizer.c \ - cell_state_sampler.c \ - cell_state_surface.c \ + cell_pipe_state.c \ cell_state_vertex.c \ cell_spu.c \ cell_surface.c \ diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index b6ba14578c..9f0ecb2be8 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -48,6 +48,7 @@ #include "cell_state.h" #include "cell_surface.h" #include "cell_spu.h" +#include "cell_pipe_state.h" #include "cell_texture.h" #include "cell_vbuf.h" @@ -198,22 +199,6 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) /* state setters */ - cell->pipe.create_blend_state = cell_create_blend_state; - cell->pipe.bind_blend_state = cell_bind_blend_state; - cell->pipe.delete_blend_state = cell_delete_blend_state; - - cell->pipe.create_sampler_state = cell_create_sampler_state; - cell->pipe.bind_sampler_state = cell_bind_sampler_state; - cell->pipe.delete_sampler_state = cell_delete_sampler_state; - - cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state; - cell->pipe.bind_depth_stencil_alpha_state = cell_bind_depth_stencil_alpha_state; - cell->pipe.delete_depth_stencil_alpha_state = cell_delete_depth_stencil_alpha_state; - - cell->pipe.create_rasterizer_state = cell_create_rasterizer_state; - cell->pipe.bind_rasterizer_state = cell_bind_rasterizer_state; - cell->pipe.delete_rasterizer_state = cell_delete_rasterizer_state; - cell->pipe.create_fs_state = cell_create_fs_state; cell->pipe.bind_fs_state = cell_bind_fs_state; cell->pipe.delete_fs_state = cell_delete_fs_state; @@ -222,16 +207,8 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->pipe.bind_vs_state = cell_bind_vs_state; cell->pipe.delete_vs_state = cell_delete_vs_state; - cell->pipe.set_blend_color = cell_set_blend_color; - cell->pipe.set_clip_state = cell_set_clip_state; cell->pipe.set_constant_buffer = cell_set_constant_buffer; - cell->pipe.set_framebuffer_state = cell_set_framebuffer_state; - - cell->pipe.set_polygon_stipple = cell_set_polygon_stipple; - cell->pipe.set_scissor_state = cell_set_scissor_state; - cell->pipe.set_viewport_state = cell_set_viewport_state; - cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; cell->pipe.set_vertex_element = cell_set_vertex_element; @@ -241,21 +218,15 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) cell->pipe.clear = cell_clear_surface; cell->pipe.flush = cell_flush; - /* textures */ - cell->pipe.texture_create = cell_texture_create; - cell->pipe.texture_release = cell_texture_release; - cell->pipe.texture_update = cell_texture_update; - cell->pipe.get_tex_surface = cell_get_tex_surface; - - cell->pipe.set_sampler_texture = cell_set_sampler_texture; - #if 0 cell->pipe.begin_query = cell_begin_query; cell->pipe.end_query = cell_end_query; cell->pipe.wait_query = cell_wait_query; #endif + cell_init_state_functions(cell); cell_init_surface_functions(cell); + cell_init_texture_functions(cell); cell->draw = cell_draw_create(cell); diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c new file mode 100644 index 0000000000..aef50725e8 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -0,0 +1,323 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "draw/draw_context.h" +#include "cell_context.h" +#include "cell_state.h" +#include "cell_texture.h" + + + +static void * +cell_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + return mem_dup(blend, sizeof(*blend)); +} + + +static void +cell_bind_blend_state(struct pipe_context *pipe, void *blend) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + cell->blend = (const struct pipe_blend_state *)blend; + + cell->dirty |= CELL_NEW_BLEND; +} + + +static void +cell_delete_blend_state(struct pipe_context *pipe, void *blend) +{ + FREE(blend); +} + + +static void +cell_set_blend_color(struct pipe_context *pipe, + const struct pipe_blend_color *blend_color) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + cell->blend_color = *blend_color; + + cell->dirty |= CELL_NEW_BLEND; +} + + + + +static void * +cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil) +{ + return mem_dup(depth_stencil, sizeof(*depth_stencil)); +} + + +static void +cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + cell->depth_stencil + = (const struct pipe_depth_stencil_alpha_state *) depth_stencil; + + cell->dirty |= CELL_NEW_DEPTH_STENCIL; +} + + +static void +cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth) +{ + FREE(depth); +} + + +static void cell_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct cell_context *cell = cell_context(pipe); + + /* pass the clip state to the draw module */ + draw_set_clip_state(cell->draw, clip); +} + + + +/* Called when driver state tracker notices changes to the viewport + * matrix: + */ +static void +cell_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct cell_context *cell = cell_context(pipe); + + cell->viewport = *viewport; /* struct copy */ + cell->dirty |= CELL_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + draw_set_viewport_state(cell->draw, viewport); + + /* Using tnl/ and vf/ modules is temporary while getting started. + * Full pipe will have vertex shader, vertex fetch of its own. + */ +} + + +static void +cell_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct cell_context *cell = cell_context(pipe); + + memcpy( &cell->scissor, scissor, sizeof(*scissor) ); + cell->dirty |= CELL_NEW_SCISSOR; +} + + +static void +cell_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ + struct cell_context *cell = cell_context(pipe); + + memcpy( &cell->poly_stipple, stipple, sizeof(*stipple) ); + cell->dirty |= CELL_NEW_STIPPLE; +} + + + +static void * +cell_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *setup) +{ + struct pipe_rasterizer_state *state + = MALLOC(sizeof(struct pipe_rasterizer_state)); + memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); + return state; +} + + +static void +cell_bind_rasterizer_state(struct pipe_context *pipe, void *setup) +{ + struct cell_context *cell = cell_context(pipe); + + /* pass-through to draw module */ + draw_set_rasterizer_state(cell->draw, setup); + + cell->rasterizer = (struct pipe_rasterizer_state *)setup; + + cell->dirty |= CELL_NEW_RASTERIZER; +} + + +static void +cell_delete_rasterizer_state(struct pipe_context *pipe, void *rasterizer) +{ + FREE(rasterizer); +} + + + +static void * +cell_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + return mem_dup(sampler, sizeof(*sampler)); +} + + +static void +cell_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + assert(unit < PIPE_MAX_SAMPLERS); + cell->sampler[unit] = (struct pipe_sampler_state *)sampler; + + cell->dirty |= CELL_NEW_SAMPLER; +} + + +static void +cell_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + FREE( sampler ); +} + + + +static void +cell_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, + struct pipe_texture *texture) +{ + struct cell_context *cell = cell_context(pipe); + + draw_flush(cell->draw); + + cell->texture[sampler] = cell_texture(texture); + + cell_update_texture_mapping(cell); + + cell->dirty |= CELL_NEW_TEXTURE; +} + + + +static void +cell_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct cell_context *cell = cell_context(pipe); + + if (1 /*memcmp(&cell->framebuffer, fb, sizeof(*fb))*/) { + struct pipe_surface *csurf = fb->cbufs[0]; + struct pipe_surface *zsurf = fb->zsbuf; + uint i; + + /* unmap old surfaces */ + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { + if (cell->framebuffer.cbufs[i] && cell->cbuf_map[i]) { + pipe_surface_unmap(cell->framebuffer.cbufs[i]); + cell->cbuf_map[i] = NULL; + } + } + + if (cell->framebuffer.zsbuf && cell->zsbuf_map) { + pipe_surface_unmap(cell->framebuffer.zsbuf); + cell->zsbuf_map = NULL; + } + + /* update my state */ + cell->framebuffer = *fb; + + /* map new surfaces */ + if (csurf) + cell->cbuf_map[0] = pipe_surface_map(csurf); + + if (zsurf) + cell->zsbuf_map = pipe_surface_map(zsurf); + + cell->dirty |= CELL_NEW_FRAMEBUFFER; + } +} + + + +void +cell_init_state_functions(struct cell_context *cell) +{ + cell->pipe.create_blend_state = cell_create_blend_state; + cell->pipe.bind_blend_state = cell_bind_blend_state; + cell->pipe.delete_blend_state = cell_delete_blend_state; + + cell->pipe.create_sampler_state = cell_create_sampler_state; + cell->pipe.bind_sampler_state = cell_bind_sampler_state; + cell->pipe.delete_sampler_state = cell_delete_sampler_state; + + cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state; + cell->pipe.bind_depth_stencil_alpha_state = cell_bind_depth_stencil_alpha_state; + cell->pipe.delete_depth_stencil_alpha_state = cell_delete_depth_stencil_alpha_state; + + cell->pipe.create_rasterizer_state = cell_create_rasterizer_state; + cell->pipe.bind_rasterizer_state = cell_bind_rasterizer_state; + cell->pipe.delete_rasterizer_state = cell_delete_rasterizer_state; + + cell->pipe.set_blend_color = cell_set_blend_color; + cell->pipe.set_clip_state = cell_set_clip_state; + cell->pipe.set_constant_buffer = cell_set_constant_buffer; + + cell->pipe.set_framebuffer_state = cell_set_framebuffer_state; + + cell->pipe.set_polygon_stipple = cell_set_polygon_stipple; + cell->pipe.set_scissor_state = cell_set_scissor_state; + cell->pipe.set_viewport_state = cell_set_viewport_state; +} + diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.h b/src/gallium/drivers/cell/ppu/cell_pipe_state.h new file mode 100644 index 0000000000..1889bd52ff --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.h @@ -0,0 +1,39 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_PIPE_STATE_H +#define CELL_PIPE_STATE_H + + +struct cell_context; + +extern void +cell_init_state_functions(struct cell_context *cell); + + +#endif /* CELL_PIPE_STATE_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_state.h b/src/gallium/drivers/cell/ppu/cell_state.h index 3a71ba14fa..2af7770ebb 100644 --- a/src/gallium/drivers/cell/ppu/cell_state.h +++ b/src/gallium/drivers/cell/ppu/cell_state.h @@ -23,42 +23,6 @@ -extern void -cell_set_framebuffer_state( struct pipe_context *, - const struct pipe_framebuffer_state * ); - - - -extern void * -cell_create_blend_state(struct pipe_context *, const struct pipe_blend_state *); -extern void cell_bind_blend_state(struct pipe_context *, void *); -extern void cell_delete_blend_state(struct pipe_context *, void *); - -extern void cell_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ); - - -void * -cell_create_sampler_state(struct pipe_context *, - const struct pipe_sampler_state *); - -extern void -cell_bind_sampler_state(struct pipe_context *, unsigned, void *); - -extern void -cell_delete_sampler_state(struct pipe_context *, void *); - - -extern void * -cell_create_depth_stencil_alpha_state(struct pipe_context *, - const struct pipe_depth_stencil_alpha_state *); - -extern void -cell_bind_depth_stencil_alpha_state(struct pipe_context *, void *); - -extern void -cell_delete_depth_stencil_alpha_state(struct pipe_context *, void *); - void *cell_create_fs_state(struct pipe_context *, const struct pipe_shader_state *); @@ -69,34 +33,11 @@ void *cell_create_vs_state(struct pipe_context *, void cell_bind_vs_state(struct pipe_context *, void *); void cell_delete_vs_state(struct pipe_context *, void *); - -void * -cell_create_rasterizer_state(struct pipe_context *, - const struct pipe_rasterizer_state *); -void cell_bind_rasterizer_state(struct pipe_context *, void *); -void cell_delete_rasterizer_state(struct pipe_context *, void *); - - -void cell_set_clip_state( struct pipe_context *, - const struct pipe_clip_state * ); - -void cell_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf); - -void cell_set_polygon_stipple( struct pipe_context *, - const struct pipe_poly_stipple * ); - void -cell_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, - struct pipe_texture *texture); - -void cell_set_scissor_state( struct pipe_context *, - const struct pipe_scissor_state * ); +cell_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf); -void cell_set_texture_state( struct pipe_context *, - unsigned unit, struct pipe_texture * ); void cell_set_vertex_element(struct pipe_context *, unsigned index, @@ -106,10 +47,6 @@ void cell_set_vertex_buffer(struct pipe_context *, unsigned index, const struct pipe_vertex_buffer *); -void cell_set_viewport_state( struct pipe_context *, - const struct pipe_viewport_state * ); - - void cell_update_derived( struct cell_context *softpipe ); #endif diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index 4629eb1320..e1b91075b2 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -79,8 +79,9 @@ cell_texture_layout(struct cell_texture * spt) } -struct pipe_texture * -cell_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) +static struct pipe_texture * +cell_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat) { struct cell_texture *spt = CALLOC_STRUCT(cell_texture); if (!spt) @@ -103,7 +104,7 @@ cell_texture_create(struct pipe_context *pipe, const struct pipe_texture *templa } -void +static void cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) { if (!*pt) @@ -128,7 +129,7 @@ cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } -void +static void cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) { /* XXX TO DO: re-tile the texture data ... */ @@ -139,7 +140,7 @@ cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) /** * Called via pipe->get_tex_surface() */ -struct pipe_surface * +static struct pipe_surface * cell_get_tex_surface(struct pipe_context *pipe, struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice) @@ -258,3 +259,13 @@ cell_update_texture_mapping(struct cell_context *cell) cell->tex_map = pipe_surface_map(cell->tex_surf); #endif } + + +void +cell_init_texture_functions(struct cell_context *cell) +{ + cell->pipe.texture_create = cell_texture_create; + cell->pipe.texture_release = cell_texture_release; + cell->pipe.texture_update = cell_texture_update; + cell->pipe.get_tex_surface = cell_get_tex_surface; +} diff --git a/src/gallium/drivers/cell/ppu/cell_texture.h b/src/gallium/drivers/cell/ppu/cell_texture.h index 07e81582f4..824fb3e20f 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.h +++ b/src/gallium/drivers/cell/ppu/cell_texture.h @@ -29,7 +29,7 @@ #define CELL_TEXTURE_H -struct pipe_context; +struct cell_context; struct pipe_texture; @@ -60,24 +60,12 @@ cell_texture(struct pipe_texture *pt) -extern struct pipe_texture * -cell_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat); - -extern void -cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); - extern void -cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture); - -extern struct pipe_surface * -cell_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice); +cell_update_texture_mapping(struct cell_context *cell); extern void -cell_update_texture_mapping(struct cell_context *cell); +cell_init_texture_functions(struct cell_context *cell); -#endif /* CELL_TEXTURE */ +#endif /* CELL_TEXTURE_H */ -- cgit v1.2.3 From acd2253ae80d133bc84a5e78909ec72464e3f901 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 14:24:46 -0700 Subject: Remove obsolete files replaced by cell_pipe_state.c --- src/gallium/drivers/cell/ppu/cell_state_blend.c | 109 --------------------- src/gallium/drivers/cell/ppu/cell_state_clip.c | 84 ---------------- .../drivers/cell/ppu/cell_state_rasterizer.c | 106 -------------------- src/gallium/drivers/cell/ppu/cell_state_sampler.c | 84 ---------------- src/gallium/drivers/cell/ppu/cell_state_surface.c | 71 -------------- 5 files changed, 454 deletions(-) delete mode 100644 src/gallium/drivers/cell/ppu/cell_state_blend.c delete mode 100644 src/gallium/drivers/cell/ppu/cell_state_clip.c delete mode 100644 src/gallium/drivers/cell/ppu/cell_state_rasterizer.c delete mode 100644 src/gallium/drivers/cell/ppu/cell_state_sampler.c delete mode 100644 src/gallium/drivers/cell/ppu/cell_state_surface.c (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_blend.c b/src/gallium/drivers/cell/ppu/cell_state_blend.c deleted file mode 100644 index b6d6d71f0c..0000000000 --- a/src/gallium/drivers/cell/ppu/cell_state_blend.c +++ /dev/null @@ -1,109 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "draw/draw_context.h" -#include "cell_context.h" -#include "cell_state.h" - - - -void * -cell_create_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend) -{ - return mem_dup(blend, sizeof(*blend)); -} - - -void -cell_bind_blend_state(struct pipe_context *pipe, void *blend) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - cell->blend = (const struct pipe_blend_state *)blend; - - cell->dirty |= CELL_NEW_BLEND; -} - - -void -cell_delete_blend_state(struct pipe_context *pipe, void *blend) -{ - FREE(blend); -} - - -void -cell_set_blend_color(struct pipe_context *pipe, - const struct pipe_blend_color *blend_color) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - cell->blend_color = *blend_color; - - cell->dirty |= CELL_NEW_BLEND; -} - - - - -void * -cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) -{ - return mem_dup(depth_stencil, sizeof(*depth_stencil)); -} - - -void -cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, - void *depth_stencil) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - cell->depth_stencil - = (const struct pipe_depth_stencil_alpha_state *) depth_stencil; - - cell->dirty |= CELL_NEW_DEPTH_STENCIL; -} - - -void -cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth) -{ - FREE(depth); -} diff --git a/src/gallium/drivers/cell/ppu/cell_state_clip.c b/src/gallium/drivers/cell/ppu/cell_state_clip.c deleted file mode 100644 index 0482f87e88..0000000000 --- a/src/gallium/drivers/cell/ppu/cell_state_clip.c +++ /dev/null @@ -1,84 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "cell_context.h" -#include "cell_state.h" -#include "draw/draw_context.h" - - -void cell_set_clip_state( struct pipe_context *pipe, - const struct pipe_clip_state *clip ) -{ - struct cell_context *cell = cell_context(pipe); - - /* pass the clip state to the draw module */ - draw_set_clip_state(cell->draw, clip); -} - - - -/* Called when driver state tracker notices changes to the viewport - * matrix: - */ -void cell_set_viewport_state( struct pipe_context *pipe, - const struct pipe_viewport_state *viewport ) -{ - struct cell_context *cell = cell_context(pipe); - - cell->viewport = *viewport; /* struct copy */ - cell->dirty |= CELL_NEW_VIEWPORT; - - /* pass the viewport info to the draw module */ - draw_set_viewport_state(cell->draw, viewport); - - /* Using tnl/ and vf/ modules is temporary while getting started. - * Full pipe will have vertex shader, vertex fetch of its own. - */ -} - - -void cell_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct cell_context *cell = cell_context(pipe); - - memcpy( &cell->scissor, scissor, sizeof(*scissor) ); - cell->dirty |= CELL_NEW_SCISSOR; -} - - -void cell_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ - struct cell_context *cell = cell_context(pipe); - - memcpy( &cell->poly_stipple, stipple, sizeof(*stipple) ); - cell->dirty |= CELL_NEW_STIPPLE; -} diff --git a/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c b/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c deleted file mode 100644 index 7eca5b5765..0000000000 --- a/src/gallium/drivers/cell/ppu/cell_state_rasterizer.c +++ /dev/null @@ -1,106 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "draw/draw_context.h" -#include "cell_context.h" -#include "cell_state.h" - - - -struct spu_rasterizer_state -{ - unsigned flatshade:1; -#if 0 - unsigned light_twoside:1; - unsigned front_winding:2; /**< PIPE_WINDING_x */ - unsigned cull_mode:2; /**< PIPE_WINDING_x */ - unsigned fill_cw:2; /**< PIPE_POLYGON_MODE_x */ - unsigned fill_ccw:2; /**< PIPE_POLYGON_MODE_x */ - unsigned offset_cw:1; - unsigned offset_ccw:1; -#endif - unsigned scissor:1; - unsigned poly_smooth:1; - unsigned poly_stipple_enable:1; - unsigned point_smooth:1; -#if 0 - unsigned point_sprite:1; - unsigned point_size_per_vertex:1; /**< size computed in vertex shader */ -#endif - unsigned multisample:1; /* XXX maybe more ms state in future */ - unsigned line_smooth:1; - unsigned line_stipple_enable:1; - unsigned line_stipple_factor:8; /**< [1..256] actually */ - unsigned line_stipple_pattern:16; -#if 0 - unsigned bypass_clipping:1; -#endif - unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ - - float line_width; - float point_size; /**< used when no per-vertex size */ -#if 0 - float offset_units; - float offset_scale; - ubyte sprite_coord_mode[PIPE_MAX_SHADER_OUTPUTS]; /**< PIPE_SPRITE_COORD_ */ -#endif -}; - - - -void * -cell_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *setup) -{ - struct pipe_rasterizer_state *state - = MALLOC(sizeof(struct pipe_rasterizer_state)); - memcpy(state, setup, sizeof(struct pipe_rasterizer_state)); - return state; -} - - -void -cell_bind_rasterizer_state(struct pipe_context *pipe, void *setup) -{ - struct cell_context *cell = cell_context(pipe); - - /* pass-through to draw module */ - draw_set_rasterizer_state(cell->draw, setup); - - cell->rasterizer = (struct pipe_rasterizer_state *)setup; - - cell->dirty |= CELL_NEW_RASTERIZER; -} - - -void -cell_delete_rasterizer_state(struct pipe_context *pipe, void *rasterizer) -{ - FREE(rasterizer); -} diff --git a/src/gallium/drivers/cell/ppu/cell_state_sampler.c b/src/gallium/drivers/cell/ppu/cell_state_sampler.c deleted file mode 100644 index a33421a4ad..0000000000 --- a/src/gallium/drivers/cell/ppu/cell_state_sampler.c +++ /dev/null @@ -1,84 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: - * Brian Paul - */ - -#include "pipe/p_util.h" -#include "draw/draw_context.h" -#include "cell_context.h" -#include "cell_state.h" -#include "cell_texture.h" - - -void * -cell_create_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) -{ - return mem_dup(sampler, sizeof(*sampler)); -} - -void -cell_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - assert(unit < PIPE_MAX_SAMPLERS); - cell->sampler[unit] = (struct pipe_sampler_state *)sampler; - - cell->dirty |= CELL_NEW_SAMPLER; -} - - -void -cell_delete_sampler_state(struct pipe_context *pipe, - void *sampler) -{ - FREE( sampler ); -} - - - -void -cell_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, - struct pipe_texture *texture) -{ - struct cell_context *cell = cell_context(pipe); - - draw_flush(cell->draw); - - cell->texture[sampler] = texture; - - cell_update_texture_mapping(cell); - - cell->dirty |= CELL_NEW_TEXTURE; -} diff --git a/src/gallium/drivers/cell/ppu/cell_state_surface.c b/src/gallium/drivers/cell/ppu/cell_state_surface.c deleted file mode 100644 index 287610b76b..0000000000 --- a/src/gallium/drivers/cell/ppu/cell_state_surface.c +++ /dev/null @@ -1,71 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "pipe/p_inlines.h" -#include "cell_context.h" -#include "cell_state.h" - - -void -cell_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb) -{ - struct cell_context *cell = cell_context(pipe); - - if (1 /*memcmp(&cell->framebuffer, fb, sizeof(*fb))*/) { - struct pipe_surface *csurf = fb->cbufs[0]; - struct pipe_surface *zsurf = fb->zsbuf; - uint i; - - /* unmap old surfaces */ - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { - if (cell->framebuffer.cbufs[i] && cell->cbuf_map[i]) { - pipe_surface_unmap(cell->framebuffer.cbufs[i]); - cell->cbuf_map[i] = NULL; - } - } - - if (cell->framebuffer.zsbuf && cell->zsbuf_map) { - pipe_surface_unmap(cell->framebuffer.zsbuf); - cell->zsbuf_map = NULL; - } - - /* update my state */ - cell->framebuffer = *fb; - - /* map new surfaces */ - if (csurf) - cell->cbuf_map[0] = pipe_surface_map(csurf); - - if (zsurf) - cell->zsbuf_map = pipe_surface_map(zsurf); - - cell->dirty |= CELL_NEW_FRAMEBUFFER; - } -} - -- cgit v1.2.3 From f6e1654e22d983ec9015fe4e7445e03df1227c71 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 14:25:07 -0700 Subject: cell: plug in cell_set_sampler_texture --- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index aef50725e8..c5ddf6a09e 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -302,6 +302,8 @@ cell_init_state_functions(struct cell_context *cell) cell->pipe.bind_sampler_state = cell_bind_sampler_state; cell->pipe.delete_sampler_state = cell_delete_sampler_state; + cell->pipe.set_sampler_texture = cell_set_sampler_texture; + cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state; cell->pipe.bind_depth_stencil_alpha_state = cell_bind_depth_stencil_alpha_state; cell->pipe.delete_depth_stencil_alpha_state = cell_delete_depth_stencil_alpha_state; -- cgit v1.2.3 From 9e57e70b42471ff587441fb8c1b5de728521fafd Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 14:27:08 -0700 Subject: cell: #includes to silence warnings --- src/gallium/drivers/cell/ppu/cell_clear.c | 1 + src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_clear.c b/src/gallium/drivers/cell/ppu/cell_clear.c index e588a30d5b..3ffe09add6 100644 --- a/src/gallium/drivers/cell/ppu/cell_clear.c +++ b/src/gallium/drivers/cell/ppu/cell_clear.c @@ -41,6 +41,7 @@ #include "cell_batch.h" #include "cell_flush.h" #include "cell_spu.h" +#include "cell_state.h" void diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index 64c7821c19..f7ef72e5a2 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -35,6 +35,7 @@ #include "cell_context.h" #include "cell_draw_arrays.h" +#include "cell_flush.h" #include "cell_spu.h" #include "cell_batch.h" -- cgit v1.2.3 From 64683473753de6eb978245e348e9b20cd1d42883 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 14:30:50 -0700 Subject: cell: init shader-related functions in cell_init_shader_functions() --- src/gallium/drivers/cell/ppu/cell_context.c | 11 +----- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 1 - src/gallium/drivers/cell/ppu/cell_state.h | 50 ++++++++++++++++---------- src/gallium/drivers/cell/ppu/cell_state_fs.c | 29 +++++++++++---- 4 files changed, 55 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index 9f0ecb2be8..98c314f45c 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -199,16 +199,6 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) /* state setters */ - cell->pipe.create_fs_state = cell_create_fs_state; - cell->pipe.bind_fs_state = cell_bind_fs_state; - cell->pipe.delete_fs_state = cell_delete_fs_state; - - cell->pipe.create_vs_state = cell_create_vs_state; - cell->pipe.bind_vs_state = cell_bind_vs_state; - cell->pipe.delete_vs_state = cell_delete_vs_state; - - cell->pipe.set_constant_buffer = cell_set_constant_buffer; - cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; cell->pipe.set_vertex_element = cell_set_vertex_element; @@ -225,6 +215,7 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) #endif cell_init_state_functions(cell); + cell_init_shader_functions(cell); cell_init_surface_functions(cell); cell_init_texture_functions(cell); diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index c5ddf6a09e..35e88f79b1 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -314,7 +314,6 @@ cell_init_state_functions(struct cell_context *cell) cell->pipe.set_blend_color = cell_set_blend_color; cell->pipe.set_clip_state = cell_set_clip_state; - cell->pipe.set_constant_buffer = cell_set_constant_buffer; cell->pipe.set_framebuffer_state = cell_set_framebuffer_state; diff --git a/src/gallium/drivers/cell/ppu/cell_state.h b/src/gallium/drivers/cell/ppu/cell_state.h index 2af7770ebb..31ce505e21 100644 --- a/src/gallium/drivers/cell/ppu/cell_state.h +++ b/src/gallium/drivers/cell/ppu/cell_state.h @@ -1,3 +1,29 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ #ifndef CELL_STATE_H @@ -22,23 +48,6 @@ #define CELL_NEW_VERTEX_INFO 0x8000 - - -void *cell_create_fs_state(struct pipe_context *, - const struct pipe_shader_state *); -void cell_bind_fs_state(struct pipe_context *, void *); -void cell_delete_fs_state(struct pipe_context *, void *); -void *cell_create_vs_state(struct pipe_context *, - const struct pipe_shader_state *); -void cell_bind_vs_state(struct pipe_context *, void *); -void cell_delete_vs_state(struct pipe_context *, void *); - -void -cell_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf); - - void cell_set_vertex_element(struct pipe_context *, unsigned index, const struct pipe_vertex_element *); @@ -49,4 +58,9 @@ void cell_set_vertex_buffer(struct pipe_context *, void cell_update_derived( struct cell_context *softpipe ); -#endif + +void +cell_init_shader_functions(struct cell_context *cell); + +#endif /* CELL_STATE_H */ + diff --git a/src/gallium/drivers/cell/ppu/cell_state_fs.c b/src/gallium/drivers/cell/ppu/cell_state_fs.c index f3958fa429..935501441b 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_fs.c +++ b/src/gallium/drivers/cell/ppu/cell_state_fs.c @@ -41,7 +41,7 @@ #include "cell_state.h" -void * +static void * cell_create_fs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { @@ -80,7 +80,7 @@ cell_create_fs_state(struct pipe_context *pipe, } -void +static void cell_bind_fs_state(struct pipe_context *pipe, void *fs) { struct cell_context *cell = cell_context(pipe); @@ -91,7 +91,7 @@ cell_bind_fs_state(struct pipe_context *pipe, void *fs) } -void +static void cell_delete_fs_state(struct pipe_context *pipe, void *fs) { struct cell_fragment_shader_state *state = @@ -101,7 +101,7 @@ cell_delete_fs_state(struct pipe_context *pipe, void *fs) } -void * +static void * cell_create_vs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { @@ -124,7 +124,7 @@ cell_create_vs_state(struct pipe_context *pipe, } -void +static void cell_bind_vs_state(struct pipe_context *pipe, void *vs) { struct cell_context *cell = cell_context(pipe); @@ -137,7 +137,7 @@ cell_bind_vs_state(struct pipe_context *pipe, void *vs) } -void +static void cell_delete_vs_state(struct pipe_context *pipe, void *vs) { struct cell_context *cell = cell_context(pipe); @@ -150,7 +150,7 @@ cell_delete_vs_state(struct pipe_context *pipe, void *vs) } -void +static void cell_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, const struct pipe_constant_buffer *buf) @@ -169,3 +169,18 @@ cell_set_constant_buffer(struct pipe_context *pipe, cell->dirty |= CELL_NEW_CONSTANTS; } + + +void +cell_init_shader_functions(struct cell_context *cell) +{ + cell->pipe.create_fs_state = cell_create_fs_state; + cell->pipe.bind_fs_state = cell_bind_fs_state; + cell->pipe.delete_fs_state = cell_delete_fs_state; + + cell->pipe.create_vs_state = cell_create_vs_state; + cell->pipe.bind_vs_state = cell_bind_vs_state; + cell->pipe.delete_vs_state = cell_delete_vs_state; + + cell->pipe.set_constant_buffer = cell_set_constant_buffer; +} -- cgit v1.2.3 From fd4bdd020a9e1999b87d553b50151405c054a619 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 14:32:43 -0700 Subject: cell: rename cell_state_fs.c -> cell_state_shader.c --- src/gallium/drivers/cell/ppu/Makefile | 2 +- src/gallium/drivers/cell/ppu/cell_state_fs.c | 186 ----------------------- src/gallium/drivers/cell/ppu/cell_state_shader.c | 186 +++++++++++++++++++++++ 3 files changed, 187 insertions(+), 187 deletions(-) delete mode 100644 src/gallium/drivers/cell/ppu/cell_state_fs.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_shader.c (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile index 6b6dfca890..164dde762c 100644 --- a/src/gallium/drivers/cell/ppu/Makefile +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -27,7 +27,7 @@ SOURCES = \ cell_flush.c \ cell_state_derived.c \ cell_state_emit.c \ - cell_state_fs.c \ + cell_state_shader.c \ cell_pipe_state.c \ cell_state_vertex.c \ cell_spu.c \ diff --git a/src/gallium/drivers/cell/ppu/cell_state_fs.c b/src/gallium/drivers/cell/ppu/cell_state_fs.c deleted file mode 100644 index 935501441b..0000000000 --- a/src/gallium/drivers/cell/ppu/cell_state_fs.c +++ /dev/null @@ -1,186 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" -#include "draw/draw_context.h" -#if 0 -#include "pipe/p_shader_tokens.h" -#include "gallivm/gallivm.h" -#include "tgsi/util/tgsi_dump.h" -#include "tgsi/exec/tgsi_sse2.h" -#endif - -#include "cell_context.h" -#include "cell_state.h" - - -static void * -cell_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - /*struct cell_context *cell = cell_context(pipe);*/ - struct cell_fragment_shader_state *state; - - state = CALLOC_STRUCT(cell_fragment_shader_state); - if (!state) - return NULL; - - state->shader = *templ; - -#if 0 - if (cell->dump_fs) { - tgsi_dump(state->shader.tokens, 0); - } - -#if defined(__i386__) || defined(__386__) - if (cell->use_sse) { - x86_init_func( &state->sse2_program ); - tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); - } -#endif - -#ifdef MESA_LLVM - state->llvm_prog = 0; - if (!gallivm_global_cpu_engine()) { - gallivm_cpu_engine_create(state->llvm_prog); - } - else - gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); -#endif -#endif - - return state; -} - - -static void -cell_bind_fs_state(struct pipe_context *pipe, void *fs) -{ - struct cell_context *cell = cell_context(pipe); - - cell->fs = (struct cell_fragment_shader_state *) fs; - - cell->dirty |= CELL_NEW_FS; -} - - -static void -cell_delete_fs_state(struct pipe_context *pipe, void *fs) -{ - struct cell_fragment_shader_state *state = - (struct cell_fragment_shader_state *) fs; - - FREE( state ); -} - - -static void * -cell_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) -{ - struct cell_context *cell = cell_context(pipe); - struct cell_vertex_shader_state *state; - - state = CALLOC_STRUCT(cell_vertex_shader_state); - if (!state) - return NULL; - - state->shader = *templ; - - state->draw_data = draw_create_vertex_shader(cell->draw, &state->shader); - if (state->draw_data == NULL) { - FREE( state ); - return NULL; - } - - return state; -} - - -static void -cell_bind_vs_state(struct pipe_context *pipe, void *vs) -{ - struct cell_context *cell = cell_context(pipe); - - cell->vs = (const struct cell_vertex_shader_state *) vs; - - draw_bind_vertex_shader(cell->draw, cell->vs->draw_data); - - cell->dirty |= CELL_NEW_VS; -} - - -static void -cell_delete_vs_state(struct pipe_context *pipe, void *vs) -{ - struct cell_context *cell = cell_context(pipe); - - struct cell_vertex_shader_state *state = - (struct cell_vertex_shader_state *) vs; - - draw_delete_vertex_shader(cell->draw, state->draw_data); - FREE( state ); -} - - -static void -cell_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) -{ - struct cell_context *cell = cell_context(pipe); - struct pipe_winsys *ws = pipe->winsys; - - assert(shader < PIPE_SHADER_TYPES); - assert(index == 0); - - /* note: reference counting */ - pipe_buffer_reference(ws, - &cell->constants[shader].buffer, - buf->buffer); - cell->constants[shader].size = buf->size; - - cell->dirty |= CELL_NEW_CONSTANTS; -} - - -void -cell_init_shader_functions(struct cell_context *cell) -{ - cell->pipe.create_fs_state = cell_create_fs_state; - cell->pipe.bind_fs_state = cell_bind_fs_state; - cell->pipe.delete_fs_state = cell_delete_fs_state; - - cell->pipe.create_vs_state = cell_create_vs_state; - cell->pipe.bind_vs_state = cell_bind_vs_state; - cell->pipe.delete_vs_state = cell_delete_vs_state; - - cell->pipe.set_constant_buffer = cell_set_constant_buffer; -} diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c new file mode 100644 index 0000000000..935501441b --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -0,0 +1,186 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "draw/draw_context.h" +#if 0 +#include "pipe/p_shader_tokens.h" +#include "gallivm/gallivm.h" +#include "tgsi/util/tgsi_dump.h" +#include "tgsi/exec/tgsi_sse2.h" +#endif + +#include "cell_context.h" +#include "cell_state.h" + + +static void * +cell_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + /*struct cell_context *cell = cell_context(pipe);*/ + struct cell_fragment_shader_state *state; + + state = CALLOC_STRUCT(cell_fragment_shader_state); + if (!state) + return NULL; + + state->shader = *templ; + +#if 0 + if (cell->dump_fs) { + tgsi_dump(state->shader.tokens, 0); + } + +#if defined(__i386__) || defined(__386__) + if (cell->use_sse) { + x86_init_func( &state->sse2_program ); + tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); + } +#endif + +#ifdef MESA_LLVM + state->llvm_prog = 0; + if (!gallivm_global_cpu_engine()) { + gallivm_cpu_engine_create(state->llvm_prog); + } + else + gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); +#endif +#endif + + return state; +} + + +static void +cell_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct cell_context *cell = cell_context(pipe); + + cell->fs = (struct cell_fragment_shader_state *) fs; + + cell->dirty |= CELL_NEW_FS; +} + + +static void +cell_delete_fs_state(struct pipe_context *pipe, void *fs) +{ + struct cell_fragment_shader_state *state = + (struct cell_fragment_shader_state *) fs; + + FREE( state ); +} + + +static void * +cell_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) +{ + struct cell_context *cell = cell_context(pipe); + struct cell_vertex_shader_state *state; + + state = CALLOC_STRUCT(cell_vertex_shader_state); + if (!state) + return NULL; + + state->shader = *templ; + + state->draw_data = draw_create_vertex_shader(cell->draw, &state->shader); + if (state->draw_data == NULL) { + FREE( state ); + return NULL; + } + + return state; +} + + +static void +cell_bind_vs_state(struct pipe_context *pipe, void *vs) +{ + struct cell_context *cell = cell_context(pipe); + + cell->vs = (const struct cell_vertex_shader_state *) vs; + + draw_bind_vertex_shader(cell->draw, cell->vs->draw_data); + + cell->dirty |= CELL_NEW_VS; +} + + +static void +cell_delete_vs_state(struct pipe_context *pipe, void *vs) +{ + struct cell_context *cell = cell_context(pipe); + + struct cell_vertex_shader_state *state = + (struct cell_vertex_shader_state *) vs; + + draw_delete_vertex_shader(cell->draw, state->draw_data); + FREE( state ); +} + + +static void +cell_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct cell_context *cell = cell_context(pipe); + struct pipe_winsys *ws = pipe->winsys; + + assert(shader < PIPE_SHADER_TYPES); + assert(index == 0); + + /* note: reference counting */ + pipe_buffer_reference(ws, + &cell->constants[shader].buffer, + buf->buffer); + cell->constants[shader].size = buf->size; + + cell->dirty |= CELL_NEW_CONSTANTS; +} + + +void +cell_init_shader_functions(struct cell_context *cell) +{ + cell->pipe.create_fs_state = cell_create_fs_state; + cell->pipe.bind_fs_state = cell_bind_fs_state; + cell->pipe.delete_fs_state = cell_delete_fs_state; + + cell->pipe.create_vs_state = cell_create_vs_state; + cell->pipe.bind_vs_state = cell_bind_vs_state; + cell->pipe.delete_vs_state = cell_delete_vs_state; + + cell->pipe.set_constant_buffer = cell_set_constant_buffer; +} -- cgit v1.2.3 From fce61f341faf6a2e1a8497ab963985ddbffa8b0a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 14:44:15 -0700 Subject: gallium: fix bad ptr comparison --- src/gallium/drivers/softpipe/sp_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 8f31f05e47..295704c05f 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -136,7 +136,7 @@ softpipe_texture_update(struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); uint unit; for (unit = 0; unit < PIPE_MAX_SAMPLERS; unit++) { - if (softpipe->texture[unit] == texture) { + if (softpipe->texture[unit] == softpipe_texture(texture)) { sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); } } -- cgit v1.2.3 From 25ea1901b44107a5bc5351487e18d52d75df8ffd Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 15:09:27 -0700 Subject: gallium: replace some ordinary assignments with pipe_reference_texture() This fixes at least one instance of dereferencing an invalid texture pointer. --- src/gallium/drivers/cell/ppu/cell_texture.c | 1 + src/gallium/drivers/i915simple/i915_state.c | 5 ++++- src/gallium/drivers/i915simple/i915_texture.c | 1 + src/gallium/drivers/i965simple/brw_state.c | 4 +++- src/gallium/drivers/i965simple/brw_tex_layout.c | 1 + src/gallium/drivers/softpipe/sp_context.h | 2 +- src/gallium/drivers/softpipe/sp_quad_fs.c | 2 +- src/gallium/drivers/softpipe/sp_state_sampler.c | 3 ++- src/gallium/drivers/softpipe/sp_texture.c | 5 ++++- src/mesa/state_tracker/st_texture.c | 9 ++++++--- 10 files changed, 24 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index e1b91075b2..0edefa5f05 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -88,6 +88,7 @@ cell_texture_create(struct pipe_context *pipe, return NULL; spt->base = *templat; + spt->base.refcount = 1; cell_texture_layout(spt); diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 294e6fad03..e055eed7e0 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -32,6 +32,7 @@ #include "draw/draw_context.h" #include "pipe/p_winsys.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "i915_context.h" #include "i915_reg.h" @@ -505,7 +506,9 @@ static void i915_set_sampler_texture(struct pipe_context *pipe, { struct i915_context *i915 = i915_context(pipe); - i915->texture[sampler] = (struct i915_texture*)texture; /* ptr, not struct */ + pipe_texture_reference(pipe, + (struct pipe_texture **) &i915->texture[sampler], + texture); i915->dirty |= I915_NEW_TEXTURE; } diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index b235fae96d..1b415a94d4 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -488,6 +488,7 @@ i915_texture_create(struct pipe_context *pipe, struct i915_context *i915 = i915_context(pipe); tex->base = *templat; + tex->base.refcount = 1; if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : i915_miptree_layout(pipe, tex)) diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index f746d1cc57..f269b2882c 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -327,7 +327,9 @@ static void brw_set_sampler_texture(struct pipe_context *pipe, { struct brw_context *brw = brw_context(pipe); - brw->attribs.Texture[unit] = (struct brw_texture*)texture; /* ptr, not struct */ + pipe_reference_texture(pipe, + (struct pipe_texture **) &brw->attribs.Texture[unit], + texture); brw->state.dirty.brw |= BRW_NEW_TEXTURE; } diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 043a2ff9a4..86ce3d0cc3 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -308,6 +308,7 @@ brw_texture_create(struct pipe_context *pipe, if (tex) { tex->base = *templat; + tex->base.refcount = 1; if (brw_miptree_layout(pipe, tex)) tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index b70d4fea85..a50cee7648 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -68,7 +68,7 @@ struct softpipe_context { struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; - struct softpipe_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index cf1b1eff75..2f40e09d5c 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -142,7 +142,7 @@ static void shade_begin(struct quad_stage *qs) /* set TGSI sampler state that varies */ for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { qss->samplers[i].state = softpipe->sampler[i]; - qss->samplers[i].texture = &softpipe->texture[i]->base; + qss->samplers[i].texture = softpipe->texture[i]; } /* find output slots for depth, color */ diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c index 9246915e19..18669a1c6e 100644 --- a/src/gallium/drivers/softpipe/sp_state_sampler.c +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -30,6 +30,7 @@ */ #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "draw/draw_context.h" @@ -82,7 +83,7 @@ softpipe_set_sampler_texture(struct pipe_context *pipe, draw_flush(softpipe->draw); assert(unit < PIPE_MAX_SAMPLERS); - softpipe->texture[unit] = softpipe_texture(texture); /* ptr, not struct */ + pipe_texture_reference(pipe, &softpipe->texture[unit], texture); sp_tile_cache_set_texture(pipe, softpipe->tex_cache[unit], texture); diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 295704c05f..6ba0f09e0a 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -89,6 +89,7 @@ softpipe_texture_create(struct pipe_context *pipe, return NULL; spt->base = *templat; + spt->base.refcount = 1; softpipe_texture_layout(spt); @@ -100,6 +101,8 @@ softpipe_texture_create(struct pipe_context *pipe, return NULL; } + assert(spt->base.refcount == 1); + return &spt->base; } @@ -136,7 +139,7 @@ softpipe_texture_update(struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); uint unit; for (unit = 0; unit < PIPE_MAX_SAMPLERS; unit++) { - if (softpipe->texture[unit] == softpipe_texture(texture)) { + if (softpipe->texture[unit] == texture) { sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); } } diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index b86f416c9b..ad284170e4 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -76,7 +76,7 @@ st_texture_create(struct st_context *st, GLuint depth0, GLuint compress_byte) { - struct pipe_texture pt; + struct pipe_texture pt, *newtex; assert(target <= PIPE_TEXTURE_CUBE); @@ -95,9 +95,12 @@ st_texture_create(struct st_context *st, pt.depth[0] = depth0; pt.compressed = compress_byte ? 1 : 0; pt.cpp = pt.compressed ? compress_byte : st_sizeof_format(format); - pt.refcount = 1; - return st->pipe->texture_create(st->pipe, &pt); + newtex = st->pipe->texture_create(st->pipe, &pt); + + assert(!newtex || newtex->refcount == 1); + + return newtex; } -- cgit v1.2.3 From e523ef72044d7f8137a298d60597b8913bae9145 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 15:13:33 -0700 Subject: cell: use pipe_texture_reference() --- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 35e88f79b1..95bfc29fbe 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -242,7 +242,9 @@ cell_set_sampler_texture(struct pipe_context *pipe, draw_flush(cell->draw); - cell->texture[sampler] = cell_texture(texture); + pipe_texture_reference(pipe, + (struct pipe_texture **) &cell->texture[sampler], + texture); cell_update_texture_mapping(cell); -- cgit v1.2.3 From 228f6b978804268a482718e762ebfccbba784949 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 15:32:41 -0700 Subject: gallium: re-fix some msvc warnings --- src/gallium/auxiliary/draw/draw_aaline.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index f1fee4f8ba..73a02a32e4 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -140,14 +140,14 @@ aa_transform_decl(struct tgsi_transform_context *ctx, aactx->colorOutput = decl->u.DeclarationRange.First; } else if (decl->Declaration.File == TGSI_FILE_SAMPLER) { - if (decl->u.DeclarationRange.Last > aactx->maxSampler) + if ((int) decl->u.DeclarationRange.Last > aactx->maxSampler) aactx->maxSampler = decl->u.DeclarationRange.Last + 1; } else if (decl->Declaration.File == TGSI_FILE_INPUT) { - if (decl->u.DeclarationRange.Last > aactx->maxInput) + if ((int) decl->u.DeclarationRange.Last > aactx->maxInput) aactx->maxInput = decl->u.DeclarationRange.Last; - if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && - decl->Semantic.SemanticIndex > aactx->maxGeneric) { + if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && + (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) { aactx->maxGeneric = decl->Semantic.SemanticIndex; } } @@ -393,7 +393,7 @@ aaline_create_texture(struct aaline_stage *aaline) for (i = 0; i < size; i++) { for (j = 0; j < size; j++) { - uint d; + ubyte d; if (size == 1) { d = 255; } @@ -494,12 +494,12 @@ aaline_line(struct draw_stage *stage, struct prim_header *header) float *pos, *tex; float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; - float a = atan2(dy, dx); - float c_a = cos(a), s_a = sin(a); + double a = atan2(dy, dx); + float c_a = (float) cos(a), s_a = (float) sin(a); uint i; /* XXX the ends of lines aren't quite perfect yet, but probably passable */ - dx = 0.5 * half_width; + dx = 0.5F * half_width; dy = half_width; /* allocate/dup new verts */ -- cgit v1.2.3 From d3b7d26b0bab6587cfad64735aefa28d8377c358 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 17:57:40 -0700 Subject: gallium: s/pipe_reference_texture/pipe_texture_reference/ --- src/gallium/drivers/i965simple/brw_state.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index f269b2882c..254e3f7245 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -327,7 +327,7 @@ static void brw_set_sampler_texture(struct pipe_context *pipe, { struct brw_context *brw = brw_context(pipe); - pipe_reference_texture(pipe, + pipe_texture_reference(pipe, (struct pipe_texture **) &brw->attribs.Texture[unit], texture); -- cgit v1.2.3 From 8be9bc08e1da31619f1b1c49aa6280d44f94c442 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 20 Feb 2008 18:00:03 -0700 Subject: gallium: include p_inlines.h --- src/gallium/drivers/i965simple/brw_state.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 254e3f7245..2fc048bde0 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -32,6 +32,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" #include "tgsi/util/tgsi_dump.h" -- cgit v1.2.3 From 7c74037852a484a8a50e8bc540b954a624de4d33 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 20 Feb 2008 14:32:25 -0800 Subject: Cell: Initial pass at unified data cache --- src/gallium/drivers/cell/common.h | 8 ++ src/gallium/drivers/cell/ppu/cell_draw_arrays.c | 6 +- src/gallium/drivers/cell/ppu/cell_flush.c | 14 ++++ src/gallium/drivers/cell/spu/spu_dcache.c | 100 ++++++++++++++++++++++++ src/gallium/drivers/cell/spu/spu_dcache.h | 34 ++++++++ src/gallium/drivers/cell/spu/spu_exec.c | 49 ++++++------ src/gallium/drivers/cell/spu/spu_main.c | 8 ++ src/gallium/drivers/cell/spu/spu_vertex_fetch.c | 70 +---------------- 8 files changed, 194 insertions(+), 95 deletions(-) create mode 100644 src/gallium/drivers/cell/spu/spu_dcache.c create mode 100644 src/gallium/drivers/cell/spu/spu_dcache.h (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index cf892206c6..f32ad5bfbe 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -93,6 +93,7 @@ #define CELL_CMD_STATE_BLEND 19 #define CELL_CMD_STATE_ATTRIB_FETCH 20 #define CELL_CMD_VS_EXECUTE 21 +#define CELL_CMD_FLUSH_BUFFER_RANGE 22 #define CELL_NUM_BUFFERS 4 @@ -144,6 +145,13 @@ struct cell_attribute_fetch_code { uint size; }; + +struct cell_buffer_range { + uint64_t base; + unsigned size; +}; + + struct cell_shader_info { uint64_t declarations; diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c index f12613649b..cbd387f014 100644 --- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c @@ -49,9 +49,12 @@ cell_map_constant_buffers(struct cell_context *sp) struct pipe_winsys *ws = sp->pipe.winsys; uint i; for (i = 0; i < 2; i++) { - if (sp->constants[i].size) + if (sp->constants[i].size) { sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, PIPE_BUFFER_USAGE_CPU_READ); + cell_flush_buffer_range(sp, sp->mapped_constants[i], + sp->constants[i].buffer->size); + } } draw_set_mapped_constant_buffer(sp->draw, @@ -124,6 +127,7 @@ cell_draw_elements(struct pipe_context *pipe, void *buf = pipe->winsys->buffer_map(pipe->winsys, sp->vertex_buffer[i].buffer, PIPE_BUFFER_USAGE_CPU_READ); + cell_flush_buffer_range(sp, buf, sp->vertex_buffer[i].buffer->size); draw_set_mapped_vertex_buffer(draw, i, buf); } } diff --git a/src/gallium/drivers/cell/ppu/cell_flush.c b/src/gallium/drivers/cell/ppu/cell_flush.c index 20f27531fc..66a5627d84 100644 --- a/src/gallium/drivers/cell/ppu/cell_flush.c +++ b/src/gallium/drivers/cell/ppu/cell_flush.c @@ -82,3 +82,17 @@ cell_flush_int(struct pipe_context *pipe, unsigned flags) flushing = FALSE; } + + +void +cell_flush_buffer_range(struct cell_context *cell, void *ptr, + unsigned size) +{ + uint64_t batch[1 + (ROUNDUP8(sizeof(struct cell_buffer_range)) / 8)]; + struct cell_buffer_range *br = (struct cell_buffer_range *) & batch[1]; + + batch[0] = CELL_CMD_FLUSH_BUFFER_RANGE; + br->base = (uintptr_t) ptr; + br->size = size; + cell_batch_append(cell, batch, sizeof(batch)); +} diff --git a/src/gallium/drivers/cell/spu/spu_dcache.c b/src/gallium/drivers/cell/spu/spu_dcache.c new file mode 100644 index 0000000000..9e30e17880 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_dcache.c @@ -0,0 +1,100 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "spu_main.h" +#include "spu_dcache.h" + +#define CACHE_NAME data +#define CACHED_TYPE qword +#define CACHE_TYPE CACHE_TYPE_RO +#define CACHE_SET_TAGID(set) TAG_VERTEX_BUFFER +#define CACHE_LOG2NNWAY 2 +#define CACHE_LOG2NSETS 6 +#include + +/* Yes folks, this is ugly. + */ +#undef CACHE_NWAY +#undef CACHE_NSETS +#define CACHE_NAME data +#define CACHE_NWAY 4 +#define CACHE_NSETS (1U << 6) + + +/** + * Fetch between arbitrary number of bytes from an unaligned address + */ +void +spu_dcache_fetch_unaligned(qword *dst, unsigned ea, unsigned size) +{ + const int shift = ea & 0x0f; + const unsigned aligned_start_ea = ea & ~0x0f; + const unsigned aligned_end_ea = (ea + size) & ~0x0f; + const unsigned num_entries = ((aligned_end_ea - aligned_start_ea) / 16) + 1; + unsigned i; + + + if (shift == 0) { + /* Data is already aligned. Fetch directly into the destination buffer. + */ + for (i = 0; i < num_entries; i++) { + dst[i] = cache_rd(data, (ea & ~0x0f) + (i * 16)); + } + } else { + qword tmp[2] ALIGN16_ATTRIB; + + + tmp[0] = cache_rd(data, (ea & ~0x0f)); + for (i = 0; i < (num_entries & ~1); i++) { + const unsigned curr = i & 1; + const unsigned next = curr ^ 1; + + tmp[next] = cache_rd(data, (ea & ~0x0f) + (next * 16)); + + dst[i] = si_or((qword) spu_slqwbyte(tmp[curr], shift), + (qword) spu_rlmaskqwbyte(tmp[next], shift - 16)); + } + + if (i < num_entries) { + dst[i] = si_or((qword) spu_slqwbyte(tmp[(i & 1)], shift), + si_il(0)); + } + } +} + + +void +spu_dcache_mark_dirty(unsigned ea, unsigned size) +{ + unsigned i; + + (void) ea; + (void) size; + + /* Invalidate the whole cache for now. + */ + for (i = 0; i < (CACHE_NWAY * CACHE_NSETS); i++) { + CACHELINE_CLEARVALID(i); + } +} diff --git a/src/gallium/drivers/cell/spu/spu_dcache.h b/src/gallium/drivers/cell/spu/spu_dcache.h new file mode 100644 index 0000000000..7a06b8c25a --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_dcache.h @@ -0,0 +1,34 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef SPU_DCACHE_H +#define SPU_DCACHE_H + +extern void +spu_dcache_fetch_unaligned(qword *dst, unsigned ea, unsigned size); + +extern void +spu_dcache_mark_dirty(unsigned ea, unsigned size); + +#endif /* SPU_DCACHE_H */ diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 0eb5ea1a3f..94ac6a2885 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -72,6 +72,7 @@ #include "spu_exec.h" #include "spu_main.h" #include "spu_vertex_shader.h" +#include "spu_dcache.h" #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 @@ -352,19 +353,17 @@ fetch_src_file_channel( case TGSI_EXTSWIZZLE_W: switch( file ) { case TGSI_FILE_CONSTANT: { - unsigned char buffer[32] ALIGN16_ATTRIB; unsigned i; for (i = 0; i < 4; i++) { const float *ptr = mach->Consts[index->i[i]]; - const uint64_t addr = (uint64_t)(uintptr_t) ptr; - const unsigned size = ((addr & 0x0f) == 0) ? 16 : 32; + float tmp[4]; - mfc_get(buffer, addr & ~0x0f, size, TAG_VERTEX_BUFFER, 0, 0); - wait_on_mask(1 << TAG_VERTEX_BUFFER); + spu_dcache_fetch_unaligned((qword *) tmp, + (uintptr_t)(ptr + swizzle), + sizeof(float)); - (void) memcpy(& chan->f[i], &buffer[(addr & 0x0f) - + (sizeof(float) * swizzle)], sizeof(float)); + chan->f[i] = tmp[0]; } break; } @@ -1899,32 +1898,30 @@ spu_exec_machine_run( struct spu_exec_machine *mach ) /* execute declarations (interpolants) */ if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { for (i = 0; i < mach->NumDeclarations; i++) { - uint8_t buffer[sizeof(struct tgsi_full_declaration) + 32] ALIGN16_ATTRIB; - struct tgsi_full_declaration decl; - unsigned long decl_addr = (unsigned long) (mach->Declarations+i); - unsigned size = ((sizeof(decl) + (decl_addr & 0x0f) + 0x0f) & ~0x0f); + union { + struct tgsi_full_declaration decl; + qword buffer[2 * ((sizeof(struct tgsi_full_declaration) + 31) + / 32)]; + } d ALIGN16_ATTRIB; + unsigned ea = (unsigned) (mach->Declarations + pc); - mfc_get(buffer, decl_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); - wait_on_mask(1 << TAG_INSTRUCTION_FETCH); + spu_dcache_fetch_unaligned(d.buffer, ea, sizeof(d.decl)); - memcpy(& decl, buffer + (decl_addr & 0x0f), sizeof(decl)); - exec_declaration( mach, &decl ); + exec_declaration( mach, &d.decl ); } } /* execute instructions, until pc is set to -1 */ while (pc != -1) { - uint8_t buffer[sizeof(struct tgsi_full_instruction) + 32] ALIGN16_ATTRIB; - struct tgsi_full_instruction inst; - unsigned long inst_addr = (unsigned long) (mach->Instructions + pc); - unsigned size = ((sizeof(inst) + (inst_addr & 0x0f) + 0x0f) & ~0x0f); - - assert(pc < mach->NumInstructions); - mfc_get(buffer, inst_addr & ~0x0f, size, TAG_INSTRUCTION_FETCH, 0, 0); - wait_on_mask(1 << TAG_INSTRUCTION_FETCH); - - memcpy(& inst, buffer + (inst_addr & 0x0f), sizeof(inst)); - exec_instruction( mach, & inst, &pc ); + union { + struct tgsi_full_instruction inst; + qword buffer[2 * ((sizeof(struct tgsi_full_instruction) + 31) + / 32)]; + } i ALIGN16_ATTRIB; + unsigned ea = (unsigned) (mach->Instructions + pc); + + spu_dcache_fetch_unaligned(i.buffer, ea, sizeof(i.inst)); + exec_instruction( mach, & i.inst, &pc ); } #if 0 diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index dbc3705c24..1136dba62d 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -462,6 +462,14 @@ cmd_batch(uint opcode) pos += (1 + ROUNDUP8(sizeof(struct cell_attribute_fetch_code)) / 8); break; } + case CELL_CMD_FLUSH_BUFFER_RANGE: { + struct cell_buffer_range *br = (struct cell_buffer_range *) + &buffer[pos+1]; + + spu_dcache_mark_dirty((unsigned) br->base, br->size); + pos += (1 + ROUNDUP8(sizeof(struct cell_buffer_range)) / 8); + break; + } default: printf("SPU %u: bad opcode: 0x%llx\n", spu.init.id, buffer[pos]); ASSERT(0); diff --git a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c index e5d9910ff3..f7e4e653e3 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c @@ -40,25 +40,7 @@ #include "spu_exec.h" #include "spu_vertex_shader.h" #include "spu_main.h" - -#define CACHE_NAME attribute -#define CACHED_TYPE qword -#define CACHE_TYPE CACHE_TYPE_RO -#define CACHE_SET_TAGID(set) TAG_VERTEX_BUFFER -#define CACHE_LOG2NNWAY 2 -#define CACHE_LOG2NSETS 6 -#include - -/* Yes folks, this is ugly. - */ -#undef CACHE_NWAY -#undef CACHE_NSETS -#define CACHE_NAME attribute -#define CACHE_NWAY 4 -#define CACHE_NSETS (1U << 6) - - -#define DRAW_DBG 0 +#include "spu_dcache.h" typedef void (*spu_fetch_func)(qword *out, const qword *in, const qword *shuffle_data); @@ -102,44 +84,6 @@ static const qword fetch_shuffle_data[5] ALIGN16_ATTRIB = { }; -/** - * Fetch between 1 and 32 bytes from an unaligned address - */ -static INLINE void -fetch_unaligned(qword *dst, unsigned ea, unsigned size) -{ - qword tmp[4] ALIGN16_ATTRIB; - const int shift = ea & 0x0f; - const unsigned aligned_start_ea = ea & ~0x0f; - const unsigned aligned_end_ea = (ea + size) & ~0x0f; - const unsigned num_entries = ((aligned_end_ea - aligned_start_ea) / 16) + 1; - unsigned i; - - - if (shift == 0) { - /* Data is already aligned. Fetch directly into the destination buffer. - */ - for (i = 0; i < num_entries; i++) { - dst[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16)); - } - } else { - /* Fetch data from the cache to the local buffer. - */ - for (i = 0; i < num_entries; i++) { - tmp[i] = cache_rd(attribute, (ea & ~0x0f) + (i * 16)); - } - - - /* Fix the alignment of the data and write to the destination buffer. - */ - for (i = 0; i < ((size + 15) / 16); i++) { - dst[i] = si_or((qword) spu_slqwbyte(tmp[i], shift), - (qword) spu_rlmaskqwbyte(tmp[i + 1], shift - 16)); - } - } -} - - /** * Fetch vertex attributes for 'count' vertices. */ @@ -182,7 +126,7 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, printf("SPU: fetching = 0x%llx\n", addr); #endif - fetch_unaligned(& in[idx], addr, bytes_per_entry); + spu_dcache_fetch_unaligned(& in[idx], addr, bytes_per_entry); idx += quads_per_entry; } @@ -200,15 +144,5 @@ static void generic_vertex_fetch(struct spu_vs_context *draw, void spu_update_vertex_fetch( struct spu_vs_context *draw ) { - unsigned i; - - - /* Invalidate the vertex cache. - */ - for (i = 0; i < (CACHE_NWAY * CACHE_NSETS); i++) { - CACHELINE_CLEARVALID(i); - } - - draw->vertex_fetch.fetch_func = generic_vertex_fetch; } -- cgit v1.2.3 From 2d1f086c12b6d64f5c3fb80474f26775aeb71370 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 20 Feb 2008 14:45:08 -0800 Subject: Cell: Fix off-by-one error in spu_dcache_fetch_unaligned An off-by-one error caused an extra qword to be fetched under certain alignment / size combinations. --- src/gallium/drivers/cell/spu/spu_dcache.c | 5 +++-- src/gallium/drivers/cell/spu/spu_exec.c | 7 +++---- 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_dcache.c b/src/gallium/drivers/cell/spu/spu_dcache.c index 9e30e17880..68aa5c4ae8 100644 --- a/src/gallium/drivers/cell/spu/spu_dcache.c +++ b/src/gallium/drivers/cell/spu/spu_dcache.c @@ -22,6 +22,7 @@ * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#include "cell/common.h" #include "spu_main.h" #include "spu_dcache.h" @@ -50,8 +51,8 @@ spu_dcache_fetch_unaligned(qword *dst, unsigned ea, unsigned size) { const int shift = ea & 0x0f; const unsigned aligned_start_ea = ea & ~0x0f; - const unsigned aligned_end_ea = (ea + size) & ~0x0f; - const unsigned num_entries = ((aligned_end_ea - aligned_start_ea) / 16) + 1; + const unsigned aligned_end_ea = ROUNDUP16(ea + size); + const unsigned num_entries = (aligned_end_ea - aligned_start_ea) / 16; unsigned i; diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 94ac6a2885..cf81bee8fd 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -73,6 +73,7 @@ #include "spu_main.h" #include "spu_vertex_shader.h" #include "spu_dcache.h" +#include "cell/common.h" #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 @@ -1900,8 +1901,7 @@ spu_exec_machine_run( struct spu_exec_machine *mach ) for (i = 0; i < mach->NumDeclarations; i++) { union { struct tgsi_full_declaration decl; - qword buffer[2 * ((sizeof(struct tgsi_full_declaration) + 31) - / 32)]; + qword buffer[ROUNDUP16(sizeof(struct tgsi_full_declaration)) / 16]; } d ALIGN16_ATTRIB; unsigned ea = (unsigned) (mach->Declarations + pc); @@ -1915,8 +1915,7 @@ spu_exec_machine_run( struct spu_exec_machine *mach ) while (pc != -1) { union { struct tgsi_full_instruction inst; - qword buffer[2 * ((sizeof(struct tgsi_full_instruction) + 31) - / 32)]; + qword buffer[ROUNDUP16(sizeof(struct tgsi_full_instruction)) / 16]; } i ALIGN16_ATTRIB; unsigned ea = (unsigned) (mach->Instructions + pc); -- cgit v1.2.3 From e78fc9f2f4d89b0cae0d56d84dd16cb76a6757dc Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 21 Feb 2008 09:03:29 -0800 Subject: Cell: Initial scalar implementation of spu_dcache_mark_dirty --- src/gallium/drivers/cell/spu/spu_dcache.c | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_dcache.c b/src/gallium/drivers/cell/spu/spu_dcache.c index 68aa5c4ae8..698a5790bb 100644 --- a/src/gallium/drivers/cell/spu/spu_dcache.c +++ b/src/gallium/drivers/cell/spu/spu_dcache.c @@ -26,6 +26,10 @@ #include "spu_main.h" #include "spu_dcache.h" +#define CACHELINE_LOG2SIZE 7 +#define LINE_SIZE (1U << 7) +#define ALIGN_MASK (~(LINE_SIZE - 1)) + #define CACHE_NAME data #define CACHED_TYPE qword #define CACHE_TYPE CACHE_TYPE_RO @@ -60,7 +64,7 @@ spu_dcache_fetch_unaligned(qword *dst, unsigned ea, unsigned size) /* Data is already aligned. Fetch directly into the destination buffer. */ for (i = 0; i < num_entries; i++) { - dst[i] = cache_rd(data, (ea & ~0x0f) + (i * 16)); + dst[i] = cache_rd(data, ea + (i * 16)); } } else { qword tmp[2] ALIGN16_ATTRIB; @@ -85,17 +89,23 @@ spu_dcache_fetch_unaligned(qword *dst, unsigned ea, unsigned size) } +/** + * Notify the cache that a range of main memory may have been modified + */ void spu_dcache_mark_dirty(unsigned ea, unsigned size) { unsigned i; + const unsigned aligned_start = (ea & ALIGN_MASK); + const unsigned aligned_end = (ea + size + (LINE_SIZE - 1)) + & ALIGN_MASK; - (void) ea; - (void) size; - /* Invalidate the whole cache for now. - */ for (i = 0; i < (CACHE_NWAY * CACHE_NSETS); i++) { - CACHELINE_CLEARVALID(i); + const unsigned entry = __cache_dir[i]; + const unsigned addr = entry & ~0x0f; + + __cache_dir[i] = ((addr >= aligned_start) && (addr < aligned_end)) + ? (entry & ~CACHELINE_VALID) : entry; } } -- cgit v1.2.3 From 6dd47c264a8642a4e3dbe0b4fc194174743c64fc Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 21 Feb 2008 10:24:29 -0800 Subject: Cell: Add spu_dcache.c to Makefile. This was erroneously missing in previous commits. --- src/gallium/drivers/cell/spu/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile index 30ef2450ec..c071de1900 100644 --- a/src/gallium/drivers/cell/spu/Makefile +++ b/src/gallium/drivers/cell/spu/Makefile @@ -18,6 +18,7 @@ PROG_SPU_EMBED_O = $(PROG)_spu-embed.o SOURCES = \ spu_main.c \ spu_blend.c \ + spu_dcache.c \ spu_render.c \ spu_texture.c \ spu_tile.c \ -- cgit v1.2.3 From de5c64e0af0b1a1ce3ee12f361341880dc260868 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 21 Feb 2008 10:32:02 -0800 Subject: Cell: Remove erroneous ALIGN16_ATTRIB attributes If a structure is marked as being aligned the SPE compiler performs extra optimizations (sadly, only -O2 is used) when reading the structure. Since most of the structures sent in batch buffers are only 8-byte aligned, this resulted in mysterous bugs with -O2. --- src/gallium/drivers/cell/common.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index f32ad5bfbe..9a4004535e 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -137,7 +137,7 @@ struct cell_array_info uint pitch; /**< Byte pitch from one entry to the next. */ uint size; uint function_offset; -} ALIGN16_ATTRIB; +}; struct cell_attribute_fetch_code { @@ -162,7 +162,7 @@ struct cell_shader_info unsigned num_declarations; unsigned num_instructions; unsigned num_immediates; -} ALIGN16_ATTRIB; +}; #define SPU_VERTS_PER_BATCH 64 @@ -175,7 +175,7 @@ struct cell_command_vs float plane[12][4]; unsigned nr_planes; unsigned nr_attrs; -} ALIGN16_ATTRIB; +}; struct cell_command_render -- cgit v1.2.3 From 1eaf7b775ba0dacff8a3debd7c0f260970e5a61d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 15 Feb 2008 18:54:00 +0000 Subject: tgsi: print debug messages on failure to codegenerate --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 29a7f842ed..779b901f2b 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -2356,11 +2356,17 @@ tgsi_emit_sse2_fs( ok = emit_instruction( func, &parse.FullToken.FullInstruction ); + + if (!ok) { + debug_printf("failed to translate tgsi opcode %d\n", + parse.FullToken.FullInstruction.Instruction.Opcode ); + } break; case TGSI_TOKEN_TYPE_IMMEDIATE: /* XXX implement this */ ok = 0; + debug_printf("failed to emit immediate value\n"); break; default: -- cgit v1.2.3 From 30479ef11004c9498c4ef09048efc56227f104cc Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 15 Feb 2008 18:56:41 +0000 Subject: draw: vertex cache rework Take a baby step to straightening out vertex paths. --- src/gallium/auxiliary/draw/draw_context.c | 8 ++-- src/gallium/auxiliary/draw/draw_prim.c | 2 + src/gallium/auxiliary/draw/draw_private.h | 13 ++++-- src/gallium/auxiliary/draw/draw_vertex_cache.c | 60 ++++++++++++++----------- src/gallium/auxiliary/draw/draw_vertex_shader.c | 7 +-- 5 files changed, 53 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index a7f3b4aecb..b90a9f474d 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -72,10 +72,10 @@ struct draw_context *draw_create( void ) { uint i; const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; - char *tmp = align_malloc(Elements(draw->vcache.vertex) * size, 16); + char *tmp = align_malloc(Elements(draw->vs.queue) * size, 16); - for (i = 0; i < Elements(draw->vcache.vertex); i++) - draw->vcache.vertex[i] = (struct vertex_header *)(tmp + i * size); + for (i = 0; i < Elements(draw->vs.queue); i++) + draw->vs.queue[i].vertex = (struct vertex_header *)(tmp + i * size); } draw->shader_queue_flush = draw_vertex_shader_queue_flush; @@ -108,7 +108,7 @@ void draw_destroy( struct draw_context *draw ) if (draw->pipeline.rasterize) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); tgsi_exec_machine_free_data(&draw->machine); - align_free( draw->vcache.vertex[0] ); /* Frees all the vertices. */ + align_free( draw->vs.queue[0].vertex ); /* Frees all the vertices. */ FREE( draw ); } diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index dd9a848863..7d6cd43410 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -114,6 +114,7 @@ static void draw_prim_queue_flush( struct draw_context *draw ) } draw->pq.queue_nr = 0; + draw->vs.post_nr = 0; draw_vertex_cache_unreference( draw ); } @@ -226,6 +227,7 @@ static void do_triangle( struct draw_context *draw, { struct prim_header *prim = get_queued_prim( draw, 3 ); +// _mesa_printf("tri %d %d %d\n", i0, i1, i2); prim->reset_line_stipple = 1; prim->edgeflags = ~0; prim->pad = 0; diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 6c7e860861..492c152ff9 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -120,7 +120,7 @@ struct draw_stage }; -#define PRIM_QUEUE_LENGTH 16 +#define PRIM_QUEUE_LENGTH 32 #define VCACHE_SIZE 32 #define VCACHE_OVERFLOW 4 #define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ @@ -244,8 +244,12 @@ struct draw_context */ struct { unsigned referenced; /**< bitfield */ - unsigned idx[VCACHE_SIZE + VCACHE_OVERFLOW]; - struct vertex_header *vertex[VCACHE_SIZE + VCACHE_OVERFLOW]; + + struct { + unsigned in; /* client array element */ + unsigned out; /* index in vs queue/array */ + } idx[VCACHE_SIZE + VCACHE_OVERFLOW]; + unsigned overflow; /** To find space in the vertex cache: */ @@ -258,9 +262,10 @@ struct draw_context struct { struct { unsigned elt; /**< index into the user's vertex arrays */ - struct vertex_header *dest; /**< points into vcache.vertex[] array */ + struct vertex_header *vertex; } queue[VS_QUEUE_LENGTH]; unsigned queue_nr; + unsigned post_nr; } vs; /** diff --git a/src/gallium/auxiliary/draw/draw_vertex_cache.c b/src/gallium/auxiliary/draw/draw_vertex_cache.c index 44427999cc..53f8bbec44 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_cache.c +++ b/src/gallium/auxiliary/draw/draw_vertex_cache.c @@ -41,7 +41,7 @@ void draw_vertex_cache_invalidate( struct draw_context *draw ) assert(draw->vs.queue_nr == 0); assert(draw->vcache.referenced == 0); - memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); +// memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); } @@ -62,43 +62,51 @@ static struct vertex_header *get_vertex( struct draw_context *draw, assert(slot < 32); /* so we don't exceed the bitfield size below */ - /* Cache miss? - */ - if (draw->vcache.idx[slot] != i) { - - /* If slot is in use, use the overflow area: + if (draw->vcache.referenced & (1<vcache.referenced & (1 << slot)) { - slot = VCACHE_SIZE + draw->vcache.overflow++; + if (draw->vcache.idx[slot].in == i) { +// _mesa_printf("HIT %d %d\n", slot, i); + assert(draw->vcache.idx[slot].out < draw->vs.queue_nr); + return draw->vs.queue[draw->vcache.idx[slot].out].vertex; } + /* Otherwise a collision + */ + slot = VCACHE_SIZE + draw->vcache.overflow++; +// _mesa_printf("XXX %d --> %d\n", i, slot); + } + + /* Deal with the cache miss: + */ + { + unsigned out; + assert(slot < Elements(draw->vcache.idx)); - draw->vcache.idx[slot] = i; +// _mesa_printf("NEW %d %d\n", slot, i); + draw->vcache.idx[slot].in = i; + draw->vcache.idx[slot].out = out = draw->vs.queue_nr++; + draw->vcache.referenced |= (1 << slot); + /* Add to vertex shader queue: */ assert(draw->vs.queue_nr < VS_QUEUE_LENGTH); - draw->vs.queue[draw->vs.queue_nr].dest = draw->vcache.vertex[slot]; - draw->vs.queue[draw->vs.queue_nr].elt = i; - draw->vs.queue_nr++; + + draw->vs.queue[out].elt = i; + draw->vs.queue[out].vertex->clipmask = 0; + draw->vs.queue[out].vertex->edgeflag = 1; /*XXX use user's edge flag! */ + draw->vs.queue[out].vertex->pad = 0; + draw->vs.queue[out].vertex->vertex_id = UNDEFINED_VERTEX_ID; /* Need to set the vertex's edge flag here. If we're being called * by do_ef_triangle(), that function needs edge flag info! */ - draw->vcache.vertex[slot]->clipmask = 0; - draw->vcache.vertex[slot]->edgeflag = 1; /*XXX use user's edge flag! */ - draw->vcache.vertex[slot]->pad = 0; - draw->vcache.vertex[slot]->vertex_id = UNDEFINED_VERTEX_ID; - } - - /* primitive flushing may have cleared the bitfield but did not - * clear the idx[] array values. Set the bit now. This fixes a - * bug found when drawing long triangle fans. - */ - draw->vcache.referenced |= (1 << slot); - return draw->vcache.vertex[slot]; + return draw->vs.queue[draw->vcache.idx[slot].out].vertex; + } } @@ -130,8 +138,8 @@ void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) { unsigned i; - for (i = 0; i < Elements(draw->vcache.vertex); i++) - draw->vcache.vertex[i]->vertex_id = UNDEFINED_VERTEX_ID; + for (i = 0; i < draw->vs.post_nr; i++) + draw->vs.queue[i].vertex->vertex_id = UNDEFINED_VERTEX_ID; } diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index f68f6e3244..5d2f5c9c43 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -55,7 +55,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) */ shader->prepare( shader, draw ); -// fprintf(stderr, " q(%d) ", draw->vs.queue_nr ); +// fprintf(stderr, "%s %d\n", __FUNCTION__, draw->vs.queue_nr ); /* run vertex shader on vertex cache entries, four per invokation */ for (i = 0; i < draw->vs.queue_nr; i += 4) { @@ -65,12 +65,12 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) for (j = 0; j < n; j++) { elts[j] = draw->vs.queue[i + j].elt; - dests[j] = draw->vs.queue[i + j].dest; + dests[j] = draw->vs.queue[i + j].vertex; } for ( ; j < 4; j++) { elts[j] = elts[0]; - dests[j] = dests[0]; + dests[j] = draw->vs.queue[i + j].vertex; } assert(n > 0); @@ -79,6 +79,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) shader->run(shader, draw, elts, n, dests); } + draw->vs.post_nr = draw->vs.queue_nr; draw->vs.queue_nr = 0; } -- cgit v1.2.3 From 4339744c1676f925d42251bd32795bba9928cd5f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 21 Feb 2008 10:07:58 +0000 Subject: [PATCH] gallium: include p_compiler.h for boolean defn --- src/gallium/include/pipe/p_shader_tokens.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 3ce35310f6..10c47e0ef0 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -5,6 +5,8 @@ extern "C" { #endif // defined __cplusplus +#include "p_compiler.h" + struct tgsi_version { unsigned MajorVersion : 8; -- cgit v1.2.3 From 20fbcbf5801c28865c0bfab3cda45302c8474a66 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 21 Feb 2008 19:07:31 +0000 Subject: [PATCH] softpipe: unbreak sp_setup_pos_vector on non-x86 systems --- src/gallium/drivers/softpipe/sp_fs_exec.c | 35 +++++++++++++++++++++++++++++++ src/gallium/drivers/softpipe/sp_fs_sse.c | 35 ------------------------------- 2 files changed, 35 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index 9ad30a7681..8cb0534342 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -44,6 +44,41 @@ struct sp_exec_fragment_shader { +/** + * Compute quad X,Y,Z,W for the four fragments in a quad. + * + * This should really be part of the compiled shader. + */ +void +sp_setup_pos_vector(const struct tgsi_interp_coef *coef, + float x, float y, + struct tgsi_exec_vector *quadpos) +{ + uint chan; + /* do X */ + quadpos->xyzw[0].f[0] = x; + quadpos->xyzw[0].f[1] = x + 1; + quadpos->xyzw[0].f[2] = x; + quadpos->xyzw[0].f[3] = x + 1; + + /* do Y */ + quadpos->xyzw[1].f[0] = y; + quadpos->xyzw[1].f[1] = y; + quadpos->xyzw[1].f[2] = y + 1; + quadpos->xyzw[1].f[3] = y + 1; + + /* do Z and W for all fragments in the quad */ + for (chan = 2; chan < 4; chan++) { + const float dadx = coef->dadx[chan]; + const float dady = coef->dady[chan]; + const float a0 = coef->a0[chan] + dadx * x + dady * y; + quadpos->xyzw[chan].f[0] = a0; + quadpos->xyzw[chan].f[1] = a0 + dadx; + quadpos->xyzw[chan].f[2] = a0 + dady; + quadpos->xyzw[chan].f[3] = a0 + dadx + dady; + } +} + static void exec_prepare( struct sp_fragment_shader *base, diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 53050b7823..8095d662ee 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -63,41 +63,6 @@ struct sp_sse_fragment_shader { }; -/** - * Compute quad X,Y,Z,W for the four fragments in a quad. - * - * This should really be part of the compiled shader. - */ -void -sp_setup_pos_vector(const struct tgsi_interp_coef *coef, - float x, float y, - struct tgsi_exec_vector *quadpos) -{ - uint chan; - /* do X */ - quadpos->xyzw[0].f[0] = x; - quadpos->xyzw[0].f[1] = x + 1; - quadpos->xyzw[0].f[2] = x; - quadpos->xyzw[0].f[3] = x + 1; - - /* do Y */ - quadpos->xyzw[1].f[0] = y; - quadpos->xyzw[1].f[1] = y; - quadpos->xyzw[1].f[2] = y + 1; - quadpos->xyzw[1].f[3] = y + 1; - - /* do Z and W for all fragments in the quad */ - for (chan = 2; chan < 4; chan++) { - const float dadx = coef->dadx[chan]; - const float dady = coef->dady[chan]; - const float a0 = coef->a0[chan] + dadx * x + dady * y; - quadpos->xyzw[chan].f[0] = a0; - quadpos->xyzw[chan].f[1] = a0 + dadx; - quadpos->xyzw[chan].f[2] = a0 + dady; - quadpos->xyzw[chan].f[3] = a0 + dadx + dady; - } -} - static void fs_sse_prepare( struct sp_fragment_shader *base, -- cgit v1.2.3 From d4d2e36a429fd015316c484fc40be7e6d2c69946 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 21 Feb 2008 08:37:49 -0700 Subject: gallium: comments, white-space clean-up --- src/gallium/include/pipe/p_state.h | 53 +++++++++++++++++++++----------------- 1 file changed, 30 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 1082343e2f..47fa78c31d 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -66,7 +66,8 @@ struct pipe_winsys; * The driver will certainly subclass this to include actual memory * management information. */ -struct pipe_buffer { +struct pipe_buffer +{ unsigned alignment; unsigned usage; unsigned size; @@ -76,8 +77,6 @@ struct pipe_buffer { }; - - /** * Primitive (point/line/tri) rasterization info */ @@ -113,18 +112,21 @@ struct pipe_rasterizer_state }; -struct pipe_poly_stipple { +struct pipe_poly_stipple +{ unsigned stipple[32]; }; -struct pipe_viewport_state { +struct pipe_viewport_state +{ float scale[4]; float translate[4]; }; -struct pipe_scissor_state { +struct pipe_scissor_state +{ unsigned minx:16; unsigned miny:16; unsigned maxx:16; @@ -132,7 +134,8 @@ struct pipe_scissor_state { }; -struct pipe_clip_state { +struct pipe_clip_state +{ float ucp[PIPE_MAX_CLIP_PLANES][4]; unsigned nr; }; @@ -141,13 +144,15 @@ struct pipe_clip_state { /** * Constants for vertex/fragment shaders */ -struct pipe_constant_buffer { +struct pipe_constant_buffer +{ struct pipe_buffer *buffer; unsigned size; /** in bytes */ }; -struct pipe_shader_state { +struct pipe_shader_state +{ const struct tgsi_token *tokens; ubyte num_inputs; ubyte num_outputs; @@ -185,7 +190,8 @@ struct pipe_depth_stencil_alpha_state }; -struct pipe_blend_state { +struct pipe_blend_state +{ unsigned blend_enable:1; unsigned rgb_func:3; /**< PIPE_BLEND_x */ @@ -204,7 +210,8 @@ struct pipe_blend_state { }; -struct pipe_blend_color { +struct pipe_blend_color +{ float color[4]; }; @@ -224,19 +231,19 @@ struct pipe_framebuffer_state */ struct pipe_sampler_state { - unsigned wrap_s:3; /**< PIPE_TEX_WRAP_x */ - unsigned wrap_t:3; /**< PIPE_TEX_WRAP_x */ - unsigned wrap_r:3; /**< PIPE_TEX_WRAP_x */ + unsigned wrap_s:3; /**< PIPE_TEX_WRAP_x */ + unsigned wrap_t:3; /**< PIPE_TEX_WRAP_x */ + unsigned wrap_r:3; /**< PIPE_TEX_WRAP_x */ unsigned min_img_filter:2; /**< PIPE_TEX_FILTER_x */ unsigned min_mip_filter:2; /**< PIPE_TEX_MIPFILTER_x */ unsigned mag_img_filter:2; /**< PIPE_TEX_FILTER_x */ - unsigned compare:1; /**< shadow/depth compare enabled? */ - unsigned compare_mode:1; /**< PIPE_TEX_COMPARE_x */ - unsigned compare_func:3; /**< PIPE_FUNC_x */ - unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */ - float shadow_ambient; /**< shadow test fail color/intensity */ - float lod_bias; /**< LOD/lambda bias */ - float min_lod, max_lod; /**< LOD clamp range, after bias */ + unsigned compare:1; /**< shadow/depth compare enabled? */ + unsigned compare_mode:1; /**< PIPE_TEX_COMPARE_x */ + unsigned compare_func:3; /**< PIPE_FUNC_x */ + unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */ + float shadow_ambient; /**< shadow test fail color/intensity */ + float lod_bias; /**< LOD/lambda bias */ + float min_lod, max_lod; /**< LOD clamp range, after bias */ float border_color[4]; float max_anisotropy; }; @@ -248,10 +255,10 @@ struct pipe_sampler_state */ struct pipe_surface { - struct pipe_buffer *buffer; /**< driver private buffer handle */ + struct pipe_buffer *buffer; /**< surface's buffer/memory */ enum pipe_format format; /**< PIPE_FORMAT_x */ unsigned status; /**< PIPE_SURFACE_STATUS_x */ - unsigned clear_value; /**< may be temporary */ + unsigned clear_value; /**< XXX may be temporary */ unsigned cpp; /**< bytes per pixel */ unsigned width; unsigned height; -- cgit v1.2.3 From a93d8bfaf2aba1b2fe3ecfbb5bc4b7ff113c305e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 21 Feb 2008 12:32:38 -0700 Subject: gallium: fill in some blend/rasterizer template fields to make sure they're all valid, even if not relevant --- src/mesa/state_tracker/st_gen_mipmap.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index a0b4b973aa..c9765b2003 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -111,14 +111,22 @@ st_init_generate_mipmap(struct st_context *st) struct pipe_rasterizer_state rasterizer; struct pipe_depth_stencil_alpha_state depthstencil; + /* we don't use blending, but need to set valid values */ memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; blend.colormask = PIPE_MASK_RGBA; st->gen_mipmap.blend_cso = pipe->create_blend_state(pipe, &blend); memset(&depthstencil, 0, sizeof(depthstencil)); st->gen_mipmap.depthstencil_cso = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil); + /* Note: we're assuming zero is valid for all non-specified fields */ memset(&rasterizer, 0, sizeof(rasterizer)); + rasterizer.front_winding = PIPE_WINDING_CW; + rasterizer.cull_mode = PIPE_WINDING_NONE; st->gen_mipmap.rasterizer_cso = pipe->create_rasterizer_state(pipe, &rasterizer); st->gen_mipmap.stfp = make_tex_fragment_program(st->ctx); -- cgit v1.2.3 From eb4dc2dd5ed62e6ccb55ccc2bc13f6a2f3fc1f76 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 21 Feb 2008 16:18:05 -0700 Subject: gallium: new AA point drawing stage AA points are drawn by converting the point to a quad, then modifying the user's fragment shader to compute a coverage value. The final fragment color's alpha is modulated by the coverage value. Fragments outside the point's radius are killed. --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/draw_aapoint.c | 875 +++++++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_context.c | 2 + src/gallium/auxiliary/draw/draw_context.h | 3 + src/gallium/auxiliary/draw/draw_private.h | 1 + src/gallium/auxiliary/draw/draw_validate.c | 5 + src/gallium/drivers/softpipe/sp_context.c | 6 +- 7 files changed, 892 insertions(+), 1 deletion(-) create mode 100644 src/gallium/auxiliary/draw/draw_aapoint.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 1ee9eca0ca..b7e71ed3ec 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -5,6 +5,7 @@ LIBNAME = draw C_SOURCES = \ draw_aaline.c \ + draw_aapoint.c \ draw_clip.c \ draw_vs_exec.c \ draw_vs_sse.c \ diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c new file mode 100644 index 0000000000..43119cc70b --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -0,0 +1,875 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * AA point stage: AA points are converted to quads and rendered with a + * special fragment shader. Another approach would be to use a texture + * map image of a point, but experiments indicate the quality isn't nearly + * as good as this approach. + * + * Note: this looks a lot like draw_aaline.c but there's actually little + * if any code that can be shared. + * + * Authors: Brian Paul + */ + + +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/util/tgsi_transform.h" +#include "tgsi/util/tgsi_dump.h" + +#include "draw_context.h" +#include "draw_private.h" + + +/* + * Enabling NORMALIZE might give _slightly_ better results. + * Basically, it controls whether we compute distance as d=sqrt(x*x+y*y) or + * d=x*x+y*y. Since we're working with a unit circle, the later seems + * close enough and saves some costly instructions. + */ +#define NORMALIZE 0 + + +/** + * Subclass of pipe_shader_state to carry extra fragment shader info. + */ +struct aapoint_fragment_shader +{ + struct pipe_shader_state state; + void *driver_fs; /**< the regular shader */ + void *aapoint_fs; /**< the aa point-augmented shader */ +}; + + +/** + * Subclass of draw_stage + */ +struct aapoint_stage +{ + struct draw_stage stage; + + int psize_slot; + float radius; + + /** this is the vertex attrib slot for the new texcoords */ + uint tex_slot; + + /* + * Currently bound state + */ + struct aapoint_fragment_shader *fs; + + /* + * Driver interface/override functions + */ + void * (*driver_create_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*driver_bind_fs_state)(struct pipe_context *, void *); + void (*driver_delete_fs_state)(struct pipe_context *, void *); + + struct pipe_context *pipe; +}; + + + +/** + * Subclass of tgsi_transform_context, used for transforming the + * user's fragment shader to add the special AA instructions. + */ +struct aa_transform_context { + struct tgsi_transform_context base; + uint tempsUsed; /**< bitmask */ + int colorOutput; /**< which output is the primary color */ + int maxInput, maxGeneric; /**< max input index found */ + int tmp0, colorTemp; /**< temp registers */ + boolean firstInstruction; +}; + + +/** + * TGSI declaration transform callback. + * Look for two free temp regs and available input reg for new texcoords. + */ +static void +aa_transform_decl(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl) +{ + struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; + + if (decl->Declaration.File == TGSI_FILE_OUTPUT && + decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR && + decl->Semantic.SemanticIndex == 0) { + aactx->colorOutput = decl->u.DeclarationRange.First; + } + else if (decl->Declaration.File == TGSI_FILE_INPUT) { + if ((int) decl->u.DeclarationRange.Last > aactx->maxInput) + aactx->maxInput = decl->u.DeclarationRange.Last; + if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && + (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) { + aactx->maxGeneric = decl->Semantic.SemanticIndex; + } + } + else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + aactx->tempsUsed |= (1 << i); + } + } + + ctx->emit_declaration(ctx, decl); +} + + +/** + * TGSI instruction transform callback. + * Replace writes to result.color w/ a temp reg. + * Upon END instruction, insert texture sampling code for antialiasing. + */ +static void +aa_transform_inst(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst) +{ + struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; + struct tgsi_full_instruction newInst; + + if (aactx->firstInstruction) { + /* emit our new declarations before the first instruction */ + + struct tgsi_full_declaration decl; + const int texInput = aactx->maxInput + 1; + int tmp0; + uint i; + + /* find two free temp regs */ + for (i = 0; i < 32; i++) { + if ((aactx->tempsUsed & (1 << i)) == 0) { + /* found a free temp */ + if (aactx->tmp0 < 0) + aactx->tmp0 = i; + else if (aactx->colorTemp < 0) + aactx->colorTemp = i; + else + break; + } + } + + assert(aactx->colorTemp != aactx->tmp0); + + tmp0 = aactx->tmp0; + + /* declare new generic input/texcoord */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; + decl.Semantic.SemanticIndex = aactx->maxGeneric + 1; + decl.Declaration.Interpolate = 1; + /* XXX this could be linear... */ + decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = texInput; + ctx->emit_declaration(ctx, &decl); + + /* declare new temp regs */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = tmp0; + ctx->emit_declaration(ctx, &decl); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->colorTemp; + ctx->emit_declaration(ctx, &decl); + + aactx->firstInstruction = FALSE; + + + /* + * Emit code to compute fragment coverage, kill if outside point radius + * + * Temp reg0 usage: + * t0.x = distance of fragment from center point + * t0.y = boolean, is t0.x > 1 ? + * t0.z = temporary for computing 1/(1-k) value + * t0.w = final coverage value + */ + + /* MUL t0.xy, tex, tex; # compute x^2, y^2 */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XY; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + ctx->emit_instruction(ctx, &newInst); + + /* ADD t0.x, t0.x, t0.y; # x^2 + y^2 */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_ADD; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + ctx->emit_instruction(ctx, &newInst); + +#if NORMALIZE /* OPTIONAL normalization of length */ + /* RSQ t0.x, t0.x; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_RSQ; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + ctx->emit_instruction(ctx, &newInst); + + /* RCP t0.x, t0.x; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_RCP; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + ctx->emit_instruction(ctx, &newInst); +#endif + + /* SGT t0.y, t0.xxxx, t0.wwww; # bool b = d > 1 (NOTE t0.w == 1) */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SGT; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; + ctx->emit_instruction(ctx, &newInst); + + /* KILP -t0.yyyy; # if b, KILL */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_KILP; + newInst.Instruction.NumDstRegs = 0; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; + ctx->emit_instruction(ctx, &newInst); + + /* SGT t0.y, t0.x, tex.z; # bool b = distance > k */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SGT; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* IF t0.y # if b then */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_IF; + newInst.Instruction.NumDstRegs = 0; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + ctx->emit_instruction(ctx, &newInst); + + { + /* compute coverage factor = (1-d)/(1-k) */ + + /* SUB t0.z, tex.w, tex.z; # m = 1 - k */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SUB; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* RCP t0.z, t0.z; # t0.z = 1 / m */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_RCP; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* SUB t0.x, 1, t0.x; # d = 1 - d */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SUB; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + ctx->emit_instruction(ctx, &newInst); + + /* MUL t0.w, t0.x, t0.z; # coverage = d * m */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + } + + /* ELSE */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_ELSE; + newInst.Instruction.NumDstRegs = 0; + newInst.Instruction.NumSrcRegs = 0; + ctx->emit_instruction(ctx, &newInst); + + { + /* MOV t0.w, tex.w; # coverage = 1.0 */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MOV; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; + ctx->emit_instruction(ctx, &newInst); + } + + /* ENDIF */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_ENDIF; + newInst.Instruction.NumDstRegs = 0; + newInst.Instruction.NumSrcRegs = 0; + ctx->emit_instruction(ctx, &newInst); + } + + if (inst->Instruction.Opcode == TGSI_OPCODE_END) { + /* add alpha modulation code at tail of program */ + + /* MOV result.color.xyz, colorTemp; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MOV; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; + ctx->emit_instruction(ctx, &newInst); + + /* MUL result.color.w, colorTemp, tmp0.w; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->tmp0; + ctx->emit_instruction(ctx, &newInst); + } + else { + /* Not an END instruction. + * Look for writes to result.color and replace with colorTemp reg. + */ + uint i; + + for (i = 0; i < inst->Instruction.NumDstRegs; i++) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + if (dst->DstRegister.File == TGSI_FILE_OUTPUT && + dst->DstRegister.Index == aactx->colorOutput) { + dst->DstRegister.File = TGSI_FILE_TEMPORARY; + dst->DstRegister.Index = aactx->colorTemp; + } + } + } + + ctx->emit_instruction(ctx, inst); +} + + +/** + * Generate the frag shader we'll use for drawing AA lines. + * This will be the user's shader plus some texture/modulate instructions. + */ +static void +generate_aapoint_fs(struct aapoint_stage *aapoint) +{ + const struct pipe_shader_state *orig_fs = &aapoint->fs->state; + struct draw_context *draw = aapoint->stage.draw; + struct pipe_shader_state aapoint_fs; + struct aa_transform_context transform; + +#define MAX 1000 + + aapoint_fs = *orig_fs; /* copy to init */ + aapoint_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); + + memset(&transform, 0, sizeof(transform)); + transform.colorOutput = -1; + transform.maxInput = -1; + transform.maxGeneric = -1; + transform.colorTemp = -1; + transform.tmp0 = -1; + transform.firstInstruction = TRUE; + transform.base.transform_instruction = aa_transform_inst; + transform.base.transform_declaration = aa_transform_decl; + + tgsi_transform_shader(orig_fs->tokens, + (struct tgsi_token *) aapoint_fs.tokens, + MAX, &transform.base); + +#if 0 /* DEBUG */ + tgsi_dump(orig_fs->tokens, 0); + tgsi_dump(aapoint_fs.tokens, 0); +#endif + + aapoint_fs.input_semantic_name[aapoint_fs.num_inputs] = TGSI_SEMANTIC_GENERIC; + aapoint_fs.input_semantic_index[aapoint_fs.num_inputs] = transform.maxGeneric + 1; + aapoint_fs.num_inputs++; + + aapoint->fs->aapoint_fs + = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs); + + /* advertise the extra post-transform vertex attributes which will have + * the texcoords. + */ + draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; + draw->extra_vp_outputs.semantic_index = transform.maxGeneric + 1; +} + + +/** + * When we're about to draw our first AA line in a batch, this function is + * called to tell the driver to bind our modified fragment shader. + */ +static void +bind_aapoint_fragment_shader(struct aapoint_stage *aapoint) +{ + if (!aapoint->fs->aapoint_fs) { + generate_aapoint_fs(aapoint); + } + aapoint->driver_bind_fs_state(aapoint->pipe, aapoint->fs->aapoint_fs); +} + + + +static INLINE struct aapoint_stage * +aapoint_stage( struct draw_stage *stage ) +{ + return (struct aapoint_stage *) stage; +} + + +static void +passthrough_line(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->line(stage->next, header); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Draw an AA point by drawing a quad. + */ +static void +aapoint_point(struct draw_stage *stage, struct prim_header *header) +{ + const struct aapoint_stage *aapoint = aapoint_stage(stage); + struct prim_header tri; + struct vertex_header *v[4]; + uint texPos = aapoint->tex_slot; + float radius, *pos, *tex; + uint i; + float k; + + if (aapoint->psize_slot >= 0) { + radius = 0.5f * header->v[0]->data[aapoint->psize_slot][0]; + } + else { + radius = aapoint->radius; + } + + /* + * Note: the texcoords (generic attrib, really) we use are special: + * The S and T components simply vary from -1 to +1. + * The R component is k, below. + * The Q component is 1.0 and will used as a handy constant in the + * fragment shader. + */ + + /* + * k is the threshold distance from the point's center at which + * we begin alpha attenuation (the coverage value). + * Operating within a unit circle, we'll compute the fragment's + * distance 'd' from the center point using the texcoords. + * IF d > 1.0 THEN + * KILL fragment + * ELSE IF d > k THEN + * compute coverage in [0,1] proportional to d in [k, 1]. + * ELSE + * coverage = 1.0; // full coverage + * ENDIF + */ + +#if !NORMALIZE + k = 1.0 / radius; + k = 1.0 - 2.0 * k + k * k; +#else + k = 1.0 - 1.0 / radius; +#endif + + /* allocate/dup new verts */ + for (i = 0; i < 4; i++) { + v[i] = dup_vert(stage, header->v[0], i); + } + + /* new verts */ + pos = v[0]->data[0]; + pos[0] -= radius; + pos[1] -= radius; + + pos = v[1]->data[0]; + pos[0] += radius; + pos[1] -= radius; + + pos = v[2]->data[0]; + pos[0] += radius; + pos[1] += radius; + + pos = v[3]->data[0]; + pos[0] -= radius; + pos[1] += radius; + + /* new texcoords */ + tex = v[0]->data[texPos]; + ASSIGN_4V(tex, -1, -1, k, 1); + + tex = v[1]->data[texPos]; + ASSIGN_4V(tex, 1, -1, k, 1); + + tex = v[2]->data[texPos]; + ASSIGN_4V(tex, 1, 1, k, 1); + + tex = v[3]->data[texPos]; + ASSIGN_4V(tex, -1, 1, k, 1); + + /* emit 2 tris for the quad strip */ + tri.v[0] = v[0]; + tri.v[1] = v[1]; + tri.v[2] = v[2]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[0]; + tri.v[1] = v[2]; + tri.v[2] = v[3]; + stage->next->tri( stage->next, &tri ); +} + + +static void +aapoint_first_point(struct draw_stage *stage, struct prim_header *header) +{ + auto struct aapoint_stage *aapoint = aapoint_stage(stage); + struct draw_context *draw = stage->draw; + + assert(draw->rasterizer->point_smooth); + + if (draw->rasterizer->point_size <= 2.0) + aapoint->radius = 1.0; + else + aapoint->radius = 0.5f * draw->rasterizer->point_size; + + aapoint->tex_slot = draw->num_vs_outputs; + assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */ + draw->extra_vp_outputs.slot = aapoint->tex_slot; + + /* + * Bind our fragprog. + */ + bind_aapoint_fragment_shader(aapoint); + + /* find psize slot in post-transform vertex */ + aapoint->psize_slot = -1; + if (draw->rasterizer->point_size_per_vertex) { + /* find PSIZ vertex output */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i; + for (i = 0; i < vs->state->num_outputs; i++) { + if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + aapoint->psize_slot = i; + break; + } + } + } + + /* now really draw first line */ + stage->point = aapoint_point; + stage->point(stage, header); +} + + +static void +aapoint_flush(struct draw_stage *stage, unsigned flags) +{ + struct draw_context *draw = stage->draw; + struct aapoint_stage *aapoint = aapoint_stage(stage); + struct pipe_context *pipe = aapoint->pipe; + + stage->point = aapoint_first_point; + stage->next->flush( stage->next, flags ); + + /* restore original frag shader */ + aapoint->driver_bind_fs_state(pipe, aapoint->fs->driver_fs); + + draw->extra_vp_outputs.slot = 0; +} + + +static void +aapoint_reset_stipple_counter(struct draw_stage *stage) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +aapoint_destroy(struct draw_stage *stage) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +static struct aapoint_stage * +draw_aapoint_stage(struct draw_context *draw) +{ + struct aapoint_stage *aapoint = CALLOC_STRUCT(aapoint_stage); + + draw_alloc_temp_verts( &aapoint->stage, 4 ); + + aapoint->stage.draw = draw; + aapoint->stage.next = NULL; + aapoint->stage.point = aapoint_first_point; + aapoint->stage.line = passthrough_line; + aapoint->stage.tri = passthrough_tri; + aapoint->stage.flush = aapoint_flush; + aapoint->stage.reset_stipple_counter = aapoint_reset_stipple_counter; + aapoint->stage.destroy = aapoint_destroy; + + return aapoint; +} + + +/* + * XXX temporary? solution to mapping a pipe_context to a aapoint_stage. + */ + +#define MAX_CONTEXTS 10 + +static struct pipe_context *Pipe[MAX_CONTEXTS]; +static struct aapoint_stage *Stage[MAX_CONTEXTS]; +static uint NumContexts; + +static void +add_aa_pipe_context(struct pipe_context *pipe, struct aapoint_stage *aa) +{ + assert(NumContexts < MAX_CONTEXTS); + Pipe[NumContexts] = pipe; + Stage[NumContexts] = aa; + NumContexts++; +} + +static struct aapoint_stage * +aapoint_stage_from_pipe(struct pipe_context *pipe) +{ + uint i; + for (i = 0; i < NumContexts; i++) { + if (Pipe[i] == pipe) + return Stage[i]; + } + assert(0); + return NULL; +} + + +/** + * This function overrides the driver's create_fs_state() function and + * will typically be called by the state tracker. + */ +static void * +aapoint_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *fs) +{ + struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); + struct aapoint_fragment_shader *aafs = CALLOC_STRUCT(aapoint_fragment_shader); + + if (aafs) { + aafs->state = *fs; + + /* pass-through */ + aafs->driver_fs = aapoint->driver_create_fs_state(aapoint->pipe, fs); + } + + return aafs; +} + + +static void +aapoint_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); + struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs; + /* save current */ + aapoint->fs = aafs; + /* pass-through */ + aapoint->driver_bind_fs_state(aapoint->pipe, aafs->driver_fs); +} + + +static void +aapoint_delete_fs_state(struct pipe_context *pipe, void *fs) +{ + struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); + struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs; + /* pass-through */ + aapoint->driver_delete_fs_state(aapoint->pipe, aafs->driver_fs); + FREE(aafs); +} + + +/** + * Called by drivers that want to install this AA point prim stage + * into the draw module's pipeline. This will not be used if the + * hardware has native support for AA points. + */ +void +draw_install_aapoint_stage(struct draw_context *draw, + struct pipe_context *pipe) +{ + struct aapoint_stage *aapoint; + + /* + * Create / install AA point drawing / prim stage + */ + aapoint = draw_aapoint_stage( draw ); + assert(aapoint); + draw->pipeline.aapoint = &aapoint->stage; + + aapoint->pipe = pipe; + + /* save original driver functions */ + aapoint->driver_create_fs_state = pipe->create_fs_state; + aapoint->driver_bind_fs_state = pipe->bind_fs_state; + aapoint->driver_delete_fs_state = pipe->delete_fs_state; + + /* override the driver's functions */ + pipe->create_fs_state = aapoint_create_fs_state; + pipe->bind_fs_state = aapoint_bind_fs_state; + pipe->delete_fs_state = aapoint_delete_fs_state; + + add_aa_pipe_context(pipe, aapoint); +} diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index b90a9f474d..c28e78d33a 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -105,6 +105,8 @@ void draw_destroy( struct draw_context *draw ) draw->pipeline.validate->destroy( draw->pipeline.validate ); if (draw->pipeline.aaline) draw->pipeline.aaline->destroy( draw->pipeline.aaline ); + if (draw->pipeline.aapoint) + draw->pipeline.aapoint->destroy( draw->pipeline.aapoint ); if (draw->pipeline.rasterize) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); tgsi_exec_machine_free_data(&draw->machine); diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index 82035aa8ad..43b58bc056 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -99,6 +99,9 @@ boolean draw_use_sse(struct draw_context *draw); void draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe); +void +draw_install_aapoint_stage(struct draw_context *draw, struct pipe_context *pipe); + int draw_find_vs_output(struct draw_context *draw, diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 492c152ff9..8c469e74b7 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -180,6 +180,7 @@ struct draw_context struct draw_stage *offset; struct draw_stage *unfilled; struct draw_stage *stipple; + struct draw_stage *aapoint; struct draw_stage *aaline; struct draw_stage *wide; struct draw_stage *rasterize; diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index 97e40e372b..5167ef1e75 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -63,6 +63,11 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) next = draw->pipeline.aaline; } + if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) { + draw->pipeline.aapoint->next = next; + next = draw->pipeline.aapoint; + } + if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines && !draw->rasterizer->line_smooth) || (draw->rasterizer->point_size != 1.0 && draw->convert_wide_points) || diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 316020cba6..de4c6c18e7 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -328,8 +328,12 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, draw_set_rasterize_stage(softpipe->draw, softpipe->setup); } - /* enable aaline stage */ + /* plug in AA line/point stages */ draw_install_aaline_stage(softpipe->draw, &softpipe->pipe); + draw_install_aapoint_stage(softpipe->draw, &softpipe->pipe); + + /* sp_prim_setup can do wide points (don't convert to quads) */ + draw_convert_wide_points(softpipe->draw, FALSE); sp_init_surface_functions(softpipe); -- cgit v1.2.3 From 446bfc32a83008e0865ec869bc80b920c907f10f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 21 Feb 2008 16:56:32 -0700 Subject: gallium: new draw stage for polygon stipple. For hardware without native polygon stipple. Create a 32x32 alpha texture that encodes the stipple pattern. Modify the user's fragment program to sample the texture (with gl_FragCoord) and kill the fragment according to the texel value. Temporarily enabled in softpipe driver, replacing the sp_quad_stipple.c step. --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/draw_context.h | 3 + src/gallium/auxiliary/draw/draw_private.h | 1 + src/gallium/auxiliary/draw/draw_pstipple.c | 717 +++++++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_validate.c | 6 + src/gallium/drivers/softpipe/sp_context.c | 5 + src/gallium/drivers/softpipe/sp_context.h | 7 + src/gallium/drivers/softpipe/sp_quad.c | 2 + 8 files changed, 742 insertions(+) create mode 100644 src/gallium/auxiliary/draw/draw_pstipple.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index b7e71ed3ec..c9980f0b83 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -16,6 +16,7 @@ C_SOURCES = \ draw_flatshade.c \ draw_offset.c \ draw_prim.c \ + draw_pstipple.c \ draw_stipple.c \ draw_twoside.c \ draw_unfilled.c \ diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index 43b58bc056..c25301f71d 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -102,6 +102,9 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe); void draw_install_aapoint_stage(struct draw_context *draw, struct pipe_context *pipe); +void +draw_install_pstipple_stage(struct draw_context *draw, struct pipe_context *pipe); + int draw_find_vs_output(struct draw_context *draw, diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 8c469e74b7..6abced139b 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -182,6 +182,7 @@ struct draw_context struct draw_stage *stipple; struct draw_stage *aapoint; struct draw_stage *aaline; + struct draw_stage *pstipple; struct draw_stage *wide; struct draw_stage *rasterize; } pipeline; diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c new file mode 100644 index 0000000000..4048abf856 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -0,0 +1,717 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Polygon stipple stage: implement polygon stipple with texture map and + * fragment program. The fragment program samples the texture and does + * a fragment kill for the stipple-failing fragments. + * + * Authors: Brian Paul + */ + + +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/util/tgsi_transform.h" +#include "tgsi/util/tgsi_dump.h" + +#include "draw_context.h" +#include "draw_private.h" + + + +/** + * Subclass of pipe_shader_state to carry extra fragment shader info. + */ +struct pstip_fragment_shader +{ + struct pipe_shader_state state; + void *driver_fs; + void *pstip_fs; +}; + + +/** + * Subclass of draw_stage + */ +struct pstip_stage +{ + struct draw_stage stage; + + void *sampler_cso; + struct pipe_texture *texture; + uint sampler_unit; + + /* + * Currently bound state + */ + struct pstip_fragment_shader *fs; + struct { + void *sampler[PIPE_MAX_SAMPLERS]; + struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + const struct pipe_poly_stipple *stipple; + } state; + + /* + * Driver interface/override functions + */ + void * (*driver_create_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*driver_bind_fs_state)(struct pipe_context *, void *); + void (*driver_delete_fs_state)(struct pipe_context *, void *); + + void (*driver_bind_sampler_state)(struct pipe_context *, unsigned, void *); + + void (*driver_set_sampler_texture)(struct pipe_context *, + unsigned sampler, + struct pipe_texture *); + + void (*driver_set_polygon_stipple)(struct pipe_context *, + const struct pipe_poly_stipple *); + + struct pipe_context *pipe; +}; + + + +/** + * Subclass of tgsi_transform_context, used for transforming the + * user's fragment shader to add the special AA instructions. + */ +struct pstip_transform_context { + struct tgsi_transform_context base; + uint tempsUsed; /**< bitmask */ + int wincoordInput; + int maxInput; + int maxSampler; /**< max sampler index found */ + int texTemp; /**< temp registers */ + int numImmed; + boolean firstInstruction; +}; + + +/** + * TGSI declaration transform callback. + * Look for a free sampler, a free input attrib, and two free temp regs. + */ +static void +pstip_transform_decl(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl) +{ + struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; + + if (decl->Declaration.File == TGSI_FILE_SAMPLER) { + if ((int) decl->u.DeclarationRange.Last > pctx->maxSampler) + pctx->maxSampler = (int) decl->u.DeclarationRange.Last; + } + else if (decl->Declaration.File == TGSI_FILE_INPUT) { + pctx->maxInput = MAX2(pctx->maxInput, decl->u.DeclarationRange.Last); + if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) + pctx->wincoordInput = (int) decl->u.DeclarationRange.First; + } + else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + pctx->tempsUsed |= (1 << i); + } + } + + ctx->emit_declaration(ctx, decl); +} + + +static void +pstip_transform_immed(struct tgsi_transform_context *ctx, + struct tgsi_full_immediate *immed) +{ + struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; + pctx->numImmed++; +} + + +/** + * TGSI instruction transform callback. + * Replace writes to result.color w/ a temp reg. + * Upon END instruction, insert texture sampling code for antialiasing. + */ +static void +pstip_transform_inst(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst) +{ + struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; + + if (pctx->firstInstruction) { + /* emit our new declarations before the first instruction */ + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction newInst; + uint i; + int wincoordInput; + const int sampler = pctx->maxSampler + 1; + + if (pctx->wincoordInput < 0) + wincoordInput = pctx->maxInput + 1; + else + wincoordInput = pctx->wincoordInput; + + /* find one free temp reg */ + for (i = 0; i < 32; i++) { + if ((pctx->tempsUsed & (1 << i)) == 0) { + /* found a free temp */ + if (pctx->texTemp < 0) + pctx->texTemp = i; + else + break; + } + } + assert(pctx->texTemp >= 0); + + if (pctx->wincoordInput < 0) { + /* declare new position input reg */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; + decl.Semantic.SemanticIndex = 0; + decl.Declaration.Interpolate = 1; + decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR; /* XXX? */ + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = wincoordInput; + ctx->emit_declaration(ctx, &decl); + } + + /* declare new sampler */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_SAMPLER; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = sampler; + ctx->emit_declaration(ctx, &decl); + + /* declare new temp regs */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = pctx->texTemp; + ctx->emit_declaration(ctx, &decl); + + /* emit immediate = {1/32, 1/32, 1, 1} + * The index/position of this immediate will be pctx->numImmed + */ + { + static const float value[4] = { 1.0/32, 1.0/32, 1.0, 1.0 }; + struct tgsi_full_immediate immed; + uint size = 4; + immed = tgsi_default_full_immediate(); + immed.Immediate.Size = 1 + size; /* one for the token itself */ + immed.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; + ctx->emit_immediate(ctx, &immed); + } + + pctx->firstInstruction = FALSE; + + + /* + * Insert new MUL/TEX/KILP instructions at start of program + * Take gl_FragCoord, divide by 32 (stipple size), sample the + * texture and kill fragment if needed. + * + * We'd like to use non-normalized texcoords to index into a RECT + * texture, but we can only use GL_REPEAT wrap mode with normalized + * texcoords. Darn. + */ + + /* MUL texTemp, INPUT[wincoord], 1/32; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = wincoordInput; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_IMMEDIATE; + newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->numImmed; + ctx->emit_instruction(ctx, &newInst); + + /* TEX texTemp, texTemp, sampler; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_TEX; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp; + newInst.Instruction.NumSrcRegs = 2; + newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + newInst.FullSrcRegisters[1].SrcRegister.Index = sampler; + ctx->emit_instruction(ctx, &newInst); + + /* KILP texTemp; # if texTemp < 0, KILL fragment */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_KILP; + newInst.Instruction.NumDstRegs = 0; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp; + newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; + ctx->emit_instruction(ctx, &newInst); + } + + /* emit this instruction */ + ctx->emit_instruction(ctx, inst); +} + + +/** + * Generate the frag shader we'll use for doing polygon stipple. + * This will be the user's shader prefixed with a TEX and KIL instruction. + */ +static void +generate_pstip_fs(struct pstip_stage *pstip) +{ + const struct pipe_shader_state *orig_fs = &pstip->fs->state; + /*struct draw_context *draw = pstip->stage.draw;*/ + struct pipe_shader_state pstip_fs; + struct pstip_transform_context transform; + +#define MAX 1000 + + pstip_fs = *orig_fs; /* copy to init */ + pstip_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); + + memset(&transform, 0, sizeof(transform)); + transform.wincoordInput = -1; + transform.maxInput = -1; + transform.maxSampler = -1; + transform.texTemp = -1; + transform.firstInstruction = TRUE; + transform.base.transform_instruction = pstip_transform_inst; + transform.base.transform_declaration = pstip_transform_decl; + transform.base.transform_immediate = pstip_transform_immed; + + tgsi_transform_shader(orig_fs->tokens, + (struct tgsi_token *) pstip_fs.tokens, + MAX, &transform.base); + +#if 1 /* DEBUG */ + tgsi_dump(orig_fs->tokens, 0); + tgsi_dump(pstip_fs.tokens, 0); +#endif + + pstip->sampler_unit = transform.maxSampler + 1; + + if (transform.wincoordInput < 0) { + pstip_fs.input_semantic_name[pstip_fs.num_inputs] = TGSI_SEMANTIC_POSITION; + pstip_fs.input_semantic_index[pstip_fs.num_inputs] = transform.maxInput; + pstip_fs.num_inputs++; + } + + pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs); +} + + +/** + * Load texture image with current stipple pattern. + */ +static void +pstip_update_texture(struct pstip_stage *pstip) +{ + static const uint bit31 = 1 << 31; + struct pipe_context *pipe = pstip->pipe; + struct pipe_surface *surface; + const uint *stipple = pstip->state.stipple->stipple; + uint i, j; + ubyte *data; + + surface = pipe->get_tex_surface(pipe, pstip->texture, 0, 0, 0); + data = pipe_surface_map(surface); + + /* + * Load alpha texture. + * Note: 0 means keep the fragment, 255 means kill it. + * We'll negate the texel value and use KILP which kills if value + * is negative. + */ + for (i = 0; i < 32; i++) { + for (j = 0; j < 32; j++) { + if (stipple[i] & (bit31 >> j)) { + /* fragment "on" */ + data[i * surface->pitch + j] = 0; + } + else { + /* fragment "off" */ + data[i * surface->pitch + j] = 255; + } + } + } + + /* unmap */ + pipe_surface_unmap(surface); + pipe_surface_reference(&surface, NULL); + pipe->texture_update(pipe, pstip->texture); +} + + +/** + * Create the texture map we'll use for stippling. + */ +static void +pstip_create_texture(struct pstip_stage *pstip) +{ + struct pipe_context *pipe = pstip->pipe; + struct pipe_texture texTemp; + + memset(&texTemp, 0, sizeof(texTemp)); + texTemp.target = PIPE_TEXTURE_2D; + texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */ + texTemp.last_level = 0; + texTemp.width[0] = 32; + texTemp.height[0] = 32; + texTemp.depth[0] = 1; + texTemp.cpp = 1; + + pstip->texture = pipe->texture_create(pipe, &texTemp); + + //pstip_update_texture(pstip); +} + + +/** + * Create the sampler CSO that'll be used for antialiasing. + * By using a mipmapped texture, we don't have to generate a different + * texture image for each line size. + */ +static void +pstip_create_sampler(struct pstip_stage *pstip) +{ + struct pipe_sampler_state sampler; + struct pipe_context *pipe = pstip->pipe; + + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_REPEAT; + sampler.wrap_t = PIPE_TEX_WRAP_REPEAT; + sampler.wrap_r = PIPE_TEX_WRAP_REPEAT; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.normalized_coords = 1; + sampler.min_lod = 0.0f; + sampler.max_lod = 0.0f; + + pstip->sampler_cso = pipe->create_sampler_state(pipe, &sampler); +} + + +/** + * When we're about to draw our first AA line in a batch, this function is + * called to tell the driver to bind our modified fragment shader. + */ +static void +bind_pstip_fragment_shader(struct pstip_stage *pstip) +{ + if (!pstip->fs->pstip_fs) { + generate_pstip_fs(pstip); + } + pstip->driver_bind_fs_state(pstip->pipe, pstip->fs->pstip_fs); +} + + + +static INLINE struct pstip_stage * +pstip_stage( struct draw_stage *stage ) +{ + return (struct pstip_stage *) stage; +} + + +static void +passthrough_point(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->point(stage->next, header); +} + + +static void +passthrough_line(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->line(stage->next, header); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + + +static void +pstip_first_tri(struct draw_stage *stage, struct prim_header *header) +{ + struct pstip_stage *pstip = pstip_stage(stage); + struct draw_context *draw = stage->draw; + struct pipe_context *pipe = pstip->pipe; + + assert(draw->rasterizer->poly_stipple_enable); + + /* + * Bind our fragprog, sampler and texture + */ + bind_pstip_fragment_shader(pstip); + + pstip->driver_bind_sampler_state(pipe, pstip->sampler_unit, pstip->sampler_cso); + pstip->driver_set_sampler_texture(pipe, pstip->sampler_unit, pstip->texture); + + /* now really draw first line */ + stage->tri = passthrough_tri; + stage->tri(stage, header); +} + + +static void +pstip_flush(struct draw_stage *stage, unsigned flags) +{ + /*struct draw_context *draw = stage->draw;*/ + struct pstip_stage *pstip = pstip_stage(stage); + struct pipe_context *pipe = pstip->pipe; + + stage->tri = pstip_first_tri; + stage->next->flush( stage->next, flags ); + + /* restore original frag shader */ + pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs); + + /* XXX restore original texture, sampler state */ + pstip->driver_bind_sampler_state(pipe, pstip->sampler_unit, + pstip->state.sampler[pstip->sampler_unit]); + pstip->driver_set_sampler_texture(pipe, pstip->sampler_unit, + pstip->state.texture[pstip->sampler_unit]); +} + + +static void +pstip_reset_stipple_counter(struct draw_stage *stage) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +pstip_destroy(struct draw_stage *stage) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +static struct pstip_stage * +draw_pstip_stage(struct draw_context *draw) +{ + struct pstip_stage *pstip = CALLOC_STRUCT(pstip_stage); + + draw_alloc_temp_verts( &pstip->stage, 8 ); + + pstip->stage.draw = draw; + pstip->stage.next = NULL; + pstip->stage.point = passthrough_point; + pstip->stage.line = passthrough_line; + pstip->stage.tri = pstip_first_tri; + pstip->stage.flush = pstip_flush; + pstip->stage.reset_stipple_counter = pstip_reset_stipple_counter; + pstip->stage.destroy = pstip_destroy; + + return pstip; +} + + +/* + * XXX temporary? solution to mapping a pipe_context to a pstip_stage. + */ + +#define MAX_CONTEXTS 10 + +static struct pipe_context *Pipe[MAX_CONTEXTS]; +static struct pstip_stage *Stage[MAX_CONTEXTS]; +static uint NumContexts; + +static void +add_pstip_pipe_context(struct pipe_context *pipe, struct pstip_stage *pstip) +{ + assert(NumContexts < MAX_CONTEXTS); + Pipe[NumContexts] = pipe; + Stage[NumContexts] = pstip; + NumContexts++; +} + +static struct pstip_stage * +pstip_stage_from_pipe(struct pipe_context *pipe) +{ + uint i; + for (i = 0; i < NumContexts; i++) { + if (Pipe[i] == pipe) + return Stage[i]; + } + assert(0); + return NULL; +} + + +/** + * This function overrides the driver's create_fs_state() function and + * will typically be called by the state tracker. + */ +static void * +pstip_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *fs) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + struct pstip_fragment_shader *aafs = CALLOC_STRUCT(pstip_fragment_shader); + + if (aafs) { + aafs->state = *fs; + + /* pass-through */ + aafs->driver_fs = pstip->driver_create_fs_state(pstip->pipe, fs); + } + + return aafs; +} + + +static void +pstip_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs; + /* save current */ + pstip->fs = aafs; + /* pass-through */ + pstip->driver_bind_fs_state(pstip->pipe, aafs->driver_fs); +} + + +static void +pstip_delete_fs_state(struct pipe_context *pipe, void *fs) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs; + /* pass-through */ + pstip->driver_delete_fs_state(pstip->pipe, aafs->driver_fs); + FREE(aafs); +} + + +static void +pstip_bind_sampler_state(struct pipe_context *pipe, + unsigned unit, void *sampler) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + /* save current */ + pstip->state.sampler[unit] = sampler; + /* pass-through */ + pstip->driver_bind_sampler_state(pstip->pipe, unit, sampler); +} + + +static void +pstip_set_sampler_texture(struct pipe_context *pipe, + unsigned sampler, struct pipe_texture *texture) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + /* save current */ + pstip->state.texture[sampler] = texture; + /* pass-through */ + pstip->driver_set_sampler_texture(pstip->pipe, sampler, texture); +} + + +static void +pstip_set_polygon_stipple(struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + /* save current */ + pstip->state.stipple = stipple; + /* pass-through */ + pstip->driver_set_polygon_stipple(pstip->pipe, stipple); + + pstip_update_texture(pstip); +} + + + +/** + * Called by drivers that want to install this AA line prim stage + * into the draw module's pipeline. This will not be used if the + * hardware has native support for AA lines. + */ +void +draw_install_pstipple_stage(struct draw_context *draw, + struct pipe_context *pipe) +{ + struct pstip_stage *pstip; + + /* + * Create / install AA line drawing / prim stage + */ + pstip = draw_pstip_stage( draw ); + assert(pstip); + draw->pipeline.pstipple = &pstip->stage; + + pstip->pipe = pipe; + + /* create special texture, sampler state */ + pstip_create_texture(pstip); + pstip_create_sampler(pstip); + + /* save original driver functions */ + pstip->driver_create_fs_state = pipe->create_fs_state; + pstip->driver_bind_fs_state = pipe->bind_fs_state; + pstip->driver_delete_fs_state = pipe->delete_fs_state; + + pstip->driver_bind_sampler_state = pipe->bind_sampler_state; + pstip->driver_set_sampler_texture = pipe->set_sampler_texture; + pstip->driver_set_polygon_stipple = pipe->set_polygon_stipple; + + /* override the driver's functions */ + pipe->create_fs_state = pstip_create_fs_state; + pipe->bind_fs_state = pstip_bind_fs_state; + pipe->delete_fs_state = pstip_delete_fs_state; + + pipe->bind_sampler_state = pstip_bind_sampler_state; + pipe->set_sampler_texture = pstip_set_sampler_texture; + pipe->set_polygon_stipple = pstip_set_polygon_stipple; + + add_pstip_pipe_context(pipe, pstip); +} diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index 5167ef1e75..efd6793f2b 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -82,6 +82,12 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) precalc_flat = 1; /* only needed for lines really */ } + if (draw->rasterizer->poly_stipple_enable + && draw->pipeline.pstipple) { + draw->pipeline.pstipple->next = next; + next = draw->pipeline.pstipple; + } + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { draw->pipeline.unfilled->next = next; diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index de4c6c18e7..2cdf3c75bf 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -332,6 +332,11 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, draw_install_aaline_stage(softpipe->draw, &softpipe->pipe); draw_install_aapoint_stage(softpipe->draw, &softpipe->pipe); +#if USE_DRAW_STAGE_PSTIPPLE + /* Do polygon stipple w/ texture map + frag prog? */ + draw_install_pstipple_stage(softpipe->draw, &softpipe->pipe); +#endif + /* sp_prim_setup can do wide points (don't convert to quads) */ draw_convert_wide_points(softpipe->draw, FALSE); diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index a50cee7648..feeafc7084 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -39,6 +39,13 @@ #include "sp_quad.h" +/** + * This is a temporary variable for testing draw-stage polygon stipple. + * If zero, do stipple in sp_quad_stipple.c + */ +#define USE_DRAW_STAGE_PSTIPPLE 1 + + struct softpipe_winsys; struct softpipe_vbuf_render; struct draw_context; diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad.c index 6bd468a51c..15b5594547 100644 --- a/src/gallium/drivers/softpipe/sp_quad.c +++ b/src/gallium/drivers/softpipe/sp_quad.c @@ -112,7 +112,9 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp_push_quad_first( sp, sp->quad.earlyz ); } +#if !USE_DRAW_STAGE_PSTIPPLE if (sp->rasterizer->poly_stipple_enable) { sp_push_quad_first( sp, sp->quad.polygon_stipple ); } +#endif } -- cgit v1.2.3 From 26c57d163092d8069c69ff47929e73682b819ab2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 23 Feb 2008 00:46:40 +0900 Subject: Avoid building problematic module/drivers on windows. --- src/SConscript | 11 ++++++----- src/gallium/SConscript | 9 +++++++-- src/gallium/winsys/SConscript | 3 ++- 3 files changed, 15 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/SConscript b/src/SConscript index 5b09943894..e2ee50dd6e 100644 --- a/src/SConscript +++ b/src/SConscript @@ -1,7 +1,8 @@ Import('*') -SConscript([ - 'gallium/SConscript', - 'mesa/SConscript', - 'gallium/winsys/SConscript', -]) +SConscript('gallium/SConscript') + +if 'mesa' in env['state_trackers']: + SConscript('mesa/SConscript') + +SConscript('gallium/winsys/SConscript') diff --git a/src/gallium/SConscript b/src/gallium/SConscript index a08b4b830e..210fd7d309 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -23,7 +23,12 @@ if llvm: SConscript([ 'drivers/softpipe/SConscript', 'drivers/i915simple/SConscript', - 'drivers/i965simple/SConscript', 'drivers/failover/SConscript', - #'drivers/cell/SConscript', ]) + +if not msvc: + SConscript(['drivers/i965simple/SConscript']) + +#if cell: +# SConscript(['drivers/cell/SConscript']) + \ No newline at end of file diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript index 32215d8d58..3f0596e19c 100644 --- a/src/gallium/winsys/SConscript +++ b/src/gallium/winsys/SConscript @@ -4,7 +4,8 @@ if dri: SConscript([ 'dri/SConscript', ]) -else: + +if env['x11'] and not dri and platform != 'winddk': SConscript([ 'xlib/SConscript', ]) -- cgit v1.2.3 From f44b30f1eb4896668c4fe91cd9b084ea63c9e915 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 20 Feb 2008 22:03:59 +0100 Subject: gallium: New file. --- src/gallium/include/pipe/p_pointer.h | 87 ++++++++++++++++++++++++++++++++++++ 1 file changed, 87 insertions(+) create mode 100644 src/gallium/include/pipe/p_pointer.h (limited to 'src') diff --git a/src/gallium/include/pipe/p_pointer.h b/src/gallium/include/pipe/p_pointer.h new file mode 100644 index 0000000000..e3927538e8 --- /dev/null +++ b/src/gallium/include/pipe/p_pointer.h @@ -0,0 +1,87 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_POINTER_H +#define P_POINTER_H + +#include "p_compiler.h" + +static INLINE intptr_t +pointer_to_intptr( const void *p ) +{ + union { + const void *p; + intptr_t i; + } pi; + pi.p = p; + return pi.i; +} + +static INLINE void * +intptr_to_pointer( intptr_t i ) +{ + union { + void *p; + intptr_t i; + } pi; + pi.i = i; + return pi.p; +} + +static INLINE uintptr_t +pointer_to_uintptr( const void *ptr ) +{ + union { + const void *p; + uintptr_t u; + } pu; + pu.p = ptr; + return pu.u; +} + +static INLINE void * +uintptr_to_pointer( uintptr_t u ) +{ + union { + void *p; + uintptr_t u; + } pu; + pu.u = u; + return pu.p; +} + +/** + * Return a pointer aligned to next multiple of N bytes. + */ +static INLINE void * +align_pointer( const void *unaligned, uintptr_t alignment ) +{ + uintptr_t aligned = (pointer_to_uintptr( unaligned ) + alignment - 1) & ~(alignment - 1); + return uintptr_to_pointer( aligned ); +} + +#endif /* P_POINTER_H */ -- cgit v1.2.3 From 8828dd1443c7d55697f78757b22be2733e059acf Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 20 Feb 2008 22:04:32 +0100 Subject: gallium: Move align_pointer() to p_pointer.h. --- src/gallium/include/pipe/p_util.h | 28 +--------------------------- 1 file changed, 1 insertion(+), 27 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index d7da2801c9..6f0dbdacd9 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -30,6 +30,7 @@ #include "p_compiler.h" #include "p_debug.h" +#include "p_pointer.h" #include @@ -114,33 +115,6 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) -/** - * Return a pointer aligned to next multiple of N bytes. - */ -static INLINE void * -align_pointer( void *unaligned, uint alignment ) -{ - if (sizeof(void *) == 64) { - union { - void *p; - uint64 u; - } pu; - pu.p = unaligned; - pu.u = (pu.u + alignment - 1) & ~(uint64) (alignment - 1); - return pu.p; - } - else { - /* 32-bit pointers */ - union { - void *p; - uint u; - } pu; - pu.p = unaligned; - pu.u = (pu.u + alignment - 1) & ~(alignment - 1); - return pu.p; - } -} - /** * Return memory on given byte alignment */ -- cgit v1.2.3 From 6f238275c7c19f7e287b47276e6b4060c270599f Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 20 Feb 2008 22:05:06 +0100 Subject: gallium: Define intptr_t for Windows platform. --- src/gallium/include/pipe/p_compiler.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index 30cd729c56..ab527f2afe 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -61,8 +61,10 @@ typedef long long int64_t; typedef unsigned long long uint64_t; #if defined(_WIN64) +typedef __int64 intptr_t; typedef unsigned __int64 uintptr_t; #else +typedef int intptr_t; typedef unsigned int uintptr_t; #endif -- cgit v1.2.3 From 69a7c9739bc0f11e66e11ab410d813fa69fe5fc9 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 20 Feb 2008 22:05:52 +0100 Subject: gallium: Silence compiler warnings on Windows. --- src/gallium/drivers/softpipe/sp_tex_sample.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 43d5085895..0ced585c7f 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -343,7 +343,7 @@ nearest_texcoord_unnorm(unsigned wrapMode, float s, unsigned size) switch (wrapMode) { case PIPE_TEX_WRAP_CLAMP: i = ifloor(s); - return CLAMP(i, 0, size-1); + return CLAMP(i, 0, (int) size-1); case PIPE_TEX_WRAP_CLAMP_TO_EDGE: /* fall-through */ case PIPE_TEX_WRAP_CLAMP_TO_BORDER: @@ -366,7 +366,7 @@ linear_texcoord_unnorm(unsigned wrapMode, float s, unsigned size, switch (wrapMode) { case PIPE_TEX_WRAP_CLAMP: /* Not exactly what the spec says, but it matches NVIDIA output */ - s = CLAMP(s - 0.5F, 0.0, (float) size - 1.0); + s = CLAMP(s - 0.5F, 0.0f, (float) size - 1.0f); *i0 = ifloor(s); *i1 = *i0 + 1; break; @@ -377,7 +377,7 @@ linear_texcoord_unnorm(unsigned wrapMode, float s, unsigned size, s -= 0.5F; *i0 = ifloor(s); *i1 = *i0 + 1; - if (*i1 > size - 1) + if (*i1 > (int) size - 1) *i1 = size - 1; break; default: -- cgit v1.2.3 From 57060bc1fa82e4e93d2affafecd98219be2f991f Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 20 Feb 2008 22:10:27 +0100 Subject: gallium: Silence compiler warnings on Windows. --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index dcbf76f600..4d33950e99 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -25,6 +25,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_debug.h" +#include "pipe/p_pointer.h" #include "rtasm_execmem.h" #include "rtasm_x86sse.h" @@ -34,7 +35,7 @@ static unsigned char *cptr( void (*label)() ) { - return (unsigned char *)(unsigned long)label; + return (unsigned char *) label; } @@ -46,7 +47,7 @@ static void do_realloc( struct x86_function *p ) p->csr = p->store; } else { - unsigned used = p->csr - p->store; + uintptr_t used = pointer_to_uintptr( p->csr ) - pointer_to_uintptr( p->store ); unsigned char *tmp = p->store; p->size *= 2; p->store = rtasm_exec_malloc(p->size); @@ -60,7 +61,7 @@ static void do_realloc( struct x86_function *p ) */ static unsigned char *reserve( struct x86_function *p, int bytes ) { - if (p->csr + bytes - p->store > p->size) + if (p->csr + bytes - p->store > (int) p->size) do_realloc(p); { @@ -135,7 +136,7 @@ static void emit_modrm( struct x86_function *p, case mod_INDIRECT: break; case mod_DISP8: - emit_1b(p, regmem.disp); + emit_1b(p, (char) regmem.disp); break; case mod_DISP32: emit_1i(p, regmem.disp); @@ -251,14 +252,14 @@ void x86_jcc( struct x86_function *p, enum x86_cc cc, unsigned char *label ) { - int offset = label - (x86_get_label(p) + 2); + intptr_t offset = pointer_to_intptr( label ) - (pointer_to_intptr( x86_get_label(p) ) + 2); if (offset <= 127 && offset >= -128) { emit_1ub(p, 0x70 + cc); emit_1b(p, (char) offset); } else { - offset = label - (x86_get_label(p) + 6); + offset = pointer_to_intptr( label ) - (pointer_to_intptr( x86_get_label(p) ) + 6); emit_2ub(p, 0x0f, 0x80 + cc); emit_1i(p, offset); } @@ -293,13 +294,13 @@ unsigned char *x86_call_forward( struct x86_function *p) void x86_fixup_fwd_jump( struct x86_function *p, unsigned char *fixup ) { - *(int *)(fixup - 4) = x86_get_label(p) - fixup; + *(int *)(fixup - 4) = pointer_to_intptr( x86_get_label(p) ) - pointer_to_intptr( fixup ); } void x86_jmp( struct x86_function *p, unsigned char *label) { emit_1ub(p, 0xe9); - emit_1i(p, label - x86_get_label(p) - 4); + emit_1i(p, pointer_to_intptr( label ) - pointer_to_intptr( x86_get_label(p) ) - 4); } #if 0 @@ -1207,7 +1208,7 @@ void (*x86_get_func( struct x86_function *p ))(void) { if (DISASSEM && p->store) debug_printf("disassemble %p %p\n", p->store, p->csr); - return (void (*)(void)) (unsigned long) p->store; + return (void (*)(void)) p->store; } #else -- cgit v1.2.3 From e0de82fbcb181f5c3e372ed66b692970a9e80766 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 20 Feb 2008 22:25:18 +0100 Subject: gallium: Fix build on Windows. --- src/gallium/drivers/softpipe/sp_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 6ba0f09e0a..3d26e8ca66 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -139,7 +139,7 @@ softpipe_texture_update(struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); uint unit; for (unit = 0; unit < PIPE_MAX_SAMPLERS; unit++) { - if (softpipe->texture[unit] == texture) { + if (&softpipe->texture[unit]->base == texture) { sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); } } -- cgit v1.2.3 From 2cf860b866d80595b7287d6991dc96abc3ca8dd3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 22 Feb 2008 16:22:41 +0900 Subject: gallium: MSVC fixes. --- src/gallium/auxiliary/cso_cache/cso_hash.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index b40217c524..e65d331a0b 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -134,7 +134,7 @@ static void cso_data_rehash(struct cso_hash_data *hash, int hint) hint = countBits(-hint); if (hint < MinNumBits) hint = MinNumBits; - hash->userNumBits = hint; + hash->userNumBits = (short)hint; while (primeForNumBits(hint) < (hash->size >> 1)) ++hint; } else if (hint < MinNumBits) { @@ -147,7 +147,7 @@ static void cso_data_rehash(struct cso_hash_data *hash, int hint) int oldNumBuckets = hash->numBuckets; int i = 0; - hash->numBits = hint; + hash->numBits = (short)hint; hash->numBuckets = primeForNumBits(hint); hash->buckets = malloc(sizeof(struct cso_node*) * hash->numBuckets); for (i = 0; i < hash->numBuckets; ++i) @@ -241,7 +241,7 @@ struct cso_hash * cso_hash_create(void) hash->data.d->buckets = 0; hash->data.d->size = 0; hash->data.d->nodeSize = sizeof(struct cso_node); - hash->data.d->userNumBits = MinNumBits; + hash->data.d->userNumBits = (short)MinNumBits; hash->data.d->numBits = 0; hash->data.d->numBuckets = 0; -- cgit v1.2.3 From 6c597238b2e168b63738ac8cc9167c1d09185aad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 22 Feb 2008 17:22:32 +0900 Subject: gallium: Add cso convenience routine (from Keith's patch). --- src/gallium/auxiliary/cso_cache/cso_cache.c | 20 ++++++++++++++++++++ src/gallium/auxiliary/cso_cache/cso_hash.h | 9 +++++++++ 2 files changed, 29 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 9e77e0774d..776ce6bacf 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -132,6 +132,26 @@ cso_find_state(struct cso_cache *sc, return cso_hash_find(hash, hash_key); } + +void *cso_hash_find_data_from_template( struct cso_hash *hash, + unsigned hash_key, + void *templ, + int size ) +{ + struct cso_hash_iter iter = cso_hash_find(hash, hash_key); + while (!cso_hash_iter_is_null(iter)) { + void *iter_data = cso_hash_iter_data(iter); + if (!memcmp(iter_data, templ, size)) { + /* Return the payload: + */ + return (unsigned char *)iter_data + size; + } + iter = cso_hash_iter_next(iter); + } + return NULL; +} + + struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, unsigned hash_key, enum cso_cache_type type, void *templ) diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h index b4aa111860..ffd99beba9 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.h +++ b/src/gallium/auxiliary/cso_cache/cso_hash.h @@ -59,4 +59,13 @@ void *cso_hash_iter_data(struct cso_hash_iter iter); struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter); struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); + +/* KW: a convenience routine: + */ +void *cso_hash_find_data_from_template( struct cso_hash *hash, + unsigned hash_key, + void *templ, + int size ); + + #endif -- cgit v1.2.3 From 901b03e84dce21f4241375da179b2199a3162e0c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 22 Feb 2008 18:28:30 +0900 Subject: gallium: Wrap decls in extern "C". --- src/gallium/auxiliary/cso_cache/cso_cache.h | 12 ++++++++++++ src/gallium/auxiliary/cso_cache/cso_hash.h | 10 ++++++++++ 2 files changed, 22 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h index 116e2eaa2c..58664cdd94 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.h +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -37,8 +37,15 @@ #include "pipe/p_state.h" +#ifdef __cplusplus +extern "C" { +#endif + + struct cso_hash; +struct cso_hash_iter; + struct cso_cache { struct cso_hash *blend_hash; struct cso_hash *depth_stencil_hash; @@ -104,4 +111,9 @@ struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, void * cso_take_state(struct cso_cache *sc, unsigned hash_key, enum cso_cache_type type); + +#ifdef __cplusplus +} +#endif + #endif diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h index ffd99beba9..2e8b69675c 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.h +++ b/src/gallium/auxiliary/cso_cache/cso_hash.h @@ -33,6 +33,12 @@ #ifndef CSO_HASH_H #define CSO_HASH_H + +#ifdef __cplusplus +extern "C" { +#endif + + struct cso_hash; struct cso_node; @@ -68,4 +74,8 @@ void *cso_hash_find_data_from_template( struct cso_hash *hash, int size ); +#ifdef __cplusplus +} +#endif + #endif -- cgit v1.2.3 From f1bef2cba2e6e4c0988b05306dd1435b8c0df3d5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 22 Feb 2008 18:45:20 +0900 Subject: gallium: Countour MSVC's pickyness for structures returned by value. --- src/gallium/auxiliary/cso_cache/cso_cache.h | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h index 58664cdd94..b4f4f3ae41 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.h +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -36,16 +36,14 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "cso_hash.h" /* for cso_hash_iter, as MSVC requires structures returned by value to be fully defined */ + #ifdef __cplusplus extern "C" { #endif -struct cso_hash; - -struct cso_hash_iter; - struct cso_cache { struct cso_hash *blend_hash; struct cso_hash *depth_stencil_hash; -- cgit v1.2.3 From b4d050ffccf46e5e7b40d8a2f3868cc73fcf110e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Feb 2008 08:50:37 -0700 Subject: cell: fix build: s/dest/vertex/ --- src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index f7ef72e5a2..42cc47cbfe 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -123,7 +123,7 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) for (j = 0; j < n; j++) { vs->elts[j] = draw->vs.queue[i + j].elt; - vs->vOut[j] = (uintptr_t) draw->vs.queue[i + j].dest; + vs->vOut[j] = (uintptr_t) draw->vs.queue[i + j].vertex; } for (/* empty */; j < SPU_VERTS_PER_BATCH; j++) { -- cgit v1.2.3 From 0bb53709e837ac709f31ace8852599e885f4d090 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Feb 2008 08:52:24 -0700 Subject: cell: added function prototypes to silence warnings --- src/gallium/drivers/cell/ppu/cell_context.h | 2 ++ src/gallium/drivers/cell/ppu/cell_flush.h | 4 ++++ 2 files changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 3b687bb868..1433a4925f 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -134,6 +134,8 @@ extern void cell_vertex_shader_queue_flush(struct draw_context *draw); +/* XXX find a better home for this */ +extern void cell_update_vertex_fetch(struct draw_context *draw); #endif /* CELL_CONTEXT_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_flush.h b/src/gallium/drivers/cell/ppu/cell_flush.h index eda351b1cb..7f940ae76b 100644 --- a/src/gallium/drivers/cell/ppu/cell_flush.h +++ b/src/gallium/drivers/cell/ppu/cell_flush.h @@ -35,4 +35,8 @@ cell_flush(struct pipe_context *pipe, unsigned flags); extern void cell_flush_int(struct pipe_context *pipe, unsigned flags); +extern void +cell_flush_buffer_range(struct cell_context *cell, void *ptr, + unsigned size); + #endif -- cgit v1.2.3 From 12c14c31b71d4bac494c4470a34e28ec66309254 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Feb 2008 08:56:55 -0700 Subject: gallium: fix brokenb build --- src/gallium/drivers/softpipe/sp_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 3d26e8ca66..6ba0f09e0a 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -139,7 +139,7 @@ softpipe_texture_update(struct pipe_context *pipe, struct softpipe_context *softpipe = softpipe_context(pipe); uint unit; for (unit = 0; unit < PIPE_MAX_SAMPLERS; unit++) { - if (&softpipe->texture[unit]->base == texture) { + if (softpipe->texture[unit] == texture) { sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); } } -- cgit v1.2.3 From 24f3d7de29c5cd30ced7c3ce7293902460a82e0b Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Feb 2008 14:07:12 -0700 Subject: i915: include p_debug.h (resolves undefined assert()) --- src/gallium/winsys/dri/intel/intel_batchbuffer.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.h b/src/gallium/winsys/dri/intel/intel_batchbuffer.h index 82feafa21f..caf6870a3c 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.h +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.h @@ -28,6 +28,7 @@ #ifndef INTEL_BATCHBUFFER_H #define INTEL_BATCHBUFFER_H +#include "pipe/p_debug.h" #include "pipe/p_compiler.h" #include "dri_bufmgr.h" -- cgit v1.2.3 From 8cd7c1d03ce045bfa39471c3f77a31030195b899 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Feb 2008 14:07:33 -0700 Subject: i915: include stdio.h --- src/gallium/winsys/dri/intel/intel_batchbuffer.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c index 49e04d81ec..5830b88b37 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.c +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.c @@ -26,6 +26,7 @@ **************************************************************************/ #include +#include #include "intel_batchbuffer.h" #include "intel_context.h" #include "intel_screen.h" -- cgit v1.2.3 From c74900ee5d80c7c2b7cbe4ed87395526a742a13e Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Feb 2008 16:48:05 -0700 Subject: gallium/i915: overhaul of fragment shader compilation, constant/immediate allocation Before, fragment shaders were translated to i915 hw code at bind time, rather than create time. Now there's an i915_fragment_shader struct with the expected contents that's created by i915_create_fs_state(). Translation to i915 code takes place there too. Immediates are handled correctly now. During program translation we keep track of which constant buffer slots are free (i.e. not referenced by the shader). Then the TGSI immediates and ancillary immediates (introduced for SIN/COS/etc) are put into those free slots. When it's time to upload the constant buffer, use the fp->constant_flags[] array to determine if we should grab an immediate from the shader, or a user-defined parameter from the gallium constant buffer. --- src/gallium/drivers/i915simple/i915_context.h | 42 ++++++- src/gallium/drivers/i915simple/i915_fpc.h | 23 ++-- src/gallium/drivers/i915simple/i915_fpc_emit.c | 101 +++++----------- .../drivers/i915simple/i915_fpc_translate.c | 132 +++++++++++++-------- src/gallium/drivers/i915simple/i915_state.c | 36 ++++-- .../drivers/i915simple/i915_state_derived.c | 3 +- src/gallium/drivers/i915simple/i915_state_emit.c | 39 ++++-- 7 files changed, 226 insertions(+), 150 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 2d876925b2..5cc38cdc85 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -79,6 +79,40 @@ #define I915_MAX_CONSTANT 32 +/** See constant_flags[] below */ +#define I915_CONSTFLAG_USER 0x1f + + +/** + * Subclass of pipe_shader_state + */ +struct i915_fragment_shader +{ + struct pipe_shader_state state; + uint *program; + uint program_len; + + /** + * constants introduced during translation. + * These are placed at the end of the constant buffer and grow toward + * the beginning (eg: slot 31, 30 29, ...) + * User-provided constants start at 0. + * This allows both types of constants to co-exist (until there's too many) + * and doesn't require regenerating/changing the fragment program to + * shuffle constants around. + */ + uint num_constants; + float constants[I915_MAX_CONSTANT][4]; + + /** + * Status of each constant + * if I915_CONSTFLAG_PARAM, the value must be taken from the corresponding + * slot of the user's constant buffer. (set by pipe->set_constant_buffer()) + * Else, the bitmask indicates which components are occupied by immediates. + */ + ubyte constant_flags[I915_MAX_CONSTANT]; +}; + struct i915_cache_context; @@ -93,11 +127,6 @@ struct i915_state float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; /** number of constants passed in through a constant buffer */ uint num_user_constants[PIPE_SHADER_TYPES]; - /** user constants, plus extra constants from shader translation */ - uint num_constants[PIPE_SHADER_TYPES]; - - uint *program; - uint program_len; /* texture sampler state */ unsigned sampler[I915_TEX_UNITS][3]; @@ -187,7 +216,8 @@ struct i915_context const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; const struct i915_depth_stencil_state *depth_stencil; const struct i915_rasterizer_state *rasterizer; - const struct pipe_shader_state *fs; + + struct i915_fragment_shader *fs; struct pipe_blend_color blend_color; struct pipe_clip_state clip; diff --git a/src/gallium/drivers/i915simple/i915_fpc.h b/src/gallium/drivers/i915simple/i915_fpc.h index 8c7b68aefb..250dfe6dbf 100644 --- a/src/gallium/drivers/i915simple/i915_fpc.h +++ b/src/gallium/drivers/i915simple/i915_fpc.h @@ -44,9 +44,16 @@ * Program translation state */ struct i915_fp_compile { - const struct pipe_shader_state *shader; + struct i915_fragment_shader *shader; /* the shader we're compiling */ - struct vertex_info *vertex_info; + boolean used_constants[I915_MAX_CONSTANT]; + + /** maps TGSI immediate index to constant slot */ + uint num_immediates; + uint immediates_map[I915_MAX_CONSTANT]; + float immediates[I915_MAX_CONSTANT][4]; + + boolean first_instruction; uint declarations[I915_PROGRAM_SIZE]; uint program[I915_PROGRAM_SIZE]; @@ -57,11 +64,6 @@ struct i915_fp_compile { uint output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; uint output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; - /** points into the i915->current.constants array: */ - float (*constants)[4]; - uint num_constants; - uint constant_flags[I915_MAX_CONSTANT]; /**< status of each constant */ - uint *csr; /**< Cursor, points into program. */ uint *decl; /**< Cursor, points into declarations. */ @@ -155,7 +157,9 @@ swizzle(int reg, uint x, uint y, uint z, uint w) /*********************************************************************** * Public interface for the compiler */ -extern void i915_translate_fragment_program( struct i915_context *i915 ); +extern void +i915_translate_fragment_program( struct i915_context *i915, + struct i915_fragment_shader *fs); @@ -206,8 +210,5 @@ extern void i915_disassemble_program(const uint * program, uint sz); extern void i915_program_error(struct i915_fp_compile *p, const char *msg, ...); -extern void -i915_translate_fragment_program(struct i915_context *i915); - #endif diff --git a/src/gallium/drivers/i915simple/i915_fpc_emit.c b/src/gallium/drivers/i915simple/i915_fpc_emit.c index 74924ff0a1..a59ee23403 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_emit.c +++ b/src/gallium/drivers/i915simple/i915_fpc_emit.c @@ -61,8 +61,6 @@ (REG_NR_MASK << UREG_NR_SHIFT)) -#define I915_CONSTFLAG_PARAM 0x1f - uint i915_get_temp(struct i915_fp_compile *p) { @@ -235,6 +233,7 @@ uint i915_emit_texld( struct i915_fp_compile *p, uint i915_emit_const1f(struct i915_fp_compile * p, float c0) { + struct i915_fragment_shader *ifs = p->shader; unsigned reg, idx; if (c0 == 0.0) @@ -243,15 +242,15 @@ i915_emit_const1f(struct i915_fp_compile * p, float c0) return swizzle(UREG(REG_TYPE_R, 0), ONE, ONE, ONE, ONE); for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == I915_CONSTFLAG_PARAM) + if (ifs->constant_flags[reg] == I915_CONSTFLAG_USER) continue; for (idx = 0; idx < 4; idx++) { - if (!(p->constant_flags[reg] & (1 << idx)) || - p->constants[reg][idx] == c0) { - p->constants[reg][idx] = c0; - p->constant_flags[reg] |= 1 << idx; - if (reg + 1 > p->num_constants) - p->num_constants = reg + 1; + if (!(ifs->constant_flags[reg] & (1 << idx)) || + ifs->constants[reg][idx] == c0) { + ifs->constants[reg][idx] = c0; + ifs->constant_flags[reg] |= 1 << idx; + if (reg + 1 > ifs->num_constants) + ifs->num_constants = reg + 1; return swizzle(UREG(REG_TYPE_CONST, reg), idx, ZERO, ZERO, ONE); } } @@ -264,6 +263,7 @@ i915_emit_const1f(struct i915_fp_compile * p, float c0) uint i915_emit_const2f(struct i915_fp_compile * p, float c0, float c1) { + struct i915_fragment_shader *ifs = p->shader; unsigned reg, idx; if (c0 == 0.0) @@ -277,16 +277,16 @@ i915_emit_const2f(struct i915_fp_compile * p, float c0, float c1) return swizzle(i915_emit_const1f(p, c0), X, ONE, Z, W); for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == 0xf || - p->constant_flags[reg] == I915_CONSTFLAG_PARAM) + if (ifs->constant_flags[reg] == 0xf || + ifs->constant_flags[reg] == I915_CONSTFLAG_USER) continue; for (idx = 0; idx < 3; idx++) { - if (!(p->constant_flags[reg] & (3 << idx))) { - p->constants[reg][idx + 0] = c0; - p->constants[reg][idx + 1] = c1; - p->constant_flags[reg] |= 3 << idx; - if (reg + 1 > p->num_constants) - p->num_constants = reg + 1; + if (!(ifs->constant_flags[reg] & (3 << idx))) { + ifs->constants[reg][idx + 0] = c0; + ifs->constants[reg][idx + 1] = c1; + ifs->constant_flags[reg] |= 3 << idx; + if (reg + 1 > ifs->num_constants) + ifs->num_constants = reg + 1; return swizzle(UREG(REG_TYPE_CONST, reg), idx, idx + 1, ZERO, ONE); } } @@ -302,25 +302,26 @@ uint i915_emit_const4f(struct i915_fp_compile * p, float c0, float c1, float c2, float c3) { + struct i915_fragment_shader *ifs = p->shader; unsigned reg; for (reg = 0; reg < I915_MAX_CONSTANT; reg++) { - if (p->constant_flags[reg] == 0xf && - p->constants[reg][0] == c0 && - p->constants[reg][1] == c1 && - p->constants[reg][2] == c2 && - p->constants[reg][3] == c3) { + if (ifs->constant_flags[reg] == 0xf && + ifs->constants[reg][0] == c0 && + ifs->constants[reg][1] == c1 && + ifs->constants[reg][2] == c2 && + ifs->constants[reg][3] == c3) { return UREG(REG_TYPE_CONST, reg); } - else if (p->constant_flags[reg] == 0) { - - p->constants[reg][0] = c0; - p->constants[reg][1] = c1; - p->constants[reg][2] = c2; - p->constants[reg][3] = c3; - p->constant_flags[reg] = 0xf; - if (reg + 1 > p->num_constants) - p->num_constants = reg + 1; + else if (ifs->constant_flags[reg] == 0) { + + ifs->constants[reg][0] = c0; + ifs->constants[reg][1] = c1; + ifs->constants[reg][2] = c2; + ifs->constants[reg][3] = c3; + ifs->constant_flags[reg] = 0xf; + if (reg + 1 > ifs->num_constants) + ifs->num_constants = reg + 1; return UREG(REG_TYPE_CONST, reg); } } @@ -335,41 +336,3 @@ i915_emit_const4fv(struct i915_fp_compile * p, const float * c) { return i915_emit_const4f(p, c[0], c[1], c[2], c[3]); } - - -#if 00000/*UNUSED*/ -/* Reserve a slot in the constant file for a Mesa state parameter. - * These will later need to be tracked on statechanges, but that is - * done elsewhere. - */ -uint -i915_emit_param4fv(struct i915_fp_compile * p, const float * values) -{ - struct i915_fragment_program *fp = p->fp; - int i; - - for (i = 0; i < fp->nr_params; i++) { - if (fp->param[i].values == values) - return UREG(REG_TYPE_CONST, fp->param[i].reg); - } - - if (p->constants->nr_constants == I915_MAX_CONSTANT || - fp->nr_params == I915_MAX_CONSTANT) { - i915_program_error(p, "i915_emit_param4fv: out of constants\n"); - return 0; - } - - { - int reg = p->constants->nr_constants++; - int i = fp->nr_params++; - - assert (p->constant_flags[reg] == 0); - p->constant_flags[reg] = I915_CONSTFLAG_PARAM; - - fp->param[i].values = values; - fp->param[i].reg = reg; - - return UREG(REG_TYPE_CONST, reg); - } -} -#endif diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 6c1524c768..afe7e25dce 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -34,6 +34,7 @@ #include "pipe/p_shader_tokens.h" #include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_dump.h" #include "draw/draw_vertex.h" @@ -97,19 +98,19 @@ negate(int reg, int x, int y, int z, int w) } +/** + * In the event of a translation failure, we'll generate a simple color + * pass-through program. + */ static void -i915_use_passthrough_shader(struct i915_context *i915) +i915_use_passthrough_shader(struct i915_fragment_shader *fs) { - debug_printf("**** Using i915 pass-through fragment shader\n"); - - i915->current.program = (uint *) MALLOC(sizeof(passthrough)); - if (i915->current.program) { - memcpy(i915->current.program, passthrough, sizeof(passthrough)); - i915->current.program_len = Elements(passthrough); + fs->program = (uint *) MALLOC(sizeof(passthrough)); + if (fs->program) { + memcpy(fs->program, passthrough, sizeof(passthrough)); + fs->program_len = Elements(passthrough); } - - i915->current.num_constants[PIPE_SHADER_FRAGMENT] = 0; - i915->current.num_user_constants[PIPE_SHADER_FRAGMENT] = 0; + fs->num_constants = 0; } @@ -161,9 +162,6 @@ src_vector(struct i915_fp_compile *p, * We also use a texture coordinate to pass wpos when possible. */ - /* use vertex format info to map a slot number to a VF attrib */ - assert(index < p->vertex_info->num_attribs); - sem_name = p->input_semantic_name[index]; sem_ind = p->input_semantic_index[index]; @@ -201,7 +199,8 @@ src_vector(struct i915_fp_compile *p, break; case TGSI_FILE_IMMEDIATE: - /* XXX unfinished - need to append immediates onto const buffer */ + assert(index < p->num_immediates); + index = p->immediates_map[index]; /* fall-through */ case TGSI_FILE_CONSTANT: src = UREG(REG_TYPE_CONST, index); @@ -896,6 +895,7 @@ static void i915_translate_instructions(struct i915_fp_compile *p, const struct tgsi_token *tokens) { + struct i915_fragment_shader *ifs = p->shader; struct tgsi_parse_context parse; tgsi_parse_init( &parse, tokens ); @@ -928,13 +928,54 @@ i915_translate_instructions(struct i915_fp_compile *p, p->output_semantic_name[ind] = sem; p->output_semantic_index[ind] = semi; } + else if (parse.FullToken.FullDeclaration.Declaration.File + == TGSI_FILE_CONSTANT) { + uint i; + for (i = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + i <= parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + i++) { + assert(ifs->constant_flags[i] == 0x0); + ifs->constant_flags[i] = I915_CONSTFLAG_USER; + ifs->num_constants = MAX2(ifs->num_constants, i + 1); + } + } break; case TGSI_TOKEN_TYPE_IMMEDIATE: - /* XXX append the immediate to the const buffer... */ + { + const struct tgsi_full_immediate *imm + = &parse.FullToken.FullImmediate; + const uint pos = p->num_immediates++; + uint j; + for (j = 0; j < imm->Immediate.Size; j++) { + p->immediates[pos][j] = imm->u.ImmediateFloat32[j].Float; + } + } break; case TGSI_TOKEN_TYPE_INSTRUCTION: + if (p->first_instruction) { + /* resolve location of immediates */ + uint i, j; + for (i = 0; i < p->num_immediates; i++) { + /* find constant slot for this immediate */ + for (j = 0; j < I915_MAX_CONSTANT; j++) { + if (ifs->constant_flags[j] == 0x0) { + memcpy(ifs->constants[j], + p->immediates[i], + 4 * sizeof(float)); + /*printf("immediate %d maps to const %d\n", i, j);*/ + ifs->constant_flags[j] = 0xf; /* all four comps used */ + p->immediates_map[i] = j; + ifs->num_constants = MAX2(ifs->num_constants, j + 1); + break; + } + } + } + + p->first_instruction = FALSE; + } + i915_translate_instruction(p, &parse.FullToken.FullInstruction); break; @@ -950,27 +991,28 @@ i915_translate_instructions(struct i915_fp_compile *p, static struct i915_fp_compile * i915_init_compile(struct i915_context *i915, - const struct pipe_shader_state *fs) + struct i915_fragment_shader *ifs) { struct i915_fp_compile *p = CALLOC_STRUCT(i915_fp_compile); - p->shader = i915->fs; + p->shader = ifs; - p->vertex_info = &i915->current.vertex_info; - - /* new constants found during translation get appended after the - * user-provided constants. + /* Put new constants at end of const buffer, growing downward. + * The problem is we don't know how many user-defined constants might + * be specified with pipe->set_constant_buffer(). + * Should pre-scan the user's program to determine the highest-numbered + * constant referenced. */ - p->constants = i915->current.constants[PIPE_SHADER_FRAGMENT]; - p->num_constants = i915->current.num_user_constants[PIPE_SHADER_FRAGMENT]; + ifs->num_constants = 0; + memset(ifs->constant_flags, 0, sizeof(ifs->constant_flags)); + + p->first_instruction = TRUE; p->nr_tex_indirect = 1; /* correct? */ p->nr_tex_insn = 0; p->nr_alu_insn = 0; p->nr_decl_insn = 0; - memset(p->constant_flags, 0, sizeof(p->constant_flags)); - p->csr = p->program; p->decl = p->declarations; p->decl_s = 0; @@ -993,6 +1035,7 @@ i915_init_compile(struct i915_context *i915, static void i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) { + struct i915_fragment_shader *ifs = p->shader; unsigned long program_size = (unsigned long) (p->csr - p->program); unsigned long decl_size = (unsigned long) (p->decl - p->declarations); @@ -1008,19 +1051,13 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) if (p->nr_decl_insn > I915_MAX_DECL_INSN) i915_program_error(p, "Exceeded max DECL instructions"); - /* free old program, if present */ - if (i915->current.program) { - FREE(i915->current.program); - i915->current.program_len = 0; - } - if (p->error) { p->NumNativeInstructions = 0; p->NumNativeAluInstructions = 0; p->NumNativeTexInstructions = 0; p->NumNativeTexIndirections = 0; - i915_use_passthrough_shader(i915); + i915_use_passthrough_shader(ifs); } else { p->NumNativeInstructions @@ -1034,24 +1071,20 @@ i915_fini_compile(struct i915_context *i915, struct i915_fp_compile *p) /* Copy compilation results to fragment program struct: */ - i915->current.program + assert(!ifs->program); + ifs->program = (uint *) MALLOC((program_size + decl_size) * sizeof(uint)); - if (i915->current.program) { - i915->current.program_len = program_size + decl_size; + if (ifs->program) { + ifs->program_len = program_size + decl_size; - memcpy(i915->current.program, + memcpy(ifs->program, p->declarations, decl_size * sizeof(uint)); - memcpy(i915->current.program + decl_size, + memcpy(ifs->program + decl_size, p->program, program_size * sizeof(uint)); } - - /* update number of constants */ - i915->current.num_constants[PIPE_SHADER_FRAGMENT] = p->num_constants; - assert(i915->current.num_constants[PIPE_SHADER_FRAGMENT] - >= i915->current.num_user_constants[PIPE_SHADER_FRAGMENT]); } /* Release the compilation struct: @@ -1085,7 +1118,7 @@ i915_find_wpos_space(struct i915_fp_compile *p) i915_program_error(p, "No free texcoord for wpos value"); } #else - if (p->shader->input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { + if (p->shader->state.input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { /* frag shader using the fragment position input */ #if 0 assert(0); @@ -1106,7 +1139,7 @@ static void i915_fixup_depth_write(struct i915_fp_compile *p) { /* XXX assuming pos/depth is always in output[0] */ - if (p->shader->output_semantic_name[0] == TGSI_SEMANTIC_POSITION) { + if (p->shader->state.output_semantic_name[0] == TGSI_SEMANTIC_POSITION) { const uint depth = UREG(REG_TYPE_OD, 0); i915_emit_arith(p, @@ -1121,13 +1154,18 @@ i915_fixup_depth_write(struct i915_fp_compile *p) void -i915_translate_fragment_program( struct i915_context *i915 ) +i915_translate_fragment_program( struct i915_context *i915, + struct i915_fragment_shader *fs) { - struct i915_fp_compile *p = i915_init_compile(i915, i915->fs); - const struct tgsi_token *tokens = i915->fs->tokens; + struct i915_fp_compile *p = i915_init_compile(i915, fs); + const struct tgsi_token *tokens = fs->state.tokens; i915_find_wpos_space(p); +#if 0 + tgsi_dump(tokens, 0); +#endif + i915_translate_instructions(p, tokens); i915_fixup_depth_write(p); diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index e055eed7e0..e4288d4e31 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -38,6 +38,7 @@ #include "i915_reg.h" #include "i915_state.h" #include "i915_state_inlines.h" +#include "i915_fpc.h" /* The i915 (and related graphics cores) do not support GL_CLAMP. The @@ -416,26 +417,47 @@ static void i915_set_polygon_stipple( struct pipe_context *pipe, } -static void * i915_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *templ) + +static void * +i915_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *templ) { - return 0; + struct i915_context *i915 = i915_context(pipe); + struct i915_fragment_shader *ifs = CALLOC_STRUCT(i915_fragment_shader); + if (!ifs) + return NULL; + + ifs->state = *templ; + + /* The shader's compiled to i915 instructions here */ + i915_translate_fragment_program(i915, ifs); + + return ifs; } -static void i915_bind_fs_state(struct pipe_context *pipe, void *fs) +static void +i915_bind_fs_state(struct pipe_context *pipe, void *shader) { struct i915_context *i915 = i915_context(pipe); - i915->fs = (struct pipe_shader_state *)fs; + i915->fs = (struct i915_fragment_shader*) shader; i915->dirty |= I915_NEW_FS; } -static void i915_delete_fs_state(struct pipe_context *pipe, void *shader) +static +void i915_delete_fs_state(struct pipe_context *pipe, void *shader) { - /*do nothing*/ + struct i915_fragment_shader *ifs = (struct i915_fragment_shader *) shader; + + if (ifs->program) + FREE(ifs->program); + ifs->program_len = 0; + + FREE(ifs); } + static void * i915_create_vs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) diff --git a/src/gallium/drivers/i915simple/i915_state_derived.c b/src/gallium/drivers/i915simple/i915_state_derived.c index 4767584fc6..f654f543cc 100644 --- a/src/gallium/drivers/i915simple/i915_state_derived.c +++ b/src/gallium/drivers/i915simple/i915_state_derived.c @@ -43,7 +43,7 @@ */ static void calculate_vertex_layout( struct i915_context *i915 ) { - const struct pipe_shader_state *fs = i915->fs; + const struct pipe_shader_state *fs = &i915->fs->state; const enum interp_mode colorInterp = i915->rasterizer->color_interp; struct vertex_info vinfo; uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; @@ -164,7 +164,6 @@ void i915_update_derived( struct i915_context *i915 ) i915_update_dynamic( i915 ); if (i915->dirty & I915_NEW_FS) { - i915_translate_fragment_program(i915); i915->hardware_dirty |= I915_HW_PROGRAM; /* XXX right? */ } diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index 3339287f49..6bbaac4e34 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -99,7 +99,11 @@ i915_emit_hardware_state(struct i915_context *i915 ) 2 + I915_TEX_UNITS*3 + 2 + I915_TEX_UNITS*3 + 2 + I915_MAX_CONSTANT*4 + +#if 0 i915->current.program_len + +#else + i915->fs->program_len + +#endif 6 ) * 3/2; /* plus 50% margin */ const unsigned relocs = ( I915_TEX_UNITS + @@ -325,15 +329,34 @@ i915_emit_hardware_state(struct i915_context *i915 ) /* 2 + I915_MAX_CONSTANT*4 dwords, 0 relocs */ if (i915->hardware_dirty & I915_HW_PROGRAM) { - const uint nr = i915->current.num_constants[PIPE_SHADER_FRAGMENT]; - assert(nr <= I915_MAX_CONSTANT); - if (nr > 0) { - const uint *c - = (const uint *) i915->current.constants[PIPE_SHADER_FRAGMENT]; + /* Collate the user-defined constants with the fragment shader's + * immediates according to the constant_flags[] array. + */ + const uint nr = i915->fs->num_constants; + if (nr) { uint i; + OUT_BATCH( _3DSTATE_PIXEL_SHADER_CONSTANTS | (nr * 4) ); OUT_BATCH( (1 << (nr - 1)) | ((1 << (nr - 1)) - 1) ); + for (i = 0; i < nr; i++) { + const uint *c; + if (i915->fs->constant_flags[i] == I915_CONSTFLAG_USER) { + /* grab user-defined constant */ + c = (uint *) i915->current.constants[PIPE_SHADER_FRAGMENT][i]; + } + else { + /* emit program constant */ + c = (uint *) i915->fs->constants[i]; + } +#if 0 /* debug */ + { + float *f = (float *) c; + printf("Const %2d: %f %f %f %f %s\n", i, f[0], f[1], f[2], f[3], + (i915->fs->constant_flags[i] == I915_CONSTFLAG_USER + ? "user" : "immediate")); + } +#endif OUT_BATCH(*c++); OUT_BATCH(*c++); OUT_BATCH(*c++); @@ -348,9 +371,9 @@ i915_emit_hardware_state(struct i915_context *i915 ) { uint i; /* we should always have, at least, a pass-through program */ - assert(i915->current.program_len > 0); - for (i = 0; i < i915->current.program_len; i++) { - OUT_BATCH(i915->current.program[i]); + assert(i915->fs->program_len > 0); + for (i = 0; i < i915->fs->program_len; i++) { + OUT_BATCH(i915->fs->program[i]); } } -- cgit v1.2.3 From e9276efafe43219d7af548ce7f5d2440e19836b0 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Feb 2008 17:22:10 -0700 Subject: gallium: fix bug in which wide point stage overrode the aapoint stage Also, simplify the logic a bit. --- src/gallium/auxiliary/draw/draw_validate.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index efd6793f2b..3a19dd4cd7 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -45,6 +45,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) struct draw_stage *next = draw->pipeline.rasterize; int need_det = 0; int precalc_flat = 0; + boolean wide_lines, wide_points; /* Set the validate's next stage to the rasterize stage, so that it * can be found later if needed for flushing. @@ -68,9 +69,18 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) next = draw->pipeline.aapoint; } - if ((draw->rasterizer->line_width != 1.0 && draw->convert_wide_lines - && !draw->rasterizer->line_smooth) || - (draw->rasterizer->point_size != 1.0 && draw->convert_wide_points) || + /* drawing wide lines? */ + wide_lines = (draw->rasterizer->line_width != 1.0 + && draw->convert_wide_lines + && !draw->rasterizer->line_smooth); + + /* drawing large points? */ + wide_points = (draw->rasterizer->point_size != 1.0 + && draw->convert_wide_points + && !draw->pipeline.aapoint); + + if (wide_lines || + wide_points || draw->rasterizer->point_sprite) { draw->pipeline.wide->next = next; next = draw->pipeline.wide; -- cgit v1.2.3 From aec315f05f860337bccfce827a7c1c80960dd476 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 22 Feb 2008 18:34:31 -0700 Subject: gallium: fix a state validation bug found w/ pointblast.c --- src/gallium/drivers/softpipe/sp_prim_setup.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 7b1e131ee1..b6a3fddb29 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -1165,6 +1165,10 @@ static void setup_begin( struct draw_stage *stage ) struct softpipe_context *sp = setup->softpipe; const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; + if (sp->dirty) { + softpipe_update_derived(sp); + } + setup->quad.nr_attrs = fs->num_inputs; sp->quad.first->begin(sp->quad.first); -- cgit v1.2.3 From bed547cb8223659a7f0d15a91c2edb5f822ba855 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 23 Feb 2008 14:14:20 +0900 Subject: Fix typo. --- src/SConscript | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/SConscript b/src/SConscript index e2ee50dd6e..8e914af4f0 100644 --- a/src/SConscript +++ b/src/SConscript @@ -2,7 +2,7 @@ Import('*') SConscript('gallium/SConscript') -if 'mesa' in env['state_trackers']: +if 'mesa' in env['statetrackers']: SConscript('mesa/SConscript') SConscript('gallium/winsys/SConscript') -- cgit v1.2.3 From e8de5c70e3370e9112a5facc870075eea60c4c46 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 23 Feb 2008 14:14:54 +0900 Subject: Bring in several forgotten MSVC fixes. --- src/gallium/auxiliary/pipebuffer/pb_buffer.h | 1 + src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c | 4 ++-- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 2 +- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 4 +++- src/gallium/drivers/softpipe/sp_fs_exec.c | 4 ++-- src/gallium/drivers/softpipe/sp_fs_llvm.c | 2 +- src/gallium/include/pipe/p_compiler.h | 8 ++++++++ 7 files changed, 18 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h index 97beb5f72a..f5b5f4052f 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h @@ -166,6 +166,7 @@ static INLINE void pb_destroy(struct pb_buffer *buf) { assert(buf); + assert(buf->vtbl); buf->vtbl->destroy(buf); } diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index ff4fd123f3..66256f3fa7 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -192,8 +192,8 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, } /* Some sanity checks */ - assert(0 <= mm_buf->block->ofs && mm_buf->block->ofs < mm->size); - assert(size <= mm_buf->block->size && mm_buf->block->ofs + mm_buf->block->size <= mm->size); + assert(0 <= (unsigned)mm_buf->block->ofs && (unsigned)mm_buf->block->ofs < mm->size); + assert(size <= (unsigned)mm_buf->block->size && (unsigned)mm_buf->block->ofs + (unsigned)mm_buf->block->size <= mm->size); _glthread_UNLOCK_MUTEX(mm->mutex); return SUPER(mm_buf); diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index d7b18dc9c5..ac52441400 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -2455,7 +2455,7 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) /* execute instructions, until pc is set to -1 */ while (pc != -1) { - assert(pc < mach->NumInstructions); + assert(pc < (int) mach->NumInstructions); exec_instruction( mach, mach->Instructions + pc, &pc ); } diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index b5c54847e0..ff74e6117c 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -299,7 +299,8 @@ static const char *TGSI_SEMANTICS[] = "SEMANTIC_BCOLOR", "SEMANTIC_FOG", "SEMANTIC_PSIZE", - "SEMANTIC_GENERIC," + "SEMANTIC_GENERIC", + "SEMANTIC_NORMAL" }; static const char *TGSI_SEMANTICS_SHORT[] = @@ -310,6 +311,7 @@ static const char *TGSI_SEMANTICS_SHORT[] = "FOG", "PSIZE", "GENERIC", + "NORMAL" }; static const char *TGSI_IMMS[] = diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index 8cb0534342..d5bd7a702f 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -81,7 +81,7 @@ sp_setup_pos_vector(const struct tgsi_interp_coef *coef, static void -exec_prepare( struct sp_fragment_shader *base, +exec_prepare( const struct sp_fragment_shader *base, struct tgsi_exec_machine *machine, struct tgsi_sampler *samplers ) { @@ -98,7 +98,7 @@ exec_prepare( struct sp_fragment_shader *base, * interface: */ static unsigned -exec_run( struct sp_fragment_shader *base, +exec_run( const struct sp_fragment_shader *base, struct tgsi_exec_machine *machine, struct quad_header *quad ) { diff --git a/src/gallium/drivers/softpipe/sp_fs_llvm.c b/src/gallium/drivers/softpipe/sp_fs_llvm.c index 22da471453..34b2b7d4e2 100644 --- a/src/gallium/drivers/softpipe/sp_fs_llvm.c +++ b/src/gallium/drivers/softpipe/sp_fs_llvm.c @@ -146,7 +146,7 @@ shade_quad_llvm(struct quad_stage *qs, unsigned -run_llvm_fs( struct sp_fragment_shader *base, +run_llvm_fs( const struct sp_fragment_shader *base, struct foo *machine ) { } diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index ab527f2afe..91f3d2ac2d 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -42,6 +42,14 @@ #endif +#if defined(__MSC__) + +/* Avoid 'expression is always true' warning */ +#pragma warning(disable: 4296) + +#endif /* __MSC__ */ + + typedef unsigned int uint; typedef unsigned char ubyte; typedef unsigned char boolean; -- cgit v1.2.3 From 58a3d7dfd94453c25607106835fbbb3a54d42306 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 23 Feb 2008 19:49:08 +0900 Subject: Revamp scons usage. --- src/gallium/SConscript | 16 ++++------------ src/gallium/winsys/SConscript | 2 +- 2 files changed, 5 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/SConscript b/src/gallium/SConscript index 210fd7d309..fa4833cbcf 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -1,3 +1,5 @@ +import os + Import('*') env = env.Clone() @@ -20,15 +22,5 @@ SConscript([ if llvm: SConscript(['auxiliary/gallivm/SConscript']) -SConscript([ - 'drivers/softpipe/SConscript', - 'drivers/i915simple/SConscript', - 'drivers/failover/SConscript', -]) - -if not msvc: - SConscript(['drivers/i965simple/SConscript']) - -#if cell: -# SConscript(['drivers/cell/SConscript']) - \ No newline at end of file +for driver in env['drivers']: + SConscript(os.path.join('drivers', driver, 'SConscript')) diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript index 3f0596e19c..635a68eea2 100644 --- a/src/gallium/winsys/SConscript +++ b/src/gallium/winsys/SConscript @@ -5,7 +5,7 @@ if dri: 'dri/SConscript', ]) -if env['x11'] and not dri and platform != 'winddk': +if 'xlib' in env['drivers'] and not dri: SConscript([ 'xlib/SConscript', ]) -- cgit v1.2.3 From e9bb63c8e20361597463b2f7f88d84fe2770c8b9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 24 Feb 2008 02:16:28 +0900 Subject: gallium: MSVC fixes. --- src/gallium/auxiliary/draw/draw_aapoint.c | 6 +++--- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 2 ++ src/gallium/drivers/softpipe/sp_state.h | 4 ++-- 3 files changed, 7 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index 43119cc70b..27a81f3e90 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -616,10 +616,10 @@ aapoint_point(struct draw_stage *stage, struct prim_header *header) */ #if !NORMALIZE - k = 1.0 / radius; - k = 1.0 - 2.0 * k + k * k; + k = 1.0f / radius; + k = 1.0f - 2.0f * k + k * k; #else - k = 1.0 - 1.0 / radius; + k = 1.0f - 1.0f / radius; #endif /* allocate/dup new verts */ diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index e2ee72ed1f..6e217eb2e0 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -285,7 +285,9 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) /* Wait on outstanding fences */ while (fenced_list->numDelayed) { _glthread_UNLOCK_MUTEX(fenced_list->mutex); +#ifndef __MSC__ sched_yield(); +#endif _fenced_buffer_list_check_free(fenced_list, 1); _glthread_LOCK_MUTEX(fenced_list->mutex); } diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index ef8cf67d4c..6ff31e601f 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -63,14 +63,14 @@ struct tgsi_exec_machine; struct sp_fragment_shader { struct pipe_shader_state shader; - void (*prepare)( struct sp_fragment_shader *shader, + void (*prepare)( const struct sp_fragment_shader *shader, struct tgsi_exec_machine *machine, struct tgsi_sampler *samplers); /* Run the shader - this interface will get cleaned up in the * future: */ - unsigned (*run)( struct sp_fragment_shader *shader, + unsigned (*run)( const struct sp_fragment_shader *shader, struct tgsi_exec_machine *machine, struct quad_header *quad ); -- cgit v1.2.3 From 1d77d6caf647424f9c1c481145be0465e96c9e3e Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 23 Feb 2008 16:15:29 -0700 Subject: gallium: added new tgsi_scan.c / tgsi_scan_shader() function Used to get information about registers, instructions used in a shader. --- src/gallium/auxiliary/tgsi/Makefile | 1 + src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 117 ++++++++++++++++++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_scan.h | 57 ++++++++++++++ 3 files changed, 175 insertions(+) create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_scan.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_scan.h (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index 71f64b747c..5555639b70 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -9,6 +9,7 @@ C_SOURCES = \ util/tgsi_build.c \ util/tgsi_dump.c \ util/tgsi_parse.c \ + util/tgsi_scan.c \ util/tgsi_transform.c \ util/tgsi_util.c diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c new file mode 100644 index 0000000000..4b99ac37cc --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -0,0 +1,117 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI program scan utility. + * Used to determine which registers and instructions are used by a shader. + * + * Authors: Brian Paul + */ + + +#include "tgsi_scan.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_build.h" + + + + +/** + */ +void +tgsi_scan_shader(const struct tgsi_token *tokens, + struct tgsi_shader_info *info) +{ + uint procType; + struct tgsi_parse_context parse; + + memset(info, 0, sizeof(*info)); + + /** + ** Setup to begin parsing input shader + **/ + if (tgsi_parse_init( &parse, tokens ) != TGSI_PARSE_OK) { + debug_printf("tgsi_parse_init() failed in tgsi_scan_shader()!\n"); + return; + } + procType = parse.FullHeader.Processor.Processor; + assert(procType == TGSI_PROCESSOR_FRAGMENT || + procType == TGSI_PROCESSOR_VERTEX || + procType == TGSI_PROCESSOR_GEOMETRY); + + + /** + ** Loop over incoming program tokens/instructions + */ + while( !tgsi_parse_end_of_tokens( &parse ) ) { + + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *fullinst + = &parse.FullToken.FullInstruction; + + assert(fullinst->Instruction.Opcode < TGSI_OPCODE_LAST); + info->opcode_count[fullinst->Instruction.Opcode]++; + } + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *fulldecl + = &parse.FullToken.FullDeclaration; + uint file = fulldecl->Declaration.File; + uint i; + for (i = fulldecl->u.DeclarationRange.First; + i <= fulldecl->u.DeclarationRange.Last; + i++) { + info->file_mask[file] |= (1 << i); + info->file_count[file]++; + + /* special case */ + if (procType == TGSI_PROCESSOR_FRAGMENT && + file == TGSI_FILE_OUTPUT && + fulldecl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) { + info->writes_z = TRUE; + } + } + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + info->immediate_count++; + break; + + default: + assert( 0 ); + } + } + + tgsi_parse_free (&parse); +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h new file mode 100644 index 0000000000..757446437c --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h @@ -0,0 +1,57 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_SCAN_H +#define TGSI_SCAN_H + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + + +/** + * Shader summary info + */ +struct tgsi_shader_info +{ + uint file_mask[TGSI_FILE_COUNT]; /**< bitmask of declared registers */ + uint file_count[TGSI_FILE_COUNT]; /**< number of declared registers */ + + uint immediate_count; /**< number of immediates declared */ + + uint opcode_count[TGSI_OPCODE_LAST]; /**< opcode histogram */ + + boolean writes_z; /**< does fragment shader write Z value? */ +}; + + +extern void +tgsi_scan_shader(const struct tgsi_token *tokens, + struct tgsi_shader_info *info); + + +#endif /* TGSI_SCAN_H */ -- cgit v1.2.3 From 35ca45daba3906ac94fb879d2374d476ba2dac47 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 23 Feb 2008 16:15:54 -0700 Subject: gallium: added TGSI_FILE_COUNT --- src/gallium/include/pipe/p_shader_tokens.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 10c47e0ef0..1806877f6c 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -50,6 +50,8 @@ struct tgsi_token #define TGSI_FILE_SAMPLER 5 #define TGSI_FILE_ADDRESS 6 #define TGSI_FILE_IMMEDIATE 7 +#define TGSI_FILE_COUNT 8 /**< how many TGSI_FILE_ types */ + #define TGSI_DECLARE_RANGE 0 #define TGSI_DECLARE_MASK 1 -- cgit v1.2.3 From 012391357fcbefd2b34e999eed91a129d5efd77c Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 23 Feb 2008 16:17:17 -0700 Subject: gallium: disable early Z test if fragment shader contains KIL instruction. Use tgsi_scan_shader() to determine if the fragment shader uses KIL or writes fragment.z --- src/gallium/drivers/softpipe/sp_quad.c | 5 +++-- src/gallium/drivers/softpipe/sp_state.h | 3 +++ src/gallium/drivers/softpipe/sp_state_fs.c | 22 +++++++++++++--------- 3 files changed, 19 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad.c index 15b5594547..142dbcc771 100644 --- a/src/gallium/drivers/softpipe/sp_quad.c +++ b/src/gallium/drivers/softpipe/sp_quad.c @@ -56,11 +56,12 @@ sp_build_depth_stencil( void sp_build_quad_pipeline(struct softpipe_context *sp) { - boolean early_depth_test = + boolean early_depth_test = sp->depth_stencil->depth.enabled && sp->framebuffer.zsbuf && !sp->depth_stencil->alpha.enabled && - sp->fs->shader.output_semantic_name[0] != TGSI_SEMANTIC_POSITION; + !sp->fs->uses_kill && + !sp->fs->writes_z; /* build up the pipeline in reverse order... */ diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 6ff31e601f..5aaa9e346b 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -63,6 +63,9 @@ struct tgsi_exec_machine; struct sp_fragment_shader { struct pipe_shader_state shader; + boolean uses_kill; + boolean writes_z; + void (*prepare)( const struct sp_fragment_shader *shader, struct tgsi_exec_machine *machine, struct tgsi_sampler *samplers); diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index b0238f8173..b184ac61bb 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -36,6 +36,7 @@ #include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" #include "tgsi/util/tgsi_dump.h" +#include "tgsi/util/tgsi_scan.h" void * @@ -44,21 +45,24 @@ softpipe_create_fs_state(struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_fragment_shader *state; + struct tgsi_shader_info info; + + tgsi_scan_shader(templ->tokens, &info); if (softpipe->dump_fs) tgsi_dump(templ->tokens, 0); state = softpipe_create_fs_llvm( softpipe, templ ); - if (state) - return state; - - state = softpipe_create_fs_sse( softpipe, templ ); - if (state) - return state; - - state = softpipe_create_fs_exec( softpipe, templ ); - + if (!state) { + state = softpipe_create_fs_sse( softpipe, templ ); + if (!state) { + state = softpipe_create_fs_exec( softpipe, templ ); + } + } assert(state); + state->uses_kill = (info.opcode_count[TGSI_OPCODE_KIL] || + info.opcode_count[TGSI_OPCODE_KILP]); + state->writes_z = info.writes_z; return state; } -- cgit v1.2.3 From 7aadb475e58b91f3c17c2a70f6700225e9ef25ed Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 24 Feb 2008 02:46:46 +0900 Subject: gallium: Fix MSVC warnings. --- src/gallium/auxiliary/draw/draw_pstipple.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index 4048abf856..1ab04cd959 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -133,7 +133,7 @@ pstip_transform_decl(struct tgsi_transform_context *ctx, pctx->maxSampler = (int) decl->u.DeclarationRange.Last; } else if (decl->Declaration.File == TGSI_FILE_INPUT) { - pctx->maxInput = MAX2(pctx->maxInput, decl->u.DeclarationRange.Last); + pctx->maxInput = MAX2(pctx->maxInput, (int) decl->u.DeclarationRange.Last); if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) pctx->wincoordInput = (int) decl->u.DeclarationRange.First; } @@ -332,7 +332,7 @@ generate_pstip_fs(struct pstip_stage *pstip) if (transform.wincoordInput < 0) { pstip_fs.input_semantic_name[pstip_fs.num_inputs] = TGSI_SEMANTIC_POSITION; - pstip_fs.input_semantic_index[pstip_fs.num_inputs] = transform.maxInput; + pstip_fs.input_semantic_index[pstip_fs.num_inputs] = (ubyte)transform.maxInput; pstip_fs.num_inputs++; } -- cgit v1.2.3 From fdcb9260eea8f9b9deaeeade2a46cffbf3dcaa59 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 24 Feb 2008 17:58:05 +0900 Subject: Add new files. --- src/gallium/auxiliary/draw/SConscript | 3 +++ src/gallium/auxiliary/tgsi/SConscript | 2 ++ src/gallium/drivers/softpipe/SConscript | 3 +++ 3 files changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 8e3a8caa74..3302dc44f7 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -3,6 +3,8 @@ Import('*') draw = env.ConvenienceLibrary( target = 'draw', source = [ + 'draw_aaline.c', + 'draw_aapoint.c', 'draw_clip.c', 'draw_vs_exec.c', 'draw_vs_sse.c', @@ -13,6 +15,7 @@ draw = env.ConvenienceLibrary( 'draw_flatshade.c', 'draw_offset.c', 'draw_prim.c', + 'draw_pstipple.c', 'draw_stipple.c', 'draw_twoside.c', 'draw_unfilled.c', diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript index 8464bfe944..4632dcc072 100644 --- a/src/gallium/auxiliary/tgsi/SConscript +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -8,6 +8,8 @@ tgsi = env.ConvenienceLibrary( 'util/tgsi_build.c', 'util/tgsi_dump.c', 'util/tgsi_parse.c', + 'util/tgsi_scan.c', + 'util/tgsi_transform.c', 'util/tgsi_util.c', ]) diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript index d581ee8d3c..4c1a6d5df0 100644 --- a/src/gallium/drivers/softpipe/SConscript +++ b/src/gallium/drivers/softpipe/SConscript @@ -5,6 +5,9 @@ env = env.Clone() softpipe = env.ConvenienceLibrary( target = 'softpipe', source = [ + 'sp_fs_exec.c', + 'sp_fs_sse.c', + 'sp_fs_llvm.c', 'sp_clear.c', 'sp_context.c', 'sp_draw_arrays.c', -- cgit v1.2.3 From b75706764b4cf18f3b41bf4a97d82b3c528064d8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 25 Feb 2008 15:18:33 +0900 Subject: Add Zack's comments about CSOs. --- src/gallium/auxiliary/cso_cache/cso_cache.h | 51 ++++++++++++++++++++++++++--- 1 file changed, 46 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h index b4f4f3ae41..3a005a376e 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.h +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -25,9 +25,48 @@ * **************************************************************************/ - /* - * Authors: - * Zack Rusin + /** + * @file + * Constant State Object (CSO) cache. + * + * The basic idea is that the states are created via the + * create_state/bind_state/delete_state semantics. The driver is expected to + * perform as much of the Gallium state translation to whatever its internal + * representation is during the create call. Gallium then has a caching + * mechanism where it stores the created states. When the pipeline needs an + * actual state change, a bind call is issued. In the bind call the driver + * gets its already translated representation. + * + * Those semantics mean that the driver doesn't do the repeated translations + * of states on every frame, but only once, when a new state is actually + * created. + * + * Even on hardware that doesn't do any kind of state cache, it makes the + * driver look a lot neater, plus it avoids all the redundant state + * translations on every frame. + * + * Currently our constant state objects are: + * - alpha test + * - blend + * - depth stencil + * - fragment shader + * - rasterizer (old setup) + * - sampler + * - vertex shader + * + * Things that are not constant state objects include: + * - blend_color + * - clip_state + * - clear_color_state + * - constant_buffer + * - feedback_state + * - framebuffer_state + * - polygon_stipple + * - scissor_state + * - texture_state + * - viewport_state + * + * @author Zack Rusin */ #ifndef CSO_CACHE_H @@ -36,7 +75,9 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" -#include "cso_hash.h" /* for cso_hash_iter, as MSVC requires structures returned by value to be fully defined */ +/* cso_hash.h is necessary for cso_hash_iter, as MSVC requires structures + * returned by value to be fully defined */ +#include "cso_hash.h" #ifdef __cplusplus -- cgit v1.2.3 From 2a0675eb75b8ca52efab739218bf93922bf884b5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 25 Feb 2008 16:39:39 +0900 Subject: Replace standand library functions by portable ones. --- src/gallium/auxiliary/cso_cache/cso_cache.c | 6 +++-- src/gallium/auxiliary/cso_cache/cso_hash.c | 34 ++++++++++++++--------------- 2 files changed, 20 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 776ce6bacf..9c32e94124 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -28,6 +28,8 @@ /* Authors: Zack Rusin */ +#include "pipe/p_util.h" + #include "cso_cache.h" #include "cso_hash.h" @@ -176,7 +178,7 @@ void * cso_take_state(struct cso_cache *sc, struct cso_cache *cso_cache_create(void) { - struct cso_cache *sc = malloc(sizeof(struct cso_cache)); + struct cso_cache *sc = MALLOC_STRUCT(cso_cache); sc->blend_hash = cso_hash_create(); sc->sampler_hash = cso_hash_create(); @@ -197,5 +199,5 @@ void cso_cache_delete(struct cso_cache *sc) cso_hash_delete(sc->rasterizer_hash); cso_hash_delete(sc->fs_hash); cso_hash_delete(sc->vs_hash); - free(sc); + FREE(sc); } diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index e65d331a0b..208fc58502 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -30,12 +30,10 @@ * Zack Rusin */ -#include "cso_hash.h" +#include "pipe/p_debug.h" +#include "pipe/p_util.h" -#include -#include -#include -#include +#include "cso_hash.h" #define MAX(a, b) ((a > b) ? (a) : (b)) @@ -98,7 +96,7 @@ struct cso_hash { static void *cso_data_allocate_node(struct cso_hash_data *hash) { - return malloc(hash->nodeSize); + return MALLOC(hash->nodeSize); } static void cso_data_free_node(struct cso_node *node) @@ -107,10 +105,10 @@ static void cso_data_free_node(struct cso_node *node) * Need to cast value ptr to original cso type, then free the * driver-specific data hanging off of it. For example: struct cso_sampler *csamp = (struct cso_sampler *) node->value; - free(csamp->data); + FREE(csamp->data); */ - free(node->value); - free(node); + FREE(node->value); + FREE(node); } static struct cso_node * @@ -149,7 +147,7 @@ static void cso_data_rehash(struct cso_hash_data *hash, int hint) hash->numBits = (short)hint; hash->numBuckets = primeForNumBits(hint); - hash->buckets = malloc(sizeof(struct cso_node*) * hash->numBuckets); + hash->buckets = MALLOC(sizeof(struct cso_node*) * hash->numBuckets); for (i = 0; i < hash->numBuckets; ++i) hash->buckets[i] = e; @@ -173,7 +171,7 @@ static void cso_data_rehash(struct cso_hash_data *hash, int hint) firstNode = afterLastNode; } } - free(oldBuckets); + FREE(oldBuckets); } } @@ -235,8 +233,8 @@ struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, struct cso_hash * cso_hash_create(void) { - struct cso_hash *hash = malloc(sizeof(struct cso_hash)); - hash->data.d = malloc(sizeof(struct cso_hash_data)); + struct cso_hash *hash = MALLOC_STRUCT(cso_hash); + hash->data.d = MALLOC_STRUCT(cso_hash_data); hash->data.d->fakeNext = 0; hash->data.d->buckets = 0; hash->data.d->size = 0; @@ -261,9 +259,9 @@ void cso_hash_delete(struct cso_hash *hash) cur = next; } } - free(hash->data.d->buckets); - free(hash->data.d); - free(hash); + FREE(hash->data.d->buckets); + FREE(hash->data.d); + FREE(hash); } struct cso_hash_iter cso_hash_find(struct cso_hash *hash, @@ -301,7 +299,7 @@ static struct cso_node *cso_hash_data_next(struct cso_node *node) a.next = node->next; if (!a.next) { - fprintf(stderr, "iterating beyond the last element\n"); + debug_printf("iterating beyond the last element\n"); return 0; } if (a.next->next) @@ -352,7 +350,7 @@ static struct cso_node *cso_hash_data_prev(struct cso_node *node) --bucket; --start; } - fprintf(stderr, "iterating backward beyond first element\n"); + debug_printf("iterating backward beyond first element\n"); return a.e; } -- cgit v1.2.3 From 2d38d1b3005c02273abf3941df5dddc245a6b792 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 25 Feb 2008 17:11:28 +0900 Subject: Remove files of unsupported build systems. --- src/descrip.mms | 43 ---- src/glu/descrip.mms | 9 - src/glu/mesa/Makefile.DJ | 100 -------- src/glu/mesa/descrip.mms | 61 ----- src/glu/mesa/mms_depend | 15 -- src/glu/sgi/Makefile.DJ | 188 --------------- src/glu/sgi/Makefile.mgw | 229 ------------------ src/glu/sgi/descrip.mms | 451 ------------------------------------ src/glut/dos/Makefile.DJ | 126 ---------- src/glut/glx/Makefile.mgw | 198 ---------------- src/glut/glx/descrip.mms | 208 ----------------- src/glut/glx/mms_depend | 72 ------ src/mesa/Makefile.DJ | 166 ------------- src/mesa/Makefile.mgw | 235 ------------------- src/mesa/descrip.mms | 26 --- src/mesa/drivers/common/descrip.mms | 38 --- src/mesa/drivers/osmesa/descrip.mms | 41 ---- src/mesa/drivers/x11/descrip.mms | 51 ---- src/mesa/glapi/descrip.mms | 37 --- src/mesa/main/descrip.mms | 221 ------------------ src/mesa/math/descrip.mms | 45 ---- src/mesa/shader/descrip.mms | 76 ------ src/mesa/shader/grammar/descrip.mms | 41 ---- src/mesa/shader/slang/descrip.mms | 65 ------ src/mesa/swrast/descrip.mms | 80 ------- src/mesa/swrast_setup/descrip.mms | 39 ---- src/mesa/tnl/descrip.mms | 75 ------ src/mesa/vbo/descrip.mms | 60 ----- 28 files changed, 2996 deletions(-) delete mode 100644 src/descrip.mms delete mode 100644 src/glu/descrip.mms delete mode 100644 src/glu/mesa/Makefile.DJ delete mode 100644 src/glu/mesa/descrip.mms delete mode 100644 src/glu/mesa/mms_depend delete mode 100644 src/glu/sgi/Makefile.DJ delete mode 100644 src/glu/sgi/Makefile.mgw delete mode 100644 src/glu/sgi/descrip.mms delete mode 100644 src/glut/dos/Makefile.DJ delete mode 100644 src/glut/glx/Makefile.mgw delete mode 100644 src/glut/glx/descrip.mms delete mode 100644 src/glut/glx/mms_depend delete mode 100644 src/mesa/Makefile.DJ delete mode 100644 src/mesa/Makefile.mgw delete mode 100644 src/mesa/descrip.mms delete mode 100644 src/mesa/drivers/common/descrip.mms delete mode 100644 src/mesa/drivers/osmesa/descrip.mms delete mode 100644 src/mesa/drivers/x11/descrip.mms delete mode 100644 src/mesa/glapi/descrip.mms delete mode 100644 src/mesa/main/descrip.mms delete mode 100644 src/mesa/math/descrip.mms delete mode 100644 src/mesa/shader/descrip.mms delete mode 100644 src/mesa/shader/grammar/descrip.mms delete mode 100644 src/mesa/shader/slang/descrip.mms delete mode 100644 src/mesa/swrast/descrip.mms delete mode 100644 src/mesa/swrast_setup/descrip.mms delete mode 100644 src/mesa/tnl/descrip.mms delete mode 100644 src/mesa/vbo/descrip.mms (limited to 'src') diff --git a/src/descrip.mms b/src/descrip.mms deleted file mode 100644 index 79c7d98d25..0000000000 --- a/src/descrip.mms +++ /dev/null @@ -1,43 +0,0 @@ -# Makefile for Mesa for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl - -.include [-]mms.config - -all : - set default [.mesa] - $(MMS)$(MMSQUALIFIERS) - set default [-] -.ifdef SHARE - $(MMS)$(MMSQUALIFIERS) [-.lib]$(GL_SHAR) -.endif - set default [.glu] - $(MMS)$(MMSQUALIFIERS) - set default [-.glut.glx] - $(MMS)$(MMSQUALIFIERS) - set default [--] - -[-.lib]$(GL_SHAR) : [-.lib]$(GL_LIB) - @ WRITE_ SYS$OUTPUT " generating libmesa.opt" - @ library/extract=* [-.lib]$(GL_LIB) - @ OPEN_/WRITE FILE libmesa.opt - @ WRITE_ FILE "!" - @ WRITE_ FILE "! libmesa.opt generated by DESCRIP.$(MMS_EXT)" - @ WRITE_ FILE "!" - @ WRITE_ FILE "IDENTIFICATION=""mesa5.1""" - @ WRITE_ FILE "GSMATCH=LEQUAL,5,1 - @ WRITE_ FILE "libmesagl.obj" - @ write_ file "sys$share:decw$xextlibshr/share" - @ write_ file "sys$share:decw$xlibshr/share" - @ write_ file "sys$share:pthread$rtl/share" - @ CLOSE_ FILE - @ $(MMS)$(MMSQUALIFIERS)/ignore=warning mesa_vms - @ WRITE_ SYS$OUTPUT " linking ..." - @ LINK_/NODEB/SHARE=[-.lib]$(GL_SHAR)/MAP=libmesa.map/FULL libmesa.opt/opt,\ - mesa_vms.opt/opt - @ delete libmesagl.obj;* - -mesa_vms : - @ WRITE_ SYS$OUTPUT " generating libmesa.map ..." - @ LINK_/NODEB/NOSHARE/NOEXE/MAP=libmesa.map/FULL libmesa.opt/OPT - @ WRITE_ SYS$OUTPUT " analyzing libmesa.map ..." - @ @[-.vms]analyze_map.com libmesa.map mesa_vms.opt diff --git a/src/glu/descrip.mms b/src/glu/descrip.mms deleted file mode 100644 index 6d5cd858da..0000000000 --- a/src/glu/descrip.mms +++ /dev/null @@ -1,9 +0,0 @@ -# Makefile for Mesa for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl - -all : -# PIPE is avalailable on VMS7.0 and higher. For lower versions split the -#command in two conditional command. JJ - if f$search("SYS$SYSTEM:CXX$COMPILER.EXE") .nes. "" then pipe set default [.sgi] ; $(MMS)$(MMSQUALIFIERS) - if f$search("SYS$SYSTEM:CXX$COMPILER.EXE") .eqs. "" then pipe set default [.mesa] ; $(MMS)$(MMSQUALIFIERS) - set default [-] diff --git a/src/glu/mesa/Makefile.DJ b/src/glu/mesa/Makefile.DJ deleted file mode 100644 index 92bcdaae94..0000000000 --- a/src/glu/mesa/Makefile.DJ +++ /dev/null @@ -1,100 +0,0 @@ -# Mesa 3-D graphics library -# Version: 4.0 -# -# Copyright (C) 1999 Brian Paul All Rights Reserved. -# -# Permission is hereby granted, free of charge, to any person obtaining a -# copy of this software and associated documentation files (the "Software"), -# to deal in the Software without restriction, including without limitation -# the rights to use, copy, modify, merge, publish, distribute, sublicense, -# and/or sell copies of the Software, and to permit persons to whom the -# Software is furnished to do so, subject to the following conditions: -# -# The above copyright notice and this permission notice shall be included -# in all copies or substantial portions of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL -# BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN -# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN -# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -# DOS/DJGPP glu makefile v1.5 for Mesa -# -# Copyright (C) 2002 - Daniel Borca -# Email : dborca@users.sourceforge.net -# Web : http://www.geocities.com/dborca - - -# -# Available options: -# -# Environment variables: -# CFLAGS -# -# Targets: -# all: build GLU -# clean: remove object files -# - - - -.PHONY: all clean - -TOP = ../../.. -LIBDIR = $(TOP)/lib -GLU_LIB = libglu.a -GLU_DXE = glu.dxe -GLU_IMP = libiglu.a - -export LD_LIBRARY_PATH := $(LD_LIBRARY_PATH);$(LIBDIR);$(GLIDE)/lib - -CC = gcc -CFLAGS += -I$(TOP)/include - -AR = ar -ARFLAGS = crus - -HAVEDXE3 = $(wildcard $(DJDIR)/bin/dxe3gen.exe) - -ifeq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),) -UNLINK = del $(subst /,\,$(1)) -else -UNLINK = $(RM) $(1) -endif - -CORE_SOURCES = \ - glu.c \ - mipmap.c \ - nurbs.c \ - nurbscrv.c \ - nurbssrf.c \ - nurbsutl.c \ - polytest.c \ - project.c \ - quadric.c \ - tess.c \ - tesselat.c - -SOURCES = $(CORE_SOURCES) - -OBJECTS = $(SOURCES:.c=.o) - -.c.o: - $(CC) -o $@ $(CFLAGS) -c $< - -all: $(LIBDIR)/$(GLU_LIB) $(LIBDIR)/$(GLU_DXE) $(LIBDIR)/$(GLU_IMP) - -$(LIBDIR)/$(GLU_LIB): $(OBJECTS) - $(AR) $(ARFLAGS) $@ $^ - -$(LIBDIR)/$(GLU_DXE) $(LIBDIR)/$(GLU_IMP): $(OBJECTS) -ifeq ($(HAVEDXE3),) - $(warning Missing DXE3 package... Skipping $(GLU_DXE)) -else - -dxe3gen -o $(LIBDIR)/$(GLU_DXE) -Y $(LIBDIR)/$(GLU_IMP) -D "MesaGLU DJGPP" -E _glu -P gl.dxe -U $^ -endif - -clean: - -$(call UNLINK,*.o) diff --git a/src/glu/mesa/descrip.mms b/src/glu/mesa/descrip.mms deleted file mode 100644 index 12eb6a437a..0000000000 --- a/src/glu/mesa/descrip.mms +++ /dev/null @@ -1,61 +0,0 @@ -# Makefile for GLU for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl - -.first - define gl [-.include.gl] - -.include [-]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = $disk2:[-.include] -LIBDIR = [-.lib] -CFLAGS = /include=$(INCDIR)/define=(FBIND=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = glu.c mipmap.c nurbs.c nurbscrv.c nurbssrf.c nurbsutl.c \ - polytest.c project.c quadric.c tess.c tesselat.c - -OBJECTS =glu.obj,mipmap.obj,nurbs.obj,nurbscrv.obj,nurbssrf.obj,nurbsutl.obj,\ - polytest.obj,project.obj,quadric.obj,tess.obj,tesselat.obj - - -##### RULES ##### - -VERSION=MesaGlu V3.2 - -##### TARGETS ##### - -# Make the library: -$(LIBDIR)$(GLU_LIB) : $(OBJECTS) -.ifdef SHARE - @ WRITE_ SYS$OUTPUT " generating mesagl1.opt" - @ OPEN_/WRITE FILE mesagl1.opt - @ WRITE_ FILE "!" - @ WRITE_ FILE "! mesagl1.opt generated by DESCRIP.$(MMS_EXT)" - @ WRITE_ FILE "!" - @ WRITE_ FILE "IDENTIFICATION=""$(VERSION)""" - @ WRITE_ FILE "GSMATCH=LEQUAL,3,2 - @ WRITE_ FILE "$(OBJECTS)" - @ WRITE_ FILE "[-.lib]libmesagl.exe/SHARE" - @ WRITE_ FILE "SYS$SHARE:DECW$XEXTLIBSHR/SHARE" - @ WRITE_ FILE "SYS$SHARE:DECW$XLIBSHR/SHARE" - @ CLOSE_ FILE - @ WRITE_ SYS$OUTPUT " generating mesagl.map ..." - @ LINK_/NODEB/NOSHARE/NOEXE/MAP=mesagl.map/FULL mesagl1.opt/OPT - @ WRITE_ SYS$OUTPUT " analyzing mesagl.map ..." - @ @[-.vms]ANALYZE_MAP.COM mesagl.map mesagl.opt - @ WRITE_ SYS$OUTPUT " linking $(GLU_LIB) ..." - @ LINK_/noinform/NODEB/SHARE=$(GLU_LIB)/MAP=mesagl.map/FULL mesagl1.opt/opt,mesagl.opt/opt -.else - @ $(MAKELIB) $(GLU_LIB) $(OBJECTS) -.endif - @ rename $(GLU_LIB)* $(LIBDIR) - -clean : - delete *.obj;* - purge - -include mms_depend. - diff --git a/src/glu/mesa/mms_depend b/src/glu/mesa/mms_depend deleted file mode 100644 index ed59ca9de8..0000000000 --- a/src/glu/mesa/mms_depend +++ /dev/null @@ -1,15 +0,0 @@ -# DO NOT DELETE THIS LINE -- make depend depends on it. - -glu.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h -mipmap.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h -nurbs.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h nurbs.h -nurbscrv.obj : nurbs.h gluP.h [-.include.gl]gl.h [-.include.gl]glu.h -nurbssrf.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h nurbs.h -nurbsutl.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h nurbs.h -project.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h -quadric.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h -tess.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h -tess_fist.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h -tess_hash.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h -tess_heap.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h -tess_clip.obj : gluP.h [-.include.gl]gl.h [-.include.gl]glu.h tess.h diff --git a/src/glu/sgi/Makefile.DJ b/src/glu/sgi/Makefile.DJ deleted file mode 100644 index b5df3e846a..0000000000 --- a/src/glu/sgi/Makefile.DJ +++ /dev/null @@ -1,188 +0,0 @@ -# Mesa 3-D graphics library -# Version: 4.0 -# -# Copyright (C) 1999 Brian Paul All Rights Reserved. -# -# Permission is hereby granted, free of charge, to any person obtaining a -# copy of this software and associated documentation files (the "Software"), -# to deal in the Software without restriction, including without limitation -# the rights to use, copy, modify, merge, publish, distribute, sublicense, -# and/or sell copies of the Software, and to permit persons to whom the -# Software is furnished to do so, subject to the following conditions: -# -# The above copyright notice and this permission notice shall be included -# in all copies or substantial portions of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL -# BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN -# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN -# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -# DOS/DJGPP glu makefile v1.5 for Mesa -# -# Copyright (C) 2002 - Daniel Borca -# Email : dborca@users.sourceforge.net -# Web : http://www.geocities.com/dborca - - -# -# Available options: -# -# Environment variables: -# CFLAGS -# -# Targets: -# all: build GLU -# clean: remove object files -# - - - -.PHONY: all clean - -TOP = ../../.. -LIBDIR = $(TOP)/lib -GLU_LIB = libglu.a -GLU_DXE = glu.dxe -GLU_IMP = libiglu.a - -export LD_LIBRARY_PATH := $(LD_LIBRARY_PATH);$(LIBDIR);$(GLIDE)/lib - -CC = gcc -CFLAGS += -DNDEBUG -DLIBRARYBUILD -I$(TOP)/include -Iinclude -CXX = gpp -CXXFLAGS = $(CFLAGS) -Ilibnurbs/internals -Ilibnurbs/interface -Ilibnurbs/nurbtess - -AR = ar -ARFLAGS = crus - -HAVEDXE3 = $(wildcard $(DJDIR)/bin/dxe3gen.exe) - -ifeq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),) -UNLINK = del $(subst /,\,$(1)) -else -UNLINK = $(RM) $(1) -endif - -C_SOURCES = \ - libutil/error.c \ - libutil/glue.c \ - libutil/mipmap.c \ - libutil/project.c \ - libutil/quad.c \ - libutil/registry.c \ - libtess/dict.c \ - libtess/geom.c \ - libtess/memalloc.c \ - libtess/mesh.c \ - libtess/normal.c \ - libtess/priorityq.c \ - libtess/render.c \ - libtess/sweep.c \ - libtess/tess.c \ - libtess/tessmono.c - -CC_SOURCES = \ - libnurbs/interface/bezierEval.cc \ - libnurbs/interface/bezierPatch.cc \ - libnurbs/interface/bezierPatchMesh.cc \ - libnurbs/interface/glcurveval.cc \ - libnurbs/interface/glinterface.cc \ - libnurbs/interface/glrenderer.cc \ - libnurbs/interface/glsurfeval.cc \ - libnurbs/interface/incurveeval.cc \ - libnurbs/interface/insurfeval.cc \ - libnurbs/internals/arc.cc \ - libnurbs/internals/arcsorter.cc \ - libnurbs/internals/arctess.cc \ - libnurbs/internals/backend.cc \ - libnurbs/internals/basiccrveval.cc \ - libnurbs/internals/basicsurfeval.cc \ - libnurbs/internals/bin.cc \ - libnurbs/internals/bufpool.cc \ - libnurbs/internals/cachingeval.cc \ - libnurbs/internals/ccw.cc \ - libnurbs/internals/coveandtiler.cc \ - libnurbs/internals/curve.cc \ - libnurbs/internals/curvelist.cc \ - libnurbs/internals/curvesub.cc \ - libnurbs/internals/dataTransform.cc \ - libnurbs/internals/displaylist.cc \ - libnurbs/internals/flist.cc \ - libnurbs/internals/flistsorter.cc \ - libnurbs/internals/hull.cc \ - libnurbs/internals/intersect.cc \ - libnurbs/internals/knotvector.cc \ - libnurbs/internals/mapdesc.cc \ - libnurbs/internals/mapdescv.cc \ - libnurbs/internals/maplist.cc \ - libnurbs/internals/mesher.cc \ - libnurbs/internals/monoTriangulationBackend.cc \ - libnurbs/internals/monotonizer.cc \ - libnurbs/internals/mycode.cc \ - libnurbs/internals/nurbsinterfac.cc \ - libnurbs/internals/nurbstess.cc \ - libnurbs/internals/patch.cc \ - libnurbs/internals/patchlist.cc \ - libnurbs/internals/quilt.cc \ - libnurbs/internals/reader.cc \ - libnurbs/internals/renderhints.cc \ - libnurbs/internals/slicer.cc \ - libnurbs/internals/sorter.cc \ - libnurbs/internals/splitarcs.cc \ - libnurbs/internals/subdivider.cc \ - libnurbs/internals/tobezier.cc \ - libnurbs/internals/trimline.cc \ - libnurbs/internals/trimregion.cc \ - libnurbs/internals/trimvertpool.cc \ - libnurbs/internals/uarray.cc \ - libnurbs/internals/varray.cc \ - libnurbs/nurbtess/directedLine.cc \ - libnurbs/nurbtess/gridWrap.cc \ - libnurbs/nurbtess/monoChain.cc \ - libnurbs/nurbtess/monoPolyPart.cc \ - libnurbs/nurbtess/monoTriangulation.cc \ - libnurbs/nurbtess/partitionX.cc \ - libnurbs/nurbtess/partitionY.cc \ - libnurbs/nurbtess/polyDBG.cc \ - libnurbs/nurbtess/polyUtil.cc \ - libnurbs/nurbtess/primitiveStream.cc \ - libnurbs/nurbtess/quicksort.cc \ - libnurbs/nurbtess/rectBlock.cc \ - libnurbs/nurbtess/sampleComp.cc \ - libnurbs/nurbtess/sampleCompBot.cc \ - libnurbs/nurbtess/sampleCompRight.cc \ - libnurbs/nurbtess/sampleCompTop.cc \ - libnurbs/nurbtess/sampleMonoPoly.cc \ - libnurbs/nurbtess/sampledLine.cc \ - libnurbs/nurbtess/searchTree.cc - -SOURCES = $(C_SOURCES) $(CC_SOURCES) - -OBJECTS = $(addsuffix .o,$(basename $(SOURCES))) - -.c.o: - $(CC) -o $@ $(CFLAGS) -c $< -.cc.o: - $(CXX) -o $@ $(CXXFLAGS) -c $< - -all: $(LIBDIR)/$(GLU_LIB) $(LIBDIR)/$(GLU_DXE) $(LIBDIR)/$(GLU_IMP) - -$(LIBDIR)/$(GLU_LIB): $(OBJECTS) - $(AR) $(ARFLAGS) $@ $^ - -$(LIBDIR)/$(GLU_DXE) $(LIBDIR)/$(GLU_IMP): $(OBJECTS) -ifeq ($(HAVEDXE3),) - $(warning Missing DXE3 package... Skipping $(GLU_DXE)) -else - -dxe3gen -o $(LIBDIR)/$(GLU_DXE) -Y $(LIBDIR)/$(GLU_IMP) -D "MesaGLU/SGI DJGPP" -E _glu -P gl.dxe -U $^ -endif - -clean: - -$(call UNLINK,libutil/*.o) - -$(call UNLINK,libtess/*.o) - -$(call UNLINK,libnurbs/interface/*.o) - -$(call UNLINK,libnurbs/internals/*.o) - -$(call UNLINK,libnurbs/nurbtess/*.o) diff --git a/src/glu/sgi/Makefile.mgw b/src/glu/sgi/Makefile.mgw deleted file mode 100644 index 43b421e737..0000000000 --- a/src/glu/sgi/Makefile.mgw +++ /dev/null @@ -1,229 +0,0 @@ -# Mesa 3-D graphics library -# Version: 5.1 -# -# Copyright (C) 1999-2003 Brian Paul All Rights Reserved. -# -# Permission is hereby granted, free of charge, to any person obtaining a -# copy of this software and associated documentation files (the "Software"), -# to deal in the Software without restriction, including without limitation -# the rights to use, copy, modify, merge, publish, distribute, sublicense, -# and/or sell copies of the Software, and to permit persons to whom the -# Software is furnished to do so, subject to the following conditions: -# -# The above copyright notice and this permission notice shall be included -# in all copies or substantial portions of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL -# BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN -# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN -# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -# MinGW core makefile v1.4 for Mesa -# -# Copyright (C) 2002 - Daniel Borca -# Email : dborca@users.sourceforge.net -# Web : http://www.geocities.com/dborca - -# MinGW core-glu makefile updated for Mesa 7.0 -# -# Updated : by Heromyth, on 2007-7-21 -# Email : zxpmyth@yahoo.com.cn -# Bugs : 1) All the default settings work fine. But the setting X86=1 can't work. -# The others havn't been tested yet. -# 2) The generated DLLs are *not* compatible with the ones built -# with the other compilers like VC8, especially for GLUT. -# 3) Although more tests are needed, it can be used individually! - -# -# Available options: -# -# Environment variables: -# CFLAGS -# -# GLIDE path to Glide3 SDK; used with FX. -# default = $(TOP)/glide3 -# FX=1 build for 3dfx Glide3. Note that this disables -# compilation of most WMesa code and requires fxMesa. -# As a consequence, you'll need the Win32 Glide3 -# library to build any application. -# default = no -# ICD=1 build the installable client driver interface -# (windows opengl driver interface) -# default = no -# X86=1 optimize for x86 (if possible, use MMX, SSE, 3DNow). -# default = no -# -# Targets: -# all: build GL -# clean: remove object files -# - - - -.PHONY: all clean -.INTERMEDIATE: x86/gen_matypes.exe -.SUFFIXES: .rc .res - -# Set this to the prefix of your build tools, i.e. mingw32- -TOOLS_PREFIX = mingw32- - -TOP = ../../.. - -LIBDIR = $(TOP)/lib - -GLU_DLL = glu32.dll -GLU_IMP = libglu32.a -GLU_DEF = glu.def - -include $(TOP)/configs/config.mgw -GL_USING_STDCALL ?= 1 - -LDLIBS = -L$(LIBDIR) -lopengl32 -LDFLAGS = -Wl,--out-implib=$(LIBDIR)/$(GLU_IMP) -Wl,--output-def=$(LIBDIR)/$(GLU_DEF) - -CFLAGS += -DBUILD_GLU32 -D_DLL - -ifeq ($(GL_USING_STDCALL),1) - LDFLAGS += -Wl,--add-stdcall-alias -else - CFLAGS += -DGL_NO_STDCALL -endif - -CC = gcc -CFLAGS += -DNDEBUG -DLIBRARYBUILD -I$(TOP)/include -Iinclude -CXX = g++ -CXXFLAGS = $(CFLAGS) -Ilibnurbs/internals -Ilibnurbs/interface -Ilibnurbs/nurbtess - -AR = ar -ARFLAGS = crus - -UNLINK = del $(subst /,\,$(1)) -ifneq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),) -UNLINK = $(RM) $(1) -endif -ifneq ($(wildcard $(addsuffix /rm,$(subst :, ,$(PATH)))),) -UNLINK = $(RM) $(1) -endif - -C_SOURCES = \ - libutil/error.c \ - libutil/glue.c \ - libutil/mipmap.c \ - libutil/project.c \ - libutil/quad.c \ - libutil/registry.c \ - libtess/dict.c \ - libtess/geom.c \ - libtess/memalloc.c \ - libtess/mesh.c \ - libtess/normal.c \ - libtess/priorityq.c \ - libtess/render.c \ - libtess/sweep.c \ - libtess/tess.c \ - libtess/tessmono.c - -CC_SOURCES = \ - libnurbs/interface/bezierEval.cc \ - libnurbs/interface/bezierPatch.cc \ - libnurbs/interface/bezierPatchMesh.cc \ - libnurbs/interface/glcurveval.cc \ - libnurbs/interface/glinterface.cc \ - libnurbs/interface/glrenderer.cc \ - libnurbs/interface/glsurfeval.cc \ - libnurbs/interface/incurveeval.cc \ - libnurbs/interface/insurfeval.cc \ - libnurbs/internals/arc.cc \ - libnurbs/internals/arcsorter.cc \ - libnurbs/internals/arctess.cc \ - libnurbs/internals/backend.cc \ - libnurbs/internals/basiccrveval.cc \ - libnurbs/internals/basicsurfeval.cc \ - libnurbs/internals/bin.cc \ - libnurbs/internals/bufpool.cc \ - libnurbs/internals/cachingeval.cc \ - libnurbs/internals/ccw.cc \ - libnurbs/internals/coveandtiler.cc \ - libnurbs/internals/curve.cc \ - libnurbs/internals/curvelist.cc \ - libnurbs/internals/curvesub.cc \ - libnurbs/internals/dataTransform.cc \ - libnurbs/internals/displaylist.cc \ - libnurbs/internals/flist.cc \ - libnurbs/internals/flistsorter.cc \ - libnurbs/internals/hull.cc \ - libnurbs/internals/intersect.cc \ - libnurbs/internals/knotvector.cc \ - libnurbs/internals/mapdesc.cc \ - libnurbs/internals/mapdescv.cc \ - libnurbs/internals/maplist.cc \ - libnurbs/internals/mesher.cc \ - libnurbs/internals/monoTriangulationBackend.cc \ - libnurbs/internals/monotonizer.cc \ - libnurbs/internals/mycode.cc \ - libnurbs/internals/nurbsinterfac.cc \ - libnurbs/internals/nurbstess.cc \ - libnurbs/internals/patch.cc \ - libnurbs/internals/patchlist.cc \ - libnurbs/internals/quilt.cc \ - libnurbs/internals/reader.cc \ - libnurbs/internals/renderhints.cc \ - libnurbs/internals/slicer.cc \ - libnurbs/internals/sorter.cc \ - libnurbs/internals/splitarcs.cc \ - libnurbs/internals/subdivider.cc \ - libnurbs/internals/tobezier.cc \ - libnurbs/internals/trimline.cc \ - libnurbs/internals/trimregion.cc \ - libnurbs/internals/trimvertpool.cc \ - libnurbs/internals/uarray.cc \ - libnurbs/internals/varray.cc \ - libnurbs/nurbtess/directedLine.cc \ - libnurbs/nurbtess/gridWrap.cc \ - libnurbs/nurbtess/monoChain.cc \ - libnurbs/nurbtess/monoPolyPart.cc \ - libnurbs/nurbtess/monoTriangulation.cc \ - libnurbs/nurbtess/partitionX.cc \ - libnurbs/nurbtess/partitionY.cc \ - libnurbs/nurbtess/polyDBG.cc \ - libnurbs/nurbtess/polyUtil.cc \ - libnurbs/nurbtess/primitiveStream.cc \ - libnurbs/nurbtess/quicksort.cc \ - libnurbs/nurbtess/rectBlock.cc \ - libnurbs/nurbtess/sampleComp.cc \ - libnurbs/nurbtess/sampleCompBot.cc \ - libnurbs/nurbtess/sampleCompRight.cc \ - libnurbs/nurbtess/sampleCompTop.cc \ - libnurbs/nurbtess/sampleMonoPoly.cc \ - libnurbs/nurbtess/sampledLine.cc \ - libnurbs/nurbtess/searchTree.cc - -SOURCES = $(C_SOURCES) $(CC_SOURCES) - -OBJECTS = $(addsuffix .o,$(basename $(SOURCES))) - -.c.o: - $(CC) -o $@ $(CFLAGS) -c $< -.cc.o: - $(CXX) -o $@ $(CXXFLAGS) -c $< - - -all: $(LIBDIR) $(LIBDIR)/$(GLU_DLL) $(LIBDIR)/$(GLU_IMP) - -$(LIBDIR): - mkdir -p $(LIBDIR) - -$(LIBDIR)/$(GLU_DLL) $(LIBDIR)/$(GLU_IMP): $(OBJECTS) - g++ -shared -fPIC -o $(LIBDIR)/$(GLU_DLL) $(LDFLAGS) \ - $^ $(LDLIBS) - - - -clean: - -$(call UNLINK,libutil/*.o) - -$(call UNLINK,libtess/*.o) - -$(call UNLINK,libnurbs/interface/*.o) - -$(call UNLINK,libnurbs/internals/*.o) - -$(call UNLINK,libnurbs/nurbtess/*.o) diff --git a/src/glu/sgi/descrip.mms b/src/glu/sgi/descrip.mms deleted file mode 100644 index 49d618e6c9..0000000000 --- a/src/glu/sgi/descrip.mms +++ /dev/null @@ -1,451 +0,0 @@ -# Makefile for GLU for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl - -.first - define gl [---.include.gl] - -.include [---]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR =([-.include],[.include],[.internals],[.libnurbs.internals],\ - [.libnurbs.interface],[.libnurbs.nurbtess]) -LIBDIR = [---.lib] -CFLAGS = /include=$(INCDIR)/name=(as_is,short)/float=ieee/ieee=denorm - -LU_OBJECTS=\ - [.libutil]error.obj, \ - [.libutil]glue.obj, \ - [.libutil]mipmap.obj,\ - [.libutil]project.obj,\ - [.libutil]quad.obj, \ - [.libutil]registry.obj - -LT_OBJECTS=[.libtess]dict.obj, \ - [.libtess]geom.obj, \ - [.libtess]memalloc.obj,\ - [.libtess]mesh.obj, \ - [.libtess]normal.obj,\ - [.libtess]priorityq.obj,\ - [.libtess]render.obj,\ - [.libtess]sweep.obj, \ - [.libtess]tess.obj, \ - [.libtess]tessmono.obj - -LI_OBJECTS=[.libnurbs.interface]bezierEval.obj, \ - [.libnurbs.interface]bezierPatch.obj, \ - [.libnurbs.interface]bezierPatchMesh.obj, \ - [.libnurbs.interface]glcurveval.obj, \ - [.libnurbs.interface]glinterface.obj - -LI_OBJECTS1=[.libnurbs.interface]glrenderer.obj, \ - [.libnurbs.interface]glsurfeval.obj, \ - [.libnurbs.interface]incurveeval.obj, \ - [.libnurbs.interface]insurfeval.obj - -LI2_OBJECTS=[.libnurbs.internals]arc.obj, \ - [.libnurbs.internals]arcsorter.obj, \ - [.libnurbs.internals]arctess.obj, \ - [.libnurbs.internals]backend.obj, \ - [.libnurbs.internals]basiccrveval.obj, \ - [.libnurbs.internals]basicsurfeval.obj - -LI2_OBJECTS1=[.libnurbs.internals]bin.obj, \ - [.libnurbs.internals]bufpool.obj, \ - [.libnurbs.internals]cachingeval.obj, \ - [.libnurbs.internals]ccw.obj, \ - [.libnurbs.internals]coveandtiler.obj, \ - [.libnurbs.internals]curve.obj, \ - [.libnurbs.internals]curvelist.obj - -LI2_OBJECTS2=[.libnurbs.internals]curvesub.obj, \ - [.libnurbs.internals]dataTransform.obj, \ - [.libnurbs.internals]displaylist.obj, \ - [.libnurbs.internals]flist.obj, \ - [.libnurbs.internals]flistsorter.obj - -LI2_OBJECTS3=[.libnurbs.internals]hull.obj, \ - [.libnurbs.internals]intersect.obj, \ - [.libnurbs.internals]knotvector.obj, \ - [.libnurbs.internals]mapdesc.obj - -LI2_OBJECTS4=[.libnurbs.internals]mapdescv.obj, \ - [.libnurbs.internals]maplist.obj, \ - [.libnurbs.internals]mesher.obj, \ - [.libnurbs.internals]monoTriangulationBackend.obj,\ - [.libnurbs.internals]monotonizer.obj - -LI2_OBJECTS5=[.libnurbs.internals]mycode.obj, \ - [.libnurbs.internals]nurbsinterfac.obj, \ - [.libnurbs.internals]nurbstess.obj, \ - [.libnurbs.internals]patch.obj - -LI2_OBJECTS6=[.libnurbs.internals]patchlist.obj, \ - [.libnurbs.internals]quilt.obj, \ - [.libnurbs.internals]reader.obj, \ - [.libnurbs.internals]renderhints.obj, \ - [.libnurbs.internals]slicer.obj - -LI2_OBJECTS7=[.libnurbs.internals]sorter.obj, \ - [.libnurbs.internals]splitarcs.obj, \ - [.libnurbs.internals]subdivider.obj, \ - [.libnurbs.internals]tobezier.obj - -LI2_OBJECTS8=[.libnurbs.internals]trimline.obj, \ - [.libnurbs.internals]trimregion.obj, \ - [.libnurbs.internals]trimvertpool.obj, \ - [.libnurbs.internals]uarray.obj, \ - [.libnurbs.internals]varray.obj - -LN_OBJECTS=[.libnurbs.nurbtess]directedLine.obj, \ - [.libnurbs.nurbtess]gridWrap.obj, \ - [.libnurbs.nurbtess]monoChain.obj, \ - [.libnurbs.nurbtess]monoPolyPart.obj, \ - [.libnurbs.nurbtess]monoTriangulation.obj - -LN_OBJECTS1=[.libnurbs.nurbtess]partitionX.obj, \ - [.libnurbs.nurbtess]partitionY.obj, \ - [.libnurbs.nurbtess]polyDBG.obj - -LN_OBJECTS2=[.libnurbs.nurbtess]polyUtil.obj, \ - [.libnurbs.nurbtess]primitiveStream.obj, \ - [.libnurbs.nurbtess]quicksort.obj, \ - [.libnurbs.nurbtess]rectBlock.obj - -LN_OBJECTS3=[.libnurbs.nurbtess]sampleComp.obj, \ - [.libnurbs.nurbtess]sampleCompBot.obj, \ - [.libnurbs.nurbtess]sampleCompRight.obj - -LN_OBJECTS4=[.libnurbs.nurbtess]sampleCompTop.obj, \ - [.libnurbs.nurbtess]sampleMonoPoly.obj,\ - [.libnurbs.nurbtess]sampledLine.obj, \ - [.libnurbs.nurbtess]searchTree.obj - -##### RULES ##### - -VERSION=MesaGlu V3.5 - -##### TARGETS ##### - -# Make the library: -$(LIBDIR)$(GLU_LIB) : $(LU_OBJECTS) $(LT_OBJECTS) $(LI_OBJECTS) $(LI_OBJECTS1)\ - $(LI2_OBJECTS) $(LI2_OBJECTS1) $(LI2_OBJECTS2)\ - $(LI2_OBJECTS3) $(LI2_OBJECTS4) $(LI2_OBJECTS5)\ - $(LI2_OBJECTS6) $(LI2_OBJECTS7) $(LI2_OBJECTS8)\ - $(LN_OBJECTS) $(LN_OBJECTS1) $(LN_OBJECTS2)\ - $(LN_OBJECTS3) $(LN_OBJECTS4) - @ $(MAKELIB) $(GLU_LIB) $(LU_OBJECTS),$(LT_OBJECTS),$(LI_OBJECTS),\ - $(LI2_OBJECTS),$(LN_OBJECTS) - @ rename $(GLU_LIB)* $(LIBDIR) -.ifdef SHARE - @ WRITE_ SYS$OUTPUT " generating mesagl1.opt" - @ OPEN_/WRITE FILE mesagl1.opt - @ WRITE_ FILE "!" - @ WRITE_ FILE "! mesagl1.opt generated by DESCRIP.$(MMS_EXT)" - @ WRITE_ FILE "!" - @ WRITE_ FILE "IDENTIFICATION=""$(VERSION)""" - @ WRITE_ FILE "GSMATCH=LEQUAL,3,5 - @ WRITE_ FILE "$(LU_OBJECTS)" - @ WRITE_ FILE "$(LT_OBJECTS)" - @ WRITE_ FILE "$(LI_OBJECTS)" - @ WRITE_ FILE "$(LI_OBJECTS1)" - @ WRITE_ FILE "$(LI2_OBJECTS)" - @ WRITE_ FILE "$(LI2_OBJECTS1)" - @ WRITE_ FILE "$(LI2_OBJECTS2)" - @ WRITE_ FILE "$(LI2_OBJECTS3)" - @ WRITE_ FILE "$(LI2_OBJECTS4)" - @ WRITE_ FILE "$(LI2_OBJECTS5)" - @ WRITE_ FILE "$(LI2_OBJECTS6)" - @ WRITE_ FILE "$(LI2_OBJECTS7)" - @ WRITE_ FILE "$(LI2_OBJECTS8)" - @ WRITE_ FILE "$(LN_OBJECTS)" - @ WRITE_ FILE "$(LN_OBJECTS1)" - @ WRITE_ FILE "$(LN_OBJECTS2)" - @ WRITE_ FILE "$(LN_OBJECTS3)" - @ WRITE_ FILE "$(LN_OBJECTS4)" - @ WRITE_ FILE "[---.lib]libmesagl.exe/SHARE" - @ WRITE_ FILE "SYS$SHARE:DECW$XEXTLIBSHR/SHARE" - @ WRITE_ FILE "SYS$SHARE:DECW$XLIBSHR/SHARE" - @ CLOSE_ FILE -# @ WRITE_ SYS$OUTPUT " generating mesagl.map ..." -# @ CXXLINK_/NODEB/NOSHARE/NOEXE/MAP=mesagl.map/FULL mesagl1.opt/OPT -# @ WRITE_ SYS$OUTPUT " analyzing mesagl.map ..." -# @ @[-.vms]ANALYZE_MAP.COM mesagl.map mesagl.opt - @ WRITE_ SYS$OUTPUT " linking $(GLU_SHAR) ..." -# @ CXXLINK_/noinform/NODEB/SHARE=$(GLU_SHAR)/MAP=mesagl.map/FULL mesagl1.opt/opt,mesagl.opt/opt - @ CXXLINK_/noinform/NODEB/SHARE=$(GLU_SHAR)/MAP=mesagl.map/FULL mesagl1.opt/opt,mesaglu.opt/opt - @ rename $(GLU_SHAR)* $(LIBDIR) -.endif - -clean : - delete [...]*.obj;* - purge - -[.libutil]error.obj : [.libutil]error.c - $(CC) $(CFLAGS) /obj=[.libutil]error.obj [.libutil]error.c - -[.libutil]glue.obj : [.libutil]glue.c - $(CC) $(CFLAGS) /obj=[.libutil]glue.obj [.libutil]glue.c - -[.libutil]mipmap.obj : [.libutil]mipmap.c - $(CC) $(CFLAGS) /obj=[.libutil]mipmap.obj [.libutil]mipmap.c - -[.libutil]project.obj : [.libutil]project.c - $(CC) $(CFLAGS) /obj=[.libutil]project.obj [.libutil]project.c - -[.libutil]quad.obj : [.libutil]quad.c - $(CC) $(CFLAGS) /obj=[.libutil]quad.obj [.libutil]quad.c - -[.libutil]registry.obj : [.libutil]registry.c - $(CC) $(CFLAGS) /obj=[.libutil]registry.obj [.libutil]registry.c - -[.libtess]dict.obj : [.libtess]dict.c - $(CC) $(CFLAGS) /obj=[.libtess]dict.obj [.libtess]dict.c - -[.libtess]geom.obj : [.libtess]geom.c - $(CC) $(CFLAGS) /obj=[.libtess]geom.obj [.libtess]geom.c - -[.libtess]memalloc.obj : [.libtess]memalloc.c - $(CC) $(CFLAGS) /obj=[.libtess]memalloc.obj [.libtess]memalloc.c - -[.libtess]mesh.obj : [.libtess]mesh.c - $(CC) $(CFLAGS) /obj=[.libtess]mesh.obj [.libtess]mesh.c - -[.libtess]normal.obj : [.libtess]normal.c - $(CC) $(CFLAGS) /obj=[.libtess]normal.obj [.libtess]normal.c - -[.libtess]priorityq.obj : [.libtess]priorityq.c - $(CC) $(CFLAGS) /obj=[.libtess]priorityq.obj [.libtess]priorityq.c - -[.libtess]render.obj : [.libtess]render.c - $(CC) $(CFLAGS) /obj=[.libtess]render.obj [.libtess]render.c - -[.libtess]sweep.obj : [.libtess]sweep.c - $(CC) $(CFLAGS) /obj=[.libtess]sweep.obj [.libtess]sweep.c - -[.libtess]tess.obj : [.libtess]tess.c - $(CC) $(CFLAGS) /obj=[.libtess]tess.obj [.libtess]tess.c - -[.libtess]tessmono.obj : [.libtess]tessmono.c - $(CC) $(CFLAGS) /obj=[.libtess]tessmono.obj [.libtess]tessmono.c - -[.libnurbs.interface]bezierEval.obj : [.libnurbs.interface]bezierEval.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.interface]bezierEval.obj [.libnurbs.interface]bezierEval.cc - -[.libnurbs.interface]bezierPatch.obj : [.libnurbs.interface]bezierPatch.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.interface]bezierPatch.obj [.libnurbs.interface]bezierPatch.cc - -[.libnurbs.interface]bezierPatchMesh.obj : [.libnurbs.interface]bezierPatchMesh.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.interface]bezierPatchMesh.obj [.libnurbs.interface]bezierPatchMesh.cc - -[.libnurbs.interface]glcurveval.obj : [.libnurbs.interface]glcurveval.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.interface]glcurveval.obj [.libnurbs.interface]glcurveval.cc - -[.libnurbs.interface]glinterface.obj : [.libnurbs.interface]glinterface.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.interface]glinterface.obj [.libnurbs.interface]glinterface.cc - -[.libnurbs.interface]glrenderer.obj : [.libnurbs.interface]glrenderer.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.interface]glrenderer.obj [.libnurbs.interface]glrenderer.cc - -[.libnurbs.interface]glsurfeval.obj : [.libnurbs.interface]glsurfeval.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.interface]glsurfeval.obj [.libnurbs.interface]glsurfeval.cc - -[.libnurbs.interface]incurveeval.obj : [.libnurbs.interface]incurveeval.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.interface]incurveeval.obj [.libnurbs.interface]incurveeval.cc - -[.libnurbs.interface]insurfeval.obj : [.libnurbs.interface]insurfeval.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.interface]insurfeval.obj [.libnurbs.interface]insurfeval.cc - -[.libnurbs.internals]arc.obj : [.libnurbs.internals]arc.cc - $(CXX) $(CFLAGS)/list/show=all /obj=[.libnurbs.internals]arc.obj [.libnurbs.internals]arc.cc - -[.libnurbs.internals]arcsorter.obj : [.libnurbs.internals]arcsorter.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]arcsorter.obj [.libnurbs.internals]arcsorter.cc - -[.libnurbs.internals]arctess.obj : [.libnurbs.internals]arctess.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]arctess.obj [.libnurbs.internals]arctess.cc - -[.libnurbs.internals]backend.obj : [.libnurbs.internals]backend.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]backend.obj [.libnurbs.internals]backend.cc - -[.libnurbs.internals]basiccrveval.obj : [.libnurbs.internals]basiccrveval.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]basiccrveval.obj [.libnurbs.internals]basiccrveval.cc - -[.libnurbs.internals]basicsurfeval.obj : [.libnurbs.internals]basicsurfeval.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]basicsurfeval.obj [.libnurbs.internals]basicsurfeval.cc - -[.libnurbs.internals]bin.obj : [.libnurbs.internals]bin.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]bin.obj [.libnurbs.internals]bin.cc - -[.libnurbs.internals]bufpool.obj : [.libnurbs.internals]bufpool.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]bufpool.obj [.libnurbs.internals]bufpool.cc - -[.libnurbs.internals]cachingeval.obj : [.libnurbs.internals]cachingeval.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]cachingeval.obj [.libnurbs.internals]cachingeval.cc - -[.libnurbs.internals]ccw.obj : [.libnurbs.internals]ccw.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]ccw.obj [.libnurbs.internals]ccw.cc - -[.libnurbs.internals]coveandtiler.obj : [.libnurbs.internals]coveandtiler.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]coveandtiler.obj [.libnurbs.internals]coveandtiler.cc - -[.libnurbs.internals]curve.obj : [.libnurbs.internals]curve.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]curve.obj [.libnurbs.internals]curve.cc - -[.libnurbs.internals]curvelist.obj : [.libnurbs.internals]curvelist.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]curvelist.obj [.libnurbs.internals]curvelist.cc - -[.libnurbs.internals]curvesub.obj : [.libnurbs.internals]curvesub.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]curvesub.obj [.libnurbs.internals]curvesub.cc - -[.libnurbs.internals]dataTransform.obj : [.libnurbs.internals]dataTransform.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]dataTransform.obj [.libnurbs.internals]dataTransform.cc - -[.libnurbs.internals]displaylist.obj : [.libnurbs.internals]displaylist.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]displaylist.obj [.libnurbs.internals]displaylist.cc - -[.libnurbs.internals]flist.obj : [.libnurbs.internals]flist.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]flist.obj [.libnurbs.internals]flist.cc - -[.libnurbs.internals]flistsorter.obj : [.libnurbs.internals]flistsorter.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]flistsorter.obj [.libnurbs.internals]flistsorter.cc - -[.libnurbs.internals]hull.obj : [.libnurbs.internals]hull.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]hull.obj [.libnurbs.internals]hull.cc - -[.libnurbs.internals]intersect.obj : [.libnurbs.internals]intersect.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]intersect.obj [.libnurbs.internals]intersect.cc - -[.libnurbs.internals]knotvector.obj : [.libnurbs.internals]knotvector.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]knotvector.obj [.libnurbs.internals]knotvector.cc - -[.libnurbs.internals]mapdesc.obj : [.libnurbs.internals]mapdesc.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]mapdesc.obj [.libnurbs.internals]mapdesc.cc - -[.libnurbs.internals]mapdescv.obj : [.libnurbs.internals]mapdescv.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]mapdescv.obj [.libnurbs.internals]mapdescv.cc - -[.libnurbs.internals]maplist.obj : [.libnurbs.internals]maplist.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]maplist.obj [.libnurbs.internals]maplist.cc - -[.libnurbs.internals]mesher.obj : [.libnurbs.internals]mesher.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]mesher.obj [.libnurbs.internals]mesher.cc - -[.libnurbs.internals]monoTriangulationBackend.obj : [.libnurbs.internals]monoTriangulationBackend.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]monoTriangulationBackend.obj [.libnurbs.internals]monoTriangulationBackend.cc - -[.libnurbs.internals]monotonizer.obj : [.libnurbs.internals]monotonizer.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]monotonizer.obj [.libnurbs.internals]monotonizer.cc - -[.libnurbs.internals]mycode.obj : [.libnurbs.internals]mycode.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]mycode.obj [.libnurbs.internals]mycode.cc - -[.libnurbs.internals]nurbsinterfac.obj : [.libnurbs.internals]nurbsinterfac.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]nurbsinterfac.obj [.libnurbs.internals]nurbsinterfac.cc - -[.libnurbs.internals]nurbstess.obj : [.libnurbs.internals]nurbstess.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]nurbstess.obj [.libnurbs.internals]nurbstess.cc - -[.libnurbs.internals]patch.obj : [.libnurbs.internals]patch.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]patch.obj [.libnurbs.internals]patch.cc - -[.libnurbs.internals]patchlist.obj : [.libnurbs.internals]patchlist.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]patchlist.obj [.libnurbs.internals]patchlist.cc - -[.libnurbs.internals]quilt.obj : [.libnurbs.internals]quilt.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]quilt.obj [.libnurbs.internals]quilt.cc - -[.libnurbs.internals]reader.obj : [.libnurbs.internals]reader.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]reader.obj [.libnurbs.internals]reader.cc - -[.libnurbs.internals]renderhints.obj : [.libnurbs.internals]renderhints.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]renderhints.obj [.libnurbs.internals]renderhints.cc - -[.libnurbs.internals]slicer.obj : [.libnurbs.internals]slicer.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]slicer.obj [.libnurbs.internals]slicer.cc - -[.libnurbs.internals]sorter.obj : [.libnurbs.internals]sorter.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]sorter.obj [.libnurbs.internals]sorter.cc - -[.libnurbs.internals]splitarcs.obj : [.libnurbs.internals]splitarcs.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]splitarcs.obj [.libnurbs.internals]splitarcs.cc - -[.libnurbs.internals]subdivider.obj : [.libnurbs.internals]subdivider.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]subdivider.obj [.libnurbs.internals]subdivider.cc - -[.libnurbs.internals]tobezier.obj : [.libnurbs.internals]tobezier.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]tobezier.obj [.libnurbs.internals]tobezier.cc - -[.libnurbs.internals]trimline.obj : [.libnurbs.internals]trimline.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]trimline.obj [.libnurbs.internals]trimline.cc - -[.libnurbs.internals]trimregion.obj : [.libnurbs.internals]trimregion.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]trimregion.obj [.libnurbs.internals]trimregion.cc - -[.libnurbs.internals]trimvertpool.obj : [.libnurbs.internals]trimvertpool.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]trimvertpool.obj [.libnurbs.internals]trimvertpool.cc - -[.libnurbs.internals]uarray.obj : [.libnurbs.internals]uarray.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]uarray.obj [.libnurbs.internals]uarray.cc - -[.libnurbs.internals]varray.obj : [.libnurbs.internals]varray.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.internals]varray.obj [.libnurbs.internals]varray.cc - -[.libnurbs.nurbtess]directedLine.obj : [.libnurbs.nurbtess]directedLine.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]directedLine.obj [.libnurbs.nurbtess]directedLine.cc - -[.libnurbs.nurbtess]gridWrap.obj : [.libnurbs.nurbtess]gridWrap.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]gridWrap.obj [.libnurbs.nurbtess]gridWrap.cc - -[.libnurbs.nurbtess]monoChain.obj : [.libnurbs.nurbtess]monoChain.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]monoChain.obj [.libnurbs.nurbtess]monoChain.cc - -[.libnurbs.nurbtess]monoPolyPart.obj : [.libnurbs.nurbtess]monoPolyPart.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]monoPolyPart.obj [.libnurbs.nurbtess]monoPolyPart.cc - -[.libnurbs.nurbtess]monoTriangulation.obj : [.libnurbs.nurbtess]monoTriangulation.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]monoTriangulation.obj [.libnurbs.nurbtess]monoTriangulation.cc - -[.libnurbs.nurbtess]partitionX.obj : [.libnurbs.nurbtess]partitionX.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]partitionX.obj [.libnurbs.nurbtess]partitionX.cc - -[.libnurbs.nurbtess]partitionY.obj : [.libnurbs.nurbtess]partitionY.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]partitionY.obj [.libnurbs.nurbtess]partitionY.cc - -[.libnurbs.nurbtess]polyDBG.obj : [.libnurbs.nurbtess]polyDBG.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]polyDBG.obj [.libnurbs.nurbtess]polyDBG.cc - -[.libnurbs.nurbtess]polyUtil.obj : [.libnurbs.nurbtess]polyUtil.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]polyUtil.obj [.libnurbs.nurbtess]polyUtil.cc - -[.libnurbs.nurbtess]primitiveStream.obj : [.libnurbs.nurbtess]primitiveStream.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]primitiveStream.obj [.libnurbs.nurbtess]primitiveStream.cc - -[.libnurbs.nurbtess]quicksort.obj : [.libnurbs.nurbtess]quicksort.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]quicksort.obj [.libnurbs.nurbtess]quicksort.cc - -[.libnurbs.nurbtess]rectBlock.obj : [.libnurbs.nurbtess]rectBlock.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]rectBlock.obj [.libnurbs.nurbtess]rectBlock.cc - -[.libnurbs.nurbtess]sampleComp.obj : [.libnurbs.nurbtess]sampleComp.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleComp.obj [.libnurbs.nurbtess]sampleComp.cc - -[.libnurbs.nurbtess]sampleCompBot.obj : [.libnurbs.nurbtess]sampleCompBot.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleCompBot.obj [.libnurbs.nurbtess]sampleCompBot.cc - -[.libnurbs.nurbtess]sampleCompRight.obj : [.libnurbs.nurbtess]sampleCompRight.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleCompRight.obj [.libnurbs.nurbtess]sampleCompRight.cc - -[.libnurbs.nurbtess]sampleCompTop.obj : [.libnurbs.nurbtess]sampleCompTop.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleCompTop.obj [.libnurbs.nurbtess]sampleCompTop.cc - -[.libnurbs.nurbtess]sampleMonoPoly.obj : [.libnurbs.nurbtess]sampleMonoPoly.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampleMonoPoly.obj [.libnurbs.nurbtess]sampleMonoPoly.cc - -[.libnurbs.nurbtess]sampledLine.obj : [.libnurbs.nurbtess]sampledLine.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]sampledLine.obj [.libnurbs.nurbtess]sampledLine.cc - -[.libnurbs.nurbtess]searchTree.obj : [.libnurbs.nurbtess]searchTree.cc - $(CXX) $(CFLAGS) /obj=[.libnurbs.nurbtess]searchTree.obj [.libnurbs.nurbtess]searchTree.cc diff --git a/src/glut/dos/Makefile.DJ b/src/glut/dos/Makefile.DJ deleted file mode 100644 index 7e4e0b8576..0000000000 --- a/src/glut/dos/Makefile.DJ +++ /dev/null @@ -1,126 +0,0 @@ -# DOS/DJGPP Mesa Utility Toolkit -# Version: 1.0 -# -# Copyright (C) 2005 Daniel Borca All Rights Reserved. -# -# Permission is hereby granted, free of charge, to any person obtaining a -# copy of this software and associated documentation files (the "Software"), -# to deal in the Software without restriction, including without limitation -# the rights to use, copy, modify, merge, publish, distribute, sublicense, -# and/or sell copies of the Software, and to permit persons to whom the -# Software is furnished to do so, subject to the following conditions: -# -# The above copyright notice and this permission notice shall be included -# in all copies or substantial portions of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL -# DANIEL BORCA BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER -# IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN -# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - -# -# Available options: -# -# Environment variables: -# CFLAGS -# -# GLIDE path to Glide3 SDK; used to resolve DXEs. -# default = $(TOP)/glide3 -# -# Targets: -# all: build GLUT -# clean: remove object files -# - - - -.PHONY: all clean - -TOP = ../../.. -GLIDE ?= $(TOP)/glide3 -LIBDIR = $(TOP)/lib -GLUT_LIB = libglut.a -GLUT_DXE = glut.dxe -GLUT_IMP = libiglut.a - -export LD_LIBRARY_PATH := $(LD_LIBRARY_PATH);$(LIBDIR);$(GLIDE)/lib - -CC = gcc -CFLAGS += -I$(TOP)/include -I. -IPC_HW -CFLAGS += -DGLUT_IMPORT_LIB - -AR = ar -ARFLAGS = crus - -HAVEDXE3 = $(wildcard $(DJDIR)/bin/dxe3gen.exe) - -ifeq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),) -UNLINK = del $(subst /,\,$(1)) -else -UNLINK = $(RM) $(1) -endif - -CORE_SOURCES = \ - loop.c \ - callback.c \ - color.c \ - extens.c \ - init.c \ - menu.c \ - mouse.c \ - overlay.c \ - state.c \ - util.c \ - window.c \ - f8x13.c \ - f9x15.c \ - hel10.c \ - hel12.c \ - hel18.c \ - tr10.c \ - tr24.c \ - mroman.c \ - roman.c \ - bitmap.c \ - stroke.c \ - teapot.c \ - shapes.c - -PC_HW_SOURCES = \ - PC_HW/pc_hw.c \ - PC_HW/pc_keyb.c \ - PC_HW/pc_mouse.c \ - PC_HW/pc_timer.c \ - PC_HW/pc_irq.S - -SOURCES = $(CORE_SOURCES) $(PC_HW_SOURCES) - -OBJECTS = $(addsuffix .o,$(basename $(SOURCES))) - -.c.o: - $(CC) -o $@ $(CFLAGS) -c $< -.S.o: - $(CC) -o $@ $(CFLAGS) -c $< -.s.o: - $(CC) -o $@ $(CFLAGS) -x assembler-with-cpp -c $< - -all: $(LIBDIR)/$(GLUT_LIB) $(LIBDIR)/$(GLUT_DXE) $(LIBDIR)/$(GLUT_IMP) - -$(LIBDIR)/$(GLUT_LIB): $(OBJECTS) - $(AR) $(ARFLAGS) $@ $^ - -$(LIBDIR)/$(GLUT_DXE) $(LIBDIR)/$(GLUT_IMP): $(OBJECTS) -ifeq ($(HAVEDXE3),) - $(warning Missing DXE3 package... Skipping $(GLUT_DXE)) -else - -dxe3gen -o $(LIBDIR)/$(GLUT_DXE) -Y $(LIBDIR)/$(GLUT_IMP) -D "MesaGLUT DJGPP" -E _glut -P gl.dxe -U $^ -endif - -clean: - -$(call UNLINK,*.o) - -$(call UNLINK,PC_HW/*.o) - --include depend diff --git a/src/glut/glx/Makefile.mgw b/src/glut/glx/Makefile.mgw deleted file mode 100644 index ae4eb6addc..0000000000 --- a/src/glut/glx/Makefile.mgw +++ /dev/null @@ -1,198 +0,0 @@ -# Mesa 3-D graphics library -# Version: 5.1 -# -# Copyright (C) 1999-2003 Brian Paul All Rights Reserved. -# -# Permission is hereby granted, free of charge, to any person obtaining a -# copy of this software and associated documentation files (the "Software"), -# to deal in the Software without restriction, including without limitation -# the rights to use, copy, modify, merge, publish, distribute, sublicense, -# and/or sell copies of the Software, and to permit persons to whom the -# Software is furnished to do so, subject to the following conditions: -# -# The above copyright notice and this permission notice shall be included -# in all copies or substantial portions of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL -# BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN -# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN -# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -# MinGW core makefile v1.4 for Mesa -# -# Copyright (C) 2002 - Daniel Borca -# Email : dborca@users.sourceforge.net -# Web : http://www.geocities.com/dborca - -# MinGW core-glut makefile updated for Mesa 7.0 -# -# Updated : by Heromyth, on 2007-7-21 -# Email : zxpmyth@yahoo.com.cn -# Bugs : 1) All the default settings work fine. But the setting X86=1 can't work. -# The others havn't been tested yet. -# 2) The generated DLLs are *not* compatible with the ones built -# with the other compilers like VC8, especially for GLUT. -# 3) Although more tests are needed, it can be used individually! - - -# -# Available options: -# -# Environment variables: -# CFLAGS -# -# GLIDE path to Glide3 SDK; used with FX. -# default = $(TOP)/glide3 -# FX=1 build for 3dfx Glide3. Note that this disables -# compilation of most WMesa code and requires fxMesa. -# As a consequence, you'll need the Win32 Glide3 -# library to build any application. -# default = no -# ICD=1 build the installable client driver interface -# (windows opengl driver interface) -# default = no -# X86=1 optimize for x86 (if possible, use MMX, SSE, 3DNow). -# default = no -# -# Targets: -# all: build GL -# clean: remove object files -# - - - -.PHONY: all clean -.INTERMEDIATE: x86/gen_matypes.exe -.SUFFIXES: .rc .res - -# Set this to the prefix of your build tools, i.e. mingw32- -TOOLS_PREFIX = mingw32- - -TOP = ../../.. - -LIBDIR = $(TOP)/lib - -GLUT_DLL = glut32.dll -GLUT_IMP = libglut32.a -GLUT_DEF = glut.def - -include $(TOP)/configs/config.mgw -GLUT_USING_STDCALL ?= 1 - - - -LDLIBS = -L$(LIBDIR) -lwinmm -lgdi32 -luser32 -lopengl32 -lglu32 -LDFLAGS = -Wl,--out-implib=$(LIBDIR)/$(GLUT_IMP) -Wl,--output-def=$(LIBDIR)/$(GLUT_DEF) - -CFLAGS += -DBUILD_GLUT32 -DGLUT_BUILDING_LIB -DMESA -D_DLL - -ifeq ($(GL_USING_STDCALL),0) - CFLAGS += -DGL_NO_STDCALL -endif - -ifeq ($(GLUT_USING_STDCALL),1) - CFLAGS += -D_STDCALL_SUPPORTED - LDFLAGS += -Wl,--add-stdcall-alias -else - CFLAGS += -DGLUT_NO_STDCALL -endif - -CFLAGS += -DNDEBUG -DLIBRARYBUILD -I$(TOP)/include - -CC = gcc -CXX = g++ -CXXFLAGS = $(CFLAGS) - -AR = ar -ARFLAGS = crus - -UNLINK = del $(subst /,\,$(1)) -ifneq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),) -UNLINK = $(RM) $(1) -endif -ifneq ($(wildcard $(addsuffix /rm,$(subst :, ,$(PATH)))),) -UNLINK = $(RM) $(1) -endif - -HDRS = glutint.h glutstroke.h glutbitmap.h glutwin32.h stroke.h win32_glx.h win32_x11.h - -SRCS = \ - glut_bitmap.c \ - glut_bwidth.c \ - glut_cindex.c \ - glut_cmap.c \ - glut_cursor.c \ - glut_dials.c \ - glut_dstr.c \ - glut_event.c \ - glut_ext.c \ - glut_fbc.c \ - glut_fullscrn.c \ - glut_gamemode.c \ - glut_get.c \ - glut_init.c \ - glut_input.c \ - glut_joy.c \ - glut_key.c \ - glut_keyctrl.c \ - glut_keyup.c \ - glut_mesa.c \ - glut_modifier.c \ - glut_overlay.c \ - glut_shapes.c \ - glut_space.c \ - glut_stroke.c \ - glut_swap.c \ - glut_swidth.c \ - glut_tablet.c \ - glut_teapot.c \ - glut_util.c \ - glut_vidresize.c \ - glut_warp.c \ - glut_win.c \ - glut_winmisc.c \ - win32_glx.c \ - win32_menu.c \ - win32_util.c \ - win32_winproc.c \ - win32_x11.c - - -SRCSSEMIGENS = \ - glut_8x13.c \ - glut_9x15.c \ - glut_hel10.c \ - glut_hel12.c \ - glut_hel18.c \ - glut_mroman.c \ - glut_roman.c \ - glut_tr10.c \ - glut_tr24.c - - - -SOURCES = $(SRCS) $(SRCSSEMIGENS) - -OBJECTS = $(addsuffix .o,$(basename $(SOURCES))) - -.c.o: - $(CC) -o $@ $(CFLAGS) -c $< -.cc.o: - $(CXX) -o $@ $(CXXFLAGS) -c $< - - -all: $(LIBDIR) $(LIBDIR)/$(GLUT_DLL) $(LIBDIR)/$(GLUT_IMP) - -$(LIBDIR): - mkdir -p $(LIBDIR) - -$(LIBDIR)/$(GLUT_DLL) $(LIBDIR)/$(GLUT_IMP): $(OBJECTS) - $(CXX) -shared -fPIC -o $(LIBDIR)/$(GLUT_DLL) $(LDFLAGS) \ - $^ $(LDLIBS) - - - -clean: - -$(call UNLINK,*.o) \ No newline at end of file diff --git a/src/glut/glx/descrip.mms b/src/glut/glx/descrip.mms deleted file mode 100644 index 358b417511..0000000000 --- a/src/glut/glx/descrip.mms +++ /dev/null @@ -1,208 +0,0 @@ -# Makefile for GLUT for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl - -.first - define gl [---.include.gl] - -.include [---]mms.config - -##### MACROS ##### -GLUT_MAJOR = 3 -GLUT_MINOR = 7 - -VPATH = RCS - -INCDIR = [---.include] -LIBDIR = [---.lib] -CFLAGS = /nowarn/include=$(INCDIR)/prefix=all/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = \ -glut_8x13.c \ -glut_9x15.c \ -glut_bitmap.c \ -glut_bwidth.c \ -glut_cindex.c \ -glut_cmap.c \ -glut_cursor.c \ -glut_dials.c \ -glut_dstr.c \ -glut_event.c \ -glut_ext.c \ -glut_fullscrn.c \ -glut_gamemode.c \ -glut_get.c \ -glut_glxext.c \ -glut_hel10.c \ -glut_hel12.c \ -glut_hel18.c \ -glut_init.c \ -glut_input.c \ -glut_joy.c \ -glut_key.c \ -glut_keyctrl.c \ -glut_keyup.c \ -glut_menu.c \ -glut_menu2.c \ -glut_mesa.c \ -glut_modifier.c \ -glut_mroman.c \ -glut_overlay.c \ -glut_roman.c \ -glut_shapes.c \ -glut_space.c \ -glut_stroke.c \ -glut_swap.c \ -glut_swidth.c \ -glut_tablet.c \ -glut_teapot.c \ -glut_tr10.c \ -glut_tr24.c \ -glut_util.c \ -glut_vidresize.c \ -glut_warp.c \ -glut_win.c \ -glut_winmisc.c \ -layerutil.c - -OBJECTS0=glut_8x13.obj,\ -glut_9x15.obj,\ -glut_bitmap.obj,\ -glut_bwidth.obj,\ -glut_cindex.obj,\ -glut_cmap.obj,\ -glut_cursor.obj,\ -glut_dials.obj,\ -glut_dstr.obj,\ -glut_event.obj,\ -glut_ext.obj,\ -glut_fullscrn.obj,\ -glut_gamemode.obj - -OBJECTS1=glut_get.obj,\ -glut_glxext.obj,\ -glut_hel10.obj,\ -glut_hel12.obj,\ -glut_hel18.obj,\ -glut_init.obj,\ -glut_input.obj,\ -glut_joy.obj,\ -glut_key.obj,\ -glut_keyctrl.obj,\ -glut_keyup.obj,\ -glut_menu.obj,\ -glut_menu2.obj,\ -glut_mesa.obj,\ -glut_modifier.obj - -OBJECTS2=glut_mroman.obj,\ -glut_overlay.obj,\ -glut_roman.obj,\ -glut_shapes.obj,\ -glut_space.obj,\ -glut_stroke.obj,\ -glut_swap.obj,\ -glut_swidth.obj,\ -glut_tablet.obj,\ -glut_teapot.obj,\ -glut_tr10.obj,\ -glut_tr24.obj,\ -glut_util.obj,\ -glut_vidresize.obj - -OBJECTS3=glut_warp.obj,\ -glut_win.obj,\ -glut_winmisc.obj,\ -layerutil.obj - -##### RULES ##### - -VERSION=Glut V3.7 - -##### TARGETS ##### - -# Make the library -$(LIBDIR)$(GLUT_LIB) : $(OBJECTS0) $(OBJECTS1) $(OBJECTS2) $(OBJECTS3) - @ $(MAKELIB) $(GLUT_LIB) $(OBJECTS0) - @ library $(GLUT_LIB) $(OBJECTS1) - @ library $(GLUT_LIB) $(OBJECTS2) - @ library $(GLUT_LIB) $(OBJECTS3) - @ rename $(GLUT_LIB)* $(LIBDIR) -.ifdef SHARE - @ WRITE_ SYS$OUTPUT " generating mesagl1.opt" - @ OPEN_/WRITE FILE mesagl1.opt - @ WRITE_ FILE "!" - @ WRITE_ FILE "! mesagl1.opt generated by DESCRIP.$(MMS_EXT)" - @ WRITE_ FILE "!" - @ WRITE_ FILE "IDENTIFICATION=""$(VERSION)""" - @ WRITE_ FILE "GSMATCH=LEQUAL,3,7 - @ WRITE_ FILE "$(OBJECTS0)" - @ WRITE_ FILE "$(OBJECTS1)" - @ WRITE_ FILE "$(OBJECTS2)" - @ WRITE_ FILE "$(OBJECTS3)" - @ WRITE_ FILE "[---.lib]libmesaglu.exe/SHARE" - @ WRITE_ FILE "[---.lib]libmesagl.exe/SHARE" - @ write file "sys$library:decw$xmulibshr.exe/share" - @ WRITE_ FILE "SYS$SHARE:DECW$XEXTLIBSHR/SHARE" - @ WRITE_ FILE "SYS$SHARE:DECW$XLIBSHR/SHARE" - @ CLOSE_ FILE - @ WRITE_ SYS$OUTPUT " generating mesagl.map ..." - @ CXXLINK_/NODEB/NOSHARE/NOEXE/MAP=mesagl.map/FULL mesagl1.opt/OPT - @ WRITE_ SYS$OUTPUT " analyzing mesagl.map ..." - @ @[---.vms]ANALYZE_MAP.COM mesagl.map mesagl.opt - @ WRITE_ SYS$OUTPUT " linking $(GLUT_SHAR) ..." - @ CXXLINK_/NODEB/SHARE=$(GLUT_SHAR)/MAP=mesagl.map/FULL mesagl1.opt/opt,mesagl.opt/opt - @ rename $(GLUT_SHAR)* $(LIBDIR) -.endif - -clean : - delete *.obj;* - purge - -include mms_depend. - -glut_8x13.obj : glut_8x13.c -glut_9x15.obj : glut_9x15.c -glut_bitmap.obj : glut_bitmap.c -glut_bwidth.obj : glut_bwidth.c -glut_cindex.obj : glut_cindex.c -glut_cmap.obj : glut_cmap.c -glut_cursor.obj : glut_cursor.c -glut_dials.obj : glut_dials.c -glut_dstr.obj : glut_dstr.c -glut_event.obj : glut_event.c -glut_ext.obj : glut_ext.c -glut_fullscrn.obj : glut_fullscrn.c -glut_gamemode.obj : glut_gamemode.c -glut_get.obj : glut_get.c -glut_glxext.obj : glut_glxext.c -glut_hel10.obj : glut_hel10.c -glut_hel12.obj : glut_hel12.c -glut_hel18.obj : glut_hel18.c -glut_init.obj : glut_init.c -glut_input.obj : glut_input.c -glut_joy.obj : glut_joy.c -glut_key.obj : glut_key.c -glut_keyctrl.obj : glut_keyctrl.c -glut_keyup.obj : glut_keyup.c -glut_menu.obj : glut_menu.c -glut_menu2.obj : glut_menu2.c -glut_mesa.obj : glut_mesa.c -glut_modifier.obj : glut_modifier.c -glut_mroman.obj : glut_mroman.c -glut_overlay.obj : glut_overlay.c -glut_roman.obj : glut_roman.c -glut_shapes.obj : glut_shapes.c -glut_space.obj : glut_space.c -glut_stroke.obj : glut_stroke.c -glut_swap.obj : glut_swap.c -glut_swidth.obj : glut_swidth.c -glut_tablet.obj : glut_tablet.c -glut_teapot.obj : glut_teapot.c -glut_tr10.obj : glut_tr10.c -glut_tr24.obj : glut_tr24.c -glut_util.obj : glut_util.c -glut_vidresize.obj : glut_vidresize.c -glut_warp.obj : glut_warp.c -glut_win.obj : glut_win.c -glut_winmisc.obj : glut_winmisc.c -layerutil.obj : layerutil.c diff --git a/src/glut/glx/mms_depend b/src/glut/glx/mms_depend deleted file mode 100644 index 98f87c29e2..0000000000 --- a/src/glut/glx/mms_depend +++ /dev/null @@ -1,72 +0,0 @@ -# DO NOT DELETE - -glut_8x13.obj : glutbitmap.h [---.include.gl]gl.h -glut_9x15.obj : glutbitmap.h [---.include.gl]gl.h -glut_bitmap.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_bitmap.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_bitmap.obj : glutbitmap.h -glut_bwidth.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_bwidth.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_bwidth.obj : glutbitmap.h -glut_cindex.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_cindex.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_cindex.obj : layerutil.h -glut_cursor.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_cursor.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_dials.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h -glut_dials.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h -glut_dstr.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h -glut_dstr.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h -glut_event.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_event.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_ext.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_ext.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_fullscrn.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h -glut_fullscrn.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h -glut_fullscrn.obj : [---.include.gl]glu.h -glut_get.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_get.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_hel10.obj : glutbitmap.h [---.include.gl]gl.h -glut_hel12.obj : glutbitmap.h [---.include.gl]gl.h -glut_hel18.obj : glutbitmap.h [---.include.gl]gl.h -glut_init.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_init.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_menu.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_menu.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h layerutil.h -glut_mesa.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h -glut_mesa.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h -glut_modifier.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h -glut_modifier.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h -glut_modifier.obj : [---.include.gl]glu.h -glut_mroman.obj : glutstroke.h -glut_overlay.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_overlay.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_overlay.obj : layerutil.h -glut_roman.obj : glutstroke.h -glut_shapes.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_shapes.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_space.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h -glut_space.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h -glut_stroke.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_stroke.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_stroke.obj : glutstroke.h -glut_swidth.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_swidth.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_swidth.obj : glutstroke.h -glut_tablet.obj : glutint.h [---.include.gl]glx.h [---.include.gl]gl.h -glut_tablet.obj : [---.include.gl]xmesa.h [---.include.gl]glut.h [---.include.gl]glu.h -glut_teapot.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_tr10.obj : glutbitmap.h [---.include.gl]gl.h -glut_tr24.obj : glutbitmap.h [---.include.gl]gl.h -glut_util.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_util.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_vidresize.obj : [---.include.gl]glx.h [---.include.gl]gl.h -glut_vidresize.obj : [---.include.gl]xmesa.h glutint.h [---.include.gl]glut.h -glut_vidresize.obj : [---.include.gl]glu.h -glut_warp.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_warp.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_win.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_win.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -glut_winmisc.obj : [---.include.gl]glut.h [---.include.gl]gl.h [---.include.gl]glu.h -glut_winmisc.obj : glutint.h [---.include.gl]glx.h [---.include.gl]xmesa.h -layerutil.obj : layerutil.h diff --git a/src/mesa/Makefile.DJ b/src/mesa/Makefile.DJ deleted file mode 100644 index 06a13fb1ab..0000000000 --- a/src/mesa/Makefile.DJ +++ /dev/null @@ -1,166 +0,0 @@ -# Mesa 3-D graphics library -# Version: 5.1 -# -# Copyright (C) 1999-2003 Brian Paul All Rights Reserved. -# -# Permission is hereby granted, free of charge, to any person obtaining a -# copy of this software and associated documentation files (the "Software"), -# to deal in the Software without restriction, including without limitation -# the rights to use, copy, modify, merge, publish, distribute, sublicense, -# and/or sell copies of the Software, and to permit persons to whom the -# Software is furnished to do so, subject to the following conditions: -# -# The above copyright notice and this permission notice shall be included -# in all copies or substantial portions of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL -# BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN -# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN -# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -# DOS/DJGPP core makefile v1.7 for Mesa -# -# Copyright (C) 2002 - Daniel Borca -# Email : dborca@users.sourceforge.net -# Web : http://www.geocities.com/dborca - - -# -# Available options: -# -# Environment variables: -# CFLAGS -# -# GLIDE path to Glide3 SDK; used with FX. -# default = $(TOP)/glide3 -# FX=1 build for 3dfx Glide3. Note that this disables -# compilation of most DMesa code and requires fxMesa. -# As a consequence, you'll need the DJGPP Glide3 -# library to build any application. -# default = no -# X86=1 optimize for x86 (if possible, use MMX, SSE, 3DNow). -# default = no -# -# Targets: -# all: build GL -# clean: remove object files -# - - - -.PHONY: all clean -.INTERMEDIATE: x86/gen_matypes.exe - -TOP = ../.. -GLIDE ?= $(TOP)/glide3 -LIBDIR = $(TOP)/lib -GL_LIB = libgl.a -GL_DXE = gl.dxe -GL_IMP = libigl.a - -export LD_LIBRARY_PATH := $(LD_LIBRARY_PATH);$(LIBDIR);$(GLIDE)/lib - -CC = gcc -CFLAGS += $(INCLUDE_DIRS) -CFLAGS += -DUSE_EXTERNAL_DXTN_LIB=1 -ifeq ($(FX),1) -CFLAGS += -D__DOS__ -CFLAGS += -I$(GLIDE)/include -DFX -LIBNAME = "Mesa/FX DJGPP" -else -LIBNAME = "Mesa DJGPP" -endif - -AR = ar -ARFLAGS = crus - -HAVEDXE3 = $(wildcard $(DJDIR)/bin/dxe3gen.exe) - -ifeq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),) -UNLINK = del $(subst /,\,$(1)) -else -UNLINK = $(RM) $(1) -endif - -include sources - -ifeq ($(X86),1) -CFLAGS += -DUSE_X86_ASM -CFLAGS += -DUSE_MMX_ASM -CFLAGS += -DUSE_SSE_ASM -CFLAGS += -DUSE_3DNOW_ASM -X86_SOURCES += $(X86_API) -else -X86_SOURCES = -endif - -DRIVER_SOURCES = \ - drivers/dos/dmesa.c -ifeq ($(FX),1) -DRIVER_SOURCES += \ - $(GLIDE_DRIVER_SOURCES) -else -DRIVER_SOURCES += \ - $(OSMESA_DRIVER_SOURCES) \ - drivers/dos/video.c \ - drivers/dos/vesa.c \ - drivers/dos/blit.S \ - drivers/dos/vga.c \ - drivers/dos/null.c \ - drivers/dos/dpmi.c -endif - -SOURCES = $(CORE_SOURCES) $(X86_SOURCES) $(COMMON_DRIVER_SOURCES) $(DRIVER_SOURCES) - -OBJECTS = $(addsuffix .o,$(basename $(SOURCES))) - -X86_OBJECTS = $(addsuffix .o,$(basename $(X86_SOURCES))) - -.c.o: - $(CC) -o $@ $(CFLAGS) -c $< -.S.o: - $(CC) -o $@ $(CFLAGS) -c $< -.s.o: - $(CC) -o $@ $(CFLAGS) -x assembler-with-cpp -c $< - -all: $(LIBDIR)/$(GL_LIB) $(LIBDIR)/$(GL_DXE) $(LIBDIR)/$(GL_IMP) - -$(LIBDIR)/$(GL_LIB): $(OBJECTS) - $(AR) $(ARFLAGS) $@ $^ - -$(LIBDIR)/$(GL_DXE) $(LIBDIR)/$(GL_IMP): $(OBJECTS) -ifeq ($(HAVEDXE3),) - $(warning Missing DXE3 package... Skipping $(GL_DXE)) -else -ifeq ($(FX),1) - -dxe3gen -o $(LIBDIR)/$(GL_DXE) -Y $(LIBDIR)/$(GL_IMP) -D $(LIBNAME) -E _gl -E _DMesa -P glide3x.dxe -U $^ -else - -dxe3gen -o $(LIBDIR)/$(GL_DXE) -Y $(LIBDIR)/$(GL_IMP) -D $(LIBNAME) -E _gl -E _DMesa -U $^ -endif -endif - -$(X86_OBJECTS): x86/matypes.h - -x86/matypes.h: x86/gen_matypes.exe - $< > $@ - -x86/gen_matypes.exe: x86/gen_matypes.c - $(CC) -o $@ $(CFLAGS) -s $< - -clean: - -$(call UNLINK,array_cache/*.o) - -$(call UNLINK,glapi/*.o) - -$(call UNLINK,main/*.o) - -$(call UNLINK,math/*.o) - -$(call UNLINK,shader/*.o) - -$(call UNLINK,sparc/*.o) - -$(call UNLINK,ppc/*.o) - -$(call UNLINK,swrast/*.o) - -$(call UNLINK,swrast_setup/*.o) - -$(call UNLINK,tnl/*.o) - -$(call UNLINK,x86/*.o) - -$(call UNLINK,drivers/common/*.o) - -$(call UNLINK,drivers/dos/*.o) - -$(call UNLINK,drivers/glide/*.o) diff --git a/src/mesa/Makefile.mgw b/src/mesa/Makefile.mgw deleted file mode 100644 index 3b52834bd1..0000000000 --- a/src/mesa/Makefile.mgw +++ /dev/null @@ -1,235 +0,0 @@ -# Mesa 3-D graphics library -# Version: 7.0 -# -# Copyright (C) 1999-2003 Brian Paul All Rights Reserved. -# -# Permission is hereby granted, free of charge, to any person obtaining a -# copy of this software and associated documentation files (the "Software"), -# to deal in the Software without restriction, including without limitation -# the rights to use, copy, modify, merge, publish, distribute, sublicense, -# and/or sell copies of the Software, and to permit persons to whom the -# Software is furnished to do so, subject to the following conditions: -# -# The above copyright notice and this permission notice shall be included -# in all copies or substantial portions of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL -# BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN -# AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN -# CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - -# MinGW core makefile v1.4 for Mesa -# -# Copyright (C) 2002 - Daniel Borca -# Email : dborca@users.sourceforge.net -# Web : http://www.geocities.com/dborca - -# MinGW core-gl makefile updated for Mesa 7.0 -# -# updated : by Heromyth, on 2007-7-21 -# Email : zxpmyth@yahoo.com.cn -# Bugs : 1) All the default settings work fine. But the setting X86=1 can't work. -# The others havn't been tested yet. -# 2) The generated DLLs are *not* compatible with the ones built -# with the other compilers like VC8, especially for GLUT. -# 3) Although more tests are needed, it can be used individually! - - -# -# Available options: -# -# Environment variables: -# CFLAGS -# -# GLIDE path to Glide3 SDK; used with FX. -# default = $(TOP)/glide3 -# FX=1 build for 3dfx Glide3. Note that this disables -# compilation of most WMesa code and requires fxMesa. -# As a consequence, you'll need the Win32 Glide3 -# library to build any application. -# default = no -# ICD=1 build the installable client driver interface -# (windows opengl driver interface) -# default = no -# X86=1 optimize for x86 (if possible, use MMX, SSE, 3DNow). -# default = no -# -# Targets: -# all: build GL -# clean: remove object files -# - - -.PHONY: all clean -.INTERMEDIATE: x86/gen_matypes.exe -.SUFFIXES: .rc .res - -# Set this to the prefix of your build tools, i.e. mingw32- -TOOLS_PREFIX = mingw32- - - - -TOP = ../.. -GLIDE ?= $(TOP)/glide3 -LIBDIR = $(TOP)/lib -ifeq ($(ICD),1) - GL_DLL = mesa32.dll - GL_IMP = libmesa32.a -else - GL_DLL = opengl32.dll - GL_IMP = libopengl32.a -endif - -GL_DEF = gl.def - -include $(TOP)/configs/config.mgw -GL_USING_STDCALL ?= 1 - -MESA_LIB = libmesa.a - -LDLIBS = -lgdi32 -luser32 -liberty -LDFLAGS = -Wl,--out-implib=$(LIBDIR)/$(GL_IMP) -Wl,--output-def=$(LIBDIR)/gl.def - -CC = $(TOOLS_PREFIX)gcc -CFLAGS += -DBUILD_GL32 -D_OPENGL32_ -D_DLL -DMESA_MINWARN -DNDEBUG -D_USRDLL -DGDI_EXPORTS - -ifeq ($(GL_USING_STDCALL),1) - LDFLAGS += -Wl,--add-stdcall-alias -else - CFLAGS += -DGL_NO_STDCALL -endif - -CFLAGS += -DUSE_EXTERNAL_DXTN_LIB=1 -ifeq ($(FX),1) - CFLAGS += -I$(GLIDE)/include -DFX - LDLIBS += -L$(GLIDE)/lib -lglide3x - GL_DEF = drivers/windows/fx/fxopengl.def - GL_RES = drivers/windows/fx/fx.rc -else - ifeq ($(ICD),1) - CFLAGS += -DUSE_MGL_NAMESPACE - GL_DEF = drivers/windows/icd/mesa.def - else - GL_DEF = drivers/windows/gdi/mesa.def - endif -endif - -AR = ar -ARFLAGS = crus - -UNLINK = del $(subst /,\,$(1)) -ifneq ($(wildcard $(addsuffix /rm.exe,$(subst ;, ,$(PATH)))),) -UNLINK = $(RM) $(1) -endif -ifneq ($(wildcard $(addsuffix /rm,$(subst :, ,$(PATH)))),) -UNLINK = $(RM) $(1) -endif - -include sources - -CFLAGS += $(INCLUDE_DIRS) - -ifeq ($(X86),1) -CFLAGS += -DUSE_X86_ASM -CFLAGS += -DUSE_MMX_ASM -CFLAGS += -DUSE_SSE_ASM -CFLAGS += -DUSE_3DNOW_ASM -X86_SOURCES += $(X86_API) -else -X86_SOURCES = -endif - -ifeq ($(FX),1) -DRIVER_SOURCES = \ - $(GLIDE_DRIVER_SOURCES) \ - drivers/windows/fx/fxwgl.c -else -ifeq ($(ICD),1) -DRIVER_SOURCES = \ - drivers/windows/gdi/wmesa.c \ - drivers/windows/icd/icd.c -else -DRIVER_SOURCES = \ - drivers/windows/gdi/wmesa.c \ - drivers/windows/gdi/wgl.c -endif -endif - -SOURCES = $(CORE_SOURCES) $(X86_SOURCES) $(COMMON_DRIVER_SOURCES) $(DRIVER_SOURCES) - -OBJECTS = $(addsuffix .o,$(basename $(SOURCES))) - -X86_OBJECTS = $(addsuffix .o,$(basename $(X86_SOURCES))) - -RESOURCE = $(GL_RES:.rc=.res) - -.c.o: - $(CC) -o $@ $(CFLAGS) -c $< -.s.o: - $(CC) -o $@ $(CFLAGS) -x assembler-with-cpp -c $< - -.rc.res: - windres -o $@ -Irc -Ocoff $< - -all: $(LIBDIR) $(LIBDIR)/$(GL_DLL) $(LIBDIR)/$(GL_IMP) - -$(LIBDIR): - mkdir -p $(LIBDIR) - -$(LIBDIR)/$(GL_DLL) $(LIBDIR)/$(GL_IMP): $(OBJECTS) $(RESOURCE) - $(CC) -shared -fPIC -o $(LIBDIR)/$(GL_DLL) $(LDFLAGS) \ - $^ $(LDLIBS) - -$(X86_OBJECTS): x86/matypes.h - -x86/matypes.h: x86/gen_matypes.exe - $(subst /,\,$< > $@) - -x86/gen_matypes.exe: x86/gen_matypes.c - $(CC) -o $@ $(CFLAGS) -s $< - -# [dBorca] -# glapi_x86.S needs some adjustments -# in order to generate correct entrypoints -# Trick: change the following condition to -# be always false if you need C entrypoints -# with USE_X86_ASM (useful for trace/debug) -ifeq (1,1) -x86/glapi_x86.o: x86/glapi_x86.S - $(CC) -o $@ $(CFLAGS) -DSTDCALL_API -c $< -else -main/dispatch.o: main/dispatch.c - $(CC) -o $@ $(CFLAGS) -UUSE_X86_ASM -c $< -glapi/glapi.o: glapi/glapi.c - $(CC) -o $@ $(CFLAGS) -UUSE_X86_ASM -c $< -endif - -# [dBorca] -# if we want codegen, we have to stdcall -tnl/t_vtx_x86_gcc.o: tnl/t_vtx_x86_gcc.S - $(CC) -o $@ $(CFLAGS) -DSTDCALL_API -c $< - -clean: - -$(call UNLINK,glapi/*.o) - -$(call UNLINK,main/*.o) - -$(call UNLINK,math/*.o) - -$(call UNLINK,vbo/*.o) - -$(call UNLINK,shader/*.o) - -$(call UNLINK,shader/slang/*.o) - -$(call UNLINK,shader/grammar/*.o) - -$(call UNLINK,sparc/*.o) - -$(call UNLINK,ppc/*.o) - -$(call UNLINK,swrast/*.o) - -$(call UNLINK,swrast_setup/*.o) - -$(call UNLINK,tnl/*.o) - -$(call UNLINK,x86/*.o) - -$(call UNLINK,x86/rtasm/*.o) - -$(call UNLINK,x86-64/*.o) - -$(call UNLINK,drivers/common/*.o) - -$(call UNLINK,drivers/glide/*.o) - -$(call UNLINK,drivers/windows/fx/*.o) - -$(call UNLINK,drivers/windows/fx/*.res) - -$(call UNLINK,drivers/windows/gdi/*.o) - -$(call UNLINK,drivers/windows/icd/*.o) diff --git a/src/mesa/descrip.mms b/src/mesa/descrip.mms deleted file mode 100644 index a12e3fc1b7..0000000000 --- a/src/mesa/descrip.mms +++ /dev/null @@ -1,26 +0,0 @@ -# Makefile for Mesa for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl - -all : - set default [.main] - $(MMS)$(MMSQUALIFIERS) - set default [-.glapi] - $(MMS)$(MMSQUALIFIERS) - set default [-.shader] - $(MMS)$(MMSQUALIFIERS) - set default [-.drivers.common] - $(MMS)$(MMSQUALIFIERS) - set default [-.x11] - $(MMS)$(MMSQUALIFIERS) - set default [-.osmesa] - $(MMS)$(MMSQUALIFIERS) - set default [--.math] - $(MMS)$(MMSQUALIFIERS) - set default [-.tnl] - $(MMS)$(MMSQUALIFIERS) - set default [-.swrast] - $(MMS)$(MMSQUALIFIERS) - set default [-.swrast_setup] - $(MMS)$(MMSQUALIFIERS) - set default [-.array_cache] - $(MMS)$(MMSQUALIFIERS) diff --git a/src/mesa/drivers/common/descrip.mms b/src/mesa/drivers/common/descrip.mms deleted file mode 100644 index c2c119db7f..0000000000 --- a/src/mesa/drivers/common/descrip.mms +++ /dev/null @@ -1,38 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.nano.tudelft.nl -# Last revision : 2 November 2005 - -.first - define gl [----.include.gl] - define math [--.math] - define tnl [--.tnl] - define swrast [--.swrast] - -.include [----]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [----.include],[--.main],[--.glapi],[--.shader] -LIBDIR = [----.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = driverfuncs.c - -OBJECTS =driverfuncs.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -driverfuncs.obj : driverfuncs.c diff --git a/src/mesa/drivers/osmesa/descrip.mms b/src/mesa/drivers/osmesa/descrip.mms deleted file mode 100644 index 5be194bcee..0000000000 --- a/src/mesa/drivers/osmesa/descrip.mms +++ /dev/null @@ -1,41 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl -# Last revision : 16 June 2003 - -.first - define gl [----.include.gl] - define math [--.math] - define tnl [--.tnl] - define swrast [--.swrast] - define swrast_setup [--.swrast_setup] - define array_cache [--.array_cache] - define drivers [-] - -.include [----]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [----.include],[--.main],[--.glapi] -LIBDIR = [----.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = osmesa.c - -OBJECTS = osmesa.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -osmesa.obj : osmesa.c diff --git a/src/mesa/drivers/x11/descrip.mms b/src/mesa/drivers/x11/descrip.mms deleted file mode 100644 index f181707a14..0000000000 --- a/src/mesa/drivers/x11/descrip.mms +++ /dev/null @@ -1,51 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl -# Last revision : 16 June 2003 - -.first - define gl [----.include.gl] - define math [--.math] - define tnl [--.tnl] - define swrast [--.swrast] - define swrast_setup [--.swrast_setup] - define array_cache [--.array_cache] - define drivers [-] - -.include [----]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [----.include],[--.main],[--.glapi] -LIBDIR = [----.lib] -CFLAGS =/include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = fakeglx.c glxapi.c xfonts.c xm_api.c xm_dd.c xm_line.c xm_span.c\ - xm_tri.c xm_buffer.c - -OBJECTS =fakeglx.obj,glxapi.obj,xfonts.obj,xm_api.obj,xm_dd.obj,xm_line.obj,\ - xm_span.obj,xm_tri.obj,xm_buffer.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -fakeglx.obj : fakeglx.c -glxapi.obj : glxapi.c -xfonts.obj : xfonts.c -xm_api.obj : xm_api.c -xm_buffer.obj : xm_buffer.c -xm_dd.obj : xm_dd.c -xm_line.obj : xm_line.c -xm_span.obj : xm_span.c -xm_tri.obj : xm_tri.c diff --git a/src/mesa/glapi/descrip.mms b/src/mesa/glapi/descrip.mms deleted file mode 100644 index 16bf6387e8..0000000000 --- a/src/mesa/glapi/descrip.mms +++ /dev/null @@ -1,37 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl -# Last revision : 16 June 2003 - -.first - define gl [---.include.gl] - -.include [---]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [---.include],[-.main] -LIBDIR = [---.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = glapi.c glthread.c - -OBJECTS = glapi.obj,glthread.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -glapi.obj : glapi.c - -glthread.obj : glthread.c diff --git a/src/mesa/main/descrip.mms b/src/mesa/main/descrip.mms deleted file mode 100644 index 2bd388be89..0000000000 --- a/src/mesa/main/descrip.mms +++ /dev/null @@ -1,221 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl -# Last revision : 10 May 2005 - -.first - define gl [---.include.gl] - define math [-.math] - define shader [-.shader] - -.include [---]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [---.include],[-.glapi],[-.shader] -LIBDIR = [---.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES =accum.c \ - api_arrayelt.c \ - api_loopback.c \ - api_noop.c \ - api_validate.c \ - attrib.c \ - arrayobj.c \ - blend.c \ - bufferobj.c \ - buffers.c \ - clip.c \ - colortab.c \ - context.c \ - convolve.c \ - debug.c \ - depth.c \ - depthstencil.c \ - dispatch.c \ - dlist.c \ - drawpix.c \ - enable.c \ - enums.c \ - eval.c \ - execmem.c \ - extensions.c \ - fbobject.c \ - feedback.c \ - fog.c \ - framebuffer.c \ - get.c \ - getstring.c \ - hash.c \ - hint.c \ - histogram.c \ - image.c \ - imports.c \ - light.c \ - lines.c \ - matrix.c \ - mipmap.c \ - mm.c \ - occlude.c \ - pixel.c \ - points.c \ - polygon.c \ - rastpos.c \ - rbadaptors.c \ - renderbuffer.c \ - state.c \ - stencil.c \ - texcompress.c \ - texcompress_fxt1.c \ - texcompress_s3tc.c \ - texenvprogram.c \ - texformat.c \ - teximage.c \ - texobj.c \ - texrender.c \ - texstate.c \ - texstore.c \ - varray.c \ - vtxfmt.c - -OBJECTS=accum.obj,\ -api_arrayelt.obj,\ -api_loopback.obj,\ -api_noop.obj,\ -api_validate.obj,\ -arrayobj.obj,\ -attrib.obj,\ -blend.obj,\ -bufferobj.obj,\ -buffers.obj,\ -clip.obj,\ -colortab.obj,\ -context.obj,\ -convolve.obj,\ -debug.obj,\ -depth.obj,\ -depthstencil.obj,\ -dispatch.obj,\ -dlist.obj,\ -drawpix.obj,\ -enable.obj,\ -enums.obj,\ -eval.obj,\ -execmem.obj,\ -extensions.obj,\ -fbobject.obj,\ -feedback.obj,\ -fog.obj,\ -framebuffer.obj,\ -get.obj,\ -getstring.obj,\ -hash.obj,\ -hint.obj,\ -histogram.obj,\ -image.obj,\ -imports.obj,\ -light.obj,\ -lines.obj,\ -matrix.obj,\ -mipmap.obj,\ -mm.obj,\ -occlude.obj,\ -pixel.obj,\ -points.obj,\ -polygon.obj,\ -rastpos.obj,\ -renderbuffer.obj,\ -state.obj,\ -stencil.obj,\ -texcompress.obj,\ -texcompress_fxt1.obj,\ -texcompress_s3tc.obj,\ -texenvprogram.obj,\ -texformat.obj,\ -teximage.obj,\ -texobj.obj,\ -texrender.obj,\ -texstate.obj,\ -texstore.obj,\ -varray.obj,\ -vtxfmt.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ $(MAKELIB) $(GL_LIB) $(OBJECTS) - @ rename $(GL_LIB)* $(LIBDIR) - -clean : - purge - delete *.obj;* - -accum.obj : accum.c -api_arrayelt.obj : api_arrayelt.c -api_loopback.obj : api_loopback.c -api_noop.obj : api_noop.c -api_validate.obj : api_validate.c -arrayobj.obj : arrayobj.c -attrib.obj : attrib.c -blend.obj : blend.c -bufferobj.obj : bufferobj.c -buffers.obj : buffers.c -clip.obj : clip.c -colortab.obj : colortab.c -context.obj : context.c -convolve.obj : convolve.c -debug.obj : debug.c -depth.obj : depth.c -depthstencil.obj : depthstencil.c -dispatch.obj : dispatch.c -dlist.obj : dlist.c -drawpix.obj : drawpix.c -enable.obj : enable.c -enums.obj : enums.c -eval.obj : eval.c -execmem.obj : execmem.c -extensions.obj : extensions.c -fbobject.obj : fbobject.c -feedback.obj : feedback.c -fog.obj : fog.c -framebuffer.obj : framebuffer.c -get.obj : get.c -getstring.obj : getstring.c -hash.obj : hash.c -hint.obj : hint.c -histogram.obj : histogram.c -image.obj : image.c -imports.obj : imports.c vsnprintf.c -light.obj : light.c -lines.obj : lines.c -matrix.obj : matrix.c -mipmap.obj : mipmap.c -mm.obj : mm.c -occlude.obj : occlude.c -pixel.obj : pixel.c -points.obj : points.c -polygon.obj : polygon.c -rastpos.obj : rastpos.c -rbadaptors.obj : rbadaptors.c -renderbuffer.obj : renderbuffer.c -state.obj : state.c -stencil.obj : stencil.c -texcompress.obj : texcompress.c -texcompress_fxt1.obj : texcompress_fxt1.c - cc$(CFLAGS)/warn=(disable=SHIFTCOUNT) texcompress_fxt1.c -texcompress_s3tc.obj : texcompress_s3tc.c -texenvprogram.obj : texenvprogram.c -texformat.obj : texformat.c -teximage.obj : teximage.c -texobj.obj : texobj.c -texrender.obj : texrender.c -texstate.obj : texstate.c -texstore.obj : texstore.c -varray.obj : varray.c -vtxfmt.obj : vtxfmt.c diff --git a/src/mesa/math/descrip.mms b/src/mesa/math/descrip.mms deleted file mode 100644 index 5b9e9915f8..0000000000 --- a/src/mesa/math/descrip.mms +++ /dev/null @@ -1,45 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl -# Last revision : 16 June 2003 - -.first - define gl [---.include.gl] - define math [-.math] - -.include [---]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [---.include],[-.main],[-.glapi] -LIBDIR = [---.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = m_debug_clip.c m_debug_norm.c m_debug_xform.c m_eval.c m_matrix.c\ - m_translate.c m_vector.c m_xform.c - -OBJECTS = m_debug_clip.obj,m_debug_norm.obj,m_debug_xform.obj,m_eval.obj,\ - m_matrix.obj,m_translate.obj,m_vector.obj,m_xform.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -m_debug_clip.obj : m_debug_clip.c -m_debug_norm.obj : m_debug_norm.c -m_debug_xform.obj : m_debug_xform.c -m_eval.obj : m_eval.c -m_matrix.obj : m_matrix.c -m_translate.obj : m_translate.c -m_vector.obj : m_vector.c -m_xform.obj : m_xform.c diff --git a/src/mesa/shader/descrip.mms b/src/mesa/shader/descrip.mms deleted file mode 100644 index f5d04cfa78..0000000000 --- a/src/mesa/shader/descrip.mms +++ /dev/null @@ -1,76 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.nano.tudelft.nl -# Last revision : 20 November 2006 -.first - define gl [---.include.gl] - define math [-.math] - define swrast [-.swrast] - define array_cache [-.array_cache] - -.include [---]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [---.include],[.grammar],[-.main],[-.glapi],[.slang] -LIBDIR = [---.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1,"__extension__=")/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = \ - atifragshader.c \ - arbprogparse.c \ - arbprogram.c \ - nvfragparse.c \ - nvprogram.c \ - nvvertexec.c \ - nvvertparse.c \ - program.c \ - shaderobjects.c \ - shaderobjects_3dlabs.c - -OBJECTS = \ - atifragshader.obj,\ - arbprogparse.obj,\ - arbprogram.obj,\ - nvfragparse.obj,\ - nvprogram.obj,\ - nvvertexec.obj,\ - nvvertparse.obj,\ - program.obj,\ - shaderobjects.obj,\ - shaderobjects_3dlabs.obj - - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -all : - $(MMS)$(MMSQUALIFIERS) $(LIBDIR)$(GL_LIB) - set def [.slang] - $(MMS)$(MMSQUALIFIERS) - set def [-.grammar] - $(MMS)$(MMSQUALIFIERS) - set def [-] - -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -atifragshader.obj : atifragshader.c -arbprogparse.obj : arbprogparse.c -arbprogram.obj : arbprogram.c -nvfragparse.obj : nvfragparse.c -nvprogram.obj : nvprogram.c -nvvertexec.obj : nvvertexec.c -nvvertparse.obj : nvvertparse.c -program.obj : program.c -shaderobjects.obj : shaderobjects.c - cc$(CFLAGS)/nowarn shaderobjects.c -shaderobjects_3dlabs.obj : shaderobjects_3dlabs.c diff --git a/src/mesa/shader/grammar/descrip.mms b/src/mesa/shader/grammar/descrip.mms deleted file mode 100644 index cff53ee872..0000000000 --- a/src/mesa/shader/grammar/descrip.mms +++ /dev/null @@ -1,41 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl -# Last revision : 1 June 2005 - -.first - define gl [----.include.gl] - define math [--.math] - define swrast [--.swrast] - define array_cache [--.array_cache] - -.include [----]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [----.include],[],[--.main],[--.glapi],[-.slang] -LIBDIR = [----.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = grammar_mesa.c - -OBJECTS = grammar_mesa.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -all : - $(MMS)$(MMSQUALIFIERS) $(LIBDIR)$(GL_LIB) - -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -grammar_mesa.obj : grammar_mesa.c grammar.c diff --git a/src/mesa/shader/slang/descrip.mms b/src/mesa/shader/slang/descrip.mms deleted file mode 100644 index 8b9cd822b8..0000000000 --- a/src/mesa/shader/slang/descrip.mms +++ /dev/null @@ -1,65 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.nano.tudelft.nl -# Last revision : 17 March 2006 - -.first - define gl [----.include.gl] - define math [--.math] - define swrast [--.swrast] - define array_cache [--.array_cache] - -.include [----]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [----.include],[--.main],[--.glapi],[-.slang],[-.grammar],[-] -LIBDIR = [----.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = \ - slang_compile.c,slang_preprocess.c - -OBJECTS = \ - slang_compile.obj,slang_preprocess.obj,slang_utility.obj,\ - slang_execute.obj,slang_assemble.obj,slang_assemble_conditional.obj,\ - slang_assemble_constructor.obj,slang_assemble_typeinfo.obj,\ - slang_storage.obj,slang_assemble_assignment.obj,\ - slang_compile_function.obj,slang_compile_struct.obj,\ - slang_compile_variable.obj,slang_compile_operation.obj,\ - slang_library_noise.obj,slang_link.obj,slang_export.obj,\ - slang_analyse.obj,slang_library_texsample.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -slang_compile.obj : slang_compile.c -slang_preprocess.obj : slang_preprocess.c -slang_utility.obj : slang_utility.c -slang_execute.obj : slang_execute.c -slang_assemble.obj : slang_assemble.c -slang_assemble_conditional.obj : slang_assemble_conditional.c -slang_assemble_constructor.obj : slang_assemble_constructor.c -slang_assemble_typeinfo.obj : slang_assemble_typeinfo.c -slang_storage.obj : slang_storage.c -slang_assemble_assignment.obj : slang_assemble_assignment.c -slang_compile_function.obj : slang_compile_function.c -slang_compile_struct.obj : slang_compile_struct.c -slang_compile_variable.obj : slang_compile_variable.c -slang_compile_operation.obj : slang_compile_operation.c -slang_library_noise.obj : slang_library_noise.c -slang_link.obj : slang_link.c -slang_export.obj : slang_export.c -slang_analyse.obj : slang_analyse.c -slang_library_texsample.obj : slang_library_texsample.c diff --git a/src/mesa/swrast/descrip.mms b/src/mesa/swrast/descrip.mms deleted file mode 100644 index 4d446600ba..0000000000 --- a/src/mesa/swrast/descrip.mms +++ /dev/null @@ -1,80 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.nano.tudelft.nl -# Last revision : 21 February 2006 - -.first - define gl [---.include.gl] - define math [-.math] - define swrast [-.swrast] - define array_cache [-.array_cache] - -.include [---]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [---.include],[-.main],[-.glapi],[-.shader],[-.shader.slang] -LIBDIR = [---.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = s_aaline.c s_aatriangle.c s_accum.c s_alpha.c \ - s_bitmap.c s_blend.c s_blit.c s_buffers.c s_context.c \ - s_copypix.c s_depth.c \ - s_drawpix.c s_feedback.c s_fog.c s_imaging.c s_lines.c s_logic.c \ - s_masking.c s_nvfragprog.c s_points.c s_readpix.c \ - s_span.c s_stencil.c s_texstore.c s_texcombine.c s_texfilter.c \ - s_triangle.c s_zoom.c s_atifragshader.c s_arbshader.c - -OBJECTS = s_aaline.obj,s_aatriangle.obj,s_accum.obj,s_alpha.obj,\ - s_bitmap.obj,s_blend.obj,s_blit.obj,s_arbshader.obj,\ - s_buffers.obj,s_context.obj,s_atifragshader.obj,\ - s_copypix.obj,s_depth.obj,s_drawpix.obj,s_feedback.obj,s_fog.obj,\ - s_imaging.obj,s_lines.obj,s_logic.obj,s_masking.obj,s_nvfragprog.obj,\ - s_points.obj,s_readpix.obj,s_span.obj,s_stencil.obj,\ - s_texstore.obj,s_texcombine.obj,s_texfilter.obj,s_triangle.obj,\ - s_zoom.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -s_atifragshader.obj : s_atifragshader.c -s_aaline.obj : s_aaline.c -s_aatriangle.obj : s_aatriangle.c -s_accum.obj : s_accum.c -s_alpha.obj : s_alpha.c -s_bitmap.obj : s_bitmap.c -s_blend.obj : s_blend.c -s_blit.obj : s_blit.c -s_buffers.obj : s_buffers.c -s_context.obj : s_context.c -s_copypix.obj : s_copypix.c -s_depth.obj : s_depth.c -s_drawpix.obj : s_drawpix.c -s_feedback.obj : s_feedback.c -s_fog.obj : s_fog.c -s_imaging.obj : s_imaging.c -s_lines.obj : s_lines.c -s_logic.obj : s_logic.c -s_masking.obj : s_masking.c -s_nvfragprog.obj : s_nvfragprog.c -s_points.obj : s_points.c -s_readpix.obj : s_readpix.c -s_span.obj : s_span.c -s_stencil.obj : s_stencil.c -s_texstore.obj : s_texstore.c -s_texcombine.obj : s_texcombine.c -s_texfilter.obj : s_texfilter.c -s_triangle.obj : s_triangle.c -s_zoom.obj : s_zoom.c -s_arbshader.obj : s_arbshader.c diff --git a/src/mesa/swrast_setup/descrip.mms b/src/mesa/swrast_setup/descrip.mms deleted file mode 100644 index 0ab81c07d3..0000000000 --- a/src/mesa/swrast_setup/descrip.mms +++ /dev/null @@ -1,39 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.stm.tudelft.nl -# Last revision : 16 June 2003 - -.first - define gl [---.include.gl] - define math [-.math] - define tnl [-.tnl] - define swrast [-.swrast] - define array_cache [-.array_cache] - -.include [---]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [---.include],[-.main],[-.glapi] -LIBDIR = [---.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = ss_context.c ss_triangle.c - -OBJECTS = ss_context.obj,ss_triangle.obj -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -ss_context.obj : ss_context.c -ss_triangle.obj : ss_triangle.c diff --git a/src/mesa/tnl/descrip.mms b/src/mesa/tnl/descrip.mms deleted file mode 100644 index a9aed89f1d..0000000000 --- a/src/mesa/tnl/descrip.mms +++ /dev/null @@ -1,75 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.nano.tudelft.nl -# Last revision : 21 February 2006 - -.first - define gl [---.include.gl] - define math [-.math] - define shader [-.shader] - define array_cache [-.array_cache] - -.include [---]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [---.include],[-.main],[-.glapi],[-.shader],[-.shader.slang] -LIBDIR = [---.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES = t_array_api.c t_array_import.c t_context.c \ - t_pipeline.c t_vb_fog.c t_save_api.c t_vtx_api.c \ - t_vb_light.c t_vb_normals.c t_vb_points.c t_vb_program.c \ - t_vb_render.c t_vb_texgen.c t_vb_texmat.c t_vb_vertex.c \ - t_vtx_eval.c t_vtx_exec.c t_save_playback.c t_save_loopback.c \ - t_vertex.c t_vtx_generic.c t_vtx_x86.c t_vertex_generic.c \ - t_vb_arbprogram.c t_vp_build.c t_vb_arbshader.c - -OBJECTS = t_array_api.obj,t_array_import.obj,t_context.obj,\ - t_pipeline.obj,t_vb_fog.obj,t_vb_light.obj,t_vb_normals.obj,\ - t_vb_points.obj,t_vb_program.obj,t_vb_render.obj,t_vb_texgen.obj,\ - t_vb_texmat.obj,t_vb_vertex.obj,t_save_api.obj,t_vtx_api.obj,\ - t_vtx_eval.obj,t_vtx_exec.obj,t_save_playback.obj,t_save_loopback.obj,\ - t_vertex.obj,t_vtx_generic.obj,t_vtx_x86.obj,t_vertex_generic.obj,\ - t_vb_arbprogram.obj,t_vp_build.obj,t_vb_arbshader.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -t_array_api.obj : t_array_api.c -t_array_import.obj : t_array_import.c -t_context.obj : t_context.c -t_pipeline.obj : t_pipeline.c -t_vb_fog.obj : t_vb_fog.c -t_vb_light.obj : t_vb_light.c -t_vb_normals.obj : t_vb_normals.c -t_vb_points.obj : t_vb_points.c -t_vb_program.obj : t_vb_program.c -t_vb_render.obj : t_vb_render.c -t_vb_texgen.obj : t_vb_texgen.c -t_vb_texmat.obj : t_vb_texmat.c -t_vb_vertex.obj : t_vb_vertex.c -t_save_api.obj : t_save_api.c -t_vtx_api.obj : t_vtx_api.c -t_vtx_eval.obj : t_vtx_eval.c -t_vtx_exec.obj : t_vtx_exec.c -t_save_playback.obj : t_save_playback.c -t_save_loopback.obj : t_save_loopback.c -t_vertex.obj : t_vertex.c -t_vtx_x86.obj : t_vtx_x86.c -t_vtx_generic.obj : t_vtx_generic.c -t_vertex_generic.obj : t_vertex_generic.c -t_vb_arbprogram.obj : t_vb_arbprogram.c -t_vp_build.obj : t_vp_build.c -t_vb_arbshader.obj : t_vb_arbshader.c diff --git a/src/mesa/vbo/descrip.mms b/src/mesa/vbo/descrip.mms deleted file mode 100644 index e00b6703aa..0000000000 --- a/src/mesa/vbo/descrip.mms +++ /dev/null @@ -1,60 +0,0 @@ -# Makefile for core library for VMS -# contributed by Jouk Jansen joukj@hrem.nano.tudelft.nl -# Last revision : 7 March 2007 - -.first - define gl [---.include.gl] - define math [-.math] - define vbo [-.vbo] - define tnl [-.tnl] - define shader [-.shader] - define swrast [-.swrast] - define swrast_setup [-.swrast_setup] - -.include [---]mms.config - -##### MACROS ##### - -VPATH = RCS - -INCDIR = [---.include],[-.main],[-.glapi],[-.shader],[-.shader.slang] -LIBDIR = [---.lib] -CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm - -SOURCES =vbo_context.c,vbo_exec.c,vbo_exec_api.c,vbo_exec_array.c,\ - vbo_exec_draw.c,vbo_exec_eval.c,vbo_rebase.c,vbo_save.c,\ - vbo_save_api.c,vbo_save_draw.c,vbo_save_loopback.c,\ - vbo_split.c,vbo_split_copy.c,vbo_split_inplace.c - -OBJECTS =vbo_context.obj,vbo_exec.obj,vbo_exec_api.obj,vbo_exec_array.obj,\ - vbo_exec_draw.obj,vbo_exec_eval.obj,vbo_rebase.obj,vbo_save.obj,\ - vbo_save_api.obj,vbo_save_draw.obj,vbo_save_loopback.obj,\ - vbo_split.obj,vbo_split_copy.obj,vbo_split_inplace.obj - -##### RULES ##### - -VERSION=Mesa V3.4 - -##### TARGETS ##### -# Make the library -$(LIBDIR)$(GL_LIB) : $(OBJECTS) - @ library $(LIBDIR)$(GL_LIB) $(OBJECTS) - -clean : - purge - delete *.obj;* - -vbo_context.obj : vbo_context.c -vbo_exec.obj : vbo_exec.c -vbo_exec_api.obj : vbo_exec_api.c -vbo_exec_array.obj : vbo_exec_array.c -vbo_exec_draw.obj : vbo_exec_draw.c -vbo_exec_eval.obj : vbo_exec_eval.c -vbo_rebase.obj : vbo_rebase.c -vbo_save.obj : vbo_save.c -vbo_save_api.obj : vbo_save_api.c -vbo_save_draw.obj : vbo_save_draw.c -vbo_save_loopback.obj : vbo_save_loopback.c -vbo_split.obj : vbo_split.c -vbo_split_copy.obj : vbo_split_copy.c -vbo_split_inplace.obj : vbo_split_inplace.c -- cgit v1.2.3 From e4e30089231831339815cccebf3a3a0ea6dcd2a9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 25 Feb 2008 20:05:41 +0900 Subject: Make the pipe headers C++ friendly. --- src/gallium/include/pipe/p_context.h | 10 ++++++++++ src/gallium/include/pipe/p_defines.h | 8 ++++++++ src/gallium/include/pipe/p_format.h | 8 ++++++++ src/gallium/include/pipe/p_inlines.h | 9 +++++++++ src/gallium/include/pipe/p_pointer.h | 8 ++++++++ src/gallium/include/pipe/p_state.h | 10 ++++++++++ src/gallium/include/pipe/p_util.h | 16 ++++++++-------- src/gallium/include/pipe/p_winsys.h | 20 ++++++++++++++------ 8 files changed, 75 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 036c4c8964..f69b52f5e3 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -31,6 +31,11 @@ #include "p_state.h" +#ifdef __cplusplus +extern "C" { +#endif + + struct pipe_state_cache; /* Opaque driver handles: @@ -226,4 +231,9 @@ struct pipe_context { unsigned flags ); }; + +#ifdef __cplusplus +} +#endif + #endif /* PIPE_CONTEXT_H */ diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index 0bf53ecb79..d84ddbc27a 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -30,6 +30,10 @@ #include "p_format.h" +#ifdef __cplusplus +extern "C" { +#endif + #define PIPE_BLENDFACTOR_ONE 0x1 #define PIPE_BLENDFACTOR_SRC_COLOR 0x2 #define PIPE_BLENDFACTOR_SRC_ALPHA 0x3 @@ -267,4 +271,8 @@ enum pipe_texture_target { #define PIPE_CAP_MAX_TEXTURE_LOD_BIAS 19 #define PIPE_CAP_BITMAP_TEXCOORD_BIAS 20 +#ifdef __cplusplus +} +#endif + #endif diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index c9ad324315..561d2e5921 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -33,6 +33,10 @@ #include "p_compiler.h" #include "p_debug.h" +#ifdef __cplusplus +extern "C" { +#endif + /** * The PIPE_FORMAT is a 32-bit wide bitfield that encodes all the information * needed to uniquely describe a pixel format. @@ -418,4 +422,8 @@ static INLINE uint pf_get_size( enum pipe_format format ) { return pf_get_bits(format) / 8; } +#ifdef __cplusplus +} +#endif + #endif diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h index ebf6ed86bc..de3fa555c5 100644 --- a/src/gallium/include/pipe/p_inlines.h +++ b/src/gallium/include/pipe/p_inlines.h @@ -33,6 +33,11 @@ #include "p_winsys.h" +#ifdef __cplusplus +extern "C" { +#endif + + static INLINE void * pipe_surface_map(struct pipe_surface *surface) { @@ -109,4 +114,8 @@ pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, } +#ifdef __cplusplus +} +#endif + #endif /* P_INLINES_H */ diff --git a/src/gallium/include/pipe/p_pointer.h b/src/gallium/include/pipe/p_pointer.h index e3927538e8..3a1e6be88e 100644 --- a/src/gallium/include/pipe/p_pointer.h +++ b/src/gallium/include/pipe/p_pointer.h @@ -30,6 +30,10 @@ #include "p_compiler.h" +#ifdef __cplusplus +extern "C" { +#endif + static INLINE intptr_t pointer_to_intptr( const void *p ) { @@ -84,4 +88,8 @@ align_pointer( const void *unaligned, uintptr_t alignment ) return uintptr_to_pointer( aligned ); } +#ifdef __cplusplus +} +#endif + #endif /* P_POINTER_H */ diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 47fa78c31d..15c88881eb 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -42,6 +42,12 @@ #include "p_defines.h" #include "p_format.h" + +#ifdef __cplusplus +extern "C" { +#endif + + /** * Implementation limits */ @@ -326,4 +332,8 @@ struct pipe_vertex_element }; +#ifdef __cplusplus +} +#endif + #endif diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 6f0dbdacd9..3b32ba1d24 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -34,13 +34,13 @@ #include -#ifdef WIN32 - #ifdef __cplusplus -extern "C" -{ +extern "C" { #endif + +#ifdef WIN32 + void * __stdcall EngAllocMem( unsigned long Flags, @@ -51,10 +51,6 @@ void __stdcall EngFreeMem( void *Mem ); -#ifdef __cplusplus -} -#endif - static INLINE void * MALLOC( unsigned size ) { @@ -379,4 +375,8 @@ extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, int src_pitch, unsigned src_x, int src_y); +#ifdef __cplusplus +} +#endif + #endif diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h index 1e81eebd78..e784c92491 100644 --- a/src/gallium/include/pipe/p_winsys.h +++ b/src/gallium/include/pipe/p_winsys.h @@ -25,12 +25,6 @@ * **************************************************************************/ -#ifndef P_WINSYS_H -#define P_WINSYS_H - - -#include "p_format.h" - /** * \file * This is the interface that Gallium3D requires any window system @@ -38,6 +32,17 @@ * which is public. */ +#ifndef P_WINSYS_H +#define P_WINSYS_H + + +#include "p_format.h" + + +#ifdef __cplusplus +extern "C" { +#endif + /** Opaque type */ struct pipe_fence_handle; @@ -156,5 +161,8 @@ struct pipe_winsys }; +#ifdef __cplusplus +} +#endif #endif /* P_WINSYS_H */ -- cgit v1.2.3 From 20839b37ed61b044d6224c0e373ce10d74be4f3d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 11:13:58 -0700 Subject: gallium/i915: added SGT/SLE opcodes --- .../drivers/i915simple/i915_fpc_translate.c | 31 ++++++++++++++++++++++ 1 file changed, 31 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index afe7e25dce..50c9e65409 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -385,6 +385,26 @@ emit_simple_arith(struct i915_fp_compile *p, arg3 ); } + +/** As above, but swap the first two src regs */ +static void +emit_simple_arith_swap2(struct i915_fp_compile *p, + const struct tgsi_full_instruction *inst, + uint opcode, uint numArgs) +{ + struct tgsi_full_instruction inst2; + + assert(numArgs == 2); + + /* transpose first two registers */ + inst2 = *inst; + inst2.FullSrcRegisters[0] = inst->FullSrcRegisters[1]; + inst2.FullSrcRegisters[1] = inst->FullSrcRegisters[0]; + + emit_simple_arith(p, &inst2, opcode, numArgs); +} + + #ifndef M_PI #define M_PI 3.14159265358979323846 #endif @@ -772,6 +792,11 @@ i915_translate_instruction(struct i915_fp_compile *p, emit_simple_arith(p, inst, A0_SGE, 2); break; + case TGSI_OPCODE_SLE: + /* like SGE, but swap reg0, reg1 */ + emit_simple_arith_swap2(p, inst, A0_SGE, 2); + break; + case TGSI_OPCODE_SIN: src0 = src_vector(p, &inst->FullSrcRegisters[0]); tmp = i915_get_utemp(p); @@ -826,6 +851,11 @@ i915_translate_instruction(struct i915_fp_compile *p, emit_simple_arith(p, inst, A0_SLT, 2); break; + case TGSI_OPCODE_SGT: + /* like SLT, but swap reg0, reg1 */ + emit_simple_arith_swap2(p, inst, A0_SLT, 2); + break; + case TGSI_OPCODE_SUB: src0 = src_vector(p, &inst->FullSrcRegisters[0]); src1 = src_vector(p, &inst->FullSrcRegisters[1]); @@ -879,6 +909,7 @@ i915_translate_instruction(struct i915_fp_compile *p, default: i915_program_error(p, "bad opcode %d", inst->Instruction.Opcode); + p->error = 1; return; } -- cgit v1.2.3 From ea02342c11eaeb700495b403caecc13a129333e8 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 11:21:03 -0700 Subject: gallium/i915: Use hardware rendering, unless INTEL_SP env var is set --- src/gallium/winsys/dri/intel/intel_context.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index c033f2a592..79b320c6bf 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -188,7 +188,8 @@ intelCreateContext(const __GLcontextModes * visual, /* * Pipe-related setup */ - if (!getenv("INTEL_HW")) { + if (getenv("INTEL_SP")) { + /* use softpipe driver instead of hw */ pipe = intel_create_softpipe( intel, intelScreen->winsys ); } else { -- cgit v1.2.3 From d6c7f7e314ee9f034402c919d142bf6ba9844ec9 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 14:46:42 -0700 Subject: gallium: modify draw_find_vs_output() to search vertex shader outputs This simplifies drivers using the draw module and removes the last dependency on vertex-shader "internals". Since the draw module is producing the post-transformed vertices, it makes sense to ask it where specific vertex attributes are located. This could also simplify some things in the state tracker code for selection, feedback, rasterpos... --- src/gallium/auxiliary/draw/draw_context.c | 24 +++++++++++++++++++++--- 1 file changed, 21 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index c28e78d33a..6e5e4d64d1 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -244,14 +244,32 @@ draw_convert_wide_lines(struct draw_context *draw, boolean enable) /** - * The draw module may sometimes generate vertices with extra attributes - * (such as texcoords for AA lines). The driver can call this function - * to find those attributes. + * Ask the draw module for the location/slot of the given vertex attribute in + * a post-transformed vertex. + * + * With this function, drivers that use the draw module should have no reason + * to track the current vertex shader. + * + * Note that the draw module may sometimes generate vertices with extra + * attributes (such as texcoords for AA lines). The driver can call this + * function to find those attributes. + * + * Zero is returned if the attribute is not found since this is + * a don't care / undefined situtation. Returning -1 would be a bit more + * work for the drivers. */ int draw_find_vs_output(struct draw_context *draw, uint semantic_name, uint semantic_index) { + const struct pipe_shader_state *vs = &draw->vertex_shader->state; + uint i; + for (i = 0; i < vs->num_outputs; i++) { + if (vs->output_semantic_name[i] == semantic_name && + vs->output_semantic_index[i] == semantic_index) + return i; + } + /* XXX there may be more than one extra vertex attrib. * For example, simulated gl_FragCoord and gl_PointCoord. */ -- cgit v1.2.3 From c037b4d45a551dc7b7dd33950c2e8df60449061c Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 14:47:13 -0700 Subject: softpipe: use draw_find_vs_output() directly --- src/gallium/drivers/softpipe/sp_state_derived.c | 39 +++++-------------------- 1 file changed, 7 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index f9f2c5eaa8..4c6313001f 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -34,33 +34,6 @@ #include "sp_state.h" -/** - * Search vertex program's outputs to find a match for the given - * semantic name/index. Return the index of the output slot. - * - * Return 0 if not found. This will cause the fragment program to use - * vertex attrib 0 (position) in the cases where the fragment program - * attempts to use a missing vertex program output. This is an undefined - * condition that users shouldn't hit anyway. - */ -static int -find_vs_output(struct softpipe_context *sp, - const struct pipe_shader_state *vs, - uint semantic_name, - uint semantic_index) -{ - uint i; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == semantic_name && - vs->output_semantic_index[i] == semantic_index) - return i; - } - - /* See if the draw module is introducing a new attribute... */ - return draw_find_vs_output(sp->draw, semantic_name, semantic_index); -} - - /** * Mark the current vertex layout as "invalid". * We'll validate the vertex layout later, when we start to actually @@ -114,24 +87,25 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) int src; switch (fs->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: - src = find_vs_output(softpipe, vs, TGSI_SEMANTIC_POSITION, 0); + src = draw_find_vs_output(softpipe->draw, + TGSI_SEMANTIC_POSITION, 0); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_POS, src); break; case TGSI_SEMANTIC_COLOR: - src = find_vs_output(softpipe, vs, TGSI_SEMANTIC_COLOR, + src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_COLOR, fs->input_semantic_index[i]); draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); break; case TGSI_SEMANTIC_FOG: - src = find_vs_output(softpipe, vs, TGSI_SEMANTIC_FOG, 0); + src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_FOG, 0); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ - src = find_vs_output(softpipe, vs, TGSI_SEMANTIC_GENERIC, + src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_GENERIC, fs->input_semantic_index[i]); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; @@ -141,7 +115,8 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) } } - softpipe->psize_slot = find_vs_output(softpipe, vs, TGSI_SEMANTIC_PSIZE, 0); + softpipe->psize_slot = draw_find_vs_output(softpipe->draw, + TGSI_SEMANTIC_PSIZE, 0); if (softpipe->psize_slot > 0) { draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_CONSTANT, softpipe->psize_slot); -- cgit v1.2.3 From 846b7fbc6c9cbd57eed01bd04b1da73109935091 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 14:48:31 -0700 Subject: gallium/i915: use draw_find_vs_output() directly, fix broken fogcoords. We now produce the correct 915 vertex layout regardless of the order in which fragment shader inputs are declared. --- .../drivers/i915simple/i915_state_derived.c | 99 ++++++++++++---------- 1 file changed, 53 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state_derived.c b/src/gallium/drivers/i915simple/i915_state_derived.c index f654f543cc..fe52fe6e6b 100644 --- a/src/gallium/drivers/i915simple/i915_state_derived.c +++ b/src/gallium/drivers/i915simple/i915_state_derived.c @@ -27,104 +27,111 @@ #include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" #include "draw/draw_vertex.h" #include "i915_context.h" #include "i915_state.h" #include "i915_reg.h" #include "i915_fpc.h" -#include "pipe/p_shader_tokens.h" + /** - * Determine which post-transform / pre-rasterization vertex attributes - * we need. - * Derived from: fs, setup states. + * Determine the hardware vertex layout. + * Depends on vertex/fragment shader state. */ static void calculate_vertex_layout( struct i915_context *i915 ) { const struct pipe_shader_state *fs = &i915->fs->state; const enum interp_mode colorInterp = i915->rasterizer->color_interp; struct vertex_info vinfo; - uint front0 = 0, back0 = 0, front1 = 0, back1 = 0; - boolean needW = 0; + boolean texCoords[8], colors[2], fog, needW; uint i; - boolean texCoords[8]; - uint src = 0; + int src; memset(texCoords, 0, sizeof(texCoords)); + colors[0] = colors[1] = fog = needW = FALSE; memset(&vinfo, 0, sizeof(vinfo)); - /* pos */ - draw_emit_vertex_attr(&vinfo, EMIT_3F, INTERP_LINEAR, src++); - /* Note: we'll set the S4_VFMT_XYZ[W] bits below */ - + /* Determine which fragment program inputs are needed. Setup HW vertex + * layout below, in the HW-specific attribute order. + */ for (i = 0; i < fs->num_inputs; i++) { switch (fs->input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: break; case TGSI_SEMANTIC_COLOR: - if (fs->input_semantic_index[i] == 0) { - front0 = draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src++); - vinfo.hwfmt[0] |= S4_VFMT_COLOR; - } - else { - assert(fs->input_semantic_index[i] == 1); - front1 = draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src++); - vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG; - } + assert(fs->input_semantic_index[i] < 2); + colors[fs->input_semantic_index[i]] = TRUE; break; case TGSI_SEMANTIC_GENERIC: /* usually a texcoord */ { const uint unit = fs->input_semantic_index[i]; - uint hwtc; + assert(unit < 8); texCoords[unit] = TRUE; - draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++); - hwtc = TEXCOORDFMT_4D; needW = TRUE; - vinfo.hwfmt[1] |= hwtc << (unit * 4); } break; case TGSI_SEMANTIC_FOG: - debug_printf("i915 fogcoord not implemented yet\n"); - draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_PERSPECTIVE, src++); + fog = TRUE; break; default: assert(0); } - } - /* finish up texcoord fields */ - for (i = 0; i < 8; i++) { - if (!texCoords[i]) { - const uint hwtc = TEXCOORDFMT_NOT_PRESENT; - vinfo.hwfmt[1] |= hwtc << (i* 4); - } - } - - /* go back and fill in the vertex position info now that we have needW */ + + /* pos */ + src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_POSITION, 0); if (needW) { + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src); vinfo.hwfmt[0] |= S4_VFMT_XYZW; vinfo.emit[0] = EMIT_4F; } else { + draw_emit_vertex_attr(&vinfo, EMIT_3F, INTERP_LINEAR, src); vinfo.hwfmt[0] |= S4_VFMT_XYZ; vinfo.emit[0] = EMIT_3F; } - /* Additional attributes required for setup: Just twosided - * lighting. Edgeflag is dealt with specially by setting bits in - * the vertex header. - */ - if (i915->rasterizer->light_twoside) { - if (front0) { - back0 = draw_emit_vertex_attr(&vinfo, EMIT_OMIT, colorInterp, src++); + /* hardware point size */ + /* XXX todo */ + + /* primary color */ + if (colors[0]) { + src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_COLOR, 0); + draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src); + vinfo.hwfmt[0] |= S4_VFMT_COLOR; + } + + /* secondary color */ + if (colors[1]) { + src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_COLOR, 1); + draw_emit_vertex_attr(&vinfo, EMIT_4UB, colorInterp, src); + vinfo.hwfmt[0] |= S4_VFMT_SPEC_FOG; + } + + /* fog coord, not fog blend factor */ + if (fog) { + src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_FOG, 0); + draw_emit_vertex_attr(&vinfo, EMIT_1F, INTERP_PERSPECTIVE, src); + vinfo.hwfmt[0] |= S4_VFMT_FOG_PARAM; + } + + /* texcoords */ + for (i = 0; i < 8; i++) { + uint hwtc; + if (texCoords[i]) { + hwtc = TEXCOORDFMT_4D; + src = draw_find_vs_output(i915->draw, TGSI_SEMANTIC_GENERIC, i); + draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); } - if (back0) { - back1 = draw_emit_vertex_attr(&vinfo, EMIT_OMIT, colorInterp, src++); + else { + hwtc = TEXCOORDFMT_NOT_PRESENT; } + vinfo.hwfmt[1] |= hwtc << (i * 4); } draw_compute_vertex_size(&vinfo); -- cgit v1.2.3 From cc0cf1154b6288d49d7a9dfd18fb666331dd7334 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 15:34:46 -0700 Subject: gallium: fix bad ptr assignment --- src/gallium/auxiliary/draw/draw_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 6e5e4d64d1..7dd1c6f6fa 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -262,7 +262,7 @@ int draw_find_vs_output(struct draw_context *draw, uint semantic_name, uint semantic_index) { - const struct pipe_shader_state *vs = &draw->vertex_shader->state; + const struct pipe_shader_state *vs = draw->vertex_shader->state; uint i; for (i = 0; i < vs->num_outputs; i++) { if (vs->output_semantic_name[i] == semantic_name && -- cgit v1.2.3 From b53110c78941e7fcaa41921cce07ca00ec117a97 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 16:16:07 -0700 Subject: gallium/i915: call draw_flush() in i915_flush() --- src/gallium/drivers/i915simple/i915_flush.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c index 3c2069b827..96a54281f1 100644 --- a/src/gallium/drivers/i915simple/i915_flush.c +++ b/src/gallium/drivers/i915simple/i915_flush.c @@ -31,6 +31,7 @@ #include "pipe/p_defines.h" +#include "draw/draw_context.h" #include "i915_context.h" #include "i915_reg.h" #include "i915_batch.h" @@ -44,6 +45,8 @@ static void i915_flush( struct pipe_context *pipe, { struct i915_context *i915 = i915_context(pipe); + draw_flush(i915->draw); + /* Do we need to emit an MI_FLUSH command to flush the hardware * caches? */ -- cgit v1.2.3 From f41e95755757cb1452697fafa1dd5288390ed57e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 16:20:04 -0700 Subject: gallium/i915: need to recompute vertex info if vertex shader changes --- src/gallium/drivers/i915simple/i915_context.h | 1 + src/gallium/drivers/i915simple/i915_state.c | 2 ++ src/gallium/drivers/i915simple/i915_state_derived.c | 2 +- 3 files changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 5cc38cdc85..d32dded6bd 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -263,6 +263,7 @@ struct i915_context #define I915_NEW_TEXTURE 0x800 #define I915_NEW_CONSTANTS 0x1000 #define I915_NEW_VBO 0x2000 +#define I915_NEW_VS 0x4000 /* Driver's internally generated state flags: diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index e4288d4e31..a35bdf941f 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -474,6 +474,8 @@ static void i915_bind_vs_state(struct pipe_context *pipe, void *shader) /* just pass-through to draw module */ draw_bind_vertex_shader(i915->draw, (struct draw_vertex_shader *) shader); + + i915->dirty |= I915_NEW_VS; } static void i915_delete_vs_state(struct pipe_context *pipe, void *shader) diff --git a/src/gallium/drivers/i915simple/i915_state_derived.c b/src/gallium/drivers/i915simple/i915_state_derived.c index fe52fe6e6b..5cf70acdf3 100644 --- a/src/gallium/drivers/i915simple/i915_state_derived.c +++ b/src/gallium/drivers/i915simple/i915_state_derived.c @@ -155,7 +155,7 @@ static void calculate_vertex_layout( struct i915_context *i915 ) */ void i915_update_derived( struct i915_context *i915 ) { - if (i915->dirty & (I915_NEW_RASTERIZER | I915_NEW_FS)) + if (i915->dirty & (I915_NEW_RASTERIZER | I915_NEW_FS | I915_NEW_VS)) calculate_vertex_layout( i915 ); if (i915->dirty & (I915_NEW_SAMPLER | I915_NEW_TEXTURE)) -- cgit v1.2.3 From 92650aeaddd1bd729f3a90383f05c8148f678066 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 16:22:58 -0700 Subject: gallium/i915: make sure state is up to date in i915_vbuf_render_get_vertex_info(), also disable bogus assertion --- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index c5bf6174f6..9d5f609220 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -83,6 +83,12 @@ i915_vbuf_render_get_vertex_info( struct vbuf_render *render ) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; + + if (i915->dirty) { + /* make sure we have up to date vertex layout */ + i915_update_derived( i915 ); + } + return &i915->current.vertex_info; } @@ -143,7 +149,8 @@ i915_vbuf_render_draw( struct vbuf_render *render, assert(nr_indices); - assert((i915->dirty & ~I915_NEW_VBO) == 0); + /* this seems to be bogus, since we validate state right after this */ + /*assert((i915->dirty & ~I915_NEW_VBO) == 0);*/ if (i915->dirty) i915_update_derived( i915 ); -- cgit v1.2.3 From 99047e0968882a4e3f9577fa2352d91733181339 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 16:24:47 -0700 Subject: gallium/i915: plug in aaline draw stage --- src/gallium/drivers/i915simple/i915_context.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index acfa349439..36f6429d46 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -298,10 +298,14 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915_init_string_functions(i915); i915_init_texture_functions(i915); + /* not working yet: + draw_install_aapoint_stage(i915->draw, &i915->pipe); + */ + draw_install_aaline_stage(i915->draw, &i915->pipe); + i915->pci_id = pci_id; i915->flags.is_i945 = is_i945; - i915->dirty = ~0; i915->hardware_dirty = ~0; -- cgit v1.2.3 From 09ba1dd4ccb0ed907510cb9403b1fb1fb0ab3658 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 16:25:24 -0700 Subject: gallium: clamp line width when creating raster state object --- src/mesa/state_tracker/st_atom_rasterizer.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 435d604af7..229839d8b2 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -30,7 +30,7 @@ * Keith Whitwell */ - +#include "main/macros.h" #include "st_context.h" #include "st_cache.h" #include "pipe/p_context.h" @@ -227,8 +227,18 @@ static void update_raster_state( struct st_context *st ) /* _NEW_LINE */ - raster.line_width = ctx->Line.Width; raster.line_smooth = ctx->Line.SmoothFlag; + if (ctx->Line.SmoothFlag) { + raster.line_width = CLAMP(ctx->Line.Width, + ctx->Const.MinLineWidthAA, + ctx->Const.MaxLineWidthAA); + } + else { + raster.line_width = CLAMP(ctx->Line.Width, + ctx->Const.MinLineWidth, + ctx->Const.MaxLineWidth); + } + raster.line_stipple_enable = ctx->Line.StippleFlag; raster.line_stipple_pattern = ctx->Line.StipplePattern; /* GL stipple factor is in [1,256], remap to [0, 255] here */ -- cgit v1.2.3 From b02fc948348db5559d658251dd3a6d4f3390d686 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 17:01:20 -0700 Subject: gallium/i915: compute vertex size _after_ state validation in emit_prim(). Fixes crash when drawing aa lines. --- src/gallium/drivers/i915simple/i915_prim_emit.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c index 44c4325936..6da42b3846 100644 --- a/src/gallium/drivers/i915simple/i915_prim_emit.c +++ b/src/gallium/drivers/i915simple/i915_prim_emit.c @@ -72,6 +72,8 @@ emit_hw_vertex( struct i915_context *i915, uint i; uint count = 0; /* for debug/sanity */ + assert(!i915->dirty); + for (i = 0; i < vinfo->num_attribs; i++) { switch (vinfo->emit[i]) { case EMIT_OMIT: @@ -122,17 +124,19 @@ emit_prim( struct draw_stage *stage, unsigned nr ) { struct i915_context *i915 = setup_stage(stage)->i915; - unsigned vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ + unsigned vertex_size; unsigned i; - assert(vertex_size >= 12); /* never smaller than 12 bytes */ - if (i915->dirty) i915_update_derived( i915 ); if (i915->hardware_dirty) i915_emit_hardware_state( i915 ); + /* need to do this after validation! */ + vertex_size = i915->current.vertex_info.size * 4; /* in bytes */ + assert(vertex_size >= 12); /* never smaller than 12 bytes */ + if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) { FLUSH_BATCH(); -- cgit v1.2.3 From 7976a084e792daf0b23c688bfa8f577de141ecca Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 21 Feb 2008 11:01:35 -0800 Subject: Cell: Use multiple DMA tags for the dcache. --- src/gallium/drivers/cell/spu/spu_dcache.c | 2 +- src/gallium/drivers/cell/spu/spu_main.h | 5 ++++- 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_dcache.c b/src/gallium/drivers/cell/spu/spu_dcache.c index 698a5790bb..3baeaea998 100644 --- a/src/gallium/drivers/cell/spu/spu_dcache.c +++ b/src/gallium/drivers/cell/spu/spu_dcache.c @@ -33,7 +33,7 @@ #define CACHE_NAME data #define CACHED_TYPE qword #define CACHE_TYPE CACHE_TYPE_RO -#define CACHE_SET_TAGID(set) TAG_VERTEX_BUFFER +#define CACHE_SET_TAGID(set) (((set) & 0x03) + TAG_DCACHE0) #define CACHE_LOG2NNWAY 2 #define CACHE_LOG2NSETS 6 #include diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index 5c95d112ac..d14f1abbe7 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -131,7 +131,10 @@ extern boolean Debug; #define TAG_BATCH_BUFFER 17 #define TAG_MISC 18 #define TAG_TEXTURE_TILE 19 -#define TAG_INSTRUCTION_FETCH 20 +#define TAG_DCACHE0 20 +#define TAG_DCACHE1 21 +#define TAG_DCACHE2 22 +#define TAG_DCACHE3 23 -- cgit v1.2.3 From fb68daceec312a90910dad3882e6ef57c370b7fd Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 21 Feb 2008 16:41:12 -0800 Subject: Cell: Remove unnecessary include files --- src/gallium/drivers/cell/spu/spu_exec.c | 2 -- src/gallium/drivers/cell/spu/spu_vertex_fetch.c | 2 -- 2 files changed, 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index cf81bee8fd..fff0114a23 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -50,8 +50,6 @@ * Brian Paul */ -#include -#include #include #include #include diff --git a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c index f7e4e653e3..219fd90cc0 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c @@ -32,8 +32,6 @@ * Ian Romanick */ -#include - #include "pipe/p_util.h" #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" -- cgit v1.2.3 From a63fd641a01a50e1be51664bf863e01ddaf61d3e Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 22 Feb 2008 16:27:39 -0800 Subject: cell: Trivial compiler warning clean-ups. --- src/gallium/drivers/cell/ppu/cell_draw_arrays.c | 1 + src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 1 - src/gallium/drivers/cell/spu/spu_exec.c | 6 ++++-- src/gallium/drivers/cell/spu/spu_main.c | 5 +++-- 4 files changed, 8 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c index cbd387f014..c839fb4d12 100644 --- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c @@ -38,6 +38,7 @@ #include "cell_context.h" #include "cell_draw_arrays.h" #include "cell_state.h" +#include "cell_flush.h" #include "draw/draw_context.h" diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index 42cc47cbfe..17141924be 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -55,7 +55,6 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) struct cell_command_vs *const vs = &cell_global.command[0].vs; uint64_t *batch; struct cell_array_info *array_info; - struct cell_shader_info *shader_info; unsigned i, j; struct cell_attribute_fetch_code *cf; diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index fff0114a23..1560c0f157 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -149,6 +149,7 @@ spu_exec_machine_init(struct spu_exec_machine *mach, const qword zero = si_il(0); const qword not_zero = si_il(~0); + (void) numSamplers; mach->Samplers = samplers; mach->Processor = processor; mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; @@ -657,9 +658,10 @@ fetch_texel( struct spu_sampler *sampler, qword rgba[4]; qword out[4]; - sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, (float *) rgba); + sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, + (float (*)[4]) rgba); - _transpose_matrix4x4(out, rgba); + _transpose_matrix4x4((vec_float4 *) out, (vec_float4 *) rgba); r->q = out[0]; g->q = out[1]; b->q = out[2]; diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 1136dba62d..cc4bafdb3a 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -38,6 +38,7 @@ #include "spu_tile.h" //#include "spu_test.h" #include "spu_vertex_shader.h" +#include "spu_dcache.h" #include "cell/common.h" #include "pipe/p_defines.h" @@ -434,7 +435,7 @@ cmd_batch(uint opcode) pos += (1 + ROUNDUP8(sizeof(struct pipe_viewport_state)) / 8); break; case CELL_CMD_STATE_UNIFORMS: - draw.constants = (float (*)[4]) (uintptr_t) buffer[pos + 1]; + draw.constants = (const float (*)[4]) (uintptr_t) buffer[pos + 1]; pos += 2; break; case CELL_CMD_STATE_VS_ARRAY_INFO: @@ -583,7 +584,7 @@ main(main_param_t speid, main_param_t argp) one_time_init(); if (Debug) - printf("SPU: main() speid=%lu\n", speid); + printf("SPU: main() speid=%lu\n", (unsigned long) speid); mfc_get(&spu.init, /* dest */ (unsigned int) argp, /* src */ -- cgit v1.2.3 From 2efa7e9489541b6a86c3d46e3d58cbf5bf399189 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 22 Feb 2008 17:51:55 -0800 Subject: cell: Fix off-by-one error in spu_dcache_fetch_unaligned This time the off-by-one error caused an extra qword to be fetched under certain circumstances when the source ea was not qword aligned. --- src/gallium/drivers/cell/spu/spu_dcache.c | 50 ++++++++++++++++++++----------- 1 file changed, 32 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_dcache.c b/src/gallium/drivers/cell/spu/spu_dcache.c index 3baeaea998..a1701d80d1 100644 --- a/src/gallium/drivers/cell/spu/spu_dcache.c +++ b/src/gallium/drivers/cell/spu/spu_dcache.c @@ -49,43 +49,57 @@ /** * Fetch between arbitrary number of bytes from an unaligned address + * + * \param dst Destination data buffer + * \param ea Main memory effective address of source data + * \param size Number of bytes to read + * + * \warning + * As is hinted by the type of the \c dst pointer, this function writes + * multiples of 16-bytes. */ void spu_dcache_fetch_unaligned(qword *dst, unsigned ea, unsigned size) { const int shift = ea & 0x0f; - const unsigned aligned_start_ea = ea & ~0x0f; - const unsigned aligned_end_ea = ROUNDUP16(ea + size); - const unsigned num_entries = (aligned_end_ea - aligned_start_ea) / 16; + const unsigned read_size = ROUNDUP16(size + shift); + const unsigned last_read = ROUNDUP16(ea + size); + const qword *const last_write = dst + (ROUNDUP16(size) / 16); unsigned i; if (shift == 0) { /* Data is already aligned. Fetch directly into the destination buffer. */ - for (i = 0; i < num_entries; i++) { - dst[i] = cache_rd(data, ea + (i * 16)); + for (i = 0; i < size; i += 16) { + *(dst++) = cache_rd(data, ea + i); } } else { - qword tmp[2] ALIGN16_ATTRIB; - + qword hi; - tmp[0] = cache_rd(data, (ea & ~0x0f)); - for (i = 0; i < (num_entries & ~1); i++) { - const unsigned curr = i & 1; - const unsigned next = curr ^ 1; - tmp[next] = cache_rd(data, (ea & ~0x0f) + (next * 16)); - - dst[i] = si_or((qword) spu_slqwbyte(tmp[curr], shift), - (qword) spu_rlmaskqwbyte(tmp[next], shift - 16)); + /* Please exercise extreme caution when modifying this code. This code + * must not read past the end of the page containing the source data, + * and it must not write more than ((size + 15) / 16) qwords to the + * destination buffer. + */ + ea &= ~0x0f; + hi = cache_rd(data, ea); + for (i = 16; i < read_size; i += 16) { + qword lo = cache_rd(data, ea + i); + + *(dst++) = si_or((qword) spu_slqwbyte(hi, shift), + (qword) spu_rlmaskqwbyte(lo, shift - 16)); + hi = lo; } - if (i < num_entries) { - dst[i] = si_or((qword) spu_slqwbyte(tmp[(i & 1)], shift), - si_il(0)); + if (dst != last_write) { + *(dst++) = si_or((qword) spu_slqwbyte(hi, shift), si_il(0)); } } + + ASSERT((ea + i) == last_read); + ASSERT(dst == last_write); } -- cgit v1.2.3 From 2fc9d0ffac4b44dad1f13443e5dedd545675a7ee Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 25 Feb 2008 16:15:59 -0800 Subject: cell: Additional changes to match changes in draw/draw_vertex_shader.c --- src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index 17141924be..f5c27852c1 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -127,7 +127,7 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) for (/* empty */; j < SPU_VERTS_PER_BATCH; j++) { vs->elts[j] = vs->elts[0]; - vs->vOut[j] = vs->vOut[0]; + vs->vOut[j] = (uintptr_t) draw->vs.queue[i + j].vertex; } vs->num_elts = n; @@ -136,5 +136,6 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) cell_flush_int(& cell->pipe, PIPE_FLUSH_WAIT); } + draw->vs.post_nr = draw->vs.queue_nr; draw->vs.queue_nr = 0; } -- cgit v1.2.3 From 0235b3252100eda553babea42c014445358a2985 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 17:59:51 -0700 Subject: gallium/i915: fix i915_emit_texld() to handle swizzled texcoords Allocate a temporary register, insert MOV instruction, etc. --- src/gallium/drivers/i915simple/i915_fpc_emit.c | 63 +++++++++++++++++----- .../drivers/i915simple/i915_fpc_translate.c | 20 +++++-- 2 files changed, 67 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_fpc_emit.c b/src/gallium/drivers/i915simple/i915_fpc_emit.c index a59ee23403..ad6dab6c11 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_emit.c +++ b/src/gallium/drivers/i915simple/i915_fpc_emit.c @@ -71,10 +71,21 @@ i915_get_temp(struct i915_fp_compile *p) } p->temp_flag |= 1 << (bit - 1); - return UREG(REG_TYPE_R, (bit - 1)); + return bit - 1; } +static void +i915_release_temp(struct i915_fp_compile *p, int reg) +{ + p->temp_flag &= ~(1 << reg); +} + + +/** + * Get unpreserved temporary, a temp whose value is not preserved between + * PS program phases. + */ uint i915_get_utemp(struct i915_fp_compile * p) { @@ -183,41 +194,63 @@ i915_emit_arith(struct i915_fp_compile * p, return dest; } + +/** + * Emit a texture load or texkill instruction. + * \param dest the dest i915 register + * \param destmask the dest register writemask + * \param sampler the i915 sampler register + * \param coord the i915 source texcoord operand + * \param opcode the instruction opcode + */ uint i915_emit_texld( struct i915_fp_compile *p, uint dest, uint destmask, uint sampler, uint coord, - uint op ) + uint opcode ) { - uint k = UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord)); + const uint k = UREG(GET_UREG_TYPE(coord), GET_UREG_NR(coord)); + int temp = -1; + if (coord != k) { - /* No real way to work around this in the general case - need to - * allocate and declare a new temporary register (a utemp won't - * do). Will fallback for now. + /* texcoord is swizzled or negated. Need to allocate a new temporary + * register (a utemp / unpreserved temp) won't do. */ - i915_program_error(p, "Can't (yet) swizzle TEX arguments"); - assert(0); - return 0; + uint tempReg; + + temp = i915_get_temp(p); /* get temp reg index */ + printf("***** ALLOC TEMP %d\n", temp); + tempReg = UREG(REG_TYPE_R, temp); /* make i915 register */ + + i915_emit_arith( p, A0_MOV, + tempReg, A0_DEST_CHANNEL_ALL, /* dest reg, writemask */ + 0, /* saturate */ + coord, 0, 0 ); /* src0, src1, src2 */ + + /* new src texcoord is tempReg */ + coord = tempReg; } /* Don't worry about saturate as we only support */ if (destmask != A0_DEST_CHANNEL_ALL) { + /* if not writing to XYZW... */ uint tmp = i915_get_utemp(p); - i915_emit_texld( p, tmp, A0_DEST_CHANNEL_ALL, sampler, coord, op ); + i915_emit_texld( p, tmp, A0_DEST_CHANNEL_ALL, sampler, coord, opcode ); i915_emit_arith( p, A0_MOV, dest, destmask, 0, tmp, 0, 0 ); - return dest; + /* XXX release utemp here? */ } else { assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST); assert(dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest))); + /* is the sampler coord a texcoord input reg? */ if (GET_UREG_TYPE(coord) != REG_TYPE_T) { p->nr_tex_indirect++; } - *(p->csr++) = (op | + *(p->csr++) = (opcode | T0_DEST( dest ) | T0_SAMPLER( sampler )); @@ -225,8 +258,12 @@ uint i915_emit_texld( struct i915_fp_compile *p, *(p->csr++) = T2_MBZ; p->nr_tex_insn++; - return dest; } + + if (temp >= 0) + i915_release_temp(p, temp); + + return dest; } diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 50c9e65409..76a2184e9a 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -575,8 +575,12 @@ i915_translate_instruction(struct i915_fp_compile *p, src0 = src_vector(p, &inst->FullSrcRegisters[0]); tmp = i915_get_utemp(p); - i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ - 0, src0, T0_TEXKILL); + i915_emit_texld(p, + tmp, /* dest reg: a dummy reg */ + A0_DEST_CHANNEL_ALL, /* dest writemask */ + 0, /* sampler */ + src0, /* coord*/ + T0_TEXKILL); /* opcode */ break; case TGSI_OPCODE_LG2: @@ -970,6 +974,16 @@ i915_translate_instructions(struct i915_fp_compile *p, ifs->num_constants = MAX2(ifs->num_constants, i + 1); } } + else if (parse.FullToken.FullDeclaration.Declaration.File + == TGSI_FILE_TEMPORARY) { + uint i; + for (i = parse.FullToken.FullDeclaration.u.DeclarationRange.First; + i <= parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + i++) { + assert(i < I915_MAX_TEMPORARY); + p->temp_flag |= (1 << i); /* mark temp as used */ + } + } break; case TGSI_TOKEN_TYPE_IMMEDIATE: @@ -1048,7 +1062,7 @@ i915_init_compile(struct i915_context *i915, p->decl = p->declarations; p->decl_s = 0; p->decl_t = 0; - p->temp_flag = 0xffff000; + p->temp_flag = ~0x0 << I915_MAX_TEMPORARY; p->utemp_flag = ~0x7; p->wpos_tex = -1; -- cgit v1.2.3 From 7ed9beef5f08554f126c64aa172cd03fd810f1db Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 18:00:14 -0700 Subject: gallium/i915: remove debug code --- src/gallium/drivers/i915simple/i915_fpc_emit.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_fpc_emit.c b/src/gallium/drivers/i915simple/i915_fpc_emit.c index ad6dab6c11..4bdeefb449 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_emit.c +++ b/src/gallium/drivers/i915simple/i915_fpc_emit.c @@ -220,7 +220,6 @@ uint i915_emit_texld( struct i915_fp_compile *p, uint tempReg; temp = i915_get_temp(p); /* get temp reg index */ - printf("***** ALLOC TEMP %d\n", temp); tempReg = UREG(REG_TYPE_R, temp); /* make i915 register */ i915_emit_arith( p, A0_MOV, -- cgit v1.2.3 From f43c44b5c9bbbd37e0d40488f911f81e5f3a0367 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 18:53:22 -0700 Subject: gallium/i915: Fix emit_hw_vertex(): need to use vinfo->src_index[] --- src/gallium/drivers/i915simple/i915_prim_emit.c | 30 +++++++++++++------------ 1 file changed, 16 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c index 6da42b3846..d8de5178f6 100644 --- a/src/gallium/drivers/i915simple/i915_prim_emit.c +++ b/src/gallium/drivers/i915simple/i915_prim_emit.c @@ -75,37 +75,39 @@ emit_hw_vertex( struct i915_context *i915, assert(!i915->dirty); for (i = 0; i < vinfo->num_attribs; i++) { + const uint j = vinfo->src_index[i]; + const float *attrib = vertex->data[j]; switch (vinfo->emit[i]) { case EMIT_OMIT: /* no-op */ break; case EMIT_1F: - OUT_BATCH( fui(vertex->data[i][0]) ); + OUT_BATCH( fui(attrib[0]) ); count++; break; case EMIT_2F: - OUT_BATCH( fui(vertex->data[i][0]) ); - OUT_BATCH( fui(vertex->data[i][1]) ); + OUT_BATCH( fui(attrib[0]) ); + OUT_BATCH( fui(attrib[1]) ); count += 2; break; case EMIT_3F: - OUT_BATCH( fui(vertex->data[i][0]) ); - OUT_BATCH( fui(vertex->data[i][1]) ); - OUT_BATCH( fui(vertex->data[i][2]) ); + OUT_BATCH( fui(attrib[0]) ); + OUT_BATCH( fui(attrib[1]) ); + OUT_BATCH( fui(attrib[2]) ); count += 3; break; case EMIT_4F: - OUT_BATCH( fui(vertex->data[i][0]) ); - OUT_BATCH( fui(vertex->data[i][1]) ); - OUT_BATCH( fui(vertex->data[i][2]) ); - OUT_BATCH( fui(vertex->data[i][3]) ); + OUT_BATCH( fui(attrib[0]) ); + OUT_BATCH( fui(attrib[1]) ); + OUT_BATCH( fui(attrib[2]) ); + OUT_BATCH( fui(attrib[3]) ); count += 4; break; case EMIT_4UB: - OUT_BATCH( pack_ub4(float_to_ubyte( vertex->data[i][2] ), - float_to_ubyte( vertex->data[i][1] ), - float_to_ubyte( vertex->data[i][0] ), - float_to_ubyte( vertex->data[i][3] )) ); + OUT_BATCH( pack_ub4(float_to_ubyte( attrib[2] ), + float_to_ubyte( attrib[1] ), + float_to_ubyte( attrib[0] ), + float_to_ubyte( attrib[3] )) ); count += 1; break; default: -- cgit v1.2.3 From ba376b33140f722b9f935960f450bbca8873439e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 18:53:57 -0700 Subject: gallium/i915: plug in aapoint draw stage --- src/gallium/drivers/i915simple/i915_context.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 36f6429d46..c3955bbd2d 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -298,10 +298,8 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915_init_string_functions(i915); i915_init_texture_functions(i915); - /* not working yet: - draw_install_aapoint_stage(i915->draw, &i915->pipe); - */ draw_install_aaline_stage(i915->draw, &i915->pipe); + draw_install_aapoint_stage(i915->draw, &i915->pipe); i915->pci_id = pci_id; i915->flags.is_i945 = is_i945; -- cgit v1.2.3 From 08a5f49644c4bfc62291c49718f2d18e58527d1d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Feb 2008 18:56:03 -0700 Subject: gallium: rewrite AA point fragment shader to use a CMP instruction instead of IF/ELSE/ENDIF Allows the shader to work on i915 hardware. --- src/gallium/auxiliary/draw/draw_aapoint.c | 196 +++++++++++++++--------------- 1 file changed, 96 insertions(+), 100 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index 27a81f3e90..cae6fcd4d2 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -222,7 +222,7 @@ aa_transform_inst(struct tgsi_transform_context *ctx, * * Temp reg0 usage: * t0.x = distance of fragment from center point - * t0.y = boolean, is t0.x > 1 ? + * t0.y = boolean, is t0.x > 1.0, also misc temp usage * t0.z = temporary for computing 1/(1-k) value * t0.w = final coverage value */ @@ -313,9 +313,73 @@ aa_transform_inst(struct tgsi_transform_context *ctx, newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; ctx->emit_instruction(ctx, &newInst); - /* SGT t0.y, t0.x, tex.z; # bool b = distance > k */ + + /* compute coverage factor = (1-d)/(1-k) */ + + /* SUB t0.z, tex.w, tex.z; # m = 1 - k */ newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_SGT; + newInst.Instruction.Opcode = TGSI_OPCODE_SUB; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* RCP t0.z, t0.z; # t0.z = 1 / m */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_RCP; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* SUB t0.y, 1, t0.x; # d = 1 - d */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SUB; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + ctx->emit_instruction(ctx, &newInst); + + /* MUL t0.w, t0.y, t0.z; # coverage = d * m */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* SLE t0.y, t0.x, tex.z; # bool b = distance <= k */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SLE; newInst.Instruction.NumDstRegs = 1; newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; newInst.FullDstRegisters[0].DstRegister.Index = tmp0; @@ -329,111 +393,40 @@ aa_transform_inst(struct tgsi_transform_context *ctx, newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Z; ctx->emit_instruction(ctx, &newInst); - /* IF t0.y # if b then */ + /* CMP t0.w, -t0.y, tex.w, t0.w; + * # if -t0.y < 0 then + * t0.w = 1 + * else + * t0.w = t0.w + */ newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_IF; - newInst.Instruction.NumDstRegs = 0; - newInst.Instruction.NumSrcRegs = 1; + newInst.Instruction.Opcode = TGSI_OPCODE_CMP; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 3; newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[2].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[2].SrcRegister.SwizzleX = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[2].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[2].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[2].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; ctx->emit_instruction(ctx, &newInst); - { - /* compute coverage factor = (1-d)/(1-k) */ - - /* SUB t0.z, tex.w, tex.z; # m = 1 - k */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_SUB; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z; - ctx->emit_instruction(ctx, &newInst); - - /* RCP t0.z, t0.z; # t0.z = 1 / m */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_RCP; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; - newInst.Instruction.NumSrcRegs = 1; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Z; - ctx->emit_instruction(ctx, &newInst); - - /* SUB t0.x, 1, t0.x; # d = 1 - d */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_SUB; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; - ctx->emit_instruction(ctx, &newInst); - - /* MUL t0.w, t0.x, t0.z; # coverage = d * m */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_MUL; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Z; - ctx->emit_instruction(ctx, &newInst); - } - - /* ELSE */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_ELSE; - newInst.Instruction.NumDstRegs = 0; - newInst.Instruction.NumSrcRegs = 0; - ctx->emit_instruction(ctx, &newInst); - - { - /* MOV t0.w, tex.w; # coverage = 1.0 */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_MOV; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; - newInst.Instruction.NumSrcRegs = 1; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; - ctx->emit_instruction(ctx, &newInst); - } - - /* ENDIF */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_ENDIF; - newInst.Instruction.NumDstRegs = 0; - newInst.Instruction.NumSrcRegs = 0; - ctx->emit_instruction(ctx, &newInst); } if (inst->Instruction.Opcode == TGSI_OPCODE_END) { @@ -516,7 +509,7 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) (struct tgsi_token *) aapoint_fs.tokens, MAX, &transform.base); -#if 0 /* DEBUG */ +#if 1 /* DEBUG */ tgsi_dump(orig_fs->tokens, 0); tgsi_dump(aapoint_fs.tokens, 0); #endif @@ -613,6 +606,9 @@ aapoint_point(struct draw_stage *stage, struct prim_header *header) * ELSE * coverage = 1.0; // full coverage * ENDIF + * + * Note: the ELSEIF and ELSE clauses are actually implemented with CMP to + * avoid using IF/ELSE/ENDIF TGSI opcodes. */ #if !NORMALIZE -- cgit v1.2.3 From 6abb82da7e676384e7e2c9732307b23f8ed7157d Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 25 Feb 2008 22:03:58 -0500 Subject: implement deleting of driver side cached state in cso's --- src/gallium/auxiliary/cso_cache/cso_cache.c | 88 +++++++++++++++++++++++++++++ src/gallium/auxiliary/cso_cache/cso_cache.h | 14 +++++ src/gallium/auxiliary/cso_cache/cso_hash.h | 1 - src/mesa/state_tracker/st_cache.c | 13 ++++- 4 files changed, 114 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 9c32e94124..9aa1a64272 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -190,9 +190,96 @@ struct cso_cache *cso_cache_create(void) return sc; } +void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type, + void (*func)(void *state, void *user_data), void *user_data) +{ + struct cso_hash *hash = 0; + struct cso_hash_iter iter; + + switch (type) { + case CSO_BLEND: + hash = sc->blend_hash; + break; + case CSO_SAMPLER: + hash = sc->sampler_hash; + break; + case CSO_DEPTH_STENCIL_ALPHA: + hash = sc->depth_stencil_hash; + break; + case CSO_RASTERIZER: + hash = sc->rasterizer_hash; + break; + case CSO_FRAGMENT_SHADER: + hash = sc->fs_hash; + break; + case CSO_VERTEX_SHADER: + hash = sc->vs_hash; + break; + } + + iter = cso_hash_first_node(hash); + while (!cso_hash_iter_is_null(iter)) { + void *state = cso_hash_iter_data(iter); + if (state) { + func(state, user_data); + } + iter = cso_hash_iter_next(iter); + } +} + +static void delete_blend_state(void *state, void *user_data) +{ + struct cso_blend *cso = (struct cso_blend *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_depth_stencil_state(void *state, void *pipe) +{ + struct cso_depth_stencil_alpha *cso = (struct cso_depth_stencil_alpha *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_sampler_state(void *state, void *pipe) +{ + struct cso_sampler *cso = (struct cso_sampler *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_rasterizer_state(void *state, void *pipe) +{ + struct cso_rasterizer *cso = (struct cso_rasterizer *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_fs_state(void *state, void *pipe) +{ + struct cso_fragment_shader *cso = (struct cso_fragment_shader *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_vs_state(void *state, void *pipe) +{ + struct cso_vertex_shader *cso = (struct cso_vertex_shader *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + void cso_cache_delete(struct cso_cache *sc) { assert(sc); + /* delete driver data */ + cso_for_each_state(sc, CSO_BLEND, delete_blend_state, 0); + cso_for_each_state(sc, CSO_DEPTH_STENCIL_ALPHA, delete_depth_stencil_state, 0); + cso_for_each_state(sc, CSO_FRAGMENT_SHADER, delete_fs_state, 0); + cso_for_each_state(sc, CSO_VERTEX_SHADER, delete_vs_state, 0); + cso_for_each_state(sc, CSO_RASTERIZER, delete_rasterizer_state, 0); + cso_for_each_state(sc, CSO_SAMPLER, delete_sampler_state, 0); + cso_hash_delete(sc->blend_hash); cso_hash_delete(sc->sampler_hash); cso_hash_delete(sc->depth_stencil_hash); @@ -201,3 +288,4 @@ void cso_cache_delete(struct cso_cache *sc) cso_hash_delete(sc->vs_hash); FREE(sc); } + diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h index 3a005a376e..f3bd4623c9 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.h +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -97,31 +97,43 @@ struct cso_cache { struct cso_blend { struct pipe_blend_state state; void *data; + void (*delete_state)(void *, void *); + void *context; }; struct cso_depth_stencil_alpha { struct pipe_depth_stencil_alpha_state state; void *data; + void (*delete_state)(void *, void *); + void *context; }; struct cso_rasterizer { struct pipe_rasterizer_state state; void *data; + void (*delete_state)(void *, void *); + void *context; }; struct cso_fragment_shader { struct pipe_shader_state state; void *data; + void (*delete_state)(void *, void *); + void *context; }; struct cso_vertex_shader { struct pipe_shader_state state; void *data; + void (*delete_state)(void *, void *); + void *context; }; struct cso_sampler { struct pipe_sampler_state state; void *data; + void (*delete_state)(void *, void *); + void *context; }; @@ -147,6 +159,8 @@ struct cso_hash_iter cso_find_state(struct cso_cache *sc, struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, unsigned hash_key, enum cso_cache_type type, void *templ); +void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type, + void (*func)(void *state, void *user_data), void *user_data); void * cso_take_state(struct cso_cache *sc, unsigned hash_key, enum cso_cache_type type); diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h index 2e8b69675c..86c62c027a 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.h +++ b/src/gallium/auxiliary/cso_cache/cso_hash.h @@ -38,7 +38,6 @@ extern "C" { #endif - struct cso_hash; struct cso_node; diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 2979e7fae5..78f282302e 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -63,6 +63,8 @@ const struct cso_blend * st_cached_blend_state(struct st_context *st, cso->data = st->pipe->create_blend_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; + cso->delete_state = st->pipe->delete_blend_state; + cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_BLEND, cso); } return ((struct cso_blend *)cso_hash_iter_data(iter)); @@ -82,6 +84,8 @@ st_cached_sampler_state(struct st_context *st, cso->data = st->pipe->create_sampler_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; + cso->delete_state = st->pipe->delete_sampler_state; + cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_SAMPLER, cso); } return (struct cso_sampler*)(cso_hash_iter_data(iter)); @@ -103,6 +107,8 @@ st_cached_depth_stencil_alpha_state(struct st_context *st, cso->data = st->pipe->create_depth_stencil_alpha_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; + cso->delete_state = st->pipe->delete_depth_stencil_alpha_state; + cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso); } return (struct cso_depth_stencil_alpha*)(cso_hash_iter_data(iter)); @@ -123,6 +129,8 @@ const struct cso_rasterizer* st_cached_rasterizer_state( cso->data = st->pipe->create_rasterizer_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; + cso->delete_state = st->pipe->delete_rasterizer_state; + cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_RASTERIZER, cso); } return (struct cso_rasterizer*)(cso_hash_iter_data(iter)); @@ -143,6 +151,8 @@ st_cached_fs_state(struct st_context *st, cso->data = st->pipe->create_fs_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; + cso->delete_state = st->pipe->delete_fs_state; + cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_FRAGMENT_SHADER, cso); } return (struct cso_fragment_shader*)(cso_hash_iter_data(iter)); @@ -163,8 +173,9 @@ st_cached_vs_state(struct st_context *st, cso->data = st->pipe->create_vs_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; + cso->delete_state = st->pipe->delete_vs_state; + cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_VERTEX_SHADER, cso); } return (struct cso_vertex_shader*)(cso_hash_iter_data(iter)); } - -- cgit v1.2.3 From bf1c2f3602038440ffacf7ae494cb4e9bacc9bb9 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 26 Feb 2008 00:15:55 -0500 Subject: hide cso cache definition and add some initial code for size limiting the caches --- src/gallium/auxiliary/cso_cache/cso_cache.c | 22 ++++++++++++++++++++++ src/gallium/auxiliary/cso_cache/cso_cache.h | 28 +++++++++++----------------- 2 files changed, 33 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 9aa1a64272..1b870c8e41 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -33,6 +33,17 @@ #include "cso_cache.h" #include "cso_hash.h" + +struct cso_cache { + struct cso_hash *blend_hash; + struct cso_hash *depth_stencil_hash; + struct cso_hash *fs_hash; + struct cso_hash *vs_hash; + struct cso_hash *rasterizer_hash; + struct cso_hash *sampler_hash; + int max_size; +}; + #if 1 static unsigned hash_key(const void *key, unsigned key_size) { @@ -180,6 +191,7 @@ struct cso_cache *cso_cache_create(void) { struct cso_cache *sc = MALLOC_STRUCT(cso_cache); + sc->max_size = 4096; sc->blend_hash = cso_hash_create(); sc->sampler_hash = cso_hash_create(); sc->depth_stencil_hash = cso_hash_create(); @@ -289,3 +301,13 @@ void cso_cache_delete(struct cso_cache *sc) FREE(sc); } +void cso_set_maximum_cache_size(struct cso_cache *sc, int number) +{ + sc->max_size = number; +} + +int cso_maximum_cache_size(const struct cso_cache *sc) +{ + return sc->max_size; +} + diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h index f3bd4623c9..72a6b9d488 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.h +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -28,7 +28,7 @@ /** * @file * Constant State Object (CSO) cache. - * + * * The basic idea is that the states are created via the * create_state/bind_state/delete_state semantics. The driver is expected to * perform as much of the Gallium state translation to whatever its internal @@ -36,15 +36,15 @@ * mechanism where it stores the created states. When the pipeline needs an * actual state change, a bind call is issued. In the bind call the driver * gets its already translated representation. - * + * * Those semantics mean that the driver doesn't do the repeated translations * of states on every frame, but only once, when a new state is actually * created. - * + * * Even on hardware that doesn't do any kind of state cache, it makes the * driver look a lot neater, plus it avoids all the redundant state * translations on every frame. - * + * * Currently our constant state objects are: * - alpha test * - blend @@ -53,7 +53,7 @@ * - rasterizer (old setup) * - sampler * - vertex shader - * + * * Things that are not constant state objects include: * - blend_color * - clip_state @@ -65,7 +65,7 @@ * - scissor_state * - texture_state * - viewport_state - * + * * @author Zack Rusin */ @@ -75,24 +75,16 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" -/* cso_hash.h is necessary for cso_hash_iter, as MSVC requires structures +/* cso_hash.h is necessary for cso_hash_iter, as MSVC requires structures * returned by value to be fully defined */ -#include "cso_hash.h" +#include "cso_hash.h" #ifdef __cplusplus extern "C" { #endif - -struct cso_cache { - struct cso_hash *blend_hash; - struct cso_hash *depth_stencil_hash; - struct cso_hash *fs_hash; - struct cso_hash *vs_hash; - struct cso_hash *rasterizer_hash; - struct cso_hash *sampler_hash; -}; +struct cso_cache; struct cso_blend { struct pipe_blend_state state; @@ -164,6 +156,8 @@ void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type, void * cso_take_state(struct cso_cache *sc, unsigned hash_key, enum cso_cache_type type); +void cso_set_maximum_cache_size(struct cso_cache *sc, int number); +int cso_maximum_cache_size(const struct cso_cache *sc); #ifdef __cplusplus } -- cgit v1.2.3 From e7985105695a18c29c13deb2b8f40c15eef72ee6 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 26 Feb 2008 00:18:54 -0500 Subject: add an explicit callback decleration and use it to silence warnings --- src/gallium/auxiliary/cso_cache/cso_cache.c | 2 +- src/gallium/auxiliary/cso_cache/cso_cache.h | 4 +++- src/mesa/state_tracker/st_cache.c | 12 ++++++------ 3 files changed, 10 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 1b870c8e41..f31042bce1 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -203,7 +203,7 @@ struct cso_cache *cso_cache_create(void) } void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type, - void (*func)(void *state, void *user_data), void *user_data) + cso_state_callback func, void *user_data) { struct cso_hash *hash = 0; struct cso_hash_iter iter; diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h index 72a6b9d488..3b0fe100b8 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.h +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -138,6 +138,8 @@ enum cso_cache_type { CSO_VERTEX_SHADER }; +typedef void (*cso_state_callback)(void *, void *); + unsigned cso_construct_key(void *item, int item_size); struct cso_cache *cso_cache_create(void); @@ -152,7 +154,7 @@ struct cso_hash_iter cso_find_state_template(struct cso_cache *sc, unsigned hash_key, enum cso_cache_type type, void *templ); void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type, - void (*func)(void *state, void *user_data), void *user_data); + cso_state_callback func, void *user_data); void * cso_take_state(struct cso_cache *sc, unsigned hash_key, enum cso_cache_type type); diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c index 78f282302e..7ee4fadc37 100644 --- a/src/mesa/state_tracker/st_cache.c +++ b/src/mesa/state_tracker/st_cache.c @@ -63,7 +63,7 @@ const struct cso_blend * st_cached_blend_state(struct st_context *st, cso->data = st->pipe->create_blend_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; - cso->delete_state = st->pipe->delete_blend_state; + cso->delete_state = (cso_state_callback)st->pipe->delete_blend_state; cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_BLEND, cso); } @@ -84,7 +84,7 @@ st_cached_sampler_state(struct st_context *st, cso->data = st->pipe->create_sampler_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; - cso->delete_state = st->pipe->delete_sampler_state; + cso->delete_state = (cso_state_callback)st->pipe->delete_sampler_state; cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_SAMPLER, cso); } @@ -107,7 +107,7 @@ st_cached_depth_stencil_alpha_state(struct st_context *st, cso->data = st->pipe->create_depth_stencil_alpha_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; - cso->delete_state = st->pipe->delete_depth_stencil_alpha_state; + cso->delete_state = (cso_state_callback)st->pipe->delete_depth_stencil_alpha_state; cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso); } @@ -129,7 +129,7 @@ const struct cso_rasterizer* st_cached_rasterizer_state( cso->data = st->pipe->create_rasterizer_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; - cso->delete_state = st->pipe->delete_rasterizer_state; + cso->delete_state = (cso_state_callback)st->pipe->delete_rasterizer_state; cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_RASTERIZER, cso); } @@ -151,7 +151,7 @@ st_cached_fs_state(struct st_context *st, cso->data = st->pipe->create_fs_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; - cso->delete_state = st->pipe->delete_fs_state; + cso->delete_state = (cso_state_callback)st->pipe->delete_fs_state; cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_FRAGMENT_SHADER, cso); } @@ -173,7 +173,7 @@ st_cached_vs_state(struct st_context *st, cso->data = st->pipe->create_vs_state(st->pipe, &cso->state); if (!cso->data) cso->data = &cso->state; - cso->delete_state = st->pipe->delete_vs_state; + cso->delete_state = (cso_state_callback)st->pipe->delete_vs_state; cso->context = st->pipe; iter = cso_insert_state(st->cache, hash_key, CSO_VERTEX_SHADER, cso); } -- cgit v1.2.3 From 7838aaffdb9d34427ebcb73aac585c85d9622018 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 26 Feb 2008 01:48:01 -0500 Subject: implement cache limits for cso by default set to 4096, which might be on the large side --- src/gallium/auxiliary/cso_cache/cso_cache.c | 144 ++++++++++++++++++++-------- src/gallium/auxiliary/cso_cache/cso_hash.c | 5 + src/gallium/auxiliary/cso_cache/cso_hash.h | 4 +- 3 files changed, 110 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index f31042bce1..a6e8469b44 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -127,12 +127,106 @@ static int _cso_size_for_type(enum cso_cache_type type) return 0; } + +static void delete_blend_state(void *state, void *data) +{ + struct cso_blend *cso = (struct cso_blend *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_depth_stencil_state(void *state, void *data) +{ + struct cso_depth_stencil_alpha *cso = (struct cso_depth_stencil_alpha *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_sampler_state(void *state, void *data) +{ + struct cso_sampler *cso = (struct cso_sampler *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_rasterizer_state(void *state, void *data) +{ + struct cso_rasterizer *cso = (struct cso_rasterizer *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_fs_state(void *state, void *data) +{ + struct cso_fragment_shader *cso = (struct cso_fragment_shader *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + +static void delete_vs_state(void *state, void *data) +{ + struct cso_vertex_shader *cso = (struct cso_vertex_shader *)state; + if (cso->delete_state && cso->data != &cso->state) + cso->delete_state(cso->context, cso->data); +} + + +static INLINE void delete_cso(void *state, enum cso_cache_type type) +{ + switch (type) { + case CSO_BLEND: { + delete_blend_state(state, 0); + } + break; + case CSO_SAMPLER: { + delete_sampler_state(state, 0); + } + break; + case CSO_DEPTH_STENCIL_ALPHA: { + delete_depth_stencil_state(state, 0); + } + break; + case CSO_RASTERIZER: { + delete_rasterizer_state(state, 0); + } + break; + case CSO_FRAGMENT_SHADER: { + delete_fs_state(state, 0); + } + break; + case CSO_VERTEX_SHADER: { + delete_vs_state(state, 0); + } + break; + } + free(state); +} + +static INLINE void sanitize_hash(struct cso_hash *hash, enum cso_cache_type type, + int max_size) +{ + /* if we're approach the maximum size, remove fourth of the entries + * otherwise every subsequent call will go through the same */ + int max_entries = (max_size > cso_hash_size(hash)) ? max_size : cso_hash_size(hash); + int to_remove = (max_size < max_entries) * max_entries/4; + while (to_remove) { + /*remove elements until we're good */ + /*fixme: currently we pick the nodes to remove at random*/ + struct cso_hash_iter iter = cso_hash_first_node(hash); + void *cso = cso_hash_take(hash, cso_hash_iter_key(iter)); + delete_cso(cso, type); + --to_remove; + } +} + struct cso_hash_iter cso_insert_state(struct cso_cache *sc, unsigned hash_key, enum cso_cache_type type, void *state) { struct cso_hash *hash = _cso_hash_for_type(sc, type); + sanitize_hash(hash, type, sc->max_size); + return cso_hash_insert(hash, hash_key, state); } @@ -239,48 +333,6 @@ void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type, } } -static void delete_blend_state(void *state, void *user_data) -{ - struct cso_blend *cso = (struct cso_blend *)state; - if (cso->delete_state && cso->data != &cso->state) - cso->delete_state(cso->context, cso->data); -} - -static void delete_depth_stencil_state(void *state, void *pipe) -{ - struct cso_depth_stencil_alpha *cso = (struct cso_depth_stencil_alpha *)state; - if (cso->delete_state && cso->data != &cso->state) - cso->delete_state(cso->context, cso->data); -} - -static void delete_sampler_state(void *state, void *pipe) -{ - struct cso_sampler *cso = (struct cso_sampler *)state; - if (cso->delete_state && cso->data != &cso->state) - cso->delete_state(cso->context, cso->data); -} - -static void delete_rasterizer_state(void *state, void *pipe) -{ - struct cso_rasterizer *cso = (struct cso_rasterizer *)state; - if (cso->delete_state && cso->data != &cso->state) - cso->delete_state(cso->context, cso->data); -} - -static void delete_fs_state(void *state, void *pipe) -{ - struct cso_fragment_shader *cso = (struct cso_fragment_shader *)state; - if (cso->delete_state && cso->data != &cso->state) - cso->delete_state(cso->context, cso->data); -} - -static void delete_vs_state(void *state, void *pipe) -{ - struct cso_vertex_shader *cso = (struct cso_vertex_shader *)state; - if (cso->delete_state && cso->data != &cso->state) - cso->delete_state(cso->context, cso->data); -} - void cso_cache_delete(struct cso_cache *sc) { assert(sc); @@ -304,6 +356,14 @@ void cso_cache_delete(struct cso_cache *sc) void cso_set_maximum_cache_size(struct cso_cache *sc, int number) { sc->max_size = number; + + sanitize_hash(sc->blend_hash, CSO_BLEND, sc->max_size); + sanitize_hash(sc->depth_stencil_hash, CSO_DEPTH_STENCIL_ALPHA, + sc->max_size); + sanitize_hash(sc->fs_hash, CSO_FRAGMENT_SHADER, sc->max_size); + sanitize_hash(sc->vs_hash, CSO_VERTEX_SHADER, sc->max_size); + sanitize_hash(sc->rasterizer_hash, CSO_RASTERIZER, sc->max_size); + sanitize_hash(sc->sampler_hash, CSO_SAMPLER, sc->max_size); } int cso_maximum_cache_size(const struct cso_cache *sc) diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index 208fc58502..b3b4d667d2 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -395,3 +395,8 @@ struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash) struct cso_hash_iter iter = {hash, cso_data_first_node(hash->data.d)}; return iter; } + +int cso_hash_size(struct cso_hash *hash) +{ + return hash->data.d->size; +} diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h index 86c62c027a..d5bca9d591 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.h +++ b/src/gallium/auxiliary/cso_cache/cso_hash.h @@ -47,7 +47,9 @@ struct cso_hash_iter { }; struct cso_hash *cso_hash_create(void); -void cso_hash_delete(struct cso_hash *hash); +void cso_hash_delete(struct cso_hash *hash); + +int cso_hash_size(struct cso_hash *hash); struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, void *data); -- cgit v1.2.3 From ad6bb870de6103ed240fa1f9f828bd13a4401a9a Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 26 Feb 2008 11:49:25 +0100 Subject: gallium: Fix build on Windows. --- src/gallium/auxiliary/cso_cache/cso_cache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index a6e8469b44..b427b509f8 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -199,7 +199,7 @@ static INLINE void delete_cso(void *state, enum cso_cache_type type) } break; } - free(state); + FREE(state); } static INLINE void sanitize_hash(struct cso_hash *hash, enum cso_cache_type type, -- cgit v1.2.3 From bb68acbf8dceb3a228c0f1ac18bfabd89a327b14 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 26 Feb 2008 14:05:31 +0900 Subject: gallium: Add missing copyright headers. --- src/mesa/state_tracker/st_atom_shader.h | 31 +++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_program.h | 26 ++++++++++++++++++++++++++ 2 files changed, 57 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.h b/src/mesa/state_tracker/st_atom_shader.h index 48e5348d73..f3124d87c0 100644 --- a/src/mesa/state_tracker/st_atom_shader.h +++ b/src/mesa/state_tracker/st_atom_shader.h @@ -1,5 +1,34 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +#ifndef ST_ATOM_SHADER_H +#define ST_ATOM_SHADER_H + extern void st_remove_vertex_program(struct st_context *, struct st_vertex_program *); @@ -7,3 +36,5 @@ st_remove_vertex_program(struct st_context *, struct st_vertex_program *); extern void st_remove_fragment_program(struct st_context *, struct st_fragment_program *); + +#endif /* ST_ATOM_SHADER_H */ diff --git a/src/mesa/state_tracker/st_cb_program.h b/src/mesa/state_tracker/st_cb_program.h index 38ec4e0a59..45dc753dff 100644 --- a/src/mesa/state_tracker/st_cb_program.h +++ b/src/mesa/state_tracker/st_cb_program.h @@ -1,3 +1,29 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ #ifndef ST_CB_PROGRAM_H #define ST_CB_PROGRAM_H -- cgit v1.2.3 From 232a41e19faa860f083e414cb1eb38c0617e9241 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 08:41:42 -0700 Subject: gallium: disable debug code --- src/gallium/auxiliary/draw/draw_aapoint.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index cae6fcd4d2..d16adb2505 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -509,7 +509,7 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) (struct tgsi_token *) aapoint_fs.tokens, MAX, &transform.base); -#if 1 /* DEBUG */ +#if 0 /* DEBUG */ tgsi_dump(orig_fs->tokens, 0); tgsi_dump(aapoint_fs.tokens, 0); #endif -- cgit v1.2.3 From ecd50ef58b034e604ff6b2fedbb0815953e510ea Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 08:43:07 -0700 Subject: gallium: remove input_map[] from pipe_shader_state --- src/gallium/include/pipe/p_state.h | 1 - src/mesa/state_tracker/st_atom_shader.c | 2 +- src/mesa/state_tracker/st_program.c | 2 +- src/mesa/state_tracker/st_program.h | 3 +++ 4 files changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 15c88881eb..ddf3c1c79b 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -162,7 +162,6 @@ struct pipe_shader_state const struct tgsi_token *tokens; ubyte num_inputs; ubyte num_outputs; - ubyte input_map[PIPE_MAX_SHADER_INPUTS]; /* XXX this may be temporary */ ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index b67b620eaa..10c131d554 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -227,7 +227,7 @@ find_translated_vp(struct st_context *st, if (fpInAttrib >= 0) { GLuint fpInSlot = stfp->input_to_slot[fpInAttrib]; if (fpInSlot != ~0) { - GLuint vpOutSlot = stfp->cso->state.input_map[fpInSlot]; + GLuint vpOutSlot = stfp->input_map[fpInSlot]; xvp->output_to_slot[outAttr] = vpOutSlot; numVpOuts++; } diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index dc992ee9c2..aa252c845a 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -311,7 +311,7 @@ st_translate_fragment_program(struct st_context *st, defaultInputMapping[attr] = slot; - fs.input_map[slot] = vslot++; + stfp->input_map[slot] = vslot++; fs.num_inputs++; diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 25cf3e94a8..31558af6ce 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -57,6 +57,9 @@ struct st_fragment_program GLuint input_to_slot[FRAG_ATTRIB_MAX]; /**< Maps FRAG_ATTRIB_x to slot */ GLuint num_input_slots; + /** map FP input back to VP output */ + GLuint input_map[PIPE_MAX_SHADER_INPUTS]; + /** The program in TGSI format */ struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; -- cgit v1.2.3 From 1410b7bb509ef37c41043b173bc1047257483af0 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 10:12:17 -0700 Subject: gallium: collect more shader info in tgsi_scan_shader() Now getting input/output semantic info so we can eventually remove those fields from pipe_shader_state. --- src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 20 ++++++++++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_scan.h | 11 +++++++++++ 2 files changed, 31 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c index 4b99ac37cc..a1cc681492 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -69,6 +69,8 @@ tgsi_scan_shader(const struct tgsi_token *tokens, */ while( !tgsi_parse_end_of_tokens( &parse ) ) { + info->num_tokens++; + tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { @@ -91,9 +93,27 @@ tgsi_scan_shader(const struct tgsi_token *tokens, for (i = fulldecl->u.DeclarationRange.First; i <= fulldecl->u.DeclarationRange.Last; i++) { + + /* only first 32 regs will appear in this bitfield */ info->file_mask[file] |= (1 << i); info->file_count[file]++; + if (file == TGSI_FILE_INPUT) { + info->input_semantic_name[info->num_inputs] + = fulldecl->Semantic.SemanticName; + info->input_semantic_index[info->num_inputs] + = fulldecl->Semantic.SemanticIndex; + info->num_inputs++; + } + + if (file == TGSI_FILE_OUTPUT) { + info->output_semantic_name[info->num_outputs] + = fulldecl->Semantic.SemanticName; + info->output_semantic_index[info->num_outputs] + = fulldecl->Semantic.SemanticIndex; + info->num_outputs++; + } + /* special case */ if (procType == TGSI_PROCESSOR_FRAGMENT && file == TGSI_FILE_OUTPUT && diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h index 757446437c..dc6dfd6d0b 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h @@ -30,6 +30,7 @@ #include "pipe/p_util.h" +#include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" @@ -38,6 +39,16 @@ */ struct tgsi_shader_info { + uint num_tokens; + + /* XXX eventually remove the corresponding fields from pipe_shader_state: */ + ubyte num_inputs; + ubyte num_outputs; + ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ + ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ + ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; + uint file_mask[TGSI_FILE_COUNT]; /**< bitmask of declared registers */ uint file_count[TGSI_FILE_COUNT]; /**< number of declared registers */ -- cgit v1.2.3 From f74279002a0ae0b106bd5410487ef9c0e9b1d8b6 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 10:13:39 -0700 Subject: gallium: added tgsi_shader_field to sp_fragment_shader Use the shader semantic info from there, instead of from pipe_shader_state. Carry this idea to draw module and other drivers... --- src/gallium/drivers/softpipe/sp_fs_llvm.c | 2 +- src/gallium/drivers/softpipe/sp_prim_setup.c | 18 +++++++++--------- src/gallium/drivers/softpipe/sp_quad.c | 2 +- src/gallium/drivers/softpipe/sp_quad_fs.c | 4 ++-- src/gallium/drivers/softpipe/sp_state.h | 6 ++++-- src/gallium/drivers/softpipe/sp_state_derived.c | 9 +++++---- src/gallium/drivers/softpipe/sp_state_fs.c | 16 ++++++++++------ 7 files changed, 32 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_fs_llvm.c b/src/gallium/drivers/softpipe/sp_fs_llvm.c index 34b2b7d4e2..07d058155f 100644 --- a/src/gallium/drivers/softpipe/sp_fs_llvm.c +++ b/src/gallium/drivers/softpipe/sp_fs_llvm.c @@ -96,7 +96,7 @@ shade_quad_llvm(struct quad_stage *qs, if (qss->colorOutSlot >= 0) { unsigned i; /* XXX need to handle multiple color outputs someday */ - allvmrt(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] + allvmrt(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] == TGSI_SEMANTIC_COLOR); for (i = 0; i < QUAD_SIZE; ++i) { quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0]; diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index b6a3fddb29..17284539b0 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -514,7 +514,7 @@ setup_fragcoord_coeff(struct setup_stage *setup, uint slot) static void setup_tri_coefficients( struct setup_stage *setup ) { struct softpipe_context *softpipe = setup->softpipe; - const struct pipe_shader_state *fs = &softpipe->fs->shader; + const struct sp_fragment_shader *spfs = softpipe->fs; const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); uint fragSlot; @@ -525,7 +525,7 @@ static void setup_tri_coefficients( struct setup_stage *setup ) /* setup interpolation for all the remaining attributes: */ - for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { + for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { const uint vertSlot = vinfo->src_index[fragSlot]; uint j; @@ -549,7 +549,7 @@ static void setup_tri_coefficients( struct setup_stage *setup ) assert(0); } - if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { /* FOG.y = front/back facing XXX fix this */ setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; setup->coef[fragSlot].dadx[1] = 0.0; @@ -757,7 +757,7 @@ static INLINE void setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) { struct softpipe_context *softpipe = setup->softpipe; - const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; + const struct sp_fragment_shader *spfs = softpipe->fs; const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); uint fragSlot; @@ -779,7 +779,7 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) /* setup interpolation for all the remaining attributes: */ - for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { + for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { const uint vertSlot = vinfo->src_index[fragSlot]; uint j; @@ -803,7 +803,7 @@ setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) assert(0); } - if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { /* FOG.y = front/back facing XXX fix this */ setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; setup->coef[fragSlot].dadx[1] = 0.0; @@ -967,7 +967,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); struct softpipe_context *softpipe = setup->softpipe; - const struct pipe_shader_state *fs = &softpipe->fs->shader; + const struct sp_fragment_shader *spfs = softpipe->fs; const struct vertex_header *v0 = prim->v[0]; const int sizeAttr = setup->softpipe->psize_slot; const float size @@ -1002,7 +1002,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) const_coeff(setup, &setup->posCoef, 0, 2); const_coeff(setup, &setup->posCoef, 0, 3); - for (fragSlot = 0; fragSlot < fs->num_inputs; fragSlot++) { + for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { const uint vertSlot = vinfo->src_index[fragSlot]; uint j; @@ -1025,7 +1025,7 @@ setup_point(struct draw_stage *stage, struct prim_header *prim) assert(0); } - if (fs->input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { /* FOG.y = front/back facing XXX fix this */ setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; setup->coef[fragSlot].dadx[1] = 0.0; diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad.c index 142dbcc771..0aaf94021f 100644 --- a/src/gallium/drivers/softpipe/sp_quad.c +++ b/src/gallium/drivers/softpipe/sp_quad.c @@ -61,7 +61,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->framebuffer.zsbuf && !sp->depth_stencil->alpha.enabled && !sp->fs->uses_kill && - !sp->fs->writes_z; + !sp->fs->info.writes_z; /* build up the pipeline in reverse order... */ diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 2f40e09d5c..1794fb5a61 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -91,7 +91,7 @@ shade_quad( /* store result color */ if (qss->colorOutSlot >= 0) { /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->shader.output_semantic_name[qss->colorOutSlot] + assert(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] == TGSI_SEMANTIC_COLOR); memcpy( quad->outputs.color, @@ -149,7 +149,7 @@ static void shade_begin(struct quad_stage *qs) qss->colorOutSlot = -1; qss->depthOutSlot = -1; for (i = 0; i < qss->stage.softpipe->fs->shader.num_outputs; i++) { - switch (qss->stage.softpipe->fs->shader.output_semantic_name[i]) { + switch (qss->stage.softpipe->fs->info.output_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: qss->depthOutSlot = i; break; diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 5aaa9e346b..b1070e185a 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -32,6 +32,7 @@ #define SP_STATE_H #include "pipe/p_state.h" +#include "tgsi/util/tgsi_scan.h" #define SP_NEW_VIEWPORT 0x1 @@ -61,10 +62,11 @@ struct tgsi_exec_machine; * This is starting to look an awful lot like a quad pipeline stage... */ struct sp_fragment_shader { - struct pipe_shader_state shader; + struct pipe_shader_state shader; + + struct tgsi_shader_info info; boolean uses_kill; - boolean writes_z; void (*prepare)( const struct sp_fragment_shader *shader, struct tgsi_exec_machine *machine, diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 4c6313001f..68702eafcf 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -62,6 +62,7 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) if (vinfo->num_attribs == 0) { /* compute vertex layout now */ const struct pipe_shader_state *vs = &softpipe->vs->shader; + const struct sp_fragment_shader *spfs = softpipe->fs; const struct pipe_shader_state *fs = &softpipe->fs->shader; const enum interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; @@ -83,9 +84,9 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) * from the vertex shader. */ vinfo->num_attribs = 0; - for (i = 0; i < fs->num_inputs; i++) { + for (i = 0; i < spfs->info.num_inputs; i++) { int src; - switch (fs->input_semantic_name[i]) { + switch (spfs->info.input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_POSITION, 0); @@ -94,7 +95,7 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) case TGSI_SEMANTIC_COLOR: src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_COLOR, - fs->input_semantic_index[i]); + spfs->info.input_semantic_index[i]); draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); break; @@ -106,7 +107,7 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ src = draw_find_vs_output(softpipe->draw, TGSI_SEMANTIC_GENERIC, - fs->input_semantic_index[i]); + spfs->info.input_semantic_index[i]); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index b184ac61bb..2715dca0f0 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -45,13 +45,12 @@ softpipe_create_fs_state(struct pipe_context *pipe, { struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_fragment_shader *state; - struct tgsi_shader_info info; - - tgsi_scan_shader(templ->tokens, &info); + /* debug */ if (softpipe->dump_fs) tgsi_dump(templ->tokens, 0); + /* codegen */ state = softpipe_create_fs_llvm( softpipe, templ ); if (!state) { state = softpipe_create_fs_sse( softpipe, templ ); @@ -59,10 +58,15 @@ softpipe_create_fs_state(struct pipe_context *pipe, state = softpipe_create_fs_exec( softpipe, templ ); } } + assert(state); - state->uses_kill = (info.opcode_count[TGSI_OPCODE_KIL] || - info.opcode_count[TGSI_OPCODE_KILP]); - state->writes_z = info.writes_z; + + /* get/save the summary info for this shader */ + tgsi_scan_shader(templ->tokens, &state->info); + + /* convenience field */ + state->uses_kill = (state->info.opcode_count[TGSI_OPCODE_KIL] || + state->info.opcode_count[TGSI_OPCODE_KILP]); return state; } -- cgit v1.2.3 From 33d213b6776701ec16b5c02b111ed31de5e93f43 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 10:44:44 -0700 Subject: gallium: remove unused var --- src/gallium/drivers/softpipe/sp_state_derived.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 68702eafcf..aa6e329116 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -63,7 +63,6 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) /* compute vertex layout now */ const struct pipe_shader_state *vs = &softpipe->vs->shader; const struct sp_fragment_shader *spfs = softpipe->fs; - const struct pipe_shader_state *fs = &softpipe->fs->shader; const enum interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; uint i; -- cgit v1.2.3 From 4901410293b35ac6bb4759142b50fcc0be8a1b25 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 10:47:42 -0700 Subject: gallium/i915: Use tgsi_scan_shader() to collect shader info No longer use semantic info in pipe_shader_state. Also, remove redundant semantic info from i915_fp_compile struct. --- src/gallium/drivers/i915simple/i915_context.h | 5 ++++ src/gallium/drivers/i915simple/i915_fpc.h | 6 ---- .../drivers/i915simple/i915_fpc_translate.c | 33 ++++------------------ src/gallium/drivers/i915simple/i915_state.c | 2 ++ .../drivers/i915simple/i915_state_derived.c | 12 ++++---- 5 files changed, 19 insertions(+), 39 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index d32dded6bd..20cf7d3c3b 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -35,6 +35,8 @@ #include "draw/draw_vertex.h" +#include "tgsi/util/tgsi_scan.h" + #define I915_TEX_UNITS 8 @@ -89,6 +91,9 @@ struct i915_fragment_shader { struct pipe_shader_state state; + + struct tgsi_shader_info info; + uint *program; uint program_len; diff --git a/src/gallium/drivers/i915simple/i915_fpc.h b/src/gallium/drivers/i915simple/i915_fpc.h index 250dfe6dbf..80a9576304 100644 --- a/src/gallium/drivers/i915simple/i915_fpc.h +++ b/src/gallium/drivers/i915simple/i915_fpc.h @@ -58,12 +58,6 @@ struct i915_fp_compile { uint declarations[I915_PROGRAM_SIZE]; uint program[I915_PROGRAM_SIZE]; - uint input_semantic_name[PIPE_MAX_SHADER_INPUTS]; - uint input_semantic_index[PIPE_MAX_SHADER_INPUTS]; - - uint output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; - uint output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; - uint *csr; /**< Cursor, points into program. */ uint *decl; /**< Cursor, points into declarations. */ diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 76a2184e9a..c2d9a93c6c 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -162,8 +162,8 @@ src_vector(struct i915_fp_compile *p, * We also use a texture coordinate to pass wpos when possible. */ - sem_name = p->input_semantic_name[index]; - sem_ind = p->input_semantic_index[index]; + sem_name = p->shader->info.input_semantic_name[index]; + sem_ind = p->shader->info.input_semantic_index[index]; switch (sem_name) { case TGSI_SEMANTIC_POSITION: @@ -265,7 +265,7 @@ get_result_vector(struct i915_fp_compile *p, switch (dest->DstRegister.File) { case TGSI_FILE_OUTPUT: { - uint sem_name = p->output_semantic_name[dest->DstRegister.Index]; + uint sem_name = p->shader->info.output_semantic_name[dest->DstRegister.Index]; switch (sem_name) { case TGSI_SEMANTIC_POSITION: return UREG(REG_TYPE_OD, 0); @@ -942,28 +942,6 @@ i915_translate_instructions(struct i915_fp_compile *p, switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: if (parse.FullToken.FullDeclaration.Declaration.File - == TGSI_FILE_INPUT) { - /* save input register info for use in src_vector() */ - uint ind, sem, semi; - ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; - semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - /*debug_printf("FS Input DECL [%u] sem %u\n", ind, sem);*/ - p->input_semantic_name[ind] = sem; - p->input_semantic_index[ind] = semi; - } - else if (parse.FullToken.FullDeclaration.Declaration.File - == TGSI_FILE_OUTPUT) { - /* save output register info for use in get_result_vector() */ - uint ind, sem, semi; - ind = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - sem = parse.FullToken.FullDeclaration.Semantic.SemanticName; - semi = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; - /*debug_printf("FS Output DECL [%u] sem %u\n", ind, sem);*/ - p->output_semantic_name[ind] = sem; - p->output_semantic_index[ind] = semi; - } - else if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_CONSTANT) { uint i; for (i = parse.FullToken.FullDeclaration.u.DeclarationRange.First; @@ -981,6 +959,7 @@ i915_translate_instructions(struct i915_fp_compile *p, i <= parse.FullToken.FullDeclaration.u.DeclarationRange.Last; i++) { assert(i < I915_MAX_TEMPORARY); + /* XXX just use shader->info->file_mask[TGSI_FILE_TEMPORARY] */ p->temp_flag |= (1 << i); /* mark temp as used */ } } @@ -1163,7 +1142,7 @@ i915_find_wpos_space(struct i915_fp_compile *p) i915_program_error(p, "No free texcoord for wpos value"); } #else - if (p->shader->state.input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { + if (p->shader->info.input_semantic_name[0] == TGSI_SEMANTIC_POSITION) { /* frag shader using the fragment position input */ #if 0 assert(0); @@ -1184,7 +1163,7 @@ static void i915_fixup_depth_write(struct i915_fp_compile *p) { /* XXX assuming pos/depth is always in output[0] */ - if (p->shader->state.output_semantic_name[0] == TGSI_SEMANTIC_POSITION) { + if (p->shader->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION) { const uint depth = UREG(REG_TYPE_OD, 0); i915_emit_arith(p, diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index a35bdf941f..9df0e12540 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -429,6 +429,8 @@ i915_create_fs_state(struct pipe_context *pipe, ifs->state = *templ; + tgsi_scan_shader(templ->tokens, &ifs->info); + /* The shader's compiled to i915 instructions here */ i915_translate_fragment_program(i915, ifs); diff --git a/src/gallium/drivers/i915simple/i915_state_derived.c b/src/gallium/drivers/i915simple/i915_state_derived.c index 5cf70acdf3..4daccec6e0 100644 --- a/src/gallium/drivers/i915simple/i915_state_derived.c +++ b/src/gallium/drivers/i915simple/i915_state_derived.c @@ -43,7 +43,7 @@ */ static void calculate_vertex_layout( struct i915_context *i915 ) { - const struct pipe_shader_state *fs = &i915->fs->state; + const struct i915_fragment_shader *fs = i915->fs; const enum interp_mode colorInterp = i915->rasterizer->color_interp; struct vertex_info vinfo; boolean texCoords[8], colors[2], fog, needW; @@ -57,18 +57,18 @@ static void calculate_vertex_layout( struct i915_context *i915 ) /* Determine which fragment program inputs are needed. Setup HW vertex * layout below, in the HW-specific attribute order. */ - for (i = 0; i < fs->num_inputs; i++) { - switch (fs->input_semantic_name[i]) { + for (i = 0; i < fs->info.num_inputs; i++) { + switch (fs->info.input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: break; case TGSI_SEMANTIC_COLOR: - assert(fs->input_semantic_index[i] < 2); - colors[fs->input_semantic_index[i]] = TRUE; + assert(fs->info.input_semantic_index[i] < 2); + colors[fs->info.input_semantic_index[i]] = TRUE; break; case TGSI_SEMANTIC_GENERIC: /* usually a texcoord */ { - const uint unit = fs->input_semantic_index[i]; + const uint unit = fs->info.input_semantic_index[i]; assert(unit < 8); texCoords[unit] = TRUE; needW = TRUE; -- cgit v1.2.3 From dbf12dcdb78aa251fe0d09b49aa661481727ecf6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 26 Feb 2008 18:48:13 +0000 Subject: intel_winsys: Fix build failure due to DEBUG_IOCTL being undefined. Include pipe/p_debug.h to consistently enable or disable the debugging code. --- src/gallium/winsys/dri/intel/intel_context.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_context.h b/src/gallium/winsys/dri/intel/intel_context.h index b01370c049..45430984d8 100644 --- a/src/gallium/winsys/dri/intel/intel_context.h +++ b/src/gallium/winsys/dri/intel/intel_context.h @@ -30,6 +30,9 @@ #include "drm.h" + +#include "pipe/p_debug.h" + #include "intel_screen.h" #include "i915_drm.h" -- cgit v1.2.3 From 75dac3959f959f0227b4e172696ac43910f9257a Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 26 Feb 2008 11:18:51 -0800 Subject: cell: Multiple rendering contexts don't work yet Log a message and forcibly exit. This prevents silly fools from thinking there's a bug...instead of just an unimplemented feature. :) --- src/gallium/drivers/cell/ppu/cell_spu.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_spu.c b/src/gallium/drivers/cell/ppu/cell_spu.c index 419e74dc40..973c0b1aa1 100644 --- a/src/gallium/drivers/cell/ppu/cell_spu.c +++ b/src/gallium/drivers/cell/ppu/cell_spu.c @@ -97,8 +97,18 @@ static void *cell_thread_function(void *arg) void cell_start_spus(struct cell_context *cell) { + static boolean one_time_init = FALSE; uint i, j; + + if (one_time_init) { + fprintf(stderr, "PPU: Multiple rendering contexts not yet supported " + "on Cell.\n"); + abort(); + } + + one_time_init = TRUE; + assert(cell->num_spus <= MAX_SPUS); ASSERT_ALIGN16(&cell_global.command[0]); -- cgit v1.2.3 From 36aa9cf781440ce685930586cbf53248cf9c0dc2 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 26 Feb 2008 20:32:42 +0100 Subject: gallium: Print texture target for short dumps. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index ff74e6117c..59be14a748 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -664,6 +664,19 @@ static const char *TGSI_TEXTURES[] = "TEXTURE_SHADOWRECT" }; +static const char *TGSI_TEXTURES_SHORT[] = +{ + "UNKNOWN", + "1D", + "2D", + "3D", + "CUBE", + "RECT", + "SHADOW1D", + "SHADOW2D", + "SHADOWRECT" +}; + static const char *TGSI_SRC_REGISTER_EXTS[] = { "SRC_REGISTER_EXT_TYPE_SWZ", @@ -1037,6 +1050,11 @@ dump_instruction_short( first_reg = FALSE; } + if (inst->InstructionExtTexture.Texture != TGSI_TEXTURE_UNKNOWN) { + TXT( ", " ); + ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES_SHORT ); + } + switch( inst->Instruction.Opcode ) { case TGSI_OPCODE_IF: case TGSI_OPCODE_ELSE: -- cgit v1.2.3 From 80efc5feb061a8ed9c1e91ad3711547927fa29e3 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 14:25:36 -0700 Subject: gallium: fix off by one rasterpos bug --- src/mesa/state_tracker/st_cb_rasterpos.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 5b0eb6022b..4d73916a35 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -145,7 +145,7 @@ rastpos_point(struct draw_stage *stage, struct prim_header *prim) /* update raster pos */ pos = prim->v[0]->data[0]; ctx->Current.RasterPos[0] = pos[0]; - ctx->Current.RasterPos[1] = height - 1 - pos[1]; + ctx->Current.RasterPos[1] = height - pos[1]; /* invert Y */ ctx->Current.RasterPos[2] = pos[2]; ctx->Current.RasterPos[3] = pos[3]; -- cgit v1.2.3 From b93cf55f4ecd94f5e9d5dda49d9092e3b769d044 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 14:26:40 -0700 Subject: gallium: fix zero-sized viewport bug If st_create_framebuffer() is called with width=0, height=0 and the program never called glViewport, the viewport wasn't properly initalized. This fixes that. --- src/mesa/state_tracker/st_framebuffer.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index bca3fa5c38..47d47daf2a 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -124,6 +124,17 @@ void st_resize_framebuffer( struct st_framebuffer *stfb, if (stfb->Base.Width != width || stfb->Base.Height != height) { GET_CURRENT_CONTEXT(ctx); if (ctx) { + if (stfb->InitWidth == 0 && stfb->InitHeight == 0) { + /* didn't have a valid size until now */ + stfb->InitWidth = width; + stfb->InitHeight = height; + if (ctx->Viewport.Width <= 1) { + /* set context's initial viewport/scissor size */ + _mesa_set_viewport(ctx, 0, 0, width, height); + _mesa_set_scissor(ctx, 0, 0, width, height); + } + } + _mesa_resize_framebuffer(ctx, &stfb->Base, width, height); assert(stfb->Base.Width == width); -- cgit v1.2.3 From 5e29aab1752c3e07ae2ebde4cb00e6550dab0eb2 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 14:29:35 -0700 Subject: gallium: replace draw_convert_wide_points() with draw_wide_point_threshold() Specifying a threshold size is a bit more flexible, and allows the option of converting even 1-pixel points to triangles (set threshold=0). Also, remove 0.25 pixel bias in wide_point(). --- src/gallium/auxiliary/draw/draw_context.c | 10 +++++----- src/gallium/auxiliary/draw/draw_private.h | 2 +- src/gallium/auxiliary/draw/draw_validate.c | 23 +++++++++++++---------- src/gallium/auxiliary/draw/draw_wide_prims.c | 7 ++++--- src/gallium/drivers/softpipe/sp_context.c | 3 --- 5 files changed, 23 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 7dd1c6f6fa..48f00946ea 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -80,7 +80,7 @@ struct draw_context *draw_create( void ) draw->shader_queue_flush = draw_vertex_shader_queue_flush; - draw->convert_wide_points = TRUE; + draw->wide_point_threshold = 1000000.0; /* infinity */ draw->convert_wide_lines = TRUE; draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ @@ -220,14 +220,14 @@ draw_set_mapped_constant_buffer(struct draw_context *draw, /** - * Tells the draw module whether to convert wide points (size != 1) - * into triangles. + * Tells the draw module to draw points with triangles if their size + * is greater than this threshold. */ void -draw_convert_wide_points(struct draw_context *draw, boolean enable) +draw_wide_point_threshold(struct draw_context *draw, float threshold) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->convert_wide_points = enable; + draw->wide_point_threshold = threshold; } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 6abced139b..03e3d3a66b 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -215,7 +215,7 @@ struct draw_context float plane[12][4]; unsigned nr_planes; - boolean convert_wide_points; /**< convert wide points to tris? */ + float wide_point_threshold; /**< convert pnts to tris if larger than this */ boolean convert_wide_lines; /**< convert wide lines to tris? */ boolean use_sse; diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index 3a19dd4cd7..ded7d10c08 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -52,6 +52,19 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) */ stage->next = next; + /* drawing wide lines? */ + wide_lines = (draw->rasterizer->line_width != 1.0 + && draw->convert_wide_lines + && !draw->rasterizer->line_smooth); + + /* drawing large points? */ + if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) + wide_points = FALSE; + else if (draw->rasterizer->point_size > draw->wide_point_threshold) + wide_points = TRUE; + else + wide_points = FALSE; + /* * NOTE: we build up the pipeline in end-to-start order. * @@ -69,16 +82,6 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) next = draw->pipeline.aapoint; } - /* drawing wide lines? */ - wide_lines = (draw->rasterizer->line_width != 1.0 - && draw->convert_wide_lines - && !draw->rasterizer->line_smooth); - - /* drawing large points? */ - wide_points = (draw->rasterizer->point_size != 1.0 - && draw->convert_wide_points - && !draw->pipeline.aapoint); - if (wide_lines || wide_points || draw->rasterizer->point_sprite) { diff --git a/src/gallium/auxiliary/draw/draw_wide_prims.c b/src/gallium/auxiliary/draw/draw_wide_prims.c index 1f8069bdca..d967810dd4 100644 --- a/src/gallium/auxiliary/draw/draw_wide_prims.c +++ b/src/gallium/auxiliary/draw/draw_wide_prims.c @@ -219,8 +219,8 @@ static void wide_point( struct draw_stage *stage, half_size = wide->half_point_size; } - left_adj = -half_size + 0.25f; - right_adj = half_size + 0.25f; + left_adj = -half_size; /* + 0.25f;*/ + right_adj = half_size; /* + 0.25f;*/ pos0[0] += left_adj; pos0[1] -= half_size; @@ -266,7 +266,8 @@ static void wide_first_point( struct draw_stage *stage, wide->half_point_size = 0.5f * draw->rasterizer->point_size; - if (draw->rasterizer->point_size != 1.0) { + /* XXX we won't know the real size if it's computed by the vertex shader! */ + if (draw->rasterizer->point_size > draw->wide_point_threshold) { stage->point = wide_point; } else { diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 2cdf3c75bf..6a884327e0 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -337,9 +337,6 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, draw_install_pstipple_stage(softpipe->draw, &softpipe->pipe); #endif - /* sp_prim_setup can do wide points (don't convert to quads) */ - draw_convert_wide_points(softpipe->draw, FALSE); - sp_init_surface_functions(softpipe); return &softpipe->pipe; -- cgit v1.2.3 From d4a4bed6638e0156324ff9b270f2248c4b5275bb Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 14:30:41 -0700 Subject: gallium: updated prototype (missed in prev commit) --- src/gallium/auxiliary/draw/draw_context.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index c25301f71d..d6685f479b 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -90,6 +90,8 @@ void draw_set_rasterizer_state( struct draw_context *draw, void draw_set_rasterize_stage( struct draw_context *draw, struct draw_stage *stage ); +void draw_wide_point_threshold(struct draw_context *draw, float threshold); + void draw_convert_wide_points(struct draw_context *draw, boolean enable); void draw_convert_wide_lines(struct draw_context *draw, boolean enable); -- cgit v1.2.3 From d5ab6507f7c368e86f653e76b759f2e71273469f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 14:34:19 -0700 Subject: gallium: #includes to silence warnings --- src/mesa/state_tracker/st_framebuffer.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 47d47daf2a..d46a9178b1 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -27,8 +27,10 @@ #include "main/imports.h" +#include "main/buffers.h" #include "main/context.h" #include "main/framebuffer.h" +#include "main/matrix.h" #include "main/renderbuffer.h" #include "st_public.h" #include "st_context.h" -- cgit v1.2.3 From 4da19dbcaa9f3e2d20fffd0145bf0bc756dd7542 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 19:31:22 -0700 Subject: gallium: remove pipe parameter from pipe_texture_reference() Added pipe field to pipe_texture (temporary, see comments). First step toward context-less texture creation... --- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 3 +-- src/gallium/drivers/i915simple/i915_state.c | 3 +-- src/gallium/drivers/i915simple/i915_texture.c | 1 + src/gallium/drivers/i965simple/brw_state.c | 3 +-- src/gallium/drivers/softpipe/sp_state_sampler.c | 2 +- src/gallium/drivers/softpipe/sp_texture.c | 1 + src/gallium/drivers/softpipe/sp_tile_cache.c | 2 +- src/gallium/include/pipe/p_inlines.h | 14 +++++++++++++- src/gallium/include/pipe/p_state.h | 5 +++++ src/mesa/state_tracker/st_cb_texture.c | 6 +++--- 10 files changed, 28 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 95bfc29fbe..075e0a0c47 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -242,8 +242,7 @@ cell_set_sampler_texture(struct pipe_context *pipe, draw_flush(cell->draw); - pipe_texture_reference(pipe, - (struct pipe_texture **) &cell->texture[sampler], + pipe_texture_reference((struct pipe_texture **) &cell->texture[sampler], texture); cell_update_texture_mapping(cell); diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 9df0e12540..27af46bea0 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -532,8 +532,7 @@ static void i915_set_sampler_texture(struct pipe_context *pipe, { struct i915_context *i915 = i915_context(pipe); - pipe_texture_reference(pipe, - (struct pipe_texture **) &i915->texture[sampler], + pipe_texture_reference((struct pipe_texture **) &i915->texture[sampler], texture); i915->dirty |= I915_NEW_TEXTURE; diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 1b415a94d4..7fcf4332e1 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -489,6 +489,7 @@ i915_texture_create(struct pipe_context *pipe, tex->base = *templat; tex->base.refcount = 1; + tex->base.pipe = pipe; if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : i915_miptree_layout(pipe, tex)) diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 2fc048bde0..7466fdc403 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -328,8 +328,7 @@ static void brw_set_sampler_texture(struct pipe_context *pipe, { struct brw_context *brw = brw_context(pipe); - pipe_texture_reference(pipe, - (struct pipe_texture **) &brw->attribs.Texture[unit], + pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[unit], texture); brw->state.dirty.brw |= BRW_NEW_TEXTURE; diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c index 18669a1c6e..1d6dd17d1d 100644 --- a/src/gallium/drivers/softpipe/sp_state_sampler.c +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -83,7 +83,7 @@ softpipe_set_sampler_texture(struct pipe_context *pipe, draw_flush(softpipe->draw); assert(unit < PIPE_MAX_SAMPLERS); - pipe_texture_reference(pipe, &softpipe->texture[unit], texture); + pipe_texture_reference(&softpipe->texture[unit], texture); sp_tile_cache_set_texture(pipe, softpipe->tex_cache[unit], texture); diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 6ba0f09e0a..a96447fa7a 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -90,6 +90,7 @@ softpipe_texture_create(struct pipe_context *pipe, spt->base = *templat; spt->base.refcount = 1; + spt->base.pipe = pipe; softpipe_texture_layout(spt); diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index da30dd6c48..0ff93c5527 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -220,7 +220,7 @@ sp_tile_cache_set_texture(struct pipe_context *pipe, assert(!tc->surface); - pipe_texture_reference(pipe, &tc->texture, texture); + pipe_texture_reference(&tc->texture, texture); if (tc->tex_surf_map) { pipe_surface_unmap(tc->tex_surf); diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h index de3fa555c5..21d4827e67 100644 --- a/src/gallium/include/pipe/p_inlines.h +++ b/src/gallium/include/pipe/p_inlines.h @@ -97,7 +97,7 @@ pipe_buffer_reference(struct pipe_winsys *winsys, * \sa pipe_surface_reference */ static INLINE void -pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, +pipe_texture_reference(struct pipe_texture **ptr, struct pipe_texture *pt) { assert(ptr); @@ -106,6 +106,7 @@ pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, pt->refcount++; if (*ptr) { + struct pipe_context *pipe = (*ptr)->pipe; pipe->texture_release(pipe, ptr); assert(!*ptr); } @@ -114,6 +115,17 @@ pipe_texture_reference(struct pipe_context *pipe, struct pipe_texture **ptr, } +static INLINE void +pipe_texture_release(struct pipe_texture **ptr) +{ + struct pipe_context *pipe; + assert(ptr); + pipe = (*ptr)->pipe; + pipe->texture_release(pipe, ptr); + *ptr = NULL; +} + + #ifdef __cplusplus } #endif diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index ddf3c1c79b..25a6fcc9e6 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -296,6 +296,11 @@ struct pipe_texture /* These are also refcounted: */ unsigned refcount; + + /**< pipe that created the texture + * XXX this'll change to a pipe_winsys (or pipe_screen)... + */ + struct pipe_context *pipe; }; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 778fb536bc..f5f956f6ea 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -573,7 +573,7 @@ st_TexImage(GLcontext * ctx, st_texture_match_image(stObj->pt, &stImage->base, stImage->face, stImage->level)) { - pipe_texture_reference(ctx->st->pipe, &stImage->pt, stObj->pt); + pipe_texture_reference(&stImage->pt, stObj->pt); assert(stImage->pt); } @@ -1371,7 +1371,7 @@ copy_image_data_to_texture(struct st_context *st, stImage->base.Data = NULL; } - pipe_texture_reference(st->pipe, &stImage->pt, stObj->pt); + pipe_texture_reference(&stImage->pt, stObj->pt); } @@ -1426,7 +1426,7 @@ st_finalize_texture(GLcontext *ctx, if (stObj->pt) ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); - pipe_texture_reference(ctx->st->pipe, &stObj->pt, firstImage->pt); + pipe_texture_reference(&stObj->pt, firstImage->pt); } if (firstImage->base.IsCompressed) { -- cgit v1.2.3 From dc2b6e2c33b44c1ffc0578b6bf52d05f7c68bb5c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 19:35:08 -0700 Subject: gallium: use pipe_texture_reference() instead of pipe->texture_release() --- src/mesa/state_tracker/st_cb_drawpixels.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 585cae3743..c837698980 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -973,7 +973,7 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, pt, stvp, stfp, color, GL_FALSE); - st->pipe->texture_release(st->pipe, &pt); + pipe_texture_reference(&pt, NULL); } } else { @@ -1124,7 +1124,7 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, pt, stvp, stfp, ctx->Current.RasterColor, GL_FALSE); - st->pipe->texture_release(st->pipe, &pt); + pipe_texture_reference(&pt, NULL); } } @@ -1282,7 +1282,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, pt, stvp, stfp, color, GL_TRUE); pipe_surface_reference(&psTex, NULL); - st->pipe->texture_release(st->pipe, &pt); + pipe_texture_reference(&pt, NULL); } -- cgit v1.2.3 From aa59a937ccf41609081d3f9a4973df5478979785 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 20:15:14 -0700 Subject: gallium: introduce 'pipe_screen' for context-independent functions This will allow creating textures before a rendering context exists, for example. Only implemented in i915 driver for now. i915pipe->texture_create() just dispatches through to the i915screen->texture_create() to avoid state tracker changes for now. --- src/gallium/drivers/i915simple/Makefile | 1 + src/gallium/drivers/i915simple/i915_context.c | 32 +----- src/gallium/drivers/i915simple/i915_context.h | 7 +- src/gallium/drivers/i915simple/i915_screen.c | 139 +++++++++++++++++++++++ src/gallium/drivers/i915simple/i915_screen.h | 60 ++++++++++ src/gallium/drivers/i915simple/i915_strings.c | 27 ++++- src/gallium/drivers/i915simple/i915_texture.c | 76 +++++++++---- src/gallium/drivers/i915simple/i915_texture.h | 7 +- src/gallium/drivers/i915simple/i915_winsys.h | 8 +- src/gallium/include/pipe/p_context.h | 3 + src/gallium/include/pipe/p_inlines.h | 11 +- src/gallium/include/pipe/p_screen.h | 98 ++++++++++++++++ src/gallium/include/pipe/p_state.h | 2 + src/gallium/winsys/dri/intel/intel_winsys_i915.c | 10 +- 14 files changed, 418 insertions(+), 63 deletions(-) create mode 100644 src/gallium/drivers/i915simple/i915_screen.c create mode 100644 src/gallium/drivers/i915simple/i915_screen.h create mode 100644 src/gallium/include/pipe/p_screen.h (limited to 'src') diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile index 2a75f5d57c..3400747a73 100644 --- a/src/gallium/drivers/i915simple/Makefile +++ b/src/gallium/drivers/i915simple/Makefile @@ -17,6 +17,7 @@ C_SOURCES = \ i915_state_derived.c \ i915_state_emit.c \ i915_state_sampler.c \ + i915_screen.c \ i915_strings.c \ i915_prim_emit.c \ i915_prim_vbuf.c \ diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index c3955bbd2d..8478cd76a5 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -234,33 +234,11 @@ static boolean i915_draw_arrays( struct pipe_context *pipe, -struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, - struct i915_winsys *i915_winsys, - unsigned pci_id ) +struct pipe_context *i915_create_context( struct pipe_screen *screen, + struct pipe_winsys *pipe_winsys, + struct i915_winsys *i915_winsys ) { struct i915_context *i915; - unsigned is_i945 = 0; - - switch (pci_id) { - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - break; - - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - is_i945 = 1; - break; - - default: - pipe_winsys->printf(pipe_winsys, - "%s: unknown pci id 0x%x, cannot create context\n", - __FUNCTION__, pci_id); - return NULL; - } i915 = CALLOC_STRUCT(i915_context); if (i915 == NULL) @@ -268,6 +246,7 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, i915->winsys = i915_winsys; i915->pipe.winsys = pipe_winsys; + i915->pipe.screen = screen; i915->pipe.destroy = i915_destroy; i915->pipe.is_format_supported = i915_is_format_supported; @@ -301,9 +280,6 @@ struct pipe_context *i915_create( struct pipe_winsys *pipe_winsys, draw_install_aaline_stage(i915->draw, &i915->pipe); draw_install_aapoint_stage(i915->draw, &i915->pipe); - i915->pci_id = pci_id; - i915->flags.is_i945 = is_i945; - i915->dirty = ~0; i915->hardware_dirty = ~0; diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 20cf7d3c3b..9fb85c122d 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -245,11 +245,6 @@ struct i915_context unsigned hardware_dirty; unsigned debug; - unsigned pci_id; - - struct { - unsigned is_i945:1; - } flags; }; /* A flag for each state_tracker state object: @@ -322,6 +317,8 @@ void i915_init_surface_functions( struct i915_context *i915 ); void i915_init_state_functions( struct i915_context *i915 ); void i915_init_flush_functions( struct i915_context *i915 ); void i915_init_string_functions( struct i915_context *i915 ); +void i915_init_screen_string_functions(struct pipe_screen *screen); + diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c new file mode 100644 index 0000000000..7e9d971d38 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -0,0 +1,139 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "i915_reg.h" +#include "i915_screen.h" +#include "i915_texture.h" + + +static const char * +i915_get_vendor( struct pipe_screen *pscreen ) +{ + return "Tungsten Graphics, Inc."; +} + + +static const char * +i915_get_name( struct pipe_screen *pscreen ) +{ + static char buffer[128]; + const char *chipset; + + switch (i915_screen(pscreen)->pci_id) { + case PCI_CHIP_I915_G: + chipset = "915G"; + break; + case PCI_CHIP_I915_GM: + chipset = "915GM"; + break; + case PCI_CHIP_I945_G: + chipset = "945G"; + break; + case PCI_CHIP_I945_GM: + chipset = "945GM"; + break; + case PCI_CHIP_I945_GME: + chipset = "945GME"; + break; + case PCI_CHIP_G33_G: + chipset = "G33"; + break; + case PCI_CHIP_Q35_G: + chipset = "Q35"; + break; + case PCI_CHIP_Q33_G: + chipset = "Q33"; + break; + default: + chipset = "unknown"; + break; + } + + sprintf(buffer, "i915 (chipset: %s)", chipset); + return buffer; +} + + + +static void +i915_destroy_screen( struct pipe_screen *screen ) +{ + FREE(screen); +} + + +/** + * Create a new i915_screen object + */ +struct pipe_screen * +i915_create_screen(struct pipe_winsys *winsys, uint pci_id) +{ + struct i915_screen *i915screen = CALLOC_STRUCT(i915_screen); + + if (!i915screen) + return NULL; + + switch (pci_id) { + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + i915screen->is_i945 = FALSE; + break; + + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + i915screen->is_i945 = TRUE; + break; + + default: + winsys->printf(winsys, + "%s: unknown pci id 0x%x, cannot create screen\n", + __FUNCTION__, pci_id); + return NULL; + } + + i915screen->pci_id = pci_id; + + i915screen->screen.winsys = winsys; + + i915screen->screen.destroy = i915_destroy_screen; + + i915screen->screen.get_name = i915_get_name; + i915screen->screen.get_vendor = i915_get_vendor; + + i915_init_screen_string_functions(&i915screen->screen); + i915_init_screen_texture_functions(&i915screen->screen); + + return &i915screen->screen; +} diff --git a/src/gallium/drivers/i915simple/i915_screen.h b/src/gallium/drivers/i915simple/i915_screen.h new file mode 100644 index 0000000000..8394ddbe89 --- /dev/null +++ b/src/gallium/drivers/i915simple/i915_screen.h @@ -0,0 +1,60 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef I915_SCREEN_H +#define I915_SCREEN_H + + +#include "pipe/p_screen.h" + + +/** + * Subclass of pipe_screen + */ +struct i915_screen +{ + struct pipe_screen screen; + + boolean is_i945; + uint pci_id; +}; + + +/** cast wrapper */ +static INLINE struct i915_screen * +i915_screen(struct pipe_screen *pscreen) +{ + return (struct i915_screen *) pscreen; +} + + +extern struct pipe_screen * +i915_create_screen(struct pipe_winsys *winsys, uint pci_id); + + +#endif /* I915_SCREEN_H */ diff --git a/src/gallium/drivers/i915simple/i915_strings.c b/src/gallium/drivers/i915simple/i915_strings.c index 301fedea19..ee62bb2e5d 100644 --- a/src/gallium/drivers/i915simple/i915_strings.c +++ b/src/gallium/drivers/i915simple/i915_strings.c @@ -26,21 +26,31 @@ **************************************************************************/ #include "i915_context.h" +#include "i915_screen.h" #include "i915_reg.h" +/** XXX temporary screen/pipe duplication here */ + + +static const char *i915_get_vendor_screen( struct pipe_screen *screen ) +{ + return "Tungsten Graphics, Inc."; +} + static const char *i915_get_vendor( struct pipe_context *pipe ) { return "Tungsten Graphics, Inc."; } -static const char *i915_get_name( struct pipe_context *pipe ) +static const char *i915_get_name_screen( struct pipe_screen *screen ) { + struct i915_screen *i915screen = i915_screen(screen); static char buffer[128]; const char *chipset; - switch (i915_context(pipe)->pci_id) { + switch (i915screen->pci_id) { case PCI_CHIP_I915_G: chipset = "915G"; break; @@ -75,9 +85,22 @@ static const char *i915_get_name( struct pipe_context *pipe ) } +static const char *i915_get_name( struct pipe_context *pipe ) +{ + return pipe->screen->get_name(pipe->screen); +} + + void i915_init_string_functions(struct i915_context *i915) { i915->pipe.get_name = i915_get_name; i915->pipe.get_vendor = i915_get_vendor; } + +void +i915_init_screen_string_functions(struct pipe_screen *screen) +{ + screen->get_name = i915_get_name_screen; + screen->get_vendor = i915_get_vendor_screen; +} diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 7fcf4332e1..3c9509dee3 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -40,6 +40,7 @@ #include "i915_context.h" #include "i915_texture.h" #include "i915_debug.h" +#include "i915_screen.h" static unsigned minify( unsigned d ) @@ -187,7 +188,7 @@ static const int step_offsets[6][2] = { static boolean -i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) +i915_miptree_layout(struct i915_texture * tex) { struct pipe_texture *pt = &tex->base; unsigned level; @@ -311,7 +312,7 @@ i915_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) static boolean -i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) +i945_miptree_layout(struct i915_texture * tex) { struct pipe_texture *pt = &tex->base; unsigned level; @@ -479,24 +480,26 @@ i945_miptree_layout(struct pipe_context *pipe, struct i915_texture * tex) static struct pipe_texture * -i915_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat) +i915_texture_create_screen(struct pipe_screen *screen, + const struct pipe_texture *templat) { struct i915_texture *tex = CALLOC_STRUCT(i915_texture); if (tex) { - struct i915_context *i915 = i915_context(pipe); + struct i915_screen *i915screen = i915_screen(screen); + struct pipe_winsys *ws = screen->winsys; tex->base = *templat; tex->base.refcount = 1; - tex->base.pipe = pipe; + tex->base.pipe = NULL; + tex->base.screen = screen; - if (i915->flags.is_i945 ? i945_miptree_layout(pipe, tex) : - i915_miptree_layout(pipe, tex)) - tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, - PIPE_BUFFER_USAGE_PIXEL, - tex->pitch * tex->base.cpp * - tex->total_height); + if (i915screen->is_i945 ? i945_miptree_layout(tex) : + i915_miptree_layout(tex)) + tex->buffer = ws->buffer_create(ws, 64, + PIPE_BUFFER_USAGE_PIXEL, + tex->pitch * tex->base.cpp * + tex->total_height); if (!tex->buffer) { FREE(tex); @@ -508,8 +511,17 @@ i915_texture_create(struct pipe_context *pipe, } +static struct pipe_texture * +i915_texture_create(struct pipe_context *pipe, + const struct pipe_texture *templat) +{ + return pipe->screen->texture_create(pipe->screen, templat); +} + + static void -i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +i915_texture_release_screen(struct pipe_screen *screen, + struct pipe_texture **pt) { if (!*pt) return; @@ -526,7 +538,7 @@ i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); */ - pipe_buffer_reference(pipe->winsys, &tex->buffer, NULL); + pipe_buffer_reference(screen->winsys, &tex->buffer, NULL); for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) if (tex->image_offset[i]) @@ -538,6 +550,13 @@ i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } +static void +i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + i915_texture_release_screen(pipe->screen, pt); +} + + static void i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) { @@ -549,11 +568,12 @@ i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) * XXX note: same as code in sp_surface.c */ static struct pipe_surface * -i915_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) +i915_get_tex_surface_screen(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) { struct i915_texture *tex = (struct i915_texture *)pt; + struct pipe_winsys *ws = screen->winsys; struct pipe_surface *ps; unsigned offset; /* in bytes */ @@ -570,11 +590,11 @@ i915_get_tex_surface(struct pipe_context *pipe, assert(zslice == 0); } - ps = pipe->winsys->surface_alloc(pipe->winsys); + ps = ws->surface_alloc(ws); if (ps) { assert(ps->refcount); assert(ps->winsys); - pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + pipe_buffer_reference(ws, &ps->buffer, tex->buffer); ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; @@ -586,6 +606,14 @@ i915_get_tex_surface(struct pipe_context *pipe, } +static struct pipe_surface * +i915_get_tex_surface(struct pipe_context *pipe, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + return i915_get_tex_surface_screen(pipe->screen, pt, face, level, zslice); +} + void i915_init_texture_functions(struct i915_context *i915) @@ -595,3 +623,13 @@ i915_init_texture_functions(struct i915_context *i915) i915->pipe.texture_update = i915_texture_update; i915->pipe.get_tex_surface = i915_get_tex_surface; } + + + +void +i915_init_screen_texture_functions(struct pipe_screen *screen) +{ + screen->texture_create = i915_texture_create_screen; + screen->texture_release = i915_texture_release_screen; + screen->get_tex_surface = i915_get_tex_surface_screen; +} diff --git a/src/gallium/drivers/i915simple/i915_texture.h b/src/gallium/drivers/i915simple/i915_texture.h index 6d8d41178f..7225016a9f 100644 --- a/src/gallium/drivers/i915simple/i915_texture.h +++ b/src/gallium/drivers/i915simple/i915_texture.h @@ -28,11 +28,16 @@ #ifndef I915_TEXTURE_H #define I915_TEXTURE_H -struct pipe_context; +struct i915_context; +struct pipe_screen; extern void i915_init_texture_functions(struct i915_context *i915); +extern void +i915_init_screen_texture_functions(struct pipe_screen *screen); + + #endif /* I915_TEXTURE_H */ diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h index fe49710852..e6b0ac9c52 100644 --- a/src/gallium/drivers/i915simple/i915_winsys.h +++ b/src/gallium/drivers/i915simple/i915_winsys.h @@ -52,6 +52,7 @@ struct pipe_buffer; struct pipe_winsys; +struct pipe_screen; /** @@ -107,9 +108,8 @@ struct i915_winsys { #define I915_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) -struct pipe_context *i915_create( struct pipe_winsys *, - struct i915_winsys *, - unsigned pci_id ); - +struct pipe_context *i915_create_context( struct pipe_screen *, + struct pipe_winsys *, + struct i915_winsys * ); #endif diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index f69b52f5e3..93fcb1c3e9 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -36,6 +36,8 @@ extern "C" { #endif +struct pipe_screen; + struct pipe_state_cache; /* Opaque driver handles: @@ -51,6 +53,7 @@ struct pipe_query; */ struct pipe_context { struct pipe_winsys *winsys; + struct pipe_screen *screen; void *priv; /** context private data (for DRI for example) */ diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h index 21d4827e67..a7e97fcd7d 100644 --- a/src/gallium/include/pipe/p_inlines.h +++ b/src/gallium/include/pipe/p_inlines.h @@ -30,6 +30,7 @@ #include "p_context.h" #include "p_defines.h" +#include "p_screen.h" #include "p_winsys.h" @@ -107,7 +108,15 @@ pipe_texture_reference(struct pipe_texture **ptr, if (*ptr) { struct pipe_context *pipe = (*ptr)->pipe; - pipe->texture_release(pipe, ptr); + /* XXX temporary mess here */ + if (pipe) { + pipe->texture_release(pipe, ptr); + } + else { + struct pipe_screen *screen = (*ptr)->screen; + screen->texture_release(screen, ptr); + } + assert(!*ptr); } diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h new file mode 100644 index 0000000000..6be9a82b68 --- /dev/null +++ b/src/gallium/include/pipe/p_screen.h @@ -0,0 +1,98 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Screen, Adapter or GPU + * + * These are driver functions/facilities that are context independent. + */ + + +#ifndef P_SCREEN_H +#define P_SCREEN_H + + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" + + + +#ifdef __cplusplus +extern "C" { +#endif + + + +/** + * Gallium screen/adapter context. Basically everything + * hardware-specific that doesn't actually require a rendering + * context. + */ +struct pipe_screen { + struct pipe_winsys *winsys; + + void (*destroy)( struct pipe_screen * ); + + + /* + * Capability queries + */ + const char *(*get_name)( struct pipe_screen * ); + + const char *(*get_vendor)( struct pipe_screen * ); + + int (*get_param)( struct pipe_screen *, int param ); + + float (*get_paramf)( struct pipe_screen *, int param ); + + boolean (*is_format_supported)( struct pipe_screen *, + enum pipe_format format, + uint type ); + + + /* + * Texture functions + */ + struct pipe_texture * (*texture_create)(struct pipe_screen *, + const struct pipe_texture *templat); + + void (*texture_release)(struct pipe_screen *, + struct pipe_texture **pt); + + /** Get a surface which is a "view" into a texture */ + struct pipe_surface *(*get_tex_surface)(struct pipe_screen *, + struct pipe_texture *texture, + unsigned face, unsigned level, + unsigned zslice); +}; + + +#ifdef __cplusplus +} +#endif + +#endif /* P_SCREEN_H */ diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 25a6fcc9e6..bb4a6cb23e 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -63,6 +63,7 @@ extern "C" { /* fwd decls */ +struct pipe_screen; struct pipe_surface; struct pipe_winsys; @@ -301,6 +302,7 @@ struct pipe_texture * XXX this'll change to a pipe_winsys (or pipe_screen)... */ struct pipe_context *pipe; + struct pipe_screen *screen; }; diff --git a/src/gallium/winsys/dri/intel/intel_winsys_i915.c b/src/gallium/winsys/dri/intel/intel_winsys_i915.c index 0ed3890e93..2def1afc31 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_i915.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_i915.c @@ -40,6 +40,7 @@ #include "pipe/p_util.h" #include "i915simple/i915_winsys.h" +#include "i915simple/i915_screen.h" struct intel_i915_winsys { @@ -135,6 +136,7 @@ intel_create_i915simple( struct intel_context *intel, struct pipe_winsys *winsys ) { struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); + struct pipe_screen *screen; /* Fill in this struct with callbacks that i915simple will need to * communicate with the window system, buffer manager, etc. @@ -146,9 +148,11 @@ intel_create_i915simple( struct intel_context *intel, iws->winsys.batch_finish = intel_i915_batch_finish; iws->intel = intel; + screen = i915_create_screen(winsys, intel->intelScreen->deviceID); + /* Create the i915simple context: */ - return i915_create( winsys, - &iws->winsys, - intel->intelScreen->deviceID ); + return i915_create_context( screen, + winsys, + &iws->winsys ); } -- cgit v1.2.3 From 94047122571ac78717874bfb42598ab210cd7f80 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 26 Feb 2008 20:12:29 -0700 Subject: cell: insert a (disabled) call to spe_cpu_info_get() Found on the Cell devel forum, but doesn't appear to be available in SDK 2.1. --- src/gallium/drivers/cell/ppu/cell_context.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index 98c314f45c..5cbc922cb9 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -228,6 +228,9 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) * SPU stuff */ cell->num_spus = 6; + /* XXX is this in SDK 3.0 only? + cell->num_spus = spe_cpu_info_get(SPE_COUNT_PHYSICAL_SPES, -1); + */ cell_start_spus(cell); -- cgit v1.2.3 From 8902ce06e85f48a436ee3794c77f7abf59f56594 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 27 Feb 2008 14:06:07 +0900 Subject: gallium: Use stricter types. VC++ won't silently convert a pointer to a function with typed pointer arguments to one with void pointer arguments. --- src/gallium/auxiliary/cso_cache/cso_cache.h | 26 +++++++++++++------------- 1 file changed, 13 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h index 3b0fe100b8..44ee128a4a 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.h +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -88,44 +88,44 @@ struct cso_cache; struct cso_blend { struct pipe_blend_state state; - void *data; - void (*delete_state)(void *, void *); - void *context; + void *data; + void (*delete_state)(struct pipe_context *, void *); + struct pipe_context *context; }; struct cso_depth_stencil_alpha { struct pipe_depth_stencil_alpha_state state; void *data; - void (*delete_state)(void *, void *); - void *context; + void (*delete_state)(struct pipe_context *, void *); + struct pipe_context *context; }; struct cso_rasterizer { struct pipe_rasterizer_state state; void *data; - void (*delete_state)(void *, void *); - void *context; + void (*delete_state)(struct pipe_context *, void *); + struct pipe_context *context; }; struct cso_fragment_shader { struct pipe_shader_state state; void *data; - void (*delete_state)(void *, void *); - void *context; + void (*delete_state)(struct pipe_context *, void *); + struct pipe_context *context; }; struct cso_vertex_shader { struct pipe_shader_state state; void *data; - void (*delete_state)(void *, void *); - void *context; + void (*delete_state)(struct pipe_context *, void *); + struct pipe_context *context; }; struct cso_sampler { struct pipe_sampler_state state; void *data; - void (*delete_state)(void *, void *); - void *context; + void (*delete_state)(struct pipe_context *, void *); + struct pipe_context *context; }; -- cgit v1.2.3 From f81b7a6285455e838adb061dcca90036c9f99522 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 27 Feb 2008 15:59:09 +0900 Subject: gallium: update for new i915_screen.c file; fix some warnings. --- src/gallium/drivers/i915simple/SConscript | 1 + src/gallium/drivers/i915simple/i915_screen.c | 1 + src/gallium/drivers/i915simple/i915_screen.h | 9 +++++++++ src/gallium/drivers/i915simple/i915_winsys.h | 9 +++++++++ 4 files changed, 20 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/SConscript b/src/gallium/drivers/i915simple/SConscript index f5fb96b995..3e1beaea6d 100644 --- a/src/gallium/drivers/i915simple/SConscript +++ b/src/gallium/drivers/i915simple/SConscript @@ -15,6 +15,7 @@ i915simple = env.ConvenienceLibrary( 'i915_fpc_translate.c', 'i915_prim_emit.c', 'i915_prim_vbuf.c', + 'i915_screen.c', 'i915_state.c', 'i915_state_derived.c', 'i915_state_dynamic.c', diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 7e9d971d38..5630440a5a 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -30,6 +30,7 @@ #include "pipe/p_winsys.h" #include "i915_reg.h" +#include "i915_context.h" #include "i915_screen.h" #include "i915_texture.h" diff --git a/src/gallium/drivers/i915simple/i915_screen.h b/src/gallium/drivers/i915simple/i915_screen.h index 8394ddbe89..73b0ff05ce 100644 --- a/src/gallium/drivers/i915simple/i915_screen.h +++ b/src/gallium/drivers/i915simple/i915_screen.h @@ -33,6 +33,11 @@ #include "pipe/p_screen.h" +#ifdef __cplusplus +extern "C" { +#endif + + /** * Subclass of pipe_screen */ @@ -57,4 +62,8 @@ extern struct pipe_screen * i915_create_screen(struct pipe_winsys *winsys, uint pci_id); +#ifdef __cplusplus +} +#endif + #endif /* I915_SCREEN_H */ diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h index e6b0ac9c52..aea3003281 100644 --- a/src/gallium/drivers/i915simple/i915_winsys.h +++ b/src/gallium/drivers/i915simple/i915_winsys.h @@ -40,6 +40,11 @@ #include "pipe/p_defines.h" +#ifdef __cplusplus +extern "C" { +#endif + + /* Pipe drivers are (meant to be!) independent of both GL and the * window system. The window system provides a buffer manager and a * set of additional hooks for things like command buffer submission, @@ -112,4 +117,8 @@ struct pipe_context *i915_create_context( struct pipe_screen *, struct pipe_winsys *, struct i915_winsys * ); +#ifdef __cplusplus +} +#endif + #endif -- cgit v1.2.3 From 94c73d1bb9037a1eead1d5174f218b15852b52d7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 27 Feb 2008 16:23:50 +0900 Subject: gallium: Remove // comments. --- src/gallium/include/pipe/p_defines.h | 2 +- src/gallium/include/pipe/p_format.h | 2 +- src/gallium/include/pipe/p_shader_tokens.h | 12 ++++++------ 3 files changed, 8 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index d84ddbc27a..0c662d6517 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -152,7 +152,7 @@ enum pipe_texture_target { */ #define PIPE_TEX_FILTER_NEAREST 0 #define PIPE_TEX_FILTER_LINEAR 1 -//#define PIPE_TEX_FILTER_ANISO 2 +/* #define PIPE_TEX_FILTER_ANISO 2 */ #define PIPE_TEX_COMPARE_NONE 0 diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 561d2e5921..f90087b3c9 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -28,7 +28,7 @@ #ifndef PIPE_FORMAT_H #define PIPE_FORMAT_H -#include // for sprintf +#include /* for sprintf */ #include "p_compiler.h" #include "p_debug.h" diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 1806877f6c..0a6145a6bf 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -1,9 +1,9 @@ #if !defined TGSI_TOKEN_H #define TGSI_TOKEN_H -#if defined __cplusplus +#ifdef __cplusplus extern "C" { -#endif // defined __cplusplus +#endif #include "p_compiler.h" @@ -802,9 +802,9 @@ struct tgsi_dst_register_ext_predicate }; -#if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus +#ifdef __cplusplus +} +#endif -#endif // !defined TGSI_TOKEN_H +#endif /* TGSI_TOKEN_H */ -- cgit v1.2.3 From 9a8a5d7c2fe7f32c8d15bc0a77f86e1f2f995ffe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 27 Feb 2008 16:42:15 +0900 Subject: gallium: Replace // comments. --- src/gallium/auxiliary/gallivm/gallivm.h | 2 +- src/gallium/auxiliary/gallivm/gallivm_p.h | 6 +++--- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2 +- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h | 8 ++++---- src/gallium/auxiliary/tgsi/util/tgsi_build.h | 8 ++++---- src/gallium/auxiliary/tgsi/util/tgsi_dump.h | 8 ++++---- src/gallium/auxiliary/tgsi/util/tgsi_parse.h | 8 ++++---- src/gallium/auxiliary/tgsi/util/tgsi_util.h | 8 ++++---- 8 files changed, 25 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/gallivm.h b/src/gallium/auxiliary/gallivm/gallivm.h index 92da4bca7f..57912a952f 100644 --- a/src/gallium/auxiliary/gallivm/gallivm.h +++ b/src/gallium/auxiliary/gallivm/gallivm.h @@ -97,7 +97,7 @@ void gallivm_cpu_engine_delete(struct gallivm_cpu_engine *ee); #endif /* MESA_LLVM */ #if defined __cplusplus -} // extern "C" +} #endif #endif diff --git a/src/gallium/auxiliary/gallivm/gallivm_p.h b/src/gallium/auxiliary/gallivm/gallivm_p.h index cfe7b1901b..ebf3e11cd5 100644 --- a/src/gallium/auxiliary/gallivm/gallivm_p.h +++ b/src/gallium/auxiliary/gallivm/gallivm_p.h @@ -32,7 +32,7 @@ struct gallivm_ir { int num_components; int num_consts; - //FIXME: this might not be enough for some shaders + /* FIXME: this might not be enough for some shaders */ struct gallivm_interpolate interpolators[32*4]; int num_interp; }; @@ -46,7 +46,7 @@ struct gallivm_prog { int num_consts; - //FIXME: this might not be enough for some shaders + /* FIXME: this might not be enough for some shaders */ struct gallivm_interpolate interpolators[32*4]; int num_interp; }; @@ -104,7 +104,7 @@ static INLINE int gallivm_w_swizzle(int swizzle) #endif /* MESA_LLVM */ #if defined __cplusplus -} // extern "C" +} #endif #endif diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 779b901f2b..57ccd86be4 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -1268,7 +1268,7 @@ emit_store( break; case TGSI_SAT_ZERO_ONE: -// assert( 0 ); + /* assert( 0 ); */ break; case TGSI_SAT_MINUS_PLUS_ONE: diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h index 9bee371766..63b8ef3911 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h @@ -3,7 +3,7 @@ #if defined __cplusplus extern "C" { -#endif // defined __cplusplus +#endif struct tgsi_token; struct x86_function; @@ -19,8 +19,8 @@ tgsi_emit_sse2_fs( struct x86_function *function ); #if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus +} +#endif -#endif // !defined TGSI_SSE2_H +#endif /* TGSI_SSE2_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.h b/src/gallium/auxiliary/tgsi/util/tgsi_build.h index 116c78abf3..607860e7fc 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.h @@ -3,7 +3,7 @@ #if defined __cplusplus extern "C" { -#endif // defined __cplusplus +#endif /* * version @@ -313,8 +313,8 @@ tgsi_build_dst_register_ext_modulate( struct tgsi_header *header ); #if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus +} +#endif -#endif // !defined TGSI_BUILD_H +#endif /* TGSI_BUILD_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h index 1adc9db251..b983b38226 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h @@ -3,7 +3,7 @@ #if defined __cplusplus extern "C" { -#endif // defined __cplusplus +#endif #define TGSI_DUMP_VERBOSE 1 #define TGSI_DUMP_NO_IGNORED 2 @@ -21,8 +21,8 @@ tgsi_dump_str( unsigned flags ); #if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus +} +#endif -#endif // !defined TGSI_DUMP_H +#endif /* TGSI_DUMP_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h index 9372da8d5d..5ccb5bfcf6 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h @@ -3,7 +3,7 @@ #if defined __cplusplus extern "C" { -#endif // defined __cplusplus +#endif struct tgsi_full_version { @@ -114,8 +114,8 @@ tgsi_parse_token( struct tgsi_parse_context *ctx ); #if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus +} +#endif -#endif // !defined TGSI_PARSE_H +#endif /* TGSI_PARSE_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_util.h b/src/gallium/auxiliary/tgsi/util/tgsi_util.h index ef14446f0e..45f5f0be0e 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_util.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_util.h @@ -3,7 +3,7 @@ #if defined __cplusplus extern "C" { -#endif // defined __cplusplus +#endif void * tgsi_align_128bit( @@ -63,8 +63,8 @@ tgsi_util_set_full_src_register_sign_mode( unsigned sign_mode ); #if defined __cplusplus -} // extern "C" -#endif // defined __cplusplus +} +#endif -#endif // !defined TGSI_UTIL_H +#endif /* TGSI_UTIL_H */ -- cgit v1.2.3 From d6229d7f1fda03d3c73998505b0facf6e3d5b882 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 27 Feb 2008 18:39:57 +0900 Subject: gallium: Make headers C++ friendly. --- src/gallium/auxiliary/pipebuffer/pb_buffer.h | 11 ++++++++++- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h | 11 ++++++++++- src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 11 ++++++++++- 3 files changed, 30 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h index f5b5f4052f..4b09c80b2a 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h @@ -37,7 +37,7 @@ * There is no obligation of a winsys driver to use this library. And a pipe * driver should be completly agnostic about it. * - * \author José Fonseca + * \author Jos� Fonseca */ #ifndef PB_BUFFER_H_ @@ -50,6 +50,11 @@ #include "pipe/p_inlines.h" +#ifdef __cplusplus +extern "C" { +#endif + + struct pb_vtbl; /** @@ -200,4 +205,8 @@ void pb_init_winsys(struct pipe_winsys *winsys); +#ifdef __cplusplus +} +#endif + #endif /*PB_BUFFER_H_*/ diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h index c40b9c75e1..50d5891bdb 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h @@ -44,7 +44,7 @@ * Between the handle's destruction, and the fence signalling, the buffer is * stored in a fenced buffer list. * - * \author José Fonseca + * \author José Fonseca */ #ifndef PB_BUFFER_FENCED_H_ @@ -54,6 +54,11 @@ #include "pipe/p_debug.h" +#ifdef __cplusplus +extern "C" { +#endif + + struct pipe_winsys; struct pipe_buffer; struct pipe_fence_handle; @@ -114,4 +119,8 @@ buffer_fence(struct pb_buffer *buf, struct pipe_fence_handle *fence); +#ifdef __cplusplus +} +#endif + #endif /*PB_BUFFER_FENCED_H_*/ diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index 1ddf784c97..0cf8e92e37 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -43,7 +43,7 @@ * - the fenced buffer manager, which will delay buffer destruction until the * the moment the card finishing processing it. * - * \author José Fonseca + * \author José Fonseca */ #ifndef PB_BUFMGR_H_ @@ -53,6 +53,11 @@ #include +#ifdef __cplusplus +extern "C" { +#endif + + struct pb_desc; struct pipe_buffer; struct pipe_winsys; @@ -123,4 +128,8 @@ fenced_bufmgr_create(struct pb_manager *provider, struct pipe_winsys *winsys); +#ifdef __cplusplus +} +#endif + #endif /*PB_BUFMGR_H_*/ -- cgit v1.2.3 From e49905b57c8a8691c093dd5dd55be7f087517416 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 27 Feb 2008 19:33:57 +0900 Subject: mesa: Add copyright headers. --- src/mesa/drivers/dri/common/dri_util.c | 26 ++++++++++++++++++++- src/mesa/drivers/dri/common/dri_util.h | 42 ++++++++++++++++------------------ 2 files changed, 45 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c index 6e8a5b5218..fff79c36ad 100644 --- a/src/mesa/drivers/dri/common/dri_util.c +++ b/src/mesa/drivers/dri/common/dri_util.c @@ -1,4 +1,28 @@ -/* $XFree86: xc/lib/GL/dri/dri_util.c,v 1.7 2003/04/28 17:01:25 dawes Exp $ */ +/* + * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + /** * \file dri_util.c * DRI utility functions. diff --git a/src/mesa/drivers/dri/common/dri_util.h b/src/mesa/drivers/dri/common/dri_util.h index 539d28d114..027cb7f461 100644 --- a/src/mesa/drivers/dri/common/dri_util.h +++ b/src/mesa/drivers/dri/common/dri_util.h @@ -1,25 +1,3 @@ -/* $XFree86: xc/lib/GL/dri/dri_util.h,v 1.1 2002/02/22 21:32:52 dawes Exp $ */ -/** - * \file dri_util.h - * DRI utility functions definitions. - * - * This module acts as glue between GLX and the actual hardware driver. A DRI - * driver doesn't really \e have to use any of this - it's optional. But, some - * useful stuff is done here that otherwise would have to be duplicated in most - * drivers. - * - * Basically, these utility functions take care of some of the dirty details of - * screen initialization, context creation, context binding, DRM setup, etc. - * - * These functions are compiled into each DRI driver so libGL.so knows nothing - * about them. - * - * \sa dri_util.c. - * - * \author Kevin E. Martin - * \author Brian Paul - */ - /* * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. * All Rights Reserved. @@ -45,6 +23,26 @@ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/** + * \file dri_util.h + * DRI utility functions definitions. + * + * This module acts as glue between GLX and the actual hardware driver. A DRI + * driver doesn't really \e have to use any of this - it's optional. But, some + * useful stuff is done here that otherwise would have to be duplicated in most + * drivers. + * + * Basically, these utility functions take care of some of the dirty details of + * screen initialization, context creation, context binding, DRM setup, etc. + * + * These functions are compiled into each DRI driver so libGL.so knows nothing + * about them. + * + * \sa dri_util.c. + * + * \author Kevin E. Martin + * \author Brian Paul + */ #ifndef _DRI_UTIL_H_ #define _DRI_UTIL_H_ -- cgit v1.2.3 From 8d94dabad8400cf65363d107a0605a013b542fd1 Mon Sep 17 00:00:00 2001 From: Jerome Glisse Date: Wed, 27 Feb 2008 07:42:48 -0700 Subject: mesa: set input read only on success --- src/mesa/shader/arbprogparse.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index d662e0b8b5..b6b3c88b14 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -1576,8 +1576,9 @@ parse_attrib_binding(GLcontext * ctx, const GLubyte ** inst, if (err) { program_error(ctx, Program->Position, "Bad attribute binding"); } - - Program->Base.InputsRead |= (1 << *inputReg); + else { + Program->Base.InputsRead |= (1 << *inputReg); + } return err; } -- cgit v1.2.3 From e8c0162fa0c4720aea612a617cbd02b83590763c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 08:58:12 -0700 Subject: gallium: softpipe_init_texture_funcs(), make texture func static --- src/gallium/drivers/softpipe/sp_context.c | 7 +------ src/gallium/drivers/softpipe/sp_texture.c | 19 +++++++++++++++---- src/gallium/drivers/softpipe/sp_texture.h | 16 +--------------- 3 files changed, 17 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 6a884327e0..78acf51433 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -279,12 +279,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->pipe.flush = softpipe_flush; softpipe_init_query_funcs( softpipe ); - - /* textures */ - softpipe->pipe.texture_create = softpipe_texture_create; - softpipe->pipe.texture_release = softpipe_texture_release; - softpipe->pipe.texture_update = softpipe_texture_update; - softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; + softpipe_init_texture_funcs( softpipe ); /* * Alloc caches for accessing drawing surfaces and textures. diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index a96447fa7a..1594521961 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -80,7 +80,7 @@ softpipe_texture_layout(struct softpipe_texture * spt) } -struct pipe_texture * +static struct pipe_texture * softpipe_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) { @@ -108,7 +108,7 @@ softpipe_texture_create(struct pipe_context *pipe, } -void +static void softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) { if (!*pt) @@ -133,7 +133,7 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) } -void +static void softpipe_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) { @@ -150,7 +150,7 @@ softpipe_texture_update(struct pipe_context *pipe, /** * Called via pipe->get_tex_surface() */ -struct pipe_surface * +static struct pipe_surface * softpipe_get_tex_surface(struct pipe_context *pipe, struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice) @@ -183,3 +183,14 @@ softpipe_get_tex_surface(struct pipe_context *pipe, } return ps; } + + + +void +softpipe_init_texture_funcs( struct softpipe_context *softpipe ) +{ + softpipe->pipe.texture_create = softpipe_texture_create; + softpipe->pipe.texture_release = softpipe_texture_release; + softpipe->pipe.texture_update = softpipe_texture_update; + softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; +} diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h index 50fc100427..01b1e28ec5 100644 --- a/src/gallium/drivers/softpipe/sp_texture.h +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -54,22 +54,8 @@ softpipe_texture(struct pipe_texture *pt) } - -extern struct pipe_texture * -softpipe_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat); - extern void -softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt); - -extern void -softpipe_texture_update(struct pipe_context *pipe, - struct pipe_texture *texture); - -extern struct pipe_surface * -softpipe_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice); +softpipe_init_texture_funcs( struct softpipe_context *softpipe ); #endif /* SP_TEXTURE */ -- cgit v1.2.3 From 31358282d4bd5a9708dba7be059dcff02233b4e1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 09:15:15 -0700 Subject: gallium: better debug messages --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 57ccd86be4..0da3b0bdb7 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -2268,7 +2268,7 @@ tgsi_emit_sse2( &parse.FullToken.FullInstruction ); if (!ok) { - debug_printf("failed to translate tgsi opcode %d\n", + debug_printf("failed to translate tgsi opcode %d to SSE\n", parse.FullToken.FullInstruction.Instruction.Opcode ); } break; @@ -2276,7 +2276,7 @@ tgsi_emit_sse2( case TGSI_TOKEN_TYPE_IMMEDIATE: /* XXX implement this */ ok = 0; - debug_printf("failed to emit immediate value\n"); + debug_printf("failed to emit immediate value to SSE\n"); break; default: @@ -2358,7 +2358,7 @@ tgsi_emit_sse2_fs( &parse.FullToken.FullInstruction ); if (!ok) { - debug_printf("failed to translate tgsi opcode %d\n", + debug_printf("failed to translate tgsi opcode %d to SSE\n", parse.FullToken.FullInstruction.Instruction.Opcode ); } break; @@ -2366,7 +2366,7 @@ tgsi_emit_sse2_fs( case TGSI_TOKEN_TYPE_IMMEDIATE: /* XXX implement this */ ok = 0; - debug_printf("failed to emit immediate value\n"); + debug_printf("failed to emit immediate value to SSE\n"); break; default: -- cgit v1.2.3 From ef6c82b0c13573df1aab7acd6f4f9ef9076f421f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 09:46:10 -0700 Subject: gallium/i915: remove some redundant code --- src/gallium/drivers/i915simple/i915_context.h | 1 - src/gallium/drivers/i915simple/i915_screen.c | 1 - src/gallium/drivers/i915simple/i915_strings.c | 55 +-------------------------- 3 files changed, 1 insertion(+), 56 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 9fb85c122d..6401112f83 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -317,7 +317,6 @@ void i915_init_surface_functions( struct i915_context *i915 ); void i915_init_state_functions( struct i915_context *i915 ); void i915_init_flush_functions( struct i915_context *i915 ); void i915_init_string_functions( struct i915_context *i915 ); -void i915_init_screen_string_functions(struct pipe_screen *screen); diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 5630440a5a..f44ff43c99 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -133,7 +133,6 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id) i915screen->screen.get_name = i915_get_name; i915screen->screen.get_vendor = i915_get_vendor; - i915_init_screen_string_functions(&i915screen->screen); i915_init_screen_texture_functions(&i915screen->screen); return &i915screen->screen; diff --git a/src/gallium/drivers/i915simple/i915_strings.c b/src/gallium/drivers/i915simple/i915_strings.c index ee62bb2e5d..14da3bcd72 100644 --- a/src/gallium/drivers/i915simple/i915_strings.c +++ b/src/gallium/drivers/i915simple/i915_strings.c @@ -33,55 +33,9 @@ /** XXX temporary screen/pipe duplication here */ -static const char *i915_get_vendor_screen( struct pipe_screen *screen ) -{ - return "Tungsten Graphics, Inc."; -} - static const char *i915_get_vendor( struct pipe_context *pipe ) { - return "Tungsten Graphics, Inc."; -} - - -static const char *i915_get_name_screen( struct pipe_screen *screen ) -{ - struct i915_screen *i915screen = i915_screen(screen); - static char buffer[128]; - const char *chipset; - - switch (i915screen->pci_id) { - case PCI_CHIP_I915_G: - chipset = "915G"; - break; - case PCI_CHIP_I915_GM: - chipset = "915GM"; - break; - case PCI_CHIP_I945_G: - chipset = "945G"; - break; - case PCI_CHIP_I945_GM: - chipset = "945GM"; - break; - case PCI_CHIP_I945_GME: - chipset = "945GME"; - break; - case PCI_CHIP_G33_G: - chipset = "G33"; - break; - case PCI_CHIP_Q35_G: - chipset = "Q35"; - break; - case PCI_CHIP_Q33_G: - chipset = "Q33"; - break; - default: - chipset = "unknown"; - break; - } - - sprintf(buffer, "i915 (chipset: %s)", chipset); - return buffer; + return pipe->screen->get_vendor(pipe->screen); } @@ -97,10 +51,3 @@ i915_init_string_functions(struct i915_context *i915) i915->pipe.get_name = i915_get_name; i915->pipe.get_vendor = i915_get_vendor; } - -void -i915_init_screen_string_functions(struct pipe_screen *screen) -{ - screen->get_name = i915_get_name_screen; - screen->get_vendor = i915_get_vendor_screen; -} -- cgit v1.2.3 From 4f36cf5858a7e53181c3578685675e15fbfcbb82 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 09:47:46 -0700 Subject: gallium: implement pipe_screen for softpipe driver --- src/gallium/drivers/softpipe/Makefile | 1 + src/gallium/drivers/softpipe/sp_context.c | 64 ++-------- src/gallium/drivers/softpipe/sp_screen.c | 141 +++++++++++++++++++++ src/gallium/drivers/softpipe/sp_texture.c | 57 +++++++-- src/gallium/drivers/softpipe/sp_texture.h | 10 +- src/gallium/drivers/softpipe/sp_winsys.h | 7 +- .../winsys/dri/intel/intel_winsys_softpipe.c | 3 +- src/gallium/winsys/xlib/xm_winsys.c | 3 +- 8 files changed, 220 insertions(+), 66 deletions(-) create mode 100644 src/gallium/drivers/softpipe/sp_screen.c (limited to 'src') diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile index 539ffb77f5..f32db35d58 100644 --- a/src/gallium/drivers/softpipe/Makefile +++ b/src/gallium/drivers/softpipe/Makefile @@ -27,6 +27,7 @@ C_SOURCES = \ sp_quad_output.c \ sp_quad_stencil.c \ sp_quad_stipple.c \ + sp_screen.c \ sp_state_blend.c \ sp_state_clip.c \ sp_state_derived.c \ diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 78acf51433..4ac1719cbb 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -143,76 +143,31 @@ static void softpipe_destroy( struct pipe_context *pipe ) } +/* XXX these will go away shortly */ static const char *softpipe_get_name( struct pipe_context *pipe ) { - return "softpipe"; + return pipe->screen->get_name(pipe->screen); } static const char *softpipe_get_vendor( struct pipe_context *pipe ) { - return "Tungsten Graphics, Inc."; + return pipe->screen->get_vendor(pipe->screen); } static int softpipe_get_param(struct pipe_context *pipe, int param) { - switch (param) { - case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; - case PIPE_CAP_NPOT_TEXTURES: - return 1; - case PIPE_CAP_TWO_SIDED_STENCIL: - return 1; - case PIPE_CAP_GLSL: - return 1; - case PIPE_CAP_S3TC: - return 0; - case PIPE_CAP_ANISOTROPIC_FILTER: - return 0; - case PIPE_CAP_POINT_SPRITE: - return 1; - case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; - case PIPE_CAP_OCCLUSION_QUERY: - return 1; - case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 1; - case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 12; /* max 2Kx2K */ - case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: - return 8; /* max 128x128x128 */ - case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 12; /* max 2Kx2K */ - default: - return 0; - } + return pipe->screen->get_param(pipe->screen, param); } static float softpipe_get_paramf(struct pipe_context *pipe, int param) { - switch (param) { - case PIPE_CAP_MAX_LINE_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_LINE_WIDTH_AA: - return 255.0; /* arbitrary */ - - case PIPE_CAP_MAX_POINT_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_POINT_WIDTH_AA: - return 255.0; /* arbitrary */ - - case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 0.0; - - case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: - return 16.0; /* arbitrary */ - - default: - return 0; - } + return pipe->screen->get_paramf(pipe->screen, param); } -struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, - struct softpipe_winsys *softpipe_winsys ) +struct pipe_context * +softpipe_create( struct pipe_screen *screen, + struct pipe_winsys *pipe_winsys, + struct softpipe_winsys *softpipe_winsys ) { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); uint i; @@ -226,6 +181,7 @@ struct pipe_context *softpipe_create( struct pipe_winsys *pipe_winsys, softpipe->dump_fs = GETENV( "GALLIUM_DUMP_FS" ) != NULL; softpipe->pipe.winsys = pipe_winsys; + softpipe->pipe.screen = screen; softpipe->pipe.destroy = softpipe_destroy; /* queries */ diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c new file mode 100644 index 0000000000..cc3b962580 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -0,0 +1,141 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" +#include "pipe/p_screen.h" + +#include "sp_texture.h" +#include "sp_winsys.h" + + +static const char * +softpipe_get_vendor(struct pipe_screen *screen) +{ + return "Tungsten Graphics, Inc."; +} + + +static const char * +softpipe_get_name(struct pipe_screen *screen) +{ + return "softpipe"; +} + + +static int +softpipe_get_param(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 1; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 1; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 1; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 12; /* max 2Kx2K */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 12; /* max 2Kx2K */ + default: + return 0; + } +} + + +static float +softpipe_get_paramf(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 255.0; /* arbitrary */ + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; /* arbitrary */ + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 0.0; + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; /* arbitrary */ + default: + return 0; + } +} + + +static void +softpipe_destroy_screen( struct pipe_screen *screen ) +{ + FREE(screen); +} + + +/** + * Create a new pipe_screen object + * Note: we're not presently subclassing pipe_screen (no softpipe_screen). + */ +struct pipe_screen * +softpipe_create_screen(struct pipe_winsys *winsys) +{ + struct pipe_screen *screen = CALLOC_STRUCT(pipe_screen); + + if (!screen) + return NULL; + + screen->winsys = winsys; + + screen->destroy = softpipe_destroy_screen; + + screen->get_name = softpipe_get_name; + screen->get_vendor = softpipe_get_vendor; + screen->get_param = softpipe_get_param; + screen->get_paramf = softpipe_get_paramf; + + softpipe_init_screen_texture_funcs(screen); + + return screen; +} diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 1594521961..f0e8350a4a 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -80,23 +80,34 @@ softpipe_texture_layout(struct softpipe_texture * spt) } +/* XXX temporary */ static struct pipe_texture * softpipe_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) { + return pipe->screen->texture_create(pipe->screen, templat); +} + + +static struct pipe_texture * +softpipe_texture_create_screen(struct pipe_screen *screen, + const struct pipe_texture *templat) +{ + struct pipe_winsys *ws = screen->winsys; struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); if (!spt) return NULL; spt->base = *templat; spt->base.refcount = 1; - spt->base.pipe = pipe; + spt->base.pipe = NULL; + spt->base.screen = screen; softpipe_texture_layout(spt); - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); + spt->buffer = ws->buffer_create(ws, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); if (!spt->buffer) { FREE(spt); return NULL; @@ -108,8 +119,17 @@ softpipe_texture_create(struct pipe_context *pipe, } +/* XXX temporary */ static void softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + return pipe->screen->texture_release(pipe->screen, pt); +} + + +static void +softpipe_texture_release_screen(struct pipe_screen *screen, + struct pipe_texture **pt) { if (!*pt) return; @@ -125,7 +145,7 @@ softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); */ - pipe_buffer_reference(pipe->winsys, &spt->buffer, NULL); + pipe_buffer_reference(screen->winsys, &spt->buffer, NULL); FREE(spt); } @@ -150,21 +170,32 @@ softpipe_texture_update(struct pipe_context *pipe, /** * Called via pipe->get_tex_surface() */ +/* XXX temporary */ static struct pipe_surface * softpipe_get_tex_surface(struct pipe_context *pipe, struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice) { + return pipe->screen->get_tex_surface(pipe->screen, pt, face, level, zslice); +} + + +static struct pipe_surface * +softpipe_get_tex_surface_screen(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct pipe_winsys *ws = screen->winsys; struct softpipe_texture *spt = softpipe_texture(pt); struct pipe_surface *ps; assert(level <= pt->last_level); - ps = pipe->winsys->surface_alloc(pipe->winsys); + ps = ws->surface_alloc(ws); if (ps) { assert(ps->refcount); assert(ps->winsys); - pipe_buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); + pipe_buffer_reference(ws, &ps->buffer, spt->buffer); ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; @@ -176,7 +207,8 @@ softpipe_get_tex_surface(struct pipe_context *pipe, ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * (pt->compressed ? ps->height/4 : ps->height) * ps->width * ps->cpp; - } else { + } + else { assert(face == 0); assert(zslice == 0); } @@ -194,3 +226,12 @@ softpipe_init_texture_funcs( struct softpipe_context *softpipe ) softpipe->pipe.texture_update = softpipe_texture_update; softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; } + + +void +softpipe_init_screen_texture_funcs(struct pipe_screen *screen) +{ + screen->texture_create = softpipe_texture_create_screen; + screen->texture_release = softpipe_texture_release_screen; + screen->get_tex_surface = softpipe_get_tex_surface_screen; +} diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h index 01b1e28ec5..a7322144e6 100644 --- a/src/gallium/drivers/softpipe/sp_texture.h +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -29,8 +29,12 @@ #define SP_TEXTURE_H +#include "pipe/p_state.h" + + struct pipe_context; -struct pipe_texture; +struct pipe_screen; +struct softpipe_context; struct softpipe_texture @@ -58,4 +62,8 @@ extern void softpipe_init_texture_funcs( struct softpipe_context *softpipe ); +extern void +softpipe_init_screen_texture_funcs(struct pipe_screen *screen); + + #endif /* SP_TEXTURE */ diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h index d6b379f58c..80f5354808 100644 --- a/src/gallium/drivers/softpipe/sp_winsys.h +++ b/src/gallium/drivers/softpipe/sp_winsys.h @@ -50,8 +50,13 @@ struct pipe_winsys; struct pipe_context; -struct pipe_context *softpipe_create( struct pipe_winsys *, +struct pipe_context *softpipe_create( struct pipe_screen *, + struct pipe_winsys *, struct softpipe_winsys * ); +struct pipe_screen * +softpipe_create_screen(struct pipe_winsys *); + + #endif /* SP_WINSYS_H */ diff --git a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c index 9e483bdc9f..0bc2dc4002 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c @@ -68,6 +68,7 @@ intel_create_softpipe( struct intel_context *intel, struct pipe_winsys *winsys ) { struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); + struct pipe_screen *screen = softpipe_create_screen(winsys); /* Fill in this struct with callbacks that softpipe will need to * communicate with the window system, buffer manager, etc. @@ -77,5 +78,5 @@ intel_create_softpipe( struct intel_context *intel, /* Create the softpipe context: */ - return softpipe_create( winsys, &isws->sws ); + return softpipe_create( screen, winsys, &isws->sws ); } diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 8da596d419..adebe4020a 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -457,7 +457,8 @@ xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) #endif { struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(pixelformat); - pipe = softpipe_create( pws, spws ); + struct pipe_screen *screen = softpipe_create_screen(pws); + pipe = softpipe_create( screen, pws, spws ); if (pipe) pipe->priv = xmesa; -- cgit v1.2.3 From f04736c8be5d30c510e1799ac0c8fa5173516513 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 09:55:17 -0700 Subject: Cell: implement pipe_screen for cell driver --- src/gallium/drivers/cell/ppu/Makefile | 1 + src/gallium/drivers/cell/ppu/cell_context.c | 68 +++---------- src/gallium/drivers/cell/ppu/cell_context.h | 2 +- src/gallium/drivers/cell/ppu/cell_screen.c | 147 ++++++++++++++++++++++++++++ src/gallium/drivers/cell/ppu/cell_screen.h | 41 ++++++++ src/gallium/drivers/cell/ppu/cell_texture.c | 49 ++++++++-- src/gallium/drivers/cell/ppu/cell_texture.h | 4 + src/gallium/winsys/xlib/xm_winsys.c | 6 +- 8 files changed, 254 insertions(+), 64 deletions(-) create mode 100644 src/gallium/drivers/cell/ppu/cell_screen.c create mode 100644 src/gallium/drivers/cell/ppu/cell_screen.h (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile index 164dde762c..d38fa6ce07 100644 --- a/src/gallium/drivers/cell/ppu/Makefile +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -29,6 +29,7 @@ SOURCES = \ cell_state_emit.c \ cell_state_shader.c \ cell_pipe_state.c \ + cell_screen.c \ cell_state_vertex.c \ cell_spu.c \ cell_surface.c \ diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index 5cbc922cb9..bb1838409f 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -37,9 +37,12 @@ #include "pipe/p_format.h" #include "pipe/p_util.h" #include "pipe/p_winsys.h" -#include "cell/common.h" +#include "pipe/p_screen.h" + #include "draw/draw_context.h" #include "draw/draw_private.h" + +#include "cell/common.h" #include "cell_clear.h" #include "cell_context.h" #include "cell_draw_arrays.h" @@ -76,73 +79,24 @@ cell_is_format_supported( struct pipe_context *pipe, static int cell_get_param(struct pipe_context *pipe, int param) { - switch (param) { - case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; - case PIPE_CAP_NPOT_TEXTURES: - return 1; - case PIPE_CAP_TWO_SIDED_STENCIL: - return 1; - case PIPE_CAP_GLSL: - return 1; - case PIPE_CAP_S3TC: - return 0; - case PIPE_CAP_ANISOTROPIC_FILTER: - return 0; - case PIPE_CAP_POINT_SPRITE: - return 1; - case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; - case PIPE_CAP_OCCLUSION_QUERY: - return 1; - case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 1; - case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 12; /* max 2Kx2K */ - case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: - return 8; /* max 128x128x128 */ - case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 12; /* max 2Kx2K */ - default: - return 0; - } + return pipe->screen->get_param(pipe->screen, param); } static float cell_get_paramf(struct pipe_context *pipe, int param) { - switch (param) { - case PIPE_CAP_MAX_LINE_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_LINE_WIDTH_AA: - return 255.0; /* arbitrary */ - - case PIPE_CAP_MAX_POINT_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_POINT_WIDTH_AA: - return 255.0; /* arbitrary */ - - case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 0.0; - - case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: - return 16.0; /* arbitrary */ - - default: - return 0; - } + return pipe->screen->get_paramf(pipe->screen, param); } - static const char * cell_get_name( struct pipe_context *pipe ) { - return "Cell"; + return pipe->screen->get_name(pipe->screen); } static const char * cell_get_vendor( struct pipe_context *pipe ) { - return "Tungsten Graphics, Inc."; + return pipe->screen->get_vendor(pipe->screen); } @@ -174,7 +128,8 @@ cell_draw_create(struct cell_context *cell) struct pipe_context * -cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) +cell_create_context(struct pipe_screen *screen, + struct cell_winsys *cws) { struct cell_context *cell; uint spu, buf; @@ -187,7 +142,8 @@ cell_create_context(struct pipe_winsys *winsys, struct cell_winsys *cws) memset(cell, 0, sizeof(*cell)); cell->winsys = cws; - cell->pipe.winsys = winsys; + cell->pipe.winsys = screen->winsys; + cell->pipe.screen = screen; cell->pipe.destroy = cell_destroy_context; /* queries */ diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 1433a4925f..bf27289f3f 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -128,7 +128,7 @@ cell_context(struct pipe_context *pipe) extern struct pipe_context * -cell_create_context(struct pipe_winsys *ws, struct cell_winsys *cws); +cell_create_context(struct pipe_screen *screen, struct cell_winsys *cws); extern void cell_vertex_shader_queue_flush(struct draw_context *draw); diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c new file mode 100644 index 0000000000..75255c0466 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_screen.c @@ -0,0 +1,147 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" +#include "pipe/p_screen.h" + +#include "cell_screen.h" +#include "cell_texture.h" +#include "cell_winsys.h" + + +static const char * +cell_get_vendor(struct pipe_screen *screen) +{ + return "Tungsten Graphics, Inc."; +} + + +static const char * +cell_get_name(struct pipe_screen *screen) +{ + return "Cell"; +} + + +static int +cell_get_param(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 1; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 1; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 1; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 12; /* max 2Kx2K */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 12; /* max 2Kx2K */ + default: + return 0; + } +} + + +static float +cell_get_paramf(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; /* arbitrary */ + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 0.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; /* arbitrary */ + + default: + return 0; + } +} + + +static void +cell_destroy_screen( struct pipe_screen *screen ) +{ + FREE(screen); +} + + +/** + * Create a new pipe_screen object + * Note: we're not presently subclassing pipe_screen (no cell_screen) but + * that would be the place to put SPU thread/context info... + */ +struct pipe_screen * +cell_create_screen(struct pipe_winsys *winsys) +{ + struct pipe_screen *screen = CALLOC_STRUCT(pipe_screen); + + if (!screen) + return NULL; + + screen->winsys = winsys; + + screen->destroy = cell_destroy_screen; + + screen->get_name = cell_get_name; + screen->get_vendor = cell_get_vendor; + screen->get_param = cell_get_param; + screen->get_paramf = cell_get_paramf; + + cell_init_screen_texture_funcs(screen); + + return screen; +} diff --git a/src/gallium/drivers/cell/ppu/cell_screen.h b/src/gallium/drivers/cell/ppu/cell_screen.h new file mode 100644 index 0000000000..c7e15889d6 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_screen.h @@ -0,0 +1,41 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CELL_SCREEN_H +#define CELL_SCREEN_H + + +struct pipe_screen; +struct pipe_winsys; + + +extern struct pipe_screen * +cell_create_screen(struct pipe_winsys *winsys); + + +#endif /* CELL_SCREEN_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index 0edefa5f05..e6398a85fa 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -83,18 +83,29 @@ static struct pipe_texture * cell_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) { + return pipe->screen->texture_create(pipe->screen, templat); + +} + +static struct pipe_texture * +cell_texture_create_screen(struct pipe_screen *screen, + const struct pipe_texture *templat) +{ + struct pipe_winsys *ws = screen->winsys; struct cell_texture *spt = CALLOC_STRUCT(cell_texture); if (!spt) return NULL; spt->base = *templat; spt->base.refcount = 1; + spt->base.pipe = NULL; + spt->base.screen = screen; cell_texture_layout(spt); - spt->buffer = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); + spt->buffer = ws->buffer_create(ws, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); if (!spt->buffer) { FREE(spt); @@ -107,6 +118,14 @@ cell_texture_create(struct pipe_context *pipe, static void cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + return pipe->screen->texture_release(pipe->screen, pt); +} + + +static void +cell_texture_release_screen(struct pipe_screen *screen, + struct pipe_texture **pt) { if (!*pt) return; @@ -122,7 +141,7 @@ cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); */ - pipe_buffer_reference(pipe->winsys, &spt->buffer, NULL); + pipe_buffer_reference(screen->winsys, &spt->buffer, NULL); FREE(spt); } @@ -146,14 +165,24 @@ cell_get_tex_surface(struct pipe_context *pipe, struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice) { + return pipe->screen->get_tex_surface(pipe->screen, pt, face, level, zslice); +} + + +static struct pipe_surface * +cell_get_tex_surface_screen(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct pipe_winsys *ws = screen->winsys; struct cell_texture *spt = cell_texture(pt); struct pipe_surface *ps; - ps = pipe->winsys->surface_alloc(pipe->winsys); + ps = ws->surface_alloc(ws); if (ps) { assert(ps->refcount); assert(ps->winsys); - pipe_buffer_reference(pipe->winsys, &ps->buffer, spt->buffer); + pipe_buffer_reference(ws, &ps->buffer, spt->buffer); ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; @@ -270,3 +299,11 @@ cell_init_texture_functions(struct cell_context *cell) cell->pipe.texture_update = cell_texture_update; cell->pipe.get_tex_surface = cell_get_tex_surface; } + +void +cell_init_screen_texture_funcs(struct pipe_screen *screen) +{ + screen->texture_create = cell_texture_create_screen; + screen->texture_release = cell_texture_release_screen; + screen->get_tex_surface = cell_get_tex_surface_screen; +} diff --git a/src/gallium/drivers/cell/ppu/cell_texture.h b/src/gallium/drivers/cell/ppu/cell_texture.h index 824fb3e20f..fcee069d05 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.h +++ b/src/gallium/drivers/cell/ppu/cell_texture.h @@ -68,4 +68,8 @@ extern void cell_init_texture_functions(struct cell_context *cell); +extern void +cell_init_screen_texture_funcs(struct pipe_screen *screen); + + #endif /* CELL_TEXTURE_H */ diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index adebe4020a..018cf01cd6 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -36,6 +36,8 @@ #include "glxheader.h" #include "xmesaP.h" +#undef ASSERT + #include "pipe/p_winsys.h" #include "pipe/p_format.h" #include "pipe/p_context.h" @@ -45,6 +47,7 @@ #ifdef GALLIUM_CELL #include "cell/ppu/cell_context.h" +#include "cell/ppu/cell_screen.h" #include "cell/ppu/cell_winsys.h" #else #define TILE_SIZE 32 /* avoid compilation errors */ @@ -448,7 +451,8 @@ xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) #ifdef GALLIUM_CELL if (!getenv("GALLIUM_NOCELL")) { struct cell_winsys *cws = cell_get_winsys(pixelformat); - pipe = cell_create_context(pws, cws); + struct pipe_screen *screen = cell_create_screen(pws); + pipe = cell_create_context(screen, cws); if (pipe) pipe->priv = xmesa; return pipe; -- cgit v1.2.3 From 17188e4d5a3707c134fc97976863f0d8e2f1f5ab Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 10:15:37 -0700 Subject: gallium/i915: hook up screen->get_param() --- src/gallium/drivers/i915simple/i915_context.c | 55 +++-------------------- src/gallium/drivers/i915simple/i915_screen.c | 63 +++++++++++++++++++++++++++ 2 files changed, 68 insertions(+), 50 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 8478cd76a5..e89db8a130 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -36,6 +36,7 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "pipe/p_util.h" +#include "pipe/p_screen.h" /** @@ -90,65 +91,19 @@ i915_is_format_supported( struct pipe_context *pipe, } +/* XXX temporary */ static int i915_get_param(struct pipe_context *pipe, int param) { - switch (param) { - case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; - case PIPE_CAP_NPOT_TEXTURES: - return 1; - case PIPE_CAP_TWO_SIDED_STENCIL: - return 1; - case PIPE_CAP_GLSL: - return 0; - case PIPE_CAP_S3TC: - return 0; - case PIPE_CAP_ANISOTROPIC_FILTER: - return 0; - case PIPE_CAP_POINT_SPRITE: - return 0; - case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; - case PIPE_CAP_OCCLUSION_QUERY: - return 0; - case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 1; - case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 11; /* max 1024x1024 */ - case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: - return 8; /* max 128x128x128 */ - case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 11; /* max 1024x1024 */ - default: - return 0; - } + return pipe->screen->get_param(pipe->screen, param); } +/* XXX temporary */ static float i915_get_paramf(struct pipe_context *pipe, int param) { - switch (param) { - case PIPE_CAP_MAX_LINE_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_LINE_WIDTH_AA: - return 7.5; - - case PIPE_CAP_MAX_POINT_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_POINT_WIDTH_AA: - return 255.0; - - case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 4.0; - - case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: - return 16.0; - - default: - return 0; - } + return pipe->screen->get_paramf(pipe->screen, param); } diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index f44ff43c99..b9b4d6b6fa 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -83,6 +83,67 @@ i915_get_name( struct pipe_screen *pscreen ) } +static int +i915_get_param(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 11; /* max 1024x1024 */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 11; /* max 1024x1024 */ + default: + return 0; + } +} + + +static float +i915_get_paramf(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 7.5; + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 4.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; + + default: + return 0; + } +} + static void i915_destroy_screen( struct pipe_screen *screen ) @@ -132,6 +193,8 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id) i915screen->screen.get_name = i915_get_name; i915screen->screen.get_vendor = i915_get_vendor; + i915screen->screen.get_param = i915_get_param; + i915screen->screen.get_paramf = i915_get_paramf; i915_init_screen_texture_functions(&i915screen->screen); -- cgit v1.2.3 From 03b5267f52d440b1b357918ed7de2ca948f314e1 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 10:17:57 -0700 Subject: gallium/i965: implement pipe_screen for i965 driver (untested) --- src/gallium/drivers/i965simple/Makefile | 1 + src/gallium/drivers/i965simple/brw_context.c | 70 ++-------- src/gallium/drivers/i965simple/brw_screen.c | 168 ++++++++++++++++++++++++ src/gallium/drivers/i965simple/brw_screen.h | 68 ++++++++++ src/gallium/drivers/i965simple/brw_strings.c | 29 +--- src/gallium/drivers/i965simple/brw_tex_layout.c | 58 ++++++-- src/gallium/drivers/i965simple/brw_tex_layout.h | 32 +++++ src/gallium/drivers/i965simple/brw_winsys.h | 4 +- src/gallium/winsys/xlib/xm_winsys_aub.c | 4 +- 9 files changed, 338 insertions(+), 96 deletions(-) create mode 100644 src/gallium/drivers/i965simple/brw_screen.c create mode 100644 src/gallium/drivers/i965simple/brw_screen.h (limited to 'src') diff --git a/src/gallium/drivers/i965simple/Makefile b/src/gallium/drivers/i965simple/Makefile index cc8580836c..7b0df0f850 100644 --- a/src/gallium/drivers/i965simple/Makefile +++ b/src/gallium/drivers/i965simple/Makefile @@ -6,6 +6,7 @@ LIBNAME = i965simple C_SOURCES = \ brw_blit.c \ brw_flush.c \ + brw_screen.c \ brw_strings.c \ brw_surface.c \ brw_cc.c \ diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c index 6fb840708e..d90b32d4b4 100644 --- a/src/gallium/drivers/i965simple/brw_context.c +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -40,15 +40,14 @@ #include "pipe/p_winsys.h" #include "pipe/p_context.h" #include "pipe/p_util.h" +#include "pipe/p_screen.h" -/*************************************** - * Mesa's Driver Functions - ***************************************/ #ifndef BRW_DEBUG int BRW_DEBUG = (0); #endif + static void brw_destroy(struct pipe_context *pipe) { struct brw_context *brw = brw_context(pipe); @@ -56,6 +55,7 @@ static void brw_destroy(struct pipe_context *pipe) FREE(brw); } + static void brw_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { @@ -74,64 +74,17 @@ static void brw_clear(struct pipe_context *pipe, struct pipe_surface *ps, static int brw_get_param(struct pipe_context *pipe, int param) { - switch (param) { - case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; - case PIPE_CAP_NPOT_TEXTURES: - return 1; - case PIPE_CAP_TWO_SIDED_STENCIL: - return 1; - case PIPE_CAP_GLSL: - return 0; - case PIPE_CAP_S3TC: - return 0; - case PIPE_CAP_ANISOTROPIC_FILTER: - return 0; - case PIPE_CAP_POINT_SPRITE: - return 0; - case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; - case PIPE_CAP_OCCLUSION_QUERY: - return 0; - case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 1; - case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: - return 11; /* max 1024x1024 */ - case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: - return 8; /* max 128x128x128 */ - case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: - return 11; /* max 1024x1024 */ - default: - return 0; - } + return pipe->screen->get_param(pipe->screen, param); } static float brw_get_paramf(struct pipe_context *pipe, int param) { - switch (param) { - case PIPE_CAP_MAX_LINE_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_LINE_WIDTH_AA: - return 7.5; - - case PIPE_CAP_MAX_POINT_WIDTH: - /* fall-through */ - case PIPE_CAP_MAX_POINT_WIDTH_AA: - return 255.0; - - case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: - return 4.0; - - case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: - return 16.0; - - default: - return 0; - } + return pipe->screen->get_paramf(pipe->screen, param); } + static boolean brw_is_format_supported( struct pipe_context *pipe, enum pipe_format format, uint type ) @@ -200,22 +153,23 @@ brw_is_format_supported( struct pipe_context *pipe, -struct pipe_context *brw_create(struct pipe_winsys *pipe_winsys, +struct pipe_context *brw_create(struct pipe_screen *screen, struct brw_winsys *brw_winsys, unsigned pci_id) { struct brw_context *brw; - pipe_winsys->printf(pipe_winsys, - "%s: creating brw_context with pci id 0x%x\n", - __FUNCTION__, pci_id); + screen->winsys->printf(screen->winsys, + "%s: creating brw_context with pci id 0x%x\n", + __FUNCTION__, pci_id); brw = CALLOC_STRUCT(brw_context); if (brw == NULL) return NULL; brw->winsys = brw_winsys; - brw->pipe.winsys = pipe_winsys; + brw->pipe.winsys = screen->winsys; + brw->pipe.screen = screen; brw->pipe.destroy = brw_destroy; brw->pipe.is_format_supported = brw_is_format_supported; diff --git a/src/gallium/drivers/i965simple/brw_screen.c b/src/gallium/drivers/i965simple/brw_screen.c new file mode 100644 index 0000000000..7afe070bd8 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_screen.c @@ -0,0 +1,168 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "brw_context.h" +#include "brw_screen.h" +#include "brw_tex_layout.h" + + +static const char * +brw_get_vendor( struct pipe_screen *screen ) +{ + return "Tungsten Graphics, Inc."; +} + + +static const char * +brw_get_name( struct pipe_screen *screen ) +{ + static char buffer[128]; + const char *chipset; + + switch (brw_screen(screen)->pci_id) { + case PCI_CHIP_I965_Q: + chipset = "Intel(R) 965Q"; + break; + case PCI_CHIP_I965_G: + case PCI_CHIP_I965_G_1: + chipset = "Intel(R) 965G"; + break; + case PCI_CHIP_I965_GM: + chipset = "Intel(R) 965GM"; + break; + case PCI_CHIP_I965_GME: + chipset = "Intel(R) 965GME/GLE"; + break; + default: + chipset = "unknown"; + break; + } + + sprintf(buffer, "i965 (chipset: %s)", chipset); + return buffer; +} + + +static int +brw_get_param(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: + return 8; + case PIPE_CAP_NPOT_TEXTURES: + return 1; + case PIPE_CAP_TWO_SIDED_STENCIL: + return 1; + case PIPE_CAP_GLSL: + return 0; + case PIPE_CAP_S3TC: + return 0; + case PIPE_CAP_ANISOTROPIC_FILTER: + return 0; + case PIPE_CAP_POINT_SPRITE: + return 0; + case PIPE_CAP_MAX_RENDER_TARGETS: + return 1; + case PIPE_CAP_OCCLUSION_QUERY: + return 0; + case PIPE_CAP_TEXTURE_SHADOW_MAP: + return 1; + case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: + return 11; /* max 1024x1024 */ + case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: + return 8; /* max 128x128x128 */ + case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS: + return 11; /* max 1024x1024 */ + default: + return 0; + } +} + + +static float +brw_get_paramf(struct pipe_screen *screen, int param) +{ + switch (param) { + case PIPE_CAP_MAX_LINE_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_LINE_WIDTH_AA: + return 7.5; + + case PIPE_CAP_MAX_POINT_WIDTH: + /* fall-through */ + case PIPE_CAP_MAX_POINT_WIDTH_AA: + return 255.0; + + case PIPE_CAP_MAX_TEXTURE_ANISOTROPY: + return 4.0; + + case PIPE_CAP_MAX_TEXTURE_LOD_BIAS: + return 16.0; + + default: + return 0; + } +} + + +static void +brw_destroy_screen( struct pipe_screen *screen ) +{ + FREE(screen); +} + + +/** + * Create a new brw_screen object + */ +struct pipe_screen * +brw_create_screen(struct pipe_winsys *winsys, uint pci_id) +{ + struct brw_screen *brwscreen = CALLOC_STRUCT(brw_screen); + + if (!brwscreen) + return NULL; + + brwscreen->pci_id = pci_id; + + brwscreen->screen.winsys = winsys; + + brwscreen->screen.destroy = brw_destroy_screen; + + brwscreen->screen.get_name = brw_get_name; + brwscreen->screen.get_vendor = brw_get_vendor; + brwscreen->screen.get_param = brw_get_param; + brwscreen->screen.get_paramf = brw_get_paramf; + + brw_init_screen_texture_funcs(&brwscreen->screen); + + return &brwscreen->screen; +} diff --git a/src/gallium/drivers/i965simple/brw_screen.h b/src/gallium/drivers/i965simple/brw_screen.h new file mode 100644 index 0000000000..d3c70387e6 --- /dev/null +++ b/src/gallium/drivers/i965simple/brw_screen.h @@ -0,0 +1,68 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef BRW_SCREEN_H +#define BRW_SCREEN_H + + +#include "pipe/p_screen.h" + + +#ifdef __cplusplus +extern "C" { +#endif + + +/** + * Subclass of pipe_screen + */ +struct brw_screen +{ + struct pipe_screen screen; + + uint pci_id; +}; + + +/** cast wrapper */ +static INLINE struct brw_screen * +brw_screen(struct pipe_screen *pscreen) +{ + return (struct brw_screen *) pscreen; +} + + +extern struct pipe_screen * +brw_create_screen(struct pipe_winsys *winsys, uint pci_id); + + +#ifdef __cplusplus +} +#endif + +#endif /* BRW_SCREEN_H */ diff --git a/src/gallium/drivers/i965simple/brw_strings.c b/src/gallium/drivers/i965simple/brw_strings.c index 3d9c50961f..a8f2ce582a 100644 --- a/src/gallium/drivers/i965simple/brw_strings.c +++ b/src/gallium/drivers/i965simple/brw_strings.c @@ -28,39 +28,18 @@ #include "brw_context.h" #include "brw_reg.h" +#include "pipe/p_screen.h" + static const char *brw_get_vendor( struct pipe_context *pipe ) { - return "Tungsten Graphics, Inc."; + return pipe->screen->get_vendor(pipe->screen); } static const char *brw_get_name( struct pipe_context *pipe ) { - static char buffer[128]; - const char *chipset; - - switch (brw_context(pipe)->pci_id) { - case PCI_CHIP_I965_Q: - chipset = "Intel(R) 965Q"; - break; - case PCI_CHIP_I965_G: - case PCI_CHIP_I965_G_1: - chipset = "Intel(R) 965G"; - break; - case PCI_CHIP_I965_GM: - chipset = "Intel(R) 965GM"; - break; - case PCI_CHIP_I965_GME: - chipset = "Intel(R) 965GME/GLE"; - break; - default: - chipset = "unknown"; - break; - } - - sprintf(buffer, "i965 (chipset: %s)", chipset); - return buffer; + return pipe->screen->get_name(pipe->screen); } diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 86ce3d0cc3..9753c50143 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -71,8 +71,6 @@ static unsigned minify( unsigned d ) } -static boolean brw_miptree_layout(struct pipe_context *, struct brw_texture *); - static void intel_miptree_set_image_offset(struct brw_texture *tex, unsigned level, unsigned img, @@ -199,7 +197,7 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) } } -static boolean brw_miptree_layout(struct pipe_context *pipe, struct brw_texture *tex) +static boolean brw_miptree_layout(struct brw_texture *tex) { struct pipe_texture *pt = &tex->base; /* XXX: these vary depending on image format: @@ -304,17 +302,26 @@ static struct pipe_texture * brw_texture_create(struct pipe_context *pipe, const struct pipe_texture *templat) { + return pipe->screen->texture_create(pipe->screen, templat); +} + + +static struct pipe_texture * +brw_texture_create_screen(struct pipe_screen *screen, + const struct pipe_texture *templat) +{ + struct pipe_winsys *ws = screen->winsys; struct brw_texture *tex = CALLOC_STRUCT(brw_texture); if (tex) { tex->base = *templat; tex->base.refcount = 1; - if (brw_miptree_layout(pipe, tex)) - tex->buffer = pipe->winsys->buffer_create(pipe->winsys, 64, - PIPE_BUFFER_USAGE_PIXEL, - tex->pitch * tex->base.cpp * - tex->total_height); + if (brw_miptree_layout(tex)) + tex->buffer = ws->buffer_create(ws, 64, + PIPE_BUFFER_USAGE_PIXEL, + tex->pitch * tex->base.cpp * + tex->total_height); if (!tex->buffer) { FREE(tex); @@ -328,6 +335,14 @@ brw_texture_create(struct pipe_context *pipe, static void brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) +{ + pipe->screen->texture_release(pipe->screen, pt); +} + + +static void +brw_texture_release_screen(struct pipe_screen *screen, + struct pipe_texture **pt) { if (!*pt) return; @@ -337,6 +352,7 @@ brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); */ if (--(*pt)->refcount <= 0) { + struct pipe_winsys *ws = screen->winsys; struct brw_texture *tex = (struct brw_texture *)*pt; uint i; @@ -344,7 +360,7 @@ brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) DBG("%s deleting %p\n", __FUNCTION__, (void *) tex); */ - pipe_buffer_reference(pipe->winsys, &tex->buffer, NULL); + pipe_buffer_reference(ws, &tex->buffer, NULL); for (i = 0; i < PIPE_MAX_TEXTURE_LEVELS; i++) if (tex->image_offset[i]) @@ -371,6 +387,16 @@ brw_get_tex_surface(struct pipe_context *pipe, struct pipe_texture *pt, unsigned face, unsigned level, unsigned zslice) { + return pipe->screen->get_tex_surface(pipe->screen, pt, face, level, zslice); +} + + +static struct pipe_surface * +brw_get_tex_surface_screen(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct pipe_winsys *ws = screen->winsys; struct brw_texture *tex = (struct brw_texture *)pt; struct pipe_surface *ps; unsigned offset; /* in bytes */ @@ -388,11 +414,11 @@ brw_get_tex_surface(struct pipe_context *pipe, assert(zslice == 0); } - ps = pipe->winsys->surface_alloc(pipe->winsys); + ps = ws->surface_alloc(ws); if (ps) { assert(ps->format); assert(ps->refcount); - pipe_buffer_reference(pipe->winsys, &ps->buffer, tex->buffer); + pipe_buffer_reference(ws, &ps->buffer, tex->buffer); ps->format = pt->format; ps->cpp = pt->cpp; ps->width = pt->width[level]; @@ -412,3 +438,13 @@ brw_init_texture_functions(struct brw_context *brw) brw->pipe.texture_update = brw_texture_update; brw->pipe.get_tex_surface = brw_get_tex_surface; } + + +void +brw_init_screen_texture_funcs(struct pipe_screen *screen) +{ + screen->texture_create = brw_texture_create_screen; + screen->texture_release = brw_texture_release_screen; + screen->get_tex_surface = brw_get_tex_surface_screen; +} + diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.h b/src/gallium/drivers/i965simple/brw_tex_layout.h index ed49baeef8..a6b6ba8146 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.h +++ b/src/gallium/drivers/i965simple/brw_tex_layout.h @@ -1,12 +1,44 @@ +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + + #ifndef BRW_TEX_LAYOUT_H #define BRW_TEX_LAYOUT_H struct brw_context; +struct pipe_screen; extern void brw_init_texture_functions(struct brw_context *brw); +extern void +brw_init_screen_texture_funcs(struct pipe_screen *screen); + #endif diff --git a/src/gallium/drivers/i965simple/brw_winsys.h b/src/gallium/drivers/i965simple/brw_winsys.h index 3523a58614..b67bd73750 100644 --- a/src/gallium/drivers/i965simple/brw_winsys.h +++ b/src/gallium/drivers/i965simple/brw_winsys.h @@ -53,6 +53,8 @@ struct pipe_buffer; struct pipe_fence_handle; struct pipe_winsys; +struct pipe_screen; + /* The pipe driver currently understands the following chipsets: */ @@ -181,7 +183,7 @@ struct brw_winsys { #define BRW_BUFFER_USAGE_LIT_VERTEX (PIPE_BUFFER_USAGE_CUSTOM << 0) -struct pipe_context *brw_create(struct pipe_winsys *, +struct pipe_context *brw_create(struct pipe_screen *, struct brw_winsys *, unsigned pci_id); diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c index dbfd37bda2..d55d8c39eb 100644 --- a/src/gallium/winsys/xlib/xm_winsys_aub.c +++ b/src/gallium/winsys/xlib/xm_winsys_aub.c @@ -40,6 +40,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "i965simple/brw_winsys.h" +#include "i965simple/brw_screen.h" #include "brw_aub.h" #include "xm_winsys_aub.h" @@ -565,6 +566,7 @@ struct pipe_context * xmesa_create_i965simple( struct pipe_winsys *winsys ) { struct aub_brw_winsys *iws = CALLOC_STRUCT( aub_brw_winsys ); + struct pipe_screen *screen = brw_create_screen(winsys, 0/* XXX pci_id */); /* Fill in this struct with callbacks that i965simple will need to * communicate with the window system, buffer manager, etc. @@ -583,7 +585,7 @@ xmesa_create_i965simple( struct pipe_winsys *winsys ) /* Create the i965simple context: */ - return brw_create( winsys, + return brw_create( screen, &iws->winsys, 0 ); } -- cgit v1.2.3 From 8383f798b41df9a305e0a33afe8afa028d5d5dfb Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 11:24:35 -0700 Subject: gallium: start removing pipe_context->get_name/vendor/param/paramf These are now per-screen functions, not per-context. State tracker updated, code in drivers and p_context.h disabled. --- src/gallium/drivers/cell/ppu/cell_context.c | 29 ---------------- src/gallium/drivers/failover/fo_context.c | 3 ++ src/gallium/drivers/i915simple/Makefile | 1 - src/gallium/drivers/i915simple/SConscript | 1 - src/gallium/drivers/i915simple/i915_context.c | 24 ------------- src/gallium/drivers/i965simple/Makefile | 1 - src/gallium/drivers/i965simple/SConscript | 1 - src/gallium/drivers/i965simple/brw_context.c | 17 ---------- src/gallium/drivers/softpipe/sp_context.c | 25 -------------- src/gallium/include/pipe/p_context.h | 3 ++ src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_cb_strings.c | 9 ++--- src/mesa/state_tracker/st_extensions.c | 49 ++++++++++++++------------- 13 files changed, 37 insertions(+), 128 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index bb1838409f..351601473d 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -77,30 +77,6 @@ cell_is_format_supported( struct pipe_context *pipe, } -static int cell_get_param(struct pipe_context *pipe, int param) -{ - return pipe->screen->get_param(pipe->screen, param); -} - -static float cell_get_paramf(struct pipe_context *pipe, int param) -{ - return pipe->screen->get_paramf(pipe->screen, param); -} - -static const char * -cell_get_name( struct pipe_context *pipe ) -{ - return pipe->screen->get_name(pipe->screen); -} - -static const char * -cell_get_vendor( struct pipe_context *pipe ) -{ - return pipe->screen->get_vendor(pipe->screen); -} - - - static void cell_destroy_context( struct pipe_context *pipe ) { @@ -148,11 +124,6 @@ cell_create_context(struct pipe_screen *screen, /* queries */ cell->pipe.is_format_supported = cell_is_format_supported; - cell->pipe.get_name = cell_get_name; - cell->pipe.get_vendor = cell_get_vendor; - cell->pipe.get_param = cell_get_param; - cell->pipe.get_paramf = cell_get_paramf; - /* state setters */ cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c index 156f7399b0..d5e54f5d61 100644 --- a/src/gallium/drivers/failover/fo_context.c +++ b/src/gallium/drivers/failover/fo_context.c @@ -117,12 +117,15 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->hw = hw; failover->sw = sw; failover->pipe.winsys = hw->winsys; + failover->pipe.screen = hw->screen; failover->pipe.destroy = failover_destroy; failover->pipe.is_format_supported = hw->is_format_supported; +#if 0 failover->pipe.get_name = hw->get_name; failover->pipe.get_vendor = hw->get_vendor; failover->pipe.get_param = hw->get_param; failover->pipe.get_paramf = hw->get_paramf; +#endif failover->pipe.draw_arrays = failover_draw_arrays; failover->pipe.draw_elements = failover_draw_elements; diff --git a/src/gallium/drivers/i915simple/Makefile b/src/gallium/drivers/i915simple/Makefile index 3400747a73..41a61a0020 100644 --- a/src/gallium/drivers/i915simple/Makefile +++ b/src/gallium/drivers/i915simple/Makefile @@ -18,7 +18,6 @@ C_SOURCES = \ i915_state_emit.c \ i915_state_sampler.c \ i915_screen.c \ - i915_strings.c \ i915_prim_emit.c \ i915_prim_vbuf.c \ i915_texture.c \ diff --git a/src/gallium/drivers/i915simple/SConscript b/src/gallium/drivers/i915simple/SConscript index 3e1beaea6d..2366e1247f 100644 --- a/src/gallium/drivers/i915simple/SConscript +++ b/src/gallium/drivers/i915simple/SConscript @@ -22,7 +22,6 @@ i915simple = env.ConvenienceLibrary( 'i915_state_emit.c', 'i915_state_immediate.c', 'i915_state_sampler.c', - 'i915_strings.c', 'i915_surface.c', 'i915_texture.c', ]) diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index e89db8a130..42355552de 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -91,22 +91,6 @@ i915_is_format_supported( struct pipe_context *pipe, } -/* XXX temporary */ -static int -i915_get_param(struct pipe_context *pipe, int param) -{ - return pipe->screen->get_param(pipe->screen, param); -} - - -/* XXX temporary */ -static float -i915_get_paramf(struct pipe_context *pipe, int param) -{ - return pipe->screen->get_paramf(pipe->screen, param); -} - - static void i915_destroy( struct pipe_context *pipe ) { struct i915_context *i915 = i915_context( pipe ); @@ -205,8 +189,6 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, i915->pipe.destroy = i915_destroy; i915->pipe.is_format_supported = i915_is_format_supported; - i915->pipe.get_param = i915_get_param; - i915->pipe.get_paramf = i915_get_paramf; i915->pipe.clear = i915_clear; @@ -229,7 +211,6 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, i915_init_surface_functions(i915); i915_init_state_functions(i915); i915_init_flush_functions(i915); - i915_init_string_functions(i915); i915_init_texture_functions(i915); draw_install_aaline_stage(i915->draw, &i915->pipe); @@ -242,11 +223,6 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, */ i915->batch_start = NULL; - /* - * XXX we could plug GL selection/feedback into the drawing pipeline - * by specifying a different setup/render stage. - */ - return &i915->pipe; } diff --git a/src/gallium/drivers/i965simple/Makefile b/src/gallium/drivers/i965simple/Makefile index 7b0df0f850..8589ebdf96 100644 --- a/src/gallium/drivers/i965simple/Makefile +++ b/src/gallium/drivers/i965simple/Makefile @@ -7,7 +7,6 @@ C_SOURCES = \ brw_blit.c \ brw_flush.c \ brw_screen.c \ - brw_strings.c \ brw_surface.c \ brw_cc.c \ brw_clip.c \ diff --git a/src/gallium/drivers/i965simple/SConscript b/src/gallium/drivers/i965simple/SConscript index 74621de84c..e0cc78826e 100644 --- a/src/gallium/drivers/i965simple/SConscript +++ b/src/gallium/drivers/i965simple/SConscript @@ -35,7 +35,6 @@ i965simple = env.ConvenienceLibrary( 'brw_state_cache.c', 'brw_state_pool.c', 'brw_state_upload.c', - 'brw_strings.c', 'brw_surface.c', 'brw_tex_layout.c', 'brw_urb.c', diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c index d90b32d4b4..e0a4037286 100644 --- a/src/gallium/drivers/i965simple/brw_context.c +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -71,20 +71,6 @@ static void brw_clear(struct pipe_context *pipe, struct pipe_surface *ps, } -static int -brw_get_param(struct pipe_context *pipe, int param) -{ - return pipe->screen->get_param(pipe->screen, param); -} - - -static float -brw_get_paramf(struct pipe_context *pipe, int param) -{ - return pipe->screen->get_paramf(pipe->screen, param); -} - - static boolean brw_is_format_supported( struct pipe_context *pipe, enum pipe_format format, uint type ) @@ -173,15 +159,12 @@ struct pipe_context *brw_create(struct pipe_screen *screen, brw->pipe.destroy = brw_destroy; brw->pipe.is_format_supported = brw_is_format_supported; - brw->pipe.get_param = brw_get_param; - brw->pipe.get_paramf = brw_get_paramf; brw->pipe.clear = brw_clear; brw_init_surface_functions(brw); brw_init_texture_functions(brw); brw_init_state_functions(brw); brw_init_flush_functions(brw); - brw_init_string_functions(brw); brw_init_draw_functions( brw ); diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 4ac1719cbb..6a88c277aa 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -143,27 +143,6 @@ static void softpipe_destroy( struct pipe_context *pipe ) } -/* XXX these will go away shortly */ -static const char *softpipe_get_name( struct pipe_context *pipe ) -{ - return pipe->screen->get_name(pipe->screen); -} - -static const char *softpipe_get_vendor( struct pipe_context *pipe ) -{ - return pipe->screen->get_vendor(pipe->screen); -} - -static int softpipe_get_param(struct pipe_context *pipe, int param) -{ - return pipe->screen->get_param(pipe->screen, param); -} - -static float softpipe_get_paramf(struct pipe_context *pipe, int param) -{ - return pipe->screen->get_paramf(pipe->screen, param); -} - struct pipe_context * softpipe_create( struct pipe_screen *screen, struct pipe_winsys *pipe_winsys, @@ -186,10 +165,6 @@ softpipe_create( struct pipe_screen *screen, /* queries */ softpipe->pipe.is_format_supported = softpipe_is_format_supported; - softpipe->pipe.get_name = softpipe_get_name; - softpipe->pipe.get_vendor = softpipe_get_vendor; - softpipe->pipe.get_param = softpipe_get_param; - softpipe->pipe.get_paramf = softpipe_get_paramf; /* state setters */ softpipe->pipe.create_blend_state = softpipe_create_blend_state; diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 93fcb1c3e9..38ee95bdad 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -66,12 +66,15 @@ struct pipe_context { boolean (*is_format_supported)( struct pipe_context *pipe, enum pipe_format format, uint type ); +#if 0 + /* XXX obsolete, moved into pipe_screen */ const char *(*get_name)( struct pipe_context *pipe ); const char *(*get_vendor)( struct pipe_context *pipe ); int (*get_param)( struct pipe_context *pipe, int param ); float (*get_paramf)( struct pipe_context *pipe, int param ); +#endif /* diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c837698980..6245ce057f 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -649,7 +649,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* XXX if DrawPixels image is larger than max texture size, break * it up into chunks. */ - maxSize = 1 << (pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1); + maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1); assert(width <= maxSize); assert(height <= maxSize); diff --git a/src/mesa/state_tracker/st_cb_strings.c b/src/mesa/state_tracker/st_cb_strings.c index c344df0ff1..247519ab3d 100644 --- a/src/mesa/state_tracker/st_cb_strings.c +++ b/src/mesa/state_tracker/st_cb_strings.c @@ -35,6 +35,7 @@ #include "main/macros.h" #include "main/version.h" #include "pipe/p_context.h" +#include "pipe/p_screen.h" #include "pipe/p_winsys.h" #include "st_context.h" #include "st_cb_strings.h" @@ -45,11 +46,11 @@ static const GLubyte * st_get_string(GLcontext * ctx, GLenum name) { struct st_context *st = st_context(ctx); - struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = st->pipe->screen; switch (name) { case GL_VENDOR: { - const char *vendor = pipe->get_vendor( pipe ); + const char *vendor = screen->get_vendor( screen ); const char *tungsten = "Tungsten Graphics, Inc."; /* Tungsten developed the state_tracker module (and much of @@ -68,8 +69,8 @@ st_get_string(GLcontext * ctx, GLenum name) case GL_RENDERER: snprintf(st->renderer, sizeof(st->renderer), "Gallium %s, %s on %s", ST_VERSION_STRING, - pipe->get_name( pipe ), - pipe->winsys->get_name( pipe->winsys )); + screen->get_name( screen ), + screen->winsys->get_name( screen->winsys )); return (GLubyte *) st->renderer; diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 97d28d77c4..99d2a5fb9e 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -32,6 +32,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_screen.h" #include "st_context.h" #include "st_extensions.h" @@ -64,19 +65,19 @@ static int clamp(int a, int min, int max) */ void st_init_limits(struct st_context *st) { - struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = st->pipe->screen; struct gl_constants *c = &st->ctx->Const; c->MaxTextureLevels - = min(pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_2D_LEVELS), + = min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS), MAX_TEXTURE_LEVELS); c->Max3DTextureLevels - = min(pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_3D_LEVELS), + = min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_3D_LEVELS), MAX_3D_TEXTURE_LEVELS); c->MaxCubeTextureLevels - = min(pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS), + = min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS), MAX_CUBE_TEXTURE_LEVELS); c->MaxTextureRectSize @@ -84,31 +85,31 @@ void st_init_limits(struct st_context *st) c->MaxTextureImageUnits = c->MaxTextureCoordUnits - = min(pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS), + = min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS), MAX_TEXTURE_IMAGE_UNITS); c->MaxDrawBuffers - = clamp(pipe->get_param(pipe, PIPE_CAP_MAX_RENDER_TARGETS), + = clamp(screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS), 1, MAX_DRAW_BUFFERS); c->MaxLineWidth - = max(1.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_LINE_WIDTH)); + = max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH)); c->MaxLineWidthAA - = max(1.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_LINE_WIDTH_AA)); + = max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH_AA)); c->MaxPointSize - = max(1.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_POINT_WIDTH)); + = max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH)); c->MaxPointSizeAA - = max(1.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_POINT_WIDTH_AA)); + = max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH_AA)); c->MaxTextureMaxAnisotropy - = max(2.0, pipe->get_paramf(pipe, PIPE_CAP_MAX_TEXTURE_ANISOTROPY)); + = max(2.0, screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_ANISOTROPY)); c->MaxTextureLodBias - = pipe->get_paramf(pipe, PIPE_CAP_MAX_TEXTURE_LOD_BIAS); + = screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_LOD_BIAS); st->bitmap_texcoord_bias - = pipe->get_paramf(pipe, PIPE_CAP_BITMAP_TEXCOORD_BIAS); + = screen->get_paramf(screen, PIPE_CAP_BITMAP_TEXCOORD_BIAS); } @@ -117,7 +118,7 @@ void st_init_limits(struct st_context *st) */ void st_init_extensions(struct st_context *st) { - struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = st->pipe->screen; GLcontext *ctx = st->ctx; /* @@ -163,11 +164,11 @@ void st_init_extensions(struct st_context *st) /* * Extensions that depend on the driver/hardware: */ - if (pipe->get_param(pipe, PIPE_CAP_MAX_RENDER_TARGETS) > 0) { + if (screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS) > 0) { ctx->Extensions.ARB_draw_buffers = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_GLSL)) { + if (screen->get_param(screen, PIPE_CAP_GLSL)) { ctx->Extensions.ARB_fragment_shader = GL_TRUE; ctx->Extensions.ARB_vertex_shader = GL_TRUE; ctx->Extensions.ARB_shader_objects = GL_TRUE; @@ -175,37 +176,37 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.ARB_shading_language_120 = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_NPOT_TEXTURES)) { + if (screen->get_param(screen, PIPE_CAP_NPOT_TEXTURES)) { ctx->Extensions.ARB_texture_non_power_of_two = GL_TRUE; ctx->Extensions.NV_texture_rectangle = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS) > 1) { + if (screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS) > 1) { ctx->Extensions.ARB_multitexture = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_TWO_SIDED_STENCIL)) { + if (screen->get_param(screen, PIPE_CAP_TWO_SIDED_STENCIL)) { ctx->Extensions.ATI_separate_stencil = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_S3TC)) { + if (screen->get_param(screen, PIPE_CAP_S3TC)) { ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_ANISOTROPIC_FILTER)) { + if (screen->get_param(screen, PIPE_CAP_ANISOTROPIC_FILTER)) { ctx->Extensions.EXT_texture_filter_anisotropic = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_POINT_SPRITE)) { + if (screen->get_param(screen, PIPE_CAP_POINT_SPRITE)) { ctx->Extensions.ARB_point_sprite = GL_TRUE; ctx->Extensions.NV_point_sprite = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_OCCLUSION_QUERY)) { + if (screen->get_param(screen, PIPE_CAP_OCCLUSION_QUERY)) { ctx->Extensions.ARB_occlusion_query = GL_TRUE; } - if (pipe->get_param(pipe, PIPE_CAP_TEXTURE_SHADOW_MAP)) { + if (screen->get_param(screen, PIPE_CAP_TEXTURE_SHADOW_MAP)) { ctx->Extensions.ARB_depth_texture = GL_TRUE; ctx->Extensions.ARB_shadow = GL_TRUE; ctx->Extensions.EXT_shadow_funcs = GL_TRUE; -- cgit v1.2.3 From 364f8cad0f8f02fd39d9c51ea0774d349121b58d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 13:58:06 -0700 Subject: gallium: move is_format_supported() to pipe_screen struct --- src/gallium/drivers/cell/ppu/cell_context.c | 23 --------- src/gallium/drivers/cell/ppu/cell_screen.c | 19 ++++++++ src/gallium/drivers/failover/fo_context.c | 2 +- src/gallium/drivers/i915simple/i915_context.c | 55 --------------------- src/gallium/drivers/i915simple/i915_screen.c | 48 +++++++++++++++++++ src/gallium/drivers/i965simple/brw_context.c | 69 --------------------------- src/gallium/drivers/i965simple/brw_screen.c | 67 ++++++++++++++++++++++++++ src/gallium/drivers/softpipe/sp_context.c | 26 ---------- src/gallium/drivers/softpipe/sp_screen.c | 24 ++++++++++ src/gallium/include/pipe/p_context.h | 19 +------- src/mesa/state_tracker/st_cb_drawpixels.c | 5 +- src/mesa/state_tracker/st_format.c | 45 +++++++++-------- src/mesa/state_tracker/st_gen_mipmap.c | 3 +- 13 files changed, 190 insertions(+), 215 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index 351601473d..2301df5ba5 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -57,26 +57,6 @@ -static boolean -cell_is_format_supported( struct pipe_context *pipe, - enum pipe_format format, uint type ) -{ - /*struct cell_context *cell = cell_context( pipe );*/ - - switch (type) { - case PIPE_TEXTURE: - /* cell supports all texture formats, XXX for now anyway */ - return TRUE; - case PIPE_SURFACE: - /* cell supports all (off-screen) surface formats, XXX for now */ - return TRUE; - default: - assert(0); - return FALSE; - } -} - - static void cell_destroy_context( struct pipe_context *pipe ) { @@ -122,9 +102,6 @@ cell_create_context(struct pipe_screen *screen, cell->pipe.screen = screen; cell->pipe.destroy = cell_destroy_context; - /* queries */ - cell->pipe.is_format_supported = cell_is_format_supported; - /* state setters */ cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; cell->pipe.set_vertex_element = cell_set_vertex_element; diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c index 75255c0466..124670df25 100644 --- a/src/gallium/drivers/cell/ppu/cell_screen.c +++ b/src/gallium/drivers/cell/ppu/cell_screen.c @@ -112,6 +112,24 @@ cell_get_paramf(struct pipe_screen *screen, int param) } +static boolean +cell_is_format_supported( struct pipe_screen *screen, + enum pipe_format format, uint type ) +{ + switch (type) { + case PIPE_TEXTURE: + /* cell supports all texture formats, XXX for now anyway */ + return TRUE; + case PIPE_SURFACE: + /* cell supports all (off-screen) surface formats, XXX for now */ + return TRUE; + default: + assert(0); + return FALSE; + } +} + + static void cell_destroy_screen( struct pipe_screen *screen ) { @@ -140,6 +158,7 @@ cell_create_screen(struct pipe_winsys *winsys) screen->get_vendor = cell_get_vendor; screen->get_param = cell_get_param; screen->get_paramf = cell_get_paramf; + screen->is_format_supported = cell_is_format_supported; cell_init_screen_texture_funcs(screen); diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c index d5e54f5d61..f559cc0d47 100644 --- a/src/gallium/drivers/failover/fo_context.c +++ b/src/gallium/drivers/failover/fo_context.c @@ -119,8 +119,8 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.winsys = hw->winsys; failover->pipe.screen = hw->screen; failover->pipe.destroy = failover_destroy; - failover->pipe.is_format_supported = hw->is_format_supported; #if 0 + failover->pipe.is_format_supported = hw->is_format_supported; failover->pipe.get_name = hw->get_name; failover->pipe.get_vendor = hw->get_vendor; failover->pipe.get_param = hw->get_param; diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 42355552de..15ff2360b7 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -39,58 +39,6 @@ #include "pipe/p_screen.h" -/** - * Query format support for creating a texture, drawing surface, etc. - * \param format the format to test - * \param type one of PIPE_TEXTURE, PIPE_SURFACE - */ -static boolean -i915_is_format_supported( struct pipe_context *pipe, - enum pipe_format format, uint type ) -{ - static const enum pipe_format tex_supported[] = { - PIPE_FORMAT_R8G8B8A8_UNORM, - PIPE_FORMAT_A8R8G8B8_UNORM, - PIPE_FORMAT_R5G6B5_UNORM, - PIPE_FORMAT_U_L8, - PIPE_FORMAT_U_A8, - PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_A8_L8, - PIPE_FORMAT_YCBCR, - PIPE_FORMAT_YCBCR_REV, - PIPE_FORMAT_S8Z24_UNORM, - PIPE_FORMAT_NONE /* list terminator */ - }; - static const enum pipe_format surface_supported[] = { - PIPE_FORMAT_A8R8G8B8_UNORM, - PIPE_FORMAT_R5G6B5_UNORM, - PIPE_FORMAT_S8Z24_UNORM, - /*PIPE_FORMAT_R16G16B16A16_SNORM,*/ - PIPE_FORMAT_NONE /* list terminator */ - }; - const enum pipe_format *list; - uint i; - - switch (type) { - case PIPE_TEXTURE: - list = tex_supported; - break; - case PIPE_SURFACE: - list = surface_supported; - break; - default: - assert(0); - } - - for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) { - if (list[i] == format) - return TRUE; - } - - return FALSE; -} - - static void i915_destroy( struct pipe_context *pipe ) { struct i915_context *i915 = i915_context( pipe ); @@ -101,8 +49,6 @@ static void i915_destroy( struct pipe_context *pipe ) } - - static boolean i915_draw_elements( struct pipe_context *pipe, struct pipe_buffer *indexBuffer, @@ -188,7 +134,6 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, i915->pipe.screen = screen; i915->pipe.destroy = i915_destroy; - i915->pipe.is_format_supported = i915_is_format_supported; i915->pipe.clear = i915_clear; diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index b9b4d6b6fa..8d7bf0b33e 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -145,6 +145,53 @@ i915_get_paramf(struct pipe_screen *screen, int param) } +static boolean +i915_is_format_supported( struct pipe_screen *screen, + enum pipe_format format, uint type ) +{ + static const enum pipe_format tex_supported[] = { + PIPE_FORMAT_R8G8B8A8_UNORM, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_U_L8, + PIPE_FORMAT_U_A8, + PIPE_FORMAT_U_I8, + PIPE_FORMAT_U_A8_L8, + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_NONE /* list terminator */ + }; + static const enum pipe_format surface_supported[] = { + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_S8Z24_UNORM, + /*PIPE_FORMAT_R16G16B16A16_SNORM,*/ + PIPE_FORMAT_NONE /* list terminator */ + }; + const enum pipe_format *list; + uint i; + + switch (type) { + case PIPE_TEXTURE: + list = tex_supported; + break; + case PIPE_SURFACE: + list = surface_supported; + break; + default: + assert(0); + } + + for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) { + if (list[i] == format) + return TRUE; + } + + return FALSE; +} + + static void i915_destroy_screen( struct pipe_screen *screen ) { @@ -195,6 +242,7 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id) i915screen->screen.get_vendor = i915_get_vendor; i915screen->screen.get_param = i915_get_param; i915screen->screen.get_paramf = i915_get_paramf; + i915screen->screen.is_format_supported = i915_is_format_supported; i915_init_screen_texture_functions(&i915screen->screen); diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c index e0a4037286..7c908da672 100644 --- a/src/gallium/drivers/i965simple/brw_context.c +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -71,74 +71,6 @@ static void brw_clear(struct pipe_context *pipe, struct pipe_surface *ps, } -static boolean -brw_is_format_supported( struct pipe_context *pipe, - enum pipe_format format, uint type ) -{ -#if 0 - /* XXX: This is broken -- rewrite if still needed. */ - static const unsigned tex_supported[] = { - PIPE_FORMAT_U_R8_G8_B8_A8, - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_U_L8, - PIPE_FORMAT_U_A8, - PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_L8_A8, - PIPE_FORMAT_YCBCR, - PIPE_FORMAT_YCBCR_REV, - PIPE_FORMAT_S8_Z24, - }; - - - /* Actually a lot more than this - add later: - */ - static const unsigned render_supported[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - }; - - /* - */ - static const unsigned z_stencil_supported[] = { - PIPE_FORMAT_U_Z16, - PIPE_FORMAT_U_Z32, - PIPE_FORMAT_S8_Z24, - }; - - switch (type) { - case PIPE_RENDER_FORMAT: - *numFormats = Elements(render_supported); - return render_supported; - - case PIPE_TEX_FORMAT: - *numFormats = Elements(tex_supported); - return render_supported; - - case PIPE_Z_STENCIL_FORMAT: - *numFormats = Elements(render_supported); - return render_supported; - - default: - *numFormats = 0; - return NULL; - } -#else - switch (format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - return TRUE; - default: - return FALSE; - }; - return FALSE; -#endif -} - - - - struct pipe_context *brw_create(struct pipe_screen *screen, struct brw_winsys *brw_winsys, unsigned pci_id) @@ -158,7 +90,6 @@ struct pipe_context *brw_create(struct pipe_screen *screen, brw->pipe.screen = screen; brw->pipe.destroy = brw_destroy; - brw->pipe.is_format_supported = brw_is_format_supported; brw->pipe.clear = brw_clear; brw_init_surface_functions(brw); diff --git a/src/gallium/drivers/i965simple/brw_screen.c b/src/gallium/drivers/i965simple/brw_screen.c index 7afe070bd8..5be369fe52 100644 --- a/src/gallium/drivers/i965simple/brw_screen.c +++ b/src/gallium/drivers/i965simple/brw_screen.c @@ -133,6 +133,72 @@ brw_get_paramf(struct pipe_screen *screen, int param) } +static boolean +brw_is_format_supported( struct pipe_screen *screen, + enum pipe_format format, uint type ) +{ +#if 0 + /* XXX: This is broken -- rewrite if still needed. */ + static const unsigned tex_supported[] = { + PIPE_FORMAT_U_R8_G8_B8_A8, + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_U_L8, + PIPE_FORMAT_U_A8, + PIPE_FORMAT_U_I8, + PIPE_FORMAT_U_L8_A8, + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + PIPE_FORMAT_S8_Z24, + }; + + + /* Actually a lot more than this - add later: + */ + static const unsigned render_supported[] = { + PIPE_FORMAT_U_A8_R8_G8_B8, + PIPE_FORMAT_U_R5_G6_B5, + }; + + /* + */ + static const unsigned z_stencil_supported[] = { + PIPE_FORMAT_U_Z16, + PIPE_FORMAT_U_Z32, + PIPE_FORMAT_S8_Z24, + }; + + switch (type) { + case PIPE_RENDER_FORMAT: + *numFormats = Elements(render_supported); + return render_supported; + + case PIPE_TEX_FORMAT: + *numFormats = Elements(tex_supported); + return render_supported; + + case PIPE_Z_STENCIL_FORMAT: + *numFormats = Elements(render_supported); + return render_supported; + + default: + *numFormats = 0; + return NULL; + } +#else + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + return TRUE; + default: + return FALSE; + }; + return FALSE; +#endif +} + + static void brw_destroy_screen( struct pipe_screen *screen ) { @@ -161,6 +227,7 @@ brw_create_screen(struct pipe_winsys *winsys, uint pci_id) brwscreen->screen.get_vendor = brw_get_vendor; brwscreen->screen.get_param = brw_get_param; brwscreen->screen.get_paramf = brw_get_paramf; + brwscreen->screen.is_format_supported = brw_is_format_supported; brw_init_screen_texture_funcs(&brwscreen->screen); diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 6a88c277aa..fa16ed94e8 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -47,29 +47,6 @@ -/** - * Query format support for creating a texture, drawing surface, etc. - * \param format the format to test - * \param type one of PIPE_TEXTURE, PIPE_SURFACE - */ -static boolean -softpipe_is_format_supported( struct pipe_context *pipe, - enum pipe_format format, uint type ) -{ - switch (type) { - case PIPE_TEXTURE: - /* softpipe supports all texture formats */ - return TRUE; - case PIPE_SURFACE: - /* softpipe supports all (off-screen) surface formats */ - return TRUE; - default: - assert(0); - return FALSE; - } -} - - /** * Map any drawing surfaces which aren't already mapped */ @@ -163,9 +140,6 @@ softpipe_create( struct pipe_screen *screen, softpipe->pipe.screen = screen; softpipe->pipe.destroy = softpipe_destroy; - /* queries */ - softpipe->pipe.is_format_supported = softpipe_is_format_supported; - /* state setters */ softpipe->pipe.create_blend_state = softpipe_create_blend_state; softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index cc3b962580..1850a1ced3 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -107,6 +107,29 @@ softpipe_get_paramf(struct pipe_screen *screen, int param) } +/** + * Query format support for creating a texture, drawing surface, etc. + * \param format the format to test + * \param type one of PIPE_TEXTURE, PIPE_SURFACE + */ +static boolean +softpipe_is_format_supported( struct pipe_screen *screen, + enum pipe_format format, uint type ) +{ + switch (type) { + case PIPE_TEXTURE: + /* softpipe supports all texture formats */ + return TRUE; + case PIPE_SURFACE: + /* softpipe supports all (off-screen) surface formats */ + return TRUE; + default: + assert(0); + return FALSE; + } +} + + static void softpipe_destroy_screen( struct pipe_screen *screen ) { @@ -134,6 +157,7 @@ softpipe_create_screen(struct pipe_winsys *winsys) screen->get_vendor = softpipe_get_vendor; screen->get_param = softpipe_get_param; screen->get_paramf = softpipe_get_paramf; + screen->is_format_supported = softpipe_is_format_supported; softpipe_init_screen_texture_funcs(screen); diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 38ee95bdad..bb345df153 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -59,24 +59,6 @@ struct pipe_context { void (*destroy)( struct pipe_context * ); - /* - * Queries - */ - /** type is one of PIPE_SURFACE, PIPE_TEXTURE, etc. */ - boolean (*is_format_supported)( struct pipe_context *pipe, - enum pipe_format format, uint type ); - -#if 0 - /* XXX obsolete, moved into pipe_screen */ - const char *(*get_name)( struct pipe_context *pipe ); - - const char *(*get_vendor)( struct pipe_context *pipe ); - - int (*get_param)( struct pipe_context *pipe, int param ); - float (*get_paramf)( struct pipe_context *pipe, int param ); -#endif - - /* * Drawing. * Return false on fallbacks (temporary??) @@ -210,6 +192,7 @@ struct pipe_context { /* * Texture functions + * XXX these are moving to pipe_screen... */ struct pipe_texture * (*texture_create)(struct pipe_context *pipe, const struct pipe_texture *templat); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 6245ce057f..0f2c6307dd 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -993,6 +993,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, const GLubyte *bitmap) { struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_surface *surface; uint format = 0, cpp, comp; ubyte *dest; @@ -1000,12 +1001,12 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, int row, col; /* find a texture format we know */ - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8, PIPE_TEXTURE )) { + if (screen->is_format_supported( screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE )) { format = PIPE_FORMAT_U_I8; cpp = 1; comp = 0; } - else if (pipe->is_format_supported( pipe, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_TEXTURE )) { + else if (screen->is_format_supported( screen, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_TEXTURE )) { format = PIPE_FORMAT_A8R8G8B8_UNORM; cpp = 4; comp = 3; /* alpha channel */ /*XXX little-endian dependency */ diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 2a23445ca2..9aeda65a5c 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -40,6 +40,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_screen.h" #include "st_context.h" #include "st_format.h" @@ -288,9 +289,10 @@ default_rgba_format(struct pipe_context *pipe, uint type) PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_R5G6B5_UNORM }; + struct pipe_screen *screen = pipe->screen; uint i; for (i = 0; i < Elements(colorFormats); i++) { - if (pipe->is_format_supported( pipe, colorFormats[i], type )) { + if (screen->is_format_supported( screen, colorFormats[i], type )) { return colorFormats[i]; } } @@ -304,7 +306,8 @@ default_rgba_format(struct pipe_context *pipe, uint type) static GLuint default_deep_rgba_format(struct pipe_context *pipe, uint type) { - if (pipe->is_format_supported(pipe, PIPE_FORMAT_R16G16B16A16_SNORM, type)) { + struct pipe_screen *screen = pipe->screen; + if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SNORM, type)) { return PIPE_FORMAT_R16G16B16A16_SNORM; } return PIPE_FORMAT_NONE; @@ -323,9 +326,10 @@ default_depth_format(struct pipe_context *pipe, uint type) PIPE_FORMAT_S8Z24_UNORM, PIPE_FORMAT_Z24S8_UNORM }; + struct pipe_screen *screen = pipe->screen; uint i; for (i = 0; i < Elements(zFormats); i++) { - if (pipe->is_format_supported( pipe, zFormats[i], type )) { + if (screen->is_format_supported( screen, zFormats[i], type )) { return zFormats[i]; } } @@ -341,6 +345,7 @@ default_depth_format(struct pipe_context *pipe, uint type) enum pipe_format st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) { + struct pipe_screen *screen = pipe->screen; uint surfType = PIPE_SURFACE; switch (internalFormat) { @@ -359,12 +364,12 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_RGBA4: case GL_RGBA2: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A4R4G4B4_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_A4R4G4B4_UNORM, surfType )) return PIPE_FORMAT_A4R4G4B4_UNORM; return default_rgba_format( pipe, surfType ); case GL_RGB5_A1: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) return PIPE_FORMAT_A1R5G5B5_UNORM; return default_rgba_format( pipe, surfType ); @@ -377,9 +382,9 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) return PIPE_FORMAT_A1R5G5B5_UNORM; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_R5G6B5_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_R5G6B5_UNORM, surfType )) return PIPE_FORMAT_R5G6B5_UNORM; return default_rgba_format( pipe, surfType ); @@ -389,7 +394,7 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_U_A8, surfType )) return PIPE_FORMAT_U_A8; return default_rgba_format( pipe, surfType ); @@ -400,7 +405,7 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_L8, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_U_L8, surfType )) return PIPE_FORMAT_U_A8; return default_rgba_format( pipe, surfType ); @@ -413,7 +418,7 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_A8_L8, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_U_A8_L8, surfType )) return PIPE_FORMAT_U_A8_L8; return default_rgba_format( pipe, surfType ); @@ -423,7 +428,7 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_I8, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_U_I8, surfType )) return PIPE_FORMAT_U_I8; return default_rgba_format( pipe, surfType ); @@ -454,17 +459,17 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) #endif case GL_DEPTH_COMPONENT16: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z16_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z16_UNORM, surfType )) return PIPE_FORMAT_Z16_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT24: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, surfType )) return PIPE_FORMAT_S8Z24_UNORM; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, surfType )) return PIPE_FORMAT_Z24S8_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT32: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z32_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z32_UNORM, surfType )) return PIPE_FORMAT_Z32_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT: @@ -475,19 +480,19 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_STENCIL_INDEX4_EXT: case GL_STENCIL_INDEX8_EXT: case GL_STENCIL_INDEX16_EXT: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_U_S8, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_U_S8, surfType )) return PIPE_FORMAT_U_S8; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, surfType )) return PIPE_FORMAT_S8Z24_UNORM; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, surfType )) return PIPE_FORMAT_Z24S8_UNORM; return PIPE_FORMAT_NONE; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - if (pipe->is_format_supported( pipe, PIPE_FORMAT_S8Z24_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, surfType )) return PIPE_FORMAT_S8Z24_UNORM; - if (pipe->is_format_supported( pipe, PIPE_FORMAT_Z24S8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, surfType )) return PIPE_FORMAT_Z24S8_UNORM; return PIPE_FORMAT_NONE; diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index c9765b2003..2b16310602 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -227,6 +227,7 @@ st_render_mipmap(struct st_context *st, uint baseLevel, uint lastLevel) { struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_framebuffer_state fb; struct pipe_sampler_state sampler; void *sampler_cso; @@ -237,7 +238,7 @@ st_render_mipmap(struct st_context *st, assert(target != GL_TEXTURE_3D); /* not done yet */ /* check if we can render in the texture's format */ - if (!pipe->is_format_supported(pipe, pt->format, PIPE_SURFACE)) { + if (!screen->is_format_supported(screen, pt->format, PIPE_SURFACE)) { return FALSE; } -- cgit v1.2.3 From 6f715dcc219071e574e363a9db4365c9c31ebbd3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 14:21:12 -0700 Subject: gallium: remove pipe_context->texture_create/release/get_tex_surface() These functions are now per-screen, not per-context. --- src/gallium/auxiliary/draw/draw_aaline.c | 5 ++- src/gallium/auxiliary/draw/draw_pstipple.c | 6 ++- src/gallium/drivers/cell/ppu/cell_texture.c | 30 ------------- src/gallium/drivers/failover/fo_context.c | 4 +- src/gallium/drivers/i915simple/i915_texture.c | 28 ------------ src/gallium/drivers/i965simple/brw_tex_layout.c | 30 ------------- src/gallium/drivers/softpipe/sp_texture.c | 60 ++++++------------------- src/gallium/drivers/softpipe/sp_tile_cache.c | 3 +- src/gallium/include/pipe/p_context.h | 18 +------- src/gallium/include/pipe/p_inlines.h | 18 +++----- src/mesa/state_tracker/st_cb_drawpixels.c | 8 ++-- src/mesa/state_tracker/st_cb_fbo.c | 9 ++-- src/mesa/state_tracker/st_cb_texture.c | 28 ++++++------ src/mesa/state_tracker/st_gen_mipmap.c | 7 +-- src/mesa/state_tracker/st_texture.c | 16 ++++--- 15 files changed, 71 insertions(+), 199 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index 73a02a32e4..be6cfd3b6a 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -362,6 +362,7 @@ static void aaline_create_texture(struct aaline_stage *aaline) { struct pipe_context *pipe = aaline->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_texture texTemp; uint level; @@ -374,7 +375,7 @@ aaline_create_texture(struct aaline_stage *aaline) texTemp.depth[0] = 1; texTemp.cpp = 1; - aaline->texture = pipe->texture_create(pipe, &texTemp); + aaline->texture = screen->texture_create(screen, &texTemp); /* Fill in mipmap images. * Basically each level is solid opaque, except for the outermost @@ -388,7 +389,7 @@ aaline_create_texture(struct aaline_stage *aaline) assert(aaline->texture->width[level] == aaline->texture->height[level]); - surface = pipe->get_tex_surface(pipe, aaline->texture, 0, level, 0); + surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0); data = pipe_surface_map(surface); for (i = 0; i < size; i++) { diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index 1ab04cd959..efc88bf038 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -348,12 +348,13 @@ pstip_update_texture(struct pstip_stage *pstip) { static const uint bit31 = 1 << 31; struct pipe_context *pipe = pstip->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_surface *surface; const uint *stipple = pstip->state.stipple->stipple; uint i, j; ubyte *data; - surface = pipe->get_tex_surface(pipe, pstip->texture, 0, 0, 0); + surface = screen->get_tex_surface(screen, pstip->texture, 0, 0, 0); data = pipe_surface_map(surface); /* @@ -389,6 +390,7 @@ static void pstip_create_texture(struct pstip_stage *pstip) { struct pipe_context *pipe = pstip->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_texture texTemp; memset(&texTemp, 0, sizeof(texTemp)); @@ -400,7 +402,7 @@ pstip_create_texture(struct pstip_stage *pstip) texTemp.depth[0] = 1; texTemp.cpp = 1; - pstip->texture = pipe->texture_create(pipe, &texTemp); + pstip->texture = screen->texture_create(screen, &texTemp); //pstip_update_texture(pstip); } diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index e6398a85fa..28cadad6ed 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -79,14 +79,6 @@ cell_texture_layout(struct cell_texture * spt) } -static struct pipe_texture * -cell_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat) -{ - return pipe->screen->texture_create(pipe->screen, templat); - -} - static struct pipe_texture * cell_texture_create_screen(struct pipe_screen *screen, const struct pipe_texture *templat) @@ -116,13 +108,6 @@ cell_texture_create_screen(struct pipe_screen *screen, } -static void -cell_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) -{ - return pipe->screen->texture_release(pipe->screen, pt); -} - - static void cell_texture_release_screen(struct pipe_screen *screen, struct pipe_texture **pt) @@ -157,18 +142,6 @@ cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) } -/** - * Called via pipe->get_tex_surface() - */ -static struct pipe_surface * -cell_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - return pipe->screen->get_tex_surface(pipe->screen, pt, face, level, zslice); -} - - static struct pipe_surface * cell_get_tex_surface_screen(struct pipe_screen *screen, struct pipe_texture *pt, @@ -294,10 +267,7 @@ cell_update_texture_mapping(struct cell_context *cell) void cell_init_texture_functions(struct cell_context *cell) { - cell->pipe.texture_create = cell_texture_create; - cell->pipe.texture_release = cell_texture_release; cell->pipe.texture_update = cell_texture_update; - cell->pipe.get_tex_surface = cell_get_tex_surface; } void diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c index f559cc0d47..afc0d7eb1e 100644 --- a/src/gallium/drivers/failover/fo_context.c +++ b/src/gallium/drivers/failover/fo_context.c @@ -143,10 +143,12 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.surface_copy = hw->surface_copy; failover->pipe.surface_fill = hw->surface_fill; +#if 0 failover->pipe.texture_create = hw->texture_create; failover->pipe.texture_release = hw->texture_release; - failover->pipe.texture_update = hw->texture_update; failover->pipe.get_tex_surface = hw->get_tex_surface; +#endif + failover->pipe.texture_update = hw->texture_update; failover->pipe.flush = hw->flush; diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 3c9509dee3..9cdf3418a9 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -511,14 +511,6 @@ i915_texture_create_screen(struct pipe_screen *screen, } -static struct pipe_texture * -i915_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat) -{ - return pipe->screen->texture_create(pipe->screen, templat); -} - - static void i915_texture_release_screen(struct pipe_screen *screen, struct pipe_texture **pt) @@ -550,13 +542,6 @@ i915_texture_release_screen(struct pipe_screen *screen, } -static void -i915_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) -{ - i915_texture_release_screen(pipe->screen, pt); -} - - static void i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) { @@ -606,26 +591,13 @@ i915_get_tex_surface_screen(struct pipe_screen *screen, } -static struct pipe_surface * -i915_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - return i915_get_tex_surface_screen(pipe->screen, pt, face, level, zslice); -} - - void i915_init_texture_functions(struct i915_context *i915) { - i915->pipe.texture_create = i915_texture_create; - i915->pipe.texture_release = i915_texture_release; i915->pipe.texture_update = i915_texture_update; - i915->pipe.get_tex_surface = i915_get_tex_surface; } - void i915_init_screen_texture_functions(struct pipe_screen *screen) { diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 9753c50143..b24ac87c37 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -298,14 +298,6 @@ static boolean brw_miptree_layout(struct brw_texture *tex) } -static struct pipe_texture * -brw_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat) -{ - return pipe->screen->texture_create(pipe->screen, templat); -} - - static struct pipe_texture * brw_texture_create_screen(struct pipe_screen *screen, const struct pipe_texture *templat) @@ -333,13 +325,6 @@ brw_texture_create_screen(struct pipe_screen *screen, } -static void -brw_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) -{ - pipe->screen->texture_release(pipe->screen, pt); -} - - static void brw_texture_release_screen(struct pipe_screen *screen, struct pipe_texture **pt) @@ -379,18 +364,6 @@ brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) } -/* - * XXX note: same as code in sp_surface.c - */ -static struct pipe_surface * -brw_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - return pipe->screen->get_tex_surface(pipe->screen, pt, face, level, zslice); -} - - static struct pipe_surface * brw_get_tex_surface_screen(struct pipe_screen *screen, struct pipe_texture *pt, @@ -433,10 +406,7 @@ brw_get_tex_surface_screen(struct pipe_screen *screen, void brw_init_texture_functions(struct brw_context *brw) { - brw->pipe.texture_create = brw_texture_create; - brw->pipe.texture_release = brw_texture_release; brw->pipe.texture_update = brw_texture_update; - brw->pipe.get_tex_surface = brw_get_tex_surface; } diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index f0e8350a4a..7c02765313 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -80,15 +80,6 @@ softpipe_texture_layout(struct softpipe_texture * spt) } -/* XXX temporary */ -static struct pipe_texture * -softpipe_texture_create(struct pipe_context *pipe, - const struct pipe_texture *templat) -{ - return pipe->screen->texture_create(pipe->screen, templat); -} - - static struct pipe_texture * softpipe_texture_create_screen(struct pipe_screen *screen, const struct pipe_texture *templat) @@ -119,14 +110,6 @@ softpipe_texture_create_screen(struct pipe_screen *screen, } -/* XXX temporary */ -static void -softpipe_texture_release(struct pipe_context *pipe, struct pipe_texture **pt) -{ - return pipe->screen->texture_release(pipe->screen, pt); -} - - static void softpipe_texture_release_screen(struct pipe_screen *screen, struct pipe_texture **pt) @@ -153,33 +136,6 @@ softpipe_texture_release_screen(struct pipe_screen *screen, } -static void -softpipe_texture_update(struct pipe_context *pipe, - struct pipe_texture *texture) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - uint unit; - for (unit = 0; unit < PIPE_MAX_SAMPLERS; unit++) { - if (softpipe->texture[unit] == texture) { - sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); - } - } -} - - -/** - * Called via pipe->get_tex_surface() - */ -/* XXX temporary */ -static struct pipe_surface * -softpipe_get_tex_surface(struct pipe_context *pipe, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - return pipe->screen->get_tex_surface(pipe->screen, pt, face, level, zslice); -} - - static struct pipe_surface * softpipe_get_tex_surface_screen(struct pipe_screen *screen, struct pipe_texture *pt, @@ -217,14 +173,24 @@ softpipe_get_tex_surface_screen(struct pipe_screen *screen, } +static void +softpipe_texture_update(struct pipe_context *pipe, + struct pipe_texture *texture) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + uint unit; + for (unit = 0; unit < PIPE_MAX_SAMPLERS; unit++) { + if (softpipe->texture[unit] == texture) { + sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); + } + } +} + void softpipe_init_texture_funcs( struct softpipe_context *softpipe ) { - softpipe->pipe.texture_create = softpipe_texture_create; - softpipe->pipe.texture_release = softpipe_texture_release; softpipe->pipe.texture_update = softpipe_texture_update; - softpipe->pipe.get_tex_surface = softpipe_get_tex_surface; } diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 0ff93c5527..4caf2dd3fc 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -489,6 +489,7 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, struct softpipe_tile_cache *tc, int x, int y, int z, int face, int level) { + struct pipe_screen *screen = pipe->screen; /* tile pos in framebuffer: */ const int tile_x = x & ~(TILE_SIZE - 1); const int tile_y = y & ~(TILE_SIZE - 1); @@ -514,7 +515,7 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, if (tc->tex_surf_map) pipe_surface_unmap(tc->tex_surf); - tc->tex_surf = pipe->get_tex_surface(pipe, tc->texture, face, level, z); + tc->tex_surf = screen->get_tex_surface(screen, tc->texture, face, level, z); tc->tex_surf_map = pipe_surface_map(tc->tex_surf); tc->tex_face = face; diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index bb345df153..d0f25d7d46 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -189,30 +189,16 @@ struct pipe_context { struct pipe_surface *ps, unsigned clearValue); - - /* - * Texture functions - * XXX these are moving to pipe_screen... - */ - struct pipe_texture * (*texture_create)(struct pipe_context *pipe, - const struct pipe_texture *templat); - - void (*texture_release)(struct pipe_context *pipe, - struct pipe_texture **pt); - /** * Called when texture data is changed. * Note: we could pass some hints about which mip levels or cube faces * have changed... + * XXX this may go away - could pass a 'write' flag to get_tex_surface() */ void (*texture_update)(struct pipe_context *pipe, struct pipe_texture *texture); - /** Get a surface which is a "view" into a texture */ - struct pipe_surface *(*get_tex_surface)(struct pipe_context *pipe, - struct pipe_texture *texture, - unsigned face, unsigned level, - unsigned zslice); + /* Flush rendering: */ diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h index a7e97fcd7d..274f76a383 100644 --- a/src/gallium/include/pipe/p_inlines.h +++ b/src/gallium/include/pipe/p_inlines.h @@ -107,15 +107,9 @@ pipe_texture_reference(struct pipe_texture **ptr, pt->refcount++; if (*ptr) { - struct pipe_context *pipe = (*ptr)->pipe; - /* XXX temporary mess here */ - if (pipe) { - pipe->texture_release(pipe, ptr); - } - else { - struct pipe_screen *screen = (*ptr)->screen; - screen->texture_release(screen, ptr); - } + struct pipe_screen *screen = (*ptr)->screen; + assert(screen); + screen->texture_release(screen, ptr); assert(!*ptr); } @@ -127,10 +121,10 @@ pipe_texture_reference(struct pipe_texture **ptr, static INLINE void pipe_texture_release(struct pipe_texture **ptr) { - struct pipe_context *pipe; + struct pipe_screen *screen; assert(ptr); - pipe = (*ptr)->pipe; - pipe->texture_release(pipe, ptr); + screen = (*ptr)->screen; + screen->texture_release(screen, ptr); *ptr = NULL; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 0f2c6307dd..ff236adc5c 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -456,6 +456,7 @@ make_texture(struct st_context *st, { GLcontext *ctx = st->ctx; struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; const struct gl_texture_format *mformat; struct pipe_texture *pt; enum pipe_format pipeFormat; @@ -493,7 +494,7 @@ make_texture(struct st_context *st, /* we'll do pixel transfer in a fragment shader */ ctx->_ImageTransferState = 0x0; - surface = pipe->get_tex_surface(pipe, pt, 0, 0, 0); + surface = screen->get_tex_surface(screen, pt, 0, 0, 0); /* map texture surface */ dest = pipe_surface_map(surface); @@ -1031,7 +1032,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, printf("st_Bitmap (sourcing from PBO not implemented yet)\n"); } - surface = pipe->get_tex_surface(pipe, pt, 0, 0, 0); + surface = screen->get_tex_surface(screen, pt, 0, 0, 0); /* map texture surface */ dest = pipe_surface_map(surface); @@ -1207,6 +1208,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, { struct st_context *st = ctx->st; struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; struct st_renderbuffer *rbRead; struct st_vertex_program *stvp; struct st_fragment_program *stfp; @@ -1248,7 +1250,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, if (!pt) return; - psTex = pipe->get_tex_surface(pipe, pt, 0, 0, 0); + psTex = screen->get_tex_surface(screen, pt, 0, 0, 0); if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { srcy = ctx->DrawBuffer->Height - srcy - height; diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 781425b546..5384252a8e 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -307,6 +307,7 @@ st_render_texture(GLcontext *ctx, struct st_renderbuffer *strb; struct gl_renderbuffer *rb; struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_texture *pt; assert(!att->Renderbuffer); @@ -332,10 +333,10 @@ st_render_texture(GLcontext *ctx, rb->Height = pt->height[att->TextureLevel]; /* the renderbuffer's surface is inside the texture */ - strb->surface = pipe->get_tex_surface(pipe, pt, - att->CubeMapFace, - att->TextureLevel, - att->Zoffset); + strb->surface = screen->get_tex_surface(screen, pt, + att->CubeMapFace, + att->TextureLevel, + att->Zoffset); assert(strb->surface); init_renderbuffer_bits(strb, pt->format); diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index f5f956f6ea..1ba3173312 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -144,12 +144,11 @@ st_NewTextureObject(GLcontext * ctx, GLuint name, GLenum target) static void st_DeleteTextureObject(GLcontext *ctx, - struct gl_texture_object *texObj) + struct gl_texture_object *texObj) { struct st_texture_object *stObj = st_texture_object(texObj); - if (stObj->pt) - ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); + pipe_texture_release(&stObj->pt); _mesa_delete_texture_object(ctx, texObj); } @@ -163,7 +162,7 @@ st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) DBG("%s\n", __FUNCTION__); if (stImage->pt) { - ctx->st->pipe->texture_release(ctx->st->pipe, &stImage->pt); + pipe_texture_release(&stImage->pt); } if (texImage->Data) { @@ -537,7 +536,7 @@ st_TexImage(GLcontext * ctx, * Release any old malloced memory. */ if (stImage->pt) { - ctx->st->pipe->texture_release(ctx->st->pipe, &stImage->pt); + pipe_texture_release(&stImage->pt); assert(!texImage->Data); } else if (texImage->Data) { @@ -556,7 +555,7 @@ st_TexImage(GLcontext * ctx, stImage->face, stImage->level)) { DBG("release it\n"); - ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); + pipe_texture_release(&stObj->pt); assert(!stObj->pt); } @@ -1025,6 +1024,7 @@ fallback_copy_texsubimage(GLcontext *ctx, GLsizei width, GLsizei height) { struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; const uint face = texture_face(target); struct pipe_texture *pt = stImage->pt; struct pipe_surface *src_surf, *dest_surf; @@ -1042,8 +1042,7 @@ fallback_copy_texsubimage(GLcontext *ctx, src_surf = strb->surface; - dest_surf = pipe->get_tex_surface(pipe, pt, - face, level, destZ); + dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ); /* buffer for one row */ data = (GLfloat *) malloc(width * 4 * sizeof(GLfloat)); @@ -1096,6 +1095,7 @@ do_copy_texsubimage(GLcontext *ctx, struct gl_framebuffer *fb = ctx->ReadBuffer; struct st_renderbuffer *strb; struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_surface *dest_surface; uint dest_format, src_format; uint do_flip = FALSE; @@ -1126,8 +1126,8 @@ do_copy_texsubimage(GLcontext *ctx, src_format = strb->surface->format; dest_format = stImage->pt->format; - dest_surface = pipe->get_tex_surface(pipe, stImage->pt, stImage->face, - stImage->level, destZ); + dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, + stImage->level, destZ); if (src_format == dest_format && ctx->_ImageTransferState == 0x0 && @@ -1352,7 +1352,7 @@ copy_image_data_to_texture(struct st_context *st, stImage->face ); - st->pipe->texture_release(st->pipe, &stImage->pt); + pipe_texture_release(&stImage->pt); } else { assert(stImage->base.Data != NULL); @@ -1408,7 +1408,7 @@ st_finalize_texture(GLcontext *ctx, */ if (firstImage->base.Border) { if (stObj->pt) { - ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); + pipe_texture_release(&stObj->pt); } return GL_FALSE; } @@ -1424,7 +1424,7 @@ st_finalize_texture(GLcontext *ctx, firstImage->pt->last_level >= stObj->lastLevel) { if (stObj->pt) - ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); + pipe_texture_release(&stObj->pt); pipe_texture_reference(&stObj->pt, firstImage->pt); } @@ -1450,7 +1450,7 @@ st_finalize_texture(GLcontext *ctx, stObj->pt->depth[0] != firstImage->base.Depth || stObj->pt->cpp != cpp || stObj->pt->compressed != firstImage->base.IsCompressed)) { - ctx->st->pipe->texture_release(ctx->st->pipe, &stObj->pt); + pipe_texture_release(&stObj->pt); } diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 2b16310602..243dc0b1d0 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -276,7 +276,7 @@ st_render_mipmap(struct st_context *st, /* * Setup framebuffer / dest surface */ - fb.cbufs[0] = pipe->get_tex_surface(pipe, pt, face, dstLevel, zslice); + fb.cbufs[0] = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); pipe->set_framebuffer_state(pipe, &fb); /* @@ -325,6 +325,7 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, struct gl_texture_object *texObj) { struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_winsys *ws = pipe->winsys; struct pipe_texture *pt = st_get_texobj_texture(texObj); const uint baseLevel = texObj->BaseLevel; @@ -345,8 +346,8 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, const ubyte *srcData; ubyte *dstData; - srcSurf = pipe->get_tex_surface(pipe, pt, face, srcLevel, zslice); - dstSurf = pipe->get_tex_surface(pipe, pt, face, dstLevel, zslice); + srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice); + dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); srcData = (ubyte *) ws->buffer_map(ws, srcSurf->buffer, PIPE_BUFFER_USAGE_CPU_READ) diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index ad284170e4..c2b0aa8a4a 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -77,6 +77,7 @@ st_texture_create(struct st_context *st, GLuint compress_byte) { struct pipe_texture pt, *newtex; + struct pipe_screen *screen = st->pipe->screen; assert(target <= PIPE_TEXTURE_CUBE); @@ -96,7 +97,7 @@ st_texture_create(struct st_context *st, pt.compressed = compress_byte ? 1 : 0; pt.cpp = pt.compressed ? compress_byte : st_sizeof_format(format); - newtex = st->pipe->texture_create(st->pipe, &pt); + newtex = screen->texture_create(screen, &pt); assert(!newtex || newtex->refcount == 1); @@ -183,11 +184,12 @@ GLubyte * st_texture_image_map(struct st_context *st, struct st_texture_image *stImage, GLuint zoffset) { + struct pipe_screen *screen = st->pipe->screen; struct pipe_texture *pt = stImage->pt; DBG("%s \n", __FUNCTION__); - stImage->surface = st->pipe->get_tex_surface(st->pipe, pt, stImage->face, - stImage->level, zoffset); + stImage->surface = screen->get_tex_surface(screen, pt, stImage->face, + stImage->level, zoffset); return pipe_surface_map(stImage->surface); } @@ -239,6 +241,7 @@ st_texture_image_data(struct pipe_context *pipe, void *src, GLuint src_row_pitch, GLuint src_image_pitch) { + struct pipe_screen *screen = pipe->screen; GLuint depth = dst->depth[level]; GLuint i; GLuint height = 0; @@ -251,7 +254,7 @@ st_texture_image_data(struct pipe_context *pipe, if(dst->compressed) height /= 4; - dst_surface = pipe->get_tex_surface(pipe, dst, face, level, i); + dst_surface = screen->get_tex_surface(screen, dst, face, level, i); st_surface_data(pipe, dst_surface, 0, 0, /* dstx, dsty */ @@ -275,6 +278,7 @@ st_texture_image_copy(struct pipe_context *pipe, struct pipe_texture *src, GLuint face) { + struct pipe_screen *screen = pipe->screen; GLuint width = dst->width[dstLevel]; GLuint height = dst->height[dstLevel]; GLuint depth = dst->depth[dstLevel]; @@ -299,8 +303,8 @@ st_texture_image_copy(struct pipe_context *pipe, assert(src->width[srcLevel] == width); assert(src->height[srcLevel] == height); - dst_surface = pipe->get_tex_surface(pipe, dst, face, dstLevel, i); - src_surface = pipe->get_tex_surface(pipe, src, face, srcLevel, i); + dst_surface = screen->get_tex_surface(screen, dst, face, dstLevel, i); + src_surface = screen->get_tex_surface(screen, src, face, srcLevel, i); pipe->surface_copy(pipe, FALSE, -- cgit v1.2.3 From cc5ffd762227345d0a5bf9e9356dd83a8b2a8b33 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 14:28:17 -0700 Subject: gallium: remove obsolete *_strings.c files --- src/gallium/drivers/i915simple/i915_strings.c | 53 --------------------------- src/gallium/drivers/i965simple/brw_strings.c | 51 -------------------------- 2 files changed, 104 deletions(-) delete mode 100644 src/gallium/drivers/i915simple/i915_strings.c delete mode 100644 src/gallium/drivers/i965simple/brw_strings.c (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_strings.c b/src/gallium/drivers/i915simple/i915_strings.c deleted file mode 100644 index 14da3bcd72..0000000000 --- a/src/gallium/drivers/i915simple/i915_strings.c +++ /dev/null @@ -1,53 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915_context.h" -#include "i915_screen.h" -#include "i915_reg.h" - - -/** XXX temporary screen/pipe duplication here */ - - -static const char *i915_get_vendor( struct pipe_context *pipe ) -{ - return pipe->screen->get_vendor(pipe->screen); -} - - -static const char *i915_get_name( struct pipe_context *pipe ) -{ - return pipe->screen->get_name(pipe->screen); -} - - -void -i915_init_string_functions(struct i915_context *i915) -{ - i915->pipe.get_name = i915_get_name; - i915->pipe.get_vendor = i915_get_vendor; -} diff --git a/src/gallium/drivers/i965simple/brw_strings.c b/src/gallium/drivers/i965simple/brw_strings.c deleted file mode 100644 index a8f2ce582a..0000000000 --- a/src/gallium/drivers/i965simple/brw_strings.c +++ /dev/null @@ -1,51 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "brw_context.h" -#include "brw_reg.h" - -#include "pipe/p_screen.h" - - -static const char *brw_get_vendor( struct pipe_context *pipe ) -{ - return pipe->screen->get_vendor(pipe->screen); -} - - -static const char *brw_get_name( struct pipe_context *pipe ) -{ - return pipe->screen->get_name(pipe->screen); -} - - -void -brw_init_string_functions(struct brw_context *brw) -{ - brw->pipe.get_name = brw_get_name; - brw->pipe.get_vendor = brw_get_vendor; -} -- cgit v1.2.3 From 1c50ea2cd9ab8752793c99b4a7a2a6656bdde1ac Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 27 Feb 2008 13:40:23 -0800 Subject: cell: Use unified data cache for textures too --- src/gallium/drivers/cell/spu/spu_main.c | 2 + src/gallium/drivers/cell/spu/spu_main.h | 3 +- src/gallium/drivers/cell/spu/spu_texture.c | 184 +++++++++++------------------ 3 files changed, 72 insertions(+), 117 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index cc4bafdb3a..59300028d4 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -286,6 +286,8 @@ cmd_state_texture(const struct cell_command_texture *texture) { spu.texture.width, spu.texture.height, 0.0, 0.0}; spu.tex_size_mask = (vector unsigned int) { spu.texture.width - 1, spu.texture.height - 1, 0, 0 }; + spu.tex_size_x_mask = spu_splats(spu.texture.width - 1); + spu.tex_size_y_mask = spu_splats(spu.texture.height - 1); } diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index d14f1abbe7..a13edd1702 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -107,6 +107,8 @@ struct spu_global vector float tex_size; vector unsigned int tex_size_mask; /**< == int(size - 1) */ + vector unsigned int tex_size_x_mask; /**< == int(size - 1) */ + vector unsigned int tex_size_y_mask; /**< == int(size - 1) */ vector float (*sample_texture)(vector float texcoord); @@ -130,7 +132,6 @@ extern boolean Debug; #define TAG_INDEX_BUFFER 16 #define TAG_BATCH_BUFFER 17 #define TAG_MISC 18 -#define TAG_TEXTURE_TILE 19 #define TAG_DCACHE0 20 #define TAG_DCACHE1 21 #define TAG_DCACHE2 22 diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c index 3962aaa4a9..67eb08196a 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.c +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -31,19 +31,7 @@ #include "spu_texture.h" #include "spu_tile.h" #include "spu_colorpack.h" - - -/** - * Number of texture tiles to cache. - * Note that this will probably be the largest consumer of SPU local store/ - * memory for this driver! - */ -#define CACHE_SIZE 16 - -static tile_t tex_tiles[CACHE_SIZE] ALIGN16_ATTRIB; - -static vector unsigned int tex_tile_xy[CACHE_SIZE]; - +#include "spu_dcache.h" /** @@ -52,78 +40,60 @@ static vector unsigned int tex_tile_xy[CACHE_SIZE]; void invalidate_tex_cache(void) { - /* XXX memset? */ - uint i; - for (i = 0; i < CACHE_SIZE; i++) { - tex_tile_xy[i] = ((vector unsigned int) { ~0U, ~0U, ~0U, ~0U }); - } + spu_dcache_mark_dirty((unsigned) spu.texture.start, + 4 * spu.texture.width * spu.texture.height); } -/** - * Return the cache pos/index which corresponds to tile (tx,ty) - */ -static INLINE uint -cache_pos(vector unsigned int txty) +static uint +get_texel(vec_uint4 coordinate) { - uint pos = (spu_extract(txty,0) + spu_extract(txty,1) * 4) % CACHE_SIZE; - return pos; + vec_uint4 tmp; + unsigned x = spu_extract(coordinate, 0); + unsigned y = spu_extract(coordinate, 1); + const unsigned tiles_per_row = spu.texture.width / TILE_SIZE; + unsigned tile_offset = sizeof(tile_t) * ((y / TILE_SIZE * tiles_per_row) + + (x / TILE_SIZE)); + unsigned texel_offset = 4 * (((y % TILE_SIZE) * TILE_SIZE) + + (x % TILE_SIZE)); + + spu_dcache_fetch_unaligned((qword *) & tmp, + spu.texture.start + tile_offset + texel_offset, + 4); + return spu_extract(tmp, 0); } -/** - * Make sure the tile for texel (i,j) is present, return its position/index - * in the cache. - */ -static uint -get_tex_tile(vector unsigned int ij) +static void +get_four_texels(vec_uint4 x, vec_uint4 y, vec_uint4 *texels) { - /* tile address: tx,ty */ - const vector unsigned int txty = spu_rlmask(ij, -5); /* divide by 32 */ - const uint pos = cache_pos(txty); - - if ((spu_extract(tex_tile_xy[pos], 0) != spu_extract(txty, 0)) || - (spu_extract(tex_tile_xy[pos], 1) != spu_extract(txty, 1))) { - - /* texture cache miss, fetch tile from main memory */ - const uint tiles_per_row = spu.texture.width / TILE_SIZE; - const uint bytes_per_tile = sizeof(tile_t); - const void *src = (const ubyte *) spu.texture.start - + (spu_extract(txty,1) * tiles_per_row + spu_extract(txty,0)) * bytes_per_tile; - - printf("SPU %u: tex cache miss at %d, %d pos=%u old=%d,%d\n", - spu.init.id, - spu_extract(txty,0), - spu_extract(txty,1), - pos, - spu_extract(tex_tile_xy[pos],0), - spu_extract(tex_tile_xy[pos],1)); - - ASSERT_ALIGN16(tex_tiles[pos].ui); - ASSERT_ALIGN16(src); - - mfc_get(tex_tiles[pos].ui, /* dest */ - (unsigned int) src, - bytes_per_tile, /* size */ - TAG_TEXTURE_TILE, - 0, /* tid */ - 0 /* rid */); - - wait_on_mask(1 << TAG_TEXTURE_TILE); - - tex_tile_xy[pos] = txty; - } - else { -#if 0 - printf("SPU %u: tex cache HIT at %d, %d\n", - spu.init.id, tx, ty); -#endif - } - - return pos; + const unsigned texture_ea = (uintptr_t) spu.texture.start; + vec_uint4 tile_x = spu_rlmask(x, -5); + vec_uint4 tile_y = spu_rlmask(y, -5); + const qword offset_x = si_andi((qword) x, 0x1f); + const qword offset_y = si_andi((qword) y, 0x1f); + + const qword tiles_per_row = (qword) spu_splats(spu.texture.width / TILE_SIZE); + const qword tile_size = (qword) spu_splats(sizeof(tile_t)); + + qword tile_offset = si_mpya((qword) tile_y, tiles_per_row, (qword) tile_x); + tile_offset = si_mpy((qword) tile_offset, tile_size); + + qword texel_offset = si_a(si_mpyui(offset_y, 32), offset_x); + texel_offset = si_mpyui(texel_offset, 4); + + vec_uint4 offset = (vec_uint4) si_a(tile_offset, texel_offset); + + spu_dcache_fetch_unaligned((qword *) & texels[0], + texture_ea + spu_extract(offset, 0), 4); + spu_dcache_fetch_unaligned((qword *) & texels[1], + texture_ea + spu_extract(offset, 1), 4); + spu_dcache_fetch_unaligned((qword *) & texels[2], + texture_ea + spu_extract(offset, 2), 4); + spu_dcache_fetch_unaligned((qword *) & texels[3], + texture_ea + spu_extract(offset, 3), 4); } - /** * Get texture sample at texcoord. * XXX this is extremely primitive for now. @@ -134,9 +104,7 @@ sample_texture_nearest(vector float texcoord) vector float tc = spu_mul(texcoord, spu.tex_size); vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ itc = spu_and(itc, spu.tex_size_mask); /* mask (GL_REPEAT) */ - vector unsigned int ij = spu_and(itc, TILE_SIZE-1); /* intra tile addr */ - uint pos = get_tex_tile(itc); - uint texel = tex_tiles[pos].ui[spu_extract(ij, 1)][spu_extract(ij, 0)]; + uint texel = get_texel(itc); return spu_unpack_A8R8G8B8(texel); } @@ -144,49 +112,33 @@ sample_texture_nearest(vector float texcoord) vector float sample_texture_bilinear(vector float texcoord) { - static const vector unsigned int offset10 = {1, 0, 0, 0}; - static const vector unsigned int offset01 = {0, 1, 0, 0}; + static const vec_uint4 offset_x = {0, 0, 1, 1}; + static const vec_uint4 offset_y = {0, 1, 0, 1}; vector float tc = spu_mul(texcoord, spu.tex_size); tc = spu_add(tc, spu_splats(-0.5f)); /* half texel bias */ /* integer texcoords S,T: */ - vector unsigned int itc00 = spu_convtu(tc, 0); /* convert to int */ - vector unsigned int itc01 = spu_add(itc00, offset01); - vector unsigned int itc10 = spu_add(itc00, offset10); - vector unsigned int itc11 = spu_add(itc10, offset01); - - /* mask (GL_REPEAT) */ - itc00 = spu_and(itc00, spu.tex_size_mask); - itc01 = spu_and(itc01, spu.tex_size_mask); - itc10 = spu_and(itc10, spu.tex_size_mask); - itc11 = spu_and(itc11, spu.tex_size_mask); - - /* intra tile addr */ - vector unsigned int ij00 = spu_and(itc00, TILE_SIZE-1); - vector unsigned int ij01 = spu_and(itc01, TILE_SIZE-1); - vector unsigned int ij10 = spu_and(itc10, TILE_SIZE-1); - vector unsigned int ij11 = spu_and(itc11, TILE_SIZE-1); - - /* get tile cache positions */ - uint pos00 = get_tex_tile(itc00); - uint pos01, pos10, pos11; - if ((spu_extract(ij00, 0) < TILE_SIZE-1) && - (spu_extract(ij00, 1) < TILE_SIZE-1)) { - /* all texels are in the same tile */ - pos01 = pos10 = pos11 = pos00; - } - else { - pos01 = get_tex_tile(itc01); - pos10 = get_tex_tile(itc10); - pos11 = get_tex_tile(itc11); - } - - /* get texels from tiles and convert to float[4] */ - vector float texel00 = spu_unpack_A8R8G8B8(tex_tiles[pos00].ui[spu_extract(ij00, 1)][spu_extract(ij00, 0)]); - vector float texel01 = spu_unpack_A8R8G8B8(tex_tiles[pos01].ui[spu_extract(ij01, 1)][spu_extract(ij01, 0)]); - vector float texel10 = spu_unpack_A8R8G8B8(tex_tiles[pos10].ui[spu_extract(ij10, 1)][spu_extract(ij10, 0)]); - vector float texel11 = spu_unpack_A8R8G8B8(tex_tiles[pos11].ui[spu_extract(ij11, 1)][spu_extract(ij11, 0)]); + vec_uint4 itc = spu_convtu(tc, 0); /* convert to int */ + + vec_uint4 texels[4]; + + vec_uint4 x = spu_splats(spu_extract(itc, 0)); + vec_uint4 y = spu_splats(spu_extract(itc, 1)); + + x = spu_add(x, offset_x); + y = spu_add(y, offset_y); + + x = spu_and(x, spu.tex_size_x_mask); + y = spu_and(y, spu.tex_size_y_mask); + + get_four_texels(x, y, texels); + + vector float texel00 = spu_unpack_A8R8G8B8(spu_extract(texels[0], 0)); + vector float texel01 = spu_unpack_A8R8G8B8(spu_extract(texels[1], 0)); + vector float texel10 = spu_unpack_A8R8G8B8(spu_extract(texels[2], 0)); + vector float texel11 = spu_unpack_A8R8G8B8(spu_extract(texels[3], 0)); + /* Compute weighting factors in [0,1] * Multiply texcoord by 1024, AND with 1023, convert back to float. -- cgit v1.2.3 From fb40c5a9c7dc91c03f80780e0a09be0cade98705 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 15:06:04 -0700 Subject: gallium: added uses_kill field to tgsi_shader_info --- src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 3 +++ src/gallium/auxiliary/tgsi/util/tgsi_scan.h | 1 + 2 files changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c index a1cc681492..a973aeb62f 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -133,5 +133,8 @@ tgsi_scan_shader(const struct tgsi_token *tokens, } } + info->uses_kill = (info->opcode_count[TGSI_OPCODE_KIL] || + info->opcode_count[TGSI_OPCODE_KILP]); + tgsi_parse_free (&parse); } diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h index dc6dfd6d0b..d10d300c4d 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h @@ -57,6 +57,7 @@ struct tgsi_shader_info uint opcode_count[TGSI_OPCODE_LAST]; /**< opcode histogram */ boolean writes_z; /**< does fragment shader write Z value? */ + boolean uses_kill; /**< KIL or KILP instruction used? */ }; -- cgit v1.2.3 From c66ec5c7a2966df3e3456dfca3eb17c294b30dd5 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 15:06:55 -0700 Subject: gallium: remove uses_kill field from softpipe_shader --- src/gallium/drivers/softpipe/sp_quad.c | 2 +- src/gallium/drivers/softpipe/sp_state.h | 2 -- src/gallium/drivers/softpipe/sp_state_fs.c | 3 --- 3 files changed, 1 insertion(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad.c index 0aaf94021f..8603c1a367 100644 --- a/src/gallium/drivers/softpipe/sp_quad.c +++ b/src/gallium/drivers/softpipe/sp_quad.c @@ -60,7 +60,7 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp->depth_stencil->depth.enabled && sp->framebuffer.zsbuf && !sp->depth_stencil->alpha.enabled && - !sp->fs->uses_kill && + !sp->fs->info.uses_kill && !sp->fs->info.writes_z; /* build up the pipeline in reverse order... */ diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index b1070e185a..895976467c 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -66,8 +66,6 @@ struct sp_fragment_shader { struct tgsi_shader_info info; - boolean uses_kill; - void (*prepare)( const struct sp_fragment_shader *shader, struct tgsi_exec_machine *machine, struct tgsi_sampler *samplers); diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 2715dca0f0..eb641ed321 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -64,9 +64,6 @@ softpipe_create_fs_state(struct pipe_context *pipe, /* get/save the summary info for this shader */ tgsi_scan_shader(templ->tokens, &state->info); - /* convenience field */ - state->uses_kill = (state->info.opcode_count[TGSI_OPCODE_KIL] || - state->info.opcode_count[TGSI_OPCODE_KILP]); return state; } -- cgit v1.2.3 From 80a9b5e1d962c17530e3bcb34c0d1ac4aae8fa7f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 15:11:12 -0700 Subject: gallium: remove unnecessary tgsi_interp_coef decl --- src/gallium/drivers/softpipe/sp_state.h | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 895976467c..3943d4ed2b 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -53,7 +53,6 @@ struct tgsi_sampler; -struct tgsi_interp_coef; struct tgsi_exec_machine; -- cgit v1.2.3 From 0e1e1f12f47d5b1d49f68930b05eadf1143e1396 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 15:45:41 -0700 Subject: gallium/i915: remove unneeded assignment --- src/gallium/drivers/i915simple/i915_texture.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 9cdf3418a9..ef5adff550 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -491,7 +491,6 @@ i915_texture_create_screen(struct pipe_screen *screen, tex->base = *templat; tex->base.refcount = 1; - tex->base.pipe = NULL; tex->base.screen = screen; if (i915screen->is_i945 ? i945_miptree_layout(tex) : -- cgit v1.2.3 From 3197ad5a56ee94773f974ac727b316c5adfe1b6f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 15:47:24 -0700 Subject: gallium: added file_max[] array to tgsi_shader_info Records the highest index of a declared register. --- src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 5 ++++- src/gallium/auxiliary/tgsi/util/tgsi_scan.h | 1 + 2 files changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c index a973aeb62f..8c886edc65 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -46,10 +46,12 @@ void tgsi_scan_shader(const struct tgsi_token *tokens, struct tgsi_shader_info *info) { - uint procType; + uint procType, i; struct tgsi_parse_context parse; memset(info, 0, sizeof(*info)); + for (i = 0; i < TGSI_FILE_COUNT; i++) + info->file_max[i] = -1; /** ** Setup to begin parsing input shader @@ -97,6 +99,7 @@ tgsi_scan_shader(const struct tgsi_token *tokens, /* only first 32 regs will appear in this bitfield */ info->file_mask[file] |= (1 << i); info->file_count[file]++; + info->file_max[file] = MAX2(info->file_max[file], i); if (file == TGSI_FILE_INPUT) { info->input_semantic_name[info->num_inputs] diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h index d10d300c4d..563ee900fa 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h @@ -51,6 +51,7 @@ struct tgsi_shader_info uint file_mask[TGSI_FILE_COUNT]; /**< bitmask of declared registers */ uint file_count[TGSI_FILE_COUNT]; /**< number of declared registers */ + int file_max[TGSI_FILE_COUNT]; /**< highest index of declared registers */ uint immediate_count; /**< number of immediates declared */ -- cgit v1.2.3 From 681b78fc60b2e60cf9f84802932bf9d2defd28c2 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 15:48:00 -0700 Subject: gallium: remove unnecessary assignment --- src/gallium/drivers/softpipe/sp_texture.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 7c02765313..c605ed925b 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -91,7 +91,6 @@ softpipe_texture_create_screen(struct pipe_screen *screen, spt->base = *templat; spt->base.refcount = 1; - spt->base.pipe = NULL; spt->base.screen = screen; softpipe_texture_layout(spt); -- cgit v1.2.3 From 679b6cf0a0e662513c8d7732049c44916e0e9e86 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:00:04 -0700 Subject: gallium: include p_compiler.h instead of p_util.h --- src/gallium/auxiliary/tgsi/util/tgsi_scan.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h index 563ee900fa..0530bc6b51 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h @@ -29,7 +29,7 @@ #define TGSI_SCAN_H -#include "pipe/p_util.h" +#include "pipe/p_compiler.h" #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" -- cgit v1.2.3 From 7df26d76d2a37e9e828296bfbcf7cec04bfbe233 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:01:35 -0700 Subject: gallium: include p_util.h --- src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c index 8c886edc65..b7bbff1689 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -37,6 +37,7 @@ #include "tgsi/util/tgsi_parse.h" #include "tgsi/util/tgsi_build.h" +#include "pipe/p_util.h" -- cgit v1.2.3 From cddeca51adf0d2b736a223e47b60f6ef3be85bff Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:02:58 -0700 Subject: gallium: remove dependencies on pipe_shader_state's semantic info Use tgsi_scan_shader() to populate a tgsi_shader_info struct and use that instead. --- src/gallium/auxiliary/draw/draw_aaline.c | 2 ++ src/gallium/auxiliary/draw/draw_aapoint.c | 6 ++++-- src/gallium/auxiliary/draw/draw_clip.c | 8 ++++---- src/gallium/auxiliary/draw/draw_context.c | 8 ++++---- src/gallium/auxiliary/draw/draw_flatshade.c | 8 ++++---- src/gallium/auxiliary/draw/draw_private.h | 3 +++ src/gallium/auxiliary/draw/draw_pstipple.c | 2 ++ src/gallium/auxiliary/draw/draw_twoside.c | 12 ++++++------ src/gallium/auxiliary/draw/draw_vertex_fetch.c | 2 +- src/gallium/auxiliary/draw/draw_vertex_shader.c | 17 +++++++++-------- src/gallium/auxiliary/draw/draw_vs_exec.c | 2 +- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- src/gallium/auxiliary/draw/draw_wide_prims.c | 8 ++++---- 13 files changed, 45 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index be6cfd3b6a..51140388f0 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -340,9 +340,11 @@ generate_aaline_fs(struct aaline_stage *aaline) tgsi_dump(aaline_fs.tokens, 0); #endif +#if 1 /* XXX remove */ aaline_fs.input_semantic_name[aaline_fs.num_inputs] = TGSI_SEMANTIC_GENERIC; aaline_fs.input_semantic_index[aaline_fs.num_inputs] = transform.maxGeneric + 1; aaline_fs.num_inputs++; +#endif aaline->fs->aaline_fs = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index d16adb2505..d48a416899 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -514,9 +514,11 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) tgsi_dump(aapoint_fs.tokens, 0); #endif +#if 1 /* XXX remove */ aapoint_fs.input_semantic_name[aapoint_fs.num_inputs] = TGSI_SEMANTIC_GENERIC; aapoint_fs.input_semantic_index[aapoint_fs.num_inputs] = transform.maxGeneric + 1; aapoint_fs.num_inputs++; +#endif aapoint->fs->aapoint_fs = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs); @@ -694,8 +696,8 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header) /* find PSIZ vertex output */ const struct draw_vertex_shader *vs = draw->vertex_shader; uint i; - for (i = 0; i < vs->state->num_outputs; i++) { - if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { aapoint->psize_slot = i; break; } diff --git a/src/gallium/auxiliary/draw/draw_clip.c b/src/gallium/auxiliary/draw/draw_clip.c index e3051507ea..200152ecab 100644 --- a/src/gallium/auxiliary/draw/draw_clip.c +++ b/src/gallium/auxiliary/draw/draw_clip.c @@ -409,13 +409,13 @@ clip_init_state( struct draw_stage *stage ) clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; if (clipper->flat) { - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + const struct draw_vertex_shader *vs = stage->draw->vertex_shader; uint i; clipper->num_color_attribs = 0; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || - vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { clipper->color_attribs[clipper->num_color_attribs++] = i; } } diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 48f00946ea..64ada8ce04 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -262,11 +262,11 @@ int draw_find_vs_output(struct draw_context *draw, uint semantic_name, uint semantic_index) { - const struct pipe_shader_state *vs = draw->vertex_shader->state; + const struct draw_vertex_shader *vs = draw->vertex_shader; uint i; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == semantic_name && - vs->output_semantic_index[i] == semantic_index) + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == semantic_name && + vs->info.output_semantic_index[i] == semantic_index) return i; } diff --git a/src/gallium/auxiliary/draw/draw_flatshade.c b/src/gallium/auxiliary/draw/draw_flatshade.c index 4398abbc60..ccad71d695 100644 --- a/src/gallium/auxiliary/draw/draw_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_flatshade.c @@ -128,14 +128,14 @@ static void flatshade_point( struct draw_stage *stage, static void flatshade_init_state( struct draw_stage *stage ) { struct flat_stage *flat = flat_stage(stage); - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + const struct draw_vertex_shader *vs = stage->draw->vertex_shader; uint i; /* Find which vertex shader outputs are colors, make a list */ flat->num_color_attribs = 0; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR || - vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { flat->color_attribs[flat->num_color_attribs++] = i; } } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 03e3d3a66b..e988e71d23 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -46,6 +46,7 @@ #include "rtasm/rtasm_x86sse.h" #include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_scan.h" struct pipe_context; @@ -134,6 +135,8 @@ struct draw_vertex_shader { */ const struct pipe_shader_state *state; + struct tgsi_shader_info info; + void (*prepare)( struct draw_vertex_shader *shader, struct draw_context *draw ); diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index efc88bf038..f6200aa820 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -330,11 +330,13 @@ generate_pstip_fs(struct pstip_stage *pstip) pstip->sampler_unit = transform.maxSampler + 1; +#if 1 /* XXX remove */ if (transform.wincoordInput < 0) { pstip_fs.input_semantic_name[pstip_fs.num_inputs] = TGSI_SEMANTIC_POSITION; pstip_fs.input_semantic_index[pstip_fs.num_inputs] = (ubyte)transform.maxInput; pstip_fs.num_inputs++; } +#endif pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs); } diff --git a/src/gallium/auxiliary/draw/draw_twoside.c b/src/gallium/auxiliary/draw/draw_twoside.c index 1c38957987..3debaac282 100644 --- a/src/gallium/auxiliary/draw/draw_twoside.c +++ b/src/gallium/auxiliary/draw/draw_twoside.c @@ -119,7 +119,7 @@ static void twoside_first_tri( struct draw_stage *stage, struct prim_header *header ) { struct twoside_stage *twoside = twoside_stage(stage); - const struct pipe_shader_state *vs = stage->draw->vertex_shader->state; + const struct draw_vertex_shader *vs = stage->draw->vertex_shader; uint i; twoside->attrib_front0 = 0; @@ -128,15 +128,15 @@ static void twoside_first_tri( struct draw_stage *stage, twoside->attrib_back1 = 0; /* Find which vertex shader outputs are front/back colors */ - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { - if (vs->output_semantic_index[i] == 0) + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { + if (vs->info.output_semantic_index[i] == 0) twoside->attrib_front0 = i; else twoside->attrib_front1 = i; } - if (vs->output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - if (vs->output_semantic_index[i] == 0) + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + if (vs->info.output_semantic_index[i] == 0) twoside->attrib_back0 = i; else twoside->attrib_back1 = i; diff --git a/src/gallium/auxiliary/draw/draw_vertex_fetch.c b/src/gallium/auxiliary/draw/draw_vertex_fetch.c index e13df04605..cb8cdd04a3 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_fetch.c +++ b/src/gallium/auxiliary/draw/draw_vertex_fetch.c @@ -473,7 +473,7 @@ void draw_update_vertex_fetch( struct draw_context *draw ) if (!draw->vertex_shader) return; - nr_attrs = draw->vertex_shader->state->num_inputs; + nr_attrs = draw->vertex_shader->info.num_inputs; for (i = 0; i < nr_attrs; i++) { unsigned buf = draw->vertex_element[i].vertex_buffer_index; diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index 5d2f5c9c43..1e95355555 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -91,15 +91,16 @@ draw_create_vertex_shader(struct draw_context *draw, struct draw_vertex_shader *vs; vs = draw_create_vs_llvm( draw, shader ); - if (vs) - return vs; + if (!vs) { + vs = draw_create_vs_sse( draw, shader ); + if (!vs) { + vs = draw_create_vs_exec( draw, shader ); + } + } + assert(vs); - vs = draw_create_vs_sse( draw, shader ); - if (vs) - return vs; + tgsi_scan_shader(shader->tokens, &vs->info); - vs = draw_create_vs_exec( draw, shader ); - assert(vs); return vs; } @@ -111,7 +112,7 @@ draw_bind_vertex_shader(struct draw_context *draw, draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); draw->vertex_shader = dvs; - draw->num_vs_outputs = dvs->state->num_outputs; + draw->num_vs_outputs = dvs->info.num_outputs; tgsi_exec_machine_init(&draw->machine); diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 8588879400..583812aadd 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -103,7 +103,7 @@ vs_exec_run( struct draw_vertex_shader *shader, const float *trans = draw->viewport.translate; assert(count <= 4); - assert(draw->vertex_shader->state->output_semantic_name[0] + assert(draw->vertex_shader->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); machine->Consts = (float (*)[4]) draw->user.constants; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 11ef0c503d..0b8bc2bf14 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -119,7 +119,7 @@ vs_sse_run( struct draw_vertex_shader *base, const float *trans = draw->viewport.translate; assert(count <= 4); - assert(draw->vertex_shader->state->output_semantic_name[0] + assert(draw->vertex_shader->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); /* Consts does not require 16 byte alignment. */ diff --git a/src/gallium/auxiliary/draw/draw_wide_prims.c b/src/gallium/auxiliary/draw/draw_wide_prims.c index d967810dd4..d6bff110b4 100644 --- a/src/gallium/auxiliary/draw/draw_wide_prims.c +++ b/src/gallium/auxiliary/draw/draw_wide_prims.c @@ -278,8 +278,8 @@ static void wide_first_point( struct draw_stage *stage, /* find vertex shader texcoord outputs */ const struct draw_vertex_shader *vs = draw->vertex_shader; uint i, j = 0; - for (i = 0; i < vs->state->num_outputs; i++) { - if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { wide->texcoord_slot[j] = i; wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; j++; @@ -294,8 +294,8 @@ static void wide_first_point( struct draw_stage *stage, /* find PSIZ vertex output */ const struct draw_vertex_shader *vs = draw->vertex_shader; uint i; - for (i = 0; i < vs->state->num_outputs; i++) { - if (vs->state->output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { wide->psize_slot = i; break; } -- cgit v1.2.3 From 6b9a7eb460fe0a9c958b837f2ed49c2d4e303ebc Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:03:40 -0700 Subject: gallium: remove dependencies on pipe_shader_state's semantic info --- src/gallium/drivers/softpipe/sp_prim_setup.c | 4 ++-- src/gallium/drivers/softpipe/sp_quad_fs.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 17284539b0..5c8b0bf69c 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -1163,13 +1163,13 @@ static void setup_begin( struct draw_stage *stage ) { struct setup_stage *setup = setup_stage(stage); struct softpipe_context *sp = setup->softpipe; - const struct pipe_shader_state *fs = &setup->softpipe->fs->shader; + const struct sp_fragment_shader *fs = &setup->softpipe->fs; if (sp->dirty) { softpipe_update_derived(sp); } - setup->quad.nr_attrs = fs->num_inputs; + setup->quad.nr_attrs = fs->info.num_inputs; sp->quad.first->begin(sp->quad.first); diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 1794fb5a61..1fbb2e38c4 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -148,7 +148,7 @@ static void shade_begin(struct quad_stage *qs) /* find output slots for depth, color */ qss->colorOutSlot = -1; qss->depthOutSlot = -1; - for (i = 0; i < qss->stage.softpipe->fs->shader.num_outputs; i++) { + for (i = 0; i < qss->stage.softpipe->fs->info.num_outputs; i++) { switch (qss->stage.softpipe->fs->info.output_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: qss->depthOutSlot = i; -- cgit v1.2.3 From ea7e86dd4d1e7dbef2642da73bb1980723ae49ef Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:05:16 -0700 Subject: gallium/i965: remove dependencies on pipe_shader_state's semantic info The brw_shader_info struct is rendundant and could be removed... --- src/gallium/drivers/i965simple/brw_context.h | 8 ++++++-- src/gallium/drivers/i965simple/brw_draw_upload.c | 2 +- src/gallium/drivers/i965simple/brw_sf.c | 2 +- src/gallium/drivers/i965simple/brw_shader_info.c | 5 +++-- src/gallium/drivers/i965simple/brw_state.c | 9 +++++++-- src/gallium/drivers/i965simple/brw_wm_decl.c | 8 ++++---- 6 files changed, 22 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index 65664d853d..fbc89c889b 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -38,6 +38,8 @@ #include "pipe/p_defines.h" #include "pipe/p_state.h" +#include "tgsi/util/tgsi_scan.h" + #include "brw_structs.h" #include "brw_winsys.h" @@ -203,7 +205,8 @@ struct brw_shader_info { struct brw_vertex_program { struct pipe_shader_state program; - struct brw_shader_info info; + struct tgsi_shader_info info; + struct brw_shader_info info2; /* XXX get rid of this */ int id; }; @@ -211,7 +214,8 @@ struct brw_vertex_program { struct brw_fragment_program { struct pipe_shader_state program; - struct brw_shader_info info; + struct tgsi_shader_info info; + struct brw_shader_info info2; /* XXX get rid of this */ boolean UsesDepth; boolean UsesKill; diff --git a/src/gallium/drivers/i965simple/brw_draw_upload.c b/src/gallium/drivers/i965simple/brw_draw_upload.c index aa85d93866..9c0c78c236 100644 --- a/src/gallium/drivers/i965simple/brw_draw_upload.c +++ b/src/gallium/drivers/i965simple/brw_draw_upload.c @@ -256,7 +256,7 @@ boolean brw_upload_vertex_elements( struct brw_context *brw ) struct brw_vertex_element_packet vep; unsigned i; - unsigned nr_enabled = brw->attribs.VertexProgram->program.num_inputs; + unsigned nr_enabled = brw->attribs.VertexProgram->info.num_inputs; memset(&vep, 0, sizeof(vep)); diff --git a/src/gallium/drivers/i965simple/brw_sf.c b/src/gallium/drivers/i965simple/brw_sf.c index 7c83b81c85..54ce5ed9f1 100644 --- a/src/gallium/drivers/i965simple/brw_sf.c +++ b/src/gallium/drivers/i965simple/brw_sf.c @@ -133,7 +133,7 @@ static void upload_sf_prog( struct brw_context *brw ) key.vp_output_count = brw->vs.prog_data->outputs_written; /* BRW_NEW_FS */ - key.fp_input_count = brw->attribs.FragmentProgram->info.nr_regs[TGSI_FILE_INPUT]; + key.fp_input_count = brw->attribs.FragmentProgram->info2.nr_regs[TGSI_FILE_INPUT]; /* BRW_NEW_REDUCED_PRIMITIVE */ diff --git a/src/gallium/drivers/i965simple/brw_shader_info.c b/src/gallium/drivers/i965simple/brw_shader_info.c index a762a870fe..e7e063dead 100644 --- a/src/gallium/drivers/i965simple/brw_shader_info.c +++ b/src/gallium/drivers/i965simple/brw_shader_info.c @@ -6,8 +6,9 @@ #include "tgsi/util/tgsi_parse.h" - - +/** + * XXX try to get rid of this. See tgsi_scan_shader() and tgsi_shader_info. + */ void brw_shader_info(const struct tgsi_token *tokens, struct brw_shader_info *info ) { diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 7466fdc403..6e46465200 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -175,8 +175,10 @@ static void * brw_create_fs_state(struct pipe_context *pipe, brw_fp->program = *shader; brw_fp->id = brw_context(pipe)->program_id++; + tgsi_scan_shader(shader->tokens, &brw_fp->info); + brw_shader_info(shader->tokens, - &brw_fp->info); + &brw_fp->info2); tgsi_dump(shader->tokens, 0); @@ -211,8 +213,11 @@ static void *brw_create_vs_state(struct pipe_context *pipe, */ brw_vp->program = *shader; brw_vp->id = brw_context(pipe)->program_id++; + + tgsi_scan_shader(shader->tokens, &brw_vp->info); + brw_shader_info(shader->tokens, - &brw_vp->info); + &brw_vp->info2); tgsi_dump(shader->tokens, 0); diff --git a/src/gallium/drivers/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c index 97418a52e7..afea042bf0 100644 --- a/src/gallium/drivers/i965simple/brw_wm_decl.c +++ b/src/gallium/drivers/i965simple/brw_wm_decl.c @@ -259,7 +259,7 @@ static void prealloc_reg(struct brw_wm_compile *c) /* Then a copy of our part of the CURBE entry: */ { - int nr_constants = c->fp->info.nr_regs[TGSI_FILE_CONSTANT]; + int nr_constants = c->fp->info2.nr_regs[TGSI_FILE_CONSTANT]; int index = 0; c->prog_data.max_const = 4*nr_constants; @@ -282,7 +282,7 @@ static void prealloc_reg(struct brw_wm_compile *c) /* Next we receive the plane coefficients for parameter * interpolation: */ - for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_INPUT]; i++) { + for (i = 0; i < c->fp->info2.nr_regs[TGSI_FILE_INPUT]; i++) { c->payload_coef[i] = brw_vec8_grf(c->reg_index, 0); c->reg_index += 2; } @@ -302,11 +302,11 @@ static void prealloc_reg(struct brw_wm_compile *c) /* Now allocate room for the interpolated inputs and staging * registers for the outputs: */ - for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_INPUT]; i++) + for (i = 0; i < c->fp->info2.nr_regs[TGSI_FILE_INPUT]; i++) for (j = 0; j < 4; j++) c->wm_regs[TGSI_FILE_INPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); - for (i = 0; i < c->fp->info.nr_regs[TGSI_FILE_OUTPUT]; i++) + for (i = 0; i < c->fp->info2.nr_regs[TGSI_FILE_OUTPUT]; i++) for (j = 0; j < 4; j++) c->wm_regs[TGSI_FILE_OUTPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); -- cgit v1.2.3 From bad54d0b4dbe62aed6fad1d2725f7fe52a987440 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:09:17 -0700 Subject: gallium/i965: remove UsesDepth, UsesKill - use tgsi_shader_info instead --- src/gallium/drivers/i965simple/brw_context.h | 2 -- src/gallium/drivers/i965simple/brw_wm.c | 4 ++-- src/gallium/drivers/i965simple/brw_wm_state.c | 4 ++-- 3 files changed, 4 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index fbc89c889b..170d50c5c3 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -218,8 +218,6 @@ struct brw_fragment_program { struct brw_shader_info info2; /* XXX get rid of this */ boolean UsesDepth; - boolean UsesKill; - boolean ComputesDepth; int id; }; diff --git a/src/gallium/drivers/i965simple/brw_wm.c b/src/gallium/drivers/i965simple/brw_wm.c index 539b170744..1c4b5b5ede 100644 --- a/src/gallium/drivers/i965simple/brw_wm.c +++ b/src/gallium/drivers/i965simple/brw_wm.c @@ -94,11 +94,11 @@ static void brw_wm_populate_key( struct brw_context *brw, /* Build the index for table lookup */ /* BRW_NEW_DEPTH_STENCIL */ - if (fp->UsesKill || + if (fp->info.uses_kill || brw->attribs.DepthStencil->alpha.enabled) lookup |= IZ_PS_KILL_ALPHATEST_BIT; - if (fp->ComputesDepth) + if (fp->info.writes_z) lookup |= IZ_PS_COMPUTES_DEPTH_BIT; if (brw->attribs.DepthStencil->depth.enabled) diff --git a/src/gallium/drivers/i965simple/brw_wm_state.c b/src/gallium/drivers/i965simple/brw_wm_state.c index 5ccd488842..f3aa36b07f 100644 --- a/src/gallium/drivers/i965simple/brw_wm_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_state.c @@ -117,11 +117,11 @@ static void upload_wm_unit(struct brw_context *brw ) if (fp->UsesDepth) wm.wm5.program_uses_depth = 1; /* as far as we can tell */ - if (fp->ComputesDepth) + if (fp->info.writes_z) wm.wm5.program_computes_depth = 1; /* BRW_NEW_ALPHA_TEST */ - if (fp->UsesKill || + if (fp->info.uses_kill || brw->attribs.DepthStencil->alpha.enabled) wm.wm5.program_uses_killpixel = 1; -- cgit v1.2.3 From 7ba1afb03308685eb07d6b88184906ac42f60c2b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:11:14 -0700 Subject: gallium/i965: added const to silence warnings --- src/gallium/drivers/i965simple/brw_context.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index 170d50c5c3..642db5b5a2 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -504,7 +504,7 @@ struct brw_context /* Arrays with buffer objects to copy non-bufferobj arrays into * for upload: */ - struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX]; + const struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX]; struct brw_vertex_element_state inputs[PIPE_ATTRIB_MAX]; -- cgit v1.2.3 From f504d87240542016213569b5da89e251adebc31d Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:11:26 -0700 Subject: gallium/i965: silence warnings --- src/gallium/drivers/i965simple/brw_surface.c | 1 + src/gallium/drivers/i965simple/brw_vs_emit.c | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index dc4846d39f..c99a91dcf7 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -164,6 +164,7 @@ brw_surface_fill(struct pipe_context *pipe, void brw_init_surface_functions(struct brw_context *brw) { + (void) brw_surface_data; /* silence warning */ brw->pipe.surface_copy = brw_surface_copy; brw->pipe.surface_fill = brw_surface_fill; } diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c index 05df4860ed..33bbdc95c6 100644 --- a/src/gallium/drivers/i965simple/brw_vs_emit.c +++ b/src/gallium/drivers/i965simple/brw_vs_emit.c @@ -1051,7 +1051,7 @@ static void process_instruction(struct brw_vs_compile *c, { struct brw_reg args[3], dst; struct brw_compile *p = &c->func; - struct brw_indirect stack_index = brw_indirect(0, 0); + /*struct brw_indirect stack_index = brw_indirect(0, 0);*/ unsigned i; unsigned index; unsigned file; -- cgit v1.2.3 From dacf91fe587a777eed95b9767bc6b4ccdc7de71c Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:22:08 -0700 Subject: gallium/i965: remove brw_shader_info struct The info it contained is now found in tgsi_shader_info. Added a few assertions to catch potential misunderstandings about register counts vs. highest register index used. --- src/gallium/drivers/i965simple/Makefile | 1 - src/gallium/drivers/i965simple/SConscript | 1 - src/gallium/drivers/i965simple/brw_context.h | 13 +------------ src/gallium/drivers/i965simple/brw_sf.c | 2 +- src/gallium/drivers/i965simple/brw_shader_info.c | 2 +- src/gallium/drivers/i965simple/brw_state.c | 5 ++++- src/gallium/drivers/i965simple/brw_state.h | 7 ------- src/gallium/drivers/i965simple/brw_wm_decl.c | 18 ++++++++++++++---- 8 files changed, 21 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/Makefile b/src/gallium/drivers/i965simple/Makefile index 8589ebdf96..e97146e57c 100644 --- a/src/gallium/drivers/i965simple/Makefile +++ b/src/gallium/drivers/i965simple/Makefile @@ -30,7 +30,6 @@ C_SOURCES = \ brw_sf.c \ brw_sf_emit.c \ brw_sf_state.c \ - brw_shader_info.c \ brw_state.c \ brw_state_batch.c \ brw_state_cache.c \ diff --git a/src/gallium/drivers/i965simple/SConscript b/src/gallium/drivers/i965simple/SConscript index e0cc78826e..c0825c4de3 100644 --- a/src/gallium/drivers/i965simple/SConscript +++ b/src/gallium/drivers/i965simple/SConscript @@ -29,7 +29,6 @@ i965simple = env.ConvenienceLibrary( 'brw_sf.c', 'brw_sf_emit.c', 'brw_sf_state.c', - 'brw_shader_info.c', 'brw_state.c', 'brw_state_batch.c', 'brw_state_cache.c', diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index 642db5b5a2..4da3a8cffc 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -197,33 +197,22 @@ struct brw_state_flags { }; -struct brw_shader_info { - int nr_regs[8]; /* TGSI_FILE_* */ -}; - - - struct brw_vertex_program { struct pipe_shader_state program; struct tgsi_shader_info info; - struct brw_shader_info info2; /* XXX get rid of this */ int id; }; - struct brw_fragment_program { struct pipe_shader_state program; struct tgsi_shader_info info; - struct brw_shader_info info2; /* XXX get rid of this */ - boolean UsesDepth; + boolean UsesDepth; /* XXX add this to tgsi_shader_info? */ int id; }; - - struct pipe_setup_linkage { struct { unsigned vp_output:5; diff --git a/src/gallium/drivers/i965simple/brw_sf.c b/src/gallium/drivers/i965simple/brw_sf.c index 54ce5ed9f1..c3b815a82b 100644 --- a/src/gallium/drivers/i965simple/brw_sf.c +++ b/src/gallium/drivers/i965simple/brw_sf.c @@ -133,7 +133,7 @@ static void upload_sf_prog( struct brw_context *brw ) key.vp_output_count = brw->vs.prog_data->outputs_written; /* BRW_NEW_FS */ - key.fp_input_count = brw->attribs.FragmentProgram->info2.nr_regs[TGSI_FILE_INPUT]; + key.fp_input_count = brw->attribs.FragmentProgram->info.file_max[TGSI_FILE_INPUT] + 1; /* BRW_NEW_REDUCED_PRIMITIVE */ diff --git a/src/gallium/drivers/i965simple/brw_shader_info.c b/src/gallium/drivers/i965simple/brw_shader_info.c index e7e063dead..f4694a4433 100644 --- a/src/gallium/drivers/i965simple/brw_shader_info.c +++ b/src/gallium/drivers/i965simple/brw_shader_info.c @@ -7,7 +7,7 @@ /** - * XXX try to get rid of this. See tgsi_scan_shader() and tgsi_shader_info. + * XXX this obsolete new and no longer compiled. */ void brw_shader_info(const struct tgsi_token *tokens, struct brw_shader_info *info ) diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 6e46465200..6744a8aa4f 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -177,8 +177,10 @@ static void * brw_create_fs_state(struct pipe_context *pipe, tgsi_scan_shader(shader->tokens, &brw_fp->info); +#if 0 brw_shader_info(shader->tokens, &brw_fp->info2); +#endif tgsi_dump(shader->tokens, 0); @@ -216,9 +218,10 @@ static void *brw_create_vs_state(struct pipe_context *pipe, tgsi_scan_shader(shader->tokens, &brw_vp->info); +#if 0 brw_shader_info(shader->tokens, &brw_vp->info2); - +#endif tgsi_dump(shader->tokens, 0); return (void *)brw_vp; diff --git a/src/gallium/drivers/i965simple/brw_state.h b/src/gallium/drivers/i965simple/brw_state.h index 258e9a556e..de0a6371b8 100644 --- a/src/gallium/drivers/i965simple/brw_state.h +++ b/src/gallium/drivers/i965simple/brw_state.h @@ -148,11 +148,4 @@ void brw_invalidate_pools( struct brw_context *brw ); void brw_clear_batch_cache_flush( struct brw_context *brw ); -/* brw_shader_info.c - */ - -void brw_shader_info(const struct tgsi_token *tokens, - struct brw_shader_info *info ); - - #endif diff --git a/src/gallium/drivers/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c index afea042bf0..b50d768062 100644 --- a/src/gallium/drivers/i965simple/brw_wm_decl.c +++ b/src/gallium/drivers/i965simple/brw_wm_decl.c @@ -259,9 +259,12 @@ static void prealloc_reg(struct brw_wm_compile *c) /* Then a copy of our part of the CURBE entry: */ { - int nr_constants = c->fp->info2.nr_regs[TGSI_FILE_CONSTANT]; + int nr_constants = c->fp->info.file_max[TGSI_FILE_CONSTANT] + 1; int index = 0; + /* XXX number of constants, or highest numbered constant? */ + assert(nr_constants == c->fp->info.file_count[TGSI_FILE_CONSTANT]); + c->prog_data.max_const = 4*nr_constants; for (i = 0; i < nr_constants; i++) { for (j = 0; j < 4; j++, index++) @@ -282,7 +285,8 @@ static void prealloc_reg(struct brw_wm_compile *c) /* Next we receive the plane coefficients for parameter * interpolation: */ - for (i = 0; i < c->fp->info2.nr_regs[TGSI_FILE_INPUT]; i++) { + assert(c->fp->info.file_max[TGSI_FILE_INPUT] == c->fp->info.num_inputs); + for (i = 0; i < c->fp->info.file_max[TGSI_FILE_INPUT] + 1; i++) { c->payload_coef[i] = brw_vec8_grf(c->reg_index, 0); c->reg_index += 2; } @@ -302,11 +306,17 @@ static void prealloc_reg(struct brw_wm_compile *c) /* Now allocate room for the interpolated inputs and staging * registers for the outputs: */ - for (i = 0; i < c->fp->info2.nr_regs[TGSI_FILE_INPUT]; i++) + /* XXX do we want to loop over the _number_ of inputs/outputs or loop + * to the highest input/output index that's used? + * Probably the same, actually. + */ + assert(c->fp->info.file_max[TGSI_FILE_INPUT] + 1 == c->fp->info.num_inputs); + assert(c->fp->info.file_max[TGSI_FILE_OUTPUT] + 1 == c->fp->info.num_outputs); + for (i = 0; i < c->fp->info.file_max[TGSI_FILE_INPUT] + 1; i++) for (j = 0; j < 4; j++) c->wm_regs[TGSI_FILE_INPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); - for (i = 0; i < c->fp->info2.nr_regs[TGSI_FILE_OUTPUT]; i++) + for (i = 0; i < c->fp->info.file_max[TGSI_FILE_OUTPUT] + 1; i++) for (j = 0; j < 4; j++) c->wm_regs[TGSI_FILE_OUTPUT][i][j] = brw_vec8_grf( c->reg_index++, 0 ); -- cgit v1.2.3 From d612b6fa9b5674e001755265e37924815646ad1a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 16:17:15 -0700 Subject: cell: fix minor get_tex_surface() breakage --- src/gallium/drivers/cell/ppu/cell_texture.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index 28cadad6ed..bf6f6d3058 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -210,6 +210,7 @@ static void cell_tile_texture(struct cell_context *cell, struct cell_texture *texture) { + struct pipe_screen *screen = cell->pipe.screen; uint face = 0, level = 0, zslice = 0; struct pipe_surface *surf; const uint w = texture->base.width[0], h = texture->base.height[0]; @@ -221,7 +222,7 @@ cell_tile_texture(struct cell_context *cell, assert(w % TILE_SIZE == 0); assert(h % TILE_SIZE == 0); - surf = cell_get_tex_surface(&cell->pipe, &texture->base, face, level, zslice); + surf = screen->get_tex_surface(screen, &texture->base, face, level, zslice); ASSERT(surf); src = (const uint *) pipe_surface_map(surf); -- cgit v1.2.3 From 308f11f9913d53944924e532e586d99070ccbe0b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 28 Feb 2008 08:17:53 +0900 Subject: scons: Build mesa source with c99. --- src/mesa/SConscript | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index db18c61fac..a77c3de6c7 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -12,6 +12,11 @@ env.Append(CPPPATH = [ '#/src/mesa/main', ]) +env.Append(CFLAGS = [ + '-std=c99', +]) + + ####################################################################### # Core sources -- cgit v1.2.3 From 46da2f42a8e1bd88086cc17afc58738956d8b699 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 17:21:01 -0700 Subject: gallium/i965: added const to silence warning --- src/gallium/drivers/i965simple/brw_vs.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_vs.h b/src/gallium/drivers/i965simple/brw_vs.h index 0e58f043b0..070f9dfcae 100644 --- a/src/gallium/drivers/i965simple/brw_vs.h +++ b/src/gallium/drivers/i965simple/brw_vs.h @@ -52,7 +52,7 @@ struct brw_vs_compile { struct brw_vs_prog_key key; struct brw_vs_prog_data prog_data; - struct brw_vertex_program *vp; + const struct brw_vertex_program *vp; unsigned nr_inputs; -- cgit v1.2.3 From 627efcaa8009bb7ed6a7f266f8122df800bb2706 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 17:21:29 -0700 Subject: gallium/i965: remove more dependencies on pipe_shader_state semantic info --- src/gallium/drivers/i965simple/brw_vs.c | 4 ++-- src/gallium/drivers/i965simple/brw_vs_emit.c | 4 ++-- src/gallium/drivers/i965simple/brw_wm_decl.c | 2 +- 3 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_vs.c b/src/gallium/drivers/i965simple/brw_vs.c index 738c6346d5..92327e896d 100644 --- a/src/gallium/drivers/i965simple/brw_vs.c +++ b/src/gallium/drivers/i965simple/brw_vs.c @@ -50,8 +50,8 @@ static void do_vs_prog( struct brw_context *brw, brw_init_compile(&c.func); c.vp = vp; - c.prog_data.outputs_written = vp->program.num_outputs; - c.prog_data.inputs_read = vp->program.num_inputs; + c.prog_data.outputs_written = vp->info.num_outputs; + c.prog_data.inputs_read = vp->info.num_inputs; #if 0 if (c.key.copy_edgeflag) { diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c index 33bbdc95c6..9020fcc001 100644 --- a/src/gallium/drivers/i965simple/brw_vs_emit.c +++ b/src/gallium/drivers/i965simple/brw_vs_emit.c @@ -86,7 +86,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c, /* Allocate input regs: */ - c->nr_inputs = c->vp->program.num_inputs; + c->nr_inputs = c->vp->info.num_inputs; for (i = 0; i < c->nr_inputs; i++) { c->regs[TGSI_FILE_INPUT][i] = brw_vec8_grf(reg, 0); reg++; @@ -99,7 +99,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c, c->nr_outputs = 0; c->first_output = reg; mrf = 4; - for (i = 0; i < c->vp->program.num_outputs; i++) { + for (i = 0; i < c->vp->info.num_outputs; i++) { c->nr_outputs++; #if 0 if (i == VERT_RESULT_HPOS) { diff --git a/src/gallium/drivers/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c index b50d768062..74ccfd494a 100644 --- a/src/gallium/drivers/i965simple/brw_wm_decl.c +++ b/src/gallium/drivers/i965simple/brw_wm_decl.c @@ -292,7 +292,7 @@ static void prealloc_reg(struct brw_wm_compile *c) } c->prog_data.first_curbe_grf = c->key.nr_depth_regs * 2; - c->prog_data.urb_read_length = (c->fp->program.num_inputs + 1) * 2; + c->prog_data.urb_read_length = (c->fp->info.num_inputs + 1) * 2; c->prog_data.curb_read_length = nr_curbe_regs; /* That's the end of the payload, now we can start allocating registers. -- cgit v1.2.3 From 1774b177b858f9f87d00e54b0bf00e9634e375e9 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 18:46:54 -0700 Subject: gallium: added draw_num_vs_outputs() to query number of post-transform vertex attribs --- src/gallium/auxiliary/draw/draw_context.c | 13 +++++++++++++ src/gallium/auxiliary/draw/draw_context.h | 4 ++++ src/gallium/drivers/softpipe/sp_state_derived.c | 4 ++-- 3 files changed, 19 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 64ada8ce04..3500c34811 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -281,6 +281,19 @@ draw_find_vs_output(struct draw_context *draw, } +/** + * Return number of vertex shader outputs. + */ +uint +draw_num_vs_outputs(struct draw_context *draw) +{ + uint count = draw->vertex_shader->info.num_outputs; + if (draw->extra_vp_outputs.slot >= 0) + count++; + return count; +} + + /** * Allocate space for temporary post-transform vertices, such as for clipping. */ diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index d6685f479b..99bfef55f4 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -112,6 +112,10 @@ int draw_find_vs_output(struct draw_context *draw, uint semantic_name, uint semantic_index); +uint +draw_num_vs_outputs(struct draw_context *draw); + + /* * Vertex shader functions diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index aa6e329116..eafbaed4b9 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -61,7 +61,6 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) if (vinfo->num_attribs == 0) { /* compute vertex layout now */ - const struct pipe_shader_state *vs = &softpipe->vs->shader; const struct sp_fragment_shader *spfs = softpipe->fs; const enum interp_mode colorInterp = softpipe->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; @@ -74,7 +73,8 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; vinfo_vbuf->num_attribs = 0; draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); - vinfo_vbuf->size = 4 * vs->num_outputs + /* size in dwords or floats */ + vinfo_vbuf->size = 4 * draw_num_vs_outputs(softpipe->draw) + sizeof(struct vertex_header) / 4; } -- cgit v1.2.3 From d8bf051c8b4a1ebe44895413a109d4ab898f7579 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 27 Feb 2008 18:49:35 -0700 Subject: gallium: remove unneeded pipe ptr from pipe_texture --- src/gallium/include/pipe/p_state.h | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index bb4a6cb23e..5fab41acbd 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -161,6 +161,7 @@ struct pipe_constant_buffer struct pipe_shader_state { const struct tgsi_token *tokens; + /* XXX these are going away */ ubyte num_inputs; ubyte num_outputs; ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ @@ -276,8 +277,7 @@ struct pipe_surface /** - * Texture. Represents one or several texture images on one or several mipmap - * levels. + * Texture object. */ struct pipe_texture { @@ -298,11 +298,7 @@ struct pipe_texture */ unsigned refcount; - /**< pipe that created the texture - * XXX this'll change to a pipe_winsys (or pipe_screen)... - */ - struct pipe_context *pipe; - struct pipe_screen *screen; + struct pipe_screen *screen; /**< screen that this texture belongs to */ }; -- cgit v1.2.3 From be2c419111b63f7e1a87a70f2714fdbff10d6c50 Mon Sep 17 00:00:00 2001 From: Jose Fonseca Date: Thu, 28 Feb 2008 12:57:18 +0900 Subject: gallium: Convert CRLF -> LF. DOS/Windows users should not commit text files with CRLF line endings but use instead git config --global core.autocrlf true --- src/mesa/state_tracker/st_mesa_to_tgsi.h | 128 +++++++++++++++---------------- 1 file changed, 64 insertions(+), 64 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index 4cd4b96a58..3ababf1339 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -1,64 +1,64 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef ST_MESA_TO_TGSI_H -#define ST_MESA_TO_TGSI_H - -#include "mtypes.h" - - -#if defined __cplusplus -extern "C" { -#endif - -struct tgsi_token; -struct gl_program; - -GLboolean -tgsi_translate_mesa_program( - uint procType, - const struct gl_program *program, - GLuint numInputs, - const GLuint inputMapping[], - const ubyte inputSemanticName[], - const ubyte inputSemanticIndex[], - const GLuint interpMode[], - GLuint numOutputs, - const GLuint outputMapping[], - const ubyte outputSemanticName[], - const ubyte outputSemanticIndex[], - struct tgsi_token *tokens, - GLuint maxTokens ); - - -#if defined __cplusplus -} /* extern "C" */ -#endif - -#endif /* ST_MESA_TO_TGSI_H */ - +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_MESA_TO_TGSI_H +#define ST_MESA_TO_TGSI_H + +#include "mtypes.h" + + +#if defined __cplusplus +extern "C" { +#endif + +struct tgsi_token; +struct gl_program; + +GLboolean +tgsi_translate_mesa_program( + uint procType, + const struct gl_program *program, + GLuint numInputs, + const GLuint inputMapping[], + const ubyte inputSemanticName[], + const ubyte inputSemanticIndex[], + const GLuint interpMode[], + GLuint numOutputs, + const GLuint outputMapping[], + const ubyte outputSemanticName[], + const ubyte outputSemanticIndex[], + struct tgsi_token *tokens, + GLuint maxTokens ); + + +#if defined __cplusplus +} /* extern "C" */ +#endif + +#endif /* ST_MESA_TO_TGSI_H */ + -- cgit v1.2.3 From 510bc3535c4af68db71e5ffd19f3e21e10ec6004 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 28 Feb 2008 11:23:01 +0900 Subject: gallium: Fix sign/unsign comparison. --- src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c index b7bbff1689..cf6de1e82f 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -100,7 +100,7 @@ tgsi_scan_shader(const struct tgsi_token *tokens, /* only first 32 regs will appear in this bitfield */ info->file_mask[file] |= (1 << i); info->file_count[file]++; - info->file_max[file] = MAX2(info->file_max[file], i); + info->file_max[file] = MAX2(info->file_max[file], (int)i); if (file == TGSI_FILE_INPUT) { info->input_semantic_name[info->num_inputs] -- cgit v1.2.3 From 5c0a089a5d13baa7a427b70852223990da5f175c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 28 Feb 2008 13:52:30 +0900 Subject: gallium: Remove extra level of indirecttion. --- src/gallium/drivers/softpipe/sp_prim_setup.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 5c8b0bf69c..2feee5c485 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -1163,7 +1163,7 @@ static void setup_begin( struct draw_stage *stage ) { struct setup_stage *setup = setup_stage(stage); struct softpipe_context *sp = setup->softpipe; - const struct sp_fragment_shader *fs = &setup->softpipe->fs; + const struct sp_fragment_shader *fs = setup->softpipe->fs; if (sp->dirty) { softpipe_update_derived(sp); -- cgit v1.2.3 From 2573f0e5d6f37f1a663bd472055babc97cfb3959 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 28 Feb 2008 15:53:13 +0900 Subject: Convert crlf->lf line endings. Windows/DOS users should enable core.autocrlf from now on. --- src/glu/mesa/Makefile.m32 | 124 +- src/glut/glx/fxglut.def | 206 +- src/glut/os2/Makefile | 840 +++--- src/glut/os2/glutOverlay.cpp | 266 +- src/glut/os2/glut_8x13.cpp | 4152 +++++++++++++-------------- src/glut/os2/glut_9x15.cpp | 4154 +++++++++++++-------------- src/glut/os2/glut_bitmap.cpp | 114 +- src/glut/os2/glut_cmap.cpp | 798 +++--- src/glut/os2/glut_cursor.cpp | 420 +-- src/glut/os2/glut_event.cpp | 2798 +++++++++--------- src/glut/os2/glut_ext.cpp | 408 +-- src/glut/os2/glut_fullscrn.cpp | 76 +- src/glut/os2/glut_get.cpp | 464 +-- src/glut/os2/glut_hel10.cpp | 3562 +++++++++++------------ src/glut/os2/glut_hel12.cpp | 3582 +++++++++++------------ src/glut/os2/glut_hel18.cpp | 3800 ++++++++++++------------- src/glut/os2/glut_init.cpp | 902 +++--- src/glut/os2/glut_input.cpp | 1256 ++++---- src/glut/os2/glut_key.cpp | 58 +- src/glut/os2/glut_keyctrl.cpp | 58 +- src/glut/os2/glut_keyup.cpp | 58 +- src/glut/os2/glut_mesa.cpp | 114 +- src/glut/os2/glut_modifier.cpp | 62 +- src/glut/os2/glut_roman.cpp | 4902 ++++++++++++++++---------------- src/glut/os2/glut_shapes.cpp | 1190 ++++---- src/glut/os2/glut_stroke.cpp | 84 +- src/glut/os2/glut_swap.cpp | 134 +- src/glut/os2/glut_teapot.cpp | 420 +-- src/glut/os2/glut_tr24.cpp | 4126 +++++++++++++-------------- src/glut/os2/glut_util.cpp | 180 +- src/glut/os2/glut_vidresize.cpp | 464 +-- src/glut/os2/glut_warp.cpp | 58 +- src/glut/os2/glutbitmap.h | 64 +- src/glut/os2/glutstroke.h | 84 +- src/glut/os2/layerutil.h | 118 +- src/glut/os2/libGlut.DEF | 188 +- src/mesa/drivers/allegro/amesa.c | 824 +++--- src/mesa/drivers/allegro/direct.h | 378 +-- src/mesa/drivers/allegro/generic.h | 468 +-- src/mesa/drivers/d3d/D3DCAPS.CPP | 500 ++-- src/mesa/drivers/d3d/D3DHAL.H | 136 +- src/mesa/drivers/d3d/D3DInit.cpp | 1782 ++++++------ src/mesa/drivers/d3d/D3DMESA.H | 166 +- src/mesa/drivers/d3d/D3DRaster.cpp | 426 +-- src/mesa/drivers/d3d/D3DTEXT.CPP | 1152 ++++---- src/mesa/drivers/d3d/D3DTextureMgr.cpp | 1894 ++++++------ src/mesa/drivers/d3d/D3DTextureMgr.h | 124 +- src/mesa/drivers/d3d/D3DUTILS.CPP | 1276 ++++----- src/mesa/drivers/d3d/DDrawPROCS.c | 798 +++--- src/mesa/drivers/d3d/DEBUG.C | 286 +- src/mesa/drivers/d3d/DEBUG.H | 180 +- src/mesa/drivers/d3d/DbgEnv.bat | 50 +- src/mesa/drivers/d3d/MAKEFILE | 202 +- src/mesa/drivers/d3d/OPENGL32.DEF | 884 +++--- src/mesa/drivers/d3d/WGL.C | 2524 ++++++++-------- 55 files changed, 27167 insertions(+), 27167 deletions(-) (limited to 'src') diff --git a/src/glu/mesa/Makefile.m32 b/src/glu/mesa/Makefile.m32 index 3e217ef5d5..e2cf9dd527 100644 --- a/src/glu/mesa/Makefile.m32 +++ b/src/glu/mesa/Makefile.m32 @@ -1,63 +1,63 @@ -# Makefile for GLU for GCC-2.95.2/Mingw32 contributed by -# Paul Garceau - -# Mesa 3-D graphics library -# Version: 3.3 -# Copyright (C) 1995-1999 Brian Paul -# -# This library is free software; you can redistribute it and/or -# modify it under the terms of the GNU Library General Public -# License as published by the Free Software Foundation; either -# version 2 of the License, or (at your option) any later version. -# -# This library is distributed in the hope that it will be useful, -# but WITHOUT ANY WARRANTY; without even the implied warranty of -# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU -# Library General Public License for more details. -# -# You should have received a copy of the GNU Library General Public -# License along with this library; if not, write to the Free -# Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. - - -MESA_MAJOR=3 -MESA_MINOR=3 -MESA_TINY=0 -VERSION=$(MESA_MAJOR).$(MESA_MINOR) - -CFLAGS = -I. -DWIN32 -D__WIN32__ -D_WINDOWS \ - -O2 -funroll-loops \ - -fexpensive-optimizations -fomit-frame-pointer -ffast-math \ - -malign-loops=2 -malign-jumps=2 -malign-functions=2 \ - -mwindows -CC = gcc -MAKELIB = ar ru -GLU_LIB = libGLU.a - - -##### MACROS ##### - -VPATH = RCS - -INCDIR = ../include -LIBDIR = ../lib - -SOURCES = glu.c mipmap.c nurbs.c nurbscrv.c nurbssrf.c nurbsutl.c \ - polytest.c project.c quadric.c tess.c tesselat.c - -OBJECTS = $(SOURCES:.c=.o) - - - -##### RULES ##### - -.c.o: - $(CC) -c -I$(INCDIR) $(CFLAGS) $< - - - -##### TARGETS ##### - -# Make the library: -$(LIBDIR)/$(GLU_LIB): $(OBJECTS) +# Makefile for GLU for GCC-2.95.2/Mingw32 contributed by +# Paul Garceau + +# Mesa 3-D graphics library +# Version: 3.3 +# Copyright (C) 1995-1999 Brian Paul +# +# This library is free software; you can redistribute it and/or +# modify it under the terms of the GNU Library General Public +# License as published by the Free Software Foundation; either +# version 2 of the License, or (at your option) any later version. +# +# This library is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU +# Library General Public License for more details. +# +# You should have received a copy of the GNU Library General Public +# License along with this library; if not, write to the Free +# Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + + +MESA_MAJOR=3 +MESA_MINOR=3 +MESA_TINY=0 +VERSION=$(MESA_MAJOR).$(MESA_MINOR) + +CFLAGS = -I. -DWIN32 -D__WIN32__ -D_WINDOWS \ + -O2 -funroll-loops \ + -fexpensive-optimizations -fomit-frame-pointer -ffast-math \ + -malign-loops=2 -malign-jumps=2 -malign-functions=2 \ + -mwindows +CC = gcc +MAKELIB = ar ru +GLU_LIB = libGLU.a + + +##### MACROS ##### + +VPATH = RCS + +INCDIR = ../include +LIBDIR = ../lib + +SOURCES = glu.c mipmap.c nurbs.c nurbscrv.c nurbssrf.c nurbsutl.c \ + polytest.c project.c quadric.c tess.c tesselat.c + +OBJECTS = $(SOURCES:.c=.o) + + + +##### RULES ##### + +.c.o: + $(CC) -c -I$(INCDIR) $(CFLAGS) $< + + + +##### TARGETS ##### + +# Make the library: +$(LIBDIR)/$(GLU_LIB): $(OBJECTS) $(MAKELIB) $(GLU_LIB) $(MAJOR) $(MINOR) $(TINY) $(OBJECTS) \ No newline at end of file diff --git a/src/glut/glx/fxglut.def b/src/glut/glx/fxglut.def index ff6ffb1d39..805a0c7222 100644 --- a/src/glut/glx/fxglut.def +++ b/src/glut/glx/fxglut.def @@ -1,103 +1,103 @@ -EXPORTS - glutInit - glutInitDisplayMode - glutInitDisplayString - glutInitWindowPosition - glutInitWindowSize - glutMainLoop - glutCreateWindow - glutCreateSubWindow - glutDestroyWindow - glutPostRedisplay - glutSwapBuffers - glutGetWindow - glutSetWindow - glutSetWindowTitle - glutSetIconTitle - glutPositionWindow - glutReshapeWindow - glutPopWindow - glutPushWindow - glutIconifyWindow - glutShowWindow - glutHideWindow - glutFullScreen - glutSetCursor - glutWarpPointer - glutEstablishOverlay - glutRemoveOverlay - glutUseLayer - glutPostOverlayRedisplay - glutShowOverlay - glutHideOverlay - glutCreateMenu - glutDestroyMenu - glutGetMenu - glutSetMenu - glutAddMenuEntry - glutAddSubMenu - glutChangeToMenuEntry - glutChangeToSubMenu - glutRemoveMenuItem - glutAttachMenu - glutDetachMenu - glutDisplayFunc - glutReshapeFunc - glutKeyboardFunc - glutMouseFunc - glutMotionFunc - glutPassiveMotionFunc - glutEntryFunc - glutVisibilityFunc - glutIdleFunc - glutTimerFunc - glutMenuStateFunc - glutSpecialFunc - glutSpaceballMotionFunc - glutSpaceballRotateFunc - glutSpaceballButtonFunc - glutButtonBoxFunc - glutDialsFunc - glutTabletMotionFunc - glutTabletButtonFunc - glutMenuStatusFunc - glutOverlayDisplayFunc - glutWindowStatusFunc - glutSetColor - glutGetColor - glutCopyColormap - glutGet - glutDeviceGet - glutExtensionSupported - glutGetModifiers - glutLayerGet - glutBitmapCharacter - glutBitmapWidth - glutStrokeCharacter - glutStrokeWidth - glutBitmapLength - glutStrokeLength - glutWireSphere - glutSolidSphere - glutWireCone - glutSolidCone - glutWireCube - glutSolidCube - glutWireTorus - glutSolidTorus - glutWireDodecahedron - glutSolidDodecahedron - glutWireTeapot - glutSolidTeapot - glutWireOctahedron - glutSolidOctahedron - glutWireTetrahedron - glutSolidTetrahedron - glutWireIcosahedron - glutSolidIcosahedron - glutVideoResizeGet - glutSetupVideoResizing - glutStopVideoResizing - glutVideoResize - glutVideoPan - glutReportErrors +EXPORTS + glutInit + glutInitDisplayMode + glutInitDisplayString + glutInitWindowPosition + glutInitWindowSize + glutMainLoop + glutCreateWindow + glutCreateSubWindow + glutDestroyWindow + glutPostRedisplay + glutSwapBuffers + glutGetWindow + glutSetWindow + glutSetWindowTitle + glutSetIconTitle + glutPositionWindow + glutReshapeWindow + glutPopWindow + glutPushWindow + glutIconifyWindow + glutShowWindow + glutHideWindow + glutFullScreen + glutSetCursor + glutWarpPointer + glutEstablishOverlay + glutRemoveOverlay + glutUseLayer + glutPostOverlayRedisplay + glutShowOverlay + glutHideOverlay + glutCreateMenu + glutDestroyMenu + glutGetMenu + glutSetMenu + glutAddMenuEntry + glutAddSubMenu + glutChangeToMenuEntry + glutChangeToSubMenu + glutRemoveMenuItem + glutAttachMenu + glutDetachMenu + glutDisplayFunc + glutReshapeFunc + glutKeyboardFunc + glutMouseFunc + glutMotionFunc + glutPassiveMotionFunc + glutEntryFunc + glutVisibilityFunc + glutIdleFunc + glutTimerFunc + glutMenuStateFunc + glutSpecialFunc + glutSpaceballMotionFunc + glutSpaceballRotateFunc + glutSpaceballButtonFunc + glutButtonBoxFunc + glutDialsFunc + glutTabletMotionFunc + glutTabletButtonFunc + glutMenuStatusFunc + glutOverlayDisplayFunc + glutWindowStatusFunc + glutSetColor + glutGetColor + glutCopyColormap + glutGet + glutDeviceGet + glutExtensionSupported + glutGetModifiers + glutLayerGet + glutBitmapCharacter + glutBitmapWidth + glutStrokeCharacter + glutStrokeWidth + glutBitmapLength + glutStrokeLength + glutWireSphere + glutSolidSphere + glutWireCone + glutSolidCone + glutWireCube + glutSolidCube + glutWireTorus + glutSolidTorus + glutWireDodecahedron + glutSolidDodecahedron + glutWireTeapot + glutSolidTeapot + glutWireOctahedron + glutSolidOctahedron + glutWireTetrahedron + glutSolidTetrahedron + glutWireIcosahedron + glutSolidIcosahedron + glutVideoResizeGet + glutSetupVideoResizing + glutStopVideoResizing + glutVideoResize + glutVideoPan + glutReportErrors diff --git a/src/glut/os2/Makefile b/src/glut/os2/Makefile index 56ad4e4c47..102a8b688d 100644 --- a/src/glut/os2/Makefile +++ b/src/glut/os2/Makefile @@ -1,420 +1,420 @@ -# Makefile -# Created by IBM WorkFrame/2 MakeMake at 12:46:25 on 3 June 2003 -# -# The actions included in this make file are: -# Compile::C++ Compiler -# Link::Linker -# Lib::Import Lib - -.SUFFIXES: - -.SUFFIXES: \ - .LIB .cpp .dll .obj - -.cpp.obj: - @echo " Compile::C++ Compiler " - icc.exe /I..\ /I..\X86 /I..\GL /I..\swrast /I..\swrast_setup /Ss /Wcmpcndcnscnvcpydclenuextgeninilanobsordparporppcprorearettrdtruund /Tx /O /Gm /Ge- /G5 /Gf /Gi /Oi /C %s - -{G:\EVGEN\MESA5\src-glut.os2pm}.cpp.obj: - @echo " Compile::C++ Compiler " - icc.exe /I..\ /I..\X86 /I..\GL /I..\swrast /I..\swrast_setup /Ss /Wcmpcndcnscnvcpydclenuextgeninilanobsordparporppcprorearettrdtruund /Tx /O /Gm /Ge- /G5 /Gf /Gi /Oi /C %s - -.dll.LIB: - @echo " Lib::Import Lib " - implib.exe %|dpfF.LIB %s - -{G:\EVGEN\MESA5\src-glut.os2pm}.dll.LIB: - @echo " Lib::Import Lib " - implib.exe %|dpfF.LIB %s - -all: \ - .\libGlut.LIB - -.\libGlut.dll: \ - .\os2_winproc.obj \ - .\WarpWin.obj \ - .\glutOverlay.obj \ - .\glut_8x13.obj \ - .\glut_9x15.obj \ - .\glut_bitmap.obj \ - .\glut_cindex.obj \ - .\glut_cmap.obj \ - .\glut_cursor.obj \ - .\glut_event.obj \ - .\glut_ext.obj \ - .\glut_fullscrn.obj \ - .\glut_gamemode.obj \ - .\glut_get.obj \ - .\glut_hel10.obj \ - .\glut_hel12.obj \ - .\glut_hel18.obj \ - .\glut_init.obj \ - .\glut_input.obj \ - .\glut_key.obj \ - .\glut_keyctrl.obj \ - .\glut_keyup.obj \ - .\glut_mesa.obj \ - .\glut_modifier.obj \ - .\glut_roman.obj \ - .\glut_shapes.obj \ - .\glut_stroke.obj \ - .\glut_swap.obj \ - .\glut_teapot.obj \ - .\glut_tr24.obj \ - .\glut_util.obj \ - .\glut_vidresize.obj \ - .\glut_warp.obj \ - .\glut_win.obj \ - .\glut_winmisc.obj \ - .\os2_glx.obj \ - .\os2_menu.obj \ - ..\si-glu\libGLU.lib \ - ..\MesaDll\MesaGL2.lib \ - ..\drv\DrvLoad\MesaDrvLoad.lib \ - {$(LIB)}libGlut.DEF - @echo " Link::Linker " - icc.exe @<< - /B" /dbgpack /exepack:2 /st:200000 /packd /optfunc" - /FelibGlut.dll - ..\si-glu\libGLU.lib - ..\MesaDll\MesaGL2.lib - ..\drv\DrvLoad\MesaDrvLoad.lib - libGlut.DEF - .\os2_winproc.obj - .\WarpWin.obj - .\glutOverlay.obj - .\glut_8x13.obj - .\glut_9x15.obj - .\glut_bitmap.obj - .\glut_cindex.obj - .\glut_cmap.obj - .\glut_cursor.obj - .\glut_event.obj - .\glut_ext.obj - .\glut_fullscrn.obj - .\glut_gamemode.obj - .\glut_get.obj - .\glut_hel10.obj - .\glut_hel12.obj - .\glut_hel18.obj - .\glut_init.obj - .\glut_input.obj - .\glut_key.obj - .\glut_keyctrl.obj - .\glut_keyup.obj - .\glut_mesa.obj - .\glut_modifier.obj - .\glut_roman.obj - .\glut_shapes.obj - .\glut_stroke.obj - .\glut_swap.obj - .\glut_teapot.obj - .\glut_tr24.obj - .\glut_util.obj - .\glut_vidresize.obj - .\glut_warp.obj - .\glut_win.obj - .\glut_winmisc.obj - .\os2_glx.obj - .\os2_menu.obj -<< - -.\os2_winproc.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\os2_winproc.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h \ - gl\os2mesa.h \ - gl\gl.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}gl_mangle.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/os2_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}os2_config.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}os2mesadef.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}context.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glapi.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}mtypes.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glheader.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}config.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glapitable.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glthread.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}math/m_matrix.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}Trace/tr_context.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}dd.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}conf.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/os2_config.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/glext.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}Xthreads.h - -.\os2_menu.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\os2_menu.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\os2_glx.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\os2_glx.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - gl\os2mesa.h \ - gl\gl.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}gl_mangle.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/os2_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}os2_config.h - -.\glut_winmisc.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_winmisc.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_win.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_win.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutos2.h \ - gl\os2mesa.h \ - gl\gl.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}gl_mangle.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/os2_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}os2_config.h - -.\glut_warp.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_warp.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_vidresize.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_vidresize.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_util.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_util.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_tr24.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_tr24.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h - -.\glut_teapot.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_teapot.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_swap.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_swap.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_stroke.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_stroke.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutstroke.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_shapes.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_shapes.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_roman.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_roman.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutstroke.h - -.\glut_modifier.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_modifier.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_mesa.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_mesa.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_keyup.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_keyup.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_keyctrl.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_keyctrl.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_key.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_key.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_input.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_input.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_init.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_init.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_hel18.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_hel18.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h - -.\glut_hel12.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_hel12.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h - -.\glut_hel10.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_hel10.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h - -.\glut_get.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_get.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_gamemode.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_gamemode.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_fullscrn.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_fullscrn.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_ext.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_ext.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_event.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_event.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_cursor.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_cursor.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_cmap.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_cmap.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}layerutil.h - -.\glut_cindex.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_cindex.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_bitmap.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_bitmap.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\glut_9x15.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_9x15.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h - -.\glut_8x13.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glut_8x13.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h - -.\glutOverlay.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\glutOverlay.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutstroke.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h - -.\WarpWin.obj: \ - G:\EVGEN\MESA5\src-glut.os2pm\WarpWin.cpp \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpWin.h \ - {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h - -.\libGlut.LIB: \ - .\libGlut.dll +# Makefile +# Created by IBM WorkFrame/2 MakeMake at 12:46:25 on 3 June 2003 +# +# The actions included in this make file are: +# Compile::C++ Compiler +# Link::Linker +# Lib::Import Lib + +.SUFFIXES: + +.SUFFIXES: \ + .LIB .cpp .dll .obj + +.cpp.obj: + @echo " Compile::C++ Compiler " + icc.exe /I..\ /I..\X86 /I..\GL /I..\swrast /I..\swrast_setup /Ss /Wcmpcndcnscnvcpydclenuextgeninilanobsordparporppcprorearettrdtruund /Tx /O /Gm /Ge- /G5 /Gf /Gi /Oi /C %s + +{G:\EVGEN\MESA5\src-glut.os2pm}.cpp.obj: + @echo " Compile::C++ Compiler " + icc.exe /I..\ /I..\X86 /I..\GL /I..\swrast /I..\swrast_setup /Ss /Wcmpcndcnscnvcpydclenuextgeninilanobsordparporppcprorearettrdtruund /Tx /O /Gm /Ge- /G5 /Gf /Gi /Oi /C %s + +.dll.LIB: + @echo " Lib::Import Lib " + implib.exe %|dpfF.LIB %s + +{G:\EVGEN\MESA5\src-glut.os2pm}.dll.LIB: + @echo " Lib::Import Lib " + implib.exe %|dpfF.LIB %s + +all: \ + .\libGlut.LIB + +.\libGlut.dll: \ + .\os2_winproc.obj \ + .\WarpWin.obj \ + .\glutOverlay.obj \ + .\glut_8x13.obj \ + .\glut_9x15.obj \ + .\glut_bitmap.obj \ + .\glut_cindex.obj \ + .\glut_cmap.obj \ + .\glut_cursor.obj \ + .\glut_event.obj \ + .\glut_ext.obj \ + .\glut_fullscrn.obj \ + .\glut_gamemode.obj \ + .\glut_get.obj \ + .\glut_hel10.obj \ + .\glut_hel12.obj \ + .\glut_hel18.obj \ + .\glut_init.obj \ + .\glut_input.obj \ + .\glut_key.obj \ + .\glut_keyctrl.obj \ + .\glut_keyup.obj \ + .\glut_mesa.obj \ + .\glut_modifier.obj \ + .\glut_roman.obj \ + .\glut_shapes.obj \ + .\glut_stroke.obj \ + .\glut_swap.obj \ + .\glut_teapot.obj \ + .\glut_tr24.obj \ + .\glut_util.obj \ + .\glut_vidresize.obj \ + .\glut_warp.obj \ + .\glut_win.obj \ + .\glut_winmisc.obj \ + .\os2_glx.obj \ + .\os2_menu.obj \ + ..\si-glu\libGLU.lib \ + ..\MesaDll\MesaGL2.lib \ + ..\drv\DrvLoad\MesaDrvLoad.lib \ + {$(LIB)}libGlut.DEF + @echo " Link::Linker " + icc.exe @<< + /B" /dbgpack /exepack:2 /st:200000 /packd /optfunc" + /FelibGlut.dll + ..\si-glu\libGLU.lib + ..\MesaDll\MesaGL2.lib + ..\drv\DrvLoad\MesaDrvLoad.lib + libGlut.DEF + .\os2_winproc.obj + .\WarpWin.obj + .\glutOverlay.obj + .\glut_8x13.obj + .\glut_9x15.obj + .\glut_bitmap.obj + .\glut_cindex.obj + .\glut_cmap.obj + .\glut_cursor.obj + .\glut_event.obj + .\glut_ext.obj + .\glut_fullscrn.obj + .\glut_gamemode.obj + .\glut_get.obj + .\glut_hel10.obj + .\glut_hel12.obj + .\glut_hel18.obj + .\glut_init.obj + .\glut_input.obj + .\glut_key.obj + .\glut_keyctrl.obj + .\glut_keyup.obj + .\glut_mesa.obj + .\glut_modifier.obj + .\glut_roman.obj + .\glut_shapes.obj + .\glut_stroke.obj + .\glut_swap.obj + .\glut_teapot.obj + .\glut_tr24.obj + .\glut_util.obj + .\glut_vidresize.obj + .\glut_warp.obj + .\glut_win.obj + .\glut_winmisc.obj + .\os2_glx.obj + .\os2_menu.obj +<< + +.\os2_winproc.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\os2_winproc.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h \ + gl\os2mesa.h \ + gl\gl.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}gl_mangle.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/os2_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}os2_config.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}os2mesadef.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}context.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glapi.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}mtypes.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glheader.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}config.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glapitable.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glthread.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}math/m_matrix.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}Trace/tr_context.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}dd.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}conf.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/os2_config.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/glext.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}Xthreads.h + +.\os2_menu.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\os2_menu.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\os2_glx.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\os2_glx.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + gl\os2mesa.h \ + gl\gl.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}gl_mangle.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/os2_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}os2_config.h + +.\glut_winmisc.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_winmisc.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_win.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_win.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutos2.h \ + gl\os2mesa.h \ + gl\gl.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}gl_mangle.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}GL/os2_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}os2_config.h + +.\glut_warp.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_warp.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_vidresize.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_vidresize.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_util.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_util.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_tr24.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_tr24.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h + +.\glut_teapot.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_teapot.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_swap.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_swap.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_stroke.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_stroke.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutstroke.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_shapes.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_shapes.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_roman.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_roman.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutstroke.h + +.\glut_modifier.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_modifier.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_mesa.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_mesa.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_keyup.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_keyup.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_keyctrl.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_keyctrl.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_key.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_key.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_input.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_input.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_init.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_init.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_hel18.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_hel18.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h + +.\glut_hel12.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_hel12.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h + +.\glut_hel10.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_hel10.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h + +.\glut_get.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_get.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_gamemode.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_gamemode.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_fullscrn.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_fullscrn.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_ext.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_ext.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_event.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_event.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_cursor.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_cursor.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_cmap.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_cmap.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}layerutil.h + +.\glut_cindex.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_cindex.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_bitmap.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_bitmap.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\glut_9x15.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_9x15.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h + +.\glut_8x13.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glut_8x13.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h + +.\glutOverlay.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\glutOverlay.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutint.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutbitmap.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutstroke.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}glutwin32.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_x11.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}win32_glx.h + +.\WarpWin.obj: \ + G:\EVGEN\MESA5\src-glut.os2pm\WarpWin.cpp \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpWin.h \ + {G:\EVGEN\MESA5\src-glut.os2pm;..\;..\X86;..\GL;..\swrast;..\swrast_setup;$(INCLUDE);}WarpGL.h + +.\libGlut.LIB: \ + .\libGlut.dll diff --git a/src/glut/os2/glutOverlay.cpp b/src/glut/os2/glutOverlay.cpp index 563708de2d..d57c4e7418 100644 --- a/src/glut/os2/glutOverlay.cpp +++ b/src/glut/os2/glutOverlay.cpp @@ -1,133 +1,133 @@ -/*********************************************************** - * Copyright (C) 1997, Be Inc. All rights reserved. - * - * FILE: glutOverlay.cpp - * - * DESCRIPTION: we don't support overlays, so this code is - * really simple - ***********************************************************/ - -/*********************************************************** - * Headers - ***********************************************************/ -#include -#include "glutint.h" -#include "glutbitmap.h" -#include "glutstroke.h" - -GLUTAPI void GLUTAPIENTRY -glutEstablishOverlay(void) -{ - __glutFatalError("OS2PM lacks overlay support."); -} - -GLUTAPI void GLUTAPIENTRY -glutUseLayer(GLenum layer) { - // ignore -} - -GLUTAPI void GLUTAPIENTRY -glutRemoveOverlay(void) { - // ignore -} - -GLUTAPI void GLUTAPIENTRY -glutPostOverlayRedisplay(void) { - // ignore -} - -GLUTAPI void GLUTAPIENTRY -glutShowOverlay(void) { - // ignore -} - -GLUTAPI void GLUTAPIENTRY glutHideOverlay(void) -{ - // ignore -} - -int GLUTAPIENTRY -glutLayerGet(GLenum param) -{ - // ignore -} - -/*********************************************************** - * Unsupported callbacks - ***********************************************************/ -GLUTAPI void GLUTAPIENTRY -glutOverlayDisplayFunc(GLUTdisplayCB displayFunc) -{ -} - -GLUTAPI void GLUTAPIENTRY -glutSpaceballMotionFunc(GLUTspaceMotionCB spaceMotionFunc) -{ -} - -GLUTAPI void GLUTAPIENTRY -glutSpaceballRotateFunc(GLUTspaceRotateCB spaceRotateFunc) -{ -} - -GLUTAPI void GLUTAPIENTRY -glutSpaceballButtonFunc(GLUTspaceButtonCB spaceButtonFunc) -{ -} - -GLUTAPI void GLUTAPIENTRY -glutButtonBoxFunc(GLUTbuttonBoxCB buttonBoxFunc) -{ -} - -GLUTAPI void GLUTAPIENTRY -glutDialsFunc(GLUTdialsCB dialsFunc) -{ -} - -GLUTAPI void GLUTAPIENTRY -glutTabletMotionFunc(GLUTtabletMotionCB tabletMotionFunc) -{ -} - -GLUTAPI void GLUTAPIENTRY -glutTabletButtonFunc(GLUTtabletButtonCB tabletButtonFunc) -{ -} -GLUTAPI void GLUTAPIENTRY -glutPostWindowOverlayRedisplay(int win) -{ // -} - -void GLUTAPIENTRY -glutInitDisplayString(const char *string) -{ // -} -void GLUTAPIENTRY -glutJoystickFunc(GLUTjoystickCB joystickFunc, int pollInterval) -{ // -} - -void GLUTAPIENTRY -glutForceJoystickFunc(void) -{ // -} - - -int GLUTAPIENTRY -glutBitmapWidth(GLUTbitmapFont font, int c) -{ return 0; -} -int GLUTAPIENTRY -glutBitmapLength(GLUTbitmapFont font, const unsigned char *string) -{ // - return 0; -} -int GLUTAPIENTRY -glutStrokeWidth(GLUTstrokeFont font, int c) -{ return 0; -} -int GLUTAPIENTRY -glutStrokeLength(GLUTstrokeFont font, const unsigned char *string) -{ return 0; -} +/*********************************************************** + * Copyright (C) 1997, Be Inc. All rights reserved. + * + * FILE: glutOverlay.cpp + * + * DESCRIPTION: we don't support overlays, so this code is + * really simple + ***********************************************************/ + +/*********************************************************** + * Headers + ***********************************************************/ +#include +#include "glutint.h" +#include "glutbitmap.h" +#include "glutstroke.h" + +GLUTAPI void GLUTAPIENTRY +glutEstablishOverlay(void) +{ + __glutFatalError("OS2PM lacks overlay support."); +} + +GLUTAPI void GLUTAPIENTRY +glutUseLayer(GLenum layer) { + // ignore +} + +GLUTAPI void GLUTAPIENTRY +glutRemoveOverlay(void) { + // ignore +} + +GLUTAPI void GLUTAPIENTRY +glutPostOverlayRedisplay(void) { + // ignore +} + +GLUTAPI void GLUTAPIENTRY +glutShowOverlay(void) { + // ignore +} + +GLUTAPI void GLUTAPIENTRY glutHideOverlay(void) +{ + // ignore +} + +int GLUTAPIENTRY +glutLayerGet(GLenum param) +{ + // ignore +} + +/*********************************************************** + * Unsupported callbacks + ***********************************************************/ +GLUTAPI void GLUTAPIENTRY +glutOverlayDisplayFunc(GLUTdisplayCB displayFunc) +{ +} + +GLUTAPI void GLUTAPIENTRY +glutSpaceballMotionFunc(GLUTspaceMotionCB spaceMotionFunc) +{ +} + +GLUTAPI void GLUTAPIENTRY +glutSpaceballRotateFunc(GLUTspaceRotateCB spaceRotateFunc) +{ +} + +GLUTAPI void GLUTAPIENTRY +glutSpaceballButtonFunc(GLUTspaceButtonCB spaceButtonFunc) +{ +} + +GLUTAPI void GLUTAPIENTRY +glutButtonBoxFunc(GLUTbuttonBoxCB buttonBoxFunc) +{ +} + +GLUTAPI void GLUTAPIENTRY +glutDialsFunc(GLUTdialsCB dialsFunc) +{ +} + +GLUTAPI void GLUTAPIENTRY +glutTabletMotionFunc(GLUTtabletMotionCB tabletMotionFunc) +{ +} + +GLUTAPI void GLUTAPIENTRY +glutTabletButtonFunc(GLUTtabletButtonCB tabletButtonFunc) +{ +} +GLUTAPI void GLUTAPIENTRY +glutPostWindowOverlayRedisplay(int win) +{ // +} + +void GLUTAPIENTRY +glutInitDisplayString(const char *string) +{ // +} +void GLUTAPIENTRY +glutJoystickFunc(GLUTjoystickCB joystickFunc, int pollInterval) +{ // +} + +void GLUTAPIENTRY +glutForceJoystickFunc(void) +{ // +} + + +int GLUTAPIENTRY +glutBitmapWidth(GLUTbitmapFont font, int c) +{ return 0; +} +int GLUTAPIENTRY +glutBitmapLength(GLUTbitmapFont font, const unsigned char *string) +{ // + return 0; +} +int GLUTAPIENTRY +glutStrokeWidth(GLUTstrokeFont font, int c) +{ return 0; +} +int GLUTAPIENTRY +glutStrokeLength(GLUTstrokeFont font, const unsigned char *string) +{ return 0; +} diff --git a/src/glut/os2/glut_8x13.cpp b/src/glut/os2/glut_8x13.cpp index e361e17cbf..29ea6ea070 100644 --- a/src/glut/os2/glut_8x13.cpp +++ b/src/glut/os2/glut_8x13.cpp @@ -1,2076 +1,2076 @@ - -/* GENERATED FILE -- DO NOT MODIFY */ - -#define glutBitmap8By13 XXX -#include "glutbitmap.h" -#undef glutBitmap8By13 - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch0data[] = { 0x0 }; -static const BitmapCharRec ch0 = {1,1,0,0,8,ch0data}; -#else -static const BitmapCharRec ch0 = {0,0,0,0,8,0}; -#endif - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch32data[] = { 0x0 }; -static const BitmapCharRec ch32 = {1,1,0,0,8,ch32data}; -#else -static const BitmapCharRec ch32 = {0,0,0,0,8,0}; -#endif - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch127data[] = { 0x0 }; -static const BitmapCharRec ch127 = {1,1,0,0,8,ch127data}; -#else -static const BitmapCharRec ch127 = {0,0,0,0,8,0}; -#endif - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch160data[] = { 0x0 }; -static const BitmapCharRec ch160 = {1,1,0,0,8,ch160data}; -#else -static const BitmapCharRec ch160 = {0,0,0,0,8,0}; -#endif - -/* char: 0xff */ - -static const GLubyte ch255data[] = { -0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x0,0x0,0x48,0x48, -}; - -static const BitmapCharRec ch255 = {6,12,-1,2,8,ch255data}; - -/* char: 0xfe */ - -static const GLubyte ch254data[] = { -0x80,0x80,0xb8,0xc4,0x84,0x84,0xc4,0xb8,0x80,0x80, -}; - -static const BitmapCharRec ch254 = {6,10,-1,2,8,ch254data}; - -/* char: 0xfd */ - -static const GLubyte ch253data[] = { -0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x0,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch253 = {6,12,-1,2,8,ch253data}; - -/* char: 0xfc */ - -static const GLubyte ch252data[] = { -0x74,0x88,0x88,0x88,0x88,0x88,0x0,0x0,0x48,0x48, -}; - -static const BitmapCharRec ch252 = {6,10,-1,0,8,ch252data}; - -/* char: 0xfb */ - -static const GLubyte ch251data[] = { -0x74,0x88,0x88,0x88,0x88,0x88,0x0,0x0,0x48,0x30, -}; - -static const BitmapCharRec ch251 = {6,10,-1,0,8,ch251data}; - -/* char: 0xfa */ - -static const GLubyte ch250data[] = { -0x74,0x88,0x88,0x88,0x88,0x88,0x0,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch250 = {6,10,-1,0,8,ch250data}; - -/* char: 0xf9 */ - -static const GLubyte ch249data[] = { -0x74,0x88,0x88,0x88,0x88,0x88,0x0,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch249 = {6,10,-1,0,8,ch249data}; - -/* char: 0xf8 */ - -static const GLubyte ch248data[] = { -0x80,0x78,0xc4,0xa4,0x94,0x8c,0x78,0x4, -}; - -static const BitmapCharRec ch248 = {6,8,-1,1,8,ch248data}; - -/* char: 0xf7 */ - -static const GLubyte ch247data[] = { -0x20,0x20,0x0,0xf8,0x0,0x20,0x20, -}; - -static const BitmapCharRec ch247 = {5,7,-1,-1,8,ch247data}; - -/* char: 0xf6 */ - -static const GLubyte ch246data[] = { -0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x48,0x48, -}; - -static const BitmapCharRec ch246 = {6,10,-1,0,8,ch246data}; - -/* char: 0xf5 */ - -static const GLubyte ch245data[] = { -0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch245 = {6,10,-1,0,8,ch245data}; - -/* char: 0xf4 */ - -static const GLubyte ch244data[] = { -0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x48,0x30, -}; - -static const BitmapCharRec ch244 = {6,10,-1,0,8,ch244data}; - -/* char: 0xf3 */ - -static const GLubyte ch243data[] = { -0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch243 = {6,10,-1,0,8,ch243data}; - -/* char: 0xf2 */ - -static const GLubyte ch242data[] = { -0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch242 = {6,10,-1,0,8,ch242data}; - -/* char: 0xf1 */ - -static const GLubyte ch241data[] = { -0x84,0x84,0x84,0x84,0xc4,0xb8,0x0,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch241 = {6,10,-1,0,8,ch241data}; - -/* char: 0xf0 */ - -static const GLubyte ch240data[] = { -0x78,0x84,0x84,0x84,0x84,0x78,0x8,0x50,0x30,0x48, -}; - -static const BitmapCharRec ch240 = {6,10,-1,0,8,ch240data}; - -/* char: 0xef */ - -static const GLubyte ch239data[] = { -0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x0,0x50,0x50, -}; - -static const BitmapCharRec ch239 = {5,10,-1,0,8,ch239data}; - -/* char: 0xee */ - -static const GLubyte ch238data[] = { -0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x0,0x90,0x60, -}; - -static const BitmapCharRec ch238 = {5,10,-1,0,8,ch238data}; - -/* char: 0xed */ - -static const GLubyte ch237data[] = { -0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x0,0x40,0x20, -}; - -static const BitmapCharRec ch237 = {5,10,-1,0,8,ch237data}; - -/* char: 0xec */ - -static const GLubyte ch236data[] = { -0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x0,0x20,0x40, -}; - -static const BitmapCharRec ch236 = {5,10,-1,0,8,ch236data}; - -/* char: 0xeb */ - -static const GLubyte ch235data[] = { -0x78,0x84,0x80,0xfc,0x84,0x78,0x0,0x0,0x48,0x48, -}; - -static const BitmapCharRec ch235 = {6,10,-1,0,8,ch235data}; - -/* char: 0xea */ - -static const GLubyte ch234data[] = { -0x78,0x84,0x80,0xfc,0x84,0x78,0x0,0x0,0x48,0x30, -}; - -static const BitmapCharRec ch234 = {6,10,-1,0,8,ch234data}; - -/* char: 0xe9 */ - -static const GLubyte ch233data[] = { -0x78,0x84,0x80,0xfc,0x84,0x78,0x0,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch233 = {6,10,-1,0,8,ch233data}; - -/* char: 0xe8 */ - -static const GLubyte ch232data[] = { -0x78,0x84,0x80,0xfc,0x84,0x78,0x0,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch232 = {6,10,-1,0,8,ch232data}; - -/* char: 0xe7 */ - -static const GLubyte ch231data[] = { -0x20,0x10,0x78,0x84,0x80,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch231 = {6,8,-1,2,8,ch231data}; - -/* char: 0xe6 */ - -static const GLubyte ch230data[] = { -0x6c,0x92,0x90,0x7c,0x12,0x6c, -}; - -static const BitmapCharRec ch230 = {7,6,0,0,8,ch230data}; - -/* char: 0xe5 */ - -static const GLubyte ch229data[] = { -0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x30,0x48,0x30, -}; - -static const BitmapCharRec ch229 = {6,10,-1,0,8,ch229data}; - -/* char: 0xe4 */ - -static const GLubyte ch228data[] = { -0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x48,0x48, -}; - -static const BitmapCharRec ch228 = {6,10,-1,0,8,ch228data}; - -/* char: 0xe3 */ - -static const GLubyte ch227data[] = { -0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch227 = {6,10,-1,0,8,ch227data}; - -/* char: 0xe2 */ - -static const GLubyte ch226data[] = { -0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x48,0x30, -}; - -static const BitmapCharRec ch226 = {6,10,-1,0,8,ch226data}; - -/* char: 0xe1 */ - -static const GLubyte ch225data[] = { -0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch225 = {6,10,-1,0,8,ch225data}; - -/* char: 0xe0 */ - -static const GLubyte ch224data[] = { -0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch224 = {6,10,-1,0,8,ch224data}; - -/* char: 0xdf */ - -static const GLubyte ch223data[] = { -0x80,0xb8,0xc4,0x84,0x84,0xf8,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch223 = {6,9,-1,1,8,ch223data}; - -/* char: 0xde */ - -static const GLubyte ch222data[] = { -0x80,0x80,0x80,0xf8,0x84,0x84,0x84,0xf8,0x80, -}; - -static const BitmapCharRec ch222 = {6,9,-1,0,8,ch222data}; - -/* char: 0xdd */ - -static const GLubyte ch221data[] = { -0x20,0x20,0x20,0x20,0x50,0x88,0x88,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch221 = {5,10,-1,0,8,ch221data}; - -/* char: 0xdc */ - -static const GLubyte ch220data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x48,0x48, -}; - -static const BitmapCharRec ch220 = {6,10,-1,0,8,ch220data}; - -/* char: 0xdb */ - -static const GLubyte ch219data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x48,0x30, -}; - -static const BitmapCharRec ch219 = {6,10,-1,0,8,ch219data}; - -/* char: 0xda */ - -static const GLubyte ch218data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch218 = {6,10,-1,0,8,ch218data}; - -/* char: 0xd9 */ - -static const GLubyte ch217data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch217 = {6,10,-1,0,8,ch217data}; - -/* char: 0xd8 */ - -static const GLubyte ch216data[] = { -0x80,0x78,0xc4,0xa4,0xa4,0xa4,0x94,0x94,0x8c,0x78,0x4, -}; - -static const BitmapCharRec ch216 = {6,11,-1,1,8,ch216data}; - -/* char: 0xd7 */ - -static const GLubyte ch215data[] = { -0x84,0x48,0x30,0x30,0x48,0x84, -}; - -static const BitmapCharRec ch215 = {6,6,-1,-1,8,ch215data}; - -/* char: 0xd6 */ - -static const GLubyte ch214data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch214 = {7,10,0,0,8,ch214data}; - -/* char: 0xd5 */ - -static const GLubyte ch213data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x28,0x14, -}; - -static const BitmapCharRec ch213 = {7,10,0,0,8,ch213data}; - -/* char: 0xd4 */ - -static const GLubyte ch212data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x24,0x18, -}; - -static const BitmapCharRec ch212 = {7,10,0,0,8,ch212data}; - -/* char: 0xd3 */ - -static const GLubyte ch211data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x10,0x8, -}; - -static const BitmapCharRec ch211 = {7,10,0,0,8,ch211data}; - -/* char: 0xd2 */ - -static const GLubyte ch210data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x8,0x10, -}; - -static const BitmapCharRec ch210 = {7,10,0,0,8,ch210data}; - -/* char: 0xd1 */ - -static const GLubyte ch209data[] = { -0x82,0x86,0x8a,0x92,0xa2,0xc2,0x82,0x0,0x28,0x14, -}; - -static const BitmapCharRec ch209 = {7,10,0,0,8,ch209data}; - -/* char: 0xd0 */ - -static const GLubyte ch208data[] = { -0xfc,0x42,0x42,0x42,0xe2,0x42,0x42,0x42,0xfc, -}; - -static const BitmapCharRec ch208 = {7,9,0,0,8,ch208data}; - -/* char: 0xcf */ - -static const GLubyte ch207data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x50,0x50, -}; - -static const BitmapCharRec ch207 = {5,10,-1,0,8,ch207data}; - -/* char: 0xce */ - -static const GLubyte ch206data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x48,0x30, -}; - -static const BitmapCharRec ch206 = {5,10,-1,0,8,ch206data}; - -/* char: 0xcd */ - -static const GLubyte ch205data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch205 = {5,10,-1,0,8,ch205data}; - -/* char: 0xcc */ - -static const GLubyte ch204data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch204 = {5,10,-1,0,8,ch204data}; - -/* char: 0xcb */ - -static const GLubyte ch203data[] = { -0xfc,0x80,0x80,0xf0,0x80,0x80,0xfc,0x0,0x48,0x48, -}; - -static const BitmapCharRec ch203 = {6,10,-1,0,8,ch203data}; - -/* char: 0xca */ - -static const GLubyte ch202data[] = { -0xfc,0x80,0x80,0xf0,0x80,0x80,0xfc,0x0,0x48,0x30, -}; - -static const BitmapCharRec ch202 = {6,10,-1,0,8,ch202data}; - -/* char: 0xc9 */ - -static const GLubyte ch201data[] = { -0xfc,0x80,0x80,0xf0,0x80,0x80,0xfc,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch201 = {6,10,-1,0,8,ch201data}; - -/* char: 0xc8 */ - -static const GLubyte ch200data[] = { -0xfc,0x80,0x80,0xf0,0x80,0x80,0xfc,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch200 = {6,10,-1,0,8,ch200data}; - -/* char: 0xc7 */ - -static const GLubyte ch199data[] = { -0x20,0x10,0x78,0x84,0x80,0x80,0x80,0x80,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch199 = {6,11,-1,2,8,ch199data}; - -/* char: 0xc6 */ - -static const GLubyte ch198data[] = { -0x9e,0x90,0x90,0xf0,0x9c,0x90,0x90,0x90,0x6e, -}; - -static const BitmapCharRec ch198 = {7,9,0,0,8,ch198data}; - -/* char: 0xc5 */ - -static const GLubyte ch197data[] = { -0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x30,0x48,0x30, -}; - -static const BitmapCharRec ch197 = {6,10,-1,0,8,ch197data}; - -/* char: 0xc4 */ - -static const GLubyte ch196data[] = { -0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x48,0x48, -}; - -static const BitmapCharRec ch196 = {6,10,-1,0,8,ch196data}; - -/* char: 0xc3 */ - -static const GLubyte ch195data[] = { -0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch195 = {6,10,-1,0,8,ch195data}; - -/* char: 0xc2 */ - -static const GLubyte ch194data[] = { -0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x48,0x30, -}; - -static const BitmapCharRec ch194 = {6,10,-1,0,8,ch194data}; - -/* char: 0xc1 */ - -static const GLubyte ch193data[] = { -0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch193 = {6,10,-1,0,8,ch193data}; - -/* char: 0xc0 */ - -static const GLubyte ch192data[] = { -0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch192 = {6,10,-1,0,8,ch192data}; - -/* char: 0xbf */ - -static const GLubyte ch191data[] = { -0x78,0x84,0x84,0x80,0x40,0x20,0x20,0x0,0x20, -}; - -static const BitmapCharRec ch191 = {6,9,-1,0,8,ch191data}; - -/* char: 0xbe */ - -static const GLubyte ch190data[] = { -0x6,0x1a,0x12,0xa,0x66,0x92,0x10,0x20,0x90,0x60, -}; - -static const BitmapCharRec ch190 = {7,10,0,0,8,ch190data}; - -/* char: 0xbd */ - -static const GLubyte ch189data[] = { -0x1e,0x10,0xc,0x2,0xf2,0x4c,0x40,0x40,0xc0,0x40, -}; - -static const BitmapCharRec ch189 = {7,10,0,0,8,ch189data}; - -/* char: 0xbc */ - -static const GLubyte ch188data[] = { -0x6,0x1a,0x12,0xa,0xe6,0x42,0x40,0x40,0xc0,0x40, -}; - -static const BitmapCharRec ch188 = {7,10,0,0,8,ch188data}; - -/* char: 0xbb */ - -static const GLubyte ch187data[] = { -0x90,0x48,0x24,0x12,0x24,0x48,0x90, -}; - -static const BitmapCharRec ch187 = {7,7,0,-1,8,ch187data}; - -/* char: 0xba */ - -static const GLubyte ch186data[] = { -0xf0,0x0,0x60,0x90,0x90,0x60, -}; - -static const BitmapCharRec ch186 = {4,6,-1,-3,8,ch186data}; - -/* char: 0xb9 */ - -static const GLubyte ch185data[] = { -0xe0,0x40,0x40,0x40,0xc0,0x40, -}; - -static const BitmapCharRec ch185 = {3,6,-1,-4,8,ch185data}; - -/* char: 0xb8 */ - -static const GLubyte ch184data[] = { -0xc0,0x40, -}; - -static const BitmapCharRec ch184 = {2,2,-3,2,8,ch184data}; - -/* char: 0xb7 */ - -static const GLubyte ch183data[] = { -0xc0, -}; - -static const BitmapCharRec ch183 = {2,1,-3,-4,8,ch183data}; - -/* char: 0xb6 */ - -static const GLubyte ch182data[] = { -0x28,0x28,0x28,0x28,0x68,0xe8,0xe8,0xe8,0x7c, -}; - -static const BitmapCharRec ch182 = {6,9,-1,0,8,ch182data}; - -/* char: 0xb5 */ - -static const GLubyte ch181data[] = { -0x80,0xb4,0xcc,0x84,0x84,0x84,0x84, -}; - -static const BitmapCharRec ch181 = {6,7,-1,1,8,ch181data}; - -/* char: 0xb4 */ - -static const GLubyte ch180data[] = { -0x80,0x40, -}; - -static const BitmapCharRec ch180 = {2,2,-3,-8,8,ch180data}; - -/* char: 0xb3 */ - -static const GLubyte ch179data[] = { -0x60,0x90,0x10,0x20,0x90,0x60, -}; - -static const BitmapCharRec ch179 = {4,6,-1,-4,8,ch179data}; - -/* char: 0xb2 */ - -static const GLubyte ch178data[] = { -0xf0,0x80,0x60,0x10,0x90,0x60, -}; - -static const BitmapCharRec ch178 = {4,6,-1,-4,8,ch178data}; - -/* char: 0xb1 */ - -static const GLubyte ch177data[] = { -0xf8,0x0,0x20,0x20,0xf8,0x20,0x20, -}; - -static const BitmapCharRec ch177 = {5,7,-1,-1,8,ch177data}; - -/* char: 0xb0 */ - -static const GLubyte ch176data[] = { -0x60,0x90,0x90,0x60, -}; - -static const BitmapCharRec ch176 = {4,4,-2,-5,8,ch176data}; - -/* char: 0xaf */ - -static const GLubyte ch175data[] = { -0xfc, -}; - -static const BitmapCharRec ch175 = {6,1,-1,-8,8,ch175data}; - -/* char: 0xae */ - -static const GLubyte ch174data[] = { -0x38,0x44,0xaa,0xb2,0xaa,0xaa,0x92,0x44,0x38, -}; - -static const BitmapCharRec ch174 = {7,9,0,-1,8,ch174data}; - -/* char: 0xad */ - -static const GLubyte ch173data[] = { -0xfc, -}; - -static const BitmapCharRec ch173 = {6,1,-1,-4,8,ch173data}; - -/* char: 0xac */ - -static const GLubyte ch172data[] = { -0x4,0x4,0x4,0xfc, -}; - -static const BitmapCharRec ch172 = {6,4,-1,-1,8,ch172data}; - -/* char: 0xab */ - -static const GLubyte ch171data[] = { -0x12,0x24,0x48,0x90,0x48,0x24,0x12, -}; - -static const BitmapCharRec ch171 = {7,7,0,-1,8,ch171data}; - -/* char: 0xaa */ - -static const GLubyte ch170data[] = { -0xf8,0x0,0x78,0x88,0x78,0x8,0x70, -}; - -static const BitmapCharRec ch170 = {5,7,-1,-2,8,ch170data}; - -/* char: 0xa9 */ - -static const GLubyte ch169data[] = { -0x38,0x44,0x92,0xaa,0xa2,0xaa,0x92,0x44,0x38, -}; - -static const BitmapCharRec ch169 = {7,9,0,-1,8,ch169data}; - -/* char: 0xa8 */ - -static const GLubyte ch168data[] = { -0xd8, -}; - -static const BitmapCharRec ch168 = {5,1,-1,-8,8,ch168data}; - -/* char: 0xa7 */ - -static const GLubyte ch167data[] = { -0x60,0x90,0x10,0x60,0x90,0x90,0x60,0x80,0x90,0x60, -}; - -static const BitmapCharRec ch167 = {4,10,-2,0,8,ch167data}; - -/* char: 0xa6 */ - -static const GLubyte ch166data[] = { -0x80,0x80,0x80,0x80,0x0,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch166 = {1,9,-3,0,8,ch166data}; - -/* char: 0xa5 */ - -static const GLubyte ch165data[] = { -0x10,0x10,0x7c,0x10,0x7c,0x28,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch165 = {7,9,0,0,8,ch165data}; - -/* char: 0xa4 */ - -static const GLubyte ch164data[] = { -0x84,0x78,0x48,0x48,0x78,0x84, -}; - -static const BitmapCharRec ch164 = {6,6,-1,-1,8,ch164data}; - -/* char: 0xa3 */ - -static const GLubyte ch163data[] = { -0xdc,0x62,0x20,0x20,0x20,0x70,0x20,0x22,0x1c, -}; - -static const BitmapCharRec ch163 = {7,9,0,0,8,ch163data}; - -/* char: 0xa2 */ - -static const GLubyte ch162data[] = { -0x20,0x70,0xa8,0xa0,0xa0,0xa8,0x70,0x20, -}; - -static const BitmapCharRec ch162 = {5,8,-1,-1,8,ch162data}; - -/* char: 0xa1 */ - -static const GLubyte ch161data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, -}; - -static const BitmapCharRec ch161 = {1,9,-3,0,8,ch161data}; - -/* char: 0x7e '~' */ - -static const GLubyte ch126data[] = { -0x90,0xa8,0x48, -}; - -static const BitmapCharRec ch126 = {5,3,-1,-6,8,ch126data}; - -/* char: 0x7d '}' */ - -static const GLubyte ch125data[] = { -0xe0,0x10,0x10,0x20,0x18,0x20,0x10,0x10,0xe0, -}; - -static const BitmapCharRec ch125 = {5,9,-1,0,8,ch125data}; - -/* char: 0x7c '|' */ - -static const GLubyte ch124data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch124 = {1,9,-3,0,8,ch124data}; - -/* char: 0x7b '{' */ - -static const GLubyte ch123data[] = { -0x38,0x40,0x40,0x20,0xc0,0x20,0x40,0x40,0x38, -}; - -static const BitmapCharRec ch123 = {5,9,-2,0,8,ch123data}; - -/* char: 0x7a 'z' */ - -static const GLubyte ch122data[] = { -0xfc,0x40,0x20,0x10,0x8,0xfc, -}; - -static const BitmapCharRec ch122 = {6,6,-1,0,8,ch122data}; - -/* char: 0x79 'y' */ - -static const GLubyte ch121data[] = { -0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84, -}; - -static const BitmapCharRec ch121 = {6,8,-1,2,8,ch121data}; - -/* char: 0x78 'x' */ - -static const GLubyte ch120data[] = { -0x84,0x48,0x30,0x30,0x48,0x84, -}; - -static const BitmapCharRec ch120 = {6,6,-1,0,8,ch120data}; - -/* char: 0x77 'w' */ - -static const GLubyte ch119data[] = { -0x44,0xaa,0x92,0x92,0x82,0x82, -}; - -static const BitmapCharRec ch119 = {7,6,0,0,8,ch119data}; - -/* char: 0x76 'v' */ - -static const GLubyte ch118data[] = { -0x20,0x50,0x50,0x88,0x88,0x88, -}; - -static const BitmapCharRec ch118 = {5,6,-1,0,8,ch118data}; - -/* char: 0x75 'u' */ - -static const GLubyte ch117data[] = { -0x74,0x88,0x88,0x88,0x88,0x88, -}; - -static const BitmapCharRec ch117 = {6,6,-1,0,8,ch117data}; - -/* char: 0x74 't' */ - -static const GLubyte ch116data[] = { -0x38,0x44,0x40,0x40,0x40,0xf8,0x40,0x40, -}; - -static const BitmapCharRec ch116 = {6,8,-1,0,8,ch116data}; - -/* char: 0x73 's' */ - -static const GLubyte ch115data[] = { -0x78,0x84,0x18,0x60,0x84,0x78, -}; - -static const BitmapCharRec ch115 = {6,6,-1,0,8,ch115data}; - -/* char: 0x72 'r' */ - -static const GLubyte ch114data[] = { -0x40,0x40,0x40,0x40,0x44,0xb8, -}; - -static const BitmapCharRec ch114 = {6,6,-1,0,8,ch114data}; - -/* char: 0x71 'q' */ - -static const GLubyte ch113data[] = { -0x4,0x4,0x4,0x74,0x8c,0x84,0x8c,0x74, -}; - -static const BitmapCharRec ch113 = {6,8,-1,2,8,ch113data}; - -/* char: 0x70 'p' */ - -static const GLubyte ch112data[] = { -0x80,0x80,0x80,0xb8,0xc4,0x84,0xc4,0xb8, -}; - -static const BitmapCharRec ch112 = {6,8,-1,2,8,ch112data}; - -/* char: 0x6f 'o' */ - -static const GLubyte ch111data[] = { -0x78,0x84,0x84,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch111 = {6,6,-1,0,8,ch111data}; - -/* char: 0x6e 'n' */ - -static const GLubyte ch110data[] = { -0x84,0x84,0x84,0x84,0xc4,0xb8, -}; - -static const BitmapCharRec ch110 = {6,6,-1,0,8,ch110data}; - -/* char: 0x6d 'm' */ - -static const GLubyte ch109data[] = { -0x82,0x92,0x92,0x92,0x92,0xec, -}; - -static const BitmapCharRec ch109 = {7,6,0,0,8,ch109data}; - -/* char: 0x6c 'l' */ - -static const GLubyte ch108data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0x60, -}; - -static const BitmapCharRec ch108 = {5,9,-1,0,8,ch108data}; - -/* char: 0x6b 'k' */ - -static const GLubyte ch107data[] = { -0x84,0x88,0x90,0xe0,0x90,0x88,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch107 = {6,9,-1,0,8,ch107data}; - -/* char: 0x6a 'j' */ - -static const GLubyte ch106data[] = { -0x70,0x88,0x88,0x8,0x8,0x8,0x8,0x18,0x0,0x8, -}; - -static const BitmapCharRec ch106 = {5,10,-1,2,8,ch106data}; - -/* char: 0x69 'i' */ - -static const GLubyte ch105data[] = { -0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x20, -}; - -static const BitmapCharRec ch105 = {5,8,-1,0,8,ch105data}; - -/* char: 0x68 'h' */ - -static const GLubyte ch104data[] = { -0x84,0x84,0x84,0x84,0xc4,0xb8,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch104 = {6,9,-1,0,8,ch104data}; - -/* char: 0x67 'g' */ - -static const GLubyte ch103data[] = { -0x78,0x84,0x78,0x80,0x70,0x88,0x88,0x74, -}; - -static const BitmapCharRec ch103 = {6,8,-1,2,8,ch103data}; - -/* char: 0x66 'f' */ - -static const GLubyte ch102data[] = { -0x40,0x40,0x40,0x40,0xf8,0x40,0x40,0x44,0x38, -}; - -static const BitmapCharRec ch102 = {6,9,-1,0,8,ch102data}; - -/* char: 0x65 'e' */ - -static const GLubyte ch101data[] = { -0x78,0x84,0x80,0xfc,0x84,0x78, -}; - -static const BitmapCharRec ch101 = {6,6,-1,0,8,ch101data}; - -/* char: 0x64 'd' */ - -static const GLubyte ch100data[] = { -0x74,0x8c,0x84,0x84,0x8c,0x74,0x4,0x4,0x4, -}; - -static const BitmapCharRec ch100 = {6,9,-1,0,8,ch100data}; - -/* char: 0x63 'c' */ - -static const GLubyte ch99data[] = { -0x78,0x84,0x80,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch99 = {6,6,-1,0,8,ch99data}; - -/* char: 0x62 'b' */ - -static const GLubyte ch98data[] = { -0xb8,0xc4,0x84,0x84,0xc4,0xb8,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch98 = {6,9,-1,0,8,ch98data}; - -/* char: 0x61 'a' */ - -static const GLubyte ch97data[] = { -0x74,0x8c,0x84,0x7c,0x4,0x78, -}; - -static const BitmapCharRec ch97 = {6,6,-1,0,8,ch97data}; - -/* char: 0x60 '`' */ - -static const GLubyte ch96data[] = { -0x10,0x60,0xe0, -}; - -static const BitmapCharRec ch96 = {4,3,-2,-6,8,ch96data}; - -/* char: 0x5f '_' */ - -static const GLubyte ch95data[] = { -0xfe, -}; - -static const BitmapCharRec ch95 = {7,1,0,1,8,ch95data}; - -/* char: 0x5e '^' */ - -static const GLubyte ch94data[] = { -0x88,0x50,0x20, -}; - -static const BitmapCharRec ch94 = {5,3,-1,-6,8,ch94data}; - -/* char: 0x5d ']' */ - -static const GLubyte ch93data[] = { -0xf0,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xf0, -}; - -static const BitmapCharRec ch93 = {4,9,-1,0,8,ch93data}; - -/* char: 0x5c '\' */ - -static const GLubyte ch92data[] = { -0x2,0x2,0x4,0x8,0x10,0x20,0x40,0x80,0x80, -}; - -static const BitmapCharRec ch92 = {7,9,0,0,8,ch92data}; - -/* char: 0x5b '[' */ - -static const GLubyte ch91data[] = { -0xf0,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0xf0, -}; - -static const BitmapCharRec ch91 = {4,9,-2,0,8,ch91data}; - -/* char: 0x5a 'Z' */ - -static const GLubyte ch90data[] = { -0xfc,0x80,0x80,0x40,0x20,0x10,0x8,0x4,0xfc, -}; - -static const BitmapCharRec ch90 = {6,9,-1,0,8,ch90data}; - -/* char: 0x59 'Y' */ - -static const GLubyte ch89data[] = { -0x10,0x10,0x10,0x10,0x10,0x28,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch89 = {7,9,0,0,8,ch89data}; - -/* char: 0x58 'X' */ - -static const GLubyte ch88data[] = { -0x82,0x82,0x44,0x28,0x10,0x28,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch88 = {7,9,0,0,8,ch88data}; - -/* char: 0x57 'W' */ - -static const GLubyte ch87data[] = { -0x44,0xaa,0x92,0x92,0x92,0x82,0x82,0x82,0x82, -}; - -static const BitmapCharRec ch87 = {7,9,0,0,8,ch87data}; - -/* char: 0x56 'V' */ - -static const GLubyte ch86data[] = { -0x10,0x28,0x28,0x28,0x44,0x44,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch86 = {7,9,0,0,8,ch86data}; - -/* char: 0x55 'U' */ - -static const GLubyte ch85data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84, -}; - -static const BitmapCharRec ch85 = {6,9,-1,0,8,ch85data}; - -/* char: 0x54 'T' */ - -static const GLubyte ch84data[] = { -0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xfe, -}; - -static const BitmapCharRec ch84 = {7,9,0,0,8,ch84data}; - -/* char: 0x53 'S' */ - -static const GLubyte ch83data[] = { -0x78,0x84,0x4,0x4,0x78,0x80,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch83 = {6,9,-1,0,8,ch83data}; - -/* char: 0x52 'R' */ - -static const GLubyte ch82data[] = { -0x84,0x88,0x90,0xa0,0xf8,0x84,0x84,0x84,0xf8, -}; - -static const BitmapCharRec ch82 = {6,9,-1,0,8,ch82data}; - -/* char: 0x51 'Q' */ - -static const GLubyte ch81data[] = { -0x4,0x78,0x94,0xa4,0x84,0x84,0x84,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch81 = {6,10,-1,1,8,ch81data}; - -/* char: 0x50 'P' */ - -static const GLubyte ch80data[] = { -0x80,0x80,0x80,0x80,0xf8,0x84,0x84,0x84,0xf8, -}; - -static const BitmapCharRec ch80 = {6,9,-1,0,8,ch80data}; - -/* char: 0x4f 'O' */ - -static const GLubyte ch79data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch79 = {6,9,-1,0,8,ch79data}; - -/* char: 0x4e 'N' */ - -static const GLubyte ch78data[] = { -0x84,0x84,0x84,0x8c,0x94,0xa4,0xc4,0x84,0x84, -}; - -static const BitmapCharRec ch78 = {6,9,-1,0,8,ch78data}; - -/* char: 0x4d 'M' */ - -static const GLubyte ch77data[] = { -0x82,0x82,0x82,0x92,0x92,0xaa,0xc6,0x82,0x82, -}; - -static const BitmapCharRec ch77 = {7,9,0,0,8,ch77data}; - -/* char: 0x4c 'L' */ - -static const GLubyte ch76data[] = { -0xfc,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch76 = {6,9,-1,0,8,ch76data}; - -/* char: 0x4b 'K' */ - -static const GLubyte ch75data[] = { -0x84,0x88,0x90,0xa0,0xc0,0xa0,0x90,0x88,0x84, -}; - -static const BitmapCharRec ch75 = {6,9,-1,0,8,ch75data}; - -/* char: 0x4a 'J' */ - -static const GLubyte ch74data[] = { -0x70,0x88,0x8,0x8,0x8,0x8,0x8,0x8,0x3c, -}; - -static const BitmapCharRec ch74 = {6,9,-1,0,8,ch74data}; - -/* char: 0x49 'I' */ - -static const GLubyte ch73data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xf8, -}; - -static const BitmapCharRec ch73 = {5,9,-1,0,8,ch73data}; - -/* char: 0x48 'H' */ - -static const GLubyte ch72data[] = { -0x84,0x84,0x84,0x84,0xfc,0x84,0x84,0x84,0x84, -}; - -static const BitmapCharRec ch72 = {6,9,-1,0,8,ch72data}; - -/* char: 0x47 'G' */ - -static const GLubyte ch71data[] = { -0x74,0x8c,0x84,0x9c,0x80,0x80,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch71 = {6,9,-1,0,8,ch71data}; - -/* char: 0x46 'F' */ - -static const GLubyte ch70data[] = { -0x80,0x80,0x80,0x80,0xf0,0x80,0x80,0x80,0xfc, -}; - -static const BitmapCharRec ch70 = {6,9,-1,0,8,ch70data}; - -/* char: 0x45 'E' */ - -static const GLubyte ch69data[] = { -0xfc,0x80,0x80,0x80,0xf0,0x80,0x80,0x80,0xfc, -}; - -static const BitmapCharRec ch69 = {6,9,-1,0,8,ch69data}; - -/* char: 0x44 'D' */ - -static const GLubyte ch68data[] = { -0xfc,0x42,0x42,0x42,0x42,0x42,0x42,0x42,0xfc, -}; - -static const BitmapCharRec ch68 = {7,9,0,0,8,ch68data}; - -/* char: 0x43 'C' */ - -static const GLubyte ch67data[] = { -0x78,0x84,0x80,0x80,0x80,0x80,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch67 = {6,9,-1,0,8,ch67data}; - -/* char: 0x42 'B' */ - -static const GLubyte ch66data[] = { -0xfc,0x42,0x42,0x42,0x7c,0x42,0x42,0x42,0xfc, -}; - -static const BitmapCharRec ch66 = {7,9,0,0,8,ch66data}; - -/* char: 0x41 'A' */ - -static const GLubyte ch65data[] = { -0x84,0x84,0x84,0xfc,0x84,0x84,0x84,0x48,0x30, -}; - -static const BitmapCharRec ch65 = {6,9,-1,0,8,ch65data}; - -/* char: 0x40 '@' */ - -static const GLubyte ch64data[] = { -0x78,0x80,0x94,0xac,0xa4,0x9c,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch64 = {6,9,-1,0,8,ch64data}; - -/* char: 0x3f '?' */ - -static const GLubyte ch63data[] = { -0x10,0x0,0x10,0x10,0x8,0x4,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch63 = {6,9,-1,0,8,ch63data}; - -/* char: 0x3e '>' */ - -static const GLubyte ch62data[] = { -0x80,0x40,0x20,0x10,0x8,0x10,0x20,0x40,0x80, -}; - -static const BitmapCharRec ch62 = {5,9,-1,0,8,ch62data}; - -/* char: 0x3d '=' */ - -static const GLubyte ch61data[] = { -0xfc,0x0,0x0,0xfc, -}; - -static const BitmapCharRec ch61 = {6,4,-1,-2,8,ch61data}; - -/* char: 0x3c '<' */ - -static const GLubyte ch60data[] = { -0x8,0x10,0x20,0x40,0x80,0x40,0x20,0x10,0x8, -}; - -static const BitmapCharRec ch60 = {5,9,-2,0,8,ch60data}; - -/* char: 0x3b ';' */ - -static const GLubyte ch59data[] = { -0x80,0x60,0x70,0x0,0x0,0x20,0x70,0x20, -}; - -static const BitmapCharRec ch59 = {4,8,-1,1,8,ch59data}; - -/* char: 0x3a ':' */ - -static const GLubyte ch58data[] = { -0x40,0xe0,0x40,0x0,0x0,0x40,0xe0,0x40, -}; - -static const BitmapCharRec ch58 = {3,8,-2,1,8,ch58data}; - -/* char: 0x39 '9' */ - -static const GLubyte ch57data[] = { -0x70,0x8,0x4,0x4,0x74,0x8c,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch57 = {6,9,-1,0,8,ch57data}; - -/* char: 0x38 '8' */ - -static const GLubyte ch56data[] = { -0x78,0x84,0x84,0x84,0x78,0x84,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch56 = {6,9,-1,0,8,ch56data}; - -/* char: 0x37 '7' */ - -static const GLubyte ch55data[] = { -0x40,0x40,0x20,0x20,0x10,0x10,0x8,0x4,0xfc, -}; - -static const BitmapCharRec ch55 = {6,9,-1,0,8,ch55data}; - -/* char: 0x36 '6' */ - -static const GLubyte ch54data[] = { -0x78,0x84,0x84,0xc4,0xb8,0x80,0x80,0x40,0x38, -}; - -static const BitmapCharRec ch54 = {6,9,-1,0,8,ch54data}; - -/* char: 0x35 '5' */ - -static const GLubyte ch53data[] = { -0x78,0x84,0x4,0x4,0xc4,0xb8,0x80,0x80,0xfc, -}; - -static const BitmapCharRec ch53 = {6,9,-1,0,8,ch53data}; - -/* char: 0x34 '4' */ - -static const GLubyte ch52data[] = { -0x8,0x8,0xfc,0x88,0x88,0x48,0x28,0x18,0x8, -}; - -static const BitmapCharRec ch52 = {6,9,-1,0,8,ch52data}; - -/* char: 0x33 '3' */ - -static const GLubyte ch51data[] = { -0x78,0x84,0x4,0x4,0x38,0x10,0x8,0x4,0xfc, -}; - -static const BitmapCharRec ch51 = {6,9,-1,0,8,ch51data}; - -/* char: 0x32 '2' */ - -static const GLubyte ch50data[] = { -0xfc,0x80,0x40,0x30,0x8,0x4,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch50 = {6,9,-1,0,8,ch50data}; - -/* char: 0x31 '1' */ - -static const GLubyte ch49data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xa0,0x60,0x20, -}; - -static const BitmapCharRec ch49 = {5,9,-1,0,8,ch49data}; - -/* char: 0x30 '0' */ - -static const GLubyte ch48data[] = { -0x30,0x48,0x84,0x84,0x84,0x84,0x84,0x48,0x30, -}; - -static const BitmapCharRec ch48 = {6,9,-1,0,8,ch48data}; - -/* char: 0x2f '/' */ - -static const GLubyte ch47data[] = { -0x80,0x80,0x40,0x20,0x10,0x8,0x4,0x2,0x2, -}; - -static const BitmapCharRec ch47 = {7,9,0,0,8,ch47data}; - -/* char: 0x2e '.' */ - -static const GLubyte ch46data[] = { -0x40,0xe0,0x40, -}; - -static const BitmapCharRec ch46 = {3,3,-2,1,8,ch46data}; - -/* char: 0x2d '-' */ - -static const GLubyte ch45data[] = { -0xfc, -}; - -static const BitmapCharRec ch45 = {6,1,-1,-4,8,ch45data}; - -/* char: 0x2c ',' */ - -static const GLubyte ch44data[] = { -0x80,0x60,0x70, -}; - -static const BitmapCharRec ch44 = {4,3,-1,1,8,ch44data}; - -/* char: 0x2b '+' */ - -static const GLubyte ch43data[] = { -0x20,0x20,0xf8,0x20,0x20, -}; - -static const BitmapCharRec ch43 = {5,5,-1,-2,8,ch43data}; - -/* char: 0x2a '*' */ - -static const GLubyte ch42data[] = { -0x48,0x30,0xfc,0x30,0x48, -}; - -static const BitmapCharRec ch42 = {6,5,-1,-2,8,ch42data}; - -/* char: 0x29 ')' */ - -static const GLubyte ch41data[] = { -0x80,0x40,0x40,0x20,0x20,0x20,0x40,0x40,0x80, -}; - -static const BitmapCharRec ch41 = {3,9,-2,0,8,ch41data}; - -/* char: 0x28 '(' */ - -static const GLubyte ch40data[] = { -0x20,0x40,0x40,0x80,0x80,0x80,0x40,0x40,0x20, -}; - -static const BitmapCharRec ch40 = {3,9,-3,0,8,ch40data}; - -/* char: 0x27 ''' */ - -static const GLubyte ch39data[] = { -0x80,0x60,0x70, -}; - -static const BitmapCharRec ch39 = {4,3,-1,-6,8,ch39data}; - -/* char: 0x26 '&' */ - -static const GLubyte ch38data[] = { -0x74,0x88,0x94,0x60,0x90,0x90,0x60, -}; - -static const BitmapCharRec ch38 = {6,7,-1,0,8,ch38data}; - -/* char: 0x25 '%' */ - -static const GLubyte ch37data[] = { -0x88,0x54,0x48,0x20,0x10,0x10,0x48,0xa4,0x44, -}; - -static const BitmapCharRec ch37 = {6,9,-1,0,8,ch37data}; - -/* char: 0x24 '$' */ - -static const GLubyte ch36data[] = { -0x20,0xf0,0x28,0x70,0xa0,0x78,0x20, -}; - -static const BitmapCharRec ch36 = {5,7,-1,-1,8,ch36data}; - -/* char: 0x23 '#' */ - -static const GLubyte ch35data[] = { -0x48,0x48,0xfc,0x48,0xfc,0x48,0x48, -}; - -static const BitmapCharRec ch35 = {6,7,-1,-1,8,ch35data}; - -/* char: 0x22 '"' */ - -static const GLubyte ch34data[] = { -0x90,0x90,0x90, -}; - -static const BitmapCharRec ch34 = {4,3,-2,-6,8,ch34data}; - -/* char: 0x21 '!' */ - -static const GLubyte ch33data[] = { -0x80,0x0,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch33 = {1,9,-3,0,8,ch33data}; - -/* char: 0x1f */ - -static const GLubyte ch31data[] = { -0x80, -}; - -static const BitmapCharRec ch31 = {1,1,-3,-3,8,ch31data}; - -/* char: 0x1e */ - -static const GLubyte ch30data[] = { -0xdc,0x62,0x20,0x20,0x20,0x70,0x20,0x22,0x1c, -}; - -static const BitmapCharRec ch30 = {7,9,0,0,8,ch30data}; - -/* char: 0x1d */ - -static const GLubyte ch29data[] = { -0x80,0x40,0xfe,0x10,0xfe,0x4,0x2, -}; - -static const BitmapCharRec ch29 = {7,7,0,0,8,ch29data}; - -/* char: 0x1c */ - -static const GLubyte ch28data[] = { -0x88,0x48,0x48,0x48,0x48,0xfc, -}; - -static const BitmapCharRec ch28 = {6,6,-1,0,8,ch28data}; - -/* char: 0x1b */ - -static const GLubyte ch27data[] = { -0xfe,0x80,0x20,0x8,0x2,0x8,0x20,0x80, -}; - -static const BitmapCharRec ch27 = {7,8,0,0,8,ch27data}; - -/* char: 0x1a */ - -static const GLubyte ch26data[] = { -0xfe,0x2,0x8,0x20,0x80,0x20,0x8,0x2, -}; - -static const BitmapCharRec ch26 = {7,8,0,0,8,ch26data}; - -/* char: 0x19 */ - -static const GLubyte ch25data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch25 = {1,13,-3,2,8,ch25data}; - -/* char: 0x18 */ - -static const GLubyte ch24data[] = { -0x10,0x10,0x10,0x10,0x10,0xff, -}; - -static const BitmapCharRec ch24 = {8,6,0,2,8,ch24data}; - -/* char: 0x17 */ - -static const GLubyte ch23data[] = { -0xff,0x10,0x10,0x10,0x10,0x10,0x10,0x10, -}; - -static const BitmapCharRec ch23 = {8,8,0,-3,8,ch23data}; - -/* char: 0x16 */ - -static const GLubyte ch22data[] = { -0x10,0x10,0x10,0x10,0x10,0xf0,0x10,0x10,0x10,0x10,0x10,0x10,0x10, -}; - -static const BitmapCharRec ch22 = {4,13,0,2,8,ch22data}; - -/* char: 0x15 */ - -static const GLubyte ch21data[] = { -0x80,0x80,0x80,0x80,0x80,0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch21 = {5,13,-3,2,8,ch21data}; - -/* char: 0x14 */ - -static const GLubyte ch20data[] = { -0xff, -}; - -static const BitmapCharRec ch20 = {8,1,0,1,8,ch20data}; - -/* char: 0x13 */ - -static const GLubyte ch19data[] = { -0xff, -}; - -static const BitmapCharRec ch19 = {8,1,0,-1,8,ch19data}; - -/* char: 0x12 */ - -static const GLubyte ch18data[] = { -0xff, -}; - -static const BitmapCharRec ch18 = {8,1,0,-3,8,ch18data}; - -/* char: 0x11 */ - -static const GLubyte ch17data[] = { -0xff, -}; - -static const BitmapCharRec ch17 = {8,1,0,-5,8,ch17data}; - -/* char: 0x10 */ - -static const GLubyte ch16data[] = { -0xff, -}; - -static const BitmapCharRec ch16 = {8,1,0,-7,8,ch16data}; - -/* char: 0xf */ - -static const GLubyte ch15data[] = { -0x10,0x10,0x10,0x10,0x10,0xff,0x10,0x10,0x10,0x10,0x10,0x10,0x10, -}; - -static const BitmapCharRec ch15 = {8,13,0,2,8,ch15data}; - -/* char: 0xe */ - -static const GLubyte ch14data[] = { -0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch14 = {5,8,-3,-3,8,ch14data}; - -/* char: 0xd */ - -static const GLubyte ch13data[] = { -0x80,0x80,0x80,0x80,0x80,0xf8, -}; - -static const BitmapCharRec ch13 = {5,6,-3,2,8,ch13data}; - -/* char: 0xc */ - -static const GLubyte ch12data[] = { -0x10,0x10,0x10,0x10,0x10,0xf0, -}; - -static const BitmapCharRec ch12 = {4,6,0,2,8,ch12data}; - -/* char: 0xb */ - -static const GLubyte ch11data[] = { -0xf0,0x10,0x10,0x10,0x10,0x10,0x10,0x10, -}; - -static const BitmapCharRec ch11 = {4,8,0,-3,8,ch11data}; - -/* char: 0xa */ - -static const GLubyte ch10data[] = { -0x8,0x8,0x8,0x8,0x3e,0x20,0x50,0x88,0x88, -}; - -static const BitmapCharRec ch10 = {7,9,0,2,8,ch10data}; - -/* char: 0x9 */ - -static const GLubyte ch9data[] = { -0x3e,0x20,0x20,0x20,0x88,0x98,0xa8,0xc8,0x88, -}; - -static const BitmapCharRec ch9 = {7,9,0,2,8,ch9data}; - -/* char: 0x8 */ - -static const GLubyte ch8data[] = { -0xfe,0x10,0x10,0xfe,0x10,0x10, -}; - -static const BitmapCharRec ch8 = {7,6,0,0,8,ch8data}; - -/* char: 0x7 */ - -static const GLubyte ch7data[] = { -0x70,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch7 = {5,4,-1,-5,8,ch7data}; - -/* char: 0x6 */ - -static const GLubyte ch6data[] = { -0x20,0x20,0x3c,0x20,0x3e,0xf8,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch6 = {7,9,0,2,8,ch6data}; - -/* char: 0x5 */ - -static const GLubyte ch5data[] = { -0x22,0x22,0x3c,0x22,0x3c,0x78,0x80,0x80,0x78, -}; - -static const BitmapCharRec ch5 = {7,9,0,2,8,ch5data}; - -/* char: 0x4 */ - -static const GLubyte ch4data[] = { -0x10,0x10,0x1c,0x10,0x9e,0x80,0xe0,0x80,0xf0, -}; - -static const BitmapCharRec ch4 = {7,9,0,2,8,ch4data}; - -/* char: 0x3 */ - -static const GLubyte ch3data[] = { -0x8,0x8,0x8,0x3e,0x88,0x88,0xf8,0x88,0x88, -}; - -static const BitmapCharRec ch3 = {7,9,0,2,8,ch3data}; - -/* char: 0x2 */ - -static const GLubyte ch2data[] = { -0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa, -}; - -static const BitmapCharRec ch2 = {8,12,0,2,8,ch2data}; - -/* char: 0x1 */ - -static const GLubyte ch1data[] = { -0x10,0x38,0x7c,0xfe,0x7c,0x38,0x10, -}; - -static const BitmapCharRec ch1 = {7,7,0,-1,8,ch1data}; - -static const BitmapCharRec * const chars[] = { -&ch0, -&ch1, -&ch2, -&ch3, -&ch4, -&ch5, -&ch6, -&ch7, -&ch8, -&ch9, -&ch10, -&ch11, -&ch12, -&ch13, -&ch14, -&ch15, -&ch16, -&ch17, -&ch18, -&ch19, -&ch20, -&ch21, -&ch22, -&ch23, -&ch24, -&ch25, -&ch26, -&ch27, -&ch28, -&ch29, -&ch30, -&ch31, -&ch32, -&ch33, -&ch34, -&ch35, -&ch36, -&ch37, -&ch38, -&ch39, -&ch40, -&ch41, -&ch42, -&ch43, -&ch44, -&ch45, -&ch46, -&ch47, -&ch48, -&ch49, -&ch50, -&ch51, -&ch52, -&ch53, -&ch54, -&ch55, -&ch56, -&ch57, -&ch58, -&ch59, -&ch60, -&ch61, -&ch62, -&ch63, -&ch64, -&ch65, -&ch66, -&ch67, -&ch68, -&ch69, -&ch70, -&ch71, -&ch72, -&ch73, -&ch74, -&ch75, -&ch76, -&ch77, -&ch78, -&ch79, -&ch80, -&ch81, -&ch82, -&ch83, -&ch84, -&ch85, -&ch86, -&ch87, -&ch88, -&ch89, -&ch90, -&ch91, -&ch92, -&ch93, -&ch94, -&ch95, -&ch96, -&ch97, -&ch98, -&ch99, -&ch100, -&ch101, -&ch102, -&ch103, -&ch104, -&ch105, -&ch106, -&ch107, -&ch108, -&ch109, -&ch110, -&ch111, -&ch112, -&ch113, -&ch114, -&ch115, -&ch116, -&ch117, -&ch118, -&ch119, -&ch120, -&ch121, -&ch122, -&ch123, -&ch124, -&ch125, -&ch126, -&ch127, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -&ch160, -&ch161, -&ch162, -&ch163, -&ch164, -&ch165, -&ch166, -&ch167, -&ch168, -&ch169, -&ch170, -&ch171, -&ch172, -&ch173, -&ch174, -&ch175, -&ch176, -&ch177, -&ch178, -&ch179, -&ch180, -&ch181, -&ch182, -&ch183, -&ch184, -&ch185, -&ch186, -&ch187, -&ch188, -&ch189, -&ch190, -&ch191, -&ch192, -&ch193, -&ch194, -&ch195, -&ch196, -&ch197, -&ch198, -&ch199, -&ch200, -&ch201, -&ch202, -&ch203, -&ch204, -&ch205, -&ch206, -&ch207, -&ch208, -&ch209, -&ch210, -&ch211, -&ch212, -&ch213, -&ch214, -&ch215, -&ch216, -&ch217, -&ch218, -&ch219, -&ch220, -&ch221, -&ch222, -&ch223, -&ch224, -&ch225, -&ch226, -&ch227, -&ch228, -&ch229, -&ch230, -&ch231, -&ch232, -&ch233, -&ch234, -&ch235, -&ch236, -&ch237, -&ch238, -&ch239, -&ch240, -&ch241, -&ch242, -&ch243, -&ch244, -&ch245, -&ch246, -&ch247, -&ch248, -&ch249, -&ch250, -&ch251, -&ch252, -&ch253, -&ch254, -&ch255, -}; - -#if !defined(__IBMCPP__) -const -#endif -BitmapFontRec glutBitmap8By13 = { -"-misc-fixed-medium-r-normal--13-120-75-75-C-80-iso8859-1", -256, -0, -chars -}; - + +/* GENERATED FILE -- DO NOT MODIFY */ + +#define glutBitmap8By13 XXX +#include "glutbitmap.h" +#undef glutBitmap8By13 + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch0data[] = { 0x0 }; +static const BitmapCharRec ch0 = {1,1,0,0,8,ch0data}; +#else +static const BitmapCharRec ch0 = {0,0,0,0,8,0}; +#endif + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch32data[] = { 0x0 }; +static const BitmapCharRec ch32 = {1,1,0,0,8,ch32data}; +#else +static const BitmapCharRec ch32 = {0,0,0,0,8,0}; +#endif + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch127data[] = { 0x0 }; +static const BitmapCharRec ch127 = {1,1,0,0,8,ch127data}; +#else +static const BitmapCharRec ch127 = {0,0,0,0,8,0}; +#endif + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch160data[] = { 0x0 }; +static const BitmapCharRec ch160 = {1,1,0,0,8,ch160data}; +#else +static const BitmapCharRec ch160 = {0,0,0,0,8,0}; +#endif + +/* char: 0xff */ + +static const GLubyte ch255data[] = { +0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x0,0x0,0x48,0x48, +}; + +static const BitmapCharRec ch255 = {6,12,-1,2,8,ch255data}; + +/* char: 0xfe */ + +static const GLubyte ch254data[] = { +0x80,0x80,0xb8,0xc4,0x84,0x84,0xc4,0xb8,0x80,0x80, +}; + +static const BitmapCharRec ch254 = {6,10,-1,2,8,ch254data}; + +/* char: 0xfd */ + +static const GLubyte ch253data[] = { +0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x0,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch253 = {6,12,-1,2,8,ch253data}; + +/* char: 0xfc */ + +static const GLubyte ch252data[] = { +0x74,0x88,0x88,0x88,0x88,0x88,0x0,0x0,0x48,0x48, +}; + +static const BitmapCharRec ch252 = {6,10,-1,0,8,ch252data}; + +/* char: 0xfb */ + +static const GLubyte ch251data[] = { +0x74,0x88,0x88,0x88,0x88,0x88,0x0,0x0,0x48,0x30, +}; + +static const BitmapCharRec ch251 = {6,10,-1,0,8,ch251data}; + +/* char: 0xfa */ + +static const GLubyte ch250data[] = { +0x74,0x88,0x88,0x88,0x88,0x88,0x0,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch250 = {6,10,-1,0,8,ch250data}; + +/* char: 0xf9 */ + +static const GLubyte ch249data[] = { +0x74,0x88,0x88,0x88,0x88,0x88,0x0,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch249 = {6,10,-1,0,8,ch249data}; + +/* char: 0xf8 */ + +static const GLubyte ch248data[] = { +0x80,0x78,0xc4,0xa4,0x94,0x8c,0x78,0x4, +}; + +static const BitmapCharRec ch248 = {6,8,-1,1,8,ch248data}; + +/* char: 0xf7 */ + +static const GLubyte ch247data[] = { +0x20,0x20,0x0,0xf8,0x0,0x20,0x20, +}; + +static const BitmapCharRec ch247 = {5,7,-1,-1,8,ch247data}; + +/* char: 0xf6 */ + +static const GLubyte ch246data[] = { +0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x48,0x48, +}; + +static const BitmapCharRec ch246 = {6,10,-1,0,8,ch246data}; + +/* char: 0xf5 */ + +static const GLubyte ch245data[] = { +0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch245 = {6,10,-1,0,8,ch245data}; + +/* char: 0xf4 */ + +static const GLubyte ch244data[] = { +0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x48,0x30, +}; + +static const BitmapCharRec ch244 = {6,10,-1,0,8,ch244data}; + +/* char: 0xf3 */ + +static const GLubyte ch243data[] = { +0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch243 = {6,10,-1,0,8,ch243data}; + +/* char: 0xf2 */ + +static const GLubyte ch242data[] = { +0x78,0x84,0x84,0x84,0x84,0x78,0x0,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch242 = {6,10,-1,0,8,ch242data}; + +/* char: 0xf1 */ + +static const GLubyte ch241data[] = { +0x84,0x84,0x84,0x84,0xc4,0xb8,0x0,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch241 = {6,10,-1,0,8,ch241data}; + +/* char: 0xf0 */ + +static const GLubyte ch240data[] = { +0x78,0x84,0x84,0x84,0x84,0x78,0x8,0x50,0x30,0x48, +}; + +static const BitmapCharRec ch240 = {6,10,-1,0,8,ch240data}; + +/* char: 0xef */ + +static const GLubyte ch239data[] = { +0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x0,0x50,0x50, +}; + +static const BitmapCharRec ch239 = {5,10,-1,0,8,ch239data}; + +/* char: 0xee */ + +static const GLubyte ch238data[] = { +0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x0,0x90,0x60, +}; + +static const BitmapCharRec ch238 = {5,10,-1,0,8,ch238data}; + +/* char: 0xed */ + +static const GLubyte ch237data[] = { +0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x0,0x40,0x20, +}; + +static const BitmapCharRec ch237 = {5,10,-1,0,8,ch237data}; + +/* char: 0xec */ + +static const GLubyte ch236data[] = { +0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x0,0x20,0x40, +}; + +static const BitmapCharRec ch236 = {5,10,-1,0,8,ch236data}; + +/* char: 0xeb */ + +static const GLubyte ch235data[] = { +0x78,0x84,0x80,0xfc,0x84,0x78,0x0,0x0,0x48,0x48, +}; + +static const BitmapCharRec ch235 = {6,10,-1,0,8,ch235data}; + +/* char: 0xea */ + +static const GLubyte ch234data[] = { +0x78,0x84,0x80,0xfc,0x84,0x78,0x0,0x0,0x48,0x30, +}; + +static const BitmapCharRec ch234 = {6,10,-1,0,8,ch234data}; + +/* char: 0xe9 */ + +static const GLubyte ch233data[] = { +0x78,0x84,0x80,0xfc,0x84,0x78,0x0,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch233 = {6,10,-1,0,8,ch233data}; + +/* char: 0xe8 */ + +static const GLubyte ch232data[] = { +0x78,0x84,0x80,0xfc,0x84,0x78,0x0,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch232 = {6,10,-1,0,8,ch232data}; + +/* char: 0xe7 */ + +static const GLubyte ch231data[] = { +0x20,0x10,0x78,0x84,0x80,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch231 = {6,8,-1,2,8,ch231data}; + +/* char: 0xe6 */ + +static const GLubyte ch230data[] = { +0x6c,0x92,0x90,0x7c,0x12,0x6c, +}; + +static const BitmapCharRec ch230 = {7,6,0,0,8,ch230data}; + +/* char: 0xe5 */ + +static const GLubyte ch229data[] = { +0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x30,0x48,0x30, +}; + +static const BitmapCharRec ch229 = {6,10,-1,0,8,ch229data}; + +/* char: 0xe4 */ + +static const GLubyte ch228data[] = { +0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x48,0x48, +}; + +static const BitmapCharRec ch228 = {6,10,-1,0,8,ch228data}; + +/* char: 0xe3 */ + +static const GLubyte ch227data[] = { +0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch227 = {6,10,-1,0,8,ch227data}; + +/* char: 0xe2 */ + +static const GLubyte ch226data[] = { +0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x48,0x30, +}; + +static const BitmapCharRec ch226 = {6,10,-1,0,8,ch226data}; + +/* char: 0xe1 */ + +static const GLubyte ch225data[] = { +0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch225 = {6,10,-1,0,8,ch225data}; + +/* char: 0xe0 */ + +static const GLubyte ch224data[] = { +0x74,0x8c,0x84,0x7c,0x4,0x78,0x0,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch224 = {6,10,-1,0,8,ch224data}; + +/* char: 0xdf */ + +static const GLubyte ch223data[] = { +0x80,0xb8,0xc4,0x84,0x84,0xf8,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch223 = {6,9,-1,1,8,ch223data}; + +/* char: 0xde */ + +static const GLubyte ch222data[] = { +0x80,0x80,0x80,0xf8,0x84,0x84,0x84,0xf8,0x80, +}; + +static const BitmapCharRec ch222 = {6,9,-1,0,8,ch222data}; + +/* char: 0xdd */ + +static const GLubyte ch221data[] = { +0x20,0x20,0x20,0x20,0x50,0x88,0x88,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch221 = {5,10,-1,0,8,ch221data}; + +/* char: 0xdc */ + +static const GLubyte ch220data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x48,0x48, +}; + +static const BitmapCharRec ch220 = {6,10,-1,0,8,ch220data}; + +/* char: 0xdb */ + +static const GLubyte ch219data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x48,0x30, +}; + +static const BitmapCharRec ch219 = {6,10,-1,0,8,ch219data}; + +/* char: 0xda */ + +static const GLubyte ch218data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch218 = {6,10,-1,0,8,ch218data}; + +/* char: 0xd9 */ + +static const GLubyte ch217data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch217 = {6,10,-1,0,8,ch217data}; + +/* char: 0xd8 */ + +static const GLubyte ch216data[] = { +0x80,0x78,0xc4,0xa4,0xa4,0xa4,0x94,0x94,0x8c,0x78,0x4, +}; + +static const BitmapCharRec ch216 = {6,11,-1,1,8,ch216data}; + +/* char: 0xd7 */ + +static const GLubyte ch215data[] = { +0x84,0x48,0x30,0x30,0x48,0x84, +}; + +static const BitmapCharRec ch215 = {6,6,-1,-1,8,ch215data}; + +/* char: 0xd6 */ + +static const GLubyte ch214data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch214 = {7,10,0,0,8,ch214data}; + +/* char: 0xd5 */ + +static const GLubyte ch213data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x28,0x14, +}; + +static const BitmapCharRec ch213 = {7,10,0,0,8,ch213data}; + +/* char: 0xd4 */ + +static const GLubyte ch212data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x24,0x18, +}; + +static const BitmapCharRec ch212 = {7,10,0,0,8,ch212data}; + +/* char: 0xd3 */ + +static const GLubyte ch211data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x10,0x8, +}; + +static const BitmapCharRec ch211 = {7,10,0,0,8,ch211data}; + +/* char: 0xd2 */ + +static const GLubyte ch210data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x8,0x10, +}; + +static const BitmapCharRec ch210 = {7,10,0,0,8,ch210data}; + +/* char: 0xd1 */ + +static const GLubyte ch209data[] = { +0x82,0x86,0x8a,0x92,0xa2,0xc2,0x82,0x0,0x28,0x14, +}; + +static const BitmapCharRec ch209 = {7,10,0,0,8,ch209data}; + +/* char: 0xd0 */ + +static const GLubyte ch208data[] = { +0xfc,0x42,0x42,0x42,0xe2,0x42,0x42,0x42,0xfc, +}; + +static const BitmapCharRec ch208 = {7,9,0,0,8,ch208data}; + +/* char: 0xcf */ + +static const GLubyte ch207data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x50,0x50, +}; + +static const BitmapCharRec ch207 = {5,10,-1,0,8,ch207data}; + +/* char: 0xce */ + +static const GLubyte ch206data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x48,0x30, +}; + +static const BitmapCharRec ch206 = {5,10,-1,0,8,ch206data}; + +/* char: 0xcd */ + +static const GLubyte ch205data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch205 = {5,10,-1,0,8,ch205data}; + +/* char: 0xcc */ + +static const GLubyte ch204data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch204 = {5,10,-1,0,8,ch204data}; + +/* char: 0xcb */ + +static const GLubyte ch203data[] = { +0xfc,0x80,0x80,0xf0,0x80,0x80,0xfc,0x0,0x48,0x48, +}; + +static const BitmapCharRec ch203 = {6,10,-1,0,8,ch203data}; + +/* char: 0xca */ + +static const GLubyte ch202data[] = { +0xfc,0x80,0x80,0xf0,0x80,0x80,0xfc,0x0,0x48,0x30, +}; + +static const BitmapCharRec ch202 = {6,10,-1,0,8,ch202data}; + +/* char: 0xc9 */ + +static const GLubyte ch201data[] = { +0xfc,0x80,0x80,0xf0,0x80,0x80,0xfc,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch201 = {6,10,-1,0,8,ch201data}; + +/* char: 0xc8 */ + +static const GLubyte ch200data[] = { +0xfc,0x80,0x80,0xf0,0x80,0x80,0xfc,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch200 = {6,10,-1,0,8,ch200data}; + +/* char: 0xc7 */ + +static const GLubyte ch199data[] = { +0x20,0x10,0x78,0x84,0x80,0x80,0x80,0x80,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch199 = {6,11,-1,2,8,ch199data}; + +/* char: 0xc6 */ + +static const GLubyte ch198data[] = { +0x9e,0x90,0x90,0xf0,0x9c,0x90,0x90,0x90,0x6e, +}; + +static const BitmapCharRec ch198 = {7,9,0,0,8,ch198data}; + +/* char: 0xc5 */ + +static const GLubyte ch197data[] = { +0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x30,0x48,0x30, +}; + +static const BitmapCharRec ch197 = {6,10,-1,0,8,ch197data}; + +/* char: 0xc4 */ + +static const GLubyte ch196data[] = { +0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x48,0x48, +}; + +static const BitmapCharRec ch196 = {6,10,-1,0,8,ch196data}; + +/* char: 0xc3 */ + +static const GLubyte ch195data[] = { +0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch195 = {6,10,-1,0,8,ch195data}; + +/* char: 0xc2 */ + +static const GLubyte ch194data[] = { +0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x48,0x30, +}; + +static const BitmapCharRec ch194 = {6,10,-1,0,8,ch194data}; + +/* char: 0xc1 */ + +static const GLubyte ch193data[] = { +0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch193 = {6,10,-1,0,8,ch193data}; + +/* char: 0xc0 */ + +static const GLubyte ch192data[] = { +0x84,0x84,0xfc,0x84,0x84,0x48,0x30,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch192 = {6,10,-1,0,8,ch192data}; + +/* char: 0xbf */ + +static const GLubyte ch191data[] = { +0x78,0x84,0x84,0x80,0x40,0x20,0x20,0x0,0x20, +}; + +static const BitmapCharRec ch191 = {6,9,-1,0,8,ch191data}; + +/* char: 0xbe */ + +static const GLubyte ch190data[] = { +0x6,0x1a,0x12,0xa,0x66,0x92,0x10,0x20,0x90,0x60, +}; + +static const BitmapCharRec ch190 = {7,10,0,0,8,ch190data}; + +/* char: 0xbd */ + +static const GLubyte ch189data[] = { +0x1e,0x10,0xc,0x2,0xf2,0x4c,0x40,0x40,0xc0,0x40, +}; + +static const BitmapCharRec ch189 = {7,10,0,0,8,ch189data}; + +/* char: 0xbc */ + +static const GLubyte ch188data[] = { +0x6,0x1a,0x12,0xa,0xe6,0x42,0x40,0x40,0xc0,0x40, +}; + +static const BitmapCharRec ch188 = {7,10,0,0,8,ch188data}; + +/* char: 0xbb */ + +static const GLubyte ch187data[] = { +0x90,0x48,0x24,0x12,0x24,0x48,0x90, +}; + +static const BitmapCharRec ch187 = {7,7,0,-1,8,ch187data}; + +/* char: 0xba */ + +static const GLubyte ch186data[] = { +0xf0,0x0,0x60,0x90,0x90,0x60, +}; + +static const BitmapCharRec ch186 = {4,6,-1,-3,8,ch186data}; + +/* char: 0xb9 */ + +static const GLubyte ch185data[] = { +0xe0,0x40,0x40,0x40,0xc0,0x40, +}; + +static const BitmapCharRec ch185 = {3,6,-1,-4,8,ch185data}; + +/* char: 0xb8 */ + +static const GLubyte ch184data[] = { +0xc0,0x40, +}; + +static const BitmapCharRec ch184 = {2,2,-3,2,8,ch184data}; + +/* char: 0xb7 */ + +static const GLubyte ch183data[] = { +0xc0, +}; + +static const BitmapCharRec ch183 = {2,1,-3,-4,8,ch183data}; + +/* char: 0xb6 */ + +static const GLubyte ch182data[] = { +0x28,0x28,0x28,0x28,0x68,0xe8,0xe8,0xe8,0x7c, +}; + +static const BitmapCharRec ch182 = {6,9,-1,0,8,ch182data}; + +/* char: 0xb5 */ + +static const GLubyte ch181data[] = { +0x80,0xb4,0xcc,0x84,0x84,0x84,0x84, +}; + +static const BitmapCharRec ch181 = {6,7,-1,1,8,ch181data}; + +/* char: 0xb4 */ + +static const GLubyte ch180data[] = { +0x80,0x40, +}; + +static const BitmapCharRec ch180 = {2,2,-3,-8,8,ch180data}; + +/* char: 0xb3 */ + +static const GLubyte ch179data[] = { +0x60,0x90,0x10,0x20,0x90,0x60, +}; + +static const BitmapCharRec ch179 = {4,6,-1,-4,8,ch179data}; + +/* char: 0xb2 */ + +static const GLubyte ch178data[] = { +0xf0,0x80,0x60,0x10,0x90,0x60, +}; + +static const BitmapCharRec ch178 = {4,6,-1,-4,8,ch178data}; + +/* char: 0xb1 */ + +static const GLubyte ch177data[] = { +0xf8,0x0,0x20,0x20,0xf8,0x20,0x20, +}; + +static const BitmapCharRec ch177 = {5,7,-1,-1,8,ch177data}; + +/* char: 0xb0 */ + +static const GLubyte ch176data[] = { +0x60,0x90,0x90,0x60, +}; + +static const BitmapCharRec ch176 = {4,4,-2,-5,8,ch176data}; + +/* char: 0xaf */ + +static const GLubyte ch175data[] = { +0xfc, +}; + +static const BitmapCharRec ch175 = {6,1,-1,-8,8,ch175data}; + +/* char: 0xae */ + +static const GLubyte ch174data[] = { +0x38,0x44,0xaa,0xb2,0xaa,0xaa,0x92,0x44,0x38, +}; + +static const BitmapCharRec ch174 = {7,9,0,-1,8,ch174data}; + +/* char: 0xad */ + +static const GLubyte ch173data[] = { +0xfc, +}; + +static const BitmapCharRec ch173 = {6,1,-1,-4,8,ch173data}; + +/* char: 0xac */ + +static const GLubyte ch172data[] = { +0x4,0x4,0x4,0xfc, +}; + +static const BitmapCharRec ch172 = {6,4,-1,-1,8,ch172data}; + +/* char: 0xab */ + +static const GLubyte ch171data[] = { +0x12,0x24,0x48,0x90,0x48,0x24,0x12, +}; + +static const BitmapCharRec ch171 = {7,7,0,-1,8,ch171data}; + +/* char: 0xaa */ + +static const GLubyte ch170data[] = { +0xf8,0x0,0x78,0x88,0x78,0x8,0x70, +}; + +static const BitmapCharRec ch170 = {5,7,-1,-2,8,ch170data}; + +/* char: 0xa9 */ + +static const GLubyte ch169data[] = { +0x38,0x44,0x92,0xaa,0xa2,0xaa,0x92,0x44,0x38, +}; + +static const BitmapCharRec ch169 = {7,9,0,-1,8,ch169data}; + +/* char: 0xa8 */ + +static const GLubyte ch168data[] = { +0xd8, +}; + +static const BitmapCharRec ch168 = {5,1,-1,-8,8,ch168data}; + +/* char: 0xa7 */ + +static const GLubyte ch167data[] = { +0x60,0x90,0x10,0x60,0x90,0x90,0x60,0x80,0x90,0x60, +}; + +static const BitmapCharRec ch167 = {4,10,-2,0,8,ch167data}; + +/* char: 0xa6 */ + +static const GLubyte ch166data[] = { +0x80,0x80,0x80,0x80,0x0,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch166 = {1,9,-3,0,8,ch166data}; + +/* char: 0xa5 */ + +static const GLubyte ch165data[] = { +0x10,0x10,0x7c,0x10,0x7c,0x28,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch165 = {7,9,0,0,8,ch165data}; + +/* char: 0xa4 */ + +static const GLubyte ch164data[] = { +0x84,0x78,0x48,0x48,0x78,0x84, +}; + +static const BitmapCharRec ch164 = {6,6,-1,-1,8,ch164data}; + +/* char: 0xa3 */ + +static const GLubyte ch163data[] = { +0xdc,0x62,0x20,0x20,0x20,0x70,0x20,0x22,0x1c, +}; + +static const BitmapCharRec ch163 = {7,9,0,0,8,ch163data}; + +/* char: 0xa2 */ + +static const GLubyte ch162data[] = { +0x20,0x70,0xa8,0xa0,0xa0,0xa8,0x70,0x20, +}; + +static const BitmapCharRec ch162 = {5,8,-1,-1,8,ch162data}; + +/* char: 0xa1 */ + +static const GLubyte ch161data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, +}; + +static const BitmapCharRec ch161 = {1,9,-3,0,8,ch161data}; + +/* char: 0x7e '~' */ + +static const GLubyte ch126data[] = { +0x90,0xa8,0x48, +}; + +static const BitmapCharRec ch126 = {5,3,-1,-6,8,ch126data}; + +/* char: 0x7d '}' */ + +static const GLubyte ch125data[] = { +0xe0,0x10,0x10,0x20,0x18,0x20,0x10,0x10,0xe0, +}; + +static const BitmapCharRec ch125 = {5,9,-1,0,8,ch125data}; + +/* char: 0x7c '|' */ + +static const GLubyte ch124data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch124 = {1,9,-3,0,8,ch124data}; + +/* char: 0x7b '{' */ + +static const GLubyte ch123data[] = { +0x38,0x40,0x40,0x20,0xc0,0x20,0x40,0x40,0x38, +}; + +static const BitmapCharRec ch123 = {5,9,-2,0,8,ch123data}; + +/* char: 0x7a 'z' */ + +static const GLubyte ch122data[] = { +0xfc,0x40,0x20,0x10,0x8,0xfc, +}; + +static const BitmapCharRec ch122 = {6,6,-1,0,8,ch122data}; + +/* char: 0x79 'y' */ + +static const GLubyte ch121data[] = { +0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84, +}; + +static const BitmapCharRec ch121 = {6,8,-1,2,8,ch121data}; + +/* char: 0x78 'x' */ + +static const GLubyte ch120data[] = { +0x84,0x48,0x30,0x30,0x48,0x84, +}; + +static const BitmapCharRec ch120 = {6,6,-1,0,8,ch120data}; + +/* char: 0x77 'w' */ + +static const GLubyte ch119data[] = { +0x44,0xaa,0x92,0x92,0x82,0x82, +}; + +static const BitmapCharRec ch119 = {7,6,0,0,8,ch119data}; + +/* char: 0x76 'v' */ + +static const GLubyte ch118data[] = { +0x20,0x50,0x50,0x88,0x88,0x88, +}; + +static const BitmapCharRec ch118 = {5,6,-1,0,8,ch118data}; + +/* char: 0x75 'u' */ + +static const GLubyte ch117data[] = { +0x74,0x88,0x88,0x88,0x88,0x88, +}; + +static const BitmapCharRec ch117 = {6,6,-1,0,8,ch117data}; + +/* char: 0x74 't' */ + +static const GLubyte ch116data[] = { +0x38,0x44,0x40,0x40,0x40,0xf8,0x40,0x40, +}; + +static const BitmapCharRec ch116 = {6,8,-1,0,8,ch116data}; + +/* char: 0x73 's' */ + +static const GLubyte ch115data[] = { +0x78,0x84,0x18,0x60,0x84,0x78, +}; + +static const BitmapCharRec ch115 = {6,6,-1,0,8,ch115data}; + +/* char: 0x72 'r' */ + +static const GLubyte ch114data[] = { +0x40,0x40,0x40,0x40,0x44,0xb8, +}; + +static const BitmapCharRec ch114 = {6,6,-1,0,8,ch114data}; + +/* char: 0x71 'q' */ + +static const GLubyte ch113data[] = { +0x4,0x4,0x4,0x74,0x8c,0x84,0x8c,0x74, +}; + +static const BitmapCharRec ch113 = {6,8,-1,2,8,ch113data}; + +/* char: 0x70 'p' */ + +static const GLubyte ch112data[] = { +0x80,0x80,0x80,0xb8,0xc4,0x84,0xc4,0xb8, +}; + +static const BitmapCharRec ch112 = {6,8,-1,2,8,ch112data}; + +/* char: 0x6f 'o' */ + +static const GLubyte ch111data[] = { +0x78,0x84,0x84,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch111 = {6,6,-1,0,8,ch111data}; + +/* char: 0x6e 'n' */ + +static const GLubyte ch110data[] = { +0x84,0x84,0x84,0x84,0xc4,0xb8, +}; + +static const BitmapCharRec ch110 = {6,6,-1,0,8,ch110data}; + +/* char: 0x6d 'm' */ + +static const GLubyte ch109data[] = { +0x82,0x92,0x92,0x92,0x92,0xec, +}; + +static const BitmapCharRec ch109 = {7,6,0,0,8,ch109data}; + +/* char: 0x6c 'l' */ + +static const GLubyte ch108data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0x60, +}; + +static const BitmapCharRec ch108 = {5,9,-1,0,8,ch108data}; + +/* char: 0x6b 'k' */ + +static const GLubyte ch107data[] = { +0x84,0x88,0x90,0xe0,0x90,0x88,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch107 = {6,9,-1,0,8,ch107data}; + +/* char: 0x6a 'j' */ + +static const GLubyte ch106data[] = { +0x70,0x88,0x88,0x8,0x8,0x8,0x8,0x18,0x0,0x8, +}; + +static const BitmapCharRec ch106 = {5,10,-1,2,8,ch106data}; + +/* char: 0x69 'i' */ + +static const GLubyte ch105data[] = { +0xf8,0x20,0x20,0x20,0x20,0x60,0x0,0x20, +}; + +static const BitmapCharRec ch105 = {5,8,-1,0,8,ch105data}; + +/* char: 0x68 'h' */ + +static const GLubyte ch104data[] = { +0x84,0x84,0x84,0x84,0xc4,0xb8,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch104 = {6,9,-1,0,8,ch104data}; + +/* char: 0x67 'g' */ + +static const GLubyte ch103data[] = { +0x78,0x84,0x78,0x80,0x70,0x88,0x88,0x74, +}; + +static const BitmapCharRec ch103 = {6,8,-1,2,8,ch103data}; + +/* char: 0x66 'f' */ + +static const GLubyte ch102data[] = { +0x40,0x40,0x40,0x40,0xf8,0x40,0x40,0x44,0x38, +}; + +static const BitmapCharRec ch102 = {6,9,-1,0,8,ch102data}; + +/* char: 0x65 'e' */ + +static const GLubyte ch101data[] = { +0x78,0x84,0x80,0xfc,0x84,0x78, +}; + +static const BitmapCharRec ch101 = {6,6,-1,0,8,ch101data}; + +/* char: 0x64 'd' */ + +static const GLubyte ch100data[] = { +0x74,0x8c,0x84,0x84,0x8c,0x74,0x4,0x4,0x4, +}; + +static const BitmapCharRec ch100 = {6,9,-1,0,8,ch100data}; + +/* char: 0x63 'c' */ + +static const GLubyte ch99data[] = { +0x78,0x84,0x80,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch99 = {6,6,-1,0,8,ch99data}; + +/* char: 0x62 'b' */ + +static const GLubyte ch98data[] = { +0xb8,0xc4,0x84,0x84,0xc4,0xb8,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch98 = {6,9,-1,0,8,ch98data}; + +/* char: 0x61 'a' */ + +static const GLubyte ch97data[] = { +0x74,0x8c,0x84,0x7c,0x4,0x78, +}; + +static const BitmapCharRec ch97 = {6,6,-1,0,8,ch97data}; + +/* char: 0x60 '`' */ + +static const GLubyte ch96data[] = { +0x10,0x60,0xe0, +}; + +static const BitmapCharRec ch96 = {4,3,-2,-6,8,ch96data}; + +/* char: 0x5f '_' */ + +static const GLubyte ch95data[] = { +0xfe, +}; + +static const BitmapCharRec ch95 = {7,1,0,1,8,ch95data}; + +/* char: 0x5e '^' */ + +static const GLubyte ch94data[] = { +0x88,0x50,0x20, +}; + +static const BitmapCharRec ch94 = {5,3,-1,-6,8,ch94data}; + +/* char: 0x5d ']' */ + +static const GLubyte ch93data[] = { +0xf0,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xf0, +}; + +static const BitmapCharRec ch93 = {4,9,-1,0,8,ch93data}; + +/* char: 0x5c '\' */ + +static const GLubyte ch92data[] = { +0x2,0x2,0x4,0x8,0x10,0x20,0x40,0x80,0x80, +}; + +static const BitmapCharRec ch92 = {7,9,0,0,8,ch92data}; + +/* char: 0x5b '[' */ + +static const GLubyte ch91data[] = { +0xf0,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0xf0, +}; + +static const BitmapCharRec ch91 = {4,9,-2,0,8,ch91data}; + +/* char: 0x5a 'Z' */ + +static const GLubyte ch90data[] = { +0xfc,0x80,0x80,0x40,0x20,0x10,0x8,0x4,0xfc, +}; + +static const BitmapCharRec ch90 = {6,9,-1,0,8,ch90data}; + +/* char: 0x59 'Y' */ + +static const GLubyte ch89data[] = { +0x10,0x10,0x10,0x10,0x10,0x28,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch89 = {7,9,0,0,8,ch89data}; + +/* char: 0x58 'X' */ + +static const GLubyte ch88data[] = { +0x82,0x82,0x44,0x28,0x10,0x28,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch88 = {7,9,0,0,8,ch88data}; + +/* char: 0x57 'W' */ + +static const GLubyte ch87data[] = { +0x44,0xaa,0x92,0x92,0x92,0x82,0x82,0x82,0x82, +}; + +static const BitmapCharRec ch87 = {7,9,0,0,8,ch87data}; + +/* char: 0x56 'V' */ + +static const GLubyte ch86data[] = { +0x10,0x28,0x28,0x28,0x44,0x44,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch86 = {7,9,0,0,8,ch86data}; + +/* char: 0x55 'U' */ + +static const GLubyte ch85data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84, +}; + +static const BitmapCharRec ch85 = {6,9,-1,0,8,ch85data}; + +/* char: 0x54 'T' */ + +static const GLubyte ch84data[] = { +0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xfe, +}; + +static const BitmapCharRec ch84 = {7,9,0,0,8,ch84data}; + +/* char: 0x53 'S' */ + +static const GLubyte ch83data[] = { +0x78,0x84,0x4,0x4,0x78,0x80,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch83 = {6,9,-1,0,8,ch83data}; + +/* char: 0x52 'R' */ + +static const GLubyte ch82data[] = { +0x84,0x88,0x90,0xa0,0xf8,0x84,0x84,0x84,0xf8, +}; + +static const BitmapCharRec ch82 = {6,9,-1,0,8,ch82data}; + +/* char: 0x51 'Q' */ + +static const GLubyte ch81data[] = { +0x4,0x78,0x94,0xa4,0x84,0x84,0x84,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch81 = {6,10,-1,1,8,ch81data}; + +/* char: 0x50 'P' */ + +static const GLubyte ch80data[] = { +0x80,0x80,0x80,0x80,0xf8,0x84,0x84,0x84,0xf8, +}; + +static const BitmapCharRec ch80 = {6,9,-1,0,8,ch80data}; + +/* char: 0x4f 'O' */ + +static const GLubyte ch79data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch79 = {6,9,-1,0,8,ch79data}; + +/* char: 0x4e 'N' */ + +static const GLubyte ch78data[] = { +0x84,0x84,0x84,0x8c,0x94,0xa4,0xc4,0x84,0x84, +}; + +static const BitmapCharRec ch78 = {6,9,-1,0,8,ch78data}; + +/* char: 0x4d 'M' */ + +static const GLubyte ch77data[] = { +0x82,0x82,0x82,0x92,0x92,0xaa,0xc6,0x82,0x82, +}; + +static const BitmapCharRec ch77 = {7,9,0,0,8,ch77data}; + +/* char: 0x4c 'L' */ + +static const GLubyte ch76data[] = { +0xfc,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch76 = {6,9,-1,0,8,ch76data}; + +/* char: 0x4b 'K' */ + +static const GLubyte ch75data[] = { +0x84,0x88,0x90,0xa0,0xc0,0xa0,0x90,0x88,0x84, +}; + +static const BitmapCharRec ch75 = {6,9,-1,0,8,ch75data}; + +/* char: 0x4a 'J' */ + +static const GLubyte ch74data[] = { +0x70,0x88,0x8,0x8,0x8,0x8,0x8,0x8,0x3c, +}; + +static const BitmapCharRec ch74 = {6,9,-1,0,8,ch74data}; + +/* char: 0x49 'I' */ + +static const GLubyte ch73data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xf8, +}; + +static const BitmapCharRec ch73 = {5,9,-1,0,8,ch73data}; + +/* char: 0x48 'H' */ + +static const GLubyte ch72data[] = { +0x84,0x84,0x84,0x84,0xfc,0x84,0x84,0x84,0x84, +}; + +static const BitmapCharRec ch72 = {6,9,-1,0,8,ch72data}; + +/* char: 0x47 'G' */ + +static const GLubyte ch71data[] = { +0x74,0x8c,0x84,0x9c,0x80,0x80,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch71 = {6,9,-1,0,8,ch71data}; + +/* char: 0x46 'F' */ + +static const GLubyte ch70data[] = { +0x80,0x80,0x80,0x80,0xf0,0x80,0x80,0x80,0xfc, +}; + +static const BitmapCharRec ch70 = {6,9,-1,0,8,ch70data}; + +/* char: 0x45 'E' */ + +static const GLubyte ch69data[] = { +0xfc,0x80,0x80,0x80,0xf0,0x80,0x80,0x80,0xfc, +}; + +static const BitmapCharRec ch69 = {6,9,-1,0,8,ch69data}; + +/* char: 0x44 'D' */ + +static const GLubyte ch68data[] = { +0xfc,0x42,0x42,0x42,0x42,0x42,0x42,0x42,0xfc, +}; + +static const BitmapCharRec ch68 = {7,9,0,0,8,ch68data}; + +/* char: 0x43 'C' */ + +static const GLubyte ch67data[] = { +0x78,0x84,0x80,0x80,0x80,0x80,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch67 = {6,9,-1,0,8,ch67data}; + +/* char: 0x42 'B' */ + +static const GLubyte ch66data[] = { +0xfc,0x42,0x42,0x42,0x7c,0x42,0x42,0x42,0xfc, +}; + +static const BitmapCharRec ch66 = {7,9,0,0,8,ch66data}; + +/* char: 0x41 'A' */ + +static const GLubyte ch65data[] = { +0x84,0x84,0x84,0xfc,0x84,0x84,0x84,0x48,0x30, +}; + +static const BitmapCharRec ch65 = {6,9,-1,0,8,ch65data}; + +/* char: 0x40 '@' */ + +static const GLubyte ch64data[] = { +0x78,0x80,0x94,0xac,0xa4,0x9c,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch64 = {6,9,-1,0,8,ch64data}; + +/* char: 0x3f '?' */ + +static const GLubyte ch63data[] = { +0x10,0x0,0x10,0x10,0x8,0x4,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch63 = {6,9,-1,0,8,ch63data}; + +/* char: 0x3e '>' */ + +static const GLubyte ch62data[] = { +0x80,0x40,0x20,0x10,0x8,0x10,0x20,0x40,0x80, +}; + +static const BitmapCharRec ch62 = {5,9,-1,0,8,ch62data}; + +/* char: 0x3d '=' */ + +static const GLubyte ch61data[] = { +0xfc,0x0,0x0,0xfc, +}; + +static const BitmapCharRec ch61 = {6,4,-1,-2,8,ch61data}; + +/* char: 0x3c '<' */ + +static const GLubyte ch60data[] = { +0x8,0x10,0x20,0x40,0x80,0x40,0x20,0x10,0x8, +}; + +static const BitmapCharRec ch60 = {5,9,-2,0,8,ch60data}; + +/* char: 0x3b ';' */ + +static const GLubyte ch59data[] = { +0x80,0x60,0x70,0x0,0x0,0x20,0x70,0x20, +}; + +static const BitmapCharRec ch59 = {4,8,-1,1,8,ch59data}; + +/* char: 0x3a ':' */ + +static const GLubyte ch58data[] = { +0x40,0xe0,0x40,0x0,0x0,0x40,0xe0,0x40, +}; + +static const BitmapCharRec ch58 = {3,8,-2,1,8,ch58data}; + +/* char: 0x39 '9' */ + +static const GLubyte ch57data[] = { +0x70,0x8,0x4,0x4,0x74,0x8c,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch57 = {6,9,-1,0,8,ch57data}; + +/* char: 0x38 '8' */ + +static const GLubyte ch56data[] = { +0x78,0x84,0x84,0x84,0x78,0x84,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch56 = {6,9,-1,0,8,ch56data}; + +/* char: 0x37 '7' */ + +static const GLubyte ch55data[] = { +0x40,0x40,0x20,0x20,0x10,0x10,0x8,0x4,0xfc, +}; + +static const BitmapCharRec ch55 = {6,9,-1,0,8,ch55data}; + +/* char: 0x36 '6' */ + +static const GLubyte ch54data[] = { +0x78,0x84,0x84,0xc4,0xb8,0x80,0x80,0x40,0x38, +}; + +static const BitmapCharRec ch54 = {6,9,-1,0,8,ch54data}; + +/* char: 0x35 '5' */ + +static const GLubyte ch53data[] = { +0x78,0x84,0x4,0x4,0xc4,0xb8,0x80,0x80,0xfc, +}; + +static const BitmapCharRec ch53 = {6,9,-1,0,8,ch53data}; + +/* char: 0x34 '4' */ + +static const GLubyte ch52data[] = { +0x8,0x8,0xfc,0x88,0x88,0x48,0x28,0x18,0x8, +}; + +static const BitmapCharRec ch52 = {6,9,-1,0,8,ch52data}; + +/* char: 0x33 '3' */ + +static const GLubyte ch51data[] = { +0x78,0x84,0x4,0x4,0x38,0x10,0x8,0x4,0xfc, +}; + +static const BitmapCharRec ch51 = {6,9,-1,0,8,ch51data}; + +/* char: 0x32 '2' */ + +static const GLubyte ch50data[] = { +0xfc,0x80,0x40,0x30,0x8,0x4,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch50 = {6,9,-1,0,8,ch50data}; + +/* char: 0x31 '1' */ + +static const GLubyte ch49data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xa0,0x60,0x20, +}; + +static const BitmapCharRec ch49 = {5,9,-1,0,8,ch49data}; + +/* char: 0x30 '0' */ + +static const GLubyte ch48data[] = { +0x30,0x48,0x84,0x84,0x84,0x84,0x84,0x48,0x30, +}; + +static const BitmapCharRec ch48 = {6,9,-1,0,8,ch48data}; + +/* char: 0x2f '/' */ + +static const GLubyte ch47data[] = { +0x80,0x80,0x40,0x20,0x10,0x8,0x4,0x2,0x2, +}; + +static const BitmapCharRec ch47 = {7,9,0,0,8,ch47data}; + +/* char: 0x2e '.' */ + +static const GLubyte ch46data[] = { +0x40,0xe0,0x40, +}; + +static const BitmapCharRec ch46 = {3,3,-2,1,8,ch46data}; + +/* char: 0x2d '-' */ + +static const GLubyte ch45data[] = { +0xfc, +}; + +static const BitmapCharRec ch45 = {6,1,-1,-4,8,ch45data}; + +/* char: 0x2c ',' */ + +static const GLubyte ch44data[] = { +0x80,0x60,0x70, +}; + +static const BitmapCharRec ch44 = {4,3,-1,1,8,ch44data}; + +/* char: 0x2b '+' */ + +static const GLubyte ch43data[] = { +0x20,0x20,0xf8,0x20,0x20, +}; + +static const BitmapCharRec ch43 = {5,5,-1,-2,8,ch43data}; + +/* char: 0x2a '*' */ + +static const GLubyte ch42data[] = { +0x48,0x30,0xfc,0x30,0x48, +}; + +static const BitmapCharRec ch42 = {6,5,-1,-2,8,ch42data}; + +/* char: 0x29 ')' */ + +static const GLubyte ch41data[] = { +0x80,0x40,0x40,0x20,0x20,0x20,0x40,0x40,0x80, +}; + +static const BitmapCharRec ch41 = {3,9,-2,0,8,ch41data}; + +/* char: 0x28 '(' */ + +static const GLubyte ch40data[] = { +0x20,0x40,0x40,0x80,0x80,0x80,0x40,0x40,0x20, +}; + +static const BitmapCharRec ch40 = {3,9,-3,0,8,ch40data}; + +/* char: 0x27 ''' */ + +static const GLubyte ch39data[] = { +0x80,0x60,0x70, +}; + +static const BitmapCharRec ch39 = {4,3,-1,-6,8,ch39data}; + +/* char: 0x26 '&' */ + +static const GLubyte ch38data[] = { +0x74,0x88,0x94,0x60,0x90,0x90,0x60, +}; + +static const BitmapCharRec ch38 = {6,7,-1,0,8,ch38data}; + +/* char: 0x25 '%' */ + +static const GLubyte ch37data[] = { +0x88,0x54,0x48,0x20,0x10,0x10,0x48,0xa4,0x44, +}; + +static const BitmapCharRec ch37 = {6,9,-1,0,8,ch37data}; + +/* char: 0x24 '$' */ + +static const GLubyte ch36data[] = { +0x20,0xf0,0x28,0x70,0xa0,0x78,0x20, +}; + +static const BitmapCharRec ch36 = {5,7,-1,-1,8,ch36data}; + +/* char: 0x23 '#' */ + +static const GLubyte ch35data[] = { +0x48,0x48,0xfc,0x48,0xfc,0x48,0x48, +}; + +static const BitmapCharRec ch35 = {6,7,-1,-1,8,ch35data}; + +/* char: 0x22 '"' */ + +static const GLubyte ch34data[] = { +0x90,0x90,0x90, +}; + +static const BitmapCharRec ch34 = {4,3,-2,-6,8,ch34data}; + +/* char: 0x21 '!' */ + +static const GLubyte ch33data[] = { +0x80,0x0,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch33 = {1,9,-3,0,8,ch33data}; + +/* char: 0x1f */ + +static const GLubyte ch31data[] = { +0x80, +}; + +static const BitmapCharRec ch31 = {1,1,-3,-3,8,ch31data}; + +/* char: 0x1e */ + +static const GLubyte ch30data[] = { +0xdc,0x62,0x20,0x20,0x20,0x70,0x20,0x22,0x1c, +}; + +static const BitmapCharRec ch30 = {7,9,0,0,8,ch30data}; + +/* char: 0x1d */ + +static const GLubyte ch29data[] = { +0x80,0x40,0xfe,0x10,0xfe,0x4,0x2, +}; + +static const BitmapCharRec ch29 = {7,7,0,0,8,ch29data}; + +/* char: 0x1c */ + +static const GLubyte ch28data[] = { +0x88,0x48,0x48,0x48,0x48,0xfc, +}; + +static const BitmapCharRec ch28 = {6,6,-1,0,8,ch28data}; + +/* char: 0x1b */ + +static const GLubyte ch27data[] = { +0xfe,0x80,0x20,0x8,0x2,0x8,0x20,0x80, +}; + +static const BitmapCharRec ch27 = {7,8,0,0,8,ch27data}; + +/* char: 0x1a */ + +static const GLubyte ch26data[] = { +0xfe,0x2,0x8,0x20,0x80,0x20,0x8,0x2, +}; + +static const BitmapCharRec ch26 = {7,8,0,0,8,ch26data}; + +/* char: 0x19 */ + +static const GLubyte ch25data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch25 = {1,13,-3,2,8,ch25data}; + +/* char: 0x18 */ + +static const GLubyte ch24data[] = { +0x10,0x10,0x10,0x10,0x10,0xff, +}; + +static const BitmapCharRec ch24 = {8,6,0,2,8,ch24data}; + +/* char: 0x17 */ + +static const GLubyte ch23data[] = { +0xff,0x10,0x10,0x10,0x10,0x10,0x10,0x10, +}; + +static const BitmapCharRec ch23 = {8,8,0,-3,8,ch23data}; + +/* char: 0x16 */ + +static const GLubyte ch22data[] = { +0x10,0x10,0x10,0x10,0x10,0xf0,0x10,0x10,0x10,0x10,0x10,0x10,0x10, +}; + +static const BitmapCharRec ch22 = {4,13,0,2,8,ch22data}; + +/* char: 0x15 */ + +static const GLubyte ch21data[] = { +0x80,0x80,0x80,0x80,0x80,0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch21 = {5,13,-3,2,8,ch21data}; + +/* char: 0x14 */ + +static const GLubyte ch20data[] = { +0xff, +}; + +static const BitmapCharRec ch20 = {8,1,0,1,8,ch20data}; + +/* char: 0x13 */ + +static const GLubyte ch19data[] = { +0xff, +}; + +static const BitmapCharRec ch19 = {8,1,0,-1,8,ch19data}; + +/* char: 0x12 */ + +static const GLubyte ch18data[] = { +0xff, +}; + +static const BitmapCharRec ch18 = {8,1,0,-3,8,ch18data}; + +/* char: 0x11 */ + +static const GLubyte ch17data[] = { +0xff, +}; + +static const BitmapCharRec ch17 = {8,1,0,-5,8,ch17data}; + +/* char: 0x10 */ + +static const GLubyte ch16data[] = { +0xff, +}; + +static const BitmapCharRec ch16 = {8,1,0,-7,8,ch16data}; + +/* char: 0xf */ + +static const GLubyte ch15data[] = { +0x10,0x10,0x10,0x10,0x10,0xff,0x10,0x10,0x10,0x10,0x10,0x10,0x10, +}; + +static const BitmapCharRec ch15 = {8,13,0,2,8,ch15data}; + +/* char: 0xe */ + +static const GLubyte ch14data[] = { +0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch14 = {5,8,-3,-3,8,ch14data}; + +/* char: 0xd */ + +static const GLubyte ch13data[] = { +0x80,0x80,0x80,0x80,0x80,0xf8, +}; + +static const BitmapCharRec ch13 = {5,6,-3,2,8,ch13data}; + +/* char: 0xc */ + +static const GLubyte ch12data[] = { +0x10,0x10,0x10,0x10,0x10,0xf0, +}; + +static const BitmapCharRec ch12 = {4,6,0,2,8,ch12data}; + +/* char: 0xb */ + +static const GLubyte ch11data[] = { +0xf0,0x10,0x10,0x10,0x10,0x10,0x10,0x10, +}; + +static const BitmapCharRec ch11 = {4,8,0,-3,8,ch11data}; + +/* char: 0xa */ + +static const GLubyte ch10data[] = { +0x8,0x8,0x8,0x8,0x3e,0x20,0x50,0x88,0x88, +}; + +static const BitmapCharRec ch10 = {7,9,0,2,8,ch10data}; + +/* char: 0x9 */ + +static const GLubyte ch9data[] = { +0x3e,0x20,0x20,0x20,0x88,0x98,0xa8,0xc8,0x88, +}; + +static const BitmapCharRec ch9 = {7,9,0,2,8,ch9data}; + +/* char: 0x8 */ + +static const GLubyte ch8data[] = { +0xfe,0x10,0x10,0xfe,0x10,0x10, +}; + +static const BitmapCharRec ch8 = {7,6,0,0,8,ch8data}; + +/* char: 0x7 */ + +static const GLubyte ch7data[] = { +0x70,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch7 = {5,4,-1,-5,8,ch7data}; + +/* char: 0x6 */ + +static const GLubyte ch6data[] = { +0x20,0x20,0x3c,0x20,0x3e,0xf8,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch6 = {7,9,0,2,8,ch6data}; + +/* char: 0x5 */ + +static const GLubyte ch5data[] = { +0x22,0x22,0x3c,0x22,0x3c,0x78,0x80,0x80,0x78, +}; + +static const BitmapCharRec ch5 = {7,9,0,2,8,ch5data}; + +/* char: 0x4 */ + +static const GLubyte ch4data[] = { +0x10,0x10,0x1c,0x10,0x9e,0x80,0xe0,0x80,0xf0, +}; + +static const BitmapCharRec ch4 = {7,9,0,2,8,ch4data}; + +/* char: 0x3 */ + +static const GLubyte ch3data[] = { +0x8,0x8,0x8,0x3e,0x88,0x88,0xf8,0x88,0x88, +}; + +static const BitmapCharRec ch3 = {7,9,0,2,8,ch3data}; + +/* char: 0x2 */ + +static const GLubyte ch2data[] = { +0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa, +}; + +static const BitmapCharRec ch2 = {8,12,0,2,8,ch2data}; + +/* char: 0x1 */ + +static const GLubyte ch1data[] = { +0x10,0x38,0x7c,0xfe,0x7c,0x38,0x10, +}; + +static const BitmapCharRec ch1 = {7,7,0,-1,8,ch1data}; + +static const BitmapCharRec * const chars[] = { +&ch0, +&ch1, +&ch2, +&ch3, +&ch4, +&ch5, +&ch6, +&ch7, +&ch8, +&ch9, +&ch10, +&ch11, +&ch12, +&ch13, +&ch14, +&ch15, +&ch16, +&ch17, +&ch18, +&ch19, +&ch20, +&ch21, +&ch22, +&ch23, +&ch24, +&ch25, +&ch26, +&ch27, +&ch28, +&ch29, +&ch30, +&ch31, +&ch32, +&ch33, +&ch34, +&ch35, +&ch36, +&ch37, +&ch38, +&ch39, +&ch40, +&ch41, +&ch42, +&ch43, +&ch44, +&ch45, +&ch46, +&ch47, +&ch48, +&ch49, +&ch50, +&ch51, +&ch52, +&ch53, +&ch54, +&ch55, +&ch56, +&ch57, +&ch58, +&ch59, +&ch60, +&ch61, +&ch62, +&ch63, +&ch64, +&ch65, +&ch66, +&ch67, +&ch68, +&ch69, +&ch70, +&ch71, +&ch72, +&ch73, +&ch74, +&ch75, +&ch76, +&ch77, +&ch78, +&ch79, +&ch80, +&ch81, +&ch82, +&ch83, +&ch84, +&ch85, +&ch86, +&ch87, +&ch88, +&ch89, +&ch90, +&ch91, +&ch92, +&ch93, +&ch94, +&ch95, +&ch96, +&ch97, +&ch98, +&ch99, +&ch100, +&ch101, +&ch102, +&ch103, +&ch104, +&ch105, +&ch106, +&ch107, +&ch108, +&ch109, +&ch110, +&ch111, +&ch112, +&ch113, +&ch114, +&ch115, +&ch116, +&ch117, +&ch118, +&ch119, +&ch120, +&ch121, +&ch122, +&ch123, +&ch124, +&ch125, +&ch126, +&ch127, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +&ch160, +&ch161, +&ch162, +&ch163, +&ch164, +&ch165, +&ch166, +&ch167, +&ch168, +&ch169, +&ch170, +&ch171, +&ch172, +&ch173, +&ch174, +&ch175, +&ch176, +&ch177, +&ch178, +&ch179, +&ch180, +&ch181, +&ch182, +&ch183, +&ch184, +&ch185, +&ch186, +&ch187, +&ch188, +&ch189, +&ch190, +&ch191, +&ch192, +&ch193, +&ch194, +&ch195, +&ch196, +&ch197, +&ch198, +&ch199, +&ch200, +&ch201, +&ch202, +&ch203, +&ch204, +&ch205, +&ch206, +&ch207, +&ch208, +&ch209, +&ch210, +&ch211, +&ch212, +&ch213, +&ch214, +&ch215, +&ch216, +&ch217, +&ch218, +&ch219, +&ch220, +&ch221, +&ch222, +&ch223, +&ch224, +&ch225, +&ch226, +&ch227, +&ch228, +&ch229, +&ch230, +&ch231, +&ch232, +&ch233, +&ch234, +&ch235, +&ch236, +&ch237, +&ch238, +&ch239, +&ch240, +&ch241, +&ch242, +&ch243, +&ch244, +&ch245, +&ch246, +&ch247, +&ch248, +&ch249, +&ch250, +&ch251, +&ch252, +&ch253, +&ch254, +&ch255, +}; + +#if !defined(__IBMCPP__) +const +#endif +BitmapFontRec glutBitmap8By13 = { +"-misc-fixed-medium-r-normal--13-120-75-75-C-80-iso8859-1", +256, +0, +chars +}; + diff --git a/src/glut/os2/glut_9x15.cpp b/src/glut/os2/glut_9x15.cpp index a3a02a4c5f..d228a2b150 100644 --- a/src/glut/os2/glut_9x15.cpp +++ b/src/glut/os2/glut_9x15.cpp @@ -1,2077 +1,2077 @@ - -/* GENERATED FILE -- DO NOT MODIFY */ - -#define glutBitmap9By15 XXX -#include "glutbitmap.h" -#undef glutBitmap9By15 - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch0data[] = { 0x0 }; -static const BitmapCharRec ch0 = {1,1,0,0,9,ch0data}; -#else -static const BitmapCharRec ch0 = {0,0,0,0,9,0}; -#endif - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch32data[] = { 0x0 }; -static const BitmapCharRec ch32 = {1,1,0,0,9,ch32data}; -#else -static const BitmapCharRec ch32 = {0,0,0,0,9,0}; -#endif - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch127data[] = { 0x0 }; -static const BitmapCharRec ch127 = {1,1,0,0,9,ch127data}; -#else -static const BitmapCharRec ch127 = {0,0,0,0,9,0}; -#endif - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch160data[] = { 0x0 }; -static const BitmapCharRec ch160 = {1,1,0,0,9,ch160data}; -#else -static const BitmapCharRec ch160 = {0,0,0,0,9,0}; -#endif - -/* char: 0xff */ - -static const GLubyte ch255data[] = { -0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch255 = {6,14,-1,3,9,ch255data}; - -/* char: 0xfe */ - -static const GLubyte ch254data[] = { -0x80,0x80,0x80,0xbc,0xc2,0x82,0x82,0x82,0xc2,0xbc,0x80,0x80, -}; - -static const BitmapCharRec ch254 = {7,12,-1,3,9,ch254data}; - -/* char: 0xfd */ - -static const GLubyte ch253data[] = { -0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch253 = {6,14,-1,3,9,ch253data}; - -/* char: 0xfc */ - -static const GLubyte ch252data[] = { -0x7a,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch252 = {7,11,-1,0,9,ch252data}; - -/* char: 0xfb */ - -static const GLubyte ch251data[] = { -0x7a,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x44,0x38, -}; - -static const BitmapCharRec ch251 = {7,11,-1,0,9,ch251data}; - -/* char: 0xfa */ - -static const GLubyte ch250data[] = { -0x7a,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch250 = {7,11,-1,0,9,ch250data}; - -/* char: 0xf9 */ - -static const GLubyte ch249data[] = { -0x7a,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x18,0x20, -}; - -static const BitmapCharRec ch249 = {7,11,-1,0,9,ch249data}; - -/* char: 0xf8 */ - -static const GLubyte ch248data[] = { -0x80,0x7c,0xa2,0xa2,0x92,0x8a,0x8a,0x7c,0x2, -}; - -static const BitmapCharRec ch248 = {7,9,-1,1,9,ch248data}; - -/* char: 0xf7 */ - -static const GLubyte ch247data[] = { -0x10,0x38,0x10,0x0,0xfe,0x0,0x10,0x38,0x10, -}; - -static const BitmapCharRec ch247 = {7,9,-1,0,9,ch247data}; - -/* char: 0xf6 */ - -static const GLubyte ch246data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch246 = {7,11,-1,0,9,ch246data}; - -/* char: 0xf5 */ - -static const GLubyte ch245data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch245 = {7,11,-1,0,9,ch245data}; - -/* char: 0xf4 */ - -static const GLubyte ch244data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x44,0x38, -}; - -static const BitmapCharRec ch244 = {7,11,-1,0,9,ch244data}; - -/* char: 0xf3 */ - -static const GLubyte ch243data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch243 = {7,11,-1,0,9,ch243data}; - -/* char: 0xf2 */ - -static const GLubyte ch242data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x18,0x20, -}; - -static const BitmapCharRec ch242 = {7,11,-1,0,9,ch242data}; - -/* char: 0xf1 */ - -static const GLubyte ch241data[] = { -0x82,0x82,0x82,0x82,0x82,0xc2,0xbc,0x0,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch241 = {7,11,-1,0,9,ch241data}; - -/* char: 0xf0 */ - -static const GLubyte ch240data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x8,0x50,0x30,0x48, -}; - -static const BitmapCharRec ch240 = {7,11,-1,0,9,ch240data}; - -/* char: 0xef */ - -static const GLubyte ch239data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x50,0x50, -}; - -static const BitmapCharRec ch239 = {5,11,-2,0,9,ch239data}; - -/* char: 0xee */ - -static const GLubyte ch238data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x90,0x60, -}; - -static const BitmapCharRec ch238 = {5,11,-2,0,9,ch238data}; - -/* char: 0xed */ - -static const GLubyte ch237data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x60,0x10, -}; - -static const BitmapCharRec ch237 = {5,11,-2,0,9,ch237data}; - -/* char: 0xec */ - -static const GLubyte ch236data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x30,0x40, -}; - -static const BitmapCharRec ch236 = {5,11,-2,0,9,ch236data}; - -/* char: 0xeb */ - -static const GLubyte ch235data[] = { -0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c,0x0,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch235 = {7,11,-1,0,9,ch235data}; - -/* char: 0xea */ - -static const GLubyte ch234data[] = { -0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c,0x0,0x0,0x44,0x38, -}; - -static const BitmapCharRec ch234 = {7,11,-1,0,9,ch234data}; - -/* char: 0xe9 */ - -static const GLubyte ch233data[] = { -0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c,0x0,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch233 = {7,11,-1,0,9,ch233data}; - -/* char: 0xe8 */ - -static const GLubyte ch232data[] = { -0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c,0x0,0x0,0x18,0x20, -}; - -static const BitmapCharRec ch232 = {7,11,-1,0,9,ch232data}; - -/* char: 0xe7 */ - -static const GLubyte ch231data[] = { -0x30,0x48,0x18,0x7c,0x82,0x80,0x80,0x80,0x82,0x7c, -}; - -static const BitmapCharRec ch231 = {7,10,-1,3,9,ch231data}; - -/* char: 0xe6 */ - -static const GLubyte ch230data[] = { -0x6e,0x92,0x90,0x7c,0x12,0x92,0x6c, -}; - -static const BitmapCharRec ch230 = {7,7,-1,0,9,ch230data}; - -/* char: 0xe5 */ - -static const GLubyte ch229data[] = { -0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x18,0x24,0x18, -}; - -static const BitmapCharRec ch229 = {7,11,-1,0,9,ch229data}; - -/* char: 0xe4 */ - -static const GLubyte ch228data[] = { -0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch228 = {7,11,-1,0,9,ch228data}; - -/* char: 0xe3 */ - -static const GLubyte ch227data[] = { -0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch227 = {7,11,-1,0,9,ch227data}; - -/* char: 0xe2 */ - -static const GLubyte ch226data[] = { -0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x44,0x38, -}; - -static const BitmapCharRec ch226 = {7,11,-1,0,9,ch226data}; - -/* char: 0xe1 */ - -static const GLubyte ch225data[] = { -0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch225 = {7,11,-1,0,9,ch225data}; - -/* char: 0xe0 */ - -static const GLubyte ch224data[] = { -0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x18,0x20, -}; - -static const BitmapCharRec ch224 = {7,11,-1,0,9,ch224data}; - -/* char: 0xdf */ - -static const GLubyte ch223data[] = { -0x80,0xbc,0xc2,0x82,0x82,0xfc,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch223 = {7,9,-1,1,9,ch223data}; - -/* char: 0xde */ - -static const GLubyte ch222data[] = { -0x80,0x80,0x80,0xfc,0x82,0x82,0x82,0xfc,0x80,0x80, -}; - -static const BitmapCharRec ch222 = {7,10,-1,0,9,ch222data}; - -/* char: 0xdd */ - -static const GLubyte ch221data[] = { -0x10,0x10,0x10,0x10,0x28,0x44,0x82,0x82,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch221 = {7,11,-1,0,9,ch221data}; - -/* char: 0xdc */ - -static const GLubyte ch220data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch220 = {7,11,-1,0,9,ch220data}; - -/* char: 0xdb */ - -static const GLubyte ch219data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x0,0x44,0x38, -}; - -static const BitmapCharRec ch219 = {7,11,-1,0,9,ch219data}; - -/* char: 0xda */ - -static const GLubyte ch218data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch218 = {7,11,-1,0,9,ch218data}; - -/* char: 0xd9 */ - -static const GLubyte ch217data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x0,0x18,0x20, -}; - -static const BitmapCharRec ch217 = {7,11,-1,0,9,ch217data}; - -/* char: 0xd8 */ - -static const GLubyte ch216data[] = { -0x80,0x7c,0xc2,0xa2,0xa2,0x92,0x92,0x8a,0x8a,0x86,0x7c,0x2, -}; - -static const BitmapCharRec ch216 = {7,12,-1,1,9,ch216data}; - -/* char: 0xd7 */ - -static const GLubyte ch215data[] = { -0x82,0x44,0x28,0x10,0x28,0x44,0x82, -}; - -static const BitmapCharRec ch215 = {7,7,-1,-1,9,ch215data}; - -/* char: 0xd6 */ - -static const GLubyte ch214data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch214 = {7,11,-1,0,9,ch214data}; - -/* char: 0xd5 */ - -static const GLubyte ch213data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch213 = {7,11,-1,0,9,ch213data}; - -/* char: 0xd4 */ - -static const GLubyte ch212data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x44,0x38, -}; - -static const BitmapCharRec ch212 = {7,11,-1,0,9,ch212data}; - -/* char: 0xd3 */ - -static const GLubyte ch211data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch211 = {7,11,-1,0,9,ch211data}; - -/* char: 0xd2 */ - -static const GLubyte ch210data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x18,0x20, -}; - -static const BitmapCharRec ch210 = {7,11,-1,0,9,ch210data}; - -/* char: 0xd1 */ - -static const GLubyte ch209data[] = { -0x82,0x86,0x8a,0x92,0x92,0xa2,0xc2,0x82,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch209 = {7,11,-1,0,9,ch209data}; - -/* char: 0xd0 */ - -static const GLubyte ch208data[] = { -0xfc,0x42,0x42,0x42,0x42,0xf2,0x42,0x42,0x42,0xfc, -}; - -static const BitmapCharRec ch208 = {7,10,-1,0,9,ch208data}; - -/* char: 0xcf */ - -static const GLubyte ch207data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x50,0x50, -}; - -static const BitmapCharRec ch207 = {5,11,-2,0,9,ch207data}; - -/* char: 0xce */ - -static const GLubyte ch206data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x88,0x70, -}; - -static const BitmapCharRec ch206 = {5,11,-2,0,9,ch206data}; - -/* char: 0xcd */ - -static const GLubyte ch205data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x60,0x10, -}; - -static const BitmapCharRec ch205 = {5,11,-2,0,9,ch205data}; - -/* char: 0xcc */ - -static const GLubyte ch204data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x30,0x40, -}; - -static const BitmapCharRec ch204 = {5,11,-2,0,9,ch204data}; - -/* char: 0xcb */ - -static const GLubyte ch203data[] = { -0xfe,0x40,0x40,0x40,0x78,0x40,0x40,0xfe,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch203 = {7,11,-1,0,9,ch203data}; - -/* char: 0xca */ - -static const GLubyte ch202data[] = { -0xfe,0x40,0x40,0x40,0x78,0x40,0x40,0xfe,0x0,0x44,0x38, -}; - -static const BitmapCharRec ch202 = {7,11,-1,0,9,ch202data}; - -/* char: 0xc9 */ - -static const GLubyte ch201data[] = { -0xfe,0x40,0x40,0x40,0x78,0x40,0x40,0xfe,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch201 = {7,11,-1,0,9,ch201data}; - -/* char: 0xc8 */ - -static const GLubyte ch200data[] = { -0xfe,0x40,0x40,0x40,0x78,0x40,0x40,0xfe,0x0,0x18,0x20, -}; - -static const BitmapCharRec ch200 = {7,11,-1,0,9,ch200data}; - -/* char: 0xc7 */ - -static const GLubyte ch199data[] = { -0x30,0x48,0x18,0x7c,0x82,0x80,0x80,0x80,0x80,0x80,0x80,0x82,0x7c, -}; - -static const BitmapCharRec ch199 = {7,13,-1,3,9,ch199data}; - -/* char: 0xc6 */ - -static const GLubyte ch198data[] = { -0x9e,0x90,0x90,0x90,0xfc,0x90,0x90,0x90,0x90,0x6e, -}; - -static const BitmapCharRec ch198 = {7,10,-1,0,9,ch198data}; - -/* char: 0xc5 */ - -static const GLubyte ch197data[] = { -0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x10,0x28,0x10, -}; - -static const BitmapCharRec ch197 = {7,11,-1,0,9,ch197data}; - -/* char: 0xc4 */ - -static const GLubyte ch196data[] = { -0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x28,0x28, -}; - -static const BitmapCharRec ch196 = {7,11,-1,0,9,ch196data}; - -/* char: 0xc3 */ - -static const GLubyte ch195data[] = { -0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch195 = {7,11,-1,0,9,ch195data}; - -/* char: 0xc2 */ - -static const GLubyte ch194data[] = { -0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x44,0x38, -}; - -static const BitmapCharRec ch194 = {7,11,-1,0,9,ch194data}; - -/* char: 0xc1 */ - -static const GLubyte ch193data[] = { -0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x30,0x8, -}; - -static const BitmapCharRec ch193 = {7,11,-1,0,9,ch193data}; - -/* char: 0xc0 */ - -static const GLubyte ch192data[] = { -0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x18,0x20, -}; - -static const BitmapCharRec ch192 = {7,11,-1,0,9,ch192data}; - -/* char: 0xbf */ - -static const GLubyte ch191data[] = { -0x7c,0x82,0x82,0x80,0x40,0x20,0x10,0x10,0x0,0x10, -}; - -static const BitmapCharRec ch191 = {7,10,-1,0,9,ch191data}; - -/* char: 0xbe */ - -static const GLubyte ch190data[] = { -0x6,0x1a,0x12,0xa,0x66,0x92,0x10,0x20,0x90,0x60, -}; - -static const BitmapCharRec ch190 = {7,10,-1,0,9,ch190data}; - -/* char: 0xbd */ - -static const GLubyte ch189data[] = { -0x1e,0x10,0xc,0x2,0xf2,0x4c,0x40,0x40,0xc0,0x40, -}; - -static const BitmapCharRec ch189 = {7,10,-1,0,9,ch189data}; - -/* char: 0xbc */ - -static const GLubyte ch188data[] = { -0x6,0x1a,0x12,0xa,0xe6,0x42,0x40,0x40,0xc0,0x40, -}; - -static const BitmapCharRec ch188 = {7,10,-1,0,9,ch188data}; - -/* char: 0xbb */ - -static const GLubyte ch187data[] = { -0x90,0x48,0x24,0x12,0x12,0x24,0x48,0x90, -}; - -static const BitmapCharRec ch187 = {7,8,-1,-1,9,ch187data}; - -/* char: 0xba */ - -static const GLubyte ch186data[] = { -0xf8,0x0,0x70,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch186 = {5,6,-1,-5,9,ch186data}; - -/* char: 0xb9 */ - -static const GLubyte ch185data[] = { -0xe0,0x40,0x40,0x40,0xc0,0x40, -}; - -static const BitmapCharRec ch185 = {3,6,-1,-4,9,ch185data}; - -/* char: 0xb8 */ - -static const GLubyte ch184data[] = { -0x60,0x90,0x30, -}; - -static const BitmapCharRec ch184 = {4,3,-2,3,9,ch184data}; - -/* char: 0xb7 */ - -static const GLubyte ch183data[] = { -0xc0,0xc0, -}; - -static const BitmapCharRec ch183 = {2,2,-4,-4,9,ch183data}; - -/* char: 0xb6 */ - -static const GLubyte ch182data[] = { -0xa,0xa,0xa,0xa,0xa,0x7a,0x8a,0x8a,0x8a,0x7e, -}; - -static const BitmapCharRec ch182 = {7,10,-1,0,9,ch182data}; - -/* char: 0xb5 */ - -static const GLubyte ch181data[] = { -0x80,0x80,0xba,0xc6,0x82,0x82,0x82,0x82,0x82, -}; - -static const BitmapCharRec ch181 = {7,9,-1,2,9,ch181data}; - -/* char: 0xb4 */ - -static const GLubyte ch180data[] = { -0xc0,0x20, -}; - -static const BitmapCharRec ch180 = {3,2,-3,-9,9,ch180data}; - -/* char: 0xb3 */ - -static const GLubyte ch179data[] = { -0x60,0x90,0x10,0x20,0x90,0x60, -}; - -static const BitmapCharRec ch179 = {4,6,-1,-4,9,ch179data}; - -/* char: 0xb2 */ - -static const GLubyte ch178data[] = { -0xf0,0x80,0x60,0x10,0x90,0x60, -}; - -static const BitmapCharRec ch178 = {4,6,-1,-4,9,ch178data}; - -/* char: 0xb1 */ - -static const GLubyte ch177data[] = { -0xfe,0x0,0x10,0x10,0x10,0xfe,0x10,0x10,0x10, -}; - -static const BitmapCharRec ch177 = {7,9,-1,-1,9,ch177data}; - -/* char: 0xb0 */ - -static const GLubyte ch176data[] = { -0x60,0x90,0x90,0x60, -}; - -static const BitmapCharRec ch176 = {4,4,-3,-6,9,ch176data}; - -/* char: 0xaf */ - -static const GLubyte ch175data[] = { -0xfc, -}; - -static const BitmapCharRec ch175 = {6,1,-1,-9,9,ch175data}; - -/* char: 0xae */ - -static const GLubyte ch174data[] = { -0x3c,0x42,0xa5,0xa9,0xbd,0xa5,0xb9,0x42,0x3c, -}; - -static const BitmapCharRec ch174 = {8,9,0,-1,9,ch174data}; - -/* char: 0xad */ - -static const GLubyte ch173data[] = { -0xfc, -}; - -static const BitmapCharRec ch173 = {6,1,-1,-4,9,ch173data}; - -/* char: 0xac */ - -static const GLubyte ch172data[] = { -0x4,0x4,0x4,0xfc, -}; - -static const BitmapCharRec ch172 = {6,4,-1,-2,9,ch172data}; - -/* char: 0xab */ - -static const GLubyte ch171data[] = { -0x12,0x24,0x48,0x90,0x90,0x48,0x24,0x12, -}; - -static const BitmapCharRec ch171 = {7,8,-1,-1,9,ch171data}; - -/* char: 0xaa */ - -static const GLubyte ch170data[] = { -0xf8,0x0,0x78,0x90,0x70,0x90,0x60, -}; - -static const BitmapCharRec ch170 = {5,7,-3,-3,9,ch170data}; - -/* char: 0xa9 */ - -static const GLubyte ch169data[] = { -0x3c,0x42,0x99,0xa5,0xa1,0xa5,0x99,0x42,0x3c, -}; - -static const BitmapCharRec ch169 = {8,9,0,-1,9,ch169data}; - -/* char: 0xa8 */ - -static const GLubyte ch168data[] = { -0xa0,0xa0, -}; - -static const BitmapCharRec ch168 = {3,2,-3,-9,9,ch168data}; - -/* char: 0xa7 */ - -static const GLubyte ch167data[] = { -0x70,0x88,0x8,0x70,0x88,0x88,0x88,0x70,0x80,0x88,0x70, -}; - -static const BitmapCharRec ch167 = {5,11,-2,1,9,ch167data}; - -/* char: 0xa6 */ - -static const GLubyte ch166data[] = { -0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch166 = {1,11,-4,1,9,ch166data}; - -/* char: 0xa5 */ - -static const GLubyte ch165data[] = { -0x10,0x10,0x10,0x7c,0x10,0x7c,0x28,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch165 = {7,10,-1,0,9,ch165data}; - -/* char: 0xa4 */ - -static const GLubyte ch164data[] = { -0x82,0x7c,0x44,0x44,0x7c,0x82, -}; - -static const BitmapCharRec ch164 = {7,6,-1,-3,9,ch164data}; - -/* char: 0xa3 */ - -static const GLubyte ch163data[] = { -0x5c,0xa2,0x60,0x20,0x20,0xf8,0x20,0x20,0x22,0x1c, -}; - -static const BitmapCharRec ch163 = {7,10,-1,0,9,ch163data}; - -/* char: 0xa2 */ - -static const GLubyte ch162data[] = { -0x40,0x78,0xa4,0xa0,0x90,0x94,0x78,0x8, -}; - -static const BitmapCharRec ch162 = {6,8,-1,0,9,ch162data}; - -/* char: 0xa1 */ - -static const GLubyte ch161data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x0,0x80,0x80, -}; - -static const BitmapCharRec ch161 = {1,11,-4,0,9,ch161data}; - -/* char: 0x7e '~' */ - -static const GLubyte ch126data[] = { -0x8c,0x92,0x62, -}; - -static const BitmapCharRec ch126 = {7,3,-1,-7,9,ch126data}; - -/* char: 0x7d '}' */ - -static const GLubyte ch125data[] = { -0xe0,0x10,0x10,0x10,0x20,0x18,0x18,0x20,0x10,0x10,0x10,0xe0, -}; - -static const BitmapCharRec ch125 = {5,12,-1,1,9,ch125data}; - -/* char: 0x7c '|' */ - -static const GLubyte ch124data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch124 = {1,12,-4,1,9,ch124data}; - -/* char: 0x7b '{' */ - -static const GLubyte ch123data[] = { -0x38,0x40,0x40,0x40,0x20,0xc0,0xc0,0x20,0x40,0x40,0x40,0x38, -}; - -static const BitmapCharRec ch123 = {5,12,-3,1,9,ch123data}; - -/* char: 0x7a 'z' */ - -static const GLubyte ch122data[] = { -0xfe,0x40,0x20,0x10,0x8,0x4,0xfe, -}; - -static const BitmapCharRec ch122 = {7,7,-1,0,9,ch122data}; - -/* char: 0x79 'y' */ - -static const GLubyte ch121data[] = { -0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x84,0x84, -}; - -static const BitmapCharRec ch121 = {6,10,-1,3,9,ch121data}; - -/* char: 0x78 'x' */ - -static const GLubyte ch120data[] = { -0x82,0x44,0x28,0x10,0x28,0x44,0x82, -}; - -static const BitmapCharRec ch120 = {7,7,-1,0,9,ch120data}; - -/* char: 0x77 'w' */ - -static const GLubyte ch119data[] = { -0x44,0xaa,0x92,0x92,0x92,0x82,0x82, -}; - -static const BitmapCharRec ch119 = {7,7,-1,0,9,ch119data}; - -/* char: 0x76 'v' */ - -static const GLubyte ch118data[] = { -0x10,0x28,0x28,0x44,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch118 = {7,7,-1,0,9,ch118data}; - -/* char: 0x75 'u' */ - -static const GLubyte ch117data[] = { -0x7a,0x84,0x84,0x84,0x84,0x84,0x84, -}; - -static const BitmapCharRec ch117 = {7,7,-1,0,9,ch117data}; - -/* char: 0x74 't' */ - -static const GLubyte ch116data[] = { -0x1c,0x22,0x20,0x20,0x20,0x20,0xfc,0x20,0x20, -}; - -static const BitmapCharRec ch116 = {7,9,-1,0,9,ch116data}; - -/* char: 0x73 's' */ - -static const GLubyte ch115data[] = { -0x7c,0x82,0x2,0x7c,0x80,0x82,0x7c, -}; - -static const BitmapCharRec ch115 = {7,7,-1,0,9,ch115data}; - -/* char: 0x72 'r' */ - -static const GLubyte ch114data[] = { -0x40,0x40,0x40,0x40,0x42,0x62,0x9c, -}; - -static const BitmapCharRec ch114 = {7,7,-1,0,9,ch114data}; - -/* char: 0x71 'q' */ - -static const GLubyte ch113data[] = { -0x2,0x2,0x2,0x7a,0x86,0x82,0x82,0x82,0x86,0x7a, -}; - -static const BitmapCharRec ch113 = {7,10,-1,3,9,ch113data}; - -/* char: 0x70 'p' */ - -static const GLubyte ch112data[] = { -0x80,0x80,0x80,0xbc,0xc2,0x82,0x82,0x82,0xc2,0xbc, -}; - -static const BitmapCharRec ch112 = {7,10,-1,3,9,ch112data}; - -/* char: 0x6f 'o' */ - -static const GLubyte ch111data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch111 = {7,7,-1,0,9,ch111data}; - -/* char: 0x6e 'n' */ - -static const GLubyte ch110data[] = { -0x82,0x82,0x82,0x82,0x82,0xc2,0xbc, -}; - -static const BitmapCharRec ch110 = {7,7,-1,0,9,ch110data}; - -/* char: 0x6d 'm' */ - -static const GLubyte ch109data[] = { -0x82,0x92,0x92,0x92,0x92,0x92,0xec, -}; - -static const BitmapCharRec ch109 = {7,7,-1,0,9,ch109data}; - -/* char: 0x6c 'l' */ - -static const GLubyte ch108data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xe0, -}; - -static const BitmapCharRec ch108 = {5,10,-2,0,9,ch108data}; - -/* char: 0x6b 'k' */ - -static const GLubyte ch107data[] = { -0x82,0x8c,0xb0,0xc0,0xb0,0x8c,0x82,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch107 = {7,10,-1,0,9,ch107data}; - -/* char: 0x6a 'j' */ - -static const GLubyte ch106data[] = { -0x78,0x84,0x84,0x84,0x4,0x4,0x4,0x4,0x4,0x1c,0x0,0x0,0xc, -}; - -static const BitmapCharRec ch106 = {6,13,-1,3,9,ch106data}; - -/* char: 0x69 'i' */ - -static const GLubyte ch105data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x60, -}; - -static const BitmapCharRec ch105 = {5,10,-2,0,9,ch105data}; - -/* char: 0x68 'h' */ - -static const GLubyte ch104data[] = { -0x82,0x82,0x82,0x82,0x82,0xc2,0xbc,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch104 = {7,10,-1,0,9,ch104data}; - -/* char: 0x67 'g' */ - -static const GLubyte ch103data[] = { -0x7c,0x82,0x82,0x7c,0x80,0x78,0x84,0x84,0x84,0x7a, -}; - -static const BitmapCharRec ch103 = {7,10,-1,3,9,ch103data}; - -/* char: 0x66 'f' */ - -static const GLubyte ch102data[] = { -0x20,0x20,0x20,0x20,0xf8,0x20,0x20,0x22,0x22,0x1c, -}; - -static const BitmapCharRec ch102 = {7,10,-1,0,9,ch102data}; - -/* char: 0x65 'e' */ - -static const GLubyte ch101data[] = { -0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch101 = {7,7,-1,0,9,ch101data}; - -/* char: 0x64 'd' */ - -static const GLubyte ch100data[] = { -0x7a,0x86,0x82,0x82,0x82,0x86,0x7a,0x2,0x2,0x2, -}; - -static const BitmapCharRec ch100 = {7,10,-1,0,9,ch100data}; - -/* char: 0x63 'c' */ - -static const GLubyte ch99data[] = { -0x7c,0x82,0x80,0x80,0x80,0x82,0x7c, -}; - -static const BitmapCharRec ch99 = {7,7,-1,0,9,ch99data}; - -/* char: 0x62 'b' */ - -static const GLubyte ch98data[] = { -0xbc,0xc2,0x82,0x82,0x82,0xc2,0xbc,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch98 = {7,10,-1,0,9,ch98data}; - -/* char: 0x61 'a' */ - -static const GLubyte ch97data[] = { -0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c, -}; - -static const BitmapCharRec ch97 = {7,7,-1,0,9,ch97data}; - -/* char: 0x60 '`' */ - -static const GLubyte ch96data[] = { -0x10,0x20,0x40,0xc0, -}; - -static const BitmapCharRec ch96 = {4,4,-3,-6,9,ch96data}; - -/* char: 0x5f '_' */ - -static const GLubyte ch95data[] = { -0xff, -}; - -static const BitmapCharRec ch95 = {8,1,0,1,9,ch95data}; - -/* char: 0x5e '^' */ - -static const GLubyte ch94data[] = { -0x82,0x44,0x28,0x10, -}; - -static const BitmapCharRec ch94 = {7,4,-1,-6,9,ch94data}; - -/* char: 0x5d ']' */ - -static const GLubyte ch93data[] = { -0xf0,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xf0, -}; - -static const BitmapCharRec ch93 = {4,12,-2,1,9,ch93data}; - -/* char: 0x5c '\' */ - -static const GLubyte ch92data[] = { -0x2,0x4,0x4,0x8,0x10,0x10,0x20,0x40,0x40,0x80, -}; - -static const BitmapCharRec ch92 = {7,10,-1,0,9,ch92data}; - -/* char: 0x5b '[' */ - -static const GLubyte ch91data[] = { -0xf0,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0xf0, -}; - -static const BitmapCharRec ch91 = {4,12,-3,1,9,ch91data}; - -/* char: 0x5a 'Z' */ - -static const GLubyte ch90data[] = { -0xfe,0x80,0x80,0x40,0x20,0x10,0x8,0x4,0x2,0xfe, -}; - -static const BitmapCharRec ch90 = {7,10,-1,0,9,ch90data}; - -/* char: 0x59 'Y' */ - -static const GLubyte ch89data[] = { -0x10,0x10,0x10,0x10,0x10,0x10,0x28,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch89 = {7,10,-1,0,9,ch89data}; - -/* char: 0x58 'X' */ - -static const GLubyte ch88data[] = { -0x82,0x82,0x44,0x28,0x10,0x10,0x28,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch88 = {7,10,-1,0,9,ch88data}; - -/* char: 0x57 'W' */ - -static const GLubyte ch87data[] = { -0x44,0xaa,0x92,0x92,0x92,0x92,0x82,0x82,0x82,0x82, -}; - -static const BitmapCharRec ch87 = {7,10,-1,0,9,ch87data}; - -/* char: 0x56 'V' */ - -static const GLubyte ch86data[] = { -0x10,0x28,0x28,0x28,0x44,0x44,0x44,0x82,0x82,0x82, -}; - -static const BitmapCharRec ch86 = {7,10,-1,0,9,ch86data}; - -/* char: 0x55 'U' */ - -static const GLubyte ch85data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x82, -}; - -static const BitmapCharRec ch85 = {7,10,-1,0,9,ch85data}; - -/* char: 0x54 'T' */ - -static const GLubyte ch84data[] = { -0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xfe, -}; - -static const BitmapCharRec ch84 = {7,10,-1,0,9,ch84data}; - -/* char: 0x53 'S' */ - -static const GLubyte ch83data[] = { -0x7c,0x82,0x82,0x2,0xc,0x70,0x80,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch83 = {7,10,-1,0,9,ch83data}; - -/* char: 0x52 'R' */ - -static const GLubyte ch82data[] = { -0x82,0x82,0x84,0x88,0x90,0xfc,0x82,0x82,0x82,0xfc, -}; - -static const BitmapCharRec ch82 = {7,10,-1,0,9,ch82data}; - -/* char: 0x51 'Q' */ - -static const GLubyte ch81data[] = { -0x6,0x8,0x7c,0x92,0xa2,0x82,0x82,0x82,0x82,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch81 = {7,12,-1,2,9,ch81data}; - -/* char: 0x50 'P' */ - -static const GLubyte ch80data[] = { -0x80,0x80,0x80,0x80,0x80,0xfc,0x82,0x82,0x82,0xfc, -}; - -static const BitmapCharRec ch80 = {7,10,-1,0,9,ch80data}; - -/* char: 0x4f 'O' */ - -static const GLubyte ch79data[] = { -0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch79 = {7,10,-1,0,9,ch79data}; - -/* char: 0x4e 'N' */ - -static const GLubyte ch78data[] = { -0x82,0x82,0x82,0x86,0x8a,0x92,0xa2,0xc2,0x82,0x82, -}; - -static const BitmapCharRec ch78 = {7,10,-1,0,9,ch78data}; - -/* char: 0x4d 'M' */ - -static const GLubyte ch77data[] = { -0x82,0x82,0x82,0x92,0x92,0xaa,0xaa,0xc6,0x82,0x82, -}; - -static const BitmapCharRec ch77 = {7,10,-1,0,9,ch77data}; - -/* char: 0x4c 'L' */ - -static const GLubyte ch76data[] = { -0xfe,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch76 = {7,10,-1,0,9,ch76data}; - -/* char: 0x4b 'K' */ - -static const GLubyte ch75data[] = { -0x82,0x84,0x88,0x90,0xa0,0xe0,0x90,0x88,0x84,0x82, -}; - -static const BitmapCharRec ch75 = {7,10,-1,0,9,ch75data}; - -/* char: 0x4a 'J' */ - -static const GLubyte ch74data[] = { -0x78,0x84,0x4,0x4,0x4,0x4,0x4,0x4,0x4,0x1e, -}; - -static const BitmapCharRec ch74 = {7,10,-1,0,9,ch74data}; - -/* char: 0x49 'I' */ - -static const GLubyte ch73data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xf8, -}; - -static const BitmapCharRec ch73 = {5,10,-2,0,9,ch73data}; - -/* char: 0x48 'H' */ - -static const GLubyte ch72data[] = { -0x82,0x82,0x82,0x82,0x82,0xfe,0x82,0x82,0x82,0x82, -}; - -static const BitmapCharRec ch72 = {7,10,-1,0,9,ch72data}; - -/* char: 0x47 'G' */ - -static const GLubyte ch71data[] = { -0x7c,0x82,0x82,0x82,0x8e,0x80,0x80,0x80,0x82,0x7c, -}; - -static const BitmapCharRec ch71 = {7,10,-1,0,9,ch71data}; - -/* char: 0x46 'F' */ - -static const GLubyte ch70data[] = { -0x40,0x40,0x40,0x40,0x40,0x78,0x40,0x40,0x40,0xfe, -}; - -static const BitmapCharRec ch70 = {7,10,-1,0,9,ch70data}; - -/* char: 0x45 'E' */ - -static const GLubyte ch69data[] = { -0xfe,0x40,0x40,0x40,0x40,0x78,0x40,0x40,0x40,0xfe, -}; - -static const BitmapCharRec ch69 = {7,10,-1,0,9,ch69data}; - -/* char: 0x44 'D' */ - -static const GLubyte ch68data[] = { -0xfc,0x42,0x42,0x42,0x42,0x42,0x42,0x42,0x42,0xfc, -}; - -static const BitmapCharRec ch68 = {7,10,-1,0,9,ch68data}; - -/* char: 0x43 'C' */ - -static const GLubyte ch67data[] = { -0x7c,0x82,0x80,0x80,0x80,0x80,0x80,0x80,0x82,0x7c, -}; - -static const BitmapCharRec ch67 = {7,10,-1,0,9,ch67data}; - -/* char: 0x42 'B' */ - -static const GLubyte ch66data[] = { -0xfc,0x42,0x42,0x42,0x42,0x7c,0x42,0x42,0x42,0xfc, -}; - -static const BitmapCharRec ch66 = {7,10,-1,0,9,ch66data}; - -/* char: 0x41 'A' */ - -static const GLubyte ch65data[] = { -0x82,0x82,0x82,0xfe,0x82,0x82,0x82,0x44,0x28,0x10, -}; - -static const BitmapCharRec ch65 = {7,10,-1,0,9,ch65data}; - -/* char: 0x40 '@' */ - -static const GLubyte ch64data[] = { -0x7c,0x80,0x80,0x9a,0xa6,0xa2,0x9e,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch64 = {7,10,-1,0,9,ch64data}; - -/* char: 0x3f '?' */ - -static const GLubyte ch63data[] = { -0x10,0x0,0x10,0x10,0x8,0x4,0x2,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch63 = {7,10,-1,0,9,ch63data}; - -/* char: 0x3e '>' */ - -static const GLubyte ch62data[] = { -0x80,0x40,0x20,0x10,0x8,0x8,0x10,0x20,0x40,0x80, -}; - -static const BitmapCharRec ch62 = {5,10,-2,0,9,ch62data}; - -/* char: 0x3d '=' */ - -static const GLubyte ch61data[] = { -0xfe,0x0,0x0,0xfe, -}; - -static const BitmapCharRec ch61 = {7,4,-1,-2,9,ch61data}; - -/* char: 0x3c '<' */ - -static const GLubyte ch60data[] = { -0x8,0x10,0x20,0x40,0x80,0x80,0x40,0x20,0x10,0x8, -}; - -static const BitmapCharRec ch60 = {5,10,-2,0,9,ch60data}; - -/* char: 0x3b ';' */ - -static const GLubyte ch59data[] = { -0x80,0x40,0x40,0xc0,0xc0,0x0,0x0,0x0,0xc0,0xc0, -}; - -static const BitmapCharRec ch59 = {2,10,-4,3,9,ch59data}; - -/* char: 0x3a ':' */ - -static const GLubyte ch58data[] = { -0xc0,0xc0,0x0,0x0,0x0,0xc0,0xc0, -}; - -static const BitmapCharRec ch58 = {2,7,-4,0,9,ch58data}; - -/* char: 0x39 '9' */ - -static const GLubyte ch57data[] = { -0x78,0x4,0x2,0x2,0x7a,0x86,0x82,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch57 = {7,10,-1,0,9,ch57data}; - -/* char: 0x38 '8' */ - -static const GLubyte ch56data[] = { -0x38,0x44,0x82,0x82,0x44,0x38,0x44,0x82,0x44,0x38, -}; - -static const BitmapCharRec ch56 = {7,10,-1,0,9,ch56data}; - -/* char: 0x37 '7' */ - -static const GLubyte ch55data[] = { -0x40,0x40,0x20,0x20,0x10,0x8,0x4,0x2,0x2,0xfe, -}; - -static const BitmapCharRec ch55 = {7,10,-1,0,9,ch55data}; - -/* char: 0x36 '6' */ - -static const GLubyte ch54data[] = { -0x7c,0x82,0x82,0x82,0xc2,0xbc,0x80,0x80,0x40,0x3c, -}; - -static const BitmapCharRec ch54 = {7,10,-1,0,9,ch54data}; - -/* char: 0x35 '5' */ - -static const GLubyte ch53data[] = { -0x7c,0x82,0x2,0x2,0x2,0xc2,0xbc,0x80,0x80,0xfe, -}; - -static const BitmapCharRec ch53 = {7,10,-1,0,9,ch53data}; - -/* char: 0x34 '4' */ - -static const GLubyte ch52data[] = { -0x4,0x4,0x4,0xfe,0x84,0x44,0x24,0x14,0xc,0x4, -}; - -static const BitmapCharRec ch52 = {7,10,-1,0,9,ch52data}; - -/* char: 0x33 '3' */ - -static const GLubyte ch51data[] = { -0x7c,0x82,0x2,0x2,0x2,0x1c,0x8,0x4,0x2,0xfe, -}; - -static const BitmapCharRec ch51 = {7,10,-1,0,9,ch51data}; - -/* char: 0x32 '2' */ - -static const GLubyte ch50data[] = { -0xfe,0x80,0x40,0x30,0x8,0x4,0x2,0x82,0x82,0x7c, -}; - -static const BitmapCharRec ch50 = {7,10,-1,0,9,ch50data}; - -/* char: 0x31 '1' */ - -static const GLubyte ch49data[] = { -0xfe,0x10,0x10,0x10,0x10,0x10,0x90,0x50,0x30,0x10, -}; - -static const BitmapCharRec ch49 = {7,10,-1,0,9,ch49data}; - -/* char: 0x30 '0' */ - -static const GLubyte ch48data[] = { -0x38,0x44,0x82,0x82,0x82,0x82,0x82,0x82,0x44,0x38, -}; - -static const BitmapCharRec ch48 = {7,10,-1,0,9,ch48data}; - -/* char: 0x2f '/' */ - -static const GLubyte ch47data[] = { -0x80,0x40,0x40,0x20,0x10,0x10,0x8,0x4,0x4,0x2, -}; - -static const BitmapCharRec ch47 = {7,10,-1,0,9,ch47data}; - -/* char: 0x2e '.' */ - -static const GLubyte ch46data[] = { -0xc0,0xc0, -}; - -static const BitmapCharRec ch46 = {2,2,-4,0,9,ch46data}; - -/* char: 0x2d '-' */ - -static const GLubyte ch45data[] = { -0xfe, -}; - -static const BitmapCharRec ch45 = {7,1,-1,-4,9,ch45data}; - -/* char: 0x2c ',' */ - -static const GLubyte ch44data[] = { -0x80,0x40,0x40,0xc0,0xc0, -}; - -static const BitmapCharRec ch44 = {2,5,-4,3,9,ch44data}; - -/* char: 0x2b '+' */ - -static const GLubyte ch43data[] = { -0x10,0x10,0x10,0xfe,0x10,0x10,0x10, -}; - -static const BitmapCharRec ch43 = {7,7,-1,-1,9,ch43data}; - -/* char: 0x2a '*' */ - -static const GLubyte ch42data[] = { -0x10,0x92,0x54,0x38,0x54,0x92,0x10, -}; - -static const BitmapCharRec ch42 = {7,7,-1,-1,9,ch42data}; - -/* char: 0x29 ')' */ - -static const GLubyte ch41data[] = { -0x80,0x40,0x40,0x20,0x20,0x20,0x20,0x20,0x20,0x40,0x40,0x80, -}; - -static const BitmapCharRec ch41 = {3,12,-3,1,9,ch41data}; - -/* char: 0x28 '(' */ - -static const GLubyte ch40data[] = { -0x20,0x40,0x40,0x80,0x80,0x80,0x80,0x80,0x80,0x40,0x40,0x20, -}; - -static const BitmapCharRec ch40 = {3,12,-3,1,9,ch40data}; - -/* char: 0x27 ''' */ - -static const GLubyte ch39data[] = { -0x80,0x40,0x20,0x30, -}; - -static const BitmapCharRec ch39 = {4,4,-3,-6,9,ch39data}; - -/* char: 0x26 '&' */ - -static const GLubyte ch38data[] = { -0x62,0x94,0x88,0x94,0x62,0x60,0x90,0x90,0x90,0x60, -}; - -static const BitmapCharRec ch38 = {7,10,-1,0,9,ch38data}; - -/* char: 0x25 '%' */ - -static const GLubyte ch37data[] = { -0x84,0x4a,0x4a,0x24,0x10,0x10,0x48,0xa4,0xa4,0x42, -}; - -static const BitmapCharRec ch37 = {7,10,-1,0,9,ch37data}; - -/* char: 0x24 '$' */ - -static const GLubyte ch36data[] = { -0x10,0x7c,0x92,0x12,0x12,0x14,0x38,0x50,0x90,0x92,0x7c,0x10, -}; - -static const BitmapCharRec ch36 = {7,12,-1,1,9,ch36data}; - -/* char: 0x23 '#' */ - -static const GLubyte ch35data[] = { -0x48,0x48,0xfc,0x48,0x48,0xfc,0x48,0x48, -}; - -static const BitmapCharRec ch35 = {6,8,-1,-1,9,ch35data}; - -/* char: 0x22 '"' */ - -static const GLubyte ch34data[] = { -0x90,0x90,0x90, -}; - -static const BitmapCharRec ch34 = {4,3,-3,-7,9,ch34data}; - -/* char: 0x21 '!' */ - -static const GLubyte ch33data[] = { -0x80,0x80,0x0,0x0,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch33 = {1,11,-4,0,9,ch33data}; - -/* char: 0x1f */ - -static const GLubyte ch31data[] = { -0xc0,0xc0, -}; - -static const BitmapCharRec ch31 = {2,2,-4,-2,9,ch31data}; - -/* char: 0x1e */ - -static const GLubyte ch30data[] = { -0x5c,0xa2,0x60,0x20,0x20,0xf8,0x20,0x20,0x22,0x1c, -}; - -static const BitmapCharRec ch30 = {7,10,-1,0,9,ch30data}; - -/* char: 0x1d */ - -static const GLubyte ch29data[] = { -0x80,0x40,0xfe,0x10,0xfe,0x4,0x2, -}; - -static const BitmapCharRec ch29 = {7,7,-1,0,9,ch29data}; - -/* char: 0x1c */ - -static const GLubyte ch28data[] = { -0x44,0x24,0x24,0x24,0x24,0x24,0xfe, -}; - -static const BitmapCharRec ch28 = {7,7,-1,0,9,ch28data}; - -/* char: 0x1b */ - -static const GLubyte ch27data[] = { -0xfe,0x0,0x80,0x40,0x20,0x10,0x8,0x8,0x10,0x20,0x40,0x80, -}; - -static const BitmapCharRec ch27 = {7,12,-1,2,9,ch27data}; - -/* char: 0x1a */ - -static const GLubyte ch26data[] = { -0xfc,0x0,0x4,0x8,0x10,0x20,0x40,0x40,0x20,0x10,0x8,0x4, -}; - -static const BitmapCharRec ch26 = {6,12,-2,2,9,ch26data}; - -/* char: 0x19 */ - -static const GLubyte ch25data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch25 = {1,15,-4,3,9,ch25data}; - -/* char: 0x18 */ - -static const GLubyte ch24data[] = { -0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0xff,0x80, -}; - -static const BitmapCharRec ch24 = {9,7,0,3,9,ch24data}; - -/* char: 0x17 */ - -static const GLubyte ch23data[] = { -0xff,0x80,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0, -0x8,0x0, -}; - -static const BitmapCharRec ch23 = {9,9,0,-3,9,ch23data}; - -/* char: 0x16 */ - -static const GLubyte ch22data[] = { -0x8,0x8,0x8,0x8,0x8,0x8,0xf8,0x8,0x8,0x8,0x8,0x8,0x8,0x8,0x8, -}; - -static const BitmapCharRec ch22 = {5,15,0,3,9,ch22data}; - -/* char: 0x15 */ - -static const GLubyte ch21data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch21 = {5,15,-4,3,9,ch21data}; - -/* char: 0x14 */ - -static const GLubyte ch20data[] = { -0xff,0x80, -}; - -static const BitmapCharRec ch20 = {9,1,0,1,9,ch20data}; - -/* char: 0x13 */ - -static const GLubyte ch19data[] = { -0xff,0x80, -}; - -static const BitmapCharRec ch19 = {9,1,0,-1,9,ch19data}; - -/* char: 0x12 */ - -static const GLubyte ch18data[] = { -0xff,0x80, -}; - -static const BitmapCharRec ch18 = {9,1,0,-3,9,ch18data}; - -/* char: 0x11 */ - -static const GLubyte ch17data[] = { -0xff,0x80, -}; - -static const BitmapCharRec ch17 = {9,1,0,-5,9,ch17data}; - -/* char: 0x10 */ - -static const GLubyte ch16data[] = { -0xff,0x80, -}; - -static const BitmapCharRec ch16 = {9,1,0,-7,9,ch16data}; - -/* char: 0xf */ - -static const GLubyte ch15data[] = { -0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0xff,0x80,0x8,0x0, -0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0, -}; - -static const BitmapCharRec ch15 = {9,15,0,3,9,ch15data}; - -/* char: 0xe */ - -static const GLubyte ch14data[] = { -0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch14 = {5,9,-4,-3,9,ch14data}; - -/* char: 0xd */ - -static const GLubyte ch13data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0xf8, -}; - -static const BitmapCharRec ch13 = {5,7,-4,3,9,ch13data}; - -/* char: 0xc */ - -static const GLubyte ch12data[] = { -0x8,0x8,0x8,0x8,0x8,0x8,0xf8, -}; - -static const BitmapCharRec ch12 = {5,7,0,3,9,ch12data}; - -/* char: 0xb */ - -static const GLubyte ch11data[] = { -0xf8,0x8,0x8,0x8,0x8,0x8,0x8,0x8,0x8, -}; - -static const BitmapCharRec ch11 = {5,9,0,-3,9,ch11data}; - -/* char: 0xa */ - -static const GLubyte ch10data[] = { -0x8,0x8,0x8,0x8,0x3e,0x0,0x20,0x50,0x88,0x88, -}; - -static const BitmapCharRec ch10 = {7,10,-1,2,9,ch10data}; - -/* char: 0x9 */ - -static const GLubyte ch9data[] = { -0x3e,0x20,0x20,0x20,0x20,0x88,0x98,0xa8,0xc8,0x88, -}; - -static const BitmapCharRec ch9 = {7,10,-1,2,9,ch9data}; - -/* char: 0x8 */ - -static const GLubyte ch8data[] = { -0xfe,0x10,0x10,0xfe,0x10,0x10, -}; - -static const BitmapCharRec ch8 = {7,6,-1,0,9,ch8data}; - -/* char: 0x7 */ - -static const GLubyte ch7data[] = { -0x70,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch7 = {5,4,-2,-6,9,ch7data}; - -/* char: 0x6 */ - -static const GLubyte ch6data[] = { -0x20,0x20,0x3c,0x20,0x3e,0x0,0xf8,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch6 = {7,10,-1,2,9,ch6data}; - -/* char: 0x5 */ - -static const GLubyte ch5data[] = { -0x22,0x22,0x3c,0x22,0x3c,0x0,0x78,0x80,0x80,0x78, -}; - -static const BitmapCharRec ch5 = {7,10,-1,2,9,ch5data}; - -/* char: 0x4 */ - -static const GLubyte ch4data[] = { -0x10,0x10,0x1c,0x10,0x1e,0x80,0x80,0xe0,0x80,0xf0, -}; - -static const BitmapCharRec ch4 = {7,10,-1,2,9,ch4data}; - -/* char: 0x3 */ - -static const GLubyte ch3data[] = { -0x8,0x8,0x8,0x3e,0x0,0x88,0x88,0xf8,0x88,0x88, -}; - -static const BitmapCharRec ch3 = {7,10,-1,2,9,ch3data}; - -/* char: 0x2 */ - -static const GLubyte ch2data[] = { -0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa, -}; - -static const BitmapCharRec ch2 = {8,14,0,3,9,ch2data}; - -/* char: 0x1 */ - -static const GLubyte ch1data[] = { -0x10,0x38,0x7c,0xfe,0x7c,0x38,0x10, -}; - -static const BitmapCharRec ch1 = {7,7,-1,0,9,ch1data}; - -static const BitmapCharRec * const chars[] = { -&ch0, -&ch1, -&ch2, -&ch3, -&ch4, -&ch5, -&ch6, -&ch7, -&ch8, -&ch9, -&ch10, -&ch11, -&ch12, -&ch13, -&ch14, -&ch15, -&ch16, -&ch17, -&ch18, -&ch19, -&ch20, -&ch21, -&ch22, -&ch23, -&ch24, -&ch25, -&ch26, -&ch27, -&ch28, -&ch29, -&ch30, -&ch31, -&ch32, -&ch33, -&ch34, -&ch35, -&ch36, -&ch37, -&ch38, -&ch39, -&ch40, -&ch41, -&ch42, -&ch43, -&ch44, -&ch45, -&ch46, -&ch47, -&ch48, -&ch49, -&ch50, -&ch51, -&ch52, -&ch53, -&ch54, -&ch55, -&ch56, -&ch57, -&ch58, -&ch59, -&ch60, -&ch61, -&ch62, -&ch63, -&ch64, -&ch65, -&ch66, -&ch67, -&ch68, -&ch69, -&ch70, -&ch71, -&ch72, -&ch73, -&ch74, -&ch75, -&ch76, -&ch77, -&ch78, -&ch79, -&ch80, -&ch81, -&ch82, -&ch83, -&ch84, -&ch85, -&ch86, -&ch87, -&ch88, -&ch89, -&ch90, -&ch91, -&ch92, -&ch93, -&ch94, -&ch95, -&ch96, -&ch97, -&ch98, -&ch99, -&ch100, -&ch101, -&ch102, -&ch103, -&ch104, -&ch105, -&ch106, -&ch107, -&ch108, -&ch109, -&ch110, -&ch111, -&ch112, -&ch113, -&ch114, -&ch115, -&ch116, -&ch117, -&ch118, -&ch119, -&ch120, -&ch121, -&ch122, -&ch123, -&ch124, -&ch125, -&ch126, -&ch127, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -&ch160, -&ch161, -&ch162, -&ch163, -&ch164, -&ch165, -&ch166, -&ch167, -&ch168, -&ch169, -&ch170, -&ch171, -&ch172, -&ch173, -&ch174, -&ch175, -&ch176, -&ch177, -&ch178, -&ch179, -&ch180, -&ch181, -&ch182, -&ch183, -&ch184, -&ch185, -&ch186, -&ch187, -&ch188, -&ch189, -&ch190, -&ch191, -&ch192, -&ch193, -&ch194, -&ch195, -&ch196, -&ch197, -&ch198, -&ch199, -&ch200, -&ch201, -&ch202, -&ch203, -&ch204, -&ch205, -&ch206, -&ch207, -&ch208, -&ch209, -&ch210, -&ch211, -&ch212, -&ch213, -&ch214, -&ch215, -&ch216, -&ch217, -&ch218, -&ch219, -&ch220, -&ch221, -&ch222, -&ch223, -&ch224, -&ch225, -&ch226, -&ch227, -&ch228, -&ch229, -&ch230, -&ch231, -&ch232, -&ch233, -&ch234, -&ch235, -&ch236, -&ch237, -&ch238, -&ch239, -&ch240, -&ch241, -&ch242, -&ch243, -&ch244, -&ch245, -&ch246, -&ch247, -&ch248, -&ch249, -&ch250, -&ch251, -&ch252, -&ch253, -&ch254, -&ch255, -}; -#if !defined(__IBMCPP__) -const -#endif - BitmapFontRec glutBitmap9By15 = { -"-misc-fixed-medium-r-normal--15-140-75-75-C-90-iso8859-1", -256, -0, -chars -}; - + +/* GENERATED FILE -- DO NOT MODIFY */ + +#define glutBitmap9By15 XXX +#include "glutbitmap.h" +#undef glutBitmap9By15 + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch0data[] = { 0x0 }; +static const BitmapCharRec ch0 = {1,1,0,0,9,ch0data}; +#else +static const BitmapCharRec ch0 = {0,0,0,0,9,0}; +#endif + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch32data[] = { 0x0 }; +static const BitmapCharRec ch32 = {1,1,0,0,9,ch32data}; +#else +static const BitmapCharRec ch32 = {0,0,0,0,9,0}; +#endif + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch127data[] = { 0x0 }; +static const BitmapCharRec ch127 = {1,1,0,0,9,ch127data}; +#else +static const BitmapCharRec ch127 = {0,0,0,0,9,0}; +#endif + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch160data[] = { 0x0 }; +static const BitmapCharRec ch160 = {1,1,0,0,9,ch160data}; +#else +static const BitmapCharRec ch160 = {0,0,0,0,9,0}; +#endif + +/* char: 0xff */ + +static const GLubyte ch255data[] = { +0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch255 = {6,14,-1,3,9,ch255data}; + +/* char: 0xfe */ + +static const GLubyte ch254data[] = { +0x80,0x80,0x80,0xbc,0xc2,0x82,0x82,0x82,0xc2,0xbc,0x80,0x80, +}; + +static const BitmapCharRec ch254 = {7,12,-1,3,9,ch254data}; + +/* char: 0xfd */ + +static const GLubyte ch253data[] = { +0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch253 = {6,14,-1,3,9,ch253data}; + +/* char: 0xfc */ + +static const GLubyte ch252data[] = { +0x7a,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch252 = {7,11,-1,0,9,ch252data}; + +/* char: 0xfb */ + +static const GLubyte ch251data[] = { +0x7a,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x44,0x38, +}; + +static const BitmapCharRec ch251 = {7,11,-1,0,9,ch251data}; + +/* char: 0xfa */ + +static const GLubyte ch250data[] = { +0x7a,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch250 = {7,11,-1,0,9,ch250data}; + +/* char: 0xf9 */ + +static const GLubyte ch249data[] = { +0x7a,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x0,0x18,0x20, +}; + +static const BitmapCharRec ch249 = {7,11,-1,0,9,ch249data}; + +/* char: 0xf8 */ + +static const GLubyte ch248data[] = { +0x80,0x7c,0xa2,0xa2,0x92,0x8a,0x8a,0x7c,0x2, +}; + +static const BitmapCharRec ch248 = {7,9,-1,1,9,ch248data}; + +/* char: 0xf7 */ + +static const GLubyte ch247data[] = { +0x10,0x38,0x10,0x0,0xfe,0x0,0x10,0x38,0x10, +}; + +static const BitmapCharRec ch247 = {7,9,-1,0,9,ch247data}; + +/* char: 0xf6 */ + +static const GLubyte ch246data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch246 = {7,11,-1,0,9,ch246data}; + +/* char: 0xf5 */ + +static const GLubyte ch245data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch245 = {7,11,-1,0,9,ch245data}; + +/* char: 0xf4 */ + +static const GLubyte ch244data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x44,0x38, +}; + +static const BitmapCharRec ch244 = {7,11,-1,0,9,ch244data}; + +/* char: 0xf3 */ + +static const GLubyte ch243data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch243 = {7,11,-1,0,9,ch243data}; + +/* char: 0xf2 */ + +static const GLubyte ch242data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x0,0x18,0x20, +}; + +static const BitmapCharRec ch242 = {7,11,-1,0,9,ch242data}; + +/* char: 0xf1 */ + +static const GLubyte ch241data[] = { +0x82,0x82,0x82,0x82,0x82,0xc2,0xbc,0x0,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch241 = {7,11,-1,0,9,ch241data}; + +/* char: 0xf0 */ + +static const GLubyte ch240data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c,0x8,0x50,0x30,0x48, +}; + +static const BitmapCharRec ch240 = {7,11,-1,0,9,ch240data}; + +/* char: 0xef */ + +static const GLubyte ch239data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x50,0x50, +}; + +static const BitmapCharRec ch239 = {5,11,-2,0,9,ch239data}; + +/* char: 0xee */ + +static const GLubyte ch238data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x90,0x60, +}; + +static const BitmapCharRec ch238 = {5,11,-2,0,9,ch238data}; + +/* char: 0xed */ + +static const GLubyte ch237data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x60,0x10, +}; + +static const BitmapCharRec ch237 = {5,11,-2,0,9,ch237data}; + +/* char: 0xec */ + +static const GLubyte ch236data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x30,0x40, +}; + +static const BitmapCharRec ch236 = {5,11,-2,0,9,ch236data}; + +/* char: 0xeb */ + +static const GLubyte ch235data[] = { +0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c,0x0,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch235 = {7,11,-1,0,9,ch235data}; + +/* char: 0xea */ + +static const GLubyte ch234data[] = { +0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c,0x0,0x0,0x44,0x38, +}; + +static const BitmapCharRec ch234 = {7,11,-1,0,9,ch234data}; + +/* char: 0xe9 */ + +static const GLubyte ch233data[] = { +0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c,0x0,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch233 = {7,11,-1,0,9,ch233data}; + +/* char: 0xe8 */ + +static const GLubyte ch232data[] = { +0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c,0x0,0x0,0x18,0x20, +}; + +static const BitmapCharRec ch232 = {7,11,-1,0,9,ch232data}; + +/* char: 0xe7 */ + +static const GLubyte ch231data[] = { +0x30,0x48,0x18,0x7c,0x82,0x80,0x80,0x80,0x82,0x7c, +}; + +static const BitmapCharRec ch231 = {7,10,-1,3,9,ch231data}; + +/* char: 0xe6 */ + +static const GLubyte ch230data[] = { +0x6e,0x92,0x90,0x7c,0x12,0x92,0x6c, +}; + +static const BitmapCharRec ch230 = {7,7,-1,0,9,ch230data}; + +/* char: 0xe5 */ + +static const GLubyte ch229data[] = { +0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x18,0x24,0x18, +}; + +static const BitmapCharRec ch229 = {7,11,-1,0,9,ch229data}; + +/* char: 0xe4 */ + +static const GLubyte ch228data[] = { +0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch228 = {7,11,-1,0,9,ch228data}; + +/* char: 0xe3 */ + +static const GLubyte ch227data[] = { +0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch227 = {7,11,-1,0,9,ch227data}; + +/* char: 0xe2 */ + +static const GLubyte ch226data[] = { +0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x44,0x38, +}; + +static const BitmapCharRec ch226 = {7,11,-1,0,9,ch226data}; + +/* char: 0xe1 */ + +static const GLubyte ch225data[] = { +0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch225 = {7,11,-1,0,9,ch225data}; + +/* char: 0xe0 */ + +static const GLubyte ch224data[] = { +0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c,0x0,0x0,0x18,0x20, +}; + +static const BitmapCharRec ch224 = {7,11,-1,0,9,ch224data}; + +/* char: 0xdf */ + +static const GLubyte ch223data[] = { +0x80,0xbc,0xc2,0x82,0x82,0xfc,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch223 = {7,9,-1,1,9,ch223data}; + +/* char: 0xde */ + +static const GLubyte ch222data[] = { +0x80,0x80,0x80,0xfc,0x82,0x82,0x82,0xfc,0x80,0x80, +}; + +static const BitmapCharRec ch222 = {7,10,-1,0,9,ch222data}; + +/* char: 0xdd */ + +static const GLubyte ch221data[] = { +0x10,0x10,0x10,0x10,0x28,0x44,0x82,0x82,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch221 = {7,11,-1,0,9,ch221data}; + +/* char: 0xdc */ + +static const GLubyte ch220data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch220 = {7,11,-1,0,9,ch220data}; + +/* char: 0xdb */ + +static const GLubyte ch219data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x0,0x44,0x38, +}; + +static const BitmapCharRec ch219 = {7,11,-1,0,9,ch219data}; + +/* char: 0xda */ + +static const GLubyte ch218data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch218 = {7,11,-1,0,9,ch218data}; + +/* char: 0xd9 */ + +static const GLubyte ch217data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x0,0x18,0x20, +}; + +static const BitmapCharRec ch217 = {7,11,-1,0,9,ch217data}; + +/* char: 0xd8 */ + +static const GLubyte ch216data[] = { +0x80,0x7c,0xc2,0xa2,0xa2,0x92,0x92,0x8a,0x8a,0x86,0x7c,0x2, +}; + +static const BitmapCharRec ch216 = {7,12,-1,1,9,ch216data}; + +/* char: 0xd7 */ + +static const GLubyte ch215data[] = { +0x82,0x44,0x28,0x10,0x28,0x44,0x82, +}; + +static const BitmapCharRec ch215 = {7,7,-1,-1,9,ch215data}; + +/* char: 0xd6 */ + +static const GLubyte ch214data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch214 = {7,11,-1,0,9,ch214data}; + +/* char: 0xd5 */ + +static const GLubyte ch213data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch213 = {7,11,-1,0,9,ch213data}; + +/* char: 0xd4 */ + +static const GLubyte ch212data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x44,0x38, +}; + +static const BitmapCharRec ch212 = {7,11,-1,0,9,ch212data}; + +/* char: 0xd3 */ + +static const GLubyte ch211data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch211 = {7,11,-1,0,9,ch211data}; + +/* char: 0xd2 */ + +static const GLubyte ch210data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x7c,0x0,0x18,0x20, +}; + +static const BitmapCharRec ch210 = {7,11,-1,0,9,ch210data}; + +/* char: 0xd1 */ + +static const GLubyte ch209data[] = { +0x82,0x86,0x8a,0x92,0x92,0xa2,0xc2,0x82,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch209 = {7,11,-1,0,9,ch209data}; + +/* char: 0xd0 */ + +static const GLubyte ch208data[] = { +0xfc,0x42,0x42,0x42,0x42,0xf2,0x42,0x42,0x42,0xfc, +}; + +static const BitmapCharRec ch208 = {7,10,-1,0,9,ch208data}; + +/* char: 0xcf */ + +static const GLubyte ch207data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x50,0x50, +}; + +static const BitmapCharRec ch207 = {5,11,-2,0,9,ch207data}; + +/* char: 0xce */ + +static const GLubyte ch206data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x88,0x70, +}; + +static const BitmapCharRec ch206 = {5,11,-2,0,9,ch206data}; + +/* char: 0xcd */ + +static const GLubyte ch205data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x60,0x10, +}; + +static const BitmapCharRec ch205 = {5,11,-2,0,9,ch205data}; + +/* char: 0xcc */ + +static const GLubyte ch204data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xf8,0x0,0x30,0x40, +}; + +static const BitmapCharRec ch204 = {5,11,-2,0,9,ch204data}; + +/* char: 0xcb */ + +static const GLubyte ch203data[] = { +0xfe,0x40,0x40,0x40,0x78,0x40,0x40,0xfe,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch203 = {7,11,-1,0,9,ch203data}; + +/* char: 0xca */ + +static const GLubyte ch202data[] = { +0xfe,0x40,0x40,0x40,0x78,0x40,0x40,0xfe,0x0,0x44,0x38, +}; + +static const BitmapCharRec ch202 = {7,11,-1,0,9,ch202data}; + +/* char: 0xc9 */ + +static const GLubyte ch201data[] = { +0xfe,0x40,0x40,0x40,0x78,0x40,0x40,0xfe,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch201 = {7,11,-1,0,9,ch201data}; + +/* char: 0xc8 */ + +static const GLubyte ch200data[] = { +0xfe,0x40,0x40,0x40,0x78,0x40,0x40,0xfe,0x0,0x18,0x20, +}; + +static const BitmapCharRec ch200 = {7,11,-1,0,9,ch200data}; + +/* char: 0xc7 */ + +static const GLubyte ch199data[] = { +0x30,0x48,0x18,0x7c,0x82,0x80,0x80,0x80,0x80,0x80,0x80,0x82,0x7c, +}; + +static const BitmapCharRec ch199 = {7,13,-1,3,9,ch199data}; + +/* char: 0xc6 */ + +static const GLubyte ch198data[] = { +0x9e,0x90,0x90,0x90,0xfc,0x90,0x90,0x90,0x90,0x6e, +}; + +static const BitmapCharRec ch198 = {7,10,-1,0,9,ch198data}; + +/* char: 0xc5 */ + +static const GLubyte ch197data[] = { +0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x10,0x28,0x10, +}; + +static const BitmapCharRec ch197 = {7,11,-1,0,9,ch197data}; + +/* char: 0xc4 */ + +static const GLubyte ch196data[] = { +0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x28,0x28, +}; + +static const BitmapCharRec ch196 = {7,11,-1,0,9,ch196data}; + +/* char: 0xc3 */ + +static const GLubyte ch195data[] = { +0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch195 = {7,11,-1,0,9,ch195data}; + +/* char: 0xc2 */ + +static const GLubyte ch194data[] = { +0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x44,0x38, +}; + +static const BitmapCharRec ch194 = {7,11,-1,0,9,ch194data}; + +/* char: 0xc1 */ + +static const GLubyte ch193data[] = { +0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x30,0x8, +}; + +static const BitmapCharRec ch193 = {7,11,-1,0,9,ch193data}; + +/* char: 0xc0 */ + +static const GLubyte ch192data[] = { +0x82,0x82,0x82,0xfe,0x82,0x82,0x44,0x38,0x0,0x18,0x20, +}; + +static const BitmapCharRec ch192 = {7,11,-1,0,9,ch192data}; + +/* char: 0xbf */ + +static const GLubyte ch191data[] = { +0x7c,0x82,0x82,0x80,0x40,0x20,0x10,0x10,0x0,0x10, +}; + +static const BitmapCharRec ch191 = {7,10,-1,0,9,ch191data}; + +/* char: 0xbe */ + +static const GLubyte ch190data[] = { +0x6,0x1a,0x12,0xa,0x66,0x92,0x10,0x20,0x90,0x60, +}; + +static const BitmapCharRec ch190 = {7,10,-1,0,9,ch190data}; + +/* char: 0xbd */ + +static const GLubyte ch189data[] = { +0x1e,0x10,0xc,0x2,0xf2,0x4c,0x40,0x40,0xc0,0x40, +}; + +static const BitmapCharRec ch189 = {7,10,-1,0,9,ch189data}; + +/* char: 0xbc */ + +static const GLubyte ch188data[] = { +0x6,0x1a,0x12,0xa,0xe6,0x42,0x40,0x40,0xc0,0x40, +}; + +static const BitmapCharRec ch188 = {7,10,-1,0,9,ch188data}; + +/* char: 0xbb */ + +static const GLubyte ch187data[] = { +0x90,0x48,0x24,0x12,0x12,0x24,0x48,0x90, +}; + +static const BitmapCharRec ch187 = {7,8,-1,-1,9,ch187data}; + +/* char: 0xba */ + +static const GLubyte ch186data[] = { +0xf8,0x0,0x70,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch186 = {5,6,-1,-5,9,ch186data}; + +/* char: 0xb9 */ + +static const GLubyte ch185data[] = { +0xe0,0x40,0x40,0x40,0xc0,0x40, +}; + +static const BitmapCharRec ch185 = {3,6,-1,-4,9,ch185data}; + +/* char: 0xb8 */ + +static const GLubyte ch184data[] = { +0x60,0x90,0x30, +}; + +static const BitmapCharRec ch184 = {4,3,-2,3,9,ch184data}; + +/* char: 0xb7 */ + +static const GLubyte ch183data[] = { +0xc0,0xc0, +}; + +static const BitmapCharRec ch183 = {2,2,-4,-4,9,ch183data}; + +/* char: 0xb6 */ + +static const GLubyte ch182data[] = { +0xa,0xa,0xa,0xa,0xa,0x7a,0x8a,0x8a,0x8a,0x7e, +}; + +static const BitmapCharRec ch182 = {7,10,-1,0,9,ch182data}; + +/* char: 0xb5 */ + +static const GLubyte ch181data[] = { +0x80,0x80,0xba,0xc6,0x82,0x82,0x82,0x82,0x82, +}; + +static const BitmapCharRec ch181 = {7,9,-1,2,9,ch181data}; + +/* char: 0xb4 */ + +static const GLubyte ch180data[] = { +0xc0,0x20, +}; + +static const BitmapCharRec ch180 = {3,2,-3,-9,9,ch180data}; + +/* char: 0xb3 */ + +static const GLubyte ch179data[] = { +0x60,0x90,0x10,0x20,0x90,0x60, +}; + +static const BitmapCharRec ch179 = {4,6,-1,-4,9,ch179data}; + +/* char: 0xb2 */ + +static const GLubyte ch178data[] = { +0xf0,0x80,0x60,0x10,0x90,0x60, +}; + +static const BitmapCharRec ch178 = {4,6,-1,-4,9,ch178data}; + +/* char: 0xb1 */ + +static const GLubyte ch177data[] = { +0xfe,0x0,0x10,0x10,0x10,0xfe,0x10,0x10,0x10, +}; + +static const BitmapCharRec ch177 = {7,9,-1,-1,9,ch177data}; + +/* char: 0xb0 */ + +static const GLubyte ch176data[] = { +0x60,0x90,0x90,0x60, +}; + +static const BitmapCharRec ch176 = {4,4,-3,-6,9,ch176data}; + +/* char: 0xaf */ + +static const GLubyte ch175data[] = { +0xfc, +}; + +static const BitmapCharRec ch175 = {6,1,-1,-9,9,ch175data}; + +/* char: 0xae */ + +static const GLubyte ch174data[] = { +0x3c,0x42,0xa5,0xa9,0xbd,0xa5,0xb9,0x42,0x3c, +}; + +static const BitmapCharRec ch174 = {8,9,0,-1,9,ch174data}; + +/* char: 0xad */ + +static const GLubyte ch173data[] = { +0xfc, +}; + +static const BitmapCharRec ch173 = {6,1,-1,-4,9,ch173data}; + +/* char: 0xac */ + +static const GLubyte ch172data[] = { +0x4,0x4,0x4,0xfc, +}; + +static const BitmapCharRec ch172 = {6,4,-1,-2,9,ch172data}; + +/* char: 0xab */ + +static const GLubyte ch171data[] = { +0x12,0x24,0x48,0x90,0x90,0x48,0x24,0x12, +}; + +static const BitmapCharRec ch171 = {7,8,-1,-1,9,ch171data}; + +/* char: 0xaa */ + +static const GLubyte ch170data[] = { +0xf8,0x0,0x78,0x90,0x70,0x90,0x60, +}; + +static const BitmapCharRec ch170 = {5,7,-3,-3,9,ch170data}; + +/* char: 0xa9 */ + +static const GLubyte ch169data[] = { +0x3c,0x42,0x99,0xa5,0xa1,0xa5,0x99,0x42,0x3c, +}; + +static const BitmapCharRec ch169 = {8,9,0,-1,9,ch169data}; + +/* char: 0xa8 */ + +static const GLubyte ch168data[] = { +0xa0,0xa0, +}; + +static const BitmapCharRec ch168 = {3,2,-3,-9,9,ch168data}; + +/* char: 0xa7 */ + +static const GLubyte ch167data[] = { +0x70,0x88,0x8,0x70,0x88,0x88,0x88,0x70,0x80,0x88,0x70, +}; + +static const BitmapCharRec ch167 = {5,11,-2,1,9,ch167data}; + +/* char: 0xa6 */ + +static const GLubyte ch166data[] = { +0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch166 = {1,11,-4,1,9,ch166data}; + +/* char: 0xa5 */ + +static const GLubyte ch165data[] = { +0x10,0x10,0x10,0x7c,0x10,0x7c,0x28,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch165 = {7,10,-1,0,9,ch165data}; + +/* char: 0xa4 */ + +static const GLubyte ch164data[] = { +0x82,0x7c,0x44,0x44,0x7c,0x82, +}; + +static const BitmapCharRec ch164 = {7,6,-1,-3,9,ch164data}; + +/* char: 0xa3 */ + +static const GLubyte ch163data[] = { +0x5c,0xa2,0x60,0x20,0x20,0xf8,0x20,0x20,0x22,0x1c, +}; + +static const BitmapCharRec ch163 = {7,10,-1,0,9,ch163data}; + +/* char: 0xa2 */ + +static const GLubyte ch162data[] = { +0x40,0x78,0xa4,0xa0,0x90,0x94,0x78,0x8, +}; + +static const BitmapCharRec ch162 = {6,8,-1,0,9,ch162data}; + +/* char: 0xa1 */ + +static const GLubyte ch161data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x0,0x80,0x80, +}; + +static const BitmapCharRec ch161 = {1,11,-4,0,9,ch161data}; + +/* char: 0x7e '~' */ + +static const GLubyte ch126data[] = { +0x8c,0x92,0x62, +}; + +static const BitmapCharRec ch126 = {7,3,-1,-7,9,ch126data}; + +/* char: 0x7d '}' */ + +static const GLubyte ch125data[] = { +0xe0,0x10,0x10,0x10,0x20,0x18,0x18,0x20,0x10,0x10,0x10,0xe0, +}; + +static const BitmapCharRec ch125 = {5,12,-1,1,9,ch125data}; + +/* char: 0x7c '|' */ + +static const GLubyte ch124data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch124 = {1,12,-4,1,9,ch124data}; + +/* char: 0x7b '{' */ + +static const GLubyte ch123data[] = { +0x38,0x40,0x40,0x40,0x20,0xc0,0xc0,0x20,0x40,0x40,0x40,0x38, +}; + +static const BitmapCharRec ch123 = {5,12,-3,1,9,ch123data}; + +/* char: 0x7a 'z' */ + +static const GLubyte ch122data[] = { +0xfe,0x40,0x20,0x10,0x8,0x4,0xfe, +}; + +static const BitmapCharRec ch122 = {7,7,-1,0,9,ch122data}; + +/* char: 0x79 'y' */ + +static const GLubyte ch121data[] = { +0x78,0x84,0x4,0x74,0x8c,0x84,0x84,0x84,0x84,0x84, +}; + +static const BitmapCharRec ch121 = {6,10,-1,3,9,ch121data}; + +/* char: 0x78 'x' */ + +static const GLubyte ch120data[] = { +0x82,0x44,0x28,0x10,0x28,0x44,0x82, +}; + +static const BitmapCharRec ch120 = {7,7,-1,0,9,ch120data}; + +/* char: 0x77 'w' */ + +static const GLubyte ch119data[] = { +0x44,0xaa,0x92,0x92,0x92,0x82,0x82, +}; + +static const BitmapCharRec ch119 = {7,7,-1,0,9,ch119data}; + +/* char: 0x76 'v' */ + +static const GLubyte ch118data[] = { +0x10,0x28,0x28,0x44,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch118 = {7,7,-1,0,9,ch118data}; + +/* char: 0x75 'u' */ + +static const GLubyte ch117data[] = { +0x7a,0x84,0x84,0x84,0x84,0x84,0x84, +}; + +static const BitmapCharRec ch117 = {7,7,-1,0,9,ch117data}; + +/* char: 0x74 't' */ + +static const GLubyte ch116data[] = { +0x1c,0x22,0x20,0x20,0x20,0x20,0xfc,0x20,0x20, +}; + +static const BitmapCharRec ch116 = {7,9,-1,0,9,ch116data}; + +/* char: 0x73 's' */ + +static const GLubyte ch115data[] = { +0x7c,0x82,0x2,0x7c,0x80,0x82,0x7c, +}; + +static const BitmapCharRec ch115 = {7,7,-1,0,9,ch115data}; + +/* char: 0x72 'r' */ + +static const GLubyte ch114data[] = { +0x40,0x40,0x40,0x40,0x42,0x62,0x9c, +}; + +static const BitmapCharRec ch114 = {7,7,-1,0,9,ch114data}; + +/* char: 0x71 'q' */ + +static const GLubyte ch113data[] = { +0x2,0x2,0x2,0x7a,0x86,0x82,0x82,0x82,0x86,0x7a, +}; + +static const BitmapCharRec ch113 = {7,10,-1,3,9,ch113data}; + +/* char: 0x70 'p' */ + +static const GLubyte ch112data[] = { +0x80,0x80,0x80,0xbc,0xc2,0x82,0x82,0x82,0xc2,0xbc, +}; + +static const BitmapCharRec ch112 = {7,10,-1,3,9,ch112data}; + +/* char: 0x6f 'o' */ + +static const GLubyte ch111data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch111 = {7,7,-1,0,9,ch111data}; + +/* char: 0x6e 'n' */ + +static const GLubyte ch110data[] = { +0x82,0x82,0x82,0x82,0x82,0xc2,0xbc, +}; + +static const BitmapCharRec ch110 = {7,7,-1,0,9,ch110data}; + +/* char: 0x6d 'm' */ + +static const GLubyte ch109data[] = { +0x82,0x92,0x92,0x92,0x92,0x92,0xec, +}; + +static const BitmapCharRec ch109 = {7,7,-1,0,9,ch109data}; + +/* char: 0x6c 'l' */ + +static const GLubyte ch108data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xe0, +}; + +static const BitmapCharRec ch108 = {5,10,-2,0,9,ch108data}; + +/* char: 0x6b 'k' */ + +static const GLubyte ch107data[] = { +0x82,0x8c,0xb0,0xc0,0xb0,0x8c,0x82,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch107 = {7,10,-1,0,9,ch107data}; + +/* char: 0x6a 'j' */ + +static const GLubyte ch106data[] = { +0x78,0x84,0x84,0x84,0x4,0x4,0x4,0x4,0x4,0x1c,0x0,0x0,0xc, +}; + +static const BitmapCharRec ch106 = {6,13,-1,3,9,ch106data}; + +/* char: 0x69 'i' */ + +static const GLubyte ch105data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0xe0,0x0,0x0,0x60, +}; + +static const BitmapCharRec ch105 = {5,10,-2,0,9,ch105data}; + +/* char: 0x68 'h' */ + +static const GLubyte ch104data[] = { +0x82,0x82,0x82,0x82,0x82,0xc2,0xbc,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch104 = {7,10,-1,0,9,ch104data}; + +/* char: 0x67 'g' */ + +static const GLubyte ch103data[] = { +0x7c,0x82,0x82,0x7c,0x80,0x78,0x84,0x84,0x84,0x7a, +}; + +static const BitmapCharRec ch103 = {7,10,-1,3,9,ch103data}; + +/* char: 0x66 'f' */ + +static const GLubyte ch102data[] = { +0x20,0x20,0x20,0x20,0xf8,0x20,0x20,0x22,0x22,0x1c, +}; + +static const BitmapCharRec ch102 = {7,10,-1,0,9,ch102data}; + +/* char: 0x65 'e' */ + +static const GLubyte ch101data[] = { +0x7c,0x80,0x80,0xfe,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch101 = {7,7,-1,0,9,ch101data}; + +/* char: 0x64 'd' */ + +static const GLubyte ch100data[] = { +0x7a,0x86,0x82,0x82,0x82,0x86,0x7a,0x2,0x2,0x2, +}; + +static const BitmapCharRec ch100 = {7,10,-1,0,9,ch100data}; + +/* char: 0x63 'c' */ + +static const GLubyte ch99data[] = { +0x7c,0x82,0x80,0x80,0x80,0x82,0x7c, +}; + +static const BitmapCharRec ch99 = {7,7,-1,0,9,ch99data}; + +/* char: 0x62 'b' */ + +static const GLubyte ch98data[] = { +0xbc,0xc2,0x82,0x82,0x82,0xc2,0xbc,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch98 = {7,10,-1,0,9,ch98data}; + +/* char: 0x61 'a' */ + +static const GLubyte ch97data[] = { +0x7a,0x86,0x82,0x7e,0x2,0x2,0x7c, +}; + +static const BitmapCharRec ch97 = {7,7,-1,0,9,ch97data}; + +/* char: 0x60 '`' */ + +static const GLubyte ch96data[] = { +0x10,0x20,0x40,0xc0, +}; + +static const BitmapCharRec ch96 = {4,4,-3,-6,9,ch96data}; + +/* char: 0x5f '_' */ + +static const GLubyte ch95data[] = { +0xff, +}; + +static const BitmapCharRec ch95 = {8,1,0,1,9,ch95data}; + +/* char: 0x5e '^' */ + +static const GLubyte ch94data[] = { +0x82,0x44,0x28,0x10, +}; + +static const BitmapCharRec ch94 = {7,4,-1,-6,9,ch94data}; + +/* char: 0x5d ']' */ + +static const GLubyte ch93data[] = { +0xf0,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xf0, +}; + +static const BitmapCharRec ch93 = {4,12,-2,1,9,ch93data}; + +/* char: 0x5c '\' */ + +static const GLubyte ch92data[] = { +0x2,0x4,0x4,0x8,0x10,0x10,0x20,0x40,0x40,0x80, +}; + +static const BitmapCharRec ch92 = {7,10,-1,0,9,ch92data}; + +/* char: 0x5b '[' */ + +static const GLubyte ch91data[] = { +0xf0,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0xf0, +}; + +static const BitmapCharRec ch91 = {4,12,-3,1,9,ch91data}; + +/* char: 0x5a 'Z' */ + +static const GLubyte ch90data[] = { +0xfe,0x80,0x80,0x40,0x20,0x10,0x8,0x4,0x2,0xfe, +}; + +static const BitmapCharRec ch90 = {7,10,-1,0,9,ch90data}; + +/* char: 0x59 'Y' */ + +static const GLubyte ch89data[] = { +0x10,0x10,0x10,0x10,0x10,0x10,0x28,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch89 = {7,10,-1,0,9,ch89data}; + +/* char: 0x58 'X' */ + +static const GLubyte ch88data[] = { +0x82,0x82,0x44,0x28,0x10,0x10,0x28,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch88 = {7,10,-1,0,9,ch88data}; + +/* char: 0x57 'W' */ + +static const GLubyte ch87data[] = { +0x44,0xaa,0x92,0x92,0x92,0x92,0x82,0x82,0x82,0x82, +}; + +static const BitmapCharRec ch87 = {7,10,-1,0,9,ch87data}; + +/* char: 0x56 'V' */ + +static const GLubyte ch86data[] = { +0x10,0x28,0x28,0x28,0x44,0x44,0x44,0x82,0x82,0x82, +}; + +static const BitmapCharRec ch86 = {7,10,-1,0,9,ch86data}; + +/* char: 0x55 'U' */ + +static const GLubyte ch85data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x82, +}; + +static const BitmapCharRec ch85 = {7,10,-1,0,9,ch85data}; + +/* char: 0x54 'T' */ + +static const GLubyte ch84data[] = { +0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xfe, +}; + +static const BitmapCharRec ch84 = {7,10,-1,0,9,ch84data}; + +/* char: 0x53 'S' */ + +static const GLubyte ch83data[] = { +0x7c,0x82,0x82,0x2,0xc,0x70,0x80,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch83 = {7,10,-1,0,9,ch83data}; + +/* char: 0x52 'R' */ + +static const GLubyte ch82data[] = { +0x82,0x82,0x84,0x88,0x90,0xfc,0x82,0x82,0x82,0xfc, +}; + +static const BitmapCharRec ch82 = {7,10,-1,0,9,ch82data}; + +/* char: 0x51 'Q' */ + +static const GLubyte ch81data[] = { +0x6,0x8,0x7c,0x92,0xa2,0x82,0x82,0x82,0x82,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch81 = {7,12,-1,2,9,ch81data}; + +/* char: 0x50 'P' */ + +static const GLubyte ch80data[] = { +0x80,0x80,0x80,0x80,0x80,0xfc,0x82,0x82,0x82,0xfc, +}; + +static const BitmapCharRec ch80 = {7,10,-1,0,9,ch80data}; + +/* char: 0x4f 'O' */ + +static const GLubyte ch79data[] = { +0x7c,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch79 = {7,10,-1,0,9,ch79data}; + +/* char: 0x4e 'N' */ + +static const GLubyte ch78data[] = { +0x82,0x82,0x82,0x86,0x8a,0x92,0xa2,0xc2,0x82,0x82, +}; + +static const BitmapCharRec ch78 = {7,10,-1,0,9,ch78data}; + +/* char: 0x4d 'M' */ + +static const GLubyte ch77data[] = { +0x82,0x82,0x82,0x92,0x92,0xaa,0xaa,0xc6,0x82,0x82, +}; + +static const BitmapCharRec ch77 = {7,10,-1,0,9,ch77data}; + +/* char: 0x4c 'L' */ + +static const GLubyte ch76data[] = { +0xfe,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch76 = {7,10,-1,0,9,ch76data}; + +/* char: 0x4b 'K' */ + +static const GLubyte ch75data[] = { +0x82,0x84,0x88,0x90,0xa0,0xe0,0x90,0x88,0x84,0x82, +}; + +static const BitmapCharRec ch75 = {7,10,-1,0,9,ch75data}; + +/* char: 0x4a 'J' */ + +static const GLubyte ch74data[] = { +0x78,0x84,0x4,0x4,0x4,0x4,0x4,0x4,0x4,0x1e, +}; + +static const BitmapCharRec ch74 = {7,10,-1,0,9,ch74data}; + +/* char: 0x49 'I' */ + +static const GLubyte ch73data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xf8, +}; + +static const BitmapCharRec ch73 = {5,10,-2,0,9,ch73data}; + +/* char: 0x48 'H' */ + +static const GLubyte ch72data[] = { +0x82,0x82,0x82,0x82,0x82,0xfe,0x82,0x82,0x82,0x82, +}; + +static const BitmapCharRec ch72 = {7,10,-1,0,9,ch72data}; + +/* char: 0x47 'G' */ + +static const GLubyte ch71data[] = { +0x7c,0x82,0x82,0x82,0x8e,0x80,0x80,0x80,0x82,0x7c, +}; + +static const BitmapCharRec ch71 = {7,10,-1,0,9,ch71data}; + +/* char: 0x46 'F' */ + +static const GLubyte ch70data[] = { +0x40,0x40,0x40,0x40,0x40,0x78,0x40,0x40,0x40,0xfe, +}; + +static const BitmapCharRec ch70 = {7,10,-1,0,9,ch70data}; + +/* char: 0x45 'E' */ + +static const GLubyte ch69data[] = { +0xfe,0x40,0x40,0x40,0x40,0x78,0x40,0x40,0x40,0xfe, +}; + +static const BitmapCharRec ch69 = {7,10,-1,0,9,ch69data}; + +/* char: 0x44 'D' */ + +static const GLubyte ch68data[] = { +0xfc,0x42,0x42,0x42,0x42,0x42,0x42,0x42,0x42,0xfc, +}; + +static const BitmapCharRec ch68 = {7,10,-1,0,9,ch68data}; + +/* char: 0x43 'C' */ + +static const GLubyte ch67data[] = { +0x7c,0x82,0x80,0x80,0x80,0x80,0x80,0x80,0x82,0x7c, +}; + +static const BitmapCharRec ch67 = {7,10,-1,0,9,ch67data}; + +/* char: 0x42 'B' */ + +static const GLubyte ch66data[] = { +0xfc,0x42,0x42,0x42,0x42,0x7c,0x42,0x42,0x42,0xfc, +}; + +static const BitmapCharRec ch66 = {7,10,-1,0,9,ch66data}; + +/* char: 0x41 'A' */ + +static const GLubyte ch65data[] = { +0x82,0x82,0x82,0xfe,0x82,0x82,0x82,0x44,0x28,0x10, +}; + +static const BitmapCharRec ch65 = {7,10,-1,0,9,ch65data}; + +/* char: 0x40 '@' */ + +static const GLubyte ch64data[] = { +0x7c,0x80,0x80,0x9a,0xa6,0xa2,0x9e,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch64 = {7,10,-1,0,9,ch64data}; + +/* char: 0x3f '?' */ + +static const GLubyte ch63data[] = { +0x10,0x0,0x10,0x10,0x8,0x4,0x2,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch63 = {7,10,-1,0,9,ch63data}; + +/* char: 0x3e '>' */ + +static const GLubyte ch62data[] = { +0x80,0x40,0x20,0x10,0x8,0x8,0x10,0x20,0x40,0x80, +}; + +static const BitmapCharRec ch62 = {5,10,-2,0,9,ch62data}; + +/* char: 0x3d '=' */ + +static const GLubyte ch61data[] = { +0xfe,0x0,0x0,0xfe, +}; + +static const BitmapCharRec ch61 = {7,4,-1,-2,9,ch61data}; + +/* char: 0x3c '<' */ + +static const GLubyte ch60data[] = { +0x8,0x10,0x20,0x40,0x80,0x80,0x40,0x20,0x10,0x8, +}; + +static const BitmapCharRec ch60 = {5,10,-2,0,9,ch60data}; + +/* char: 0x3b ';' */ + +static const GLubyte ch59data[] = { +0x80,0x40,0x40,0xc0,0xc0,0x0,0x0,0x0,0xc0,0xc0, +}; + +static const BitmapCharRec ch59 = {2,10,-4,3,9,ch59data}; + +/* char: 0x3a ':' */ + +static const GLubyte ch58data[] = { +0xc0,0xc0,0x0,0x0,0x0,0xc0,0xc0, +}; + +static const BitmapCharRec ch58 = {2,7,-4,0,9,ch58data}; + +/* char: 0x39 '9' */ + +static const GLubyte ch57data[] = { +0x78,0x4,0x2,0x2,0x7a,0x86,0x82,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch57 = {7,10,-1,0,9,ch57data}; + +/* char: 0x38 '8' */ + +static const GLubyte ch56data[] = { +0x38,0x44,0x82,0x82,0x44,0x38,0x44,0x82,0x44,0x38, +}; + +static const BitmapCharRec ch56 = {7,10,-1,0,9,ch56data}; + +/* char: 0x37 '7' */ + +static const GLubyte ch55data[] = { +0x40,0x40,0x20,0x20,0x10,0x8,0x4,0x2,0x2,0xfe, +}; + +static const BitmapCharRec ch55 = {7,10,-1,0,9,ch55data}; + +/* char: 0x36 '6' */ + +static const GLubyte ch54data[] = { +0x7c,0x82,0x82,0x82,0xc2,0xbc,0x80,0x80,0x40,0x3c, +}; + +static const BitmapCharRec ch54 = {7,10,-1,0,9,ch54data}; + +/* char: 0x35 '5' */ + +static const GLubyte ch53data[] = { +0x7c,0x82,0x2,0x2,0x2,0xc2,0xbc,0x80,0x80,0xfe, +}; + +static const BitmapCharRec ch53 = {7,10,-1,0,9,ch53data}; + +/* char: 0x34 '4' */ + +static const GLubyte ch52data[] = { +0x4,0x4,0x4,0xfe,0x84,0x44,0x24,0x14,0xc,0x4, +}; + +static const BitmapCharRec ch52 = {7,10,-1,0,9,ch52data}; + +/* char: 0x33 '3' */ + +static const GLubyte ch51data[] = { +0x7c,0x82,0x2,0x2,0x2,0x1c,0x8,0x4,0x2,0xfe, +}; + +static const BitmapCharRec ch51 = {7,10,-1,0,9,ch51data}; + +/* char: 0x32 '2' */ + +static const GLubyte ch50data[] = { +0xfe,0x80,0x40,0x30,0x8,0x4,0x2,0x82,0x82,0x7c, +}; + +static const BitmapCharRec ch50 = {7,10,-1,0,9,ch50data}; + +/* char: 0x31 '1' */ + +static const GLubyte ch49data[] = { +0xfe,0x10,0x10,0x10,0x10,0x10,0x90,0x50,0x30,0x10, +}; + +static const BitmapCharRec ch49 = {7,10,-1,0,9,ch49data}; + +/* char: 0x30 '0' */ + +static const GLubyte ch48data[] = { +0x38,0x44,0x82,0x82,0x82,0x82,0x82,0x82,0x44,0x38, +}; + +static const BitmapCharRec ch48 = {7,10,-1,0,9,ch48data}; + +/* char: 0x2f '/' */ + +static const GLubyte ch47data[] = { +0x80,0x40,0x40,0x20,0x10,0x10,0x8,0x4,0x4,0x2, +}; + +static const BitmapCharRec ch47 = {7,10,-1,0,9,ch47data}; + +/* char: 0x2e '.' */ + +static const GLubyte ch46data[] = { +0xc0,0xc0, +}; + +static const BitmapCharRec ch46 = {2,2,-4,0,9,ch46data}; + +/* char: 0x2d '-' */ + +static const GLubyte ch45data[] = { +0xfe, +}; + +static const BitmapCharRec ch45 = {7,1,-1,-4,9,ch45data}; + +/* char: 0x2c ',' */ + +static const GLubyte ch44data[] = { +0x80,0x40,0x40,0xc0,0xc0, +}; + +static const BitmapCharRec ch44 = {2,5,-4,3,9,ch44data}; + +/* char: 0x2b '+' */ + +static const GLubyte ch43data[] = { +0x10,0x10,0x10,0xfe,0x10,0x10,0x10, +}; + +static const BitmapCharRec ch43 = {7,7,-1,-1,9,ch43data}; + +/* char: 0x2a '*' */ + +static const GLubyte ch42data[] = { +0x10,0x92,0x54,0x38,0x54,0x92,0x10, +}; + +static const BitmapCharRec ch42 = {7,7,-1,-1,9,ch42data}; + +/* char: 0x29 ')' */ + +static const GLubyte ch41data[] = { +0x80,0x40,0x40,0x20,0x20,0x20,0x20,0x20,0x20,0x40,0x40,0x80, +}; + +static const BitmapCharRec ch41 = {3,12,-3,1,9,ch41data}; + +/* char: 0x28 '(' */ + +static const GLubyte ch40data[] = { +0x20,0x40,0x40,0x80,0x80,0x80,0x80,0x80,0x80,0x40,0x40,0x20, +}; + +static const BitmapCharRec ch40 = {3,12,-3,1,9,ch40data}; + +/* char: 0x27 ''' */ + +static const GLubyte ch39data[] = { +0x80,0x40,0x20,0x30, +}; + +static const BitmapCharRec ch39 = {4,4,-3,-6,9,ch39data}; + +/* char: 0x26 '&' */ + +static const GLubyte ch38data[] = { +0x62,0x94,0x88,0x94,0x62,0x60,0x90,0x90,0x90,0x60, +}; + +static const BitmapCharRec ch38 = {7,10,-1,0,9,ch38data}; + +/* char: 0x25 '%' */ + +static const GLubyte ch37data[] = { +0x84,0x4a,0x4a,0x24,0x10,0x10,0x48,0xa4,0xa4,0x42, +}; + +static const BitmapCharRec ch37 = {7,10,-1,0,9,ch37data}; + +/* char: 0x24 '$' */ + +static const GLubyte ch36data[] = { +0x10,0x7c,0x92,0x12,0x12,0x14,0x38,0x50,0x90,0x92,0x7c,0x10, +}; + +static const BitmapCharRec ch36 = {7,12,-1,1,9,ch36data}; + +/* char: 0x23 '#' */ + +static const GLubyte ch35data[] = { +0x48,0x48,0xfc,0x48,0x48,0xfc,0x48,0x48, +}; + +static const BitmapCharRec ch35 = {6,8,-1,-1,9,ch35data}; + +/* char: 0x22 '"' */ + +static const GLubyte ch34data[] = { +0x90,0x90,0x90, +}; + +static const BitmapCharRec ch34 = {4,3,-3,-7,9,ch34data}; + +/* char: 0x21 '!' */ + +static const GLubyte ch33data[] = { +0x80,0x80,0x0,0x0,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch33 = {1,11,-4,0,9,ch33data}; + +/* char: 0x1f */ + +static const GLubyte ch31data[] = { +0xc0,0xc0, +}; + +static const BitmapCharRec ch31 = {2,2,-4,-2,9,ch31data}; + +/* char: 0x1e */ + +static const GLubyte ch30data[] = { +0x5c,0xa2,0x60,0x20,0x20,0xf8,0x20,0x20,0x22,0x1c, +}; + +static const BitmapCharRec ch30 = {7,10,-1,0,9,ch30data}; + +/* char: 0x1d */ + +static const GLubyte ch29data[] = { +0x80,0x40,0xfe,0x10,0xfe,0x4,0x2, +}; + +static const BitmapCharRec ch29 = {7,7,-1,0,9,ch29data}; + +/* char: 0x1c */ + +static const GLubyte ch28data[] = { +0x44,0x24,0x24,0x24,0x24,0x24,0xfe, +}; + +static const BitmapCharRec ch28 = {7,7,-1,0,9,ch28data}; + +/* char: 0x1b */ + +static const GLubyte ch27data[] = { +0xfe,0x0,0x80,0x40,0x20,0x10,0x8,0x8,0x10,0x20,0x40,0x80, +}; + +static const BitmapCharRec ch27 = {7,12,-1,2,9,ch27data}; + +/* char: 0x1a */ + +static const GLubyte ch26data[] = { +0xfc,0x0,0x4,0x8,0x10,0x20,0x40,0x40,0x20,0x10,0x8,0x4, +}; + +static const BitmapCharRec ch26 = {6,12,-2,2,9,ch26data}; + +/* char: 0x19 */ + +static const GLubyte ch25data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch25 = {1,15,-4,3,9,ch25data}; + +/* char: 0x18 */ + +static const GLubyte ch24data[] = { +0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0xff,0x80, +}; + +static const BitmapCharRec ch24 = {9,7,0,3,9,ch24data}; + +/* char: 0x17 */ + +static const GLubyte ch23data[] = { +0xff,0x80,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0, +0x8,0x0, +}; + +static const BitmapCharRec ch23 = {9,9,0,-3,9,ch23data}; + +/* char: 0x16 */ + +static const GLubyte ch22data[] = { +0x8,0x8,0x8,0x8,0x8,0x8,0xf8,0x8,0x8,0x8,0x8,0x8,0x8,0x8,0x8, +}; + +static const BitmapCharRec ch22 = {5,15,0,3,9,ch22data}; + +/* char: 0x15 */ + +static const GLubyte ch21data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch21 = {5,15,-4,3,9,ch21data}; + +/* char: 0x14 */ + +static const GLubyte ch20data[] = { +0xff,0x80, +}; + +static const BitmapCharRec ch20 = {9,1,0,1,9,ch20data}; + +/* char: 0x13 */ + +static const GLubyte ch19data[] = { +0xff,0x80, +}; + +static const BitmapCharRec ch19 = {9,1,0,-1,9,ch19data}; + +/* char: 0x12 */ + +static const GLubyte ch18data[] = { +0xff,0x80, +}; + +static const BitmapCharRec ch18 = {9,1,0,-3,9,ch18data}; + +/* char: 0x11 */ + +static const GLubyte ch17data[] = { +0xff,0x80, +}; + +static const BitmapCharRec ch17 = {9,1,0,-5,9,ch17data}; + +/* char: 0x10 */ + +static const GLubyte ch16data[] = { +0xff,0x80, +}; + +static const BitmapCharRec ch16 = {9,1,0,-7,9,ch16data}; + +/* char: 0xf */ + +static const GLubyte ch15data[] = { +0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0xff,0x80,0x8,0x0, +0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0,0x8,0x0, +}; + +static const BitmapCharRec ch15 = {9,15,0,3,9,ch15data}; + +/* char: 0xe */ + +static const GLubyte ch14data[] = { +0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch14 = {5,9,-4,-3,9,ch14data}; + +/* char: 0xd */ + +static const GLubyte ch13data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0xf8, +}; + +static const BitmapCharRec ch13 = {5,7,-4,3,9,ch13data}; + +/* char: 0xc */ + +static const GLubyte ch12data[] = { +0x8,0x8,0x8,0x8,0x8,0x8,0xf8, +}; + +static const BitmapCharRec ch12 = {5,7,0,3,9,ch12data}; + +/* char: 0xb */ + +static const GLubyte ch11data[] = { +0xf8,0x8,0x8,0x8,0x8,0x8,0x8,0x8,0x8, +}; + +static const BitmapCharRec ch11 = {5,9,0,-3,9,ch11data}; + +/* char: 0xa */ + +static const GLubyte ch10data[] = { +0x8,0x8,0x8,0x8,0x3e,0x0,0x20,0x50,0x88,0x88, +}; + +static const BitmapCharRec ch10 = {7,10,-1,2,9,ch10data}; + +/* char: 0x9 */ + +static const GLubyte ch9data[] = { +0x3e,0x20,0x20,0x20,0x20,0x88,0x98,0xa8,0xc8,0x88, +}; + +static const BitmapCharRec ch9 = {7,10,-1,2,9,ch9data}; + +/* char: 0x8 */ + +static const GLubyte ch8data[] = { +0xfe,0x10,0x10,0xfe,0x10,0x10, +}; + +static const BitmapCharRec ch8 = {7,6,-1,0,9,ch8data}; + +/* char: 0x7 */ + +static const GLubyte ch7data[] = { +0x70,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch7 = {5,4,-2,-6,9,ch7data}; + +/* char: 0x6 */ + +static const GLubyte ch6data[] = { +0x20,0x20,0x3c,0x20,0x3e,0x0,0xf8,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch6 = {7,10,-1,2,9,ch6data}; + +/* char: 0x5 */ + +static const GLubyte ch5data[] = { +0x22,0x22,0x3c,0x22,0x3c,0x0,0x78,0x80,0x80,0x78, +}; + +static const BitmapCharRec ch5 = {7,10,-1,2,9,ch5data}; + +/* char: 0x4 */ + +static const GLubyte ch4data[] = { +0x10,0x10,0x1c,0x10,0x1e,0x80,0x80,0xe0,0x80,0xf0, +}; + +static const BitmapCharRec ch4 = {7,10,-1,2,9,ch4data}; + +/* char: 0x3 */ + +static const GLubyte ch3data[] = { +0x8,0x8,0x8,0x3e,0x0,0x88,0x88,0xf8,0x88,0x88, +}; + +static const BitmapCharRec ch3 = {7,10,-1,2,9,ch3data}; + +/* char: 0x2 */ + +static const GLubyte ch2data[] = { +0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa,0x55,0xaa, +}; + +static const BitmapCharRec ch2 = {8,14,0,3,9,ch2data}; + +/* char: 0x1 */ + +static const GLubyte ch1data[] = { +0x10,0x38,0x7c,0xfe,0x7c,0x38,0x10, +}; + +static const BitmapCharRec ch1 = {7,7,-1,0,9,ch1data}; + +static const BitmapCharRec * const chars[] = { +&ch0, +&ch1, +&ch2, +&ch3, +&ch4, +&ch5, +&ch6, +&ch7, +&ch8, +&ch9, +&ch10, +&ch11, +&ch12, +&ch13, +&ch14, +&ch15, +&ch16, +&ch17, +&ch18, +&ch19, +&ch20, +&ch21, +&ch22, +&ch23, +&ch24, +&ch25, +&ch26, +&ch27, +&ch28, +&ch29, +&ch30, +&ch31, +&ch32, +&ch33, +&ch34, +&ch35, +&ch36, +&ch37, +&ch38, +&ch39, +&ch40, +&ch41, +&ch42, +&ch43, +&ch44, +&ch45, +&ch46, +&ch47, +&ch48, +&ch49, +&ch50, +&ch51, +&ch52, +&ch53, +&ch54, +&ch55, +&ch56, +&ch57, +&ch58, +&ch59, +&ch60, +&ch61, +&ch62, +&ch63, +&ch64, +&ch65, +&ch66, +&ch67, +&ch68, +&ch69, +&ch70, +&ch71, +&ch72, +&ch73, +&ch74, +&ch75, +&ch76, +&ch77, +&ch78, +&ch79, +&ch80, +&ch81, +&ch82, +&ch83, +&ch84, +&ch85, +&ch86, +&ch87, +&ch88, +&ch89, +&ch90, +&ch91, +&ch92, +&ch93, +&ch94, +&ch95, +&ch96, +&ch97, +&ch98, +&ch99, +&ch100, +&ch101, +&ch102, +&ch103, +&ch104, +&ch105, +&ch106, +&ch107, +&ch108, +&ch109, +&ch110, +&ch111, +&ch112, +&ch113, +&ch114, +&ch115, +&ch116, +&ch117, +&ch118, +&ch119, +&ch120, +&ch121, +&ch122, +&ch123, +&ch124, +&ch125, +&ch126, +&ch127, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +&ch160, +&ch161, +&ch162, +&ch163, +&ch164, +&ch165, +&ch166, +&ch167, +&ch168, +&ch169, +&ch170, +&ch171, +&ch172, +&ch173, +&ch174, +&ch175, +&ch176, +&ch177, +&ch178, +&ch179, +&ch180, +&ch181, +&ch182, +&ch183, +&ch184, +&ch185, +&ch186, +&ch187, +&ch188, +&ch189, +&ch190, +&ch191, +&ch192, +&ch193, +&ch194, +&ch195, +&ch196, +&ch197, +&ch198, +&ch199, +&ch200, +&ch201, +&ch202, +&ch203, +&ch204, +&ch205, +&ch206, +&ch207, +&ch208, +&ch209, +&ch210, +&ch211, +&ch212, +&ch213, +&ch214, +&ch215, +&ch216, +&ch217, +&ch218, +&ch219, +&ch220, +&ch221, +&ch222, +&ch223, +&ch224, +&ch225, +&ch226, +&ch227, +&ch228, +&ch229, +&ch230, +&ch231, +&ch232, +&ch233, +&ch234, +&ch235, +&ch236, +&ch237, +&ch238, +&ch239, +&ch240, +&ch241, +&ch242, +&ch243, +&ch244, +&ch245, +&ch246, +&ch247, +&ch248, +&ch249, +&ch250, +&ch251, +&ch252, +&ch253, +&ch254, +&ch255, +}; +#if !defined(__IBMCPP__) +const +#endif + BitmapFontRec glutBitmap9By15 = { +"-misc-fixed-medium-r-normal--15-140-75-75-C-90-iso8859-1", +256, +0, +chars +}; + diff --git a/src/glut/os2/glut_bitmap.cpp b/src/glut/os2/glut_bitmap.cpp index 789a89ae73..40530b69c1 100644 --- a/src/glut/os2/glut_bitmap.cpp +++ b/src/glut/os2/glut_bitmap.cpp @@ -1,57 +1,57 @@ - -/* Copyright (c) Mark J. Kilgard, 1994. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include "glutint.h" -#include "glutbitmap.h" - -void GLUTAPIENTRY -glutBitmapCharacter(GLUTbitmapFont font, int c) -{ - const BitmapCharRec *ch; - BitmapFontPtr fontinfo; - GLint swapbytes, lsbfirst, rowlength; - GLint skiprows, skippixels, alignment; - -#if defined(_WIN32) - fontinfo = (BitmapFontPtr) __glutFont(font); -#else - fontinfo = (BitmapFontPtr) font; -#endif - - if (c < fontinfo->first || - c >= fontinfo->first + fontinfo->num_chars) - return; - ch = fontinfo->ch[c - fontinfo->first]; - if (ch) { - /* Save current modes. */ - glGetIntegerv(GL_UNPACK_SWAP_BYTES, &swapbytes); - glGetIntegerv(GL_UNPACK_LSB_FIRST, &lsbfirst); - glGetIntegerv(GL_UNPACK_ROW_LENGTH, &rowlength); - glGetIntegerv(GL_UNPACK_SKIP_ROWS, &skiprows); - glGetIntegerv(GL_UNPACK_SKIP_PIXELS, &skippixels); - glGetIntegerv(GL_UNPACK_ALIGNMENT, &alignment); - /* Little endian machines (DEC Alpha for example) could - benefit from setting GL_UNPACK_LSB_FIRST to GL_TRUE - instead of GL_FALSE, but this would require changing the - generated bitmaps too. */ - glPixelStorei(GL_UNPACK_SWAP_BYTES, GL_FALSE); - glPixelStorei(GL_UNPACK_LSB_FIRST, GL_FALSE); - glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); - glPixelStorei(GL_UNPACK_SKIP_ROWS, 0); - glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0); - glPixelStorei(GL_UNPACK_ALIGNMENT, 1); - glBitmap(ch->width, ch->height, ch->xorig, ch->yorig, - ch->advance, 0, ch->bitmap); - /* Restore saved modes. */ - glPixelStorei(GL_UNPACK_SWAP_BYTES, swapbytes); - glPixelStorei(GL_UNPACK_LSB_FIRST, lsbfirst); - glPixelStorei(GL_UNPACK_ROW_LENGTH, rowlength); - glPixelStorei(GL_UNPACK_SKIP_ROWS, skiprows); - glPixelStorei(GL_UNPACK_SKIP_PIXELS, skippixels); - glPixelStorei(GL_UNPACK_ALIGNMENT, alignment); - } -} + +/* Copyright (c) Mark J. Kilgard, 1994. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include "glutint.h" +#include "glutbitmap.h" + +void GLUTAPIENTRY +glutBitmapCharacter(GLUTbitmapFont font, int c) +{ + const BitmapCharRec *ch; + BitmapFontPtr fontinfo; + GLint swapbytes, lsbfirst, rowlength; + GLint skiprows, skippixels, alignment; + +#if defined(_WIN32) + fontinfo = (BitmapFontPtr) __glutFont(font); +#else + fontinfo = (BitmapFontPtr) font; +#endif + + if (c < fontinfo->first || + c >= fontinfo->first + fontinfo->num_chars) + return; + ch = fontinfo->ch[c - fontinfo->first]; + if (ch) { + /* Save current modes. */ + glGetIntegerv(GL_UNPACK_SWAP_BYTES, &swapbytes); + glGetIntegerv(GL_UNPACK_LSB_FIRST, &lsbfirst); + glGetIntegerv(GL_UNPACK_ROW_LENGTH, &rowlength); + glGetIntegerv(GL_UNPACK_SKIP_ROWS, &skiprows); + glGetIntegerv(GL_UNPACK_SKIP_PIXELS, &skippixels); + glGetIntegerv(GL_UNPACK_ALIGNMENT, &alignment); + /* Little endian machines (DEC Alpha for example) could + benefit from setting GL_UNPACK_LSB_FIRST to GL_TRUE + instead of GL_FALSE, but this would require changing the + generated bitmaps too. */ + glPixelStorei(GL_UNPACK_SWAP_BYTES, GL_FALSE); + glPixelStorei(GL_UNPACK_LSB_FIRST, GL_FALSE); + glPixelStorei(GL_UNPACK_ROW_LENGTH, 0); + glPixelStorei(GL_UNPACK_SKIP_ROWS, 0); + glPixelStorei(GL_UNPACK_SKIP_PIXELS, 0); + glPixelStorei(GL_UNPACK_ALIGNMENT, 1); + glBitmap(ch->width, ch->height, ch->xorig, ch->yorig, + ch->advance, 0, ch->bitmap); + /* Restore saved modes. */ + glPixelStorei(GL_UNPACK_SWAP_BYTES, swapbytes); + glPixelStorei(GL_UNPACK_LSB_FIRST, lsbfirst); + glPixelStorei(GL_UNPACK_ROW_LENGTH, rowlength); + glPixelStorei(GL_UNPACK_SKIP_ROWS, skiprows); + glPixelStorei(GL_UNPACK_SKIP_PIXELS, skippixels); + glPixelStorei(GL_UNPACK_ALIGNMENT, alignment); + } +} diff --git a/src/glut/os2/glut_cmap.cpp b/src/glut/os2/glut_cmap.cpp index d5b8d436a5..afe44f0aca 100644 --- a/src/glut/os2/glut_cmap.cpp +++ b/src/glut/os2/glut_cmap.cpp @@ -1,399 +1,399 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1996, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#ifdef __VMS -//EK#include -#endif - -#include -#include -#include /* SunOS multithreaded assert() needs . Lame. */ -#include -#if !defined(_WIN32) && !defined(__OS2__) - #include - #include - #include /* for XA_RGB_DEFAULT_MAP atom */ - #if defined(__vms) - #include /* for XmuLookupStandardColormap */ - #else - #include /* for XmuLookupStandardColormap */ - #endif -#endif - -/* SGI optimization introduced in IRIX 6.3 to avoid X server - round trips for interning common X atoms. */ -#if defined(_SGI_EXTRA_PREDEFINES) && !defined(NO_FAST_ATOMS) -#include -#else -#define XSGIFastInternAtom(dpy,string,fast_name,how) XInternAtom(dpy,string,how) -#endif - -#include "glutint.h" -#include "layerutil.h" - -GLUTcolormap *__glutColormapList = NULL; - -GLUTcolormap * -__glutAssociateNewColormap(XVisualInfo * vis) -{ - GLUTcolormap *cmap; - int transparentPixel, i; - unsigned long pixels[255]; - - cmap = (GLUTcolormap *) malloc(sizeof(GLUTcolormap)); - if (!cmap) - __glutFatalError("out of memory."); -#if defined(_WIN32) || defined(__OS2__) - pixels[0] = 0; /* avoid compilation warnings on win32 */ - cmap->visual = 0; - cmap->size = 256; /* always assume 256 on Win32 */ -#else - cmap->visual = vis->visual; - cmap->size = vis->visual->map_entries; -#endif - cmap->refcnt = 1; - cmap->cells = (GLUTcolorcell *) - malloc(sizeof(GLUTcolorcell) * cmap->size); - if (!cmap->cells) - __glutFatalError("out of memory."); - /* make all color cell entries be invalid */ - for (i = cmap->size - 1; i >= 0; i--) { - cmap->cells[i].component[GLUT_RED] = -1.0; - cmap->cells[i].component[GLUT_GREEN] = -1.0; - cmap->cells[i].component[GLUT_BLUE] = -1.0; - } - transparentPixel = __glutGetTransparentPixel(__glutDisplay, vis); - if (transparentPixel == -1 || transparentPixel >= cmap->size) { - - /* If there is no transparent pixel or if the transparent - pixel is outside the range of valid colormap cells (HP - can implement their overlays this smart way since their - transparent pixel is 255), we can AllocAll the colormap. - See note below. */ - - cmap->cmap = XCreateColormap(__glutDisplay, - __glutRoot, cmap->visual, AllocAll); - } else { - - /* On machines where zero (or some other value in the range - of 0 through map_entries-1), BadAlloc may be generated - when an AllocAll overlay colormap is allocated since the - transparent pixel precludes all the cells in the colormap - being allocated (the transparent pixel is pre-allocated). - So in this case, use XAllocColorCells to allocate - map_entries-1 pixels (that is, all but the transparent - pixel. */ - -#if defined(_WIN32) || defined(__OS2__) - cmap->cmap = XCreateColormap(__glutDisplay, - __glutRoot, 0, AllocNone); -#else - cmap->cmap = XCreateColormap(__glutDisplay, - __glutRoot, vis->visual, AllocNone); - XAllocColorCells(__glutDisplay, cmap->cmap, False, 0, 0, - pixels, cmap->size - 1); -#endif - } - cmap->next = __glutColormapList; - __glutColormapList = cmap; - return cmap; -} - -static GLUTcolormap * -associateColormap(XVisualInfo * vis) -{ -#if !defined(_WIN32) && !defined(__OS2__) - GLUTcolormap *cmap = __glutColormapList; - - while (cmap != NULL) { - /* Play safe: compare visual IDs, not Visual*'s. */ - if (cmap->visual->visualid == vis->visual->visualid) { - /* Already have created colormap for the visual. */ - cmap->refcnt++; - return cmap; - } - cmap = cmap->next; - } -#endif - return __glutAssociateNewColormap(vis); -} - -void -__glutSetupColormap(XVisualInfo * vi, GLUTcolormap ** colormap, Colormap * cmap) -{ -#if defined(_WIN32) || defined(__OS2__) - if (vi->dwFlags & PFD_NEED_PALETTE || vi->iPixelType == PFD_TYPE_COLORINDEX) { - *colormap = associateColormap(vi); - *cmap = (*colormap)->cmap; - } else { - *colormap = NULL; - *cmap = 0; - } -#else - Status status; - XStandardColormap *standardCmaps; - int i, numCmaps; - static Atom hpColorRecoveryAtom = -1; - int isRGB, visualClass, rc; - -#if defined(__cplusplus) || defined(c_plusplus) - visualClass = vi->c_class; -#else - visualClass = vi->class; -#endif - switch (visualClass) { - case PseudoColor: - /* Mesa might return a PseudoColor visual for RGB mode. */ - rc = glXGetConfig(__glutDisplay, vi, GLX_RGBA, &isRGB); - if (rc == 0 && isRGB) { - /* Must be Mesa. */ - *colormap = NULL; - if (MaxCmapsOfScreen(DefaultScreenOfDisplay(__glutDisplay)) == 1 - && vi->visual == DefaultVisual(__glutDisplay, __glutScreen)) { - char *privateCmap = getenv("MESA_PRIVATE_CMAP"); - - if (privateCmap) { - /* User doesn't want to share colormaps. */ - *cmap = XCreateColormap(__glutDisplay, __glutRoot, - vi->visual, AllocNone); - } else { - /* Share the root colormap. */ - *cmap = DefaultColormap(__glutDisplay, __glutScreen); - } - } else { - /* Get our own PseudoColor colormap. */ - *cmap = XCreateColormap(__glutDisplay, __glutRoot, - vi->visual, AllocNone); - } - } else { - /* CI mode, real GLX never returns a PseudoColor visual - for RGB mode. */ - *colormap = associateColormap(vi); - *cmap = (*colormap)->cmap; - } - break; - case TrueColor: - case DirectColor: - *colormap = NULL; /* NULL if RGBA */ - - /* Hewlett-Packard supports a feature called "HP Color - Recovery". Mesa has code to use HP Color Recovery. For - Mesa to use this feature, the atom - _HP_RGB_SMOOTH_MAP_LIST must be defined on the root - window AND the colormap obtainable by XGetRGBColormaps - for that atom must be set on the window. If that - colormap is not set, the output will look stripy. */ - - if (hpColorRecoveryAtom == -1) { - char *xvendor; - -#define VENDOR_HP "Hewlett-Packard" - - /* Only makes sense to make XInternAtom round-trip if we - know that we are connected to an HP X server. */ - xvendor = ServerVendor(__glutDisplay); - if (!strncmp(xvendor, VENDOR_HP, sizeof(VENDOR_HP) - 1)) { - hpColorRecoveryAtom = XInternAtom(__glutDisplay, "_HP_RGB_SMOOTH_MAP_LIST", True); - } else { - hpColorRecoveryAtom = None; - } - } - if (hpColorRecoveryAtom != None) { - status = XGetRGBColormaps(__glutDisplay, __glutRoot, - &standardCmaps, &numCmaps, hpColorRecoveryAtom); - if (status == 1) { - for (i = 0; i < numCmaps; i++) { - if (standardCmaps[i].visualid == vi->visualid) { - *cmap = standardCmaps[i].colormap; - XFree(standardCmaps); - return; - } - } - XFree(standardCmaps); - } - } -#ifndef SOLARIS_2_4_BUG - /* Solaris 2.4 and 2.5 have a bug in their - XmuLookupStandardColormap implementations. Please - compile your Solaris 2.4 or 2.5 version of GLUT with - -DSOLARIS_2_4_BUG to work around this bug. The symptom - of the bug is that programs will get a BadMatch error - from X_CreateWindow when creating a GLUT window because - Solaris 2.4 and 2.5 create a corrupted RGB_DEFAULT_MAP - property. Note that this workaround prevents Colormap - sharing between applications, perhaps leading - unnecessary colormap installations or colormap flashing. - Sun fixed this bug in Solaris 2.6. */ - status = XmuLookupStandardColormap(__glutDisplay, - vi->screen, vi->visualid, vi->depth, XA_RGB_DEFAULT_MAP, - /* replace */ False, /* retain */ True); - if (status == 1) { - status = XGetRGBColormaps(__glutDisplay, __glutRoot, - &standardCmaps, &numCmaps, XA_RGB_DEFAULT_MAP); - if (status == 1) { - for (i = 0; i < numCmaps; i++) { - if (standardCmaps[i].visualid == vi->visualid) { - *cmap = standardCmaps[i].colormap; - XFree(standardCmaps); - return; - } - } - XFree(standardCmaps); - } - } -#endif - /* If no standard colormap but TrueColor, just make a - private one. */ - /* XXX Should do a better job of internal sharing for - privately allocated TrueColor colormaps. */ - /* XXX DirectColor probably needs ramps hand initialized! */ - *cmap = XCreateColormap(__glutDisplay, __glutRoot, - vi->visual, AllocNone); - break; - case StaticColor: - case StaticGray: - case GrayScale: - /* Mesa supports these visuals */ - *colormap = NULL; - *cmap = XCreateColormap(__glutDisplay, __glutRoot, - vi->visual, AllocNone); - break; - default: - __glutFatalError( - "could not allocate colormap for visual type: %d.", - visualClass); - } - return; -#endif -} - -#if !defined(_WIN32) && !defined(__OS2__) -static int -findColormaps(GLUTwindow * window, - Window * winlist, Colormap * cmaplist, int num, int max) -{ - GLUTwindow *child; - int i; - - /* Do not allow more entries that maximum number of - colormaps! */ - if (num >= max) - return num; - /* Is cmap for this window already on the list? */ - for (i = 0; i < num; i++) { - if (cmaplist[i] == window->cmap) - goto normalColormapAlreadyListed; - } - /* Not found on the list; add colormap and window. */ - winlist[num] = window->win; - cmaplist[num] = window->cmap; - num++; - -normalColormapAlreadyListed: - - /* Repeat above but for the overlay colormap if there one. */ - if (window->overlay) { - if (num >= max) - return num; - for (i = 0; i < num; i++) { - if (cmaplist[i] == window->overlay->cmap) - goto overlayColormapAlreadyListed; - } - winlist[num] = window->overlay->win; - cmaplist[num] = window->overlay->cmap; - num++; - } -overlayColormapAlreadyListed: - - /* Recursively search children. */ - child = window->children; - while (child) { - num = findColormaps(child, winlist, cmaplist, num, max); - child = child->siblings; - } - return num; -} - -void -__glutEstablishColormapsProperty(GLUTwindow * window) -{ - /* this routine is strictly X. Win32 doesn't need to do - anything of this sort (but has to do other wacky stuff - later). */ - static Atom wmColormapWindows = None; - Window *winlist; - Colormap *cmaplist; - Status status; - int maxcmaps, num; - - assert(!window->parent); - maxcmaps = MaxCmapsOfScreen(ScreenOfDisplay(__glutDisplay, - __glutScreen)); - /* For portability reasons we don't use alloca for winlist - and cmaplist, but we could. */ - winlist = (Window *) malloc(maxcmaps * sizeof(Window)); - cmaplist = (Colormap *) malloc(maxcmaps * sizeof(Colormap)); - num = findColormaps(window, winlist, cmaplist, 0, maxcmaps); - if (num < 2) { - /* Property no longer needed; remove it. */ - wmColormapWindows = XSGIFastInternAtom(__glutDisplay, - "WM_COLORMAP_WINDOWS", SGI_XA_WM_COLORMAP_WINDOWS, False); - if (wmColormapWindows == None) { - __glutWarning("Could not intern X atom for WM_COLORMAP_WINDOWS."); - return; - } - XDeleteProperty(__glutDisplay, window->win, wmColormapWindows); - } else { - status = XSetWMColormapWindows(__glutDisplay, window->win, - winlist, num); - /* XSetWMColormapWindows should always work unless the - WM_COLORMAP_WINDOWS property cannot be intern'ed. We - check to be safe. */ - if (status == False) - __glutFatalError("XSetWMColormapWindows returned False."); - } - /* For portability reasons we don't use alloca for winlist - and cmaplist, but we could. */ - free(winlist); - free(cmaplist); -} - -GLUTwindow * -__glutToplevelOf(GLUTwindow * window) -{ - while (window->parent) { - window = window->parent; - } - return window; -} -#endif - -void -__glutFreeColormap(GLUTcolormap * cmap) -{ - GLUTcolormap *cur, **prev; - - cmap->refcnt--; - if (cmap->refcnt == 0) { - /* remove from colormap list */ - cur = __glutColormapList; - prev = &__glutColormapList; - while (cur) { - if (cur == cmap) { - *prev = cmap->next; - break; - } - prev = &(cur->next); - cur = cur->next; - } - /* actually free colormap */ - XFreeColormap(__glutDisplay, cmap->cmap); - free(cmap->cells); - free(cmap); - } -} - + +/* Copyright (c) Mark J. Kilgard, 1994, 1996, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#ifdef __VMS +//EK#include +#endif + +#include +#include +#include /* SunOS multithreaded assert() needs . Lame. */ +#include +#if !defined(_WIN32) && !defined(__OS2__) + #include + #include + #include /* for XA_RGB_DEFAULT_MAP atom */ + #if defined(__vms) + #include /* for XmuLookupStandardColormap */ + #else + #include /* for XmuLookupStandardColormap */ + #endif +#endif + +/* SGI optimization introduced in IRIX 6.3 to avoid X server + round trips for interning common X atoms. */ +#if defined(_SGI_EXTRA_PREDEFINES) && !defined(NO_FAST_ATOMS) +#include +#else +#define XSGIFastInternAtom(dpy,string,fast_name,how) XInternAtom(dpy,string,how) +#endif + +#include "glutint.h" +#include "layerutil.h" + +GLUTcolormap *__glutColormapList = NULL; + +GLUTcolormap * +__glutAssociateNewColormap(XVisualInfo * vis) +{ + GLUTcolormap *cmap; + int transparentPixel, i; + unsigned long pixels[255]; + + cmap = (GLUTcolormap *) malloc(sizeof(GLUTcolormap)); + if (!cmap) + __glutFatalError("out of memory."); +#if defined(_WIN32) || defined(__OS2__) + pixels[0] = 0; /* avoid compilation warnings on win32 */ + cmap->visual = 0; + cmap->size = 256; /* always assume 256 on Win32 */ +#else + cmap->visual = vis->visual; + cmap->size = vis->visual->map_entries; +#endif + cmap->refcnt = 1; + cmap->cells = (GLUTcolorcell *) + malloc(sizeof(GLUTcolorcell) * cmap->size); + if (!cmap->cells) + __glutFatalError("out of memory."); + /* make all color cell entries be invalid */ + for (i = cmap->size - 1; i >= 0; i--) { + cmap->cells[i].component[GLUT_RED] = -1.0; + cmap->cells[i].component[GLUT_GREEN] = -1.0; + cmap->cells[i].component[GLUT_BLUE] = -1.0; + } + transparentPixel = __glutGetTransparentPixel(__glutDisplay, vis); + if (transparentPixel == -1 || transparentPixel >= cmap->size) { + + /* If there is no transparent pixel or if the transparent + pixel is outside the range of valid colormap cells (HP + can implement their overlays this smart way since their + transparent pixel is 255), we can AllocAll the colormap. + See note below. */ + + cmap->cmap = XCreateColormap(__glutDisplay, + __glutRoot, cmap->visual, AllocAll); + } else { + + /* On machines where zero (or some other value in the range + of 0 through map_entries-1), BadAlloc may be generated + when an AllocAll overlay colormap is allocated since the + transparent pixel precludes all the cells in the colormap + being allocated (the transparent pixel is pre-allocated). + So in this case, use XAllocColorCells to allocate + map_entries-1 pixels (that is, all but the transparent + pixel. */ + +#if defined(_WIN32) || defined(__OS2__) + cmap->cmap = XCreateColormap(__glutDisplay, + __glutRoot, 0, AllocNone); +#else + cmap->cmap = XCreateColormap(__glutDisplay, + __glutRoot, vis->visual, AllocNone); + XAllocColorCells(__glutDisplay, cmap->cmap, False, 0, 0, + pixels, cmap->size - 1); +#endif + } + cmap->next = __glutColormapList; + __glutColormapList = cmap; + return cmap; +} + +static GLUTcolormap * +associateColormap(XVisualInfo * vis) +{ +#if !defined(_WIN32) && !defined(__OS2__) + GLUTcolormap *cmap = __glutColormapList; + + while (cmap != NULL) { + /* Play safe: compare visual IDs, not Visual*'s. */ + if (cmap->visual->visualid == vis->visual->visualid) { + /* Already have created colormap for the visual. */ + cmap->refcnt++; + return cmap; + } + cmap = cmap->next; + } +#endif + return __glutAssociateNewColormap(vis); +} + +void +__glutSetupColormap(XVisualInfo * vi, GLUTcolormap ** colormap, Colormap * cmap) +{ +#if defined(_WIN32) || defined(__OS2__) + if (vi->dwFlags & PFD_NEED_PALETTE || vi->iPixelType == PFD_TYPE_COLORINDEX) { + *colormap = associateColormap(vi); + *cmap = (*colormap)->cmap; + } else { + *colormap = NULL; + *cmap = 0; + } +#else + Status status; + XStandardColormap *standardCmaps; + int i, numCmaps; + static Atom hpColorRecoveryAtom = -1; + int isRGB, visualClass, rc; + +#if defined(__cplusplus) || defined(c_plusplus) + visualClass = vi->c_class; +#else + visualClass = vi->class; +#endif + switch (visualClass) { + case PseudoColor: + /* Mesa might return a PseudoColor visual for RGB mode. */ + rc = glXGetConfig(__glutDisplay, vi, GLX_RGBA, &isRGB); + if (rc == 0 && isRGB) { + /* Must be Mesa. */ + *colormap = NULL; + if (MaxCmapsOfScreen(DefaultScreenOfDisplay(__glutDisplay)) == 1 + && vi->visual == DefaultVisual(__glutDisplay, __glutScreen)) { + char *privateCmap = getenv("MESA_PRIVATE_CMAP"); + + if (privateCmap) { + /* User doesn't want to share colormaps. */ + *cmap = XCreateColormap(__glutDisplay, __glutRoot, + vi->visual, AllocNone); + } else { + /* Share the root colormap. */ + *cmap = DefaultColormap(__glutDisplay, __glutScreen); + } + } else { + /* Get our own PseudoColor colormap. */ + *cmap = XCreateColormap(__glutDisplay, __glutRoot, + vi->visual, AllocNone); + } + } else { + /* CI mode, real GLX never returns a PseudoColor visual + for RGB mode. */ + *colormap = associateColormap(vi); + *cmap = (*colormap)->cmap; + } + break; + case TrueColor: + case DirectColor: + *colormap = NULL; /* NULL if RGBA */ + + /* Hewlett-Packard supports a feature called "HP Color + Recovery". Mesa has code to use HP Color Recovery. For + Mesa to use this feature, the atom + _HP_RGB_SMOOTH_MAP_LIST must be defined on the root + window AND the colormap obtainable by XGetRGBColormaps + for that atom must be set on the window. If that + colormap is not set, the output will look stripy. */ + + if (hpColorRecoveryAtom == -1) { + char *xvendor; + +#define VENDOR_HP "Hewlett-Packard" + + /* Only makes sense to make XInternAtom round-trip if we + know that we are connected to an HP X server. */ + xvendor = ServerVendor(__glutDisplay); + if (!strncmp(xvendor, VENDOR_HP, sizeof(VENDOR_HP) - 1)) { + hpColorRecoveryAtom = XInternAtom(__glutDisplay, "_HP_RGB_SMOOTH_MAP_LIST", True); + } else { + hpColorRecoveryAtom = None; + } + } + if (hpColorRecoveryAtom != None) { + status = XGetRGBColormaps(__glutDisplay, __glutRoot, + &standardCmaps, &numCmaps, hpColorRecoveryAtom); + if (status == 1) { + for (i = 0; i < numCmaps; i++) { + if (standardCmaps[i].visualid == vi->visualid) { + *cmap = standardCmaps[i].colormap; + XFree(standardCmaps); + return; + } + } + XFree(standardCmaps); + } + } +#ifndef SOLARIS_2_4_BUG + /* Solaris 2.4 and 2.5 have a bug in their + XmuLookupStandardColormap implementations. Please + compile your Solaris 2.4 or 2.5 version of GLUT with + -DSOLARIS_2_4_BUG to work around this bug. The symptom + of the bug is that programs will get a BadMatch error + from X_CreateWindow when creating a GLUT window because + Solaris 2.4 and 2.5 create a corrupted RGB_DEFAULT_MAP + property. Note that this workaround prevents Colormap + sharing between applications, perhaps leading + unnecessary colormap installations or colormap flashing. + Sun fixed this bug in Solaris 2.6. */ + status = XmuLookupStandardColormap(__glutDisplay, + vi->screen, vi->visualid, vi->depth, XA_RGB_DEFAULT_MAP, + /* replace */ False, /* retain */ True); + if (status == 1) { + status = XGetRGBColormaps(__glutDisplay, __glutRoot, + &standardCmaps, &numCmaps, XA_RGB_DEFAULT_MAP); + if (status == 1) { + for (i = 0; i < numCmaps; i++) { + if (standardCmaps[i].visualid == vi->visualid) { + *cmap = standardCmaps[i].colormap; + XFree(standardCmaps); + return; + } + } + XFree(standardCmaps); + } + } +#endif + /* If no standard colormap but TrueColor, just make a + private one. */ + /* XXX Should do a better job of internal sharing for + privately allocated TrueColor colormaps. */ + /* XXX DirectColor probably needs ramps hand initialized! */ + *cmap = XCreateColormap(__glutDisplay, __glutRoot, + vi->visual, AllocNone); + break; + case StaticColor: + case StaticGray: + case GrayScale: + /* Mesa supports these visuals */ + *colormap = NULL; + *cmap = XCreateColormap(__glutDisplay, __glutRoot, + vi->visual, AllocNone); + break; + default: + __glutFatalError( + "could not allocate colormap for visual type: %d.", + visualClass); + } + return; +#endif +} + +#if !defined(_WIN32) && !defined(__OS2__) +static int +findColormaps(GLUTwindow * window, + Window * winlist, Colormap * cmaplist, int num, int max) +{ + GLUTwindow *child; + int i; + + /* Do not allow more entries that maximum number of + colormaps! */ + if (num >= max) + return num; + /* Is cmap for this window already on the list? */ + for (i = 0; i < num; i++) { + if (cmaplist[i] == window->cmap) + goto normalColormapAlreadyListed; + } + /* Not found on the list; add colormap and window. */ + winlist[num] = window->win; + cmaplist[num] = window->cmap; + num++; + +normalColormapAlreadyListed: + + /* Repeat above but for the overlay colormap if there one. */ + if (window->overlay) { + if (num >= max) + return num; + for (i = 0; i < num; i++) { + if (cmaplist[i] == window->overlay->cmap) + goto overlayColormapAlreadyListed; + } + winlist[num] = window->overlay->win; + cmaplist[num] = window->overlay->cmap; + num++; + } +overlayColormapAlreadyListed: + + /* Recursively search children. */ + child = window->children; + while (child) { + num = findColormaps(child, winlist, cmaplist, num, max); + child = child->siblings; + } + return num; +} + +void +__glutEstablishColormapsProperty(GLUTwindow * window) +{ + /* this routine is strictly X. Win32 doesn't need to do + anything of this sort (but has to do other wacky stuff + later). */ + static Atom wmColormapWindows = None; + Window *winlist; + Colormap *cmaplist; + Status status; + int maxcmaps, num; + + assert(!window->parent); + maxcmaps = MaxCmapsOfScreen(ScreenOfDisplay(__glutDisplay, + __glutScreen)); + /* For portability reasons we don't use alloca for winlist + and cmaplist, but we could. */ + winlist = (Window *) malloc(maxcmaps * sizeof(Window)); + cmaplist = (Colormap *) malloc(maxcmaps * sizeof(Colormap)); + num = findColormaps(window, winlist, cmaplist, 0, maxcmaps); + if (num < 2) { + /* Property no longer needed; remove it. */ + wmColormapWindows = XSGIFastInternAtom(__glutDisplay, + "WM_COLORMAP_WINDOWS", SGI_XA_WM_COLORMAP_WINDOWS, False); + if (wmColormapWindows == None) { + __glutWarning("Could not intern X atom for WM_COLORMAP_WINDOWS."); + return; + } + XDeleteProperty(__glutDisplay, window->win, wmColormapWindows); + } else { + status = XSetWMColormapWindows(__glutDisplay, window->win, + winlist, num); + /* XSetWMColormapWindows should always work unless the + WM_COLORMAP_WINDOWS property cannot be intern'ed. We + check to be safe. */ + if (status == False) + __glutFatalError("XSetWMColormapWindows returned False."); + } + /* For portability reasons we don't use alloca for winlist + and cmaplist, but we could. */ + free(winlist); + free(cmaplist); +} + +GLUTwindow * +__glutToplevelOf(GLUTwindow * window) +{ + while (window->parent) { + window = window->parent; + } + return window; +} +#endif + +void +__glutFreeColormap(GLUTcolormap * cmap) +{ + GLUTcolormap *cur, **prev; + + cmap->refcnt--; + if (cmap->refcnt == 0) { + /* remove from colormap list */ + cur = __glutColormapList; + prev = &__glutColormapList; + while (cur) { + if (cur == cmap) { + *prev = cmap->next; + break; + } + prev = &(cur->next); + cur = cur->next; + } + /* actually free colormap */ + XFreeColormap(__glutDisplay, cmap->cmap); + free(cmap->cells); + free(cmap); + } +} + diff --git a/src/glut/os2/glut_cursor.cpp b/src/glut/os2/glut_cursor.cpp index 9784e9b622..0db0aa64a4 100644 --- a/src/glut/os2/glut_cursor.cpp +++ b/src/glut/os2/glut_cursor.cpp @@ -1,210 +1,210 @@ - -/* Copyright (c) Mark J. Kilgard, 1995, 1998. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include "glutint.h" - -#if !defined(_WIN32) && !defined(__OS2PM__) -#include /* For XA_CURSOR */ -#include -#endif - -typedef struct _CursorTable { -#if defined(_WIN32) - char* glyph; -#else - int glyph; -#endif - Cursor cursor; -} CursorTable; -/* *INDENT-OFF* */ - -static CursorTable cursorTable[] = { - {XC_arrow, None}, /* GLUT_CURSOR_RIGHT_ARROW */ - {XC_top_left_arrow, None}, /* GLUT_CURSOR_LEFT_ARROW */ - {XC_hand1, None}, /* GLUT_CURSOR_INFO */ - {XC_pirate, None}, /* GLUT_CURSOR_DESTROY */ - {XC_question_arrow, None}, /* GLUT_CURSOR_HELP */ - {XC_exchange, None}, /* GLUT_CURSOR_CYCLE */ - {XC_spraycan, None}, /* GLUT_CURSOR_SPRAY */ - {XC_watch, None}, /* GLUT_CURSOR_WAIT */ - {XC_xterm, None}, /* GLUT_CURSOR_TEXT */ - {XC_crosshair, None}, /* GLUT_CURSOR_CROSSHAIR */ - {XC_sb_v_double_arrow, None}, /* GLUT_CURSOR_UP_DOWN */ - {XC_sb_h_double_arrow, None}, /* GLUT_CURSOR_LEFT_RIGHT */ - {XC_top_side, None}, /* GLUT_CURSOR_TOP_SIDE */ - {XC_bottom_side, None}, /* GLUT_CURSOR_BOTTOM_SIDE */ - {XC_left_side, None}, /* GLUT_CURSOR_LEFT_SIDE */ - {XC_right_side, None}, /* GLUT_CURSOR_RIGHT_SIDE */ - {XC_top_left_corner, None}, /* GLUT_CURSOR_TOP_LEFT_CORNER */ - {XC_top_right_corner, None}, /* GLUT_CURSOR_TOP_RIGHT_CORNER */ - {XC_bottom_right_corner, None}, /* GLUT_CURSOR_BOTTOM_RIGHT_CORNER */ - {XC_bottom_left_corner, None}, /* GLUT_CURSOR_BOTTOM_LEFT_CORNER */ -}; -/* *INDENT-ON* */ - -#if !defined(_WIN32) && !defined(__OS2PM__) -static Cursor blankCursor = None; -static Cursor fullCrosshairCusor = None; - -/* SGI X server's support a special property called the - _SGI_CROSSHAIR_CURSOR that when installed as a window's - cursor, becomes a full screen crosshair cursor. SGI - has special cursor generation hardware for this case. */ -static Cursor -getFullCrosshairCursor(void) -{ - Cursor cursor; - Atom crosshairAtom, actualType; - int rc, actualFormat; - unsigned long n, left; - unsigned long *value; - - if (fullCrosshairCusor == None) { - crosshairAtom = XInternAtom(__glutDisplay, - "_SGI_CROSSHAIR_CURSOR", True); - if (crosshairAtom != None) { - value = 0; /* Make compiler happy. */ - rc = XGetWindowProperty(__glutDisplay, __glutRoot, - crosshairAtom, 0, 1, False, XA_CURSOR, &actualType, - &actualFormat, &n, &left, (unsigned char **) &value); - if (rc == Success && actualFormat == 32 && n >= 1) { - cursor = value[0]; - XFree(value); - return cursor; - } - } - } - return XCreateFontCursor(__glutDisplay, XC_crosshair); -} - -/* X11 forces you to create a blank cursor if you want - to disable the cursor. */ -static Cursor -makeBlankCursor(void) -{ - static char data[1] = - {0}; - Cursor cursor; - Pixmap blank; - XColor dummy; - - blank = XCreateBitmapFromData(__glutDisplay, __glutRoot, - data, 1, 1); - if (blank == None) - __glutFatalError("out of memory."); - cursor = XCreatePixmapCursor(__glutDisplay, blank, blank, - &dummy, &dummy, 0, 0); - XFreePixmap(__glutDisplay, blank); - - return cursor; -} -#endif /* !_WIN32 && !__OS2PM__*/ - -/* Win32 and X11 use this same function to accomplish - fairly different tasks. X11 lets you just define the - cursor for a window and the window system takes care - of making sure that the window's cursor is installed - when the mouse is in the window. Win32 requires the - application to handle a WM_SETCURSOR message to install - the right cursor when windows are entered. Think of - the Win32 __glutSetCursor (called from __glutWindowProc) - as "install cursor". Think of the X11 __glutSetCursor - (called from glutSetCursor) as "define cursor". */ -void -__glutSetCursor(GLUTwindow *window) -{ - int cursor = window->cursor; - Cursor xcursor = 0; - - if (cursor >= 0 && - cursor < sizeof(cursorTable) / sizeof(cursorTable[0])) { - if (cursorTable[cursor].cursor == None) { - cursorTable[cursor].cursor = XCreateFontCursor(__glutDisplay, - cursorTable[cursor].glyph); - } - xcursor = cursorTable[cursor].cursor; - } else { - /* Special cases. */ - switch (cursor) { - case GLUT_CURSOR_INHERIT: -#if defined(_WIN32) - while (window->parent) { - window = window->parent; - if (window->cursor != GLUT_CURSOR_INHERIT) { - __glutSetCursor(window); - return; - } - } - /* XXX Default to an arrow cursor. Is this - right or should we be letting the default - window proc be installing some system cursor? */ - xcursor = cursorTable[0].cursor; - if (xcursor == NULL) { - xcursor = - cursorTable[0].cursor = - LoadCursor(NULL, cursorTable[0].glyph); - } - -#elif defined(__OS2PM__) -//todo - xcursor = None; - -#else - xcursor = None; -#endif - break; - case GLUT_CURSOR_NONE: -#if defined(_WIN32) || defined(__OS2PM__) - xcursor = NULL; -#else - if (blankCursor == None) { - blankCursor = makeBlankCursor(); - } - xcursor = blankCursor; -#endif - break; - case GLUT_CURSOR_FULL_CROSSHAIR: -#if defined(_WIN32) - xcursor = (HICON) IDC_CROSS; -#elif defined(__OS2PM__) -//todo -#else - if (fullCrosshairCusor == None) { - fullCrosshairCusor = getFullCrosshairCursor(); - } - xcursor = fullCrosshairCusor; -#endif - break; - } - } - XDefineCursor(__glutDisplay, - window->win, xcursor); - XFlush(__glutDisplay); -} - -/* CENTRY */ -void GLUTAPIENTRY -glutSetCursor(int cursor) -{ -#ifdef _WIN32 - POINT point; - - __glutCurrentWindow->cursor = cursor; - /* Are we in the window right now? If so, - install the cursor. */ - GetCursorPos(&point); - if (__glutCurrentWindow->win == WindowFromPoint(point)) { - __glutSetCursor(__glutCurrentWindow); - } -#elif defined(__OS2PM__) -//todo -#else - __glutCurrentWindow->cursor = cursor; - __glutSetCursor(__glutCurrentWindow); -#endif -} -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1995, 1998. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include "glutint.h" + +#if !defined(_WIN32) && !defined(__OS2PM__) +#include /* For XA_CURSOR */ +#include +#endif + +typedef struct _CursorTable { +#if defined(_WIN32) + char* glyph; +#else + int glyph; +#endif + Cursor cursor; +} CursorTable; +/* *INDENT-OFF* */ + +static CursorTable cursorTable[] = { + {XC_arrow, None}, /* GLUT_CURSOR_RIGHT_ARROW */ + {XC_top_left_arrow, None}, /* GLUT_CURSOR_LEFT_ARROW */ + {XC_hand1, None}, /* GLUT_CURSOR_INFO */ + {XC_pirate, None}, /* GLUT_CURSOR_DESTROY */ + {XC_question_arrow, None}, /* GLUT_CURSOR_HELP */ + {XC_exchange, None}, /* GLUT_CURSOR_CYCLE */ + {XC_spraycan, None}, /* GLUT_CURSOR_SPRAY */ + {XC_watch, None}, /* GLUT_CURSOR_WAIT */ + {XC_xterm, None}, /* GLUT_CURSOR_TEXT */ + {XC_crosshair, None}, /* GLUT_CURSOR_CROSSHAIR */ + {XC_sb_v_double_arrow, None}, /* GLUT_CURSOR_UP_DOWN */ + {XC_sb_h_double_arrow, None}, /* GLUT_CURSOR_LEFT_RIGHT */ + {XC_top_side, None}, /* GLUT_CURSOR_TOP_SIDE */ + {XC_bottom_side, None}, /* GLUT_CURSOR_BOTTOM_SIDE */ + {XC_left_side, None}, /* GLUT_CURSOR_LEFT_SIDE */ + {XC_right_side, None}, /* GLUT_CURSOR_RIGHT_SIDE */ + {XC_top_left_corner, None}, /* GLUT_CURSOR_TOP_LEFT_CORNER */ + {XC_top_right_corner, None}, /* GLUT_CURSOR_TOP_RIGHT_CORNER */ + {XC_bottom_right_corner, None}, /* GLUT_CURSOR_BOTTOM_RIGHT_CORNER */ + {XC_bottom_left_corner, None}, /* GLUT_CURSOR_BOTTOM_LEFT_CORNER */ +}; +/* *INDENT-ON* */ + +#if !defined(_WIN32) && !defined(__OS2PM__) +static Cursor blankCursor = None; +static Cursor fullCrosshairCusor = None; + +/* SGI X server's support a special property called the + _SGI_CROSSHAIR_CURSOR that when installed as a window's + cursor, becomes a full screen crosshair cursor. SGI + has special cursor generation hardware for this case. */ +static Cursor +getFullCrosshairCursor(void) +{ + Cursor cursor; + Atom crosshairAtom, actualType; + int rc, actualFormat; + unsigned long n, left; + unsigned long *value; + + if (fullCrosshairCusor == None) { + crosshairAtom = XInternAtom(__glutDisplay, + "_SGI_CROSSHAIR_CURSOR", True); + if (crosshairAtom != None) { + value = 0; /* Make compiler happy. */ + rc = XGetWindowProperty(__glutDisplay, __glutRoot, + crosshairAtom, 0, 1, False, XA_CURSOR, &actualType, + &actualFormat, &n, &left, (unsigned char **) &value); + if (rc == Success && actualFormat == 32 && n >= 1) { + cursor = value[0]; + XFree(value); + return cursor; + } + } + } + return XCreateFontCursor(__glutDisplay, XC_crosshair); +} + +/* X11 forces you to create a blank cursor if you want + to disable the cursor. */ +static Cursor +makeBlankCursor(void) +{ + static char data[1] = + {0}; + Cursor cursor; + Pixmap blank; + XColor dummy; + + blank = XCreateBitmapFromData(__glutDisplay, __glutRoot, + data, 1, 1); + if (blank == None) + __glutFatalError("out of memory."); + cursor = XCreatePixmapCursor(__glutDisplay, blank, blank, + &dummy, &dummy, 0, 0); + XFreePixmap(__glutDisplay, blank); + + return cursor; +} +#endif /* !_WIN32 && !__OS2PM__*/ + +/* Win32 and X11 use this same function to accomplish + fairly different tasks. X11 lets you just define the + cursor for a window and the window system takes care + of making sure that the window's cursor is installed + when the mouse is in the window. Win32 requires the + application to handle a WM_SETCURSOR message to install + the right cursor when windows are entered. Think of + the Win32 __glutSetCursor (called from __glutWindowProc) + as "install cursor". Think of the X11 __glutSetCursor + (called from glutSetCursor) as "define cursor". */ +void +__glutSetCursor(GLUTwindow *window) +{ + int cursor = window->cursor; + Cursor xcursor = 0; + + if (cursor >= 0 && + cursor < sizeof(cursorTable) / sizeof(cursorTable[0])) { + if (cursorTable[cursor].cursor == None) { + cursorTable[cursor].cursor = XCreateFontCursor(__glutDisplay, + cursorTable[cursor].glyph); + } + xcursor = cursorTable[cursor].cursor; + } else { + /* Special cases. */ + switch (cursor) { + case GLUT_CURSOR_INHERIT: +#if defined(_WIN32) + while (window->parent) { + window = window->parent; + if (window->cursor != GLUT_CURSOR_INHERIT) { + __glutSetCursor(window); + return; + } + } + /* XXX Default to an arrow cursor. Is this + right or should we be letting the default + window proc be installing some system cursor? */ + xcursor = cursorTable[0].cursor; + if (xcursor == NULL) { + xcursor = + cursorTable[0].cursor = + LoadCursor(NULL, cursorTable[0].glyph); + } + +#elif defined(__OS2PM__) +//todo + xcursor = None; + +#else + xcursor = None; +#endif + break; + case GLUT_CURSOR_NONE: +#if defined(_WIN32) || defined(__OS2PM__) + xcursor = NULL; +#else + if (blankCursor == None) { + blankCursor = makeBlankCursor(); + } + xcursor = blankCursor; +#endif + break; + case GLUT_CURSOR_FULL_CROSSHAIR: +#if defined(_WIN32) + xcursor = (HICON) IDC_CROSS; +#elif defined(__OS2PM__) +//todo +#else + if (fullCrosshairCusor == None) { + fullCrosshairCusor = getFullCrosshairCursor(); + } + xcursor = fullCrosshairCusor; +#endif + break; + } + } + XDefineCursor(__glutDisplay, + window->win, xcursor); + XFlush(__glutDisplay); +} + +/* CENTRY */ +void GLUTAPIENTRY +glutSetCursor(int cursor) +{ +#ifdef _WIN32 + POINT point; + + __glutCurrentWindow->cursor = cursor; + /* Are we in the window right now? If so, + install the cursor. */ + GetCursorPos(&point); + if (__glutCurrentWindow->win == WindowFromPoint(point)) { + __glutSetCursor(__glutCurrentWindow); + } +#elif defined(__OS2PM__) +//todo +#else + __glutCurrentWindow->cursor = cursor; + __glutSetCursor(__glutCurrentWindow); +#endif +} +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_event.cpp b/src/glut/os2/glut_event.cpp index c556ef2826..4ece82e52b 100644 --- a/src/glut/os2/glut_event.cpp +++ b/src/glut/os2/glut_event.cpp @@ -1,1399 +1,1399 @@ -/* glut_event.c */ -/* Copyright (c) Mark J. Kilgard, 1994, 1995, 1996, 1997, 1998. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#ifdef __VMS -#include -#endif - -#include -#include -#include -#include -#include /* Some FD_ZERO macros use memset without - prototyping memset. */ - -/* Much of the following #ifdef logic to include the proper - prototypes for the select system call is based on logic - from the X11R6.3 version of . */ - -#if !defined(_WIN32) -# ifdef __sgi -# include /* prototype for bzero used by FD_ZERO */ -# endif -# if (defined(SVR4) || defined(CRAY) || defined(AIXV3)) && !defined(FD_SETSIZE) -# include /* select system call interface */ -# ifdef luna -# include -# endif -# endif - /* AIX 4.2 fubar-ed , so go to heroic measures to get it */ -# if defined(AIXV4) && !defined(NFDBITS) -# include -# endif -#endif /* !_WIN32 */ - -#include - -#if defined(__OS2__) -//??? ­ ä¨£  ?? # include -#elif !defined(_WIN32) -# if defined(__vms) && ( __VMS_VER < 70000000 ) -# include -# else -# ifndef __vms -# include -# endif -# endif -# include -# include -# include -#else -# ifdef __CYGWIN32__ -# include -# else -# include -# endif -# ifdef __hpux - /* XXX Bert Gijsbers reports that HP-UX - needs different keysyms for the End, Insert, and Delete keys - to work on an HP 715. It would be better if HP generated - standard keysyms for standard keys. */ -# include -# endif -#endif /* !_WIN32 */ - -#include "glutint.h" - -#if defined(__vms) && ( __VMS_VER < 70000000 ) -#include -#include -extern int SYS$CLREF(int efn); -extern int SYS$SETIMR(unsigned int efn, struct timeval6 *timeout, void *ast, - unsigned int request_id, unsigned int flags); -extern int SYS$WFLOR(unsigned int efn, unsigned int mask); -extern int SYS$CANTIM(unsigned int request_id, unsigned int mode); -#endif /* __vms, VMs 6.2 or earlier */ - -static GLUTtimer *freeTimerList = NULL; - -GLUTidleCB __glutIdleFunc = NULL; -GLUTtimer *__glutTimerList = NULL; -#ifdef SUPPORT_FORTRAN -GLUTtimer *__glutNewTimer; -#endif -GLUTwindow *__glutWindowWorkList = NULL; -GLUTmenu *__glutMappedMenu; -GLUTmenu *__glutCurrentMenu = NULL; - -void (*__glutUpdateInputDeviceMaskFunc) (GLUTwindow *); -#if !defined(_WIN32) && !defined(__OS2__) -void (*__glutMenuItemEnterOrLeave)(GLUTmenuItem * item, int num, int type) = NULL; -void (*__glutFinishMenu)(Window win, int x, int y); -void (*__glutPaintMenu)(GLUTmenu * menu); -void (*__glutStartMenu)(GLUTmenu * menu, GLUTwindow * window, int x, int y, int x_win, int y_win); -GLUTmenu * (*__glutGetMenuByNum)(int menunum); -GLUTmenuItem * (*__glutGetMenuItem)(GLUTmenu * menu, Window win, int *which); -GLUTmenu * (*__glutGetMenu)(Window win); -#endif - -Atom __glutMotifHints = None; -/* Modifier mask of ~0 implies not in core input callback. */ -unsigned int __glutModifierMask = (unsigned int) ~0; -int __glutWindowDamaged = 0; - -void GLUTAPIENTRY -glutIdleFunc(GLUTidleCB idleFunc) -{ - __glutIdleFunc = idleFunc; -} - -void GLUTAPIENTRY -glutTimerFunc(unsigned int interval, GLUTtimerCB timerFunc, int value) -{ - GLUTtimer *timer, *other; - GLUTtimer **prevptr; -#ifdef OLD_VMS - struct timeval6 now; -#else - struct timeval now; -#endif - - if (!timerFunc) - return; - - if (freeTimerList) { - timer = freeTimerList; - freeTimerList = timer->next; - } else { - timer = (GLUTtimer *) malloc(sizeof(GLUTtimer)); - if (!timer) - __glutFatalError("out of memory."); - } - - timer->func = timerFunc; -#if defined(__vms) && ( __VMS_VER < 70000000 ) - /* VMS time is expressed in units of 100 ns */ - timer->timeout.val = interval * TICKS_PER_MILLISECOND; -#else - timer->timeout.tv_sec = (int) interval / 1000; - timer->timeout.tv_usec = (int) (interval % 1000) * 1000; -#endif - timer->value = value; - timer->next = NULL; - GETTIMEOFDAY(&now); - ADD_TIME(timer->timeout, timer->timeout, now); - prevptr = &__glutTimerList; - other = *prevptr; - while (other && IS_AFTER(other->timeout, timer->timeout)) { - prevptr = &other->next; - other = *prevptr; - } - timer->next = other; -#ifdef SUPPORT_FORTRAN - __glutNewTimer = timer; /* for Fortran binding! */ -#endif - *prevptr = timer; -} - -void -handleTimeouts(void) -{ -#ifdef OLD_VMS - struct timeval6 now; -#else - struct timeval now; -#endif - GLUTtimer *timer; - - /* Assumption is that __glutTimerList is already determined - to be non-NULL. */ - GETTIMEOFDAY(&now); - while (IS_AT_OR_AFTER(__glutTimerList->timeout, now)) { - timer = __glutTimerList; - timer->func(timer->value); - __glutTimerList = timer->next; - timer->next = freeTimerList; - freeTimerList = timer; - if (!__glutTimerList) - break; - } -} - -void -__glutPutOnWorkList(GLUTwindow * window, int workMask) -{ - if (window->workMask) { - /* Already on list; just OR in new workMask. */ - window->workMask |= workMask; - } else { - /* Update work mask and add to window work list. */ - window->workMask = workMask; - /* Assert that if the window does not have a - workMask already, the window should definitely - not be the head of the work list. */ - assert(window != __glutWindowWorkList); - window->prevWorkWin = __glutWindowWorkList; - __glutWindowWorkList = window; - } -} - -void -__glutPostRedisplay(GLUTwindow * window, int layerMask) -{ - int shown = (layerMask & (GLUT_REDISPLAY_WORK | GLUT_REPAIR_WORK)) ? - window->shownState : window->overlay->shownState; - - /* Post a redisplay if the window is visible (or the - visibility of the window is unknown, ie. window->visState - == -1) _and_ the layer is known to be shown. */ - if (window->visState != GLUT_HIDDEN - && window->visState != GLUT_FULLY_COVERED && shown) { - __glutPutOnWorkList(window, layerMask); - } -} - -/* CENTRY */ -void GLUTAPIENTRY -glutPostRedisplay(void) -{ - __glutPostRedisplay(__glutCurrentWindow, GLUT_REDISPLAY_WORK); -} - -/* The advantage of this routine is that it saves the cost of a - glutSetWindow call (entailing an expensive OpenGL context switch), - particularly useful when multiple windows need redisplays posted at - the same times. See also glutPostWindowOverlayRedisplay. */ -void GLUTAPIENTRY -glutPostWindowRedisplay(int win) -{ - __glutPostRedisplay(__glutWindowList[win - 1], GLUT_REDISPLAY_WORK); -} - -/* ENDCENTRY */ -static GLUTeventParser *eventParserList = NULL; - -/* __glutRegisterEventParser allows another module to register - to intercept X events types not otherwise acted on by the - GLUT processEventsAndTimeouts routine. The X Input - extension support code uses an event parser for handling X - Input extension events. */ - -void -__glutRegisterEventParser(GLUTeventParser * parser) -{ - parser->next = eventParserList; - eventParserList = parser; -} - -static void -markWindowHidden(GLUTwindow * window) -{ - if (GLUT_HIDDEN != window->visState) { - GLUTwindow *child; - - if (window->windowStatus) { - window->visState = GLUT_HIDDEN; - __glutSetWindow(window); - window->windowStatus(GLUT_HIDDEN); - } - /* An unmap is only reported on a single window; its - descendents need to know they are no longer visible. */ - child = window->children; - while (child) { - markWindowHidden(child); - child = child->siblings; - } - } -} - -#if !defined(_WIN32) && !defined(__OS2__) - -static void -purgeStaleWindow(Window win) -{ - GLUTstale **pEntry = &__glutStaleWindowList; - GLUTstale *entry = __glutStaleWindowList; - - /* Tranverse singly-linked stale window list look for the - window ID. */ - while (entry) { - if (entry->win == win) { - /* Found it; delete it. */ - *pEntry = entry->next; - free(entry); - return; - } else { - pEntry = &entry->next; - entry = *pEntry; - } - } -} - -/* Unlike XNextEvent, if a signal arrives, - interruptibleXNextEvent will return (with a zero return - value). This helps GLUT drop out of XNextEvent if a signal - is delivered. The intent is so that a GLUT program can call - glutIdleFunc in a signal handler to register an idle func - and then immediately get dropped into the idle func (after - returning from the signal handler). The idea is to make - GLUT's main loop reliably interruptible by signals. */ -static int -interruptibleXNextEvent(Display * dpy, XEvent * event) -{ - fd_set fds; - int rc; - - /* Flush X protocol since XPending does not do this - implicitly. */ - XFlush(__glutDisplay); - for (;;) { - if (XPending(__glutDisplay)) { - XNextEvent(dpy, event); - return 1; - } -#ifndef VMS - /* the combination ConectionNumber-select is buggy on VMS. Sometimes it - * fails. This part of the code hangs the program on VMS7.2. But even - * without it the program seems to run correctly. - * Note that this is a bug in the VMS/DECWindows run-time-libraries. - * Compaq engeneering does not want or is not able to make a fix. - * (last sentence is a quotation from Compaq when I reported the - * problem January 2000) */ - FD_ZERO(&fds); - FD_SET(__glutConnectionFD, &fds); - rc = select(__glutConnectionFD + 1, &fds, NULL, NULL, NULL); - if (rc < 0) { - if (errno == EINTR) { - return 0; - } else { - __glutFatalError("select error."); - } - } -#endif - } -} - -#endif - -static void -processEventsAndTimeouts(void) -{ - do { -#if defined(__OS2__) - QMSG qmsg; /* message from message queue */ - extern HAB hab; /* PM anchor block handle */ - - if(! WinGetMsg( hab, &qmsg, 0UL, 0UL, 0UL ) ) - exit(0); - WinDispatchMsg( hab, /* PM anchor block handle */ - &qmsg ); /* pointer to message */ - -#elif defined(_WIN32) - MSG event; - - if(!GetMessage(&event, NULL, 0, 0)) /* bail if no more messages */ - exit(0); - TranslateMessage(&event); /* translate virtual-key messages */ - DispatchMessage(&event); /* call the window proc */ - /* see win32_event.c for event (message) processing procedures */ -#else - static int mappedMenuButton; - GLUTeventParser *parser; - XEvent event, ahead; - GLUTwindow *window; - GLUTkeyboardCB keyboard; - GLUTspecialCB special; - int gotEvent, width, height; - - gotEvent = interruptibleXNextEvent(__glutDisplay, &event); - if (gotEvent) { - switch (event.type) { - case MappingNotify: - XRefreshKeyboardMapping((XMappingEvent *) & event); - break; - case ConfigureNotify: - window = __glutGetWindow(event.xconfigure.window); - if (window) { - if (window->win != event.xconfigure.window) { - /* Ignore ConfigureNotify sent to the overlay - planes. GLUT could get here because overlays - select for StructureNotify events to receive - DestroyNotify. */ - break; - } - width = event.xconfigure.width; - height = event.xconfigure.height; - if (width != window->width || height != window->height) { - if (window->overlay) { - XResizeWindow(__glutDisplay, window->overlay->win, width, height); - } - window->width = width; - window->height = height; - __glutSetWindow(window); - /* Do not execute OpenGL out of sequence with - respect to the XResizeWindow request! */ - glXWaitX(); - window->reshape(width, height); - window->forceReshape = False; - /* A reshape should be considered like posting a - repair; this is necessary for the "Mesa - glXSwapBuffers to repair damage" hack to operate - correctly. Without it, there's not an initial - back buffer render from which to blit from when - damage happens to the window. */ - __glutPostRedisplay(window, GLUT_REPAIR_WORK); - } - } - break; - case Expose: - /* compress expose events */ - while (XEventsQueued(__glutDisplay, QueuedAfterReading) - > 0) { - XPeekEvent(__glutDisplay, &ahead); - if (ahead.type != Expose || - ahead.xexpose.window != event.xexpose.window) { - break; - } - XNextEvent(__glutDisplay, &event); - } - if (event.xexpose.count == 0) { - GLUTmenu *menu; - - if (__glutMappedMenu && - (menu = __glutGetMenu(event.xexpose.window))) { - __glutPaintMenu(menu); - } else { - window = __glutGetWindow(event.xexpose.window); - if (window) { - if (window->win == event.xexpose.window) { - __glutPostRedisplay(window, GLUT_REPAIR_WORK); - } else if (window->overlay && window->overlay->win == event.xexpose.window) { - __glutPostRedisplay(window, GLUT_OVERLAY_REPAIR_WORK); - } - } - } - } else { - /* there are more exposes to read; wait to redisplay */ - } - break; - case ButtonPress: - case ButtonRelease: - if (__glutMappedMenu && event.type == ButtonRelease - && mappedMenuButton == event.xbutton.button) { - /* Menu is currently popped up and its button is - released. */ - __glutFinishMenu(event.xbutton.window, event.xbutton.x, event.xbutton.y); - } else { - window = __glutGetWindow(event.xbutton.window); - if (window) { - GLUTmenu *menu; - int menuNum; - - menuNum = window->menu[event.xbutton.button - 1]; - /* Make sure that __glutGetMenuByNum is only called if there - really is a menu present. */ - if ((menuNum > 0) && (menu = __glutGetMenuByNum(menuNum))) { - if (event.type == ButtonPress && !__glutMappedMenu) { - __glutStartMenu(menu, window, - event.xbutton.x_root, event.xbutton.y_root, - event.xbutton.x, event.xbutton.y); - mappedMenuButton = event.xbutton.button; - } else { - /* Ignore a release of a button with a menu - attatched to it when no menu is popped up, - or ignore a press when another menu is - already popped up. */ - } - } else if (window->mouse) { - __glutSetWindow(window); - __glutModifierMask = event.xbutton.state; - window->mouse(event.xbutton.button - 1, - event.type == ButtonRelease ? - GLUT_UP : GLUT_DOWN, - event.xbutton.x, event.xbutton.y); - __glutModifierMask = ~0; - } else { - /* Stray mouse events. Ignore. */ - } - } else { - /* Window might have been destroyed and all the - events for the window may not yet be received. */ - } - } - break; - case MotionNotify: - if (!__glutMappedMenu) { - window = __glutGetWindow(event.xmotion.window); - if (window) { - /* If motion function registered _and_ buttons held - * down, call motion function... */ - if (window->motion && event.xmotion.state & - (Button1Mask | Button2Mask | Button3Mask)) { - __glutSetWindow(window); - window->motion(event.xmotion.x, event.xmotion.y); - } - /* If passive motion function registered _and_ - buttons not held down, call passive motion - function... */ - else if (window->passive && - ((event.xmotion.state & - (Button1Mask | Button2Mask | Button3Mask)) == - 0)) { - __glutSetWindow(window); - window->passive(event.xmotion.x, - event.xmotion.y); - } - } - } else { - /* Motion events are thrown away when a pop up menu - is active. */ - } - break; - case KeyPress: - case KeyRelease: - window = __glutGetWindow(event.xkey.window); - if (!window) { - break; - } - if (event.type == KeyPress) { - keyboard = window->keyboard; - } else { - - /* If we are ignoring auto repeated keys for this window, - check if the next event in the X event queue is a KeyPress - for the exact same key (and at the exact same time) as the - key being released. The X11 protocol will send auto - repeated keys as such KeyRelease/KeyPress pairs. */ - - if (window->ignoreKeyRepeat) { - if (XEventsQueued(__glutDisplay, QueuedAfterReading)) { - XPeekEvent(__glutDisplay, &ahead); - if (ahead.type == KeyPress - && ahead.xkey.window == event.xkey.window - && ahead.xkey.keycode == event.xkey.keycode - && ahead.xkey.time == event.xkey.time) { - /* Pop off the repeated KeyPress and ignore - the auto repeated KeyRelease/KeyPress pair. */ - XNextEvent(__glutDisplay, &event); - break; - } - } - } - keyboard = window->keyboardUp; - } - if (keyboard) { - char tmp[1]; - int rc; - - rc = XLookupString(&event.xkey, tmp, sizeof(tmp), - NULL, NULL); - if (rc) { - __glutSetWindow(window); - __glutModifierMask = event.xkey.state; - keyboard(tmp[0], - event.xkey.x, event.xkey.y); - __glutModifierMask = ~0; - break; - } - } - if (event.type == KeyPress) { - special = window->special; - } else { - special = window->specialUp; - } - if (special) { - KeySym ks; - int key; - -/* Introduced in X11R6: (Partial list of) Keypad Functions. Define - in place in case compiling against an older pre-X11R6 - X11/keysymdef.h file. */ -#ifndef XK_KP_Home -#define XK_KP_Home 0xFF95 -#endif -#ifndef XK_KP_Left -#define XK_KP_Left 0xFF96 -#endif -#ifndef XK_KP_Up -#define XK_KP_Up 0xFF97 -#endif -#ifndef XK_KP_Right -#define XK_KP_Right 0xFF98 -#endif -#ifndef XK_KP_Down -#define XK_KP_Down 0xFF99 -#endif -#ifndef XK_KP_Prior -#define XK_KP_Prior 0xFF9A -#endif -#ifndef XK_KP_Next -#define XK_KP_Next 0xFF9B -#endif -#ifndef XK_KP_End -#define XK_KP_End 0xFF9C -#endif -#ifndef XK_KP_Insert -#define XK_KP_Insert 0xFF9E -#endif -#ifndef XK_KP_Delete -#define XK_KP_Delete 0xFF9F -#endif - - ks = XLookupKeysym((XKeyEvent *) & event, 0); - /* XXX Verbose, but makes no assumptions about keysym - layout. */ - switch (ks) { -/* *INDENT-OFF* */ - /* function keys */ - case XK_F1: key = GLUT_KEY_F1; break; - case XK_F2: key = GLUT_KEY_F2; break; - case XK_F3: key = GLUT_KEY_F3; break; - case XK_F4: key = GLUT_KEY_F4; break; - case XK_F5: key = GLUT_KEY_F5; break; - case XK_F6: key = GLUT_KEY_F6; break; - case XK_F7: key = GLUT_KEY_F7; break; - case XK_F8: key = GLUT_KEY_F8; break; - case XK_F9: key = GLUT_KEY_F9; break; - case XK_F10: key = GLUT_KEY_F10; break; - case XK_F11: key = GLUT_KEY_F11; break; - case XK_F12: key = GLUT_KEY_F12; break; - /* directional keys */ - case XK_KP_Left: - case XK_Left: key = GLUT_KEY_LEFT; break; - case XK_KP_Up: /* Introduced in X11R6. */ - case XK_Up: key = GLUT_KEY_UP; break; - case XK_KP_Right: /* Introduced in X11R6. */ - case XK_Right: key = GLUT_KEY_RIGHT; break; - case XK_KP_Down: /* Introduced in X11R6. */ - case XK_Down: key = GLUT_KEY_DOWN; break; -/* *INDENT-ON* */ - - case XK_KP_Prior: /* Introduced in X11R6. */ - case XK_Prior: - /* XK_Prior same as X11R6's XK_Page_Up */ - key = GLUT_KEY_PAGE_UP; - break; - case XK_KP_Next: /* Introduced in X11R6. */ - case XK_Next: - /* XK_Next same as X11R6's XK_Page_Down */ - key = GLUT_KEY_PAGE_DOWN; - break; - case XK_KP_Home: /* Introduced in X11R6. */ - case XK_Home: - key = GLUT_KEY_HOME; - break; -#ifdef __hpux - case XK_Select: -#endif - case XK_KP_End: /* Introduced in X11R6. */ - case XK_End: - key = GLUT_KEY_END; - break; -#ifdef __hpux - case XK_InsertChar: -#endif - case XK_KP_Insert: /* Introduced in X11R6. */ - case XK_Insert: - key = GLUT_KEY_INSERT; - break; -#ifdef __hpux - case XK_DeleteChar: -#endif - case XK_KP_Delete: /* Introduced in X11R6. */ - /* The Delete character is really an ASCII key. */ - __glutSetWindow(window); - keyboard(127, /* ASCII Delete character. */ - event.xkey.x, event.xkey.y); - goto skip; - default: - goto skip; - } - __glutSetWindow(window); - __glutModifierMask = event.xkey.state; - special(key, event.xkey.x, event.xkey.y); - __glutModifierMask = ~0; - skip:; - } - break; - case EnterNotify: - case LeaveNotify: - if (event.xcrossing.mode != NotifyNormal || - event.xcrossing.detail == NotifyNonlinearVirtual || - event.xcrossing.detail == NotifyVirtual) { - - /* Careful to ignore Enter/LeaveNotify events that - come from the pop-up menu pointer grab and ungrab. - Also, ignore "virtual" Enter/LeaveNotify events - since they represent the pointer passing through - the window hierarchy without actually entering or - leaving the actual real estate of a window. */ - - break; - } - if (__glutMappedMenu) { - GLUTmenuItem *item; - int num; - - item = __glutGetMenuItem(__glutMappedMenu, - event.xcrossing.window, &num); - if (item) { - __glutMenuItemEnterOrLeave(item, num, event.type); - break; - } - } - window = __glutGetWindow(event.xcrossing.window); - if (window) { - if (window->entry) { - if (event.type == EnterNotify) { - - /* With overlays established, X can report two - enter events for both the overlay and normal - plane window. Do not generate a second enter - callback if we reported one without an - intervening leave. */ - - if (window->entryState != EnterNotify) { - int num = window->num; - Window xid = window->win; - - window->entryState = EnterNotify; - __glutSetWindow(window); - window->entry(GLUT_ENTERED); - - if (__glutMappedMenu) { - - /* Do not generate any passive motion events - when menus are in use. */ - - } else { - - /* An EnterNotify event can result in a - "compound" callback if a passive motion - callback is also registered. In this case, - be a little paranoid about the possibility - the window could have been destroyed in the - entry callback. */ - - window = __glutWindowList[num]; - if (window && window->passive && window->win == xid) { - __glutSetWindow(window); - window->passive(event.xcrossing.x, event.xcrossing.y); - } - } - } - } else { - if (window->entryState != LeaveNotify) { - - /* When an overlay is established for a window - already mapped and with the pointer in it, - the X server will generate a leave/enter - event pair as the pointer leaves (without - moving) from the normal plane X window to - the newly mapped overlay X window (or vice - versa). This enter/leave pair should not be - reported to the GLUT program since the pair - is a consequence of creating (or destroying) - the overlay, not an actual leave from the - GLUT window. */ - - if (XEventsQueued(__glutDisplay, QueuedAfterReading)) { - XPeekEvent(__glutDisplay, &ahead); - if (ahead.type == EnterNotify && - __glutGetWindow(ahead.xcrossing.window) == window) { - XNextEvent(__glutDisplay, &event); - break; - } - } - window->entryState = LeaveNotify; - __glutSetWindow(window); - window->entry(GLUT_LEFT); - } - } - } else if (window->passive) { - __glutSetWindow(window); - window->passive(event.xcrossing.x, event.xcrossing.y); - } - } - break; - case UnmapNotify: - /* MapNotify events are not needed to maintain - visibility state since VisibilityNotify events will - be delivered when a window becomes visible from - mapping. However, VisibilityNotify events are not - delivered when a window is unmapped (for the window - or its children). */ - window = __glutGetWindow(event.xunmap.window); - if (window) { - if (window->win != event.xconfigure.window) { - /* Ignore UnmapNotify sent to the overlay planes. - GLUT could get here because overlays select for - StructureNotify events to receive DestroyNotify. - */ - break; - } - markWindowHidden(window); - } - break; - case VisibilityNotify: - window = __glutGetWindow(event.xvisibility.window); - if (window) { - /* VisibilityUnobscured+1 = GLUT_FULLY_RETAINED, - VisibilityPartiallyObscured+1 = - GLUT_PARTIALLY_RETAINED, VisibilityFullyObscured+1 - = GLUT_FULLY_COVERED. */ - int visState = event.xvisibility.state + 1; - - if (visState != window->visState) { - if (window->windowStatus) { - window->visState = visState; - __glutSetWindow(window); - window->windowStatus(visState); - } - } - } - break; - case ClientMessage: - if (event.xclient.data.l[0] == __glutWMDeleteWindow) - exit(0); - break; - case DestroyNotify: - purgeStaleWindow(event.xdestroywindow.window); - break; - case CirculateNotify: - case CreateNotify: - case GravityNotify: - case ReparentNotify: - /* Uninteresting to GLUT (but possible for GLUT to - receive). */ - break; - default: - /* Pass events not directly handled by the GLUT main - event loop to any event parsers that have been - registered. In this way, X Input extension events - are passed to the correct handler without forcing - all GLUT programs to support X Input event handling. - */ - parser = eventParserList; - while (parser) { - if (parser->func(&event)) - break; - parser = parser->next; - } - break; - } - } -#endif /* _WIN32 */ - if (__glutTimerList) { - handleTimeouts(); - } - } - while (XPending(__glutDisplay)); -} - -static void -waitForSomething(void) -{ -#if defined(__vms) && ( __VMS_VER < 70000000 ) - static struct timeval6 zerotime = - {0}; - unsigned int timer_efn; -#define timer_id 'glut' /* random :-) number */ - unsigned int wait_mask; -#else - static struct timeval zerotime = - {0, 0}; -#if defined(__OS2__) - -#elif !defined(_WIN32) - fd_set fds; -#endif -#endif -#ifdef OLD_VMS - struct timeval6 now, timeout, waittime; -#else - struct timeval now, timeout, waittime; -#endif -#if !defined(_WIN32) - int rc; -#endif - - /* Flush X protocol since XPending does not do this - implicitly. */ - XFlush(__glutDisplay); - if (XPending(__glutDisplay)) { - /* It is possible (but quite rare) that XFlush may have - needed to wait for a writable X connection file - descriptor, and in the process, may have had to read off - X protocol from the file descriptor. If XPending is true, - this case occured and we should avoid waiting in select - since X protocol buffered within Xlib is due to be - processed and potentially no more X protocol is on the - file descriptor, so we would risk waiting improperly in - select. */ - goto immediatelyHandleXinput; - } -#if defined(__vms) && ( __VMS_VER < 70000000 ) - timeout = __glutTimerList->timeout; - GETTIMEOFDAY(&now); - wait_mask = 1 << (__glutConnectionFD & 31); - if (IS_AFTER(now, timeout)) { - /* We need an event flag for the timer. */ - /* XXX The `right' way to do this is to use LIB$GET_EF, but - since it needs to be in the same cluster as the EFN for - the display, we will have hack it. */ - timer_efn = __glutConnectionFD - 1; - if ((timer_efn / 32) != (__glutConnectionFD / 32)) { - timer_efn = __glutConnectionFD + 1; - } - rc = SYS$CLREF(timer_efn); - rc = SYS$SETIMR(timer_efn, &timeout, NULL, timer_id, 0); - wait_mask |= 1 << (timer_efn & 31); - } else { - timer_efn = 0; - } - rc = SYS$WFLOR(__glutConnectionFD, wait_mask); - if (timer_efn != 0 && SYS$CLREF(timer_efn) == SS$_WASCLR) { - rc = SYS$CANTIM(timer_id, PSL$C_USER); - } - /* XXX There does not seem to be checking of "rc" in the code - above. Can any of the SYS$ routines above fail? */ -#else /* not vms6.2 or lower */ -#if defined(__OS2__) - -#elif !defined(_WIN32) - FD_ZERO(&fds); - FD_SET(__glutConnectionFD, &fds); -#endif - timeout = __glutTimerList->timeout; - GETTIMEOFDAY(&now); - if (IS_AFTER(now, timeout)) { - TIMEDELTA(waittime, timeout, now); - } else { - waittime = zerotime; - } - -#if defined(__OS2__) - DosSleep(0); -#elif !defined(_WIN32) - rc = select(__glutConnectionFD + 1, &fds, - NULL, NULL, &waittime); - if (rc < 0 && errno != EINTR) - __glutFatalError("select error."); -#else - - MsgWaitForMultipleObjects(0, NULL, FALSE, - waittime.tv_sec*1000 + waittime.tv_usec/1000, QS_ALLINPUT); - -#endif -#endif /* not vms6.2 or lower */ - /* Without considering the cause of select unblocking, check - for pending X events and handle any timeouts (by calling - processEventsAndTimeouts). We always look for X events - even if select returned with 0 (indicating a timeout); - otherwise we risk starving X event processing by continous - timeouts. */ - if (XPending(__glutDisplay)) { - immediatelyHandleXinput: - processEventsAndTimeouts(); - } else { - if (__glutTimerList) - handleTimeouts(); - } -} - -static void -idleWait(void) -{ - if (XPending(__glutDisplay)) { - processEventsAndTimeouts(); - } else { - if (__glutTimerList) { - handleTimeouts(); - } - } - /* Make sure idle func still exists! */ - if (__glutIdleFunc) { - __glutIdleFunc(); - } -} - -static GLUTwindow **beforeEnd; - -static GLUTwindow * -processWindowWorkList(GLUTwindow * window) -{ - int workMask; - - if (window->prevWorkWin) { - window->prevWorkWin = processWindowWorkList(window->prevWorkWin); - } else { - beforeEnd = &window->prevWorkWin; - } - - /* Capture work mask for work that needs to be done to this - window, then clear the window's work mask (excepting the - dummy work bit, see below). Then, process the captured - work mask. This allows callbacks in the processing the - captured work mask to set the window's work mask for - subsequent processing. */ - - workMask = window->workMask; - assert((workMask & GLUT_DUMMY_WORK) == 0); - - /* Set the dummy work bit, clearing all other bits, to - indicate that the window is currently on the window work - list _and_ that the window's work mask is currently being - processed. This convinces __glutPutOnWorkList that this - window is on the work list still. */ - window->workMask = GLUT_DUMMY_WORK; - - /* Optimization: most of the time, the work to do is a - redisplay and not these other types of work. Check for - the following cases as a group to before checking each one - individually one by one. This saves about 25 MIPS - instructions in the common redisplay only case. */ - if (workMask & (GLUT_EVENT_MASK_WORK | GLUT_DEVICE_MASK_WORK | - GLUT_CONFIGURE_WORK | GLUT_COLORMAP_WORK | GLUT_MAP_WORK)) { - -#if !defined(_WIN32) && !defined(__OS2__) - /* Be sure to set event mask BEFORE map window is done. */ - if (workMask & GLUT_EVENT_MASK_WORK) { - long eventMask; - - /* Make sure children are not propogating events this - window is selecting for. Be sure to do this before - enabling events on the children's parent. */ - if (window->children) { - GLUTwindow *child = window->children; - unsigned long attribMask = CWDontPropagate; - XSetWindowAttributes wa; - - wa.do_not_propagate_mask = window->eventMask & GLUT_DONT_PROPAGATE_FILTER_MASK; - if (window->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK) { - wa.event_mask = child->eventMask | (window->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK); - attribMask |= CWEventMask; - } - do { - XChangeWindowAttributes(__glutDisplay, child->win, - attribMask, &wa); - child = child->siblings; - } while (child); - } - eventMask = window->eventMask; - if (window->parent && window->parent->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK) - eventMask |= (window->parent->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK); - XSelectInput(__glutDisplay, window->win, eventMask); - if (window->overlay) - XSelectInput(__glutDisplay, window->overlay->win, - window->eventMask & GLUT_OVERLAY_EVENT_FILTER_MASK); - } -#endif /* !_WIN32 */ - /* Be sure to set device mask BEFORE map window is done. */ - if (workMask & GLUT_DEVICE_MASK_WORK) { - __glutUpdateInputDeviceMaskFunc(window); - } - /* Be sure to configure window BEFORE map window is done. */ - if (workMask & GLUT_CONFIGURE_WORK) { -#if defined(__OS2__) - RECTL changes; - -#elif defined(_WIN32) - RECT changes; - POINT point; - UINT flags = SWP_NOACTIVATE | SWP_NOMOVE | SWP_NOOWNERZORDER - | SWP_NOSENDCHANGING | SWP_NOSIZE | SWP_NOZORDER; - - GetClientRect(window->win, &changes); - - /* If this window is a toplevel window, translate the 0,0 client - coordinate into a screen coordinate for proper placement. */ - if (!window->parent) { - point.x = 0; - point.y = 0; - ClientToScreen(window->win, &point); - changes.left = point.x; - changes.top = point.y; - } - if (window->desiredConfMask & (CWX | CWY)) { - changes.left = window->desiredX; - changes.top = window->desiredY; - flags &= ~SWP_NOMOVE; - } - if (window->desiredConfMask & (CWWidth | CWHeight)) { - changes.right = changes.left + window->desiredWidth; - changes.bottom = changes.top + window->desiredHeight; - flags &= ~SWP_NOSIZE; - /* XXX If overlay exists, resize the overlay here, ie. - if (window->overlay) ... */ - } - if (window->desiredConfMask & CWStackMode) { - flags &= ~SWP_NOZORDER; - /* XXX Overlay support might require something special here. */ - } - - /* Adjust the window rectangle because Win32 thinks that the x, y, - width & height are the WHOLE window (including decorations), - whereas GLUT treats the x, y, width & height as only the CLIENT - area of the window. Only do this to top level windows - that are not in game mode (since game mode windows do - not have any decorations). */ - if (!window->parent && window != __glutGameModeWindow) { - AdjustWindowRect(&changes, - WS_OVERLAPPEDWINDOW | WS_CLIPSIBLINGS | WS_CLIPCHILDREN, - FALSE); - } - - /* Do the repositioning, moving, and push/pop. */ - SetWindowPos(window->win, - window->desiredStack == Above ? HWND_TOP : HWND_NOTOPMOST, - changes.left, changes.top, - changes.right - changes.left, changes.bottom - changes.top, - flags); - - /* Zero out the mask. */ - window->desiredConfMask = 0; - - /* This hack causes the window to go back to the right position - when it is taken out of fullscreen mode. */ - if (workMask & GLUT_FULL_SCREEN_WORK) { - window->desiredConfMask |= CWX | CWY; - window->desiredX = point.x; - window->desiredY = point.y; - } -#else /* !_WIN32 */ - XWindowChanges changes; - - changes.x = window->desiredX; - changes.y = window->desiredY; - if (window->desiredConfMask & (CWWidth | CWHeight)) { - changes.width = window->desiredWidth; - changes.height = window->desiredHeight; - if (window->overlay) - XResizeWindow(__glutDisplay, window->overlay->win, - window->desiredWidth, window->desiredHeight); - if (__glutMotifHints != None) { - if (workMask & GLUT_FULL_SCREEN_WORK) { - MotifWmHints hints; - - hints.flags = MWM_HINTS_DECORATIONS; - hints.decorations = 0; /* Absolutely no - decorations. */ - XChangeProperty(__glutDisplay, window->win, - __glutMotifHints, __glutMotifHints, 32, - PropModeReplace, (unsigned char *) &hints, 4); - if (workMask & GLUT_MAP_WORK) { - /* Handle case where glutFullScreen is called - before the first time that the window is - mapped. Some window managers will randomly or - interactively position the window the first - time it is mapped if the window's - WM_NORMAL_HINTS property does not request an - explicit position. We don't want any such - window manager interaction when going - fullscreen. Overwrite the WM_NORMAL_HINTS - property installed by glutCreateWindow's - XSetWMProperties property with one explicitly - requesting a fullscreen window. */ - XSizeHints hints; - - hints.flags = USPosition | USSize; - hints.x = 0; - hints.y = 0; - hints.width = window->desiredWidth; - hints.height = window->desiredHeight; - XSetWMNormalHints(__glutDisplay, window->win, &hints); - } - } else { - XDeleteProperty(__glutDisplay, window->win, __glutMotifHints); - } - } - } - if (window->desiredConfMask & CWStackMode) { - changes.stack_mode = window->desiredStack; - /* Do not let glutPushWindow push window beneath the - underlay. */ - if (window->parent && window->parent->overlay - && window->desiredStack == Below) { - changes.stack_mode = Above; - changes.sibling = window->parent->overlay->win; - window->desiredConfMask |= CWSibling; - } - } - XConfigureWindow(__glutDisplay, window->win, - window->desiredConfMask, &changes); - window->desiredConfMask = 0; -#endif - } -#if !defined(_WIN32) && !defined(__OS2__) - /* Be sure to establish the colormaps BEFORE map window is - done. */ - if (workMask & GLUT_COLORMAP_WORK) { - __glutEstablishColormapsProperty(window); - } -#endif - if (workMask & GLUT_MAP_WORK) { - switch (window->desiredMapState) { - case WithdrawnState: - if (window->parent) { - XUnmapWindow(__glutDisplay, window->win); - } else { - XWithdrawWindow(__glutDisplay, window->win, - __glutScreen); - } - window->shownState = 0; - break; - case NormalState: - XMapWindow(__glutDisplay, window->win); - window->shownState = 1; - break; -#ifdef _WIN32 - case GameModeState: /* Not an Xlib value. */ - ShowWindow(window->win, SW_SHOW); - window->shownState = 1; - break; -#endif - case IconicState: - XIconifyWindow(__glutDisplay, window->win, __glutScreen); - window->shownState = 0; - break; - } - } - } - if (workMask & (GLUT_REDISPLAY_WORK | GLUT_OVERLAY_REDISPLAY_WORK | GLUT_REPAIR_WORK | GLUT_OVERLAY_REPAIR_WORK)) { - if (window->forceReshape) { - /* Guarantee that before a display callback is generated - for a window, a reshape callback must be generated. */ - __glutSetWindow(window); - window->reshape(window->width, window->height); - window->forceReshape = False; - - /* Setting the redisplay bit on the first reshape is - necessary to make the "Mesa glXSwapBuffers to repair - damage" hack operate correctly. Without indicating a - redisplay is necessary, there's not an initial back - buffer render from which to blit from when damage - happens to the window. */ - workMask |= GLUT_REDISPLAY_WORK; - } - /* The code below is more involved than otherwise necessary - because it is paranoid about the overlay or entire window - being removed or destroyed in the course of the callbacks. - Notice how the global __glutWindowDamaged is used to record - the layers' damage status. See the code in glutLayerGet for - how __glutWindowDamaged is used. The point is to not have to - update the "damaged" field after the callback since the - window (or overlay) may be destroyed (or removed) when the - callback returns. */ - - if (window->overlay && window->overlay->display) { - int num = window->num; - Window xid = window->overlay ? window->overlay->win : None; - - /* If an overlay display callback is registered, we - differentiate between a redisplay needed for the - overlay and/or normal plane. If there is no overlay - display callback registered, we simply use the - standard display callback. */ - - if (workMask & (GLUT_REDISPLAY_WORK | GLUT_REPAIR_WORK)) { - if (__glutMesaSwapHackSupport) { - if (window->usedSwapBuffers) { - if ((workMask & (GLUT_REPAIR_WORK | GLUT_REDISPLAY_WORK)) == GLUT_REPAIR_WORK) { - SWAP_BUFFERS_WINDOW(window); - goto skippedDisplayCallback1; - } - } - } - /* Render to normal plane. */ -#ifdef _WIN32 - window->renderDc = window->hdc; -#endif - window->renderWin = window->win; - window->renderCtx = window->ctx; - __glutWindowDamaged = (workMask & GLUT_REPAIR_WORK); - __glutSetWindow(window); - window->usedSwapBuffers = 0; - window->display(); - __glutWindowDamaged = 0; - - skippedDisplayCallback1:; - } - if (workMask & (GLUT_OVERLAY_REDISPLAY_WORK | GLUT_OVERLAY_REPAIR_WORK)) { - window = __glutWindowList[num]; - if (window && window->overlay && - window->overlay->win == xid && window->overlay->display) { - - /* Render to overlay. */ -#ifdef _WIN32 - window->renderDc = window->overlay->hdc; -#endif - window->renderWin = window->overlay->win; - window->renderCtx = window->overlay->ctx; - __glutWindowDamaged = (workMask & GLUT_OVERLAY_REPAIR_WORK); - __glutSetWindow(window); - window->overlay->display(); - __glutWindowDamaged = 0; - } else { - /* Overlay may have since been destroyed or the - overlay callback may have been disabled during - normal display callback. */ - } - } - } else { - if (__glutMesaSwapHackSupport) { - if (!window->overlay && window->usedSwapBuffers) { - if ((workMask & (GLUT_REPAIR_WORK | GLUT_REDISPLAY_WORK)) == GLUT_REPAIR_WORK) { - SWAP_BUFFERS_WINDOW(window); - goto skippedDisplayCallback2; - } - } - } - /* Render to normal plane (and possibly overlay). */ - __glutWindowDamaged = (workMask & (GLUT_OVERLAY_REPAIR_WORK | GLUT_REPAIR_WORK)); - __glutSetWindow(window); - window->usedSwapBuffers = 0; - window->display(); - __glutWindowDamaged = 0; - - skippedDisplayCallback2:; - } - } - /* Combine workMask with window->workMask to determine what - finish and debug work there is. */ - workMask |= window->workMask; - - if (workMask & GLUT_FINISH_WORK) { - /* Finish work makes sure a glFinish gets done to indirect - rendering contexts. Indirect contexts tend to have much - longer latency because lots of OpenGL extension requests - can queue up in the X protocol stream. __glutSetWindow - is where the finish works gets queued for indirect - contexts. */ - __glutSetWindow(window); - glFinish(); - } - if (workMask & GLUT_DEBUG_WORK) { - __glutSetWindow(window); - glutReportErrors(); - } - /* Strip out dummy, finish, and debug work bits. */ - window->workMask &= ~(GLUT_DUMMY_WORK | GLUT_FINISH_WORK | GLUT_DEBUG_WORK); - if (window->workMask) { - /* Leave on work list. */ - return window; - } else { - /* Remove current window from work list. */ - return window->prevWorkWin; - } -} - -#ifndef _WIN32 -static /* X11 implementations do not need this global. */ -#endif -void -__glutProcessWindowWorkLists(void) -{ - if (__glutWindowWorkList) { - GLUTwindow *remainder, *work; - - work = __glutWindowWorkList; - __glutWindowWorkList = NULL; - if (work) { - remainder = processWindowWorkList(work); - if (remainder) { - *beforeEnd = __glutWindowWorkList; - __glutWindowWorkList = remainder; - } - } - } -} - -/* CENTRY */ -void GLUTAPIENTRY -glutMainLoop(void) -{ -#if !defined(_WIN32) - if (!__glutDisplay) - __glutFatalUsage("main loop entered with out proper initialization."); -#endif - if (!__glutWindowListSize) - __glutFatalUsage( - "main loop entered with no windows created."); - for (;;) { - __glutProcessWindowWorkLists(); - if (__glutIdleFunc || __glutWindowWorkList) { - idleWait(); - } else { - if (__glutTimerList) { - waitForSomething(); - } else { - processEventsAndTimeouts(); - } - } - } -} -/* ENDCENTRY */ +/* glut_event.c */ +/* Copyright (c) Mark J. Kilgard, 1994, 1995, 1996, 1997, 1998. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#ifdef __VMS +#include +#endif + +#include +#include +#include +#include +#include /* Some FD_ZERO macros use memset without + prototyping memset. */ + +/* Much of the following #ifdef logic to include the proper + prototypes for the select system call is based on logic + from the X11R6.3 version of . */ + +#if !defined(_WIN32) +# ifdef __sgi +# include /* prototype for bzero used by FD_ZERO */ +# endif +# if (defined(SVR4) || defined(CRAY) || defined(AIXV3)) && !defined(FD_SETSIZE) +# include /* select system call interface */ +# ifdef luna +# include +# endif +# endif + /* AIX 4.2 fubar-ed , so go to heroic measures to get it */ +# if defined(AIXV4) && !defined(NFDBITS) +# include +# endif +#endif /* !_WIN32 */ + +#include + +#if defined(__OS2__) +//??? ­ ä¨£  ?? # include +#elif !defined(_WIN32) +# if defined(__vms) && ( __VMS_VER < 70000000 ) +# include +# else +# ifndef __vms +# include +# endif +# endif +# include +# include +# include +#else +# ifdef __CYGWIN32__ +# include +# else +# include +# endif +# ifdef __hpux + /* XXX Bert Gijsbers reports that HP-UX + needs different keysyms for the End, Insert, and Delete keys + to work on an HP 715. It would be better if HP generated + standard keysyms for standard keys. */ +# include +# endif +#endif /* !_WIN32 */ + +#include "glutint.h" + +#if defined(__vms) && ( __VMS_VER < 70000000 ) +#include +#include +extern int SYS$CLREF(int efn); +extern int SYS$SETIMR(unsigned int efn, struct timeval6 *timeout, void *ast, + unsigned int request_id, unsigned int flags); +extern int SYS$WFLOR(unsigned int efn, unsigned int mask); +extern int SYS$CANTIM(unsigned int request_id, unsigned int mode); +#endif /* __vms, VMs 6.2 or earlier */ + +static GLUTtimer *freeTimerList = NULL; + +GLUTidleCB __glutIdleFunc = NULL; +GLUTtimer *__glutTimerList = NULL; +#ifdef SUPPORT_FORTRAN +GLUTtimer *__glutNewTimer; +#endif +GLUTwindow *__glutWindowWorkList = NULL; +GLUTmenu *__glutMappedMenu; +GLUTmenu *__glutCurrentMenu = NULL; + +void (*__glutUpdateInputDeviceMaskFunc) (GLUTwindow *); +#if !defined(_WIN32) && !defined(__OS2__) +void (*__glutMenuItemEnterOrLeave)(GLUTmenuItem * item, int num, int type) = NULL; +void (*__glutFinishMenu)(Window win, int x, int y); +void (*__glutPaintMenu)(GLUTmenu * menu); +void (*__glutStartMenu)(GLUTmenu * menu, GLUTwindow * window, int x, int y, int x_win, int y_win); +GLUTmenu * (*__glutGetMenuByNum)(int menunum); +GLUTmenuItem * (*__glutGetMenuItem)(GLUTmenu * menu, Window win, int *which); +GLUTmenu * (*__glutGetMenu)(Window win); +#endif + +Atom __glutMotifHints = None; +/* Modifier mask of ~0 implies not in core input callback. */ +unsigned int __glutModifierMask = (unsigned int) ~0; +int __glutWindowDamaged = 0; + +void GLUTAPIENTRY +glutIdleFunc(GLUTidleCB idleFunc) +{ + __glutIdleFunc = idleFunc; +} + +void GLUTAPIENTRY +glutTimerFunc(unsigned int interval, GLUTtimerCB timerFunc, int value) +{ + GLUTtimer *timer, *other; + GLUTtimer **prevptr; +#ifdef OLD_VMS + struct timeval6 now; +#else + struct timeval now; +#endif + + if (!timerFunc) + return; + + if (freeTimerList) { + timer = freeTimerList; + freeTimerList = timer->next; + } else { + timer = (GLUTtimer *) malloc(sizeof(GLUTtimer)); + if (!timer) + __glutFatalError("out of memory."); + } + + timer->func = timerFunc; +#if defined(__vms) && ( __VMS_VER < 70000000 ) + /* VMS time is expressed in units of 100 ns */ + timer->timeout.val = interval * TICKS_PER_MILLISECOND; +#else + timer->timeout.tv_sec = (int) interval / 1000; + timer->timeout.tv_usec = (int) (interval % 1000) * 1000; +#endif + timer->value = value; + timer->next = NULL; + GETTIMEOFDAY(&now); + ADD_TIME(timer->timeout, timer->timeout, now); + prevptr = &__glutTimerList; + other = *prevptr; + while (other && IS_AFTER(other->timeout, timer->timeout)) { + prevptr = &other->next; + other = *prevptr; + } + timer->next = other; +#ifdef SUPPORT_FORTRAN + __glutNewTimer = timer; /* for Fortran binding! */ +#endif + *prevptr = timer; +} + +void +handleTimeouts(void) +{ +#ifdef OLD_VMS + struct timeval6 now; +#else + struct timeval now; +#endif + GLUTtimer *timer; + + /* Assumption is that __glutTimerList is already determined + to be non-NULL. */ + GETTIMEOFDAY(&now); + while (IS_AT_OR_AFTER(__glutTimerList->timeout, now)) { + timer = __glutTimerList; + timer->func(timer->value); + __glutTimerList = timer->next; + timer->next = freeTimerList; + freeTimerList = timer; + if (!__glutTimerList) + break; + } +} + +void +__glutPutOnWorkList(GLUTwindow * window, int workMask) +{ + if (window->workMask) { + /* Already on list; just OR in new workMask. */ + window->workMask |= workMask; + } else { + /* Update work mask and add to window work list. */ + window->workMask = workMask; + /* Assert that if the window does not have a + workMask already, the window should definitely + not be the head of the work list. */ + assert(window != __glutWindowWorkList); + window->prevWorkWin = __glutWindowWorkList; + __glutWindowWorkList = window; + } +} + +void +__glutPostRedisplay(GLUTwindow * window, int layerMask) +{ + int shown = (layerMask & (GLUT_REDISPLAY_WORK | GLUT_REPAIR_WORK)) ? + window->shownState : window->overlay->shownState; + + /* Post a redisplay if the window is visible (or the + visibility of the window is unknown, ie. window->visState + == -1) _and_ the layer is known to be shown. */ + if (window->visState != GLUT_HIDDEN + && window->visState != GLUT_FULLY_COVERED && shown) { + __glutPutOnWorkList(window, layerMask); + } +} + +/* CENTRY */ +void GLUTAPIENTRY +glutPostRedisplay(void) +{ + __glutPostRedisplay(__glutCurrentWindow, GLUT_REDISPLAY_WORK); +} + +/* The advantage of this routine is that it saves the cost of a + glutSetWindow call (entailing an expensive OpenGL context switch), + particularly useful when multiple windows need redisplays posted at + the same times. See also glutPostWindowOverlayRedisplay. */ +void GLUTAPIENTRY +glutPostWindowRedisplay(int win) +{ + __glutPostRedisplay(__glutWindowList[win - 1], GLUT_REDISPLAY_WORK); +} + +/* ENDCENTRY */ +static GLUTeventParser *eventParserList = NULL; + +/* __glutRegisterEventParser allows another module to register + to intercept X events types not otherwise acted on by the + GLUT processEventsAndTimeouts routine. The X Input + extension support code uses an event parser for handling X + Input extension events. */ + +void +__glutRegisterEventParser(GLUTeventParser * parser) +{ + parser->next = eventParserList; + eventParserList = parser; +} + +static void +markWindowHidden(GLUTwindow * window) +{ + if (GLUT_HIDDEN != window->visState) { + GLUTwindow *child; + + if (window->windowStatus) { + window->visState = GLUT_HIDDEN; + __glutSetWindow(window); + window->windowStatus(GLUT_HIDDEN); + } + /* An unmap is only reported on a single window; its + descendents need to know they are no longer visible. */ + child = window->children; + while (child) { + markWindowHidden(child); + child = child->siblings; + } + } +} + +#if !defined(_WIN32) && !defined(__OS2__) + +static void +purgeStaleWindow(Window win) +{ + GLUTstale **pEntry = &__glutStaleWindowList; + GLUTstale *entry = __glutStaleWindowList; + + /* Tranverse singly-linked stale window list look for the + window ID. */ + while (entry) { + if (entry->win == win) { + /* Found it; delete it. */ + *pEntry = entry->next; + free(entry); + return; + } else { + pEntry = &entry->next; + entry = *pEntry; + } + } +} + +/* Unlike XNextEvent, if a signal arrives, + interruptibleXNextEvent will return (with a zero return + value). This helps GLUT drop out of XNextEvent if a signal + is delivered. The intent is so that a GLUT program can call + glutIdleFunc in a signal handler to register an idle func + and then immediately get dropped into the idle func (after + returning from the signal handler). The idea is to make + GLUT's main loop reliably interruptible by signals. */ +static int +interruptibleXNextEvent(Display * dpy, XEvent * event) +{ + fd_set fds; + int rc; + + /* Flush X protocol since XPending does not do this + implicitly. */ + XFlush(__glutDisplay); + for (;;) { + if (XPending(__glutDisplay)) { + XNextEvent(dpy, event); + return 1; + } +#ifndef VMS + /* the combination ConectionNumber-select is buggy on VMS. Sometimes it + * fails. This part of the code hangs the program on VMS7.2. But even + * without it the program seems to run correctly. + * Note that this is a bug in the VMS/DECWindows run-time-libraries. + * Compaq engeneering does not want or is not able to make a fix. + * (last sentence is a quotation from Compaq when I reported the + * problem January 2000) */ + FD_ZERO(&fds); + FD_SET(__glutConnectionFD, &fds); + rc = select(__glutConnectionFD + 1, &fds, NULL, NULL, NULL); + if (rc < 0) { + if (errno == EINTR) { + return 0; + } else { + __glutFatalError("select error."); + } + } +#endif + } +} + +#endif + +static void +processEventsAndTimeouts(void) +{ + do { +#if defined(__OS2__) + QMSG qmsg; /* message from message queue */ + extern HAB hab; /* PM anchor block handle */ + + if(! WinGetMsg( hab, &qmsg, 0UL, 0UL, 0UL ) ) + exit(0); + WinDispatchMsg( hab, /* PM anchor block handle */ + &qmsg ); /* pointer to message */ + +#elif defined(_WIN32) + MSG event; + + if(!GetMessage(&event, NULL, 0, 0)) /* bail if no more messages */ + exit(0); + TranslateMessage(&event); /* translate virtual-key messages */ + DispatchMessage(&event); /* call the window proc */ + /* see win32_event.c for event (message) processing procedures */ +#else + static int mappedMenuButton; + GLUTeventParser *parser; + XEvent event, ahead; + GLUTwindow *window; + GLUTkeyboardCB keyboard; + GLUTspecialCB special; + int gotEvent, width, height; + + gotEvent = interruptibleXNextEvent(__glutDisplay, &event); + if (gotEvent) { + switch (event.type) { + case MappingNotify: + XRefreshKeyboardMapping((XMappingEvent *) & event); + break; + case ConfigureNotify: + window = __glutGetWindow(event.xconfigure.window); + if (window) { + if (window->win != event.xconfigure.window) { + /* Ignore ConfigureNotify sent to the overlay + planes. GLUT could get here because overlays + select for StructureNotify events to receive + DestroyNotify. */ + break; + } + width = event.xconfigure.width; + height = event.xconfigure.height; + if (width != window->width || height != window->height) { + if (window->overlay) { + XResizeWindow(__glutDisplay, window->overlay->win, width, height); + } + window->width = width; + window->height = height; + __glutSetWindow(window); + /* Do not execute OpenGL out of sequence with + respect to the XResizeWindow request! */ + glXWaitX(); + window->reshape(width, height); + window->forceReshape = False; + /* A reshape should be considered like posting a + repair; this is necessary for the "Mesa + glXSwapBuffers to repair damage" hack to operate + correctly. Without it, there's not an initial + back buffer render from which to blit from when + damage happens to the window. */ + __glutPostRedisplay(window, GLUT_REPAIR_WORK); + } + } + break; + case Expose: + /* compress expose events */ + while (XEventsQueued(__glutDisplay, QueuedAfterReading) + > 0) { + XPeekEvent(__glutDisplay, &ahead); + if (ahead.type != Expose || + ahead.xexpose.window != event.xexpose.window) { + break; + } + XNextEvent(__glutDisplay, &event); + } + if (event.xexpose.count == 0) { + GLUTmenu *menu; + + if (__glutMappedMenu && + (menu = __glutGetMenu(event.xexpose.window))) { + __glutPaintMenu(menu); + } else { + window = __glutGetWindow(event.xexpose.window); + if (window) { + if (window->win == event.xexpose.window) { + __glutPostRedisplay(window, GLUT_REPAIR_WORK); + } else if (window->overlay && window->overlay->win == event.xexpose.window) { + __glutPostRedisplay(window, GLUT_OVERLAY_REPAIR_WORK); + } + } + } + } else { + /* there are more exposes to read; wait to redisplay */ + } + break; + case ButtonPress: + case ButtonRelease: + if (__glutMappedMenu && event.type == ButtonRelease + && mappedMenuButton == event.xbutton.button) { + /* Menu is currently popped up and its button is + released. */ + __glutFinishMenu(event.xbutton.window, event.xbutton.x, event.xbutton.y); + } else { + window = __glutGetWindow(event.xbutton.window); + if (window) { + GLUTmenu *menu; + int menuNum; + + menuNum = window->menu[event.xbutton.button - 1]; + /* Make sure that __glutGetMenuByNum is only called if there + really is a menu present. */ + if ((menuNum > 0) && (menu = __glutGetMenuByNum(menuNum))) { + if (event.type == ButtonPress && !__glutMappedMenu) { + __glutStartMenu(menu, window, + event.xbutton.x_root, event.xbutton.y_root, + event.xbutton.x, event.xbutton.y); + mappedMenuButton = event.xbutton.button; + } else { + /* Ignore a release of a button with a menu + attatched to it when no menu is popped up, + or ignore a press when another menu is + already popped up. */ + } + } else if (window->mouse) { + __glutSetWindow(window); + __glutModifierMask = event.xbutton.state; + window->mouse(event.xbutton.button - 1, + event.type == ButtonRelease ? + GLUT_UP : GLUT_DOWN, + event.xbutton.x, event.xbutton.y); + __glutModifierMask = ~0; + } else { + /* Stray mouse events. Ignore. */ + } + } else { + /* Window might have been destroyed and all the + events for the window may not yet be received. */ + } + } + break; + case MotionNotify: + if (!__glutMappedMenu) { + window = __glutGetWindow(event.xmotion.window); + if (window) { + /* If motion function registered _and_ buttons held + * down, call motion function... */ + if (window->motion && event.xmotion.state & + (Button1Mask | Button2Mask | Button3Mask)) { + __glutSetWindow(window); + window->motion(event.xmotion.x, event.xmotion.y); + } + /* If passive motion function registered _and_ + buttons not held down, call passive motion + function... */ + else if (window->passive && + ((event.xmotion.state & + (Button1Mask | Button2Mask | Button3Mask)) == + 0)) { + __glutSetWindow(window); + window->passive(event.xmotion.x, + event.xmotion.y); + } + } + } else { + /* Motion events are thrown away when a pop up menu + is active. */ + } + break; + case KeyPress: + case KeyRelease: + window = __glutGetWindow(event.xkey.window); + if (!window) { + break; + } + if (event.type == KeyPress) { + keyboard = window->keyboard; + } else { + + /* If we are ignoring auto repeated keys for this window, + check if the next event in the X event queue is a KeyPress + for the exact same key (and at the exact same time) as the + key being released. The X11 protocol will send auto + repeated keys as such KeyRelease/KeyPress pairs. */ + + if (window->ignoreKeyRepeat) { + if (XEventsQueued(__glutDisplay, QueuedAfterReading)) { + XPeekEvent(__glutDisplay, &ahead); + if (ahead.type == KeyPress + && ahead.xkey.window == event.xkey.window + && ahead.xkey.keycode == event.xkey.keycode + && ahead.xkey.time == event.xkey.time) { + /* Pop off the repeated KeyPress and ignore + the auto repeated KeyRelease/KeyPress pair. */ + XNextEvent(__glutDisplay, &event); + break; + } + } + } + keyboard = window->keyboardUp; + } + if (keyboard) { + char tmp[1]; + int rc; + + rc = XLookupString(&event.xkey, tmp, sizeof(tmp), + NULL, NULL); + if (rc) { + __glutSetWindow(window); + __glutModifierMask = event.xkey.state; + keyboard(tmp[0], + event.xkey.x, event.xkey.y); + __glutModifierMask = ~0; + break; + } + } + if (event.type == KeyPress) { + special = window->special; + } else { + special = window->specialUp; + } + if (special) { + KeySym ks; + int key; + +/* Introduced in X11R6: (Partial list of) Keypad Functions. Define + in place in case compiling against an older pre-X11R6 + X11/keysymdef.h file. */ +#ifndef XK_KP_Home +#define XK_KP_Home 0xFF95 +#endif +#ifndef XK_KP_Left +#define XK_KP_Left 0xFF96 +#endif +#ifndef XK_KP_Up +#define XK_KP_Up 0xFF97 +#endif +#ifndef XK_KP_Right +#define XK_KP_Right 0xFF98 +#endif +#ifndef XK_KP_Down +#define XK_KP_Down 0xFF99 +#endif +#ifndef XK_KP_Prior +#define XK_KP_Prior 0xFF9A +#endif +#ifndef XK_KP_Next +#define XK_KP_Next 0xFF9B +#endif +#ifndef XK_KP_End +#define XK_KP_End 0xFF9C +#endif +#ifndef XK_KP_Insert +#define XK_KP_Insert 0xFF9E +#endif +#ifndef XK_KP_Delete +#define XK_KP_Delete 0xFF9F +#endif + + ks = XLookupKeysym((XKeyEvent *) & event, 0); + /* XXX Verbose, but makes no assumptions about keysym + layout. */ + switch (ks) { +/* *INDENT-OFF* */ + /* function keys */ + case XK_F1: key = GLUT_KEY_F1; break; + case XK_F2: key = GLUT_KEY_F2; break; + case XK_F3: key = GLUT_KEY_F3; break; + case XK_F4: key = GLUT_KEY_F4; break; + case XK_F5: key = GLUT_KEY_F5; break; + case XK_F6: key = GLUT_KEY_F6; break; + case XK_F7: key = GLUT_KEY_F7; break; + case XK_F8: key = GLUT_KEY_F8; break; + case XK_F9: key = GLUT_KEY_F9; break; + case XK_F10: key = GLUT_KEY_F10; break; + case XK_F11: key = GLUT_KEY_F11; break; + case XK_F12: key = GLUT_KEY_F12; break; + /* directional keys */ + case XK_KP_Left: + case XK_Left: key = GLUT_KEY_LEFT; break; + case XK_KP_Up: /* Introduced in X11R6. */ + case XK_Up: key = GLUT_KEY_UP; break; + case XK_KP_Right: /* Introduced in X11R6. */ + case XK_Right: key = GLUT_KEY_RIGHT; break; + case XK_KP_Down: /* Introduced in X11R6. */ + case XK_Down: key = GLUT_KEY_DOWN; break; +/* *INDENT-ON* */ + + case XK_KP_Prior: /* Introduced in X11R6. */ + case XK_Prior: + /* XK_Prior same as X11R6's XK_Page_Up */ + key = GLUT_KEY_PAGE_UP; + break; + case XK_KP_Next: /* Introduced in X11R6. */ + case XK_Next: + /* XK_Next same as X11R6's XK_Page_Down */ + key = GLUT_KEY_PAGE_DOWN; + break; + case XK_KP_Home: /* Introduced in X11R6. */ + case XK_Home: + key = GLUT_KEY_HOME; + break; +#ifdef __hpux + case XK_Select: +#endif + case XK_KP_End: /* Introduced in X11R6. */ + case XK_End: + key = GLUT_KEY_END; + break; +#ifdef __hpux + case XK_InsertChar: +#endif + case XK_KP_Insert: /* Introduced in X11R6. */ + case XK_Insert: + key = GLUT_KEY_INSERT; + break; +#ifdef __hpux + case XK_DeleteChar: +#endif + case XK_KP_Delete: /* Introduced in X11R6. */ + /* The Delete character is really an ASCII key. */ + __glutSetWindow(window); + keyboard(127, /* ASCII Delete character. */ + event.xkey.x, event.xkey.y); + goto skip; + default: + goto skip; + } + __glutSetWindow(window); + __glutModifierMask = event.xkey.state; + special(key, event.xkey.x, event.xkey.y); + __glutModifierMask = ~0; + skip:; + } + break; + case EnterNotify: + case LeaveNotify: + if (event.xcrossing.mode != NotifyNormal || + event.xcrossing.detail == NotifyNonlinearVirtual || + event.xcrossing.detail == NotifyVirtual) { + + /* Careful to ignore Enter/LeaveNotify events that + come from the pop-up menu pointer grab and ungrab. + Also, ignore "virtual" Enter/LeaveNotify events + since they represent the pointer passing through + the window hierarchy without actually entering or + leaving the actual real estate of a window. */ + + break; + } + if (__glutMappedMenu) { + GLUTmenuItem *item; + int num; + + item = __glutGetMenuItem(__glutMappedMenu, + event.xcrossing.window, &num); + if (item) { + __glutMenuItemEnterOrLeave(item, num, event.type); + break; + } + } + window = __glutGetWindow(event.xcrossing.window); + if (window) { + if (window->entry) { + if (event.type == EnterNotify) { + + /* With overlays established, X can report two + enter events for both the overlay and normal + plane window. Do not generate a second enter + callback if we reported one without an + intervening leave. */ + + if (window->entryState != EnterNotify) { + int num = window->num; + Window xid = window->win; + + window->entryState = EnterNotify; + __glutSetWindow(window); + window->entry(GLUT_ENTERED); + + if (__glutMappedMenu) { + + /* Do not generate any passive motion events + when menus are in use. */ + + } else { + + /* An EnterNotify event can result in a + "compound" callback if a passive motion + callback is also registered. In this case, + be a little paranoid about the possibility + the window could have been destroyed in the + entry callback. */ + + window = __glutWindowList[num]; + if (window && window->passive && window->win == xid) { + __glutSetWindow(window); + window->passive(event.xcrossing.x, event.xcrossing.y); + } + } + } + } else { + if (window->entryState != LeaveNotify) { + + /* When an overlay is established for a window + already mapped and with the pointer in it, + the X server will generate a leave/enter + event pair as the pointer leaves (without + moving) from the normal plane X window to + the newly mapped overlay X window (or vice + versa). This enter/leave pair should not be + reported to the GLUT program since the pair + is a consequence of creating (or destroying) + the overlay, not an actual leave from the + GLUT window. */ + + if (XEventsQueued(__glutDisplay, QueuedAfterReading)) { + XPeekEvent(__glutDisplay, &ahead); + if (ahead.type == EnterNotify && + __glutGetWindow(ahead.xcrossing.window) == window) { + XNextEvent(__glutDisplay, &event); + break; + } + } + window->entryState = LeaveNotify; + __glutSetWindow(window); + window->entry(GLUT_LEFT); + } + } + } else if (window->passive) { + __glutSetWindow(window); + window->passive(event.xcrossing.x, event.xcrossing.y); + } + } + break; + case UnmapNotify: + /* MapNotify events are not needed to maintain + visibility state since VisibilityNotify events will + be delivered when a window becomes visible from + mapping. However, VisibilityNotify events are not + delivered when a window is unmapped (for the window + or its children). */ + window = __glutGetWindow(event.xunmap.window); + if (window) { + if (window->win != event.xconfigure.window) { + /* Ignore UnmapNotify sent to the overlay planes. + GLUT could get here because overlays select for + StructureNotify events to receive DestroyNotify. + */ + break; + } + markWindowHidden(window); + } + break; + case VisibilityNotify: + window = __glutGetWindow(event.xvisibility.window); + if (window) { + /* VisibilityUnobscured+1 = GLUT_FULLY_RETAINED, + VisibilityPartiallyObscured+1 = + GLUT_PARTIALLY_RETAINED, VisibilityFullyObscured+1 + = GLUT_FULLY_COVERED. */ + int visState = event.xvisibility.state + 1; + + if (visState != window->visState) { + if (window->windowStatus) { + window->visState = visState; + __glutSetWindow(window); + window->windowStatus(visState); + } + } + } + break; + case ClientMessage: + if (event.xclient.data.l[0] == __glutWMDeleteWindow) + exit(0); + break; + case DestroyNotify: + purgeStaleWindow(event.xdestroywindow.window); + break; + case CirculateNotify: + case CreateNotify: + case GravityNotify: + case ReparentNotify: + /* Uninteresting to GLUT (but possible for GLUT to + receive). */ + break; + default: + /* Pass events not directly handled by the GLUT main + event loop to any event parsers that have been + registered. In this way, X Input extension events + are passed to the correct handler without forcing + all GLUT programs to support X Input event handling. + */ + parser = eventParserList; + while (parser) { + if (parser->func(&event)) + break; + parser = parser->next; + } + break; + } + } +#endif /* _WIN32 */ + if (__glutTimerList) { + handleTimeouts(); + } + } + while (XPending(__glutDisplay)); +} + +static void +waitForSomething(void) +{ +#if defined(__vms) && ( __VMS_VER < 70000000 ) + static struct timeval6 zerotime = + {0}; + unsigned int timer_efn; +#define timer_id 'glut' /* random :-) number */ + unsigned int wait_mask; +#else + static struct timeval zerotime = + {0, 0}; +#if defined(__OS2__) + +#elif !defined(_WIN32) + fd_set fds; +#endif +#endif +#ifdef OLD_VMS + struct timeval6 now, timeout, waittime; +#else + struct timeval now, timeout, waittime; +#endif +#if !defined(_WIN32) + int rc; +#endif + + /* Flush X protocol since XPending does not do this + implicitly. */ + XFlush(__glutDisplay); + if (XPending(__glutDisplay)) { + /* It is possible (but quite rare) that XFlush may have + needed to wait for a writable X connection file + descriptor, and in the process, may have had to read off + X protocol from the file descriptor. If XPending is true, + this case occured and we should avoid waiting in select + since X protocol buffered within Xlib is due to be + processed and potentially no more X protocol is on the + file descriptor, so we would risk waiting improperly in + select. */ + goto immediatelyHandleXinput; + } +#if defined(__vms) && ( __VMS_VER < 70000000 ) + timeout = __glutTimerList->timeout; + GETTIMEOFDAY(&now); + wait_mask = 1 << (__glutConnectionFD & 31); + if (IS_AFTER(now, timeout)) { + /* We need an event flag for the timer. */ + /* XXX The `right' way to do this is to use LIB$GET_EF, but + since it needs to be in the same cluster as the EFN for + the display, we will have hack it. */ + timer_efn = __glutConnectionFD - 1; + if ((timer_efn / 32) != (__glutConnectionFD / 32)) { + timer_efn = __glutConnectionFD + 1; + } + rc = SYS$CLREF(timer_efn); + rc = SYS$SETIMR(timer_efn, &timeout, NULL, timer_id, 0); + wait_mask |= 1 << (timer_efn & 31); + } else { + timer_efn = 0; + } + rc = SYS$WFLOR(__glutConnectionFD, wait_mask); + if (timer_efn != 0 && SYS$CLREF(timer_efn) == SS$_WASCLR) { + rc = SYS$CANTIM(timer_id, PSL$C_USER); + } + /* XXX There does not seem to be checking of "rc" in the code + above. Can any of the SYS$ routines above fail? */ +#else /* not vms6.2 or lower */ +#if defined(__OS2__) + +#elif !defined(_WIN32) + FD_ZERO(&fds); + FD_SET(__glutConnectionFD, &fds); +#endif + timeout = __glutTimerList->timeout; + GETTIMEOFDAY(&now); + if (IS_AFTER(now, timeout)) { + TIMEDELTA(waittime, timeout, now); + } else { + waittime = zerotime; + } + +#if defined(__OS2__) + DosSleep(0); +#elif !defined(_WIN32) + rc = select(__glutConnectionFD + 1, &fds, + NULL, NULL, &waittime); + if (rc < 0 && errno != EINTR) + __glutFatalError("select error."); +#else + + MsgWaitForMultipleObjects(0, NULL, FALSE, + waittime.tv_sec*1000 + waittime.tv_usec/1000, QS_ALLINPUT); + +#endif +#endif /* not vms6.2 or lower */ + /* Without considering the cause of select unblocking, check + for pending X events and handle any timeouts (by calling + processEventsAndTimeouts). We always look for X events + even if select returned with 0 (indicating a timeout); + otherwise we risk starving X event processing by continous + timeouts. */ + if (XPending(__glutDisplay)) { + immediatelyHandleXinput: + processEventsAndTimeouts(); + } else { + if (__glutTimerList) + handleTimeouts(); + } +} + +static void +idleWait(void) +{ + if (XPending(__glutDisplay)) { + processEventsAndTimeouts(); + } else { + if (__glutTimerList) { + handleTimeouts(); + } + } + /* Make sure idle func still exists! */ + if (__glutIdleFunc) { + __glutIdleFunc(); + } +} + +static GLUTwindow **beforeEnd; + +static GLUTwindow * +processWindowWorkList(GLUTwindow * window) +{ + int workMask; + + if (window->prevWorkWin) { + window->prevWorkWin = processWindowWorkList(window->prevWorkWin); + } else { + beforeEnd = &window->prevWorkWin; + } + + /* Capture work mask for work that needs to be done to this + window, then clear the window's work mask (excepting the + dummy work bit, see below). Then, process the captured + work mask. This allows callbacks in the processing the + captured work mask to set the window's work mask for + subsequent processing. */ + + workMask = window->workMask; + assert((workMask & GLUT_DUMMY_WORK) == 0); + + /* Set the dummy work bit, clearing all other bits, to + indicate that the window is currently on the window work + list _and_ that the window's work mask is currently being + processed. This convinces __glutPutOnWorkList that this + window is on the work list still. */ + window->workMask = GLUT_DUMMY_WORK; + + /* Optimization: most of the time, the work to do is a + redisplay and not these other types of work. Check for + the following cases as a group to before checking each one + individually one by one. This saves about 25 MIPS + instructions in the common redisplay only case. */ + if (workMask & (GLUT_EVENT_MASK_WORK | GLUT_DEVICE_MASK_WORK | + GLUT_CONFIGURE_WORK | GLUT_COLORMAP_WORK | GLUT_MAP_WORK)) { + +#if !defined(_WIN32) && !defined(__OS2__) + /* Be sure to set event mask BEFORE map window is done. */ + if (workMask & GLUT_EVENT_MASK_WORK) { + long eventMask; + + /* Make sure children are not propogating events this + window is selecting for. Be sure to do this before + enabling events on the children's parent. */ + if (window->children) { + GLUTwindow *child = window->children; + unsigned long attribMask = CWDontPropagate; + XSetWindowAttributes wa; + + wa.do_not_propagate_mask = window->eventMask & GLUT_DONT_PROPAGATE_FILTER_MASK; + if (window->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK) { + wa.event_mask = child->eventMask | (window->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK); + attribMask |= CWEventMask; + } + do { + XChangeWindowAttributes(__glutDisplay, child->win, + attribMask, &wa); + child = child->siblings; + } while (child); + } + eventMask = window->eventMask; + if (window->parent && window->parent->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK) + eventMask |= (window->parent->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK); + XSelectInput(__glutDisplay, window->win, eventMask); + if (window->overlay) + XSelectInput(__glutDisplay, window->overlay->win, + window->eventMask & GLUT_OVERLAY_EVENT_FILTER_MASK); + } +#endif /* !_WIN32 */ + /* Be sure to set device mask BEFORE map window is done. */ + if (workMask & GLUT_DEVICE_MASK_WORK) { + __glutUpdateInputDeviceMaskFunc(window); + } + /* Be sure to configure window BEFORE map window is done. */ + if (workMask & GLUT_CONFIGURE_WORK) { +#if defined(__OS2__) + RECTL changes; + +#elif defined(_WIN32) + RECT changes; + POINT point; + UINT flags = SWP_NOACTIVATE | SWP_NOMOVE | SWP_NOOWNERZORDER + | SWP_NOSENDCHANGING | SWP_NOSIZE | SWP_NOZORDER; + + GetClientRect(window->win, &changes); + + /* If this window is a toplevel window, translate the 0,0 client + coordinate into a screen coordinate for proper placement. */ + if (!window->parent) { + point.x = 0; + point.y = 0; + ClientToScreen(window->win, &point); + changes.left = point.x; + changes.top = point.y; + } + if (window->desiredConfMask & (CWX | CWY)) { + changes.left = window->desiredX; + changes.top = window->desiredY; + flags &= ~SWP_NOMOVE; + } + if (window->desiredConfMask & (CWWidth | CWHeight)) { + changes.right = changes.left + window->desiredWidth; + changes.bottom = changes.top + window->desiredHeight; + flags &= ~SWP_NOSIZE; + /* XXX If overlay exists, resize the overlay here, ie. + if (window->overlay) ... */ + } + if (window->desiredConfMask & CWStackMode) { + flags &= ~SWP_NOZORDER; + /* XXX Overlay support might require something special here. */ + } + + /* Adjust the window rectangle because Win32 thinks that the x, y, + width & height are the WHOLE window (including decorations), + whereas GLUT treats the x, y, width & height as only the CLIENT + area of the window. Only do this to top level windows + that are not in game mode (since game mode windows do + not have any decorations). */ + if (!window->parent && window != __glutGameModeWindow) { + AdjustWindowRect(&changes, + WS_OVERLAPPEDWINDOW | WS_CLIPSIBLINGS | WS_CLIPCHILDREN, + FALSE); + } + + /* Do the repositioning, moving, and push/pop. */ + SetWindowPos(window->win, + window->desiredStack == Above ? HWND_TOP : HWND_NOTOPMOST, + changes.left, changes.top, + changes.right - changes.left, changes.bottom - changes.top, + flags); + + /* Zero out the mask. */ + window->desiredConfMask = 0; + + /* This hack causes the window to go back to the right position + when it is taken out of fullscreen mode. */ + if (workMask & GLUT_FULL_SCREEN_WORK) { + window->desiredConfMask |= CWX | CWY; + window->desiredX = point.x; + window->desiredY = point.y; + } +#else /* !_WIN32 */ + XWindowChanges changes; + + changes.x = window->desiredX; + changes.y = window->desiredY; + if (window->desiredConfMask & (CWWidth | CWHeight)) { + changes.width = window->desiredWidth; + changes.height = window->desiredHeight; + if (window->overlay) + XResizeWindow(__glutDisplay, window->overlay->win, + window->desiredWidth, window->desiredHeight); + if (__glutMotifHints != None) { + if (workMask & GLUT_FULL_SCREEN_WORK) { + MotifWmHints hints; + + hints.flags = MWM_HINTS_DECORATIONS; + hints.decorations = 0; /* Absolutely no + decorations. */ + XChangeProperty(__glutDisplay, window->win, + __glutMotifHints, __glutMotifHints, 32, + PropModeReplace, (unsigned char *) &hints, 4); + if (workMask & GLUT_MAP_WORK) { + /* Handle case where glutFullScreen is called + before the first time that the window is + mapped. Some window managers will randomly or + interactively position the window the first + time it is mapped if the window's + WM_NORMAL_HINTS property does not request an + explicit position. We don't want any such + window manager interaction when going + fullscreen. Overwrite the WM_NORMAL_HINTS + property installed by glutCreateWindow's + XSetWMProperties property with one explicitly + requesting a fullscreen window. */ + XSizeHints hints; + + hints.flags = USPosition | USSize; + hints.x = 0; + hints.y = 0; + hints.width = window->desiredWidth; + hints.height = window->desiredHeight; + XSetWMNormalHints(__glutDisplay, window->win, &hints); + } + } else { + XDeleteProperty(__glutDisplay, window->win, __glutMotifHints); + } + } + } + if (window->desiredConfMask & CWStackMode) { + changes.stack_mode = window->desiredStack; + /* Do not let glutPushWindow push window beneath the + underlay. */ + if (window->parent && window->parent->overlay + && window->desiredStack == Below) { + changes.stack_mode = Above; + changes.sibling = window->parent->overlay->win; + window->desiredConfMask |= CWSibling; + } + } + XConfigureWindow(__glutDisplay, window->win, + window->desiredConfMask, &changes); + window->desiredConfMask = 0; +#endif + } +#if !defined(_WIN32) && !defined(__OS2__) + /* Be sure to establish the colormaps BEFORE map window is + done. */ + if (workMask & GLUT_COLORMAP_WORK) { + __glutEstablishColormapsProperty(window); + } +#endif + if (workMask & GLUT_MAP_WORK) { + switch (window->desiredMapState) { + case WithdrawnState: + if (window->parent) { + XUnmapWindow(__glutDisplay, window->win); + } else { + XWithdrawWindow(__glutDisplay, window->win, + __glutScreen); + } + window->shownState = 0; + break; + case NormalState: + XMapWindow(__glutDisplay, window->win); + window->shownState = 1; + break; +#ifdef _WIN32 + case GameModeState: /* Not an Xlib value. */ + ShowWindow(window->win, SW_SHOW); + window->shownState = 1; + break; +#endif + case IconicState: + XIconifyWindow(__glutDisplay, window->win, __glutScreen); + window->shownState = 0; + break; + } + } + } + if (workMask & (GLUT_REDISPLAY_WORK | GLUT_OVERLAY_REDISPLAY_WORK | GLUT_REPAIR_WORK | GLUT_OVERLAY_REPAIR_WORK)) { + if (window->forceReshape) { + /* Guarantee that before a display callback is generated + for a window, a reshape callback must be generated. */ + __glutSetWindow(window); + window->reshape(window->width, window->height); + window->forceReshape = False; + + /* Setting the redisplay bit on the first reshape is + necessary to make the "Mesa glXSwapBuffers to repair + damage" hack operate correctly. Without indicating a + redisplay is necessary, there's not an initial back + buffer render from which to blit from when damage + happens to the window. */ + workMask |= GLUT_REDISPLAY_WORK; + } + /* The code below is more involved than otherwise necessary + because it is paranoid about the overlay or entire window + being removed or destroyed in the course of the callbacks. + Notice how the global __glutWindowDamaged is used to record + the layers' damage status. See the code in glutLayerGet for + how __glutWindowDamaged is used. The point is to not have to + update the "damaged" field after the callback since the + window (or overlay) may be destroyed (or removed) when the + callback returns. */ + + if (window->overlay && window->overlay->display) { + int num = window->num; + Window xid = window->overlay ? window->overlay->win : None; + + /* If an overlay display callback is registered, we + differentiate between a redisplay needed for the + overlay and/or normal plane. If there is no overlay + display callback registered, we simply use the + standard display callback. */ + + if (workMask & (GLUT_REDISPLAY_WORK | GLUT_REPAIR_WORK)) { + if (__glutMesaSwapHackSupport) { + if (window->usedSwapBuffers) { + if ((workMask & (GLUT_REPAIR_WORK | GLUT_REDISPLAY_WORK)) == GLUT_REPAIR_WORK) { + SWAP_BUFFERS_WINDOW(window); + goto skippedDisplayCallback1; + } + } + } + /* Render to normal plane. */ +#ifdef _WIN32 + window->renderDc = window->hdc; +#endif + window->renderWin = window->win; + window->renderCtx = window->ctx; + __glutWindowDamaged = (workMask & GLUT_REPAIR_WORK); + __glutSetWindow(window); + window->usedSwapBuffers = 0; + window->display(); + __glutWindowDamaged = 0; + + skippedDisplayCallback1:; + } + if (workMask & (GLUT_OVERLAY_REDISPLAY_WORK | GLUT_OVERLAY_REPAIR_WORK)) { + window = __glutWindowList[num]; + if (window && window->overlay && + window->overlay->win == xid && window->overlay->display) { + + /* Render to overlay. */ +#ifdef _WIN32 + window->renderDc = window->overlay->hdc; +#endif + window->renderWin = window->overlay->win; + window->renderCtx = window->overlay->ctx; + __glutWindowDamaged = (workMask & GLUT_OVERLAY_REPAIR_WORK); + __glutSetWindow(window); + window->overlay->display(); + __glutWindowDamaged = 0; + } else { + /* Overlay may have since been destroyed or the + overlay callback may have been disabled during + normal display callback. */ + } + } + } else { + if (__glutMesaSwapHackSupport) { + if (!window->overlay && window->usedSwapBuffers) { + if ((workMask & (GLUT_REPAIR_WORK | GLUT_REDISPLAY_WORK)) == GLUT_REPAIR_WORK) { + SWAP_BUFFERS_WINDOW(window); + goto skippedDisplayCallback2; + } + } + } + /* Render to normal plane (and possibly overlay). */ + __glutWindowDamaged = (workMask & (GLUT_OVERLAY_REPAIR_WORK | GLUT_REPAIR_WORK)); + __glutSetWindow(window); + window->usedSwapBuffers = 0; + window->display(); + __glutWindowDamaged = 0; + + skippedDisplayCallback2:; + } + } + /* Combine workMask with window->workMask to determine what + finish and debug work there is. */ + workMask |= window->workMask; + + if (workMask & GLUT_FINISH_WORK) { + /* Finish work makes sure a glFinish gets done to indirect + rendering contexts. Indirect contexts tend to have much + longer latency because lots of OpenGL extension requests + can queue up in the X protocol stream. __glutSetWindow + is where the finish works gets queued for indirect + contexts. */ + __glutSetWindow(window); + glFinish(); + } + if (workMask & GLUT_DEBUG_WORK) { + __glutSetWindow(window); + glutReportErrors(); + } + /* Strip out dummy, finish, and debug work bits. */ + window->workMask &= ~(GLUT_DUMMY_WORK | GLUT_FINISH_WORK | GLUT_DEBUG_WORK); + if (window->workMask) { + /* Leave on work list. */ + return window; + } else { + /* Remove current window from work list. */ + return window->prevWorkWin; + } +} + +#ifndef _WIN32 +static /* X11 implementations do not need this global. */ +#endif +void +__glutProcessWindowWorkLists(void) +{ + if (__glutWindowWorkList) { + GLUTwindow *remainder, *work; + + work = __glutWindowWorkList; + __glutWindowWorkList = NULL; + if (work) { + remainder = processWindowWorkList(work); + if (remainder) { + *beforeEnd = __glutWindowWorkList; + __glutWindowWorkList = remainder; + } + } + } +} + +/* CENTRY */ +void GLUTAPIENTRY +glutMainLoop(void) +{ +#if !defined(_WIN32) + if (!__glutDisplay) + __glutFatalUsage("main loop entered with out proper initialization."); +#endif + if (!__glutWindowListSize) + __glutFatalUsage( + "main loop entered with no windows created."); + for (;;) { + __glutProcessWindowWorkLists(); + if (__glutIdleFunc || __glutWindowWorkList) { + idleWait(); + } else { + if (__glutTimerList) { + waitForSomething(); + } else { + processEventsAndTimeouts(); + } + } + } +} +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_ext.cpp b/src/glut/os2/glut_ext.cpp index feb6fbfa20..6617fd6edb 100644 --- a/src/glut/os2/glut_ext.cpp +++ b/src/glut/os2/glut_ext.cpp @@ -1,204 +1,204 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include -#include - -#include "glutint.h" - -/* CENTRY */ -int GLUTAPIENTRY -glutExtensionSupported(const char *extension) -{ - static const GLubyte *extensions = NULL; - const GLubyte *start; - GLubyte *where, *terminator; - - /* Extension names should not have spaces. */ - where = (GLubyte *) strchr(extension, ' '); - if (where || *extension == '\0') - return 0; - - if (!extensions) { - extensions = glGetString(GL_EXTENSIONS); - } - /* It takes a bit of care to be fool-proof about parsing the - OpenGL extensions string. Don't be fooled by sub-strings, - etc. */ - start = extensions; - for (;;) { - /* If your application crashes in the strstr routine below, - you are probably calling glutExtensionSupported without - having a current window. Calling glGetString without - a current OpenGL context has unpredictable results. - Please fix your program. */ - where = (GLubyte *) strstr((const char *) start, extension); - if (!where) - break; - terminator = where + strlen(extension); - if (where == start || *(where - 1) == ' ') { - if (*terminator == ' ' || *terminator == '\0') { - return 1; - } - } - start = terminator; - } - return 0; -} - - -struct name_address_pair { - const char *name; - const void *address; -}; - -static struct name_address_pair glut_functions[] = { - { "glutInit", (const void *) glutInit }, - { "glutInitDisplayMode", (const void *) glutInitDisplayMode }, - { "glutInitDisplayString", (const void *) glutInitDisplayString }, - { "glutInitWindowPosition", (const void *) glutInitWindowPosition }, - { "glutInitWindowSize", (const void *) glutInitWindowSize }, - { "glutMainLoop", (const void *) glutMainLoop }, - { "glutCreateWindow", (const void *) glutCreateWindow }, - { "glutCreateSubWindow", (const void *) glutCreateSubWindow }, - { "glutDestroyWindow", (const void *) glutDestroyWindow }, - { "glutPostRedisplay", (const void *) glutPostRedisplay }, - { "glutPostWindowRedisplay", (const void *) glutPostWindowRedisplay }, - { "glutSwapBuffers", (const void *) glutSwapBuffers }, - { "glutGetWindow", (const void *) glutGetWindow }, - { "glutSetWindow", (const void *) glutSetWindow }, - { "glutSetWindowTitle", (const void *) glutSetWindowTitle }, - { "glutSetIconTitle", (const void *) glutSetIconTitle }, - { "glutPositionWindow", (const void *) glutPositionWindow }, - { "glutReshapeWindow", (const void *) glutReshapeWindow }, - { "glutPopWindow", (const void *) glutPopWindow }, - { "glutPushWindow", (const void *) glutPushWindow }, - { "glutIconifyWindow", (const void *) glutIconifyWindow }, - { "glutShowWindow", (const void *) glutShowWindow }, - { "glutHideWindow", (const void *) glutHideWindow }, - { "glutFullScreen", (const void *) glutFullScreen }, - { "glutSetCursor", (const void *) glutSetCursor }, - { "glutWarpPointer", (const void *) glutWarpPointer }, - { "glutEstablishOverlay", (const void *) glutEstablishOverlay }, - { "glutRemoveOverlay", (const void *) glutRemoveOverlay }, - { "glutUseLayer", (const void *) glutUseLayer }, - { "glutPostOverlayRedisplay", (const void *) glutPostOverlayRedisplay }, - { "glutPostWindowOverlayRedisplay", (const void *) glutPostWindowOverlayRedisplay }, - { "glutShowOverlay", (const void *) glutShowOverlay }, - { "glutHideOverlay", (const void *) glutHideOverlay }, - { "glutCreateMenu", (const void *) glutCreateMenu }, - { "glutDestroyMenu", (const void *) glutDestroyMenu }, - { "glutGetMenu", (const void *) glutGetMenu }, - { "glutSetMenu", (const void *) glutSetMenu }, - { "glutAddMenuEntry", (const void *) glutAddMenuEntry }, - { "glutAddSubMenu", (const void *) glutAddSubMenu }, - { "glutChangeToMenuEntry", (const void *) glutChangeToMenuEntry }, - { "glutChangeToSubMenu", (const void *) glutChangeToSubMenu }, - { "glutRemoveMenuItem", (const void *) glutRemoveMenuItem }, - { "glutAttachMenu", (const void *) glutAttachMenu }, - { "glutDetachMenu", (const void *) glutDetachMenu }, - { "glutDisplayFunc", (const void *) glutDisplayFunc }, - { "glutReshapeFunc", (const void *) glutReshapeFunc }, - { "glutKeyboardFunc", (const void *) glutKeyboardFunc }, - { "glutMouseFunc", (const void *) glutMouseFunc }, - { "glutMotionFunc", (const void *) glutMotionFunc }, - { "glutPassiveMotionFunc", (const void *) glutPassiveMotionFunc }, - { "glutEntryFunc", (const void *) glutEntryFunc }, - { "glutVisibilityFunc", (const void *) glutVisibilityFunc }, - { "glutIdleFunc", (const void *) glutIdleFunc }, - { "glutTimerFunc", (const void *) glutTimerFunc }, - { "glutMenuStateFunc", (const void *) glutMenuStateFunc }, - { "glutSpecialFunc", (const void *) glutSpecialFunc }, - { "glutSpaceballMotionFunc", (const void *) glutSpaceballMotionFunc }, - { "glutSpaceballRotateFunc", (const void *) glutSpaceballRotateFunc }, - { "glutSpaceballButtonFunc", (const void *) glutSpaceballButtonFunc }, - { "glutButtonBoxFunc", (const void *) glutButtonBoxFunc }, - { "glutDialsFunc", (const void *) glutDialsFunc }, - { "glutTabletMotionFunc", (const void *) glutTabletMotionFunc }, - { "glutTabletButtonFunc", (const void *) glutTabletButtonFunc }, - { "glutMenuStatusFunc", (const void *) glutMenuStatusFunc }, - { "glutOverlayDisplayFunc", (const void *) glutOverlayDisplayFunc }, - { "glutWindowStatusFunc", (const void *) glutWindowStatusFunc }, - { "glutKeyboardUpFunc", (const void *) glutKeyboardUpFunc }, - { "glutSpecialUpFunc", (const void *) glutSpecialUpFunc }, - { "glutJoystickFunc", (const void *) glutJoystickFunc }, - { "glutSetColor", (const void *) glutSetColor }, - { "glutGetColor", (const void *) glutGetColor }, - { "glutCopyColormap", (const void *) glutCopyColormap }, - { "glutGet", (const void *) glutGet }, - { "glutDeviceGet", (const void *) glutDeviceGet }, - { "glutExtensionSupported", (const void *) glutExtensionSupported }, - { "glutGetModifiers", (const void *) glutGetModifiers }, - { "glutLayerGet", (const void *) glutLayerGet }, - { "glutGetProcAddress", (const void *) glutGetProcAddress }, - { "glutBitmapCharacter", (const void *) glutBitmapCharacter }, - { "glutBitmapWidth", (const void *) glutBitmapWidth }, - { "glutStrokeCharacter", (const void *) glutStrokeCharacter }, - { "glutStrokeWidth", (const void *) glutStrokeWidth }, - { "glutBitmapLength", (const void *) glutBitmapLength }, - { "glutStrokeLength", (const void *) glutStrokeLength }, - { "glutWireSphere", (const void *) glutWireSphere }, - { "glutSolidSphere", (const void *) glutSolidSphere }, - { "glutWireCone", (const void *) glutWireCone }, - { "glutSolidCone", (const void *) glutSolidCone }, - { "glutWireCube", (const void *) glutWireCube }, - { "glutSolidCube", (const void *) glutSolidCube }, - { "glutWireTorus", (const void *) glutWireTorus }, - { "glutSolidTorus", (const void *) glutSolidTorus }, - { "glutWireDodecahedron", (const void *) glutWireDodecahedron }, - { "glutSolidDodecahedron", (const void *) glutSolidDodecahedron }, - { "glutWireTeapot", (const void *) glutWireTeapot }, - { "glutSolidTeapot", (const void *) glutSolidTeapot }, - { "glutWireOctahedron", (const void *) glutWireOctahedron }, - { "glutSolidOctahedron", (const void *) glutSolidOctahedron }, - { "glutWireTetrahedron", (const void *) glutWireTetrahedron }, - { "glutSolidTetrahedron", (const void *) glutSolidTetrahedron }, - { "glutWireIcosahedron", (const void *) glutWireIcosahedron }, - { "glutSolidIcosahedron", (const void *) glutSolidIcosahedron }, - { "glutVideoResizeGet", (const void *) glutVideoResizeGet }, - { "glutSetupVideoResizing", (const void *) glutSetupVideoResizing }, - { "glutStopVideoResizing", (const void *) glutStopVideoResizing }, - { "glutVideoResize", (const void *) glutVideoResize }, - { "glutVideoPan", (const void *) glutVideoPan }, - { "glutReportErrors", (const void *) glutReportErrors }, - { "glutIgnoreKeyRepeat", (const void *) glutIgnoreKeyRepeat }, - { "glutSetKeyRepeat", (const void *) glutSetKeyRepeat }, - { "glutForceJoystickFunc", (const void *) glutForceJoystickFunc }, - { "glutGameModeString", (const void *) glutGameModeString }, - { "glutEnterGameMode", (const void *) glutEnterGameMode }, - { "glutLeaveGameMode", (const void *) glutLeaveGameMode }, - { "glutGameModeGet", (const void *) glutGameModeGet }, - { NULL, NULL } -}; - - -/* XXX This isn't an official GLUT function, yet */ -void * GLUTAPIENTRY -glutGetProcAddress(const char *procName) -{ - /* Try GLUT functions first */ - int i; - for (i = 0; glut_functions[i].name; i++) { - if (strcmp(glut_functions[i].name, procName) == 0) - return (void *) glut_functions[i].address; - } - - /* Try core GL functions */ -#if defined(_WIN32) - return (void *) wglGetProcAddress((LPCSTR) procName); - -#elif defined(__OS2PM__) - return (void *) wglGetProcAddress((char *) procName); -#elif defined(GLX_ARB_get_proc_address) - return (void *) glXGetProcAddressARB((const GLubyte *) procName); -#else - return NULL; -#endif -} - - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include +#include + +#include "glutint.h" + +/* CENTRY */ +int GLUTAPIENTRY +glutExtensionSupported(const char *extension) +{ + static const GLubyte *extensions = NULL; + const GLubyte *start; + GLubyte *where, *terminator; + + /* Extension names should not have spaces. */ + where = (GLubyte *) strchr(extension, ' '); + if (where || *extension == '\0') + return 0; + + if (!extensions) { + extensions = glGetString(GL_EXTENSIONS); + } + /* It takes a bit of care to be fool-proof about parsing the + OpenGL extensions string. Don't be fooled by sub-strings, + etc. */ + start = extensions; + for (;;) { + /* If your application crashes in the strstr routine below, + you are probably calling glutExtensionSupported without + having a current window. Calling glGetString without + a current OpenGL context has unpredictable results. + Please fix your program. */ + where = (GLubyte *) strstr((const char *) start, extension); + if (!where) + break; + terminator = where + strlen(extension); + if (where == start || *(where - 1) == ' ') { + if (*terminator == ' ' || *terminator == '\0') { + return 1; + } + } + start = terminator; + } + return 0; +} + + +struct name_address_pair { + const char *name; + const void *address; +}; + +static struct name_address_pair glut_functions[] = { + { "glutInit", (const void *) glutInit }, + { "glutInitDisplayMode", (const void *) glutInitDisplayMode }, + { "glutInitDisplayString", (const void *) glutInitDisplayString }, + { "glutInitWindowPosition", (const void *) glutInitWindowPosition }, + { "glutInitWindowSize", (const void *) glutInitWindowSize }, + { "glutMainLoop", (const void *) glutMainLoop }, + { "glutCreateWindow", (const void *) glutCreateWindow }, + { "glutCreateSubWindow", (const void *) glutCreateSubWindow }, + { "glutDestroyWindow", (const void *) glutDestroyWindow }, + { "glutPostRedisplay", (const void *) glutPostRedisplay }, + { "glutPostWindowRedisplay", (const void *) glutPostWindowRedisplay }, + { "glutSwapBuffers", (const void *) glutSwapBuffers }, + { "glutGetWindow", (const void *) glutGetWindow }, + { "glutSetWindow", (const void *) glutSetWindow }, + { "glutSetWindowTitle", (const void *) glutSetWindowTitle }, + { "glutSetIconTitle", (const void *) glutSetIconTitle }, + { "glutPositionWindow", (const void *) glutPositionWindow }, + { "glutReshapeWindow", (const void *) glutReshapeWindow }, + { "glutPopWindow", (const void *) glutPopWindow }, + { "glutPushWindow", (const void *) glutPushWindow }, + { "glutIconifyWindow", (const void *) glutIconifyWindow }, + { "glutShowWindow", (const void *) glutShowWindow }, + { "glutHideWindow", (const void *) glutHideWindow }, + { "glutFullScreen", (const void *) glutFullScreen }, + { "glutSetCursor", (const void *) glutSetCursor }, + { "glutWarpPointer", (const void *) glutWarpPointer }, + { "glutEstablishOverlay", (const void *) glutEstablishOverlay }, + { "glutRemoveOverlay", (const void *) glutRemoveOverlay }, + { "glutUseLayer", (const void *) glutUseLayer }, + { "glutPostOverlayRedisplay", (const void *) glutPostOverlayRedisplay }, + { "glutPostWindowOverlayRedisplay", (const void *) glutPostWindowOverlayRedisplay }, + { "glutShowOverlay", (const void *) glutShowOverlay }, + { "glutHideOverlay", (const void *) glutHideOverlay }, + { "glutCreateMenu", (const void *) glutCreateMenu }, + { "glutDestroyMenu", (const void *) glutDestroyMenu }, + { "glutGetMenu", (const void *) glutGetMenu }, + { "glutSetMenu", (const void *) glutSetMenu }, + { "glutAddMenuEntry", (const void *) glutAddMenuEntry }, + { "glutAddSubMenu", (const void *) glutAddSubMenu }, + { "glutChangeToMenuEntry", (const void *) glutChangeToMenuEntry }, + { "glutChangeToSubMenu", (const void *) glutChangeToSubMenu }, + { "glutRemoveMenuItem", (const void *) glutRemoveMenuItem }, + { "glutAttachMenu", (const void *) glutAttachMenu }, + { "glutDetachMenu", (const void *) glutDetachMenu }, + { "glutDisplayFunc", (const void *) glutDisplayFunc }, + { "glutReshapeFunc", (const void *) glutReshapeFunc }, + { "glutKeyboardFunc", (const void *) glutKeyboardFunc }, + { "glutMouseFunc", (const void *) glutMouseFunc }, + { "glutMotionFunc", (const void *) glutMotionFunc }, + { "glutPassiveMotionFunc", (const void *) glutPassiveMotionFunc }, + { "glutEntryFunc", (const void *) glutEntryFunc }, + { "glutVisibilityFunc", (const void *) glutVisibilityFunc }, + { "glutIdleFunc", (const void *) glutIdleFunc }, + { "glutTimerFunc", (const void *) glutTimerFunc }, + { "glutMenuStateFunc", (const void *) glutMenuStateFunc }, + { "glutSpecialFunc", (const void *) glutSpecialFunc }, + { "glutSpaceballMotionFunc", (const void *) glutSpaceballMotionFunc }, + { "glutSpaceballRotateFunc", (const void *) glutSpaceballRotateFunc }, + { "glutSpaceballButtonFunc", (const void *) glutSpaceballButtonFunc }, + { "glutButtonBoxFunc", (const void *) glutButtonBoxFunc }, + { "glutDialsFunc", (const void *) glutDialsFunc }, + { "glutTabletMotionFunc", (const void *) glutTabletMotionFunc }, + { "glutTabletButtonFunc", (const void *) glutTabletButtonFunc }, + { "glutMenuStatusFunc", (const void *) glutMenuStatusFunc }, + { "glutOverlayDisplayFunc", (const void *) glutOverlayDisplayFunc }, + { "glutWindowStatusFunc", (const void *) glutWindowStatusFunc }, + { "glutKeyboardUpFunc", (const void *) glutKeyboardUpFunc }, + { "glutSpecialUpFunc", (const void *) glutSpecialUpFunc }, + { "glutJoystickFunc", (const void *) glutJoystickFunc }, + { "glutSetColor", (const void *) glutSetColor }, + { "glutGetColor", (const void *) glutGetColor }, + { "glutCopyColormap", (const void *) glutCopyColormap }, + { "glutGet", (const void *) glutGet }, + { "glutDeviceGet", (const void *) glutDeviceGet }, + { "glutExtensionSupported", (const void *) glutExtensionSupported }, + { "glutGetModifiers", (const void *) glutGetModifiers }, + { "glutLayerGet", (const void *) glutLayerGet }, + { "glutGetProcAddress", (const void *) glutGetProcAddress }, + { "glutBitmapCharacter", (const void *) glutBitmapCharacter }, + { "glutBitmapWidth", (const void *) glutBitmapWidth }, + { "glutStrokeCharacter", (const void *) glutStrokeCharacter }, + { "glutStrokeWidth", (const void *) glutStrokeWidth }, + { "glutBitmapLength", (const void *) glutBitmapLength }, + { "glutStrokeLength", (const void *) glutStrokeLength }, + { "glutWireSphere", (const void *) glutWireSphere }, + { "glutSolidSphere", (const void *) glutSolidSphere }, + { "glutWireCone", (const void *) glutWireCone }, + { "glutSolidCone", (const void *) glutSolidCone }, + { "glutWireCube", (const void *) glutWireCube }, + { "glutSolidCube", (const void *) glutSolidCube }, + { "glutWireTorus", (const void *) glutWireTorus }, + { "glutSolidTorus", (const void *) glutSolidTorus }, + { "glutWireDodecahedron", (const void *) glutWireDodecahedron }, + { "glutSolidDodecahedron", (const void *) glutSolidDodecahedron }, + { "glutWireTeapot", (const void *) glutWireTeapot }, + { "glutSolidTeapot", (const void *) glutSolidTeapot }, + { "glutWireOctahedron", (const void *) glutWireOctahedron }, + { "glutSolidOctahedron", (const void *) glutSolidOctahedron }, + { "glutWireTetrahedron", (const void *) glutWireTetrahedron }, + { "glutSolidTetrahedron", (const void *) glutSolidTetrahedron }, + { "glutWireIcosahedron", (const void *) glutWireIcosahedron }, + { "glutSolidIcosahedron", (const void *) glutSolidIcosahedron }, + { "glutVideoResizeGet", (const void *) glutVideoResizeGet }, + { "glutSetupVideoResizing", (const void *) glutSetupVideoResizing }, + { "glutStopVideoResizing", (const void *) glutStopVideoResizing }, + { "glutVideoResize", (const void *) glutVideoResize }, + { "glutVideoPan", (const void *) glutVideoPan }, + { "glutReportErrors", (const void *) glutReportErrors }, + { "glutIgnoreKeyRepeat", (const void *) glutIgnoreKeyRepeat }, + { "glutSetKeyRepeat", (const void *) glutSetKeyRepeat }, + { "glutForceJoystickFunc", (const void *) glutForceJoystickFunc }, + { "glutGameModeString", (const void *) glutGameModeString }, + { "glutEnterGameMode", (const void *) glutEnterGameMode }, + { "glutLeaveGameMode", (const void *) glutLeaveGameMode }, + { "glutGameModeGet", (const void *) glutGameModeGet }, + { NULL, NULL } +}; + + +/* XXX This isn't an official GLUT function, yet */ +void * GLUTAPIENTRY +glutGetProcAddress(const char *procName) +{ + /* Try GLUT functions first */ + int i; + for (i = 0; glut_functions[i].name; i++) { + if (strcmp(glut_functions[i].name, procName) == 0) + return (void *) glut_functions[i].address; + } + + /* Try core GL functions */ +#if defined(_WIN32) + return (void *) wglGetProcAddress((LPCSTR) procName); + +#elif defined(__OS2PM__) + return (void *) wglGetProcAddress((char *) procName); +#elif defined(GLX_ARB_get_proc_address) + return (void *) glXGetProcAddressARB((const GLubyte *) procName); +#else + return NULL; +#endif +} + + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_fullscrn.cpp b/src/glut/os2/glut_fullscrn.cpp index d6efb68b3a..584aba4fd8 100644 --- a/src/glut/os2/glut_fullscrn.cpp +++ b/src/glut/os2/glut_fullscrn.cpp @@ -1,38 +1,38 @@ - -/* Copyright (c) Mark J. Kilgard, 1995, 1998. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include - -#include "glutint.h" - -/* CENTRY */ -void GLUTAPIENTRY -glutFullScreen(void) -{ - assert(!__glutCurrentWindow->parent); - IGNORE_IN_GAME_MODE(); -#if !defined(_WIN32) && !defined(__OS2PM__) - if (__glutMotifHints == None) { - __glutMotifHints = XSGIFastInternAtom(__glutDisplay, "_MOTIF_WM_HINTS", - SGI_XA__MOTIF_WM_HINTS, 0); - if (__glutMotifHints == None) { - __glutWarning("Could not intern X atom for _MOTIF_WM_HINTS."); - } - } -#endif - - __glutCurrentWindow->desiredX = 0; - __glutCurrentWindow->desiredY = 0; - __glutCurrentWindow->desiredWidth = __glutScreenWidth; - __glutCurrentWindow->desiredHeight = __glutScreenHeight; - __glutCurrentWindow->desiredConfMask |= CWX | CWY | CWWidth | CWHeight; - - __glutPutOnWorkList(__glutCurrentWindow, - GLUT_CONFIGURE_WORK | GLUT_FULL_SCREEN_WORK); -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1995, 1998. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include + +#include "glutint.h" + +/* CENTRY */ +void GLUTAPIENTRY +glutFullScreen(void) +{ + assert(!__glutCurrentWindow->parent); + IGNORE_IN_GAME_MODE(); +#if !defined(_WIN32) && !defined(__OS2PM__) + if (__glutMotifHints == None) { + __glutMotifHints = XSGIFastInternAtom(__glutDisplay, "_MOTIF_WM_HINTS", + SGI_XA__MOTIF_WM_HINTS, 0); + if (__glutMotifHints == None) { + __glutWarning("Could not intern X atom for _MOTIF_WM_HINTS."); + } + } +#endif + + __glutCurrentWindow->desiredX = 0; + __glutCurrentWindow->desiredY = 0; + __glutCurrentWindow->desiredWidth = __glutScreenWidth; + __glutCurrentWindow->desiredHeight = __glutScreenHeight; + __glutCurrentWindow->desiredConfMask |= CWX | CWY | CWWidth | CWHeight; + + __glutPutOnWorkList(__glutCurrentWindow, + GLUT_CONFIGURE_WORK | GLUT_FULL_SCREEN_WORK); +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_get.cpp b/src/glut/os2/glut_get.cpp index c846f48658..8921e9ab03 100644 --- a/src/glut/os2/glut_get.cpp +++ b/src/glut/os2/glut_get.cpp @@ -1,232 +1,232 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1997, 1998. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include -#include /* SunOS 4 needs NULL defined for GETTIMEOFDAY macro. */ -#include "glutint.h" - -/* CENTRY */ -int GLUTAPIENTRY -glutGet(GLenum param) -{ - Window win, root; - int x, y, value; - unsigned int width, height, border, depth; - - switch (param) { - case GLUT_INIT_WINDOW_X: - return __glutInitX; - case GLUT_INIT_WINDOW_Y: - return __glutInitY; - case GLUT_INIT_WINDOW_WIDTH: - return __glutInitWidth; - case GLUT_INIT_WINDOW_HEIGHT: - return __glutInitHeight; - case GLUT_INIT_DISPLAY_MODE: - return __glutDisplayMode; - case GLUT_WINDOW_X: - XTranslateCoordinates(__glutDisplay, __glutCurrentWindow->win, - __glutRoot, 0, 0, &x, &y, &win); - return x; - case GLUT_WINDOW_Y: - XTranslateCoordinates(__glutDisplay, __glutCurrentWindow->win, - __glutRoot, 0, 0, &x, &y, &win); - return y; - case GLUT_WINDOW_WIDTH: - if (!__glutCurrentWindow->reshape) { - XGetGeometry(__glutDisplay, __glutCurrentWindow->win, - &root, &x, &y, - &width, &height, &border, &depth); - return width; - } - return __glutCurrentWindow->width; - case GLUT_WINDOW_HEIGHT: - if (!__glutCurrentWindow->reshape) { - XGetGeometry(__glutDisplay, __glutCurrentWindow->win, - &root, &x, &y, - &width, &height, &border, &depth); - return height; - } - return __glutCurrentWindow->height; -#ifdef __OS2__ - #define GET_CONFIG(attrib) \ - { if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) \ - glXGetConfig( __glutCurrentWindow->vis, attrib, &value); \ - else \ - glXGetConfig(__glutCurrentWindow->overlay->vis, attrib, &value); \ - } \ - -#else - -#define GET_CONFIG(attrib) { \ - if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { \ - glXGetConfig(__glutDisplay, __glutCurrentWindow->vis, \ - attrib, &value); \ - } else { \ - glXGetConfig(__glutDisplay, __glutCurrentWindow->overlay->vis, \ - attrib, &value); \ - } \ -} -#endif - - case GLUT_WINDOW_BUFFER_SIZE: - GET_CONFIG(GLX_BUFFER_SIZE); - return value; - case GLUT_WINDOW_STENCIL_SIZE: - GET_CONFIG(GLX_STENCIL_SIZE); - return value; - case GLUT_WINDOW_DEPTH_SIZE: - GET_CONFIG(GLX_DEPTH_SIZE); - return value; - case GLUT_WINDOW_RED_SIZE: - GET_CONFIG(GLX_RED_SIZE); - return value; - case GLUT_WINDOW_GREEN_SIZE: - GET_CONFIG(GLX_GREEN_SIZE); - return value; - case GLUT_WINDOW_BLUE_SIZE: - GET_CONFIG(GLX_BLUE_SIZE); - return value; - case GLUT_WINDOW_ALPHA_SIZE: - GET_CONFIG(GLX_ALPHA_SIZE); - return value; - case GLUT_WINDOW_ACCUM_RED_SIZE: - GET_CONFIG(GLX_ACCUM_RED_SIZE); - return value; - case GLUT_WINDOW_ACCUM_GREEN_SIZE: - GET_CONFIG(GLX_ACCUM_GREEN_SIZE); - return value; - case GLUT_WINDOW_ACCUM_BLUE_SIZE: - GET_CONFIG(GLX_ACCUM_BLUE_SIZE); - return value; - case GLUT_WINDOW_ACCUM_ALPHA_SIZE: - GET_CONFIG(GLX_ACCUM_ALPHA_SIZE); - return value; - case GLUT_WINDOW_DOUBLEBUFFER: - GET_CONFIG(GLX_DOUBLEBUFFER); - return value; - case GLUT_WINDOW_RGBA: - GET_CONFIG(GLX_RGBA); - return value; - case GLUT_WINDOW_COLORMAP_SIZE: - GET_CONFIG(GLX_RGBA); - if (value) { - return 0; - } else { -#if defined(_WIN32) || defined(__OS2__) - /* KLUDGE: we always assume 256 colors in CI mode on - Win32 */ - return 256; -#else - if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { - return __glutCurrentWindow->vis->visual->map_entries; - } else { - return __glutCurrentWindow->overlay->vis->visual->map_entries; - } -#endif /* _WIN32 */ - } - case GLUT_WINDOW_PARENT: - return __glutCurrentWindow->parent ? - __glutCurrentWindow->parent->num + 1 : 0; - case GLUT_WINDOW_NUM_CHILDREN: - { - int num = 0; - GLUTwindow *children = __glutCurrentWindow->children; - - while (children) { - num++; - children = children->siblings; - } - return num; - } - case GLUT_WINDOW_NUM_SAMPLES: -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIS_multisample) - if (__glutIsSupportedByGLX("GLX_SGIS_multisample")) { - GET_CONFIG(GLX_SAMPLES_SGIS); - return value; - } else { - return 0; - } -#else - /* Independent of GLX server support, multisampling not - supported by GLX client-side. */ - return 0; -#endif - case GLUT_WINDOW_STEREO: - GET_CONFIG(GLX_STEREO); - return value; - case GLUT_WINDOW_CURSOR: - return __glutCurrentWindow->cursor; - case GLUT_SCREEN_WIDTH: - return DisplayWidth(__glutDisplay, __glutScreen); - case GLUT_SCREEN_HEIGHT: - return DisplayHeight(__glutDisplay, __glutScreen); - case GLUT_SCREEN_WIDTH_MM: - return DisplayWidthMM(__glutDisplay, __glutScreen); - case GLUT_SCREEN_HEIGHT_MM: - return DisplayHeightMM(__glutDisplay, __glutScreen); - case GLUT_MENU_NUM_ITEMS: - return __glutCurrentMenu->num; - case GLUT_DISPLAY_MODE_POSSIBLE: - { - XVisualInfo *vi; - Bool dummy, visAlloced; - void *fbc; - -#if defined(_WIN32) - /* Our fake glXChooseVisual (which is called by - __glutDetermineVisual) needs an HDC to work with, so grab one - from the "root" window. */ - XHDC = GetDC(GetDesktopWindow()); -#endif - vi = __glutDetermineWindowVisual(&dummy, &visAlloced, &fbc); -#if defined(_WIN32) - ReleaseDC(GetDesktopWindow(), XHDC); -#endif - if (vi) { - if (visAlloced) - XFree(vi); - return 1; - } - return 0; - } - case GLUT_ELAPSED_TIME: - { -#ifdef OLD_VMS - struct timeval6 elapsed, beginning, now; -#else - struct timeval elapsed, beginning, now; -#endif - - __glutInitTime(&beginning); - GETTIMEOFDAY(&now); - TIMEDELTA(elapsed, now, beginning); - /* Return elapsed milliseconds. */ -#if defined(__vms) && ( __VMS_VER < 70000000 ) - return (int) (elapsed.val / TICKS_PER_MILLISECOND); -#else - return (int) ((elapsed.tv_sec * 1000) + (elapsed.tv_usec / 1000)); -#endif - } - case GLUT_WINDOW_FORMAT_ID: -#if defined(__OS2__) - return wglGetPixelFormat(__glutCurrentWindow->hdc); -#elif defined(_WIN32) - return GetPixelFormat(__glutCurrentWindow->hdc); -#else - if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { - return (int) __glutCurrentWindow->vis->visualid; - } else { - return (int) __glutCurrentWindow->overlay->vis->visualid; - } -#endif - default: - __glutWarning("invalid glutGet parameter: %d", param); - return -1; - } -} -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994, 1997, 1998. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include +#include /* SunOS 4 needs NULL defined for GETTIMEOFDAY macro. */ +#include "glutint.h" + +/* CENTRY */ +int GLUTAPIENTRY +glutGet(GLenum param) +{ + Window win, root; + int x, y, value; + unsigned int width, height, border, depth; + + switch (param) { + case GLUT_INIT_WINDOW_X: + return __glutInitX; + case GLUT_INIT_WINDOW_Y: + return __glutInitY; + case GLUT_INIT_WINDOW_WIDTH: + return __glutInitWidth; + case GLUT_INIT_WINDOW_HEIGHT: + return __glutInitHeight; + case GLUT_INIT_DISPLAY_MODE: + return __glutDisplayMode; + case GLUT_WINDOW_X: + XTranslateCoordinates(__glutDisplay, __glutCurrentWindow->win, + __glutRoot, 0, 0, &x, &y, &win); + return x; + case GLUT_WINDOW_Y: + XTranslateCoordinates(__glutDisplay, __glutCurrentWindow->win, + __glutRoot, 0, 0, &x, &y, &win); + return y; + case GLUT_WINDOW_WIDTH: + if (!__glutCurrentWindow->reshape) { + XGetGeometry(__glutDisplay, __glutCurrentWindow->win, + &root, &x, &y, + &width, &height, &border, &depth); + return width; + } + return __glutCurrentWindow->width; + case GLUT_WINDOW_HEIGHT: + if (!__glutCurrentWindow->reshape) { + XGetGeometry(__glutDisplay, __glutCurrentWindow->win, + &root, &x, &y, + &width, &height, &border, &depth); + return height; + } + return __glutCurrentWindow->height; +#ifdef __OS2__ + #define GET_CONFIG(attrib) \ + { if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) \ + glXGetConfig( __glutCurrentWindow->vis, attrib, &value); \ + else \ + glXGetConfig(__glutCurrentWindow->overlay->vis, attrib, &value); \ + } \ + +#else + +#define GET_CONFIG(attrib) { \ + if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { \ + glXGetConfig(__glutDisplay, __glutCurrentWindow->vis, \ + attrib, &value); \ + } else { \ + glXGetConfig(__glutDisplay, __glutCurrentWindow->overlay->vis, \ + attrib, &value); \ + } \ +} +#endif + + case GLUT_WINDOW_BUFFER_SIZE: + GET_CONFIG(GLX_BUFFER_SIZE); + return value; + case GLUT_WINDOW_STENCIL_SIZE: + GET_CONFIG(GLX_STENCIL_SIZE); + return value; + case GLUT_WINDOW_DEPTH_SIZE: + GET_CONFIG(GLX_DEPTH_SIZE); + return value; + case GLUT_WINDOW_RED_SIZE: + GET_CONFIG(GLX_RED_SIZE); + return value; + case GLUT_WINDOW_GREEN_SIZE: + GET_CONFIG(GLX_GREEN_SIZE); + return value; + case GLUT_WINDOW_BLUE_SIZE: + GET_CONFIG(GLX_BLUE_SIZE); + return value; + case GLUT_WINDOW_ALPHA_SIZE: + GET_CONFIG(GLX_ALPHA_SIZE); + return value; + case GLUT_WINDOW_ACCUM_RED_SIZE: + GET_CONFIG(GLX_ACCUM_RED_SIZE); + return value; + case GLUT_WINDOW_ACCUM_GREEN_SIZE: + GET_CONFIG(GLX_ACCUM_GREEN_SIZE); + return value; + case GLUT_WINDOW_ACCUM_BLUE_SIZE: + GET_CONFIG(GLX_ACCUM_BLUE_SIZE); + return value; + case GLUT_WINDOW_ACCUM_ALPHA_SIZE: + GET_CONFIG(GLX_ACCUM_ALPHA_SIZE); + return value; + case GLUT_WINDOW_DOUBLEBUFFER: + GET_CONFIG(GLX_DOUBLEBUFFER); + return value; + case GLUT_WINDOW_RGBA: + GET_CONFIG(GLX_RGBA); + return value; + case GLUT_WINDOW_COLORMAP_SIZE: + GET_CONFIG(GLX_RGBA); + if (value) { + return 0; + } else { +#if defined(_WIN32) || defined(__OS2__) + /* KLUDGE: we always assume 256 colors in CI mode on + Win32 */ + return 256; +#else + if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { + return __glutCurrentWindow->vis->visual->map_entries; + } else { + return __glutCurrentWindow->overlay->vis->visual->map_entries; + } +#endif /* _WIN32 */ + } + case GLUT_WINDOW_PARENT: + return __glutCurrentWindow->parent ? + __glutCurrentWindow->parent->num + 1 : 0; + case GLUT_WINDOW_NUM_CHILDREN: + { + int num = 0; + GLUTwindow *children = __glutCurrentWindow->children; + + while (children) { + num++; + children = children->siblings; + } + return num; + } + case GLUT_WINDOW_NUM_SAMPLES: +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIS_multisample) + if (__glutIsSupportedByGLX("GLX_SGIS_multisample")) { + GET_CONFIG(GLX_SAMPLES_SGIS); + return value; + } else { + return 0; + } +#else + /* Independent of GLX server support, multisampling not + supported by GLX client-side. */ + return 0; +#endif + case GLUT_WINDOW_STEREO: + GET_CONFIG(GLX_STEREO); + return value; + case GLUT_WINDOW_CURSOR: + return __glutCurrentWindow->cursor; + case GLUT_SCREEN_WIDTH: + return DisplayWidth(__glutDisplay, __glutScreen); + case GLUT_SCREEN_HEIGHT: + return DisplayHeight(__glutDisplay, __glutScreen); + case GLUT_SCREEN_WIDTH_MM: + return DisplayWidthMM(__glutDisplay, __glutScreen); + case GLUT_SCREEN_HEIGHT_MM: + return DisplayHeightMM(__glutDisplay, __glutScreen); + case GLUT_MENU_NUM_ITEMS: + return __glutCurrentMenu->num; + case GLUT_DISPLAY_MODE_POSSIBLE: + { + XVisualInfo *vi; + Bool dummy, visAlloced; + void *fbc; + +#if defined(_WIN32) + /* Our fake glXChooseVisual (which is called by + __glutDetermineVisual) needs an HDC to work with, so grab one + from the "root" window. */ + XHDC = GetDC(GetDesktopWindow()); +#endif + vi = __glutDetermineWindowVisual(&dummy, &visAlloced, &fbc); +#if defined(_WIN32) + ReleaseDC(GetDesktopWindow(), XHDC); +#endif + if (vi) { + if (visAlloced) + XFree(vi); + return 1; + } + return 0; + } + case GLUT_ELAPSED_TIME: + { +#ifdef OLD_VMS + struct timeval6 elapsed, beginning, now; +#else + struct timeval elapsed, beginning, now; +#endif + + __glutInitTime(&beginning); + GETTIMEOFDAY(&now); + TIMEDELTA(elapsed, now, beginning); + /* Return elapsed milliseconds. */ +#if defined(__vms) && ( __VMS_VER < 70000000 ) + return (int) (elapsed.val / TICKS_PER_MILLISECOND); +#else + return (int) ((elapsed.tv_sec * 1000) + (elapsed.tv_usec / 1000)); +#endif + } + case GLUT_WINDOW_FORMAT_ID: +#if defined(__OS2__) + return wglGetPixelFormat(__glutCurrentWindow->hdc); +#elif defined(_WIN32) + return GetPixelFormat(__glutCurrentWindow->hdc); +#else + if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { + return (int) __glutCurrentWindow->vis->visualid; + } else { + return (int) __glutCurrentWindow->overlay->vis->visualid; + } +#endif + default: + __glutWarning("invalid glutGet parameter: %d", param); + return -1; + } +} +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_hel10.cpp b/src/glut/os2/glut_hel10.cpp index 6409f6181a..a9f08ff133 100644 --- a/src/glut/os2/glut_hel10.cpp +++ b/src/glut/os2/glut_hel10.cpp @@ -1,1781 +1,1781 @@ - -/* GENERATED FILE -- DO NOT MODIFY */ - -#define glutBitmapHelvetica10 XXX -#include "glutbitmap.h" -#undef glutBitmapHelvetica10 - -/* char: 0xff */ - -static const GLubyte ch255data[] = { -0x80,0x40,0x40,0x60,0xa0,0xa0,0x90,0x90,0x0,0x50, -}; - -static const BitmapCharRec ch255 = {4,10,0,2,5,ch255data}; - -/* char: 0xfe */ - -static const GLubyte ch254data[] = { -0x80,0x80,0xb0,0xc8,0x88,0x88,0xc8,0xb0,0x80,0x80, -}; - -static const BitmapCharRec ch254 = {5,10,0,2,6,ch254data}; - -/* char: 0xfd */ - -static const GLubyte ch253data[] = { -0x80,0x40,0x40,0x60,0xa0,0xa0,0x90,0x90,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch253 = {4,11,0,2,5,ch253data}; - -/* char: 0xfc */ - -static const GLubyte ch252data[] = { -0x70,0x90,0x90,0x90,0x90,0x90,0x0,0x50, -}; - -static const BitmapCharRec ch252 = {4,8,0,0,5,ch252data}; - -/* char: 0xfb */ - -static const GLubyte ch251data[] = { -0x70,0x90,0x90,0x90,0x90,0x90,0x0,0x50,0x20, -}; - -static const BitmapCharRec ch251 = {4,9,0,0,5,ch251data}; - -/* char: 0xfa */ - -static const GLubyte ch250data[] = { -0x70,0x90,0x90,0x90,0x90,0x90,0x0,0x40,0x20, -}; - -static const BitmapCharRec ch250 = {4,9,0,0,5,ch250data}; - -/* char: 0xf9 */ - -static const GLubyte ch249data[] = { -0x70,0x90,0x90,0x90,0x90,0x90,0x0,0x20,0x40, -}; - -static const BitmapCharRec ch249 = {4,9,0,0,5,ch249data}; - -/* char: 0xf8 */ - -static const GLubyte ch248data[] = { -0x70,0x88,0xc8,0xa8,0x98,0x74, -}; - -static const BitmapCharRec ch248 = {6,6,0,0,6,ch248data}; - -/* char: 0xf7 */ - -static const GLubyte ch247data[] = { -0x20,0x0,0xf8,0x0,0x20, -}; - -static const BitmapCharRec ch247 = {5,5,0,-1,6,ch247data}; - -/* char: 0xf6 */ - -static const GLubyte ch246data[] = { -0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x50, -}; - -static const BitmapCharRec ch246 = {5,8,0,0,6,ch246data}; - -/* char: 0xf5 */ - -static const GLubyte ch245data[] = { -0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch245 = {5,9,0,0,6,ch245data}; - -/* char: 0xf4 */ - -static const GLubyte ch244data[] = { -0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x50,0x20, -}; - -static const BitmapCharRec ch244 = {5,9,0,0,6,ch244data}; - -/* char: 0xf3 */ - -static const GLubyte ch243data[] = { -0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch243 = {5,9,0,0,6,ch243data}; - -/* char: 0xf2 */ - -static const GLubyte ch242data[] = { -0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x20,0x40, -}; - -static const BitmapCharRec ch242 = {5,9,0,0,6,ch242data}; - -/* char: 0xf1 */ - -static const GLubyte ch241data[] = { -0x90,0x90,0x90,0x90,0x90,0xe0,0x0,0xa0,0x50, -}; - -static const BitmapCharRec ch241 = {4,9,0,0,5,ch241data}; - -/* char: 0xf0 */ - -static const GLubyte ch240data[] = { -0x70,0x88,0x88,0x88,0x88,0x78,0x90,0x60,0x50, -}; - -static const BitmapCharRec ch240 = {5,9,0,0,6,ch240data}; - -/* char: 0xef */ - -static const GLubyte ch239data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0, -}; - -static const BitmapCharRec ch239 = {3,8,0,0,2,ch239data}; - -/* char: 0xee */ - -static const GLubyte ch238data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0,0x40, -}; - -static const BitmapCharRec ch238 = {3,9,1,0,2,ch238data}; - -/* char: 0xed */ - -static const GLubyte ch237data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x40, -}; - -static const BitmapCharRec ch237 = {2,9,0,0,2,ch237data}; - -/* char: 0xec */ - -static const GLubyte ch236data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40,0x80, -}; - -static const BitmapCharRec ch236 = {2,9,1,0,2,ch236data}; - -/* char: 0xeb */ - -static const GLubyte ch235data[] = { -0x60,0x90,0x80,0xf0,0x90,0x60,0x0,0x50, -}; - -static const BitmapCharRec ch235 = {4,8,0,0,5,ch235data}; - -/* char: 0xea */ - -static const GLubyte ch234data[] = { -0x60,0x90,0x80,0xf0,0x90,0x60,0x0,0x50,0x20, -}; - -static const BitmapCharRec ch234 = {4,9,0,0,5,ch234data}; - -/* char: 0xe9 */ - -static const GLubyte ch233data[] = { -0x60,0x90,0x80,0xf0,0x90,0x60,0x0,0x40,0x20, -}; - -static const BitmapCharRec ch233 = {4,9,0,0,5,ch233data}; - -/* char: 0xe8 */ - -static const GLubyte ch232data[] = { -0x60,0x90,0x80,0xf0,0x90,0x60,0x0,0x20,0x40, -}; - -static const BitmapCharRec ch232 = {4,9,0,0,5,ch232data}; - -/* char: 0xe7 */ - -static const GLubyte ch231data[] = { -0x60,0x20,0x60,0x90,0x80,0x80,0x90,0x60, -}; - -static const BitmapCharRec ch231 = {4,8,0,2,5,ch231data}; - -/* char: 0xe6 */ - -static const GLubyte ch230data[] = { -0x6c,0x92,0x90,0x7e,0x12,0xec, -}; - -static const BitmapCharRec ch230 = {7,6,0,0,8,ch230data}; - -/* char: 0xe5 */ - -static const GLubyte ch229data[] = { -0x68,0x90,0x90,0x70,0x10,0xe0,0x20,0x50,0x20, -}; - -static const BitmapCharRec ch229 = {5,9,0,0,5,ch229data}; - -/* char: 0xe4 */ - -static const GLubyte ch228data[] = { -0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0x50, -}; - -static const BitmapCharRec ch228 = {5,8,0,0,5,ch228data}; - -/* char: 0xe3 */ - -static const GLubyte ch227data[] = { -0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0xa0,0x50, -}; - -static const BitmapCharRec ch227 = {5,9,0,0,5,ch227data}; - -/* char: 0xe2 */ - -static const GLubyte ch226data[] = { -0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0x50,0x20, -}; - -static const BitmapCharRec ch226 = {5,9,0,0,5,ch226data}; - -/* char: 0xe1 */ - -static const GLubyte ch225data[] = { -0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch225 = {5,9,0,0,5,ch225data}; - -/* char: 0xe0 */ - -static const GLubyte ch224data[] = { -0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0x20,0x40, -}; - -static const BitmapCharRec ch224 = {5,9,0,0,5,ch224data}; - -/* char: 0xdf */ - -static const GLubyte ch223data[] = { -0xa0,0x90,0x90,0x90,0xa0,0x90,0x90,0x60, -}; - -static const BitmapCharRec ch223 = {4,8,0,0,5,ch223data}; - -/* char: 0xde */ - -static const GLubyte ch222data[] = { -0x80,0x80,0xf0,0x88,0x88,0xf0,0x80,0x80, -}; - -static const BitmapCharRec ch222 = {5,8,-1,0,7,ch222data}; - -/* char: 0xdd */ - -static const GLubyte ch221data[] = { -0x10,0x10,0x10,0x28,0x28,0x44,0x44,0x82,0x0,0x10,0x8, -}; - -static const BitmapCharRec ch221 = {7,11,0,0,7,ch221data}; - -/* char: 0xdc */ - -static const GLubyte ch220data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x48, -}; - -static const BitmapCharRec ch220 = {6,10,-1,0,8,ch220data}; - -/* char: 0xdb */ - -static const GLubyte ch219data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x28,0x10, -}; - -static const BitmapCharRec ch219 = {6,11,-1,0,8,ch219data}; - -/* char: 0xda */ - -static const GLubyte ch218data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch218 = {6,11,-1,0,8,ch218data}; - -/* char: 0xd9 */ - -static const GLubyte ch217data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch217 = {6,11,-1,0,8,ch217data}; - -/* char: 0xd8 */ - -static const GLubyte ch216data[] = { -0x80,0x78,0xc4,0xa4,0xa4,0x94,0x94,0x8c,0x78,0x4, -}; - -static const BitmapCharRec ch216 = {6,10,-1,1,8,ch216data}; - -/* char: 0xd7 */ - -static const GLubyte ch215data[] = { -0x88,0x50,0x20,0x50,0x88, -}; - -static const BitmapCharRec ch215 = {5,5,0,-1,6,ch215data}; - -/* char: 0xd6 */ - -static const GLubyte ch214data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x48, -}; - -static const BitmapCharRec ch214 = {6,10,-1,0,8,ch214data}; - -/* char: 0xd5 */ - -static const GLubyte ch213data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch213 = {6,11,-1,0,8,ch213data}; - -/* char: 0xd4 */ - -static const GLubyte ch212data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x28,0x10, -}; - -static const BitmapCharRec ch212 = {6,11,-1,0,8,ch212data}; - -/* char: 0xd3 */ - -static const GLubyte ch211data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x10,0x8, -}; - -static const BitmapCharRec ch211 = {6,11,-1,0,8,ch211data}; - -/* char: 0xd2 */ - -static const GLubyte ch210data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch210 = {6,11,-1,0,8,ch210data}; - -/* char: 0xd1 */ - -static const GLubyte ch209data[] = { -0x8c,0x8c,0x94,0x94,0xa4,0xa4,0xc4,0xc4,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch209 = {6,11,-1,0,8,ch209data}; - -/* char: 0xd0 */ - -static const GLubyte ch208data[] = { -0x78,0x44,0x42,0x42,0xf2,0x42,0x44,0x78, -}; - -static const BitmapCharRec ch208 = {7,8,0,0,8,ch208data}; - -/* char: 0xcf */ - -static const GLubyte ch207data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0, -}; - -static const BitmapCharRec ch207 = {3,10,0,0,3,ch207data}; - -/* char: 0xce */ - -static const GLubyte ch206data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0,0x40, -}; - -static const BitmapCharRec ch206 = {3,11,0,0,3,ch206data}; - -/* char: 0xcd */ - -static const GLubyte ch205data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x40, -}; - -static const BitmapCharRec ch205 = {2,11,-1,0,3,ch205data}; - -/* char: 0xcc */ - -static const GLubyte ch204data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40,0x80, -}; - -static const BitmapCharRec ch204 = {2,11,0,0,3,ch204data}; - -/* char: 0xcb */ - -static const GLubyte ch203data[] = { -0xf8,0x80,0x80,0x80,0xf8,0x80,0x80,0xf8,0x0,0x50, -}; - -static const BitmapCharRec ch203 = {5,10,-1,0,7,ch203data}; - -/* char: 0xca */ - -static const GLubyte ch202data[] = { -0xf8,0x80,0x80,0xf8,0x80,0x80,0x80,0xf8,0x0,0x50,0x20, -}; - -static const BitmapCharRec ch202 = {5,11,-1,0,7,ch202data}; - -/* char: 0xc9 */ - -static const GLubyte ch201data[] = { -0xf8,0x80,0x80,0x80,0xf8,0x80,0x80,0xf8,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch201 = {5,11,-1,0,7,ch201data}; - -/* char: 0xc8 */ - -static const GLubyte ch200data[] = { -0xf8,0x80,0x80,0x80,0xf8,0x80,0x80,0xf8,0x0,0x20,0x40, -}; - -static const BitmapCharRec ch200 = {5,11,-1,0,7,ch200data}; - -/* char: 0xc7 */ - -static const GLubyte ch199data[] = { -0x30,0x10,0x78,0x84,0x80,0x80,0x80,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch199 = {6,10,-1,2,8,ch199data}; - -/* char: 0xc6 */ - -static const GLubyte ch198data[] = { -0x8f,0x80,0x88,0x0,0x78,0x0,0x48,0x0,0x2f,0x80,0x28,0x0,0x18,0x0,0x1f,0x80, -}; - -static const BitmapCharRec ch198 = {9,8,0,0,10,ch198data}; - -/* char: 0xc5 */ - -static const GLubyte ch197data[] = { -0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x10,0x28,0x10, -}; - -static const BitmapCharRec ch197 = {7,11,0,0,7,ch197data}; - -/* char: 0xc4 */ - -static const GLubyte ch196data[] = { -0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x28, -}; - -static const BitmapCharRec ch196 = {7,10,0,0,7,ch196data}; - -/* char: 0xc3 */ - -static const GLubyte ch195data[] = { -0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x28,0x14, -}; - -static const BitmapCharRec ch195 = {7,11,0,0,7,ch195data}; - -/* char: 0xc2 */ - -static const GLubyte ch194data[] = { -0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x28,0x10, -}; - -static const BitmapCharRec ch194 = {7,11,0,0,7,ch194data}; - -/* char: 0xc1 */ - -static const GLubyte ch193data[] = { -0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x10,0x8, -}; - -static const BitmapCharRec ch193 = {7,11,0,0,7,ch193data}; - -/* char: 0xc0 */ - -static const GLubyte ch192data[] = { -0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch192 = {7,11,0,0,7,ch192data}; - -/* char: 0xbf */ - -static const GLubyte ch191data[] = { -0x60,0x90,0x80,0x40,0x20,0x20,0x0,0x20, -}; - -static const BitmapCharRec ch191 = {4,8,-1,2,6,ch191data}; - -/* char: 0xbe */ - -static const GLubyte ch190data[] = { -0x21,0x0,0x17,0x80,0x13,0x0,0x9,0x0,0xc8,0x0,0x24,0x0,0x44,0x0,0xe2,0x0, -}; - -static const BitmapCharRec ch190 = {9,8,0,0,9,ch190data}; - -/* char: 0xbd */ - -static const GLubyte ch189data[] = { -0x27,0x12,0x15,0xb,0x48,0x44,0xc4,0x42, -}; - -static const BitmapCharRec ch189 = {8,8,0,0,9,ch189data}; - -/* char: 0xbc */ - -static const GLubyte ch188data[] = { -0x21,0x0,0x17,0x80,0x13,0x0,0x9,0x0,0x48,0x0,0x44,0x0,0xc4,0x0,0x42,0x0, -}; - -static const BitmapCharRec ch188 = {9,8,0,0,9,ch188data}; - -/* char: 0xbb */ - -static const GLubyte ch187data[] = { -0xa0,0x50,0x28,0x50,0xa0, -}; - -static const BitmapCharRec ch187 = {5,5,0,0,6,ch187data}; - -/* char: 0xba */ - -static const GLubyte ch186data[] = { -0xe0,0x0,0xe0,0xa0,0xe0, -}; - -static const BitmapCharRec ch186 = {3,5,0,-3,4,ch186data}; - -/* char: 0xb9 */ - -static const GLubyte ch185data[] = { -0x40,0x40,0xc0,0x40, -}; - -static const BitmapCharRec ch185 = {2,4,0,-3,3,ch185data}; - -/* char: 0xb8 */ - -static const GLubyte ch184data[] = { -0xc0,0x40, -}; - -static const BitmapCharRec ch184 = {2,2,0,2,3,ch184data}; - -/* char: 0xb7 */ - -static const GLubyte ch183data[] = { -0xc0, -}; - -static const BitmapCharRec ch183 = {2,1,0,-3,3,ch183data}; - -/* char: 0xb6 */ - -static const GLubyte ch182data[] = { -0x28,0x28,0x28,0x28,0x28,0x68,0xe8,0xe8,0xe8,0x7c, -}; - -static const BitmapCharRec ch182 = {6,10,0,2,6,ch182data}; - -/* char: 0xb5 */ - -static const GLubyte ch181data[] = { -0x80,0x80,0xf0,0x90,0x90,0x90,0x90,0x90, -}; - -static const BitmapCharRec ch181 = {4,8,0,2,5,ch181data}; - -/* char: 0xb4 */ - -static const GLubyte ch180data[] = { -0x80,0x40, -}; - -static const BitmapCharRec ch180 = {2,2,0,-6,3,ch180data}; - -/* char: 0xb3 */ - -static const GLubyte ch179data[] = { -0xc0,0x20,0x40,0xe0, -}; - -static const BitmapCharRec ch179 = {3,4,0,-3,3,ch179data}; - -/* char: 0xb2 */ - -static const GLubyte ch178data[] = { -0xe0,0x40,0xa0,0x60, -}; - -static const BitmapCharRec ch178 = {3,4,0,-3,3,ch178data}; - -/* char: 0xb1 */ - -static const GLubyte ch177data[] = { -0xf8,0x0,0x20,0x20,0xf8,0x20,0x20, -}; - -static const BitmapCharRec ch177 = {5,7,0,0,6,ch177data}; - -/* char: 0xb0 */ - -static const GLubyte ch176data[] = { -0x60,0x90,0x90,0x60, -}; - -static const BitmapCharRec ch176 = {4,4,0,-3,4,ch176data}; - -/* char: 0xaf */ - -static const GLubyte ch175data[] = { -0xe0, -}; - -static const BitmapCharRec ch175 = {3,1,0,-7,3,ch175data}; - -/* char: 0xae */ - -static const GLubyte ch174data[] = { -0x38,0x44,0xaa,0xb2,0xba,0x44,0x38, -}; - -static const BitmapCharRec ch174 = {7,7,-1,0,9,ch174data}; - -/* char: 0xad */ - -static const GLubyte ch173data[] = { -0xe0, -}; - -static const BitmapCharRec ch173 = {3,1,0,-3,4,ch173data}; - -/* char: 0xac */ - -static const GLubyte ch172data[] = { -0x8,0x8,0xf8, -}; - -static const BitmapCharRec ch172 = {5,3,-1,-2,7,ch172data}; - -/* char: 0xab */ - -static const GLubyte ch171data[] = { -0x28,0x50,0xa0,0x50,0x28, -}; - -static const BitmapCharRec ch171 = {5,5,0,0,6,ch171data}; - -/* char: 0xaa */ - -static const GLubyte ch170data[] = { -0xe0,0x0,0xa0,0x20,0xe0, -}; - -static const BitmapCharRec ch170 = {3,5,0,-3,4,ch170data}; - -/* char: 0xa9 */ - -static const GLubyte ch169data[] = { -0x38,0x44,0x9a,0xa2,0x9a,0x44,0x38, -}; - -static const BitmapCharRec ch169 = {7,7,-1,0,9,ch169data}; - -/* char: 0xa8 */ - -static const GLubyte ch168data[] = { -0xa0, -}; - -static const BitmapCharRec ch168 = {3,1,0,-7,3,ch168data}; - -/* char: 0xa7 */ - -static const GLubyte ch167data[] = { -0x70,0x88,0x18,0x70,0xc8,0x98,0x70,0xc0,0x88,0x70, -}; - -static const BitmapCharRec ch167 = {5,10,0,2,6,ch167data}; - -/* char: 0xa6 */ - -static const GLubyte ch166data[] = { -0x80,0x80,0x80,0x80,0x0,0x0,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch166 = {1,10,-1,2,3,ch166data}; - -/* char: 0xa5 */ - -static const GLubyte ch165data[] = { -0x20,0xf8,0x20,0xf8,0x50,0x50,0x88,0x88, -}; - -static const BitmapCharRec ch165 = {5,8,0,0,6,ch165data}; - -/* char: 0xa4 */ - -static const GLubyte ch164data[] = { -0x90,0x60,0x90,0x90,0x60,0x90, -}; - -static const BitmapCharRec ch164 = {4,6,0,-1,5,ch164data}; - -/* char: 0xa3 */ - -static const GLubyte ch163data[] = { -0xb0,0x48,0x40,0x40,0xe0,0x40,0x48,0x30, -}; - -static const BitmapCharRec ch163 = {5,8,0,0,6,ch163data}; - -/* char: 0xa2 */ - -static const GLubyte ch162data[] = { -0x40,0x70,0xa8,0xa0,0xa0,0xa8,0x70,0x10, -}; - -static const BitmapCharRec ch162 = {5,8,0,1,6,ch162data}; - -/* char: 0xa1 */ - -static const GLubyte ch161data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, -}; - -static const BitmapCharRec ch161 = {1,8,-1,2,3,ch161data}; - -/* char: 0xa0 */ - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch160data[] = { 0x0 }; -static const BitmapCharRec ch160 = {1,1,0,0,3,ch160data}; -#else -static const BitmapCharRec ch160 = {0,0,0,0,3,0}; -#endif - -/* char: 0x7e '~' */ - -static const GLubyte ch126data[] = { -0x98,0x64, -}; - -static const BitmapCharRec ch126 = {6,2,0,-3,7,ch126data}; - -/* char: 0x7d '}' */ - -static const GLubyte ch125data[] = { -0x80,0x40,0x40,0x40,0x40,0x20,0x40,0x40,0x40,0x80, -}; - -static const BitmapCharRec ch125 = {3,10,0,2,3,ch125data}; - -/* char: 0x7c '|' */ - -static const GLubyte ch124data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch124 = {1,10,-1,2,3,ch124data}; - -/* char: 0x7b '{' */ - -static const GLubyte ch123data[] = { -0x20,0x40,0x40,0x40,0x40,0x80,0x40,0x40,0x40,0x20, -}; - -static const BitmapCharRec ch123 = {3,10,0,2,3,ch123data}; - -/* char: 0x7a 'z' */ - -static const GLubyte ch122data[] = { -0xf0,0x80,0x40,0x20,0x10,0xf0, -}; - -static const BitmapCharRec ch122 = {4,6,0,0,5,ch122data}; - -/* char: 0x79 'y' */ - -static const GLubyte ch121data[] = { -0x80,0x40,0x40,0x60,0xa0,0xa0,0x90,0x90, -}; - -static const BitmapCharRec ch121 = {4,8,0,2,5,ch121data}; - -/* char: 0x78 'x' */ - -static const GLubyte ch120data[] = { -0x88,0x88,0x50,0x20,0x50,0x88, -}; - -static const BitmapCharRec ch120 = {5,6,0,0,6,ch120data}; - -/* char: 0x77 'w' */ - -static const GLubyte ch119data[] = { -0x28,0x28,0x54,0x54,0x92,0x92, -}; - -static const BitmapCharRec ch119 = {7,6,0,0,8,ch119data}; - -/* char: 0x76 'v' */ - -static const GLubyte ch118data[] = { -0x20,0x20,0x50,0x50,0x88,0x88, -}; - -static const BitmapCharRec ch118 = {5,6,0,0,6,ch118data}; - -/* char: 0x75 'u' */ - -static const GLubyte ch117data[] = { -0x70,0x90,0x90,0x90,0x90,0x90, -}; - -static const BitmapCharRec ch117 = {4,6,0,0,5,ch117data}; - -/* char: 0x74 't' */ - -static const GLubyte ch116data[] = { -0x60,0x40,0x40,0x40,0x40,0xe0,0x40,0x40, -}; - -static const BitmapCharRec ch116 = {3,8,0,0,4,ch116data}; - -/* char: 0x73 's' */ - -static const GLubyte ch115data[] = { -0x60,0x90,0x10,0x60,0x90,0x60, -}; - -static const BitmapCharRec ch115 = {4,6,0,0,5,ch115data}; - -/* char: 0x72 'r' */ - -static const GLubyte ch114data[] = { -0x80,0x80,0x80,0x80,0xc0,0xa0, -}; - -static const BitmapCharRec ch114 = {3,6,0,0,4,ch114data}; - -/* char: 0x71 'q' */ - -static const GLubyte ch113data[] = { -0x8,0x8,0x68,0x98,0x88,0x88,0x98,0x68, -}; - -static const BitmapCharRec ch113 = {5,8,0,2,6,ch113data}; - -/* char: 0x70 'p' */ - -static const GLubyte ch112data[] = { -0x80,0x80,0xb0,0xc8,0x88,0x88,0xc8,0xb0, -}; - -static const BitmapCharRec ch112 = {5,8,0,2,6,ch112data}; - -/* char: 0x6f 'o' */ - -static const GLubyte ch111data[] = { -0x70,0x88,0x88,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch111 = {5,6,0,0,6,ch111data}; - -/* char: 0x6e 'n' */ - -static const GLubyte ch110data[] = { -0x88,0x88,0x88,0x88,0xc8,0xb0, -}; - -static const BitmapCharRec ch110 = {5,6,0,0,6,ch110data}; - -/* char: 0x6d 'm' */ - -static const GLubyte ch109data[] = { -0x92,0x92,0x92,0x92,0x92,0xec, -}; - -static const BitmapCharRec ch109 = {7,6,0,0,8,ch109data}; - -/* char: 0x6c 'l' */ - -static const GLubyte ch108data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch108 = {1,8,0,0,2,ch108data}; - -/* char: 0x6b 'k' */ - -static const GLubyte ch107data[] = { -0x90,0x90,0xa0,0xc0,0xa0,0x90,0x80,0x80, -}; - -static const BitmapCharRec ch107 = {4,8,0,0,5,ch107data}; - -/* char: 0x6a 'j' */ - -static const GLubyte ch106data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, -}; - -static const BitmapCharRec ch106 = {1,9,0,1,2,ch106data}; - -/* char: 0x69 'i' */ - -static const GLubyte ch105data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, -}; - -static const BitmapCharRec ch105 = {1,8,0,0,2,ch105data}; - -/* char: 0x68 'h' */ - -static const GLubyte ch104data[] = { -0x88,0x88,0x88,0x88,0xc8,0xb0,0x80,0x80, -}; - -static const BitmapCharRec ch104 = {5,8,0,0,6,ch104data}; - -/* char: 0x67 'g' */ - -static const GLubyte ch103data[] = { -0x70,0x8,0x68,0x98,0x88,0x88,0x98,0x68, -}; - -static const BitmapCharRec ch103 = {5,8,0,2,6,ch103data}; - -/* char: 0x66 'f' */ - -static const GLubyte ch102data[] = { -0x40,0x40,0x40,0x40,0x40,0xe0,0x40,0x30, -}; - -static const BitmapCharRec ch102 = {4,8,0,0,4,ch102data}; - -/* char: 0x65 'e' */ - -static const GLubyte ch101data[] = { -0x60,0x90,0x80,0xf0,0x90,0x60, -}; - -static const BitmapCharRec ch101 = {4,6,0,0,5,ch101data}; - -/* char: 0x64 'd' */ - -static const GLubyte ch100data[] = { -0x68,0x98,0x88,0x88,0x98,0x68,0x8,0x8, -}; - -static const BitmapCharRec ch100 = {5,8,0,0,6,ch100data}; - -/* char: 0x63 'c' */ - -static const GLubyte ch99data[] = { -0x60,0x90,0x80,0x80,0x90,0x60, -}; - -static const BitmapCharRec ch99 = {4,6,0,0,5,ch99data}; - -/* char: 0x62 'b' */ - -static const GLubyte ch98data[] = { -0xb0,0xc8,0x88,0x88,0xc8,0xb0,0x80,0x80, -}; - -static const BitmapCharRec ch98 = {5,8,0,0,6,ch98data}; - -/* char: 0x61 'a' */ - -static const GLubyte ch97data[] = { -0x68,0x90,0x90,0x70,0x10,0xe0, -}; - -static const BitmapCharRec ch97 = {5,6,0,0,5,ch97data}; - -/* char: 0x60 '`' */ - -static const GLubyte ch96data[] = { -0x80,0x80,0x40, -}; - -static const BitmapCharRec ch96 = {2,3,0,-5,3,ch96data}; - -/* char: 0x5f '_' */ - -static const GLubyte ch95data[] = { -0xfc, -}; - -static const BitmapCharRec ch95 = {6,1,0,2,6,ch95data}; - -/* char: 0x5e '^' */ - -static const GLubyte ch94data[] = { -0x88,0x50,0x50,0x20,0x20, -}; - -static const BitmapCharRec ch94 = {5,5,0,-3,6,ch94data}; - -/* char: 0x5d ']' */ - -static const GLubyte ch93data[] = { -0xc0,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0xc0, -}; - -static const BitmapCharRec ch93 = {2,10,0,2,3,ch93data}; - -/* char: 0x5c '\' */ - -static const GLubyte ch92data[] = { -0x20,0x20,0x40,0x40,0x40,0x40,0x80,0x80, -}; - -static const BitmapCharRec ch92 = {3,8,0,0,3,ch92data}; - -/* char: 0x5b '[' */ - -static const GLubyte ch91data[] = { -0xc0,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0xc0, -}; - -static const BitmapCharRec ch91 = {2,10,-1,2,3,ch91data}; - -/* char: 0x5a 'Z' */ - -static const GLubyte ch90data[] = { -0xf8,0x80,0x40,0x20,0x20,0x10,0x8,0xf8, -}; - -static const BitmapCharRec ch90 = {5,8,-1,0,7,ch90data}; - -/* char: 0x59 'Y' */ - -static const GLubyte ch89data[] = { -0x10,0x10,0x10,0x28,0x28,0x44,0x44,0x82, -}; - -static const BitmapCharRec ch89 = {7,8,0,0,7,ch89data}; - -/* char: 0x58 'X' */ - -static const GLubyte ch88data[] = { -0x88,0x88,0x50,0x50,0x20,0x50,0x88,0x88, -}; - -static const BitmapCharRec ch88 = {5,8,-1,0,7,ch88data}; - -/* char: 0x57 'W' */ - -static const GLubyte ch87data[] = { -0x22,0x0,0x22,0x0,0x22,0x0,0x55,0x0,0x49,0x0,0x49,0x0,0x88,0x80,0x88,0x80, -}; - -static const BitmapCharRec ch87 = {9,8,0,0,9,ch87data}; - -/* char: 0x56 'V' */ - -static const GLubyte ch86data[] = { -0x10,0x28,0x28,0x44,0x44,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch86 = {7,8,0,0,7,ch86data}; - -/* char: 0x55 'U' */ - -static const GLubyte ch85data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84, -}; - -static const BitmapCharRec ch85 = {6,8,-1,0,8,ch85data}; - -/* char: 0x54 'T' */ - -static const GLubyte ch84data[] = { -0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xf8, -}; - -static const BitmapCharRec ch84 = {5,8,0,0,5,ch84data}; - -/* char: 0x53 'S' */ - -static const GLubyte ch83data[] = { -0x70,0x88,0x88,0x8,0x70,0x80,0x88,0x70, -}; - -static const BitmapCharRec ch83 = {5,8,-1,0,7,ch83data}; - -/* char: 0x52 'R' */ - -static const GLubyte ch82data[] = { -0x88,0x88,0x88,0x88,0xf0,0x88,0x88,0xf0, -}; - -static const BitmapCharRec ch82 = {5,8,-1,0,7,ch82data}; - -/* char: 0x51 'Q' */ - -static const GLubyte ch81data[] = { -0x2,0x7c,0x8c,0x94,0x84,0x84,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch81 = {7,9,-1,1,8,ch81data}; - -/* char: 0x50 'P' */ - -static const GLubyte ch80data[] = { -0x80,0x80,0x80,0x80,0xf0,0x88,0x88,0xf0, -}; - -static const BitmapCharRec ch80 = {5,8,-1,0,7,ch80data}; - -/* char: 0x4f 'O' */ - -static const GLubyte ch79data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78, -}; - -static const BitmapCharRec ch79 = {6,8,-1,0,8,ch79data}; - -/* char: 0x4e 'N' */ - -static const GLubyte ch78data[] = { -0x8c,0x8c,0x94,0x94,0xa4,0xa4,0xc4,0xc4, -}; - -static const BitmapCharRec ch78 = {6,8,-1,0,8,ch78data}; - -/* char: 0x4d 'M' */ - -static const GLubyte ch77data[] = { -0x92,0x92,0x92,0xaa,0xaa,0xc6,0xc6,0x82, -}; - -static const BitmapCharRec ch77 = {7,8,-1,0,9,ch77data}; - -/* char: 0x4c 'L' */ - -static const GLubyte ch76data[] = { -0xf0,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch76 = {4,8,-1,0,6,ch76data}; - -/* char: 0x4b 'K' */ - -static const GLubyte ch75data[] = { -0x88,0x88,0x90,0x90,0xe0,0xa0,0x90,0x88, -}; - -static const BitmapCharRec ch75 = {5,8,-1,0,7,ch75data}; - -/* char: 0x4a 'J' */ - -static const GLubyte ch74data[] = { -0x60,0x90,0x10,0x10,0x10,0x10,0x10,0x10, -}; - -static const BitmapCharRec ch74 = {4,8,0,0,5,ch74data}; - -/* char: 0x49 'I' */ - -static const GLubyte ch73data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch73 = {1,8,-1,0,3,ch73data}; - -/* char: 0x48 'H' */ - -static const GLubyte ch72data[] = { -0x84,0x84,0x84,0x84,0xfc,0x84,0x84,0x84, -}; - -static const BitmapCharRec ch72 = {6,8,-1,0,8,ch72data}; - -/* char: 0x47 'G' */ - -static const GLubyte ch71data[] = { -0x74,0x8c,0x84,0x8c,0x80,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch71 = {6,8,-1,0,8,ch71data}; - -/* char: 0x46 'F' */ - -static const GLubyte ch70data[] = { -0x80,0x80,0x80,0x80,0xf0,0x80,0x80,0xf8, -}; - -static const BitmapCharRec ch70 = {5,8,-1,0,6,ch70data}; - -/* char: 0x45 'E' */ - -static const GLubyte ch69data[] = { -0xf8,0x80,0x80,0x80,0xf8,0x80,0x80,0xf8, -}; - -static const BitmapCharRec ch69 = {5,8,-1,0,7,ch69data}; - -/* char: 0x44 'D' */ - -static const GLubyte ch68data[] = { -0xf0,0x88,0x84,0x84,0x84,0x84,0x88,0xf0, -}; - -static const BitmapCharRec ch68 = {6,8,-1,0,8,ch68data}; - -/* char: 0x43 'C' */ - -static const GLubyte ch67data[] = { -0x78,0x84,0x80,0x80,0x80,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch67 = {6,8,-1,0,8,ch67data}; - -/* char: 0x42 'B' */ - -static const GLubyte ch66data[] = { -0xf0,0x88,0x88,0x88,0xf0,0x88,0x88,0xf0, -}; - -static const BitmapCharRec ch66 = {5,8,-1,0,7,ch66data}; - -/* char: 0x41 'A' */ - -static const GLubyte ch65data[] = { -0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10, -}; - -static const BitmapCharRec ch65 = {7,8,0,0,7,ch65data}; - -/* char: 0x40 '@' */ - -static const GLubyte ch64data[] = { -0x3e,0x0,0x40,0x0,0x9b,0x0,0xa4,0x80,0xa4,0x80,0xa2,0x40,0x92,0x40,0x4d,0x40, -0x20,0x80,0x1f,0x0, -}; - -static const BitmapCharRec ch64 = {10,10,0,2,11,ch64data}; - -/* char: 0x3f '?' */ - -static const GLubyte ch63data[] = { -0x40,0x0,0x40,0x40,0x20,0x10,0x90,0x60, -}; - -static const BitmapCharRec ch63 = {4,8,-1,0,6,ch63data}; - -/* char: 0x3e '>' */ - -static const GLubyte ch62data[] = { -0x80,0x40,0x20,0x40,0x80, -}; - -static const BitmapCharRec ch62 = {3,5,-1,-1,6,ch62data}; - -/* char: 0x3d '=' */ - -static const GLubyte ch61data[] = { -0xf0,0x0,0xf0, -}; - -static const BitmapCharRec ch61 = {4,3,0,-2,5,ch61data}; - -/* char: 0x3c '<' */ - -static const GLubyte ch60data[] = { -0x20,0x40,0x80,0x40,0x20, -}; - -static const BitmapCharRec ch60 = {3,5,-1,-1,6,ch60data}; - -/* char: 0x3b ';' */ - -static const GLubyte ch59data[] = { -0x80,0x40,0x40,0x0,0x0,0x0,0x0,0x40, -}; - -static const BitmapCharRec ch59 = {2,8,0,2,3,ch59data}; - -/* char: 0x3a ':' */ - -static const GLubyte ch58data[] = { -0x80,0x0,0x0,0x0,0x0,0x80, -}; - -static const BitmapCharRec ch58 = {1,6,-1,0,3,ch58data}; - -/* char: 0x39 '9' */ - -static const GLubyte ch57data[] = { -0x70,0x88,0x8,0x68,0x98,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch57 = {5,8,0,0,6,ch57data}; - -/* char: 0x38 '8' */ - -static const GLubyte ch56data[] = { -0x70,0x88,0x88,0x88,0x70,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch56 = {5,8,0,0,6,ch56data}; - -/* char: 0x37 '7' */ - -static const GLubyte ch55data[] = { -0x40,0x40,0x20,0x20,0x10,0x10,0x8,0xf8, -}; - -static const BitmapCharRec ch55 = {5,8,0,0,6,ch55data}; - -/* char: 0x36 '6' */ - -static const GLubyte ch54data[] = { -0x70,0x88,0x88,0xc8,0xb0,0x80,0x88,0x70, -}; - -static const BitmapCharRec ch54 = {5,8,0,0,6,ch54data}; - -/* char: 0x35 '5' */ - -static const GLubyte ch53data[] = { -0x70,0x88,0x8,0x8,0xf0,0x80,0x80,0xf8, -}; - -static const BitmapCharRec ch53 = {5,8,0,0,6,ch53data}; - -/* char: 0x34 '4' */ - -static const GLubyte ch52data[] = { -0x10,0x10,0xf8,0x90,0x50,0x50,0x30,0x10, -}; - -static const BitmapCharRec ch52 = {5,8,0,0,6,ch52data}; - -/* char: 0x33 '3' */ - -static const GLubyte ch51data[] = { -0x70,0x88,0x8,0x8,0x30,0x8,0x88,0x70, -}; - -static const BitmapCharRec ch51 = {5,8,0,0,6,ch51data}; - -/* char: 0x32 '2' */ - -static const GLubyte ch50data[] = { -0xf8,0x80,0x40,0x30,0x8,0x8,0x88,0x70, -}; - -static const BitmapCharRec ch50 = {5,8,0,0,6,ch50data}; - -/* char: 0x31 '1' */ - -static const GLubyte ch49data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0xc0,0x40, -}; - -static const BitmapCharRec ch49 = {2,8,-1,0,6,ch49data}; - -/* char: 0x30 '0' */ - -static const GLubyte ch48data[] = { -0x70,0x88,0x88,0x88,0x88,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch48 = {5,8,0,0,6,ch48data}; - -/* char: 0x2f '/' */ - -static const GLubyte ch47data[] = { -0x80,0x80,0x40,0x40,0x40,0x40,0x20,0x20, -}; - -static const BitmapCharRec ch47 = {3,8,0,0,3,ch47data}; - -/* char: 0x2e '.' */ - -static const GLubyte ch46data[] = { -0x80, -}; - -static const BitmapCharRec ch46 = {1,1,-1,0,3,ch46data}; - -/* char: 0x2d '-' */ - -static const GLubyte ch45data[] = { -0xf8, -}; - -static const BitmapCharRec ch45 = {5,1,-1,-3,7,ch45data}; - -/* char: 0x2c ',' */ - -static const GLubyte ch44data[] = { -0x80,0x40,0x40, -}; - -static const BitmapCharRec ch44 = {2,3,0,2,3,ch44data}; - -/* char: 0x2b '+' */ - -static const GLubyte ch43data[] = { -0x20,0x20,0xf8,0x20,0x20, -}; - -static const BitmapCharRec ch43 = {5,5,0,-1,6,ch43data}; - -/* char: 0x2a '*' */ - -static const GLubyte ch42data[] = { -0xa0,0x40,0xa0, -}; - -static const BitmapCharRec ch42 = {3,3,0,-5,4,ch42data}; - -/* char: 0x29 ')' */ - -static const GLubyte ch41data[] = { -0x80,0x40,0x40,0x20,0x20,0x20,0x20,0x40,0x40,0x80, -}; - -static const BitmapCharRec ch41 = {3,10,-1,2,4,ch41data}; - -/* char: 0x28 '(' */ - -static const GLubyte ch40data[] = { -0x20,0x40,0x40,0x80,0x80,0x80,0x80,0x40,0x40,0x20, -}; - -static const BitmapCharRec ch40 = {3,10,0,2,4,ch40data}; - -/* char: 0x27 ''' */ - -static const GLubyte ch39data[] = { -0x80,0x40,0x40, -}; - -static const BitmapCharRec ch39 = {2,3,-1,-5,3,ch39data}; - -/* char: 0x26 '&' */ - -static const GLubyte ch38data[] = { -0x64,0x98,0x98,0xa4,0x60,0x50,0x50,0x20, -}; - -static const BitmapCharRec ch38 = {6,8,-1,0,8,ch38data}; - -/* char: 0x25 '%' */ - -static const GLubyte ch37data[] = { -0x26,0x29,0x16,0x10,0x8,0x68,0x94,0x64, -}; - -static const BitmapCharRec ch37 = {8,8,0,0,9,ch37data}; - -/* char: 0x24 '$' */ - -static const GLubyte ch36data[] = { -0x20,0x70,0xa8,0x28,0x70,0xa0,0xa8,0x70,0x20, -}; - -static const BitmapCharRec ch36 = {5,9,0,1,6,ch36data}; - -/* char: 0x23 '#' */ - -static const GLubyte ch35data[] = { -0x50,0x50,0xf8,0x28,0x7c,0x28,0x28, -}; - -static const BitmapCharRec ch35 = {6,7,0,0,6,ch35data}; - -/* char: 0x22 '"' */ - -static const GLubyte ch34data[] = { -0xa0,0xa0, -}; - -static const BitmapCharRec ch34 = {3,2,-1,-6,4,ch34data}; - -/* char: 0x21 '!' */ - -static const GLubyte ch33data[] = { -0x80,0x0,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch33 = {1,8,-1,0,3,ch33data}; - -/* char: 0x20 ' ' */ - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch32data[] = { 0x0 }; -static const BitmapCharRec ch32 = {0,0,0,0,3,ch32data}; -#else -static const BitmapCharRec ch32 = {0,0,0,0,3,0}; -#endif - -static const BitmapCharRec * const chars[] = { -&ch32, -&ch33, -&ch34, -&ch35, -&ch36, -&ch37, -&ch38, -&ch39, -&ch40, -&ch41, -&ch42, -&ch43, -&ch44, -&ch45, -&ch46, -&ch47, -&ch48, -&ch49, -&ch50, -&ch51, -&ch52, -&ch53, -&ch54, -&ch55, -&ch56, -&ch57, -&ch58, -&ch59, -&ch60, -&ch61, -&ch62, -&ch63, -&ch64, -&ch65, -&ch66, -&ch67, -&ch68, -&ch69, -&ch70, -&ch71, -&ch72, -&ch73, -&ch74, -&ch75, -&ch76, -&ch77, -&ch78, -&ch79, -&ch80, -&ch81, -&ch82, -&ch83, -&ch84, -&ch85, -&ch86, -&ch87, -&ch88, -&ch89, -&ch90, -&ch91, -&ch92, -&ch93, -&ch94, -&ch95, -&ch96, -&ch97, -&ch98, -&ch99, -&ch100, -&ch101, -&ch102, -&ch103, -&ch104, -&ch105, -&ch106, -&ch107, -&ch108, -&ch109, -&ch110, -&ch111, -&ch112, -&ch113, -&ch114, -&ch115, -&ch116, -&ch117, -&ch118, -&ch119, -&ch120, -&ch121, -&ch122, -&ch123, -&ch124, -&ch125, -&ch126, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -&ch160, -&ch161, -&ch162, -&ch163, -&ch164, -&ch165, -&ch166, -&ch167, -&ch168, -&ch169, -&ch170, -&ch171, -&ch172, -&ch173, -&ch174, -&ch175, -&ch176, -&ch177, -&ch178, -&ch179, -&ch180, -&ch181, -&ch182, -&ch183, -&ch184, -&ch185, -&ch186, -&ch187, -&ch188, -&ch189, -&ch190, -&ch191, -&ch192, -&ch193, -&ch194, -&ch195, -&ch196, -&ch197, -&ch198, -&ch199, -&ch200, -&ch201, -&ch202, -&ch203, -&ch204, -&ch205, -&ch206, -&ch207, -&ch208, -&ch209, -&ch210, -&ch211, -&ch212, -&ch213, -&ch214, -&ch215, -&ch216, -&ch217, -&ch218, -&ch219, -&ch220, -&ch221, -&ch222, -&ch223, -&ch224, -&ch225, -&ch226, -&ch227, -&ch228, -&ch229, -&ch230, -&ch231, -&ch232, -&ch233, -&ch234, -&ch235, -&ch236, -&ch237, -&ch238, -&ch239, -&ch240, -&ch241, -&ch242, -&ch243, -&ch244, -&ch245, -&ch246, -&ch247, -&ch248, -&ch249, -&ch250, -&ch251, -&ch252, -&ch253, -&ch254, -&ch255, -}; - -#if !defined(__IBMCPP__) -const -#endif -BitmapFontRec glutBitmapHelvetica10 = { -"-adobe-helvetica-medium-r-normal--10-100-75-75-p-56-iso8859-1", -224, -32, -chars -}; - + +/* GENERATED FILE -- DO NOT MODIFY */ + +#define glutBitmapHelvetica10 XXX +#include "glutbitmap.h" +#undef glutBitmapHelvetica10 + +/* char: 0xff */ + +static const GLubyte ch255data[] = { +0x80,0x40,0x40,0x60,0xa0,0xa0,0x90,0x90,0x0,0x50, +}; + +static const BitmapCharRec ch255 = {4,10,0,2,5,ch255data}; + +/* char: 0xfe */ + +static const GLubyte ch254data[] = { +0x80,0x80,0xb0,0xc8,0x88,0x88,0xc8,0xb0,0x80,0x80, +}; + +static const BitmapCharRec ch254 = {5,10,0,2,6,ch254data}; + +/* char: 0xfd */ + +static const GLubyte ch253data[] = { +0x80,0x40,0x40,0x60,0xa0,0xa0,0x90,0x90,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch253 = {4,11,0,2,5,ch253data}; + +/* char: 0xfc */ + +static const GLubyte ch252data[] = { +0x70,0x90,0x90,0x90,0x90,0x90,0x0,0x50, +}; + +static const BitmapCharRec ch252 = {4,8,0,0,5,ch252data}; + +/* char: 0xfb */ + +static const GLubyte ch251data[] = { +0x70,0x90,0x90,0x90,0x90,0x90,0x0,0x50,0x20, +}; + +static const BitmapCharRec ch251 = {4,9,0,0,5,ch251data}; + +/* char: 0xfa */ + +static const GLubyte ch250data[] = { +0x70,0x90,0x90,0x90,0x90,0x90,0x0,0x40,0x20, +}; + +static const BitmapCharRec ch250 = {4,9,0,0,5,ch250data}; + +/* char: 0xf9 */ + +static const GLubyte ch249data[] = { +0x70,0x90,0x90,0x90,0x90,0x90,0x0,0x20,0x40, +}; + +static const BitmapCharRec ch249 = {4,9,0,0,5,ch249data}; + +/* char: 0xf8 */ + +static const GLubyte ch248data[] = { +0x70,0x88,0xc8,0xa8,0x98,0x74, +}; + +static const BitmapCharRec ch248 = {6,6,0,0,6,ch248data}; + +/* char: 0xf7 */ + +static const GLubyte ch247data[] = { +0x20,0x0,0xf8,0x0,0x20, +}; + +static const BitmapCharRec ch247 = {5,5,0,-1,6,ch247data}; + +/* char: 0xf6 */ + +static const GLubyte ch246data[] = { +0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x50, +}; + +static const BitmapCharRec ch246 = {5,8,0,0,6,ch246data}; + +/* char: 0xf5 */ + +static const GLubyte ch245data[] = { +0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch245 = {5,9,0,0,6,ch245data}; + +/* char: 0xf4 */ + +static const GLubyte ch244data[] = { +0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x50,0x20, +}; + +static const BitmapCharRec ch244 = {5,9,0,0,6,ch244data}; + +/* char: 0xf3 */ + +static const GLubyte ch243data[] = { +0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch243 = {5,9,0,0,6,ch243data}; + +/* char: 0xf2 */ + +static const GLubyte ch242data[] = { +0x70,0x88,0x88,0x88,0x88,0x70,0x0,0x20,0x40, +}; + +static const BitmapCharRec ch242 = {5,9,0,0,6,ch242data}; + +/* char: 0xf1 */ + +static const GLubyte ch241data[] = { +0x90,0x90,0x90,0x90,0x90,0xe0,0x0,0xa0,0x50, +}; + +static const BitmapCharRec ch241 = {4,9,0,0,5,ch241data}; + +/* char: 0xf0 */ + +static const GLubyte ch240data[] = { +0x70,0x88,0x88,0x88,0x88,0x78,0x90,0x60,0x50, +}; + +static const BitmapCharRec ch240 = {5,9,0,0,6,ch240data}; + +/* char: 0xef */ + +static const GLubyte ch239data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0, +}; + +static const BitmapCharRec ch239 = {3,8,0,0,2,ch239data}; + +/* char: 0xee */ + +static const GLubyte ch238data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0,0x40, +}; + +static const BitmapCharRec ch238 = {3,9,1,0,2,ch238data}; + +/* char: 0xed */ + +static const GLubyte ch237data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x40, +}; + +static const BitmapCharRec ch237 = {2,9,0,0,2,ch237data}; + +/* char: 0xec */ + +static const GLubyte ch236data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40,0x80, +}; + +static const BitmapCharRec ch236 = {2,9,1,0,2,ch236data}; + +/* char: 0xeb */ + +static const GLubyte ch235data[] = { +0x60,0x90,0x80,0xf0,0x90,0x60,0x0,0x50, +}; + +static const BitmapCharRec ch235 = {4,8,0,0,5,ch235data}; + +/* char: 0xea */ + +static const GLubyte ch234data[] = { +0x60,0x90,0x80,0xf0,0x90,0x60,0x0,0x50,0x20, +}; + +static const BitmapCharRec ch234 = {4,9,0,0,5,ch234data}; + +/* char: 0xe9 */ + +static const GLubyte ch233data[] = { +0x60,0x90,0x80,0xf0,0x90,0x60,0x0,0x40,0x20, +}; + +static const BitmapCharRec ch233 = {4,9,0,0,5,ch233data}; + +/* char: 0xe8 */ + +static const GLubyte ch232data[] = { +0x60,0x90,0x80,0xf0,0x90,0x60,0x0,0x20,0x40, +}; + +static const BitmapCharRec ch232 = {4,9,0,0,5,ch232data}; + +/* char: 0xe7 */ + +static const GLubyte ch231data[] = { +0x60,0x20,0x60,0x90,0x80,0x80,0x90,0x60, +}; + +static const BitmapCharRec ch231 = {4,8,0,2,5,ch231data}; + +/* char: 0xe6 */ + +static const GLubyte ch230data[] = { +0x6c,0x92,0x90,0x7e,0x12,0xec, +}; + +static const BitmapCharRec ch230 = {7,6,0,0,8,ch230data}; + +/* char: 0xe5 */ + +static const GLubyte ch229data[] = { +0x68,0x90,0x90,0x70,0x10,0xe0,0x20,0x50,0x20, +}; + +static const BitmapCharRec ch229 = {5,9,0,0,5,ch229data}; + +/* char: 0xe4 */ + +static const GLubyte ch228data[] = { +0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0x50, +}; + +static const BitmapCharRec ch228 = {5,8,0,0,5,ch228data}; + +/* char: 0xe3 */ + +static const GLubyte ch227data[] = { +0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0xa0,0x50, +}; + +static const BitmapCharRec ch227 = {5,9,0,0,5,ch227data}; + +/* char: 0xe2 */ + +static const GLubyte ch226data[] = { +0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0x50,0x20, +}; + +static const BitmapCharRec ch226 = {5,9,0,0,5,ch226data}; + +/* char: 0xe1 */ + +static const GLubyte ch225data[] = { +0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch225 = {5,9,0,0,5,ch225data}; + +/* char: 0xe0 */ + +static const GLubyte ch224data[] = { +0x68,0x90,0x90,0x70,0x10,0xe0,0x0,0x20,0x40, +}; + +static const BitmapCharRec ch224 = {5,9,0,0,5,ch224data}; + +/* char: 0xdf */ + +static const GLubyte ch223data[] = { +0xa0,0x90,0x90,0x90,0xa0,0x90,0x90,0x60, +}; + +static const BitmapCharRec ch223 = {4,8,0,0,5,ch223data}; + +/* char: 0xde */ + +static const GLubyte ch222data[] = { +0x80,0x80,0xf0,0x88,0x88,0xf0,0x80,0x80, +}; + +static const BitmapCharRec ch222 = {5,8,-1,0,7,ch222data}; + +/* char: 0xdd */ + +static const GLubyte ch221data[] = { +0x10,0x10,0x10,0x28,0x28,0x44,0x44,0x82,0x0,0x10,0x8, +}; + +static const BitmapCharRec ch221 = {7,11,0,0,7,ch221data}; + +/* char: 0xdc */ + +static const GLubyte ch220data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x48, +}; + +static const BitmapCharRec ch220 = {6,10,-1,0,8,ch220data}; + +/* char: 0xdb */ + +static const GLubyte ch219data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x28,0x10, +}; + +static const BitmapCharRec ch219 = {6,11,-1,0,8,ch219data}; + +/* char: 0xda */ + +static const GLubyte ch218data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch218 = {6,11,-1,0,8,ch218data}; + +/* char: 0xd9 */ + +static const GLubyte ch217data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch217 = {6,11,-1,0,8,ch217data}; + +/* char: 0xd8 */ + +static const GLubyte ch216data[] = { +0x80,0x78,0xc4,0xa4,0xa4,0x94,0x94,0x8c,0x78,0x4, +}; + +static const BitmapCharRec ch216 = {6,10,-1,1,8,ch216data}; + +/* char: 0xd7 */ + +static const GLubyte ch215data[] = { +0x88,0x50,0x20,0x50,0x88, +}; + +static const BitmapCharRec ch215 = {5,5,0,-1,6,ch215data}; + +/* char: 0xd6 */ + +static const GLubyte ch214data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x48, +}; + +static const BitmapCharRec ch214 = {6,10,-1,0,8,ch214data}; + +/* char: 0xd5 */ + +static const GLubyte ch213data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch213 = {6,11,-1,0,8,ch213data}; + +/* char: 0xd4 */ + +static const GLubyte ch212data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x28,0x10, +}; + +static const BitmapCharRec ch212 = {6,11,-1,0,8,ch212data}; + +/* char: 0xd3 */ + +static const GLubyte ch211data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x10,0x8, +}; + +static const BitmapCharRec ch211 = {6,11,-1,0,8,ch211data}; + +/* char: 0xd2 */ + +static const GLubyte ch210data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch210 = {6,11,-1,0,8,ch210data}; + +/* char: 0xd1 */ + +static const GLubyte ch209data[] = { +0x8c,0x8c,0x94,0x94,0xa4,0xa4,0xc4,0xc4,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch209 = {6,11,-1,0,8,ch209data}; + +/* char: 0xd0 */ + +static const GLubyte ch208data[] = { +0x78,0x44,0x42,0x42,0xf2,0x42,0x44,0x78, +}; + +static const BitmapCharRec ch208 = {7,8,0,0,8,ch208data}; + +/* char: 0xcf */ + +static const GLubyte ch207data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0, +}; + +static const BitmapCharRec ch207 = {3,10,0,0,3,ch207data}; + +/* char: 0xce */ + +static const GLubyte ch206data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0,0x40, +}; + +static const BitmapCharRec ch206 = {3,11,0,0,3,ch206data}; + +/* char: 0xcd */ + +static const GLubyte ch205data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x40, +}; + +static const BitmapCharRec ch205 = {2,11,-1,0,3,ch205data}; + +/* char: 0xcc */ + +static const GLubyte ch204data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40,0x80, +}; + +static const BitmapCharRec ch204 = {2,11,0,0,3,ch204data}; + +/* char: 0xcb */ + +static const GLubyte ch203data[] = { +0xf8,0x80,0x80,0x80,0xf8,0x80,0x80,0xf8,0x0,0x50, +}; + +static const BitmapCharRec ch203 = {5,10,-1,0,7,ch203data}; + +/* char: 0xca */ + +static const GLubyte ch202data[] = { +0xf8,0x80,0x80,0xf8,0x80,0x80,0x80,0xf8,0x0,0x50,0x20, +}; + +static const BitmapCharRec ch202 = {5,11,-1,0,7,ch202data}; + +/* char: 0xc9 */ + +static const GLubyte ch201data[] = { +0xf8,0x80,0x80,0x80,0xf8,0x80,0x80,0xf8,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch201 = {5,11,-1,0,7,ch201data}; + +/* char: 0xc8 */ + +static const GLubyte ch200data[] = { +0xf8,0x80,0x80,0x80,0xf8,0x80,0x80,0xf8,0x0,0x20,0x40, +}; + +static const BitmapCharRec ch200 = {5,11,-1,0,7,ch200data}; + +/* char: 0xc7 */ + +static const GLubyte ch199data[] = { +0x30,0x10,0x78,0x84,0x80,0x80,0x80,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch199 = {6,10,-1,2,8,ch199data}; + +/* char: 0xc6 */ + +static const GLubyte ch198data[] = { +0x8f,0x80,0x88,0x0,0x78,0x0,0x48,0x0,0x2f,0x80,0x28,0x0,0x18,0x0,0x1f,0x80, +}; + +static const BitmapCharRec ch198 = {9,8,0,0,10,ch198data}; + +/* char: 0xc5 */ + +static const GLubyte ch197data[] = { +0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x10,0x28,0x10, +}; + +static const BitmapCharRec ch197 = {7,11,0,0,7,ch197data}; + +/* char: 0xc4 */ + +static const GLubyte ch196data[] = { +0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x28, +}; + +static const BitmapCharRec ch196 = {7,10,0,0,7,ch196data}; + +/* char: 0xc3 */ + +static const GLubyte ch195data[] = { +0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x28,0x14, +}; + +static const BitmapCharRec ch195 = {7,11,0,0,7,ch195data}; + +/* char: 0xc2 */ + +static const GLubyte ch194data[] = { +0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x28,0x10, +}; + +static const BitmapCharRec ch194 = {7,11,0,0,7,ch194data}; + +/* char: 0xc1 */ + +static const GLubyte ch193data[] = { +0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x10,0x8, +}; + +static const BitmapCharRec ch193 = {7,11,0,0,7,ch193data}; + +/* char: 0xc0 */ + +static const GLubyte ch192data[] = { +0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch192 = {7,11,0,0,7,ch192data}; + +/* char: 0xbf */ + +static const GLubyte ch191data[] = { +0x60,0x90,0x80,0x40,0x20,0x20,0x0,0x20, +}; + +static const BitmapCharRec ch191 = {4,8,-1,2,6,ch191data}; + +/* char: 0xbe */ + +static const GLubyte ch190data[] = { +0x21,0x0,0x17,0x80,0x13,0x0,0x9,0x0,0xc8,0x0,0x24,0x0,0x44,0x0,0xe2,0x0, +}; + +static const BitmapCharRec ch190 = {9,8,0,0,9,ch190data}; + +/* char: 0xbd */ + +static const GLubyte ch189data[] = { +0x27,0x12,0x15,0xb,0x48,0x44,0xc4,0x42, +}; + +static const BitmapCharRec ch189 = {8,8,0,0,9,ch189data}; + +/* char: 0xbc */ + +static const GLubyte ch188data[] = { +0x21,0x0,0x17,0x80,0x13,0x0,0x9,0x0,0x48,0x0,0x44,0x0,0xc4,0x0,0x42,0x0, +}; + +static const BitmapCharRec ch188 = {9,8,0,0,9,ch188data}; + +/* char: 0xbb */ + +static const GLubyte ch187data[] = { +0xa0,0x50,0x28,0x50,0xa0, +}; + +static const BitmapCharRec ch187 = {5,5,0,0,6,ch187data}; + +/* char: 0xba */ + +static const GLubyte ch186data[] = { +0xe0,0x0,0xe0,0xa0,0xe0, +}; + +static const BitmapCharRec ch186 = {3,5,0,-3,4,ch186data}; + +/* char: 0xb9 */ + +static const GLubyte ch185data[] = { +0x40,0x40,0xc0,0x40, +}; + +static const BitmapCharRec ch185 = {2,4,0,-3,3,ch185data}; + +/* char: 0xb8 */ + +static const GLubyte ch184data[] = { +0xc0,0x40, +}; + +static const BitmapCharRec ch184 = {2,2,0,2,3,ch184data}; + +/* char: 0xb7 */ + +static const GLubyte ch183data[] = { +0xc0, +}; + +static const BitmapCharRec ch183 = {2,1,0,-3,3,ch183data}; + +/* char: 0xb6 */ + +static const GLubyte ch182data[] = { +0x28,0x28,0x28,0x28,0x28,0x68,0xe8,0xe8,0xe8,0x7c, +}; + +static const BitmapCharRec ch182 = {6,10,0,2,6,ch182data}; + +/* char: 0xb5 */ + +static const GLubyte ch181data[] = { +0x80,0x80,0xf0,0x90,0x90,0x90,0x90,0x90, +}; + +static const BitmapCharRec ch181 = {4,8,0,2,5,ch181data}; + +/* char: 0xb4 */ + +static const GLubyte ch180data[] = { +0x80,0x40, +}; + +static const BitmapCharRec ch180 = {2,2,0,-6,3,ch180data}; + +/* char: 0xb3 */ + +static const GLubyte ch179data[] = { +0xc0,0x20,0x40,0xe0, +}; + +static const BitmapCharRec ch179 = {3,4,0,-3,3,ch179data}; + +/* char: 0xb2 */ + +static const GLubyte ch178data[] = { +0xe0,0x40,0xa0,0x60, +}; + +static const BitmapCharRec ch178 = {3,4,0,-3,3,ch178data}; + +/* char: 0xb1 */ + +static const GLubyte ch177data[] = { +0xf8,0x0,0x20,0x20,0xf8,0x20,0x20, +}; + +static const BitmapCharRec ch177 = {5,7,0,0,6,ch177data}; + +/* char: 0xb0 */ + +static const GLubyte ch176data[] = { +0x60,0x90,0x90,0x60, +}; + +static const BitmapCharRec ch176 = {4,4,0,-3,4,ch176data}; + +/* char: 0xaf */ + +static const GLubyte ch175data[] = { +0xe0, +}; + +static const BitmapCharRec ch175 = {3,1,0,-7,3,ch175data}; + +/* char: 0xae */ + +static const GLubyte ch174data[] = { +0x38,0x44,0xaa,0xb2,0xba,0x44,0x38, +}; + +static const BitmapCharRec ch174 = {7,7,-1,0,9,ch174data}; + +/* char: 0xad */ + +static const GLubyte ch173data[] = { +0xe0, +}; + +static const BitmapCharRec ch173 = {3,1,0,-3,4,ch173data}; + +/* char: 0xac */ + +static const GLubyte ch172data[] = { +0x8,0x8,0xf8, +}; + +static const BitmapCharRec ch172 = {5,3,-1,-2,7,ch172data}; + +/* char: 0xab */ + +static const GLubyte ch171data[] = { +0x28,0x50,0xa0,0x50,0x28, +}; + +static const BitmapCharRec ch171 = {5,5,0,0,6,ch171data}; + +/* char: 0xaa */ + +static const GLubyte ch170data[] = { +0xe0,0x0,0xa0,0x20,0xe0, +}; + +static const BitmapCharRec ch170 = {3,5,0,-3,4,ch170data}; + +/* char: 0xa9 */ + +static const GLubyte ch169data[] = { +0x38,0x44,0x9a,0xa2,0x9a,0x44,0x38, +}; + +static const BitmapCharRec ch169 = {7,7,-1,0,9,ch169data}; + +/* char: 0xa8 */ + +static const GLubyte ch168data[] = { +0xa0, +}; + +static const BitmapCharRec ch168 = {3,1,0,-7,3,ch168data}; + +/* char: 0xa7 */ + +static const GLubyte ch167data[] = { +0x70,0x88,0x18,0x70,0xc8,0x98,0x70,0xc0,0x88,0x70, +}; + +static const BitmapCharRec ch167 = {5,10,0,2,6,ch167data}; + +/* char: 0xa6 */ + +static const GLubyte ch166data[] = { +0x80,0x80,0x80,0x80,0x0,0x0,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch166 = {1,10,-1,2,3,ch166data}; + +/* char: 0xa5 */ + +static const GLubyte ch165data[] = { +0x20,0xf8,0x20,0xf8,0x50,0x50,0x88,0x88, +}; + +static const BitmapCharRec ch165 = {5,8,0,0,6,ch165data}; + +/* char: 0xa4 */ + +static const GLubyte ch164data[] = { +0x90,0x60,0x90,0x90,0x60,0x90, +}; + +static const BitmapCharRec ch164 = {4,6,0,-1,5,ch164data}; + +/* char: 0xa3 */ + +static const GLubyte ch163data[] = { +0xb0,0x48,0x40,0x40,0xe0,0x40,0x48,0x30, +}; + +static const BitmapCharRec ch163 = {5,8,0,0,6,ch163data}; + +/* char: 0xa2 */ + +static const GLubyte ch162data[] = { +0x40,0x70,0xa8,0xa0,0xa0,0xa8,0x70,0x10, +}; + +static const BitmapCharRec ch162 = {5,8,0,1,6,ch162data}; + +/* char: 0xa1 */ + +static const GLubyte ch161data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, +}; + +static const BitmapCharRec ch161 = {1,8,-1,2,3,ch161data}; + +/* char: 0xa0 */ + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch160data[] = { 0x0 }; +static const BitmapCharRec ch160 = {1,1,0,0,3,ch160data}; +#else +static const BitmapCharRec ch160 = {0,0,0,0,3,0}; +#endif + +/* char: 0x7e '~' */ + +static const GLubyte ch126data[] = { +0x98,0x64, +}; + +static const BitmapCharRec ch126 = {6,2,0,-3,7,ch126data}; + +/* char: 0x7d '}' */ + +static const GLubyte ch125data[] = { +0x80,0x40,0x40,0x40,0x40,0x20,0x40,0x40,0x40,0x80, +}; + +static const BitmapCharRec ch125 = {3,10,0,2,3,ch125data}; + +/* char: 0x7c '|' */ + +static const GLubyte ch124data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch124 = {1,10,-1,2,3,ch124data}; + +/* char: 0x7b '{' */ + +static const GLubyte ch123data[] = { +0x20,0x40,0x40,0x40,0x40,0x80,0x40,0x40,0x40,0x20, +}; + +static const BitmapCharRec ch123 = {3,10,0,2,3,ch123data}; + +/* char: 0x7a 'z' */ + +static const GLubyte ch122data[] = { +0xf0,0x80,0x40,0x20,0x10,0xf0, +}; + +static const BitmapCharRec ch122 = {4,6,0,0,5,ch122data}; + +/* char: 0x79 'y' */ + +static const GLubyte ch121data[] = { +0x80,0x40,0x40,0x60,0xa0,0xa0,0x90,0x90, +}; + +static const BitmapCharRec ch121 = {4,8,0,2,5,ch121data}; + +/* char: 0x78 'x' */ + +static const GLubyte ch120data[] = { +0x88,0x88,0x50,0x20,0x50,0x88, +}; + +static const BitmapCharRec ch120 = {5,6,0,0,6,ch120data}; + +/* char: 0x77 'w' */ + +static const GLubyte ch119data[] = { +0x28,0x28,0x54,0x54,0x92,0x92, +}; + +static const BitmapCharRec ch119 = {7,6,0,0,8,ch119data}; + +/* char: 0x76 'v' */ + +static const GLubyte ch118data[] = { +0x20,0x20,0x50,0x50,0x88,0x88, +}; + +static const BitmapCharRec ch118 = {5,6,0,0,6,ch118data}; + +/* char: 0x75 'u' */ + +static const GLubyte ch117data[] = { +0x70,0x90,0x90,0x90,0x90,0x90, +}; + +static const BitmapCharRec ch117 = {4,6,0,0,5,ch117data}; + +/* char: 0x74 't' */ + +static const GLubyte ch116data[] = { +0x60,0x40,0x40,0x40,0x40,0xe0,0x40,0x40, +}; + +static const BitmapCharRec ch116 = {3,8,0,0,4,ch116data}; + +/* char: 0x73 's' */ + +static const GLubyte ch115data[] = { +0x60,0x90,0x10,0x60,0x90,0x60, +}; + +static const BitmapCharRec ch115 = {4,6,0,0,5,ch115data}; + +/* char: 0x72 'r' */ + +static const GLubyte ch114data[] = { +0x80,0x80,0x80,0x80,0xc0,0xa0, +}; + +static const BitmapCharRec ch114 = {3,6,0,0,4,ch114data}; + +/* char: 0x71 'q' */ + +static const GLubyte ch113data[] = { +0x8,0x8,0x68,0x98,0x88,0x88,0x98,0x68, +}; + +static const BitmapCharRec ch113 = {5,8,0,2,6,ch113data}; + +/* char: 0x70 'p' */ + +static const GLubyte ch112data[] = { +0x80,0x80,0xb0,0xc8,0x88,0x88,0xc8,0xb0, +}; + +static const BitmapCharRec ch112 = {5,8,0,2,6,ch112data}; + +/* char: 0x6f 'o' */ + +static const GLubyte ch111data[] = { +0x70,0x88,0x88,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch111 = {5,6,0,0,6,ch111data}; + +/* char: 0x6e 'n' */ + +static const GLubyte ch110data[] = { +0x88,0x88,0x88,0x88,0xc8,0xb0, +}; + +static const BitmapCharRec ch110 = {5,6,0,0,6,ch110data}; + +/* char: 0x6d 'm' */ + +static const GLubyte ch109data[] = { +0x92,0x92,0x92,0x92,0x92,0xec, +}; + +static const BitmapCharRec ch109 = {7,6,0,0,8,ch109data}; + +/* char: 0x6c 'l' */ + +static const GLubyte ch108data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch108 = {1,8,0,0,2,ch108data}; + +/* char: 0x6b 'k' */ + +static const GLubyte ch107data[] = { +0x90,0x90,0xa0,0xc0,0xa0,0x90,0x80,0x80, +}; + +static const BitmapCharRec ch107 = {4,8,0,0,5,ch107data}; + +/* char: 0x6a 'j' */ + +static const GLubyte ch106data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, +}; + +static const BitmapCharRec ch106 = {1,9,0,1,2,ch106data}; + +/* char: 0x69 'i' */ + +static const GLubyte ch105data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, +}; + +static const BitmapCharRec ch105 = {1,8,0,0,2,ch105data}; + +/* char: 0x68 'h' */ + +static const GLubyte ch104data[] = { +0x88,0x88,0x88,0x88,0xc8,0xb0,0x80,0x80, +}; + +static const BitmapCharRec ch104 = {5,8,0,0,6,ch104data}; + +/* char: 0x67 'g' */ + +static const GLubyte ch103data[] = { +0x70,0x8,0x68,0x98,0x88,0x88,0x98,0x68, +}; + +static const BitmapCharRec ch103 = {5,8,0,2,6,ch103data}; + +/* char: 0x66 'f' */ + +static const GLubyte ch102data[] = { +0x40,0x40,0x40,0x40,0x40,0xe0,0x40,0x30, +}; + +static const BitmapCharRec ch102 = {4,8,0,0,4,ch102data}; + +/* char: 0x65 'e' */ + +static const GLubyte ch101data[] = { +0x60,0x90,0x80,0xf0,0x90,0x60, +}; + +static const BitmapCharRec ch101 = {4,6,0,0,5,ch101data}; + +/* char: 0x64 'd' */ + +static const GLubyte ch100data[] = { +0x68,0x98,0x88,0x88,0x98,0x68,0x8,0x8, +}; + +static const BitmapCharRec ch100 = {5,8,0,0,6,ch100data}; + +/* char: 0x63 'c' */ + +static const GLubyte ch99data[] = { +0x60,0x90,0x80,0x80,0x90,0x60, +}; + +static const BitmapCharRec ch99 = {4,6,0,0,5,ch99data}; + +/* char: 0x62 'b' */ + +static const GLubyte ch98data[] = { +0xb0,0xc8,0x88,0x88,0xc8,0xb0,0x80,0x80, +}; + +static const BitmapCharRec ch98 = {5,8,0,0,6,ch98data}; + +/* char: 0x61 'a' */ + +static const GLubyte ch97data[] = { +0x68,0x90,0x90,0x70,0x10,0xe0, +}; + +static const BitmapCharRec ch97 = {5,6,0,0,5,ch97data}; + +/* char: 0x60 '`' */ + +static const GLubyte ch96data[] = { +0x80,0x80,0x40, +}; + +static const BitmapCharRec ch96 = {2,3,0,-5,3,ch96data}; + +/* char: 0x5f '_' */ + +static const GLubyte ch95data[] = { +0xfc, +}; + +static const BitmapCharRec ch95 = {6,1,0,2,6,ch95data}; + +/* char: 0x5e '^' */ + +static const GLubyte ch94data[] = { +0x88,0x50,0x50,0x20,0x20, +}; + +static const BitmapCharRec ch94 = {5,5,0,-3,6,ch94data}; + +/* char: 0x5d ']' */ + +static const GLubyte ch93data[] = { +0xc0,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0xc0, +}; + +static const BitmapCharRec ch93 = {2,10,0,2,3,ch93data}; + +/* char: 0x5c '\' */ + +static const GLubyte ch92data[] = { +0x20,0x20,0x40,0x40,0x40,0x40,0x80,0x80, +}; + +static const BitmapCharRec ch92 = {3,8,0,0,3,ch92data}; + +/* char: 0x5b '[' */ + +static const GLubyte ch91data[] = { +0xc0,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0xc0, +}; + +static const BitmapCharRec ch91 = {2,10,-1,2,3,ch91data}; + +/* char: 0x5a 'Z' */ + +static const GLubyte ch90data[] = { +0xf8,0x80,0x40,0x20,0x20,0x10,0x8,0xf8, +}; + +static const BitmapCharRec ch90 = {5,8,-1,0,7,ch90data}; + +/* char: 0x59 'Y' */ + +static const GLubyte ch89data[] = { +0x10,0x10,0x10,0x28,0x28,0x44,0x44,0x82, +}; + +static const BitmapCharRec ch89 = {7,8,0,0,7,ch89data}; + +/* char: 0x58 'X' */ + +static const GLubyte ch88data[] = { +0x88,0x88,0x50,0x50,0x20,0x50,0x88,0x88, +}; + +static const BitmapCharRec ch88 = {5,8,-1,0,7,ch88data}; + +/* char: 0x57 'W' */ + +static const GLubyte ch87data[] = { +0x22,0x0,0x22,0x0,0x22,0x0,0x55,0x0,0x49,0x0,0x49,0x0,0x88,0x80,0x88,0x80, +}; + +static const BitmapCharRec ch87 = {9,8,0,0,9,ch87data}; + +/* char: 0x56 'V' */ + +static const GLubyte ch86data[] = { +0x10,0x28,0x28,0x44,0x44,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch86 = {7,8,0,0,7,ch86data}; + +/* char: 0x55 'U' */ + +static const GLubyte ch85data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84, +}; + +static const BitmapCharRec ch85 = {6,8,-1,0,8,ch85data}; + +/* char: 0x54 'T' */ + +static const GLubyte ch84data[] = { +0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xf8, +}; + +static const BitmapCharRec ch84 = {5,8,0,0,5,ch84data}; + +/* char: 0x53 'S' */ + +static const GLubyte ch83data[] = { +0x70,0x88,0x88,0x8,0x70,0x80,0x88,0x70, +}; + +static const BitmapCharRec ch83 = {5,8,-1,0,7,ch83data}; + +/* char: 0x52 'R' */ + +static const GLubyte ch82data[] = { +0x88,0x88,0x88,0x88,0xf0,0x88,0x88,0xf0, +}; + +static const BitmapCharRec ch82 = {5,8,-1,0,7,ch82data}; + +/* char: 0x51 'Q' */ + +static const GLubyte ch81data[] = { +0x2,0x7c,0x8c,0x94,0x84,0x84,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch81 = {7,9,-1,1,8,ch81data}; + +/* char: 0x50 'P' */ + +static const GLubyte ch80data[] = { +0x80,0x80,0x80,0x80,0xf0,0x88,0x88,0xf0, +}; + +static const BitmapCharRec ch80 = {5,8,-1,0,7,ch80data}; + +/* char: 0x4f 'O' */ + +static const GLubyte ch79data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x78, +}; + +static const BitmapCharRec ch79 = {6,8,-1,0,8,ch79data}; + +/* char: 0x4e 'N' */ + +static const GLubyte ch78data[] = { +0x8c,0x8c,0x94,0x94,0xa4,0xa4,0xc4,0xc4, +}; + +static const BitmapCharRec ch78 = {6,8,-1,0,8,ch78data}; + +/* char: 0x4d 'M' */ + +static const GLubyte ch77data[] = { +0x92,0x92,0x92,0xaa,0xaa,0xc6,0xc6,0x82, +}; + +static const BitmapCharRec ch77 = {7,8,-1,0,9,ch77data}; + +/* char: 0x4c 'L' */ + +static const GLubyte ch76data[] = { +0xf0,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch76 = {4,8,-1,0,6,ch76data}; + +/* char: 0x4b 'K' */ + +static const GLubyte ch75data[] = { +0x88,0x88,0x90,0x90,0xe0,0xa0,0x90,0x88, +}; + +static const BitmapCharRec ch75 = {5,8,-1,0,7,ch75data}; + +/* char: 0x4a 'J' */ + +static const GLubyte ch74data[] = { +0x60,0x90,0x10,0x10,0x10,0x10,0x10,0x10, +}; + +static const BitmapCharRec ch74 = {4,8,0,0,5,ch74data}; + +/* char: 0x49 'I' */ + +static const GLubyte ch73data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch73 = {1,8,-1,0,3,ch73data}; + +/* char: 0x48 'H' */ + +static const GLubyte ch72data[] = { +0x84,0x84,0x84,0x84,0xfc,0x84,0x84,0x84, +}; + +static const BitmapCharRec ch72 = {6,8,-1,0,8,ch72data}; + +/* char: 0x47 'G' */ + +static const GLubyte ch71data[] = { +0x74,0x8c,0x84,0x8c,0x80,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch71 = {6,8,-1,0,8,ch71data}; + +/* char: 0x46 'F' */ + +static const GLubyte ch70data[] = { +0x80,0x80,0x80,0x80,0xf0,0x80,0x80,0xf8, +}; + +static const BitmapCharRec ch70 = {5,8,-1,0,6,ch70data}; + +/* char: 0x45 'E' */ + +static const GLubyte ch69data[] = { +0xf8,0x80,0x80,0x80,0xf8,0x80,0x80,0xf8, +}; + +static const BitmapCharRec ch69 = {5,8,-1,0,7,ch69data}; + +/* char: 0x44 'D' */ + +static const GLubyte ch68data[] = { +0xf0,0x88,0x84,0x84,0x84,0x84,0x88,0xf0, +}; + +static const BitmapCharRec ch68 = {6,8,-1,0,8,ch68data}; + +/* char: 0x43 'C' */ + +static const GLubyte ch67data[] = { +0x78,0x84,0x80,0x80,0x80,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch67 = {6,8,-1,0,8,ch67data}; + +/* char: 0x42 'B' */ + +static const GLubyte ch66data[] = { +0xf0,0x88,0x88,0x88,0xf0,0x88,0x88,0xf0, +}; + +static const BitmapCharRec ch66 = {5,8,-1,0,7,ch66data}; + +/* char: 0x41 'A' */ + +static const GLubyte ch65data[] = { +0x82,0x82,0x7c,0x44,0x28,0x28,0x10,0x10, +}; + +static const BitmapCharRec ch65 = {7,8,0,0,7,ch65data}; + +/* char: 0x40 '@' */ + +static const GLubyte ch64data[] = { +0x3e,0x0,0x40,0x0,0x9b,0x0,0xa4,0x80,0xa4,0x80,0xa2,0x40,0x92,0x40,0x4d,0x40, +0x20,0x80,0x1f,0x0, +}; + +static const BitmapCharRec ch64 = {10,10,0,2,11,ch64data}; + +/* char: 0x3f '?' */ + +static const GLubyte ch63data[] = { +0x40,0x0,0x40,0x40,0x20,0x10,0x90,0x60, +}; + +static const BitmapCharRec ch63 = {4,8,-1,0,6,ch63data}; + +/* char: 0x3e '>' */ + +static const GLubyte ch62data[] = { +0x80,0x40,0x20,0x40,0x80, +}; + +static const BitmapCharRec ch62 = {3,5,-1,-1,6,ch62data}; + +/* char: 0x3d '=' */ + +static const GLubyte ch61data[] = { +0xf0,0x0,0xf0, +}; + +static const BitmapCharRec ch61 = {4,3,0,-2,5,ch61data}; + +/* char: 0x3c '<' */ + +static const GLubyte ch60data[] = { +0x20,0x40,0x80,0x40,0x20, +}; + +static const BitmapCharRec ch60 = {3,5,-1,-1,6,ch60data}; + +/* char: 0x3b ';' */ + +static const GLubyte ch59data[] = { +0x80,0x40,0x40,0x0,0x0,0x0,0x0,0x40, +}; + +static const BitmapCharRec ch59 = {2,8,0,2,3,ch59data}; + +/* char: 0x3a ':' */ + +static const GLubyte ch58data[] = { +0x80,0x0,0x0,0x0,0x0,0x80, +}; + +static const BitmapCharRec ch58 = {1,6,-1,0,3,ch58data}; + +/* char: 0x39 '9' */ + +static const GLubyte ch57data[] = { +0x70,0x88,0x8,0x68,0x98,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch57 = {5,8,0,0,6,ch57data}; + +/* char: 0x38 '8' */ + +static const GLubyte ch56data[] = { +0x70,0x88,0x88,0x88,0x70,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch56 = {5,8,0,0,6,ch56data}; + +/* char: 0x37 '7' */ + +static const GLubyte ch55data[] = { +0x40,0x40,0x20,0x20,0x10,0x10,0x8,0xf8, +}; + +static const BitmapCharRec ch55 = {5,8,0,0,6,ch55data}; + +/* char: 0x36 '6' */ + +static const GLubyte ch54data[] = { +0x70,0x88,0x88,0xc8,0xb0,0x80,0x88,0x70, +}; + +static const BitmapCharRec ch54 = {5,8,0,0,6,ch54data}; + +/* char: 0x35 '5' */ + +static const GLubyte ch53data[] = { +0x70,0x88,0x8,0x8,0xf0,0x80,0x80,0xf8, +}; + +static const BitmapCharRec ch53 = {5,8,0,0,6,ch53data}; + +/* char: 0x34 '4' */ + +static const GLubyte ch52data[] = { +0x10,0x10,0xf8,0x90,0x50,0x50,0x30,0x10, +}; + +static const BitmapCharRec ch52 = {5,8,0,0,6,ch52data}; + +/* char: 0x33 '3' */ + +static const GLubyte ch51data[] = { +0x70,0x88,0x8,0x8,0x30,0x8,0x88,0x70, +}; + +static const BitmapCharRec ch51 = {5,8,0,0,6,ch51data}; + +/* char: 0x32 '2' */ + +static const GLubyte ch50data[] = { +0xf8,0x80,0x40,0x30,0x8,0x8,0x88,0x70, +}; + +static const BitmapCharRec ch50 = {5,8,0,0,6,ch50data}; + +/* char: 0x31 '1' */ + +static const GLubyte ch49data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0xc0,0x40, +}; + +static const BitmapCharRec ch49 = {2,8,-1,0,6,ch49data}; + +/* char: 0x30 '0' */ + +static const GLubyte ch48data[] = { +0x70,0x88,0x88,0x88,0x88,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch48 = {5,8,0,0,6,ch48data}; + +/* char: 0x2f '/' */ + +static const GLubyte ch47data[] = { +0x80,0x80,0x40,0x40,0x40,0x40,0x20,0x20, +}; + +static const BitmapCharRec ch47 = {3,8,0,0,3,ch47data}; + +/* char: 0x2e '.' */ + +static const GLubyte ch46data[] = { +0x80, +}; + +static const BitmapCharRec ch46 = {1,1,-1,0,3,ch46data}; + +/* char: 0x2d '-' */ + +static const GLubyte ch45data[] = { +0xf8, +}; + +static const BitmapCharRec ch45 = {5,1,-1,-3,7,ch45data}; + +/* char: 0x2c ',' */ + +static const GLubyte ch44data[] = { +0x80,0x40,0x40, +}; + +static const BitmapCharRec ch44 = {2,3,0,2,3,ch44data}; + +/* char: 0x2b '+' */ + +static const GLubyte ch43data[] = { +0x20,0x20,0xf8,0x20,0x20, +}; + +static const BitmapCharRec ch43 = {5,5,0,-1,6,ch43data}; + +/* char: 0x2a '*' */ + +static const GLubyte ch42data[] = { +0xa0,0x40,0xa0, +}; + +static const BitmapCharRec ch42 = {3,3,0,-5,4,ch42data}; + +/* char: 0x29 ')' */ + +static const GLubyte ch41data[] = { +0x80,0x40,0x40,0x20,0x20,0x20,0x20,0x40,0x40,0x80, +}; + +static const BitmapCharRec ch41 = {3,10,-1,2,4,ch41data}; + +/* char: 0x28 '(' */ + +static const GLubyte ch40data[] = { +0x20,0x40,0x40,0x80,0x80,0x80,0x80,0x40,0x40,0x20, +}; + +static const BitmapCharRec ch40 = {3,10,0,2,4,ch40data}; + +/* char: 0x27 ''' */ + +static const GLubyte ch39data[] = { +0x80,0x40,0x40, +}; + +static const BitmapCharRec ch39 = {2,3,-1,-5,3,ch39data}; + +/* char: 0x26 '&' */ + +static const GLubyte ch38data[] = { +0x64,0x98,0x98,0xa4,0x60,0x50,0x50,0x20, +}; + +static const BitmapCharRec ch38 = {6,8,-1,0,8,ch38data}; + +/* char: 0x25 '%' */ + +static const GLubyte ch37data[] = { +0x26,0x29,0x16,0x10,0x8,0x68,0x94,0x64, +}; + +static const BitmapCharRec ch37 = {8,8,0,0,9,ch37data}; + +/* char: 0x24 '$' */ + +static const GLubyte ch36data[] = { +0x20,0x70,0xa8,0x28,0x70,0xa0,0xa8,0x70,0x20, +}; + +static const BitmapCharRec ch36 = {5,9,0,1,6,ch36data}; + +/* char: 0x23 '#' */ + +static const GLubyte ch35data[] = { +0x50,0x50,0xf8,0x28,0x7c,0x28,0x28, +}; + +static const BitmapCharRec ch35 = {6,7,0,0,6,ch35data}; + +/* char: 0x22 '"' */ + +static const GLubyte ch34data[] = { +0xa0,0xa0, +}; + +static const BitmapCharRec ch34 = {3,2,-1,-6,4,ch34data}; + +/* char: 0x21 '!' */ + +static const GLubyte ch33data[] = { +0x80,0x0,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch33 = {1,8,-1,0,3,ch33data}; + +/* char: 0x20 ' ' */ + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch32data[] = { 0x0 }; +static const BitmapCharRec ch32 = {0,0,0,0,3,ch32data}; +#else +static const BitmapCharRec ch32 = {0,0,0,0,3,0}; +#endif + +static const BitmapCharRec * const chars[] = { +&ch32, +&ch33, +&ch34, +&ch35, +&ch36, +&ch37, +&ch38, +&ch39, +&ch40, +&ch41, +&ch42, +&ch43, +&ch44, +&ch45, +&ch46, +&ch47, +&ch48, +&ch49, +&ch50, +&ch51, +&ch52, +&ch53, +&ch54, +&ch55, +&ch56, +&ch57, +&ch58, +&ch59, +&ch60, +&ch61, +&ch62, +&ch63, +&ch64, +&ch65, +&ch66, +&ch67, +&ch68, +&ch69, +&ch70, +&ch71, +&ch72, +&ch73, +&ch74, +&ch75, +&ch76, +&ch77, +&ch78, +&ch79, +&ch80, +&ch81, +&ch82, +&ch83, +&ch84, +&ch85, +&ch86, +&ch87, +&ch88, +&ch89, +&ch90, +&ch91, +&ch92, +&ch93, +&ch94, +&ch95, +&ch96, +&ch97, +&ch98, +&ch99, +&ch100, +&ch101, +&ch102, +&ch103, +&ch104, +&ch105, +&ch106, +&ch107, +&ch108, +&ch109, +&ch110, +&ch111, +&ch112, +&ch113, +&ch114, +&ch115, +&ch116, +&ch117, +&ch118, +&ch119, +&ch120, +&ch121, +&ch122, +&ch123, +&ch124, +&ch125, +&ch126, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +&ch160, +&ch161, +&ch162, +&ch163, +&ch164, +&ch165, +&ch166, +&ch167, +&ch168, +&ch169, +&ch170, +&ch171, +&ch172, +&ch173, +&ch174, +&ch175, +&ch176, +&ch177, +&ch178, +&ch179, +&ch180, +&ch181, +&ch182, +&ch183, +&ch184, +&ch185, +&ch186, +&ch187, +&ch188, +&ch189, +&ch190, +&ch191, +&ch192, +&ch193, +&ch194, +&ch195, +&ch196, +&ch197, +&ch198, +&ch199, +&ch200, +&ch201, +&ch202, +&ch203, +&ch204, +&ch205, +&ch206, +&ch207, +&ch208, +&ch209, +&ch210, +&ch211, +&ch212, +&ch213, +&ch214, +&ch215, +&ch216, +&ch217, +&ch218, +&ch219, +&ch220, +&ch221, +&ch222, +&ch223, +&ch224, +&ch225, +&ch226, +&ch227, +&ch228, +&ch229, +&ch230, +&ch231, +&ch232, +&ch233, +&ch234, +&ch235, +&ch236, +&ch237, +&ch238, +&ch239, +&ch240, +&ch241, +&ch242, +&ch243, +&ch244, +&ch245, +&ch246, +&ch247, +&ch248, +&ch249, +&ch250, +&ch251, +&ch252, +&ch253, +&ch254, +&ch255, +}; + +#if !defined(__IBMCPP__) +const +#endif +BitmapFontRec glutBitmapHelvetica10 = { +"-adobe-helvetica-medium-r-normal--10-100-75-75-p-56-iso8859-1", +224, +32, +chars +}; + diff --git a/src/glut/os2/glut_hel12.cpp b/src/glut/os2/glut_hel12.cpp index cbd0548b93..f615fbb2f8 100644 --- a/src/glut/os2/glut_hel12.cpp +++ b/src/glut/os2/glut_hel12.cpp @@ -1,1791 +1,1791 @@ - -/* GENERATED FILE -- DO NOT MODIFY */ - -#define glutBitmapHelvetica12 XXX -#include "glutbitmap.h" -#undef glutBitmapHelvetica12 - -/* char: 0xff */ - -static const GLubyte ch255data[] = { -0xc0,0x20,0x20,0x20,0x30,0x50,0x50,0x48,0x88,0x88,0x0,0x50, -}; - -static const BitmapCharRec ch255 = {5,12,-1,3,7,ch255data}; - -/* char: 0xfe */ - -static const GLubyte ch254data[] = { -0x80,0x80,0x80,0xb0,0xc8,0x88,0x88,0x88,0xc8,0xb0,0x80,0x80, -}; - -static const BitmapCharRec ch254 = {5,12,-1,3,7,ch254data}; - -/* char: 0xfd */ - -static const GLubyte ch253data[] = { -0x80,0x40,0x20,0x20,0x50,0x50,0x90,0x88,0x88,0x88,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch253 = {5,13,-1,3,7,ch253data}; - -/* char: 0xfc */ - -static const GLubyte ch252data[] = { -0x68,0x98,0x88,0x88,0x88,0x88,0x88,0x0,0x50, -}; - -static const BitmapCharRec ch252 = {5,9,-1,0,7,ch252data}; - -/* char: 0xfb */ - -static const GLubyte ch251data[] = { -0x68,0x98,0x88,0x88,0x88,0x88,0x88,0x0,0x50,0x20, -}; - -static const BitmapCharRec ch251 = {5,10,-1,0,7,ch251data}; - -/* char: 0xfa */ - -static const GLubyte ch250data[] = { -0x68,0x98,0x88,0x88,0x88,0x88,0x88,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch250 = {5,10,-1,0,7,ch250data}; - -/* char: 0xf9 */ - -static const GLubyte ch249data[] = { -0x68,0x98,0x88,0x88,0x88,0x88,0x88,0x0,0x20,0x40, -}; - -static const BitmapCharRec ch249 = {5,10,-1,0,7,ch249data}; - -/* char: 0xf8 */ - -static const GLubyte ch248data[] = { -0xb8,0x44,0x64,0x54,0x4c,0x44,0x3a, -}; - -static const BitmapCharRec ch248 = {7,7,0,0,7,ch248data}; - -/* char: 0xf7 */ - -static const GLubyte ch247data[] = { -0x20,0x0,0xf8,0x0,0x20, -}; - -static const BitmapCharRec ch247 = {5,5,-1,-1,7,ch247data}; - -/* char: 0xf6 */ - -static const GLubyte ch246data[] = { -0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x50, -}; - -static const BitmapCharRec ch246 = {5,9,-1,0,7,ch246data}; - -/* char: 0xf5 */ - -static const GLubyte ch245data[] = { -0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch245 = {5,10,-1,0,7,ch245data}; - -/* char: 0xf4 */ - -static const GLubyte ch244data[] = { -0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x50,0x20, -}; - -static const BitmapCharRec ch244 = {5,10,-1,0,7,ch244data}; - -/* char: 0xf3 */ - -static const GLubyte ch243data[] = { -0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch243 = {5,10,-1,0,7,ch243data}; - -/* char: 0xf2 */ - -static const GLubyte ch242data[] = { -0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x20,0x40, -}; - -static const BitmapCharRec ch242 = {5,10,-1,0,7,ch242data}; - -/* char: 0xf1 */ - -static const GLubyte ch241data[] = { -0x88,0x88,0x88,0x88,0x88,0xc8,0xb0,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch241 = {5,10,-1,0,7,ch241data}; - -/* char: 0xf0 */ - -static const GLubyte ch240data[] = { -0x70,0x88,0x88,0x88,0x88,0x78,0x8,0x50,0x30,0x68, -}; - -static const BitmapCharRec ch240 = {5,10,-1,0,7,ch240data}; - -/* char: 0xef */ - -static const GLubyte ch239data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0, -}; - -static const BitmapCharRec ch239 = {3,9,0,0,3,ch239data}; - -/* char: 0xee */ - -static const GLubyte ch238data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0,0x40, -}; - -static const BitmapCharRec ch238 = {3,10,0,0,3,ch238data}; - -/* char: 0xed */ - -static const GLubyte ch237data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x40, -}; - -static const BitmapCharRec ch237 = {2,10,-1,0,3,ch237data}; - -/* char: 0xec */ - -static const GLubyte ch236data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40,0x80, -}; - -static const BitmapCharRec ch236 = {2,10,0,0,3,ch236data}; - -/* char: 0xeb */ - -static const GLubyte ch235data[] = { -0x70,0x88,0x80,0xf8,0x88,0x88,0x70,0x0,0x50, -}; - -static const BitmapCharRec ch235 = {5,9,-1,0,7,ch235data}; - -/* char: 0xea */ - -static const GLubyte ch234data[] = { -0x70,0x88,0x80,0xf8,0x88,0x88,0x70,0x0,0x50,0x20, -}; - -static const BitmapCharRec ch234 = {5,10,-1,0,7,ch234data}; - -/* char: 0xe9 */ - -static const GLubyte ch233data[] = { -0x70,0x88,0x80,0xf8,0x88,0x88,0x70,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch233 = {5,10,-1,0,7,ch233data}; - -/* char: 0xe8 */ - -static const GLubyte ch232data[] = { -0x70,0x88,0x80,0xf8,0x88,0x88,0x70,0x0,0x20,0x40, -}; - -static const BitmapCharRec ch232 = {5,10,-1,0,7,ch232data}; - -/* char: 0xe7 */ - -static const GLubyte ch231data[] = { -0x60,0x10,0x20,0x70,0x88,0x80,0x80,0x80,0x88,0x70, -}; - -static const BitmapCharRec ch231 = {5,10,-1,3,7,ch231data}; - -/* char: 0xe6 */ - -static const GLubyte ch230data[] = { -0x77,0x0,0x88,0x80,0x88,0x0,0x7f,0x80,0x8,0x80,0x88,0x80,0x77,0x0, -}; - -static const BitmapCharRec ch230 = {9,7,-1,0,11,ch230data}; - -/* char: 0xe5 */ - -static const GLubyte ch229data[] = { -0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x30,0x48,0x30, -}; - -static const BitmapCharRec ch229 = {6,10,-1,0,7,ch229data}; - -/* char: 0xe4 */ - -static const GLubyte ch228data[] = { -0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x50, -}; - -static const BitmapCharRec ch228 = {6,9,-1,0,7,ch228data}; - -/* char: 0xe3 */ - -static const GLubyte ch227data[] = { -0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x50,0x28, -}; - -static const BitmapCharRec ch227 = {6,10,-1,0,7,ch227data}; - -/* char: 0xe2 */ - -static const GLubyte ch226data[] = { -0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x50,0x20, -}; - -static const BitmapCharRec ch226 = {6,10,-1,0,7,ch226data}; - -/* char: 0xe1 */ - -static const GLubyte ch225data[] = { -0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x20,0x10, -}; - -static const BitmapCharRec ch225 = {6,10,-1,0,7,ch225data}; - -/* char: 0xe0 */ - -static const GLubyte ch224data[] = { -0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch224 = {6,10,-1,0,7,ch224data}; - -/* char: 0xdf */ - -static const GLubyte ch223data[] = { -0xb0,0x88,0x88,0x88,0xb0,0x88,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch223 = {5,9,-1,0,7,ch223data}; - -/* char: 0xde */ - -static const GLubyte ch222data[] = { -0x80,0x80,0xf8,0x84,0x84,0x84,0xf8,0x80,0x80, -}; - -static const BitmapCharRec ch222 = {6,9,-1,0,8,ch222data}; - -/* char: 0xdd */ - -static const GLubyte ch221data[] = { -0x10,0x10,0x10,0x10,0x28,0x44,0x44,0x82,0x82,0x0,0x10,0x8, -}; - -static const BitmapCharRec ch221 = {7,12,-1,0,9,ch221data}; - -/* char: 0xdc */ - -static const GLubyte ch220data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x48, -}; - -static const BitmapCharRec ch220 = {6,11,-1,0,8,ch220data}; - -/* char: 0xdb */ - -static const GLubyte ch219data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x28,0x10, -}; - -static const BitmapCharRec ch219 = {6,12,-1,0,8,ch219data}; - -/* char: 0xda */ - -static const GLubyte ch218data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x10,0x8, -}; - -static const BitmapCharRec ch218 = {6,12,-1,0,8,ch218data}; - -/* char: 0xd9 */ - -static const GLubyte ch217data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch217 = {6,12,-1,0,8,ch217data}; - -/* char: 0xd8 */ - -static const GLubyte ch216data[] = { -0x80,0x0,0x5e,0x0,0x21,0x0,0x50,0x80,0x48,0x80,0x44,0x80,0x44,0x80,0x42,0x80, -0x21,0x0,0x1e,0x80,0x0,0x40, -}; - -static const BitmapCharRec ch216 = {10,11,0,1,10,ch216data}; - -/* char: 0xd7 */ - -static const GLubyte ch215data[] = { -0x88,0x50,0x20,0x50,0x88, -}; - -static const BitmapCharRec ch215 = {5,5,-1,-1,7,ch215data}; - -/* char: 0xd6 */ - -static const GLubyte ch214data[] = { -0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x24, -}; - -static const BitmapCharRec ch214 = {8,11,-1,0,10,ch214data}; - -/* char: 0xd5 */ - -static const GLubyte ch213data[] = { -0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x28,0x14, -}; - -static const BitmapCharRec ch213 = {8,12,-1,0,10,ch213data}; - -/* char: 0xd4 */ - -static const GLubyte ch212data[] = { -0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x14,0x8, -}; - -static const BitmapCharRec ch212 = {8,12,-1,0,10,ch212data}; - -/* char: 0xd3 */ - -static const GLubyte ch211data[] = { -0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x8,0x4, -}; - -static const BitmapCharRec ch211 = {8,12,-1,0,10,ch211data}; - -/* char: 0xd2 */ - -static const GLubyte ch210data[] = { -0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x8,0x10, -}; - -static const BitmapCharRec ch210 = {8,12,-1,0,10,ch210data}; - -/* char: 0xd1 */ - -static const GLubyte ch209data[] = { -0x82,0x86,0x8a,0x8a,0x92,0xa2,0xa2,0xc2,0x82,0x0,0x28,0x14, -}; - -static const BitmapCharRec ch209 = {7,12,-1,0,9,ch209data}; - -/* char: 0xd0 */ - -static const GLubyte ch208data[] = { -0x7c,0x42,0x41,0x41,0xf1,0x41,0x41,0x42,0x7c, -}; - -static const BitmapCharRec ch208 = {8,9,0,0,9,ch208data}; - -/* char: 0xcf */ - -static const GLubyte ch207data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0, -}; - -static const BitmapCharRec ch207 = {3,11,0,0,3,ch207data}; - -/* char: 0xce */ - -static const GLubyte ch206data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0,0x40, -}; - -static const BitmapCharRec ch206 = {3,12,0,0,3,ch206data}; - -/* char: 0xcd */ - -static const GLubyte ch205data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x40, -}; - -static const BitmapCharRec ch205 = {2,12,-1,0,3,ch205data}; - -/* char: 0xcc */ - -static const GLubyte ch204data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40,0x80, -}; - -static const BitmapCharRec ch204 = {2,12,0,0,3,ch204data}; - -/* char: 0xcb */ - -static const GLubyte ch203data[] = { -0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc,0x0,0x28, -}; - -static const BitmapCharRec ch203 = {6,11,-1,0,8,ch203data}; - -/* char: 0xca */ - -static const GLubyte ch202data[] = { -0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc,0x0,0x28,0x10, -}; - -static const BitmapCharRec ch202 = {6,12,-1,0,8,ch202data}; - -/* char: 0xc9 */ - -static const GLubyte ch201data[] = { -0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc,0x0,0x10,0x8, -}; - -static const BitmapCharRec ch201 = {6,12,-1,0,8,ch201data}; - -/* char: 0xc8 */ - -static const GLubyte ch200data[] = { -0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch200 = {6,12,-1,0,8,ch200data}; - -/* char: 0xc7 */ - -static const GLubyte ch199data[] = { -0x30,0x8,0x8,0x3c,0x42,0x80,0x80,0x80,0x80,0x80,0x42,0x3c, -}; - -static const BitmapCharRec ch199 = {7,12,-1,3,9,ch199data}; - -/* char: 0xc6 */ - -static const GLubyte ch198data[] = { -0x8f,0x80,0x88,0x0,0x88,0x0,0x78,0x0,0x4f,0x80,0x48,0x0,0x28,0x0,0x28,0x0, -0x1f,0x80, -}; - -static const BitmapCharRec ch198 = {9,9,-1,0,11,ch198data}; - -/* char: 0xc5 */ - -static const GLubyte ch197data[] = { -0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x10,0x28,0x10, -}; - -static const BitmapCharRec ch197 = {7,12,-1,0,9,ch197data}; - -/* char: 0xc4 */ - -static const GLubyte ch196data[] = { -0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x28, -}; - -static const BitmapCharRec ch196 = {7,11,-1,0,9,ch196data}; - -/* char: 0xc3 */ - -static const GLubyte ch195data[] = { -0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x28,0x14, -}; - -static const BitmapCharRec ch195 = {7,12,-1,0,9,ch195data}; - -/* char: 0xc2 */ - -static const GLubyte ch194data[] = { -0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x28,0x10, -}; - -static const BitmapCharRec ch194 = {7,12,-1,0,9,ch194data}; - -/* char: 0xc1 */ - -static const GLubyte ch193data[] = { -0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x10,0x8, -}; - -static const BitmapCharRec ch193 = {7,12,-1,0,9,ch193data}; - -/* char: 0xc0 */ - -static const GLubyte ch192data[] = { -0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x10,0x20, -}; - -static const BitmapCharRec ch192 = {7,12,-1,0,9,ch192data}; - -/* char: 0xbf */ - -static const GLubyte ch191data[] = { -0x70,0x88,0x88,0x40,0x40,0x20,0x20,0x0,0x20, -}; - -static const BitmapCharRec ch191 = {5,9,-1,3,7,ch191data}; - -/* char: 0xbe */ - -static const GLubyte ch190data[] = { -0x21,0x0,0x17,0x80,0x15,0x0,0xb,0x0,0xc9,0x0,0x24,0x0,0x44,0x0,0x22,0x0, -0xe1,0x0, -}; - -static const BitmapCharRec ch190 = {9,9,0,0,10,ch190data}; - -/* char: 0xbd */ - -static const GLubyte ch189data[] = { -0x47,0x80,0x22,0x0,0x11,0x0,0x14,0x80,0x4b,0x0,0x48,0x0,0x44,0x0,0xc2,0x0, -0x41,0x0, -}; - -static const BitmapCharRec ch189 = {9,9,0,0,10,ch189data}; - -/* char: 0xbc */ - -static const GLubyte ch188data[] = { -0x41,0x0,0x27,0x80,0x15,0x0,0x13,0x0,0x49,0x0,0x44,0x0,0x44,0x0,0xc2,0x0, -0x41,0x0, -}; - -static const BitmapCharRec ch188 = {9,9,0,0,10,ch188data}; - -/* char: 0xbb */ - -static const GLubyte ch187data[] = { -0xa0,0x50,0x28,0x50,0xa0, -}; - -static const BitmapCharRec ch187 = {5,5,-1,-1,7,ch187data}; - -/* char: 0xba */ - -static const GLubyte ch186data[] = { -0xe0,0x0,0xe0,0xa0,0xe0, -}; - -static const BitmapCharRec ch186 = {3,5,-1,-4,5,ch186data}; - -/* char: 0xb9 */ - -static const GLubyte ch185data[] = { -0x40,0x40,0x40,0xc0,0x40, -}; - -static const BitmapCharRec ch185 = {2,5,-1,-3,4,ch185data}; - -/* char: 0xb8 */ - -static const GLubyte ch184data[] = { -0xc0,0x20,0x20,0x40, -}; - -static const BitmapCharRec ch184 = {3,4,0,3,3,ch184data}; - -/* char: 0xb7 */ - -static const GLubyte ch183data[] = { -0x80, -}; - -static const BitmapCharRec ch183 = {1,1,-1,-3,3,ch183data}; - -/* char: 0xb6 */ - -static const GLubyte ch182data[] = { -0x28,0x28,0x28,0x28,0x28,0x28,0x68,0xe8,0xe8,0xe8,0x68,0x3c, -}; - -static const BitmapCharRec ch182 = {6,12,0,3,7,ch182data}; - -/* char: 0xb5 */ - -static const GLubyte ch181data[] = { -0x80,0x80,0x80,0xe8,0x98,0x88,0x88,0x88,0x88,0x88, -}; - -static const BitmapCharRec ch181 = {5,10,-1,3,7,ch181data}; - -/* char: 0xb4 */ - -static const GLubyte ch180data[] = { -0x80,0x40, -}; - -static const BitmapCharRec ch180 = {2,2,0,-8,2,ch180data}; - -/* char: 0xb3 */ - -static const GLubyte ch179data[] = { -0xc0,0x20,0x40,0x20,0xe0, -}; - -static const BitmapCharRec ch179 = {3,5,0,-3,4,ch179data}; - -/* char: 0xb2 */ - -static const GLubyte ch178data[] = { -0xf0,0x40,0x20,0x90,0x60, -}; - -static const BitmapCharRec ch178 = {4,5,0,-3,4,ch178data}; - -/* char: 0xb1 */ - -static const GLubyte ch177data[] = { -0xf8,0x0,0x20,0x20,0xf8,0x20,0x20, -}; - -static const BitmapCharRec ch177 = {5,7,-1,0,7,ch177data}; - -/* char: 0xb0 */ - -static const GLubyte ch176data[] = { -0x60,0x90,0x90,0x60, -}; - -static const BitmapCharRec ch176 = {4,4,0,-4,5,ch176data}; - -/* char: 0xaf */ - -static const GLubyte ch175data[] = { -0xf0, -}; - -static const BitmapCharRec ch175 = {4,1,0,-8,4,ch175data}; - -/* char: 0xae */ - -static const GLubyte ch174data[] = { -0x3e,0x0,0x41,0x0,0x94,0x80,0x94,0x80,0x98,0x80,0x94,0x80,0x9c,0x80,0x41,0x0, -0x3e,0x0, -}; - -static const BitmapCharRec ch174 = {9,9,-1,0,11,ch174data}; - -/* char: 0xad */ - -static const GLubyte ch173data[] = { -0xf0, -}; - -static const BitmapCharRec ch173 = {4,1,0,-3,5,ch173data}; - -/* char: 0xac */ - -static const GLubyte ch172data[] = { -0x4,0x4,0x4,0xfc, -}; - -static const BitmapCharRec ch172 = {6,4,-1,-2,8,ch172data}; - -/* char: 0xab */ - -static const GLubyte ch171data[] = { -0x28,0x50,0xa0,0x50,0x28, -}; - -static const BitmapCharRec ch171 = {5,5,-1,-1,7,ch171data}; - -/* char: 0xaa */ - -static const GLubyte ch170data[] = { -0xe0,0x0,0xa0,0x20,0xe0, -}; - -static const BitmapCharRec ch170 = {3,5,-1,-4,5,ch170data}; - -/* char: 0xa9 */ - -static const GLubyte ch169data[] = { -0x3e,0x0,0x41,0x0,0x9c,0x80,0xa2,0x80,0xa0,0x80,0xa2,0x80,0x9c,0x80,0x41,0x0, -0x3e,0x0, -}; - -static const BitmapCharRec ch169 = {9,9,-1,0,11,ch169data}; - -/* char: 0xa8 */ - -static const GLubyte ch168data[] = { -0xa0, -}; - -static const BitmapCharRec ch168 = {3,1,0,-8,3,ch168data}; - -/* char: 0xa7 */ - -static const GLubyte ch167data[] = { -0x70,0x88,0x8,0x30,0x48,0x88,0x88,0x90,0x60,0x80,0x88,0x70, -}; - -static const BitmapCharRec ch167 = {5,12,0,3,6,ch167data}; - -/* char: 0xa6 */ - -static const GLubyte ch166data[] = { -0x80,0x80,0x80,0x80,0x0,0x0,0x0,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch166 = {1,11,-1,2,3,ch166data}; - -/* char: 0xa5 */ - -static const GLubyte ch165data[] = { -0x20,0x20,0xf8,0x20,0xf8,0x20,0x50,0x88,0x88, -}; - -static const BitmapCharRec ch165 = {5,9,-1,0,7,ch165data}; - -/* char: 0xa4 */ - -static const GLubyte ch164data[] = { -0x84,0x78,0x48,0x48,0x78,0x84, -}; - -static const BitmapCharRec ch164 = {6,6,0,-1,7,ch164data}; - -/* char: 0xa3 */ - -static const GLubyte ch163data[] = { -0xb0,0x48,0x20,0x20,0xf0,0x40,0x40,0x48,0x30, -}; - -static const BitmapCharRec ch163 = {5,9,-1,0,7,ch163data}; - -/* char: 0xa2 */ - -static const GLubyte ch162data[] = { -0x40,0x70,0xc8,0xa0,0xa0,0xa0,0xa8,0x70,0x10, -}; - -static const BitmapCharRec ch162 = {5,9,-1,1,7,ch162data}; - -/* char: 0xa1 */ - -static const GLubyte ch161data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, -}; - -static const BitmapCharRec ch161 = {1,10,-1,3,3,ch161data}; - -/* char: 0xa0 */ - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch160data[] = { 0x0 }; -static const BitmapCharRec ch160 = {1,1,0,0,4,ch160data}; -#else -static const BitmapCharRec ch160 = {0,0,0,0,4,0}; -#endif - -/* char: 0x7e '~' */ - -static const GLubyte ch126data[] = { -0x98,0x64, -}; - -static const BitmapCharRec ch126 = {6,2,0,-3,7,ch126data}; - -/* char: 0x7d '}' */ - -static const GLubyte ch125data[] = { -0xc0,0x20,0x20,0x20,0x20,0x20,0x10,0x20,0x20,0x20,0x20,0xc0, -}; - -static const BitmapCharRec ch125 = {4,12,0,3,4,ch125data}; - -/* char: 0x7c '|' */ - -static const GLubyte ch124data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch124 = {1,12,-1,3,3,ch124data}; - -/* char: 0x7b '{' */ - -static const GLubyte ch123data[] = { -0x30,0x40,0x40,0x40,0x40,0x40,0x80,0x40,0x40,0x40,0x40,0x30, -}; - -static const BitmapCharRec ch123 = {4,12,0,3,4,ch123data}; - -/* char: 0x7a 'z' */ - -static const GLubyte ch122data[] = { -0xf0,0x80,0x40,0x40,0x20,0x10,0xf0, -}; - -static const BitmapCharRec ch122 = {4,7,-1,0,6,ch122data}; - -/* char: 0x79 'y' */ - -static const GLubyte ch121data[] = { -0x80,0x40,0x20,0x20,0x50,0x50,0x90,0x88,0x88,0x88, -}; - -static const BitmapCharRec ch121 = {5,10,-1,3,7,ch121data}; - -/* char: 0x78 'x' */ - -static const GLubyte ch120data[] = { -0x84,0x84,0x48,0x30,0x30,0x48,0x84, -}; - -static const BitmapCharRec ch120 = {6,7,0,0,6,ch120data}; - -/* char: 0x77 'w' */ - -static const GLubyte ch119data[] = { -0x22,0x0,0x22,0x0,0x55,0x0,0x49,0x0,0x49,0x0,0x88,0x80,0x88,0x80, -}; - -static const BitmapCharRec ch119 = {9,7,0,0,9,ch119data}; - -/* char: 0x76 'v' */ - -static const GLubyte ch118data[] = { -0x20,0x20,0x50,0x50,0x88,0x88,0x88, -}; - -static const BitmapCharRec ch118 = {5,7,-1,0,7,ch118data}; - -/* char: 0x75 'u' */ - -static const GLubyte ch117data[] = { -0x68,0x98,0x88,0x88,0x88,0x88,0x88, -}; - -static const BitmapCharRec ch117 = {5,7,-1,0,7,ch117data}; - -/* char: 0x74 't' */ - -static const GLubyte ch116data[] = { -0x60,0x40,0x40,0x40,0x40,0x40,0xe0,0x40,0x40, -}; - -static const BitmapCharRec ch116 = {3,9,0,0,3,ch116data}; - -/* char: 0x73 's' */ - -static const GLubyte ch115data[] = { -0x60,0x90,0x10,0x60,0x80,0x90,0x60, -}; - -static const BitmapCharRec ch115 = {4,7,-1,0,6,ch115data}; - -/* char: 0x72 'r' */ - -static const GLubyte ch114data[] = { -0x80,0x80,0x80,0x80,0x80,0xc0,0xa0, -}; - -static const BitmapCharRec ch114 = {3,7,-1,0,4,ch114data}; - -/* char: 0x71 'q' */ - -static const GLubyte ch113data[] = { -0x8,0x8,0x8,0x68,0x98,0x88,0x88,0x88,0x98,0x68, -}; - -static const BitmapCharRec ch113 = {5,10,-1,3,7,ch113data}; - -/* char: 0x70 'p' */ - -static const GLubyte ch112data[] = { -0x80,0x80,0x80,0xb0,0xc8,0x88,0x88,0x88,0xc8,0xb0, -}; - -static const BitmapCharRec ch112 = {5,10,-1,3,7,ch112data}; - -/* char: 0x6f 'o' */ - -static const GLubyte ch111data[] = { -0x70,0x88,0x88,0x88,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch111 = {5,7,-1,0,7,ch111data}; - -/* char: 0x6e 'n' */ - -static const GLubyte ch110data[] = { -0x88,0x88,0x88,0x88,0x88,0xc8,0xb0, -}; - -static const BitmapCharRec ch110 = {5,7,-1,0,7,ch110data}; - -/* char: 0x6d 'm' */ - -static const GLubyte ch109data[] = { -0x92,0x92,0x92,0x92,0x92,0xda,0xa4, -}; - -static const BitmapCharRec ch109 = {7,7,-1,0,9,ch109data}; - -/* char: 0x6c 'l' */ - -static const GLubyte ch108data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch108 = {1,9,-1,0,3,ch108data}; - -/* char: 0x6b 'k' */ - -static const GLubyte ch107data[] = { -0x88,0x90,0xa0,0xc0,0xc0,0xa0,0x90,0x80,0x80, -}; - -static const BitmapCharRec ch107 = {5,9,-1,0,6,ch107data}; - -/* char: 0x6a 'j' */ - -static const GLubyte ch106data[] = { -0x80,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40, -}; - -static const BitmapCharRec ch106 = {2,12,0,3,3,ch106data}; - -/* char: 0x69 'i' */ - -static const GLubyte ch105data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, -}; - -static const BitmapCharRec ch105 = {1,9,-1,0,3,ch105data}; - -/* char: 0x68 'h' */ - -static const GLubyte ch104data[] = { -0x88,0x88,0x88,0x88,0x88,0xc8,0xb0,0x80,0x80, -}; - -static const BitmapCharRec ch104 = {5,9,-1,0,7,ch104data}; - -/* char: 0x67 'g' */ - -static const GLubyte ch103data[] = { -0x70,0x88,0x8,0x68,0x98,0x88,0x88,0x88,0x98,0x68, -}; - -static const BitmapCharRec ch103 = {5,10,-1,3,7,ch103data}; - -/* char: 0x66 'f' */ - -static const GLubyte ch102data[] = { -0x40,0x40,0x40,0x40,0x40,0x40,0xe0,0x40,0x30, -}; - -static const BitmapCharRec ch102 = {4,9,0,0,3,ch102data}; - -/* char: 0x65 'e' */ - -static const GLubyte ch101data[] = { -0x70,0x88,0x80,0xf8,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch101 = {5,7,-1,0,7,ch101data}; - -/* char: 0x64 'd' */ - -static const GLubyte ch100data[] = { -0x68,0x98,0x88,0x88,0x88,0x98,0x68,0x8,0x8, -}; - -static const BitmapCharRec ch100 = {5,9,-1,0,7,ch100data}; - -/* char: 0x63 'c' */ - -static const GLubyte ch99data[] = { -0x70,0x88,0x80,0x80,0x80,0x88,0x70, -}; - -static const BitmapCharRec ch99 = {5,7,-1,0,7,ch99data}; - -/* char: 0x62 'b' */ - -static const GLubyte ch98data[] = { -0xb0,0xc8,0x88,0x88,0x88,0xc8,0xb0,0x80,0x80, -}; - -static const BitmapCharRec ch98 = {5,9,-1,0,7,ch98data}; - -/* char: 0x61 'a' */ - -static const GLubyte ch97data[] = { -0x74,0x88,0x88,0x78,0x8,0x88,0x70, -}; - -static const BitmapCharRec ch97 = {6,7,-1,0,7,ch97data}; - -/* char: 0x60 '`' */ - -static const GLubyte ch96data[] = { -0xc0,0x80,0x40, -}; - -static const BitmapCharRec ch96 = {2,3,0,-6,3,ch96data}; - -/* char: 0x5f '_' */ - -static const GLubyte ch95data[] = { -0xfe, -}; - -static const BitmapCharRec ch95 = {7,1,0,2,7,ch95data}; - -/* char: 0x5e '^' */ - -static const GLubyte ch94data[] = { -0x88,0x50,0x20, -}; - -static const BitmapCharRec ch94 = {5,3,0,-5,6,ch94data}; - -/* char: 0x5d ']' */ - -static const GLubyte ch93data[] = { -0xc0,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0xc0, -}; - -static const BitmapCharRec ch93 = {2,12,0,3,3,ch93data}; - -/* char: 0x5c '\' */ - -static const GLubyte ch92data[] = { -0x10,0x10,0x20,0x20,0x20,0x40,0x40,0x80,0x80, -}; - -static const BitmapCharRec ch92 = {4,9,0,0,4,ch92data}; - -/* char: 0x5b '[' */ - -static const GLubyte ch91data[] = { -0xc0,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0xc0, -}; - -static const BitmapCharRec ch91 = {2,12,-1,3,3,ch91data}; - -/* char: 0x5a 'Z' */ - -static const GLubyte ch90data[] = { -0xfe,0x80,0x40,0x20,0x10,0x8,0x4,0x2,0xfe, -}; - -static const BitmapCharRec ch90 = {7,9,-1,0,9,ch90data}; - -/* char: 0x59 'Y' */ - -static const GLubyte ch89data[] = { -0x10,0x10,0x10,0x10,0x28,0x44,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch89 = {7,9,-1,0,9,ch89data}; - -/* char: 0x58 'X' */ - -static const GLubyte ch88data[] = { -0x82,0x44,0x44,0x28,0x10,0x28,0x44,0x44,0x82, -}; - -static const BitmapCharRec ch88 = {7,9,-1,0,9,ch88data}; - -/* char: 0x57 'W' */ - -static const GLubyte ch87data[] = { -0x22,0x0,0x22,0x0,0x22,0x0,0x55,0x0,0x55,0x0,0x49,0x0,0x88,0x80,0x88,0x80, -0x88,0x80, -}; - -static const BitmapCharRec ch87 = {9,9,-1,0,11,ch87data}; - -/* char: 0x56 'V' */ - -static const GLubyte ch86data[] = { -0x10,0x10,0x28,0x28,0x44,0x44,0x44,0x82,0x82, -}; - -static const BitmapCharRec ch86 = {7,9,-1,0,9,ch86data}; - -/* char: 0x55 'U' */ - -static const GLubyte ch85data[] = { -0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84, -}; - -static const BitmapCharRec ch85 = {6,9,-1,0,8,ch85data}; - -/* char: 0x54 'T' */ - -static const GLubyte ch84data[] = { -0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xfe, -}; - -static const BitmapCharRec ch84 = {7,9,0,0,7,ch84data}; - -/* char: 0x53 'S' */ - -static const GLubyte ch83data[] = { -0x78,0x84,0x84,0x4,0x18,0x60,0x80,0x84,0x78, -}; - -static const BitmapCharRec ch83 = {6,9,-1,0,8,ch83data}; - -/* char: 0x52 'R' */ - -static const GLubyte ch82data[] = { -0x84,0x84,0x84,0x88,0xf8,0x84,0x84,0x84,0xf8, -}; - -static const BitmapCharRec ch82 = {6,9,-1,0,8,ch82data}; - -/* char: 0x51 'Q' */ - -static const GLubyte ch81data[] = { -0x3d,0x42,0x85,0x89,0x81,0x81,0x81,0x42,0x3c, -}; - -static const BitmapCharRec ch81 = {8,9,-1,0,10,ch81data}; - -/* char: 0x50 'P' */ - -static const GLubyte ch80data[] = { -0x80,0x80,0x80,0x80,0xf8,0x84,0x84,0x84,0xf8, -}; - -static const BitmapCharRec ch80 = {6,9,-1,0,8,ch80data}; - -/* char: 0x4f 'O' */ - -static const GLubyte ch79data[] = { -0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c, -}; - -static const BitmapCharRec ch79 = {8,9,-1,0,10,ch79data}; - -/* char: 0x4e 'N' */ - -static const GLubyte ch78data[] = { -0x82,0x86,0x8a,0x8a,0x92,0xa2,0xa2,0xc2,0x82, -}; - -static const BitmapCharRec ch78 = {7,9,-1,0,9,ch78data}; - -/* char: 0x4d 'M' */ - -static const GLubyte ch77data[] = { -0x88,0x80,0x88,0x80,0x94,0x80,0x94,0x80,0xa2,0x80,0xa2,0x80,0xc1,0x80,0xc1,0x80, -0x80,0x80, -}; - -static const BitmapCharRec ch77 = {9,9,-1,0,11,ch77data}; - -/* char: 0x4c 'L' */ - -static const GLubyte ch76data[] = { -0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch76 = {5,9,-1,0,7,ch76data}; - -/* char: 0x4b 'K' */ - -static const GLubyte ch75data[] = { -0x82,0x84,0x88,0x90,0xe0,0xa0,0x90,0x88,0x84, -}; - -static const BitmapCharRec ch75 = {7,9,-1,0,8,ch75data}; - -/* char: 0x4a 'J' */ - -static const GLubyte ch74data[] = { -0x70,0x88,0x88,0x8,0x8,0x8,0x8,0x8,0x8, -}; - -static const BitmapCharRec ch74 = {5,9,-1,0,7,ch74data}; - -/* char: 0x49 'I' */ - -static const GLubyte ch73data[] = { -0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch73 = {1,9,-1,0,3,ch73data}; - -/* char: 0x48 'H' */ - -static const GLubyte ch72data[] = { -0x82,0x82,0x82,0x82,0xfe,0x82,0x82,0x82,0x82, -}; - -static const BitmapCharRec ch72 = {7,9,-1,0,9,ch72data}; - -/* char: 0x47 'G' */ - -static const GLubyte ch71data[] = { -0x3a,0x46,0x82,0x82,0x8e,0x80,0x80,0x42,0x3c, -}; - -static const BitmapCharRec ch71 = {7,9,-1,0,9,ch71data}; - -/* char: 0x46 'F' */ - -static const GLubyte ch70data[] = { -0x80,0x80,0x80,0x80,0xf8,0x80,0x80,0x80,0xfc, -}; - -static const BitmapCharRec ch70 = {6,9,-1,0,8,ch70data}; - -/* char: 0x45 'E' */ - -static const GLubyte ch69data[] = { -0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc, -}; - -static const BitmapCharRec ch69 = {6,9,-1,0,8,ch69data}; - -/* char: 0x44 'D' */ - -static const GLubyte ch68data[] = { -0xf8,0x84,0x82,0x82,0x82,0x82,0x82,0x84,0xf8, -}; - -static const BitmapCharRec ch68 = {7,9,-1,0,9,ch68data}; - -/* char: 0x43 'C' */ - -static const GLubyte ch67data[] = { -0x3c,0x42,0x80,0x80,0x80,0x80,0x80,0x42,0x3c, -}; - -static const BitmapCharRec ch67 = {7,9,-1,0,9,ch67data}; - -/* char: 0x42 'B' */ - -static const GLubyte ch66data[] = { -0xf8,0x84,0x84,0x84,0xf8,0x84,0x84,0x84,0xf8, -}; - -static const BitmapCharRec ch66 = {6,9,-1,0,8,ch66data}; - -/* char: 0x41 'A' */ - -static const GLubyte ch65data[] = { -0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x28,0x10, -}; - -static const BitmapCharRec ch65 = {7,9,-1,0,9,ch65data}; - -/* char: 0x40 '@' */ - -static const GLubyte ch64data[] = { -0x3e,0x0,0x40,0x0,0x9b,0x0,0xa6,0x80,0xa2,0x40,0xa2,0x40,0x92,0x40,0x4d,0x40, -0x60,0x80,0x1f,0x0, -}; - -static const BitmapCharRec ch64 = {10,10,-1,1,12,ch64data}; - -/* char: 0x3f '?' */ - -static const GLubyte ch63data[] = { -0x20,0x0,0x20,0x20,0x10,0x10,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch63 = {5,9,-1,0,7,ch63data}; - -/* char: 0x3e '>' */ - -static const GLubyte ch62data[] = { -0xc0,0x30,0xc,0x30,0xc0, -}; - -static const BitmapCharRec ch62 = {6,5,-1,-1,7,ch62data}; - -/* char: 0x3d '=' */ - -static const GLubyte ch61data[] = { -0xf8,0x0,0xf8, -}; - -static const BitmapCharRec ch61 = {5,3,-1,-2,7,ch61data}; - -/* char: 0x3c '<' */ - -static const GLubyte ch60data[] = { -0xc,0x30,0xc0,0x30,0xc, -}; - -static const BitmapCharRec ch60 = {6,5,0,-1,7,ch60data}; - -/* char: 0x3b ';' */ - -static const GLubyte ch59data[] = { -0x80,0x40,0x40,0x0,0x0,0x0,0x0,0x40, -}; - -static const BitmapCharRec ch59 = {2,8,0,2,3,ch59data}; - -/* char: 0x3a ':' */ - -static const GLubyte ch58data[] = { -0x80,0x0,0x0,0x0,0x0,0x80, -}; - -static const BitmapCharRec ch58 = {1,6,-1,0,3,ch58data}; - -/* char: 0x39 '9' */ - -static const GLubyte ch57data[] = { -0x70,0x88,0x8,0x8,0x78,0x88,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch57 = {5,9,-1,0,7,ch57data}; - -/* char: 0x38 '8' */ - -static const GLubyte ch56data[] = { -0x70,0x88,0x88,0x88,0x88,0x70,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch56 = {5,9,-1,0,7,ch56data}; - -/* char: 0x37 '7' */ - -static const GLubyte ch55data[] = { -0x40,0x40,0x20,0x20,0x20,0x10,0x10,0x8,0xf8, -}; - -static const BitmapCharRec ch55 = {5,9,-1,0,7,ch55data}; - -/* char: 0x36 '6' */ - -static const GLubyte ch54data[] = { -0x70,0x88,0x88,0x88,0xc8,0xb0,0x80,0x88,0x70, -}; - -static const BitmapCharRec ch54 = {5,9,-1,0,7,ch54data}; - -/* char: 0x35 '5' */ - -static const GLubyte ch53data[] = { -0x70,0x88,0x88,0x8,0x8,0xf0,0x80,0x80,0xf8, -}; - -static const BitmapCharRec ch53 = {5,9,-1,0,7,ch53data}; - -/* char: 0x34 '4' */ - -static const GLubyte ch52data[] = { -0x8,0x8,0xfc,0x88,0x48,0x28,0x28,0x18,0x8, -}; - -static const BitmapCharRec ch52 = {6,9,0,0,7,ch52data}; - -/* char: 0x33 '3' */ - -static const GLubyte ch51data[] = { -0x70,0x88,0x88,0x8,0x8,0x30,0x8,0x88,0x70, -}; - -static const BitmapCharRec ch51 = {5,9,-1,0,7,ch51data}; - -/* char: 0x32 '2' */ - -static const GLubyte ch50data[] = { -0xf8,0x80,0x80,0x40,0x20,0x10,0x8,0x88,0x70, -}; - -static const BitmapCharRec ch50 = {5,9,-1,0,7,ch50data}; - -/* char: 0x31 '1' */ - -static const GLubyte ch49data[] = { -0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xe0,0x20, -}; - -static const BitmapCharRec ch49 = {3,9,-1,0,7,ch49data}; - -/* char: 0x30 '0' */ - -static const GLubyte ch48data[] = { -0x70,0x88,0x88,0x88,0x88,0x88,0x88,0x88,0x70, -}; - -static const BitmapCharRec ch48 = {5,9,-1,0,7,ch48data}; - -/* char: 0x2f '/' */ - -static const GLubyte ch47data[] = { -0x80,0x80,0x40,0x40,0x40,0x20,0x20,0x10,0x10, -}; - -static const BitmapCharRec ch47 = {4,9,0,0,4,ch47data}; - -/* char: 0x2e '.' */ - -static const GLubyte ch46data[] = { -0x80, -}; - -static const BitmapCharRec ch46 = {1,1,-1,0,3,ch46data}; - -/* char: 0x2d '-' */ - -static const GLubyte ch45data[] = { -0xf8, -}; - -static const BitmapCharRec ch45 = {5,1,-1,-3,8,ch45data}; - -/* char: 0x2c ',' */ - -static const GLubyte ch44data[] = { -0x80,0x40,0x40, -}; - -static const BitmapCharRec ch44 = {2,3,-1,2,4,ch44data}; - -/* char: 0x2b '+' */ - -static const GLubyte ch43data[] = { -0x20,0x20,0xf8,0x20,0x20, -}; - -static const BitmapCharRec ch43 = {5,5,-1,-1,7,ch43data}; - -/* char: 0x2a '*' */ - -static const GLubyte ch42data[] = { -0xa0,0x40,0xa0, -}; - -static const BitmapCharRec ch42 = {3,3,-1,-6,5,ch42data}; - -/* char: 0x29 ')' */ - -static const GLubyte ch41data[] = { -0x80,0x40,0x40,0x20,0x20,0x20,0x20,0x20,0x20,0x40,0x40,0x80, -}; - -static const BitmapCharRec ch41 = {3,12,0,3,4,ch41data}; - -/* char: 0x28 '(' */ - -static const GLubyte ch40data[] = { -0x20,0x40,0x40,0x80,0x80,0x80,0x80,0x80,0x80,0x40,0x40,0x20, -}; - -static const BitmapCharRec ch40 = {3,12,-1,3,4,ch40data}; - -/* char: 0x27 ''' */ - -static const GLubyte ch39data[] = { -0x80,0x40,0xc0, -}; - -static const BitmapCharRec ch39 = {2,3,-1,-6,3,ch39data}; - -/* char: 0x26 '&' */ - -static const GLubyte ch38data[] = { -0x72,0x8c,0x84,0x8a,0x50,0x30,0x48,0x48,0x30, -}; - -static const BitmapCharRec ch38 = {7,9,-1,0,9,ch38data}; - -/* char: 0x25 '%' */ - -static const GLubyte ch37data[] = { -0x23,0x0,0x14,0x80,0x14,0x80,0x13,0x0,0x8,0x0,0x68,0x0,0x94,0x0,0x94,0x0, -0x62,0x0, -}; - -static const BitmapCharRec ch37 = {9,9,-1,0,11,ch37data}; - -/* char: 0x24 '$' */ - -static const GLubyte ch36data[] = { -0x20,0x70,0xa8,0xa8,0x28,0x70,0xa0,0xa8,0x70,0x20, -}; - -static const BitmapCharRec ch36 = {5,10,-1,1,7,ch36data}; - -/* char: 0x23 '#' */ - -static const GLubyte ch35data[] = { -0x50,0x50,0x50,0xfc,0x28,0xfc,0x28,0x28, -}; - -static const BitmapCharRec ch35 = {6,8,0,0,7,ch35data}; - -/* char: 0x22 '"' */ - -static const GLubyte ch34data[] = { -0xa0,0xa0,0xa0, -}; - -static const BitmapCharRec ch34 = {3,3,-1,-6,5,ch34data}; - -/* char: 0x21 '!' */ - -static const GLubyte ch33data[] = { -0x80,0x0,0x80,0x80,0x80,0x80,0x80,0x80,0x80, -}; - -static const BitmapCharRec ch33 = {1,9,-1,0,3,ch33data}; - -/* char: 0x20 ' ' */ - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch32data[] = { 0x0 }; -static const BitmapCharRec ch32 = {1,1,0,0,4,ch32data}; -#else -static const BitmapCharRec ch32 = {0,0,0,0,4,0}; -#endif - -static const BitmapCharRec * const chars[] = { -&ch32, -&ch33, -&ch34, -&ch35, -&ch36, -&ch37, -&ch38, -&ch39, -&ch40, -&ch41, -&ch42, -&ch43, -&ch44, -&ch45, -&ch46, -&ch47, -&ch48, -&ch49, -&ch50, -&ch51, -&ch52, -&ch53, -&ch54, -&ch55, -&ch56, -&ch57, -&ch58, -&ch59, -&ch60, -&ch61, -&ch62, -&ch63, -&ch64, -&ch65, -&ch66, -&ch67, -&ch68, -&ch69, -&ch70, -&ch71, -&ch72, -&ch73, -&ch74, -&ch75, -&ch76, -&ch77, -&ch78, -&ch79, -&ch80, -&ch81, -&ch82, -&ch83, -&ch84, -&ch85, -&ch86, -&ch87, -&ch88, -&ch89, -&ch90, -&ch91, -&ch92, -&ch93, -&ch94, -&ch95, -&ch96, -&ch97, -&ch98, -&ch99, -&ch100, -&ch101, -&ch102, -&ch103, -&ch104, -&ch105, -&ch106, -&ch107, -&ch108, -&ch109, -&ch110, -&ch111, -&ch112, -&ch113, -&ch114, -&ch115, -&ch116, -&ch117, -&ch118, -&ch119, -&ch120, -&ch121, -&ch122, -&ch123, -&ch124, -&ch125, -&ch126, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -&ch160, -&ch161, -&ch162, -&ch163, -&ch164, -&ch165, -&ch166, -&ch167, -&ch168, -&ch169, -&ch170, -&ch171, -&ch172, -&ch173, -&ch174, -&ch175, -&ch176, -&ch177, -&ch178, -&ch179, -&ch180, -&ch181, -&ch182, -&ch183, -&ch184, -&ch185, -&ch186, -&ch187, -&ch188, -&ch189, -&ch190, -&ch191, -&ch192, -&ch193, -&ch194, -&ch195, -&ch196, -&ch197, -&ch198, -&ch199, -&ch200, -&ch201, -&ch202, -&ch203, -&ch204, -&ch205, -&ch206, -&ch207, -&ch208, -&ch209, -&ch210, -&ch211, -&ch212, -&ch213, -&ch214, -&ch215, -&ch216, -&ch217, -&ch218, -&ch219, -&ch220, -&ch221, -&ch222, -&ch223, -&ch224, -&ch225, -&ch226, -&ch227, -&ch228, -&ch229, -&ch230, -&ch231, -&ch232, -&ch233, -&ch234, -&ch235, -&ch236, -&ch237, -&ch238, -&ch239, -&ch240, -&ch241, -&ch242, -&ch243, -&ch244, -&ch245, -&ch246, -&ch247, -&ch248, -&ch249, -&ch250, -&ch251, -&ch252, -&ch253, -&ch254, -&ch255, -}; - -#if !defined(__IBMCPP__) -const -#endif -BitmapFontRec glutBitmapHelvetica12 = { -"-adobe-helvetica-medium-r-normal--12-120-75-75-p-67-iso8859-1", -224, -32, -chars -}; - + +/* GENERATED FILE -- DO NOT MODIFY */ + +#define glutBitmapHelvetica12 XXX +#include "glutbitmap.h" +#undef glutBitmapHelvetica12 + +/* char: 0xff */ + +static const GLubyte ch255data[] = { +0xc0,0x20,0x20,0x20,0x30,0x50,0x50,0x48,0x88,0x88,0x0,0x50, +}; + +static const BitmapCharRec ch255 = {5,12,-1,3,7,ch255data}; + +/* char: 0xfe */ + +static const GLubyte ch254data[] = { +0x80,0x80,0x80,0xb0,0xc8,0x88,0x88,0x88,0xc8,0xb0,0x80,0x80, +}; + +static const BitmapCharRec ch254 = {5,12,-1,3,7,ch254data}; + +/* char: 0xfd */ + +static const GLubyte ch253data[] = { +0x80,0x40,0x20,0x20,0x50,0x50,0x90,0x88,0x88,0x88,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch253 = {5,13,-1,3,7,ch253data}; + +/* char: 0xfc */ + +static const GLubyte ch252data[] = { +0x68,0x98,0x88,0x88,0x88,0x88,0x88,0x0,0x50, +}; + +static const BitmapCharRec ch252 = {5,9,-1,0,7,ch252data}; + +/* char: 0xfb */ + +static const GLubyte ch251data[] = { +0x68,0x98,0x88,0x88,0x88,0x88,0x88,0x0,0x50,0x20, +}; + +static const BitmapCharRec ch251 = {5,10,-1,0,7,ch251data}; + +/* char: 0xfa */ + +static const GLubyte ch250data[] = { +0x68,0x98,0x88,0x88,0x88,0x88,0x88,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch250 = {5,10,-1,0,7,ch250data}; + +/* char: 0xf9 */ + +static const GLubyte ch249data[] = { +0x68,0x98,0x88,0x88,0x88,0x88,0x88,0x0,0x20,0x40, +}; + +static const BitmapCharRec ch249 = {5,10,-1,0,7,ch249data}; + +/* char: 0xf8 */ + +static const GLubyte ch248data[] = { +0xb8,0x44,0x64,0x54,0x4c,0x44,0x3a, +}; + +static const BitmapCharRec ch248 = {7,7,0,0,7,ch248data}; + +/* char: 0xf7 */ + +static const GLubyte ch247data[] = { +0x20,0x0,0xf8,0x0,0x20, +}; + +static const BitmapCharRec ch247 = {5,5,-1,-1,7,ch247data}; + +/* char: 0xf6 */ + +static const GLubyte ch246data[] = { +0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x50, +}; + +static const BitmapCharRec ch246 = {5,9,-1,0,7,ch246data}; + +/* char: 0xf5 */ + +static const GLubyte ch245data[] = { +0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch245 = {5,10,-1,0,7,ch245data}; + +/* char: 0xf4 */ + +static const GLubyte ch244data[] = { +0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x50,0x20, +}; + +static const BitmapCharRec ch244 = {5,10,-1,0,7,ch244data}; + +/* char: 0xf3 */ + +static const GLubyte ch243data[] = { +0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch243 = {5,10,-1,0,7,ch243data}; + +/* char: 0xf2 */ + +static const GLubyte ch242data[] = { +0x70,0x88,0x88,0x88,0x88,0x88,0x70,0x0,0x20,0x40, +}; + +static const BitmapCharRec ch242 = {5,10,-1,0,7,ch242data}; + +/* char: 0xf1 */ + +static const GLubyte ch241data[] = { +0x88,0x88,0x88,0x88,0x88,0xc8,0xb0,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch241 = {5,10,-1,0,7,ch241data}; + +/* char: 0xf0 */ + +static const GLubyte ch240data[] = { +0x70,0x88,0x88,0x88,0x88,0x78,0x8,0x50,0x30,0x68, +}; + +static const BitmapCharRec ch240 = {5,10,-1,0,7,ch240data}; + +/* char: 0xef */ + +static const GLubyte ch239data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0, +}; + +static const BitmapCharRec ch239 = {3,9,0,0,3,ch239data}; + +/* char: 0xee */ + +static const GLubyte ch238data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0,0x40, +}; + +static const BitmapCharRec ch238 = {3,10,0,0,3,ch238data}; + +/* char: 0xed */ + +static const GLubyte ch237data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x40, +}; + +static const BitmapCharRec ch237 = {2,10,-1,0,3,ch237data}; + +/* char: 0xec */ + +static const GLubyte ch236data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40,0x80, +}; + +static const BitmapCharRec ch236 = {2,10,0,0,3,ch236data}; + +/* char: 0xeb */ + +static const GLubyte ch235data[] = { +0x70,0x88,0x80,0xf8,0x88,0x88,0x70,0x0,0x50, +}; + +static const BitmapCharRec ch235 = {5,9,-1,0,7,ch235data}; + +/* char: 0xea */ + +static const GLubyte ch234data[] = { +0x70,0x88,0x80,0xf8,0x88,0x88,0x70,0x0,0x50,0x20, +}; + +static const BitmapCharRec ch234 = {5,10,-1,0,7,ch234data}; + +/* char: 0xe9 */ + +static const GLubyte ch233data[] = { +0x70,0x88,0x80,0xf8,0x88,0x88,0x70,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch233 = {5,10,-1,0,7,ch233data}; + +/* char: 0xe8 */ + +static const GLubyte ch232data[] = { +0x70,0x88,0x80,0xf8,0x88,0x88,0x70,0x0,0x20,0x40, +}; + +static const BitmapCharRec ch232 = {5,10,-1,0,7,ch232data}; + +/* char: 0xe7 */ + +static const GLubyte ch231data[] = { +0x60,0x10,0x20,0x70,0x88,0x80,0x80,0x80,0x88,0x70, +}; + +static const BitmapCharRec ch231 = {5,10,-1,3,7,ch231data}; + +/* char: 0xe6 */ + +static const GLubyte ch230data[] = { +0x77,0x0,0x88,0x80,0x88,0x0,0x7f,0x80,0x8,0x80,0x88,0x80,0x77,0x0, +}; + +static const BitmapCharRec ch230 = {9,7,-1,0,11,ch230data}; + +/* char: 0xe5 */ + +static const GLubyte ch229data[] = { +0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x30,0x48,0x30, +}; + +static const BitmapCharRec ch229 = {6,10,-1,0,7,ch229data}; + +/* char: 0xe4 */ + +static const GLubyte ch228data[] = { +0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x50, +}; + +static const BitmapCharRec ch228 = {6,9,-1,0,7,ch228data}; + +/* char: 0xe3 */ + +static const GLubyte ch227data[] = { +0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x50,0x28, +}; + +static const BitmapCharRec ch227 = {6,10,-1,0,7,ch227data}; + +/* char: 0xe2 */ + +static const GLubyte ch226data[] = { +0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x50,0x20, +}; + +static const BitmapCharRec ch226 = {6,10,-1,0,7,ch226data}; + +/* char: 0xe1 */ + +static const GLubyte ch225data[] = { +0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x20,0x10, +}; + +static const BitmapCharRec ch225 = {6,10,-1,0,7,ch225data}; + +/* char: 0xe0 */ + +static const GLubyte ch224data[] = { +0x74,0x88,0x88,0x78,0x8,0x88,0x70,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch224 = {6,10,-1,0,7,ch224data}; + +/* char: 0xdf */ + +static const GLubyte ch223data[] = { +0xb0,0x88,0x88,0x88,0xb0,0x88,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch223 = {5,9,-1,0,7,ch223data}; + +/* char: 0xde */ + +static const GLubyte ch222data[] = { +0x80,0x80,0xf8,0x84,0x84,0x84,0xf8,0x80,0x80, +}; + +static const BitmapCharRec ch222 = {6,9,-1,0,8,ch222data}; + +/* char: 0xdd */ + +static const GLubyte ch221data[] = { +0x10,0x10,0x10,0x10,0x28,0x44,0x44,0x82,0x82,0x0,0x10,0x8, +}; + +static const BitmapCharRec ch221 = {7,12,-1,0,9,ch221data}; + +/* char: 0xdc */ + +static const GLubyte ch220data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x48, +}; + +static const BitmapCharRec ch220 = {6,11,-1,0,8,ch220data}; + +/* char: 0xdb */ + +static const GLubyte ch219data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x28,0x10, +}; + +static const BitmapCharRec ch219 = {6,12,-1,0,8,ch219data}; + +/* char: 0xda */ + +static const GLubyte ch218data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x10,0x8, +}; + +static const BitmapCharRec ch218 = {6,12,-1,0,8,ch218data}; + +/* char: 0xd9 */ + +static const GLubyte ch217data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch217 = {6,12,-1,0,8,ch217data}; + +/* char: 0xd8 */ + +static const GLubyte ch216data[] = { +0x80,0x0,0x5e,0x0,0x21,0x0,0x50,0x80,0x48,0x80,0x44,0x80,0x44,0x80,0x42,0x80, +0x21,0x0,0x1e,0x80,0x0,0x40, +}; + +static const BitmapCharRec ch216 = {10,11,0,1,10,ch216data}; + +/* char: 0xd7 */ + +static const GLubyte ch215data[] = { +0x88,0x50,0x20,0x50,0x88, +}; + +static const BitmapCharRec ch215 = {5,5,-1,-1,7,ch215data}; + +/* char: 0xd6 */ + +static const GLubyte ch214data[] = { +0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x24, +}; + +static const BitmapCharRec ch214 = {8,11,-1,0,10,ch214data}; + +/* char: 0xd5 */ + +static const GLubyte ch213data[] = { +0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x28,0x14, +}; + +static const BitmapCharRec ch213 = {8,12,-1,0,10,ch213data}; + +/* char: 0xd4 */ + +static const GLubyte ch212data[] = { +0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x14,0x8, +}; + +static const BitmapCharRec ch212 = {8,12,-1,0,10,ch212data}; + +/* char: 0xd3 */ + +static const GLubyte ch211data[] = { +0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x8,0x4, +}; + +static const BitmapCharRec ch211 = {8,12,-1,0,10,ch211data}; + +/* char: 0xd2 */ + +static const GLubyte ch210data[] = { +0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c,0x0,0x8,0x10, +}; + +static const BitmapCharRec ch210 = {8,12,-1,0,10,ch210data}; + +/* char: 0xd1 */ + +static const GLubyte ch209data[] = { +0x82,0x86,0x8a,0x8a,0x92,0xa2,0xa2,0xc2,0x82,0x0,0x28,0x14, +}; + +static const BitmapCharRec ch209 = {7,12,-1,0,9,ch209data}; + +/* char: 0xd0 */ + +static const GLubyte ch208data[] = { +0x7c,0x42,0x41,0x41,0xf1,0x41,0x41,0x42,0x7c, +}; + +static const BitmapCharRec ch208 = {8,9,0,0,9,ch208data}; + +/* char: 0xcf */ + +static const GLubyte ch207data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0, +}; + +static const BitmapCharRec ch207 = {3,11,0,0,3,ch207data}; + +/* char: 0xce */ + +static const GLubyte ch206data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0xa0,0x40, +}; + +static const BitmapCharRec ch206 = {3,12,0,0,3,ch206data}; + +/* char: 0xcd */ + +static const GLubyte ch205data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80,0x40, +}; + +static const BitmapCharRec ch205 = {2,12,-1,0,3,ch205data}; + +/* char: 0xcc */ + +static const GLubyte ch204data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40,0x80, +}; + +static const BitmapCharRec ch204 = {2,12,0,0,3,ch204data}; + +/* char: 0xcb */ + +static const GLubyte ch203data[] = { +0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc,0x0,0x28, +}; + +static const BitmapCharRec ch203 = {6,11,-1,0,8,ch203data}; + +/* char: 0xca */ + +static const GLubyte ch202data[] = { +0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc,0x0,0x28,0x10, +}; + +static const BitmapCharRec ch202 = {6,12,-1,0,8,ch202data}; + +/* char: 0xc9 */ + +static const GLubyte ch201data[] = { +0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc,0x0,0x10,0x8, +}; + +static const BitmapCharRec ch201 = {6,12,-1,0,8,ch201data}; + +/* char: 0xc8 */ + +static const GLubyte ch200data[] = { +0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch200 = {6,12,-1,0,8,ch200data}; + +/* char: 0xc7 */ + +static const GLubyte ch199data[] = { +0x30,0x8,0x8,0x3c,0x42,0x80,0x80,0x80,0x80,0x80,0x42,0x3c, +}; + +static const BitmapCharRec ch199 = {7,12,-1,3,9,ch199data}; + +/* char: 0xc6 */ + +static const GLubyte ch198data[] = { +0x8f,0x80,0x88,0x0,0x88,0x0,0x78,0x0,0x4f,0x80,0x48,0x0,0x28,0x0,0x28,0x0, +0x1f,0x80, +}; + +static const BitmapCharRec ch198 = {9,9,-1,0,11,ch198data}; + +/* char: 0xc5 */ + +static const GLubyte ch197data[] = { +0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x10,0x28,0x10, +}; + +static const BitmapCharRec ch197 = {7,12,-1,0,9,ch197data}; + +/* char: 0xc4 */ + +static const GLubyte ch196data[] = { +0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x28, +}; + +static const BitmapCharRec ch196 = {7,11,-1,0,9,ch196data}; + +/* char: 0xc3 */ + +static const GLubyte ch195data[] = { +0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x28,0x14, +}; + +static const BitmapCharRec ch195 = {7,12,-1,0,9,ch195data}; + +/* char: 0xc2 */ + +static const GLubyte ch194data[] = { +0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x28,0x10, +}; + +static const BitmapCharRec ch194 = {7,12,-1,0,9,ch194data}; + +/* char: 0xc1 */ + +static const GLubyte ch193data[] = { +0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x10,0x8, +}; + +static const BitmapCharRec ch193 = {7,12,-1,0,9,ch193data}; + +/* char: 0xc0 */ + +static const GLubyte ch192data[] = { +0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x10,0x10,0x0,0x10,0x20, +}; + +static const BitmapCharRec ch192 = {7,12,-1,0,9,ch192data}; + +/* char: 0xbf */ + +static const GLubyte ch191data[] = { +0x70,0x88,0x88,0x40,0x40,0x20,0x20,0x0,0x20, +}; + +static const BitmapCharRec ch191 = {5,9,-1,3,7,ch191data}; + +/* char: 0xbe */ + +static const GLubyte ch190data[] = { +0x21,0x0,0x17,0x80,0x15,0x0,0xb,0x0,0xc9,0x0,0x24,0x0,0x44,0x0,0x22,0x0, +0xe1,0x0, +}; + +static const BitmapCharRec ch190 = {9,9,0,0,10,ch190data}; + +/* char: 0xbd */ + +static const GLubyte ch189data[] = { +0x47,0x80,0x22,0x0,0x11,0x0,0x14,0x80,0x4b,0x0,0x48,0x0,0x44,0x0,0xc2,0x0, +0x41,0x0, +}; + +static const BitmapCharRec ch189 = {9,9,0,0,10,ch189data}; + +/* char: 0xbc */ + +static const GLubyte ch188data[] = { +0x41,0x0,0x27,0x80,0x15,0x0,0x13,0x0,0x49,0x0,0x44,0x0,0x44,0x0,0xc2,0x0, +0x41,0x0, +}; + +static const BitmapCharRec ch188 = {9,9,0,0,10,ch188data}; + +/* char: 0xbb */ + +static const GLubyte ch187data[] = { +0xa0,0x50,0x28,0x50,0xa0, +}; + +static const BitmapCharRec ch187 = {5,5,-1,-1,7,ch187data}; + +/* char: 0xba */ + +static const GLubyte ch186data[] = { +0xe0,0x0,0xe0,0xa0,0xe0, +}; + +static const BitmapCharRec ch186 = {3,5,-1,-4,5,ch186data}; + +/* char: 0xb9 */ + +static const GLubyte ch185data[] = { +0x40,0x40,0x40,0xc0,0x40, +}; + +static const BitmapCharRec ch185 = {2,5,-1,-3,4,ch185data}; + +/* char: 0xb8 */ + +static const GLubyte ch184data[] = { +0xc0,0x20,0x20,0x40, +}; + +static const BitmapCharRec ch184 = {3,4,0,3,3,ch184data}; + +/* char: 0xb7 */ + +static const GLubyte ch183data[] = { +0x80, +}; + +static const BitmapCharRec ch183 = {1,1,-1,-3,3,ch183data}; + +/* char: 0xb6 */ + +static const GLubyte ch182data[] = { +0x28,0x28,0x28,0x28,0x28,0x28,0x68,0xe8,0xe8,0xe8,0x68,0x3c, +}; + +static const BitmapCharRec ch182 = {6,12,0,3,7,ch182data}; + +/* char: 0xb5 */ + +static const GLubyte ch181data[] = { +0x80,0x80,0x80,0xe8,0x98,0x88,0x88,0x88,0x88,0x88, +}; + +static const BitmapCharRec ch181 = {5,10,-1,3,7,ch181data}; + +/* char: 0xb4 */ + +static const GLubyte ch180data[] = { +0x80,0x40, +}; + +static const BitmapCharRec ch180 = {2,2,0,-8,2,ch180data}; + +/* char: 0xb3 */ + +static const GLubyte ch179data[] = { +0xc0,0x20,0x40,0x20,0xe0, +}; + +static const BitmapCharRec ch179 = {3,5,0,-3,4,ch179data}; + +/* char: 0xb2 */ + +static const GLubyte ch178data[] = { +0xf0,0x40,0x20,0x90,0x60, +}; + +static const BitmapCharRec ch178 = {4,5,0,-3,4,ch178data}; + +/* char: 0xb1 */ + +static const GLubyte ch177data[] = { +0xf8,0x0,0x20,0x20,0xf8,0x20,0x20, +}; + +static const BitmapCharRec ch177 = {5,7,-1,0,7,ch177data}; + +/* char: 0xb0 */ + +static const GLubyte ch176data[] = { +0x60,0x90,0x90,0x60, +}; + +static const BitmapCharRec ch176 = {4,4,0,-4,5,ch176data}; + +/* char: 0xaf */ + +static const GLubyte ch175data[] = { +0xf0, +}; + +static const BitmapCharRec ch175 = {4,1,0,-8,4,ch175data}; + +/* char: 0xae */ + +static const GLubyte ch174data[] = { +0x3e,0x0,0x41,0x0,0x94,0x80,0x94,0x80,0x98,0x80,0x94,0x80,0x9c,0x80,0x41,0x0, +0x3e,0x0, +}; + +static const BitmapCharRec ch174 = {9,9,-1,0,11,ch174data}; + +/* char: 0xad */ + +static const GLubyte ch173data[] = { +0xf0, +}; + +static const BitmapCharRec ch173 = {4,1,0,-3,5,ch173data}; + +/* char: 0xac */ + +static const GLubyte ch172data[] = { +0x4,0x4,0x4,0xfc, +}; + +static const BitmapCharRec ch172 = {6,4,-1,-2,8,ch172data}; + +/* char: 0xab */ + +static const GLubyte ch171data[] = { +0x28,0x50,0xa0,0x50,0x28, +}; + +static const BitmapCharRec ch171 = {5,5,-1,-1,7,ch171data}; + +/* char: 0xaa */ + +static const GLubyte ch170data[] = { +0xe0,0x0,0xa0,0x20,0xe0, +}; + +static const BitmapCharRec ch170 = {3,5,-1,-4,5,ch170data}; + +/* char: 0xa9 */ + +static const GLubyte ch169data[] = { +0x3e,0x0,0x41,0x0,0x9c,0x80,0xa2,0x80,0xa0,0x80,0xa2,0x80,0x9c,0x80,0x41,0x0, +0x3e,0x0, +}; + +static const BitmapCharRec ch169 = {9,9,-1,0,11,ch169data}; + +/* char: 0xa8 */ + +static const GLubyte ch168data[] = { +0xa0, +}; + +static const BitmapCharRec ch168 = {3,1,0,-8,3,ch168data}; + +/* char: 0xa7 */ + +static const GLubyte ch167data[] = { +0x70,0x88,0x8,0x30,0x48,0x88,0x88,0x90,0x60,0x80,0x88,0x70, +}; + +static const BitmapCharRec ch167 = {5,12,0,3,6,ch167data}; + +/* char: 0xa6 */ + +static const GLubyte ch166data[] = { +0x80,0x80,0x80,0x80,0x0,0x0,0x0,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch166 = {1,11,-1,2,3,ch166data}; + +/* char: 0xa5 */ + +static const GLubyte ch165data[] = { +0x20,0x20,0xf8,0x20,0xf8,0x20,0x50,0x88,0x88, +}; + +static const BitmapCharRec ch165 = {5,9,-1,0,7,ch165data}; + +/* char: 0xa4 */ + +static const GLubyte ch164data[] = { +0x84,0x78,0x48,0x48,0x78,0x84, +}; + +static const BitmapCharRec ch164 = {6,6,0,-1,7,ch164data}; + +/* char: 0xa3 */ + +static const GLubyte ch163data[] = { +0xb0,0x48,0x20,0x20,0xf0,0x40,0x40,0x48,0x30, +}; + +static const BitmapCharRec ch163 = {5,9,-1,0,7,ch163data}; + +/* char: 0xa2 */ + +static const GLubyte ch162data[] = { +0x40,0x70,0xc8,0xa0,0xa0,0xa0,0xa8,0x70,0x10, +}; + +static const BitmapCharRec ch162 = {5,9,-1,1,7,ch162data}; + +/* char: 0xa1 */ + +static const GLubyte ch161data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, +}; + +static const BitmapCharRec ch161 = {1,10,-1,3,3,ch161data}; + +/* char: 0xa0 */ + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch160data[] = { 0x0 }; +static const BitmapCharRec ch160 = {1,1,0,0,4,ch160data}; +#else +static const BitmapCharRec ch160 = {0,0,0,0,4,0}; +#endif + +/* char: 0x7e '~' */ + +static const GLubyte ch126data[] = { +0x98,0x64, +}; + +static const BitmapCharRec ch126 = {6,2,0,-3,7,ch126data}; + +/* char: 0x7d '}' */ + +static const GLubyte ch125data[] = { +0xc0,0x20,0x20,0x20,0x20,0x20,0x10,0x20,0x20,0x20,0x20,0xc0, +}; + +static const BitmapCharRec ch125 = {4,12,0,3,4,ch125data}; + +/* char: 0x7c '|' */ + +static const GLubyte ch124data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch124 = {1,12,-1,3,3,ch124data}; + +/* char: 0x7b '{' */ + +static const GLubyte ch123data[] = { +0x30,0x40,0x40,0x40,0x40,0x40,0x80,0x40,0x40,0x40,0x40,0x30, +}; + +static const BitmapCharRec ch123 = {4,12,0,3,4,ch123data}; + +/* char: 0x7a 'z' */ + +static const GLubyte ch122data[] = { +0xf0,0x80,0x40,0x40,0x20,0x10,0xf0, +}; + +static const BitmapCharRec ch122 = {4,7,-1,0,6,ch122data}; + +/* char: 0x79 'y' */ + +static const GLubyte ch121data[] = { +0x80,0x40,0x20,0x20,0x50,0x50,0x90,0x88,0x88,0x88, +}; + +static const BitmapCharRec ch121 = {5,10,-1,3,7,ch121data}; + +/* char: 0x78 'x' */ + +static const GLubyte ch120data[] = { +0x84,0x84,0x48,0x30,0x30,0x48,0x84, +}; + +static const BitmapCharRec ch120 = {6,7,0,0,6,ch120data}; + +/* char: 0x77 'w' */ + +static const GLubyte ch119data[] = { +0x22,0x0,0x22,0x0,0x55,0x0,0x49,0x0,0x49,0x0,0x88,0x80,0x88,0x80, +}; + +static const BitmapCharRec ch119 = {9,7,0,0,9,ch119data}; + +/* char: 0x76 'v' */ + +static const GLubyte ch118data[] = { +0x20,0x20,0x50,0x50,0x88,0x88,0x88, +}; + +static const BitmapCharRec ch118 = {5,7,-1,0,7,ch118data}; + +/* char: 0x75 'u' */ + +static const GLubyte ch117data[] = { +0x68,0x98,0x88,0x88,0x88,0x88,0x88, +}; + +static const BitmapCharRec ch117 = {5,7,-1,0,7,ch117data}; + +/* char: 0x74 't' */ + +static const GLubyte ch116data[] = { +0x60,0x40,0x40,0x40,0x40,0x40,0xe0,0x40,0x40, +}; + +static const BitmapCharRec ch116 = {3,9,0,0,3,ch116data}; + +/* char: 0x73 's' */ + +static const GLubyte ch115data[] = { +0x60,0x90,0x10,0x60,0x80,0x90,0x60, +}; + +static const BitmapCharRec ch115 = {4,7,-1,0,6,ch115data}; + +/* char: 0x72 'r' */ + +static const GLubyte ch114data[] = { +0x80,0x80,0x80,0x80,0x80,0xc0,0xa0, +}; + +static const BitmapCharRec ch114 = {3,7,-1,0,4,ch114data}; + +/* char: 0x71 'q' */ + +static const GLubyte ch113data[] = { +0x8,0x8,0x8,0x68,0x98,0x88,0x88,0x88,0x98,0x68, +}; + +static const BitmapCharRec ch113 = {5,10,-1,3,7,ch113data}; + +/* char: 0x70 'p' */ + +static const GLubyte ch112data[] = { +0x80,0x80,0x80,0xb0,0xc8,0x88,0x88,0x88,0xc8,0xb0, +}; + +static const BitmapCharRec ch112 = {5,10,-1,3,7,ch112data}; + +/* char: 0x6f 'o' */ + +static const GLubyte ch111data[] = { +0x70,0x88,0x88,0x88,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch111 = {5,7,-1,0,7,ch111data}; + +/* char: 0x6e 'n' */ + +static const GLubyte ch110data[] = { +0x88,0x88,0x88,0x88,0x88,0xc8,0xb0, +}; + +static const BitmapCharRec ch110 = {5,7,-1,0,7,ch110data}; + +/* char: 0x6d 'm' */ + +static const GLubyte ch109data[] = { +0x92,0x92,0x92,0x92,0x92,0xda,0xa4, +}; + +static const BitmapCharRec ch109 = {7,7,-1,0,9,ch109data}; + +/* char: 0x6c 'l' */ + +static const GLubyte ch108data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch108 = {1,9,-1,0,3,ch108data}; + +/* char: 0x6b 'k' */ + +static const GLubyte ch107data[] = { +0x88,0x90,0xa0,0xc0,0xc0,0xa0,0x90,0x80,0x80, +}; + +static const BitmapCharRec ch107 = {5,9,-1,0,6,ch107data}; + +/* char: 0x6a 'j' */ + +static const GLubyte ch106data[] = { +0x80,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x0,0x40, +}; + +static const BitmapCharRec ch106 = {2,12,0,3,3,ch106data}; + +/* char: 0x69 'i' */ + +static const GLubyte ch105data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x0,0x80, +}; + +static const BitmapCharRec ch105 = {1,9,-1,0,3,ch105data}; + +/* char: 0x68 'h' */ + +static const GLubyte ch104data[] = { +0x88,0x88,0x88,0x88,0x88,0xc8,0xb0,0x80,0x80, +}; + +static const BitmapCharRec ch104 = {5,9,-1,0,7,ch104data}; + +/* char: 0x67 'g' */ + +static const GLubyte ch103data[] = { +0x70,0x88,0x8,0x68,0x98,0x88,0x88,0x88,0x98,0x68, +}; + +static const BitmapCharRec ch103 = {5,10,-1,3,7,ch103data}; + +/* char: 0x66 'f' */ + +static const GLubyte ch102data[] = { +0x40,0x40,0x40,0x40,0x40,0x40,0xe0,0x40,0x30, +}; + +static const BitmapCharRec ch102 = {4,9,0,0,3,ch102data}; + +/* char: 0x65 'e' */ + +static const GLubyte ch101data[] = { +0x70,0x88,0x80,0xf8,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch101 = {5,7,-1,0,7,ch101data}; + +/* char: 0x64 'd' */ + +static const GLubyte ch100data[] = { +0x68,0x98,0x88,0x88,0x88,0x98,0x68,0x8,0x8, +}; + +static const BitmapCharRec ch100 = {5,9,-1,0,7,ch100data}; + +/* char: 0x63 'c' */ + +static const GLubyte ch99data[] = { +0x70,0x88,0x80,0x80,0x80,0x88,0x70, +}; + +static const BitmapCharRec ch99 = {5,7,-1,0,7,ch99data}; + +/* char: 0x62 'b' */ + +static const GLubyte ch98data[] = { +0xb0,0xc8,0x88,0x88,0x88,0xc8,0xb0,0x80,0x80, +}; + +static const BitmapCharRec ch98 = {5,9,-1,0,7,ch98data}; + +/* char: 0x61 'a' */ + +static const GLubyte ch97data[] = { +0x74,0x88,0x88,0x78,0x8,0x88,0x70, +}; + +static const BitmapCharRec ch97 = {6,7,-1,0,7,ch97data}; + +/* char: 0x60 '`' */ + +static const GLubyte ch96data[] = { +0xc0,0x80,0x40, +}; + +static const BitmapCharRec ch96 = {2,3,0,-6,3,ch96data}; + +/* char: 0x5f '_' */ + +static const GLubyte ch95data[] = { +0xfe, +}; + +static const BitmapCharRec ch95 = {7,1,0,2,7,ch95data}; + +/* char: 0x5e '^' */ + +static const GLubyte ch94data[] = { +0x88,0x50,0x20, +}; + +static const BitmapCharRec ch94 = {5,3,0,-5,6,ch94data}; + +/* char: 0x5d ']' */ + +static const GLubyte ch93data[] = { +0xc0,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0x40,0xc0, +}; + +static const BitmapCharRec ch93 = {2,12,0,3,3,ch93data}; + +/* char: 0x5c '\' */ + +static const GLubyte ch92data[] = { +0x10,0x10,0x20,0x20,0x20,0x40,0x40,0x80,0x80, +}; + +static const BitmapCharRec ch92 = {4,9,0,0,4,ch92data}; + +/* char: 0x5b '[' */ + +static const GLubyte ch91data[] = { +0xc0,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0xc0, +}; + +static const BitmapCharRec ch91 = {2,12,-1,3,3,ch91data}; + +/* char: 0x5a 'Z' */ + +static const GLubyte ch90data[] = { +0xfe,0x80,0x40,0x20,0x10,0x8,0x4,0x2,0xfe, +}; + +static const BitmapCharRec ch90 = {7,9,-1,0,9,ch90data}; + +/* char: 0x59 'Y' */ + +static const GLubyte ch89data[] = { +0x10,0x10,0x10,0x10,0x28,0x44,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch89 = {7,9,-1,0,9,ch89data}; + +/* char: 0x58 'X' */ + +static const GLubyte ch88data[] = { +0x82,0x44,0x44,0x28,0x10,0x28,0x44,0x44,0x82, +}; + +static const BitmapCharRec ch88 = {7,9,-1,0,9,ch88data}; + +/* char: 0x57 'W' */ + +static const GLubyte ch87data[] = { +0x22,0x0,0x22,0x0,0x22,0x0,0x55,0x0,0x55,0x0,0x49,0x0,0x88,0x80,0x88,0x80, +0x88,0x80, +}; + +static const BitmapCharRec ch87 = {9,9,-1,0,11,ch87data}; + +/* char: 0x56 'V' */ + +static const GLubyte ch86data[] = { +0x10,0x10,0x28,0x28,0x44,0x44,0x44,0x82,0x82, +}; + +static const BitmapCharRec ch86 = {7,9,-1,0,9,ch86data}; + +/* char: 0x55 'U' */ + +static const GLubyte ch85data[] = { +0x78,0x84,0x84,0x84,0x84,0x84,0x84,0x84,0x84, +}; + +static const BitmapCharRec ch85 = {6,9,-1,0,8,ch85data}; + +/* char: 0x54 'T' */ + +static const GLubyte ch84data[] = { +0x10,0x10,0x10,0x10,0x10,0x10,0x10,0x10,0xfe, +}; + +static const BitmapCharRec ch84 = {7,9,0,0,7,ch84data}; + +/* char: 0x53 'S' */ + +static const GLubyte ch83data[] = { +0x78,0x84,0x84,0x4,0x18,0x60,0x80,0x84,0x78, +}; + +static const BitmapCharRec ch83 = {6,9,-1,0,8,ch83data}; + +/* char: 0x52 'R' */ + +static const GLubyte ch82data[] = { +0x84,0x84,0x84,0x88,0xf8,0x84,0x84,0x84,0xf8, +}; + +static const BitmapCharRec ch82 = {6,9,-1,0,8,ch82data}; + +/* char: 0x51 'Q' */ + +static const GLubyte ch81data[] = { +0x3d,0x42,0x85,0x89,0x81,0x81,0x81,0x42,0x3c, +}; + +static const BitmapCharRec ch81 = {8,9,-1,0,10,ch81data}; + +/* char: 0x50 'P' */ + +static const GLubyte ch80data[] = { +0x80,0x80,0x80,0x80,0xf8,0x84,0x84,0x84,0xf8, +}; + +static const BitmapCharRec ch80 = {6,9,-1,0,8,ch80data}; + +/* char: 0x4f 'O' */ + +static const GLubyte ch79data[] = { +0x3c,0x42,0x81,0x81,0x81,0x81,0x81,0x42,0x3c, +}; + +static const BitmapCharRec ch79 = {8,9,-1,0,10,ch79data}; + +/* char: 0x4e 'N' */ + +static const GLubyte ch78data[] = { +0x82,0x86,0x8a,0x8a,0x92,0xa2,0xa2,0xc2,0x82, +}; + +static const BitmapCharRec ch78 = {7,9,-1,0,9,ch78data}; + +/* char: 0x4d 'M' */ + +static const GLubyte ch77data[] = { +0x88,0x80,0x88,0x80,0x94,0x80,0x94,0x80,0xa2,0x80,0xa2,0x80,0xc1,0x80,0xc1,0x80, +0x80,0x80, +}; + +static const BitmapCharRec ch77 = {9,9,-1,0,11,ch77data}; + +/* char: 0x4c 'L' */ + +static const GLubyte ch76data[] = { +0xf8,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch76 = {5,9,-1,0,7,ch76data}; + +/* char: 0x4b 'K' */ + +static const GLubyte ch75data[] = { +0x82,0x84,0x88,0x90,0xe0,0xa0,0x90,0x88,0x84, +}; + +static const BitmapCharRec ch75 = {7,9,-1,0,8,ch75data}; + +/* char: 0x4a 'J' */ + +static const GLubyte ch74data[] = { +0x70,0x88,0x88,0x8,0x8,0x8,0x8,0x8,0x8, +}; + +static const BitmapCharRec ch74 = {5,9,-1,0,7,ch74data}; + +/* char: 0x49 'I' */ + +static const GLubyte ch73data[] = { +0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch73 = {1,9,-1,0,3,ch73data}; + +/* char: 0x48 'H' */ + +static const GLubyte ch72data[] = { +0x82,0x82,0x82,0x82,0xfe,0x82,0x82,0x82,0x82, +}; + +static const BitmapCharRec ch72 = {7,9,-1,0,9,ch72data}; + +/* char: 0x47 'G' */ + +static const GLubyte ch71data[] = { +0x3a,0x46,0x82,0x82,0x8e,0x80,0x80,0x42,0x3c, +}; + +static const BitmapCharRec ch71 = {7,9,-1,0,9,ch71data}; + +/* char: 0x46 'F' */ + +static const GLubyte ch70data[] = { +0x80,0x80,0x80,0x80,0xf8,0x80,0x80,0x80,0xfc, +}; + +static const BitmapCharRec ch70 = {6,9,-1,0,8,ch70data}; + +/* char: 0x45 'E' */ + +static const GLubyte ch69data[] = { +0xfc,0x80,0x80,0x80,0xfc,0x80,0x80,0x80,0xfc, +}; + +static const BitmapCharRec ch69 = {6,9,-1,0,8,ch69data}; + +/* char: 0x44 'D' */ + +static const GLubyte ch68data[] = { +0xf8,0x84,0x82,0x82,0x82,0x82,0x82,0x84,0xf8, +}; + +static const BitmapCharRec ch68 = {7,9,-1,0,9,ch68data}; + +/* char: 0x43 'C' */ + +static const GLubyte ch67data[] = { +0x3c,0x42,0x80,0x80,0x80,0x80,0x80,0x42,0x3c, +}; + +static const BitmapCharRec ch67 = {7,9,-1,0,9,ch67data}; + +/* char: 0x42 'B' */ + +static const GLubyte ch66data[] = { +0xf8,0x84,0x84,0x84,0xf8,0x84,0x84,0x84,0xf8, +}; + +static const BitmapCharRec ch66 = {6,9,-1,0,8,ch66data}; + +/* char: 0x41 'A' */ + +static const GLubyte ch65data[] = { +0x82,0x82,0x82,0x7c,0x44,0x44,0x28,0x28,0x10, +}; + +static const BitmapCharRec ch65 = {7,9,-1,0,9,ch65data}; + +/* char: 0x40 '@' */ + +static const GLubyte ch64data[] = { +0x3e,0x0,0x40,0x0,0x9b,0x0,0xa6,0x80,0xa2,0x40,0xa2,0x40,0x92,0x40,0x4d,0x40, +0x60,0x80,0x1f,0x0, +}; + +static const BitmapCharRec ch64 = {10,10,-1,1,12,ch64data}; + +/* char: 0x3f '?' */ + +static const GLubyte ch63data[] = { +0x20,0x0,0x20,0x20,0x10,0x10,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch63 = {5,9,-1,0,7,ch63data}; + +/* char: 0x3e '>' */ + +static const GLubyte ch62data[] = { +0xc0,0x30,0xc,0x30,0xc0, +}; + +static const BitmapCharRec ch62 = {6,5,-1,-1,7,ch62data}; + +/* char: 0x3d '=' */ + +static const GLubyte ch61data[] = { +0xf8,0x0,0xf8, +}; + +static const BitmapCharRec ch61 = {5,3,-1,-2,7,ch61data}; + +/* char: 0x3c '<' */ + +static const GLubyte ch60data[] = { +0xc,0x30,0xc0,0x30,0xc, +}; + +static const BitmapCharRec ch60 = {6,5,0,-1,7,ch60data}; + +/* char: 0x3b ';' */ + +static const GLubyte ch59data[] = { +0x80,0x40,0x40,0x0,0x0,0x0,0x0,0x40, +}; + +static const BitmapCharRec ch59 = {2,8,0,2,3,ch59data}; + +/* char: 0x3a ':' */ + +static const GLubyte ch58data[] = { +0x80,0x0,0x0,0x0,0x0,0x80, +}; + +static const BitmapCharRec ch58 = {1,6,-1,0,3,ch58data}; + +/* char: 0x39 '9' */ + +static const GLubyte ch57data[] = { +0x70,0x88,0x8,0x8,0x78,0x88,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch57 = {5,9,-1,0,7,ch57data}; + +/* char: 0x38 '8' */ + +static const GLubyte ch56data[] = { +0x70,0x88,0x88,0x88,0x88,0x70,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch56 = {5,9,-1,0,7,ch56data}; + +/* char: 0x37 '7' */ + +static const GLubyte ch55data[] = { +0x40,0x40,0x20,0x20,0x20,0x10,0x10,0x8,0xf8, +}; + +static const BitmapCharRec ch55 = {5,9,-1,0,7,ch55data}; + +/* char: 0x36 '6' */ + +static const GLubyte ch54data[] = { +0x70,0x88,0x88,0x88,0xc8,0xb0,0x80,0x88,0x70, +}; + +static const BitmapCharRec ch54 = {5,9,-1,0,7,ch54data}; + +/* char: 0x35 '5' */ + +static const GLubyte ch53data[] = { +0x70,0x88,0x88,0x8,0x8,0xf0,0x80,0x80,0xf8, +}; + +static const BitmapCharRec ch53 = {5,9,-1,0,7,ch53data}; + +/* char: 0x34 '4' */ + +static const GLubyte ch52data[] = { +0x8,0x8,0xfc,0x88,0x48,0x28,0x28,0x18,0x8, +}; + +static const BitmapCharRec ch52 = {6,9,0,0,7,ch52data}; + +/* char: 0x33 '3' */ + +static const GLubyte ch51data[] = { +0x70,0x88,0x88,0x8,0x8,0x30,0x8,0x88,0x70, +}; + +static const BitmapCharRec ch51 = {5,9,-1,0,7,ch51data}; + +/* char: 0x32 '2' */ + +static const GLubyte ch50data[] = { +0xf8,0x80,0x80,0x40,0x20,0x10,0x8,0x88,0x70, +}; + +static const BitmapCharRec ch50 = {5,9,-1,0,7,ch50data}; + +/* char: 0x31 '1' */ + +static const GLubyte ch49data[] = { +0x20,0x20,0x20,0x20,0x20,0x20,0x20,0xe0,0x20, +}; + +static const BitmapCharRec ch49 = {3,9,-1,0,7,ch49data}; + +/* char: 0x30 '0' */ + +static const GLubyte ch48data[] = { +0x70,0x88,0x88,0x88,0x88,0x88,0x88,0x88,0x70, +}; + +static const BitmapCharRec ch48 = {5,9,-1,0,7,ch48data}; + +/* char: 0x2f '/' */ + +static const GLubyte ch47data[] = { +0x80,0x80,0x40,0x40,0x40,0x20,0x20,0x10,0x10, +}; + +static const BitmapCharRec ch47 = {4,9,0,0,4,ch47data}; + +/* char: 0x2e '.' */ + +static const GLubyte ch46data[] = { +0x80, +}; + +static const BitmapCharRec ch46 = {1,1,-1,0,3,ch46data}; + +/* char: 0x2d '-' */ + +static const GLubyte ch45data[] = { +0xf8, +}; + +static const BitmapCharRec ch45 = {5,1,-1,-3,8,ch45data}; + +/* char: 0x2c ',' */ + +static const GLubyte ch44data[] = { +0x80,0x40,0x40, +}; + +static const BitmapCharRec ch44 = {2,3,-1,2,4,ch44data}; + +/* char: 0x2b '+' */ + +static const GLubyte ch43data[] = { +0x20,0x20,0xf8,0x20,0x20, +}; + +static const BitmapCharRec ch43 = {5,5,-1,-1,7,ch43data}; + +/* char: 0x2a '*' */ + +static const GLubyte ch42data[] = { +0xa0,0x40,0xa0, +}; + +static const BitmapCharRec ch42 = {3,3,-1,-6,5,ch42data}; + +/* char: 0x29 ')' */ + +static const GLubyte ch41data[] = { +0x80,0x40,0x40,0x20,0x20,0x20,0x20,0x20,0x20,0x40,0x40,0x80, +}; + +static const BitmapCharRec ch41 = {3,12,0,3,4,ch41data}; + +/* char: 0x28 '(' */ + +static const GLubyte ch40data[] = { +0x20,0x40,0x40,0x80,0x80,0x80,0x80,0x80,0x80,0x40,0x40,0x20, +}; + +static const BitmapCharRec ch40 = {3,12,-1,3,4,ch40data}; + +/* char: 0x27 ''' */ + +static const GLubyte ch39data[] = { +0x80,0x40,0xc0, +}; + +static const BitmapCharRec ch39 = {2,3,-1,-6,3,ch39data}; + +/* char: 0x26 '&' */ + +static const GLubyte ch38data[] = { +0x72,0x8c,0x84,0x8a,0x50,0x30,0x48,0x48,0x30, +}; + +static const BitmapCharRec ch38 = {7,9,-1,0,9,ch38data}; + +/* char: 0x25 '%' */ + +static const GLubyte ch37data[] = { +0x23,0x0,0x14,0x80,0x14,0x80,0x13,0x0,0x8,0x0,0x68,0x0,0x94,0x0,0x94,0x0, +0x62,0x0, +}; + +static const BitmapCharRec ch37 = {9,9,-1,0,11,ch37data}; + +/* char: 0x24 '$' */ + +static const GLubyte ch36data[] = { +0x20,0x70,0xa8,0xa8,0x28,0x70,0xa0,0xa8,0x70,0x20, +}; + +static const BitmapCharRec ch36 = {5,10,-1,1,7,ch36data}; + +/* char: 0x23 '#' */ + +static const GLubyte ch35data[] = { +0x50,0x50,0x50,0xfc,0x28,0xfc,0x28,0x28, +}; + +static const BitmapCharRec ch35 = {6,8,0,0,7,ch35data}; + +/* char: 0x22 '"' */ + +static const GLubyte ch34data[] = { +0xa0,0xa0,0xa0, +}; + +static const BitmapCharRec ch34 = {3,3,-1,-6,5,ch34data}; + +/* char: 0x21 '!' */ + +static const GLubyte ch33data[] = { +0x80,0x0,0x80,0x80,0x80,0x80,0x80,0x80,0x80, +}; + +static const BitmapCharRec ch33 = {1,9,-1,0,3,ch33data}; + +/* char: 0x20 ' ' */ + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch32data[] = { 0x0 }; +static const BitmapCharRec ch32 = {1,1,0,0,4,ch32data}; +#else +static const BitmapCharRec ch32 = {0,0,0,0,4,0}; +#endif + +static const BitmapCharRec * const chars[] = { +&ch32, +&ch33, +&ch34, +&ch35, +&ch36, +&ch37, +&ch38, +&ch39, +&ch40, +&ch41, +&ch42, +&ch43, +&ch44, +&ch45, +&ch46, +&ch47, +&ch48, +&ch49, +&ch50, +&ch51, +&ch52, +&ch53, +&ch54, +&ch55, +&ch56, +&ch57, +&ch58, +&ch59, +&ch60, +&ch61, +&ch62, +&ch63, +&ch64, +&ch65, +&ch66, +&ch67, +&ch68, +&ch69, +&ch70, +&ch71, +&ch72, +&ch73, +&ch74, +&ch75, +&ch76, +&ch77, +&ch78, +&ch79, +&ch80, +&ch81, +&ch82, +&ch83, +&ch84, +&ch85, +&ch86, +&ch87, +&ch88, +&ch89, +&ch90, +&ch91, +&ch92, +&ch93, +&ch94, +&ch95, +&ch96, +&ch97, +&ch98, +&ch99, +&ch100, +&ch101, +&ch102, +&ch103, +&ch104, +&ch105, +&ch106, +&ch107, +&ch108, +&ch109, +&ch110, +&ch111, +&ch112, +&ch113, +&ch114, +&ch115, +&ch116, +&ch117, +&ch118, +&ch119, +&ch120, +&ch121, +&ch122, +&ch123, +&ch124, +&ch125, +&ch126, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +&ch160, +&ch161, +&ch162, +&ch163, +&ch164, +&ch165, +&ch166, +&ch167, +&ch168, +&ch169, +&ch170, +&ch171, +&ch172, +&ch173, +&ch174, +&ch175, +&ch176, +&ch177, +&ch178, +&ch179, +&ch180, +&ch181, +&ch182, +&ch183, +&ch184, +&ch185, +&ch186, +&ch187, +&ch188, +&ch189, +&ch190, +&ch191, +&ch192, +&ch193, +&ch194, +&ch195, +&ch196, +&ch197, +&ch198, +&ch199, +&ch200, +&ch201, +&ch202, +&ch203, +&ch204, +&ch205, +&ch206, +&ch207, +&ch208, +&ch209, +&ch210, +&ch211, +&ch212, +&ch213, +&ch214, +&ch215, +&ch216, +&ch217, +&ch218, +&ch219, +&ch220, +&ch221, +&ch222, +&ch223, +&ch224, +&ch225, +&ch226, +&ch227, +&ch228, +&ch229, +&ch230, +&ch231, +&ch232, +&ch233, +&ch234, +&ch235, +&ch236, +&ch237, +&ch238, +&ch239, +&ch240, +&ch241, +&ch242, +&ch243, +&ch244, +&ch245, +&ch246, +&ch247, +&ch248, +&ch249, +&ch250, +&ch251, +&ch252, +&ch253, +&ch254, +&ch255, +}; + +#if !defined(__IBMCPP__) +const +#endif +BitmapFontRec glutBitmapHelvetica12 = { +"-adobe-helvetica-medium-r-normal--12-120-75-75-p-67-iso8859-1", +224, +32, +chars +}; + diff --git a/src/glut/os2/glut_hel18.cpp b/src/glut/os2/glut_hel18.cpp index 26f88ec505..8fe76332b2 100644 --- a/src/glut/os2/glut_hel18.cpp +++ b/src/glut/os2/glut_hel18.cpp @@ -1,1900 +1,1900 @@ - -/* GENERATED FILE -- DO NOT MODIFY */ - -#define glutBitmapHelvetica18 XXX -#include "glutbitmap.h" -#undef glutBitmapHelvetica18 - -/* char: 0xff */ - -static const GLubyte ch255data[] = { -0x70,0x70,0x18,0x18,0x18,0x18,0x3c,0x24,0x66,0x66,0x66,0xc3,0xc3,0xc3,0x0,0x66, -0x66, -}; - -static const BitmapCharRec ch255 = {8,17,-1,4,10,ch255data}; - -/* char: 0xfe */ - -static const GLubyte ch254data[] = { -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xde,0x0,0xff,0x0,0xe3,0x0,0xc1,0x80, -0xc1,0x80,0xc1,0x80,0xc1,0x80,0xe3,0x0,0xff,0x0,0xde,0x0,0xc0,0x0,0xc0,0x0, -0xc0,0x0,0xc0,0x0, -}; - -static const BitmapCharRec ch254 = {9,18,-1,4,11,ch254data}; - -/* char: 0xfd */ - -static const GLubyte ch253data[] = { -0x70,0x70,0x18,0x18,0x18,0x18,0x3c,0x24,0x66,0x66,0x66,0xc3,0xc3,0xc3,0x0,0x18, -0xc,0x6, -}; - -static const BitmapCharRec ch253 = {8,18,-1,4,10,ch253data}; - -/* char: 0xfc */ - -static const GLubyte ch252data[] = { -0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x0,0x66,0x66, -}; - -static const BitmapCharRec ch252 = {8,13,-1,0,10,ch252data}; - -/* char: 0xfb */ - -static const GLubyte ch251data[] = { -0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x0,0x66,0x3c,0x18, -}; - -static const BitmapCharRec ch251 = {8,14,-1,0,10,ch251data}; - -/* char: 0xfa */ - -static const GLubyte ch250data[] = { -0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x0,0x18,0xc,0x6, -}; - -static const BitmapCharRec ch250 = {8,14,-1,0,10,ch250data}; - -/* char: 0xf9 */ - -static const GLubyte ch249data[] = { -0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x0,0xc,0x18,0x30, -}; - -static const BitmapCharRec ch249 = {8,14,-1,0,10,ch249data}; - -/* char: 0xf8 */ - -static const GLubyte ch248data[] = { -0xce,0x0,0x7f,0x80,0x31,0x80,0x78,0xc0,0x6c,0xc0,0x66,0xc0,0x63,0xc0,0x31,0x80, -0x3f,0xc0,0xe,0x60, -}; - -static const BitmapCharRec ch248 = {11,10,0,0,11,ch248data}; - -/* char: 0xf7 */ - -static const GLubyte ch247data[] = { -0x18,0x18,0x0,0xff,0xff,0x0,0x18,0x18, -}; - -static const BitmapCharRec ch247 = {8,8,-1,-1,10,ch247data}; - -/* char: 0xf6 */ - -static const GLubyte ch246data[] = { -0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, -0x7f,0x0,0x3e,0x0,0x0,0x0,0x36,0x0,0x36,0x0, -}; - -static const BitmapCharRec ch246 = {9,13,-1,0,11,ch246data}; - -/* char: 0xf5 */ - -static const GLubyte ch245data[] = { -0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, -0x7f,0x0,0x3e,0x0,0x0,0x0,0x26,0x0,0x2d,0x0,0x19,0x0, -}; - -static const BitmapCharRec ch245 = {9,14,-1,0,11,ch245data}; - -/* char: 0xf4 */ - -static const GLubyte ch244data[] = { -0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, -0x7f,0x0,0x3e,0x0,0x0,0x0,0x33,0x0,0x1e,0x0,0xc,0x0, -}; - -static const BitmapCharRec ch244 = {9,14,-1,0,11,ch244data}; - -/* char: 0xf3 */ - -static const GLubyte ch243data[] = { -0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, -0x7f,0x0,0x3e,0x0,0x0,0x0,0x18,0x0,0xc,0x0,0x6,0x0, -}; - -static const BitmapCharRec ch243 = {9,14,-1,0,11,ch243data}; - -/* char: 0xf2 */ - -static const GLubyte ch242data[] = { -0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, -0x7f,0x0,0x3e,0x0,0x0,0x0,0xc,0x0,0x18,0x0,0x30,0x0, -}; - -static const BitmapCharRec ch242 = {9,14,-1,0,11,ch242data}; - -/* char: 0xf1 */ - -static const GLubyte ch241data[] = { -0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xe3,0xdf,0xce,0x0,0x4c,0x5a,0x32, -}; - -static const BitmapCharRec ch241 = {8,14,-1,0,10,ch241data}; - -/* char: 0xf0 */ - -static const GLubyte ch240data[] = { -0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, -0x7f,0x0,0x3e,0x0,0x4c,0x0,0x38,0x0,0x36,0x0,0x60,0x0, -}; - -static const BitmapCharRec ch240 = {9,14,-1,0,11,ch240data}; - -/* char: 0xef */ - -static const GLubyte ch239data[] = { -0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x0,0xd8,0xd8, -}; - -static const BitmapCharRec ch239 = {5,13,0,0,4,ch239data}; - -/* char: 0xee */ - -static const GLubyte ch238data[] = { -0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0xcc,0x78,0x30, -}; - -static const BitmapCharRec ch238 = {6,14,1,0,4,ch238data}; - -/* char: 0xed */ - -static const GLubyte ch237data[] = { -0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x0,0xc0,0x60,0x30, -}; - -static const BitmapCharRec ch237 = {4,14,0,0,4,ch237data}; - -/* char: 0xec */ - -static const GLubyte ch236data[] = { -0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x0,0x30,0x60,0xc0, -}; - -static const BitmapCharRec ch236 = {4,14,0,0,4,ch236data}; - -/* char: 0xeb */ - -static const GLubyte ch235data[] = { -0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c,0x0,0x36,0x36, -}; - -static const BitmapCharRec ch235 = {8,13,-1,0,10,ch235data}; - -/* char: 0xea */ - -static const GLubyte ch234data[] = { -0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c,0x0,0x66,0x3c,0x18, -}; - -static const BitmapCharRec ch234 = {8,14,-1,0,10,ch234data}; - -/* char: 0xe9 */ - -static const GLubyte ch233data[] = { -0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c,0x0,0x18,0xc,0x6, -}; - -static const BitmapCharRec ch233 = {8,14,-1,0,10,ch233data}; - -/* char: 0xe8 */ - -static const GLubyte ch232data[] = { -0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c,0x0,0x18,0x30,0x60, -}; - -static const BitmapCharRec ch232 = {8,14,-1,0,10,ch232data}; - -/* char: 0xe7 */ - -static const GLubyte ch231data[] = { -0x78,0x6c,0xc,0x38,0x3e,0x7f,0x63,0xc0,0xc0,0xc0,0xc0,0x63,0x7f,0x3e, -}; - -static const BitmapCharRec ch231 = {8,14,-1,4,10,ch231data}; - -/* char: 0xe6 */ - -static const GLubyte ch230data[] = { -0x75,0xe0,0xef,0xf8,0xc7,0x18,0xc6,0x0,0xe6,0x0,0x7f,0xf8,0xe,0x18,0xc6,0x18, -0xef,0xf0,0x7d,0xe0, -}; - -static const BitmapCharRec ch230 = {13,10,-1,0,15,ch230data}; - -/* char: 0xe5 */ - -static const GLubyte ch229data[] = { -0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x38,0x6c,0x6c,0x38, -}; - -static const BitmapCharRec ch229 = {7,14,-1,0,9,ch229data}; - -/* char: 0xe4 */ - -static const GLubyte ch228data[] = { -0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x6c,0x6c, -}; - -static const BitmapCharRec ch228 = {7,13,-1,0,9,ch228data}; - -/* char: 0xe3 */ - -static const GLubyte ch227data[] = { -0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x4c,0x5a,0x32, -}; - -static const BitmapCharRec ch227 = {7,14,-1,0,9,ch227data}; - -/* char: 0xe2 */ - -static const GLubyte ch226data[] = { -0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x66,0x3c,0x18, -}; - -static const BitmapCharRec ch226 = {7,14,-1,0,9,ch226data}; - -/* char: 0xe1 */ - -static const GLubyte ch225data[] = { -0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x30,0x18,0xc, -}; - -static const BitmapCharRec ch225 = {7,14,-1,0,9,ch225data}; - -/* char: 0xe0 */ - -static const GLubyte ch224data[] = { -0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x18,0x30,0x60, -}; - -static const BitmapCharRec ch224 = {7,14,-1,0,9,ch224data}; - -/* char: 0xdf */ - -static const GLubyte ch223data[] = { -0xdc,0xde,0xc6,0xc6,0xc6,0xc6,0xdc,0xdc,0xc6,0xc6,0xc6,0xc6,0x7c,0x38, -}; - -static const BitmapCharRec ch223 = {7,14,-1,0,9,ch223data}; - -/* char: 0xde */ - -static const GLubyte ch222data[] = { -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x80,0xc1,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc1,0xc0,0xff,0x80,0xff,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, -}; - -static const BitmapCharRec ch222 = {10,14,-1,0,12,ch222data}; - -/* char: 0xdd */ - -static const GLubyte ch221data[] = { -0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0xf,0x0,0x19,0x80, -0x30,0xc0,0x30,0xc0,0x60,0x60,0x60,0x60,0xc0,0x30,0xc0,0x30,0x0,0x0,0x6,0x0, -0x3,0x0,0x1,0x80, -}; - -static const BitmapCharRec ch221 = {12,18,-1,0,14,ch221data}; - -/* char: 0xdc */ - -static const GLubyte ch220data[] = { -0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, -0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0x0,0x0,0x19,0x80, -0x19,0x80, -}; - -static const BitmapCharRec ch220 = {11,17,-1,0,13,ch220data}; - -/* char: 0xdb */ - -static const GLubyte ch219data[] = { -0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, -0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0x0,0x0,0x19,0x80, -0xf,0x0,0x6,0x0, -}; - -static const BitmapCharRec ch219 = {11,18,-1,0,13,ch219data}; - -/* char: 0xda */ - -static const GLubyte ch218data[] = { -0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, -0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0x0,0x0,0xc,0x0, -0x6,0x0,0x3,0x0, -}; - -static const BitmapCharRec ch218 = {11,18,-1,0,13,ch218data}; - -/* char: 0xd9 */ - -static const GLubyte ch217data[] = { -0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, -0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0x0,0x0,0x6,0x0, -0xc,0x0,0x18,0x0, -}; - -static const BitmapCharRec ch217 = {11,18,-1,0,13,ch217data}; - -/* char: 0xd8 */ - -static const GLubyte ch216data[] = { -0xc7,0xc0,0xff,0xf0,0x78,0x38,0x38,0x18,0x6c,0x1c,0x6e,0xc,0x67,0xc,0x63,0x8c, -0x61,0xcc,0x70,0xdc,0x30,0x78,0x38,0x38,0x1f,0xfc,0x7,0xcc, -}; - -static const BitmapCharRec ch216 = {14,14,0,0,15,ch216data}; - -/* char: 0xd7 */ - -static const GLubyte ch215data[] = { -0xc0,0xc0,0x61,0x80,0x33,0x0,0x1e,0x0,0xc,0x0,0x1e,0x0,0x33,0x0,0x61,0x80, -0xc0,0xc0, -}; - -static const BitmapCharRec ch215 = {10,9,0,0,10,ch215data}; - -/* char: 0xd6 */ - -static const GLubyte ch214data[] = { -0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, -0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0xd,0x80, -0xd,0x80, -}; - -static const BitmapCharRec ch214 = {13,17,-1,0,15,ch214data}; - -/* char: 0xd5 */ - -static const GLubyte ch213data[] = { -0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, -0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0x9,0x80, -0xb,0x40,0x6,0x40, -}; - -static const BitmapCharRec ch213 = {13,18,-1,0,15,ch213data}; - -/* char: 0xd4 */ - -static const GLubyte ch212data[] = { -0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, -0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0xc,0xc0, -0x7,0x80,0x3,0x0, -}; - -static const BitmapCharRec ch212 = {13,18,-1,0,15,ch212data}; - -/* char: 0xd3 */ - -static const GLubyte ch211data[] = { -0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, -0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0x3,0x0, -0x1,0x80,0x0,0xc0, -}; - -static const BitmapCharRec ch211 = {13,18,-1,0,15,ch211data}; - -/* char: 0xd2 */ - -static const GLubyte ch210data[] = { -0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, -0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0x3,0x0, -0x6,0x0,0xc,0x0, -}; - -static const BitmapCharRec ch210 = {13,18,-1,0,15,ch210data}; - -/* char: 0xd1 */ - -static const GLubyte ch209data[] = { -0xc0,0x60,0xc0,0xe0,0xc1,0xe0,0xc1,0xe0,0xc3,0x60,0xc6,0x60,0xc6,0x60,0xcc,0x60, -0xcc,0x60,0xd8,0x60,0xd8,0x60,0xf0,0x60,0xe0,0x60,0xe0,0x60,0x0,0x0,0x13,0x0, -0x16,0x80,0xc,0x80, -}; - -static const BitmapCharRec ch209 = {11,18,-1,0,13,ch209data}; - -/* char: 0xd0 */ - -static const GLubyte ch208data[] = { -0x7f,0x80,0x7f,0xc0,0x60,0xe0,0x60,0x60,0x60,0x30,0x60,0x30,0xfc,0x30,0xfc,0x30, -0x60,0x30,0x60,0x30,0x60,0x60,0x60,0xe0,0x7f,0xc0,0x7f,0x80, -}; - -static const BitmapCharRec ch208 = {12,14,0,0,13,ch208data}; - -/* char: 0xcf */ - -static const GLubyte ch207data[] = { -0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0xcc, -0xcc, -}; - -static const BitmapCharRec ch207 = {6,17,0,0,6,ch207data}; - -/* char: 0xce */ - -static const GLubyte ch206data[] = { -0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0xcc, -0x78,0x30, -}; - -static const BitmapCharRec ch206 = {6,18,0,0,6,ch206data}; - -/* char: 0xcd */ - -static const GLubyte ch205data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0xc0, -0x60,0x30, -}; - -static const BitmapCharRec ch205 = {4,18,-2,0,6,ch205data}; - -/* char: 0xcc */ - -static const GLubyte ch204data[] = { -0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0x30, -0x60,0xc0, -}; - -static const BitmapCharRec ch204 = {4,18,0,0,6,ch204data}; - -/* char: 0xcb */ - -static const GLubyte ch203data[] = { -0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80,0x0,0x0,0x33,0x0, -0x33,0x0, -}; - -static const BitmapCharRec ch203 = {9,17,-1,0,11,ch203data}; - -/* char: 0xca */ - -static const GLubyte ch202data[] = { -0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80,0x0,0x0,0x33,0x0, -0x1e,0x0,0xc,0x0, -}; - -static const BitmapCharRec ch202 = {9,18,-1,0,11,ch202data}; - -/* char: 0xc9 */ - -static const GLubyte ch201data[] = { -0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80,0x0,0x0,0xc,0x0, -0x6,0x0,0x3,0x0, -}; - -static const BitmapCharRec ch201 = {9,18,-1,0,11,ch201data}; - -/* char: 0xc8 */ - -static const GLubyte ch200data[] = { -0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80,0x0,0x0,0xc,0x0, -0x18,0x0,0x30,0x0, -}; - -static const BitmapCharRec ch200 = {9,18,-1,0,11,ch200data}; - -/* char: 0xc7 */ - -static const GLubyte ch199data[] = { -0x1e,0x0,0x1b,0x0,0x3,0x0,0xe,0x0,0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30, -0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xe0,0x0,0x60,0x30,0x70,0x70, -0x3f,0xe0,0xf,0x80, -}; - -static const BitmapCharRec ch199 = {12,18,-1,4,14,ch199data}; - -/* char: 0xc6 */ - -static const GLubyte ch198data[] = { -0xc1,0xff,0xc1,0xff,0x61,0x80,0x61,0x80,0x7f,0x80,0x3f,0x80,0x31,0xfe,0x31,0xfe, -0x19,0x80,0x19,0x80,0xd,0x80,0xd,0x80,0x7,0xff,0x7,0xff, -}; - -static const BitmapCharRec ch198 = {16,14,-1,0,18,ch198data}; - -/* char: 0xc5 */ - -static const GLubyte ch197data[] = { -0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, -0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0xf,0x0,0x19,0x80, -0x19,0x80,0xf,0x0, -}; - -static const BitmapCharRec ch197 = {12,18,0,0,12,ch197data}; - -/* char: 0xc4 */ - -static const GLubyte ch196data[] = { -0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, -0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x19,0x80, -0x19,0x80, -}; - -static const BitmapCharRec ch196 = {12,17,0,0,12,ch196data}; - -/* char: 0xc3 */ - -static const GLubyte ch195data[] = { -0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, -0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x13,0x0, -0x16,0x80,0xc,0x80, -}; - -static const BitmapCharRec ch195 = {12,18,0,0,12,ch195data}; - -/* char: 0xc2 */ - -static const GLubyte ch194data[] = { -0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, -0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x19,0x80, -0xf,0x0,0x6,0x0, -}; - -static const BitmapCharRec ch194 = {12,18,0,0,12,ch194data}; - -/* char: 0xc1 */ - -static const GLubyte ch193data[] = { -0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, -0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x6,0x0, -0x3,0x0,0x1,0x80, -}; - -static const BitmapCharRec ch193 = {12,18,0,0,12,ch193data}; - -/* char: 0xc0 */ - -static const GLubyte ch192data[] = { -0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, -0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x6,0x0, -0xc,0x0,0x18,0x0, -}; - -static const BitmapCharRec ch192 = {12,18,0,0,12,ch192data}; - -/* char: 0xbf */ - -static const GLubyte ch191data[] = { -0x7c,0xfe,0xc6,0xc6,0xe0,0x70,0x38,0x18,0x18,0x18,0x0,0x0,0x18,0x18, -}; - -static const BitmapCharRec ch191 = {7,14,-1,4,10,ch191data}; - -/* char: 0xbe */ - -static const GLubyte ch190data[] = { -0x18,0x18,0x18,0x18,0xc,0xfc,0x6,0xd8,0x6,0x78,0x73,0x38,0xf9,0x18,0x99,0x88, -0x30,0xc0,0x30,0xc0,0x98,0x60,0xf8,0x30,0x70,0x30, -}; - -static const BitmapCharRec ch190 = {14,13,0,0,15,ch190data}; - -/* char: 0xbd */ - -static const GLubyte ch189data[] = { -0x30,0xf8,0x30,0xf8,0x18,0x60,0xc,0x30,0xc,0x18,0x66,0x98,0x62,0xf8,0x63,0x70, -0x61,0x80,0x61,0x80,0xe0,0xc0,0xe0,0x60,0x60,0x60, -}; - -static const BitmapCharRec ch189 = {13,13,-1,0,15,ch189data}; - -/* char: 0xbc */ - -static const GLubyte ch188data[] = { -0x30,0x30,0x30,0x30,0x19,0xf8,0xd,0xb0,0xc,0xf0,0x66,0x70,0x62,0x30,0x63,0x10, -0x61,0x80,0x61,0x80,0xe0,0xc0,0xe0,0x60,0x60,0x60, -}; - -static const BitmapCharRec ch188 = {13,13,-1,0,15,ch188data}; - -/* char: 0xbb */ - -static const GLubyte ch187data[] = { -0x90,0xd8,0x6c,0x36,0x36,0x6c,0xd8,0x90, -}; - -static const BitmapCharRec ch187 = {7,8,-1,-1,9,ch187data}; - -/* char: 0xba */ - -static const GLubyte ch186data[] = { -0xf8,0x0,0x70,0xd8,0x88,0x88,0xd8,0x70, -}; - -static const BitmapCharRec ch186 = {5,8,-1,-6,7,ch186data}; - -/* char: 0xb9 */ - -static const GLubyte ch185data[] = { -0x60,0x60,0x60,0x60,0x60,0xe0,0xe0,0x60, -}; - -static const BitmapCharRec ch185 = {3,8,-1,-5,6,ch185data}; - -/* char: 0xb8 */ - -static const GLubyte ch184data[] = { -0xf0,0xd8,0x18,0x70,0x60, -}; - -static const BitmapCharRec ch184 = {5,5,0,4,5,ch184data}; - -/* char: 0xb7 */ - -static const GLubyte ch183data[] = { -0xc0,0xc0, -}; - -static const BitmapCharRec ch183 = {2,2,-1,-4,4,ch183data}; - -/* char: 0xb6 */ - -static const GLubyte ch182data[] = { -0x12,0x12,0x12,0x12,0x12,0x12,0x12,0x12,0x12,0x12,0x32,0x72,0xf2,0xf2,0xf2,0xf2, -0x72,0x3f, -}; - -static const BitmapCharRec ch182 = {8,18,-1,4,10,ch182data}; - -/* char: 0xb5 */ - -static const GLubyte ch181data[] = { -0xc0,0xc0,0xc0,0xc0,0xdb,0xff,0xe7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3, -}; - -static const BitmapCharRec ch181 = {8,14,-1,4,10,ch181data}; - -/* char: 0xb4 */ - -static const GLubyte ch180data[] = { -0xc0,0x60,0x30, -}; - -static const BitmapCharRec ch180 = {4,3,0,-11,4,ch180data}; - -/* char: 0xb3 */ - -static const GLubyte ch179data[] = { -0x70,0xf8,0x98,0x30,0x30,0x98,0xf8,0x70, -}; - -static const BitmapCharRec ch179 = {5,8,0,-5,6,ch179data}; - -/* char: 0xb2 */ - -static const GLubyte ch178data[] = { -0xf8,0xf8,0x60,0x30,0x18,0x98,0xf8,0x70, -}; - -static const BitmapCharRec ch178 = {5,8,0,-5,6,ch178data}; - -/* char: 0xb1 */ - -static const GLubyte ch177data[] = { -0xff,0xff,0x0,0x18,0x18,0x18,0xff,0xff,0x18,0x18,0x18, -}; - -static const BitmapCharRec ch177 = {8,11,-1,0,10,ch177data}; - -/* char: 0xb0 */ - -static const GLubyte ch176data[] = { -0x70,0xd8,0x88,0xd8,0x70, -}; - -static const BitmapCharRec ch176 = {5,5,-1,-8,7,ch176data}; - -/* char: 0xaf */ - -static const GLubyte ch175data[] = { -0xf8, -}; - -static const BitmapCharRec ch175 = {5,1,0,-12,5,ch175data}; - -/* char: 0xae */ - -static const GLubyte ch174data[] = { -0xf,0x80,0x30,0x60,0x40,0x10,0x48,0x50,0x88,0x88,0x89,0x8,0x8f,0x88,0x88,0x48, -0x88,0x48,0x4f,0x90,0x40,0x10,0x30,0x60,0xf,0x80, -}; - -static const BitmapCharRec ch174 = {13,13,-1,0,14,ch174data}; - -/* char: 0xad */ - -static const GLubyte ch173data[] = { -0xf8,0xf8, -}; - -static const BitmapCharRec ch173 = {5,2,-1,-4,7,ch173data}; - -/* char: 0xac */ - -static const GLubyte ch172data[] = { -0x1,0x80,0x1,0x80,0x1,0x80,0xff,0x80,0xff,0x80, -}; - -static const BitmapCharRec ch172 = {9,5,-1,-3,11,ch172data}; - -/* char: 0xab */ - -static const GLubyte ch171data[] = { -0x12,0x36,0x6c,0xd8,0xd8,0x6c,0x36,0x12, -}; - -static const BitmapCharRec ch171 = {7,8,-1,-1,9,ch171data}; - -/* char: 0xaa */ - -static const GLubyte ch170data[] = { -0xf8,0x0,0x68,0xd8,0x48,0x38,0xc8,0x70, -}; - -static const BitmapCharRec ch170 = {5,8,-1,-6,7,ch170data}; - -/* char: 0xa9 */ - -static const GLubyte ch169data[] = { -0xf,0x80,0x30,0x60,0x40,0x10,0x47,0x10,0x88,0x88,0x90,0x8,0x90,0x8,0x90,0x8, -0x88,0x88,0x47,0x10,0x40,0x10,0x30,0x60,0xf,0x80, -}; - -static const BitmapCharRec ch169 = {13,13,-1,0,15,ch169data}; - -/* char: 0xa8 */ - -static const GLubyte ch168data[] = { -0xd8,0xd8, -}; - -static const BitmapCharRec ch168 = {5,2,0,-11,6,ch168data}; - -/* char: 0xa7 */ - -static const GLubyte ch167data[] = { -0x3c,0x7e,0xc3,0xc3,0x7,0xe,0x3e,0x73,0xe3,0xc3,0xc7,0x6e,0x7c,0xf0,0xc3,0xc3, -0x7e,0x3c, -}; - -static const BitmapCharRec ch167 = {8,18,-1,4,10,ch167data}; - -/* char: 0xa6 */ - -static const GLubyte ch166data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0x0,0x0,0x0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0, -}; - -static const BitmapCharRec ch166 = {2,17,-1,3,4,ch166data}; - -/* char: 0xa5 */ - -static const GLubyte ch165data[] = { -0x18,0x18,0x18,0x18,0xff,0x18,0xff,0x3c,0x66,0x66,0x66,0xc3,0xc3, -}; - -static const BitmapCharRec ch165 = {8,13,-1,0,10,ch165data}; - -/* char: 0xa4 */ - -static const GLubyte ch164data[] = { -0xc3,0xff,0x66,0x66,0x66,0xff,0xc3, -}; - -static const BitmapCharRec ch164 = {8,7,-1,-3,10,ch164data}; - -/* char: 0xa3 */ - -static const GLubyte ch163data[] = { -0xdf,0x0,0xff,0x80,0x60,0x80,0x30,0x0,0x18,0x0,0x18,0x0,0x7e,0x0,0x30,0x0, -0x60,0x0,0x61,0x80,0x61,0x80,0x3f,0x0,0x1e,0x0, -}; - -static const BitmapCharRec ch163 = {9,13,0,0,10,ch163data}; - -/* char: 0xa2 */ - -static const GLubyte ch162data[] = { -0x10,0x10,0x3e,0x7f,0x6b,0xc8,0xc8,0xc8,0xc8,0x6b,0x7f,0x3e,0x4,0x4, -}; - -static const BitmapCharRec ch162 = {8,14,-1,2,10,ch162data}; - -/* char: 0xa1 */ - -static const GLubyte ch161data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x40,0x40,0x0,0x0,0xc0,0xc0, -}; - -static const BitmapCharRec ch161 = {2,14,-2,4,6,ch161data}; - -/* char: 0xa0 */ - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch160data[] = { 0x0 }; -static const BitmapCharRec ch160 = {1,1,0,0,5,ch160data}; -#else -static const BitmapCharRec ch160 = {0,0,0,0,5,0}; -#endif - -/* char: 0x7e '~' */ - -static const GLubyte ch126data[] = { -0xcc,0x7e,0x33, -}; - -static const BitmapCharRec ch126 = {8,3,-1,-4,10,ch126data}; - -/* char: 0x7d '}' */ - -static const GLubyte ch125data[] = { -0xc0,0x60,0x30,0x30,0x30,0x30,0x30,0x30,0x18,0xc,0x18,0x30,0x30,0x30,0x30,0x30, -0x60,0xc0, -}; - -static const BitmapCharRec ch125 = {6,18,0,4,6,ch125data}; - -/* char: 0x7c '|' */ - -static const GLubyte ch124data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0, -}; - -static const BitmapCharRec ch124 = {2,18,-1,4,4,ch124data}; - -/* char: 0x7b '{' */ - -static const GLubyte ch123data[] = { -0xc,0x18,0x30,0x30,0x30,0x30,0x30,0x30,0x60,0xc0,0x60,0x30,0x30,0x30,0x30,0x30, -0x18,0xc, -}; - -static const BitmapCharRec ch123 = {6,18,0,4,6,ch123data}; - -/* char: 0x7a 'z' */ - -static const GLubyte ch122data[] = { -0xfe,0xfe,0xc0,0x60,0x30,0x18,0xc,0x6,0xfe,0xfe, -}; - -static const BitmapCharRec ch122 = {7,10,-1,0,9,ch122data}; - -/* char: 0x79 'y' */ - -static const GLubyte ch121data[] = { -0x70,0x70,0x18,0x18,0x18,0x18,0x3c,0x24,0x66,0x66,0x66,0xc3,0xc3,0xc3, -}; - -static const BitmapCharRec ch121 = {8,14,-1,4,10,ch121data}; - -/* char: 0x78 'x' */ - -static const GLubyte ch120data[] = { -0xc3,0xe7,0x66,0x3c,0x18,0x18,0x3c,0x66,0xe7,0xc3, -}; - -static const BitmapCharRec ch120 = {8,10,-1,0,10,ch120data}; - -/* char: 0x77 'w' */ - -static const GLubyte ch119data[] = { -0x19,0x80,0x19,0x80,0x39,0xc0,0x29,0x40,0x69,0x60,0x66,0x60,0x66,0x60,0xc6,0x30, -0xc6,0x30,0xc6,0x30, -}; - -static const BitmapCharRec ch119 = {12,10,-1,0,14,ch119data}; - -/* char: 0x76 'v' */ - -static const GLubyte ch118data[] = { -0x18,0x18,0x3c,0x24,0x66,0x66,0x66,0xc3,0xc3,0xc3, -}; - -static const BitmapCharRec ch118 = {8,10,-1,0,10,ch118data}; - -/* char: 0x75 'u' */ - -static const GLubyte ch117data[] = { -0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3, -}; - -static const BitmapCharRec ch117 = {8,10,-1,0,10,ch117data}; - -/* char: 0x74 't' */ - -static const GLubyte ch116data[] = { -0x18,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0xfc,0xfc,0x30,0x30,0x30, -}; - -static const BitmapCharRec ch116 = {6,13,0,0,6,ch116data}; - -/* char: 0x73 's' */ - -static const GLubyte ch115data[] = { -0x78,0xfc,0xc6,0x6,0x3e,0xfc,0xc0,0xc6,0x7e,0x3c, -}; - -static const BitmapCharRec ch115 = {7,10,-1,0,9,ch115data}; - -/* char: 0x72 'r' */ - -static const GLubyte ch114data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xe0,0xd8,0xd8, -}; - -static const BitmapCharRec ch114 = {5,10,-1,0,6,ch114data}; - -/* char: 0x71 'q' */ - -static const GLubyte ch113data[] = { -0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x3d,0x80,0x7f,0x80,0x63,0x80,0xc1,0x80, -0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x80,0x7f,0x80,0x3d,0x80, -}; - -static const BitmapCharRec ch113 = {9,14,-1,4,11,ch113data}; - -/* char: 0x70 'p' */ - -static const GLubyte ch112data[] = { -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xde,0x0,0xff,0x0,0xe3,0x0,0xc1,0x80, -0xc1,0x80,0xc1,0x80,0xc1,0x80,0xe3,0x0,0xff,0x0,0xde,0x0, -}; - -static const BitmapCharRec ch112 = {9,14,-1,4,11,ch112data}; - -/* char: 0x6f 'o' */ - -static const GLubyte ch111data[] = { -0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, -0x7f,0x0,0x3e,0x0, -}; - -static const BitmapCharRec ch111 = {9,10,-1,0,11,ch111data}; - -/* char: 0x6e 'n' */ - -static const GLubyte ch110data[] = { -0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xe3,0xdf,0xce, -}; - -static const BitmapCharRec ch110 = {8,10,-1,0,10,ch110data}; - -/* char: 0x6d 'm' */ - -static const GLubyte ch109data[] = { -0xc6,0x30,0xc6,0x30,0xc6,0x30,0xc6,0x30,0xc6,0x30,0xc6,0x30,0xc6,0x30,0xe7,0x30, -0xde,0xf0,0xcc,0x60, -}; - -static const BitmapCharRec ch109 = {12,10,-1,0,14,ch109data}; - -/* char: 0x6c 'l' */ - -static const GLubyte ch108data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -}; - -static const BitmapCharRec ch108 = {2,14,-1,0,4,ch108data}; - -/* char: 0x6b 'k' */ - -static const GLubyte ch107data[] = { -0xc7,0xc6,0xce,0xcc,0xd8,0xf8,0xf0,0xd8,0xcc,0xc6,0xc0,0xc0,0xc0,0xc0, -}; - -static const BitmapCharRec ch107 = {8,14,-1,0,9,ch107data}; - -/* char: 0x6a 'j' */ - -static const GLubyte ch106data[] = { -0xe0,0xf0,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0x0, -0x30,0x30, -}; - -static const BitmapCharRec ch106 = {4,18,1,4,4,ch106data}; - -/* char: 0x69 'i' */ - -static const GLubyte ch105data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0x0,0xc0,0xc0, -}; - -static const BitmapCharRec ch105 = {2,14,-1,0,4,ch105data}; - -/* char: 0x68 'h' */ - -static const GLubyte ch104data[] = { -0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xe3,0xdf,0xce,0xc0,0xc0,0xc0,0xc0, -}; - -static const BitmapCharRec ch104 = {8,14,-1,0,10,ch104data}; - -/* char: 0x67 'g' */ - -static const GLubyte ch103data[] = { -0x1c,0x0,0x7f,0x0,0x63,0x0,0x1,0x80,0x3d,0x80,0x7f,0x80,0x63,0x80,0xc1,0x80, -0xc1,0x80,0xc1,0x80,0xc1,0x80,0x61,0x80,0x7f,0x80,0x3d,0x80, -}; - -static const BitmapCharRec ch103 = {9,14,-1,4,11,ch103data}; - -/* char: 0x66 'f' */ - -static const GLubyte ch102data[] = { -0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0xfc,0xfc,0x30,0x30,0x3c,0x1c, -}; - -static const BitmapCharRec ch102 = {6,14,0,0,6,ch102data}; - -/* char: 0x65 'e' */ - -static const GLubyte ch101data[] = { -0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c, -}; - -static const BitmapCharRec ch101 = {8,10,-1,0,10,ch101data}; - -/* char: 0x64 'd' */ - -static const GLubyte ch100data[] = { -0x3d,0x80,0x7f,0x80,0x63,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x80, -0x7f,0x80,0x3d,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80, -}; - -static const BitmapCharRec ch100 = {9,14,-1,0,11,ch100data}; - -/* char: 0x63 'c' */ - -static const GLubyte ch99data[] = { -0x3e,0x7f,0x63,0xc0,0xc0,0xc0,0xc0,0x63,0x7f,0x3e, -}; - -static const BitmapCharRec ch99 = {8,10,-1,0,10,ch99data}; - -/* char: 0x62 'b' */ - -static const GLubyte ch98data[] = { -0xde,0x0,0xff,0x0,0xe3,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xe3,0x0, -0xff,0x0,0xde,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, -}; - -static const BitmapCharRec ch98 = {9,14,-1,0,11,ch98data}; - -/* char: 0x61 'a' */ - -static const GLubyte ch97data[] = { -0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c, -}; - -static const BitmapCharRec ch97 = {7,10,-1,0,9,ch97data}; - -/* char: 0x60 '`' */ - -static const GLubyte ch96data[] = { -0xc0,0xc0,0x80,0x80,0x40, -}; - -static const BitmapCharRec ch96 = {2,5,-1,-9,4,ch96data}; - -/* char: 0x5f '_' */ - -static const GLubyte ch95data[] = { -0xff,0xc0,0xff,0xc0, -}; - -static const BitmapCharRec ch95 = {10,2,0,4,10,ch95data}; - -/* char: 0x5e '^' */ - -static const GLubyte ch94data[] = { -0x82,0xc6,0x6c,0x38,0x10, -}; - -static const BitmapCharRec ch94 = {7,5,-1,-8,9,ch94data}; - -/* char: 0x5d ']' */ - -static const GLubyte ch93data[] = { -0xf0,0xf0,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, -0xf0,0xf0, -}; - -static const BitmapCharRec ch93 = {4,18,0,4,5,ch93data}; - -/* char: 0x5c '\' */ - -static const GLubyte ch92data[] = { -0x18,0x18,0x10,0x10,0x30,0x30,0x20,0x20,0x60,0x60,0x40,0x40,0xc0,0xc0, -}; - -static const BitmapCharRec ch92 = {5,14,0,0,5,ch92data}; - -/* char: 0x5b '[' */ - -static const GLubyte ch91data[] = { -0xf0,0xf0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xf0,0xf0, -}; - -static const BitmapCharRec ch91 = {4,18,-1,4,5,ch91data}; - -/* char: 0x5a 'Z' */ - -static const GLubyte ch90data[] = { -0xff,0xc0,0xff,0xc0,0xc0,0x0,0x60,0x0,0x30,0x0,0x18,0x0,0x1c,0x0,0xc,0x0, -0x6,0x0,0x3,0x0,0x1,0x80,0x0,0xc0,0xff,0xc0,0xff,0xc0, -}; - -static const BitmapCharRec ch90 = {10,14,-1,0,12,ch90data}; - -/* char: 0x59 'Y' */ - -static const GLubyte ch89data[] = { -0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0xf,0x0,0x19,0x80, -0x30,0xc0,0x30,0xc0,0x60,0x60,0x60,0x60,0xc0,0x30,0xc0,0x30, -}; - -static const BitmapCharRec ch89 = {12,14,-1,0,14,ch89data}; - -/* char: 0x58 'X' */ - -static const GLubyte ch88data[] = { -0xc0,0x60,0xe0,0xe0,0x60,0xc0,0x71,0xc0,0x31,0x80,0x1b,0x0,0xe,0x0,0xe,0x0, -0x1b,0x0,0x31,0x80,0x71,0xc0,0x60,0xc0,0xe0,0xe0,0xc0,0x60, -}; - -static const BitmapCharRec ch88 = {11,14,-1,0,13,ch88data}; - -/* char: 0x57 'W' */ - -static const GLubyte ch87data[] = { -0x18,0x18,0x18,0x18,0x1c,0x38,0x34,0x2c,0x36,0x6c,0x36,0x6c,0x66,0x66,0x66,0x66, -0x62,0x46,0x63,0xc6,0xc3,0xc3,0xc1,0x83,0xc1,0x83,0xc1,0x83, -}; - -static const BitmapCharRec ch87 = {16,14,-1,0,18,ch87data}; - -/* char: 0x56 'V' */ - -static const GLubyte ch86data[] = { -0x6,0x0,0xf,0x0,0xf,0x0,0x19,0x80,0x19,0x80,0x19,0x80,0x30,0xc0,0x30,0xc0, -0x30,0xc0,0x60,0x60,0x60,0x60,0x60,0x60,0xc0,0x30,0xc0,0x30, -}; - -static const BitmapCharRec ch86 = {12,14,-1,0,14,ch86data}; - -/* char: 0x55 'U' */ - -static const GLubyte ch85data[] = { -0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, -0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, -}; - -static const BitmapCharRec ch85 = {11,14,-1,0,13,ch85data}; - -/* char: 0x54 'T' */ - -static const GLubyte ch84data[] = { -0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0, -0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xff,0xc0,0xff,0xc0, -}; - -static const BitmapCharRec ch84 = {10,14,-1,0,12,ch84data}; - -/* char: 0x53 'S' */ - -static const GLubyte ch83data[] = { -0x3f,0x0,0x7f,0xc0,0xe0,0xe0,0xc0,0x60,0x0,0x60,0x0,0xe0,0x3,0xc0,0x1f,0x0, -0x7c,0x0,0xe0,0x0,0xc0,0x60,0xe0,0xe0,0x7f,0xc0,0x1f,0x0, -}; - -static const BitmapCharRec ch83 = {11,14,-1,0,13,ch83data}; - -/* char: 0x52 'R' */ - -static const GLubyte ch82data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc1,0x80,0xc1,0x80,0xff,0x0,0xff,0x80, -0xc1,0xc0,0xc0,0xc0,0xc0,0xc0,0xc1,0xc0,0xff,0x80,0xff,0x0, -}; - -static const BitmapCharRec ch82 = {10,14,-1,0,12,ch82data}; - -/* char: 0x51 'Q' */ - -static const GLubyte ch81data[] = { -0x0,0x30,0xf,0xb0,0x3f,0xe0,0x70,0xf0,0x61,0xb0,0xe1,0xb8,0xc0,0x18,0xc0,0x18, -0xc0,0x18,0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80, -}; - -static const BitmapCharRec ch81 = {13,15,-1,1,15,ch81data}; - -/* char: 0x50 'P' */ - -static const GLubyte ch80data[] = { -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x80, -0xc1,0xc0,0xc0,0xc0,0xc0,0xc0,0xc1,0xc0,0xff,0x80,0xff,0x0, -}; - -static const BitmapCharRec ch80 = {10,14,-1,0,12,ch80data}; - -/* char: 0x4f 'O' */ - -static const GLubyte ch79data[] = { -0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, -0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80, -}; - -static const BitmapCharRec ch79 = {13,14,-1,0,15,ch79data}; - -/* char: 0x4e 'N' */ - -static const GLubyte ch78data[] = { -0xc0,0x60,0xc0,0xe0,0xc1,0xe0,0xc1,0xe0,0xc3,0x60,0xc6,0x60,0xc6,0x60,0xcc,0x60, -0xcc,0x60,0xd8,0x60,0xf0,0x60,0xf0,0x60,0xe0,0x60,0xc0,0x60, -}; - -static const BitmapCharRec ch78 = {11,14,-1,0,13,ch78data}; - -/* char: 0x4d 'M' */ - -static const GLubyte ch77data[] = { -0xc3,0xc,0xc3,0xc,0xc7,0x8c,0xc4,0x8c,0xcc,0xcc,0xcc,0xcc,0xd8,0x6c,0xd8,0x6c, -0xf0,0x3c,0xf0,0x3c,0xe0,0x1c,0xe0,0x1c,0xc0,0xc,0xc0,0xc, -}; - -static const BitmapCharRec ch77 = {14,14,-1,0,16,ch77data}; - -/* char: 0x4c 'L' */ - -static const GLubyte ch76data[] = { -0xff,0xff,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -}; - -static const BitmapCharRec ch76 = {8,14,-1,0,10,ch76data}; - -/* char: 0x4b 'K' */ - -static const GLubyte ch75data[] = { -0xc0,0x70,0xc0,0xe0,0xc1,0xc0,0xc3,0x80,0xc7,0x0,0xce,0x0,0xfc,0x0,0xf8,0x0, -0xdc,0x0,0xce,0x0,0xc7,0x0,0xc3,0x80,0xc1,0xc0,0xc0,0xe0, -}; - -static const BitmapCharRec ch75 = {12,14,-1,0,13,ch75data}; - -/* char: 0x4a 'J' */ - -static const GLubyte ch74data[] = { -0x3c,0x7e,0xe7,0xc3,0xc3,0x3,0x3,0x3,0x3,0x3,0x3,0x3,0x3,0x3, -}; - -static const BitmapCharRec ch74 = {8,14,-1,0,10,ch74data}; - -/* char: 0x49 'I' */ - -static const GLubyte ch73data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -}; - -static const BitmapCharRec ch73 = {2,14,-2,0,6,ch73data}; - -/* char: 0x48 'H' */ - -static const GLubyte ch72data[] = { -0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xff,0xe0,0xff,0xe0, -0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, -}; - -static const BitmapCharRec ch72 = {11,14,-1,0,13,ch72data}; - -/* char: 0x47 'G' */ - -static const GLubyte ch71data[] = { -0xf,0xb0,0x3f,0xf0,0x70,0x70,0x60,0x30,0xe0,0x30,0xc1,0xf0,0xc1,0xf0,0xc0,0x0, -0xc0,0x0,0xe0,0x30,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80, -}; - -static const BitmapCharRec ch71 = {12,14,-1,0,14,ch71data}; - -/* char: 0x46 'F' */ - -static const GLubyte ch70data[] = { -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80, -}; - -static const BitmapCharRec ch70 = {9,14,-1,0,11,ch70data}; - -/* char: 0x45 'E' */ - -static const GLubyte ch69data[] = { -0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80, -}; - -static const BitmapCharRec ch69 = {9,14,-1,0,11,ch69data}; - -/* char: 0x44 'D' */ - -static const GLubyte ch68data[] = { -0xff,0x0,0xff,0x80,0xc1,0xc0,0xc0,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, -0xc0,0x60,0xc0,0x60,0xc0,0xc0,0xc1,0xc0,0xff,0x80,0xff,0x0, -}; - -static const BitmapCharRec ch68 = {11,14,-1,0,13,ch68data}; - -/* char: 0x43 'C' */ - -static const GLubyte ch67data[] = { -0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, -0xc0,0x0,0xe0,0x0,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80, -}; - -static const BitmapCharRec ch67 = {12,14,-1,0,14,ch67data}; - -/* char: 0x42 'B' */ - -static const GLubyte ch66data[] = { -0xff,0x80,0xff,0xc0,0xc0,0xe0,0xc0,0x60,0xc0,0x60,0xc0,0xe0,0xff,0xc0,0xff,0x80, -0xc1,0x80,0xc0,0xc0,0xc0,0xc0,0xc1,0xc0,0xff,0x80,0xff,0x0, -}; - -static const BitmapCharRec ch66 = {11,14,-1,0,13,ch66data}; - -/* char: 0x41 'A' */ - -static const GLubyte ch65data[] = { -0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, -0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0, -}; - -static const BitmapCharRec ch65 = {12,14,0,0,12,ch65data}; - -/* char: 0x40 '@' */ - -static const GLubyte ch64data[] = { -0x7,0xe0,0x1f,0xf0,0x38,0x0,0x70,0x0,0x67,0x70,0xcf,0xf8,0xcc,0xcc,0xcc,0x66, -0xcc,0x66,0xcc,0x63,0xc6,0x33,0x67,0x73,0x63,0xb3,0x30,0x6,0x1c,0xe,0xf,0xfc, -0x3,0xf0, -}; - -static const BitmapCharRec ch64 = {16,17,-1,3,18,ch64data}; - -/* char: 0x3f '?' */ - -static const GLubyte ch63data[] = { -0x30,0x30,0x0,0x0,0x30,0x30,0x30,0x38,0x1c,0xe,0xc6,0xc6,0xfe,0x7c, -}; - -static const BitmapCharRec ch63 = {7,14,-1,0,10,ch63data}; - -/* char: 0x3e '>' */ - -static const GLubyte ch62data[] = { -0xc0,0xf0,0x3c,0xe,0x3,0xe,0x3c,0xf0,0xc0, -}; - -static const BitmapCharRec ch62 = {8,9,-1,0,10,ch62data}; - -/* char: 0x3d '=' */ - -static const GLubyte ch61data[] = { -0xfe,0xfe,0x0,0x0,0xfe,0xfe, -}; - -static const BitmapCharRec ch61 = {7,6,-2,-2,11,ch61data}; - -/* char: 0x3c '<' */ - -static const GLubyte ch60data[] = { -0x3,0xf,0x3c,0x70,0xc0,0x70,0x3c,0xf,0x3, -}; - -static const BitmapCharRec ch60 = {8,9,-1,0,10,ch60data}; - -/* char: 0x3b ';' */ - -static const GLubyte ch59data[] = { -0x80,0x40,0x40,0xc0,0xc0,0x0,0x0,0x0,0x0,0x0,0x0,0xc0,0xc0, -}; - -static const BitmapCharRec ch59 = {2,13,-1,3,5,ch59data}; - -/* char: 0x3a ':' */ - -static const GLubyte ch58data[] = { -0xc0,0xc0,0x0,0x0,0x0,0x0,0x0,0x0,0xc0,0xc0, -}; - -static const BitmapCharRec ch58 = {2,10,-1,0,5,ch58data}; - -/* char: 0x39 '9' */ - -static const GLubyte ch57data[] = { -0x7c,0xfe,0xc6,0x3,0x3,0x3b,0x7f,0xc3,0xc3,0xc3,0xc7,0x7e,0x3c, -}; - -static const BitmapCharRec ch57 = {8,13,-1,0,10,ch57data}; - -/* char: 0x38 '8' */ - -static const GLubyte ch56data[] = { -0x3c,0x7e,0xe7,0xc3,0xc3,0x66,0x7e,0x66,0xc3,0xc3,0xe7,0x7e,0x3c, -}; - -static const BitmapCharRec ch56 = {8,13,-1,0,10,ch56data}; - -/* char: 0x37 '7' */ - -static const GLubyte ch55data[] = { -0x60,0x60,0x30,0x30,0x30,0x18,0x18,0xc,0xc,0x6,0x3,0xff,0xff, -}; - -static const BitmapCharRec ch55 = {8,13,-1,0,10,ch55data}; - -/* char: 0x36 '6' */ - -static const GLubyte ch54data[] = { -0x3c,0x7e,0xe3,0xc3,0xc3,0xc3,0xfe,0xdc,0xc0,0xc0,0x63,0x7f,0x3c, -}; - -static const BitmapCharRec ch54 = {8,13,-1,0,10,ch54data}; - -/* char: 0x35 '5' */ - -static const GLubyte ch53data[] = { -0x7c,0xfe,0xc7,0xc3,0x3,0x3,0xc7,0xfe,0xfc,0xc0,0xc0,0xfe,0xfe, -}; - -static const BitmapCharRec ch53 = {8,13,-1,0,10,ch53data}; - -/* char: 0x34 '4' */ - -static const GLubyte ch52data[] = { -0x3,0x0,0x3,0x0,0x3,0x0,0xff,0x80,0xff,0x80,0xc3,0x0,0x63,0x0,0x33,0x0, -0x33,0x0,0x1b,0x0,0xf,0x0,0x7,0x0,0x3,0x0, -}; - -static const BitmapCharRec ch52 = {9,13,-1,0,10,ch52data}; - -/* char: 0x33 '3' */ - -static const GLubyte ch51data[] = { -0x3c,0x7e,0xc7,0xc3,0x3,0x7,0x1e,0x1c,0x6,0xc3,0xc3,0x7e,0x3c, -}; - -static const BitmapCharRec ch51 = {8,13,-1,0,10,ch51data}; - -/* char: 0x32 '2' */ - -static const GLubyte ch50data[] = { -0xff,0xff,0xc0,0xe0,0x70,0x38,0x1c,0xe,0x7,0x3,0xc3,0xfe,0x3c, -}; - -static const BitmapCharRec ch50 = {8,13,-1,0,10,ch50data}; - -/* char: 0x31 '1' */ - -static const GLubyte ch49data[] = { -0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0xf8,0xf8,0x18, -}; - -static const BitmapCharRec ch49 = {5,13,-2,0,10,ch49data}; - -/* char: 0x30 '0' */ - -static const GLubyte ch48data[] = { -0x3c,0x7e,0x66,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x66,0x7e,0x3c, -}; - -static const BitmapCharRec ch48 = {8,13,-1,0,10,ch48data}; - -/* char: 0x2f '/' */ - -static const GLubyte ch47data[] = { -0xc0,0xc0,0x40,0x40,0x60,0x60,0x20,0x20,0x30,0x30,0x10,0x10,0x18,0x18, -}; - -static const BitmapCharRec ch47 = {5,14,0,0,5,ch47data}; - -/* char: 0x2e '.' */ - -static const GLubyte ch46data[] = { -0xc0,0xc0, -}; - -static const BitmapCharRec ch46 = {2,2,-1,0,5,ch46data}; - -/* char: 0x2d '-' */ - -static const GLubyte ch45data[] = { -0xff,0xff, -}; - -static const BitmapCharRec ch45 = {8,2,-1,-4,11,ch45data}; - -/* char: 0x2c ',' */ - -static const GLubyte ch44data[] = { -0x80,0x40,0x40,0xc0,0xc0, -}; - -static const BitmapCharRec ch44 = {2,5,-1,3,5,ch44data}; - -/* char: 0x2b '+' */ - -static const GLubyte ch43data[] = { -0x18,0x18,0x18,0x18,0xff,0xff,0x18,0x18,0x18,0x18, -}; - -static const BitmapCharRec ch43 = {8,10,-1,0,10,ch43data}; - -/* char: 0x2a '*' */ - -static const GLubyte ch42data[] = { -0x88,0x70,0x70,0xf8,0x20,0x20, -}; - -static const BitmapCharRec ch42 = {5,6,-1,-8,7,ch42data}; - -/* char: 0x29 ')' */ - -static const GLubyte ch41data[] = { -0x80,0xc0,0x60,0x60,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x60,0x60, -0xc0,0x80, -}; - -static const BitmapCharRec ch41 = {4,18,-1,4,6,ch41data}; - -/* char: 0x28 '(' */ - -static const GLubyte ch40data[] = { -0x10,0x30,0x60,0x60,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x60,0x60, -0x30,0x10, -}; - -static const BitmapCharRec ch40 = {4,18,-1,4,6,ch40data}; - -/* char: 0x27 ''' */ - -static const GLubyte ch39data[] = { -0x80,0x40,0x40,0xc0,0xc0, -}; - -static const BitmapCharRec ch39 = {2,5,-1,-9,4,ch39data}; - -/* char: 0x26 '&' */ - -static const GLubyte ch38data[] = { -0x3c,0x70,0x7e,0xe0,0xe7,0xc0,0xc3,0x80,0xc3,0xc0,0xc6,0xc0,0xee,0xc0,0x7c,0x0, -0x3c,0x0,0x66,0x0,0x66,0x0,0x7e,0x0,0x3c,0x0, -}; - -static const BitmapCharRec ch38 = {12,13,-1,0,13,ch38data}; - -/* char: 0x25 '%' */ - -static const GLubyte ch37data[] = { -0x18,0x78,0x18,0xfc,0xc,0xcc,0xc,0xcc,0x6,0xfc,0x6,0x78,0x3,0x0,0x7b,0x0, -0xfd,0x80,0xcd,0x80,0xcc,0xc0,0xfc,0xc0,0x78,0x60, -}; - -static const BitmapCharRec ch37 = {14,13,-1,0,16,ch37data}; - -/* char: 0x24 '$' */ - -static const GLubyte ch36data[] = { -0x8,0x0,0x8,0x0,0x3e,0x0,0x7f,0x0,0xeb,0x80,0xc9,0x80,0x9,0x80,0xf,0x0, -0x3e,0x0,0x78,0x0,0xe8,0x0,0xc8,0x0,0xcb,0x0,0x7f,0x0,0x3e,0x0,0x8,0x0, -}; - -static const BitmapCharRec ch36 = {9,16,-1,2,10,ch36data}; - -/* char: 0x23 '#' */ - -static const GLubyte ch35data[] = { -0x24,0x0,0x24,0x0,0x24,0x0,0xff,0x80,0xff,0x80,0x12,0x0,0x12,0x0,0x12,0x0, -0x7f,0xc0,0x7f,0xc0,0x9,0x0,0x9,0x0,0x9,0x0, -}; - -static const BitmapCharRec ch35 = {10,13,0,0,10,ch35data}; - -/* char: 0x22 '"' */ - -static const GLubyte ch34data[] = { -0x90,0x90,0xd8,0xd8,0xd8, -}; - -static const BitmapCharRec ch34 = {5,5,0,-9,5,ch34data}; - -/* char: 0x21 '!' */ - -static const GLubyte ch33data[] = { -0xc0,0xc0,0x0,0x0,0x80,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -}; - -static const BitmapCharRec ch33 = {2,14,-2,0,6,ch33data}; - -/* char: 0x20 ' ' */ - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch32data[] = { 0x0 }; -static const BitmapCharRec ch32 = {1,1,0,0,5,ch32data}; -#else -static const BitmapCharRec ch32 = {0,0,0,0,5,0}; -#endif - -static const BitmapCharRec * const chars[] = { -&ch32, -&ch33, -&ch34, -&ch35, -&ch36, -&ch37, -&ch38, -&ch39, -&ch40, -&ch41, -&ch42, -&ch43, -&ch44, -&ch45, -&ch46, -&ch47, -&ch48, -&ch49, -&ch50, -&ch51, -&ch52, -&ch53, -&ch54, -&ch55, -&ch56, -&ch57, -&ch58, -&ch59, -&ch60, -&ch61, -&ch62, -&ch63, -&ch64, -&ch65, -&ch66, -&ch67, -&ch68, -&ch69, -&ch70, -&ch71, -&ch72, -&ch73, -&ch74, -&ch75, -&ch76, -&ch77, -&ch78, -&ch79, -&ch80, -&ch81, -&ch82, -&ch83, -&ch84, -&ch85, -&ch86, -&ch87, -&ch88, -&ch89, -&ch90, -&ch91, -&ch92, -&ch93, -&ch94, -&ch95, -&ch96, -&ch97, -&ch98, -&ch99, -&ch100, -&ch101, -&ch102, -&ch103, -&ch104, -&ch105, -&ch106, -&ch107, -&ch108, -&ch109, -&ch110, -&ch111, -&ch112, -&ch113, -&ch114, -&ch115, -&ch116, -&ch117, -&ch118, -&ch119, -&ch120, -&ch121, -&ch122, -&ch123, -&ch124, -&ch125, -&ch126, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -&ch160, -&ch161, -&ch162, -&ch163, -&ch164, -&ch165, -&ch166, -&ch167, -&ch168, -&ch169, -&ch170, -&ch171, -&ch172, -&ch173, -&ch174, -&ch175, -&ch176, -&ch177, -&ch178, -&ch179, -&ch180, -&ch181, -&ch182, -&ch183, -&ch184, -&ch185, -&ch186, -&ch187, -&ch188, -&ch189, -&ch190, -&ch191, -&ch192, -&ch193, -&ch194, -&ch195, -&ch196, -&ch197, -&ch198, -&ch199, -&ch200, -&ch201, -&ch202, -&ch203, -&ch204, -&ch205, -&ch206, -&ch207, -&ch208, -&ch209, -&ch210, -&ch211, -&ch212, -&ch213, -&ch214, -&ch215, -&ch216, -&ch217, -&ch218, -&ch219, -&ch220, -&ch221, -&ch222, -&ch223, -&ch224, -&ch225, -&ch226, -&ch227, -&ch228, -&ch229, -&ch230, -&ch231, -&ch232, -&ch233, -&ch234, -&ch235, -&ch236, -&ch237, -&ch238, -&ch239, -&ch240, -&ch241, -&ch242, -&ch243, -&ch244, -&ch245, -&ch246, -&ch247, -&ch248, -&ch249, -&ch250, -&ch251, -&ch252, -&ch253, -&ch254, -&ch255, -}; - -#if !defined(__IBMCPP__) -const -#endif -BitmapFontRec glutBitmapHelvetica18 = { -"-adobe-helvetica-medium-r-normal--18-180-75-75-p-98-iso8859-1", -224, -32, -chars -}; - + +/* GENERATED FILE -- DO NOT MODIFY */ + +#define glutBitmapHelvetica18 XXX +#include "glutbitmap.h" +#undef glutBitmapHelvetica18 + +/* char: 0xff */ + +static const GLubyte ch255data[] = { +0x70,0x70,0x18,0x18,0x18,0x18,0x3c,0x24,0x66,0x66,0x66,0xc3,0xc3,0xc3,0x0,0x66, +0x66, +}; + +static const BitmapCharRec ch255 = {8,17,-1,4,10,ch255data}; + +/* char: 0xfe */ + +static const GLubyte ch254data[] = { +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xde,0x0,0xff,0x0,0xe3,0x0,0xc1,0x80, +0xc1,0x80,0xc1,0x80,0xc1,0x80,0xe3,0x0,0xff,0x0,0xde,0x0,0xc0,0x0,0xc0,0x0, +0xc0,0x0,0xc0,0x0, +}; + +static const BitmapCharRec ch254 = {9,18,-1,4,11,ch254data}; + +/* char: 0xfd */ + +static const GLubyte ch253data[] = { +0x70,0x70,0x18,0x18,0x18,0x18,0x3c,0x24,0x66,0x66,0x66,0xc3,0xc3,0xc3,0x0,0x18, +0xc,0x6, +}; + +static const BitmapCharRec ch253 = {8,18,-1,4,10,ch253data}; + +/* char: 0xfc */ + +static const GLubyte ch252data[] = { +0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x0,0x66,0x66, +}; + +static const BitmapCharRec ch252 = {8,13,-1,0,10,ch252data}; + +/* char: 0xfb */ + +static const GLubyte ch251data[] = { +0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x0,0x66,0x3c,0x18, +}; + +static const BitmapCharRec ch251 = {8,14,-1,0,10,ch251data}; + +/* char: 0xfa */ + +static const GLubyte ch250data[] = { +0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x0,0x18,0xc,0x6, +}; + +static const BitmapCharRec ch250 = {8,14,-1,0,10,ch250data}; + +/* char: 0xf9 */ + +static const GLubyte ch249data[] = { +0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x0,0xc,0x18,0x30, +}; + +static const BitmapCharRec ch249 = {8,14,-1,0,10,ch249data}; + +/* char: 0xf8 */ + +static const GLubyte ch248data[] = { +0xce,0x0,0x7f,0x80,0x31,0x80,0x78,0xc0,0x6c,0xc0,0x66,0xc0,0x63,0xc0,0x31,0x80, +0x3f,0xc0,0xe,0x60, +}; + +static const BitmapCharRec ch248 = {11,10,0,0,11,ch248data}; + +/* char: 0xf7 */ + +static const GLubyte ch247data[] = { +0x18,0x18,0x0,0xff,0xff,0x0,0x18,0x18, +}; + +static const BitmapCharRec ch247 = {8,8,-1,-1,10,ch247data}; + +/* char: 0xf6 */ + +static const GLubyte ch246data[] = { +0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, +0x7f,0x0,0x3e,0x0,0x0,0x0,0x36,0x0,0x36,0x0, +}; + +static const BitmapCharRec ch246 = {9,13,-1,0,11,ch246data}; + +/* char: 0xf5 */ + +static const GLubyte ch245data[] = { +0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, +0x7f,0x0,0x3e,0x0,0x0,0x0,0x26,0x0,0x2d,0x0,0x19,0x0, +}; + +static const BitmapCharRec ch245 = {9,14,-1,0,11,ch245data}; + +/* char: 0xf4 */ + +static const GLubyte ch244data[] = { +0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, +0x7f,0x0,0x3e,0x0,0x0,0x0,0x33,0x0,0x1e,0x0,0xc,0x0, +}; + +static const BitmapCharRec ch244 = {9,14,-1,0,11,ch244data}; + +/* char: 0xf3 */ + +static const GLubyte ch243data[] = { +0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, +0x7f,0x0,0x3e,0x0,0x0,0x0,0x18,0x0,0xc,0x0,0x6,0x0, +}; + +static const BitmapCharRec ch243 = {9,14,-1,0,11,ch243data}; + +/* char: 0xf2 */ + +static const GLubyte ch242data[] = { +0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, +0x7f,0x0,0x3e,0x0,0x0,0x0,0xc,0x0,0x18,0x0,0x30,0x0, +}; + +static const BitmapCharRec ch242 = {9,14,-1,0,11,ch242data}; + +/* char: 0xf1 */ + +static const GLubyte ch241data[] = { +0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xe3,0xdf,0xce,0x0,0x4c,0x5a,0x32, +}; + +static const BitmapCharRec ch241 = {8,14,-1,0,10,ch241data}; + +/* char: 0xf0 */ + +static const GLubyte ch240data[] = { +0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, +0x7f,0x0,0x3e,0x0,0x4c,0x0,0x38,0x0,0x36,0x0,0x60,0x0, +}; + +static const BitmapCharRec ch240 = {9,14,-1,0,11,ch240data}; + +/* char: 0xef */ + +static const GLubyte ch239data[] = { +0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x0,0xd8,0xd8, +}; + +static const BitmapCharRec ch239 = {5,13,0,0,4,ch239data}; + +/* char: 0xee */ + +static const GLubyte ch238data[] = { +0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0xcc,0x78,0x30, +}; + +static const BitmapCharRec ch238 = {6,14,1,0,4,ch238data}; + +/* char: 0xed */ + +static const GLubyte ch237data[] = { +0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x0,0xc0,0x60,0x30, +}; + +static const BitmapCharRec ch237 = {4,14,0,0,4,ch237data}; + +/* char: 0xec */ + +static const GLubyte ch236data[] = { +0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x0,0x30,0x60,0xc0, +}; + +static const BitmapCharRec ch236 = {4,14,0,0,4,ch236data}; + +/* char: 0xeb */ + +static const GLubyte ch235data[] = { +0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c,0x0,0x36,0x36, +}; + +static const BitmapCharRec ch235 = {8,13,-1,0,10,ch235data}; + +/* char: 0xea */ + +static const GLubyte ch234data[] = { +0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c,0x0,0x66,0x3c,0x18, +}; + +static const BitmapCharRec ch234 = {8,14,-1,0,10,ch234data}; + +/* char: 0xe9 */ + +static const GLubyte ch233data[] = { +0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c,0x0,0x18,0xc,0x6, +}; + +static const BitmapCharRec ch233 = {8,14,-1,0,10,ch233data}; + +/* char: 0xe8 */ + +static const GLubyte ch232data[] = { +0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c,0x0,0x18,0x30,0x60, +}; + +static const BitmapCharRec ch232 = {8,14,-1,0,10,ch232data}; + +/* char: 0xe7 */ + +static const GLubyte ch231data[] = { +0x78,0x6c,0xc,0x38,0x3e,0x7f,0x63,0xc0,0xc0,0xc0,0xc0,0x63,0x7f,0x3e, +}; + +static const BitmapCharRec ch231 = {8,14,-1,4,10,ch231data}; + +/* char: 0xe6 */ + +static const GLubyte ch230data[] = { +0x75,0xe0,0xef,0xf8,0xc7,0x18,0xc6,0x0,0xe6,0x0,0x7f,0xf8,0xe,0x18,0xc6,0x18, +0xef,0xf0,0x7d,0xe0, +}; + +static const BitmapCharRec ch230 = {13,10,-1,0,15,ch230data}; + +/* char: 0xe5 */ + +static const GLubyte ch229data[] = { +0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x38,0x6c,0x6c,0x38, +}; + +static const BitmapCharRec ch229 = {7,14,-1,0,9,ch229data}; + +/* char: 0xe4 */ + +static const GLubyte ch228data[] = { +0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x6c,0x6c, +}; + +static const BitmapCharRec ch228 = {7,13,-1,0,9,ch228data}; + +/* char: 0xe3 */ + +static const GLubyte ch227data[] = { +0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x4c,0x5a,0x32, +}; + +static const BitmapCharRec ch227 = {7,14,-1,0,9,ch227data}; + +/* char: 0xe2 */ + +static const GLubyte ch226data[] = { +0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x66,0x3c,0x18, +}; + +static const BitmapCharRec ch226 = {7,14,-1,0,9,ch226data}; + +/* char: 0xe1 */ + +static const GLubyte ch225data[] = { +0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x30,0x18,0xc, +}; + +static const BitmapCharRec ch225 = {7,14,-1,0,9,ch225data}; + +/* char: 0xe0 */ + +static const GLubyte ch224data[] = { +0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c,0x0,0x18,0x30,0x60, +}; + +static const BitmapCharRec ch224 = {7,14,-1,0,9,ch224data}; + +/* char: 0xdf */ + +static const GLubyte ch223data[] = { +0xdc,0xde,0xc6,0xc6,0xc6,0xc6,0xdc,0xdc,0xc6,0xc6,0xc6,0xc6,0x7c,0x38, +}; + +static const BitmapCharRec ch223 = {7,14,-1,0,9,ch223data}; + +/* char: 0xde */ + +static const GLubyte ch222data[] = { +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x80,0xc1,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc1,0xc0,0xff,0x80,0xff,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, +}; + +static const BitmapCharRec ch222 = {10,14,-1,0,12,ch222data}; + +/* char: 0xdd */ + +static const GLubyte ch221data[] = { +0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0xf,0x0,0x19,0x80, +0x30,0xc0,0x30,0xc0,0x60,0x60,0x60,0x60,0xc0,0x30,0xc0,0x30,0x0,0x0,0x6,0x0, +0x3,0x0,0x1,0x80, +}; + +static const BitmapCharRec ch221 = {12,18,-1,0,14,ch221data}; + +/* char: 0xdc */ + +static const GLubyte ch220data[] = { +0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, +0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0x0,0x0,0x19,0x80, +0x19,0x80, +}; + +static const BitmapCharRec ch220 = {11,17,-1,0,13,ch220data}; + +/* char: 0xdb */ + +static const GLubyte ch219data[] = { +0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, +0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0x0,0x0,0x19,0x80, +0xf,0x0,0x6,0x0, +}; + +static const BitmapCharRec ch219 = {11,18,-1,0,13,ch219data}; + +/* char: 0xda */ + +static const GLubyte ch218data[] = { +0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, +0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0x0,0x0,0xc,0x0, +0x6,0x0,0x3,0x0, +}; + +static const BitmapCharRec ch218 = {11,18,-1,0,13,ch218data}; + +/* char: 0xd9 */ + +static const GLubyte ch217data[] = { +0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, +0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0x0,0x0,0x6,0x0, +0xc,0x0,0x18,0x0, +}; + +static const BitmapCharRec ch217 = {11,18,-1,0,13,ch217data}; + +/* char: 0xd8 */ + +static const GLubyte ch216data[] = { +0xc7,0xc0,0xff,0xf0,0x78,0x38,0x38,0x18,0x6c,0x1c,0x6e,0xc,0x67,0xc,0x63,0x8c, +0x61,0xcc,0x70,0xdc,0x30,0x78,0x38,0x38,0x1f,0xfc,0x7,0xcc, +}; + +static const BitmapCharRec ch216 = {14,14,0,0,15,ch216data}; + +/* char: 0xd7 */ + +static const GLubyte ch215data[] = { +0xc0,0xc0,0x61,0x80,0x33,0x0,0x1e,0x0,0xc,0x0,0x1e,0x0,0x33,0x0,0x61,0x80, +0xc0,0xc0, +}; + +static const BitmapCharRec ch215 = {10,9,0,0,10,ch215data}; + +/* char: 0xd6 */ + +static const GLubyte ch214data[] = { +0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, +0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0xd,0x80, +0xd,0x80, +}; + +static const BitmapCharRec ch214 = {13,17,-1,0,15,ch214data}; + +/* char: 0xd5 */ + +static const GLubyte ch213data[] = { +0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, +0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0x9,0x80, +0xb,0x40,0x6,0x40, +}; + +static const BitmapCharRec ch213 = {13,18,-1,0,15,ch213data}; + +/* char: 0xd4 */ + +static const GLubyte ch212data[] = { +0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, +0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0xc,0xc0, +0x7,0x80,0x3,0x0, +}; + +static const BitmapCharRec ch212 = {13,18,-1,0,15,ch212data}; + +/* char: 0xd3 */ + +static const GLubyte ch211data[] = { +0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, +0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0x3,0x0, +0x1,0x80,0x0,0xc0, +}; + +static const BitmapCharRec ch211 = {13,18,-1,0,15,ch211data}; + +/* char: 0xd2 */ + +static const GLubyte ch210data[] = { +0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, +0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80,0x0,0x0,0x3,0x0, +0x6,0x0,0xc,0x0, +}; + +static const BitmapCharRec ch210 = {13,18,-1,0,15,ch210data}; + +/* char: 0xd1 */ + +static const GLubyte ch209data[] = { +0xc0,0x60,0xc0,0xe0,0xc1,0xe0,0xc1,0xe0,0xc3,0x60,0xc6,0x60,0xc6,0x60,0xcc,0x60, +0xcc,0x60,0xd8,0x60,0xd8,0x60,0xf0,0x60,0xe0,0x60,0xe0,0x60,0x0,0x0,0x13,0x0, +0x16,0x80,0xc,0x80, +}; + +static const BitmapCharRec ch209 = {11,18,-1,0,13,ch209data}; + +/* char: 0xd0 */ + +static const GLubyte ch208data[] = { +0x7f,0x80,0x7f,0xc0,0x60,0xe0,0x60,0x60,0x60,0x30,0x60,0x30,0xfc,0x30,0xfc,0x30, +0x60,0x30,0x60,0x30,0x60,0x60,0x60,0xe0,0x7f,0xc0,0x7f,0x80, +}; + +static const BitmapCharRec ch208 = {12,14,0,0,13,ch208data}; + +/* char: 0xcf */ + +static const GLubyte ch207data[] = { +0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0xcc, +0xcc, +}; + +static const BitmapCharRec ch207 = {6,17,0,0,6,ch207data}; + +/* char: 0xce */ + +static const GLubyte ch206data[] = { +0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0xcc, +0x78,0x30, +}; + +static const BitmapCharRec ch206 = {6,18,0,0,6,ch206data}; + +/* char: 0xcd */ + +static const GLubyte ch205data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0xc0, +0x60,0x30, +}; + +static const BitmapCharRec ch205 = {4,18,-2,0,6,ch205data}; + +/* char: 0xcc */ + +static const GLubyte ch204data[] = { +0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0x30, +0x60,0xc0, +}; + +static const BitmapCharRec ch204 = {4,18,0,0,6,ch204data}; + +/* char: 0xcb */ + +static const GLubyte ch203data[] = { +0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80,0x0,0x0,0x33,0x0, +0x33,0x0, +}; + +static const BitmapCharRec ch203 = {9,17,-1,0,11,ch203data}; + +/* char: 0xca */ + +static const GLubyte ch202data[] = { +0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80,0x0,0x0,0x33,0x0, +0x1e,0x0,0xc,0x0, +}; + +static const BitmapCharRec ch202 = {9,18,-1,0,11,ch202data}; + +/* char: 0xc9 */ + +static const GLubyte ch201data[] = { +0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80,0x0,0x0,0xc,0x0, +0x6,0x0,0x3,0x0, +}; + +static const BitmapCharRec ch201 = {9,18,-1,0,11,ch201data}; + +/* char: 0xc8 */ + +static const GLubyte ch200data[] = { +0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80,0x0,0x0,0xc,0x0, +0x18,0x0,0x30,0x0, +}; + +static const BitmapCharRec ch200 = {9,18,-1,0,11,ch200data}; + +/* char: 0xc7 */ + +static const GLubyte ch199data[] = { +0x1e,0x0,0x1b,0x0,0x3,0x0,0xe,0x0,0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30, +0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xe0,0x0,0x60,0x30,0x70,0x70, +0x3f,0xe0,0xf,0x80, +}; + +static const BitmapCharRec ch199 = {12,18,-1,4,14,ch199data}; + +/* char: 0xc6 */ + +static const GLubyte ch198data[] = { +0xc1,0xff,0xc1,0xff,0x61,0x80,0x61,0x80,0x7f,0x80,0x3f,0x80,0x31,0xfe,0x31,0xfe, +0x19,0x80,0x19,0x80,0xd,0x80,0xd,0x80,0x7,0xff,0x7,0xff, +}; + +static const BitmapCharRec ch198 = {16,14,-1,0,18,ch198data}; + +/* char: 0xc5 */ + +static const GLubyte ch197data[] = { +0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, +0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0xf,0x0,0x19,0x80, +0x19,0x80,0xf,0x0, +}; + +static const BitmapCharRec ch197 = {12,18,0,0,12,ch197data}; + +/* char: 0xc4 */ + +static const GLubyte ch196data[] = { +0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, +0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x19,0x80, +0x19,0x80, +}; + +static const BitmapCharRec ch196 = {12,17,0,0,12,ch196data}; + +/* char: 0xc3 */ + +static const GLubyte ch195data[] = { +0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, +0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x13,0x0, +0x16,0x80,0xc,0x80, +}; + +static const BitmapCharRec ch195 = {12,18,0,0,12,ch195data}; + +/* char: 0xc2 */ + +static const GLubyte ch194data[] = { +0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, +0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x19,0x80, +0xf,0x0,0x6,0x0, +}; + +static const BitmapCharRec ch194 = {12,18,0,0,12,ch194data}; + +/* char: 0xc1 */ + +static const GLubyte ch193data[] = { +0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, +0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x6,0x0, +0x3,0x0,0x1,0x80, +}; + +static const BitmapCharRec ch193 = {12,18,0,0,12,ch193data}; + +/* char: 0xc0 */ + +static const GLubyte ch192data[] = { +0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, +0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0,0x0,0x0,0x6,0x0, +0xc,0x0,0x18,0x0, +}; + +static const BitmapCharRec ch192 = {12,18,0,0,12,ch192data}; + +/* char: 0xbf */ + +static const GLubyte ch191data[] = { +0x7c,0xfe,0xc6,0xc6,0xe0,0x70,0x38,0x18,0x18,0x18,0x0,0x0,0x18,0x18, +}; + +static const BitmapCharRec ch191 = {7,14,-1,4,10,ch191data}; + +/* char: 0xbe */ + +static const GLubyte ch190data[] = { +0x18,0x18,0x18,0x18,0xc,0xfc,0x6,0xd8,0x6,0x78,0x73,0x38,0xf9,0x18,0x99,0x88, +0x30,0xc0,0x30,0xc0,0x98,0x60,0xf8,0x30,0x70,0x30, +}; + +static const BitmapCharRec ch190 = {14,13,0,0,15,ch190data}; + +/* char: 0xbd */ + +static const GLubyte ch189data[] = { +0x30,0xf8,0x30,0xf8,0x18,0x60,0xc,0x30,0xc,0x18,0x66,0x98,0x62,0xf8,0x63,0x70, +0x61,0x80,0x61,0x80,0xe0,0xc0,0xe0,0x60,0x60,0x60, +}; + +static const BitmapCharRec ch189 = {13,13,-1,0,15,ch189data}; + +/* char: 0xbc */ + +static const GLubyte ch188data[] = { +0x30,0x30,0x30,0x30,0x19,0xf8,0xd,0xb0,0xc,0xf0,0x66,0x70,0x62,0x30,0x63,0x10, +0x61,0x80,0x61,0x80,0xe0,0xc0,0xe0,0x60,0x60,0x60, +}; + +static const BitmapCharRec ch188 = {13,13,-1,0,15,ch188data}; + +/* char: 0xbb */ + +static const GLubyte ch187data[] = { +0x90,0xd8,0x6c,0x36,0x36,0x6c,0xd8,0x90, +}; + +static const BitmapCharRec ch187 = {7,8,-1,-1,9,ch187data}; + +/* char: 0xba */ + +static const GLubyte ch186data[] = { +0xf8,0x0,0x70,0xd8,0x88,0x88,0xd8,0x70, +}; + +static const BitmapCharRec ch186 = {5,8,-1,-6,7,ch186data}; + +/* char: 0xb9 */ + +static const GLubyte ch185data[] = { +0x60,0x60,0x60,0x60,0x60,0xe0,0xe0,0x60, +}; + +static const BitmapCharRec ch185 = {3,8,-1,-5,6,ch185data}; + +/* char: 0xb8 */ + +static const GLubyte ch184data[] = { +0xf0,0xd8,0x18,0x70,0x60, +}; + +static const BitmapCharRec ch184 = {5,5,0,4,5,ch184data}; + +/* char: 0xb7 */ + +static const GLubyte ch183data[] = { +0xc0,0xc0, +}; + +static const BitmapCharRec ch183 = {2,2,-1,-4,4,ch183data}; + +/* char: 0xb6 */ + +static const GLubyte ch182data[] = { +0x12,0x12,0x12,0x12,0x12,0x12,0x12,0x12,0x12,0x12,0x32,0x72,0xf2,0xf2,0xf2,0xf2, +0x72,0x3f, +}; + +static const BitmapCharRec ch182 = {8,18,-1,4,10,ch182data}; + +/* char: 0xb5 */ + +static const GLubyte ch181data[] = { +0xc0,0xc0,0xc0,0xc0,0xdb,0xff,0xe7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3, +}; + +static const BitmapCharRec ch181 = {8,14,-1,4,10,ch181data}; + +/* char: 0xb4 */ + +static const GLubyte ch180data[] = { +0xc0,0x60,0x30, +}; + +static const BitmapCharRec ch180 = {4,3,0,-11,4,ch180data}; + +/* char: 0xb3 */ + +static const GLubyte ch179data[] = { +0x70,0xf8,0x98,0x30,0x30,0x98,0xf8,0x70, +}; + +static const BitmapCharRec ch179 = {5,8,0,-5,6,ch179data}; + +/* char: 0xb2 */ + +static const GLubyte ch178data[] = { +0xf8,0xf8,0x60,0x30,0x18,0x98,0xf8,0x70, +}; + +static const BitmapCharRec ch178 = {5,8,0,-5,6,ch178data}; + +/* char: 0xb1 */ + +static const GLubyte ch177data[] = { +0xff,0xff,0x0,0x18,0x18,0x18,0xff,0xff,0x18,0x18,0x18, +}; + +static const BitmapCharRec ch177 = {8,11,-1,0,10,ch177data}; + +/* char: 0xb0 */ + +static const GLubyte ch176data[] = { +0x70,0xd8,0x88,0xd8,0x70, +}; + +static const BitmapCharRec ch176 = {5,5,-1,-8,7,ch176data}; + +/* char: 0xaf */ + +static const GLubyte ch175data[] = { +0xf8, +}; + +static const BitmapCharRec ch175 = {5,1,0,-12,5,ch175data}; + +/* char: 0xae */ + +static const GLubyte ch174data[] = { +0xf,0x80,0x30,0x60,0x40,0x10,0x48,0x50,0x88,0x88,0x89,0x8,0x8f,0x88,0x88,0x48, +0x88,0x48,0x4f,0x90,0x40,0x10,0x30,0x60,0xf,0x80, +}; + +static const BitmapCharRec ch174 = {13,13,-1,0,14,ch174data}; + +/* char: 0xad */ + +static const GLubyte ch173data[] = { +0xf8,0xf8, +}; + +static const BitmapCharRec ch173 = {5,2,-1,-4,7,ch173data}; + +/* char: 0xac */ + +static const GLubyte ch172data[] = { +0x1,0x80,0x1,0x80,0x1,0x80,0xff,0x80,0xff,0x80, +}; + +static const BitmapCharRec ch172 = {9,5,-1,-3,11,ch172data}; + +/* char: 0xab */ + +static const GLubyte ch171data[] = { +0x12,0x36,0x6c,0xd8,0xd8,0x6c,0x36,0x12, +}; + +static const BitmapCharRec ch171 = {7,8,-1,-1,9,ch171data}; + +/* char: 0xaa */ + +static const GLubyte ch170data[] = { +0xf8,0x0,0x68,0xd8,0x48,0x38,0xc8,0x70, +}; + +static const BitmapCharRec ch170 = {5,8,-1,-6,7,ch170data}; + +/* char: 0xa9 */ + +static const GLubyte ch169data[] = { +0xf,0x80,0x30,0x60,0x40,0x10,0x47,0x10,0x88,0x88,0x90,0x8,0x90,0x8,0x90,0x8, +0x88,0x88,0x47,0x10,0x40,0x10,0x30,0x60,0xf,0x80, +}; + +static const BitmapCharRec ch169 = {13,13,-1,0,15,ch169data}; + +/* char: 0xa8 */ + +static const GLubyte ch168data[] = { +0xd8,0xd8, +}; + +static const BitmapCharRec ch168 = {5,2,0,-11,6,ch168data}; + +/* char: 0xa7 */ + +static const GLubyte ch167data[] = { +0x3c,0x7e,0xc3,0xc3,0x7,0xe,0x3e,0x73,0xe3,0xc3,0xc7,0x6e,0x7c,0xf0,0xc3,0xc3, +0x7e,0x3c, +}; + +static const BitmapCharRec ch167 = {8,18,-1,4,10,ch167data}; + +/* char: 0xa6 */ + +static const GLubyte ch166data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0x0,0x0,0x0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0, +}; + +static const BitmapCharRec ch166 = {2,17,-1,3,4,ch166data}; + +/* char: 0xa5 */ + +static const GLubyte ch165data[] = { +0x18,0x18,0x18,0x18,0xff,0x18,0xff,0x3c,0x66,0x66,0x66,0xc3,0xc3, +}; + +static const BitmapCharRec ch165 = {8,13,-1,0,10,ch165data}; + +/* char: 0xa4 */ + +static const GLubyte ch164data[] = { +0xc3,0xff,0x66,0x66,0x66,0xff,0xc3, +}; + +static const BitmapCharRec ch164 = {8,7,-1,-3,10,ch164data}; + +/* char: 0xa3 */ + +static const GLubyte ch163data[] = { +0xdf,0x0,0xff,0x80,0x60,0x80,0x30,0x0,0x18,0x0,0x18,0x0,0x7e,0x0,0x30,0x0, +0x60,0x0,0x61,0x80,0x61,0x80,0x3f,0x0,0x1e,0x0, +}; + +static const BitmapCharRec ch163 = {9,13,0,0,10,ch163data}; + +/* char: 0xa2 */ + +static const GLubyte ch162data[] = { +0x10,0x10,0x3e,0x7f,0x6b,0xc8,0xc8,0xc8,0xc8,0x6b,0x7f,0x3e,0x4,0x4, +}; + +static const BitmapCharRec ch162 = {8,14,-1,2,10,ch162data}; + +/* char: 0xa1 */ + +static const GLubyte ch161data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x40,0x40,0x0,0x0,0xc0,0xc0, +}; + +static const BitmapCharRec ch161 = {2,14,-2,4,6,ch161data}; + +/* char: 0xa0 */ + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch160data[] = { 0x0 }; +static const BitmapCharRec ch160 = {1,1,0,0,5,ch160data}; +#else +static const BitmapCharRec ch160 = {0,0,0,0,5,0}; +#endif + +/* char: 0x7e '~' */ + +static const GLubyte ch126data[] = { +0xcc,0x7e,0x33, +}; + +static const BitmapCharRec ch126 = {8,3,-1,-4,10,ch126data}; + +/* char: 0x7d '}' */ + +static const GLubyte ch125data[] = { +0xc0,0x60,0x30,0x30,0x30,0x30,0x30,0x30,0x18,0xc,0x18,0x30,0x30,0x30,0x30,0x30, +0x60,0xc0, +}; + +static const BitmapCharRec ch125 = {6,18,0,4,6,ch125data}; + +/* char: 0x7c '|' */ + +static const GLubyte ch124data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0, +}; + +static const BitmapCharRec ch124 = {2,18,-1,4,4,ch124data}; + +/* char: 0x7b '{' */ + +static const GLubyte ch123data[] = { +0xc,0x18,0x30,0x30,0x30,0x30,0x30,0x30,0x60,0xc0,0x60,0x30,0x30,0x30,0x30,0x30, +0x18,0xc, +}; + +static const BitmapCharRec ch123 = {6,18,0,4,6,ch123data}; + +/* char: 0x7a 'z' */ + +static const GLubyte ch122data[] = { +0xfe,0xfe,0xc0,0x60,0x30,0x18,0xc,0x6,0xfe,0xfe, +}; + +static const BitmapCharRec ch122 = {7,10,-1,0,9,ch122data}; + +/* char: 0x79 'y' */ + +static const GLubyte ch121data[] = { +0x70,0x70,0x18,0x18,0x18,0x18,0x3c,0x24,0x66,0x66,0x66,0xc3,0xc3,0xc3, +}; + +static const BitmapCharRec ch121 = {8,14,-1,4,10,ch121data}; + +/* char: 0x78 'x' */ + +static const GLubyte ch120data[] = { +0xc3,0xe7,0x66,0x3c,0x18,0x18,0x3c,0x66,0xe7,0xc3, +}; + +static const BitmapCharRec ch120 = {8,10,-1,0,10,ch120data}; + +/* char: 0x77 'w' */ + +static const GLubyte ch119data[] = { +0x19,0x80,0x19,0x80,0x39,0xc0,0x29,0x40,0x69,0x60,0x66,0x60,0x66,0x60,0xc6,0x30, +0xc6,0x30,0xc6,0x30, +}; + +static const BitmapCharRec ch119 = {12,10,-1,0,14,ch119data}; + +/* char: 0x76 'v' */ + +static const GLubyte ch118data[] = { +0x18,0x18,0x3c,0x24,0x66,0x66,0x66,0xc3,0xc3,0xc3, +}; + +static const BitmapCharRec ch118 = {8,10,-1,0,10,ch118data}; + +/* char: 0x75 'u' */ + +static const GLubyte ch117data[] = { +0x73,0xfb,0xc7,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3, +}; + +static const BitmapCharRec ch117 = {8,10,-1,0,10,ch117data}; + +/* char: 0x74 't' */ + +static const GLubyte ch116data[] = { +0x18,0x38,0x30,0x30,0x30,0x30,0x30,0x30,0xfc,0xfc,0x30,0x30,0x30, +}; + +static const BitmapCharRec ch116 = {6,13,0,0,6,ch116data}; + +/* char: 0x73 's' */ + +static const GLubyte ch115data[] = { +0x78,0xfc,0xc6,0x6,0x3e,0xfc,0xc0,0xc6,0x7e,0x3c, +}; + +static const BitmapCharRec ch115 = {7,10,-1,0,9,ch115data}; + +/* char: 0x72 'r' */ + +static const GLubyte ch114data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xe0,0xd8,0xd8, +}; + +static const BitmapCharRec ch114 = {5,10,-1,0,6,ch114data}; + +/* char: 0x71 'q' */ + +static const GLubyte ch113data[] = { +0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x3d,0x80,0x7f,0x80,0x63,0x80,0xc1,0x80, +0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x80,0x7f,0x80,0x3d,0x80, +}; + +static const BitmapCharRec ch113 = {9,14,-1,4,11,ch113data}; + +/* char: 0x70 'p' */ + +static const GLubyte ch112data[] = { +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xde,0x0,0xff,0x0,0xe3,0x0,0xc1,0x80, +0xc1,0x80,0xc1,0x80,0xc1,0x80,0xe3,0x0,0xff,0x0,0xde,0x0, +}; + +static const BitmapCharRec ch112 = {9,14,-1,4,11,ch112data}; + +/* char: 0x6f 'o' */ + +static const GLubyte ch111data[] = { +0x3e,0x0,0x7f,0x0,0x63,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x0, +0x7f,0x0,0x3e,0x0, +}; + +static const BitmapCharRec ch111 = {9,10,-1,0,11,ch111data}; + +/* char: 0x6e 'n' */ + +static const GLubyte ch110data[] = { +0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xe3,0xdf,0xce, +}; + +static const BitmapCharRec ch110 = {8,10,-1,0,10,ch110data}; + +/* char: 0x6d 'm' */ + +static const GLubyte ch109data[] = { +0xc6,0x30,0xc6,0x30,0xc6,0x30,0xc6,0x30,0xc6,0x30,0xc6,0x30,0xc6,0x30,0xe7,0x30, +0xde,0xf0,0xcc,0x60, +}; + +static const BitmapCharRec ch109 = {12,10,-1,0,14,ch109data}; + +/* char: 0x6c 'l' */ + +static const GLubyte ch108data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +}; + +static const BitmapCharRec ch108 = {2,14,-1,0,4,ch108data}; + +/* char: 0x6b 'k' */ + +static const GLubyte ch107data[] = { +0xc7,0xc6,0xce,0xcc,0xd8,0xf8,0xf0,0xd8,0xcc,0xc6,0xc0,0xc0,0xc0,0xc0, +}; + +static const BitmapCharRec ch107 = {8,14,-1,0,9,ch107data}; + +/* char: 0x6a 'j' */ + +static const GLubyte ch106data[] = { +0xe0,0xf0,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x0,0x0, +0x30,0x30, +}; + +static const BitmapCharRec ch106 = {4,18,1,4,4,ch106data}; + +/* char: 0x69 'i' */ + +static const GLubyte ch105data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0x0,0xc0,0xc0, +}; + +static const BitmapCharRec ch105 = {2,14,-1,0,4,ch105data}; + +/* char: 0x68 'h' */ + +static const GLubyte ch104data[] = { +0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xe3,0xdf,0xce,0xc0,0xc0,0xc0,0xc0, +}; + +static const BitmapCharRec ch104 = {8,14,-1,0,10,ch104data}; + +/* char: 0x67 'g' */ + +static const GLubyte ch103data[] = { +0x1c,0x0,0x7f,0x0,0x63,0x0,0x1,0x80,0x3d,0x80,0x7f,0x80,0x63,0x80,0xc1,0x80, +0xc1,0x80,0xc1,0x80,0xc1,0x80,0x61,0x80,0x7f,0x80,0x3d,0x80, +}; + +static const BitmapCharRec ch103 = {9,14,-1,4,11,ch103data}; + +/* char: 0x66 'f' */ + +static const GLubyte ch102data[] = { +0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0xfc,0xfc,0x30,0x30,0x3c,0x1c, +}; + +static const BitmapCharRec ch102 = {6,14,0,0,6,ch102data}; + +/* char: 0x65 'e' */ + +static const GLubyte ch101data[] = { +0x3c,0x7f,0xe3,0xc0,0xc0,0xff,0xc3,0xc3,0x7e,0x3c, +}; + +static const BitmapCharRec ch101 = {8,10,-1,0,10,ch101data}; + +/* char: 0x64 'd' */ + +static const GLubyte ch100data[] = { +0x3d,0x80,0x7f,0x80,0x63,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x63,0x80, +0x7f,0x80,0x3d,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80, +}; + +static const BitmapCharRec ch100 = {9,14,-1,0,11,ch100data}; + +/* char: 0x63 'c' */ + +static const GLubyte ch99data[] = { +0x3e,0x7f,0x63,0xc0,0xc0,0xc0,0xc0,0x63,0x7f,0x3e, +}; + +static const BitmapCharRec ch99 = {8,10,-1,0,10,ch99data}; + +/* char: 0x62 'b' */ + +static const GLubyte ch98data[] = { +0xde,0x0,0xff,0x0,0xe3,0x0,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xe3,0x0, +0xff,0x0,0xde,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, +}; + +static const BitmapCharRec ch98 = {9,14,-1,0,11,ch98data}; + +/* char: 0x61 'a' */ + +static const GLubyte ch97data[] = { +0x76,0xee,0xc6,0xc6,0xe6,0x7e,0xe,0xc6,0xee,0x7c, +}; + +static const BitmapCharRec ch97 = {7,10,-1,0,9,ch97data}; + +/* char: 0x60 '`' */ + +static const GLubyte ch96data[] = { +0xc0,0xc0,0x80,0x80,0x40, +}; + +static const BitmapCharRec ch96 = {2,5,-1,-9,4,ch96data}; + +/* char: 0x5f '_' */ + +static const GLubyte ch95data[] = { +0xff,0xc0,0xff,0xc0, +}; + +static const BitmapCharRec ch95 = {10,2,0,4,10,ch95data}; + +/* char: 0x5e '^' */ + +static const GLubyte ch94data[] = { +0x82,0xc6,0x6c,0x38,0x10, +}; + +static const BitmapCharRec ch94 = {7,5,-1,-8,9,ch94data}; + +/* char: 0x5d ']' */ + +static const GLubyte ch93data[] = { +0xf0,0xf0,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, +0xf0,0xf0, +}; + +static const BitmapCharRec ch93 = {4,18,0,4,5,ch93data}; + +/* char: 0x5c '\' */ + +static const GLubyte ch92data[] = { +0x18,0x18,0x10,0x10,0x30,0x30,0x20,0x20,0x60,0x60,0x40,0x40,0xc0,0xc0, +}; + +static const BitmapCharRec ch92 = {5,14,0,0,5,ch92data}; + +/* char: 0x5b '[' */ + +static const GLubyte ch91data[] = { +0xf0,0xf0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xf0,0xf0, +}; + +static const BitmapCharRec ch91 = {4,18,-1,4,5,ch91data}; + +/* char: 0x5a 'Z' */ + +static const GLubyte ch90data[] = { +0xff,0xc0,0xff,0xc0,0xc0,0x0,0x60,0x0,0x30,0x0,0x18,0x0,0x1c,0x0,0xc,0x0, +0x6,0x0,0x3,0x0,0x1,0x80,0x0,0xc0,0xff,0xc0,0xff,0xc0, +}; + +static const BitmapCharRec ch90 = {10,14,-1,0,12,ch90data}; + +/* char: 0x59 'Y' */ + +static const GLubyte ch89data[] = { +0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0xf,0x0,0x19,0x80, +0x30,0xc0,0x30,0xc0,0x60,0x60,0x60,0x60,0xc0,0x30,0xc0,0x30, +}; + +static const BitmapCharRec ch89 = {12,14,-1,0,14,ch89data}; + +/* char: 0x58 'X' */ + +static const GLubyte ch88data[] = { +0xc0,0x60,0xe0,0xe0,0x60,0xc0,0x71,0xc0,0x31,0x80,0x1b,0x0,0xe,0x0,0xe,0x0, +0x1b,0x0,0x31,0x80,0x71,0xc0,0x60,0xc0,0xe0,0xe0,0xc0,0x60, +}; + +static const BitmapCharRec ch88 = {11,14,-1,0,13,ch88data}; + +/* char: 0x57 'W' */ + +static const GLubyte ch87data[] = { +0x18,0x18,0x18,0x18,0x1c,0x38,0x34,0x2c,0x36,0x6c,0x36,0x6c,0x66,0x66,0x66,0x66, +0x62,0x46,0x63,0xc6,0xc3,0xc3,0xc1,0x83,0xc1,0x83,0xc1,0x83, +}; + +static const BitmapCharRec ch87 = {16,14,-1,0,18,ch87data}; + +/* char: 0x56 'V' */ + +static const GLubyte ch86data[] = { +0x6,0x0,0xf,0x0,0xf,0x0,0x19,0x80,0x19,0x80,0x19,0x80,0x30,0xc0,0x30,0xc0, +0x30,0xc0,0x60,0x60,0x60,0x60,0x60,0x60,0xc0,0x30,0xc0,0x30, +}; + +static const BitmapCharRec ch86 = {12,14,-1,0,14,ch86data}; + +/* char: 0x55 'U' */ + +static const GLubyte ch85data[] = { +0x1f,0x0,0x7f,0xc0,0x60,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, +0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, +}; + +static const BitmapCharRec ch85 = {11,14,-1,0,13,ch85data}; + +/* char: 0x54 'T' */ + +static const GLubyte ch84data[] = { +0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0, +0xc,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0xff,0xc0,0xff,0xc0, +}; + +static const BitmapCharRec ch84 = {10,14,-1,0,12,ch84data}; + +/* char: 0x53 'S' */ + +static const GLubyte ch83data[] = { +0x3f,0x0,0x7f,0xc0,0xe0,0xe0,0xc0,0x60,0x0,0x60,0x0,0xe0,0x3,0xc0,0x1f,0x0, +0x7c,0x0,0xe0,0x0,0xc0,0x60,0xe0,0xe0,0x7f,0xc0,0x1f,0x0, +}; + +static const BitmapCharRec ch83 = {11,14,-1,0,13,ch83data}; + +/* char: 0x52 'R' */ + +static const GLubyte ch82data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc1,0x80,0xc1,0x80,0xff,0x0,0xff,0x80, +0xc1,0xc0,0xc0,0xc0,0xc0,0xc0,0xc1,0xc0,0xff,0x80,0xff,0x0, +}; + +static const BitmapCharRec ch82 = {10,14,-1,0,12,ch82data}; + +/* char: 0x51 'Q' */ + +static const GLubyte ch81data[] = { +0x0,0x30,0xf,0xb0,0x3f,0xe0,0x70,0xf0,0x61,0xb0,0xe1,0xb8,0xc0,0x18,0xc0,0x18, +0xc0,0x18,0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80, +}; + +static const BitmapCharRec ch81 = {13,15,-1,1,15,ch81data}; + +/* char: 0x50 'P' */ + +static const GLubyte ch80data[] = { +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x80, +0xc1,0xc0,0xc0,0xc0,0xc0,0xc0,0xc1,0xc0,0xff,0x80,0xff,0x0, +}; + +static const BitmapCharRec ch80 = {10,14,-1,0,12,ch80data}; + +/* char: 0x4f 'O' */ + +static const GLubyte ch79data[] = { +0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x38,0xc0,0x18,0xc0,0x18,0xc0,0x18, +0xc0,0x18,0xe0,0x38,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80, +}; + +static const BitmapCharRec ch79 = {13,14,-1,0,15,ch79data}; + +/* char: 0x4e 'N' */ + +static const GLubyte ch78data[] = { +0xc0,0x60,0xc0,0xe0,0xc1,0xe0,0xc1,0xe0,0xc3,0x60,0xc6,0x60,0xc6,0x60,0xcc,0x60, +0xcc,0x60,0xd8,0x60,0xf0,0x60,0xf0,0x60,0xe0,0x60,0xc0,0x60, +}; + +static const BitmapCharRec ch78 = {11,14,-1,0,13,ch78data}; + +/* char: 0x4d 'M' */ + +static const GLubyte ch77data[] = { +0xc3,0xc,0xc3,0xc,0xc7,0x8c,0xc4,0x8c,0xcc,0xcc,0xcc,0xcc,0xd8,0x6c,0xd8,0x6c, +0xf0,0x3c,0xf0,0x3c,0xe0,0x1c,0xe0,0x1c,0xc0,0xc,0xc0,0xc, +}; + +static const BitmapCharRec ch77 = {14,14,-1,0,16,ch77data}; + +/* char: 0x4c 'L' */ + +static const GLubyte ch76data[] = { +0xff,0xff,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +}; + +static const BitmapCharRec ch76 = {8,14,-1,0,10,ch76data}; + +/* char: 0x4b 'K' */ + +static const GLubyte ch75data[] = { +0xc0,0x70,0xc0,0xe0,0xc1,0xc0,0xc3,0x80,0xc7,0x0,0xce,0x0,0xfc,0x0,0xf8,0x0, +0xdc,0x0,0xce,0x0,0xc7,0x0,0xc3,0x80,0xc1,0xc0,0xc0,0xe0, +}; + +static const BitmapCharRec ch75 = {12,14,-1,0,13,ch75data}; + +/* char: 0x4a 'J' */ + +static const GLubyte ch74data[] = { +0x3c,0x7e,0xe7,0xc3,0xc3,0x3,0x3,0x3,0x3,0x3,0x3,0x3,0x3,0x3, +}; + +static const BitmapCharRec ch74 = {8,14,-1,0,10,ch74data}; + +/* char: 0x49 'I' */ + +static const GLubyte ch73data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +}; + +static const BitmapCharRec ch73 = {2,14,-2,0,6,ch73data}; + +/* char: 0x48 'H' */ + +static const GLubyte ch72data[] = { +0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xff,0xe0,0xff,0xe0, +0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, +}; + +static const BitmapCharRec ch72 = {11,14,-1,0,13,ch72data}; + +/* char: 0x47 'G' */ + +static const GLubyte ch71data[] = { +0xf,0xb0,0x3f,0xf0,0x70,0x70,0x60,0x30,0xe0,0x30,0xc1,0xf0,0xc1,0xf0,0xc0,0x0, +0xc0,0x0,0xe0,0x30,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80, +}; + +static const BitmapCharRec ch71 = {12,14,-1,0,14,ch71data}; + +/* char: 0x46 'F' */ + +static const GLubyte ch70data[] = { +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80, +}; + +static const BitmapCharRec ch70 = {9,14,-1,0,11,ch70data}; + +/* char: 0x45 'E' */ + +static const GLubyte ch69data[] = { +0xff,0x80,0xff,0x80,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x0,0xff,0x0, +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80,0xff,0x80, +}; + +static const BitmapCharRec ch69 = {9,14,-1,0,11,ch69data}; + +/* char: 0x44 'D' */ + +static const GLubyte ch68data[] = { +0xff,0x0,0xff,0x80,0xc1,0xc0,0xc0,0xc0,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60, +0xc0,0x60,0xc0,0x60,0xc0,0xc0,0xc1,0xc0,0xff,0x80,0xff,0x0, +}; + +static const BitmapCharRec ch68 = {11,14,-1,0,13,ch68data}; + +/* char: 0x43 'C' */ + +static const GLubyte ch67data[] = { +0xf,0x80,0x3f,0xe0,0x70,0x70,0x60,0x30,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, +0xc0,0x0,0xe0,0x0,0x60,0x30,0x70,0x70,0x3f,0xe0,0xf,0x80, +}; + +static const BitmapCharRec ch67 = {12,14,-1,0,14,ch67data}; + +/* char: 0x42 'B' */ + +static const GLubyte ch66data[] = { +0xff,0x80,0xff,0xc0,0xc0,0xe0,0xc0,0x60,0xc0,0x60,0xc0,0xe0,0xff,0xc0,0xff,0x80, +0xc1,0x80,0xc0,0xc0,0xc0,0xc0,0xc1,0xc0,0xff,0x80,0xff,0x0, +}; + +static const BitmapCharRec ch66 = {11,14,-1,0,13,ch66data}; + +/* char: 0x41 'A' */ + +static const GLubyte ch65data[] = { +0xc0,0x30,0xc0,0x30,0x60,0x60,0x60,0x60,0x7f,0xe0,0x3f,0xc0,0x30,0xc0,0x30,0xc0, +0x19,0x80,0x19,0x80,0xf,0x0,0xf,0x0,0x6,0x0,0x6,0x0, +}; + +static const BitmapCharRec ch65 = {12,14,0,0,12,ch65data}; + +/* char: 0x40 '@' */ + +static const GLubyte ch64data[] = { +0x7,0xe0,0x1f,0xf0,0x38,0x0,0x70,0x0,0x67,0x70,0xcf,0xf8,0xcc,0xcc,0xcc,0x66, +0xcc,0x66,0xcc,0x63,0xc6,0x33,0x67,0x73,0x63,0xb3,0x30,0x6,0x1c,0xe,0xf,0xfc, +0x3,0xf0, +}; + +static const BitmapCharRec ch64 = {16,17,-1,3,18,ch64data}; + +/* char: 0x3f '?' */ + +static const GLubyte ch63data[] = { +0x30,0x30,0x0,0x0,0x30,0x30,0x30,0x38,0x1c,0xe,0xc6,0xc6,0xfe,0x7c, +}; + +static const BitmapCharRec ch63 = {7,14,-1,0,10,ch63data}; + +/* char: 0x3e '>' */ + +static const GLubyte ch62data[] = { +0xc0,0xf0,0x3c,0xe,0x3,0xe,0x3c,0xf0,0xc0, +}; + +static const BitmapCharRec ch62 = {8,9,-1,0,10,ch62data}; + +/* char: 0x3d '=' */ + +static const GLubyte ch61data[] = { +0xfe,0xfe,0x0,0x0,0xfe,0xfe, +}; + +static const BitmapCharRec ch61 = {7,6,-2,-2,11,ch61data}; + +/* char: 0x3c '<' */ + +static const GLubyte ch60data[] = { +0x3,0xf,0x3c,0x70,0xc0,0x70,0x3c,0xf,0x3, +}; + +static const BitmapCharRec ch60 = {8,9,-1,0,10,ch60data}; + +/* char: 0x3b ';' */ + +static const GLubyte ch59data[] = { +0x80,0x40,0x40,0xc0,0xc0,0x0,0x0,0x0,0x0,0x0,0x0,0xc0,0xc0, +}; + +static const BitmapCharRec ch59 = {2,13,-1,3,5,ch59data}; + +/* char: 0x3a ':' */ + +static const GLubyte ch58data[] = { +0xc0,0xc0,0x0,0x0,0x0,0x0,0x0,0x0,0xc0,0xc0, +}; + +static const BitmapCharRec ch58 = {2,10,-1,0,5,ch58data}; + +/* char: 0x39 '9' */ + +static const GLubyte ch57data[] = { +0x7c,0xfe,0xc6,0x3,0x3,0x3b,0x7f,0xc3,0xc3,0xc3,0xc7,0x7e,0x3c, +}; + +static const BitmapCharRec ch57 = {8,13,-1,0,10,ch57data}; + +/* char: 0x38 '8' */ + +static const GLubyte ch56data[] = { +0x3c,0x7e,0xe7,0xc3,0xc3,0x66,0x7e,0x66,0xc3,0xc3,0xe7,0x7e,0x3c, +}; + +static const BitmapCharRec ch56 = {8,13,-1,0,10,ch56data}; + +/* char: 0x37 '7' */ + +static const GLubyte ch55data[] = { +0x60,0x60,0x30,0x30,0x30,0x18,0x18,0xc,0xc,0x6,0x3,0xff,0xff, +}; + +static const BitmapCharRec ch55 = {8,13,-1,0,10,ch55data}; + +/* char: 0x36 '6' */ + +static const GLubyte ch54data[] = { +0x3c,0x7e,0xe3,0xc3,0xc3,0xc3,0xfe,0xdc,0xc0,0xc0,0x63,0x7f,0x3c, +}; + +static const BitmapCharRec ch54 = {8,13,-1,0,10,ch54data}; + +/* char: 0x35 '5' */ + +static const GLubyte ch53data[] = { +0x7c,0xfe,0xc7,0xc3,0x3,0x3,0xc7,0xfe,0xfc,0xc0,0xc0,0xfe,0xfe, +}; + +static const BitmapCharRec ch53 = {8,13,-1,0,10,ch53data}; + +/* char: 0x34 '4' */ + +static const GLubyte ch52data[] = { +0x3,0x0,0x3,0x0,0x3,0x0,0xff,0x80,0xff,0x80,0xc3,0x0,0x63,0x0,0x33,0x0, +0x33,0x0,0x1b,0x0,0xf,0x0,0x7,0x0,0x3,0x0, +}; + +static const BitmapCharRec ch52 = {9,13,-1,0,10,ch52data}; + +/* char: 0x33 '3' */ + +static const GLubyte ch51data[] = { +0x3c,0x7e,0xc7,0xc3,0x3,0x7,0x1e,0x1c,0x6,0xc3,0xc3,0x7e,0x3c, +}; + +static const BitmapCharRec ch51 = {8,13,-1,0,10,ch51data}; + +/* char: 0x32 '2' */ + +static const GLubyte ch50data[] = { +0xff,0xff,0xc0,0xe0,0x70,0x38,0x1c,0xe,0x7,0x3,0xc3,0xfe,0x3c, +}; + +static const BitmapCharRec ch50 = {8,13,-1,0,10,ch50data}; + +/* char: 0x31 '1' */ + +static const GLubyte ch49data[] = { +0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0xf8,0xf8,0x18, +}; + +static const BitmapCharRec ch49 = {5,13,-2,0,10,ch49data}; + +/* char: 0x30 '0' */ + +static const GLubyte ch48data[] = { +0x3c,0x7e,0x66,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0xc3,0x66,0x7e,0x3c, +}; + +static const BitmapCharRec ch48 = {8,13,-1,0,10,ch48data}; + +/* char: 0x2f '/' */ + +static const GLubyte ch47data[] = { +0xc0,0xc0,0x40,0x40,0x60,0x60,0x20,0x20,0x30,0x30,0x10,0x10,0x18,0x18, +}; + +static const BitmapCharRec ch47 = {5,14,0,0,5,ch47data}; + +/* char: 0x2e '.' */ + +static const GLubyte ch46data[] = { +0xc0,0xc0, +}; + +static const BitmapCharRec ch46 = {2,2,-1,0,5,ch46data}; + +/* char: 0x2d '-' */ + +static const GLubyte ch45data[] = { +0xff,0xff, +}; + +static const BitmapCharRec ch45 = {8,2,-1,-4,11,ch45data}; + +/* char: 0x2c ',' */ + +static const GLubyte ch44data[] = { +0x80,0x40,0x40,0xc0,0xc0, +}; + +static const BitmapCharRec ch44 = {2,5,-1,3,5,ch44data}; + +/* char: 0x2b '+' */ + +static const GLubyte ch43data[] = { +0x18,0x18,0x18,0x18,0xff,0xff,0x18,0x18,0x18,0x18, +}; + +static const BitmapCharRec ch43 = {8,10,-1,0,10,ch43data}; + +/* char: 0x2a '*' */ + +static const GLubyte ch42data[] = { +0x88,0x70,0x70,0xf8,0x20,0x20, +}; + +static const BitmapCharRec ch42 = {5,6,-1,-8,7,ch42data}; + +/* char: 0x29 ')' */ + +static const GLubyte ch41data[] = { +0x80,0xc0,0x60,0x60,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x60,0x60, +0xc0,0x80, +}; + +static const BitmapCharRec ch41 = {4,18,-1,4,6,ch41data}; + +/* char: 0x28 '(' */ + +static const GLubyte ch40data[] = { +0x10,0x30,0x60,0x60,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x60,0x60, +0x30,0x10, +}; + +static const BitmapCharRec ch40 = {4,18,-1,4,6,ch40data}; + +/* char: 0x27 ''' */ + +static const GLubyte ch39data[] = { +0x80,0x40,0x40,0xc0,0xc0, +}; + +static const BitmapCharRec ch39 = {2,5,-1,-9,4,ch39data}; + +/* char: 0x26 '&' */ + +static const GLubyte ch38data[] = { +0x3c,0x70,0x7e,0xe0,0xe7,0xc0,0xc3,0x80,0xc3,0xc0,0xc6,0xc0,0xee,0xc0,0x7c,0x0, +0x3c,0x0,0x66,0x0,0x66,0x0,0x7e,0x0,0x3c,0x0, +}; + +static const BitmapCharRec ch38 = {12,13,-1,0,13,ch38data}; + +/* char: 0x25 '%' */ + +static const GLubyte ch37data[] = { +0x18,0x78,0x18,0xfc,0xc,0xcc,0xc,0xcc,0x6,0xfc,0x6,0x78,0x3,0x0,0x7b,0x0, +0xfd,0x80,0xcd,0x80,0xcc,0xc0,0xfc,0xc0,0x78,0x60, +}; + +static const BitmapCharRec ch37 = {14,13,-1,0,16,ch37data}; + +/* char: 0x24 '$' */ + +static const GLubyte ch36data[] = { +0x8,0x0,0x8,0x0,0x3e,0x0,0x7f,0x0,0xeb,0x80,0xc9,0x80,0x9,0x80,0xf,0x0, +0x3e,0x0,0x78,0x0,0xe8,0x0,0xc8,0x0,0xcb,0x0,0x7f,0x0,0x3e,0x0,0x8,0x0, +}; + +static const BitmapCharRec ch36 = {9,16,-1,2,10,ch36data}; + +/* char: 0x23 '#' */ + +static const GLubyte ch35data[] = { +0x24,0x0,0x24,0x0,0x24,0x0,0xff,0x80,0xff,0x80,0x12,0x0,0x12,0x0,0x12,0x0, +0x7f,0xc0,0x7f,0xc0,0x9,0x0,0x9,0x0,0x9,0x0, +}; + +static const BitmapCharRec ch35 = {10,13,0,0,10,ch35data}; + +/* char: 0x22 '"' */ + +static const GLubyte ch34data[] = { +0x90,0x90,0xd8,0xd8,0xd8, +}; + +static const BitmapCharRec ch34 = {5,5,0,-9,5,ch34data}; + +/* char: 0x21 '!' */ + +static const GLubyte ch33data[] = { +0xc0,0xc0,0x0,0x0,0x80,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +}; + +static const BitmapCharRec ch33 = {2,14,-2,0,6,ch33data}; + +/* char: 0x20 ' ' */ + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch32data[] = { 0x0 }; +static const BitmapCharRec ch32 = {1,1,0,0,5,ch32data}; +#else +static const BitmapCharRec ch32 = {0,0,0,0,5,0}; +#endif + +static const BitmapCharRec * const chars[] = { +&ch32, +&ch33, +&ch34, +&ch35, +&ch36, +&ch37, +&ch38, +&ch39, +&ch40, +&ch41, +&ch42, +&ch43, +&ch44, +&ch45, +&ch46, +&ch47, +&ch48, +&ch49, +&ch50, +&ch51, +&ch52, +&ch53, +&ch54, +&ch55, +&ch56, +&ch57, +&ch58, +&ch59, +&ch60, +&ch61, +&ch62, +&ch63, +&ch64, +&ch65, +&ch66, +&ch67, +&ch68, +&ch69, +&ch70, +&ch71, +&ch72, +&ch73, +&ch74, +&ch75, +&ch76, +&ch77, +&ch78, +&ch79, +&ch80, +&ch81, +&ch82, +&ch83, +&ch84, +&ch85, +&ch86, +&ch87, +&ch88, +&ch89, +&ch90, +&ch91, +&ch92, +&ch93, +&ch94, +&ch95, +&ch96, +&ch97, +&ch98, +&ch99, +&ch100, +&ch101, +&ch102, +&ch103, +&ch104, +&ch105, +&ch106, +&ch107, +&ch108, +&ch109, +&ch110, +&ch111, +&ch112, +&ch113, +&ch114, +&ch115, +&ch116, +&ch117, +&ch118, +&ch119, +&ch120, +&ch121, +&ch122, +&ch123, +&ch124, +&ch125, +&ch126, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +&ch160, +&ch161, +&ch162, +&ch163, +&ch164, +&ch165, +&ch166, +&ch167, +&ch168, +&ch169, +&ch170, +&ch171, +&ch172, +&ch173, +&ch174, +&ch175, +&ch176, +&ch177, +&ch178, +&ch179, +&ch180, +&ch181, +&ch182, +&ch183, +&ch184, +&ch185, +&ch186, +&ch187, +&ch188, +&ch189, +&ch190, +&ch191, +&ch192, +&ch193, +&ch194, +&ch195, +&ch196, +&ch197, +&ch198, +&ch199, +&ch200, +&ch201, +&ch202, +&ch203, +&ch204, +&ch205, +&ch206, +&ch207, +&ch208, +&ch209, +&ch210, +&ch211, +&ch212, +&ch213, +&ch214, +&ch215, +&ch216, +&ch217, +&ch218, +&ch219, +&ch220, +&ch221, +&ch222, +&ch223, +&ch224, +&ch225, +&ch226, +&ch227, +&ch228, +&ch229, +&ch230, +&ch231, +&ch232, +&ch233, +&ch234, +&ch235, +&ch236, +&ch237, +&ch238, +&ch239, +&ch240, +&ch241, +&ch242, +&ch243, +&ch244, +&ch245, +&ch246, +&ch247, +&ch248, +&ch249, +&ch250, +&ch251, +&ch252, +&ch253, +&ch254, +&ch255, +}; + +#if !defined(__IBMCPP__) +const +#endif +BitmapFontRec glutBitmapHelvetica18 = { +"-adobe-helvetica-medium-r-normal--18-180-75-75-p-98-iso8859-1", +224, +32, +chars +}; + diff --git a/src/glut/os2/glut_init.cpp b/src/glut/os2/glut_init.cpp index da5d0735e2..9e9f2e366e 100644 --- a/src/glut/os2/glut_init.cpp +++ b/src/glut/os2/glut_init.cpp @@ -1,451 +1,451 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#ifdef __VMS -#include -#endif - -#include -#include -#include - -#if !defined(_WIN32) && !defined(__OS2__) - #include - #include -#endif - -/* SGI optimization introduced in IRIX 6.3 to avoid X server - round trips for interning common X atoms. */ -#if defined(_SGI_EXTRA_PREDEFINES) && !defined(NO_FAST_ATOMS) -#include -#else -#define XSGIFastInternAtom(dpy,string,fast_name,how) XInternAtom(dpy,string,how) -#endif - -#include "glutint.h" - -/* GLUT inter-file variables */ -/* *INDENT-OFF* */ -char *__glutProgramName = NULL; -int __glutArgc = 0; -char **__glutArgv = NULL; -char *__glutGeometry = NULL; -Display *__glutDisplay = NULL; -int __glutScreen; -Window __glutRoot; -int __glutScreenHeight; -int __glutScreenWidth; -GLboolean __glutIconic = GL_FALSE; -GLboolean __glutDebug = GL_FALSE; -unsigned int __glutDisplayMode = - GLUT_RGB | GLUT_SINGLE | GLUT_DEPTH; -char *__glutDisplayString = NULL; -int __glutConnectionFD; -XSizeHints __glutSizeHints = {0}; -int __glutInitWidth = 300, __glutInitHeight = 300; -int __glutInitX = -1, __glutInitY = -1; -GLboolean __glutForceDirect = GL_FALSE, - __glutTryDirect = GL_TRUE; -Atom __glutWMDeleteWindow; -/* *INDENT-ON* */ - -#ifdef _WIN32 -void (__cdecl *__glutExitFunc)(int retval) = NULL; -#endif - -static Bool synchronize = False; - -#if defined(__OS2__) - -MRESULT EXPENTRY GlutWindowProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ); -MRESULT EXPENTRY GlutWindowChildProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ); - - -void __glutOpenOS2Connection(char* display) -{ - static char *classname=NULL; -extern HAB hab; /* PM anchor block handle */ -ERRORID erridErrorCode;/* last error id code */ -int ii; - - /* Make sure we register the window only once. */ - if(classname) - return; - - classname = "GLUT"; - - if ( !WinRegisterClass( hab, /* PM anchor block handle */ - classname,/* window class name */ - GlutWindowProc,/* address of window procedure*/ - CS_SIZEREDRAW, /* |CS_SYNCPAINT size changes cause redrawing */ - 0UL ) ) /* window data */ - { erridErrorCode = WinGetLastError(hab); - ii = erridErrorCode; - return; - } - - classname = "GLUTCHILD"; - - if ( !WinRegisterClass( hab, /* PM anchor block handle */ - classname,/* window class name */ - GlutWindowChildProc,/* address of window procedure*/ - CS_SIZEREDRAW, /* size changes cause redrawing */ - 0UL ) ) /* window data */ - { erridErrorCode = WinGetLastError(hab); - ii = erridErrorCode; - return; - } - - __glutScreenWidth = GetSystemMetrics(SM_CXSCREEN); - __glutScreenHeight = GetSystemMetrics(SM_CYSCREEN); - - /* Set the root window to NULL because windows creates a top-level - window when the parent is NULL. X creates a top-level window - when the parent is the root window. */ - __glutRoot = NULLHANDLE; - - /* Set the display to 1 -- we shouldn't be using this anywhere - (except as an argument to X calls). */ - __glutDisplay = (Display*)1; - - /* There isn't any concept of multiple screens in Win32, therefore, - we don't need to keep track of the screen we're on... it's always - the same one. */ - __glutScreen = 0; -} - -#elif defined(_WIN32) - -#ifdef __BORLANDC__ -#include /* For masking floating point exceptions. */ -#endif - -void -__glutOpenWin32Connection(char* display) -{ - static char *classname; - WNDCLASS wc; - HINSTANCE hInstance = GetModuleHandle(NULL); - - /* Make sure we register the window only once. */ - if(classname) - return; - -#ifdef __BORLANDC__ - /* Under certain conditions (e.g. while rendering solid surfaces with - lighting enabled) Microsoft OpenGL libraries cause some illegal - operations like floating point overflow or division by zero. The - default behaviour of Microsoft compilers is to mask (ignore) - floating point exceptions, while Borland compilers do not. The - following function of Borland RTL allows to mask exceptions. - Advice from Pier Giorgio Esposito (mc2172@mclink.it). */ - _control87(MCW_EM,MCW_EM); -#endif - - classname = "GLUT"; - - /* Clear (important!) and then fill in the window class structure. */ - memset(&wc, 0, sizeof(WNDCLASS)); - wc.style = CS_OWNDC; - wc.lpfnWndProc = (WNDPROC)__glutWindowProc; - wc.hInstance = hInstance; - wc.hIcon = LoadIcon(hInstance, "GLUT_ICON"); - wc.hCursor = LoadCursor(hInstance, IDC_ARROW); - wc.hbrBackground = NULL; - wc.lpszMenuName = NULL; - wc.lpszClassName = classname; - - /* Fill in a default icon if one isn't specified as a resource. */ - if(!wc.hIcon) - wc.hIcon = LoadIcon(NULL, IDI_WINLOGO); - - if(!RegisterClass(&wc)) { - __glutFatalError("RegisterClass() failed:" - "Cannot register GLUT window class."); - } - - __glutScreenWidth = GetSystemMetrics(SM_CXSCREEN); - __glutScreenHeight = GetSystemMetrics(SM_CYSCREEN); - - /* Set the root window to NULL because windows creates a top-level - window when the parent is NULL. X creates a top-level window - when the parent is the root window. */ - __glutRoot = NULL; - - /* Set the display to 1 -- we shouldn't be using this anywhere - (except as an argument to X calls). */ - __glutDisplay = (Display*)1; - - /* There isn't any concept of multiple screens in Win32, therefore, - we don't need to keep track of the screen we're on... it's always - the same one. */ - __glutScreen = 0; -} -#else /* !_WIN32 */ -void -__glutOpenXConnection(char *display) -{ - int errorBase, eventBase; - - __glutDisplay = XOpenDisplay(display); - if (!__glutDisplay) - __glutFatalError("could not open display: %s", - XDisplayName(display)); - if (synchronize) - XSynchronize(__glutDisplay, True); - if (!glXQueryExtension(__glutDisplay, &errorBase, &eventBase)) - __glutFatalError( - "OpenGL GLX extension not supported by display: %s", - XDisplayName(display)); - __glutScreen = DefaultScreen(__glutDisplay); - __glutRoot = RootWindow(__glutDisplay, __glutScreen); - __glutScreenWidth = DisplayWidth(__glutDisplay, __glutScreen); - __glutScreenHeight = DisplayHeight(__glutDisplay, - __glutScreen); - __glutConnectionFD = ConnectionNumber(__glutDisplay); - __glutWMDeleteWindow = XSGIFastInternAtom(__glutDisplay, - "WM_DELETE_WINDOW", SGI_XA_WM_DELETE_WINDOW, False); -} -#endif /* _WIN32 */ - -void -#ifdef OLD_VMS - __glutInitTime(struct timeval6 *beginning) -#else - __glutInitTime(struct timeval *beginning) -#endif -{ - static int beenhere = 0; -#ifdef OLD_VMS - static struct timeval6 genesis; -#else - static struct timeval genesis; -#endif - - if (!beenhere) { - GETTIMEOFDAY(&genesis); - beenhere = 1; - } - *beginning = genesis; -} - -static void -removeArgs(int *argcp, char **argv, int numToRemove) -{ - int i, j; - - for (i = 0, j = numToRemove; argv[j]; i++, j++) { - argv[i] = argv[j]; - } - argv[i] = NULL; - *argcp -= numToRemove; -} - -void GLUTAPIENTRY -glutInit(int *argcp, char **argv) -{ - char *display = NULL; - char *str, *geometry = NULL; -#ifdef OLD_VMS - struct timeval6 unused; -#else - struct timeval unused; -#endif - int i; - - if (__glutDisplay) { - __glutWarning("glutInit being called a second time."); - return; - } - /* Determine temporary program name. */ - str = strrchr(argv[0], '/'); - if (str == NULL) { - __glutProgramName = argv[0]; - } else { - __glutProgramName = str + 1; - } - - /* Make private copy of command line arguments. */ - __glutArgc = *argcp; - __glutArgv = (char **) malloc(__glutArgc * sizeof(char *)); - if (!__glutArgv) - __glutFatalError("out of memory."); - for (i = 0; i < __glutArgc; i++) { - __glutArgv[i] = __glutStrdup(argv[i]); - if (!__glutArgv[i]) - __glutFatalError("out of memory."); - } - - /* determine permanent program name */ - str = strrchr(__glutArgv[0], '/'); - if (str == NULL) { - __glutProgramName = __glutArgv[0]; - } else { - __glutProgramName = str + 1; - } - - /* parse arguments for standard options */ - for (i = 1; i < __glutArgc; i++) { - if (!strcmp(__glutArgv[i], "-display")) { -#if defined(_WIN32) - __glutWarning("-display option not supported by Win32 GLUT."); -#endif - if (++i >= __glutArgc) { - __glutFatalError( - "follow -display option with X display name."); - } - display = __glutArgv[i]; - removeArgs(argcp, &argv[1], 2); - } else if (!strcmp(__glutArgv[i], "-geometry")) { - if (++i >= __glutArgc) { - __glutFatalError( - "follow -geometry option with geometry parameter."); - } - geometry = __glutArgv[i]; - removeArgs(argcp, &argv[1], 2); - } else if (!strcmp(__glutArgv[i], "-direct")) { -#if defined(_WIN32) - __glutWarning("-direct option not supported by Win32 GLUT."); -#endif - if (!__glutTryDirect) - __glutFatalError( - "cannot force both direct and indirect rendering."); - __glutForceDirect = GL_TRUE; - removeArgs(argcp, &argv[1], 1); - } else if (!strcmp(__glutArgv[i], "-indirect")) { -#if defined(_WIN32) - __glutWarning("-indirect option not supported by Win32 GLUT."); -#endif - if (__glutForceDirect) - __glutFatalError( - "cannot force both direct and indirect rendering."); - __glutTryDirect = GL_FALSE; - removeArgs(argcp, &argv[1], 1); - } else if (!strcmp(__glutArgv[i], "-iconic")) { - __glutIconic = GL_TRUE; - removeArgs(argcp, &argv[1], 1); - } else if (!strcmp(__glutArgv[i], "-gldebug")) { - __glutDebug = GL_TRUE; - removeArgs(argcp, &argv[1], 1); - } else if (!strcmp(__glutArgv[i], "-sync")) { -#if defined(_WIN32) - __glutWarning("-sync option not supported by Win32 GLUT."); -#endif - synchronize = GL_TRUE; - removeArgs(argcp, &argv[1], 1); - } else { - /* Once unknown option encountered, stop command line - processing. */ - break; - } - } -#if defined(__OS2__) - __glutOpenOS2Connection(display); -#elif defined(_WIN32) - __glutOpenWin32Connection(display); -#else - __glutOpenXConnection(display); -#endif - if (geometry) { - int flags, x, y, width, height; - - /* Fix bogus "{width|height} may be used before set" - warning */ - width = 0; - height = 0; - - flags = XParseGeometry(geometry, &x, &y, - (unsigned int *) &width, (unsigned int *) &height); - if (WidthValue & flags) { - /* Careful because X does not allow zero or negative - width windows */ - if (width > 0) - __glutInitWidth = width; - } - if (HeightValue & flags) { - /* Careful because X does not allow zero or negative - height windows */ - if (height > 0) - __glutInitHeight = height; - } - glutInitWindowSize(__glutInitWidth, __glutInitHeight); - if (XValue & flags) { - if (XNegative & flags) - x = DisplayWidth(__glutDisplay, __glutScreen) + - x - __glutSizeHints.width; - /* Play safe: reject negative X locations */ - if (x >= 0) - __glutInitX = x; - } - if (YValue & flags) { - if (YNegative & flags) - y = DisplayHeight(__glutDisplay, __glutScreen) + - y - __glutSizeHints.height; - /* Play safe: reject negative Y locations */ - if (y >= 0) - __glutInitY = y; - } - glutInitWindowPosition(__glutInitX, __glutInitY); - } - __glutInitTime(&unused); - - /* check if GLUT_FPS env var is set */ - { - const char *fps = getenv("GLUT_FPS"); - if (fps) { - sscanf(fps, "%d", &__glutFPS); - if (__glutFPS <= 0) - __glutFPS = 5000; /* 5000 milliseconds */ - } - } -} - -#ifdef _WIN32 -void APIENTRY -__glutInitWithExit(int *argcp, char **argv, void (__cdecl *exitfunc)(int)) -{ - __glutExitFunc = exitfunc; - glutInit(argcp, argv); -} -#endif - -/* CENTRY */ -void GLUTAPIENTRY -glutInitWindowPosition(int x, int y) -{ - __glutInitX = x; - __glutInitY = y; - if (x >= 0 && y >= 0) { - __glutSizeHints.x = x; - __glutSizeHints.y = y; - __glutSizeHints.flags |= USPosition; - } else { - __glutSizeHints.flags &= ~USPosition; - } -} - -void GLUTAPIENTRY -glutInitWindowSize(int width, int height) -{ - __glutInitWidth = width; - __glutInitHeight = height; - if (width > 0 && height > 0) { - __glutSizeHints.width = width; - __glutSizeHints.height = height; - __glutSizeHints.flags |= USSize; - } else { - __glutSizeHints.flags &= ~USSize; - } -} - -void GLUTAPIENTRY -glutInitDisplayMode(unsigned int mask) -{ - __glutDisplayMode = mask; -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#ifdef __VMS +#include +#endif + +#include +#include +#include + +#if !defined(_WIN32) && !defined(__OS2__) + #include + #include +#endif + +/* SGI optimization introduced in IRIX 6.3 to avoid X server + round trips for interning common X atoms. */ +#if defined(_SGI_EXTRA_PREDEFINES) && !defined(NO_FAST_ATOMS) +#include +#else +#define XSGIFastInternAtom(dpy,string,fast_name,how) XInternAtom(dpy,string,how) +#endif + +#include "glutint.h" + +/* GLUT inter-file variables */ +/* *INDENT-OFF* */ +char *__glutProgramName = NULL; +int __glutArgc = 0; +char **__glutArgv = NULL; +char *__glutGeometry = NULL; +Display *__glutDisplay = NULL; +int __glutScreen; +Window __glutRoot; +int __glutScreenHeight; +int __glutScreenWidth; +GLboolean __glutIconic = GL_FALSE; +GLboolean __glutDebug = GL_FALSE; +unsigned int __glutDisplayMode = + GLUT_RGB | GLUT_SINGLE | GLUT_DEPTH; +char *__glutDisplayString = NULL; +int __glutConnectionFD; +XSizeHints __glutSizeHints = {0}; +int __glutInitWidth = 300, __glutInitHeight = 300; +int __glutInitX = -1, __glutInitY = -1; +GLboolean __glutForceDirect = GL_FALSE, + __glutTryDirect = GL_TRUE; +Atom __glutWMDeleteWindow; +/* *INDENT-ON* */ + +#ifdef _WIN32 +void (__cdecl *__glutExitFunc)(int retval) = NULL; +#endif + +static Bool synchronize = False; + +#if defined(__OS2__) + +MRESULT EXPENTRY GlutWindowProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ); +MRESULT EXPENTRY GlutWindowChildProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ); + + +void __glutOpenOS2Connection(char* display) +{ + static char *classname=NULL; +extern HAB hab; /* PM anchor block handle */ +ERRORID erridErrorCode;/* last error id code */ +int ii; + + /* Make sure we register the window only once. */ + if(classname) + return; + + classname = "GLUT"; + + if ( !WinRegisterClass( hab, /* PM anchor block handle */ + classname,/* window class name */ + GlutWindowProc,/* address of window procedure*/ + CS_SIZEREDRAW, /* |CS_SYNCPAINT size changes cause redrawing */ + 0UL ) ) /* window data */ + { erridErrorCode = WinGetLastError(hab); + ii = erridErrorCode; + return; + } + + classname = "GLUTCHILD"; + + if ( !WinRegisterClass( hab, /* PM anchor block handle */ + classname,/* window class name */ + GlutWindowChildProc,/* address of window procedure*/ + CS_SIZEREDRAW, /* size changes cause redrawing */ + 0UL ) ) /* window data */ + { erridErrorCode = WinGetLastError(hab); + ii = erridErrorCode; + return; + } + + __glutScreenWidth = GetSystemMetrics(SM_CXSCREEN); + __glutScreenHeight = GetSystemMetrics(SM_CYSCREEN); + + /* Set the root window to NULL because windows creates a top-level + window when the parent is NULL. X creates a top-level window + when the parent is the root window. */ + __glutRoot = NULLHANDLE; + + /* Set the display to 1 -- we shouldn't be using this anywhere + (except as an argument to X calls). */ + __glutDisplay = (Display*)1; + + /* There isn't any concept of multiple screens in Win32, therefore, + we don't need to keep track of the screen we're on... it's always + the same one. */ + __glutScreen = 0; +} + +#elif defined(_WIN32) + +#ifdef __BORLANDC__ +#include /* For masking floating point exceptions. */ +#endif + +void +__glutOpenWin32Connection(char* display) +{ + static char *classname; + WNDCLASS wc; + HINSTANCE hInstance = GetModuleHandle(NULL); + + /* Make sure we register the window only once. */ + if(classname) + return; + +#ifdef __BORLANDC__ + /* Under certain conditions (e.g. while rendering solid surfaces with + lighting enabled) Microsoft OpenGL libraries cause some illegal + operations like floating point overflow or division by zero. The + default behaviour of Microsoft compilers is to mask (ignore) + floating point exceptions, while Borland compilers do not. The + following function of Borland RTL allows to mask exceptions. + Advice from Pier Giorgio Esposito (mc2172@mclink.it). */ + _control87(MCW_EM,MCW_EM); +#endif + + classname = "GLUT"; + + /* Clear (important!) and then fill in the window class structure. */ + memset(&wc, 0, sizeof(WNDCLASS)); + wc.style = CS_OWNDC; + wc.lpfnWndProc = (WNDPROC)__glutWindowProc; + wc.hInstance = hInstance; + wc.hIcon = LoadIcon(hInstance, "GLUT_ICON"); + wc.hCursor = LoadCursor(hInstance, IDC_ARROW); + wc.hbrBackground = NULL; + wc.lpszMenuName = NULL; + wc.lpszClassName = classname; + + /* Fill in a default icon if one isn't specified as a resource. */ + if(!wc.hIcon) + wc.hIcon = LoadIcon(NULL, IDI_WINLOGO); + + if(!RegisterClass(&wc)) { + __glutFatalError("RegisterClass() failed:" + "Cannot register GLUT window class."); + } + + __glutScreenWidth = GetSystemMetrics(SM_CXSCREEN); + __glutScreenHeight = GetSystemMetrics(SM_CYSCREEN); + + /* Set the root window to NULL because windows creates a top-level + window when the parent is NULL. X creates a top-level window + when the parent is the root window. */ + __glutRoot = NULL; + + /* Set the display to 1 -- we shouldn't be using this anywhere + (except as an argument to X calls). */ + __glutDisplay = (Display*)1; + + /* There isn't any concept of multiple screens in Win32, therefore, + we don't need to keep track of the screen we're on... it's always + the same one. */ + __glutScreen = 0; +} +#else /* !_WIN32 */ +void +__glutOpenXConnection(char *display) +{ + int errorBase, eventBase; + + __glutDisplay = XOpenDisplay(display); + if (!__glutDisplay) + __glutFatalError("could not open display: %s", + XDisplayName(display)); + if (synchronize) + XSynchronize(__glutDisplay, True); + if (!glXQueryExtension(__glutDisplay, &errorBase, &eventBase)) + __glutFatalError( + "OpenGL GLX extension not supported by display: %s", + XDisplayName(display)); + __glutScreen = DefaultScreen(__glutDisplay); + __glutRoot = RootWindow(__glutDisplay, __glutScreen); + __glutScreenWidth = DisplayWidth(__glutDisplay, __glutScreen); + __glutScreenHeight = DisplayHeight(__glutDisplay, + __glutScreen); + __glutConnectionFD = ConnectionNumber(__glutDisplay); + __glutWMDeleteWindow = XSGIFastInternAtom(__glutDisplay, + "WM_DELETE_WINDOW", SGI_XA_WM_DELETE_WINDOW, False); +} +#endif /* _WIN32 */ + +void +#ifdef OLD_VMS + __glutInitTime(struct timeval6 *beginning) +#else + __glutInitTime(struct timeval *beginning) +#endif +{ + static int beenhere = 0; +#ifdef OLD_VMS + static struct timeval6 genesis; +#else + static struct timeval genesis; +#endif + + if (!beenhere) { + GETTIMEOFDAY(&genesis); + beenhere = 1; + } + *beginning = genesis; +} + +static void +removeArgs(int *argcp, char **argv, int numToRemove) +{ + int i, j; + + for (i = 0, j = numToRemove; argv[j]; i++, j++) { + argv[i] = argv[j]; + } + argv[i] = NULL; + *argcp -= numToRemove; +} + +void GLUTAPIENTRY +glutInit(int *argcp, char **argv) +{ + char *display = NULL; + char *str, *geometry = NULL; +#ifdef OLD_VMS + struct timeval6 unused; +#else + struct timeval unused; +#endif + int i; + + if (__glutDisplay) { + __glutWarning("glutInit being called a second time."); + return; + } + /* Determine temporary program name. */ + str = strrchr(argv[0], '/'); + if (str == NULL) { + __glutProgramName = argv[0]; + } else { + __glutProgramName = str + 1; + } + + /* Make private copy of command line arguments. */ + __glutArgc = *argcp; + __glutArgv = (char **) malloc(__glutArgc * sizeof(char *)); + if (!__glutArgv) + __glutFatalError("out of memory."); + for (i = 0; i < __glutArgc; i++) { + __glutArgv[i] = __glutStrdup(argv[i]); + if (!__glutArgv[i]) + __glutFatalError("out of memory."); + } + + /* determine permanent program name */ + str = strrchr(__glutArgv[0], '/'); + if (str == NULL) { + __glutProgramName = __glutArgv[0]; + } else { + __glutProgramName = str + 1; + } + + /* parse arguments for standard options */ + for (i = 1; i < __glutArgc; i++) { + if (!strcmp(__glutArgv[i], "-display")) { +#if defined(_WIN32) + __glutWarning("-display option not supported by Win32 GLUT."); +#endif + if (++i >= __glutArgc) { + __glutFatalError( + "follow -display option with X display name."); + } + display = __glutArgv[i]; + removeArgs(argcp, &argv[1], 2); + } else if (!strcmp(__glutArgv[i], "-geometry")) { + if (++i >= __glutArgc) { + __glutFatalError( + "follow -geometry option with geometry parameter."); + } + geometry = __glutArgv[i]; + removeArgs(argcp, &argv[1], 2); + } else if (!strcmp(__glutArgv[i], "-direct")) { +#if defined(_WIN32) + __glutWarning("-direct option not supported by Win32 GLUT."); +#endif + if (!__glutTryDirect) + __glutFatalError( + "cannot force both direct and indirect rendering."); + __glutForceDirect = GL_TRUE; + removeArgs(argcp, &argv[1], 1); + } else if (!strcmp(__glutArgv[i], "-indirect")) { +#if defined(_WIN32) + __glutWarning("-indirect option not supported by Win32 GLUT."); +#endif + if (__glutForceDirect) + __glutFatalError( + "cannot force both direct and indirect rendering."); + __glutTryDirect = GL_FALSE; + removeArgs(argcp, &argv[1], 1); + } else if (!strcmp(__glutArgv[i], "-iconic")) { + __glutIconic = GL_TRUE; + removeArgs(argcp, &argv[1], 1); + } else if (!strcmp(__glutArgv[i], "-gldebug")) { + __glutDebug = GL_TRUE; + removeArgs(argcp, &argv[1], 1); + } else if (!strcmp(__glutArgv[i], "-sync")) { +#if defined(_WIN32) + __glutWarning("-sync option not supported by Win32 GLUT."); +#endif + synchronize = GL_TRUE; + removeArgs(argcp, &argv[1], 1); + } else { + /* Once unknown option encountered, stop command line + processing. */ + break; + } + } +#if defined(__OS2__) + __glutOpenOS2Connection(display); +#elif defined(_WIN32) + __glutOpenWin32Connection(display); +#else + __glutOpenXConnection(display); +#endif + if (geometry) { + int flags, x, y, width, height; + + /* Fix bogus "{width|height} may be used before set" + warning */ + width = 0; + height = 0; + + flags = XParseGeometry(geometry, &x, &y, + (unsigned int *) &width, (unsigned int *) &height); + if (WidthValue & flags) { + /* Careful because X does not allow zero or negative + width windows */ + if (width > 0) + __glutInitWidth = width; + } + if (HeightValue & flags) { + /* Careful because X does not allow zero or negative + height windows */ + if (height > 0) + __glutInitHeight = height; + } + glutInitWindowSize(__glutInitWidth, __glutInitHeight); + if (XValue & flags) { + if (XNegative & flags) + x = DisplayWidth(__glutDisplay, __glutScreen) + + x - __glutSizeHints.width; + /* Play safe: reject negative X locations */ + if (x >= 0) + __glutInitX = x; + } + if (YValue & flags) { + if (YNegative & flags) + y = DisplayHeight(__glutDisplay, __glutScreen) + + y - __glutSizeHints.height; + /* Play safe: reject negative Y locations */ + if (y >= 0) + __glutInitY = y; + } + glutInitWindowPosition(__glutInitX, __glutInitY); + } + __glutInitTime(&unused); + + /* check if GLUT_FPS env var is set */ + { + const char *fps = getenv("GLUT_FPS"); + if (fps) { + sscanf(fps, "%d", &__glutFPS); + if (__glutFPS <= 0) + __glutFPS = 5000; /* 5000 milliseconds */ + } + } +} + +#ifdef _WIN32 +void APIENTRY +__glutInitWithExit(int *argcp, char **argv, void (__cdecl *exitfunc)(int)) +{ + __glutExitFunc = exitfunc; + glutInit(argcp, argv); +} +#endif + +/* CENTRY */ +void GLUTAPIENTRY +glutInitWindowPosition(int x, int y) +{ + __glutInitX = x; + __glutInitY = y; + if (x >= 0 && y >= 0) { + __glutSizeHints.x = x; + __glutSizeHints.y = y; + __glutSizeHints.flags |= USPosition; + } else { + __glutSizeHints.flags &= ~USPosition; + } +} + +void GLUTAPIENTRY +glutInitWindowSize(int width, int height) +{ + __glutInitWidth = width; + __glutInitHeight = height; + if (width > 0 && height > 0) { + __glutSizeHints.width = width; + __glutSizeHints.height = height; + __glutSizeHints.flags |= USSize; + } else { + __glutSizeHints.flags &= ~USSize; + } +} + +void GLUTAPIENTRY +glutInitDisplayMode(unsigned int mask) +{ + __glutDisplayMode = mask; +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_input.cpp b/src/glut/os2/glut_input.cpp index c517fe1249..e65b691cb1 100644 --- a/src/glut/os2/glut_input.cpp +++ b/src/glut/os2/glut_input.cpp @@ -1,628 +1,628 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1997, 1998. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include -#include -#include -#include - -#include "glutint.h" -#define POFIG 0 -#if POFIG - -int __glutNumDials = 0; -int __glutNumSpaceballButtons = 0; -int __glutNumButtonBoxButtons = 0; -int __glutNumTabletButtons = 0; -int __glutNumMouseButtons = 3; /* Good guess. */ -XDevice *__glutTablet = NULL; -XDevice *__glutDials = NULL; -XDevice *__glutSpaceball = NULL; - -int __glutHasJoystick = 0; -int __glutNumJoystickButtons = 0; -int __glutNumJoystickAxes = 0; - -#if !defined(_WIN32) -typedef struct _Range { - int min; - int range; -} Range; - -#define NUM_SPACEBALL_AXIS 6 -#define NUM_TABLET_AXIS 2 -#define NUM_DIALS_AXIS 8 - -Range __glutSpaceballRange[NUM_SPACEBALL_AXIS]; -Range __glutTabletRange[NUM_TABLET_AXIS]; -int *__glutDialsResolution; - -/* Safely assumes 0 is an illegal event type for X Input - extension events. */ -int __glutDeviceMotionNotify = 0; -int __glutDeviceButtonPress = 0; -int __glutDeviceButtonPressGrab = 0; -int __glutDeviceButtonRelease = 0; -int __glutDeviceStateNotify = 0; - -static int -normalizeTabletPos(int axis, int rawValue) -{ - assert(rawValue >= __glutTabletRange[axis].min); - assert(rawValue <= __glutTabletRange[axis].min - + __glutTabletRange[axis].range); - /* Normalize rawValue to between 0 and 4000. */ - return ((rawValue - __glutTabletRange[axis].min) * 4000) / - __glutTabletRange[axis].range; -} - -static int -normalizeDialAngle(int axis, int rawValue) -{ - /* XXX Assumption made that the resolution of the device is - number of clicks for one complete dial revolution. This - is true for SGI's dial & button box. */ - return (rawValue * 360.0) / __glutDialsResolution[axis]; -} - -static int -normalizeSpaceballAngle(int axis, int rawValue) -{ - assert(rawValue >= __glutSpaceballRange[axis].min); - assert(rawValue <= __glutSpaceballRange[axis].min + - __glutSpaceballRange[axis].range); - /* Normalize rawValue to between -1800 and 1800. */ - return ((rawValue - __glutSpaceballRange[axis].min) * 3600) / - __glutSpaceballRange[axis].range - 1800; -} - -static int -normalizeSpaceballDelta(int axis, int rawValue) -{ - assert(rawValue >= __glutSpaceballRange[axis].min); - assert(rawValue <= __glutSpaceballRange[axis].min + - __glutSpaceballRange[axis].range); - /* Normalize rawValue to between -1000 and 1000. */ - return ((rawValue - __glutSpaceballRange[axis].min) * 2000) / - __glutSpaceballRange[axis].range - 1000; -} - -static void -queryTabletPos(GLUTwindow * window) -{ - XDeviceState *state; - XInputClass *any; - XValuatorState *v; - int i; - - state = XQueryDeviceState(__glutDisplay, __glutTablet); - any = state->data; - for (i = 0; i < state->num_classes; i++) { -#if defined(__cplusplus) || defined(c_plusplus) - switch (any->c_class) { -#else - switch (any->class) { -#endif - case ValuatorClass: - v = (XValuatorState *) any; - if (v->num_valuators < 2) - goto end; - if (window->tabletPos[0] == -1) - window->tabletPos[0] = normalizeTabletPos(0, v->valuators[0]); - if (window->tabletPos[1] == -1) - window->tabletPos[1] = normalizeTabletPos(1, v->valuators[1]); - } - any = (XInputClass *) ((char *) any + any->length); - } -end: - XFreeDeviceState(state); -} - -static void -tabletPosChange(GLUTwindow * window, int first, int count, int *data) -{ - int i, value, genEvent = 0; - - for (i = first; i < first + count; i++) { - switch (i) { - case 0: /* X axis */ - case 1: /* Y axis */ - value = normalizeTabletPos(i, data[i - first]); - if (value != window->tabletPos[i]) { - window->tabletPos[i] = value; - genEvent = 1; - } - break; - } - } - if (window->tabletPos[0] == -1 || window->tabletPos[1] == -1) - queryTabletPos(window); - if (genEvent) - window->tabletMotion(window->tabletPos[0], window->tabletPos[1]); -} -#endif /* !_WIN32 */ - -static int -__glutProcessDeviceEvents(XEvent * event) -{ -#if !defined(_WIN32) - GLUTwindow *window; - - /* XXX Ugly code fan out. */ - - /* Can't use switch/case since X Input event types are - dynamic. */ - - if (__glutDeviceMotionNotify && event->type == __glutDeviceMotionNotify) { - XDeviceMotionEvent *devmot = (XDeviceMotionEvent *) event; - - window = __glutGetWindow(devmot->window); - if (window) { - if (__glutTablet - && devmot->deviceid == __glutTablet->device_id - && window->tabletMotion) { - tabletPosChange(window, devmot->first_axis, devmot->axes_count, - devmot->axis_data); - } else if (__glutDials - && devmot->deviceid == __glutDials->device_id - && window->dials) { - int i, first = devmot->first_axis, count = devmot->axes_count; - - for (i = first; i < first + count; i++) - window->dials(i + 1, - normalizeDialAngle(i, devmot->axis_data[i - first])); - } else if (__glutSpaceball - && devmot->deviceid == __glutSpaceball->device_id) { - /* XXX Assume that space ball motion events come in as - all the first 6 axes. Assume first 3 axes are XYZ - translations; second 3 axes are XYZ rotations. */ - if (devmot->first_axis == 0 && devmot->axes_count == 6) { - if (window->spaceMotion) - window->spaceMotion( - normalizeSpaceballDelta(0, devmot->axis_data[0]), - normalizeSpaceballDelta(1, devmot->axis_data[1]), - normalizeSpaceballDelta(2, devmot->axis_data[2])); - if (window->spaceRotate) - window->spaceRotate( - normalizeSpaceballAngle(3, devmot->axis_data[3]), - normalizeSpaceballAngle(4, devmot->axis_data[4]), - normalizeSpaceballAngle(5, devmot->axis_data[5])); - } - } - return 1; - } - } else if (__glutDeviceButtonPress - && event->type == __glutDeviceButtonPress) { - XDeviceButtonEvent *devbtn = (XDeviceButtonEvent *) event; - - window = __glutGetWindow(devbtn->window); - if (window) { - if (__glutTablet - && devbtn->deviceid == __glutTablet->device_id - && window->tabletButton - && devbtn->first_axis == 0 - && devbtn->axes_count == 2) { - tabletPosChange(window, devbtn->first_axis, devbtn->axes_count, - devbtn->axis_data); - window->tabletButton(devbtn->button, GLUT_DOWN, - window->tabletPos[0], window->tabletPos[1]); - } else if (__glutDials - && devbtn->deviceid == __glutDials->device_id - && window->buttonBox) { - window->buttonBox(devbtn->button, GLUT_DOWN); - } else if (__glutSpaceball - && devbtn->deviceid == __glutSpaceball->device_id - && window->spaceButton) { - window->spaceButton(devbtn->button, GLUT_DOWN); - } - return 1; - } - } else if (__glutDeviceButtonRelease - && event->type == __glutDeviceButtonRelease) { - XDeviceButtonEvent *devbtn = (XDeviceButtonEvent *) event; - - window = __glutGetWindow(devbtn->window); - if (window) { - if (__glutTablet - && devbtn->deviceid == __glutTablet->device_id - && window->tabletButton - && devbtn->first_axis == 0 - && devbtn->axes_count == 2) { - tabletPosChange(window, devbtn->first_axis, devbtn->axes_count, - devbtn->axis_data); - window->tabletButton(devbtn->button, GLUT_UP, - window->tabletPos[0], window->tabletPos[1]); - } else if (__glutDials - && devbtn->deviceid == __glutDials->device_id - && window->buttonBox) { - window->buttonBox(devbtn->button, GLUT_UP); - } else if (__glutSpaceball - && devbtn->deviceid == __glutSpaceball->device_id - && window->spaceButton) { - window->spaceButton(devbtn->button, GLUT_UP); - } - return 1; - } - } -#else - { - JOYINFOEX info; - JOYCAPS joyCaps; - - memset(&info, 0, sizeof(JOYINFOEX)); - info.dwSize = sizeof(JOYINFOEX); - info.dwFlags = JOY_RETURNALL; - - if (joyGetPosEx(JOYSTICKID1,&info) != JOYERR_NOERROR) { - __glutHasJoystick = 1; - joyGetDevCaps(JOYSTICKID1, &joyCaps, sizeof(joyCaps)); - __glutNumJoystickButtons = joyCaps.wNumButtons; - __glutNumJoystickAxes = joyCaps.wNumAxes; - } else { - __glutHasJoystick = 0; - __glutNumJoystickButtons = 0; - __glutNumJoystickAxes = 0; - } - } -#endif /* !_WIN32 */ - return 0; -} - -static GLUTeventParser eventParser = -{__glutProcessDeviceEvents, NULL}; - -static void -addDeviceEventParser(void) -{ - static Bool been_here = False; - - if (been_here) - return; - been_here = True; - __glutRegisterEventParser(&eventParser); -} - -static int -probeDevices(void) -{ - static Bool been_here = False; - static int support; -#if !defined(_WIN32) - XExtensionVersion *version; - XDeviceInfoPtr device_info, device; - XAnyClassPtr any; - XButtonInfoPtr b; - XValuatorInfoPtr v; - XAxisInfoPtr a; - int num_dev = 0, btns = 0, dials = 0; - int i, j, k; -#endif /* !_WIN32 */ - - if (been_here) { - return support; - } - been_here = True; - -#if !defined(_WIN32) - version = XGetExtensionVersion(__glutDisplay, "XInputExtension"); - /* Ugh. XInput extension API forces annoying cast of a pointer - to a long so it can be compared with the NoSuchExtension - value (#defined to 1). */ - if (version == NULL || ((long) version) == NoSuchExtension) { - support = 0; - return support; - } - XFree(version); - device_info = XListInputDevices(__glutDisplay, &num_dev); - if (device_info) { - for (i = 0; i < num_dev; i++) { - /* XXX These are SGI names for these devices; - unfortunately, no good standard exists for standard - types of X input extension devices. */ - - device = &device_info[i]; - any = (XAnyClassPtr) device->inputclassinfo; - - if (!__glutSpaceball && !strcmp(device->name, "spaceball")) { - v = NULL; - b = NULL; - for (j = 0; j < device->num_classes; j++) { -#if defined(__cplusplus) || defined(c_plusplus) - switch (any->c_class) { -#else - switch (any->class) { -#endif - case ButtonClass: - b = (XButtonInfoPtr) any; - btns = b->num_buttons; - break; - case ValuatorClass: - v = (XValuatorInfoPtr) any; - /* Sanity check: at least 6 valuators? */ - if (v->num_axes < NUM_SPACEBALL_AXIS) - goto skip_device; - a = (XAxisInfoPtr) ((char *) v + sizeof(XValuatorInfo)); - for (k = 0; k < NUM_SPACEBALL_AXIS; k++, a++) { - __glutSpaceballRange[k].min = a->min_value; - __glutSpaceballRange[k].range = a->max_value - a->min_value; - } - break; - } - any = (XAnyClassPtr) ((char *) any + any->length); - } - if (v) { - __glutSpaceball = XOpenDevice(__glutDisplay, device->id); - if (__glutSpaceball) { - __glutNumSpaceballButtons = btns; - addDeviceEventParser(); - } - } - } else if (!__glutDials && !strcmp(device->name, "dial+buttons")) { - v = NULL; - b = NULL; - for (j = 0; j < device->num_classes; j++) { -#if defined(__cplusplus) || defined(c_plusplus) - switch (any->c_class) { -#else - switch (any->class) { -#endif - case ButtonClass: - b = (XButtonInfoPtr) any; - btns = b->num_buttons; - break; - case ValuatorClass: - v = (XValuatorInfoPtr) any; - /* Sanity check: at least 8 valuators? */ - if (v->num_axes < NUM_DIALS_AXIS) - goto skip_device; - dials = v->num_axes; - __glutDialsResolution = (int *) malloc(sizeof(int) * dials); - a = (XAxisInfoPtr) ((char *) v + sizeof(XValuatorInfo)); - for (k = 0; k < dials; k++, a++) { - __glutDialsResolution[k] = a->resolution; - } - break; - } - any = (XAnyClassPtr) ((char *) any + any->length); - } - if (v) { - __glutDials = XOpenDevice(__glutDisplay, device->id); - if (__glutDials) { - __glutNumButtonBoxButtons = btns; - __glutNumDials = dials; - addDeviceEventParser(); - } - } - } else if (!__glutTablet && !strcmp(device->name, "tablet")) { - v = NULL; - b = NULL; - for (j = 0; j < device->num_classes; j++) { -#if defined(__cplusplus) || defined(c_plusplus) - switch (any->c_class) { -#else - switch (any->class) { -#endif - case ButtonClass: - b = (XButtonInfoPtr) any; - btns = b->num_buttons; - break; - case ValuatorClass: - v = (XValuatorInfoPtr) any; - /* Sanity check: exactly 2 valuators? */ - if (v->num_axes != NUM_TABLET_AXIS) - goto skip_device; - a = (XAxisInfoPtr) ((char *) v + sizeof(XValuatorInfo)); - for (k = 0; k < NUM_TABLET_AXIS; k++, a++) { - __glutTabletRange[k].min = a->min_value; - __glutTabletRange[k].range = a->max_value - a->min_value; - } - break; - } - any = (XAnyClassPtr) ((char *) any + any->length); - } - if (v) { - __glutTablet = XOpenDevice(__glutDisplay, device->id); - if (__glutTablet) { - __glutNumTabletButtons = btns; - addDeviceEventParser(); - } - } - } else if (!strcmp(device->name, "mouse")) { - for (j = 0; j < device->num_classes; j++) { -#if defined(__cplusplus) || defined(c_plusplus) - if (any->c_class == ButtonClass) { -#else - if (any->class == ButtonClass) { -#endif - b = (XButtonInfoPtr) any; - __glutNumMouseButtons = b->num_buttons; - } - any = (XAnyClassPtr) ((char *) any + any->length); - } - } - skip_device:; - } - XFreeDeviceList(device_info); - } -#else /* _WIN32 */ - __glutNumMouseButtons = GetSystemMetrics(SM_CMOUSEBUTTONS); -#endif /* !_WIN32 */ - /* X Input extension might be supported, but only if there is - a tablet, dials, or spaceball do we claim devices are - supported. */ - support = __glutTablet || __glutDials || __glutSpaceball; - return support; -} - -void -__glutUpdateInputDeviceMask(GLUTwindow * window) -{ -#if !defined(_WIN32) - /* 5 (dial and buttons) + 5 (tablet locator and buttons) + 5 - (Spaceball buttons and axis) = 15 */ - XEventClass eventList[15]; - int rc, numEvents; - - rc = probeDevices(); - if (rc) { - numEvents = 0; - if (__glutTablet) { - if (window->tabletMotion) { - DeviceMotionNotify(__glutTablet, __glutDeviceMotionNotify, - eventList[numEvents]); - numEvents++; - } - if (window->tabletButton) { - DeviceButtonPress(__glutTablet, __glutDeviceButtonPress, - eventList[numEvents]); - numEvents++; - DeviceButtonPressGrab(__glutTablet, __glutDeviceButtonPressGrab, - eventList[numEvents]); - numEvents++; - DeviceButtonRelease(__glutTablet, __glutDeviceButtonRelease, - eventList[numEvents]); - numEvents++; - } - if (window->tabletMotion || window->tabletButton) { - DeviceStateNotify(__glutTablet, __glutDeviceStateNotify, - eventList[numEvents]); - numEvents++; - } - } - if (__glutDials) { - if (window->dials) { - DeviceMotionNotify(__glutDials, __glutDeviceMotionNotify, - eventList[numEvents]); - numEvents++; - } - if (window->buttonBox) { - DeviceButtonPress(__glutDials, __glutDeviceButtonPress, - eventList[numEvents]); - numEvents++; - DeviceButtonPressGrab(__glutDials, __glutDeviceButtonPressGrab, - eventList[numEvents]); - numEvents++; - DeviceButtonRelease(__glutDials, __glutDeviceButtonRelease, - eventList[numEvents]); - numEvents++; - } - if (window->dials || window->buttonBox) { - DeviceStateNotify(__glutDials, __glutDeviceStateNotify, - eventList[numEvents]); - numEvents++; - } - } - if (__glutSpaceball) { - if (window->spaceMotion || window->spaceRotate) { - DeviceMotionNotify(__glutSpaceball, __glutDeviceMotionNotify, - eventList[numEvents]); - numEvents++; - } - if (window->spaceButton) { - DeviceButtonPress(__glutSpaceball, __glutDeviceButtonPress, - eventList[numEvents]); - numEvents++; - DeviceButtonPressGrab(__glutSpaceball, __glutDeviceButtonPressGrab, - eventList[numEvents]); - numEvents++; - DeviceButtonRelease(__glutSpaceball, __glutDeviceButtonRelease, - eventList[numEvents]); - numEvents++; - } - if (window->spaceMotion || window->spaceRotate || window->spaceButton) { - DeviceStateNotify(__glutSpaceball, __glutDeviceStateNotify, - eventList[numEvents]); - numEvents++; - } - } -#if 0 - if (window->children) { - GLUTwindow *child = window->children; - - do { - XChangeDeviceDontPropagateList(__glutDisplay, child->win, - numEvents, eventList, AddToList); - child = child->siblings; - } while (child); - } -#endif - XSelectExtensionEvent(__glutDisplay, window->win, - eventList, numEvents); - if (window->overlay) { - XSelectExtensionEvent(__glutDisplay, window->overlay->win, - eventList, numEvents); - } - } else { - /* X Input extension not supported; no chance for exotic - input devices. */ - } -#endif /* !_WIN32 */ -} - -#endif //POFIG - -/* CENTRY */ -int GLUTAPIENTRY -glutDeviceGet(GLenum param) -{ -#if POFIG - probeDevices(); -#endif - switch (param) { - case GLUT_HAS_KEYBOARD: - case GLUT_HAS_MOUSE: - /* Assume window system always has mouse and keyboard. */ - return 1; -#if POFIG - case GLUT_HAS_SPACEBALL: - return __glutSpaceball != NULL; - case GLUT_HAS_DIAL_AND_BUTTON_BOX: - return __glutDials != NULL; - case GLUT_HAS_TABLET: - return __glutTablet != NULL; - case GLUT_NUM_MOUSE_BUTTONS: - return __glutNumMouseButtons; - case GLUT_NUM_SPACEBALL_BUTTONS: - return __glutNumSpaceballButtons; - case GLUT_NUM_BUTTON_BOX_BUTTONS: - return __glutNumButtonBoxButtons; - case GLUT_NUM_DIALS: - return __glutNumDials; - case GLUT_NUM_TABLET_BUTTONS: - return __glutNumTabletButtons; - case GLUT_DEVICE_IGNORE_KEY_REPEAT: - return __glutCurrentWindow->ignoreKeyRepeat; -#ifndef _WIN32 - case GLUT_DEVICE_KEY_REPEAT: - { - XKeyboardState state; - - XGetKeyboardControl(__glutDisplay, &state); - return state.global_auto_repeat; - } - case GLUT_JOYSTICK_POLL_RATE: - return 0; -#else - case GLUT_DEVICE_KEY_REPEAT: - /* Win32 cannot globally disable key repeat. */ - return GLUT_KEY_REPEAT_ON; - case GLUT_JOYSTICK_POLL_RATE: - return __glutCurrentWindow->joyPollInterval; -#endif - case GLUT_HAS_JOYSTICK: - return __glutHasJoystick; - case GLUT_JOYSTICK_BUTTONS: - return __glutNumJoystickButtons; - case GLUT_JOYSTICK_AXES: - return __glutNumJoystickAxes; -#endif //POFIG - default: - __glutWarning("invalid glutDeviceGet parameter: %d", param); - return -1; - } -} -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994, 1997, 1998. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include +#include +#include +#include + +#include "glutint.h" +#define POFIG 0 +#if POFIG + +int __glutNumDials = 0; +int __glutNumSpaceballButtons = 0; +int __glutNumButtonBoxButtons = 0; +int __glutNumTabletButtons = 0; +int __glutNumMouseButtons = 3; /* Good guess. */ +XDevice *__glutTablet = NULL; +XDevice *__glutDials = NULL; +XDevice *__glutSpaceball = NULL; + +int __glutHasJoystick = 0; +int __glutNumJoystickButtons = 0; +int __glutNumJoystickAxes = 0; + +#if !defined(_WIN32) +typedef struct _Range { + int min; + int range; +} Range; + +#define NUM_SPACEBALL_AXIS 6 +#define NUM_TABLET_AXIS 2 +#define NUM_DIALS_AXIS 8 + +Range __glutSpaceballRange[NUM_SPACEBALL_AXIS]; +Range __glutTabletRange[NUM_TABLET_AXIS]; +int *__glutDialsResolution; + +/* Safely assumes 0 is an illegal event type for X Input + extension events. */ +int __glutDeviceMotionNotify = 0; +int __glutDeviceButtonPress = 0; +int __glutDeviceButtonPressGrab = 0; +int __glutDeviceButtonRelease = 0; +int __glutDeviceStateNotify = 0; + +static int +normalizeTabletPos(int axis, int rawValue) +{ + assert(rawValue >= __glutTabletRange[axis].min); + assert(rawValue <= __glutTabletRange[axis].min + + __glutTabletRange[axis].range); + /* Normalize rawValue to between 0 and 4000. */ + return ((rawValue - __glutTabletRange[axis].min) * 4000) / + __glutTabletRange[axis].range; +} + +static int +normalizeDialAngle(int axis, int rawValue) +{ + /* XXX Assumption made that the resolution of the device is + number of clicks for one complete dial revolution. This + is true for SGI's dial & button box. */ + return (rawValue * 360.0) / __glutDialsResolution[axis]; +} + +static int +normalizeSpaceballAngle(int axis, int rawValue) +{ + assert(rawValue >= __glutSpaceballRange[axis].min); + assert(rawValue <= __glutSpaceballRange[axis].min + + __glutSpaceballRange[axis].range); + /* Normalize rawValue to between -1800 and 1800. */ + return ((rawValue - __glutSpaceballRange[axis].min) * 3600) / + __glutSpaceballRange[axis].range - 1800; +} + +static int +normalizeSpaceballDelta(int axis, int rawValue) +{ + assert(rawValue >= __glutSpaceballRange[axis].min); + assert(rawValue <= __glutSpaceballRange[axis].min + + __glutSpaceballRange[axis].range); + /* Normalize rawValue to between -1000 and 1000. */ + return ((rawValue - __glutSpaceballRange[axis].min) * 2000) / + __glutSpaceballRange[axis].range - 1000; +} + +static void +queryTabletPos(GLUTwindow * window) +{ + XDeviceState *state; + XInputClass *any; + XValuatorState *v; + int i; + + state = XQueryDeviceState(__glutDisplay, __glutTablet); + any = state->data; + for (i = 0; i < state->num_classes; i++) { +#if defined(__cplusplus) || defined(c_plusplus) + switch (any->c_class) { +#else + switch (any->class) { +#endif + case ValuatorClass: + v = (XValuatorState *) any; + if (v->num_valuators < 2) + goto end; + if (window->tabletPos[0] == -1) + window->tabletPos[0] = normalizeTabletPos(0, v->valuators[0]); + if (window->tabletPos[1] == -1) + window->tabletPos[1] = normalizeTabletPos(1, v->valuators[1]); + } + any = (XInputClass *) ((char *) any + any->length); + } +end: + XFreeDeviceState(state); +} + +static void +tabletPosChange(GLUTwindow * window, int first, int count, int *data) +{ + int i, value, genEvent = 0; + + for (i = first; i < first + count; i++) { + switch (i) { + case 0: /* X axis */ + case 1: /* Y axis */ + value = normalizeTabletPos(i, data[i - first]); + if (value != window->tabletPos[i]) { + window->tabletPos[i] = value; + genEvent = 1; + } + break; + } + } + if (window->tabletPos[0] == -1 || window->tabletPos[1] == -1) + queryTabletPos(window); + if (genEvent) + window->tabletMotion(window->tabletPos[0], window->tabletPos[1]); +} +#endif /* !_WIN32 */ + +static int +__glutProcessDeviceEvents(XEvent * event) +{ +#if !defined(_WIN32) + GLUTwindow *window; + + /* XXX Ugly code fan out. */ + + /* Can't use switch/case since X Input event types are + dynamic. */ + + if (__glutDeviceMotionNotify && event->type == __glutDeviceMotionNotify) { + XDeviceMotionEvent *devmot = (XDeviceMotionEvent *) event; + + window = __glutGetWindow(devmot->window); + if (window) { + if (__glutTablet + && devmot->deviceid == __glutTablet->device_id + && window->tabletMotion) { + tabletPosChange(window, devmot->first_axis, devmot->axes_count, + devmot->axis_data); + } else if (__glutDials + && devmot->deviceid == __glutDials->device_id + && window->dials) { + int i, first = devmot->first_axis, count = devmot->axes_count; + + for (i = first; i < first + count; i++) + window->dials(i + 1, + normalizeDialAngle(i, devmot->axis_data[i - first])); + } else if (__glutSpaceball + && devmot->deviceid == __glutSpaceball->device_id) { + /* XXX Assume that space ball motion events come in as + all the first 6 axes. Assume first 3 axes are XYZ + translations; second 3 axes are XYZ rotations. */ + if (devmot->first_axis == 0 && devmot->axes_count == 6) { + if (window->spaceMotion) + window->spaceMotion( + normalizeSpaceballDelta(0, devmot->axis_data[0]), + normalizeSpaceballDelta(1, devmot->axis_data[1]), + normalizeSpaceballDelta(2, devmot->axis_data[2])); + if (window->spaceRotate) + window->spaceRotate( + normalizeSpaceballAngle(3, devmot->axis_data[3]), + normalizeSpaceballAngle(4, devmot->axis_data[4]), + normalizeSpaceballAngle(5, devmot->axis_data[5])); + } + } + return 1; + } + } else if (__glutDeviceButtonPress + && event->type == __glutDeviceButtonPress) { + XDeviceButtonEvent *devbtn = (XDeviceButtonEvent *) event; + + window = __glutGetWindow(devbtn->window); + if (window) { + if (__glutTablet + && devbtn->deviceid == __glutTablet->device_id + && window->tabletButton + && devbtn->first_axis == 0 + && devbtn->axes_count == 2) { + tabletPosChange(window, devbtn->first_axis, devbtn->axes_count, + devbtn->axis_data); + window->tabletButton(devbtn->button, GLUT_DOWN, + window->tabletPos[0], window->tabletPos[1]); + } else if (__glutDials + && devbtn->deviceid == __glutDials->device_id + && window->buttonBox) { + window->buttonBox(devbtn->button, GLUT_DOWN); + } else if (__glutSpaceball + && devbtn->deviceid == __glutSpaceball->device_id + && window->spaceButton) { + window->spaceButton(devbtn->button, GLUT_DOWN); + } + return 1; + } + } else if (__glutDeviceButtonRelease + && event->type == __glutDeviceButtonRelease) { + XDeviceButtonEvent *devbtn = (XDeviceButtonEvent *) event; + + window = __glutGetWindow(devbtn->window); + if (window) { + if (__glutTablet + && devbtn->deviceid == __glutTablet->device_id + && window->tabletButton + && devbtn->first_axis == 0 + && devbtn->axes_count == 2) { + tabletPosChange(window, devbtn->first_axis, devbtn->axes_count, + devbtn->axis_data); + window->tabletButton(devbtn->button, GLUT_UP, + window->tabletPos[0], window->tabletPos[1]); + } else if (__glutDials + && devbtn->deviceid == __glutDials->device_id + && window->buttonBox) { + window->buttonBox(devbtn->button, GLUT_UP); + } else if (__glutSpaceball + && devbtn->deviceid == __glutSpaceball->device_id + && window->spaceButton) { + window->spaceButton(devbtn->button, GLUT_UP); + } + return 1; + } + } +#else + { + JOYINFOEX info; + JOYCAPS joyCaps; + + memset(&info, 0, sizeof(JOYINFOEX)); + info.dwSize = sizeof(JOYINFOEX); + info.dwFlags = JOY_RETURNALL; + + if (joyGetPosEx(JOYSTICKID1,&info) != JOYERR_NOERROR) { + __glutHasJoystick = 1; + joyGetDevCaps(JOYSTICKID1, &joyCaps, sizeof(joyCaps)); + __glutNumJoystickButtons = joyCaps.wNumButtons; + __glutNumJoystickAxes = joyCaps.wNumAxes; + } else { + __glutHasJoystick = 0; + __glutNumJoystickButtons = 0; + __glutNumJoystickAxes = 0; + } + } +#endif /* !_WIN32 */ + return 0; +} + +static GLUTeventParser eventParser = +{__glutProcessDeviceEvents, NULL}; + +static void +addDeviceEventParser(void) +{ + static Bool been_here = False; + + if (been_here) + return; + been_here = True; + __glutRegisterEventParser(&eventParser); +} + +static int +probeDevices(void) +{ + static Bool been_here = False; + static int support; +#if !defined(_WIN32) + XExtensionVersion *version; + XDeviceInfoPtr device_info, device; + XAnyClassPtr any; + XButtonInfoPtr b; + XValuatorInfoPtr v; + XAxisInfoPtr a; + int num_dev = 0, btns = 0, dials = 0; + int i, j, k; +#endif /* !_WIN32 */ + + if (been_here) { + return support; + } + been_here = True; + +#if !defined(_WIN32) + version = XGetExtensionVersion(__glutDisplay, "XInputExtension"); + /* Ugh. XInput extension API forces annoying cast of a pointer + to a long so it can be compared with the NoSuchExtension + value (#defined to 1). */ + if (version == NULL || ((long) version) == NoSuchExtension) { + support = 0; + return support; + } + XFree(version); + device_info = XListInputDevices(__glutDisplay, &num_dev); + if (device_info) { + for (i = 0; i < num_dev; i++) { + /* XXX These are SGI names for these devices; + unfortunately, no good standard exists for standard + types of X input extension devices. */ + + device = &device_info[i]; + any = (XAnyClassPtr) device->inputclassinfo; + + if (!__glutSpaceball && !strcmp(device->name, "spaceball")) { + v = NULL; + b = NULL; + for (j = 0; j < device->num_classes; j++) { +#if defined(__cplusplus) || defined(c_plusplus) + switch (any->c_class) { +#else + switch (any->class) { +#endif + case ButtonClass: + b = (XButtonInfoPtr) any; + btns = b->num_buttons; + break; + case ValuatorClass: + v = (XValuatorInfoPtr) any; + /* Sanity check: at least 6 valuators? */ + if (v->num_axes < NUM_SPACEBALL_AXIS) + goto skip_device; + a = (XAxisInfoPtr) ((char *) v + sizeof(XValuatorInfo)); + for (k = 0; k < NUM_SPACEBALL_AXIS; k++, a++) { + __glutSpaceballRange[k].min = a->min_value; + __glutSpaceballRange[k].range = a->max_value - a->min_value; + } + break; + } + any = (XAnyClassPtr) ((char *) any + any->length); + } + if (v) { + __glutSpaceball = XOpenDevice(__glutDisplay, device->id); + if (__glutSpaceball) { + __glutNumSpaceballButtons = btns; + addDeviceEventParser(); + } + } + } else if (!__glutDials && !strcmp(device->name, "dial+buttons")) { + v = NULL; + b = NULL; + for (j = 0; j < device->num_classes; j++) { +#if defined(__cplusplus) || defined(c_plusplus) + switch (any->c_class) { +#else + switch (any->class) { +#endif + case ButtonClass: + b = (XButtonInfoPtr) any; + btns = b->num_buttons; + break; + case ValuatorClass: + v = (XValuatorInfoPtr) any; + /* Sanity check: at least 8 valuators? */ + if (v->num_axes < NUM_DIALS_AXIS) + goto skip_device; + dials = v->num_axes; + __glutDialsResolution = (int *) malloc(sizeof(int) * dials); + a = (XAxisInfoPtr) ((char *) v + sizeof(XValuatorInfo)); + for (k = 0; k < dials; k++, a++) { + __glutDialsResolution[k] = a->resolution; + } + break; + } + any = (XAnyClassPtr) ((char *) any + any->length); + } + if (v) { + __glutDials = XOpenDevice(__glutDisplay, device->id); + if (__glutDials) { + __glutNumButtonBoxButtons = btns; + __glutNumDials = dials; + addDeviceEventParser(); + } + } + } else if (!__glutTablet && !strcmp(device->name, "tablet")) { + v = NULL; + b = NULL; + for (j = 0; j < device->num_classes; j++) { +#if defined(__cplusplus) || defined(c_plusplus) + switch (any->c_class) { +#else + switch (any->class) { +#endif + case ButtonClass: + b = (XButtonInfoPtr) any; + btns = b->num_buttons; + break; + case ValuatorClass: + v = (XValuatorInfoPtr) any; + /* Sanity check: exactly 2 valuators? */ + if (v->num_axes != NUM_TABLET_AXIS) + goto skip_device; + a = (XAxisInfoPtr) ((char *) v + sizeof(XValuatorInfo)); + for (k = 0; k < NUM_TABLET_AXIS; k++, a++) { + __glutTabletRange[k].min = a->min_value; + __glutTabletRange[k].range = a->max_value - a->min_value; + } + break; + } + any = (XAnyClassPtr) ((char *) any + any->length); + } + if (v) { + __glutTablet = XOpenDevice(__glutDisplay, device->id); + if (__glutTablet) { + __glutNumTabletButtons = btns; + addDeviceEventParser(); + } + } + } else if (!strcmp(device->name, "mouse")) { + for (j = 0; j < device->num_classes; j++) { +#if defined(__cplusplus) || defined(c_plusplus) + if (any->c_class == ButtonClass) { +#else + if (any->class == ButtonClass) { +#endif + b = (XButtonInfoPtr) any; + __glutNumMouseButtons = b->num_buttons; + } + any = (XAnyClassPtr) ((char *) any + any->length); + } + } + skip_device:; + } + XFreeDeviceList(device_info); + } +#else /* _WIN32 */ + __glutNumMouseButtons = GetSystemMetrics(SM_CMOUSEBUTTONS); +#endif /* !_WIN32 */ + /* X Input extension might be supported, but only if there is + a tablet, dials, or spaceball do we claim devices are + supported. */ + support = __glutTablet || __glutDials || __glutSpaceball; + return support; +} + +void +__glutUpdateInputDeviceMask(GLUTwindow * window) +{ +#if !defined(_WIN32) + /* 5 (dial and buttons) + 5 (tablet locator and buttons) + 5 + (Spaceball buttons and axis) = 15 */ + XEventClass eventList[15]; + int rc, numEvents; + + rc = probeDevices(); + if (rc) { + numEvents = 0; + if (__glutTablet) { + if (window->tabletMotion) { + DeviceMotionNotify(__glutTablet, __glutDeviceMotionNotify, + eventList[numEvents]); + numEvents++; + } + if (window->tabletButton) { + DeviceButtonPress(__glutTablet, __glutDeviceButtonPress, + eventList[numEvents]); + numEvents++; + DeviceButtonPressGrab(__glutTablet, __glutDeviceButtonPressGrab, + eventList[numEvents]); + numEvents++; + DeviceButtonRelease(__glutTablet, __glutDeviceButtonRelease, + eventList[numEvents]); + numEvents++; + } + if (window->tabletMotion || window->tabletButton) { + DeviceStateNotify(__glutTablet, __glutDeviceStateNotify, + eventList[numEvents]); + numEvents++; + } + } + if (__glutDials) { + if (window->dials) { + DeviceMotionNotify(__glutDials, __glutDeviceMotionNotify, + eventList[numEvents]); + numEvents++; + } + if (window->buttonBox) { + DeviceButtonPress(__glutDials, __glutDeviceButtonPress, + eventList[numEvents]); + numEvents++; + DeviceButtonPressGrab(__glutDials, __glutDeviceButtonPressGrab, + eventList[numEvents]); + numEvents++; + DeviceButtonRelease(__glutDials, __glutDeviceButtonRelease, + eventList[numEvents]); + numEvents++; + } + if (window->dials || window->buttonBox) { + DeviceStateNotify(__glutDials, __glutDeviceStateNotify, + eventList[numEvents]); + numEvents++; + } + } + if (__glutSpaceball) { + if (window->spaceMotion || window->spaceRotate) { + DeviceMotionNotify(__glutSpaceball, __glutDeviceMotionNotify, + eventList[numEvents]); + numEvents++; + } + if (window->spaceButton) { + DeviceButtonPress(__glutSpaceball, __glutDeviceButtonPress, + eventList[numEvents]); + numEvents++; + DeviceButtonPressGrab(__glutSpaceball, __glutDeviceButtonPressGrab, + eventList[numEvents]); + numEvents++; + DeviceButtonRelease(__glutSpaceball, __glutDeviceButtonRelease, + eventList[numEvents]); + numEvents++; + } + if (window->spaceMotion || window->spaceRotate || window->spaceButton) { + DeviceStateNotify(__glutSpaceball, __glutDeviceStateNotify, + eventList[numEvents]); + numEvents++; + } + } +#if 0 + if (window->children) { + GLUTwindow *child = window->children; + + do { + XChangeDeviceDontPropagateList(__glutDisplay, child->win, + numEvents, eventList, AddToList); + child = child->siblings; + } while (child); + } +#endif + XSelectExtensionEvent(__glutDisplay, window->win, + eventList, numEvents); + if (window->overlay) { + XSelectExtensionEvent(__glutDisplay, window->overlay->win, + eventList, numEvents); + } + } else { + /* X Input extension not supported; no chance for exotic + input devices. */ + } +#endif /* !_WIN32 */ +} + +#endif //POFIG + +/* CENTRY */ +int GLUTAPIENTRY +glutDeviceGet(GLenum param) +{ +#if POFIG + probeDevices(); +#endif + switch (param) { + case GLUT_HAS_KEYBOARD: + case GLUT_HAS_MOUSE: + /* Assume window system always has mouse and keyboard. */ + return 1; +#if POFIG + case GLUT_HAS_SPACEBALL: + return __glutSpaceball != NULL; + case GLUT_HAS_DIAL_AND_BUTTON_BOX: + return __glutDials != NULL; + case GLUT_HAS_TABLET: + return __glutTablet != NULL; + case GLUT_NUM_MOUSE_BUTTONS: + return __glutNumMouseButtons; + case GLUT_NUM_SPACEBALL_BUTTONS: + return __glutNumSpaceballButtons; + case GLUT_NUM_BUTTON_BOX_BUTTONS: + return __glutNumButtonBoxButtons; + case GLUT_NUM_DIALS: + return __glutNumDials; + case GLUT_NUM_TABLET_BUTTONS: + return __glutNumTabletButtons; + case GLUT_DEVICE_IGNORE_KEY_REPEAT: + return __glutCurrentWindow->ignoreKeyRepeat; +#ifndef _WIN32 + case GLUT_DEVICE_KEY_REPEAT: + { + XKeyboardState state; + + XGetKeyboardControl(__glutDisplay, &state); + return state.global_auto_repeat; + } + case GLUT_JOYSTICK_POLL_RATE: + return 0; +#else + case GLUT_DEVICE_KEY_REPEAT: + /* Win32 cannot globally disable key repeat. */ + return GLUT_KEY_REPEAT_ON; + case GLUT_JOYSTICK_POLL_RATE: + return __glutCurrentWindow->joyPollInterval; +#endif + case GLUT_HAS_JOYSTICK: + return __glutHasJoystick; + case GLUT_JOYSTICK_BUTTONS: + return __glutNumJoystickButtons; + case GLUT_JOYSTICK_AXES: + return __glutNumJoystickAxes; +#endif //POFIG + default: + __glutWarning("invalid glutDeviceGet parameter: %d", param); + return -1; + } +} +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_key.cpp b/src/glut/os2/glut_key.cpp index a1b9919abc..3e1f7e52c1 100644 --- a/src/glut/os2/glut_key.cpp +++ b/src/glut/os2/glut_key.cpp @@ -1,29 +1,29 @@ - -/* Copyright (c) Mark J. Kilgard, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include - -#include "glutint.h" - -/* CENTRY */ -void GLUTAPIENTRY -glutKeyboardFunc(GLUTkeyboardCB keyboardFunc) -{ - __glutChangeWindowEventMask(KeyPressMask, - keyboardFunc != NULL || __glutCurrentWindow->special != NULL); - __glutCurrentWindow->keyboard = keyboardFunc; -} - -void GLUTAPIENTRY -glutSpecialFunc(GLUTspecialCB specialFunc) -{ - __glutChangeWindowEventMask(KeyPressMask, - specialFunc != NULL || __glutCurrentWindow->keyboard != NULL); - __glutCurrentWindow->special = specialFunc; -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include + +#include "glutint.h" + +/* CENTRY */ +void GLUTAPIENTRY +glutKeyboardFunc(GLUTkeyboardCB keyboardFunc) +{ + __glutChangeWindowEventMask(KeyPressMask, + keyboardFunc != NULL || __glutCurrentWindow->special != NULL); + __glutCurrentWindow->keyboard = keyboardFunc; +} + +void GLUTAPIENTRY +glutSpecialFunc(GLUTspecialCB specialFunc) +{ + __glutChangeWindowEventMask(KeyPressMask, + specialFunc != NULL || __glutCurrentWindow->keyboard != NULL); + __glutCurrentWindow->special = specialFunc; +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_keyctrl.cpp b/src/glut/os2/glut_keyctrl.cpp index b0f3a28f7d..ddac630509 100644 --- a/src/glut/os2/glut_keyctrl.cpp +++ b/src/glut/os2/glut_keyctrl.cpp @@ -1,29 +1,29 @@ - -/* Copyright (c) Mark J. Kilgard, 1996, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include "glutint.h" - -/* CENTRY */ -void GLUTAPIENTRY -glutIgnoreKeyRepeat(int ignore) -{ - __glutCurrentWindow->ignoreKeyRepeat = ignore; -} - -void GLUTAPIENTRY -glutSetKeyRepeat(int repeatMode) -{ -#if !defined(_WIN32) && !defined(__OS2PM__) - XKeyboardControl values; - - /* GLUT's repeatMode #define's match the Xlib API values. */ - values.auto_repeat_mode = repeatMode; - XChangeKeyboardControl(__glutDisplay, KBAutoRepeatMode, &values); -#endif -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1996, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include "glutint.h" + +/* CENTRY */ +void GLUTAPIENTRY +glutIgnoreKeyRepeat(int ignore) +{ + __glutCurrentWindow->ignoreKeyRepeat = ignore; +} + +void GLUTAPIENTRY +glutSetKeyRepeat(int repeatMode) +{ +#if !defined(_WIN32) && !defined(__OS2PM__) + XKeyboardControl values; + + /* GLUT's repeatMode #define's match the Xlib API values. */ + values.auto_repeat_mode = repeatMode; + XChangeKeyboardControl(__glutDisplay, KBAutoRepeatMode, &values); +#endif +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_keyup.cpp b/src/glut/os2/glut_keyup.cpp index ada758e277..c33770f1be 100644 --- a/src/glut/os2/glut_keyup.cpp +++ b/src/glut/os2/glut_keyup.cpp @@ -1,29 +1,29 @@ - -/* Copyright (c) Mark J. Kilgard, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include - -#include "glutint.h" - -/* CENTRY */ -void GLUTAPIENTRY -glutKeyboardUpFunc(GLUTkeyboardCB keyboardUpFunc) -{ - __glutChangeWindowEventMask(KeyReleaseMask, - keyboardUpFunc != NULL || __glutCurrentWindow->specialUp != NULL); - __glutCurrentWindow->keyboardUp = keyboardUpFunc; -} - -void GLUTAPIENTRY -glutSpecialUpFunc(GLUTspecialCB specialUpFunc) -{ - __glutChangeWindowEventMask(KeyReleaseMask, - specialUpFunc != NULL || __glutCurrentWindow->keyboardUp != NULL); - __glutCurrentWindow->specialUp = specialUpFunc; -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include + +#include "glutint.h" + +/* CENTRY */ +void GLUTAPIENTRY +glutKeyboardUpFunc(GLUTkeyboardCB keyboardUpFunc) +{ + __glutChangeWindowEventMask(KeyReleaseMask, + keyboardUpFunc != NULL || __glutCurrentWindow->specialUp != NULL); + __glutCurrentWindow->keyboardUp = keyboardUpFunc; +} + +void GLUTAPIENTRY +glutSpecialUpFunc(GLUTspecialCB specialUpFunc) +{ + __glutChangeWindowEventMask(KeyReleaseMask, + specialUpFunc != NULL || __glutCurrentWindow->keyboardUp != NULL); + __glutCurrentWindow->specialUp = specialUpFunc; +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_mesa.cpp b/src/glut/os2/glut_mesa.cpp index 73682a5067..7e202fa078 100644 --- a/src/glut/os2/glut_mesa.cpp +++ b/src/glut/os2/glut_mesa.cpp @@ -1,57 +1,57 @@ - -/* Copyright (c) Mark J. Kilgard, 1996. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include -#include -#include "glutint.h" - -int __glutMesaSwapHackSupport = 0; /* Not supported until - proven otherwise. */ - -/* Use the "Mesa swap hack" if reasonable if and only if - MESA_SWAP_HACK is set to something whose first character is - not "N" or "n" AND "Brian Paul" is the vendor string AND - "Mesa X11"* (or "Mesa" for backward compatibility) is the - renderer string. - - Anyone who modifies Mesa so that glXSwapBuffers does not - simply blit the previously rendered back buffer should - change either their vendor or renderer string to avoid - confusing GLUT. */ - -void -__glutDetermineMesaSwapHackSupport(void) -{ - static int doneAlready = 0; - char *env, *vendor, *renderer; - - if (doneAlready) - return; - env = getenv("MESA_SWAP_HACK"); - if (env) { - if ((env[0] != 'n') && (env[0] != 'N')) { - vendor = (char *) glGetString(GL_VENDOR); - renderer = (char *) glGetString(GL_RENDERER); - - /* Old versions of X11 Mesa uses the renderer string - "Mesa"; Brian plans to start using "Mesa X11" to - distinguish the X version of Mesa from other flavor - such as Windows or 3Dfx. */ - -#define MESA_X11 "Mesa X11" - - /* XXX At some point in the future, eliminate the - backward compatibility for the old "Mesa" renderer - string. */ - - if (!strcmp(vendor, "Brian Paul") && (!strcmp(renderer, "Mesa") || - !strncmp(renderer, MESA_X11, sizeof(MESA_X11) - 1))) - __glutMesaSwapHackSupport = 1; - } - } - doneAlready = 1; -} + +/* Copyright (c) Mark J. Kilgard, 1996. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include +#include +#include "glutint.h" + +int __glutMesaSwapHackSupport = 0; /* Not supported until + proven otherwise. */ + +/* Use the "Mesa swap hack" if reasonable if and only if + MESA_SWAP_HACK is set to something whose first character is + not "N" or "n" AND "Brian Paul" is the vendor string AND + "Mesa X11"* (or "Mesa" for backward compatibility) is the + renderer string. + + Anyone who modifies Mesa so that glXSwapBuffers does not + simply blit the previously rendered back buffer should + change either their vendor or renderer string to avoid + confusing GLUT. */ + +void +__glutDetermineMesaSwapHackSupport(void) +{ + static int doneAlready = 0; + char *env, *vendor, *renderer; + + if (doneAlready) + return; + env = getenv("MESA_SWAP_HACK"); + if (env) { + if ((env[0] != 'n') && (env[0] != 'N')) { + vendor = (char *) glGetString(GL_VENDOR); + renderer = (char *) glGetString(GL_RENDERER); + + /* Old versions of X11 Mesa uses the renderer string + "Mesa"; Brian plans to start using "Mesa X11" to + distinguish the X version of Mesa from other flavor + such as Windows or 3Dfx. */ + +#define MESA_X11 "Mesa X11" + + /* XXX At some point in the future, eliminate the + backward compatibility for the old "Mesa" renderer + string. */ + + if (!strcmp(vendor, "Brian Paul") && (!strcmp(renderer, "Mesa") || + !strncmp(renderer, MESA_X11, sizeof(MESA_X11) - 1))) + __glutMesaSwapHackSupport = 1; + } + } + doneAlready = 1; +} diff --git a/src/glut/os2/glut_modifier.cpp b/src/glut/os2/glut_modifier.cpp index 0ffa5a6801..7e0e67a21f 100644 --- a/src/glut/os2/glut_modifier.cpp +++ b/src/glut/os2/glut_modifier.cpp @@ -1,31 +1,31 @@ - -/* Copyright (c) Mark J. Kilgard, 1994. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include "glutint.h" - -/* CENTRY */ -int GLUTAPIENTRY -glutGetModifiers(void) -{ - int modifiers; - - if(__glutModifierMask == (unsigned int) ~0) { - __glutWarning( - "glutCurrentModifiers: do not call outside core input callback."); - return 0; - } - modifiers = 0; - if(__glutModifierMask & (ShiftMask|LockMask)) - modifiers |= GLUT_ACTIVE_SHIFT; - if(__glutModifierMask & ControlMask) - modifiers |= GLUT_ACTIVE_CTRL; - if(__glutModifierMask & Mod1Mask) - modifiers |= GLUT_ACTIVE_ALT; - return modifiers; -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include "glutint.h" + +/* CENTRY */ +int GLUTAPIENTRY +glutGetModifiers(void) +{ + int modifiers; + + if(__glutModifierMask == (unsigned int) ~0) { + __glutWarning( + "glutCurrentModifiers: do not call outside core input callback."); + return 0; + } + modifiers = 0; + if(__glutModifierMask & (ShiftMask|LockMask)) + modifiers |= GLUT_ACTIVE_SHIFT; + if(__glutModifierMask & ControlMask) + modifiers |= GLUT_ACTIVE_CTRL; + if(__glutModifierMask & Mod1Mask) + modifiers |= GLUT_ACTIVE_ALT; + return modifiers; +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_roman.cpp b/src/glut/os2/glut_roman.cpp index 8e25de8ac1..af2b4ec956 100644 --- a/src/glut/os2/glut_roman.cpp +++ b/src/glut/os2/glut_roman.cpp @@ -1,2451 +1,2451 @@ - -/* GENERATED FILE -- DO NOT MODIFY */ - -#include "glutstroke.h" - -/* char: 33 '!' */ - -static const CoordRec char33_stroke0[] = { - { 13.3819, 100 }, - { 13.3819, 33.3333 }, -}; - -static const CoordRec char33_stroke1[] = { - { 13.3819, 9.5238 }, - { 8.62, 4.7619 }, - { 13.3819, 0 }, - { 18.1438, 4.7619 }, - { 13.3819, 9.5238 }, -}; - -static const StrokeRec char33[] = { - { 2, char33_stroke0 }, - { 5, char33_stroke1 }, -}; - -/* char: 34 '"' */ - -static const CoordRec char34_stroke0[] = { - { 4.02, 100 }, - { 4.02, 66.6667 }, -}; - -static const CoordRec char34_stroke1[] = { - { 42.1152, 100 }, - { 42.1152, 66.6667 }, -}; - -static const StrokeRec char34[] = { - { 2, char34_stroke0 }, - { 2, char34_stroke1 }, -}; - -/* char: 35 '#' */ - -static const CoordRec char35_stroke0[] = { - { 41.2952, 119.048 }, - { 7.9619, -33.3333 }, -}; - -static const CoordRec char35_stroke1[] = { - { 69.8667, 119.048 }, - { 36.5333, -33.3333 }, -}; - -static const CoordRec char35_stroke2[] = { - { 7.9619, 57.1429 }, - { 74.6286, 57.1429 }, -}; - -static const CoordRec char35_stroke3[] = { - { 3.2, 28.5714 }, - { 69.8667, 28.5714 }, -}; - -static const StrokeRec char35[] = { - { 2, char35_stroke0 }, - { 2, char35_stroke1 }, - { 2, char35_stroke2 }, - { 2, char35_stroke3 }, -}; - -/* char: 36 '$' */ - -static const CoordRec char36_stroke0[] = { - { 28.6295, 119.048 }, - { 28.6295, -19.0476 }, -}; - -static const CoordRec char36_stroke1[] = { - { 47.6771, 119.048 }, - { 47.6771, -19.0476 }, -}; - -static const CoordRec char36_stroke2[] = { - { 71.4867, 85.7143 }, - { 61.9629, 95.2381 }, - { 47.6771, 100 }, - { 28.6295, 100 }, - { 14.3438, 95.2381 }, - { 4.82, 85.7143 }, - { 4.82, 76.1905 }, - { 9.5819, 66.6667 }, - { 14.3438, 61.9048 }, - { 23.8676, 57.1429 }, - { 52.439, 47.619 }, - { 61.9629, 42.8571 }, - { 66.7248, 38.0952 }, - { 71.4867, 28.5714 }, - { 71.4867, 14.2857 }, - { 61.9629, 4.7619 }, - { 47.6771, 0 }, - { 28.6295, 0 }, - { 14.3438, 4.7619 }, - { 4.82, 14.2857 }, -}; - -static const StrokeRec char36[] = { - { 2, char36_stroke0 }, - { 2, char36_stroke1 }, - { 20, char36_stroke2 }, -}; - -/* char: 37 '%' */ - -static const CoordRec char37_stroke0[] = { - { 92.0743, 100 }, - { 6.36, 0 }, -}; - -static const CoordRec char37_stroke1[] = { - { 30.1695, 100 }, - { 39.6933, 90.4762 }, - { 39.6933, 80.9524 }, - { 34.9314, 71.4286 }, - { 25.4076, 66.6667 }, - { 15.8838, 66.6667 }, - { 6.36, 76.1905 }, - { 6.36, 85.7143 }, - { 11.1219, 95.2381 }, - { 20.6457, 100 }, - { 30.1695, 100 }, - { 39.6933, 95.2381 }, - { 53.979, 90.4762 }, - { 68.2648, 90.4762 }, - { 82.5505, 95.2381 }, - { 92.0743, 100 }, -}; - -static const CoordRec char37_stroke2[] = { - { 73.0267, 33.3333 }, - { 63.5029, 28.5714 }, - { 58.741, 19.0476 }, - { 58.741, 9.5238 }, - { 68.2648, 0 }, - { 77.7886, 0 }, - { 87.3124, 4.7619 }, - { 92.0743, 14.2857 }, - { 92.0743, 23.8095 }, - { 82.5505, 33.3333 }, - { 73.0267, 33.3333 }, -}; - -static const StrokeRec char37[] = { - { 2, char37_stroke0 }, - { 16, char37_stroke1 }, - { 11, char37_stroke2 }, -}; - -/* char: 38 '&' */ - -static const CoordRec char38_stroke0[] = { - { 101.218, 57.1429 }, - { 101.218, 61.9048 }, - { 96.4562, 66.6667 }, - { 91.6943, 66.6667 }, - { 86.9324, 61.9048 }, - { 82.1705, 52.381 }, - { 72.6467, 28.5714 }, - { 63.1229, 14.2857 }, - { 53.599, 4.7619 }, - { 44.0752, 0 }, - { 25.0276, 0 }, - { 15.5038, 4.7619 }, - { 10.7419, 9.5238 }, - { 5.98, 19.0476 }, - { 5.98, 28.5714 }, - { 10.7419, 38.0952 }, - { 15.5038, 42.8571 }, - { 48.8371, 61.9048 }, - { 53.599, 66.6667 }, - { 58.361, 76.1905 }, - { 58.361, 85.7143 }, - { 53.599, 95.2381 }, - { 44.0752, 100 }, - { 34.5514, 95.2381 }, - { 29.7895, 85.7143 }, - { 29.7895, 76.1905 }, - { 34.5514, 61.9048 }, - { 44.0752, 47.619 }, - { 67.8848, 14.2857 }, - { 77.4086, 4.7619 }, - { 86.9324, 0 }, - { 96.4562, 0 }, - { 101.218, 4.7619 }, - { 101.218, 9.5238 }, -}; - -static const StrokeRec char38[] = { - { 34, char38_stroke0 }, -}; - -/* char: 39 ''' */ - -static const CoordRec char39_stroke0[] = { - { 4.44, 100 }, - { 4.44, 66.6667 }, -}; - -static const StrokeRec char39[] = { - { 2, char39_stroke0 }, -}; - -/* char: 40 '(' */ - -static const CoordRec char40_stroke0[] = { - { 40.9133, 119.048 }, - { 31.3895, 109.524 }, - { 21.8657, 95.2381 }, - { 12.3419, 76.1905 }, - { 7.58, 52.381 }, - { 7.58, 33.3333 }, - { 12.3419, 9.5238 }, - { 21.8657, -9.5238 }, - { 31.3895, -23.8095 }, - { 40.9133, -33.3333 }, -}; - -static const StrokeRec char40[] = { - { 10, char40_stroke0 }, -}; - -/* char: 41 ')' */ - -static const CoordRec char41_stroke0[] = { - { 5.28, 119.048 }, - { 14.8038, 109.524 }, - { 24.3276, 95.2381 }, - { 33.8514, 76.1905 }, - { 38.6133, 52.381 }, - { 38.6133, 33.3333 }, - { 33.8514, 9.5238 }, - { 24.3276, -9.5238 }, - { 14.8038, -23.8095 }, - { 5.28, -33.3333 }, -}; - -static const StrokeRec char41[] = { - { 10, char41_stroke0 }, -}; - -/* char: 42 '*' */ - -static const CoordRec char42_stroke0[] = { - { 30.7695, 71.4286 }, - { 30.7695, 14.2857 }, -}; - -static const CoordRec char42_stroke1[] = { - { 6.96, 57.1429 }, - { 54.579, 28.5714 }, -}; - -static const CoordRec char42_stroke2[] = { - { 54.579, 57.1429 }, - { 6.96, 28.5714 }, -}; - -static const StrokeRec char42[] = { - { 2, char42_stroke0 }, - { 2, char42_stroke1 }, - { 2, char42_stroke2 }, -}; - -/* char: 43 '+' */ - -static const CoordRec char43_stroke0[] = { - { 48.8371, 85.7143 }, - { 48.8371, 0 }, -}; - -static const CoordRec char43_stroke1[] = { - { 5.98, 42.8571 }, - { 91.6943, 42.8571 }, -}; - -static const StrokeRec char43[] = { - { 2, char43_stroke0 }, - { 2, char43_stroke1 }, -}; - -/* char: 44 ',' */ - -static const CoordRec char44_stroke0[] = { - { 18.2838, 4.7619 }, - { 13.5219, 0 }, - { 8.76, 4.7619 }, - { 13.5219, 9.5238 }, - { 18.2838, 4.7619 }, - { 18.2838, -4.7619 }, - { 13.5219, -14.2857 }, - { 8.76, -19.0476 }, -}; - -static const StrokeRec char44[] = { - { 8, char44_stroke0 }, -}; - -/* char: 45 '-' */ - -static const CoordRec char45_stroke0[] = { - { 7.38, 42.8571 }, - { 93.0943, 42.8571 }, -}; - -static const StrokeRec char45[] = { - { 2, char45_stroke0 }, -}; - -/* char: 46 '.' */ - -static const CoordRec char46_stroke0[] = { - { 13.1019, 9.5238 }, - { 8.34, 4.7619 }, - { 13.1019, 0 }, - { 17.8638, 4.7619 }, - { 13.1019, 9.5238 }, -}; - -static const StrokeRec char46[] = { - { 5, char46_stroke0 }, -}; - -/* char: 47 '/' */ - -static const CoordRec char47_stroke0[] = { - { 7.24, -14.2857 }, - { 73.9067, 100 }, -}; - -static const StrokeRec char47[] = { - { 2, char47_stroke0 }, -}; - -/* char: 48 '0' */ - -static const CoordRec char48_stroke0[] = { - { 33.5514, 100 }, - { 19.2657, 95.2381 }, - { 9.7419, 80.9524 }, - { 4.98, 57.1429 }, - { 4.98, 42.8571 }, - { 9.7419, 19.0476 }, - { 19.2657, 4.7619 }, - { 33.5514, 0 }, - { 43.0752, 0 }, - { 57.361, 4.7619 }, - { 66.8848, 19.0476 }, - { 71.6467, 42.8571 }, - { 71.6467, 57.1429 }, - { 66.8848, 80.9524 }, - { 57.361, 95.2381 }, - { 43.0752, 100 }, - { 33.5514, 100 }, -}; - -static const StrokeRec char48[] = { - { 17, char48_stroke0 }, -}; - -/* char: 49 '1' */ - -static const CoordRec char49_stroke0[] = { - { 11.82, 80.9524 }, - { 21.3438, 85.7143 }, - { 35.6295, 100 }, - { 35.6295, 0 }, -}; - -static const StrokeRec char49[] = { - { 4, char49_stroke0 }, -}; - -/* char: 50 '2' */ - -static const CoordRec char50_stroke0[] = { - { 10.1819, 76.1905 }, - { 10.1819, 80.9524 }, - { 14.9438, 90.4762 }, - { 19.7057, 95.2381 }, - { 29.2295, 100 }, - { 48.2771, 100 }, - { 57.801, 95.2381 }, - { 62.5629, 90.4762 }, - { 67.3248, 80.9524 }, - { 67.3248, 71.4286 }, - { 62.5629, 61.9048 }, - { 53.039, 47.619 }, - { 5.42, 0 }, - { 72.0867, 0 }, -}; - -static const StrokeRec char50[] = { - { 14, char50_stroke0 }, -}; - -/* char: 51 '3' */ - -static const CoordRec char51_stroke0[] = { - { 14.5238, 100 }, - { 66.9048, 100 }, - { 38.3333, 61.9048 }, - { 52.619, 61.9048 }, - { 62.1429, 57.1429 }, - { 66.9048, 52.381 }, - { 71.6667, 38.0952 }, - { 71.6667, 28.5714 }, - { 66.9048, 14.2857 }, - { 57.381, 4.7619 }, - { 43.0952, 0 }, - { 28.8095, 0 }, - { 14.5238, 4.7619 }, - { 9.7619, 9.5238 }, - { 5, 19.0476 }, -}; - -static const StrokeRec char51[] = { - { 15, char51_stroke0 }, -}; - -/* char: 52 '4' */ - -static const CoordRec char52_stroke0[] = { - { 51.499, 100 }, - { 3.88, 33.3333 }, - { 75.3086, 33.3333 }, -}; - -static const CoordRec char52_stroke1[] = { - { 51.499, 100 }, - { 51.499, 0 }, -}; - -static const StrokeRec char52[] = { - { 3, char52_stroke0 }, - { 2, char52_stroke1 }, -}; - -/* char: 53 '5' */ - -static const CoordRec char53_stroke0[] = { - { 62.0029, 100 }, - { 14.3838, 100 }, - { 9.6219, 57.1429 }, - { 14.3838, 61.9048 }, - { 28.6695, 66.6667 }, - { 42.9552, 66.6667 }, - { 57.241, 61.9048 }, - { 66.7648, 52.381 }, - { 71.5267, 38.0952 }, - { 71.5267, 28.5714 }, - { 66.7648, 14.2857 }, - { 57.241, 4.7619 }, - { 42.9552, 0 }, - { 28.6695, 0 }, - { 14.3838, 4.7619 }, - { 9.6219, 9.5238 }, - { 4.86, 19.0476 }, -}; - -static const StrokeRec char53[] = { - { 17, char53_stroke0 }, -}; - -/* char: 54 '6' */ - -static const CoordRec char54_stroke0[] = { - { 62.7229, 85.7143 }, - { 57.961, 95.2381 }, - { 43.6752, 100 }, - { 34.1514, 100 }, - { 19.8657, 95.2381 }, - { 10.3419, 80.9524 }, - { 5.58, 57.1429 }, - { 5.58, 33.3333 }, - { 10.3419, 14.2857 }, - { 19.8657, 4.7619 }, - { 34.1514, 0 }, - { 38.9133, 0 }, - { 53.199, 4.7619 }, - { 62.7229, 14.2857 }, - { 67.4848, 28.5714 }, - { 67.4848, 33.3333 }, - { 62.7229, 47.619 }, - { 53.199, 57.1429 }, - { 38.9133, 61.9048 }, - { 34.1514, 61.9048 }, - { 19.8657, 57.1429 }, - { 10.3419, 47.619 }, - { 5.58, 33.3333 }, -}; - -static const StrokeRec char54[] = { - { 23, char54_stroke0 }, -}; - -/* char: 55 '7' */ - -static const CoordRec char55_stroke0[] = { - { 72.2267, 100 }, - { 24.6076, 0 }, -}; - -static const CoordRec char55_stroke1[] = { - { 5.56, 100 }, - { 72.2267, 100 }, -}; - -static const StrokeRec char55[] = { - { 2, char55_stroke0 }, - { 2, char55_stroke1 }, -}; - -/* char: 56 '8' */ - -static const CoordRec char56_stroke0[] = { - { 29.4095, 100 }, - { 15.1238, 95.2381 }, - { 10.3619, 85.7143 }, - { 10.3619, 76.1905 }, - { 15.1238, 66.6667 }, - { 24.6476, 61.9048 }, - { 43.6952, 57.1429 }, - { 57.981, 52.381 }, - { 67.5048, 42.8571 }, - { 72.2667, 33.3333 }, - { 72.2667, 19.0476 }, - { 67.5048, 9.5238 }, - { 62.7429, 4.7619 }, - { 48.4571, 0 }, - { 29.4095, 0 }, - { 15.1238, 4.7619 }, - { 10.3619, 9.5238 }, - { 5.6, 19.0476 }, - { 5.6, 33.3333 }, - { 10.3619, 42.8571 }, - { 19.8857, 52.381 }, - { 34.1714, 57.1429 }, - { 53.219, 61.9048 }, - { 62.7429, 66.6667 }, - { 67.5048, 76.1905 }, - { 67.5048, 85.7143 }, - { 62.7429, 95.2381 }, - { 48.4571, 100 }, - { 29.4095, 100 }, -}; - -static const StrokeRec char56[] = { - { 29, char56_stroke0 }, -}; - -/* char: 57 '9' */ - -static const CoordRec char57_stroke0[] = { - { 68.5048, 66.6667 }, - { 63.7429, 52.381 }, - { 54.219, 42.8571 }, - { 39.9333, 38.0952 }, - { 35.1714, 38.0952 }, - { 20.8857, 42.8571 }, - { 11.3619, 52.381 }, - { 6.6, 66.6667 }, - { 6.6, 71.4286 }, - { 11.3619, 85.7143 }, - { 20.8857, 95.2381 }, - { 35.1714, 100 }, - { 39.9333, 100 }, - { 54.219, 95.2381 }, - { 63.7429, 85.7143 }, - { 68.5048, 66.6667 }, - { 68.5048, 42.8571 }, - { 63.7429, 19.0476 }, - { 54.219, 4.7619 }, - { 39.9333, 0 }, - { 30.4095, 0 }, - { 16.1238, 4.7619 }, - { 11.3619, 14.2857 }, -}; - -static const StrokeRec char57[] = { - { 23, char57_stroke0 }, -}; - -/* char: 58 ':' */ - -static const CoordRec char58_stroke0[] = { - { 14.0819, 66.6667 }, - { 9.32, 61.9048 }, - { 14.0819, 57.1429 }, - { 18.8438, 61.9048 }, - { 14.0819, 66.6667 }, -}; - -static const CoordRec char58_stroke1[] = { - { 14.0819, 9.5238 }, - { 9.32, 4.7619 }, - { 14.0819, 0 }, - { 18.8438, 4.7619 }, - { 14.0819, 9.5238 }, -}; - -static const StrokeRec char58[] = { - { 5, char58_stroke0 }, - { 5, char58_stroke1 }, -}; - -/* char: 59 ';' */ - -static const CoordRec char59_stroke0[] = { - { 12.9619, 66.6667 }, - { 8.2, 61.9048 }, - { 12.9619, 57.1429 }, - { 17.7238, 61.9048 }, - { 12.9619, 66.6667 }, -}; - -static const CoordRec char59_stroke1[] = { - { 17.7238, 4.7619 }, - { 12.9619, 0 }, - { 8.2, 4.7619 }, - { 12.9619, 9.5238 }, - { 17.7238, 4.7619 }, - { 17.7238, -4.7619 }, - { 12.9619, -14.2857 }, - { 8.2, -19.0476 }, -}; - -static const StrokeRec char59[] = { - { 5, char59_stroke0 }, - { 8, char59_stroke1 }, -}; - -/* char: 60 '<' */ - -static const CoordRec char60_stroke0[] = { - { 79.2505, 85.7143 }, - { 3.06, 42.8571 }, - { 79.2505, 0 }, -}; - -static const StrokeRec char60[] = { - { 3, char60_stroke0 }, -}; - -/* char: 61 '=' */ - -static const CoordRec char61_stroke0[] = { - { 5.7, 57.1429 }, - { 91.4143, 57.1429 }, -}; - -static const CoordRec char61_stroke1[] = { - { 5.7, 28.5714 }, - { 91.4143, 28.5714 }, -}; - -static const StrokeRec char61[] = { - { 2, char61_stroke0 }, - { 2, char61_stroke1 }, -}; - -/* char: 62 '>' */ - -static const CoordRec char62_stroke0[] = { - { 2.78, 85.7143 }, - { 78.9705, 42.8571 }, - { 2.78, 0 }, -}; - -static const StrokeRec char62[] = { - { 3, char62_stroke0 }, -}; - -/* char: 63 '?' */ - -static const CoordRec char63_stroke0[] = { - { 8.42, 76.1905 }, - { 8.42, 80.9524 }, - { 13.1819, 90.4762 }, - { 17.9438, 95.2381 }, - { 27.4676, 100 }, - { 46.5152, 100 }, - { 56.039, 95.2381 }, - { 60.801, 90.4762 }, - { 65.5629, 80.9524 }, - { 65.5629, 71.4286 }, - { 60.801, 61.9048 }, - { 56.039, 57.1429 }, - { 36.9914, 47.619 }, - { 36.9914, 33.3333 }, -}; - -static const CoordRec char63_stroke1[] = { - { 36.9914, 9.5238 }, - { 32.2295, 4.7619 }, - { 36.9914, 0 }, - { 41.7533, 4.7619 }, - { 36.9914, 9.5238 }, -}; - -static const StrokeRec char63[] = { - { 14, char63_stroke0 }, - { 5, char63_stroke1 }, -}; - -/* char: 64 '@' */ - -static const CoordRec char64_stroke0[] = { - { 49.2171, 52.381 }, - { 39.6933, 57.1429 }, - { 30.1695, 57.1429 }, - { 25.4076, 47.619 }, - { 25.4076, 42.8571 }, - { 30.1695, 33.3333 }, - { 39.6933, 33.3333 }, - { 49.2171, 38.0952 }, -}; - -static const CoordRec char64_stroke1[] = { - { 49.2171, 57.1429 }, - { 49.2171, 38.0952 }, - { 53.979, 33.3333 }, - { 63.5029, 33.3333 }, - { 68.2648, 42.8571 }, - { 68.2648, 47.619 }, - { 63.5029, 61.9048 }, - { 53.979, 71.4286 }, - { 39.6933, 76.1905 }, - { 34.9314, 76.1905 }, - { 20.6457, 71.4286 }, - { 11.1219, 61.9048 }, - { 6.36, 47.619 }, - { 6.36, 42.8571 }, - { 11.1219, 28.5714 }, - { 20.6457, 19.0476 }, - { 34.9314, 14.2857 }, - { 39.6933, 14.2857 }, - { 53.979, 19.0476 }, -}; - -static const StrokeRec char64[] = { - { 8, char64_stroke0 }, - { 19, char64_stroke1 }, -}; - -/* char: 65 'A' */ - -static const CoordRec char65_stroke0[] = { - { 40.5952, 100 }, - { 2.5, 0 }, -}; - -static const CoordRec char65_stroke1[] = { - { 40.5952, 100 }, - { 78.6905, 0 }, -}; - -static const CoordRec char65_stroke2[] = { - { 16.7857, 33.3333 }, - { 64.4048, 33.3333 }, -}; - -static const StrokeRec char65[] = { - { 2, char65_stroke0 }, - { 2, char65_stroke1 }, - { 2, char65_stroke2 }, -}; - -/* char: 66 'B' */ - -static const CoordRec char66_stroke0[] = { - { 11.42, 100 }, - { 11.42, 0 }, -}; - -static const CoordRec char66_stroke1[] = { - { 11.42, 100 }, - { 54.2771, 100 }, - { 68.5629, 95.2381 }, - { 73.3248, 90.4762 }, - { 78.0867, 80.9524 }, - { 78.0867, 71.4286 }, - { 73.3248, 61.9048 }, - { 68.5629, 57.1429 }, - { 54.2771, 52.381 }, -}; - -static const CoordRec char66_stroke2[] = { - { 11.42, 52.381 }, - { 54.2771, 52.381 }, - { 68.5629, 47.619 }, - { 73.3248, 42.8571 }, - { 78.0867, 33.3333 }, - { 78.0867, 19.0476 }, - { 73.3248, 9.5238 }, - { 68.5629, 4.7619 }, - { 54.2771, 0 }, - { 11.42, 0 }, -}; - -static const StrokeRec char66[] = { - { 2, char66_stroke0 }, - { 9, char66_stroke1 }, - { 10, char66_stroke2 }, -}; - -/* char: 67 'C' */ - -static const CoordRec char67_stroke0[] = { - { 78.0886, 76.1905 }, - { 73.3267, 85.7143 }, - { 63.8029, 95.2381 }, - { 54.279, 100 }, - { 35.2314, 100 }, - { 25.7076, 95.2381 }, - { 16.1838, 85.7143 }, - { 11.4219, 76.1905 }, - { 6.66, 61.9048 }, - { 6.66, 38.0952 }, - { 11.4219, 23.8095 }, - { 16.1838, 14.2857 }, - { 25.7076, 4.7619 }, - { 35.2314, 0 }, - { 54.279, 0 }, - { 63.8029, 4.7619 }, - { 73.3267, 14.2857 }, - { 78.0886, 23.8095 }, -}; - -static const StrokeRec char67[] = { - { 18, char67_stroke0 }, -}; - -/* char: 68 'D' */ - -static const CoordRec char68_stroke0[] = { - { 11.96, 100 }, - { 11.96, 0 }, -}; - -static const CoordRec char68_stroke1[] = { - { 11.96, 100 }, - { 45.2933, 100 }, - { 59.579, 95.2381 }, - { 69.1029, 85.7143 }, - { 73.8648, 76.1905 }, - { 78.6267, 61.9048 }, - { 78.6267, 38.0952 }, - { 73.8648, 23.8095 }, - { 69.1029, 14.2857 }, - { 59.579, 4.7619 }, - { 45.2933, 0 }, - { 11.96, 0 }, -}; - -static const StrokeRec char68[] = { - { 2, char68_stroke0 }, - { 12, char68_stroke1 }, -}; - -/* char: 69 'E' */ - -static const CoordRec char69_stroke0[] = { - { 11.42, 100 }, - { 11.42, 0 }, -}; - -static const CoordRec char69_stroke1[] = { - { 11.42, 100 }, - { 73.3248, 100 }, -}; - -static const CoordRec char69_stroke2[] = { - { 11.42, 52.381 }, - { 49.5152, 52.381 }, -}; - -static const CoordRec char69_stroke3[] = { - { 11.42, 0 }, - { 73.3248, 0 }, -}; - -static const StrokeRec char69[] = { - { 2, char69_stroke0 }, - { 2, char69_stroke1 }, - { 2, char69_stroke2 }, - { 2, char69_stroke3 }, -}; - -/* char: 70 'F' */ - -static const CoordRec char70_stroke0[] = { - { 11.42, 100 }, - { 11.42, 0 }, -}; - -static const CoordRec char70_stroke1[] = { - { 11.42, 100 }, - { 73.3248, 100 }, -}; - -static const CoordRec char70_stroke2[] = { - { 11.42, 52.381 }, - { 49.5152, 52.381 }, -}; - -static const StrokeRec char70[] = { - { 2, char70_stroke0 }, - { 2, char70_stroke1 }, - { 2, char70_stroke2 }, -}; - -/* char: 71 'G' */ - -static const CoordRec char71_stroke0[] = { - { 78.4886, 76.1905 }, - { 73.7267, 85.7143 }, - { 64.2029, 95.2381 }, - { 54.679, 100 }, - { 35.6314, 100 }, - { 26.1076, 95.2381 }, - { 16.5838, 85.7143 }, - { 11.8219, 76.1905 }, - { 7.06, 61.9048 }, - { 7.06, 38.0952 }, - { 11.8219, 23.8095 }, - { 16.5838, 14.2857 }, - { 26.1076, 4.7619 }, - { 35.6314, 0 }, - { 54.679, 0 }, - { 64.2029, 4.7619 }, - { 73.7267, 14.2857 }, - { 78.4886, 23.8095 }, - { 78.4886, 38.0952 }, -}; - -static const CoordRec char71_stroke1[] = { - { 54.679, 38.0952 }, - { 78.4886, 38.0952 }, -}; - -static const StrokeRec char71[] = { - { 19, char71_stroke0 }, - { 2, char71_stroke1 }, -}; - -/* char: 72 'H' */ - -static const CoordRec char72_stroke0[] = { - { 11.42, 100 }, - { 11.42, 0 }, -}; - -static const CoordRec char72_stroke1[] = { - { 78.0867, 100 }, - { 78.0867, 0 }, -}; - -static const CoordRec char72_stroke2[] = { - { 11.42, 52.381 }, - { 78.0867, 52.381 }, -}; - -static const StrokeRec char72[] = { - { 2, char72_stroke0 }, - { 2, char72_stroke1 }, - { 2, char72_stroke2 }, -}; - -/* char: 73 'I' */ - -static const CoordRec char73_stroke0[] = { - { 10.86, 100 }, - { 10.86, 0 }, -}; - -static const StrokeRec char73[] = { - { 2, char73_stroke0 }, -}; - -/* char: 74 'J' */ - -static const CoordRec char74_stroke0[] = { - { 50.119, 100 }, - { 50.119, 23.8095 }, - { 45.3571, 9.5238 }, - { 40.5952, 4.7619 }, - { 31.0714, 0 }, - { 21.5476, 0 }, - { 12.0238, 4.7619 }, - { 7.2619, 9.5238 }, - { 2.5, 23.8095 }, - { 2.5, 33.3333 }, -}; - -static const StrokeRec char74[] = { - { 10, char74_stroke0 }, -}; - -/* char: 75 'K' */ - -static const CoordRec char75_stroke0[] = { - { 11.28, 100 }, - { 11.28, 0 }, -}; - -static const CoordRec char75_stroke1[] = { - { 77.9467, 100 }, - { 11.28, 33.3333 }, -}; - -static const CoordRec char75_stroke2[] = { - { 35.0895, 57.1429 }, - { 77.9467, 0 }, -}; - -static const StrokeRec char75[] = { - { 2, char75_stroke0 }, - { 2, char75_stroke1 }, - { 2, char75_stroke2 }, -}; - -/* char: 76 'L' */ - -static const CoordRec char76_stroke0[] = { - { 11.68, 100 }, - { 11.68, 0 }, -}; - -static const CoordRec char76_stroke1[] = { - { 11.68, 0 }, - { 68.8229, 0 }, -}; - -static const StrokeRec char76[] = { - { 2, char76_stroke0 }, - { 2, char76_stroke1 }, -}; - -/* char: 77 'M' */ - -static const CoordRec char77_stroke0[] = { - { 10.86, 100 }, - { 10.86, 0 }, -}; - -static const CoordRec char77_stroke1[] = { - { 10.86, 100 }, - { 48.9552, 0 }, -}; - -static const CoordRec char77_stroke2[] = { - { 87.0505, 100 }, - { 48.9552, 0 }, -}; - -static const CoordRec char77_stroke3[] = { - { 87.0505, 100 }, - { 87.0505, 0 }, -}; - -static const StrokeRec char77[] = { - { 2, char77_stroke0 }, - { 2, char77_stroke1 }, - { 2, char77_stroke2 }, - { 2, char77_stroke3 }, -}; - -/* char: 78 'N' */ - -static const CoordRec char78_stroke0[] = { - { 11.14, 100 }, - { 11.14, 0 }, -}; - -static const CoordRec char78_stroke1[] = { - { 11.14, 100 }, - { 77.8067, 0 }, -}; - -static const CoordRec char78_stroke2[] = { - { 77.8067, 100 }, - { 77.8067, 0 }, -}; - -static const StrokeRec char78[] = { - { 2, char78_stroke0 }, - { 2, char78_stroke1 }, - { 2, char78_stroke2 }, -}; - -/* char: 79 'O' */ - -static const CoordRec char79_stroke0[] = { - { 34.8114, 100 }, - { 25.2876, 95.2381 }, - { 15.7638, 85.7143 }, - { 11.0019, 76.1905 }, - { 6.24, 61.9048 }, - { 6.24, 38.0952 }, - { 11.0019, 23.8095 }, - { 15.7638, 14.2857 }, - { 25.2876, 4.7619 }, - { 34.8114, 0 }, - { 53.859, 0 }, - { 63.3829, 4.7619 }, - { 72.9067, 14.2857 }, - { 77.6686, 23.8095 }, - { 82.4305, 38.0952 }, - { 82.4305, 61.9048 }, - { 77.6686, 76.1905 }, - { 72.9067, 85.7143 }, - { 63.3829, 95.2381 }, - { 53.859, 100 }, - { 34.8114, 100 }, -}; - -static const StrokeRec char79[] = { - { 21, char79_stroke0 }, -}; - -/* char: 80 'P' */ - -static const CoordRec char80_stroke0[] = { - { 12.1, 100 }, - { 12.1, 0 }, -}; - -static const CoordRec char80_stroke1[] = { - { 12.1, 100 }, - { 54.9571, 100 }, - { 69.2429, 95.2381 }, - { 74.0048, 90.4762 }, - { 78.7667, 80.9524 }, - { 78.7667, 66.6667 }, - { 74.0048, 57.1429 }, - { 69.2429, 52.381 }, - { 54.9571, 47.619 }, - { 12.1, 47.619 }, -}; - -static const StrokeRec char80[] = { - { 2, char80_stroke0 }, - { 10, char80_stroke1 }, -}; - -/* char: 81 'Q' */ - -static const CoordRec char81_stroke0[] = { - { 33.8714, 100 }, - { 24.3476, 95.2381 }, - { 14.8238, 85.7143 }, - { 10.0619, 76.1905 }, - { 5.3, 61.9048 }, - { 5.3, 38.0952 }, - { 10.0619, 23.8095 }, - { 14.8238, 14.2857 }, - { 24.3476, 4.7619 }, - { 33.8714, 0 }, - { 52.919, 0 }, - { 62.4429, 4.7619 }, - { 71.9667, 14.2857 }, - { 76.7286, 23.8095 }, - { 81.4905, 38.0952 }, - { 81.4905, 61.9048 }, - { 76.7286, 76.1905 }, - { 71.9667, 85.7143 }, - { 62.4429, 95.2381 }, - { 52.919, 100 }, - { 33.8714, 100 }, -}; - -static const CoordRec char81_stroke1[] = { - { 48.1571, 19.0476 }, - { 76.7286, -9.5238 }, -}; - -static const StrokeRec char81[] = { - { 21, char81_stroke0 }, - { 2, char81_stroke1 }, -}; - -/* char: 82 'R' */ - -static const CoordRec char82_stroke0[] = { - { 11.68, 100 }, - { 11.68, 0 }, -}; - -static const CoordRec char82_stroke1[] = { - { 11.68, 100 }, - { 54.5371, 100 }, - { 68.8229, 95.2381 }, - { 73.5848, 90.4762 }, - { 78.3467, 80.9524 }, - { 78.3467, 71.4286 }, - { 73.5848, 61.9048 }, - { 68.8229, 57.1429 }, - { 54.5371, 52.381 }, - { 11.68, 52.381 }, -}; - -static const CoordRec char82_stroke2[] = { - { 45.0133, 52.381 }, - { 78.3467, 0 }, -}; - -static const StrokeRec char82[] = { - { 2, char82_stroke0 }, - { 10, char82_stroke1 }, - { 2, char82_stroke2 }, -}; - -/* char: 83 'S' */ - -static const CoordRec char83_stroke0[] = { - { 74.6667, 85.7143 }, - { 65.1429, 95.2381 }, - { 50.8571, 100 }, - { 31.8095, 100 }, - { 17.5238, 95.2381 }, - { 8, 85.7143 }, - { 8, 76.1905 }, - { 12.7619, 66.6667 }, - { 17.5238, 61.9048 }, - { 27.0476, 57.1429 }, - { 55.619, 47.619 }, - { 65.1429, 42.8571 }, - { 69.9048, 38.0952 }, - { 74.6667, 28.5714 }, - { 74.6667, 14.2857 }, - { 65.1429, 4.7619 }, - { 50.8571, 0 }, - { 31.8095, 0 }, - { 17.5238, 4.7619 }, - { 8, 14.2857 }, -}; - -static const StrokeRec char83[] = { - { 20, char83_stroke0 }, -}; - -/* char: 84 'T' */ - -static const CoordRec char84_stroke0[] = { - { 35.6933, 100 }, - { 35.6933, 0 }, -}; - -static const CoordRec char84_stroke1[] = { - { 2.36, 100 }, - { 69.0267, 100 }, -}; - -static const StrokeRec char84[] = { - { 2, char84_stroke0 }, - { 2, char84_stroke1 }, -}; - -/* char: 85 'U' */ - -static const CoordRec char85_stroke0[] = { - { 11.54, 100 }, - { 11.54, 28.5714 }, - { 16.3019, 14.2857 }, - { 25.8257, 4.7619 }, - { 40.1114, 0 }, - { 49.6352, 0 }, - { 63.921, 4.7619 }, - { 73.4448, 14.2857 }, - { 78.2067, 28.5714 }, - { 78.2067, 100 }, -}; - -static const StrokeRec char85[] = { - { 10, char85_stroke0 }, -}; - -/* char: 86 'V' */ - -static const CoordRec char86_stroke0[] = { - { 2.36, 100 }, - { 40.4552, 0 }, -}; - -static const CoordRec char86_stroke1[] = { - { 78.5505, 100 }, - { 40.4552, 0 }, -}; - -static const StrokeRec char86[] = { - { 2, char86_stroke0 }, - { 2, char86_stroke1 }, -}; - -/* char: 87 'W' */ - -static const CoordRec char87_stroke0[] = { - { 2.22, 100 }, - { 26.0295, 0 }, -}; - -static const CoordRec char87_stroke1[] = { - { 49.839, 100 }, - { 26.0295, 0 }, -}; - -static const CoordRec char87_stroke2[] = { - { 49.839, 100 }, - { 73.6486, 0 }, -}; - -static const CoordRec char87_stroke3[] = { - { 97.4581, 100 }, - { 73.6486, 0 }, -}; - -static const StrokeRec char87[] = { - { 2, char87_stroke0 }, - { 2, char87_stroke1 }, - { 2, char87_stroke2 }, - { 2, char87_stroke3 }, -}; - -/* char: 88 'X' */ - -static const CoordRec char88_stroke0[] = { - { 2.5, 100 }, - { 69.1667, 0 }, -}; - -static const CoordRec char88_stroke1[] = { - { 69.1667, 100 }, - { 2.5, 0 }, -}; - -static const StrokeRec char88[] = { - { 2, char88_stroke0 }, - { 2, char88_stroke1 }, -}; - -/* char: 89 'Y' */ - -static const CoordRec char89_stroke0[] = { - { 1.52, 100 }, - { 39.6152, 52.381 }, - { 39.6152, 0 }, -}; - -static const CoordRec char89_stroke1[] = { - { 77.7105, 100 }, - { 39.6152, 52.381 }, -}; - -static const StrokeRec char89[] = { - { 3, char89_stroke0 }, - { 2, char89_stroke1 }, -}; - -/* char: 90 'Z' */ - -static const CoordRec char90_stroke0[] = { - { 69.1667, 100 }, - { 2.5, 0 }, -}; - -static const CoordRec char90_stroke1[] = { - { 2.5, 100 }, - { 69.1667, 100 }, -}; - -static const CoordRec char90_stroke2[] = { - { 2.5, 0 }, - { 69.1667, 0 }, -}; - -static const StrokeRec char90[] = { - { 2, char90_stroke0 }, - { 2, char90_stroke1 }, - { 2, char90_stroke2 }, -}; - -/* char: 91 '[' */ - -static const CoordRec char91_stroke0[] = { - { 7.78, 119.048 }, - { 7.78, -33.3333 }, -}; - -static const CoordRec char91_stroke1[] = { - { 12.5419, 119.048 }, - { 12.5419, -33.3333 }, -}; - -static const CoordRec char91_stroke2[] = { - { 7.78, 119.048 }, - { 41.1133, 119.048 }, -}; - -static const CoordRec char91_stroke3[] = { - { 7.78, -33.3333 }, - { 41.1133, -33.3333 }, -}; - -static const StrokeRec char91[] = { - { 2, char91_stroke0 }, - { 2, char91_stroke1 }, - { 2, char91_stroke2 }, - { 2, char91_stroke3 }, -}; - -/* char: 92 '\' */ - -static const CoordRec char92_stroke0[] = { - { 5.84, 100 }, - { 72.5067, -14.2857 }, -}; - -static const StrokeRec char92[] = { - { 2, char92_stroke0 }, -}; - -/* char: 93 ']' */ - -static const CoordRec char93_stroke0[] = { - { 33.0114, 119.048 }, - { 33.0114, -33.3333 }, -}; - -static const CoordRec char93_stroke1[] = { - { 37.7733, 119.048 }, - { 37.7733, -33.3333 }, -}; - -static const CoordRec char93_stroke2[] = { - { 4.44, 119.048 }, - { 37.7733, 119.048 }, -}; - -static const CoordRec char93_stroke3[] = { - { 4.44, -33.3333 }, - { 37.7733, -33.3333 }, -}; - -static const StrokeRec char93[] = { - { 2, char93_stroke0 }, - { 2, char93_stroke1 }, - { 2, char93_stroke2 }, - { 2, char93_stroke3 }, -}; - -/* char: 94 '^' */ - -static const CoordRec char94_stroke0[] = { - { 44.0752, 109.524 }, - { 5.98, 42.8571 }, -}; - -static const CoordRec char94_stroke1[] = { - { 44.0752, 109.524 }, - { 82.1705, 42.8571 }, -}; - -static const StrokeRec char94[] = { - { 2, char94_stroke0 }, - { 2, char94_stroke1 }, -}; - -/* char: 95 '_' */ - -static const CoordRec char95_stroke0[] = { - { -1.1, -33.3333 }, - { 103.662, -33.3333 }, - { 103.662, -28.5714 }, - { -1.1, -28.5714 }, - { -1.1, -33.3333 }, -}; - -static const StrokeRec char95[] = { - { 5, char95_stroke0 }, -}; - -/* char: 96 '`' */ - -static const CoordRec char96_stroke0[] = { - { 33.0219, 100 }, - { 56.8314, 71.4286 }, -}; - -static const CoordRec char96_stroke1[] = { - { 33.0219, 100 }, - { 28.26, 95.2381 }, - { 56.8314, 71.4286 }, -}; - -static const StrokeRec char96[] = { - { 2, char96_stroke0 }, - { 3, char96_stroke1 }, -}; - -/* char: 97 'a' */ - -static const CoordRec char97_stroke0[] = { - { 63.8229, 66.6667 }, - { 63.8229, 0 }, -}; - -static const CoordRec char97_stroke1[] = { - { 63.8229, 52.381 }, - { 54.299, 61.9048 }, - { 44.7752, 66.6667 }, - { 30.4895, 66.6667 }, - { 20.9657, 61.9048 }, - { 11.4419, 52.381 }, - { 6.68, 38.0952 }, - { 6.68, 28.5714 }, - { 11.4419, 14.2857 }, - { 20.9657, 4.7619 }, - { 30.4895, 0 }, - { 44.7752, 0 }, - { 54.299, 4.7619 }, - { 63.8229, 14.2857 }, -}; - -static const StrokeRec char97[] = { - { 2, char97_stroke0 }, - { 14, char97_stroke1 }, -}; - -/* char: 98 'b' */ - -static const CoordRec char98_stroke0[] = { - { 8.76, 100 }, - { 8.76, 0 }, -}; - -static const CoordRec char98_stroke1[] = { - { 8.76, 52.381 }, - { 18.2838, 61.9048 }, - { 27.8076, 66.6667 }, - { 42.0933, 66.6667 }, - { 51.6171, 61.9048 }, - { 61.141, 52.381 }, - { 65.9029, 38.0952 }, - { 65.9029, 28.5714 }, - { 61.141, 14.2857 }, - { 51.6171, 4.7619 }, - { 42.0933, 0 }, - { 27.8076, 0 }, - { 18.2838, 4.7619 }, - { 8.76, 14.2857 }, -}; - -static const StrokeRec char98[] = { - { 2, char98_stroke0 }, - { 14, char98_stroke1 }, -}; - -/* char: 99 'c' */ - -static const CoordRec char99_stroke0[] = { - { 62.6629, 52.381 }, - { 53.139, 61.9048 }, - { 43.6152, 66.6667 }, - { 29.3295, 66.6667 }, - { 19.8057, 61.9048 }, - { 10.2819, 52.381 }, - { 5.52, 38.0952 }, - { 5.52, 28.5714 }, - { 10.2819, 14.2857 }, - { 19.8057, 4.7619 }, - { 29.3295, 0 }, - { 43.6152, 0 }, - { 53.139, 4.7619 }, - { 62.6629, 14.2857 }, -}; - -static const StrokeRec char99[] = { - { 14, char99_stroke0 }, -}; - -/* char: 100 'd' */ - -static const CoordRec char100_stroke0[] = { - { 61.7829, 100 }, - { 61.7829, 0 }, -}; - -static const CoordRec char100_stroke1[] = { - { 61.7829, 52.381 }, - { 52.259, 61.9048 }, - { 42.7352, 66.6667 }, - { 28.4495, 66.6667 }, - { 18.9257, 61.9048 }, - { 9.4019, 52.381 }, - { 4.64, 38.0952 }, - { 4.64, 28.5714 }, - { 9.4019, 14.2857 }, - { 18.9257, 4.7619 }, - { 28.4495, 0 }, - { 42.7352, 0 }, - { 52.259, 4.7619 }, - { 61.7829, 14.2857 }, -}; - -static const StrokeRec char100[] = { - { 2, char100_stroke0 }, - { 14, char100_stroke1 }, -}; - -/* char: 101 'e' */ - -static const CoordRec char101_stroke0[] = { - { 5.72, 38.0952 }, - { 62.8629, 38.0952 }, - { 62.8629, 47.619 }, - { 58.101, 57.1429 }, - { 53.339, 61.9048 }, - { 43.8152, 66.6667 }, - { 29.5295, 66.6667 }, - { 20.0057, 61.9048 }, - { 10.4819, 52.381 }, - { 5.72, 38.0952 }, - { 5.72, 28.5714 }, - { 10.4819, 14.2857 }, - { 20.0057, 4.7619 }, - { 29.5295, 0 }, - { 43.8152, 0 }, - { 53.339, 4.7619 }, - { 62.8629, 14.2857 }, -}; - -static const StrokeRec char101[] = { - { 17, char101_stroke0 }, -}; - -/* char: 102 'f' */ - -static const CoordRec char102_stroke0[] = { - { 38.7752, 100 }, - { 29.2514, 100 }, - { 19.7276, 95.2381 }, - { 14.9657, 80.9524 }, - { 14.9657, 0 }, -}; - -static const CoordRec char102_stroke1[] = { - { 0.68, 66.6667 }, - { 34.0133, 66.6667 }, -}; - -static const StrokeRec char102[] = { - { 5, char102_stroke0 }, - { 2, char102_stroke1 }, -}; - -/* char: 103 'g' */ - -static const CoordRec char103_stroke0[] = { - { 62.5029, 66.6667 }, - { 62.5029, -9.5238 }, - { 57.741, -23.8095 }, - { 52.979, -28.5714 }, - { 43.4552, -33.3333 }, - { 29.1695, -33.3333 }, - { 19.6457, -28.5714 }, -}; - -static const CoordRec char103_stroke1[] = { - { 62.5029, 52.381 }, - { 52.979, 61.9048 }, - { 43.4552, 66.6667 }, - { 29.1695, 66.6667 }, - { 19.6457, 61.9048 }, - { 10.1219, 52.381 }, - { 5.36, 38.0952 }, - { 5.36, 28.5714 }, - { 10.1219, 14.2857 }, - { 19.6457, 4.7619 }, - { 29.1695, 0 }, - { 43.4552, 0 }, - { 52.979, 4.7619 }, - { 62.5029, 14.2857 }, -}; - -static const StrokeRec char103[] = { - { 7, char103_stroke0 }, - { 14, char103_stroke1 }, -}; - -/* char: 104 'h' */ - -static const CoordRec char104_stroke0[] = { - { 9.6, 100 }, - { 9.6, 0 }, -}; - -static const CoordRec char104_stroke1[] = { - { 9.6, 47.619 }, - { 23.8857, 61.9048 }, - { 33.4095, 66.6667 }, - { 47.6952, 66.6667 }, - { 57.219, 61.9048 }, - { 61.981, 47.619 }, - { 61.981, 0 }, -}; - -static const StrokeRec char104[] = { - { 2, char104_stroke0 }, - { 7, char104_stroke1 }, -}; - -/* char: 105 'i' */ - -static const CoordRec char105_stroke0[] = { - { 10.02, 100 }, - { 14.7819, 95.2381 }, - { 19.5438, 100 }, - { 14.7819, 104.762 }, - { 10.02, 100 }, -}; - -static const CoordRec char105_stroke1[] = { - { 14.7819, 66.6667 }, - { 14.7819, 0 }, -}; - -static const StrokeRec char105[] = { - { 5, char105_stroke0 }, - { 2, char105_stroke1 }, -}; - -/* char: 106 'j' */ - -static const CoordRec char106_stroke0[] = { - { 17.3876, 100 }, - { 22.1495, 95.2381 }, - { 26.9114, 100 }, - { 22.1495, 104.762 }, - { 17.3876, 100 }, -}; - -static const CoordRec char106_stroke1[] = { - { 22.1495, 66.6667 }, - { 22.1495, -14.2857 }, - { 17.3876, -28.5714 }, - { 7.8638, -33.3333 }, - { -1.66, -33.3333 }, -}; - -static const StrokeRec char106[] = { - { 5, char106_stroke0 }, - { 5, char106_stroke1 }, -}; - -/* char: 107 'k' */ - -static const CoordRec char107_stroke0[] = { - { 9.6, 100 }, - { 9.6, 0 }, -}; - -static const CoordRec char107_stroke1[] = { - { 57.219, 66.6667 }, - { 9.6, 19.0476 }, -}; - -static const CoordRec char107_stroke2[] = { - { 28.6476, 38.0952 }, - { 61.981, 0 }, -}; - -static const StrokeRec char107[] = { - { 2, char107_stroke0 }, - { 2, char107_stroke1 }, - { 2, char107_stroke2 }, -}; - -/* char: 108 'l' */ - -static const CoordRec char108_stroke0[] = { - { 10.02, 100 }, - { 10.02, 0 }, -}; - -static const StrokeRec char108[] = { - { 2, char108_stroke0 }, -}; - -/* char: 109 'm' */ - -static const CoordRec char109_stroke0[] = { - { 9.6, 66.6667 }, - { 9.6, 0 }, -}; - -static const CoordRec char109_stroke1[] = { - { 9.6, 47.619 }, - { 23.8857, 61.9048 }, - { 33.4095, 66.6667 }, - { 47.6952, 66.6667 }, - { 57.219, 61.9048 }, - { 61.981, 47.619 }, - { 61.981, 0 }, -}; - -static const CoordRec char109_stroke2[] = { - { 61.981, 47.619 }, - { 76.2667, 61.9048 }, - { 85.7905, 66.6667 }, - { 100.076, 66.6667 }, - { 109.6, 61.9048 }, - { 114.362, 47.619 }, - { 114.362, 0 }, -}; - -static const StrokeRec char109[] = { - { 2, char109_stroke0 }, - { 7, char109_stroke1 }, - { 7, char109_stroke2 }, -}; - -/* char: 110 'n' */ - -static const CoordRec char110_stroke0[] = { - { 9.18, 66.6667 }, - { 9.18, 0 }, -}; - -static const CoordRec char110_stroke1[] = { - { 9.18, 47.619 }, - { 23.4657, 61.9048 }, - { 32.9895, 66.6667 }, - { 47.2752, 66.6667 }, - { 56.799, 61.9048 }, - { 61.561, 47.619 }, - { 61.561, 0 }, -}; - -static const StrokeRec char110[] = { - { 2, char110_stroke0 }, - { 7, char110_stroke1 }, -}; - -/* char: 111 'o' */ - -static const CoordRec char111_stroke0[] = { - { 28.7895, 66.6667 }, - { 19.2657, 61.9048 }, - { 9.7419, 52.381 }, - { 4.98, 38.0952 }, - { 4.98, 28.5714 }, - { 9.7419, 14.2857 }, - { 19.2657, 4.7619 }, - { 28.7895, 0 }, - { 43.0752, 0 }, - { 52.599, 4.7619 }, - { 62.1229, 14.2857 }, - { 66.8848, 28.5714 }, - { 66.8848, 38.0952 }, - { 62.1229, 52.381 }, - { 52.599, 61.9048 }, - { 43.0752, 66.6667 }, - { 28.7895, 66.6667 }, -}; - -static const StrokeRec char111[] = { - { 17, char111_stroke0 }, -}; - -/* char: 112 'p' */ - -static const CoordRec char112_stroke0[] = { - { 9.46, 66.6667 }, - { 9.46, -33.3333 }, -}; - -static const CoordRec char112_stroke1[] = { - { 9.46, 52.381 }, - { 18.9838, 61.9048 }, - { 28.5076, 66.6667 }, - { 42.7933, 66.6667 }, - { 52.3171, 61.9048 }, - { 61.841, 52.381 }, - { 66.6029, 38.0952 }, - { 66.6029, 28.5714 }, - { 61.841, 14.2857 }, - { 52.3171, 4.7619 }, - { 42.7933, 0 }, - { 28.5076, 0 }, - { 18.9838, 4.7619 }, - { 9.46, 14.2857 }, -}; - -static const StrokeRec char112[] = { - { 2, char112_stroke0 }, - { 14, char112_stroke1 }, -}; - -/* char: 113 'q' */ - -static const CoordRec char113_stroke0[] = { - { 61.9829, 66.6667 }, - { 61.9829, -33.3333 }, -}; - -static const CoordRec char113_stroke1[] = { - { 61.9829, 52.381 }, - { 52.459, 61.9048 }, - { 42.9352, 66.6667 }, - { 28.6495, 66.6667 }, - { 19.1257, 61.9048 }, - { 9.6019, 52.381 }, - { 4.84, 38.0952 }, - { 4.84, 28.5714 }, - { 9.6019, 14.2857 }, - { 19.1257, 4.7619 }, - { 28.6495, 0 }, - { 42.9352, 0 }, - { 52.459, 4.7619 }, - { 61.9829, 14.2857 }, -}; - -static const StrokeRec char113[] = { - { 2, char113_stroke0 }, - { 14, char113_stroke1 }, -}; - -/* char: 114 'r' */ - -static const CoordRec char114_stroke0[] = { - { 9.46, 66.6667 }, - { 9.46, 0 }, -}; - -static const CoordRec char114_stroke1[] = { - { 9.46, 38.0952 }, - { 14.2219, 52.381 }, - { 23.7457, 61.9048 }, - { 33.2695, 66.6667 }, - { 47.5552, 66.6667 }, -}; - -static const StrokeRec char114[] = { - { 2, char114_stroke0 }, - { 5, char114_stroke1 }, -}; - -/* char: 115 's' */ - -static const CoordRec char115_stroke0[] = { - { 57.081, 52.381 }, - { 52.319, 61.9048 }, - { 38.0333, 66.6667 }, - { 23.7476, 66.6667 }, - { 9.4619, 61.9048 }, - { 4.7, 52.381 }, - { 9.4619, 42.8571 }, - { 18.9857, 38.0952 }, - { 42.7952, 33.3333 }, - { 52.319, 28.5714 }, - { 57.081, 19.0476 }, - { 57.081, 14.2857 }, - { 52.319, 4.7619 }, - { 38.0333, 0 }, - { 23.7476, 0 }, - { 9.4619, 4.7619 }, - { 4.7, 14.2857 }, -}; - -static const StrokeRec char115[] = { - { 17, char115_stroke0 }, -}; - -/* char: 116 't' */ - -static const CoordRec char116_stroke0[] = { - { 14.8257, 100 }, - { 14.8257, 19.0476 }, - { 19.5876, 4.7619 }, - { 29.1114, 0 }, - { 38.6352, 0 }, -}; - -static const CoordRec char116_stroke1[] = { - { 0.54, 66.6667 }, - { 33.8733, 66.6667 }, -}; - -static const StrokeRec char116[] = { - { 5, char116_stroke0 }, - { 2, char116_stroke1 }, -}; - -/* char: 117 'u' */ - -static const CoordRec char117_stroke0[] = { - { 9.46, 66.6667 }, - { 9.46, 19.0476 }, - { 14.2219, 4.7619 }, - { 23.7457, 0 }, - { 38.0314, 0 }, - { 47.5552, 4.7619 }, - { 61.841, 19.0476 }, -}; - -static const CoordRec char117_stroke1[] = { - { 61.841, 66.6667 }, - { 61.841, 0 }, -}; - -static const StrokeRec char117[] = { - { 7, char117_stroke0 }, - { 2, char117_stroke1 }, -}; - -/* char: 118 'v' */ - -static const CoordRec char118_stroke0[] = { - { 1.8, 66.6667 }, - { 30.3714, 0 }, -}; - -static const CoordRec char118_stroke1[] = { - { 58.9429, 66.6667 }, - { 30.3714, 0 }, -}; - -static const StrokeRec char118[] = { - { 2, char118_stroke0 }, - { 2, char118_stroke1 }, -}; - -/* char: 119 'w' */ - -static const CoordRec char119_stroke0[] = { - { 2.5, 66.6667 }, - { 21.5476, 0 }, -}; - -static const CoordRec char119_stroke1[] = { - { 40.5952, 66.6667 }, - { 21.5476, 0 }, -}; - -static const CoordRec char119_stroke2[] = { - { 40.5952, 66.6667 }, - { 59.6429, 0 }, -}; - -static const CoordRec char119_stroke3[] = { - { 78.6905, 66.6667 }, - { 59.6429, 0 }, -}; - -static const StrokeRec char119[] = { - { 2, char119_stroke0 }, - { 2, char119_stroke1 }, - { 2, char119_stroke2 }, - { 2, char119_stroke3 }, -}; - -/* char: 120 'x' */ - -static const CoordRec char120_stroke0[] = { - { 1.66, 66.6667 }, - { 54.041, 0 }, -}; - -static const CoordRec char120_stroke1[] = { - { 54.041, 66.6667 }, - { 1.66, 0 }, -}; - -static const StrokeRec char120[] = { - { 2, char120_stroke0 }, - { 2, char120_stroke1 }, -}; - -/* char: 121 'y' */ - -static const CoordRec char121_stroke0[] = { - { 6.5619, 66.6667 }, - { 35.1333, 0 }, -}; - -static const CoordRec char121_stroke1[] = { - { 63.7048, 66.6667 }, - { 35.1333, 0 }, - { 25.6095, -19.0476 }, - { 16.0857, -28.5714 }, - { 6.5619, -33.3333 }, - { 1.8, -33.3333 }, -}; - -static const StrokeRec char121[] = { - { 2, char121_stroke0 }, - { 6, char121_stroke1 }, -}; - -/* char: 122 'z' */ - -static const CoordRec char122_stroke0[] = { - { 56.821, 66.6667 }, - { 4.44, 0 }, -}; - -static const CoordRec char122_stroke1[] = { - { 4.44, 66.6667 }, - { 56.821, 66.6667 }, -}; - -static const CoordRec char122_stroke2[] = { - { 4.44, 0 }, - { 56.821, 0 }, -}; - -static const StrokeRec char122[] = { - { 2, char122_stroke0 }, - { 2, char122_stroke1 }, - { 2, char122_stroke2 }, -}; - -/* char: 123 '{' */ - -static const CoordRec char123_stroke0[] = { - { 31.1895, 119.048 }, - { 21.6657, 114.286 }, - { 16.9038, 109.524 }, - { 12.1419, 100 }, - { 12.1419, 90.4762 }, - { 16.9038, 80.9524 }, - { 21.6657, 76.1905 }, - { 26.4276, 66.6667 }, - { 26.4276, 57.1429 }, - { 16.9038, 47.619 }, -}; - -static const CoordRec char123_stroke1[] = { - { 21.6657, 114.286 }, - { 16.9038, 104.762 }, - { 16.9038, 95.2381 }, - { 21.6657, 85.7143 }, - { 26.4276, 80.9524 }, - { 31.1895, 71.4286 }, - { 31.1895, 61.9048 }, - { 26.4276, 52.381 }, - { 7.38, 42.8571 }, - { 26.4276, 33.3333 }, - { 31.1895, 23.8095 }, - { 31.1895, 14.2857 }, - { 26.4276, 4.7619 }, - { 21.6657, 0 }, - { 16.9038, -9.5238 }, - { 16.9038, -19.0476 }, - { 21.6657, -28.5714 }, -}; - -static const CoordRec char123_stroke2[] = { - { 16.9038, 38.0952 }, - { 26.4276, 28.5714 }, - { 26.4276, 19.0476 }, - { 21.6657, 9.5238 }, - { 16.9038, 4.7619 }, - { 12.1419, -4.7619 }, - { 12.1419, -14.2857 }, - { 16.9038, -23.8095 }, - { 21.6657, -28.5714 }, - { 31.1895, -33.3333 }, -}; - -static const StrokeRec char123[] = { - { 10, char123_stroke0 }, - { 17, char123_stroke1 }, - { 10, char123_stroke2 }, -}; - -/* char: 124 '|' */ - -static const CoordRec char124_stroke0[] = { - { 11.54, 119.048 }, - { 11.54, -33.3333 }, -}; - -static const StrokeRec char124[] = { - { 2, char124_stroke0 }, -}; - -/* char: 125 '}' */ - -static const CoordRec char125_stroke0[] = { - { 9.18, 119.048 }, - { 18.7038, 114.286 }, - { 23.4657, 109.524 }, - { 28.2276, 100 }, - { 28.2276, 90.4762 }, - { 23.4657, 80.9524 }, - { 18.7038, 76.1905 }, - { 13.9419, 66.6667 }, - { 13.9419, 57.1429 }, - { 23.4657, 47.619 }, -}; - -static const CoordRec char125_stroke1[] = { - { 18.7038, 114.286 }, - { 23.4657, 104.762 }, - { 23.4657, 95.2381 }, - { 18.7038, 85.7143 }, - { 13.9419, 80.9524 }, - { 9.18, 71.4286 }, - { 9.18, 61.9048 }, - { 13.9419, 52.381 }, - { 32.9895, 42.8571 }, - { 13.9419, 33.3333 }, - { 9.18, 23.8095 }, - { 9.18, 14.2857 }, - { 13.9419, 4.7619 }, - { 18.7038, 0 }, - { 23.4657, -9.5238 }, - { 23.4657, -19.0476 }, - { 18.7038, -28.5714 }, -}; - -static const CoordRec char125_stroke2[] = { - { 23.4657, 38.0952 }, - { 13.9419, 28.5714 }, - { 13.9419, 19.0476 }, - { 18.7038, 9.5238 }, - { 23.4657, 4.7619 }, - { 28.2276, -4.7619 }, - { 28.2276, -14.2857 }, - { 23.4657, -23.8095 }, - { 18.7038, -28.5714 }, - { 9.18, -33.3333 }, -}; - -static const StrokeRec char125[] = { - { 10, char125_stroke0 }, - { 17, char125_stroke1 }, - { 10, char125_stroke2 }, -}; - -/* char: 126 '~' */ - -static const CoordRec char126_stroke0[] = { - { 2.92, 28.5714 }, - { 2.92, 38.0952 }, - { 7.6819, 52.381 }, - { 17.2057, 57.1429 }, - { 26.7295, 57.1429 }, - { 36.2533, 52.381 }, - { 55.301, 38.0952 }, - { 64.8248, 33.3333 }, - { 74.3486, 33.3333 }, - { 83.8724, 38.0952 }, - { 88.6343, 47.619 }, -}; - -static const CoordRec char126_stroke1[] = { - { 2.92, 38.0952 }, - { 7.6819, 47.619 }, - { 17.2057, 52.381 }, - { 26.7295, 52.381 }, - { 36.2533, 47.619 }, - { 55.301, 33.3333 }, - { 64.8248, 28.5714 }, - { 74.3486, 28.5714 }, - { 83.8724, 33.3333 }, - { 88.6343, 47.619 }, - { 88.6343, 57.1429 }, -}; - -static const StrokeRec char126[] = { - { 11, char126_stroke0 }, - { 11, char126_stroke1 }, -}; - -/* char: 127 */ - -static const CoordRec char127_stroke0[] = { - { 52.381, 100 }, - { 14.2857, -33.3333 }, -}; - -static const CoordRec char127_stroke1[] = { - { 28.5714, 66.6667 }, - { 14.2857, 61.9048 }, - { 4.7619, 52.381 }, - { 0, 38.0952 }, - { 0, 23.8095 }, - { 4.7619, 14.2857 }, - { 14.2857, 4.7619 }, - { 28.5714, 0 }, - { 38.0952, 0 }, - { 52.381, 4.7619 }, - { 61.9048, 14.2857 }, - { 66.6667, 28.5714 }, - { 66.6667, 42.8571 }, - { 61.9048, 52.381 }, - { 52.381, 61.9048 }, - { 38.0952, 66.6667 }, - { 28.5714, 66.6667 }, -}; - -static const StrokeRec char127[] = { - { 2, char127_stroke0 }, - { 17, char127_stroke1 }, -}; - -static const StrokeCharRec chars[] = { - { 0, /* char0 */ 0, 0, 0 }, - { 0, /* char1 */ 0, 0, 0 }, - { 0, /* char2 */ 0, 0, 0 }, - { 0, /* char3 */ 0, 0, 0 }, - { 0, /* char4 */ 0, 0, 0 }, - { 0, /* char5 */ 0, 0, 0 }, - { 0, /* char6 */ 0, 0, 0 }, - { 0, /* char7 */ 0, 0, 0 }, - { 0, /* char8 */ 0, 0, 0 }, - { 0, /* char9 */ 0, 0, 0 }, - { 0, /* char10 */ 0, 0, 0 }, - { 0, /* char11 */ 0, 0, 0 }, - { 0, /* char12 */ 0, 0, 0 }, - { 0, /* char13 */ 0, 0, 0 }, - { 0, /* char14 */ 0, 0, 0 }, - { 0, /* char15 */ 0, 0, 0 }, - { 0, /* char16 */ 0, 0, 0 }, - { 0, /* char17 */ 0, 0, 0 }, - { 0, /* char18 */ 0, 0, 0 }, - { 0, /* char19 */ 0, 0, 0 }, - { 0, /* char20 */ 0, 0, 0 }, - { 0, /* char21 */ 0, 0, 0 }, - { 0, /* char22 */ 0, 0, 0 }, - { 0, /* char23 */ 0, 0, 0 }, - { 0, /* char24 */ 0, 0, 0 }, - { 0, /* char25 */ 0, 0, 0 }, - { 0, /* char26 */ 0, 0, 0 }, - { 0, /* char27 */ 0, 0, 0 }, - { 0, /* char28 */ 0, 0, 0 }, - { 0, /* char29 */ 0, 0, 0 }, - { 0, /* char30 */ 0, 0, 0 }, - { 0, /* char31 */ 0, 0, 0 }, - { 0, /* char32 */ 0, 52.381, 104.762 }, - { 2, char33, 13.3819, 26.6238 }, - { 2, char34, 23.0676, 51.4352 }, - { 4, char35, 36.5333, 79.4886 }, - { 3, char36, 38.1533, 76.2067 }, - { 3, char37, 49.2171, 96.5743 }, - { 1, char38, 53.599, 101.758 }, - { 1, char39, 4.44, 13.62 }, - { 1, char40, 21.8657, 47.1733 }, - { 1, char41, 24.3276, 47.5333 }, - { 3, char42, 30.7695, 59.439 }, - { 2, char43, 48.8371, 97.2543 }, - { 1, char44, 13.5219, 26.0638 }, - { 1, char45, 50.2371, 100.754 }, - { 1, char46, 13.1019, 26.4838 }, - { 1, char47, 40.5733, 82.1067 }, - { 1, char48, 38.3133, 77.0667 }, - { 1, char49, 30.8676, 66.5295 }, - { 1, char50, 38.7533, 77.6467 }, - { 1, char51, 38.3333, 77.0467 }, - { 2, char52, 37.2133, 80.1686 }, - { 1, char53, 38.1933, 77.6867 }, - { 1, char54, 34.1514, 73.8048 }, - { 2, char55, 38.8933, 77.2267 }, - { 1, char56, 38.9333, 77.6667 }, - { 1, char57, 39.9333, 74.0648 }, - { 2, char58, 14.0819, 26.2238 }, - { 2, char59, 12.9619, 26.3038 }, - { 1, char60, 41.1552, 81.6105 }, - { 2, char61, 48.5571, 97.2543 }, - { 1, char62, 40.8752, 81.6105 }, - { 2, char63, 36.9914, 73.9029 }, - { 2, char64, 34.9314, 74.3648 }, - { 3, char65, 40.5952, 80.4905 }, - { 3, char66, 44.7533, 83.6267 }, - { 1, char67, 39.9933, 84.4886 }, - { 2, char68, 45.2933, 85.2867 }, - { 4, char69, 39.9914, 78.1848 }, - { 3, char70, 39.9914, 78.7448 }, - { 2, char71, 40.3933, 89.7686 }, - { 3, char72, 44.7533, 89.0867 }, - { 1, char73, 10.86, 21.3 }, - { 1, char74, 31.0714, 59.999 }, - { 3, char75, 44.6133, 79.3267 }, - { 2, char76, 40.2514, 71.3229 }, - { 4, char77, 48.9552, 97.2105 }, - { 3, char78, 44.4733, 88.8067 }, - { 1, char79, 44.3352, 88.8305 }, - { 2, char80, 45.4333, 85.6667 }, - { 2, char81, 43.3952, 88.0905 }, - { 3, char82, 45.0133, 82.3667 }, - { 1, char83, 41.3333, 80.8267 }, - { 2, char84, 35.6933, 71.9467 }, - { 1, char85, 44.8733, 89.4867 }, - { 2, char86, 40.4552, 81.6105 }, - { 4, char87, 49.839, 100.518 }, - { 2, char88, 35.8333, 72.3667 }, - { 2, char89, 39.6152, 79.6505 }, - { 3, char90, 35.8333, 73.7467 }, - { 4, char91, 22.0657, 46.1133 }, - { 1, char92, 39.1733, 78.2067 }, - { 4, char93, 23.4876, 46.3933 }, - { 2, char94, 44.0752, 90.2305 }, - { 1, char95, 51.281, 104.062 }, - { 2, char96, 42.5457, 83.5714 }, - { 2, char97, 35.2514, 66.6029 }, - { 2, char98, 37.3314, 70.4629 }, - { 1, char99, 34.0914, 68.9229 }, - { 2, char100, 33.2114, 70.2629 }, - { 1, char101, 34.2914, 68.5229 }, - { 2, char102, 14.9657, 38.6552 }, - { 2, char103, 33.9314, 70.9829 }, - { 2, char104, 33.4095, 71.021 }, - { 2, char105, 14.7819, 28.8638 }, - { 2, char106, 17.3876, 36.2314 }, - { 3, char107, 33.4095, 62.521 }, - { 1, char108, 10.02, 19.34 }, - { 3, char109, 61.981, 123.962 }, - { 2, char110, 32.9895, 70.881 }, - { 1, char111, 33.5514, 71.7448 }, - { 2, char112, 38.0314, 70.8029 }, - { 2, char113, 33.4114, 70.7429 }, - { 2, char114, 23.7457, 49.4952 }, - { 1, char115, 28.5095, 62.321 }, - { 2, char116, 14.8257, 39.3152 }, - { 2, char117, 33.2695, 71.161 }, - { 2, char118, 30.3714, 60.6029 }, - { 4, char119, 40.5952, 80.4905 }, - { 2, char120, 25.4695, 56.401 }, - { 2, char121, 35.1333, 66.0648 }, - { 3, char122, 28.2495, 61.821 }, - { 3, char123, 21.6657, 41.6295 }, - { 1, char124, 11.54, 23.78 }, - { 3, char125, 18.7038, 41.4695 }, - { 2, char126, 45.7771, 91.2743 }, - { 2, char127, 33.3333, 66.6667 }, -}; - -StrokeFontRec glutStrokeRoman = { "Roman", 128, chars, 119.048, -33.3333 }; - + +/* GENERATED FILE -- DO NOT MODIFY */ + +#include "glutstroke.h" + +/* char: 33 '!' */ + +static const CoordRec char33_stroke0[] = { + { 13.3819, 100 }, + { 13.3819, 33.3333 }, +}; + +static const CoordRec char33_stroke1[] = { + { 13.3819, 9.5238 }, + { 8.62, 4.7619 }, + { 13.3819, 0 }, + { 18.1438, 4.7619 }, + { 13.3819, 9.5238 }, +}; + +static const StrokeRec char33[] = { + { 2, char33_stroke0 }, + { 5, char33_stroke1 }, +}; + +/* char: 34 '"' */ + +static const CoordRec char34_stroke0[] = { + { 4.02, 100 }, + { 4.02, 66.6667 }, +}; + +static const CoordRec char34_stroke1[] = { + { 42.1152, 100 }, + { 42.1152, 66.6667 }, +}; + +static const StrokeRec char34[] = { + { 2, char34_stroke0 }, + { 2, char34_stroke1 }, +}; + +/* char: 35 '#' */ + +static const CoordRec char35_stroke0[] = { + { 41.2952, 119.048 }, + { 7.9619, -33.3333 }, +}; + +static const CoordRec char35_stroke1[] = { + { 69.8667, 119.048 }, + { 36.5333, -33.3333 }, +}; + +static const CoordRec char35_stroke2[] = { + { 7.9619, 57.1429 }, + { 74.6286, 57.1429 }, +}; + +static const CoordRec char35_stroke3[] = { + { 3.2, 28.5714 }, + { 69.8667, 28.5714 }, +}; + +static const StrokeRec char35[] = { + { 2, char35_stroke0 }, + { 2, char35_stroke1 }, + { 2, char35_stroke2 }, + { 2, char35_stroke3 }, +}; + +/* char: 36 '$' */ + +static const CoordRec char36_stroke0[] = { + { 28.6295, 119.048 }, + { 28.6295, -19.0476 }, +}; + +static const CoordRec char36_stroke1[] = { + { 47.6771, 119.048 }, + { 47.6771, -19.0476 }, +}; + +static const CoordRec char36_stroke2[] = { + { 71.4867, 85.7143 }, + { 61.9629, 95.2381 }, + { 47.6771, 100 }, + { 28.6295, 100 }, + { 14.3438, 95.2381 }, + { 4.82, 85.7143 }, + { 4.82, 76.1905 }, + { 9.5819, 66.6667 }, + { 14.3438, 61.9048 }, + { 23.8676, 57.1429 }, + { 52.439, 47.619 }, + { 61.9629, 42.8571 }, + { 66.7248, 38.0952 }, + { 71.4867, 28.5714 }, + { 71.4867, 14.2857 }, + { 61.9629, 4.7619 }, + { 47.6771, 0 }, + { 28.6295, 0 }, + { 14.3438, 4.7619 }, + { 4.82, 14.2857 }, +}; + +static const StrokeRec char36[] = { + { 2, char36_stroke0 }, + { 2, char36_stroke1 }, + { 20, char36_stroke2 }, +}; + +/* char: 37 '%' */ + +static const CoordRec char37_stroke0[] = { + { 92.0743, 100 }, + { 6.36, 0 }, +}; + +static const CoordRec char37_stroke1[] = { + { 30.1695, 100 }, + { 39.6933, 90.4762 }, + { 39.6933, 80.9524 }, + { 34.9314, 71.4286 }, + { 25.4076, 66.6667 }, + { 15.8838, 66.6667 }, + { 6.36, 76.1905 }, + { 6.36, 85.7143 }, + { 11.1219, 95.2381 }, + { 20.6457, 100 }, + { 30.1695, 100 }, + { 39.6933, 95.2381 }, + { 53.979, 90.4762 }, + { 68.2648, 90.4762 }, + { 82.5505, 95.2381 }, + { 92.0743, 100 }, +}; + +static const CoordRec char37_stroke2[] = { + { 73.0267, 33.3333 }, + { 63.5029, 28.5714 }, + { 58.741, 19.0476 }, + { 58.741, 9.5238 }, + { 68.2648, 0 }, + { 77.7886, 0 }, + { 87.3124, 4.7619 }, + { 92.0743, 14.2857 }, + { 92.0743, 23.8095 }, + { 82.5505, 33.3333 }, + { 73.0267, 33.3333 }, +}; + +static const StrokeRec char37[] = { + { 2, char37_stroke0 }, + { 16, char37_stroke1 }, + { 11, char37_stroke2 }, +}; + +/* char: 38 '&' */ + +static const CoordRec char38_stroke0[] = { + { 101.218, 57.1429 }, + { 101.218, 61.9048 }, + { 96.4562, 66.6667 }, + { 91.6943, 66.6667 }, + { 86.9324, 61.9048 }, + { 82.1705, 52.381 }, + { 72.6467, 28.5714 }, + { 63.1229, 14.2857 }, + { 53.599, 4.7619 }, + { 44.0752, 0 }, + { 25.0276, 0 }, + { 15.5038, 4.7619 }, + { 10.7419, 9.5238 }, + { 5.98, 19.0476 }, + { 5.98, 28.5714 }, + { 10.7419, 38.0952 }, + { 15.5038, 42.8571 }, + { 48.8371, 61.9048 }, + { 53.599, 66.6667 }, + { 58.361, 76.1905 }, + { 58.361, 85.7143 }, + { 53.599, 95.2381 }, + { 44.0752, 100 }, + { 34.5514, 95.2381 }, + { 29.7895, 85.7143 }, + { 29.7895, 76.1905 }, + { 34.5514, 61.9048 }, + { 44.0752, 47.619 }, + { 67.8848, 14.2857 }, + { 77.4086, 4.7619 }, + { 86.9324, 0 }, + { 96.4562, 0 }, + { 101.218, 4.7619 }, + { 101.218, 9.5238 }, +}; + +static const StrokeRec char38[] = { + { 34, char38_stroke0 }, +}; + +/* char: 39 ''' */ + +static const CoordRec char39_stroke0[] = { + { 4.44, 100 }, + { 4.44, 66.6667 }, +}; + +static const StrokeRec char39[] = { + { 2, char39_stroke0 }, +}; + +/* char: 40 '(' */ + +static const CoordRec char40_stroke0[] = { + { 40.9133, 119.048 }, + { 31.3895, 109.524 }, + { 21.8657, 95.2381 }, + { 12.3419, 76.1905 }, + { 7.58, 52.381 }, + { 7.58, 33.3333 }, + { 12.3419, 9.5238 }, + { 21.8657, -9.5238 }, + { 31.3895, -23.8095 }, + { 40.9133, -33.3333 }, +}; + +static const StrokeRec char40[] = { + { 10, char40_stroke0 }, +}; + +/* char: 41 ')' */ + +static const CoordRec char41_stroke0[] = { + { 5.28, 119.048 }, + { 14.8038, 109.524 }, + { 24.3276, 95.2381 }, + { 33.8514, 76.1905 }, + { 38.6133, 52.381 }, + { 38.6133, 33.3333 }, + { 33.8514, 9.5238 }, + { 24.3276, -9.5238 }, + { 14.8038, -23.8095 }, + { 5.28, -33.3333 }, +}; + +static const StrokeRec char41[] = { + { 10, char41_stroke0 }, +}; + +/* char: 42 '*' */ + +static const CoordRec char42_stroke0[] = { + { 30.7695, 71.4286 }, + { 30.7695, 14.2857 }, +}; + +static const CoordRec char42_stroke1[] = { + { 6.96, 57.1429 }, + { 54.579, 28.5714 }, +}; + +static const CoordRec char42_stroke2[] = { + { 54.579, 57.1429 }, + { 6.96, 28.5714 }, +}; + +static const StrokeRec char42[] = { + { 2, char42_stroke0 }, + { 2, char42_stroke1 }, + { 2, char42_stroke2 }, +}; + +/* char: 43 '+' */ + +static const CoordRec char43_stroke0[] = { + { 48.8371, 85.7143 }, + { 48.8371, 0 }, +}; + +static const CoordRec char43_stroke1[] = { + { 5.98, 42.8571 }, + { 91.6943, 42.8571 }, +}; + +static const StrokeRec char43[] = { + { 2, char43_stroke0 }, + { 2, char43_stroke1 }, +}; + +/* char: 44 ',' */ + +static const CoordRec char44_stroke0[] = { + { 18.2838, 4.7619 }, + { 13.5219, 0 }, + { 8.76, 4.7619 }, + { 13.5219, 9.5238 }, + { 18.2838, 4.7619 }, + { 18.2838, -4.7619 }, + { 13.5219, -14.2857 }, + { 8.76, -19.0476 }, +}; + +static const StrokeRec char44[] = { + { 8, char44_stroke0 }, +}; + +/* char: 45 '-' */ + +static const CoordRec char45_stroke0[] = { + { 7.38, 42.8571 }, + { 93.0943, 42.8571 }, +}; + +static const StrokeRec char45[] = { + { 2, char45_stroke0 }, +}; + +/* char: 46 '.' */ + +static const CoordRec char46_stroke0[] = { + { 13.1019, 9.5238 }, + { 8.34, 4.7619 }, + { 13.1019, 0 }, + { 17.8638, 4.7619 }, + { 13.1019, 9.5238 }, +}; + +static const StrokeRec char46[] = { + { 5, char46_stroke0 }, +}; + +/* char: 47 '/' */ + +static const CoordRec char47_stroke0[] = { + { 7.24, -14.2857 }, + { 73.9067, 100 }, +}; + +static const StrokeRec char47[] = { + { 2, char47_stroke0 }, +}; + +/* char: 48 '0' */ + +static const CoordRec char48_stroke0[] = { + { 33.5514, 100 }, + { 19.2657, 95.2381 }, + { 9.7419, 80.9524 }, + { 4.98, 57.1429 }, + { 4.98, 42.8571 }, + { 9.7419, 19.0476 }, + { 19.2657, 4.7619 }, + { 33.5514, 0 }, + { 43.0752, 0 }, + { 57.361, 4.7619 }, + { 66.8848, 19.0476 }, + { 71.6467, 42.8571 }, + { 71.6467, 57.1429 }, + { 66.8848, 80.9524 }, + { 57.361, 95.2381 }, + { 43.0752, 100 }, + { 33.5514, 100 }, +}; + +static const StrokeRec char48[] = { + { 17, char48_stroke0 }, +}; + +/* char: 49 '1' */ + +static const CoordRec char49_stroke0[] = { + { 11.82, 80.9524 }, + { 21.3438, 85.7143 }, + { 35.6295, 100 }, + { 35.6295, 0 }, +}; + +static const StrokeRec char49[] = { + { 4, char49_stroke0 }, +}; + +/* char: 50 '2' */ + +static const CoordRec char50_stroke0[] = { + { 10.1819, 76.1905 }, + { 10.1819, 80.9524 }, + { 14.9438, 90.4762 }, + { 19.7057, 95.2381 }, + { 29.2295, 100 }, + { 48.2771, 100 }, + { 57.801, 95.2381 }, + { 62.5629, 90.4762 }, + { 67.3248, 80.9524 }, + { 67.3248, 71.4286 }, + { 62.5629, 61.9048 }, + { 53.039, 47.619 }, + { 5.42, 0 }, + { 72.0867, 0 }, +}; + +static const StrokeRec char50[] = { + { 14, char50_stroke0 }, +}; + +/* char: 51 '3' */ + +static const CoordRec char51_stroke0[] = { + { 14.5238, 100 }, + { 66.9048, 100 }, + { 38.3333, 61.9048 }, + { 52.619, 61.9048 }, + { 62.1429, 57.1429 }, + { 66.9048, 52.381 }, + { 71.6667, 38.0952 }, + { 71.6667, 28.5714 }, + { 66.9048, 14.2857 }, + { 57.381, 4.7619 }, + { 43.0952, 0 }, + { 28.8095, 0 }, + { 14.5238, 4.7619 }, + { 9.7619, 9.5238 }, + { 5, 19.0476 }, +}; + +static const StrokeRec char51[] = { + { 15, char51_stroke0 }, +}; + +/* char: 52 '4' */ + +static const CoordRec char52_stroke0[] = { + { 51.499, 100 }, + { 3.88, 33.3333 }, + { 75.3086, 33.3333 }, +}; + +static const CoordRec char52_stroke1[] = { + { 51.499, 100 }, + { 51.499, 0 }, +}; + +static const StrokeRec char52[] = { + { 3, char52_stroke0 }, + { 2, char52_stroke1 }, +}; + +/* char: 53 '5' */ + +static const CoordRec char53_stroke0[] = { + { 62.0029, 100 }, + { 14.3838, 100 }, + { 9.6219, 57.1429 }, + { 14.3838, 61.9048 }, + { 28.6695, 66.6667 }, + { 42.9552, 66.6667 }, + { 57.241, 61.9048 }, + { 66.7648, 52.381 }, + { 71.5267, 38.0952 }, + { 71.5267, 28.5714 }, + { 66.7648, 14.2857 }, + { 57.241, 4.7619 }, + { 42.9552, 0 }, + { 28.6695, 0 }, + { 14.3838, 4.7619 }, + { 9.6219, 9.5238 }, + { 4.86, 19.0476 }, +}; + +static const StrokeRec char53[] = { + { 17, char53_stroke0 }, +}; + +/* char: 54 '6' */ + +static const CoordRec char54_stroke0[] = { + { 62.7229, 85.7143 }, + { 57.961, 95.2381 }, + { 43.6752, 100 }, + { 34.1514, 100 }, + { 19.8657, 95.2381 }, + { 10.3419, 80.9524 }, + { 5.58, 57.1429 }, + { 5.58, 33.3333 }, + { 10.3419, 14.2857 }, + { 19.8657, 4.7619 }, + { 34.1514, 0 }, + { 38.9133, 0 }, + { 53.199, 4.7619 }, + { 62.7229, 14.2857 }, + { 67.4848, 28.5714 }, + { 67.4848, 33.3333 }, + { 62.7229, 47.619 }, + { 53.199, 57.1429 }, + { 38.9133, 61.9048 }, + { 34.1514, 61.9048 }, + { 19.8657, 57.1429 }, + { 10.3419, 47.619 }, + { 5.58, 33.3333 }, +}; + +static const StrokeRec char54[] = { + { 23, char54_stroke0 }, +}; + +/* char: 55 '7' */ + +static const CoordRec char55_stroke0[] = { + { 72.2267, 100 }, + { 24.6076, 0 }, +}; + +static const CoordRec char55_stroke1[] = { + { 5.56, 100 }, + { 72.2267, 100 }, +}; + +static const StrokeRec char55[] = { + { 2, char55_stroke0 }, + { 2, char55_stroke1 }, +}; + +/* char: 56 '8' */ + +static const CoordRec char56_stroke0[] = { + { 29.4095, 100 }, + { 15.1238, 95.2381 }, + { 10.3619, 85.7143 }, + { 10.3619, 76.1905 }, + { 15.1238, 66.6667 }, + { 24.6476, 61.9048 }, + { 43.6952, 57.1429 }, + { 57.981, 52.381 }, + { 67.5048, 42.8571 }, + { 72.2667, 33.3333 }, + { 72.2667, 19.0476 }, + { 67.5048, 9.5238 }, + { 62.7429, 4.7619 }, + { 48.4571, 0 }, + { 29.4095, 0 }, + { 15.1238, 4.7619 }, + { 10.3619, 9.5238 }, + { 5.6, 19.0476 }, + { 5.6, 33.3333 }, + { 10.3619, 42.8571 }, + { 19.8857, 52.381 }, + { 34.1714, 57.1429 }, + { 53.219, 61.9048 }, + { 62.7429, 66.6667 }, + { 67.5048, 76.1905 }, + { 67.5048, 85.7143 }, + { 62.7429, 95.2381 }, + { 48.4571, 100 }, + { 29.4095, 100 }, +}; + +static const StrokeRec char56[] = { + { 29, char56_stroke0 }, +}; + +/* char: 57 '9' */ + +static const CoordRec char57_stroke0[] = { + { 68.5048, 66.6667 }, + { 63.7429, 52.381 }, + { 54.219, 42.8571 }, + { 39.9333, 38.0952 }, + { 35.1714, 38.0952 }, + { 20.8857, 42.8571 }, + { 11.3619, 52.381 }, + { 6.6, 66.6667 }, + { 6.6, 71.4286 }, + { 11.3619, 85.7143 }, + { 20.8857, 95.2381 }, + { 35.1714, 100 }, + { 39.9333, 100 }, + { 54.219, 95.2381 }, + { 63.7429, 85.7143 }, + { 68.5048, 66.6667 }, + { 68.5048, 42.8571 }, + { 63.7429, 19.0476 }, + { 54.219, 4.7619 }, + { 39.9333, 0 }, + { 30.4095, 0 }, + { 16.1238, 4.7619 }, + { 11.3619, 14.2857 }, +}; + +static const StrokeRec char57[] = { + { 23, char57_stroke0 }, +}; + +/* char: 58 ':' */ + +static const CoordRec char58_stroke0[] = { + { 14.0819, 66.6667 }, + { 9.32, 61.9048 }, + { 14.0819, 57.1429 }, + { 18.8438, 61.9048 }, + { 14.0819, 66.6667 }, +}; + +static const CoordRec char58_stroke1[] = { + { 14.0819, 9.5238 }, + { 9.32, 4.7619 }, + { 14.0819, 0 }, + { 18.8438, 4.7619 }, + { 14.0819, 9.5238 }, +}; + +static const StrokeRec char58[] = { + { 5, char58_stroke0 }, + { 5, char58_stroke1 }, +}; + +/* char: 59 ';' */ + +static const CoordRec char59_stroke0[] = { + { 12.9619, 66.6667 }, + { 8.2, 61.9048 }, + { 12.9619, 57.1429 }, + { 17.7238, 61.9048 }, + { 12.9619, 66.6667 }, +}; + +static const CoordRec char59_stroke1[] = { + { 17.7238, 4.7619 }, + { 12.9619, 0 }, + { 8.2, 4.7619 }, + { 12.9619, 9.5238 }, + { 17.7238, 4.7619 }, + { 17.7238, -4.7619 }, + { 12.9619, -14.2857 }, + { 8.2, -19.0476 }, +}; + +static const StrokeRec char59[] = { + { 5, char59_stroke0 }, + { 8, char59_stroke1 }, +}; + +/* char: 60 '<' */ + +static const CoordRec char60_stroke0[] = { + { 79.2505, 85.7143 }, + { 3.06, 42.8571 }, + { 79.2505, 0 }, +}; + +static const StrokeRec char60[] = { + { 3, char60_stroke0 }, +}; + +/* char: 61 '=' */ + +static const CoordRec char61_stroke0[] = { + { 5.7, 57.1429 }, + { 91.4143, 57.1429 }, +}; + +static const CoordRec char61_stroke1[] = { + { 5.7, 28.5714 }, + { 91.4143, 28.5714 }, +}; + +static const StrokeRec char61[] = { + { 2, char61_stroke0 }, + { 2, char61_stroke1 }, +}; + +/* char: 62 '>' */ + +static const CoordRec char62_stroke0[] = { + { 2.78, 85.7143 }, + { 78.9705, 42.8571 }, + { 2.78, 0 }, +}; + +static const StrokeRec char62[] = { + { 3, char62_stroke0 }, +}; + +/* char: 63 '?' */ + +static const CoordRec char63_stroke0[] = { + { 8.42, 76.1905 }, + { 8.42, 80.9524 }, + { 13.1819, 90.4762 }, + { 17.9438, 95.2381 }, + { 27.4676, 100 }, + { 46.5152, 100 }, + { 56.039, 95.2381 }, + { 60.801, 90.4762 }, + { 65.5629, 80.9524 }, + { 65.5629, 71.4286 }, + { 60.801, 61.9048 }, + { 56.039, 57.1429 }, + { 36.9914, 47.619 }, + { 36.9914, 33.3333 }, +}; + +static const CoordRec char63_stroke1[] = { + { 36.9914, 9.5238 }, + { 32.2295, 4.7619 }, + { 36.9914, 0 }, + { 41.7533, 4.7619 }, + { 36.9914, 9.5238 }, +}; + +static const StrokeRec char63[] = { + { 14, char63_stroke0 }, + { 5, char63_stroke1 }, +}; + +/* char: 64 '@' */ + +static const CoordRec char64_stroke0[] = { + { 49.2171, 52.381 }, + { 39.6933, 57.1429 }, + { 30.1695, 57.1429 }, + { 25.4076, 47.619 }, + { 25.4076, 42.8571 }, + { 30.1695, 33.3333 }, + { 39.6933, 33.3333 }, + { 49.2171, 38.0952 }, +}; + +static const CoordRec char64_stroke1[] = { + { 49.2171, 57.1429 }, + { 49.2171, 38.0952 }, + { 53.979, 33.3333 }, + { 63.5029, 33.3333 }, + { 68.2648, 42.8571 }, + { 68.2648, 47.619 }, + { 63.5029, 61.9048 }, + { 53.979, 71.4286 }, + { 39.6933, 76.1905 }, + { 34.9314, 76.1905 }, + { 20.6457, 71.4286 }, + { 11.1219, 61.9048 }, + { 6.36, 47.619 }, + { 6.36, 42.8571 }, + { 11.1219, 28.5714 }, + { 20.6457, 19.0476 }, + { 34.9314, 14.2857 }, + { 39.6933, 14.2857 }, + { 53.979, 19.0476 }, +}; + +static const StrokeRec char64[] = { + { 8, char64_stroke0 }, + { 19, char64_stroke1 }, +}; + +/* char: 65 'A' */ + +static const CoordRec char65_stroke0[] = { + { 40.5952, 100 }, + { 2.5, 0 }, +}; + +static const CoordRec char65_stroke1[] = { + { 40.5952, 100 }, + { 78.6905, 0 }, +}; + +static const CoordRec char65_stroke2[] = { + { 16.7857, 33.3333 }, + { 64.4048, 33.3333 }, +}; + +static const StrokeRec char65[] = { + { 2, char65_stroke0 }, + { 2, char65_stroke1 }, + { 2, char65_stroke2 }, +}; + +/* char: 66 'B' */ + +static const CoordRec char66_stroke0[] = { + { 11.42, 100 }, + { 11.42, 0 }, +}; + +static const CoordRec char66_stroke1[] = { + { 11.42, 100 }, + { 54.2771, 100 }, + { 68.5629, 95.2381 }, + { 73.3248, 90.4762 }, + { 78.0867, 80.9524 }, + { 78.0867, 71.4286 }, + { 73.3248, 61.9048 }, + { 68.5629, 57.1429 }, + { 54.2771, 52.381 }, +}; + +static const CoordRec char66_stroke2[] = { + { 11.42, 52.381 }, + { 54.2771, 52.381 }, + { 68.5629, 47.619 }, + { 73.3248, 42.8571 }, + { 78.0867, 33.3333 }, + { 78.0867, 19.0476 }, + { 73.3248, 9.5238 }, + { 68.5629, 4.7619 }, + { 54.2771, 0 }, + { 11.42, 0 }, +}; + +static const StrokeRec char66[] = { + { 2, char66_stroke0 }, + { 9, char66_stroke1 }, + { 10, char66_stroke2 }, +}; + +/* char: 67 'C' */ + +static const CoordRec char67_stroke0[] = { + { 78.0886, 76.1905 }, + { 73.3267, 85.7143 }, + { 63.8029, 95.2381 }, + { 54.279, 100 }, + { 35.2314, 100 }, + { 25.7076, 95.2381 }, + { 16.1838, 85.7143 }, + { 11.4219, 76.1905 }, + { 6.66, 61.9048 }, + { 6.66, 38.0952 }, + { 11.4219, 23.8095 }, + { 16.1838, 14.2857 }, + { 25.7076, 4.7619 }, + { 35.2314, 0 }, + { 54.279, 0 }, + { 63.8029, 4.7619 }, + { 73.3267, 14.2857 }, + { 78.0886, 23.8095 }, +}; + +static const StrokeRec char67[] = { + { 18, char67_stroke0 }, +}; + +/* char: 68 'D' */ + +static const CoordRec char68_stroke0[] = { + { 11.96, 100 }, + { 11.96, 0 }, +}; + +static const CoordRec char68_stroke1[] = { + { 11.96, 100 }, + { 45.2933, 100 }, + { 59.579, 95.2381 }, + { 69.1029, 85.7143 }, + { 73.8648, 76.1905 }, + { 78.6267, 61.9048 }, + { 78.6267, 38.0952 }, + { 73.8648, 23.8095 }, + { 69.1029, 14.2857 }, + { 59.579, 4.7619 }, + { 45.2933, 0 }, + { 11.96, 0 }, +}; + +static const StrokeRec char68[] = { + { 2, char68_stroke0 }, + { 12, char68_stroke1 }, +}; + +/* char: 69 'E' */ + +static const CoordRec char69_stroke0[] = { + { 11.42, 100 }, + { 11.42, 0 }, +}; + +static const CoordRec char69_stroke1[] = { + { 11.42, 100 }, + { 73.3248, 100 }, +}; + +static const CoordRec char69_stroke2[] = { + { 11.42, 52.381 }, + { 49.5152, 52.381 }, +}; + +static const CoordRec char69_stroke3[] = { + { 11.42, 0 }, + { 73.3248, 0 }, +}; + +static const StrokeRec char69[] = { + { 2, char69_stroke0 }, + { 2, char69_stroke1 }, + { 2, char69_stroke2 }, + { 2, char69_stroke3 }, +}; + +/* char: 70 'F' */ + +static const CoordRec char70_stroke0[] = { + { 11.42, 100 }, + { 11.42, 0 }, +}; + +static const CoordRec char70_stroke1[] = { + { 11.42, 100 }, + { 73.3248, 100 }, +}; + +static const CoordRec char70_stroke2[] = { + { 11.42, 52.381 }, + { 49.5152, 52.381 }, +}; + +static const StrokeRec char70[] = { + { 2, char70_stroke0 }, + { 2, char70_stroke1 }, + { 2, char70_stroke2 }, +}; + +/* char: 71 'G' */ + +static const CoordRec char71_stroke0[] = { + { 78.4886, 76.1905 }, + { 73.7267, 85.7143 }, + { 64.2029, 95.2381 }, + { 54.679, 100 }, + { 35.6314, 100 }, + { 26.1076, 95.2381 }, + { 16.5838, 85.7143 }, + { 11.8219, 76.1905 }, + { 7.06, 61.9048 }, + { 7.06, 38.0952 }, + { 11.8219, 23.8095 }, + { 16.5838, 14.2857 }, + { 26.1076, 4.7619 }, + { 35.6314, 0 }, + { 54.679, 0 }, + { 64.2029, 4.7619 }, + { 73.7267, 14.2857 }, + { 78.4886, 23.8095 }, + { 78.4886, 38.0952 }, +}; + +static const CoordRec char71_stroke1[] = { + { 54.679, 38.0952 }, + { 78.4886, 38.0952 }, +}; + +static const StrokeRec char71[] = { + { 19, char71_stroke0 }, + { 2, char71_stroke1 }, +}; + +/* char: 72 'H' */ + +static const CoordRec char72_stroke0[] = { + { 11.42, 100 }, + { 11.42, 0 }, +}; + +static const CoordRec char72_stroke1[] = { + { 78.0867, 100 }, + { 78.0867, 0 }, +}; + +static const CoordRec char72_stroke2[] = { + { 11.42, 52.381 }, + { 78.0867, 52.381 }, +}; + +static const StrokeRec char72[] = { + { 2, char72_stroke0 }, + { 2, char72_stroke1 }, + { 2, char72_stroke2 }, +}; + +/* char: 73 'I' */ + +static const CoordRec char73_stroke0[] = { + { 10.86, 100 }, + { 10.86, 0 }, +}; + +static const StrokeRec char73[] = { + { 2, char73_stroke0 }, +}; + +/* char: 74 'J' */ + +static const CoordRec char74_stroke0[] = { + { 50.119, 100 }, + { 50.119, 23.8095 }, + { 45.3571, 9.5238 }, + { 40.5952, 4.7619 }, + { 31.0714, 0 }, + { 21.5476, 0 }, + { 12.0238, 4.7619 }, + { 7.2619, 9.5238 }, + { 2.5, 23.8095 }, + { 2.5, 33.3333 }, +}; + +static const StrokeRec char74[] = { + { 10, char74_stroke0 }, +}; + +/* char: 75 'K' */ + +static const CoordRec char75_stroke0[] = { + { 11.28, 100 }, + { 11.28, 0 }, +}; + +static const CoordRec char75_stroke1[] = { + { 77.9467, 100 }, + { 11.28, 33.3333 }, +}; + +static const CoordRec char75_stroke2[] = { + { 35.0895, 57.1429 }, + { 77.9467, 0 }, +}; + +static const StrokeRec char75[] = { + { 2, char75_stroke0 }, + { 2, char75_stroke1 }, + { 2, char75_stroke2 }, +}; + +/* char: 76 'L' */ + +static const CoordRec char76_stroke0[] = { + { 11.68, 100 }, + { 11.68, 0 }, +}; + +static const CoordRec char76_stroke1[] = { + { 11.68, 0 }, + { 68.8229, 0 }, +}; + +static const StrokeRec char76[] = { + { 2, char76_stroke0 }, + { 2, char76_stroke1 }, +}; + +/* char: 77 'M' */ + +static const CoordRec char77_stroke0[] = { + { 10.86, 100 }, + { 10.86, 0 }, +}; + +static const CoordRec char77_stroke1[] = { + { 10.86, 100 }, + { 48.9552, 0 }, +}; + +static const CoordRec char77_stroke2[] = { + { 87.0505, 100 }, + { 48.9552, 0 }, +}; + +static const CoordRec char77_stroke3[] = { + { 87.0505, 100 }, + { 87.0505, 0 }, +}; + +static const StrokeRec char77[] = { + { 2, char77_stroke0 }, + { 2, char77_stroke1 }, + { 2, char77_stroke2 }, + { 2, char77_stroke3 }, +}; + +/* char: 78 'N' */ + +static const CoordRec char78_stroke0[] = { + { 11.14, 100 }, + { 11.14, 0 }, +}; + +static const CoordRec char78_stroke1[] = { + { 11.14, 100 }, + { 77.8067, 0 }, +}; + +static const CoordRec char78_stroke2[] = { + { 77.8067, 100 }, + { 77.8067, 0 }, +}; + +static const StrokeRec char78[] = { + { 2, char78_stroke0 }, + { 2, char78_stroke1 }, + { 2, char78_stroke2 }, +}; + +/* char: 79 'O' */ + +static const CoordRec char79_stroke0[] = { + { 34.8114, 100 }, + { 25.2876, 95.2381 }, + { 15.7638, 85.7143 }, + { 11.0019, 76.1905 }, + { 6.24, 61.9048 }, + { 6.24, 38.0952 }, + { 11.0019, 23.8095 }, + { 15.7638, 14.2857 }, + { 25.2876, 4.7619 }, + { 34.8114, 0 }, + { 53.859, 0 }, + { 63.3829, 4.7619 }, + { 72.9067, 14.2857 }, + { 77.6686, 23.8095 }, + { 82.4305, 38.0952 }, + { 82.4305, 61.9048 }, + { 77.6686, 76.1905 }, + { 72.9067, 85.7143 }, + { 63.3829, 95.2381 }, + { 53.859, 100 }, + { 34.8114, 100 }, +}; + +static const StrokeRec char79[] = { + { 21, char79_stroke0 }, +}; + +/* char: 80 'P' */ + +static const CoordRec char80_stroke0[] = { + { 12.1, 100 }, + { 12.1, 0 }, +}; + +static const CoordRec char80_stroke1[] = { + { 12.1, 100 }, + { 54.9571, 100 }, + { 69.2429, 95.2381 }, + { 74.0048, 90.4762 }, + { 78.7667, 80.9524 }, + { 78.7667, 66.6667 }, + { 74.0048, 57.1429 }, + { 69.2429, 52.381 }, + { 54.9571, 47.619 }, + { 12.1, 47.619 }, +}; + +static const StrokeRec char80[] = { + { 2, char80_stroke0 }, + { 10, char80_stroke1 }, +}; + +/* char: 81 'Q' */ + +static const CoordRec char81_stroke0[] = { + { 33.8714, 100 }, + { 24.3476, 95.2381 }, + { 14.8238, 85.7143 }, + { 10.0619, 76.1905 }, + { 5.3, 61.9048 }, + { 5.3, 38.0952 }, + { 10.0619, 23.8095 }, + { 14.8238, 14.2857 }, + { 24.3476, 4.7619 }, + { 33.8714, 0 }, + { 52.919, 0 }, + { 62.4429, 4.7619 }, + { 71.9667, 14.2857 }, + { 76.7286, 23.8095 }, + { 81.4905, 38.0952 }, + { 81.4905, 61.9048 }, + { 76.7286, 76.1905 }, + { 71.9667, 85.7143 }, + { 62.4429, 95.2381 }, + { 52.919, 100 }, + { 33.8714, 100 }, +}; + +static const CoordRec char81_stroke1[] = { + { 48.1571, 19.0476 }, + { 76.7286, -9.5238 }, +}; + +static const StrokeRec char81[] = { + { 21, char81_stroke0 }, + { 2, char81_stroke1 }, +}; + +/* char: 82 'R' */ + +static const CoordRec char82_stroke0[] = { + { 11.68, 100 }, + { 11.68, 0 }, +}; + +static const CoordRec char82_stroke1[] = { + { 11.68, 100 }, + { 54.5371, 100 }, + { 68.8229, 95.2381 }, + { 73.5848, 90.4762 }, + { 78.3467, 80.9524 }, + { 78.3467, 71.4286 }, + { 73.5848, 61.9048 }, + { 68.8229, 57.1429 }, + { 54.5371, 52.381 }, + { 11.68, 52.381 }, +}; + +static const CoordRec char82_stroke2[] = { + { 45.0133, 52.381 }, + { 78.3467, 0 }, +}; + +static const StrokeRec char82[] = { + { 2, char82_stroke0 }, + { 10, char82_stroke1 }, + { 2, char82_stroke2 }, +}; + +/* char: 83 'S' */ + +static const CoordRec char83_stroke0[] = { + { 74.6667, 85.7143 }, + { 65.1429, 95.2381 }, + { 50.8571, 100 }, + { 31.8095, 100 }, + { 17.5238, 95.2381 }, + { 8, 85.7143 }, + { 8, 76.1905 }, + { 12.7619, 66.6667 }, + { 17.5238, 61.9048 }, + { 27.0476, 57.1429 }, + { 55.619, 47.619 }, + { 65.1429, 42.8571 }, + { 69.9048, 38.0952 }, + { 74.6667, 28.5714 }, + { 74.6667, 14.2857 }, + { 65.1429, 4.7619 }, + { 50.8571, 0 }, + { 31.8095, 0 }, + { 17.5238, 4.7619 }, + { 8, 14.2857 }, +}; + +static const StrokeRec char83[] = { + { 20, char83_stroke0 }, +}; + +/* char: 84 'T' */ + +static const CoordRec char84_stroke0[] = { + { 35.6933, 100 }, + { 35.6933, 0 }, +}; + +static const CoordRec char84_stroke1[] = { + { 2.36, 100 }, + { 69.0267, 100 }, +}; + +static const StrokeRec char84[] = { + { 2, char84_stroke0 }, + { 2, char84_stroke1 }, +}; + +/* char: 85 'U' */ + +static const CoordRec char85_stroke0[] = { + { 11.54, 100 }, + { 11.54, 28.5714 }, + { 16.3019, 14.2857 }, + { 25.8257, 4.7619 }, + { 40.1114, 0 }, + { 49.6352, 0 }, + { 63.921, 4.7619 }, + { 73.4448, 14.2857 }, + { 78.2067, 28.5714 }, + { 78.2067, 100 }, +}; + +static const StrokeRec char85[] = { + { 10, char85_stroke0 }, +}; + +/* char: 86 'V' */ + +static const CoordRec char86_stroke0[] = { + { 2.36, 100 }, + { 40.4552, 0 }, +}; + +static const CoordRec char86_stroke1[] = { + { 78.5505, 100 }, + { 40.4552, 0 }, +}; + +static const StrokeRec char86[] = { + { 2, char86_stroke0 }, + { 2, char86_stroke1 }, +}; + +/* char: 87 'W' */ + +static const CoordRec char87_stroke0[] = { + { 2.22, 100 }, + { 26.0295, 0 }, +}; + +static const CoordRec char87_stroke1[] = { + { 49.839, 100 }, + { 26.0295, 0 }, +}; + +static const CoordRec char87_stroke2[] = { + { 49.839, 100 }, + { 73.6486, 0 }, +}; + +static const CoordRec char87_stroke3[] = { + { 97.4581, 100 }, + { 73.6486, 0 }, +}; + +static const StrokeRec char87[] = { + { 2, char87_stroke0 }, + { 2, char87_stroke1 }, + { 2, char87_stroke2 }, + { 2, char87_stroke3 }, +}; + +/* char: 88 'X' */ + +static const CoordRec char88_stroke0[] = { + { 2.5, 100 }, + { 69.1667, 0 }, +}; + +static const CoordRec char88_stroke1[] = { + { 69.1667, 100 }, + { 2.5, 0 }, +}; + +static const StrokeRec char88[] = { + { 2, char88_stroke0 }, + { 2, char88_stroke1 }, +}; + +/* char: 89 'Y' */ + +static const CoordRec char89_stroke0[] = { + { 1.52, 100 }, + { 39.6152, 52.381 }, + { 39.6152, 0 }, +}; + +static const CoordRec char89_stroke1[] = { + { 77.7105, 100 }, + { 39.6152, 52.381 }, +}; + +static const StrokeRec char89[] = { + { 3, char89_stroke0 }, + { 2, char89_stroke1 }, +}; + +/* char: 90 'Z' */ + +static const CoordRec char90_stroke0[] = { + { 69.1667, 100 }, + { 2.5, 0 }, +}; + +static const CoordRec char90_stroke1[] = { + { 2.5, 100 }, + { 69.1667, 100 }, +}; + +static const CoordRec char90_stroke2[] = { + { 2.5, 0 }, + { 69.1667, 0 }, +}; + +static const StrokeRec char90[] = { + { 2, char90_stroke0 }, + { 2, char90_stroke1 }, + { 2, char90_stroke2 }, +}; + +/* char: 91 '[' */ + +static const CoordRec char91_stroke0[] = { + { 7.78, 119.048 }, + { 7.78, -33.3333 }, +}; + +static const CoordRec char91_stroke1[] = { + { 12.5419, 119.048 }, + { 12.5419, -33.3333 }, +}; + +static const CoordRec char91_stroke2[] = { + { 7.78, 119.048 }, + { 41.1133, 119.048 }, +}; + +static const CoordRec char91_stroke3[] = { + { 7.78, -33.3333 }, + { 41.1133, -33.3333 }, +}; + +static const StrokeRec char91[] = { + { 2, char91_stroke0 }, + { 2, char91_stroke1 }, + { 2, char91_stroke2 }, + { 2, char91_stroke3 }, +}; + +/* char: 92 '\' */ + +static const CoordRec char92_stroke0[] = { + { 5.84, 100 }, + { 72.5067, -14.2857 }, +}; + +static const StrokeRec char92[] = { + { 2, char92_stroke0 }, +}; + +/* char: 93 ']' */ + +static const CoordRec char93_stroke0[] = { + { 33.0114, 119.048 }, + { 33.0114, -33.3333 }, +}; + +static const CoordRec char93_stroke1[] = { + { 37.7733, 119.048 }, + { 37.7733, -33.3333 }, +}; + +static const CoordRec char93_stroke2[] = { + { 4.44, 119.048 }, + { 37.7733, 119.048 }, +}; + +static const CoordRec char93_stroke3[] = { + { 4.44, -33.3333 }, + { 37.7733, -33.3333 }, +}; + +static const StrokeRec char93[] = { + { 2, char93_stroke0 }, + { 2, char93_stroke1 }, + { 2, char93_stroke2 }, + { 2, char93_stroke3 }, +}; + +/* char: 94 '^' */ + +static const CoordRec char94_stroke0[] = { + { 44.0752, 109.524 }, + { 5.98, 42.8571 }, +}; + +static const CoordRec char94_stroke1[] = { + { 44.0752, 109.524 }, + { 82.1705, 42.8571 }, +}; + +static const StrokeRec char94[] = { + { 2, char94_stroke0 }, + { 2, char94_stroke1 }, +}; + +/* char: 95 '_' */ + +static const CoordRec char95_stroke0[] = { + { -1.1, -33.3333 }, + { 103.662, -33.3333 }, + { 103.662, -28.5714 }, + { -1.1, -28.5714 }, + { -1.1, -33.3333 }, +}; + +static const StrokeRec char95[] = { + { 5, char95_stroke0 }, +}; + +/* char: 96 '`' */ + +static const CoordRec char96_stroke0[] = { + { 33.0219, 100 }, + { 56.8314, 71.4286 }, +}; + +static const CoordRec char96_stroke1[] = { + { 33.0219, 100 }, + { 28.26, 95.2381 }, + { 56.8314, 71.4286 }, +}; + +static const StrokeRec char96[] = { + { 2, char96_stroke0 }, + { 3, char96_stroke1 }, +}; + +/* char: 97 'a' */ + +static const CoordRec char97_stroke0[] = { + { 63.8229, 66.6667 }, + { 63.8229, 0 }, +}; + +static const CoordRec char97_stroke1[] = { + { 63.8229, 52.381 }, + { 54.299, 61.9048 }, + { 44.7752, 66.6667 }, + { 30.4895, 66.6667 }, + { 20.9657, 61.9048 }, + { 11.4419, 52.381 }, + { 6.68, 38.0952 }, + { 6.68, 28.5714 }, + { 11.4419, 14.2857 }, + { 20.9657, 4.7619 }, + { 30.4895, 0 }, + { 44.7752, 0 }, + { 54.299, 4.7619 }, + { 63.8229, 14.2857 }, +}; + +static const StrokeRec char97[] = { + { 2, char97_stroke0 }, + { 14, char97_stroke1 }, +}; + +/* char: 98 'b' */ + +static const CoordRec char98_stroke0[] = { + { 8.76, 100 }, + { 8.76, 0 }, +}; + +static const CoordRec char98_stroke1[] = { + { 8.76, 52.381 }, + { 18.2838, 61.9048 }, + { 27.8076, 66.6667 }, + { 42.0933, 66.6667 }, + { 51.6171, 61.9048 }, + { 61.141, 52.381 }, + { 65.9029, 38.0952 }, + { 65.9029, 28.5714 }, + { 61.141, 14.2857 }, + { 51.6171, 4.7619 }, + { 42.0933, 0 }, + { 27.8076, 0 }, + { 18.2838, 4.7619 }, + { 8.76, 14.2857 }, +}; + +static const StrokeRec char98[] = { + { 2, char98_stroke0 }, + { 14, char98_stroke1 }, +}; + +/* char: 99 'c' */ + +static const CoordRec char99_stroke0[] = { + { 62.6629, 52.381 }, + { 53.139, 61.9048 }, + { 43.6152, 66.6667 }, + { 29.3295, 66.6667 }, + { 19.8057, 61.9048 }, + { 10.2819, 52.381 }, + { 5.52, 38.0952 }, + { 5.52, 28.5714 }, + { 10.2819, 14.2857 }, + { 19.8057, 4.7619 }, + { 29.3295, 0 }, + { 43.6152, 0 }, + { 53.139, 4.7619 }, + { 62.6629, 14.2857 }, +}; + +static const StrokeRec char99[] = { + { 14, char99_stroke0 }, +}; + +/* char: 100 'd' */ + +static const CoordRec char100_stroke0[] = { + { 61.7829, 100 }, + { 61.7829, 0 }, +}; + +static const CoordRec char100_stroke1[] = { + { 61.7829, 52.381 }, + { 52.259, 61.9048 }, + { 42.7352, 66.6667 }, + { 28.4495, 66.6667 }, + { 18.9257, 61.9048 }, + { 9.4019, 52.381 }, + { 4.64, 38.0952 }, + { 4.64, 28.5714 }, + { 9.4019, 14.2857 }, + { 18.9257, 4.7619 }, + { 28.4495, 0 }, + { 42.7352, 0 }, + { 52.259, 4.7619 }, + { 61.7829, 14.2857 }, +}; + +static const StrokeRec char100[] = { + { 2, char100_stroke0 }, + { 14, char100_stroke1 }, +}; + +/* char: 101 'e' */ + +static const CoordRec char101_stroke0[] = { + { 5.72, 38.0952 }, + { 62.8629, 38.0952 }, + { 62.8629, 47.619 }, + { 58.101, 57.1429 }, + { 53.339, 61.9048 }, + { 43.8152, 66.6667 }, + { 29.5295, 66.6667 }, + { 20.0057, 61.9048 }, + { 10.4819, 52.381 }, + { 5.72, 38.0952 }, + { 5.72, 28.5714 }, + { 10.4819, 14.2857 }, + { 20.0057, 4.7619 }, + { 29.5295, 0 }, + { 43.8152, 0 }, + { 53.339, 4.7619 }, + { 62.8629, 14.2857 }, +}; + +static const StrokeRec char101[] = { + { 17, char101_stroke0 }, +}; + +/* char: 102 'f' */ + +static const CoordRec char102_stroke0[] = { + { 38.7752, 100 }, + { 29.2514, 100 }, + { 19.7276, 95.2381 }, + { 14.9657, 80.9524 }, + { 14.9657, 0 }, +}; + +static const CoordRec char102_stroke1[] = { + { 0.68, 66.6667 }, + { 34.0133, 66.6667 }, +}; + +static const StrokeRec char102[] = { + { 5, char102_stroke0 }, + { 2, char102_stroke1 }, +}; + +/* char: 103 'g' */ + +static const CoordRec char103_stroke0[] = { + { 62.5029, 66.6667 }, + { 62.5029, -9.5238 }, + { 57.741, -23.8095 }, + { 52.979, -28.5714 }, + { 43.4552, -33.3333 }, + { 29.1695, -33.3333 }, + { 19.6457, -28.5714 }, +}; + +static const CoordRec char103_stroke1[] = { + { 62.5029, 52.381 }, + { 52.979, 61.9048 }, + { 43.4552, 66.6667 }, + { 29.1695, 66.6667 }, + { 19.6457, 61.9048 }, + { 10.1219, 52.381 }, + { 5.36, 38.0952 }, + { 5.36, 28.5714 }, + { 10.1219, 14.2857 }, + { 19.6457, 4.7619 }, + { 29.1695, 0 }, + { 43.4552, 0 }, + { 52.979, 4.7619 }, + { 62.5029, 14.2857 }, +}; + +static const StrokeRec char103[] = { + { 7, char103_stroke0 }, + { 14, char103_stroke1 }, +}; + +/* char: 104 'h' */ + +static const CoordRec char104_stroke0[] = { + { 9.6, 100 }, + { 9.6, 0 }, +}; + +static const CoordRec char104_stroke1[] = { + { 9.6, 47.619 }, + { 23.8857, 61.9048 }, + { 33.4095, 66.6667 }, + { 47.6952, 66.6667 }, + { 57.219, 61.9048 }, + { 61.981, 47.619 }, + { 61.981, 0 }, +}; + +static const StrokeRec char104[] = { + { 2, char104_stroke0 }, + { 7, char104_stroke1 }, +}; + +/* char: 105 'i' */ + +static const CoordRec char105_stroke0[] = { + { 10.02, 100 }, + { 14.7819, 95.2381 }, + { 19.5438, 100 }, + { 14.7819, 104.762 }, + { 10.02, 100 }, +}; + +static const CoordRec char105_stroke1[] = { + { 14.7819, 66.6667 }, + { 14.7819, 0 }, +}; + +static const StrokeRec char105[] = { + { 5, char105_stroke0 }, + { 2, char105_stroke1 }, +}; + +/* char: 106 'j' */ + +static const CoordRec char106_stroke0[] = { + { 17.3876, 100 }, + { 22.1495, 95.2381 }, + { 26.9114, 100 }, + { 22.1495, 104.762 }, + { 17.3876, 100 }, +}; + +static const CoordRec char106_stroke1[] = { + { 22.1495, 66.6667 }, + { 22.1495, -14.2857 }, + { 17.3876, -28.5714 }, + { 7.8638, -33.3333 }, + { -1.66, -33.3333 }, +}; + +static const StrokeRec char106[] = { + { 5, char106_stroke0 }, + { 5, char106_stroke1 }, +}; + +/* char: 107 'k' */ + +static const CoordRec char107_stroke0[] = { + { 9.6, 100 }, + { 9.6, 0 }, +}; + +static const CoordRec char107_stroke1[] = { + { 57.219, 66.6667 }, + { 9.6, 19.0476 }, +}; + +static const CoordRec char107_stroke2[] = { + { 28.6476, 38.0952 }, + { 61.981, 0 }, +}; + +static const StrokeRec char107[] = { + { 2, char107_stroke0 }, + { 2, char107_stroke1 }, + { 2, char107_stroke2 }, +}; + +/* char: 108 'l' */ + +static const CoordRec char108_stroke0[] = { + { 10.02, 100 }, + { 10.02, 0 }, +}; + +static const StrokeRec char108[] = { + { 2, char108_stroke0 }, +}; + +/* char: 109 'm' */ + +static const CoordRec char109_stroke0[] = { + { 9.6, 66.6667 }, + { 9.6, 0 }, +}; + +static const CoordRec char109_stroke1[] = { + { 9.6, 47.619 }, + { 23.8857, 61.9048 }, + { 33.4095, 66.6667 }, + { 47.6952, 66.6667 }, + { 57.219, 61.9048 }, + { 61.981, 47.619 }, + { 61.981, 0 }, +}; + +static const CoordRec char109_stroke2[] = { + { 61.981, 47.619 }, + { 76.2667, 61.9048 }, + { 85.7905, 66.6667 }, + { 100.076, 66.6667 }, + { 109.6, 61.9048 }, + { 114.362, 47.619 }, + { 114.362, 0 }, +}; + +static const StrokeRec char109[] = { + { 2, char109_stroke0 }, + { 7, char109_stroke1 }, + { 7, char109_stroke2 }, +}; + +/* char: 110 'n' */ + +static const CoordRec char110_stroke0[] = { + { 9.18, 66.6667 }, + { 9.18, 0 }, +}; + +static const CoordRec char110_stroke1[] = { + { 9.18, 47.619 }, + { 23.4657, 61.9048 }, + { 32.9895, 66.6667 }, + { 47.2752, 66.6667 }, + { 56.799, 61.9048 }, + { 61.561, 47.619 }, + { 61.561, 0 }, +}; + +static const StrokeRec char110[] = { + { 2, char110_stroke0 }, + { 7, char110_stroke1 }, +}; + +/* char: 111 'o' */ + +static const CoordRec char111_stroke0[] = { + { 28.7895, 66.6667 }, + { 19.2657, 61.9048 }, + { 9.7419, 52.381 }, + { 4.98, 38.0952 }, + { 4.98, 28.5714 }, + { 9.7419, 14.2857 }, + { 19.2657, 4.7619 }, + { 28.7895, 0 }, + { 43.0752, 0 }, + { 52.599, 4.7619 }, + { 62.1229, 14.2857 }, + { 66.8848, 28.5714 }, + { 66.8848, 38.0952 }, + { 62.1229, 52.381 }, + { 52.599, 61.9048 }, + { 43.0752, 66.6667 }, + { 28.7895, 66.6667 }, +}; + +static const StrokeRec char111[] = { + { 17, char111_stroke0 }, +}; + +/* char: 112 'p' */ + +static const CoordRec char112_stroke0[] = { + { 9.46, 66.6667 }, + { 9.46, -33.3333 }, +}; + +static const CoordRec char112_stroke1[] = { + { 9.46, 52.381 }, + { 18.9838, 61.9048 }, + { 28.5076, 66.6667 }, + { 42.7933, 66.6667 }, + { 52.3171, 61.9048 }, + { 61.841, 52.381 }, + { 66.6029, 38.0952 }, + { 66.6029, 28.5714 }, + { 61.841, 14.2857 }, + { 52.3171, 4.7619 }, + { 42.7933, 0 }, + { 28.5076, 0 }, + { 18.9838, 4.7619 }, + { 9.46, 14.2857 }, +}; + +static const StrokeRec char112[] = { + { 2, char112_stroke0 }, + { 14, char112_stroke1 }, +}; + +/* char: 113 'q' */ + +static const CoordRec char113_stroke0[] = { + { 61.9829, 66.6667 }, + { 61.9829, -33.3333 }, +}; + +static const CoordRec char113_stroke1[] = { + { 61.9829, 52.381 }, + { 52.459, 61.9048 }, + { 42.9352, 66.6667 }, + { 28.6495, 66.6667 }, + { 19.1257, 61.9048 }, + { 9.6019, 52.381 }, + { 4.84, 38.0952 }, + { 4.84, 28.5714 }, + { 9.6019, 14.2857 }, + { 19.1257, 4.7619 }, + { 28.6495, 0 }, + { 42.9352, 0 }, + { 52.459, 4.7619 }, + { 61.9829, 14.2857 }, +}; + +static const StrokeRec char113[] = { + { 2, char113_stroke0 }, + { 14, char113_stroke1 }, +}; + +/* char: 114 'r' */ + +static const CoordRec char114_stroke0[] = { + { 9.46, 66.6667 }, + { 9.46, 0 }, +}; + +static const CoordRec char114_stroke1[] = { + { 9.46, 38.0952 }, + { 14.2219, 52.381 }, + { 23.7457, 61.9048 }, + { 33.2695, 66.6667 }, + { 47.5552, 66.6667 }, +}; + +static const StrokeRec char114[] = { + { 2, char114_stroke0 }, + { 5, char114_stroke1 }, +}; + +/* char: 115 's' */ + +static const CoordRec char115_stroke0[] = { + { 57.081, 52.381 }, + { 52.319, 61.9048 }, + { 38.0333, 66.6667 }, + { 23.7476, 66.6667 }, + { 9.4619, 61.9048 }, + { 4.7, 52.381 }, + { 9.4619, 42.8571 }, + { 18.9857, 38.0952 }, + { 42.7952, 33.3333 }, + { 52.319, 28.5714 }, + { 57.081, 19.0476 }, + { 57.081, 14.2857 }, + { 52.319, 4.7619 }, + { 38.0333, 0 }, + { 23.7476, 0 }, + { 9.4619, 4.7619 }, + { 4.7, 14.2857 }, +}; + +static const StrokeRec char115[] = { + { 17, char115_stroke0 }, +}; + +/* char: 116 't' */ + +static const CoordRec char116_stroke0[] = { + { 14.8257, 100 }, + { 14.8257, 19.0476 }, + { 19.5876, 4.7619 }, + { 29.1114, 0 }, + { 38.6352, 0 }, +}; + +static const CoordRec char116_stroke1[] = { + { 0.54, 66.6667 }, + { 33.8733, 66.6667 }, +}; + +static const StrokeRec char116[] = { + { 5, char116_stroke0 }, + { 2, char116_stroke1 }, +}; + +/* char: 117 'u' */ + +static const CoordRec char117_stroke0[] = { + { 9.46, 66.6667 }, + { 9.46, 19.0476 }, + { 14.2219, 4.7619 }, + { 23.7457, 0 }, + { 38.0314, 0 }, + { 47.5552, 4.7619 }, + { 61.841, 19.0476 }, +}; + +static const CoordRec char117_stroke1[] = { + { 61.841, 66.6667 }, + { 61.841, 0 }, +}; + +static const StrokeRec char117[] = { + { 7, char117_stroke0 }, + { 2, char117_stroke1 }, +}; + +/* char: 118 'v' */ + +static const CoordRec char118_stroke0[] = { + { 1.8, 66.6667 }, + { 30.3714, 0 }, +}; + +static const CoordRec char118_stroke1[] = { + { 58.9429, 66.6667 }, + { 30.3714, 0 }, +}; + +static const StrokeRec char118[] = { + { 2, char118_stroke0 }, + { 2, char118_stroke1 }, +}; + +/* char: 119 'w' */ + +static const CoordRec char119_stroke0[] = { + { 2.5, 66.6667 }, + { 21.5476, 0 }, +}; + +static const CoordRec char119_stroke1[] = { + { 40.5952, 66.6667 }, + { 21.5476, 0 }, +}; + +static const CoordRec char119_stroke2[] = { + { 40.5952, 66.6667 }, + { 59.6429, 0 }, +}; + +static const CoordRec char119_stroke3[] = { + { 78.6905, 66.6667 }, + { 59.6429, 0 }, +}; + +static const StrokeRec char119[] = { + { 2, char119_stroke0 }, + { 2, char119_stroke1 }, + { 2, char119_stroke2 }, + { 2, char119_stroke3 }, +}; + +/* char: 120 'x' */ + +static const CoordRec char120_stroke0[] = { + { 1.66, 66.6667 }, + { 54.041, 0 }, +}; + +static const CoordRec char120_stroke1[] = { + { 54.041, 66.6667 }, + { 1.66, 0 }, +}; + +static const StrokeRec char120[] = { + { 2, char120_stroke0 }, + { 2, char120_stroke1 }, +}; + +/* char: 121 'y' */ + +static const CoordRec char121_stroke0[] = { + { 6.5619, 66.6667 }, + { 35.1333, 0 }, +}; + +static const CoordRec char121_stroke1[] = { + { 63.7048, 66.6667 }, + { 35.1333, 0 }, + { 25.6095, -19.0476 }, + { 16.0857, -28.5714 }, + { 6.5619, -33.3333 }, + { 1.8, -33.3333 }, +}; + +static const StrokeRec char121[] = { + { 2, char121_stroke0 }, + { 6, char121_stroke1 }, +}; + +/* char: 122 'z' */ + +static const CoordRec char122_stroke0[] = { + { 56.821, 66.6667 }, + { 4.44, 0 }, +}; + +static const CoordRec char122_stroke1[] = { + { 4.44, 66.6667 }, + { 56.821, 66.6667 }, +}; + +static const CoordRec char122_stroke2[] = { + { 4.44, 0 }, + { 56.821, 0 }, +}; + +static const StrokeRec char122[] = { + { 2, char122_stroke0 }, + { 2, char122_stroke1 }, + { 2, char122_stroke2 }, +}; + +/* char: 123 '{' */ + +static const CoordRec char123_stroke0[] = { + { 31.1895, 119.048 }, + { 21.6657, 114.286 }, + { 16.9038, 109.524 }, + { 12.1419, 100 }, + { 12.1419, 90.4762 }, + { 16.9038, 80.9524 }, + { 21.6657, 76.1905 }, + { 26.4276, 66.6667 }, + { 26.4276, 57.1429 }, + { 16.9038, 47.619 }, +}; + +static const CoordRec char123_stroke1[] = { + { 21.6657, 114.286 }, + { 16.9038, 104.762 }, + { 16.9038, 95.2381 }, + { 21.6657, 85.7143 }, + { 26.4276, 80.9524 }, + { 31.1895, 71.4286 }, + { 31.1895, 61.9048 }, + { 26.4276, 52.381 }, + { 7.38, 42.8571 }, + { 26.4276, 33.3333 }, + { 31.1895, 23.8095 }, + { 31.1895, 14.2857 }, + { 26.4276, 4.7619 }, + { 21.6657, 0 }, + { 16.9038, -9.5238 }, + { 16.9038, -19.0476 }, + { 21.6657, -28.5714 }, +}; + +static const CoordRec char123_stroke2[] = { + { 16.9038, 38.0952 }, + { 26.4276, 28.5714 }, + { 26.4276, 19.0476 }, + { 21.6657, 9.5238 }, + { 16.9038, 4.7619 }, + { 12.1419, -4.7619 }, + { 12.1419, -14.2857 }, + { 16.9038, -23.8095 }, + { 21.6657, -28.5714 }, + { 31.1895, -33.3333 }, +}; + +static const StrokeRec char123[] = { + { 10, char123_stroke0 }, + { 17, char123_stroke1 }, + { 10, char123_stroke2 }, +}; + +/* char: 124 '|' */ + +static const CoordRec char124_stroke0[] = { + { 11.54, 119.048 }, + { 11.54, -33.3333 }, +}; + +static const StrokeRec char124[] = { + { 2, char124_stroke0 }, +}; + +/* char: 125 '}' */ + +static const CoordRec char125_stroke0[] = { + { 9.18, 119.048 }, + { 18.7038, 114.286 }, + { 23.4657, 109.524 }, + { 28.2276, 100 }, + { 28.2276, 90.4762 }, + { 23.4657, 80.9524 }, + { 18.7038, 76.1905 }, + { 13.9419, 66.6667 }, + { 13.9419, 57.1429 }, + { 23.4657, 47.619 }, +}; + +static const CoordRec char125_stroke1[] = { + { 18.7038, 114.286 }, + { 23.4657, 104.762 }, + { 23.4657, 95.2381 }, + { 18.7038, 85.7143 }, + { 13.9419, 80.9524 }, + { 9.18, 71.4286 }, + { 9.18, 61.9048 }, + { 13.9419, 52.381 }, + { 32.9895, 42.8571 }, + { 13.9419, 33.3333 }, + { 9.18, 23.8095 }, + { 9.18, 14.2857 }, + { 13.9419, 4.7619 }, + { 18.7038, 0 }, + { 23.4657, -9.5238 }, + { 23.4657, -19.0476 }, + { 18.7038, -28.5714 }, +}; + +static const CoordRec char125_stroke2[] = { + { 23.4657, 38.0952 }, + { 13.9419, 28.5714 }, + { 13.9419, 19.0476 }, + { 18.7038, 9.5238 }, + { 23.4657, 4.7619 }, + { 28.2276, -4.7619 }, + { 28.2276, -14.2857 }, + { 23.4657, -23.8095 }, + { 18.7038, -28.5714 }, + { 9.18, -33.3333 }, +}; + +static const StrokeRec char125[] = { + { 10, char125_stroke0 }, + { 17, char125_stroke1 }, + { 10, char125_stroke2 }, +}; + +/* char: 126 '~' */ + +static const CoordRec char126_stroke0[] = { + { 2.92, 28.5714 }, + { 2.92, 38.0952 }, + { 7.6819, 52.381 }, + { 17.2057, 57.1429 }, + { 26.7295, 57.1429 }, + { 36.2533, 52.381 }, + { 55.301, 38.0952 }, + { 64.8248, 33.3333 }, + { 74.3486, 33.3333 }, + { 83.8724, 38.0952 }, + { 88.6343, 47.619 }, +}; + +static const CoordRec char126_stroke1[] = { + { 2.92, 38.0952 }, + { 7.6819, 47.619 }, + { 17.2057, 52.381 }, + { 26.7295, 52.381 }, + { 36.2533, 47.619 }, + { 55.301, 33.3333 }, + { 64.8248, 28.5714 }, + { 74.3486, 28.5714 }, + { 83.8724, 33.3333 }, + { 88.6343, 47.619 }, + { 88.6343, 57.1429 }, +}; + +static const StrokeRec char126[] = { + { 11, char126_stroke0 }, + { 11, char126_stroke1 }, +}; + +/* char: 127 */ + +static const CoordRec char127_stroke0[] = { + { 52.381, 100 }, + { 14.2857, -33.3333 }, +}; + +static const CoordRec char127_stroke1[] = { + { 28.5714, 66.6667 }, + { 14.2857, 61.9048 }, + { 4.7619, 52.381 }, + { 0, 38.0952 }, + { 0, 23.8095 }, + { 4.7619, 14.2857 }, + { 14.2857, 4.7619 }, + { 28.5714, 0 }, + { 38.0952, 0 }, + { 52.381, 4.7619 }, + { 61.9048, 14.2857 }, + { 66.6667, 28.5714 }, + { 66.6667, 42.8571 }, + { 61.9048, 52.381 }, + { 52.381, 61.9048 }, + { 38.0952, 66.6667 }, + { 28.5714, 66.6667 }, +}; + +static const StrokeRec char127[] = { + { 2, char127_stroke0 }, + { 17, char127_stroke1 }, +}; + +static const StrokeCharRec chars[] = { + { 0, /* char0 */ 0, 0, 0 }, + { 0, /* char1 */ 0, 0, 0 }, + { 0, /* char2 */ 0, 0, 0 }, + { 0, /* char3 */ 0, 0, 0 }, + { 0, /* char4 */ 0, 0, 0 }, + { 0, /* char5 */ 0, 0, 0 }, + { 0, /* char6 */ 0, 0, 0 }, + { 0, /* char7 */ 0, 0, 0 }, + { 0, /* char8 */ 0, 0, 0 }, + { 0, /* char9 */ 0, 0, 0 }, + { 0, /* char10 */ 0, 0, 0 }, + { 0, /* char11 */ 0, 0, 0 }, + { 0, /* char12 */ 0, 0, 0 }, + { 0, /* char13 */ 0, 0, 0 }, + { 0, /* char14 */ 0, 0, 0 }, + { 0, /* char15 */ 0, 0, 0 }, + { 0, /* char16 */ 0, 0, 0 }, + { 0, /* char17 */ 0, 0, 0 }, + { 0, /* char18 */ 0, 0, 0 }, + { 0, /* char19 */ 0, 0, 0 }, + { 0, /* char20 */ 0, 0, 0 }, + { 0, /* char21 */ 0, 0, 0 }, + { 0, /* char22 */ 0, 0, 0 }, + { 0, /* char23 */ 0, 0, 0 }, + { 0, /* char24 */ 0, 0, 0 }, + { 0, /* char25 */ 0, 0, 0 }, + { 0, /* char26 */ 0, 0, 0 }, + { 0, /* char27 */ 0, 0, 0 }, + { 0, /* char28 */ 0, 0, 0 }, + { 0, /* char29 */ 0, 0, 0 }, + { 0, /* char30 */ 0, 0, 0 }, + { 0, /* char31 */ 0, 0, 0 }, + { 0, /* char32 */ 0, 52.381, 104.762 }, + { 2, char33, 13.3819, 26.6238 }, + { 2, char34, 23.0676, 51.4352 }, + { 4, char35, 36.5333, 79.4886 }, + { 3, char36, 38.1533, 76.2067 }, + { 3, char37, 49.2171, 96.5743 }, + { 1, char38, 53.599, 101.758 }, + { 1, char39, 4.44, 13.62 }, + { 1, char40, 21.8657, 47.1733 }, + { 1, char41, 24.3276, 47.5333 }, + { 3, char42, 30.7695, 59.439 }, + { 2, char43, 48.8371, 97.2543 }, + { 1, char44, 13.5219, 26.0638 }, + { 1, char45, 50.2371, 100.754 }, + { 1, char46, 13.1019, 26.4838 }, + { 1, char47, 40.5733, 82.1067 }, + { 1, char48, 38.3133, 77.0667 }, + { 1, char49, 30.8676, 66.5295 }, + { 1, char50, 38.7533, 77.6467 }, + { 1, char51, 38.3333, 77.0467 }, + { 2, char52, 37.2133, 80.1686 }, + { 1, char53, 38.1933, 77.6867 }, + { 1, char54, 34.1514, 73.8048 }, + { 2, char55, 38.8933, 77.2267 }, + { 1, char56, 38.9333, 77.6667 }, + { 1, char57, 39.9333, 74.0648 }, + { 2, char58, 14.0819, 26.2238 }, + { 2, char59, 12.9619, 26.3038 }, + { 1, char60, 41.1552, 81.6105 }, + { 2, char61, 48.5571, 97.2543 }, + { 1, char62, 40.8752, 81.6105 }, + { 2, char63, 36.9914, 73.9029 }, + { 2, char64, 34.9314, 74.3648 }, + { 3, char65, 40.5952, 80.4905 }, + { 3, char66, 44.7533, 83.6267 }, + { 1, char67, 39.9933, 84.4886 }, + { 2, char68, 45.2933, 85.2867 }, + { 4, char69, 39.9914, 78.1848 }, + { 3, char70, 39.9914, 78.7448 }, + { 2, char71, 40.3933, 89.7686 }, + { 3, char72, 44.7533, 89.0867 }, + { 1, char73, 10.86, 21.3 }, + { 1, char74, 31.0714, 59.999 }, + { 3, char75, 44.6133, 79.3267 }, + { 2, char76, 40.2514, 71.3229 }, + { 4, char77, 48.9552, 97.2105 }, + { 3, char78, 44.4733, 88.8067 }, + { 1, char79, 44.3352, 88.8305 }, + { 2, char80, 45.4333, 85.6667 }, + { 2, char81, 43.3952, 88.0905 }, + { 3, char82, 45.0133, 82.3667 }, + { 1, char83, 41.3333, 80.8267 }, + { 2, char84, 35.6933, 71.9467 }, + { 1, char85, 44.8733, 89.4867 }, + { 2, char86, 40.4552, 81.6105 }, + { 4, char87, 49.839, 100.518 }, + { 2, char88, 35.8333, 72.3667 }, + { 2, char89, 39.6152, 79.6505 }, + { 3, char90, 35.8333, 73.7467 }, + { 4, char91, 22.0657, 46.1133 }, + { 1, char92, 39.1733, 78.2067 }, + { 4, char93, 23.4876, 46.3933 }, + { 2, char94, 44.0752, 90.2305 }, + { 1, char95, 51.281, 104.062 }, + { 2, char96, 42.5457, 83.5714 }, + { 2, char97, 35.2514, 66.6029 }, + { 2, char98, 37.3314, 70.4629 }, + { 1, char99, 34.0914, 68.9229 }, + { 2, char100, 33.2114, 70.2629 }, + { 1, char101, 34.2914, 68.5229 }, + { 2, char102, 14.9657, 38.6552 }, + { 2, char103, 33.9314, 70.9829 }, + { 2, char104, 33.4095, 71.021 }, + { 2, char105, 14.7819, 28.8638 }, + { 2, char106, 17.3876, 36.2314 }, + { 3, char107, 33.4095, 62.521 }, + { 1, char108, 10.02, 19.34 }, + { 3, char109, 61.981, 123.962 }, + { 2, char110, 32.9895, 70.881 }, + { 1, char111, 33.5514, 71.7448 }, + { 2, char112, 38.0314, 70.8029 }, + { 2, char113, 33.4114, 70.7429 }, + { 2, char114, 23.7457, 49.4952 }, + { 1, char115, 28.5095, 62.321 }, + { 2, char116, 14.8257, 39.3152 }, + { 2, char117, 33.2695, 71.161 }, + { 2, char118, 30.3714, 60.6029 }, + { 4, char119, 40.5952, 80.4905 }, + { 2, char120, 25.4695, 56.401 }, + { 2, char121, 35.1333, 66.0648 }, + { 3, char122, 28.2495, 61.821 }, + { 3, char123, 21.6657, 41.6295 }, + { 1, char124, 11.54, 23.78 }, + { 3, char125, 18.7038, 41.4695 }, + { 2, char126, 45.7771, 91.2743 }, + { 2, char127, 33.3333, 66.6667 }, +}; + +StrokeFontRec glutStrokeRoman = { "Roman", 128, chars, 119.048, -33.3333 }; + diff --git a/src/glut/os2/glut_shapes.cpp b/src/glut/os2/glut_shapes.cpp index 76b0809511..111e2d48e9 100644 --- a/src/glut/os2/glut_shapes.cpp +++ b/src/glut/os2/glut_shapes.cpp @@ -1,595 +1,595 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ - -/** -(c) Copyright 1993, Silicon Graphics, Inc. - -ALL RIGHTS RESERVED - -Permission to use, copy, modify, and distribute this software -for any purpose and without fee is hereby granted, provided -that the above copyright notice appear in all copies and that -both the copyright notice and this permission notice appear in -supporting documentation, and that the name of Silicon -Graphics, Inc. not be used in advertising or publicity -pertaining to distribution of the software without specific, -written prior permission. - -THE MATERIAL EMBODIED ON THIS SOFTWARE IS PROVIDED TO YOU -"AS-IS" AND WITHOUT WARRANTY OF ANY KIND, EXPRESS, IMPLIED OR -OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY WARRANTY OF -MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. IN NO -EVENT SHALL SILICON GRAPHICS, INC. BE LIABLE TO YOU OR ANYONE -ELSE FOR ANY DIRECT, SPECIAL, INCIDENTAL, INDIRECT OR -CONSEQUENTIAL DAMAGES OF ANY KIND, OR ANY DAMAGES WHATSOEVER, -INCLUDING WITHOUT LIMITATION, LOSS OF PROFIT, LOSS OF USE, -SAVINGS OR REVENUE, OR THE CLAIMS OF THIRD PARTIES, WHETHER OR -NOT SILICON GRAPHICS, INC. HAS BEEN ADVISED OF THE POSSIBILITY -OF SUCH LOSS, HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, -ARISING OUT OF OR IN CONNECTION WITH THE POSSESSION, USE OR -PERFORMANCE OF THIS SOFTWARE. - -US Government Users Restricted Rights - -Use, duplication, or disclosure by the Government is subject to -restrictions set forth in FAR 52.227.19(c)(2) or subparagraph -(c)(1)(ii) of the Rights in Technical Data and Computer -Software clause at DFARS 252.227-7013 and/or in similar or -successor clauses in the FAR or the DOD or NASA FAR -Supplement. Unpublished-- rights reserved under the copyright -laws of the United States. Contractor/manufacturer is Silicon -Graphics, Inc., 2011 N. Shoreline Blvd., Mountain View, CA -94039-7311. - -OpenGL(TM) is a trademark of Silicon Graphics, Inc. -*/ - -#include -#include "glutint.h" - -/* Some files do not define M_PI... */ -#ifndef M_PI -#define M_PI 3.14159265358979323846 -#endif - -static GLUquadricObj *quadObj; - -#define QUAD_OBJ_INIT() { if(!quadObj) initQuadObj(); } - -static void -initQuadObj(void) -{ - quadObj = gluNewQuadric(); - if (!quadObj) - __glutFatalError("out of memory."); -} - -/* CENTRY */ -void GLUTAPIENTRY -glutWireSphere(GLdouble radius, GLint slices, GLint stacks) -{ - QUAD_OBJ_INIT(); - gluQuadricDrawStyle(quadObj, GLU_LINE); - gluQuadricNormals(quadObj, GLU_SMOOTH); - /* If we ever changed/used the texture or orientation state - of quadObj, we'd need to change it to the defaults here - with gluQuadricTexture and/or gluQuadricOrientation. */ - gluSphere(quadObj, radius, slices, stacks); -} - -void GLUTAPIENTRY -glutSolidSphere(GLdouble radius, GLint slices, GLint stacks) -{ - QUAD_OBJ_INIT(); - gluQuadricDrawStyle(quadObj, GLU_FILL); - gluQuadricNormals(quadObj, GLU_SMOOTH); - /* If we ever changed/used the texture or orientation state - of quadObj, we'd need to change it to the defaults here - with gluQuadricTexture and/or gluQuadricOrientation. */ - gluSphere(quadObj, radius, slices, stacks); -} - -void GLUTAPIENTRY -glutWireCone(GLdouble base, GLdouble height, - GLint slices, GLint stacks) -{ - QUAD_OBJ_INIT(); - gluQuadricDrawStyle(quadObj, GLU_LINE); - gluQuadricNormals(quadObj, GLU_SMOOTH); - /* If we ever changed/used the texture or orientation state - of quadObj, we'd need to change it to the defaults here - with gluQuadricTexture and/or gluQuadricOrientation. */ - gluCylinder(quadObj, base, 0.0, height, slices, stacks); -} - -void GLUTAPIENTRY -glutSolidCone(GLdouble base, GLdouble height, - GLint slices, GLint stacks) -{ - QUAD_OBJ_INIT(); - gluQuadricDrawStyle(quadObj, GLU_FILL); - gluQuadricNormals(quadObj, GLU_SMOOTH); - /* If we ever changed/used the texture or orientation state - of quadObj, we'd need to change it to the defaults here - with gluQuadricTexture and/or gluQuadricOrientation. */ - gluCylinder(quadObj, base, 0.0, height, slices, stacks); -} - -/* ENDCENTRY */ - -static void -drawBox(GLfloat size, GLenum type) -{ - static GLfloat n[6][3] = - { - {-1.0, 0.0, 0.0}, - {0.0, 1.0, 0.0}, - {1.0, 0.0, 0.0}, - {0.0, -1.0, 0.0}, - {0.0, 0.0, 1.0}, - {0.0, 0.0, -1.0} - }; - static GLint faces[6][4] = - { - {0, 1, 2, 3}, - {3, 2, 6, 7}, - {7, 6, 5, 4}, - {4, 5, 1, 0}, - {5, 6, 2, 1}, - {7, 4, 0, 3} - }; - GLfloat v[8][3]; - GLint i; - - v[0][0] = v[1][0] = v[2][0] = v[3][0] = -size / 2; - v[4][0] = v[5][0] = v[6][0] = v[7][0] = size / 2; - v[0][1] = v[1][1] = v[4][1] = v[5][1] = -size / 2; - v[2][1] = v[3][1] = v[6][1] = v[7][1] = size / 2; - v[0][2] = v[3][2] = v[4][2] = v[7][2] = -size / 2; - v[1][2] = v[2][2] = v[5][2] = v[6][2] = size / 2; - - for (i = 5; i >= 0; i--) { - glBegin(type); - glNormal3fv(&n[i][0]); - glVertex3fv(&v[faces[i][0]][0]); - glVertex3fv(&v[faces[i][1]][0]); - glVertex3fv(&v[faces[i][2]][0]); - glVertex3fv(&v[faces[i][3]][0]); - glEnd(); - } -} - -/* CENTRY */ -void GLUTAPIENTRY -glutWireCube(GLdouble size) -{ - drawBox(size, GL_LINE_LOOP); -} - -void GLUTAPIENTRY -glutSolidCube(GLdouble size) -{ - drawBox(size, GL_QUADS); -} - -/* ENDCENTRY */ - -static void -doughnut(GLfloat r, GLfloat R, GLint nsides, GLint rings) -{ - int i, j; - GLfloat theta, phi, theta1; - GLfloat cosTheta, sinTheta; - GLfloat cosTheta1, sinTheta1; - GLfloat ringDelta, sideDelta; - - ringDelta = 2.0 * M_PI / rings; - sideDelta = 2.0 * M_PI / nsides; - - theta = 0.0; - cosTheta = 1.0; - sinTheta = 0.0; - for (i = rings - 1; i >= 0; i--) { - theta1 = theta + ringDelta; - cosTheta1 = cos(theta1); - sinTheta1 = sin(theta1); - glBegin(GL_QUAD_STRIP); - phi = 0.0; - for (j = nsides; j >= 0; j--) { - GLfloat cosPhi, sinPhi, dist; - - phi += sideDelta; - cosPhi = cos(phi); - sinPhi = sin(phi); - dist = R + r * cosPhi; - - glNormal3f(cosTheta1 * cosPhi, -sinTheta1 * cosPhi, sinPhi); - glVertex3f(cosTheta1 * dist, -sinTheta1 * dist, r * sinPhi); - glNormal3f(cosTheta * cosPhi, -sinTheta * cosPhi, sinPhi); - glVertex3f(cosTheta * dist, -sinTheta * dist, r * sinPhi); - } - glEnd(); - theta = theta1; - cosTheta = cosTheta1; - sinTheta = sinTheta1; - } -} - -/* CENTRY */ -void GLUTAPIENTRY -glutWireTorus(GLdouble innerRadius, GLdouble outerRadius, - GLint nsides, GLint rings) -{ - glPushAttrib(GL_POLYGON_BIT); - glPolygonMode(GL_FRONT_AND_BACK, GL_LINE); - doughnut(innerRadius, outerRadius, nsides, rings); - glPopAttrib(); -} - -void GLUTAPIENTRY -glutSolidTorus(GLdouble innerRadius, GLdouble outerRadius, - GLint nsides, GLint rings) -{ - doughnut(innerRadius, outerRadius, nsides, rings); -} - -/* ENDCENTRY */ - -static GLfloat dodec[20][3]; - -static void -initDodecahedron(void) -{ - GLfloat alpha, beta; - - alpha = sqrt(2.0 / (3.0 + sqrt(5.0))); - beta = 1.0 + sqrt(6.0 / (3.0 + sqrt(5.0)) - - 2.0 + 2.0 * sqrt(2.0 / (3.0 + sqrt(5.0)))); - /* *INDENT-OFF* */ - dodec[0][0] = -alpha; dodec[0][1] = 0; dodec[0][2] = beta; - dodec[1][0] = alpha; dodec[1][1] = 0; dodec[1][2] = beta; - dodec[2][0] = -1; dodec[2][1] = -1; dodec[2][2] = -1; - dodec[3][0] = -1; dodec[3][1] = -1; dodec[3][2] = 1; - dodec[4][0] = -1; dodec[4][1] = 1; dodec[4][2] = -1; - dodec[5][0] = -1; dodec[5][1] = 1; dodec[5][2] = 1; - dodec[6][0] = 1; dodec[6][1] = -1; dodec[6][2] = -1; - dodec[7][0] = 1; dodec[7][1] = -1; dodec[7][2] = 1; - dodec[8][0] = 1; dodec[8][1] = 1; dodec[8][2] = -1; - dodec[9][0] = 1; dodec[9][1] = 1; dodec[9][2] = 1; - dodec[10][0] = beta; dodec[10][1] = alpha; dodec[10][2] = 0; - dodec[11][0] = beta; dodec[11][1] = -alpha; dodec[11][2] = 0; - dodec[12][0] = -beta; dodec[12][1] = alpha; dodec[12][2] = 0; - dodec[13][0] = -beta; dodec[13][1] = -alpha; dodec[13][2] = 0; - dodec[14][0] = -alpha; dodec[14][1] = 0; dodec[14][2] = -beta; - dodec[15][0] = alpha; dodec[15][1] = 0; dodec[15][2] = -beta; - dodec[16][0] = 0; dodec[16][1] = beta; dodec[16][2] = alpha; - dodec[17][0] = 0; dodec[17][1] = beta; dodec[17][2] = -alpha; - dodec[18][0] = 0; dodec[18][1] = -beta; dodec[18][2] = alpha; - dodec[19][0] = 0; dodec[19][1] = -beta; dodec[19][2] = -alpha; - /* *INDENT-ON* */ - -} - -#define DIFF3(_a,_b,_c) { \ - (_c)[0] = (_a)[0] - (_b)[0]; \ - (_c)[1] = (_a)[1] - (_b)[1]; \ - (_c)[2] = (_a)[2] - (_b)[2]; \ -} - -static void -crossprod(GLfloat v1[3], GLfloat v2[3], GLfloat prod[3]) -{ - GLfloat p[3]; /* in case prod == v1 or v2 */ - - p[0] = v1[1] * v2[2] - v2[1] * v1[2]; - p[1] = v1[2] * v2[0] - v2[2] * v1[0]; - p[2] = v1[0] * v2[1] - v2[0] * v1[1]; - prod[0] = p[0]; - prod[1] = p[1]; - prod[2] = p[2]; -} - -static void -normalize(GLfloat v[3]) -{ - GLfloat d; - - d = sqrt(v[0] * v[0] + v[1] * v[1] + v[2] * v[2]); - if (d == 0.0) { - __glutWarning("normalize: zero length vector"); - v[0] = d = 1.0; - } - d = 1 / d; - v[0] *= d; - v[1] *= d; - v[2] *= d; -} - -static void -pentagon(int a, int b, int c, int d, int e, GLenum shadeType) -{ - GLfloat n0[3], d1[3], d2[3]; - - DIFF3(dodec[a], dodec[b], d1); - DIFF3(dodec[b], dodec[c], d2); - crossprod(d1, d2, n0); - normalize(n0); - - glBegin(shadeType); - glNormal3fv(n0); - glVertex3fv(&dodec[a][0]); - glVertex3fv(&dodec[b][0]); - glVertex3fv(&dodec[c][0]); - glVertex3fv(&dodec[d][0]); - glVertex3fv(&dodec[e][0]); - glEnd(); -} - -static void -dodecahedron(GLenum type) -{ - static int inited = 0; - - if (inited == 0) { - inited = 1; - initDodecahedron(); - } - pentagon(0, 1, 9, 16, 5, type); - pentagon(1, 0, 3, 18, 7, type); - pentagon(1, 7, 11, 10, 9, type); - pentagon(11, 7, 18, 19, 6, type); - pentagon(8, 17, 16, 9, 10, type); - pentagon(2, 14, 15, 6, 19, type); - pentagon(2, 13, 12, 4, 14, type); - pentagon(2, 19, 18, 3, 13, type); - pentagon(3, 0, 5, 12, 13, type); - pentagon(6, 15, 8, 10, 11, type); - pentagon(4, 17, 8, 15, 14, type); - pentagon(4, 12, 5, 16, 17, type); -} - -/* CENTRY */ -void GLUTAPIENTRY glutWireDodecahedron(void) -{ - dodecahedron(GL_LINE_LOOP); -} - -void GLUTAPIENTRY -glutSolidDodecahedron(void) -{ - dodecahedron(GL_TRIANGLE_FAN); -} - -/* ENDCENTRY */ - -static void -recorditem(GLfloat * n1, GLfloat * n2, GLfloat * n3, - GLenum shadeType) -{ - GLfloat q0[3], q1[3]; - - DIFF3(n1, n2, q0); - DIFF3(n2, n3, q1); - crossprod(q0, q1, q1); - normalize(q1); - - glBegin(shadeType); - glNormal3fv(q1); - glVertex3fv(n1); - glVertex3fv(n2); - glVertex3fv(n3); - glEnd(); -} - -static void -subdivide(GLfloat * v0, GLfloat * v1, GLfloat * v2, - GLenum shadeType) -{ - int depth; - GLfloat w0[3], w1[3], w2[3]; - GLfloat l; - int i, j, k, n; - - depth = 1; - for (i = 0; i < depth; i++) { - for (j = 0; i + j < depth; j++) { - k = depth - i - j; - for (n = 0; n < 3; n++) { - w0[n] = (i * v0[n] + j * v1[n] + k * v2[n]) / depth; - w1[n] = ((i + 1) * v0[n] + j * v1[n] + (k - 1) * v2[n]) - / depth; - w2[n] = (i * v0[n] + (j + 1) * v1[n] + (k - 1) * v2[n]) - / depth; - } - l = sqrt(w0[0] * w0[0] + w0[1] * w0[1] + w0[2] * w0[2]); - w0[0] /= l; - w0[1] /= l; - w0[2] /= l; - l = sqrt(w1[0] * w1[0] + w1[1] * w1[1] + w1[2] * w1[2]); - w1[0] /= l; - w1[1] /= l; - w1[2] /= l; - l = sqrt(w2[0] * w2[0] + w2[1] * w2[1] + w2[2] * w2[2]); - w2[0] /= l; - w2[1] /= l; - w2[2] /= l; - recorditem(w1, w0, w2, shadeType); - } - } -} - -static void -drawtriangle(int i, GLfloat data[][3], int ndx[][3], - GLenum shadeType) -{ - GLfloat *x0, *x1, *x2; - - x0 = data[ndx[i][0]]; - x1 = data[ndx[i][1]]; - x2 = data[ndx[i][2]]; - subdivide(x0, x1, x2, shadeType); -} - -/* octahedron data: The octahedron produced is centered at the - origin and has radius 1.0 */ -static GLfloat odata[6][3] = -{ - {1.0, 0.0, 0.0}, - {-1.0, 0.0, 0.0}, - {0.0, 1.0, 0.0}, - {0.0, -1.0, 0.0}, - {0.0, 0.0, 1.0}, - {0.0, 0.0, -1.0} -}; - -static int ondex[8][3] = -{ - {0, 4, 2}, - {1, 2, 4}, - {0, 3, 4}, - {1, 4, 3}, - {0, 2, 5}, - {1, 5, 2}, - {0, 5, 3}, - {1, 3, 5} -}; - -static void -octahedron(GLenum shadeType) -{ - int i; - - for (i = 7; i >= 0; i--) { - drawtriangle(i, odata, ondex, shadeType); - } -} - -/* CENTRY */ -void GLUTAPIENTRY -glutWireOctahedron(void) -{ - octahedron(GL_LINE_LOOP); -} - -void GLUTAPIENTRY -glutSolidOctahedron(void) -{ - octahedron(GL_TRIANGLES); -} - -/* ENDCENTRY */ - -/* icosahedron data: These numbers are rigged to make an - icosahedron of radius 1.0 */ - -#define X .525731112119133606 -#define Z .850650808352039932 - -static GLfloat idata[12][3] = -{ - {-X, 0, Z}, - {X, 0, Z}, - {-X, 0, -Z}, - {X, 0, -Z}, - {0, Z, X}, - {0, Z, -X}, - {0, -Z, X}, - {0, -Z, -X}, - {Z, X, 0}, - {-Z, X, 0}, - {Z, -X, 0}, - {-Z, -X, 0} -}; - -static int index[20][3] = -{ - {0, 4, 1}, - {0, 9, 4}, - {9, 5, 4}, - {4, 5, 8}, - {4, 8, 1}, - {8, 10, 1}, - {8, 3, 10}, - {5, 3, 8}, - {5, 2, 3}, - {2, 7, 3}, - {7, 10, 3}, - {7, 6, 10}, - {7, 11, 6}, - {11, 0, 6}, - {0, 1, 6}, - {6, 1, 10}, - {9, 0, 11}, - {9, 11, 2}, - {9, 2, 5}, - {7, 2, 11}, -}; - -static void -icosahedron(GLenum shadeType) -{ - int i; - - for (i = 19; i >= 0; i--) { - drawtriangle(i, idata, index, shadeType); - } -} - -/* CENTRY */ -void GLUTAPIENTRY -glutWireIcosahedron(void) -{ - icosahedron(GL_LINE_LOOP); -} - -void GLUTAPIENTRY -glutSolidIcosahedron(void) -{ - icosahedron(GL_TRIANGLES); -} - -/* ENDCENTRY */ - -/* tetrahedron data: */ - -#define T 1.73205080756887729 - -static GLfloat tdata[4][3] = -{ - {T, T, T}, - {T, -T, -T}, - {-T, T, -T}, - {-T, -T, T} -}; - -static int tndex[4][3] = -{ - {0, 1, 3}, - {2, 1, 0}, - {3, 2, 0}, - {1, 2, 3} -}; - -static void -tetrahedron(GLenum shadeType) -{ - int i; - - for (i = 3; i >= 0; i--) - drawtriangle(i, tdata, tndex, shadeType); -} - -/* CENTRY */ -void GLUTAPIENTRY -glutWireTetrahedron(void) -{ - tetrahedron(GL_LINE_LOOP); -} - -void GLUTAPIENTRY -glutSolidTetrahedron(void) -{ - tetrahedron(GL_TRIANGLES); -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ + +/** +(c) Copyright 1993, Silicon Graphics, Inc. + +ALL RIGHTS RESERVED + +Permission to use, copy, modify, and distribute this software +for any purpose and without fee is hereby granted, provided +that the above copyright notice appear in all copies and that +both the copyright notice and this permission notice appear in +supporting documentation, and that the name of Silicon +Graphics, Inc. not be used in advertising or publicity +pertaining to distribution of the software without specific, +written prior permission. + +THE MATERIAL EMBODIED ON THIS SOFTWARE IS PROVIDED TO YOU +"AS-IS" AND WITHOUT WARRANTY OF ANY KIND, EXPRESS, IMPLIED OR +OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY WARRANTY OF +MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. IN NO +EVENT SHALL SILICON GRAPHICS, INC. BE LIABLE TO YOU OR ANYONE +ELSE FOR ANY DIRECT, SPECIAL, INCIDENTAL, INDIRECT OR +CONSEQUENTIAL DAMAGES OF ANY KIND, OR ANY DAMAGES WHATSOEVER, +INCLUDING WITHOUT LIMITATION, LOSS OF PROFIT, LOSS OF USE, +SAVINGS OR REVENUE, OR THE CLAIMS OF THIRD PARTIES, WHETHER OR +NOT SILICON GRAPHICS, INC. HAS BEEN ADVISED OF THE POSSIBILITY +OF SUCH LOSS, HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, +ARISING OUT OF OR IN CONNECTION WITH THE POSSESSION, USE OR +PERFORMANCE OF THIS SOFTWARE. + +US Government Users Restricted Rights + +Use, duplication, or disclosure by the Government is subject to +restrictions set forth in FAR 52.227.19(c)(2) or subparagraph +(c)(1)(ii) of the Rights in Technical Data and Computer +Software clause at DFARS 252.227-7013 and/or in similar or +successor clauses in the FAR or the DOD or NASA FAR +Supplement. Unpublished-- rights reserved under the copyright +laws of the United States. Contractor/manufacturer is Silicon +Graphics, Inc., 2011 N. Shoreline Blvd., Mountain View, CA +94039-7311. + +OpenGL(TM) is a trademark of Silicon Graphics, Inc. +*/ + +#include +#include "glutint.h" + +/* Some files do not define M_PI... */ +#ifndef M_PI +#define M_PI 3.14159265358979323846 +#endif + +static GLUquadricObj *quadObj; + +#define QUAD_OBJ_INIT() { if(!quadObj) initQuadObj(); } + +static void +initQuadObj(void) +{ + quadObj = gluNewQuadric(); + if (!quadObj) + __glutFatalError("out of memory."); +} + +/* CENTRY */ +void GLUTAPIENTRY +glutWireSphere(GLdouble radius, GLint slices, GLint stacks) +{ + QUAD_OBJ_INIT(); + gluQuadricDrawStyle(quadObj, GLU_LINE); + gluQuadricNormals(quadObj, GLU_SMOOTH); + /* If we ever changed/used the texture or orientation state + of quadObj, we'd need to change it to the defaults here + with gluQuadricTexture and/or gluQuadricOrientation. */ + gluSphere(quadObj, radius, slices, stacks); +} + +void GLUTAPIENTRY +glutSolidSphere(GLdouble radius, GLint slices, GLint stacks) +{ + QUAD_OBJ_INIT(); + gluQuadricDrawStyle(quadObj, GLU_FILL); + gluQuadricNormals(quadObj, GLU_SMOOTH); + /* If we ever changed/used the texture or orientation state + of quadObj, we'd need to change it to the defaults here + with gluQuadricTexture and/or gluQuadricOrientation. */ + gluSphere(quadObj, radius, slices, stacks); +} + +void GLUTAPIENTRY +glutWireCone(GLdouble base, GLdouble height, + GLint slices, GLint stacks) +{ + QUAD_OBJ_INIT(); + gluQuadricDrawStyle(quadObj, GLU_LINE); + gluQuadricNormals(quadObj, GLU_SMOOTH); + /* If we ever changed/used the texture or orientation state + of quadObj, we'd need to change it to the defaults here + with gluQuadricTexture and/or gluQuadricOrientation. */ + gluCylinder(quadObj, base, 0.0, height, slices, stacks); +} + +void GLUTAPIENTRY +glutSolidCone(GLdouble base, GLdouble height, + GLint slices, GLint stacks) +{ + QUAD_OBJ_INIT(); + gluQuadricDrawStyle(quadObj, GLU_FILL); + gluQuadricNormals(quadObj, GLU_SMOOTH); + /* If we ever changed/used the texture or orientation state + of quadObj, we'd need to change it to the defaults here + with gluQuadricTexture and/or gluQuadricOrientation. */ + gluCylinder(quadObj, base, 0.0, height, slices, stacks); +} + +/* ENDCENTRY */ + +static void +drawBox(GLfloat size, GLenum type) +{ + static GLfloat n[6][3] = + { + {-1.0, 0.0, 0.0}, + {0.0, 1.0, 0.0}, + {1.0, 0.0, 0.0}, + {0.0, -1.0, 0.0}, + {0.0, 0.0, 1.0}, + {0.0, 0.0, -1.0} + }; + static GLint faces[6][4] = + { + {0, 1, 2, 3}, + {3, 2, 6, 7}, + {7, 6, 5, 4}, + {4, 5, 1, 0}, + {5, 6, 2, 1}, + {7, 4, 0, 3} + }; + GLfloat v[8][3]; + GLint i; + + v[0][0] = v[1][0] = v[2][0] = v[3][0] = -size / 2; + v[4][0] = v[5][0] = v[6][0] = v[7][0] = size / 2; + v[0][1] = v[1][1] = v[4][1] = v[5][1] = -size / 2; + v[2][1] = v[3][1] = v[6][1] = v[7][1] = size / 2; + v[0][2] = v[3][2] = v[4][2] = v[7][2] = -size / 2; + v[1][2] = v[2][2] = v[5][2] = v[6][2] = size / 2; + + for (i = 5; i >= 0; i--) { + glBegin(type); + glNormal3fv(&n[i][0]); + glVertex3fv(&v[faces[i][0]][0]); + glVertex3fv(&v[faces[i][1]][0]); + glVertex3fv(&v[faces[i][2]][0]); + glVertex3fv(&v[faces[i][3]][0]); + glEnd(); + } +} + +/* CENTRY */ +void GLUTAPIENTRY +glutWireCube(GLdouble size) +{ + drawBox(size, GL_LINE_LOOP); +} + +void GLUTAPIENTRY +glutSolidCube(GLdouble size) +{ + drawBox(size, GL_QUADS); +} + +/* ENDCENTRY */ + +static void +doughnut(GLfloat r, GLfloat R, GLint nsides, GLint rings) +{ + int i, j; + GLfloat theta, phi, theta1; + GLfloat cosTheta, sinTheta; + GLfloat cosTheta1, sinTheta1; + GLfloat ringDelta, sideDelta; + + ringDelta = 2.0 * M_PI / rings; + sideDelta = 2.0 * M_PI / nsides; + + theta = 0.0; + cosTheta = 1.0; + sinTheta = 0.0; + for (i = rings - 1; i >= 0; i--) { + theta1 = theta + ringDelta; + cosTheta1 = cos(theta1); + sinTheta1 = sin(theta1); + glBegin(GL_QUAD_STRIP); + phi = 0.0; + for (j = nsides; j >= 0; j--) { + GLfloat cosPhi, sinPhi, dist; + + phi += sideDelta; + cosPhi = cos(phi); + sinPhi = sin(phi); + dist = R + r * cosPhi; + + glNormal3f(cosTheta1 * cosPhi, -sinTheta1 * cosPhi, sinPhi); + glVertex3f(cosTheta1 * dist, -sinTheta1 * dist, r * sinPhi); + glNormal3f(cosTheta * cosPhi, -sinTheta * cosPhi, sinPhi); + glVertex3f(cosTheta * dist, -sinTheta * dist, r * sinPhi); + } + glEnd(); + theta = theta1; + cosTheta = cosTheta1; + sinTheta = sinTheta1; + } +} + +/* CENTRY */ +void GLUTAPIENTRY +glutWireTorus(GLdouble innerRadius, GLdouble outerRadius, + GLint nsides, GLint rings) +{ + glPushAttrib(GL_POLYGON_BIT); + glPolygonMode(GL_FRONT_AND_BACK, GL_LINE); + doughnut(innerRadius, outerRadius, nsides, rings); + glPopAttrib(); +} + +void GLUTAPIENTRY +glutSolidTorus(GLdouble innerRadius, GLdouble outerRadius, + GLint nsides, GLint rings) +{ + doughnut(innerRadius, outerRadius, nsides, rings); +} + +/* ENDCENTRY */ + +static GLfloat dodec[20][3]; + +static void +initDodecahedron(void) +{ + GLfloat alpha, beta; + + alpha = sqrt(2.0 / (3.0 + sqrt(5.0))); + beta = 1.0 + sqrt(6.0 / (3.0 + sqrt(5.0)) - + 2.0 + 2.0 * sqrt(2.0 / (3.0 + sqrt(5.0)))); + /* *INDENT-OFF* */ + dodec[0][0] = -alpha; dodec[0][1] = 0; dodec[0][2] = beta; + dodec[1][0] = alpha; dodec[1][1] = 0; dodec[1][2] = beta; + dodec[2][0] = -1; dodec[2][1] = -1; dodec[2][2] = -1; + dodec[3][0] = -1; dodec[3][1] = -1; dodec[3][2] = 1; + dodec[4][0] = -1; dodec[4][1] = 1; dodec[4][2] = -1; + dodec[5][0] = -1; dodec[5][1] = 1; dodec[5][2] = 1; + dodec[6][0] = 1; dodec[6][1] = -1; dodec[6][2] = -1; + dodec[7][0] = 1; dodec[7][1] = -1; dodec[7][2] = 1; + dodec[8][0] = 1; dodec[8][1] = 1; dodec[8][2] = -1; + dodec[9][0] = 1; dodec[9][1] = 1; dodec[9][2] = 1; + dodec[10][0] = beta; dodec[10][1] = alpha; dodec[10][2] = 0; + dodec[11][0] = beta; dodec[11][1] = -alpha; dodec[11][2] = 0; + dodec[12][0] = -beta; dodec[12][1] = alpha; dodec[12][2] = 0; + dodec[13][0] = -beta; dodec[13][1] = -alpha; dodec[13][2] = 0; + dodec[14][0] = -alpha; dodec[14][1] = 0; dodec[14][2] = -beta; + dodec[15][0] = alpha; dodec[15][1] = 0; dodec[15][2] = -beta; + dodec[16][0] = 0; dodec[16][1] = beta; dodec[16][2] = alpha; + dodec[17][0] = 0; dodec[17][1] = beta; dodec[17][2] = -alpha; + dodec[18][0] = 0; dodec[18][1] = -beta; dodec[18][2] = alpha; + dodec[19][0] = 0; dodec[19][1] = -beta; dodec[19][2] = -alpha; + /* *INDENT-ON* */ + +} + +#define DIFF3(_a,_b,_c) { \ + (_c)[0] = (_a)[0] - (_b)[0]; \ + (_c)[1] = (_a)[1] - (_b)[1]; \ + (_c)[2] = (_a)[2] - (_b)[2]; \ +} + +static void +crossprod(GLfloat v1[3], GLfloat v2[3], GLfloat prod[3]) +{ + GLfloat p[3]; /* in case prod == v1 or v2 */ + + p[0] = v1[1] * v2[2] - v2[1] * v1[2]; + p[1] = v1[2] * v2[0] - v2[2] * v1[0]; + p[2] = v1[0] * v2[1] - v2[0] * v1[1]; + prod[0] = p[0]; + prod[1] = p[1]; + prod[2] = p[2]; +} + +static void +normalize(GLfloat v[3]) +{ + GLfloat d; + + d = sqrt(v[0] * v[0] + v[1] * v[1] + v[2] * v[2]); + if (d == 0.0) { + __glutWarning("normalize: zero length vector"); + v[0] = d = 1.0; + } + d = 1 / d; + v[0] *= d; + v[1] *= d; + v[2] *= d; +} + +static void +pentagon(int a, int b, int c, int d, int e, GLenum shadeType) +{ + GLfloat n0[3], d1[3], d2[3]; + + DIFF3(dodec[a], dodec[b], d1); + DIFF3(dodec[b], dodec[c], d2); + crossprod(d1, d2, n0); + normalize(n0); + + glBegin(shadeType); + glNormal3fv(n0); + glVertex3fv(&dodec[a][0]); + glVertex3fv(&dodec[b][0]); + glVertex3fv(&dodec[c][0]); + glVertex3fv(&dodec[d][0]); + glVertex3fv(&dodec[e][0]); + glEnd(); +} + +static void +dodecahedron(GLenum type) +{ + static int inited = 0; + + if (inited == 0) { + inited = 1; + initDodecahedron(); + } + pentagon(0, 1, 9, 16, 5, type); + pentagon(1, 0, 3, 18, 7, type); + pentagon(1, 7, 11, 10, 9, type); + pentagon(11, 7, 18, 19, 6, type); + pentagon(8, 17, 16, 9, 10, type); + pentagon(2, 14, 15, 6, 19, type); + pentagon(2, 13, 12, 4, 14, type); + pentagon(2, 19, 18, 3, 13, type); + pentagon(3, 0, 5, 12, 13, type); + pentagon(6, 15, 8, 10, 11, type); + pentagon(4, 17, 8, 15, 14, type); + pentagon(4, 12, 5, 16, 17, type); +} + +/* CENTRY */ +void GLUTAPIENTRY glutWireDodecahedron(void) +{ + dodecahedron(GL_LINE_LOOP); +} + +void GLUTAPIENTRY +glutSolidDodecahedron(void) +{ + dodecahedron(GL_TRIANGLE_FAN); +} + +/* ENDCENTRY */ + +static void +recorditem(GLfloat * n1, GLfloat * n2, GLfloat * n3, + GLenum shadeType) +{ + GLfloat q0[3], q1[3]; + + DIFF3(n1, n2, q0); + DIFF3(n2, n3, q1); + crossprod(q0, q1, q1); + normalize(q1); + + glBegin(shadeType); + glNormal3fv(q1); + glVertex3fv(n1); + glVertex3fv(n2); + glVertex3fv(n3); + glEnd(); +} + +static void +subdivide(GLfloat * v0, GLfloat * v1, GLfloat * v2, + GLenum shadeType) +{ + int depth; + GLfloat w0[3], w1[3], w2[3]; + GLfloat l; + int i, j, k, n; + + depth = 1; + for (i = 0; i < depth; i++) { + for (j = 0; i + j < depth; j++) { + k = depth - i - j; + for (n = 0; n < 3; n++) { + w0[n] = (i * v0[n] + j * v1[n] + k * v2[n]) / depth; + w1[n] = ((i + 1) * v0[n] + j * v1[n] + (k - 1) * v2[n]) + / depth; + w2[n] = (i * v0[n] + (j + 1) * v1[n] + (k - 1) * v2[n]) + / depth; + } + l = sqrt(w0[0] * w0[0] + w0[1] * w0[1] + w0[2] * w0[2]); + w0[0] /= l; + w0[1] /= l; + w0[2] /= l; + l = sqrt(w1[0] * w1[0] + w1[1] * w1[1] + w1[2] * w1[2]); + w1[0] /= l; + w1[1] /= l; + w1[2] /= l; + l = sqrt(w2[0] * w2[0] + w2[1] * w2[1] + w2[2] * w2[2]); + w2[0] /= l; + w2[1] /= l; + w2[2] /= l; + recorditem(w1, w0, w2, shadeType); + } + } +} + +static void +drawtriangle(int i, GLfloat data[][3], int ndx[][3], + GLenum shadeType) +{ + GLfloat *x0, *x1, *x2; + + x0 = data[ndx[i][0]]; + x1 = data[ndx[i][1]]; + x2 = data[ndx[i][2]]; + subdivide(x0, x1, x2, shadeType); +} + +/* octahedron data: The octahedron produced is centered at the + origin and has radius 1.0 */ +static GLfloat odata[6][3] = +{ + {1.0, 0.0, 0.0}, + {-1.0, 0.0, 0.0}, + {0.0, 1.0, 0.0}, + {0.0, -1.0, 0.0}, + {0.0, 0.0, 1.0}, + {0.0, 0.0, -1.0} +}; + +static int ondex[8][3] = +{ + {0, 4, 2}, + {1, 2, 4}, + {0, 3, 4}, + {1, 4, 3}, + {0, 2, 5}, + {1, 5, 2}, + {0, 5, 3}, + {1, 3, 5} +}; + +static void +octahedron(GLenum shadeType) +{ + int i; + + for (i = 7; i >= 0; i--) { + drawtriangle(i, odata, ondex, shadeType); + } +} + +/* CENTRY */ +void GLUTAPIENTRY +glutWireOctahedron(void) +{ + octahedron(GL_LINE_LOOP); +} + +void GLUTAPIENTRY +glutSolidOctahedron(void) +{ + octahedron(GL_TRIANGLES); +} + +/* ENDCENTRY */ + +/* icosahedron data: These numbers are rigged to make an + icosahedron of radius 1.0 */ + +#define X .525731112119133606 +#define Z .850650808352039932 + +static GLfloat idata[12][3] = +{ + {-X, 0, Z}, + {X, 0, Z}, + {-X, 0, -Z}, + {X, 0, -Z}, + {0, Z, X}, + {0, Z, -X}, + {0, -Z, X}, + {0, -Z, -X}, + {Z, X, 0}, + {-Z, X, 0}, + {Z, -X, 0}, + {-Z, -X, 0} +}; + +static int index[20][3] = +{ + {0, 4, 1}, + {0, 9, 4}, + {9, 5, 4}, + {4, 5, 8}, + {4, 8, 1}, + {8, 10, 1}, + {8, 3, 10}, + {5, 3, 8}, + {5, 2, 3}, + {2, 7, 3}, + {7, 10, 3}, + {7, 6, 10}, + {7, 11, 6}, + {11, 0, 6}, + {0, 1, 6}, + {6, 1, 10}, + {9, 0, 11}, + {9, 11, 2}, + {9, 2, 5}, + {7, 2, 11}, +}; + +static void +icosahedron(GLenum shadeType) +{ + int i; + + for (i = 19; i >= 0; i--) { + drawtriangle(i, idata, index, shadeType); + } +} + +/* CENTRY */ +void GLUTAPIENTRY +glutWireIcosahedron(void) +{ + icosahedron(GL_LINE_LOOP); +} + +void GLUTAPIENTRY +glutSolidIcosahedron(void) +{ + icosahedron(GL_TRIANGLES); +} + +/* ENDCENTRY */ + +/* tetrahedron data: */ + +#define T 1.73205080756887729 + +static GLfloat tdata[4][3] = +{ + {T, T, T}, + {T, -T, -T}, + {-T, T, -T}, + {-T, -T, T} +}; + +static int tndex[4][3] = +{ + {0, 1, 3}, + {2, 1, 0}, + {3, 2, 0}, + {1, 2, 3} +}; + +static void +tetrahedron(GLenum shadeType) +{ + int i; + + for (i = 3; i >= 0; i--) + drawtriangle(i, tdata, tndex, shadeType); +} + +/* CENTRY */ +void GLUTAPIENTRY +glutWireTetrahedron(void) +{ + tetrahedron(GL_LINE_LOOP); +} + +void GLUTAPIENTRY +glutSolidTetrahedron(void) +{ + tetrahedron(GL_TRIANGLES); +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_stroke.cpp b/src/glut/os2/glut_stroke.cpp index 4823283c8c..ac231e4973 100644 --- a/src/glut/os2/glut_stroke.cpp +++ b/src/glut/os2/glut_stroke.cpp @@ -1,42 +1,42 @@ - -/* Copyright (c) Mark J. Kilgard, 1994. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include "glutint.h" -#include "glutstroke.h" - -void GLUTAPIENTRY -glutStrokeCharacter(GLUTstrokeFont font, int c) -{ - const StrokeCharRec *ch; - const StrokeRec *stroke; - const CoordRec *coord; - StrokeFontPtr fontinfo; - int i, j; - - -#if defined(_WIN32) - fontinfo = (StrokeFontPtr) __glutFont(font); -#else - fontinfo = (StrokeFontPtr) font; -#endif - - if (c < 0 || c >= fontinfo->num_chars) - return; - ch = &(fontinfo->ch[c]); - if (ch) { - for (i = ch->num_strokes, stroke = ch->stroke; - i > 0; i--, stroke++) { - glBegin(GL_LINE_STRIP); - for (j = stroke->num_coords, coord = stroke->coord; - j > 0; j--, coord++) { - glVertex2f(coord->x, coord->y); - } - glEnd(); - } - glTranslatef(ch->right, 0.0, 0.0); - } -} + +/* Copyright (c) Mark J. Kilgard, 1994. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include "glutint.h" +#include "glutstroke.h" + +void GLUTAPIENTRY +glutStrokeCharacter(GLUTstrokeFont font, int c) +{ + const StrokeCharRec *ch; + const StrokeRec *stroke; + const CoordRec *coord; + StrokeFontPtr fontinfo; + int i, j; + + +#if defined(_WIN32) + fontinfo = (StrokeFontPtr) __glutFont(font); +#else + fontinfo = (StrokeFontPtr) font; +#endif + + if (c < 0 || c >= fontinfo->num_chars) + return; + ch = &(fontinfo->ch[c]); + if (ch) { + for (i = ch->num_strokes, stroke = ch->stroke; + i > 0; i--, stroke++) { + glBegin(GL_LINE_STRIP); + for (j = stroke->num_coords, coord = stroke->coord; + j > 0; j--, coord++) { + glVertex2f(coord->x, coord->y); + } + glEnd(); + } + glTranslatef(ch->right, 0.0, 0.0); + } +} diff --git a/src/glut/os2/glut_swap.cpp b/src/glut/os2/glut_swap.cpp index 210a7fc37e..4831b8eb35 100644 --- a/src/glut/os2/glut_swap.cpp +++ b/src/glut/os2/glut_swap.cpp @@ -1,67 +1,67 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include -#include "glutint.h" - -GLint __glutFPS = 0; -GLint __glutSwapCount = 0; -GLint __glutSwapTime = 0; - -/* CENTRY */ -void GLUTAPIENTRY -glutSwapBuffers(void) -{ - GLUTwindow *window = __glutCurrentWindow; - - if (window->renderWin == window->win) { - if (__glutCurrentWindow->treatAsSingle) { - /* Pretend the double buffered window is single buffered, - so treat glutSwapBuffers as a no-op. */ - return; - } - } else { - if (__glutCurrentWindow->overlay->treatAsSingle) { - /* Pretend the double buffered overlay is single - buffered, so treat glutSwapBuffers as a no-op. */ - return; - } - } - - /* For the MESA_SWAP_HACK. */ - window->usedSwapBuffers = 1; - - SWAP_BUFFERS_LAYER(__glutCurrentWindow); - - /* I considered putting the window being swapped on the - GLUT_FINISH_WORK work list because you could call - glutSwapBuffers from an idle callback which doesn't call - __glutSetWindow which normally adds indirect rendering - windows to the GLUT_FINISH_WORK work list. Not being put - on the list could lead to the buffering up of multiple - redisplays and buffer swaps and hamper interactivity. I - consider this an application bug due to not using - glutPostRedisplay to trigger redraws. If - glutPostRedisplay were used, __glutSetWindow would be - called and a glFinish to throttle buffering would occur. */ - - if (__glutFPS) { - GLint t = glutGet(GLUT_ELAPSED_TIME); - __glutSwapCount++; - if (__glutSwapTime == 0) - __glutSwapTime = t; - else if (t - __glutSwapTime > __glutFPS) { - float time = 0.001 * (t - __glutSwapTime); - float fps = (float) __glutSwapCount / time; - fprintf(stderr, "GLUT: %d frames in %.2f seconds = %.2f FPS\n", - __glutSwapCount, time, fps); - __glutSwapTime = t; - __glutSwapCount = 0; - } - } -} -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include +#include "glutint.h" + +GLint __glutFPS = 0; +GLint __glutSwapCount = 0; +GLint __glutSwapTime = 0; + +/* CENTRY */ +void GLUTAPIENTRY +glutSwapBuffers(void) +{ + GLUTwindow *window = __glutCurrentWindow; + + if (window->renderWin == window->win) { + if (__glutCurrentWindow->treatAsSingle) { + /* Pretend the double buffered window is single buffered, + so treat glutSwapBuffers as a no-op. */ + return; + } + } else { + if (__glutCurrentWindow->overlay->treatAsSingle) { + /* Pretend the double buffered overlay is single + buffered, so treat glutSwapBuffers as a no-op. */ + return; + } + } + + /* For the MESA_SWAP_HACK. */ + window->usedSwapBuffers = 1; + + SWAP_BUFFERS_LAYER(__glutCurrentWindow); + + /* I considered putting the window being swapped on the + GLUT_FINISH_WORK work list because you could call + glutSwapBuffers from an idle callback which doesn't call + __glutSetWindow which normally adds indirect rendering + windows to the GLUT_FINISH_WORK work list. Not being put + on the list could lead to the buffering up of multiple + redisplays and buffer swaps and hamper interactivity. I + consider this an application bug due to not using + glutPostRedisplay to trigger redraws. If + glutPostRedisplay were used, __glutSetWindow would be + called and a glFinish to throttle buffering would occur. */ + + if (__glutFPS) { + GLint t = glutGet(GLUT_ELAPSED_TIME); + __glutSwapCount++; + if (__glutSwapTime == 0) + __glutSwapTime = t; + else if (t - __glutSwapTime > __glutFPS) { + float time = 0.001 * (t - __glutSwapTime); + float fps = (float) __glutSwapCount / time; + fprintf(stderr, "GLUT: %d frames in %.2f seconds = %.2f FPS\n", + __glutSwapCount, time, fps); + __glutSwapTime = t; + __glutSwapCount = 0; + } + } +} +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_teapot.cpp b/src/glut/os2/glut_teapot.cpp index e2fcb53152..31bbbbd6df 100644 --- a/src/glut/os2/glut_teapot.cpp +++ b/src/glut/os2/glut_teapot.cpp @@ -1,210 +1,210 @@ - -/* Copyright (c) Mark J. Kilgard, 1994. */ - -/** -(c) Copyright 1993, Silicon Graphics, Inc. - -ALL RIGHTS RESERVED - -Permission to use, copy, modify, and distribute this software -for any purpose and without fee is hereby granted, provided -that the above copyright notice appear in all copies and that -both the copyright notice and this permission notice appear in -supporting documentation, and that the name of Silicon -Graphics, Inc. not be used in advertising or publicity -pertaining to distribution of the software without specific, -written prior permission. - -THE MATERIAL EMBODIED ON THIS SOFTWARE IS PROVIDED TO YOU -"AS-IS" AND WITHOUT WARRANTY OF ANY KIND, EXPRESS, IMPLIED OR -OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY WARRANTY OF -MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. IN NO -EVENT SHALL SILICON GRAPHICS, INC. BE LIABLE TO YOU OR ANYONE -ELSE FOR ANY DIRECT, SPECIAL, INCIDENTAL, INDIRECT OR -CONSEQUENTIAL DAMAGES OF ANY KIND, OR ANY DAMAGES WHATSOEVER, -INCLUDING WITHOUT LIMITATION, LOSS OF PROFIT, LOSS OF USE, -SAVINGS OR REVENUE, OR THE CLAIMS OF THIRD PARTIES, WHETHER OR -NOT SILICON GRAPHICS, INC. HAS BEEN ADVISED OF THE POSSIBILITY -OF SUCH LOSS, HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, -ARISING OUT OF OR IN CONNECTION WITH THE POSSESSION, USE OR -PERFORMANCE OF THIS SOFTWARE. - -US Government Users Restricted Rights - -Use, duplication, or disclosure by the Government is subject to -restrictions set forth in FAR 52.227.19(c)(2) or subparagraph -(c)(1)(ii) of the Rights in Technical Data and Computer -Software clause at DFARS 252.227-7013 and/or in similar or -successor clauses in the FAR or the DOD or NASA FAR -Supplement. Unpublished-- rights reserved under the copyright -laws of the United States. Contractor/manufacturer is Silicon -Graphics, Inc., 2011 N. Shoreline Blvd., Mountain View, CA -94039-7311. - -OpenGL(TM) is a trademark of Silicon Graphics, Inc. -*/ - -#include "glutint.h" - -/* Rim, body, lid, and bottom data must be reflected in x and - y; handle and spout data across the y axis only. */ - -static int patchdata[][16] = -{ - /* rim */ - {102, 103, 104, 105, 4, 5, 6, 7, 8, 9, 10, 11, - 12, 13, 14, 15}, - /* body */ - {12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, - 24, 25, 26, 27}, - {24, 25, 26, 27, 29, 30, 31, 32, 33, 34, 35, 36, - 37, 38, 39, 40}, - /* lid */ - {96, 96, 96, 96, 97, 98, 99, 100, 101, 101, 101, - 101, 0, 1, 2, 3,}, - {0, 1, 2, 3, 106, 107, 108, 109, 110, 111, 112, - 113, 114, 115, 116, 117}, - /* bottom */ - {118, 118, 118, 118, 124, 122, 119, 121, 123, 126, - 125, 120, 40, 39, 38, 37}, - /* handle */ - {41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, - 53, 54, 55, 56}, - {53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, - 28, 65, 66, 67}, - /* spout */ - {68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, - 80, 81, 82, 83}, - {80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, - 92, 93, 94, 95} -}; -/* *INDENT-OFF* */ - -static float cpdata[][3] = -{ - {0.2, 0, 2.7}, {0.2, -0.112, 2.7}, {0.112, -0.2, 2.7}, {0, - -0.2, 2.7}, {1.3375, 0, 2.53125}, {1.3375, -0.749, 2.53125}, - {0.749, -1.3375, 2.53125}, {0, -1.3375, 2.53125}, {1.4375, - 0, 2.53125}, {1.4375, -0.805, 2.53125}, {0.805, -1.4375, - 2.53125}, {0, -1.4375, 2.53125}, {1.5, 0, 2.4}, {1.5, -0.84, - 2.4}, {0.84, -1.5, 2.4}, {0, -1.5, 2.4}, {1.75, 0, 1.875}, - {1.75, -0.98, 1.875}, {0.98, -1.75, 1.875}, {0, -1.75, - 1.875}, {2, 0, 1.35}, {2, -1.12, 1.35}, {1.12, -2, 1.35}, - {0, -2, 1.35}, {2, 0, 0.9}, {2, -1.12, 0.9}, {1.12, -2, - 0.9}, {0, -2, 0.9}, {-2, 0, 0.9}, {2, 0, 0.45}, {2, -1.12, - 0.45}, {1.12, -2, 0.45}, {0, -2, 0.45}, {1.5, 0, 0.225}, - {1.5, -0.84, 0.225}, {0.84, -1.5, 0.225}, {0, -1.5, 0.225}, - {1.5, 0, 0.15}, {1.5, -0.84, 0.15}, {0.84, -1.5, 0.15}, {0, - -1.5, 0.15}, {-1.6, 0, 2.025}, {-1.6, -0.3, 2.025}, {-1.5, - -0.3, 2.25}, {-1.5, 0, 2.25}, {-2.3, 0, 2.025}, {-2.3, -0.3, - 2.025}, {-2.5, -0.3, 2.25}, {-2.5, 0, 2.25}, {-2.7, 0, - 2.025}, {-2.7, -0.3, 2.025}, {-3, -0.3, 2.25}, {-3, 0, - 2.25}, {-2.7, 0, 1.8}, {-2.7, -0.3, 1.8}, {-3, -0.3, 1.8}, - {-3, 0, 1.8}, {-2.7, 0, 1.575}, {-2.7, -0.3, 1.575}, {-3, - -0.3, 1.35}, {-3, 0, 1.35}, {-2.5, 0, 1.125}, {-2.5, -0.3, - 1.125}, {-2.65, -0.3, 0.9375}, {-2.65, 0, 0.9375}, {-2, - -0.3, 0.9}, {-1.9, -0.3, 0.6}, {-1.9, 0, 0.6}, {1.7, 0, - 1.425}, {1.7, -0.66, 1.425}, {1.7, -0.66, 0.6}, {1.7, 0, - 0.6}, {2.6, 0, 1.425}, {2.6, -0.66, 1.425}, {3.1, -0.66, - 0.825}, {3.1, 0, 0.825}, {2.3, 0, 2.1}, {2.3, -0.25, 2.1}, - {2.4, -0.25, 2.025}, {2.4, 0, 2.025}, {2.7, 0, 2.4}, {2.7, - -0.25, 2.4}, {3.3, -0.25, 2.4}, {3.3, 0, 2.4}, {2.8, 0, - 2.475}, {2.8, -0.25, 2.475}, {3.525, -0.25, 2.49375}, - {3.525, 0, 2.49375}, {2.9, 0, 2.475}, {2.9, -0.15, 2.475}, - {3.45, -0.15, 2.5125}, {3.45, 0, 2.5125}, {2.8, 0, 2.4}, - {2.8, -0.15, 2.4}, {3.2, -0.15, 2.4}, {3.2, 0, 2.4}, {0, 0, - 3.15}, {0.8, 0, 3.15}, {0.8, -0.45, 3.15}, {0.45, -0.8, - 3.15}, {0, -0.8, 3.15}, {0, 0, 2.85}, {1.4, 0, 2.4}, {1.4, - -0.784, 2.4}, {0.784, -1.4, 2.4}, {0, -1.4, 2.4}, {0.4, 0, - 2.55}, {0.4, -0.224, 2.55}, {0.224, -0.4, 2.55}, {0, -0.4, - 2.55}, {1.3, 0, 2.55}, {1.3, -0.728, 2.55}, {0.728, -1.3, - 2.55}, {0, -1.3, 2.55}, {1.3, 0, 2.4}, {1.3, -0.728, 2.4}, - {0.728, -1.3, 2.4}, {0, -1.3, 2.4}, {0, 0, 0}, {1.425, - -0.798, 0}, {1.5, 0, 0.075}, {1.425, 0, 0}, {0.798, -1.425, - 0}, {0, -1.5, 0.075}, {0, -1.425, 0}, {1.5, -0.84, 0.075}, - {0.84, -1.5, 0.075} -}; - -static float tex[2][2][2] = -{ - { {0, 0}, - {1, 0}}, - { {0, 1}, - {1, 1}} -}; - -/* *INDENT-ON* */ - -static void -teapot(GLint grid, GLdouble scale, GLenum type) -{ - float p[4][4][3], q[4][4][3], r[4][4][3], s[4][4][3]; - long i, j, k, l; - - glPushAttrib(GL_ENABLE_BIT | GL_EVAL_BIT); - glEnable(GL_AUTO_NORMAL); - glEnable(GL_NORMALIZE); - glEnable(GL_MAP2_VERTEX_3); - glEnable(GL_MAP2_TEXTURE_COORD_2); - glPushMatrix(); - glRotatef(270.0, 1.0, 0.0, 0.0); - glScalef(0.5 * scale, 0.5 * scale, 0.5 * scale); - glTranslatef(0.0, 0.0, -1.5); - for (i = 0; i < 10; i++) { - for (j = 0; j < 4; j++) { - for (k = 0; k < 4; k++) { - for (l = 0; l < 3; l++) { - p[j][k][l] = cpdata[patchdata[i][j * 4 + k]][l]; - q[j][k][l] = cpdata[patchdata[i][j * 4 + (3 - k)]][l]; - if (l == 1) - q[j][k][l] *= -1.0; - if (i < 6) { - r[j][k][l] = - cpdata[patchdata[i][j * 4 + (3 - k)]][l]; - if (l == 0) - r[j][k][l] *= -1.0; - s[j][k][l] = cpdata[patchdata[i][j * 4 + k]][l]; - if (l == 0) - s[j][k][l] *= -1.0; - if (l == 1) - s[j][k][l] *= -1.0; - } - } - } - } - glMap2f(GL_MAP2_TEXTURE_COORD_2, 0, 1, 2, 2, 0, 1, 4, 2, - &tex[0][0][0]); - glMap2f(GL_MAP2_VERTEX_3, 0, 1, 3, 4, 0, 1, 12, 4, - &p[0][0][0]); - glMapGrid2f(grid, 0.0, 1.0, grid, 0.0, 1.0); - glEvalMesh2(type, 0, grid, 0, grid); - glMap2f(GL_MAP2_VERTEX_3, 0, 1, 3, 4, 0, 1, 12, 4, - &q[0][0][0]); - glEvalMesh2(type, 0, grid, 0, grid); - if (i < 6) { - glMap2f(GL_MAP2_VERTEX_3, 0, 1, 3, 4, 0, 1, 12, 4, - &r[0][0][0]); - glEvalMesh2(type, 0, grid, 0, grid); - glMap2f(GL_MAP2_VERTEX_3, 0, 1, 3, 4, 0, 1, 12, 4, - &s[0][0][0]); - glEvalMesh2(type, 0, grid, 0, grid); - } - } - glPopMatrix(); - glPopAttrib(); -} - -/* CENTRY */ -void GLUTAPIENTRY -glutSolidTeapot(GLdouble scale) -{ - teapot(7, scale, GL_FILL); -} - -void GLUTAPIENTRY -glutWireTeapot(GLdouble scale) -{ - teapot(10, scale, GL_LINE); -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994. */ + +/** +(c) Copyright 1993, Silicon Graphics, Inc. + +ALL RIGHTS RESERVED + +Permission to use, copy, modify, and distribute this software +for any purpose and without fee is hereby granted, provided +that the above copyright notice appear in all copies and that +both the copyright notice and this permission notice appear in +supporting documentation, and that the name of Silicon +Graphics, Inc. not be used in advertising or publicity +pertaining to distribution of the software without specific, +written prior permission. + +THE MATERIAL EMBODIED ON THIS SOFTWARE IS PROVIDED TO YOU +"AS-IS" AND WITHOUT WARRANTY OF ANY KIND, EXPRESS, IMPLIED OR +OTHERWISE, INCLUDING WITHOUT LIMITATION, ANY WARRANTY OF +MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE. IN NO +EVENT SHALL SILICON GRAPHICS, INC. BE LIABLE TO YOU OR ANYONE +ELSE FOR ANY DIRECT, SPECIAL, INCIDENTAL, INDIRECT OR +CONSEQUENTIAL DAMAGES OF ANY KIND, OR ANY DAMAGES WHATSOEVER, +INCLUDING WITHOUT LIMITATION, LOSS OF PROFIT, LOSS OF USE, +SAVINGS OR REVENUE, OR THE CLAIMS OF THIRD PARTIES, WHETHER OR +NOT SILICON GRAPHICS, INC. HAS BEEN ADVISED OF THE POSSIBILITY +OF SUCH LOSS, HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, +ARISING OUT OF OR IN CONNECTION WITH THE POSSESSION, USE OR +PERFORMANCE OF THIS SOFTWARE. + +US Government Users Restricted Rights + +Use, duplication, or disclosure by the Government is subject to +restrictions set forth in FAR 52.227.19(c)(2) or subparagraph +(c)(1)(ii) of the Rights in Technical Data and Computer +Software clause at DFARS 252.227-7013 and/or in similar or +successor clauses in the FAR or the DOD or NASA FAR +Supplement. Unpublished-- rights reserved under the copyright +laws of the United States. Contractor/manufacturer is Silicon +Graphics, Inc., 2011 N. Shoreline Blvd., Mountain View, CA +94039-7311. + +OpenGL(TM) is a trademark of Silicon Graphics, Inc. +*/ + +#include "glutint.h" + +/* Rim, body, lid, and bottom data must be reflected in x and + y; handle and spout data across the y axis only. */ + +static int patchdata[][16] = +{ + /* rim */ + {102, 103, 104, 105, 4, 5, 6, 7, 8, 9, 10, 11, + 12, 13, 14, 15}, + /* body */ + {12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, + 24, 25, 26, 27}, + {24, 25, 26, 27, 29, 30, 31, 32, 33, 34, 35, 36, + 37, 38, 39, 40}, + /* lid */ + {96, 96, 96, 96, 97, 98, 99, 100, 101, 101, 101, + 101, 0, 1, 2, 3,}, + {0, 1, 2, 3, 106, 107, 108, 109, 110, 111, 112, + 113, 114, 115, 116, 117}, + /* bottom */ + {118, 118, 118, 118, 124, 122, 119, 121, 123, 126, + 125, 120, 40, 39, 38, 37}, + /* handle */ + {41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, + 53, 54, 55, 56}, + {53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, + 28, 65, 66, 67}, + /* spout */ + {68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, + 80, 81, 82, 83}, + {80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, + 92, 93, 94, 95} +}; +/* *INDENT-OFF* */ + +static float cpdata[][3] = +{ + {0.2, 0, 2.7}, {0.2, -0.112, 2.7}, {0.112, -0.2, 2.7}, {0, + -0.2, 2.7}, {1.3375, 0, 2.53125}, {1.3375, -0.749, 2.53125}, + {0.749, -1.3375, 2.53125}, {0, -1.3375, 2.53125}, {1.4375, + 0, 2.53125}, {1.4375, -0.805, 2.53125}, {0.805, -1.4375, + 2.53125}, {0, -1.4375, 2.53125}, {1.5, 0, 2.4}, {1.5, -0.84, + 2.4}, {0.84, -1.5, 2.4}, {0, -1.5, 2.4}, {1.75, 0, 1.875}, + {1.75, -0.98, 1.875}, {0.98, -1.75, 1.875}, {0, -1.75, + 1.875}, {2, 0, 1.35}, {2, -1.12, 1.35}, {1.12, -2, 1.35}, + {0, -2, 1.35}, {2, 0, 0.9}, {2, -1.12, 0.9}, {1.12, -2, + 0.9}, {0, -2, 0.9}, {-2, 0, 0.9}, {2, 0, 0.45}, {2, -1.12, + 0.45}, {1.12, -2, 0.45}, {0, -2, 0.45}, {1.5, 0, 0.225}, + {1.5, -0.84, 0.225}, {0.84, -1.5, 0.225}, {0, -1.5, 0.225}, + {1.5, 0, 0.15}, {1.5, -0.84, 0.15}, {0.84, -1.5, 0.15}, {0, + -1.5, 0.15}, {-1.6, 0, 2.025}, {-1.6, -0.3, 2.025}, {-1.5, + -0.3, 2.25}, {-1.5, 0, 2.25}, {-2.3, 0, 2.025}, {-2.3, -0.3, + 2.025}, {-2.5, -0.3, 2.25}, {-2.5, 0, 2.25}, {-2.7, 0, + 2.025}, {-2.7, -0.3, 2.025}, {-3, -0.3, 2.25}, {-3, 0, + 2.25}, {-2.7, 0, 1.8}, {-2.7, -0.3, 1.8}, {-3, -0.3, 1.8}, + {-3, 0, 1.8}, {-2.7, 0, 1.575}, {-2.7, -0.3, 1.575}, {-3, + -0.3, 1.35}, {-3, 0, 1.35}, {-2.5, 0, 1.125}, {-2.5, -0.3, + 1.125}, {-2.65, -0.3, 0.9375}, {-2.65, 0, 0.9375}, {-2, + -0.3, 0.9}, {-1.9, -0.3, 0.6}, {-1.9, 0, 0.6}, {1.7, 0, + 1.425}, {1.7, -0.66, 1.425}, {1.7, -0.66, 0.6}, {1.7, 0, + 0.6}, {2.6, 0, 1.425}, {2.6, -0.66, 1.425}, {3.1, -0.66, + 0.825}, {3.1, 0, 0.825}, {2.3, 0, 2.1}, {2.3, -0.25, 2.1}, + {2.4, -0.25, 2.025}, {2.4, 0, 2.025}, {2.7, 0, 2.4}, {2.7, + -0.25, 2.4}, {3.3, -0.25, 2.4}, {3.3, 0, 2.4}, {2.8, 0, + 2.475}, {2.8, -0.25, 2.475}, {3.525, -0.25, 2.49375}, + {3.525, 0, 2.49375}, {2.9, 0, 2.475}, {2.9, -0.15, 2.475}, + {3.45, -0.15, 2.5125}, {3.45, 0, 2.5125}, {2.8, 0, 2.4}, + {2.8, -0.15, 2.4}, {3.2, -0.15, 2.4}, {3.2, 0, 2.4}, {0, 0, + 3.15}, {0.8, 0, 3.15}, {0.8, -0.45, 3.15}, {0.45, -0.8, + 3.15}, {0, -0.8, 3.15}, {0, 0, 2.85}, {1.4, 0, 2.4}, {1.4, + -0.784, 2.4}, {0.784, -1.4, 2.4}, {0, -1.4, 2.4}, {0.4, 0, + 2.55}, {0.4, -0.224, 2.55}, {0.224, -0.4, 2.55}, {0, -0.4, + 2.55}, {1.3, 0, 2.55}, {1.3, -0.728, 2.55}, {0.728, -1.3, + 2.55}, {0, -1.3, 2.55}, {1.3, 0, 2.4}, {1.3, -0.728, 2.4}, + {0.728, -1.3, 2.4}, {0, -1.3, 2.4}, {0, 0, 0}, {1.425, + -0.798, 0}, {1.5, 0, 0.075}, {1.425, 0, 0}, {0.798, -1.425, + 0}, {0, -1.5, 0.075}, {0, -1.425, 0}, {1.5, -0.84, 0.075}, + {0.84, -1.5, 0.075} +}; + +static float tex[2][2][2] = +{ + { {0, 0}, + {1, 0}}, + { {0, 1}, + {1, 1}} +}; + +/* *INDENT-ON* */ + +static void +teapot(GLint grid, GLdouble scale, GLenum type) +{ + float p[4][4][3], q[4][4][3], r[4][4][3], s[4][4][3]; + long i, j, k, l; + + glPushAttrib(GL_ENABLE_BIT | GL_EVAL_BIT); + glEnable(GL_AUTO_NORMAL); + glEnable(GL_NORMALIZE); + glEnable(GL_MAP2_VERTEX_3); + glEnable(GL_MAP2_TEXTURE_COORD_2); + glPushMatrix(); + glRotatef(270.0, 1.0, 0.0, 0.0); + glScalef(0.5 * scale, 0.5 * scale, 0.5 * scale); + glTranslatef(0.0, 0.0, -1.5); + for (i = 0; i < 10; i++) { + for (j = 0; j < 4; j++) { + for (k = 0; k < 4; k++) { + for (l = 0; l < 3; l++) { + p[j][k][l] = cpdata[patchdata[i][j * 4 + k]][l]; + q[j][k][l] = cpdata[patchdata[i][j * 4 + (3 - k)]][l]; + if (l == 1) + q[j][k][l] *= -1.0; + if (i < 6) { + r[j][k][l] = + cpdata[patchdata[i][j * 4 + (3 - k)]][l]; + if (l == 0) + r[j][k][l] *= -1.0; + s[j][k][l] = cpdata[patchdata[i][j * 4 + k]][l]; + if (l == 0) + s[j][k][l] *= -1.0; + if (l == 1) + s[j][k][l] *= -1.0; + } + } + } + } + glMap2f(GL_MAP2_TEXTURE_COORD_2, 0, 1, 2, 2, 0, 1, 4, 2, + &tex[0][0][0]); + glMap2f(GL_MAP2_VERTEX_3, 0, 1, 3, 4, 0, 1, 12, 4, + &p[0][0][0]); + glMapGrid2f(grid, 0.0, 1.0, grid, 0.0, 1.0); + glEvalMesh2(type, 0, grid, 0, grid); + glMap2f(GL_MAP2_VERTEX_3, 0, 1, 3, 4, 0, 1, 12, 4, + &q[0][0][0]); + glEvalMesh2(type, 0, grid, 0, grid); + if (i < 6) { + glMap2f(GL_MAP2_VERTEX_3, 0, 1, 3, 4, 0, 1, 12, 4, + &r[0][0][0]); + glEvalMesh2(type, 0, grid, 0, grid); + glMap2f(GL_MAP2_VERTEX_3, 0, 1, 3, 4, 0, 1, 12, 4, + &s[0][0][0]); + glEvalMesh2(type, 0, grid, 0, grid); + } + } + glPopMatrix(); + glPopAttrib(); +} + +/* CENTRY */ +void GLUTAPIENTRY +glutSolidTeapot(GLdouble scale) +{ + teapot(7, scale, GL_FILL); +} + +void GLUTAPIENTRY +glutWireTeapot(GLdouble scale) +{ + teapot(10, scale, GL_LINE); +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_tr24.cpp b/src/glut/os2/glut_tr24.cpp index 9e1234644d..a3cbf4b62b 100644 --- a/src/glut/os2/glut_tr24.cpp +++ b/src/glut/os2/glut_tr24.cpp @@ -1,2063 +1,2063 @@ - -/* GENERATED FILE -- DO NOT MODIFY */ - -#define glutBitmapTimesRoman24 XXX -#include "glutbitmap.h" -#undef glutBitmapTimesRoman24 - -/* char: 0xff */ - -static const GLubyte ch255data[] = { -0xe0,0x0,0xf0,0x0,0x18,0x0,0x8,0x0,0xc,0x0,0x4,0x0,0xe,0x0,0xe,0x0, -0x1a,0x0,0x19,0x0,0x19,0x0,0x31,0x0,0x30,0x80,0x30,0x80,0x60,0x80,0x60,0xc0, -0xf1,0xe0,0x0,0x0,0x0,0x0,0x33,0x0,0x33,0x0, -}; - -static const BitmapCharRec ch255 = {11,21,0,5,11,ch255data}; - -/* char: 0xfe */ - -static const GLubyte ch254data[] = { -0xf0,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x6e,0x0,0x73,0x80,0x61,0x80, -0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x61,0x80,0x73,0x80, -0x6e,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0xe0,0x0, -}; - -static const BitmapCharRec ch254 = {10,22,-1,5,12,ch254data}; - -/* char: 0xfd */ - -static const GLubyte ch253data[] = { -0xe0,0x0,0xf0,0x0,0x18,0x0,0x8,0x0,0xc,0x0,0x4,0x0,0xe,0x0,0xe,0x0, -0x1a,0x0,0x19,0x0,0x19,0x0,0x31,0x0,0x30,0x80,0x30,0x80,0x60,0x80,0x60,0xc0, -0xf1,0xe0,0x0,0x0,0x8,0x0,0x6,0x0,0x3,0x80,0x1,0x80, -}; - -static const BitmapCharRec ch253 = {11,22,0,5,11,ch253data}; - -/* char: 0xfc */ - -static const GLubyte ch252data[] = { -0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0,0x0,0x0,0x0,0x0,0x33,0x0,0x33,0x0, -}; - -static const BitmapCharRec ch252 = {11,16,-1,0,13,ch252data}; - -/* char: 0xfb */ - -static const GLubyte ch251data[] = { -0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0,0x0,0x0,0x21,0x0,0x12,0x0,0x1e,0x0, -0xc,0x0, -}; - -static const BitmapCharRec ch251 = {11,17,-1,0,13,ch251data}; - -/* char: 0xfa */ - -static const GLubyte ch250data[] = { -0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0,0x0,0x0,0x8,0x0,0x6,0x0,0x3,0x80, -0x1,0x80, -}; - -static const BitmapCharRec ch250 = {11,17,-1,0,13,ch250data}; - -/* char: 0xf9 */ - -static const GLubyte ch249data[] = { -0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0,0x0,0x0,0x2,0x0,0xc,0x0,0x38,0x0, -0x30,0x0, -}; - -static const BitmapCharRec ch249 = {11,17,-1,0,13,ch249data}; - -/* char: 0xf8 */ - -static const GLubyte ch248data[] = { -0xc0,0x0,0xde,0x0,0x73,0x80,0x71,0x80,0xd0,0xc0,0xd8,0xc0,0xc8,0xc0,0xcc,0xc0, -0xc4,0xc0,0xc6,0xc0,0x63,0x80,0x73,0x80,0x1e,0xc0,0x0,0xc0, -}; - -static const BitmapCharRec ch248 = {10,14,-1,1,12,ch248data}; - -/* char: 0xf7 */ - -static const GLubyte ch247data[] = { -0x6,0x0,0x6,0x0,0x0,0x0,0x0,0x0,0xff,0xf0,0xff,0xf0,0x0,0x0,0x0,0x0, -0x6,0x0,0x6,0x0, -}; - -static const BitmapCharRec ch247 = {12,10,-1,-2,14,ch247data}; - -/* char: 0xf6 */ - -static const GLubyte ch246data[] = { -0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x0,0x0,0x33,0x0,0x33,0x0, -}; - -static const BitmapCharRec ch246 = {10,16,-1,0,12,ch246data}; - -/* char: 0xf5 */ - -static const GLubyte ch245data[] = { -0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x0,0x0,0x27,0x0,0x1c,0x80, -}; - -static const BitmapCharRec ch245 = {10,16,-1,0,12,ch245data}; - -/* char: 0xf4 */ - -static const GLubyte ch244data[] = { -0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x21,0x0,0x12,0x0,0x1e,0x0, -0xc,0x0, -}; - -static const BitmapCharRec ch244 = {10,17,-1,0,12,ch244data}; - -/* char: 0xf3 */ - -static const GLubyte ch243data[] = { -0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x8,0x0,0x6,0x0,0x3,0x80, -0x1,0x80, -}; - -static const BitmapCharRec ch243 = {10,17,-1,0,12,ch243data}; - -/* char: 0xf2 */ - -static const GLubyte ch242data[] = { -0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x2,0x0,0xc,0x0,0x38,0x0, -0x30,0x0, -}; - -static const BitmapCharRec ch242 = {10,17,-1,0,12,ch242data}; - -/* char: 0xf1 */ - -static const GLubyte ch241data[] = { -0xf1,0xe0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x60,0xc0,0x71,0xc0,0x6f,0x80,0xe7,0x0,0x0,0x0,0x0,0x0,0x27,0x0,0x1c,0x80, -}; - -static const BitmapCharRec ch241 = {11,16,-1,0,13,ch241data}; - -/* char: 0xf0 */ - -static const GLubyte ch240data[] = { -0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0,0x61,0x80,0x73,0x80,0x1f,0x0,0xc6,0x0,0x3c,0x0,0x1e,0x0,0x71,0x80, -0xc0,0x0, -}; - -static const BitmapCharRec ch240 = {10,17,-1,0,12,ch240data}; - -/* char: 0xef */ - -static const GLubyte ch239data[] = { -0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x70,0x0,0x0,0xcc,0xcc, -}; - -static const BitmapCharRec ch239 = {6,16,0,0,6,ch239data}; - -/* char: 0xee */ - -static const GLubyte ch238data[] = { -0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x70,0x0,0x84,0x48,0x78, -0x30, -}; - -static const BitmapCharRec ch238 = {6,17,0,0,6,ch238data}; - -/* char: 0xed */ - -static const GLubyte ch237data[] = { -0xf0,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0xe0,0x0,0x80,0x60,0x38, -0x18, -}; - -static const BitmapCharRec ch237 = {5,17,-1,0,6,ch237data}; - -/* char: 0xec */ - -static const GLubyte ch236data[] = { -0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x70,0x0,0x8,0x30,0xe0, -0xc0, -}; - -static const BitmapCharRec ch236 = {5,17,0,0,6,ch236data}; - -/* char: 0xeb */ - -static const GLubyte ch235data[] = { -0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, -0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0,0x0,0x0,0x0,0x0,0x33,0x0,0x33,0x0, -}; - -static const BitmapCharRec ch235 = {9,16,-1,0,11,ch235data}; - -/* char: 0xea */ - -static const GLubyte ch234data[] = { -0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, -0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0,0x0,0x0,0x21,0x0,0x12,0x0,0x1e,0x0, -0xc,0x0, -}; - -static const BitmapCharRec ch234 = {9,17,-1,0,11,ch234data}; - -/* char: 0xe9 */ - -static const GLubyte ch233data[] = { -0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, -0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0,0x0,0x0,0x10,0x0,0xc,0x0,0x7,0x0, -0x3,0x0, -}; - -static const BitmapCharRec ch233 = {9,17,-1,0,11,ch233data}; - -/* char: 0xe8 */ - -static const GLubyte ch232data[] = { -0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, -0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0,0x0,0x0,0x4,0x0,0x18,0x0,0x70,0x0, -0x60,0x0, -}; - -static const BitmapCharRec ch232 = {9,17,-1,0,11,ch232data}; - -/* char: 0xe7 */ - -static const GLubyte ch231data[] = { -0x3c,0x0,0x66,0x0,0x6,0x0,0x1e,0x0,0x18,0x0,0x8,0x0,0x1e,0x0,0x7f,0x0, -0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0x41,0x80, -0x63,0x80,0x1f,0x0, -}; - -static const BitmapCharRec ch231 = {9,18,-1,6,11,ch231data}; - -/* char: 0xe6 */ - -static const GLubyte ch230data[] = { -0x70,0xf0,0xfb,0xf8,0xc7,0x84,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0xfc, -0x3,0xc,0x63,0xc,0x67,0x98,0x3c,0xf0, -}; - -static const BitmapCharRec ch230 = {14,12,-1,0,16,ch230data}; - -/* char: 0xe5 */ - -static const GLubyte ch229data[] = { -0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, -0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x1c,0x0,0x22,0x0,0x22,0x0, -0x1c,0x0, -}; - -static const BitmapCharRec ch229 = {9,17,-1,0,11,ch229data}; - -/* char: 0xe4 */ - -static const GLubyte ch228data[] = { -0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, -0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x0,0x0,0x66,0x0,0x66,0x0, -}; - -static const BitmapCharRec ch228 = {9,16,-1,0,11,ch228data}; - -/* char: 0xe3 */ - -static const GLubyte ch227data[] = { -0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, -0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x0,0x0,0x5c,0x0,0x3a,0x0, -}; - -static const BitmapCharRec ch227 = {9,16,-1,0,11,ch227data}; - -/* char: 0xe2 */ - -static const GLubyte ch226data[] = { -0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, -0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x42,0x0,0x24,0x0,0x3c,0x0, -0x18,0x0, -}; - -static const BitmapCharRec ch226 = {9,17,-1,0,11,ch226data}; - -/* char: 0xe1 */ - -static const GLubyte ch225data[] = { -0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, -0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x10,0x0,0xc,0x0,0x7,0x0, -0x3,0x0, -}; - -static const BitmapCharRec ch225 = {9,17,-1,0,11,ch225data}; - -/* char: 0xe0 */ - -static const GLubyte ch224data[] = { -0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, -0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x4,0x0,0x18,0x0,0x70,0x0, -0x60,0x0, -}; - -static const BitmapCharRec ch224 = {9,17,-1,0,11,ch224data}; - -/* char: 0xdf */ - -static const GLubyte ch223data[] = { -0xe7,0x0,0x6c,0x80,0x6c,0xc0,0x60,0xc0,0x60,0xc0,0x61,0xc0,0x61,0x80,0x63,0x80, -0x67,0x0,0x6c,0x0,0x63,0x0,0x61,0x80,0x61,0x80,0x61,0x80,0x61,0x80,0x33,0x0, -0x1e,0x0, -}; - -static const BitmapCharRec ch223 = {10,17,-1,0,12,ch223data}; - -/* char: 0xde */ - -static const GLubyte ch222data[] = { -0xfc,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x3f,0xc0,0x30,0x70,0x30,0x30,0x30,0x18, -0x30,0x18,0x30,0x18,0x30,0x30,0x30,0x70,0x3f,0xc0,0x30,0x0,0x30,0x0,0x30,0x0, -0xfc,0x0, -}; - -static const BitmapCharRec ch222 = {13,17,-1,0,15,ch222data}; - -/* char: 0xdd */ - -static const GLubyte ch221data[] = { -0x7,0xe0,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x3,0xc0, -0x3,0x40,0x6,0x60,0x6,0x20,0xc,0x30,0x1c,0x10,0x18,0x18,0x38,0x8,0x30,0xc, -0xfc,0x3f,0x0,0x0,0x1,0x0,0x0,0xc0,0x0,0x70,0x0,0x30, -}; - -static const BitmapCharRec ch221 = {16,22,0,0,16,ch221data}; - -/* char: 0xdc */ - -static const GLubyte ch220data[] = { -0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0xfc,0x1f,0x0,0x0,0x0,0x0,0x6,0x30,0x6,0x30, -}; - -static const BitmapCharRec ch220 = {16,21,-1,0,18,ch220data}; - -/* char: 0xdb */ - -static const GLubyte ch219data[] = { -0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0xfc,0x1f,0x0,0x0,0x8,0x10,0x6,0x60,0x3,0xc0,0x1,0x80, -}; - -static const BitmapCharRec ch219 = {16,22,-1,0,18,ch219data}; - -/* char: 0xda */ - -static const GLubyte ch218data[] = { -0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0xfc,0x1f,0x0,0x0,0x1,0x0,0x0,0xc0,0x0,0x70,0x0,0x30, -}; - -static const BitmapCharRec ch218 = {16,22,-1,0,18,ch218data}; - -/* char: 0xd9 */ - -static const GLubyte ch217data[] = { -0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0xfc,0x1f,0x0,0x0,0x0,0x40,0x1,0x80,0x7,0x0,0x6,0x0, -}; - -static const BitmapCharRec ch217 = {16,22,-1,0,18,ch217data}; - -/* char: 0xd8 */ - -static const GLubyte ch216data[] = { -0x20,0x0,0x27,0xe0,0x1c,0x38,0x38,0x1c,0x68,0x6,0x64,0x6,0xc2,0x3,0xc2,0x3, -0xc1,0x3,0xc1,0x3,0xc0,0x83,0xc0,0x83,0xc0,0x43,0x60,0x46,0x60,0x26,0x38,0x1c, -0x1c,0x38,0x7,0xe4,0x0,0x4, -}; - -static const BitmapCharRec ch216 = {16,19,-1,1,18,ch216data}; - -/* char: 0xd7 */ - -static const GLubyte ch215data[] = { -0x80,0x40,0xc0,0xc0,0x61,0x80,0x33,0x0,0x1e,0x0,0xc,0x0,0x1e,0x0,0x33,0x0, -0x61,0x80,0xc0,0xc0,0x80,0x40, -}; - -static const BitmapCharRec ch215 = {10,11,-2,-1,14,ch215data}; - -/* char: 0xd6 */ - -static const GLubyte ch214data[] = { -0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, -0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, -0x7,0xe0,0x0,0x0,0x0,0x0,0x6,0x60,0x6,0x60, -}; - -static const BitmapCharRec ch214 = {16,21,-1,0,18,ch214data}; - -/* char: 0xd5 */ - -static const GLubyte ch213data[] = { -0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, -0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, -0x7,0xe0,0x0,0x0,0x0,0x0,0x4,0xe0,0x3,0x90, -}; - -static const BitmapCharRec ch213 = {16,21,-1,0,18,ch213data}; - -/* char: 0xd4 */ - -static const GLubyte ch212data[] = { -0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, -0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, -0x7,0xe0,0x0,0x0,0x8,0x10,0x6,0x60,0x3,0xc0,0x1,0x80, -}; - -static const BitmapCharRec ch212 = {16,22,-1,0,18,ch212data}; - -/* char: 0xd3 */ - -static const GLubyte ch211data[] = { -0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, -0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, -0x7,0xe0,0x0,0x0,0x1,0x0,0x0,0xc0,0x0,0x70,0x0,0x30, -}; - -static const BitmapCharRec ch211 = {16,22,-1,0,18,ch211data}; - -/* char: 0xd2 */ - -static const GLubyte ch210data[] = { -0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, -0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, -0x7,0xe0,0x0,0x0,0x0,0x40,0x1,0x80,0x7,0x0,0x6,0x0, -}; - -static const BitmapCharRec ch210 = {16,22,-1,0,18,ch210data}; - -/* char: 0xd1 */ - -static const GLubyte ch209data[] = { -0xf8,0xc,0x20,0x1c,0x20,0x1c,0x20,0x34,0x20,0x64,0x20,0x64,0x20,0xc4,0x21,0x84, -0x21,0x84,0x23,0x4,0x26,0x4,0x26,0x4,0x2c,0x4,0x38,0x4,0x38,0x4,0x30,0x4, -0xf0,0x1f,0x0,0x0,0x0,0x0,0x4,0xe0,0x3,0x90, -}; - -static const BitmapCharRec ch209 = {16,21,-1,0,18,ch209data}; - -/* char: 0xd0 */ - -static const GLubyte ch208data[] = { -0x7f,0xe0,0x18,0x38,0x18,0x1c,0x18,0x6,0x18,0x6,0x18,0x3,0x18,0x3,0x18,0x3, -0xff,0x3,0x18,0x3,0x18,0x3,0x18,0x3,0x18,0x6,0x18,0x6,0x18,0x1c,0x18,0x38, -0x7f,0xe0, -}; - -static const BitmapCharRec ch208 = {16,17,0,0,17,ch208data}; - -/* char: 0xcf */ - -static const GLubyte ch207data[] = { -0xfc,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, -0xfc,0x0,0x0,0xcc,0xcc, -}; - -static const BitmapCharRec ch207 = {6,21,-1,0,8,ch207data}; - -/* char: 0xce */ - -static const GLubyte ch206data[] = { -0x7e,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18, -0x7e,0x0,0x81,0x66,0x3c,0x18, -}; - -static const BitmapCharRec ch206 = {8,22,-1,0,8,ch206data}; - -/* char: 0xcd */ - -static const GLubyte ch205data[] = { -0xfc,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, -0xfc,0x0,0x40,0x30,0x1c,0xc, -}; - -static const BitmapCharRec ch205 = {6,22,-1,0,8,ch205data}; - -/* char: 0xcc */ - -static const GLubyte ch204data[] = { -0xfc,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, -0xfc,0x0,0x8,0x30,0xe0,0xc0, -}; - -static const BitmapCharRec ch204 = {6,22,-1,0,8,ch204data}; - -/* char: 0xcb */ - -static const GLubyte ch203data[] = { -0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, -0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, -0xff,0xf0,0x0,0x0,0x0,0x0,0x19,0x80,0x19,0x80, -}; - -static const BitmapCharRec ch203 = {13,21,-1,0,15,ch203data}; - -/* char: 0xca */ - -static const GLubyte ch202data[] = { -0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, -0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, -0xff,0xf0,0x0,0x0,0x10,0x20,0xc,0xc0,0x7,0x80,0x3,0x0, -}; - -static const BitmapCharRec ch202 = {13,22,-1,0,15,ch202data}; - -/* char: 0xc9 */ - -static const GLubyte ch201data[] = { -0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, -0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, -0xff,0xf0,0x0,0x0,0x4,0x0,0x3,0x0,0x1,0xc0,0x0,0xc0, -}; - -static const BitmapCharRec ch201 = {13,22,-1,0,15,ch201data}; - -/* char: 0xc8 */ - -static const GLubyte ch200data[] = { -0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, -0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, -0xff,0xf0,0x0,0x0,0x1,0x0,0x6,0x0,0x1c,0x0,0x18,0x0, -}; - -static const BitmapCharRec ch200 = {13,22,-1,0,15,ch200data}; - -/* char: 0xc7 */ - -static const GLubyte ch199data[] = { -0x7,0x80,0xc,0xc0,0x0,0xc0,0x3,0xc0,0x3,0x0,0x1,0x0,0x7,0xe0,0x1e,0x38, -0x38,0x8,0x60,0x4,0x60,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, -0xc0,0x0,0xc0,0x0,0x60,0x4,0x60,0x4,0x38,0xc,0x1c,0x3c,0x7,0xe4, -}; - -static const BitmapCharRec ch199 = {14,23,-1,6,16,ch199data}; - -/* char: 0xc6 */ - -static const GLubyte ch198data[] = { -0xf9,0xff,0xf0,0x30,0x60,0x30,0x10,0x60,0x10,0x10,0x60,0x10,0x18,0x60,0x0,0x8, -0x60,0x0,0xf,0xe0,0x80,0xc,0x60,0x80,0x4,0x7f,0x80,0x4,0x60,0x80,0x6,0x60, -0x80,0x2,0x60,0x0,0x2,0x60,0x0,0x1,0x60,0x20,0x1,0x60,0x20,0x1,0xe0,0x60, -0x3,0xff,0xe0, -}; - -static const BitmapCharRec ch198 = {20,17,0,0,21,ch198data}; - -/* char: 0xc5 */ - -static const GLubyte ch197data[] = { -0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, -0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, -0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, -0x0,0x80,0x0,0x1,0xc0,0x0,0x2,0x20,0x0,0x2,0x20,0x0,0x1,0xc0,0x0, -}; - -static const BitmapCharRec ch197 = {17,21,0,0,17,ch197data}; - -/* char: 0xc4 */ - -static const GLubyte ch196data[] = { -0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, -0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, -0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, -0x0,0x80,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x6,0x30,0x0,0x6,0x30,0x0, -}; - -static const BitmapCharRec ch196 = {17,21,0,0,17,ch196data}; - -/* char: 0xc3 */ - -static const GLubyte ch195data[] = { -0xfc,0x1f,0x80,0x30,0x7,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, -0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, -0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, -0x0,0x80,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x4,0xe0,0x0,0x3,0x90,0x0, -}; - -static const BitmapCharRec ch195 = {17,21,0,0,17,ch195data}; - -/* char: 0xc2 */ - -static const GLubyte ch194data[] = { -0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, -0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, -0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, -0x0,0x80,0x0,0x0,0x0,0x0,0x8,0x10,0x0,0x6,0x60,0x0,0x3,0xc0,0x0,0x1, -0x80,0x0, -}; - -static const BitmapCharRec ch194 = {17,22,0,0,17,ch194data}; - -/* char: 0xc1 */ - -static const GLubyte ch193data[] = { -0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, -0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, -0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, -0x0,0x80,0x0,0x0,0x0,0x0,0x1,0x0,0x0,0x0,0xc0,0x0,0x0,0x70,0x0,0x0, -0x30,0x0, -}; - -static const BitmapCharRec ch193 = {17,22,0,0,17,ch193data}; - -/* char: 0xc0 */ - -static const GLubyte ch192data[] = { -0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, -0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, -0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, -0x0,0x80,0x0,0x0,0x0,0x0,0x0,0x20,0x0,0x0,0xc0,0x0,0x3,0x80,0x0,0x3, -0x0,0x0, -}; - -static const BitmapCharRec ch192 = {17,22,0,0,17,ch192data}; - -/* char: 0xbf */ - -static const GLubyte ch191data[] = { -0x3e,0x63,0xc1,0xc3,0xc3,0xe0,0x70,0x30,0x38,0x18,0x18,0x8,0x8,0x0,0x0,0xc, -0xc, -}; - -static const BitmapCharRec ch191 = {8,17,-1,5,11,ch191data}; - -/* char: 0xbe */ - -static const GLubyte ch190data[] = { -0x18,0x2,0x0,0x8,0x2,0x0,0xc,0x7f,0x80,0x4,0x22,0x0,0x6,0x32,0x0,0x3, -0x12,0x0,0x1,0xa,0x0,0x71,0x8e,0x0,0x88,0x86,0x0,0x8c,0xc2,0x0,0xc,0x60, -0x0,0x8,0x20,0x0,0x30,0x30,0x0,0x8,0x10,0x0,0x8c,0x18,0x0,0x4c,0xc,0x0, -0x38,0x4,0x0, -}; - -static const BitmapCharRec ch190 = {17,17,0,0,18,ch190data}; - -/* char: 0xbd */ - -static const GLubyte ch189data[] = { -0x30,0x7e,0x10,0x22,0x18,0x10,0x8,0x18,0xc,0x8,0x6,0x4,0x2,0x6,0xfb,0x46, -0x21,0x26,0x21,0x9c,0x20,0xc0,0x20,0x40,0x20,0x60,0x20,0x20,0xa0,0x30,0x60,0x18, -0x20,0x8, -}; - -static const BitmapCharRec ch189 = {15,17,-1,0,18,ch189data}; - -/* char: 0xbc */ - -static const GLubyte ch188data[] = { -0x30,0x4,0x10,0x4,0x18,0xff,0x8,0x44,0xc,0x64,0x6,0x24,0x2,0x14,0xfb,0x1c, -0x21,0xc,0x21,0x84,0x20,0xc0,0x20,0x40,0x20,0x60,0x20,0x20,0xa0,0x30,0x60,0x18, -0x20,0x8, -}; - -static const BitmapCharRec ch188 = {16,17,-1,0,18,ch188data}; - -/* char: 0xbb */ - -static const GLubyte ch187data[] = { -0x88,0x0,0xcc,0x0,0x66,0x0,0x33,0x0,0x19,0x80,0x19,0x80,0x33,0x0,0x66,0x0, -0xcc,0x0,0x88,0x0, -}; - -static const BitmapCharRec ch187 = {9,10,-2,-1,12,ch187data}; - -/* char: 0xba */ - -static const GLubyte ch186data[] = { -0xfc,0x0,0x78,0xcc,0xcc,0xcc,0xcc,0xcc,0x78, -}; - -static const BitmapCharRec ch186 = {6,9,-1,-8,8,ch186data}; - -/* char: 0xb9 */ - -static const GLubyte ch185data[] = { -0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xa0,0x60,0x20, -}; - -static const BitmapCharRec ch185 = {5,10,-1,-7,7,ch185data}; - -/* char: 0xb8 */ - -static const GLubyte ch184data[] = { -0x78,0xcc,0xc,0x3c,0x30,0x10, -}; - -static const BitmapCharRec ch184 = {6,6,-1,6,8,ch184data}; - -/* char: 0xb7 */ - -static const GLubyte ch183data[] = { -0xc0,0xc0, -}; - -static const BitmapCharRec ch183 = {2,2,-2,-6,6,ch183data}; - -/* char: 0xb6 */ - -static const GLubyte ch182data[] = { -0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0, -0x9,0x0,0x9,0x0,0x9,0x0,0x19,0x0,0x39,0x0,0x79,0x0,0x79,0x0,0xf9,0x0, -0xf9,0x0,0xf9,0x0,0x79,0x0,0x79,0x0,0x39,0x0,0x1f,0x80, -}; - -static const BitmapCharRec ch182 = {9,22,-1,5,11,ch182data}; - -/* char: 0xb5 */ - -static const GLubyte ch181data[] = { -0x40,0x0,0xe0,0x0,0xc0,0x0,0x40,0x0,0x40,0x0,0x5c,0xe0,0x7e,0xc0,0x71,0xc0, -0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0xe1,0xc0, -}; - -static const BitmapCharRec ch181 = {11,17,-1,5,13,ch181data}; - -/* char: 0xb4 */ - -static const GLubyte ch180data[] = { -0x80,0x60,0x38,0x18, -}; - -static const BitmapCharRec ch180 = {5,4,-2,-13,8,ch180data}; - -/* char: 0xb3 */ - -static const GLubyte ch179data[] = { -0x70,0x88,0x8c,0xc,0x8,0x30,0x8,0x8c,0x4c,0x38, -}; - -static const BitmapCharRec ch179 = {6,10,0,-7,7,ch179data}; - -/* char: 0xb2 */ - -static const GLubyte ch178data[] = { -0xfc,0x44,0x20,0x30,0x10,0x8,0xc,0x8c,0x4c,0x38, -}; - -static const BitmapCharRec ch178 = {6,10,0,-7,7,ch178data}; - -/* char: 0xb1 */ - -static const GLubyte ch177data[] = { -0xff,0xf0,0xff,0xf0,0x0,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, -0xff,0xf0,0xff,0xf0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, -}; - -static const BitmapCharRec ch177 = {12,15,-1,0,14,ch177data}; - -/* char: 0xb0 */ - -static const GLubyte ch176data[] = { -0x38,0x44,0x82,0x82,0x82,0x44,0x38, -}; - -static const BitmapCharRec ch176 = {7,7,-1,-10,9,ch176data}; - -/* char: 0xaf */ - -static const GLubyte ch175data[] = { -0xfc,0xfc, -}; - -static const BitmapCharRec ch175 = {6,2,-1,-14,8,ch175data}; - -/* char: 0xae */ - -static const GLubyte ch174data[] = { -0x7,0xf0,0x0,0x1c,0x1c,0x0,0x30,0x6,0x0,0x60,0x3,0x0,0x47,0x19,0x0,0xc2, -0x31,0x80,0x82,0x20,0x80,0x82,0x40,0x80,0x83,0xe0,0x80,0x82,0x30,0x80,0x82,0x10, -0x80,0xc2,0x11,0x80,0x42,0x31,0x0,0x67,0xe3,0x0,0x30,0x6,0x0,0x1c,0x1c,0x0, -0x7,0xf0,0x0, -}; - -static const BitmapCharRec ch174 = {17,17,-1,0,19,ch174data}; - -/* char: 0xad */ - -static const GLubyte ch173data[] = { -0xfe,0xfe, -}; - -static const BitmapCharRec ch173 = {7,2,-1,-5,9,ch173data}; - -/* char: 0xac */ - -static const GLubyte ch172data[] = { -0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0xff,0xf0,0xff,0xf0, -}; - -static const BitmapCharRec ch172 = {12,7,-1,-3,14,ch172data}; - -/* char: 0xab */ - -static const GLubyte ch171data[] = { -0x8,0x80,0x19,0x80,0x33,0x0,0x66,0x0,0xcc,0x0,0xcc,0x0,0x66,0x0,0x33,0x0, -0x19,0x80,0x8,0x80, -}; - -static const BitmapCharRec ch171 = {9,10,-2,-1,13,ch171data}; - -/* char: 0xaa */ - -static const GLubyte ch170data[] = { -0x7e,0x0,0x76,0xcc,0xcc,0x7c,0xc,0xcc,0x78, -}; - -static const BitmapCharRec ch170 = {7,9,0,-8,8,ch170data}; - -/* char: 0xa9 */ - -static const GLubyte ch169data[] = { -0x7,0xf0,0x0,0x1c,0x1c,0x0,0x30,0x6,0x0,0x61,0xc3,0x0,0x47,0x71,0x0,0xc4, -0x19,0x80,0x8c,0x0,0x80,0x88,0x0,0x80,0x88,0x0,0x80,0x88,0x0,0x80,0x8c,0x0, -0x80,0xc4,0x19,0x80,0x47,0x31,0x0,0x61,0xe3,0x0,0x30,0x6,0x0,0x1c,0x1c,0x0, -0x7,0xf0,0x0, -}; - -static const BitmapCharRec ch169 = {17,17,-1,0,19,ch169data}; - -/* char: 0xa8 */ - -static const GLubyte ch168data[] = { -0xcc,0xcc, -}; - -static const BitmapCharRec ch168 = {6,2,-1,-14,8,ch168data}; - -/* char: 0xa7 */ - -static const GLubyte ch167data[] = { -0x38,0x64,0x62,0x6,0xe,0x1c,0x38,0x74,0xe2,0xc3,0x83,0x87,0x4e,0x3c,0x38,0x70, -0x60,0x46,0x26,0x1c, -}; - -static const BitmapCharRec ch167 = {8,20,-2,2,12,ch167data}; - -/* char: 0xa6 */ - -static const GLubyte ch166data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0x0,0x0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0, -}; - -static const BitmapCharRec ch166 = {2,17,-2,0,6,ch166data}; - -/* char: 0xa5 */ - -static const GLubyte ch165data[] = { -0xf,0xc0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x1f,0xe0,0x3,0x0,0x1f,0xe0, -0x3,0x0,0x7,0x80,0xc,0x80,0xc,0xc0,0x18,0x40,0x18,0x60,0x30,0x20,0x70,0x30, -0xf8,0x7c, -}; - -static const BitmapCharRec ch165 = {14,17,0,0,14,ch165data}; - -/* char: 0xa4 */ - -static const GLubyte ch164data[] = { -0xc0,0x60,0xee,0xe0,0x7f,0xc0,0x31,0x80,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x31,0x80,0x7f,0xc0,0xee,0xe0,0xc0,0x60, -}; - -static const BitmapCharRec ch164 = {11,12,-1,-3,13,ch164data}; - -/* char: 0xa3 */ - -static const GLubyte ch163data[] = { -0xe7,0x80,0xbe,0xc0,0x78,0x40,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0, -0x30,0x0,0xfc,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x31,0x80,0x19,0x80, -0xf,0x0, -}; - -static const BitmapCharRec ch163 = {10,17,-1,0,12,ch163data}; - -/* char: 0xa2 */ - -static const GLubyte ch162data[] = { -0x40,0x0,0x40,0x0,0x3e,0x0,0x7f,0x0,0x70,0x80,0xd0,0x0,0xc8,0x0,0xc8,0x0, -0xc8,0x0,0xc4,0x0,0xc4,0x0,0x43,0x80,0x63,0x80,0x1f,0x0,0x1,0x0,0x1,0x0, -}; - -static const BitmapCharRec ch162 = {9,16,-1,2,12,ch162data}; - -/* char: 0xa1 */ - -static const GLubyte ch161data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0x0,0x0,0xc0, -0xc0, -}; - -static const BitmapCharRec ch161 = {2,17,-4,5,8,ch161data}; - -/* char: 0xa0 */ - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch160data[] = { 0x0 }; -static const BitmapCharRec ch160 = {1,1,0,0,6,ch160data}; -#else -static const BitmapCharRec ch160 = {0,0,0,0,6,0}; -#endif - -/* char: 0x7e '~' */ - -static const GLubyte ch126data[] = { -0x83,0x80,0xc7,0xc0,0x7c,0x60,0x38,0x20, -}; - -static const BitmapCharRec ch126 = {11,4,-1,-5,13,ch126data}; - -/* char: 0x7d '}' */ - -static const GLubyte ch125data[] = { -0xe0,0x30,0x18,0x18,0x18,0x18,0x18,0x18,0x8,0xc,0x4,0x3,0x4,0xc,0x8,0x18, -0x18,0x18,0x18,0x18,0x30,0xe0, -}; - -static const BitmapCharRec ch125 = {8,22,-1,5,10,ch125data}; - -/* char: 0x7c '|' */ - -static const GLubyte ch124data[] = { -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0, -}; - -static const BitmapCharRec ch124 = {2,17,-2,0,6,ch124data}; - -/* char: 0x7b '{' */ - -static const GLubyte ch123data[] = { -0x7,0xc,0x18,0x18,0x18,0x18,0x18,0x18,0x10,0x30,0x20,0xc0,0x20,0x30,0x10,0x18, -0x18,0x18,0x18,0x18,0xc,0x7, -}; - -static const BitmapCharRec ch123 = {8,22,-1,5,10,ch123data}; - -/* char: 0x7a 'z' */ - -static const GLubyte ch122data[] = { -0xff,0xc3,0x61,0x70,0x30,0x38,0x18,0x1c,0xe,0x86,0xc3,0xff, -}; - -static const BitmapCharRec ch122 = {8,12,-1,0,10,ch122data}; - -/* char: 0x79 'y' */ - -static const GLubyte ch121data[] = { -0xe0,0x0,0xf0,0x0,0x18,0x0,0x8,0x0,0xc,0x0,0x4,0x0,0xe,0x0,0xe,0x0, -0x1a,0x0,0x19,0x0,0x19,0x0,0x31,0x0,0x30,0x80,0x30,0x80,0x60,0x80,0x60,0xc0, -0xf1,0xe0, -}; - -static const BitmapCharRec ch121 = {11,17,0,5,11,ch121data}; - -/* char: 0x78 'x' */ - -static const GLubyte ch120data[] = { -0xf1,0xe0,0x60,0xc0,0x21,0x80,0x33,0x80,0x1b,0x0,0xe,0x0,0xc,0x0,0x1a,0x0, -0x39,0x0,0x31,0x80,0x60,0xc0,0xf1,0xe0, -}; - -static const BitmapCharRec ch120 = {11,12,-1,0,13,ch120data}; - -/* char: 0x77 'w' */ - -static const GLubyte ch119data[] = { -0x4,0x10,0x0,0xe,0x38,0x0,0xe,0x38,0x0,0x1a,0x28,0x0,0x1a,0x64,0x0,0x19, -0x64,0x0,0x31,0x64,0x0,0x30,0xc2,0x0,0x30,0xc2,0x0,0x60,0xc2,0x0,0x60,0xc3, -0x0,0xf1,0xe7,0x80, -}; - -static const BitmapCharRec ch119 = {17,12,0,0,17,ch119data}; - -/* char: 0x76 'v' */ - -static const GLubyte ch118data[] = { -0x4,0x0,0xe,0x0,0xe,0x0,0x1a,0x0,0x19,0x0,0x19,0x0,0x31,0x0,0x30,0x80, -0x30,0x80,0x60,0x80,0x60,0xc0,0xf1,0xe0, -}; - -static const BitmapCharRec ch118 = {11,12,0,0,11,ch118data}; - -/* char: 0x75 'u' */ - -static const GLubyte ch117data[] = { -0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0, -}; - -static const BitmapCharRec ch117 = {11,12,-1,0,13,ch117data}; - -/* char: 0x74 't' */ - -static const GLubyte ch116data[] = { -0x1c,0x32,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0xfe,0x70,0x30,0x10, -}; - -static const BitmapCharRec ch116 = {7,15,0,0,7,ch116data}; - -/* char: 0x73 's' */ - -static const GLubyte ch115data[] = { -0xf8,0xc6,0x83,0x3,0x7,0x1e,0x7c,0x70,0xe0,0xc2,0x66,0x3e, -}; - -static const BitmapCharRec ch115 = {8,12,-1,0,10,ch115data}; - -/* char: 0x72 'r' */ - -static const GLubyte ch114data[] = { -0xf0,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x76,0x6e,0xe6, -}; - -static const BitmapCharRec ch114 = {7,12,-1,0,8,ch114data}; - -/* char: 0x71 'q' */ - -static const GLubyte ch113data[] = { -0x3,0xc0,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1d,0x80,0x73,0x80,0x61,0x80, -0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x61,0x80,0x73,0x80, -0x1d,0x80, -}; - -static const BitmapCharRec ch113 = {10,17,-1,5,12,ch113data}; - -/* char: 0x70 'p' */ - -static const GLubyte ch112data[] = { -0xf0,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x6e,0x0,0x73,0x80,0x61,0x80, -0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x61,0x80,0x73,0x80, -0xee,0x0, -}; - -static const BitmapCharRec ch112 = {10,17,-1,5,12,ch112data}; - -/* char: 0x6f 'o' */ - -static const GLubyte ch111data[] = { -0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0, -}; - -static const BitmapCharRec ch111 = {10,12,-1,0,12,ch111data}; - -/* char: 0x6e 'n' */ - -static const GLubyte ch110data[] = { -0xf1,0xe0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x60,0xc0,0x71,0xc0,0x6f,0x80,0xe7,0x0, -}; - -static const BitmapCharRec ch110 = {11,12,-1,0,13,ch110data}; - -/* char: 0x6d 'm' */ - -static const GLubyte ch109data[] = { -0xf1,0xe3,0xc0,0x60,0xc1,0x80,0x60,0xc1,0x80,0x60,0xc1,0x80,0x60,0xc1,0x80,0x60, -0xc1,0x80,0x60,0xc1,0x80,0x60,0xc1,0x80,0x60,0xc1,0x80,0x71,0xe3,0x80,0x6f,0x9f, -0x0,0xe7,0xe,0x0, -}; - -static const BitmapCharRec ch109 = {18,12,-1,0,20,ch109data}; - -/* char: 0x6c 'l' */ - -static const GLubyte ch108data[] = { -0xf0,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60, -0xe0, -}; - -static const BitmapCharRec ch108 = {4,17,-1,0,6,ch108data}; - -/* char: 0x6b 'k' */ - -static const GLubyte ch107data[] = { -0xf3,0xe0,0x61,0xc0,0x63,0x80,0x67,0x0,0x6e,0x0,0x6c,0x0,0x78,0x0,0x68,0x0, -0x64,0x0,0x66,0x0,0x63,0x0,0x67,0xc0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0, -0xe0,0x0, -}; - -static const BitmapCharRec ch107 = {11,17,-1,0,12,ch107data}; - -/* char: 0x6a 'j' */ - -static const GLubyte ch106data[] = { -0xc0,0xe0,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, -0x70,0x0,0x0,0x0,0x30,0x30, -}; - -static const BitmapCharRec ch106 = {4,22,0,5,6,ch106data}; - -/* char: 0x69 'i' */ - -static const GLubyte ch105data[] = { -0xf0,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0xe0,0x0,0x0,0x0,0x60, -0x60, -}; - -static const BitmapCharRec ch105 = {4,17,-1,0,6,ch105data}; - -/* char: 0x68 'h' */ - -static const GLubyte ch104data[] = { -0xf1,0xe0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x60,0xc0,0x71,0xc0,0x6f,0x80,0x67,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0, -0xe0,0x0, -}; - -static const BitmapCharRec ch104 = {11,17,-1,0,13,ch104data}; - -/* char: 0x67 'g' */ - -static const GLubyte ch103data[] = { -0x3f,0x0,0xf1,0xc0,0xc0,0x60,0xc0,0x20,0x60,0x60,0x3f,0xc0,0x7f,0x0,0x60,0x0, -0x30,0x0,0x3e,0x0,0x33,0x0,0x61,0x80,0x61,0x80,0x61,0x80,0x61,0x80,0x33,0x0, -0x1f,0xc0, -}; - -static const BitmapCharRec ch103 = {11,17,-1,5,12,ch103data}; - -/* char: 0x66 'f' */ - -static const GLubyte ch102data[] = { -0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0xfe,0x30,0x30,0x30,0x16, -0xe, -}; - -static const BitmapCharRec ch102 = {7,17,0,0,7,ch102data}; - -/* char: 0x65 'e' */ - -static const GLubyte ch101data[] = { -0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, -0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0, -}; - -static const BitmapCharRec ch101 = {9,12,-1,0,11,ch101data}; - -/* char: 0x64 'd' */ - -static const GLubyte ch100data[] = { -0x1e,0xc0,0x73,0x80,0x61,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80, -0xc1,0x80,0x61,0x80,0x73,0x80,0x1d,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80, -0x3,0x80, -}; - -static const BitmapCharRec ch100 = {10,17,-1,0,12,ch100data}; - -/* char: 0x63 'c' */ - -static const GLubyte ch99data[] = { -0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, -0xc0,0x0,0x41,0x80,0x63,0x80,0x1f,0x0, -}; - -static const BitmapCharRec ch99 = {9,12,-1,0,11,ch99data}; - -/* char: 0x62 'b' */ - -static const GLubyte ch98data[] = { -0x5e,0x0,0x73,0x80,0x61,0x80,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, -0x60,0xc0,0x61,0x80,0x73,0x80,0x6e,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0, -0xe0,0x0, -}; - -static const BitmapCharRec ch98 = {10,17,-1,0,12,ch98data}; - -/* char: 0x61 'a' */ - -static const GLubyte ch97data[] = { -0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, -0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0, -}; - -static const BitmapCharRec ch97 = {9,12,-1,0,11,ch97data}; - -/* char: 0x60 '`' */ - -static const GLubyte ch96data[] = { -0x60,0xe0,0x80,0xc0,0x60, -}; - -static const BitmapCharRec ch96 = {3,5,-2,-12,7,ch96data}; - -/* char: 0x5f '_' */ - -static const GLubyte ch95data[] = { -0xff,0xf8,0xff,0xf8, -}; - -static const BitmapCharRec ch95 = {13,2,0,5,13,ch95data}; - -/* char: 0x5e '^' */ - -static const GLubyte ch94data[] = { -0x80,0x80,0xc1,0x80,0x41,0x0,0x63,0x0,0x22,0x0,0x36,0x0,0x14,0x0,0x1c,0x0, -0x8,0x0, -}; - -static const BitmapCharRec ch94 = {9,9,-1,-8,11,ch94data}; - -/* char: 0x5d ']' */ - -static const GLubyte ch93data[] = { -0xf8,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18, -0x18,0x18,0x18,0x18,0xf8, -}; - -static const BitmapCharRec ch93 = {5,21,-1,4,8,ch93data}; - -/* char: 0x5c '\' */ - -static const GLubyte ch92data[] = { -0x6,0x6,0x4,0xc,0xc,0x8,0x18,0x18,0x10,0x30,0x30,0x20,0x60,0x60,0x40,0xc0, -0xc0, -}; - -static const BitmapCharRec ch92 = {7,17,0,0,7,ch92data}; - -/* char: 0x5b '[' */ - -static const GLubyte ch91data[] = { -0xf8,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0,0xc0,0xc0,0xf8, -}; - -static const BitmapCharRec ch91 = {5,21,-2,4,8,ch91data}; - -/* char: 0x5a 'Z' */ - -static const GLubyte ch90data[] = { -0xff,0xf8,0xe0,0x18,0x70,0x8,0x30,0x8,0x38,0x0,0x18,0x0,0x1c,0x0,0xe,0x0, -0x6,0x0,0x7,0x0,0x3,0x0,0x3,0x80,0x1,0xc0,0x80,0xc0,0x80,0xe0,0xc0,0x70, -0xff,0xf0, -}; - -static const BitmapCharRec ch90 = {13,17,-1,0,15,ch90data}; - -/* char: 0x59 'Y' */ - -static const GLubyte ch89data[] = { -0x7,0xe0,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x3,0xc0, -0x3,0x40,0x6,0x60,0x6,0x20,0xc,0x30,0x1c,0x10,0x18,0x18,0x38,0x8,0x30,0xc, -0xfc,0x3f, -}; - -static const BitmapCharRec ch89 = {16,17,0,0,16,ch89data}; - -/* char: 0x58 'X' */ - -static const GLubyte ch88data[] = { -0xfc,0xf,0xc0,0x30,0x3,0x80,0x18,0x7,0x0,0x8,0xe,0x0,0x4,0xc,0x0,0x6, -0x18,0x0,0x2,0x38,0x0,0x1,0x70,0x0,0x0,0xe0,0x0,0x0,0xc0,0x0,0x1,0xc0, -0x0,0x3,0xa0,0x0,0x3,0x10,0x0,0x6,0x8,0x0,0xe,0xc,0x0,0x1c,0x6,0x0, -0x7e,0xf,0x80, -}; - -static const BitmapCharRec ch88 = {18,17,0,0,18,ch88data}; - -/* char: 0x57 'W' */ - -static const GLubyte ch87data[] = { -0x1,0x83,0x0,0x1,0x83,0x0,0x1,0x83,0x80,0x3,0x87,0x80,0x3,0x46,0x80,0x3, -0x46,0xc0,0x6,0x46,0x40,0x6,0x4c,0x40,0x6,0x4c,0x60,0xc,0x2c,0x60,0xc,0x2c, -0x20,0x18,0x2c,0x20,0x18,0x18,0x30,0x18,0x18,0x10,0x30,0x18,0x10,0x30,0x18,0x18, -0xfc,0x7e,0x7e, -}; - -static const BitmapCharRec ch87 = {23,17,0,0,23,ch87data}; - -/* char: 0x56 'V' */ - -static const GLubyte ch86data[] = { -0x1,0x80,0x0,0x1,0x80,0x0,0x1,0x80,0x0,0x3,0xc0,0x0,0x3,0x40,0x0,0x3, -0x60,0x0,0x6,0x20,0x0,0x6,0x20,0x0,0x6,0x30,0x0,0xc,0x10,0x0,0xc,0x18, -0x0,0x18,0x8,0x0,0x18,0x8,0x0,0x18,0xc,0x0,0x30,0x4,0x0,0x30,0x6,0x0, -0xfc,0x1f,0x80, -}; - -static const BitmapCharRec ch86 = {17,17,0,0,17,ch86data}; - -/* char: 0x55 'U' */ - -static const GLubyte ch85data[] = { -0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, -0xfc,0x1f, -}; - -static const BitmapCharRec ch85 = {16,17,-1,0,18,ch85data}; - -/* char: 0x54 'T' */ - -static const GLubyte ch84data[] = { -0xf,0xc0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0, -0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x83,0x4,0x83,0x4,0xc3,0xc, -0xff,0xfc, -}; - -static const BitmapCharRec ch84 = {14,17,-1,0,16,ch84data}; - -/* char: 0x53 'S' */ - -static const GLubyte ch83data[] = { -0x9e,0x0,0xf1,0x80,0xc0,0xc0,0x80,0x60,0x80,0x60,0x0,0x60,0x0,0xe0,0x3,0xc0, -0xf,0x80,0x1e,0x0,0x78,0x0,0xe0,0x0,0xc0,0x40,0xc0,0x40,0xc0,0xc0,0x63,0xc0, -0x1e,0x40, -}; - -static const BitmapCharRec ch83 = {11,17,-1,0,13,ch83data}; - -/* char: 0x52 'R' */ - -static const GLubyte ch82data[] = { -0xfc,0x1e,0x30,0x1c,0x30,0x38,0x30,0x70,0x30,0x60,0x30,0xc0,0x31,0xc0,0x33,0x80, -0x3f,0xc0,0x30,0x70,0x30,0x30,0x30,0x38,0x30,0x18,0x30,0x38,0x30,0x30,0x30,0x70, -0xff,0xc0, -}; - -static const BitmapCharRec ch82 = {15,17,-1,0,16,ch82data}; - -/* char: 0x51 'Q' */ - -static const GLubyte ch81data[] = { -0x0,0xf,0x0,0x38,0x0,0x70,0x0,0xe0,0x1,0xc0,0x7,0xe0,0x1c,0x38,0x38,0x1c, -0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3, -0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38,0x7,0xe0, -}; - -static const BitmapCharRec ch81 = {16,22,-1,5,18,ch81data}; - -/* char: 0x50 'P' */ - -static const GLubyte ch80data[] = { -0xfc,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0, -0x3f,0xc0,0x30,0x70,0x30,0x30,0x30,0x18,0x30,0x18,0x30,0x18,0x30,0x30,0x30,0x70, -0xff,0xc0, -}; - -static const BitmapCharRec ch80 = {13,17,-1,0,15,ch80data}; - -/* char: 0x4f 'O' */ - -static const GLubyte ch79data[] = { -0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, -0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, -0x7,0xe0, -}; - -static const BitmapCharRec ch79 = {16,17,-1,0,18,ch79data}; - -/* char: 0x4e 'N' */ - -static const GLubyte ch78data[] = { -0xf8,0xc,0x20,0x1c,0x20,0x1c,0x20,0x34,0x20,0x64,0x20,0x64,0x20,0xc4,0x21,0x84, -0x21,0x84,0x23,0x4,0x26,0x4,0x26,0x4,0x2c,0x4,0x38,0x4,0x38,0x4,0x30,0x4, -0xf0,0x1f, -}; - -static const BitmapCharRec ch78 = {16,17,-1,0,18,ch78data}; - -/* char: 0x4d 'M' */ - -static const GLubyte ch77data[] = { -0xf8,0x21,0xf8,0x20,0x60,0x60,0x20,0x60,0x60,0x20,0xd0,0x60,0x20,0xd0,0x60,0x21, -0x88,0x60,0x21,0x88,0x60,0x23,0x8,0x60,0x23,0x4,0x60,0x26,0x4,0x60,0x26,0x2, -0x60,0x2c,0x2,0x60,0x2c,0x2,0x60,0x38,0x1,0x60,0x38,0x1,0x60,0x30,0x0,0xe0, -0xf0,0x0,0xf8, -}; - -static const BitmapCharRec ch77 = {21,17,-1,0,22,ch77data}; - -/* char: 0x4c 'L' */ - -static const GLubyte ch76data[] = { -0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0, -0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0, -0xfc,0x0, -}; - -static const BitmapCharRec ch76 = {13,17,-1,0,14,ch76data}; - -/* char: 0x4b 'K' */ - -static const GLubyte ch75data[] = { -0xfc,0x1f,0x30,0xe,0x30,0x1c,0x30,0x38,0x30,0x70,0x30,0xe0,0x31,0xc0,0x33,0x80, -0x3f,0x0,0x3e,0x0,0x33,0x0,0x31,0x80,0x30,0xc0,0x30,0x60,0x30,0x30,0x30,0x18, -0xfc,0x7e, -}; - -static const BitmapCharRec ch75 = {16,17,-1,0,17,ch75data}; - -/* char: 0x4a 'J' */ - -static const GLubyte ch74data[] = { -0x78,0x0,0xcc,0x0,0xc6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, -0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, -0x1f,0x80, -}; - -static const BitmapCharRec ch74 = {9,17,-1,0,11,ch74data}; - -/* char: 0x49 'I' */ - -static const GLubyte ch73data[] = { -0xfc,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, -0xfc, -}; - -static const BitmapCharRec ch73 = {6,17,-1,0,8,ch73data}; - -/* char: 0x48 'H' */ - -static const GLubyte ch72data[] = { -0xfc,0x1f,0x80,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30, -0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x3f,0xfe,0x0,0x30,0x6,0x0,0x30,0x6, -0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0, -0xfc,0x1f,0x80, -}; - -static const BitmapCharRec ch72 = {17,17,-1,0,19,ch72data}; - -/* char: 0x47 'G' */ - -static const GLubyte ch71data[] = { -0x7,0xe0,0x1e,0x38,0x38,0x1c,0x60,0xc,0x60,0xc,0xc0,0xc,0xc0,0xc,0xc0,0x3f, -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0x60,0x4,0x60,0x4,0x38,0xc,0x1c,0x3c, -0x7,0xe4, -}; - -static const BitmapCharRec ch71 = {16,17,-1,0,18,ch71data}; - -/* char: 0x46 'F' */ - -static const GLubyte ch70data[] = { -0xfc,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x20,0x30,0x20, -0x3f,0xe0,0x30,0x20,0x30,0x20,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, -0xff,0xf0, -}; - -static const BitmapCharRec ch70 = {12,17,-1,0,14,ch70data}; - -/* char: 0x45 'E' */ - -static const GLubyte ch69data[] = { -0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, -0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, -0xff,0xf0, -}; - -static const BitmapCharRec ch69 = {13,17,-1,0,15,ch69data}; - -/* char: 0x44 'D' */ - -static const GLubyte ch68data[] = { -0xff,0xc0,0x30,0x70,0x30,0x38,0x30,0xc,0x30,0xc,0x30,0x6,0x30,0x6,0x30,0x6, -0x30,0x6,0x30,0x6,0x30,0x6,0x30,0x6,0x30,0xc,0x30,0xc,0x30,0x38,0x30,0x70, -0xff,0xc0, -}; - -static const BitmapCharRec ch68 = {15,17,-1,0,17,ch68data}; - -/* char: 0x43 'C' */ - -static const GLubyte ch67data[] = { -0x7,0xe0,0x1e,0x38,0x38,0x8,0x60,0x4,0x60,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, -0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0x60,0x4,0x60,0x4,0x38,0xc,0x1c,0x3c, -0x7,0xe4, -}; - -static const BitmapCharRec ch67 = {14,17,-1,0,16,ch67data}; - -/* char: 0x42 'B' */ - -static const GLubyte ch66data[] = { -0xff,0xe0,0x30,0x78,0x30,0x18,0x30,0xc,0x30,0xc,0x30,0xc,0x30,0x18,0x30,0x38, -0x3f,0xe0,0x30,0x40,0x30,0x30,0x30,0x18,0x30,0x18,0x30,0x18,0x30,0x30,0x30,0x70, -0xff,0xc0, -}; - -static const BitmapCharRec ch66 = {14,17,-1,0,16,ch66data}; - -/* char: 0x41 'A' */ - -static const GLubyte ch65data[] = { -0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, -0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, -0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, -0x0,0x80,0x0, -}; - -static const BitmapCharRec ch65 = {17,17,0,0,17,ch65data}; - -/* char: 0x40 '@' */ - -static const GLubyte ch64data[] = { -0x3,0xf0,0x0,0xe,0xc,0x0,0x18,0x0,0x0,0x30,0x0,0x0,0x61,0xde,0x0,0x63, -0x7b,0x0,0xc6,0x39,0x80,0xc6,0x18,0x80,0xc6,0x18,0xc0,0xc6,0x18,0x40,0xc6,0xc, -0x40,0xc3,0xc,0x40,0xc3,0x8c,0x40,0xe1,0xfc,0x40,0x60,0xec,0xc0,0x70,0x0,0x80, -0x38,0x1,0x80,0x1c,0x3,0x0,0xf,0xe,0x0,0x3,0xf8,0x0, -}; - -static const BitmapCharRec ch64 = {18,20,-2,3,22,ch64data}; - -/* char: 0x3f '?' */ - -static const GLubyte ch63data[] = { -0x30,0x30,0x0,0x0,0x10,0x10,0x10,0x18,0x18,0xc,0xe,0x7,0xc3,0xc3,0x83,0xc6, -0x7c, -}; - -static const BitmapCharRec ch63 = {8,17,-2,0,11,ch63data}; - -/* char: 0x3e '>' */ - -static const GLubyte ch62data[] = { -0xc0,0x0,0x70,0x0,0x1c,0x0,0x7,0x0,0x1,0xc0,0x0,0x60,0x1,0xc0,0x7,0x0, -0x1c,0x0,0x70,0x0,0xc0,0x0, -}; - -static const BitmapCharRec ch62 = {11,11,-1,-1,13,ch62data}; - -/* char: 0x3d '=' */ - -static const GLubyte ch61data[] = { -0xff,0xf0,0xff,0xf0,0x0,0x0,0x0,0x0,0xff,0xf0,0xff,0xf0, -}; - -static const BitmapCharRec ch61 = {12,6,-1,-4,14,ch61data}; - -/* char: 0x3c '<' */ - -static const GLubyte ch60data[] = { -0x0,0x60,0x1,0xc0,0x7,0x0,0x1c,0x0,0x70,0x0,0xc0,0x0,0x70,0x0,0x1c,0x0, -0x7,0x0,0x1,0xc0,0x0,0x60, -}; - -static const BitmapCharRec ch60 = {11,11,-1,-1,13,ch60data}; - -/* char: 0x3b ';' */ - -static const GLubyte ch59data[] = { -0xc0,0x60,0x20,0xe0,0xc0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0xc0,0xc0, -}; - -static const BitmapCharRec ch59 = {3,14,-2,3,7,ch59data}; - -/* char: 0x3a ':' */ - -static const GLubyte ch58data[] = { -0xc0,0xc0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0xc0,0xc0, -}; - -static const BitmapCharRec ch58 = {2,11,-2,0,6,ch58data}; - -/* char: 0x39 '9' */ - -static const GLubyte ch57data[] = { -0xf0,0x0,0x1c,0x0,0x6,0x0,0x3,0x0,0x3,0x80,0x1,0x80,0x1d,0x80,0x73,0xc0, -0x61,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc1,0xc0,0x61,0x80,0x77,0x80, -0x1e,0x0, -}; - -static const BitmapCharRec ch57 = {10,17,-1,0,12,ch57data}; - -/* char: 0x38 '8' */ - -static const GLubyte ch56data[] = { -0x1e,0x0,0x73,0x80,0xe1,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x41,0xc0,0x61,0x80, -0x37,0x0,0x1e,0x0,0x1e,0x0,0x33,0x0,0x61,0x80,0x61,0x80,0x61,0x80,0x33,0x0, -0x1e,0x0, -}; - -static const BitmapCharRec ch56 = {10,17,-1,0,12,ch56data}; - -/* char: 0x37 '7' */ - -static const GLubyte ch55data[] = { -0x18,0x0,0x18,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0x4,0x0,0x6,0x0,0x6,0x0, -0x2,0x0,0x3,0x0,0x3,0x0,0x1,0x0,0x1,0x80,0x81,0x80,0xc0,0xc0,0xff,0xc0, -0x7f,0xc0, -}; - -static const BitmapCharRec ch55 = {10,17,-1,0,12,ch55data}; - -/* char: 0x36 '6' */ - -static const GLubyte ch54data[] = { -0x1e,0x0,0x7b,0x80,0x61,0x80,0xe0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc1,0x80,0xf3,0x80,0xee,0x0,0x60,0x0,0x70,0x0,0x30,0x0,0x18,0x0,0xe,0x0, -0x3,0xc0, -}; - -static const BitmapCharRec ch54 = {10,17,-1,0,12,ch54data}; - -/* char: 0x35 '5' */ - -static const GLubyte ch53data[] = { -0x7e,0x0,0xe3,0x80,0xc1,0x80,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x1,0xc0, -0x3,0x80,0xf,0x80,0x7e,0x0,0x78,0x0,0x60,0x0,0x20,0x0,0x20,0x0,0x1f,0x80, -0x1f,0xc0, -}; - -static const BitmapCharRec ch53 = {10,17,-1,0,12,ch53data}; - -/* char: 0x34 '4' */ - -static const GLubyte ch52data[] = { -0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0xff,0xc0,0xff,0xc0,0xc3,0x0,0x43,0x0, -0x63,0x0,0x23,0x0,0x33,0x0,0x13,0x0,0x1b,0x0,0xb,0x0,0x7,0x0,0x7,0x0, -0x3,0x0, -}; - -static const BitmapCharRec ch52 = {10,17,-1,0,12,ch52data}; - -/* char: 0x33 '3' */ - -static const GLubyte ch51data[] = { -0x78,0x0,0xe6,0x0,0xc3,0x0,0x1,0x0,0x1,0x80,0x1,0x80,0x1,0x80,0x3,0x80, -0x7,0x0,0x1e,0x0,0xc,0x0,0x6,0x0,0x83,0x0,0x83,0x0,0x47,0x0,0x7e,0x0, -0x1c,0x0, -}; - -static const BitmapCharRec ch51 = {9,17,-1,0,12,ch51data}; - -/* char: 0x32 '2' */ - -static const GLubyte ch50data[] = { -0xff,0x80,0xff,0xc0,0x60,0x40,0x30,0x0,0x18,0x0,0xc,0x0,0x4,0x0,0x6,0x0, -0x3,0x0,0x3,0x0,0x1,0x80,0x1,0x80,0x81,0x80,0x81,0x80,0x43,0x80,0x7f,0x0, -0x1c,0x0, -}; - -static const BitmapCharRec ch50 = {10,17,-1,0,12,ch50data}; - -/* char: 0x31 '1' */ - -static const GLubyte ch49data[] = { -0xff,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x78,0x18, -0x8, -}; - -static const BitmapCharRec ch49 = {8,17,-2,0,12,ch49data}; - -/* char: 0x30 '0' */ - -static const GLubyte ch48data[] = { -0x1e,0x0,0x33,0x0,0x61,0x80,0x61,0x80,0xe1,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x61,0x80,0x61,0x80,0x33,0x0, -0x1e,0x0, -}; - -static const BitmapCharRec ch48 = {10,17,-1,0,12,ch48data}; - -/* char: 0x2f '/' */ - -static const GLubyte ch47data[] = { -0xc0,0xc0,0xc0,0x60,0x60,0x20,0x30,0x30,0x10,0x18,0x18,0x8,0xc,0xc,0x4,0x6, -0x6,0x3,0x3,0x3, -}; - -static const BitmapCharRec ch47 = {8,20,1,3,7,ch47data}; - -/* char: 0x2e '.' */ - -static const GLubyte ch46data[] = { -0xc0,0xc0, -}; - -static const BitmapCharRec ch46 = {2,2,-2,0,6,ch46data}; - -/* char: 0x2d '-' */ - -static const GLubyte ch45data[] = { -0xff,0xf0,0xff,0xf0, -}; - -static const BitmapCharRec ch45 = {12,2,-1,-6,14,ch45data}; - -/* char: 0x2c ',' */ - -static const GLubyte ch44data[] = { -0xc0,0x60,0x20,0xe0,0xc0, -}; - -static const BitmapCharRec ch44 = {3,5,-2,3,7,ch44data}; - -/* char: 0x2b '+' */ - -static const GLubyte ch43data[] = { -0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0xff,0xf0,0xff,0xf0,0x6,0x0, -0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, -}; - -static const BitmapCharRec ch43 = {12,12,-1,-1,14,ch43data}; - -/* char: 0x2a '*' */ - -static const GLubyte ch42data[] = { -0x8,0x0,0x1c,0x0,0xc9,0x80,0xeb,0x80,0x1c,0x0,0xeb,0x80,0xc9,0x80,0x1c,0x0, -0x8,0x0, -}; - -static const BitmapCharRec ch42 = {9,9,-2,-8,12,ch42data}; - -/* char: 0x29 ')' */ - -static const GLubyte ch41data[] = { -0x80,0x40,0x20,0x30,0x10,0x18,0x18,0xc,0xc,0xc,0xc,0xc,0xc,0xc,0xc,0x18, -0x18,0x10,0x30,0x20,0x40,0x80, -}; - -static const BitmapCharRec ch41 = {6,22,-1,5,8,ch41data}; - -/* char: 0x28 '(' */ - -static const GLubyte ch40data[] = { -0x4,0x8,0x10,0x30,0x20,0x60,0x60,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x60, -0x60,0x20,0x30,0x10,0x8,0x4, -}; - -static const BitmapCharRec ch40 = {6,22,-1,5,8,ch40data}; - -/* char: 0x27 ''' */ - -static const GLubyte ch39data[] = { -0xc0,0x60,0x20,0xe0,0xc0, -}; - -static const BitmapCharRec ch39 = {3,5,-3,-12,8,ch39data}; - -/* char: 0x26 '&' */ - -static const GLubyte ch38data[] = { -0x3c,0x3c,0x7f,0x7e,0xe1,0xe1,0xc0,0xc0,0xc1,0xc0,0xc1,0xa0,0x63,0x20,0x37,0x10, -0x1e,0x18,0xe,0x3e,0xf,0x0,0x1d,0x80,0x18,0xc0,0x18,0x40,0x18,0x40,0xc,0xc0, -0x7,0x80, -}; - -static const BitmapCharRec ch38 = {16,17,-1,0,18,ch38data}; - -/* char: 0x25 '%' */ - -static const GLubyte ch37data[] = { -0x30,0x3c,0x0,0x18,0x72,0x0,0xc,0x61,0x0,0x4,0x60,0x80,0x6,0x60,0x80,0x3, -0x30,0x80,0x1,0x19,0x80,0x1,0x8f,0x0,0x78,0xc0,0x0,0xe4,0x40,0x0,0xc2,0x60, -0x0,0xc1,0x30,0x0,0xc1,0x10,0x0,0x61,0x18,0x0,0x33,0xfc,0x0,0x1e,0xc,0x0, -}; - -static const BitmapCharRec ch37 = {17,16,-1,0,19,ch37data}; - -/* char: 0x24 '$' */ - -static const GLubyte ch36data[] = { -0x4,0x0,0x4,0x0,0x3f,0x0,0xe5,0xc0,0xc4,0xc0,0x84,0x60,0x84,0x60,0x4,0x60, -0x4,0xe0,0x7,0xc0,0x7,0x80,0x1e,0x0,0x3c,0x0,0x74,0x0,0x64,0x0,0x64,0x20, -0x64,0x60,0x34,0xe0,0x1f,0x80,0x4,0x0,0x4,0x0, -}; - -static const BitmapCharRec ch36 = {11,21,0,2,12,ch36data}; - -/* char: 0x23 '#' */ - -static const GLubyte ch35data[] = { -0x22,0x0,0x22,0x0,0x22,0x0,0x22,0x0,0x22,0x0,0xff,0xc0,0xff,0xc0,0x11,0x0, -0x11,0x0,0x11,0x0,0x7f,0xe0,0x7f,0xe0,0x8,0x80,0x8,0x80,0x8,0x80,0x8,0x80, -0x8,0x80, -}; - -static const BitmapCharRec ch35 = {11,17,-1,0,13,ch35data}; - -/* char: 0x22 '"' */ - -static const GLubyte ch34data[] = { -0x88,0xcc,0xcc,0xcc,0xcc, -}; - -static const BitmapCharRec ch34 = {6,5,-1,-12,10,ch34data}; - -/* char: 0x21 '!' */ - -static const GLubyte ch33data[] = { -0xc0,0xc0,0x0,0x0,0x0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, -0xc0, -}; - -static const BitmapCharRec ch33 = {2,17,-3,0,8,ch33data}; - -/* char: 0x20 ' ' */ - -#ifdef _WIN32 -/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with - a height or width of zero does not advance the raster position - as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ -static const GLubyte ch32data[] = { 0x0 }; -static const BitmapCharRec ch32 = {1,1,0,0,6,ch32data}; -#else -static const BitmapCharRec ch32 = {0,0,0,0,6,0}; -#endif - -static const BitmapCharRec * const chars[] = { -&ch32, -&ch33, -&ch34, -&ch35, -&ch36, -&ch37, -&ch38, -&ch39, -&ch40, -&ch41, -&ch42, -&ch43, -&ch44, -&ch45, -&ch46, -&ch47, -&ch48, -&ch49, -&ch50, -&ch51, -&ch52, -&ch53, -&ch54, -&ch55, -&ch56, -&ch57, -&ch58, -&ch59, -&ch60, -&ch61, -&ch62, -&ch63, -&ch64, -&ch65, -&ch66, -&ch67, -&ch68, -&ch69, -&ch70, -&ch71, -&ch72, -&ch73, -&ch74, -&ch75, -&ch76, -&ch77, -&ch78, -&ch79, -&ch80, -&ch81, -&ch82, -&ch83, -&ch84, -&ch85, -&ch86, -&ch87, -&ch88, -&ch89, -&ch90, -&ch91, -&ch92, -&ch93, -&ch94, -&ch95, -&ch96, -&ch97, -&ch98, -&ch99, -&ch100, -&ch101, -&ch102, -&ch103, -&ch104, -&ch105, -&ch106, -&ch107, -&ch108, -&ch109, -&ch110, -&ch111, -&ch112, -&ch113, -&ch114, -&ch115, -&ch116, -&ch117, -&ch118, -&ch119, -&ch120, -&ch121, -&ch122, -&ch123, -&ch124, -&ch125, -&ch126, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -0, -&ch160, -&ch161, -&ch162, -&ch163, -&ch164, -&ch165, -&ch166, -&ch167, -&ch168, -&ch169, -&ch170, -&ch171, -&ch172, -&ch173, -&ch174, -&ch175, -&ch176, -&ch177, -&ch178, -&ch179, -&ch180, -&ch181, -&ch182, -&ch183, -&ch184, -&ch185, -&ch186, -&ch187, -&ch188, -&ch189, -&ch190, -&ch191, -&ch192, -&ch193, -&ch194, -&ch195, -&ch196, -&ch197, -&ch198, -&ch199, -&ch200, -&ch201, -&ch202, -&ch203, -&ch204, -&ch205, -&ch206, -&ch207, -&ch208, -&ch209, -&ch210, -&ch211, -&ch212, -&ch213, -&ch214, -&ch215, -&ch216, -&ch217, -&ch218, -&ch219, -&ch220, -&ch221, -&ch222, -&ch223, -&ch224, -&ch225, -&ch226, -&ch227, -&ch228, -&ch229, -&ch230, -&ch231, -&ch232, -&ch233, -&ch234, -&ch235, -&ch236, -&ch237, -&ch238, -&ch239, -&ch240, -&ch241, -&ch242, -&ch243, -&ch244, -&ch245, -&ch246, -&ch247, -&ch248, -&ch249, -&ch250, -&ch251, -&ch252, -&ch253, -&ch254, -&ch255, -}; - -#if !defined(__IBMCPP__) -const -#endif -BitmapFontRec glutBitmapTimesRoman24 = { -"-adobe-times-medium-r-normal--24-240-75-75-p-124-iso8859-1", -224, -32, -chars -}; - + +/* GENERATED FILE -- DO NOT MODIFY */ + +#define glutBitmapTimesRoman24 XXX +#include "glutbitmap.h" +#undef glutBitmapTimesRoman24 + +/* char: 0xff */ + +static const GLubyte ch255data[] = { +0xe0,0x0,0xf0,0x0,0x18,0x0,0x8,0x0,0xc,0x0,0x4,0x0,0xe,0x0,0xe,0x0, +0x1a,0x0,0x19,0x0,0x19,0x0,0x31,0x0,0x30,0x80,0x30,0x80,0x60,0x80,0x60,0xc0, +0xf1,0xe0,0x0,0x0,0x0,0x0,0x33,0x0,0x33,0x0, +}; + +static const BitmapCharRec ch255 = {11,21,0,5,11,ch255data}; + +/* char: 0xfe */ + +static const GLubyte ch254data[] = { +0xf0,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x6e,0x0,0x73,0x80,0x61,0x80, +0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x61,0x80,0x73,0x80, +0x6e,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0xe0,0x0, +}; + +static const BitmapCharRec ch254 = {10,22,-1,5,12,ch254data}; + +/* char: 0xfd */ + +static const GLubyte ch253data[] = { +0xe0,0x0,0xf0,0x0,0x18,0x0,0x8,0x0,0xc,0x0,0x4,0x0,0xe,0x0,0xe,0x0, +0x1a,0x0,0x19,0x0,0x19,0x0,0x31,0x0,0x30,0x80,0x30,0x80,0x60,0x80,0x60,0xc0, +0xf1,0xe0,0x0,0x0,0x8,0x0,0x6,0x0,0x3,0x80,0x1,0x80, +}; + +static const BitmapCharRec ch253 = {11,22,0,5,11,ch253data}; + +/* char: 0xfc */ + +static const GLubyte ch252data[] = { +0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0,0x0,0x0,0x0,0x0,0x33,0x0,0x33,0x0, +}; + +static const BitmapCharRec ch252 = {11,16,-1,0,13,ch252data}; + +/* char: 0xfb */ + +static const GLubyte ch251data[] = { +0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0,0x0,0x0,0x21,0x0,0x12,0x0,0x1e,0x0, +0xc,0x0, +}; + +static const BitmapCharRec ch251 = {11,17,-1,0,13,ch251data}; + +/* char: 0xfa */ + +static const GLubyte ch250data[] = { +0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0,0x0,0x0,0x8,0x0,0x6,0x0,0x3,0x80, +0x1,0x80, +}; + +static const BitmapCharRec ch250 = {11,17,-1,0,13,ch250data}; + +/* char: 0xf9 */ + +static const GLubyte ch249data[] = { +0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0,0x0,0x0,0x2,0x0,0xc,0x0,0x38,0x0, +0x30,0x0, +}; + +static const BitmapCharRec ch249 = {11,17,-1,0,13,ch249data}; + +/* char: 0xf8 */ + +static const GLubyte ch248data[] = { +0xc0,0x0,0xde,0x0,0x73,0x80,0x71,0x80,0xd0,0xc0,0xd8,0xc0,0xc8,0xc0,0xcc,0xc0, +0xc4,0xc0,0xc6,0xc0,0x63,0x80,0x73,0x80,0x1e,0xc0,0x0,0xc0, +}; + +static const BitmapCharRec ch248 = {10,14,-1,1,12,ch248data}; + +/* char: 0xf7 */ + +static const GLubyte ch247data[] = { +0x6,0x0,0x6,0x0,0x0,0x0,0x0,0x0,0xff,0xf0,0xff,0xf0,0x0,0x0,0x0,0x0, +0x6,0x0,0x6,0x0, +}; + +static const BitmapCharRec ch247 = {12,10,-1,-2,14,ch247data}; + +/* char: 0xf6 */ + +static const GLubyte ch246data[] = { +0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x0,0x0,0x33,0x0,0x33,0x0, +}; + +static const BitmapCharRec ch246 = {10,16,-1,0,12,ch246data}; + +/* char: 0xf5 */ + +static const GLubyte ch245data[] = { +0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x0,0x0,0x27,0x0,0x1c,0x80, +}; + +static const BitmapCharRec ch245 = {10,16,-1,0,12,ch245data}; + +/* char: 0xf4 */ + +static const GLubyte ch244data[] = { +0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x21,0x0,0x12,0x0,0x1e,0x0, +0xc,0x0, +}; + +static const BitmapCharRec ch244 = {10,17,-1,0,12,ch244data}; + +/* char: 0xf3 */ + +static const GLubyte ch243data[] = { +0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x8,0x0,0x6,0x0,0x3,0x80, +0x1,0x80, +}; + +static const BitmapCharRec ch243 = {10,17,-1,0,12,ch243data}; + +/* char: 0xf2 */ + +static const GLubyte ch242data[] = { +0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0,0x0,0x0,0x2,0x0,0xc,0x0,0x38,0x0, +0x30,0x0, +}; + +static const BitmapCharRec ch242 = {10,17,-1,0,12,ch242data}; + +/* char: 0xf1 */ + +static const GLubyte ch241data[] = { +0xf1,0xe0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x60,0xc0,0x71,0xc0,0x6f,0x80,0xe7,0x0,0x0,0x0,0x0,0x0,0x27,0x0,0x1c,0x80, +}; + +static const BitmapCharRec ch241 = {11,16,-1,0,13,ch241data}; + +/* char: 0xf0 */ + +static const GLubyte ch240data[] = { +0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0,0x61,0x80,0x73,0x80,0x1f,0x0,0xc6,0x0,0x3c,0x0,0x1e,0x0,0x71,0x80, +0xc0,0x0, +}; + +static const BitmapCharRec ch240 = {10,17,-1,0,12,ch240data}; + +/* char: 0xef */ + +static const GLubyte ch239data[] = { +0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x70,0x0,0x0,0xcc,0xcc, +}; + +static const BitmapCharRec ch239 = {6,16,0,0,6,ch239data}; + +/* char: 0xee */ + +static const GLubyte ch238data[] = { +0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x70,0x0,0x84,0x48,0x78, +0x30, +}; + +static const BitmapCharRec ch238 = {6,17,0,0,6,ch238data}; + +/* char: 0xed */ + +static const GLubyte ch237data[] = { +0xf0,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0xe0,0x0,0x80,0x60,0x38, +0x18, +}; + +static const BitmapCharRec ch237 = {5,17,-1,0,6,ch237data}; + +/* char: 0xec */ + +static const GLubyte ch236data[] = { +0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x70,0x0,0x8,0x30,0xe0, +0xc0, +}; + +static const BitmapCharRec ch236 = {5,17,0,0,6,ch236data}; + +/* char: 0xeb */ + +static const GLubyte ch235data[] = { +0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, +0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0,0x0,0x0,0x0,0x0,0x33,0x0,0x33,0x0, +}; + +static const BitmapCharRec ch235 = {9,16,-1,0,11,ch235data}; + +/* char: 0xea */ + +static const GLubyte ch234data[] = { +0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, +0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0,0x0,0x0,0x21,0x0,0x12,0x0,0x1e,0x0, +0xc,0x0, +}; + +static const BitmapCharRec ch234 = {9,17,-1,0,11,ch234data}; + +/* char: 0xe9 */ + +static const GLubyte ch233data[] = { +0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, +0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0,0x0,0x0,0x10,0x0,0xc,0x0,0x7,0x0, +0x3,0x0, +}; + +static const BitmapCharRec ch233 = {9,17,-1,0,11,ch233data}; + +/* char: 0xe8 */ + +static const GLubyte ch232data[] = { +0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, +0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0,0x0,0x0,0x4,0x0,0x18,0x0,0x70,0x0, +0x60,0x0, +}; + +static const BitmapCharRec ch232 = {9,17,-1,0,11,ch232data}; + +/* char: 0xe7 */ + +static const GLubyte ch231data[] = { +0x3c,0x0,0x66,0x0,0x6,0x0,0x1e,0x0,0x18,0x0,0x8,0x0,0x1e,0x0,0x7f,0x0, +0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0x41,0x80, +0x63,0x80,0x1f,0x0, +}; + +static const BitmapCharRec ch231 = {9,18,-1,6,11,ch231data}; + +/* char: 0xe6 */ + +static const GLubyte ch230data[] = { +0x70,0xf0,0xfb,0xf8,0xc7,0x84,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0xfc, +0x3,0xc,0x63,0xc,0x67,0x98,0x3c,0xf0, +}; + +static const BitmapCharRec ch230 = {14,12,-1,0,16,ch230data}; + +/* char: 0xe5 */ + +static const GLubyte ch229data[] = { +0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, +0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x1c,0x0,0x22,0x0,0x22,0x0, +0x1c,0x0, +}; + +static const BitmapCharRec ch229 = {9,17,-1,0,11,ch229data}; + +/* char: 0xe4 */ + +static const GLubyte ch228data[] = { +0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, +0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x0,0x0,0x66,0x0,0x66,0x0, +}; + +static const BitmapCharRec ch228 = {9,16,-1,0,11,ch228data}; + +/* char: 0xe3 */ + +static const GLubyte ch227data[] = { +0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, +0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x0,0x0,0x5c,0x0,0x3a,0x0, +}; + +static const BitmapCharRec ch227 = {9,16,-1,0,11,ch227data}; + +/* char: 0xe2 */ + +static const GLubyte ch226data[] = { +0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, +0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x42,0x0,0x24,0x0,0x3c,0x0, +0x18,0x0, +}; + +static const BitmapCharRec ch226 = {9,17,-1,0,11,ch226data}; + +/* char: 0xe1 */ + +static const GLubyte ch225data[] = { +0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, +0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x10,0x0,0xc,0x0,0x7,0x0, +0x3,0x0, +}; + +static const BitmapCharRec ch225 = {9,17,-1,0,11,ch225data}; + +/* char: 0xe0 */ + +static const GLubyte ch224data[] = { +0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, +0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0,0x0,0x0,0x4,0x0,0x18,0x0,0x70,0x0, +0x60,0x0, +}; + +static const BitmapCharRec ch224 = {9,17,-1,0,11,ch224data}; + +/* char: 0xdf */ + +static const GLubyte ch223data[] = { +0xe7,0x0,0x6c,0x80,0x6c,0xc0,0x60,0xc0,0x60,0xc0,0x61,0xc0,0x61,0x80,0x63,0x80, +0x67,0x0,0x6c,0x0,0x63,0x0,0x61,0x80,0x61,0x80,0x61,0x80,0x61,0x80,0x33,0x0, +0x1e,0x0, +}; + +static const BitmapCharRec ch223 = {10,17,-1,0,12,ch223data}; + +/* char: 0xde */ + +static const GLubyte ch222data[] = { +0xfc,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x3f,0xc0,0x30,0x70,0x30,0x30,0x30,0x18, +0x30,0x18,0x30,0x18,0x30,0x30,0x30,0x70,0x3f,0xc0,0x30,0x0,0x30,0x0,0x30,0x0, +0xfc,0x0, +}; + +static const BitmapCharRec ch222 = {13,17,-1,0,15,ch222data}; + +/* char: 0xdd */ + +static const GLubyte ch221data[] = { +0x7,0xe0,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x3,0xc0, +0x3,0x40,0x6,0x60,0x6,0x20,0xc,0x30,0x1c,0x10,0x18,0x18,0x38,0x8,0x30,0xc, +0xfc,0x3f,0x0,0x0,0x1,0x0,0x0,0xc0,0x0,0x70,0x0,0x30, +}; + +static const BitmapCharRec ch221 = {16,22,0,0,16,ch221data}; + +/* char: 0xdc */ + +static const GLubyte ch220data[] = { +0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0xfc,0x1f,0x0,0x0,0x0,0x0,0x6,0x30,0x6,0x30, +}; + +static const BitmapCharRec ch220 = {16,21,-1,0,18,ch220data}; + +/* char: 0xdb */ + +static const GLubyte ch219data[] = { +0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0xfc,0x1f,0x0,0x0,0x8,0x10,0x6,0x60,0x3,0xc0,0x1,0x80, +}; + +static const BitmapCharRec ch219 = {16,22,-1,0,18,ch219data}; + +/* char: 0xda */ + +static const GLubyte ch218data[] = { +0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0xfc,0x1f,0x0,0x0,0x1,0x0,0x0,0xc0,0x0,0x70,0x0,0x30, +}; + +static const BitmapCharRec ch218 = {16,22,-1,0,18,ch218data}; + +/* char: 0xd9 */ + +static const GLubyte ch217data[] = { +0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0xfc,0x1f,0x0,0x0,0x0,0x40,0x1,0x80,0x7,0x0,0x6,0x0, +}; + +static const BitmapCharRec ch217 = {16,22,-1,0,18,ch217data}; + +/* char: 0xd8 */ + +static const GLubyte ch216data[] = { +0x20,0x0,0x27,0xe0,0x1c,0x38,0x38,0x1c,0x68,0x6,0x64,0x6,0xc2,0x3,0xc2,0x3, +0xc1,0x3,0xc1,0x3,0xc0,0x83,0xc0,0x83,0xc0,0x43,0x60,0x46,0x60,0x26,0x38,0x1c, +0x1c,0x38,0x7,0xe4,0x0,0x4, +}; + +static const BitmapCharRec ch216 = {16,19,-1,1,18,ch216data}; + +/* char: 0xd7 */ + +static const GLubyte ch215data[] = { +0x80,0x40,0xc0,0xc0,0x61,0x80,0x33,0x0,0x1e,0x0,0xc,0x0,0x1e,0x0,0x33,0x0, +0x61,0x80,0xc0,0xc0,0x80,0x40, +}; + +static const BitmapCharRec ch215 = {10,11,-2,-1,14,ch215data}; + +/* char: 0xd6 */ + +static const GLubyte ch214data[] = { +0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, +0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, +0x7,0xe0,0x0,0x0,0x0,0x0,0x6,0x60,0x6,0x60, +}; + +static const BitmapCharRec ch214 = {16,21,-1,0,18,ch214data}; + +/* char: 0xd5 */ + +static const GLubyte ch213data[] = { +0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, +0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, +0x7,0xe0,0x0,0x0,0x0,0x0,0x4,0xe0,0x3,0x90, +}; + +static const BitmapCharRec ch213 = {16,21,-1,0,18,ch213data}; + +/* char: 0xd4 */ + +static const GLubyte ch212data[] = { +0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, +0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, +0x7,0xe0,0x0,0x0,0x8,0x10,0x6,0x60,0x3,0xc0,0x1,0x80, +}; + +static const BitmapCharRec ch212 = {16,22,-1,0,18,ch212data}; + +/* char: 0xd3 */ + +static const GLubyte ch211data[] = { +0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, +0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, +0x7,0xe0,0x0,0x0,0x1,0x0,0x0,0xc0,0x0,0x70,0x0,0x30, +}; + +static const BitmapCharRec ch211 = {16,22,-1,0,18,ch211data}; + +/* char: 0xd2 */ + +static const GLubyte ch210data[] = { +0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, +0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, +0x7,0xe0,0x0,0x0,0x0,0x40,0x1,0x80,0x7,0x0,0x6,0x0, +}; + +static const BitmapCharRec ch210 = {16,22,-1,0,18,ch210data}; + +/* char: 0xd1 */ + +static const GLubyte ch209data[] = { +0xf8,0xc,0x20,0x1c,0x20,0x1c,0x20,0x34,0x20,0x64,0x20,0x64,0x20,0xc4,0x21,0x84, +0x21,0x84,0x23,0x4,0x26,0x4,0x26,0x4,0x2c,0x4,0x38,0x4,0x38,0x4,0x30,0x4, +0xf0,0x1f,0x0,0x0,0x0,0x0,0x4,0xe0,0x3,0x90, +}; + +static const BitmapCharRec ch209 = {16,21,-1,0,18,ch209data}; + +/* char: 0xd0 */ + +static const GLubyte ch208data[] = { +0x7f,0xe0,0x18,0x38,0x18,0x1c,0x18,0x6,0x18,0x6,0x18,0x3,0x18,0x3,0x18,0x3, +0xff,0x3,0x18,0x3,0x18,0x3,0x18,0x3,0x18,0x6,0x18,0x6,0x18,0x1c,0x18,0x38, +0x7f,0xe0, +}; + +static const BitmapCharRec ch208 = {16,17,0,0,17,ch208data}; + +/* char: 0xcf */ + +static const GLubyte ch207data[] = { +0xfc,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, +0xfc,0x0,0x0,0xcc,0xcc, +}; + +static const BitmapCharRec ch207 = {6,21,-1,0,8,ch207data}; + +/* char: 0xce */ + +static const GLubyte ch206data[] = { +0x7e,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18, +0x7e,0x0,0x81,0x66,0x3c,0x18, +}; + +static const BitmapCharRec ch206 = {8,22,-1,0,8,ch206data}; + +/* char: 0xcd */ + +static const GLubyte ch205data[] = { +0xfc,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, +0xfc,0x0,0x40,0x30,0x1c,0xc, +}; + +static const BitmapCharRec ch205 = {6,22,-1,0,8,ch205data}; + +/* char: 0xcc */ + +static const GLubyte ch204data[] = { +0xfc,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, +0xfc,0x0,0x8,0x30,0xe0,0xc0, +}; + +static const BitmapCharRec ch204 = {6,22,-1,0,8,ch204data}; + +/* char: 0xcb */ + +static const GLubyte ch203data[] = { +0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, +0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, +0xff,0xf0,0x0,0x0,0x0,0x0,0x19,0x80,0x19,0x80, +}; + +static const BitmapCharRec ch203 = {13,21,-1,0,15,ch203data}; + +/* char: 0xca */ + +static const GLubyte ch202data[] = { +0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, +0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, +0xff,0xf0,0x0,0x0,0x10,0x20,0xc,0xc0,0x7,0x80,0x3,0x0, +}; + +static const BitmapCharRec ch202 = {13,22,-1,0,15,ch202data}; + +/* char: 0xc9 */ + +static const GLubyte ch201data[] = { +0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, +0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, +0xff,0xf0,0x0,0x0,0x4,0x0,0x3,0x0,0x1,0xc0,0x0,0xc0, +}; + +static const BitmapCharRec ch201 = {13,22,-1,0,15,ch201data}; + +/* char: 0xc8 */ + +static const GLubyte ch200data[] = { +0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, +0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, +0xff,0xf0,0x0,0x0,0x1,0x0,0x6,0x0,0x1c,0x0,0x18,0x0, +}; + +static const BitmapCharRec ch200 = {13,22,-1,0,15,ch200data}; + +/* char: 0xc7 */ + +static const GLubyte ch199data[] = { +0x7,0x80,0xc,0xc0,0x0,0xc0,0x3,0xc0,0x3,0x0,0x1,0x0,0x7,0xe0,0x1e,0x38, +0x38,0x8,0x60,0x4,0x60,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, +0xc0,0x0,0xc0,0x0,0x60,0x4,0x60,0x4,0x38,0xc,0x1c,0x3c,0x7,0xe4, +}; + +static const BitmapCharRec ch199 = {14,23,-1,6,16,ch199data}; + +/* char: 0xc6 */ + +static const GLubyte ch198data[] = { +0xf9,0xff,0xf0,0x30,0x60,0x30,0x10,0x60,0x10,0x10,0x60,0x10,0x18,0x60,0x0,0x8, +0x60,0x0,0xf,0xe0,0x80,0xc,0x60,0x80,0x4,0x7f,0x80,0x4,0x60,0x80,0x6,0x60, +0x80,0x2,0x60,0x0,0x2,0x60,0x0,0x1,0x60,0x20,0x1,0x60,0x20,0x1,0xe0,0x60, +0x3,0xff,0xe0, +}; + +static const BitmapCharRec ch198 = {20,17,0,0,21,ch198data}; + +/* char: 0xc5 */ + +static const GLubyte ch197data[] = { +0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, +0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, +0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, +0x0,0x80,0x0,0x1,0xc0,0x0,0x2,0x20,0x0,0x2,0x20,0x0,0x1,0xc0,0x0, +}; + +static const BitmapCharRec ch197 = {17,21,0,0,17,ch197data}; + +/* char: 0xc4 */ + +static const GLubyte ch196data[] = { +0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, +0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, +0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, +0x0,0x80,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x6,0x30,0x0,0x6,0x30,0x0, +}; + +static const BitmapCharRec ch196 = {17,21,0,0,17,ch196data}; + +/* char: 0xc3 */ + +static const GLubyte ch195data[] = { +0xfc,0x1f,0x80,0x30,0x7,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, +0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, +0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, +0x0,0x80,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0x4,0xe0,0x0,0x3,0x90,0x0, +}; + +static const BitmapCharRec ch195 = {17,21,0,0,17,ch195data}; + +/* char: 0xc2 */ + +static const GLubyte ch194data[] = { +0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, +0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, +0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, +0x0,0x80,0x0,0x0,0x0,0x0,0x8,0x10,0x0,0x6,0x60,0x0,0x3,0xc0,0x0,0x1, +0x80,0x0, +}; + +static const BitmapCharRec ch194 = {17,22,0,0,17,ch194data}; + +/* char: 0xc1 */ + +static const GLubyte ch193data[] = { +0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, +0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, +0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, +0x0,0x80,0x0,0x0,0x0,0x0,0x1,0x0,0x0,0x0,0xc0,0x0,0x0,0x70,0x0,0x0, +0x30,0x0, +}; + +static const BitmapCharRec ch193 = {17,22,0,0,17,ch193data}; + +/* char: 0xc0 */ + +static const GLubyte ch192data[] = { +0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, +0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, +0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, +0x0,0x80,0x0,0x0,0x0,0x0,0x0,0x20,0x0,0x0,0xc0,0x0,0x3,0x80,0x0,0x3, +0x0,0x0, +}; + +static const BitmapCharRec ch192 = {17,22,0,0,17,ch192data}; + +/* char: 0xbf */ + +static const GLubyte ch191data[] = { +0x3e,0x63,0xc1,0xc3,0xc3,0xe0,0x70,0x30,0x38,0x18,0x18,0x8,0x8,0x0,0x0,0xc, +0xc, +}; + +static const BitmapCharRec ch191 = {8,17,-1,5,11,ch191data}; + +/* char: 0xbe */ + +static const GLubyte ch190data[] = { +0x18,0x2,0x0,0x8,0x2,0x0,0xc,0x7f,0x80,0x4,0x22,0x0,0x6,0x32,0x0,0x3, +0x12,0x0,0x1,0xa,0x0,0x71,0x8e,0x0,0x88,0x86,0x0,0x8c,0xc2,0x0,0xc,0x60, +0x0,0x8,0x20,0x0,0x30,0x30,0x0,0x8,0x10,0x0,0x8c,0x18,0x0,0x4c,0xc,0x0, +0x38,0x4,0x0, +}; + +static const BitmapCharRec ch190 = {17,17,0,0,18,ch190data}; + +/* char: 0xbd */ + +static const GLubyte ch189data[] = { +0x30,0x7e,0x10,0x22,0x18,0x10,0x8,0x18,0xc,0x8,0x6,0x4,0x2,0x6,0xfb,0x46, +0x21,0x26,0x21,0x9c,0x20,0xc0,0x20,0x40,0x20,0x60,0x20,0x20,0xa0,0x30,0x60,0x18, +0x20,0x8, +}; + +static const BitmapCharRec ch189 = {15,17,-1,0,18,ch189data}; + +/* char: 0xbc */ + +static const GLubyte ch188data[] = { +0x30,0x4,0x10,0x4,0x18,0xff,0x8,0x44,0xc,0x64,0x6,0x24,0x2,0x14,0xfb,0x1c, +0x21,0xc,0x21,0x84,0x20,0xc0,0x20,0x40,0x20,0x60,0x20,0x20,0xa0,0x30,0x60,0x18, +0x20,0x8, +}; + +static const BitmapCharRec ch188 = {16,17,-1,0,18,ch188data}; + +/* char: 0xbb */ + +static const GLubyte ch187data[] = { +0x88,0x0,0xcc,0x0,0x66,0x0,0x33,0x0,0x19,0x80,0x19,0x80,0x33,0x0,0x66,0x0, +0xcc,0x0,0x88,0x0, +}; + +static const BitmapCharRec ch187 = {9,10,-2,-1,12,ch187data}; + +/* char: 0xba */ + +static const GLubyte ch186data[] = { +0xfc,0x0,0x78,0xcc,0xcc,0xcc,0xcc,0xcc,0x78, +}; + +static const BitmapCharRec ch186 = {6,9,-1,-8,8,ch186data}; + +/* char: 0xb9 */ + +static const GLubyte ch185data[] = { +0xf8,0x20,0x20,0x20,0x20,0x20,0x20,0xa0,0x60,0x20, +}; + +static const BitmapCharRec ch185 = {5,10,-1,-7,7,ch185data}; + +/* char: 0xb8 */ + +static const GLubyte ch184data[] = { +0x78,0xcc,0xc,0x3c,0x30,0x10, +}; + +static const BitmapCharRec ch184 = {6,6,-1,6,8,ch184data}; + +/* char: 0xb7 */ + +static const GLubyte ch183data[] = { +0xc0,0xc0, +}; + +static const BitmapCharRec ch183 = {2,2,-2,-6,6,ch183data}; + +/* char: 0xb6 */ + +static const GLubyte ch182data[] = { +0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0,0x9,0x0, +0x9,0x0,0x9,0x0,0x9,0x0,0x19,0x0,0x39,0x0,0x79,0x0,0x79,0x0,0xf9,0x0, +0xf9,0x0,0xf9,0x0,0x79,0x0,0x79,0x0,0x39,0x0,0x1f,0x80, +}; + +static const BitmapCharRec ch182 = {9,22,-1,5,11,ch182data}; + +/* char: 0xb5 */ + +static const GLubyte ch181data[] = { +0x40,0x0,0xe0,0x0,0xc0,0x0,0x40,0x0,0x40,0x0,0x5c,0xe0,0x7e,0xc0,0x71,0xc0, +0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0xe1,0xc0, +}; + +static const BitmapCharRec ch181 = {11,17,-1,5,13,ch181data}; + +/* char: 0xb4 */ + +static const GLubyte ch180data[] = { +0x80,0x60,0x38,0x18, +}; + +static const BitmapCharRec ch180 = {5,4,-2,-13,8,ch180data}; + +/* char: 0xb3 */ + +static const GLubyte ch179data[] = { +0x70,0x88,0x8c,0xc,0x8,0x30,0x8,0x8c,0x4c,0x38, +}; + +static const BitmapCharRec ch179 = {6,10,0,-7,7,ch179data}; + +/* char: 0xb2 */ + +static const GLubyte ch178data[] = { +0xfc,0x44,0x20,0x30,0x10,0x8,0xc,0x8c,0x4c,0x38, +}; + +static const BitmapCharRec ch178 = {6,10,0,-7,7,ch178data}; + +/* char: 0xb1 */ + +static const GLubyte ch177data[] = { +0xff,0xf0,0xff,0xf0,0x0,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, +0xff,0xf0,0xff,0xf0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, +}; + +static const BitmapCharRec ch177 = {12,15,-1,0,14,ch177data}; + +/* char: 0xb0 */ + +static const GLubyte ch176data[] = { +0x38,0x44,0x82,0x82,0x82,0x44,0x38, +}; + +static const BitmapCharRec ch176 = {7,7,-1,-10,9,ch176data}; + +/* char: 0xaf */ + +static const GLubyte ch175data[] = { +0xfc,0xfc, +}; + +static const BitmapCharRec ch175 = {6,2,-1,-14,8,ch175data}; + +/* char: 0xae */ + +static const GLubyte ch174data[] = { +0x7,0xf0,0x0,0x1c,0x1c,0x0,0x30,0x6,0x0,0x60,0x3,0x0,0x47,0x19,0x0,0xc2, +0x31,0x80,0x82,0x20,0x80,0x82,0x40,0x80,0x83,0xe0,0x80,0x82,0x30,0x80,0x82,0x10, +0x80,0xc2,0x11,0x80,0x42,0x31,0x0,0x67,0xe3,0x0,0x30,0x6,0x0,0x1c,0x1c,0x0, +0x7,0xf0,0x0, +}; + +static const BitmapCharRec ch174 = {17,17,-1,0,19,ch174data}; + +/* char: 0xad */ + +static const GLubyte ch173data[] = { +0xfe,0xfe, +}; + +static const BitmapCharRec ch173 = {7,2,-1,-5,9,ch173data}; + +/* char: 0xac */ + +static const GLubyte ch172data[] = { +0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0xff,0xf0,0xff,0xf0, +}; + +static const BitmapCharRec ch172 = {12,7,-1,-3,14,ch172data}; + +/* char: 0xab */ + +static const GLubyte ch171data[] = { +0x8,0x80,0x19,0x80,0x33,0x0,0x66,0x0,0xcc,0x0,0xcc,0x0,0x66,0x0,0x33,0x0, +0x19,0x80,0x8,0x80, +}; + +static const BitmapCharRec ch171 = {9,10,-2,-1,13,ch171data}; + +/* char: 0xaa */ + +static const GLubyte ch170data[] = { +0x7e,0x0,0x76,0xcc,0xcc,0x7c,0xc,0xcc,0x78, +}; + +static const BitmapCharRec ch170 = {7,9,0,-8,8,ch170data}; + +/* char: 0xa9 */ + +static const GLubyte ch169data[] = { +0x7,0xf0,0x0,0x1c,0x1c,0x0,0x30,0x6,0x0,0x61,0xc3,0x0,0x47,0x71,0x0,0xc4, +0x19,0x80,0x8c,0x0,0x80,0x88,0x0,0x80,0x88,0x0,0x80,0x88,0x0,0x80,0x8c,0x0, +0x80,0xc4,0x19,0x80,0x47,0x31,0x0,0x61,0xe3,0x0,0x30,0x6,0x0,0x1c,0x1c,0x0, +0x7,0xf0,0x0, +}; + +static const BitmapCharRec ch169 = {17,17,-1,0,19,ch169data}; + +/* char: 0xa8 */ + +static const GLubyte ch168data[] = { +0xcc,0xcc, +}; + +static const BitmapCharRec ch168 = {6,2,-1,-14,8,ch168data}; + +/* char: 0xa7 */ + +static const GLubyte ch167data[] = { +0x38,0x64,0x62,0x6,0xe,0x1c,0x38,0x74,0xe2,0xc3,0x83,0x87,0x4e,0x3c,0x38,0x70, +0x60,0x46,0x26,0x1c, +}; + +static const BitmapCharRec ch167 = {8,20,-2,2,12,ch167data}; + +/* char: 0xa6 */ + +static const GLubyte ch166data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0x0,0x0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0, +}; + +static const BitmapCharRec ch166 = {2,17,-2,0,6,ch166data}; + +/* char: 0xa5 */ + +static const GLubyte ch165data[] = { +0xf,0xc0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x1f,0xe0,0x3,0x0,0x1f,0xe0, +0x3,0x0,0x7,0x80,0xc,0x80,0xc,0xc0,0x18,0x40,0x18,0x60,0x30,0x20,0x70,0x30, +0xf8,0x7c, +}; + +static const BitmapCharRec ch165 = {14,17,0,0,14,ch165data}; + +/* char: 0xa4 */ + +static const GLubyte ch164data[] = { +0xc0,0x60,0xee,0xe0,0x7f,0xc0,0x31,0x80,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x31,0x80,0x7f,0xc0,0xee,0xe0,0xc0,0x60, +}; + +static const BitmapCharRec ch164 = {11,12,-1,-3,13,ch164data}; + +/* char: 0xa3 */ + +static const GLubyte ch163data[] = { +0xe7,0x80,0xbe,0xc0,0x78,0x40,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0, +0x30,0x0,0xfc,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x31,0x80,0x19,0x80, +0xf,0x0, +}; + +static const BitmapCharRec ch163 = {10,17,-1,0,12,ch163data}; + +/* char: 0xa2 */ + +static const GLubyte ch162data[] = { +0x40,0x0,0x40,0x0,0x3e,0x0,0x7f,0x0,0x70,0x80,0xd0,0x0,0xc8,0x0,0xc8,0x0, +0xc8,0x0,0xc4,0x0,0xc4,0x0,0x43,0x80,0x63,0x80,0x1f,0x0,0x1,0x0,0x1,0x0, +}; + +static const BitmapCharRec ch162 = {9,16,-1,2,12,ch162data}; + +/* char: 0xa1 */ + +static const GLubyte ch161data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x0,0x0,0x0,0xc0, +0xc0, +}; + +static const BitmapCharRec ch161 = {2,17,-4,5,8,ch161data}; + +/* char: 0xa0 */ + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch160data[] = { 0x0 }; +static const BitmapCharRec ch160 = {1,1,0,0,6,ch160data}; +#else +static const BitmapCharRec ch160 = {0,0,0,0,6,0}; +#endif + +/* char: 0x7e '~' */ + +static const GLubyte ch126data[] = { +0x83,0x80,0xc7,0xc0,0x7c,0x60,0x38,0x20, +}; + +static const BitmapCharRec ch126 = {11,4,-1,-5,13,ch126data}; + +/* char: 0x7d '}' */ + +static const GLubyte ch125data[] = { +0xe0,0x30,0x18,0x18,0x18,0x18,0x18,0x18,0x8,0xc,0x4,0x3,0x4,0xc,0x8,0x18, +0x18,0x18,0x18,0x18,0x30,0xe0, +}; + +static const BitmapCharRec ch125 = {8,22,-1,5,10,ch125data}; + +/* char: 0x7c '|' */ + +static const GLubyte ch124data[] = { +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0, +}; + +static const BitmapCharRec ch124 = {2,17,-2,0,6,ch124data}; + +/* char: 0x7b '{' */ + +static const GLubyte ch123data[] = { +0x7,0xc,0x18,0x18,0x18,0x18,0x18,0x18,0x10,0x30,0x20,0xc0,0x20,0x30,0x10,0x18, +0x18,0x18,0x18,0x18,0xc,0x7, +}; + +static const BitmapCharRec ch123 = {8,22,-1,5,10,ch123data}; + +/* char: 0x7a 'z' */ + +static const GLubyte ch122data[] = { +0xff,0xc3,0x61,0x70,0x30,0x38,0x18,0x1c,0xe,0x86,0xc3,0xff, +}; + +static const BitmapCharRec ch122 = {8,12,-1,0,10,ch122data}; + +/* char: 0x79 'y' */ + +static const GLubyte ch121data[] = { +0xe0,0x0,0xf0,0x0,0x18,0x0,0x8,0x0,0xc,0x0,0x4,0x0,0xe,0x0,0xe,0x0, +0x1a,0x0,0x19,0x0,0x19,0x0,0x31,0x0,0x30,0x80,0x30,0x80,0x60,0x80,0x60,0xc0, +0xf1,0xe0, +}; + +static const BitmapCharRec ch121 = {11,17,0,5,11,ch121data}; + +/* char: 0x78 'x' */ + +static const GLubyte ch120data[] = { +0xf1,0xe0,0x60,0xc0,0x21,0x80,0x33,0x80,0x1b,0x0,0xe,0x0,0xc,0x0,0x1a,0x0, +0x39,0x0,0x31,0x80,0x60,0xc0,0xf1,0xe0, +}; + +static const BitmapCharRec ch120 = {11,12,-1,0,13,ch120data}; + +/* char: 0x77 'w' */ + +static const GLubyte ch119data[] = { +0x4,0x10,0x0,0xe,0x38,0x0,0xe,0x38,0x0,0x1a,0x28,0x0,0x1a,0x64,0x0,0x19, +0x64,0x0,0x31,0x64,0x0,0x30,0xc2,0x0,0x30,0xc2,0x0,0x60,0xc2,0x0,0x60,0xc3, +0x0,0xf1,0xe7,0x80, +}; + +static const BitmapCharRec ch119 = {17,12,0,0,17,ch119data}; + +/* char: 0x76 'v' */ + +static const GLubyte ch118data[] = { +0x4,0x0,0xe,0x0,0xe,0x0,0x1a,0x0,0x19,0x0,0x19,0x0,0x31,0x0,0x30,0x80, +0x30,0x80,0x60,0x80,0x60,0xc0,0xf1,0xe0, +}; + +static const BitmapCharRec ch118 = {11,12,0,0,11,ch118data}; + +/* char: 0x75 'u' */ + +static const GLubyte ch117data[] = { +0x1c,0xe0,0x3e,0xc0,0x71,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x60,0xc0,0x60,0xc0,0x60,0xc0,0xe1,0xc0, +}; + +static const BitmapCharRec ch117 = {11,12,-1,0,13,ch117data}; + +/* char: 0x74 't' */ + +static const GLubyte ch116data[] = { +0x1c,0x32,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0xfe,0x70,0x30,0x10, +}; + +static const BitmapCharRec ch116 = {7,15,0,0,7,ch116data}; + +/* char: 0x73 's' */ + +static const GLubyte ch115data[] = { +0xf8,0xc6,0x83,0x3,0x7,0x1e,0x7c,0x70,0xe0,0xc2,0x66,0x3e, +}; + +static const BitmapCharRec ch115 = {8,12,-1,0,10,ch115data}; + +/* char: 0x72 'r' */ + +static const GLubyte ch114data[] = { +0xf0,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x76,0x6e,0xe6, +}; + +static const BitmapCharRec ch114 = {7,12,-1,0,8,ch114data}; + +/* char: 0x71 'q' */ + +static const GLubyte ch113data[] = { +0x3,0xc0,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1d,0x80,0x73,0x80,0x61,0x80, +0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0x61,0x80,0x73,0x80, +0x1d,0x80, +}; + +static const BitmapCharRec ch113 = {10,17,-1,5,12,ch113data}; + +/* char: 0x70 'p' */ + +static const GLubyte ch112data[] = { +0xf0,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x6e,0x0,0x73,0x80,0x61,0x80, +0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x61,0x80,0x73,0x80, +0xee,0x0, +}; + +static const BitmapCharRec ch112 = {10,17,-1,5,12,ch112data}; + +/* char: 0x6f 'o' */ + +static const GLubyte ch111data[] = { +0x1e,0x0,0x73,0x80,0x61,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0,0x61,0x80,0x73,0x80,0x1e,0x0, +}; + +static const BitmapCharRec ch111 = {10,12,-1,0,12,ch111data}; + +/* char: 0x6e 'n' */ + +static const GLubyte ch110data[] = { +0xf1,0xe0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x60,0xc0,0x71,0xc0,0x6f,0x80,0xe7,0x0, +}; + +static const BitmapCharRec ch110 = {11,12,-1,0,13,ch110data}; + +/* char: 0x6d 'm' */ + +static const GLubyte ch109data[] = { +0xf1,0xe3,0xc0,0x60,0xc1,0x80,0x60,0xc1,0x80,0x60,0xc1,0x80,0x60,0xc1,0x80,0x60, +0xc1,0x80,0x60,0xc1,0x80,0x60,0xc1,0x80,0x60,0xc1,0x80,0x71,0xe3,0x80,0x6f,0x9f, +0x0,0xe7,0xe,0x0, +}; + +static const BitmapCharRec ch109 = {18,12,-1,0,20,ch109data}; + +/* char: 0x6c 'l' */ + +static const GLubyte ch108data[] = { +0xf0,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60, +0xe0, +}; + +static const BitmapCharRec ch108 = {4,17,-1,0,6,ch108data}; + +/* char: 0x6b 'k' */ + +static const GLubyte ch107data[] = { +0xf3,0xe0,0x61,0xc0,0x63,0x80,0x67,0x0,0x6e,0x0,0x6c,0x0,0x78,0x0,0x68,0x0, +0x64,0x0,0x66,0x0,0x63,0x0,0x67,0xc0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0, +0xe0,0x0, +}; + +static const BitmapCharRec ch107 = {11,17,-1,0,12,ch107data}; + +/* char: 0x6a 'j' */ + +static const GLubyte ch106data[] = { +0xc0,0xe0,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, +0x70,0x0,0x0,0x0,0x30,0x30, +}; + +static const BitmapCharRec ch106 = {4,22,0,5,6,ch106data}; + +/* char: 0x69 'i' */ + +static const GLubyte ch105data[] = { +0xf0,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0x60,0xe0,0x0,0x0,0x0,0x60, +0x60, +}; + +static const BitmapCharRec ch105 = {4,17,-1,0,6,ch105data}; + +/* char: 0x68 'h' */ + +static const GLubyte ch104data[] = { +0xf1,0xe0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x60,0xc0,0x71,0xc0,0x6f,0x80,0x67,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0, +0xe0,0x0, +}; + +static const BitmapCharRec ch104 = {11,17,-1,0,13,ch104data}; + +/* char: 0x67 'g' */ + +static const GLubyte ch103data[] = { +0x3f,0x0,0xf1,0xc0,0xc0,0x60,0xc0,0x20,0x60,0x60,0x3f,0xc0,0x7f,0x0,0x60,0x0, +0x30,0x0,0x3e,0x0,0x33,0x0,0x61,0x80,0x61,0x80,0x61,0x80,0x61,0x80,0x33,0x0, +0x1f,0xc0, +}; + +static const BitmapCharRec ch103 = {11,17,-1,5,12,ch103data}; + +/* char: 0x66 'f' */ + +static const GLubyte ch102data[] = { +0x78,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0xfe,0x30,0x30,0x30,0x16, +0xe, +}; + +static const BitmapCharRec ch102 = {7,17,0,0,7,ch102data}; + +/* char: 0x65 'e' */ + +static const GLubyte ch101data[] = { +0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xff,0x80, +0xc1,0x80,0x41,0x80,0x63,0x0,0x1e,0x0, +}; + +static const BitmapCharRec ch101 = {9,12,-1,0,11,ch101data}; + +/* char: 0x64 'd' */ + +static const GLubyte ch100data[] = { +0x1e,0xc0,0x73,0x80,0x61,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80,0xc1,0x80, +0xc1,0x80,0x61,0x80,0x73,0x80,0x1d,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80, +0x3,0x80, +}; + +static const BitmapCharRec ch100 = {10,17,-1,0,12,ch100data}; + +/* char: 0x63 'c' */ + +static const GLubyte ch99data[] = { +0x1e,0x0,0x7f,0x0,0x70,0x80,0xe0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, +0xc0,0x0,0x41,0x80,0x63,0x80,0x1f,0x0, +}; + +static const BitmapCharRec ch99 = {9,12,-1,0,11,ch99data}; + +/* char: 0x62 'b' */ + +static const GLubyte ch98data[] = { +0x5e,0x0,0x73,0x80,0x61,0x80,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0,0x60,0xc0, +0x60,0xc0,0x61,0x80,0x73,0x80,0x6e,0x0,0x60,0x0,0x60,0x0,0x60,0x0,0x60,0x0, +0xe0,0x0, +}; + +static const BitmapCharRec ch98 = {10,17,-1,0,12,ch98data}; + +/* char: 0x61 'a' */ + +static const GLubyte ch97data[] = { +0x71,0x80,0xfb,0x0,0xc7,0x0,0xc3,0x0,0xc3,0x0,0x63,0x0,0x3b,0x0,0xf,0x0, +0x3,0x0,0x63,0x0,0x67,0x0,0x3e,0x0, +}; + +static const BitmapCharRec ch97 = {9,12,-1,0,11,ch97data}; + +/* char: 0x60 '`' */ + +static const GLubyte ch96data[] = { +0x60,0xe0,0x80,0xc0,0x60, +}; + +static const BitmapCharRec ch96 = {3,5,-2,-12,7,ch96data}; + +/* char: 0x5f '_' */ + +static const GLubyte ch95data[] = { +0xff,0xf8,0xff,0xf8, +}; + +static const BitmapCharRec ch95 = {13,2,0,5,13,ch95data}; + +/* char: 0x5e '^' */ + +static const GLubyte ch94data[] = { +0x80,0x80,0xc1,0x80,0x41,0x0,0x63,0x0,0x22,0x0,0x36,0x0,0x14,0x0,0x1c,0x0, +0x8,0x0, +}; + +static const BitmapCharRec ch94 = {9,9,-1,-8,11,ch94data}; + +/* char: 0x5d ']' */ + +static const GLubyte ch93data[] = { +0xf8,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18, +0x18,0x18,0x18,0x18,0xf8, +}; + +static const BitmapCharRec ch93 = {5,21,-1,4,8,ch93data}; + +/* char: 0x5c '\' */ + +static const GLubyte ch92data[] = { +0x6,0x6,0x4,0xc,0xc,0x8,0x18,0x18,0x10,0x30,0x30,0x20,0x60,0x60,0x40,0xc0, +0xc0, +}; + +static const BitmapCharRec ch92 = {7,17,0,0,7,ch92data}; + +/* char: 0x5b '[' */ + +static const GLubyte ch91data[] = { +0xf8,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0,0xc0,0xc0,0xf8, +}; + +static const BitmapCharRec ch91 = {5,21,-2,4,8,ch91data}; + +/* char: 0x5a 'Z' */ + +static const GLubyte ch90data[] = { +0xff,0xf8,0xe0,0x18,0x70,0x8,0x30,0x8,0x38,0x0,0x18,0x0,0x1c,0x0,0xe,0x0, +0x6,0x0,0x7,0x0,0x3,0x0,0x3,0x80,0x1,0xc0,0x80,0xc0,0x80,0xe0,0xc0,0x70, +0xff,0xf0, +}; + +static const BitmapCharRec ch90 = {13,17,-1,0,15,ch90data}; + +/* char: 0x59 'Y' */ + +static const GLubyte ch89data[] = { +0x7,0xe0,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x1,0x80,0x3,0xc0, +0x3,0x40,0x6,0x60,0x6,0x20,0xc,0x30,0x1c,0x10,0x18,0x18,0x38,0x8,0x30,0xc, +0xfc,0x3f, +}; + +static const BitmapCharRec ch89 = {16,17,0,0,16,ch89data}; + +/* char: 0x58 'X' */ + +static const GLubyte ch88data[] = { +0xfc,0xf,0xc0,0x30,0x3,0x80,0x18,0x7,0x0,0x8,0xe,0x0,0x4,0xc,0x0,0x6, +0x18,0x0,0x2,0x38,0x0,0x1,0x70,0x0,0x0,0xe0,0x0,0x0,0xc0,0x0,0x1,0xc0, +0x0,0x3,0xa0,0x0,0x3,0x10,0x0,0x6,0x8,0x0,0xe,0xc,0x0,0x1c,0x6,0x0, +0x7e,0xf,0x80, +}; + +static const BitmapCharRec ch88 = {18,17,0,0,18,ch88data}; + +/* char: 0x57 'W' */ + +static const GLubyte ch87data[] = { +0x1,0x83,0x0,0x1,0x83,0x0,0x1,0x83,0x80,0x3,0x87,0x80,0x3,0x46,0x80,0x3, +0x46,0xc0,0x6,0x46,0x40,0x6,0x4c,0x40,0x6,0x4c,0x60,0xc,0x2c,0x60,0xc,0x2c, +0x20,0x18,0x2c,0x20,0x18,0x18,0x30,0x18,0x18,0x10,0x30,0x18,0x10,0x30,0x18,0x18, +0xfc,0x7e,0x7e, +}; + +static const BitmapCharRec ch87 = {23,17,0,0,23,ch87data}; + +/* char: 0x56 'V' */ + +static const GLubyte ch86data[] = { +0x1,0x80,0x0,0x1,0x80,0x0,0x1,0x80,0x0,0x3,0xc0,0x0,0x3,0x40,0x0,0x3, +0x60,0x0,0x6,0x20,0x0,0x6,0x20,0x0,0x6,0x30,0x0,0xc,0x10,0x0,0xc,0x18, +0x0,0x18,0x8,0x0,0x18,0x8,0x0,0x18,0xc,0x0,0x30,0x4,0x0,0x30,0x6,0x0, +0xfc,0x1f,0x80, +}; + +static const BitmapCharRec ch86 = {17,17,0,0,17,ch86data}; + +/* char: 0x55 'U' */ + +static const GLubyte ch85data[] = { +0x7,0xe0,0x1c,0x30,0x18,0x8,0x30,0x8,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4,0x30,0x4, +0xfc,0x1f, +}; + +static const BitmapCharRec ch85 = {16,17,-1,0,18,ch85data}; + +/* char: 0x54 'T' */ + +static const GLubyte ch84data[] = { +0xf,0xc0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0, +0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0x83,0x4,0x83,0x4,0xc3,0xc, +0xff,0xfc, +}; + +static const BitmapCharRec ch84 = {14,17,-1,0,16,ch84data}; + +/* char: 0x53 'S' */ + +static const GLubyte ch83data[] = { +0x9e,0x0,0xf1,0x80,0xc0,0xc0,0x80,0x60,0x80,0x60,0x0,0x60,0x0,0xe0,0x3,0xc0, +0xf,0x80,0x1e,0x0,0x78,0x0,0xe0,0x0,0xc0,0x40,0xc0,0x40,0xc0,0xc0,0x63,0xc0, +0x1e,0x40, +}; + +static const BitmapCharRec ch83 = {11,17,-1,0,13,ch83data}; + +/* char: 0x52 'R' */ + +static const GLubyte ch82data[] = { +0xfc,0x1e,0x30,0x1c,0x30,0x38,0x30,0x70,0x30,0x60,0x30,0xc0,0x31,0xc0,0x33,0x80, +0x3f,0xc0,0x30,0x70,0x30,0x30,0x30,0x38,0x30,0x18,0x30,0x38,0x30,0x30,0x30,0x70, +0xff,0xc0, +}; + +static const BitmapCharRec ch82 = {15,17,-1,0,16,ch82data}; + +/* char: 0x51 'Q' */ + +static const GLubyte ch81data[] = { +0x0,0xf,0x0,0x38,0x0,0x70,0x0,0xe0,0x1,0xc0,0x7,0xe0,0x1c,0x38,0x38,0x1c, +0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3, +0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38,0x7,0xe0, +}; + +static const BitmapCharRec ch81 = {16,22,-1,5,18,ch81data}; + +/* char: 0x50 'P' */ + +static const GLubyte ch80data[] = { +0xfc,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0, +0x3f,0xc0,0x30,0x70,0x30,0x30,0x30,0x18,0x30,0x18,0x30,0x18,0x30,0x30,0x30,0x70, +0xff,0xc0, +}; + +static const BitmapCharRec ch80 = {13,17,-1,0,15,ch80data}; + +/* char: 0x4f 'O' */ + +static const GLubyte ch79data[] = { +0x7,0xe0,0x1c,0x38,0x38,0x1c,0x60,0x6,0x60,0x6,0xc0,0x3,0xc0,0x3,0xc0,0x3, +0xc0,0x3,0xc0,0x3,0xc0,0x3,0xc0,0x3,0x60,0x6,0x60,0x6,0x38,0x1c,0x1c,0x38, +0x7,0xe0, +}; + +static const BitmapCharRec ch79 = {16,17,-1,0,18,ch79data}; + +/* char: 0x4e 'N' */ + +static const GLubyte ch78data[] = { +0xf8,0xc,0x20,0x1c,0x20,0x1c,0x20,0x34,0x20,0x64,0x20,0x64,0x20,0xc4,0x21,0x84, +0x21,0x84,0x23,0x4,0x26,0x4,0x26,0x4,0x2c,0x4,0x38,0x4,0x38,0x4,0x30,0x4, +0xf0,0x1f, +}; + +static const BitmapCharRec ch78 = {16,17,-1,0,18,ch78data}; + +/* char: 0x4d 'M' */ + +static const GLubyte ch77data[] = { +0xf8,0x21,0xf8,0x20,0x60,0x60,0x20,0x60,0x60,0x20,0xd0,0x60,0x20,0xd0,0x60,0x21, +0x88,0x60,0x21,0x88,0x60,0x23,0x8,0x60,0x23,0x4,0x60,0x26,0x4,0x60,0x26,0x2, +0x60,0x2c,0x2,0x60,0x2c,0x2,0x60,0x38,0x1,0x60,0x38,0x1,0x60,0x30,0x0,0xe0, +0xf0,0x0,0xf8, +}; + +static const BitmapCharRec ch77 = {21,17,-1,0,22,ch77data}; + +/* char: 0x4c 'L' */ + +static const GLubyte ch76data[] = { +0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0, +0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0, +0xfc,0x0, +}; + +static const BitmapCharRec ch76 = {13,17,-1,0,14,ch76data}; + +/* char: 0x4b 'K' */ + +static const GLubyte ch75data[] = { +0xfc,0x1f,0x30,0xe,0x30,0x1c,0x30,0x38,0x30,0x70,0x30,0xe0,0x31,0xc0,0x33,0x80, +0x3f,0x0,0x3e,0x0,0x33,0x0,0x31,0x80,0x30,0xc0,0x30,0x60,0x30,0x30,0x30,0x18, +0xfc,0x7e, +}; + +static const BitmapCharRec ch75 = {16,17,-1,0,17,ch75data}; + +/* char: 0x4a 'J' */ + +static const GLubyte ch74data[] = { +0x78,0x0,0xcc,0x0,0xc6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, +0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, +0x1f,0x80, +}; + +static const BitmapCharRec ch74 = {9,17,-1,0,11,ch74data}; + +/* char: 0x49 'I' */ + +static const GLubyte ch73data[] = { +0xfc,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30,0x30, +0xfc, +}; + +static const BitmapCharRec ch73 = {6,17,-1,0,8,ch73data}; + +/* char: 0x48 'H' */ + +static const GLubyte ch72data[] = { +0xfc,0x1f,0x80,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30, +0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x3f,0xfe,0x0,0x30,0x6,0x0,0x30,0x6, +0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0,0x30,0x6,0x0, +0xfc,0x1f,0x80, +}; + +static const BitmapCharRec ch72 = {17,17,-1,0,19,ch72data}; + +/* char: 0x47 'G' */ + +static const GLubyte ch71data[] = { +0x7,0xe0,0x1e,0x38,0x38,0x1c,0x60,0xc,0x60,0xc,0xc0,0xc,0xc0,0xc,0xc0,0x3f, +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0x60,0x4,0x60,0x4,0x38,0xc,0x1c,0x3c, +0x7,0xe4, +}; + +static const BitmapCharRec ch71 = {16,17,-1,0,18,ch71data}; + +/* char: 0x46 'F' */ + +static const GLubyte ch70data[] = { +0xfc,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x0,0x30,0x20,0x30,0x20, +0x3f,0xe0,0x30,0x20,0x30,0x20,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, +0xff,0xf0, +}; + +static const BitmapCharRec ch70 = {12,17,-1,0,14,ch70data}; + +/* char: 0x45 'E' */ + +static const GLubyte ch69data[] = { +0xff,0xf8,0x30,0x18,0x30,0x8,0x30,0x8,0x30,0x0,0x30,0x0,0x30,0x40,0x30,0x40, +0x3f,0xc0,0x30,0x40,0x30,0x40,0x30,0x0,0x30,0x0,0x30,0x10,0x30,0x10,0x30,0x30, +0xff,0xf0, +}; + +static const BitmapCharRec ch69 = {13,17,-1,0,15,ch69data}; + +/* char: 0x44 'D' */ + +static const GLubyte ch68data[] = { +0xff,0xc0,0x30,0x70,0x30,0x38,0x30,0xc,0x30,0xc,0x30,0x6,0x30,0x6,0x30,0x6, +0x30,0x6,0x30,0x6,0x30,0x6,0x30,0x6,0x30,0xc,0x30,0xc,0x30,0x38,0x30,0x70, +0xff,0xc0, +}; + +static const BitmapCharRec ch68 = {15,17,-1,0,17,ch68data}; + +/* char: 0x43 'C' */ + +static const GLubyte ch67data[] = { +0x7,0xe0,0x1e,0x38,0x38,0x8,0x60,0x4,0x60,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0, +0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0x60,0x4,0x60,0x4,0x38,0xc,0x1c,0x3c, +0x7,0xe4, +}; + +static const BitmapCharRec ch67 = {14,17,-1,0,16,ch67data}; + +/* char: 0x42 'B' */ + +static const GLubyte ch66data[] = { +0xff,0xe0,0x30,0x78,0x30,0x18,0x30,0xc,0x30,0xc,0x30,0xc,0x30,0x18,0x30,0x38, +0x3f,0xe0,0x30,0x40,0x30,0x30,0x30,0x18,0x30,0x18,0x30,0x18,0x30,0x30,0x30,0x70, +0xff,0xc0, +}; + +static const BitmapCharRec ch66 = {14,17,-1,0,16,ch66data}; + +/* char: 0x41 'A' */ + +static const GLubyte ch65data[] = { +0xfc,0x1f,0x80,0x30,0x6,0x0,0x10,0x6,0x0,0x10,0xc,0x0,0x18,0xc,0x0,0x8, +0xc,0x0,0xf,0xf8,0x0,0xc,0x18,0x0,0x4,0x18,0x0,0x4,0x30,0x0,0x6,0x30, +0x0,0x2,0x30,0x0,0x2,0x60,0x0,0x1,0x60,0x0,0x1,0xc0,0x0,0x1,0xc0,0x0, +0x0,0x80,0x0, +}; + +static const BitmapCharRec ch65 = {17,17,0,0,17,ch65data}; + +/* char: 0x40 '@' */ + +static const GLubyte ch64data[] = { +0x3,0xf0,0x0,0xe,0xc,0x0,0x18,0x0,0x0,0x30,0x0,0x0,0x61,0xde,0x0,0x63, +0x7b,0x0,0xc6,0x39,0x80,0xc6,0x18,0x80,0xc6,0x18,0xc0,0xc6,0x18,0x40,0xc6,0xc, +0x40,0xc3,0xc,0x40,0xc3,0x8c,0x40,0xe1,0xfc,0x40,0x60,0xec,0xc0,0x70,0x0,0x80, +0x38,0x1,0x80,0x1c,0x3,0x0,0xf,0xe,0x0,0x3,0xf8,0x0, +}; + +static const BitmapCharRec ch64 = {18,20,-2,3,22,ch64data}; + +/* char: 0x3f '?' */ + +static const GLubyte ch63data[] = { +0x30,0x30,0x0,0x0,0x10,0x10,0x10,0x18,0x18,0xc,0xe,0x7,0xc3,0xc3,0x83,0xc6, +0x7c, +}; + +static const BitmapCharRec ch63 = {8,17,-2,0,11,ch63data}; + +/* char: 0x3e '>' */ + +static const GLubyte ch62data[] = { +0xc0,0x0,0x70,0x0,0x1c,0x0,0x7,0x0,0x1,0xc0,0x0,0x60,0x1,0xc0,0x7,0x0, +0x1c,0x0,0x70,0x0,0xc0,0x0, +}; + +static const BitmapCharRec ch62 = {11,11,-1,-1,13,ch62data}; + +/* char: 0x3d '=' */ + +static const GLubyte ch61data[] = { +0xff,0xf0,0xff,0xf0,0x0,0x0,0x0,0x0,0xff,0xf0,0xff,0xf0, +}; + +static const BitmapCharRec ch61 = {12,6,-1,-4,14,ch61data}; + +/* char: 0x3c '<' */ + +static const GLubyte ch60data[] = { +0x0,0x60,0x1,0xc0,0x7,0x0,0x1c,0x0,0x70,0x0,0xc0,0x0,0x70,0x0,0x1c,0x0, +0x7,0x0,0x1,0xc0,0x0,0x60, +}; + +static const BitmapCharRec ch60 = {11,11,-1,-1,13,ch60data}; + +/* char: 0x3b ';' */ + +static const GLubyte ch59data[] = { +0xc0,0x60,0x20,0xe0,0xc0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0xc0,0xc0, +}; + +static const BitmapCharRec ch59 = {3,14,-2,3,7,ch59data}; + +/* char: 0x3a ':' */ + +static const GLubyte ch58data[] = { +0xc0,0xc0,0x0,0x0,0x0,0x0,0x0,0x0,0x0,0xc0,0xc0, +}; + +static const BitmapCharRec ch58 = {2,11,-2,0,6,ch58data}; + +/* char: 0x39 '9' */ + +static const GLubyte ch57data[] = { +0xf0,0x0,0x1c,0x0,0x6,0x0,0x3,0x0,0x3,0x80,0x1,0x80,0x1d,0x80,0x73,0xc0, +0x61,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc1,0xc0,0x61,0x80,0x77,0x80, +0x1e,0x0, +}; + +static const BitmapCharRec ch57 = {10,17,-1,0,12,ch57data}; + +/* char: 0x38 '8' */ + +static const GLubyte ch56data[] = { +0x1e,0x0,0x73,0x80,0xe1,0x80,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x41,0xc0,0x61,0x80, +0x37,0x0,0x1e,0x0,0x1e,0x0,0x33,0x0,0x61,0x80,0x61,0x80,0x61,0x80,0x33,0x0, +0x1e,0x0, +}; + +static const BitmapCharRec ch56 = {10,17,-1,0,12,ch56data}; + +/* char: 0x37 '7' */ + +static const GLubyte ch55data[] = { +0x18,0x0,0x18,0x0,0xc,0x0,0xc,0x0,0xc,0x0,0x4,0x0,0x6,0x0,0x6,0x0, +0x2,0x0,0x3,0x0,0x3,0x0,0x1,0x0,0x1,0x80,0x81,0x80,0xc0,0xc0,0xff,0xc0, +0x7f,0xc0, +}; + +static const BitmapCharRec ch55 = {10,17,-1,0,12,ch55data}; + +/* char: 0x36 '6' */ + +static const GLubyte ch54data[] = { +0x1e,0x0,0x7b,0x80,0x61,0x80,0xe0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc1,0x80,0xf3,0x80,0xee,0x0,0x60,0x0,0x70,0x0,0x30,0x0,0x18,0x0,0xe,0x0, +0x3,0xc0, +}; + +static const BitmapCharRec ch54 = {10,17,-1,0,12,ch54data}; + +/* char: 0x35 '5' */ + +static const GLubyte ch53data[] = { +0x7e,0x0,0xe3,0x80,0xc1,0x80,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x0,0xc0,0x1,0xc0, +0x3,0x80,0xf,0x80,0x7e,0x0,0x78,0x0,0x60,0x0,0x20,0x0,0x20,0x0,0x1f,0x80, +0x1f,0xc0, +}; + +static const BitmapCharRec ch53 = {10,17,-1,0,12,ch53data}; + +/* char: 0x34 '4' */ + +static const GLubyte ch52data[] = { +0x3,0x0,0x3,0x0,0x3,0x0,0x3,0x0,0xff,0xc0,0xff,0xc0,0xc3,0x0,0x43,0x0, +0x63,0x0,0x23,0x0,0x33,0x0,0x13,0x0,0x1b,0x0,0xb,0x0,0x7,0x0,0x7,0x0, +0x3,0x0, +}; + +static const BitmapCharRec ch52 = {10,17,-1,0,12,ch52data}; + +/* char: 0x33 '3' */ + +static const GLubyte ch51data[] = { +0x78,0x0,0xe6,0x0,0xc3,0x0,0x1,0x0,0x1,0x80,0x1,0x80,0x1,0x80,0x3,0x80, +0x7,0x0,0x1e,0x0,0xc,0x0,0x6,0x0,0x83,0x0,0x83,0x0,0x47,0x0,0x7e,0x0, +0x1c,0x0, +}; + +static const BitmapCharRec ch51 = {9,17,-1,0,12,ch51data}; + +/* char: 0x32 '2' */ + +static const GLubyte ch50data[] = { +0xff,0x80,0xff,0xc0,0x60,0x40,0x30,0x0,0x18,0x0,0xc,0x0,0x4,0x0,0x6,0x0, +0x3,0x0,0x3,0x0,0x1,0x80,0x1,0x80,0x81,0x80,0x81,0x80,0x43,0x80,0x7f,0x0, +0x1c,0x0, +}; + +static const BitmapCharRec ch50 = {10,17,-1,0,12,ch50data}; + +/* char: 0x31 '1' */ + +static const GLubyte ch49data[] = { +0xff,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x18,0x78,0x18, +0x8, +}; + +static const BitmapCharRec ch49 = {8,17,-2,0,12,ch49data}; + +/* char: 0x30 '0' */ + +static const GLubyte ch48data[] = { +0x1e,0x0,0x33,0x0,0x61,0x80,0x61,0x80,0xe1,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x61,0x80,0x61,0x80,0x33,0x0, +0x1e,0x0, +}; + +static const BitmapCharRec ch48 = {10,17,-1,0,12,ch48data}; + +/* char: 0x2f '/' */ + +static const GLubyte ch47data[] = { +0xc0,0xc0,0xc0,0x60,0x60,0x20,0x30,0x30,0x10,0x18,0x18,0x8,0xc,0xc,0x4,0x6, +0x6,0x3,0x3,0x3, +}; + +static const BitmapCharRec ch47 = {8,20,1,3,7,ch47data}; + +/* char: 0x2e '.' */ + +static const GLubyte ch46data[] = { +0xc0,0xc0, +}; + +static const BitmapCharRec ch46 = {2,2,-2,0,6,ch46data}; + +/* char: 0x2d '-' */ + +static const GLubyte ch45data[] = { +0xff,0xf0,0xff,0xf0, +}; + +static const BitmapCharRec ch45 = {12,2,-1,-6,14,ch45data}; + +/* char: 0x2c ',' */ + +static const GLubyte ch44data[] = { +0xc0,0x60,0x20,0xe0,0xc0, +}; + +static const BitmapCharRec ch44 = {3,5,-2,3,7,ch44data}; + +/* char: 0x2b '+' */ + +static const GLubyte ch43data[] = { +0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0,0xff,0xf0,0xff,0xf0,0x6,0x0, +0x6,0x0,0x6,0x0,0x6,0x0,0x6,0x0, +}; + +static const BitmapCharRec ch43 = {12,12,-1,-1,14,ch43data}; + +/* char: 0x2a '*' */ + +static const GLubyte ch42data[] = { +0x8,0x0,0x1c,0x0,0xc9,0x80,0xeb,0x80,0x1c,0x0,0xeb,0x80,0xc9,0x80,0x1c,0x0, +0x8,0x0, +}; + +static const BitmapCharRec ch42 = {9,9,-2,-8,12,ch42data}; + +/* char: 0x29 ')' */ + +static const GLubyte ch41data[] = { +0x80,0x40,0x20,0x30,0x10,0x18,0x18,0xc,0xc,0xc,0xc,0xc,0xc,0xc,0xc,0x18, +0x18,0x10,0x30,0x20,0x40,0x80, +}; + +static const BitmapCharRec ch41 = {6,22,-1,5,8,ch41data}; + +/* char: 0x28 '(' */ + +static const GLubyte ch40data[] = { +0x4,0x8,0x10,0x30,0x20,0x60,0x60,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0x60, +0x60,0x20,0x30,0x10,0x8,0x4, +}; + +static const BitmapCharRec ch40 = {6,22,-1,5,8,ch40data}; + +/* char: 0x27 ''' */ + +static const GLubyte ch39data[] = { +0xc0,0x60,0x20,0xe0,0xc0, +}; + +static const BitmapCharRec ch39 = {3,5,-3,-12,8,ch39data}; + +/* char: 0x26 '&' */ + +static const GLubyte ch38data[] = { +0x3c,0x3c,0x7f,0x7e,0xe1,0xe1,0xc0,0xc0,0xc1,0xc0,0xc1,0xa0,0x63,0x20,0x37,0x10, +0x1e,0x18,0xe,0x3e,0xf,0x0,0x1d,0x80,0x18,0xc0,0x18,0x40,0x18,0x40,0xc,0xc0, +0x7,0x80, +}; + +static const BitmapCharRec ch38 = {16,17,-1,0,18,ch38data}; + +/* char: 0x25 '%' */ + +static const GLubyte ch37data[] = { +0x30,0x3c,0x0,0x18,0x72,0x0,0xc,0x61,0x0,0x4,0x60,0x80,0x6,0x60,0x80,0x3, +0x30,0x80,0x1,0x19,0x80,0x1,0x8f,0x0,0x78,0xc0,0x0,0xe4,0x40,0x0,0xc2,0x60, +0x0,0xc1,0x30,0x0,0xc1,0x10,0x0,0x61,0x18,0x0,0x33,0xfc,0x0,0x1e,0xc,0x0, +}; + +static const BitmapCharRec ch37 = {17,16,-1,0,19,ch37data}; + +/* char: 0x24 '$' */ + +static const GLubyte ch36data[] = { +0x4,0x0,0x4,0x0,0x3f,0x0,0xe5,0xc0,0xc4,0xc0,0x84,0x60,0x84,0x60,0x4,0x60, +0x4,0xe0,0x7,0xc0,0x7,0x80,0x1e,0x0,0x3c,0x0,0x74,0x0,0x64,0x0,0x64,0x20, +0x64,0x60,0x34,0xe0,0x1f,0x80,0x4,0x0,0x4,0x0, +}; + +static const BitmapCharRec ch36 = {11,21,0,2,12,ch36data}; + +/* char: 0x23 '#' */ + +static const GLubyte ch35data[] = { +0x22,0x0,0x22,0x0,0x22,0x0,0x22,0x0,0x22,0x0,0xff,0xc0,0xff,0xc0,0x11,0x0, +0x11,0x0,0x11,0x0,0x7f,0xe0,0x7f,0xe0,0x8,0x80,0x8,0x80,0x8,0x80,0x8,0x80, +0x8,0x80, +}; + +static const BitmapCharRec ch35 = {11,17,-1,0,13,ch35data}; + +/* char: 0x22 '"' */ + +static const GLubyte ch34data[] = { +0x88,0xcc,0xcc,0xcc,0xcc, +}; + +static const BitmapCharRec ch34 = {6,5,-1,-12,10,ch34data}; + +/* char: 0x21 '!' */ + +static const GLubyte ch33data[] = { +0xc0,0xc0,0x0,0x0,0x0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0,0xc0, +0xc0, +}; + +static const BitmapCharRec ch33 = {2,17,-3,0,8,ch33data}; + +/* char: 0x20 ' ' */ + +#ifdef _WIN32 +/* XXX Work around Microsoft OpenGL 1.1 bug where glBitmap with + a height or width of zero does not advance the raster position + as specified by OpenGL. (Cosmo OpenGL does not have this bug.) */ +static const GLubyte ch32data[] = { 0x0 }; +static const BitmapCharRec ch32 = {1,1,0,0,6,ch32data}; +#else +static const BitmapCharRec ch32 = {0,0,0,0,6,0}; +#endif + +static const BitmapCharRec * const chars[] = { +&ch32, +&ch33, +&ch34, +&ch35, +&ch36, +&ch37, +&ch38, +&ch39, +&ch40, +&ch41, +&ch42, +&ch43, +&ch44, +&ch45, +&ch46, +&ch47, +&ch48, +&ch49, +&ch50, +&ch51, +&ch52, +&ch53, +&ch54, +&ch55, +&ch56, +&ch57, +&ch58, +&ch59, +&ch60, +&ch61, +&ch62, +&ch63, +&ch64, +&ch65, +&ch66, +&ch67, +&ch68, +&ch69, +&ch70, +&ch71, +&ch72, +&ch73, +&ch74, +&ch75, +&ch76, +&ch77, +&ch78, +&ch79, +&ch80, +&ch81, +&ch82, +&ch83, +&ch84, +&ch85, +&ch86, +&ch87, +&ch88, +&ch89, +&ch90, +&ch91, +&ch92, +&ch93, +&ch94, +&ch95, +&ch96, +&ch97, +&ch98, +&ch99, +&ch100, +&ch101, +&ch102, +&ch103, +&ch104, +&ch105, +&ch106, +&ch107, +&ch108, +&ch109, +&ch110, +&ch111, +&ch112, +&ch113, +&ch114, +&ch115, +&ch116, +&ch117, +&ch118, +&ch119, +&ch120, +&ch121, +&ch122, +&ch123, +&ch124, +&ch125, +&ch126, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +0, +&ch160, +&ch161, +&ch162, +&ch163, +&ch164, +&ch165, +&ch166, +&ch167, +&ch168, +&ch169, +&ch170, +&ch171, +&ch172, +&ch173, +&ch174, +&ch175, +&ch176, +&ch177, +&ch178, +&ch179, +&ch180, +&ch181, +&ch182, +&ch183, +&ch184, +&ch185, +&ch186, +&ch187, +&ch188, +&ch189, +&ch190, +&ch191, +&ch192, +&ch193, +&ch194, +&ch195, +&ch196, +&ch197, +&ch198, +&ch199, +&ch200, +&ch201, +&ch202, +&ch203, +&ch204, +&ch205, +&ch206, +&ch207, +&ch208, +&ch209, +&ch210, +&ch211, +&ch212, +&ch213, +&ch214, +&ch215, +&ch216, +&ch217, +&ch218, +&ch219, +&ch220, +&ch221, +&ch222, +&ch223, +&ch224, +&ch225, +&ch226, +&ch227, +&ch228, +&ch229, +&ch230, +&ch231, +&ch232, +&ch233, +&ch234, +&ch235, +&ch236, +&ch237, +&ch238, +&ch239, +&ch240, +&ch241, +&ch242, +&ch243, +&ch244, +&ch245, +&ch246, +&ch247, +&ch248, +&ch249, +&ch250, +&ch251, +&ch252, +&ch253, +&ch254, +&ch255, +}; + +#if !defined(__IBMCPP__) +const +#endif +BitmapFontRec glutBitmapTimesRoman24 = { +"-adobe-times-medium-r-normal--24-240-75-75-p-124-iso8859-1", +224, +32, +chars +}; + diff --git a/src/glut/os2/glut_util.cpp b/src/glut/os2/glut_util.cpp index 76f25020af..2a50a7400e 100644 --- a/src/glut/os2/glut_util.cpp +++ b/src/glut/os2/glut_util.cpp @@ -1,90 +1,90 @@ - -/* Copyright (c) Mark J. Kilgard, 1994. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include -#include -#include -#include - -#include "glutint.h" - -#if !defined(__OS2__) - -/* strdup is actually not a standard ANSI C or POSIX routine - so implement a private one for GLUT. OpenVMS does not have a - strdup; Linux's standard libc doesn't declare strdup by default - (unless BSD or SVID interfaces are requested). */ -char * -__glutStrdup(const char *string) -{ - char *copy; - - copy = (char*) malloc(strlen(string) + 1); - if (copy == NULL) - return NULL; - strcpy(copy, string); - return copy; -} -#endif - -void -__glutWarning(char *format,...) -{ - va_list args; - - va_start(args, format); - fprintf(stderr, "GLUT: Warning in %s: ", - __glutProgramName ? __glutProgramName : "(unamed)"); - vfprintf(stderr, format, args); - va_end(args); - putc('\n', stderr); -} - -/* CENTRY */ -void GLUTAPIENTRY -glutReportErrors(void) -{ - GLenum error; - - while ((error = glGetError()) != GL_NO_ERROR) - __glutWarning("GL error: %s", gluErrorString(error)); -} -/* ENDCENTRY */ - -void -__glutFatalError(char *format,...) -{ - va_list args; - - va_start(args, format); - fprintf(stderr, "GLUT: Fatal Error in %s: ", - __glutProgramName ? __glutProgramName : "(unamed)"); - vfprintf(stderr, format, args); - va_end(args); - putc('\n', stderr); -/* || defined(__OS2__) */ -#if defined(_WIN32) - if (__glutExitFunc) { - __glutExitFunc(1); - } -#endif - exit(1); -} - -void -__glutFatalUsage(char *format,...) -{ - va_list args; - - va_start(args, format); - fprintf(stderr, "GLUT: Fatal API Usage in %s: ", - __glutProgramName ? __glutProgramName : "(unamed)"); - vfprintf(stderr, format, args); - va_end(args); - putc('\n', stderr); - abort(); -} + +/* Copyright (c) Mark J. Kilgard, 1994. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include +#include +#include +#include + +#include "glutint.h" + +#if !defined(__OS2__) + +/* strdup is actually not a standard ANSI C or POSIX routine + so implement a private one for GLUT. OpenVMS does not have a + strdup; Linux's standard libc doesn't declare strdup by default + (unless BSD or SVID interfaces are requested). */ +char * +__glutStrdup(const char *string) +{ + char *copy; + + copy = (char*) malloc(strlen(string) + 1); + if (copy == NULL) + return NULL; + strcpy(copy, string); + return copy; +} +#endif + +void +__glutWarning(char *format,...) +{ + va_list args; + + va_start(args, format); + fprintf(stderr, "GLUT: Warning in %s: ", + __glutProgramName ? __glutProgramName : "(unamed)"); + vfprintf(stderr, format, args); + va_end(args); + putc('\n', stderr); +} + +/* CENTRY */ +void GLUTAPIENTRY +glutReportErrors(void) +{ + GLenum error; + + while ((error = glGetError()) != GL_NO_ERROR) + __glutWarning("GL error: %s", gluErrorString(error)); +} +/* ENDCENTRY */ + +void +__glutFatalError(char *format,...) +{ + va_list args; + + va_start(args, format); + fprintf(stderr, "GLUT: Fatal Error in %s: ", + __glutProgramName ? __glutProgramName : "(unamed)"); + vfprintf(stderr, format, args); + va_end(args); + putc('\n', stderr); +/* || defined(__OS2__) */ +#if defined(_WIN32) + if (__glutExitFunc) { + __glutExitFunc(1); + } +#endif + exit(1); +} + +void +__glutFatalUsage(char *format,...) +{ + va_list args; + + va_start(args, format); + fprintf(stderr, "GLUT: Fatal API Usage in %s: ", + __glutProgramName ? __glutProgramName : "(unamed)"); + vfprintf(stderr, format, args); + va_end(args); + putc('\n', stderr); + abort(); +} diff --git a/src/glut/os2/glut_vidresize.cpp b/src/glut/os2/glut_vidresize.cpp index f9d345dcf5..9f9b6eef58 100644 --- a/src/glut/os2/glut_vidresize.cpp +++ b/src/glut/os2/glut_vidresize.cpp @@ -1,232 +1,232 @@ - -/* Copyright (c) Mark J. Kilgard, 1996. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include - -#ifdef __sgi -#include -#endif - -#include "glutint.h" - -/* Grumble. The IRIX 6.3 and early IRIX 6.4 OpenGL headers - support the video resize extension, but failed to define - GLX_SGIX_video_resize. */ -#if 0 -#ifdef GLX_SYNC_FRAME_SGIX -#define GLX_SGIX_video_resize 1 -#endif -#endif - -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) -static int canVideoResize = -1; -static int videoResizeChannel; -#else -static int canVideoResize = 0; -#endif -static int videoResizeInUse = 0; -static int dx = -1, dy = -1, dw = -1, dh = -1; - -/* XXX Note that IRIX 6.2, 6.3, and some 6.4 versions have a - bug where programs seg-fault when they attempt video - resizing from an indirect OpenGL context (either local or - over a network). */ - -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) - -static volatile int errorCaught; - -/* ARGSUSED */ -static int -catchXSGIvcErrors(Display * dpy, XErrorEvent * event) -{ - errorCaught = 1; - return 0; -} -#endif - -/* CENTRY */ - -int GLUTAPIENTRY -glutVideoResizeGet(GLenum param) -{ -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) - if (canVideoResize < 0) { - canVideoResize = __glutIsSupportedByGLX("GLX_SGIX_video_resize"); - if (canVideoResize) { -#if defined(__sgi) && __sgi - /* This is a hack because IRIX 6.2, 6.3, and some 6.4 - versions were released with GLX_SGIX_video_resize - being advertised by the X server though the video - resize extension is not actually supported. We try to - determine if the libGL.so we are using actually has a - video resize entrypoint before we try to use the - feature. */ - void (*func) (void); - void *glxDso = dlopen("libGL.so", RTLD_LAZY); - - func = (void (*)(void)) dlsym(glxDso, "glXQueryChannelDeltasSGIX"); - if (!func) { - canVideoResize = 0; - } else -#endif - { - char *channelString; - int (*handler) (Display *, XErrorEvent *); - - channelString = getenv("GLUT_VIDEO_RESIZE_CHANNEL"); - videoResizeChannel = channelString ? atoi(channelString) : 0; - - /* Work around another annoying problem with SGI's - GLX_SGIX_video_resize implementation. Early IRIX - 6.4 OpenGL's advertise the extension and have the - video resize API, but an XSGIvc X protocol errors - result trying to use the API. Set up an error - handler to intercept what would otherwise be a fatal - error. If an error was recieved, do not report that - video resize is possible. */ - handler = XSetErrorHandler(catchXSGIvcErrors); - - errorCaught = 0; - -#if defined(GLX_GLXEXT_PROTOTYPES) -#endif - - __glut_glXQueryChannelDeltasSGIX(__glutDisplay, __glutScreen, - videoResizeChannel, &dx, &dy, &dw, &dh); - - /* glXQueryChannelDeltasSGIX is an inherent X server - round-trip so we know we will have gotten either the - correct reply or and error by this time. */ - XSetErrorHandler(handler); - - /* Still yet another work around. In IRIX 6.4 betas, - glXQueryChannelDeltasSGIX will return as if it - succeeded, but the values are filled with junk. - Watch to make sure the delta variables really make - sense. */ - if (errorCaught || - dx < 0 || dy < 0 || dw < 0 || dh < 0 || - dx > 2048 || dy > 2048 || dw > 2048 || dh > 2048) { - canVideoResize = 0; - } - } - } - } -#endif /* GLX_SGIX_video_resize */ - - switch (param) { - case GLUT_VIDEO_RESIZE_POSSIBLE: - return canVideoResize; - case GLUT_VIDEO_RESIZE_IN_USE: - return videoResizeInUse; - case GLUT_VIDEO_RESIZE_X_DELTA: - return dx; - case GLUT_VIDEO_RESIZE_Y_DELTA: - return dy; - case GLUT_VIDEO_RESIZE_WIDTH_DELTA: - return dw; - case GLUT_VIDEO_RESIZE_HEIGHT_DELTA: - return dh; - case GLUT_VIDEO_RESIZE_X: - case GLUT_VIDEO_RESIZE_Y: - case GLUT_VIDEO_RESIZE_WIDTH: - case GLUT_VIDEO_RESIZE_HEIGHT: -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) - if (videoResizeInUse) { - int x, y, width, height; - - __glut_glXQueryChannelRectSGIX(__glutDisplay, __glutScreen, - videoResizeChannel, &x, &y, &width, &height); - switch (param) { - case GLUT_VIDEO_RESIZE_X: - return x; - case GLUT_VIDEO_RESIZE_Y: - return y; - case GLUT_VIDEO_RESIZE_WIDTH: - return width; - case GLUT_VIDEO_RESIZE_HEIGHT: - return height; - } - } -#endif - return -1; - default: - __glutWarning("invalid glutVideoResizeGet parameter: %d", param); - return -1; - } -} - -void GLUTAPIENTRY -glutSetupVideoResizing(void) -{ -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) - if (glutVideoResizeGet(GLUT_VIDEO_RESIZE_POSSIBLE)) { - __glut_glXBindChannelToWindowSGIX(__glutDisplay, __glutScreen, - videoResizeChannel, __glutCurrentWindow->win); - videoResizeInUse = 1; - } else -#endif - __glutFatalError("glutEstablishVideoResizing: video resizing not possible.\n"); -} - -void GLUTAPIENTRY -glutStopVideoResizing(void) -{ -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) - if (glutVideoResizeGet(GLUT_VIDEO_RESIZE_POSSIBLE)) { - if (videoResizeInUse) { - __glut_glXBindChannelToWindowSGIX(__glutDisplay, __glutScreen, - videoResizeChannel, None); - videoResizeInUse = 0; - } - } -#endif -} - -/* ARGSUSED */ -void GLUTAPIENTRY -glutVideoResize(int x, int y, int width, int height) -{ -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) - if (videoResizeInUse) { -#ifdef GLX_SYNC_SWAP_SGIX - /* glXChannelRectSyncSGIX introduced in a patch to IRIX - 6.2; the original unpatched IRIX 6.2 behavior is always - GLX_SYNC_SWAP_SGIX. */ - __glut_glXChannelRectSyncSGIX(__glutDisplay, __glutScreen, - videoResizeChannel, GLX_SYNC_SWAP_SGIX); -#endif - __glut_glXChannelRectSGIX(__glutDisplay, __glutScreen, - videoResizeChannel, x, y, width, height); - } -#endif -} - -/* ARGSUSED */ -void GLUTAPIENTRY -glutVideoPan(int x, int y, int width, int height) -{ -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) - if (videoResizeInUse) { -#ifdef GLX_SYNC_FRAME_SGIX - /* glXChannelRectSyncSGIX introduced in a patch to IRIX - 6.2; the original unpatched IRIX 6.2 behavior is always - GLX_SYNC_SWAP_SGIX. We just ignore that we cannot - accomplish GLX_SYNC_FRAME_SGIX on IRIX unpatched 6.2; - this means you'd need a glutSwapBuffers to actually - realize the video resize. */ - __glut_glXChannelRectSyncSGIX(__glutDisplay, __glutScreen, - videoResizeChannel, GLX_SYNC_FRAME_SGIX); -#endif - __glut_glXChannelRectSGIX(__glutDisplay, __glutScreen, - videoResizeChannel, x, y, width, height); - } -#endif -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1996. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include + +#ifdef __sgi +#include +#endif + +#include "glutint.h" + +/* Grumble. The IRIX 6.3 and early IRIX 6.4 OpenGL headers + support the video resize extension, but failed to define + GLX_SGIX_video_resize. */ +#if 0 +#ifdef GLX_SYNC_FRAME_SGIX +#define GLX_SGIX_video_resize 1 +#endif +#endif + +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) +static int canVideoResize = -1; +static int videoResizeChannel; +#else +static int canVideoResize = 0; +#endif +static int videoResizeInUse = 0; +static int dx = -1, dy = -1, dw = -1, dh = -1; + +/* XXX Note that IRIX 6.2, 6.3, and some 6.4 versions have a + bug where programs seg-fault when they attempt video + resizing from an indirect OpenGL context (either local or + over a network). */ + +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) + +static volatile int errorCaught; + +/* ARGSUSED */ +static int +catchXSGIvcErrors(Display * dpy, XErrorEvent * event) +{ + errorCaught = 1; + return 0; +} +#endif + +/* CENTRY */ + +int GLUTAPIENTRY +glutVideoResizeGet(GLenum param) +{ +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) + if (canVideoResize < 0) { + canVideoResize = __glutIsSupportedByGLX("GLX_SGIX_video_resize"); + if (canVideoResize) { +#if defined(__sgi) && __sgi + /* This is a hack because IRIX 6.2, 6.3, and some 6.4 + versions were released with GLX_SGIX_video_resize + being advertised by the X server though the video + resize extension is not actually supported. We try to + determine if the libGL.so we are using actually has a + video resize entrypoint before we try to use the + feature. */ + void (*func) (void); + void *glxDso = dlopen("libGL.so", RTLD_LAZY); + + func = (void (*)(void)) dlsym(glxDso, "glXQueryChannelDeltasSGIX"); + if (!func) { + canVideoResize = 0; + } else +#endif + { + char *channelString; + int (*handler) (Display *, XErrorEvent *); + + channelString = getenv("GLUT_VIDEO_RESIZE_CHANNEL"); + videoResizeChannel = channelString ? atoi(channelString) : 0; + + /* Work around another annoying problem with SGI's + GLX_SGIX_video_resize implementation. Early IRIX + 6.4 OpenGL's advertise the extension and have the + video resize API, but an XSGIvc X protocol errors + result trying to use the API. Set up an error + handler to intercept what would otherwise be a fatal + error. If an error was recieved, do not report that + video resize is possible. */ + handler = XSetErrorHandler(catchXSGIvcErrors); + + errorCaught = 0; + +#if defined(GLX_GLXEXT_PROTOTYPES) +#endif + + __glut_glXQueryChannelDeltasSGIX(__glutDisplay, __glutScreen, + videoResizeChannel, &dx, &dy, &dw, &dh); + + /* glXQueryChannelDeltasSGIX is an inherent X server + round-trip so we know we will have gotten either the + correct reply or and error by this time. */ + XSetErrorHandler(handler); + + /* Still yet another work around. In IRIX 6.4 betas, + glXQueryChannelDeltasSGIX will return as if it + succeeded, but the values are filled with junk. + Watch to make sure the delta variables really make + sense. */ + if (errorCaught || + dx < 0 || dy < 0 || dw < 0 || dh < 0 || + dx > 2048 || dy > 2048 || dw > 2048 || dh > 2048) { + canVideoResize = 0; + } + } + } + } +#endif /* GLX_SGIX_video_resize */ + + switch (param) { + case GLUT_VIDEO_RESIZE_POSSIBLE: + return canVideoResize; + case GLUT_VIDEO_RESIZE_IN_USE: + return videoResizeInUse; + case GLUT_VIDEO_RESIZE_X_DELTA: + return dx; + case GLUT_VIDEO_RESIZE_Y_DELTA: + return dy; + case GLUT_VIDEO_RESIZE_WIDTH_DELTA: + return dw; + case GLUT_VIDEO_RESIZE_HEIGHT_DELTA: + return dh; + case GLUT_VIDEO_RESIZE_X: + case GLUT_VIDEO_RESIZE_Y: + case GLUT_VIDEO_RESIZE_WIDTH: + case GLUT_VIDEO_RESIZE_HEIGHT: +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) + if (videoResizeInUse) { + int x, y, width, height; + + __glut_glXQueryChannelRectSGIX(__glutDisplay, __glutScreen, + videoResizeChannel, &x, &y, &width, &height); + switch (param) { + case GLUT_VIDEO_RESIZE_X: + return x; + case GLUT_VIDEO_RESIZE_Y: + return y; + case GLUT_VIDEO_RESIZE_WIDTH: + return width; + case GLUT_VIDEO_RESIZE_HEIGHT: + return height; + } + } +#endif + return -1; + default: + __glutWarning("invalid glutVideoResizeGet parameter: %d", param); + return -1; + } +} + +void GLUTAPIENTRY +glutSetupVideoResizing(void) +{ +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) + if (glutVideoResizeGet(GLUT_VIDEO_RESIZE_POSSIBLE)) { + __glut_glXBindChannelToWindowSGIX(__glutDisplay, __glutScreen, + videoResizeChannel, __glutCurrentWindow->win); + videoResizeInUse = 1; + } else +#endif + __glutFatalError("glutEstablishVideoResizing: video resizing not possible.\n"); +} + +void GLUTAPIENTRY +glutStopVideoResizing(void) +{ +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) + if (glutVideoResizeGet(GLUT_VIDEO_RESIZE_POSSIBLE)) { + if (videoResizeInUse) { + __glut_glXBindChannelToWindowSGIX(__glutDisplay, __glutScreen, + videoResizeChannel, None); + videoResizeInUse = 0; + } + } +#endif +} + +/* ARGSUSED */ +void GLUTAPIENTRY +glutVideoResize(int x, int y, int width, int height) +{ +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) + if (videoResizeInUse) { +#ifdef GLX_SYNC_SWAP_SGIX + /* glXChannelRectSyncSGIX introduced in a patch to IRIX + 6.2; the original unpatched IRIX 6.2 behavior is always + GLX_SYNC_SWAP_SGIX. */ + __glut_glXChannelRectSyncSGIX(__glutDisplay, __glutScreen, + videoResizeChannel, GLX_SYNC_SWAP_SGIX); +#endif + __glut_glXChannelRectSGIX(__glutDisplay, __glutScreen, + videoResizeChannel, x, y, width, height); + } +#endif +} + +/* ARGSUSED */ +void GLUTAPIENTRY +glutVideoPan(int x, int y, int width, int height) +{ +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_video_resize) + if (videoResizeInUse) { +#ifdef GLX_SYNC_FRAME_SGIX + /* glXChannelRectSyncSGIX introduced in a patch to IRIX + 6.2; the original unpatched IRIX 6.2 behavior is always + GLX_SYNC_SWAP_SGIX. We just ignore that we cannot + accomplish GLX_SYNC_FRAME_SGIX on IRIX unpatched 6.2; + this means you'd need a glutSwapBuffers to actually + realize the video resize. */ + __glut_glXChannelRectSyncSGIX(__glutDisplay, __glutScreen, + videoResizeChannel, GLX_SYNC_FRAME_SGIX); +#endif + __glut_glXChannelRectSGIX(__glutDisplay, __glutScreen, + videoResizeChannel, x, y, width, height); + } +#endif +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glut_warp.cpp b/src/glut/os2/glut_warp.cpp index 885e169d3f..4ef6f7bf39 100644 --- a/src/glut/os2/glut_warp.cpp +++ b/src/glut/os2/glut_warp.cpp @@ -1,29 +1,29 @@ - -/* Copyright (c) Mark J. Kilgard, 1996, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include -#include -#include - -#include "glutint.h" - -/* CENTRY */ -void GLUTAPIENTRY -glutWarpPointer(int x, int y) -{ -// XWarpPointer(__glutDisplay, None, __glutCurrentWindow->win, -// 0, 0, 0, 0, x, y); - POINTL point; - point.x = x; - point.y = y; - WinMapWindowPoints(__glutCurrentWindow->win,HWND_DESKTOP,&point,1); - WinSetPointerPos(HWND_DESKTOP, point.x, point.y); - - XFlush(__glutDisplay); -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1996, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include +#include +#include + +#include "glutint.h" + +/* CENTRY */ +void GLUTAPIENTRY +glutWarpPointer(int x, int y) +{ +// XWarpPointer(__glutDisplay, None, __glutCurrentWindow->win, +// 0, 0, 0, 0, x, y); + POINTL point; + point.x = x; + point.y = y; + WinMapWindowPoints(__glutCurrentWindow->win,HWND_DESKTOP,&point,1); + WinSetPointerPos(HWND_DESKTOP, point.x, point.y); + + XFlush(__glutDisplay); +} + +/* ENDCENTRY */ diff --git a/src/glut/os2/glutbitmap.h b/src/glut/os2/glutbitmap.h index dc1b61af19..9584bb129d 100644 --- a/src/glut/os2/glutbitmap.h +++ b/src/glut/os2/glutbitmap.h @@ -1,32 +1,32 @@ -#ifndef __glutbitmap_h__ -#define __glutbitmap_h__ - -/* Copyright (c) Mark J. Kilgard, 1994, 1998. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#define GLUT_NO_LIB_PRAGMA /* Avoid auto library linking when building - the GLUT library itself. */ -#include - -typedef struct { - const GLsizei width; - const GLsizei height; - const GLfloat xorig; - const GLfloat yorig; - const GLfloat advance; - const GLubyte *bitmap; -} BitmapCharRec, *BitmapCharPtr; - -typedef struct { - const char *name; - const int num_chars; - const int first; - const BitmapCharRec * const *ch; -} BitmapFontRec, *BitmapFontPtr; - -typedef void *GLUTbitmapFont; - -#endif /* __glutbitmap_h__ */ +#ifndef __glutbitmap_h__ +#define __glutbitmap_h__ + +/* Copyright (c) Mark J. Kilgard, 1994, 1998. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#define GLUT_NO_LIB_PRAGMA /* Avoid auto library linking when building + the GLUT library itself. */ +#include + +typedef struct { + const GLsizei width; + const GLsizei height; + const GLfloat xorig; + const GLfloat yorig; + const GLfloat advance; + const GLubyte *bitmap; +} BitmapCharRec, *BitmapCharPtr; + +typedef struct { + const char *name; + const int num_chars; + const int first; + const BitmapCharRec * const *ch; +} BitmapFontRec, *BitmapFontPtr; + +typedef void *GLUTbitmapFont; + +#endif /* __glutbitmap_h__ */ diff --git a/src/glut/os2/glutstroke.h b/src/glut/os2/glutstroke.h index fbbc70d6cc..cbc9e156fe 100644 --- a/src/glut/os2/glutstroke.h +++ b/src/glut/os2/glutstroke.h @@ -1,42 +1,42 @@ -#ifndef __glutstroke_h__ -#define __glutstroke_h__ - -/* Copyright (c) Mark J. Kilgard, 1994. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#if defined(_WIN32) -#pragma warning (disable:4244) /* disable bogus conversion warnings */ -#pragma warning (disable:4305) /* VC++ 5.0 version of above warning. */ -#endif - -typedef struct { - float x; - float y; -} CoordRec, *CoordPtr; - -typedef struct { - int num_coords; - const CoordRec *coord; -} StrokeRec, *StrokePtr; - -typedef struct { - int num_strokes; - const StrokeRec *stroke; - float center; - float right; -} StrokeCharRec, *StrokeCharPtr; - -typedef struct { - const char *name; - int num_chars; - const StrokeCharRec *ch; - float top; - float bottom; -} StrokeFontRec, *StrokeFontPtr; - -typedef void *GLUTstrokeFont; - -#endif /* __glutstroke_h__ */ +#ifndef __glutstroke_h__ +#define __glutstroke_h__ + +/* Copyright (c) Mark J. Kilgard, 1994. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#if defined(_WIN32) +#pragma warning (disable:4244) /* disable bogus conversion warnings */ +#pragma warning (disable:4305) /* VC++ 5.0 version of above warning. */ +#endif + +typedef struct { + float x; + float y; +} CoordRec, *CoordPtr; + +typedef struct { + int num_coords; + const CoordRec *coord; +} StrokeRec, *StrokePtr; + +typedef struct { + int num_strokes; + const StrokeRec *stroke; + float center; + float right; +} StrokeCharRec, *StrokeCharPtr; + +typedef struct { + const char *name; + int num_chars; + const StrokeCharRec *ch; + float top; + float bottom; +} StrokeFontRec, *StrokeFontPtr; + +typedef void *GLUTstrokeFont; + +#endif /* __glutstroke_h__ */ diff --git a/src/glut/os2/layerutil.h b/src/glut/os2/layerutil.h index 3786c4d871..3f62031939 100644 --- a/src/glut/os2/layerutil.h +++ b/src/glut/os2/layerutil.h @@ -1,59 +1,59 @@ -#ifndef __layerutil_h__ -#define __layerutil_h__ - -/* Copyright (c) Mark J. Kilgard, 1993, 1994. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -/* Based on XLayerUtil.h: Revision: 1.3 */ - -#ifdef __VMS -#include -#endif - -#if !defined(_WIN32) && !defined(__OS2__) -#include -#include -#include -#endif /* !_WIN32 */ - -/* Transparent type values */ -/* None 0 */ -#define TransparentPixel 1 -#define TransparentMask 2 - -/* layered visual info template flags */ -#define VisualLayerMask 0x200 -#define VisualTransparentType 0x400 -#define VisualTransparentValue 0x800 -#define VisualAllLayerMask 0xFFF - -/* layered visual info structure */ -typedef struct _XLayerVisualInfo { - XVisualInfo vinfo; - long layer; - long type; - unsigned long value; -} XLayerVisualInfo; - -/* SERVER_OVERLAY_VISUALS property element */ -typedef struct _OverlayInfo { - /* Avoid 64-bit portability problems by being careful to use - longs due to the way XGetWindowProperty is specified. Note - that these parameters are passed as CARD32s over X - protocol. */ - long overlay_visual; - long transparent_type; - long value; - long layer; -} OverlayInfo; - -extern int __glutGetTransparentPixel(Display *, XVisualInfo *); -extern XLayerVisualInfo *__glutXGetLayerVisualInfo(Display *, - long, XLayerVisualInfo *, int *); -extern Status __glutXMatchLayerVisualInfo(Display *, - int, int, int, int, XLayerVisualInfo *); - -#endif /* __layerutil_h__ */ +#ifndef __layerutil_h__ +#define __layerutil_h__ + +/* Copyright (c) Mark J. Kilgard, 1993, 1994. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +/* Based on XLayerUtil.h: Revision: 1.3 */ + +#ifdef __VMS +#include +#endif + +#if !defined(_WIN32) && !defined(__OS2__) +#include +#include +#include +#endif /* !_WIN32 */ + +/* Transparent type values */ +/* None 0 */ +#define TransparentPixel 1 +#define TransparentMask 2 + +/* layered visual info template flags */ +#define VisualLayerMask 0x200 +#define VisualTransparentType 0x400 +#define VisualTransparentValue 0x800 +#define VisualAllLayerMask 0xFFF + +/* layered visual info structure */ +typedef struct _XLayerVisualInfo { + XVisualInfo vinfo; + long layer; + long type; + unsigned long value; +} XLayerVisualInfo; + +/* SERVER_OVERLAY_VISUALS property element */ +typedef struct _OverlayInfo { + /* Avoid 64-bit portability problems by being careful to use + longs due to the way XGetWindowProperty is specified. Note + that these parameters are passed as CARD32s over X + protocol. */ + long overlay_visual; + long transparent_type; + long value; + long layer; +} OverlayInfo; + +extern int __glutGetTransparentPixel(Display *, XVisualInfo *); +extern XLayerVisualInfo *__glutXGetLayerVisualInfo(Display *, + long, XLayerVisualInfo *, int *); +extern Status __glutXMatchLayerVisualInfo(Display *, + int, int, int, int, XLayerVisualInfo *); + +#endif /* __layerutil_h__ */ diff --git a/src/glut/os2/libGlut.DEF b/src/glut/os2/libGlut.DEF index fb92544ab8..a0e184eef0 100644 --- a/src/glut/os2/libGlut.DEF +++ b/src/glut/os2/libGlut.DEF @@ -1,94 +1,94 @@ - -;----------------------------------------------------- -; def Module definition file for the DLL -;----------------------------------------------------- - -LIBRARY libGlut INITINSTANCE TERMINSTANCE -PROTMODE -DATA MULTIPLE NONSHARED READWRITE LOADONCALL -CODE LOADONCALL -EXPORTS ; Names of exported functions and data - -;********************************* -;MesaGLUT - glutCommandFunc @3001 - glutCreateWindow @3002 - glutCreateSubWindow @3003 - glutDestroyWindow @3004 - glutDisplayFunc @3005 - glutEntryFunc @3006 - glutExtensionSupported @3007 - glutIdleFunc @3008 - glutInit @3009 - glutInitDisplayMode @3010 - glutInitWindowPosition @3011 - glutInitWindowSize @3012 - glutMainLoop @3013 - glutGet @3014 - glutGetColor @3015 - glutGetWindow @3016 - glutKeyboardFunc @3017 - glutMouseFunc @3018 - glutSpecialFunc @3019 - glutStrokeCharacter @3020 - glutSetColor @3021 - glutSetIconTitle @3022 - glutSetWindow @3023 - glutSetWindowTitle @3024 - - glutReshapeFunc @3025 - glutReshapeWindow @3026 - - - glutSwapBuffers @3027 - glutPostRedisplay @3028 - glutPositionWindow @3029 - glutVisibilityFunc @3030 - glutTimerFunc @3031 - glutMotionFunc @3032 - - glutCreateMenu @3033 - glutAttachMenu @3034 - glutDestroyMenu @3035 - glutAddMenuEntry @3036 - glutPassiveMotionFunc @3037 - - glutSolidCone @3038 - glutSolidCube @3039 - glutSolidDodecahedron @3040 - glutSolidOctahedron @3041 - glutSolidIcosahedron @3042 - glutSolidSphere @3043 - glutSolidTeapot @3044 - glutSolidTetrahedron @3045 - glutSolidTorus @3046 - - glutWireCone @3047 - glutWireCube @3048 - glutWireDodecahedron @3049 - glutWireIcosahedron @3050 - glutWireOctahedron @3051 - glutWireSphere @3052 - glutWireTetrahedron @3053 - glutWireTorus @3054 - glutWireTeapot @3055 - -;GL_GLEXT_PROTOTYPES - glutBitmapCharacter @3101 - glutBitmap9By15 @3102 - glutBitmapHelvetica10 @3103 - glutBitmapHelvetica12 @3104 - glutBitmapHelvetica18 @3105 - glutBitmapTimesRoman24 @3106 - glutStrokeRoman @3107 - glutBitmap8By13 @3108 - -;Global constants -; hab @12001 -; /* PM anchor block handle */ -; hpsCurrent @12002 - XHDC @12004 - XHWND @12005 - - - + +;----------------------------------------------------- +; def Module definition file for the DLL +;----------------------------------------------------- + +LIBRARY libGlut INITINSTANCE TERMINSTANCE +PROTMODE +DATA MULTIPLE NONSHARED READWRITE LOADONCALL +CODE LOADONCALL +EXPORTS ; Names of exported functions and data + +;********************************* +;MesaGLUT + glutCommandFunc @3001 + glutCreateWindow @3002 + glutCreateSubWindow @3003 + glutDestroyWindow @3004 + glutDisplayFunc @3005 + glutEntryFunc @3006 + glutExtensionSupported @3007 + glutIdleFunc @3008 + glutInit @3009 + glutInitDisplayMode @3010 + glutInitWindowPosition @3011 + glutInitWindowSize @3012 + glutMainLoop @3013 + glutGet @3014 + glutGetColor @3015 + glutGetWindow @3016 + glutKeyboardFunc @3017 + glutMouseFunc @3018 + glutSpecialFunc @3019 + glutStrokeCharacter @3020 + glutSetColor @3021 + glutSetIconTitle @3022 + glutSetWindow @3023 + glutSetWindowTitle @3024 + + glutReshapeFunc @3025 + glutReshapeWindow @3026 + + + glutSwapBuffers @3027 + glutPostRedisplay @3028 + glutPositionWindow @3029 + glutVisibilityFunc @3030 + glutTimerFunc @3031 + glutMotionFunc @3032 + + glutCreateMenu @3033 + glutAttachMenu @3034 + glutDestroyMenu @3035 + glutAddMenuEntry @3036 + glutPassiveMotionFunc @3037 + + glutSolidCone @3038 + glutSolidCube @3039 + glutSolidDodecahedron @3040 + glutSolidOctahedron @3041 + glutSolidIcosahedron @3042 + glutSolidSphere @3043 + glutSolidTeapot @3044 + glutSolidTetrahedron @3045 + glutSolidTorus @3046 + + glutWireCone @3047 + glutWireCube @3048 + glutWireDodecahedron @3049 + glutWireIcosahedron @3050 + glutWireOctahedron @3051 + glutWireSphere @3052 + glutWireTetrahedron @3053 + glutWireTorus @3054 + glutWireTeapot @3055 + +;GL_GLEXT_PROTOTYPES + glutBitmapCharacter @3101 + glutBitmap9By15 @3102 + glutBitmapHelvetica10 @3103 + glutBitmapHelvetica12 @3104 + glutBitmapHelvetica18 @3105 + glutBitmapTimesRoman24 @3106 + glutStrokeRoman @3107 + glutBitmap8By13 @3108 + +;Global constants +; hab @12001 +; /* PM anchor block handle */ +; hpsCurrent @12002 + XHDC @12004 + XHWND @12005 + + + diff --git a/src/mesa/drivers/allegro/amesa.c b/src/mesa/drivers/allegro/amesa.c index 518211ccf4..c12d42148b 100644 --- a/src/mesa/drivers/allegro/amesa.c +++ b/src/mesa/drivers/allegro/amesa.c @@ -1,414 +1,414 @@ -/* - * Mesa 3-D graphics library - * Version: 3.0 - * Copyright (C) 1995-1998 Brian Paul - * - * This library is free software; you can redistribute it and/or - * modify it under the terms of the GNU Library General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This library is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * Library General Public License for more details. - * - * You should have received a copy of the GNU Library General Public - * License along with this library; if not, write to the Free - * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. - */ - -#include -#include -#include -#include "buffers.h" -#include "context.h" -#include "imports.h" -#include "matrix.h" +/* + * Mesa 3-D graphics library + * Version: 3.0 + * Copyright (C) 1995-1998 Brian Paul + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Library General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Library General Public License for more details. + * + * You should have received a copy of the GNU Library General Public + * License along with this library; if not, write to the Free + * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#include +#include +#include +#include "buffers.h" +#include "context.h" +#include "imports.h" +#include "matrix.h" #include "mtypes.h" -#include "GL/amesa.h" - - -struct amesa_visual - { - GLvisual *GLVisual; /* inherit from GLvisual */ - GLboolean DBFlag; /* double buffered? */ - GLuint Depth; /* bits per pixel ( >= 15 ) */ - }; - - -struct amesa_buffer - { - GLframebuffer *GLBuffer; /* inherit from GLframebuffer */ - GLuint Width, Height; - BITMAP *Screen; - BITMAP *Background; - BITMAP *Active; - }; - - -struct amesa_context - { - GLcontext *GLContext; /* inherit from GLcontext */ - AMesaVisual Visual; - AMesaBuffer Buffer; - GLuint ClearColor; - GLuint CurrentColor; - }; - - -static void setup_dd_pointers(GLcontext *ctx); - - -/**********************************************************************/ -/***** drawing functions *****/ -/**********************************************************************/ - -#define FLIP(context, y) (context->Buffer->Height - (y) - 1) - -#include "allegro/generic.h" -#include "allegro/direct.h" - - -/**********************************************************************/ -/***** 15-bit accelerated drawing funcs *****/ -/**********************************************************************/ - -IMPLEMENT_WRITE_RGBA_SPAN(15, unsigned short) -IMPLEMENT_WRITE_RGB_SPAN(15, unsigned short) -IMPLEMENT_WRITE_MONO_RGBA_SPAN(15, unsigned short) -IMPLEMENT_READ_RGBA_SPAN(15, unsigned short) -IMPLEMENT_WRITE_RGBA_PIXELS(15, unsigned short) -IMPLEMENT_WRITE_MONO_RGBA_PIXELS(15, unsigned short) -IMPLEMENT_READ_RGBA_PIXELS(15, unsigned short) - - -/**********************************************************************/ -/***** 16-bit accelerated drawing funcs *****/ -/**********************************************************************/ - -IMPLEMENT_WRITE_RGBA_SPAN(16, unsigned short) -IMPLEMENT_WRITE_RGB_SPAN(16, unsigned short) -IMPLEMENT_WRITE_MONO_RGBA_SPAN(16, unsigned short) -IMPLEMENT_READ_RGBA_SPAN(16, unsigned short) -IMPLEMENT_WRITE_RGBA_PIXELS(16, unsigned short) -IMPLEMENT_WRITE_MONO_RGBA_PIXELS(16, unsigned short) -IMPLEMENT_READ_RGBA_PIXELS(16, unsigned short) - - -/**********************************************************************/ -/***** 32-bit accelerated drawing funcs *****/ -/**********************************************************************/ - -IMPLEMENT_WRITE_RGBA_SPAN(32, unsigned long) -IMPLEMENT_WRITE_RGB_SPAN(32, unsigned long) -IMPLEMENT_WRITE_MONO_RGBA_SPAN(32, unsigned long) -IMPLEMENT_READ_RGBA_SPAN(32, unsigned long) -IMPLEMENT_WRITE_RGBA_PIXELS(32, unsigned long) -IMPLEMENT_WRITE_MONO_RGBA_PIXELS(32, unsigned long) -IMPLEMENT_READ_RGBA_PIXELS(32, unsigned long) - - -/**********************************************************************/ -/***** Miscellaneous device driver funcs *****/ -/**********************************************************************/ - -static GLboolean set_buffer(GLcontext *ctx, GLframebuffer *buffer, GLuint bit) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - GLboolean ok = GL_TRUE; - - if (bit == DD_FRONT_LEFT_BIT) - context->Buffer->Active = context->Buffer->Screen; - - else if (bit == DD_BACK_LEFT) - { - if (context->Buffer->Background) - context->Buffer->Active = context->Buffer->Background; - else - ok = GL_FALSE; - } - - else - ok = GL_FALSE; - - return ok; - } - - -static void get_buffer_size(GLcontext *ctx, GLuint *width, GLuint *height) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - - *width = context->Buffer->Width; - *height = context->Buffer->Height; - } - - -/** - * We only implement this function as a mechanism to check if the - * framebuffer size has changed (and update corresponding state). - */ -static void viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) -{ - /* poll for window size change and realloc software Z/stencil/etc if needed */ - GLuint newWidth, newHeight; - GLframebuffer *buffer = ctx->WinSysDrawBuffer; - get_buffer_size( &newWidth, &newHeight ); - if (buffer->Width != newWidth || buffer->Height != newHeight) { - _mesa_resize_framebuffer(ctx, buffer, newWidth, newHeight ); - } - -} - - -/**********************************************************************/ -/**********************************************************************/ - -static void setup_dd_pointers(GLcontext *ctx) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - - /* Initialize all the pointers in the driver struct. Do this whenever */ - /* a new context is made current or we change buffers via set_buffer! */ - - ctx->Driver.UpdateState = setup_dd_pointers; - ctx->Driver.SetBuffer = set_buffer; - ctx->Driver.GetBufferSize = get_buffer_size; - ctx->Driver.Viewport = viewport; - - ctx->Driver.Color = set_color_generic; - ctx->Driver.ClearColor = clear_color_generic; - ctx->Driver.Clear = clear_generic; - ctx->Driver.WriteRGBASpan = write_rgba_span_generic; - ctx->Driver.WriteRGBSpan = write_rgb_span_generic; - ctx->Driver.WriteMonoRGBASpan = write_mono_rgba_span_generic; - ctx->Driver.WriteRGBAPixels = write_rgba_pixels_generic; - ctx->Driver.WriteMonoRGBAPixels = write_mono_rgba_pixels_generic; - ctx->Driver.ReadRGBASpan = read_rgba_span_generic; - ctx->Driver.ReadRGBAPixels = read_rgba_pixels_generic; - - if (context->Buffer->Active != screen) - { - switch (context->Visual->Depth) - { - case 15: - ctx->Driver.WriteRGBASpan = write_rgba_span_15; - ctx->Driver.WriteRGBSpan = write_rgb_span_15; - ctx->Driver.WriteMonoRGBASpan = write_mono_rgba_span_15; - ctx->Driver.WriteRGBAPixels = write_rgba_pixels_15; - ctx->Driver.WriteMonoRGBAPixels = write_mono_rgba_pixels_15; - ctx->Driver.ReadRGBASpan = read_rgba_span_15; - ctx->Driver.ReadRGBAPixels = read_rgba_pixels_15; - break; - - case 16: - ctx->Driver.WriteRGBASpan = write_rgba_span_16; - ctx->Driver.WriteRGBSpan = write_rgb_span_16; - ctx->Driver.WriteMonoRGBASpan = write_mono_rgba_span_16; - ctx->Driver.WriteRGBAPixels = write_rgba_pixels_16; - ctx->Driver.WriteMonoRGBAPixels = write_mono_rgba_pixels_16; - ctx->Driver.ReadRGBASpan = read_rgba_span_16; - ctx->Driver.ReadRGBAPixels = read_rgba_pixels_16; - break; - - case 32: - ctx->Driver.WriteRGBASpan = write_rgba_span_32; - ctx->Driver.WriteRGBSpan = write_rgb_span_32; - ctx->Driver.WriteMonoRGBASpan = write_mono_rgba_span_32; - ctx->Driver.WriteRGBAPixels = write_rgba_pixels_32; - ctx->Driver.WriteMonoRGBAPixels = write_mono_rgba_pixels_32; - ctx->Driver.ReadRGBASpan = read_rgba_span_32; - ctx->Driver.ReadRGBAPixels = read_rgba_pixels_32; - break; - } - } - } - - -/**********************************************************************/ -/***** AMesa Public API Functions *****/ -/**********************************************************************/ - - -AMesaVisual AMesaCreateVisual(GLboolean dbFlag, GLint depth, - GLint depthSize, GLint stencilSize, GLint accumSize) - { - AMesaVisual visual; - GLbyte redBits, greenBits, blueBits; - - visual = (AMesaVisual)calloc(1, sizeof(struct amesa_visual)); - if (!visual) - return NULL; - - switch (depth) - { - case 15: - redBits = 5; - greenBits = 5; - blueBits = 5; - break; - - case 16: - redBits = 5; - greenBits = 6; - blueBits = 5; - break; - - case 24: case 32: - redBits = 8; - greenBits = 8; - blueBits = 8; - break; - - default: - free(visual); - return NULL; - } - - visual->DBFlag = dbFlag; - visual->Depth = depth; - visual->GLVisual = _mesa_create_visual(GL_TRUE, /* rgb mode */ - dbFlag, /* db_flag */ - GL_FALSE, /* stereo */ - redBits, greenBits, blueBits, 8, - 0, /* index bits */ - depthSize, /* depth bits */ - stencilSize,/* stencil bits */ - accumSize, /* accum bits */ - accumSize, /* accum bits */ - accumSize, /* accum bits */ - accumSize, /* accum bits */ - 1 ); - if (!visual->GLVisual) - { - free(visual); - return NULL; - } - - return visual; - } - - -void AMesaDestroyVisual(AMesaVisual visual) - { - _mesa_destroy_visual(visual->GLVisual); - free(visual); - } - - -AMesaBuffer AMesaCreateBuffer(AMesaVisual visual, - GLint width, GLint height) - { - AMesaBuffer buffer; - - buffer = (AMesaBuffer)calloc(1, sizeof(struct amesa_buffer)); - if (!buffer) - return NULL; - - buffer->Screen = NULL; - buffer->Background = NULL; - buffer->Active = NULL; - buffer->Width = width; - buffer->Height = height; - - if (visual->DBFlag) - { - buffer->Background = create_bitmap_ex(visual->Depth, width, height); - if (!buffer->Background) - { - free(buffer); - return NULL; - } - } - - buffer->GLBuffer = _mesa_create_framebuffer(visual->GLVisual); - if (!buffer->GLBuffer) - { - if (buffer->Background) destroy_bitmap(buffer->Background); - free(buffer); - return NULL; - } - - return buffer; - } - - -void AMesaDestroyBuffer(AMesaBuffer buffer) -{ - if (buffer->Screen) destroy_bitmap(buffer->Screen); - if (buffer->Background) destroy_bitmap(buffer->Background); - _mesa_unreference_framebuffer(&buffer->GLBuffer); - free(buffer); -} - - -AMesaContext AMesaCreateContext(AMesaVisual visual, - AMesaContext share) -{ - AMesaContext context; - GLboolean direct = GL_FALSE; - - context = (AMesaContext)calloc(1, sizeof(struct amesa_context)); - if (!context) - return NULL; - - context->Visual = visual; - context->Buffer = NULL; - context->ClearColor = 0; - context->CurrentColor = 0; - context->GLContext = _mesa_create_context(visual->GLVisual, - share ? share->GLContext : NULL, - (void *) context, GL_FALSE ); - if (!context->GLContext) - { - free(context); - return NULL; - } - - return context; -} - - -void AMesaDestroyContext(AMesaContext context) -{ - _mesa_destroy_context(context->GLContext); - free(context); -} - - -GLboolean AMesaMakeCurrent(AMesaContext context, AMesaBuffer buffer) -{ - if (context && buffer) { - set_color_depth(context->Visual->Depth); - if (set_gfx_mode(GFX_AUTODETECT, buffer->Width, buffer->Height, 0, 0) != 0) - return GL_FALSE; - - context->Buffer = buffer; - buffer->Screen = screen; - buffer->Active = buffer->Background ? buffer->Background : screen; - - setup_dd_pointers(context->GLContext); - _mesa_make_current(context->GLContext, buffer->GLBuffer); - } - else { +#include "GL/amesa.h" + + +struct amesa_visual + { + GLvisual *GLVisual; /* inherit from GLvisual */ + GLboolean DBFlag; /* double buffered? */ + GLuint Depth; /* bits per pixel ( >= 15 ) */ + }; + + +struct amesa_buffer + { + GLframebuffer *GLBuffer; /* inherit from GLframebuffer */ + GLuint Width, Height; + BITMAP *Screen; + BITMAP *Background; + BITMAP *Active; + }; + + +struct amesa_context + { + GLcontext *GLContext; /* inherit from GLcontext */ + AMesaVisual Visual; + AMesaBuffer Buffer; + GLuint ClearColor; + GLuint CurrentColor; + }; + + +static void setup_dd_pointers(GLcontext *ctx); + + +/**********************************************************************/ +/***** drawing functions *****/ +/**********************************************************************/ + +#define FLIP(context, y) (context->Buffer->Height - (y) - 1) + +#include "allegro/generic.h" +#include "allegro/direct.h" + + +/**********************************************************************/ +/***** 15-bit accelerated drawing funcs *****/ +/**********************************************************************/ + +IMPLEMENT_WRITE_RGBA_SPAN(15, unsigned short) +IMPLEMENT_WRITE_RGB_SPAN(15, unsigned short) +IMPLEMENT_WRITE_MONO_RGBA_SPAN(15, unsigned short) +IMPLEMENT_READ_RGBA_SPAN(15, unsigned short) +IMPLEMENT_WRITE_RGBA_PIXELS(15, unsigned short) +IMPLEMENT_WRITE_MONO_RGBA_PIXELS(15, unsigned short) +IMPLEMENT_READ_RGBA_PIXELS(15, unsigned short) + + +/**********************************************************************/ +/***** 16-bit accelerated drawing funcs *****/ +/**********************************************************************/ + +IMPLEMENT_WRITE_RGBA_SPAN(16, unsigned short) +IMPLEMENT_WRITE_RGB_SPAN(16, unsigned short) +IMPLEMENT_WRITE_MONO_RGBA_SPAN(16, unsigned short) +IMPLEMENT_READ_RGBA_SPAN(16, unsigned short) +IMPLEMENT_WRITE_RGBA_PIXELS(16, unsigned short) +IMPLEMENT_WRITE_MONO_RGBA_PIXELS(16, unsigned short) +IMPLEMENT_READ_RGBA_PIXELS(16, unsigned short) + + +/**********************************************************************/ +/***** 32-bit accelerated drawing funcs *****/ +/**********************************************************************/ + +IMPLEMENT_WRITE_RGBA_SPAN(32, unsigned long) +IMPLEMENT_WRITE_RGB_SPAN(32, unsigned long) +IMPLEMENT_WRITE_MONO_RGBA_SPAN(32, unsigned long) +IMPLEMENT_READ_RGBA_SPAN(32, unsigned long) +IMPLEMENT_WRITE_RGBA_PIXELS(32, unsigned long) +IMPLEMENT_WRITE_MONO_RGBA_PIXELS(32, unsigned long) +IMPLEMENT_READ_RGBA_PIXELS(32, unsigned long) + + +/**********************************************************************/ +/***** Miscellaneous device driver funcs *****/ +/**********************************************************************/ + +static GLboolean set_buffer(GLcontext *ctx, GLframebuffer *buffer, GLuint bit) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + GLboolean ok = GL_TRUE; + + if (bit == DD_FRONT_LEFT_BIT) + context->Buffer->Active = context->Buffer->Screen; + + else if (bit == DD_BACK_LEFT) + { + if (context->Buffer->Background) + context->Buffer->Active = context->Buffer->Background; + else + ok = GL_FALSE; + } + + else + ok = GL_FALSE; + + return ok; + } + + +static void get_buffer_size(GLcontext *ctx, GLuint *width, GLuint *height) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + + *width = context->Buffer->Width; + *height = context->Buffer->Height; + } + + +/** + * We only implement this function as a mechanism to check if the + * framebuffer size has changed (and update corresponding state). + */ +static void viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h) +{ + /* poll for window size change and realloc software Z/stencil/etc if needed */ + GLuint newWidth, newHeight; + GLframebuffer *buffer = ctx->WinSysDrawBuffer; + get_buffer_size( &newWidth, &newHeight ); + if (buffer->Width != newWidth || buffer->Height != newHeight) { + _mesa_resize_framebuffer(ctx, buffer, newWidth, newHeight ); + } + +} + + +/**********************************************************************/ +/**********************************************************************/ + +static void setup_dd_pointers(GLcontext *ctx) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + + /* Initialize all the pointers in the driver struct. Do this whenever */ + /* a new context is made current or we change buffers via set_buffer! */ + + ctx->Driver.UpdateState = setup_dd_pointers; + ctx->Driver.SetBuffer = set_buffer; + ctx->Driver.GetBufferSize = get_buffer_size; + ctx->Driver.Viewport = viewport; + + ctx->Driver.Color = set_color_generic; + ctx->Driver.ClearColor = clear_color_generic; + ctx->Driver.Clear = clear_generic; + ctx->Driver.WriteRGBASpan = write_rgba_span_generic; + ctx->Driver.WriteRGBSpan = write_rgb_span_generic; + ctx->Driver.WriteMonoRGBASpan = write_mono_rgba_span_generic; + ctx->Driver.WriteRGBAPixels = write_rgba_pixels_generic; + ctx->Driver.WriteMonoRGBAPixels = write_mono_rgba_pixels_generic; + ctx->Driver.ReadRGBASpan = read_rgba_span_generic; + ctx->Driver.ReadRGBAPixels = read_rgba_pixels_generic; + + if (context->Buffer->Active != screen) + { + switch (context->Visual->Depth) + { + case 15: + ctx->Driver.WriteRGBASpan = write_rgba_span_15; + ctx->Driver.WriteRGBSpan = write_rgb_span_15; + ctx->Driver.WriteMonoRGBASpan = write_mono_rgba_span_15; + ctx->Driver.WriteRGBAPixels = write_rgba_pixels_15; + ctx->Driver.WriteMonoRGBAPixels = write_mono_rgba_pixels_15; + ctx->Driver.ReadRGBASpan = read_rgba_span_15; + ctx->Driver.ReadRGBAPixels = read_rgba_pixels_15; + break; + + case 16: + ctx->Driver.WriteRGBASpan = write_rgba_span_16; + ctx->Driver.WriteRGBSpan = write_rgb_span_16; + ctx->Driver.WriteMonoRGBASpan = write_mono_rgba_span_16; + ctx->Driver.WriteRGBAPixels = write_rgba_pixels_16; + ctx->Driver.WriteMonoRGBAPixels = write_mono_rgba_pixels_16; + ctx->Driver.ReadRGBASpan = read_rgba_span_16; + ctx->Driver.ReadRGBAPixels = read_rgba_pixels_16; + break; + + case 32: + ctx->Driver.WriteRGBASpan = write_rgba_span_32; + ctx->Driver.WriteRGBSpan = write_rgb_span_32; + ctx->Driver.WriteMonoRGBASpan = write_mono_rgba_span_32; + ctx->Driver.WriteRGBAPixels = write_rgba_pixels_32; + ctx->Driver.WriteMonoRGBAPixels = write_mono_rgba_pixels_32; + ctx->Driver.ReadRGBASpan = read_rgba_span_32; + ctx->Driver.ReadRGBAPixels = read_rgba_pixels_32; + break; + } + } + } + + +/**********************************************************************/ +/***** AMesa Public API Functions *****/ +/**********************************************************************/ + + +AMesaVisual AMesaCreateVisual(GLboolean dbFlag, GLint depth, + GLint depthSize, GLint stencilSize, GLint accumSize) + { + AMesaVisual visual; + GLbyte redBits, greenBits, blueBits; + + visual = (AMesaVisual)calloc(1, sizeof(struct amesa_visual)); + if (!visual) + return NULL; + + switch (depth) + { + case 15: + redBits = 5; + greenBits = 5; + blueBits = 5; + break; + + case 16: + redBits = 5; + greenBits = 6; + blueBits = 5; + break; + + case 24: case 32: + redBits = 8; + greenBits = 8; + blueBits = 8; + break; + + default: + free(visual); + return NULL; + } + + visual->DBFlag = dbFlag; + visual->Depth = depth; + visual->GLVisual = _mesa_create_visual(GL_TRUE, /* rgb mode */ + dbFlag, /* db_flag */ + GL_FALSE, /* stereo */ + redBits, greenBits, blueBits, 8, + 0, /* index bits */ + depthSize, /* depth bits */ + stencilSize,/* stencil bits */ + accumSize, /* accum bits */ + accumSize, /* accum bits */ + accumSize, /* accum bits */ + accumSize, /* accum bits */ + 1 ); + if (!visual->GLVisual) + { + free(visual); + return NULL; + } + + return visual; + } + + +void AMesaDestroyVisual(AMesaVisual visual) + { + _mesa_destroy_visual(visual->GLVisual); + free(visual); + } + + +AMesaBuffer AMesaCreateBuffer(AMesaVisual visual, + GLint width, GLint height) + { + AMesaBuffer buffer; + + buffer = (AMesaBuffer)calloc(1, sizeof(struct amesa_buffer)); + if (!buffer) + return NULL; + + buffer->Screen = NULL; + buffer->Background = NULL; + buffer->Active = NULL; + buffer->Width = width; + buffer->Height = height; + + if (visual->DBFlag) + { + buffer->Background = create_bitmap_ex(visual->Depth, width, height); + if (!buffer->Background) + { + free(buffer); + return NULL; + } + } + + buffer->GLBuffer = _mesa_create_framebuffer(visual->GLVisual); + if (!buffer->GLBuffer) + { + if (buffer->Background) destroy_bitmap(buffer->Background); + free(buffer); + return NULL; + } + + return buffer; + } + + +void AMesaDestroyBuffer(AMesaBuffer buffer) +{ + if (buffer->Screen) destroy_bitmap(buffer->Screen); + if (buffer->Background) destroy_bitmap(buffer->Background); + _mesa_unreference_framebuffer(&buffer->GLBuffer); + free(buffer); +} + + +AMesaContext AMesaCreateContext(AMesaVisual visual, + AMesaContext share) +{ + AMesaContext context; + GLboolean direct = GL_FALSE; + + context = (AMesaContext)calloc(1, sizeof(struct amesa_context)); + if (!context) + return NULL; + + context->Visual = visual; + context->Buffer = NULL; + context->ClearColor = 0; + context->CurrentColor = 0; + context->GLContext = _mesa_create_context(visual->GLVisual, + share ? share->GLContext : NULL, + (void *) context, GL_FALSE ); + if (!context->GLContext) + { + free(context); + return NULL; + } + + return context; +} + + +void AMesaDestroyContext(AMesaContext context) +{ + _mesa_destroy_context(context->GLContext); + free(context); +} + + +GLboolean AMesaMakeCurrent(AMesaContext context, AMesaBuffer buffer) +{ + if (context && buffer) { + set_color_depth(context->Visual->Depth); + if (set_gfx_mode(GFX_AUTODETECT, buffer->Width, buffer->Height, 0, 0) != 0) + return GL_FALSE; + + context->Buffer = buffer; + buffer->Screen = screen; + buffer->Active = buffer->Background ? buffer->Background : screen; + + setup_dd_pointers(context->GLContext); + _mesa_make_current(context->GLContext, buffer->GLBuffer); + } + else { /* XXX I don't think you want to destroy anything here! */ - destroy_bitmap(context->Buffer->Screen); - context->Buffer->Screen = NULL; - context->Buffer->Active = NULL; - context->Buffer = NULL; - _mesa_make_current(NULL, NULL); - } - - return GL_TRUE; -} - - -void AMesaSwapBuffers(AMesaBuffer buffer) -{ - if (buffer->Background) { - blit(buffer->Background, buffer->Screen, - 0, 0, 0, 0, - buffer->Width, buffer->Height); - } -} + destroy_bitmap(context->Buffer->Screen); + context->Buffer->Screen = NULL; + context->Buffer->Active = NULL; + context->Buffer = NULL; + _mesa_make_current(NULL, NULL); + } + + return GL_TRUE; +} + + +void AMesaSwapBuffers(AMesaBuffer buffer) +{ + if (buffer->Background) { + blit(buffer->Background, buffer->Screen, + 0, 0, 0, 0, + buffer->Width, buffer->Height); + } +} diff --git a/src/mesa/drivers/allegro/direct.h b/src/mesa/drivers/allegro/direct.h index 3998fc19d7..bd8b5eb49d 100644 --- a/src/mesa/drivers/allegro/direct.h +++ b/src/mesa/drivers/allegro/direct.h @@ -1,189 +1,189 @@ -/* - * Mesa 3-D graphics library - * Version: 3.0 - * Copyright (C) 1995-1998 Brian Paul - * - * This library is free software; you can redistribute it and/or - * modify it under the terms of the GNU Library General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This library is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * Library General Public License for more details. - * - * You should have received a copy of the GNU Library General Public - * License along with this library; if not, write to the Free - * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. - */ - - -#define DESTINATION(BMP, X, Y, TYPE) \ - ({ \ - BITMAP *_bmp = BMP; \ - \ - (TYPE*)(_bmp->line[_bmp->h - (Y) - 1]) + (X); \ - }) - - -#define IMPLEMENT_WRITE_RGBA_SPAN(DEPTH, TYPE) \ -static void write_rgba_span_##DEPTH (const GLcontext *ctx, \ - GLuint n, GLint x, GLint y, \ - const GLubyte rgba[][4], \ - const GLubyte mask[]) \ - { \ - AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ - TYPE *d = DESTINATION(context->Buffer->Active, x, y, TYPE); \ - \ - if (mask) \ - { \ - while (n--) \ - { \ - if (mask[0]) d[0] = makecol##DEPTH(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP]); \ - d++; rgba++; mask++; \ - } \ - } \ - else \ - { \ - while (n--) \ - { \ - d[0] = makecol##DEPTH(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP]); \ - d++; rgba++; \ - } \ - } \ - } - - -#define IMPLEMENT_WRITE_RGB_SPAN(DEPTH, TYPE) \ -static void write_rgb_span_##DEPTH (const GLcontext *ctx, \ - GLuint n, GLint x, GLint y, \ - const GLubyte rgb[][3], \ - const GLubyte mask[]) \ - { \ - AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ - TYPE *d = DESTINATION(context->Buffer->Active, x, y, TYPE); \ - \ - if (mask) \ - { \ - while (n--) \ - { \ - if (mask[0]) d[0] = makecol##DEPTH(rgb[0][RCOMP], rgb[0][GCOMP], rgb[0][BCOMP]); \ - d++; rgb++; mask++; \ - } \ - } \ - else \ - { \ - while (n--) \ - { \ - d[0] = makecol##DEPTH(rgb[0][RCOMP], rgb[0][GCOMP], rgb[0][BCOMP]); \ - d++; rgb++; \ - } \ - } \ - } - - -#define IMPLEMENT_WRITE_MONO_RGBA_SPAN(DEPTH, TYPE) \ -static void write_mono_rgba_span_##DEPTH (const GLcontext *ctx, \ - GLuint n, GLint x, GLint y, \ - const GLubyte mask[]) \ - { \ - AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ - TYPE color = context->CurrentColor; \ - TYPE *d = DESTINATION(context->Buffer->Active, x, y, TYPE); \ - \ - while (n--) \ - { \ - if (mask[0]) d[0] = color; \ - d++; mask++; \ - } \ - } - - -#define IMPLEMENT_READ_RGBA_SPAN(DEPTH, TYPE) \ -static void read_rgba_span_##DEPTH (const GLcontext *ctx, \ - GLuint n, GLint x, GLint y, \ - GLubyte rgba[][4]) \ - { \ - AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ - BITMAP *bmp = context->Buffer->Active; \ - TYPE *d = DESTINATION(bmp, x, y, TYPE); \ - \ - while (n--) \ - { \ - rgba[0][RCOMP] = getr##DEPTH(d[0]); \ - rgba[0][GCOMP] = getg##DEPTH(d[0]); \ - rgba[0][BCOMP] = getb##DEPTH(d[0]); \ - rgba[0][ACOMP] = 255; \ - \ - d++; rgba++; \ - } \ - } - - -#define IMPLEMENT_WRITE_RGBA_PIXELS(DEPTH, TYPE) \ -static void write_rgba_pixels_##DEPTH (const GLcontext *ctx, \ - GLuint n, \ - const GLint x[], \ - const GLint y[], \ - const GLubyte rgba[][4], \ - const GLubyte mask[]) \ - { \ - AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ - BITMAP *bmp = context->Buffer->Active; \ - \ - while (n--) \ - { \ - if (mask[0]) *DESTINATION(bmp, x[0], y[0], TYPE) = makecol##DEPTH(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP]); \ - rgba++; x++; y++; mask++; \ - } \ - } - - - -#define IMPLEMENT_WRITE_MONO_RGBA_PIXELS(DEPTH, TYPE) \ -static void write_mono_rgba_pixels_##DEPTH (const GLcontext *ctx, \ - GLuint n, \ - const GLint x[], \ - const GLint y[], \ - const GLubyte mask[]) \ - { \ - AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ - TYPE color = context->CurrentColor; \ - BITMAP *bmp = context->Buffer->Active; \ - \ - while (n--) \ - { \ - if (mask[0]) *DESTINATION(bmp, x[0], y[0], TYPE) = color; \ - x++; y++; mask++; \ - } \ - } - - -#define IMPLEMENT_READ_RGBA_PIXELS(DEPTH, TYPE) \ -static void read_rgba_pixels_##DEPTH (const GLcontext *ctx, \ - GLuint n, \ - const GLint x[], \ - const GLint y[], \ - GLubyte rgba[][4], \ - const GLubyte mask[]) \ - { \ - AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ - BITMAP *bmp = context->Buffer->Active; \ - \ - while (n--) \ - { \ - if (mask[0]) \ - { \ - int color = *DESTINATION(bmp, x[0], y[0], TYPE); \ - \ - rgba[0][RCOMP] = getr##DEPTH(color); \ - rgba[0][GCOMP] = getg##DEPTH(color); \ - rgba[0][BCOMP] = getb##DEPTH(color); \ - rgba[0][ACOMP] = 255; \ - } \ - \ - x++; y++; rgba++; mask++; \ - } \ - } - +/* + * Mesa 3-D graphics library + * Version: 3.0 + * Copyright (C) 1995-1998 Brian Paul + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Library General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Library General Public License for more details. + * + * You should have received a copy of the GNU Library General Public + * License along with this library; if not, write to the Free + * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + + +#define DESTINATION(BMP, X, Y, TYPE) \ + ({ \ + BITMAP *_bmp = BMP; \ + \ + (TYPE*)(_bmp->line[_bmp->h - (Y) - 1]) + (X); \ + }) + + +#define IMPLEMENT_WRITE_RGBA_SPAN(DEPTH, TYPE) \ +static void write_rgba_span_##DEPTH (const GLcontext *ctx, \ + GLuint n, GLint x, GLint y, \ + const GLubyte rgba[][4], \ + const GLubyte mask[]) \ + { \ + AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ + TYPE *d = DESTINATION(context->Buffer->Active, x, y, TYPE); \ + \ + if (mask) \ + { \ + while (n--) \ + { \ + if (mask[0]) d[0] = makecol##DEPTH(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP]); \ + d++; rgba++; mask++; \ + } \ + } \ + else \ + { \ + while (n--) \ + { \ + d[0] = makecol##DEPTH(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP]); \ + d++; rgba++; \ + } \ + } \ + } + + +#define IMPLEMENT_WRITE_RGB_SPAN(DEPTH, TYPE) \ +static void write_rgb_span_##DEPTH (const GLcontext *ctx, \ + GLuint n, GLint x, GLint y, \ + const GLubyte rgb[][3], \ + const GLubyte mask[]) \ + { \ + AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ + TYPE *d = DESTINATION(context->Buffer->Active, x, y, TYPE); \ + \ + if (mask) \ + { \ + while (n--) \ + { \ + if (mask[0]) d[0] = makecol##DEPTH(rgb[0][RCOMP], rgb[0][GCOMP], rgb[0][BCOMP]); \ + d++; rgb++; mask++; \ + } \ + } \ + else \ + { \ + while (n--) \ + { \ + d[0] = makecol##DEPTH(rgb[0][RCOMP], rgb[0][GCOMP], rgb[0][BCOMP]); \ + d++; rgb++; \ + } \ + } \ + } + + +#define IMPLEMENT_WRITE_MONO_RGBA_SPAN(DEPTH, TYPE) \ +static void write_mono_rgba_span_##DEPTH (const GLcontext *ctx, \ + GLuint n, GLint x, GLint y, \ + const GLubyte mask[]) \ + { \ + AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ + TYPE color = context->CurrentColor; \ + TYPE *d = DESTINATION(context->Buffer->Active, x, y, TYPE); \ + \ + while (n--) \ + { \ + if (mask[0]) d[0] = color; \ + d++; mask++; \ + } \ + } + + +#define IMPLEMENT_READ_RGBA_SPAN(DEPTH, TYPE) \ +static void read_rgba_span_##DEPTH (const GLcontext *ctx, \ + GLuint n, GLint x, GLint y, \ + GLubyte rgba[][4]) \ + { \ + AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ + BITMAP *bmp = context->Buffer->Active; \ + TYPE *d = DESTINATION(bmp, x, y, TYPE); \ + \ + while (n--) \ + { \ + rgba[0][RCOMP] = getr##DEPTH(d[0]); \ + rgba[0][GCOMP] = getg##DEPTH(d[0]); \ + rgba[0][BCOMP] = getb##DEPTH(d[0]); \ + rgba[0][ACOMP] = 255; \ + \ + d++; rgba++; \ + } \ + } + + +#define IMPLEMENT_WRITE_RGBA_PIXELS(DEPTH, TYPE) \ +static void write_rgba_pixels_##DEPTH (const GLcontext *ctx, \ + GLuint n, \ + const GLint x[], \ + const GLint y[], \ + const GLubyte rgba[][4], \ + const GLubyte mask[]) \ + { \ + AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ + BITMAP *bmp = context->Buffer->Active; \ + \ + while (n--) \ + { \ + if (mask[0]) *DESTINATION(bmp, x[0], y[0], TYPE) = makecol##DEPTH(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP]); \ + rgba++; x++; y++; mask++; \ + } \ + } + + + +#define IMPLEMENT_WRITE_MONO_RGBA_PIXELS(DEPTH, TYPE) \ +static void write_mono_rgba_pixels_##DEPTH (const GLcontext *ctx, \ + GLuint n, \ + const GLint x[], \ + const GLint y[], \ + const GLubyte mask[]) \ + { \ + AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ + TYPE color = context->CurrentColor; \ + BITMAP *bmp = context->Buffer->Active; \ + \ + while (n--) \ + { \ + if (mask[0]) *DESTINATION(bmp, x[0], y[0], TYPE) = color; \ + x++; y++; mask++; \ + } \ + } + + +#define IMPLEMENT_READ_RGBA_PIXELS(DEPTH, TYPE) \ +static void read_rgba_pixels_##DEPTH (const GLcontext *ctx, \ + GLuint n, \ + const GLint x[], \ + const GLint y[], \ + GLubyte rgba[][4], \ + const GLubyte mask[]) \ + { \ + AMesaContext context = (AMesaContext)(ctx->DriverCtx); \ + BITMAP *bmp = context->Buffer->Active; \ + \ + while (n--) \ + { \ + if (mask[0]) \ + { \ + int color = *DESTINATION(bmp, x[0], y[0], TYPE); \ + \ + rgba[0][RCOMP] = getr##DEPTH(color); \ + rgba[0][GCOMP] = getg##DEPTH(color); \ + rgba[0][BCOMP] = getb##DEPTH(color); \ + rgba[0][ACOMP] = 255; \ + } \ + \ + x++; y++; rgba++; mask++; \ + } \ + } + diff --git a/src/mesa/drivers/allegro/generic.h b/src/mesa/drivers/allegro/generic.h index cbdf5ff2b3..4c8af9b95c 100644 --- a/src/mesa/drivers/allegro/generic.h +++ b/src/mesa/drivers/allegro/generic.h @@ -1,234 +1,234 @@ -/* - * Mesa 3-D graphics library - * Version: 3.0 - * Copyright (C) 1995-1998 Brian Paul - * - * This library is free software; you can redistribute it and/or - * modify it under the terms of the GNU Library General Public - * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. - * - * This library is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU - * Library General Public License for more details. - * - * You should have received a copy of the GNU Library General Public - * License along with this library; if not, write to the Free - * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. - */ - -static void clear_color_generic(GLcontext *ctx, const GLfloat color[4]) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - GLubyte r, g, b; - CLAMPED_FLOAT_TO_UBYTE(r, color[0]); - CLAMPED_FLOAT_TO_UBYTE(g, color[1]); - CLAMPED_FLOAT_TO_UBYTE(b, color[2]); - context->ClearColor = makecol(r, g, b); - } - - -static void set_color_generic(GLcontext *ctx, - GLubyte red, GLubyte green, - GLubyte blue, GLubyte alpha) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - - context->CurrentColor = makecol(red, green, blue); - } - - -static GLbitfield clear_generic(GLcontext *ctx, - GLbitfield mask, GLboolean all, - GLint x, GLint y, - GLint width, GLint height) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - - if (mask & GL_COLOR_BUFFER_BIT) - { - if (all) - clear_to_color(context->Buffer->Active, context->ClearColor); - else - rect(context->Buffer->Active, - x, y, x+width-1, y+height-1, - context->ClearColor); - } - - return mask & (~GL_COLOR_BUFFER_BIT); - } - - -static void write_rgba_span_generic(const GLcontext *ctx, - GLuint n, GLint x, GLint y, - const GLubyte rgba[][4], - const GLubyte mask[]) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - BITMAP *bmp = context->Buffer->Active; - - y = FLIP(context, y); - - if (mask) - { - while (n--) - { - if (mask[0]) putpixel(bmp, x, y, makecol(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP])); - x++; mask++; rgba++; - } - } - else - { - while (n--) - { - putpixel(bmp, x, y, makecol(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP])); - x++; rgba++; - } - } - } - - -static void write_rgb_span_generic(const GLcontext *ctx, - GLuint n, GLint x, GLint y, - const GLubyte rgb[][3], - const GLubyte mask[]) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - BITMAP *bmp = context->Buffer->Active; - - y = FLIP(context, y); - - if (mask) - { - while(n--) - { - if (mask[0]) putpixel(bmp, x, y, makecol(rgb[0][RCOMP], rgb[0][GCOMP], rgb[0][BCOMP])); - x++; mask++; rgb++; - } - } - else - { - while (n--) - { - putpixel(bmp, x, y, makecol(rgb[0][RCOMP], rgb[0][GCOMP], rgb[0][BCOMP])); - x++; rgb++; - } - } - } - - -static void write_mono_rgba_span_generic(const GLcontext *ctx, - GLuint n, GLint x, GLint y, - const GLubyte mask[]) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - BITMAP *bmp = context->Buffer->Active; - int color = context->CurrentColor; - - y = FLIP(context, y); - - if (mask) - { - while(n--) - { - if (mask[0]) putpixel(bmp, x, y, color); - x++; mask++; - } - } - else - { - while(n--) - { - putpixel(bmp, x, y, color); - x++; - } - } - } - - -static void read_rgba_span_generic(const GLcontext *ctx, - GLuint n, GLint x, GLint y, - GLubyte rgba[][4]) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - BITMAP *bmp = context->Buffer->Active; - - y = FLIP(context, y); - - while (n--) - { - int color = getpixel(bmp, x, y); - - rgba[0][RCOMP] = getr(color); - rgba[0][GCOMP] = getg(color); - rgba[0][BCOMP] = getb(color); - rgba[0][ACOMP] = 255; - - x++; rgba++; - } - } - - -static void write_rgba_pixels_generic(const GLcontext *ctx, - GLuint n, - const GLint x[], - const GLint y[], - const GLubyte rgba[][4], - const GLubyte mask[]) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - BITMAP *bmp = context->Buffer->Active; - - while (n--) - { - if (mask[0]) putpixel(bmp, x[0], FLIP(context, y[0]), makecol(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP])); - x++; y++; mask++; - } - } - - -static void write_mono_rgba_pixels_generic(const GLcontext *ctx, - GLuint n, - const GLint x[], - const GLint y[], - const GLubyte mask[]) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - BITMAP *bmp = context->Buffer->Active; - int color = context->CurrentColor; - - while (n--) - { - if (mask[0]) putpixel(bmp, x[0], FLIP(context, y[0]), color); - x++; y++; mask++; - } - } - - -static void read_rgba_pixels_generic(const GLcontext *ctx, - GLuint n, - const GLint x[], - const GLint y[], - GLubyte rgba[][4], - const GLubyte mask[]) - { - AMesaContext context = (AMesaContext)(ctx->DriverCtx); - BITMAP *bmp = context->Buffer->Active; - - while (n--) - { - if (mask[0]) - { - int color = getpixel(bmp, x[0], FLIP(context, y[0])); - - rgba[0][RCOMP] = getr(color); - rgba[0][GCOMP] = getg(color); - rgba[0][BCOMP] = getb(color); - rgba[0][ACOMP] = 255; - } - - x++; y++; mask++; rgba++; - } - } - +/* + * Mesa 3-D graphics library + * Version: 3.0 + * Copyright (C) 1995-1998 Brian Paul + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Library General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Library General Public License for more details. + * + * You should have received a copy of the GNU Library General Public + * License along with this library; if not, write to the Free + * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +static void clear_color_generic(GLcontext *ctx, const GLfloat color[4]) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + GLubyte r, g, b; + CLAMPED_FLOAT_TO_UBYTE(r, color[0]); + CLAMPED_FLOAT_TO_UBYTE(g, color[1]); + CLAMPED_FLOAT_TO_UBYTE(b, color[2]); + context->ClearColor = makecol(r, g, b); + } + + +static void set_color_generic(GLcontext *ctx, + GLubyte red, GLubyte green, + GLubyte blue, GLubyte alpha) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + + context->CurrentColor = makecol(red, green, blue); + } + + +static GLbitfield clear_generic(GLcontext *ctx, + GLbitfield mask, GLboolean all, + GLint x, GLint y, + GLint width, GLint height) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + + if (mask & GL_COLOR_BUFFER_BIT) + { + if (all) + clear_to_color(context->Buffer->Active, context->ClearColor); + else + rect(context->Buffer->Active, + x, y, x+width-1, y+height-1, + context->ClearColor); + } + + return mask & (~GL_COLOR_BUFFER_BIT); + } + + +static void write_rgba_span_generic(const GLcontext *ctx, + GLuint n, GLint x, GLint y, + const GLubyte rgba[][4], + const GLubyte mask[]) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + BITMAP *bmp = context->Buffer->Active; + + y = FLIP(context, y); + + if (mask) + { + while (n--) + { + if (mask[0]) putpixel(bmp, x, y, makecol(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP])); + x++; mask++; rgba++; + } + } + else + { + while (n--) + { + putpixel(bmp, x, y, makecol(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP])); + x++; rgba++; + } + } + } + + +static void write_rgb_span_generic(const GLcontext *ctx, + GLuint n, GLint x, GLint y, + const GLubyte rgb[][3], + const GLubyte mask[]) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + BITMAP *bmp = context->Buffer->Active; + + y = FLIP(context, y); + + if (mask) + { + while(n--) + { + if (mask[0]) putpixel(bmp, x, y, makecol(rgb[0][RCOMP], rgb[0][GCOMP], rgb[0][BCOMP])); + x++; mask++; rgb++; + } + } + else + { + while (n--) + { + putpixel(bmp, x, y, makecol(rgb[0][RCOMP], rgb[0][GCOMP], rgb[0][BCOMP])); + x++; rgb++; + } + } + } + + +static void write_mono_rgba_span_generic(const GLcontext *ctx, + GLuint n, GLint x, GLint y, + const GLubyte mask[]) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + BITMAP *bmp = context->Buffer->Active; + int color = context->CurrentColor; + + y = FLIP(context, y); + + if (mask) + { + while(n--) + { + if (mask[0]) putpixel(bmp, x, y, color); + x++; mask++; + } + } + else + { + while(n--) + { + putpixel(bmp, x, y, color); + x++; + } + } + } + + +static void read_rgba_span_generic(const GLcontext *ctx, + GLuint n, GLint x, GLint y, + GLubyte rgba[][4]) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + BITMAP *bmp = context->Buffer->Active; + + y = FLIP(context, y); + + while (n--) + { + int color = getpixel(bmp, x, y); + + rgba[0][RCOMP] = getr(color); + rgba[0][GCOMP] = getg(color); + rgba[0][BCOMP] = getb(color); + rgba[0][ACOMP] = 255; + + x++; rgba++; + } + } + + +static void write_rgba_pixels_generic(const GLcontext *ctx, + GLuint n, + const GLint x[], + const GLint y[], + const GLubyte rgba[][4], + const GLubyte mask[]) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + BITMAP *bmp = context->Buffer->Active; + + while (n--) + { + if (mask[0]) putpixel(bmp, x[0], FLIP(context, y[0]), makecol(rgba[0][RCOMP], rgba[0][GCOMP], rgba[0][BCOMP])); + x++; y++; mask++; + } + } + + +static void write_mono_rgba_pixels_generic(const GLcontext *ctx, + GLuint n, + const GLint x[], + const GLint y[], + const GLubyte mask[]) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + BITMAP *bmp = context->Buffer->Active; + int color = context->CurrentColor; + + while (n--) + { + if (mask[0]) putpixel(bmp, x[0], FLIP(context, y[0]), color); + x++; y++; mask++; + } + } + + +static void read_rgba_pixels_generic(const GLcontext *ctx, + GLuint n, + const GLint x[], + const GLint y[], + GLubyte rgba[][4], + const GLubyte mask[]) + { + AMesaContext context = (AMesaContext)(ctx->DriverCtx); + BITMAP *bmp = context->Buffer->Active; + + while (n--) + { + if (mask[0]) + { + int color = getpixel(bmp, x[0], FLIP(context, y[0])); + + rgba[0][RCOMP] = getr(color); + rgba[0][GCOMP] = getg(color); + rgba[0][BCOMP] = getb(color); + rgba[0][ACOMP] = 255; + } + + x++; y++; mask++; rgba++; + } + } + diff --git a/src/mesa/drivers/d3d/D3DCAPS.CPP b/src/mesa/drivers/d3d/D3DCAPS.CPP index cd95ef0f1e..80ee91d922 100644 --- a/src/mesa/drivers/d3d/D3DCAPS.CPP +++ b/src/mesa/drivers/d3d/D3DCAPS.CPP @@ -1,250 +1,250 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver Build 5 */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#include "D3DHAL.h" -/*===========================================================================*/ -/* Macros. */ -/*===========================================================================*/ -#define SRCBLEND_MAP(gl,d3d,fall) if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwSrcBlendCaps & d3d ) \ - { \ - sprintf( buffer, "SRC Blend: %s -> %s", # gl, # d3d ); \ - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), buffer )); \ - pShared->dwSrcBlendCaps[index] = d3d; \ - } \ - else \ - { \ - sprintf( buffer, "SRC Blend: %s -> %s", # gl, # fall ); \ - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), buffer )); \ - pShared->dwSrcBlendCaps[index] = fall; \ - } -#define DSTBLEND_MAP(gl,d3d,fall) if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwDestBlendCaps & d3d ) \ - { \ - sprintf( buffer, "DST Blend: %s -> %s", # gl, # d3d ); \ - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), buffer )); \ - pShared->dwDestBlendCaps[index] = d3d; \ - } \ - else \ - { \ - sprintf( buffer, "DST Blend: %s -> %s", # gl, # fall ); \ - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), buffer )); \ - pShared->dwDestBlendCaps[index] = fall; \ - } - -/*===========================================================================*/ -/* I use this function to handle the fact that the D3D texture blending and */ -/* OpenGL texture blending functions don't map one to one. Also there is the*/ -/* problem with cards not supporting all the D3D functions. So I use the CAPS*/ -/* of the card to make a table of functions that will have defaults for the */ -/* unsupported functions. */ -/* So first I fill the table with the fallback function then I check to see */ -/* if the card supports the requested function. If it does I replace the */ -/* default thats already in the array. Now order does matter as I used an */ -/* enum type in D3DShared.h so that the mapping would be a little easier. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void AlphaBlendTableHAL( PMESAD3DHAL pHAL ) -{ - PMESAD3DSHARED pShared = &pHAL->shared; - int index; - char buffer[128]; - - DPF(( DBG_FUNC, "AlphaBlendTableHAL();" )); - - /* Make the fallback for the Source blend. */ - for( index = 0; index < 14; index++ ) - { - switch( index ) - { - case s_zero: - SRCBLEND_MAP( GL_ZERO, D3DBLEND_ZERO, D3DBLEND_ONE ); - break; - case s_one: - SRCBLEND_MAP( GL_ONE, D3DBLEND_ONE, D3DBLEND_ONE ); - break; - case s_dst_color: - SRCBLEND_MAP( GL_DST_COLOR, D3DBLEND_DESTCOLOR, D3DBLEND_ONE ); - break; - case s_one_minus_dst_color: - SRCBLEND_MAP( GL_ONE_MINUS_DST_COLOR, D3DBLEND_INVDESTCOLOR, D3DBLEND_ONE ); - break; - case s_src_alpha: - SRCBLEND_MAP( GL_SRC_ALPHA, D3DBLEND_SRCALPHA, D3DBLEND_ONE ); - break; - case s_one_minus_src_alpha: - SRCBLEND_MAP( GL_ONE_MINUS_SRC_ALPHA, D3DBLEND_INVSRCALPHA, D3DBLEND_ONE ); - break; - case s_dst_alpha: - SRCBLEND_MAP( GL_DST_ALPHA, D3DBLEND_DESTALPHA, D3DBLEND_ONE ); - break; - case s_one_minus_dst_alpha: - SRCBLEND_MAP( GL_ONE_MINUS_DST_ALPHA, D3DBLEND_INVDESTALPHA, D3DBLEND_ONE ); - break; - case s_src_alpha_saturate: - SRCBLEND_MAP( GL_SRC_ALPHA_SATURATE, D3DBLEND_SRCALPHASAT, D3DBLEND_ONE ); - break; - case s_constant_color: - SRCBLEND_MAP( GL_CONSTANT_COLOR, D3DBLEND_SRCCOLOR, D3DBLEND_ONE ); - break; - case s_one_minus_constant_color: - SRCBLEND_MAP( GL_ONE_MINUS_CONSTANT_COLOR, D3DBLEND_INVSRCCOLOR, D3DBLEND_ONE ); - break; - case s_constant_alpha: - SRCBLEND_MAP( GL_CONSTANT_ALPHA, D3DBLEND_BOTHSRCALPHA, D3DBLEND_ONE ); - break; - case s_one_minus_constant_alpha: - SRCBLEND_MAP( GL_ONE_MINUS_CONSTANT_ALPHA, D3DBLEND_BOTHINVSRCALPHA, D3DBLEND_ONE ); - break; - } - } - - /* Make the fallback for the Destination blend. */ - for( index = 0; index < 14; index++ ) - { - switch( index ) - { - case d_zero: - DSTBLEND_MAP( GL_ZERO, D3DBLEND_ZERO, D3DBLEND_ONE ); - break; - case d_one: - DSTBLEND_MAP( GL_ONE, D3DBLEND_ONE, D3DBLEND_ONE ); - break; - case d_src_color: - DSTBLEND_MAP( GL_SRC_COLOR, D3DBLEND_SRCCOLOR, D3DBLEND_ONE ); - break; - case d_one_minus_src_color: - DSTBLEND_MAP( GL_ONE_MINUS_SRC_COLOR, D3DBLEND_INVSRCCOLOR, D3DBLEND_ONE ); - break; - case d_src_alpha: - DSTBLEND_MAP( GL_SRC_ALPHA, D3DBLEND_SRCALPHA, D3DBLEND_ONE ); - break; - case d_one_minus_src_alpha: - DSTBLEND_MAP( GL_ONE_MINUS_SRC_ALPHA, D3DBLEND_INVSRCALPHA, D3DBLEND_ONE ); - break; - case d_dst_alpha: - DSTBLEND_MAP( GL_DST_ALPHA, D3DBLEND_DESTALPHA, D3DBLEND_ONE ); - break; - case d_one_minus_dst_alpha: - DSTBLEND_MAP( GL_ONE_MINUS_DST_ALPHA, D3DBLEND_INVDESTALPHA, D3DBLEND_ONE ); - break; - case d_constant_color: - DSTBLEND_MAP( GL_CONSTANT_COLOR, D3DBLEND_DESTCOLOR, D3DBLEND_ONE ); - break; - case d_one_minus_constant_color: - DSTBLEND_MAP( GL_ONE_MINUS_CONSTANT_COLOR, D3DBLEND_INVDESTCOLOR, D3DBLEND_ONE ); - break; - case d_constant_alpha: - DSTBLEND_MAP( GL_CONSTANT_ALPHAR, D3DBLEND_BOTHSRCALPHA, D3DBLEND_ONE ); - break; - case d_one_minus_constant_alpha: - DSTBLEND_MAP( GL_ONE_MINUS_CONSTANT_ALPHA, D3DBLEND_BOTHINVSRCALPHA, D3DBLEND_ONE ); - break; - } - } - - /* Make the fallbacks for the texture functions. */ - for( index = 0; index < 4; index++ ) - { - switch( index ) - { - case d3dtblend_decal: - if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECAL ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECAL -> D3DTBLEND_DECAL" )); - pShared->dwTexFunc[index] = D3DTBLEND_DECAL; - } - else - { - if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATE ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECAL -> D3DTBLEND_MODULATE" )); - pShared->dwTexFunc[index] = D3DTBLEND_MODULATE; - } - else - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECAL -> D3DTBLEND_ADD" )); - pShared->dwTexFunc[index] = D3DTBLEND_ADD; - } - } - break; - case d3dtblend_decalalpha: - if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECALALPHA ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECALALPHA -> D3DTBLEND_DECALALPHA" )); - pShared->dwTexFunc[index] = D3DTBLEND_DECALALPHA; - } - else - { - if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECAL ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECALALPA -> D3DTBLEND_DECAL" )); - pShared->dwTexFunc[index] = D3DTBLEND_DECAL; - } - else - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECALALPHA -> D3DTBLEND_ADD" )); - pShared->dwTexFunc[index] = D3DTBLEND_ADD; - } - } - break; - case d3dtblend_modulate: - if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATE ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATE -> D3DTBLEND_MODULATE" )); - pShared->dwTexFunc[index] = D3DTBLEND_MODULATE; - } - else - { - if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATEALPHA ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATE -> D3DTBLEND_MODULATEALPHA" )); - pShared->dwTexFunc[index] = D3DTBLEND_MODULATEALPHA; - } - else if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECAL ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATE -> D3DTBLEND_DECAL" )); - pShared->dwTexFunc[index] = D3DTBLEND_DECAL; - } - else - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATE -> D3DTBLEND_ADD" )); - pShared->dwTexFunc[index] = D3DTBLEND_ADD; - } - } - break; - case d3dtblend_modulatealpha: - if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATEALPHA ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATEALPHA -> D3DTBLEND_MODULATEALPHA" )); - pShared->dwTexFunc[index] = D3DTBLEND_MODULATEALPHA; - } - else - { - if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATE ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATEALPHA -> D3DTBLEND_MODULATE" )); - pShared->dwTexFunc[index] = D3DTBLEND_MODULATE; - } - else if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECAL ) - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATEALPHA -> D3DTBLEND_DECALE" )); - pShared->dwTexFunc[index] = D3DTBLEND_DECAL; - } - else - { - DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATEALPHA -> D3DTBLEND_ADD" )); - pShared->dwTexFunc[index] = D3DTBLEND_ADD; - } - } - break; - } - } -} - +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver Build 5 */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#include "D3DHAL.h" +/*===========================================================================*/ +/* Macros. */ +/*===========================================================================*/ +#define SRCBLEND_MAP(gl,d3d,fall) if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwSrcBlendCaps & d3d ) \ + { \ + sprintf( buffer, "SRC Blend: %s -> %s", # gl, # d3d ); \ + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), buffer )); \ + pShared->dwSrcBlendCaps[index] = d3d; \ + } \ + else \ + { \ + sprintf( buffer, "SRC Blend: %s -> %s", # gl, # fall ); \ + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), buffer )); \ + pShared->dwSrcBlendCaps[index] = fall; \ + } +#define DSTBLEND_MAP(gl,d3d,fall) if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwDestBlendCaps & d3d ) \ + { \ + sprintf( buffer, "DST Blend: %s -> %s", # gl, # d3d ); \ + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), buffer )); \ + pShared->dwDestBlendCaps[index] = d3d; \ + } \ + else \ + { \ + sprintf( buffer, "DST Blend: %s -> %s", # gl, # fall ); \ + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), buffer )); \ + pShared->dwDestBlendCaps[index] = fall; \ + } + +/*===========================================================================*/ +/* I use this function to handle the fact that the D3D texture blending and */ +/* OpenGL texture blending functions don't map one to one. Also there is the*/ +/* problem with cards not supporting all the D3D functions. So I use the CAPS*/ +/* of the card to make a table of functions that will have defaults for the */ +/* unsupported functions. */ +/* So first I fill the table with the fallback function then I check to see */ +/* if the card supports the requested function. If it does I replace the */ +/* default thats already in the array. Now order does matter as I used an */ +/* enum type in D3DShared.h so that the mapping would be a little easier. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void AlphaBlendTableHAL( PMESAD3DHAL pHAL ) +{ + PMESAD3DSHARED pShared = &pHAL->shared; + int index; + char buffer[128]; + + DPF(( DBG_FUNC, "AlphaBlendTableHAL();" )); + + /* Make the fallback for the Source blend. */ + for( index = 0; index < 14; index++ ) + { + switch( index ) + { + case s_zero: + SRCBLEND_MAP( GL_ZERO, D3DBLEND_ZERO, D3DBLEND_ONE ); + break; + case s_one: + SRCBLEND_MAP( GL_ONE, D3DBLEND_ONE, D3DBLEND_ONE ); + break; + case s_dst_color: + SRCBLEND_MAP( GL_DST_COLOR, D3DBLEND_DESTCOLOR, D3DBLEND_ONE ); + break; + case s_one_minus_dst_color: + SRCBLEND_MAP( GL_ONE_MINUS_DST_COLOR, D3DBLEND_INVDESTCOLOR, D3DBLEND_ONE ); + break; + case s_src_alpha: + SRCBLEND_MAP( GL_SRC_ALPHA, D3DBLEND_SRCALPHA, D3DBLEND_ONE ); + break; + case s_one_minus_src_alpha: + SRCBLEND_MAP( GL_ONE_MINUS_SRC_ALPHA, D3DBLEND_INVSRCALPHA, D3DBLEND_ONE ); + break; + case s_dst_alpha: + SRCBLEND_MAP( GL_DST_ALPHA, D3DBLEND_DESTALPHA, D3DBLEND_ONE ); + break; + case s_one_minus_dst_alpha: + SRCBLEND_MAP( GL_ONE_MINUS_DST_ALPHA, D3DBLEND_INVDESTALPHA, D3DBLEND_ONE ); + break; + case s_src_alpha_saturate: + SRCBLEND_MAP( GL_SRC_ALPHA_SATURATE, D3DBLEND_SRCALPHASAT, D3DBLEND_ONE ); + break; + case s_constant_color: + SRCBLEND_MAP( GL_CONSTANT_COLOR, D3DBLEND_SRCCOLOR, D3DBLEND_ONE ); + break; + case s_one_minus_constant_color: + SRCBLEND_MAP( GL_ONE_MINUS_CONSTANT_COLOR, D3DBLEND_INVSRCCOLOR, D3DBLEND_ONE ); + break; + case s_constant_alpha: + SRCBLEND_MAP( GL_CONSTANT_ALPHA, D3DBLEND_BOTHSRCALPHA, D3DBLEND_ONE ); + break; + case s_one_minus_constant_alpha: + SRCBLEND_MAP( GL_ONE_MINUS_CONSTANT_ALPHA, D3DBLEND_BOTHINVSRCALPHA, D3DBLEND_ONE ); + break; + } + } + + /* Make the fallback for the Destination blend. */ + for( index = 0; index < 14; index++ ) + { + switch( index ) + { + case d_zero: + DSTBLEND_MAP( GL_ZERO, D3DBLEND_ZERO, D3DBLEND_ONE ); + break; + case d_one: + DSTBLEND_MAP( GL_ONE, D3DBLEND_ONE, D3DBLEND_ONE ); + break; + case d_src_color: + DSTBLEND_MAP( GL_SRC_COLOR, D3DBLEND_SRCCOLOR, D3DBLEND_ONE ); + break; + case d_one_minus_src_color: + DSTBLEND_MAP( GL_ONE_MINUS_SRC_COLOR, D3DBLEND_INVSRCCOLOR, D3DBLEND_ONE ); + break; + case d_src_alpha: + DSTBLEND_MAP( GL_SRC_ALPHA, D3DBLEND_SRCALPHA, D3DBLEND_ONE ); + break; + case d_one_minus_src_alpha: + DSTBLEND_MAP( GL_ONE_MINUS_SRC_ALPHA, D3DBLEND_INVSRCALPHA, D3DBLEND_ONE ); + break; + case d_dst_alpha: + DSTBLEND_MAP( GL_DST_ALPHA, D3DBLEND_DESTALPHA, D3DBLEND_ONE ); + break; + case d_one_minus_dst_alpha: + DSTBLEND_MAP( GL_ONE_MINUS_DST_ALPHA, D3DBLEND_INVDESTALPHA, D3DBLEND_ONE ); + break; + case d_constant_color: + DSTBLEND_MAP( GL_CONSTANT_COLOR, D3DBLEND_DESTCOLOR, D3DBLEND_ONE ); + break; + case d_one_minus_constant_color: + DSTBLEND_MAP( GL_ONE_MINUS_CONSTANT_COLOR, D3DBLEND_INVDESTCOLOR, D3DBLEND_ONE ); + break; + case d_constant_alpha: + DSTBLEND_MAP( GL_CONSTANT_ALPHAR, D3DBLEND_BOTHSRCALPHA, D3DBLEND_ONE ); + break; + case d_one_minus_constant_alpha: + DSTBLEND_MAP( GL_ONE_MINUS_CONSTANT_ALPHA, D3DBLEND_BOTHINVSRCALPHA, D3DBLEND_ONE ); + break; + } + } + + /* Make the fallbacks for the texture functions. */ + for( index = 0; index < 4; index++ ) + { + switch( index ) + { + case d3dtblend_decal: + if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECAL ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECAL -> D3DTBLEND_DECAL" )); + pShared->dwTexFunc[index] = D3DTBLEND_DECAL; + } + else + { + if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATE ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECAL -> D3DTBLEND_MODULATE" )); + pShared->dwTexFunc[index] = D3DTBLEND_MODULATE; + } + else + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECAL -> D3DTBLEND_ADD" )); + pShared->dwTexFunc[index] = D3DTBLEND_ADD; + } + } + break; + case d3dtblend_decalalpha: + if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECALALPHA ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECALALPHA -> D3DTBLEND_DECALALPHA" )); + pShared->dwTexFunc[index] = D3DTBLEND_DECALALPHA; + } + else + { + if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECAL ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECALALPA -> D3DTBLEND_DECAL" )); + pShared->dwTexFunc[index] = D3DTBLEND_DECAL; + } + else + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_DECALALPHA -> D3DTBLEND_ADD" )); + pShared->dwTexFunc[index] = D3DTBLEND_ADD; + } + } + break; + case d3dtblend_modulate: + if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATE ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATE -> D3DTBLEND_MODULATE" )); + pShared->dwTexFunc[index] = D3DTBLEND_MODULATE; + } + else + { + if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATEALPHA ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATE -> D3DTBLEND_MODULATEALPHA" )); + pShared->dwTexFunc[index] = D3DTBLEND_MODULATEALPHA; + } + else if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECAL ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATE -> D3DTBLEND_DECAL" )); + pShared->dwTexFunc[index] = D3DTBLEND_DECAL; + } + else + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATE -> D3DTBLEND_ADD" )); + pShared->dwTexFunc[index] = D3DTBLEND_ADD; + } + } + break; + case d3dtblend_modulatealpha: + if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATEALPHA ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATEALPHA -> D3DTBLEND_MODULATEALPHA" )); + pShared->dwTexFunc[index] = D3DTBLEND_MODULATEALPHA; + } + else + { + if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_MODULATE ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATEALPHA -> D3DTBLEND_MODULATE" )); + pShared->dwTexFunc[index] = D3DTBLEND_MODULATE; + } + else if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureBlendCaps & D3DTBLEND_DECAL ) + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATEALPHA -> D3DTBLEND_DECALE" )); + pShared->dwTexFunc[index] = D3DTBLEND_DECAL; + } + else + { + DPF(( (DBG_CNTX_INFO|DBG_TXT_INFO), "D3DTBLEND_MODULATEALPHA -> D3DTBLEND_ADD" )); + pShared->dwTexFunc[index] = D3DTBLEND_ADD; + } + } + break; + } + } +} + diff --git a/src/mesa/drivers/d3d/D3DHAL.H b/src/mesa/drivers/d3d/D3DHAL.H index 5295520d7b..2496d164e5 100644 --- a/src/mesa/drivers/d3d/D3DHAL.H +++ b/src/mesa/drivers/d3d/D3DHAL.H @@ -1,68 +1,68 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#ifndef _D3D_HAL_INC -#define _D3D_HAL_INC - -/*===========================================================================*/ -/* Includes. */ -/*===========================================================================*/ -#include -#include -#include -#include -#include -#include "D3DShared.h" -#include "D3DTextureMgr.h" -#include "Debug.h" -/*===========================================================================*/ -/* Defines. */ -/*===========================================================================*/ -#define DX_RESTORE(ps) if ( (ps) && (ps)->IsLost() ) (ps)->Restore(); -/*===========================================================================*/ -/* Type defines. */ -/*===========================================================================*/ -typedef struct _d3d_hal_struct -{ - MESAD3DSHARED shared; - - GUID guid; - LPDIRECTDRAW lpDD; - LPDIRECTDRAW4 lpDD4; - LPDIRECT3D3 lpD3D3; - LPDIRECT3DDEVICE3 lpD3DDevice; - D3DDEVICEDESC D3DHWDevDesc; - LPDIRECTDRAWSURFACE4 lpDDSPrimary, - lpDDSRender, - lpDDSZbuffer; - LPDIRECT3DVIEWPORT3 lpViewport; - LPDIRECTDRAWCLIPPER lpClipper; - DDPIXELFORMAT ddpf, - ddpfZBuffer; - PTM_OBJECT pTMList; - -} MESAD3DHAL, *PMESAD3DHAL; -/*===========================================================================*/ -/* External function prototypes. */ -/*===========================================================================*/ -extern BOOL InitTMgrHAL( PMESAD3DHAL pHAL ); -extern void TermTMgrHAL( PMESAD3DHAL pHAL ); -extern void AlphaBlendTableHAL( PMESAD3DHAL pHAL ); - -extern void Solve8BitChannelPixelFormat( DDPIXELFORMAT *pddpf, PPIXELINFO pPixel ); -extern char *ErrorStringD3D( HRESULT hr ); -extern void FatalShutDown( PMESAD3DHAL pHAL ); -/*===========================================================================*/ -/* Global variables. */ -/*===========================================================================*/ -extern char *errorMsg; - -#endif - +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#ifndef _D3D_HAL_INC +#define _D3D_HAL_INC + +/*===========================================================================*/ +/* Includes. */ +/*===========================================================================*/ +#include +#include +#include +#include +#include +#include "D3DShared.h" +#include "D3DTextureMgr.h" +#include "Debug.h" +/*===========================================================================*/ +/* Defines. */ +/*===========================================================================*/ +#define DX_RESTORE(ps) if ( (ps) && (ps)->IsLost() ) (ps)->Restore(); +/*===========================================================================*/ +/* Type defines. */ +/*===========================================================================*/ +typedef struct _d3d_hal_struct +{ + MESAD3DSHARED shared; + + GUID guid; + LPDIRECTDRAW lpDD; + LPDIRECTDRAW4 lpDD4; + LPDIRECT3D3 lpD3D3; + LPDIRECT3DDEVICE3 lpD3DDevice; + D3DDEVICEDESC D3DHWDevDesc; + LPDIRECTDRAWSURFACE4 lpDDSPrimary, + lpDDSRender, + lpDDSZbuffer; + LPDIRECT3DVIEWPORT3 lpViewport; + LPDIRECTDRAWCLIPPER lpClipper; + DDPIXELFORMAT ddpf, + ddpfZBuffer; + PTM_OBJECT pTMList; + +} MESAD3DHAL, *PMESAD3DHAL; +/*===========================================================================*/ +/* External function prototypes. */ +/*===========================================================================*/ +extern BOOL InitTMgrHAL( PMESAD3DHAL pHAL ); +extern void TermTMgrHAL( PMESAD3DHAL pHAL ); +extern void AlphaBlendTableHAL( PMESAD3DHAL pHAL ); + +extern void Solve8BitChannelPixelFormat( DDPIXELFORMAT *pddpf, PPIXELINFO pPixel ); +extern char *ErrorStringD3D( HRESULT hr ); +extern void FatalShutDown( PMESAD3DHAL pHAL ); +/*===========================================================================*/ +/* Global variables. */ +/*===========================================================================*/ +extern char *errorMsg; + +#endif + diff --git a/src/mesa/drivers/d3d/D3DInit.cpp b/src/mesa/drivers/d3d/D3DInit.cpp index ba7891612e..e21cd21e25 100644 --- a/src/mesa/drivers/d3d/D3DInit.cpp +++ b/src/mesa/drivers/d3d/D3DInit.cpp @@ -1,891 +1,891 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver Build 5 */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#include "D3DHAL.h" -/*===========================================================================*/ -/* Local function prototypes. */ -/*===========================================================================*/ -static void DestroyAllSurfaces( PMESAD3DHAL pHAL ); -static void DestroyDevice( PMESAD3DHAL pHAL ); -static void DestroyInterfaces( PMESAD3DHAL pHAL ); - -HRESULT WINAPI EnumSurfacesHook( LPDIRECTDRAWSURFACE4 lpDDS, LPDDSURFACEDESC2 lpDDSDesc, LPVOID pVoid ); -HRESULT CALLBACK EnumZBufferHook( DDPIXELFORMAT* pddpf, VOID *pVoid ); -HRESULT CALLBACK EnumDeviceHook( GUID FAR* lpGuid, LPSTR lpDesc, LPSTR lpName, LPD3DDEVICEDESC lpD3DHWDesc, LPD3DDEVICEDESC lpD3DHELDesc, void *pVoid ); -/*===========================================================================*/ -/* Globals. */ -/*===========================================================================*/ -//char *errorMsg; -/*===========================================================================*/ -/* This function is responable for allocating the actual MESAD3DHAL struct. */ -/* Each Mesa context will have its own MESAD3DHAL struct so its like a mini */ -/* context to some extent. All one time allocations/operations get done here.*/ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -extern "C" PMESAD3DSHARED InitHAL( HWND hwnd ) -{ - PMESAD3DHAL pHAL; - ULONG rc; - - DPF(( DBG_FUNC, "InitHAL();" )); - DPF(( DBG_CNTX_INFO, "hwnd: %d", hwnd )); - - /* Allocate the structure and zero it out. */ - pHAL = (PMESAD3DHAL)ALLOC( sizeof(MESAD3DHAL) ); - if ( pHAL == NULL ) - { - RIP( pHAL, "InitHAL->", "Memory Allocation" ); - return (PMESAD3DSHARED)NULL; - } - memset( pHAL, 0, sizeof(MESAD3DHAL) ); - - /* Get the texture manager going. */ - rc = InitTMgrHAL( pHAL ); - if ( rc == FALSE ) - { - RIP( pHAL, "InitTMgrHAL->", "Failed" ); - return (PMESAD3DSHARED)NULL; - } - - /* Fill in the window parameters if we can. */ - pHAL->shared.hwnd = hwnd; - - /* Parse the user's enviroment variables to generate a debug mask. */ - ReadDBGEnv(); - - return (PMESAD3DSHARED)pHAL; -} -/*===========================================================================*/ -/* This function will unload all the resources that the MESAD3DHAL struct */ -/* has bound to it. The actual structure itself will be freed. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void TermHAL( PMESAD3DSHARED pShared ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - - DPF(( DBG_FUNC, "TermHAL();" )); - - /* Check for an empty wrapper structure. */ - if ( pHAL == NULL ) - return; - - /* Kill this texture manager. */ - TermTMgrHAL( pHAL ); - - /* Kill any DDraw stuff if exists. */ - DestroyDevice( pHAL ); - DestroyAllSurfaces( pHAL ); - DestroyInterfaces( pHAL ); - - FREE( pHAL ); -} -/*===========================================================================*/ -/* This function is used to init and resize the rendering surface as the two*/ -/* are almost the same. First the device and all the surfaces are destoryed */ -/* if they already exist. Next we create a OffScreen rendering surface and */ -/* save some pixelformat info to do color convertions. Next we start to take */ -/* care of getting the most out of the hardware. I use bHardware to determine*/ -/* the state of the device we found in the device enumeration. The enum proc*/ -/* will try for hardware first. I next use a bForceSW to make the enum proc */ -/* choose a software device. So I will try some combinations with HW first */ -/* until I feel I have to set the bForceSW and call this function again. If */ -/* this function is called with no width or height then use the internals. */ -/* NOTE: The worst case is that all will be in SW (RGBDevice) and really */ -/* I should forget the whole thing and fall back to a DDraw span type*/ -/* rendering but what is the point. This way I always know I have a */ -/* D3DDevice and that makes things easier. I do impliment the span */ -/* rendering function for stuff that I haven't done support for such */ -/* as points and lines. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE */ -/*===========================================================================*/ -extern "C" BOOL CreateHAL( PMESAD3DSHARED pShared ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - DDSURFACEDESC2 ddsd2; - D3DDEVICEDESC D3DSWDevDesc; - DDSCAPS2 ddscaps; - DWORD dwCoopFlags, - dwWidth, - dwHeight; - ULONG rc; - - DPF(( DBG_FUNC, "CreateHAL();" )); - -#define InitDDSD2(f) memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); \ - ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); \ - ddsd2.dwFlags = f; - - if ( pHAL == NULL ) - return FALSE; - - /* Use the internal rectangle struct. */ - dwWidth = pShared->rectW.right - pShared->rectW.left; - dwHeight = pShared->rectW.bottom - pShared->rectW.top; - - DPF(( DBG_CNTX_INFO, "Width: %d Height: %d", dwWidth, dwHeight )); - - /* The dimensions might still be the same so just leave. */ - if ( (dwWidth == pShared->dwWidth) && (dwHeight == pShared->dwHeight) ) - { - DPF(( DBG_CNTX_WARN, "Context size hasn't changed" )); - return TRUE; - } - - /* If one of the dimensions are zero then leave. WM_SIZE should get us back here. */ - if ( (dwWidth == 0) || (dwHeight == 0) ) - return TRUE; - - /* Save the renders dimensions. */ - pShared->dwWidth = dwWidth; - pShared->dwHeight = dwHeight; - - DPF(( DBG_CNTX_INFO, "Creating Context:\n cx:%d cy:%d", pShared->dwWidth, pShared->dwHeight )); - - /*=================================*/ - /* Create all required interfaces. */ - /*=================================*/ - - /* Kill any DDraw stuff if exists. */ - DestroyDevice( pHAL ); - DestroyAllSurfaces( pHAL ); - DestroyInterfaces( pHAL ); - - /* Create a instance of DDraw using the Primary display driver. */ - rc = DirectDrawCreate( NULL, &pHAL->lpDD, NULL ); - if( FAILED(rc) ) - { - RIP( pHAL, "DirectDrawCreate->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Get the DDraw4 interface. */ - rc = pHAL->lpDD->QueryInterface( IID_IDirectDraw4, (void **)&pHAL->lpDD4 ); - if( FAILED(rc) ) - { - RIP( pHAL, "QueryInterface (IID_IDirectDraw4) ->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Get the Direct3D3 interface. */ - rc = pHAL->lpDD4->QueryInterface( IID_IDirect3D3, (void **)&pHAL->lpD3D3 ); - if( FAILED(rc) ) - { - RIP( pHAL, "QueryInterface (IID_IDirect3D3) ->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Set the Cooperative level. NOTE: we need to know if we are FS at this point.*/ - dwCoopFlags = (pShared->bWindow == TRUE) ? DDSCL_NORMAL : (DDSCL_EXCLUSIVE | DDSCL_FULLSCREEN); - rc = pHAL->lpDD4->SetCooperativeLevel( pShared->hwnd, dwCoopFlags ); - if ( FAILED(rc) ) - { - RIP( pHAL, "SetCooperativeLevel->", ErrorStringD3D(rc) ); - return FALSE; - } - - /*==================================================================*/ - /* Get the best device we can and note whether its hardware or not. */ - /*==================================================================*/ - pShared->bForceSW = FALSE; - pHAL->lpD3D3->EnumDevices( EnumDeviceHook, (void *)pHAL ); - pShared->bHardware = IsEqualIID( pHAL->guid, IID_IDirect3DHALDevice ); - DPF(( DBG_CNTX_INFO, "bHardware: %s", (pShared->bHardware) ? "TRUE" : "FALSE" )); - DPF(( DBG_CNTX_INFO, "bWindowed: %s", (pShared->bWindow) ? "TRUE" : "FALSE" )); - - /*========================================================================*/ - /* HARDWARE was found. */ - /*========================================================================*/ - if ( pShared->bHardware == TRUE ) - { - /*===================================*/ - /* HARDWARE -> Z-BUFFER. */ - /*===================================*/ - - /* Get a Z-Buffer pixelformat. */ - memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); - ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); - rc = pHAL->lpD3D3->EnumZBufferFormats( pHAL->guid, EnumZBufferHook, (VOID*)&ddsd2.ddpfPixelFormat ); - if ( FAILED(rc) ) - { - RIP( pHAL, "EnumZBufferFormatsl->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Setup our request structure for the Z-buffer surface. */ - ddsd2.dwFlags = DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT | DDSD_PIXELFORMAT; - ddsd2.ddsCaps.dwCaps = DDSCAPS_ZBUFFER | DDSCAPS_VIDEOMEMORY; - ddsd2.dwWidth = dwWidth; - ddsd2.dwHeight = dwHeight; - rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSZbuffer, NULL ); - if ( !FAILED(rc) ) - { - DPF(( DBG_CNTX_INFO, "HW ZBuffer" )); - - /*===================================*/ - /* HARDWARE -> Z-BUFFER -> FLIPABLE */ - /*===================================*/ - if ( pShared->bWindow == FALSE ) - { - InitDDSD2( DDSD_CAPS | DDSD_BACKBUFFERCOUNT ); - ddsd2.dwBackBufferCount = 1; - ddsd2.ddsCaps.dwCaps = DDSCAPS_PRIMARYSURFACE | DDSCAPS_3DDEVICE | DDSCAPS_FLIP | DDSCAPS_COMPLEX; - rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSPrimary, NULL ); - if ( FAILED(rc) ) - { - /* Make sure we try the next fall back. */ - DPF(( DBG_CNTX_WARN, "HW Flip/Complex not available" )); - pHAL->lpDDSPrimary = NULL; - } - else - { - /* Get the back buffer that was created. */ - ddscaps.dwCaps = DDSCAPS_BACKBUFFER; - rc = pHAL->lpDDSPrimary->GetAttachedSurface( &ddscaps, &pHAL->lpDDSRender ); - if ( FAILED(rc) ) - { - DPF(( DBG_CNTX_WARN, "GetAttachedSurface failed -> HW Flip/Complex" )); - - /* Make sure we try the next fall back. */ - pHAL->lpDDSPrimary->Release(); - pHAL->lpDDSPrimary = NULL; - } - else - { - /* I have had problems when a complex surface comes back */ - /* with the back buffer being created in SW. Not sure why */ - /* or how this is possable but I'm checking for it here. */ - memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); - ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); - DX_RESTORE( pHAL->lpDDSRender ); - rc = pHAL->lpDDSRender->GetSurfaceDesc( &ddsd2 ); - if ( FAILED(rc) ) - { - RIP( pHAL, "GetSurfaceDesc (RENDER) ->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* If the surface is in VID then we are happy with are Flipable. */ - if ( ddsd2.ddsCaps.dwCaps & DDSCAPS_LOCALVIDMEM ) - { - pShared->bFlipable = TRUE; - DPF(( DBG_CNTX_INFO, "HW Flip/Complex!" )); - } - else - { - /* Kill this setup. */ - pHAL->lpDDSPrimary->Release(); - pHAL->lpDDSPrimary = NULL; - } - } - } - } - - /*===================================*/ - /* HARDWARE -> Z-BUFFER -> BLT */ - /*===================================*/ - if ( pHAL->lpDDSPrimary == NULL ) - { - pShared->bFlipable = FALSE; - - /* Create the Primary (front buffer). */ - InitDDSD2( DDSD_CAPS ); - ddsd2.ddsCaps.dwCaps = DDSCAPS_PRIMARYSURFACE; - rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSPrimary, NULL ); - if ( FAILED(rc) ) - { - /* This is an error as we should be able to do this at minimum. */ - RIP( pHAL, "CreateSurface (PRIMARY) ->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Create the Render (back buffer). */ - InitDDSD2( DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT ); - ddsd2.dwWidth = dwWidth; - ddsd2.dwHeight = dwHeight; - ddsd2.ddsCaps.dwCaps = DDSCAPS_OFFSCREENPLAIN | DDSCAPS_3DDEVICE; - rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSRender, NULL ); - if ( FAILED(rc) ) - { - DPF(( DBG_CNTX_WARN, "Failed HW Offscreen surface" )); - - /* Make sure we try the next fall back. */ - pHAL->lpDDSPrimary->Release(); - pHAL->lpDDSPrimary = NULL; - } - else - { - /* Might as well check here too see if this surface is in */ - /* hardware. If nothing else just to be consistant. */ - memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); - ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); - DX_RESTORE( pHAL->lpDDSRender ); - rc = pHAL->lpDDSRender->GetSurfaceDesc( &ddsd2 ); - if ( FAILED(rc) ) - { - RIP( pHAL, "GetSurfaceDesc (RENDER) ->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* If the surface is in VID then we are happy. */ - if ( ddsd2.ddsCaps.dwCaps & DDSCAPS_LOCALVIDMEM ) - { - /* Create a clipper object so that DDraw will be able to blt windows that */ - /* have been clipped by the screen or other windows. */ - pHAL->lpDD4->CreateClipper( 0, &pHAL->lpClipper, NULL ); - pHAL->lpClipper->SetHWnd( 0, pShared->hwnd ); - pHAL->lpDDSPrimary->SetClipper( pHAL->lpClipper ); - pHAL->lpClipper->Release(); - DPF(( DBG_CNTX_INFO, "HW RENDER surface" )); - } - else - { - /* Kill this setup. */ - pHAL->lpDDSRender->Release(); - pHAL->lpDDSRender = NULL; - pHAL->lpDDSPrimary->Release(); - pHAL->lpDDSPrimary = NULL; - } - } - } - - /*===================================*/ - /* Create D3DDEVICE -> HARDWARE. */ - /*===================================*/ - if ( pHAL->lpDDSZbuffer && pHAL->lpDDSPrimary && pHAL->lpDDSRender ) - { - DX_RESTORE( pHAL->lpDDSRender ); - DX_RESTORE( pHAL->lpDDSZbuffer ); - - rc = pHAL->lpDDSRender->AddAttachedSurface( pHAL->lpDDSZbuffer ); - if ( FAILED(rc) ) - { - RIP( pHAL, "AddAttachedSurface (ZBUFFER) ->", ErrorStringD3D(rc) ); - return FALSE; - } - - rc = pHAL->lpD3D3->CreateDevice( IID_IDirect3DHALDevice, pHAL->lpDDSRender, &pHAL->lpD3DDevice, NULL ); - if ( rc != D3D_OK ) - { - DPF(( DBG_CNTX_WARN, "Failed HW Device" )); - pHAL->lpD3DDevice = NULL; - } - else - { - DPF(( DBG_CNTX_INFO, "HW Device" )); - } - } - } - } - - /*========================================================================*/ - /* SOFTWARE fallback. */ - /*========================================================================*/ - if ( pHAL->lpD3DDevice == NULL ) - { - DPF(( DBG_CNTX_INFO, "SW fallback :(" )); - - /* Make sure we have no surfaces allocated. Just incase. */ - DestroyAllSurfaces( pHAL ); - - /* Get a software device. */ - pShared->bFlipable = FALSE; - pShared->bForceSW = TRUE; - pHAL->lpD3D3->EnumDevices( EnumDeviceHook, (void *)pHAL ); - pShared->bHardware = IsEqualIID( pHAL->guid, IID_IDirect3DHALDevice ); - - /*===================================*/ - /* SOFTWARE -> Z-BUFFER. */ - /*===================================*/ - - /*===================================*/ - /* SOFTWARE -> Z-BUFFER -> FLIPABLE */ - /*===================================*/ - if ( pShared->bWindow == FALSE ) - { - InitDDSD2( DDSD_CAPS | DDSD_BACKBUFFERCOUNT ); - ddsd2.dwBackBufferCount = 1; - ddsd2.ddsCaps.dwCaps = DDSCAPS_PRIMARYSURFACE | DDSCAPS_3DDEVICE | DDSCAPS_FLIP | DDSCAPS_COMPLEX; - ddsd2.ddpfPixelFormat.dwSize = sizeof( DDPIXELFORMAT ); - ddsd2.ddpfPixelFormat.dwFlags = (DDPF_RGB | DDPF_ALPHAPIXELS); - rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSPrimary, NULL ); - if ( FAILED(rc) ) - { - DPF(( DBG_CNTX_WARN, "Failed SW Flip/Complex" )); - - /* Make sure we try the next fall back. */ - pHAL->lpDDSPrimary = NULL; - } - else - { - ddscaps.dwCaps = DDSCAPS_BACKBUFFER; - rc = pHAL->lpDDSPrimary->GetAttachedSurface( &ddscaps, &pHAL->lpDDSRender ); - if ( FAILED(rc) ) - { - /* Make sure we try the next fall back. */ - DPF(( DBG_CNTX_WARN, "GetAttachedSurface failed -> SW Flip/Complex" )); - pHAL->lpDDSPrimary->Release(); - pHAL->lpDDSPrimary = NULL; - } - else - { - DPF(( DBG_CNTX_INFO, "SW Flip/Complex" )); - pShared->bFlipable = TRUE; - } - } - } - - /*===================================*/ - /* SOFTWARE -> Z-BUFFER -> BLT */ - /*===================================*/ - if ( pHAL->lpDDSPrimary == NULL ) - { - /* Create the Primary (front buffer). */ - InitDDSD2( DDSD_CAPS ); - ddsd2.ddsCaps.dwCaps = DDSCAPS_PRIMARYSURFACE; - rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSPrimary, NULL ); - if ( FAILED(rc) ) - { - /* This is an error as we should be able to do this at minimum. */ - RIP( pHAL, "CreateSurface (PRIMARY) ->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Create the Render (back buffer). */ - InitDDSD2( DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT ); - ddsd2.dwWidth = dwWidth; - ddsd2.dwHeight = dwHeight; - ddsd2.ddsCaps.dwCaps = DDSCAPS_OFFSCREENPLAIN | DDSCAPS_3DDEVICE; - ddsd2.ddpfPixelFormat.dwSize = sizeof( DDPIXELFORMAT ); - ddsd2.ddpfPixelFormat.dwFlags = (DDPF_RGB | DDPF_ALPHAPIXELS); - rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSRender, NULL ); - if ( FAILED(rc) ) - { - /* That was our last hope. */ - RIP( pHAL, "CreateSurface (RENDER) ->", ErrorStringD3D(rc) ); - return FALSE; - } - else - { - DPF(( DBG_CNTX_INFO, "SW RENDER surface" )); - - /* Create a clipper object so that DDraw will be able to blt windows that */ - /* have been clipped by the screen or other windows. */ - pHAL->lpDD4->CreateClipper( 0, &pHAL->lpClipper, NULL ); - pHAL->lpClipper->SetHWnd( 0, pShared->hwnd ); - pHAL->lpDDSPrimary->SetClipper( pHAL->lpClipper ); - pHAL->lpClipper->Release(); - } - } - - /*===================================*/ - /* Create D3DDEVICE -> SOFTWARE. */ - /*===================================*/ - if ( pHAL->lpDDSPrimary && pHAL->lpDDSRender ) - { - DX_RESTORE( pHAL->lpDDSRender ); - rc = pHAL->lpD3D3->CreateDevice( IID_IDirect3DRGBDevice, pHAL->lpDDSRender, &pHAL->lpD3DDevice, NULL ); - if ( rc != D3D_OK ) - { - /* That was our last hope. */ - RIP( pHAL, "CreateDevice (IID_IDirect3DRGBDevice) ->", ErrorStringD3D(rc) ); - return FALSE; - } - - DPF(( DBG_CNTX_INFO, "SW Device" )); - } - } - - /*==============================================================================*/ - /* Get a copy of the render pixelformat so that wgl.c can call GetPixelInfoD3D. */ - /*==============================================================================*/ - memset( &pHAL->ddpf, 0, sizeof(DDPIXELFORMAT) ); - pHAL->ddpf.dwSize = sizeof( DDPIXELFORMAT ); - rc = pHAL->lpDDSRender->GetPixelFormat( &pHAL->ddpf ); - if ( FAILED(rc) ) - { - RIP( pHAL, "GetPixelFormat ->", ErrorStringD3D(rc) ); - return FALSE; - } - DebugPixelFormat( "Using OFFSCREEN", &pHAL->ddpf ); - DebugPixelFormat( "Using ZBUFFER", &ddsd2.ddpfPixelFormat ); - - /* Get a copy of what the D3DDevice supports for later use. */ - memset( &D3DSWDevDesc, 0, sizeof(D3DDEVICEDESC) ); - memset( &pHAL->D3DHWDevDesc, 0, sizeof(D3DDEVICEDESC) ); - D3DSWDevDesc.dwSize = sizeof( D3DDEVICEDESC ); - pHAL->D3DHWDevDesc.dwSize = sizeof( D3DDEVICEDESC ); - rc = pHAL->lpD3DDevice->GetCaps( &pHAL->D3DHWDevDesc, &D3DSWDevDesc ); - if ( FAILED(rc) ) - { - RIP( pHAL, "GetCaps ->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Get a copy of the pixel convertion stuff for direct buffer access. */ - Solve8BitChannelPixelFormat( &pHAL->ddpf, &pShared->pixel ); - AlphaBlendTableHAL( pHAL ); - - /* We must prime the Begin/End scene for SwapBuffers to work. */ - rc = pHAL->lpD3DDevice->BeginScene(); - if ( FAILED(rc) ) - { - RIP( pHAL, "BeginScene ->", ErrorStringD3D(rc) ); - return FALSE; - } - -#undef InitDDSD2 - - return TRUE; -} -/*===========================================================================*/ -/* This function will make sure a viewport is created and set for the device*/ -/* in the supplied structure. If a rect is supplied then it will be used for*/ -/* the viewport otherwise the current setting in the strucute will be used. */ -/* Note that the rect is relative to the window. So left/top must be 0,0 to */ -/* use the whole window else there is scissoring going down. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -extern "C" BOOL SetViewportHAL( PMESAD3DSHARED pShared, RECT *pRect, float minZ, float maxZ ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - D3DVIEWPORT2 vdData; - ULONG rc; - POINT pt; - - DPF(( DBG_FUNC, "SetViewportHAL();" )); - - /* Make sure we have enough info. */ - if ( !pHAL || !pHAL->lpDDSPrimary || !pHAL->lpD3DDevice ) - { - DPF(( DBG_CNTX_WARN, "SetViewport() -> NULL Pointer" )); - return FALSE; - } - - /* TODO: this is just a temp fix to stop redundant changes. */ - if ( pRect && - (pShared->rectV.left == pRect->left) && - (pShared->rectV.right == pRect->right) && - (pShared->rectV.top == pRect->top) && - (pShared->rectV.bottom == pRect->bottom) ) - { - DPF(( DBG_CNTX_WARN, "Redundant viewport" )); - return TRUE; - } - - DPF(( DBG_CNTX_INFO, "Current Viewport:" )); - DPF(( DBG_CNTX_INFO, "x: %d y: %d", pShared->rectV.left, pShared->rectV.top )); - DPF(( DBG_CNTX_INFO, "cx: %d cy: %d", (pShared->rectV.right-pShared->rectV.left), (pShared->rectV.bottom-pShared->rectV.top) )); - DPF(( DBG_CNTX_INFO, "New Viewport:" )); - DPF(( DBG_CNTX_INFO, "x: %d y: %d", pRect->left, pRect->top )); - DPF(( DBG_CNTX_INFO, "cx: %d cy: %d", (pRect->right-pRect->left), (pRect->bottom-pRect->top) )); - - /* Update the current viewport rect if one is supplied. */ - if ( pRect ) - memcpy( &pShared->rectV, pRect, sizeof(RECT) ); - - /* Build the request structure. */ - memset( &vdData, 0, sizeof(D3DVIEWPORT2) ); - vdData.dwSize = sizeof(D3DVIEWPORT2); - vdData.dwX = pShared->rectV.left; - vdData.dwY = pShared->rectV.top; - vdData.dwWidth = (pShared->rectV.right - pShared->rectV.left); - vdData.dwHeight = (pShared->rectV.bottom - pShared->rectV.top); - - if ( !vdData.dwWidth || !vdData.dwHeight ) - { - GetClientRect( pShared->hwnd, &pShared->rectW ); - pt.x = pt.y = 0; - ClientToScreen( pShared->hwnd, &pt ); - OffsetRect( &pShared->rectW, pt.x, pt.y); - vdData.dwX = pShared->rectW.left; - vdData.dwY = pShared->rectW.top; - vdData.dwWidth = (pShared->rectW.right - pShared->rectW.left); - vdData.dwHeight = (pShared->rectW.bottom - pShared->rectW.top); - memcpy( &pShared->rectV, &pShared->rectW, sizeof(RECT) ); - } - - // The dvClipX, dvClipY, dvClipWidth, dvClipHeight, dvMinZ, - // and dvMaxZ members define the non-normalized post-perspective - // 3-D view volume which is visible to the viewer. In most cases, - // dvClipX is set to -1.0 and dvClipY is set to the inverse of - // the viewport's aspect ratio on the target surface, which can be - // calculated by dividing the dwHeight member by dwWidth. Similarly, - // the dvClipWidth member is typically 2.0 and dvClipHeight is set - // to twice the aspect ratio set in dwClipY. The dvMinZ and dvMaxZ - // are usually set to 0.0 and 1.0. - vdData.dvClipX = -1.0f; - vdData.dvClipWidth = 2.0f; - vdData.dvClipY = 1.0f; - vdData.dvClipHeight = 2.0f; - vdData.dvMaxZ = maxZ; - vdData.dvMinZ = minZ; - - DPF(( DBG_CNTX_INFO, "zMin: %f zMax: %f", minZ, maxZ )); - - /* I'm going to destroy the viewport everytime as when we size we will */ - /* have a new D3DDevice. As this area doesn't need to be fast... */ - if ( pHAL->lpViewport ) - { - DPF(( DBG_CNTX_INFO, "DeleteViewport" )); - - pHAL->lpD3DDevice->DeleteViewport( pHAL->lpViewport ); - rc = pHAL->lpViewport->Release(); - pHAL->lpViewport = NULL; - } - - rc = pHAL->lpD3D3->CreateViewport( &pHAL->lpViewport, NULL ); - if ( rc != D3D_OK ) - { - DPF(( DBG_CNTX_ERROR, "CreateViewport Failed" )); - return FALSE; - } - - /* Update the device with the new viewport. */ - pHAL->lpD3DDevice->AddViewport( pHAL->lpViewport ); - pHAL->lpViewport->SetViewport2( &vdData ); - pHAL->lpD3DDevice->SetCurrentViewport( pHAL->lpViewport ); - - return TRUE; -} -/*===========================================================================*/ -/* */ -/* */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -HRESULT WINAPI EnumSurfacesHook( LPDIRECTDRAWSURFACE4 lpDDS, LPDDSURFACEDESC2 lpDDSDesc, LPVOID pVoid ) -{ - DDSURFACEDESC2 *pddsd2 = (DDSURFACEDESC2 *)pVoid; - - DPF(( DBG_FUNC, "EnumSurfacesHook();" )); - - if ( (lpDDSDesc->ddpfPixelFormat.dwFlags == pddsd2->ddpfPixelFormat.dwFlags) && (lpDDSDesc->ddsCaps.dwCaps == pddsd2->ddsCaps.dwCaps) ) - { - /* Save the pixelformat now so that we know we have one. */ - memcpy( pddsd2, lpDDSDesc, sizeof(DDSURFACEDESC2) ); - - return D3DENUMRET_CANCEL; - } - - return D3DENUMRET_OK; -} -/*===========================================================================*/ -/* This is the callback proc to get a Z-Buffer. Thats it. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -HRESULT CALLBACK EnumZBufferHook( DDPIXELFORMAT* pddpf, VOID *pVoid ) -{ - DDPIXELFORMAT *pddpfChoice = (DDPIXELFORMAT *)pVoid; - - DPF(( DBG_FUNC, "EnumZBufferHook();" )); - - /* If this is ANY type of depth-buffer, stop. */ - if( pddpf->dwFlags == DDPF_ZBUFFER ) - { - /* Save the pixelformat now so that we know we have one. */ - memcpy( pddpfChoice, pddpf, sizeof(DDPIXELFORMAT) ); - - /* I feel if the hardware supports this low then lets use it. Could get ugly. */ - if( pddpf->dwZBufferBitDepth >= 8 ) - { - return D3DENUMRET_CANCEL; - } - } - - return D3DENUMRET_OK; -} -/*===========================================================================*/ -/* This function handles the callback for the D3DDevice enumeration. Good */ -/* god who's idea was this? The D3D wrapper has two variable related to what*/ -/* kind of device we want and have. First we have a Bool that is set if we */ -/* have allocated a HW device. We always look for the HW device first. The */ -/* other variable is used to force SW. If we have run into a case that we */ -/* want to fallback to SW then we set this. We will fallback if we cannot */ -/* texture in video memory (among others). */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -HRESULT CALLBACK EnumDeviceHook( GUID FAR* lpGuid, LPSTR lpDesc, LPSTR lpName, LPD3DDEVICEDESC lpD3DHWDesc, LPD3DDEVICEDESC lpD3DHELDesc, void *pVoid ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pVoid; - LPD3DDEVICEDESC pChoice = lpD3DHWDesc; - - DPF(( DBG_FUNC, "EnumDeviceHook();" )); - - /* Determine if which device description is valid. */ - if ( pChoice->dcmColorModel == 0 ) - pChoice = lpD3DHELDesc; - - /* Make sure we always have a GUID. */ - memcpy( &pHAL->guid, lpGuid, sizeof(GUID) ); - - /* This controls whether we will except HW or not. */ - if ( pHAL->shared.bForceSW == TRUE ) - { - return (pChoice == lpD3DHELDesc) ? D3DENUMRET_CANCEL : D3DENUMRET_OK; - } - - /* Always try for hardware. */ - if ( pChoice == lpD3DHWDesc ) - { - return D3DENUMRET_CANCEL; - } - - return D3DENUMRET_OK; -} -/*===========================================================================*/ -/* This function will destroy any and all surfaces that this context has */ -/* allocated. If there is a clipper object then it will also be destoryed as*/ -/* it is part of the Primary Surface. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void DestroyAllSurfaces( PMESAD3DHAL pHAL ) -{ - LONG refCount; - - DPF(( DBG_FUNC, "DestroyAllSurfaces();" )); - - DX_RESTORE( pHAL->lpDDSPrimary ); - DX_RESTORE( pHAL->lpDDSRender ); - DX_RESTORE( pHAL->lpDDSZbuffer); - - if ( pHAL->lpDDSRender ) - { - pHAL->lpDDSRender->Unlock( NULL ); - - /* If this isn't a Flipable surface then we must clean up the render. */ - if ( pHAL->shared.bFlipable == FALSE) - { - if ( pHAL->lpDDSZbuffer ) - { - DPF(( DBG_CNTX_INFO, "Remove attached surfaces from RENDER" )); - pHAL->lpDDSRender->DeleteAttachedSurface( 0, NULL ); - } - - DPF(( DBG_CNTX_INFO, "Release RENDER" )); - refCount = pHAL->lpDDSRender->Release(); - pHAL->lpDDSRender = NULL; - } - } - - if ( pHAL->lpDDSZbuffer ) - { - DPF(( DBG_CNTX_INFO, "Release ZBuffer" )); - pHAL->lpDDSZbuffer->Unlock( NULL ); - refCount = pHAL->lpDDSZbuffer->Release(); - pHAL->lpDDSZbuffer = NULL; - } - - if ( pHAL->lpClipper ) - { - DPF(( DBG_CNTX_INFO, "Release Clipper" )); - refCount = pHAL->lpClipper->Release(); - pHAL->lpClipper = NULL; - } - - if ( pHAL->lpDDSPrimary ) - { - pHAL->lpDDSPrimary->Unlock( NULL ); - - DPF(( DBG_CNTX_INFO, "Release PRIMARY" )); - refCount = pHAL->lpDDSPrimary->Release(); - pHAL->lpDDSPrimary = NULL; - } -} -/*===========================================================================*/ -/* This function will destroy the current D3DDevice and any resources that */ -/* belong to it. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void DestroyDevice( PMESAD3DHAL pHAL ) -{ - LONG refCount; - - DPF(( DBG_FUNC, "DestroyDevice();" )); - - /* Kill the D3D stuff if exists. */ - if ( pHAL->lpViewport ) - { - DPF(( DBG_CNTX_INFO, "Delete Viewport" )); - pHAL->lpD3DDevice->DeleteViewport( pHAL->lpViewport ); - - DPF(( DBG_CNTX_INFO, "Release Viewport" )); - refCount = pHAL->lpViewport->Release(); - pHAL->lpViewport = NULL; - } - - if ( pHAL->lpD3DDevice != NULL ) - { - DPF(( DBG_CNTX_INFO, "Release D3DDevice" )); - refCount = pHAL->lpD3DDevice->EndScene(); - refCount = pHAL->lpD3DDevice->Release(); - pHAL->lpD3DDevice = NULL; - } -} -/*===========================================================================*/ -/* This function will destroy the current D3DDevice and any resources that */ -/* belong to it. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void DestroyInterfaces( PMESAD3DHAL pHAL ) -{ - LONG refCount; - - DPF(( DBG_FUNC, "DestroyInterfaces();" )); - - if ( pHAL->lpD3D3 != NULL ) - { - DPF(( DBG_CNTX_INFO, "Release Direct3D3" )); - refCount = pHAL->lpD3D3->Release(); - pHAL->lpD3D3 = NULL; - } - - if ( pHAL->lpDD4 != NULL ) - { - DPF(( DBG_CNTX_INFO, "Release DDraw4" )); - refCount = pHAL->lpDD4->Release(); - pHAL->lpDD4 = NULL; - } - - if ( pHAL->lpDD != NULL ) - { - DPF(( DBG_CNTX_INFO, "Release DDraw" )); - refCount = pHAL->lpDD->Release(); - pHAL->lpDD = NULL; - } -} -/*===========================================================================*/ -/* This function will first send (not post) a message to the client window */ -/* that this context is using. The client will respond by unbinding itself */ -/* and binding the 'default' context. This allows the API to be supported */ -/* until the window can be destroyed. Finally we post the quit message to */ -/* the client in hopes to end the application. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void FatalShutDown( PMESAD3DHAL pHAL ) -{ - /* Whip this baby in too try and support the API until we die... */ - if ( pHAL ) - SendMessage( pHAL->shared.hwnd, UM_FATALSHUTDOWN, 0L, 0L ); - - /* Close the client application down. */ - PostQuitMessage( 0 ); -} - +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver Build 5 */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#include "D3DHAL.h" +/*===========================================================================*/ +/* Local function prototypes. */ +/*===========================================================================*/ +static void DestroyAllSurfaces( PMESAD3DHAL pHAL ); +static void DestroyDevice( PMESAD3DHAL pHAL ); +static void DestroyInterfaces( PMESAD3DHAL pHAL ); + +HRESULT WINAPI EnumSurfacesHook( LPDIRECTDRAWSURFACE4 lpDDS, LPDDSURFACEDESC2 lpDDSDesc, LPVOID pVoid ); +HRESULT CALLBACK EnumZBufferHook( DDPIXELFORMAT* pddpf, VOID *pVoid ); +HRESULT CALLBACK EnumDeviceHook( GUID FAR* lpGuid, LPSTR lpDesc, LPSTR lpName, LPD3DDEVICEDESC lpD3DHWDesc, LPD3DDEVICEDESC lpD3DHELDesc, void *pVoid ); +/*===========================================================================*/ +/* Globals. */ +/*===========================================================================*/ +//char *errorMsg; +/*===========================================================================*/ +/* This function is responable for allocating the actual MESAD3DHAL struct. */ +/* Each Mesa context will have its own MESAD3DHAL struct so its like a mini */ +/* context to some extent. All one time allocations/operations get done here.*/ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +extern "C" PMESAD3DSHARED InitHAL( HWND hwnd ) +{ + PMESAD3DHAL pHAL; + ULONG rc; + + DPF(( DBG_FUNC, "InitHAL();" )); + DPF(( DBG_CNTX_INFO, "hwnd: %d", hwnd )); + + /* Allocate the structure and zero it out. */ + pHAL = (PMESAD3DHAL)ALLOC( sizeof(MESAD3DHAL) ); + if ( pHAL == NULL ) + { + RIP( pHAL, "InitHAL->", "Memory Allocation" ); + return (PMESAD3DSHARED)NULL; + } + memset( pHAL, 0, sizeof(MESAD3DHAL) ); + + /* Get the texture manager going. */ + rc = InitTMgrHAL( pHAL ); + if ( rc == FALSE ) + { + RIP( pHAL, "InitTMgrHAL->", "Failed" ); + return (PMESAD3DSHARED)NULL; + } + + /* Fill in the window parameters if we can. */ + pHAL->shared.hwnd = hwnd; + + /* Parse the user's enviroment variables to generate a debug mask. */ + ReadDBGEnv(); + + return (PMESAD3DSHARED)pHAL; +} +/*===========================================================================*/ +/* This function will unload all the resources that the MESAD3DHAL struct */ +/* has bound to it. The actual structure itself will be freed. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void TermHAL( PMESAD3DSHARED pShared ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + + DPF(( DBG_FUNC, "TermHAL();" )); + + /* Check for an empty wrapper structure. */ + if ( pHAL == NULL ) + return; + + /* Kill this texture manager. */ + TermTMgrHAL( pHAL ); + + /* Kill any DDraw stuff if exists. */ + DestroyDevice( pHAL ); + DestroyAllSurfaces( pHAL ); + DestroyInterfaces( pHAL ); + + FREE( pHAL ); +} +/*===========================================================================*/ +/* This function is used to init and resize the rendering surface as the two*/ +/* are almost the same. First the device and all the surfaces are destoryed */ +/* if they already exist. Next we create a OffScreen rendering surface and */ +/* save some pixelformat info to do color convertions. Next we start to take */ +/* care of getting the most out of the hardware. I use bHardware to determine*/ +/* the state of the device we found in the device enumeration. The enum proc*/ +/* will try for hardware first. I next use a bForceSW to make the enum proc */ +/* choose a software device. So I will try some combinations with HW first */ +/* until I feel I have to set the bForceSW and call this function again. If */ +/* this function is called with no width or height then use the internals. */ +/* NOTE: The worst case is that all will be in SW (RGBDevice) and really */ +/* I should forget the whole thing and fall back to a DDraw span type*/ +/* rendering but what is the point. This way I always know I have a */ +/* D3DDevice and that makes things easier. I do impliment the span */ +/* rendering function for stuff that I haven't done support for such */ +/* as points and lines. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE */ +/*===========================================================================*/ +extern "C" BOOL CreateHAL( PMESAD3DSHARED pShared ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + DDSURFACEDESC2 ddsd2; + D3DDEVICEDESC D3DSWDevDesc; + DDSCAPS2 ddscaps; + DWORD dwCoopFlags, + dwWidth, + dwHeight; + ULONG rc; + + DPF(( DBG_FUNC, "CreateHAL();" )); + +#define InitDDSD2(f) memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); \ + ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); \ + ddsd2.dwFlags = f; + + if ( pHAL == NULL ) + return FALSE; + + /* Use the internal rectangle struct. */ + dwWidth = pShared->rectW.right - pShared->rectW.left; + dwHeight = pShared->rectW.bottom - pShared->rectW.top; + + DPF(( DBG_CNTX_INFO, "Width: %d Height: %d", dwWidth, dwHeight )); + + /* The dimensions might still be the same so just leave. */ + if ( (dwWidth == pShared->dwWidth) && (dwHeight == pShared->dwHeight) ) + { + DPF(( DBG_CNTX_WARN, "Context size hasn't changed" )); + return TRUE; + } + + /* If one of the dimensions are zero then leave. WM_SIZE should get us back here. */ + if ( (dwWidth == 0) || (dwHeight == 0) ) + return TRUE; + + /* Save the renders dimensions. */ + pShared->dwWidth = dwWidth; + pShared->dwHeight = dwHeight; + + DPF(( DBG_CNTX_INFO, "Creating Context:\n cx:%d cy:%d", pShared->dwWidth, pShared->dwHeight )); + + /*=================================*/ + /* Create all required interfaces. */ + /*=================================*/ + + /* Kill any DDraw stuff if exists. */ + DestroyDevice( pHAL ); + DestroyAllSurfaces( pHAL ); + DestroyInterfaces( pHAL ); + + /* Create a instance of DDraw using the Primary display driver. */ + rc = DirectDrawCreate( NULL, &pHAL->lpDD, NULL ); + if( FAILED(rc) ) + { + RIP( pHAL, "DirectDrawCreate->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Get the DDraw4 interface. */ + rc = pHAL->lpDD->QueryInterface( IID_IDirectDraw4, (void **)&pHAL->lpDD4 ); + if( FAILED(rc) ) + { + RIP( pHAL, "QueryInterface (IID_IDirectDraw4) ->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Get the Direct3D3 interface. */ + rc = pHAL->lpDD4->QueryInterface( IID_IDirect3D3, (void **)&pHAL->lpD3D3 ); + if( FAILED(rc) ) + { + RIP( pHAL, "QueryInterface (IID_IDirect3D3) ->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Set the Cooperative level. NOTE: we need to know if we are FS at this point.*/ + dwCoopFlags = (pShared->bWindow == TRUE) ? DDSCL_NORMAL : (DDSCL_EXCLUSIVE | DDSCL_FULLSCREEN); + rc = pHAL->lpDD4->SetCooperativeLevel( pShared->hwnd, dwCoopFlags ); + if ( FAILED(rc) ) + { + RIP( pHAL, "SetCooperativeLevel->", ErrorStringD3D(rc) ); + return FALSE; + } + + /*==================================================================*/ + /* Get the best device we can and note whether its hardware or not. */ + /*==================================================================*/ + pShared->bForceSW = FALSE; + pHAL->lpD3D3->EnumDevices( EnumDeviceHook, (void *)pHAL ); + pShared->bHardware = IsEqualIID( pHAL->guid, IID_IDirect3DHALDevice ); + DPF(( DBG_CNTX_INFO, "bHardware: %s", (pShared->bHardware) ? "TRUE" : "FALSE" )); + DPF(( DBG_CNTX_INFO, "bWindowed: %s", (pShared->bWindow) ? "TRUE" : "FALSE" )); + + /*========================================================================*/ + /* HARDWARE was found. */ + /*========================================================================*/ + if ( pShared->bHardware == TRUE ) + { + /*===================================*/ + /* HARDWARE -> Z-BUFFER. */ + /*===================================*/ + + /* Get a Z-Buffer pixelformat. */ + memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); + ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); + rc = pHAL->lpD3D3->EnumZBufferFormats( pHAL->guid, EnumZBufferHook, (VOID*)&ddsd2.ddpfPixelFormat ); + if ( FAILED(rc) ) + { + RIP( pHAL, "EnumZBufferFormatsl->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Setup our request structure for the Z-buffer surface. */ + ddsd2.dwFlags = DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT | DDSD_PIXELFORMAT; + ddsd2.ddsCaps.dwCaps = DDSCAPS_ZBUFFER | DDSCAPS_VIDEOMEMORY; + ddsd2.dwWidth = dwWidth; + ddsd2.dwHeight = dwHeight; + rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSZbuffer, NULL ); + if ( !FAILED(rc) ) + { + DPF(( DBG_CNTX_INFO, "HW ZBuffer" )); + + /*===================================*/ + /* HARDWARE -> Z-BUFFER -> FLIPABLE */ + /*===================================*/ + if ( pShared->bWindow == FALSE ) + { + InitDDSD2( DDSD_CAPS | DDSD_BACKBUFFERCOUNT ); + ddsd2.dwBackBufferCount = 1; + ddsd2.ddsCaps.dwCaps = DDSCAPS_PRIMARYSURFACE | DDSCAPS_3DDEVICE | DDSCAPS_FLIP | DDSCAPS_COMPLEX; + rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSPrimary, NULL ); + if ( FAILED(rc) ) + { + /* Make sure we try the next fall back. */ + DPF(( DBG_CNTX_WARN, "HW Flip/Complex not available" )); + pHAL->lpDDSPrimary = NULL; + } + else + { + /* Get the back buffer that was created. */ + ddscaps.dwCaps = DDSCAPS_BACKBUFFER; + rc = pHAL->lpDDSPrimary->GetAttachedSurface( &ddscaps, &pHAL->lpDDSRender ); + if ( FAILED(rc) ) + { + DPF(( DBG_CNTX_WARN, "GetAttachedSurface failed -> HW Flip/Complex" )); + + /* Make sure we try the next fall back. */ + pHAL->lpDDSPrimary->Release(); + pHAL->lpDDSPrimary = NULL; + } + else + { + /* I have had problems when a complex surface comes back */ + /* with the back buffer being created in SW. Not sure why */ + /* or how this is possable but I'm checking for it here. */ + memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); + ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); + DX_RESTORE( pHAL->lpDDSRender ); + rc = pHAL->lpDDSRender->GetSurfaceDesc( &ddsd2 ); + if ( FAILED(rc) ) + { + RIP( pHAL, "GetSurfaceDesc (RENDER) ->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* If the surface is in VID then we are happy with are Flipable. */ + if ( ddsd2.ddsCaps.dwCaps & DDSCAPS_LOCALVIDMEM ) + { + pShared->bFlipable = TRUE; + DPF(( DBG_CNTX_INFO, "HW Flip/Complex!" )); + } + else + { + /* Kill this setup. */ + pHAL->lpDDSPrimary->Release(); + pHAL->lpDDSPrimary = NULL; + } + } + } + } + + /*===================================*/ + /* HARDWARE -> Z-BUFFER -> BLT */ + /*===================================*/ + if ( pHAL->lpDDSPrimary == NULL ) + { + pShared->bFlipable = FALSE; + + /* Create the Primary (front buffer). */ + InitDDSD2( DDSD_CAPS ); + ddsd2.ddsCaps.dwCaps = DDSCAPS_PRIMARYSURFACE; + rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSPrimary, NULL ); + if ( FAILED(rc) ) + { + /* This is an error as we should be able to do this at minimum. */ + RIP( pHAL, "CreateSurface (PRIMARY) ->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Create the Render (back buffer). */ + InitDDSD2( DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT ); + ddsd2.dwWidth = dwWidth; + ddsd2.dwHeight = dwHeight; + ddsd2.ddsCaps.dwCaps = DDSCAPS_OFFSCREENPLAIN | DDSCAPS_3DDEVICE; + rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSRender, NULL ); + if ( FAILED(rc) ) + { + DPF(( DBG_CNTX_WARN, "Failed HW Offscreen surface" )); + + /* Make sure we try the next fall back. */ + pHAL->lpDDSPrimary->Release(); + pHAL->lpDDSPrimary = NULL; + } + else + { + /* Might as well check here too see if this surface is in */ + /* hardware. If nothing else just to be consistant. */ + memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); + ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); + DX_RESTORE( pHAL->lpDDSRender ); + rc = pHAL->lpDDSRender->GetSurfaceDesc( &ddsd2 ); + if ( FAILED(rc) ) + { + RIP( pHAL, "GetSurfaceDesc (RENDER) ->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* If the surface is in VID then we are happy. */ + if ( ddsd2.ddsCaps.dwCaps & DDSCAPS_LOCALVIDMEM ) + { + /* Create a clipper object so that DDraw will be able to blt windows that */ + /* have been clipped by the screen or other windows. */ + pHAL->lpDD4->CreateClipper( 0, &pHAL->lpClipper, NULL ); + pHAL->lpClipper->SetHWnd( 0, pShared->hwnd ); + pHAL->lpDDSPrimary->SetClipper( pHAL->lpClipper ); + pHAL->lpClipper->Release(); + DPF(( DBG_CNTX_INFO, "HW RENDER surface" )); + } + else + { + /* Kill this setup. */ + pHAL->lpDDSRender->Release(); + pHAL->lpDDSRender = NULL; + pHAL->lpDDSPrimary->Release(); + pHAL->lpDDSPrimary = NULL; + } + } + } + + /*===================================*/ + /* Create D3DDEVICE -> HARDWARE. */ + /*===================================*/ + if ( pHAL->lpDDSZbuffer && pHAL->lpDDSPrimary && pHAL->lpDDSRender ) + { + DX_RESTORE( pHAL->lpDDSRender ); + DX_RESTORE( pHAL->lpDDSZbuffer ); + + rc = pHAL->lpDDSRender->AddAttachedSurface( pHAL->lpDDSZbuffer ); + if ( FAILED(rc) ) + { + RIP( pHAL, "AddAttachedSurface (ZBUFFER) ->", ErrorStringD3D(rc) ); + return FALSE; + } + + rc = pHAL->lpD3D3->CreateDevice( IID_IDirect3DHALDevice, pHAL->lpDDSRender, &pHAL->lpD3DDevice, NULL ); + if ( rc != D3D_OK ) + { + DPF(( DBG_CNTX_WARN, "Failed HW Device" )); + pHAL->lpD3DDevice = NULL; + } + else + { + DPF(( DBG_CNTX_INFO, "HW Device" )); + } + } + } + } + + /*========================================================================*/ + /* SOFTWARE fallback. */ + /*========================================================================*/ + if ( pHAL->lpD3DDevice == NULL ) + { + DPF(( DBG_CNTX_INFO, "SW fallback :(" )); + + /* Make sure we have no surfaces allocated. Just incase. */ + DestroyAllSurfaces( pHAL ); + + /* Get a software device. */ + pShared->bFlipable = FALSE; + pShared->bForceSW = TRUE; + pHAL->lpD3D3->EnumDevices( EnumDeviceHook, (void *)pHAL ); + pShared->bHardware = IsEqualIID( pHAL->guid, IID_IDirect3DHALDevice ); + + /*===================================*/ + /* SOFTWARE -> Z-BUFFER. */ + /*===================================*/ + + /*===================================*/ + /* SOFTWARE -> Z-BUFFER -> FLIPABLE */ + /*===================================*/ + if ( pShared->bWindow == FALSE ) + { + InitDDSD2( DDSD_CAPS | DDSD_BACKBUFFERCOUNT ); + ddsd2.dwBackBufferCount = 1; + ddsd2.ddsCaps.dwCaps = DDSCAPS_PRIMARYSURFACE | DDSCAPS_3DDEVICE | DDSCAPS_FLIP | DDSCAPS_COMPLEX; + ddsd2.ddpfPixelFormat.dwSize = sizeof( DDPIXELFORMAT ); + ddsd2.ddpfPixelFormat.dwFlags = (DDPF_RGB | DDPF_ALPHAPIXELS); + rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSPrimary, NULL ); + if ( FAILED(rc) ) + { + DPF(( DBG_CNTX_WARN, "Failed SW Flip/Complex" )); + + /* Make sure we try the next fall back. */ + pHAL->lpDDSPrimary = NULL; + } + else + { + ddscaps.dwCaps = DDSCAPS_BACKBUFFER; + rc = pHAL->lpDDSPrimary->GetAttachedSurface( &ddscaps, &pHAL->lpDDSRender ); + if ( FAILED(rc) ) + { + /* Make sure we try the next fall back. */ + DPF(( DBG_CNTX_WARN, "GetAttachedSurface failed -> SW Flip/Complex" )); + pHAL->lpDDSPrimary->Release(); + pHAL->lpDDSPrimary = NULL; + } + else + { + DPF(( DBG_CNTX_INFO, "SW Flip/Complex" )); + pShared->bFlipable = TRUE; + } + } + } + + /*===================================*/ + /* SOFTWARE -> Z-BUFFER -> BLT */ + /*===================================*/ + if ( pHAL->lpDDSPrimary == NULL ) + { + /* Create the Primary (front buffer). */ + InitDDSD2( DDSD_CAPS ); + ddsd2.ddsCaps.dwCaps = DDSCAPS_PRIMARYSURFACE; + rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSPrimary, NULL ); + if ( FAILED(rc) ) + { + /* This is an error as we should be able to do this at minimum. */ + RIP( pHAL, "CreateSurface (PRIMARY) ->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Create the Render (back buffer). */ + InitDDSD2( DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT ); + ddsd2.dwWidth = dwWidth; + ddsd2.dwHeight = dwHeight; + ddsd2.ddsCaps.dwCaps = DDSCAPS_OFFSCREENPLAIN | DDSCAPS_3DDEVICE; + ddsd2.ddpfPixelFormat.dwSize = sizeof( DDPIXELFORMAT ); + ddsd2.ddpfPixelFormat.dwFlags = (DDPF_RGB | DDPF_ALPHAPIXELS); + rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pHAL->lpDDSRender, NULL ); + if ( FAILED(rc) ) + { + /* That was our last hope. */ + RIP( pHAL, "CreateSurface (RENDER) ->", ErrorStringD3D(rc) ); + return FALSE; + } + else + { + DPF(( DBG_CNTX_INFO, "SW RENDER surface" )); + + /* Create a clipper object so that DDraw will be able to blt windows that */ + /* have been clipped by the screen or other windows. */ + pHAL->lpDD4->CreateClipper( 0, &pHAL->lpClipper, NULL ); + pHAL->lpClipper->SetHWnd( 0, pShared->hwnd ); + pHAL->lpDDSPrimary->SetClipper( pHAL->lpClipper ); + pHAL->lpClipper->Release(); + } + } + + /*===================================*/ + /* Create D3DDEVICE -> SOFTWARE. */ + /*===================================*/ + if ( pHAL->lpDDSPrimary && pHAL->lpDDSRender ) + { + DX_RESTORE( pHAL->lpDDSRender ); + rc = pHAL->lpD3D3->CreateDevice( IID_IDirect3DRGBDevice, pHAL->lpDDSRender, &pHAL->lpD3DDevice, NULL ); + if ( rc != D3D_OK ) + { + /* That was our last hope. */ + RIP( pHAL, "CreateDevice (IID_IDirect3DRGBDevice) ->", ErrorStringD3D(rc) ); + return FALSE; + } + + DPF(( DBG_CNTX_INFO, "SW Device" )); + } + } + + /*==============================================================================*/ + /* Get a copy of the render pixelformat so that wgl.c can call GetPixelInfoD3D. */ + /*==============================================================================*/ + memset( &pHAL->ddpf, 0, sizeof(DDPIXELFORMAT) ); + pHAL->ddpf.dwSize = sizeof( DDPIXELFORMAT ); + rc = pHAL->lpDDSRender->GetPixelFormat( &pHAL->ddpf ); + if ( FAILED(rc) ) + { + RIP( pHAL, "GetPixelFormat ->", ErrorStringD3D(rc) ); + return FALSE; + } + DebugPixelFormat( "Using OFFSCREEN", &pHAL->ddpf ); + DebugPixelFormat( "Using ZBUFFER", &ddsd2.ddpfPixelFormat ); + + /* Get a copy of what the D3DDevice supports for later use. */ + memset( &D3DSWDevDesc, 0, sizeof(D3DDEVICEDESC) ); + memset( &pHAL->D3DHWDevDesc, 0, sizeof(D3DDEVICEDESC) ); + D3DSWDevDesc.dwSize = sizeof( D3DDEVICEDESC ); + pHAL->D3DHWDevDesc.dwSize = sizeof( D3DDEVICEDESC ); + rc = pHAL->lpD3DDevice->GetCaps( &pHAL->D3DHWDevDesc, &D3DSWDevDesc ); + if ( FAILED(rc) ) + { + RIP( pHAL, "GetCaps ->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Get a copy of the pixel convertion stuff for direct buffer access. */ + Solve8BitChannelPixelFormat( &pHAL->ddpf, &pShared->pixel ); + AlphaBlendTableHAL( pHAL ); + + /* We must prime the Begin/End scene for SwapBuffers to work. */ + rc = pHAL->lpD3DDevice->BeginScene(); + if ( FAILED(rc) ) + { + RIP( pHAL, "BeginScene ->", ErrorStringD3D(rc) ); + return FALSE; + } + +#undef InitDDSD2 + + return TRUE; +} +/*===========================================================================*/ +/* This function will make sure a viewport is created and set for the device*/ +/* in the supplied structure. If a rect is supplied then it will be used for*/ +/* the viewport otherwise the current setting in the strucute will be used. */ +/* Note that the rect is relative to the window. So left/top must be 0,0 to */ +/* use the whole window else there is scissoring going down. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +extern "C" BOOL SetViewportHAL( PMESAD3DSHARED pShared, RECT *pRect, float minZ, float maxZ ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + D3DVIEWPORT2 vdData; + ULONG rc; + POINT pt; + + DPF(( DBG_FUNC, "SetViewportHAL();" )); + + /* Make sure we have enough info. */ + if ( !pHAL || !pHAL->lpDDSPrimary || !pHAL->lpD3DDevice ) + { + DPF(( DBG_CNTX_WARN, "SetViewport() -> NULL Pointer" )); + return FALSE; + } + + /* TODO: this is just a temp fix to stop redundant changes. */ + if ( pRect && + (pShared->rectV.left == pRect->left) && + (pShared->rectV.right == pRect->right) && + (pShared->rectV.top == pRect->top) && + (pShared->rectV.bottom == pRect->bottom) ) + { + DPF(( DBG_CNTX_WARN, "Redundant viewport" )); + return TRUE; + } + + DPF(( DBG_CNTX_INFO, "Current Viewport:" )); + DPF(( DBG_CNTX_INFO, "x: %d y: %d", pShared->rectV.left, pShared->rectV.top )); + DPF(( DBG_CNTX_INFO, "cx: %d cy: %d", (pShared->rectV.right-pShared->rectV.left), (pShared->rectV.bottom-pShared->rectV.top) )); + DPF(( DBG_CNTX_INFO, "New Viewport:" )); + DPF(( DBG_CNTX_INFO, "x: %d y: %d", pRect->left, pRect->top )); + DPF(( DBG_CNTX_INFO, "cx: %d cy: %d", (pRect->right-pRect->left), (pRect->bottom-pRect->top) )); + + /* Update the current viewport rect if one is supplied. */ + if ( pRect ) + memcpy( &pShared->rectV, pRect, sizeof(RECT) ); + + /* Build the request structure. */ + memset( &vdData, 0, sizeof(D3DVIEWPORT2) ); + vdData.dwSize = sizeof(D3DVIEWPORT2); + vdData.dwX = pShared->rectV.left; + vdData.dwY = pShared->rectV.top; + vdData.dwWidth = (pShared->rectV.right - pShared->rectV.left); + vdData.dwHeight = (pShared->rectV.bottom - pShared->rectV.top); + + if ( !vdData.dwWidth || !vdData.dwHeight ) + { + GetClientRect( pShared->hwnd, &pShared->rectW ); + pt.x = pt.y = 0; + ClientToScreen( pShared->hwnd, &pt ); + OffsetRect( &pShared->rectW, pt.x, pt.y); + vdData.dwX = pShared->rectW.left; + vdData.dwY = pShared->rectW.top; + vdData.dwWidth = (pShared->rectW.right - pShared->rectW.left); + vdData.dwHeight = (pShared->rectW.bottom - pShared->rectW.top); + memcpy( &pShared->rectV, &pShared->rectW, sizeof(RECT) ); + } + + // The dvClipX, dvClipY, dvClipWidth, dvClipHeight, dvMinZ, + // and dvMaxZ members define the non-normalized post-perspective + // 3-D view volume which is visible to the viewer. In most cases, + // dvClipX is set to -1.0 and dvClipY is set to the inverse of + // the viewport's aspect ratio on the target surface, which can be + // calculated by dividing the dwHeight member by dwWidth. Similarly, + // the dvClipWidth member is typically 2.0 and dvClipHeight is set + // to twice the aspect ratio set in dwClipY. The dvMinZ and dvMaxZ + // are usually set to 0.0 and 1.0. + vdData.dvClipX = -1.0f; + vdData.dvClipWidth = 2.0f; + vdData.dvClipY = 1.0f; + vdData.dvClipHeight = 2.0f; + vdData.dvMaxZ = maxZ; + vdData.dvMinZ = minZ; + + DPF(( DBG_CNTX_INFO, "zMin: %f zMax: %f", minZ, maxZ )); + + /* I'm going to destroy the viewport everytime as when we size we will */ + /* have a new D3DDevice. As this area doesn't need to be fast... */ + if ( pHAL->lpViewport ) + { + DPF(( DBG_CNTX_INFO, "DeleteViewport" )); + + pHAL->lpD3DDevice->DeleteViewport( pHAL->lpViewport ); + rc = pHAL->lpViewport->Release(); + pHAL->lpViewport = NULL; + } + + rc = pHAL->lpD3D3->CreateViewport( &pHAL->lpViewport, NULL ); + if ( rc != D3D_OK ) + { + DPF(( DBG_CNTX_ERROR, "CreateViewport Failed" )); + return FALSE; + } + + /* Update the device with the new viewport. */ + pHAL->lpD3DDevice->AddViewport( pHAL->lpViewport ); + pHAL->lpViewport->SetViewport2( &vdData ); + pHAL->lpD3DDevice->SetCurrentViewport( pHAL->lpViewport ); + + return TRUE; +} +/*===========================================================================*/ +/* */ +/* */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +HRESULT WINAPI EnumSurfacesHook( LPDIRECTDRAWSURFACE4 lpDDS, LPDDSURFACEDESC2 lpDDSDesc, LPVOID pVoid ) +{ + DDSURFACEDESC2 *pddsd2 = (DDSURFACEDESC2 *)pVoid; + + DPF(( DBG_FUNC, "EnumSurfacesHook();" )); + + if ( (lpDDSDesc->ddpfPixelFormat.dwFlags == pddsd2->ddpfPixelFormat.dwFlags) && (lpDDSDesc->ddsCaps.dwCaps == pddsd2->ddsCaps.dwCaps) ) + { + /* Save the pixelformat now so that we know we have one. */ + memcpy( pddsd2, lpDDSDesc, sizeof(DDSURFACEDESC2) ); + + return D3DENUMRET_CANCEL; + } + + return D3DENUMRET_OK; +} +/*===========================================================================*/ +/* This is the callback proc to get a Z-Buffer. Thats it. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +HRESULT CALLBACK EnumZBufferHook( DDPIXELFORMAT* pddpf, VOID *pVoid ) +{ + DDPIXELFORMAT *pddpfChoice = (DDPIXELFORMAT *)pVoid; + + DPF(( DBG_FUNC, "EnumZBufferHook();" )); + + /* If this is ANY type of depth-buffer, stop. */ + if( pddpf->dwFlags == DDPF_ZBUFFER ) + { + /* Save the pixelformat now so that we know we have one. */ + memcpy( pddpfChoice, pddpf, sizeof(DDPIXELFORMAT) ); + + /* I feel if the hardware supports this low then lets use it. Could get ugly. */ + if( pddpf->dwZBufferBitDepth >= 8 ) + { + return D3DENUMRET_CANCEL; + } + } + + return D3DENUMRET_OK; +} +/*===========================================================================*/ +/* This function handles the callback for the D3DDevice enumeration. Good */ +/* god who's idea was this? The D3D wrapper has two variable related to what*/ +/* kind of device we want and have. First we have a Bool that is set if we */ +/* have allocated a HW device. We always look for the HW device first. The */ +/* other variable is used to force SW. If we have run into a case that we */ +/* want to fallback to SW then we set this. We will fallback if we cannot */ +/* texture in video memory (among others). */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +HRESULT CALLBACK EnumDeviceHook( GUID FAR* lpGuid, LPSTR lpDesc, LPSTR lpName, LPD3DDEVICEDESC lpD3DHWDesc, LPD3DDEVICEDESC lpD3DHELDesc, void *pVoid ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pVoid; + LPD3DDEVICEDESC pChoice = lpD3DHWDesc; + + DPF(( DBG_FUNC, "EnumDeviceHook();" )); + + /* Determine if which device description is valid. */ + if ( pChoice->dcmColorModel == 0 ) + pChoice = lpD3DHELDesc; + + /* Make sure we always have a GUID. */ + memcpy( &pHAL->guid, lpGuid, sizeof(GUID) ); + + /* This controls whether we will except HW or not. */ + if ( pHAL->shared.bForceSW == TRUE ) + { + return (pChoice == lpD3DHELDesc) ? D3DENUMRET_CANCEL : D3DENUMRET_OK; + } + + /* Always try for hardware. */ + if ( pChoice == lpD3DHWDesc ) + { + return D3DENUMRET_CANCEL; + } + + return D3DENUMRET_OK; +} +/*===========================================================================*/ +/* This function will destroy any and all surfaces that this context has */ +/* allocated. If there is a clipper object then it will also be destoryed as*/ +/* it is part of the Primary Surface. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void DestroyAllSurfaces( PMESAD3DHAL pHAL ) +{ + LONG refCount; + + DPF(( DBG_FUNC, "DestroyAllSurfaces();" )); + + DX_RESTORE( pHAL->lpDDSPrimary ); + DX_RESTORE( pHAL->lpDDSRender ); + DX_RESTORE( pHAL->lpDDSZbuffer); + + if ( pHAL->lpDDSRender ) + { + pHAL->lpDDSRender->Unlock( NULL ); + + /* If this isn't a Flipable surface then we must clean up the render. */ + if ( pHAL->shared.bFlipable == FALSE) + { + if ( pHAL->lpDDSZbuffer ) + { + DPF(( DBG_CNTX_INFO, "Remove attached surfaces from RENDER" )); + pHAL->lpDDSRender->DeleteAttachedSurface( 0, NULL ); + } + + DPF(( DBG_CNTX_INFO, "Release RENDER" )); + refCount = pHAL->lpDDSRender->Release(); + pHAL->lpDDSRender = NULL; + } + } + + if ( pHAL->lpDDSZbuffer ) + { + DPF(( DBG_CNTX_INFO, "Release ZBuffer" )); + pHAL->lpDDSZbuffer->Unlock( NULL ); + refCount = pHAL->lpDDSZbuffer->Release(); + pHAL->lpDDSZbuffer = NULL; + } + + if ( pHAL->lpClipper ) + { + DPF(( DBG_CNTX_INFO, "Release Clipper" )); + refCount = pHAL->lpClipper->Release(); + pHAL->lpClipper = NULL; + } + + if ( pHAL->lpDDSPrimary ) + { + pHAL->lpDDSPrimary->Unlock( NULL ); + + DPF(( DBG_CNTX_INFO, "Release PRIMARY" )); + refCount = pHAL->lpDDSPrimary->Release(); + pHAL->lpDDSPrimary = NULL; + } +} +/*===========================================================================*/ +/* This function will destroy the current D3DDevice and any resources that */ +/* belong to it. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void DestroyDevice( PMESAD3DHAL pHAL ) +{ + LONG refCount; + + DPF(( DBG_FUNC, "DestroyDevice();" )); + + /* Kill the D3D stuff if exists. */ + if ( pHAL->lpViewport ) + { + DPF(( DBG_CNTX_INFO, "Delete Viewport" )); + pHAL->lpD3DDevice->DeleteViewport( pHAL->lpViewport ); + + DPF(( DBG_CNTX_INFO, "Release Viewport" )); + refCount = pHAL->lpViewport->Release(); + pHAL->lpViewport = NULL; + } + + if ( pHAL->lpD3DDevice != NULL ) + { + DPF(( DBG_CNTX_INFO, "Release D3DDevice" )); + refCount = pHAL->lpD3DDevice->EndScene(); + refCount = pHAL->lpD3DDevice->Release(); + pHAL->lpD3DDevice = NULL; + } +} +/*===========================================================================*/ +/* This function will destroy the current D3DDevice and any resources that */ +/* belong to it. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void DestroyInterfaces( PMESAD3DHAL pHAL ) +{ + LONG refCount; + + DPF(( DBG_FUNC, "DestroyInterfaces();" )); + + if ( pHAL->lpD3D3 != NULL ) + { + DPF(( DBG_CNTX_INFO, "Release Direct3D3" )); + refCount = pHAL->lpD3D3->Release(); + pHAL->lpD3D3 = NULL; + } + + if ( pHAL->lpDD4 != NULL ) + { + DPF(( DBG_CNTX_INFO, "Release DDraw4" )); + refCount = pHAL->lpDD4->Release(); + pHAL->lpDD4 = NULL; + } + + if ( pHAL->lpDD != NULL ) + { + DPF(( DBG_CNTX_INFO, "Release DDraw" )); + refCount = pHAL->lpDD->Release(); + pHAL->lpDD = NULL; + } +} +/*===========================================================================*/ +/* This function will first send (not post) a message to the client window */ +/* that this context is using. The client will respond by unbinding itself */ +/* and binding the 'default' context. This allows the API to be supported */ +/* until the window can be destroyed. Finally we post the quit message to */ +/* the client in hopes to end the application. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void FatalShutDown( PMESAD3DHAL pHAL ) +{ + /* Whip this baby in too try and support the API until we die... */ + if ( pHAL ) + SendMessage( pHAL->shared.hwnd, UM_FATALSHUTDOWN, 0L, 0L ); + + /* Close the client application down. */ + PostQuitMessage( 0 ); +} + diff --git a/src/mesa/drivers/d3d/D3DMESA.H b/src/mesa/drivers/d3d/D3DMESA.H index 074ceca816..71a7cd8aed 100644 --- a/src/mesa/drivers/d3d/D3DMESA.H +++ b/src/mesa/drivers/d3d/D3DMESA.H @@ -1,84 +1,84 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#ifndef D3D_MESA_H -#define D3D_MESA_H -/*===========================================================================*/ -/* Includes. */ -/*===========================================================================*/ -#include -#include -#include -#include "matrix.h" -#include "context.h" +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#ifndef D3D_MESA_H +#define D3D_MESA_H +/*===========================================================================*/ +/* Includes. */ +/*===========================================================================*/ +#include +#include +#include +#include "matrix.h" +#include "context.h" #include "mtypes.h" -#include "vb.h" -#include "D3DShared.h" -#include "Debug.h" -#include "NULLProcs.h" -/*===========================================================================*/ -/* Macros. */ -/*===========================================================================*/ -#define FLIP(h,y) (h-y) -/*===========================================================================*/ -/* Magic numbers. */ -/*===========================================================================*/ -/*===========================================================================*/ -/* Type defines. */ -/*===========================================================================*/ -struct __extensions__ -{ - PROC proc; - char *name; -}; - -typedef GLbitfield (*ClearPROC)( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ); -typedef void (*WSpanRGBPROC)( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgb[][3], const GLubyte mask[] ); -typedef void (*WSpanRGBAPROC)( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgba[][4], const GLubyte mask[] ); -typedef void (*WSpanRGBAMonoPROC)( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte mask[] ); -typedef void (*WPixelsRGBAPROC)( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte rgba[][4], const GLubyte mask[] ); -typedef void (*WPixelsRGBAMonoPROC)( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte mask[] ); -typedef void (*RSpanRGBAPROC)( const GLcontext* ctx, GLuint n, GLint x, GLint y, GLubyte rgba[][4] ); -typedef void (*RPixelsRGBAPROC)( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], GLubyte rgba[][4], const GLubyte mask[] ); - -typedef struct D3D_mesa_context -{ - PMESAD3DSHARED pShared; - - GLcontext *gl_ctx; /* The core GL/Mesa context */ - GLvisual *gl_visual; /* Describes the buffers */ - GLframebuffer *gl_buffer; /* Depth, stencil, accum, etc buffers */ - - HDC hdc; - WNDPROC hOldProc; - - UCHAR rClear, /* Current clear colors. */ - gClear, - bClear, - aClear, - rCurrent, /* Current rendering colors. */ - gCurrent, - bCurrent, - aCurrent; - - struct D3D_mesa_context *next; - -} D3DMESACONTEXT, *PD3DMESACONTEXT; -/*===========================================================================*/ -/* Extern function prototypes. */ -/*===========================================================================*/ -extern void gl_Viewport( GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height ); -/*===========================================================================*/ -/* Global variables. */ -/*===========================================================================*/ -extern D3DTLVERTEX D3DTLVertices[(VB_MAX*6)]; - -#endif - +#include "vb.h" +#include "D3DShared.h" +#include "Debug.h" +#include "NULLProcs.h" +/*===========================================================================*/ +/* Macros. */ +/*===========================================================================*/ +#define FLIP(h,y) (h-y) +/*===========================================================================*/ +/* Magic numbers. */ +/*===========================================================================*/ +/*===========================================================================*/ +/* Type defines. */ +/*===========================================================================*/ +struct __extensions__ +{ + PROC proc; + char *name; +}; + +typedef GLbitfield (*ClearPROC)( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ); +typedef void (*WSpanRGBPROC)( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgb[][3], const GLubyte mask[] ); +typedef void (*WSpanRGBAPROC)( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgba[][4], const GLubyte mask[] ); +typedef void (*WSpanRGBAMonoPROC)( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte mask[] ); +typedef void (*WPixelsRGBAPROC)( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte rgba[][4], const GLubyte mask[] ); +typedef void (*WPixelsRGBAMonoPROC)( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte mask[] ); +typedef void (*RSpanRGBAPROC)( const GLcontext* ctx, GLuint n, GLint x, GLint y, GLubyte rgba[][4] ); +typedef void (*RPixelsRGBAPROC)( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], GLubyte rgba[][4], const GLubyte mask[] ); + +typedef struct D3D_mesa_context +{ + PMESAD3DSHARED pShared; + + GLcontext *gl_ctx; /* The core GL/Mesa context */ + GLvisual *gl_visual; /* Describes the buffers */ + GLframebuffer *gl_buffer; /* Depth, stencil, accum, etc buffers */ + + HDC hdc; + WNDPROC hOldProc; + + UCHAR rClear, /* Current clear colors. */ + gClear, + bClear, + aClear, + rCurrent, /* Current rendering colors. */ + gCurrent, + bCurrent, + aCurrent; + + struct D3D_mesa_context *next; + +} D3DMESACONTEXT, *PD3DMESACONTEXT; +/*===========================================================================*/ +/* Extern function prototypes. */ +/*===========================================================================*/ +extern void gl_Viewport( GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height ); +/*===========================================================================*/ +/* Global variables. */ +/*===========================================================================*/ +extern D3DTLVERTEX D3DTLVertices[(VB_MAX*6)]; + +#endif + diff --git a/src/mesa/drivers/d3d/D3DRaster.cpp b/src/mesa/drivers/d3d/D3DRaster.cpp index 004bb77360..41e42ad94c 100644 --- a/src/mesa/drivers/d3d/D3DRaster.cpp +++ b/src/mesa/drivers/d3d/D3DRaster.cpp @@ -1,213 +1,213 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#include "D3DHAL.h" -/*===========================================================================*/ -/* This function clears the context bound to the supplied shared context. */ -/* The function takes the D3D flags D3DCLEAR_TARGET, D3DCLEAR_STENCIL and */ -/* D3DCLEAR_ZBUFFER. Set bAll to TRUE for a full clear else supply the coord*/ -/* of the rect to be cleared relative to the window. The color is always a */ -/* 32bit value (RGBA). Fill in the z-value and stencil if needed. */ -/* */ -/* TODO: this can be redone to be called by Mesa directly. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void ClearHAL( PMESAD3DSHARED pShared, DWORD dwFlags, BOOL bAll, int x, int y, int cx, int cy, DWORD dwColor, float zv, DWORD dwStencil ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - D3DRECT d3dRect; - -#ifdef D3D_DEBUG - HRESULT rc; - - DPF(( DBG_FUNC, "CleaHAL();" )); - - /* Make sure we have enough info. */ - if ( (pHAL == NULL) || (pHAL->lpViewport == NULL) ) - return; -#endif - - if ( bAll ) - { - /* I assume my viewport is valid. */ - d3dRect.lX1 = pShared->rectV.left; - d3dRect.lY1 = pShared->rectV.top; - d3dRect.lX2 = pShared->rectV.right; - d3dRect.lY2 = pShared->rectV.bottom; - } - else - { - d3dRect.lX1 = pShared->rectV.left + x; - d3dRect.lY1 = pShared->rectV.top + y; - d3dRect.lX2 = d3dRect.lX1 + cx; - d3dRect.lY2 = d3dRect.lY1 + cy; - } - -#ifdef D3D_DEBUG - rc = pHAL->lpViewport->Clear2( 1, &d3dRect, dwFlags, dwColor, zv, dwStencil ); - if ( FAILED(rc) ) - { - RIP( pHAL, "Clear2 ->", ErrorStringD3D(rc) ); - } -#else - pHAL->lpViewport->Clear2( 1, &d3dRect, dwFlags, dwColor, zv, dwStencil ); -#endif -} -/*===========================================================================*/ -/* Well this is the guts of it all. Here we rasterize the primitives that */ -/* are in their final form. OpenGL has done all the lighting, transfomations*/ -/* and clipping at this point. */ -/* */ -/* TODO: I'm not sure if I want to bother to check for errors on this call. */ -/* The overhead kills me... */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void DrawPrimitiveHAL( PMESAD3DSHARED pShared, D3DPRIMITIVETYPE dptPrimitiveType, D3DTLVERTEX *pVertices, DWORD dwCount ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - -#ifdef D3D_DEBUG - HRESULT rc; - - DPF(( DBG_FUNC, "DrawPrimitveHAL();" )); - - /* Make sure we have enough info. */ - if ( (pHAL == NULL) || (pHAL->lpD3DDevice == NULL) ) - return; - - DPF(( DBG_PRIM_INFO, "DP( %d )", dwCount )); - - rc = pHAL->lpD3DDevice->DrawPrimitive( dptPrimitiveType, - D3DFVF_TLVERTEX, - (LPVOID)pVertices, - dwCount, - (D3DDP_DONOTCLIP | D3DDP_DONOTLIGHT) ); - if ( FAILED(rc) ) - { - RIP( pHAL, "DrawPrimitive ->", ErrorStringD3D(rc) ); - } -#else - pHAL->lpD3DDevice->DrawPrimitive( dptPrimitiveType, - D3DFVF_TLVERTEX, - (LPVOID)pVertices, - dwCount, - (D3DDP_DONOTCLIP | D3DDP_DONOTLIGHT) ); -#endif -} -/*===========================================================================*/ -/* This call will handle the swapping of the buffers. Now I didn't bother */ -/* to support single buffered so this will be used for glFlush() as its all */ -/* the same. So first we do an EndScene as we are always considered to be in*/ -/* a BeginScene because when we leave we do a BeginScene. Now note that when*/ -/* the context is created in the first place we do a BeginScene also just to */ -/* get things going. The call will use either Flip/blt based on the type of */ -/* surface was created for rendering. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void SwapBuffersHAL( PMESAD3DSHARED pShared ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - -#ifdef D3D_DEBUG - HRESULT rc; - - DPF(( DBG_FUNC, "SwapBuffersHAL();" )); - DPF(( DBG_ALL_PROFILE, "=================SWAP===================" )); - - /* Make sure we have enough info. */ - if ( (pHAL == NULL) || (pHAL->lpD3DDevice == NULL) ) - return; - - /* Make sure we have enough info. */ - if ( pHAL->lpDDSPrimary != NULL ) - { - rc = pHAL->lpD3DDevice->EndScene(); - if ( FAILED(rc) ) - { - RIP( pHAL, "EndScene ->", ErrorStringD3D(rc) ); - } - - if ( pShared->bFlipable ) - { - DPF(( DBG_CNTX_PROFILE, "Swap->FLIP" )); - rc = pHAL->lpDDSPrimary->Flip( NULL, DDFLIP_WAIT ); - } - else - { - DPF(( DBG_CNTX_PROFILE, "Swap->Blt" )); - rc = pHAL->lpDDSPrimary->Blt( &pShared->rectW, pHAL->lpDDSRender, NULL, DDBLT_WAIT, NULL ); - } - if ( FAILED(rc) ) - { - RIP( pHAL, "Blt (RENDER/PRIMARY) ->", ErrorStringD3D(rc) ); - } - - rc = pHAL->lpD3DDevice->BeginScene(); - if ( FAILED(rc) ) - { - RIP( pHAL, "BeginScene ->", ErrorStringD3D(rc) ); - } - } -#else - pHAL->lpD3DDevice->EndScene(); - - if ( pShared->bFlipable ) - pHAL->lpDDSPrimary->Flip( NULL, DDFLIP_WAIT ); - else - pHAL->lpDDSPrimary->Blt( &pShared->rectW, pHAL->lpDDSRender, NULL, DDBLT_WAIT, NULL ); - - pHAL->lpD3DDevice->BeginScene(); - -#endif -} -/*===========================================================================*/ -/* This function is a very thin wrapper for the D3D call 'SetRenderState'. */ -/* Using this function requires all the types to be defined by including the */ -/* D3D header file. */ -/* */ -/* TODO: would be much better to get ride of all these calls per VBRender. */ -/* I feel I should get this call into SetRenderStates() the RenderVB. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void SetStateHAL( PMESAD3DSHARED pShared, DWORD dwType, DWORD dwState ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - -#ifdef D3D_DEBUG - HRESULT rc; - - DPF(( DBG_FUNC, "SetStateHAL();" )); - - /* Make sure we have enough info. */ - if ( (pHAL == NULL) || (pHAL->lpD3DDevice == NULL) ) - return; - - rc = pHAL->lpD3DDevice->SetRenderState( (D3DRENDERSTATETYPE)dwType, dwState ); - if ( FAILED(rc) ) - { - RIP( pHAL, "SetRenderState ->", ErrorStringD3D(rc) ); - } - -#else - pHAL->lpD3DDevice->SetRenderState( (D3DRENDERSTATETYPE)dwType, dwState ); -#endif -} - - - - - - - - +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#include "D3DHAL.h" +/*===========================================================================*/ +/* This function clears the context bound to the supplied shared context. */ +/* The function takes the D3D flags D3DCLEAR_TARGET, D3DCLEAR_STENCIL and */ +/* D3DCLEAR_ZBUFFER. Set bAll to TRUE for a full clear else supply the coord*/ +/* of the rect to be cleared relative to the window. The color is always a */ +/* 32bit value (RGBA). Fill in the z-value and stencil if needed. */ +/* */ +/* TODO: this can be redone to be called by Mesa directly. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void ClearHAL( PMESAD3DSHARED pShared, DWORD dwFlags, BOOL bAll, int x, int y, int cx, int cy, DWORD dwColor, float zv, DWORD dwStencil ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + D3DRECT d3dRect; + +#ifdef D3D_DEBUG + HRESULT rc; + + DPF(( DBG_FUNC, "CleaHAL();" )); + + /* Make sure we have enough info. */ + if ( (pHAL == NULL) || (pHAL->lpViewport == NULL) ) + return; +#endif + + if ( bAll ) + { + /* I assume my viewport is valid. */ + d3dRect.lX1 = pShared->rectV.left; + d3dRect.lY1 = pShared->rectV.top; + d3dRect.lX2 = pShared->rectV.right; + d3dRect.lY2 = pShared->rectV.bottom; + } + else + { + d3dRect.lX1 = pShared->rectV.left + x; + d3dRect.lY1 = pShared->rectV.top + y; + d3dRect.lX2 = d3dRect.lX1 + cx; + d3dRect.lY2 = d3dRect.lY1 + cy; + } + +#ifdef D3D_DEBUG + rc = pHAL->lpViewport->Clear2( 1, &d3dRect, dwFlags, dwColor, zv, dwStencil ); + if ( FAILED(rc) ) + { + RIP( pHAL, "Clear2 ->", ErrorStringD3D(rc) ); + } +#else + pHAL->lpViewport->Clear2( 1, &d3dRect, dwFlags, dwColor, zv, dwStencil ); +#endif +} +/*===========================================================================*/ +/* Well this is the guts of it all. Here we rasterize the primitives that */ +/* are in their final form. OpenGL has done all the lighting, transfomations*/ +/* and clipping at this point. */ +/* */ +/* TODO: I'm not sure if I want to bother to check for errors on this call. */ +/* The overhead kills me... */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void DrawPrimitiveHAL( PMESAD3DSHARED pShared, D3DPRIMITIVETYPE dptPrimitiveType, D3DTLVERTEX *pVertices, DWORD dwCount ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + +#ifdef D3D_DEBUG + HRESULT rc; + + DPF(( DBG_FUNC, "DrawPrimitveHAL();" )); + + /* Make sure we have enough info. */ + if ( (pHAL == NULL) || (pHAL->lpD3DDevice == NULL) ) + return; + + DPF(( DBG_PRIM_INFO, "DP( %d )", dwCount )); + + rc = pHAL->lpD3DDevice->DrawPrimitive( dptPrimitiveType, + D3DFVF_TLVERTEX, + (LPVOID)pVertices, + dwCount, + (D3DDP_DONOTCLIP | D3DDP_DONOTLIGHT) ); + if ( FAILED(rc) ) + { + RIP( pHAL, "DrawPrimitive ->", ErrorStringD3D(rc) ); + } +#else + pHAL->lpD3DDevice->DrawPrimitive( dptPrimitiveType, + D3DFVF_TLVERTEX, + (LPVOID)pVertices, + dwCount, + (D3DDP_DONOTCLIP | D3DDP_DONOTLIGHT) ); +#endif +} +/*===========================================================================*/ +/* This call will handle the swapping of the buffers. Now I didn't bother */ +/* to support single buffered so this will be used for glFlush() as its all */ +/* the same. So first we do an EndScene as we are always considered to be in*/ +/* a BeginScene because when we leave we do a BeginScene. Now note that when*/ +/* the context is created in the first place we do a BeginScene also just to */ +/* get things going. The call will use either Flip/blt based on the type of */ +/* surface was created for rendering. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void SwapBuffersHAL( PMESAD3DSHARED pShared ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + +#ifdef D3D_DEBUG + HRESULT rc; + + DPF(( DBG_FUNC, "SwapBuffersHAL();" )); + DPF(( DBG_ALL_PROFILE, "=================SWAP===================" )); + + /* Make sure we have enough info. */ + if ( (pHAL == NULL) || (pHAL->lpD3DDevice == NULL) ) + return; + + /* Make sure we have enough info. */ + if ( pHAL->lpDDSPrimary != NULL ) + { + rc = pHAL->lpD3DDevice->EndScene(); + if ( FAILED(rc) ) + { + RIP( pHAL, "EndScene ->", ErrorStringD3D(rc) ); + } + + if ( pShared->bFlipable ) + { + DPF(( DBG_CNTX_PROFILE, "Swap->FLIP" )); + rc = pHAL->lpDDSPrimary->Flip( NULL, DDFLIP_WAIT ); + } + else + { + DPF(( DBG_CNTX_PROFILE, "Swap->Blt" )); + rc = pHAL->lpDDSPrimary->Blt( &pShared->rectW, pHAL->lpDDSRender, NULL, DDBLT_WAIT, NULL ); + } + if ( FAILED(rc) ) + { + RIP( pHAL, "Blt (RENDER/PRIMARY) ->", ErrorStringD3D(rc) ); + } + + rc = pHAL->lpD3DDevice->BeginScene(); + if ( FAILED(rc) ) + { + RIP( pHAL, "BeginScene ->", ErrorStringD3D(rc) ); + } + } +#else + pHAL->lpD3DDevice->EndScene(); + + if ( pShared->bFlipable ) + pHAL->lpDDSPrimary->Flip( NULL, DDFLIP_WAIT ); + else + pHAL->lpDDSPrimary->Blt( &pShared->rectW, pHAL->lpDDSRender, NULL, DDBLT_WAIT, NULL ); + + pHAL->lpD3DDevice->BeginScene(); + +#endif +} +/*===========================================================================*/ +/* This function is a very thin wrapper for the D3D call 'SetRenderState'. */ +/* Using this function requires all the types to be defined by including the */ +/* D3D header file. */ +/* */ +/* TODO: would be much better to get ride of all these calls per VBRender. */ +/* I feel I should get this call into SetRenderStates() the RenderVB. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void SetStateHAL( PMESAD3DSHARED pShared, DWORD dwType, DWORD dwState ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + +#ifdef D3D_DEBUG + HRESULT rc; + + DPF(( DBG_FUNC, "SetStateHAL();" )); + + /* Make sure we have enough info. */ + if ( (pHAL == NULL) || (pHAL->lpD3DDevice == NULL) ) + return; + + rc = pHAL->lpD3DDevice->SetRenderState( (D3DRENDERSTATETYPE)dwType, dwState ); + if ( FAILED(rc) ) + { + RIP( pHAL, "SetRenderState ->", ErrorStringD3D(rc) ); + } + +#else + pHAL->lpD3DDevice->SetRenderState( (D3DRENDERSTATETYPE)dwType, dwState ); +#endif +} + + + + + + + + diff --git a/src/mesa/drivers/d3d/D3DTEXT.CPP b/src/mesa/drivers/d3d/D3DTEXT.CPP index e6ff864561..39855b65fa 100644 --- a/src/mesa/drivers/d3d/D3DTEXT.CPP +++ b/src/mesa/drivers/d3d/D3DTEXT.CPP @@ -1,576 +1,576 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#include "d3dText.h" - -/*============================================================================= - - 1 - ------ - | | - 6 | | 2 - | 7 | - ------ - | | - 5 | | 3 - | | - ------ - 4 - - TL_0 TR_0 -TLL TL_1 TR_1 TRR - -MLL_0 ML_0 MR_0 MRR_0 -MLL_1 ML_1 MR_1 MRR_1 - -BLL BL_0 BR_0 BRR - BL_1 BR_1 - -=============================================================================*/ - -#define TLL 0 -#define TRR 1 -#define TL_0 2 -#define TL_1 3 -#define TR_0 4 -#define TR_1 5 - -#define MLL_0 6 -#define MLL_1 7 -#define MRR_0 8 -#define MRR_1 9 - -#define ML_0 10 -#define ML_1 11 -#define MR_0 12 -#define MR_1 13 - -#define BL_0 14 -#define BL_1 15 -#define BR_0 16 -#define BR_1 17 -#define BLL 18 -#define BRR 19 - -#define BIT1 0x00000001 -#define BIT2 0x00000002 -#define BIT3 0x00000004 -#define BIT4 0x00000008 -#define BIT5 0x00000010 -#define BIT6 0x00000020 -#define BIT7 0x00000040 - -#define TOP BIT4 -#define MIDDLE BIT7 -#define BOTTOM BIT1 -#define TLEFT BIT5 -#define BLEFT BIT6 -#define LEFT (TLEFT|BLEFT) -#define TRIGHT BIT3 -#define BRIGHT BIT2 -#define RIGHT (TRIGHT|BRIGHT) -#define ALL 0xFFFFFFFF - -/*===========================================================================*/ -/* This is the static array that will map the ASCII value of the character */ -/* being draw to the bit mask that will be scan converted to the LED display.*/ -/*===========================================================================*/ -DWORD textBitMasks[] = -{ - 0xFFFFFFFF, // 000 - 0xFFFFFFFF, // 001 - 0xFFFFFFFF, // 002 - 0xFFFFFFFF, // 003 - 0xFFFFFFFF, // 004 - 0xFFFFFFFF, // 005 - 0xFFFFFFFF, // 006 - 0xFFFFFFFF, // 007 - 0xFFFFFFFF, // 008 - 0xFFFFFFFF, // 009 - 0xFFFFFFFF, // 010 - 0xFFFFFFFF, // 011 - 0xFFFFFFFF, // 012 - 0xFFFFFFFF, // 013 - 0xFFFFFFFF, // 014 - 0xFFFFFFFF, // 015 - 0xFFFFFFFF, // 016 - 0xFFFFFFFF, // 017 - 0xFFFFFFFF, // 018 - 0xFFFFFFFF, // 019 - 0xFFFFFFFF, // 020 - 0xFFFFFFFF, // 021 - 0xFFFFFFFF, // 022 - 0xFFFFFFFF, // 023 - 0xFFFFFFFF, // 024 - 0xFFFFFFFF, // 025 - 0xFFFFFFFF, // 026 - 0xFFFFFFFF, // 027 - 0xFFFFFFFF, // 028 - 0xFFFFFFFF, // 029 - 0xFFFFFFFF, // 030 - 0XFFFFFFFF, // 031 - 0x00000000, // 032 'SPC' - 0xFFFFFFFF, // 033 - 0xFFFFFFFF, // 034 - 0xFFFFFFFF, // 035 - 0xFFFFFFFF, // 036 - 0xFFFFFFFF, // 037 - 0xFFFFFFFF, // 038 - 0xFFFFFFFF, // 039 - 0xFFFFFFFF, // 040 - 0xFFFFFFFF, // 041 - 0xFFFFFFFF, // 042 - 0xFFFFFFFF, // 043 - 0xFFFFFFFF, // 044 - 0xFFFFFFFF, // 045 - 0xFFFFFFFF, // 046 - 0xFFFFFFFF, // 047 - (ALL &~ MIDDLE), // 048 '0' - (RIGHT), // 049 '1' - (ALL &~ TLEFT &~ BRIGHT), // 050 '2' - (ALL &~ LEFT), // 051 '3' - (TLEFT | MIDDLE | RIGHT), // 052 '4' - (ALL &~ TRIGHT &~ BLEFT), // 053 '5' - (ALL &~ TRIGHT), // 054 '6' - (TOP | RIGHT), // 055 '7' - (ALL), // 056 '8' - (ALL &~ BOTTOM &~ BLEFT), // 057 '9' - 0xFFFFFFFF, // 058 - 0xFFFFFFFF, // 059 - 0xFFFFFFFF, // 060 - 0XFFFFFFFF, // 061 - 0xFFFFFFFF, // 062 - 0xFFFFFFFF, // 063 - 0xFFFFFFFF, // 064 - (ALL &~ BOTTOM), // 065 'A' - (ALL), // 066 'B' - (TOP | LEFT | BOTTOM), // 067 'C' - (ALL &~ MIDDLE), // 068 'D' - (ALL &~ RIGHT), // 069 'E' - (LEFT | TOP | MIDDLE), // 070 'F' - 0x00000000, // 071 'G' - (ALL &~ TOP &~ BOTTOM), // 072 'H' - (RIGHT), // 073 'I' - (RIGHT | BOTTOM), // 074 'J' - 0x00000000, // 075 'K' - (LEFT | BOTTOM), // 076 'L' - 0x00000000, // 088 'M' - 0x00000000, // 089 'N' - (ALL &~ MIDDLE), // 090 'O' - (ALL &~ BRIGHT &~ BOTTOM),// 091 'P' - 0x00000000, // 092 'Q' - 0x00000000, // 093 'R' - (ALL &~ TRIGHT &~ BLEFT), // 094 'S' - 0X00000000, // 095 'T' - (LEFT | RIGHT | BOTTOM), // 096 'U' - 0x00000000, // 097 'V' - 0x00000000, // 098 'W' - 0x00000000, // 099 'X' - 0x00000000, // 1000 'Z' - 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 100 - 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 104 - 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 108 - 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 112 - 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 116 - 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 120 - 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF // 124 -}; - -#define CT 1.0f -#define CX 7.0f -#define CY 13.0f -#define CM ((CY-(CT*3.0f))/2.0f) - -float lCoords[][2] = -{ - /* Top outsides. */ - { 0, (CY-CT) }, - { CX, (CY-CT) }, - - /* Top Line. */ - { CT, CY }, - { CT, (CY-CT) }, - { (CX-CT), CY }, - { (CX-CT), (CY-CT) }, - - /* Middle outsides. */ - { 0.0f, (CT+CM+CT) }, - { 0.0f, (CT+CM) }, - { CX, (CT+CM+CT) }, - { CX, (CT+CM) }, - - /* Middle Line. */ - { CT, (CT+CM+CT) }, - { CT, (CT+CM) }, - { (CX-CT), (CT+CM+CT) }, - { (CX-CT), (CT+CM) }, - - /* Bottom line. */ - { CT, CT }, - { CT, 0.0f }, - { (CX-CT), CT }, - { (CX-CT), 0.0f }, - - /* Bottom outsides. */ - { 0.0f, CT}, - { CX, CT } -}; - -static int ConvertCharacter( char *c, int cIndex, PD3DFONTMETRICS pfntMetrics ); - -D3DTLVERTEX TextVertices[MAX_VERTICES]; -/*===========================================================================*/ -/* When we attach I will zero out the whole D3D vertex buffer I'm using for */ -/* the text. This way I don't need to set all the redundant values. I also */ -/* set all the oow values to 1 as I will be doing direct rendering. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -extern "C" BOOL InitD3DText( void ) -{ - int index; - - /* Set the D3D Vertex Buffer up once so we don't do redundant changes. */ - memset( &TextVertices[0], 0, sizeof(TextVertices) ); - for( index = 0; index < MAX_VERTICES; index++ ) - TextVertices[index].rhw = D3DVAL( 1.0 ); - - return TRUE; -} -/*===========================================================================*/ -/* This function takes a single character and draw it using the supplied */ -/* fontmetrics structure. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void d3dTextDrawString( char *pszString, int x, int y, PD3DFONTMETRICS pfntMetrics ) -{ - int cIndex, - nIndex, - index; - float cWidth = CX, - cHeight = CY; - - /* Find the max width/height of a character and add the spacing so */ - /* that we can use this value to calculate the x,y of the character.*/ - cWidth = (cWidth * pfntMetrics->fntXScale) + pfntMetrics->fntXSpacing; - cHeight = (cHeight * pfntMetrics->fntYScale) + pfntMetrics->fntYSpacing; - - /* Walk the string. This must be NULL terminated. */ - for( cIndex = 0, nIndex = 0; *pszString; pszString++, cIndex = nIndex, x++ ) - { - /* Convert the character and get the index into the text vertex buffer. */ - nIndex = ConvertCharacter( &pszString[0], cIndex, pfntMetrics ); - if ( (nIndex - cIndex) > 2 ) - { - /* Modify the text vertex buffer based on the fntMetrics structure. */ - for( index = cIndex; index < nIndex; index++ ) - { - /* Scale the character. */ - TextVertices[index].sx *= pfntMetrics->fntXScale; - TextVertices[index].sy *= pfntMetrics->fntYScale; - - /* Move the character. */ - TextVertices[index].sx += (cWidth*x); - TextVertices[index].sy += (cHeight*y); - - /* Set the color. */ - TextVertices[index].color = pfntMetrics->dwColor; - } - } - } - - if ( nIndex < 3 ) - return; - - /* Set the states that slim things down. */ - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_CULLMODE, D3DCULL_NONE ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_FILLMODE, D3DFILL_SOLID ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ZENABLE, FALSE ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ZWRITEENABLE , FALSE ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ALPHATESTENABLE, FALSE ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ALPHABLENDENABLE, FALSE ); - - /* Blast them baby... */ - pfntMetrics->lpD3DDevice->DrawPrimitive( D3DPT_TRIANGLELIST, - D3DFVF_TLVERTEX, - (LPVOID)&TextVertices[0], - nIndex, - (D3DDP_DONOTCLIP | D3DDP_DONOTLIGHT) ); -} -/*===========================================================================*/ -/* This function takes a single character and draw it directly to the screen*/ -/* unsing the supplied fntMetrics structure. The character will be drawn at */ -/* the supplied x,y. The x,y position is relative to the top left and uses */ -/* the spacing in the fntMetrics structure. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void d3dTextDrawCharacter( char *c, int x, int y, PD3DFONTMETRICS pfntMetrics ) -{ - int cIndex = 0, - index; - float cWidth = CX, - cHeight = CY; - - /* Convert the character and get the index into the text vertex buffer. */ - cIndex = ConvertCharacter( c, 0, pfntMetrics ); - if ( cIndex < 3 ) - return; - - /* Find the max width/height of a character and add the spacing so */ - /* that we can use this value to calculate the x,y of the character.*/ - cWidth = (cWidth * pfntMetrics->fntXScale) + pfntMetrics->fntXSpacing; - cHeight = (cHeight * pfntMetrics->fntYScale) + pfntMetrics->fntYSpacing; - - /* Modify the text vertex buffer based on the fntMetrics structure. */ - for( index = 0; index < cIndex; index++ ) - { - /* Scale the character. */ - TextVertices[index].sx *= pfntMetrics->fntXScale; - TextVertices[index].sy *= pfntMetrics->fntYScale; - - /* Move the character. */ - TextVertices[index].sx += (cWidth*x); - TextVertices[index].sy += (cHeight*y); - - /* Set the color. */ - TextVertices[index].color = pfntMetrics->dwColor; - } - - - /* Set the states that slim things down. */ - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_CULLMODE, D3DCULL_NONE ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_FILLMODE, D3DFILL_SOLID ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ZENABLE, FALSE ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ZWRITEENABLE , FALSE ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ALPHATESTENABLE, FALSE ); - pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ALPHABLENDENABLE, FALSE ); - - /* Blast them baby... */ - pfntMetrics->lpD3DDevice->DrawPrimitive( D3DPT_TRIANGLELIST, - D3DFVF_TLVERTEX, - (LPVOID)&TextVertices[0], - cIndex, - (D3DDP_DONOTCLIP | D3DDP_DONOTLIGHT) ); -} -/*===========================================================================*/ -/* This function takes a single character and draw it using the supplied */ -/* fontmetrics structure. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static int ConvertCharacter( char *c, int cIndex, PD3DFONTMETRICS pfntMetrics ) -{ - DWORD asciiChar = (int)(*c); - - /* Handle the TOP line. */ - if ( textBitMasks[asciiChar] & BIT1 ) - { - TextVertices[cIndex].sx = D3DVAL( lCoords[TL_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TL_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TR_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TL_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TL_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TL_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TL_0][1] ); - } - - /* Handle the TOP/BOTTOM RIGHT lines. */ - // if ( textBitMasks[index] & (BIT2|BIT3) ) - if ( 1 == 0 ) - { - TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TRR][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TRR][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BRR][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BRR][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BRR][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BRR][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BR_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BR_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_1][1] ); - } - else - { - if ( textBitMasks[asciiChar] & BIT2 ) - { - TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[TR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TRR][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[TRR][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MRR_0][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[MRR_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MRR_0][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[MRR_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MR_0][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[MR_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[TR_1][1] ); - } - if ( textBitMasks[asciiChar] & BIT3 ) - { - TextVertices[cIndex].sx = D3DVAL( lCoords[MR_1][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[MR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MRR_1][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[MRR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BRR][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[BRR][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BRR][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[BRR][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BR_0][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[BR_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MR_1][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[MR_1][1] ); - } - } - - /* Handle the TOP/BOTTOM LEFT lines. */ - // if ( textBitMasks[asciiChar] & (BIT5|BIT6) ) - if ( 1 == 0 ) - { - TextVertices[cIndex].sx = D3DVAL( lCoords[TLL][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TLL][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TL_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TL_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BLL][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BLL][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TLL][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[TLL][1] ); - } - else - { - if ( textBitMasks[asciiChar] & BIT5 ) - { - TextVertices[cIndex].sx = D3DVAL( lCoords[MLL_1][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[MLL_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[ML_1][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[ML_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[BL_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[BL_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BLL][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[BLL][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MLL_1][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[MLL_1][1] ); - } - if ( textBitMasks[asciiChar] & BIT6 ) - { - TextVertices[cIndex].sx = D3DVAL( lCoords[TLL][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[TLL][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TL_1][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[TL_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[ML_0][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[ML_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[ML_0][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[ML_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MLL_0][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[MLL_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[TLL][0] ); - TextVertices[cIndex++].sy = D3DVAL( lCoords[TLL][1] ); - } - } - - /* Handle the MIDDLE line. */ - if ( textBitMasks[asciiChar] & BIT7 ) - { - TextVertices[cIndex].sx = D3DVAL( lCoords[ML_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[ML_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MR_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[MR_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MR_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[MR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[MR_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[MR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[ML_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[ML_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[ML_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[ML_0][1] ); - } - - /* Handle the BOTTOM line. */ - if ( textBitMasks[asciiChar] & BIT4 ) - { - TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BR_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BR_0][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BR_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BR_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BR_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BL_1][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_1][1] ); - TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); - TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_0][1] ); - } - - return cIndex; -} - -#undef CM -#undef CY -#undef CX -#undef CT - -#undef TLL -#undef TRR -#undef TL_0 -#undef TL_1 -#undef TR_0 -#undef TR_1 - -#undef MLL_0 -#undef MLL_1 -#undef MRR_0 -#undef MRR_1 - -#undef ML_0 -#undef ML_1 -#undef MR_0 -#undef MR_1 - -#undef BL_0 -#undef BL_1 -#undef BR_0 -#undef BR_1 -#undef BLL -#undef BRR - -#undef BIT1 -#undef BIT2 -#undef BIT3 -#undef BIT4 -#undef BIT5 -#undef BIT6 -#undef BIT7 - -#undef TOP -#undef MIDDLE -#undef BOTTOM -#undef TLEFT -#undef BLEFT -#undef LEFT -#undef TRIGHT -#undef BRIGHT -#undef RIGHT -#undef ALL +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#include "d3dText.h" + +/*============================================================================= + + 1 + ------ + | | + 6 | | 2 + | 7 | + ------ + | | + 5 | | 3 + | | + ------ + 4 + + TL_0 TR_0 +TLL TL_1 TR_1 TRR + +MLL_0 ML_0 MR_0 MRR_0 +MLL_1 ML_1 MR_1 MRR_1 + +BLL BL_0 BR_0 BRR + BL_1 BR_1 + +=============================================================================*/ + +#define TLL 0 +#define TRR 1 +#define TL_0 2 +#define TL_1 3 +#define TR_0 4 +#define TR_1 5 + +#define MLL_0 6 +#define MLL_1 7 +#define MRR_0 8 +#define MRR_1 9 + +#define ML_0 10 +#define ML_1 11 +#define MR_0 12 +#define MR_1 13 + +#define BL_0 14 +#define BL_1 15 +#define BR_0 16 +#define BR_1 17 +#define BLL 18 +#define BRR 19 + +#define BIT1 0x00000001 +#define BIT2 0x00000002 +#define BIT3 0x00000004 +#define BIT4 0x00000008 +#define BIT5 0x00000010 +#define BIT6 0x00000020 +#define BIT7 0x00000040 + +#define TOP BIT4 +#define MIDDLE BIT7 +#define BOTTOM BIT1 +#define TLEFT BIT5 +#define BLEFT BIT6 +#define LEFT (TLEFT|BLEFT) +#define TRIGHT BIT3 +#define BRIGHT BIT2 +#define RIGHT (TRIGHT|BRIGHT) +#define ALL 0xFFFFFFFF + +/*===========================================================================*/ +/* This is the static array that will map the ASCII value of the character */ +/* being draw to the bit mask that will be scan converted to the LED display.*/ +/*===========================================================================*/ +DWORD textBitMasks[] = +{ + 0xFFFFFFFF, // 000 + 0xFFFFFFFF, // 001 + 0xFFFFFFFF, // 002 + 0xFFFFFFFF, // 003 + 0xFFFFFFFF, // 004 + 0xFFFFFFFF, // 005 + 0xFFFFFFFF, // 006 + 0xFFFFFFFF, // 007 + 0xFFFFFFFF, // 008 + 0xFFFFFFFF, // 009 + 0xFFFFFFFF, // 010 + 0xFFFFFFFF, // 011 + 0xFFFFFFFF, // 012 + 0xFFFFFFFF, // 013 + 0xFFFFFFFF, // 014 + 0xFFFFFFFF, // 015 + 0xFFFFFFFF, // 016 + 0xFFFFFFFF, // 017 + 0xFFFFFFFF, // 018 + 0xFFFFFFFF, // 019 + 0xFFFFFFFF, // 020 + 0xFFFFFFFF, // 021 + 0xFFFFFFFF, // 022 + 0xFFFFFFFF, // 023 + 0xFFFFFFFF, // 024 + 0xFFFFFFFF, // 025 + 0xFFFFFFFF, // 026 + 0xFFFFFFFF, // 027 + 0xFFFFFFFF, // 028 + 0xFFFFFFFF, // 029 + 0xFFFFFFFF, // 030 + 0XFFFFFFFF, // 031 + 0x00000000, // 032 'SPC' + 0xFFFFFFFF, // 033 + 0xFFFFFFFF, // 034 + 0xFFFFFFFF, // 035 + 0xFFFFFFFF, // 036 + 0xFFFFFFFF, // 037 + 0xFFFFFFFF, // 038 + 0xFFFFFFFF, // 039 + 0xFFFFFFFF, // 040 + 0xFFFFFFFF, // 041 + 0xFFFFFFFF, // 042 + 0xFFFFFFFF, // 043 + 0xFFFFFFFF, // 044 + 0xFFFFFFFF, // 045 + 0xFFFFFFFF, // 046 + 0xFFFFFFFF, // 047 + (ALL &~ MIDDLE), // 048 '0' + (RIGHT), // 049 '1' + (ALL &~ TLEFT &~ BRIGHT), // 050 '2' + (ALL &~ LEFT), // 051 '3' + (TLEFT | MIDDLE | RIGHT), // 052 '4' + (ALL &~ TRIGHT &~ BLEFT), // 053 '5' + (ALL &~ TRIGHT), // 054 '6' + (TOP | RIGHT), // 055 '7' + (ALL), // 056 '8' + (ALL &~ BOTTOM &~ BLEFT), // 057 '9' + 0xFFFFFFFF, // 058 + 0xFFFFFFFF, // 059 + 0xFFFFFFFF, // 060 + 0XFFFFFFFF, // 061 + 0xFFFFFFFF, // 062 + 0xFFFFFFFF, // 063 + 0xFFFFFFFF, // 064 + (ALL &~ BOTTOM), // 065 'A' + (ALL), // 066 'B' + (TOP | LEFT | BOTTOM), // 067 'C' + (ALL &~ MIDDLE), // 068 'D' + (ALL &~ RIGHT), // 069 'E' + (LEFT | TOP | MIDDLE), // 070 'F' + 0x00000000, // 071 'G' + (ALL &~ TOP &~ BOTTOM), // 072 'H' + (RIGHT), // 073 'I' + (RIGHT | BOTTOM), // 074 'J' + 0x00000000, // 075 'K' + (LEFT | BOTTOM), // 076 'L' + 0x00000000, // 088 'M' + 0x00000000, // 089 'N' + (ALL &~ MIDDLE), // 090 'O' + (ALL &~ BRIGHT &~ BOTTOM),// 091 'P' + 0x00000000, // 092 'Q' + 0x00000000, // 093 'R' + (ALL &~ TRIGHT &~ BLEFT), // 094 'S' + 0X00000000, // 095 'T' + (LEFT | RIGHT | BOTTOM), // 096 'U' + 0x00000000, // 097 'V' + 0x00000000, // 098 'W' + 0x00000000, // 099 'X' + 0x00000000, // 1000 'Z' + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 100 + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 104 + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 108 + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 112 + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 116 + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, // 120 + 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF, 0xFFFFFFFF // 124 +}; + +#define CT 1.0f +#define CX 7.0f +#define CY 13.0f +#define CM ((CY-(CT*3.0f))/2.0f) + +float lCoords[][2] = +{ + /* Top outsides. */ + { 0, (CY-CT) }, + { CX, (CY-CT) }, + + /* Top Line. */ + { CT, CY }, + { CT, (CY-CT) }, + { (CX-CT), CY }, + { (CX-CT), (CY-CT) }, + + /* Middle outsides. */ + { 0.0f, (CT+CM+CT) }, + { 0.0f, (CT+CM) }, + { CX, (CT+CM+CT) }, + { CX, (CT+CM) }, + + /* Middle Line. */ + { CT, (CT+CM+CT) }, + { CT, (CT+CM) }, + { (CX-CT), (CT+CM+CT) }, + { (CX-CT), (CT+CM) }, + + /* Bottom line. */ + { CT, CT }, + { CT, 0.0f }, + { (CX-CT), CT }, + { (CX-CT), 0.0f }, + + /* Bottom outsides. */ + { 0.0f, CT}, + { CX, CT } +}; + +static int ConvertCharacter( char *c, int cIndex, PD3DFONTMETRICS pfntMetrics ); + +D3DTLVERTEX TextVertices[MAX_VERTICES]; +/*===========================================================================*/ +/* When we attach I will zero out the whole D3D vertex buffer I'm using for */ +/* the text. This way I don't need to set all the redundant values. I also */ +/* set all the oow values to 1 as I will be doing direct rendering. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +extern "C" BOOL InitD3DText( void ) +{ + int index; + + /* Set the D3D Vertex Buffer up once so we don't do redundant changes. */ + memset( &TextVertices[0], 0, sizeof(TextVertices) ); + for( index = 0; index < MAX_VERTICES; index++ ) + TextVertices[index].rhw = D3DVAL( 1.0 ); + + return TRUE; +} +/*===========================================================================*/ +/* This function takes a single character and draw it using the supplied */ +/* fontmetrics structure. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void d3dTextDrawString( char *pszString, int x, int y, PD3DFONTMETRICS pfntMetrics ) +{ + int cIndex, + nIndex, + index; + float cWidth = CX, + cHeight = CY; + + /* Find the max width/height of a character and add the spacing so */ + /* that we can use this value to calculate the x,y of the character.*/ + cWidth = (cWidth * pfntMetrics->fntXScale) + pfntMetrics->fntXSpacing; + cHeight = (cHeight * pfntMetrics->fntYScale) + pfntMetrics->fntYSpacing; + + /* Walk the string. This must be NULL terminated. */ + for( cIndex = 0, nIndex = 0; *pszString; pszString++, cIndex = nIndex, x++ ) + { + /* Convert the character and get the index into the text vertex buffer. */ + nIndex = ConvertCharacter( &pszString[0], cIndex, pfntMetrics ); + if ( (nIndex - cIndex) > 2 ) + { + /* Modify the text vertex buffer based on the fntMetrics structure. */ + for( index = cIndex; index < nIndex; index++ ) + { + /* Scale the character. */ + TextVertices[index].sx *= pfntMetrics->fntXScale; + TextVertices[index].sy *= pfntMetrics->fntYScale; + + /* Move the character. */ + TextVertices[index].sx += (cWidth*x); + TextVertices[index].sy += (cHeight*y); + + /* Set the color. */ + TextVertices[index].color = pfntMetrics->dwColor; + } + } + } + + if ( nIndex < 3 ) + return; + + /* Set the states that slim things down. */ + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_CULLMODE, D3DCULL_NONE ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_FILLMODE, D3DFILL_SOLID ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ZENABLE, FALSE ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ZWRITEENABLE , FALSE ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ALPHATESTENABLE, FALSE ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ALPHABLENDENABLE, FALSE ); + + /* Blast them baby... */ + pfntMetrics->lpD3DDevice->DrawPrimitive( D3DPT_TRIANGLELIST, + D3DFVF_TLVERTEX, + (LPVOID)&TextVertices[0], + nIndex, + (D3DDP_DONOTCLIP | D3DDP_DONOTLIGHT) ); +} +/*===========================================================================*/ +/* This function takes a single character and draw it directly to the screen*/ +/* unsing the supplied fntMetrics structure. The character will be drawn at */ +/* the supplied x,y. The x,y position is relative to the top left and uses */ +/* the spacing in the fntMetrics structure. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void d3dTextDrawCharacter( char *c, int x, int y, PD3DFONTMETRICS pfntMetrics ) +{ + int cIndex = 0, + index; + float cWidth = CX, + cHeight = CY; + + /* Convert the character and get the index into the text vertex buffer. */ + cIndex = ConvertCharacter( c, 0, pfntMetrics ); + if ( cIndex < 3 ) + return; + + /* Find the max width/height of a character and add the spacing so */ + /* that we can use this value to calculate the x,y of the character.*/ + cWidth = (cWidth * pfntMetrics->fntXScale) + pfntMetrics->fntXSpacing; + cHeight = (cHeight * pfntMetrics->fntYScale) + pfntMetrics->fntYSpacing; + + /* Modify the text vertex buffer based on the fntMetrics structure. */ + for( index = 0; index < cIndex; index++ ) + { + /* Scale the character. */ + TextVertices[index].sx *= pfntMetrics->fntXScale; + TextVertices[index].sy *= pfntMetrics->fntYScale; + + /* Move the character. */ + TextVertices[index].sx += (cWidth*x); + TextVertices[index].sy += (cHeight*y); + + /* Set the color. */ + TextVertices[index].color = pfntMetrics->dwColor; + } + + + /* Set the states that slim things down. */ + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_CULLMODE, D3DCULL_NONE ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_FILLMODE, D3DFILL_SOLID ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ZENABLE, FALSE ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ZWRITEENABLE , FALSE ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ALPHATESTENABLE, FALSE ); + pfntMetrics->lpD3DDevice->SetRenderState( D3DRENDERSTATE_ALPHABLENDENABLE, FALSE ); + + /* Blast them baby... */ + pfntMetrics->lpD3DDevice->DrawPrimitive( D3DPT_TRIANGLELIST, + D3DFVF_TLVERTEX, + (LPVOID)&TextVertices[0], + cIndex, + (D3DDP_DONOTCLIP | D3DDP_DONOTLIGHT) ); +} +/*===========================================================================*/ +/* This function takes a single character and draw it using the supplied */ +/* fontmetrics structure. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static int ConvertCharacter( char *c, int cIndex, PD3DFONTMETRICS pfntMetrics ) +{ + DWORD asciiChar = (int)(*c); + + /* Handle the TOP line. */ + if ( textBitMasks[asciiChar] & BIT1 ) + { + TextVertices[cIndex].sx = D3DVAL( lCoords[TL_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TL_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TR_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TL_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TL_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TL_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TL_0][1] ); + } + + /* Handle the TOP/BOTTOM RIGHT lines. */ + // if ( textBitMasks[index] & (BIT2|BIT3) ) + if ( 1 == 0 ) + { + TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TRR][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TRR][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BRR][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BRR][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BRR][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BRR][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BR_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BR_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TR_1][1] ); + } + else + { + if ( textBitMasks[asciiChar] & BIT2 ) + { + TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[TR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TRR][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[TRR][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MRR_0][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[MRR_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MRR_0][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[MRR_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MR_0][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[MR_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TR_1][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[TR_1][1] ); + } + if ( textBitMasks[asciiChar] & BIT3 ) + { + TextVertices[cIndex].sx = D3DVAL( lCoords[MR_1][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[MR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MRR_1][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[MRR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BRR][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[BRR][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BRR][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[BRR][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BR_0][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[BR_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MR_1][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[MR_1][1] ); + } + } + + /* Handle the TOP/BOTTOM LEFT lines. */ + // if ( textBitMasks[asciiChar] & (BIT5|BIT6) ) + if ( 1 == 0 ) + { + TextVertices[cIndex].sx = D3DVAL( lCoords[TLL][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TLL][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TL_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TL_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BLL][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BLL][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TLL][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[TLL][1] ); + } + else + { + if ( textBitMasks[asciiChar] & BIT5 ) + { + TextVertices[cIndex].sx = D3DVAL( lCoords[MLL_1][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[MLL_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[ML_1][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[ML_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[BL_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[BL_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BLL][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[BLL][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MLL_1][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[MLL_1][1] ); + } + if ( textBitMasks[asciiChar] & BIT6 ) + { + TextVertices[cIndex].sx = D3DVAL( lCoords[TLL][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[TLL][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TL_1][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[TL_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[ML_0][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[ML_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[ML_0][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[ML_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MLL_0][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[MLL_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[TLL][0] ); + TextVertices[cIndex++].sy = D3DVAL( lCoords[TLL][1] ); + } + } + + /* Handle the MIDDLE line. */ + if ( textBitMasks[asciiChar] & BIT7 ) + { + TextVertices[cIndex].sx = D3DVAL( lCoords[ML_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[ML_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MR_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[MR_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MR_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[MR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[MR_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[MR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[ML_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[ML_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[ML_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[ML_0][1] ); + } + + /* Handle the BOTTOM line. */ + if ( textBitMasks[asciiChar] & BIT4 ) + { + TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BR_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BR_0][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BR_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BR_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BR_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BL_1][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_1][1] ); + TextVertices[cIndex].sx = D3DVAL( lCoords[BL_0][0] ); + TextVertices[cIndex++].sy= D3DVAL( lCoords[BL_0][1] ); + } + + return cIndex; +} + +#undef CM +#undef CY +#undef CX +#undef CT + +#undef TLL +#undef TRR +#undef TL_0 +#undef TL_1 +#undef TR_0 +#undef TR_1 + +#undef MLL_0 +#undef MLL_1 +#undef MRR_0 +#undef MRR_1 + +#undef ML_0 +#undef ML_1 +#undef MR_0 +#undef MR_1 + +#undef BL_0 +#undef BL_1 +#undef BR_0 +#undef BR_1 +#undef BLL +#undef BRR + +#undef BIT1 +#undef BIT2 +#undef BIT3 +#undef BIT4 +#undef BIT5 +#undef BIT6 +#undef BIT7 + +#undef TOP +#undef MIDDLE +#undef BOTTOM +#undef TLEFT +#undef BLEFT +#undef LEFT +#undef TRIGHT +#undef BRIGHT +#undef RIGHT +#undef ALL diff --git a/src/mesa/drivers/d3d/D3DTextureMgr.cpp b/src/mesa/drivers/d3d/D3DTextureMgr.cpp index ac9d2621fc..a6600cab02 100644 --- a/src/mesa/drivers/d3d/D3DTextureMgr.cpp +++ b/src/mesa/drivers/d3d/D3DTextureMgr.cpp @@ -1,947 +1,947 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#include "D3DHAL.h" -/*===========================================================================*/ -/* Local function prototypes. */ -/*===========================================================================*/ -static void UpdateTexture( PTM_OBJECT pTMObj, BOOL bVideo, RECT *pRect, UCHAR *pixels ); -static BOOL LoadTextureInVideo( PMESAD3DHAL pHAL, PTM_OBJECT pTMObj ); -static BOOL FreeTextureMemory( PMESAD3DHAL pHAL, PTM_OBJECT pTMObject ); -static BOOL DestroyTextureObject( PMESAD3DHAL pHAL, PTM_OBJECT pTMObject ); -HRESULT CALLBACK EnumPFHook( LPDDPIXELFORMAT lpDDPixFmt, LPVOID lpContext ); -/*===========================================================================*/ -/* This function will simply set the top of stack to NULL. I only used it */ -/* just incase I want to add something later. */ -/*===========================================================================*/ -/* RETURN: TRUE. */ -/*===========================================================================*/ -BOOL InitTMgrHAL( PMESAD3DHAL pHAL ) -{ - DPF(( DBG_FUNC, "InitTMgrHAL();" )); - - /* Be clean my friend. */ - pHAL->pTMList = NULL; - - return TRUE; -} -/*===========================================================================*/ -/* This function will walk the Texture Managers linked list and destroy all */ -/* surfaces (SYSTEM/VIDEO). The texture objects themselves also will be */ -/* freed. */ -/* NOTE: this is per/context. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void TermTMgrHAL( PMESAD3DHAL pHAL ) -{ - DPF(( DBG_FUNC, "TermTMgrHAL();" )); - - if ( pHAL && pHAL->pTMList ) - { - /* Destroy the surface and remove the TMO from the stack. */ - while( DestroyTextureObject(pHAL,NULL) ); - - /* Be clean my friend. */ - pHAL->pTMList = NULL; - } -} -/*===========================================================================*/ -/* This function is a HACK as I don't know how I can disable a texture with-*/ -/* out booting it out. Is there know state change? */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void DisableTMgrHAL( PMESAD3DSHARED pShared ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - - DPF(( DBG_FUNC, "DisableTMgrHAL();" )); - - /* Check too see that we have a valid context. */ - if ( (pHAL == NULL) && (pHAL->lpD3DDevice != NULL) ) - { - DPF(( DBG_TXT_WARN, "Null HAL/Direct3D Device!" )); - return; - } - - // TODO: This is a hack to shut off textures. - pHAL->lpD3DDevice->SetTexture( 0, NULL ); -} -/*===========================================================================*/ -/* This function is the only entry into the TextureManager that Mesa/wgl */ -/* will see. It uses a dwAction to specify what we are doing. I did this as*/ -/* depending on the cards resources the action taken can change. */ -/* When this function is called we will always search the Texture Managers */ -/* linked list (per context remember) and try and find a structure that has */ -/* the same dwName. If we have a match we pull it out of the list and put it*/ -/* at the top of the list (TOL). If we don't find one then we create a struc*/ -/* and put it a TOL. This TOL idea makes for some caching as we will always */ -/* destroy Texture Surfaces from the bottom up... */ -/* All texture objects at this point will create a texture surface in System*/ -/* memory (SMEM). Then we will copy the Mesa texture into the surface using */ -/* the 'pixel' struc to get the translation info. So now this means that all*/ -/* textures that Mesa gives me I will have a Surface with a copy. If Mesa */ -/* changes the texture the I update the surface in (SMEM). */ -/* Now we have a texture struc and a Texture Surface in SMEM. At this point*/ -/* we create another surface on the card (VMEM). Finally we blt from the */ -/* SMEM to the VMEM and set the texture as current. Why do I need two? First*/ -/* this solves square textures. If the cards CAPS is square textures only */ -/* then I change the dimensions of the VMEM surface and the blt solves it for*/ -/* me. Second it saves me from filling D3D textures over and over if the */ -/* card needs to be creating and destroying surfaces because of low memory. */ -/* The surface in SMEM is expected to work always. When a surface has to be*/ -/* created in VMEM then we put it in a loop that tries to create the surface.*/ -/* If we create the surface ok then we brake from the loop. If we fail then */ -/* we will call 'FreeTextureMemory' that will return TRUE/FALSE as to whether*/ -/* memory was freed. If memory was freed then we can try again. If no memory*/ -/* was freed then it just can't fit. */ -/* 'FreeTextureMemory' will find the end of the list and start freeing VMEM */ -/* (never SMEM) surfaces that are not locked. */ -/* BIND - when we bind and there is a texture struct with a texture surface */ -/* in VMEM then we just make it current. If we have a struct and a surface */ -/* in SMEM but no VMEM surface then we create the surface in VMEM and blt */ -/* from the SMEM surface. If we have nothing its just like a creation... */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -extern "C" BOOL CreateTMgrHAL( PMESAD3DSHARED pShared, DWORD dwName, int level, DWORD dwRequestFlags, - RECT *rectDirty, DWORD dwWidth, DWORD dwHeight, DWORD dwAction, void *pPixels ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - PTM_OBJECT pTMObj, - pTemp; - DDSURFACEDESC2 ddsd2; - HRESULT rc; - - - DPF(( DBG_FUNC, "CreateTMgrHAL();" )); - - DPF(( DBG_TXT_INFO, "Texture:" )); - DPF(( DBG_TXT_INFO, "cx: %d cy: %d", dwWidth, dwHeight )); - DPF(( DBG_TXT_INFO, "Rect:" )); - if ( rectDirty ) - { - DPF(( DBG_TXT_INFO, "x0: %d y0: %d", rectDirty->left, rectDirty->top )); - DPF(( DBG_TXT_INFO, "x1: %d y1: %d", rectDirty->right, rectDirty->bottom )); - } - - /* Check too see that we have a valid context. */ - if ( (pHAL == NULL) && (pHAL->lpD3DDevice != NULL) ) - { - DPF(( DBG_TXT_WARN, "Null HAL/Direct3D Device!" )); - return FALSE; - } - - /*=================================================*/ - /* See if we can find this texture object by name. */ - /*=================================================*/ - for( pTMObj = pHAL->pTMList; pTMObj && (pTMObj->dwName != dwName); pTMObj = pTMObj->next ); - - /*=========================================================*/ - /* Allocate a new object if we didn't get a matching name. */ - /*=========================================================*/ - if ( pTMObj == NULL ) - { - pTMObj = (PTM_OBJECT)ALLOC( sizeof(TM_OBJECT) ); - if ( pTMObj == NULL ) - return FALSE; - memset( pTMObj, 0, sizeof(TM_OBJECT) ); - - /* Put the object at the beginning of the list. */ - pTMObj->next = pHAL->pTMList; - if ( pTMObj->next ) - { - pTemp = pTMObj->next; - pTemp->prev = pTMObj; - } - pHAL->pTMList = pTMObj; - } - else - { - /*===============================================================*/ - /* Make some caching happen by pulling this object to the front. */ - /*===============================================================*/ - if ( pHAL->pTMList != pTMObj ) - { - /* Pull the object out of the list. */ - if ( pTMObj->prev ) - { - pTemp = pTMObj->prev; - pTemp->next = pTMObj->next; - } - if ( pTMObj->next ) - { - pTemp = pTMObj->next; - pTemp->prev = pTMObj->prev; - } - - pTMObj->prev = NULL; - pTMObj->next = NULL; - - /* Put the object at the front of the list. */ - pTMObj->next = pHAL->pTMList; - if ( pTMObj->next ) - { - pTemp = pTMObj->next; - pTemp->prev = pTMObj; - } - pHAL->pTMList = pTMObj; - } - } - - /*========================================================*/ - /* If we are doing BIND and the texture is in VID memory. */ - /*========================================================*/ - if ( (dwAction == TM_ACTION_BIND) && pTMObj->lpDDS_Video ) - { - DPF(( DBG_TXT_PROFILE, "Cache HIT (%d)", dwName )); - - /* Make this the current texture. */ - rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); - if ( FAILED(rc) ) - { - DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); - pHAL->lpD3DDevice->SetTexture( 0, NULL ); - return FALSE; - } - - return TRUE; - } - - /*=================================================================*/ - /* If we are doing BIND and the texture is at least in SYS memory. */ - /*=================================================================*/ - if ( (dwAction == TM_ACTION_BIND) && pTMObj->lpDDS_System ) - { - DPF(( DBG_TXT_PROFILE, "Cache MISS (%d)", dwName )); - - /* Create the texture on the card. */ - rc = LoadTextureInVideo( pHAL, pTMObj ); - if ( rc == FALSE ) - return FALSE; - - /* Make this the current texture. */ - rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); - if ( FAILED(rc) ) - { - DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); - pHAL->lpD3DDevice->SetTexture( 0, NULL ); - return FALSE; - } - - return TRUE; - } - - /*=========================================================*/ - /* If we are doing UPDATE then try in VID first for speed. */ - /*=========================================================*/ - if ( (dwAction == TM_ACTION_UPDATE) && pTMObj->lpDDS_Video && - !(pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureCaps & D3DPTEXTURECAPS_SQUAREONLY) ) - { - DPF(( DBG_TXT_INFO, "Fix the SubTexture update Leigh!" )); - - /* Update the texture on the card. */ - UpdateTexture( pTMObj, TRUE, rectDirty, (UCHAR *)pPixels ); - - /* We updated the texture in VID so kill the SYS so we know its dirty. */ - if ( pTMObj->lpDDS_System ) - { - DPF(( DBG_TXT_INFO, "Release texture (SYS)" )); - DX_RESTORE( pTMObj->lpDDS_System ); - pTMObj->lpDDS_System->Release(); - pTMObj->lpDDS_System = NULL; - } - - /* Make this the current texture. */ - rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); - if ( FAILED(rc) ) - { - DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); - pHAL->lpD3DDevice->SetTexture( 0, NULL ); - return FALSE; - } - - return TRUE; - } - - /*===========================================================*/ - /* If we are doing UPDATE then try in SYS still gives speed. */ - /*===========================================================*/ - if ( (dwAction == TM_ACTION_UPDATE) && pTMObj->lpDDS_System ) - { - DPF(( DBG_TXT_INFO, "Fix the SubTexture update Leigh!" )); - - /* Update the texture in SYS. */ - UpdateTexture( pTMObj, FALSE, NULL, (UCHAR *)pPixels ); - - /* We updated the SYS texture only so now blt to the VID. */ - rc = LoadTextureInVideo( pHAL, pTMObj ); - if ( rc == FALSE ) - return FALSE; - - /* Make this the current texture. */ - rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); - if ( FAILED(rc) ) - { - DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); - pHAL->lpD3DDevice->SetTexture( 0, NULL ); - return FALSE; - } - - return TRUE; - } - - /* At this point we have a valid Texture Manager Object with updated */ - /* links. We now need to create or update a texture surface that is */ - /* in system memory. Every texture has a copy in system so we can use*/ - /* blt to solve problems with textures allocated on the card (square */ - /* only textures, pixelformats...). */ - - // TODO: make support for update also. Dirty rectangle basicly... - - /* Kill the interface if we have one no matter what. */ - if ( pTMObj->lpD3DTexture2 ) - { - DPF(( DBG_TXT_INFO, "Release Texture2" )); - pTMObj->lpD3DTexture2->Release(); - pTMObj->lpD3DTexture2 = NULL; - } - - /* Kill the system surface. TODO: should try to get the SubIMage going again */ - if ( pTMObj->lpDDS_System ) - { - DPF(( DBG_TXT_INFO, "Release texture (SYS)" )); - DX_RESTORE( pTMObj->lpDDS_System ); - pTMObj->lpDDS_System->Release(); - pTMObj->lpDDS_System = NULL; - } - - /* Kill the Video surface. TODO: need some reuse system... */ - if ( pTMObj->lpDDS_Video ) - { - DPF(( DBG_TXT_INFO, "Release texture (VID)" )); - DX_RESTORE( pTMObj->lpDDS_Video ); - pTMObj->lpDDS_Video->Release(); - pTMObj->lpDDS_Video = NULL; - } - - /*================================================================*/ - /* Translate the the Mesa/OpenGL pixel channels to the D3D flags. */ - /*================================================================*/ - switch( dwRequestFlags ) - { - case GL_ALPHA: - dwRequestFlags = DDPF_ALPHA; - DPF(( DBG_TXT_WARN, "GL_ALPHA not supported!)" )); - return FALSE; - - case GL_INTENSITY: - case GL_LUMINANCE: - DPF(( DBG_TXT_WARN, "GL_INTENSITY/GL_LUMINANCE not supported!)" )); - dwRequestFlags = DDPF_LUMINANCE; - return FALSE; - - case GL_LUMINANCE_ALPHA: - DPF(( DBG_TXT_WARN, "GL_LUMINANCE_ALPHA not supported!)" )); - dwRequestFlags = DDPF_LUMINANCE | DDPF_ALPHAPIXELS; - return FALSE; - - case GL_RGB: - DPF(( DBG_TXT_INFO, "Texture -> GL_RGB" )); - dwRequestFlags = DDPF_RGB; - break; - - case GL_RGBA: - DPF(( DBG_TXT_INFO, "Texture -> GL_RGBA" )); - dwRequestFlags = DDPF_RGB | DDPF_ALPHAPIXELS; - break; - } - - /*==============================*/ - /* Populate the texture object. */ - /*==============================*/ - pTMObj->dwName = dwName; - pTMObj->lpD3DDevice = pHAL->lpD3DDevice; - pTMObj->dwFlags = dwRequestFlags; - if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureCaps & D3DPTEXTURECAPS_SQUAREONLY ) - { - DPF(( DBG_TXT_INFO, "Convert to Square..." )); - pTMObj->dwSHeight = dwHeight; - pTMObj->dwSWidth = dwWidth; - - /* Shrink non-square textures. */ - pTMObj->dwVHeight = (dwHeight > dwWidth) ? dwWidth : dwHeight; - pTMObj->dwVWidth = (dwHeight > dwWidth) ? dwWidth : dwHeight; - } - else - { - pTMObj->dwSHeight = dwHeight; - pTMObj->dwSWidth = dwWidth; - pTMObj->dwVHeight = dwHeight; - pTMObj->dwVWidth = dwWidth; - } - - /*========================*/ - /* Create SYSTEM surface. */ - /*========================*/ - - /* Request a surface in system memory. */ - memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); - ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); - ddsd2.dwWidth = pTMObj->dwSWidth; - ddsd2.dwHeight = pTMObj->dwSHeight; - ddsd2.dwFlags = DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT | DDSD_PIXELFORMAT; - ddsd2.ddsCaps.dwCaps = DDSCAPS_TEXTURE | DDSCAPS_SYSTEMMEMORY; - ddsd2.ddsCaps.dwCaps2 = 0L; - memset( &ddsd2.ddpfPixelFormat, 0, sizeof(DDPIXELFORMAT) ); - ddsd2.ddpfPixelFormat.dwSize = sizeof( DDPIXELFORMAT ); - ddsd2.ddpfPixelFormat.dwFlags = dwRequestFlags; - rc = pHAL->lpD3DDevice->EnumTextureFormats( EnumPFHook, &ddsd2.ddpfPixelFormat ); - if ( FAILED(rc) ) - { - RIP( pHAL, "EnumerTextureFormats (SYSTEM)->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Create the surface using the enumerated pixelformat. */ - rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pTMObj->lpDDS_System, NULL ); - if ( FAILED(rc) ) - { - RIP( pHAL, "CreateSurface (TEXTURE/SYSTEM)->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Solve the pixel mapping info using the surface pixelformat. */ - Solve8BitChannelPixelFormat( &ddsd2.ddpfPixelFormat, &pTMObj->pixel ); - - /*===================================================================*/ - /* Fill the texture using the PixelInfo structure to do the mapping. */ - /*===================================================================*/ - UpdateTexture( pTMObj, FALSE, NULL, (UCHAR *)pPixels ); - - /*=======================*/ - /* Create VIDEO surface. */ - /*=======================*/ - rc = LoadTextureInVideo( pHAL, pTMObj ); - if ( rc == FALSE ) - return FALSE; - - /* Make this the current texture. */ - rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); - if ( FAILED(rc) ) - { - DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); - pHAL->lpD3DDevice->SetTexture( 0, NULL ); - return FALSE; - } - - return TRUE; -} -/*===========================================================================*/ -/* This function will handle the creation and destruction of the texture */ -/* surfaces on the card. Using the dw'V'Width/Height dimensions the call */ -/* try and create the texture on the card and keep using FreeTextureMemory */ -/* until the surace can be created. Once the surface is created we get the */ -/* interface that we will use to make it the current texture. I didn't put */ -/* the code to make the texture current in this function as BIND needs to */ -/* use the same code and this function doesn't always get called when we do a*/ -/* bind. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -static BOOL LoadTextureInVideo( PMESAD3DHAL pHAL, PTM_OBJECT pTMObj ) -{ - DDSURFACEDESC2 ddsd2; - HRESULT rc; - - DPF(( DBG_FUNC, "LoadTextureInVideo();" )); - - /* Kill the interface if we have one no matter what. */ - if ( pTMObj->lpD3DTexture2 ) - { - DPF(( DBG_TXT_INFO, "Release Texture2" )); - pTMObj->lpD3DTexture2->Release(); - pTMObj->lpD3DTexture2 = NULL; - } - - /* Kill the Video surface. TODO: need some reuse system... */ - if ( pTMObj->lpDDS_Video ) - { - DPF(( DBG_TXT_INFO, "Release texture (VID)" )); - DX_RESTORE( pTMObj->lpDDS_Video ); - pTMObj->lpDDS_Video->Release(); - pTMObj->lpDDS_Video = NULL; - } - - /* Request a surface in Video memory. */ - memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); - ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); - ddsd2.dwWidth = pTMObj->dwVWidth; - ddsd2.dwHeight = pTMObj->dwVHeight; - ddsd2.dwFlags = DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT | DDSD_PIXELFORMAT; - ddsd2.ddsCaps.dwCaps = DDSCAPS_TEXTURE | DDSCAPS_VIDEOMEMORY; - ddsd2.ddsCaps.dwCaps2 = 0L; - memset( &ddsd2.ddpfPixelFormat, 0, sizeof(DDPIXELFORMAT) ); - ddsd2.ddpfPixelFormat.dwSize = sizeof( DDPIXELFORMAT ); - ddsd2.ddpfPixelFormat.dwFlags = pTMObj->dwFlags; - rc = pHAL->lpD3DDevice->EnumTextureFormats( EnumPFHook, &ddsd2.ddpfPixelFormat ); - if ( FAILED(rc) ) - { - RIP( pHAL, "EnumerTextureFormats ->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Make sure we lock so we don't nuke this texture trying to free memory for it. */ - pTMObj->bLock = TRUE; - - /* Start a loop that will free all textures until we have created the texture */ - /* surface or we can't free up more memory. */ - do - { - /* Try to create the texture surface. */ - rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pTMObj->lpDDS_Video, NULL ); - if ( !FAILED(rc) ) - break; - - DPF(( DBG_TXT_INFO, "Free Texture Memory" )); - - /* DestroyTexture will return TRUE if a surface was freed. */ - } while( FreeTextureMemory(pHAL,NULL) ); - - /* Make sure we unlock or we won't be able to nuke the TMO later. */ - pTMObj->bLock = FALSE; - - /* Did we create a valid texture surface? */ - if ( FAILED(rc) ) - { - DPF(( DBG_TXT_WARN, "Failed to load texture" )); - pHAL->lpD3DDevice->SetTexture( 0, NULL ); - return FALSE; - } - - DX_RESTORE( pTMObj->lpDDS_System ); - DX_RESTORE( pTMObj->lpDDS_Video ); - - DPF(( DBG_TXT_INFO, "Texture Blt SYSTEM -> VID" )); - - /* Now blt the texture in system memory to the card. */ - rc = pTMObj->lpDDS_Video->Blt( NULL, pTMObj->lpDDS_System, NULL, DDBLT_WAIT, NULL ); - if ( FAILED(rc) ) - { - RIP( pHAL, "Blt (TEXTURE) ->", ErrorStringD3D(rc) ); - return FALSE; - } - - /* Get the Texture interface that is used to render with. */ - pTMObj->lpDDS_Video->QueryInterface( IID_IDirect3DTexture2, (void **)&pTMObj->lpD3DTexture2 ); - if ( pTMObj->lpD3DTexture2 == NULL ) - { - DPF(( DBG_TXT_WARN, "Failed QueryTextureInterface" )); - pHAL->lpD3DDevice->SetTexture( 0, NULL ); - return FALSE; - } - - return TRUE; -} -/*===========================================================================*/ -/* If this function gets a texture object struc then we will try and free */ -/* it. If we get a NULL then we will search from the bottom up and free one */ -/* VMEM surface. I can only free when the surface isn't locked and of course*/ -/* there must be a VMEM surface. We never free SMEM surfaces as that isn't */ -/* the point. */ -/* TODO: should have a pointer to the bottom of the stack really. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static BOOL FreeTextureMemory( PMESAD3DHAL pHAL, PTM_OBJECT pTMObject ) -{ - PTM_OBJECT pCurrent; - BOOL bFreed = FALSE; - - DPF(( DBG_FUNC, "FreeTextureMemory();" )); - DPF(( DBG_TXT_WARN, "FREE TEXTURE!" )); - - /* Just to be safe. */ - if ( !pHAL || !pHAL->pTMList ) - { - DPF(( DBG_TXT_WARN, "FreeTextureMemory() -> NULL pHAL/pHAL->pTMList" )); - return FALSE; - } - - /* Free the last texture in the list. */ - if ( pTMObject == NULL ) - { - DPF(( DBG_TXT_INFO, "Free Last texture in cache" )); - - /* Find the last texture object. */ - for( pCurrent = pHAL->pTMList; pCurrent->next; pCurrent = pCurrent->next ); - - /* Now backup until we find a texture on the card. */ - while( pCurrent && (pCurrent->lpDDS_Video == NULL) && (pCurrent->bLock == FALSE) ) - pCurrent = pCurrent->prev; - - /* Didn't find anything. */ - if ( pCurrent == NULL ) - { - DPF(( DBG_TXT_INFO, "No texture memory freed" )); - return FALSE; - } - } - else - { - /* See if we can find this texture object. */ - for( pCurrent = pHAL->pTMList; pCurrent && (pCurrent != pTMObject); pCurrent = pCurrent->next ); - - /* Didn't find anything. */ - if ( pCurrent == NULL ) - { - DPF(( DBG_TXT_INFO, "Requested texture to be freed NOT FOUND" )); - return FALSE; - } - } - - /* Can't free this baby. */ - if ( pCurrent->bLock == TRUE ) - { - DPF(( DBG_TXT_WARN, "Requested texture LOCKED" )); - return FALSE; - } - - /* Free the texture memory. */ - if ( pCurrent->lpD3DTexture2 ) - { - DPF(( DBG_TXT_INFO, "Release Texture2" )); - pCurrent->lpD3DTexture2->Release(); - pCurrent->lpD3DTexture2 = NULL; - bFreed = TRUE; - } - if ( pCurrent->lpDDS_Video ) - { - DPF(( DBG_TXT_INFO, "Release texture (VID):" )); - DPF(( DBG_TXT_INFO, "dwName: %d", pCurrent->dwName )); - DPF(( DBG_TXT_INFO, "cx: %d, cy: %d", pCurrent->dwVWidth, pCurrent->dwVHeight )); - pCurrent->lpDDS_Video->Release(); - pCurrent->lpDDS_Video = NULL; - bFreed = TRUE; - } - - return bFreed; -} -/*===========================================================================*/ -/* This function searches the linked list of texture objects in the supplied*/ -/* D3Dwrapper structure. If it finds a match it will free it and pull it out*/ -/* of the linked list. The function works on the bases of a matching pointer*/ -/* to the object (not matching content). */ -/* If the function gets passed a NULL then we want to free the last texture */ -/* object in the list. Used in a loop to destory all. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -static BOOL DestroyTextureObject( PMESAD3DHAL pHAL, PTM_OBJECT pTMObject ) -{ - PTM_OBJECT pCurrent; - - DPF(( DBG_FUNC, "DestoryTextureObject();" )); - - /* Just to be safe. */ - if ( !pHAL || !pHAL->pTMList ) - { - DPF(( DBG_TXT_WARN, "DestroyTextureObject() -> NULL pHAL/pHAL->pTMList" )); - return FALSE; - } - - /* Free the last texture in the list. */ - if ( pTMObject == NULL ) - { - /* Find the last texture object. */ - for( pCurrent = pHAL->pTMList; pCurrent->next; pCurrent = pCurrent->next ); - } - else - { - /* See if we can find this texture object. */ - for( pCurrent = pHAL->pTMList; pCurrent && (pCurrent != pTMObject); pCurrent = pCurrent->next ); - - /* Didn't find anything. */ - if ( pCurrent == NULL ) - { - DPF(( DBG_TXT_WARN, "No textures to be freed" )); - return FALSE; - } - } - - /* Can't free this baby. */ - if ( pCurrent->bLock == TRUE ) - { - DPF(( DBG_TXT_WARN, "Requested texture to be freed LOCKED" )); - return FALSE; - } - - /* Free the texture memory. */ - if ( pCurrent->lpD3DTexture2 ) - { - DPF(( DBG_TXT_INFO, "Release Texture2" )); - pCurrent->lpD3DTexture2->Release(); - pCurrent->lpD3DTexture2 = NULL; - } - if ( pCurrent->lpDDS_Video ) - { - DPF(( DBG_TXT_INFO, "Release texture (VID):" )); - pCurrent->lpDDS_Video->Release(); - pCurrent->lpDDS_Video = NULL; - } - if ( pCurrent->lpDDS_System ) - { - DPF(( DBG_TXT_INFO, "Release texture (SYS):" )); - pCurrent->lpDDS_System->Release(); - pCurrent->lpDDS_System = NULL; - } - - /* Pull this texture out of the list. */ - if ( pCurrent == pHAL->pTMList ) - pHAL->pTMList = NULL; - if ( pCurrent->prev ) - (pCurrent->prev)->next = pCurrent->next; - if ( pCurrent->next ) - (pCurrent->next)->prev = pCurrent->prev; - FREE( pCurrent ); - - return TRUE; -} -/*===========================================================================*/ -/* This function is the callback function that gets called when we are doing*/ -/* an enumeration of the texture formats supported by this device. The choice*/ -/* is made by checking to see if we have a match with the supplied D3D pixel-*/ -/* format. So the enumeration has to pass a desired D3D PF as the user var. */ -/*===========================================================================*/ -/* RETURN: D3DENUMRET_OK, D3DENUMRET_CANCEL. */ -/*===========================================================================*/ -static void UpdateTexture( PTM_OBJECT pTMObj, BOOL bVideo, RECT *pRect, UCHAR *pixels ) -{ - LPDIRECTDRAWSURFACE4 lpDDS; - DDSURFACEDESC2 ddsd2; - DWORD srcPitch, - dwHeight, - dwWidth, - dwCol, - dwColor; - UCHAR *pSrc, - *pSrcRow, - *pDest, - *pDestRow; - int rc; - - // TODO: Do I need to pass the h/w when its in the object! - DPF(( DBG_FUNC, "UpdateTexture();" )); - - /* Get the surface pointer we are looking for. */ - lpDDS = (bVideo) ? pTMObj->lpDDS_Video : pTMObj->lpDDS_System; - - /*===================================================================*/ - /* Fill the texture using the PixelInfo structure to do the mapping. */ - /*===================================================================*/ - - /* Get the surface pointer. */ - memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); - ddsd2.dwSize = sizeof(DDSURFACEDESC2); - rc = lpDDS->Lock( NULL, &ddsd2, DDLOCK_WAIT, NULL ); - if ( FAILED(rc) ) - { - RIP( NULL, "Lock (TEXTURE/SYSTEM)->", ErrorStringD3D(rc) ); - return; - } - - /* For now we are only updating the system surface so use its dimensions. */ - dwWidth = (bVideo) ? pTMObj->dwVWidth : pTMObj->dwSWidth; - dwHeight = (bVideo) ? pTMObj->dwVHeight : pTMObj->dwSHeight; - - /* If we are updating the whole surface then the pDest/pSrc will */ - /* always be the same. */ - if ( pRect == NULL ) - { - pDest = (UCHAR *)ddsd2.lpSurface; - pSrc = pixels; - } - - /* Fill the texture surface based on the pixelformat flags. */ - if ( pTMObj->dwFlags == (DDPF_RGB | DDPF_ALPHAPIXELS) ) - { - srcPitch = dwWidth * 4; - if ( pRect ) - { - pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); - pSrc = pixels + (pRect->top * dwWidth * 4) + (pRect->left * 4); - dwHeight = (pRect->bottom - pRect->top); - dwWidth = (pRect->right - pRect->left); - } - - for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) - { - for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) - { - dwColor = ( ((DWORD)(*(pSrc ) * pTMObj->pixel.rScale)) << pTMObj->pixel.rShift ); - dwColor |= ( ((DWORD)(*(pSrc+1) * pTMObj->pixel.gScale)) << pTMObj->pixel.gShift ); - dwColor |= ( ((DWORD)(*(pSrc+2) * pTMObj->pixel.bScale)) << pTMObj->pixel.bShift ); - if ( pTMObj->pixel.aScale == -1.0 ) - dwColor |= ( (*(pSrc+3) & 0x80) ? (1 << pTMObj->pixel.aShift) : 0 ); - else - dwColor |= ( ((DWORD)(*(pSrc+3) * pTMObj->pixel.aScale)) << pTMObj->pixel.aShift ); - memcpy( pDest, &dwColor, pTMObj->pixel.cb ); - pDest += pTMObj->pixel.cb; - pSrc += 4; - } - } - } - else if ( pTMObj->dwFlags == DDPF_RGB ) - { - srcPitch = dwWidth * 3; - if ( pRect ) - { - pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); - pSrc = pixels + (pRect->top * dwWidth * 3) + (pRect->left * 3); - dwHeight = (pRect->bottom - pRect->top); - dwWidth = (pRect->right - pRect->left); - } - - for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) - { - for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) - { - dwColor = ( ((DWORD)(*(pSrc ) * pTMObj->pixel.rScale)) << pTMObj->pixel.rShift ); - dwColor |= ( ((DWORD)(*(pSrc+1) * pTMObj->pixel.gScale)) << pTMObj->pixel.gShift ); - dwColor |= ( ((DWORD)(*(pSrc+2) * pTMObj->pixel.bScale)) << pTMObj->pixel.bShift ); - memcpy( pDest, &dwColor, pTMObj->pixel.cb ); - pDest += pTMObj->pixel.cb; - pSrc += 3; - } - } - } - else if ( pTMObj->dwFlags == (DDPF_LUMINANCE | DDPF_ALPHAPIXELS) ) - { - srcPitch = dwWidth * 2; - if ( pRect ) - { - pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); - pSrc = pixels + (pRect->top * dwWidth * 2) + (pRect->left * 2); - dwHeight = (pRect->bottom - pRect->top); - dwWidth = (pRect->right - pRect->left); - } - - for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) - { - for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) - { - dwColor = ( ((DWORD)(*(pSrc ) * pTMObj->pixel.rScale)) << pTMObj->pixel.rShift ); - if ( pTMObj->pixel.aScale == -1.0 ) - dwColor |= ( (*(pSrc+1) & 0x80) ? (1 << pTMObj->pixel.aShift) : 0 ); - else - dwColor |= ( ((DWORD)(*(pSrc+1) * pTMObj->pixel.aScale)) << pTMObj->pixel.aShift ); - memcpy( pDest, &dwColor, pTMObj->pixel.cb ); - pDest += pTMObj->pixel.cb; - pSrc += 2; - } - } - } - else if ( pTMObj->dwFlags == DDPF_LUMINANCE ) - { - srcPitch = dwWidth; - if ( pRect ) - { - pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); - pSrc = pixels + (pRect->top * dwWidth) + (pRect->left); - dwHeight = (pRect->bottom - pRect->top); - dwWidth = (pRect->right - pRect->left); - } - - for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) - { - for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) - { - dwColor = ( ((DWORD)(*pSrc * pTMObj->pixel.rScale)) << pTMObj->pixel.rShift ); - memcpy( pDest, &dwColor, pTMObj->pixel.cb ); - pDest += pTMObj->pixel.cb; - pSrc++; - } - } - } - else if ( pTMObj->dwFlags == DDPF_ALPHAPIXELS ) - { - srcPitch = dwWidth; - if ( pRect ) - { - pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); - pSrc = pixels + (pRect->top * dwWidth) + (pRect->left); - dwHeight = (pRect->bottom - pRect->top); - dwWidth = (pRect->right - pRect->left); - } - - for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) - { - for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) - { - if ( pTMObj->pixel.aScale == -1.0 ) - dwColor = ( (*pSrc & 0x80) ? (1 << pTMObj->pixel.aShift) : 0 ); - else - dwColor = ( ((DWORD)(*pSrc * pTMObj->pixel.aScale)) << pTMObj->pixel.aShift ); - memcpy( pDest, &dwColor, pTMObj->pixel.cb ); - pDest += pTMObj->pixel.cb; - pSrc++; - } - } - } - - /* Unlock the surface. */ - rc = lpDDS->Unlock( NULL ); - if ( FAILED(rc) ) - { - RIP( NULL, "Unlock (TEXTURE/SYSTEM)->", ErrorStringD3D(rc) ); - } -} -/*===========================================================================*/ -/* This function is the callback function that gets called when we are doing*/ -/* an enumeration of the texture formats supported by this device. The choice*/ -/* is made by checking to see if we have a match with the supplied D3D pixel-*/ -/* format. So the enumeration has to pass a desired D3D PF as the user var. */ -/*===========================================================================*/ -/* RETURN: D3DENUMRET_OK, D3DENUMRET_CANCEL. */ -/*===========================================================================*/ -HRESULT CALLBACK EnumPFHook( LPDDPIXELFORMAT lpDDPixFmt, LPVOID lpContext ) -{ - LPDDPIXELFORMAT lpDDPixFmtRequest = (LPDDPIXELFORMAT)lpContext; - PIXELINFO pixel; - - DPF(( DBG_FUNC, "EnumPFHook();" )); - - if ( lpDDPixFmt->dwFlags == lpDDPixFmtRequest->dwFlags ) - { - /* Are we looking for an alpha channel? */ - if ( lpDDPixFmtRequest->dwFlags & DDPF_ALPHAPIXELS ) - { - /* Try for something that has more then 1bits of Alpha. */ - Solve8BitChannelPixelFormat( lpDDPixFmt, &pixel ); - if ( pixel.aScale == -1.0 ) - { - /* Save this format no matter what as its a match of sorts. */ - memcpy( lpDDPixFmtRequest, lpDDPixFmt, sizeof(DDPIXELFORMAT) ); - return D3DENUMRET_OK; - } - } - - /* Save this format as its a good match. */ - memcpy( lpDDPixFmtRequest, lpDDPixFmt, sizeof(DDPIXELFORMAT) ); - - /* We are happy at this point so lets leave. */ - return D3DENUMRET_CANCEL; - } - - return D3DENUMRET_OK; -} - - +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#include "D3DHAL.h" +/*===========================================================================*/ +/* Local function prototypes. */ +/*===========================================================================*/ +static void UpdateTexture( PTM_OBJECT pTMObj, BOOL bVideo, RECT *pRect, UCHAR *pixels ); +static BOOL LoadTextureInVideo( PMESAD3DHAL pHAL, PTM_OBJECT pTMObj ); +static BOOL FreeTextureMemory( PMESAD3DHAL pHAL, PTM_OBJECT pTMObject ); +static BOOL DestroyTextureObject( PMESAD3DHAL pHAL, PTM_OBJECT pTMObject ); +HRESULT CALLBACK EnumPFHook( LPDDPIXELFORMAT lpDDPixFmt, LPVOID lpContext ); +/*===========================================================================*/ +/* This function will simply set the top of stack to NULL. I only used it */ +/* just incase I want to add something later. */ +/*===========================================================================*/ +/* RETURN: TRUE. */ +/*===========================================================================*/ +BOOL InitTMgrHAL( PMESAD3DHAL pHAL ) +{ + DPF(( DBG_FUNC, "InitTMgrHAL();" )); + + /* Be clean my friend. */ + pHAL->pTMList = NULL; + + return TRUE; +} +/*===========================================================================*/ +/* This function will walk the Texture Managers linked list and destroy all */ +/* surfaces (SYSTEM/VIDEO). The texture objects themselves also will be */ +/* freed. */ +/* NOTE: this is per/context. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void TermTMgrHAL( PMESAD3DHAL pHAL ) +{ + DPF(( DBG_FUNC, "TermTMgrHAL();" )); + + if ( pHAL && pHAL->pTMList ) + { + /* Destroy the surface and remove the TMO from the stack. */ + while( DestroyTextureObject(pHAL,NULL) ); + + /* Be clean my friend. */ + pHAL->pTMList = NULL; + } +} +/*===========================================================================*/ +/* This function is a HACK as I don't know how I can disable a texture with-*/ +/* out booting it out. Is there know state change? */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void DisableTMgrHAL( PMESAD3DSHARED pShared ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + + DPF(( DBG_FUNC, "DisableTMgrHAL();" )); + + /* Check too see that we have a valid context. */ + if ( (pHAL == NULL) && (pHAL->lpD3DDevice != NULL) ) + { + DPF(( DBG_TXT_WARN, "Null HAL/Direct3D Device!" )); + return; + } + + // TODO: This is a hack to shut off textures. + pHAL->lpD3DDevice->SetTexture( 0, NULL ); +} +/*===========================================================================*/ +/* This function is the only entry into the TextureManager that Mesa/wgl */ +/* will see. It uses a dwAction to specify what we are doing. I did this as*/ +/* depending on the cards resources the action taken can change. */ +/* When this function is called we will always search the Texture Managers */ +/* linked list (per context remember) and try and find a structure that has */ +/* the same dwName. If we have a match we pull it out of the list and put it*/ +/* at the top of the list (TOL). If we don't find one then we create a struc*/ +/* and put it a TOL. This TOL idea makes for some caching as we will always */ +/* destroy Texture Surfaces from the bottom up... */ +/* All texture objects at this point will create a texture surface in System*/ +/* memory (SMEM). Then we will copy the Mesa texture into the surface using */ +/* the 'pixel' struc to get the translation info. So now this means that all*/ +/* textures that Mesa gives me I will have a Surface with a copy. If Mesa */ +/* changes the texture the I update the surface in (SMEM). */ +/* Now we have a texture struc and a Texture Surface in SMEM. At this point*/ +/* we create another surface on the card (VMEM). Finally we blt from the */ +/* SMEM to the VMEM and set the texture as current. Why do I need two? First*/ +/* this solves square textures. If the cards CAPS is square textures only */ +/* then I change the dimensions of the VMEM surface and the blt solves it for*/ +/* me. Second it saves me from filling D3D textures over and over if the */ +/* card needs to be creating and destroying surfaces because of low memory. */ +/* The surface in SMEM is expected to work always. When a surface has to be*/ +/* created in VMEM then we put it in a loop that tries to create the surface.*/ +/* If we create the surface ok then we brake from the loop. If we fail then */ +/* we will call 'FreeTextureMemory' that will return TRUE/FALSE as to whether*/ +/* memory was freed. If memory was freed then we can try again. If no memory*/ +/* was freed then it just can't fit. */ +/* 'FreeTextureMemory' will find the end of the list and start freeing VMEM */ +/* (never SMEM) surfaces that are not locked. */ +/* BIND - when we bind and there is a texture struct with a texture surface */ +/* in VMEM then we just make it current. If we have a struct and a surface */ +/* in SMEM but no VMEM surface then we create the surface in VMEM and blt */ +/* from the SMEM surface. If we have nothing its just like a creation... */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +extern "C" BOOL CreateTMgrHAL( PMESAD3DSHARED pShared, DWORD dwName, int level, DWORD dwRequestFlags, + RECT *rectDirty, DWORD dwWidth, DWORD dwHeight, DWORD dwAction, void *pPixels ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + PTM_OBJECT pTMObj, + pTemp; + DDSURFACEDESC2 ddsd2; + HRESULT rc; + + + DPF(( DBG_FUNC, "CreateTMgrHAL();" )); + + DPF(( DBG_TXT_INFO, "Texture:" )); + DPF(( DBG_TXT_INFO, "cx: %d cy: %d", dwWidth, dwHeight )); + DPF(( DBG_TXT_INFO, "Rect:" )); + if ( rectDirty ) + { + DPF(( DBG_TXT_INFO, "x0: %d y0: %d", rectDirty->left, rectDirty->top )); + DPF(( DBG_TXT_INFO, "x1: %d y1: %d", rectDirty->right, rectDirty->bottom )); + } + + /* Check too see that we have a valid context. */ + if ( (pHAL == NULL) && (pHAL->lpD3DDevice != NULL) ) + { + DPF(( DBG_TXT_WARN, "Null HAL/Direct3D Device!" )); + return FALSE; + } + + /*=================================================*/ + /* See if we can find this texture object by name. */ + /*=================================================*/ + for( pTMObj = pHAL->pTMList; pTMObj && (pTMObj->dwName != dwName); pTMObj = pTMObj->next ); + + /*=========================================================*/ + /* Allocate a new object if we didn't get a matching name. */ + /*=========================================================*/ + if ( pTMObj == NULL ) + { + pTMObj = (PTM_OBJECT)ALLOC( sizeof(TM_OBJECT) ); + if ( pTMObj == NULL ) + return FALSE; + memset( pTMObj, 0, sizeof(TM_OBJECT) ); + + /* Put the object at the beginning of the list. */ + pTMObj->next = pHAL->pTMList; + if ( pTMObj->next ) + { + pTemp = pTMObj->next; + pTemp->prev = pTMObj; + } + pHAL->pTMList = pTMObj; + } + else + { + /*===============================================================*/ + /* Make some caching happen by pulling this object to the front. */ + /*===============================================================*/ + if ( pHAL->pTMList != pTMObj ) + { + /* Pull the object out of the list. */ + if ( pTMObj->prev ) + { + pTemp = pTMObj->prev; + pTemp->next = pTMObj->next; + } + if ( pTMObj->next ) + { + pTemp = pTMObj->next; + pTemp->prev = pTMObj->prev; + } + + pTMObj->prev = NULL; + pTMObj->next = NULL; + + /* Put the object at the front of the list. */ + pTMObj->next = pHAL->pTMList; + if ( pTMObj->next ) + { + pTemp = pTMObj->next; + pTemp->prev = pTMObj; + } + pHAL->pTMList = pTMObj; + } + } + + /*========================================================*/ + /* If we are doing BIND and the texture is in VID memory. */ + /*========================================================*/ + if ( (dwAction == TM_ACTION_BIND) && pTMObj->lpDDS_Video ) + { + DPF(( DBG_TXT_PROFILE, "Cache HIT (%d)", dwName )); + + /* Make this the current texture. */ + rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); + if ( FAILED(rc) ) + { + DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); + pHAL->lpD3DDevice->SetTexture( 0, NULL ); + return FALSE; + } + + return TRUE; + } + + /*=================================================================*/ + /* If we are doing BIND and the texture is at least in SYS memory. */ + /*=================================================================*/ + if ( (dwAction == TM_ACTION_BIND) && pTMObj->lpDDS_System ) + { + DPF(( DBG_TXT_PROFILE, "Cache MISS (%d)", dwName )); + + /* Create the texture on the card. */ + rc = LoadTextureInVideo( pHAL, pTMObj ); + if ( rc == FALSE ) + return FALSE; + + /* Make this the current texture. */ + rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); + if ( FAILED(rc) ) + { + DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); + pHAL->lpD3DDevice->SetTexture( 0, NULL ); + return FALSE; + } + + return TRUE; + } + + /*=========================================================*/ + /* If we are doing UPDATE then try in VID first for speed. */ + /*=========================================================*/ + if ( (dwAction == TM_ACTION_UPDATE) && pTMObj->lpDDS_Video && + !(pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureCaps & D3DPTEXTURECAPS_SQUAREONLY) ) + { + DPF(( DBG_TXT_INFO, "Fix the SubTexture update Leigh!" )); + + /* Update the texture on the card. */ + UpdateTexture( pTMObj, TRUE, rectDirty, (UCHAR *)pPixels ); + + /* We updated the texture in VID so kill the SYS so we know its dirty. */ + if ( pTMObj->lpDDS_System ) + { + DPF(( DBG_TXT_INFO, "Release texture (SYS)" )); + DX_RESTORE( pTMObj->lpDDS_System ); + pTMObj->lpDDS_System->Release(); + pTMObj->lpDDS_System = NULL; + } + + /* Make this the current texture. */ + rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); + if ( FAILED(rc) ) + { + DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); + pHAL->lpD3DDevice->SetTexture( 0, NULL ); + return FALSE; + } + + return TRUE; + } + + /*===========================================================*/ + /* If we are doing UPDATE then try in SYS still gives speed. */ + /*===========================================================*/ + if ( (dwAction == TM_ACTION_UPDATE) && pTMObj->lpDDS_System ) + { + DPF(( DBG_TXT_INFO, "Fix the SubTexture update Leigh!" )); + + /* Update the texture in SYS. */ + UpdateTexture( pTMObj, FALSE, NULL, (UCHAR *)pPixels ); + + /* We updated the SYS texture only so now blt to the VID. */ + rc = LoadTextureInVideo( pHAL, pTMObj ); + if ( rc == FALSE ) + return FALSE; + + /* Make this the current texture. */ + rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); + if ( FAILED(rc) ) + { + DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); + pHAL->lpD3DDevice->SetTexture( 0, NULL ); + return FALSE; + } + + return TRUE; + } + + /* At this point we have a valid Texture Manager Object with updated */ + /* links. We now need to create or update a texture surface that is */ + /* in system memory. Every texture has a copy in system so we can use*/ + /* blt to solve problems with textures allocated on the card (square */ + /* only textures, pixelformats...). */ + + // TODO: make support for update also. Dirty rectangle basicly... + + /* Kill the interface if we have one no matter what. */ + if ( pTMObj->lpD3DTexture2 ) + { + DPF(( DBG_TXT_INFO, "Release Texture2" )); + pTMObj->lpD3DTexture2->Release(); + pTMObj->lpD3DTexture2 = NULL; + } + + /* Kill the system surface. TODO: should try to get the SubIMage going again */ + if ( pTMObj->lpDDS_System ) + { + DPF(( DBG_TXT_INFO, "Release texture (SYS)" )); + DX_RESTORE( pTMObj->lpDDS_System ); + pTMObj->lpDDS_System->Release(); + pTMObj->lpDDS_System = NULL; + } + + /* Kill the Video surface. TODO: need some reuse system... */ + if ( pTMObj->lpDDS_Video ) + { + DPF(( DBG_TXT_INFO, "Release texture (VID)" )); + DX_RESTORE( pTMObj->lpDDS_Video ); + pTMObj->lpDDS_Video->Release(); + pTMObj->lpDDS_Video = NULL; + } + + /*================================================================*/ + /* Translate the the Mesa/OpenGL pixel channels to the D3D flags. */ + /*================================================================*/ + switch( dwRequestFlags ) + { + case GL_ALPHA: + dwRequestFlags = DDPF_ALPHA; + DPF(( DBG_TXT_WARN, "GL_ALPHA not supported!)" )); + return FALSE; + + case GL_INTENSITY: + case GL_LUMINANCE: + DPF(( DBG_TXT_WARN, "GL_INTENSITY/GL_LUMINANCE not supported!)" )); + dwRequestFlags = DDPF_LUMINANCE; + return FALSE; + + case GL_LUMINANCE_ALPHA: + DPF(( DBG_TXT_WARN, "GL_LUMINANCE_ALPHA not supported!)" )); + dwRequestFlags = DDPF_LUMINANCE | DDPF_ALPHAPIXELS; + return FALSE; + + case GL_RGB: + DPF(( DBG_TXT_INFO, "Texture -> GL_RGB" )); + dwRequestFlags = DDPF_RGB; + break; + + case GL_RGBA: + DPF(( DBG_TXT_INFO, "Texture -> GL_RGBA" )); + dwRequestFlags = DDPF_RGB | DDPF_ALPHAPIXELS; + break; + } + + /*==============================*/ + /* Populate the texture object. */ + /*==============================*/ + pTMObj->dwName = dwName; + pTMObj->lpD3DDevice = pHAL->lpD3DDevice; + pTMObj->dwFlags = dwRequestFlags; + if ( pHAL->D3DHWDevDesc.dpcTriCaps.dwTextureCaps & D3DPTEXTURECAPS_SQUAREONLY ) + { + DPF(( DBG_TXT_INFO, "Convert to Square..." )); + pTMObj->dwSHeight = dwHeight; + pTMObj->dwSWidth = dwWidth; + + /* Shrink non-square textures. */ + pTMObj->dwVHeight = (dwHeight > dwWidth) ? dwWidth : dwHeight; + pTMObj->dwVWidth = (dwHeight > dwWidth) ? dwWidth : dwHeight; + } + else + { + pTMObj->dwSHeight = dwHeight; + pTMObj->dwSWidth = dwWidth; + pTMObj->dwVHeight = dwHeight; + pTMObj->dwVWidth = dwWidth; + } + + /*========================*/ + /* Create SYSTEM surface. */ + /*========================*/ + + /* Request a surface in system memory. */ + memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); + ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); + ddsd2.dwWidth = pTMObj->dwSWidth; + ddsd2.dwHeight = pTMObj->dwSHeight; + ddsd2.dwFlags = DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT | DDSD_PIXELFORMAT; + ddsd2.ddsCaps.dwCaps = DDSCAPS_TEXTURE | DDSCAPS_SYSTEMMEMORY; + ddsd2.ddsCaps.dwCaps2 = 0L; + memset( &ddsd2.ddpfPixelFormat, 0, sizeof(DDPIXELFORMAT) ); + ddsd2.ddpfPixelFormat.dwSize = sizeof( DDPIXELFORMAT ); + ddsd2.ddpfPixelFormat.dwFlags = dwRequestFlags; + rc = pHAL->lpD3DDevice->EnumTextureFormats( EnumPFHook, &ddsd2.ddpfPixelFormat ); + if ( FAILED(rc) ) + { + RIP( pHAL, "EnumerTextureFormats (SYSTEM)->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Create the surface using the enumerated pixelformat. */ + rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pTMObj->lpDDS_System, NULL ); + if ( FAILED(rc) ) + { + RIP( pHAL, "CreateSurface (TEXTURE/SYSTEM)->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Solve the pixel mapping info using the surface pixelformat. */ + Solve8BitChannelPixelFormat( &ddsd2.ddpfPixelFormat, &pTMObj->pixel ); + + /*===================================================================*/ + /* Fill the texture using the PixelInfo structure to do the mapping. */ + /*===================================================================*/ + UpdateTexture( pTMObj, FALSE, NULL, (UCHAR *)pPixels ); + + /*=======================*/ + /* Create VIDEO surface. */ + /*=======================*/ + rc = LoadTextureInVideo( pHAL, pTMObj ); + if ( rc == FALSE ) + return FALSE; + + /* Make this the current texture. */ + rc = pHAL->lpD3DDevice->SetTexture( 0, pTMObj->lpD3DTexture2 ); + if ( FAILED(rc) ) + { + DPF(( DBG_TXT_WARN, "Failed SetTexture() (%s)", ErrorStringD3D(rc) )); + pHAL->lpD3DDevice->SetTexture( 0, NULL ); + return FALSE; + } + + return TRUE; +} +/*===========================================================================*/ +/* This function will handle the creation and destruction of the texture */ +/* surfaces on the card. Using the dw'V'Width/Height dimensions the call */ +/* try and create the texture on the card and keep using FreeTextureMemory */ +/* until the surace can be created. Once the surface is created we get the */ +/* interface that we will use to make it the current texture. I didn't put */ +/* the code to make the texture current in this function as BIND needs to */ +/* use the same code and this function doesn't always get called when we do a*/ +/* bind. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +static BOOL LoadTextureInVideo( PMESAD3DHAL pHAL, PTM_OBJECT pTMObj ) +{ + DDSURFACEDESC2 ddsd2; + HRESULT rc; + + DPF(( DBG_FUNC, "LoadTextureInVideo();" )); + + /* Kill the interface if we have one no matter what. */ + if ( pTMObj->lpD3DTexture2 ) + { + DPF(( DBG_TXT_INFO, "Release Texture2" )); + pTMObj->lpD3DTexture2->Release(); + pTMObj->lpD3DTexture2 = NULL; + } + + /* Kill the Video surface. TODO: need some reuse system... */ + if ( pTMObj->lpDDS_Video ) + { + DPF(( DBG_TXT_INFO, "Release texture (VID)" )); + DX_RESTORE( pTMObj->lpDDS_Video ); + pTMObj->lpDDS_Video->Release(); + pTMObj->lpDDS_Video = NULL; + } + + /* Request a surface in Video memory. */ + memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); + ddsd2.dwSize = sizeof( DDSURFACEDESC2 ); + ddsd2.dwWidth = pTMObj->dwVWidth; + ddsd2.dwHeight = pTMObj->dwVHeight; + ddsd2.dwFlags = DDSD_CAPS | DDSD_WIDTH | DDSD_HEIGHT | DDSD_PIXELFORMAT; + ddsd2.ddsCaps.dwCaps = DDSCAPS_TEXTURE | DDSCAPS_VIDEOMEMORY; + ddsd2.ddsCaps.dwCaps2 = 0L; + memset( &ddsd2.ddpfPixelFormat, 0, sizeof(DDPIXELFORMAT) ); + ddsd2.ddpfPixelFormat.dwSize = sizeof( DDPIXELFORMAT ); + ddsd2.ddpfPixelFormat.dwFlags = pTMObj->dwFlags; + rc = pHAL->lpD3DDevice->EnumTextureFormats( EnumPFHook, &ddsd2.ddpfPixelFormat ); + if ( FAILED(rc) ) + { + RIP( pHAL, "EnumerTextureFormats ->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Make sure we lock so we don't nuke this texture trying to free memory for it. */ + pTMObj->bLock = TRUE; + + /* Start a loop that will free all textures until we have created the texture */ + /* surface or we can't free up more memory. */ + do + { + /* Try to create the texture surface. */ + rc = pHAL->lpDD4->CreateSurface( &ddsd2, &pTMObj->lpDDS_Video, NULL ); + if ( !FAILED(rc) ) + break; + + DPF(( DBG_TXT_INFO, "Free Texture Memory" )); + + /* DestroyTexture will return TRUE if a surface was freed. */ + } while( FreeTextureMemory(pHAL,NULL) ); + + /* Make sure we unlock or we won't be able to nuke the TMO later. */ + pTMObj->bLock = FALSE; + + /* Did we create a valid texture surface? */ + if ( FAILED(rc) ) + { + DPF(( DBG_TXT_WARN, "Failed to load texture" )); + pHAL->lpD3DDevice->SetTexture( 0, NULL ); + return FALSE; + } + + DX_RESTORE( pTMObj->lpDDS_System ); + DX_RESTORE( pTMObj->lpDDS_Video ); + + DPF(( DBG_TXT_INFO, "Texture Blt SYSTEM -> VID" )); + + /* Now blt the texture in system memory to the card. */ + rc = pTMObj->lpDDS_Video->Blt( NULL, pTMObj->lpDDS_System, NULL, DDBLT_WAIT, NULL ); + if ( FAILED(rc) ) + { + RIP( pHAL, "Blt (TEXTURE) ->", ErrorStringD3D(rc) ); + return FALSE; + } + + /* Get the Texture interface that is used to render with. */ + pTMObj->lpDDS_Video->QueryInterface( IID_IDirect3DTexture2, (void **)&pTMObj->lpD3DTexture2 ); + if ( pTMObj->lpD3DTexture2 == NULL ) + { + DPF(( DBG_TXT_WARN, "Failed QueryTextureInterface" )); + pHAL->lpD3DDevice->SetTexture( 0, NULL ); + return FALSE; + } + + return TRUE; +} +/*===========================================================================*/ +/* If this function gets a texture object struc then we will try and free */ +/* it. If we get a NULL then we will search from the bottom up and free one */ +/* VMEM surface. I can only free when the surface isn't locked and of course*/ +/* there must be a VMEM surface. We never free SMEM surfaces as that isn't */ +/* the point. */ +/* TODO: should have a pointer to the bottom of the stack really. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static BOOL FreeTextureMemory( PMESAD3DHAL pHAL, PTM_OBJECT pTMObject ) +{ + PTM_OBJECT pCurrent; + BOOL bFreed = FALSE; + + DPF(( DBG_FUNC, "FreeTextureMemory();" )); + DPF(( DBG_TXT_WARN, "FREE TEXTURE!" )); + + /* Just to be safe. */ + if ( !pHAL || !pHAL->pTMList ) + { + DPF(( DBG_TXT_WARN, "FreeTextureMemory() -> NULL pHAL/pHAL->pTMList" )); + return FALSE; + } + + /* Free the last texture in the list. */ + if ( pTMObject == NULL ) + { + DPF(( DBG_TXT_INFO, "Free Last texture in cache" )); + + /* Find the last texture object. */ + for( pCurrent = pHAL->pTMList; pCurrent->next; pCurrent = pCurrent->next ); + + /* Now backup until we find a texture on the card. */ + while( pCurrent && (pCurrent->lpDDS_Video == NULL) && (pCurrent->bLock == FALSE) ) + pCurrent = pCurrent->prev; + + /* Didn't find anything. */ + if ( pCurrent == NULL ) + { + DPF(( DBG_TXT_INFO, "No texture memory freed" )); + return FALSE; + } + } + else + { + /* See if we can find this texture object. */ + for( pCurrent = pHAL->pTMList; pCurrent && (pCurrent != pTMObject); pCurrent = pCurrent->next ); + + /* Didn't find anything. */ + if ( pCurrent == NULL ) + { + DPF(( DBG_TXT_INFO, "Requested texture to be freed NOT FOUND" )); + return FALSE; + } + } + + /* Can't free this baby. */ + if ( pCurrent->bLock == TRUE ) + { + DPF(( DBG_TXT_WARN, "Requested texture LOCKED" )); + return FALSE; + } + + /* Free the texture memory. */ + if ( pCurrent->lpD3DTexture2 ) + { + DPF(( DBG_TXT_INFO, "Release Texture2" )); + pCurrent->lpD3DTexture2->Release(); + pCurrent->lpD3DTexture2 = NULL; + bFreed = TRUE; + } + if ( pCurrent->lpDDS_Video ) + { + DPF(( DBG_TXT_INFO, "Release texture (VID):" )); + DPF(( DBG_TXT_INFO, "dwName: %d", pCurrent->dwName )); + DPF(( DBG_TXT_INFO, "cx: %d, cy: %d", pCurrent->dwVWidth, pCurrent->dwVHeight )); + pCurrent->lpDDS_Video->Release(); + pCurrent->lpDDS_Video = NULL; + bFreed = TRUE; + } + + return bFreed; +} +/*===========================================================================*/ +/* This function searches the linked list of texture objects in the supplied*/ +/* D3Dwrapper structure. If it finds a match it will free it and pull it out*/ +/* of the linked list. The function works on the bases of a matching pointer*/ +/* to the object (not matching content). */ +/* If the function gets passed a NULL then we want to free the last texture */ +/* object in the list. Used in a loop to destory all. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +static BOOL DestroyTextureObject( PMESAD3DHAL pHAL, PTM_OBJECT pTMObject ) +{ + PTM_OBJECT pCurrent; + + DPF(( DBG_FUNC, "DestoryTextureObject();" )); + + /* Just to be safe. */ + if ( !pHAL || !pHAL->pTMList ) + { + DPF(( DBG_TXT_WARN, "DestroyTextureObject() -> NULL pHAL/pHAL->pTMList" )); + return FALSE; + } + + /* Free the last texture in the list. */ + if ( pTMObject == NULL ) + { + /* Find the last texture object. */ + for( pCurrent = pHAL->pTMList; pCurrent->next; pCurrent = pCurrent->next ); + } + else + { + /* See if we can find this texture object. */ + for( pCurrent = pHAL->pTMList; pCurrent && (pCurrent != pTMObject); pCurrent = pCurrent->next ); + + /* Didn't find anything. */ + if ( pCurrent == NULL ) + { + DPF(( DBG_TXT_WARN, "No textures to be freed" )); + return FALSE; + } + } + + /* Can't free this baby. */ + if ( pCurrent->bLock == TRUE ) + { + DPF(( DBG_TXT_WARN, "Requested texture to be freed LOCKED" )); + return FALSE; + } + + /* Free the texture memory. */ + if ( pCurrent->lpD3DTexture2 ) + { + DPF(( DBG_TXT_INFO, "Release Texture2" )); + pCurrent->lpD3DTexture2->Release(); + pCurrent->lpD3DTexture2 = NULL; + } + if ( pCurrent->lpDDS_Video ) + { + DPF(( DBG_TXT_INFO, "Release texture (VID):" )); + pCurrent->lpDDS_Video->Release(); + pCurrent->lpDDS_Video = NULL; + } + if ( pCurrent->lpDDS_System ) + { + DPF(( DBG_TXT_INFO, "Release texture (SYS):" )); + pCurrent->lpDDS_System->Release(); + pCurrent->lpDDS_System = NULL; + } + + /* Pull this texture out of the list. */ + if ( pCurrent == pHAL->pTMList ) + pHAL->pTMList = NULL; + if ( pCurrent->prev ) + (pCurrent->prev)->next = pCurrent->next; + if ( pCurrent->next ) + (pCurrent->next)->prev = pCurrent->prev; + FREE( pCurrent ); + + return TRUE; +} +/*===========================================================================*/ +/* This function is the callback function that gets called when we are doing*/ +/* an enumeration of the texture formats supported by this device. The choice*/ +/* is made by checking to see if we have a match with the supplied D3D pixel-*/ +/* format. So the enumeration has to pass a desired D3D PF as the user var. */ +/*===========================================================================*/ +/* RETURN: D3DENUMRET_OK, D3DENUMRET_CANCEL. */ +/*===========================================================================*/ +static void UpdateTexture( PTM_OBJECT pTMObj, BOOL bVideo, RECT *pRect, UCHAR *pixels ) +{ + LPDIRECTDRAWSURFACE4 lpDDS; + DDSURFACEDESC2 ddsd2; + DWORD srcPitch, + dwHeight, + dwWidth, + dwCol, + dwColor; + UCHAR *pSrc, + *pSrcRow, + *pDest, + *pDestRow; + int rc; + + // TODO: Do I need to pass the h/w when its in the object! + DPF(( DBG_FUNC, "UpdateTexture();" )); + + /* Get the surface pointer we are looking for. */ + lpDDS = (bVideo) ? pTMObj->lpDDS_Video : pTMObj->lpDDS_System; + + /*===================================================================*/ + /* Fill the texture using the PixelInfo structure to do the mapping. */ + /*===================================================================*/ + + /* Get the surface pointer. */ + memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); + ddsd2.dwSize = sizeof(DDSURFACEDESC2); + rc = lpDDS->Lock( NULL, &ddsd2, DDLOCK_WAIT, NULL ); + if ( FAILED(rc) ) + { + RIP( NULL, "Lock (TEXTURE/SYSTEM)->", ErrorStringD3D(rc) ); + return; + } + + /* For now we are only updating the system surface so use its dimensions. */ + dwWidth = (bVideo) ? pTMObj->dwVWidth : pTMObj->dwSWidth; + dwHeight = (bVideo) ? pTMObj->dwVHeight : pTMObj->dwSHeight; + + /* If we are updating the whole surface then the pDest/pSrc will */ + /* always be the same. */ + if ( pRect == NULL ) + { + pDest = (UCHAR *)ddsd2.lpSurface; + pSrc = pixels; + } + + /* Fill the texture surface based on the pixelformat flags. */ + if ( pTMObj->dwFlags == (DDPF_RGB | DDPF_ALPHAPIXELS) ) + { + srcPitch = dwWidth * 4; + if ( pRect ) + { + pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); + pSrc = pixels + (pRect->top * dwWidth * 4) + (pRect->left * 4); + dwHeight = (pRect->bottom - pRect->top); + dwWidth = (pRect->right - pRect->left); + } + + for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) + { + for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) + { + dwColor = ( ((DWORD)(*(pSrc ) * pTMObj->pixel.rScale)) << pTMObj->pixel.rShift ); + dwColor |= ( ((DWORD)(*(pSrc+1) * pTMObj->pixel.gScale)) << pTMObj->pixel.gShift ); + dwColor |= ( ((DWORD)(*(pSrc+2) * pTMObj->pixel.bScale)) << pTMObj->pixel.bShift ); + if ( pTMObj->pixel.aScale == -1.0 ) + dwColor |= ( (*(pSrc+3) & 0x80) ? (1 << pTMObj->pixel.aShift) : 0 ); + else + dwColor |= ( ((DWORD)(*(pSrc+3) * pTMObj->pixel.aScale)) << pTMObj->pixel.aShift ); + memcpy( pDest, &dwColor, pTMObj->pixel.cb ); + pDest += pTMObj->pixel.cb; + pSrc += 4; + } + } + } + else if ( pTMObj->dwFlags == DDPF_RGB ) + { + srcPitch = dwWidth * 3; + if ( pRect ) + { + pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); + pSrc = pixels + (pRect->top * dwWidth * 3) + (pRect->left * 3); + dwHeight = (pRect->bottom - pRect->top); + dwWidth = (pRect->right - pRect->left); + } + + for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) + { + for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) + { + dwColor = ( ((DWORD)(*(pSrc ) * pTMObj->pixel.rScale)) << pTMObj->pixel.rShift ); + dwColor |= ( ((DWORD)(*(pSrc+1) * pTMObj->pixel.gScale)) << pTMObj->pixel.gShift ); + dwColor |= ( ((DWORD)(*(pSrc+2) * pTMObj->pixel.bScale)) << pTMObj->pixel.bShift ); + memcpy( pDest, &dwColor, pTMObj->pixel.cb ); + pDest += pTMObj->pixel.cb; + pSrc += 3; + } + } + } + else if ( pTMObj->dwFlags == (DDPF_LUMINANCE | DDPF_ALPHAPIXELS) ) + { + srcPitch = dwWidth * 2; + if ( pRect ) + { + pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); + pSrc = pixels + (pRect->top * dwWidth * 2) + (pRect->left * 2); + dwHeight = (pRect->bottom - pRect->top); + dwWidth = (pRect->right - pRect->left); + } + + for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) + { + for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) + { + dwColor = ( ((DWORD)(*(pSrc ) * pTMObj->pixel.rScale)) << pTMObj->pixel.rShift ); + if ( pTMObj->pixel.aScale == -1.0 ) + dwColor |= ( (*(pSrc+1) & 0x80) ? (1 << pTMObj->pixel.aShift) : 0 ); + else + dwColor |= ( ((DWORD)(*(pSrc+1) * pTMObj->pixel.aScale)) << pTMObj->pixel.aShift ); + memcpy( pDest, &dwColor, pTMObj->pixel.cb ); + pDest += pTMObj->pixel.cb; + pSrc += 2; + } + } + } + else if ( pTMObj->dwFlags == DDPF_LUMINANCE ) + { + srcPitch = dwWidth; + if ( pRect ) + { + pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); + pSrc = pixels + (pRect->top * dwWidth) + (pRect->left); + dwHeight = (pRect->bottom - pRect->top); + dwWidth = (pRect->right - pRect->left); + } + + for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) + { + for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) + { + dwColor = ( ((DWORD)(*pSrc * pTMObj->pixel.rScale)) << pTMObj->pixel.rShift ); + memcpy( pDest, &dwColor, pTMObj->pixel.cb ); + pDest += pTMObj->pixel.cb; + pSrc++; + } + } + } + else if ( pTMObj->dwFlags == DDPF_ALPHAPIXELS ) + { + srcPitch = dwWidth; + if ( pRect ) + { + pDest = ((UCHAR *)ddsd2.lpSurface) + (pRect->top * ddsd2.lPitch) + (pRect->left * pTMObj->pixel.cb); + pSrc = pixels + (pRect->top * dwWidth) + (pRect->left); + dwHeight = (pRect->bottom - pRect->top); + dwWidth = (pRect->right - pRect->left); + } + + for( pDestRow = pDest, pSrcRow = pSrc; dwHeight > 0; dwHeight--, pDestRow += ddsd2.lPitch, pSrcRow += srcPitch ) + { + for( dwCol = 0, pDest = pDestRow, pSrc = pSrcRow; dwCol < dwWidth; dwCol++ ) + { + if ( pTMObj->pixel.aScale == -1.0 ) + dwColor = ( (*pSrc & 0x80) ? (1 << pTMObj->pixel.aShift) : 0 ); + else + dwColor = ( ((DWORD)(*pSrc * pTMObj->pixel.aScale)) << pTMObj->pixel.aShift ); + memcpy( pDest, &dwColor, pTMObj->pixel.cb ); + pDest += pTMObj->pixel.cb; + pSrc++; + } + } + } + + /* Unlock the surface. */ + rc = lpDDS->Unlock( NULL ); + if ( FAILED(rc) ) + { + RIP( NULL, "Unlock (TEXTURE/SYSTEM)->", ErrorStringD3D(rc) ); + } +} +/*===========================================================================*/ +/* This function is the callback function that gets called when we are doing*/ +/* an enumeration of the texture formats supported by this device. The choice*/ +/* is made by checking to see if we have a match with the supplied D3D pixel-*/ +/* format. So the enumeration has to pass a desired D3D PF as the user var. */ +/*===========================================================================*/ +/* RETURN: D3DENUMRET_OK, D3DENUMRET_CANCEL. */ +/*===========================================================================*/ +HRESULT CALLBACK EnumPFHook( LPDDPIXELFORMAT lpDDPixFmt, LPVOID lpContext ) +{ + LPDDPIXELFORMAT lpDDPixFmtRequest = (LPDDPIXELFORMAT)lpContext; + PIXELINFO pixel; + + DPF(( DBG_FUNC, "EnumPFHook();" )); + + if ( lpDDPixFmt->dwFlags == lpDDPixFmtRequest->dwFlags ) + { + /* Are we looking for an alpha channel? */ + if ( lpDDPixFmtRequest->dwFlags & DDPF_ALPHAPIXELS ) + { + /* Try for something that has more then 1bits of Alpha. */ + Solve8BitChannelPixelFormat( lpDDPixFmt, &pixel ); + if ( pixel.aScale == -1.0 ) + { + /* Save this format no matter what as its a match of sorts. */ + memcpy( lpDDPixFmtRequest, lpDDPixFmt, sizeof(DDPIXELFORMAT) ); + return D3DENUMRET_OK; + } + } + + /* Save this format as its a good match. */ + memcpy( lpDDPixFmtRequest, lpDDPixFmt, sizeof(DDPIXELFORMAT) ); + + /* We are happy at this point so lets leave. */ + return D3DENUMRET_CANCEL; + } + + return D3DENUMRET_OK; +} + + diff --git a/src/mesa/drivers/d3d/D3DTextureMgr.h b/src/mesa/drivers/d3d/D3DTextureMgr.h index f4a4154917..8a64ceaaec 100644 --- a/src/mesa/drivers/d3d/D3DTextureMgr.h +++ b/src/mesa/drivers/d3d/D3DTextureMgr.h @@ -1,62 +1,62 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#ifndef _TEXTURE_MGR_INC -#define _TEXTURE_MGR_INC - -/*===========================================================================*/ -/* Includes. */ -/*===========================================================================*/ -#include -#include -#include -#include -#include -#include "GL/gl.h" -/*========================================================================*/ -/* Defines. */ -/*========================================================================*/ -/*========================================================================*/ -/* Type defines. */ -/*========================================================================*/ -typedef struct _local_texture_object -{ - DWORD dwName, - dwPriority, - dwFlags, - dwSWidth, - dwSHeight, - dwVWidth, - dwVHeight; - BOOL bLock, - bDirty; /* I only update VID on SubImage calls so the system */ - /* texture can get invalid. */ - - LPDIRECT3DDEVICE3 lpD3DDevice; /* If the device changes we must get new handles... */ - LPDIRECTDRAWSURFACE4 lpDDS_System, - lpDDS_Video; - LPDIRECT3DTEXTURE2 lpD3DTexture2; - - PIXELINFO pixel; - - struct _local_texture_object *next; - struct _local_texture_object *prev; - -} TM_OBJECT, *PTM_OBJECT; -/*========================================================================*/ -/* Function prototypes. */ -/*========================================================================*/ -void APIENTRY InitTMD3D( void *pVoid ); -void APIENTRY TermTMD3D( void *pVoid ); -/*========================================================================*/ -/* Global variables declaration. */ -/*========================================================================*/ - -#endif +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#ifndef _TEXTURE_MGR_INC +#define _TEXTURE_MGR_INC + +/*===========================================================================*/ +/* Includes. */ +/*===========================================================================*/ +#include +#include +#include +#include +#include +#include "GL/gl.h" +/*========================================================================*/ +/* Defines. */ +/*========================================================================*/ +/*========================================================================*/ +/* Type defines. */ +/*========================================================================*/ +typedef struct _local_texture_object +{ + DWORD dwName, + dwPriority, + dwFlags, + dwSWidth, + dwSHeight, + dwVWidth, + dwVHeight; + BOOL bLock, + bDirty; /* I only update VID on SubImage calls so the system */ + /* texture can get invalid. */ + + LPDIRECT3DDEVICE3 lpD3DDevice; /* If the device changes we must get new handles... */ + LPDIRECTDRAWSURFACE4 lpDDS_System, + lpDDS_Video; + LPDIRECT3DTEXTURE2 lpD3DTexture2; + + PIXELINFO pixel; + + struct _local_texture_object *next; + struct _local_texture_object *prev; + +} TM_OBJECT, *PTM_OBJECT; +/*========================================================================*/ +/* Function prototypes. */ +/*========================================================================*/ +void APIENTRY InitTMD3D( void *pVoid ); +void APIENTRY TermTMD3D( void *pVoid ); +/*========================================================================*/ +/* Global variables declaration. */ +/*========================================================================*/ + +#endif diff --git a/src/mesa/drivers/d3d/D3DUTILS.CPP b/src/mesa/drivers/d3d/D3DUTILS.CPP index c13d89cd71..02664f96e4 100644 --- a/src/mesa/drivers/d3d/D3DUTILS.CPP +++ b/src/mesa/drivers/d3d/D3DUTILS.CPP @@ -1,638 +1,638 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#include "D3DHAL.h" -/*===========================================================================*/ -/* Local only functions. */ -/*===========================================================================*/ -static int CountTrailingZeros( DWORD dwMask ); -/*===========================================================================*/ -/* This function is used to get the pointer to the surface and the pitch for*/ -/* the scanline rendering functions. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" DDSURFACEDESC2 *LockHAL( PMESAD3DSHARED pShared, BOOL bBack ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - static DDSURFACEDESC2 ddsd2; - HRESULT rc; - - DPF(( DBG_FUNC, "LockHAL();" )); - - /* Set the request structure up first. */ - memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); - ddsd2.dwSize = sizeof(DDSURFACEDESC2); - - /* Make sure we have enough info. */ - if ( pHAL ) - { - rc = pHAL->lpDDSRender->Lock( NULL, &ddsd2, DDLOCK_WAIT, NULL ); - if ( FAILED(rc) ) - { - RIP( pHAL, "Lock (RENDER) ->", ErrorStringD3D(rc) ); - } - } - - return &ddsd2; -} -/*===========================================================================*/ -/* This is just a simple wrapper. I probably don't need to do any error */ -/* checking as the Lock must have worked inorder to get here... */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void UnlockHAL( PMESAD3DSHARED pShared, BOOL bBack ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - HRESULT rc; - - DPF(( DBG_FUNC, "UnlockHAL();" )); - - /* Make sure we have enough info. */ - if ( pHAL ) - { - rc = pHAL->lpDDSRender->Unlock( NULL ); - if ( FAILED(rc) ) - { - RIP( pHAL, "Unlock (RENDER) ->", ErrorStringD3D(rc) ); - } - } -} -/*===========================================================================*/ -/* This function will track the main/Primary window that will be used as the*/ -/* target for the Blt in SwapBuffers. As a side effect the call will check */ -/* to see if the primary surface is the same size and position as the screen.*/ -/* If they are the same size we will call it fullscreen... */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -extern "C" void UpdateScreenPosHAL( PMESAD3DSHARED pShared ) -{ - PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; - POINT pt; - DWORD dwWidth, dwHeight; - - DPF(( DBG_FUNC, "UpdateScreenPosHAL();" )); - - /* Make sure we have enough info. */ - if ( pHAL != NULL ) - { - /* Update the windows screen position. */ - GetClientRect( pShared->hwnd, &pShared->rectW ); - pt.x = pt.y = 0; - ClientToScreen( pShared->hwnd, &pt ); - OffsetRect( &pShared->rectW, pt.x, pt.y); - - /* Compare the primary to the screen. */ - dwWidth = GetSystemMetrics( SM_CXSCREEN ); - dwHeight = GetSystemMetrics( SM_CYSCREEN ); - if ( (pShared->rectW.left > 0) || (pShared->rectW.top > 0) || - (pShared->rectW.right > dwWidth) || (pShared->rectW.bottom > dwHeight) ) - pShared->bWindow = TRUE; - else - pShared->bWindow = FALSE; - } -} -/*===========================================================================*/ -/* This function will fill in the pixel info structure defined in D3Dshared.*/ -/* Basicly it will take a DirectDraw pixelformat structure and make scaling */ -/* values that will convert from 8bit channels to whatever the supplied ddpf */ -/* uses. Also we will generate shift values that will be used to get move */ -/* each component of the pixel into place. */ -/* I have now added a special case for a 1bit alpha channel. If I find a 1b*/ -/* alpha then I will set the scale to -1.0 which should be unique. Later I */ -/* can check the alpha scale value too see if its -1.0 and thus handle it. I*/ -/* was finding that the case was not working tom my advantage so this is my */ -/* HACK for the day. As a TODO I should work on this... */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void Solve8BitChannelPixelFormat( DDPIXELFORMAT *pddpf, PPIXELINFO pPixel ) -{ - DPF(( DBG_FUNC, "Solve8BitChannelPixelFromat();" )); - - memset( pPixel, 0, sizeof(PPIXELINFO) ); - - /* Check too see if the color space is valid in the PF. */ - if ( pddpf->dwFlags & DDPF_RGB ) - { - /* Solve the red stuff. */ - pPixel->dwRMask = pddpf->dwRBitMask; - pPixel->rShift = CountTrailingZeros( pPixel->dwRMask ); - pPixel->rScale = (float)0.00392156 * (float)(pPixel->dwRMask >> pPixel->rShift); - - /* Solve the green thingy's. */ - pPixel->dwGMask = pddpf->dwGBitMask; - pPixel->gShift = CountTrailingZeros( pPixel->dwGMask ); - pPixel->gScale = (float)0.00392156 * (float)(pPixel->dwGMask >> pPixel->gShift); - - /* Solve the blues. */ - pPixel->dwBMask = pddpf->dwBBitMask; - pPixel->bShift = CountTrailingZeros( pddpf->dwBBitMask ); - pPixel->bScale = (float)0.00392156 * (float)(pddpf->dwBBitMask >> pPixel->bShift); - } - - /* Do the alpha channel if there is one. */ - if ( pddpf->dwFlags & DDPF_ALPHAPIXELS ) - { - pPixel->dwAMask = pddpf->dwRGBAlphaBitMask; - pPixel->aShift = CountTrailingZeros( pPixel->dwAMask ); - - /* Special case a 1bit alpha. */ - if ( (pPixel->dwAMask >> pPixel->aShift) == 1 ) - pPixel->aScale = -1.0; - else - pPixel->aScale = (float)0.00392156 * (float)(pPixel->dwAMask >> pPixel->aShift); - } - - /* Get the size of the pixel in bytes. Should work as dwRGBBitCount is in a union. */ - pPixel->cb = pddpf->dwRGBBitCount / 8; -} -/*===========================================================================*/ -/* See RETURN :) */ -/*===========================================================================*/ -/* RETURN: number of contiguous zeros starting from the right. */ -/*===========================================================================*/ -static int CountTrailingZeros( DWORD dwMask ) -{ - DWORD Mask; - - if ( dwMask == 0 ) - return 32; - - /* Can't take credit for this one! */ - Mask = dwMask & -(int)dwMask; - return ((Mask & 0xFFFF0000)!=0) << 4 - | ((Mask & 0xFF00FF00)!=0) << 3 - | ((Mask & 0xF0F0F0F0)!=0) << 2 - | ((Mask & 0xCCCCCCCC)!=0) << 1 - | ((Mask & 0xAAAAAAAA)!=0); -} -/*===========================================================================*/ -/* This function will convert the DDraw error code to its macro string. The*/ -/* returned pointer is static so you need not worry about memory managemnet */ -/* but the error message gets written over from call to call... */ -/*===========================================================================*/ -/* RETURN: pointer to the single static buffer that hold the error message. */ -/*===========================================================================*/ -char *ErrorStringD3D( HRESULT hr ) -{ - static char errorString[128]; - - switch( hr ) - { - case DDERR_ALREADYINITIALIZED: - strcpy( errorString, "DDERR_ALREADYINITIALIZED" ); - break; - - case DDERR_CANNOTATTACHSURFACE: - strcpy( errorString, "DDERR_CANNOTATTACHSURFACE" ); - break; - - case DDERR_CANNOTDETACHSURFACE: - strcpy( errorString, "DDERR_CANNOTDETACHSURFACE" ); - break; - - case DDERR_CURRENTLYNOTAVAIL: - strcpy( errorString, "DDERR_CURRENTLYNOTAVAIL" ); - break; - - case DDERR_EXCEPTION: - strcpy( errorString, "DDERR_EXCEPTION" ); - break; - - case DDERR_GENERIC: - strcpy( errorString, "DDERR_GENERIC" ); - break; - - case DDERR_HEIGHTALIGN: - strcpy( errorString, "DDERR_HEIGHTALIGN" ); - break; - - case DDERR_INCOMPATIBLEPRIMARY: - strcpy( errorString, "DDERR_INCOMPATIBLEPRIMARY" ); - break; - - case DDERR_INVALIDCAPS: - strcpy( errorString, "DDERR_INVALIDCAPS" ); - break; - - case DDERR_INVALIDCLIPLIST: - strcpy( errorString, "DDERR_INVALIDCLIPLIST" ); - break; - - case DDERR_INVALIDMODE: - strcpy( errorString, "DDERR_INVALIDMODE" ); - break; - - case DDERR_INVALIDOBJECT: - strcpy( errorString, "DDERR_INVALIDOBJECT" ); - break; - - case DDERR_INVALIDPARAMS: - strcpy( errorString, "DDERR_INVALIDPARAMS" ); - break; - - case DDERR_INVALIDPIXELFORMAT: - strcpy( errorString, "DDERR_INVALIDPIXELFORMAT" ); - break; - - case DDERR_INVALIDRECT: - strcpy( errorString, "DDERR_INVALIDRECT" ); - break; - - case DDERR_LOCKEDSURFACES: - strcpy( errorString, "DDERR_LOCKEDSURFACES" ); - break; - - case DDERR_NO3D: - strcpy( errorString, "DDERR_NO3D" ); - break; - - case DDERR_NOALPHAHW: - strcpy( errorString, "DDERR_NOALPHAHW" ); - break; - - case DDERR_NOCLIPLIST: - strcpy( errorString, "DDERR_NOCLIPLIST" ); - break; - - case DDERR_NOCOLORCONVHW: - strcpy( errorString, "DDERR_NOCOLORCONVHW" ); - break; - - case DDERR_NOCOOPERATIVELEVELSET: - strcpy( errorString, "DDERR_NOCOOPERATIVELEVELSET" ); - break; - - case DDERR_NOCOLORKEY: - strcpy( errorString, "DDERR_NOCOLORKEY" ); - break; - - case DDERR_NOCOLORKEYHW: - strcpy( errorString, "DDERR_NOCOLORKEYHW" ); - break; - - case DDERR_NODIRECTDRAWSUPPORT: - strcpy( errorString, "DDERR_NODIRECTDRAWSUPPORT" ); - break; - - case DDERR_NOEXCLUSIVEMODE: - strcpy( errorString, "DDERR_NOEXCLUSIVEMODE" ); - break; - - case DDERR_NOFLIPHW: - strcpy( errorString, "DDERR_NOFLIPHW" ); - break; - - case DDERR_NOGDI: - strcpy( errorString, "DDERR_NOGDI" ); - break; - - case DDERR_NOMIRRORHW: - strcpy( errorString, "DDERR_NOMIRRORHW" ); - break; - - case DDERR_NOTFOUND: - strcpy( errorString, "DDERR_NOTFOUND" ); - break; - - case DDERR_NOOVERLAYHW: - strcpy( errorString, "DDERR_NOOVERLAYHW" ); - break; - - case DDERR_OVERLAPPINGRECTS: - strcpy( errorString, "DDERR_OVERLAPPINGRECTS" ); - break; - - case DDERR_NORASTEROPHW: - strcpy( errorString, "DDERR_NORASTEROPHW" ); - break; - - case DDERR_NOROTATIONHW: - strcpy( errorString, "DDERR_NOROTATIONHW" ); - break; - - case DDERR_NOSTRETCHHW: - strcpy( errorString, "DDERR_NOSTRETCHHW" ); - break; - - case DDERR_NOT4BITCOLOR: - strcpy( errorString, "DDERR_NOT4BITCOLOR" ); - break; - - case DDERR_NOT4BITCOLORINDEX: - strcpy( errorString, "DDERR_NOT4BITCOLORINDEX" ); - break; - - case DDERR_NOT8BITCOLOR: - strcpy( errorString, "DDERR_NOT8BITCOLOR" ); - break; - - case DDERR_NOTEXTUREHW: - strcpy( errorString, "DDERR_NOTEXTUREHW" ); - break; - - case DDERR_NOVSYNCHW: - strcpy( errorString, "DDERR_NOVSYNCHW" ); - break; - - case DDERR_NOZBUFFERHW: - strcpy( errorString, "DDERR_NOZBUFFERHW" ); - break; - - case DDERR_NOZOVERLAYHW: - strcpy( errorString, "DDERR_NOZOVERLAYHW" ); - break; - - case DDERR_OUTOFCAPS: - strcpy( errorString, "DDERR_OUTOFCAPS" ); - break; - - case DDERR_OUTOFMEMORY: - strcpy( errorString, "DDERR_OUTOFMEMORY" ); - break; - - case DDERR_OUTOFVIDEOMEMORY: - strcpy( errorString, "DDERR_OUTOFVIDEOMEMORY" ); - break; - - case DDERR_OVERLAYCANTCLIP: - strcpy( errorString, "DDERR_OVERLAYCANTCLIP" ); - break; - - case DDERR_OVERLAYCOLORKEYONLYONEACTIVE: - strcpy( errorString, "DDERR_OVERLAYCOLORKEYONLYONEACTIVE" ); - break; - - case DDERR_PALETTEBUSY: - strcpy( errorString, "DDERR_PALETTEBUSY" ); - break; - - case DDERR_COLORKEYNOTSET: - strcpy( errorString, "DDERR_COLORKEYNOTSET" ); - break; - - case DDERR_SURFACEALREADYATTACHED: - strcpy( errorString, "DDERR_SURFACEALREADYATTACHED" ); - break; - - case DDERR_SURFACEALREADYDEPENDENT: - strcpy( errorString, "DDERR_SURFACEALREADYDEPENDENT" ); - break; - - case DDERR_SURFACEBUSY: - strcpy( errorString, "DDERR_SURFACEBUSY" ); - break; - - case DDERR_CANTLOCKSURFACE: - strcpy( errorString, "DDERR_CANTLOCKSURFACE" ); - break; - - case DDERR_SURFACEISOBSCURED: - strcpy( errorString, "DDERR_SURFACEISOBSCURED" ); - break; - - case DDERR_SURFACELOST: - strcpy( errorString, "DDERR_SURFACELOST" ); - break; - - case DDERR_SURFACENOTATTACHED: - strcpy( errorString, "DDERR_SURFACENOTATTACHED" ); - break; - - case DDERR_TOOBIGHEIGHT: - strcpy( errorString, "DDERR_TOOBIGHEIGHT" ); - break; - - case DDERR_TOOBIGSIZE: - strcpy( errorString, "DDERR_TOOBIGSIZE" ); - break; - - case DDERR_TOOBIGWIDTH: - strcpy( errorString, "DDERR_TOOBIGWIDTH" ); - break; - - case DDERR_UNSUPPORTED: - strcpy( errorString, "DDERR_UNSUPPORTED" ); - break; - - case DDERR_UNSUPPORTEDFORMAT: - strcpy( errorString, "DDERR_UNSUPPORTEDFORMAT" ); - break; - - case DDERR_UNSUPPORTEDMASK: - strcpy( errorString, "DDERR_UNSUPPORTEDMASK" ); - break; - - case DDERR_INVALIDSTREAM: - strcpy( errorString, "DDERR_INVALIDSTREAM" ); - break; - - case DDERR_VERTICALBLANKINPROGRESS: - strcpy( errorString, "DDERR_VERTICALBLANKINPROGRESS" ); - break; - - case DDERR_WASSTILLDRAWING: - strcpy( errorString, "DDERR_WASSTILLDRAWING" ); - break; - - case DDERR_XALIGN: - strcpy( errorString, "DDERR_XALIGN" ); - break; - - case DDERR_INVALIDDIRECTDRAWGUID: - strcpy( errorString, "DDERR_INVALIDDIRECTDRAWGUID" ); - break; - - case DDERR_DIRECTDRAWALREADYCREATED: - strcpy( errorString, "DDERR_DIRECTDRAWALREADYCREATED" ); - break; - - case DDERR_NODIRECTDRAWHW: - strcpy( errorString, "DDERR_NODIRECTDRAWHW" ); - break; - - case DDERR_PRIMARYSURFACEALREADYEXISTS: - strcpy( errorString, "DDERR_PRIMARYSURFACEALREADYEXISTS" ); - break; - - case DDERR_NOEMULATION: - strcpy( errorString, "DDERR_NOEMULATION" ); - break; - - case DDERR_REGIONTOOSMALL: - strcpy( errorString, "DDERR_REGIONTOOSMALL" ); - break; - - case DDERR_CLIPPERISUSINGHWND: - strcpy( errorString, "DDERR_CLIPPERISUSINGHWND" ); - break; - - case DDERR_NOCLIPPERATTACHED: - strcpy( errorString, "DDERR_NOCLIPPERATTACHED" ); - break; - - case DDERR_NOHWND: - strcpy( errorString, "DDERR_NOHWND" ); - break; - - case DDERR_HWNDSUBCLASSED: - strcpy( errorString, "DDERR_HWNDSUBCLASSED" ); - break; - - case DDERR_HWNDALREADYSET: - strcpy( errorString, "DDERR_HWNDALREADYSET" ); - break; - - case DDERR_NOPALETTEATTACHED: - strcpy( errorString, "DDERR_NOPALETTEATTACHED" ); - break; - - case DDERR_NOPALETTEHW: - strcpy( errorString, "DDERR_NOPALETTEHW" ); - break; - - case DDERR_BLTFASTCANTCLIP: - strcpy( errorString, "DDERR_BLTFASTCANTCLIP" ); - break; - - case DDERR_NOBLTHW: - strcpy( errorString, "DDERR_NOBLTHW" ); - break; - - case DDERR_NODDROPSHW: - strcpy( errorString, "DDERR_NODDROPSHW" ); - break; - - case DDERR_OVERLAYNOTVISIBLE: - strcpy( errorString, "DDERR_OVERLAYNOTVISIBLE" ); - break; - - case DDERR_NOOVERLAYDEST: - strcpy( errorString, "DDERR_NOOVERLAYDEST" ); - break; - - case DDERR_INVALIDPOSITION: - strcpy( errorString, "DDERR_INVALIDPOSITION" ); - break; - - case DDERR_NOTAOVERLAYSURFACE: - strcpy( errorString, "DDERR_NOTAOVERLAYSURFACE" ); - break; - - case DDERR_EXCLUSIVEMODEALREADYSET: - strcpy( errorString, "DDERR_EXCLUSIVEMODEALREADYSET" ); - break; - - case DDERR_NOTFLIPPABLE: - strcpy( errorString, "DDERR_NOTFLIPPABLE" ); - break; - - case DDERR_CANTDUPLICATE: - strcpy( errorString, "DDERR_CANTDUPLICATE" ); - break; - - case DDERR_NOTLOCKED: - strcpy( errorString, "DDERR_NOTLOCKED" ); - break; - - case DDERR_CANTCREATEDC: - strcpy( errorString, "DDERR_CANTCREATEDC" ); - break; - - case DDERR_NODC: - strcpy( errorString, "DDERR_NODC" ); - break; - - case DDERR_WRONGMODE: - strcpy( errorString, "DDERR_WRONGMODE" ); - break; - - case DDERR_IMPLICITLYCREATED: - strcpy( errorString, "DDERR_IMPLICITLYCREATED" ); - break; - - case DDERR_NOTPALETTIZED: - strcpy( errorString, "DDERR_NOTPALETTIZED" ); - break; - - case DDERR_UNSUPPORTEDMODE: - strcpy( errorString, "DDERR_UNSUPPORTEDMODE" ); - break; - - case DDERR_NOMIPMAPHW: - strcpy( errorString, "DDERR_NOMIPMAPHW" ); - break; - - case DDERR_INVALIDSURFACETYPE: - strcpy( errorString, "DDERR_INVALIDSURFACETYPE" ); - break; - - case DDERR_NOOPTIMIZEHW: - strcpy( errorString, "DDERR_NOOPTIMIZEHW" ); - break; - - case DDERR_NOTLOADED: - strcpy( errorString, "DDERR_NOTLOADED" ); - break; - - case DDERR_NOFOCUSWINDOW: - strcpy( errorString, "DDERR_NOFOCUSWINDOW" ); - break; - - case DDERR_DCALREADYCREATED: - strcpy( errorString, "DDERR_DCALREADYCREATED" ); - break; - - case DDERR_NONONLOCALVIDMEM: - strcpy( errorString, "DDERR_NONONLOCALVIDMEM" ); - break; - - case DDERR_CANTPAGELOCK: - strcpy( errorString, "DDERR_CANTPAGELOCK" ); - break; - - case DDERR_CANTPAGEUNLOCK: - strcpy( errorString, "DDERR_CANTPAGEUNLOCK" ); - break; - - case DDERR_NOTPAGELOCKED: - strcpy( errorString, "DDERR_NOTPAGELOCKED" ); - break; - - case DDERR_MOREDATA: - strcpy( errorString, "DDERR_MOREDATA" ); - break; - - case DDERR_EXPIRED: - strcpy( errorString, "DDERR_EXPIRED" ); - break; - - case DDERR_VIDEONOTACTIVE: - strcpy( errorString, "DDERR_VIDEONOTACTIVE" ); - break; - - case DDERR_DEVICEDOESNTOWNSURFACE: - strcpy( errorString, "DDERR_DEVICEDOESNTOWNSURFACE" ); - break; - - case DDERR_NOTINITIALIZED: - strcpy( errorString, "DDERR_NOTINITIALIZED" ); - break; - - default: - strcpy( errorString, "" ); - break; - } - - return &errorString[0]; -} +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#include "D3DHAL.h" +/*===========================================================================*/ +/* Local only functions. */ +/*===========================================================================*/ +static int CountTrailingZeros( DWORD dwMask ); +/*===========================================================================*/ +/* This function is used to get the pointer to the surface and the pitch for*/ +/* the scanline rendering functions. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" DDSURFACEDESC2 *LockHAL( PMESAD3DSHARED pShared, BOOL bBack ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + static DDSURFACEDESC2 ddsd2; + HRESULT rc; + + DPF(( DBG_FUNC, "LockHAL();" )); + + /* Set the request structure up first. */ + memset( &ddsd2, 0, sizeof(DDSURFACEDESC2) ); + ddsd2.dwSize = sizeof(DDSURFACEDESC2); + + /* Make sure we have enough info. */ + if ( pHAL ) + { + rc = pHAL->lpDDSRender->Lock( NULL, &ddsd2, DDLOCK_WAIT, NULL ); + if ( FAILED(rc) ) + { + RIP( pHAL, "Lock (RENDER) ->", ErrorStringD3D(rc) ); + } + } + + return &ddsd2; +} +/*===========================================================================*/ +/* This is just a simple wrapper. I probably don't need to do any error */ +/* checking as the Lock must have worked inorder to get here... */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void UnlockHAL( PMESAD3DSHARED pShared, BOOL bBack ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + HRESULT rc; + + DPF(( DBG_FUNC, "UnlockHAL();" )); + + /* Make sure we have enough info. */ + if ( pHAL ) + { + rc = pHAL->lpDDSRender->Unlock( NULL ); + if ( FAILED(rc) ) + { + RIP( pHAL, "Unlock (RENDER) ->", ErrorStringD3D(rc) ); + } + } +} +/*===========================================================================*/ +/* This function will track the main/Primary window that will be used as the*/ +/* target for the Blt in SwapBuffers. As a side effect the call will check */ +/* to see if the primary surface is the same size and position as the screen.*/ +/* If they are the same size we will call it fullscreen... */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +extern "C" void UpdateScreenPosHAL( PMESAD3DSHARED pShared ) +{ + PMESAD3DHAL pHAL = (PMESAD3DHAL)pShared; + POINT pt; + DWORD dwWidth, dwHeight; + + DPF(( DBG_FUNC, "UpdateScreenPosHAL();" )); + + /* Make sure we have enough info. */ + if ( pHAL != NULL ) + { + /* Update the windows screen position. */ + GetClientRect( pShared->hwnd, &pShared->rectW ); + pt.x = pt.y = 0; + ClientToScreen( pShared->hwnd, &pt ); + OffsetRect( &pShared->rectW, pt.x, pt.y); + + /* Compare the primary to the screen. */ + dwWidth = GetSystemMetrics( SM_CXSCREEN ); + dwHeight = GetSystemMetrics( SM_CYSCREEN ); + if ( (pShared->rectW.left > 0) || (pShared->rectW.top > 0) || + (pShared->rectW.right > dwWidth) || (pShared->rectW.bottom > dwHeight) ) + pShared->bWindow = TRUE; + else + pShared->bWindow = FALSE; + } +} +/*===========================================================================*/ +/* This function will fill in the pixel info structure defined in D3Dshared.*/ +/* Basicly it will take a DirectDraw pixelformat structure and make scaling */ +/* values that will convert from 8bit channels to whatever the supplied ddpf */ +/* uses. Also we will generate shift values that will be used to get move */ +/* each component of the pixel into place. */ +/* I have now added a special case for a 1bit alpha channel. If I find a 1b*/ +/* alpha then I will set the scale to -1.0 which should be unique. Later I */ +/* can check the alpha scale value too see if its -1.0 and thus handle it. I*/ +/* was finding that the case was not working tom my advantage so this is my */ +/* HACK for the day. As a TODO I should work on this... */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void Solve8BitChannelPixelFormat( DDPIXELFORMAT *pddpf, PPIXELINFO pPixel ) +{ + DPF(( DBG_FUNC, "Solve8BitChannelPixelFromat();" )); + + memset( pPixel, 0, sizeof(PPIXELINFO) ); + + /* Check too see if the color space is valid in the PF. */ + if ( pddpf->dwFlags & DDPF_RGB ) + { + /* Solve the red stuff. */ + pPixel->dwRMask = pddpf->dwRBitMask; + pPixel->rShift = CountTrailingZeros( pPixel->dwRMask ); + pPixel->rScale = (float)0.00392156 * (float)(pPixel->dwRMask >> pPixel->rShift); + + /* Solve the green thingy's. */ + pPixel->dwGMask = pddpf->dwGBitMask; + pPixel->gShift = CountTrailingZeros( pPixel->dwGMask ); + pPixel->gScale = (float)0.00392156 * (float)(pPixel->dwGMask >> pPixel->gShift); + + /* Solve the blues. */ + pPixel->dwBMask = pddpf->dwBBitMask; + pPixel->bShift = CountTrailingZeros( pddpf->dwBBitMask ); + pPixel->bScale = (float)0.00392156 * (float)(pddpf->dwBBitMask >> pPixel->bShift); + } + + /* Do the alpha channel if there is one. */ + if ( pddpf->dwFlags & DDPF_ALPHAPIXELS ) + { + pPixel->dwAMask = pddpf->dwRGBAlphaBitMask; + pPixel->aShift = CountTrailingZeros( pPixel->dwAMask ); + + /* Special case a 1bit alpha. */ + if ( (pPixel->dwAMask >> pPixel->aShift) == 1 ) + pPixel->aScale = -1.0; + else + pPixel->aScale = (float)0.00392156 * (float)(pPixel->dwAMask >> pPixel->aShift); + } + + /* Get the size of the pixel in bytes. Should work as dwRGBBitCount is in a union. */ + pPixel->cb = pddpf->dwRGBBitCount / 8; +} +/*===========================================================================*/ +/* See RETURN :) */ +/*===========================================================================*/ +/* RETURN: number of contiguous zeros starting from the right. */ +/*===========================================================================*/ +static int CountTrailingZeros( DWORD dwMask ) +{ + DWORD Mask; + + if ( dwMask == 0 ) + return 32; + + /* Can't take credit for this one! */ + Mask = dwMask & -(int)dwMask; + return ((Mask & 0xFFFF0000)!=0) << 4 + | ((Mask & 0xFF00FF00)!=0) << 3 + | ((Mask & 0xF0F0F0F0)!=0) << 2 + | ((Mask & 0xCCCCCCCC)!=0) << 1 + | ((Mask & 0xAAAAAAAA)!=0); +} +/*===========================================================================*/ +/* This function will convert the DDraw error code to its macro string. The*/ +/* returned pointer is static so you need not worry about memory managemnet */ +/* but the error message gets written over from call to call... */ +/*===========================================================================*/ +/* RETURN: pointer to the single static buffer that hold the error message. */ +/*===========================================================================*/ +char *ErrorStringD3D( HRESULT hr ) +{ + static char errorString[128]; + + switch( hr ) + { + case DDERR_ALREADYINITIALIZED: + strcpy( errorString, "DDERR_ALREADYINITIALIZED" ); + break; + + case DDERR_CANNOTATTACHSURFACE: + strcpy( errorString, "DDERR_CANNOTATTACHSURFACE" ); + break; + + case DDERR_CANNOTDETACHSURFACE: + strcpy( errorString, "DDERR_CANNOTDETACHSURFACE" ); + break; + + case DDERR_CURRENTLYNOTAVAIL: + strcpy( errorString, "DDERR_CURRENTLYNOTAVAIL" ); + break; + + case DDERR_EXCEPTION: + strcpy( errorString, "DDERR_EXCEPTION" ); + break; + + case DDERR_GENERIC: + strcpy( errorString, "DDERR_GENERIC" ); + break; + + case DDERR_HEIGHTALIGN: + strcpy( errorString, "DDERR_HEIGHTALIGN" ); + break; + + case DDERR_INCOMPATIBLEPRIMARY: + strcpy( errorString, "DDERR_INCOMPATIBLEPRIMARY" ); + break; + + case DDERR_INVALIDCAPS: + strcpy( errorString, "DDERR_INVALIDCAPS" ); + break; + + case DDERR_INVALIDCLIPLIST: + strcpy( errorString, "DDERR_INVALIDCLIPLIST" ); + break; + + case DDERR_INVALIDMODE: + strcpy( errorString, "DDERR_INVALIDMODE" ); + break; + + case DDERR_INVALIDOBJECT: + strcpy( errorString, "DDERR_INVALIDOBJECT" ); + break; + + case DDERR_INVALIDPARAMS: + strcpy( errorString, "DDERR_INVALIDPARAMS" ); + break; + + case DDERR_INVALIDPIXELFORMAT: + strcpy( errorString, "DDERR_INVALIDPIXELFORMAT" ); + break; + + case DDERR_INVALIDRECT: + strcpy( errorString, "DDERR_INVALIDRECT" ); + break; + + case DDERR_LOCKEDSURFACES: + strcpy( errorString, "DDERR_LOCKEDSURFACES" ); + break; + + case DDERR_NO3D: + strcpy( errorString, "DDERR_NO3D" ); + break; + + case DDERR_NOALPHAHW: + strcpy( errorString, "DDERR_NOALPHAHW" ); + break; + + case DDERR_NOCLIPLIST: + strcpy( errorString, "DDERR_NOCLIPLIST" ); + break; + + case DDERR_NOCOLORCONVHW: + strcpy( errorString, "DDERR_NOCOLORCONVHW" ); + break; + + case DDERR_NOCOOPERATIVELEVELSET: + strcpy( errorString, "DDERR_NOCOOPERATIVELEVELSET" ); + break; + + case DDERR_NOCOLORKEY: + strcpy( errorString, "DDERR_NOCOLORKEY" ); + break; + + case DDERR_NOCOLORKEYHW: + strcpy( errorString, "DDERR_NOCOLORKEYHW" ); + break; + + case DDERR_NODIRECTDRAWSUPPORT: + strcpy( errorString, "DDERR_NODIRECTDRAWSUPPORT" ); + break; + + case DDERR_NOEXCLUSIVEMODE: + strcpy( errorString, "DDERR_NOEXCLUSIVEMODE" ); + break; + + case DDERR_NOFLIPHW: + strcpy( errorString, "DDERR_NOFLIPHW" ); + break; + + case DDERR_NOGDI: + strcpy( errorString, "DDERR_NOGDI" ); + break; + + case DDERR_NOMIRRORHW: + strcpy( errorString, "DDERR_NOMIRRORHW" ); + break; + + case DDERR_NOTFOUND: + strcpy( errorString, "DDERR_NOTFOUND" ); + break; + + case DDERR_NOOVERLAYHW: + strcpy( errorString, "DDERR_NOOVERLAYHW" ); + break; + + case DDERR_OVERLAPPINGRECTS: + strcpy( errorString, "DDERR_OVERLAPPINGRECTS" ); + break; + + case DDERR_NORASTEROPHW: + strcpy( errorString, "DDERR_NORASTEROPHW" ); + break; + + case DDERR_NOROTATIONHW: + strcpy( errorString, "DDERR_NOROTATIONHW" ); + break; + + case DDERR_NOSTRETCHHW: + strcpy( errorString, "DDERR_NOSTRETCHHW" ); + break; + + case DDERR_NOT4BITCOLOR: + strcpy( errorString, "DDERR_NOT4BITCOLOR" ); + break; + + case DDERR_NOT4BITCOLORINDEX: + strcpy( errorString, "DDERR_NOT4BITCOLORINDEX" ); + break; + + case DDERR_NOT8BITCOLOR: + strcpy( errorString, "DDERR_NOT8BITCOLOR" ); + break; + + case DDERR_NOTEXTUREHW: + strcpy( errorString, "DDERR_NOTEXTUREHW" ); + break; + + case DDERR_NOVSYNCHW: + strcpy( errorString, "DDERR_NOVSYNCHW" ); + break; + + case DDERR_NOZBUFFERHW: + strcpy( errorString, "DDERR_NOZBUFFERHW" ); + break; + + case DDERR_NOZOVERLAYHW: + strcpy( errorString, "DDERR_NOZOVERLAYHW" ); + break; + + case DDERR_OUTOFCAPS: + strcpy( errorString, "DDERR_OUTOFCAPS" ); + break; + + case DDERR_OUTOFMEMORY: + strcpy( errorString, "DDERR_OUTOFMEMORY" ); + break; + + case DDERR_OUTOFVIDEOMEMORY: + strcpy( errorString, "DDERR_OUTOFVIDEOMEMORY" ); + break; + + case DDERR_OVERLAYCANTCLIP: + strcpy( errorString, "DDERR_OVERLAYCANTCLIP" ); + break; + + case DDERR_OVERLAYCOLORKEYONLYONEACTIVE: + strcpy( errorString, "DDERR_OVERLAYCOLORKEYONLYONEACTIVE" ); + break; + + case DDERR_PALETTEBUSY: + strcpy( errorString, "DDERR_PALETTEBUSY" ); + break; + + case DDERR_COLORKEYNOTSET: + strcpy( errorString, "DDERR_COLORKEYNOTSET" ); + break; + + case DDERR_SURFACEALREADYATTACHED: + strcpy( errorString, "DDERR_SURFACEALREADYATTACHED" ); + break; + + case DDERR_SURFACEALREADYDEPENDENT: + strcpy( errorString, "DDERR_SURFACEALREADYDEPENDENT" ); + break; + + case DDERR_SURFACEBUSY: + strcpy( errorString, "DDERR_SURFACEBUSY" ); + break; + + case DDERR_CANTLOCKSURFACE: + strcpy( errorString, "DDERR_CANTLOCKSURFACE" ); + break; + + case DDERR_SURFACEISOBSCURED: + strcpy( errorString, "DDERR_SURFACEISOBSCURED" ); + break; + + case DDERR_SURFACELOST: + strcpy( errorString, "DDERR_SURFACELOST" ); + break; + + case DDERR_SURFACENOTATTACHED: + strcpy( errorString, "DDERR_SURFACENOTATTACHED" ); + break; + + case DDERR_TOOBIGHEIGHT: + strcpy( errorString, "DDERR_TOOBIGHEIGHT" ); + break; + + case DDERR_TOOBIGSIZE: + strcpy( errorString, "DDERR_TOOBIGSIZE" ); + break; + + case DDERR_TOOBIGWIDTH: + strcpy( errorString, "DDERR_TOOBIGWIDTH" ); + break; + + case DDERR_UNSUPPORTED: + strcpy( errorString, "DDERR_UNSUPPORTED" ); + break; + + case DDERR_UNSUPPORTEDFORMAT: + strcpy( errorString, "DDERR_UNSUPPORTEDFORMAT" ); + break; + + case DDERR_UNSUPPORTEDMASK: + strcpy( errorString, "DDERR_UNSUPPORTEDMASK" ); + break; + + case DDERR_INVALIDSTREAM: + strcpy( errorString, "DDERR_INVALIDSTREAM" ); + break; + + case DDERR_VERTICALBLANKINPROGRESS: + strcpy( errorString, "DDERR_VERTICALBLANKINPROGRESS" ); + break; + + case DDERR_WASSTILLDRAWING: + strcpy( errorString, "DDERR_WASSTILLDRAWING" ); + break; + + case DDERR_XALIGN: + strcpy( errorString, "DDERR_XALIGN" ); + break; + + case DDERR_INVALIDDIRECTDRAWGUID: + strcpy( errorString, "DDERR_INVALIDDIRECTDRAWGUID" ); + break; + + case DDERR_DIRECTDRAWALREADYCREATED: + strcpy( errorString, "DDERR_DIRECTDRAWALREADYCREATED" ); + break; + + case DDERR_NODIRECTDRAWHW: + strcpy( errorString, "DDERR_NODIRECTDRAWHW" ); + break; + + case DDERR_PRIMARYSURFACEALREADYEXISTS: + strcpy( errorString, "DDERR_PRIMARYSURFACEALREADYEXISTS" ); + break; + + case DDERR_NOEMULATION: + strcpy( errorString, "DDERR_NOEMULATION" ); + break; + + case DDERR_REGIONTOOSMALL: + strcpy( errorString, "DDERR_REGIONTOOSMALL" ); + break; + + case DDERR_CLIPPERISUSINGHWND: + strcpy( errorString, "DDERR_CLIPPERISUSINGHWND" ); + break; + + case DDERR_NOCLIPPERATTACHED: + strcpy( errorString, "DDERR_NOCLIPPERATTACHED" ); + break; + + case DDERR_NOHWND: + strcpy( errorString, "DDERR_NOHWND" ); + break; + + case DDERR_HWNDSUBCLASSED: + strcpy( errorString, "DDERR_HWNDSUBCLASSED" ); + break; + + case DDERR_HWNDALREADYSET: + strcpy( errorString, "DDERR_HWNDALREADYSET" ); + break; + + case DDERR_NOPALETTEATTACHED: + strcpy( errorString, "DDERR_NOPALETTEATTACHED" ); + break; + + case DDERR_NOPALETTEHW: + strcpy( errorString, "DDERR_NOPALETTEHW" ); + break; + + case DDERR_BLTFASTCANTCLIP: + strcpy( errorString, "DDERR_BLTFASTCANTCLIP" ); + break; + + case DDERR_NOBLTHW: + strcpy( errorString, "DDERR_NOBLTHW" ); + break; + + case DDERR_NODDROPSHW: + strcpy( errorString, "DDERR_NODDROPSHW" ); + break; + + case DDERR_OVERLAYNOTVISIBLE: + strcpy( errorString, "DDERR_OVERLAYNOTVISIBLE" ); + break; + + case DDERR_NOOVERLAYDEST: + strcpy( errorString, "DDERR_NOOVERLAYDEST" ); + break; + + case DDERR_INVALIDPOSITION: + strcpy( errorString, "DDERR_INVALIDPOSITION" ); + break; + + case DDERR_NOTAOVERLAYSURFACE: + strcpy( errorString, "DDERR_NOTAOVERLAYSURFACE" ); + break; + + case DDERR_EXCLUSIVEMODEALREADYSET: + strcpy( errorString, "DDERR_EXCLUSIVEMODEALREADYSET" ); + break; + + case DDERR_NOTFLIPPABLE: + strcpy( errorString, "DDERR_NOTFLIPPABLE" ); + break; + + case DDERR_CANTDUPLICATE: + strcpy( errorString, "DDERR_CANTDUPLICATE" ); + break; + + case DDERR_NOTLOCKED: + strcpy( errorString, "DDERR_NOTLOCKED" ); + break; + + case DDERR_CANTCREATEDC: + strcpy( errorString, "DDERR_CANTCREATEDC" ); + break; + + case DDERR_NODC: + strcpy( errorString, "DDERR_NODC" ); + break; + + case DDERR_WRONGMODE: + strcpy( errorString, "DDERR_WRONGMODE" ); + break; + + case DDERR_IMPLICITLYCREATED: + strcpy( errorString, "DDERR_IMPLICITLYCREATED" ); + break; + + case DDERR_NOTPALETTIZED: + strcpy( errorString, "DDERR_NOTPALETTIZED" ); + break; + + case DDERR_UNSUPPORTEDMODE: + strcpy( errorString, "DDERR_UNSUPPORTEDMODE" ); + break; + + case DDERR_NOMIPMAPHW: + strcpy( errorString, "DDERR_NOMIPMAPHW" ); + break; + + case DDERR_INVALIDSURFACETYPE: + strcpy( errorString, "DDERR_INVALIDSURFACETYPE" ); + break; + + case DDERR_NOOPTIMIZEHW: + strcpy( errorString, "DDERR_NOOPTIMIZEHW" ); + break; + + case DDERR_NOTLOADED: + strcpy( errorString, "DDERR_NOTLOADED" ); + break; + + case DDERR_NOFOCUSWINDOW: + strcpy( errorString, "DDERR_NOFOCUSWINDOW" ); + break; + + case DDERR_DCALREADYCREATED: + strcpy( errorString, "DDERR_DCALREADYCREATED" ); + break; + + case DDERR_NONONLOCALVIDMEM: + strcpy( errorString, "DDERR_NONONLOCALVIDMEM" ); + break; + + case DDERR_CANTPAGELOCK: + strcpy( errorString, "DDERR_CANTPAGELOCK" ); + break; + + case DDERR_CANTPAGEUNLOCK: + strcpy( errorString, "DDERR_CANTPAGEUNLOCK" ); + break; + + case DDERR_NOTPAGELOCKED: + strcpy( errorString, "DDERR_NOTPAGELOCKED" ); + break; + + case DDERR_MOREDATA: + strcpy( errorString, "DDERR_MOREDATA" ); + break; + + case DDERR_EXPIRED: + strcpy( errorString, "DDERR_EXPIRED" ); + break; + + case DDERR_VIDEONOTACTIVE: + strcpy( errorString, "DDERR_VIDEONOTACTIVE" ); + break; + + case DDERR_DEVICEDOESNTOWNSURFACE: + strcpy( errorString, "DDERR_DEVICEDOESNTOWNSURFACE" ); + break; + + case DDERR_NOTINITIALIZED: + strcpy( errorString, "DDERR_NOTINITIALIZED" ); + break; + + default: + strcpy( errorString, "" ); + break; + } + + return &errorString[0]; +} diff --git a/src/mesa/drivers/d3d/DDrawPROCS.c b/src/mesa/drivers/d3d/DDrawPROCS.c index a02a89debb..10dcfdbabb 100644 --- a/src/mesa/drivers/d3d/DDrawPROCS.c +++ b/src/mesa/drivers/d3d/DDrawPROCS.c @@ -1,399 +1,399 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#include "D3DMesa.h" -/*===========================================================================*/ -/* This call will clear the render surface using the pixel info built from */ -/* the surface at creation time. The call uses Lock/Unlock to access the */ -/* surface. The call also special cases a full clear or a dirty rectangle. */ -/* Finally the call returns the new clear mask that reflects that the color */ -/* buffer was cleared. */ -/*===========================================================================*/ -/* RETURN: the original mask with the bits cleared that represents the buffer*/ -/* or buffers we just cleared. */ -/*===========================================================================*/ -GLbitfield ClearBuffers( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - DDSURFACEDESC2 *pddsd2; - UCHAR *pBuffer, - *pScanLine; - int index, - index2; - DWORD dwColor; - - if ( mask & GL_COLOR_BUFFER_BIT ) - { - /* Lock the surface to get the surface pointer. */ - pddsd2 = LockHAL( pContext->pShared, TRUE ); - - /* Solve the color once only. */ - dwColor = ( ((DWORD)((float)pContext->rClear * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); - dwColor |= ( ((DWORD)((float)pContext->gClear * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); - dwColor |= ( ((DWORD)((float)pContext->bClear * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); - - if ( all ) - { - for( index = 0, pScanLine = (UCHAR *)pddsd2->lpSurface; index < pContext->pShared->dwHeight; index++, pScanLine += pddsd2->lPitch ) - for( pBuffer = pScanLine, index2 = 0; index2 < pContext->pShared->dwWidth; index2++, pBuffer += pContext->pShared->pixel.cb ) - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - else - { - pScanLine = ((UCHAR *)pddsd2->lpSurface) + - ( (FLIP( pContext->pShared->dwHeight, (y+height)) * pddsd2->lPitch) + (x * pContext->pShared->pixel.cb) ); - - for( index = 0; index < height; index++, pScanLine += pddsd2->lPitch ) - { - for( index2 = 0, pBuffer = pScanLine; index2 < width; index2++, pBuffer += pContext->pShared->pixel.cb ) - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - } - - UnlockHAL( pContext->pShared, TRUE ); - } - - return (mask & ~GL_COLOR_BUFFER_BIT); -} -/*===========================================================================*/ -/* This proc (as all others) has been written for the general case. I use */ -/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ -/* Screen render surface uses. The alpha is ignored as Mesa does it in SW. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void WSpanRGB( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgb[][3], const GLubyte mask[] ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - DDSURFACEDESC2 *pddsd2; - UCHAR *pBuffer; - int index; - DWORD dwColor; - - /* Get the surface pointer and the pitch. */ - pddsd2 = LockHAL( pContext->pShared, TRUE ); - - /* Find the start of the span. Invert y for Windows. */ - pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y) * pddsd2->lPitch) + (x*pContext->pShared->pixel.cb); - - if ( mask ) - { - for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) - { - if ( mask[index] ) - { - /* Pack the color components. */ - dwColor = ( ((DWORD)((float)rgb[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); - dwColor |= ( ((DWORD)((float)rgb[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); - dwColor |= ( ((DWORD)((float)rgb[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - } - } - else - { - for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) - { - /* Pack the color components. */ - dwColor = ( ((DWORD)((float)rgb[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); - dwColor |= ( ((DWORD)((float)rgb[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); - dwColor |= ( ((DWORD)((float)rgb[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - } - - /* Giver back. */ - UnlockHAL( pContext->pShared, TRUE ); -} -/*===========================================================================*/ -/* This proc (as all others) has been written for the general case. I use */ -/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ -/* Screen render surface uses. The alpha is ignored as Mesa does it in SW. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void WSpanRGBA( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgba[][4], const GLubyte mask[] ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - DDSURFACEDESC2 *pddsd2; - UCHAR *pBuffer; - int index; - DWORD dwColor; - - /* Get the surface pointer and the pitch. */ - pddsd2 = LockHAL( pContext->pShared, TRUE ); - - /* Find the start of the span. Invert y for Windows. */ - pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y) * pddsd2->lPitch) + (x*pContext->pShared->pixel.cb); - - if ( mask ) - { - for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) - { - if ( mask[index] ) - { - /* Pack the color components. */ - dwColor = ( ((DWORD)((float)rgba[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); - dwColor |= ( ((DWORD)((float)rgba[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); - dwColor |= ( ((DWORD)((float)rgba[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - } - } - else - { - for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) - { - /* Pack the color components. */ - dwColor = ( ((DWORD)((float)rgba[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); - dwColor |= ( ((DWORD)((float)rgba[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); - dwColor |= ( ((DWORD)((float)rgba[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - } - - /* Giver back. */ - UnlockHAL( pContext->pShared, TRUE ); -} -/*===========================================================================*/ -/* This proc (as all others) has been written for the general case. I use */ -/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ -/* Screen render surface uses. The color is solved once from the current */ -/* color components. The alpha is ignored as Mesa is doing it in SW. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void WSpanRGBAMono( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte mask[] ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - DDSURFACEDESC2 *pddsd2; - UCHAR *pBuffer; - int index; - DWORD dwColor; - - /* Lock the surface to get the surface pointer and the pitch. */ - pddsd2 = LockHAL( pContext->pShared, TRUE ); - - /* Solve the color once only. (no alpha) */ - dwColor = ( ((DWORD)((float)pContext->rCurrent * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); - dwColor |= ( ((DWORD)((float)pContext->gCurrent * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); - dwColor |= ( ((DWORD)((float)pContext->bCurrent * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); - - /* Find the start of the span. Invert y for Windows. */ - pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y) * pddsd2->lPitch) + (x*pContext->pShared->pixel.cb); - - if ( mask ) - { - for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) - if ( mask[index] ) - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - else - { - for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - - /* Giver back. */ - UnlockHAL( pContext->pShared, TRUE ); -} -/*===========================================================================*/ -/* This proc (as all others) has been written for the general case. I use */ -/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ -/* Screen render surface uses. The alpha is ignored as Mesa does it in SW. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void WPixelsRGBA( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte rgba[][4], const GLubyte mask[] ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - DDSURFACEDESC2 *pddsd2; - UCHAR *pBuffer; - int index; - DWORD dwColor; - - /* Get the surface pointer and the pitch. */ - pddsd2 = LockHAL( pContext->pShared, TRUE ); - - if ( mask ) - { - for( index = 0; index < n; index++ ) - { - if ( mask[index] ) - { - /* Pack the color components. */ - dwColor = ( ((DWORD)((float)rgba[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); - dwColor |= ( ((DWORD)((float)rgba[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); - dwColor |= ( ((DWORD)((float)rgba[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); - - /* Find the pixel. Invert y for Windows. */ - pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb); - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - } - } - else - { - for( index = 0; index < n; index++ ) - { - /* Pack the color components. */ - dwColor = ( ((DWORD)((float)rgba[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); - dwColor |= ( ((DWORD)((float)rgba[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); - dwColor |= ( ((DWORD)((float)rgba[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); - - /* Find the pixel. Invert y for Windows. */ - pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb); - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - } - - /* Giver back. */ - UnlockHAL( pContext->pShared, TRUE ); -} -/*===========================================================================*/ -/* This proc (as all others) has been written for the general case. I use */ -/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ -/* Screen render surface uses. The color is solved once from the current */ -/* color components. The alpha is ignored as Mesa is doing it in SW. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void WPixelsRGBAMono( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte mask[] ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - DDSURFACEDESC2 *pddsd2; - UCHAR *pBuffer; - int index; - DWORD dwColor; - - /* Get the surface pointer and the pitch. */ - pddsd2 = LockHAL( pContext->pShared, TRUE ); - - /* Solve the color once only. I don't uses the alpha. */ - dwColor = ( ((DWORD)((float)pContext->rCurrent * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); - dwColor |= ( ((DWORD)((float)pContext->gCurrent * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); - dwColor |= ( ((DWORD)((float)pContext->bCurrent * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); - - if ( mask ) - { - /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ - for( index = 0; index < n; index++ ) - { - if ( mask[index] ) - { - /* Find the pixel. Invert y for Windows. */ - pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb); - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - } - } - else - { - /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ - for( index = 0; index < n; index++ ) - { - /* Find the pixel. Invert y for Windows. */ - pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb); - memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); - } - } - - /* Giver back. */ - UnlockHAL( pContext->pShared, TRUE ); -} -/*===========================================================================*/ -/* This proc isn't written for speed rather its to handle the general case. */ -/* I grab each pixel from the surface and unpack the info using the PIXELINFO*/ -/* structure that was generated from the OffScreen surface pixelformat. The */ -/* function will not fill in the alpha value as Mesa I have Mesa allocate its*/ -/* own alpha channel when the context was created. I did this as I didn't */ -/* feel that it was worth the effort to try and get HW to work (bus bound). */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void RSpanRGBA( const GLcontext* ctx, GLuint n, GLint x, GLint y, GLubyte rgba[][4] ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - DDSURFACEDESC2 *pddsd2; - UCHAR *pBuffer; - int index; - DWORD *pdwColor; - - /* Get the surface pointer and the pitch. */ - pddsd2 = LockHAL( pContext->pShared, TRUE ); - - /* Find the start of the span. Invert y for Windows. */ - pBuffer = (UCHAR *)pddsd2->lpSurface + - (FLIP(pContext->pShared->dwHeight,y) * pddsd2->lPitch) + - (x*pContext->pShared->pixel.cb); - - /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ - for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) - { - pdwColor = (DWORD *)pBuffer; - rgba[index][RCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwRMask) >> pContext->pShared->pixel.rShift) / pContext->pShared->pixel.rScale); - rgba[index][GCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwGMask) >> pContext->pShared->pixel.gShift) / pContext->pShared->pixel.gScale); - rgba[index][BCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwBMask) >> pContext->pShared->pixel.bShift) / pContext->pShared->pixel.bScale); - } - - /* Giver back. */ - UnlockHAL( pContext->pShared, TRUE ); -} -/*===========================================================================*/ -/* This proc isn't written for speed rather its to handle the general case. */ -/* I grab each pixel from the surface and unpack the info using the PIXELINFO*/ -/* structure that was generated from the OffScreen surface pixelformat. The */ -/* function will not fill in the alpha value as Mesa I have Mesa allocate its*/ -/* own alpha channel when the context was created. I did this as I didn't */ -/* feel that it was worth the effort to try and get HW to work (bus bound). */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void RPixelsRGBA( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], GLubyte rgba[][4], const GLubyte mask[] ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - DDSURFACEDESC2 *pddsd2; - int index; - DWORD *pdwColor; - - /* Get the surface pointer and the pitch. */ - pddsd2 = LockHAL( pContext->pShared, TRUE ); - - if ( mask ) - { - /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ - for( index = 0; index < n; index++ ) - { - if ( mask[index] ) - { - /* Find the start of the pixel. Invert y for Windows. */ - pdwColor = (DWORD *)((UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb)); - rgba[index][RCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwRMask) >> pContext->pShared->pixel.rShift) / pContext->pShared->pixel.rScale); - rgba[index][GCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwGMask) >> pContext->pShared->pixel.gShift) / pContext->pShared->pixel.gScale); - rgba[index][BCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwBMask) >> pContext->pShared->pixel.bShift) / pContext->pShared->pixel.bScale); - } - } - } - else - { - /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ - for( index = 0; index < n; index++ ) - { - /* Find the start of the pixel. Invert y for Windows. */ - pdwColor = (DWORD *)((UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb)); - rgba[index][RCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwRMask) >> pContext->pShared->pixel.rShift) / pContext->pShared->pixel.rScale); - rgba[index][GCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwGMask) >> pContext->pShared->pixel.gShift) / pContext->pShared->pixel.gScale); - rgba[index][BCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwBMask) >> pContext->pShared->pixel.bShift) / pContext->pShared->pixel.bScale); - } - } - - /* Giver back. */ - UnlockHAL( pContext->pShared, TRUE ); -} +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#include "D3DMesa.h" +/*===========================================================================*/ +/* This call will clear the render surface using the pixel info built from */ +/* the surface at creation time. The call uses Lock/Unlock to access the */ +/* surface. The call also special cases a full clear or a dirty rectangle. */ +/* Finally the call returns the new clear mask that reflects that the color */ +/* buffer was cleared. */ +/*===========================================================================*/ +/* RETURN: the original mask with the bits cleared that represents the buffer*/ +/* or buffers we just cleared. */ +/*===========================================================================*/ +GLbitfield ClearBuffers( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + DDSURFACEDESC2 *pddsd2; + UCHAR *pBuffer, + *pScanLine; + int index, + index2; + DWORD dwColor; + + if ( mask & GL_COLOR_BUFFER_BIT ) + { + /* Lock the surface to get the surface pointer. */ + pddsd2 = LockHAL( pContext->pShared, TRUE ); + + /* Solve the color once only. */ + dwColor = ( ((DWORD)((float)pContext->rClear * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); + dwColor |= ( ((DWORD)((float)pContext->gClear * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); + dwColor |= ( ((DWORD)((float)pContext->bClear * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); + + if ( all ) + { + for( index = 0, pScanLine = (UCHAR *)pddsd2->lpSurface; index < pContext->pShared->dwHeight; index++, pScanLine += pddsd2->lPitch ) + for( pBuffer = pScanLine, index2 = 0; index2 < pContext->pShared->dwWidth; index2++, pBuffer += pContext->pShared->pixel.cb ) + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + else + { + pScanLine = ((UCHAR *)pddsd2->lpSurface) + + ( (FLIP( pContext->pShared->dwHeight, (y+height)) * pddsd2->lPitch) + (x * pContext->pShared->pixel.cb) ); + + for( index = 0; index < height; index++, pScanLine += pddsd2->lPitch ) + { + for( index2 = 0, pBuffer = pScanLine; index2 < width; index2++, pBuffer += pContext->pShared->pixel.cb ) + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + } + + UnlockHAL( pContext->pShared, TRUE ); + } + + return (mask & ~GL_COLOR_BUFFER_BIT); +} +/*===========================================================================*/ +/* This proc (as all others) has been written for the general case. I use */ +/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ +/* Screen render surface uses. The alpha is ignored as Mesa does it in SW. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void WSpanRGB( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgb[][3], const GLubyte mask[] ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + DDSURFACEDESC2 *pddsd2; + UCHAR *pBuffer; + int index; + DWORD dwColor; + + /* Get the surface pointer and the pitch. */ + pddsd2 = LockHAL( pContext->pShared, TRUE ); + + /* Find the start of the span. Invert y for Windows. */ + pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y) * pddsd2->lPitch) + (x*pContext->pShared->pixel.cb); + + if ( mask ) + { + for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) + { + if ( mask[index] ) + { + /* Pack the color components. */ + dwColor = ( ((DWORD)((float)rgb[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); + dwColor |= ( ((DWORD)((float)rgb[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); + dwColor |= ( ((DWORD)((float)rgb[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + } + } + else + { + for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) + { + /* Pack the color components. */ + dwColor = ( ((DWORD)((float)rgb[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); + dwColor |= ( ((DWORD)((float)rgb[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); + dwColor |= ( ((DWORD)((float)rgb[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + } + + /* Giver back. */ + UnlockHAL( pContext->pShared, TRUE ); +} +/*===========================================================================*/ +/* This proc (as all others) has been written for the general case. I use */ +/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ +/* Screen render surface uses. The alpha is ignored as Mesa does it in SW. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void WSpanRGBA( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgba[][4], const GLubyte mask[] ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + DDSURFACEDESC2 *pddsd2; + UCHAR *pBuffer; + int index; + DWORD dwColor; + + /* Get the surface pointer and the pitch. */ + pddsd2 = LockHAL( pContext->pShared, TRUE ); + + /* Find the start of the span. Invert y for Windows. */ + pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y) * pddsd2->lPitch) + (x*pContext->pShared->pixel.cb); + + if ( mask ) + { + for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) + { + if ( mask[index] ) + { + /* Pack the color components. */ + dwColor = ( ((DWORD)((float)rgba[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); + dwColor |= ( ((DWORD)((float)rgba[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); + dwColor |= ( ((DWORD)((float)rgba[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + } + } + else + { + for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) + { + /* Pack the color components. */ + dwColor = ( ((DWORD)((float)rgba[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); + dwColor |= ( ((DWORD)((float)rgba[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); + dwColor |= ( ((DWORD)((float)rgba[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + } + + /* Giver back. */ + UnlockHAL( pContext->pShared, TRUE ); +} +/*===========================================================================*/ +/* This proc (as all others) has been written for the general case. I use */ +/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ +/* Screen render surface uses. The color is solved once from the current */ +/* color components. The alpha is ignored as Mesa is doing it in SW. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void WSpanRGBAMono( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte mask[] ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + DDSURFACEDESC2 *pddsd2; + UCHAR *pBuffer; + int index; + DWORD dwColor; + + /* Lock the surface to get the surface pointer and the pitch. */ + pddsd2 = LockHAL( pContext->pShared, TRUE ); + + /* Solve the color once only. (no alpha) */ + dwColor = ( ((DWORD)((float)pContext->rCurrent * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); + dwColor |= ( ((DWORD)((float)pContext->gCurrent * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); + dwColor |= ( ((DWORD)((float)pContext->bCurrent * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); + + /* Find the start of the span. Invert y for Windows. */ + pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y) * pddsd2->lPitch) + (x*pContext->pShared->pixel.cb); + + if ( mask ) + { + for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) + if ( mask[index] ) + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + else + { + for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + + /* Giver back. */ + UnlockHAL( pContext->pShared, TRUE ); +} +/*===========================================================================*/ +/* This proc (as all others) has been written for the general case. I use */ +/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ +/* Screen render surface uses. The alpha is ignored as Mesa does it in SW. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void WPixelsRGBA( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte rgba[][4], const GLubyte mask[] ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + DDSURFACEDESC2 *pddsd2; + UCHAR *pBuffer; + int index; + DWORD dwColor; + + /* Get the surface pointer and the pitch. */ + pddsd2 = LockHAL( pContext->pShared, TRUE ); + + if ( mask ) + { + for( index = 0; index < n; index++ ) + { + if ( mask[index] ) + { + /* Pack the color components. */ + dwColor = ( ((DWORD)((float)rgba[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); + dwColor |= ( ((DWORD)((float)rgba[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); + dwColor |= ( ((DWORD)((float)rgba[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); + + /* Find the pixel. Invert y for Windows. */ + pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb); + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + } + } + else + { + for( index = 0; index < n; index++ ) + { + /* Pack the color components. */ + dwColor = ( ((DWORD)((float)rgba[index][RCOMP] * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); + dwColor |= ( ((DWORD)((float)rgba[index][GCOMP] * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); + dwColor |= ( ((DWORD)((float)rgba[index][BCOMP] * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); + + /* Find the pixel. Invert y for Windows. */ + pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb); + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + } + + /* Giver back. */ + UnlockHAL( pContext->pShared, TRUE ); +} +/*===========================================================================*/ +/* This proc (as all others) has been written for the general case. I use */ +/* the PIXELINFO structure to pack the pixel from RGB24 to whatever the Off- */ +/* Screen render surface uses. The color is solved once from the current */ +/* color components. The alpha is ignored as Mesa is doing it in SW. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void WPixelsRGBAMono( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte mask[] ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + DDSURFACEDESC2 *pddsd2; + UCHAR *pBuffer; + int index; + DWORD dwColor; + + /* Get the surface pointer and the pitch. */ + pddsd2 = LockHAL( pContext->pShared, TRUE ); + + /* Solve the color once only. I don't uses the alpha. */ + dwColor = ( ((DWORD)((float)pContext->rCurrent * pContext->pShared->pixel.rScale)) << pContext->pShared->pixel.rShift ); + dwColor |= ( ((DWORD)((float)pContext->gCurrent * pContext->pShared->pixel.gScale)) << pContext->pShared->pixel.gShift ); + dwColor |= ( ((DWORD)((float)pContext->bCurrent * pContext->pShared->pixel.bScale)) << pContext->pShared->pixel.bShift ); + + if ( mask ) + { + /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ + for( index = 0; index < n; index++ ) + { + if ( mask[index] ) + { + /* Find the pixel. Invert y for Windows. */ + pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb); + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + } + } + else + { + /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ + for( index = 0; index < n; index++ ) + { + /* Find the pixel. Invert y for Windows. */ + pBuffer = (UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb); + memcpy( pBuffer, &dwColor, pContext->pShared->pixel.cb ); + } + } + + /* Giver back. */ + UnlockHAL( pContext->pShared, TRUE ); +} +/*===========================================================================*/ +/* This proc isn't written for speed rather its to handle the general case. */ +/* I grab each pixel from the surface and unpack the info using the PIXELINFO*/ +/* structure that was generated from the OffScreen surface pixelformat. The */ +/* function will not fill in the alpha value as Mesa I have Mesa allocate its*/ +/* own alpha channel when the context was created. I did this as I didn't */ +/* feel that it was worth the effort to try and get HW to work (bus bound). */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void RSpanRGBA( const GLcontext* ctx, GLuint n, GLint x, GLint y, GLubyte rgba[][4] ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + DDSURFACEDESC2 *pddsd2; + UCHAR *pBuffer; + int index; + DWORD *pdwColor; + + /* Get the surface pointer and the pitch. */ + pddsd2 = LockHAL( pContext->pShared, TRUE ); + + /* Find the start of the span. Invert y for Windows. */ + pBuffer = (UCHAR *)pddsd2->lpSurface + + (FLIP(pContext->pShared->dwHeight,y) * pddsd2->lPitch) + + (x*pContext->pShared->pixel.cb); + + /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ + for( index = 0; index < n; index++, pBuffer += pContext->pShared->pixel.cb ) + { + pdwColor = (DWORD *)pBuffer; + rgba[index][RCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwRMask) >> pContext->pShared->pixel.rShift) / pContext->pShared->pixel.rScale); + rgba[index][GCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwGMask) >> pContext->pShared->pixel.gShift) / pContext->pShared->pixel.gScale); + rgba[index][BCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwBMask) >> pContext->pShared->pixel.bShift) / pContext->pShared->pixel.bScale); + } + + /* Giver back. */ + UnlockHAL( pContext->pShared, TRUE ); +} +/*===========================================================================*/ +/* This proc isn't written for speed rather its to handle the general case. */ +/* I grab each pixel from the surface and unpack the info using the PIXELINFO*/ +/* structure that was generated from the OffScreen surface pixelformat. The */ +/* function will not fill in the alpha value as Mesa I have Mesa allocate its*/ +/* own alpha channel when the context was created. I did this as I didn't */ +/* feel that it was worth the effort to try and get HW to work (bus bound). */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void RPixelsRGBA( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], GLubyte rgba[][4], const GLubyte mask[] ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + DDSURFACEDESC2 *pddsd2; + int index; + DWORD *pdwColor; + + /* Get the surface pointer and the pitch. */ + pddsd2 = LockHAL( pContext->pShared, TRUE ); + + if ( mask ) + { + /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ + for( index = 0; index < n; index++ ) + { + if ( mask[index] ) + { + /* Find the start of the pixel. Invert y for Windows. */ + pdwColor = (DWORD *)((UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb)); + rgba[index][RCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwRMask) >> pContext->pShared->pixel.rShift) / pContext->pShared->pixel.rScale); + rgba[index][GCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwGMask) >> pContext->pShared->pixel.gShift) / pContext->pShared->pixel.gScale); + rgba[index][BCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwBMask) >> pContext->pShared->pixel.bShift) / pContext->pShared->pixel.bScale); + } + } + } + else + { + /* We store the surface pointer as a UCHAR so that pixel.cb (count in btyles) will work for all. */ + for( index = 0; index < n; index++ ) + { + /* Find the start of the pixel. Invert y for Windows. */ + pdwColor = (DWORD *)((UCHAR *)pddsd2->lpSurface + (FLIP(pContext->pShared->dwHeight,y[index]) * pddsd2->lPitch) + (x[index]*pContext->pShared->pixel.cb)); + rgba[index][RCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwRMask) >> pContext->pShared->pixel.rShift) / pContext->pShared->pixel.rScale); + rgba[index][GCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwGMask) >> pContext->pShared->pixel.gShift) / pContext->pShared->pixel.gScale); + rgba[index][BCOMP] = (GLubyte)((float)((*pdwColor & pContext->pShared->pixel.dwBMask) >> pContext->pShared->pixel.bShift) / pContext->pShared->pixel.bScale); + } + } + + /* Giver back. */ + UnlockHAL( pContext->pShared, TRUE ); +} diff --git a/src/mesa/drivers/d3d/DEBUG.C b/src/mesa/drivers/d3d/DEBUG.C index dfa524bf31..79e273903a 100644 --- a/src/mesa/drivers/d3d/DEBUG.C +++ b/src/mesa/drivers/d3d/DEBUG.C @@ -1,143 +1,143 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#include "Debug.h" -/*===========================================================================*/ -/* Global variables. */ -/*===========================================================================*/ -DWORD g_DBGMask = DBG_ALL_ERROR; -/*===========================================================================*/ -/* This is your basic DPF function with printf like support. The function */ -/* also works with a global debug mask variable. I have written support that*/ -/* allows for the user's enviroment variable space to be read and set the */ -/* masks. This is done when the dll starts and is only in the debug version.*/ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void _cdecl DebugPrint( int mask, char *pszFormat, ... ) -{ - char buffer[512]; - va_list args; - - /* A mask of 0 will always pass. Easy to remeber. */ - if ( (mask == 0) || (mask & g_DBGMask) ) - { - va_start( args, pszFormat ); - - if ( mask & DBG_ALL_ERROR ) - OutputDebugString( "MesaD3D: (ERROR)" ); - else - OutputDebugString( "MesaD3D: " ); - - vsprintf( buffer, pszFormat, args ); - strcat( buffer, "\n" ); - OutputDebugString( buffer ); - - va_end( args ); - } -} -/*===========================================================================*/ -/* This call reads the users enviroment variables and sets any debug mask */ -/* that they have set to TRUE. Now the value must be "TRUE". */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void ReadDBGEnv( void ) -{ - g_DBGMask = DBG_ALL_ERROR; - -#define IS_VAR_SET(v) if ( getenv( # v ) && !strcmp(getenv( # v ),"TRUE") ) g_DBGMask |= v; - - IS_VAR_SET( DBG_FUNC ); - IS_VAR_SET( DBG_STATES ); - - IS_VAR_SET( DBG_CNTX_INFO ); - IS_VAR_SET( DBG_CNTX_WARN ); - IS_VAR_SET( DBG_CNTX_PROFILE ); - IS_VAR_SET( DBG_CNTX_ERROR ); - IS_VAR_SET( DBG_CNTX_ALL ); - - IS_VAR_SET( DBG_PRIM_INFO ); - IS_VAR_SET( DBG_PRIM_WARN ); - IS_VAR_SET( DBG_PRIM_PROFILE ); - IS_VAR_SET( DBG_PRIM_ERROR ); - IS_VAR_SET( DBG_PRIM_ALL ); - - IS_VAR_SET( DBG_TXT_INFO ); - IS_VAR_SET( DBG_TXT_WARN ); - IS_VAR_SET( DBG_TXT_PROFILE ); - IS_VAR_SET( DBG_TXT_ERROR ); - IS_VAR_SET( DBG_TXT_ALL ); - - IS_VAR_SET( DBG_ALL_INFO ); - IS_VAR_SET( DBG_ALL_WARN ); - IS_VAR_SET( DBG_ALL_PROFILE ); - IS_VAR_SET( DBG_ALL_ERROR ); - IS_VAR_SET( DBG_ALL ); - -#undef IS_VAR_SET -} -/*===========================================================================*/ -/* This function will take a pointer to a DDSURFACEDESC2 structure & display*/ -/* the parsed information using a DPF call. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -void DebugPixelFormat( char *pszSurfaceName, DDPIXELFORMAT *pddpf ) -{ - char buffer[256]; - - /* Parse the flag type and write the string equivalent. */ - if ( pddpf->dwFlags & DDPF_ALPHA ) - strcat( buffer, "DDPF_ALPHA " ); - if ( pddpf->dwFlags & DDPF_ALPHAPIXELS ) - strcat( buffer, "DDPF_ALPHAPIXELS " ); - if ( pddpf->dwFlags & DDPF_ALPHAPREMULT ) - strcat( buffer, "DDPF_ALPHAPREMULT " ); - if ( pddpf->dwFlags & DDPF_BUMPLUMINANCE ) - strcat( buffer, "DDPF_BUMPLUMINANCE " ); - if ( pddpf->dwFlags & DDPF_BUMPDUDV ) - strcat( buffer, "DDPF_BUMPDUDV " ); - if ( pddpf->dwFlags & DDPF_COMPRESSED ) - strcat( buffer, "DDPF_COMPRESSED " ); - if ( pddpf->dwFlags & DDPF_FOURCC ) - strcat( buffer, "DDPF_FOURCC " ); - if ( pddpf->dwFlags & DDPF_LUMINANCE ) - strcat( buffer, "DDPF_LUMINANCE " ); - if ( pddpf->dwFlags & DDPF_PALETTEINDEXED1 ) - strcat( buffer, "DDPF_PALETTEINDEXED1 " ); - if ( pddpf->dwFlags & DDPF_PALETTEINDEXED2 ) - strcat( buffer, "DDPF_PALETTEINDEXED2 " ); - if ( pddpf->dwFlags & DDPF_PALETTEINDEXED4 ) - strcat( buffer, "DDPF_PALETTEINDEXED4 " ); - if ( pddpf->dwFlags & DDPF_PALETTEINDEXED8 ) - strcat( buffer, "DDPF_PALETTEINDEXED8 " ); - if ( pddpf->dwFlags & DDPF_PALETTEINDEXEDTO8 ) - strcat( buffer, "DDPF_PALETTEINDEXEDTO8 " ); - if ( pddpf->dwFlags & DDPF_RGB ) - strcat( buffer, "DDPF_RGB " ); - if ( pddpf->dwFlags & DDPF_RGBTOYUV ) - strcat( buffer, "DDPF_RGBTOYUV " ); - if ( pddpf->dwFlags & DDPF_STENCILBUFFER ) - strcat( buffer, "DDPF_STENCILBUFFER " ); - if ( pddpf->dwFlags & DDPF_YUV ) - strcat( buffer, "DDPF_YUV " ); - if ( pddpf->dwFlags & DDPF_ZBUFFER ) - strcat( buffer, "DDPF_ZBUFFER " ); - if ( pddpf->dwFlags & DDPF_ZPIXELS ) - strcat( buffer, "DDPF_ZPIXELS " ); - - DPF(( (DBG_TXT_INFO|DBG_CNTX_INFO),"%s", buffer )); -} - - - - - +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#include "Debug.h" +/*===========================================================================*/ +/* Global variables. */ +/*===========================================================================*/ +DWORD g_DBGMask = DBG_ALL_ERROR; +/*===========================================================================*/ +/* This is your basic DPF function with printf like support. The function */ +/* also works with a global debug mask variable. I have written support that*/ +/* allows for the user's enviroment variable space to be read and set the */ +/* masks. This is done when the dll starts and is only in the debug version.*/ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void _cdecl DebugPrint( int mask, char *pszFormat, ... ) +{ + char buffer[512]; + va_list args; + + /* A mask of 0 will always pass. Easy to remeber. */ + if ( (mask == 0) || (mask & g_DBGMask) ) + { + va_start( args, pszFormat ); + + if ( mask & DBG_ALL_ERROR ) + OutputDebugString( "MesaD3D: (ERROR)" ); + else + OutputDebugString( "MesaD3D: " ); + + vsprintf( buffer, pszFormat, args ); + strcat( buffer, "\n" ); + OutputDebugString( buffer ); + + va_end( args ); + } +} +/*===========================================================================*/ +/* This call reads the users enviroment variables and sets any debug mask */ +/* that they have set to TRUE. Now the value must be "TRUE". */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void ReadDBGEnv( void ) +{ + g_DBGMask = DBG_ALL_ERROR; + +#define IS_VAR_SET(v) if ( getenv( # v ) && !strcmp(getenv( # v ),"TRUE") ) g_DBGMask |= v; + + IS_VAR_SET( DBG_FUNC ); + IS_VAR_SET( DBG_STATES ); + + IS_VAR_SET( DBG_CNTX_INFO ); + IS_VAR_SET( DBG_CNTX_WARN ); + IS_VAR_SET( DBG_CNTX_PROFILE ); + IS_VAR_SET( DBG_CNTX_ERROR ); + IS_VAR_SET( DBG_CNTX_ALL ); + + IS_VAR_SET( DBG_PRIM_INFO ); + IS_VAR_SET( DBG_PRIM_WARN ); + IS_VAR_SET( DBG_PRIM_PROFILE ); + IS_VAR_SET( DBG_PRIM_ERROR ); + IS_VAR_SET( DBG_PRIM_ALL ); + + IS_VAR_SET( DBG_TXT_INFO ); + IS_VAR_SET( DBG_TXT_WARN ); + IS_VAR_SET( DBG_TXT_PROFILE ); + IS_VAR_SET( DBG_TXT_ERROR ); + IS_VAR_SET( DBG_TXT_ALL ); + + IS_VAR_SET( DBG_ALL_INFO ); + IS_VAR_SET( DBG_ALL_WARN ); + IS_VAR_SET( DBG_ALL_PROFILE ); + IS_VAR_SET( DBG_ALL_ERROR ); + IS_VAR_SET( DBG_ALL ); + +#undef IS_VAR_SET +} +/*===========================================================================*/ +/* This function will take a pointer to a DDSURFACEDESC2 structure & display*/ +/* the parsed information using a DPF call. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +void DebugPixelFormat( char *pszSurfaceName, DDPIXELFORMAT *pddpf ) +{ + char buffer[256]; + + /* Parse the flag type and write the string equivalent. */ + if ( pddpf->dwFlags & DDPF_ALPHA ) + strcat( buffer, "DDPF_ALPHA " ); + if ( pddpf->dwFlags & DDPF_ALPHAPIXELS ) + strcat( buffer, "DDPF_ALPHAPIXELS " ); + if ( pddpf->dwFlags & DDPF_ALPHAPREMULT ) + strcat( buffer, "DDPF_ALPHAPREMULT " ); + if ( pddpf->dwFlags & DDPF_BUMPLUMINANCE ) + strcat( buffer, "DDPF_BUMPLUMINANCE " ); + if ( pddpf->dwFlags & DDPF_BUMPDUDV ) + strcat( buffer, "DDPF_BUMPDUDV " ); + if ( pddpf->dwFlags & DDPF_COMPRESSED ) + strcat( buffer, "DDPF_COMPRESSED " ); + if ( pddpf->dwFlags & DDPF_FOURCC ) + strcat( buffer, "DDPF_FOURCC " ); + if ( pddpf->dwFlags & DDPF_LUMINANCE ) + strcat( buffer, "DDPF_LUMINANCE " ); + if ( pddpf->dwFlags & DDPF_PALETTEINDEXED1 ) + strcat( buffer, "DDPF_PALETTEINDEXED1 " ); + if ( pddpf->dwFlags & DDPF_PALETTEINDEXED2 ) + strcat( buffer, "DDPF_PALETTEINDEXED2 " ); + if ( pddpf->dwFlags & DDPF_PALETTEINDEXED4 ) + strcat( buffer, "DDPF_PALETTEINDEXED4 " ); + if ( pddpf->dwFlags & DDPF_PALETTEINDEXED8 ) + strcat( buffer, "DDPF_PALETTEINDEXED8 " ); + if ( pddpf->dwFlags & DDPF_PALETTEINDEXEDTO8 ) + strcat( buffer, "DDPF_PALETTEINDEXEDTO8 " ); + if ( pddpf->dwFlags & DDPF_RGB ) + strcat( buffer, "DDPF_RGB " ); + if ( pddpf->dwFlags & DDPF_RGBTOYUV ) + strcat( buffer, "DDPF_RGBTOYUV " ); + if ( pddpf->dwFlags & DDPF_STENCILBUFFER ) + strcat( buffer, "DDPF_STENCILBUFFER " ); + if ( pddpf->dwFlags & DDPF_YUV ) + strcat( buffer, "DDPF_YUV " ); + if ( pddpf->dwFlags & DDPF_ZBUFFER ) + strcat( buffer, "DDPF_ZBUFFER " ); + if ( pddpf->dwFlags & DDPF_ZPIXELS ) + strcat( buffer, "DDPF_ZPIXELS " ); + + DPF(( (DBG_TXT_INFO|DBG_CNTX_INFO),"%s", buffer )); +} + + + + + diff --git a/src/mesa/drivers/d3d/DEBUG.H b/src/mesa/drivers/d3d/DEBUG.H index 76122b9121..daeeb72278 100644 --- a/src/mesa/drivers/d3d/DEBUG.H +++ b/src/mesa/drivers/d3d/DEBUG.H @@ -1,90 +1,90 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 DirectX 6 Driver */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#ifndef _DEBUG_H -#define _DEBUG_H - -#ifdef __cplusplus -extern "C" { -#endif - -/*===========================================================================*/ -/* Includes. */ -/*===========================================================================*/ -#include -#include -#include -#include -#include "D3DShared.h" -/*===========================================================================*/ -/* Magic numbers. */ -/*===========================================================================*/ -/*===========================================================================*/ -/* Macros defines. */ -/*===========================================================================*/ -#define DBG_FUNC 0x00000001 -#define DBG_STATES 0x00000002 - -#define DBG_CNTX_INFO 0x00000010 -#define DBG_CNTX_WARN 0x00000020 -#define DBG_CNTX_PROFILE 0x00000040 -#define DBG_CNTX_ERROR 0x00000080 -#define DBG_CNTX_ALL 0x000000F0 - -#define DBG_PRIM_INFO 0x00000100 -#define DBG_PRIM_WARN 0x00000200 -#define DBG_PRIM_PROFILE 0x00000400 -#define DBG_PRIM_ERROR 0x00000800 -#define DBG_PRIM_ALL 0x00000F00 - -#define DBG_TXT_INFO 0x00001000 -#define DBG_TXT_WARN 0x00002000 -#define DBG_TXT_PROFILE 0x00004000 -#define DBG_TXT_ERROR 0x00008000 -#define DBG_TXT_ALL 0x0000F000 - -#define DBG_ALL_INFO 0x11111110 -#define DBG_ALL_WARN 0x22222220 -#define DBG_ALL_PROFILE 0x44444440 -#define DBG_ALL_ERROR 0x88888880 -#define DBG_ALL 0xFFFFFFFF - -#ifdef D3D_DEBUG -# define DPF(arg) DebugPrint arg -# define RIP(pH,msg,err) OutputDebugString(msg); \ - OutputDebugString(err); \ - OutputDebugString("\n"); \ - FatalShutDown(pH) -#else -# define DPF(arg) -# define RIP(pH,msg,err) FatalShutDown(pH) -#endif -/*===========================================================================*/ -/* Type defines. */ -/*===========================================================================*/ -/*===========================================================================*/ -/* Function prototypes. */ -/*===========================================================================*/ -extern void ReadDBGEnv( void ); -extern void _cdecl DebugPrint( int mask, char *pszFormat, ... ); -extern void DebugPixelFormat( char *pszSurfaceName, DDPIXELFORMAT *pddpf ); -/*===========================================================================*/ -/* Global variables. */ -/*===========================================================================*/ -extern DWORD g_DBGMask; - -#ifdef __cplusplus -} -#endif - -#endif - - - +/*===========================================================================*/ +/* */ +/* Mesa-3.0 DirectX 6 Driver */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1999-1998 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#ifndef _DEBUG_H +#define _DEBUG_H + +#ifdef __cplusplus +extern "C" { +#endif + +/*===========================================================================*/ +/* Includes. */ +/*===========================================================================*/ +#include +#include +#include +#include +#include "D3DShared.h" +/*===========================================================================*/ +/* Magic numbers. */ +/*===========================================================================*/ +/*===========================================================================*/ +/* Macros defines. */ +/*===========================================================================*/ +#define DBG_FUNC 0x00000001 +#define DBG_STATES 0x00000002 + +#define DBG_CNTX_INFO 0x00000010 +#define DBG_CNTX_WARN 0x00000020 +#define DBG_CNTX_PROFILE 0x00000040 +#define DBG_CNTX_ERROR 0x00000080 +#define DBG_CNTX_ALL 0x000000F0 + +#define DBG_PRIM_INFO 0x00000100 +#define DBG_PRIM_WARN 0x00000200 +#define DBG_PRIM_PROFILE 0x00000400 +#define DBG_PRIM_ERROR 0x00000800 +#define DBG_PRIM_ALL 0x00000F00 + +#define DBG_TXT_INFO 0x00001000 +#define DBG_TXT_WARN 0x00002000 +#define DBG_TXT_PROFILE 0x00004000 +#define DBG_TXT_ERROR 0x00008000 +#define DBG_TXT_ALL 0x0000F000 + +#define DBG_ALL_INFO 0x11111110 +#define DBG_ALL_WARN 0x22222220 +#define DBG_ALL_PROFILE 0x44444440 +#define DBG_ALL_ERROR 0x88888880 +#define DBG_ALL 0xFFFFFFFF + +#ifdef D3D_DEBUG +# define DPF(arg) DebugPrint arg +# define RIP(pH,msg,err) OutputDebugString(msg); \ + OutputDebugString(err); \ + OutputDebugString("\n"); \ + FatalShutDown(pH) +#else +# define DPF(arg) +# define RIP(pH,msg,err) FatalShutDown(pH) +#endif +/*===========================================================================*/ +/* Type defines. */ +/*===========================================================================*/ +/*===========================================================================*/ +/* Function prototypes. */ +/*===========================================================================*/ +extern void ReadDBGEnv( void ); +extern void _cdecl DebugPrint( int mask, char *pszFormat, ... ); +extern void DebugPixelFormat( char *pszSurfaceName, DDPIXELFORMAT *pddpf ); +/*===========================================================================*/ +/* Global variables. */ +/*===========================================================================*/ +extern DWORD g_DBGMask; + +#ifdef __cplusplus +} +#endif + +#endif + + + diff --git a/src/mesa/drivers/d3d/DbgEnv.bat b/src/mesa/drivers/d3d/DbgEnv.bat index 40858e6684..acea045856 100644 --- a/src/mesa/drivers/d3d/DbgEnv.bat +++ b/src/mesa/drivers/d3d/DbgEnv.bat @@ -1,25 +1,25 @@ -SET DBG_FUNC=FALSE - -SET DBG_CNTX_INFO=TRUE -SET DBG_CNTX_WARN=TRUE -SET DBG_CNTX_PROFILE=FALSE -SET DBG_CNTX_ERROR=TRUE -SET DBG_CNTX_ALL=TRUE - -SET DBG_PRIM_INFO=FALSE -SET DBG_PRIM_WARN=FALSE -SET DBG_PRIM_PROFILE=FALSE -SET DBG_PRIM_ERROR=TRUE -SET DBG_PRIM_ALL=FALSE - -SET DBG_TXT_INFO=FALSE -SET DBG_TXT_WARN=TRUE -SET DBG_TXT_PROFILE=FALSE -SET DBG_TXT_ERROR=TRUE -SET DBG_TXT_ALL=FALSE - -SET DBG_ALL_INFO=FALSE -SET DBG_ALL_WARN=TRUE -SET DBG_ALL_PROFILE=FALSE -SET DBG_ALL_ERROR=TRUE -SET DBG_ALL=FALSE +SET DBG_FUNC=FALSE + +SET DBG_CNTX_INFO=TRUE +SET DBG_CNTX_WARN=TRUE +SET DBG_CNTX_PROFILE=FALSE +SET DBG_CNTX_ERROR=TRUE +SET DBG_CNTX_ALL=TRUE + +SET DBG_PRIM_INFO=FALSE +SET DBG_PRIM_WARN=FALSE +SET DBG_PRIM_PROFILE=FALSE +SET DBG_PRIM_ERROR=TRUE +SET DBG_PRIM_ALL=FALSE + +SET DBG_TXT_INFO=FALSE +SET DBG_TXT_WARN=TRUE +SET DBG_TXT_PROFILE=FALSE +SET DBG_TXT_ERROR=TRUE +SET DBG_TXT_ALL=FALSE + +SET DBG_ALL_INFO=FALSE +SET DBG_ALL_WARN=TRUE +SET DBG_ALL_PROFILE=FALSE +SET DBG_ALL_ERROR=TRUE +SET DBG_ALL=FALSE diff --git a/src/mesa/drivers/d3d/MAKEFILE b/src/mesa/drivers/d3d/MAKEFILE index ad1d40dc27..6aa88f3cef 100644 --- a/src/mesa/drivers/d3d/MAKEFILE +++ b/src/mesa/drivers/d3d/MAKEFILE @@ -1,101 +1,101 @@ -############################################################################## -# -# Mesa-3.0 Makefile for DirectX 6 Driver -# -# By Leigh McRae -# -# http://www.altsoftware.com/ -# -# Copyright (c) 1999-1998 alt.software inc. All Rights Reserved -############################################################################## -NAME= -TARGET= WGL Driver (D3DHAL) - -D3D_DIR=$(MAKEDIR)\D3D -TARGET_DIR=e:\WinNT\System32 -TEMP_DIR=c:\Temp - -SPACE=- -LINKER=link.exe - -INCLUDE=$(SDKROOT)\include;$(INCLUDE) -LIB=$(SDKROOT)\lib;$(LIB) -############################################################################## -CFLAGS = /c /nologo /W1 /G5 /I..\ /I..\..\Include \ - /D "_WIN32" /D "WIN32" /D "_WINDOWS" /D "__WIN32__" /D "__MSC__" /D "MESAD3D" -CPPFLAGS= /c /nologo /W1 /G5 /I..\ /I..\..\Include \ - /D "_WIN32" /D "WIN32" /D "_WINDOWS" /D "__WIN32__" /D "__MSC__" /D "MESAD3D" - -!IF "$(DEBUG)" == "1" - -CFLAGS = /MTd /Od /Z7 /Yd /D "_DEBUG" /D "D3D_DEBUG" $(CFLAGS) -CPPFLAGS = /MTd /Od /Z7 /Yd /D "_DEBUG" /D "D3D_DEBUG" $(CPPFLAGS) -BUILD_TYPE=debug - -!ELSE - -CFLAGS = /MT /Ox /D "NDEBUG" $(CFLAGS) -CPPFLAGS = /MT /Ox /D "NDEBUG" $(CPPFLAGS) -BUILD_TYPE=release - -!ENDIF -############################################################################## -SRCS_WGL = wgl.c D3Dvbrender.c DDrawPROCS.c NULLProcs.c Debug.c -SRCS_HAL = D3DInit.cpp D3DRaster.cpp D3DTextureMgr.cpp D3DUtils.cpp D3DCaps.cpp -OBJS_WGL = $(SRCS_WGL:.c=.obj) -OBJS_HAL = $(SRCS_HAL:.cpp=.obj) - -WINLIBS = kernel32.lib user32.lib gdi32.lib oldnames.lib -DXLIBS = -LIBS = $(WINLIBS) $(DXLIBS) -############################################################################### -# Primary Targets # -############################################################################### - -default: header WGL HAL footer - -all: default - -WGL : $(OBJS_WGL) - -HAL : $(OBJS_HAL) - -install : forceit - @echo $(SPACE) - @echo ======================================== - @echo Install files created. - @echo ======================================== - - -############################################################################### -# Secondary Targets # -############################################################################### - -clean: - @echo ======================================== - @echo Cleaning $(TARGET) - @del *.obj - @del *.dep - @del *.exp - @del *.ncb - @del *.plg - @del *.lib - @echo ======================================== - -header: - @echo ============================================================ - @echo Building $(TARGET) ($(BUILD_TYPE) version) - @echo ============================================================ - @echo $(SPACE) - -footer: - @echo $(SPACE) - @echo ============================================================ - @echo DONE building $(TARGET) ($(BUILD_TYPE) version) - @echo ============================================================ - -forceit: - - - - +############################################################################## +# +# Mesa-3.0 Makefile for DirectX 6 Driver +# +# By Leigh McRae +# +# http://www.altsoftware.com/ +# +# Copyright (c) 1999-1998 alt.software inc. All Rights Reserved +############################################################################## +NAME= +TARGET= WGL Driver (D3DHAL) + +D3D_DIR=$(MAKEDIR)\D3D +TARGET_DIR=e:\WinNT\System32 +TEMP_DIR=c:\Temp + +SPACE=- +LINKER=link.exe + +INCLUDE=$(SDKROOT)\include;$(INCLUDE) +LIB=$(SDKROOT)\lib;$(LIB) +############################################################################## +CFLAGS = /c /nologo /W1 /G5 /I..\ /I..\..\Include \ + /D "_WIN32" /D "WIN32" /D "_WINDOWS" /D "__WIN32__" /D "__MSC__" /D "MESAD3D" +CPPFLAGS= /c /nologo /W1 /G5 /I..\ /I..\..\Include \ + /D "_WIN32" /D "WIN32" /D "_WINDOWS" /D "__WIN32__" /D "__MSC__" /D "MESAD3D" + +!IF "$(DEBUG)" == "1" + +CFLAGS = /MTd /Od /Z7 /Yd /D "_DEBUG" /D "D3D_DEBUG" $(CFLAGS) +CPPFLAGS = /MTd /Od /Z7 /Yd /D "_DEBUG" /D "D3D_DEBUG" $(CPPFLAGS) +BUILD_TYPE=debug + +!ELSE + +CFLAGS = /MT /Ox /D "NDEBUG" $(CFLAGS) +CPPFLAGS = /MT /Ox /D "NDEBUG" $(CPPFLAGS) +BUILD_TYPE=release + +!ENDIF +############################################################################## +SRCS_WGL = wgl.c D3Dvbrender.c DDrawPROCS.c NULLProcs.c Debug.c +SRCS_HAL = D3DInit.cpp D3DRaster.cpp D3DTextureMgr.cpp D3DUtils.cpp D3DCaps.cpp +OBJS_WGL = $(SRCS_WGL:.c=.obj) +OBJS_HAL = $(SRCS_HAL:.cpp=.obj) + +WINLIBS = kernel32.lib user32.lib gdi32.lib oldnames.lib +DXLIBS = +LIBS = $(WINLIBS) $(DXLIBS) +############################################################################### +# Primary Targets # +############################################################################### + +default: header WGL HAL footer + +all: default + +WGL : $(OBJS_WGL) + +HAL : $(OBJS_HAL) + +install : forceit + @echo $(SPACE) + @echo ======================================== + @echo Install files created. + @echo ======================================== + + +############################################################################### +# Secondary Targets # +############################################################################### + +clean: + @echo ======================================== + @echo Cleaning $(TARGET) + @del *.obj + @del *.dep + @del *.exp + @del *.ncb + @del *.plg + @del *.lib + @echo ======================================== + +header: + @echo ============================================================ + @echo Building $(TARGET) ($(BUILD_TYPE) version) + @echo ============================================================ + @echo $(SPACE) + +footer: + @echo $(SPACE) + @echo ============================================================ + @echo DONE building $(TARGET) ($(BUILD_TYPE) version) + @echo ============================================================ + +forceit: + + + + diff --git a/src/mesa/drivers/d3d/OPENGL32.DEF b/src/mesa/drivers/d3d/OPENGL32.DEF index b32bd1fef8..bc0304da33 100644 --- a/src/mesa/drivers/d3d/OPENGL32.DEF +++ b/src/mesa/drivers/d3d/OPENGL32.DEF @@ -1,442 +1,442 @@ -;=========================================================================== -; -; Mesa-3.0 DirectX 6 Driver -; -; By Leigh McRae -; -; http://www.altsoftware.com/ -; -; Copyright (c) 1999-1998 alt.software inc. All Rights Reserved -;=========================================================================== -NAME OpenGL32.DLL -DESCRIPTION "Mesa-3.0 DX6 Driver Version 0.5" - -EXPORTS - DllMain - glAccum - glAlphaFunc - glAreTexturesResident - glAreTexturesResidentEXT - glArrayElement - glArrayElementEXT - glBegin - glBindTexture - glBindTextureEXT - glBitmap - glBlendColorEXT - glBlendEquationEXT - glBlendFunc - glCallList - glCallLists - glClear - glClearAccum - glClearColor - glClearDepth - glClearIndex - glClearStencil - glClipPlane - glColor3b - glColor3bv - glColor3d - glColor3dv - glColor3f - glColor3fv - glColor3i - glColor3iv - glColor3s - glColor3sv - glColor3ub - glColor3ubv - glColor3ui - glColor3uiv - glColor3us - glColor3usv - glColor4b - glColor4bv - glColor4d - glColor4dv - glColor4f - glColor4fv - glColor4i - glColor4iv - glColor4s - glColor4sv - glColor4ub - glColor4ubv - glColor4ui - glColor4uiv - glColor4us - glColor4usv - glColorMask - glColorMaterial - glColorPointer - glColorPointerEXT - glColorSubTableEXT - glColorTableEXT - glCopyPixels - glCopyTexImage1D - glCopyTexImage2D - glCopyTexSubImage1D - glCopyTexSubImage2D - glCopyTexSubImage3DEXT - glCullFace - glDeleteLists - glDeleteTextures - glDeleteTexturesEXT - glDepthFunc - glDepthMask - glDepthRange - glDisable - glDisableClientState - glDrawArrays - glDrawArraysEXT - glDrawBuffer - glDrawElements - glDrawPixels - glEdgeFlag - glEdgeFlagPointer - glEdgeFlagPointerEXT - glEdgeFlagv - glEnable - glEnableClientState - glEnd - glEndList - glEvalCoord1d - glEvalCoord1dv - glEvalCoord1f - glEvalCoord1fv - glEvalCoord2d - glEvalCoord2dv - glEvalCoord2f - glEvalCoord2fv - glEvalMesh1 - glEvalMesh2 - glEvalPoint1 - glEvalPoint2 - glFeedbackBuffer - glFinish - glFlush - glFogf - glFogfv - glFogi - glFogiv - glFrontFace - glFrustum - glGenLists - glGenTextures - glGenTexturesEXT - glGetBooleanv - glGetClipPlane - glGetColorTableEXT - glGetColorTableParameterfvEXT - glGetColorTableParameterivEXT - glGetDoublev - glGetError - glGetFloatv - glGetIntegerv - glGetLightfv - glGetLightiv - glGetMapdv - glGetMapfv - glGetMapiv - glGetMaterialfv - glGetMaterialiv - glGetPixelMapfv - glGetPixelMapuiv - glGetPixelMapusv - glGetPointerv - glGetPointervEXT - glGetPolygonStipple - glGetString - glGetTexEnvfv - glGetTexEnviv - glGetTexGendv - glGetTexGenfv - glGetTexGeniv - glGetTexImage - glGetTexLevelParameterfv - glGetTexLevelParameteriv - glGetTexParameterfv - glGetTexParameteriv - glHint - glIndexd - glIndexdv - glIndexf - glIndexfv - glIndexi - glIndexiv - glIndexMask - glIndexPointer - glIndexPointerEXT - glIndexs - glIndexsv - glIndexub - glIndexubv - glInitNames - glInterleavedArrays - glIsEnabled - glIsList - glIsTexture - glIsTextureEXT - glLightf - glLightfv - glLighti - glLightiv - glLightModelf - glLightModelfv - glLightModeli - glLightModeliv - glLineStipple - glLineWidth - glListBase - glLoadIdentity - glLoadMatrixd - glLoadMatrixf - glLoadName - glLogicOp - glMap1d - glMap1f - glMap2d - glMap2f - glMapGrid1d - glMapGrid1f - glMapGrid2d - glMapGrid2f - glMaterialf - glMaterialfv - glMateriali - glMaterialiv - glMatrixMode - glMultMatrixd - glMultMatrixf - glNewList - glNormal3b - glNormal3bv - glNormal3d - glNormal3dv - glNormal3f - glNormal3fv - glNormal3i - glNormal3iv - glNormal3s - glNormal3sv - glNormalPointer - glNormalPointerEXT - glOrtho - glPassThrough - glPixelMapfv - glPixelMapuiv - glPixelMapusv - glPixelStoref - glPixelStorei - glPixelTransferf - glPixelTransferi - glPixelZoom - glPointParameterfEXT - glPointParameterfvEXT - glPointSize - glPolygonMode - glPolygonOffset - glPolygonOffsetEXT - glPolygonStipple - glPopAttrib - glPopClientAttrib - glPopMatrix - glPopName - glPrioritizeTextures - glPrioritizeTexturesEXT - glPushAttrib - glPushClientAttrib - glPushMatrix - glPushName - glRasterPos2d - glRasterPos2dv - glRasterPos2f - glRasterPos2fv - glRasterPos2i - glRasterPos2iv - glRasterPos2s - glRasterPos2sv - glRasterPos3d - glRasterPos3dv - glRasterPos3f - glRasterPos3fv - glRasterPos3i - glRasterPos3iv - glRasterPos3s - glRasterPos3sv - glRasterPos4d - glRasterPos4dv - glRasterPos4f - glRasterPos4fv - glRasterPos4i - glRasterPos4iv - glRasterPos4s - glRasterPos4sv - glReadBuffer - glReadPixels - glRectd - glRectdv - glRectf - glRectfv - glRecti - glRectiv - glRects - glRectsv - glRenderMode - glResizeBuffersMESA - glRotated - glRotatef - glScaled - glScalef - glScissor - glSelectBuffer - glShadeModel - glStencilFunc - glStencilMask - glStencilOp - glTexCoord1d - glTexCoord1dv - glTexCoord1f - glTexCoord1fv - glTexCoord1i - glTexCoord1iv - glTexCoord1s - glTexCoord1sv - glTexCoord2d - glTexCoord2dv - glTexCoord2f - glTexCoord2fv - glTexCoord2i - glTexCoord2iv - glTexCoord2s - glTexCoord2sv - glTexCoord3d - glTexCoord3dv - glTexCoord3f - glTexCoord3fv - glTexCoord3i - glTexCoord3iv - glTexCoord3s - glTexCoord3sv - glTexCoord4d - glTexCoord4dv - glTexCoord4f - glTexCoord4fv - glTexCoord4i - glTexCoord4iv - glTexCoord4s - glTexCoord4sv - glTexCoordPointer - glTexCoordPointerEXT - glTexEnvf - glTexEnvfv - glTexEnvi - glTexEnviv - glTexGend - glTexGendv - glTexGenf - glTexGenfv - glTexGeni - glTexGeniv - glTexImage1D - glTexImage2D - glTexImage3DEXT - glTexParameterf - glTexParameterfv - glTexParameteri - glTexParameteriv - glTexSubImage1D - glTexSubImage2D - glTexSubImage3DEXT - glTranslated - glTranslatef - glVertex2d - glVertex2dv - glVertex2f - glVertex2fv - glVertex2i - glVertex2iv - glVertex2s - glVertex2sv - glVertex3d - glVertex3dv - glVertex3f - glVertex3fv - glVertex3i - glVertex3iv - glVertex3s - glVertex3sv - glVertex4d - glVertex4dv - glVertex4f - glVertex4fv - glVertex4i - glVertex4iv - glVertex4s - glVertex4sv - glVertexPointer - glVertexPointerEXT - glViewport - glWindowPos2dMESA - glWindowPos2dvMESA - glWindowPos2fMESA - glWindowPos2fvMESA - glWindowPos2iMESA - glWindowPos2ivMESA - glWindowPos2sMESA - glWindowPos2svMESA - glWindowPos3dMESA - glWindowPos3dvMESA - glWindowPos3fMESA - glWindowPos3fvMESA - glWindowPos3iMESA - glWindowPos3ivMESA - glWindowPos3sMESA - glWindowPos3svMESA - glWindowPos4dMESA - glWindowPos4dvMESA - glWindowPos4fMESA - glWindowPos4fvMESA - glWindowPos4iMESA - glWindowPos4ivMESA - glWindowPos4sMESA - glWindowPos4svMESA -; WMesaCreateContext -; WMesaDestroyContext -; WMesaMakeCurrent -; WMesaPaletteChange -; WMesaSwapBuffers -; OSMesaCreateContext -; OSMesaDestroyContext -; OSMesaMakeCurrent -; OSMesaGetCurrentContext -; OSMesaPixelStore -; OSMesaGetIntegerv -; OSMesaGetDepthBuffer - wglCopyContext - wglCreateContext - wglCreateLayerContext - wglDeleteContext -; wglDescribeLayerPlane - wglGetCurrentContext - wglGetCurrentDC -; wglGetLayerPaletteEntries - wglGetProcAddress - wglMakeCurrent -; wglRealizeLayerPalette -; wglSetLayerPaletteEntries - wglShareLists - wglSwapLayerBuffers - wglUseFontBitmapsA - wglUseFontBitmapsW - wglUseFontOutlinesA - wglUseFontOutlinesW - wglChoosePixelFormat - wglDescribePixelFormat - wglGetPixelFormat - wglSetPixelFormat - wglSwapBuffers - - - +;=========================================================================== +; +; Mesa-3.0 DirectX 6 Driver +; +; By Leigh McRae +; +; http://www.altsoftware.com/ +; +; Copyright (c) 1999-1998 alt.software inc. All Rights Reserved +;=========================================================================== +NAME OpenGL32.DLL +DESCRIPTION "Mesa-3.0 DX6 Driver Version 0.5" + +EXPORTS + DllMain + glAccum + glAlphaFunc + glAreTexturesResident + glAreTexturesResidentEXT + glArrayElement + glArrayElementEXT + glBegin + glBindTexture + glBindTextureEXT + glBitmap + glBlendColorEXT + glBlendEquationEXT + glBlendFunc + glCallList + glCallLists + glClear + glClearAccum + glClearColor + glClearDepth + glClearIndex + glClearStencil + glClipPlane + glColor3b + glColor3bv + glColor3d + glColor3dv + glColor3f + glColor3fv + glColor3i + glColor3iv + glColor3s + glColor3sv + glColor3ub + glColor3ubv + glColor3ui + glColor3uiv + glColor3us + glColor3usv + glColor4b + glColor4bv + glColor4d + glColor4dv + glColor4f + glColor4fv + glColor4i + glColor4iv + glColor4s + glColor4sv + glColor4ub + glColor4ubv + glColor4ui + glColor4uiv + glColor4us + glColor4usv + glColorMask + glColorMaterial + glColorPointer + glColorPointerEXT + glColorSubTableEXT + glColorTableEXT + glCopyPixels + glCopyTexImage1D + glCopyTexImage2D + glCopyTexSubImage1D + glCopyTexSubImage2D + glCopyTexSubImage3DEXT + glCullFace + glDeleteLists + glDeleteTextures + glDeleteTexturesEXT + glDepthFunc + glDepthMask + glDepthRange + glDisable + glDisableClientState + glDrawArrays + glDrawArraysEXT + glDrawBuffer + glDrawElements + glDrawPixels + glEdgeFlag + glEdgeFlagPointer + glEdgeFlagPointerEXT + glEdgeFlagv + glEnable + glEnableClientState + glEnd + glEndList + glEvalCoord1d + glEvalCoord1dv + glEvalCoord1f + glEvalCoord1fv + glEvalCoord2d + glEvalCoord2dv + glEvalCoord2f + glEvalCoord2fv + glEvalMesh1 + glEvalMesh2 + glEvalPoint1 + glEvalPoint2 + glFeedbackBuffer + glFinish + glFlush + glFogf + glFogfv + glFogi + glFogiv + glFrontFace + glFrustum + glGenLists + glGenTextures + glGenTexturesEXT + glGetBooleanv + glGetClipPlane + glGetColorTableEXT + glGetColorTableParameterfvEXT + glGetColorTableParameterivEXT + glGetDoublev + glGetError + glGetFloatv + glGetIntegerv + glGetLightfv + glGetLightiv + glGetMapdv + glGetMapfv + glGetMapiv + glGetMaterialfv + glGetMaterialiv + glGetPixelMapfv + glGetPixelMapuiv + glGetPixelMapusv + glGetPointerv + glGetPointervEXT + glGetPolygonStipple + glGetString + glGetTexEnvfv + glGetTexEnviv + glGetTexGendv + glGetTexGenfv + glGetTexGeniv + glGetTexImage + glGetTexLevelParameterfv + glGetTexLevelParameteriv + glGetTexParameterfv + glGetTexParameteriv + glHint + glIndexd + glIndexdv + glIndexf + glIndexfv + glIndexi + glIndexiv + glIndexMask + glIndexPointer + glIndexPointerEXT + glIndexs + glIndexsv + glIndexub + glIndexubv + glInitNames + glInterleavedArrays + glIsEnabled + glIsList + glIsTexture + glIsTextureEXT + glLightf + glLightfv + glLighti + glLightiv + glLightModelf + glLightModelfv + glLightModeli + glLightModeliv + glLineStipple + glLineWidth + glListBase + glLoadIdentity + glLoadMatrixd + glLoadMatrixf + glLoadName + glLogicOp + glMap1d + glMap1f + glMap2d + glMap2f + glMapGrid1d + glMapGrid1f + glMapGrid2d + glMapGrid2f + glMaterialf + glMaterialfv + glMateriali + glMaterialiv + glMatrixMode + glMultMatrixd + glMultMatrixf + glNewList + glNormal3b + glNormal3bv + glNormal3d + glNormal3dv + glNormal3f + glNormal3fv + glNormal3i + glNormal3iv + glNormal3s + glNormal3sv + glNormalPointer + glNormalPointerEXT + glOrtho + glPassThrough + glPixelMapfv + glPixelMapuiv + glPixelMapusv + glPixelStoref + glPixelStorei + glPixelTransferf + glPixelTransferi + glPixelZoom + glPointParameterfEXT + glPointParameterfvEXT + glPointSize + glPolygonMode + glPolygonOffset + glPolygonOffsetEXT + glPolygonStipple + glPopAttrib + glPopClientAttrib + glPopMatrix + glPopName + glPrioritizeTextures + glPrioritizeTexturesEXT + glPushAttrib + glPushClientAttrib + glPushMatrix + glPushName + glRasterPos2d + glRasterPos2dv + glRasterPos2f + glRasterPos2fv + glRasterPos2i + glRasterPos2iv + glRasterPos2s + glRasterPos2sv + glRasterPos3d + glRasterPos3dv + glRasterPos3f + glRasterPos3fv + glRasterPos3i + glRasterPos3iv + glRasterPos3s + glRasterPos3sv + glRasterPos4d + glRasterPos4dv + glRasterPos4f + glRasterPos4fv + glRasterPos4i + glRasterPos4iv + glRasterPos4s + glRasterPos4sv + glReadBuffer + glReadPixels + glRectd + glRectdv + glRectf + glRectfv + glRecti + glRectiv + glRects + glRectsv + glRenderMode + glResizeBuffersMESA + glRotated + glRotatef + glScaled + glScalef + glScissor + glSelectBuffer + glShadeModel + glStencilFunc + glStencilMask + glStencilOp + glTexCoord1d + glTexCoord1dv + glTexCoord1f + glTexCoord1fv + glTexCoord1i + glTexCoord1iv + glTexCoord1s + glTexCoord1sv + glTexCoord2d + glTexCoord2dv + glTexCoord2f + glTexCoord2fv + glTexCoord2i + glTexCoord2iv + glTexCoord2s + glTexCoord2sv + glTexCoord3d + glTexCoord3dv + glTexCoord3f + glTexCoord3fv + glTexCoord3i + glTexCoord3iv + glTexCoord3s + glTexCoord3sv + glTexCoord4d + glTexCoord4dv + glTexCoord4f + glTexCoord4fv + glTexCoord4i + glTexCoord4iv + glTexCoord4s + glTexCoord4sv + glTexCoordPointer + glTexCoordPointerEXT + glTexEnvf + glTexEnvfv + glTexEnvi + glTexEnviv + glTexGend + glTexGendv + glTexGenf + glTexGenfv + glTexGeni + glTexGeniv + glTexImage1D + glTexImage2D + glTexImage3DEXT + glTexParameterf + glTexParameterfv + glTexParameteri + glTexParameteriv + glTexSubImage1D + glTexSubImage2D + glTexSubImage3DEXT + glTranslated + glTranslatef + glVertex2d + glVertex2dv + glVertex2f + glVertex2fv + glVertex2i + glVertex2iv + glVertex2s + glVertex2sv + glVertex3d + glVertex3dv + glVertex3f + glVertex3fv + glVertex3i + glVertex3iv + glVertex3s + glVertex3sv + glVertex4d + glVertex4dv + glVertex4f + glVertex4fv + glVertex4i + glVertex4iv + glVertex4s + glVertex4sv + glVertexPointer + glVertexPointerEXT + glViewport + glWindowPos2dMESA + glWindowPos2dvMESA + glWindowPos2fMESA + glWindowPos2fvMESA + glWindowPos2iMESA + glWindowPos2ivMESA + glWindowPos2sMESA + glWindowPos2svMESA + glWindowPos3dMESA + glWindowPos3dvMESA + glWindowPos3fMESA + glWindowPos3fvMESA + glWindowPos3iMESA + glWindowPos3ivMESA + glWindowPos3sMESA + glWindowPos3svMESA + glWindowPos4dMESA + glWindowPos4dvMESA + glWindowPos4fMESA + glWindowPos4fvMESA + glWindowPos4iMESA + glWindowPos4ivMESA + glWindowPos4sMESA + glWindowPos4svMESA +; WMesaCreateContext +; WMesaDestroyContext +; WMesaMakeCurrent +; WMesaPaletteChange +; WMesaSwapBuffers +; OSMesaCreateContext +; OSMesaDestroyContext +; OSMesaMakeCurrent +; OSMesaGetCurrentContext +; OSMesaPixelStore +; OSMesaGetIntegerv +; OSMesaGetDepthBuffer + wglCopyContext + wglCreateContext + wglCreateLayerContext + wglDeleteContext +; wglDescribeLayerPlane + wglGetCurrentContext + wglGetCurrentDC +; wglGetLayerPaletteEntries + wglGetProcAddress + wglMakeCurrent +; wglRealizeLayerPalette +; wglSetLayerPaletteEntries + wglShareLists + wglSwapLayerBuffers + wglUseFontBitmapsA + wglUseFontBitmapsW + wglUseFontOutlinesA + wglUseFontOutlinesW + wglChoosePixelFormat + wglDescribePixelFormat + wglGetPixelFormat + wglSetPixelFormat + wglSwapBuffers + + + diff --git a/src/mesa/drivers/d3d/WGL.C b/src/mesa/drivers/d3d/WGL.C index e3b95e1de7..170d094ed4 100644 --- a/src/mesa/drivers/d3d/WGL.C +++ b/src/mesa/drivers/d3d/WGL.C @@ -1,1262 +1,1262 @@ -/*===========================================================================*/ -/* */ -/* Mesa-3.0 Makefile for DirectX 6 */ -/* */ -/* By Leigh McRae */ -/* */ -/* http://www.altsoftware.com/ */ -/* */ -/* Copyright (c) 1998-1997 alt.software inc. All Rights Reserved */ -/*===========================================================================*/ -#include "D3DMesa.h" -/*===========================================================================*/ -/* Window managment. */ -/*===========================================================================*/ -static BOOL InitOpenGL( HINSTANCE hInst ); -static BOOL TermOpenGL( HINSTANCE hInst ); -static BOOL ResizeContext( GLcontext *ctx ); -static BOOL MakeCurrent( D3DMESACONTEXT *pContext ); -static void DestroyContext( D3DMESACONTEXT *pContext ); -static BOOL UnBindWindow( D3DMESACONTEXT *pContext ); -LONG APIENTRY wglMonitorProc( HWND hwnd, UINT message, UINT wParam, LONG lParam ); -/*===========================================================================*/ -/* Mesa hooks. */ -/*===========================================================================*/ -static void SetupDDPointers( GLcontext *ctx ); -static void SetupSWDDPointers( GLcontext *ctx ); -static void SetupHWDDPointers( GLcontext *ctx ); -static void SetupNULLDDPointers( GLcontext *ctx ); -static const char *RendererString( void ); - -/* State Management hooks. */ -static void SetColor( GLcontext *ctx, GLubyte r, GLubyte g, GLubyte b, GLubyte a ); -static void ClearColor( GLcontext *ctx, GLubyte r, GLubyte g, GLubyte b, GLubyte a ); -static GLboolean SetBuffer( GLcontext *ctx, GLenum buffer ); - -/* Window Management hooks. */ -static void GetBufferSize( GLcontext *ctx, GLuint *width, GLuint *height ); -static void SetViewport( GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h ); -static void Flush( GLcontext *ctx ); - -/* Span rendering hooks. */ -void WSpanRGB( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgb[][3], const GLubyte mask[] ); -void WSpanRGBA( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgba[][4], const GLubyte mask[] ); -void WSpanRGBAMono( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte mask[] ); -void WPixelsRGBA( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte rgba[][4], const GLubyte mask[] ); -void WPixelsRGBAMono( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte mask[] ); -void RSpanRGBA( const GLcontext* ctx, GLuint n, GLint x, GLint y, GLubyte rgba[][4] ); -void RPixelsRGBA( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], GLubyte rgba[][4], const GLubyte mask[] ); -GLbitfield ClearBuffers( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ); - -/* Primitve rendering hooks. */ -GLboolean RenderVertexBuffer( GLcontext *ctx, GLboolean allDone ); -void RenderOneTriangle( GLcontext *ctx, GLuint v1, GLuint v2, GLuint v3, GLuint pv ); -void RenderOneLine( GLcontext *ctx, GLuint v1, GLuint v2, GLuint pv ); -GLbitfield ClearBuffersD3D( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ); - -/* Texture Management hooks. */ -static void TextureBind( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj ); -static void TextureLoad( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj, GLint level, GLint internalFormat, const struct gl_texture_image *image ); -static void TextureSubImage( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLint internalFormat, const struct gl_texture_image *image ); -/*===========================================================================*/ -/* Global variables. */ -/*===========================================================================*/ -D3DMESACONTEXT *pD3DCurrent, - *pD3DDefault; /* Thin support context. */ - -struct __extensions__ ext[] = { - - { (PROC)glPolygonOffsetEXT, "glPolygonOffsetEXT" }, - { (PROC)glBlendEquationEXT, "glBlendEquationEXT" }, - { (PROC)glBlendColorEXT, "glBlendColorExt" }, - { (PROC)glVertexPointerEXT, "glVertexPointerEXT" }, - { (PROC)glNormalPointerEXT, "glNormalPointerEXT" }, - { (PROC)glColorPointerEXT, "glColorPointerEXT" }, - { (PROC)glIndexPointerEXT, "glIndexPointerEXT" }, - { (PROC)glTexCoordPointerEXT, "glTexCoordPointer" }, - { (PROC)glEdgeFlagPointerEXT, "glEdgeFlagPointerEXT" }, - { (PROC)glGetPointervEXT, "glGetPointervEXT" }, - { (PROC)glArrayElementEXT, "glArrayElementEXT" }, - { (PROC)glDrawArraysEXT, "glDrawArrayEXT" }, - { (PROC)glAreTexturesResidentEXT, "glAreTexturesResidentEXT" }, - { (PROC)glBindTextureEXT, "glBindTextureEXT" }, - { (PROC)glDeleteTexturesEXT, "glDeleteTexturesEXT" }, - { (PROC)glGenTexturesEXT, "glGenTexturesEXT" }, - { (PROC)glIsTextureEXT, "glIsTextureEXT" }, - { (PROC)glPrioritizeTexturesEXT, "glPrioritizeTexturesEXT" }, - { (PROC)glCopyTexSubImage3DEXT, "glCopyTexSubImage3DEXT" }, - { (PROC)glTexImage3DEXT, "glTexImage3DEXT" }, - { (PROC)glTexSubImage3DEXT, "glTexSubImage3DEXT" }, -}; - -int qt_ext = sizeof(ext) / sizeof(ext[0]); -float g_DepthScale, - g_MaxDepth; -/*===========================================================================*/ -/* When a process loads this DLL we will setup the linked list for context */ -/* management and create a default context that will support the API until */ -/* the user creates and binds thier own. This THIN default context is useful*/ -/* to have around. */ -/* When the process terminates we will clean up all resources here. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY DllMain( HINSTANCE hInst, DWORD reason, LPVOID reserved ) -{ - switch( reason ) - { - case DLL_PROCESS_ATTACH: - return InitOpenGL( hInst ); - - case DLL_PROCESS_DETACH: - return TermOpenGL( hInst ); - } - - return TRUE; -} -/*===========================================================================*/ -/* The first thing we do when this dll is hit is connect to the dll that has*/ -/* handles all the DirectX 6 rendering. I decided to use another dll as DX6 */ -/* is all C++ and Mesa-3.0 is C (thats a good thing). This way I can write */ -/* the DX6 in C++ and Mesa-3.0 in C without having to worry about linkage. */ -/* I feel this is easy and better then using static wrappers as it is likely */ -/* faster and it allows me to just develope the one without compiling the */ -/* other. */ -/* NOTE that at this point we don't have much other than a very thin context*/ -/* that will support the API calls only to the point of not causing the app */ -/* to crash from the API table being empty. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -static BOOL InitOpenGL( HINSTANCE hInst ) -{ - /* Allocate and clear the default context. */ - pD3DDefault = (PD3DMESACONTEXT)ALLOC( sizeof(D3DMESACONTEXT) ); - if ( pD3DDefault == NULL ) - return FALSE; - memset( pD3DDefault, 0, sizeof(D3DMESACONTEXT) ); - - /* Clear the D3D vertex buffer so that values not used will be zero. This */ - /* save me from some redundant work. */ - memset( &D3DTLVertices, 0, sizeof(D3DTLVertices) ); - - /* Update the link. We uses a circular list so that it is easy to */ - /* add and search. This context will also be used for head and tail.*/ - pD3DDefault->next = pD3DDefault; - - /*========================================================================*/ - /* Do all core Mesa stuff. */ - /*========================================================================*/ - pD3DDefault->gl_visual = _mesa_create_visual( TRUE, - FALSE, /* db_flag */ - GL_FALSE, /* stereo */ - 8,8,8,8, /* r, g, b, a bits */ - 0, /* index bits */ - 16, /* depth_bits */ - 8, /* stencil_bits */ - 8,8,8,8, /* accum_bits */ - 1 ); - - if ( pD3DDefault->gl_visual == NULL) - { - FREE( pD3DDefault ); - return FALSE; - } - - /* Allocate a new Mesa context */ - pD3DDefault->gl_ctx = _mesa_create_context( pD3DDefault->gl_visual, NULL, pD3DDefault, GL_TRUE ); - if ( pD3DDefault->gl_ctx == NULL ) - { - _mesa_destroy_visual( pD3DDefault->gl_visual ); - FREE( pD3DDefault ); - return FALSE; - } - - /* Allocate a new Mesa frame buffer */ - pD3DDefault->gl_buffer = _mesa_create_framebuffer( pD3DDefault->gl_visual ); - if ( pD3DDefault->gl_buffer == NULL ) - { - _mesa_destroy_visual( pD3DDefault->gl_visual ); - _mesa_destroy_context( pD3DDefault->gl_ctx ); - FREE( pD3DDefault ); - return FALSE; - } - SetupDDPointers( pD3DDefault->gl_ctx ); - _mesa_make_current( pD3DDefault->gl_ctx, pD3DDefault->gl_buffer ); - - return TRUE; -} -/*===========================================================================*/ -/* This function will create a new D3D context but will not create the D3D */ -/* surfaces or even an instance of D3D (see at GetBufferSize). The only stuff*/ -/* done here is the internal Mesa stuff and some Win32 handles. */ -/*===========================================================================*/ -/* RETURN: casted pointer to the context, NULL. */ -/*===========================================================================*/ -HGLRC APIENTRY wglCreateContext( HDC hdc ) -{ - D3DMESACONTEXT *pNewContext; - DWORD dwCoopFlags = DDSCL_NORMAL; - RECT rectClient; - POINT pt; - - /* ALLOC and clear the new context. */ - pNewContext = (PD3DMESACONTEXT)ALLOC( sizeof(D3DMESACONTEXT) ); - if ( pNewContext == NULL ) - { - SetLastError( 0 ); - return (HGLRC)NULL; - } - memset( pNewContext, 0, sizeof(D3DMESACONTEXT) ); - - /*========================================================================*/ - /* Do all core Mesa stuff. */ - /*========================================================================*/ - - /* TODO: support more then one visual. */ - pNewContext->gl_visual = _mesa_create_visual( TRUE, - TRUE, /* db_flag */ - GL_FALSE, /* stereo */ - 8,8,8,8, /* r, g, b, a bits */ - 0, /* index bits */ - 16, /* depth_bits */ - 8, /* stencil_bits */ - 16,16,16,16,/* accum_bits */ - 1 ); - if ( pNewContext->gl_visual == NULL) - { - FREE( pNewContext ); - SetLastError( 0 ); - return (HGLRC)NULL; - } - - /* Allocate a new Mesa context */ - pNewContext->gl_ctx = _mesa_create_context( pNewContext->gl_visual, NULL, pNewContext, GL_TRUE ); - if ( pNewContext->gl_ctx == NULL ) - { - _mesa_destroy_visual( pNewContext->gl_visual ); - FREE( pNewContext ); - SetLastError( 0 ); - return (HGLRC)NULL; - } - - /* Allocate a new Mesa frame buffer */ - pNewContext->gl_buffer = _mesa_create_framebuffer( pNewContext->gl_visual ); - if ( pNewContext->gl_buffer == NULL ) - { - _mesa_destroy_visual( pNewContext->gl_visual ); - _mesa_destroy_context( pNewContext->gl_ctx ); - FREE( pNewContext ); - SetLastError( 0 ); - return (HGLRC)NULL; - } - - /*========================================================================*/ - /* Do all the driver stuff. */ - /*========================================================================*/ - pNewContext->hdc = hdc; - pNewContext->next = pD3DDefault->next; - pD3DDefault->next = pNewContext; /* Add to circular list. */ - - /* Create the HAL for the new context. */ - pNewContext->pShared = InitHAL( WindowFromDC(hdc) ); - - return (HGLRC)pNewContext; -} -/*===========================================================================*/ -/* This is a wrapper function that is supported by MakeCurrent. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY wglMakeCurrent( HDC hdc, HGLRC hglrc ) -{ - return MakeCurrent((D3DMESACONTEXT *)hglrc); -} -/*===========================================================================*/ -/* MakeCurrent will unbind whatever context is current (if any) & then bind */ -/* the supplied context. A context that is bound has it's window proc hooked*/ -/* with the wglMonitorProc and the context pointer is saved in pD3DCurrent. */ -/* Once the context is bound we update the Mesa-3.0 hooks (SetDDPointers) and*/ -/* the viewport (Mesa-.30 and DX6). */ -/* */ -/* TODO: this function can't fail. */ -/*===========================================================================*/ -/* RETURN: TRUE */ -/*===========================================================================*/ -static BOOL MakeCurrent( D3DMESACONTEXT *pContext ) -{ - D3DMESACONTEXT *pNext; - - /*====================================================================*/ - /* This is a special case that is a request to have no context bound. */ - /*====================================================================*/ - if ( pContext == NULL ) - { - /* Walk the whole list. We start and end at the Default context. */ - for( pNext = pD3DDefault->next; pNext != pD3DDefault; pNext = pNext->next ) - UnBindWindow( pNext ); - - return TRUE; - } - - /*=================================================*/ - /* Make for a fast redundant use of this function. */ - /*=================================================*/ - if ( pD3DCurrent == pContext ) - return TRUE; - - /*=============================*/ - /* Unbind the current context. */ - /*=============================*/ - UnBindWindow( pD3DCurrent ); - - /*=====================================*/ - /* Let Mesa-3.0 we have a new context. */ - /*=====================================*/ - SetupDDPointers( pContext->gl_ctx ); - _mesa_make_current( pContext->gl_ctx, pContext->gl_buffer ); - - /* We are done so set the internal current context. */ - if ( pContext != pD3DDefault ) - { - ResizeContext( pContext->gl_ctx ); - pContext->hOldProc = (WNDPROC)GetWindowLong( pContext->pShared->hwnd, GWL_WNDPROC ); - SetWindowLong( pContext->pShared->hwnd, GWL_WNDPROC, (LONG)wglMonitorProc ); - } - pD3DCurrent = pContext; - - return TRUE; -} -/*===========================================================================*/ -/* This function will only return the current window size. I have re-done */ -/* this function so that it doesn't check the current size and react to it as*/ -/* I should be able to have all the react code in the WM_SIZE message. The */ -/* old version would check the current window size and create/resize the HAL */ -/* surfaces if they have changed. I needed to delay the creation if the */ -/* surfaces because sometimes I wouldn't have a window size so this is where */ -/* I delayed it. If you are reading this then all went ok! */ -/* The default context will return a zero sized window and I'm not sure if */ -/* this is ok at this point (TODO). */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void GetBufferSize( GLcontext *ctx, GLuint *width, GLuint *height ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - - /* Fall through for the default because that is one of the uses for it. */ - if ( pContext == pD3DDefault ) - { - *width = 0; - *height = 0; - } - else - { - *width = pContext->pShared->dwWidth; - *height = pContext->pShared->dwHeight; - } -} -/*===========================================================================*/ -/* */ -/* */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static BOOL ResizeContext( GLcontext *ctx ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx, - *pCurrentTemp; - RECT rectClient; - POINT pt; - DWORD dwWidth, - dwHeight; - static BOOL bDDrawLock = FALSE; - - /* Make sure we have some values. */ - if ( (pContext->hdc == NULL ) || - (pContext->pShared->hwnd != WindowFromDC(pContext->hdc)) || - (pContext == pD3DDefault) ) - return FALSE; - - /* Having problems with DDraw sending resize messages before I was done. */ - if( bDDrawLock == TRUE ) - return FALSE; - - // TODO: don't think I need this anymore. - pCurrentTemp = pD3DCurrent; - pD3DCurrent = pD3DDefault; - bDDrawLock = TRUE; - - /* Get the current window dimentions. */ - UpdateScreenPosHAL( pContext->pShared ); - dwWidth = pContext->pShared->rectW.right - pContext->pShared->rectW.left; - dwHeight = pContext->pShared->rectW.bottom - pContext->pShared->rectW.top; - - /* Is the size of the OffScreen Render different? */ - if ( (dwWidth != pContext->pShared->dwWidth) || (dwHeight != pContext->pShared->dwHeight) ) - { - /* Create all the D3D surfaces and device. */ - CreateHAL( pContext->pShared ); - - /* I did this so that software rendering would still work as */ - /* I don't need to scale the z values twice. */ - g_DepthScale = (pContext->pShared->bHardware) ? 1.0 : ((float)0x00FFFFFF); - g_MaxDepth = (pContext->pShared->bHardware) ? 1.0 : ((float)0x00FFFFFF); - gl_DepthRange( pContext->gl_ctx, ctx->Viewport.Near, ctx->Viewport.Far ); - - /* Make sure we have a viewport. */ - gl_Viewport( pContext->gl_ctx, 0, 0, dwWidth, dwHeight ); - - /* Update Mesa as we might have changed from SW <-> HW. */ - SetupDDPointers( pContext->gl_ctx ); - _mesa_make_current( pContext->gl_ctx, pContext->gl_buffer ); - - /* If we are in HW we need to load the current texture if there is one already. */ - // if ( (ctx->Texture.Set[ctx->Texture.CurrentSet].Current != NULL) && - // (pContext->pShared->bHardware == TRUE) ) - // { - // CreateTMgrHAL( pContext->pShared, - // ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Name, - // 0, - // ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Image[0]->Format, - // (RECT *)NULL, - // ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Image[0]->Width, - // ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Image[0]->Height, - // TM_ACTION_BIND, - // (void *)ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Image[0]->Data ); - // } - } - - // TODO: don't think I need this anymore. - pD3DCurrent = pCurrentTemp; - bDDrawLock = FALSE; - - return TRUE; -} - -/*===========================================================================* -/* This function will Blt the render buffer to the PRIMARY surface. I repeat*/ -/* this code for the other SwapBuffer like functions and the flush (didn't */ -/* want the function calling overhead). Thsi could have been a macro... */ -/* */ -/* TODO: there are some problems with viewport/scissoring. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY wglSwapBuffers( HDC hdc ) -{ - /* Fall through for the default because that is one of the uses for it. */ - if ( pD3DCurrent == pD3DDefault ) - return FALSE; - - SwapBuffersHAL( pD3DCurrent->pShared ); - - return TRUE; -} -/*===========================================================================*/ -/* Same as wglSwapBuffers. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY SwapBuffers( HDC hdc ) -{ - /* Fall through for the default because that is one of the uses for it. */ - if ( pD3DCurrent == pD3DDefault ) - return FALSE; - - SwapBuffersHAL( pD3DCurrent->pShared ); - - return TRUE; -} -/*===========================================================================*/ -/* This should be ok as none of the SwapBuffers will cause a redundant Blt */ -/* as none of my Swap functions will call flush. This should also allow */ -/* sinlge buffered applications to work (not really worried though). Some */ -/* applications may flush then swap but then this is there fault IMHO. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void Flush( GLcontext *ctx ) -{ - /* Fall through for the default because that is one of the uses for it. */ - if ( pD3DCurrent == pD3DDefault ) - return; - - SwapBuffersHAL( pD3DCurrent->pShared ); -} -/*===========================================================================*/ -/* For now this function will ignore the supplied PF. If I'm going to allow */ -/* the user to choice the mode and device at startup I'm going to have to do */ -/* something different. */ -/* */ -/* TODO: use the linked list of modes to build a pixel format to be returned */ -/* to the caller. */ -/*===========================================================================*/ -/* RETURN: 1. */ -/*===========================================================================*/ -int APIENTRY wglChoosePixelFormat( HDC hdc, CONST PIXELFORMATDESCRIPTOR *ppfd ) -{ - return 1; -} -/*===========================================================================*/ -/* See wglChoosePixelFormat. */ -/*===========================================================================*/ -/* RETURN: 1. */ -/*===========================================================================*/ -int APIENTRY ChoosePixelFormat( HDC hdc, CONST PIXELFORMATDESCRIPTOR *ppfd ) -{ - return wglChoosePixelFormat(hdc,ppfd); -} -/*===========================================================================*/ -/* This function (for now) returns a static PF everytime. This is just to */ -/* allow things to continue. */ -/*===========================================================================*/ -/* RETURN: 1. */ -/*===========================================================================*/ -int APIENTRY wglDescribePixelFormat( HDC hdc, int iPixelFormat, UINT nBytes, LPPIXELFORMATDESCRIPTOR ppfd ) -{ - static PIXELFORMATDESCRIPTOR pfd = - { - sizeof(PIXELFORMATDESCRIPTOR), /* size */ - 1, /* version */ - PFD_SUPPORT_OPENGL | - PFD_DRAW_TO_WINDOW | - PFD_DOUBLEBUFFER, /* support double-buffering */ - PFD_TYPE_RGBA, /* color type */ - 16, /* prefered color depth */ - 0, 0, 0, 0, 0, 0, /* color bits (ignored) */ - 0, /* no alpha buffer */ - 0, /* alpha bits (ignored) */ - 0, /* no accumulation buffer */ - 0, 0, 0, 0, /* accum bits (ignored) */ - 16, /* depth buffer */ - 0, /* no stencil buffer */ - 0, /* no auxiliary buffers */ - PFD_MAIN_PLANE, /* main layer */ - 0, /* reserved */ - 0, 0, 0, /* no layer, visible, damage masks */ - }; - - /* Return the address of this static PF if one was requested. */ - if ( ppfd != NULL ) - memcpy( ppfd, &pfd, sizeof(PIXELFORMATDESCRIPTOR) ); - - return 1; -} -/*===========================================================================*/ -/* See wglDescribePixelFormat. */ -/*===========================================================================*/ -/* RETURN: 1. */ -/*===========================================================================*/ -int APIENTRY DescribePixelFormat( HDC hdc, int iPixelFormat, UINT nBytes, LPPIXELFORMATDESCRIPTOR ppfd ) -{ - return wglDescribePixelFormat(hdc,iPixelFormat,nBytes,ppfd); -} -/*===========================================================================*/ -/* This function will always return 1 for now. Just to allow for support. */ -/*===========================================================================*/ -/* RETURN: 1. */ -/*===========================================================================*/ -int APIENTRY wglGetPixelFormat( HDC hdc ) -{ - return 1; -} -/*===========================================================================*/ -/* See wglGetPixelFormat. */ -/*===========================================================================*/ -/* RETURN: 1. */ -/*===========================================================================*/ -int APIENTRY GetPixelFormat( HDC hdc ) -{ - return wglGetPixelFormat(hdc); -} -/*===========================================================================*/ -/* This will aways work for now. */ -/*===========================================================================*/ -/* RETURN: TRUE. */ -/*===========================================================================*/ -BOOL APIENTRY wglSetPixelFormat( HDC hdc, int iPixelFormat, CONST PIXELFORMATDESCRIPTOR *ppfd ) -{ - return TRUE; -} -/*===========================================================================*/ -/* See wglSetPixelFormat. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY SetPixelFormat( HDC hdc, int iPixelFormat, CONST PIXELFORMATDESCRIPTOR *ppfd ) -{ - return wglSetPixelFormat(hdc,iPixelFormat,ppfd); -} -/*===========================================================================*/ -/* This is a wrapper function that is supported by my own internal function.*/ -/* that takes my own D3D Mesa context structure. This so I can reuse the */ -/* function (no need for speed). */ -/*===========================================================================*/ -/* RETURN: TRUE. */ -/*===========================================================================*/ -BOOL APIENTRY wglDeleteContext( HGLRC hglrc ) -{ - DestroyContext( (D3DMESACONTEXT *)hglrc ); - - return TRUE; -} -/*===========================================================================*/ -/* Simple getter function that uses a cast. */ -/*===========================================================================*/ -/* RETURN: casted pointer to the context, NULL. */ -/*===========================================================================*/ -HGLRC APIENTRY wglGetCurrentContext( VOID ) -{ - return (pD3DCurrent) ? (HGLRC)pD3DCurrent : (HGLRC)NULL; -} -/*===========================================================================*/ -/* No support. */ -/*===========================================================================*/ -/* RETURN: FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY wglCopyContext( HGLRC hglrcSrc, HGLRC hglrcDst, UINT mask ) -{ - SetLastError( 0 ); - return FALSE; -} -/*===========================================================================*/ -/* No support. */ -/*===========================================================================*/ -/* RETURN: NULL. */ -/*===========================================================================*/ -HGLRC APIENTRY wglCreateLayerContext( HDC hdc,int iLayerPlane ) -{ - SetLastError( 0 ); - return (HGLRC)NULL; -} -/*===========================================================================*/ -/* Simple getter function. */ -/*===========================================================================*/ -/* RETURN: FALSE. */ -/*===========================================================================*/ -HDC APIENTRY wglGetCurrentDC( VOID ) -{ - return (pD3DCurrent) ? pD3DCurrent->hdc : (HDC)NULL; -} -/*===========================================================================*/ -/* Simply call that searches the supported extensions for a match & returns */ -/* the pointer to the function that lends support. */ -/*===========================================================================*/ -/* RETURN: pointer to API call, NULL. */ -/*===========================================================================*/ -PROC APIENTRY wglGetProcAddress( LPCSTR lpszProc ) -{ - int index; - - for( index = 0; index < qt_ext; index++ ) - if( !strcmp(lpszProc,ext[index].name) ) - return ext[index].proc; - - SetLastError( 0 ); - return NULL; -} -/*===========================================================================*/ -/* No support. */ -/*===========================================================================*/ -/* RETURN: FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY wglShareLists( HGLRC hglrc1, HGLRC hglrc2 ) -{ - SetLastError( 0 ); - return FALSE; -} -/*===========================================================================*/ -/* No support. */ -/*===========================================================================*/ -/* RETURN: FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY wglUseFontBitmaps( HDC fontDevice, DWORD firstChar, DWORD numChars, DWORD listBase ) -{ - SetLastError( 0 ); - return FALSE; -} -/*===========================================================================*/ -/* No support. */ -/*===========================================================================*/ -/* RETURN: FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY wglUseFontBitmapsW( HDC hdc,DWORD first,DWORD count,DWORD listBase ) -{ - SetLastError( 0 ); - return FALSE; -} -/*===========================================================================*/ -/* No support. */ -/*===========================================================================*/ -/* RETURN: FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY wglUseFontOutlinesA( HDC hdc, DWORD first, DWORD count, DWORD listBase, FLOAT deviation, FLOAT extrusion, int format, LPGLYPHMETRICSFLOAT lpgmf ) -{ - SetLastError( 0 ); - return FALSE; -} -/*===========================================================================*/ -/* No support. */ -/*===========================================================================*/ -/* RETURN: FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY wglUseFontOutlinesW( HDC hdc,DWORD first,DWORD count, DWORD listBase,FLOAT deviation, FLOAT extrusion,int format, LPGLYPHMETRICSFLOAT lpgmf ) -{ - SetLastError( 0 ); - return FALSE ; -} -/*===========================================================================*/ -/* No support. */ -/*===========================================================================*/ -/* RETURN: FALSE. */ -/*===========================================================================*/ -BOOL APIENTRY wglSwapLayerBuffers( HDC hdc, UINT fuPlanes ) -{ - SetLastError( 0 ); - return FALSE; -} -/*===========================================================================*/ -/* This function will be hooked into the window that has been bound. Right */ -/* now it is used to track the window size and position. Also the we clean */ -/* up the currrent context when the window is close/destroyed. */ -/* */ -/* TODO: there might be something wrong here as some games (Heretic II) don't*/ -/* track the window quit right. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -LONG APIENTRY wglMonitorProc( HWND hwnd, UINT message, UINT wParam, LONG lParam ) -{ - WNDPROC hOldProc; - GLint width, - height; - - switch( message ) - { -// case WM_PAINT: -// break; -// case WM_ACTIVATE: -// break; -// case WM_SHOWWINDOW: -// break; - - case UM_FATALSHUTDOWN: - /* Support the API until we die... */ - MakeCurrent( pD3DDefault ); - break; - - case WM_MOVE: - case WM_DISPLAYCHANGE: - case WM_SIZE: - ResizeContext( pD3DCurrent->gl_ctx ); - break; - - case WM_CLOSE: - case WM_DESTROY: - /* Support the API until we die... */ - hOldProc = pD3DCurrent->hOldProc; - DestroyContext( pD3DCurrent ); - return (hOldProc)(hwnd,message,wParam,lParam); - } - - return (pD3DCurrent->hOldProc)(hwnd,message,wParam,lParam); -} - -/**********************************************************************/ -/***** Miscellaneous device driver funcs *****/ -/**********************************************************************/ - -/*===========================================================================*/ -/* Not reacting to this as I'm only supporting drawing to the back buffer */ -/* right now. */ -/*===========================================================================*/ -/* RETURN: TRUE. */ -/*===========================================================================*/ -static GLboolean SetBuffer( GLcontext *ctx, GLenum buffer ) -{ - if (buffer == GL_BACK_LEFT) - return GL_TRUE; - else - return GL_FALSE; -} -/*===========================================================================*/ -/* This proc will be called by Mesa when the viewport has been set. So if */ -/* we have a context and it isn't the default then we should let D3D know of */ -/* the change. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void SetViewport( GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - RECT rect; - - /* Make sure we can set a viewport. */ - if ( pContext->pShared && (pContext != pD3DDefault) ) - { - // TODO: might be needed. - UpdateScreenPosHAL( pContext->pShared ); - rect.left = x; - rect.right = x + w; - rect.top = y; - rect.bottom = y + h; - - // TODO: shared struct should make this call smaller - SetViewportHAL( pContext->pShared, &rect, 0.0F, 1.0F ); - } -} -/*===========================================================================*/ -/* This function could be better I guess but I decided just to grab the four*/ -/* components and store then seperately. Makes it easier to use IMHO. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void ClearColor( GLcontext *ctx, GLubyte r, GLubyte g, GLubyte b, GLubyte a ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - - pContext->aClear = a; - pContext->bClear = b; - pContext->gClear = g; - pContext->rClear = r; -} -/*===========================================================================*/ -/* This function could be better I guess but I decided just to grab the four*/ -/* components and store then seperately. Makes it easier to use IMHO. */ -/* (is there an echo in here?) */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void SetColor( GLcontext *ctx, GLubyte r, GLubyte g, GLubyte b, GLubyte a ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - - pContext->aCurrent = a; - pContext->bCurrent = b; - pContext->gCurrent = g; - pContext->rCurrent = r; -} -/*===========================================================================*/ -/* */ -/* */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static const char *RendererString( void ) -{ - static char pszRender[64]; - - strcpy( pszRender, "altD3D " ); - - if ( pD3DCurrent->pShared->bHardware ) - strcat( pszRender, "(HW)"); - else - strcat( pszRender, "(SW)"); - - return (const char *)pszRender; -} -/*===========================================================================*/ -/* This function will choose which set of pointers Mesa will use based on */ -/* whether we hard using hardware or software. I have added another set of */ -/* pointers that will do nothing but stop the API from crashing. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void SetupDDPointers( GLcontext *ctx ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - - // TODO: write a generic NULL support for the span render. - if ( pContext->pShared && pContext->pShared->bHardware ) - { - ctx->Driver.UpdateState = SetupHWDDPointers; - } - else if ( pContext == pD3DDefault ) - { - ctx->Driver.UpdateState = SetupNULLDDPointers; - } - else - { - ctx->Driver.UpdateState = SetupSWDDPointers; - } -} -/*===========================================================================*/ -/* This function will populate all the Mesa driver hooks. This version of */ -/* hooks will do nothing but support the API when we don't have a valid */ -/* context bound. This is mostly for applications that don't behave right */ -/* and also to help exit as clean as possable when we have a FatalError. */ -/*===========================================================================*/ -/* RETURN: pointer to the specific function. */ -/*===========================================================================*/ -static void SetupNULLDDPointers( GLcontext *ctx ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - - /* Initialize all the pointers in the DD struct. Do this whenever */ - /* a new context is made current or we change buffers via set_buffer! */ - ctx->Driver.UpdateState = SetupNULLDDPointers; - - /* State management hooks. */ - ctx->Driver.Color = NULLSetColor; - ctx->Driver.ClearColor = NULLClearColor; - ctx->Driver.Clear = NULLClearBuffers; - ctx->Driver.SetBuffer = NULLSetBuffer; - - /* Window management hooks. */ - ctx->Driver.GetBufferSize = NULLGetBufferSize; - - /* Primitive rendering hooks. */ - ctx->Driver.TriangleFunc = NULL; - ctx->Driver.RenderVB = NULL; - - /* Pixel/span writing functions: */ - ctx->Driver.WriteRGBASpan = NULLWrSpRGBA; - ctx->Driver.WriteRGBSpan = NULLWrSpRGB; - ctx->Driver.WriteMonoRGBASpan = NULLWrSpRGBAMono; - ctx->Driver.WriteRGBAPixels = NULLWrPiRGBA; - ctx->Driver.WriteMonoRGBAPixels = NULLWrPiRGBAMono; - - /* Pixel/span reading functions: */ - ctx->Driver.ReadRGBASpan = NULLReSpRGBA; - ctx->Driver.ReadRGBAPixels = NULLRePiRGBA; - - /* Misc. hooks. */ - ctx->Driver.RendererString = RendererString; -} -/*===========================================================================*/ -/* This function will populate all the Mesa driver hooks. There are two of */ -/* these functions. One if we have hardware support and one is there is only*/ -/* software. These functions will be called by Mesa and by the wgl.c when we*/ -/* have resized (or created) the buffers. The thing is that if a window gets*/ -/* resized we may loose hardware support or gain it... */ -/*===========================================================================*/ -/* RETURN: pointer to the specific function. */ -/*===========================================================================*/ -static void SetupSWDDPointers( GLcontext *ctx ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - - /* Initialize all the pointers in the DD struct. Do this whenever */ - /* a new context is made current or we change buffers via set_buffer! */ - ctx->Driver.UpdateState = SetupSWDDPointers; - - /* State management hooks. */ - ctx->Driver.Color = SetColor; - ctx->Driver.ClearColor = ClearColor; - ctx->Driver.Clear = ClearBuffers; - ctx->Driver.SetBuffer = SetBuffer; - - /* Window management hooks. */ - ctx->Driver.GetBufferSize = GetBufferSize; - ctx->Driver.Viewport = SetViewport; - - /* Primitive rendering hooks. */ - ctx->Driver.TriangleFunc = NULL; - ctx->Driver.RenderVB = NULL; - - /* Texture management hooks. */ - - /* Pixel/span writing functions: */ - ctx->Driver.WriteRGBASpan = WSpanRGBA; - ctx->Driver.WriteRGBSpan = WSpanRGB; - ctx->Driver.WriteMonoRGBASpan = WSpanRGBAMono; - ctx->Driver.WriteRGBAPixels = WPixelsRGBA; - ctx->Driver.WriteMonoRGBAPixels = WPixelsRGBAMono; - - /* Pixel/span reading functions: */ - ctx->Driver.ReadRGBASpan = RSpanRGBA; - ctx->Driver.ReadRGBAPixels = RPixelsRGBA; - - /* Misc. hooks. */ - ctx->Driver.Flush = Flush; - ctx->Driver.RendererString = RendererString; -} -/*===========================================================================*/ -/* This function will populate all the Mesa driver hooks. There are two of */ -/* these functions. One if we have hardware support and one is there is only*/ -/* software. These functions will be called by Mesa and by the wgl.c when we*/ -/* have resized (or created) the buffers. The thing is that if a window gets*/ -/* resized we may loose hardware support or gain it... */ -/*===========================================================================*/ -/* RETURN: pointer to the specific function. */ -/*===========================================================================*/ -static void SetupHWDDPointers( GLcontext *ctx ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - - /* Initialize all the pointers in the DD struct. Do this whenever */ - /* a new context is made current or we change buffers via set_buffer! */ - ctx->Driver.UpdateState = SetupHWDDPointers; - - /* State management hooks. */ - ctx->Driver.Color = SetColor; - ctx->Driver.ClearColor = ClearColor; - ctx->Driver.Clear = ClearBuffersD3D; - ctx->Driver.SetBuffer = SetBuffer; - - /* Window management hooks. */ - ctx->Driver.GetBufferSize = GetBufferSize; - ctx->Driver.Viewport = SetViewport; - - /* Primitive rendering hooks. */ - ctx->Driver.TriangleFunc = RenderOneTriangle; - ctx->Driver.LineFunc = RenderOneLine; - ctx->Driver.RenderVB = RenderVertexBuffer; - - /* Pixel/span writing functions: */ - ctx->Driver.WriteRGBASpan = WSpanRGBA; - ctx->Driver.WriteRGBSpan = WSpanRGB; - ctx->Driver.WriteMonoRGBASpan = WSpanRGBAMono; - ctx->Driver.WriteRGBAPixels = WPixelsRGBA; - ctx->Driver.WriteMonoRGBAPixels = WPixelsRGBAMono; - - /* Pixel/span reading functions: */ - ctx->Driver.ReadRGBASpan = RSpanRGBA; - ctx->Driver.ReadRGBAPixels = RPixelsRGBA; - - /* Texture management hooks. */ - // ctx->Driver.BindTexture = TextureBind; - ctx->Driver.TexImage = TextureLoad; - ctx->Driver.TexSubImage = TextureSubImage; - - /* Misc. hooks. */ - ctx->Driver.Flush = Flush; - ctx->Driver.RendererString = RendererString; -} -/*===========================================================================*/ -/* This function will release all resources used by the DLL. Every context */ -/* will be clobbered by releaseing all driver desources and then freeing the */ -/* context memory. Most all the work is done in DestroyContext. */ -/*===========================================================================*/ -/* RETURN: TRUE. */ -/*===========================================================================*/ -static BOOL TermOpenGL( HINSTANCE hInst ) -{ - D3DMESACONTEXT *pTmp, - *pNext; - - /* Just incase we are still getting paint msg. */ - MakeCurrent( pD3DDefault ); - - /* Walk the list until we get back to the default context. */ - for( pTmp = pD3DDefault->next; pTmp != pD3DDefault; pTmp = pNext ) - { - pNext = pTmp->next; - DestroyContext( pTmp ); - } - DestroyContext( pD3DDefault ); - - return TRUE; -} -/*===========================================================================*/ -/* This function is an internal function that will clean up all the Mesa */ -/* context bound to this D3D context. Also any D3D stuff that this context */ -/* uses will be unloaded. */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -static void DestroyContext( D3DMESACONTEXT *pContext ) -{ - D3DMESACONTEXT *pTmp; - - /* Walk the list until we find the context before this one. */ - for( pTmp = pD3DDefault; pTmp && (pTmp->next != pContext); pTmp = pTmp->next ) - if ( pTmp == pTmp->next ) - break; - - /* If we never found it it must already be deleted. */ - if ( pTmp->next != pContext ) - return; - - /* Make sure we are not using this context. */ - if ( pContext == pD3DCurrent ) - MakeCurrent( pD3DDefault ); - - /* Free the Mesa stuff. */ - if ( pContext->gl_visual ) - { - _mesa_destroy_visual( pContext->gl_visual ); - pContext->gl_visual = NULL; - } - if ( pContext->gl_buffer ) - { - _mesa_destroy_framebuffer( pContext->gl_buffer ); - pContext->gl_buffer = NULL; - } - if ( pContext->gl_ctx ) - { - _mesa_destroy_context( pContext->gl_ctx ); - pContext->gl_ctx = NULL; - } - - /* Now dump the D3D. */ - if ( pContext->pShared ) - TermHAL( pContext->pShared ); - - /* Update the previous context's link. */ - pTmp->next = pContext->next; - - /* Gonzo. */ - FREE( pContext ); -} -/*===========================================================================*/ -/* This function will pull the supplied context away from Win32. Basicly it*/ -/* will remove the hook from the window Proc. */ -/* */ -/* TODO: might want to serialize this stuff... */ -/*===========================================================================*/ -/* RETURN: TRUE, FALSE. */ -/*===========================================================================*/ -static BOOL UnBindWindow( D3DMESACONTEXT *pContext ) -{ - if ( pContext == NULL ) - return FALSE; - - if ( pContext == pD3DDefault ) - return TRUE; - - /* Make sure we always have a context bound. */ - if ( pContext == pD3DCurrent ) - pD3DCurrent = pD3DDefault; - - SetWindowLong( pContext->pShared->hwnd, GWL_WNDPROC, (LONG)pContext->hOldProc ); - pContext->hOldProc = NULL; - - return TRUE; -} -/*===========================================================================*/ -/* There are two cases that allow for a faster clear when we know that the */ -/* whole buffer is cleared and that there is no clipping. */ -/*===========================================================================*/ -/* RETURN: the original mask with the bits cleared that represents the buffer* -/* or buffers we just cleared. */ -/*===========================================================================*/ -GLbitfield ClearBuffersD3D( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - DWORD dwFlags = 0; - - if ( mask & GL_COLOR_BUFFER_BIT ) - { - dwFlags |= D3DCLEAR_TARGET; - mask &= ~GL_COLOR_BUFFER_BIT; - } - if ( mask & GL_DEPTH_BUFFER_BIT ) - { - dwFlags |= D3DCLEAR_ZBUFFER; - mask &= ~GL_DEPTH_BUFFER_BIT; - } - if ( dwFlags == 0 ) - return mask; - - ClearHAL( pContext->pShared, - dwFlags, - all, - x, y, - width, height, - ((pContext->aClear<<24) | (pContext->rClear<<16) | (pContext->gClear<<8) | (pContext->bClear)), - ctx->Depth.Clear, - 0 ); - - return mask; -} - - - -/*===========================================================================*/ -/* TEXTURE MANAGER: ok here is how I did textures. Mesa-3.0 will keep track*/ -/* of all the textures for us. So this means that at anytime we can go to */ -/* the Mesa context and get the current texture. With this in mind this is */ -/* what I did. I really don't care about what textures get or are loaded */ -/* until I actually have to draw a tri that is textured. At this point I */ -/* must have the texture so I demand the texture by destorying all other */ -/* texture surfaces if need be and load the current one. This allows for the*/ -/* best preformance on low memory cards as time is not wasted loading and */ -/* unload textures. */ -/*===========================================================================*/ - - - - - -/*===========================================================================*/ -/* TextureLoad will try and create a D3D surface from the supplied texture */ -/* object if its level 0 (first). The surface will be fully filled with the */ -/* texture. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void TextureLoad( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj, GLint level, GLint internalFormat, const struct gl_texture_image *image ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - - /* TODO: only doing first LOD. */ - if ( (ctx->DriverCtx == NULL) || (level != 0) ) - return; - - CreateTMgrHAL( pContext->pShared, - tObj->Name, - level, - tObj->Image[level]->Format, - (RECT *)NULL, - tObj->Image[level]->Width, - tObj->Image[level]->Height, - TM_ACTION_LOAD, - (void *)tObj->Image[level]->Data ); -} -/*===========================================================================*/ -/* TextureBind make sure that the texture is on the card. Thats it. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void TextureBind( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - - /* TODO: only doing first LOD. */ - if ( (tObj->Image[0] == NULL) || (ctx->DriverCtx == NULL) ) - return; - - CreateTMgrHAL( pContext->pShared, - tObj->Name, - 0, - tObj->Image[0]->Format, - (RECT *)NULL, - tObj->Image[0]->Width, - tObj->Image[0]->Height, - TM_ACTION_BIND, - (void *)tObj->Image[0]->Data ); -} -/*===========================================================================*/ -/* TextureSubImage will make sure that the texture being updated is updated */ -/* if its on the card. */ -/*===========================================================================*/ -/* RETURN: */ -/*===========================================================================*/ -static void TextureSubImage( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLint internalFormat, const struct gl_texture_image *image ) -{ - D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; - RECT rect; - - /* TODO: only doing first LOD. */ - if ( (ctx->DriverCtx == NULL) || (level > 0) ) - return; - - /* Create a dirty rectangle structure. */ - rect.left = xoffset; - rect.right = xoffset + width; - rect.top = yoffset; - rect.bottom = yoffset + height; - - CreateTMgrHAL( pContext->pShared, - tObj->Name, - 0, - tObj->Image[0]->Format, - &rect, - tObj->Image[0]->Width, - tObj->Image[0]->Height, - TM_ACTION_UPDATE, - (void *)tObj->Image[0]->Data ); -} - +/*===========================================================================*/ +/* */ +/* Mesa-3.0 Makefile for DirectX 6 */ +/* */ +/* By Leigh McRae */ +/* */ +/* http://www.altsoftware.com/ */ +/* */ +/* Copyright (c) 1998-1997 alt.software inc. All Rights Reserved */ +/*===========================================================================*/ +#include "D3DMesa.h" +/*===========================================================================*/ +/* Window managment. */ +/*===========================================================================*/ +static BOOL InitOpenGL( HINSTANCE hInst ); +static BOOL TermOpenGL( HINSTANCE hInst ); +static BOOL ResizeContext( GLcontext *ctx ); +static BOOL MakeCurrent( D3DMESACONTEXT *pContext ); +static void DestroyContext( D3DMESACONTEXT *pContext ); +static BOOL UnBindWindow( D3DMESACONTEXT *pContext ); +LONG APIENTRY wglMonitorProc( HWND hwnd, UINT message, UINT wParam, LONG lParam ); +/*===========================================================================*/ +/* Mesa hooks. */ +/*===========================================================================*/ +static void SetupDDPointers( GLcontext *ctx ); +static void SetupSWDDPointers( GLcontext *ctx ); +static void SetupHWDDPointers( GLcontext *ctx ); +static void SetupNULLDDPointers( GLcontext *ctx ); +static const char *RendererString( void ); + +/* State Management hooks. */ +static void SetColor( GLcontext *ctx, GLubyte r, GLubyte g, GLubyte b, GLubyte a ); +static void ClearColor( GLcontext *ctx, GLubyte r, GLubyte g, GLubyte b, GLubyte a ); +static GLboolean SetBuffer( GLcontext *ctx, GLenum buffer ); + +/* Window Management hooks. */ +static void GetBufferSize( GLcontext *ctx, GLuint *width, GLuint *height ); +static void SetViewport( GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h ); +static void Flush( GLcontext *ctx ); + +/* Span rendering hooks. */ +void WSpanRGB( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgb[][3], const GLubyte mask[] ); +void WSpanRGBA( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte rgba[][4], const GLubyte mask[] ); +void WSpanRGBAMono( const GLcontext* ctx, GLuint n, GLint x, GLint y, const GLubyte mask[] ); +void WPixelsRGBA( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte rgba[][4], const GLubyte mask[] ); +void WPixelsRGBAMono( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], const GLubyte mask[] ); +void RSpanRGBA( const GLcontext* ctx, GLuint n, GLint x, GLint y, GLubyte rgba[][4] ); +void RPixelsRGBA( const GLcontext* ctx, GLuint n, const GLint x[], const GLint y[], GLubyte rgba[][4], const GLubyte mask[] ); +GLbitfield ClearBuffers( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ); + +/* Primitve rendering hooks. */ +GLboolean RenderVertexBuffer( GLcontext *ctx, GLboolean allDone ); +void RenderOneTriangle( GLcontext *ctx, GLuint v1, GLuint v2, GLuint v3, GLuint pv ); +void RenderOneLine( GLcontext *ctx, GLuint v1, GLuint v2, GLuint pv ); +GLbitfield ClearBuffersD3D( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ); + +/* Texture Management hooks. */ +static void TextureBind( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj ); +static void TextureLoad( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj, GLint level, GLint internalFormat, const struct gl_texture_image *image ); +static void TextureSubImage( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLint internalFormat, const struct gl_texture_image *image ); +/*===========================================================================*/ +/* Global variables. */ +/*===========================================================================*/ +D3DMESACONTEXT *pD3DCurrent, + *pD3DDefault; /* Thin support context. */ + +struct __extensions__ ext[] = { + + { (PROC)glPolygonOffsetEXT, "glPolygonOffsetEXT" }, + { (PROC)glBlendEquationEXT, "glBlendEquationEXT" }, + { (PROC)glBlendColorEXT, "glBlendColorExt" }, + { (PROC)glVertexPointerEXT, "glVertexPointerEXT" }, + { (PROC)glNormalPointerEXT, "glNormalPointerEXT" }, + { (PROC)glColorPointerEXT, "glColorPointerEXT" }, + { (PROC)glIndexPointerEXT, "glIndexPointerEXT" }, + { (PROC)glTexCoordPointerEXT, "glTexCoordPointer" }, + { (PROC)glEdgeFlagPointerEXT, "glEdgeFlagPointerEXT" }, + { (PROC)glGetPointervEXT, "glGetPointervEXT" }, + { (PROC)glArrayElementEXT, "glArrayElementEXT" }, + { (PROC)glDrawArraysEXT, "glDrawArrayEXT" }, + { (PROC)glAreTexturesResidentEXT, "glAreTexturesResidentEXT" }, + { (PROC)glBindTextureEXT, "glBindTextureEXT" }, + { (PROC)glDeleteTexturesEXT, "glDeleteTexturesEXT" }, + { (PROC)glGenTexturesEXT, "glGenTexturesEXT" }, + { (PROC)glIsTextureEXT, "glIsTextureEXT" }, + { (PROC)glPrioritizeTexturesEXT, "glPrioritizeTexturesEXT" }, + { (PROC)glCopyTexSubImage3DEXT, "glCopyTexSubImage3DEXT" }, + { (PROC)glTexImage3DEXT, "glTexImage3DEXT" }, + { (PROC)glTexSubImage3DEXT, "glTexSubImage3DEXT" }, +}; + +int qt_ext = sizeof(ext) / sizeof(ext[0]); +float g_DepthScale, + g_MaxDepth; +/*===========================================================================*/ +/* When a process loads this DLL we will setup the linked list for context */ +/* management and create a default context that will support the API until */ +/* the user creates and binds thier own. This THIN default context is useful*/ +/* to have around. */ +/* When the process terminates we will clean up all resources here. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY DllMain( HINSTANCE hInst, DWORD reason, LPVOID reserved ) +{ + switch( reason ) + { + case DLL_PROCESS_ATTACH: + return InitOpenGL( hInst ); + + case DLL_PROCESS_DETACH: + return TermOpenGL( hInst ); + } + + return TRUE; +} +/*===========================================================================*/ +/* The first thing we do when this dll is hit is connect to the dll that has*/ +/* handles all the DirectX 6 rendering. I decided to use another dll as DX6 */ +/* is all C++ and Mesa-3.0 is C (thats a good thing). This way I can write */ +/* the DX6 in C++ and Mesa-3.0 in C without having to worry about linkage. */ +/* I feel this is easy and better then using static wrappers as it is likely */ +/* faster and it allows me to just develope the one without compiling the */ +/* other. */ +/* NOTE that at this point we don't have much other than a very thin context*/ +/* that will support the API calls only to the point of not causing the app */ +/* to crash from the API table being empty. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +static BOOL InitOpenGL( HINSTANCE hInst ) +{ + /* Allocate and clear the default context. */ + pD3DDefault = (PD3DMESACONTEXT)ALLOC( sizeof(D3DMESACONTEXT) ); + if ( pD3DDefault == NULL ) + return FALSE; + memset( pD3DDefault, 0, sizeof(D3DMESACONTEXT) ); + + /* Clear the D3D vertex buffer so that values not used will be zero. This */ + /* save me from some redundant work. */ + memset( &D3DTLVertices, 0, sizeof(D3DTLVertices) ); + + /* Update the link. We uses a circular list so that it is easy to */ + /* add and search. This context will also be used for head and tail.*/ + pD3DDefault->next = pD3DDefault; + + /*========================================================================*/ + /* Do all core Mesa stuff. */ + /*========================================================================*/ + pD3DDefault->gl_visual = _mesa_create_visual( TRUE, + FALSE, /* db_flag */ + GL_FALSE, /* stereo */ + 8,8,8,8, /* r, g, b, a bits */ + 0, /* index bits */ + 16, /* depth_bits */ + 8, /* stencil_bits */ + 8,8,8,8, /* accum_bits */ + 1 ); + + if ( pD3DDefault->gl_visual == NULL) + { + FREE( pD3DDefault ); + return FALSE; + } + + /* Allocate a new Mesa context */ + pD3DDefault->gl_ctx = _mesa_create_context( pD3DDefault->gl_visual, NULL, pD3DDefault, GL_TRUE ); + if ( pD3DDefault->gl_ctx == NULL ) + { + _mesa_destroy_visual( pD3DDefault->gl_visual ); + FREE( pD3DDefault ); + return FALSE; + } + + /* Allocate a new Mesa frame buffer */ + pD3DDefault->gl_buffer = _mesa_create_framebuffer( pD3DDefault->gl_visual ); + if ( pD3DDefault->gl_buffer == NULL ) + { + _mesa_destroy_visual( pD3DDefault->gl_visual ); + _mesa_destroy_context( pD3DDefault->gl_ctx ); + FREE( pD3DDefault ); + return FALSE; + } + SetupDDPointers( pD3DDefault->gl_ctx ); + _mesa_make_current( pD3DDefault->gl_ctx, pD3DDefault->gl_buffer ); + + return TRUE; +} +/*===========================================================================*/ +/* This function will create a new D3D context but will not create the D3D */ +/* surfaces or even an instance of D3D (see at GetBufferSize). The only stuff*/ +/* done here is the internal Mesa stuff and some Win32 handles. */ +/*===========================================================================*/ +/* RETURN: casted pointer to the context, NULL. */ +/*===========================================================================*/ +HGLRC APIENTRY wglCreateContext( HDC hdc ) +{ + D3DMESACONTEXT *pNewContext; + DWORD dwCoopFlags = DDSCL_NORMAL; + RECT rectClient; + POINT pt; + + /* ALLOC and clear the new context. */ + pNewContext = (PD3DMESACONTEXT)ALLOC( sizeof(D3DMESACONTEXT) ); + if ( pNewContext == NULL ) + { + SetLastError( 0 ); + return (HGLRC)NULL; + } + memset( pNewContext, 0, sizeof(D3DMESACONTEXT) ); + + /*========================================================================*/ + /* Do all core Mesa stuff. */ + /*========================================================================*/ + + /* TODO: support more then one visual. */ + pNewContext->gl_visual = _mesa_create_visual( TRUE, + TRUE, /* db_flag */ + GL_FALSE, /* stereo */ + 8,8,8,8, /* r, g, b, a bits */ + 0, /* index bits */ + 16, /* depth_bits */ + 8, /* stencil_bits */ + 16,16,16,16,/* accum_bits */ + 1 ); + if ( pNewContext->gl_visual == NULL) + { + FREE( pNewContext ); + SetLastError( 0 ); + return (HGLRC)NULL; + } + + /* Allocate a new Mesa context */ + pNewContext->gl_ctx = _mesa_create_context( pNewContext->gl_visual, NULL, pNewContext, GL_TRUE ); + if ( pNewContext->gl_ctx == NULL ) + { + _mesa_destroy_visual( pNewContext->gl_visual ); + FREE( pNewContext ); + SetLastError( 0 ); + return (HGLRC)NULL; + } + + /* Allocate a new Mesa frame buffer */ + pNewContext->gl_buffer = _mesa_create_framebuffer( pNewContext->gl_visual ); + if ( pNewContext->gl_buffer == NULL ) + { + _mesa_destroy_visual( pNewContext->gl_visual ); + _mesa_destroy_context( pNewContext->gl_ctx ); + FREE( pNewContext ); + SetLastError( 0 ); + return (HGLRC)NULL; + } + + /*========================================================================*/ + /* Do all the driver stuff. */ + /*========================================================================*/ + pNewContext->hdc = hdc; + pNewContext->next = pD3DDefault->next; + pD3DDefault->next = pNewContext; /* Add to circular list. */ + + /* Create the HAL for the new context. */ + pNewContext->pShared = InitHAL( WindowFromDC(hdc) ); + + return (HGLRC)pNewContext; +} +/*===========================================================================*/ +/* This is a wrapper function that is supported by MakeCurrent. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY wglMakeCurrent( HDC hdc, HGLRC hglrc ) +{ + return MakeCurrent((D3DMESACONTEXT *)hglrc); +} +/*===========================================================================*/ +/* MakeCurrent will unbind whatever context is current (if any) & then bind */ +/* the supplied context. A context that is bound has it's window proc hooked*/ +/* with the wglMonitorProc and the context pointer is saved in pD3DCurrent. */ +/* Once the context is bound we update the Mesa-3.0 hooks (SetDDPointers) and*/ +/* the viewport (Mesa-.30 and DX6). */ +/* */ +/* TODO: this function can't fail. */ +/*===========================================================================*/ +/* RETURN: TRUE */ +/*===========================================================================*/ +static BOOL MakeCurrent( D3DMESACONTEXT *pContext ) +{ + D3DMESACONTEXT *pNext; + + /*====================================================================*/ + /* This is a special case that is a request to have no context bound. */ + /*====================================================================*/ + if ( pContext == NULL ) + { + /* Walk the whole list. We start and end at the Default context. */ + for( pNext = pD3DDefault->next; pNext != pD3DDefault; pNext = pNext->next ) + UnBindWindow( pNext ); + + return TRUE; + } + + /*=================================================*/ + /* Make for a fast redundant use of this function. */ + /*=================================================*/ + if ( pD3DCurrent == pContext ) + return TRUE; + + /*=============================*/ + /* Unbind the current context. */ + /*=============================*/ + UnBindWindow( pD3DCurrent ); + + /*=====================================*/ + /* Let Mesa-3.0 we have a new context. */ + /*=====================================*/ + SetupDDPointers( pContext->gl_ctx ); + _mesa_make_current( pContext->gl_ctx, pContext->gl_buffer ); + + /* We are done so set the internal current context. */ + if ( pContext != pD3DDefault ) + { + ResizeContext( pContext->gl_ctx ); + pContext->hOldProc = (WNDPROC)GetWindowLong( pContext->pShared->hwnd, GWL_WNDPROC ); + SetWindowLong( pContext->pShared->hwnd, GWL_WNDPROC, (LONG)wglMonitorProc ); + } + pD3DCurrent = pContext; + + return TRUE; +} +/*===========================================================================*/ +/* This function will only return the current window size. I have re-done */ +/* this function so that it doesn't check the current size and react to it as*/ +/* I should be able to have all the react code in the WM_SIZE message. The */ +/* old version would check the current window size and create/resize the HAL */ +/* surfaces if they have changed. I needed to delay the creation if the */ +/* surfaces because sometimes I wouldn't have a window size so this is where */ +/* I delayed it. If you are reading this then all went ok! */ +/* The default context will return a zero sized window and I'm not sure if */ +/* this is ok at this point (TODO). */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void GetBufferSize( GLcontext *ctx, GLuint *width, GLuint *height ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + + /* Fall through for the default because that is one of the uses for it. */ + if ( pContext == pD3DDefault ) + { + *width = 0; + *height = 0; + } + else + { + *width = pContext->pShared->dwWidth; + *height = pContext->pShared->dwHeight; + } +} +/*===========================================================================*/ +/* */ +/* */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static BOOL ResizeContext( GLcontext *ctx ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx, + *pCurrentTemp; + RECT rectClient; + POINT pt; + DWORD dwWidth, + dwHeight; + static BOOL bDDrawLock = FALSE; + + /* Make sure we have some values. */ + if ( (pContext->hdc == NULL ) || + (pContext->pShared->hwnd != WindowFromDC(pContext->hdc)) || + (pContext == pD3DDefault) ) + return FALSE; + + /* Having problems with DDraw sending resize messages before I was done. */ + if( bDDrawLock == TRUE ) + return FALSE; + + // TODO: don't think I need this anymore. + pCurrentTemp = pD3DCurrent; + pD3DCurrent = pD3DDefault; + bDDrawLock = TRUE; + + /* Get the current window dimentions. */ + UpdateScreenPosHAL( pContext->pShared ); + dwWidth = pContext->pShared->rectW.right - pContext->pShared->rectW.left; + dwHeight = pContext->pShared->rectW.bottom - pContext->pShared->rectW.top; + + /* Is the size of the OffScreen Render different? */ + if ( (dwWidth != pContext->pShared->dwWidth) || (dwHeight != pContext->pShared->dwHeight) ) + { + /* Create all the D3D surfaces and device. */ + CreateHAL( pContext->pShared ); + + /* I did this so that software rendering would still work as */ + /* I don't need to scale the z values twice. */ + g_DepthScale = (pContext->pShared->bHardware) ? 1.0 : ((float)0x00FFFFFF); + g_MaxDepth = (pContext->pShared->bHardware) ? 1.0 : ((float)0x00FFFFFF); + gl_DepthRange( pContext->gl_ctx, ctx->Viewport.Near, ctx->Viewport.Far ); + + /* Make sure we have a viewport. */ + gl_Viewport( pContext->gl_ctx, 0, 0, dwWidth, dwHeight ); + + /* Update Mesa as we might have changed from SW <-> HW. */ + SetupDDPointers( pContext->gl_ctx ); + _mesa_make_current( pContext->gl_ctx, pContext->gl_buffer ); + + /* If we are in HW we need to load the current texture if there is one already. */ + // if ( (ctx->Texture.Set[ctx->Texture.CurrentSet].Current != NULL) && + // (pContext->pShared->bHardware == TRUE) ) + // { + // CreateTMgrHAL( pContext->pShared, + // ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Name, + // 0, + // ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Image[0]->Format, + // (RECT *)NULL, + // ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Image[0]->Width, + // ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Image[0]->Height, + // TM_ACTION_BIND, + // (void *)ctx->Texture.Set[ctx->Texture.CurrentSet].Current->Image[0]->Data ); + // } + } + + // TODO: don't think I need this anymore. + pD3DCurrent = pCurrentTemp; + bDDrawLock = FALSE; + + return TRUE; +} + +/*===========================================================================* +/* This function will Blt the render buffer to the PRIMARY surface. I repeat*/ +/* this code for the other SwapBuffer like functions and the flush (didn't */ +/* want the function calling overhead). Thsi could have been a macro... */ +/* */ +/* TODO: there are some problems with viewport/scissoring. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY wglSwapBuffers( HDC hdc ) +{ + /* Fall through for the default because that is one of the uses for it. */ + if ( pD3DCurrent == pD3DDefault ) + return FALSE; + + SwapBuffersHAL( pD3DCurrent->pShared ); + + return TRUE; +} +/*===========================================================================*/ +/* Same as wglSwapBuffers. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY SwapBuffers( HDC hdc ) +{ + /* Fall through for the default because that is one of the uses for it. */ + if ( pD3DCurrent == pD3DDefault ) + return FALSE; + + SwapBuffersHAL( pD3DCurrent->pShared ); + + return TRUE; +} +/*===========================================================================*/ +/* This should be ok as none of the SwapBuffers will cause a redundant Blt */ +/* as none of my Swap functions will call flush. This should also allow */ +/* sinlge buffered applications to work (not really worried though). Some */ +/* applications may flush then swap but then this is there fault IMHO. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void Flush( GLcontext *ctx ) +{ + /* Fall through for the default because that is one of the uses for it. */ + if ( pD3DCurrent == pD3DDefault ) + return; + + SwapBuffersHAL( pD3DCurrent->pShared ); +} +/*===========================================================================*/ +/* For now this function will ignore the supplied PF. If I'm going to allow */ +/* the user to choice the mode and device at startup I'm going to have to do */ +/* something different. */ +/* */ +/* TODO: use the linked list of modes to build a pixel format to be returned */ +/* to the caller. */ +/*===========================================================================*/ +/* RETURN: 1. */ +/*===========================================================================*/ +int APIENTRY wglChoosePixelFormat( HDC hdc, CONST PIXELFORMATDESCRIPTOR *ppfd ) +{ + return 1; +} +/*===========================================================================*/ +/* See wglChoosePixelFormat. */ +/*===========================================================================*/ +/* RETURN: 1. */ +/*===========================================================================*/ +int APIENTRY ChoosePixelFormat( HDC hdc, CONST PIXELFORMATDESCRIPTOR *ppfd ) +{ + return wglChoosePixelFormat(hdc,ppfd); +} +/*===========================================================================*/ +/* This function (for now) returns a static PF everytime. This is just to */ +/* allow things to continue. */ +/*===========================================================================*/ +/* RETURN: 1. */ +/*===========================================================================*/ +int APIENTRY wglDescribePixelFormat( HDC hdc, int iPixelFormat, UINT nBytes, LPPIXELFORMATDESCRIPTOR ppfd ) +{ + static PIXELFORMATDESCRIPTOR pfd = + { + sizeof(PIXELFORMATDESCRIPTOR), /* size */ + 1, /* version */ + PFD_SUPPORT_OPENGL | + PFD_DRAW_TO_WINDOW | + PFD_DOUBLEBUFFER, /* support double-buffering */ + PFD_TYPE_RGBA, /* color type */ + 16, /* prefered color depth */ + 0, 0, 0, 0, 0, 0, /* color bits (ignored) */ + 0, /* no alpha buffer */ + 0, /* alpha bits (ignored) */ + 0, /* no accumulation buffer */ + 0, 0, 0, 0, /* accum bits (ignored) */ + 16, /* depth buffer */ + 0, /* no stencil buffer */ + 0, /* no auxiliary buffers */ + PFD_MAIN_PLANE, /* main layer */ + 0, /* reserved */ + 0, 0, 0, /* no layer, visible, damage masks */ + }; + + /* Return the address of this static PF if one was requested. */ + if ( ppfd != NULL ) + memcpy( ppfd, &pfd, sizeof(PIXELFORMATDESCRIPTOR) ); + + return 1; +} +/*===========================================================================*/ +/* See wglDescribePixelFormat. */ +/*===========================================================================*/ +/* RETURN: 1. */ +/*===========================================================================*/ +int APIENTRY DescribePixelFormat( HDC hdc, int iPixelFormat, UINT nBytes, LPPIXELFORMATDESCRIPTOR ppfd ) +{ + return wglDescribePixelFormat(hdc,iPixelFormat,nBytes,ppfd); +} +/*===========================================================================*/ +/* This function will always return 1 for now. Just to allow for support. */ +/*===========================================================================*/ +/* RETURN: 1. */ +/*===========================================================================*/ +int APIENTRY wglGetPixelFormat( HDC hdc ) +{ + return 1; +} +/*===========================================================================*/ +/* See wglGetPixelFormat. */ +/*===========================================================================*/ +/* RETURN: 1. */ +/*===========================================================================*/ +int APIENTRY GetPixelFormat( HDC hdc ) +{ + return wglGetPixelFormat(hdc); +} +/*===========================================================================*/ +/* This will aways work for now. */ +/*===========================================================================*/ +/* RETURN: TRUE. */ +/*===========================================================================*/ +BOOL APIENTRY wglSetPixelFormat( HDC hdc, int iPixelFormat, CONST PIXELFORMATDESCRIPTOR *ppfd ) +{ + return TRUE; +} +/*===========================================================================*/ +/* See wglSetPixelFormat. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY SetPixelFormat( HDC hdc, int iPixelFormat, CONST PIXELFORMATDESCRIPTOR *ppfd ) +{ + return wglSetPixelFormat(hdc,iPixelFormat,ppfd); +} +/*===========================================================================*/ +/* This is a wrapper function that is supported by my own internal function.*/ +/* that takes my own D3D Mesa context structure. This so I can reuse the */ +/* function (no need for speed). */ +/*===========================================================================*/ +/* RETURN: TRUE. */ +/*===========================================================================*/ +BOOL APIENTRY wglDeleteContext( HGLRC hglrc ) +{ + DestroyContext( (D3DMESACONTEXT *)hglrc ); + + return TRUE; +} +/*===========================================================================*/ +/* Simple getter function that uses a cast. */ +/*===========================================================================*/ +/* RETURN: casted pointer to the context, NULL. */ +/*===========================================================================*/ +HGLRC APIENTRY wglGetCurrentContext( VOID ) +{ + return (pD3DCurrent) ? (HGLRC)pD3DCurrent : (HGLRC)NULL; +} +/*===========================================================================*/ +/* No support. */ +/*===========================================================================*/ +/* RETURN: FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY wglCopyContext( HGLRC hglrcSrc, HGLRC hglrcDst, UINT mask ) +{ + SetLastError( 0 ); + return FALSE; +} +/*===========================================================================*/ +/* No support. */ +/*===========================================================================*/ +/* RETURN: NULL. */ +/*===========================================================================*/ +HGLRC APIENTRY wglCreateLayerContext( HDC hdc,int iLayerPlane ) +{ + SetLastError( 0 ); + return (HGLRC)NULL; +} +/*===========================================================================*/ +/* Simple getter function. */ +/*===========================================================================*/ +/* RETURN: FALSE. */ +/*===========================================================================*/ +HDC APIENTRY wglGetCurrentDC( VOID ) +{ + return (pD3DCurrent) ? pD3DCurrent->hdc : (HDC)NULL; +} +/*===========================================================================*/ +/* Simply call that searches the supported extensions for a match & returns */ +/* the pointer to the function that lends support. */ +/*===========================================================================*/ +/* RETURN: pointer to API call, NULL. */ +/*===========================================================================*/ +PROC APIENTRY wglGetProcAddress( LPCSTR lpszProc ) +{ + int index; + + for( index = 0; index < qt_ext; index++ ) + if( !strcmp(lpszProc,ext[index].name) ) + return ext[index].proc; + + SetLastError( 0 ); + return NULL; +} +/*===========================================================================*/ +/* No support. */ +/*===========================================================================*/ +/* RETURN: FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY wglShareLists( HGLRC hglrc1, HGLRC hglrc2 ) +{ + SetLastError( 0 ); + return FALSE; +} +/*===========================================================================*/ +/* No support. */ +/*===========================================================================*/ +/* RETURN: FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY wglUseFontBitmaps( HDC fontDevice, DWORD firstChar, DWORD numChars, DWORD listBase ) +{ + SetLastError( 0 ); + return FALSE; +} +/*===========================================================================*/ +/* No support. */ +/*===========================================================================*/ +/* RETURN: FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY wglUseFontBitmapsW( HDC hdc,DWORD first,DWORD count,DWORD listBase ) +{ + SetLastError( 0 ); + return FALSE; +} +/*===========================================================================*/ +/* No support. */ +/*===========================================================================*/ +/* RETURN: FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY wglUseFontOutlinesA( HDC hdc, DWORD first, DWORD count, DWORD listBase, FLOAT deviation, FLOAT extrusion, int format, LPGLYPHMETRICSFLOAT lpgmf ) +{ + SetLastError( 0 ); + return FALSE; +} +/*===========================================================================*/ +/* No support. */ +/*===========================================================================*/ +/* RETURN: FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY wglUseFontOutlinesW( HDC hdc,DWORD first,DWORD count, DWORD listBase,FLOAT deviation, FLOAT extrusion,int format, LPGLYPHMETRICSFLOAT lpgmf ) +{ + SetLastError( 0 ); + return FALSE ; +} +/*===========================================================================*/ +/* No support. */ +/*===========================================================================*/ +/* RETURN: FALSE. */ +/*===========================================================================*/ +BOOL APIENTRY wglSwapLayerBuffers( HDC hdc, UINT fuPlanes ) +{ + SetLastError( 0 ); + return FALSE; +} +/*===========================================================================*/ +/* This function will be hooked into the window that has been bound. Right */ +/* now it is used to track the window size and position. Also the we clean */ +/* up the currrent context when the window is close/destroyed. */ +/* */ +/* TODO: there might be something wrong here as some games (Heretic II) don't*/ +/* track the window quit right. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +LONG APIENTRY wglMonitorProc( HWND hwnd, UINT message, UINT wParam, LONG lParam ) +{ + WNDPROC hOldProc; + GLint width, + height; + + switch( message ) + { +// case WM_PAINT: +// break; +// case WM_ACTIVATE: +// break; +// case WM_SHOWWINDOW: +// break; + + case UM_FATALSHUTDOWN: + /* Support the API until we die... */ + MakeCurrent( pD3DDefault ); + break; + + case WM_MOVE: + case WM_DISPLAYCHANGE: + case WM_SIZE: + ResizeContext( pD3DCurrent->gl_ctx ); + break; + + case WM_CLOSE: + case WM_DESTROY: + /* Support the API until we die... */ + hOldProc = pD3DCurrent->hOldProc; + DestroyContext( pD3DCurrent ); + return (hOldProc)(hwnd,message,wParam,lParam); + } + + return (pD3DCurrent->hOldProc)(hwnd,message,wParam,lParam); +} + +/**********************************************************************/ +/***** Miscellaneous device driver funcs *****/ +/**********************************************************************/ + +/*===========================================================================*/ +/* Not reacting to this as I'm only supporting drawing to the back buffer */ +/* right now. */ +/*===========================================================================*/ +/* RETURN: TRUE. */ +/*===========================================================================*/ +static GLboolean SetBuffer( GLcontext *ctx, GLenum buffer ) +{ + if (buffer == GL_BACK_LEFT) + return GL_TRUE; + else + return GL_FALSE; +} +/*===========================================================================*/ +/* This proc will be called by Mesa when the viewport has been set. So if */ +/* we have a context and it isn't the default then we should let D3D know of */ +/* the change. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void SetViewport( GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + RECT rect; + + /* Make sure we can set a viewport. */ + if ( pContext->pShared && (pContext != pD3DDefault) ) + { + // TODO: might be needed. + UpdateScreenPosHAL( pContext->pShared ); + rect.left = x; + rect.right = x + w; + rect.top = y; + rect.bottom = y + h; + + // TODO: shared struct should make this call smaller + SetViewportHAL( pContext->pShared, &rect, 0.0F, 1.0F ); + } +} +/*===========================================================================*/ +/* This function could be better I guess but I decided just to grab the four*/ +/* components and store then seperately. Makes it easier to use IMHO. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void ClearColor( GLcontext *ctx, GLubyte r, GLubyte g, GLubyte b, GLubyte a ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + + pContext->aClear = a; + pContext->bClear = b; + pContext->gClear = g; + pContext->rClear = r; +} +/*===========================================================================*/ +/* This function could be better I guess but I decided just to grab the four*/ +/* components and store then seperately. Makes it easier to use IMHO. */ +/* (is there an echo in here?) */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void SetColor( GLcontext *ctx, GLubyte r, GLubyte g, GLubyte b, GLubyte a ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + + pContext->aCurrent = a; + pContext->bCurrent = b; + pContext->gCurrent = g; + pContext->rCurrent = r; +} +/*===========================================================================*/ +/* */ +/* */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static const char *RendererString( void ) +{ + static char pszRender[64]; + + strcpy( pszRender, "altD3D " ); + + if ( pD3DCurrent->pShared->bHardware ) + strcat( pszRender, "(HW)"); + else + strcat( pszRender, "(SW)"); + + return (const char *)pszRender; +} +/*===========================================================================*/ +/* This function will choose which set of pointers Mesa will use based on */ +/* whether we hard using hardware or software. I have added another set of */ +/* pointers that will do nothing but stop the API from crashing. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void SetupDDPointers( GLcontext *ctx ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + + // TODO: write a generic NULL support for the span render. + if ( pContext->pShared && pContext->pShared->bHardware ) + { + ctx->Driver.UpdateState = SetupHWDDPointers; + } + else if ( pContext == pD3DDefault ) + { + ctx->Driver.UpdateState = SetupNULLDDPointers; + } + else + { + ctx->Driver.UpdateState = SetupSWDDPointers; + } +} +/*===========================================================================*/ +/* This function will populate all the Mesa driver hooks. This version of */ +/* hooks will do nothing but support the API when we don't have a valid */ +/* context bound. This is mostly for applications that don't behave right */ +/* and also to help exit as clean as possable when we have a FatalError. */ +/*===========================================================================*/ +/* RETURN: pointer to the specific function. */ +/*===========================================================================*/ +static void SetupNULLDDPointers( GLcontext *ctx ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + + /* Initialize all the pointers in the DD struct. Do this whenever */ + /* a new context is made current or we change buffers via set_buffer! */ + ctx->Driver.UpdateState = SetupNULLDDPointers; + + /* State management hooks. */ + ctx->Driver.Color = NULLSetColor; + ctx->Driver.ClearColor = NULLClearColor; + ctx->Driver.Clear = NULLClearBuffers; + ctx->Driver.SetBuffer = NULLSetBuffer; + + /* Window management hooks. */ + ctx->Driver.GetBufferSize = NULLGetBufferSize; + + /* Primitive rendering hooks. */ + ctx->Driver.TriangleFunc = NULL; + ctx->Driver.RenderVB = NULL; + + /* Pixel/span writing functions: */ + ctx->Driver.WriteRGBASpan = NULLWrSpRGBA; + ctx->Driver.WriteRGBSpan = NULLWrSpRGB; + ctx->Driver.WriteMonoRGBASpan = NULLWrSpRGBAMono; + ctx->Driver.WriteRGBAPixels = NULLWrPiRGBA; + ctx->Driver.WriteMonoRGBAPixels = NULLWrPiRGBAMono; + + /* Pixel/span reading functions: */ + ctx->Driver.ReadRGBASpan = NULLReSpRGBA; + ctx->Driver.ReadRGBAPixels = NULLRePiRGBA; + + /* Misc. hooks. */ + ctx->Driver.RendererString = RendererString; +} +/*===========================================================================*/ +/* This function will populate all the Mesa driver hooks. There are two of */ +/* these functions. One if we have hardware support and one is there is only*/ +/* software. These functions will be called by Mesa and by the wgl.c when we*/ +/* have resized (or created) the buffers. The thing is that if a window gets*/ +/* resized we may loose hardware support or gain it... */ +/*===========================================================================*/ +/* RETURN: pointer to the specific function. */ +/*===========================================================================*/ +static void SetupSWDDPointers( GLcontext *ctx ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + + /* Initialize all the pointers in the DD struct. Do this whenever */ + /* a new context is made current or we change buffers via set_buffer! */ + ctx->Driver.UpdateState = SetupSWDDPointers; + + /* State management hooks. */ + ctx->Driver.Color = SetColor; + ctx->Driver.ClearColor = ClearColor; + ctx->Driver.Clear = ClearBuffers; + ctx->Driver.SetBuffer = SetBuffer; + + /* Window management hooks. */ + ctx->Driver.GetBufferSize = GetBufferSize; + ctx->Driver.Viewport = SetViewport; + + /* Primitive rendering hooks. */ + ctx->Driver.TriangleFunc = NULL; + ctx->Driver.RenderVB = NULL; + + /* Texture management hooks. */ + + /* Pixel/span writing functions: */ + ctx->Driver.WriteRGBASpan = WSpanRGBA; + ctx->Driver.WriteRGBSpan = WSpanRGB; + ctx->Driver.WriteMonoRGBASpan = WSpanRGBAMono; + ctx->Driver.WriteRGBAPixels = WPixelsRGBA; + ctx->Driver.WriteMonoRGBAPixels = WPixelsRGBAMono; + + /* Pixel/span reading functions: */ + ctx->Driver.ReadRGBASpan = RSpanRGBA; + ctx->Driver.ReadRGBAPixels = RPixelsRGBA; + + /* Misc. hooks. */ + ctx->Driver.Flush = Flush; + ctx->Driver.RendererString = RendererString; +} +/*===========================================================================*/ +/* This function will populate all the Mesa driver hooks. There are two of */ +/* these functions. One if we have hardware support and one is there is only*/ +/* software. These functions will be called by Mesa and by the wgl.c when we*/ +/* have resized (or created) the buffers. The thing is that if a window gets*/ +/* resized we may loose hardware support or gain it... */ +/*===========================================================================*/ +/* RETURN: pointer to the specific function. */ +/*===========================================================================*/ +static void SetupHWDDPointers( GLcontext *ctx ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + + /* Initialize all the pointers in the DD struct. Do this whenever */ + /* a new context is made current or we change buffers via set_buffer! */ + ctx->Driver.UpdateState = SetupHWDDPointers; + + /* State management hooks. */ + ctx->Driver.Color = SetColor; + ctx->Driver.ClearColor = ClearColor; + ctx->Driver.Clear = ClearBuffersD3D; + ctx->Driver.SetBuffer = SetBuffer; + + /* Window management hooks. */ + ctx->Driver.GetBufferSize = GetBufferSize; + ctx->Driver.Viewport = SetViewport; + + /* Primitive rendering hooks. */ + ctx->Driver.TriangleFunc = RenderOneTriangle; + ctx->Driver.LineFunc = RenderOneLine; + ctx->Driver.RenderVB = RenderVertexBuffer; + + /* Pixel/span writing functions: */ + ctx->Driver.WriteRGBASpan = WSpanRGBA; + ctx->Driver.WriteRGBSpan = WSpanRGB; + ctx->Driver.WriteMonoRGBASpan = WSpanRGBAMono; + ctx->Driver.WriteRGBAPixels = WPixelsRGBA; + ctx->Driver.WriteMonoRGBAPixels = WPixelsRGBAMono; + + /* Pixel/span reading functions: */ + ctx->Driver.ReadRGBASpan = RSpanRGBA; + ctx->Driver.ReadRGBAPixels = RPixelsRGBA; + + /* Texture management hooks. */ + // ctx->Driver.BindTexture = TextureBind; + ctx->Driver.TexImage = TextureLoad; + ctx->Driver.TexSubImage = TextureSubImage; + + /* Misc. hooks. */ + ctx->Driver.Flush = Flush; + ctx->Driver.RendererString = RendererString; +} +/*===========================================================================*/ +/* This function will release all resources used by the DLL. Every context */ +/* will be clobbered by releaseing all driver desources and then freeing the */ +/* context memory. Most all the work is done in DestroyContext. */ +/*===========================================================================*/ +/* RETURN: TRUE. */ +/*===========================================================================*/ +static BOOL TermOpenGL( HINSTANCE hInst ) +{ + D3DMESACONTEXT *pTmp, + *pNext; + + /* Just incase we are still getting paint msg. */ + MakeCurrent( pD3DDefault ); + + /* Walk the list until we get back to the default context. */ + for( pTmp = pD3DDefault->next; pTmp != pD3DDefault; pTmp = pNext ) + { + pNext = pTmp->next; + DestroyContext( pTmp ); + } + DestroyContext( pD3DDefault ); + + return TRUE; +} +/*===========================================================================*/ +/* This function is an internal function that will clean up all the Mesa */ +/* context bound to this D3D context. Also any D3D stuff that this context */ +/* uses will be unloaded. */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +static void DestroyContext( D3DMESACONTEXT *pContext ) +{ + D3DMESACONTEXT *pTmp; + + /* Walk the list until we find the context before this one. */ + for( pTmp = pD3DDefault; pTmp && (pTmp->next != pContext); pTmp = pTmp->next ) + if ( pTmp == pTmp->next ) + break; + + /* If we never found it it must already be deleted. */ + if ( pTmp->next != pContext ) + return; + + /* Make sure we are not using this context. */ + if ( pContext == pD3DCurrent ) + MakeCurrent( pD3DDefault ); + + /* Free the Mesa stuff. */ + if ( pContext->gl_visual ) + { + _mesa_destroy_visual( pContext->gl_visual ); + pContext->gl_visual = NULL; + } + if ( pContext->gl_buffer ) + { + _mesa_destroy_framebuffer( pContext->gl_buffer ); + pContext->gl_buffer = NULL; + } + if ( pContext->gl_ctx ) + { + _mesa_destroy_context( pContext->gl_ctx ); + pContext->gl_ctx = NULL; + } + + /* Now dump the D3D. */ + if ( pContext->pShared ) + TermHAL( pContext->pShared ); + + /* Update the previous context's link. */ + pTmp->next = pContext->next; + + /* Gonzo. */ + FREE( pContext ); +} +/*===========================================================================*/ +/* This function will pull the supplied context away from Win32. Basicly it*/ +/* will remove the hook from the window Proc. */ +/* */ +/* TODO: might want to serialize this stuff... */ +/*===========================================================================*/ +/* RETURN: TRUE, FALSE. */ +/*===========================================================================*/ +static BOOL UnBindWindow( D3DMESACONTEXT *pContext ) +{ + if ( pContext == NULL ) + return FALSE; + + if ( pContext == pD3DDefault ) + return TRUE; + + /* Make sure we always have a context bound. */ + if ( pContext == pD3DCurrent ) + pD3DCurrent = pD3DDefault; + + SetWindowLong( pContext->pShared->hwnd, GWL_WNDPROC, (LONG)pContext->hOldProc ); + pContext->hOldProc = NULL; + + return TRUE; +} +/*===========================================================================*/ +/* There are two cases that allow for a faster clear when we know that the */ +/* whole buffer is cleared and that there is no clipping. */ +/*===========================================================================*/ +/* RETURN: the original mask with the bits cleared that represents the buffer* +/* or buffers we just cleared. */ +/*===========================================================================*/ +GLbitfield ClearBuffersD3D( GLcontext *ctx, GLbitfield mask, GLboolean all, GLint x, GLint y, GLint width, GLint height ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + DWORD dwFlags = 0; + + if ( mask & GL_COLOR_BUFFER_BIT ) + { + dwFlags |= D3DCLEAR_TARGET; + mask &= ~GL_COLOR_BUFFER_BIT; + } + if ( mask & GL_DEPTH_BUFFER_BIT ) + { + dwFlags |= D3DCLEAR_ZBUFFER; + mask &= ~GL_DEPTH_BUFFER_BIT; + } + if ( dwFlags == 0 ) + return mask; + + ClearHAL( pContext->pShared, + dwFlags, + all, + x, y, + width, height, + ((pContext->aClear<<24) | (pContext->rClear<<16) | (pContext->gClear<<8) | (pContext->bClear)), + ctx->Depth.Clear, + 0 ); + + return mask; +} + + + +/*===========================================================================*/ +/* TEXTURE MANAGER: ok here is how I did textures. Mesa-3.0 will keep track*/ +/* of all the textures for us. So this means that at anytime we can go to */ +/* the Mesa context and get the current texture. With this in mind this is */ +/* what I did. I really don't care about what textures get or are loaded */ +/* until I actually have to draw a tri that is textured. At this point I */ +/* must have the texture so I demand the texture by destorying all other */ +/* texture surfaces if need be and load the current one. This allows for the*/ +/* best preformance on low memory cards as time is not wasted loading and */ +/* unload textures. */ +/*===========================================================================*/ + + + + + +/*===========================================================================*/ +/* TextureLoad will try and create a D3D surface from the supplied texture */ +/* object if its level 0 (first). The surface will be fully filled with the */ +/* texture. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void TextureLoad( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj, GLint level, GLint internalFormat, const struct gl_texture_image *image ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + + /* TODO: only doing first LOD. */ + if ( (ctx->DriverCtx == NULL) || (level != 0) ) + return; + + CreateTMgrHAL( pContext->pShared, + tObj->Name, + level, + tObj->Image[level]->Format, + (RECT *)NULL, + tObj->Image[level]->Width, + tObj->Image[level]->Height, + TM_ACTION_LOAD, + (void *)tObj->Image[level]->Data ); +} +/*===========================================================================*/ +/* TextureBind make sure that the texture is on the card. Thats it. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void TextureBind( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + + /* TODO: only doing first LOD. */ + if ( (tObj->Image[0] == NULL) || (ctx->DriverCtx == NULL) ) + return; + + CreateTMgrHAL( pContext->pShared, + tObj->Name, + 0, + tObj->Image[0]->Format, + (RECT *)NULL, + tObj->Image[0]->Width, + tObj->Image[0]->Height, + TM_ACTION_BIND, + (void *)tObj->Image[0]->Data ); +} +/*===========================================================================*/ +/* TextureSubImage will make sure that the texture being updated is updated */ +/* if its on the card. */ +/*===========================================================================*/ +/* RETURN: */ +/*===========================================================================*/ +static void TextureSubImage( GLcontext *ctx, GLenum target, struct gl_texture_object *tObj, GLint level, GLint xoffset, GLint yoffset, GLsizei width, GLsizei height, GLint internalFormat, const struct gl_texture_image *image ) +{ + D3DMESACONTEXT *pContext = (D3DMESACONTEXT *)ctx->DriverCtx; + RECT rect; + + /* TODO: only doing first LOD. */ + if ( (ctx->DriverCtx == NULL) || (level > 0) ) + return; + + /* Create a dirty rectangle structure. */ + rect.left = xoffset; + rect.right = xoffset + width; + rect.top = yoffset; + rect.bottom = yoffset + height; + + CreateTMgrHAL( pContext->pShared, + tObj->Name, + 0, + tObj->Image[0]->Format, + &rect, + tObj->Image[0]->Width, + tObj->Image[0]->Height, + TM_ACTION_UPDATE, + (void *)tObj->Image[0]->Data ); +} + -- cgit v1.2.3 From 10d83df3a9cb76a3db76ec9970d7108cf9255d77 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 28 Feb 2008 09:07:03 +0000 Subject: gallium: State tracker cleanups wrt clears. --- src/mesa/state_tracker/st_cb_clear.c | 40 +++++++++++++++--------------------- 1 file changed, 16 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 78baf772f4..e712fd84cd 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -502,37 +502,30 @@ check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) static void clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { - if (!check_clear_color_with_quad( ctx, rb )) - { + if (check_clear_color_with_quad( ctx, rb )) { + /* masking or scissoring */ + clear_with_quad(ctx, GL_TRUE, GL_FALSE, GL_FALSE); + } + else { struct st_renderbuffer *strb = st_renderbuffer(rb); /* clear whole buffer w/out masking */ - GLuint clearValue - = color_value(strb->surface->format, ctx->Color.ClearColor); + uint clearValue = color_value(strb->surface->format, ctx->Color.ClearColor); ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } - else { - /* masking or scissoring */ - clear_with_quad(ctx, GL_TRUE, GL_FALSE, GL_FALSE); - } } static void clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { - struct st_renderbuffer *strb = st_renderbuffer(rb); - /* - const GLboolean isDS = is_depth_stencil_format(strb->surface->format); - */ - - assert(strb->surface->format); - if (check_clear_depth_with_quad(ctx, rb)) { /* scissoring or we have a combined depth/stencil buffer */ clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_FALSE); } else { + struct st_renderbuffer *strb = st_renderbuffer(rb); + /* simple clear of whole buffer */ uint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); @@ -543,13 +536,13 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) static void clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { - struct st_renderbuffer *strb = st_renderbuffer(rb); - if (check_clear_stencil_with_quad(ctx, rb)) { /* masking or scissoring or combined depth/stencil buffer */ clear_with_quad(ctx, GL_FALSE, GL_FALSE, GL_TRUE); } else { + struct st_renderbuffer *strb = st_renderbuffer(rb); + /* simple clear of whole buffer */ GLuint clearValue = ctx->Stencil.Clear; ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); @@ -560,11 +553,14 @@ clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) static void clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { - struct st_renderbuffer *strb = st_renderbuffer(rb); - - assert(is_depth_stencil_format(strb->surface->format)); if (check_clear_depth_stencil_with_quad(ctx, rb)) { + /* masking or scissoring */ + clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_TRUE); + } + else { + struct st_renderbuffer *strb = st_renderbuffer(rb); + /* clear whole buffer w/out masking */ GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); @@ -581,10 +577,6 @@ clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } - else { - /* masking or scissoring */ - clear_with_quad(ctx, GL_FALSE, GL_TRUE, GL_TRUE); - } } -- cgit v1.2.3 From 626b8d177b5fdacfac70c09216c02b1c833b91ea Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 28 Feb 2008 09:07:38 +0000 Subject: Make sure struct pipe_screen is declared. --- src/gallium/drivers/softpipe/sp_winsys.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h index 80f5354808..fc372dba27 100644 --- a/src/gallium/drivers/softpipe/sp_winsys.h +++ b/src/gallium/drivers/softpipe/sp_winsys.h @@ -46,6 +46,7 @@ struct softpipe_winsys { }; +struct pipe_screen; struct pipe_winsys; struct pipe_context; -- cgit v1.2.3 From e280bd50cc46ddbf5ac7fbaafc934d1048d77ba2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 28 Feb 2008 21:25:54 +0900 Subject: gallium: Fix MSVC warnings. --- src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c index cf6de1e82f..ea4a72967d 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -104,17 +104,17 @@ tgsi_scan_shader(const struct tgsi_token *tokens, if (file == TGSI_FILE_INPUT) { info->input_semantic_name[info->num_inputs] - = fulldecl->Semantic.SemanticName; + = (ubyte)fulldecl->Semantic.SemanticName; info->input_semantic_index[info->num_inputs] - = fulldecl->Semantic.SemanticIndex; + = (ubyte)fulldecl->Semantic.SemanticIndex; info->num_inputs++; } if (file == TGSI_FILE_OUTPUT) { info->output_semantic_name[info->num_outputs] - = fulldecl->Semantic.SemanticName; + = (ubyte)fulldecl->Semantic.SemanticName; info->output_semantic_index[info->num_outputs] - = fulldecl->Semantic.SemanticIndex; + = (ubyte)fulldecl->Semantic.SemanticIndex; info->num_outputs++; } -- cgit v1.2.3 From 9a264a056abc376a70e01f097934d590a36df887 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 28 Feb 2008 09:44:09 -0700 Subject: gallium: in GL_SELECT mode, update hitflag in rasterpos --- src/mesa/state_tracker/st_cb_rasterpos.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 4d73916a35..2ed228778e 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -38,6 +38,7 @@ #include "main/imports.h" #include "main/macros.h" +#include "main/feedback.h" #include "st_context.h" #include "st_atom.h" @@ -163,6 +164,10 @@ rastpos_point(struct draw_stage *stage, struct prim_header *prim) ctx->Current.RasterTexCoords[i], VERT_RESULT_TEX0 + i, VERT_ATTRIB_TEX0 + i); } + + if (ctx->RenderMode == GL_SELECT) { + _mesa_update_hitflag( ctx, ctx->Current.RasterPos[2] ); + } } -- cgit v1.2.3 From 476cc931f2bba69d4075b7c44985b145c3999af5 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 28 Feb 2008 11:04:39 -0700 Subject: galliums: s/uint/enum pipe_format/ --- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index ff236adc5c..e0dc657b45 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1216,7 +1216,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, struct pipe_surface *psTex; struct pipe_texture *pt; GLfloat *color; - uint format; + enum pipe_format format; /* make sure rendering has completed */ pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); -- cgit v1.2.3 From 01e53be2b80d5bcb48102f7f1be507b6a5a5832a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 28 Feb 2008 11:17:30 -0700 Subject: gallium: added a texture format sanity check/assertion in st_texture_create() make sure the tex format is actually supported by the driver. --- src/mesa/state_tracker/st_texture.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index c2b0aa8a4a..cbc6f84998 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -86,6 +86,7 @@ st_texture_create(struct st_context *st, _mesa_lookup_enum_by_nr(format), last_level); assert(format); + assert(screen->is_format_supported(screen, format, PIPE_TEXTURE)); memset(&pt, 0, sizeof(pt)); pt.target = target; -- cgit v1.2.3 From 70126588cf78b8a835dfced2b7ca7f1e05afeb67 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 28 Feb 2008 11:31:56 -0700 Subject: gallium: fix surface/texture format conversion in st_CopyPixels If the src buffer's format isn't supported as a texture format, find a different texture format. --- src/mesa/state_tracker/st_cb_drawpixels.c | 40 +++++++++++++++++++++++++++---- 1 file changed, 35 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index e0dc657b45..6c0d75cc55 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1216,7 +1216,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, struct pipe_surface *psTex; struct pipe_texture *pt; GLfloat *color; - enum pipe_format format; + enum pipe_format srcFormat, texFormat; /* make sure rendering has completed */ pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); @@ -1236,6 +1236,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, stvp = st_make_passthrough_vertex_shader(ctx->st, GL_FALSE); } else { + assert(type == GL_DEPTH); rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer); color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0]; stfp = make_fragment_shader_z(ctx->st); @@ -1243,10 +1244,39 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, } psRead = rbRead->surface; - format = psRead->format; + srcFormat = psRead->format; - pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, width, height, - 1, 0); + if (screen->is_format_supported(screen, srcFormat, PIPE_TEXTURE)) { + texFormat = srcFormat; + } + else { + /* srcFormat can't be used as a texture format */ + if (type == GL_DEPTH) { + static const enum pipe_format zFormats[] = { + PIPE_FORMAT_Z16_UNORM, + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_Z24S8_UNORM + }; + uint i; + texFormat = 0; + for (i = 0; i < Elements(zFormats); i++) { + if (screen->is_format_supported(screen, zFormats[i], + PIPE_TEXTURE)) { + texFormat = zFormats[i]; + break; + } + } + assert(texFormat); /* XXX no depth texture formats??? */ + } + else { + /* todo */ + assert(0); + } + } + + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, texFormat, 0, + width, height, 1, 0); if (!pt) return; @@ -1260,7 +1290,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, * front/back color buffers as surfaces (they're XImages and Pixmaps). * So, this var tells us if we can use surface_copy here... */ - if (st->haveFramebufferSurfaces) { + if (st->haveFramebufferSurfaces && srcFormat == texFormat) { /* copy source framebuffer surface into mipmap/texture */ pipe->surface_copy(pipe, FALSE, -- cgit v1.2.3 From a1a13954885cd469faab49633b5386e5c889e3df Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 28 Feb 2008 17:49:22 -0700 Subject: gallium: split draw_wide_prim stage into separate point/line stages. This fixes a validation/code-path problem. Enabling the stage for the sake of wide points also inadvertantly caused wide lines to be converted to tris when we actually want them passed through, such as for the AA line stage. This is just cleaner now. Also, replace draw_convert_wide_lines() with draw_wide_line_threshold() as was done for points. Allows for 1-pixel lines to be converted too if needed. --- src/gallium/auxiliary/draw/Makefile | 4 +++- src/gallium/auxiliary/draw/SConscript | 3 ++- src/gallium/auxiliary/draw/draw_context.c | 17 ++++++++++------- src/gallium/auxiliary/draw/draw_context.h | 4 +--- src/gallium/auxiliary/draw/draw_private.h | 8 +++++--- src/gallium/auxiliary/draw/draw_validate.c | 16 +++++++++------- 6 files changed, 30 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index c9980f0b83..2daa1636f3 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -29,7 +29,9 @@ C_SOURCES = \ draw_vf.c \ draw_vf_generic.c \ draw_vf_sse.c \ - draw_wide_prims.c + draw_wide_line.c \ + draw_wide_point.c + include ../../Makefile.template diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 3302dc44f7..c18dcb2927 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -28,7 +28,8 @@ draw = env.ConvenienceLibrary( 'draw_vf.c', 'draw_vf_generic.c', 'draw_vf_sse.c', - 'draw_wide_prims.c', + 'draw_wide_point.c', + 'draw_wide_line.c' ]) auxiliaries.insert(0, draw) diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 3500c34811..428b6209e0 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -48,7 +48,8 @@ struct draw_context *draw_create( void ) #endif /* create pipeline stages */ - draw->pipeline.wide = draw_wide_stage( draw ); + draw->pipeline.wide_line = draw_wide_line_stage( draw ); + draw->pipeline.wide_point = draw_wide_point_stage( draw ); draw->pipeline.stipple = draw_stipple_stage( draw ); draw->pipeline.unfilled = draw_unfilled_stage( draw ); draw->pipeline.twoside = draw_twoside_stage( draw ); @@ -80,8 +81,9 @@ struct draw_context *draw_create( void ) draw->shader_queue_flush = draw_vertex_shader_queue_flush; + /* these defaults are oriented toward the needs of softpipe */ draw->wide_point_threshold = 1000000.0; /* infinity */ - draw->convert_wide_lines = TRUE; + draw->wide_line_threshold = 1.0; draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ @@ -94,7 +96,8 @@ struct draw_context *draw_create( void ) void draw_destroy( struct draw_context *draw ) { - draw->pipeline.wide->destroy( draw->pipeline.wide ); + draw->pipeline.wide_line->destroy( draw->pipeline.wide_line ); + draw->pipeline.wide_point->destroy( draw->pipeline.wide_point ); draw->pipeline.stipple->destroy( draw->pipeline.stipple ); draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); draw->pipeline.twoside->destroy( draw->pipeline.twoside ); @@ -232,14 +235,14 @@ draw_wide_point_threshold(struct draw_context *draw, float threshold) /** - * Tells the draw module whether to convert wide lines (width != 1) - * into triangles. + * Tells the draw module to draw lines with triangles if their width + * is greater than this threshold. */ void -draw_convert_wide_lines(struct draw_context *draw, boolean enable) +draw_wide_line_threshold(struct draw_context *draw, float threshold) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->convert_wide_lines = enable; + draw->wide_line_threshold = threshold; } diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index 99bfef55f4..ab87b4127c 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -92,9 +92,7 @@ void draw_set_rasterize_stage( struct draw_context *draw, void draw_wide_point_threshold(struct draw_context *draw, float threshold); -void draw_convert_wide_points(struct draw_context *draw, boolean enable); - -void draw_convert_wide_lines(struct draw_context *draw, boolean enable); +void draw_wide_line_threshold(struct draw_context *draw, float threshold); boolean draw_use_sse(struct draw_context *draw); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index e988e71d23..c732d723a7 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -186,7 +186,8 @@ struct draw_context struct draw_stage *aapoint; struct draw_stage *aaline; struct draw_stage *pstipple; - struct draw_stage *wide; + struct draw_stage *wide_line; + struct draw_stage *wide_point; struct draw_stage *rasterize; } pipeline; @@ -219,7 +220,7 @@ struct draw_context unsigned nr_planes; float wide_point_threshold; /**< convert pnts to tris if larger than this */ - boolean convert_wide_lines; /**< convert wide lines to tris? */ + float wide_line_threshold; /**< convert lines to tris if wider than this */ boolean use_sse; /* If a prim stage introduces new vertex attributes, they'll be stored here @@ -304,7 +305,8 @@ extern struct draw_stage *draw_clip_stage( struct draw_context *context ); extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); extern struct draw_stage *draw_cull_stage( struct draw_context *context ); extern struct draw_stage *draw_stipple_stage( struct draw_context *context ); -extern struct draw_stage *draw_wide_stage( struct draw_context *context ); +extern struct draw_stage *draw_wide_line_stage( struct draw_context *context ); +extern struct draw_stage *draw_wide_point_stage( struct draw_context *context ); extern struct draw_stage *draw_validate_stage( struct draw_context *context ); diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index ded7d10c08..084eee9b6e 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -53,8 +53,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) stage->next = next; /* drawing wide lines? */ - wide_lines = (draw->rasterizer->line_width != 1.0 - && draw->convert_wide_lines + wide_lines = (draw->rasterizer->line_width > draw->wide_line_threshold && !draw->rasterizer->line_smooth); /* drawing large points? */ @@ -82,11 +81,14 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) next = draw->pipeline.aapoint; } - if (wide_lines || - wide_points || - draw->rasterizer->point_sprite) { - draw->pipeline.wide->next = next; - next = draw->pipeline.wide; + if (wide_lines) { + draw->pipeline.wide_line->next = next; + next = draw->pipeline.wide_line; + } + + if (wide_points || draw->rasterizer->point_sprite) { + draw->pipeline.wide_point->next = next; + next = draw->pipeline.wide_point; } if (draw->rasterizer->line_stipple_enable) { -- cgit v1.2.3 From b233b1e2dc2fca2f45b3cb5df167e3675b8cc1fb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 28 Feb 2008 17:54:42 -0700 Subject: gallium: new wide point/line stages (missed in prev commit) --- src/gallium/auxiliary/draw/draw_wide_line.c | 187 +++++++++++++++++++ src/gallium/auxiliary/draw/draw_wide_point.c | 257 +++++++++++++++++++++++++++ 2 files changed, 444 insertions(+) create mode 100644 src/gallium/auxiliary/draw/draw_wide_line.c create mode 100644 src/gallium/auxiliary/draw/draw_wide_point.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_wide_line.c b/src/gallium/auxiliary/draw/draw_wide_line.c new file mode 100644 index 0000000000..946a983f00 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_wide_line.c @@ -0,0 +1,187 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +struct wideline_stage { + struct draw_stage stage; + + float half_line_width; +}; + + + +static INLINE struct wideline_stage *wideline_stage( struct draw_stage *stage ) +{ + return (struct wideline_stage *)stage; +} + + +static void wideline_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void wideline_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Draw a wide line by drawing a quad (two triangles). + * XXX need to disable polygon stipple. + */ +static void wideline_line( struct draw_stage *stage, + struct prim_header *header ) +{ + /*const struct wideline_stage *wide = wideline_stage(stage);*/ + const float half_width = 0.5f * stage->draw->rasterizer->line_width; + + struct prim_header tri; + + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + const float dx = FABSF(pos0[0] - pos2[0]); + const float dy = FABSF(pos0[1] - pos2[1]); + + /* + * Draw wide line as a quad (two tris) by "stretching" the line along + * X or Y. + * We need to tweak coords in several ways to be conformant here. + */ + + if (dx > dy) { + /* x-major line */ + pos0[1] = pos0[1] - half_width - 0.25f; + pos1[1] = pos1[1] + half_width - 0.25f; + pos2[1] = pos2[1] - half_width - 0.25f; + pos3[1] = pos3[1] + half_width - 0.25f; + if (pos0[0] < pos2[0]) { + /* left to right line */ + pos0[0] -= 0.5f; + pos1[0] -= 0.5f; + pos2[0] -= 0.5f; + pos3[0] -= 0.5f; + } + else { + /* right to left line */ + pos0[0] += 0.5f; + pos1[0] += 0.5f; + pos2[0] += 0.5f; + pos3[0] += 0.5f; + } + } + else { + /* y-major line */ + pos0[0] = pos0[0] - half_width + 0.25f; + pos1[0] = pos1[0] + half_width + 0.25f; + pos2[0] = pos2[0] - half_width + 0.25f; + pos3[0] = pos3[0] + half_width + 0.25f; + if (pos0[1] < pos2[1]) { + /* top to bottom line */ + pos0[1] -= 0.5f; + pos1[1] -= 0.5f; + pos2[1] -= 0.5f; + pos3[1] -= 0.5f; + } + else { + /* bottom to top line */ + pos0[1] += 0.5f; + pos1[1] += 0.5f; + pos2[1] += 0.5f; + pos3[1] += 0.5f; + } + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +static void wideline_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->next->flush( stage->next, flags ); +} + + +static void wideline_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void wideline_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +struct draw_stage *draw_wide_line_stage( struct draw_context *draw ) +{ + struct wideline_stage *wide = CALLOC_STRUCT(wideline_stage); + + draw_alloc_temp_verts( &wide->stage, 4 ); + + wide->stage.draw = draw; + wide->stage.next = NULL; + wide->stage.point = wideline_point; + wide->stage.line = wideline_line; + wide->stage.tri = wideline_tri; + wide->stage.flush = wideline_flush; + wide->stage.reset_stipple_counter = wideline_reset_stipple_counter; + wide->stage.destroy = wideline_destroy; + + return &wide->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_wide_point.c b/src/gallium/auxiliary/draw/draw_wide_point.c new file mode 100644 index 0000000000..8f877a102a --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_wide_point.c @@ -0,0 +1,257 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +struct widepoint_stage { + struct draw_stage stage; + + float half_point_size; + + uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; + uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; + uint num_texcoords; + + int psize_slot; +}; + + + +static INLINE struct widepoint_stage * +widepoint_stage( struct draw_stage *stage ) +{ + return (struct widepoint_stage *)stage; +} + + +static void passthrough_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void widepoint_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line(stage->next, header); +} + +static void widepoint_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Set the vertex texcoords for sprite mode. + * Coords may be left untouched or set to a right-side-up or upside-down + * orientation. + */ +static void set_texcoords(const struct widepoint_stage *wide, + struct vertex_header *v, const float tc[4]) +{ + uint i; + for (i = 0; i < wide->num_texcoords; i++) { + if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { + uint j = wide->texcoord_slot[i]; + v->data[j][0] = tc[0]; + if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) + v->data[j][1] = 1.0f - tc[1]; + else + v->data[j][1] = tc[1]; + v->data[j][2] = tc[2]; + v->data[j][3] = tc[3]; + } + } +} + + +/* If there are lots of sprite points (and why wouldn't there be?) it + * would probably be more sensible to change hardware setup to + * optimize this rather than doing the whole thing in software like + * this. + */ +static void widepoint_point( struct draw_stage *stage, + struct prim_header *header ) +{ + const struct widepoint_stage *wide = widepoint_stage(stage); + const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; + float half_size; + float left_adj, right_adj; + + struct prim_header tri; + + /* four dups of original vertex */ + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + /* point size is either per-vertex or fixed size */ + if (wide->psize_slot >= 0) { + half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; + } + else { + half_size = wide->half_point_size; + } + + left_adj = -half_size; /* + 0.25f;*/ + right_adj = half_size; /* + 0.25f;*/ + + pos0[0] += left_adj; + pos0[1] -= half_size; + + pos1[0] += left_adj; + pos1[1] += half_size; + + pos2[0] += right_adj; + pos2[1] -= half_size; + + pos3[0] += right_adj; + pos3[1] += half_size; + + if (sprite) { + static const float tex00[4] = { 0, 0, 0, 1 }; + static const float tex01[4] = { 0, 1, 0, 1 }; + static const float tex11[4] = { 1, 1, 0, 1 }; + static const float tex10[4] = { 1, 0, 0, 1 }; + set_texcoords( wide, v0, tex00 ); + set_texcoords( wide, v1, tex01 ); + set_texcoords( wide, v2, tex10 ); + set_texcoords( wide, v3, tex11 ); + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +static void widepoint_first_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct widepoint_stage *wide = widepoint_stage(stage); + struct draw_context *draw = stage->draw; + + wide->half_point_size = 0.5f * draw->rasterizer->point_size; + + /* XXX we won't know the real size if it's computed by the vertex shader! */ + if (draw->rasterizer->point_size > draw->wide_point_threshold) { + stage->point = widepoint_point; + } + else { + stage->point = passthrough_point; + } + + if (draw->rasterizer->point_sprite) { + /* find vertex shader texcoord outputs */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i, j = 0; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { + wide->texcoord_slot[j] = i; + wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; + j++; + } + } + wide->num_texcoords = j; + } + + wide->psize_slot = -1; + if (draw->rasterizer->point_size_per_vertex) { + /* find PSIZ vertex output */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + wide->psize_slot = i; + break; + } + } + } + + stage->point( stage, header ); +} + + +static void widepoint_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->point = widepoint_first_point; + stage->next->flush( stage->next, flags ); +} + + +static void widepoint_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void widepoint_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +struct draw_stage *draw_wide_point_stage( struct draw_context *draw ) +{ + struct widepoint_stage *wide = CALLOC_STRUCT(widepoint_stage); + + draw_alloc_temp_verts( &wide->stage, 4 ); + + wide->stage.draw = draw; + wide->stage.next = NULL; + wide->stage.point = widepoint_first_point; + wide->stage.line = widepoint_line; + wide->stage.tri = widepoint_tri; + wide->stage.flush = widepoint_flush; + wide->stage.reset_stipple_counter = widepoint_reset_stipple_counter; + wide->stage.destroy = widepoint_destroy; + + return &wide->stage; +} -- cgit v1.2.3 From 6144c2bd65974f4e5b74936e26451a1ad75cb349 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 28 Feb 2008 17:57:54 -0700 Subject: cell: remove obsolete texture field --- src/gallium/drivers/cell/ppu/cell_texture.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index bf6f6d3058..e235421107 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -90,7 +90,6 @@ cell_texture_create_screen(struct pipe_screen *screen, spt->base = *templat; spt->base.refcount = 1; - spt->base.pipe = NULL; spt->base.screen = screen; cell_texture_layout(spt); -- cgit v1.2.3 From ebe3b34ad225e320a09bb4069ce4d24808386327 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 28 Feb 2008 18:02:05 -0700 Subject: cell: convert all points/lines to tris for the time being Allows more programs to look correct. We'll want native points/lines someday. --- src/gallium/drivers/cell/ppu/cell_context.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index 2301df5ba5..ccbbd1d331 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -128,6 +128,10 @@ cell_create_context(struct pipe_screen *screen, cell_init_vbuf(cell); draw_set_rasterize_stage(cell->draw, cell->vbuf); + /* convert all points/lines to tris for the time being */ + draw_wide_point_threshold(cell->draw, 0.0); + draw_wide_line_threshold(cell->draw, 0.0); + /* * SPU stuff */ -- cgit v1.2.3 From 78220aea864c36038f8425ad9d8467d2a2bdea58 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 29 Feb 2008 10:07:44 -0700 Subject: gallium: remove the ugly pipe->draw stage lookup code in aaline/point/pstipple stages Added a void *draw ptr to pipe_context. Probably look for a better solution someday. --- src/gallium/auxiliary/draw/draw_aaline.c | 32 ++++-------------------------- src/gallium/auxiliary/draw/draw_aapoint.c | 32 ++++-------------------------- src/gallium/auxiliary/draw/draw_pstipple.c | 32 ++++-------------------------- src/gallium/include/pipe/p_context.h | 1 + 4 files changed, 13 insertions(+), 84 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index 51140388f0..7660e56fe6 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -691,35 +691,11 @@ draw_aaline_stage(struct draw_context *draw) } -/* - * XXX temporary? solution to mapping a pipe_context to a aaline_stage. - */ - -#define MAX_CONTEXTS 10 - -static struct pipe_context *Pipe[MAX_CONTEXTS]; -static struct aaline_stage *Stage[MAX_CONTEXTS]; -static uint NumContexts; - -static void -add_aa_pipe_context(struct pipe_context *pipe, struct aaline_stage *aa) -{ - assert(NumContexts < MAX_CONTEXTS); - Pipe[NumContexts] = pipe; - Stage[NumContexts] = aa; - NumContexts++; -} - static struct aaline_stage * aaline_stage_from_pipe(struct pipe_context *pipe) { - uint i; - for (i = 0; i < NumContexts; i++) { - if (Pipe[i] == pipe) - return Stage[i]; - } - assert(0); - return NULL; + struct draw_context *draw = (struct draw_context *) pipe->draw; + return aaline_stage(draw->pipeline.aaline); } @@ -802,6 +778,8 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) { struct aaline_stage *aaline; + pipe->draw = (void *) draw; + /* * Create / install AA line drawing / prim stage */ @@ -830,6 +808,4 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) pipe->bind_sampler_state = aaline_bind_sampler_state; pipe->set_sampler_texture = aaline_set_sampler_texture; - - add_aa_pipe_context(pipe, aaline); } diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index d48a416899..70f696475f 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -762,35 +762,11 @@ draw_aapoint_stage(struct draw_context *draw) } -/* - * XXX temporary? solution to mapping a pipe_context to a aapoint_stage. - */ - -#define MAX_CONTEXTS 10 - -static struct pipe_context *Pipe[MAX_CONTEXTS]; -static struct aapoint_stage *Stage[MAX_CONTEXTS]; -static uint NumContexts; - -static void -add_aa_pipe_context(struct pipe_context *pipe, struct aapoint_stage *aa) -{ - assert(NumContexts < MAX_CONTEXTS); - Pipe[NumContexts] = pipe; - Stage[NumContexts] = aa; - NumContexts++; -} - static struct aapoint_stage * aapoint_stage_from_pipe(struct pipe_context *pipe) { - uint i; - for (i = 0; i < NumContexts; i++) { - if (Pipe[i] == pipe) - return Stage[i]; - } - assert(0); - return NULL; + struct draw_context *draw = (struct draw_context *) pipe->draw; + return aapoint_stage(draw->pipeline.aapoint); } @@ -850,6 +826,8 @@ draw_install_aapoint_stage(struct draw_context *draw, { struct aapoint_stage *aapoint; + pipe->draw = (void *) draw; + /* * Create / install AA point drawing / prim stage */ @@ -868,6 +846,4 @@ draw_install_aapoint_stage(struct draw_context *draw, pipe->create_fs_state = aapoint_create_fs_state; pipe->bind_fs_state = aapoint_bind_fs_state; pipe->delete_fs_state = aapoint_delete_fs_state; - - add_aa_pipe_context(pipe, aapoint); } diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index f6200aa820..2cfeb813b3 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -559,35 +559,11 @@ draw_pstip_stage(struct draw_context *draw) } -/* - * XXX temporary? solution to mapping a pipe_context to a pstip_stage. - */ - -#define MAX_CONTEXTS 10 - -static struct pipe_context *Pipe[MAX_CONTEXTS]; -static struct pstip_stage *Stage[MAX_CONTEXTS]; -static uint NumContexts; - -static void -add_pstip_pipe_context(struct pipe_context *pipe, struct pstip_stage *pstip) -{ - assert(NumContexts < MAX_CONTEXTS); - Pipe[NumContexts] = pipe; - Stage[NumContexts] = pstip; - NumContexts++; -} - static struct pstip_stage * pstip_stage_from_pipe(struct pipe_context *pipe) { - uint i; - for (i = 0; i < NumContexts; i++) { - if (Pipe[i] == pipe) - return Stage[i]; - } - assert(0); - return NULL; + struct draw_context *draw = (struct draw_context *) pipe->draw; + return pstip_stage(draw->pipeline.pstipple); } @@ -686,6 +662,8 @@ draw_install_pstipple_stage(struct draw_context *draw, { struct pstip_stage *pstip; + pipe->draw = (void *) draw; + /* * Create / install AA line drawing / prim stage */ @@ -716,6 +694,4 @@ draw_install_pstipple_stage(struct draw_context *draw, pipe->bind_sampler_state = pstip_bind_sampler_state; pipe->set_sampler_texture = pstip_set_sampler_texture; pipe->set_polygon_stipple = pstip_set_polygon_stipple; - - add_pstip_pipe_context(pipe, pstip); } diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index d0f25d7d46..1501b52f3e 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -56,6 +56,7 @@ struct pipe_context { struct pipe_screen *screen; void *priv; /** context private data (for DRI for example) */ + void *draw; /** private, for draw module (temporary? */ void (*destroy)( struct pipe_context * ); -- cgit v1.2.3 From a41b77f4fedc7d956d8cb44547d812b5449f8641 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 29 Feb 2008 11:09:02 -0700 Subject: gallium: added pipe_get/put_tile_z() functions --- src/gallium/auxiliary/util/p_tile.c | 124 ++++++++++++++++++++++++++++++++++++ src/gallium/auxiliary/util/p_tile.h | 14 ++++ 2 files changed, 138 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 3f795a3898..287d783981 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -697,3 +697,127 @@ pipe_put_tile_rgba(struct pipe_context *pipe, FREE(packed); } + + +/** + * Get a block of Z values, converted to 32-bit range. + */ +void +pipe_get_tile_z(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + uint *z) +{ + const uint dstStride = w; + uint *pDest = z; + uint i, j; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + switch (ps->format) { + case PIPE_FORMAT_Z32_UNORM: + { + const uint *pSrc + = (const uint *) pipe_surface_map(ps) + (y * ps->pitch + x); + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, 4 * w); + pDest += dstStride; + pSrc += ps->pitch; + } + } + break; + case PIPE_FORMAT_S8Z24_UNORM: + { + const uint *pSrc + = (const uint *) pipe_surface_map(ps) + (y * ps->pitch + x); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 24-bit Z to 32-bit Z */ + pDest[j] = (pSrc[j] << 8) | (pSrc[j] & 0xff); + } + pDest += dstStride; + pSrc += ps->pitch; + } + } + break; + case PIPE_FORMAT_Z16_UNORM: + { + const ushort *pSrc + = (const ushort *) pipe_surface_map(ps) + (y * ps->pitch + x); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 16-bit Z to 32-bit Z */ + pDest[j] = (pSrc[j] << 16) | pSrc[j]; + } + pDest += dstStride; + pSrc += ps->pitch; + } + } + break; + default: + assert(0); + } + + pipe_surface_unmap(ps); +} + + +void +pipe_put_tile_z(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const uint *zSrc) +{ + const uint srcStride = w; + const uint *pSrc = zSrc; + uint i, j; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + switch (ps->format) { + case PIPE_FORMAT_Z32_UNORM: + { + uint *pDest = (uint *) pipe_surface_map(ps) + (y * ps->pitch + x); + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, 4 * w); + pDest += ps->pitch; + pSrc += srcStride; + } + } + break; + case PIPE_FORMAT_S8Z24_UNORM: + { + uint *pDest = (uint *) pipe_surface_map(ps) + (y * ps->pitch + x); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 32-bit Z to 24-bit Z (0 stencil) */ + pDest[j] = pSrc[j] >> 8; + } + pDest += ps->pitch; + pSrc += srcStride; + } + } + break; + case PIPE_FORMAT_Z16_UNORM: + { + ushort *pDest = (ushort *) pipe_surface_map(ps) + (y * ps->pitch + x); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 32-bit Z to 16-bit Z */ + pDest[j] = pSrc[j] >> 16; + } + pDest += ps->pitch; + pSrc += srcStride; + } + } + break; + default: + assert(0); + } + + pipe_surface_unmap(ps); +} + + diff --git a/src/gallium/auxiliary/util/p_tile.h b/src/gallium/auxiliary/util/p_tile.h index cd8124bf11..318b6d11a6 100644 --- a/src/gallium/auxiliary/util/p_tile.h +++ b/src/gallium/auxiliary/util/p_tile.h @@ -78,4 +78,18 @@ pipe_put_tile_rgba(struct pipe_context *pipe, uint x, uint y, uint w, uint h, const float *p); + +extern void +pipe_get_tile_z(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + uint *z); + +extern void +pipe_put_tile_z(struct pipe_context *pipe, + struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const uint *z); + + #endif -- cgit v1.2.3 From 0b47eb4808aa47e2ab276ab60493e28774cef21a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 29 Feb 2008 11:09:45 -0700 Subject: gallium: fix glCopyPixels(GL_DEPTH) when Z format conversion is needed --- src/mesa/state_tracker/st_cb_drawpixels.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 6c0d75cc55..de99882ab7 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1299,7 +1299,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, psRead, srcx, srcy, width, height); } - else { + else if (type == GL_COLOR) { /* alternate path using get/put_tile() */ GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); @@ -1308,6 +1308,13 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, free(buf); } + else { + /* GL_DEPTH */ + GLuint *buf = (GLuint *) malloc(width * height * sizeof(GLuint)); + pipe_get_tile_z(pipe, psRead, srcx, srcy, width, height, buf); + pipe_put_tile_z(pipe, psTex, 0, 0, width, height, buf); + free(buf); + } /* draw textured quad */ draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2], -- cgit v1.2.3 From 2a121e8e2289d2ca136f511c5a6ef049f9c99ec4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 29 Feb 2008 11:37:12 -0700 Subject: gallium: tweak coords for wide lines --- src/gallium/auxiliary/draw/draw_wide_line.c | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_wide_line.c b/src/gallium/auxiliary/draw/draw_wide_line.c index 946a983f00..9a168ce8bd 100644 --- a/src/gallium/auxiliary/draw/draw_wide_line.c +++ b/src/gallium/auxiliary/draw/draw_wide_line.c @@ -86,7 +86,10 @@ static void wideline_line( struct draw_stage *stage, const float dx = FABSF(pos0[0] - pos2[0]); const float dy = FABSF(pos0[1] - pos2[1]); - + + /* small tweak to meet GL specification */ + const float bias = 0.125f; + /* * Draw wide line as a quad (two tris) by "stretching" the line along * X or Y. @@ -95,10 +98,10 @@ static void wideline_line( struct draw_stage *stage, if (dx > dy) { /* x-major line */ - pos0[1] = pos0[1] - half_width - 0.25f; - pos1[1] = pos1[1] + half_width - 0.25f; - pos2[1] = pos2[1] - half_width - 0.25f; - pos3[1] = pos3[1] + half_width - 0.25f; + pos0[1] = pos0[1] - half_width - bias; + pos1[1] = pos1[1] + half_width - bias; + pos2[1] = pos2[1] - half_width - bias; + pos3[1] = pos3[1] + half_width - bias; if (pos0[0] < pos2[0]) { /* left to right line */ pos0[0] -= 0.5f; @@ -116,10 +119,10 @@ static void wideline_line( struct draw_stage *stage, } else { /* y-major line */ - pos0[0] = pos0[0] - half_width + 0.25f; - pos1[0] = pos1[0] + half_width + 0.25f; - pos2[0] = pos2[0] - half_width + 0.25f; - pos3[0] = pos3[0] + half_width + 0.25f; + pos0[0] = pos0[0] - half_width + bias; + pos1[0] = pos1[0] + half_width + bias; + pos2[0] = pos2[0] - half_width + bias; + pos3[0] = pos3[0] + half_width + bias; if (pos0[1] < pos2[1]) { /* top to bottom line */ pos0[1] -= 0.5f; -- cgit v1.2.3 From 4d22330837278574c7f06bdc9e12ffffb659f43d Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 29 Feb 2008 10:32:25 +0100 Subject: scons: List sp_screen.c. --- src/gallium/drivers/softpipe/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript index 4c1a6d5df0..88c21ee3b0 100644 --- a/src/gallium/drivers/softpipe/SConscript +++ b/src/gallium/drivers/softpipe/SConscript @@ -28,6 +28,7 @@ softpipe = env.ConvenienceLibrary( 'sp_quad_stencil.c', 'sp_quad_stipple.c', 'sp_query.c', + 'sp_screen.c', 'sp_state_blend.c', 'sp_state_clip.c', 'sp_state_derived.c', -- cgit v1.2.3 From b8ee90e05a98d5a167c1fdb5a8fc3bc0cb4a6a78 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 29 Feb 2008 12:21:42 -0700 Subject: gallium: need precalc_flat=1 for wide lines --- src/gallium/auxiliary/draw/draw_validate.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index 084eee9b6e..b43295b586 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -84,6 +84,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) if (wide_lines) { draw->pipeline.wide_line->next = next; next = draw->pipeline.wide_line; + precalc_flat = 1; } if (wide_points || draw->rasterizer->point_sprite) { -- cgit v1.2.3 From 6da943d204eeb488895933c45e174042cb69c92d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 29 Feb 2008 12:22:04 -0700 Subject: gallium: point rast coord tweak --- src/gallium/auxiliary/draw/draw_wide_point.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_wide_point.c b/src/gallium/auxiliary/draw/draw_wide_point.c index 8f877a102a..65bd50f2b8 100644 --- a/src/gallium/auxiliary/draw/draw_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_wide_point.c @@ -109,7 +109,7 @@ static void widepoint_point( struct draw_stage *stage, const struct widepoint_stage *wide = widepoint_stage(stage); const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; float half_size; - float left_adj, right_adj; + float left_adj, right_adj, bot_adj, top_adj; struct prim_header tri; @@ -124,6 +124,8 @@ static void widepoint_point( struct draw_stage *stage, float *pos2 = v2->data[0]; float *pos3 = v3->data[0]; + const float xbias = 0.0, ybias = -0.125; + /* point size is either per-vertex or fixed size */ if (wide->psize_slot >= 0) { half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; @@ -132,20 +134,22 @@ static void widepoint_point( struct draw_stage *stage, half_size = wide->half_point_size; } - left_adj = -half_size; /* + 0.25f;*/ - right_adj = half_size; /* + 0.25f;*/ + left_adj = -half_size + xbias; + right_adj = half_size + xbias; + bot_adj = half_size + ybias; + top_adj = -half_size + ybias; pos0[0] += left_adj; - pos0[1] -= half_size; + pos0[1] += top_adj; pos1[0] += left_adj; - pos1[1] += half_size; + pos1[1] += bot_adj; pos2[0] += right_adj; - pos2[1] -= half_size; + pos2[1] += top_adj; pos3[0] += right_adj; - pos3[1] += half_size; + pos3[1] += bot_adj; if (sprite) { static const float tex00[4] = { 0, 0, 0, 1 }; -- cgit v1.2.3 From 5240cebb23f4862f4f7458a1b397957e4460b527 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 29 Feb 2008 13:00:17 -0700 Subject: gallium: fix line emit order for unfilled tris A tri drawn with GL_LINE_LOOP and GL_POLYGON w/ fillmode=GL_LINE should produce the same results when line stipple is enabled. Results are correct now. --- src/gallium/auxiliary/draw/draw_unfilled.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_unfilled.c b/src/gallium/auxiliary/draw/draw_unfilled.c index 8777cfdfc8..4d718d514c 100644 --- a/src/gallium/auxiliary/draw/draw_unfilled.c +++ b/src/gallium/auxiliary/draw/draw_unfilled.c @@ -101,9 +101,9 @@ static void lines( struct draw_stage *stage, assert(((header->edgeflags & 0x4) >> 2) == header->v[2]->edgeflag); #endif + if (header->edgeflags & 0x4) line( stage, v2, v0 ); if (header->edgeflags & 0x1) line( stage, v0, v1 ); if (header->edgeflags & 0x2) line( stage, v1, v2 ); - if (header->edgeflags & 0x4) line( stage, v2, v0 ); } -- cgit v1.2.3 From c8bca8d8a94b7bce532b40fd5c422063632b26c4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 29 Feb 2008 15:38:43 -0700 Subject: gallium: fix update_textures() bug in pipe_texture binding logic The 'st' pointer might not have changed, but st->pt might have. The dirtyData flag will indicate when that's happened, so check it. --- src/mesa/state_tracker/st_atom_texture.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index a4ac726816..0e1c396aeb 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -67,7 +67,9 @@ update_textures(struct st_context *st) * this table before being deleted, otherwise the pointer * comparison below could fail. */ - if (st->state.sampler_texture[unit] != stObj) { + if (st->state.sampler_texture[unit] != stObj || + (stObj && stObj->dirtyData)) { + /* should really test if the bound 'pt' is changing */ struct pipe_texture *pt = st_get_stobj_texture(stObj); st->state.sampler_texture[unit] = stObj; st->pipe->set_sampler_texture(st->pipe, unit, pt); -- cgit v1.2.3 From 07d6347e8a51fc7bbd5c586a5739f17c68c5eafd Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 29 Feb 2008 16:16:16 -0700 Subject: gallium: change st->state.sampler_texture[] to store pipe_texture pointers This is a better fix for the previous check-in. Fixes texadd.c conform test, and probably other bugs. --- src/mesa/state_tracker/st_atom_texture.c | 14 ++++++-------- src/mesa/state_tracker/st_cb_drawpixels.c | 3 +-- src/mesa/state_tracker/st_context.h | 2 +- src/mesa/state_tracker/st_gen_mipmap.c | 3 +-- 4 files changed, 9 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 0e1c396aeb..697d2cdfb4 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -55,6 +55,7 @@ update_textures(struct st_context *st) const GLuint su = fprog->Base.SamplerUnits[unit]; struct gl_texture_object *texObj = st->ctx->Texture.Unit[su]._Current; struct st_texture_object *stObj = st_texture_object(texObj); + struct pipe_texture *pt; if (texObj) { GLboolean flush, retval; @@ -67,18 +68,15 @@ update_textures(struct st_context *st) * this table before being deleted, otherwise the pointer * comparison below could fail. */ - if (st->state.sampler_texture[unit] != stObj || - (stObj && stObj->dirtyData)) { - /* should really test if the bound 'pt' is changing */ - struct pipe_texture *pt = st_get_stobj_texture(stObj); - st->state.sampler_texture[unit] = stObj; + + pt = st_get_stobj_texture(stObj); + + if (st->state.sampler_texture[unit] != pt) { + st->state.sampler_texture[unit] = pt; st->pipe->set_sampler_texture(st->pipe, unit, pt); } - stObj = st->state.sampler_texture[unit]; - if (stObj && stObj->dirtyData) { - struct pipe_texture *pt = st_get_stobj_texture(stObj); st->pipe->texture_update(st->pipe, pt); stObj->dirtyData = GL_FALSE; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index de99882ab7..65c9fda9cb 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -727,8 +727,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); pipe->bind_fs_state(pipe, ctx->st->state.fs->data); pipe->bind_vs_state(pipe, ctx->st->state.vs->cso->data); - pipe->set_sampler_texture(pipe, unit, - st_get_stobj_texture(ctx->st->state.sampler_texture[unit])); + pipe->set_sampler_texture(pipe, unit, ctx->st->state.sampler_texture[unit]); pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]->data); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 5be4769be4..1fbf9721e7 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -86,7 +86,7 @@ struct st_context struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; struct pipe_framebuffer_state framebuffer; - struct st_texture_object *sampler_texture[PIPE_MAX_SAMPLERS]; + struct pipe_texture *sampler_texture[PIPE_MAX_SAMPLERS]; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_viewport_state viewport; diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 243dc0b1d0..841d77abbc 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -312,8 +312,7 @@ st_render_mipmap(struct st_context *st, pipe->bind_vs_state(pipe, st->state.vs->cso->data); if (st->state.sampler[0]) pipe->bind_sampler_state(pipe, 0, st->state.sampler[0]->data); - pipe->set_sampler_texture(pipe, 0, - st_get_stobj_texture(st->state.sampler_texture[0])); + pipe->set_sampler_texture(pipe, 0, st->state.sampler_texture[0]); pipe->set_viewport_state(pipe, &st->state.viewport); return TRUE; -- cgit v1.2.3 From e884c7ed9a14aabaa86f6710c594d20812ed11d9 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sat, 1 Mar 2008 08:04:21 -0500 Subject: start implementing start of bultins --- src/gallium/auxiliary/gallivm/Makefile | 7 +- src/gallium/auxiliary/gallivm/instructions.cpp | 10 ++ src/gallium/auxiliary/gallivm/instructionssoa.cpp | 116 ++++++++++++++++++++++ src/gallium/auxiliary/gallivm/instructionssoa.h | 14 ++- src/gallium/auxiliary/gallivm/soabuiltins.c | 58 +++++++++++ src/gallium/auxiliary/gallivm/storagesoa.cpp | 28 +++--- src/gallium/auxiliary/gallivm/storagesoa.h | 15 +-- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 38 ++----- src/gallium/include/pipe/p_shader_tokens.h | 20 ++-- 9 files changed, 239 insertions(+), 67 deletions(-) create mode 100644 src/gallium/auxiliary/gallivm/soabuiltins.c (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/Makefile b/src/gallium/auxiliary/gallivm/Makefile index 39fac6ea4a..c24e19e062 100644 --- a/src/gallium/auxiliary/gallivm/Makefile +++ b/src/gallium/auxiliary/gallivm/Makefile @@ -15,7 +15,7 @@ GALLIVM_SOURCES = \ storagesoa.cpp \ instructionssoa.cpp -INC_SOURCES = gallivm_builtins.cpp +INC_SOURCES = gallivm_builtins.cpp gallivmsoabuiltins.cpp CPP_SOURCES = \ $(GALLIVM_SOURCES) @@ -65,8 +65,10 @@ depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) gallivm_builtins.cpp: llvm_builtins.c - clang --emit-llvm $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins + clang --emit-llvm < $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins +gallivmsoabuiltins.cpp: soabuiltins.c + clang --emit-llvm < $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-module -o=$@ -f -for=shader -funcname=createSoaBuiltins # Emacs tags tags: @@ -78,6 +80,7 @@ clean: -rm -f *.o */*.o *~ *.so *~ server/*.o -rm -f depend depend.bak -rm -f gallivm_builtins.cpp + -rm -f gallivmsoabuiltins.cpp symlinks: diff --git a/src/gallium/auxiliary/gallivm/instructions.cpp b/src/gallium/auxiliary/gallivm/instructions.cpp index 55d39fa5f1..8919491792 100644 --- a/src/gallium/auxiliary/gallivm/instructions.cpp +++ b/src/gallium/auxiliary/gallivm/instructions.cpp @@ -42,6 +42,7 @@ #include #include #include +#include #include #include @@ -51,6 +52,15 @@ using namespace llvm; #include "gallivm_builtins.cpp" +#if 0 + +llvm::Value *arrayFromChannels(std::vector &vals) +{ + VectorType *vectorType = VectorType::get(Type::FloatTy, 4); + ArrayType *vectorArray = ArrayType::get(vectorType, 4); +} +#endif + static inline std::string createFuncName(int label) { std::ostringstream stream; diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index a4d5046637..5739cf0cde 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -2,9 +2,28 @@ #include "storagesoa.h" +#include "pipe/p_shader_tokens.h" + +#include #include +#include +#include +#include +#include +#include + +#include + +/* disable some warnings. this file is autogenerated */ +#if defined(__GNUC__) +#pragma GCC diagnostic ignored "-Wunused-variable" +#endif using namespace llvm; +#include "gallivmsoabuiltins.cpp" +#if defined(__GNUC__) +#pragma GCC diagnostic warning "-Wunused-variable" +#endif InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, StorageSoa *storage) @@ -12,6 +31,8 @@ InstructionsSoa::InstructionsSoa(llvm::Module *mod, llvm::Function *func, m_storage(storage), m_idx(0) { + createFunctionMap(); + createBuiltins(); } const char * InstructionsSoa::name(const char *prefix) const @@ -119,3 +140,98 @@ std::vector InstructionsSoa::extractVector(llvm::Value *vector) return res; } + +void InstructionsSoa::createFunctionMap() +{ + m_functionsMap[TGSI_OPCODE_DP3] = "dp3"; +} + +llvm::Function * InstructionsSoa::function(int op) +{ + if (m_functions.find(op) != m_functions.end()) + return m_functions[op]; + + std::string name = m_functionsMap[op]; + + llvm::Function *originalFunc = m_builtins->getFunction(name); + llvm::Function *func = CloneFunction(originalFunc); + currentModule()->getFunctionList().push_back(func); + std::cout << "Func parent is "<getParent() + <<", cur is "<setParent(currentModule()); + m_functions[op] = func; + return func; +} + +llvm::Module * InstructionsSoa::currentModule() const +{ + BasicBlock *block = m_builder.GetInsertBlock(); + if (!block || !block->getParent()) + return 0; + + return block->getParent()->getParent(); +} + +void InstructionsSoa::createBuiltins() +{ + m_builtins = createSoaBuiltins(); +} + +std::vector InstructionsSoa::dp3(const std::vector in1, + const std::vector in2) +{ + llvm::Function *func = function(TGSI_OPCODE_DP3); + std::vector params; + + llvm::Value *tmp = allocaTemp(); + params.push_back(tmp); + + params.push_back(in1[0]); + params.push_back(in1[1]); + params.push_back(in1[2]); + params.push_back(in1[3]); + params.push_back(in2[0]); + params.push_back(in2[1]); + params.push_back(in2[2]); + params.push_back(in2[3]); + CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), + name("dp3")); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + + std::vector indices; + indices.push_back(m_storage->constantInt(0)); + indices.push_back(m_storage->constantInt(0)); + + GetElementPtrInst *getElem = new GetElementPtrInst(tmp, + indices.begin(), + indices.end(), + name("allocaPtr"), + m_builder.GetInsertBlock()); + indices = std::vector(); + indices.push_back(m_storage->constantInt(0)); + GetElementPtrInst *xElemPtr = new GetElementPtrInst(getElem, + m_storage->constantInt(0), + name("xPtr"), + m_builder.GetInsertBlock()); + GetElementPtrInst *yElemPtr = new GetElementPtrInst(getElem, + m_storage->constantInt(1), + name("yPtr"), + m_builder.GetInsertBlock()); + GetElementPtrInst *zElemPtr = new GetElementPtrInst(getElem, + m_storage->constantInt(2), + name("zPtr"), + m_builder.GetInsertBlock()); + GetElementPtrInst *wElemPtr = new GetElementPtrInst(getElem, + m_storage->constantInt(3), + name("wPtr"), + m_builder.GetInsertBlock()); + + std::vector res(4); + res[0] = new LoadInst(xElemPtr); + res[1] = new LoadInst(yElemPtr); + res[2] = new LoadInst(zElemPtr); + res[3] = new LoadInst(wElemPtr); + + return res; +} diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h index 4169dcbb2e..5c26687150 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.h +++ b/src/gallium/auxiliary/gallivm/instructionssoa.h @@ -30,6 +30,7 @@ #include +#include #include namespace llvm { @@ -47,9 +48,10 @@ public: llvm::BasicBlock *block, StorageSoa *storage); std::vector arl(const std::vector in); - std::vector add(const std::vector in1, const std::vector in2); + std::vector dp3(const std::vector in1, + const std::vector in2); std::vector madd(const std::vector in1, const std::vector in2, const std::vector in3); @@ -62,9 +64,19 @@ private: const char * name(const char *prefix) const; llvm::Value *vectorFromVals(llvm::Value *x, llvm::Value *y, llvm::Value *z, llvm::Value *w); + void createFunctionMap(); + void createBuiltins(); + llvm::Function *function(int); + llvm::Module *currentModule() const; + llvm::Value *allocaTemp(); private: llvm::LLVMFoldingBuilder m_builder; StorageSoa *m_storage; + + std::map m_functionsMap; + std::map m_functions; + llvm::Module *m_builtins; + private: mutable int m_idx; mutable char m_name[32]; diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c new file mode 100644 index 0000000000..0b428a750f --- /dev/null +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -0,0 +1,58 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * This file is compiled with clang into the LLVM bitcode + * + * Authors: + * Zack Rusin zack@tungstengraphics.com + */ +typedef __attribute__(( ocu_vector_type(4) )) float float4; + +void dp3(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, + float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) +{ + float4 dot = (tmp0x * tmp1x) + (tmp0y * tmp1y) + + (tmp0z * tmp1z); + + res[0] = dot; + res[1] = dot; + res[2] = dot; + res[3] = dot; +} + +#if 0 +void yo(float4 *out, float4 *in) +{ + float4 res[4]; + + dp3(res, in[0], in[1], in[2], in[3], + in[4], in[5], in[6], in[7]); + out[1] = res[1]; +} +#endif diff --git a/src/gallium/auxiliary/gallivm/storagesoa.cpp b/src/gallium/auxiliary/gallivm/storagesoa.cpp index ed0674a96f..bb6fe3d7e1 100644 --- a/src/gallium/auxiliary/gallivm/storagesoa.cpp +++ b/src/gallium/auxiliary/gallivm/storagesoa.cpp @@ -277,7 +277,7 @@ llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector &v return constVector; } -std::vector StorageSoa::load(Argument type, int idx, int swizzle, +std::vector StorageSoa::load(enum tgsi_file_type type, int idx, int swizzle, llvm::Value *indIdx) { std::vector val(4); @@ -292,25 +292,29 @@ std::vector StorageSoa::load(Argument type, int idx, int swizzle, debug_printf("XXXXXXXXX realIdx = %p, indIdx = %p\n", realIndex, indIdx); switch(type) { - case Input: + case TGSI_FILE_INPUT: val = inputElement(realIndex); break; - case Output: + case TGSI_FILE_OUTPUT: val = outputElement(realIndex); break; - case Temp: + case TGSI_FILE_TEMPORARY: val = tempElement(realIndex); break; - case Const: + case TGSI_FILE_CONSTANT: val = constElement(realIndex); break; - case Immediate: + case TGSI_FILE_IMMEDIATE: val = immediateElement(realIndex); break; - case Address: + case TGSI_FILE_ADDRESS: debug_printf("Address not handled in the load phase!\n"); assert(0); break; + default: + debug_printf("Unknown load!\n"); + assert(0); + break; } if (!gallivm_is_swizzle(swizzle)) return val; @@ -324,21 +328,21 @@ std::vector StorageSoa::load(Argument type, int idx, int swizzle, return res; } -void StorageSoa::store(Argument type, int idx, const std::vector &val, +void StorageSoa::store(enum tgsi_file_type type, int idx, const std::vector &val, int mask) { llvm::Value *out = 0; switch(type) { - case Output: + case TGSI_FILE_OUTPUT: out = m_output; break; - case Temp: + case TGSI_FILE_TEMPORARY: out = m_temps; break; - case Input: + case TGSI_FILE_INPUT: out = m_input; break; - case Address: { + case TGSI_FILE_ADDRESS: { llvm::Value *addr = m_addresses[idx]; if (!addr) { addAddress(idx); diff --git a/src/gallium/auxiliary/gallivm/storagesoa.h b/src/gallium/auxiliary/gallivm/storagesoa.h index 6443351f27..ae2fc7c6ae 100644 --- a/src/gallium/auxiliary/gallivm/storagesoa.h +++ b/src/gallium/auxiliary/gallivm/storagesoa.h @@ -28,6 +28,8 @@ #ifndef STORAGESOA_H #define STORAGESOA_H +#include + #include #include #include @@ -45,15 +47,6 @@ namespace llvm { class StorageSoa { -public: - enum Argument { - Input, - Output, - Temp, - Const, - Immediate, - Address - }; public: StorageSoa(llvm::BasicBlock *block, llvm::Value *input, @@ -62,9 +55,9 @@ public: llvm::Value *temps); - std::vector load(Argument type, int idx, int swizzle, + std::vector load(enum tgsi_file_type type, int idx, int swizzle, llvm::Value *indIdx =0); - void store(Argument type, int idx, const std::vector &val, + void store(enum tgsi_file_type type, int idx, const std::vector &val, int mask); void addImmediate(float *vec); diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index 2cb4acce32..a52ee26434 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -708,25 +708,9 @@ translate_instructionir(llvm::Module *module, if (src->SrcRegister.Indirect) { indIdx = storage->addrElement(src->SrcRegisterInd.Index); } - if (src->SrcRegister.File == TGSI_FILE_CONSTANT) { - val = storage->load(StorageSoa::Const, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_INPUT) { - val = storage->load(StorageSoa::Input, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_TEMPORARY) { - val = storage->load(StorageSoa::Temp, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_OUTPUT) { - val = storage->load(StorageSoa::Output, - src->SrcRegister.Index, swizzle, indIdx); - } else if (src->SrcRegister.File == TGSI_FILE_IMMEDIATE) { - val = storage->load(StorageSoa::Immediate, - src->SrcRegister.Index, swizzle, indIdx); - } else { - fprintf(stderr, "ERROR: not supported llvm source %d\n", src->SrcRegister.File); - return; - } + + val = storage->load((enum tgsi_file_type)src->SrcRegister.File, + src->SrcRegister.Index, swizzle, indIdx); inputs[i] = val; } @@ -763,6 +747,7 @@ translate_instructionir(llvm::Module *module, } break; case TGSI_OPCODE_DP3: { + out = instr->dp3(inputs[0], inputs[1]); } break; case TGSI_OPCODE_DP4: { @@ -1067,19 +1052,8 @@ translate_instructionir(llvm::Module *module, for (int i = 0; i < inst->Instruction.NumDstRegs; ++i) { struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - if (dst->DstRegister.File == TGSI_FILE_OUTPUT) { - storage->store(StorageSoa::Output, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_TEMPORARY) { - storage->store(StorageSoa::Temp, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else if (dst->DstRegister.File == TGSI_FILE_ADDRESS) { - storage->store(StorageSoa::Address, - dst->DstRegister.Index, out, dst->DstRegister.WriteMask); - } else { - fprintf(stderr, "ERROR: unsupported LLVM destination!"); - assert(!"wrong destination"); - } + storage->store((enum tgsi_file_type)dst->DstRegister.File, + dst->DstRegister.Index, out, dst->DstRegister.WriteMask); } } diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 0a6145a6bf..b110f01291 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -42,15 +42,17 @@ struct tgsi_token unsigned Extended : 1; /* BOOL */ }; -#define TGSI_FILE_NULL 0 -#define TGSI_FILE_CONSTANT 1 -#define TGSI_FILE_INPUT 2 -#define TGSI_FILE_OUTPUT 3 -#define TGSI_FILE_TEMPORARY 4 -#define TGSI_FILE_SAMPLER 5 -#define TGSI_FILE_ADDRESS 6 -#define TGSI_FILE_IMMEDIATE 7 -#define TGSI_FILE_COUNT 8 /**< how many TGSI_FILE_ types */ +enum tgsi_file_type { + TGSI_FILE_NULL =0, + TGSI_FILE_CONSTANT =1, + TGSI_FILE_INPUT =2, + TGSI_FILE_OUTPUT =3, + TGSI_FILE_TEMPORARY =4, + TGSI_FILE_SAMPLER =5, + TGSI_FILE_ADDRESS =6, + TGSI_FILE_IMMEDIATE =7, + TGSI_FILE_COUNT /**< how many TGSI_FILE_ types */ +}; #define TGSI_DECLARE_RANGE 0 -- cgit v1.2.3 From 17f543fc4529ca4ce7f73a840ed0fb50d1fec925 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sat, 1 Mar 2008 08:32:31 -0500 Subject: make the first builtin work (dp3) --- src/gallium/auxiliary/gallivm/gallivm.cpp | 10 +++++- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 41 +++++++++++++---------- 2 files changed, 33 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/gallivm.cpp b/src/gallium/auxiliary/gallivm/gallivm.cpp index d14bb3b99a..b6f641a3f8 100644 --- a/src/gallium/auxiliary/gallivm/gallivm.cpp +++ b/src/gallium/auxiliary/gallivm/gallivm.cpp @@ -306,11 +306,19 @@ struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) { struct gallivm_prog *prog = (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); + + std::cout << "Before optimizations:"<module->dump(); + std::cout<<"-------------------------------"<module); llvm::Module *mod = llvm::CloneModule(ir->module); prog->num_consts = ir->num_consts; memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators)); prog->num_interp = ir->num_interp; - + /* Run optimization passes over it */ PassManager passes; passes.add(new TargetData(mod)); diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 5739cf0cde..3fcfce8ce0 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -158,6 +158,7 @@ llvm::Function * InstructionsSoa::function(int op) currentModule()->getFunctionList().push_back(func); std::cout << "Func parent is "<getParent() <<", cur is "<dump(); //func->setParent(currentModule()); m_functions[op] = func; return func; @@ -184,8 +185,15 @@ std::vector InstructionsSoa::dp3(const std::vector i std::vector params; llvm::Value *tmp = allocaTemp(); - params.push_back(tmp); - + std::vector indices; + indices.push_back(m_storage->constantInt(0)); + indices.push_back(m_storage->constantInt(0)); + GetElementPtrInst *getElem = new GetElementPtrInst(tmp, + indices.begin(), + indices.end(), + name("allocaPtr"), + m_builder.GetInsertBlock()); + params.push_back(getElem); params.push_back(in1[0]); params.push_back(in1[1]); params.push_back(in1[2]); @@ -194,20 +202,10 @@ std::vector InstructionsSoa::dp3(const std::vector i params.push_back(in2[1]); params.push_back(in2[2]); params.push_back(in2[3]); - CallInst *call = m_builder.CreateCall(func, params.begin(), params.end(), - name("dp3")); + CallInst *call = m_builder.CreateCall(func, params.begin(), params.end()); call->setCallingConv(CallingConv::C); call->setTailCall(false); - std::vector indices; - indices.push_back(m_storage->constantInt(0)); - indices.push_back(m_storage->constantInt(0)); - - GetElementPtrInst *getElem = new GetElementPtrInst(tmp, - indices.begin(), - indices.end(), - name("allocaPtr"), - m_builder.GetInsertBlock()); indices = std::vector(); indices.push_back(m_storage->constantInt(0)); GetElementPtrInst *xElemPtr = new GetElementPtrInst(getElem, @@ -228,10 +226,19 @@ std::vector InstructionsSoa::dp3(const std::vector i m_builder.GetInsertBlock()); std::vector res(4); - res[0] = new LoadInst(xElemPtr); - res[1] = new LoadInst(yElemPtr); - res[2] = new LoadInst(zElemPtr); - res[3] = new LoadInst(wElemPtr); + res[0] = new LoadInst(xElemPtr, name("xRes"), false, m_builder.GetInsertBlock()); + res[1] = new LoadInst(yElemPtr, name("yRes"), false, m_builder.GetInsertBlock()); + res[2] = new LoadInst(zElemPtr, name("zRes"), false, m_builder.GetInsertBlock()); + res[3] = new LoadInst(wElemPtr, name("wRes"), false, m_builder.GetInsertBlock()); return res; } + +llvm::Value * InstructionsSoa::allocaTemp() +{ + VectorType *vector = VectorType::get(Type::FloatTy, 4); + ArrayType *vecArray = ArrayType::get(vector, 4); + AllocaInst *alloca = new AllocaInst(vecArray, name("tmpRes"), + m_builder.GetInsertBlock()); + return alloca; +} -- cgit v1.2.3 From a9c40f833ead8459788b86603c7f2b94632b1109 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sat, 1 Mar 2008 09:50:41 -0500 Subject: refactor code calling builtins and implement dp4 --- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 116 +++++++++++++++++----- src/gallium/auxiliary/gallivm/instructionssoa.h | 12 +++ src/gallium/auxiliary/gallivm/soabuiltins.c | 14 +++ src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 1 + 4 files changed, 116 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 3fcfce8ce0..89d513afd0 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -144,6 +144,7 @@ std::vector InstructionsSoa::extractVector(llvm::Value *vector) void InstructionsSoa::createFunctionMap() { m_functionsMap[TGSI_OPCODE_DP3] = "dp3"; + m_functionsMap[TGSI_OPCODE_DP4] = "dp4"; } llvm::Function * InstructionsSoa::function(int op) @@ -182,45 +183,42 @@ std::vector InstructionsSoa::dp3(const std::vector i const std::vector in2) { llvm::Function *func = function(TGSI_OPCODE_DP3); - std::vector params; + return callBuiltin(func, in1, in2); +} + +llvm::Value * InstructionsSoa::allocaTemp() +{ + VectorType *vector = VectorType::get(Type::FloatTy, 4); + ArrayType *vecArray = ArrayType::get(vector, 4); + AllocaInst *alloca = new AllocaInst(vecArray, name("tmpRes"), + m_builder.GetInsertBlock()); - llvm::Value *tmp = allocaTemp(); std::vector indices; indices.push_back(m_storage->constantInt(0)); indices.push_back(m_storage->constantInt(0)); - GetElementPtrInst *getElem = new GetElementPtrInst(tmp, + GetElementPtrInst *getElem = new GetElementPtrInst(alloca, indices.begin(), indices.end(), name("allocaPtr"), m_builder.GetInsertBlock()); - params.push_back(getElem); - params.push_back(in1[0]); - params.push_back(in1[1]); - params.push_back(in1[2]); - params.push_back(in1[3]); - params.push_back(in2[0]); - params.push_back(in2[1]); - params.push_back(in2[2]); - params.push_back(in2[3]); - CallInst *call = m_builder.CreateCall(func, params.begin(), params.end()); - call->setCallingConv(CallingConv::C); - call->setTailCall(false); + return getElem; +} - indices = std::vector(); - indices.push_back(m_storage->constantInt(0)); - GetElementPtrInst *xElemPtr = new GetElementPtrInst(getElem, +std::vector InstructionsSoa::allocaToResult(llvm::Value *allocaPtr) +{ + GetElementPtrInst *xElemPtr = new GetElementPtrInst(allocaPtr, m_storage->constantInt(0), name("xPtr"), m_builder.GetInsertBlock()); - GetElementPtrInst *yElemPtr = new GetElementPtrInst(getElem, + GetElementPtrInst *yElemPtr = new GetElementPtrInst(allocaPtr, m_storage->constantInt(1), name("yPtr"), m_builder.GetInsertBlock()); - GetElementPtrInst *zElemPtr = new GetElementPtrInst(getElem, + GetElementPtrInst *zElemPtr = new GetElementPtrInst(allocaPtr, m_storage->constantInt(2), name("zPtr"), m_builder.GetInsertBlock()); - GetElementPtrInst *wElemPtr = new GetElementPtrInst(getElem, + GetElementPtrInst *wElemPtr = new GetElementPtrInst(allocaPtr, m_storage->constantInt(3), name("wPtr"), m_builder.GetInsertBlock()); @@ -234,11 +232,75 @@ std::vector InstructionsSoa::dp3(const std::vector i return res; } -llvm::Value * InstructionsSoa::allocaTemp() +std::vector InstructionsSoa::dp4(const std::vector in1, + const std::vector in2) { - VectorType *vector = VectorType::get(Type::FloatTy, 4); - ArrayType *vecArray = ArrayType::get(vector, 4); - AllocaInst *alloca = new AllocaInst(vecArray, name("tmpRes"), - m_builder.GetInsertBlock()); - return alloca; + llvm::Function *func = function(TGSI_OPCODE_DP4); + return callBuiltin(func, in1, in2); +} + +std::vector InstructionsSoa::callBuiltin(llvm::Function *func, const std::vector in1) +{ + std::vector params; + + llvm::Value *allocaPtr = allocaTemp(); + params.push_back(allocaPtr); + params.push_back(in1[0]); + params.push_back(in1[1]); + params.push_back(in1[2]); + params.push_back(in1[3]); + CallInst *call = m_builder.CreateCall(func, params.begin(), params.end()); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + + return allocaToResult(allocaPtr); +} + +std::vector InstructionsSoa::callBuiltin(llvm::Function *func, const std::vector in1, + const std::vector in2) +{ + std::vector params; + + llvm::Value *allocaPtr = allocaTemp(); + params.push_back(allocaPtr); + params.push_back(in1[0]); + params.push_back(in1[1]); + params.push_back(in1[2]); + params.push_back(in1[3]); + params.push_back(in2[0]); + params.push_back(in2[1]); + params.push_back(in2[2]); + params.push_back(in2[3]); + CallInst *call = m_builder.CreateCall(func, params.begin(), params.end()); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + + return allocaToResult(allocaPtr); +} + +std::vector InstructionsSoa::callBuiltin(llvm::Function *func, const std::vector in1, + const std::vector in2, + const std::vector in3) +{ + std::vector params; + + llvm::Value *allocaPtr = allocaTemp(); + params.push_back(allocaPtr); + params.push_back(in1[0]); + params.push_back(in1[1]); + params.push_back(in1[2]); + params.push_back(in1[3]); + params.push_back(in2[0]); + params.push_back(in2[1]); + params.push_back(in2[2]); + params.push_back(in2[3]); + params.push_back(in3[0]); + params.push_back(in3[1]); + params.push_back(in3[2]); + params.push_back(in3[3]); + CallInst *call = m_builder.CreateCall(func, params.begin(), params.end()); + call->setCallingConv(CallingConv::C); + call->setTailCall(false); + + return allocaToResult(allocaPtr); } diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h index 5c26687150..3ef51dcaff 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.h +++ b/src/gallium/auxiliary/gallivm/instructionssoa.h @@ -52,6 +52,8 @@ public: const std::vector in2); std::vector dp3(const std::vector in1, const std::vector in2); + std::vector dp4(const std::vector in1, + const std::vector in2); std::vector madd(const std::vector in1, const std::vector in2, const std::vector in3); @@ -69,6 +71,16 @@ private: llvm::Function *function(int); llvm::Module *currentModule() const; llvm::Value *allocaTemp(); + std::vector allocaToResult(llvm::Value *allocaPtr); + std::vector callBuiltin(llvm::Function *func, + const std::vector in1); + std::vector callBuiltin(llvm::Function *func, + const std::vector in1, + const std::vector in2); + std::vector callBuiltin(llvm::Function *func, + const std::vector in1, + const std::vector in2, + const std::vector in3); private: llvm::LLVMFoldingBuilder m_builder; StorageSoa *m_storage; diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index 0b428a750f..24c14e1b69 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -46,6 +46,20 @@ void dp3(float4 *res, res[3] = dot; } + +void dp4(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, + float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) +{ + float4 dot = (tmp0x * tmp1x) + (tmp0y * tmp1y) + + (tmp0z * tmp1z) + (tmp0w * tmp1w); + + res[0] = dot; + res[1] = dot; + res[2] = dot; + res[3] = dot; +} + #if 0 void yo(float4 *out, float4 *in) { diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index a52ee26434..3f65865a5a 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -751,6 +751,7 @@ translate_instructionir(llvm::Module *module, } break; case TGSI_OPCODE_DP4: { + out = instr->dp4(inputs[0], inputs[1]); } break; case TGSI_OPCODE_DST: { -- cgit v1.2.3 From f4e91c3432eaf653757193a8a1ac438372ea64a6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 27 Feb 2008 13:06:18 +0000 Subject: gallium: document user_buffer_create a little --- src/gallium/include/pipe/p_winsys.h | 31 +++++++++++++++++++++++++++---- 1 file changed, 27 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h index e784c92491..1383bd0544 100644 --- a/src/gallium/include/pipe/p_winsys.h +++ b/src/gallium/include/pipe/p_winsys.h @@ -104,13 +104,36 @@ struct pipe_winsys * usage is a bitmask of PIPE_BUFFER_USAGE_PIXEL/VERTEX/INDEX/CONSTANT. This * usage argument is only an optimization hint, not a guarantee, therefore * proper behavior must be observed in all circumstances. + * + * alignment indicates the client's alignment requirements, eg for + * SSE instructions. */ struct pipe_buffer *(*buffer_create)( struct pipe_winsys *sws, - unsigned alignment, - unsigned usage, - unsigned size ); + unsigned alignment, + unsigned usage, + unsigned size ); - /** Create a buffer that wraps user-space data */ + /** + * Create a buffer that wraps user-space data. + * + * Effectively this schedules a delayed call to buffer_create + * followed by an upload of the data at *some point in the future*, + * or perhaps never. Basically the allocate/upload is delayed + * until the buffer is actually passed to hardware. + * + * The intention is to provide a quick way to turn regular data + * into a buffer, and secondly to avoid a copy operation if that + * data subsequently turns out to be only accessed by the CPU. + * + * Common example is OpenGL vertex buffers that are subsequently + * processed either by software TNL in the driver or by passing to + * hardware. + * + * XXX: What happens if the delayed call to buffer_create() fails? + * + * Note that ptr may be accessed at any time upto the time when the + * buffer is destroyed, so the data must not be freed before then. + */ struct pipe_buffer *(*user_buffer_create)(struct pipe_winsys *sws, void *ptr, unsigned bytes); -- cgit v1.2.3 From 9506ac823593387aa2f3a19f48ea07a91d4b6bb9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 28 Feb 2008 17:08:37 +0000 Subject: gallium: remove obsolete comment --- src/gallium/include/pipe/p_state.h | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 5fab41acbd..49a003b923 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -281,8 +281,6 @@ struct pipe_surface */ struct pipe_texture { - /* Effectively the key: - */ enum pipe_texture_target target; /**< PIPE_TEXTURE_x */ enum pipe_format format; /**< PIPE_FORMAT_x */ -- cgit v1.2.3 From 800d13df726b9f82f796c86fe7ae6d18231820ec Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 3 Mar 2008 17:44:04 +0100 Subject: draw: add fetch for bgra ubyte surfaces --- src/gallium/auxiliary/draw/draw_vertex_fetch.c | 44 +++++++++++++++++--------- 1 file changed, 29 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vertex_fetch.c b/src/gallium/auxiliary/draw/draw_vertex_fetch.c index cb8cdd04a3..b56d85396d 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_fetch.c +++ b/src/gallium/auxiliary/draw/draw_vertex_fetch.c @@ -54,7 +54,7 @@ fetch_##NAME(const void *ptr, float *attrib) \ int i; \ \ for (i = 0; i < SZ; i++) { \ - attrib[i] = CVT; \ + attrib[i] = CVT(i); \ } \ \ for (; i < 4; i++) { \ @@ -62,24 +62,24 @@ fetch_##NAME(const void *ptr, float *attrib) \ } \ } -#define CVT_64_FLOAT (float) ((double *) ptr)[i] -#define CVT_32_FLOAT ((float *) ptr)[i] +#define CVT_64_FLOAT(i) (float) ((double *) ptr)[i] +#define CVT_32_FLOAT(i) ((float *) ptr)[i] -#define CVT_8_USCALED (float) ((unsigned char *) ptr)[i] -#define CVT_16_USCALED (float) ((unsigned short *) ptr)[i] -#define CVT_32_USCALED (float) ((unsigned int *) ptr)[i] +#define CVT_8_USCALED(i) (float) ((unsigned char *) ptr)[i] +#define CVT_16_USCALED(i) (float) ((unsigned short *) ptr)[i] +#define CVT_32_USCALED(i) (float) ((unsigned int *) ptr)[i] -#define CVT_8_SSCALED (float) ((char *) ptr)[i] -#define CVT_16_SSCALED (float) ((short *) ptr)[i] -#define CVT_32_SSCALED (float) ((int *) ptr)[i] +#define CVT_8_SSCALED(i) (float) ((char *) ptr)[i] +#define CVT_16_SSCALED(i) (float) ((short *) ptr)[i] +#define CVT_32_SSCALED(i) (float) ((int *) ptr)[i] -#define CVT_8_UNORM (float) ((unsigned char *) ptr)[i] / 255.0f -#define CVT_16_UNORM (float) ((unsigned short *) ptr)[i] / 65535.0f -#define CVT_32_UNORM (float) ((unsigned int *) ptr)[i] / 4294967295.0f +#define CVT_8_UNORM(i) (float) ((unsigned char *) ptr)[i] / 255.0f +#define CVT_16_UNORM(i) (float) ((unsigned short *) ptr)[i] / 65535.0f +#define CVT_32_UNORM(i) (float) ((unsigned int *) ptr)[i] / 4294967295.0f -#define CVT_8_SNORM (float) ((char *) ptr)[i] / 127.0f -#define CVT_16_SNORM (float) ((short *) ptr)[i] / 32767.0f -#define CVT_32_SNORM (float) ((int *) ptr)[i] / 2147483647.0f +#define CVT_8_SNORM(i) (float) ((char *) ptr)[i] / 127.0f +#define CVT_16_SNORM(i) (float) ((short *) ptr)[i] / 32767.0f +#define CVT_32_SNORM(i) (float) ((int *) ptr)[i] / 2147483647.0f FETCH_ATTRIB( R64G64B64A64_FLOAT, 4, CVT_64_FLOAT ) FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) @@ -156,6 +156,16 @@ FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) +static void +fetch_B8G8R8A8_UNORM(const void *ptr, float *attrib) +{ + attrib[2] = CVT_8_UNORM(0); + attrib[1] = CVT_8_UNORM(1); + attrib[0] = CVT_8_UNORM(2); + attrib[3] = CVT_8_UNORM(3); +} + + static fetch_func get_fetch_func( enum pipe_format format ) { #if 0 @@ -296,6 +306,10 @@ static fetch_func get_fetch_func( enum pipe_format format ) case PIPE_FORMAT_A8R8G8B8_UNORM: return fetch_A8R8G8B8_UNORM; + + case PIPE_FORMAT_B8G8R8A8_UNORM: + return fetch_B8G8R8A8_UNORM; + case 0: return NULL; /* not sure why this is needed */ -- cgit v1.2.3 From 689e1c5d501eb2f557f85dd3279ac5d91e53b0ad Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 3 Mar 2008 17:49:38 +0100 Subject: win32: don't prepend all debug with gallium3d --- src/gallium/auxiliary/util/p_debug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index b9607a6ba7..0da3b66a4d 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -43,7 +43,7 @@ void debug_vprintf(const char *format, va_list ap) { #ifdef WIN32 - EngDebugPrint("Gallium3D: ", (PCHAR)format, ap); + EngDebugPrint("", (PCHAR)format, ap); #else vfprintf(stderr, format, ap); #endif -- cgit v1.2.3 From 1cd2623a53cb02b491c725f101ee70824ab26a12 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 4 Mar 2008 18:30:12 +0100 Subject: gallium: Make scons build gallivm before other auxiliary modules. This ensures that the gallivm symbols referenced by the draw module are resolved properly. --- src/gallium/SConscript | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/SConscript b/src/gallium/SConscript index fa4833cbcf..f09778ce99 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -9,6 +9,9 @@ auxiliaries = [] Export('auxiliaries') +if llvm: + SConscript(['auxiliary/gallivm/SConscript']) + SConscript([ # NOTE: order matters! 'auxiliary/util/SConscript', @@ -19,8 +22,5 @@ SConscript([ 'auxiliary/pipebuffer/SConscript', ]) -if llvm: - SConscript(['auxiliary/gallivm/SConscript']) - for driver in env['drivers']: SConscript(os.path.join('drivers', driver, 'SConscript')) -- cgit v1.2.3 From 5b9b5c850f3b34d8891890d808a520a64dd779bc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 4 Mar 2008 18:32:16 +0100 Subject: gallium: Fix scons condition for building the xlib winsys. --- src/gallium/winsys/SConscript | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript index 635a68eea2..f18d3bd2f8 100644 --- a/src/gallium/winsys/SConscript +++ b/src/gallium/winsys/SConscript @@ -5,7 +5,7 @@ if dri: 'dri/SConscript', ]) -if 'xlib' in env['drivers'] and not dri: +if 'xlib' in env['winsys'] and not dri: SConscript([ 'xlib/SConscript', ]) -- cgit v1.2.3 From e2a91c294edb198d6ebedd30fb4a39d722c8ee9f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 4 Mar 2008 18:32:43 +0100 Subject: gallium: Add missing file for scons build. --- src/mesa/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index a77c3de6c7..678e4ad0e8 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -161,6 +161,7 @@ STATETRACKER_SOURCES = [ 'state_tracker/st_extensions.c', 'state_tracker/st_format.c', 'state_tracker/st_framebuffer.c', + 'state_tracker/st_gen_mipmap.c', 'state_tracker/st_mesa_to_tgsi.c', 'state_tracker/st_program.c', 'state_tracker/st_texture.c', -- cgit v1.2.3 From 19cc2e363185b50d19e4f257dd3558896b9b49d3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 4 Mar 2008 17:55:02 +0100 Subject: draw: dont' compute clipmask or apply viewport when not clipping (rename bypass_clipping to coords_in_window_space? --- src/gallium/auxiliary/draw/draw_vs_exec.c | 38 +++++++++++++++++++------------ 1 file changed, 24 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 583812aadd..514303e0ea 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -132,20 +132,30 @@ vs_exec_run( struct draw_vertex_shader *shader, z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; + if (!draw->rasterizer->bypass_clipping) { + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + } + else { + vOut[j]->clipmask = 0; + vOut[j]->edgeflag = 1; + vOut[j]->data[0][0] = x; + vOut[j]->data[0][1] = y; + vOut[j]->data[0][2] = z; + vOut[j]->data[0][3] = w; + } /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. -- cgit v1.2.3 From b1922de9f3478869c6788ef4e954c06c20e7aa9c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 5 Mar 2008 11:38:21 +0100 Subject: gallium: Use custom vsnprintf in WINDDK. EngDebugPrint does not handle float point arguments, so we need to use our own vsnprintf implementation. --- src/gallium/auxiliary/util/SConscript | 1 + src/gallium/auxiliary/util/p_debug.c | 20 +- src/gallium/auxiliary/util/u_snprintf.c | 1482 +++++++++++++++++++++++++++++++ 3 files changed, 1502 insertions(+), 1 deletion(-) create mode 100644 src/gallium/auxiliary/util/u_snprintf.c (limited to 'src') diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 4717941434..642088b962 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -7,6 +7,7 @@ util = env.ConvenienceLibrary( 'p_tile.c', 'p_util.c', 'u_mm.c', + 'u_snprintf.c', ]) auxiliaries.insert(0, util) diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 0da3b66a4d..7b3c030956 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -40,10 +40,28 @@ #include "pipe/p_compiler.h" +#ifdef WIN32 +static INLINE void +rpl_EngDebugPrint(const char *format, ...) +{ + va_list ap; + va_start(ap, format); + EngDebugPrint("", (PCHAR)format, ap); + va_end(ap); +} + +int rpl_vsnprintf(char *, size_t, const char *, va_list); +#endif + + void debug_vprintf(const char *format, va_list ap) { #ifdef WIN32 - EngDebugPrint("", (PCHAR)format, ap); + /* EngDebugPrint does not handle float point arguments, so we need to use + * our own vsnprintf implementation */ + char buf[512 + 1]; + rpl_vsnprintf(buf, sizeof(buf), format, ap); + rpl_EngDebugPrint("%s", buf); #else vfprintf(stderr, format, ap); #endif diff --git a/src/gallium/auxiliary/util/u_snprintf.c b/src/gallium/auxiliary/util/u_snprintf.c new file mode 100644 index 0000000000..5e19a2ddb2 --- /dev/null +++ b/src/gallium/auxiliary/util/u_snprintf.c @@ -0,0 +1,1482 @@ +/* + * Copyright (c) 1995 Patrick Powell. + * + * This code is based on code written by Patrick Powell . + * It may be used for any purpose as long as this notice remains intact on all + * source code distributions. + */ + +/* + * Copyright (c) 2008 Holger Weiss. + * + * This version of the code is maintained by Holger Weiss . + * My changes to the code may freely be used, modified and/or redistributed for + * any purpose. It would be nice if additions and fixes to this file (including + * trivial code cleanups) would be sent back in order to let me include them in + * the version available at . + * However, this is not a requirement for using or redistributing (possibly + * modified) versions of this file, nor is leaving this notice intact mandatory. + */ + +/* + * History + * + * 2008-01-20 Holger Weiss for C99-snprintf 1.1: + * + * Fixed the detection of infinite floating point values on IRIX (and + * possibly other systems) and applied another few minor cleanups. + * + * 2008-01-06 Holger Weiss for C99-snprintf 1.0: + * + * Added a lot of new features, fixed many bugs, and incorporated various + * improvements done by Andrew Tridgell , Russ Allbery + * , Hrvoje Niksic , Damien Miller + * , and others for the Samba, INN, Wget, and OpenSSH + * projects. The additions include: support the "e", "E", "g", "G", and + * "F" conversion specifiers (and use conversion style "f" or "F" for the + * still unsupported "a" and "A" specifiers); support the "hh", "ll", "j", + * "t", and "z" length modifiers; support the "#" flag and the (non-C99) + * "'" flag; use localeconv(3) (if available) to get both the current + * locale's decimal point character and the separator between groups of + * digits; fix the handling of various corner cases of field width and + * precision specifications; fix various floating point conversion bugs; + * handle infinite and NaN floating point values; don't attempt to write to + * the output buffer (which may be NULL) if a size of zero was specified; + * check for integer overflow of the field width, precision, and return + * values and during the floating point conversion; use the OUTCHAR() macro + * instead of a function for better performance; provide asprintf(3) and + * vasprintf(3) functions; add new test cases. The replacement functions + * have been renamed to use an "rpl_" prefix, the function calls in the + * main project (and in this file) must be redefined accordingly for each + * replacement function which is needed (by using Autoconf or other means). + * Various other minor improvements have been applied and the coding style + * was cleaned up for consistency. + * + * 2007-07-23 Holger Weiss for Mutt 1.5.13: + * + * C99 compliant snprintf(3) and vsnprintf(3) functions return the number + * of characters that would have been written to a sufficiently sized + * buffer (excluding the '\0'). The original code simply returned the + * length of the resulting output string, so that's been fixed. + * + * 1998-03-05 Michael Elkins for Mutt 0.90.8: + * + * The original code assumed that both snprintf(3) and vsnprintf(3) were + * missing. Some systems only have snprintf(3) but not vsnprintf(3), so + * the code is now broken down under HAVE_SNPRINTF and HAVE_VSNPRINTF. + * + * 1998-01-27 Thomas Roessler for Mutt 0.89i: + * + * The PGP code was using unsigned hexadecimal formats. Unfortunately, + * unsigned formats simply didn't work. + * + * 1997-10-22 Brandon Long for Mutt 0.87.1: + * + * Ok, added some minimal floating point support, which means this probably + * requires libm on most operating systems. Don't yet support the exponent + * (e,E) and sigfig (g,G). Also, fmtint() was pretty badly broken, it just + * wasn't being exercised in ways which showed it, so that's been fixed. + * Also, formatted the code to Mutt conventions, and removed dead code left + * over from the original. Also, there is now a builtin-test, run with: + * gcc -DTEST_SNPRINTF -o snprintf snprintf.c -lm && ./snprintf + * + * 2996-09-15 Brandon Long for Mutt 0.43: + * + * This was ugly. It is still ugly. I opted out of floating point + * numbers, but the formatter understands just about everything from the + * normal C string format, at least as far as I can tell from the Solaris + * 2.5 printf(3S) man page. + */ + +/* + * ToDo + * + * - Add wide character support. + * - Add support for "%a" and "%A" conversions. + * - Create test routines which predefine the expected results. Our test cases + * usually expose bugs in system implementations rather than in ours :-) + */ + +/* + * Usage + * + * 1) The following preprocessor macros should be defined to 1 if the feature or + * file in question is available on the target system (by using Autoconf or + * other means), though basic functionality should be available as long as + * HAVE_STDARG_H and HAVE_STDLIB_H are defined correctly: + * + * HAVE_VSNPRINTF + * HAVE_SNPRINTF + * HAVE_VASPRINTF + * HAVE_ASPRINTF + * HAVE_STDARG_H + * HAVE_STDDEF_H + * HAVE_STDINT_H + * HAVE_STDLIB_H + * HAVE_INTTYPES_H + * HAVE_LOCALE_H + * HAVE_LOCALECONV + * HAVE_LCONV_DECIMAL_POINT + * HAVE_LCONV_THOUSANDS_SEP + * HAVE_LONG_DOUBLE + * HAVE_LONG_LONG_INT + * HAVE_UNSIGNED_LONG_LONG_INT + * HAVE_INTMAX_T + * HAVE_UINTMAX_T + * HAVE_UINTPTR_T + * HAVE_PTRDIFF_T + * HAVE_VA_COPY + * HAVE___VA_COPY + * + * 2) The calls to the functions which should be replaced must be redefined + * throughout the project files (by using Autoconf or other means): + * + * #define vsnprintf rpl_vsnprintf + * #define snprintf rpl_snprintf + * #define vasprintf rpl_vasprintf + * #define asprintf rpl_asprintf + * + * 3) The required replacement functions should be declared in some header file + * included throughout the project files: + * + * #if HAVE_CONFIG_H + * #include + * #endif + * #if HAVE_STDARG_H + * #include + * #if !HAVE_VSNPRINTF + * int rpl_vsnprintf(char *, size_t, const char *, va_list); + * #endif + * #if !HAVE_SNPRINTF + * int rpl_snprintf(char *, size_t, const char *, ...); + * #endif + * #if !HAVE_VASPRINTF + * int rpl_vasprintf(char **, const char *, va_list); + * #endif + * #if !HAVE_ASPRINTF + * int rpl_asprintf(char **, const char *, ...); + * #endif + * #endif + * + * Autoconf macros for handling step 1 and step 2 are available at + * . + */ + +#if HAVE_CONFIG_H +#include +#else +#ifdef WIN32 +#define vsnprintf rpl_vsnprintf +#define snprintf rpl_snprintf +#define HAVE_VSNPRINTF 0 +#define HAVE_SNPRINTF 0 +#define HAVE_VASPRINTF 1 /* not needed */ +#define HAVE_ASPRINTF 1 /* not needed */ +#define HAVE_STDARG_H 1 +#define HAVE_STDDEF_H 1 +#define HAVE_STDINT_H 0 +#define HAVE_STDLIB_H 1 +#define HAVE_INTTYPES_H 0 +#define HAVE_LOCALE_H 0 +#define HAVE_LOCALECONV 0 +#define HAVE_LCONV_DECIMAL_POINT 0 +#define HAVE_LCONV_THOUSANDS_SEP 0 +#define HAVE_LONG_DOUBLE 0 +#define HAVE_LONG_LONG_INT 1 +#define HAVE_UNSIGNED_LONG_LONG_INT 1 +#define HAVE_INTMAX_T 0 +#define HAVE_UINTMAX_T 0 +#define HAVE_UINTPTR_T 1 +#define HAVE_PTRDIFF_T 1 +#define HAVE_VA_COPY 0 +#define HAVE___VA_COPY 0 +#else +#define HAVE_VSNPRINTF 1 +#define HAVE_SNPRINTF 1 +#define HAVE_VASPRINTF 1 +#define HAVE_ASPRINTF 1 +#endif +#endif /* HAVE_CONFIG_H */ + +#if !HAVE_SNPRINTF || !HAVE_VSNPRINTF || !HAVE_ASPRINTF || !HAVE_VASPRINTF +#include /* For NULL, size_t, vsnprintf(3), and vasprintf(3). */ +#ifdef VA_START +#undef VA_START +#endif /* defined(VA_START) */ +#ifdef VA_SHIFT +#undef VA_SHIFT +#endif /* defined(VA_SHIFT) */ +#if HAVE_STDARG_H +#include +#define VA_START(ap, last) va_start(ap, last) +#define VA_SHIFT(ap, value, type) /* No-op for ANSI C. */ +#else /* Assume is available. */ +#include +#define VA_START(ap, last) va_start(ap) /* "last" is ignored. */ +#define VA_SHIFT(ap, value, type) value = va_arg(ap, type) +#endif /* HAVE_STDARG_H */ + +#if !HAVE_VASPRINTF +#if HAVE_STDLIB_H +#include /* For malloc(3). */ +#endif /* HAVE_STDLIB_H */ +#ifdef VA_COPY +#undef VA_COPY +#endif /* defined(VA_COPY) */ +#ifdef VA_END_COPY +#undef VA_END_COPY +#endif /* defined(VA_END_COPY) */ +#if HAVE_VA_COPY +#define VA_COPY(dest, src) va_copy(dest, src) +#define VA_END_COPY(ap) va_end(ap) +#elif HAVE___VA_COPY +#define VA_COPY(dest, src) __va_copy(dest, src) +#define VA_END_COPY(ap) va_end(ap) +#else +#define VA_COPY(dest, src) (void)mymemcpy(&dest, &src, sizeof(va_list)) +#define VA_END_COPY(ap) /* No-op. */ +#define NEED_MYMEMCPY 1 +static void *mymemcpy(void *, void *, size_t); +#endif /* HAVE_VA_COPY */ +#endif /* !HAVE_VASPRINTF */ + +#if !HAVE_VSNPRINTF +#include /* For ERANGE and errno. */ +#include /* For *_MAX. */ +#if HAVE_INTTYPES_H +#include /* For intmax_t (if not defined in ). */ +#endif /* HAVE_INTTYPES_H */ +#if HAVE_LOCALE_H +#include /* For localeconv(3). */ +#endif /* HAVE_LOCALE_H */ +#if HAVE_STDDEF_H +#include /* For ptrdiff_t. */ +#endif /* HAVE_STDDEF_H */ +#if HAVE_STDINT_H +#include /* For intmax_t. */ +#endif /* HAVE_STDINT_H */ + +/* Support for unsigned long long int. We may also need ULLONG_MAX. */ +#ifndef ULONG_MAX /* We may need ULONG_MAX as a fallback. */ +#ifdef UINT_MAX +#define ULONG_MAX UINT_MAX +#else +#define ULONG_MAX INT_MAX +#endif /* defined(UINT_MAX) */ +#endif /* !defined(ULONG_MAX) */ +#ifdef ULLONG +#undef ULLONG +#endif /* defined(ULLONG) */ +#if HAVE_UNSIGNED_LONG_LONG_INT +#define ULLONG unsigned long long int +#ifndef ULLONG_MAX +#define ULLONG_MAX ULONG_MAX +#endif /* !defined(ULLONG_MAX) */ +#else +#define ULLONG unsigned long int +#ifdef ULLONG_MAX +#undef ULLONG_MAX +#endif /* defined(ULLONG_MAX) */ +#define ULLONG_MAX ULONG_MAX +#endif /* HAVE_LONG_LONG_INT */ + +/* Support for uintmax_t. We also need UINTMAX_MAX. */ +#ifdef UINTMAX_T +#undef UINTMAX_T +#endif /* defined(UINTMAX_T) */ +#if HAVE_UINTMAX_T || defined(uintmax_t) +#define UINTMAX_T uintmax_t +#ifndef UINTMAX_MAX +#define UINTMAX_MAX ULLONG_MAX +#endif /* !defined(UINTMAX_MAX) */ +#else +#define UINTMAX_T ULLONG +#ifdef UINTMAX_MAX +#undef UINTMAX_MAX +#endif /* defined(UINTMAX_MAX) */ +#define UINTMAX_MAX ULLONG_MAX +#endif /* HAVE_UINTMAX_T || defined(uintmax_t) */ + +/* Support for long double. */ +#ifndef LDOUBLE +#if HAVE_LONG_DOUBLE +#define LDOUBLE long double +#else +#define LDOUBLE double +#endif /* HAVE_LONG_DOUBLE */ +#endif /* !defined(LDOUBLE) */ + +/* Support for long long int. */ +#ifndef LLONG +#if HAVE_LONG_LONG_INT +#define LLONG long long int +#else +#define LLONG long int +#endif /* HAVE_LONG_LONG_INT */ +#endif /* !defined(LLONG) */ + +/* Support for intmax_t. */ +#ifndef INTMAX_T +#if HAVE_INTMAX_T || defined(intmax_t) +#define INTMAX_T intmax_t +#else +#define INTMAX_T LLONG +#endif /* HAVE_INTMAX_T || defined(intmax_t) */ +#endif /* !defined(INTMAX_T) */ + +/* Support for uintptr_t. */ +#ifndef UINTPTR_T +#if HAVE_UINTPTR_T || defined(uintptr_t) +#define UINTPTR_T uintptr_t +#else +#define UINTPTR_T unsigned long int +#endif /* HAVE_UINTPTR_T || defined(uintptr_t) */ +#endif /* !defined(UINTPTR_T) */ + +/* Support for ptrdiff_t. */ +#ifndef PTRDIFF_T +#if HAVE_PTRDIFF_T || defined(ptrdiff_t) +#define PTRDIFF_T ptrdiff_t +#else +#define PTRDIFF_T long int +#endif /* HAVE_PTRDIFF_T || defined(ptrdiff_t) */ +#endif /* !defined(PTRDIFF_T) */ + +/* + * We need an unsigned integer type corresponding to ptrdiff_t (cf. C99: + * 7.19.6.1, 7). However, we'll simply use PTRDIFF_T and convert it to an + * unsigned type if necessary. This should work just fine in practice. + */ +#ifndef UPTRDIFF_T +#define UPTRDIFF_T PTRDIFF_T +#endif /* !defined(UPTRDIFF_T) */ + +/* + * We need a signed integer type corresponding to size_t (cf. C99: 7.19.6.1, 7). + * However, we'll simply use size_t and convert it to a signed type if + * necessary. This should work just fine in practice. + */ +#ifndef SSIZE_T +#define SSIZE_T size_t +#endif /* !defined(SSIZE_T) */ + +/* Either ERANGE or E2BIG should be available everywhere. */ +#ifndef ERANGE +#define ERANGE E2BIG +#endif /* !defined(ERANGE) */ +#ifndef EOVERFLOW +#define EOVERFLOW ERANGE +#endif /* !defined(EOVERFLOW) */ + +/* + * Buffer size to hold the octal string representation of UINT128_MAX without + * nul-termination ("3777777777777777777777777777777777777777777"). + */ +#ifdef MAX_CONVERT_LENGTH +#undef MAX_CONVERT_LENGTH +#endif /* defined(MAX_CONVERT_LENGTH) */ +#define MAX_CONVERT_LENGTH 43 + +/* Format read states. */ +#define PRINT_S_DEFAULT 0 +#define PRINT_S_FLAGS 1 +#define PRINT_S_WIDTH 2 +#define PRINT_S_DOT 3 +#define PRINT_S_PRECISION 4 +#define PRINT_S_MOD 5 +#define PRINT_S_CONV 6 + +/* Format flags. */ +#define PRINT_F_MINUS (1 << 0) +#define PRINT_F_PLUS (1 << 1) +#define PRINT_F_SPACE (1 << 2) +#define PRINT_F_NUM (1 << 3) +#define PRINT_F_ZERO (1 << 4) +#define PRINT_F_QUOTE (1 << 5) +#define PRINT_F_UP (1 << 6) +#define PRINT_F_UNSIGNED (1 << 7) +#define PRINT_F_TYPE_G (1 << 8) +#define PRINT_F_TYPE_E (1 << 9) + +/* Conversion flags. */ +#define PRINT_C_CHAR 1 +#define PRINT_C_SHORT 2 +#define PRINT_C_LONG 3 +#define PRINT_C_LLONG 4 +#define PRINT_C_LDOUBLE 5 +#define PRINT_C_SIZE 6 +#define PRINT_C_PTRDIFF 7 +#define PRINT_C_INTMAX 8 + +#ifndef MAX +#define MAX(x, y) ((x >= y) ? x : y) +#endif /* !defined(MAX) */ +#ifndef CHARTOINT +#define CHARTOINT(ch) (ch - '0') +#endif /* !defined(CHARTOINT) */ +#ifndef ISDIGIT +#define ISDIGIT(ch) ('0' <= (unsigned char)ch && (unsigned char)ch <= '9') +#endif /* !defined(ISDIGIT) */ +#ifndef ISNAN +#define ISNAN(x) (x != x) +#endif /* !defined(ISNAN) */ +#ifndef ISINF +#define ISINF(x) (x != 0.0 && x + x == x) +#endif /* !defined(ISINF) */ + +#ifdef OUTCHAR +#undef OUTCHAR +#endif /* defined(OUTCHAR) */ +#define OUTCHAR(str, len, size, ch) \ +do { \ + if (len + 1 < size) \ + str[len] = ch; \ + (len)++; \ +} while (/* CONSTCOND */ 0) + +static void fmtstr(char *, size_t *, size_t, const char *, int, int, int); +static void fmtint(char *, size_t *, size_t, INTMAX_T, int, int, int, int); +static void fmtflt(char *, size_t *, size_t, LDOUBLE, int, int, int, int *); +static void printsep(char *, size_t *, size_t); +static int getnumsep(int); +static int getexponent(LDOUBLE); +static int convert(UINTMAX_T, char *, size_t, int, int); +static UINTMAX_T cast(LDOUBLE); +static UINTMAX_T myround(LDOUBLE); +static LDOUBLE mypow10(int); + +extern int errno; + +int +rpl_vsnprintf(char *str, size_t size, const char *format, va_list args) +{ + LDOUBLE fvalue; + INTMAX_T value; + unsigned char cvalue; + const char *strvalue; + INTMAX_T *intmaxptr; + PTRDIFF_T *ptrdiffptr; + SSIZE_T *sizeptr; + LLONG *llongptr; + long int *longptr; + int *intptr; + short int *shortptr; + signed char *charptr; + size_t len = 0; + int overflow = 0; + int base = 0; + int cflags = 0; + int flags = 0; + int width = 0; + int precision = -1; + int state = PRINT_S_DEFAULT; + char ch = *format++; + + /* + * C99 says: "If `n' is zero, nothing is written, and `s' may be a null + * pointer." (7.19.6.5, 2) We're forgiving and allow a NULL pointer + * even if a size larger than zero was specified. At least NetBSD's + * snprintf(3) does the same, as well as other versions of this file. + * (Though some of these versions will write to a non-NULL buffer even + * if a size of zero was specified, which violates the standard.) + */ + if (str == NULL && size != 0) + size = 0; + + while (ch != '\0') + switch (state) { + case PRINT_S_DEFAULT: + if (ch == '%') + state = PRINT_S_FLAGS; + else + OUTCHAR(str, len, size, ch); + ch = *format++; + break; + case PRINT_S_FLAGS: + switch (ch) { + case '-': + flags |= PRINT_F_MINUS; + ch = *format++; + break; + case '+': + flags |= PRINT_F_PLUS; + ch = *format++; + break; + case ' ': + flags |= PRINT_F_SPACE; + ch = *format++; + break; + case '#': + flags |= PRINT_F_NUM; + ch = *format++; + break; + case '0': + flags |= PRINT_F_ZERO; + ch = *format++; + break; + case '\'': /* SUSv2 flag (not in C99). */ + flags |= PRINT_F_QUOTE; + ch = *format++; + break; + default: + state = PRINT_S_WIDTH; + break; + } + break; + case PRINT_S_WIDTH: + if (ISDIGIT(ch)) { + ch = CHARTOINT(ch); + if (width > (INT_MAX - ch) / 10) { + overflow = 1; + goto out; + } + width = 10 * width + ch; + ch = *format++; + } else if (ch == '*') { + /* + * C99 says: "A negative field width argument is + * taken as a `-' flag followed by a positive + * field width." (7.19.6.1, 5) + */ + if ((width = va_arg(args, int)) < 0) { + flags |= PRINT_F_MINUS; + width = -width; + } + ch = *format++; + state = PRINT_S_DOT; + } else + state = PRINT_S_DOT; + break; + case PRINT_S_DOT: + if (ch == '.') { + state = PRINT_S_PRECISION; + ch = *format++; + } else + state = PRINT_S_MOD; + break; + case PRINT_S_PRECISION: + if (precision == -1) + precision = 0; + if (ISDIGIT(ch)) { + ch = CHARTOINT(ch); + if (precision > (INT_MAX - ch) / 10) { + overflow = 1; + goto out; + } + precision = 10 * precision + ch; + ch = *format++; + } else if (ch == '*') { + /* + * C99 says: "A negative precision argument is + * taken as if the precision were omitted." + * (7.19.6.1, 5) + */ + if ((precision = va_arg(args, int)) < 0) + precision = -1; + ch = *format++; + state = PRINT_S_MOD; + } else + state = PRINT_S_MOD; + break; + case PRINT_S_MOD: + switch (ch) { + case 'h': + ch = *format++; + if (ch == 'h') { /* It's a char. */ + ch = *format++; + cflags = PRINT_C_CHAR; + } else + cflags = PRINT_C_SHORT; + break; + case 'l': + ch = *format++; + if (ch == 'l') { /* It's a long long. */ + ch = *format++; + cflags = PRINT_C_LLONG; + } else + cflags = PRINT_C_LONG; + break; + case 'L': + cflags = PRINT_C_LDOUBLE; + ch = *format++; + break; + case 'j': + cflags = PRINT_C_INTMAX; + ch = *format++; + break; + case 't': + cflags = PRINT_C_PTRDIFF; + ch = *format++; + break; + case 'z': + cflags = PRINT_C_SIZE; + ch = *format++; + break; + } + state = PRINT_S_CONV; + break; + case PRINT_S_CONV: + switch (ch) { + case 'd': + /* FALLTHROUGH */ + case 'i': + switch (cflags) { + case PRINT_C_CHAR: + value = (signed char)va_arg(args, int); + break; + case PRINT_C_SHORT: + value = (short int)va_arg(args, int); + break; + case PRINT_C_LONG: + value = va_arg(args, long int); + break; + case PRINT_C_LLONG: + value = va_arg(args, LLONG); + break; + case PRINT_C_SIZE: + value = va_arg(args, SSIZE_T); + break; + case PRINT_C_INTMAX: + value = va_arg(args, INTMAX_T); + break; + case PRINT_C_PTRDIFF: + value = va_arg(args, PTRDIFF_T); + break; + default: + value = va_arg(args, int); + break; + } + fmtint(str, &len, size, value, 10, width, + precision, flags); + break; + case 'X': + flags |= PRINT_F_UP; + /* FALLTHROUGH */ + case 'x': + base = 16; + /* FALLTHROUGH */ + case 'o': + if (base == 0) + base = 8; + /* FALLTHROUGH */ + case 'u': + if (base == 0) + base = 10; + flags |= PRINT_F_UNSIGNED; + switch (cflags) { + case PRINT_C_CHAR: + value = (unsigned char)va_arg(args, + unsigned int); + break; + case PRINT_C_SHORT: + value = (unsigned short int)va_arg(args, + unsigned int); + break; + case PRINT_C_LONG: + value = va_arg(args, unsigned long int); + break; + case PRINT_C_LLONG: + value = va_arg(args, ULLONG); + break; + case PRINT_C_SIZE: + value = va_arg(args, size_t); + break; + case PRINT_C_INTMAX: + value = va_arg(args, UINTMAX_T); + break; + case PRINT_C_PTRDIFF: + value = va_arg(args, UPTRDIFF_T); + break; + default: + value = va_arg(args, unsigned int); + break; + } + fmtint(str, &len, size, value, base, width, + precision, flags); + break; + case 'A': + /* Not yet supported, we'll use "%F". */ + /* FALLTHROUGH */ + case 'F': + flags |= PRINT_F_UP; + case 'a': + /* Not yet supported, we'll use "%f". */ + /* FALLTHROUGH */ + case 'f': + if (cflags == PRINT_C_LDOUBLE) + fvalue = va_arg(args, LDOUBLE); + else + fvalue = va_arg(args, double); + fmtflt(str, &len, size, fvalue, width, + precision, flags, &overflow); + if (overflow) + goto out; + break; + case 'E': + flags |= PRINT_F_UP; + /* FALLTHROUGH */ + case 'e': + flags |= PRINT_F_TYPE_E; + if (cflags == PRINT_C_LDOUBLE) + fvalue = va_arg(args, LDOUBLE); + else + fvalue = va_arg(args, double); + fmtflt(str, &len, size, fvalue, width, + precision, flags, &overflow); + if (overflow) + goto out; + break; + case 'G': + flags |= PRINT_F_UP; + /* FALLTHROUGH */ + case 'g': + flags |= PRINT_F_TYPE_G; + if (cflags == PRINT_C_LDOUBLE) + fvalue = va_arg(args, LDOUBLE); + else + fvalue = va_arg(args, double); + /* + * If the precision is zero, it is treated as + * one (cf. C99: 7.19.6.1, 8). + */ + if (precision == 0) + precision = 1; + fmtflt(str, &len, size, fvalue, width, + precision, flags, &overflow); + if (overflow) + goto out; + break; + case 'c': + cvalue = va_arg(args, int); + OUTCHAR(str, len, size, cvalue); + break; + case 's': + strvalue = va_arg(args, char *); + fmtstr(str, &len, size, strvalue, width, + precision, flags); + break; + case 'p': + /* + * C99 says: "The value of the pointer is + * converted to a sequence of printing + * characters, in an implementation-defined + * manner." (C99: 7.19.6.1, 8) + */ + if ((strvalue = va_arg(args, void *)) == NULL) + /* + * We use the glibc format. BSD prints + * "0x0", SysV "0". + */ + fmtstr(str, &len, size, "(nil)", width, + -1, flags); + else { + /* + * We use the BSD/glibc format. SysV + * omits the "0x" prefix (which we emit + * using the PRINT_F_NUM flag). + */ + flags |= PRINT_F_NUM; + flags |= PRINT_F_UNSIGNED; + fmtint(str, &len, size, + (UINTPTR_T)strvalue, 16, width, + precision, flags); + } + break; + case 'n': + switch (cflags) { + case PRINT_C_CHAR: + charptr = va_arg(args, signed char *); + *charptr = len; + break; + case PRINT_C_SHORT: + shortptr = va_arg(args, short int *); + *shortptr = len; + break; + case PRINT_C_LONG: + longptr = va_arg(args, long int *); + *longptr = len; + break; + case PRINT_C_LLONG: + llongptr = va_arg(args, LLONG *); + *llongptr = len; + break; + case PRINT_C_SIZE: + /* + * C99 says that with the "z" length + * modifier, "a following `n' conversion + * specifier applies to a pointer to a + * signed integer type corresponding to + * size_t argument." (7.19.6.1, 7) + */ + sizeptr = va_arg(args, SSIZE_T *); + *sizeptr = len; + break; + case PRINT_C_INTMAX: + intmaxptr = va_arg(args, INTMAX_T *); + *intmaxptr = len; + break; + case PRINT_C_PTRDIFF: + ptrdiffptr = va_arg(args, PTRDIFF_T *); + *ptrdiffptr = len; + break; + default: + intptr = va_arg(args, int *); + *intptr = len; + break; + } + break; + case '%': /* Print a "%" character verbatim. */ + OUTCHAR(str, len, size, ch); + break; + default: /* Skip other characters. */ + break; + } + ch = *format++; + state = PRINT_S_DEFAULT; + base = cflags = flags = width = 0; + precision = -1; + break; + } +out: + if (len < size) + str[len] = '\0'; + else if (size > 0) + str[size - 1] = '\0'; + + if (overflow || len >= INT_MAX) { + errno = overflow ? EOVERFLOW : ERANGE; + return -1; + } + return (int)len; +} + +static void +fmtstr(char *str, size_t *len, size_t size, const char *value, int width, + int precision, int flags) +{ + int padlen, strln; /* Amount to pad. */ + int noprecision = (precision == -1); + + if (value == NULL) /* We're forgiving. */ + value = "(null)"; + + /* If a precision was specified, don't read the string past it. */ + for (strln = 0; value[strln] != '\0' && + (noprecision || strln < precision); strln++) + continue; + + if ((padlen = width - strln) < 0) + padlen = 0; + if (flags & PRINT_F_MINUS) /* Left justify. */ + padlen = -padlen; + + while (padlen > 0) { /* Leading spaces. */ + OUTCHAR(str, *len, size, ' '); + padlen--; + } + while (*value != '\0' && (noprecision || precision-- > 0)) { + OUTCHAR(str, *len, size, *value); + value++; + } + while (padlen < 0) { /* Trailing spaces. */ + OUTCHAR(str, *len, size, ' '); + padlen++; + } +} + +static void +fmtint(char *str, size_t *len, size_t size, INTMAX_T value, int base, int width, + int precision, int flags) +{ + UINTMAX_T uvalue; + char iconvert[MAX_CONVERT_LENGTH]; + char sign = 0; + char hexprefix = 0; + int spadlen = 0; /* Amount to space pad. */ + int zpadlen = 0; /* Amount to zero pad. */ + int pos; + int separators = (flags & PRINT_F_QUOTE); + int noprecision = (precision == -1); + + if (flags & PRINT_F_UNSIGNED) + uvalue = value; + else { + uvalue = (value >= 0) ? value : -value; + if (value < 0) + sign = '-'; + else if (flags & PRINT_F_PLUS) /* Do a sign. */ + sign = '+'; + else if (flags & PRINT_F_SPACE) + sign = ' '; + } + + pos = convert(uvalue, iconvert, sizeof(iconvert), base, + flags & PRINT_F_UP); + + if (flags & PRINT_F_NUM && uvalue != 0) { + /* + * C99 says: "The result is converted to an `alternative form'. + * For `o' conversion, it increases the precision, if and only + * if necessary, to force the first digit of the result to be a + * zero (if the value and precision are both 0, a single 0 is + * printed). For `x' (or `X') conversion, a nonzero result has + * `0x' (or `0X') prefixed to it." (7.19.6.1, 6) + */ + switch (base) { + case 8: + if (precision <= pos) + precision = pos + 1; + break; + case 16: + hexprefix = (flags & PRINT_F_UP) ? 'X' : 'x'; + break; + } + } + + if (separators) /* Get the number of group separators we'll print. */ + separators = getnumsep(pos); + + zpadlen = precision - pos - separators; + spadlen = width /* Minimum field width. */ + - separators /* Number of separators. */ + - MAX(precision, pos) /* Number of integer digits. */ + - ((sign != 0) ? 1 : 0) /* Will we print a sign? */ + - ((hexprefix != 0) ? 2 : 0); /* Will we print a prefix? */ + + if (zpadlen < 0) + zpadlen = 0; + if (spadlen < 0) + spadlen = 0; + + /* + * C99 says: "If the `0' and `-' flags both appear, the `0' flag is + * ignored. For `d', `i', `o', `u', `x', and `X' conversions, if a + * precision is specified, the `0' flag is ignored." (7.19.6.1, 6) + */ + if (flags & PRINT_F_MINUS) /* Left justify. */ + spadlen = -spadlen; + else if (flags & PRINT_F_ZERO && noprecision) { + zpadlen += spadlen; + spadlen = 0; + } + while (spadlen > 0) { /* Leading spaces. */ + OUTCHAR(str, *len, size, ' '); + spadlen--; + } + if (sign != 0) /* Sign. */ + OUTCHAR(str, *len, size, sign); + if (hexprefix != 0) { /* A "0x" or "0X" prefix. */ + OUTCHAR(str, *len, size, '0'); + OUTCHAR(str, *len, size, hexprefix); + } + while (zpadlen > 0) { /* Leading zeros. */ + OUTCHAR(str, *len, size, '0'); + zpadlen--; + } + while (pos > 0) { /* The actual digits. */ + pos--; + OUTCHAR(str, *len, size, iconvert[pos]); + if (separators > 0 && pos > 0 && pos % 3 == 0) + printsep(str, len, size); + } + while (spadlen < 0) { /* Trailing spaces. */ + OUTCHAR(str, *len, size, ' '); + spadlen++; + } +} + +static void +fmtflt(char *str, size_t *len, size_t size, LDOUBLE fvalue, int width, + int precision, int flags, int *overflow) +{ + LDOUBLE ufvalue; + UINTMAX_T intpart; + UINTMAX_T fracpart; + UINTMAX_T mask; + const char *infnan = NULL; + char iconvert[MAX_CONVERT_LENGTH]; + char fconvert[MAX_CONVERT_LENGTH]; + char econvert[4]; /* "e-12" (without nul-termination). */ + char esign = 0; + char sign = 0; + int leadfraczeros = 0; + int exponent = 0; + int emitpoint = 0; + int omitzeros = 0; + int omitcount = 0; + int padlen = 0; + int epos = 0; + int fpos = 0; + int ipos = 0; + int separators = (flags & PRINT_F_QUOTE); + int estyle = (flags & PRINT_F_TYPE_E); +#if HAVE_LOCALECONV && HAVE_LCONV_DECIMAL_POINT + struct lconv *lc = localeconv(); +#endif /* HAVE_LOCALECONV && HAVE_LCONV_DECIMAL_POINT */ + + /* + * AIX' man page says the default is 0, but C99 and at least Solaris' + * and NetBSD's man pages say the default is 6, and sprintf(3) on AIX + * defaults to 6. + */ + if (precision == -1) + precision = 6; + + if (fvalue < 0.0) + sign = '-'; + else if (flags & PRINT_F_PLUS) /* Do a sign. */ + sign = '+'; + else if (flags & PRINT_F_SPACE) + sign = ' '; + + if (ISNAN(fvalue)) + infnan = (flags & PRINT_F_UP) ? "NAN" : "nan"; + else if (ISINF(fvalue)) + infnan = (flags & PRINT_F_UP) ? "INF" : "inf"; + + if (infnan != NULL) { + if (sign != 0) + iconvert[ipos++] = sign; + while (*infnan != '\0') + iconvert[ipos++] = *infnan++; + fmtstr(str, len, size, iconvert, width, ipos, flags); + return; + } + + /* "%e" (or "%E") or "%g" (or "%G") conversion. */ + if (flags & PRINT_F_TYPE_E || flags & PRINT_F_TYPE_G) { + if (flags & PRINT_F_TYPE_G) { + /* + * For "%g" (and "%G") conversions, the precision + * specifies the number of significant digits, which + * includes the digits in the integer part. The + * conversion will or will not be using "e-style" (like + * "%e" or "%E" conversions) depending on the precision + * and on the exponent. However, the exponent can be + * affected by rounding the converted value, so we'll + * leave this decision for later. Until then, we'll + * assume that we're going to do an "e-style" conversion + * (in order to get the exponent calculated). For + * "e-style", the precision must be decremented by one. + */ + precision--; + /* + * For "%g" (and "%G") conversions, trailing zeros are + * removed from the fractional portion of the result + * unless the "#" flag was specified. + */ + if (!(flags & PRINT_F_NUM)) + omitzeros = 1; + } + exponent = getexponent(fvalue); + estyle = 1; + } + +again: + /* + * Sorry, we only support 9, 19, or 38 digits (that is, the number of + * digits of the 32-bit, the 64-bit, or the 128-bit UINTMAX_MAX value + * minus one) past the decimal point due to our conversion method. + */ + switch (sizeof(UINTMAX_T)) { + case 16: + if (precision > 38) + precision = 38; + break; + case 8: + if (precision > 19) + precision = 19; + break; + default: + if (precision > 9) + precision = 9; + break; + } + + ufvalue = (fvalue >= 0.0) ? fvalue : -fvalue; + if (estyle) /* We want exactly one integer digit. */ + ufvalue /= mypow10(exponent); + + if ((intpart = cast(ufvalue)) == UINTMAX_MAX) { + *overflow = 1; + return; + } + + /* + * Factor of ten with the number of digits needed for the fractional + * part. For example, if the precision is 3, the mask will be 1000. + */ + mask = mypow10(precision); + /* + * We "cheat" by converting the fractional part to integer by + * multiplying by a factor of ten. + */ + if ((fracpart = myround(mask * (ufvalue - intpart))) >= mask) { + /* + * For example, ufvalue = 2.99962, intpart = 2, and mask = 1000 + * (because precision = 3). Now, myround(1000 * 0.99962) will + * return 1000. So, the integer part must be incremented by one + * and the fractional part must be set to zero. + */ + intpart++; + fracpart = 0; + if (estyle && intpart == 10) { + /* + * The value was rounded up to ten, but we only want one + * integer digit if using "e-style". So, the integer + * part must be set to one and the exponent must be + * incremented by one. + */ + intpart = 1; + exponent++; + } + } + + /* + * Now that we know the real exponent, we can check whether or not to + * use "e-style" for "%g" (and "%G") conversions. If we don't need + * "e-style", the precision must be adjusted and the integer and + * fractional parts must be recalculated from the original value. + * + * C99 says: "Let P equal the precision if nonzero, 6 if the precision + * is omitted, or 1 if the precision is zero. Then, if a conversion + * with style `E' would have an exponent of X: + * + * - if P > X >= -4, the conversion is with style `f' (or `F') and + * precision P - (X + 1). + * + * - otherwise, the conversion is with style `e' (or `E') and precision + * P - 1." (7.19.6.1, 8) + * + * Note that we had decremented the precision by one. + */ + if (flags & PRINT_F_TYPE_G && estyle && + precision + 1 > exponent && exponent >= -4) { + precision -= exponent; + estyle = 0; + goto again; + } + + if (estyle) { + if (exponent < 0) { + exponent = -exponent; + esign = '-'; + } else + esign = '+'; + + /* + * Convert the exponent. The sizeof(econvert) is 4. So, the + * econvert buffer can hold e.g. "e+99" and "e-99". We don't + * support an exponent which contains more than two digits. + * Therefore, the following stores are safe. + */ + epos = convert(exponent, econvert, 2, 10, 0); + /* + * C99 says: "The exponent always contains at least two digits, + * and only as many more digits as necessary to represent the + * exponent." (7.19.6.1, 8) + */ + if (epos == 1) + econvert[epos++] = '0'; + econvert[epos++] = esign; + econvert[epos++] = (flags & PRINT_F_UP) ? 'E' : 'e'; + } + + /* Convert the integer part and the fractional part. */ + ipos = convert(intpart, iconvert, sizeof(iconvert), 10, 0); + if (fracpart != 0) /* convert() would return 1 if fracpart == 0. */ + fpos = convert(fracpart, fconvert, sizeof(fconvert), 10, 0); + + leadfraczeros = precision - fpos; + + if (omitzeros) { + if (fpos > 0) /* Omit trailing fractional part zeros. */ + while (omitcount < fpos && fconvert[omitcount] == '0') + omitcount++; + else { /* The fractional part is zero, omit it completely. */ + omitcount = precision; + leadfraczeros = 0; + } + precision -= omitcount; + } + + /* + * Print a decimal point if either the fractional part is non-zero + * and/or the "#" flag was specified. + */ + if (precision > 0 || flags & PRINT_F_NUM) + emitpoint = 1; + if (separators) /* Get the number of group separators we'll print. */ + separators = getnumsep(ipos); + + padlen = width /* Minimum field width. */ + - ipos /* Number of integer digits. */ + - epos /* Number of exponent characters. */ + - precision /* Number of fractional digits. */ + - separators /* Number of group separators. */ + - (emitpoint ? 1 : 0) /* Will we print a decimal point? */ + - ((sign != 0) ? 1 : 0); /* Will we print a sign character? */ + + if (padlen < 0) + padlen = 0; + + /* + * C99 says: "If the `0' and `-' flags both appear, the `0' flag is + * ignored." (7.19.6.1, 6) + */ + if (flags & PRINT_F_MINUS) /* Left justifty. */ + padlen = -padlen; + else if (flags & PRINT_F_ZERO && padlen > 0) { + if (sign != 0) { /* Sign. */ + OUTCHAR(str, *len, size, sign); + sign = 0; + } + while (padlen > 0) { /* Leading zeros. */ + OUTCHAR(str, *len, size, '0'); + padlen--; + } + } + while (padlen > 0) { /* Leading spaces. */ + OUTCHAR(str, *len, size, ' '); + padlen--; + } + if (sign != 0) /* Sign. */ + OUTCHAR(str, *len, size, sign); + while (ipos > 0) { /* Integer part. */ + ipos--; + OUTCHAR(str, *len, size, iconvert[ipos]); + if (separators > 0 && ipos > 0 && ipos % 3 == 0) + printsep(str, len, size); + } + if (emitpoint) { /* Decimal point. */ +#if HAVE_LOCALECONV && HAVE_LCONV_DECIMAL_POINT + if (lc->decimal_point != NULL && *lc->decimal_point != '\0') + OUTCHAR(str, *len, size, *lc->decimal_point); + else /* We'll always print some decimal point character. */ +#endif /* HAVE_LOCALECONV && HAVE_LCONV_DECIMAL_POINT */ + OUTCHAR(str, *len, size, '.'); + } + while (leadfraczeros > 0) { /* Leading fractional part zeros. */ + OUTCHAR(str, *len, size, '0'); + leadfraczeros--; + } + while (fpos > omitcount) { /* The remaining fractional part. */ + fpos--; + OUTCHAR(str, *len, size, fconvert[fpos]); + } + while (epos > 0) { /* Exponent. */ + epos--; + OUTCHAR(str, *len, size, econvert[epos]); + } + while (padlen < 0) { /* Trailing spaces. */ + OUTCHAR(str, *len, size, ' '); + padlen++; + } +} + +static void +printsep(char *str, size_t *len, size_t size) +{ +#if HAVE_LOCALECONV && HAVE_LCONV_THOUSANDS_SEP + struct lconv *lc = localeconv(); + int i; + + if (lc->thousands_sep != NULL) + for (i = 0; lc->thousands_sep[i] != '\0'; i++) + OUTCHAR(str, *len, size, lc->thousands_sep[i]); + else +#endif /* HAVE_LOCALECONV && HAVE_LCONV_THOUSANDS_SEP */ + OUTCHAR(str, *len, size, ','); +} + +static int +getnumsep(int digits) +{ + int separators = (digits - ((digits % 3 == 0) ? 1 : 0)) / 3; +#if HAVE_LOCALECONV && HAVE_LCONV_THOUSANDS_SEP + int strln; + struct lconv *lc = localeconv(); + + /* We support an arbitrary separator length (including zero). */ + if (lc->thousands_sep != NULL) { + for (strln = 0; lc->thousands_sep[strln] != '\0'; strln++) + continue; + separators *= strln; + } +#endif /* HAVE_LOCALECONV && HAVE_LCONV_THOUSANDS_SEP */ + return separators; +} + +static int +getexponent(LDOUBLE value) +{ + LDOUBLE tmp = (value >= 0.0) ? value : -value; + int exponent = 0; + + /* + * We check for 99 > exponent > -99 in order to work around possible + * endless loops which could happen (at least) in the second loop (at + * least) if we're called with an infinite value. However, we checked + * for infinity before calling this function using our ISINF() macro, so + * this might be somewhat paranoid. + */ + while (tmp < 1.0 && tmp > 0.0 && --exponent > -99) + tmp *= 10; + while (tmp >= 10.0 && ++exponent < 99) + tmp /= 10; + + return exponent; +} + +static int +convert(UINTMAX_T value, char *buf, size_t size, int base, int caps) +{ + const char *digits = caps ? "0123456789ABCDEF" : "0123456789abcdef"; + size_t pos = 0; + + /* We return an unterminated buffer with the digits in reverse order. */ + do { + buf[pos++] = digits[value % base]; + value /= base; + } while (value != 0 && pos < size); + + return (int)pos; +} + +static UINTMAX_T +cast(LDOUBLE value) +{ + UINTMAX_T result; + + /* + * We check for ">=" and not for ">" because if UINTMAX_MAX cannot be + * represented exactly as an LDOUBLE value (but is less than LDBL_MAX), + * it may be increased to the nearest higher representable value for the + * comparison (cf. C99: 6.3.1.4, 2). It might then equal the LDOUBLE + * value although converting the latter to UINTMAX_T would overflow. + */ + if (value >= UINTMAX_MAX) + return UINTMAX_MAX; + + result = value; + /* + * At least on NetBSD/sparc64 3.0.2 and 4.99.30, casting long double to + * an integer type converts e.g. 1.9 to 2 instead of 1 (which violates + * the standard). Sigh. + */ + return (result <= value) ? result : result - 1; +} + +static UINTMAX_T +myround(LDOUBLE value) +{ + UINTMAX_T intpart = cast(value); + + return ((value -= intpart) < 0.5) ? intpart : intpart + 1; +} + +static LDOUBLE +mypow10(int exponent) +{ + LDOUBLE result = 1; + + while (exponent > 0) { + result *= 10; + exponent--; + } + while (exponent < 0) { + result /= 10; + exponent++; + } + return result; +} +#endif /* !HAVE_VSNPRINTF */ + +#if !HAVE_VASPRINTF +#if NEED_MYMEMCPY +void * +mymemcpy(void *dst, void *src, size_t len) +{ + const char *from = src; + char *to = dst; + + /* No need for optimization, we use this only to replace va_copy(3). */ + while (len-- > 0) + *to++ = *from++; + return dst; +} +#endif /* NEED_MYMEMCPY */ + +int +rpl_vasprintf(char **ret, const char *format, va_list ap) +{ + size_t size; + int len; + va_list aq; + + VA_COPY(aq, ap); + len = vsnprintf(NULL, 0, format, aq); + VA_END_COPY(aq); + if (len < 0 || (*ret = malloc(size = len + 1)) == NULL) + return -1; + return vsnprintf(*ret, size, format, ap); +} +#endif /* !HAVE_VASPRINTF */ + +#if !HAVE_SNPRINTF +#if HAVE_STDARG_H +int +rpl_snprintf(char *str, size_t size, const char *format, ...) +#else +int +rpl_snprintf(va_alist) va_dcl +#endif /* HAVE_STDARG_H */ +{ +#if !HAVE_STDARG_H + char *str; + size_t size; + char *format; +#endif /* HAVE_STDARG_H */ + va_list ap; + int len; + + VA_START(ap, format); + VA_SHIFT(ap, str, char *); + VA_SHIFT(ap, size, size_t); + VA_SHIFT(ap, format, const char *); + len = vsnprintf(str, size, format, ap); + va_end(ap); + return len; +} +#endif /* !HAVE_SNPRINTF */ + +#if !HAVE_ASPRINTF +#if HAVE_STDARG_H +int +rpl_asprintf(char **ret, const char *format, ...) +#else +int +rpl_asprintf(va_alist) va_dcl +#endif /* HAVE_STDARG_H */ +{ +#if !HAVE_STDARG_H + char **ret; + char *format; +#endif /* HAVE_STDARG_H */ + va_list ap; + int len; + + VA_START(ap, format); + VA_SHIFT(ap, ret, char **); + VA_SHIFT(ap, format, const char *); + len = vasprintf(ret, format, ap); + va_end(ap); + return len; +} +#endif /* !HAVE_ASPRINTF */ +#else /* Dummy declaration to avoid empty translation unit warnings. */ +int main(void); +#endif /* !HAVE_SNPRINTF || !HAVE_VSNPRINTF || !HAVE_ASPRINTF || [...] */ + + +/* vim: set joinspaces textwidth=80: */ -- cgit v1.2.3 From 4528287e040415c2071012d02f20979ff995c754 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 5 Mar 2008 10:50:14 +0100 Subject: gallium: michel's patch to rework texture/sampler binding interface Bind all the samplers/textures at once rather than piecemeal. This is easier for drivers to understand. --- src/gallium/auxiliary/draw/draw_aaline.c | 54 +- src/gallium/auxiliary/draw/draw_pstipple.c | 51 +- src/gallium/drivers/failover/fo_context.h | 7 +- src/gallium/drivers/failover/fo_state.c | 59 +- src/gallium/drivers/failover/fo_state_emit.c | 18 +- src/gallium/drivers/i915simple/i915_context.h | 675 +++++----- src/gallium/drivers/i915simple/i915_state.c | 46 +- src/gallium/drivers/i915simple/i915_state_emit.c | 6 - .../drivers/i915simple/i915_state_sampler.c | 6 +- src/gallium/drivers/i965simple/brw_context.h | 1365 ++++++++++---------- src/gallium/drivers/i965simple/brw_state.c | 896 ++++++------- .../drivers/i965simple/brw_wm_sampler_state.c | 3 +- .../drivers/i965simple/brw_wm_surface_state.c | 2 +- src/gallium/drivers/softpipe/sp_context.c | 486 +++---- src/gallium/drivers/softpipe/sp_context.h | 3 + src/gallium/drivers/softpipe/sp_quad_fs.c | 417 +++--- src/gallium/drivers/softpipe/sp_state.h | 390 +++--- src/gallium/drivers/softpipe/sp_state_sampler.c | 56 +- src/gallium/drivers/softpipe/sp_texture.c | 404 +++--- src/gallium/include/pipe/p_context.h | 8 +- src/mesa/state_tracker/st_atom_sampler.c | 16 +- src/mesa/state_tracker/st_atom_texture.c | 15 +- src/mesa/state_tracker/st_cb_drawpixels.c | 11 +- src/mesa/state_tracker/st_context.h | 5 +- src/mesa/state_tracker/st_gen_mipmap.c | 11 +- 25 files changed, 2561 insertions(+), 2449 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index 7660e56fe6..3ec73b0800 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -78,7 +78,8 @@ struct aaline_stage void *sampler_cso; struct pipe_texture *texture; - uint sampler_unit; + uint num_samplers; + uint num_textures; /* @@ -98,11 +99,10 @@ struct aaline_stage void (*driver_bind_fs_state)(struct pipe_context *, void *); void (*driver_delete_fs_state)(struct pipe_context *, void *); - void (*driver_bind_sampler_state)(struct pipe_context *, unsigned, void *); - - void (*driver_set_sampler_texture)(struct pipe_context *, - unsigned sampler, - struct pipe_texture *); + void (*driver_bind_sampler_states)(struct pipe_context *, unsigned, + void **); + void (*driver_set_sampler_textures)(struct pipe_context *, unsigned, + struct pipe_texture **); struct pipe_context *pipe; }; @@ -607,6 +607,7 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) auto struct aaline_stage *aaline = aaline_stage(stage); struct draw_context *draw = stage->draw; struct pipe_context *pipe = aaline->pipe; + uint num = MAX2(aaline->num_textures, aaline->num_samplers); assert(draw->rasterizer->line_smooth); @@ -624,8 +625,11 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) */ bind_aaline_fragment_shader(aaline); - aaline->driver_bind_sampler_state(pipe, aaline->sampler_unit, aaline->sampler_cso); - aaline->driver_set_sampler_texture(pipe, aaline->sampler_unit, aaline->texture); + aaline->state.sampler[num] = aaline->sampler_cso; + aaline->state.texture[num] = aaline->texture; + + aaline->driver_bind_sampler_states(pipe, num + 1, aaline->state.sampler); + aaline->driver_set_sampler_textures(pipe, num + 1, aaline->state.texture); /* now really draw first line */ stage->line = aaline_line; @@ -647,10 +651,10 @@ aaline_flush(struct draw_stage *stage, unsigned flags) aaline->driver_bind_fs_state(pipe, aaline->fs->driver_fs); /* XXX restore original texture, sampler state */ - aaline->driver_bind_sampler_state(pipe, aaline->sampler_unit, - aaline->state.sampler[aaline->sampler_unit]); - aaline->driver_set_sampler_texture(pipe, aaline->sampler_unit, - aaline->state.texture[aaline->sampler_unit]); + aaline->driver_bind_sampler_states(pipe, aaline->num_samplers, + aaline->state.sampler); + aaline->driver_set_sampler_textures(pipe, aaline->num_textures, + aaline->state.texture); draw->extra_vp_outputs.slot = 0; } @@ -745,26 +749,28 @@ aaline_delete_fs_state(struct pipe_context *pipe, void *fs) static void -aaline_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) +aaline_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **sampler) { struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); /* save current */ - aaline->state.sampler[unit] = sampler; + memcpy(aaline->state.sampler, sampler, num * sizeof(void *)); + aaline->num_samplers = num; /* pass-through */ - aaline->driver_bind_sampler_state(aaline->pipe, unit, sampler); + aaline->driver_bind_sampler_states(aaline->pipe, num, sampler); } static void -aaline_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, struct pipe_texture *texture) +aaline_set_sampler_textures(struct pipe_context *pipe, + unsigned num, struct pipe_texture **texture) { struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); /* save current */ - aaline->state.texture[sampler] = texture; + memcpy(aaline->state.texture, texture, num * sizeof(struct pipe_texture *)); + aaline->num_textures = num; /* pass-through */ - aaline->driver_set_sampler_texture(aaline->pipe, sampler, texture); + aaline->driver_set_sampler_textures(aaline->pipe, num, texture); } @@ -798,14 +804,14 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) aaline->driver_bind_fs_state = pipe->bind_fs_state; aaline->driver_delete_fs_state = pipe->delete_fs_state; - aaline->driver_bind_sampler_state = pipe->bind_sampler_state; - aaline->driver_set_sampler_texture = pipe->set_sampler_texture; + aaline->driver_bind_sampler_states = pipe->bind_sampler_states; + aaline->driver_set_sampler_textures = pipe->set_sampler_textures; /* override the driver's functions */ pipe->create_fs_state = aaline_create_fs_state; pipe->bind_fs_state = aaline_bind_fs_state; pipe->delete_fs_state = aaline_delete_fs_state; - pipe->bind_sampler_state = aaline_bind_sampler_state; - pipe->set_sampler_texture = aaline_set_sampler_texture; + pipe->bind_sampler_states = aaline_bind_sampler_states; + pipe->set_sampler_textures = aaline_set_sampler_textures; } diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index 2cfeb813b3..894b136f2c 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -68,7 +68,8 @@ struct pstip_stage void *sampler_cso; struct pipe_texture *texture; - uint sampler_unit; + uint num_samplers; + uint num_textures; /* * Currently bound state @@ -88,11 +89,10 @@ struct pstip_stage void (*driver_bind_fs_state)(struct pipe_context *, void *); void (*driver_delete_fs_state)(struct pipe_context *, void *); - void (*driver_bind_sampler_state)(struct pipe_context *, unsigned, void *); + void (*driver_bind_sampler_states)(struct pipe_context *, unsigned, void **); - void (*driver_set_sampler_texture)(struct pipe_context *, - unsigned sampler, - struct pipe_texture *); + void (*driver_set_sampler_textures)(struct pipe_context *, unsigned, + struct pipe_texture **); void (*driver_set_polygon_stipple)(struct pipe_context *, const struct pipe_poly_stipple *); @@ -328,8 +328,6 @@ generate_pstip_fs(struct pstip_stage *pstip) tgsi_dump(pstip_fs.tokens, 0); #endif - pstip->sampler_unit = transform.maxSampler + 1; - #if 1 /* XXX remove */ if (transform.wincoordInput < 0) { pstip_fs.input_semantic_name[pstip_fs.num_inputs] = TGSI_SEMANTIC_POSITION; @@ -486,6 +484,7 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) struct pstip_stage *pstip = pstip_stage(stage); struct draw_context *draw = stage->draw; struct pipe_context *pipe = pstip->pipe; + uint num = MAX2(pstip->num_textures, pstip->num_samplers); assert(draw->rasterizer->poly_stipple_enable); @@ -494,8 +493,8 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) */ bind_pstip_fragment_shader(pstip); - pstip->driver_bind_sampler_state(pipe, pstip->sampler_unit, pstip->sampler_cso); - pstip->driver_set_sampler_texture(pipe, pstip->sampler_unit, pstip->texture); + pstip->driver_bind_sampler_states(pipe, num + 1, pstip->state.sampler); + pstip->driver_set_sampler_textures(pipe, num + 1, pstip->state.texture); /* now really draw first line */ stage->tri = passthrough_tri; @@ -517,10 +516,10 @@ pstip_flush(struct draw_stage *stage, unsigned flags) pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs); /* XXX restore original texture, sampler state */ - pstip->driver_bind_sampler_state(pipe, pstip->sampler_unit, - pstip->state.sampler[pstip->sampler_unit]); - pstip->driver_set_sampler_texture(pipe, pstip->sampler_unit, - pstip->state.texture[pstip->sampler_unit]); + pstip->driver_bind_sampler_states(pipe, pstip->num_samplers, + pstip->state.sampler); + pstip->driver_set_sampler_textures(pipe, pstip->num_textures, + pstip->state.texture); } @@ -613,26 +612,28 @@ pstip_delete_fs_state(struct pipe_context *pipe, void *fs) static void -pstip_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) +pstip_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **sampler) { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); /* save current */ - pstip->state.sampler[unit] = sampler; + memcpy(pstip->state.sampler, sampler, num * sizeof(void *)); + pstip->num_samplers = num; /* pass-through */ - pstip->driver_bind_sampler_state(pstip->pipe, unit, sampler); + pstip->driver_bind_sampler_states(pstip->pipe, num, sampler); } static void -pstip_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, struct pipe_texture *texture) +pstip_set_sampler_textures(struct pipe_context *pipe, + unsigned num, struct pipe_texture **texture) { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); /* save current */ - pstip->state.texture[sampler] = texture; + memcpy(pstip->state.texture, texture, num * sizeof(struct pipe_texture *)); + pstip->num_textures = num; /* pass-through */ - pstip->driver_set_sampler_texture(pstip->pipe, sampler, texture); + pstip->driver_set_sampler_textures(pstip->pipe, num, texture); } @@ -682,8 +683,8 @@ draw_install_pstipple_stage(struct draw_context *draw, pstip->driver_bind_fs_state = pipe->bind_fs_state; pstip->driver_delete_fs_state = pipe->delete_fs_state; - pstip->driver_bind_sampler_state = pipe->bind_sampler_state; - pstip->driver_set_sampler_texture = pipe->set_sampler_texture; + pstip->driver_bind_sampler_states = pipe->bind_sampler_states; + pstip->driver_set_sampler_textures = pipe->set_sampler_textures; pstip->driver_set_polygon_stipple = pipe->set_polygon_stipple; /* override the driver's functions */ @@ -691,7 +692,7 @@ draw_install_pstipple_stage(struct draw_context *draw, pipe->bind_fs_state = pstip_bind_fs_state; pipe->delete_fs_state = pstip_delete_fs_state; - pipe->bind_sampler_state = pstip_bind_sampler_state; - pipe->set_sampler_texture = pstip_set_sampler_texture; + pipe->bind_sampler_states = pstip_bind_sampler_states; + pipe->set_sampler_textures = pstip_set_sampler_textures; pipe->set_polygon_stipple = pstip_set_polygon_stipple; } diff --git a/src/gallium/drivers/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h index 1dc87291c9..8f3ad3ee79 100644 --- a/src/gallium/drivers/failover/fo_context.h +++ b/src/gallium/drivers/failover/fo_context.h @@ -87,12 +87,15 @@ struct failover_context { struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + void *sw_sampler_state[PIPE_MAX_SAMPLERS]; + void *hw_sampler_state[PIPE_MAX_SAMPLERS]; + unsigned dirty; - unsigned dirty_sampler; - unsigned dirty_texture; unsigned dirty_vertex_buffer; unsigned dirty_vertex_element; + unsigned num_samplers; + unsigned num_textures; unsigned mode; struct pipe_context *hw; diff --git a/src/gallium/drivers/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c index 0fc5568da1..11eec2714e 100644 --- a/src/gallium/drivers/failover/fo_state.c +++ b/src/gallium/drivers/failover/fo_state.c @@ -28,6 +28,8 @@ /* Authors: Keith Whitwell */ +#include "pipe/p_inlines.h" + #include "fo_context.h" @@ -322,18 +324,27 @@ failover_create_sampler_state(struct pipe_context *pipe, } static void -failover_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) +failover_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **sampler) { struct failover_context *failover = failover_context(pipe); struct fo_state *state = (struct fo_state*)sampler; - failover->sampler[unit] = state; + uint i; + assert(num <= PIPE_MAX_SAMPLERS); + /* Check for no-op */ + if (num == failover->num_samplers && + !memcmp(failover->sampler, sampler, num * sizeof(void *))) + return; + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + failover->sw_sampler_state[i] = i < num ? state[i].sw_state : NULL; + failover->hw_sampler_state[i] = i < num ? state[i].hw_state : NULL; + } failover->dirty |= FO_NEW_SAMPLER; - failover->dirty_sampler |= (1<sw->bind_sampler_state(failover->sw, unit, - state->sw_state); - failover->hw->bind_sampler_state(failover->hw, unit, - state->hw_state); + failover->num_samplers = num; + failover->sw->bind_sampler_states(failover->sw, num, + failover->sw_sampler_state); + failover->hw->bind_sampler_states(failover->hw, num, + failover->hw_sampler_state); } static void @@ -351,17 +362,29 @@ failover_delete_sampler_state(struct pipe_context *pipe, void *sampler) static void -failover_set_sampler_texture(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) +failover_set_sampler_textures(struct pipe_context *pipe, + unsigned num, + struct pipe_texture **texture) { struct failover_context *failover = failover_context(pipe); - - failover->texture[unit] = texture; + uint i; + + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == failover->num_textures && + !memcmp(failover->texture, texture, num * sizeof(struct pipe_texture *))) + return; + for (i = 0; i < num; i++) + pipe_texture_reference((struct pipe_texture **) &failover->texture[i], + texture[i]); + for (i = num; i < failover->num_textures; i++) + pipe_texture_reference((struct pipe_texture **) &failover->texture[i], + NULL); failover->dirty |= FO_NEW_TEXTURE; - failover->dirty_texture |= (1<sw->set_sampler_texture( failover->sw, unit, texture ); - failover->hw->set_sampler_texture( failover->hw, unit, texture ); + failover->num_textures = num; + failover->sw->set_sampler_textures( failover->sw, num, texture ); + failover->hw->set_sampler_textures( failover->hw, num, texture ); } @@ -429,7 +452,7 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.bind_blend_state = failover_bind_blend_state; failover->pipe.delete_blend_state = failover_delete_blend_state; failover->pipe.create_sampler_state = failover_create_sampler_state; - failover->pipe.bind_sampler_state = failover_bind_sampler_state; + failover->pipe.bind_sampler_states = failover_bind_sampler_states; failover->pipe.delete_sampler_state = failover_delete_sampler_state; failover->pipe.create_depth_stencil_alpha_state = failover_create_depth_stencil_state; failover->pipe.bind_depth_stencil_alpha_state = failover_bind_depth_stencil_state; @@ -449,7 +472,7 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.set_framebuffer_state = failover_set_framebuffer_state; failover->pipe.set_polygon_stipple = failover_set_polygon_stipple; failover->pipe.set_scissor_state = failover_set_scissor_state; - failover->pipe.set_sampler_texture = failover_set_sampler_texture; + failover->pipe.set_sampler_textures = failover_set_sampler_textures; failover->pipe.set_viewport_state = failover_set_viewport_state; failover->pipe.set_vertex_buffer = failover_set_vertex_buffer; failover->pipe.set_vertex_element = failover_set_vertex_element; diff --git a/src/gallium/drivers/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c index c663dd4947..3de931e04e 100644 --- a/src/gallium/drivers/failover/fo_state_emit.c +++ b/src/gallium/drivers/failover/fo_state_emit.c @@ -94,21 +94,13 @@ failover_state_emit( struct failover_context *failover ) failover->sw->set_viewport_state( failover->sw, &failover->viewport ); if (failover->dirty & FO_NEW_SAMPLER) { - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - if (failover->dirty_sampler & (1<sw->bind_sampler_state( failover->sw, i, - failover->sampler[i]->sw_state ); - } - } + failover->sw->bind_sampler_states( failover->sw, failover->num_samplers, + failover->sw_sampler_state ); } if (failover->dirty & FO_NEW_TEXTURE) { - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - if (failover->dirty_texture & (1<sw->set_sampler_texture( failover->sw, i, - failover->texture[i] ); - } - } + failover->sw->set_sampler_textures( failover->sw, failover->num_textures, + failover->texture ); } if (failover->dirty & FO_NEW_VERTEX_BUFFER) { @@ -132,6 +124,4 @@ failover_state_emit( struct failover_context *failover ) failover->dirty = 0; failover->dirty_vertex_element = 0; failover->dirty_vertex_buffer = 0; - failover->dirty_texture = 0; - failover->dirty_sampler = 0; } diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 6401112f83..746f18ba38 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -1,336 +1,339 @@ - /************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef I915_CONTEXT_H -#define I915_CONTEXT_H - - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" - -#include "draw/draw_vertex.h" - -#include "tgsi/util/tgsi_scan.h" - - -#define I915_TEX_UNITS 8 - -#define I915_DYNAMIC_MODES4 0 -#define I915_DYNAMIC_DEPTHSCALE_0 1 /* just the header */ -#define I915_DYNAMIC_DEPTHSCALE_1 2 -#define I915_DYNAMIC_IAB 3 -#define I915_DYNAMIC_BC_0 4 /* just the header */ -#define I915_DYNAMIC_BC_1 5 -#define I915_DYNAMIC_BFO_0 6 -#define I915_DYNAMIC_BFO_1 7 -#define I915_DYNAMIC_STP_0 8 -#define I915_DYNAMIC_STP_1 9 -#define I915_DYNAMIC_SC_ENA_0 10 -#define I915_DYNAMIC_SC_RECT_0 11 -#define I915_DYNAMIC_SC_RECT_1 12 -#define I915_DYNAMIC_SC_RECT_2 13 -#define I915_MAX_DYNAMIC 14 - - -#define I915_IMMEDIATE_S0 0 -#define I915_IMMEDIATE_S1 1 -#define I915_IMMEDIATE_S2 2 -#define I915_IMMEDIATE_S3 3 -#define I915_IMMEDIATE_S4 4 -#define I915_IMMEDIATE_S5 5 -#define I915_IMMEDIATE_S6 6 -#define I915_IMMEDIATE_S7 7 -#define I915_MAX_IMMEDIATE 8 - -/* These must mach the order of LI0_STATE_* bits, as they will be used - * to generate hardware packets: - */ -#define I915_CACHE_STATIC 0 -#define I915_CACHE_DYNAMIC 1 /* handled specially */ -#define I915_CACHE_SAMPLER 2 -#define I915_CACHE_MAP 3 -#define I915_CACHE_PROGRAM 4 -#define I915_CACHE_CONSTANTS 5 -#define I915_MAX_CACHE 6 - -#define I915_MAX_CONSTANT 32 - - -/** See constant_flags[] below */ -#define I915_CONSTFLAG_USER 0x1f - - -/** - * Subclass of pipe_shader_state - */ -struct i915_fragment_shader -{ - struct pipe_shader_state state; - - struct tgsi_shader_info info; - - uint *program; - uint program_len; - - /** - * constants introduced during translation. - * These are placed at the end of the constant buffer and grow toward - * the beginning (eg: slot 31, 30 29, ...) - * User-provided constants start at 0. - * This allows both types of constants to co-exist (until there's too many) - * and doesn't require regenerating/changing the fragment program to - * shuffle constants around. - */ - uint num_constants; - float constants[I915_MAX_CONSTANT][4]; - - /** - * Status of each constant - * if I915_CONSTFLAG_PARAM, the value must be taken from the corresponding - * slot of the user's constant buffer. (set by pipe->set_constant_buffer()) - * Else, the bitmask indicates which components are occupied by immediates. - */ - ubyte constant_flags[I915_MAX_CONSTANT]; -}; - - -struct i915_cache_context; - -/* Use to calculate differences between state emitted to hardware and - * current driver-calculated state. - */ -struct i915_state -{ - unsigned immediate[I915_MAX_IMMEDIATE]; - unsigned dynamic[I915_MAX_DYNAMIC]; - - float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; - /** number of constants passed in through a constant buffer */ - uint num_user_constants[PIPE_SHADER_TYPES]; - - /* texture sampler state */ - unsigned sampler[I915_TEX_UNITS][3]; - unsigned sampler_enable_flags; - unsigned sampler_enable_nr; - - /* texture image buffers */ - unsigned texbuffer[I915_TEX_UNITS][2]; - - /** Describes the current hardware vertex layout */ - struct vertex_info vertex_info; - - unsigned id; /* track lost context events */ -}; - -struct i915_blend_state { - unsigned iab; - unsigned modes4; - unsigned LIS5; - unsigned LIS6; -}; - -struct i915_depth_stencil_state { - unsigned stencil_modes4; - unsigned bfo[2]; - unsigned stencil_LIS5; - unsigned depth_LIS6; -}; - -struct i915_rasterizer_state { - int light_twoside : 1; - unsigned st; - enum interp_mode color_interp; - - unsigned LIS4; - unsigned LIS7; - unsigned sc[1]; - - const struct pipe_rasterizer_state *templ; - - union { float f; unsigned u; } ds[2]; -}; - -struct i915_sampler_state { - unsigned state[3]; - const struct pipe_sampler_state *templ; -}; - - -struct i915_texture { - struct pipe_texture base; - - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ - unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; -}; - -struct i915_context -{ - struct pipe_context pipe; - struct i915_winsys *winsys; - struct draw_context *draw; - - /* The most recent drawing state as set by the driver: - */ - const struct i915_blend_state *blend; - const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct i915_depth_stencil_state *depth_stencil; - const struct i915_rasterizer_state *rasterizer; - - struct i915_fragment_shader *fs; - - struct pipe_blend_color blend_color; - struct pipe_clip_state clip; - struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; - struct pipe_framebuffer_state framebuffer; - struct pipe_poly_stipple poly_stipple; - struct pipe_scissor_state scissor; - struct i915_texture *texture[PIPE_MAX_SAMPLERS]; - struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - - unsigned dirty; - - unsigned *batch_start; - - /** Vertex buffer */ - struct pipe_buffer *vbo; - - struct i915_state current; - unsigned hardware_dirty; - - unsigned debug; -}; - -/* A flag for each state_tracker state object: - */ -#define I915_NEW_VIEWPORT 0x1 -#define I915_NEW_RASTERIZER 0x2 -#define I915_NEW_FS 0x4 -#define I915_NEW_BLEND 0x8 -#define I915_NEW_CLIP 0x10 -#define I915_NEW_SCISSOR 0x20 -#define I915_NEW_STIPPLE 0x40 -#define I915_NEW_FRAMEBUFFER 0x80 -#define I915_NEW_ALPHA_TEST 0x100 -#define I915_NEW_DEPTH_STENCIL 0x200 -#define I915_NEW_SAMPLER 0x400 -#define I915_NEW_TEXTURE 0x800 -#define I915_NEW_CONSTANTS 0x1000 -#define I915_NEW_VBO 0x2000 -#define I915_NEW_VS 0x4000 - - -/* Driver's internally generated state flags: - */ -#define I915_NEW_VERTEX_FORMAT 0x10000 - - -/* Dirty flags for hardware emit - */ -#define I915_HW_STATIC (1<set_constant_buffer()) + * Else, the bitmask indicates which components are occupied by immediates. + */ + ubyte constant_flags[I915_MAX_CONSTANT]; +}; + + +struct i915_cache_context; + +/* Use to calculate differences between state emitted to hardware and + * current driver-calculated state. + */ +struct i915_state +{ + unsigned immediate[I915_MAX_IMMEDIATE]; + unsigned dynamic[I915_MAX_DYNAMIC]; + + float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; + /** number of constants passed in through a constant buffer */ + uint num_user_constants[PIPE_SHADER_TYPES]; + + /* texture sampler state */ + unsigned sampler[I915_TEX_UNITS][3]; + unsigned sampler_enable_flags; + unsigned sampler_enable_nr; + + /* texture image buffers */ + unsigned texbuffer[I915_TEX_UNITS][2]; + + /** Describes the current hardware vertex layout */ + struct vertex_info vertex_info; + + unsigned id; /* track lost context events */ +}; + +struct i915_blend_state { + unsigned iab; + unsigned modes4; + unsigned LIS5; + unsigned LIS6; +}; + +struct i915_depth_stencil_state { + unsigned stencil_modes4; + unsigned bfo[2]; + unsigned stencil_LIS5; + unsigned depth_LIS6; +}; + +struct i915_rasterizer_state { + int light_twoside : 1; + unsigned st; + enum interp_mode color_interp; + + unsigned LIS4; + unsigned LIS7; + unsigned sc[1]; + + const struct pipe_rasterizer_state *templ; + + union { float f; unsigned u; } ds[2]; +}; + +struct i915_sampler_state { + unsigned state[3]; + const struct pipe_sampler_state *templ; +}; + + +struct i915_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; +}; + +struct i915_context +{ + struct pipe_context pipe; + struct i915_winsys *winsys; + struct draw_context *draw; + + /* The most recent drawing state as set by the driver: + */ + const struct i915_blend_state *blend; + const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct i915_depth_stencil_state *depth_stencil; + const struct i915_rasterizer_state *rasterizer; + + struct i915_fragment_shader *fs; + + struct pipe_blend_color blend_color; + struct pipe_clip_state clip; + struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; + struct pipe_framebuffer_state framebuffer; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct i915_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + + unsigned dirty; + + unsigned num_samplers; + unsigned num_textures; + + unsigned *batch_start; + + /** Vertex buffer */ + struct pipe_buffer *vbo; + + struct i915_state current; + unsigned hardware_dirty; + + unsigned debug; +}; + +/* A flag for each state_tracker state object: + */ +#define I915_NEW_VIEWPORT 0x1 +#define I915_NEW_RASTERIZER 0x2 +#define I915_NEW_FS 0x4 +#define I915_NEW_BLEND 0x8 +#define I915_NEW_CLIP 0x10 +#define I915_NEW_SCISSOR 0x20 +#define I915_NEW_STIPPLE 0x40 +#define I915_NEW_FRAMEBUFFER 0x80 +#define I915_NEW_ALPHA_TEST 0x100 +#define I915_NEW_DEPTH_STENCIL 0x200 +#define I915_NEW_SAMPLER 0x400 +#define I915_NEW_TEXTURE 0x800 +#define I915_NEW_CONSTANTS 0x1000 +#define I915_NEW_VBO 0x2000 +#define I915_NEW_VS 0x4000 + + +/* Driver's internally generated state flags: + */ +#define I915_NEW_VERTEX_FORMAT 0x10000 + + +/* Dirty flags for hardware emit + */ +#define I915_HW_STATIC (1<sampler[unit] = (const struct i915_sampler_state*)sampler; + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == i915->num_samplers && + !memcmp(i915->sampler, sampler, num * sizeof(void *))) + return; + + memcpy(i915->sampler, sampler, num * sizeof(void *)); + memset(&i915->sampler[num], 0, (PIPE_MAX_SAMPLERS - num) * sizeof(void *)); + + i915->num_samplers = num; i915->dirty |= I915_NEW_SAMPLER; } @@ -526,14 +535,29 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, } -static void i915_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, - struct pipe_texture *texture) +static void i915_set_sampler_textures(struct pipe_context *pipe, + unsigned num, + struct pipe_texture **texture) { struct i915_context *i915 = i915_context(pipe); + uint i; + + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == i915->num_textures && + !memcmp(i915->texture, texture, num * sizeof(struct pipe_texture *))) + return; + + for (i = 0; i < num; i++) + pipe_texture_reference((struct pipe_texture **) &i915->texture[i], + texture[i]); + + for (i = num; i < i915->num_textures; i++) + pipe_texture_reference((struct pipe_texture **) &i915->texture[i], + NULL); - pipe_texture_reference((struct pipe_texture **) &i915->texture[sampler], - texture); + i915->num_textures = num; i915->dirty |= I915_NEW_TEXTURE; } @@ -691,7 +715,7 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.delete_blend_state = i915_delete_blend_state; i915->pipe.create_sampler_state = i915_create_sampler_state; - i915->pipe.bind_sampler_state = i915_bind_sampler_state; + i915->pipe.bind_sampler_states = i915_bind_sampler_states; i915->pipe.delete_sampler_state = i915_delete_sampler_state; i915->pipe.create_depth_stencil_alpha_state = i915_create_depth_stencil_state; @@ -715,7 +739,7 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_polygon_stipple = i915_set_polygon_stipple; i915->pipe.set_scissor_state = i915_set_scissor_state; - i915->pipe.set_sampler_texture = i915_set_sampler_texture; + i915->pipe.set_sampler_textures = i915_set_sampler_textures; i915->pipe.set_viewport_state = i915_set_viewport_state; i915->pipe.set_vertex_buffer = i915_set_vertex_buffer; i915->pipe.set_vertex_element = i915_set_vertex_element; diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index 6bbaac4e34..a7498d22b7 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -267,12 +267,6 @@ i915_emit_hardware_state(struct i915_context *i915 ) /* 2 + I915_TEX_UNITS*3 dwords, I915_TEX_UNITS relocs */ if (i915->hardware_dirty & (I915_HW_MAP | I915_HW_SAMPLER)) { - /* XXX: we were refering to sampler state - * (current.sampler_enable_nr) below, but only checking - * I915_HW_MAP above. Should probably calculate the enabled - * flags separately - but there will be further rework of - * state so perhaps not necessary yet. - */ const uint nr = i915->current.sampler_enable_nr; if (nr) { const uint enabled = i915->current.sampler_enable_flags; diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index 9c1a5bbbd6..9dbb1b1b23 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -106,7 +106,8 @@ void i915_update_samplers( struct i915_context *i915 ) i915->current.sampler_enable_nr = 0; i915->current.sampler_enable_flags = 0x0; - for (unit = 0; unit < I915_TEX_UNITS; unit++) { + for (unit = 0; unit < i915->num_textures && unit < i915->num_samplers; + unit++) { /* determine unit enable/disable by looking for a bound texture */ /* could also examine the fragment program? */ if (i915->texture[unit]) { @@ -219,7 +220,8 @@ i915_update_textures(struct i915_context *i915) { uint unit; - for (unit = 0; unit < I915_TEX_UNITS; unit++) { + for (unit = 0; unit < i915->num_textures && unit < i915->num_samplers; + unit++) { /* determine unit enable/disable by looking for a bound texture */ /* could also examine the fragment program? */ if (i915->texture[unit]) { diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index 4da3a8cffc..b83a13c3b6 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -1,681 +1,684 @@ -/* - Copyright (C) Intel Corp. 2006. All Rights Reserved. - Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to - develop this 3D driver. - - Permission is hereby granted, free of charge, to any person obtaining - a copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sublicense, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial - portions of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, - EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE - LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION - OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION - WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - - **********************************************************************/ - /* - * Authors: - * Keith Whitwell - */ - - -#ifndef BRWCONTEXT_INC -#define BRWCONTEXT_INC - - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" - -#include "tgsi/util/tgsi_scan.h" - -#include "brw_structs.h" -#include "brw_winsys.h" - - -/* Glossary: - * - * URB - uniform resource buffer. A mid-sized buffer which is - * partitioned between the fixed function units and used for passing - * values (vertices, primitives, constants) between them. - * - * CURBE - constant URB entry. An urb region (entry) used to hold - * constant values which the fixed function units can be instructed to - * preload into the GRF when spawining a thread. - * - * VUE - vertex URB entry. An urb entry holding a vertex and usually - * a vertex header. The header contains control information and - * things like primitive type, Begin/end flags and clip codes. - * - * PUE - primitive URB entry. An urb entry produced by the setup (SF) - * unit holding rasterization and interpolation parameters. - * - * GRF - general register file. One of several register files - * addressable by programmed threads. The inputs (r0, payload, curbe, - * urb) of the thread are preloaded to this area before the thread is - * spawned. The registers are individually 8 dwords wide and suitable - * for general usage. Registers holding thread input values are not - * special and may be overwritten. - * - * MRF - message register file. Threads communicate (and terminate) - * by sending messages. Message parameters are placed in contigous - * MRF registers. All program output is via these messages. URB - * entries are populated by sending a message to the shared URB - * function containing the new data, together with a control word, - * often an unmodified copy of R0. - * - * R0 - GRF register 0. Typically holds control information used when - * sending messages to other threads. - * - * EU or GEN4 EU: The name of the programmable subsystem of the - * i965 hardware. Threads are executed by the EU, the registers - * described above are part of the EU architecture. - * - * Fixed function units: - * - * CS - Command streamer. Notional first unit, little software - * interaction. Holds the URB entries used for constant data, ie the - * CURBEs. - * - * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of - * this unit is responsible for pulling vertices out of vertex buffers - * in vram and injecting them into the processing pipe as VUEs. If - * enabled, it first passes them to a VS thread which is a good place - * for the driver to implement any active vertex shader. - * - * GS - Geometry Shader. This corresponds to a new DX10 concept. If - * enabled, incoming strips etc are passed to GS threads in individual - * line/triangle/point units. The GS thread may perform arbitary - * computation and emit whatever primtives with whatever vertices it - * chooses. This makes GS an excellent place to implement GL's - * unfilled polygon modes, though of course it is capable of much - * more. Additionally, GS is used to translate away primitives not - * handled by latter units, including Quads and Lineloops. - * - * CS - Clipper. Mesa's clipping algorithms are imported to run on - * this unit. The fixed function part performs cliptesting against - * the 6 fixed clipplanes and makes descisions on whether or not the - * incoming primitive needs to be passed to a thread for clipping. - * User clip planes are handled via cooperation with the VS thread. - * - * SF - Strips Fans or Setup: Triangles are prepared for - * rasterization. Interpolation coefficients are calculated. - * Flatshading and two-side lighting usually performed here. - * - * WM - Windower. Interpolation of vertex attributes performed here. - * Fragment shader implemented here. SIMD aspects of EU taken full - * advantage of, as pixels are processed in blocks of 16. - * - * CC - Color Calculator. No EU threads associated with this unit. - * Handles blending and (presumably) depth and stencil testing. - */ - -#define BRW_MAX_CURBE (32*16) - -struct brw_context; -struct brw_winsys; - - -/* Raised when we receive new state across the pipe interface: - */ -#define BRW_NEW_VIEWPORT 0x1 -#define BRW_NEW_RASTERIZER 0x2 -#define BRW_NEW_FS 0x4 -#define BRW_NEW_BLEND 0x8 -#define BRW_NEW_CLIP 0x10 -#define BRW_NEW_SCISSOR 0x20 -#define BRW_NEW_STIPPLE 0x40 -#define BRW_NEW_FRAMEBUFFER 0x80 -#define BRW_NEW_ALPHA_TEST 0x100 -#define BRW_NEW_DEPTH_STENCIL 0x200 -#define BRW_NEW_SAMPLER 0x400 -#define BRW_NEW_TEXTURE 0x800 -#define BRW_NEW_CONSTANTS 0x1000 -#define BRW_NEW_VBO 0x2000 -#define BRW_NEW_VS 0x4000 - -/* Raised for other internal events: - */ -#define BRW_NEW_URB_FENCE 0x10000 -#define BRW_NEW_PSP 0x20000 -#define BRW_NEW_CURBE_OFFSETS 0x40000 -#define BRW_NEW_REDUCED_PRIMITIVE 0x80000 -#define BRW_NEW_PRIMITIVE 0x100000 -#define BRW_NEW_SCENE 0x200000 -#define BRW_NEW_SF_LINKAGE 0x400000 - -extern int BRW_DEBUG; - -#define DEBUG_TEXTURE 0x1 -#define DEBUG_STATE 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_PRIMS 0x8 -#define DEBUG_VERTS 0x10 -#define DEBUG_FALLBACKS 0x20 -#define DEBUG_VERBOSE 0x40 -#define DEBUG_DRI 0x80 -#define DEBUG_DMA 0x100 -#define DEBUG_SANITY 0x200 -#define DEBUG_SYNC 0x400 -#define DEBUG_SLEEP 0x800 -#define DEBUG_PIXEL 0x1000 -#define DEBUG_STATS 0x2000 -#define DEBUG_TILE 0x4000 -#define DEBUG_SINGLE_THREAD 0x8000 -#define DEBUG_WM 0x10000 -#define DEBUG_URB 0x20000 -#define DEBUG_VS 0x40000 -#define DEBUG_BATCH 0x80000 -#define DEBUG_BUFMGR 0x100000 -#define DEBUG_BLIT 0x200000 -#define DEBUG_REGION 0x400000 -#define DEBUG_MIPTREE 0x800000 - -#define DBG(...) do { \ - if (BRW_DEBUG & FILE_DEBUG_FLAG) \ - brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ -} while(0) - -#define PRINT(...) do { \ - brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ -} while(0) - -struct brw_state_flags { - unsigned cache; - unsigned brw; -}; - - -struct brw_vertex_program { - struct pipe_shader_state program; - struct tgsi_shader_info info; - int id; -}; - - -struct brw_fragment_program { - struct pipe_shader_state program; - struct tgsi_shader_info info; - - boolean UsesDepth; /* XXX add this to tgsi_shader_info? */ - int id; -}; - - -struct pipe_setup_linkage { - struct { - unsigned vp_output:5; - unsigned interp_mode:4; - unsigned bf_vp_output:5; - } fp_input[PIPE_MAX_SHADER_INPUTS]; - - unsigned fp_input_count:5; - unsigned max_vp_output:5; -}; - - - -struct brw_texture { - struct pipe_texture base; - - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ - unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; -}; - -/* Data about a particular attempt to compile a program. Note that - * there can be many of these, each in a different GL state - * corresponding to a different brw_wm_prog_key struct, with different - * compiled programs: - */ -/* Data about a particular attempt to compile a program. Note that - * there can be many of these, each in a different GL state - * corresponding to a different brw_wm_prog_key struct, with different - * compiled programs: - */ - -struct brw_wm_prog_data { - unsigned curb_read_length; - unsigned urb_read_length; - - unsigned first_curbe_grf; - unsigned total_grf; - unsigned total_scratch; - - /* Internally generated constants for the CURBE. These are loaded - * ahead of the data from the constant buffer. - */ - const float internal_const[8]; - unsigned nr_internal_consts; - unsigned max_const; - - boolean error; -}; - -struct brw_sf_prog_data { - unsigned urb_read_length; - unsigned total_grf; - - /* Each vertex may have upto 12 attributes, 4 components each, - * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 - * rows. - * - * Actually we use 4 for each, so call it 12 rows. - */ - unsigned urb_entry_size; -}; - -struct brw_clip_prog_data { - unsigned curb_read_length; /* user planes? */ - unsigned clip_mode; - unsigned urb_read_length; - unsigned total_grf; -}; - -struct brw_gs_prog_data { - unsigned urb_read_length; - unsigned total_grf; -}; - -struct brw_vs_prog_data { - unsigned curb_read_length; - unsigned urb_read_length; - unsigned total_grf; - unsigned outputs_written; - - unsigned inputs_read; - - unsigned max_const; - - float imm_buf[PIPE_MAX_CONSTANT][4]; - unsigned num_imm; - unsigned num_consts; - - /* Used for calculating urb partitions: - */ - unsigned urb_entry_size; -}; - - -#define BRW_MAX_TEX_UNIT 8 -#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 - -/* Create a fixed sized struct for caching binding tables: - */ -struct brw_surface_binding_table { - unsigned surf_ss_offset[BRW_WM_MAX_SURF]; -}; - - -struct brw_cache; - -struct brw_mem_pool { - struct pipe_buffer *buffer; - - unsigned size; - unsigned offset; /* offset of first free byte */ - - struct brw_context *brw; -}; - -struct brw_cache_item { - unsigned hash; - unsigned key_size; /* for variable-sized keys */ - const void *key; - - unsigned offset; /* offset within pool's buffer */ - unsigned data_size; - - struct brw_cache_item *next; -}; - - - -struct brw_cache { - unsigned id; - - const char *name; - - struct brw_context *brw; - struct brw_mem_pool *pool; - - struct brw_cache_item **items; - unsigned size, n_items; - - unsigned key_size; /* for fixed-size keys */ - unsigned aux_size; - - unsigned last_addr; /* offset of active item */ -}; - - - - -/* Considered adding a member to this struct to document which flags - * an update might raise so that ordering of the state atoms can be - * checked or derived at runtime. Dropped the idea in favor of having - * a debug mode where the state is monitored for flags which are - * raised that have already been tested against. - */ -struct brw_tracked_state { - struct brw_state_flags dirty; - void (*update)( struct brw_context *brw ); -}; - - -/* Flags for brw->state.cache. - */ -#define CACHE_NEW_CC_VP (1< 32. Wouldn't life - * be easier if C allowed arrays of packed elements? - */ -#define ATTRIB_BIT_DWORDS ((PIPE_ATTRIB_MAX+31)/32) - - - - -struct brw_vertex_info { - unsigned varying; /* varying:1[PIPE_ATTRIB_MAX] */ - unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_ATTRIB_MAX] */ -}; - - - - - -struct brw_context -{ - struct pipe_context pipe; - struct brw_winsys *winsys; - - unsigned primitive; - unsigned reduced_primitive; - - boolean emit_state_always; - - struct { - struct brw_state_flags dirty; - } state; - - - struct { - const struct pipe_blend_state *Blend; - const struct pipe_depth_stencil_alpha_state *DepthStencil; - const struct pipe_poly_stipple *PolygonStipple; - const struct pipe_rasterizer_state *Raster; - const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; - const struct brw_vertex_program *VertexProgram; - const struct brw_fragment_program *FragmentProgram; - - struct pipe_clip_state Clip; - struct pipe_blend_color BlendColor; - struct pipe_scissor_state Scissor; - struct pipe_viewport_state Viewport; - struct pipe_framebuffer_state FrameBuffer; - - const struct pipe_constant_buffer *Constants[2]; - const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; - } attribs; - - struct brw_mem_pool pool[BRW_MAX_POOL]; - struct brw_cache cache[BRW_MAX_CACHE]; - struct brw_cached_batch_item *cached_batch_items; - - struct { - - /* Arrays with buffer objects to copy non-bufferobj arrays into - * for upload: - */ - const struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX]; - - struct brw_vertex_element_state inputs[PIPE_ATTRIB_MAX]; - -#define BRW_NR_UPLOAD_BUFS 17 -#define BRW_UPLOAD_INIT_SIZE (128*1024) - - /* Summary of size and varying of active arrays, so we can check - * for changes to this state: - */ - struct brw_vertex_info info; - } vb; - - - unsigned hardware_dirty; - unsigned dirty; - unsigned pci_id; - /* BRW_NEW_URB_ALLOCATIONS: - */ - struct { - unsigned vsize; /* vertex size plus header in urb registers */ - unsigned csize; /* constant buffer size in urb registers */ - unsigned sfsize; /* setup data size in urb registers */ - - boolean constrained; - - unsigned nr_vs_entries; - unsigned nr_gs_entries; - unsigned nr_clip_entries; - unsigned nr_sf_entries; - unsigned nr_cs_entries; - -/* unsigned vs_size; */ -/* unsigned gs_size; */ -/* unsigned clip_size; */ -/* unsigned sf_size; */ -/* unsigned cs_size; */ - - unsigned vs_start; - unsigned gs_start; - unsigned clip_start; - unsigned sf_start; - unsigned cs_start; - } urb; - - - /* BRW_NEW_CURBE_OFFSETS: - */ - struct { - unsigned wm_start; - unsigned wm_size; - unsigned clip_start; - unsigned clip_size; - unsigned vs_start; - unsigned vs_size; - unsigned total_size; - - unsigned gs_offset; - - float *last_buf; - unsigned last_bufsz; - } curbe; - - struct { - struct brw_vs_prog_data *prog_data; - - unsigned prog_gs_offset; - unsigned state_gs_offset; - } vs; - - struct { - struct brw_gs_prog_data *prog_data; - - boolean prog_active; - unsigned prog_gs_offset; - unsigned state_gs_offset; - } gs; - - struct { - struct brw_clip_prog_data *prog_data; - - unsigned prog_gs_offset; - unsigned vp_gs_offset; - unsigned state_gs_offset; - } clip; - - - struct { - struct brw_sf_prog_data *prog_data; - - struct pipe_setup_linkage linkage; - - unsigned prog_gs_offset; - unsigned vp_gs_offset; - unsigned state_gs_offset; - } sf; - - struct { - struct brw_wm_prog_data *prog_data; - -// struct brw_wm_compiler *compile_data; - - - /** - * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER - * cache - */ - struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; - - unsigned render_surf; - unsigned nr_surfaces; - - unsigned max_threads; - struct pipe_buffer *scratch_buffer; - unsigned scratch_buffer_size; - - unsigned sampler_count; - unsigned sampler_gs_offset; - - struct brw_surface_binding_table bind; - unsigned bind_ss_offset; - - unsigned prog_gs_offset; - unsigned state_gs_offset; - } wm; - - - struct { - unsigned vp_gs_offset; - unsigned state_gs_offset; - } cc; - - - /* Used to give every program string a unique id - */ - unsigned program_id; -}; - - -#define BRW_PACKCOLOR8888(r,g,b,a) ((r<<24) | (g<<16) | (b<<8) | a) - - -/*====================================================================== - * brw_vtbl.c - */ -void brw_do_flush( struct brw_context *brw, - unsigned flags ); - - -/*====================================================================== - * brw_state.c - */ -void brw_validate_state(struct brw_context *brw); -void brw_init_state(struct brw_context *brw); -void brw_destroy_state(struct brw_context *brw); - - -/*====================================================================== - * brw_tex.c - */ -void brwUpdateTextureState( struct brw_context *brw ); - - -/* brw_urb.c - */ -void brw_upload_urb_fence(struct brw_context *brw); - -void brw_upload_constant_buffer_state(struct brw_context *brw); - -void brw_init_surface_functions(struct brw_context *brw); -void brw_init_state_functions(struct brw_context *brw); -void brw_init_flush_functions(struct brw_context *brw); -void brw_init_string_functions(struct brw_context *brw); - -/*====================================================================== - * Inline conversion functions. These are better-typed than the - * macros used previously: - */ -static inline struct brw_context * -brw_context( struct pipe_context *ctx ) -{ - return (struct brw_context *)ctx; -} - -#endif - +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRWCONTEXT_INC +#define BRWCONTEXT_INC + + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "tgsi/util/tgsi_scan.h" + +#include "brw_structs.h" +#include "brw_winsys.h" + + +/* Glossary: + * + * URB - uniform resource buffer. A mid-sized buffer which is + * partitioned between the fixed function units and used for passing + * values (vertices, primitives, constants) between them. + * + * CURBE - constant URB entry. An urb region (entry) used to hold + * constant values which the fixed function units can be instructed to + * preload into the GRF when spawining a thread. + * + * VUE - vertex URB entry. An urb entry holding a vertex and usually + * a vertex header. The header contains control information and + * things like primitive type, Begin/end flags and clip codes. + * + * PUE - primitive URB entry. An urb entry produced by the setup (SF) + * unit holding rasterization and interpolation parameters. + * + * GRF - general register file. One of several register files + * addressable by programmed threads. The inputs (r0, payload, curbe, + * urb) of the thread are preloaded to this area before the thread is + * spawned. The registers are individually 8 dwords wide and suitable + * for general usage. Registers holding thread input values are not + * special and may be overwritten. + * + * MRF - message register file. Threads communicate (and terminate) + * by sending messages. Message parameters are placed in contigous + * MRF registers. All program output is via these messages. URB + * entries are populated by sending a message to the shared URB + * function containing the new data, together with a control word, + * often an unmodified copy of R0. + * + * R0 - GRF register 0. Typically holds control information used when + * sending messages to other threads. + * + * EU or GEN4 EU: The name of the programmable subsystem of the + * i965 hardware. Threads are executed by the EU, the registers + * described above are part of the EU architecture. + * + * Fixed function units: + * + * CS - Command streamer. Notional first unit, little software + * interaction. Holds the URB entries used for constant data, ie the + * CURBEs. + * + * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of + * this unit is responsible for pulling vertices out of vertex buffers + * in vram and injecting them into the processing pipe as VUEs. If + * enabled, it first passes them to a VS thread which is a good place + * for the driver to implement any active vertex shader. + * + * GS - Geometry Shader. This corresponds to a new DX10 concept. If + * enabled, incoming strips etc are passed to GS threads in individual + * line/triangle/point units. The GS thread may perform arbitary + * computation and emit whatever primtives with whatever vertices it + * chooses. This makes GS an excellent place to implement GL's + * unfilled polygon modes, though of course it is capable of much + * more. Additionally, GS is used to translate away primitives not + * handled by latter units, including Quads and Lineloops. + * + * CS - Clipper. Mesa's clipping algorithms are imported to run on + * this unit. The fixed function part performs cliptesting against + * the 6 fixed clipplanes and makes descisions on whether or not the + * incoming primitive needs to be passed to a thread for clipping. + * User clip planes are handled via cooperation with the VS thread. + * + * SF - Strips Fans or Setup: Triangles are prepared for + * rasterization. Interpolation coefficients are calculated. + * Flatshading and two-side lighting usually performed here. + * + * WM - Windower. Interpolation of vertex attributes performed here. + * Fragment shader implemented here. SIMD aspects of EU taken full + * advantage of, as pixels are processed in blocks of 16. + * + * CC - Color Calculator. No EU threads associated with this unit. + * Handles blending and (presumably) depth and stencil testing. + */ + +#define BRW_MAX_CURBE (32*16) + +struct brw_context; +struct brw_winsys; + + +/* Raised when we receive new state across the pipe interface: + */ +#define BRW_NEW_VIEWPORT 0x1 +#define BRW_NEW_RASTERIZER 0x2 +#define BRW_NEW_FS 0x4 +#define BRW_NEW_BLEND 0x8 +#define BRW_NEW_CLIP 0x10 +#define BRW_NEW_SCISSOR 0x20 +#define BRW_NEW_STIPPLE 0x40 +#define BRW_NEW_FRAMEBUFFER 0x80 +#define BRW_NEW_ALPHA_TEST 0x100 +#define BRW_NEW_DEPTH_STENCIL 0x200 +#define BRW_NEW_SAMPLER 0x400 +#define BRW_NEW_TEXTURE 0x800 +#define BRW_NEW_CONSTANTS 0x1000 +#define BRW_NEW_VBO 0x2000 +#define BRW_NEW_VS 0x4000 + +/* Raised for other internal events: + */ +#define BRW_NEW_URB_FENCE 0x10000 +#define BRW_NEW_PSP 0x20000 +#define BRW_NEW_CURBE_OFFSETS 0x40000 +#define BRW_NEW_REDUCED_PRIMITIVE 0x80000 +#define BRW_NEW_PRIMITIVE 0x100000 +#define BRW_NEW_SCENE 0x200000 +#define BRW_NEW_SF_LINKAGE 0x400000 + +extern int BRW_DEBUG; + +#define DEBUG_TEXTURE 0x1 +#define DEBUG_STATE 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_PRIMS 0x8 +#define DEBUG_VERTS 0x10 +#define DEBUG_FALLBACKS 0x20 +#define DEBUG_VERBOSE 0x40 +#define DEBUG_DRI 0x80 +#define DEBUG_DMA 0x100 +#define DEBUG_SANITY 0x200 +#define DEBUG_SYNC 0x400 +#define DEBUG_SLEEP 0x800 +#define DEBUG_PIXEL 0x1000 +#define DEBUG_STATS 0x2000 +#define DEBUG_TILE 0x4000 +#define DEBUG_SINGLE_THREAD 0x8000 +#define DEBUG_WM 0x10000 +#define DEBUG_URB 0x20000 +#define DEBUG_VS 0x40000 +#define DEBUG_BATCH 0x80000 +#define DEBUG_BUFMGR 0x100000 +#define DEBUG_BLIT 0x200000 +#define DEBUG_REGION 0x400000 +#define DEBUG_MIPTREE 0x800000 + +#define DBG(...) do { \ + if (BRW_DEBUG & FILE_DEBUG_FLAG) \ + brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ +} while(0) + +#define PRINT(...) do { \ + brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ +} while(0) + +struct brw_state_flags { + unsigned cache; + unsigned brw; +}; + + +struct brw_vertex_program { + struct pipe_shader_state program; + struct tgsi_shader_info info; + int id; +}; + + +struct brw_fragment_program { + struct pipe_shader_state program; + struct tgsi_shader_info info; + + boolean UsesDepth; /* XXX add this to tgsi_shader_info? */ + int id; +}; + + +struct pipe_setup_linkage { + struct { + unsigned vp_output:5; + unsigned interp_mode:4; + unsigned bf_vp_output:5; + } fp_input[PIPE_MAX_SHADER_INPUTS]; + + unsigned fp_input_count:5; + unsigned max_vp_output:5; +}; + + + +struct brw_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; +}; + +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ + +struct brw_wm_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + + unsigned first_curbe_grf; + unsigned total_grf; + unsigned total_scratch; + + /* Internally generated constants for the CURBE. These are loaded + * ahead of the data from the constant buffer. + */ + const float internal_const[8]; + unsigned nr_internal_consts; + unsigned max_const; + + boolean error; +}; + +struct brw_sf_prog_data { + unsigned urb_read_length; + unsigned total_grf; + + /* Each vertex may have upto 12 attributes, 4 components each, + * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 + * rows. + * + * Actually we use 4 for each, so call it 12 rows. + */ + unsigned urb_entry_size; +}; + +struct brw_clip_prog_data { + unsigned curb_read_length; /* user planes? */ + unsigned clip_mode; + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_gs_prog_data { + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_vs_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + unsigned total_grf; + unsigned outputs_written; + + unsigned inputs_read; + + unsigned max_const; + + float imm_buf[PIPE_MAX_CONSTANT][4]; + unsigned num_imm; + unsigned num_consts; + + /* Used for calculating urb partitions: + */ + unsigned urb_entry_size; +}; + + +#define BRW_MAX_TEX_UNIT 8 +#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 + +/* Create a fixed sized struct for caching binding tables: + */ +struct brw_surface_binding_table { + unsigned surf_ss_offset[BRW_WM_MAX_SURF]; +}; + + +struct brw_cache; + +struct brw_mem_pool { + struct pipe_buffer *buffer; + + unsigned size; + unsigned offset; /* offset of first free byte */ + + struct brw_context *brw; +}; + +struct brw_cache_item { + unsigned hash; + unsigned key_size; /* for variable-sized keys */ + const void *key; + + unsigned offset; /* offset within pool's buffer */ + unsigned data_size; + + struct brw_cache_item *next; +}; + + + +struct brw_cache { + unsigned id; + + const char *name; + + struct brw_context *brw; + struct brw_mem_pool *pool; + + struct brw_cache_item **items; + unsigned size, n_items; + + unsigned key_size; /* for fixed-size keys */ + unsigned aux_size; + + unsigned last_addr; /* offset of active item */ +}; + + + + +/* Considered adding a member to this struct to document which flags + * an update might raise so that ordering of the state atoms can be + * checked or derived at runtime. Dropped the idea in favor of having + * a debug mode where the state is monitored for flags which are + * raised that have already been tested against. + */ +struct brw_tracked_state { + struct brw_state_flags dirty; + void (*update)( struct brw_context *brw ); +}; + + +/* Flags for brw->state.cache. + */ +#define CACHE_NEW_CC_VP (1< 32. Wouldn't life + * be easier if C allowed arrays of packed elements? + */ +#define ATTRIB_BIT_DWORDS ((PIPE_ATTRIB_MAX+31)/32) + + + + +struct brw_vertex_info { + unsigned varying; /* varying:1[PIPE_ATTRIB_MAX] */ + unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_ATTRIB_MAX] */ +}; + + + + + +struct brw_context +{ + struct pipe_context pipe; + struct brw_winsys *winsys; + + unsigned primitive; + unsigned reduced_primitive; + + boolean emit_state_always; + + struct { + struct brw_state_flags dirty; + } state; + + + struct { + const struct pipe_blend_state *Blend; + const struct pipe_depth_stencil_alpha_state *DepthStencil; + const struct pipe_poly_stipple *PolygonStipple; + const struct pipe_rasterizer_state *Raster; + const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; + const struct brw_vertex_program *VertexProgram; + const struct brw_fragment_program *FragmentProgram; + + struct pipe_clip_state Clip; + struct pipe_blend_color BlendColor; + struct pipe_scissor_state Scissor; + struct pipe_viewport_state Viewport; + struct pipe_framebuffer_state FrameBuffer; + + const struct pipe_constant_buffer *Constants[2]; + const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; + } attribs; + + unsigned num_samplers; + unsigned num_textures; + + struct brw_mem_pool pool[BRW_MAX_POOL]; + struct brw_cache cache[BRW_MAX_CACHE]; + struct brw_cached_batch_item *cached_batch_items; + + struct { + + /* Arrays with buffer objects to copy non-bufferobj arrays into + * for upload: + */ + const struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX]; + + struct brw_vertex_element_state inputs[PIPE_ATTRIB_MAX]; + +#define BRW_NR_UPLOAD_BUFS 17 +#define BRW_UPLOAD_INIT_SIZE (128*1024) + + /* Summary of size and varying of active arrays, so we can check + * for changes to this state: + */ + struct brw_vertex_info info; + } vb; + + + unsigned hardware_dirty; + unsigned dirty; + unsigned pci_id; + /* BRW_NEW_URB_ALLOCATIONS: + */ + struct { + unsigned vsize; /* vertex size plus header in urb registers */ + unsigned csize; /* constant buffer size in urb registers */ + unsigned sfsize; /* setup data size in urb registers */ + + boolean constrained; + + unsigned nr_vs_entries; + unsigned nr_gs_entries; + unsigned nr_clip_entries; + unsigned nr_sf_entries; + unsigned nr_cs_entries; + +/* unsigned vs_size; */ +/* unsigned gs_size; */ +/* unsigned clip_size; */ +/* unsigned sf_size; */ +/* unsigned cs_size; */ + + unsigned vs_start; + unsigned gs_start; + unsigned clip_start; + unsigned sf_start; + unsigned cs_start; + } urb; + + + /* BRW_NEW_CURBE_OFFSETS: + */ + struct { + unsigned wm_start; + unsigned wm_size; + unsigned clip_start; + unsigned clip_size; + unsigned vs_start; + unsigned vs_size; + unsigned total_size; + + unsigned gs_offset; + + float *last_buf; + unsigned last_bufsz; + } curbe; + + struct { + struct brw_vs_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } vs; + + struct { + struct brw_gs_prog_data *prog_data; + + boolean prog_active; + unsigned prog_gs_offset; + unsigned state_gs_offset; + } gs; + + struct { + struct brw_clip_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } clip; + + + struct { + struct brw_sf_prog_data *prog_data; + + struct pipe_setup_linkage linkage; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } sf; + + struct { + struct brw_wm_prog_data *prog_data; + +// struct brw_wm_compiler *compile_data; + + + /** + * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER + * cache + */ + struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; + + unsigned render_surf; + unsigned nr_surfaces; + + unsigned max_threads; + struct pipe_buffer *scratch_buffer; + unsigned scratch_buffer_size; + + unsigned sampler_count; + unsigned sampler_gs_offset; + + struct brw_surface_binding_table bind; + unsigned bind_ss_offset; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } wm; + + + struct { + unsigned vp_gs_offset; + unsigned state_gs_offset; + } cc; + + + /* Used to give every program string a unique id + */ + unsigned program_id; +}; + + +#define BRW_PACKCOLOR8888(r,g,b,a) ((r<<24) | (g<<16) | (b<<8) | a) + + +/*====================================================================== + * brw_vtbl.c + */ +void brw_do_flush( struct brw_context *brw, + unsigned flags ); + + +/*====================================================================== + * brw_state.c + */ +void brw_validate_state(struct brw_context *brw); +void brw_init_state(struct brw_context *brw); +void brw_destroy_state(struct brw_context *brw); + + +/*====================================================================== + * brw_tex.c + */ +void brwUpdateTextureState( struct brw_context *brw ); + + +/* brw_urb.c + */ +void brw_upload_urb_fence(struct brw_context *brw); + +void brw_upload_constant_buffer_state(struct brw_context *brw); + +void brw_init_surface_functions(struct brw_context *brw); +void brw_init_state_functions(struct brw_context *brw); +void brw_init_flush_functions(struct brw_context *brw); +void brw_init_string_functions(struct brw_context *brw); + +/*====================================================================== + * Inline conversion functions. These are better-typed than the + * macros used previously: + */ +static inline struct brw_context * +brw_context( struct pipe_context *ctx ) +{ + return (struct brw_context *)ctx; +} + +#endif + diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 6744a8aa4f..f5efe9fc06 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -1,434 +1,462 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Zack Rusin - * Keith Whitwell - */ - - -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_dump.h" - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_state.h" -#include "brw_draw.h" - - -#define DUP( TYPE, VAL ) \ -do { \ - struct TYPE *x = malloc(sizeof(*x)); \ - memcpy(x, VAL, sizeof(*x) ); \ - return x; \ -} while (0) - -/************************************************************************ - * Blend - */ -static void * -brw_create_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend) -{ - DUP( pipe_blend_state, blend ); -} - -static void brw_bind_blend_state(struct pipe_context *pipe, - void *blend) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Blend = (struct pipe_blend_state*)blend; - brw->state.dirty.brw |= BRW_NEW_BLEND; -} - - -static void brw_delete_blend_state(struct pipe_context *pipe, void *blend) -{ - free(blend); -} - -static void brw_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.BlendColor = *blend_color; - - brw->state.dirty.brw |= BRW_NEW_BLEND; -} - -/************************************************************************ - * Sampler - */ - -static void * -brw_create_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) -{ - DUP( pipe_sampler_state, sampler ); -} - -static void brw_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Samplers[unit] = sampler; - brw->state.dirty.brw |= BRW_NEW_SAMPLER; -} - -static void brw_delete_sampler_state(struct pipe_context *pipe, - void *sampler) -{ - free(sampler); -} - - -/************************************************************************ - * Depth stencil - */ - -static void * -brw_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) -{ - DUP( pipe_depth_stencil_alpha_state, depth_stencil ); -} - -static void brw_bind_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; - - brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; -} - -static void brw_delete_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - free(depth_stencil); -} - -/************************************************************************ - * Scissor - */ -static void brw_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct brw_context *brw = brw_context(pipe); - - memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) ); - brw->state.dirty.brw |= BRW_NEW_SCISSOR; -} - - -/************************************************************************ - * Stipple - */ - -static void brw_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ -} - - -/************************************************************************ - * Fragment shader - */ - -static void * brw_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *shader) -{ - struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); - - /* XXX: Do I have to duplicate the tokens as well?? - */ - brw_fp->program = *shader; - brw_fp->id = brw_context(pipe)->program_id++; - - tgsi_scan_shader(shader->tokens, &brw_fp->info); - -#if 0 - brw_shader_info(shader->tokens, - &brw_fp->info2); -#endif - - tgsi_dump(shader->tokens, 0); - - - return (void *)brw_fp; -} - -static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader; - brw->state.dirty.brw |= BRW_NEW_FS; -} - -static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) -{ - FREE(shader); -} - - -/************************************************************************ - * Vertex shader and other TNL state - */ - -static void *brw_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *shader) -{ - struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); - - /* XXX: Do I have to duplicate the tokens as well?? - */ - brw_vp->program = *shader; - brw_vp->id = brw_context(pipe)->program_id++; - - tgsi_scan_shader(shader->tokens, &brw_vp->info); - -#if 0 - brw_shader_info(shader->tokens, - &brw_vp->info2); -#endif - tgsi_dump(shader->tokens, 0); - - return (void *)brw_vp; -} - -static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; - brw->state.dirty.brw |= BRW_NEW_VS; - - debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); -} - -static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) -{ - FREE(shader); -} - - -static void brw_set_clip_state( struct pipe_context *pipe, - const struct pipe_clip_state *clip ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Clip = *clip; -} - - -static void brw_set_viewport_state( struct pipe_context *pipe, - const struct pipe_viewport_state *viewport ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Viewport = *viewport; /* struct copy */ - brw->state.dirty.brw |= BRW_NEW_VIEWPORT; - - /* pass the viewport info to the draw module */ - //draw_set_viewport_state(brw->draw, viewport); -} - - -static void brw_set_vertex_buffer( struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer ) -{ - struct brw_context *brw = brw_context(pipe); - brw->vb.vbo_array[index] = buffer; -} - -static void brw_set_vertex_element(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_element *element) -{ - /* flush ? */ - struct brw_context *brw = brw_context(pipe); - - assert(index < PIPE_ATTRIB_MAX); - struct brw_vertex_element_state el; - memset(&el, 0, sizeof(el)); - - el.ve0.src_offset = element->src_offset; - el.ve0.src_format = brw_translate_surface_format(element->src_format); - el.ve0.valid = 1; - el.ve0.vertex_buffer_index = element->vertex_buffer_index; - - el.ve1.dst_offset = index * 4; - - el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; - - switch (element->nr_components) { - case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; - case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; - case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; - break; - } - - brw->vb.inputs[index] = el; -} - - - -/************************************************************************ - * Constant buffers - */ - -static void brw_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) -{ - struct brw_context *brw = brw_context(pipe); - - assert(buf == 0 || index == 0); - - brw->attribs.Constants[shader] = buf; - brw->state.dirty.brw |= BRW_NEW_CONSTANTS; -} - - -/************************************************************************ - * Texture surfaces - */ - - -static void brw_set_sampler_texture(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) -{ - struct brw_context *brw = brw_context(pipe); - - pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[unit], - texture); - - brw->state.dirty.brw |= BRW_NEW_TEXTURE; -} - - -/************************************************************************ - * Render targets, etc - */ - -static void brw_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.FrameBuffer = *fb; /* struct copy */ - - brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER; -} - - - -/************************************************************************ - * Rasterizer state - */ - -static void * -brw_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *rasterizer) -{ - DUP(pipe_rasterizer_state, rasterizer); -} - -static void brw_bind_rasterizer_state( struct pipe_context *pipe, - void *setup ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Raster = (struct pipe_rasterizer_state *)setup; - - /* Also pass-through to draw module: - */ - //draw_set_rasterizer_state(brw->draw, setup); - - brw->state.dirty.brw |= BRW_NEW_RASTERIZER; -} - -static void brw_delete_rasterizer_state(struct pipe_context *pipe, - void *setup) -{ - free(setup); -} - - - -void -brw_init_state_functions( struct brw_context *brw ) -{ - brw->pipe.create_blend_state = brw_create_blend_state; - brw->pipe.bind_blend_state = brw_bind_blend_state; - brw->pipe.delete_blend_state = brw_delete_blend_state; - - brw->pipe.create_sampler_state = brw_create_sampler_state; - brw->pipe.bind_sampler_state = brw_bind_sampler_state; - brw->pipe.delete_sampler_state = brw_delete_sampler_state; - - brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; - brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; - brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; - - brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; - brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; - brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state; - brw->pipe.create_fs_state = brw_create_fs_state; - brw->pipe.bind_fs_state = brw_bind_fs_state; - brw->pipe.delete_fs_state = brw_delete_fs_state; - brw->pipe.create_vs_state = brw_create_vs_state; - brw->pipe.bind_vs_state = brw_bind_vs_state; - brw->pipe.delete_vs_state = brw_delete_vs_state; - - brw->pipe.set_blend_color = brw_set_blend_color; - brw->pipe.set_clip_state = brw_set_clip_state; - brw->pipe.set_constant_buffer = brw_set_constant_buffer; - brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; - -// brw->pipe.set_feedback_state = brw_set_feedback_state; -// brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; - - brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; - brw->pipe.set_scissor_state = brw_set_scissor_state; - brw->pipe.set_sampler_texture = brw_set_sampler_texture; - brw->pipe.set_viewport_state = brw_set_viewport_state; - brw->pipe.set_vertex_buffer = brw_set_vertex_buffer; - brw->pipe.set_vertex_element = brw_set_vertex_element; -} +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Zack Rusin + * Keith Whitwell + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_dump.h" + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_state.h" +#include "brw_draw.h" + + +#define DUP( TYPE, VAL ) \ +do { \ + struct TYPE *x = malloc(sizeof(*x)); \ + memcpy(x, VAL, sizeof(*x) ); \ + return x; \ +} while (0) + +/************************************************************************ + * Blend + */ +static void * +brw_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + DUP( pipe_blend_state, blend ); +} + +static void brw_bind_blend_state(struct pipe_context *pipe, + void *blend) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Blend = (struct pipe_blend_state*)blend; + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + + +static void brw_delete_blend_state(struct pipe_context *pipe, void *blend) +{ + free(blend); +} + +static void brw_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.BlendColor = *blend_color; + + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + +/************************************************************************ + * Sampler + */ + +static void * +brw_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + DUP( pipe_sampler_state, sampler ); +} + +static void brw_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **sampler) +{ + struct brw_context *brw = brw_context(pipe); + + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == brw->num_samplers && + !memcmp(brw->attribs.Samplers, sampler, num * sizeof(void *))) + return; + + memcpy(brw->attribs.Samplers, sampler, num * sizeof(void *)); + memset(&brw->attribs.Samplers[num], 0, (PIPE_MAX_SAMPLERS - num) * + sizeof(void *)); + + brw->num_samplers = num; + + brw->state.dirty.brw |= BRW_NEW_SAMPLER; +} + +static void brw_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + free(sampler); +} + + +/************************************************************************ + * Depth stencil + */ + +static void * +brw_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil) +{ + DUP( pipe_depth_stencil_alpha_state, depth_stencil ); +} + +static void brw_bind_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; + + brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; +} + +static void brw_delete_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + free(depth_stencil); +} + +/************************************************************************ + * Scissor + */ +static void brw_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct brw_context *brw = brw_context(pipe); + + memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) ); + brw->state.dirty.brw |= BRW_NEW_SCISSOR; +} + + +/************************************************************************ + * Stipple + */ + +static void brw_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ +} + + +/************************************************************************ + * Fragment shader + */ + +static void * brw_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); + + /* XXX: Do I have to duplicate the tokens as well?? + */ + brw_fp->program = *shader; + brw_fp->id = brw_context(pipe)->program_id++; + + tgsi_scan_shader(shader->tokens, &brw_fp->info); + +#if 0 + brw_shader_info(shader->tokens, + &brw_fp->info2); +#endif + + tgsi_dump(shader->tokens, 0); + + + return (void *)brw_fp; +} + +static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader; + brw->state.dirty.brw |= BRW_NEW_FS; +} + +static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) +{ + FREE(shader); +} + + +/************************************************************************ + * Vertex shader and other TNL state + */ + +static void *brw_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); + + /* XXX: Do I have to duplicate the tokens as well?? + */ + brw_vp->program = *shader; + brw_vp->id = brw_context(pipe)->program_id++; + + tgsi_scan_shader(shader->tokens, &brw_vp->info); + +#if 0 + brw_shader_info(shader->tokens, + &brw_vp->info2); +#endif + tgsi_dump(shader->tokens, 0); + + return (void *)brw_vp; +} + +static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; + brw->state.dirty.brw |= BRW_NEW_VS; + + debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); +} + +static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) +{ + FREE(shader); +} + + +static void brw_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Clip = *clip; +} + + +static void brw_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Viewport = *viewport; /* struct copy */ + brw->state.dirty.brw |= BRW_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + //draw_set_viewport_state(brw->draw, viewport); +} + + +static void brw_set_vertex_buffer( struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_buffer *buffer ) +{ + struct brw_context *brw = brw_context(pipe); + brw->vb.vbo_array[index] = buffer; +} + +static void brw_set_vertex_element(struct pipe_context *pipe, + unsigned index, + const struct pipe_vertex_element *element) +{ + /* flush ? */ + struct brw_context *brw = brw_context(pipe); + + assert(index < PIPE_ATTRIB_MAX); + struct brw_vertex_element_state el; + memset(&el, 0, sizeof(el)); + + el.ve0.src_offset = element->src_offset; + el.ve0.src_format = brw_translate_surface_format(element->src_format); + el.ve0.valid = 1; + el.ve0.vertex_buffer_index = element->vertex_buffer_index; + + el.ve1.dst_offset = index * 4; + + el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; + + switch (element->nr_components) { + case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; + case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; + case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; + break; + } + + brw->vb.inputs[index] = el; +} + + + +/************************************************************************ + * Constant buffers + */ + +static void brw_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct brw_context *brw = brw_context(pipe); + + assert(buf == 0 || index == 0); + + brw->attribs.Constants[shader] = buf; + brw->state.dirty.brw |= BRW_NEW_CONSTANTS; +} + + +/************************************************************************ + * Texture surfaces + */ + + +static void brw_set_sampler_textures(struct pipe_context *pipe, + unsigned num, + struct pipe_texture **texture) +{ + struct brw_context *brw = brw_context(pipe); + uint i; + + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == brw->num_textures && + !memcmp(brw->attribs.Texture, texture, num * + sizeof(struct pipe_texture *))) + return; + + for (i = 0; i < num; i++) + pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], + texture[i]); + + for (i = num; i < brw->num_textures; i++) + pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], + NULL); + + brw->num_textures = num; + + brw->state.dirty.brw |= BRW_NEW_TEXTURE; +} + + +/************************************************************************ + * Render targets, etc + */ + +static void brw_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FrameBuffer = *fb; /* struct copy */ + + brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER; +} + + + +/************************************************************************ + * Rasterizer state + */ + +static void * +brw_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *rasterizer) +{ + DUP(pipe_rasterizer_state, rasterizer); +} + +static void brw_bind_rasterizer_state( struct pipe_context *pipe, + void *setup ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Raster = (struct pipe_rasterizer_state *)setup; + + /* Also pass-through to draw module: + */ + //draw_set_rasterizer_state(brw->draw, setup); + + brw->state.dirty.brw |= BRW_NEW_RASTERIZER; +} + +static void brw_delete_rasterizer_state(struct pipe_context *pipe, + void *setup) +{ + free(setup); +} + + + +void +brw_init_state_functions( struct brw_context *brw ) +{ + brw->pipe.create_blend_state = brw_create_blend_state; + brw->pipe.bind_blend_state = brw_bind_blend_state; + brw->pipe.delete_blend_state = brw_delete_blend_state; + + brw->pipe.create_sampler_state = brw_create_sampler_state; + brw->pipe.bind_sampler_states = brw_bind_sampler_states; + brw->pipe.delete_sampler_state = brw_delete_sampler_state; + + brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; + brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; + brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; + + brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; + brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; + brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state; + brw->pipe.create_fs_state = brw_create_fs_state; + brw->pipe.bind_fs_state = brw_bind_fs_state; + brw->pipe.delete_fs_state = brw_delete_fs_state; + brw->pipe.create_vs_state = brw_create_vs_state; + brw->pipe.bind_vs_state = brw_bind_vs_state; + brw->pipe.delete_vs_state = brw_delete_vs_state; + + brw->pipe.set_blend_color = brw_set_blend_color; + brw->pipe.set_clip_state = brw_set_clip_state; + brw->pipe.set_constant_buffer = brw_set_constant_buffer; + brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; + +// brw->pipe.set_feedback_state = brw_set_feedback_state; +// brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; + + brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; + brw->pipe.set_scissor_state = brw_set_scissor_state; + brw->pipe.set_sampler_textures = brw_set_sampler_textures; + brw->pipe.set_viewport_state = brw_set_viewport_state; + brw->pipe.set_vertex_buffer = brw_set_vertex_buffer; + brw->pipe.set_vertex_element = brw_set_vertex_element; +} diff --git a/src/gallium/drivers/i965simple/brw_wm_sampler_state.c b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c index de42ffc5b1..ff5ba7e7c7 100644 --- a/src/gallium/drivers/i965simple/brw_wm_sampler_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c @@ -235,7 +235,8 @@ static void upload_wm_samplers(struct brw_context *brw) unsigned sampler_count = 0; /* BRW_NEW_SAMPLER */ - for (unit = 0; unit < BRW_MAX_TEX_UNIT; unit++) { + for (unit = 0; unit < brw->num_textures && unit < brw->num_samplers; + unit++) { /* determine unit enable/disable by looking for a bound texture */ if (brw->attribs.Texture[unit]) { const struct pipe_sampler_state *sampler = brw->attribs.Samplers[unit]; diff --git a/src/gallium/drivers/i965simple/brw_wm_surface_state.c b/src/gallium/drivers/i965simple/brw_wm_surface_state.c index d16d919bce..853c743ccf 100644 --- a/src/gallium/drivers/i965simple/brw_wm_surface_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_surface_state.c @@ -237,7 +237,7 @@ static void upload_wm_surfaces(struct brw_context *brw ) /* BRW_NEW_TEXTURE */ - for (i = 0; i < BRW_MAX_TEX_UNIT; i++) { + for (i = 0; i < brw->num_textures && i < brw->num_samplers; i++) { const struct brw_texture *texUnit = brw->attribs.Texture[i]; if (texUnit && diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index fa16ed94e8..316ae552b8 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -1,243 +1,243 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - -#include "draw/draw_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "sp_clear.h" -#include "sp_context.h" -#include "sp_flush.h" -#include "sp_prim_setup.h" -#include "sp_prim_vbuf.h" -#include "sp_state.h" -#include "sp_surface.h" -#include "sp_tile_cache.h" -#include "sp_texture.h" -#include "sp_winsys.h" -#include "sp_query.h" - - - -/** - * Map any drawing surfaces which aren't already mapped - */ -void -softpipe_map_surfaces(struct softpipe_context *sp) -{ - unsigned i; - - for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); - } - - sp_tile_cache_map_surfaces(sp->zsbuf_cache); -} - - -/** - * Unmap any mapped drawing surfaces - */ -void -softpipe_unmap_surfaces(struct softpipe_context *sp) -{ - uint i; - - for (i = 0; i < sp->framebuffer.num_cbufs; i++) - sp_flush_tile_cache(sp, sp->cbuf_cache[i]); - sp_flush_tile_cache(sp, sp->zsbuf_cache); - - for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); - } - sp_tile_cache_unmap_surfaces(sp->zsbuf_cache); -} - - -static void softpipe_destroy( struct pipe_context *pipe ) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - struct pipe_winsys *ws = pipe->winsys; - uint i; - - draw_destroy( softpipe->draw ); - - softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); - softpipe->quad.earlyz->destroy( softpipe->quad.earlyz ); - softpipe->quad.shade->destroy( softpipe->quad.shade ); - softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); - softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); - softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); - softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); - softpipe->quad.coverage->destroy( softpipe->quad.coverage ); - softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); - softpipe->quad.blend->destroy( softpipe->quad.blend ); - softpipe->quad.colormask->destroy( softpipe->quad.colormask ); - softpipe->quad.output->destroy( softpipe->quad.output ); - - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) - sp_destroy_tile_cache(softpipe->cbuf_cache[i]); - sp_destroy_tile_cache(softpipe->zsbuf_cache); - - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) - sp_destroy_tile_cache(softpipe->tex_cache[i]); - - for (i = 0; i < Elements(softpipe->constants); i++) { - if (softpipe->constants[i].buffer) { - pipe_buffer_reference(ws, &softpipe->constants[i].buffer, NULL); - } - } - - FREE( softpipe ); -} - - -struct pipe_context * -softpipe_create( struct pipe_screen *screen, - struct pipe_winsys *pipe_winsys, - struct softpipe_winsys *softpipe_winsys ) -{ - struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); - uint i; - -#if defined(__i386__) || defined(__386__) - softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; -#else - softpipe->use_sse = FALSE; -#endif - - softpipe->dump_fs = GETENV( "GALLIUM_DUMP_FS" ) != NULL; - - softpipe->pipe.winsys = pipe_winsys; - softpipe->pipe.screen = screen; - softpipe->pipe.destroy = softpipe_destroy; - - /* state setters */ - softpipe->pipe.create_blend_state = softpipe_create_blend_state; - softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; - softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; - - softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; - softpipe->pipe.bind_sampler_state = softpipe_bind_sampler_state; - softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; - - softpipe->pipe.create_depth_stencil_alpha_state = softpipe_create_depth_stencil_state; - softpipe->pipe.bind_depth_stencil_alpha_state = softpipe_bind_depth_stencil_state; - softpipe->pipe.delete_depth_stencil_alpha_state = softpipe_delete_depth_stencil_state; - - softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; - softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; - softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; - - softpipe->pipe.create_fs_state = softpipe_create_fs_state; - softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; - softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; - - softpipe->pipe.create_vs_state = softpipe_create_vs_state; - softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; - softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; - - softpipe->pipe.set_blend_color = softpipe_set_blend_color; - softpipe->pipe.set_clip_state = softpipe_set_clip_state; - softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; - softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; - softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; - softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; - softpipe->pipe.set_sampler_texture = softpipe_set_sampler_texture; - softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; - - softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; - softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; - - softpipe->pipe.draw_arrays = softpipe_draw_arrays; - softpipe->pipe.draw_elements = softpipe_draw_elements; - - softpipe->pipe.clear = softpipe_clear; - softpipe->pipe.flush = softpipe_flush; - - softpipe_init_query_funcs( softpipe ); - softpipe_init_texture_funcs( softpipe ); - - /* - * Alloc caches for accessing drawing surfaces and textures. - * Must be before quad stage setup! - */ - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) - softpipe->cbuf_cache[i] = sp_create_tile_cache(); - softpipe->zsbuf_cache = sp_create_tile_cache(); - - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) - softpipe->tex_cache[i] = sp_create_tile_cache(); - - - /* setup quad rendering stages */ - softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); - softpipe->quad.earlyz = sp_quad_earlyz_stage(softpipe); - softpipe->quad.shade = sp_quad_shade_stage(softpipe); - softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); - softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); - softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); - softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); - softpipe->quad.coverage = sp_quad_coverage_stage(softpipe); - softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); - softpipe->quad.blend = sp_quad_blend_stage(softpipe); - softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); - softpipe->quad.output = sp_quad_output_stage(softpipe); - - softpipe->winsys = softpipe_winsys; - - /* - * Create drawing context and plug our rendering stage into it. - */ - softpipe->draw = draw_create(); - assert(softpipe->draw); - softpipe->setup = sp_draw_render_stage(softpipe); - - if (GETENV( "SP_VBUF" ) != NULL) { - sp_init_vbuf(softpipe); - } - else { - draw_set_rasterize_stage(softpipe->draw, softpipe->setup); - } - - /* plug in AA line/point stages */ - draw_install_aaline_stage(softpipe->draw, &softpipe->pipe); - draw_install_aapoint_stage(softpipe->draw, &softpipe->pipe); - -#if USE_DRAW_STAGE_PSTIPPLE - /* Do polygon stipple w/ texture map + frag prog? */ - draw_install_pstipple_stage(softpipe->draw, &softpipe->pipe); -#endif - - sp_init_surface_functions(softpipe); - - return &softpipe->pipe; -} +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#include "draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "sp_clear.h" +#include "sp_context.h" +#include "sp_flush.h" +#include "sp_prim_setup.h" +#include "sp_prim_vbuf.h" +#include "sp_state.h" +#include "sp_surface.h" +#include "sp_tile_cache.h" +#include "sp_texture.h" +#include "sp_winsys.h" +#include "sp_query.h" + + + +/** + * Map any drawing surfaces which aren't already mapped + */ +void +softpipe_map_surfaces(struct softpipe_context *sp) +{ + unsigned i; + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) { + sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); + } + + sp_tile_cache_map_surfaces(sp->zsbuf_cache); +} + + +/** + * Unmap any mapped drawing surfaces + */ +void +softpipe_unmap_surfaces(struct softpipe_context *sp) +{ + uint i; + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) + sp_flush_tile_cache(sp, sp->cbuf_cache[i]); + sp_flush_tile_cache(sp, sp->zsbuf_cache); + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) { + sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); + } + sp_tile_cache_unmap_surfaces(sp->zsbuf_cache); +} + + +static void softpipe_destroy( struct pipe_context *pipe ) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + struct pipe_winsys *ws = pipe->winsys; + uint i; + + draw_destroy( softpipe->draw ); + + softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); + softpipe->quad.earlyz->destroy( softpipe->quad.earlyz ); + softpipe->quad.shade->destroy( softpipe->quad.shade ); + softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); + softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); + softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); + softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); + softpipe->quad.coverage->destroy( softpipe->quad.coverage ); + softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); + softpipe->quad.blend->destroy( softpipe->quad.blend ); + softpipe->quad.colormask->destroy( softpipe->quad.colormask ); + softpipe->quad.output->destroy( softpipe->quad.output ); + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + sp_destroy_tile_cache(softpipe->cbuf_cache[i]); + sp_destroy_tile_cache(softpipe->zsbuf_cache); + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + sp_destroy_tile_cache(softpipe->tex_cache[i]); + + for (i = 0; i < Elements(softpipe->constants); i++) { + if (softpipe->constants[i].buffer) { + pipe_buffer_reference(ws, &softpipe->constants[i].buffer, NULL); + } + } + + FREE( softpipe ); +} + + +struct pipe_context * +softpipe_create( struct pipe_screen *screen, + struct pipe_winsys *pipe_winsys, + struct softpipe_winsys *softpipe_winsys ) +{ + struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); + uint i; + +#if defined(__i386__) || defined(__386__) + softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; +#else + softpipe->use_sse = FALSE; +#endif + + softpipe->dump_fs = GETENV( "GALLIUM_DUMP_FS" ) != NULL; + + softpipe->pipe.winsys = pipe_winsys; + softpipe->pipe.screen = screen; + softpipe->pipe.destroy = softpipe_destroy; + + /* state setters */ + softpipe->pipe.create_blend_state = softpipe_create_blend_state; + softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; + softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; + + softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; + softpipe->pipe.bind_sampler_states = softpipe_bind_sampler_states; + softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; + + softpipe->pipe.create_depth_stencil_alpha_state = softpipe_create_depth_stencil_state; + softpipe->pipe.bind_depth_stencil_alpha_state = softpipe_bind_depth_stencil_state; + softpipe->pipe.delete_depth_stencil_alpha_state = softpipe_delete_depth_stencil_state; + + softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; + softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; + softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; + + softpipe->pipe.create_fs_state = softpipe_create_fs_state; + softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; + softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; + + softpipe->pipe.create_vs_state = softpipe_create_vs_state; + softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; + softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; + + softpipe->pipe.set_blend_color = softpipe_set_blend_color; + softpipe->pipe.set_clip_state = softpipe_set_clip_state; + softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; + softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; + softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; + softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; + softpipe->pipe.set_sampler_textures = softpipe_set_sampler_textures; + softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; + + softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; + softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; + + softpipe->pipe.draw_arrays = softpipe_draw_arrays; + softpipe->pipe.draw_elements = softpipe_draw_elements; + + softpipe->pipe.clear = softpipe_clear; + softpipe->pipe.flush = softpipe_flush; + + softpipe_init_query_funcs( softpipe ); + softpipe_init_texture_funcs( softpipe ); + + /* + * Alloc caches for accessing drawing surfaces and textures. + * Must be before quad stage setup! + */ + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + softpipe->cbuf_cache[i] = sp_create_tile_cache(); + softpipe->zsbuf_cache = sp_create_tile_cache(); + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + softpipe->tex_cache[i] = sp_create_tile_cache(); + + + /* setup quad rendering stages */ + softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); + softpipe->quad.earlyz = sp_quad_earlyz_stage(softpipe); + softpipe->quad.shade = sp_quad_shade_stage(softpipe); + softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); + softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); + softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); + softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); + softpipe->quad.coverage = sp_quad_coverage_stage(softpipe); + softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); + softpipe->quad.blend = sp_quad_blend_stage(softpipe); + softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); + softpipe->quad.output = sp_quad_output_stage(softpipe); + + softpipe->winsys = softpipe_winsys; + + /* + * Create drawing context and plug our rendering stage into it. + */ + softpipe->draw = draw_create(); + assert(softpipe->draw); + softpipe->setup = sp_draw_render_stage(softpipe); + + if (GETENV( "SP_VBUF" ) != NULL) { + sp_init_vbuf(softpipe); + } + else { + draw_set_rasterize_stage(softpipe->draw, softpipe->setup); + } + + /* plug in AA line/point stages */ + draw_install_aaline_stage(softpipe->draw, &softpipe->pipe); + draw_install_aapoint_stage(softpipe->draw, &softpipe->pipe); + +#if USE_DRAW_STAGE_PSTIPPLE + /* Do polygon stipple w/ texture map + frag prog? */ + draw_install_pstipple_stage(softpipe->draw, &softpipe->pipe); +#endif + + sp_init_surface_functions(softpipe); + + return &softpipe->pipe; +} diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index feeafc7084..19e6cfaf02 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -81,6 +81,9 @@ struct softpipe_context { struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; unsigned dirty; + unsigned num_samplers; + unsigned num_textures; + /* Counter for occlusion queries. Note this supports overlapping * queries. */ diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 1fbb2e38c4..9198198db3 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -1,208 +1,209 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_headers.h" -#include "sp_quad.h" -#include "sp_texture.h" -#include "sp_tex_sample.h" - - -struct quad_shade_stage -{ - struct quad_stage stage; - struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; - struct tgsi_exec_machine machine; - struct tgsi_exec_vector *inputs, *outputs; - int colorOutSlot, depthOutSlot; -}; - - -/** cast wrapper */ -static INLINE struct quad_shade_stage * -quad_shade_stage(struct quad_stage *qs) -{ - return (struct quad_shade_stage *) qs; -} - - - -/** - * Execute fragment shader for the four fragments in the quad. - */ -static void -shade_quad( - struct quad_stage *qs, - struct quad_header *quad ) -{ - struct quad_shade_stage *qss = quad_shade_stage( qs ); - struct softpipe_context *softpipe = qs->softpipe; - struct tgsi_exec_machine *machine = &qss->machine; - - /* Consts do not require 16 byte alignment. */ - machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; - - machine->InterpCoefs = quad->coef; - - /* run shader */ - quad->mask &= softpipe->fs->run( softpipe->fs, - &qss->machine, - quad ); - - /* store result color */ - if (qss->colorOutSlot >= 0) { - /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] - == TGSI_SEMANTIC_COLOR); - memcpy( - quad->outputs.color, - &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); - } - - /* - * XXX the following code for updating quad->outputs.depth - * isn't really needed if we did early z testing. - */ - - /* store result Z */ - if (qss->depthOutSlot >= 0) { - /* output[slot] is new Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; - } - } - else { - /* copy input Z (which was interpolated by the executor) to output Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; - /* XXX not sure the above line is always correct. The following - * might be better: - quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i]; - */ - } - } - - /* shader may cull fragments */ - if( quad->mask ) { - qs->next->run( qs->next, quad ); - } -} - -/** - * Per-primitive (or per-begin?) setup - */ -static void shade_begin(struct quad_stage *qs) -{ - struct quad_shade_stage *qss = quad_shade_stage(qs); - struct softpipe_context *softpipe = qs->softpipe; - unsigned i; - - /* set TGSI sampler state that varies */ - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - qss->samplers[i].state = softpipe->sampler[i]; - qss->samplers[i].texture = softpipe->texture[i]; - } - - /* find output slots for depth, color */ - qss->colorOutSlot = -1; - qss->depthOutSlot = -1; - for (i = 0; i < qss->stage.softpipe->fs->info.num_outputs; i++) { - switch (qss->stage.softpipe->fs->info.output_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - qss->depthOutSlot = i; - break; - case TGSI_SEMANTIC_COLOR: - qss->colorOutSlot = i; - break; - } - } - - softpipe->fs->prepare( softpipe->fs, - &qss->machine, - qss->samplers ); - - qs->next->begin(qs->next); -} - - -static void shade_destroy(struct quad_stage *qs) -{ - struct quad_shade_stage *qss = (struct quad_shade_stage *) qs; - - tgsi_exec_machine_free_data(&qss->machine); - FREE( qss->inputs ); - FREE( qss->outputs ); - FREE( qs ); -} - - -struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) -{ - struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); - uint i; - - /* allocate storage for program inputs/outputs, aligned to 16 bytes */ - qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); - qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); - qss->machine.Inputs = align16(qss->inputs); - qss->machine.Outputs = align16(qss->outputs); - - qss->stage.softpipe = softpipe; - qss->stage.begin = shade_begin; - qss->stage.run = shade_quad; - qss->stage.destroy = shade_destroy; - - /* set TGSI sampler state that's constant */ - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - assert(softpipe->tex_cache[i]); - qss->samplers[i].get_samples = sp_get_samples; - qss->samplers[i].pipe = &softpipe->pipe; - qss->samplers[i].cache = softpipe->tex_cache[i]; - } - - tgsi_exec_machine_init( &qss->machine ); - - return &qss->stage; -} +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "sp_texture.h" +#include "sp_tex_sample.h" + + +struct quad_shade_stage +{ + struct quad_stage stage; + struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; + struct tgsi_exec_machine machine; + struct tgsi_exec_vector *inputs, *outputs; + int colorOutSlot, depthOutSlot; +}; + + +/** cast wrapper */ +static INLINE struct quad_shade_stage * +quad_shade_stage(struct quad_stage *qs) +{ + return (struct quad_shade_stage *) qs; +} + + + +/** + * Execute fragment shader for the four fragments in the quad. + */ +static void +shade_quad( + struct quad_stage *qs, + struct quad_header *quad ) +{ + struct quad_shade_stage *qss = quad_shade_stage( qs ); + struct softpipe_context *softpipe = qs->softpipe; + struct tgsi_exec_machine *machine = &qss->machine; + + /* Consts do not require 16 byte alignment. */ + machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; + + machine->InterpCoefs = quad->coef; + + /* run shader */ + quad->mask &= softpipe->fs->run( softpipe->fs, + &qss->machine, + quad ); + + /* store result color */ + if (qss->colorOutSlot >= 0) { + /* XXX need to handle multiple color outputs someday */ + assert(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] + == TGSI_SEMANTIC_COLOR); + memcpy( + quad->outputs.color, + &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); + } + + /* + * XXX the following code for updating quad->outputs.depth + * isn't really needed if we did early z testing. + */ + + /* store result Z */ + if (qss->depthOutSlot >= 0) { + /* output[slot] is new Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; + } + } + else { + /* copy input Z (which was interpolated by the executor) to output Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; + /* XXX not sure the above line is always correct. The following + * might be better: + quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i]; + */ + } + } + + /* shader may cull fragments */ + if( quad->mask ) { + qs->next->run( qs->next, quad ); + } +} + +/** + * Per-primitive (or per-begin?) setup + */ +static void shade_begin(struct quad_stage *qs) +{ + struct quad_shade_stage *qss = quad_shade_stage(qs); + struct softpipe_context *softpipe = qs->softpipe; + unsigned i; + unsigned num = MAX2(softpipe->num_textures, softpipe->num_samplers); + + /* set TGSI sampler state that varies */ + for (i = 0; i < num; i++) { + qss->samplers[i].state = softpipe->sampler[i]; + qss->samplers[i].texture = softpipe->texture[i]; + } + + /* find output slots for depth, color */ + qss->colorOutSlot = -1; + qss->depthOutSlot = -1; + for (i = 0; i < qss->stage.softpipe->fs->info.num_outputs; i++) { + switch (qss->stage.softpipe->fs->info.output_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + qss->depthOutSlot = i; + break; + case TGSI_SEMANTIC_COLOR: + qss->colorOutSlot = i; + break; + } + } + + softpipe->fs->prepare( softpipe->fs, + &qss->machine, + qss->samplers ); + + qs->next->begin(qs->next); +} + + +static void shade_destroy(struct quad_stage *qs) +{ + struct quad_shade_stage *qss = (struct quad_shade_stage *) qs; + + tgsi_exec_machine_free_data(&qss->machine); + FREE( qss->inputs ); + FREE( qss->outputs ); + FREE( qs ); +} + + +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) +{ + struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); + uint i; + + /* allocate storage for program inputs/outputs, aligned to 16 bytes */ + qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); + qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); + qss->machine.Inputs = align16(qss->inputs); + qss->machine.Outputs = align16(qss->outputs); + + qss->stage.softpipe = softpipe; + qss->stage.begin = shade_begin; + qss->stage.run = shade_quad; + qss->stage.destroy = shade_destroy; + + /* set TGSI sampler state that's constant */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + assert(softpipe->tex_cache[i]); + qss->samplers[i].get_samples = sp_get_samples; + qss->samplers[i].pipe = &softpipe->pipe; + qss->samplers[i].cache = softpipe->tex_cache[i]; + } + + tgsi_exec_machine_init( &qss->machine ); + + return &qss->stage; +} diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 3943d4ed2b..45d159143f 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -1,195 +1,195 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef SP_STATE_H -#define SP_STATE_H - -#include "pipe/p_state.h" -#include "tgsi/util/tgsi_scan.h" - - -#define SP_NEW_VIEWPORT 0x1 -#define SP_NEW_RASTERIZER 0x2 -#define SP_NEW_FS 0x4 -#define SP_NEW_BLEND 0x8 -#define SP_NEW_CLIP 0x10 -#define SP_NEW_SCISSOR 0x20 -#define SP_NEW_STIPPLE 0x40 -#define SP_NEW_FRAMEBUFFER 0x80 -#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100 -#define SP_NEW_CONSTANTS 0x200 -#define SP_NEW_SAMPLER 0x400 -#define SP_NEW_TEXTURE 0x800 -#define SP_NEW_VERTEX 0x1000 -#define SP_NEW_VS 0x2000 -#define SP_NEW_QUERY 0x4000 - - -struct tgsi_sampler; -struct tgsi_exec_machine; - - -/** Subclass of pipe_shader_state (though it doesn't really need to be). - * - * This is starting to look an awful lot like a quad pipeline stage... - */ -struct sp_fragment_shader { - struct pipe_shader_state shader; - - struct tgsi_shader_info info; - - void (*prepare)( const struct sp_fragment_shader *shader, - struct tgsi_exec_machine *machine, - struct tgsi_sampler *samplers); - - /* Run the shader - this interface will get cleaned up in the - * future: - */ - unsigned (*run)( const struct sp_fragment_shader *shader, - struct tgsi_exec_machine *machine, - struct quad_header *quad ); - - - void (*delete)( struct sp_fragment_shader * ); -}; - -struct vertex_info; - -/** Subclass of pipe_shader_state */ -struct sp_vertex_shader { - struct pipe_shader_state shader; - struct draw_vertex_shader *draw_data; -}; - - - -void * -softpipe_create_blend_state(struct pipe_context *, - const struct pipe_blend_state *); -void softpipe_bind_blend_state(struct pipe_context *, - void *); -void softpipe_delete_blend_state(struct pipe_context *, - void *); - -void * -softpipe_create_sampler_state(struct pipe_context *, - const struct pipe_sampler_state *); -void softpipe_bind_sampler_state(struct pipe_context *, unsigned, void *); -void softpipe_delete_sampler_state(struct pipe_context *, void *); - -void * -softpipe_create_depth_stencil_state(struct pipe_context *, - const struct pipe_depth_stencil_alpha_state *); -void softpipe_bind_depth_stencil_state(struct pipe_context *, void *); -void softpipe_delete_depth_stencil_state(struct pipe_context *, void *); - -void * -softpipe_create_rasterizer_state(struct pipe_context *, - const struct pipe_rasterizer_state *); -void softpipe_bind_rasterizer_state(struct pipe_context *, void *); -void softpipe_delete_rasterizer_state(struct pipe_context *, void *); - -void softpipe_set_framebuffer_state( struct pipe_context *, - const struct pipe_framebuffer_state * ); - -void softpipe_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ); - -void softpipe_set_clip_state( struct pipe_context *, - const struct pipe_clip_state * ); - -void softpipe_set_constant_buffer(struct pipe_context *, - uint shader, uint index, - const struct pipe_constant_buffer *buf); - -void *softpipe_create_fs_state(struct pipe_context *, - const struct pipe_shader_state *); -void softpipe_bind_fs_state(struct pipe_context *, void *); -void softpipe_delete_fs_state(struct pipe_context *, void *); -void *softpipe_create_vs_state(struct pipe_context *, - const struct pipe_shader_state *); -void softpipe_bind_vs_state(struct pipe_context *, void *); -void softpipe_delete_vs_state(struct pipe_context *, void *); - -void softpipe_set_polygon_stipple( struct pipe_context *, - const struct pipe_poly_stipple * ); - -void softpipe_set_scissor_state( struct pipe_context *, - const struct pipe_scissor_state * ); - -void softpipe_set_sampler_texture( struct pipe_context *, - unsigned unit, - struct pipe_texture * ); - -void softpipe_set_viewport_state( struct pipe_context *, - const struct pipe_viewport_state * ); - -void softpipe_set_vertex_element(struct pipe_context *, - unsigned index, - const struct pipe_vertex_element *); - -void softpipe_set_vertex_buffer(struct pipe_context *, - unsigned index, - const struct pipe_vertex_buffer *); - - -void softpipe_update_derived( struct softpipe_context *softpipe ); - - -boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, - unsigned start, unsigned count); - -boolean softpipe_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count); - - -void -softpipe_map_surfaces(struct softpipe_context *sp); - -void -softpipe_unmap_surfaces(struct softpipe_context *sp); - -void -softpipe_map_texture_surfaces(struct softpipe_context *sp); - -void -softpipe_unmap_texture_surfaces(struct softpipe_context *sp); - - -struct vertex_info * -softpipe_get_vertex_info(struct softpipe_context *softpipe); - -struct vertex_info * -softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe); - - -#endif +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef SP_STATE_H +#define SP_STATE_H + +#include "pipe/p_state.h" +#include "tgsi/util/tgsi_scan.h" + + +#define SP_NEW_VIEWPORT 0x1 +#define SP_NEW_RASTERIZER 0x2 +#define SP_NEW_FS 0x4 +#define SP_NEW_BLEND 0x8 +#define SP_NEW_CLIP 0x10 +#define SP_NEW_SCISSOR 0x20 +#define SP_NEW_STIPPLE 0x40 +#define SP_NEW_FRAMEBUFFER 0x80 +#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100 +#define SP_NEW_CONSTANTS 0x200 +#define SP_NEW_SAMPLER 0x400 +#define SP_NEW_TEXTURE 0x800 +#define SP_NEW_VERTEX 0x1000 +#define SP_NEW_VS 0x2000 +#define SP_NEW_QUERY 0x4000 + + +struct tgsi_sampler; +struct tgsi_exec_machine; + + +/** Subclass of pipe_shader_state (though it doesn't really need to be). + * + * This is starting to look an awful lot like a quad pipeline stage... + */ +struct sp_fragment_shader { + struct pipe_shader_state shader; + + struct tgsi_shader_info info; + + void (*prepare)( const struct sp_fragment_shader *shader, + struct tgsi_exec_machine *machine, + struct tgsi_sampler *samplers); + + /* Run the shader - this interface will get cleaned up in the + * future: + */ + unsigned (*run)( const struct sp_fragment_shader *shader, + struct tgsi_exec_machine *machine, + struct quad_header *quad ); + + + void (*delete)( struct sp_fragment_shader * ); +}; + +struct vertex_info; + +/** Subclass of pipe_shader_state */ +struct sp_vertex_shader { + struct pipe_shader_state shader; + struct draw_vertex_shader *draw_data; +}; + + + +void * +softpipe_create_blend_state(struct pipe_context *, + const struct pipe_blend_state *); +void softpipe_bind_blend_state(struct pipe_context *, + void *); +void softpipe_delete_blend_state(struct pipe_context *, + void *); + +void * +softpipe_create_sampler_state(struct pipe_context *, + const struct pipe_sampler_state *); +void softpipe_bind_sampler_states(struct pipe_context *, unsigned, void **); +void softpipe_delete_sampler_state(struct pipe_context *, void *); + +void * +softpipe_create_depth_stencil_state(struct pipe_context *, + const struct pipe_depth_stencil_alpha_state *); +void softpipe_bind_depth_stencil_state(struct pipe_context *, void *); +void softpipe_delete_depth_stencil_state(struct pipe_context *, void *); + +void * +softpipe_create_rasterizer_state(struct pipe_context *, + const struct pipe_rasterizer_state *); +void softpipe_bind_rasterizer_state(struct pipe_context *, void *); +void softpipe_delete_rasterizer_state(struct pipe_context *, void *); + +void softpipe_set_framebuffer_state( struct pipe_context *, + const struct pipe_framebuffer_state * ); + +void softpipe_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ); + +void softpipe_set_clip_state( struct pipe_context *, + const struct pipe_clip_state * ); + +void softpipe_set_constant_buffer(struct pipe_context *, + uint shader, uint index, + const struct pipe_constant_buffer *buf); + +void *softpipe_create_fs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_fs_state(struct pipe_context *, void *); +void softpipe_delete_fs_state(struct pipe_context *, void *); +void *softpipe_create_vs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_vs_state(struct pipe_context *, void *); +void softpipe_delete_vs_state(struct pipe_context *, void *); + +void softpipe_set_polygon_stipple( struct pipe_context *, + const struct pipe_poly_stipple * ); + +void softpipe_set_scissor_state( struct pipe_context *, + const struct pipe_scissor_state * ); + +void softpipe_set_sampler_textures( struct pipe_context *, + unsigned num, + struct pipe_texture ** ); + +void softpipe_set_viewport_state( struct pipe_context *, + const struct pipe_viewport_state * ); + +void softpipe_set_vertex_element(struct pipe_context *, + unsigned index, + const struct pipe_vertex_element *); + +void softpipe_set_vertex_buffer(struct pipe_context *, + unsigned index, + const struct pipe_vertex_buffer *); + + +void softpipe_update_derived( struct softpipe_context *softpipe ); + + +boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count); + +boolean softpipe_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); + + +void +softpipe_map_surfaces(struct softpipe_context *sp); + +void +softpipe_unmap_surfaces(struct softpipe_context *sp); + +void +softpipe_map_texture_surfaces(struct softpipe_context *sp); + +void +softpipe_unmap_texture_surfaces(struct softpipe_context *sp); + + +struct vertex_info * +softpipe_get_vertex_info(struct softpipe_context *softpipe); + +struct vertex_info * +softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe); + + +#endif diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c index 1d6dd17d1d..7cf85b9207 100644 --- a/src/gallium/drivers/softpipe/sp_state_sampler.c +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -50,45 +50,67 @@ softpipe_create_sampler_state(struct pipe_context *pipe, return mem_dup(sampler, sizeof(*sampler)); } + void -softpipe_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) +softpipe_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **sampler) { struct softpipe_context *softpipe = softpipe_context(pipe); + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == softpipe->num_samplers && + !memcmp(softpipe->sampler, sampler, num * sizeof(void *))) + return; + draw_flush(softpipe->draw); - assert(unit < PIPE_MAX_SAMPLERS); - softpipe->sampler[unit] = (struct pipe_sampler_state *)sampler; + memcpy(softpipe->sampler, sampler, num * sizeof(void *)); + memset(&softpipe->sampler[num], 0, (PIPE_MAX_SAMPLERS - num) * + sizeof(void *)); + + softpipe->num_samplers = num; softpipe->dirty |= SP_NEW_SAMPLER; } void -softpipe_delete_sampler_state(struct pipe_context *pipe, - void *sampler) +softpipe_set_sampler_textures(struct pipe_context *pipe, + unsigned num, struct pipe_texture **texture) { - FREE( sampler ); -} + struct softpipe_context *softpipe = softpipe_context(pipe); + uint i; + assert(num <= PIPE_MAX_SAMPLERS); -void -softpipe_set_sampler_texture(struct pipe_context *pipe, - unsigned unit, - struct pipe_texture *texture) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); + /* Check for no-op */ + if (num == softpipe->num_textures && + !memcmp(softpipe->texture, texture, num * sizeof(struct pipe_texture *))) + return; draw_flush(softpipe->draw); - assert(unit < PIPE_MAX_SAMPLERS); - pipe_texture_reference(&softpipe->texture[unit], texture); + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + struct pipe_texture *tex = i < num ? texture[i] : NULL; + + pipe_texture_reference(&softpipe->texture[i], tex); + sp_tile_cache_set_texture(pipe, softpipe->tex_cache[i], tex); + } - sp_tile_cache_set_texture(pipe, softpipe->tex_cache[unit], texture); + softpipe->num_textures = num; softpipe->dirty |= SP_NEW_TEXTURE; } +void +softpipe_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + FREE( sampler ); +} + + diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index c605ed925b..bbf2d80308 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -1,202 +1,202 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Michel Dänzer - */ - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "pipe/p_winsys.h" - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_texture.h" -#include "sp_tile_cache.h" - - -/* Simple, maximally packed layout. - */ - -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - - -static void -softpipe_texture_layout(struct softpipe_texture * spt) -{ - struct pipe_texture *pt = &spt->base; - unsigned level; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - - spt->buffer_size = 0; - - for (level = 0; level <= pt->last_level; level++) { - pt->width[level] = width; - pt->height[level] = height; - pt->depth[level] = depth; - - spt->level_offset[level] = spt->buffer_size; - - spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * - ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * - width * pt->cpp; - - width = minify(width); - height = minify(height); - depth = minify(depth); - } -} - - -static struct pipe_texture * -softpipe_texture_create_screen(struct pipe_screen *screen, - const struct pipe_texture *templat) -{ - struct pipe_winsys *ws = screen->winsys; - struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); - if (!spt) - return NULL; - - spt->base = *templat; - spt->base.refcount = 1; - spt->base.screen = screen; - - softpipe_texture_layout(spt); - - spt->buffer = ws->buffer_create(ws, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); - if (!spt->buffer) { - FREE(spt); - return NULL; - } - - assert(spt->base.refcount == 1); - - return &spt->base; -} - - -static void -softpipe_texture_release_screen(struct pipe_screen *screen, - struct pipe_texture **pt) -{ - if (!*pt) - return; - - /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); - */ - if (--(*pt)->refcount <= 0) { - struct softpipe_texture *spt = softpipe_texture(*pt); - - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); - */ - - pipe_buffer_reference(screen->winsys, &spt->buffer, NULL); - - FREE(spt); - } - *pt = NULL; -} - - -static struct pipe_surface * -softpipe_get_tex_surface_screen(struct pipe_screen *screen, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - struct pipe_winsys *ws = screen->winsys; - struct softpipe_texture *spt = softpipe_texture(pt); - struct pipe_surface *ps; - - assert(level <= pt->last_level); - - ps = ws->surface_alloc(ws); - if (ps) { - assert(ps->refcount); - assert(ps->winsys); - pipe_buffer_reference(ws, &ps->buffer, spt->buffer); - ps->format = pt->format; - ps->cpp = pt->cpp; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->pitch = ps->width; - ps->offset = spt->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { - ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - (pt->compressed ? ps->height/4 : ps->height) * - ps->width * ps->cpp; - } - else { - assert(face == 0); - assert(zslice == 0); - } - } - return ps; -} - - -static void -softpipe_texture_update(struct pipe_context *pipe, - struct pipe_texture *texture) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - uint unit; - for (unit = 0; unit < PIPE_MAX_SAMPLERS; unit++) { - if (softpipe->texture[unit] == texture) { - sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); - } - } -} - - -void -softpipe_init_texture_funcs( struct softpipe_context *softpipe ) -{ - softpipe->pipe.texture_update = softpipe_texture_update; -} - - -void -softpipe_init_screen_texture_funcs(struct pipe_screen *screen) -{ - screen->texture_create = softpipe_texture_create_screen; - screen->texture_release = softpipe_texture_release_screen; - screen->get_tex_surface = softpipe_get_tex_surface_screen; -} +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_texture.h" +#include "sp_tile_cache.h" + + +/* Simple, maximally packed layout. + */ + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + + +static void +softpipe_texture_layout(struct softpipe_texture * spt) +{ + struct pipe_texture *pt = &spt->base; + unsigned level; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + + spt->buffer_size = 0; + + for (level = 0; level <= pt->last_level; level++) { + pt->width[level] = width; + pt->height[level] = height; + pt->depth[level] = depth; + + spt->level_offset[level] = spt->buffer_size; + + spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * + ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * + width * pt->cpp; + + width = minify(width); + height = minify(height); + depth = minify(depth); + } +} + + +static struct pipe_texture * +softpipe_texture_create_screen(struct pipe_screen *screen, + const struct pipe_texture *templat) +{ + struct pipe_winsys *ws = screen->winsys; + struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); + if (!spt) + return NULL; + + spt->base = *templat; + spt->base.refcount = 1; + spt->base.screen = screen; + + softpipe_texture_layout(spt); + + spt->buffer = ws->buffer_create(ws, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); + if (!spt->buffer) { + FREE(spt); + return NULL; + } + + assert(spt->base.refcount == 1); + + return &spt->base; +} + + +static void +softpipe_texture_release_screen(struct pipe_screen *screen, + struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct softpipe_texture *spt = softpipe_texture(*pt); + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); + */ + + pipe_buffer_reference(screen->winsys, &spt->buffer, NULL); + + FREE(spt); + } + *pt = NULL; +} + + +static struct pipe_surface * +softpipe_get_tex_surface_screen(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct pipe_winsys *ws = screen->winsys; + struct softpipe_texture *spt = softpipe_texture(pt); + struct pipe_surface *ps; + + assert(level <= pt->last_level); + + ps = ws->surface_alloc(ws); + if (ps) { + assert(ps->refcount); + assert(ps->winsys); + pipe_buffer_reference(ws, &ps->buffer, spt->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = ps->width; + ps->offset = spt->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { + ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * + (pt->compressed ? ps->height/4 : ps->height) * + ps->width * ps->cpp; + } + else { + assert(face == 0); + assert(zslice == 0); + } + } + return ps; +} + + +static void +softpipe_texture_update(struct pipe_context *pipe, + struct pipe_texture *texture) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + uint unit; + for (unit = 0; unit < softpipe->num_textures; unit++) { + if (softpipe->texture[unit] == texture) { + sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); + } + } +} + + +void +softpipe_init_texture_funcs( struct softpipe_context *softpipe ) +{ + softpipe->pipe.texture_update = softpipe_texture_update; +} + + +void +softpipe_init_screen_texture_funcs(struct pipe_screen *screen) +{ + screen->texture_create = softpipe_texture_create_screen; + screen->texture_release = softpipe_texture_release_screen; + screen->get_tex_surface = softpipe_get_tex_surface_screen; +} diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 1501b52f3e..b64948f0f3 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -100,7 +100,7 @@ struct pipe_context { void * (*create_sampler_state)(struct pipe_context *, const struct pipe_sampler_state *); - void (*bind_sampler_state)(struct pipe_context *, unsigned unit, void *); + void (*bind_sampler_states)(struct pipe_context *, unsigned num, void **); void (*delete_sampler_state)(struct pipe_context *, void *); void * (*create_rasterizer_state)(struct pipe_context *, @@ -148,9 +148,9 @@ struct pipe_context { /* Currently a sampler is constrained to sample from a single texture: */ - void (*set_sampler_texture)( struct pipe_context *, - unsigned sampler, - struct pipe_texture * ); + void (*set_sampler_textures)( struct pipe_context *, + unsigned num, + struct pipe_texture ** ); void (*set_viewport_state)( struct pipe_context *, const struct pipe_viewport_state * ); diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 92263cb688..1000f98ffc 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -120,10 +120,11 @@ update_samplers(struct st_context *st) const struct st_fragment_program *fs = st->fp; GLuint su; + st->state.num_samplers = 0; + /* loop over sampler units (aka tex image units) */ for (su = 0; su < st->ctx->Const.MaxTextureImageUnits; su++) { struct pipe_sampler_state sampler; - const struct cso_sampler *cso; memset(&sampler, 0, sizeof(sampler)); @@ -168,17 +169,16 @@ update_samplers(struct st_context *st) = st_compare_func_to_pipe(texobj->CompareFunc); } + st->state.num_samplers = su + 1; + /* XXX more sampler state here */ } - cso = st_cached_sampler_state(st, &sampler); - - if (cso != st->state.sampler[su]) { - /* state has changed */ - st->state.sampler[su] = cso; - st->pipe->bind_sampler_state(st->pipe, su, cso->data); - } + st->state.sampler[su] = st_cached_sampler_state(st, &sampler)->data; } + + st->pipe->bind_sampler_states(st->pipe, st->state.num_samplers, + st->state.sampler); } diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 697d2cdfb4..e53a897637 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -37,6 +37,7 @@ #include "st_texture.h" #include "st_cb_texture.h" #include "pipe/p_context.h" +#include "pipe/p_inlines.h" /** @@ -51,6 +52,8 @@ update_textures(struct st_context *st) struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current; GLuint unit; + st->state.num_textures = 0; + for (unit = 0; unit < st->ctx->Const.MaxTextureCoordUnits; unit++) { const GLuint su = fprog->Base.SamplerUnits[unit]; struct gl_texture_object *texObj = st->ctx->Texture.Unit[su]._Current; @@ -62,6 +65,8 @@ update_textures(struct st_context *st) retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); /* XXX retval indicates whether there's a texture border */ + + st->state.num_textures = unit + 1; } /* XXX: need to ensure that textures are unbound/removed from @@ -70,18 +75,16 @@ update_textures(struct st_context *st) */ pt = st_get_stobj_texture(stObj); - - if (st->state.sampler_texture[unit] != pt) { - st->state.sampler_texture[unit] = pt; - st->pipe->set_sampler_texture(st->pipe, unit, pt); - } + pipe_texture_reference(&st->state.sampler_texture[unit], pt); if (stObj && stObj->dirtyData) { st->pipe->texture_update(st->pipe, pt); stObj->dirtyData = GL_FALSE; } - } + + st->pipe->set_sampler_textures(st->pipe, st->state.num_textures, + st->state.sampler_texture); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 65c9fda9cb..dee4c4132a 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -641,7 +641,6 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, const GLfloat *color, GLboolean invertTex) { - const GLuint unit = 0; struct pipe_context *pipe = ctx->st->pipe; GLfloat x0, y0, x1, y1; GLuint maxSize; @@ -684,7 +683,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; sampler.normalized_coords = 1; cso = st_cached_sampler_state(ctx->st, &sampler); - pipe->bind_sampler_state(pipe, unit, cso->data); + pipe->bind_sampler_states(pipe, 1, (void**)&cso->data); } /* viewport state: viewport matching window dims */ @@ -705,7 +704,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* texture state: */ { - pipe->set_sampler_texture(pipe, unit, pt); + pipe->set_sampler_textures(pipe, 1, &pt); } /* Compute window coords (y=0=bottom) with pixel zoom. @@ -727,8 +726,10 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); pipe->bind_fs_state(pipe, ctx->st->state.fs->data); pipe->bind_vs_state(pipe, ctx->st->state.vs->cso->data); - pipe->set_sampler_texture(pipe, unit, ctx->st->state.sampler_texture[unit]); - pipe->bind_sampler_state(pipe, unit, ctx->st->state.sampler[unit]->data); + pipe->set_sampler_textures(pipe, ctx->st->state.num_textures, + ctx->st->state.sampler_texture); + pipe->bind_sampler_states(pipe, ctx->st->state.num_samplers, + ctx->st->state.sampler); pipe->set_viewport_state(pipe, &ctx->st->state.viewport); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 1fbf9721e7..897a5109b7 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -76,7 +76,7 @@ struct st_context struct { const struct cso_alpha_test *alpha_test; const struct cso_blend *blend; - const struct cso_sampler *sampler[PIPE_MAX_SAMPLERS]; + void *sampler[PIPE_MAX_SAMPLERS]; const struct cso_depth_stencil_alpha *depth_stencil; const struct cso_rasterizer *rasterizer; const struct cso_fragment_shader *fs; @@ -90,6 +90,9 @@ struct st_context struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct pipe_viewport_state viewport; + + GLuint num_samplers; + GLuint num_textures; } state; struct { diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 841d77abbc..3723e26d45 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -284,7 +284,7 @@ st_render_mipmap(struct st_context *st, */ sampler.min_lod = sampler.max_lod = srcLevel; sampler_cso = pipe->create_sampler_state(pipe, &sampler); - pipe->bind_sampler_state(pipe, 0, sampler_cso); + pipe->bind_sampler_states(pipe, 1, &sampler_cso); simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]); @@ -293,7 +293,7 @@ st_render_mipmap(struct st_context *st, * the right mipmap level. */ /*pt->first_level = srcLevel;*/ - pipe->set_sampler_texture(pipe, 0, pt); + pipe->set_sampler_textures(pipe, 1, &pt); draw_quad(st->ctx); @@ -310,9 +310,10 @@ st_render_mipmap(struct st_context *st, pipe->bind_fs_state(pipe, st->state.fs->data); if (st->state.vs) pipe->bind_vs_state(pipe, st->state.vs->cso->data); - if (st->state.sampler[0]) - pipe->bind_sampler_state(pipe, 0, st->state.sampler[0]->data); - pipe->set_sampler_texture(pipe, 0, st->state.sampler_texture[0]); + pipe->bind_sampler_states(pipe, st->state.num_samplers, + st->state.sampler); + pipe->set_sampler_textures(pipe, st->state.num_textures, + st->state.sampler_texture); pipe->set_viewport_state(pipe, &st->state.viewport); return TRUE; -- cgit v1.2.3 From 8143adafddabb6ac3a21c18927ae41425f26bfff Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 6 Mar 2008 19:57:41 +0100 Subject: gallium: Surround externs with extern "C". --- src/gallium/auxiliary/util/p_tile.h | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.h b/src/gallium/auxiliary/util/p_tile.h index 318b6d11a6..fdc80a13b3 100644 --- a/src/gallium/auxiliary/util/p_tile.h +++ b/src/gallium/auxiliary/util/p_tile.h @@ -52,44 +52,50 @@ pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps) return FALSE; } +#ifdef __cplusplus +extern "C" { +#endif -extern void +void pipe_get_tile_raw(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, void *p, int dst_stride); -extern void +void pipe_put_tile_raw(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const void *p, int src_stride); -extern void +void pipe_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p); -extern void +void pipe_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p); -extern void +void pipe_get_tile_z(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, uint *z); -extern void +void pipe_put_tile_z(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, const uint *z); +#ifdef __cplusplus +} +#endif #endif -- cgit v1.2.3 From fc96aec9b7aceb4a0e7471e797abe8a00fc40cf2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 8 Mar 2008 16:29:12 +0000 Subject: gallium: Document debug_printf usage. --- src/gallium/auxiliary/util/Makefile | 3 ++- src/gallium/include/pipe/p_debug.h | 14 ++++++++++++++ 2 files changed, 16 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 906a46d6b4..2a3a9380b3 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -7,7 +7,8 @@ C_SOURCES = \ p_debug.c \ p_tile.c \ p_util.c \ - u_mm.c + u_mm.c \ + u_snprintf.c include ../../Makefile.template diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 2a11627b36..a14a1fc5f6 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -58,8 +58,22 @@ extern "C" { #endif +/** + * Print debug messages. + * + * A debug message will be printed regardless of the DEBUG/NDEBUG macros. + * + * The actual channel used to output debug message is platform specific. To + * avoid misformating or truncation, follow these rules of thumb: + * - output whole lines + * - avoid outputing large strings (512 bytes is the current maximum length + * that is guaranteed to be printed in all platforms) + */ void debug_printf(const char *format, ...); +/** + * @sa debug_printf + */ void debug_vprintf(const char *format, va_list ap); void debug_assert_fail(const char *expr, const char *file, unsigned line); -- cgit v1.2.3 From 99691f38c278f1d4aeb0617b149109644e3571fb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 9 Mar 2008 15:07:09 +0100 Subject: gallium: add some commonly implemented bits of hw state --- src/gallium/include/pipe/p_state.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 49a003b923..e338a27383 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -108,6 +108,7 @@ struct pipe_rasterizer_state unsigned line_stipple_enable:1; unsigned line_stipple_factor:8; /**< [1..256] actually */ unsigned line_stipple_pattern:16; + unsigned line_last_pixel:1; unsigned bypass_clipping:1; unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ @@ -154,7 +155,7 @@ struct pipe_clip_state struct pipe_constant_buffer { struct pipe_buffer *buffer; - unsigned size; /** in bytes */ + unsigned size; /** in bytes (XXX: redundant!) */ }; @@ -248,6 +249,7 @@ struct pipe_sampler_state unsigned compare_mode:1; /**< PIPE_TEX_COMPARE_x */ unsigned compare_func:3; /**< PIPE_FUNC_x */ unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */ + unsigned prefilter:4; /**< Wierd sampling state exposed by some api's */ float shadow_ambient; /**< shadow test fail color/intensity */ float lod_bias; /**< LOD/lambda bias */ float min_lod, max_lod; /**< LOD clamp range, after bias */ -- cgit v1.2.3 From 5d802d8c8460cecf306b130eb29ef05069173e30 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 9 Mar 2008 15:09:55 +0100 Subject: cso: add a higher-level interface which does all pipe interactions to set a given state --- src/gallium/auxiliary/cso_cache/Makefile | 1 + src/gallium/auxiliary/cso_cache/SConscript | 1 + src/gallium/auxiliary/cso_cache/cso_context.c | 340 ++++++++++++++++++++++++++ src/gallium/auxiliary/cso_cache/cso_context.h | 85 +++++++ 4 files changed, 427 insertions(+) create mode 100644 src/gallium/auxiliary/cso_cache/cso_context.c create mode 100644 src/gallium/auxiliary/cso_cache/cso_context.h (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/Makefile b/src/gallium/auxiliary/cso_cache/Makefile index 3e49266163..6bd6602088 100644 --- a/src/gallium/auxiliary/cso_cache/Makefile +++ b/src/gallium/auxiliary/cso_cache/Makefile @@ -4,6 +4,7 @@ include $(TOP)/configs/current LIBNAME = cso_cache C_SOURCES = \ + cso_context.c \ cso_cache.c \ cso_hash.c diff --git a/src/gallium/auxiliary/cso_cache/SConscript b/src/gallium/auxiliary/cso_cache/SConscript index 9751881613..651e68a191 100644 --- a/src/gallium/auxiliary/cso_cache/SConscript +++ b/src/gallium/auxiliary/cso_cache/SConscript @@ -3,6 +3,7 @@ Import('*') cso_cache = env.ConvenienceLibrary( target = 'cso_cache', source = [ + 'cso_context.c', 'cso_cache.c', 'cso_hash.c', ]) diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c new file mode 100644 index 0000000000..77237464c5 --- /dev/null +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -0,0 +1,340 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* Wrap the cso cache & hash mechanisms in a simplified + * pipe-driver-specific interface. + * + * Authors: + * Zack Rusin + * Keith Whitwell + */ + +#include "pipe/p_state.h" +#include "pipe/p_util.h" + +#include "cso_cache/cso_context.h" +#include "cso_cache/cso_cache.h" +#include "cso_cache/cso_hash.h" + +struct cso_context { + struct pipe_context *pipe; + struct cso_cache *cache; + + struct { + void *samplers[PIPE_MAX_SAMPLERS]; + unsigned nr_samplers; + } hw; + + void *samplers[PIPE_MAX_SAMPLERS]; + unsigned nr_samplers; + + void *blend; + void *depth_stencil; + void *rasterizer; + void *fragment_shader; + void *vertex_shader; +}; + + +struct cso_context *cso_create_context( struct pipe_context *pipe ) +{ + struct cso_context *ctx = CALLOC_STRUCT(cso_context); + if (ctx == NULL) + goto out; + + ctx->cache = cso_cache_create(); + if (ctx->cache == NULL) + goto out; + + ctx->pipe = pipe; + + return ctx; + +out: + cso_destroy_context( ctx ); + return NULL; +} + + +void cso_destroy_context( struct cso_context *ctx ) +{ + if (ctx == NULL) + return; + +/* + if (ctx->pipe) + ctx->pipe->flush( ctx->pipe, PIPE_FLUSH_UNBIND_ALL ); +*/ + + if (ctx->cache) + cso_cache_delete( ctx->cache ); + + FREE( ctx ); +} + + +/* Those function will either find the state of the given template + * in the cache or they will create a new state from the given + * template, insert it in the cache and return it. + */ + +/* + * If the driver returns 0 from the create method then they will assign + * the data member of the cso to be the template itself. + */ + +void cso_set_blend(struct cso_context *ctx, + const struct pipe_blend_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_blend_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_BLEND, + (void*)templ); + void *handle; + + if (cso_hash_iter_is_null(iter)) { + struct cso_blend *cso = malloc(sizeof(struct cso_blend)); + + cso->state = *templ; + cso->data = ctx->pipe->create_blend_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_blend_state; + cso->context = ctx->pipe; + + iter = cso_insert_state(ctx->cache, hash_key, CSO_BLEND, cso); + handle = cso->data; + } + else { + handle = ((struct cso_blend *)cso_hash_iter_data(iter))->data; + } + + if (ctx->blend != handle) { + ctx->blend = handle; + ctx->pipe->bind_blend_state(ctx->pipe, handle); + } +} + +void cso_single_sampler(struct cso_context *ctx, + unsigned idx, + const struct pipe_sampler_state *templ) +{ + void *handle = NULL; + + if (templ != NULL) { + unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_sampler_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_SAMPLER, + (void*)templ); + + if (cso_hash_iter_is_null(iter)) { + struct cso_sampler *cso = malloc(sizeof(struct cso_sampler)); + + cso->state = *templ; + cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state; + cso->context = ctx->pipe; + + iter = cso_insert_state(ctx->cache, hash_key, CSO_SAMPLER, cso); + handle = cso->data; + } + else { + handle = ((struct cso_sampler *)cso_hash_iter_data(iter))->data; + } + } + + ctx->samplers[idx] = handle; +} + +void cso_single_sampler_done( struct cso_context *ctx ) +{ + unsigned i; + + for (i = 0; i < 8; i++) + if (ctx->samplers[i] == NULL) + break; + + ctx->nr_samplers = i; + + if (ctx->hw.nr_samplers != ctx->nr_samplers || + memcmp(ctx->hw.samplers, + ctx->samplers, + ctx->nr_samplers * sizeof(void *)) != 0) + { + memcpy(ctx->hw.samplers, ctx->samplers, ctx->nr_samplers * sizeof(void *)); + ctx->hw.nr_samplers = ctx->nr_samplers; + + ctx->pipe->bind_sampler_states(ctx->pipe, ctx->nr_samplers, ctx->samplers); + } +} + +void cso_set_samplers( struct cso_context *ctx, + unsigned nr, + const struct pipe_sampler_state **templates ) +{ + unsigned i; + + /* TODO: fastpath + */ + + for (i = 0; i < nr; i++) + cso_single_sampler( ctx, i, templates[i] ); + + for ( ; i < ctx->nr_samplers; i++) + cso_single_sampler( ctx, i, NULL ); + + cso_single_sampler_done( ctx ); +} + +void cso_set_depth_stencil_alpha(struct cso_context *ctx, + const struct pipe_depth_stencil_alpha_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_depth_stencil_alpha_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, + CSO_DEPTH_STENCIL_ALPHA, + (void*)templ); + void *handle; + + if (cso_hash_iter_is_null(iter)) { + struct cso_depth_stencil_alpha *cso = malloc(sizeof(struct cso_depth_stencil_alpha)); + + cso->state = *templ; + cso->data = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_depth_stencil_alpha_state; + cso->context = ctx->pipe; + + cso_insert_state(ctx->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso); + handle = cso->data; + } + else { + handle = ((struct cso_depth_stencil_alpha *)cso_hash_iter_data(iter))->data; + } + + if (ctx->depth_stencil != handle) { + ctx->depth_stencil = handle; + ctx->pipe->bind_depth_stencil_alpha_state(ctx->pipe, handle); + } +} + + + +void cso_set_rasterizer(struct cso_context *ctx, + const struct pipe_rasterizer_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_rasterizer_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_RASTERIZER, + (void*)templ); + void *handle = NULL; + + if (cso_hash_iter_is_null(iter)) { + struct cso_rasterizer *cso = malloc(sizeof(struct cso_rasterizer)); + + cso->state = *templ; + cso->data = ctx->pipe->create_rasterizer_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_rasterizer_state; + cso->context = ctx->pipe; + + cso_insert_state(ctx->cache, hash_key, CSO_RASTERIZER, cso); + handle = cso->data; + } + else { + handle = ((struct cso_rasterizer *)cso_hash_iter_data(iter))->data; + } + + if (ctx->rasterizer != handle) { + ctx->rasterizer = handle; + ctx->pipe->bind_rasterizer_state(ctx->pipe, handle); + } +} + + + + + +void cso_set_fragment_shader(struct cso_context *ctx, + const struct pipe_shader_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_shader_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_FRAGMENT_SHADER, + (void*)templ); + void *handle = NULL; + + if (cso_hash_iter_is_null(iter)) { + struct cso_fragment_shader *cso = malloc(sizeof(struct cso_fragment_shader)); + + cso->state = *templ; + cso->data = ctx->pipe->create_fs_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_fs_state; + cso->context = ctx->pipe; + + iter = cso_insert_state(ctx->cache, hash_key, CSO_FRAGMENT_SHADER, cso); + handle = cso->data; + } + else { + handle = ((struct cso_fragment_shader *)cso_hash_iter_data(iter))->data; + } + + if (ctx->fragment_shader != handle) { + ctx->fragment_shader = handle; + ctx->pipe->bind_fs_state(ctx->pipe, handle); + } +} + +void cso_set_vertex_shader(struct cso_context *ctx, + const struct pipe_shader_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_shader_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_VERTEX_SHADER, + (void*)templ); + void *handle = NULL; + + if (cso_hash_iter_is_null(iter)) { + struct cso_vertex_shader *cso = malloc(sizeof(struct cso_vertex_shader)); + + cso->state = *templ; + cso->data = ctx->pipe->create_vs_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_vs_state; + cso->context = ctx->pipe; + + iter = cso_insert_state(ctx->cache, hash_key, CSO_VERTEX_SHADER, cso); + handle = cso->data; + } + else { + handle = ((struct cso_vertex_shader *)cso_hash_iter_data(iter))->data; + } + + if (ctx->vertex_shader != handle) { + ctx->vertex_shader = handle; + ctx->pipe->bind_fs_state(ctx->pipe, handle); + } +} diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h new file mode 100644 index 0000000000..1f2a630804 --- /dev/null +++ b/src/gallium/auxiliary/cso_cache/cso_context.h @@ -0,0 +1,85 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef CSO_CONTEXT_H +#define CSO_CONTEXT_H + +#include "pipe/p_context.h" +#include "pipe/p_state.h" + + +#ifdef __cplusplus +extern "C" { +#endif + +struct cso_context; + +struct cso_context *cso_create_context( struct pipe_context *pipe ); + +void cso_set_blend( struct cso_context *cso, + const struct pipe_blend_state *blend ); + +void cso_set_depth_stencil_alpha( struct cso_context *cso, + const struct pipe_depth_stencil_alpha_state *dsa ); + +void cso_set_rasterizer( struct cso_context *cso, + const struct pipe_rasterizer_state *rasterizer ); + +void cso_set_samplers( struct cso_context *cso, + unsigned count, + const struct pipe_sampler_state **states ); + +/* Alternate interface to support state trackers that like to modify + * samplers one at a time: + */ +void cso_single_sampler( struct cso_context *cso, + unsigned nr, + const struct pipe_sampler_state *states ); + +void cso_single_sampler_done( struct cso_context *cso ); + + +/* These aren't really sensible -- most of the time the api provides + * object semantics for shaders anyway, and the cases where it doesn't + * (eg mesa's internall-generated texenv programs), it will be up to + * the state tracker to implement their own specialized caching. + */ +void cso_set_fragment_shader( struct cso_context *cso, + const struct pipe_shader_state *shader ); + +void cso_set_vertex_shader( struct cso_context *cso, + const struct pipe_shader_state *shader ); + +void cso_destroy_context( struct cso_context *cso ); + + +#ifdef __cplusplus +} +#endif + +#endif -- cgit v1.2.3 From ac87bc18359825890a53d4dbfda5c6eecd916afd Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 9 Mar 2008 17:05:46 +0100 Subject: cso: Use MALLOC --- src/gallium/auxiliary/cso_cache/cso_context.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 77237464c5..051e7d96d3 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -117,7 +117,7 @@ void cso_set_blend(struct cso_context *ctx, void *handle; if (cso_hash_iter_is_null(iter)) { - struct cso_blend *cso = malloc(sizeof(struct cso_blend)); + struct cso_blend *cso = MALLOC(sizeof(struct cso_blend)); cso->state = *templ; cso->data = ctx->pipe->create_blend_state(ctx->pipe, &cso->state); @@ -150,7 +150,7 @@ void cso_single_sampler(struct cso_context *ctx, (void*)templ); if (cso_hash_iter_is_null(iter)) { - struct cso_sampler *cso = malloc(sizeof(struct cso_sampler)); + struct cso_sampler *cso = MALLOC(sizeof(struct cso_sampler)); cso->state = *templ; cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state); @@ -220,7 +220,7 @@ void cso_set_depth_stencil_alpha(struct cso_context *ctx, void *handle; if (cso_hash_iter_is_null(iter)) { - struct cso_depth_stencil_alpha *cso = malloc(sizeof(struct cso_depth_stencil_alpha)); + struct cso_depth_stencil_alpha *cso = MALLOC(sizeof(struct cso_depth_stencil_alpha)); cso->state = *templ; cso->data = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &cso->state); @@ -253,7 +253,7 @@ void cso_set_rasterizer(struct cso_context *ctx, void *handle = NULL; if (cso_hash_iter_is_null(iter)) { - struct cso_rasterizer *cso = malloc(sizeof(struct cso_rasterizer)); + struct cso_rasterizer *cso = MALLOC(sizeof(struct cso_rasterizer)); cso->state = *templ; cso->data = ctx->pipe->create_rasterizer_state(ctx->pipe, &cso->state); @@ -288,7 +288,7 @@ void cso_set_fragment_shader(struct cso_context *ctx, void *handle = NULL; if (cso_hash_iter_is_null(iter)) { - struct cso_fragment_shader *cso = malloc(sizeof(struct cso_fragment_shader)); + struct cso_fragment_shader *cso = MALLOC(sizeof(struct cso_fragment_shader)); cso->state = *templ; cso->data = ctx->pipe->create_fs_state(ctx->pipe, &cso->state); @@ -319,7 +319,7 @@ void cso_set_vertex_shader(struct cso_context *ctx, void *handle = NULL; if (cso_hash_iter_is_null(iter)) { - struct cso_vertex_shader *cso = malloc(sizeof(struct cso_vertex_shader)); + struct cso_vertex_shader *cso = MALLOC(sizeof(struct cso_vertex_shader)); cso->state = *templ; cso->data = ctx->pipe->create_vs_state(ctx->pipe, &cso->state); -- cgit v1.2.3 From d8d6569e288fe3324473fb19ade798502dfbba8e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 9 Mar 2008 17:06:20 +0100 Subject: cso: fix line endings --- src/gallium/auxiliary/cso_cache/cso_context.c | 680 +++++++++++++------------- 1 file changed, 340 insertions(+), 340 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 051e7d96d3..596e5a9ad6 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -1,340 +1,340 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* Wrap the cso cache & hash mechanisms in a simplified - * pipe-driver-specific interface. - * - * Authors: - * Zack Rusin - * Keith Whitwell - */ - -#include "pipe/p_state.h" -#include "pipe/p_util.h" - -#include "cso_cache/cso_context.h" -#include "cso_cache/cso_cache.h" -#include "cso_cache/cso_hash.h" - -struct cso_context { - struct pipe_context *pipe; - struct cso_cache *cache; - - struct { - void *samplers[PIPE_MAX_SAMPLERS]; - unsigned nr_samplers; - } hw; - - void *samplers[PIPE_MAX_SAMPLERS]; - unsigned nr_samplers; - - void *blend; - void *depth_stencil; - void *rasterizer; - void *fragment_shader; - void *vertex_shader; -}; - - -struct cso_context *cso_create_context( struct pipe_context *pipe ) -{ - struct cso_context *ctx = CALLOC_STRUCT(cso_context); - if (ctx == NULL) - goto out; - - ctx->cache = cso_cache_create(); - if (ctx->cache == NULL) - goto out; - - ctx->pipe = pipe; - - return ctx; - -out: - cso_destroy_context( ctx ); - return NULL; -} - - -void cso_destroy_context( struct cso_context *ctx ) -{ - if (ctx == NULL) - return; - -/* - if (ctx->pipe) - ctx->pipe->flush( ctx->pipe, PIPE_FLUSH_UNBIND_ALL ); -*/ - - if (ctx->cache) - cso_cache_delete( ctx->cache ); - - FREE( ctx ); -} - - -/* Those function will either find the state of the given template - * in the cache or they will create a new state from the given - * template, insert it in the cache and return it. - */ - -/* - * If the driver returns 0 from the create method then they will assign - * the data member of the cso to be the template itself. - */ - -void cso_set_blend(struct cso_context *ctx, - const struct pipe_blend_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_blend_state)); - struct cso_hash_iter iter = cso_find_state_template(ctx->cache, - hash_key, CSO_BLEND, - (void*)templ); - void *handle; - - if (cso_hash_iter_is_null(iter)) { - struct cso_blend *cso = MALLOC(sizeof(struct cso_blend)); - - cso->state = *templ; - cso->data = ctx->pipe->create_blend_state(ctx->pipe, &cso->state); - cso->delete_state = (cso_state_callback)ctx->pipe->delete_blend_state; - cso->context = ctx->pipe; - - iter = cso_insert_state(ctx->cache, hash_key, CSO_BLEND, cso); - handle = cso->data; - } - else { - handle = ((struct cso_blend *)cso_hash_iter_data(iter))->data; - } - - if (ctx->blend != handle) { - ctx->blend = handle; - ctx->pipe->bind_blend_state(ctx->pipe, handle); - } -} - -void cso_single_sampler(struct cso_context *ctx, - unsigned idx, - const struct pipe_sampler_state *templ) -{ - void *handle = NULL; - - if (templ != NULL) { - unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_sampler_state)); - struct cso_hash_iter iter = cso_find_state_template(ctx->cache, - hash_key, CSO_SAMPLER, - (void*)templ); - - if (cso_hash_iter_is_null(iter)) { - struct cso_sampler *cso = MALLOC(sizeof(struct cso_sampler)); - - cso->state = *templ; - cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state); - cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state; - cso->context = ctx->pipe; - - iter = cso_insert_state(ctx->cache, hash_key, CSO_SAMPLER, cso); - handle = cso->data; - } - else { - handle = ((struct cso_sampler *)cso_hash_iter_data(iter))->data; - } - } - - ctx->samplers[idx] = handle; -} - -void cso_single_sampler_done( struct cso_context *ctx ) -{ - unsigned i; - - for (i = 0; i < 8; i++) - if (ctx->samplers[i] == NULL) - break; - - ctx->nr_samplers = i; - - if (ctx->hw.nr_samplers != ctx->nr_samplers || - memcmp(ctx->hw.samplers, - ctx->samplers, - ctx->nr_samplers * sizeof(void *)) != 0) - { - memcpy(ctx->hw.samplers, ctx->samplers, ctx->nr_samplers * sizeof(void *)); - ctx->hw.nr_samplers = ctx->nr_samplers; - - ctx->pipe->bind_sampler_states(ctx->pipe, ctx->nr_samplers, ctx->samplers); - } -} - -void cso_set_samplers( struct cso_context *ctx, - unsigned nr, - const struct pipe_sampler_state **templates ) -{ - unsigned i; - - /* TODO: fastpath - */ - - for (i = 0; i < nr; i++) - cso_single_sampler( ctx, i, templates[i] ); - - for ( ; i < ctx->nr_samplers; i++) - cso_single_sampler( ctx, i, NULL ); - - cso_single_sampler_done( ctx ); -} - -void cso_set_depth_stencil_alpha(struct cso_context *ctx, - const struct pipe_depth_stencil_alpha_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_depth_stencil_alpha_state)); - struct cso_hash_iter iter = cso_find_state_template(ctx->cache, - hash_key, - CSO_DEPTH_STENCIL_ALPHA, - (void*)templ); - void *handle; - - if (cso_hash_iter_is_null(iter)) { - struct cso_depth_stencil_alpha *cso = MALLOC(sizeof(struct cso_depth_stencil_alpha)); - - cso->state = *templ; - cso->data = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &cso->state); - cso->delete_state = (cso_state_callback)ctx->pipe->delete_depth_stencil_alpha_state; - cso->context = ctx->pipe; - - cso_insert_state(ctx->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso); - handle = cso->data; - } - else { - handle = ((struct cso_depth_stencil_alpha *)cso_hash_iter_data(iter))->data; - } - - if (ctx->depth_stencil != handle) { - ctx->depth_stencil = handle; - ctx->pipe->bind_depth_stencil_alpha_state(ctx->pipe, handle); - } -} - - - -void cso_set_rasterizer(struct cso_context *ctx, - const struct pipe_rasterizer_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_rasterizer_state)); - struct cso_hash_iter iter = cso_find_state_template(ctx->cache, - hash_key, CSO_RASTERIZER, - (void*)templ); - void *handle = NULL; - - if (cso_hash_iter_is_null(iter)) { - struct cso_rasterizer *cso = MALLOC(sizeof(struct cso_rasterizer)); - - cso->state = *templ; - cso->data = ctx->pipe->create_rasterizer_state(ctx->pipe, &cso->state); - cso->delete_state = (cso_state_callback)ctx->pipe->delete_rasterizer_state; - cso->context = ctx->pipe; - - cso_insert_state(ctx->cache, hash_key, CSO_RASTERIZER, cso); - handle = cso->data; - } - else { - handle = ((struct cso_rasterizer *)cso_hash_iter_data(iter))->data; - } - - if (ctx->rasterizer != handle) { - ctx->rasterizer = handle; - ctx->pipe->bind_rasterizer_state(ctx->pipe, handle); - } -} - - - - - -void cso_set_fragment_shader(struct cso_context *ctx, - const struct pipe_shader_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_shader_state)); - struct cso_hash_iter iter = cso_find_state_template(ctx->cache, - hash_key, CSO_FRAGMENT_SHADER, - (void*)templ); - void *handle = NULL; - - if (cso_hash_iter_is_null(iter)) { - struct cso_fragment_shader *cso = MALLOC(sizeof(struct cso_fragment_shader)); - - cso->state = *templ; - cso->data = ctx->pipe->create_fs_state(ctx->pipe, &cso->state); - cso->delete_state = (cso_state_callback)ctx->pipe->delete_fs_state; - cso->context = ctx->pipe; - - iter = cso_insert_state(ctx->cache, hash_key, CSO_FRAGMENT_SHADER, cso); - handle = cso->data; - } - else { - handle = ((struct cso_fragment_shader *)cso_hash_iter_data(iter))->data; - } - - if (ctx->fragment_shader != handle) { - ctx->fragment_shader = handle; - ctx->pipe->bind_fs_state(ctx->pipe, handle); - } -} - -void cso_set_vertex_shader(struct cso_context *ctx, - const struct pipe_shader_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_shader_state)); - struct cso_hash_iter iter = cso_find_state_template(ctx->cache, - hash_key, CSO_VERTEX_SHADER, - (void*)templ); - void *handle = NULL; - - if (cso_hash_iter_is_null(iter)) { - struct cso_vertex_shader *cso = MALLOC(sizeof(struct cso_vertex_shader)); - - cso->state = *templ; - cso->data = ctx->pipe->create_vs_state(ctx->pipe, &cso->state); - cso->delete_state = (cso_state_callback)ctx->pipe->delete_vs_state; - cso->context = ctx->pipe; - - iter = cso_insert_state(ctx->cache, hash_key, CSO_VERTEX_SHADER, cso); - handle = cso->data; - } - else { - handle = ((struct cso_vertex_shader *)cso_hash_iter_data(iter))->data; - } - - if (ctx->vertex_shader != handle) { - ctx->vertex_shader = handle; - ctx->pipe->bind_fs_state(ctx->pipe, handle); - } -} +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* Wrap the cso cache & hash mechanisms in a simplified + * pipe-driver-specific interface. + * + * Authors: + * Zack Rusin + * Keith Whitwell + */ + +#include "pipe/p_state.h" +#include "pipe/p_util.h" + +#include "cso_cache/cso_context.h" +#include "cso_cache/cso_cache.h" +#include "cso_cache/cso_hash.h" + +struct cso_context { + struct pipe_context *pipe; + struct cso_cache *cache; + + struct { + void *samplers[PIPE_MAX_SAMPLERS]; + unsigned nr_samplers; + } hw; + + void *samplers[PIPE_MAX_SAMPLERS]; + unsigned nr_samplers; + + void *blend; + void *depth_stencil; + void *rasterizer; + void *fragment_shader; + void *vertex_shader; +}; + + +struct cso_context *cso_create_context( struct pipe_context *pipe ) +{ + struct cso_context *ctx = CALLOC_STRUCT(cso_context); + if (ctx == NULL) + goto out; + + ctx->cache = cso_cache_create(); + if (ctx->cache == NULL) + goto out; + + ctx->pipe = pipe; + + return ctx; + +out: + cso_destroy_context( ctx ); + return NULL; +} + + +void cso_destroy_context( struct cso_context *ctx ) +{ + if (ctx == NULL) + return; + +/* + if (ctx->pipe) + ctx->pipe->flush( ctx->pipe, PIPE_FLUSH_UNBIND_ALL ); +*/ + + if (ctx->cache) + cso_cache_delete( ctx->cache ); + + FREE( ctx ); +} + + +/* Those function will either find the state of the given template + * in the cache or they will create a new state from the given + * template, insert it in the cache and return it. + */ + +/* + * If the driver returns 0 from the create method then they will assign + * the data member of the cso to be the template itself. + */ + +void cso_set_blend(struct cso_context *ctx, + const struct pipe_blend_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_blend_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_BLEND, + (void*)templ); + void *handle; + + if (cso_hash_iter_is_null(iter)) { + struct cso_blend *cso = MALLOC(sizeof(struct cso_blend)); + + cso->state = *templ; + cso->data = ctx->pipe->create_blend_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_blend_state; + cso->context = ctx->pipe; + + iter = cso_insert_state(ctx->cache, hash_key, CSO_BLEND, cso); + handle = cso->data; + } + else { + handle = ((struct cso_blend *)cso_hash_iter_data(iter))->data; + } + + if (ctx->blend != handle) { + ctx->blend = handle; + ctx->pipe->bind_blend_state(ctx->pipe, handle); + } +} + +void cso_single_sampler(struct cso_context *ctx, + unsigned idx, + const struct pipe_sampler_state *templ) +{ + void *handle = NULL; + + if (templ != NULL) { + unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_sampler_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_SAMPLER, + (void*)templ); + + if (cso_hash_iter_is_null(iter)) { + struct cso_sampler *cso = MALLOC(sizeof(struct cso_sampler)); + + cso->state = *templ; + cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state; + cso->context = ctx->pipe; + + iter = cso_insert_state(ctx->cache, hash_key, CSO_SAMPLER, cso); + handle = cso->data; + } + else { + handle = ((struct cso_sampler *)cso_hash_iter_data(iter))->data; + } + } + + ctx->samplers[idx] = handle; +} + +void cso_single_sampler_done( struct cso_context *ctx ) +{ + unsigned i; + + for (i = 0; i < 8; i++) + if (ctx->samplers[i] == NULL) + break; + + ctx->nr_samplers = i; + + if (ctx->hw.nr_samplers != ctx->nr_samplers || + memcmp(ctx->hw.samplers, + ctx->samplers, + ctx->nr_samplers * sizeof(void *)) != 0) + { + memcpy(ctx->hw.samplers, ctx->samplers, ctx->nr_samplers * sizeof(void *)); + ctx->hw.nr_samplers = ctx->nr_samplers; + + ctx->pipe->bind_sampler_states(ctx->pipe, ctx->nr_samplers, ctx->samplers); + } +} + +void cso_set_samplers( struct cso_context *ctx, + unsigned nr, + const struct pipe_sampler_state **templates ) +{ + unsigned i; + + /* TODO: fastpath + */ + + for (i = 0; i < nr; i++) + cso_single_sampler( ctx, i, templates[i] ); + + for ( ; i < ctx->nr_samplers; i++) + cso_single_sampler( ctx, i, NULL ); + + cso_single_sampler_done( ctx ); +} + +void cso_set_depth_stencil_alpha(struct cso_context *ctx, + const struct pipe_depth_stencil_alpha_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_depth_stencil_alpha_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, + CSO_DEPTH_STENCIL_ALPHA, + (void*)templ); + void *handle; + + if (cso_hash_iter_is_null(iter)) { + struct cso_depth_stencil_alpha *cso = MALLOC(sizeof(struct cso_depth_stencil_alpha)); + + cso->state = *templ; + cso->data = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_depth_stencil_alpha_state; + cso->context = ctx->pipe; + + cso_insert_state(ctx->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso); + handle = cso->data; + } + else { + handle = ((struct cso_depth_stencil_alpha *)cso_hash_iter_data(iter))->data; + } + + if (ctx->depth_stencil != handle) { + ctx->depth_stencil = handle; + ctx->pipe->bind_depth_stencil_alpha_state(ctx->pipe, handle); + } +} + + + +void cso_set_rasterizer(struct cso_context *ctx, + const struct pipe_rasterizer_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_rasterizer_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_RASTERIZER, + (void*)templ); + void *handle = NULL; + + if (cso_hash_iter_is_null(iter)) { + struct cso_rasterizer *cso = MALLOC(sizeof(struct cso_rasterizer)); + + cso->state = *templ; + cso->data = ctx->pipe->create_rasterizer_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_rasterizer_state; + cso->context = ctx->pipe; + + cso_insert_state(ctx->cache, hash_key, CSO_RASTERIZER, cso); + handle = cso->data; + } + else { + handle = ((struct cso_rasterizer *)cso_hash_iter_data(iter))->data; + } + + if (ctx->rasterizer != handle) { + ctx->rasterizer = handle; + ctx->pipe->bind_rasterizer_state(ctx->pipe, handle); + } +} + + + + + +void cso_set_fragment_shader(struct cso_context *ctx, + const struct pipe_shader_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_shader_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_FRAGMENT_SHADER, + (void*)templ); + void *handle = NULL; + + if (cso_hash_iter_is_null(iter)) { + struct cso_fragment_shader *cso = MALLOC(sizeof(struct cso_fragment_shader)); + + cso->state = *templ; + cso->data = ctx->pipe->create_fs_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_fs_state; + cso->context = ctx->pipe; + + iter = cso_insert_state(ctx->cache, hash_key, CSO_FRAGMENT_SHADER, cso); + handle = cso->data; + } + else { + handle = ((struct cso_fragment_shader *)cso_hash_iter_data(iter))->data; + } + + if (ctx->fragment_shader != handle) { + ctx->fragment_shader = handle; + ctx->pipe->bind_fs_state(ctx->pipe, handle); + } +} + +void cso_set_vertex_shader(struct cso_context *ctx, + const struct pipe_shader_state *templ) +{ + unsigned hash_key = cso_construct_key((void*)templ, + sizeof(struct pipe_shader_state)); + struct cso_hash_iter iter = cso_find_state_template(ctx->cache, + hash_key, CSO_VERTEX_SHADER, + (void*)templ); + void *handle = NULL; + + if (cso_hash_iter_is_null(iter)) { + struct cso_vertex_shader *cso = MALLOC(sizeof(struct cso_vertex_shader)); + + cso->state = *templ; + cso->data = ctx->pipe->create_vs_state(ctx->pipe, &cso->state); + cso->delete_state = (cso_state_callback)ctx->pipe->delete_vs_state; + cso->context = ctx->pipe; + + iter = cso_insert_state(ctx->cache, hash_key, CSO_VERTEX_SHADER, cso); + handle = cso->data; + } + else { + handle = ((struct cso_vertex_shader *)cso_hash_iter_data(iter))->data; + } + + if (ctx->vertex_shader != handle) { + ctx->vertex_shader = handle; + ctx->pipe->bind_fs_state(ctx->pipe, handle); + } +} -- cgit v1.2.3 From aff4cf19a753baf0428d2bf53614900e5afea8a3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 9 Mar 2008 20:17:02 +0000 Subject: draw: cope with binding NULL vertex shader (on context delete, for instance) --- src/gallium/auxiliary/draw/draw_vertex_shader.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index 1e95355555..133418baca 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -110,13 +110,20 @@ draw_bind_vertex_shader(struct draw_context *draw, struct draw_vertex_shader *dvs) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + if (dvs) + { + draw->vertex_shader = dvs; + draw->num_vs_outputs = dvs->info.num_outputs; - draw->vertex_shader = dvs; - draw->num_vs_outputs = dvs->info.num_outputs; + tgsi_exec_machine_init(&draw->machine); - tgsi_exec_machine_init(&draw->machine); - - dvs->prepare( dvs, draw ); + dvs->prepare( dvs, draw ); + } + else { + draw->vertex_shader = NULL; + draw->num_vs_outputs = 0; + } } -- cgit v1.2.3 From b041dbe9019ff8cb16ff15d0baaa803c7dc654db Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 9 Mar 2008 20:21:45 +0000 Subject: gallium: avoid deleting currently-bound CSO's on cache destruction --- src/gallium/auxiliary/cso_cache/cso_context.c | 34 +++++++++++++++++++-------- 1 file changed, 24 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 596e5a9ad6..fbb26ca511 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -72,6 +72,9 @@ struct cso_context *cso_create_context( struct pipe_context *pipe ) ctx->pipe = pipe; + /* Enable for testing: */ + if (0) cso_set_maximum_cache_size( ctx->cache, 4 ); + return ctx; out: @@ -79,20 +82,31 @@ out: return NULL; } +static void cso_release_all( struct cso_context *ctx ) +{ + if (ctx->pipe) { + ctx->pipe->bind_blend_state( ctx->pipe, NULL ); + ctx->pipe->bind_rasterizer_state( ctx->pipe, NULL ); + ctx->pipe->bind_sampler_states( ctx->pipe, 0, NULL ); + ctx->pipe->bind_depth_stencil_alpha_state( ctx->pipe, NULL ); + ctx->pipe->bind_fs_state( ctx->pipe, NULL ); + ctx->pipe->bind_vs_state( ctx->pipe, NULL ); + } + + if (ctx->cache) { + cso_cache_delete( ctx->cache ); + ctx->cache = NULL; + } +} + void cso_destroy_context( struct cso_context *ctx ) { - if (ctx == NULL) - return; - -/* - if (ctx->pipe) - ctx->pipe->flush( ctx->pipe, PIPE_FLUSH_UNBIND_ALL ); -*/ + debug_printf("%s\n", __FUNCTION__); + + if (ctx) + cso_release_all( ctx ); - if (ctx->cache) - cso_cache_delete( ctx->cache ); - FREE( ctx ); } -- cgit v1.2.3 From b721bc8792f6add71dede11924d7060bbce72f0e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 10 Mar 2008 13:04:13 +0000 Subject: gallium: WinCE portability fixes. --- src/gallium/auxiliary/draw/draw_unfilled.c | 2 +- src/gallium/auxiliary/util/p_debug.c | 8 ++++++++ src/gallium/drivers/softpipe/sp_quad_alpha_test.c | 2 +- src/gallium/drivers/softpipe/sp_quad_blend.c | 6 +++--- src/gallium/drivers/softpipe/sp_quad_depth_test.c | 2 +- src/gallium/include/pipe/p_util.h | 10 ++++++++++ 6 files changed, 24 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_unfilled.c b/src/gallium/auxiliary/draw/draw_unfilled.c index 4d718d514c..b07860cd9e 100644 --- a/src/gallium/auxiliary/draw/draw_unfilled.c +++ b/src/gallium/auxiliary/draw/draw_unfilled.c @@ -129,7 +129,7 @@ static void unfilled_tri( struct draw_stage *stage, points( stage, header ); break; default: - abort(); + assert(0); } } diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 7b3c030956..6255d716a6 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -57,11 +57,15 @@ int rpl_vsnprintf(char *, size_t, const char *, va_list); void debug_vprintf(const char *format, va_list ap) { #ifdef WIN32 +#ifndef WINCE /* EngDebugPrint does not handle float point arguments, so we need to use * our own vsnprintf implementation */ char buf[512 + 1]; rpl_vsnprintf(buf, sizeof(buf), format, ap); rpl_EngDebugPrint("%s", buf); +#else + /* TODO: Implement debug print for WINCE */ +#endif #else vfprintf(stderr, format, ap); #endif @@ -80,7 +84,11 @@ void debug_printf(const char *format, ...) static INLINE void debug_abort(void) { #ifdef WIN32 +#ifndef WINCE EngDebugBreak(); +#else + _asm {int 3}; +#endif #else abort(); #endif diff --git a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c index 4ffeac35e1..318916fe30 100644 --- a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c @@ -72,7 +72,7 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) passMask = MASK_ALL; break; default: - abort(); + assert(0); } quad->mask &= passMask; diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index 17f3ecd0b8..0b79cfa1ed 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -392,7 +392,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) assert(0); /* to do */ break; default: - abort(); + assert(0); } /* @@ -464,7 +464,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) } break; default: - abort(); + assert(0); } @@ -716,7 +716,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) VEC4_MAX(quadColor[3], source[3], dest[3]); /* A */ break; default: - abort(); + assert(0); } /* pass blended quad to next stage */ diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c index a9a0754f27..a1859f9883 100644 --- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c @@ -185,7 +185,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) zmask = MASK_ALL; break; default: - abort(); + assert(0); } quad->mask &= zmask; diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 3b32ba1d24..ef36ce75f7 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -54,7 +54,12 @@ EngFreeMem( static INLINE void * MALLOC( unsigned size ) { +#ifdef WINCE + /* TODO: Need to abstract this */ + return malloc( size ); +#else return EngAllocMem( 0, size, 'D3AG' ); +#endif } static INLINE void * @@ -71,7 +76,12 @@ static INLINE void FREE( void *ptr ) { if( ptr ) { +#ifdef WINCE + /* TODO: Need to abstract this */ + free( ptr ); +#else EngFreeMem( ptr ); +#endif } } -- cgit v1.2.3 From 01bd21eef8f572944c09771f44e3006e2991280e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 10 Mar 2008 16:45:19 +0000 Subject: gallium: Import Dennis Smit cpu detection code. It still needs a slight code massasing to integrate with the rest of gallium (namely mapping the OS_* ARCH_* defines), but I'm commiting anyway so that it is available to be used when somebody needs it. --- src/gallium/auxiliary/util/u_cpu_detect.c | 506 ++++++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_cpu_detect.h | 78 +++++ 2 files changed, 584 insertions(+) create mode 100644 src/gallium/auxiliary/util/u_cpu_detect.c create mode 100644 src/gallium/auxiliary/util/u_cpu_detect.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_cpu_detect.c b/src/gallium/auxiliary/util/u_cpu_detect.c new file mode 100644 index 0000000000..d9f2f8fc28 --- /dev/null +++ b/src/gallium/auxiliary/util/u_cpu_detect.c @@ -0,0 +1,506 @@ +/************************************************************************** + * + * Copyright 2008 Dennis Smit + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * Based on the work of Eric Anholt + */ + +/* FIXME: clean this entire file up */ + +#include "u_cpu_detect.h" + +#ifdef __linux__ +#define OS_LINUX +#endif +#ifdef WIN32 +#define OS_WIN32 +#endif + +#if defined(ARCH_POWERPC) +#if defined(OS_DARWIN) +#include +#else +#include +#include +#endif +#endif + +#if defined(OS_NETBSD) || defined(OS_OPENBSD) +#include +#include +#include +#endif + +#if defined(OS_FREEBSD) +#include +#include +#endif + +#if defined(OS_LINUX) +#include +#endif + +#if defined(OS_WIN32) +#include +#endif + +#include +#include +#include +#include + + +static struct cpu_detect_caps __cpu_detect_caps; +static int __cpu_detect_initialized = 0; + +static int has_cpuid(void); +static int cpuid(unsigned int ax, unsigned int *p); + +/* The sigill handlers */ +#if defined(ARCH_X86) /* x86 (linux katmai handler check thing) */ +#if defined(OS_LINUX) && defined(_POSIX_SOURCE) && defined(X86_FXSR_MAGIC) +static void sigill_handler_sse(int signal, struct sigcontext sc) +{ + /* Both the "xorps %%xmm0,%%xmm0" and "divps %xmm0,%%xmm1" + * instructions are 3 bytes long. We must increment the instruction + * pointer manually to avoid repeated execution of the offending + * instruction. + * + * If the SIGILL is caused by a divide-by-zero when unmasked + * exceptions aren't supported, the SIMD FPU status and control + * word will be restored at the end of the test, so we don't need + * to worry about doing it here. Besides, we may not be able to... + */ + sc.eip += 3; + + __cpu_detect_caps.hasSSE=0; +} + +static void sigfpe_handler_sse(int signal, struct sigcontext sc) +{ + if (sc.fpstate->magic != 0xffff) { + /* Our signal context has the extended FPU state, so reset the + * divide-by-zero exception mask and clear the divide-by-zero + * exception bit. + */ + sc.fpstate->mxcsr |= 0x00000200; + sc.fpstate->mxcsr &= 0xfffffffb; + } else { + /* If we ever get here, we're completely hosed. + */ + } +} +#endif +#endif /* OS_LINUX && _POSIX_SOURCE && X86_FXSR_MAGIC */ + +#if defined(OS_WIN32) +LONG CALLBACK win32_sig_handler_sse(EXCEPTION_POINTERS* ep) +{ + if(ep->ExceptionRecord->ExceptionCode==EXCEPTION_ILLEGAL_INSTRUCTION){ + ep->ContextRecord->Eip +=3; + __cpu_detect_caps.hasSSE=0; + return EXCEPTION_CONTINUE_EXECUTION; + } + return EXCEPTION_CONTINUE_SEARCH; +} +#endif /* OS_WIN32 */ + + +#if defined(ARCH_POWERPC) && !defined(OS_DARWIN) +static sigjmp_buf __lv_powerpc_jmpbuf; +static volatile sig_atomic_t __lv_powerpc_canjump = 0; + +static void sigill_handler (int sig); + +static void sigill_handler (int sig) +{ + if (!__lv_powerpc_canjump) { + signal (sig, SIG_DFL); + raise (sig); + } + + __lv_powerpc_canjump = 0; + siglongjmp(__lv_powerpc_jmpbuf, 1); +} + +static void check_os_altivec_support(void) +{ +#if defined(OS_DARWIN) + int sels[2] = {CTL_HW, HW_VECTORUNIT}; + int has_vu = 0; + int len = sizeof (has_vu); + int err; + + err = sysctl(sels, 2, &has_vu, &len, NULL, 0); + + if (err == 0) { + if (has_vu != 0) { + __cpu_detect_caps.hasAltiVec = 1; + } + } +#else /* !OS_DARWIN */ + /* no Darwin, do it the brute-force way */ + /* this is borrowed from the libmpeg2 library */ + signal(SIGILL, sigill_handler); + if (sigsetjmp(__lv_powerpc_jmpbuf, 1)) { + signal(SIGILL, SIG_DFL); + } else { + __lv_powerpc_canjump = 1; + + __asm __volatile + ("mtspr 256, %0\n\t" + "vand %%v0, %%v0, %%v0" + : + : "r" (-1)); + + signal(SIGILL, SIG_DFL); + __cpu_detect_caps.hasAltiVec = 1; + } +#endif +} +#endif + +/* If we're running on a processor that can do SSE, let's see if we + * are allowed to or not. This will catch 2.4.0 or later kernels that + * haven't been configured for a Pentium III but are running on one, + * and RedHat patched 2.2 kernels that have broken exception handling + * support for user space apps that do SSE. + */ +static void check_os_katmai_support(void) +{ +#if defined(ARCH_X86) +#if defined(OS_FREEBSD) + int has_sse=0, ret; + int len = sizeof (has_sse); + + ret = sysctlbyname("hw.instruction_sse", &has_sse, &len, NULL, 0); + if (ret || !has_sse) + __cpu_detect_caps.hasSSE=0; + +#elif defined(OS_NETBSD) || defined(OS_OPENBSD) + int has_sse, has_sse2, ret, mib[2]; + int varlen; + + mib[0] = CTL_MACHDEP; + mib[1] = CPU_SSE; + varlen = sizeof (has_sse); + + ret = sysctl(mib, 2, &has_sse, &varlen, NULL, 0); + if (ret < 0 || !has_sse) { + __cpu_detect_caps.hasSSE = 0; + } else { + __cpu_detect_caps.hasSSE = 1; + } + + mib[1] = CPU_SSE2; + varlen = sizeof (has_sse2); + ret = sysctl(mib, 2, &has_sse2, &varlen, NULL, 0); + if (ret < 0 || !has_sse2) { + __cpu_detect_caps.hasSSE2 = 0; + } else { + __cpu_detect_caps.hasSSE2 = 1; + } + __cpu_detect_caps.hasSSE = 0; /* FIXME ?!?!? */ + +#elif defined(OS_WIN32) + LPTOP_LEVEL_EXCEPTION_FILTER exc_fil; + if (__cpu_detect_caps.hasSSE) { + exc_fil = SetUnhandledExceptionFilter(win32_sig_handler_sse); + __asm __volatile ("xorps %xmm0, %xmm0"); + SetUnhandledExceptionFilter(exc_fil); + } +#elif defined(OS_LINUX) + struct sigaction saved_sigill; + struct sigaction saved_sigfpe; + + /* Save the original signal handlers. + */ + sigaction(SIGILL, NULL, &saved_sigill); + sigaction(SIGFPE, NULL, &saved_sigfpe); + + signal(SIGILL, (void (*)(int))sigill_handler_sse); + signal(SIGFPE, (void (*)(int))sigfpe_handler_sse); + + /* Emulate test for OSFXSR in CR4. The OS will set this bit if it + * supports the extended FPU save and restore required for SSE. If + * we execute an SSE instruction on a PIII and get a SIGILL, the OS + * doesn't support Streaming SIMD Exceptions, even if the processor + * does. + */ + if (__cpu_detect_caps.hasSSE) { + __asm __volatile ("xorps %xmm1, %xmm0"); + } + + /* Emulate test for OSXMMEXCPT in CR4. The OS will set this bit if + * it supports unmasked SIMD FPU exceptions. If we unmask the + * exceptions, do a SIMD divide-by-zero and get a SIGILL, the OS + * doesn't support unmasked SIMD FPU exceptions. If we get a SIGFPE + * as expected, we're okay but we need to clean up after it. + * + * Are we being too stringent in our requirement that the OS support + * unmasked exceptions? Certain RedHat 2.2 kernels enable SSE by + * setting CR4.OSFXSR but don't support unmasked exceptions. Win98 + * doesn't even support them. We at least know the user-space SSE + * support is good in kernels that do support unmasked exceptions, + * and therefore to be safe I'm going to leave this test in here. + */ + if (__cpu_detect_caps.hasSSE) { + // test_os_katmai_exception_support(); + } + + /* Restore the original signal handlers. + */ + sigaction(SIGILL, &saved_sigill, NULL); + sigaction(SIGFPE, &saved_sigfpe, NULL); + +#else + /* We can't use POSIX signal handling to test the availability of + * SSE, so we disable it by default. + */ + __cpu_detect_caps.hasSSE = 0; +#endif /* __linux__ */ +#endif +} + + +static int has_cpuid(void) +{ +#if defined(ARCH_X86) + int a, c; + + __asm __volatile + ("pushf\n" + "popl %0\n" + "movl %0, %1\n" + "xorl $0x200000, %0\n" + "push %0\n" + "popf\n" + "pushf\n" + "popl %0\n" + : "=a" (a), "=c" (c) + : + : "cc"); + + return a != c; +#else + return 0; +#endif +} + +static int cpuid(unsigned int ax, unsigned int *p) +{ +#if defined(ARCH_X86) + unsigned int flags; + + __asm __volatile + ("movl %%ebx, %%esi\n\t" + "cpuid\n\t" + "xchgl %%ebx, %%esi" + : "=a" (p[0]), "=S" (p[1]), + "=c" (p[2]), "=d" (p[3]) + : "0" (ax)); + + return 0; +#else + return -1; +#endif +} + +void cpu_detect_initialize() +{ + unsigned int regs[4]; + unsigned int regs2[4]; + + int mib[2], ncpu; + int len; + + memset(&__cpu_detect_caps, 0, sizeof (struct cpu_detect_caps)); + + /* Check for arch type */ +#if defined(ARCH_MIPS) + __cpu_detect_caps.type = CPU_DETECT_TYPE_MIPS; +#elif defined(ARCH_ALPHA) + __cpu_detect_caps.type = CPU_DETECT_TYPE_ALPHA; +#elif defined(ARCH_SPARC) + __cpu_detect_caps.type = CPU_DETECT_TYPE_SPARC; +#elif defined(ARCH_X86) + __cpu_detect_caps.type = CPU_DETECT_TYPE_X86; +#elif defined(ARCH_POWERPC) + __cpu_detect_caps.type = CPU_DETECT_TYPE_POWERPC; +#else + __cpu_detect_caps.type = CPU_DETECT_TYPE_OTHER; +#endif + + /* Count the number of CPUs in system */ +#if !defined(OS_WIN32) && !defined(OS_UNKNOWN) && defined(_SC_NPROCESSORS_ONLN) + __cpu_detect_caps.nrcpu = sysconf(_SC_NPROCESSORS_ONLN); + if (__cpu_detect_caps.nrcpu == -1) + __cpu_detect_caps.nrcpu = 1; + +#elif defined(OS_NETBSD) || defined(OS_FREEBSD) || defined(OS_OPENBSD) + + mib[0] = CTL_HW; + mib[1] = HW_NCPU; + + len = sizeof (ncpu); + sysctl(mib, 2, &ncpu, &len, NULL, 0); + __cpu_detect_caps.nrcpu = ncpu; + +#else + __cpu_detect_caps.nrcpu = 1; +#endif + +#if defined(ARCH_X86) + /* No cpuid, old 486 or lower */ + if (has_cpuid() == 0) + return; + + __cpu_detect_caps.cacheline = 32; + + /* Get max cpuid level */ + cpuid(0x00000000, regs); + + if (regs[0] >= 0x00000001) { + unsigned int cacheline; + + cpuid (0x00000001, regs2); + + __cpu_detect_caps.x86cpuType = (regs2[0] >> 8) & 0xf; + if (__cpu_detect_caps.x86cpuType == 0xf) + __cpu_detect_caps.x86cpuType = 8 + ((regs2[0] >> 20) & 255); /* use extended family (P4, IA64) */ + + /* general feature flags */ + __cpu_detect_caps.hasTSC = (regs2[3] & (1 << 8 )) >> 8; /* 0x0000010 */ + __cpu_detect_caps.hasMMX = (regs2[3] & (1 << 23 )) >> 23; /* 0x0800000 */ + __cpu_detect_caps.hasSSE = (regs2[3] & (1 << 25 )) >> 25; /* 0x2000000 */ + __cpu_detect_caps.hasSSE2 = (regs2[3] & (1 << 26 )) >> 26; /* 0x4000000 */ + __cpu_detect_caps.hasSSE3 = (regs2[2] & (1)); /* 0x0000001 */ + __cpu_detect_caps.hasSSSE3 = (regs2[2] & (1 << 9 )) >> 9; /* 0x0000020 */ + __cpu_detect_caps.hasMMX2 = __cpu_detect_caps.hasSSE; /* SSE cpus supports mmxext too */ + + cacheline = ((regs2[1] >> 8) & 0xFF) * 8; + if (cacheline > 0) + __cpu_detect_caps.cacheline = cacheline; + } + + cpuid(0x80000000, regs); + + if (regs[0] >= 0x80000001) { + + cpuid(0x80000001, regs2); + + __cpu_detect_caps.hasMMX |= (regs2[3] & (1 << 23 )) >> 23; /* 0x0800000 */ + __cpu_detect_caps.hasMMX2 |= (regs2[3] & (1 << 22 )) >> 22; /* 0x400000 */ + __cpu_detect_caps.has3DNow = (regs2[3] & (1 << 31 )) >> 31; /* 0x80000000 */ + __cpu_detect_caps.has3DNowExt = (regs2[3] & (1 << 30 )) >> 30; + } + + if (regs[0] >= 0x80000006) { + cpuid(0x80000006, regs2); + __cpu_detect_caps.cacheline = regs2[2] & 0xFF; + } + + +#if defined(OS_LINUX) || defined(OS_FREEBSD) || defined(OS_NETBSD) || defined(OS_CYGWIN) || defined(OS_OPENBSD) + if (__cpu_detect_caps.hasSSE) + check_os_katmai_support(); + + if (!__cpu_detect_caps.hasSSE) { + __cpu_detect_caps.hasSSE2 = 0; + __cpu_detect_caps.hasSSE3 = 0; + __cpu_detect_caps.hasSSSE3 = 0; + } +#else + __cpu_detect_caps.hasSSE = 0; + __cpu_detect_caps.hasSSE2 = 0; + __cpu_detect_caps.hasSSE3 = 0; + __cpu_detect_caps.hasSSSE3 = 0; +#endif +#endif /* ARCH_X86 */ + +#if defined(ARCH_POWERPC) + check_os_altivec_support(); +#endif /* ARCH_POWERPC */ + + __cpu_detect_initialized = 1; +} + +struct cpu_detect_caps *cpu_detect_get_caps() +{ + return &__cpu_detect_caps; +} + +/* The getters and setters for feature flags */ +int cpu_detect_get_tsc() +{ + return __cpu_detect_caps.hasTSC; +} + +int cpu_detect_get_mmx() +{ + return __cpu_detect_caps.hasMMX; +} + +int cpu_detect_get_mmx2() +{ + return __cpu_detect_caps.hasMMX2; +} + +int cpu_detect_get_sse() +{ + return __cpu_detect_caps.hasSSE; +} + +int cpu_detect_get_sse2() +{ + return __cpu_detect_caps.hasSSE2; +} + +int cpu_detect_get_sse3() +{ + return __cpu_detect_caps.hasSSE3; +} + +int cpu_detect_get_ssse3() +{ + return __cpu_detect_caps.hasSSSE3; +} + +int cpu_detect_get_3dnow() +{ + return __cpu_detect_caps.has3DNow; +} + +int cpu_detect_get_3dnow2() +{ + return __cpu_detect_caps.has3DNowExt; +} + +int cpu_detect_get_altivec() +{ + return __cpu_detect_caps.hasAltiVec; +} + diff --git a/src/gallium/auxiliary/util/u_cpu_detect.h b/src/gallium/auxiliary/util/u_cpu_detect.h new file mode 100644 index 0000000000..1612d49286 --- /dev/null +++ b/src/gallium/auxiliary/util/u_cpu_detect.h @@ -0,0 +1,78 @@ +/************************************************************************** + * + * Copyright 2008 Dennis Smit + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + ***************************************************************************/ + +/* + * Based on the work of Eric Anholt + */ + +#ifndef _CPU_DETECT_H +#define _CPU_DETECT_H + +typedef enum { + CPU_DETECT_TYPE_MIPS, + CPU_DETECT_TYPE_ALPHA, + CPU_DETECT_TYPE_SPARC, + CPU_DETECT_TYPE_X86, + CPU_DETECT_TYPE_POWERPC, + CPU_DETECT_TYPE_OTHER +} cpu_detect_type; + +struct cpu_detect_caps { + cpu_detect_type type; + int nrcpu; + + /* Feature flags */ + int x86cpuType; + int cacheline; + + int hasTSC; + int hasMMX; + int hasMMX2; + int hasSSE; + int hasSSE2; + int hasSSE3; + int hasSSSE3; + int has3DNow; + int has3DNowExt; + int hasAltiVec; +}; + +/* prototypes */ +void cpu_detect_initialize(void); +struct cpu_detect_caps *cpu_detect_get_caps(void); + +int cpu_detect_get_tsc(void); +int cpu_detect_get_mmx(void); +int cpu_detect_get_mmx2(void); +int cpu_detect_get_sse(void); +int cpu_detect_get_sse2(void); +int cpu_detect_get_sse3(void); +int cpu_detect_get_ssse3(void); +int cpu_detect_get_3dnow(void); +int cpu_detect_get_3dnow2(void); +int cpu_detect_get_altivec(void); + +#endif /* _CPU_DETECT_H */ -- cgit v1.2.3 From 34497eabc803fd979c58824d4737cabde1fb47e5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 10 Mar 2008 21:15:31 +0000 Subject: Some notes about debugging --- src/gallium/README.portability | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/README.portability b/src/gallium/README.portability index c70ca774da..ab0c197847 100644 --- a/src/gallium/README.portability +++ b/src/gallium/README.portability @@ -39,5 +39,6 @@ portable way. == Debugging == -TODO +* Use the functions/macros in p_debug.h. +* Don't include assert.h, call abort, printf, etc. -- cgit v1.2.3 From 31022681543932ae4fce2bae5bc9d024c8178f1a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 10 Mar 2008 21:18:00 +0000 Subject: gallium: Use hardcoded breakpoints on x86 targets. --- src/gallium/auxiliary/util/p_debug.c | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 6255d716a6..93bfaea393 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -81,14 +81,17 @@ void debug_printf(const char *format, ...) } -static INLINE void debug_abort(void) +/* TODO: implement a debug_abort that calls EngBugCheckEx on WIN32 */ + + +static INLINE void debug_break(void) { -#ifdef WIN32 -#ifndef WINCE - EngDebugBreak(); -#else +#if (defined(__i386__) || defined(__386__)) && defined(__GNUC__) + __asm("int3"); +#elif (defined(__i386__) || defined(__386__)) && defined(__MSC__) _asm {int 3}; -#endif +#elif defined(WIN32) && !defined(WINCE) + EngDebugBreak(); #else abort(); #endif @@ -98,5 +101,5 @@ static INLINE void debug_abort(void) void debug_assert_fail(const char *expr, const char *file, unsigned line) { debug_printf("%s:%i: Assertion `%s' failed.\n", file, line, expr); - debug_abort(); + debug_break(); } -- cgit v1.2.3 From 5b82d551b7a4954b24059585fea207f3a250ec0f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Mar 2008 16:43:36 -0600 Subject: cell: sync up with sampler/texture state-setting changes --- src/gallium/drivers/cell/ppu/cell_context.h | 2 ++ src/gallium/drivers/cell/ppu/cell_pipe_state.c | 32 ++++++++++++++++++-------- src/gallium/drivers/cell/ppu/cell_state_emit.c | 6 +++-- 3 files changed, 28 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index bf27289f3f..c568922cbd 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -63,6 +63,7 @@ struct cell_context const struct pipe_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + uint num_samplers; const struct pipe_depth_stencil_alpha_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct cell_vertex_shader_state *vs; @@ -75,6 +76,7 @@ struct cell_context struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; struct cell_texture *texture[PIPE_MAX_SAMPLERS]; + uint num_textures; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 075e0a0c47..025ed3bbbf 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -210,15 +210,19 @@ cell_create_sampler_state(struct pipe_context *pipe, static void -cell_bind_sampler_state(struct pipe_context *pipe, - unsigned unit, void *sampler) +cell_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **samplers) { struct cell_context *cell = cell_context(pipe); draw_flush(cell->draw); assert(unit < PIPE_MAX_SAMPLERS); - cell->sampler[unit] = (struct pipe_sampler_state *)sampler; + + memcpy(cell->sampler, samplers, num * sizeof(void *)); + memset(&cell->sampler[num], 0, (PIPE_MAX_SAMPLERS - num) * + sizeof(void *)); + cell->num_samplers = num; cell->dirty |= CELL_NEW_SAMPLER; } @@ -234,16 +238,24 @@ cell_delete_sampler_state(struct pipe_context *pipe, static void -cell_set_sampler_texture(struct pipe_context *pipe, - unsigned sampler, - struct pipe_texture *texture) +cell_set_sampler_textures(struct pipe_context *pipe, + unsigned num, struct pipe_texture **texture) { struct cell_context *cell = cell_context(pipe); + uint i; + + /* Check for no-op */ + if (num == cell->num_textures && + !memcmp(cell->texture, texture, num * sizeof(struct pipe_texture *))) + return; draw_flush(cell->draw); - pipe_texture_reference((struct pipe_texture **) &cell->texture[sampler], - texture); + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + struct pipe_texture *tex = i < num ? texture[i] : NULL; + + pipe_texture_reference((struct pipe_texture **) &cell->texture[i], tex); + } cell_update_texture_mapping(cell); @@ -300,10 +312,10 @@ cell_init_state_functions(struct cell_context *cell) cell->pipe.delete_blend_state = cell_delete_blend_state; cell->pipe.create_sampler_state = cell_create_sampler_state; - cell->pipe.bind_sampler_state = cell_bind_sampler_state; + cell->pipe.bind_sampler_states = cell_bind_sampler_states; cell->pipe.delete_sampler_state = cell_delete_sampler_state; - cell->pipe.set_sampler_texture = cell_set_sampler_texture; + cell->pipe.set_sampler_textures = cell_set_sampler_textures; cell->pipe.create_depth_stencil_alpha_state = cell_create_depth_stencil_alpha_state; cell->pipe.bind_depth_stencil_alpha_state = cell_bind_depth_stencil_alpha_state; diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 49c0d130c5..670eb26bdd 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -77,8 +77,10 @@ cell_emit_state(struct cell_context *cell) } if (cell->dirty & CELL_NEW_SAMPLER) { - emit_state_cmd(cell, CELL_CMD_STATE_SAMPLER, - cell->sampler[0], sizeof(struct pipe_sampler_state)); + if (cell->sampler[0]) { + emit_state_cmd(cell, CELL_CMD_STATE_SAMPLER, + cell->sampler[0], sizeof(struct pipe_sampler_state)); + } } if (cell->dirty & CELL_NEW_TEXTURE) { -- cgit v1.2.3 From d5692cb349fb74e8f9d3a18f5bbd788b09b93581 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Mar 2008 17:21:43 -0600 Subject: gallium: new surface/context tracker (sct) module Will be used for tracking the surfaces and textures which are bound/used by contexts. --- src/gallium/auxiliary/sct/Makefile | 12 + src/gallium/auxiliary/sct/SConscript | 9 + src/gallium/auxiliary/sct/sct.c | 453 +++++++++++++++++++++++++++++++++++ src/gallium/auxiliary/sct/sct.h | 123 ++++++++++ src/gallium/auxiliary/sct/usage.c | 61 +++++ 5 files changed, 658 insertions(+) create mode 100644 src/gallium/auxiliary/sct/Makefile create mode 100644 src/gallium/auxiliary/sct/SConscript create mode 100644 src/gallium/auxiliary/sct/sct.c create mode 100644 src/gallium/auxiliary/sct/sct.h create mode 100644 src/gallium/auxiliary/sct/usage.c (limited to 'src') diff --git a/src/gallium/auxiliary/sct/Makefile b/src/gallium/auxiliary/sct/Makefile new file mode 100644 index 0000000000..516d1756cf --- /dev/null +++ b/src/gallium/auxiliary/sct/Makefile @@ -0,0 +1,12 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = sct + +C_SOURCES = \ + sct.c + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/auxiliary/sct/SConscript b/src/gallium/auxiliary/sct/SConscript new file mode 100644 index 0000000000..76927d973f --- /dev/null +++ b/src/gallium/auxiliary/sct/SConscript @@ -0,0 +1,9 @@ +Import('*') + +sct = env.ConvenienceLibrary( + target = 'sct', + source = [ + 'sct.c' + ]) + +auxiliaries.insert(0, sct) diff --git a/src/gallium/auxiliary/sct/sct.c b/src/gallium/auxiliary/sct/sct.c new file mode 100644 index 0000000000..97ee5882a1 --- /dev/null +++ b/src/gallium/auxiliary/sct/sct.c @@ -0,0 +1,453 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" +#include "sct.h" + + +struct texture_list +{ + struct pipe_texture *texture; + struct texture_list *next; +}; + + + +#define MAX_SURFACES ((PIPE_MAX_COLOR_BUFS) + 1) + +struct sct_context +{ + const struct pipe_context *context; + + /** surfaces the context is drawing into */ + struct pipe_surface *surfaces[MAX_SURFACES]; + + /** currently bound textures */ + struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; + + /** previously bound textures, used but not flushed */ + struct texture_list *textures_used; + + boolean needs_flush; + + struct sct_context *next; +}; + + + +struct sct_surface +{ + const struct pipe_surface *surface; + + /** list of contexts drawing to this surface */ + struct sct_context_list *contexts; + + struct sct_surface *next; +}; + + + +/** + * Find the surface_info for the given pipe_surface + */ +static struct sct_surface * +find_surface_info(struct surface_context_tracker *sct, + const struct pipe_surface *surface) +{ + struct sct_surface *si; + for (si = sct->surfaces; si; si = si->next) + if (si->surface == surface) + return si; + return NULL; +} + + +/** + * As above, but create new surface_info if surface is new. + */ +static struct sct_surface * +find_create_surface_info(struct surface_context_tracker *sct, + const struct pipe_surface *surface) +{ + struct sct_surface *si = find_surface_info(sct, surface); + if (si) + return si; + + /* alloc new */ + si = CALLOC_STRUCT(sct_surface); + if (si) { + si->surface = surface; + + /* insert at head */ + si->next = sct->surfaces; + sct->surfaces = si; + } + + return si; +} + + +/** + * Find a context_info for the given context. + */ +static struct sct_context * +find_context_info(struct surface_context_tracker *sct, + const struct pipe_context *context) +{ + struct sct_context *ci; + for (ci = sct->contexts; ci; ci = ci->next) + if (ci->context == context) + return ci; + return NULL; +} + + +/** + * As above, but create new context_info if context is new. + */ +static struct sct_context * +find_create_context_info(struct surface_context_tracker *sct, + const struct pipe_context *context) +{ + struct sct_context *ci = find_context_info(sct, context); + if (ci) + return ci; + + /* alloc new */ + ci = CALLOC_STRUCT(sct_context); + if (ci) { + ci->context = context; + + /* insert at head */ + ci->next = sct->contexts; + sct->contexts = ci; + } + + return ci; +} + + +/** + * Is the context already bound to the surface? + */ +static boolean +find_surface_context(const struct sct_surface *si, + const struct pipe_context *context) +{ + const struct sct_context_list *cl; + for (cl = si->contexts; cl; cl = cl->next) { + if (cl->context == context) { + return TRUE; + } + } + return FALSE; +} + + +/** + * Add a context to the list of contexts associated with a surface. + */ +static void +add_context_to_surface(struct sct_surface *si, + const struct pipe_context *context) +{ + struct sct_context_list *cl = CALLOC_STRUCT(sct_context_list); + if (cl) { + cl->context = context; + /* insert at head of list of contexts */ + cl->next = si->contexts; + si->contexts = cl; + } +} + + +/** + * Remove a context from the list of contexts associated with a surface. + */ +static void +remove_context_from_surface(struct sct_surface *si, + const struct pipe_context *context) +{ + struct sct_context_list *prev = NULL, *curr, *next; + + for (curr = si->contexts; curr; curr = next) { + if (curr->context == context) { + /* remove */ + if (prev) + prev->next = curr->next; + else + si->contexts = curr->next; + next = curr->next; + FREE(curr); + } + else { + prev = curr; + } + } +} + + +/** + * Unbind context from surface. + */ +static void +unbind_context_surface(struct surface_context_tracker *sct, + struct pipe_context *context, + struct pipe_surface *surface) +{ + struct sct_surface *si = find_surface_info(sct, surface); + if (si) { + remove_context_from_surface(si, context); + } +} + + +/** + * Bind context to a set of surfaces (color + Z). + * Like MakeCurrent(). + */ +void +sct_bind_surfaces(struct surface_context_tracker *sct, + struct pipe_context *context, + uint num_surf, + struct pipe_surface **surfaces) +{ + struct sct_context *ci = find_create_context_info(sct, context); + uint i; + + if (!ci) { + return; /* out of memory */ + } + + /* unbind currently bound surfaces */ + for (i = 0; i < MAX_SURFACES; i++) { + if (ci->surfaces[i]) { + unbind_context_surface(sct, context, ci->surfaces[i]); + } + } + + /* bind new surfaces */ + for (i = 0; i < num_surf; i++) { + struct sct_surface *si = find_create_surface_info(sct, surfaces[i]); + if (!find_surface_context(si, context)) { + add_context_to_surface(si, context); + } + } +} + + +/** + * Return list of contexts bound to a surface. + */ +const struct sct_context_list * +sct_get_surface_contexts(struct surface_context_tracker *sct, + const struct pipe_surface *surface) +{ + const struct sct_surface *si = find_surface_info(sct, surface); + return si->contexts; +} + + + +static boolean +find_texture(const struct sct_context *ci, + const struct pipe_texture *texture) +{ + const struct texture_list *tl; + + for (tl = ci->textures_used; tl; tl = tl->next) { + if (tl->texture == texture) { + return TRUE; + } + } + return FALSE; +} + + +/** + * Add the given texture to the context's list of used textures. + */ +static void +add_texture_used(struct sct_context *ci, + struct pipe_texture *texture) +{ + if (!find_texture(ci, texture)) { + /* add to list */ + struct texture_list *tl = CALLOC_STRUCT(texture_list); + if (tl) { + pipe_texture_reference(&tl->texture, texture); + /* insert at head */ + tl->next = ci->textures_used; + ci->textures_used = tl; + } + } +} + + +/** + * Bind a texture to a rendering context. + */ +void +sct_bind_texture(struct surface_context_tracker *sct, + struct pipe_context *context, + uint unit, + struct pipe_texture *tex) +{ + struct sct_context *ci = find_context_info(sct, context); + + if (ci->textures[unit] != tex) { + /* put texture on the 'used' list */ + add_texture_used(ci, tex); + /* bind new */ + pipe_texture_reference(&ci->textures[unit], tex); + } +} + + +/** + * Check if the given texture has been used by the rendering context + * since the last call to sct_flush_textures(). + */ +boolean +sct_is_texture_used(struct surface_context_tracker *sct, + const struct pipe_context *context, + const struct pipe_texture *texture) +{ + const struct sct_context *ci = find_context_info(sct, context); + return find_texture(ci, texture); +} + + +/** + * To be called when the image contents of a texture are changed, such + * as for gl[Copy]TexSubImage(). + * XXX this may not be needed + */ +void +sct_update_texture(struct pipe_texture *tex) +{ + +} + + +/** + * When a scene is flushed/rendered we can release the list of + * used textures. + */ +void +sct_flush_textures(struct surface_context_tracker *sct, + struct pipe_context *context) +{ + struct sct_context *ci = find_context_info(sct, context); + struct texture_list *tl, *next; + uint i; + + for (tl = ci->textures_used; tl; tl = next) { + next = tl->next; + pipe_texture_release(&tl->texture); + FREE(tl); + } + ci->textures_used = NULL; + + /* put the currently bound textures on the 'used' list */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + add_texture_used(ci, ci->textures[i]); + } +} + + + +void +sct_destroy_context(struct surface_context_tracker *sct, + struct pipe_context *context) +{ + /* XXX should we require an unbinding first? */ + { + struct sct_surface *si; + for (si = sct->surfaces; si; si = si->next) { + remove_context_from_surface(si, context); + } + } + + /* remove context from context_info list */ + { + struct sct_context *ci, *next, *prev = NULL; + for (ci = sct->contexts; ci; ci = next) { + next = ci->next; + if (ci->context == context) { + if (prev) + prev->next = ci->next; + else + sct->contexts = ci->next; + FREE(ci); + } + else { + prev = ci; + } + } + } + +} + + +void +sct_destroy_surface(struct surface_context_tracker *sct, + struct pipe_surface *surface) +{ + if (1) { + /* debug/sanity: no context should be bound to surface */ + struct sct_context *ci; + uint i; + for (ci = sct->contexts; ci; ci = ci->next) { + for (i = 0; i < MAX_SURFACES; i++) { + assert(ci->surfaces[i] != surface); + } + } + } + + /* remove surface from sct_surface list */ + { + struct sct_surface *si, *next, *prev = NULL; + for (si = sct->surfaces; si; si = next) { + next = si->next; + if (si->surface == surface) { + /* unlink */ + if (prev) + prev->next = si->next; + else + sct->surfaces = si->next; + FREE(si); + } + else { + prev = si; + } + } + } +} diff --git a/src/gallium/auxiliary/sct/sct.h b/src/gallium/auxiliary/sct/sct.h new file mode 100644 index 0000000000..cf7c4d3bdf --- /dev/null +++ b/src/gallium/auxiliary/sct/sct.h @@ -0,0 +1,123 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Surface/Context Tracking + * + * For some drivers, we need to monitor the binding between contexts and + * surfaces/textures. + * This code may evolve quite a bit... + */ + + +#ifndef SCT_H +#define SCT_H + + +#ifdef __cplusplus +extern "C" { +#endif + + +struct pipe_context; +struct pipe_surface; + +struct sct_context; +struct sct_surface; + + +/** + * Per-device info, basically + */ +struct surface_context_tracker +{ + struct sct_context *contexts; + struct sct_surface *surfaces; +}; + + + +/** + * Simple linked list of contexts + */ +struct sct_context_list +{ + const struct pipe_context *context; + struct sct_context_list *next; +}; + + + +extern void +sct_bind_surfaces(struct surface_context_tracker *sct, + struct pipe_context *context, + uint num_surf, + struct pipe_surface **surfaces); + + +extern void +sct_bind_texture(struct surface_context_tracker *sct, + struct pipe_context *context, + uint unit, + struct pipe_texture *texture); + + +extern void +sct_update_texture(struct pipe_texture *tex); + + +extern boolean +sct_is_texture_used(struct surface_context_tracker *sct, + const struct pipe_context *context, + const struct pipe_texture *texture); + +extern void +sct_flush_textures(struct surface_context_tracker *sct, + struct pipe_context *context); + + +extern const struct sct_context_list * +sct_get_surface_contexts(struct surface_context_tracker *sct, + const struct pipe_surface *surf); + + +extern void +sct_destroy_context(struct surface_context_tracker *sct, + struct pipe_context *context); + + +extern void +sct_destroy_surface(struct surface_context_tracker *sct, + struct pipe_surface *surface); + + + +#ifdef __cplusplus +} +#endif + +#endif /* SCT_H */ diff --git a/src/gallium/auxiliary/sct/usage.c b/src/gallium/auxiliary/sct/usage.c new file mode 100644 index 0000000000..6227f19962 --- /dev/null +++ b/src/gallium/auxiliary/sct/usage.c @@ -0,0 +1,61 @@ +/* surface / context tracking */ + + +/* + +context A: + render to texture T + +context B: + texture from T + +----------------------- + +flush surface: + which contexts are bound to the surface? + +----------------------- + +glTexSubImage(): + which contexts need to be flushed? + + */ + + +/* + +in MakeCurrent(): + + call sct_bind_surfaces(context, list of surfaces) to update the + dependencies between context and surfaces + + +in SurfaceFlush(), or whatever it is in D3D: + + call sct_get_surface_contexts(surface) to get a list of contexts + which are currently bound to the surface. + + + +in BindTexture(): + + call sct_bind_texture(context, texture) to indicate that the texture + is used in the scene. + + +in glTexSubImage() or RenderToTexture(): + + call sct_is_texture_used(context, texture) to determine if the texture + has been used in the scene, but the scene's not flushed. If TRUE is + returned it means the scene has to be rendered/flushed before the contents + of the texture can be changed. + + +in psb_scene_flush/terminate(): + + call sct_flush_textures(context) to tell the SCT that the textures which + were used in the scene can be released. + + + +*/ -- cgit v1.2.3 From 6d5ee6d9a4a705cce80117f90ee334986f5e5e26 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 10 Mar 2008 19:37:32 +0000 Subject: gallium: enable bug workaround in draw_vertex_cache_invalidate --- src/gallium/auxiliary/draw/draw_vertex_cache.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vertex_cache.c b/src/gallium/auxiliary/draw/draw_vertex_cache.c index 53f8bbec44..161b247d4e 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_cache.c +++ b/src/gallium/auxiliary/draw/draw_vertex_cache.c @@ -41,7 +41,11 @@ void draw_vertex_cache_invalidate( struct draw_context *draw ) assert(draw->vs.queue_nr == 0); assert(draw->vcache.referenced == 0); -// memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); + /* There's an error somewhere in the vcache code that requires this + * memset. The bug is exposed in q3demo demo001, but probably + * elsewhere as well. Will track it down later. + */ + memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); } -- cgit v1.2.3 From 3f5b9f4ba49df57e7bbab04eab55a17a99bb5046 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 10 Mar 2008 19:41:12 +0000 Subject: gallium: use the same bypass_clipping logic on all vs paths --- src/gallium/auxiliary/draw/draw_vs_llvm.c | 48 +++++++++++++++++++------------ src/gallium/auxiliary/draw/draw_vs_sse.c | 38 +++++++++++++++--------- 2 files changed, 53 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 0fd557d667..53c260be53 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -135,25 +135,35 @@ vs_llvm_run( struct draw_vertex_shader *base, unsigned slot; float x, y, z, w; - x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; + if (!draw->rasterizer->bypass_clipping) { + x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + } + else { + vOut[j]->clipmask = 0; + vOut[j]->edgeflag = 1; + vOut[j]->data[0][0] = x; + vOut[j]->data[0][1] = y; + vOut[j]->data[0][2] = z; + vOut[j]->data[0][3] = w; + } /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 0b8bc2bf14..e5c1a40cca 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -158,20 +158,30 @@ vs_sse_run( struct draw_vertex_shader *base, z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; + if (!draw->rasterizer->bypass_clipping) { + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + vOut[j]->edgeflag = 1; + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + + /* Viewport mapping */ + vOut[j]->data[0][0] = x * scale[0] + trans[0]; + vOut[j]->data[0][1] = y * scale[1] + trans[1]; + vOut[j]->data[0][2] = z * scale[2] + trans[2]; + vOut[j]->data[0][3] = w; + } + else { + vOut[j]->clipmask = 0; + vOut[j]->edgeflag = 1; + vOut[j]->data[0][0] = x; + vOut[j]->data[0][1] = y; + vOut[j]->data[0][2] = z; + vOut[j]->data[0][3] = w; + } /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. -- cgit v1.2.3 From 7375369fb32e203023fbacf948169aad3f4c3a1d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 10 Mar 2008 19:41:51 +0000 Subject: gallium: fix compiler warning --- src/gallium/auxiliary/draw/draw_pstipple.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index 894b136f2c..b3e52dc1c7 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -482,11 +482,10 @@ static void pstip_first_tri(struct draw_stage *stage, struct prim_header *header) { struct pstip_stage *pstip = pstip_stage(stage); - struct draw_context *draw = stage->draw; struct pipe_context *pipe = pstip->pipe; uint num = MAX2(pstip->num_textures, pstip->num_samplers); - assert(draw->rasterizer->poly_stipple_enable); + assert(stage->draw->rasterizer->poly_stipple_enable); /* * Bind our fragprog, sampler and texture -- cgit v1.2.3 From d08cd68d3f5456b69ad504ede2b090c0bb6474db Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 10 Mar 2008 19:44:54 +0000 Subject: mesa: fast-track glColor and similar calls when not immediate mode rendering Often these are mixed in with draw arrays calls, etc. Try not to get the whole immediate rendering state machine going when we receive one of these on their own. --- src/mesa/vbo/vbo_exec_api.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index b7f4d8a307..35dc0e768f 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -309,16 +309,23 @@ static void vbo_exec_fixup_vertex( GLcontext *ctx, GLuint attr, GLuint sz ) { struct vbo_exec_context *exec = &vbo_context(ctx)->exec; + static const GLfloat id[4] = { 0, 0, 0, 1 }; int i; - if (sz > exec->vtx.attrsz[attr]) { + if (exec->vtx.prim_count == 0) { + GLfloat *current = (GLfloat *)vbo_context(ctx)->currval[attr].Ptr; + exec->vtx.attrptr[attr] = current; + memcpy(current, id, sizeof(id)); + ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT; + return; + } + else if (sz > exec->vtx.attrsz[attr]) { /* New size is larger. Need to flush existing vertices and get * an enlarged vertex format. */ vbo_exec_wrap_upgrade_vertex( exec, attr, sz ); } else if (sz < exec->vtx.active_sz[attr]) { - static const GLfloat id[4] = { 0, 0, 0, 1 }; /* New size is smaller - just need to fill in some * zeros. Don't need to flush or wrap. -- cgit v1.2.3 From 297b3be25a7f097fb9b1a79e332acddc12dcc3fe Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 10 Mar 2008 19:49:15 +0000 Subject: draw: placeholder/prototype code for a passthrough draw path --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/SConscript | 1 + src/gallium/auxiliary/draw/draw_context.c | 12 ++++++++++++ src/gallium/auxiliary/draw/draw_context.h | 5 +++++ src/gallium/auxiliary/draw/draw_private.h | 27 +++++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vbuf.h | 6 ++++++ 6 files changed, 52 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 2daa1636f3..ce6667d8ec 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -4,6 +4,7 @@ include $(TOP)/configs/current LIBNAME = draw C_SOURCES = \ + draw_passthrough.c \ draw_aaline.c \ draw_aapoint.c \ draw_clip.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index c18dcb2927..5cb7664c85 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -16,6 +16,7 @@ draw = env.ConvenienceLibrary( 'draw_offset.c', 'draw_prim.c', 'draw_pstipple.c', + 'draw_passthrough.c', 'draw_stipple.c', 'draw_twoside.c', 'draw_unfilled.c', diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 428b6209e0..bb64b50a17 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -34,6 +34,7 @@ #include "pipe/p_util.h" #include "draw_context.h" #include "draw_private.h" +#include "draw_vbuf.h" @@ -114,6 +115,10 @@ void draw_destroy( struct draw_context *draw ) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); tgsi_exec_machine_free_data(&draw->machine); align_free( draw->vs.queue[0].vertex ); /* Frees all the vertices. */ + + if (draw->render) + draw->render->destroy( draw->render ); + FREE( draw ); } @@ -349,3 +354,10 @@ void draw_reset_vertex_ids(struct draw_context *draw) draw_vertex_cache_reset_vertex_ids(draw); } + + +void draw_set_render( struct draw_context *draw, + struct vbuf_render *render ) +{ + draw->render = render; +} diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index ab87b4127c..df63e91a22 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -168,4 +168,9 @@ unsigned draw_trim_prim( unsigned mode, unsigned count ); + +struct vbuf_render; +void draw_set_render( struct draw_context *draw, + struct vbuf_render *render ); + #endif /* DRAW_CONTEXT_H */ diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index c732d723a7..25fa8c09c2 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -162,8 +162,14 @@ typedef void (*full_fetch_func)( struct draw_context *draw, const unsigned *elts, unsigned count ); +typedef void (*pt_fetch_func)( struct draw_context *draw, + float *out, + unsigned start, + unsigned count ); +struct vbuf_render; + /** * Private context for the drawing module. */ @@ -191,6 +197,17 @@ struct draw_context struct draw_stage *rasterize; } pipeline; + + struct vbuf_render *render; + + /* Support prototype passthrough path: + */ + struct { + unsigned prim; + unsigned hw_vertex_size; + } pt; + + /* pipe state that we need: */ const struct pipe_rasterizer_state *rasterizer; struct pipe_viewport_state viewport; @@ -244,6 +261,7 @@ struct draw_context fetch_func fetch[PIPE_ATTRIB_MAX]; unsigned nr_attrs; full_fetch_func fetch_func; + pt_fetch_func pt_fetch; } vertex_fetch; /* Post-tnl vertex cache: @@ -331,6 +349,15 @@ struct tgsi_exec_machine; extern void draw_update_vertex_fetch( struct draw_context *draw ); +/* Prototype/hack + */ +boolean +draw_passthrough_arrays(struct draw_context *draw, + unsigned prim, + unsigned start, + unsigned count); + + #define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ #define DRAW_FLUSH_PRIM_QUEUE 0x2 #define DRAW_FLUSH_VERTEX_CACHE 0x4 diff --git a/src/gallium/auxiliary/draw/draw_vbuf.h b/src/gallium/auxiliary/draw/draw_vbuf.h index cfd2b9820c..5e7de905c1 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.h +++ b/src/gallium/auxiliary/draw/draw_vbuf.h @@ -85,6 +85,12 @@ struct vbuf_render { const ushort *indices, uint nr_indices ); + /* Draw Arrays path too. + */ + void (*draw_arrays)( struct vbuf_render *, + unsigned start, + uint nr ); + /** * Called when vbuf is done with this set of vertices: */ -- cgit v1.2.3 From b1525662b330ca8b4cdd930775f3642bfec3b58f Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 10 Mar 2008 16:28:54 -0700 Subject: Move SPE register allocator to rtasm code Move the register allocator to a common location. There is more code on the way that will make use of this interface. --- src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c | 47 +++++++++++ src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h | 16 ++++ src/gallium/drivers/cell/ppu/cell_vertex_fetch.c | 101 +++++++---------------- 3 files changed, 92 insertions(+), 72 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c index 95a2d6fcbb..a996218ce7 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c @@ -306,6 +306,11 @@ void spe_init_func(struct spe_function *p, unsigned code_size) { p->store = align_malloc(code_size, 16); p->csr = p->store; + + /* Conservatively treat R0 - R2 and R80 - R127 as non-volatile. + */ + p->regs[0] = ~7; + p->regs[1] = (1U << (80 - 64)) - 1; } @@ -317,6 +322,48 @@ void spe_release_func(struct spe_function *p) } +int spe_allocate_available_register(struct spe_function *p) +{ + unsigned i; + for (i = 0; i < 128; i++) { + const uint64_t mask = (1ULL << (i % 128)); + const unsigned idx = i / 128; + + if ((p->regs[idx] & mask) != 0) { + p->regs[idx] &= ~mask; + return i; + } + } + + return -1; +} + + +int spe_allocate_register(struct spe_function *p, int reg) +{ + const unsigned idx = reg / 128; + const unsigned bit = reg % 128; + + assert((p->regs[idx] & (1ULL << bit)) != 0); + + p->regs[idx] &= ~(1ULL << bit); + return reg; +} + + +void spe_release_register(struct spe_function *p, int reg) +{ + const unsigned idx = reg / 128; + const unsigned bit = reg % 128; + + assert((p->regs[idx] & (1ULL << bit)) == 0); + + p->regs[idx] |= (1ULL << bit); +} + + + + void spe_bi(struct spe_function *p, unsigned rA, int d, int e) { emit_RI7(p, 0x1a8, 0, rA, (d << 5) | (e << 4)); diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h index 10ce44b3a0..5a1eb1ed8d 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h @@ -39,11 +39,27 @@ struct spe_function { uint32_t *store; uint32_t *csr; const char *fn; + + /** + * Mask of used / unused registers + * + * Each set bit corresponds to an available register. Each cleared bit + * corresponds to an allocated register. + * + * \sa + * spe_allocate_register, spe_allocate_available_register, + * spe_release_register + */ + uint64_t regs[2]; }; extern void spe_init_func(struct spe_function *p, unsigned code_size); extern void spe_release_func(struct spe_function *p); +extern int spe_allocate_available_register(struct spe_function *p); +extern int spe_allocate_register(struct spe_function *p, int reg); +extern void spe_release_register(struct spe_function *p, int reg); + #endif /* RTASM_PPC_SPE_H */ #ifndef EMIT_ diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c index 9cf74bab47..4828a8023b 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c @@ -33,46 +33,11 @@ #include "cell_context.h" #include "rtasm/rtasm_ppc_spe.h" -typedef uint64_t register_mask; - -int allocate_available_register(register_mask *m) -{ - unsigned i; - for (i = 0; i < 64; i++) { - const uint64_t mask = (1ULL << i); - - if ((m[0] & mask) != 0) { - m[0] &= ~mask; - return i; - } - } - - return -1; -} - - -int allocate_register(register_mask *m, unsigned reg) -{ - assert((m[0] & (1ULL << reg)) != 0); - - m[0] &= ~(1ULL << reg); - return reg; -} - - -void release_register(register_mask *m, unsigned reg) -{ - assert((m[0] & (1ULL << reg)) == 0); - - m[0] |= (1ULL << reg); -} - /** * Emit a 4x4 matrix transpose operation * * \param p Function that the transpose operation is to be appended to - * \param m Live register mask * \param row0 Register containing row 0 of the source matrix * \param row1 Register containing row 1 of the source matrix * \param row2 Register containing row 2 of the source matrix @@ -91,15 +56,15 @@ void release_register(register_mask *m, unsigned reg) * This function requires that four temporary are available on entry. */ static void -emit_matrix_transpose(struct spe_function *p, register_mask *m, +emit_matrix_transpose(struct spe_function *p, unsigned row0, unsigned row1, unsigned row2, unsigned row3, unsigned dest_ptr, unsigned shuf_ptr, unsigned count) { - int shuf_hi = allocate_available_register(m); - int shuf_lo = allocate_available_register(m); - int t1 = allocate_available_register(m); - int t2 = allocate_available_register(m); + int shuf_hi = spe_allocate_available_register(p); + int shuf_lo = spe_allocate_available_register(p); + int t1 = spe_allocate_available_register(p); + int t2 = spe_allocate_available_register(p); int t3; int t4; int col0; @@ -169,19 +134,19 @@ emit_matrix_transpose(struct spe_function *p, register_mask *m, /* Release all of the temporary registers used. */ - release_register(m, col0); - release_register(m, col1); - release_register(m, col2); - release_register(m, col3); - release_register(m, shuf_hi); - release_register(m, shuf_lo); - release_register(m, t2); - release_register(m, t4); + spe_release_register(p, col0); + spe_release_register(p, col1); + spe_release_register(p, col2); + spe_release_register(p, col3); + spe_release_register(p, shuf_hi); + spe_release_register(p, shuf_lo); + spe_release_register(p, t2); + spe_release_register(p, t4); } static void -emit_fetch(struct spe_function *p, register_mask *m, +emit_fetch(struct spe_function *p, unsigned in_ptr, unsigned *offset, unsigned out_ptr, unsigned shuf_ptr, enum pipe_format format) @@ -191,11 +156,11 @@ emit_fetch(struct spe_function *p, register_mask *m, const unsigned type = pf_type(format); const unsigned bytes = pf_size_x(format); - int v0 = allocate_available_register(m); - int v1 = allocate_available_register(m); - int v2 = allocate_available_register(m); - int v3 = allocate_available_register(m); - int tmp = allocate_available_register(m); + int v0 = spe_allocate_available_register(p); + int v1 = spe_allocate_available_register(p); + int v2 = spe_allocate_available_register(p); + int v3 = spe_allocate_available_register(p); + int tmp = spe_allocate_available_register(p); int float_zero = -1; int float_one = -1; float scale_signed = 0.0; @@ -260,19 +225,19 @@ emit_fetch(struct spe_function *p, register_mask *m, if (count < 4) { - float_one = allocate_available_register(m); + float_one = spe_allocate_available_register(p); spe_il(p, float_one, 1); spe_cuflt(p, float_one, float_one, 0); if (count < 3) { - float_zero = allocate_available_register(m); + float_zero = spe_allocate_available_register(p); spe_il(p, float_zero, 0); } } - release_register(m, tmp); + spe_release_register(p, tmp); - emit_matrix_transpose(p, m, v0, v1, v2, v3, out_ptr, shuf_ptr, count); + emit_matrix_transpose(p, v0, v1, v2, v3, out_ptr, shuf_ptr, count); switch (count) { case 1: @@ -284,11 +249,11 @@ emit_fetch(struct spe_function *p, register_mask *m, } if (float_zero != -1) { - release_register(m, float_zero); + spe_release_register(p, float_zero); } if (float_one != -1) { - release_register(m, float_one); + spe_release_register(p, float_one); } } @@ -297,7 +262,6 @@ void cell_update_vertex_fetch(struct draw_context *draw) { struct cell_context *const cell = (struct cell_context *) draw->driver_private; - register_mask m = ~0; struct spe_function *p = &cell->attrib_fetch; unsigned function_index[PIPE_ATTRIB_MAX]; unsigned unique_attr_formats; @@ -338,18 +302,11 @@ void cell_update_vertex_fetch(struct draw_context *draw) spe_init_func(p, 136 * unique_attr_formats); - /* Registers 0, 1, and 2 are reserved by the ABI. - */ - allocate_register(&m, 0); - allocate_register(&m, 1); - allocate_register(&m, 2); - - /* Allocate registers for the function's input parameters. */ - out_ptr = allocate_register(&m, 3); - in_ptr = allocate_register(&m, 4); - shuf_ptr = allocate_register(&m, 5); + out_ptr = spe_allocate_register(p, 3); + in_ptr = spe_allocate_register(p, 4); + shuf_ptr = spe_allocate_register(p, 5); /* Generate code for the individual attribute fetch functions. @@ -362,7 +319,7 @@ void cell_update_vertex_fetch(struct draw_context *draw) - (void *) p->store); offset = 0; - emit_fetch(p, & m, in_ptr, &offset, out_ptr, shuf_ptr, + emit_fetch(p, in_ptr, &offset, out_ptr, shuf_ptr, draw->vertex_element[i].src_format); spe_bi(p, 0, 0, 0); -- cgit v1.2.3 From 7585b4ceb8fed862c07f50af8030a6f0eb8a8321 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 4 Mar 2008 08:38:54 -0700 Subject: gallium: disable an unneeded assertion (hit with texwrap.c) --- src/mesa/state_tracker/st_cb_texture.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 1ba3173312..d4731c7737 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -529,7 +529,8 @@ st_TexImage(GLcontext * ctx, texImage->RowStride = postConvWidth; } - assert(texImage->RowStride == postConvWidth); + /* we'll set RowStride elsewhere when the texture is a "mapped" state */ + /*assert(texImage->RowStride == postConvWidth);*/ } /* Release the reference to a potentially orphaned buffer. -- cgit v1.2.3 From c813b545ab4726fc5030f123ec6255224d64ad82 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 10 Mar 2008 17:41:00 -0600 Subject: fix Height2/Depth2 init problem when using texture borders --- src/mesa/main/teximage.c | 23 +++++++++++++++++------ 1 file changed, 17 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c index 5c96be9216..f15e404527 100644 --- a/src/mesa/main/teximage.c +++ b/src/mesa/main/teximage.c @@ -1214,19 +1214,30 @@ _mesa_init_teximage_fields(GLcontext *ctx, GLenum target, img->Width = width; img->Height = height; img->Depth = depth; + img->Width2 = width - 2 * border; /* == 1 << img->WidthLog2; */ - img->Height2 = height - 2 * border; /* == 1 << img->HeightLog2; */ - img->Depth2 = depth - 2 * border; /* == 1 << img->DepthLog2; */ img->WidthLog2 = logbase2(img->Width2); - if (height == 1) /* 1-D texture */ + + if (height == 1) { /* 1-D texture */ + img->Height2 = 1; img->HeightLog2 = 0; - else + } + else { + img->Height2 = height - 2 * border; /* == 1 << img->HeightLog2; */ img->HeightLog2 = logbase2(img->Height2); - if (depth == 1) /* 2-D texture */ + } + + if (depth == 1) { /* 2-D texture */ + img->Depth2 = 1; img->DepthLog2 = 0; - else + } + else { + img->Depth2 = depth - 2 * border; /* == 1 << img->DepthLog2; */ img->DepthLog2 = logbase2(img->Depth2); + } + img->MaxLog2 = MAX2(img->WidthLog2, img->HeightLog2); + img->IsCompressed = GL_FALSE; img->CompressedSize = 0; -- cgit v1.2.3 From d9d2ca7a07469a7d5cdc183f2daa6cf9e30938fe Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 10 Mar 2008 22:10:07 -0400 Subject: fix compilation --- src/gallium/auxiliary/draw/Makefile | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index ce6667d8ec..2daa1636f3 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -4,7 +4,6 @@ include $(TOP)/configs/current LIBNAME = draw C_SOURCES = \ - draw_passthrough.c \ draw_aaline.c \ draw_aapoint.c \ draw_clip.c \ -- cgit v1.2.3 From be9a2457388d99a2185f258aeb5ef5183ccfbbb2 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 10 Mar 2008 22:10:18 -0400 Subject: fix double deletion plus, if the current hash is bigger than max size make sure we delete enough from it --- src/gallium/auxiliary/cso_cache/cso_cache.c | 5 ++++- src/gallium/auxiliary/cso_cache/cso_hash.c | 7 ------- 2 files changed, 4 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index b427b509f8..a2764b4265 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -207,8 +207,11 @@ static INLINE void sanitize_hash(struct cso_hash *hash, enum cso_cache_type type { /* if we're approach the maximum size, remove fourth of the entries * otherwise every subsequent call will go through the same */ - int max_entries = (max_size > cso_hash_size(hash)) ? max_size : cso_hash_size(hash); + int hash_size = cso_hash_size(hash); + int max_entries = (max_size > hash_size) ? max_size : hash_size; int to_remove = (max_size < max_entries) * max_entries/4; + if (hash_size > max_size) + to_remove += hash_size - max_size; while (to_remove) { /*remove elements until we're good */ /*fixme: currently we pick the nodes to remove at random*/ diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index b3b4d667d2..5cad5d3be7 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -101,13 +101,6 @@ static void *cso_data_allocate_node(struct cso_hash_data *hash) static void cso_data_free_node(struct cso_node *node) { - /* XXX still a leak here. - * Need to cast value ptr to original cso type, then free the - * driver-specific data hanging off of it. For example: - struct cso_sampler *csamp = (struct cso_sampler *) node->value; - FREE(csamp->data); - */ - FREE(node->value); FREE(node); } -- cgit v1.2.3 From ff3c7a3243e4f3fc60e6cfcfc6a2711e9ea5cf65 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 11 Mar 2008 08:42:49 +0000 Subject: gallium: missing file --- src/gallium/auxiliary/draw/draw_passthrough.c | 222 ++++++++++++++++++++++++++ 1 file changed, 222 insertions(+) create mode 100644 src/gallium/auxiliary/draw/draw_passthrough.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_passthrough.c b/src/gallium/auxiliary/draw/draw_passthrough.c new file mode 100644 index 0000000000..fc2dde38ba --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_passthrough.c @@ -0,0 +1,222 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +/* This code is a prototype of what a passhthrough vertex shader might + * look like. + * + * Probably the best approach for us is to do: + * - vertex fetch + * - vertex shader + * - cliptest / viewport transform + * + * in one step, then examine the clipOrMask & choose between two paths: + * + * Either: + * - build primitive headers + * - clip and the primitive path + * - build clipped vertex buffers, + * - vertex-emit to vbuf buffers + * + * Or, if no clipping: + * - vertex-emit directly to vbuf buffers + * + * But when bypass clipping is enabled, we just take the latter + * choice. If (some new) passthrough-vertex-shader flag is also set, + * the pipeline degenerates to: + * + * - vertex fetch + * - vertex emit to vbuf buffers + * + * Which is what is prototyped here. + */ +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" + + + +/* Example of a fetch/emit passthrough shader which could be + * generated when bypass_clipping is enabled on a passthrough vertex + * shader. + */ +static void fetch_xyz_rgb_st( struct draw_context *draw, + float *out, + unsigned start, + unsigned count ) +{ + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + int i; + + const ubyte *xyzw = src[0] + start * pitch[0]; + const ubyte *rgba = src[1] + start * pitch[1]; + const ubyte *st = src[2] + start * pitch[2]; + + /* loop over vertex attributes (vertex shader inputs) + */ + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyzw; xyzw += pitch[0]; + /* decode input, encode output. Assume both are float[4] */ + out[0] = in[0]; + out[1] = in[1]; + out[2] = in[2]; + out[3] = in[3]; + } + + { + const float *in = (const float *)rgba; rgba += pitch[1]; + /* decode input, encode output. Assume both are float[4] */ + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = in[3]; + } + + { + const float *in = (const float *)st; st += pitch[2]; + /* decode input, encode output. Assume both are float[2] */ + out[8] = in[0]; + out[9] = in[1]; + } + + out += 10; + } +} + + +static boolean update_shader( struct draw_context *draw ) +{ + const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); + + unsigned nr_attrs = vinfo->num_attribs; + unsigned i; + + for (i = 0; i < nr_attrs; i++) { + unsigned buf = draw->vertex_element[i].vertex_buffer_index; + + draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[i].src_offset; + + draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; + draw->vertex_fetch.fetch[i] = NULL; + } + + draw->vertex_fetch.nr_attrs = nr_attrs; + draw->vertex_fetch.fetch_func = NULL; + draw->vertex_fetch.pt_fetch = NULL; + + draw->pt.hw_vertex_size = vinfo->size * 4; + + /* Just trying to figure out how this would work: + */ + if (nr_attrs == 3 && + 0 /* some other tests */) + { + draw->vertex_fetch.pt_fetch = fetch_xyz_rgb_st; + assert(vinfo->size == 10); + return TRUE; + } + + return FALSE; +} + + + +static boolean set_prim( struct draw_context *draw, + unsigned prim ) +{ + assert(!draw->user.elts); + + draw->pt.prim = prim; + + switch (prim) { + case PIPE_PRIM_LINE_LOOP: + case PIPE_PRIM_QUADS: + case PIPE_PRIM_QUAD_STRIP: + return FALSE; + default: + draw->render->set_primitive( draw->render, prim ); + return TRUE; + } +} + + + +boolean +draw_passthrough_arrays(struct draw_context *draw, + unsigned prim, + unsigned start, + unsigned count) +{ + float *hw_verts; + + if (!set_prim(draw, prim)) + return FALSE; + + if (!update_shader( draw )) + return FALSE; + + hw_verts = draw->render->allocate_vertices( draw->render, + draw->pt.hw_vertex_size, + count ); + + if (!hw_verts) + return FALSE; + + /* Single routine to fetch vertices, run shader and emit HW verts. + * Clipping and viewport transformation are done on hardware. + */ + draw->vertex_fetch.pt_fetch( draw, + hw_verts, + start, count ); + + /* Draw arrays path to avoid re-emitting index list again and + * again. + */ + draw->render->draw_arrays( draw->render, + start, + count ); + + + draw->render->release_vertices( draw->render, + hw_verts, + draw->pt.hw_vertex_size, + count ); + + return TRUE; +} + -- cgit v1.2.3 From 5038c20795cb2e49d72c1f43a8b705056592356c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 11 Mar 2008 14:23:08 +0000 Subject: draw: don't free our copy of the render stage -- just borrowing it from vbuf stage --- src/gallium/auxiliary/draw/draw_context.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index bb64b50a17..fed2b6e759 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -116,8 +116,11 @@ void draw_destroy( struct draw_context *draw ) tgsi_exec_machine_free_data(&draw->machine); align_free( draw->vs.queue[0].vertex ); /* Frees all the vertices. */ + /* Not so fast -- we're just borrowing this at the moment. + * if (draw->render) draw->render->destroy( draw->render ); + */ FREE( draw ); } -- cgit v1.2.3 From 30fab81de8ea7bf81181db7bd605f376d4e4fca2 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Mar 2008 14:31:17 -0600 Subject: gallium: fix fs/vs typo in cso_set_vertex_shader() --- src/gallium/auxiliary/cso_cache/cso_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index fbb26ca511..f7f4aebb16 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -349,6 +349,6 @@ void cso_set_vertex_shader(struct cso_context *ctx, if (ctx->vertex_shader != handle) { ctx->vertex_shader = handle; - ctx->pipe->bind_fs_state(ctx->pipe, handle); + ctx->pipe->bind_vs_state(ctx->pipe, handle); } } -- cgit v1.2.3 From 34a0ac7f2b93d6d0f3fc85106e7dacb38c4229b6 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Mar 2008 15:01:52 -0600 Subject: gallium: fix some cso_state_callback cast warnings --- src/gallium/auxiliary/cso_cache/cso_cache.h | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h index 44ee128a4a..e5edbbb556 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.h +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -84,47 +84,49 @@ extern "C" { #endif +typedef void (*cso_state_callback)(void *ctx, void *obj); + struct cso_cache; struct cso_blend { struct pipe_blend_state state; void *data; - void (*delete_state)(struct pipe_context *, void *); + cso_state_callback delete_state; struct pipe_context *context; }; struct cso_depth_stencil_alpha { struct pipe_depth_stencil_alpha_state state; void *data; - void (*delete_state)(struct pipe_context *, void *); + cso_state_callback delete_state; struct pipe_context *context; }; struct cso_rasterizer { struct pipe_rasterizer_state state; void *data; - void (*delete_state)(struct pipe_context *, void *); + cso_state_callback delete_state; struct pipe_context *context; }; struct cso_fragment_shader { struct pipe_shader_state state; void *data; - void (*delete_state)(struct pipe_context *, void *); + cso_state_callback delete_state; struct pipe_context *context; }; struct cso_vertex_shader { struct pipe_shader_state state; void *data; - void (*delete_state)(struct pipe_context *, void *); + cso_state_callback delete_state; struct pipe_context *context; }; struct cso_sampler { struct pipe_sampler_state state; void *data; - void (*delete_state)(struct pipe_context *, void *); + cso_state_callback delete_state; struct pipe_context *context; }; @@ -138,8 +140,6 @@ enum cso_cache_type { CSO_VERTEX_SHADER }; -typedef void (*cso_state_callback)(void *, void *); - unsigned cso_construct_key(void *item, int item_size); struct cso_cache *cso_cache_create(void); -- cgit v1.2.3 From 130b3154544701be2a57ac1c57432f153e363572 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 11 Mar 2008 23:21:36 +0100 Subject: tgsi: Map OPCODE_TEXKILL to OPCODE_KILP. --- src/gallium/include/pipe/p_shader_tokens.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index b110f01291..e4d89971f9 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -341,7 +341,7 @@ struct tgsi_immediate_float32 * ps_1_1 */ #define TGSI_OPCODE_TEXCOORD TGSI_OPCODE_NOP -#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL +#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KILP #define TGSI_OPCODE_TEXBEM 107 #define TGSI_OPCODE_TEXBEML 108 #define TGSI_OPCODE_TEXREG2AR 109 -- cgit v1.2.3 From 45c59895113f997e5f2b7e346f95e46099fa3566 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 11 Mar 2008 12:03:11 +0000 Subject: gallium: Conditional debugging output. Generalize the conditional debugging output code found trhought the gallium drivers. --- src/gallium/auxiliary/util/p_debug.c | 34 +++++++++++++++++++ src/gallium/include/pipe/p_debug.h | 64 ++++++++++++++++++++++++++++++++++++ 2 files changed, 98 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 93bfaea393..04e55dd91d 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -103,3 +103,37 @@ void debug_assert_fail(const char *expr, const char *file, unsigned line) debug_printf("%s:%i: Assertion `%s' failed.\n", file, line, expr); debug_break(); } + + +#define DEBUG_MASK_TABLE_SIZE 256 + + +/** + * Mask hash table. + * + * For now we just take the lower bits of the key, and do no attempt to solve + * collisions. Use a proper hash table when we have dozens of drivers. + */ +static uint32_t debug_mask_table[DEBUG_MASK_TABLE_SIZE]; + + +void debug_mask_set(uint32_t uuid, uint32_t mask) +{ + unsigned hash = uuid & (DEBUG_MASK_TABLE_SIZE - 1); + debug_mask_table[hash] = mask; +} + + +uint32_t debug_mask_get(uint32_t uuid) +{ + unsigned hash = uuid & (DEBUG_MASK_TABLE_SIZE - 1); + return debug_mask_table[hash]; +} + + +void debug_mask_vprintf(uint32_t uuid, uint32_t what, const char *format, va_list ap) +{ + uint32_t mask = debug_mask_get(uuid); + if(mask & what) + debug_vprintf(format, ap); +} diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index a14a1fc5f6..f45363f355 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -41,6 +41,8 @@ #include +#include "p_compiler.h" + #ifdef __cplusplus extern "C" { @@ -93,6 +95,68 @@ void debug_assert_fail(const char *expr, const char *file, unsigned line); #define assert(expr) debug_assert(expr) +/** + * Set a channel's debug mask. + * + * uuid is just a random 32 bit integer that uniquely identifies the debugging + * channel. + * + * @note Due to current implementation issues, make sure the lower 8 bits of + * UUID are unique. + */ +void debug_mask_set(uint32_t uuid, uint32_t mask); + + +uint32_t debug_mask_get(uint32_t uuid); + + +/** + * Conditional debug output. + * + * This is just a generalization of the debug filtering mechanism used + * throughout Gallium. + * + * You use this function as: + * + * @code + * #define MYDRIVER_UUID 0x12345678 // random 32 bit identifier + * + * static inline mydriver_debug(uint32_t what, const char *format, ...) + * { + * #ifdef DEBUG + * va_list ap; + * va_start(ap, format); + * debug_mask_vprintf(MYDRIVER_UUID, what, format, ap); + * va_end(ap); + * #endif + * } + * + * ... + * + * debug_mask_set(MYDRIVER_UUID, + * MYDRIVER_DEBUG_THIS | + * MYDRIVER_DEBUG_THAT | + * ... ); + * + * ... + * + * mydriver_debug(MYDRIVER_DEBUG_THIS, + * "this and this happened\n"); + * + * mydriver_debug(MYDRIVER_DEBUG_THAT, + * "that = %f\n", that); + * ... + * @endcode + * + * You can also define several variants of mydriver_debug, with hardcoded what. + * Note that although macros with variable number of arguments would accomplish + * more in less code, they are not portable. + */ +void debug_mask_vprintf(uint32_t uuid, + uint32_t what, + const char *format, + va_list ap); + #ifdef __cplusplus } #endif -- cgit v1.2.3 From 21ff00306131cd5598f95285badaaabc98021e11 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 11 Mar 2008 23:51:27 +0000 Subject: gallium: Silence MSVC warnings. --- src/gallium/auxiliary/util/u_snprintf.c | 10 +++------- src/gallium/drivers/i915simple/i915_state.c | 7 +++++-- src/gallium/drivers/softpipe/sp_state_sampler.c | 8 +++++--- 3 files changed, 13 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_snprintf.c b/src/gallium/auxiliary/util/u_snprintf.c index 5e19a2ddb2..61c20b48f7 100644 --- a/src/gallium/auxiliary/util/u_snprintf.c +++ b/src/gallium/auxiliary/util/u_snprintf.c @@ -241,7 +241,6 @@ static void *mymemcpy(void *, void *, size_t); #endif /* !HAVE_VASPRINTF */ #if !HAVE_VSNPRINTF -#include /* For ERANGE and errno. */ #include /* For *_MAX. */ #if HAVE_INTTYPES_H #include /* For intmax_t (if not defined in ). */ @@ -445,8 +444,6 @@ static UINTMAX_T cast(LDOUBLE); static UINTMAX_T myround(LDOUBLE); static LDOUBLE mypow10(int); -extern int errno; - int rpl_vsnprintf(char *str, size_t size, const char *format, va_list args) { @@ -747,7 +744,7 @@ rpl_vsnprintf(char *str, size_t size, const char *format, va_list args) goto out; break; case 'c': - cvalue = va_arg(args, int); + cvalue = (unsigned char)va_arg(args, int); OUTCHAR(str, len, size, cvalue); break; case 's': @@ -844,7 +841,6 @@ out: str[size - 1] = '\0'; if (overflow || len >= INT_MAX) { - errno = overflow ? EOVERFLOW : ERANGE; return -1; } return (int)len; @@ -1106,7 +1102,7 @@ again: * Factor of ten with the number of digits needed for the fractional * part. For example, if the precision is 3, the mask will be 1000. */ - mask = mypow10(precision); + mask = (UINTMAX_T)mypow10(precision); /* * We "cheat" by converting the fractional part to integer by * multiplying by a factor of ten. @@ -1358,7 +1354,7 @@ cast(LDOUBLE value) if (value >= UINTMAX_MAX) return UINTMAX_MAX; - result = value; + result = (UINTMAX_T)value; /* * At least on NetBSD/sparc64 3.0.2 and 4.99.30, casting long double to * an integer type converts e.g. 1.9 to 2 instead of 1 (which violates diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 24143243d3..d9ab483bfc 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -273,6 +273,7 @@ static void i915_bind_sampler_states(struct pipe_context *pipe, unsigned num, void **sampler) { struct i915_context *i915 = i915_context(pipe); + unsigned i; assert(num <= PIPE_MAX_SAMPLERS); @@ -281,8 +282,10 @@ static void i915_bind_sampler_states(struct pipe_context *pipe, !memcmp(i915->sampler, sampler, num * sizeof(void *))) return; - memcpy(i915->sampler, sampler, num * sizeof(void *)); - memset(&i915->sampler[num], 0, (PIPE_MAX_SAMPLERS - num) * sizeof(void *)); + for (i = 0; i < num; ++i) + i915->sampler[i] = sampler[i]; + for (i = num; i < PIPE_MAX_SAMPLERS; ++i) + i915->sampler[i] = NULL; i915->num_samplers = num; diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c index 7cf85b9207..033288a0aa 100644 --- a/src/gallium/drivers/softpipe/sp_state_sampler.c +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -56,6 +56,7 @@ softpipe_bind_sampler_states(struct pipe_context *pipe, unsigned num, void **sampler) { struct softpipe_context *softpipe = softpipe_context(pipe); + unsigned i; assert(num <= PIPE_MAX_SAMPLERS); @@ -66,9 +67,10 @@ softpipe_bind_sampler_states(struct pipe_context *pipe, draw_flush(softpipe->draw); - memcpy(softpipe->sampler, sampler, num * sizeof(void *)); - memset(&softpipe->sampler[num], 0, (PIPE_MAX_SAMPLERS - num) * - sizeof(void *)); + for (i = 0; i < num; ++i) + softpipe->sampler[i] = sampler[i]; + for (i = num; i < PIPE_MAX_SAMPLERS; ++i) + softpipe->sampler[i] = NULL; softpipe->num_samplers = num; -- cgit v1.2.3 From 339e7ec6805e6de8794514c0a935081b5d36d38f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Mar 2008 18:54:31 -0600 Subject: gallium: rework CSO-related code in state tracker Use the code in cso_context.c rather than st_cache.c. Basically, binding of state objects now goes through the CSO module. But Vertex/fragment shaders go through pipe->bind_fs/vs_state() since they're not cached by the CSO module at this time. Also, update softpipe driver to handle NULL state objects in various places. This happens during context destruction. May need to update other drivers... --- src/gallium/auxiliary/draw/draw_aaline.c | 3 +- src/gallium/auxiliary/draw/draw_aapoint.c | 3 +- src/gallium/auxiliary/draw/draw_pstipple.c | 3 +- src/gallium/drivers/softpipe/sp_state_fs.c | 3 +- src/mesa/sources | 1 - src/mesa/state_tracker/st_atom_blend.c | 52 ++++++-------- src/mesa/state_tracker/st_atom_depth.c | 60 +++++++--------- src/mesa/state_tracker/st_atom_rasterizer.c | 108 ++++++++++++++-------------- src/mesa/state_tracker/st_atom_sampler.c | 43 ++++++----- src/mesa/state_tracker/st_atom_shader.c | 28 ++------ src/mesa/state_tracker/st_cb_accum.c | 1 - src/mesa/state_tracker/st_cb_clear.c | 47 ++++++------ src/mesa/state_tracker/st_cb_drawpixels.c | 64 +++++++++-------- src/mesa/state_tracker/st_cb_program.c | 5 +- src/mesa/state_tracker/st_context.c | 9 ++- src/mesa/state_tracker/st_context.h | 19 ++--- src/mesa/state_tracker/st_debug.c | 6 +- src/mesa/state_tracker/st_draw.c | 13 ++-- src/mesa/state_tracker/st_gen_mipmap.c | 34 ++++----- src/mesa/state_tracker/st_mesa_to_tgsi.c | 5 +- src/mesa/state_tracker/st_mesa_to_tgsi.h | 2 +- src/mesa/state_tracker/st_program.c | 59 +++++++++------ src/mesa/state_tracker/st_program.h | 22 ++---- 23 files changed, 292 insertions(+), 298 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index 3ec73b0800..6b1e640ae9 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -733,7 +733,8 @@ aaline_bind_fs_state(struct pipe_context *pipe, void *fs) /* save current */ aaline->fs = aafs; /* pass-through */ - aaline->driver_bind_fs_state(aaline->pipe, aafs->driver_fs); + aaline->driver_bind_fs_state(aaline->pipe, + (aafs ? aafs->driver_fs : NULL)); } diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index 70f696475f..99e9e9fe34 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -800,7 +800,8 @@ aapoint_bind_fs_state(struct pipe_context *pipe, void *fs) /* save current */ aapoint->fs = aafs; /* pass-through */ - aapoint->driver_bind_fs_state(aapoint->pipe, aafs->driver_fs); + aapoint->driver_bind_fs_state(aapoint->pipe, + (aafs ? aafs->driver_fs : NULL)); } diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index b3e52dc1c7..ed50d0805a 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -595,7 +595,8 @@ pstip_bind_fs_state(struct pipe_context *pipe, void *fs) /* save current */ pstip->fs = aafs; /* pass-through */ - pstip->driver_bind_fs_state(pstip->pipe, aafs->driver_fs); + pstip->driver_bind_fs_state(pstip->pipe, + (aafs ? aafs->driver_fs : NULL)); } diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index eb641ed321..4eefd1d61f 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -120,7 +120,8 @@ softpipe_bind_vs_state(struct pipe_context *pipe, void *vs) softpipe->vs = (const struct sp_vertex_shader *)vs; - draw_bind_vertex_shader(softpipe->draw, softpipe->vs->draw_data); + draw_bind_vertex_shader(softpipe->draw, + (softpipe->vs ? softpipe->vs->draw_data : NULL)); softpipe->dirty |= SP_NEW_VS; } diff --git a/src/mesa/sources b/src/mesa/sources index f0bf7b31fb..e3d5f22849 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -184,7 +184,6 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_readpixels.c \ state_tracker/st_cb_strings.c \ state_tracker/st_cb_texture.c \ - state_tracker/st_cache.c \ state_tracker/st_context.c \ state_tracker/st_debug.c \ state_tracker/st_draw.c \ diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index 2a9d209153..6c13fc8141 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -33,11 +33,11 @@ #include "st_context.h" -#include "st_cache.h" #include "st_atom.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "cso_cache/cso_context.h" /** @@ -155,44 +155,43 @@ translate_logicop(GLenum logicop) static void update_blend( struct st_context *st ) { - struct pipe_blend_state blend; - const struct cso_blend *cso; + struct pipe_blend_state *blend = &st->state.blend; - memset(&blend, 0, sizeof(blend)); + memset(blend, 0, sizeof(*blend)); if (st->ctx->Color.ColorLogicOpEnabled || (st->ctx->Color.BlendEnabled && st->ctx->Color.BlendEquationRGB == GL_LOGIC_OP)) { /* logicop enabled */ - blend.logicop_enable = 1; - blend.logicop_func = translate_logicop(st->ctx->Color.LogicOp); + blend->logicop_enable = 1; + blend->logicop_func = translate_logicop(st->ctx->Color.LogicOp); } else if (st->ctx->Color.BlendEnabled) { /* blending enabled */ - blend.blend_enable = 1; + blend->blend_enable = 1; - blend.rgb_func = translate_blend(st->ctx->Color.BlendEquationRGB); + blend->rgb_func = translate_blend(st->ctx->Color.BlendEquationRGB); if (st->ctx->Color.BlendEquationRGB == GL_MIN || st->ctx->Color.BlendEquationRGB == GL_MAX) { /* Min/max are special */ - blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; - blend.rgb_dst_factor = PIPE_BLENDFACTOR_ONE; + blend->rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend->rgb_dst_factor = PIPE_BLENDFACTOR_ONE; } else { - blend.rgb_src_factor = translate_blend(st->ctx->Color.BlendSrcRGB); - blend.rgb_dst_factor = translate_blend(st->ctx->Color.BlendDstRGB); + blend->rgb_src_factor = translate_blend(st->ctx->Color.BlendSrcRGB); + blend->rgb_dst_factor = translate_blend(st->ctx->Color.BlendDstRGB); } - blend.alpha_func = translate_blend(st->ctx->Color.BlendEquationA); + blend->alpha_func = translate_blend(st->ctx->Color.BlendEquationA); if (st->ctx->Color.BlendEquationA == GL_MIN || st->ctx->Color.BlendEquationA == GL_MAX) { /* Min/max are special */ - blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; - blend.alpha_dst_factor = PIPE_BLENDFACTOR_ONE; + blend->alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend->alpha_dst_factor = PIPE_BLENDFACTOR_ONE; } else { - blend.alpha_src_factor = translate_blend(st->ctx->Color.BlendSrcA); - blend.alpha_dst_factor = translate_blend(st->ctx->Color.BlendDstA); + blend->alpha_src_factor = translate_blend(st->ctx->Color.BlendSrcA); + blend->alpha_dst_factor = translate_blend(st->ctx->Color.BlendDstA); } } else { @@ -201,25 +200,18 @@ update_blend( struct st_context *st ) /* Colormask - maybe reverse these bits? */ if (st->ctx->Color.ColorMask[0]) - blend.colormask |= PIPE_MASK_R; + blend->colormask |= PIPE_MASK_R; if (st->ctx->Color.ColorMask[1]) - blend.colormask |= PIPE_MASK_G; + blend->colormask |= PIPE_MASK_G; if (st->ctx->Color.ColorMask[2]) - blend.colormask |= PIPE_MASK_B; + blend->colormask |= PIPE_MASK_B; if (st->ctx->Color.ColorMask[3]) - blend.colormask |= PIPE_MASK_A; + blend->colormask |= PIPE_MASK_A; if (st->ctx->Color.DitherFlag) - blend.dither = 1; + blend->dither = 1; - cso = st_cached_blend_state(st, &blend); - - if (st->state.blend != cso) { - /* state has changed */ - st->state.blend = cso; - /* bind new state */ - st->pipe->bind_blend_state(st->pipe, cso->data); - } + cso_set_blend(st->cso_context, blend); if (memcmp(st->ctx->Color.BlendColor, &st->state.blend_color, 4 * sizeof(GLfloat)) != 0) { /* state has changed */ diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index 7aecdbfbcc..827ad3b548 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -34,10 +34,10 @@ #include "st_context.h" -#include "st_cache.h" #include "st_atom.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "cso_cache/cso_context.h" /** @@ -93,53 +93,47 @@ gl_stencil_op_to_pipe(GLenum func) static void update_depth_stencil_alpha(struct st_context *st) { - struct pipe_depth_stencil_alpha_state depth_stencil; - const struct cso_depth_stencil_alpha *cso; + struct pipe_depth_stencil_alpha_state *dsa = &st->state.depth_stencil; - memset(&depth_stencil, 0, sizeof(depth_stencil)); + memset(dsa, 0, sizeof(*dsa)); - depth_stencil.depth.enabled = st->ctx->Depth.Test; - depth_stencil.depth.writemask = st->ctx->Depth.Mask; - depth_stencil.depth.func = st_compare_func_to_pipe(st->ctx->Depth.Func); + dsa->depth.enabled = st->ctx->Depth.Test; + dsa->depth.writemask = st->ctx->Depth.Mask; + dsa->depth.func = st_compare_func_to_pipe(st->ctx->Depth.Func); if (st->ctx->Query.CurrentOcclusionObject && st->ctx->Query.CurrentOcclusionObject->Active) - depth_stencil.depth.occlusion_count = 1; + dsa->depth.occlusion_count = 1; if (st->ctx->Stencil.Enabled) { - depth_stencil.stencil[0].enabled = 1; - depth_stencil.stencil[0].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[0]); - depth_stencil.stencil[0].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[0]); - depth_stencil.stencil[0].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[0]); - depth_stencil.stencil[0].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[0]); - depth_stencil.stencil[0].ref_value = st->ctx->Stencil.Ref[0] & 0xff; - depth_stencil.stencil[0].value_mask = st->ctx->Stencil.ValueMask[0] & 0xff; - depth_stencil.stencil[0].write_mask = st->ctx->Stencil.WriteMask[0] & 0xff; + dsa->stencil[0].enabled = 1; + dsa->stencil[0].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[0]); + dsa->stencil[0].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[0]); + dsa->stencil[0].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[0]); + dsa->stencil[0].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[0]); + dsa->stencil[0].ref_value = st->ctx->Stencil.Ref[0] & 0xff; + dsa->stencil[0].value_mask = st->ctx->Stencil.ValueMask[0] & 0xff; + dsa->stencil[0].write_mask = st->ctx->Stencil.WriteMask[0] & 0xff; if (st->ctx->Stencil.TestTwoSide) { - depth_stencil.stencil[1].enabled = 1; - depth_stencil.stencil[1].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[1]); - depth_stencil.stencil[1].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[1]); - depth_stencil.stencil[1].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[1]); - depth_stencil.stencil[1].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[1]); - depth_stencil.stencil[1].ref_value = st->ctx->Stencil.Ref[1] & 0xff; - depth_stencil.stencil[1].value_mask = st->ctx->Stencil.ValueMask[1] & 0xff; - depth_stencil.stencil[1].write_mask = st->ctx->Stencil.WriteMask[1] & 0xff; + dsa->stencil[1].enabled = 1; + dsa->stencil[1].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[1]); + dsa->stencil[1].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[1]); + dsa->stencil[1].zfail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZFailFunc[1]); + dsa->stencil[1].zpass_op = gl_stencil_op_to_pipe(st->ctx->Stencil.ZPassFunc[1]); + dsa->stencil[1].ref_value = st->ctx->Stencil.Ref[1] & 0xff; + dsa->stencil[1].value_mask = st->ctx->Stencil.ValueMask[1] & 0xff; + dsa->stencil[1].write_mask = st->ctx->Stencil.WriteMask[1] & 0xff; } } if (st->ctx->Color.AlphaEnabled) { - depth_stencil.alpha.enabled = 1; - depth_stencil.alpha.func = st_compare_func_to_pipe(st->ctx->Color.AlphaFunc); - depth_stencil.alpha.ref = st->ctx->Color.AlphaRef; + dsa->alpha.enabled = 1; + dsa->alpha.func = st_compare_func_to_pipe(st->ctx->Color.AlphaFunc); + dsa->alpha.ref = st->ctx->Color.AlphaRef; } - cso = st_cached_depth_stencil_alpha_state(st, &depth_stencil); - if (st->state.depth_stencil != cso) { - /* state has changed */ - st->state.depth_stencil = cso; - st->pipe->bind_depth_stencil_alpha_state(st->pipe, cso->data); /* bind new state */ - } + cso_set_depth_stencil_alpha(st->cso_context, dsa); } diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 229839d8b2..77cef9236b 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -32,10 +32,11 @@ #include "main/macros.h" #include "st_context.h" -#include "st_cache.h" +#include "st_atom.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "st_atom.h" +#include "cso_cache/cso_context.h" + static GLuint translate_fill( GLenum mode ) { @@ -72,22 +73,21 @@ static GLboolean get_offset_flag( GLuint fill_mode, static void update_raster_state( struct st_context *st ) { GLcontext *ctx = st->ctx; - struct pipe_rasterizer_state raster; - const struct cso_rasterizer *cso; + struct pipe_rasterizer_state *raster = &st->state.rasterizer; const struct gl_vertex_program *vertProg = ctx->VertexProgram._Current; uint i; - memset(&raster, 0, sizeof(raster)); + memset(raster, 0, sizeof(*raster)); - raster.origin_lower_left = 1; /* Always true for OpenGL */ + raster->origin_lower_left = 1; /* Always true for OpenGL */ /* _NEW_POLYGON, _NEW_BUFFERS */ { if (ctx->Polygon.FrontFace == GL_CCW) - raster.front_winding = PIPE_WINDING_CCW; + raster->front_winding = PIPE_WINDING_CCW; else - raster.front_winding = PIPE_WINDING_CW; + raster->front_winding = PIPE_WINDING_CW; /* XXX * I think the intention here is that user-created framebuffer objects @@ -96,13 +96,13 @@ static void update_raster_state( struct st_context *st ) * But this is an implementation/driver-specific artifact - remove... */ if (ctx->DrawBuffer && ctx->DrawBuffer->Name != 0) - raster.front_winding ^= PIPE_WINDING_BOTH; + raster->front_winding ^= PIPE_WINDING_BOTH; } /* _NEW_LIGHT */ if (ctx->Light.ShadeModel == GL_FLAT) - raster.flatshade = 1; + raster->flatshade = 1; /* _NEW_LIGHT | _NEW_PROGRAM * @@ -113,28 +113,28 @@ static void update_raster_state( struct st_context *st ) if (ctx->VertexProgram._Current) { if (ctx->VertexProgram._Enabled) { /* user-defined program */ - raster.light_twoside = ctx->VertexProgram.TwoSideEnabled; + raster->light_twoside = ctx->VertexProgram.TwoSideEnabled; } else { /* TNL-generated program */ - raster.light_twoside = ctx->Light.Enabled && ctx->Light.Model.TwoSide; + raster->light_twoside = ctx->Light.Enabled && ctx->Light.Model.TwoSide; } } else if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) { - raster.light_twoside = 1; + raster->light_twoside = 1; } /* _NEW_POLYGON */ if (ctx->Polygon.CullFlag) { if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK) { - raster.cull_mode = PIPE_WINDING_BOTH; + raster->cull_mode = PIPE_WINDING_BOTH; } else if (ctx->Polygon.CullFaceMode == GL_FRONT) { - raster.cull_mode = raster.front_winding; + raster->cull_mode = raster->front_winding; } else { - raster.cull_mode = raster.front_winding ^ PIPE_WINDING_BOTH; + raster->cull_mode = raster->front_winding ^ PIPE_WINDING_BOTH; } } @@ -144,23 +144,23 @@ static void update_raster_state( struct st_context *st ) GLuint fill_front = translate_fill( ctx->Polygon.FrontMode ); GLuint fill_back = translate_fill( ctx->Polygon.BackMode ); - if (raster.front_winding == PIPE_WINDING_CW) { - raster.fill_cw = fill_front; - raster.fill_ccw = fill_back; + if (raster->front_winding == PIPE_WINDING_CW) { + raster->fill_cw = fill_front; + raster->fill_ccw = fill_back; } else { - raster.fill_cw = fill_back; - raster.fill_ccw = fill_front; + raster->fill_cw = fill_back; + raster->fill_ccw = fill_front; } /* Simplify when culling is active: */ - if (raster.cull_mode & PIPE_WINDING_CW) { - raster.fill_cw = raster.fill_ccw; + if (raster->cull_mode & PIPE_WINDING_CW) { + raster->fill_cw = raster->fill_ccw; } - if (raster.cull_mode & PIPE_WINDING_CCW) { - raster.fill_ccw = raster.fill_cw; + if (raster->cull_mode & PIPE_WINDING_CCW) { + raster->fill_ccw = raster->fill_cw; } } @@ -168,95 +168,91 @@ static void update_raster_state( struct st_context *st ) */ if (ctx->Polygon.OffsetUnits != 0.0 || ctx->Polygon.OffsetFactor != 0.0) { - raster.offset_cw = get_offset_flag( raster.fill_cw, &ctx->Polygon ); - raster.offset_ccw = get_offset_flag( raster.fill_ccw, &ctx->Polygon ); - raster.offset_units = ctx->Polygon.OffsetUnits; - raster.offset_scale = ctx->Polygon.OffsetFactor; + raster->offset_cw = get_offset_flag( raster->fill_cw, &ctx->Polygon ); + raster->offset_ccw = get_offset_flag( raster->fill_ccw, &ctx->Polygon ); + raster->offset_units = ctx->Polygon.OffsetUnits; + raster->offset_scale = ctx->Polygon.OffsetFactor; } if (ctx->Polygon.SmoothFlag) - raster.poly_smooth = 1; + raster->poly_smooth = 1; if (ctx->Polygon.StippleFlag) - raster.poly_stipple_enable = 1; + raster->poly_stipple_enable = 1; /* _NEW_BUFFERS, _NEW_POLYGON */ - if (raster.fill_cw != PIPE_POLYGON_MODE_FILL || - raster.fill_ccw != PIPE_POLYGON_MODE_FILL) + if (raster->fill_cw != PIPE_POLYGON_MODE_FILL || + raster->fill_ccw != PIPE_POLYGON_MODE_FILL) { GLfloat mrd = (ctx->DrawBuffer ? ctx->DrawBuffer->_MRD : 1.0); - raster.offset_units = ctx->Polygon.OffsetFactor * mrd; - raster.offset_scale = (ctx->Polygon.OffsetUnits * mrd * + raster->offset_units = ctx->Polygon.OffsetFactor * mrd; + raster->offset_scale = (ctx->Polygon.OffsetUnits * mrd * st->polygon_offset_scale); } /* _NEW_POINT */ - raster.point_size = ctx->Point.Size; - raster.point_smooth = ctx->Point.SmoothFlag; - raster.point_sprite = ctx->Point.PointSprite; + raster->point_size = ctx->Point.Size; + raster->point_smooth = ctx->Point.SmoothFlag; + raster->point_sprite = ctx->Point.PointSprite; for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) { if (ctx->Point.CoordReplace[i]) { if (ctx->Point.SpriteOrigin == GL_UPPER_LEFT) - raster.sprite_coord_mode[i] = PIPE_SPRITE_COORD_UPPER_LEFT; + raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_UPPER_LEFT; else - raster.sprite_coord_mode[i] = PIPE_SPRITE_COORD_LOWER_LEFT; + raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_LOWER_LEFT; } else { - raster.sprite_coord_mode[i] = PIPE_SPRITE_COORD_NONE; + raster->sprite_coord_mode[i] = PIPE_SPRITE_COORD_NONE; } } if (vertProg) { if (vertProg->Base.Id == 0) { if (vertProg->Base.OutputsWritten & (1 << VERT_RESULT_PSIZ)) { /* generated program which emits point size */ - raster.point_size_per_vertex = TRUE; + raster->point_size_per_vertex = TRUE; } } else if (ctx->VertexProgram.PointSizeEnabled) { /* user-defined program and GL_VERTEX_PROGRAM_POINT_SIZE set */ - raster.point_size_per_vertex = ctx->VertexProgram.PointSizeEnabled; + raster->point_size_per_vertex = ctx->VertexProgram.PointSizeEnabled; } } /* _NEW_LINE */ - raster.line_smooth = ctx->Line.SmoothFlag; + raster->line_smooth = ctx->Line.SmoothFlag; if (ctx->Line.SmoothFlag) { - raster.line_width = CLAMP(ctx->Line.Width, + raster->line_width = CLAMP(ctx->Line.Width, ctx->Const.MinLineWidthAA, ctx->Const.MaxLineWidthAA); } else { - raster.line_width = CLAMP(ctx->Line.Width, + raster->line_width = CLAMP(ctx->Line.Width, ctx->Const.MinLineWidth, ctx->Const.MaxLineWidth); } - raster.line_stipple_enable = ctx->Line.StippleFlag; - raster.line_stipple_pattern = ctx->Line.StipplePattern; + raster->line_stipple_enable = ctx->Line.StippleFlag; + raster->line_stipple_pattern = ctx->Line.StipplePattern; /* GL stipple factor is in [1,256], remap to [0, 255] here */ - raster.line_stipple_factor = ctx->Line.StippleFactor - 1; + raster->line_stipple_factor = ctx->Line.StippleFactor - 1; /* _NEW_MULTISAMPLE */ if (ctx->Multisample.Enabled) - raster.multisample = 1; + raster->multisample = 1; /* _NEW_SCISSOR */ if (ctx->Scissor.Enabled) - raster.scissor = 1; + raster->scissor = 1; - cso = st_cached_rasterizer_state(st, &raster); - if (st->state.rasterizer != cso) { - st->state.rasterizer = cso; - st->pipe->bind_rasterizer_state(st->pipe, cso->data); - } + cso_set_rasterizer(st->cso_context, raster); } const struct st_tracked_state st_update_rasterizer = { diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 1000f98ffc..1babba9b4f 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -33,11 +33,11 @@ #include "st_context.h" -#include "st_cache.h" #include "st_atom.h" #include "st_program.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "cso_cache/cso_context.h" /** @@ -124,9 +124,9 @@ update_samplers(struct st_context *st) /* loop over sampler units (aka tex image units) */ for (su = 0; su < st->ctx->Const.MaxTextureImageUnits; su++) { - struct pipe_sampler_state sampler; + struct pipe_sampler_state *sampler = st->state.samplers + su; - memset(&sampler, 0, sizeof(sampler)); + memset(sampler, 0, sizeof(*sampler)); if (fs->Base.Base.SamplersUsed & (1 << su)) { GLuint texUnit = fs->Base.Base.SamplerUnits[su]; @@ -135,37 +135,37 @@ update_samplers(struct st_context *st) assert(texobj); - sampler.wrap_s = gl_wrap_to_sp(texobj->WrapS); - sampler.wrap_t = gl_wrap_to_sp(texobj->WrapT); - sampler.wrap_r = gl_wrap_to_sp(texobj->WrapR); + sampler->wrap_s = gl_wrap_to_sp(texobj->WrapS); + sampler->wrap_t = gl_wrap_to_sp(texobj->WrapT); + sampler->wrap_r = gl_wrap_to_sp(texobj->WrapR); - sampler.min_img_filter = gl_filter_to_img_filter(texobj->MinFilter); - sampler.min_mip_filter = gl_filter_to_mip_filter(texobj->MinFilter); - sampler.mag_img_filter = gl_filter_to_img_filter(texobj->MagFilter); + sampler->min_img_filter = gl_filter_to_img_filter(texobj->MinFilter); + sampler->min_mip_filter = gl_filter_to_mip_filter(texobj->MinFilter); + sampler->mag_img_filter = gl_filter_to_img_filter(texobj->MagFilter); if (texobj->Target != GL_TEXTURE_RECTANGLE_ARB) - sampler.normalized_coords = 1; + sampler->normalized_coords = 1; - sampler.lod_bias = st->ctx->Texture.Unit[su].LodBias; + sampler->lod_bias = st->ctx->Texture.Unit[su].LodBias; #if 1 - sampler.min_lod = (texobj->MinLod) < 0.0 ? 0.0 : texobj->MinLod; - sampler.max_lod = texobj->MaxLod; + sampler->min_lod = (texobj->MinLod) < 0.0 ? 0.0 : texobj->MinLod; + sampler->max_lod = texobj->MaxLod; #else /* min/max lod should really be as follows (untested). * Also, calculate_first_last_level() needs to be overhauled * since today's hardware had real support for LOD clamping. */ - sampler.min_lod = MAX2(texobj->BaseLevel, texobj->MinLod); - sampler.max_lod = MIN2(texobj->MaxLevel, texobj->MaxLod); + sampler->min_lod = MAX2(texobj->BaseLevel, texobj->MinLod); + sampler->max_lod = MIN2(texobj->MaxLevel, texobj->MaxLod); #endif - sampler.max_anisotropy = texobj->MaxAnisotropy; + sampler->max_anisotropy = texobj->MaxAnisotropy; /* only care about ARB_shadow, not SGI shadow */ if (texobj->CompareMode == GL_COMPARE_R_TO_TEXTURE) { - sampler.compare = 1; - sampler.compare_mode = PIPE_TEX_COMPARE_R_TO_TEXTURE; - sampler.compare_func + sampler->compare = 1; + sampler->compare_mode = PIPE_TEX_COMPARE_R_TO_TEXTURE; + sampler->compare_func = st_compare_func_to_pipe(texobj->CompareFunc); } @@ -174,11 +174,10 @@ update_samplers(struct st_context *st) /* XXX more sampler state here */ } - st->state.sampler[su] = st_cached_sampler_state(st, &sampler)->data; + cso_single_sampler(st->cso_context, su, sampler); } - st->pipe->bind_sampler_states(st->pipe, st->state.num_samplers, - st->state.sampler); + cso_single_sampler_done(st->cso_context); } diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 10c131d554..0726688493 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -43,10 +43,9 @@ #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" -#include "cso_cache/cso_cache.h" +#include "cso_cache/cso_context.h" #include "st_context.h" -#include "st_cache.h" #include "st_atom.h" #include "st_program.h" #include "st_atom_shader.h" @@ -70,9 +69,6 @@ struct translated_vertex_program /** Maps VERT_RESULT_x to slot */ GLuint output_to_slot[VERT_RESULT_MAX]; - /** The program in TGSI format */ - struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; - /** Pointer to the translated vertex program */ struct st_vertex_program *vp; @@ -158,7 +154,7 @@ find_translated_vp(struct st_context *st, /* * Translate fragment program if needed. */ - if (!stfp->cso) { + if (!stfp->state.tokens) { GLuint inAttr, numIn = 0; for (inAttr = 0; inAttr < FRAG_ATTRIB_MAX; inAttr++) { @@ -175,11 +171,7 @@ find_translated_vp(struct st_context *st, assert(stfp->Base.Base.NumInstructions > 1); - (void) st_translate_fragment_program(st, stfp, - stfp->input_to_slot, - stfp->tokens, - ST_MAX_SHADER_TOKENS); - assert(stfp->cso); + st_translate_fragment_program(st, stfp, stfp->input_to_slot); } @@ -261,12 +253,8 @@ find_translated_vp(struct st_context *st, assert(stvp->Base.Base.NumInstructions > 1); - st_translate_vertex_program(st, stvp, - xvp->output_to_slot, - xvp->tokens, - ST_MAX_SHADER_TOKENS); + st_translate_vertex_program(st, stvp, xvp->output_to_slot); - assert(stvp->cso); xvp->vp = stvp; /* translated VP is up to date now */ @@ -296,12 +284,10 @@ update_linkage( struct st_context *st ) xvp = find_translated_vp(st, stvp, stfp); st->vp = stvp; - st->state.vs = xvp->vp; - st->pipe->bind_vs_state(st->pipe, st->state.vs->cso->data); - st->fp = stfp; - st->state.fs = stfp->cso; - st->pipe->bind_fs_state(st->pipe, st->state.fs->data); + + st->pipe->bind_vs_state(st->pipe, stvp->driver_shader); + st->pipe->bind_fs_state(st->pipe, stfp->driver_shader); st->vertex_result_to_slot = xvp->output_to_slot; } diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 663c4f205d..f1fddc4e02 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -35,7 +35,6 @@ #include "main/macros.h" #include "st_context.h" -#include "st_cache.h" #include "st_cb_accum.h" #include "st_cb_fbo.h" #include "st_draw.h" diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index e712fd84cd..eae40f2a4f 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -35,7 +35,6 @@ #include "main/macros.h" #include "shader/prog_instruction.h" #include "st_atom.h" -#include "st_cache.h" #include "st_context.h" #include "st_cb_accum.h" #include "st_cb_clear.h" @@ -50,6 +49,8 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" +#include "cso_cache/cso_context.h" + @@ -157,8 +158,7 @@ make_frag_shader(struct st_context *st) p->OutputsWritten = (1 << FRAG_RESULT_COLR); stfp = (struct st_fragment_program *) p; - st_translate_fragment_program(st, stfp, NULL, - stfp->tokens, ST_MAX_SHADER_TOKENS); + st_translate_fragment_program(st, stfp, NULL); return stfp; } @@ -206,9 +206,10 @@ make_vertex_shader(struct st_context *st) (1 << VERT_RESULT_HPOS)); stvp = (struct st_vertex_program *) p; - st_translate_vertex_program(st, stvp, NULL, - stvp->tokens, ST_MAX_SHADER_TOKENS); + st_translate_vertex_program(st, stvp, NULL); +#if 0 assert(stvp->cso); +#endif return stvp; } @@ -284,7 +285,6 @@ clear_with_quad(GLcontext *ctx, /* blend state: RGBA masking */ { struct pipe_blend_state blend; - const struct cso_blend *cso; memset(&blend, 0, sizeof(blend)); if (color) { if (ctx->Color.ColorMask[0]) @@ -298,14 +298,12 @@ clear_with_quad(GLcontext *ctx, if (st->ctx->Color.DitherFlag) blend.dither = 1; } - cso = st_cached_blend_state(st, &blend); - pipe->bind_blend_state(pipe, cso->data); + cso_set_blend(st->cso_context, &blend); } /* depth_stencil state: always pass/set to ref value */ { struct pipe_depth_stencil_alpha_state depth_stencil; - const struct cso_depth_stencil_alpha *cso; memset(&depth_stencil, 0, sizeof(depth_stencil)); if (depth) { depth_stencil.depth.enabled = 1; @@ -323,14 +321,13 @@ clear_with_quad(GLcontext *ctx, depth_stencil.stencil[0].value_mask = 0xff; depth_stencil.stencil[0].write_mask = ctx->Stencil.WriteMask[0] & 0xff; } - cso = st_cached_depth_stencil_alpha_state(st, &depth_stencil); - pipe->bind_depth_stencil_alpha_state(pipe, cso->data); + + cso_set_depth_stencil_alpha(st->cso_context, &depth_stencil); } /* rasterizer state: nothing */ { struct pipe_rasterizer_state raster; - const struct cso_rasterizer *cso; memset(&raster, 0, sizeof(raster)); #if 0 /* don't do per-pixel scissor; we'll just draw a PIPE_PRIM_QUAD @@ -339,8 +336,7 @@ clear_with_quad(GLcontext *ctx, if (ctx->Scissor.Enabled) raster.scissor = 1; #endif - cso = st_cached_rasterizer_state(st, &raster); - pipe->bind_rasterizer_state(pipe, cso->data); + cso_set_rasterizer(st->cso_context, &raster); } /* fragment shader state: color pass-through program */ @@ -349,7 +345,7 @@ clear_with_quad(GLcontext *ctx, if (!stfp) { stfp = make_frag_shader(st); } - pipe->bind_fs_state(pipe, stfp->cso->data); + pipe->bind_fs_state(pipe, stfp->driver_shader); } /* vertex shader state: color/position pass-through */ @@ -358,7 +354,7 @@ clear_with_quad(GLcontext *ctx, if (!stvp) { stvp = make_vertex_shader(st); } - pipe->bind_vs_state(pipe, stvp->cso->data); + pipe->bind_vs_state(pipe, stvp->driver_shader); } /* viewport state: viewport matching window dims */ @@ -380,16 +376,19 @@ clear_with_quad(GLcontext *ctx, /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); +#if 0 + /* Can't depend on old state objects still existing -- may have + * been deleted to make room in the hash, etc. (Should get + * fixed...) + */ + st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); +#else /* Restore pipe state */ - pipe->bind_blend_state(pipe, st->state.blend->data); - pipe->bind_depth_stencil_alpha_state(pipe, st->state.depth_stencil->data); - pipe->bind_fs_state(pipe, st->state.fs->data); - pipe->bind_vs_state(pipe, st->state.vs->cso->data); - pipe->bind_rasterizer_state(pipe, st->state.rasterizer->data); + cso_set_rasterizer(st->cso_context, &st->state.rasterizer); + pipe->bind_fs_state(pipe, st->fp->driver_shader); + pipe->bind_vs_state(pipe, st->vp->driver_shader); +#endif pipe->set_viewport_state(pipe, &st->state.viewport); - /* OR: - st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); - */ } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index dee4c4132a..18ec9645c4 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -41,7 +41,6 @@ #include "st_context.h" #include "st_atom.h" #include "st_atom_constbuf.h" -#include "st_cache.h" #include "st_draw.h" #include "st_program.h" #include "st_cb_drawpixels.h" @@ -58,6 +57,7 @@ #include "pipe/p_winsys.h" #include "util/p_tile.h" #include "shader/prog_instruction.h" +#include "cso_cache/cso_context.h" /** @@ -159,8 +159,7 @@ make_bitmap_fragment_program(GLcontext *ctx) stfp = (struct st_fragment_program *) p; stfp->Base.UsesKill = GL_TRUE; - st_translate_fragment_program(ctx->st, stfp, NULL, - stfp->tokens, ST_MAX_SHADER_TOKENS); + st_translate_fragment_program(ctx->st, stfp, NULL); return stfp; } @@ -204,8 +203,7 @@ combined_bitmap_fragment_program(GLcontext *ctx) #endif /* translate to TGSI tokens */ - st_translate_fragment_program(st, stfp, NULL, - stfp->tokens, ST_MAX_SHADER_TOKENS); + st_translate_fragment_program(st, stfp, NULL); /* save new program, update serial numbers */ st->bitmap.user_prog_sn = st->fp->serialNo; @@ -266,8 +264,7 @@ combined_drawpix_fragment_program(GLcontext *ctx) #endif /* translate to TGSI tokens */ - st_translate_fragment_program(st, stfp, NULL, - stfp->tokens, ST_MAX_SHADER_TOKENS); + st_translate_fragment_program(st, stfp, NULL); /* save new program, update serial numbers */ st->pixel_xfer.xfer_prog_sn = st->pixel_xfer.program->serialNo; @@ -343,8 +340,7 @@ make_fragment_shader_z(struct st_context *st) p->OutputsWritten = (1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR); stfp = (struct st_fragment_program *) p; - st_translate_fragment_program(st, stfp, NULL, - stfp->tokens, ST_MAX_SHADER_TOKENS); + st_translate_fragment_program(st, stfp, NULL); return stfp; } @@ -421,8 +417,7 @@ st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor) } stvp = (struct st_vertex_program *) p; - st_translate_vertex_program(st, stvp, NULL, - stvp->tokens, ST_MAX_SHADER_TOKENS); + st_translate_vertex_program(st, stvp, NULL); progs[passColor] = stvp; @@ -641,7 +636,9 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, const GLfloat *color, GLboolean invertTex) { + struct st_context *st = ctx->st; struct pipe_context *pipe = ctx->st->pipe; + struct cso_context *cso = ctx->st->cso_context; GLfloat x0, y0, x1, y1; GLuint maxSize; @@ -656,24 +653,23 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* setup state: just scissor */ { struct pipe_rasterizer_state setup; - const struct cso_rasterizer *cso; memset(&setup, 0, sizeof(setup)); if (ctx->Scissor.Enabled) setup.scissor = 1; - cso = st_cached_rasterizer_state(ctx->st, &setup); - pipe->bind_rasterizer_state(pipe, cso->data); + + cso_set_rasterizer(cso, &setup); } /* fragment shader state: TEX lookup program */ - pipe->bind_fs_state(pipe, stfp->cso->data); + pipe->bind_fs_state(pipe, stfp->driver_shader); /* vertex shader state: position + texcoord pass-through */ - pipe->bind_vs_state(pipe, stvp->cso->data); + pipe->bind_vs_state(pipe, stvp->driver_shader); + /* texture sampling state: */ { struct pipe_sampler_state sampler; - const struct cso_sampler *cso; memset(&sampler, 0, sizeof(sampler)); sampler.wrap_s = PIPE_TEX_WRAP_CLAMP; sampler.wrap_t = PIPE_TEX_WRAP_CLAMP; @@ -682,8 +678,9 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; sampler.normalized_coords = 1; - cso = st_cached_sampler_state(ctx->st, &sampler); - pipe->bind_sampler_states(pipe, 1, (void**)&cso->data); + + cso_single_sampler(cso, 0, &sampler); + cso_single_sampler_done(cso); } /* viewport state: viewport matching window dims */ @@ -723,14 +720,25 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, draw_quad(ctx, x0, y0, z, x1, y1, invertTex); /* restore GL state */ - pipe->bind_rasterizer_state(pipe, ctx->st->state.rasterizer->data); - pipe->bind_fs_state(pipe, ctx->st->state.fs->data); - pipe->bind_vs_state(pipe, ctx->st->state.vs->cso->data); pipe->set_sampler_textures(pipe, ctx->st->state.num_textures, ctx->st->state.sampler_texture); - pipe->bind_sampler_states(pipe, ctx->st->state.num_samplers, - ctx->st->state.sampler); + pipe->set_viewport_state(pipe, &ctx->st->state.viewport); + +#if 0 + /* Can't depend on old state objects still existing -- may have + * been deleted to make room in the hash, etc. (Should get + * fixed...) + */ + st_invalidate_state(ctx, _NEW_COLOR | _NEW_TEXTURE); +#else + /* restore state */ + pipe->bind_fs_state(pipe, st->fp->driver_shader); + pipe->bind_vs_state(pipe, st->vp->driver_shader); + cso_set_rasterizer(cso, &st->state.rasterizer); + cso_set_samplers(cso, PIPE_MAX_SAMPLERS, + (const struct pipe_sampler_state **) st->state.sampler_list); +#endif } @@ -798,10 +806,10 @@ compatible_formats(GLenum format, GLenum type, enum pipe_format pipeFormat) static GLboolean any_fragment_ops(const struct st_context *st) { - if (st->state.depth_stencil->state.alpha.enabled || - st->state.blend->state.blend_enable || - st->state.blend->state.logicop_enable || - st->state.depth_stencil->state.depth.enabled) + if (st->state.depth_stencil.alpha.enabled || + st->state.depth_stencil.depth.enabled || + st->state.blend.blend_enable || + st->state.blend.logicop_enable) /* XXX more checks */ return GL_TRUE; else diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 61d4f4c41c..4dc76f19b1 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -168,11 +168,13 @@ static void st_program_string_notify( GLcontext *ctx, stfp->serialNo++; +#if 0 if (stfp->cso) { /* free the TGSI code */ // cso_delete(stfp->vs); stfp->cso = NULL; } +#endif stfp->param_state = stfp->Base.Base.Parameters->StateFlags; @@ -184,13 +186,14 @@ static void st_program_string_notify( GLcontext *ctx, stvp->serialNo++; +#if 0 if (stvp->cso) { /* free the CSO data */ st->pipe->delete_vs_state(st->pipe, stvp->cso->data); FREE((void *) stvp->cso); stvp->cso = NULL; } - +#endif if (stvp->draw_shader) { draw_delete_vertex_shader(st->draw, stvp->draw_shader); stvp->draw_shader = NULL; diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 09e389f9dc..5458ab420e 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -56,6 +56,7 @@ #include "pipe/p_inlines.h" #include "draw/draw_context.h" #include "cso_cache/cso_cache.h" +#include "cso_cache/cso_context.h" /** @@ -78,6 +79,7 @@ void st_invalidate_state(GLcontext * ctx, GLuint new_state) static struct st_context * st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) { + uint i; struct st_context *st = CALLOC_STRUCT( st_context ); ctx->st = st; @@ -93,12 +95,15 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st->dirty.mesa = ~0; st->dirty.st = ~0; - st->cache = cso_cache_create(); + st->cso_context = cso_create_context(pipe); st_init_atoms( st ); st_init_draw( st ); st_init_generate_mipmap(st); + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + st->state.sampler_list[i] = &st->state.samplers[i]; + /* we want all vertex data to be placed in buffer objects */ vbo_use_buffer_objects(ctx); @@ -149,7 +154,7 @@ static void st_destroy_context_priv( struct st_context *st ) _vbo_DestroyContext(st->ctx); - cso_cache_delete( st->cache ); + cso_destroy_context(st->cso_context); _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 897a5109b7..e81aebba3d 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -74,14 +74,11 @@ struct st_context * Other state is just parameter values. */ struct { - const struct cso_alpha_test *alpha_test; - const struct cso_blend *blend; - void *sampler[PIPE_MAX_SAMPLERS]; - const struct cso_depth_stencil_alpha *depth_stencil; - const struct cso_rasterizer *rasterizer; - const struct cso_fragment_shader *fs; - struct st_vertex_program *vs; - + struct pipe_blend_state blend; + struct pipe_depth_stencil_alpha_state depth_stencil; + struct pipe_rasterizer_state rasterizer; + struct pipe_sampler_state samplers[PIPE_MAX_SAMPLERS]; + struct pipe_sampler_state *sampler_list[PIPE_MAX_SAMPLERS]; struct pipe_blend_color blend_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; @@ -151,6 +148,10 @@ struct st_context /** For gen/render mipmap feature */ struct { + struct pipe_blend_state blend; + struct pipe_depth_stencil_alpha_state depthstencil; + struct pipe_rasterizer_state rasterizer; + void *blend_cso; void *depthstencil_cso; void *rasterizer_cso; @@ -158,7 +159,7 @@ struct st_context struct st_vertex_program *stvp; } gen_mipmap; - struct cso_cache *cache; + struct cso_context *cso_context; }; diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index 5888bcb98a..9c13010da8 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -53,15 +53,15 @@ st_print_current(void) int i; printf("Vertex Transform Inputs:\n"); - for (i = 0; i < st->state.vs->cso->state.num_inputs; i++) { + for (i = 0; i < st->vp->state.num_inputs; i++) { printf(" Slot %d: VERT_ATTRIB_%d\n", i, st->vp->index_to_input[i]); } - tgsi_dump( st->state.vs->cso->state.tokens, 0 ); + tgsi_dump( st->vp->state.tokens, 0 ); if (st->vp->Base.Base.Parameters) _mesa_print_parameter_list(st->vp->Base.Base.Parameters); - tgsi_dump( st->state.fs->state.tokens, 0 ); + tgsi_dump( st->fp->state.tokens, 0 ); if (st->fp->Base.Base.Parameters) _mesa_print_parameter_list(st->fp->Base.Base.Parameters); } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 1c0fa8c6aa..98504e46c1 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -36,7 +36,6 @@ #include "vbo/vbo.h" #include "st_atom.h" -#include "st_cache.h" #include "st_context.h" #include "st_cb_bufferobjects.h" #include "st_draw.h" @@ -219,7 +218,7 @@ st_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = &ctx->st->state.vs->cso->state; + vs = &ctx->st->vp->state; /* loop over TGSI shader inputs to determine vertex buffer * and attribute info @@ -481,10 +480,10 @@ st_feedback_draw_vbo(GLcontext *ctx, /* must get these after state validation! */ vp = ctx->st->vp; - vs = &ctx->st->state.vs->cso->state; + vs = &st->vp->state; - if (!st->state.vs->draw_shader) { - st->state.vs->draw_shader = draw_create_vertex_shader(draw, vs); + if (!st->vp->draw_shader) { + st->vp->draw_shader = draw_create_vertex_shader(draw, vs); } /* @@ -496,8 +495,8 @@ st_feedback_draw_vbo(GLcontext *ctx, assert(draw); draw_set_viewport_state(draw, &st->state.viewport); draw_set_clip_state(draw, &st->state.clip); - draw_set_rasterizer_state(draw, &st->state.rasterizer->state); - draw_bind_vertex_shader(draw, st->state.vs->draw_shader); + draw_set_rasterizer_state(draw, &st->state.rasterizer); + draw_bind_vertex_shader(draw, st->vp->draw_shader); set_feedback_vertex_format(ctx); /* loop over TGSI shader inputs to determine vertex buffer diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 3723e26d45..9c4e1032ef 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -38,6 +38,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "cso_cache/cso_cache.h" +#include "cso_cache/cso_context.h" #include "st_context.h" #include "st_draw.h" @@ -89,8 +90,7 @@ make_tex_fragment_program(GLcontext *ctx) stfp = (struct st_fragment_program *) p; - st_translate_fragment_program(ctx->st, stfp, NULL, - stfp->tokens, ST_MAX_SHADER_TOKENS); + st_translate_fragment_program(ctx->st, stfp, NULL); return stfp; } @@ -118,6 +118,7 @@ st_init_generate_mipmap(struct st_context *st) blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; blend.colormask = PIPE_MASK_RGBA; + st->gen_mipmap.blend = blend; st->gen_mipmap.blend_cso = pipe->create_blend_state(pipe, &blend); memset(&depthstencil, 0, sizeof(depthstencil)); @@ -257,14 +258,14 @@ st_render_mipmap(struct st_context *st, sampler.normalized_coords = 1; - /* bind CSOs */ - pipe->bind_blend_state(pipe, st->gen_mipmap.blend_cso); - pipe->bind_depth_stencil_alpha_state(pipe, st->gen_mipmap.depthstencil_cso); - pipe->bind_rasterizer_state(pipe, st->gen_mipmap.rasterizer_cso); + /* bind state */ + cso_set_blend(st->cso_context, &st->gen_mipmap.blend); + cso_set_depth_stencil_alpha(st->cso_context, &st->gen_mipmap.depthstencil); + cso_set_rasterizer(st->cso_context, &st->gen_mipmap.rasterizer); /* bind shaders */ - pipe->bind_fs_state(pipe, st->gen_mipmap.stfp->cso->data); - pipe->bind_vs_state(pipe, st->gen_mipmap.stvp->cso->data); + pipe->bind_fs_state(pipe, st->gen_mipmap.stfp->driver_shader); + pipe->bind_vs_state(pipe, st->gen_mipmap.stvp->driver_shader); /* * XXX for small mipmap levels, it may be faster to use the software @@ -304,17 +305,18 @@ st_render_mipmap(struct st_context *st, /*pt->first_level = first_level_save;*/ /* restore pipe state */ - if (st->state.rasterizer) - pipe->bind_rasterizer_state(pipe, st->state.rasterizer->data); - if (st->state.fs) - pipe->bind_fs_state(pipe, st->state.fs->data); - if (st->state.vs) - pipe->bind_vs_state(pipe, st->state.vs->cso->data); - pipe->bind_sampler_states(pipe, st->state.num_samplers, - st->state.sampler); +#if 0 + cso_set_rasterizer(st->cso_context, &st->state.rasterizer); + cso_set_samplers(st->cso_context, st->state.samplers_list); + pipe->bind_fs_state(pipe, st->fp->shader_program); + pipe->bind_vs_state(pipe, st->vp->shader_program); pipe->set_sampler_textures(pipe, st->state.num_textures, st->state.sampler_texture); pipe->set_viewport_state(pipe, &st->state.viewport); +#else + /* XXX is this sufficient? */ + st_invalidate_state(st->ctx, _NEW_COLOR | _NEW_TEXTURE); +#endif return TRUE; } diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 97206752af..9446b012ad 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -683,8 +683,9 @@ find_temporaries(const struct gl_program *program, * \param tokens array to store translated tokens in * \param maxTokens size of the tokens array * + * \return number of tokens placed in 'tokens' buffer, or zero if error */ -GLboolean +GLuint tgsi_translate_mesa_program( uint procType, const struct gl_program *program, @@ -918,6 +919,6 @@ tgsi_translate_mesa_program( maxTokens - ti ); } - return GL_TRUE; + return ti; } diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index 3ababf1339..63fc855b53 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -39,7 +39,7 @@ extern "C" { struct tgsi_token; struct gl_program; -GLboolean +GLuint tgsi_translate_mesa_program( uint procType, const struct gl_program *program, diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index aa252c845a..0f8784e132 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -42,15 +42,29 @@ #include "tgsi/util/tgsi_dump.h" #include "st_context.h" -#include "st_cache.h" #include "st_atom.h" #include "st_program.h" #include "st_mesa_to_tgsi.h" +#include "cso_cache/cso_context.h" #define TGSI_DEBUG 0 +/** XXX we should use the version of this from p_util.h but including + * that header causes symbol collisions. + */ +static INLINE void * +mem_dup(const void *src, uint size) +{ + void *dup = MALLOC(size); + if (dup) + memcpy(dup, src, size); + return dup; +} + + + /** * Translate a Mesa vertex shader into a TGSI shader. * \param outputMapping to map vertex program output registers to TGSI @@ -61,15 +75,15 @@ void st_translate_vertex_program(struct st_context *st, struct st_vertex_program *stvp, - const GLuint outputMapping[], - struct tgsi_token *tokensOut, - GLuint maxTokens) + const GLuint outputMapping[]) { + struct pipe_context *pipe = st->pipe; + struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; GLuint defaultOutputMapping[VERT_RESULT_MAX]; struct pipe_shader_state vs; - const struct cso_vertex_shader *cso; GLuint attr, i; GLuint num_generic = 0; + GLuint num_tokens; memset(&vs, 0, sizeof(vs)); @@ -240,7 +254,7 @@ st_translate_vertex_program(struct st_context *st, /* XXX: fix static allocation of tokens: */ - tgsi_translate_mesa_program( TGSI_PROCESSOR_VERTEX, + num_tokens = tgsi_translate_mesa_program( TGSI_PROCESSOR_VERTEX, &stvp->Base.Base, /* inputs */ vs.num_inputs, @@ -252,20 +266,21 @@ st_translate_vertex_program(struct st_context *st, vs.num_outputs, outputMapping, vs.output_semantic_name, - vs.output_semantic_index, + vs.output_semantic_index, /* tokenized result */ - tokensOut, maxTokens); + tokens, ST_MAX_SHADER_TOKENS); - vs.tokens = tokensOut; + vs.tokens = (struct tgsi_token *) + mem_dup(tokens, num_tokens * sizeof(tokens[0])); - cso = st_cached_vs_state(st, &vs); - stvp->cso = cso; + stvp->state = vs; /* struct copy */ + stvp->driver_shader = pipe->create_vs_state(pipe, &vs); if (0) _mesa_print_program(&stvp->Base.Base); if (TGSI_DEBUG) - tgsi_dump( tokensOut, 0 ); + tgsi_dump( vs.tokens, 0 ); } @@ -280,10 +295,10 @@ st_translate_vertex_program(struct st_context *st, const struct cso_fragment_shader * st_translate_fragment_program(struct st_context *st, struct st_fragment_program *stfp, - const GLuint inputMapping[], - struct tgsi_token *tokensOut, - GLuint maxTokens) + const GLuint inputMapping[]) { + struct pipe_context *pipe = st->pipe; + struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; GLuint outputMapping[FRAG_RESULT_MAX]; GLuint defaultInputMapping[FRAG_ATTRIB_MAX]; struct pipe_shader_state fs; @@ -293,6 +308,7 @@ st_translate_fragment_program(struct st_context *st, const GLbitfield inputsRead = stfp->Base.Base.InputsRead; GLuint vslot = 0; GLuint num_generic = 0; + GLuint num_tokens; memset(&fs, 0, sizeof(fs)); @@ -401,7 +417,7 @@ st_translate_fragment_program(struct st_context *st, /* XXX: fix static allocation of tokens: */ - tgsi_translate_mesa_program( TGSI_PROCESSOR_FRAGMENT, + num_tokens = tgsi_translate_mesa_program( TGSI_PROCESSOR_FRAGMENT, &stfp->Base.Base, /* inputs */ fs.num_inputs, @@ -415,18 +431,19 @@ st_translate_fragment_program(struct st_context *st, fs.output_semantic_name, fs.output_semantic_index, /* tokenized result */ - tokensOut, maxTokens); + tokens, ST_MAX_SHADER_TOKENS); - fs.tokens = tokensOut; + fs.tokens = (struct tgsi_token *) + mem_dup(tokens, num_tokens * sizeof(tokens[0])); - cso = st_cached_fs_state(st, &fs); - stfp->cso = cso; + stfp->state = fs; /* struct copy */ + stfp->driver_shader = pipe->create_fs_state(pipe, &fs); if (0) _mesa_print_program(&stfp->Base.Base); if (TGSI_DEBUG) - tgsi_dump( tokensOut, 0/*TGSI_DUMP_VERBOSE*/ ); + tgsi_dump( fs.tokens, 0/*TGSI_DUMP_VERBOSE*/ ); return cso; } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 31558af6ce..78786dcbb6 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -60,11 +60,8 @@ struct st_fragment_program /** map FP input back to VP output */ GLuint input_map[PIPE_MAX_SHADER_INPUTS]; - /** The program in TGSI format */ - struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; - - /** Pointer to the corresponding cached shader */ - const struct cso_fragment_shader *cso; + struct pipe_shader_state state; + struct pipe_shader_state *driver_shader; GLuint param_state; @@ -88,11 +85,8 @@ struct st_vertex_program /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ GLuint index_to_input[PIPE_MAX_SHADER_INPUTS]; - /** The program in TGSI format */ - struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; - - /** Pointer to the corresponding cached shader */ - const struct cso_vertex_shader *cso; + struct pipe_shader_state state; + struct pipe_shader_state *driver_shader; /** For using our private draw module (glRasterPos) */ struct draw_vertex_shader *draw_shader; @@ -122,16 +116,12 @@ st_vertex_program( struct gl_vertex_program *vp ) extern const struct cso_fragment_shader * st_translate_fragment_program(struct st_context *st, struct st_fragment_program *fp, - const GLuint inputMapping[], - struct tgsi_token *tokens, - GLuint maxTokens); + const GLuint inputMapping[]); extern void st_translate_vertex_program(struct st_context *st, struct st_vertex_program *vp, - const GLuint vert_output_to_slot[], - struct tgsi_token *tokens, - GLuint maxTokens); + const GLuint vert_output_to_slot[]); #endif -- cgit v1.2.3 From 7aa34eb40c05a9ccdbe5cef3fd426def0c295a60 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Mar 2008 19:02:51 -0600 Subject: gallium: dummy install target --- src/gallium/Makefile | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/Makefile b/src/gallium/Makefile index 568747e157..aa77021daf 100644 --- a/src/gallium/Makefile +++ b/src/gallium/Makefile @@ -19,3 +19,7 @@ subdirs: clean: rm -f `find . -name \*.[oa]` rm -f `find . -name depend` + + +# Dummy install target +install: -- cgit v1.2.3 From 221adbd60116d6334996a6b71a8dd133e229a3e9 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 11 Mar 2008 20:03:37 -0600 Subject: cell: check for NULL shader pointer in cell_bind_vs_state() --- src/gallium/drivers/cell/ppu/cell_state_shader.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c index 935501441b..269a5c15ba 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -131,7 +131,8 @@ cell_bind_vs_state(struct pipe_context *pipe, void *vs) cell->vs = (const struct cell_vertex_shader_state *) vs; - draw_bind_vertex_shader(cell->draw, cell->vs->draw_data); + draw_bind_vertex_shader(cell->draw, + (cell->vs ? cell->vs->draw_data : NULL)); cell->dirty |= CELL_NEW_VS; } -- cgit v1.2.3 From feb02084a88ca6e23c34fa06e963765c890f0b65 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 12 Mar 2008 11:37:02 +0100 Subject: tgsi: Dump source register divide component. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 59be14a748..a393c65da6 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -1047,6 +1047,11 @@ dump_instruction_short( CHR( ')' ); } + if (src->SrcRegisterExtSwz.ExtDivide != TGSI_EXTSWIZZLE_ONE) { + CHR( '/' ); + ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES_SHORT ); + } + first_reg = FALSE; } -- cgit v1.2.3 From 98ae83d5cc73b61826823c915b5c59746c2e85c7 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 12 Mar 2008 10:39:25 +0000 Subject: gallium: Add TEX_FILTER_ANISO img filter Hardware almost universally expects us to set a special filtering mode when anisotropic filtering is enabled, as opposed to varying a max-aniso values. Do this once in the state tracker & simplify the driver code. --- src/gallium/drivers/i915simple/i915_state.c | 19 +++++------ .../drivers/i965simple/brw_wm_sampler_state.c | 38 +++++++++++----------- src/gallium/drivers/softpipe/sp_tex_sample.c | 3 ++ src/gallium/include/pipe/p_defines.h | 2 +- src/mesa/state_tracker/st_atom_sampler.c | 4 +++ 5 files changed, 35 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index d9ab483bfc..57b195ea8d 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -71,6 +71,8 @@ static unsigned translate_img_filter( unsigned filter ) return FILTER_NEAREST; case PIPE_TEX_FILTER_LINEAR: return FILTER_LINEAR; + case PIPE_TEX_FILTER_ANISO: + return FILTER_ANISOTROPIC; default: assert(0); return FILTER_NEAREST; @@ -84,7 +86,7 @@ static unsigned translate_mip_filter( unsigned filter ) return MIPFILTER_NONE; case PIPE_TEX_MIPFILTER_NEAREST: return MIPFILTER_NEAREST; - case PIPE_TEX_FILTER_LINEAR: + case PIPE_TEX_MIPFILTER_LINEAR: return MIPFILTER_LINEAR; default: assert(0); @@ -211,16 +213,11 @@ i915_create_sampler_state(struct pipe_context *pipe, cso->templ = sampler; mipFilt = translate_mip_filter(sampler->min_mip_filter); - if (sampler->max_anisotropy > 1.0) { - minFilt = FILTER_ANISOTROPIC; - magFilt = FILTER_ANISOTROPIC; - if (sampler->max_anisotropy > 2.0) { - cso->state[0] |= SS2_MAX_ANISO_4; - } - } - else { - minFilt = translate_img_filter( sampler->min_img_filter ); - magFilt = translate_img_filter( sampler->mag_img_filter ); + minFilt = translate_img_filter( sampler->min_img_filter ); + magFilt = translate_img_filter( sampler->mag_img_filter ); + + if (sampler->max_anisotropy > 2.0) { + cso->state[0] |= SS2_MAX_ANISO_4; } { diff --git a/src/gallium/drivers/i965simple/brw_wm_sampler_state.c b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c index ff5ba7e7c7..b9eaee56ee 100644 --- a/src/gallium/drivers/i965simple/brw_wm_sampler_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c @@ -136,6 +136,9 @@ static void brw_update_sampler_state( const struct pipe_sampler_state *pipe_samp case PIPE_TEX_FILTER_LINEAR: sampler->ss0.min_filter = BRW_MAPFILTER_LINEAR; break; + case PIPE_TEX_FILTER_ANISO: + sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC; + break; default: break; } @@ -155,26 +158,23 @@ static void brw_update_sampler_state( const struct pipe_sampler_state *pipe_samp } /* Set Anisotropy: */ - if (pipe_sampler->max_anisotropy > 1.0) { - sampler->ss0.min_filter = BRW_MAPFILTER_ANISOTROPIC; - sampler->ss0.mag_filter = BRW_MAPFILTER_ANISOTROPIC; - - if (pipe_sampler->max_anisotropy > 2.0) { - sampler->ss3.max_aniso = MAX2((pipe_sampler->max_anisotropy - 2) / 2, - BRW_ANISORATIO_16); - } + switch (pipe_sampler->mag_img_filter) { + case PIPE_TEX_FILTER_NEAREST: + sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST; + break; + case PIPE_TEX_FILTER_LINEAR: + sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR; + break; + case PIPE_TEX_FILTER_ANISO: + sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR; + break; + default: + break; } - else { - switch (pipe_sampler->mag_img_filter) { - case PIPE_TEX_FILTER_NEAREST: - sampler->ss0.mag_filter = BRW_MAPFILTER_NEAREST; - break; - case PIPE_TEX_FILTER_LINEAR: - sampler->ss0.mag_filter = BRW_MAPFILTER_LINEAR; - break; - default: - break; - } + + if (pipe_sampler->max_anisotropy > 2.0) { + sampler->ss3.max_aniso = MAX2((pipe_sampler->max_anisotropy - 2) / 2, + BRW_ANISORATIO_16); } sampler->ss1.s_wrap_mode = translate_wrap_mode(pipe_sampler->wrap_s); diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 0ced585c7f..34da6356d7 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -714,6 +714,7 @@ sp_get_samples_2d_common(struct tgsi_sampler *sampler, } break; case PIPE_TEX_FILTER_LINEAR: + case PIPE_TEX_FILTER_ANISO: for (j = 0; j < QUAD_SIZE; j++) { float tx[4][4], a, b; int x0, y0, x1, y1, c; @@ -846,6 +847,7 @@ sp_get_samples_3d(struct tgsi_sampler *sampler, } break; case PIPE_TEX_FILTER_LINEAR: + case PIPE_TEX_FILTER_ANISO: for (j = 0; j < QUAD_SIZE; j++) { float texel0[4][4], texel1[4][4]; float xw, yw, zw; /* interpolation weights */ @@ -972,6 +974,7 @@ sp_get_samples_rect(struct tgsi_sampler *sampler, } break; case PIPE_TEX_FILTER_LINEAR: + case PIPE_TEX_FILTER_ANISO: for (j = 0; j < QUAD_SIZE; j++) { float tx[4][4], a, b; int x0, y0, x1, y1, c; diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index 0c662d6517..bc938ba253 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -152,7 +152,7 @@ enum pipe_texture_target { */ #define PIPE_TEX_FILTER_NEAREST 0 #define PIPE_TEX_FILTER_LINEAR 1 -/* #define PIPE_TEX_FILTER_ANISO 2 */ +#define PIPE_TEX_FILTER_ANISO 2 #define PIPE_TEX_COMPARE_NONE 0 diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 1babba9b4f..d376480c91 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -160,6 +160,10 @@ update_samplers(struct st_context *st) #endif sampler->max_anisotropy = texobj->MaxAnisotropy; + if (sampler->max_anisotropy > 1.0) { + sampler->min_img_filter = PIPE_TEX_FILTER_ANISO; + sampler->mag_img_filter = PIPE_TEX_FILTER_ANISO; + } /* only care about ARB_shadow, not SGI shadow */ if (texobj->CompareMode == GL_COMPARE_R_TO_TEXTURE) { -- cgit v1.2.3 From 8fd633b5cfa36e0cf0acef096315c9250015aba7 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 12 Mar 2008 10:43:53 +0000 Subject: gallium: reduce signed/unsigned warnings --- src/gallium/auxiliary/draw/draw_passthrough.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_passthrough.c b/src/gallium/auxiliary/draw/draw_passthrough.c index fc2dde38ba..a51fa0ab23 100644 --- a/src/gallium/auxiliary/draw/draw_passthrough.c +++ b/src/gallium/auxiliary/draw/draw_passthrough.c @@ -78,7 +78,7 @@ static void fetch_xyz_rgb_st( struct draw_context *draw, { const unsigned *pitch = draw->vertex_fetch.pitch; const ubyte **src = draw->vertex_fetch.src_ptr; - int i; + unsigned i; const ubyte *xyzw = src[0] + start * pitch[0]; const ubyte *rgba = src[1] + start * pitch[1]; -- cgit v1.2.3 From 70ae7f09c739465242b0c6255196dae1de9dd8d3 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 12 Mar 2008 14:54:43 +0100 Subject: tgsi: Introduce OPCODE_TXP. Depricate ExdDivide field. --- src/gallium/include/pipe/p_shader_tokens.h | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index e4d89971f9..f0bb43bc5b 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -227,6 +227,7 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_STR 51 #define TGSI_OPCODE_TEX 52 #define TGSI_OPCODE_TXD 53 +#define TGSI_OPCODE_TXP 134 #define TGSI_OPCODE_UP2H 54 #define TGSI_OPCODE_UP2US 55 #define TGSI_OPCODE_UP4B 56 @@ -421,7 +422,7 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_KIL 132 /* unpredicated kill */ #define TGSI_OPCODE_END 133 /* aka HALT */ -#define TGSI_OPCODE_LAST 134 +#define TGSI_OPCODE_LAST 135 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ @@ -670,7 +671,14 @@ struct tgsi_src_register_ext_swz unsigned NegateY : 1; /* BOOL */ unsigned NegateZ : 1; /* BOOL */ unsigned NegateW : 1; /* BOOL */ + + /* + * XXX: Do not use. This field has been depricated. + * XXX: If using in conjunction with OPCODE_TEX, please use OPCODE_TXP + * XXX: and, if needed, perform a swizzle on the texture coordinate. + */ unsigned ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ + unsigned Padding : 3; unsigned Extended : 1; /* BOOL */ }; -- cgit v1.2.3 From a2ea51ed828d8b503492a7b42ac937d2642ac4f1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 12 Mar 2008 15:01:18 +0000 Subject: gallium: Change assert behavior on runtime (Mark Mueller). --- src/gallium/auxiliary/util/p_debug.c | 38 ++++++++++++++++++++++++++++++++++-- 1 file changed, 36 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 04e55dd91d..09cabdae25 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -36,8 +36,9 @@ #include #endif -#include "pipe/p_debug.h" #include "pipe/p_compiler.h" +#include "pipe/p_util.h" +#include "pipe/p_debug.h" #ifdef WIN32 @@ -97,11 +98,44 @@ static INLINE void debug_break(void) #endif } +#if defined(WIN32) +ULONG_PTR debug_config_file = 0; +void *mapped_config_file = 0; + +enum { + eAssertAbortEn = 0x1, +}; + +/* Check for aborts enabled. */ +static unsigned abort_en() +{ + if (!mapped_config_file) + { + /* Open an 8 byte file for configuration data. */ + mapped_config_file = EngMapFile(L"\\??\\c:\\gaDebug.cfg", 8, &debug_config_file); + } + /* An value of "0" (ascii) in the configuration file will clear the first 8 bits in the test byte. */ + /* An value of "1" (ascii) in the configuration file will set the first bit in the test byte. */ + /* An value of "2" (ascii) in the configuration file will set the second bit in the test byte. */ + return ((((char *)mapped_config_file)[0]) - 0x30) & eAssertAbortEn; +} +#else /* WIN32 */ +static unsigned abort_en() +{ + return !GETENV("GALLIUM_ABORT_ON_ASSERT"); +} +#endif void debug_assert_fail(const char *expr, const char *file, unsigned line) { debug_printf("%s:%i: Assertion `%s' failed.\n", file, line, expr); - debug_break(); + if (abort_en()) + { + debug_break(); + } else + { + debug_printf("continuing...\n"); + } } -- cgit v1.2.3 From 8901a46a742b0cfecde0b981fc65160bf3e8d019 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 12 Mar 2008 15:02:11 +0000 Subject: gallium: Generic handle table. --- src/gallium/auxiliary/util/Makefile | 1 + src/gallium/auxiliary/util/SConscript | 1 + src/gallium/auxiliary/util/u_handle_table.c | 207 ++++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_handle_table.h | 96 +++++++++++++ 4 files changed, 305 insertions(+) create mode 100644 src/gallium/auxiliary/util/u_handle_table.c create mode 100644 src/gallium/auxiliary/util/u_handle_table.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 2a3a9380b3..2abbe9500e 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -7,6 +7,7 @@ C_SOURCES = \ p_debug.c \ p_tile.c \ p_util.c \ + u_handle_table.c \ u_mm.c \ u_snprintf.c diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 642088b962..2030214aa7 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -6,6 +6,7 @@ util = env.ConvenienceLibrary( 'p_debug.c', 'p_tile.c', 'p_util.c', + 'u_handle_table.c', 'u_mm.c', 'u_snprintf.c', ]) diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c new file mode 100644 index 0000000000..8a298f7c41 --- /dev/null +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -0,0 +1,207 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Generic handle table implementation. + * + * @author José Fonseca + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_util.h" + +#include "u_handle_table.h" + + +#define HANDLE_TABLE_INITIAL_SIZE 16 + + +struct handle_table +{ + /** Object array. Empty handles have a null object */ + void **objects; + + /** Number of objects the handle can currently hold */ + unsigned size; + /** Number of consecutive objects allocated at the start of the table */ + unsigned filled; + + /** Optional object destructor */ + void (*destroy)(void *object); +}; + + +struct handle_table * +handle_table_create(void) +{ + struct handle_table *ht; + + ht = MALLOC_STRUCT(handle_table); + if(!ht) + return NULL; + + ht->objects = (void **)CALLOC(HANDLE_TABLE_INITIAL_SIZE, sizeof(void *)); + if(!ht->objects) { + FREE(ht); + return NULL; + } + + ht->size = HANDLE_TABLE_INITIAL_SIZE; + ht->filled = 0; + + ht->destroy = NULL; + + return ht; +} + + +void +handle_table_set_destroy(struct handle_table *ht, + void (*destroy)(void *object)) +{ + assert(ht); + ht->destroy = destroy; +} + + +unsigned +handle_table_add(struct handle_table *ht, + void *object) +{ + unsigned index; + unsigned handle; + + assert(ht); + assert(object); + if(!object) + return 0; + + /* linear search for an empty handle */ + while(ht->filled < ht->size) { + if(!ht->objects[ht->filled]) + break; + ++ht->filled; + } + + /* grow the table */ + if(ht->filled == ht->size) { + unsigned new_size; + void **new_objects; + + new_size = ht->size*2; + assert(new_size); + + new_objects = (void **)REALLOC((void *)ht->objects, + ht->size*sizeof(void *), + new_size*sizeof(void *)); + if(!new_objects) + return 0; + + memset(new_objects + ht->size, 0, (new_size - ht->size)*sizeof(void *)); + + ht->size = new_size; + ht->objects = new_objects; + } + + index = ht->filled; + + handle = index + 1; + + /* check integer overflow */ + if(!handle) + return 0; + + assert(!ht->objects[index]); + ht->objects[index] = object; + ++ht->filled; + + return handle; +} + + +void * +handle_table_get(struct handle_table *ht, + unsigned handle) +{ + void *object; + + assert(ht); + assert(handle > 0); + assert(handle <= ht->size); + if(!handle || handle > ht->size) + return NULL; + + object = ht->objects[handle - 1]; + assert(object); + + return object; +} + + +void +handle_table_remove(struct handle_table *ht, + unsigned handle) +{ + void *object; + unsigned index; + + assert(ht); + assert(handle > 0); + assert(handle <= ht->size); + if(!handle || handle > ht->size) + return; + + index = handle - 1; + object = ht->objects[index]; + assert(object); + + if(object && ht->destroy) + ht->destroy(object); + + ht->objects[index] = NULL; + if(index < ht->filled) + ht->filled = index; +} + + +void +handle_table_destroy(struct handle_table *ht) +{ + unsigned index; + assert(ht); + + if(ht->destroy) + for(index = 0; index < ht->size; ++index) + if(ht->objects[index]) + ht->destroy(ht->objects[index]); + + FREE(ht->objects); + FREE(ht); +} + diff --git a/src/gallium/auxiliary/util/u_handle_table.h b/src/gallium/auxiliary/util/u_handle_table.h new file mode 100644 index 0000000000..51fc273865 --- /dev/null +++ b/src/gallium/auxiliary/util/u_handle_table.h @@ -0,0 +1,96 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Generic handle table. + * + * @author José Fonseca + */ + +#ifndef U_HANDLE_TABLE_H_ +#define U_HANDLE_TABLE_H_ + + +#ifdef __cplusplus +extern "C" { +#endif + + +/** + * Abstract data type to map integer handles to objects. + */ +struct handle_table; + + +struct handle_table * +handle_table_create(void); + + +/** + * Set an optional destructor callback. + * + * If set, it will be called during handle_table_remove and + * handle_table_destroy calls. + */ +void +handle_table_set_destroy(struct handle_table *ht, + void (*destroy)(void *object)); + + +/** + * Add a new object. + * + * Returns a zero handle on failure (out of memory). + */ +unsigned +handle_table_add(struct handle_table *ht, + void *object); + +/** + * Fetch an existing object. + * + * Returns NULL for an invalid handle. + */ +void * +handle_table_get(struct handle_table *ht, + unsigned handle); + + +void +handle_table_remove(struct handle_table *ht, + unsigned handle); + + +void +handle_table_destroy(struct handle_table *ht); + + +#ifdef __cplusplus +} +#endif + +#endif /* U_HANDLE_TABLE_H_ */ -- cgit v1.2.3 From e5b1a53c9f9ad247272415e0e21e83cfe00728a9 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 12 Mar 2008 15:29:39 +0100 Subject: tgsi: Dump TXP opcode. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index a393c65da6..1913d482f1 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -459,7 +459,8 @@ static const char *TGSI_OPCODES[] = "OPCODE_IFC", "OPCODE_BREAKC", "OPCODE_KIL", - "OPCODE_END" + "OPCODE_END", + "OPCODE_TXP" }; static const char *TGSI_OPCODES_SHORT[] = @@ -597,7 +598,8 @@ static const char *TGSI_OPCODES_SHORT[] = "IFC", "BREAKC", "KIL", - "END" + "END", + "TXP" }; static const char *TGSI_SATS[] = -- cgit v1.2.3 From ba75e82b6ebaf88dd2e4a8f764b2d296d715bf8a Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 12 Mar 2008 16:41:25 +0100 Subject: tgsi: Remove ExtDivide field from existence. Implement OPCODE_TXP. --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 46 +++++++--------------- src/gallium/auxiliary/tgsi/util/tgsi_build.c | 5 --- src/gallium/auxiliary/tgsi/util/tgsi_build.h | 1 - src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 9 ----- .../drivers/i915simple/i915_fpc_translate.c | 12 +++--- src/gallium/include/pipe/p_shader_tokens.h | 13 +----- 6 files changed, 21 insertions(+), 65 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index ac52441400..f2ed9e0353 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -1220,7 +1220,8 @@ fetch_texel( struct tgsi_sampler *sampler, static void exec_tex(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst, - boolean biasLod) + boolean biasLod, + boolean projected) { const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; union tgsi_exec_channel r[8]; @@ -1234,17 +1235,9 @@ exec_tex(struct tgsi_exec_machine *mach, FETCH(&r[0], 0, CHAN_X); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: + if (projected) { FETCH(&r[1], 0, CHAN_W); micro_div( &r[0], &r[0], &r[1] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); } if (biasLod) { @@ -1266,19 +1259,11 @@ exec_tex(struct tgsi_exec_machine *mach, FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 0, CHAN_Z); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: + if (projected) { FETCH(&r[3], 0, CHAN_W); micro_div( &r[0], &r[0], &r[3] ); micro_div( &r[1], &r[1], &r[3] ); micro_div( &r[2], &r[2], &r[3] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); } if (biasLod) { @@ -1300,19 +1285,11 @@ exec_tex(struct tgsi_exec_machine *mach, FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 0, CHAN_Z); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: + if (projected) { FETCH(&r[3], 0, CHAN_W); micro_div( &r[0], &r[0], &r[3] ); micro_div( &r[1], &r[1], &r[3] ); micro_div( &r[2], &r[2], &r[3] ); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); } if (biasLod) { @@ -2007,14 +1984,14 @@ exec_instruction( /* simple texture lookup */ /* src[0] = texcoord */ /* src[1] = sampler unit */ - exec_tex(mach, inst, FALSE); + exec_tex(mach, inst, FALSE, FALSE); break; case TGSI_OPCODE_TXB: /* Texture lookup with lod bias */ /* src[0] = texcoord (src[0].w = LOD bias) */ /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); + exec_tex(mach, inst, TRUE, FALSE); break; case TGSI_OPCODE_TXD: @@ -2030,7 +2007,14 @@ exec_instruction( /* Texture lookup with explit LOD */ /* src[0] = texcoord (src[0].w = LOD) */ /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); + exec_tex(mach, inst, TRUE, FALSE); + break; + + case TGSI_OPCODE_TXP: + /* Texture lookup with projection */ + /* src[0] = texcoord (src[0].w = projection) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, FALSE, TRUE); break; case TGSI_OPCODE_UP2H: diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.c b/src/gallium/auxiliary/tgsi/util/tgsi_build.c index a00ff1c2a5..9c883ab704 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.c @@ -719,7 +719,6 @@ tgsi_build_full_instruction( reg->SrcRegisterExtSwz.NegateY, reg->SrcRegisterExtSwz.NegateZ, reg->SrcRegisterExtSwz.NegateW, - reg->SrcRegisterExtSwz.ExtDivide, prev_token, instruction, header ); @@ -1057,7 +1056,6 @@ tgsi_default_src_register_ext_swz( void ) src_register_ext_swz.NegateY = 0; src_register_ext_swz.NegateZ = 0; src_register_ext_swz.NegateW = 0; - src_register_ext_swz.ExtDivide = TGSI_EXTSWIZZLE_ONE; src_register_ext_swz.Padding = 0; src_register_ext_swz.Extended = 0; @@ -1084,7 +1082,6 @@ tgsi_build_src_register_ext_swz( unsigned negate_y, unsigned negate_z, unsigned negate_w, - unsigned ext_divide, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ) @@ -1099,7 +1096,6 @@ tgsi_build_src_register_ext_swz( assert( negate_y <= 1 ); assert( negate_z <= 1 ); assert( negate_w <= 1 ); - assert( ext_divide <= TGSI_EXTSWIZZLE_ONE ); src_register_ext_swz = tgsi_default_src_register_ext_swz(); src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; @@ -1110,7 +1106,6 @@ tgsi_build_src_register_ext_swz( src_register_ext_swz.NegateY = negate_y; src_register_ext_swz.NegateZ = negate_z; src_register_ext_swz.NegateW = negate_w; - src_register_ext_swz.ExtDivide = ext_divide; prev_token->Extended = 1; instruction_grow( instruction, header ); diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.h b/src/gallium/auxiliary/tgsi/util/tgsi_build.h index 607860e7fc..80bffc4ae7 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.h @@ -229,7 +229,6 @@ tgsi_build_src_register_ext_swz( unsigned negate_y, unsigned negate_z, unsigned negate_w, - unsigned ext_divide, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 1913d482f1..ceb407b884 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -1049,11 +1049,6 @@ dump_instruction_short( CHR( ')' ); } - if (src->SrcRegisterExtSwz.ExtDivide != TGSI_EXTSWIZZLE_ONE) { - CHR( '/' ); - ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES_SHORT ); - } - first_reg = FALSE; } @@ -1368,10 +1363,6 @@ dump_instruction_verbose( TXT( "\nNegateW : " ); UID( src->SrcRegisterExtSwz.NegateW ); } - if( deflt || fs->SrcRegisterExtSwz.ExtDivide != src->SrcRegisterExtSwz.ExtDivide ) { - TXT( "\nExtDivide : " ); - ENM( src->SrcRegisterExtSwz.ExtDivide, TGSI_EXTSWIZZLES ); - } if( ignored ) { TXT( "\nPadding : " ); UIX( src->SrcRegisterExtSwz.Padding ); diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index c2d9a93c6c..7b4fca5db1 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -872,19 +872,17 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_TEX: - if (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide - == TGSI_EXTSWIZZLE_W) { - emit_tex(p, inst, T0_TEXLDP); - } - else { - emit_tex(p, inst, T0_TEXLD); - } + emit_tex(p, inst, T0_TEXLD); break; case TGSI_OPCODE_TXB: emit_tex(p, inst, T0_TEXLDB); break; + case TGSI_OPCODE_TXP: + emit_tex(p, inst, T0_TEXLDP); + break; + case TGSI_OPCODE_XPD: /* Cross product: * result.x = src0.y * src1.z - src0.z * src1.y; diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index f0bb43bc5b..210169f54f 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -655,9 +655,6 @@ struct tgsi_src_register_ext * * NegateX, NegateY, NegateZ and NegateW negate individual components of the * source register. - * - * ExtDivide specifies which component is used to divide all components of the - * source register. */ struct tgsi_src_register_ext_swz @@ -671,15 +668,7 @@ struct tgsi_src_register_ext_swz unsigned NegateY : 1; /* BOOL */ unsigned NegateZ : 1; /* BOOL */ unsigned NegateW : 1; /* BOOL */ - - /* - * XXX: Do not use. This field has been depricated. - * XXX: If using in conjunction with OPCODE_TEX, please use OPCODE_TXP - * XXX: and, if needed, perform a swizzle on the texture coordinate. - */ - unsigned ExtDivide : 4; /* TGSI_EXTSWIZZLE_ */ - - unsigned Padding : 3; + unsigned Padding : 7; unsigned Extended : 1; /* BOOL */ }; -- cgit v1.2.3 From 12ab5f97013e398b9f6485b97d6691c3c170447a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Mar 2008 13:20:29 -0600 Subject: gallium: change draw_vertex_shader->state from pointer to struct We were sometimes keeping a pointer to a stack-allocated object. Now make a copy of the pipe_shader_state object. This should fix some seemingly random memory errors/crashes. --- src/gallium/auxiliary/draw/draw_private.h | 2 +- src/gallium/auxiliary/draw/draw_vs_exec.c | 4 ++-- src/gallium/auxiliary/draw/draw_vs_sse.c | 4 ++-- 3 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 25fa8c09c2..4147472d45 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -133,7 +133,7 @@ struct draw_vertex_shader { /* This member will disappear shortly: */ - const struct pipe_shader_state *state; + struct pipe_shader_state state; struct tgsi_shader_info info; diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 514303e0ea..55bec14116 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -71,7 +71,7 @@ vs_exec_prepare( struct draw_vertex_shader *shader, { /* specify the vertex program to interpret/execute */ tgsi_exec_machine_bind_shader(&draw->machine, - shader->state->tokens, + shader->state.tokens, PIPE_MAX_SAMPLERS, NULL /*samplers*/ ); @@ -187,7 +187,7 @@ draw_create_vs_exec(struct draw_context *draw, if (vs == NULL) return NULL; - vs->state = state; + vs->state = *state; vs->prepare = vs_exec_prepare; vs->run = vs_exec_run; vs->delete = vs_exec_delete; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index e5c1a40cca..5ee2adb344 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -221,14 +221,14 @@ draw_create_vs_sse(struct draw_context *draw, if (vs == NULL) return NULL; - vs->base.state = templ; + vs->base.state = *templ; vs->base.prepare = vs_sse_prepare; vs->base.run = vs_sse_run; vs->base.delete = vs_sse_delete; x86_init_func( &vs->sse2_program ); - if (!tgsi_emit_sse2( (struct tgsi_token *) vs->base.state->tokens, + if (!tgsi_emit_sse2( (struct tgsi_token *) vs->base.state.tokens, &vs->sse2_program )) goto fail; -- cgit v1.2.3 From a2b917c1cdb9cdd9b016e5cb9637e90c58fa911b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Mar 2008 13:21:41 -0600 Subject: gallium: use TXP rather than ExtDivide flag --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 9446b012ad..f3cfda0bfb 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -513,10 +513,9 @@ compile_instruction( case OPCODE_TXP: /* texture lookup with divide by Q component */ /* convert to TEX w/ special flag for division */ - fullinst->Instruction.Opcode = TGSI_OPCODE_TEX; + fullinst->Instruction.Opcode = TGSI_OPCODE_TXP; fullinst->Instruction.NumSrcRegs = 2; fullinst->InstructionExtTexture.Texture = map_texture_target( inst->TexSrcTarget ); - fullinst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide = TGSI_EXTSWIZZLE_W; fullinst->FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; fullinst->FullSrcRegisters[1].SrcRegister.Index = inst->TexSrcUnit; break; -- cgit v1.2.3 From 51809bc1bc719c8d988cd6e06b6c96af43d12026 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Mar 2008 13:22:58 -0600 Subject: remove reference to obsolete ExtDivide --- src/gallium/drivers/i965simple/brw_wm_glsl.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_wm_glsl.c b/src/gallium/drivers/i965simple/brw_wm_glsl.c index 44f946ea74..5c90583824 100644 --- a/src/gallium/drivers/i965simple/brw_wm_glsl.c +++ b/src/gallium/drivers/i965simple/brw_wm_glsl.c @@ -124,10 +124,6 @@ static struct brw_reg get_src_reg(struct brw_wm_compile *c, */ component = get_ext_swz(src->SrcRegisterExtSwz, component); - /* Can't handle this, don't know if we need to: - */ - assert(src->SrcRegisterExtSwz.ExtDivide == TGSI_EXTSWIZZLE_ONE); - /* Not handling indirect lookups yet: */ assert(src->SrcRegister.Indirect == 0); -- cgit v1.2.3 From 0dd79011b9c407191e40c2697c68946f35bc557a Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Mar 2008 14:20:54 -0600 Subject: gallium: pass NULL to cso_single_sampler() when the sampler isn't used. This fixes an AA line crash/regression. The aaline stage needs to find a free/unused sampler to do its thing. --- src/mesa/state_tracker/st_atom_sampler.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index d376480c91..5787a7492c 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -176,9 +176,12 @@ update_samplers(struct st_context *st) st->state.num_samplers = su + 1; /* XXX more sampler state here */ - } - cso_single_sampler(st->cso_context, su, sampler); + cso_single_sampler(st->cso_context, su, sampler); + } + else { + cso_single_sampler(st->cso_context, su, NULL); + } } cso_single_sampler_done(st->cso_context); -- cgit v1.2.3 From d1284d3be2bcf5e1607458cd9fb0d8c32fad13ea Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Mar 2008 15:33:41 -0600 Subject: mesa: set SamplersUsed bitmask when parsing ARB fragment programs --- src/mesa/shader/arbprogparse.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index b6b3c88b14..e385b9d997 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -3910,8 +3910,11 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target, program->Base.NumNativeTexIndirections = ap.Base.NumTexIndirections; program->Base.InputsRead = ap.Base.InputsRead; program->Base.OutputsWritten = ap.Base.OutputsWritten; - for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++) + for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++) { program->Base.TexturesUsed[i] = ap.TexturesUsed[i]; + if (ap.TexturesUsed[i]) + program->Base.SamplersUsed |= (1 << i); + } program->Base.ShadowSamplers = ap.ShadowSamplers; program->FogOption = ap.FogOption; program->UsesKill = ap.UsesKill; -- cgit v1.2.3 From 2109ba4c5d22e1f7effa33a6ff26ce587ce46fe3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Mar 2008 16:56:12 -0600 Subject: i915: handle NULL object in i915_bind_rasterizer_state() --- src/gallium/drivers/i915simple/i915_state.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 57b195ea8d..503c092400 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -668,22 +668,23 @@ i915_create_rasterizer_state(struct pipe_context *pipe, } static void i915_bind_rasterizer_state( struct pipe_context *pipe, - void *setup ) + void *raster ) { struct i915_context *i915 = i915_context(pipe); - i915->rasterizer = (struct i915_rasterizer_state *)setup; + i915->rasterizer = (struct i915_rasterizer_state *)raster; /* pass-through to draw module */ - draw_set_rasterizer_state(i915->draw, i915->rasterizer->templ); + draw_set_rasterizer_state(i915->draw, + (i915->rasterizer ? i915->rasterizer->templ : NULL)); i915->dirty |= I915_NEW_RASTERIZER; } static void i915_delete_rasterizer_state(struct pipe_context *pipe, - void *setup) + void *raster) { - FREE(setup); + FREE(raster); } static void i915_set_vertex_buffer( struct pipe_context *pipe, -- cgit v1.2.3 From d91e62e9c6ff81c91b83af883281c1e205c6bf35 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Mar 2008 18:24:46 -0600 Subject: gallium: in clear_stencil_buffer() check surface format to determine stencil clear value ... as we do for the Z and Z+stencil cases --- src/mesa/state_tracker/st_cb_clear.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index eae40f2a4f..cc8a136292 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -544,6 +544,15 @@ clear_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) /* simple clear of whole buffer */ GLuint clearValue = ctx->Stencil.Clear; + + switch (strb->surface->format) { + case PIPE_FORMAT_S8Z24_UNORM: + clearValue <<= 24; + break; + default: + ; /* no-op, stencil value is in least significant bits */ + } + ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } } -- cgit v1.2.3 From 329c5431348117e5b99adf14936d2f57f2ef5f1f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 12 Mar 2008 19:29:30 -0600 Subject: gallium: fix polygon stipple Was broken by commit 4528287e040415c2071012d02f20979ff995c754 (bind all samplers/texures at once). --- src/gallium/auxiliary/draw/draw_pstipple.c | 34 +++++++++++++++++++----------- 1 file changed, 22 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index ed50d0805a..8b3e84a9a0 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -67,6 +67,7 @@ struct pstip_stage struct draw_stage stage; void *sampler_cso; + uint sampler_unit; struct pipe_texture *texture; uint num_samplers; uint num_textures; @@ -76,8 +77,8 @@ struct pstip_stage */ struct pstip_fragment_shader *fs; struct { - void *sampler[PIPE_MAX_SAMPLERS]; - struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + void *samplers[PIPE_MAX_SAMPLERS]; + struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; const struct pipe_poly_stipple *stipple; } state; @@ -328,6 +329,8 @@ generate_pstip_fs(struct pstip_stage *pstip) tgsi_dump(pstip_fs.tokens, 0); #endif + pstip->sampler_unit = transform.maxSampler + 1; + #if 1 /* XXX remove */ if (transform.wincoordInput < 0) { pstip_fs.input_semantic_name[pstip_fs.num_inputs] = TGSI_SEMANTIC_POSITION; @@ -483,17 +486,24 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) { struct pstip_stage *pstip = pstip_stage(stage); struct pipe_context *pipe = pstip->pipe; - uint num = MAX2(pstip->num_textures, pstip->num_samplers); + uint num_samplers; + + /* how many samplers? */ + /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ + num_samplers = MAX2(pstip->num_textures, pstip->num_samplers); + num_samplers = MAX2(num_samplers, pstip->sampler_unit + 1); assert(stage->draw->rasterizer->poly_stipple_enable); - /* - * Bind our fragprog, sampler and texture - */ + /* bind our fragprog */ bind_pstip_fragment_shader(pstip); - pstip->driver_bind_sampler_states(pipe, num + 1, pstip->state.sampler); - pstip->driver_set_sampler_textures(pipe, num + 1, pstip->state.texture); + /* plug in our sampler, texture */ + pstip->state.samplers[pstip->sampler_unit] = pstip->sampler_cso; + pstip->state.textures[pstip->sampler_unit] = pstip->texture; + + pstip->driver_bind_sampler_states(pipe, num_samplers, pstip->state.samplers); + pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures); /* now really draw first line */ stage->tri = passthrough_tri; @@ -516,9 +526,9 @@ pstip_flush(struct draw_stage *stage, unsigned flags) /* XXX restore original texture, sampler state */ pstip->driver_bind_sampler_states(pipe, pstip->num_samplers, - pstip->state.sampler); + pstip->state.samplers); pstip->driver_set_sampler_textures(pipe, pstip->num_textures, - pstip->state.texture); + pstip->state.textures); } @@ -617,7 +627,7 @@ pstip_bind_sampler_states(struct pipe_context *pipe, { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); /* save current */ - memcpy(pstip->state.sampler, sampler, num * sizeof(void *)); + memcpy(pstip->state.samplers, sampler, num * sizeof(void *)); pstip->num_samplers = num; /* pass-through */ pstip->driver_bind_sampler_states(pstip->pipe, num, sampler); @@ -630,7 +640,7 @@ pstip_set_sampler_textures(struct pipe_context *pipe, { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); /* save current */ - memcpy(pstip->state.texture, texture, num * sizeof(struct pipe_texture *)); + memcpy(pstip->state.textures, texture, num * sizeof(struct pipe_texture *)); pstip->num_textures = num; /* pass-through */ pstip->driver_set_sampler_textures(pstip->pipe, num, texture); -- cgit v1.2.3 From a1d56728655a3fc87360b45ac8b348bcfdf6ac15 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 12 Mar 2008 21:42:33 -0400 Subject: document hash collision resolutions --- src/gallium/auxiliary/cso_cache/cso_hash.h | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h index d5bca9d591..84b45a5963 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.h +++ b/src/gallium/auxiliary/cso_cache/cso_hash.h @@ -25,6 +25,16 @@ * **************************************************************************/ +/** + This file provides a hash implementation that is capable of dealing + with collisions. It stores colliding entries in linked list. All + functions operating on the hash return an iterator. The iterator + itself points to the collision list. If there wasn't any collision + the list will have just one entry, otherwise client code should + iterate over the entries to find the exact entry among ones that + had the same key (e.g. memcmp could be used on the data to check + that) +*/ /* * Authors: * Zack Rusin -- cgit v1.2.3 From 2366bb1baf2e9ae5b6ecf19f66ae9e0a4b0d2f36 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 12 Mar 2008 22:06:51 -0400 Subject: Add some basic documentation for gallivm code --- src/gallium/auxiliary/gallivm/gallivm.h | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/gallivm.h b/src/gallium/auxiliary/gallivm/gallivm.h index 57912a952f..b4d6555d2f 100644 --- a/src/gallium/auxiliary/gallivm/gallivm.h +++ b/src/gallium/auxiliary/gallivm/gallivm.h @@ -33,6 +33,16 @@ #ifndef GALLIVM_H #define GALLIVM_H +/* + LLVM representation consists of two stages - layout independent + intermediate representation gallivm_ir and driver specific + gallivm_prog. TGSI is first being translated into gallivm_ir + after that driver can set number of options on gallivm_ir and + have it compiled into gallivm_prog. gallivm_prog can be either + executed (assuming there's LLVM JIT backend for the current + target) or machine code generation can be done (assuming there's + a LLVM code generator for thecurrent target) + */ #if defined __cplusplus extern "C" { #endif -- cgit v1.2.3 From cac037d36d451e8cafbb4a759d0edf9fa8b1ca81 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 12 Mar 2008 22:51:57 -0400 Subject: add code handling dependencies between generated code --- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 96 ++++++++++++++++++++--- src/gallium/auxiliary/gallivm/instructionssoa.h | 6 ++ src/gallium/auxiliary/gallivm/soabuiltins.c | 18 +++++ src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 1 + 4 files changed, 111 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 89d513afd0..6f83b56a72 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -143,8 +143,16 @@ std::vector InstructionsSoa::extractVector(llvm::Value *vector) void InstructionsSoa::createFunctionMap() { - m_functionsMap[TGSI_OPCODE_DP3] = "dp3"; - m_functionsMap[TGSI_OPCODE_DP4] = "dp4"; + m_functionsMap[TGSI_OPCODE_DP3] = "dp3"; + m_functionsMap[TGSI_OPCODE_DP4] = "dp4"; + m_functionsMap[TGSI_OPCODE_POWER] = "pow"; +} + +void InstructionsSoa::createDependencies() +{ + std::vector powDeps(1); + powDeps[0] = "powf"; + m_builtinDependencies["pow"] = powDeps; } llvm::Function * InstructionsSoa::function(int op) @@ -154,15 +162,14 @@ llvm::Function * InstructionsSoa::function(int op) std::string name = m_functionsMap[op]; + std::vector deps = m_builtinDependencies[name]; + for (unsigned int i = 0; i < deps.size(); ++i) { + injectFunction(m_builtins->getFunction(deps[i])); + } + llvm::Function *originalFunc = m_builtins->getFunction(name); - llvm::Function *func = CloneFunction(originalFunc); - currentModule()->getFunctionList().push_back(func); - std::cout << "Func parent is "<getParent() - <<", cur is "<dump(); - //func->setParent(currentModule()); - m_functions[op] = func; - return func; + injectFunction(originalFunc, op); + return m_functions[op]; } llvm::Module * InstructionsSoa::currentModule() const @@ -177,6 +184,7 @@ llvm::Module * InstructionsSoa::currentModule() const void InstructionsSoa::createBuiltins() { m_builtins = createSoaBuiltins(); + createDependencies(); } std::vector InstructionsSoa::dp3(const std::vector in1, @@ -304,3 +312,71 @@ std::vector InstructionsSoa::callBuiltin(llvm::Function *func, const std return allocaToResult(allocaPtr); } + +std::vector InstructionsSoa::pow(const std::vector in1, + const std::vector in2) +{ + llvm::Function *func = function(TGSI_OPCODE_POWER); + return callBuiltin(func, in1, in2); +} + +void checkFunction(Function *func) +{ + for (Function::const_iterator BI = func->begin(), BE = func->end(); + BI != BE; ++BI) { + const BasicBlock &BB = *BI; + for (BasicBlock::const_iterator II = BB.begin(), IE = BB.end(); + II != IE; ++II) { + const Instruction &I = *II; + std::cout<< "Instr = "<setOperand(op, V); + } + } + } +} + +void InstructionsSoa::injectFunction(llvm::Function *originalFunc, int op) +{ + assert(originalFunc); + std::cout << "injecting function originalFunc " <getName() <getFunction(originalFunc->getName()); + if (func) { + m_functions[op] = func; + return; + } + } + llvm::Function *func = 0; + if (originalFunc->isDeclaration()) { + std::cout << "function decleration" <getFunctionType(), GlobalValue::ExternalLinkage, + originalFunc->getName(), currentModule()); + func->setCallingConv(CallingConv::C); + const ParamAttrsList *pal = 0; + func->setParamAttrs(pal); + currentModule()->dump(); + } else { + DenseMap val; + val[m_builtins->getFunction("powf")] = currentModule()->getFunction("powf"); + std::cout <<" replacing "<getFunction("powf") + <<", with " <getFunction("powf")<getFunctionList().push_back(func); + std::cout << "Func parent is "<getParent() + <<", cur is "< #include #include @@ -59,6 +60,8 @@ public: const std::vector in3); std::vector mul(const std::vector in1, const std::vector in2); + std::vector pow(const std::vector in1, + const std::vector in2); void end(); std::vector extractVector(llvm::Value *vector); @@ -68,6 +71,7 @@ private: llvm::Value *z, llvm::Value *w); void createFunctionMap(); void createBuiltins(); + void createDependencies(); llvm::Function *function(int); llvm::Module *currentModule() const; llvm::Value *allocaTemp(); @@ -81,6 +85,7 @@ private: const std::vector in1, const std::vector in2, const std::vector in3); + void injectFunction(llvm::Function *originalFunc, int op = TGSI_OPCODE_LAST); private: llvm::LLVMFoldingBuilder m_builder; StorageSoa *m_storage; @@ -88,6 +93,7 @@ private: std::map m_functionsMap; std::map m_functions; llvm::Module *m_builtins; + std::map > m_builtinDependencies; private: mutable int m_idx; diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index 24c14e1b69..4d658be520 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -60,6 +60,24 @@ void dp4(float4 *res, res[3] = dot; } +extern float powf(float num, float p); + +void pow(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, + float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) +{ + float4 p; + p.x = powf(tmp0x.x, tmp1x.x); + p.y = powf(tmp0x.y, tmp1x.y); + p.z = powf(tmp0x.z, tmp1x.z); + p.w = powf(tmp0x.w, tmp1x.w); + + res[0] = p; + res[1] = p; + res[2] = p; + res[3] = p; +} + #if 0 void yo(float4 *out, float4 *in) { diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index 3f65865a5a..d2b747ffd1 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -806,6 +806,7 @@ translate_instructionir(llvm::Module *module, } break; case TGSI_OPCODE_POWER: { + out = instr->pow(inputs[0], inputs[1]); } break; case TGSI_OPCODE_CROSSPRODUCT: { -- cgit v1.2.3 From 5ba2f0a507b8d413eea2eb7da09c304ab5a8d3f1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Mar 2008 09:57:01 +0000 Subject: tgsi: bump MAX_SRC_REGS to 4, for TXD --- src/gallium/auxiliary/tgsi/util/tgsi_parse.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h index 5ccb5bfcf6..40083728d6 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h @@ -56,7 +56,7 @@ struct tgsi_full_immediate }; #define TGSI_FULL_MAX_DST_REGISTERS 2 -#define TGSI_FULL_MAX_SRC_REGISTERS 3 +#define TGSI_FULL_MAX_SRC_REGISTERS 4 /* TXD has 4 */ struct tgsi_full_instruction { -- cgit v1.2.3 From ddb4e5cbaced3e96117a97fe362ab890794f5ab7 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Mar 2008 10:01:38 +0000 Subject: tgsi: replace erroneous use of FETCH with emit_tempf --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 18 +++++++++++++++--- 1 file changed, 15 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 0da3b0bdb7..4e80597b3f 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -1821,7 +1821,11 @@ emit_instruction( STORE( func, *inst, 5, 0, CHAN_Z ); } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); STORE( func, *inst, 0, 0, CHAN_W ); } break; @@ -2021,11 +2025,19 @@ emit_instruction( STORE( func, *inst, 0, 0, CHAN_Y ); } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - FETCH( func, *inst, 0, TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ); + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ); STORE( func, *inst, 0, 0, CHAN_Z ); } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - FETCH( func, *inst, 0, TGSI_EXEC_TEMP_ONE_I, TGSI_EXEC_TEMP_ONE_C ); + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); STORE( func, *inst, 0, 0, CHAN_W ); } break; -- cgit v1.2.3 From 92523ad0fd11ff532f1e0642410d0a623fe53b06 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Mar 2008 10:14:21 +0000 Subject: mesa: fix (harmless?) assignment in assert --- src/mesa/shader/slang/slang_codegen.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 767ba3ffb4..0de2b0a57e 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1087,7 +1087,7 @@ slang_inline_function_call(slang_assemble_ctx * A, slang_function *fun, slang_operation_copy(inlined, fun->body); /*** XXX review this */ - assert(inlined->type = SLANG_OPER_BLOCK_NO_NEW_SCOPE); + assert(inlined->type == SLANG_OPER_BLOCK_NO_NEW_SCOPE); inlined->type = SLANG_OPER_BLOCK_NEW_SCOPE; #if 0 -- cgit v1.2.3 From 192d1cbbdf9f8e2527ef38761195a87517c2d244 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 12 Mar 2008 22:39:13 +0000 Subject: gallium: Add a new handle_table_set that accepts an arbitrary handle. --- src/gallium/auxiliary/util/u_handle_table.c | 85 ++++++++++++++++++++++------- src/gallium/auxiliary/util/u_handle_table.h | 11 ++++ 2 files changed, 75 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index 8a298f7c41..ab427ee371 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -26,6 +26,7 @@ **************************************************************************/ /** + * @file * Generic handle table implementation. * * @author José Fonseca @@ -90,6 +91,39 @@ handle_table_set_destroy(struct handle_table *ht, } +/** + * Resize the table if necessary + */ +static INLINE int +handle_table_resize(struct handle_table *ht, + unsigned minimum_size) +{ + unsigned new_size; + void **new_objects; + + if(ht->size > minimum_size) + return ht->size; + + new_size = ht->size; + while(!(new_size > minimum_size)) + new_size *= 2; + assert(new_size); + + new_objects = (void **)REALLOC((void *)ht->objects, + ht->size*sizeof(void *), + new_size*sizeof(void *)); + if(!new_objects) + return 0; + + memset(new_objects + ht->size, 0, (new_size - ht->size)*sizeof(void *)); + + ht->size = new_size; + ht->objects = new_objects; + + return ht->size; +} + + unsigned handle_table_add(struct handle_table *ht, void *object) @@ -109,34 +143,17 @@ handle_table_add(struct handle_table *ht, ++ht->filled; } - /* grow the table */ - if(ht->filled == ht->size) { - unsigned new_size; - void **new_objects; - - new_size = ht->size*2; - assert(new_size); - - new_objects = (void **)REALLOC((void *)ht->objects, - ht->size*sizeof(void *), - new_size*sizeof(void *)); - if(!new_objects) - return 0; - - memset(new_objects + ht->size, 0, (new_size - ht->size)*sizeof(void *)); - - ht->size = new_size; - ht->objects = new_objects; - } - index = ht->filled; - handle = index + 1; /* check integer overflow */ if(!handle) return 0; + /* grow the table if necessary */ + if(!handle_table_resize(ht, index)) + return 0; + assert(!ht->objects[index]); ht->objects[index] = object; ++ht->filled; @@ -145,6 +162,32 @@ handle_table_add(struct handle_table *ht, } +unsigned +handle_table_set(struct handle_table *ht, + unsigned handle, + void *object) +{ + unsigned index; + + assert(ht); + assert(handle > 0); + assert(handle <= ht->size); + if(!handle || handle > ht->size) + return 0; + + index = handle - 1; + + /* grow the table if necessary */ + if(!handle_table_resize(ht, index)) + return 0; + + assert(!ht->objects[index]); + ht->objects[index] = object; + + return handle; +} + + void * handle_table_get(struct handle_table *ht, unsigned handle) diff --git a/src/gallium/auxiliary/util/u_handle_table.h b/src/gallium/auxiliary/util/u_handle_table.h index 51fc273865..a2f1f604ad 100644 --- a/src/gallium/auxiliary/util/u_handle_table.h +++ b/src/gallium/auxiliary/util/u_handle_table.h @@ -26,6 +26,7 @@ **************************************************************************/ /** + * @file * Generic handle table. * * @author José Fonseca @@ -42,6 +43,8 @@ extern "C" { /** * Abstract data type to map integer handles to objects. + * + * Also referred as "pointer array". */ struct handle_table; @@ -70,6 +73,14 @@ unsigned handle_table_add(struct handle_table *ht, void *object); +/** + * Returns zero on failure (out of memory). + */ +unsigned +handle_table_set(struct handle_table *ht, + unsigned handle, + void *object); + /** * Fetch an existing object. * -- cgit v1.2.3 From 8506e41dc080e20286709ab93b728aa5162f3c87 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 12 Mar 2008 22:39:44 +0000 Subject: gallium: Fix debug_mask_vprintf's example. --- src/gallium/include/pipe/p_debug.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index f45363f355..f971ad3adc 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -121,7 +121,8 @@ uint32_t debug_mask_get(uint32_t uuid); * @code * #define MYDRIVER_UUID 0x12345678 // random 32 bit identifier * - * static inline mydriver_debug(uint32_t what, const char *format, ...) + * static void inline + * mydriver_debug(uint32_t what, const char *format, ...) * { * #ifdef DEBUG * va_list ap; -- cgit v1.2.3 From e584eb888fac90783c5f62333a39e6735be3e488 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 13 Mar 2008 09:58:29 +0000 Subject: gallium: Add a bit of documentation to cso_hash. --- src/gallium/auxiliary/cso_cache/cso_hash.h | 49 +++++++++++++++++++++--------- 1 file changed, 34 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h index 84b45a5963..a3a65b68c8 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.h +++ b/src/gallium/auxiliary/cso_cache/cso_hash.h @@ -26,19 +26,20 @@ **************************************************************************/ /** - This file provides a hash implementation that is capable of dealing - with collisions. It stores colliding entries in linked list. All - functions operating on the hash return an iterator. The iterator - itself points to the collision list. If there wasn't any collision - the list will have just one entry, otherwise client code should - iterate over the entries to find the exact entry among ones that - had the same key (e.g. memcmp could be used on the data to check - that) -*/ - /* - * Authors: - * Zack Rusin - */ + * @file + * Hash table implementation. + * + * This file provides a hash implementation that is capable of dealing + * with collisions. It stores colliding entries in linked list. All + * functions operating on the hash return an iterator. The iterator + * itself points to the collision list. If there wasn't any collision + * the list will have just one entry, otherwise client code should + * iterate over the entries to find the exact entry among ones that + * had the same key (e.g. memcmp could be used on the data to check + * that) + * + * @author Zack Rusin + */ #ifndef CSO_HASH_H #define CSO_HASH_H @@ -48,24 +49,38 @@ extern "C" { #endif + struct cso_hash; struct cso_node; + struct cso_hash_iter { struct cso_hash *hash; struct cso_node *node; }; + struct cso_hash *cso_hash_create(void); void cso_hash_delete(struct cso_hash *hash); + int cso_hash_size(struct cso_hash *hash); + +/** + * Create a list of objects and just add entry with the same key to the list. + */ struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, void *data); + void *cso_hash_take(struct cso_hash *hash, unsigned key); + struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash); + +/** + * Return an iterator pointing to the first entry in the collision list. + */ struct cso_hash_iter cso_hash_find(struct cso_hash *hash, unsigned key); @@ -73,12 +88,16 @@ int cso_hash_iter_is_null(struct cso_hash_iter iter); unsigned cso_hash_iter_key(struct cso_hash_iter iter); void *cso_hash_iter_data(struct cso_hash_iter iter); + struct cso_hash_iter cso_hash_iter_next(struct cso_hash_iter iter); struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); -/* KW: a convenience routine: - */ +/** + * Convenience routine to iterate over the collision list while doing a memory + * comparison to see which entry in the list is a direct copy of our template + * and returns that entry. + */ void *cso_hash_find_data_from_template( struct cso_hash *hash, unsigned hash_key, void *templ, -- cgit v1.2.3 From 734ccee565efd274e47f95ea314220726a38a512 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 13 Mar 2008 09:59:29 +0000 Subject: gallium: Standardize most important error codes. --- src/gallium/include/pipe/p_error.h | 65 ++++++++++++++++++++++++++++++++++++++ 1 file changed, 65 insertions(+) create mode 100644 src/gallium/include/pipe/p_error.h (limited to 'src') diff --git a/src/gallium/include/pipe/p_error.h b/src/gallium/include/pipe/p_error.h new file mode 100644 index 0000000000..b865b22635 --- /dev/null +++ b/src/gallium/include/pipe/p_error.h @@ -0,0 +1,65 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Gallium error codes. + * + * @author José Fonseca + */ + +#ifndef P_ERROR_H_ +#define P_ERROR_H_ + + +#ifdef __cplusplus +extern "C" { +#endif + + +/** + * Gallium error codes. + * + * - A zero value always means success. + * - A negative value always means failure. + * - The meaning of a positive value is function dependent. + */ +enum pipe_error { + PIPE_OK = 0, + PIPE_ERROR = -1, /**< Generic error */ + PIPE_ERROR_BAD_INPUT = -2, + PIPE_ERROR_OUT_OF_MEMORY = -3, + PIPE_ERROR_RETRY = -4 + /* TODO */ +}; + + +#ifdef __cplusplus +} +#endif + +#endif /* P_ERROR_H_ */ -- cgit v1.2.3 From 42f28684168ff88942f51fb01377702570d9d4d7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 13 Mar 2008 10:19:23 +0000 Subject: gallium: General purpose hash table, which is actually just a convenient frontend to cso_hash. --- src/gallium/auxiliary/util/Makefile | 1 + src/gallium/auxiliary/util/SConscript | 1 + src/gallium/auxiliary/util/u_hash_table.c | 199 ++++++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_hash_table.h | 86 +++++++++++++ 4 files changed, 287 insertions(+) create mode 100644 src/gallium/auxiliary/util/u_hash_table.c create mode 100644 src/gallium/auxiliary/util/u_hash_table.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 2abbe9500e..2016c6fb1f 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -8,6 +8,7 @@ C_SOURCES = \ p_tile.c \ p_util.c \ u_handle_table.c \ + u_hash_table.c \ u_mm.c \ u_snprintf.c diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 2030214aa7..154a3eca8c 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -7,6 +7,7 @@ util = env.ConvenienceLibrary( 'p_tile.c', 'p_util.c', 'u_handle_table.c', + 'u_hash_table.c', 'u_mm.c', 'u_snprintf.c', ]) diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c new file mode 100644 index 0000000000..ac2cb1b540 --- /dev/null +++ b/src/gallium/auxiliary/util/u_hash_table.c @@ -0,0 +1,199 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * General purpose hash table implementation. + * + * Just uses the cso_hash for now, but it might be better switch to a linear + * probing hash table implementation at some point -- as it is said they have + * better lookup and cache performance and it appears to be possible to write + * a lock-free implementation of such hash tables . + * + * @author José Fonseca + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_util.h" + +#include "cso_cache/cso_hash.h" +#include "u_hash_table.h" + + +struct hash_table +{ + struct cso_hash *cso; + + /** Hash function */ + unsigned (*hash)(void *key); + + /** Compare two keys */ + int (*compare)(void *key1, void *key2); + + /* TODO: key, value destructors? */ +}; + + +struct hash_table_item +{ + void *key; + void *value; +}; + + +struct hash_table * +hash_table_create(unsigned (*hash)(void *key), + int (*compare)(void *key1, void *key2)) +{ + struct hash_table *ht; + + ht = MALLOC_STRUCT(hash_table); + if(!ht) + return NULL; + + ht->cso = cso_hash_create(); + if(!ht->cso) { + FREE(ht); + return NULL; + } + + ht->hash = hash; + ht->compare = compare; + + return ht; +} + + +static struct hash_table_item * +hash_table_find_item(struct hash_table *ht, + void *key, + unsigned key_hash) +{ + struct cso_hash_iter iter; + struct hash_table_item *item; + + iter = cso_hash_find(ht->cso, key_hash); + while (!cso_hash_iter_is_null(iter)) { + item = (struct hash_table_item *)cso_hash_iter_data(iter); + if (!ht->compare(item->key, key)) + return item; + iter = cso_hash_iter_next(iter); + } + + return NULL; +} + + +enum pipe_error +hash_table_set(struct hash_table *ht, + void *key, + void *value) +{ + unsigned key_hash; + struct hash_table_item *item; + + assert(ht); + + key_hash = ht->hash(key); + + item = hash_table_find_item(ht, key, key_hash); + if(item) { + /* TODO: key/value destruction? */ + item->value = value; + return PIPE_OK; + } + + item = MALLOC_STRUCT(hash_table_item); + if(!item) + return PIPE_ERROR_OUT_OF_MEMORY; + + item->key = key; + item->value = value; + + cso_hash_insert(ht->cso, key_hash, item); + /* FIXME: there is no OOM propagation in cso_hash */ + if(0) { + FREE(item); + return PIPE_ERROR_OUT_OF_MEMORY; + } + + return PIPE_OK; +} + + +void * +hash_table_get(struct hash_table *ht, + void *key) +{ + unsigned key_hash; + struct hash_table_item *item; + + assert(ht); + + key_hash = ht->hash(key); + + item = hash_table_find_item(ht, key, key_hash); + if(!item) + return NULL; + + return item->value; +} + + +void +hash_table_remove(struct hash_table *ht, + void *key) +{ + unsigned key_hash; + struct hash_table_item *item; + + assert(ht); + + key_hash = ht->hash(key); + + item = hash_table_find_item(ht, key, key_hash); + if(!item) + return; + + /* FIXME: cso_hash_take takes the first element of the collision list + * indiscriminately, so we can not take the item down. */ + item->value = NULL; +} + + +void +hash_table_destroy(struct hash_table *ht) +{ + assert(ht); + + cso_hash_delete(ht->cso); + + FREE(ht); +} + diff --git a/src/gallium/auxiliary/util/u_hash_table.h b/src/gallium/auxiliary/util/u_hash_table.h new file mode 100644 index 0000000000..d941f2c6b1 --- /dev/null +++ b/src/gallium/auxiliary/util/u_hash_table.h @@ -0,0 +1,86 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * General purpose hash table. + * + * @author José Fonseca + */ + +#ifndef U_HASH_TABLE_H_ +#define U_HASH_TABLE_H_ + + +#include "pipe/p_error.h" + + +#ifdef __cplusplus +extern "C" { +#endif + + +/** + * Generic purpose hash table. + */ +struct hash_table; + + +/** + * Create an hash table. + * + * @param hash hash function + * @param compare should return 0 for two equal keys. + */ +struct hash_table * +hash_table_create(unsigned (*hash)(void *key), + int (*compare)(void *key1, void *key2)); + + +enum pipe_error +hash_table_set(struct hash_table *ht, + void *key, + void *value); + +void * +hash_table_get(struct hash_table *ht, + void *key); + + +void +hash_table_remove(struct hash_table *ht, + void *key); + + +void +hash_table_destroy(struct hash_table *ht); + + +#ifdef __cplusplus +} +#endif + +#endif /* U_HASH_TABLE_H_ */ -- cgit v1.2.3 From 78ddfbd1306c113c3276c79712e5d0e2da82d136 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 13 Mar 2008 12:59:48 +0100 Subject: tgsi: Remove OPCODE_TEXCOORD, OPCODE_TEXCRD aliases. --- src/gallium/include/pipe/p_shader_tokens.h | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 210169f54f..e3947f663b 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -341,7 +341,6 @@ struct tgsi_immediate_float32 /* * ps_1_1 */ -#define TGSI_OPCODE_TEXCOORD TGSI_OPCODE_NOP #define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KILP #define TGSI_OPCODE_TEXBEM 107 #define TGSI_OPCODE_TEXBEML 108 @@ -372,7 +371,6 @@ struct tgsi_immediate_float32 /* * ps_1_4 */ -#define TGSI_OPCODE_TEXCRD TGSI_OPCODE_TEXCOORD #define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX #define TGSI_OPCODE_TEXDEPTH 122 #define TGSI_OPCODE_BEM 123 -- cgit v1.2.3 From bcb454e7a6e2f7efae114321c65bf98e91d5892f Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 13 Mar 2008 18:12:36 +0100 Subject: tgsi: Drop pre-ps_2_0 opcodes. --- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 68 ------------- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 46 ++------- src/gallium/include/pipe/p_shader_tokens.h | 147 ++++++++++++--------------- 3 files changed, 71 insertions(+), 190 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index d2b747ffd1..ab9e7a06fb 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -588,40 +588,6 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_NOP: break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; case TGSI_OPCODE_M4X3: break; case TGSI_OPCODE_M3X4: @@ -981,40 +947,6 @@ translate_instructionir(llvm::Module *module, break; case TGSI_OPCODE_NOP: break; - case TGSI_OPCODE_TEXBEM: - break; - case TGSI_OPCODE_TEXBEML: - break; - case TGSI_OPCODE_TEXREG2AR: - break; - case TGSI_OPCODE_TEXM3X2PAD: - break; - case TGSI_OPCODE_TEXM3X2TEX: - break; - case TGSI_OPCODE_TEXM3X3PAD: - break; - case TGSI_OPCODE_TEXM3X3TEX: - break; - case TGSI_OPCODE_TEXM3X3SPEC: - break; - case TGSI_OPCODE_TEXM3X3VSPEC: - break; - case TGSI_OPCODE_TEXREG2GB: - break; - case TGSI_OPCODE_TEXREG2RGB: - break; - case TGSI_OPCODE_TEXDP3TEX: - break; - case TGSI_OPCODE_TEXDP3: - break; - case TGSI_OPCODE_TEXM3X3: - break; - case TGSI_OPCODE_TEXM3X2DEPTH: - break; - case TGSI_OPCODE_TEXDEPTH: - break; - case TGSI_OPCODE_BEM: - break; case TGSI_OPCODE_M4X3: break; case TGSI_OPCODE_M3X4: diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index ceb407b884..aa78278700 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -324,7 +324,7 @@ static const char *TGSI_IMMS_SHORT[] = "FLT32" }; -static const char *TGSI_OPCODES[] = +static const char *TGSI_OPCODES[TGSI_OPCODE_LAST] = { "OPCODE_ARL", "OPCODE_MOV", @@ -380,6 +380,7 @@ static const char *TGSI_OPCODES[] = "OPCODE_STR", "OPCODE_TEX", "OPCODE_TXD", + "OPCODE_TXP", "OPCODE_UP2H", "OPCODE_UP2US", "OPCODE_UP4B", @@ -433,23 +434,6 @@ static const char *TGSI_OPCODES[] = "OPCODE_NOISE3", "OPCODE_NOISE4", "OPCODE_NOP", - "OPCODE_TEXBEM", - "OPCODE_TEXBEML", - "OPCODE_TEXREG2AR", - "OPCODE_TEXM3X2PAD", - "OPCODE_TEXM3X2TEX", - "OPCODE_TEXM3X3PAD", - "OPCODE_TEXM3X3TEX", - "OPCODE_TEXM3X3SPEC", - "OPCODE_TEXM3X3VSPEC", - "OPCODE_TEXREG2GB", - "OPCODE_TEXREG2RGB", - "OPCODE_TEXDP3TEX", - "OPCODE_TEXDP3", - "OPCODE_TEXM3X3", - "OPCODE_TEXM3X2DEPTH", - "OPCODE_TEXDEPTH", - "OPCODE_BEM", "OPCODE_M4X3", "OPCODE_M3X4", "OPCODE_M3X3", @@ -459,11 +443,10 @@ static const char *TGSI_OPCODES[] = "OPCODE_IFC", "OPCODE_BREAKC", "OPCODE_KIL", - "OPCODE_END", - "OPCODE_TXP" + "OPCODE_END" }; -static const char *TGSI_OPCODES_SHORT[] = +static const char *TGSI_OPCODES_SHORT[TGSI_OPCODE_LAST] = { "ARL", "MOV", @@ -519,6 +502,7 @@ static const char *TGSI_OPCODES_SHORT[] = "STR", "TEX", "TXD", + "TXP", "UP2H", "UP2US", "UP4B", @@ -572,23 +556,6 @@ static const char *TGSI_OPCODES_SHORT[] = "NOISE3", "NOISE4", "NOP", - "TEXBEM", - "TEXBEML", - "TEXREG2AR", - "TEXM3X2PAD", - "TEXM3X2TEX", - "TEXM3X3PAD", - "TEXM3X3TEX", - "TEXM3X3SPEC", - "TEXM3X3VSPEC", - "TEXREG2GB", - "TEXREG2RGB", - "TEXDP3TEX", - "TEXDP3", - "TEXM3X3", - "TEXM3X2DEPTH", - "TEXDEPTH", - "BEM", "M4X3", "M3X4", "M3X3", @@ -598,8 +565,7 @@ static const char *TGSI_OPCODES_SHORT[] = "IFC", "BREAKC", "KIL", - "END", - "TXP" + "END" }; static const char *TGSI_SATS[] = diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index e3947f663b..f05e1a34b3 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -227,22 +227,22 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_STR 51 #define TGSI_OPCODE_TEX 52 #define TGSI_OPCODE_TXD 53 -#define TGSI_OPCODE_TXP 134 -#define TGSI_OPCODE_UP2H 54 -#define TGSI_OPCODE_UP2US 55 -#define TGSI_OPCODE_UP4B 56 -#define TGSI_OPCODE_UP4UB 57 -#define TGSI_OPCODE_X2D 58 +#define TGSI_OPCODE_TXP 54 +#define TGSI_OPCODE_UP2H 55 +#define TGSI_OPCODE_UP2US 56 +#define TGSI_OPCODE_UP4B 57 +#define TGSI_OPCODE_UP4UB 58 +#define TGSI_OPCODE_X2D 59 /* * GL_NV_vertex_program2 */ -#define TGSI_OPCODE_ARA 59 -#define TGSI_OPCODE_ARR 60 -#define TGSI_OPCODE_BRA 61 -#define TGSI_OPCODE_CAL 62 -#define TGSI_OPCODE_RET 63 -#define TGSI_OPCODE_SSG 64 +#define TGSI_OPCODE_ARA 60 +#define TGSI_OPCODE_ARR 61 +#define TGSI_OPCODE_BRA 62 +#define TGSI_OPCODE_CAL 63 +#define TGSI_OPCODE_RET 64 +#define TGSI_OPCODE_SSG 65 /* * GL_ARB_vertex_program @@ -253,9 +253,9 @@ struct tgsi_immediate_float32 /* * GL_ARB_fragment_program */ -#define TGSI_OPCODE_CMP 65 -#define TGSI_OPCODE_SCS 66 -#define TGSI_OPCODE_TXB 67 +#define TGSI_OPCODE_CMP 66 +#define TGSI_OPCODE_SCS 67 +#define TGSI_OPCODE_TXB 68 /* * GL_NV_fragment_program_option @@ -265,19 +265,19 @@ struct tgsi_immediate_float32 /* * GL_NV_fragment_program2 */ -#define TGSI_OPCODE_NRM 68 -#define TGSI_OPCODE_DIV 69 -#define TGSI_OPCODE_DP2 70 +#define TGSI_OPCODE_NRM 69 +#define TGSI_OPCODE_DIV 70 +#define TGSI_OPCODE_DP2 71 #define TGSI_OPCODE_DP2A TGSI_OPCODE_DOT2ADD -#define TGSI_OPCODE_TXL 71 -#define TGSI_OPCODE_BRK 72 -#define TGSI_OPCODE_IF 73 -#define TGSI_OPCODE_LOOP 74 -#define TGSI_OPCODE_REP 75 -#define TGSI_OPCODE_ELSE 76 -#define TGSI_OPCODE_ENDIF 77 -#define TGSI_OPCODE_ENDLOOP 78 -#define TGSI_OPCODE_ENDREP 79 +#define TGSI_OPCODE_TXL 72 +#define TGSI_OPCODE_BRK 73 +#define TGSI_OPCODE_IF 74 +#define TGSI_OPCODE_LOOP 75 +#define TGSI_OPCODE_REP 76 +#define TGSI_OPCODE_ELSE 77 +#define TGSI_OPCODE_ENDIF 78 +#define TGSI_OPCODE_ENDLOOP 79 +#define TGSI_OPCODE_ENDREP 80 /* * GL_NV_vertex_program2_option @@ -286,26 +286,26 @@ struct tgsi_immediate_float32 /* * GL_NV_vertex_program3 */ -#define TGSI_OPCODE_PUSHA 80 -#define TGSI_OPCODE_POPA 81 +#define TGSI_OPCODE_PUSHA 81 +#define TGSI_OPCODE_POPA 82 /* * GL_NV_gpu_program4 */ -#define TGSI_OPCODE_CEIL 82 -#define TGSI_OPCODE_I2F 83 -#define TGSI_OPCODE_NOT 84 -#define TGSI_OPCODE_TRUNC 85 -#define TGSI_OPCODE_SHL 86 -#define TGSI_OPCODE_SHR 87 -#define TGSI_OPCODE_AND 88 -#define TGSI_OPCODE_OR 89 -#define TGSI_OPCODE_MOD 90 -#define TGSI_OPCODE_XOR 91 -#define TGSI_OPCODE_SAD 92 -#define TGSI_OPCODE_TXF 93 -#define TGSI_OPCODE_TXQ 94 -#define TGSI_OPCODE_CONT 95 +#define TGSI_OPCODE_CEIL 83 +#define TGSI_OPCODE_I2F 84 +#define TGSI_OPCODE_NOT 85 +#define TGSI_OPCODE_TRUNC 86 +#define TGSI_OPCODE_SHL 87 +#define TGSI_OPCODE_SHR 88 +#define TGSI_OPCODE_AND 89 +#define TGSI_OPCODE_OR 90 +#define TGSI_OPCODE_MOD 91 +#define TGSI_OPCODE_XOR 92 +#define TGSI_OPCODE_SAD 93 +#define TGSI_OPCODE_TXF 94 +#define TGSI_OPCODE_TXQ 95 +#define TGSI_OPCODE_CONT 96 /* * GL_NV_vertex_program4 @@ -321,70 +321,53 @@ struct tgsi_immediate_float32 * GL_NV_geometry_program4 */ /* Same as GL_NV_gpu_program4 */ -#define TGSI_OPCODE_EMIT 96 -#define TGSI_OPCODE_ENDPRIM 97 +#define TGSI_OPCODE_EMIT 97 +#define TGSI_OPCODE_ENDPRIM 98 /* * GLSL */ -#define TGSI_OPCODE_BGNLOOP2 98 -#define TGSI_OPCODE_BGNSUB 99 -#define TGSI_OPCODE_ENDLOOP2 100 -#define TGSI_OPCODE_ENDSUB 101 +#define TGSI_OPCODE_BGNLOOP2 99 +#define TGSI_OPCODE_BGNSUB 100 +#define TGSI_OPCODE_ENDLOOP2 101 +#define TGSI_OPCODE_ENDSUB 102 #define TGSI_OPCODE_INT TGSI_OPCODE_TRUNC -#define TGSI_OPCODE_NOISE1 102 -#define TGSI_OPCODE_NOISE2 103 -#define TGSI_OPCODE_NOISE3 104 -#define TGSI_OPCODE_NOISE4 105 -#define TGSI_OPCODE_NOP 106 +#define TGSI_OPCODE_NOISE1 103 +#define TGSI_OPCODE_NOISE2 104 +#define TGSI_OPCODE_NOISE3 105 +#define TGSI_OPCODE_NOISE4 106 +#define TGSI_OPCODE_NOP 107 /* * ps_1_1 */ #define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KILP -#define TGSI_OPCODE_TEXBEM 107 -#define TGSI_OPCODE_TEXBEML 108 -#define TGSI_OPCODE_TEXREG2AR 109 -#define TGSI_OPCODE_TEXM3X2PAD 110 -#define TGSI_OPCODE_TEXM3X2TEX 111 -#define TGSI_OPCODE_TEXM3X3PAD 112 -#define TGSI_OPCODE_TEXM3X3TEX 113 -#define TGSI_OPCODE_TEXM3X3SPEC 114 -#define TGSI_OPCODE_TEXM3X3VSPEC 115 /* * ps_1_2 */ -#define TGSI_OPCODE_TEXREG2GB 116 -#define TGSI_OPCODE_TEXREG2RGB 117 -#define TGSI_OPCODE_TEXDP3TEX 118 -#define TGSI_OPCODE_TEXDP3 119 -#define TGSI_OPCODE_TEXM3X3 120 /* CMP - use TGSI_OPCODE_CND0 */ /* * ps_1_3 */ -#define TGSI_OPCODE_TEXM3X2DEPTH 121 /* CMP - use TGSI_OPCODE_CND0 */ /* * ps_1_4 */ #define TGSI_OPCODE_TEXLD TGSI_OPCODE_TEX -#define TGSI_OPCODE_TEXDEPTH 122 -#define TGSI_OPCODE_BEM 123 /* * ps_2_0 */ #define TGSI_OPCODE_M4X4 TGSI_OPCODE_MULTIPLYMATRIX -#define TGSI_OPCODE_M4X3 124 -#define TGSI_OPCODE_M3X4 125 -#define TGSI_OPCODE_M3X3 126 -#define TGSI_OPCODE_M3X2 127 +#define TGSI_OPCODE_M4X3 108 +#define TGSI_OPCODE_M3X4 109 +#define TGSI_OPCODE_M3X3 110 +#define TGSI_OPCODE_M3X2 111 #define TGSI_OPCODE_CRS TGSI_OPCODE_XPD -#define TGSI_OPCODE_NRM4 128 +#define TGSI_OPCODE_NRM4 112 #define TGSI_OPCODE_SINCOS TGSI_OPCODE_SCS #define TGSI_OPCODE_TEXLDB TGSI_OPCODE_TXB #define TGSI_OPCODE_DP2ADD TGSI_OPCODE_DP2A @@ -393,10 +376,10 @@ struct tgsi_immediate_float32 * ps_2_x */ #define TGSI_OPCODE_CALL TGSI_OPCODE_CAL -#define TGSI_OPCODE_CALLNZ 129 -#define TGSI_OPCODE_IFC 130 +#define TGSI_OPCODE_CALLNZ 113 +#define TGSI_OPCODE_IFC 114 #define TGSI_OPCODE_BREAK TGSI_OPCODE_BRK -#define TGSI_OPCODE_BREAKC 131 +#define TGSI_OPCODE_BREAKC 115 #define TGSI_OPCODE_DSX TGSI_OPCODE_DDX #define TGSI_OPCODE_DSY TGSI_OPCODE_DDY #define TGSI_OPCODE_TEXLDD TGSI_OPCODE_TXD @@ -417,10 +400,10 @@ struct tgsi_immediate_float32 * vs_2_x */ -#define TGSI_OPCODE_KIL 132 /* unpredicated kill */ -#define TGSI_OPCODE_END 133 /* aka HALT */ +#define TGSI_OPCODE_KIL 116 /* unpredicated kill */ +#define TGSI_OPCODE_END 117 /* aka HALT */ -#define TGSI_OPCODE_LAST 135 +#define TGSI_OPCODE_LAST 118 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ -- cgit v1.2.3 From b0d5519b449feda7b048bc59d4fede54e43f5ae1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 13 Mar 2008 15:23:04 +0000 Subject: gallium: make the windows config function more readable with 80-ish columns --- src/gallium/auxiliary/util/p_debug.c | 30 ++++++++++++++++++++---------- 1 file changed, 20 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 09cabdae25..bb84e8096d 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -109,20 +109,30 @@ enum { /* Check for aborts enabled. */ static unsigned abort_en() { - if (!mapped_config_file) - { - /* Open an 8 byte file for configuration data. */ - mapped_config_file = EngMapFile(L"\\??\\c:\\gaDebug.cfg", 8, &debug_config_file); - } - /* An value of "0" (ascii) in the configuration file will clear the first 8 bits in the test byte. */ - /* An value of "1" (ascii) in the configuration file will set the first bit in the test byte. */ - /* An value of "2" (ascii) in the configuration file will set the second bit in the test byte. */ - return ((((char *)mapped_config_file)[0]) - 0x30) & eAssertAbortEn; + if (!mapped_config_file) + { + /* Open an 8 byte file for configuration data. */ + mapped_config_file = EngMapFile(L"\\??\\c:\\gaDebug.cfg", 8, &debug_config_file); + } + + /* A value of "0" (ascii) in the configuration file will clear the + * first 8 bits in the test byte. + * + * A value of "1" (ascii) in the configuration file will set the + * first bit in the test byte. + * + * A value of "2" (ascii) in the configuration file will set the + * second bit in the test byte. + * + * Currently the only interesting values are 0 and 1, which clear + * and set abort-on-assert behaviour respectively. + */ + return ((((char *)mapped_config_file)[0]) - 0x30) & eAssertAbortEn; } #else /* WIN32 */ static unsigned abort_en() { - return !GETENV("GALLIUM_ABORT_ON_ASSERT"); + return !GETENV("GALLIUM_ABORT_ON_ASSERT"); } #endif -- cgit v1.2.3 From fa9e7e9a8debb68611909ac2ffab527c6c39a3e5 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 13 Mar 2008 18:08:18 +0000 Subject: gallium: remove semantic info from pipe_shader_state Brian's patch to clean up the shader interfaces. --- src/gallium/auxiliary/draw/draw_aaline.c | 2 +- src/gallium/auxiliary/draw/draw_aapoint.c | 2 +- src/gallium/auxiliary/draw/draw_pstipple.c | 2 +- src/gallium/include/pipe/p_state.h | 2 + src/mesa/state_tracker/st_debug.c | 2 + src/mesa/state_tracker/st_draw.c | 6 +- src/mesa/state_tracker/st_program.c | 167 ++++++++++++++++------------- src/mesa/state_tracker/st_program.h | 2 + 8 files changed, 104 insertions(+), 81 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index 6b1e640ae9..bed709bad0 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -340,7 +340,7 @@ generate_aaline_fs(struct aaline_stage *aaline) tgsi_dump(aaline_fs.tokens, 0); #endif -#if 1 /* XXX remove */ +#if 0 /* XXX remove */ aaline_fs.input_semantic_name[aaline_fs.num_inputs] = TGSI_SEMANTIC_GENERIC; aaline_fs.input_semantic_index[aaline_fs.num_inputs] = transform.maxGeneric + 1; aaline_fs.num_inputs++; diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index 99e9e9fe34..4bc2b5f3e2 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -514,7 +514,7 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) tgsi_dump(aapoint_fs.tokens, 0); #endif -#if 1 /* XXX remove */ +#if 0 /* XXX remove */ aapoint_fs.input_semantic_name[aapoint_fs.num_inputs] = TGSI_SEMANTIC_GENERIC; aapoint_fs.input_semantic_index[aapoint_fs.num_inputs] = transform.maxGeneric + 1; aapoint_fs.num_inputs++; diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index 8b3e84a9a0..d7475b457c 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -331,7 +331,7 @@ generate_pstip_fs(struct pstip_stage *pstip) pstip->sampler_unit = transform.maxSampler + 1; -#if 1 /* XXX remove */ +#if 0 /* XXX remove */ if (transform.wincoordInput < 0) { pstip_fs.input_semantic_name[pstip_fs.num_inputs] = TGSI_SEMANTIC_POSITION; pstip_fs.input_semantic_index[pstip_fs.num_inputs] = (ubyte)transform.maxInput; diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index e338a27383..bb5fab5d5f 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -162,6 +162,7 @@ struct pipe_constant_buffer struct pipe_shader_state { const struct tgsi_token *tokens; +#if 0 /* XXX these are going away */ ubyte num_inputs; ubyte num_outputs; @@ -169,6 +170,7 @@ struct pipe_shader_state ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; +#endif }; diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index 9c13010da8..8b9b91aa34 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -52,10 +52,12 @@ st_print_current(void) struct st_context *st = ctx->st; int i; +#if 0 printf("Vertex Transform Inputs:\n"); for (i = 0; i < st->vp->state.num_inputs; i++) { printf(" Slot %d: VERT_ATTRIB_%d\n", i, st->vp->index_to_input[i]); } +#endif tgsi_dump( st->vp->state.tokens, 0 ); if (st->vp->Base.Base.Parameters) diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 98504e46c1..5ebd0bbcce 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -223,7 +223,7 @@ st_draw_vbo(GLcontext *ctx, /* loop over TGSI shader inputs to determine vertex buffer * and attribute info */ - for (attr = 0; attr < vs->num_inputs; attr++) { + for (attr = 0; attr < /*vs*/vp->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; struct pipe_vertex_element velement; @@ -333,7 +333,7 @@ st_draw_vbo(GLcontext *ctx, } /* unreference buffers (frees wrapped user-space buffer objects) */ - for (attr = 0; attr < vs->num_inputs; attr++) { + for (attr = 0; attr < /*vs*/vp->num_inputs; attr++) { pipe_buffer_reference(winsys, &vbuffer[attr].buffer, NULL); assert(!vbuffer[attr].buffer); pipe->set_vertex_buffer(pipe, attr, &vbuffer[attr]); @@ -502,7 +502,7 @@ st_feedback_draw_vbo(GLcontext *ctx, /* loop over TGSI shader inputs to determine vertex buffer * and attribute info */ - for (attr = 0; attr < vs->num_inputs; attr++) { + for (attr = 0; attr < /*vs*/vp->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; struct pipe_vertex_element velement; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 0f8784e132..d9d11ee0e8 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -85,6 +85,14 @@ st_translate_vertex_program(struct st_context *st, GLuint num_generic = 0; GLuint num_tokens; + ubyte vs_input_semantic_name[PIPE_MAX_SHADER_INPUTS]; + ubyte vs_input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + uint vs_num_inputs = 0; + + ubyte vs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; + ubyte vs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; + uint vs_num_outputs = 0; + memset(&vs, 0, sizeof(vs)); /* @@ -93,36 +101,36 @@ st_translate_vertex_program(struct st_context *st, */ for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { if (stvp->Base.Base.InputsRead & (1 << attr)) { - const GLuint slot = vs.num_inputs; + const GLuint slot = vs_num_inputs; - vs.num_inputs++; + vs_num_inputs++; stvp->input_to_index[attr] = slot; stvp->index_to_input[slot] = attr; switch (attr) { case VERT_ATTRIB_POS: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - vs.input_semantic_index[slot] = 0; + vs_input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + vs_input_semantic_index[slot] = 0; break; case VERT_ATTRIB_WEIGHT: /* fall-through */ case VERT_ATTRIB_NORMAL: /* just label as a generic */ - vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[slot] = 0; + vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs_input_semantic_index[slot] = 0; break; case VERT_ATTRIB_COLOR0: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs.input_semantic_index[slot] = 0; + vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs_input_semantic_index[slot] = 0; break; case VERT_ATTRIB_COLOR1: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs.input_semantic_index[slot] = 1; + vs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs_input_semantic_index[slot] = 1; break; case VERT_ATTRIB_FOG: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_FOG; - vs.input_semantic_index[slot] = 0; + vs_input_semantic_name[slot] = TGSI_SEMANTIC_FOG; + vs_input_semantic_index[slot] = 0; break; case VERT_ATTRIB_TEX0: case VERT_ATTRIB_TEX1: @@ -132,8 +140,8 @@ st_translate_vertex_program(struct st_context *st, case VERT_ATTRIB_TEX5: case VERT_ATTRIB_TEX6: case VERT_ATTRIB_TEX7: - vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[slot] = num_generic++; + vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs_input_semantic_index[slot] = num_generic++; break; case VERT_ATTRIB_GENERIC0: case VERT_ATTRIB_GENERIC1: @@ -144,8 +152,8 @@ st_translate_vertex_program(struct st_context *st, case VERT_ATTRIB_GENERIC6: case VERT_ATTRIB_GENERIC7: assert(attr < VERT_ATTRIB_MAX); - vs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.input_semantic_index[slot] = num_generic++; + vs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs_input_semantic_index[slot] = num_generic++; break; default: assert(0); @@ -155,8 +163,8 @@ st_translate_vertex_program(struct st_context *st, /* initialize output semantics to defaults */ for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) { - vs.output_semantic_name[i] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[i] = 0; + vs_output_semantic_name[i] = TGSI_SEMANTIC_GENERIC; + vs_output_semantic_index[i] = 0; } num_generic = 0; @@ -173,8 +181,8 @@ st_translate_vertex_program(struct st_context *st, assert(slot != ~0); } else { - slot = vs.num_outputs; - vs.num_outputs++; + slot = vs_num_outputs; + vs_num_outputs++; defaultOutputMapping[attr] = slot; } @@ -185,32 +193,32 @@ st_translate_vertex_program(struct st_context *st, switch (attr) { case VERT_RESULT_HPOS: assert(slot == 0); - vs.output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - vs.output_semantic_index[slot] = 0; + vs_output_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + vs_output_semantic_index[slot] = 0; break; case VERT_RESULT_COL0: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs.output_semantic_index[slot] = 0; + vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs_output_semantic_index[slot] = 0; break; case VERT_RESULT_COL1: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - vs.output_semantic_index[slot] = 1; + vs_output_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + vs_output_semantic_index[slot] = 1; break; case VERT_RESULT_BFC0: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; - vs.output_semantic_index[slot] = 0; + vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + vs_output_semantic_index[slot] = 0; break; case VERT_RESULT_BFC1: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; - vs.output_semantic_index[slot] = 1; + vs_output_semantic_name[slot] = TGSI_SEMANTIC_BCOLOR; + vs_output_semantic_index[slot] = 1; break; case VERT_RESULT_FOGC: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_FOG; - vs.output_semantic_index[slot] = 0; + vs_output_semantic_name[slot] = TGSI_SEMANTIC_FOG; + vs_output_semantic_index[slot] = 0; break; case VERT_RESULT_PSIZ: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; - vs.output_semantic_index[slot] = 0; + vs_output_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; + vs_output_semantic_index[slot] = 0; break; case VERT_RESULT_EDGE: assert(0); @@ -223,30 +231,30 @@ st_translate_vertex_program(struct st_context *st, case VERT_RESULT_TEX5: case VERT_RESULT_TEX6: case VERT_RESULT_TEX7: - vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[slot] = num_generic++; + vs_output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs_output_semantic_index[slot] = num_generic++; break; case VERT_RESULT_VAR0: /* fall-through */ default: assert(attr - VERT_RESULT_VAR0 < MAX_VARYING); - vs.output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs.output_semantic_index[slot] = num_generic++; + vs_output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs_output_semantic_index[slot] = num_generic++; } } } - assert(vs.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); + assert(vs_output_semantic_name[0] == TGSI_SEMANTIC_POSITION); if (outputMapping) { - /* find max output slot referenced to compute vs.num_outputs */ + /* find max output slot referenced to compute vs_num_outputs */ GLuint maxSlot = 0; for (attr = 0; attr < VERT_RESULT_MAX; attr++) { if (outputMapping[attr] != ~0 && outputMapping[attr] > maxSlot) maxSlot = outputMapping[attr]; } - vs.num_outputs = maxSlot + 1; + vs_num_outputs = maxSlot + 1; } else { outputMapping = defaultOutputMapping; @@ -257,22 +265,23 @@ st_translate_vertex_program(struct st_context *st, num_tokens = tgsi_translate_mesa_program( TGSI_PROCESSOR_VERTEX, &stvp->Base.Base, /* inputs */ - vs.num_inputs, + vs_num_inputs, stvp->input_to_index, - vs.input_semantic_name, - vs.input_semantic_index, + vs_input_semantic_name, + vs_input_semantic_index, NULL, /* outputs */ - vs.num_outputs, + vs_num_outputs, outputMapping, - vs.output_semantic_name, - vs.output_semantic_index, + vs_output_semantic_name, + vs_output_semantic_index, /* tokenized result */ tokens, ST_MAX_SHADER_TOKENS); vs.tokens = (struct tgsi_token *) mem_dup(tokens, num_tokens * sizeof(tokens[0])); + stvp->num_inputs = vs_num_inputs; stvp->state = vs; /* struct copy */ stvp->driver_shader = pipe->create_vs_state(pipe, &vs); @@ -310,6 +319,14 @@ st_translate_fragment_program(struct st_context *st, GLuint num_generic = 0; GLuint num_tokens; + ubyte fs_input_semantic_name[PIPE_MAX_SHADER_INPUTS]; + ubyte fs_input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + uint fs_num_inputs = 0; + + ubyte fs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; + ubyte fs_output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; + uint fs_num_outputs = 0; + memset(&fs, 0, sizeof(fs)); /* which vertex output goes to the first fragment input: */ @@ -323,33 +340,33 @@ st_translate_fragment_program(struct st_context *st, */ for (attr = 0; attr < FRAG_ATTRIB_MAX; attr++) { if (inputsRead & (1 << attr)) { - const GLuint slot = fs.num_inputs; + const GLuint slot = fs_num_inputs; defaultInputMapping[attr] = slot; stfp->input_map[slot] = vslot++; - fs.num_inputs++; + fs_num_inputs++; switch (attr) { case FRAG_ATTRIB_WPOS: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - fs.input_semantic_index[slot] = 0; + fs_input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + fs_input_semantic_index[slot] = 0; interpMode[slot] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_COL0: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - fs.input_semantic_index[slot] = 0; + fs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + fs_input_semantic_index[slot] = 0; interpMode[slot] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_COL1: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - fs.input_semantic_index[slot] = 1; + fs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + fs_input_semantic_index[slot] = 1; interpMode[slot] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_FOGC: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_FOG; - fs.input_semantic_index[slot] = 0; + fs_input_semantic_name[slot] = TGSI_SEMANTIC_FOG; + fs_input_semantic_index[slot] = 0; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; case FRAG_ATTRIB_TEX0: @@ -360,15 +377,15 @@ st_translate_fragment_program(struct st_context *st, case FRAG_ATTRIB_TEX5: case FRAG_ATTRIB_TEX6: case FRAG_ATTRIB_TEX7: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - fs.input_semantic_index[slot] = num_generic++; + fs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + fs_input_semantic_index[slot] = num_generic++; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; case FRAG_ATTRIB_VAR0: /* fall-through */ default: - fs.input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - fs.input_semantic_index[slot] = num_generic++; + fs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + fs_input_semantic_index[slot] = num_generic++; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; } } @@ -383,10 +400,10 @@ st_translate_fragment_program(struct st_context *st, /* if z is written, emit that first */ if (outputsWritten & (1 << FRAG_RESULT_DEPR)) { - fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_POSITION; - fs.output_semantic_index[fs.num_outputs] = 0; - outputMapping[FRAG_RESULT_DEPR] = fs.num_outputs; - fs.num_outputs++; + fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_POSITION; + fs_output_semantic_index[fs_num_outputs] = 0; + outputMapping[FRAG_RESULT_DEPR] = fs_num_outputs; + fs_num_outputs++; outputsWritten &= ~(1 << FRAG_RESULT_DEPR); } @@ -399,15 +416,15 @@ st_translate_fragment_program(struct st_context *st, assert(0); break; case FRAG_RESULT_COLR: - fs.output_semantic_name[fs.num_outputs] = TGSI_SEMANTIC_COLOR; - fs.output_semantic_index[fs.num_outputs] = numColors; - outputMapping[attr] = fs.num_outputs; + fs_output_semantic_name[fs_num_outputs] = TGSI_SEMANTIC_COLOR; + fs_output_semantic_index[fs_num_outputs] = numColors; + outputMapping[attr] = fs_num_outputs; numColors++; break; default: assert(0); } - fs.num_outputs++; + fs_num_outputs++; } } } @@ -420,16 +437,16 @@ st_translate_fragment_program(struct st_context *st, num_tokens = tgsi_translate_mesa_program( TGSI_PROCESSOR_FRAGMENT, &stfp->Base.Base, /* inputs */ - fs.num_inputs, + fs_num_inputs, inputMapping, - fs.input_semantic_name, - fs.input_semantic_index, + fs_input_semantic_name, + fs_input_semantic_index, interpMode, /* outputs */ - fs.num_outputs, + fs_num_outputs, outputMapping, - fs.output_semantic_name, - fs.output_semantic_index, + fs_output_semantic_name, + fs_output_semantic_index, /* tokenized result */ tokens, ST_MAX_SHADER_TOKENS); diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 78786dcbb6..9ef2a07eaa 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -85,6 +85,8 @@ struct st_vertex_program /** maps a TGSI input index back to a Mesa VERT_ATTRIB_x */ GLuint index_to_input[PIPE_MAX_SHADER_INPUTS]; + GLuint num_inputs; + struct pipe_shader_state state; struct pipe_shader_state *driver_shader; -- cgit v1.2.3 From 647213804582a6b6cfd4fbfeb1c9874ef53307f3 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 13 Mar 2008 11:19:50 -0700 Subject: Replicate TXP changes in the SPU version of TGSI exec Replicate changes from commit ba75e82b6ebaf88dd2e4a8f764b2d296d715bf8a in spu_exec.c --- src/gallium/drivers/cell/spu/spu_exec.c | 45 ++++++++++----------------------- 1 file changed, 14 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 1560c0f157..061fbebf61 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -672,7 +672,7 @@ fetch_texel( struct spu_sampler *sampler, static void exec_tex(struct spu_exec_machine *mach, const struct tgsi_full_instruction *inst, - boolean biasLod) + boolean biasLod, boolean projected) { const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; union spu_exec_channel r[8]; @@ -686,17 +686,9 @@ exec_tex(struct spu_exec_machine *mach, FETCH(&r[0], 0, CHAN_X); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: + if (projected) { FETCH(&r[1], 0, CHAN_W); r[0].q = micro_div(r[0].q, r[1].q); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); } if (biasLod) { @@ -718,19 +710,11 @@ exec_tex(struct spu_exec_machine *mach, FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 0, CHAN_Z); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: + if (projected) { FETCH(&r[3], 0, CHAN_W); r[0].q = micro_div(r[0].q, r[3].q); r[1].q = micro_div(r[1].q, r[3].q); r[2].q = micro_div(r[2].q, r[3].q); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); } if (biasLod) { @@ -752,19 +736,11 @@ exec_tex(struct spu_exec_machine *mach, FETCH(&r[1], 0, CHAN_Y); FETCH(&r[2], 0, CHAN_Z); - switch (inst->FullSrcRegisters[0].SrcRegisterExtSwz.ExtDivide) { - case TGSI_EXTSWIZZLE_W: + if (projected) { FETCH(&r[3], 0, CHAN_W); r[0].q = micro_div(r[0].q, r[3].q); r[1].q = micro_div(r[1].q, r[3].q); r[2].q = micro_div(r[2].q, r[3].q); - break; - - case TGSI_EXTSWIZZLE_ONE: - break; - - default: - assert (0); } if (biasLod) { @@ -1450,14 +1426,14 @@ exec_instruction( /* simple texture lookup */ /* src[0] = texcoord */ /* src[1] = sampler unit */ - exec_tex(mach, inst, FALSE); + exec_tex(mach, inst, FALSE, FALSE); break; case TGSI_OPCODE_TXB: /* Texture lookup with lod bias */ /* src[0] = texcoord (src[0].w = load bias) */ /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); + exec_tex(mach, inst, TRUE, FALSE); break; case TGSI_OPCODE_TXD: @@ -1473,7 +1449,14 @@ exec_instruction( /* Texture lookup with explit LOD */ /* src[0] = texcoord (src[0].w = load bias) */ /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE); + exec_tex(mach, inst, TRUE, FALSE); + break; + + case TGSI_OPCODE_TXP: + /* Texture lookup with projection + /* src[0] = texcoord (src[0].w = projection) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE, TRUE); break; case TGSI_OPCODE_UP2H: -- cgit v1.2.3 From cb294542bcaca7d9847067ce502a68fd8e92f42e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 12 Mar 2008 16:28:33 -0600 Subject: gallium: remove a debug printf --- src/gallium/auxiliary/cso_cache/cso_context.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index f7f4aebb16..fd86bfaca9 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -102,8 +102,6 @@ static void cso_release_all( struct cso_context *ctx ) void cso_destroy_context( struct cso_context *ctx ) { - debug_printf("%s\n", __FUNCTION__); - if (ctx) cso_release_all( ctx ); -- cgit v1.2.3 From 7ffbaebce1da7fec36a38f424f266806a3a0fc6a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 13 Mar 2008 13:22:24 -0600 Subject: gallium: fix bug in stip_first_tri() Need to compute num_samplers after binding/creating the fragment shader. --- src/gallium/auxiliary/draw/draw_pstipple.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index d7475b457c..08f06bf222 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -488,16 +488,16 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) struct pipe_context *pipe = pstip->pipe; uint num_samplers; - /* how many samplers? */ - /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ - num_samplers = MAX2(pstip->num_textures, pstip->num_samplers); - num_samplers = MAX2(num_samplers, pstip->sampler_unit + 1); - assert(stage->draw->rasterizer->poly_stipple_enable); /* bind our fragprog */ bind_pstip_fragment_shader(pstip); + /* how many samplers? */ + /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ + num_samplers = MAX2(pstip->num_textures, pstip->num_samplers); + num_samplers = MAX2(num_samplers, pstip->sampler_unit + 1); + /* plug in our sampler, texture */ pstip->state.samplers[pstip->sampler_unit] = pstip->sampler_cso; pstip->state.textures[pstip->sampler_unit] = pstip->texture; -- cgit v1.2.3 From b12a28db96d3bc7f01b6cdc9ee909f95a8c9ccc2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 13 Mar 2008 14:05:30 -0600 Subject: gallium: remove dead code related to shader semantic input/output info --- src/gallium/auxiliary/draw/draw_aaline.c | 6 ------ src/gallium/auxiliary/draw/draw_aapoint.c | 6 ------ src/gallium/auxiliary/draw/draw_pstipple.c | 10 +--------- src/gallium/include/pipe/p_state.h | 9 --------- 4 files changed, 1 insertion(+), 30 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index bed709bad0..f2b983374e 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -340,12 +340,6 @@ generate_aaline_fs(struct aaline_stage *aaline) tgsi_dump(aaline_fs.tokens, 0); #endif -#if 0 /* XXX remove */ - aaline_fs.input_semantic_name[aaline_fs.num_inputs] = TGSI_SEMANTIC_GENERIC; - aaline_fs.input_semantic_index[aaline_fs.num_inputs] = transform.maxGeneric + 1; - aaline_fs.num_inputs++; -#endif - aaline->fs->aaline_fs = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index 4bc2b5f3e2..67a7a8ebab 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -514,12 +514,6 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) tgsi_dump(aapoint_fs.tokens, 0); #endif -#if 0 /* XXX remove */ - aapoint_fs.input_semantic_name[aapoint_fs.num_inputs] = TGSI_SEMANTIC_GENERIC; - aapoint_fs.input_semantic_index[aapoint_fs.num_inputs] = transform.maxGeneric + 1; - aapoint_fs.num_inputs++; -#endif - aapoint->fs->aapoint_fs = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs); diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index 08f06bf222..09d542002f 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -324,21 +324,13 @@ generate_pstip_fs(struct pstip_stage *pstip) (struct tgsi_token *) pstip_fs.tokens, MAX, &transform.base); -#if 1 /* DEBUG */ +#if 0 /* DEBUG */ tgsi_dump(orig_fs->tokens, 0); tgsi_dump(pstip_fs.tokens, 0); #endif pstip->sampler_unit = transform.maxSampler + 1; -#if 0 /* XXX remove */ - if (transform.wincoordInput < 0) { - pstip_fs.input_semantic_name[pstip_fs.num_inputs] = TGSI_SEMANTIC_POSITION; - pstip_fs.input_semantic_index[pstip_fs.num_inputs] = (ubyte)transform.maxInput; - pstip_fs.num_inputs++; - } -#endif - pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs); } diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index bb5fab5d5f..02c354322d 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -162,15 +162,6 @@ struct pipe_constant_buffer struct pipe_shader_state { const struct tgsi_token *tokens; -#if 0 - /* XXX these are going away */ - ubyte num_inputs; - ubyte num_outputs; - ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ - ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; - ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ - ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; -#endif }; -- cgit v1.2.3 From 3115e8c968b51d22962b1b92b13946956dddd98e Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 13 Mar 2008 13:03:22 -0700 Subject: cell: Fix to work with commit fa9e7e9a8debb68611909ac2ffab527c6c39a3e5 --- src/gallium/drivers/cell/ppu/cell_context.h | 4 +++- src/gallium/drivers/cell/ppu/cell_state_derived.c | 20 ++++++++++---------- src/gallium/drivers/cell/ppu/cell_state_shader.c | 23 ++--------------------- 3 files changed, 15 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index c568922cbd..b221424323 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -37,13 +37,14 @@ #include "cell_winsys.h" #include "cell/common.h" #include "rtasm/rtasm_ppc_spe.h" - +#include "tgsi/util/tgsi_scan.h" struct cell_vbuf_render; struct cell_vertex_shader_state { struct pipe_shader_state shader; + struct tgsi_shader_info info; void *draw_data; }; @@ -51,6 +52,7 @@ struct cell_vertex_shader_state struct cell_fragment_shader_state { struct pipe_shader_state shader; + struct tgsi_shader_info info; void *data; }; diff --git a/src/gallium/drivers/cell/ppu/cell_state_derived.c b/src/gallium/drivers/cell/ppu/cell_state_derived.c index 0c46829258..5c240a55c0 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_derived.c +++ b/src/gallium/drivers/cell/ppu/cell_state_derived.c @@ -36,14 +36,14 @@ static int -find_vs_output(const struct pipe_shader_state *vs, +find_vs_output(const struct cell_vertex_shader_state *vs, uint semantic_name, uint semantic_index) { uint i; - for (i = 0; i < vs->num_outputs; i++) { - if (vs->output_semantic_name[i] == semantic_name && - vs->output_semantic_index[i] == semantic_index) + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == semantic_name && + vs->info.output_semantic_index[i] == semantic_index) return i; } return -1; @@ -58,8 +58,8 @@ find_vs_output(const struct pipe_shader_state *vs, static void calculate_vertex_layout( struct cell_context *cell ) { - const struct pipe_shader_state *vs = &cell->vs->shader; - const struct pipe_shader_state *fs = &cell->fs->shader; + const struct cell_vertex_shader_state *vs = cell->vs; + const struct cell_fragment_shader_state *fs = cell->fs; const enum interp_mode colorInterp = cell->rasterizer->flatshade ? INTERP_CONSTANT : INTERP_LINEAR; struct vertex_info *vinfo = &cell->vertex_info; @@ -91,15 +91,15 @@ calculate_vertex_layout( struct cell_context *cell ) * Loop over fragment shader inputs, searching for the matching output * from the vertex shader. */ - for (i = 0; i < fs->num_inputs; i++) { - switch (fs->input_semantic_name[i]) { + for (i = 0; i < fs->info.num_inputs; i++) { + switch (fs->info.input_semantic_name[i]) { case TGSI_SEMANTIC_POSITION: /* already done above */ break; case TGSI_SEMANTIC_COLOR: src = find_vs_output(vs, TGSI_SEMANTIC_COLOR, - fs->input_semantic_index[i]); + fs->info.input_semantic_index[i]); assert(src >= 0); draw_emit_vertex_attr(vinfo, EMIT_4F, colorInterp, src); break; @@ -117,7 +117,7 @@ calculate_vertex_layout( struct cell_context *cell ) case TGSI_SEMANTIC_GENERIC: /* this includes texcoords and varying vars */ src = find_vs_output(vs, TGSI_SEMANTIC_GENERIC, - fs->input_semantic_index[i]); + fs->info.input_semantic_index[i]); assert(src >= 0); draw_emit_vertex_attr(vinfo, EMIT_4F, INTERP_PERSPECTIVE, src); break; diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c index 269a5c15ba..fb2e940348 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -54,27 +54,7 @@ cell_create_fs_state(struct pipe_context *pipe, state->shader = *templ; -#if 0 - if (cell->dump_fs) { - tgsi_dump(state->shader.tokens, 0); - } - -#if defined(__i386__) || defined(__386__) - if (cell->use_sse) { - x86_init_func( &state->sse2_program ); - tgsi_emit_sse2_fs( state->shader.tokens, &state->sse2_program ); - } -#endif - -#ifdef MESA_LLVM - state->llvm_prog = 0; - if (!gallivm_global_cpu_engine()) { - gallivm_cpu_engine_create(state->llvm_prog); - } - else - gallivm_cpu_jit_compile(gallivm_global_cpu_engine(), state->llvm_prog); -#endif -#endif + tgsi_scan_shader(templ->tokens, &state->info); return state; } @@ -113,6 +93,7 @@ cell_create_vs_state(struct pipe_context *pipe, return NULL; state->shader = *templ; + tgsi_scan_shader(templ->tokens, &state->info); state->draw_data = draw_create_vertex_shader(cell->draw, &state->shader); if (state->draw_data == NULL) { -- cgit v1.2.3 From b6ed165748c6585f1368be37c0d0289cead419c9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 14:32:20 -0600 Subject: gallium: added bypass_vs flag to rasterizer state (may be temporary) --- src/gallium/include/pipe/p_state.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 02c354322d..5791a10119 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -110,6 +110,7 @@ struct pipe_rasterizer_state unsigned line_stipple_pattern:16; unsigned line_last_pixel:1; unsigned bypass_clipping:1; + unsigned bypass_vs:1; /**< vertices are already fully transformed */ unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ float line_width; -- cgit v1.2.3 From 8b8c9acdb747499149e633179a8ad10b0e4206b1 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 14:33:57 -0600 Subject: gallium: added draw_enable_line_stipple() function Allows drivers that implement line stipple to turn off this drawing stage. --- src/gallium/auxiliary/draw/draw_context.c | 12 ++++++++++++ src/gallium/auxiliary/draw/draw_context.h | 3 +++ src/gallium/auxiliary/draw/draw_private.h | 5 +++-- 3 files changed, 18 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index fed2b6e759..3d09d95851 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -85,6 +85,7 @@ struct draw_context *draw_create( void ) /* these defaults are oriented toward the needs of softpipe */ draw->wide_point_threshold = 1000000.0; /* infinity */ draw->wide_line_threshold = 1.0; + draw->line_stipple = TRUE; draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ @@ -254,6 +255,17 @@ draw_wide_line_threshold(struct draw_context *draw, float threshold) } +/** + * Tells the draw module whether or not to implement line stipple. + */ +void +draw_enable_line_stipple(struct draw_context *draw, boolean enable) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->line_stipple = enable; +} + + /** * Ask the draw module for the location/slot of the given vertex attribute in * a post-transformed vertex. diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index df63e91a22..e8e2b56bae 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -94,6 +94,9 @@ void draw_wide_point_threshold(struct draw_context *draw, float threshold); void draw_wide_line_threshold(struct draw_context *draw, float threshold); +void draw_enable_line_stipple(struct draw_context *draw, boolean enable); + + boolean draw_use_sse(struct draw_context *draw); void diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 4147472d45..379b6c3206 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -238,6 +238,7 @@ struct draw_context float wide_point_threshold; /**< convert pnts to tris if larger than this */ float wide_line_threshold; /**< convert lines to tris if wider than this */ + boolean line_stipple; /**< do line stipple? */ boolean use_sse; /* If a prim stage introduces new vertex attributes, they'll be stored here @@ -344,10 +345,10 @@ extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); -struct tgsi_exec_machine; - extern void draw_update_vertex_fetch( struct draw_context *draw ); +extern boolean draw_need_pipeline(const struct draw_context *draw); + /* Prototype/hack */ -- cgit v1.2.3 From 3faf6230ff4b63833c072ac7afeb43c25d3cba22 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 14:34:35 -0600 Subject: gallium: added draw_need_pipeline() predicate function To test if we need any pipeline stage, or whether we can go into passthrough mode. --- src/gallium/auxiliary/draw/draw_validate.c | 55 +++++++++++++++++++++++++++++- 1 file changed, 54 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index b43295b586..602fa3429d 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -33,6 +33,59 @@ #include "draw_private.h" +/** + * Check if we need any special pipeline stages, or whether prims/verts + * can go through untouched. + */ +boolean +draw_need_pipeline(const struct draw_context *draw) +{ + /* clipping */ + if (!draw->rasterizer->bypass_clipping) + return TRUE; + + /* vertex shader */ + if (!draw->rasterizer->bypass_vs) + return TRUE; + + /* line stipple */ + if (draw->rasterizer->line_stipple_enable && draw->line_stipple) + return TRUE; + + /* wide lines */ + if (draw->rasterizer->line_width > draw->wide_line_threshold) + return TRUE; + + /* large points */ + if (draw->rasterizer->point_size > draw->wide_point_threshold) + return TRUE; + + /* AA lines */ + if (draw->rasterizer->line_smooth && draw->pipeline.aaline) + return TRUE; + + /* AA points */ + if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) + return TRUE; + + /* polygon stipple */ + if (draw->rasterizer->poly_stipple_enable && draw->pipeline.pstipple) + return TRUE; + + /* polygon offset */ + if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) + return TRUE; + + /* two-side lighting */ + if (draw->rasterizer->light_twoside) + return TRUE; + + /* polygon cull */ + if (draw->rasterizer->cull_mode) + return TRUE; + + return FALSE; +} @@ -92,7 +145,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) next = draw->pipeline.wide_point; } - if (draw->rasterizer->line_stipple_enable) { + if (draw->rasterizer->line_stipple_enable && draw->line_stipple) { draw->pipeline.stipple->next = next; next = draw->pipeline.stipple; precalc_flat = 1; /* only needed for lines really */ -- cgit v1.2.3 From a889928d85ac8ba7e1a7fe15393858a9422cf750 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 13 Mar 2008 16:41:12 -0400 Subject: add a way of removing an exact iterator from the hash --- src/gallium/auxiliary/cso_cache/cso_hash.c | 25 ++++++++++++++++++++++--- src/gallium/auxiliary/cso_cache/cso_hash.h | 16 ++++++++++++++-- 2 files changed, 36 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index 5cad5d3be7..ddce3822f7 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -99,7 +99,7 @@ static void *cso_data_allocate_node(struct cso_hash_data *hash) return MALLOC(hash->nodeSize); } -static void cso_data_free_node(struct cso_node *node) +static void cso_free_node(struct cso_node *node) { FREE(node); } @@ -248,7 +248,7 @@ void cso_hash_delete(struct cso_hash *hash) struct cso_node *cur = *bucket++; while (cur != e_for_x) { struct cso_node *next = cur->next; - cso_data_free_node(cur); + cso_free_node(cur); cur = next; } } @@ -367,7 +367,7 @@ void * cso_hash_take(struct cso_hash *hash, if (*node != hash->data.e) { void *t = (*node)->value; struct cso_node *next = (*node)->next; - cso_data_free_node(*node); + cso_free_node(*node); *node = next; --hash->data.d->size; cso_data_has_shrunk(hash->data.d); @@ -393,3 +393,22 @@ int cso_hash_size(struct cso_hash *hash) { return hash->data.d->size; } + +struct cso_hash_iter cso_hash_erase(struct cso_hash *hash, struct cso_hash_iter iter) +{ + struct cso_hash_iter ret = iter; + struct cso_node *node = iter.node; + struct cso_node **node_ptr; + + if (node == hash->data.e) + return iter; + + ret = cso_hash_iter_next(ret); + node_ptr = (struct cso_node**)(&hash->data.d->buckets[node->key % hash->data.d->numBuckets]); + while (*node_ptr != node) + node_ptr = &(*node_ptr)->next; + *node_ptr = node->next; + cso_free_node(node); + --hash->data.d->size; + return ret; +} diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h index a3a65b68c8..73c4742006 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.h +++ b/src/gallium/auxiliary/cso_cache/cso_hash.h @@ -68,14 +68,26 @@ int cso_hash_size(struct cso_hash *hash); /** - * Create a list of objects and just add entry with the same key to the list. + * Adds a data with the given key to the hash. If entry with the given + * key is already in the hash, this current entry is instered before it + * in the collision list. + * Function returns iterator pointing to the inserted item in the hash. */ struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, unsigned key, void *data); +/** + * Removes the item pointed to by the current iterator from the hash. + * Note that the data itself is not erased and if it was a malloc'ed pointer + * it will have to be freed after calling this function by the callee. + * Function returns iterator pointing to the item after the removed one in + * the hash. + */ +struct cso_hash_iter cso_hash_erase(struct cso_hash *hash, struct cso_hash_iter iter); void *cso_hash_take(struct cso_hash *hash, unsigned key); + struct cso_hash_iter cso_hash_first_node(struct cso_hash *hash); /** @@ -97,7 +109,7 @@ struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); * Convenience routine to iterate over the collision list while doing a memory * comparison to see which entry in the list is a direct copy of our template * and returns that entry. - */ + */ void *cso_hash_find_data_from_template( struct cso_hash *hash, unsigned hash_key, void *templ, -- cgit v1.2.3 From 69c39b9ae28764194a6d310d58aa36b7ac596aa9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 14:57:19 -0600 Subject: gallium: set surface status to CLEAR or DEFINED in clearing/drawing code. Otherwise, we were never setting these flags. This confused the state tracker. Fixes progs/demos/texenv.c, probably others. --- src/gallium/drivers/softpipe/sp_clear.c | 2 ++ src/gallium/drivers/softpipe/sp_prim_setup.c | 11 +++++++++++ 2 files changed, 13 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_clear.c b/src/gallium/drivers/softpipe/sp_clear.c index 8d295a30ca..39aed151c7 100644 --- a/src/gallium/drivers/softpipe/sp_clear.c +++ b/src/gallium/drivers/softpipe/sp_clear.c @@ -55,6 +55,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, if (ps == sp_tile_cache_get_surface(softpipe->zsbuf_cache)) { sp_tile_cache_clear(softpipe->zsbuf_cache, clearValue); + softpipe->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_CLEAR; #if TILE_CLEAR_OPTIMIZATION return; #endif @@ -63,6 +64,7 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { if (ps == sp_tile_cache_get_surface(softpipe->cbuf_cache[i])) { sp_tile_cache_clear(softpipe->cbuf_cache[i], clearValue); + softpipe->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_CLEAR; } } diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 2feee5c485..7aa9cf8e85 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -1164,11 +1164,22 @@ static void setup_begin( struct draw_stage *stage ) struct setup_stage *setup = setup_stage(stage); struct softpipe_context *sp = setup->softpipe; const struct sp_fragment_shader *fs = setup->softpipe->fs; + uint i; if (sp->dirty) { softpipe_update_derived(sp); } + /* Mark surfaces as defined now */ + for (i = 0; i < sp->framebuffer.num_cbufs; i++){ + if (sp->framebuffer.cbufs[i]) { + sp->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED; + } + } + if (sp->framebuffer.zsbuf) { + sp->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_DEFINED; + } + setup->quad.nr_attrs = fs->info.num_inputs; sp->quad.first->begin(sp->quad.first); -- cgit v1.2.3 From 3e625ce18e35b2b0343962f93480abf4bf9b188a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 15:07:14 -0600 Subject: gallium: add some temporary code for testing draw module vertex passthrough Set TEST_DRAW_PASSTHROUGH=1, run progs/trivial/clear-scissor --- src/mesa/state_tracker/st_cb_clear.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index cc8a136292..4fe6195a07 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -52,7 +52,8 @@ #include "cso_cache/cso_context.h" - +/* XXX for testing draw module vertex passthrough: */ +#define TEST_DRAW_PASSTHROUGH 0 static GLuint @@ -229,6 +230,12 @@ draw_quad(GLcontext *ctx, GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ GLuint i; +#if TEST_DRAW_PASSTHROUGH + /* invert Y coords (may be off by one pixel) */ + y0 = ctx->DrawBuffer->Height - y0; + y1 = ctx->DrawBuffer->Height - y1; +#endif + /* positions */ verts[0][0][0] = x0; verts[0][0][1] = y0; @@ -335,6 +342,10 @@ clear_with_quad(GLcontext *ctx, */ if (ctx->Scissor.Enabled) raster.scissor = 1; +#endif +#if TEST_DRAW_PASSTHROUGH + raster.bypass_clipping = 1; + raster.bypass_vs = 1; #endif cso_set_rasterizer(st->cso_context, &raster); } @@ -348,6 +359,7 @@ clear_with_quad(GLcontext *ctx, pipe->bind_fs_state(pipe, stfp->driver_shader); } +#if !TEST_DRAW_PASSTHROUGH /* vertex shader state: color/position pass-through */ { static struct st_vertex_program *stvp = NULL; @@ -356,7 +368,9 @@ clear_with_quad(GLcontext *ctx, } pipe->bind_vs_state(pipe, stvp->driver_shader); } +#endif +#if !TEST_DRAW_PASSTHROUGH /* viewport state: viewport matching window dims */ { const float width = ctx->DrawBuffer->Width; @@ -372,6 +386,7 @@ clear_with_quad(GLcontext *ctx, vp.translate[3] = 0.0; pipe->set_viewport_state(pipe, &vp); } +#endif /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); -- cgit v1.2.3 From 7d5e38a55ae99a4c28873377572f77f383ce0c3e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 16:53:44 -0600 Subject: gallium: plug in new sp_vbuf_draw_arrays() function Will be used for pass-through mode. Also, call draw_set_render() to register the vbuf stage. Should probably rename that function to something like draw_set_vbuf_stage(). --- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 57 +++++++++++++++++++++++++++++ 1 file changed, 57 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 69bea8a8f5..184aac16f4 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -183,6 +183,60 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) } +/** + * This function is hit when the draw module is working in pass-through mode. + * It's up to us to convert the vertex array into point/line/tri prims. + */ +static void +sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) +{ + struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + struct softpipe_context *softpipe = cvbr->softpipe; + struct draw_stage *setup = softpipe->setup; + struct prim_header prim; + const void *vertex_buffer = cvbr->vertex_buffer; + const unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); + unsigned i, j; + + prim.det = 0; + prim.reset_line_stipple = 0; + prim.edgeflags = 0; + prim.pad = 0; + +#define VERTEX(I) \ + (struct vertex_header *) ((char *) vertex_buffer + (I) * vertex_size) + + switch (cvbr->prim) { + case PIPE_PRIM_TRIANGLES: + assert(nr % 3 == 0); + for (i = 0; i < nr; i += 3) { + prim.v[0] = VERTEX(i + 0); + prim.v[1] = VERTEX(i + 1); + prim.v[2] = VERTEX(i + 2); + calc_det(&prim); + setup->tri( setup, &prim ); + } + break; + case PIPE_PRIM_POLYGON: + /* draw as tri fan */ + for (i = 2; i < nr; i++) { + prim.v[0] = VERTEX(0); + prim.v[1] = VERTEX(i - 1); + prim.v[2] = VERTEX(i); + calc_det(&prim); + setup->tri( setup, &prim ); + } + break; + default: + /* XXX finish remaining prim types */ + assert(0); + } + +#undef VERTEX +} + + + static void sp_vbuf_destroy(struct vbuf_render *vbr) { @@ -210,6 +264,7 @@ sp_init_vbuf(struct softpipe_context *sp) sp->vbuf_render->base.allocate_vertices = sp_vbuf_allocate_vertices; sp->vbuf_render->base.set_primitive = sp_vbuf_set_primitive; sp->vbuf_render->base.draw = sp_vbuf_draw; + sp->vbuf_render->base.draw_arrays = sp_vbuf_draw_arrays; sp->vbuf_render->base.release_vertices = sp_vbuf_release_vertices; sp->vbuf_render->base.destroy = sp_vbuf_destroy; @@ -218,4 +273,6 @@ sp_init_vbuf(struct softpipe_context *sp) sp->vbuf = draw_vbuf_stage(sp->draw, &sp->vbuf_render->base); draw_set_rasterize_stage(sp->draw, sp->vbuf); + + draw_set_render(sp->draw, &sp->vbuf_render->base); } -- cgit v1.2.3 From b9518a4e39f739a31dd3f62d67563944f8c266a9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 16:55:48 -0600 Subject: gallium: added new EMIT_HEADER token Used to emit the struct vertex_header info for softpipe. Before we were using the EMIT_ALL token but that's insufficient for the draw pass-through mode. EMIT_ALL might get removed soon... --- src/gallium/auxiliary/draw/draw_vertex.c | 3 +++ src/gallium/auxiliary/draw/draw_vertex.h | 1 + 2 files changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vertex.c b/src/gallium/auxiliary/draw/draw_vertex.c index daf1ef4b80..970adc95e7 100644 --- a/src/gallium/auxiliary/draw/draw_vertex.c +++ b/src/gallium/auxiliary/draw/draw_vertex.c @@ -52,6 +52,9 @@ draw_compute_vertex_size(struct vertex_info *vinfo) switch (vinfo->emit[i]) { case EMIT_OMIT: break; + case EMIT_HEADER: + vinfo->size += sizeof(struct vertex_header) / 4; + break; case EMIT_4UB: /* fall-through */ case EMIT_1F_PSIZE: diff --git a/src/gallium/auxiliary/draw/draw_vertex.h b/src/gallium/auxiliary/draw/draw_vertex.h index 267c74203b..abd2017ed3 100644 --- a/src/gallium/auxiliary/draw/draw_vertex.h +++ b/src/gallium/auxiliary/draw/draw_vertex.h @@ -48,6 +48,7 @@ enum attrib_emit { EMIT_OMIT, /**< don't emit the attribute */ EMIT_ALL, /**< emit whole post-xform vertex, w/ header */ + EMIT_HEADER, /**< emit vertex_header struct (XXX temp?) */ EMIT_1F, EMIT_1F_PSIZE, /**< insert constant point size */ EMIT_2F, -- cgit v1.2.3 From 13334c8dd2744402d43f8ea0a9d2c0e5e76ac28e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 16:57:37 -0600 Subject: gallium: added EMIT_HEADER case in emit_vertex() --- src/gallium/auxiliary/draw/draw_vbuf.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vbuf.c b/src/gallium/auxiliary/draw/draw_vbuf.c index 71ac73912b..f83b441e93 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_vbuf.c @@ -225,6 +225,11 @@ emit_vertex( struct vbuf_stage *vbuf, vbuf->vertex_ptr += vinfo->size; count += vinfo->size; break; + case EMIT_HEADER: + memcpy(vbuf->vertex_ptr, vertex, sizeof(*vertex)); + *vbuf->vertex_ptr += sizeof(*vertex) / 4; + count += sizeof(*vertex) / 4; + break; case EMIT_1F: *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); count++; -- cgit v1.2.3 From 269fbeb5459952532f5d188dd3653fa6b7425cfe Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 16:57:57 -0600 Subject: gallium: in softpipe_get_vertex_info() generate a vbuf vertex_info with real attribs Can't use the EMIT_ALL shortcut/optimization anymore because of passthrough mode. --- src/gallium/drivers/softpipe/sp_state_derived.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index eafbaed4b9..10483675ea 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -72,10 +72,22 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) */ struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; vinfo_vbuf->num_attribs = 0; +#if 0 + /* special-case to allow memcpy of whole vertex */ draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); /* size in dwords or floats */ vinfo_vbuf->size = 4 * draw_num_vs_outputs(softpipe->draw) + sizeof(struct vertex_header) / 4; +#else + /* for pass-through mode, we need a more explicit list of attribs */ + const uint num = draw_num_vs_outputs(softpipe->draw); + uint i; + draw_emit_vertex_attr(vinfo_vbuf, EMIT_HEADER, INTERP_NONE, 0); + for (i = 0; i < num; i++) { + draw_emit_vertex_attr(vinfo_vbuf, EMIT_4F, INTERP_PERSPECTIVE, i); + } + draw_compute_vertex_size(vinfo_vbuf); +#endif } /* -- cgit v1.2.3 From cf106789abd4a84e8f07dc6ca12d2261e9bf92cd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 17:04:04 -0600 Subject: gallium: added EMIT_HEADER case --- src/gallium/auxiliary/draw/draw_vf.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vf.c b/src/gallium/auxiliary/draw/draw_vf.c index 901ff20a7e..4b5b2a86f6 100644 --- a/src/gallium/auxiliary/draw/draw_vf.c +++ b/src/gallium/auxiliary/draw/draw_vf.c @@ -263,6 +263,10 @@ void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, } break; } + case EMIT_HEADER: + /* XXX emit new DRAW_EMIT_HEADER attribute??? */ + count += sizeof(struct vertex_header) / 4; + break; case EMIT_1F: attrs[nr_attrs].attrib = j; attrs[nr_attrs].format = DRAW_EMIT_1F; -- cgit v1.2.3 From ce49c4c24bf72ea642015d566ff687d512574fd5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 17:04:56 -0600 Subject: gallium: fix bug in draw_num_vs_outputs() --- src/gallium/auxiliary/draw/draw_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 3d09d95851..e8473a4c57 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -311,7 +311,7 @@ uint draw_num_vs_outputs(struct draw_context *draw) { uint count = draw->vertex_shader->info.num_outputs; - if (draw->extra_vp_outputs.slot >= 0) + if (draw->extra_vp_outputs.slot > 0) count++; return count; } -- cgit v1.2.3 From 5bae5871f03c96eb173cb55d7e8a846a2c4bd4a0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 17:06:13 -0600 Subject: gallium: for TEST_DRAW_PASSTHROUGH, pass inClipCoords=FALSE to st_draw_vertices() When pass-through mode is fully supported we'll clean this up more. --- src/mesa/state_tracker/st_cb_clear.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 4fe6195a07..5865071439 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -259,7 +259,13 @@ draw_quad(GLcontext *ctx, verts[i][1][3] = color[3]; } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, GL_FALSE); + st_draw_vertices(ctx, PIPE_PRIM_POLYGON, 4, (float *) verts, 2, +#if TEST_DRAW_PASSTHROUGH + GL_TRUE +#else + GL_FALSE +#endif + ); } -- cgit v1.2.3 From d088d640fca415261a208d3cbede94a6522ebb6b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 17:10:28 -0600 Subject: gallium: plug in vertex passthrough code Based on a patch from Zack. Basically, implement a new draw_arrays function that copies the incoming user-vertices to the hardware vertex buffer, doing format/type conversion as needed. The vertex fetch/store code is totally temporary for now. --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/SConscript | 1 + src/gallium/auxiliary/draw/draw_passthrough.c | 106 ++++++++++++++++++++++++-- src/gallium/auxiliary/draw/draw_prim.c | 47 +++++++++++- 4 files changed, 147 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 2daa1636f3..21e9f737b7 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -15,6 +15,7 @@ C_SOURCES = \ draw_debug.c \ draw_flatshade.c \ draw_offset.c \ + draw_passthrough.c \ draw_prim.c \ draw_pstipple.c \ draw_stipple.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 5cb7664c85..64b444dbd5 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -15,6 +15,7 @@ draw = env.ConvenienceLibrary( 'draw_flatshade.c', 'draw_offset.c', 'draw_prim.c', + 'draw_passthrough.c', 'draw_pstipple.c', 'draw_passthrough.c', 'draw_stipple.c', diff --git a/src/gallium/auxiliary/draw/draw_passthrough.c b/src/gallium/auxiliary/draw/draw_passthrough.c index a51fa0ab23..d16f056191 100644 --- a/src/gallium/auxiliary/draw/draw_passthrough.c +++ b/src/gallium/auxiliary/draw/draw_passthrough.c @@ -66,6 +66,95 @@ #include "draw/draw_vertex.h" +/** + * General-purpose fetch from user's vertex arrays, emit to driver's + * vertex buffer. + * + * XXX this is totally temporary. + */ +static void +fetch_store_general( struct draw_context *draw, + float *out, + unsigned start, + unsigned count ) +{ + const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); + const unsigned nr_attrs = vinfo->num_attribs; + uint i, j; + + const unsigned *pitch = draw->vertex_fetch.pitch; + const ubyte **src = draw->vertex_fetch.src_ptr; + + for (i = start; i < count; i++) { + for (j = 0; j < nr_attrs; j++) { + const uint jj = vinfo->src_index[j]; + const enum pipe_format srcFormat = draw->vertex_element[jj].src_format; + const ubyte *from = src[jj] + i * pitch[jj]; + float attrib[4]; + + switch (srcFormat) { + case PIPE_FORMAT_R32G32B32A32_FLOAT: + { + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = f[1]; + attrib[2] = f[2]; + attrib[3] = f[3]; + } + break; + case PIPE_FORMAT_R32G32B32_FLOAT: + { + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = f[1]; + attrib[2] = f[2]; + attrib[3] = 1.0; + } + break; + case PIPE_FORMAT_R32G32_FLOAT: + { + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = f[1]; + attrib[2] = 0.0; + attrib[3] = 1.0; + } + break; + case PIPE_FORMAT_R32_FLOAT: + { + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = 0.0; + attrib[2] = 0.0; + attrib[3] = 1.0; + } + break; + default: + abort(); + } + + /* XXX this will probably only work for softpipe */ + switch (vinfo->emit[j]) { + case EMIT_HEADER: + memset(out, 0, sizeof(struct vertex_header)); + out += sizeof(struct vertex_header) / 4; + break; + case EMIT_4F: + out[0] = attrib[0]; + out[1] = attrib[1]; + out[2] = attrib[2]; + out[3] = attrib[3]; + out += 4; + break; + default: + abort(); + } + + } + } +} + + /* Example of a fetch/emit passthrough shader which could be * generated when bypass_clipping is enabled on a passthrough vertex @@ -116,7 +205,6 @@ static void fetch_xyz_rgb_st( struct draw_context *draw, } } - static boolean update_shader( struct draw_context *draw ) { const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); @@ -143,11 +231,15 @@ static boolean update_shader( struct draw_context *draw ) /* Just trying to figure out how this would work: */ - if (nr_attrs == 3 && - 0 /* some other tests */) + if (draw->rasterizer->bypass_vs || + (nr_attrs == 3 && 0 /* some other tests */)) { +#if 0 draw->vertex_fetch.pt_fetch = fetch_xyz_rgb_st; - assert(vinfo->size == 10); +#else + draw->vertex_fetch.pt_fetch = fetch_store_general; +#endif + /*assert(vinfo->size == 10);*/ return TRUE; } @@ -175,7 +267,6 @@ static boolean set_prim( struct draw_context *draw, } - boolean draw_passthrough_arrays(struct draw_context *draw, unsigned prim, @@ -184,10 +275,13 @@ draw_passthrough_arrays(struct draw_context *draw, { float *hw_verts; + if (draw_need_pipeline(draw)) + return FALSE; + if (!set_prim(draw, prim)) return FALSE; - if (!update_shader( draw )) + if (!update_shader(draw)) return FALSE; hw_verts = draw->render->allocate_vertices( draw->render, diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index 7d6cd43410..ff71ba9b73 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -31,6 +31,7 @@ */ #include "pipe/p_debug.h" +#include "pipe/p_util.h" #include "draw_private.h" #include "draw_context.h" @@ -118,7 +119,42 @@ static void draw_prim_queue_flush( struct draw_context *draw ) draw_vertex_cache_unreference( draw ); } +static INLINE void fetch_and_store(struct draw_context *draw) +{ + unsigned i; + + /* run vertex shader on vertex cache entries, four per invokation */ +#if 0 + { + const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); + memcpy(draw->vs.queue[0].vertex, draw->vs.queue[i + j].elt, + count * vinfo->size); + } +#elif 0 + draw_update_vertex_fetch(draw); + for (i = 0; i < draw->vs.queue_nr; i += 4) { + struct vertex_header *dests[4]; + unsigned elts[4]; + struct tgsi_exec_machine *machine = &draw->machine; + int j, n = MIN2(4, draw->vs.queue_nr - i); + + for (j = 0; j < n; j++) { + elts[j] = draw->vs.queue[i + j].elt; + dests[j] = draw->vs.queue[i + j].vertex; + } + for ( ; j < 4; j++) { + elts[j] = elts[0]; + dests[j] = draw->vs.queue[i + j].vertex; + } + //fetch directly into dests + draw->vertex_fetch.fetch_func(draw, machine, dests, count); + } +#endif + + draw->vs.post_nr = draw->vs.queue_nr; + draw->vs.queue_nr = 0; +} void draw_do_flush( struct draw_context *draw, unsigned flags ) { @@ -134,7 +170,10 @@ void draw_do_flush( struct draw_context *draw, unsigned flags ) if (flags >= DRAW_FLUSH_SHADER_QUEUE) { if (draw->vs.queue_nr) - (*draw->shader_queue_flush)(draw); + if (draw->rasterizer->bypass_vs) + fetch_and_store(draw); + else + (*draw->shader_queue_flush)(draw); if (flags >= DRAW_FLUSH_PRIM_QUEUE) { if (draw->pq.queue_nr) @@ -485,7 +524,11 @@ draw_arrays(struct draw_context *draw, unsigned prim, } /* drawing done here: */ - draw_prim(draw, prim, start, count); + if (!draw->rasterizer->bypass_vs || + !draw_passthrough_arrays(draw, prim, start, count)) { + /* we have to run the whole pipeline */ + draw_prim(draw, prim, start, count); + } } -- cgit v1.2.3 From 34be3969505d378c5a6734a134f03d094a865c56 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 17:39:30 -0600 Subject: gallium: fix EMIT_HEADER case in draw_vf_set_vertex_info() --- src/gallium/auxiliary/draw/draw_vf.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vf.c b/src/gallium/auxiliary/draw/draw_vf.c index 4b5b2a86f6..f4e29a6293 100644 --- a/src/gallium/auxiliary/draw/draw_vf.c +++ b/src/gallium/auxiliary/draw/draw_vf.c @@ -265,7 +265,11 @@ void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, } case EMIT_HEADER: /* XXX emit new DRAW_EMIT_HEADER attribute??? */ - count += sizeof(struct vertex_header) / 4; + attrs[nr_attrs].attrib = 0; + attrs[nr_attrs].format = DRAW_EMIT_PAD; + attrs[nr_attrs].offset = offsetof(struct vertex_header, data); + count += offsetof(struct vertex_header, data)/4; + nr_attrs++; break; case EMIT_1F: attrs[nr_attrs].attrib = j; -- cgit v1.2.3 From 78302c7ca30d27ef3d087deb4d1a22e83858ce4b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 13 Mar 2008 18:19:47 -0600 Subject: gallium: need to all draw_flush() in softpipe_unmap_constant_buffers() Otherwise, we won't have our constants when we run the fragment shader. Fixes crash in glsl tests when SP_VBUF=1. --- src/gallium/drivers/softpipe/sp_draw_arrays.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 2049afda34..5b5a0fe573 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -62,6 +62,14 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp) { struct pipe_winsys *ws = sp->pipe.winsys; uint i; + + /* really need to flush all prims since the vert/frag shaders const buffers + * are going away now. + */ + draw_flush(sp->draw); + + draw_set_mapped_constant_buffer(sp->draw, NULL); + for (i = 0; i < 2; i++) { if (sp->constants[i].size) ws->buffer_unmap(ws, sp->constants[i].buffer); -- cgit v1.2.3 From d34bc880a4d17420ec20d422dcb461783457c473 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 14 Mar 2008 08:42:08 +0000 Subject: tgsi: add debug_printf version of tgsi_dump --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 22 ++++++++++++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_dump.h | 4 ++++ 2 files changed, 26 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index aa78278700..7d292778ad 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -1563,3 +1563,25 @@ tgsi_dump_str( *str = dump.text; } + + +void tgsi_debug_dump( struct tgsi_token *tokens ) +{ + char *str, *p; + + tgsi_dump_str( &str, tokens, 0 ); + + p = str; + while (p != NULL) + { + char *end = strchr( p, '\n' ); + if (end != NULL) + { + *end++ = '\0'; + } + debug_printf( "%s\n", p ); + p = end; + } + + FREE( str ); +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h index b983b38226..51d79a0362 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h @@ -20,6 +20,10 @@ tgsi_dump_str( const struct tgsi_token *tokens, unsigned flags ); +/* Dump to debug_printf() + */ +void tgsi_debug_dump( struct tgsi_token *tokens ); + #if defined __cplusplus } #endif -- cgit v1.2.3 From e4cdce43cebe6a2b38f7ea5145474ca2b12c57bb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 14 Mar 2008 08:42:45 +0000 Subject: gallium: fix some compiler warnings --- src/gallium/auxiliary/draw/draw_prim.c | 4 ++-- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 2 +- src/gallium/drivers/softpipe/sp_state_derived.c | 4 +++- 3 files changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index ff71ba9b73..888fa536ea 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -121,8 +121,6 @@ static void draw_prim_queue_flush( struct draw_context *draw ) static INLINE void fetch_and_store(struct draw_context *draw) { - unsigned i; - /* run vertex shader on vertex cache entries, four per invokation */ #if 0 { @@ -131,6 +129,8 @@ static INLINE void fetch_and_store(struct draw_context *draw) count * vinfo->size); } #elif 0 + unsigned i; + draw_update_vertex_fetch(draw); for (i = 0; i < draw->vs.queue_nr; i += 4) { struct vertex_header *dests[4]; diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 184aac16f4..db0913cb2b 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -196,7 +196,7 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) struct prim_header prim; const void *vertex_buffer = cvbr->vertex_buffer; const unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); - unsigned i, j; + unsigned i; prim.det = 0; prim.reset_line_stipple = 0; diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 10483675ea..82cb31ece7 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -71,8 +71,8 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) * simply emit the whole post-xform vertex as-is: */ struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; - vinfo_vbuf->num_attribs = 0; #if 0 + vinfo_vbuf->num_attribs = 0; /* special-case to allow memcpy of whole vertex */ draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); /* size in dwords or floats */ @@ -82,6 +82,8 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) /* for pass-through mode, we need a more explicit list of attribs */ const uint num = draw_num_vs_outputs(softpipe->draw); uint i; + + vinfo_vbuf->num_attribs = 0; draw_emit_vertex_attr(vinfo_vbuf, EMIT_HEADER, INTERP_NONE, 0); for (i = 0; i < num; i++) { draw_emit_vertex_attr(vinfo_vbuf, EMIT_4F, INTERP_PERSPECTIVE, i); -- cgit v1.2.3 From ee55eaa609aee9ffd7ddbc839d7d6161727cee3a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Mar 2008 10:19:19 -0600 Subject: gallium: in st_program_string_notify() free shader tokens. When the shader instructions change, free the TGSI tokens. This signals the state tracker to recompile the shader. Fixes a regression caused by recent shader/state-tracker changes. --- src/mesa/state_tracker/st_cb_program.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 4dc76f19b1..0a62e20e0c 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -178,6 +178,11 @@ static void st_program_string_notify( GLcontext *ctx, stfp->param_state = stfp->Base.Base.Parameters->StateFlags; + if (stfp->state.tokens) { + FREE((void *) stfp->state.tokens); + stfp->state.tokens = NULL; + } + if (st->fp == stfp) st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; } @@ -201,6 +206,11 @@ static void st_program_string_notify( GLcontext *ctx, stvp->param_state = stvp->Base.Base.Parameters->StateFlags; + if (stvp->state.tokens) { + FREE((void *) stvp->state.tokens); + stvp->state.tokens = NULL; + } + if (st->vp == stvp) st->dirty.st |= ST_NEW_VERTEX_PROGRAM; } -- cgit v1.2.3 From c45a525cdc4bd5f9b35c4f6abae6a65f3279e2c4 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Mar 2008 10:20:29 -0600 Subject: mesa: revert fast-track glColor and similar calls when not immediate mode rendering Revert commit d08cd68d3f5456b69ad504ede2b090c0bb6474db This change caused some glean tests to fail. The alpha value of colors was always 1.0 instead of the value from glColor4fv(). The mesa color array found in st_draw() has size=3 instead of size=4. --- src/mesa/vbo/vbo_exec_api.c | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index 35dc0e768f..b7f4d8a307 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -309,23 +309,16 @@ static void vbo_exec_fixup_vertex( GLcontext *ctx, GLuint attr, GLuint sz ) { struct vbo_exec_context *exec = &vbo_context(ctx)->exec; - static const GLfloat id[4] = { 0, 0, 0, 1 }; int i; - if (exec->vtx.prim_count == 0) { - GLfloat *current = (GLfloat *)vbo_context(ctx)->currval[attr].Ptr; - exec->vtx.attrptr[attr] = current; - memcpy(current, id, sizeof(id)); - ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT; - return; - } - else if (sz > exec->vtx.attrsz[attr]) { + if (sz > exec->vtx.attrsz[attr]) { /* New size is larger. Need to flush existing vertices and get * an enlarged vertex format. */ vbo_exec_wrap_upgrade_vertex( exec, attr, sz ); } else if (sz < exec->vtx.active_sz[attr]) { + static const GLfloat id[4] = { 0, 0, 0, 1 }; /* New size is smaller - just need to fill in some * zeros. Don't need to flush or wrap. -- cgit v1.2.3 From a254d0c36870724e51e91de3a3da9220b67af850 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Mar 2008 10:22:53 -0600 Subject: gallium: added dummy install target --- src/gallium/winsys/Makefile | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/Makefile b/src/gallium/winsys/Makefile index 3dc5ea5744..2360a6a94a 100644 --- a/src/gallium/winsys/Makefile +++ b/src/gallium/winsys/Makefile @@ -18,3 +18,7 @@ subdirs: clean: rm -f `find . -name \*.[oa]` + + +# Dummy install target +install: -- cgit v1.2.3 From 027433176cddec58821d625fb2df45cfd95f1e33 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Mar 2008 10:23:11 -0600 Subject: i915: check for NULL const buffer ptr --- src/gallium/drivers/i915simple/i915_state.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 503c092400..1cec36e206 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -516,7 +516,7 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, * stays the same. In that case we should only be updating the first * N constants, leaving any extras from shader translation alone. */ - { + if (buf) { void *mapped; if (buf->size && (mapped = ws->buffer_map(ws, buf->buffer, -- cgit v1.2.3 From 3088eb59497ec8621e003ce3bc87025f257c0a92 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Mar 2008 10:23:39 -0600 Subject: gallium: added some debug code (disabled) --- src/gallium/auxiliary/draw/draw_vs_exec.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 55bec14116..364693e0b4 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -166,6 +166,19 @@ vs_exec_run( struct draw_vertex_shader *shader, vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; } + +#if 0 /*DEBUG*/ + printf("Post xform vert:\n"); + for (slot = 0; slot < draw->num_vs_outputs; slot++) { + printf("%d: %f %f %f %f\n", slot, + vOut[j]->data[slot][0], + vOut[j]->data[slot][1], + vOut[j]->data[slot][2], + vOut[j]->data[slot][3]); + } +#endif + + } /* loop over vertices */ } -- cgit v1.2.3 From 9de9e1fe8c3f87fe672aed074348f07107fa3cec Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 14 Mar 2008 11:24:28 -0600 Subject: gallium: print warning rather than assert(0) for LOG/EXP opcodes Glean vertProg1 runs all the way through, rather than aborting. --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 34 +++++++++++++++++++++++++++-- 1 file changed, 32 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index f2ed9e0353..ad871d2bdf 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -1516,11 +1516,41 @@ exec_instruction( break; case TGSI_OPCODE_EXP: - assert (0); + printf("TGSI: EXP opcode not implemented\n"); + /* from ARB_v_p: + tmp = ScalarLoad(op0); + result.x = 2^floor(tmp); + result.y = tmp - floor(tmp); + result.z = RoughApprox2ToX(tmp); + result.w = 1.0; + */ +#if 0 + /* something like this: */ + FETCH( &r[0], 0, CHAN_X ); + micro_exp2( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } +#endif break; case TGSI_OPCODE_LOG: - assert (0); + printf("TGSI: LOG opcode not implemented\n"); + /* from ARB_v_p: + tmp = fabs(ScalarLoad(op0)); + result.x = floor(log2(tmp)); + result.y = tmp / 2^(floor(log2(tmp))); + result.z = RoughApproxLog2(tmp); + result.w = 1.0; + */ +#if 0 + /* something like this: */ + FETCH( &r[0], 0, CHAN_X ); + micro_lg2( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } +#endif break; case TGSI_OPCODE_MUL: -- cgit v1.2.3 From 7dc449d406a88fbb92aedfacfa3869176ba2cb31 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 14 Mar 2008 13:50:01 -0600 Subject: mesa: fix emit_clamp() so that we don't use an output register as temporary IR_CLAMP is decomposed into OPCODE_MIN+OPCODE_MAX. Allocate a temporary register for the intermediate value so we don't inadvertantly use an output register (which are write-only on some GPUs). --- src/mesa/shader/slang/slang_emit.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 2b08e7020f..3763b56705 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -677,6 +677,7 @@ static struct prog_instruction * emit_clamp(slang_emit_info *emitInfo, slang_ir_node *n) { struct prog_instruction *inst; + slang_ir_node tmpNode; assert(n->Opcode == IR_CLAMP); /* ch[0] = value @@ -722,18 +723,26 @@ emit_clamp(slang_emit_info *emitInfo, slang_ir_node *n) emit(emitInfo, n->Children[1]); emit(emitInfo, n->Children[2]); + /* Some GPUs don't allow reading from output registers. So if the + * dest for this clamp() is an output reg, we can't use that reg for + * the intermediate result. Use a temp register instead. + */ + alloc_temp_storage(emitInfo, &tmpNode, n->Store->Size); + /* tmp = max(ch[0], ch[1]) */ inst = new_instruction(emitInfo, OPCODE_MAX); - storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); + storage_to_dst_reg(&inst->DstReg, tmpNode.Store, n->Writemask); storage_to_src_reg(&inst->SrcReg[0], n->Children[0]->Store); storage_to_src_reg(&inst->SrcReg[1], n->Children[1]->Store); - /* tmp = min(tmp, ch[2]) */ + /* n->dest = min(tmp, ch[2]) */ inst = new_instruction(emitInfo, OPCODE_MIN); storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); - storage_to_src_reg(&inst->SrcReg[0], n->Store); + storage_to_src_reg(&inst->SrcReg[0], tmpNode.Store); storage_to_src_reg(&inst->SrcReg[1], n->Children[2]->Store); + free_temp_storage(emitInfo->vt, &tmpNode); + return inst; } -- cgit v1.2.3 From a9aef5fccd0a71bc29845c5af9aacdab5e2234ea Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 14 Mar 2008 14:14:13 -0600 Subject: mesa: init tmpNode to zeros --- src/mesa/shader/slang/slang_emit.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 3763b56705..382b13ff05 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -727,6 +727,7 @@ emit_clamp(slang_emit_info *emitInfo, slang_ir_node *n) * dest for this clamp() is an output reg, we can't use that reg for * the intermediate result. Use a temp register instead. */ + _mesa_bzero(&tmpNode, sizeof(tmpNode)); alloc_temp_storage(emitInfo, &tmpNode, n->Store->Size); /* tmp = max(ch[0], ch[1]) */ -- cgit v1.2.3 From 2b8f31a6daf6a52086a3454a5dfd1f8bac046804 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 14 Mar 2008 21:40:02 +0100 Subject: scons: Remove second occurence of draw_passthrough.c. --- src/gallium/auxiliary/draw/SConscript | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 64b444dbd5..d7fb86d992 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -14,10 +14,9 @@ draw = env.ConvenienceLibrary( 'draw_debug.c', 'draw_flatshade.c', 'draw_offset.c', - 'draw_prim.c', 'draw_passthrough.c', + 'draw_prim.c', 'draw_pstipple.c', - 'draw_passthrough.c', 'draw_stipple.c', 'draw_twoside.c', 'draw_unfilled.c', -- cgit v1.2.3 From f23207ca57095b620febaf723815cc3eef3e87bd Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 14 Mar 2008 21:44:06 +0100 Subject: tgsi: Use debug_printf(). --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index ad871d2bdf..78e7dec569 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -1516,7 +1516,7 @@ exec_instruction( break; case TGSI_OPCODE_EXP: - printf("TGSI: EXP opcode not implemented\n"); + debug_printf("TGSI: EXP opcode not implemented\n"); /* from ARB_v_p: tmp = ScalarLoad(op0); result.x = 2^floor(tmp); @@ -1535,7 +1535,7 @@ exec_instruction( break; case TGSI_OPCODE_LOG: - printf("TGSI: LOG opcode not implemented\n"); + debug_printf("TGSI: LOG opcode not implemented\n"); /* from ARB_v_p: tmp = fabs(ScalarLoad(op0)); result.x = floor(log2(tmp)); -- cgit v1.2.3 From 08e341e5dc2e15d8a6c4ba870c9d293295df9467 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Mar 2008 15:54:02 -0600 Subject: gallium: remove DOS carriage returns --- src/gallium/drivers/softpipe/sp_state.h | 390 ++++++++++++++++---------------- 1 file changed, 195 insertions(+), 195 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 45d159143f..0bb1095aec 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -1,195 +1,195 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#ifndef SP_STATE_H -#define SP_STATE_H - -#include "pipe/p_state.h" -#include "tgsi/util/tgsi_scan.h" - - -#define SP_NEW_VIEWPORT 0x1 -#define SP_NEW_RASTERIZER 0x2 -#define SP_NEW_FS 0x4 -#define SP_NEW_BLEND 0x8 -#define SP_NEW_CLIP 0x10 -#define SP_NEW_SCISSOR 0x20 -#define SP_NEW_STIPPLE 0x40 -#define SP_NEW_FRAMEBUFFER 0x80 -#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100 -#define SP_NEW_CONSTANTS 0x200 -#define SP_NEW_SAMPLER 0x400 -#define SP_NEW_TEXTURE 0x800 -#define SP_NEW_VERTEX 0x1000 -#define SP_NEW_VS 0x2000 -#define SP_NEW_QUERY 0x4000 - - -struct tgsi_sampler; -struct tgsi_exec_machine; - - -/** Subclass of pipe_shader_state (though it doesn't really need to be). - * - * This is starting to look an awful lot like a quad pipeline stage... - */ -struct sp_fragment_shader { - struct pipe_shader_state shader; - - struct tgsi_shader_info info; - - void (*prepare)( const struct sp_fragment_shader *shader, - struct tgsi_exec_machine *machine, - struct tgsi_sampler *samplers); - - /* Run the shader - this interface will get cleaned up in the - * future: - */ - unsigned (*run)( const struct sp_fragment_shader *shader, - struct tgsi_exec_machine *machine, - struct quad_header *quad ); - - - void (*delete)( struct sp_fragment_shader * ); -}; - -struct vertex_info; - -/** Subclass of pipe_shader_state */ -struct sp_vertex_shader { - struct pipe_shader_state shader; - struct draw_vertex_shader *draw_data; -}; - - - -void * -softpipe_create_blend_state(struct pipe_context *, - const struct pipe_blend_state *); -void softpipe_bind_blend_state(struct pipe_context *, - void *); -void softpipe_delete_blend_state(struct pipe_context *, - void *); - -void * -softpipe_create_sampler_state(struct pipe_context *, - const struct pipe_sampler_state *); -void softpipe_bind_sampler_states(struct pipe_context *, unsigned, void **); -void softpipe_delete_sampler_state(struct pipe_context *, void *); - -void * -softpipe_create_depth_stencil_state(struct pipe_context *, - const struct pipe_depth_stencil_alpha_state *); -void softpipe_bind_depth_stencil_state(struct pipe_context *, void *); -void softpipe_delete_depth_stencil_state(struct pipe_context *, void *); - -void * -softpipe_create_rasterizer_state(struct pipe_context *, - const struct pipe_rasterizer_state *); -void softpipe_bind_rasterizer_state(struct pipe_context *, void *); -void softpipe_delete_rasterizer_state(struct pipe_context *, void *); - -void softpipe_set_framebuffer_state( struct pipe_context *, - const struct pipe_framebuffer_state * ); - -void softpipe_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ); - -void softpipe_set_clip_state( struct pipe_context *, - const struct pipe_clip_state * ); - -void softpipe_set_constant_buffer(struct pipe_context *, - uint shader, uint index, - const struct pipe_constant_buffer *buf); - -void *softpipe_create_fs_state(struct pipe_context *, - const struct pipe_shader_state *); -void softpipe_bind_fs_state(struct pipe_context *, void *); -void softpipe_delete_fs_state(struct pipe_context *, void *); -void *softpipe_create_vs_state(struct pipe_context *, - const struct pipe_shader_state *); -void softpipe_bind_vs_state(struct pipe_context *, void *); -void softpipe_delete_vs_state(struct pipe_context *, void *); - -void softpipe_set_polygon_stipple( struct pipe_context *, - const struct pipe_poly_stipple * ); - -void softpipe_set_scissor_state( struct pipe_context *, - const struct pipe_scissor_state * ); - -void softpipe_set_sampler_textures( struct pipe_context *, - unsigned num, - struct pipe_texture ** ); - -void softpipe_set_viewport_state( struct pipe_context *, - const struct pipe_viewport_state * ); - -void softpipe_set_vertex_element(struct pipe_context *, - unsigned index, - const struct pipe_vertex_element *); - -void softpipe_set_vertex_buffer(struct pipe_context *, - unsigned index, - const struct pipe_vertex_buffer *); - - -void softpipe_update_derived( struct softpipe_context *softpipe ); - - -boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, - unsigned start, unsigned count); - -boolean softpipe_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count); - - -void -softpipe_map_surfaces(struct softpipe_context *sp); - -void -softpipe_unmap_surfaces(struct softpipe_context *sp); - -void -softpipe_map_texture_surfaces(struct softpipe_context *sp); - -void -softpipe_unmap_texture_surfaces(struct softpipe_context *sp); - - -struct vertex_info * -softpipe_get_vertex_info(struct softpipe_context *softpipe); - -struct vertex_info * -softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe); - - -#endif +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef SP_STATE_H +#define SP_STATE_H + +#include "pipe/p_state.h" +#include "tgsi/util/tgsi_scan.h" + + +#define SP_NEW_VIEWPORT 0x1 +#define SP_NEW_RASTERIZER 0x2 +#define SP_NEW_FS 0x4 +#define SP_NEW_BLEND 0x8 +#define SP_NEW_CLIP 0x10 +#define SP_NEW_SCISSOR 0x20 +#define SP_NEW_STIPPLE 0x40 +#define SP_NEW_FRAMEBUFFER 0x80 +#define SP_NEW_DEPTH_STENCIL_ALPHA 0x100 +#define SP_NEW_CONSTANTS 0x200 +#define SP_NEW_SAMPLER 0x400 +#define SP_NEW_TEXTURE 0x800 +#define SP_NEW_VERTEX 0x1000 +#define SP_NEW_VS 0x2000 +#define SP_NEW_QUERY 0x4000 + + +struct tgsi_sampler; +struct tgsi_exec_machine; + + +/** Subclass of pipe_shader_state (though it doesn't really need to be). + * + * This is starting to look an awful lot like a quad pipeline stage... + */ +struct sp_fragment_shader { + struct pipe_shader_state shader; + + struct tgsi_shader_info info; + + void (*prepare)( const struct sp_fragment_shader *shader, + struct tgsi_exec_machine *machine, + struct tgsi_sampler *samplers); + + /* Run the shader - this interface will get cleaned up in the + * future: + */ + unsigned (*run)( const struct sp_fragment_shader *shader, + struct tgsi_exec_machine *machine, + struct quad_header *quad ); + + + void (*delete)( struct sp_fragment_shader * ); +}; + +struct vertex_info; + +/** Subclass of pipe_shader_state */ +struct sp_vertex_shader { + struct pipe_shader_state shader; + struct draw_vertex_shader *draw_data; +}; + + + +void * +softpipe_create_blend_state(struct pipe_context *, + const struct pipe_blend_state *); +void softpipe_bind_blend_state(struct pipe_context *, + void *); +void softpipe_delete_blend_state(struct pipe_context *, + void *); + +void * +softpipe_create_sampler_state(struct pipe_context *, + const struct pipe_sampler_state *); +void softpipe_bind_sampler_states(struct pipe_context *, unsigned, void **); +void softpipe_delete_sampler_state(struct pipe_context *, void *); + +void * +softpipe_create_depth_stencil_state(struct pipe_context *, + const struct pipe_depth_stencil_alpha_state *); +void softpipe_bind_depth_stencil_state(struct pipe_context *, void *); +void softpipe_delete_depth_stencil_state(struct pipe_context *, void *); + +void * +softpipe_create_rasterizer_state(struct pipe_context *, + const struct pipe_rasterizer_state *); +void softpipe_bind_rasterizer_state(struct pipe_context *, void *); +void softpipe_delete_rasterizer_state(struct pipe_context *, void *); + +void softpipe_set_framebuffer_state( struct pipe_context *, + const struct pipe_framebuffer_state * ); + +void softpipe_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ); + +void softpipe_set_clip_state( struct pipe_context *, + const struct pipe_clip_state * ); + +void softpipe_set_constant_buffer(struct pipe_context *, + uint shader, uint index, + const struct pipe_constant_buffer *buf); + +void *softpipe_create_fs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_fs_state(struct pipe_context *, void *); +void softpipe_delete_fs_state(struct pipe_context *, void *); +void *softpipe_create_vs_state(struct pipe_context *, + const struct pipe_shader_state *); +void softpipe_bind_vs_state(struct pipe_context *, void *); +void softpipe_delete_vs_state(struct pipe_context *, void *); + +void softpipe_set_polygon_stipple( struct pipe_context *, + const struct pipe_poly_stipple * ); + +void softpipe_set_scissor_state( struct pipe_context *, + const struct pipe_scissor_state * ); + +void softpipe_set_sampler_textures( struct pipe_context *, + unsigned num, + struct pipe_texture ** ); + +void softpipe_set_viewport_state( struct pipe_context *, + const struct pipe_viewport_state * ); + +void softpipe_set_vertex_element(struct pipe_context *, + unsigned index, + const struct pipe_vertex_element *); + +void softpipe_set_vertex_buffer(struct pipe_context *, + unsigned index, + const struct pipe_vertex_buffer *); + + +void softpipe_update_derived( struct softpipe_context *softpipe ); + + +boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count); + +boolean softpipe_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); + + +void +softpipe_map_surfaces(struct softpipe_context *sp); + +void +softpipe_unmap_surfaces(struct softpipe_context *sp); + +void +softpipe_map_texture_surfaces(struct softpipe_context *sp); + +void +softpipe_unmap_texture_surfaces(struct softpipe_context *sp); + + +struct vertex_info * +softpipe_get_vertex_info(struct softpipe_context *softpipe); + +struct vertex_info * +softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe); + + +#endif -- cgit v1.2.3 From 344356a0edee932604027386591c82f6666e607c Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Mar 2008 16:00:15 -0600 Subject: gallium: remove DOS carriage returns --- src/gallium/drivers/softpipe/sp_context.c | 486 +++++++++++++++--------------- src/gallium/drivers/softpipe/sp_quad_fs.c | 418 ++++++++++++------------- src/gallium/drivers/softpipe/sp_texture.c | 404 ++++++++++++------------- 3 files changed, 654 insertions(+), 654 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 316ae552b8..16fb06f176 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -1,243 +1,243 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Author: - * Keith Whitwell - */ - -#include "draw/draw_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "sp_clear.h" -#include "sp_context.h" -#include "sp_flush.h" -#include "sp_prim_setup.h" -#include "sp_prim_vbuf.h" -#include "sp_state.h" -#include "sp_surface.h" -#include "sp_tile_cache.h" -#include "sp_texture.h" -#include "sp_winsys.h" -#include "sp_query.h" - - - -/** - * Map any drawing surfaces which aren't already mapped - */ -void -softpipe_map_surfaces(struct softpipe_context *sp) -{ - unsigned i; - - for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); - } - - sp_tile_cache_map_surfaces(sp->zsbuf_cache); -} - - -/** - * Unmap any mapped drawing surfaces - */ -void -softpipe_unmap_surfaces(struct softpipe_context *sp) -{ - uint i; - - for (i = 0; i < sp->framebuffer.num_cbufs; i++) - sp_flush_tile_cache(sp, sp->cbuf_cache[i]); - sp_flush_tile_cache(sp, sp->zsbuf_cache); - - for (i = 0; i < sp->framebuffer.num_cbufs; i++) { - sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); - } - sp_tile_cache_unmap_surfaces(sp->zsbuf_cache); -} - - -static void softpipe_destroy( struct pipe_context *pipe ) -{ - struct softpipe_context *softpipe = softpipe_context( pipe ); - struct pipe_winsys *ws = pipe->winsys; - uint i; - - draw_destroy( softpipe->draw ); - - softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); - softpipe->quad.earlyz->destroy( softpipe->quad.earlyz ); - softpipe->quad.shade->destroy( softpipe->quad.shade ); - softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); - softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); - softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); - softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); - softpipe->quad.coverage->destroy( softpipe->quad.coverage ); - softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); - softpipe->quad.blend->destroy( softpipe->quad.blend ); - softpipe->quad.colormask->destroy( softpipe->quad.colormask ); - softpipe->quad.output->destroy( softpipe->quad.output ); - - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) - sp_destroy_tile_cache(softpipe->cbuf_cache[i]); - sp_destroy_tile_cache(softpipe->zsbuf_cache); - - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) - sp_destroy_tile_cache(softpipe->tex_cache[i]); - - for (i = 0; i < Elements(softpipe->constants); i++) { - if (softpipe->constants[i].buffer) { - pipe_buffer_reference(ws, &softpipe->constants[i].buffer, NULL); - } - } - - FREE( softpipe ); -} - - -struct pipe_context * -softpipe_create( struct pipe_screen *screen, - struct pipe_winsys *pipe_winsys, - struct softpipe_winsys *softpipe_winsys ) -{ - struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); - uint i; - -#if defined(__i386__) || defined(__386__) - softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; -#else - softpipe->use_sse = FALSE; -#endif - - softpipe->dump_fs = GETENV( "GALLIUM_DUMP_FS" ) != NULL; - - softpipe->pipe.winsys = pipe_winsys; - softpipe->pipe.screen = screen; - softpipe->pipe.destroy = softpipe_destroy; - - /* state setters */ - softpipe->pipe.create_blend_state = softpipe_create_blend_state; - softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; - softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; - - softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; - softpipe->pipe.bind_sampler_states = softpipe_bind_sampler_states; - softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; - - softpipe->pipe.create_depth_stencil_alpha_state = softpipe_create_depth_stencil_state; - softpipe->pipe.bind_depth_stencil_alpha_state = softpipe_bind_depth_stencil_state; - softpipe->pipe.delete_depth_stencil_alpha_state = softpipe_delete_depth_stencil_state; - - softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; - softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; - softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; - - softpipe->pipe.create_fs_state = softpipe_create_fs_state; - softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; - softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; - - softpipe->pipe.create_vs_state = softpipe_create_vs_state; - softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; - softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; - - softpipe->pipe.set_blend_color = softpipe_set_blend_color; - softpipe->pipe.set_clip_state = softpipe_set_clip_state; - softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; - softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; - softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; - softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; - softpipe->pipe.set_sampler_textures = softpipe_set_sampler_textures; - softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; - - softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; - softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; - - softpipe->pipe.draw_arrays = softpipe_draw_arrays; - softpipe->pipe.draw_elements = softpipe_draw_elements; - - softpipe->pipe.clear = softpipe_clear; - softpipe->pipe.flush = softpipe_flush; - - softpipe_init_query_funcs( softpipe ); - softpipe_init_texture_funcs( softpipe ); - - /* - * Alloc caches for accessing drawing surfaces and textures. - * Must be before quad stage setup! - */ - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) - softpipe->cbuf_cache[i] = sp_create_tile_cache(); - softpipe->zsbuf_cache = sp_create_tile_cache(); - - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) - softpipe->tex_cache[i] = sp_create_tile_cache(); - - - /* setup quad rendering stages */ - softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); - softpipe->quad.earlyz = sp_quad_earlyz_stage(softpipe); - softpipe->quad.shade = sp_quad_shade_stage(softpipe); - softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); - softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); - softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); - softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); - softpipe->quad.coverage = sp_quad_coverage_stage(softpipe); - softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); - softpipe->quad.blend = sp_quad_blend_stage(softpipe); - softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); - softpipe->quad.output = sp_quad_output_stage(softpipe); - - softpipe->winsys = softpipe_winsys; - - /* - * Create drawing context and plug our rendering stage into it. - */ - softpipe->draw = draw_create(); - assert(softpipe->draw); - softpipe->setup = sp_draw_render_stage(softpipe); - - if (GETENV( "SP_VBUF" ) != NULL) { - sp_init_vbuf(softpipe); - } - else { - draw_set_rasterize_stage(softpipe->draw, softpipe->setup); - } - - /* plug in AA line/point stages */ - draw_install_aaline_stage(softpipe->draw, &softpipe->pipe); - draw_install_aapoint_stage(softpipe->draw, &softpipe->pipe); - -#if USE_DRAW_STAGE_PSTIPPLE - /* Do polygon stipple w/ texture map + frag prog? */ - draw_install_pstipple_stage(softpipe->draw, &softpipe->pipe); -#endif - - sp_init_surface_functions(softpipe); - - return &softpipe->pipe; -} +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Author: + * Keith Whitwell + */ + +#include "draw/draw_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "sp_clear.h" +#include "sp_context.h" +#include "sp_flush.h" +#include "sp_prim_setup.h" +#include "sp_prim_vbuf.h" +#include "sp_state.h" +#include "sp_surface.h" +#include "sp_tile_cache.h" +#include "sp_texture.h" +#include "sp_winsys.h" +#include "sp_query.h" + + + +/** + * Map any drawing surfaces which aren't already mapped + */ +void +softpipe_map_surfaces(struct softpipe_context *sp) +{ + unsigned i; + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) { + sp_tile_cache_map_surfaces(sp->cbuf_cache[i]); + } + + sp_tile_cache_map_surfaces(sp->zsbuf_cache); +} + + +/** + * Unmap any mapped drawing surfaces + */ +void +softpipe_unmap_surfaces(struct softpipe_context *sp) +{ + uint i; + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) + sp_flush_tile_cache(sp, sp->cbuf_cache[i]); + sp_flush_tile_cache(sp, sp->zsbuf_cache); + + for (i = 0; i < sp->framebuffer.num_cbufs; i++) { + sp_tile_cache_unmap_surfaces(sp->cbuf_cache[i]); + } + sp_tile_cache_unmap_surfaces(sp->zsbuf_cache); +} + + +static void softpipe_destroy( struct pipe_context *pipe ) +{ + struct softpipe_context *softpipe = softpipe_context( pipe ); + struct pipe_winsys *ws = pipe->winsys; + uint i; + + draw_destroy( softpipe->draw ); + + softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); + softpipe->quad.earlyz->destroy( softpipe->quad.earlyz ); + softpipe->quad.shade->destroy( softpipe->quad.shade ); + softpipe->quad.alpha_test->destroy( softpipe->quad.alpha_test ); + softpipe->quad.depth_test->destroy( softpipe->quad.depth_test ); + softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); + softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); + softpipe->quad.coverage->destroy( softpipe->quad.coverage ); + softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); + softpipe->quad.blend->destroy( softpipe->quad.blend ); + softpipe->quad.colormask->destroy( softpipe->quad.colormask ); + softpipe->quad.output->destroy( softpipe->quad.output ); + + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + sp_destroy_tile_cache(softpipe->cbuf_cache[i]); + sp_destroy_tile_cache(softpipe->zsbuf_cache); + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + sp_destroy_tile_cache(softpipe->tex_cache[i]); + + for (i = 0; i < Elements(softpipe->constants); i++) { + if (softpipe->constants[i].buffer) { + pipe_buffer_reference(ws, &softpipe->constants[i].buffer, NULL); + } + } + + FREE( softpipe ); +} + + +struct pipe_context * +softpipe_create( struct pipe_screen *screen, + struct pipe_winsys *pipe_winsys, + struct softpipe_winsys *softpipe_winsys ) +{ + struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); + uint i; + +#if defined(__i386__) || defined(__386__) + softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; +#else + softpipe->use_sse = FALSE; +#endif + + softpipe->dump_fs = GETENV( "GALLIUM_DUMP_FS" ) != NULL; + + softpipe->pipe.winsys = pipe_winsys; + softpipe->pipe.screen = screen; + softpipe->pipe.destroy = softpipe_destroy; + + /* state setters */ + softpipe->pipe.create_blend_state = softpipe_create_blend_state; + softpipe->pipe.bind_blend_state = softpipe_bind_blend_state; + softpipe->pipe.delete_blend_state = softpipe_delete_blend_state; + + softpipe->pipe.create_sampler_state = softpipe_create_sampler_state; + softpipe->pipe.bind_sampler_states = softpipe_bind_sampler_states; + softpipe->pipe.delete_sampler_state = softpipe_delete_sampler_state; + + softpipe->pipe.create_depth_stencil_alpha_state = softpipe_create_depth_stencil_state; + softpipe->pipe.bind_depth_stencil_alpha_state = softpipe_bind_depth_stencil_state; + softpipe->pipe.delete_depth_stencil_alpha_state = softpipe_delete_depth_stencil_state; + + softpipe->pipe.create_rasterizer_state = softpipe_create_rasterizer_state; + softpipe->pipe.bind_rasterizer_state = softpipe_bind_rasterizer_state; + softpipe->pipe.delete_rasterizer_state = softpipe_delete_rasterizer_state; + + softpipe->pipe.create_fs_state = softpipe_create_fs_state; + softpipe->pipe.bind_fs_state = softpipe_bind_fs_state; + softpipe->pipe.delete_fs_state = softpipe_delete_fs_state; + + softpipe->pipe.create_vs_state = softpipe_create_vs_state; + softpipe->pipe.bind_vs_state = softpipe_bind_vs_state; + softpipe->pipe.delete_vs_state = softpipe_delete_vs_state; + + softpipe->pipe.set_blend_color = softpipe_set_blend_color; + softpipe->pipe.set_clip_state = softpipe_set_clip_state; + softpipe->pipe.set_constant_buffer = softpipe_set_constant_buffer; + softpipe->pipe.set_framebuffer_state = softpipe_set_framebuffer_state; + softpipe->pipe.set_polygon_stipple = softpipe_set_polygon_stipple; + softpipe->pipe.set_scissor_state = softpipe_set_scissor_state; + softpipe->pipe.set_sampler_textures = softpipe_set_sampler_textures; + softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; + + softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; + softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; + + softpipe->pipe.draw_arrays = softpipe_draw_arrays; + softpipe->pipe.draw_elements = softpipe_draw_elements; + + softpipe->pipe.clear = softpipe_clear; + softpipe->pipe.flush = softpipe_flush; + + softpipe_init_query_funcs( softpipe ); + softpipe_init_texture_funcs( softpipe ); + + /* + * Alloc caches for accessing drawing surfaces and textures. + * Must be before quad stage setup! + */ + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) + softpipe->cbuf_cache[i] = sp_create_tile_cache(); + softpipe->zsbuf_cache = sp_create_tile_cache(); + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + softpipe->tex_cache[i] = sp_create_tile_cache(); + + + /* setup quad rendering stages */ + softpipe->quad.polygon_stipple = sp_quad_polygon_stipple_stage(softpipe); + softpipe->quad.earlyz = sp_quad_earlyz_stage(softpipe); + softpipe->quad.shade = sp_quad_shade_stage(softpipe); + softpipe->quad.alpha_test = sp_quad_alpha_test_stage(softpipe); + softpipe->quad.depth_test = sp_quad_depth_test_stage(softpipe); + softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); + softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); + softpipe->quad.coverage = sp_quad_coverage_stage(softpipe); + softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); + softpipe->quad.blend = sp_quad_blend_stage(softpipe); + softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); + softpipe->quad.output = sp_quad_output_stage(softpipe); + + softpipe->winsys = softpipe_winsys; + + /* + * Create drawing context and plug our rendering stage into it. + */ + softpipe->draw = draw_create(); + assert(softpipe->draw); + softpipe->setup = sp_draw_render_stage(softpipe); + + if (GETENV( "SP_VBUF" ) != NULL) { + sp_init_vbuf(softpipe); + } + else { + draw_set_rasterize_stage(softpipe->draw, softpipe->setup); + } + + /* plug in AA line/point stages */ + draw_install_aaline_stage(softpipe->draw, &softpipe->pipe); + draw_install_aapoint_stage(softpipe->draw, &softpipe->pipe); + +#if USE_DRAW_STAGE_PSTIPPLE + /* Do polygon stipple w/ texture map + frag prog? */ + draw_install_pstipple_stage(softpipe->draw, &softpipe->pipe); +#endif + + sp_init_surface_functions(softpipe); + + return &softpipe->pipe; +} diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 9198198db3..861285101f 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -1,209 +1,209 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Vertices are just an array of floats, with all the attributes - * packed. We currently assume a layout like: - * - * attr[0][0..3] - window position - * attr[1..n][0..3] - remaining attributes. - * - * Attributes are assumed to be 4 floats wide but are packed so that - * all the enabled attributes run contiguously. - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_headers.h" -#include "sp_quad.h" -#include "sp_texture.h" -#include "sp_tex_sample.h" - - -struct quad_shade_stage -{ - struct quad_stage stage; - struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; - struct tgsi_exec_machine machine; - struct tgsi_exec_vector *inputs, *outputs; - int colorOutSlot, depthOutSlot; -}; - - -/** cast wrapper */ -static INLINE struct quad_shade_stage * -quad_shade_stage(struct quad_stage *qs) -{ - return (struct quad_shade_stage *) qs; -} - - - -/** - * Execute fragment shader for the four fragments in the quad. - */ -static void -shade_quad( - struct quad_stage *qs, - struct quad_header *quad ) -{ - struct quad_shade_stage *qss = quad_shade_stage( qs ); - struct softpipe_context *softpipe = qs->softpipe; - struct tgsi_exec_machine *machine = &qss->machine; - - /* Consts do not require 16 byte alignment. */ - machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; - - machine->InterpCoefs = quad->coef; - - /* run shader */ - quad->mask &= softpipe->fs->run( softpipe->fs, - &qss->machine, - quad ); - - /* store result color */ - if (qss->colorOutSlot >= 0) { - /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] - == TGSI_SEMANTIC_COLOR); - memcpy( - quad->outputs.color, - &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); - } - - /* - * XXX the following code for updating quad->outputs.depth - * isn't really needed if we did early z testing. - */ - - /* store result Z */ - if (qss->depthOutSlot >= 0) { - /* output[slot] is new Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; - } - } - else { - /* copy input Z (which was interpolated by the executor) to output Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; - /* XXX not sure the above line is always correct. The following - * might be better: - quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i]; - */ - } - } - - /* shader may cull fragments */ - if( quad->mask ) { - qs->next->run( qs->next, quad ); - } -} - -/** - * Per-primitive (or per-begin?) setup - */ -static void shade_begin(struct quad_stage *qs) -{ - struct quad_shade_stage *qss = quad_shade_stage(qs); - struct softpipe_context *softpipe = qs->softpipe; - unsigned i; - unsigned num = MAX2(softpipe->num_textures, softpipe->num_samplers); - - /* set TGSI sampler state that varies */ - for (i = 0; i < num; i++) { - qss->samplers[i].state = softpipe->sampler[i]; - qss->samplers[i].texture = softpipe->texture[i]; - } - - /* find output slots for depth, color */ - qss->colorOutSlot = -1; - qss->depthOutSlot = -1; - for (i = 0; i < qss->stage.softpipe->fs->info.num_outputs; i++) { - switch (qss->stage.softpipe->fs->info.output_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - qss->depthOutSlot = i; - break; - case TGSI_SEMANTIC_COLOR: - qss->colorOutSlot = i; - break; - } - } - - softpipe->fs->prepare( softpipe->fs, - &qss->machine, - qss->samplers ); - - qs->next->begin(qs->next); -} - - -static void shade_destroy(struct quad_stage *qs) -{ - struct quad_shade_stage *qss = (struct quad_shade_stage *) qs; - - tgsi_exec_machine_free_data(&qss->machine); - FREE( qss->inputs ); - FREE( qss->outputs ); - FREE( qs ); -} - - -struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) -{ - struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); - uint i; - - /* allocate storage for program inputs/outputs, aligned to 16 bytes */ - qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); - qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); - qss->machine.Inputs = align16(qss->inputs); - qss->machine.Outputs = align16(qss->outputs); - - qss->stage.softpipe = softpipe; - qss->stage.begin = shade_begin; - qss->stage.run = shade_quad; - qss->stage.destroy = shade_destroy; - - /* set TGSI sampler state that's constant */ - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { - assert(softpipe->tex_cache[i]); - qss->samplers[i].get_samples = sp_get_samples; - qss->samplers[i].pipe = &softpipe->pipe; - qss->samplers[i].cache = softpipe->tex_cache[i]; - } - - tgsi_exec_machine_init( &qss->machine ); - - return &qss->stage; -} +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Vertices are just an array of floats, with all the attributes + * packed. We currently assume a layout like: + * + * attr[0][0..3] - window position + * attr[1..n][0..3] - remaining attributes. + * + * Attributes are assumed to be 4 floats wide but are packed so that + * all the enabled attributes run contiguously. + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "sp_texture.h" +#include "sp_tex_sample.h" + + +struct quad_shade_stage +{ + struct quad_stage stage; + struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; + struct tgsi_exec_machine machine; + struct tgsi_exec_vector *inputs, *outputs; + int colorOutSlot, depthOutSlot; +}; + + +/** cast wrapper */ +static INLINE struct quad_shade_stage * +quad_shade_stage(struct quad_stage *qs) +{ + return (struct quad_shade_stage *) qs; +} + + + +/** + * Execute fragment shader for the four fragments in the quad. + */ +static void +shade_quad( + struct quad_stage *qs, + struct quad_header *quad ) +{ + struct quad_shade_stage *qss = quad_shade_stage( qs ); + struct softpipe_context *softpipe = qs->softpipe; + struct tgsi_exec_machine *machine = &qss->machine; + + /* Consts do not require 16 byte alignment. */ + machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; + + machine->InterpCoefs = quad->coef; + + /* run shader */ + quad->mask &= softpipe->fs->run( softpipe->fs, + &qss->machine, + quad ); + + /* store result color */ + if (qss->colorOutSlot >= 0) { + /* XXX need to handle multiple color outputs someday */ + assert(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] + == TGSI_SEMANTIC_COLOR); + memcpy( + quad->outputs.color, + &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], + sizeof( quad->outputs.color ) ); + } + + /* + * XXX the following code for updating quad->outputs.depth + * isn't really needed if we did early z testing. + */ + + /* store result Z */ + if (qss->depthOutSlot >= 0) { + /* output[slot] is new Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; + } + } + else { + /* copy input Z (which was interpolated by the executor) to output Z */ + uint i; + for (i = 0; i < 4; i++) { + quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; + /* XXX not sure the above line is always correct. The following + * might be better: + quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i]; + */ + } + } + + /* shader may cull fragments */ + if( quad->mask ) { + qs->next->run( qs->next, quad ); + } +} + +/** + * Per-primitive (or per-begin?) setup + */ +static void shade_begin(struct quad_stage *qs) +{ + struct quad_shade_stage *qss = quad_shade_stage(qs); + struct softpipe_context *softpipe = qs->softpipe; + unsigned i; + unsigned num = MAX2(softpipe->num_textures, softpipe->num_samplers); + + /* set TGSI sampler state that varies */ + for (i = 0; i < num; i++) { + qss->samplers[i].state = softpipe->sampler[i]; + qss->samplers[i].texture = softpipe->texture[i]; + } + + /* find output slots for depth, color */ + qss->colorOutSlot = -1; + qss->depthOutSlot = -1; + for (i = 0; i < qss->stage.softpipe->fs->info.num_outputs; i++) { + switch (qss->stage.softpipe->fs->info.output_semantic_name[i]) { + case TGSI_SEMANTIC_POSITION: + qss->depthOutSlot = i; + break; + case TGSI_SEMANTIC_COLOR: + qss->colorOutSlot = i; + break; + } + } + + softpipe->fs->prepare( softpipe->fs, + &qss->machine, + qss->samplers ); + + qs->next->begin(qs->next); +} + + +static void shade_destroy(struct quad_stage *qs) +{ + struct quad_shade_stage *qss = (struct quad_shade_stage *) qs; + + tgsi_exec_machine_free_data(&qss->machine); + FREE( qss->inputs ); + FREE( qss->outputs ); + FREE( qs ); +} + + +struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) +{ + struct quad_shade_stage *qss = CALLOC_STRUCT(quad_shade_stage); + uint i; + + /* allocate storage for program inputs/outputs, aligned to 16 bytes */ + qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); + qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); + qss->machine.Inputs = align16(qss->inputs); + qss->machine.Outputs = align16(qss->outputs); + + qss->stage.softpipe = softpipe; + qss->stage.begin = shade_begin; + qss->stage.run = shade_quad; + qss->stage.destroy = shade_destroy; + + /* set TGSI sampler state that's constant */ + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + assert(softpipe->tex_cache[i]); + qss->samplers[i].get_samples = sp_get_samples; + qss->samplers[i].pipe = &softpipe->pipe; + qss->samplers[i].cache = softpipe->tex_cache[i]; + } + + tgsi_exec_machine_init( &qss->machine ); + + return &qss->stage; +} diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index bbf2d80308..64bf353aa4 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -1,202 +1,202 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Michel Dänzer - */ - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_inlines.h" -#include "pipe/p_util.h" -#include "pipe/p_winsys.h" - -#include "sp_context.h" -#include "sp_state.h" -#include "sp_texture.h" -#include "sp_tile_cache.h" - - -/* Simple, maximally packed layout. - */ - -static unsigned minify( unsigned d ) -{ - return MAX2(1, d>>1); -} - - -static void -softpipe_texture_layout(struct softpipe_texture * spt) -{ - struct pipe_texture *pt = &spt->base; - unsigned level; - unsigned width = pt->width[0]; - unsigned height = pt->height[0]; - unsigned depth = pt->depth[0]; - - spt->buffer_size = 0; - - for (level = 0; level <= pt->last_level; level++) { - pt->width[level] = width; - pt->height[level] = height; - pt->depth[level] = depth; - - spt->level_offset[level] = spt->buffer_size; - - spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * - ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * - width * pt->cpp; - - width = minify(width); - height = minify(height); - depth = minify(depth); - } -} - - -static struct pipe_texture * -softpipe_texture_create_screen(struct pipe_screen *screen, - const struct pipe_texture *templat) -{ - struct pipe_winsys *ws = screen->winsys; - struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); - if (!spt) - return NULL; - - spt->base = *templat; - spt->base.refcount = 1; - spt->base.screen = screen; - - softpipe_texture_layout(spt); - - spt->buffer = ws->buffer_create(ws, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); - if (!spt->buffer) { - FREE(spt); - return NULL; - } - - assert(spt->base.refcount == 1); - - return &spt->base; -} - - -static void -softpipe_texture_release_screen(struct pipe_screen *screen, - struct pipe_texture **pt) -{ - if (!*pt) - return; - - /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); - */ - if (--(*pt)->refcount <= 0) { - struct softpipe_texture *spt = softpipe_texture(*pt); - - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); - */ - - pipe_buffer_reference(screen->winsys, &spt->buffer, NULL); - - FREE(spt); - } - *pt = NULL; -} - - -static struct pipe_surface * -softpipe_get_tex_surface_screen(struct pipe_screen *screen, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) -{ - struct pipe_winsys *ws = screen->winsys; - struct softpipe_texture *spt = softpipe_texture(pt); - struct pipe_surface *ps; - - assert(level <= pt->last_level); - - ps = ws->surface_alloc(ws); - if (ps) { - assert(ps->refcount); - assert(ps->winsys); - pipe_buffer_reference(ws, &ps->buffer, spt->buffer); - ps->format = pt->format; - ps->cpp = pt->cpp; - ps->width = pt->width[level]; - ps->height = pt->height[level]; - ps->pitch = ps->width; - ps->offset = spt->level_offset[level]; - - if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { - ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - (pt->compressed ? ps->height/4 : ps->height) * - ps->width * ps->cpp; - } - else { - assert(face == 0); - assert(zslice == 0); - } - } - return ps; -} - - -static void -softpipe_texture_update(struct pipe_context *pipe, - struct pipe_texture *texture) -{ - struct softpipe_context *softpipe = softpipe_context(pipe); - uint unit; - for (unit = 0; unit < softpipe->num_textures; unit++) { - if (softpipe->texture[unit] == texture) { - sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); - } - } -} - - -void -softpipe_init_texture_funcs( struct softpipe_context *softpipe ) -{ - softpipe->pipe.texture_update = softpipe_texture_update; -} - - -void -softpipe_init_screen_texture_funcs(struct pipe_screen *screen) -{ - screen->texture_create = softpipe_texture_create_screen; - screen->texture_release = softpipe_texture_release_screen; - screen->get_tex_surface = softpipe_get_tex_surface_screen; -} +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /* + * Authors: + * Keith Whitwell + * Michel Dänzer + */ + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" + +#include "sp_context.h" +#include "sp_state.h" +#include "sp_texture.h" +#include "sp_tile_cache.h" + + +/* Simple, maximally packed layout. + */ + +static unsigned minify( unsigned d ) +{ + return MAX2(1, d>>1); +} + + +static void +softpipe_texture_layout(struct softpipe_texture * spt) +{ + struct pipe_texture *pt = &spt->base; + unsigned level; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; + unsigned depth = pt->depth[0]; + + spt->buffer_size = 0; + + for (level = 0; level <= pt->last_level; level++) { + pt->width[level] = width; + pt->height[level] = height; + pt->depth[level] = depth; + + spt->level_offset[level] = spt->buffer_size; + + spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * + ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * + width * pt->cpp; + + width = minify(width); + height = minify(height); + depth = minify(depth); + } +} + + +static struct pipe_texture * +softpipe_texture_create_screen(struct pipe_screen *screen, + const struct pipe_texture *templat) +{ + struct pipe_winsys *ws = screen->winsys; + struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); + if (!spt) + return NULL; + + spt->base = *templat; + spt->base.refcount = 1; + spt->base.screen = screen; + + softpipe_texture_layout(spt); + + spt->buffer = ws->buffer_create(ws, 32, + PIPE_BUFFER_USAGE_PIXEL, + spt->buffer_size); + if (!spt->buffer) { + FREE(spt); + return NULL; + } + + assert(spt->base.refcount == 1); + + return &spt->base; +} + + +static void +softpipe_texture_release_screen(struct pipe_screen *screen, + struct pipe_texture **pt) +{ + if (!*pt) + return; + + /* + DBG("%s %p refcount will be %d\n", + __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); + */ + if (--(*pt)->refcount <= 0) { + struct softpipe_texture *spt = softpipe_texture(*pt); + + /* + DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); + */ + + pipe_buffer_reference(screen->winsys, &spt->buffer, NULL); + + FREE(spt); + } + *pt = NULL; +} + + +static struct pipe_surface * +softpipe_get_tex_surface_screen(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) +{ + struct pipe_winsys *ws = screen->winsys; + struct softpipe_texture *spt = softpipe_texture(pt); + struct pipe_surface *ps; + + assert(level <= pt->last_level); + + ps = ws->surface_alloc(ws); + if (ps) { + assert(ps->refcount); + assert(ps->winsys); + pipe_buffer_reference(ws, &ps->buffer, spt->buffer); + ps->format = pt->format; + ps->cpp = pt->cpp; + ps->width = pt->width[level]; + ps->height = pt->height[level]; + ps->pitch = ps->width; + ps->offset = spt->level_offset[level]; + + if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { + ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * + (pt->compressed ? ps->height/4 : ps->height) * + ps->width * ps->cpp; + } + else { + assert(face == 0); + assert(zslice == 0); + } + } + return ps; +} + + +static void +softpipe_texture_update(struct pipe_context *pipe, + struct pipe_texture *texture) +{ + struct softpipe_context *softpipe = softpipe_context(pipe); + uint unit; + for (unit = 0; unit < softpipe->num_textures; unit++) { + if (softpipe->texture[unit] == texture) { + sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); + } + } +} + + +void +softpipe_init_texture_funcs( struct softpipe_context *softpipe ) +{ + softpipe->pipe.texture_update = softpipe_texture_update; +} + + +void +softpipe_init_screen_texture_funcs(struct pipe_screen *screen) +{ + screen->texture_create = softpipe_texture_create_screen; + screen->texture_release = softpipe_texture_release_screen; + screen->get_tex_surface = softpipe_get_tex_surface_screen; +} -- cgit v1.2.3 From 5a09ad8248ce452136ed96a3d46532b03c877618 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 14 Mar 2008 16:13:35 -0600 Subject: gallium: add explicit control for point sprites (convert points to textured quads) New draw_enable_point_sprites() function. Fixes spriteblast.c demo --- src/gallium/auxiliary/draw/draw_context.c | 12 ++++++++++++ src/gallium/auxiliary/draw/draw_context.h | 2 ++ src/gallium/auxiliary/draw/draw_private.h | 1 + src/gallium/auxiliary/draw/draw_validate.c | 8 +++++++- src/gallium/auxiliary/draw/draw_wide_point.c | 3 ++- 5 files changed, 24 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index e8473a4c57..4cca965ac1 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -86,6 +86,7 @@ struct draw_context *draw_create( void ) draw->wide_point_threshold = 1000000.0; /* infinity */ draw->wide_line_threshold = 1.0; draw->line_stipple = TRUE; + draw->point_sprite = TRUE; draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ @@ -266,6 +267,17 @@ draw_enable_line_stipple(struct draw_context *draw, boolean enable) } +/** + * Tells draw module whether to convert points to quads for sprite mode. + */ +void +draw_enable_point_sprites(struct draw_context *draw, boolean enable) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->point_sprite = enable; +} + + /** * Ask the draw module for the location/slot of the given vertex attribute in * a post-transformed vertex. diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index e8e2b56bae..dae687e590 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -96,6 +96,8 @@ void draw_wide_line_threshold(struct draw_context *draw, float threshold); void draw_enable_line_stipple(struct draw_context *draw, boolean enable); +void draw_enable_point_sprites(struct draw_context *draw, boolean enable); + boolean draw_use_sse(struct draw_context *draw); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 379b6c3206..1c65c3d1b2 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -239,6 +239,7 @@ struct draw_context float wide_point_threshold; /**< convert pnts to tris if larger than this */ float wide_line_threshold; /**< convert lines to tris if wider than this */ boolean line_stipple; /**< do line stipple? */ + boolean point_sprite; /**< convert points to quads for sprites? */ boolean use_sse; /* If a prim stage introduces new vertex attributes, they'll be stored here diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index 602fa3429d..33e5559508 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -76,6 +76,10 @@ draw_need_pipeline(const struct draw_context *draw) if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) return TRUE; + /* point sprites */ + if (draw->rasterizer->point_sprite && draw->point_sprite) + return TRUE; + /* two-side lighting */ if (draw->rasterizer->light_twoside) return TRUE; @@ -110,7 +114,9 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) && !draw->rasterizer->line_smooth); /* drawing large points? */ - if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) + if (draw->rasterizer->point_sprite && draw->point_sprite) + wide_points = TRUE; + else if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) wide_points = FALSE; else if (draw->rasterizer->point_size > draw->wide_point_threshold) wide_points = TRUE; diff --git a/src/gallium/auxiliary/draw/draw_wide_point.c b/src/gallium/auxiliary/draw/draw_wide_point.c index 65bd50f2b8..c53f7e6cb3 100644 --- a/src/gallium/auxiliary/draw/draw_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_wide_point.c @@ -184,7 +184,8 @@ static void widepoint_first_point( struct draw_stage *stage, wide->half_point_size = 0.5f * draw->rasterizer->point_size; /* XXX we won't know the real size if it's computed by the vertex shader! */ - if (draw->rasterizer->point_size > draw->wide_point_threshold) { + if ((draw->rasterizer->point_size > draw->wide_point_threshold) || + (draw->rasterizer->point_sprite && draw->point_sprite)) { stage->point = widepoint_point; } else { -- cgit v1.2.3 From cbf42c45a15d6a132c7f7d4946c422ee3de35e6c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 14 Mar 2008 17:45:27 -0600 Subject: gallium: if point size not computed per vertex, apply size clamp immediately. Fixes glean pointAtten failure. --- src/mesa/state_tracker/st_atom_rasterizer.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 77cef9236b..17d77f90ae 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -223,7 +223,12 @@ static void update_raster_state( struct st_context *st ) raster->point_size_per_vertex = ctx->VertexProgram.PointSizeEnabled; } } - + if (!raster->point_size_per_vertex) { + /* clamp size now */ + raster->point_size = CLAMP(ctx->Point.Size, + ctx->Point.MinSize, + ctx->Point.MaxSize); + } /* _NEW_LINE */ -- cgit v1.2.3 From c2bf23b8372607a5507a44ea3654c8af1083529a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 14 Mar 2008 17:47:11 -0600 Subject: mesa: clamp point size in vertex program when computing attenuated size --- src/mesa/main/ffvertex_prog.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 04ece4cda8..f648d081c0 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -1398,12 +1398,13 @@ static void build_pointsize( struct tnl_program *p ) /* 1 / sqrt(factor) */ emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut ); -#if 1 +#if 0 /* out = pointSize / sqrt(factor) */ emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size); #else - /* not sure, might make sense to do clamping here, - but it's not done in t_vb_points neither */ + /* this is a good place to clamp the point size since there's likely + * no hardware registers to clamp point size at rasterization time. + */ emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size); emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y)); emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z)); -- cgit v1.2.3 From cb98f71d42e4c714dfb0c3e29d28d8418a1ee86b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 15 Mar 2008 00:55:28 +0000 Subject: gallium: Ensure we don't add null objects to the table, as they mark empty handles. --- src/gallium/auxiliary/util/u_handle_table.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index ab427ee371..d25872972a 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -175,6 +175,10 @@ handle_table_set(struct handle_table *ht, if(!handle || handle > ht->size) return 0; + assert(object); + if(!object) + return 0; + index = handle - 1; /* grow the table if necessary */ -- cgit v1.2.3 From 49a687882a659bd03fd09ca7a7d592818914597a Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 16 Mar 2008 10:33:59 -0600 Subject: gallium: finish remaining prim types for sp_vbuf_draw_arrays() Not totally tested, but easily fixed if glitches are found. --- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 73 +++++++++++++++++++++++++++++ 1 file changed, 73 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index db0913cb2b..e7d0cb2b87 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -207,6 +207,27 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) (struct vertex_header *) ((char *) vertex_buffer + (I) * vertex_size) switch (cvbr->prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < nr; i++) { + prim.v[0] = VERTEX(i); + setup->point( setup, &prim ); + } + break; + case PIPE_PRIM_LINES: + assert(nr % 2 == 0); + for (i = 0; i < nr; i += 2) { + prim.v[0] = VERTEX(i); + prim.v[1] = VERTEX(i + 1); + setup->line( setup, &prim ); + } + break; + case PIPE_PRIM_LINE_STRIP: + for (i = 1; i < nr; i++) { + prim.v[0] = VERTEX(i - 1); + prim.v[1] = VERTEX(i); + setup->line( setup, &prim ); + } + break; case PIPE_PRIM_TRIANGLES: assert(nr % 3 == 0); for (i = 0; i < nr; i += 3) { @@ -217,6 +238,58 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) setup->tri( setup, &prim ); } break; + case PIPE_PRIM_TRIANGLE_STRIP: + assert(nr >= 3); + for (i = 2; i < nr; i++) { + prim.v[0] = VERTEX(i - 2); + prim.v[1] = VERTEX(i - 1); + prim.v[2] = VERTEX(i); + calc_det(&prim); + setup->tri( setup, &prim ); + } + break; + case PIPE_PRIM_TRIANGLE_FAN: + assert(nr >= 3); + for (i = 2; i < nr; i++) { + prim.v[0] = VERTEX(0); + prim.v[1] = VERTEX(i - 1); + prim.v[2] = VERTEX(i); + calc_det(&prim); + setup->tri( setup, &prim ); + } + break; + case PIPE_PRIM_QUADS: + assert(nr % 4 == 0); + for (i = 0; i < nr; i += 4) { + prim.v[0] = VERTEX(i + 0); + prim.v[1] = VERTEX(i + 1); + prim.v[2] = VERTEX(i + 2); + calc_det(&prim); + setup->tri( setup, &prim ); + + prim.v[0] = VERTEX(i + 0); + prim.v[1] = VERTEX(i + 2); + prim.v[2] = VERTEX(i + 3); + calc_det(&prim); + setup->tri( setup, &prim ); + } + break; + case PIPE_PRIM_QUAD_STRIP: + assert(nr >= 4); + for (i = 2; i < nr; i += 2) { + prim.v[0] = VERTEX(i - 2); + prim.v[1] = VERTEX(i); + prim.v[2] = VERTEX(i + 1); + calc_det(&prim); + setup->tri( setup, &prim ); + + prim.v[0] = VERTEX(i - 2); + prim.v[1] = VERTEX(i + 1); + prim.v[2] = VERTEX(i - 1); + calc_det(&prim); + setup->tri( setup, &prim ); + } + break; case PIPE_PRIM_POLYGON: /* draw as tri fan */ for (i = 2; i < nr; i++) { -- cgit v1.2.3 From 9425a702bef7d3f601e9ddc2b801f00a3d52dbb8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 17 Mar 2008 10:10:45 +0000 Subject: gallium: improvements, or extensions at least, to the passthrough path Passthrough is actually more tricky than you'd think... --- src/gallium/auxiliary/draw/draw_passthrough.c | 349 +++++++++++++++++------- src/gallium/auxiliary/draw/draw_vbuf.h | 6 +- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 12 +- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 13 +- 4 files changed, 275 insertions(+), 105 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_passthrough.c b/src/gallium/auxiliary/draw/draw_passthrough.c index d16f056191..fdec6a591b 100644 --- a/src/gallium/auxiliary/draw/draw_passthrough.c +++ b/src/gallium/auxiliary/draw/draw_passthrough.c @@ -85,14 +85,52 @@ fetch_store_general( struct draw_context *draw, const unsigned *pitch = draw->vertex_fetch.pitch; const ubyte **src = draw->vertex_fetch.src_ptr; - for (i = start; i < count; i++) { + for (i = start; i < start + count; i++) { for (j = 0; j < nr_attrs; j++) { + /* vinfo->src_index is the output of the vertex shader + * matching this hw-vertex component. + * + * In passthrough, we require a 1:1 mapping between vertex + * shader outputs and inputs, which in turn correspond to + * vertex elements in the state. So, this is the vertex + * element we're interested in... + */ const uint jj = vinfo->src_index[j]; const enum pipe_format srcFormat = draw->vertex_element[jj].src_format; const ubyte *from = src[jj] + i * pitch[jj]; float attrib[4]; + /* Except... When we're not. Two cases EMIT_HEADER & + * EMIT_1F_PSIZE don't consume an input. Should have some + * method for indicating this, or change the logic here + * somewhat so it doesn't matter. + * + * Just hack this up now, do something better about it later. + */ + if (vinfo->emit[j] == EMIT_HEADER) { + memset(out, 0, sizeof(struct vertex_header)); + out += sizeof(struct vertex_header) / 4; + continue; + } + else if (vinfo->emit[j] == EMIT_1F_PSIZE) { + out[0] = 1.0; /* xxx */ + out += 1; + continue; + } + + + /* The normal fetch/emit code: + */ switch (srcFormat) { + case PIPE_FORMAT_B8G8R8A8_UNORM: + { + ubyte *ub = (ubyte *) from; + attrib[0] = UBYTE_TO_FLOAT(ub[0]); + attrib[1] = UBYTE_TO_FLOAT(ub[1]); + attrib[2] = UBYTE_TO_FLOAT(ub[2]); + attrib[3] = UBYTE_TO_FLOAT(ub[3]); + } + break; case PIPE_FORMAT_R32G32B32A32_FLOAT: { float *f = (float *) from; @@ -130,14 +168,21 @@ fetch_store_general( struct draw_context *draw, } break; default: - abort(); + assert(0); } - /* XXX this will probably only work for softpipe */ + debug_printf("attrib %d: %f %f %f %f\n", j, + attrib[0], attrib[1], attrib[2], attrib[3]); + switch (vinfo->emit[j]) { - case EMIT_HEADER: - memset(out, 0, sizeof(struct vertex_header)); - out += sizeof(struct vertex_header) / 4; + case EMIT_1F: + out[0] = attrib[0]; + out += 1; + break; + case EMIT_2F: + out[0] = attrib[0]; + out[1] = attrib[1]; + out += 2; break; case EMIT_4F: out[0] = attrib[0]; @@ -147,64 +192,15 @@ fetch_store_general( struct draw_context *draw, out += 4; break; default: - abort(); + assert(0); } - } + debug_printf("\n"); } } -/* Example of a fetch/emit passthrough shader which could be - * generated when bypass_clipping is enabled on a passthrough vertex - * shader. - */ -static void fetch_xyz_rgb_st( struct draw_context *draw, - float *out, - unsigned start, - unsigned count ) -{ - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - unsigned i; - - const ubyte *xyzw = src[0] + start * pitch[0]; - const ubyte *rgba = src[1] + start * pitch[1]; - const ubyte *st = src[2] + start * pitch[2]; - - /* loop over vertex attributes (vertex shader inputs) - */ - for (i = 0; i < count; i++) { - { - const float *in = (const float *)xyzw; xyzw += pitch[0]; - /* decode input, encode output. Assume both are float[4] */ - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; - out[3] = in[3]; - } - - { - const float *in = (const float *)rgba; rgba += pitch[1]; - /* decode input, encode output. Assume both are float[4] */ - out[4] = in[0]; - out[5] = in[1]; - out[6] = in[2]; - out[7] = in[3]; - } - - { - const float *in = (const float *)st; st += pitch[2]; - /* decode input, encode output. Assume both are float[2] */ - out[8] = in[0]; - out[9] = in[1]; - } - - out += 10; - } -} - static boolean update_shader( struct draw_context *draw ) { const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); @@ -229,70 +225,166 @@ static boolean update_shader( struct draw_context *draw ) draw->pt.hw_vertex_size = vinfo->size * 4; - /* Just trying to figure out how this would work: - */ - if (draw->rasterizer->bypass_vs || - (nr_attrs == 3 && 0 /* some other tests */)) - { -#if 0 - draw->vertex_fetch.pt_fetch = fetch_xyz_rgb_st; -#else - draw->vertex_fetch.pt_fetch = fetch_store_general; -#endif - /*assert(vinfo->size == 10);*/ + draw->vertex_fetch.pt_fetch = fetch_store_general; + return TRUE; +} + + + + +static boolean split_prim_inplace(unsigned prim, unsigned *first, unsigned *incr) +{ + switch (prim) { + case PIPE_PRIM_POINTS: + *first = 1; + *incr = 1; + return TRUE; + case PIPE_PRIM_LINES: + *first = 2; + *incr = 2; return TRUE; + case PIPE_PRIM_LINE_STRIP: + *first = 2; + *incr = 1; + return TRUE; + case PIPE_PRIM_TRIANGLES: + *first = 3; + *incr = 3; + return TRUE; + case PIPE_PRIM_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + return TRUE; + case PIPE_PRIM_QUADS: + *first = 4; + *incr = 4; + return TRUE; + case PIPE_PRIM_QUAD_STRIP: + *first = 4; + *incr = 2; + return TRUE; + default: + *first = 0; + *incr = 1; /* set to one so that count % incr works */ + return FALSE; } - - return FALSE; } static boolean set_prim( struct draw_context *draw, - unsigned prim ) + unsigned prim, + unsigned count ) { assert(!draw->user.elts); - draw->pt.prim = prim; - switch (prim) { case PIPE_PRIM_LINE_LOOP: + if (count > 1024) + return FALSE; + return draw->render->set_primitive( draw->render, PIPE_PRIM_LINE_STRIP ); + + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + if (count > 1024) + return FALSE; + return draw->render->set_primitive( draw->render, prim ); + case PIPE_PRIM_QUADS: case PIPE_PRIM_QUAD_STRIP: - return FALSE; + return draw->render->set_primitive( draw->render, PIPE_PRIM_TRIANGLES ); + default: - draw->render->set_primitive( draw->render, prim ); - return TRUE; + return draw->render->set_primitive( draw->render, prim ); + break; } + + return TRUE; } -boolean -draw_passthrough_arrays(struct draw_context *draw, - unsigned prim, - unsigned start, - unsigned count) + +#define INDEX(i) (start + (i)) +static void pt_draw_arrays( struct draw_context *draw, + unsigned start, + unsigned length ) { - float *hw_verts; + ushort *tmp = NULL; + unsigned i, j; - if (draw_need_pipeline(draw)) - return FALSE; + switch (draw->pt.prim) { + case PIPE_PRIM_LINE_LOOP: + tmp = MALLOC( sizeof(ushort) * (length + 1) ); - if (!set_prim(draw, prim)) - return FALSE; + for (i = 0; i < length; i++) + tmp[i] = INDEX(i); + tmp[length] = 0; - if (!update_shader(draw)) - return FALSE; + draw->render->draw( draw->render, + tmp, + length+1 ); + break; - hw_verts = draw->render->allocate_vertices( draw->render, - draw->pt.hw_vertex_size, - count ); + + case PIPE_PRIM_QUAD_STRIP: + tmp = MALLOC( sizeof(ushort) * (length / 2 * 6) ); + + for (j = i = 0; i + 3 < length; i += 2, j += 6) { + tmp[j+0] = INDEX(i+0); + tmp[j+1] = INDEX(i+1); + tmp[j+2] = INDEX(i+3); + + tmp[j+3] = INDEX(i+2); + tmp[j+4] = INDEX(i+0); + tmp[j+5] = INDEX(i+3); + } + + if (j) + draw->render->draw( draw->render, tmp, j ); + break; + + case PIPE_PRIM_QUADS: + tmp = MALLOC( sizeof(int) * (length / 4 * 6) ); + + for (j = i = 0; i + 3 < length; i += 4, j += 6) { + tmp[j+0] = INDEX(i+0); + tmp[j+1] = INDEX(i+1); + tmp[j+2] = INDEX(i+3); + + tmp[j+3] = INDEX(i+1); + tmp[j+4] = INDEX(i+2); + tmp[j+5] = INDEX(i+3); + } + + if (j) + draw->render->draw( draw->render, tmp, j ); + break; + + default: + draw->render->draw_arrays( draw->render, + start, + length ); + break; + } + + if (tmp) + FREE(tmp); +} + + + +static boolean do_draw( struct draw_context *draw, + unsigned start, unsigned count ) +{ + float *hw_verts = + draw->render->allocate_vertices( draw->render, + (ushort)draw->pt.hw_vertex_size, + (ushort)count ); if (!hw_verts) return FALSE; - /* Single routine to fetch vertices, run shader and emit HW verts. - * Clipping and viewport transformation are done on hardware. + /* Single routine to fetch vertices and emit HW verts. */ draw->vertex_fetch.pt_fetch( draw, hw_verts, @@ -301,9 +393,9 @@ draw_passthrough_arrays(struct draw_context *draw, /* Draw arrays path to avoid re-emitting index list again and * again. */ - draw->render->draw_arrays( draw->render, - start, - count ); + pt_draw_arrays( draw, + 0, + count ); draw->render->release_vertices( draw->render, @@ -314,3 +406,68 @@ draw_passthrough_arrays(struct draw_context *draw, return TRUE; } + +boolean +draw_passthrough_arrays(struct draw_context *draw, + unsigned prim, + unsigned start, + unsigned count) +{ + unsigned i = 0; + unsigned first, incr; + + //debug_printf("%s prim %d start %d count %d\n", __FUNCTION__, prim, start, count); + + split_prim_inplace(prim, &first, &incr); + + count -= (count - first) % incr; + + debug_printf("%s %d %d %d\n", __FUNCTION__, prim, start, count); + + if (draw_need_pipeline(draw)) + return FALSE; + + debug_printf("%s AAA\n", __FUNCTION__); + + if (!set_prim(draw, prim, count)) + return FALSE; + + /* XXX: need a single value that reflects the most recent call to + * driver->set_primitive: + */ + draw->pt.prim = prim; + + debug_printf("%s BBB\n", __FUNCTION__); + + if (!update_shader(draw)) + return FALSE; + + debug_printf("%s CCC\n", __FUNCTION__); + + /* Chop this up into bite-sized pieces that a driver should be able + * to devour -- problem is we don't have a quick way to query the + * driver on the maximum size for this chunk in the current state. + */ + while (i + first <= count) { + int nr = MIN2( count - i, 1024 ); + + /* snap to prim boundary + */ + nr -= (nr - first) % incr; + + if (!do_draw( draw, start + i, nr )) { + assert(0); + return FALSE; + } + + /* increment allowing for repeated vertices + */ + i += nr - (first - incr); + } + + + debug_printf("%s DONE\n", __FUNCTION__); + return TRUE; +} + + diff --git a/src/gallium/auxiliary/draw/draw_vbuf.h b/src/gallium/auxiliary/draw/draw_vbuf.h index 5e7de905c1..e90f37872a 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.h +++ b/src/gallium/auxiliary/draw/draw_vbuf.h @@ -74,9 +74,11 @@ struct vbuf_render { /** * Notify the renderer of the current primitive when it changes. - * Prim is restricted to TRIANGLES, LINES and POINTS. + * Must succeed for TRIANGLES, LINES and POINTS. Other prims at + * the discretion of the driver, for the benefit of the passthrough + * path. */ - void (*set_primitive)( struct vbuf_render *, unsigned prim ); + boolean (*set_primitive)( struct vbuf_render *, unsigned prim ); /** * DrawElements, note indices are ushort: diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index 9d5f609220..eb64f51943 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -116,7 +116,7 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, } -static void +static boolean i915_vbuf_render_set_primitive( struct vbuf_render *render, unsigned prim ) { @@ -125,15 +125,17 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render, switch(prim) { case PIPE_PRIM_POINTS: i915_render->hwprim = PRIM3D_POINTLIST; - break; + return TRUE; case PIPE_PRIM_LINES: i915_render->hwprim = PRIM3D_LINELIST; - break; + return TRUE; case PIPE_PRIM_TRIANGLES: i915_render->hwprim = PRIM3D_TRILIST; - break; + return TRUE; default: - assert(0); + /* Actually, can handle a lot more just fine... Fixme. + */ + return FALSE; } } diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index e7d0cb2b87..d940718ed2 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -101,11 +101,20 @@ sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, } -static void +static boolean sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); - cvbr->prim = prim; + if (prim == PIPE_PRIM_TRIANGLES || + prim == PIPE_PRIM_LINES || + prim == PIPE_PRIM_POINTS) { + cvbr->prim = prim; + return TRUE; + } + else { + return FALSE; + } + } -- cgit v1.2.3 From f3c490634ffb8cd2c5329ba1ebfe9c738c2b7441 Mon Sep 17 00:00:00 2001 From: Markus Amsler Date: Mon, 17 Mar 2008 08:35:37 -0600 Subject: only set InputsRead bit if input is really used --- src/mesa/shader/arbprogparse.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index e385b9d997..9e5169eff1 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -1576,9 +1576,6 @@ parse_attrib_binding(GLcontext * ctx, const GLubyte ** inst, if (err) { program_error(ctx, Program->Position, "Bad attribute binding"); } - else { - Program->Base.InputsRead |= (1 << *inputReg); - } return err; } @@ -2557,6 +2554,11 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst, return 1; } + /* Add attributes to InputsRead only if they are used the program. + * This avoids the handling of unused ATTRIB declarations in the drivers. */ + if (*File == PROGRAM_INPUT) + Program->Base.InputsRead |= (1 << *Index); + return 0; } -- cgit v1.2.3 From a2ab6930df2c26e19a723b43a757718ecbc89bcd Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 17 Mar 2008 09:42:08 -0600 Subject: gallium: in gen_mipmap, also set sampler lod_bias and do a flush() after rendering each level --- src/mesa/state_tracker/st_gen_mipmap.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 9c4e1032ef..6ae235df6f 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -282,8 +282,13 @@ st_render_mipmap(struct st_context *st, /* * Setup sampler state + * Note: we should only have to set the min/max LOD clamps to ensure + * we grab texels from the right mipmap level. But some hardware + * has trouble with min clamping so we also setting the lod_bias to + * try to work around that. */ sampler.min_lod = sampler.max_lod = srcLevel; + sampler.lod_bias = srcLevel; sampler_cso = pipe->create_sampler_state(pipe, &sampler); pipe->bind_sampler_states(pipe, 1, &sampler_cso); @@ -298,6 +303,7 @@ st_render_mipmap(struct st_context *st, draw_quad(st->ctx); + pipe->flush(pipe, PIPE_FLUSH_WAIT); pipe->delete_sampler_state(pipe, sampler_cso); } -- cgit v1.2.3 From 22b9cc3f5f6c900559332f7a30c9cf869e3490d6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 17 Mar 2008 12:38:24 -0600 Subject: gallium: set min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST Plus, comments, clean-ups. --- src/mesa/state_tracker/st_gen_mipmap.c | 15 +++++---------- 1 file changed, 5 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 6ae235df6f..e878925990 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -252,7 +252,7 @@ st_render_mipmap(struct st_context *st, sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; sampler.normalized_coords = 1; @@ -284,7 +284,7 @@ st_render_mipmap(struct st_context *st, * Setup sampler state * Note: we should only have to set the min/max LOD clamps to ensure * we grab texels from the right mipmap level. But some hardware - * has trouble with min clamping so we also setting the lod_bias to + * has trouble with min clamping so we also set the lod_bias to * try to work around that. */ sampler.min_lod = sampler.max_lod = srcLevel; @@ -294,22 +294,17 @@ st_render_mipmap(struct st_context *st, simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]); - /* - * Setup src texture, override pt->first_level so we sample from - * the right mipmap level. - */ - /*pt->first_level = srcLevel;*/ pipe->set_sampler_textures(pipe, 1, &pt); draw_quad(st->ctx); pipe->flush(pipe, PIPE_FLUSH_WAIT); + + /*pipe->texture_update(pipe, pt); not really needed */ + pipe->delete_sampler_state(pipe, sampler_cso); } - /* restore first_level */ - /*pt->first_level = first_level_save;*/ - /* restore pipe state */ #if 0 cso_set_rasterizer(st->cso_context, &st->state.rasterizer); -- cgit v1.2.3 From e3a747f1fefc147e8dca5729bcc8d68f419c595a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Mar 2008 15:37:42 -0600 Subject: gallium: new util_draw_texquad() function. --- src/gallium/auxiliary/util/u_draw_quad.c | 112 +++++++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_draw_quad.h | 37 ++++++++++ 2 files changed, 149 insertions(+) create mode 100644 src/gallium/auxiliary/util/u_draw_quad.c create mode 100644 src/gallium/auxiliary/util/u_draw_quad.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c new file mode 100644 index 0000000000..79a69de633 --- /dev/null +++ b/src/gallium/auxiliary/util/u_draw_quad.c @@ -0,0 +1,112 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "util/u_draw_quad.h" + + +/** + * Draw screen-aligned textured quad. + */ +void +util_draw_texquad(struct pipe_context *pipe, + float x0, float y0, float x1, float y1, float z) +{ + struct pipe_buffer *vbuf; + struct pipe_vertex_buffer vbuffer; + struct pipe_vertex_element velement; + uint numAttribs = 2, vertexBytes, i, j; + float *v; + + vertexBytes = 4 * (4 * numAttribs * sizeof(float)); + + /* XXX create one-time */ + vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_VERTEX, vertexBytes); + assert(vbuf); + + v = (float *) pipe->winsys->buffer_map(pipe->winsys, vbuf, + PIPE_BUFFER_USAGE_CPU_WRITE); + + /* + * Load vertex buffer + */ + for (i = j = 0; i < 4; i++) { + v[j + 2] = z; /* z */ + v[j + 3] = 1.0; /* w */ + v[j + 6] = 0.0; /* r */ + v[j + 7] = 1.0; /* q */ + j += 8; + } + + v[0] = x0; + v[1] = y0; + v[4] = 0.0; /*s*/ + v[5] = 0.0; /*t*/ + + v[8] = x1; + v[9] = y0; + v[12] = 1.0; + v[13] = 0.0; + + v[16] = x1; + v[17] = y1; + v[20] = 1.0; + v[21] = 1.0; + + v[24] = x0; + v[25] = y1; + v[28] = 0.0; + v[29] = 1.0; + + pipe->winsys->buffer_unmap(pipe->winsys, vbuf); + + /* tell pipe about the vertex buffer */ + vbuffer.buffer = vbuf; + vbuffer.pitch = numAttribs * 4 * sizeof(float); /* vertex size */ + vbuffer.buffer_offset = 0; + pipe->set_vertex_buffer(pipe, 0, &vbuffer); + + /* tell pipe about the vertex attributes */ + for (i = 0; i < numAttribs; i++) { + velement.src_offset = i * 4 * sizeof(float); + velement.vertex_buffer_index = 0; + velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + velement.nr_components = 4; + pipe->set_vertex_element(pipe, i, &velement); + } + + /* draw */ + pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_FAN, 0, 4); + + /* XXX: do one-time */ + pipe_buffer_reference(pipe->winsys, &vbuf, NULL); +} diff --git a/src/gallium/auxiliary/util/u_draw_quad.h b/src/gallium/auxiliary/util/u_draw_quad.h new file mode 100644 index 0000000000..a97f55d2ef --- /dev/null +++ b/src/gallium/auxiliary/util/u_draw_quad.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef U_DRAWQUAD_H +#define U_DRAWQUAD_H + + +extern void +util_draw_texquad(struct pipe_context *pipe, + float x0, float y0, float x1, float y1, float z); + + +#endif -- cgit v1.2.3 From f1b42595d02828bcc617bc88b4ce6cf38a69ddbf Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Mar 2008 15:38:14 -0600 Subject: gallium: new mipmap generation code Based on code from Mesa's state tracker. Still need to implement fallbacks for those texture formats which can't generally be rendered to. --- src/gallium/auxiliary/util/u_gen_mipmap.c | 495 ++++++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_gen_mipmap.h | 63 ++++ 2 files changed, 558 insertions(+) create mode 100644 src/gallium/auxiliary/util/u_gen_mipmap.c create mode 100644 src/gallium/auxiliary/util/u_gen_mipmap.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c new file mode 100644 index 0000000000..f0b822929e --- /dev/null +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -0,0 +1,495 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Mipmap generation utility + * + * @author Brian Paul + */ + + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#include "pipe/p_shader_tokens.h" + +#include "util/u_draw_quad.h" +#include "util/u_gen_mipmap.h" + +#include "tgsi/util/tgsi_build.h" +#include "tgsi/util/tgsi_dump.h" +#include "tgsi/util/tgsi_parse.h" + + +/** + * Make simple fragment shader: + * TEX OUT[0], IN[0], SAMP[0], 2D; + * END; + */ +static void +make_fragment_shader(struct gen_mipmap_state *ctx) +{ + struct pipe_context *pipe = ctx->pipe; + uint maxTokens = 100; + struct tgsi_token *tokens; + struct tgsi_header *header; + struct tgsi_processor *processor; + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + const uint procType = TGSI_PROCESSOR_FRAGMENT; + uint ti = 0; + struct pipe_shader_state shader; + + tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); + + /* shader header + */ + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header(); + + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor( procType, header ); + + ti = 3; + + /* declare TEX[0] input */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; + decl.Semantic.SemanticIndex = 0; + /* XXX this could be linear... */ + decl.Declaration.Interpolate = 1; + decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* declare color[0] output */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; + decl.Semantic.SemanticIndex = 0; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* declare sampler */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_SAMPLER; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* TEX instruction */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_TEX; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + inst.FullDstRegisters[0].DstRegister.Index = 0; + inst.Instruction.NumSrcRegs = 2; + inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; + inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + inst.FullSrcRegisters[0].SrcRegister.Index = 0; + inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + inst.FullSrcRegisters[1].SrcRegister.Index = 0; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + + /* END instruction */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_END; + inst.Instruction.NumDstRegs = 0; + inst.Instruction.NumSrcRegs = 0; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + +#if 0 /*debug*/ + tgsi_dump(tokens, 0); +#endif + + shader.tokens = tokens; + ctx->fs = pipe->create_fs_state(pipe, &shader); +} + + +/** + * Make simple fragment shader: + * MOV OUT[0], IN[0]; + * MOV OUT[1], IN[1]; + * END; + * + * XXX eliminate this when vertex passthrough-mode is more solid. + */ +static void +make_vertex_shader(struct gen_mipmap_state *ctx) +{ + struct pipe_context *pipe = ctx->pipe; + uint maxTokens = 100; + struct tgsi_token *tokens; + struct tgsi_header *header; + struct tgsi_processor *processor; + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + const uint procType = TGSI_PROCESSOR_VERTEX; + uint ti = 0; + struct pipe_shader_state shader; + + tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); + + /* shader header + */ + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header(); + + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor( procType, header ); + + ti = 3; + + /* declare POS input */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + /* + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; + decl.Semantic.SemanticIndex = 0; + */ + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + /* declare TEX[0] input */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + /* + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; + decl.Semantic.SemanticIndex = 0; + */ + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 1; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* declare POS output */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; + decl.Semantic.SemanticIndex = 0; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* declare TEX[0] output */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; + decl.Semantic.SemanticIndex = 0; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 1; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* MOVE out[0], in[0]; # POS */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_MOV; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + inst.FullDstRegisters[0].DstRegister.Index = 0; + inst.Instruction.NumSrcRegs = 1; + inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + inst.FullSrcRegisters[0].SrcRegister.Index = 0; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + + /* MOVE out[1], in[1]; # TEX */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_MOV; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + inst.FullDstRegisters[0].DstRegister.Index = 1; + inst.Instruction.NumSrcRegs = 1; + inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + inst.FullSrcRegisters[0].SrcRegister.Index = 1; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + + /* END instruction */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_END; + inst.Instruction.NumDstRegs = 0; + inst.Instruction.NumSrcRegs = 0; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + +#if 0 /*debug*/ + tgsi_dump(tokens, 0); +#endif + + shader.tokens = tokens; + ctx->vs = pipe->create_vs_state(pipe, &shader); +} + + +/** + * Create a mipmap generation context. + * The idea is to create one of these and re-use it each time we need to + * generate a mipmap. + */ +struct gen_mipmap_state * +util_create_gen_mipmap(struct pipe_context *pipe) +{ + struct pipe_blend_state blend; + struct pipe_depth_stencil_alpha_state depthstencil; + struct pipe_rasterizer_state rasterizer; + struct gen_mipmap_state *ctx; + + ctx = CALLOC_STRUCT(gen_mipmap_state); + if (!ctx) + return NULL; + + ctx->pipe = pipe; + + /* we don't use blending, but need to set valid values */ + memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.colormask = PIPE_MASK_RGBA; + ctx->blend = pipe->create_blend_state(pipe, &blend); + + /* depth/stencil/alpha */ + memset(&depthstencil, 0, sizeof(depthstencil)); + ctx->depthstencil = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil); + + /* rasterizer */ + memset(&rasterizer, 0, sizeof(rasterizer)); + rasterizer.front_winding = PIPE_WINDING_CW; + rasterizer.cull_mode = PIPE_WINDING_NONE; + rasterizer.bypass_clipping = 1; /* bypasses viewport too */ + //rasterizer.bypass_vs = 1; + ctx->rasterizer = pipe->create_rasterizer_state(pipe, &rasterizer); + +#if 0 + /* viewport */ + ctx->viewport.scale[0] = 1.0; + ctx->viewport.scale[1] = 1.0; + ctx->viewport.scale[2] = 1.0; + ctx->viewport.scale[3] = 1.0; + ctx->viewport.translate[0] = 0.0; + ctx->viewport.translate[1] = 0.0; + ctx->viewport.translate[2] = 0.0; + ctx->viewport.translate[3] = 0.0; +#endif + + make_vertex_shader(ctx); + make_fragment_shader(ctx); + + return ctx; +} + + +/** + * Destroy a mipmap generation context + */ +void +util_destroy_gen_mipmap(struct gen_mipmap_state *ctx) +{ + struct pipe_context *pipe = ctx->pipe; + + pipe->delete_blend_state(pipe, ctx->blend); + pipe->delete_depth_stencil_alpha_state(pipe, ctx->depthstencil); + pipe->delete_rasterizer_state(pipe, ctx->rasterizer); + pipe->delete_vs_state(pipe, ctx->vs); + pipe->delete_fs_state(pipe, ctx->fs); + + FREE(ctx); +} + + +#if 0 +static void +simple_viewport(struct pipe_context *pipe, uint width, uint height) +{ + struct pipe_viewport_state vp; + + vp.scale[0] = 0.5 * width; + vp.scale[1] = -0.5 * height; + vp.scale[2] = 1.0; + vp.scale[3] = 1.0; + vp.translate[0] = 0.5 * width; + vp.translate[1] = 0.5 * height; + vp.translate[2] = 0.0; + vp.translate[3] = 0.0; + + pipe->set_viewport_state(pipe, &vp); +} +#endif + + +/** + * Generate mipmap images. It's assumed all needed texture memory is + * already allocated. + * + * \param pt the texture to generate mipmap levels for + * \param face which cube face to generate mipmaps for (0 for non-cube maps) + * \param baseLevel the first mipmap level to use as a src + * \param lastLevel the last mipmap level to generate + */ +void +util_gen_mipmap(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_framebuffer_state fb; + struct pipe_sampler_state sampler; + void *sampler_cso; + uint dstLevel; + uint zslice = 0; + + /* check if we can render in the texture's format */ + if (!screen->is_format_supported(screen, pt->format, PIPE_SURFACE)) { +#if 0 + fallback_gen_mipmap(ctx, pt, face, baseLevel, lastLevel); +#endif + return; + } + + /* init framebuffer state */ + memset(&fb, 0, sizeof(fb)); + fb.num_cbufs = 1; + + /* sampler state */ + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; + sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.normalized_coords = 1; + + /* bind our state */ + pipe->bind_blend_state(pipe, ctx->blend); + pipe->bind_depth_stencil_alpha_state(pipe, ctx->depthstencil); + pipe->bind_rasterizer_state(pipe, ctx->rasterizer); + pipe->bind_vs_state(pipe, ctx->vs); + pipe->bind_fs_state(pipe, ctx->fs); +#if 0 + pipe->set_viewport_state(pipe, &ctx->viewport); +#endif + + /* + * XXX for small mipmap levels, it may be faster to use the software + * fallback path... + */ + for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { + const uint srcLevel = dstLevel - 1; + + /* + * Setup framebuffer / dest surface + */ + fb.cbufs[0] = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); + pipe->set_framebuffer_state(pipe, &fb); + + /* + * Setup sampler state + * Note: we should only have to set the min/max LOD clamps to ensure + * we grab texels from the right mipmap level. But some hardware + * has trouble with min clamping so we also set the lod_bias to + * try to work around that. + */ + sampler.min_lod = sampler.max_lod = srcLevel; + sampler.lod_bias = srcLevel; + sampler_cso = pipe->create_sampler_state(pipe, &sampler); + pipe->bind_sampler_states(pipe, 1, &sampler_cso); + +#if 0 + simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]); +#endif + + pipe->set_sampler_textures(pipe, 1, &pt); + + /* quad coords in window coords (bypassing clipping, viewport mapping) */ + util_draw_texquad(pipe, + 0.0F, 0.0F, /* x0, y0 */ + (float) pt->width[dstLevel], /* x1 */ + (float) pt->height[dstLevel], /* y1 */ + 0.0F); /* z */ + + + pipe->flush(pipe, PIPE_FLUSH_WAIT); + + /*pipe->texture_update(pipe, pt); not really needed */ + + pipe->delete_sampler_state(pipe, sampler_cso); + } + + /* Note: caller must restore pipe/gallium state at this time */ +} diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.h b/src/gallium/auxiliary/util/u_gen_mipmap.h new file mode 100644 index 0000000000..6ee909f0a6 --- /dev/null +++ b/src/gallium/auxiliary/util/u_gen_mipmap.h @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef U_GENMIPMAP_H +#define U_GENMIPMAP_H + +#include "pipe/p_state.h" + + +struct gen_mipmap_state +{ + struct pipe_context *pipe; + + void *blend; + void *depthstencil; + void *rasterizer; + /*struct pipe_viewport_state viewport;*/ + struct pipe_sampler_state *vs; + struct pipe_sampler_state *fs; +}; + + + +extern struct gen_mipmap_state * +util_create_gen_mipmap(struct pipe_context *pipe); + + +extern void +util_destroy_gen_mipmap(struct gen_mipmap_state *ctx); + + + +extern void +util_gen_mipmap(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel); + + +#endif -- cgit v1.2.3 From 11d723cda68d775fa20e43db16ac2ce7e00430b1 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Mar 2008 15:39:19 -0600 Subject: gallium: added new u_draw_quad.c and u_gen_mipmap.c files. --- src/gallium/auxiliary/util/Makefile | 2 ++ src/gallium/auxiliary/util/SConscript | 2 ++ 2 files changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 2016c6fb1f..d11fa6e803 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -7,6 +7,8 @@ C_SOURCES = \ p_debug.c \ p_tile.c \ p_util.c \ + u_draw_quad.c \ + u_gen_mipmap.c \ u_handle_table.c \ u_hash_table.c \ u_mm.c \ diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 154a3eca8c..6d4ba8643d 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -6,6 +6,8 @@ util = env.ConvenienceLibrary( 'p_debug.c', 'p_tile.c', 'p_util.c', + 'u_draw_quad.c', + 'u_gen_mipmap.c', 'u_handle_table.c', 'u_hash_table.c', 'u_mm.c', -- cgit v1.2.3 From 088c6404fcae07dec6dcf16d2cb0777aa7b446ad Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Mar 2008 15:48:13 -0600 Subject: gallium: use new gallium utility code for generating mipmaps --- src/mesa/state_tracker/st_context.h | 13 +- src/mesa/state_tracker/st_gen_mipmap.c | 225 +-------------------------------- 2 files changed, 8 insertions(+), 230 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index e81aebba3d..09bab7bfb1 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -146,18 +146,7 @@ struct st_context struct st_fragment_program *combined_prog; } bitmap; - /** For gen/render mipmap feature */ - struct { - struct pipe_blend_state blend; - struct pipe_depth_stencil_alpha_state depthstencil; - struct pipe_rasterizer_state rasterizer; - - void *blend_cso; - void *depthstencil_cso; - void *rasterizer_cso; - struct st_fragment_program *stfp; - struct st_vertex_program *stvp; - } gen_mipmap; + struct gen_mipmap_state *gen_mipmap; struct cso_context *cso_context; }; diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index e878925990..5f09d9bb06 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -37,6 +37,8 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" +#include "util/u_gen_mipmap.h" + #include "cso_cache/cso_cache.h" #include "cso_cache/cso_context.h" @@ -49,55 +51,6 @@ #include "st_cb_texture.h" - -static struct st_fragment_program * -make_tex_fragment_program(GLcontext *ctx) -{ - struct st_fragment_program *stfp; - struct gl_program *p; - GLuint ic = 0; - - p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); - if (!p) - return NULL; - - p->NumInstructions = 2; - - p->Instructions = _mesa_alloc_instructions(p->NumInstructions); - if (!p->Instructions) { - ctx->Driver.DeleteProgram(ctx, p); - return NULL; - } - _mesa_init_instructions(p->Instructions, p->NumInstructions); - - /* TEX result.color, fragment.texcoord[0], texture[0], 2D; */ - p->Instructions[ic].Opcode = OPCODE_TEX; - p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLR; - p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; - p->Instructions[ic].TexSrcUnit = 0; - p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; - ic++; - - /* END; */ - p->Instructions[ic++].Opcode = OPCODE_END; - - assert(ic == p->NumInstructions); - - p->InputsRead = FRAG_BIT_TEX0; - p->OutputsWritten = (1 << FRAG_RESULT_COLR); - - stfp = (struct st_fragment_program *) p; - - st_translate_fragment_program(ctx->st, stfp, NULL); - - return stfp; -} - - - - /** * one-time init for generate mipmap * XXX Note: there may be other times we need no-op/simple state like this. @@ -106,117 +59,18 @@ make_tex_fragment_program(GLcontext *ctx) void st_init_generate_mipmap(struct st_context *st) { - struct pipe_context *pipe = st->pipe; - struct pipe_blend_state blend; - struct pipe_rasterizer_state rasterizer; - struct pipe_depth_stencil_alpha_state depthstencil; - - /* we don't use blending, but need to set valid values */ - memset(&blend, 0, sizeof(blend)); - blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; - blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; - blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; - blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; - blend.colormask = PIPE_MASK_RGBA; - st->gen_mipmap.blend = blend; - st->gen_mipmap.blend_cso = pipe->create_blend_state(pipe, &blend); - - memset(&depthstencil, 0, sizeof(depthstencil)); - st->gen_mipmap.depthstencil_cso = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil); - - /* Note: we're assuming zero is valid for all non-specified fields */ - memset(&rasterizer, 0, sizeof(rasterizer)); - rasterizer.front_winding = PIPE_WINDING_CW; - rasterizer.cull_mode = PIPE_WINDING_NONE; - st->gen_mipmap.rasterizer_cso = pipe->create_rasterizer_state(pipe, &rasterizer); - - st->gen_mipmap.stfp = make_tex_fragment_program(st->ctx); - st->gen_mipmap.stvp = st_make_passthrough_vertex_shader(st, GL_FALSE); + st->gen_mipmap = util_create_gen_mipmap(st->pipe); } void st_destroy_generate_mipmpap(struct st_context *st) { - struct pipe_context *pipe = st->pipe; - - pipe->delete_blend_state(pipe, st->gen_mipmap.blend_cso); - pipe->delete_depth_stencil_alpha_state(pipe, st->gen_mipmap.depthstencil_cso); - pipe->delete_rasterizer_state(pipe, st->gen_mipmap.rasterizer_cso); - - /* XXX free stfp, stvp */ + util_destroy_gen_mipmap(st->gen_mipmap); + st->gen_mipmap = NULL; } -static void -simple_viewport(struct pipe_context *pipe, uint width, uint height) -{ - struct pipe_viewport_state vp; - - vp.scale[0] = 0.5 * width; - vp.scale[1] = -0.5 * height; - vp.scale[2] = 1.0; - vp.scale[3] = 1.0; - vp.translate[0] = 0.5 * width; - vp.translate[1] = 0.5 * height; - vp.translate[2] = 0.0; - vp.translate[3] = 0.0; - - pipe->set_viewport_state(pipe, &vp); -} - - - -/* - * Draw simple [-1,1]x[-1,1] quad - */ -static void -draw_quad(GLcontext *ctx) -{ - GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ - GLuint i; - GLfloat sLeft = 0.0, sRight = 1.0; - GLfloat tTop = 1.0, tBot = 0.0; - GLfloat x0 = -1.0, x1 = 1.0; - GLfloat y0 = -1.0, y1 = 1.0; - - /* upper-left */ - verts[0][0][0] = x0; /* attr[0].x */ - verts[0][0][1] = y0; /* attr[0].y */ - verts[0][1][0] = sLeft; /* attr[1].s */ - verts[0][1][1] = tTop; /* attr[1].t */ - - /* upper-right */ - verts[1][0][0] = x1; - verts[1][0][1] = y0; - verts[1][1][0] = sRight; - verts[1][1][1] = tTop; - - /* lower-right */ - verts[2][0][0] = x1; - verts[2][0][1] = y1; - verts[2][1][0] = sRight; - verts[2][1][1] = tBot; - - /* lower-left */ - verts[3][0][0] = x0; - verts[3][0][1] = y1; - verts[3][1][0] = sLeft; - verts[3][1][1] = tBot; - - /* same for all verts: */ - for (i = 0; i < 4; i++) { - verts[i][0][2] = 0.0; /*Z*/ - verts[i][0][3] = 1.0; /*W*/ - verts[i][1][2] = 0.0; /*R*/ - verts[i][1][3] = 1.0; /*Q*/ - } - - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, GL_TRUE); -} - - - /** * Generate mipmap levels using hardware rendering. * \return TRUE if successful, FALSE if not possible @@ -229,12 +83,7 @@ st_render_mipmap(struct st_context *st, { struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; - struct pipe_framebuffer_state fb; - struct pipe_sampler_state sampler; - void *sampler_cso; - const uint face = _mesa_tex_target_to_face(target), zslice = 0; - /*const uint first_level_save = pt->first_level;*/ - uint dstLevel; + const uint face = _mesa_tex_target_to_face(target); assert(target != GL_TEXTURE_3D); /* not done yet */ @@ -243,67 +92,7 @@ st_render_mipmap(struct st_context *st, return FALSE; } - /* init framebuffer state */ - memset(&fb, 0, sizeof(fb)); - fb.num_cbufs = 1; - - /* sampler state */ - memset(&sampler, 0, sizeof(sampler)); - sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; - sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; - sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; - sampler.normalized_coords = 1; - - - /* bind state */ - cso_set_blend(st->cso_context, &st->gen_mipmap.blend); - cso_set_depth_stencil_alpha(st->cso_context, &st->gen_mipmap.depthstencil); - cso_set_rasterizer(st->cso_context, &st->gen_mipmap.rasterizer); - - /* bind shaders */ - pipe->bind_fs_state(pipe, st->gen_mipmap.stfp->driver_shader); - pipe->bind_vs_state(pipe, st->gen_mipmap.stvp->driver_shader); - - /* - * XXX for small mipmap levels, it may be faster to use the software - * fallback path... - */ - for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { - const uint srcLevel = dstLevel - 1; - - /* - * Setup framebuffer / dest surface - */ - fb.cbufs[0] = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); - pipe->set_framebuffer_state(pipe, &fb); - - /* - * Setup sampler state - * Note: we should only have to set the min/max LOD clamps to ensure - * we grab texels from the right mipmap level. But some hardware - * has trouble with min clamping so we also set the lod_bias to - * try to work around that. - */ - sampler.min_lod = sampler.max_lod = srcLevel; - sampler.lod_bias = srcLevel; - sampler_cso = pipe->create_sampler_state(pipe, &sampler); - pipe->bind_sampler_states(pipe, 1, &sampler_cso); - - simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]); - - pipe->set_sampler_textures(pipe, 1, &pt); - - draw_quad(st->ctx); - - pipe->flush(pipe, PIPE_FLUSH_WAIT); - - /*pipe->texture_update(pipe, pt); not really needed */ - - pipe->delete_sampler_state(pipe, sampler_cso); - } + util_gen_mipmap(st->gen_mipmap, pt, face, baseLevel, lastLevel); /* restore pipe state */ #if 0 -- cgit v1.2.3 From 0c715de39fa8337a2753dacd77ed280000416c1a Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 17 Mar 2008 15:37:09 -0700 Subject: cell: Fix simple register allocator THere are 64-bits in a uint64_t, not 128. Duh. --- src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c index a996218ce7..842d713f84 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c @@ -326,8 +326,8 @@ int spe_allocate_available_register(struct spe_function *p) { unsigned i; for (i = 0; i < 128; i++) { - const uint64_t mask = (1ULL << (i % 128)); - const unsigned idx = i / 128; + const uint64_t mask = (1ULL << (i % 64)); + const unsigned idx = i / 64; if ((p->regs[idx] & mask) != 0) { p->regs[idx] &= ~mask; @@ -341,8 +341,8 @@ int spe_allocate_available_register(struct spe_function *p) int spe_allocate_register(struct spe_function *p, int reg) { - const unsigned idx = reg / 128; - const unsigned bit = reg % 128; + const unsigned idx = reg / 64; + const unsigned bit = reg % 64; assert((p->regs[idx] & (1ULL << bit)) != 0); @@ -353,8 +353,8 @@ int spe_allocate_register(struct spe_function *p, int reg) void spe_release_register(struct spe_function *p, int reg) { - const unsigned idx = reg / 128; - const unsigned bit = reg % 128; + const unsigned idx = reg / 64; + const unsigned bit = reg % 64; assert((p->regs[idx] & (1ULL << bit)) == 0); -- cgit v1.2.3 From 1936e4bdfd776f78f9fe44f77ce66066fd166360 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 17 Mar 2008 15:45:52 -0700 Subject: cell: Initial code-gen for alpha / stencil / depth testing Alpha test is currently broken because all per-fragment testing occurs before alpha is calculated. Stencil test is currently broken because the Z-clear code asserts if there is a stencil buffer. --- src/gallium/drivers/cell/common.h | 10 + src/gallium/drivers/cell/ppu/Makefile | 1 + src/gallium/drivers/cell/ppu/cell_context.h | 25 +- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 37 +- src/gallium/drivers/cell/ppu/cell_state_emit.c | 24 +- .../drivers/cell/ppu/cell_state_per_fragment.c | 1020 ++++++++++++++++++++ .../drivers/cell/ppu/cell_state_per_fragment.h | 35 + src/gallium/drivers/cell/spu/Makefile | 1 + src/gallium/drivers/cell/spu/spu_main.c | 25 +- src/gallium/drivers/cell/spu/spu_main.h | 16 +- src/gallium/drivers/cell/spu/spu_per_fragment_op.c | 191 ++++ src/gallium/drivers/cell/spu/spu_per_fragment_op.h | 32 + src/gallium/drivers/cell/spu/spu_render.c | 4 +- src/gallium/drivers/cell/spu/spu_tri.c | 23 +- src/gallium/drivers/cell/spu/spu_ztest.h | 135 --- 15 files changed, 1409 insertions(+), 170 deletions(-) create mode 100644 src/gallium/drivers/cell/ppu/cell_state_per_fragment.c create mode 100644 src/gallium/drivers/cell/ppu/cell_state_per_fragment.h create mode 100644 src/gallium/drivers/cell/spu/spu_per_fragment_op.c create mode 100644 src/gallium/drivers/cell/spu/spu_per_fragment_op.h delete mode 100644 src/gallium/drivers/cell/spu/spu_ztest.h (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index 9a4004535e..fe93fd8e1a 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -104,6 +104,16 @@ +/** + */ +struct cell_command_depth_stencil_alpha_test { + uint64_t base; /**< Effective address of code start. */ + unsigned size; /**< Size in bytes of test code. */ + unsigned read_depth; /**< Flag: should depth be read? */ + unsigned read_stencil; /**< Flag: should stencil be read? */ +}; + + /** * Tell SPUs about the framebuffer size, location */ diff --git a/src/gallium/drivers/cell/ppu/Makefile b/src/gallium/drivers/cell/ppu/Makefile index d38fa6ce07..0389a9554c 100644 --- a/src/gallium/drivers/cell/ppu/Makefile +++ b/src/gallium/drivers/cell/ppu/Makefile @@ -27,6 +27,7 @@ SOURCES = \ cell_flush.c \ cell_state_derived.c \ cell_state_emit.c \ + cell_state_per_fragment.c \ cell_state_shader.c \ cell_pipe_state.c \ cell_screen.c \ diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index b221424323..9e79db0ace 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -57,16 +57,37 @@ struct cell_fragment_shader_state }; +struct cell_blend_state { + struct pipe_blend_state base; + + /** + * Generated code to perform alpha blending + */ + struct spe_function code; +}; + + +struct cell_depth_stencil_alpha_state { + struct pipe_depth_stencil_alpha_state base; + + /** + * Generated code to perform alpha, stencil, and depth testing on the SPE + */ + struct spe_function code; + +}; + + struct cell_context { struct pipe_context pipe; struct cell_winsys *winsys; - const struct pipe_blend_state *blend; + const struct cell_blend_state *blend; const struct pipe_sampler_state *sampler[PIPE_MAX_SAMPLERS]; uint num_samplers; - const struct pipe_depth_stencil_alpha_state *depth_stencil; + const struct cell_depth_stencil_alpha_state *depth_stencil; const struct pipe_rasterizer_state *rasterizer; const struct cell_vertex_shader_state *vs; const struct cell_fragment_shader_state *fs; diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 025ed3bbbf..66ede99d13 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -36,6 +36,7 @@ #include "cell_context.h" #include "cell_state.h" #include "cell_texture.h" +#include "cell_state_per_fragment.h" @@ -43,7 +44,12 @@ static void * cell_create_blend_state(struct pipe_context *pipe, const struct pipe_blend_state *blend) { - return mem_dup(blend, sizeof(*blend)); + struct cell_blend_state *cb = MALLOC(sizeof(struct cell_blend_state)); + + (void) memcpy(cb, blend, sizeof(*blend)); + cb->code.store = NULL; + + return cb; } @@ -54,7 +60,7 @@ cell_bind_blend_state(struct pipe_context *pipe, void *blend) draw_flush(cell->draw); - cell->blend = (const struct pipe_blend_state *)blend; + cell->blend = (const struct cell_blend_state *)blend; cell->dirty |= CELL_NEW_BLEND; } @@ -63,7 +69,10 @@ cell_bind_blend_state(struct pipe_context *pipe, void *blend) static void cell_delete_blend_state(struct pipe_context *pipe, void *blend) { - FREE(blend); + struct cell_blend_state *cb = (struct cell_blend_state *) blend; + + spe_release_func(& cb->code); + FREE(cb); } @@ -87,7 +96,13 @@ static void * cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, const struct pipe_depth_stencil_alpha_state *depth_stencil) { - return mem_dup(depth_stencil, sizeof(*depth_stencil)); + struct cell_depth_stencil_alpha_state *cdsa = + MALLOC(sizeof(struct cell_depth_stencil_alpha_state)); + + (void) memcpy(cdsa, depth_stencil, sizeof(*depth_stencil)); + cdsa->code.store = NULL; + + return cdsa; } @@ -96,12 +111,16 @@ cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth_stencil) { struct cell_context *cell = cell_context(pipe); + struct cell_depth_stencil_alpha_state *cdsa = + (struct cell_depth_stencil_alpha_state *) depth_stencil; draw_flush(cell->draw); - cell->depth_stencil - = (const struct pipe_depth_stencil_alpha_state *) depth_stencil; + if (cdsa->code.store == NULL) { + cell_generate_depth_stencil_test(cdsa); + } + cell->depth_stencil = cdsa; cell->dirty |= CELL_NEW_DEPTH_STENCIL; } @@ -109,7 +128,11 @@ cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, static void cell_delete_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth) { - FREE(depth); + struct cell_depth_stencil_alpha_state *cdsa = + (struct cell_depth_stencil_alpha_state *) depth; + + spe_release_func(& cdsa->code); + FREE(cdsa); } diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 670eb26bdd..c8d5fdf709 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -71,9 +71,27 @@ cell_emit_state(struct cell_context *cell) } if (cell->dirty & CELL_NEW_DEPTH_STENCIL) { - emit_state_cmd(cell, CELL_CMD_STATE_DEPTH_STENCIL, - cell->depth_stencil, - sizeof(struct pipe_depth_stencil_alpha_state)); + struct cell_command_depth_stencil_alpha_test dsat; + + + dsat.base = (intptr_t) cell->depth_stencil->code.store; + dsat.size = (char *) cell->depth_stencil->code.csr + - (char *) cell->depth_stencil->code.store; + dsat.read_depth = TRUE; + dsat.read_stencil = FALSE; + + { + uint32_t *p = cell->depth_stencil->code.store; + + printf("\t.text\n"); + for (/* empty */; p < cell->depth_stencil->code.csr; p++) { + printf("\t.long\t0x%04x\n", *p); + } + fflush(stdout); + } + + emit_state_cmd(cell, CELL_CMD_STATE_DEPTH_STENCIL, &dsat, + sizeof(dsat)); } if (cell->dirty & CELL_NEW_SAMPLER) { diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c new file mode 100644 index 0000000000..60b64438d8 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -0,0 +1,1020 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file + * Generate code to perform all per-fragment operations. + * + * Code generated by these functions perform both alpha, depth, and stencil + * testing as well as alpha blending. + * + * \note + * Occlusion query is not supported, but this is the right place to add that + * support. + * + * \author Ian Romanick + */ + +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "cell_context.h" + +#include "rtasm/rtasm_ppc_spe.h" + + +/** + * Generate code to perform alpha testing. + * + * The code generated by this function uses the register specificed by + * \c mask as both an input and an output. + * + * \param dsa Current alpha-test state + * \param f Function to which code should be appended + * \param mask Index of register containing active fragment mask + * \param alphas Index of register containing per-fragment alpha values + * + * \note Emits a maximum of 6 instructions. + */ +static void +emit_alpha_test(struct pipe_depth_stencil_alpha_state *dsa, + struct spe_function *f, int mask, int alphas) +{ + /* If the alpha function is either NEVER or ALWAYS, there is no need to + * load the reference value into a register. ALWAYS is a fairly common + * case, and this optimization saves 2 instructions. + */ + if (dsa->alpha.enabled + && (dsa->alpha.func != PIPE_FUNC_NEVER) + && (dsa->alpha.func != PIPE_FUNC_ALWAYS)) { + int ref = spe_allocate_available_register(f); + int tmp_a = spe_allocate_available_register(f); + int tmp_b = spe_allocate_available_register(f); + union { + float f; + unsigned u; + } ref_val; + boolean complement = FALSE; + + ref_val.f = dsa->alpha.ref; + + spe_il(f, ref, ref_val.u & 0x0000ffff); + spe_ilh(f, ref, ref_val.u >> 16); + + switch (dsa->alpha.func) { + case PIPE_FUNC_NOTEQUAL: + complement = TRUE; + /* FALLTHROUGH */ + + case PIPE_FUNC_EQUAL: + spe_fceq(f, tmp_a, ref, alphas); + break; + + case PIPE_FUNC_LEQUAL: + complement = TRUE; + /* FALLTHROUGH */ + + case PIPE_FUNC_GREATER: + spe_fcgt(f, tmp_a, ref, alphas); + break; + + case PIPE_FUNC_LESS: + complement = TRUE; + /* FALLTHROUGH */ + + case PIPE_FUNC_GEQUAL: + spe_fcgt(f, tmp_a, ref, alphas); + spe_fceq(f, tmp_b, ref, alphas); + spe_or(f, tmp_a, tmp_b, tmp_a); + break; + + case PIPE_FUNC_ALWAYS: + case PIPE_FUNC_NEVER: + default: + assert(0); + break; + } + + if (complement) { + spe_andc(f, mask, mask, tmp_a); + } else { + spe_and(f, mask, mask, tmp_a); + } + + spe_release_register(f, ref); + spe_release_register(f, tmp_a); + spe_release_register(f, tmp_b); + } else if (dsa->alpha.enabled && (dsa->alpha.func == PIPE_FUNC_NEVER)) { + spe_il(f, mask, 0); + } +} + + +/** + * \param dsa Current depth-test state + * \param f Function to which code should be appended + * \param m Mask of allocated / free SPE registers + * \param mask Index of register to contain depth-pass mask + * \param stored Index of register containing values from depth buffer + * \param calculated Index of register containing per-fragment depth values + * + * \return + * If the calculated depth comparison mask is the actual mask, \c FALSE is + * returned. If the calculated depth comparison mask is the compliment of + * the actual mask, \c TRUE is returned. + * + * \note Emits a maximum of 3 instructions. + */ +static boolean +emit_depth_test(struct pipe_depth_stencil_alpha_state *dsa, + struct spe_function *f, int mask, int stored, int calculated) +{ + unsigned func = (dsa->depth.enabled) + ? dsa->depth.func : PIPE_FUNC_ALWAYS; + int tmp = spe_allocate_available_register(f); + boolean compliment = FALSE; + + switch (func) { + case PIPE_FUNC_NEVER: + spe_il(f, mask, 0); + break; + + case PIPE_FUNC_NOTEQUAL: + compliment = TRUE; + /* FALLTHROUGH */ + case PIPE_FUNC_EQUAL: + spe_ceq(f, mask, calculated, stored); + break; + + case PIPE_FUNC_LEQUAL: + compliment = TRUE; + /* FALLTHROUGH */ + case PIPE_FUNC_GREATER: + spe_clgt(f, mask, calculated, stored); + break; + + case PIPE_FUNC_LESS: + compliment = TRUE; + /* FALLTHROUGH */ + case PIPE_FUNC_GEQUAL: + spe_clgt(f, mask, calculated, stored); + spe_ceq(f, tmp, calculated, stored); + spe_or(f, mask, mask, tmp); + break; + + case PIPE_FUNC_ALWAYS: + spe_il(f, mask, ~0); + break; + + default: + assert(0); + break; + } + + spe_release_register(f, tmp); + return compliment; +} + + +/** + * \note Emits a maximum of 5 instructions. + */ +static void +emit_stencil_op(struct spe_function *f, + int out, int in, int mask, unsigned op, unsigned ref) +{ + const int clamp = spe_allocate_available_register(f); + const int tmp = spe_allocate_available_register(f); + + switch(op) { + case PIPE_STENCIL_OP_KEEP: + assert(0); + case PIPE_STENCIL_OP_ZERO: + spe_il(f, out, 0); + break; + case PIPE_STENCIL_OP_REPLACE: + spe_il(f, out, ref); + break; + case PIPE_STENCIL_OP_INCR: + spe_il(f, clamp, 0x0ff); + spe_ai(f, out, in, 1); + spe_cgti(f, tmp, out, clamp); + spe_selb(f, out, out, clamp, tmp); + break; + case PIPE_STENCIL_OP_DECR: + spe_il(f, clamp, 0); + spe_ai(f, out, in, -1); + spe_cgti(f, tmp, out, clamp); + spe_selb(f, out, clamp, out, tmp); + break; + case PIPE_STENCIL_OP_INCR_WRAP: + spe_ai(f, out, in, 1); + break; + case PIPE_STENCIL_OP_DECR_WRAP: + spe_ai(f, out, in, -1); + break; + case PIPE_STENCIL_OP_INVERT: + spe_nor(f, out, in, in); + break; + default: + assert(0); + } + + spe_release_register(f, tmp); + spe_release_register(f, clamp); + + spe_selb(f, out, in, out, mask); +} + + +/** + * \param dsa Depth / stencil test state + * \param face 0 for front face, 1 for back face + * \param f Function to append instructions to + * \param reg_mask Mask of allocated registers + * \param mask Register containing mask of fragments passing the + * alpha test + * \param depth_mask Register containing mask of fragments passing the + * depth test + * \param depth_compliment Is \c depth_mask the compliment of the actual mask? + * \param stencil Register containing values from stencil buffer + * \param depth_pass Register to store mask of fragments passing stencil test + * and depth test + * + * \note + * Emits a maximum of 10 + (3 * 5) = 25 instructions. + */ +static int +emit_stencil_test(struct pipe_depth_stencil_alpha_state *dsa, + unsigned face, + struct spe_function *f, + int mask, + int depth_mask, + boolean depth_complement, + int stencil, + int depth_pass) +{ + int stencil_fail = spe_allocate_available_register(f); + int depth_fail = spe_allocate_available_register(f); + int stencil_mask = spe_allocate_available_register(f); + int stencil_pass = spe_allocate_available_register(f); + int face_stencil = spe_allocate_available_register(f); + int stencil_src = stencil; + const unsigned ref = (dsa->stencil[face].ref_value + & dsa->stencil[face].value_mask); + boolean complement = FALSE; + int stored = spe_allocate_available_register(f); + int tmp = spe_allocate_available_register(f); + + + if ((dsa->stencil[face].func != PIPE_FUNC_NEVER) + && (dsa->stencil[face].func != PIPE_FUNC_ALWAYS) + && (dsa->stencil[face].value_mask != 0x0ff)) { + spe_andi(f, stored, stencil, dsa->stencil[face].value_mask); + } + + + switch (dsa->stencil[face].func) { + case PIPE_FUNC_NEVER: + spe_il(f, stencil_mask, 0); + break; + + case PIPE_FUNC_NOTEQUAL: + complement = TRUE; + /* FALLTHROUGH */ + case PIPE_FUNC_EQUAL: + spe_ceqi(f, stencil_mask, stored, ref); + break; + + case PIPE_FUNC_LEQUAL: + complement = TRUE; + /* FALLTHROUGH */ + case PIPE_FUNC_GREATER: + spe_clgti(f, stencil_mask, stored, ref); + break; + + case PIPE_FUNC_LESS: + complement = TRUE; + /* FALLTHROUGH */ + case PIPE_FUNC_GEQUAL: + spe_clgti(f, stencil_mask, stored, ref); + spe_ceqi(f, tmp, stored, ref); + spe_or(f, stencil_mask, stencil_mask, tmp); + break; + + case PIPE_FUNC_ALWAYS: + /* See comment below. */ + break; + + default: + assert(0); + break; + } + + spe_release_register(f, stored); + spe_release_register(f, tmp); + + + /* ALWAYS is a very common stencil-test, so some effort is applied to + * optimize that case. The stencil-pass mask is the same as the input + * fragment mask. This makes the stencil-test (above) a no-op, and the + * input fragment mask can be "renamed" the stencil-pass mask. + */ + if (dsa->stencil[face].func == PIPE_FUNC_ALWAYS) { + spe_release_register(f, stencil_pass); + stencil_pass = mask; + } else { + if (complement) { + spe_andc(f, stencil_pass, mask, stencil_mask); + } else { + spe_and(f, stencil_pass, mask, stencil_mask); + } + } + + if (depth_complement) { + spe_andc(f, depth_pass, stencil_pass, depth_mask); + } else { + spe_and(f, depth_pass, stencil_pass, depth_mask); + } + + + /* Conditionally emit code to update the stencil value under various + * condititons. Note that there is no need to generate code under the + * following circumstances: + * + * - Stencil write mask is zero. + * - For stencil-fail if the stencil test is ALWAYS + * - For depth-fail if the stencil test is NEVER + * - For depth-pass if the stencil test is NEVER + * - Any of the 3 conditions if the operation is KEEP + */ + if (dsa->stencil[face].write_mask != 0) { + if ((dsa->stencil[face].func != PIPE_FUNC_ALWAYS) + && (dsa->stencil[face].fail_op != PIPE_STENCIL_OP_KEEP)) { + if (complement) { + spe_and(f, stencil_fail, mask, stencil_mask); + } else { + spe_andc(f, stencil_fail, mask, stencil_mask); + } + + emit_stencil_op(f, face_stencil, stencil_src, stencil_fail, + dsa->stencil[face].fail_op, + dsa->stencil[face].ref_value); + + stencil_src = face_stencil; + } + + if ((dsa->stencil[face].func != PIPE_FUNC_NEVER) + && (dsa->stencil[face].zfail_op != PIPE_STENCIL_OP_KEEP)) { + if (depth_complement) { + spe_and(f, depth_fail, stencil_pass, depth_mask); + } else { + spe_andc(f, depth_fail, stencil_pass, depth_mask); + } + + emit_stencil_op(f, face_stencil, stencil_src, depth_fail, + dsa->stencil[face].zfail_op, + dsa->stencil[face].ref_value); + stencil_src = face_stencil; + } + + if ((dsa->stencil[face].func != PIPE_FUNC_NEVER) + && (dsa->stencil[face].zpass_op != PIPE_STENCIL_OP_KEEP)) { + emit_stencil_op(f, face_stencil, stencil_src, depth_pass, + dsa->stencil[face].zpass_op, + dsa->stencil[face].ref_value); + stencil_src = face_stencil; + } + } + + spe_release_register(f, stencil_fail); + spe_release_register(f, depth_fail); + spe_release_register(f, stencil_mask); + if (stencil_pass != mask) { + spe_release_register(f, stencil_pass); + } + + /* If all of the stencil operations were KEEP or the stencil write mask was + * zero, "stencil_src" will still be set to "stencil". In this case + * release the "face_stencil" register. Otherwise apply the stencil write + * mask to select bits from the calculated stencil value and the previous + * stencil value. + */ + if (stencil_src == stencil) { + spe_release_register(f, face_stencil); + } else if (dsa->stencil[face].write_mask != 0x0ff) { + int tmp = spe_allocate_available_register(f); + + spe_il(f, tmp, dsa->stencil[face].write_mask); + spe_selb(f, stencil_src, stencil, stencil_src, tmp); + + spe_release_register(f, tmp); + } + + return stencil_src; +} + + +void +cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa) +{ + struct pipe_depth_stencil_alpha_state *const dsa = &cdsa->base; + struct spe_function *const f = &cdsa->code; + + /* This code generates a maximum of 6 (alpha test) + 3 (depth test) + * + 25 (front stencil) + 25 (back stencil) + 4 = 63 instructions. Round + * up to 64 to make it a happy power-of-two. + */ + spe_init_func(f, 4 * 64); + + + /* Allocate registers for the function's input parameters. Cleverly (and + * clever code is usually dangerous, but I couldn't resist) the generated + * function returns a structure. Returned structures start with register + * 3, and the structure fields are ordered to match up exactly with the + * input parameters. + */ + int mask = spe_allocate_register(f, 3); + int depth = spe_allocate_register(f, 4); + int stencil = spe_allocate_register(f, 5); + int zvals = spe_allocate_register(f, 6); + int frag_a = spe_allocate_register(f, 7); + int facing = spe_allocate_register(f, 8); + + int depth_mask = spe_allocate_available_register(f); + + boolean depth_complement; + + + emit_alpha_test(dsa, f, mask, frag_a); + + depth_complement = emit_depth_test(dsa, f, depth_mask, depth, zvals); + + if (dsa->stencil[0].enabled) { + const int front_depth_pass = spe_allocate_available_register(f); + int front_stencil = emit_stencil_test(dsa, 0, f, mask, + depth_mask, depth_complement, + stencil, front_depth_pass); + + if (dsa->stencil[1].enabled) { + const int back_depth_pass = spe_allocate_available_register(f); + int back_stencil = emit_stencil_test(dsa, 1, f, mask, + depth_mask, depth_complement, + stencil, back_depth_pass); + + /* If the front facing stencil value and the back facing stencil + * value are stored in the same register, there is no need to select + * a value based on the facing. This can happen if the stencil value + * was not modified due to the write masks being zero, the stencil + * operations being KEEP, etc. + */ + if (front_stencil != back_stencil) { + spe_selb(f, stencil, back_stencil, front_stencil, facing); + } + + if (back_stencil != stencil) { + spe_release_register(f, back_stencil); + } + + if (front_stencil != stencil) { + spe_release_register(f, front_stencil); + } + + spe_selb(f, mask, back_depth_pass, front_depth_pass, facing); + + spe_release_register(f, back_depth_pass); + } else { + if (front_stencil != stencil) { + spe_or(f, stencil, front_stencil, front_stencil); + spe_release_register(f, front_stencil); + } + } + + spe_release_register(f, front_depth_pass); + } else if (dsa->depth.enabled) { + if (depth_complement) { + spe_andc(f, mask, mask, depth_mask); + } else { + spe_and(f, mask, mask, depth_mask); + } + } + + if (dsa->depth.writemask) { + spe_selb(f, depth, depth, zvals, mask); + } + + spe_bi(f, 0, 0, 0); +} + + +/** + * \note Emits a maximum of 3 instructions + */ +static int +emit_alpha_factor_calculation(struct spe_function *f, + unsigned factor, float const_alpha, + int src_alpha, int dst_alpha) +{ + union { + float f; + unsigned u; + } alpha; + int factor_reg; + int tmp; + + + alpha.f = const_alpha; + + switch (factor) { + case PIPE_BLENDFACTOR_ONE: + factor_reg = -1; + break; + + case PIPE_BLENDFACTOR_SRC_ALPHA: + factor_reg = spe_allocate_available_register(f); + + spe_or(f, factor_reg, src_alpha, src_alpha); + break; + + case PIPE_BLENDFACTOR_DST_ALPHA: + factor_reg = dst_alpha; + break; + + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + factor_reg = -1; + break; + + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + const_alpha = 1.0 - const_alpha; + /* FALLTHROUGH */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + factor_reg = spe_allocate_available_register(f); + + spe_il(f, factor_reg, alpha.u & 0x0ffff); + spe_ilh(f, factor_reg, alpha.u >> 16); + break; + + case PIPE_BLENDFACTOR_ZERO: + factor_reg = -1; + break; + + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + tmp = spe_allocate_available_register(f); + factor_reg = spe_allocate_available_register(f); + + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + spe_fs(f, factor_reg, tmp, src_alpha); + + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + tmp = spe_allocate_available_register(f); + factor_reg = spe_allocate_available_register(f); + + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + spe_fs(f, factor_reg, tmp, dst_alpha); + + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_SRC1_ALPHA: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + default: + assert(0); + factor_reg = -1; + break; + } + + return factor_reg; +} + + +/** + * \note Emits a maximum of 5 instructions + */ +static void +emit_color_factor_calculation(struct spe_function *f, + unsigned sF, unsigned mask, + const struct pipe_blend_color *blend_color, + const int *src, + const int *dst, + int *factor) +{ + union { + float f[4]; + unsigned u[4]; + } color; + int tmp; + unsigned i; + + + color.f[0] = blend_color->color[0]; + color.f[1] = blend_color->color[1]; + color.f[2] = blend_color->color[2]; + color.f[3] = blend_color->color[3]; + + factor[0] = -1; + factor[1] = -1; + factor[2] = -1; + factor[3] = -1; + + switch (sF) { + case PIPE_BLENDFACTOR_ONE: + break; + + case PIPE_BLENDFACTOR_SRC_COLOR: + for (i = 0; i < 3; ++i) { + if ((mask & (1U << i)) != 0) { + factor[i] = spe_allocate_available_register(f); + spe_or(f, factor[i], src[i], src[i]); + } + } + break; + + case PIPE_BLENDFACTOR_SRC_ALPHA: + factor[0] = spe_allocate_available_register(f); + factor[1] = factor[0]; + factor[2] = factor[0]; + + spe_or(f, factor[0], src[3], src[3]); + break; + + case PIPE_BLENDFACTOR_DST_ALPHA: + factor[0] = dst[3]; + factor[1] = dst[3]; + factor[2] = dst[3]; + break; + + case PIPE_BLENDFACTOR_DST_COLOR: + factor[0] = dst[0]; + factor[1] = dst[1]; + factor[2] = dst[2]; + break; + + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + tmp = spe_allocate_available_register(f); + factor[0] = spe_allocate_available_register(f); + factor[1] = factor[0]; + factor[2] = factor[0]; + + /* Alpha saturate means min(As, 1-Ad). + */ + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + spe_fs(f, tmp, tmp, dst[3]); + spe_fcgt(f, factor[0], tmp, src[3]); + spe_selb(f, factor[0], src[3], tmp, factor[0]); + + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + color.f[0] = 1.0 - color.f[0]; + color.f[1] = 1.0 - color.f[1]; + color.f[2] = 1.0 - color.f[2]; + /* FALLTHROUGH */ + case PIPE_BLENDFACTOR_CONST_COLOR: + for (i = 0; i < 3; i++) { + factor[i] = spe_allocate_available_register(f); + + spe_il(f, factor[i], color.u[i] & 0x0ffff); + spe_ilh(f, factor[i], color.u[i] >> 16); + } + break; + + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + color.f[3] = 1.0 - color.f[3]; + /* FALLTHROUGH */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + factor[0] = spe_allocate_available_register(f); + factor[1] = factor[0]; + factor[2] = factor[0]; + + spe_il(f, factor[0], color.u[3] & 0x0ffff); + spe_ilh(f, factor[0], color.u[3] >> 16); + break; + + case PIPE_BLENDFACTOR_ZERO: + break; + + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + tmp = spe_allocate_available_register(f); + + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + + for (i = 0; i < 3; ++i) { + if ((mask & (1U << i)) != 0) { + factor[i] = spe_allocate_available_register(f); + spe_fs(f, factor[i], tmp, src[i]); + } + } + + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + tmp = spe_allocate_available_register(f); + factor[0] = spe_allocate_available_register(f); + factor[1] = factor[0]; + factor[2] = factor[0]; + + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + spe_fs(f, factor[0], tmp, src[3]); + + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + tmp = spe_allocate_available_register(f); + factor[0] = spe_allocate_available_register(f); + factor[1] = factor[0]; + factor[2] = factor[0]; + + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + spe_fs(f, factor[0], tmp, dst[3]); + + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_INV_DST_COLOR: + tmp = spe_allocate_available_register(f); + + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + + for (i = 0; i < 3; ++i) { + if ((mask & (1U << i)) != 0) { + factor[i] = spe_allocate_available_register(f); + spe_fs(f, factor[i], tmp, dst[i]); + } + } + + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + default: + assert(0); + } +} + + +static void +emit_blend_calculation(struct spe_function *f, + unsigned func, unsigned sF, unsigned dF, + int src, int src_factor, int dst, int dst_factor) +{ + int tmp = spe_allocate_available_register(f); + + switch (func) { + case PIPE_BLEND_ADD: + if (sF == PIPE_BLENDFACTOR_ONE) { + if (dF == PIPE_BLENDFACTOR_ZERO) { + /* Do nothing. */ + } else if (dF == PIPE_BLENDFACTOR_ONE) { + spe_fa(f, src, src, dst); + } + } else if (sF == PIPE_BLENDFACTOR_ZERO) { + if (dF == PIPE_BLENDFACTOR_ZERO) { + spe_il(f, src, 0); + } else if (dF == PIPE_BLENDFACTOR_ONE) { + spe_or(f, src, dst, dst); + } + } else { + spe_fm(f, tmp, dst, dst_factor); + spe_fma(f, src, src, src_factor, tmp); + } + break; + + case PIPE_BLEND_SUBTRACT: + if (sF == PIPE_BLENDFACTOR_ONE) { + if (dF == PIPE_BLENDFACTOR_ZERO) { + /* Do nothing. */ + } else if (dF == PIPE_BLENDFACTOR_ONE) { + spe_fs(f, src, src, dst); + } + } else if (sF == PIPE_BLENDFACTOR_ZERO) { + if (dF == PIPE_BLENDFACTOR_ZERO) { + spe_il(f, src, 0); + } else if (dF == PIPE_BLENDFACTOR_ONE) { + spe_il(f, tmp, 0); + spe_fs(f, src, tmp, dst); + } + } else { + spe_fm(f, tmp, dst, dst_factor); + spe_fms(f, src, src, src_factor, tmp); + } + break; + + case PIPE_BLEND_REVERSE_SUBTRACT: + if (sF == PIPE_BLENDFACTOR_ONE) { + if (dF == PIPE_BLENDFACTOR_ZERO) { + spe_il(f, tmp, 0); + spe_fs(f, src, tmp, src); + } else if (dF == PIPE_BLENDFACTOR_ONE) { + spe_fs(f, src, dst, src); + } + } else if (sF == PIPE_BLENDFACTOR_ZERO) { + if (dF == PIPE_BLENDFACTOR_ZERO) { + spe_il(f, src, 0); + } else if (dF == PIPE_BLENDFACTOR_ONE) { + spe_or(f, src, dst, dst); + } + } else { + spe_fm(f, tmp, src, src_factor); + spe_fms(f, src, src, dst_factor, tmp); + } + break; + + case PIPE_BLEND_MIN: + spe_cgt(f, tmp, src, dst); + spe_selb(f, src, dst, src, tmp); + break; + + case PIPE_BLEND_MAX: + spe_cgt(f, tmp, src, dst); + spe_selb(f, src, src, dst, tmp); + break; + + default: + assert(0); + } + + spe_release_register(f, tmp); +} + + +/** + * Generate code to perform alpha blending on the SPE + */ +void +cell_generate_alpha_blend(struct cell_blend_state *cb, + const struct pipe_blend_color *blend_color) +{ + struct pipe_blend_state *const b = &cb->base; + struct spe_function *const f = &cb->code; + + /* This code generates a maximum of 3 (source alpha factor) + * + 3 (destination alpha factor) + (3 * 5) (source color factor) + * + (3 * 5) (destination color factor) + (4 * 2) (blend equation) + * + 4 (fragment mask) + 1 (return) = 49 instlructions. Round up to 64 to + * make it a happy power-of-two. + */ + spe_init_func(f, 4 * 64); + + + const int frag[4] = { + spe_allocate_register(f, 3), + spe_allocate_register(f, 4), + spe_allocate_register(f, 5), + spe_allocate_register(f, 6), + }; + const int pixel[4] = { + spe_allocate_register(f, 7), + spe_allocate_register(f, 8), + spe_allocate_register(f, 9), + spe_allocate_register(f, 10), + }; + const int mask = spe_allocate_register(f, 11); + unsigned func[4]; + unsigned sF[4]; + unsigned dF[4]; + unsigned i; + int src_factor[4]; + int dst_factor[4]; + + + /* Does the selected blend mode make use of the source / destination + * color (RGB) blend factors? + */ + boolean need_color_factor = b->blend_enable + && (b->rgb_func != PIPE_BLEND_MIN) + && (b->rgb_func != PIPE_BLEND_MAX); + + /* Does the selected blend mode make use of the source / destination + * alpha blend factors? + */ + boolean need_alpha_factor = b->blend_enable + && (b->alpha_func != PIPE_BLEND_MIN) + && (b->alpha_func != PIPE_BLEND_MAX); + + + sF[0] = b->rgb_src_factor; + sF[1] = sF[0]; + sF[2] = sF[0]; + sF[3] = (b->alpha_src_factor == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE) + ? PIPE_BLENDFACTOR_ONE : b->alpha_src_factor; + + dF[0] = b->rgb_dst_factor; + dF[1] = dF[0]; + dF[2] = dF[0]; + dF[3] = b->rgb_dst_factor; + + + /* If alpha writing is enabled and the alpha blend mode requires use of + * the alpha factor, calculate the alpha factor. + */ + if (((b->colormask & 8) != 0) && need_alpha_factor) { + src_factor[3] = emit_alpha_factor_calculation(f, sF[3], + blend_color->color[3], + frag[3], pixel[3]); + + /* If the alpha destination blend factor is the same as the alpha source + * blend factor, re-use the previously calculated value. + */ + dst_factor[3] = (dF[3] == sF[3]) + ? src_factor[3] + : emit_alpha_factor_calculation(f, dF[3], + blend_color->color[3], + frag[3], pixel[3]); + } + + + if (sF[0] == sF[3]) { + src_factor[0] = src_factor[3]; + src_factor[1] = src_factor[3]; + src_factor[2] = src_factor[3]; + } else if (sF[0] == dF[3]) { + src_factor[0] = dst_factor[3]; + src_factor[1] = dst_factor[3]; + src_factor[2] = dst_factor[3]; + } else if (need_color_factor) { + emit_color_factor_calculation(f, + b->rgb_src_factor, + b->colormask, + blend_color, + frag, pixel, src_factor); + } + + + if (dF[0] == sF[3]) { + dst_factor[0] = src_factor[3]; + dst_factor[1] = src_factor[3]; + dst_factor[2] = src_factor[3]; + } else if (dF[0] == dF[3]) { + dst_factor[0] = dst_factor[3]; + dst_factor[1] = dst_factor[3]; + dst_factor[2] = dst_factor[3]; + } else if (dF[0] == sF[0]) { + dst_factor[0] = src_factor[0]; + dst_factor[1] = src_factor[1]; + dst_factor[2] = src_factor[2]; + } else if (need_color_factor) { + emit_color_factor_calculation(f, + b->rgb_dst_factor, + b->colormask, + blend_color, + frag, pixel, dst_factor); + } + + + + func[0] = b->rgb_func; + func[1] = func[0]; + func[2] = func[0]; + func[3] = b->alpha_func; + + for (i = 0; i < 4; ++i) { + if ((b->colormask & (1U << i)) != 0) { + emit_blend_calculation(f, + func[i], sF[i], dF[i], + frag[i], src_factor[i], + pixel[i], dst_factor[i]); + spe_selb(f, frag[i], pixel[i], frag[i], mask); + } else { + spe_or(f, frag[i], pixel[i], pixel[i]); + } + } + + spe_bi(f, 0, 0, 0); +} diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h new file mode 100644 index 0000000000..541c3b3be0 --- /dev/null +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h @@ -0,0 +1,35 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef CELL_STATE_PER_FRAGMENT_H +#define CELL_STATE_PER_FRAGMENT_H + +extern void +cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa); + +extern void +cell_generate_alpha_blend(struct cell_blend_state *cb, + const struct pipe_blend_color *blend_color); + +#endif /* CELL_STATE_PER_FRAGMENT_H */ diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile index c071de1900..115ca8cd90 100644 --- a/src/gallium/drivers/cell/spu/Makefile +++ b/src/gallium/drivers/cell/spu/Makefile @@ -19,6 +19,7 @@ SOURCES = \ spu_main.c \ spu_blend.c \ spu_dcache.c \ + spu_per_fragment_op.c \ spu_render.c \ spu_texture.c \ spu_tile.c \ diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 59300028d4..8e46f6e471 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -58,6 +58,9 @@ struct spu_vs_context draw; static unsigned char attribute_fetch_code_buffer[136 * PIPE_ATTRIB_MAX] ALIGN16_ATTRIB; +static unsigned char depth_stencil_code_buffer[4 * 64] + ALIGN16_ATTRIB; + /** * Tell the PPU that this SPU has finished copying a buffer to * local store and that it may be reused by the PPU. @@ -248,14 +251,26 @@ cmd_state_blend(const struct pipe_blend_state *state) static void -cmd_state_depth_stencil(const struct pipe_depth_stencil_alpha_state *state) +cmd_state_depth_stencil(const struct cell_command_depth_stencil_alpha_test *state) { if (Debug) printf("SPU %u: DEPTH_STENCIL: ztest %d\n", spu.init.id, - state->depth.enabled); + state->read_depth); + + ASSERT_ALIGN16(state->base); + + mfc_get(depth_stencil_code_buffer, + (unsigned int) state->base, /* src */ + ROUNDUP16(state->size), + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); - memcpy(&spu.depth_stencil, state, sizeof(*state)); + spu.frag_test = (frag_test_func) depth_stencil_code_buffer; + spu.read_depth = state->read_depth; + spu.read_stencil = state->read_stencil; } @@ -415,9 +430,9 @@ cmd_batch(uint opcode) pos += (1 + ROUNDUP8(sizeof(struct pipe_blend_state)) / 8); break; case CELL_CMD_STATE_DEPTH_STENCIL: - cmd_state_depth_stencil((struct pipe_depth_stencil_alpha_state *) + cmd_state_depth_stencil((struct cell_command_depth_stencil_alpha_test *) &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct pipe_depth_stencil_alpha_state)) / 8); + pos += (1 + ROUNDUP8(sizeof(struct cell_command_depth_stencil_alpha_test)) / 8); break; case CELL_CMD_STATE_SAMPLER: cmd_state_sampler((struct pipe_sampler_state *) &buffer[pos+1]); diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index a13edd1702..444e218645 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -56,6 +56,17 @@ typedef union { #define TILE_STATUS_GETTING 5 /**< mfc_get() called but not yet arrived */ +struct spu_frag_test_results { + qword mask; + qword depth; + qword stencil; +}; + +typedef struct spu_frag_test_results (*frag_test_func)(qword frag_mask, + qword pixel_depth, qword pixel_stencil, qword frag_depth, + qword frag_alpha, qword facing); + + struct spu_framebuffer { void *color_start; /**< addr of color surface in main memory */ void *depth_start; /**< addr of depth surface in main memory */ @@ -79,8 +90,9 @@ struct spu_global struct cell_init_info init; struct spu_framebuffer fb; - struct pipe_blend_state blend_stencil; - struct pipe_depth_stencil_alpha_state depth_stencil; + boolean read_depth; + boolean read_stencil; + frag_test_func frag_test; struct pipe_blend_state blend; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct cell_command_texture texture; diff --git a/src/gallium/drivers/cell/spu/spu_per_fragment_op.c b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c new file mode 100644 index 0000000000..d42b522b41 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c @@ -0,0 +1,191 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file spu_per_fragment_op.c + * SPU implementation various per-fragment operations. + * + * \author Ian Romanick + */ + +#include "pipe/p_format.h" +#include "spu_main.h" +#include "spu_per_fragment_op.h" + +#define ZERO 0x80 + +static void +read_ds_quad(tile_t *buffer, unsigned x, unsigned y, + enum pipe_format depth_format, qword *depth, + qword *stencil) +{ + const int ix = x / 2; + const int iy = y / 2; + + switch (depth_format) { + case PIPE_FORMAT_Z16_UNORM: { + qword *ptr = (qword *) &buffer->us8[iy][ix / 2]; + + const qword shuf_vec = (qword) { + ZERO, ZERO, 0, 1, ZERO, ZERO, 2, 3, + ZERO, ZERO, 4, 5, ZERO, ZERO, 6, 7 + }; + + + /* At even X values we want the first 4 shorts, and at odd X values we + * want the second 4 shorts. + */ + qword bias = (qword) spu_splats((unsigned char) ((ix & 0x01) << 3)); + qword bias_mask = si_fsmbi(0x3333); + qword sv = si_a(shuf_vec, si_and(bias_mask, bias)); + + *depth = si_shufb(*ptr, *ptr, sv); + *stencil = si_il(0); + break; + } + + + case PIPE_FORMAT_Z32_UNORM: { + qword *ptr = (qword *) &buffer->ui4[iy][ix]; + + *depth = *ptr; + *stencil = si_il(0); + break; + } + + + case PIPE_FORMAT_Z24S8_UNORM: { + qword *ptr = (qword *) &buffer->ui4[iy][ix]; + qword mask = si_fsmbi(0x7777); + + *depth = si_and(*ptr, mask); + *stencil = si_rotmai(si_andc(*ptr, mask), -24); + break; + } + + + default: + assert(0); + break; + } +} + + +static void +write_ds_quad(tile_t *buffer, unsigned x, unsigned y, + enum pipe_format depth_format, + qword depth, qword stencil) +{ + const int ix = x / 2; + const int iy = y / 2; + + (void) stencil; + + switch (depth_format) { + case PIPE_FORMAT_Z16_UNORM: { + qword *ptr = (qword *) &buffer->us8[iy][ix / 2]; + + qword sv = ((ix & 0x01) == 0) + ? (qword) { 2, 3, 6, 7, 10, 11, 14, 15, + 24, 25, 26, 27, 28, 29, 30, 31 } + : (qword) { 16, 17, 18, 19, 20 , 21, 22, 23, + 2, 3, 6, 7, 10, 11, 14, 15 }; + *ptr = si_shufb(depth, *ptr, sv); + break; + } + + + case PIPE_FORMAT_Z32_UNORM: { + qword *ptr = (qword *) &buffer->ui4[iy][ix]; + *ptr = depth; + break; + } + + + case PIPE_FORMAT_Z24S8_UNORM: { + qword *ptr = (qword *) &buffer->ui4[iy][ix]; + qword mask = si_fsmbi(0x7777); + + stencil = si_rotmai(stencil, 24); + *ptr = si_selb(stencil, depth, mask); + break; + } + + + default: + assert(0); + break; + } +} + + +qword +spu_do_depth_stencil(int x, int y, + qword frag_mask, qword frag_depth, qword frag_alpha, + qword facing) +{ + struct spu_frag_test_results result; + qword pixel_depth; + qword pixel_stencil; + + /* All of this preable code (everthing before the call to frag_test) should + * be generated on the PPU and upload to the SPU. + */ + if (spu.read_depth || spu.read_stencil) { + read_ds_quad(&spu.ztile, x, y, spu.fb.depth_format, + &pixel_depth, &pixel_stencil); + } + + switch (spu.fb.depth_format) { + case PIPE_FORMAT_Z16_UNORM: + frag_depth = si_fm(frag_depth, (qword)spu_splats((float)(0x0000ffffu))); + frag_depth = si_cfltu(frag_depth, 0); + break; + case PIPE_FORMAT_Z32_UNORM: + frag_depth = si_fm(frag_depth, (qword)spu_splats((float)(0xffffffffu))); + frag_depth = si_cfltu(frag_depth, 0); + break; + case PIPE_FORMAT_Z24S8_UNORM: + frag_depth = si_fm(frag_depth, (qword)spu_splats((float)(0x00ffffffu))); + frag_depth = si_cfltu(frag_depth, 0); + break; + default: + assert(0); + break; + } + + result = (*spu.frag_test)(frag_mask, pixel_depth, pixel_stencil, + frag_depth, frag_alpha, facing); + + + /* This code (everthing after the call to frag_test) should + * be generated on the PPU and upload to the SPU. + */ + if (spu.read_depth || spu.read_stencil) { + write_ds_quad(&spu.ztile, x, y, spu.fb.depth_format, + result.depth, result.stencil); + } + + return result.mask; +} diff --git a/src/gallium/drivers/cell/spu/spu_per_fragment_op.h b/src/gallium/drivers/cell/spu/spu_per_fragment_op.h new file mode 100644 index 0000000000..6571258699 --- /dev/null +++ b/src/gallium/drivers/cell/spu/spu_per_fragment_op.h @@ -0,0 +1,32 @@ +/* + * (C) Copyright IBM Corporation 2008 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * AUTHORS, COPYRIGHT HOLDERS, AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef SPU_PER_FRAGMENT_OP +#define SPU_PER_FRAGMENT_OP + +extern qword +spu_do_depth_stencil(int x, int y, qword frag_mask, qword frag_depth, + qword frag_alpha, qword facing); + +#endif /* SPU_PER_FRAGMENT_OP */ diff --git a/src/gallium/drivers/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c index 20e77aa2e6..6df59abd36 100644 --- a/src/gallium/drivers/cell/spu/spu_render.c +++ b/src/gallium/drivers/cell/spu/spu_render.c @@ -98,7 +98,7 @@ my_tile(uint tx, uint ty) static INLINE void get_cz_tiles(uint tx, uint ty) { - if (spu.depth_stencil.depth.enabled) { + if (spu.read_depth) { if (spu.cur_ztile_status != TILE_STATUS_CLEAR) { //printf("SPU %u: getting Z tile %u, %u\n", spu.init.id, tx, ty); get_tile(tx, ty, &spu.ztile, TAG_READ_TILE_Z, 1); @@ -153,7 +153,7 @@ static INLINE void wait_put_cz_tiles(void) { wait_on_mask(1 << TAG_WRITE_TILE_COLOR); - if (spu.depth_stencil.depth.enabled) { + if (spu.read_depth) { wait_on_mask(1 << TAG_WRITE_TILE_Z); } } diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index be9624cf7d..81823f2463 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -38,8 +38,7 @@ #include "spu_texture.h" #include "spu_tile.h" #include "spu_tri.h" - -#include "spu_ztest.h" +#include "spu_per_fragment_op.h" /** Masks are uint[4] vectors with each element being 0 or 0xffffffff */ @@ -264,16 +263,12 @@ do_depth_test(int x, int y, mask_t quadmask) zvals.v = eval_z((float) x, (float) y); - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { - int ix = (x - setup.cliprect_minx) / 4; - int iy = (y - setup.cliprect_miny) / 2; - mask = spu_z16_test_less(zvals.v, &spu.ztile.us8[iy][ix], x>>1, quadmask); - } - else { - int ix = (x - setup.cliprect_minx) / 2; - int iy = (y - setup.cliprect_miny) / 2; - mask = spu_z32_test_less(zvals.v, &spu.ztile.ui4[iy][ix], quadmask); - } + mask = (mask_t) spu_do_depth_stencil(x - setup.cliprect_minx, + y - setup.cliprect_miny, + (qword) quadmask, + (qword) zvals.v, + (qword) spu_splats((unsigned char) 0x0ffu), + (qword) spu_splats((unsigned int) 0x01u)); if (spu_extract(spu_orx(mask), 0)) spu.cur_ztile_status = TILE_STATUS_DIRTY; @@ -299,7 +294,7 @@ emit_quad( int x, int y, mask_t mask ) sp->quad.first->run(sp->quad.first, &setup.quad); #else - if (spu.depth_stencil.depth.enabled) { + if (spu.read_depth) { mask = do_depth_test(x, y, mask); } @@ -434,7 +429,7 @@ static void flush_spans( void ) } ASSERT(spu.cur_ctile_status != TILE_STATUS_DEFINED); - if (spu.depth_stencil.depth.enabled) { + if (spu.read_depth) { if (spu.cur_ztile_status == TILE_STATUS_GETTING) { /* wait for mfc_get() to complete */ //printf("SPU: %u: waiting for ztile\n", spu.init.id); diff --git a/src/gallium/drivers/cell/spu/spu_ztest.h b/src/gallium/drivers/cell/spu/spu_ztest.h deleted file mode 100644 index ce8ad00339..0000000000 --- a/src/gallium/drivers/cell/spu/spu_ztest.h +++ /dev/null @@ -1,135 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -/** - * Zbuffer/depth test code. - */ - - -#ifndef SPU_ZTEST_H -#define SPU_ZTEST_H - - -#ifdef __SPU__ -#include -#endif - - - -/** - * Perform Z testing for a 16-bit/value Z buffer. - * - * \param zvals vector of four fragment zvalues as floats - * \param zbuf ptr to vector of ushort[8] zbuffer values. Note that this - * contains the Z values for 2 quads, 8 pixels. - * \param x x coordinate of quad (only lsbit is significant) - * \param inMask indicates which fragments in the quad are alive - * \return new mask indicating which fragments are alive after ztest - */ -static INLINE vector unsigned int -spu_z16_test_less(vector float zvals, vector unsigned short *zbuf, - uint x, vector unsigned int inMask) -{ -#define ZERO 0x80 - vector unsigned int zvals_ui4, zbuf_ui4, mask; - - /* convert floats to uints in [0, 65535] */ - zvals_ui4 = spu_convtu(zvals, 32); /* convert to [0, 2^32] */ - zvals_ui4 = spu_rlmask(zvals_ui4, -16); /* right shift 16 */ - - /* XXX this conditional could be removed with a bit of work */ - if (x & 1) { - /* convert zbuffer values from ushorts to uints */ - /* gather lower four ushorts */ - zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, - (vector unsigned int) *zbuf, - ((vector unsigned char) { - ZERO, ZERO, 8, 9, ZERO, ZERO, 10, 11, - ZERO, ZERO, 12, 13, ZERO, ZERO, 14, 15})); - /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ - mask = spu_cmpgt(zbuf_ui4, zvals_ui4); - /* mask &= inMask */ - mask = spu_and(mask, inMask); - /* zbuf = mask ? zval : zbuf */ - zbuf_ui4 = spu_sel(zbuf_ui4, zvals_ui4, mask); - /* convert zbuffer values from uints back to ushorts, preserve lower 4 */ - *zbuf = (vector unsigned short) - spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, - ((vector unsigned char) { - 16, 17, 18, 19, 20, 21, 22, 23, - 2, 3, 6, 7, 10, 11, 14, 15})); - } - else { - /* convert zbuffer values from ushorts to uints */ - /* gather upper four ushorts */ - zbuf_ui4 = spu_shuffle((vector unsigned int) *zbuf, - (vector unsigned int) *zbuf, - ((vector unsigned char) { - ZERO, ZERO, 0, 1, ZERO, ZERO, 2, 3, - ZERO, ZERO, 4, 5, ZERO, ZERO, 6, 7})); - /* mask = (zbuf_ui4 < zvals_ui4) ? ~0 : 0 */ - mask = spu_cmpgt(zbuf_ui4, zvals_ui4); - /* mask &= inMask */ - mask = spu_and(mask, inMask); - /* zbuf = mask ? zval : zbuf */ - zbuf_ui4 = spu_sel(zbuf_ui4, zvals_ui4, mask); - /* convert zbuffer values from uints back to ushorts, preserve upper 4 */ - *zbuf = (vector unsigned short) - spu_shuffle(zbuf_ui4, (vector unsigned int) *zbuf, - ((vector unsigned char) { - 2, 3, 6, 7, 10, 11, 14, 15, - 24, 25, 26, 27, 28, 29, 30, 31})); - } - return mask; -#undef ZERO -} - - -/** - * As above, but Zbuffer values as 32-bit uints - */ -static INLINE vector unsigned int -spu_z32_test_less(vector float zvals, vector unsigned int *zbuf_ptr, - vector unsigned int inMask) -{ - vector unsigned int zvals_ui4, mask, zbuf = *zbuf_ptr; - - /* convert floats to uints in [0, 0xffffffff] */ - zvals_ui4 = spu_convtu(zvals, 32); - /* mask = (zbuf < zvals_ui4) ? ~0 : 0 */ - mask = spu_cmpgt(zbuf, zvals_ui4); - /* mask &= inMask */ - mask = spu_and(mask, inMask); - /* zbuf = mask ? zval : zbuf */ - *zbuf_ptr = spu_sel(zbuf, zvals_ui4, mask); - - return mask; -} - - -#endif /* SPU_ZTEST_H */ -- cgit v1.2.3 From 9f106a8683ec89b873f0237fbb6930a63b89dfa0 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 17 Mar 2008 16:07:54 -0700 Subject: cell: Don't free NULL code pointers --- src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c index 842d713f84..24be65bff9 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c @@ -316,7 +316,9 @@ void spe_init_func(struct spe_function *p, unsigned code_size) void spe_release_func(struct spe_function *p) { - align_free(p->store); + if (p->store != NULL) { + align_free(p->store); + } p->store = NULL; p->csr = NULL; } -- cgit v1.2.3 From 9f93e6701902312edf48821bb4c0558c3d62aaa3 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 17 Mar 2008 16:09:28 -0700 Subject: cell: Don't segfault when unbinding alpha / stencil / depth test state --- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 2 +- src/gallium/drivers/cell/ppu/cell_state_emit.c | 17 ++++++++++++----- src/gallium/drivers/cell/spu/spu_main.c | 23 ++++++++++++++++------- 3 files changed, 29 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 66ede99d13..c880760e4b 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -116,7 +116,7 @@ cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, draw_flush(cell->draw); - if (cdsa->code.store == NULL) { + if ((cdsa != NULL) && (cdsa->code.store == NULL)) { cell_generate_depth_stencil_test(cdsa); } diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index c8d5fdf709..e2cc9de48a 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -74,11 +74,18 @@ cell_emit_state(struct cell_context *cell) struct cell_command_depth_stencil_alpha_test dsat; - dsat.base = (intptr_t) cell->depth_stencil->code.store; - dsat.size = (char *) cell->depth_stencil->code.csr - - (char *) cell->depth_stencil->code.store; - dsat.read_depth = TRUE; - dsat.read_stencil = FALSE; + if (cell->depth_stencil != NULL) { + dsat.base = (intptr_t) cell->depth_stencil->code.store; + dsat.size = (char *) cell->depth_stencil->code.csr + - (char *) cell->depth_stencil->code.store; + dsat.read_depth = TRUE; + dsat.read_stencil = FALSE; + } else { + dsat.base = 0; + dsat.size = 0; + dsat.read_depth = FALSE; + dsat.read_stencil = FALSE; + } { uint32_t *p = cell->depth_stencil->code.store; diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 8e46f6e471..b8bb8e9449 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -260,13 +260,22 @@ cmd_state_depth_stencil(const struct cell_command_depth_stencil_alpha_test *stat ASSERT_ALIGN16(state->base); - mfc_get(depth_stencil_code_buffer, - (unsigned int) state->base, /* src */ - ROUNDUP16(state->size), - TAG_BATCH_BUFFER, - 0, /* tid */ - 0 /* rid */); - wait_on_mask(1 << TAG_BATCH_BUFFER); + if (state->size != 0) { + mfc_get(depth_stencil_code_buffer, + (unsigned int) state->base, /* src */ + ROUNDUP16(state->size), + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); + } else { + /* If there is no code, emit a return instruction. + */ + depth_stencil_code_buffer[0] = 0x35; + depth_stencil_code_buffer[1] = 0x00; + depth_stencil_code_buffer[2] = 0x00; + depth_stencil_code_buffer[3] = 0x00; + } spu.frag_test = (frag_test_func) depth_stencil_code_buffer; spu.read_depth = state->read_depth; -- cgit v1.2.3 From de779c8d319b6269705cd7e6f2009243a771a2b9 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 17 Mar 2008 17:45:00 -0600 Subject: gallium: initial gen mipmap s/w fallback code --- src/gallium/auxiliary/util/u_gen_mipmap.c | 609 +++++++++++++++++++++++++++++- 1 file changed, 607 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index f0b822929e..8ca06281ab 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -34,6 +34,7 @@ #include "pipe/p_context.h" +#include "pipe/p_debug.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_util.h" @@ -48,6 +49,612 @@ #include "tgsi/util/tgsi_parse.h" + +enum dtype +{ + UBYTE, + UBYTE_3_3_2, + USHORT, + USHORT_4_4_4_4, + USHORT_5_6_5, + USHORT_1_5_5_5_REV, + UINT, + FLOAT, + HALF_FLOAT +}; + + +typedef ushort half_float; + + +#if 0 +extern half_float +float_to_half(float f); + +extern float +half_to_float(half_float h); +#endif + + +/** + * Average together two rows of a source image to produce a single new + * row in the dest image. It's legal for the two source rows to point + * to the same data. The source width must be equal to either the + * dest width or two times the dest width. + * \param datatype GL_UNSIGNED_BYTE, GL_UNSIGNED_SHORT, GL_FLOAT, etc. + * \param comps number of components per pixel (1..4) + */ +static void +do_row(enum dtype datatype, uint comps, int srcWidth, + const void *srcRowA, const void *srcRowB, + int dstWidth, void *dstRow) +{ + const uint k0 = (srcWidth == dstWidth) ? 0 : 1; + const uint colStride = (srcWidth == dstWidth) ? 1 : 2; + + assert(comps >= 1); + assert(comps <= 4); + + /* This assertion is no longer valid with non-power-of-2 textures + assert(srcWidth == dstWidth || srcWidth == 2 * dstWidth); + */ + + if (datatype == UBYTE && comps == 4) { + uint i, j, k; + const ubyte(*rowA)[4] = (const ubyte(*)[4]) srcRowA; + const ubyte(*rowB)[4] = (const ubyte(*)[4]) srcRowB; + ubyte(*dst)[4] = (ubyte(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; + } + } + else if (datatype == UBYTE && comps == 3) { + uint i, j, k; + const ubyte(*rowA)[3] = (const ubyte(*)[3]) srcRowA; + const ubyte(*rowB)[3] = (const ubyte(*)[3]) srcRowB; + ubyte(*dst)[3] = (ubyte(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + } + } + else if (datatype == UBYTE && comps == 2) { + uint i, j, k; + const ubyte(*rowA)[2] = (const ubyte(*)[2]) srcRowA; + const ubyte(*rowB)[2] = (const ubyte(*)[2]) srcRowB; + ubyte(*dst)[2] = (ubyte(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) >> 2; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) >> 2; + } + } + else if (datatype == UBYTE && comps == 1) { + uint i, j, k; + const ubyte *rowA = (const ubyte *) srcRowA; + const ubyte *rowB = (const ubyte *) srcRowB; + ubyte *dst = (ubyte *) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) >> 2; + } + } + + else if (datatype == USHORT && comps == 4) { + uint i, j, k; + const ushort(*rowA)[4] = (const ushort(*)[4]) srcRowA; + const ushort(*rowB)[4] = (const ushort(*)[4]) srcRowB; + ushort(*dst)[4] = (ushort(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + dst[i][3] = (rowA[j][3] + rowA[k][3] + rowB[j][3] + rowB[k][3]) / 4; + } + } + else if (datatype == USHORT && comps == 3) { + uint i, j, k; + const ushort(*rowA)[3] = (const ushort(*)[3]) srcRowA; + const ushort(*rowB)[3] = (const ushort(*)[3]) srcRowB; + ushort(*dst)[3] = (ushort(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + dst[i][2] = (rowA[j][2] + rowA[k][2] + rowB[j][2] + rowB[k][2]) / 4; + } + } + else if (datatype == USHORT && comps == 2) { + uint i, j, k; + const ushort(*rowA)[2] = (const ushort(*)[2]) srcRowA; + const ushort(*rowB)[2] = (const ushort(*)[2]) srcRowB; + ushort(*dst)[2] = (ushort(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + rowB[j][0] + rowB[k][0]) / 4; + dst[i][1] = (rowA[j][1] + rowA[k][1] + rowB[j][1] + rowB[k][1]) / 4; + } + } + else if (datatype == USHORT && comps == 1) { + uint i, j, k; + const ushort *rowA = (const ushort *) srcRowA; + const ushort *rowB = (const ushort *) srcRowB; + ushort *dst = (ushort *) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) / 4; + } + } + + else if (datatype == FLOAT && comps == 4) { + uint i, j, k; + const float(*rowA)[4] = (const float(*)[4]) srcRowA; + const float(*rowB)[4] = (const float(*)[4]) srcRowB; + float(*dst)[4] = (float(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + + rowB[j][0] + rowB[k][0]) * 0.25F; + dst[i][1] = (rowA[j][1] + rowA[k][1] + + rowB[j][1] + rowB[k][1]) * 0.25F; + dst[i][2] = (rowA[j][2] + rowA[k][2] + + rowB[j][2] + rowB[k][2]) * 0.25F; + dst[i][3] = (rowA[j][3] + rowA[k][3] + + rowB[j][3] + rowB[k][3]) * 0.25F; + } + } + else if (datatype == FLOAT && comps == 3) { + uint i, j, k; + const float(*rowA)[3] = (const float(*)[3]) srcRowA; + const float(*rowB)[3] = (const float(*)[3]) srcRowB; + float(*dst)[3] = (float(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + + rowB[j][0] + rowB[k][0]) * 0.25F; + dst[i][1] = (rowA[j][1] + rowA[k][1] + + rowB[j][1] + rowB[k][1]) * 0.25F; + dst[i][2] = (rowA[j][2] + rowA[k][2] + + rowB[j][2] + rowB[k][2]) * 0.25F; + } + } + else if (datatype == FLOAT && comps == 2) { + uint i, j, k; + const float(*rowA)[2] = (const float(*)[2]) srcRowA; + const float(*rowB)[2] = (const float(*)[2]) srcRowB; + float(*dst)[2] = (float(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i][0] = (rowA[j][0] + rowA[k][0] + + rowB[j][0] + rowB[k][0]) * 0.25F; + dst[i][1] = (rowA[j][1] + rowA[k][1] + + rowB[j][1] + rowB[k][1]) * 0.25F; + } + } + else if (datatype == FLOAT && comps == 1) { + uint i, j, k; + const float *rowA = (const float *) srcRowA; + const float *rowB = (const float *) srcRowB; + float *dst = (float *) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i] = (rowA[j] + rowA[k] + rowB[j] + rowB[k]) * 0.25F; + } + } + +#if 0 + else if (datatype == HALF_FLOAT && comps == 4) { + uint i, j, k, comp; + const half_float(*rowA)[4] = (const half_float(*)[4]) srcRowA; + const half_float(*rowB)[4] = (const half_float(*)[4]) srcRowB; + half_float(*dst)[4] = (half_float(*)[4]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + for (comp = 0; comp < 4; comp++) { + float aj, ak, bj, bk; + aj = half_to_float(rowA[j][comp]); + ak = half_to_float(rowA[k][comp]); + bj = half_to_float(rowB[j][comp]); + bk = half_to_float(rowB[k][comp]); + dst[i][comp] = float_to_half((aj + ak + bj + bk) * 0.25F); + } + } + } + else if (datatype == HALF_FLOAT && comps == 3) { + uint i, j, k, comp; + const half_float(*rowA)[3] = (const half_float(*)[3]) srcRowA; + const half_float(*rowB)[3] = (const half_float(*)[3]) srcRowB; + half_float(*dst)[3] = (half_float(*)[3]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + for (comp = 0; comp < 3; comp++) { + float aj, ak, bj, bk; + aj = half_to_float(rowA[j][comp]); + ak = half_to_float(rowA[k][comp]); + bj = half_to_float(rowB[j][comp]); + bk = half_to_float(rowB[k][comp]); + dst[i][comp] = float_to_half((aj + ak + bj + bk) * 0.25F); + } + } + } + else if (datatype == HALF_FLOAT && comps == 2) { + uint i, j, k, comp; + const half_float(*rowA)[2] = (const half_float(*)[2]) srcRowA; + const half_float(*rowB)[2] = (const half_float(*)[2]) srcRowB; + half_float(*dst)[2] = (half_float(*)[2]) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + for (comp = 0; comp < 2; comp++) { + float aj, ak, bj, bk; + aj = half_to_float(rowA[j][comp]); + ak = half_to_float(rowA[k][comp]); + bj = half_to_float(rowB[j][comp]); + bk = half_to_float(rowB[k][comp]); + dst[i][comp] = float_to_half((aj + ak + bj + bk) * 0.25F); + } + } + } + else if (datatype == HALF_FLOAT && comps == 1) { + uint i, j, k; + const half_float *rowA = (const half_float *) srcRowA; + const half_float *rowB = (const half_float *) srcRowB; + half_float *dst = (half_float *) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + float aj, ak, bj, bk; + aj = half_to_float(rowA[j]); + ak = half_to_float(rowA[k]); + bj = half_to_float(rowB[j]); + bk = half_to_float(rowB[k]); + dst[i] = float_to_half((aj + ak + bj + bk) * 0.25F); + } + } +#endif + + else if (datatype == UINT && comps == 1) { + uint i, j, k; + const uint *rowA = (const uint *) srcRowA; + const uint *rowB = (const uint *) srcRowB; + float *dst = (float *) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; + } + } + + else if (datatype == USHORT_5_6_5 && comps == 3) { + uint i, j, k; + const ushort *rowA = (const ushort *) srcRowA; + const ushort *rowB = (const ushort *) srcRowB; + ushort *dst = (ushort *) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + const int rowAr0 = rowA[j] & 0x1f; + const int rowAr1 = rowA[k] & 0x1f; + const int rowBr0 = rowB[j] & 0x1f; + const int rowBr1 = rowB[k] & 0x1f; + const int rowAg0 = (rowA[j] >> 5) & 0x3f; + const int rowAg1 = (rowA[k] >> 5) & 0x3f; + const int rowBg0 = (rowB[j] >> 5) & 0x3f; + const int rowBg1 = (rowB[k] >> 5) & 0x3f; + const int rowAb0 = (rowA[j] >> 11) & 0x1f; + const int rowAb1 = (rowA[k] >> 11) & 0x1f; + const int rowBb0 = (rowB[j] >> 11) & 0x1f; + const int rowBb1 = (rowB[k] >> 11) & 0x1f; + const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; + const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; + const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; + dst[i] = (blue << 11) | (green << 5) | red; + } + } + else if (datatype == USHORT_4_4_4_4 && comps == 4) { + uint i, j, k; + const ushort *rowA = (const ushort *) srcRowA; + const ushort *rowB = (const ushort *) srcRowB; + ushort *dst = (ushort *) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + const int rowAr0 = rowA[j] & 0xf; + const int rowAr1 = rowA[k] & 0xf; + const int rowBr0 = rowB[j] & 0xf; + const int rowBr1 = rowB[k] & 0xf; + const int rowAg0 = (rowA[j] >> 4) & 0xf; + const int rowAg1 = (rowA[k] >> 4) & 0xf; + const int rowBg0 = (rowB[j] >> 4) & 0xf; + const int rowBg1 = (rowB[k] >> 4) & 0xf; + const int rowAb0 = (rowA[j] >> 8) & 0xf; + const int rowAb1 = (rowA[k] >> 8) & 0xf; + const int rowBb0 = (rowB[j] >> 8) & 0xf; + const int rowBb1 = (rowB[k] >> 8) & 0xf; + const int rowAa0 = (rowA[j] >> 12) & 0xf; + const int rowAa1 = (rowA[k] >> 12) & 0xf; + const int rowBa0 = (rowB[j] >> 12) & 0xf; + const int rowBa1 = (rowB[k] >> 12) & 0xf; + const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; + const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; + const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; + const int alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2; + dst[i] = (alpha << 12) | (blue << 8) | (green << 4) | red; + } + } + else if (datatype == USHORT_1_5_5_5_REV && comps == 4) { + uint i, j, k; + const ushort *rowA = (const ushort *) srcRowA; + const ushort *rowB = (const ushort *) srcRowB; + ushort *dst = (ushort *) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + const int rowAr0 = rowA[j] & 0x1f; + const int rowAr1 = rowA[k] & 0x1f; + const int rowBr0 = rowB[j] & 0x1f; + const int rowBr1 = rowB[k] & 0xf; + const int rowAg0 = (rowA[j] >> 5) & 0x1f; + const int rowAg1 = (rowA[k] >> 5) & 0x1f; + const int rowBg0 = (rowB[j] >> 5) & 0x1f; + const int rowBg1 = (rowB[k] >> 5) & 0x1f; + const int rowAb0 = (rowA[j] >> 10) & 0x1f; + const int rowAb1 = (rowA[k] >> 10) & 0x1f; + const int rowBb0 = (rowB[j] >> 10) & 0x1f; + const int rowBb1 = (rowB[k] >> 10) & 0x1f; + const int rowAa0 = (rowA[j] >> 15) & 0x1; + const int rowAa1 = (rowA[k] >> 15) & 0x1; + const int rowBa0 = (rowB[j] >> 15) & 0x1; + const int rowBa1 = (rowB[k] >> 15) & 0x1; + const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; + const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; + const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; + const int alpha = (rowAa0 + rowAa1 + rowBa0 + rowBa1) >> 2; + dst[i] = (alpha << 15) | (blue << 10) | (green << 5) | red; + } + } + else if (datatype == UBYTE_3_3_2 && comps == 3) { + uint i, j, k; + const ubyte *rowA = (const ubyte *) srcRowA; + const ubyte *rowB = (const ubyte *) srcRowB; + ubyte *dst = (ubyte *) dstRow; + for (i = j = 0, k = k0; i < (uint) dstWidth; + i++, j += colStride, k += colStride) { + const int rowAr0 = rowA[j] & 0x3; + const int rowAr1 = rowA[k] & 0x3; + const int rowBr0 = rowB[j] & 0x3; + const int rowBr1 = rowB[k] & 0x3; + const int rowAg0 = (rowA[j] >> 2) & 0x7; + const int rowAg1 = (rowA[k] >> 2) & 0x7; + const int rowBg0 = (rowB[j] >> 2) & 0x7; + const int rowBg1 = (rowB[k] >> 2) & 0x7; + const int rowAb0 = (rowA[j] >> 5) & 0x7; + const int rowAb1 = (rowA[k] >> 5) & 0x7; + const int rowBb0 = (rowB[j] >> 5) & 0x7; + const int rowBb1 = (rowB[k] >> 5) & 0x7; + const int red = (rowAr0 + rowAr1 + rowBr0 + rowBr1) >> 2; + const int green = (rowAg0 + rowAg1 + rowBg0 + rowBg1) >> 2; + const int blue = (rowAb0 + rowAb1 + rowBb0 + rowBb1) >> 2; + dst[i] = (blue << 5) | (green << 2) | red; + } + } + else { + debug_printf("bad format in do_row()"); + } +} + + +static void +format_to_type_comps(enum pipe_format pformat, + enum dtype *datatype, uint *comps) +{ + switch (pformat) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_B8G8R8A8_UNORM: + *datatype = UBYTE; + *comps = 4; + return; + case PIPE_FORMAT_A1R5G5B5_UNORM: + *datatype = USHORT_1_5_5_5_REV; + *comps = 4; + return; + case PIPE_FORMAT_A4R4G4B4_UNORM: + *datatype = USHORT_4_4_4_4; + *comps = 4; + return; + case PIPE_FORMAT_R5G6B5_UNORM: + *datatype = USHORT_5_6_5; + *comps = 3; + return; + case PIPE_FORMAT_U_L8: + case PIPE_FORMAT_U_A8: + case PIPE_FORMAT_U_I8: + *datatype = UBYTE; + *comps = 1; + return; + case PIPE_FORMAT_U_A8_L8: + *datatype = UBYTE; + *comps = 2; + return; + default: + assert(0); + } +} + + +static void +reduce_1d(enum pipe_format pformat, + int srcWidth, const ubyte *srcPtr, + int dstWidth, ubyte *dstPtr) +{ + enum dtype datatype; + uint comps; + + format_to_type_comps(pformat, &datatype, &comps); + + /* we just duplicate the input row, kind of hack, saves code */ + do_row(datatype, comps, + srcWidth, srcPtr, srcPtr, + dstWidth, dstPtr); +} + + +/** + * Strides are in bytes. If zero, it'll be computed as width * bpp. + */ +static void +reduce_2d(enum pipe_format pformat, + int srcWidth, int srcHeight, + int srcRowStride, const ubyte *srcPtr, + int dstWidth, int dstHeight, + int dstRowStride, ubyte *dstPtr) +{ + enum dtype datatype; + uint comps; + const int bpt = pf_get_size(pformat); + const ubyte *srcA, *srcB; + ubyte *dst; + int row; + + format_to_type_comps(pformat, &datatype, &comps); + + if (!srcRowStride) + srcRowStride = bpt * srcWidth; + + if (!dstRowStride) + dstRowStride = bpt * dstWidth; + + /* Compute src and dst pointers */ + srcA = srcPtr; + if (srcHeight > 1) + srcB = srcA + srcRowStride; + else + srcB = srcA; + dst = dstPtr; + + for (row = 0; row < dstHeight; row++) { + do_row(datatype, comps, + srcWidth, srcA, srcB, + dstWidth, dst); + srcA += 2 * srcRowStride; + srcB += 2 * srcRowStride; + dst += dstRowStride; + } +} + + +static void +make_1d_mipmap(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_winsys *winsys = pipe->winsys; + const uint zslice = 0; + uint dstLevel; + + for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { + const uint srcLevel = dstLevel - 1; + struct pipe_surface *srcSurf, *dstSurf; + void *srcMap, *dstMap; + + srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice); + dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); + + srcMap = ((ubyte *) winsys->buffer_map(winsys, srcSurf->buffer, + PIPE_BUFFER_USAGE_CPU_READ) + + srcSurf->offset); + dstMap = ((ubyte *) winsys->buffer_map(winsys, dstSurf->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE) + + dstSurf->offset); + + reduce_1d(pt->format, + srcSurf->width, srcMap, + dstSurf->width, dstMap); + + winsys->buffer_unmap(winsys, srcSurf->buffer); + winsys->buffer_unmap(winsys, dstSurf->buffer); + + pipe_surface_reference(&srcSurf, NULL); + pipe_surface_reference(&dstSurf, NULL); + } +} + + +static void +make_2d_mipmap(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_winsys *winsys = pipe->winsys; + const uint zslice = 0; + uint dstLevel; + const int bpt = pf_get_size(pt->format); + + for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { + const uint srcLevel = dstLevel - 1; + struct pipe_surface *srcSurf, *dstSurf; + ubyte *srcMap, *dstMap; + + srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice); + dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); + + srcMap = ((ubyte *) winsys->buffer_map(winsys, srcSurf->buffer, + PIPE_BUFFER_USAGE_CPU_READ) + + srcSurf->offset); + dstMap = ((ubyte *) winsys->buffer_map(winsys, dstSurf->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE) + + dstSurf->offset); + + reduce_2d(pt->format, + srcSurf->width, srcSurf->height, + srcSurf->pitch * bpt, srcMap, + dstSurf->width, dstSurf->height, + dstSurf->pitch * bpt, dstMap); + + winsys->buffer_unmap(winsys, srcSurf->buffer); + winsys->buffer_unmap(winsys, dstSurf->buffer); + + pipe_surface_reference(&srcSurf, NULL); + pipe_surface_reference(&dstSurf, NULL); + } +} + + +static void +make_3d_mipmap(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel) +{ +} + + +static void +fallback_gen_mipmap(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel) +{ + switch (pt->target) { + case PIPE_TEXTURE_1D: + make_1d_mipmap(ctx, pt, face, baseLevel, lastLevel); + break; + case PIPE_TEXTURE_2D: + case PIPE_TEXTURE_CUBE: + make_2d_mipmap(ctx, pt, face, baseLevel, lastLevel); + break; + case PIPE_TEXTURE_3D: + make_3d_mipmap(ctx, pt, face, baseLevel, lastLevel); + break; + default: + assert(0); + } +} + + /** * Make simple fragment shader: * TEX OUT[0], IN[0], SAMP[0], 2D; @@ -415,9 +1022,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, /* check if we can render in the texture's format */ if (!screen->is_format_supported(screen, pt->format, PIPE_SURFACE)) { -#if 0 fallback_gen_mipmap(ctx, pt, face, baseLevel, lastLevel); -#endif return; } -- cgit v1.2.3 From b2b905b04c09dd5e701a43b0fecb73921b8f2866 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 18 Mar 2008 15:59:55 +0100 Subject: gallium: Silencium warnings on Windows. --- src/gallium/auxiliary/util/u_gen_mipmap.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 8ca06281ab..d6d8c766ce 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -325,7 +325,7 @@ do_row(enum dtype datatype, uint comps, int srcWidth, float *dst = (float *) dstRow; for (i = j = 0, k = k0; i < (uint) dstWidth; i++, j += colStride, k += colStride) { - dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; + dst[i] = (float) rowA[j] / 4 + (float) rowA[k] / 4 + (float) rowB[j] / 4 + (float) rowB[k] / 4; } } @@ -1070,8 +1070,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, * has trouble with min clamping so we also set the lod_bias to * try to work around that. */ - sampler.min_lod = sampler.max_lod = srcLevel; - sampler.lod_bias = srcLevel; + sampler.min_lod = sampler.max_lod = (float) srcLevel; + sampler.lod_bias = (float) srcLevel; sampler_cso = pipe->create_sampler_state(pipe, &sampler); pipe->bind_sampler_states(pipe, 1, &sampler_cso); -- cgit v1.2.3 From 544c27bcefd8a10318800c3cc2019514ee01a15b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 09:22:41 -0600 Subject: gallium: utility for packing color, z values according to pipe_format --- src/gallium/auxiliary/util/u_pack_color.h | 125 ++++++++++++++++++++++++++++++ 1 file changed, 125 insertions(+) create mode 100644 src/gallium/auxiliary/util/u_pack_color.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h new file mode 100644 index 0000000000..93a18c1c7e --- /dev/null +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -0,0 +1,125 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Functions to produce packed colors/Z from floats. + */ + + +#include "pipe/p_compiler.h" + + +/** + * Note rgba outside [0,1] will be clamped for int pixel formats. + */ +static INLINE void +util_pack_color(const float rgba[4], enum pipe_format format, void *dest) +{ + ubyte r, g, b, a; + + if (pf_size_x(format) <= 8) { + /* format uses 8-bit components or less */ + UNCLAMPED_FLOAT_TO_UBYTE(r, rgba[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, rgba[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, rgba[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, rgba[3]); + } + + switch (format) { + case PIPE_FORMAT_R8G8B8A8_UNORM: + { + uint *d = (uint *) dest; + *d = (r << 24) | (g << 16) | (b << 8) | a; + } + return; + case PIPE_FORMAT_A8R8G8B8_UNORM: + { + uint *d = (uint *) dest; + *d = (a << 24) | (r << 16) | (g << 8) | b; + } + return; + case PIPE_FORMAT_B8G8R8A8_UNORM: + { + uint *d = (uint *) dest; + *d = (b << 24) | (g << 16) | (r << 8) | a; + } + return; + case PIPE_FORMAT_R5G6B5_UNORM: + { + ushort *d = (ushort *) dest; + *d = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); + } + return; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + { + float *d = (float *) dest; + d[0] = rgba[0]; + d[1] = rgba[1]; + d[2] = rgba[2]; + d[3] = rgba[3]; + } + return; + case PIPE_FORMAT_R32G32B32_FLOAT: + { + float *d = (float *) dest; + d[0] = rgba[0]; + d[1] = rgba[1]; + d[2] = rgba[2]; + } + return; + /* XXX lots more cases to add */ + default: + debug_printf("gallium: unhandled format in util_pack_color()"); + } +} + + +/** + * Note: it's assumed that z is in [0,1] + */ +static INLINE uint +util_pack_z(enum pipe_format format, double z) +{ + switch (format) { + case PIPE_FORMAT_Z16_UNORM: + return (uint) (z * 0xffff); + case PIPE_FORMAT_Z32_UNORM: + /* special-case to avoid overflow */ + if (z == 1.0) + return 0xffffffff; + else + return (uint) (z * 0xffffffff); + case PIPE_FORMAT_S8Z24_UNORM: + return (uint) (z * 0xffffff); + case PIPE_FORMAT_Z24S8_UNORM: + return ((uint) (z * 0xffffff)) << 8; + default: + debug_printf("gallium: unhandled fomrat in util_pack_z()"); + return 0; + } +} -- cgit v1.2.3 From f1cfb1e3676fc9e1fea6698ed5e8e79d2b094dae Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 09:23:35 -0600 Subject: gallium: use new color packing utility functions --- src/mesa/state_tracker/st_cb_clear.c | 61 ++++-------------------------------- 1 file changed, 6 insertions(+), 55 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 5865071439..8223784d2e 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -48,6 +48,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" +#include "util/u_pack_color.h" #include "cso_cache/cso_context.h" @@ -56,55 +57,6 @@ #define TEST_DRAW_PASSTHROUGH 0 -static GLuint -color_value(enum pipe_format pipeFormat, const GLfloat color[4]) -{ - GLubyte r, g, b, a; - - UNCLAMPED_FLOAT_TO_UBYTE(r, color[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, color[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, color[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, color[3]); - - switch (pipeFormat) { - case PIPE_FORMAT_R8G8B8A8_UNORM: - return (r << 24) | (g << 16) | (b << 8) | a; - case PIPE_FORMAT_A8R8G8B8_UNORM: - return (a << 24) | (r << 16) | (g << 8) | b; - case PIPE_FORMAT_B8G8R8A8_UNORM: - return (b << 24) | (g << 16) | (r << 8) | a; - case PIPE_FORMAT_R5G6B5_UNORM: - return ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); - default: - assert(0); - return 0; - } -} - - -static uint -depth_value(enum pipe_format pipeFormat, GLfloat value) -{ - switch (pipeFormat) { - case PIPE_FORMAT_Z16_UNORM: - return (uint) (value * 0xffff); - case PIPE_FORMAT_Z32_UNORM: - /* special-case to avoid overflow */ - if (value == 1.0) - return 0xffffffff; - else - return (uint) (value * 0xffffffff); - case PIPE_FORMAT_S8Z24_UNORM: - return (uint) (value * 0xffffff); - case PIPE_FORMAT_Z24S8_UNORM: - return ((uint) (value * 0xffffff)) << 8; - default: - assert(0); - return 0; - } -} - - static GLboolean is_depth_stencil_format(enum pipe_format pipeFormat) { @@ -518,7 +470,6 @@ check_clear_stencil_with_quad(GLcontext *ctx, struct gl_renderbuffer *rb) - static void clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { @@ -527,10 +478,10 @@ clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) clear_with_quad(ctx, GL_TRUE, GL_FALSE, GL_FALSE); } else { - struct st_renderbuffer *strb = st_renderbuffer(rb); - /* clear whole buffer w/out masking */ - uint clearValue = color_value(strb->surface->format, ctx->Color.ClearColor); + struct st_renderbuffer *strb = st_renderbuffer(rb); + uint clearValue; + util_pack_color(ctx->Color.ClearColor, strb->surface->format, &clearValue); ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } } @@ -547,7 +498,7 @@ clear_depth_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) struct st_renderbuffer *strb = st_renderbuffer(rb); /* simple clear of whole buffer */ - uint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); + uint clearValue = util_pack_z(strb->surface->format, ctx->Depth.Clear); ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } } @@ -591,7 +542,7 @@ clear_depth_stencil_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) struct st_renderbuffer *strb = st_renderbuffer(rb); /* clear whole buffer w/out masking */ - GLuint clearValue = depth_value(strb->surface->format, ctx->Depth.Clear); + GLuint clearValue = util_pack_z(strb->surface->format, ctx->Depth.Clear); switch (strb->surface->format) { case PIPE_FORMAT_S8Z24_UNORM: -- cgit v1.2.3 From e8f8b12bb940794cef8eff52ae8c908ad0604161 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 09:28:04 -0600 Subject: gallium: fix a mix-up in the uint[1] do_row() case --- src/gallium/auxiliary/util/u_gen_mipmap.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index d6d8c766ce..d47c677074 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -322,10 +322,10 @@ do_row(enum dtype datatype, uint comps, int srcWidth, uint i, j, k; const uint *rowA = (const uint *) srcRowA; const uint *rowB = (const uint *) srcRowB; - float *dst = (float *) dstRow; + uint *dst = (uint *) dstRow; for (i = j = 0, k = k0; i < (uint) dstWidth; i++, j += colStride, k += colStride) { - dst[i] = (float) rowA[j] / 4 + (float) rowA[k] / 4 + (float) rowB[j] / 4 + (float) rowB[k] / 4; + dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; } } -- cgit v1.2.3 From d1ca951cc4f3392aeec2817e97fb9ade2c1b7881 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 18 Mar 2008 11:49:02 +0000 Subject: gallium: Convenience debug_warning function. --- src/gallium/include/pipe/p_debug.h | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index f971ad3adc..f3dfa06216 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -158,6 +158,16 @@ void debug_mask_vprintf(uint32_t uuid, const char *format, va_list ap); + +#ifdef DEBUG +#define debug_warning(__msg) \ + debug_printf("%s:%i:warning: %s\n", __FILE__, __LINE__, (__msg)) +#else +#define debug_warning(__msg) \ + ((void)0) +#endif + + #ifdef __cplusplus } #endif -- cgit v1.2.3 From 56ac9eb1f6e3826e4e8f7ab0f1fdbeed06c41c9f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 18 Mar 2008 11:49:29 +0000 Subject: gallium: Don't be pedantic about removing non exiting items from the table. --- src/gallium/auxiliary/util/u_handle_table.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index d25872972a..fae4ac4039 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -226,9 +226,13 @@ handle_table_remove(struct handle_table *ht, index = handle - 1; object = ht->objects[index]; - assert(object); + if(!object) { + /* XXX: this warning may be noisy for legitimate use -- remove later */ + debug_warning("removing empty handle"); + return; + } - if(object && ht->destroy) + if(ht->destroy) ht->destroy(object); ht->objects[index] = NULL; -- cgit v1.2.3 From d26139d6a19aaf8b4dbbaa1ee937fed2283923e4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 18 Mar 2008 12:01:14 +0000 Subject: d3d: Add function to walk through all items in the hash table. --- src/gallium/auxiliary/util/u_hash_table.c | 23 ++++++++++++++++++++++- src/gallium/auxiliary/util/u_hash_table.h | 5 +++++ 2 files changed, 27 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c index ac2cb1b540..f3f16a8d94 100644 --- a/src/gallium/auxiliary/util/u_hash_table.c +++ b/src/gallium/auxiliary/util/u_hash_table.c @@ -187,6 +187,28 @@ hash_table_remove(struct hash_table *ht, } +enum pipe_error +hash_table_foreach(struct hash_table *ht, + enum pipe_error (*callback)(void *key, void *value, void *data), + void *data) +{ + struct cso_hash_iter iter; + struct hash_table_item *item; + enum pipe_error result; + + iter = cso_hash_first_node(ht->cso); + while (!cso_hash_iter_is_null(iter)) { + item = (struct hash_table_item *)cso_hash_iter_data(iter); + result = callback(item->key, item->value, data); + if(result != PIPE_OK) + return result; + iter = cso_hash_iter_next(iter); + } + + return PIPE_OK; +} + + void hash_table_destroy(struct hash_table *ht) { @@ -196,4 +218,3 @@ hash_table_destroy(struct hash_table *ht) FREE(ht); } - diff --git a/src/gallium/auxiliary/util/u_hash_table.h b/src/gallium/auxiliary/util/u_hash_table.h index d941f2c6b1..1583bd7548 100644 --- a/src/gallium/auxiliary/util/u_hash_table.h +++ b/src/gallium/auxiliary/util/u_hash_table.h @@ -75,6 +75,11 @@ hash_table_remove(struct hash_table *ht, void *key); +enum pipe_error +hash_table_foreach(struct hash_table *ht, + enum pipe_error (*callback)(void *key, void *value, void *data), + void *data); + void hash_table_destroy(struct hash_table *ht); -- cgit v1.2.3 From 527e30c53baadb396e5503e5188f0a9f1b2d2501 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 18 Mar 2008 12:46:24 +0000 Subject: d3d: Allow to iterate over the handle table. --- src/gallium/auxiliary/util/u_handle_table.c | 22 ++++++++++++++++++++++ src/gallium/auxiliary/util/u_handle_table.h | 9 +++++++++ 2 files changed, 31 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index fae4ac4039..0bfb9e1b4a 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -241,6 +241,28 @@ handle_table_remove(struct handle_table *ht, } +unsigned +handle_table_get_next_handle(struct handle_table *ht, + unsigned handle) +{ + unsigned index; + + for(index = handle; index < ht->size; ++index) { + if(!ht->objects[index]) + return index + 1; + } + + return 0; +} + + +unsigned +handle_table_get_first_handle(struct handle_table *ht) +{ + return handle_table_get_next_handle(ht, 0); +} + + void handle_table_destroy(struct handle_table *ht) { diff --git a/src/gallium/auxiliary/util/u_handle_table.h b/src/gallium/auxiliary/util/u_handle_table.h index a2f1f604ad..d080135c9f 100644 --- a/src/gallium/auxiliary/util/u_handle_table.h +++ b/src/gallium/auxiliary/util/u_handle_table.h @@ -100,6 +100,15 @@ void handle_table_destroy(struct handle_table *ht); +unsigned +handle_table_get_first_handle(struct handle_table *ht); + + +unsigned +handle_table_get_next_handle(struct handle_table *ht, + unsigned handle); + + #ifdef __cplusplus } #endif -- cgit v1.2.3 From f432ac5a1126997f10072760fe5afbf1b96f44f7 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 18 Mar 2008 10:23:39 -0700 Subject: cell: Move and (conditionally) silence debug code --- src/gallium/drivers/cell/ppu/cell_state_emit.c | 10 ------ .../drivers/cell/ppu/cell_state_per_fragment.c | 39 ++++++++++++++++++++++ 2 files changed, 39 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index e2cc9de48a..4d589bcdbf 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -87,16 +87,6 @@ cell_emit_state(struct cell_context *cell) dsat.read_stencil = FALSE; } - { - uint32_t *p = cell->depth_stencil->code.store; - - printf("\t.text\n"); - for (/* empty */; p < cell->depth_stencil->code.csr; p++) { - printf("\t.long\t0x%04x\n", *p); - } - fflush(stdout); - } - emit_state_cmd(cell, CELL_CMD_STATE_DEPTH_STENCIL, &dsat, sizeof(dsat)); } diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c index 60b64438d8..0ad9344cd4 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -525,6 +525,45 @@ cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa) } spe_bi(f, 0, 0, 0); + + +#if 0 + { + const uint32_t *p = f->store; + unsigned i; + + printf("# alpha (%sabled)\n", + (dsa->alpha.enabled) ? "en" : "dis"); + printf("# func: %u\n", dsa->alpha.func); + printf("# ref: %.2f\n", dsa->alpha.ref); + + printf("# depth (%sabled)\n", + (dsa->depth.enabled) ? "en" : "dis"); + printf("# func: %u\n", dsa->depth.func); + + for (i = 0; i < 2; i++) { + printf("# %s stencil (%sabled)\n", + (i == 0) ? "front" : "back", + (dsa->stencil[i].enabled) ? "en" : "dis"); + + printf("# func: %u\n", dsa->stencil[i].func); + printf("# op (sf, zf, zp): %u %u %u\n", + dsa->stencil[i].fail_op, + dsa->stencil[i].zfail_op, + dsa->stencil[i].zpass_op); + printf("# ref value / value mask / write mask: %02x %02x %02x\n", + dsa->stencil[i].ref_value, + dsa->stencil[i].value_mask, + dsa->stencil[i].write_mask); + } + + printf("\t.text\n"); + for (/* empty */; p < f->csr; p++) { + printf("\t.long\t0x%04x\n", *p); + } + fflush(stdout); + } +#endif } -- cgit v1.2.3 From 5fdaebc51c5433ebc73f89690fd435dae2fcef60 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 18 Mar 2008 10:26:45 -0700 Subject: cell: Minor changes to make stencil not crash I'm not sure these are quite correct. The reflect demo doesn't assert anymore, but it doesn't produce correct results either. SPE-based vertex shader code needs to be disabled for relfect to run. --- src/gallium/drivers/cell/spu/spu_main.c | 12 +++++++++--- src/gallium/drivers/cell/spu/spu_tile.h | 4 ++-- 2 files changed, 11 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index b8bb8e9449..122cf337a6 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -218,12 +218,18 @@ cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) spu.fb.width_tiles = (spu.fb.width + TILE_SIZE - 1) / TILE_SIZE; spu.fb.height_tiles = (spu.fb.height + TILE_SIZE - 1) / TILE_SIZE; - if (spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM) + switch (spu.fb.depth_format) { + case PIPE_FORMAT_Z32_UNORM: + case PIPE_FORMAT_Z24S8_UNORM: spu.fb.zsize = 4; - else if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) + break; + case PIPE_FORMAT_Z16_UNORM: spu.fb.zsize = 2; - else + break; + default: spu.fb.zsize = 0; + break; + } if (spu.fb.color_format == PIPE_FORMAT_A8R8G8B8_UNORM) spu.color_shuffle = ((vector unsigned char) { diff --git a/src/gallium/drivers/cell/spu/spu_tile.h b/src/gallium/drivers/cell/spu/spu_tile.h index 3105b848fd..1b5491112d 100644 --- a/src/gallium/drivers/cell/spu/spu_tile.h +++ b/src/gallium/drivers/cell/spu/spu_tile.h @@ -56,13 +56,13 @@ clear_c_tile(tile_t *ctile) static INLINE void clear_z_tile(tile_t *ztile) { - if (spu.fb.depth_format == PIPE_FORMAT_Z16_UNORM) { + if (spu.fb.zsize == 2) { memset16((ushort*) ztile->us, spu.fb.depth_clear_value, TILE_SIZE * TILE_SIZE); } else { - ASSERT(spu.fb.depth_format == PIPE_FORMAT_Z32_UNORM); + ASSERT(spu.fb.zsize != 0); memset32((uint*) ztile->ui, spu.fb.depth_clear_value, TILE_SIZE * TILE_SIZE); -- cgit v1.2.3 From f3e686d24a54f65b98c9a61a952577faaf451148 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 18 Mar 2008 10:29:14 -0700 Subject: cell: Fix added whitespace errors. --- src/gallium/drivers/cell/ppu/cell_state_per_fragment.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c index 0ad9344cd4..11c0aa5b23 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -532,12 +532,12 @@ cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa) const uint32_t *p = f->store; unsigned i; - printf("# alpha (%sabled)\n", + printf("# alpha (%sabled)\n", (dsa->alpha.enabled) ? "en" : "dis"); printf("# func: %u\n", dsa->alpha.func); printf("# ref: %.2f\n", dsa->alpha.ref); - printf("# depth (%sabled)\n", + printf("# depth (%sabled)\n", (dsa->depth.enabled) ? "en" : "dis"); printf("# func: %u\n", dsa->depth.func); -- cgit v1.2.3 From 17b234ae3319d8a36afc44d0cceb30fea6b42d67 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 18 Mar 2008 11:47:37 -0700 Subject: cell: Fix depth read / write for s8z24. Stencil is still broken. --- src/gallium/drivers/cell/spu/spu_main.c | 1 + src/gallium/drivers/cell/spu/spu_per_fragment_op.c | 25 ++++++++++++++++++++-- 2 files changed, 24 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 122cf337a6..937962285d 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -221,6 +221,7 @@ cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) switch (spu.fb.depth_format) { case PIPE_FORMAT_Z32_UNORM: case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: spu.fb.zsize = 4; break; case PIPE_FORMAT_Z16_UNORM: diff --git a/src/gallium/drivers/cell/spu/spu_per_fragment_op.c b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c index d42b522b41..06d68f5604 100644 --- a/src/gallium/drivers/cell/spu/spu_per_fragment_op.c +++ b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c @@ -76,6 +76,16 @@ read_ds_quad(tile_t *buffer, unsigned x, unsigned y, case PIPE_FORMAT_Z24S8_UNORM: { + qword *ptr = (qword *) &buffer->ui4[iy][ix]; + qword mask = si_fsmbi(0xEEEE); + + *depth = si_rotmai(si_and(*ptr, mask), -8); + *stencil = si_andc(*ptr, mask); + break; + } + + + case PIPE_FORMAT_S8Z24_UNORM: { qword *ptr = (qword *) &buffer->ui4[iy][ix]; qword mask = si_fsmbi(0x7777); @@ -124,10 +134,20 @@ write_ds_quad(tile_t *buffer, unsigned x, unsigned y, case PIPE_FORMAT_Z24S8_UNORM: { + qword *ptr = (qword *) &buffer->ui4[iy][ix]; + qword mask = si_fsmbi(0xEEEE); + + depth = si_shli(depth, 8); + *ptr = si_selb(stencil, depth, mask); + break; + } + + + case PIPE_FORMAT_S8Z24_UNORM: { qword *ptr = (qword *) &buffer->ui4[iy][ix]; qword mask = si_fsmbi(0x7777); - stencil = si_rotmai(stencil, 24); + stencil = si_shli(stencil, 24); *ptr = si_selb(stencil, depth, mask); break; } @@ -167,11 +187,12 @@ spu_do_depth_stencil(int x, int y, frag_depth = si_cfltu(frag_depth, 0); break; case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: frag_depth = si_fm(frag_depth, (qword)spu_splats((float)(0x00ffffffu))); frag_depth = si_cfltu(frag_depth, 0); break; default: - assert(0); + ASSERT(0); break; } -- cgit v1.2.3 From 35a1ec53a7728311de22124c14b93dbbdee2ce90 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 18 Mar 2008 13:10:51 +0000 Subject: gallium: make REALLOC a bit more like realloc --- src/gallium/include/pipe/p_util.h | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index ef36ce75f7..c2e0f8c6a5 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -88,14 +88,16 @@ FREE( void *ptr ) static INLINE void * REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) { - void *new_ptr; - if( new_size <= old_size ) { - return old_ptr; - } - new_ptr = MALLOC( new_size ); - if( new_ptr ) { - memcpy( new_ptr, old_ptr, old_size ); + void *new_ptr = NULL; + + if (new_size != 0) { + new_ptr = MALLOC( new_size ); + + if( new_ptr && old_ptr ) { + memcpy( new_ptr, old_ptr, old_size ); + } } + FREE( old_ptr ); return new_ptr; } -- cgit v1.2.3 From fa69a6e1bb7a1fe96848456255e5370f1904706d Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 18 Mar 2008 15:59:06 -0700 Subject: cell: Correctly load stencil for PIPE_FORMAT_S8Z24_UNORM --- src/gallium/drivers/cell/spu/spu_per_fragment_op.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_per_fragment_op.c b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c index 06d68f5604..b4cffeeb32 100644 --- a/src/gallium/drivers/cell/spu/spu_per_fragment_op.c +++ b/src/gallium/drivers/cell/spu/spu_per_fragment_op.c @@ -87,10 +87,9 @@ read_ds_quad(tile_t *buffer, unsigned x, unsigned y, case PIPE_FORMAT_S8Z24_UNORM: { qword *ptr = (qword *) &buffer->ui4[iy][ix]; - qword mask = si_fsmbi(0x7777); - *depth = si_and(*ptr, mask); - *stencil = si_rotmai(si_andc(*ptr, mask), -24); + *depth = si_and(*ptr, si_fsmbi(0x7777)); + *stencil = si_andi(si_roti(*ptr, 8), 0x0ff); break; } -- cgit v1.2.3 From fe40dae02d3f3a83ee5bb66819c1ed4a3f66da80 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 18 Mar 2008 15:59:33 -0700 Subject: cell: Fix various stencil test code-gen bugs --- .../drivers/cell/ppu/cell_state_per_fragment.c | 70 +++++++++++++--------- 1 file changed, 43 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c index 11c0aa5b23..9c47968459 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -199,52 +199,62 @@ emit_depth_test(struct pipe_depth_stencil_alpha_state *dsa, /** * \note Emits a maximum of 5 instructions. + * + * \warning + * Since \c out and \c in might be the same register, this routine cannot + * generate code that uses \c out as a temporary. */ static void emit_stencil_op(struct spe_function *f, int out, int in, int mask, unsigned op, unsigned ref) { const int clamp = spe_allocate_available_register(f); - const int tmp = spe_allocate_available_register(f); + const int clamp_mask = spe_allocate_available_register(f); + const int result = spe_allocate_available_register(f); switch(op) { case PIPE_STENCIL_OP_KEEP: assert(0); case PIPE_STENCIL_OP_ZERO: - spe_il(f, out, 0); + spe_il(f, result, 0); break; case PIPE_STENCIL_OP_REPLACE: - spe_il(f, out, ref); + spe_il(f, result, ref); break; case PIPE_STENCIL_OP_INCR: spe_il(f, clamp, 0x0ff); - spe_ai(f, out, in, 1); - spe_cgti(f, tmp, out, clamp); - spe_selb(f, out, out, clamp, tmp); + spe_ai(f, result, in, 1); + spe_clgti(f, clamp_mask, result, 0x0ff); + spe_selb(f, result, result, clamp, clamp_mask); break; case PIPE_STENCIL_OP_DECR: spe_il(f, clamp, 0); - spe_ai(f, out, in, -1); - spe_cgti(f, tmp, out, clamp); - spe_selb(f, out, clamp, out, tmp); + spe_ai(f, result, in, -1); + + /* If "(s-1) < 0" in signed arithemtic, then "(s-1) > MAX" in unsigned + * arithmetic. + */ + spe_clgti(f, clamp_mask, result, 0x0ff); + spe_selb(f, result, result, clamp, clamp_mask); break; case PIPE_STENCIL_OP_INCR_WRAP: - spe_ai(f, out, in, 1); + spe_ai(f, result, in, 1); break; case PIPE_STENCIL_OP_DECR_WRAP: - spe_ai(f, out, in, -1); + spe_ai(f, result, in, -1); break; case PIPE_STENCIL_OP_INVERT: - spe_nor(f, out, in, in); + spe_nor(f, result, in, in); break; default: assert(0); } - spe_release_register(f, tmp); - spe_release_register(f, clamp); + spe_selb(f, out, in, result, mask); - spe_selb(f, out, in, out, mask); + spe_release_register(f, result); + spe_release_register(f, clamp_mask); + spe_release_register(f, clamp); } @@ -261,13 +271,13 @@ emit_stencil_op(struct spe_function *f, * \param stencil Register containing values from stencil buffer * \param depth_pass Register to store mask of fragments passing stencil test * and depth test - * + * * \note * Emits a maximum of 10 + (3 * 5) = 25 instructions. */ static int emit_stencil_test(struct pipe_depth_stencil_alpha_state *dsa, - unsigned face, + unsigned face, struct spe_function *f, int mask, int depth_mask, @@ -284,14 +294,17 @@ emit_stencil_test(struct pipe_depth_stencil_alpha_state *dsa, const unsigned ref = (dsa->stencil[face].ref_value & dsa->stencil[face].value_mask); boolean complement = FALSE; - int stored = spe_allocate_available_register(f); + int stored; int tmp = spe_allocate_available_register(f); if ((dsa->stencil[face].func != PIPE_FUNC_NEVER) && (dsa->stencil[face].func != PIPE_FUNC_ALWAYS) && (dsa->stencil[face].value_mask != 0x0ff)) { + stored = spe_allocate_available_register(f); spe_andi(f, stored, stencil, dsa->stencil[face].value_mask); + } else { + stored = stencil; } @@ -332,7 +345,9 @@ emit_stencil_test(struct pipe_depth_stencil_alpha_state *dsa, break; } - spe_release_register(f, stored); + if (stored != stencil) { + spe_release_register(f, stored); + } spe_release_register(f, tmp); @@ -362,7 +377,7 @@ emit_stencil_test(struct pipe_depth_stencil_alpha_state *dsa, /* Conditionally emit code to update the stencil value under various * condititons. Note that there is no need to generate code under the * following circumstances: - * + * * - Stencil write mask is zero. * - For stencil-fail if the stencil test is ALWAYS * - For depth-fail if the stencil test is NEVER @@ -425,7 +440,7 @@ emit_stencil_test(struct pipe_depth_stencil_alpha_state *dsa, spe_release_register(f, face_stencil); } else if (dsa->stencil[face].write_mask != 0x0ff) { int tmp = spe_allocate_available_register(f); - + spe_il(f, tmp, dsa->stencil[face].write_mask); spe_selb(f, stencil_src, stencil, stencil_src, tmp); @@ -492,12 +507,12 @@ cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa) if (front_stencil != back_stencil) { spe_selb(f, stencil, back_stencil, front_stencil, facing); } - - if (back_stencil != stencil) { + + if (back_stencil != stencil) { spe_release_register(f, back_stencil); } - if (front_stencil != stencil) { + if (front_stencil != stencil) { spe_release_register(f, front_stencil); } @@ -505,10 +520,11 @@ cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa) spe_release_register(f, back_depth_pass); } else { - if (front_stencil != stencil) { + if (front_stencil != stencil) { spe_or(f, stencil, front_stencil, front_stencil); spe_release_register(f, front_stencil); } + spe_or(f, mask, front_depth_pass, front_depth_pass); } spe_release_register(f, front_depth_pass); @@ -997,7 +1013,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, blend_color->color[3], frag[3], pixel[3]); } - + if (sF[0] == sF[3]) { src_factor[0] = src_factor[3]; @@ -1036,7 +1052,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, frag, pixel, dst_factor); } - + func[0] = b->rgb_func; func[1] = func[0]; -- cgit v1.2.3 From bab9209e12ec16ef3b33d46be8e6154f8c8f182d Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 15:15:02 -0600 Subject: gallium: restore additional state after clearing with quad --- src/mesa/state_tracker/st_cb_clear.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 8223784d2e..693cddedf7 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -357,6 +357,8 @@ clear_with_quad(GLcontext *ctx, st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); #else /* Restore pipe state */ + cso_set_blend(st->cso_context, &st->state.blend); + cso_set_depth_stencil_alpha(st->cso_context, &st->state.depth_stencil); cso_set_rasterizer(st->cso_context, &st->state.rasterizer); pipe->bind_fs_state(pipe, st->fp->driver_shader); pipe->bind_vs_state(pipe, st->vp->driver_shader); -- cgit v1.2.3 From 31970c4633c5000916b0a36022ff761038f5cf5a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 16:56:55 -0600 Subject: gallium: utilities for creating simple vertex/fragment shaders --- src/gallium/auxiliary/util/u_simple_shaders.c | 263 ++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_simple_shaders.h | 52 +++++ 2 files changed, 315 insertions(+) create mode 100644 src/gallium/auxiliary/util/u_simple_shaders.c create mode 100644 src/gallium/auxiliary/util/u_simple_shaders.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c new file mode 100644 index 0000000000..88e2ab05bd --- /dev/null +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -0,0 +1,263 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Simple vertex/fragment shader generators. + * + * @author Brian Paul + */ + + +#include "pipe/p_context.h" +#include "pipe/p_debug.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#include "pipe/p_shader_tokens.h" + +#include "util/u_simple_shaders.h" + +#include "tgsi/util/tgsi_build.h" +#include "tgsi/util/tgsi_dump.h" +#include "tgsi/util/tgsi_parse.h" + + + +/** + * Make simple vertex pass-through shader. + */ +void * +util_make_vertex_passthrough_shader(struct pipe_context *pipe, + uint num_attribs, + const uint *semantic_names, + const uint *semantic_indexes) +{ + uint maxTokens = 100; + struct tgsi_token *tokens; + struct tgsi_header *header; + struct tgsi_processor *processor; + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + const uint procType = TGSI_PROCESSOR_VERTEX; + uint ti, i; + struct pipe_shader_state shader; + + tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); + + /* shader header + */ + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header(); + + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor( procType, header ); + + ti = 3; + + /* declare inputs */ + for (i = 0; i < num_attribs; i++) { + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + /* + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; + decl.Semantic.SemanticIndex = 0; + */ + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + } + + /* declare outputs */ + for (i = 0; i < num_attribs; i++) { + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = semantic_names[i]; + decl.Semantic.SemanticIndex = semantic_indexes[i]; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + } + + /* emit MOV instructions */ + for (i = 0; i < num_attribs; i++) { + /* MOVE out[i], in[i]; */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_MOV; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + inst.FullDstRegisters[0].DstRegister.Index = i; + inst.Instruction.NumSrcRegs = 1; + inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + inst.FullSrcRegisters[0].SrcRegister.Index = i; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + } + + /* END instruction */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_END; + inst.Instruction.NumDstRegs = 0; + inst.Instruction.NumSrcRegs = 0; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + +#if 0 /*debug*/ + tgsi_dump(tokens, 0); +#endif + + shader.tokens = tokens; + return pipe->create_vs_state(pipe, &shader); +} + + + + +/** + * Make simple fragment texture shader: + * TEX OUT[0], IN[0], SAMP[0], 2D; + * END; + */ +void * +util_make_fragment_tex_shader(struct pipe_context *pipe) +{ + uint maxTokens = 100; + struct tgsi_token *tokens; + struct tgsi_header *header; + struct tgsi_processor *processor; + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + const uint procType = TGSI_PROCESSOR_FRAGMENT; + uint ti; + struct pipe_shader_state shader; + + tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); + + /* shader header + */ + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header(); + + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor( procType, header ); + + ti = 3; + + /* declare TEX[0] input */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; + decl.Semantic.SemanticIndex = 0; + /* XXX this could be linear... */ + decl.Declaration.Interpolate = 1; + decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* declare color[0] output */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; + decl.Semantic.SemanticIndex = 0; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* declare sampler */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_SAMPLER; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* TEX instruction */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_TEX; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + inst.FullDstRegisters[0].DstRegister.Index = 0; + inst.Instruction.NumSrcRegs = 2; + inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; + inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + inst.FullSrcRegisters[0].SrcRegister.Index = 0; + inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + inst.FullSrcRegisters[1].SrcRegister.Index = 0; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + + /* END instruction */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_END; + inst.Instruction.NumDstRegs = 0; + inst.Instruction.NumSrcRegs = 0; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + +#if 0 /*debug*/ + tgsi_dump(tokens, 0); +#endif + + shader.tokens = tokens; + return pipe->create_fs_state(pipe, &shader); +} + diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h new file mode 100644 index 0000000000..3ef4f28801 --- /dev/null +++ b/src/gallium/auxiliary/util/u_simple_shaders.h @@ -0,0 +1,52 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef U_SIMPLE_SHADERS_H +#define U_SIMPLE_SHADERS_H + + +#include "pipe/p_compiler.h" + + +struct pipe_context; + + +extern void * +util_make_vertex_passthrough_shader(struct pipe_context *pipe, + uint num_attribs, + const uint *semantic_names, + const uint *semantic_indexes); + + +extern void * +util_make_fragment_tex_shader(struct pipe_context *pipe); + + +#endif + + -- cgit v1.2.3 From 4bd2b74441092154f0d0048822c6e36cfcc183af Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 16:57:23 -0600 Subject: gallium: use new simple shader utility routines --- src/gallium/auxiliary/util/u_gen_mipmap.c | 260 ++---------------------------- 1 file changed, 12 insertions(+), 248 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index d47c677074..e32ce768c4 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -43,6 +43,7 @@ #include "util/u_draw_quad.h" #include "util/u_gen_mipmap.h" +#include "util/u_simple_shaders.h" #include "tgsi/util/tgsi_build.h" #include "tgsi/util/tgsi_dump.h" @@ -655,252 +656,6 @@ fallback_gen_mipmap(struct gen_mipmap_state *ctx, } -/** - * Make simple fragment shader: - * TEX OUT[0], IN[0], SAMP[0], 2D; - * END; - */ -static void -make_fragment_shader(struct gen_mipmap_state *ctx) -{ - struct pipe_context *pipe = ctx->pipe; - uint maxTokens = 100; - struct tgsi_token *tokens; - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_declaration decl; - struct tgsi_full_instruction inst; - const uint procType = TGSI_PROCESSOR_FRAGMENT; - uint ti = 0; - struct pipe_shader_state shader; - - tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); - - /* shader header - */ - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); - - ti = 3; - - /* declare TEX[0] input */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; - decl.Semantic.SemanticIndex = 0; - /* XXX this could be linear... */ - decl.Declaration.Interpolate = 1; - decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - maxTokens - ti); - - /* declare color[0] output */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; - decl.Semantic.SemanticIndex = 0; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - maxTokens - ti); - - /* declare sampler */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_SAMPLER; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - maxTokens - ti); - - /* TEX instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_TEX; - inst.Instruction.NumDstRegs = 1; - inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - inst.FullDstRegisters[0].DstRegister.Index = 0; - inst.Instruction.NumSrcRegs = 2; - inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; - inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - inst.FullSrcRegisters[0].SrcRegister.Index = 0; - inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - inst.FullSrcRegisters[1].SrcRegister.Index = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - maxTokens - ti ); - - /* END instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_END; - inst.Instruction.NumDstRegs = 0; - inst.Instruction.NumSrcRegs = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - maxTokens - ti ); - -#if 0 /*debug*/ - tgsi_dump(tokens, 0); -#endif - - shader.tokens = tokens; - ctx->fs = pipe->create_fs_state(pipe, &shader); -} - - -/** - * Make simple fragment shader: - * MOV OUT[0], IN[0]; - * MOV OUT[1], IN[1]; - * END; - * - * XXX eliminate this when vertex passthrough-mode is more solid. - */ -static void -make_vertex_shader(struct gen_mipmap_state *ctx) -{ - struct pipe_context *pipe = ctx->pipe; - uint maxTokens = 100; - struct tgsi_token *tokens; - struct tgsi_header *header; - struct tgsi_processor *processor; - struct tgsi_full_declaration decl; - struct tgsi_full_instruction inst; - const uint procType = TGSI_PROCESSOR_VERTEX; - uint ti = 0; - struct pipe_shader_state shader; - - tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); - - /* shader header - */ - *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); - - header = (struct tgsi_header *) &tokens[1]; - *header = tgsi_build_header(); - - processor = (struct tgsi_processor *) &tokens[2]; - *processor = tgsi_build_processor( procType, header ); - - ti = 3; - - /* declare POS input */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - /* - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; - decl.Semantic.SemanticIndex = 0; - */ - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - maxTokens - ti); - /* declare TEX[0] input */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - /* - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; - decl.Semantic.SemanticIndex = 0; - */ - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 1; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - maxTokens - ti); - - /* declare POS output */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; - decl.Semantic.SemanticIndex = 0; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - maxTokens - ti); - - /* declare TEX[0] output */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; - decl.Semantic.SemanticIndex = 0; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 1; - ti += tgsi_build_full_declaration(&decl, - &tokens[ti], - header, - maxTokens - ti); - - /* MOVE out[0], in[0]; # POS */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_MOV; - inst.Instruction.NumDstRegs = 1; - inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - inst.FullDstRegisters[0].DstRegister.Index = 0; - inst.Instruction.NumSrcRegs = 1; - inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - inst.FullSrcRegisters[0].SrcRegister.Index = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - maxTokens - ti ); - - /* MOVE out[1], in[1]; # TEX */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_MOV; - inst.Instruction.NumDstRegs = 1; - inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - inst.FullDstRegisters[0].DstRegister.Index = 1; - inst.Instruction.NumSrcRegs = 1; - inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - inst.FullSrcRegisters[0].SrcRegister.Index = 1; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - maxTokens - ti ); - - /* END instruction */ - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = TGSI_OPCODE_END; - inst.Instruction.NumDstRegs = 0; - inst.Instruction.NumSrcRegs = 0; - ti += tgsi_build_full_instruction(&inst, - &tokens[ti], - header, - maxTokens - ti ); - -#if 0 /*debug*/ - tgsi_dump(tokens, 0); -#endif - - shader.tokens = tokens; - ctx->vs = pipe->create_vs_state(pipe, &shader); -} - - /** * Create a mipmap generation context. * The idea is to create one of these and re-use it each time we need to @@ -953,8 +708,17 @@ util_create_gen_mipmap(struct pipe_context *pipe) ctx->viewport.translate[3] = 0.0; #endif - make_vertex_shader(ctx); - make_fragment_shader(ctx); + /* vertex shader */ + { + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, + TGSI_SEMANTIC_GENERIC }; + const uint semantic_indexes[] = { 0, 0 }; + ctx->vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names, + semantic_indexes); + } + + /* fragment shader */ + ctx->fs = util_make_fragment_tex_shader(pipe); return ctx; } -- cgit v1.2.3 From e5a20499d84ba8d270c3144f2c1c7615eeb077c7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 16:58:15 -0600 Subject: gallium: make the gen_mipmap_state struct private --- src/gallium/auxiliary/util/u_gen_mipmap.c | 13 +++++++++++++ src/gallium/auxiliary/util/u_gen_mipmap.h | 13 +------------ 2 files changed, 14 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index e32ce768c4..e18f8ab72a 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -50,6 +50,19 @@ #include "tgsi/util/tgsi_parse.h" +struct gen_mipmap_state +{ + struct pipe_context *pipe; + + void *blend; + void *depthstencil; + void *rasterizer; + /*struct pipe_viewport_state viewport;*/ + struct pipe_sampler_state *vs; + struct pipe_sampler_state *fs; +}; + + enum dtype { diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.h b/src/gallium/auxiliary/util/u_gen_mipmap.h index 6ee909f0a6..80496140a2 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.h +++ b/src/gallium/auxiliary/util/u_gen_mipmap.h @@ -31,18 +31,7 @@ #include "pipe/p_state.h" -struct gen_mipmap_state -{ - struct pipe_context *pipe; - - void *blend; - void *depthstencil; - void *rasterizer; - /*struct pipe_viewport_state viewport;*/ - struct pipe_sampler_state *vs; - struct pipe_sampler_state *fs; -}; - +struct gen_mipmap_state; extern struct gen_mipmap_state * -- cgit v1.2.3 From 8de9503d5d290c475a1940a71e58086ab3735f83 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 16:59:15 -0600 Subject: added u_simple_shaders.c --- src/gallium/auxiliary/util/Makefile | 1 + src/gallium/auxiliary/util/SConscript | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index d11fa6e803..7af9b35d4b 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -12,6 +12,7 @@ C_SOURCES = \ u_handle_table.c \ u_hash_table.c \ u_mm.c \ + u_simple_shaders.c \ u_snprintf.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 6d4ba8643d..0eb1b8d9da 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -11,6 +11,7 @@ util = env.ConvenienceLibrary( 'u_handle_table.c', 'u_hash_table.c', 'u_mm.c', + 'u_simple_shaders.c', 'u_snprintf.c', ]) -- cgit v1.2.3 From 18f9fa9e71abdd999e49ef78729bfe3d92304312 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 16:59:46 -0600 Subject: gallium: protect against multi-#include --- src/gallium/auxiliary/util/u_pack_color.h | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 93a18c1c7e..cd13823985 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -31,7 +31,12 @@ */ +#ifndef U_PACK_COLOR_H +#define U_PACK_COLOR_H + + #include "pipe/p_compiler.h" +#include "pipe/p_format.h" /** @@ -123,3 +128,6 @@ util_pack_z(enum pipe_format format, double z) return 0; } } + + +#endif /* U_PACK_COLOR_H */ -- cgit v1.2.3 From 0abe462128a8a0725e006751e553f89fee2d7fa7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 17:04:59 -0600 Subject: gallium: new pixel blit code Copy rectangular region from one surface to another w/ scaling. Disables most fragment operations. --- src/gallium/auxiliary/util/Makefile | 1 + src/gallium/auxiliary/util/SConscript | 1 + src/gallium/auxiliary/util/u_blit.c | 276 ++++++++++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_blit.h | 60 ++++++++ 4 files changed, 338 insertions(+) create mode 100644 src/gallium/auxiliary/util/u_blit.c create mode 100644 src/gallium/auxiliary/util/u_blit.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 7af9b35d4b..9b6c2708b6 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -7,6 +7,7 @@ C_SOURCES = \ p_debug.c \ p_tile.c \ p_util.c \ + u_blit.c \ u_draw_quad.c \ u_gen_mipmap.c \ u_handle_table.c \ diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 0eb1b8d9da..b44f2d5e39 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -6,6 +6,7 @@ util = env.ConvenienceLibrary( 'p_debug.c', 'p_tile.c', 'p_util.c', + 'u_blit.c', 'u_draw_quad.c', 'u_gen_mipmap.c', 'u_handle_table.c', diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c new file mode 100644 index 0000000000..4b4ab8185f --- /dev/null +++ b/src/gallium/auxiliary/util/u_blit.c @@ -0,0 +1,276 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Copy/blit pixel rect between surfaces + * + * @author Brian Paul + */ + + +#include "pipe/p_context.h" +#include "pipe/p_debug.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#include "pipe/p_shader_tokens.h" + +#include "util/u_draw_quad.h" +#include "util/u_blit.h" +#include "util/u_simple_shaders.h" + + +struct blit_state +{ + struct pipe_context *pipe; + + void *blend; + void *depthstencil; + void *rasterizer; + void *samplers[2]; /* one for linear, one for nearest sampling */ + + /*struct pipe_viewport_state viewport;*/ + struct pipe_sampler_state *vs; + struct pipe_sampler_state *fs; +}; + + +/** + * Create state object for blit. + * Intended to be created once and re-used for many blit() calls. + */ +struct blit_state * +util_create_blit(struct pipe_context *pipe) +{ + struct pipe_blend_state blend; + struct pipe_depth_stencil_alpha_state depthstencil; + struct pipe_rasterizer_state rasterizer; + struct blit_state *ctx; + struct pipe_sampler_state sampler; + + ctx = CALLOC_STRUCT(blit_state); + if (!ctx) + return NULL; + + ctx->pipe = pipe; + + /* we don't use blending, but need to set valid values */ + memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.colormask = PIPE_MASK_RGBA; + ctx->blend = pipe->create_blend_state(pipe, &blend); + + /* depth/stencil/alpha */ + memset(&depthstencil, 0, sizeof(depthstencil)); + ctx->depthstencil = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil); + + /* rasterizer */ + memset(&rasterizer, 0, sizeof(rasterizer)); + rasterizer.front_winding = PIPE_WINDING_CW; + rasterizer.cull_mode = PIPE_WINDING_NONE; + rasterizer.bypass_clipping = 1; /* bypasses viewport too */ + /*rasterizer.bypass_vs = 1;*/ + ctx->rasterizer = pipe->create_rasterizer_state(pipe, &rasterizer); + + /* samplers */ + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + sampler.normalized_coords = 1; + ctx->samplers[0] = pipe->create_sampler_state(pipe, &sampler); + + sampler.min_img_filter = PIPE_TEX_MIPFILTER_LINEAR; + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_LINEAR; + ctx->samplers[1] = pipe->create_sampler_state(pipe, &sampler); + + +#if 0 + /* viewport */ + ctx->viewport.scale[0] = 1.0; + ctx->viewport.scale[1] = 1.0; + ctx->viewport.scale[2] = 1.0; + ctx->viewport.scale[3] = 1.0; + ctx->viewport.translate[0] = 0.0; + ctx->viewport.translate[1] = 0.0; + ctx->viewport.translate[2] = 0.0; + ctx->viewport.translate[3] = 0.0; +#endif + + /* vertex shader */ + { + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, + TGSI_SEMANTIC_GENERIC }; + const uint semantic_indexes[] = { 0, 0 }; + ctx->vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names, + semantic_indexes); + } + + /* fragment shader */ + ctx->fs = util_make_fragment_tex_shader(pipe); + + return ctx; +} + + +/** + * Destroy a blit context + */ +void +util_destroy_blit(struct blit_state *ctx) +{ + struct pipe_context *pipe = ctx->pipe; + + pipe->delete_blend_state(pipe, ctx->blend); + pipe->delete_depth_stencil_alpha_state(pipe, ctx->depthstencil); + pipe->delete_rasterizer_state(pipe, ctx->rasterizer); + pipe->delete_sampler_state(pipe, ctx->samplers[0]); + pipe->delete_sampler_state(pipe, ctx->samplers[1]); + + pipe->delete_vs_state(pipe, ctx->vs); + pipe->delete_fs_state(pipe, ctx->fs); + + FREE(ctx); +} + + +/** + * Copy pixel block from src surface to dst surface. + * Overlapping regions are acceptable. + * XXX need some control over blitting Z and/or stencil. + */ +void +util_blit_pixels(struct blit_state *ctx, + struct pipe_surface *src, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_texture texTemp, *tex; + struct pipe_surface *texSurf; + struct pipe_framebuffer_state fb; + const int srcW = abs(srcX1 - srcX0); + const int srcH = abs(srcY1 - srcY0); + const int srcLeft = MIN2(srcX0, srcX1); + const int srcTop = MIN2(srcY0, srcY1); + + assert(filter == PIPE_TEX_MIPFILTER_NEAREST || + filter == PIPE_TEX_MIPFILTER_LINEAR); + + if (srcLeft != srcX0) { + /* left-right flip */ + int tmp = dstX0; + dstX0 = dstX1; + dstX1 = tmp; + } + + if (srcTop != srcY0) { + /* up-down flip */ + int tmp = dstY0; + dstY0 = dstY1; + dstY1 = tmp; + } + + /* + * XXX for now we're always creating a temporary texture. + * Strictly speaking that's not always needed. + */ + + /* create temp texture */ + memset(&texTemp, 0, sizeof(texTemp)); + texTemp.target = PIPE_TEXTURE_2D; + texTemp.format = src->format; + texTemp.last_level = 0; + texTemp.width[0] = srcW; + texTemp.height[0] = srcH; + texTemp.depth[0] = 1; + texTemp.compressed = 0; + texTemp.cpp = pf_get_bits(src->format) / 8; + + tex = screen->texture_create(screen, &texTemp); + if (!tex) + return; + + texSurf = screen->get_tex_surface(screen, tex, 0, 0, 0); + + /* load temp texture */ + pipe->surface_copy(pipe, FALSE, + texSurf, 0, 0, /* dest */ + src, srcLeft, srcTop, /* src */ + srcW, srcH); /* size */ + + /* drawing dest */ + memset(&fb, 0, sizeof(fb)); + fb.num_cbufs = 1; + fb.cbufs[0] = dst; + pipe->set_framebuffer_state(pipe, &fb); + + /* sampler */ + if (filter == PIPE_TEX_MIPFILTER_NEAREST) + pipe->bind_sampler_states(pipe, 1, &ctx->samplers[0]); + else + pipe->bind_sampler_states(pipe, 1, &ctx->samplers[1]); + + /* texture */ + pipe->set_sampler_textures(pipe, 1, &tex); + + /* shaders */ + pipe->bind_fs_state(pipe, ctx->fs); + pipe->bind_vs_state(pipe, ctx->vs); + + /* misc state */ + pipe->bind_blend_state(pipe, ctx->blend); + pipe->bind_depth_stencil_alpha_state(pipe, ctx->depthstencil); + pipe->bind_rasterizer_state(pipe, ctx->rasterizer); + + /* draw quad */ + util_draw_texquad(pipe, dstX0, dstY0, dstX1, dstY1, z); + + /* unbind */ + pipe->set_sampler_textures(pipe, 0, NULL); + pipe->bind_sampler_states(pipe, 0, NULL); + + /* free stuff */ + pipe_surface_reference(&texSurf, NULL); + screen->texture_release(screen, &tex); + + /* Note: caller must restore pipe/gallium state at this time */ +} + diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h new file mode 100644 index 0000000000..0a8e05f88b --- /dev/null +++ b/src/gallium/auxiliary/util/u_blit.h @@ -0,0 +1,60 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef U_BLIT_H +#define U_BLIT_H + + +struct pipe_context; +struct pipe_surface; + + +struct copy_pixels_state; + + +extern struct blit_state * +util_create_blit(struct pipe_context *pipe); + + +extern void +util_destroy_blit(struct blit_state *ctx); + + + +extern void +util_blit_pixels(struct blit_state *ctx, + struct pipe_surface *src, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter); + + +#endif -- cgit v1.2.3 From 320da13c87c683cbe6e8145a9258ea2b7ef674cd Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 17:12:51 -0600 Subject: gallium: fix typos --- src/mesa/state_tracker/st_gen_mipmap.c | 2 +- src/mesa/state_tracker/st_gen_mipmap.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 5f09d9bb06..6c3afca1ba 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -64,7 +64,7 @@ st_init_generate_mipmap(struct st_context *st) void -st_destroy_generate_mipmpap(struct st_context *st) +st_destroy_generate_mipmap(struct st_context *st) { util_destroy_gen_mipmap(st->gen_mipmap); st->gen_mipmap = NULL; diff --git a/src/mesa/state_tracker/st_gen_mipmap.h b/src/mesa/state_tracker/st_gen_mipmap.h index 7668c1e44e..00fbae9302 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.h +++ b/src/mesa/state_tracker/st_gen_mipmap.h @@ -35,7 +35,7 @@ st_init_generate_mipmap(struct st_context *st); extern void -st_destroy_generate_mipmpap(struct st_context *st); +st_destroy_generate_mipmap(struct st_context *st); extern void -- cgit v1.2.3 From 84836ffce1ca322f91aaae25c4de2be20e55a4b5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 17:13:14 -0600 Subject: gallium: glFramebufferBlitEXT() implementation --- src/mesa/state_tracker/st_cb_blit.c | 125 ++++++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_blit.h | 46 +++++++++++++ 2 files changed, 171 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_blit.c create mode 100644 src/mesa/state_tracker/st_cb_blit.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c new file mode 100644 index 0000000000..dfa79c975c --- /dev/null +++ b/src/mesa/state_tracker/st_cb_blit.c @@ -0,0 +1,125 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Brian Paul + */ + +#include "main/imports.h" +#include "main/image.h" +#include "main/macros.h" +#include "main/texformat.h" +#include "shader/program.h" +#include "shader/prog_parameter.h" +#include "shader/prog_print.h" + +#include "st_context.h" +#include "st_program.h" +#include "st_cb_drawpixels.h" +#include "st_cb_blit.h" +#include "st_cb_fbo.h" + +#include "util/u_blit.h" + +#include "cso_cache/cso_context.h" + + +void +st_init_blit(struct st_context *st) +{ + st->blit = util_create_blit(st->pipe); +} + + +void +st_destroy_blit(struct st_context *st) +{ + util_destroy_blit(st->blit); + st->blit = NULL; +} + + +static void +st_BlitFramebuffer(GLcontext *ctx, + GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, + GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, + GLbitfield mask, GLenum filter) +{ + struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; + + const uint pFilter = ((filter == GL_NEAREST) + ? PIPE_TEX_MIPFILTER_NEAREST + : PIPE_TEX_MIPFILTER_LINEAR); + + if (mask & GL_COLOR_BUFFER_BIT) { + struct st_renderbuffer *srcRb = + st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + struct st_renderbuffer *dstRb = + st_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0][0]); + struct pipe_surface *srcSurf = srcRb->surface; + struct pipe_surface *dstSurf = dstRb->surface; + + srcY0 = srcRb->Base.Height - srcY0; + srcY1 = srcRb->Base.Height - srcY1; + + dstY0 = dstRb->Base.Height - dstY0; + dstY1 = dstRb->Base.Height - dstY1; + + util_blit_pixels(st->blit, + srcSurf, srcX0, srcY0, srcX1, srcY1, + dstSurf, dstX0, dstY0, dstX1, dstY1, + 0.0, pFilter); + + } + +#if 0 + /* XXX is this sufficient? */ + st_invalidate_state(ctx, _NEW_COLOR | _NEW_TEXTURE); +#else + /* need to "unset" cso state because we went behind the back of the cso + * tracker. Without unset, the _set_ calls would be no-ops. + */ + cso_unset_blend(st->cso_context); + cso_unset_depth_stencil_alpha(st->cso_context); + cso_unset_rasterizer(st->cso_context); + cso_set_blend(st->cso_context, &st->state.blend); + cso_set_depth_stencil_alpha(st->cso_context, &st->state.depth_stencil); + cso_set_rasterizer(st->cso_context, &st->state.rasterizer); + pipe->bind_fs_state(pipe, st->fp->driver_shader); + pipe->bind_vs_state(pipe, st->vp->driver_shader); +#endif +} + + + +void +st_init_blit_functions(struct dd_function_table *functions) +{ + functions->BlitFramebuffer = st_BlitFramebuffer; +} diff --git a/src/mesa/state_tracker/st_cb_blit.h b/src/mesa/state_tracker/st_cb_blit.h new file mode 100644 index 0000000000..ed22986b53 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_blit.h @@ -0,0 +1,46 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef ST_CB_BLIT_H +#define ST_CB_BLIT_H + + +#include "st_context.h" + + + +extern void +st_init_blit(struct st_context *st); + +extern void +st_destroy_blit(struct st_context *st); + +extern void +st_init_blit_functions(struct dd_function_table *functions); + + +#endif /* ST_CB_BLIT_H */ -- cgit v1.2.3 From e5b19a0f833b5a3d5ffcf50d25a620d00bd8914b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 17:14:05 -0600 Subject: gallium: added cso_unset_*() functions If we go behind the CSO context's back and set pipe state directly we need to invalidate the CSO's 'current' pointers. This will be revisited... --- src/gallium/auxiliary/cso_cache/cso_context.c | 36 ++++++++++++++++++++++++++- src/gallium/auxiliary/cso_cache/cso_context.h | 13 ++++++++++ 2 files changed, 48 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index fd86bfaca9..294ac82281 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -149,6 +149,12 @@ void cso_set_blend(struct cso_context *ctx, } } +void cso_unset_blend(struct cso_context *ctx) +{ + ctx->blend = NULL; +} + + void cso_single_sampler(struct cso_context *ctx, unsigned idx, const struct pipe_sampler_state *templ) @@ -220,6 +226,15 @@ void cso_set_samplers( struct cso_context *ctx, cso_single_sampler_done( ctx ); } +void cso_unset_samplers( struct cso_context *ctx ) +{ + uint i; + for (i = 0; i < ctx->nr_samplers; i++) + ctx->samplers[i] = NULL; +} + + + void cso_set_depth_stencil_alpha(struct cso_context *ctx, const struct pipe_depth_stencil_alpha_state *templ) { @@ -252,6 +267,11 @@ void cso_set_depth_stencil_alpha(struct cso_context *ctx, } } +void cso_unset_depth_stencil_alpha(struct cso_context *ctx) +{ + ctx->depth_stencil = NULL; +} + void cso_set_rasterizer(struct cso_context *ctx, @@ -285,7 +305,10 @@ void cso_set_rasterizer(struct cso_context *ctx, } } - +void cso_unset_rasterizer(struct cso_context *ctx) +{ + ctx->rasterizer = NULL; +} @@ -320,6 +343,12 @@ void cso_set_fragment_shader(struct cso_context *ctx, } } +void cso_unset_fragment_shader(struct cso_context *ctx) +{ + ctx->fragment_shader = NULL; +} + + void cso_set_vertex_shader(struct cso_context *ctx, const struct pipe_shader_state *templ) { @@ -350,3 +379,8 @@ void cso_set_vertex_shader(struct cso_context *ctx, ctx->pipe->bind_vs_state(ctx->pipe, handle); } } + +void cso_unset_vertex_shader(struct cso_context *ctx) +{ + ctx->vertex_shader = NULL; +} diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h index 1f2a630804..6aa619abf5 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.h +++ b/src/gallium/auxiliary/cso_cache/cso_context.h @@ -44,16 +44,25 @@ struct cso_context *cso_create_context( struct pipe_context *pipe ); void cso_set_blend( struct cso_context *cso, const struct pipe_blend_state *blend ); +void cso_unset_blend(struct cso_context *cso); + void cso_set_depth_stencil_alpha( struct cso_context *cso, const struct pipe_depth_stencil_alpha_state *dsa ); +void cso_unset_depth_stencil_alpha( struct cso_context *cso ); + void cso_set_rasterizer( struct cso_context *cso, const struct pipe_rasterizer_state *rasterizer ); +void cso_unset_rasterizer( struct cso_context *cso ); + void cso_set_samplers( struct cso_context *cso, unsigned count, const struct pipe_sampler_state **states ); +void cso_unset_samplers( struct cso_context *cso ); + + /* Alternate interface to support state trackers that like to modify * samplers one at a time: */ @@ -72,9 +81,13 @@ void cso_single_sampler_done( struct cso_context *cso ); void cso_set_fragment_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); +void cso_unset_fragment_shader( struct cso_context *cso ); + void cso_set_vertex_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); +void cso_unset_vertex_shader( struct cso_context *cso ); + void cso_destroy_context( struct cso_context *cso ); -- cgit v1.2.3 From d946b55543e483a15d8efbe4022d895c8444f6ac Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 17:15:25 -0600 Subject: gallium: s/copy_pixels_state/blit_state --- src/gallium/auxiliary/util/u_blit.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h index 0a8e05f88b..a349be99ad 100644 --- a/src/gallium/auxiliary/util/u_blit.h +++ b/src/gallium/auxiliary/util/u_blit.h @@ -34,7 +34,7 @@ struct pipe_context; struct pipe_surface; -struct copy_pixels_state; +struct blit_state; extern struct blit_state * -- cgit v1.2.3 From ecb873b2b5754adbd47cee4856a549e48e48505f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 17:16:12 -0600 Subject: gallium: added st_cb_blit.c --- src/mesa/sources | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index e3d5f22849..287af7121a 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -172,6 +172,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_texture.c \ state_tracker/st_atom_viewport.c \ state_tracker/st_cb_accum.c \ + state_tracker/st_cb_blit.c \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_flush.c \ -- cgit v1.2.3 From 0df877a0ee68359d106ba84e3dd3b7c42ae20d5a Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 17:16:23 -0600 Subject: gallium: plug in and init GL_EXT_framebuffer_blit function/extension --- src/mesa/state_tracker/st_context.c | 5 +++++ src/mesa/state_tracker/st_context.h | 4 ++++ src/mesa/state_tracker/st_extensions.c | 1 + 3 files changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 5458ab420e..e1fc885e0e 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -36,6 +36,7 @@ #include "st_context.h" #include "st_cb_accum.h" #include "st_cb_bufferobjects.h" +#include "st_cb_blit.h" #include "st_cb_clear.h" #include "st_cb_drawpixels.h" #include "st_cb_fbo.h" @@ -100,6 +101,7 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st_init_atoms( st ); st_init_draw( st ); st_init_generate_mipmap(st); + st_init_blit(st); for (i = 0; i < PIPE_MAX_SAMPLERS; i++) st->state.sampler_list[i] = &st->state.samplers[i]; @@ -151,6 +153,8 @@ static void st_destroy_context_priv( struct st_context *st ) draw_destroy(st->draw); st_destroy_atoms( st ); st_destroy_draw( st ); + st_destroy_generate_mipmap(st); + st_destroy_blit(st); _vbo_DestroyContext(st->ctx); @@ -217,6 +221,7 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_accum_functions(functions); st_init_bufferobject_functions(functions); + st_init_blit_functions(functions); st_init_clear_functions(functions); st_init_drawpixels_functions(functions); st_init_fbo_functions(functions); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 09bab7bfb1..63150dbeaf 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -40,6 +40,9 @@ struct draw_context; struct draw_stage; struct cso_cache; struct cso_blend; +struct gen_mipmap_state; +struct blit_state; + #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 @@ -147,6 +150,7 @@ struct st_context } bitmap; struct gen_mipmap_state *gen_mipmap; + struct blit_state *blit; struct cso_context *cso_context; }; diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 99d2a5fb9e..0962b5f74c 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -143,6 +143,7 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.EXT_blend_logic_op = GL_TRUE; ctx->Extensions.EXT_blend_minmax = GL_TRUE; ctx->Extensions.EXT_blend_subtract = GL_TRUE; + ctx->Extensions.EXT_framebuffer_blit = GL_TRUE; ctx->Extensions.EXT_framebuffer_object = GL_TRUE; ctx->Extensions.EXT_fog_coord = GL_TRUE; ctx->Extensions.EXT_multi_draw_arrays = GL_TRUE; -- cgit v1.2.3 From 9f50a6a24c922926904c215249c3a0426dd433aa Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 18 Mar 2008 17:19:57 -0600 Subject: gallium: added braces to silence warning --- src/gallium/auxiliary/draw/draw_prim.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index 888fa536ea..cb0277fb6c 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -169,11 +169,12 @@ void draw_do_flush( struct draw_context *draw, unsigned flags ) flushing = TRUE; if (flags >= DRAW_FLUSH_SHADER_QUEUE) { - if (draw->vs.queue_nr) + if (draw->vs.queue_nr) { if (draw->rasterizer->bypass_vs) fetch_and_store(draw); else (*draw->shader_queue_flush)(draw); + } if (flags >= DRAW_FLUSH_PRIM_QUEUE) { if (draw->pq.queue_nr) -- cgit v1.2.3 From 6a8bbdba188fc299c46e8d8f11777046737cd183 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Wed, 19 Mar 2008 00:16:16 +0000 Subject: remove redundant init --- src/mesa/tnl/t_vertex_sse.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vertex_sse.c b/src/mesa/tnl/t_vertex_sse.c index ad4cc62d5f..5160a8c3ae 100644 --- a/src/mesa/tnl/t_vertex_sse.c +++ b/src/mesa/tnl/t_vertex_sse.c @@ -348,8 +348,6 @@ static GLboolean build_vertex_emit( struct x86_program *p ) struct x86_reg vp1 = x86_make_reg(file_XMM, 2); GLubyte *fixup, *label; - x86_init_func(&p->func); - /* Push a few regs? */ x86_push(&p->func, countEBP); -- cgit v1.2.3 From ec890533c2852fa62366d449e6fbc899fb0498be Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Wed, 19 Mar 2008 00:18:07 +0000 Subject: set outputs_safe to 0 as it's possible for the code generation to slip over the allocated memory for the vb. --- src/mesa/tnl/t_vertex_sse.c | 2 +- src/mesa/vf/vf_sse.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/tnl/t_vertex_sse.c b/src/mesa/tnl/t_vertex_sse.c index 5160a8c3ae..a180441a5a 100644 --- a/src/mesa/tnl/t_vertex_sse.c +++ b/src/mesa/tnl/t_vertex_sse.c @@ -639,7 +639,7 @@ void _tnl_generate_sse_emit( GLcontext *ctx ) p.ctx = ctx; p.inputs_safe = 0; /* for now */ - p.outputs_safe = 1; /* for now */ + p.outputs_safe = 0; /* for now */ p.have_sse2 = cpu_has_xmm2; p.identity = x86_make_reg(file_XMM, 6); p.chan0 = x86_make_reg(file_XMM, 7); diff --git a/src/mesa/vf/vf_sse.c b/src/mesa/vf/vf_sse.c index 3ce76e2b04..c3a2166578 100644 --- a/src/mesa/vf/vf_sse.c +++ b/src/mesa/vf/vf_sse.c @@ -633,7 +633,7 @@ void vf_generate_sse_emit( struct vertex_fetch *vf ) p.vf = vf; p.inputs_safe = 0; /* for now */ - p.outputs_safe = 1; /* for now */ + p.outputs_safe = 0; /* for now */ p.have_sse2 = cpu_has_xmm2; p.identity = x86_make_reg(file_XMM, 6); p.chan0 = x86_make_reg(file_XMM, 7); -- cgit v1.2.3 From e08501b45763cf177f03fb34b737050d23ba4bc0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 19 Mar 2008 16:41:07 +0000 Subject: gallium: Add generic enum and flags dumping utility functions. --- src/gallium/auxiliary/util/p_debug.c | 59 ++++++++++++++++++++++++++++++++++++ src/gallium/include/pipe/p_debug.h | 49 ++++++++++++++++++++++++++++++ 2 files changed, 108 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index bb84e8096d..bd3a0221ea 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -52,6 +52,9 @@ rpl_EngDebugPrint(const char *format, ...) } int rpl_vsnprintf(char *, size_t, const char *, va_list); +int rpl_snprintf(char *str, size_t size, const char *format, ...); +#define vsnprintf rpl_vsnprintf +#define snprintf rpl_snprintf #endif @@ -181,3 +184,59 @@ void debug_mask_vprintf(uint32_t uuid, uint32_t what, const char *format, va_lis if(mask & what) debug_vprintf(format, ap); } + + +const char * +debug_dump_enum(const struct debug_named_value *names, + unsigned long value) +{ + static char rest[256]; + + while(names->name) { + if(names->value == value) + return names->name; + ++names; + } + + snprintf(rest, sizeof(rest), "0x%08x", value); + return rest; +} + + +const char * +debug_dump_flags(const struct debug_named_value *names, + unsigned long value) +{ + static char output[4096]; + static char rest[256]; + int first = 1; + + output[0] = '\0'; + + while(names->name) { + if((names->value & value) == names->value) { + if (!first) + strncat(output, "|", sizeof(output)); + else + first = 0; + strncat(output, names->name, sizeof(output)); + value &= ~names->value; + } + ++names; + } + + if (value) { + if (!first) + strncat(output, "|", sizeof(output)); + else + first = 0; + + snprintf(rest, sizeof(rest), "0x%08x", value); + strncat(output, rest, sizeof(output)); + } + + if(first) + return "0"; + + return output; +} diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index f3dfa06216..e924c1ef60 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -168,6 +168,55 @@ void debug_mask_vprintf(uint32_t uuid, #endif +/** + * Used by debug_dump_enum and debug_dump_flags to describe symbols. + */ +struct debug_named_value +{ + const char *name; + unsigned long value; +}; + + +/** + * Some C pre-processor magic to simplify creating named values. + * + * Example: + * @code + * static const debug_named_value my_names[] = { + * DEBUG_NAMED_VALUE(MY_ENUM_VALUE_X), + * DEBUG_NAMED_VALUE(MY_ENUM_VALUE_Y), + * DEBUG_NAMED_VALUE(MY_ENUM_VALUE_Z), + * DEBUG_NAMED_VALUE_END + * }; + * + * ... + * debug_printf("%s = %s\n", + * name, + * debug_dump_enum(my_names, my_value)); + * ... + * @endcode + */ +#define DEBUG_NAMED_VALUE(__symbol) {#__symbol, (unsigned long)__symbol} +#define DEBUG_NAMED_VALUE_END {NULL, 0} + + +/** + * Convert a enum value to a string. + */ +const char * +debug_dump_enum(const struct debug_named_value *names, + unsigned long value); + + +/** + * Convert binary flags value to a string. + */ +const char * +debug_dump_flags(const struct debug_named_value *names, + unsigned long value); + + #ifdef __cplusplus } #endif -- cgit v1.2.3 From 4b39ba72166c6468525b19797c3d8a058814d789 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Mar 2008 08:53:02 -0600 Subject: gallium: set tc->surface_map = NULL after unmapping --- src/gallium/drivers/softpipe/sp_tile_cache.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 4caf2dd3fc..19f71887e7 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -156,6 +156,7 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, if (tc->surface_map) { /*assert(tc->surface != ps);*/ pipe_surface_unmap(tc->surface); + tc->surface_map = NULL; } pipe_surface_reference(&tc->surface, ps); -- cgit v1.2.3 From 1213c7257335d577cf0217e34edafddf0451ba1b Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Mar 2008 10:43:35 -0600 Subject: gallium: remove old commented-out code --- src/mesa/state_tracker/st_draw.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 5ebd0bbcce..20af90df7d 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -223,7 +223,7 @@ st_draw_vbo(GLcontext *ctx, /* loop over TGSI shader inputs to determine vertex buffer * and attribute info */ - for (attr = 0; attr < /*vs*/vp->num_inputs; attr++) { + for (attr = 0; attr < vp->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; struct pipe_vertex_element velement; @@ -333,7 +333,7 @@ st_draw_vbo(GLcontext *ctx, } /* unreference buffers (frees wrapped user-space buffer objects) */ - for (attr = 0; attr < /*vs*/vp->num_inputs; attr++) { + for (attr = 0; attr < vp->num_inputs; attr++) { pipe_buffer_reference(winsys, &vbuffer[attr].buffer, NULL); assert(!vbuffer[attr].buffer); pipe->set_vertex_buffer(pipe, attr, &vbuffer[attr]); @@ -502,7 +502,7 @@ st_feedback_draw_vbo(GLcontext *ctx, /* loop over TGSI shader inputs to determine vertex buffer * and attribute info */ - for (attr = 0; attr < /*vs*/vp->num_inputs; attr++) { + for (attr = 0; attr < vp->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; struct pipe_vertex_element velement; -- cgit v1.2.3 From 7d95efde0a0e13e13c59444703bc47eb13926385 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 19 Mar 2008 11:12:48 -0600 Subject: gallium: implement CSO save/restore functions for use by meta operations (blit, gen-mipmaps, quad-clear, etc) Also, additional cso_set_*() functions for viewport, framebuffer, blend color, etc. state. --- src/gallium/auxiliary/cso_cache/cso_context.c | 202 +++++++++++++++++++++++--- src/gallium/auxiliary/cso_cache/cso_context.h | 53 +++++-- src/gallium/auxiliary/util/u_blit.c | 125 ++++++++-------- src/gallium/auxiliary/util/u_blit.h | 5 +- src/gallium/auxiliary/util/u_gen_mipmap.c | 116 ++++++++------- src/gallium/auxiliary/util/u_gen_mipmap.h | 6 +- src/mesa/state_tracker/st_atom_blend.c | 12 +- src/mesa/state_tracker/st_atom_framebuffer.c | 9 +- src/mesa/state_tracker/st_atom_viewport.c | 26 ++-- src/mesa/state_tracker/st_cb_blit.c | 17 +-- src/mesa/state_tracker/st_cb_clear.c | 35 ++--- src/mesa/state_tracker/st_cb_drawpixels.c | 25 ++-- src/mesa/state_tracker/st_context.h | 1 - src/mesa/state_tracker/st_gen_mipmap.c | 20 +-- 14 files changed, 408 insertions(+), 244 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 294ac82281..53d05ae6ce 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -52,11 +52,27 @@ struct cso_context { void *samplers[PIPE_MAX_SAMPLERS]; unsigned nr_samplers; - void *blend; - void *depth_stencil; - void *rasterizer; - void *fragment_shader; - void *vertex_shader; + void *samplers_saved[PIPE_MAX_SAMPLERS]; + unsigned nr_samplers_saved; + + struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; + uint nr_textures; + + struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS]; + uint nr_textures_saved; + + /** Current and saved state. + * The saved state is used as a 1-deep stack. + */ + void *blend, *blend_saved; + void *depth_stencil, *depth_stencil_saved; + void *rasterizer, *rasterizer_saved; + void *fragment_shader, *fragment_shader_saved; + void *vertex_shader, *vertex_shader_saved; + + struct pipe_framebuffer_state fb, fb_saved; + struct pipe_viewport_state vp, vp_saved; + struct pipe_blend_color blend_color; }; @@ -149,12 +165,23 @@ void cso_set_blend(struct cso_context *ctx, } } -void cso_unset_blend(struct cso_context *ctx) +void cso_save_blend(struct cso_context *ctx) +{ + assert(!ctx->blend_saved); + ctx->blend_saved = ctx->blend; +} + +void cso_restore_blend(struct cso_context *ctx) { - ctx->blend = NULL; + if (ctx->blend != ctx->blend_saved) { + ctx->blend = ctx->blend_saved; + ctx->pipe->bind_blend_state(ctx->pipe, ctx->blend_saved); + } + ctx->blend_saved = NULL; } + void cso_single_sampler(struct cso_context *ctx, unsigned idx, const struct pipe_sampler_state *templ) @@ -226,13 +253,48 @@ void cso_set_samplers( struct cso_context *ctx, cso_single_sampler_done( ctx ); } -void cso_unset_samplers( struct cso_context *ctx ) +void cso_save_samplers(struct cso_context *ctx) +{ + ctx->nr_samplers_saved = ctx->nr_samplers; + memcpy(ctx->samplers_saved, ctx->samplers, sizeof(ctx->samplers)); +} + +void cso_restore_samplers(struct cso_context *ctx) +{ + cso_set_samplers(ctx, ctx->nr_samplers_saved, + (const struct pipe_sampler_state **) ctx->samplers_saved); +} + + +void cso_set_sampler_textures( struct cso_context *ctx, + uint count, + struct pipe_texture **textures ) { uint i; - for (i = 0; i < ctx->nr_samplers; i++) - ctx->samplers[i] = NULL; + + ctx->nr_textures = count; + + for (i = 0; i < count; i++) + ctx->textures[i] = textures[i]; + for ( ; i < PIPE_MAX_SAMPLERS; i++) + ctx->textures[i] = NULL; + + ctx->pipe->set_sampler_textures(ctx->pipe, count, textures); } +void cso_save_sampler_textures( struct cso_context *ctx ) +{ + ctx->nr_textures_saved = ctx->nr_textures; + memcpy(ctx->textures_saved, ctx->textures, sizeof(ctx->textures)); +} + +void cso_restore_sampler_textures( struct cso_context *ctx ) +{ + cso_set_sampler_textures(ctx, ctx->nr_textures_saved, ctx->textures_saved); + ctx->nr_textures_saved = 0; +} + + void cso_set_depth_stencil_alpha(struct cso_context *ctx, @@ -267,15 +329,25 @@ void cso_set_depth_stencil_alpha(struct cso_context *ctx, } } -void cso_unset_depth_stencil_alpha(struct cso_context *ctx) +void cso_save_depth_stencil_alpha(struct cso_context *ctx) { - ctx->depth_stencil = NULL; + assert(!ctx->depth_stencil_saved); + ctx->depth_stencil_saved = ctx->depth_stencil; +} + +void cso_restore_depth_stencil_alpha(struct cso_context *ctx) +{ + if (ctx->depth_stencil != ctx->depth_stencil_saved) { + ctx->depth_stencil = ctx->depth_stencil_saved; + ctx->pipe->bind_depth_stencil_alpha_state(ctx->pipe, ctx->depth_stencil_saved); + } + ctx->depth_stencil_saved = NULL; } void cso_set_rasterizer(struct cso_context *ctx, - const struct pipe_rasterizer_state *templ) + const struct pipe_rasterizer_state *templ) { unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_rasterizer_state)); @@ -305,11 +377,20 @@ void cso_set_rasterizer(struct cso_context *ctx, } } -void cso_unset_rasterizer(struct cso_context *ctx) +void cso_save_rasterizer(struct cso_context *ctx) { - ctx->rasterizer = NULL; + assert(!ctx->rasterizer_saved); + ctx->rasterizer_saved = ctx->rasterizer; } +void cso_restore_rasterizer(struct cso_context *ctx) +{ + if (ctx->rasterizer != ctx->rasterizer_saved) { + ctx->rasterizer = ctx->rasterizer_saved; + ctx->pipe->bind_rasterizer_state(ctx->pipe, ctx->rasterizer_saved); + } + ctx->rasterizer_saved = NULL; +} void cso_set_fragment_shader(struct cso_context *ctx, @@ -343,11 +424,23 @@ void cso_set_fragment_shader(struct cso_context *ctx, } } -void cso_unset_fragment_shader(struct cso_context *ctx) +void cso_save_fragment_shader(struct cso_context *ctx) { - ctx->fragment_shader = NULL; + assert(!ctx->fragment_shader_saved); + ctx->fragment_shader_saved = ctx->fragment_shader; } +void cso_restore_fragment_shader(struct cso_context *ctx) +{ + assert(ctx->fragment_shader_saved); + if (ctx->fragment_shader_saved != ctx->fragment_shader) { + ctx->pipe->bind_fs_state(ctx->pipe, ctx->fragment_shader_saved); + ctx->fragment_shader = ctx->fragment_shader_saved; + } + ctx->fragment_shader_saved = NULL; +} + + void cso_set_vertex_shader(struct cso_context *ctx, const struct pipe_shader_state *templ) @@ -380,7 +473,78 @@ void cso_set_vertex_shader(struct cso_context *ctx, } } -void cso_unset_vertex_shader(struct cso_context *ctx) +void cso_save_vertex_shader(struct cso_context *ctx) { - ctx->vertex_shader = NULL; + assert(!ctx->vertex_shader_saved); + ctx->vertex_shader_saved = ctx->vertex_shader; +} + +void cso_restore_vertex_shader(struct cso_context *ctx) +{ + assert(ctx->vertex_shader_saved); + if (ctx->vertex_shader_saved != ctx->vertex_shader) { + ctx->pipe->bind_fs_state(ctx->pipe, ctx->vertex_shader_saved); + ctx->vertex_shader = ctx->vertex_shader_saved; + } + ctx->vertex_shader_saved = NULL; +} + + + +void cso_set_framebuffer(struct cso_context *ctx, + const struct pipe_framebuffer_state *fb) +{ + if (memcmp(&ctx->fb, fb, sizeof(*fb))) { + ctx->fb = *fb; + ctx->pipe->set_framebuffer_state(ctx->pipe, fb); + } +} + +void cso_save_framebuffer(struct cso_context *ctx) +{ + ctx->fb_saved = ctx->fb; +} + +void cso_restore_framebuffer(struct cso_context *ctx) +{ + if (memcmp(&ctx->fb, &ctx->fb_saved, sizeof(ctx->fb))) { + ctx->fb = ctx->fb_saved; + ctx->pipe->set_framebuffer_state(ctx->pipe, &ctx->fb); + } +} + + +void cso_set_viewport(struct cso_context *ctx, + const struct pipe_viewport_state *vp) +{ + if (memcmp(&ctx->vp, vp, sizeof(*vp))) { + ctx->vp = *vp; + ctx->pipe->set_viewport_state(ctx->pipe, vp); + } +} + +void cso_save_viewport(struct cso_context *ctx) +{ + ctx->vp_saved = ctx->vp; +} + + +void cso_restore_viewport(struct cso_context *ctx) +{ + if (memcmp(&ctx->vp, &ctx->vp_saved, sizeof(ctx->vp))) { + ctx->vp = ctx->vp_saved; + ctx->pipe->set_viewport_state(ctx->pipe, &ctx->vp); + } +} + + + + +void cso_set_blend_color(struct cso_context *ctx, + const struct pipe_blend_color *bc) +{ + if (memcmp(&ctx->blend_color, bc, sizeof(ctx->blend_color))) { + ctx->blend_color = *bc; + ctx->pipe->set_blend_color(ctx->pipe, bc); + } } diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h index 6aa619abf5..665e8d9911 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.h +++ b/src/gallium/auxiliary/cso_cache/cso_context.h @@ -41,27 +41,36 @@ struct cso_context; struct cso_context *cso_create_context( struct pipe_context *pipe ); +void cso_destroy_context( struct cso_context *cso ); + + + void cso_set_blend( struct cso_context *cso, const struct pipe_blend_state *blend ); +void cso_save_blend(struct cso_context *cso); +void cso_restore_blend(struct cso_context *cso); + -void cso_unset_blend(struct cso_context *cso); void cso_set_depth_stencil_alpha( struct cso_context *cso, const struct pipe_depth_stencil_alpha_state *dsa ); +void cso_save_depth_stencil_alpha(struct cso_context *cso); +void cso_restore_depth_stencil_alpha(struct cso_context *cso); + -void cso_unset_depth_stencil_alpha( struct cso_context *cso ); void cso_set_rasterizer( struct cso_context *cso, const struct pipe_rasterizer_state *rasterizer ); +void cso_save_rasterizer(struct cso_context *cso); +void cso_restore_rasterizer(struct cso_context *cso); + -void cso_unset_rasterizer( struct cso_context *cso ); void cso_set_samplers( struct cso_context *cso, unsigned count, const struct pipe_sampler_state **states ); - -void cso_unset_samplers( struct cso_context *cso ); - +void cso_save_samplers(struct cso_context *cso); +void cso_restore_samplers(struct cso_context *cso); /* Alternate interface to support state trackers that like to modify * samplers one at a time: @@ -73,6 +82,15 @@ void cso_single_sampler( struct cso_context *cso, void cso_single_sampler_done( struct cso_context *cso ); + +void cso_set_sampler_textures( struct cso_context *cso, + uint count, + struct pipe_texture **textures ); +void cso_save_sampler_textures( struct cso_context *cso ); +void cso_restore_sampler_textures( struct cso_context *cso ); + + + /* These aren't really sensible -- most of the time the api provides * object semantics for shaders anyway, and the cases where it doesn't * (eg mesa's internall-generated texenv programs), it will be up to @@ -80,15 +98,32 @@ void cso_single_sampler_done( struct cso_context *cso ); */ void cso_set_fragment_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); +void cso_save_fragment_shader(struct cso_context *cso); +void cso_restore_fragment_shader(struct cso_context *cso); + -void cso_unset_fragment_shader( struct cso_context *cso ); void cso_set_vertex_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); +void cso_save_vertex_shader(struct cso_context *cso); +void cso_restore_vertex_shader(struct cso_context *cso); -void cso_unset_vertex_shader( struct cso_context *cso ); -void cso_destroy_context( struct cso_context *cso ); + +void cso_set_framebuffer(struct cso_context *cso, + const struct pipe_framebuffer_state *fb); +void cso_save_framebuffer(struct cso_context *cso); +void cso_restore_framebuffer(struct cso_context *cso); + + +void cso_set_viewport(struct cso_context *cso, + const struct pipe_viewport_state *vp); +void cso_save_viewport(struct cso_context *cso); +void cso_restore_viewport(struct cso_context *cso); + + +void cso_set_blend_color(struct cso_context *cso, + const struct pipe_blend_color *bc); #ifdef __cplusplus diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 4b4ab8185f..123304fe68 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -45,15 +45,18 @@ #include "util/u_blit.h" #include "util/u_simple_shaders.h" +#include "cso_cache/cso_context.h" + struct blit_state { struct pipe_context *pipe; + struct cso_context *cso; - void *blend; - void *depthstencil; - void *rasterizer; - void *samplers[2]; /* one for linear, one for nearest sampling */ + struct pipe_blend_state blend; + struct pipe_depth_stencil_alpha_state depthstencil; + struct pipe_rasterizer_state rasterizer; + struct pipe_sampler_state sampler; /*struct pipe_viewport_state viewport;*/ struct pipe_sampler_state *vs; @@ -66,56 +69,44 @@ struct blit_state * Intended to be created once and re-used for many blit() calls. */ struct blit_state * -util_create_blit(struct pipe_context *pipe) +util_create_blit(struct pipe_context *pipe, struct cso_context *cso) { - struct pipe_blend_state blend; - struct pipe_depth_stencil_alpha_state depthstencil; - struct pipe_rasterizer_state rasterizer; struct blit_state *ctx; - struct pipe_sampler_state sampler; ctx = CALLOC_STRUCT(blit_state); if (!ctx) return NULL; ctx->pipe = pipe; + ctx->cso = cso; - /* we don't use blending, but need to set valid values */ - memset(&blend, 0, sizeof(blend)); - blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; - blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; - blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; - blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; - blend.colormask = PIPE_MASK_RGBA; - ctx->blend = pipe->create_blend_state(pipe, &blend); + /* disabled blending/masking */ + memset(&ctx->blend, 0, sizeof(ctx->blend)); + ctx->blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + ctx->blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + ctx->blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + ctx->blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + ctx->blend.colormask = PIPE_MASK_RGBA; - /* depth/stencil/alpha */ - memset(&depthstencil, 0, sizeof(depthstencil)); - ctx->depthstencil = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil); + /* no-op depth/stencil/alpha */ + memset(&ctx->depthstencil, 0, sizeof(ctx->depthstencil)); /* rasterizer */ - memset(&rasterizer, 0, sizeof(rasterizer)); - rasterizer.front_winding = PIPE_WINDING_CW; - rasterizer.cull_mode = PIPE_WINDING_NONE; - rasterizer.bypass_clipping = 1; /* bypasses viewport too */ - /*rasterizer.bypass_vs = 1;*/ - ctx->rasterizer = pipe->create_rasterizer_state(pipe, &rasterizer); + memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer)); + ctx->rasterizer.front_winding = PIPE_WINDING_CW; + ctx->rasterizer.cull_mode = PIPE_WINDING_NONE; + ctx->rasterizer.bypass_clipping = 1; /* bypasses viewport too */ + /*ctx->rasterizer.bypass_vs = 1;*/ /* samplers */ - memset(&sampler, 0, sizeof(sampler)); - sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; - sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; - sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; - sampler.normalized_coords = 1; - ctx->samplers[0] = pipe->create_sampler_state(pipe, &sampler); - - sampler.min_img_filter = PIPE_TEX_MIPFILTER_LINEAR; - sampler.mag_img_filter = PIPE_TEX_MIPFILTER_LINEAR; - ctx->samplers[1] = pipe->create_sampler_state(pipe, &sampler); - + memset(&ctx->sampler, 0, sizeof(ctx->sampler)); + ctx->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + ctx->sampler.min_img_filter = 0; /* set later */ + ctx->sampler.mag_img_filter = 0; /* set later */ + ctx->sampler.normalized_coords = 1; #if 0 /* viewport */ @@ -153,12 +144,6 @@ util_destroy_blit(struct blit_state *ctx) { struct pipe_context *pipe = ctx->pipe; - pipe->delete_blend_state(pipe, ctx->blend); - pipe->delete_depth_stencil_alpha_state(pipe, ctx->depthstencil); - pipe->delete_rasterizer_state(pipe, ctx->rasterizer); - pipe->delete_sampler_state(pipe, ctx->samplers[0]); - pipe->delete_sampler_state(pipe, ctx->samplers[1]); - pipe->delete_vs_state(pipe, ctx->vs); pipe->delete_fs_state(pipe, ctx->fs); @@ -236,17 +221,24 @@ util_blit_pixels(struct blit_state *ctx, src, srcLeft, srcTop, /* src */ srcW, srcH); /* size */ - /* drawing dest */ - memset(&fb, 0, sizeof(fb)); - fb.num_cbufs = 1; - fb.cbufs[0] = dst; - pipe->set_framebuffer_state(pipe, &fb); + /* save state (restored below) */ + cso_save_blend(ctx->cso); + cso_save_depth_stencil_alpha(ctx->cso); + cso_save_rasterizer(ctx->cso); + cso_save_samplers(ctx->cso); + cso_save_sampler_textures(ctx->cso); + cso_save_framebuffer(ctx->cso); + + /* set misc state we care about */ + cso_set_blend(ctx->cso, &ctx->blend); + cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); + cso_set_rasterizer(ctx->cso, &ctx->rasterizer); /* sampler */ - if (filter == PIPE_TEX_MIPFILTER_NEAREST) - pipe->bind_sampler_states(pipe, 1, &ctx->samplers[0]); - else - pipe->bind_sampler_states(pipe, 1, &ctx->samplers[1]); + ctx->sampler.min_img_filter = filter; + ctx->sampler.mag_img_filter = filter; + cso_single_sampler(ctx->cso, 0, &ctx->sampler); + cso_single_sampler_done(ctx->cso); /* texture */ pipe->set_sampler_textures(pipe, 1, &tex); @@ -255,22 +247,25 @@ util_blit_pixels(struct blit_state *ctx, pipe->bind_fs_state(pipe, ctx->fs); pipe->bind_vs_state(pipe, ctx->vs); - /* misc state */ - pipe->bind_blend_state(pipe, ctx->blend); - pipe->bind_depth_stencil_alpha_state(pipe, ctx->depthstencil); - pipe->bind_rasterizer_state(pipe, ctx->rasterizer); + /* drawing dest */ + memset(&fb, 0, sizeof(fb)); + fb.num_cbufs = 1; + fb.cbufs[0] = dst; + cso_set_framebuffer(ctx->cso, &fb); /* draw quad */ util_draw_texquad(pipe, dstX0, dstY0, dstX1, dstY1, z); - /* unbind */ - pipe->set_sampler_textures(pipe, 0, NULL); - pipe->bind_sampler_states(pipe, 0, NULL); + /* restore state we changed */ + cso_restore_blend(ctx->cso); + cso_restore_depth_stencil_alpha(ctx->cso); + cso_restore_rasterizer(ctx->cso); + cso_restore_samplers(ctx->cso); + cso_restore_sampler_textures(ctx->cso); + cso_restore_framebuffer(ctx->cso); - /* free stuff */ + /* free the texture */ pipe_surface_reference(&texSurf, NULL); screen->texture_release(screen, &tex); - - /* Note: caller must restore pipe/gallium state at this time */ } diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h index a349be99ad..61f1d9bb32 100644 --- a/src/gallium/auxiliary/util/u_blit.h +++ b/src/gallium/auxiliary/util/u_blit.h @@ -30,15 +30,16 @@ #define U_BLIT_H + struct pipe_context; struct pipe_surface; +struct cso_context; struct blit_state; - extern struct blit_state * -util_create_blit(struct pipe_context *pipe); +util_create_blit(struct pipe_context *pipe, struct cso_context *cso); extern void diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index e18f8ab72a..27141c4d13 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -49,14 +49,18 @@ #include "tgsi/util/tgsi_dump.h" #include "tgsi/util/tgsi_parse.h" +#include "cso_cache/cso_context.h" + struct gen_mipmap_state { struct pipe_context *pipe; + struct cso_context *cso; - void *blend; - void *depthstencil; - void *rasterizer; + struct pipe_blend_state blend; + struct pipe_depth_stencil_alpha_state depthstencil; + struct pipe_rasterizer_state rasterizer; + struct pipe_sampler_state sampler; /*struct pipe_viewport_state viewport;*/ struct pipe_sampler_state *vs; struct pipe_sampler_state *fs; @@ -675,11 +679,9 @@ fallback_gen_mipmap(struct gen_mipmap_state *ctx, * generate a mipmap. */ struct gen_mipmap_state * -util_create_gen_mipmap(struct pipe_context *pipe) +util_create_gen_mipmap(struct pipe_context *pipe, + struct cso_context *cso) { - struct pipe_blend_state blend; - struct pipe_depth_stencil_alpha_state depthstencil; - struct pipe_rasterizer_state rasterizer; struct gen_mipmap_state *ctx; ctx = CALLOC_STRUCT(gen_mipmap_state); @@ -687,27 +689,36 @@ util_create_gen_mipmap(struct pipe_context *pipe) return NULL; ctx->pipe = pipe; + ctx->cso = cso; - /* we don't use blending, but need to set valid values */ - memset(&blend, 0, sizeof(blend)); - blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; - blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; - blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; - blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; - blend.colormask = PIPE_MASK_RGBA; - ctx->blend = pipe->create_blend_state(pipe, &blend); + /* disabled blending/masking */ + memset(&ctx->blend, 0, sizeof(ctx->blend)); + ctx->blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + ctx->blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + ctx->blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + ctx->blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + ctx->blend.colormask = PIPE_MASK_RGBA; - /* depth/stencil/alpha */ - memset(&depthstencil, 0, sizeof(depthstencil)); - ctx->depthstencil = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil); + /* no-op depth/stencil/alpha */ + memset(&ctx->depthstencil, 0, sizeof(ctx->depthstencil)); /* rasterizer */ - memset(&rasterizer, 0, sizeof(rasterizer)); - rasterizer.front_winding = PIPE_WINDING_CW; - rasterizer.cull_mode = PIPE_WINDING_NONE; - rasterizer.bypass_clipping = 1; /* bypasses viewport too */ - //rasterizer.bypass_vs = 1; - ctx->rasterizer = pipe->create_rasterizer_state(pipe, &rasterizer); + memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer)); + ctx->rasterizer.front_winding = PIPE_WINDING_CW; + ctx->rasterizer.cull_mode = PIPE_WINDING_NONE; + ctx->rasterizer.bypass_clipping = 1; /* bypasses viewport too */ + /*ctx->rasterizer.bypass_vs = 1;*/ + + /* sampler state */ + memset(&ctx->sampler, 0, sizeof(ctx->sampler)); + ctx->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; + ctx->sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; + ctx->sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; + ctx->sampler.normalized_coords = 1; + #if 0 /* viewport */ @@ -745,9 +756,6 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx) { struct pipe_context *pipe = ctx->pipe; - pipe->delete_blend_state(pipe, ctx->blend); - pipe->delete_depth_stencil_alpha_state(pipe, ctx->depthstencil); - pipe->delete_rasterizer_state(pipe, ctx->rasterizer); pipe->delete_vs_state(pipe, ctx->vs); pipe->delete_fs_state(pipe, ctx->fs); @@ -792,8 +800,6 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, struct pipe_context *pipe = ctx->pipe; struct pipe_screen *screen = pipe->screen; struct pipe_framebuffer_state fb; - struct pipe_sampler_state sampler; - void *sampler_cso; uint dstLevel; uint zslice = 0; @@ -803,30 +809,29 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, return; } - /* init framebuffer state */ - memset(&fb, 0, sizeof(fb)); - fb.num_cbufs = 1; - - /* sampler state */ - memset(&sampler, 0, sizeof(sampler)); - sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; - sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; - sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; - sampler.normalized_coords = 1; + /* save state (restored below) */ + cso_save_blend(ctx->cso); + cso_save_depth_stencil_alpha(ctx->cso); + cso_save_rasterizer(ctx->cso); + cso_save_samplers(ctx->cso); + cso_save_sampler_textures(ctx->cso); + cso_save_framebuffer(ctx->cso); /* bind our state */ - pipe->bind_blend_state(pipe, ctx->blend); - pipe->bind_depth_stencil_alpha_state(pipe, ctx->depthstencil); - pipe->bind_rasterizer_state(pipe, ctx->rasterizer); + cso_set_blend(ctx->cso, &ctx->blend); + cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); + cso_set_rasterizer(ctx->cso, &ctx->rasterizer); + pipe->bind_vs_state(pipe, ctx->vs); pipe->bind_fs_state(pipe, ctx->fs); #if 0 pipe->set_viewport_state(pipe, &ctx->viewport); #endif + /* init framebuffer state */ + memset(&fb, 0, sizeof(fb)); + fb.num_cbufs = 1; + /* * XXX for small mipmap levels, it may be faster to use the software * fallback path... @@ -838,7 +843,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, * Setup framebuffer / dest surface */ fb.cbufs[0] = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); - pipe->set_framebuffer_state(pipe, &fb); + cso_set_framebuffer(ctx->cso, &fb); /* * Setup sampler state @@ -847,11 +852,10 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, * has trouble with min clamping so we also set the lod_bias to * try to work around that. */ - sampler.min_lod = sampler.max_lod = (float) srcLevel; - sampler.lod_bias = (float) srcLevel; - sampler_cso = pipe->create_sampler_state(pipe, &sampler); - pipe->bind_sampler_states(pipe, 1, &sampler_cso); - + ctx->sampler.min_lod = ctx->sampler.max_lod = (float) srcLevel; + ctx->sampler.lod_bias = (float) srcLevel; + cso_single_sampler(ctx->cso, 0, &ctx->sampler); + cso_single_sampler_done(ctx->cso); #if 0 simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]); #endif @@ -869,9 +873,13 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, pipe->flush(pipe, PIPE_FLUSH_WAIT); /*pipe->texture_update(pipe, pt); not really needed */ - - pipe->delete_sampler_state(pipe, sampler_cso); } - /* Note: caller must restore pipe/gallium state at this time */ + /* restore state we changed */ + cso_restore_blend(ctx->cso); + cso_restore_depth_stencil_alpha(ctx->cso); + cso_restore_rasterizer(ctx->cso); + cso_restore_samplers(ctx->cso); + cso_restore_sampler_textures(ctx->cso); + cso_restore_framebuffer(ctx->cso); } diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.h b/src/gallium/auxiliary/util/u_gen_mipmap.h index 80496140a2..eeabf3bf07 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.h +++ b/src/gallium/auxiliary/util/u_gen_mipmap.h @@ -31,11 +31,15 @@ #include "pipe/p_state.h" +struct pipe_context; +struct pipe_texture; +struct cso_context; + struct gen_mipmap_state; extern struct gen_mipmap_state * -util_create_gen_mipmap(struct pipe_context *pipe); +util_create_gen_mipmap(struct pipe_context *pipe, struct cso_context *cso); extern void diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c index 6c13fc8141..2a0e92245c 100644 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -39,6 +39,7 @@ #include "pipe/p_defines.h" #include "cso_cache/cso_context.h" +#include "main/macros.h" /** * Convert GLenum blend tokens to pipe tokens. @@ -213,13 +214,10 @@ update_blend( struct st_context *st ) cso_set_blend(st->cso_context, blend); - if (memcmp(st->ctx->Color.BlendColor, &st->state.blend_color, 4 * sizeof(GLfloat)) != 0) { - /* state has changed */ - st->state.blend_color.color[0] = st->ctx->Color.BlendColor[0]; - st->state.blend_color.color[1] = st->ctx->Color.BlendColor[1]; - st->state.blend_color.color[2] = st->ctx->Color.BlendColor[2]; - st->state.blend_color.color[3] = st->ctx->Color.BlendColor[3]; - st->pipe->set_blend_color(st->pipe, (struct pipe_blend_color *) st->ctx->Color.BlendColor); + { + struct pipe_blend_color bc; + COPY_4FV(bc.color, st->ctx->Color.BlendColor); + cso_set_blend_color(st->cso_context, &bc); } } diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 3e58d49f1f..c8fa0cbdfb 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -35,6 +35,7 @@ #include "st_atom.h" #include "st_cb_fbo.h" #include "pipe/p_context.h" +#include "cso_cache/cso_context.h" /** @@ -76,13 +77,7 @@ update_framebuffer_state( struct st_context *st ) } } - /* XXX: The memcmp is insufficient for eliminating redundant state changes, - * but we should probably do more work here to that end. - */ - if (1 /*memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0*/) { - st->state.framebuffer = framebuffer; - st->pipe->set_framebuffer_state( st->pipe, &framebuffer ); - } + cso_set_framebuffer(st->cso_context, &framebuffer); } diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index 147aa3c51a..eb3f62cfbe 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -29,9 +29,9 @@ #include "context.h" #include "colormac.h" #include "st_context.h" -#include "pipe/p_context.h" #include "st_atom.h" - +#include "pipe/p_context.h" +#include "cso_cache/cso_context.h" /** * Update the viewport transformation matrix. Depends on: @@ -65,22 +65,18 @@ update_viewport( struct st_context *st ) GLfloat half_width = ctx->Viewport.Width / 2.0; GLfloat half_height = ctx->Viewport.Height / 2.0; GLfloat half_depth = (ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; - struct pipe_viewport_state vp; - vp.scale[0] = half_width; - vp.scale[1] = half_height * yScale; - vp.scale[2] = half_depth; - vp.scale[3] = 1.0; + st->state.viewport.scale[0] = half_width; + st->state.viewport.scale[1] = half_height * yScale; + st->state.viewport.scale[2] = half_depth; + st->state.viewport.scale[3] = 1.0; - vp.translate[0] = half_width + x; - vp.translate[1] = (half_height + y) * yScale + yBias; - vp.translate[2] = half_depth + z; - vp.translate[3] = 0.0; + st->state.viewport.translate[0] = half_width + x; + st->state.viewport.translate[1] = (half_height + y) * yScale + yBias; + st->state.viewport.translate[2] = half_depth + z; + st->state.viewport.translate[3] = 0.0; - if (memcmp(&vp, &st->state.viewport, sizeof(vp)) != 0) { - st->state.viewport = vp; - st->pipe->set_viewport_state(st->pipe, &vp); - } + cso_set_viewport(st->cso_context, &st->state.viewport); } } diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c index dfa79c975c..64314a5078 100644 --- a/src/mesa/state_tracker/st_cb_blit.c +++ b/src/mesa/state_tracker/st_cb_blit.c @@ -52,7 +52,7 @@ void st_init_blit(struct st_context *st) { - st->blit = util_create_blit(st->pipe); + st->blit = util_create_blit(st->pipe, st->cso_context); } @@ -98,22 +98,9 @@ st_BlitFramebuffer(GLcontext *ctx, } -#if 0 - /* XXX is this sufficient? */ - st_invalidate_state(ctx, _NEW_COLOR | _NEW_TEXTURE); -#else - /* need to "unset" cso state because we went behind the back of the cso - * tracker. Without unset, the _set_ calls would be no-ops. - */ - cso_unset_blend(st->cso_context); - cso_unset_depth_stencil_alpha(st->cso_context); - cso_unset_rasterizer(st->cso_context); - cso_set_blend(st->cso_context, &st->state.blend); - cso_set_depth_stencil_alpha(st->cso_context, &st->state.depth_stencil); - cso_set_rasterizer(st->cso_context, &st->state.rasterizer); + /* shaders don't go through CSO yet */ pipe->bind_fs_state(pipe, st->fp->driver_shader); pipe->bind_vs_state(pipe, st->vp->driver_shader); -#endif } diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 693cddedf7..c23938dc68 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -247,10 +247,19 @@ clear_with_quad(GLcontext *ctx, x1, y1); */ + cso_save_blend(st->cso_context); + cso_save_depth_stencil_alpha(st->cso_context); + cso_save_rasterizer(st->cso_context); + cso_save_viewport(st->cso_context); + /* blend state: RGBA masking */ { struct pipe_blend_state blend; memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; if (color) { if (ctx->Color.ColorMask[0]) blend.colormask |= PIPE_MASK_R; @@ -294,13 +303,6 @@ clear_with_quad(GLcontext *ctx, { struct pipe_rasterizer_state raster; memset(&raster, 0, sizeof(raster)); -#if 0 - /* don't do per-pixel scissor; we'll just draw a PIPE_PRIM_QUAD - * that matches the scissor bounds. - */ - if (ctx->Scissor.Enabled) - raster.scissor = 1; -#endif #if TEST_DRAW_PASSTHROUGH raster.bypass_clipping = 1; raster.bypass_vs = 1; @@ -342,28 +344,21 @@ clear_with_quad(GLcontext *ctx, vp.translate[1] = 0.5 * height; vp.translate[2] = 0.0; vp.translate[3] = 0.0; - pipe->set_viewport_state(pipe, &vp); + cso_set_viewport(st->cso_context, &vp); } #endif /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); -#if 0 - /* Can't depend on old state objects still existing -- may have - * been deleted to make room in the hash, etc. (Should get - * fixed...) - */ - st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL); -#else /* Restore pipe state */ - cso_set_blend(st->cso_context, &st->state.blend); - cso_set_depth_stencil_alpha(st->cso_context, &st->state.depth_stencil); - cso_set_rasterizer(st->cso_context, &st->state.rasterizer); + cso_restore_blend(st->cso_context); + cso_restore_depth_stencil_alpha(st->cso_context); + cso_restore_rasterizer(st->cso_context); + cso_restore_viewport(st->cso_context); + /* these don't go through cso yet */ pipe->bind_fs_state(pipe, st->fp->driver_shader); pipe->bind_vs_state(pipe, st->vp->driver_shader); -#endif - pipe->set_viewport_state(pipe, &st->state.viewport); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 18ec9645c4..33d34445ee 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -650,6 +650,9 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, assert(width <= maxSize); assert(height <= maxSize); + cso_save_rasterizer(cso); + cso_save_viewport(cso); + /* setup state: just scissor */ { struct pipe_rasterizer_state setup; @@ -696,7 +699,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, vp.translate[1] = 0.5 * height; vp.translate[2] = 0.0; vp.translate[3] = 0.0; - pipe->set_viewport_state(pipe, &vp); + cso_set_viewport(cso, &vp); } /* texture state: */ @@ -719,26 +722,18 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, else draw_quad(ctx, x0, y0, z, x1, y1, invertTex); - /* restore GL state */ - pipe->set_sampler_textures(pipe, ctx->st->state.num_textures, - ctx->st->state.sampler_texture); - - pipe->set_viewport_state(pipe, &ctx->st->state.viewport); - -#if 0 - /* Can't depend on old state objects still existing -- may have - * been deleted to make room in the hash, etc. (Should get - * fixed...) - */ - st_invalidate_state(ctx, _NEW_COLOR | _NEW_TEXTURE); -#else /* restore state */ + cso_restore_rasterizer(cso); + cso_restore_viewport(cso); + /* shaders don't go through cso yet */ pipe->bind_fs_state(pipe, st->fp->driver_shader); pipe->bind_vs_state(pipe, st->vp->driver_shader); + cso_set_rasterizer(cso, &st->state.rasterizer); cso_set_samplers(cso, PIPE_MAX_SAMPLERS, (const struct pipe_sampler_state **) st->state.sampler_list); -#endif + pipe->set_sampler_textures(pipe, ctx->st->state.num_textures, + ctx->st->state.sampler_texture); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 63150dbeaf..ca8307c4ba 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -82,7 +82,6 @@ struct st_context struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state samplers[PIPE_MAX_SAMPLERS]; struct pipe_sampler_state *sampler_list[PIPE_MAX_SAMPLERS]; - struct pipe_blend_color blend_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; struct pipe_framebuffer_state framebuffer; diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 6c3afca1ba..61e1d9621c 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -59,7 +59,7 @@ void st_init_generate_mipmap(struct st_context *st) { - st->gen_mipmap = util_create_gen_mipmap(st->pipe); + st->gen_mipmap = util_create_gen_mipmap(st->pipe, st->cso_context); } @@ -94,19 +94,11 @@ st_render_mipmap(struct st_context *st, util_gen_mipmap(st->gen_mipmap, pt, face, baseLevel, lastLevel); - /* restore pipe state */ -#if 0 - cso_set_rasterizer(st->cso_context, &st->state.rasterizer); - cso_set_samplers(st->cso_context, st->state.samplers_list); - pipe->bind_fs_state(pipe, st->fp->shader_program); - pipe->bind_vs_state(pipe, st->vp->shader_program); - pipe->set_sampler_textures(pipe, st->state.num_textures, - st->state.sampler_texture); - pipe->set_viewport_state(pipe, &st->state.viewport); -#else - /* XXX is this sufficient? */ - st_invalidate_state(st->ctx, _NEW_COLOR | _NEW_TEXTURE); -#endif + /* shaders don't go through CSO yet */ + if (st->fp) + pipe->bind_fs_state(pipe, st->fp->driver_shader); + if (st->vp) + pipe->bind_vs_state(pipe, st->vp->driver_shader); return TRUE; } -- cgit v1.2.3 From b4f03d0c98674c83d06edfa767a7898eca5d4ef8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 19 Mar 2008 17:35:42 +0000 Subject: gallium: explict float casts --- src/gallium/auxiliary/util/u_blit.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 123304fe68..d05dae3af8 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -254,7 +254,11 @@ util_blit_pixels(struct blit_state *ctx, cso_set_framebuffer(ctx->cso, &fb); /* draw quad */ - util_draw_texquad(pipe, dstX0, dstY0, dstX1, dstY1, z); + util_draw_texquad(pipe, + (float)dstX0, + (float)dstY0, + (float)dstX1, + (float)dstY1, z); /* restore state we changed */ cso_restore_blend(ctx->cso); -- cgit v1.2.3 From df5ba799fa929d4c739be9d11d3f1000afc265b2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 19 Mar 2008 20:46:08 +0000 Subject: gallium: Fix broken logic. --- src/gallium/auxiliary/util/u_handle_table.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index 0bfb9e1b4a..8b0f7fca4b 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -248,7 +248,7 @@ handle_table_get_next_handle(struct handle_table *ht, unsigned index; for(index = handle; index < ht->size; ++index) { - if(!ht->objects[index]) + if(ht->objects[index]) return index + 1; } -- cgit v1.2.3 From b45669283fe4b9af9f2e78ac3c0c84207cf63775 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 19 Mar 2008 16:41:54 -0600 Subject: gallium: fix bug in cso_single_sampler_done() in computation of nr_samplers Need to find highest used sampler so search from end toward beginning. --- src/gallium/auxiliary/cso_cache/cso_context.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 53d05ae6ce..01fe216447 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -217,9 +217,11 @@ void cso_single_sampler_done( struct cso_context *ctx ) { unsigned i; - for (i = 0; i < 8; i++) - if (ctx->samplers[i] == NULL) + /* find highest non-null sampler */ + for (i = PIPE_MAX_SAMPLERS; i > 0; i--) { + if (ctx->samplers[i - 1] != NULL) break; + } ctx->nr_samplers = i; -- cgit v1.2.3 From 11c34dc644fe58b8178ab9142929a6685e3c0848 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 19 Mar 2008 17:04:49 -0600 Subject: gallium: need to set/save framebuffer state in st_context since we use it elsewhere --- src/mesa/state_tracker/st_atom_framebuffer.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index c8fa0cbdfb..43259c3ecb 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -45,39 +45,39 @@ static void update_framebuffer_state( struct st_context *st ) { - struct pipe_framebuffer_state framebuffer; + struct pipe_framebuffer_state *framebuffer = &st->state.framebuffer; struct gl_framebuffer *fb = st->ctx->DrawBuffer; struct st_renderbuffer *strb; GLuint i; - memset(&framebuffer, 0, sizeof(framebuffer)); + memset(framebuffer, 0, sizeof(*framebuffer)); /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state * to determine which surfaces to draw to */ - framebuffer.num_cbufs = fb->_NumColorDrawBuffers[0]; - for (i = 0; i < framebuffer.num_cbufs; i++) { + framebuffer->num_cbufs = fb->_NumColorDrawBuffers[0]; + for (i = 0; i < framebuffer->num_cbufs; i++) { strb = st_renderbuffer(fb->_ColorDrawBuffers[0][i]); assert(strb->surface); - framebuffer.cbufs[i] = strb->surface; + framebuffer->cbufs[i] = strb->surface; } strb = st_renderbuffer(fb->Attachment[BUFFER_DEPTH].Renderbuffer); if (strb) { strb = st_renderbuffer(strb->Base.Wrapped); assert(strb->surface); - framebuffer.zsbuf = strb->surface; + framebuffer->zsbuf = strb->surface; } else { strb = st_renderbuffer(fb->Attachment[BUFFER_STENCIL].Renderbuffer); if (strb) { strb = st_renderbuffer(strb->Base.Wrapped); assert(strb->surface); - framebuffer.zsbuf = strb->surface; + framebuffer->zsbuf = strb->surface; } } - cso_set_framebuffer(st->cso_context, &framebuffer); + cso_set_framebuffer(st->cso_context, framebuffer); } -- cgit v1.2.3 From 4984487bc3338fc351a0631eaa4515e4adbb86a9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 19 Mar 2008 17:08:16 -0600 Subject: gallium: add face, dirtyLevels params to pipe->texture_update() This provides better information about which images in texture object have changed. Also, call texture_update() from more places previously missed. --- src/gallium/auxiliary/draw/draw_aaline.c | 1 + src/gallium/auxiliary/draw/draw_pstipple.c | 2 +- src/gallium/auxiliary/util/u_gen_mipmap.c | 3 ++- src/gallium/drivers/cell/ppu/cell_texture.c | 3 ++- src/gallium/drivers/i915simple/i915_texture.c | 3 ++- src/gallium/drivers/i965simple/brw_tex_layout.c | 3 ++- src/gallium/drivers/softpipe/sp_texture.c | 3 ++- src/gallium/include/pipe/p_context.h | 6 ++---- src/mesa/state_tracker/st_atom_texture.c | 5 ----- src/mesa/state_tracker/st_cb_drawpixels.c | 3 +++ src/mesa/state_tracker/st_cb_texture.c | 13 +++++++------ src/mesa/state_tracker/st_texture.h | 2 -- 12 files changed, 24 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index f2b983374e..b4fa6bd967 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -410,6 +410,7 @@ aaline_create_texture(struct aaline_stage *aaline) /* unmap */ pipe_surface_unmap(surface); pipe_surface_reference(&surface, NULL); + pipe->texture_update(pipe, aaline->texture, 0, (1 << level)); } } diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index 09d542002f..9d154a6838 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -374,7 +374,7 @@ pstip_update_texture(struct pstip_stage *pstip) /* unmap */ pipe_surface_unmap(surface); pipe_surface_reference(&surface, NULL); - pipe->texture_update(pipe, pstip->texture); + pipe->texture_update(pipe, pstip->texture, 0, 0x1); } diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 27141c4d13..028b180a77 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -872,7 +872,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, pipe->flush(pipe, PIPE_FLUSH_WAIT); - /*pipe->texture_update(pipe, pt); not really needed */ + /* need to signal that the texture has changed _after_ rendering to it */ + pipe->texture_update(pipe, pt, face, (1 << dstLevel)); } /* restore state we changed */ diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index e235421107..9c694e136d 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -134,7 +134,8 @@ cell_texture_release_screen(struct pipe_screen *screen, static void -cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) +cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture, + uint face, uint levelsMask) { /* XXX TO DO: re-tile the texture data ... */ diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index ef5adff550..c39e747705 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -542,7 +542,8 @@ i915_texture_release_screen(struct pipe_screen *screen, static void -i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) +i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture, + uint face, uint levelsMask) { /* no-op? */ } diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index b24ac87c37..b580f98204 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -358,7 +358,8 @@ brw_texture_release_screen(struct pipe_screen *screen, static void -brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture) +brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture, + uint face, uint levelsMask) { /* no-op? */ } diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 64bf353aa4..a98b3b1a4a 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -174,7 +174,8 @@ softpipe_get_tex_surface_screen(struct pipe_screen *screen, static void softpipe_texture_update(struct pipe_context *pipe, - struct pipe_texture *texture) + struct pipe_texture *texture, + uint face, uint levelsMask) { struct softpipe_context *softpipe = softpipe_context(pipe); uint unit; diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index b64948f0f3..a3824601be 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -192,12 +192,10 @@ struct pipe_context { /** * Called when texture data is changed. - * Note: we could pass some hints about which mip levels or cube faces - * have changed... - * XXX this may go away - could pass a 'write' flag to get_tex_surface() */ void (*texture_update)(struct pipe_context *pipe, - struct pipe_texture *texture); + struct pipe_texture *texture, + uint face, uint dirtyLevelsMask); diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index e53a897637..2a711e513d 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -76,11 +76,6 @@ update_textures(struct st_context *st) pt = st_get_stobj_texture(stObj); pipe_texture_reference(&st->state.sampler_texture[unit], pt); - - if (stObj && stObj->dirtyData) { - st->pipe->texture_update(st->pipe, pt); - stObj->dirtyData = GL_FALSE; - } } st->pipe->set_sampler_textures(st->pipe, st->state.num_textures, diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 33d34445ee..99d5e3e848 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -513,6 +513,8 @@ make_texture(struct st_context *st, /* unmap */ pipe_surface_unmap(surface); pipe_surface_reference(&surface, NULL); + pipe->texture_update(pipe, pt, 0, 0x1); + assert(success); /* restore */ @@ -1100,6 +1102,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, /* Release surface */ pipe_surface_unmap(surface); pipe_surface_reference(&surface, NULL); + pipe->texture_update(pipe, pt, 0, 0x1); pt->format = format; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index d4731c7737..306b27c423 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -476,6 +476,7 @@ st_TexImage(GLcontext * ctx, struct gl_texture_image *texImage, GLsizei imageSize, int compressed) { + struct pipe_context *pipe = ctx->st->pipe; struct st_texture_object *stObj = st_texture_object(texObj); struct st_texture_image *stImage = st_texture_image(texImage); GLint postConvWidth, postConvHeight; @@ -690,8 +691,8 @@ st_TexImage(GLcontext * ctx, texImage->Data = NULL; } - /* flag data as dirty */ - stObj->dirtyData = GL_TRUE; + if (stObj->pt) + pipe->texture_update(pipe, stObj->pt, stImage->face, (1 << level)); if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); @@ -866,6 +867,7 @@ st_TexSubimage(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { + struct pipe_context *pipe = ctx->st->pipe; struct st_texture_object *stObj = st_texture_object(texObj); struct st_texture_image *stImage = st_texture_image(texImage); GLuint dstRowStride; @@ -924,8 +926,7 @@ st_TexSubimage(GLcontext * ctx, texImage->Data = NULL; } - /* flag data as dirty */ - stObj->dirtyData = GL_TRUE; + pipe->texture_update(pipe, stObj->pt, stImage->face, (1 << level)); } @@ -1179,8 +1180,7 @@ do_copy_texsubimage(GLcontext *ctx, pipe_surface_reference(&dest_surface, NULL); - /* flag data as dirty */ - stObj->dirtyData = GL_TRUE; + pipe->texture_update(pipe, stObj->pt, stImage->face, (1 << level)); if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); @@ -1481,6 +1481,7 @@ st_finalize_texture(GLcontext *ctx, if (stImage && stObj->pt != stImage->pt) { copy_image_data_to_texture(ctx->st, stObj, level, stImage); *needFlush = GL_TRUE; + pipe->texture_update(pipe, stObj->pt, face, (1 << level)); } } } diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 78f5f451ed..7abccb3a69 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -71,8 +71,6 @@ struct st_texture_object GLboolean imageOverride; GLint depthOverride; GLuint pitchOverride; - - GLboolean dirtyData; }; -- cgit v1.2.3 From 122ed506f4b808503b230bade421018614dbe696 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 19 Mar 2008 18:10:09 -0600 Subject: gallium: added fb_width/height fields to softpipe context These are convenience fields. Otherwise, we have to check cbuf[0] or zsbuf in various places. --- src/gallium/drivers/softpipe/sp_context.h | 2 ++ src/gallium/drivers/softpipe/sp_prim_setup.c | 2 +- src/gallium/drivers/softpipe/sp_quad_stipple.c | 2 +- src/gallium/drivers/softpipe/sp_state_derived.c | 13 ++----------- src/gallium/drivers/softpipe/sp_state_surface.c | 12 ++++++++++++ 5 files changed, 18 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 19e6cfaf02..31d7062dcc 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -84,6 +84,8 @@ struct softpipe_context { unsigned num_samplers; unsigned num_textures; + uint fb_width, fb_height; + /* Counter for occlusion queries. Note this supports overlapping * queries. */ diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 7aa9cf8e85..f2d6043e2e 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -485,7 +485,7 @@ setup_fragcoord_coeff(struct setup_stage *setup, uint slot) /*Y*/ if (setup->softpipe->rasterizer->origin_lower_left) { /* y=0=bottom */ - const int winHeight = setup->softpipe->framebuffer.cbufs[0]->height; + const int winHeight = setup->softpipe->fb_height; setup->coef[slot].a0[1] = (float) (winHeight - 1); setup->coef[slot].dady[1] = -1.0; } diff --git a/src/gallium/drivers/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c index 8660432259..9a39249576 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stipple.c +++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c @@ -25,7 +25,7 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) int y0, y1; uint stipple0, stipple1; if (softpipe->rasterizer->origin_lower_left) { - y0 = softpipe->framebuffer.cbufs[0]->height - 1 - quad->y0; + y0 = softpipe->fb_height - 1 - quad->y0; y1 = y0 - 1; } else { diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 82cb31ece7..53b2056b8a 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -171,17 +171,8 @@ softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe) static void compute_cliprect(struct softpipe_context *sp) { - unsigned surfWidth, surfHeight; - - if (sp->framebuffer.num_cbufs > 0) { - surfWidth = sp->framebuffer.cbufs[0]->width; - surfHeight = sp->framebuffer.cbufs[0]->height; - } - else { - /* no surface? */ - surfWidth = sp->scissor.maxx; - surfHeight = sp->scissor.maxy; - } + uint surfWidth = sp->fb_width; + uint surfHeight = sp->fb_height; if (sp->rasterizer->scissor) { /* clip to scissor rect */ diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c index 124b18b708..2f55684464 100644 --- a/src/gallium/drivers/softpipe/sp_state_surface.c +++ b/src/gallium/drivers/softpipe/sp_state_surface.c @@ -48,6 +48,9 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, struct softpipe_context *sp = softpipe_context(pipe); uint i; + /* updated below */ + sp->fb_width = sp->fb_height = 0; + for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { /* check if changing cbuf */ if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) { @@ -60,6 +63,10 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* update cache */ sp_tile_cache_set_surface(sp->cbuf_cache[i], fb->cbufs[i]); } + if (fb->cbufs[i]) { + sp->fb_width = fb->cbufs[i]->width; + sp->fb_height = fb->cbufs[i]->height; + } } sp->framebuffer.num_cbufs = fb->num_cbufs; @@ -74,6 +81,11 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* update cache */ sp_tile_cache_set_surface(sp->zsbuf_cache, fb->zsbuf); + + if (!sp->fb_width && fb->zsbuf) { + sp->fb_width = fb->zsbuf->width; + sp->fb_height = fb->zsbuf->height; + } } #if 0 -- cgit v1.2.3 From a88202d3b02a24a3bfff95c5e375ead44dae4c5e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 20 Mar 2008 13:10:32 +0000 Subject: gallium: add debug facility to dump random blobs as hex --- src/gallium/auxiliary/util/p_debug.c | 18 ++++++++++++++++++ src/gallium/include/pipe/p_debug.h | 8 ++++++++ 2 files changed, 26 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index bd3a0221ea..c51e9e6a69 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -85,6 +85,22 @@ void debug_printf(const char *format, ...) } +void debug_print_blob( const char *name, + const void *blob, + unsigned size ) +{ + const unsigned *ublob = (const unsigned *)blob; + unsigned i; + + debug_printf("%s (%d dwords%s)\n", name, size/4, + size%4 ? "... plus a few bytes" : ""); + + for (i = 0; i < size/4; i++) { + debug_printf("%d:\t%08x\n", i, ublob[i]); + } +} + + /* TODO: implement a debug_abort that calls EngBugCheckEx on WIN32 */ @@ -240,3 +256,5 @@ debug_dump_flags(const struct debug_named_value *names, return output; } + + diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index e924c1ef60..14f8056924 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -73,6 +73,14 @@ extern "C" { */ void debug_printf(const char *format, ...); + +/* Dump a blob in hex to the same place that debug_printf sends its + * messages: + */ +void debug_print_blob( const char *name, + const void *blob, + unsigned size ); + /** * @sa debug_printf */ -- cgit v1.2.3 From 6a9a3afcf923ec5c67069cdb1656f52675cd8ede Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 09:11:27 -0600 Subject: gallium: added util_make_fragment_passthrough_shader() --- src/gallium/auxiliary/util/u_simple_shaders.c | 95 +++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_simple_shaders.h | 12 ++++ 2 files changed, 107 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index 88e2ab05bd..fca6f11119 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -261,3 +261,98 @@ util_make_fragment_tex_shader(struct pipe_context *pipe) return pipe->create_fs_state(pipe, &shader); } + + + + +/** + * Make simple fragment color pass-through shader. + */ +void * +util_make_fragment_passthrough_shader(struct pipe_context *pipe) +{ + uint maxTokens = 40; + struct tgsi_token *tokens; + struct tgsi_header *header; + struct tgsi_processor *processor; + struct tgsi_full_declaration decl; + struct tgsi_full_instruction inst; + const uint procType = TGSI_PROCESSOR_FRAGMENT; + uint ti, i; + struct pipe_shader_state shader; + + tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); + + /* shader header + */ + *(struct tgsi_version *) &tokens[0] = tgsi_build_version(); + + header = (struct tgsi_header *) &tokens[1]; + *header = tgsi_build_header(); + + processor = (struct tgsi_processor *) &tokens[2]; + *processor = tgsi_build_processor( procType, header ); + + ti = 3; + + /* declare input */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; + decl.Semantic.SemanticIndex = 0; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + /* declare output */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_OUTPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; + decl.Semantic.SemanticIndex = 0; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = 0; + ti += tgsi_build_full_declaration(&decl, + &tokens[ti], + header, + maxTokens - ti); + + + /* MOVE out[0], in[0]; */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_MOV; + inst.Instruction.NumDstRegs = 1; + inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + inst.FullDstRegisters[0].DstRegister.Index = 0; + inst.Instruction.NumSrcRegs = 1; + inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + inst.FullSrcRegisters[0].SrcRegister.Index = 0; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + + /* END instruction */ + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = TGSI_OPCODE_END; + inst.Instruction.NumDstRegs = 0; + inst.Instruction.NumSrcRegs = 0; + ti += tgsi_build_full_instruction(&inst, + &tokens[ti], + header, + maxTokens - ti ); + + assert(ti < maxTokens); + +#if 0 /*debug*/ + tgsi_dump(tokens, 0); +#endif + + shader.tokens = tokens; + return pipe->create_fs_state(pipe, &shader); +} + diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h index 3ef4f28801..ca219a092c 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.h +++ b/src/gallium/auxiliary/util/u_simple_shaders.h @@ -36,6 +36,11 @@ struct pipe_context; +#ifdef __cplusplus +extern "C" { +#endif + + extern void * util_make_vertex_passthrough_shader(struct pipe_context *pipe, uint num_attribs, @@ -47,6 +52,13 @@ extern void * util_make_fragment_tex_shader(struct pipe_context *pipe); +extern void * +util_make_fragment_passthrough_shader(struct pipe_context *pipe); + + +#ifdef __cplusplus +} #endif +#endif -- cgit v1.2.3 From 85e4ec6d118e340eaccd176aa622221642a2e754 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 09:13:51 -0600 Subject: gallium: use the utility pasthrough shaders This avoids the Mesa->TGSI translation step. --- src/mesa/state_tracker/st_cb_clear.c | 137 +++++++---------------------------- src/mesa/state_tracker/st_cb_clear.h | 4 + src/mesa/state_tracker/st_context.c | 1 + src/mesa/state_tracker/st_context.h | 7 ++ 4 files changed, 39 insertions(+), 110 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index c23938dc68..8cc02f63bb 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -49,6 +49,7 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" #include "util/u_pack_color.h" +#include "util/u_simple_shaders.h" #include "cso_cache/cso_context.h" @@ -57,6 +58,20 @@ #define TEST_DRAW_PASSTHROUGH 0 +void +st_destroy_clear(struct st_context *st) +{ + if (st->clear.fs) { + st->pipe->delete_fs_state(st->pipe, st->clear.fs); + st->clear.fs = NULL; + } + if (st->clear.vs) { + st->pipe->delete_vs_state(st->pipe, st->clear.vs); + st->clear.vs = NULL; + } +} + + static GLboolean is_depth_stencil_format(enum pipe_format pipeFormat) { @@ -71,104 +86,6 @@ is_depth_stencil_format(enum pipe_format pipeFormat) -/** - * Create a simple fragment shader that just passes through the fragment color. - */ -static struct st_fragment_program * -make_frag_shader(struct st_context *st) -{ - GLcontext *ctx = st->ctx; - struct st_fragment_program *stfp; - struct gl_program *p; - GLuint interpMode[16]; - GLuint i; - - /* XXX temporary */ - for (i = 0; i < 16; i++) - interpMode[i] = TGSI_INTERPOLATE_LINEAR; - - p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); - if (!p) - return NULL; - - p->NumInstructions = 2; - p->Instructions = _mesa_alloc_instructions(2); - if (!p->Instructions) { - ctx->Driver.DeleteProgram(ctx, p); - return NULL; - } - _mesa_init_instructions(p->Instructions, 2); - /* MOV result.color, fragment.color; */ - p->Instructions[0].Opcode = OPCODE_MOV; - p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[0].DstReg.Index = FRAG_RESULT_COLR; - p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[0].SrcReg[0].Index = FRAG_ATTRIB_COL0; - /* END; */ - p->Instructions[1].Opcode = OPCODE_END; - - p->InputsRead = FRAG_BIT_COL0; - p->OutputsWritten = (1 << FRAG_RESULT_COLR); - - stfp = (struct st_fragment_program *) p; - st_translate_fragment_program(st, stfp, NULL); - - return stfp; -} - - -/** - * Create a simple vertex shader that just passes through the - * vertex position and color. - */ -static struct st_vertex_program * -make_vertex_shader(struct st_context *st) -{ - GLcontext *ctx = st->ctx; - struct st_vertex_program *stvp; - struct gl_program *p; - - p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); - if (!p) - return NULL; - - p->NumInstructions = 3; - p->Instructions = _mesa_alloc_instructions(3); - if (!p->Instructions) { - ctx->Driver.DeleteProgram(ctx, p); - return NULL; - } - _mesa_init_instructions(p->Instructions, 3); - /* MOV result.pos, vertex.pos; */ - p->Instructions[0].Opcode = OPCODE_MOV; - p->Instructions[0].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[0].DstReg.Index = VERT_RESULT_HPOS; - p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[0].SrcReg[0].Index = VERT_ATTRIB_POS; - /* MOV result.color, vertex.color; */ - p->Instructions[1].Opcode = OPCODE_MOV; - p->Instructions[1].DstReg.File = PROGRAM_OUTPUT; - p->Instructions[1].DstReg.Index = VERT_RESULT_COL0; - p->Instructions[1].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[1].SrcReg[0].Index = VERT_ATTRIB_COLOR0; - /* END; */ - p->Instructions[2].Opcode = OPCODE_END; - - p->InputsRead = VERT_BIT_POS | VERT_BIT_COLOR0; - p->OutputsWritten = ((1 << VERT_RESULT_COL0) | - (1 << VERT_RESULT_HPOS)); - - stvp = (struct st_vertex_program *) p; - st_translate_vertex_program(st, stvp, NULL); -#if 0 - assert(stvp->cso); -#endif - - return stvp; -} - - - /** * Draw a screen-aligned quadrilateral. * Coords are window coords with y=0=bottom. These coords will be transformed @@ -311,23 +228,23 @@ clear_with_quad(GLcontext *ctx, } /* fragment shader state: color pass-through program */ - { - static struct st_fragment_program *stfp = NULL; - if (!stfp) { - stfp = make_frag_shader(st); - } - pipe->bind_fs_state(pipe, stfp->driver_shader); + if (!st->clear.fs) { + st->clear.fs = util_make_fragment_passthrough_shader(pipe); } + pipe->bind_fs_state(pipe, st->clear.fs); + #if !TEST_DRAW_PASSTHROUGH /* vertex shader state: color/position pass-through */ - { - static struct st_vertex_program *stvp = NULL; - if (!stvp) { - stvp = make_vertex_shader(st); - } - pipe->bind_vs_state(pipe, stvp->driver_shader); + if (!st->clear.vs) { + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, + TGSI_SEMANTIC_COLOR }; + const uint semantic_indexes[] = { 0, 0 }; + st->clear.vs = util_make_vertex_passthrough_shader(pipe, 2, + semantic_names, + semantic_indexes); } + pipe->bind_vs_state(pipe, st->clear.vs); #endif #if !TEST_DRAW_PASSTHROUGH diff --git a/src/mesa/state_tracker/st_cb_clear.h b/src/mesa/state_tracker/st_cb_clear.h index c715e56bd5..dfa4033faa 100644 --- a/src/mesa/state_tracker/st_cb_clear.h +++ b/src/mesa/state_tracker/st_cb_clear.h @@ -30,6 +30,10 @@ #define ST_CB_CLEAR_H +extern void +st_destroy_clear(struct st_context *st); + + extern void st_init_clear_functions(struct dd_function_table *functions); diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index e1fc885e0e..b1681be2eb 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -155,6 +155,7 @@ static void st_destroy_context_priv( struct st_context *st ) st_destroy_draw( st ); st_destroy_generate_mipmap(st); st_destroy_blit(st); + st_destroy_clear(st); _vbo_DestroyContext(st->ctx); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index ca8307c4ba..8058086d87 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -142,12 +142,19 @@ struct st_context GLuint combined_prog_sn; } pixel_xfer; + /** for glBitmap */ struct { struct st_fragment_program *program; /**< bitmap tex/kil program */ GLuint user_prog_sn; /**< user fragment program serial no. */ struct st_fragment_program *combined_prog; } bitmap; + /** for glClear */ + struct { + void *vs; + void *fs; + } clear; + struct gen_mipmap_state *gen_mipmap; struct blit_state *blit; -- cgit v1.2.3 From 8dd90ee19d97c4b032c2b057d96b3e674be3e1fd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 09:15:24 -0600 Subject: gallium: temporarily disable the memcmp() in cso_set_framebuffer() The memcmp() fails to detect buffer size changes... --- src/gallium/auxiliary/cso_cache/cso_context.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 01fe216447..4a1a6cb79c 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -496,7 +496,8 @@ void cso_restore_vertex_shader(struct cso_context *ctx) void cso_set_framebuffer(struct cso_context *ctx, const struct pipe_framebuffer_state *fb) { - if (memcmp(&ctx->fb, fb, sizeof(*fb))) { + /* XXX this memcmp() fails to detect buffer size changes */ + if (1/*memcmp(&ctx->fb, fb, sizeof(*fb))*/) { ctx->fb = *fb; ctx->pipe->set_framebuffer_state(ctx->pipe, fb); } -- cgit v1.2.3 From 8c71406c74b80fa2d2b1a488938c3b9dfc156343 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 10:44:27 -0600 Subject: gallium: added util_draw_vertex_buffer() --- src/gallium/auxiliary/util/u_draw_quad.c | 59 +++++++++++++++++++++----------- src/gallium/auxiliary/util/u_draw_quad.h | 16 +++++++++ 2 files changed, 55 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c index 79a69de633..37e8533609 100644 --- a/src/gallium/auxiliary/util/u_draw_quad.c +++ b/src/gallium/auxiliary/util/u_draw_quad.c @@ -33,16 +33,52 @@ #include "util/u_draw_quad.h" +/** + * Draw a simple vertex buffer / primitive. + * Limited to float[4] vertex attribs, tightly packed. + */ +void +util_draw_vertex_buffer(struct pipe_context *pipe, + struct pipe_buffer *vbuf, + uint prim_type, + uint num_verts, + uint num_attribs) +{ + struct pipe_vertex_buffer vbuffer; + struct pipe_vertex_element velement; + uint i; + + /* tell pipe about the vertex buffer */ + vbuffer.buffer = vbuf; + vbuffer.pitch = num_attribs * 4 * sizeof(float); /* vertex size */ + vbuffer.buffer_offset = 0; + pipe->set_vertex_buffer(pipe, 0, &vbuffer); + + /* tell pipe about the vertex attributes */ + for (i = 0; i < num_attribs; i++) { + velement.src_offset = i * 4 * sizeof(float); + velement.vertex_buffer_index = 0; + velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + velement.nr_components = 4; + pipe->set_vertex_element(pipe, i, &velement); + } + + /* draw */ + pipe->draw_arrays(pipe, prim_type, 0, num_verts); +} + + + /** * Draw screen-aligned textured quad. + * Note: this function allocs/destroys a vertex buffer and isn't especially + * efficient. */ void util_draw_texquad(struct pipe_context *pipe, float x0, float y0, float x1, float y1, float z) { struct pipe_buffer *vbuf; - struct pipe_vertex_buffer vbuffer; - struct pipe_vertex_element velement; uint numAttribs = 2, vertexBytes, i, j; float *v; @@ -89,24 +125,7 @@ util_draw_texquad(struct pipe_context *pipe, pipe->winsys->buffer_unmap(pipe->winsys, vbuf); - /* tell pipe about the vertex buffer */ - vbuffer.buffer = vbuf; - vbuffer.pitch = numAttribs * 4 * sizeof(float); /* vertex size */ - vbuffer.buffer_offset = 0; - pipe->set_vertex_buffer(pipe, 0, &vbuffer); - - /* tell pipe about the vertex attributes */ - for (i = 0; i < numAttribs; i++) { - velement.src_offset = i * 4 * sizeof(float); - velement.vertex_buffer_index = 0; - velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - velement.nr_components = 4; - pipe->set_vertex_element(pipe, i, &velement); - } - - /* draw */ - pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_FAN, 0, 4); + util_draw_vertex_buffer(pipe, vbuf, PIPE_PRIM_TRIANGLE_FAN, 4, 2); - /* XXX: do one-time */ pipe_buffer_reference(pipe->winsys, &vbuf, NULL); } diff --git a/src/gallium/auxiliary/util/u_draw_quad.h b/src/gallium/auxiliary/util/u_draw_quad.h index a97f55d2ef..5b6539a99c 100644 --- a/src/gallium/auxiliary/util/u_draw_quad.h +++ b/src/gallium/auxiliary/util/u_draw_quad.h @@ -29,9 +29,25 @@ #define U_DRAWQUAD_H +#ifdef __cplusplus +extern "C" { +#endif + + +extern void +util_draw_vertex_buffer(struct pipe_context *pipe, + struct pipe_buffer *vbuf, + uint num_attribs, uint num_verts, uint prim_type); + + extern void util_draw_texquad(struct pipe_context *pipe, float x0, float y0, float x1, float y1, float z); +#ifdef __cplusplus +} +#endif + + #endif -- cgit v1.2.3 From 309d6e52c5c1cdeca1434cbe29e869b9176e5fa5 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 10:44:51 -0600 Subject: gallium: create vertex buffer once and re-use. --- src/gallium/auxiliary/util/u_gen_mipmap.c | 71 ++++++++++++++++++++++++++++--- 1 file changed, 65 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 028b180a77..5824843069 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -64,6 +64,9 @@ struct gen_mipmap_state /*struct pipe_viewport_state viewport;*/ struct pipe_sampler_state *vs; struct pipe_sampler_state *fs; + + struct pipe_buffer *vbuf; /**< quad vertices */ + float vertices[4][2][4]; /**< vertex/texcoords for quad */ }; @@ -683,6 +686,7 @@ util_create_gen_mipmap(struct pipe_context *pipe, struct cso_context *cso) { struct gen_mipmap_state *ctx; + uint i; ctx = CALLOC_STRUCT(gen_mipmap_state); if (!ctx) @@ -744,10 +748,62 @@ util_create_gen_mipmap(struct pipe_context *pipe, /* fragment shader */ ctx->fs = util_make_fragment_tex_shader(pipe); + ctx->vbuf = pipe->winsys->buffer_create(pipe->winsys, + 32, + PIPE_BUFFER_USAGE_VERTEX, + 4 * 2 * 4 * sizeof(float)); + if (!ctx->vbuf) { + FREE(ctx); + return NULL; + } + + /* vertex data that doesn't change */ + for (i = 0; i < 4; i++) { + ctx->vertices[i][0][2] = 0.0f; /* z */ + ctx->vertices[i][0][3] = 1.0f; /* w */ + ctx->vertices[i][1][2] = 0.0f; /* r */ + ctx->vertices[i][1][3] = 1.0f; /* q */ + } + return ctx; } +static void +set_vertex_data(struct gen_mipmap_state *ctx, float width, float height) +{ + void *buf; + + ctx->vertices[0][0][0] = 0.0f; /*x*/ + ctx->vertices[0][0][1] = 0.0f; /*y*/ + ctx->vertices[0][1][0] = 0.0f; /*s*/ + ctx->vertices[0][1][1] = 0.0f; /*t*/ + + ctx->vertices[1][0][0] = width; /*x*/ + ctx->vertices[1][0][1] = 0.0f; /*y*/ + ctx->vertices[1][1][0] = 1.0f; /*s*/ + ctx->vertices[1][1][1] = 0.0f; /*t*/ + + ctx->vertices[2][0][0] = width; + ctx->vertices[2][0][1] = height; + ctx->vertices[2][1][0] = 1.0f; + ctx->vertices[2][1][1] = 1.0f; + + ctx->vertices[3][0][0] = 0.0f; + ctx->vertices[3][0][1] = height; + ctx->vertices[3][1][0] = 0.0f; + ctx->vertices[3][1][1] = 1.0f; + + buf = ctx->pipe->winsys->buffer_map(ctx->pipe->winsys, ctx->vbuf, + PIPE_BUFFER_USAGE_CPU_WRITE); + + memcpy(buf, ctx->vertices, sizeof(ctx->vertices)); + + ctx->pipe->winsys->buffer_unmap(ctx->pipe->winsys, ctx->vbuf); +} + + + /** * Destroy a mipmap generation context */ @@ -759,6 +815,8 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx) pipe->delete_vs_state(pipe, ctx->vs); pipe->delete_fs_state(pipe, ctx->fs); + pipe->winsys->buffer_destroy(pipe->winsys, ctx->vbuf); + FREE(ctx); } @@ -863,12 +921,13 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, pipe->set_sampler_textures(pipe, 1, &pt); /* quad coords in window coords (bypassing clipping, viewport mapping) */ - util_draw_texquad(pipe, - 0.0F, 0.0F, /* x0, y0 */ - (float) pt->width[dstLevel], /* x1 */ - (float) pt->height[dstLevel], /* y1 */ - 0.0F); /* z */ - + set_vertex_data(ctx, + (float) pt->width[dstLevel], + (float) pt->height[dstLevel]); + util_draw_vertex_buffer(ctx->pipe, ctx->vbuf, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ pipe->flush(pipe, PIPE_FLUSH_WAIT); -- cgit v1.2.3 From e1406c8d2366dccac0037e5329217d1c8c265eaf Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 10:46:44 -0600 Subject: gallium: create one vertex buffer and re-use, also enable bypass_clipping Quad clears should be a little more efficient now. --- src/mesa/state_tracker/st_cb_clear.c | 76 +++++++++++++++++++++--------------- src/mesa/state_tracker/st_context.h | 2 + 2 files changed, 47 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 8cc02f63bb..dc0d87acd3 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -50,25 +50,33 @@ #include "pipe/p_winsys.h" #include "util/u_pack_color.h" #include "util/u_simple_shaders.h" +#include "util/u_draw_quad.h" #include "cso_cache/cso_context.h" /* XXX for testing draw module vertex passthrough: */ +/* XXX this hack is broken now */ #define TEST_DRAW_PASSTHROUGH 0 void st_destroy_clear(struct st_context *st) { + struct pipe_context *pipe = st->pipe; + if (st->clear.fs) { - st->pipe->delete_fs_state(st->pipe, st->clear.fs); + pipe->delete_fs_state(pipe, st->clear.fs); st->clear.fs = NULL; } if (st->clear.vs) { - st->pipe->delete_vs_state(st->pipe, st->clear.vs); + pipe->delete_vs_state(pipe, st->clear.vs); st->clear.vs = NULL; } + if (st->clear.vbuf) { + pipe->winsys->buffer_destroy(pipe->winsys, st->clear.vbuf); + st->clear.vbuf = NULL; + } } @@ -96,45 +104,51 @@ draw_quad(GLcontext *ctx, float x0, float y0, float x1, float y1, GLfloat z, const GLfloat color[4]) { - GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ + struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; GLuint i; + void *buf; -#if TEST_DRAW_PASSTHROUGH - /* invert Y coords (may be off by one pixel) */ - y0 = ctx->DrawBuffer->Height - y0; - y1 = ctx->DrawBuffer->Height - y1; -#endif + if (!st->clear.vbuf) { + st->clear.vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(st->clear.vertices)); + } /* positions */ - verts[0][0][0] = x0; - verts[0][0][1] = y0; + st->clear.vertices[0][0][0] = x0; + st->clear.vertices[0][0][1] = y0; - verts[1][0][0] = x1; - verts[1][0][1] = y0; + st->clear.vertices[1][0][0] = x1; + st->clear.vertices[1][0][1] = y0; - verts[2][0][0] = x1; - verts[2][0][1] = y1; + st->clear.vertices[2][0][0] = x1; + st->clear.vertices[2][0][1] = y1; - verts[3][0][0] = x0; - verts[3][0][1] = y1; + st->clear.vertices[3][0][0] = x0; + st->clear.vertices[3][0][1] = y1; /* same for all verts: */ for (i = 0; i < 4; i++) { - verts[i][0][2] = z; - verts[i][0][3] = 1.0; - verts[i][1][0] = color[0]; - verts[i][1][1] = color[1]; - verts[i][1][2] = color[2]; - verts[i][1][3] = color[3]; + st->clear.vertices[i][0][2] = z; + st->clear.vertices[i][0][3] = 1.0; + st->clear.vertices[i][1][0] = color[0]; + st->clear.vertices[i][1][1] = color[1]; + st->clear.vertices[i][1][2] = color[2]; + st->clear.vertices[i][1][3] = color[3]; } - st_draw_vertices(ctx, PIPE_PRIM_POLYGON, 4, (float *) verts, 2, -#if TEST_DRAW_PASSTHROUGH - GL_TRUE -#else - GL_FALSE -#endif - ); + /* put vertex data into vbuf */ + buf = pipe->winsys->buffer_map(pipe->winsys, st->clear.vbuf, + PIPE_BUFFER_USAGE_CPU_WRITE); + memcpy(buf, st->clear.vertices, sizeof(st->clear.vertices)); + pipe->winsys->buffer_unmap(pipe->winsys, st->clear.vbuf); + + /* draw */ + util_draw_vertex_buffer(pipe, st->clear.vbuf, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ } @@ -216,12 +230,12 @@ clear_with_quad(GLcontext *ctx, cso_set_depth_stencil_alpha(st->cso_context, &depth_stencil); } - /* rasterizer state: nothing */ + /* rasterizer state: bypass clipping */ { struct pipe_rasterizer_state raster; memset(&raster, 0, sizeof(raster)); -#if TEST_DRAW_PASSTHROUGH raster.bypass_clipping = 1; +#if TEST_DRAW_PASSTHROUGH raster.bypass_vs = 1; #endif cso_set_rasterizer(st->cso_context, &raster); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 8058086d87..2563c7fed0 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -153,6 +153,8 @@ struct st_context struct { void *vs; void *fs; + float vertices[4][2][4]; /**< vertex pos + color */ + struct pipe_buffer *vbuf; } clear; struct gen_mipmap_state *gen_mipmap; -- cgit v1.2.3 From f259ea0347754e0e8c93fd16796fc1db72b03372 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 20 Mar 2008 19:25:44 +0000 Subject: gallium: remove unused local var --- src/gallium/auxiliary/util/u_simple_shaders.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index fca6f11119..442467ecad 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -278,7 +278,7 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe) struct tgsi_full_declaration decl; struct tgsi_full_instruction inst; const uint procType = TGSI_PROCESSOR_FRAGMENT; - uint ti, i; + uint ti; struct pipe_shader_state shader; tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); -- cgit v1.2.3 From 482f4995253a0c295dc02e34e58a138ac8822c54 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 20 Mar 2008 20:25:40 +0100 Subject: gallium: Fix build on Windows. --- src/gallium/auxiliary/util/u_simple_shaders.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index 442467ecad..a2c918b7fe 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -68,7 +68,7 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, uint ti, i; struct pipe_shader_state shader; - tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); + tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0])); /* shader header */ @@ -173,7 +173,7 @@ util_make_fragment_tex_shader(struct pipe_context *pipe) uint ti; struct pipe_shader_state shader; - tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); + tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0])); /* shader header */ -- cgit v1.2.3 From 400b12b4ceda32cc35b60d0484dfd333f1749b8e Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 20 Mar 2008 20:34:03 +0100 Subject: gallium: Fix build on Windows. --- src/gallium/auxiliary/util/u_simple_shaders.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index a2c918b7fe..8cb74e81d2 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -281,7 +281,7 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe) uint ti; struct pipe_shader_state shader; - tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0])); + tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0])); /* shader header */ -- cgit v1.2.3 From 9fa88fb3c57f6868dc169b692528cf23ceced76c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 20 Mar 2008 20:34:46 +0000 Subject: gallium: Give some chance for the table to actually grow. --- src/gallium/auxiliary/util/u_handle_table.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index 8b0f7fca4b..5a731a6b96 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -171,8 +171,7 @@ handle_table_set(struct handle_table *ht, assert(ht); assert(handle > 0); - assert(handle <= ht->size); - if(!handle || handle > ht->size) + if(!handle) return 0; assert(object); -- cgit v1.2.3 From c2044eaca96abfae153651ec609b5af2fd0b6fb0 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 14:17:06 -0600 Subject: gallium: in clear_with_quad() check fb orientation, invert Y if needed --- src/mesa/state_tracker/st_cb_clear.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index dc0d87acd3..5ca15df602 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -165,9 +165,17 @@ clear_with_quad(GLcontext *ctx, struct st_context *st = ctx->st; struct pipe_context *pipe = st->pipe; const GLfloat x0 = ctx->DrawBuffer->_Xmin; - const GLfloat y0 = ctx->DrawBuffer->_Ymin; const GLfloat x1 = ctx->DrawBuffer->_Xmax; - const GLfloat y1 = ctx->DrawBuffer->_Ymax; + GLfloat y0, y1; + + if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { + y0 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax; + y1 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin; + } + else { + y0 = ctx->DrawBuffer->_Ymin; + y1 = ctx->DrawBuffer->_Ymax; + } /* printf("%s %s%s%s %f,%f %f,%f\n", __FUNCTION__, -- cgit v1.2.3 From 3aa8c660ffa9763dbdfdf2d75e991a08e41c29b6 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 14:20:25 -0600 Subject: gallium: Y-invert blit depending on framebuffer orientation --- src/mesa/state_tracker/st_cb_blit.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c index 64314a5078..63211d8b66 100644 --- a/src/mesa/state_tracker/st_cb_blit.c +++ b/src/mesa/state_tracker/st_cb_blit.c @@ -85,11 +85,14 @@ st_BlitFramebuffer(GLcontext *ctx, struct pipe_surface *srcSurf = srcRb->surface; struct pipe_surface *dstSurf = dstRb->surface; - srcY0 = srcRb->Base.Height - srcY0; - srcY1 = srcRb->Base.Height - srcY1; - - dstY0 = dstRb->Base.Height - dstY0; - dstY1 = dstRb->Base.Height - dstY1; + if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { + /* invert Y */ + srcY0 = srcRb->Base.Height - srcY0; + srcY1 = srcRb->Base.Height - srcY1; + + dstY0 = dstRb->Base.Height - dstY0; + dstY1 = dstRb->Base.Height - dstY1; + } util_blit_pixels(st->blit, srcSurf, srcX0, srcY0, srcX1, srcY1, -- cgit v1.2.3 From 18be9a588acb46dad52321c91eab9927bc332756 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 14:26:51 -0600 Subject: gallium: use sizeof(vertex buffer) --- src/gallium/auxiliary/util/u_gen_mipmap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 5824843069..b84752e0b4 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -751,7 +751,7 @@ util_create_gen_mipmap(struct pipe_context *pipe, ctx->vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, PIPE_BUFFER_USAGE_VERTEX, - 4 * 2 * 4 * sizeof(float)); + sizeof(ctx->vertices)); if (!ctx->vbuf) { FREE(ctx); return NULL; -- cgit v1.2.3 From b028d32af4f5e7beeb4e659a4726e0d35d105ccc Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 14:27:14 -0600 Subject: gallium: create one vertex buffer and re-use --- src/gallium/auxiliary/util/u_blit.c | 80 ++++++++++++++++++++++++++++++++++--- 1 file changed, 75 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index d05dae3af8..bb316554a4 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -61,6 +61,9 @@ struct blit_state /*struct pipe_viewport_state viewport;*/ struct pipe_sampler_state *vs; struct pipe_sampler_state *fs; + + struct pipe_buffer *vbuf; /**< quad vertices */ + float vertices[4][2][4]; /**< vertex/texcoords for quad */ }; @@ -72,6 +75,7 @@ struct blit_state * util_create_blit(struct pipe_context *pipe, struct cso_context *cso) { struct blit_state *ctx; + uint i; ctx = CALLOC_STRUCT(blit_state); if (!ctx) @@ -132,6 +136,24 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) /* fragment shader */ ctx->fs = util_make_fragment_tex_shader(pipe); + ctx->vbuf = pipe->winsys->buffer_create(pipe->winsys, + 32, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(ctx->vertices)); + if (!ctx->vbuf) { + FREE(ctx); + ctx->pipe->delete_fs_state(ctx->pipe, ctx->fs); + ctx->pipe->delete_vs_state(ctx->pipe, ctx->vs); + return NULL; + } + + /* init vertex data that doesn't change */ + for (i = 0; i < 4; i++) { + ctx->vertices[i][0][3] = 1.0f; /* w */ + ctx->vertices[i][1][2] = 0.0f; /* r */ + ctx->vertices[i][1][3] = 1.0f; /* q */ + } + return ctx; } @@ -147,10 +169,55 @@ util_destroy_blit(struct blit_state *ctx) pipe->delete_vs_state(pipe, ctx->vs); pipe->delete_fs_state(pipe, ctx->fs); + pipe->winsys->buffer_destroy(pipe->winsys, ctx->vbuf); + FREE(ctx); } +/** + * Setup vertex data for the textured quad we'll draw. + * Note: y=0=top + */ +static void +setup_vertex_data(struct blit_state *ctx, + float x0, float y0, float x1, float y1, float z) +{ + void *buf; + + ctx->vertices[0][0][0] = x0; + ctx->vertices[0][0][1] = y0; + ctx->vertices[0][0][2] = z; + ctx->vertices[0][1][0] = 0.0f; /*s*/ + ctx->vertices[0][1][1] = 0.0f; /*t*/ + + ctx->vertices[1][0][0] = x1; + ctx->vertices[1][0][1] = y0; + ctx->vertices[1][0][2] = z; + ctx->vertices[1][1][0] = 1.0f; /*s*/ + ctx->vertices[1][1][1] = 0.0f; /*t*/ + + ctx->vertices[2][0][0] = x1; + ctx->vertices[2][0][1] = y1; + ctx->vertices[2][0][2] = z; + ctx->vertices[2][1][0] = 1.0f; + ctx->vertices[2][1][1] = 1.0f; + + ctx->vertices[3][0][0] = x0; + ctx->vertices[3][0][1] = y1; + ctx->vertices[3][0][2] = z; + ctx->vertices[3][1][0] = 0.0f; + ctx->vertices[3][1][1] = 1.0f; + + buf = ctx->pipe->winsys->buffer_map(ctx->pipe->winsys, ctx->vbuf, + PIPE_BUFFER_USAGE_CPU_WRITE); + + memcpy(buf, ctx->vertices, sizeof(ctx->vertices)); + + ctx->pipe->winsys->buffer_unmap(ctx->pipe->winsys, ctx->vbuf); +} + + /** * Copy pixel block from src surface to dst surface. * Overlapping regions are acceptable. @@ -254,11 +321,14 @@ util_blit_pixels(struct blit_state *ctx, cso_set_framebuffer(ctx->cso, &fb); /* draw quad */ - util_draw_texquad(pipe, - (float)dstX0, - (float)dstY0, - (float)dstX1, - (float)dstY1, z); + setup_vertex_data(ctx, + (float) dstX0, (float) dstY0, + (float) dstX1, (float) dstY1, z); + + util_draw_vertex_buffer(ctx->pipe, ctx->vbuf, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ /* restore state we changed */ cso_restore_blend(ctx->cso); -- cgit v1.2.3 From 3ece9ace540447c4f0d340a6f9f27980b54f83fb Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 14:49:56 -0600 Subject: gallium: glBitmap code now separated from glDraw/CopyPixels code --- src/mesa/state_tracker/st_cb_bitmap.c | 524 ++++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_bitmap.h | 41 +++ 2 files changed, 565 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_bitmap.c create mode 100644 src/mesa/state_tracker/st_cb_bitmap.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c new file mode 100644 index 0000000000..33256196bb --- /dev/null +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -0,0 +1,524 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Brian Paul + */ + +#include "main/imports.h" +#include "main/image.h" +#include "main/macros.h" +#include "main/texformat.h" +#include "shader/program.h" +#include "shader/prog_parameter.h" +#include "shader/prog_print.h" + +#include "st_context.h" +#include "st_atom.h" +#include "st_atom_constbuf.h" +#include "st_program.h" +#include "st_cb_bitmap.h" +#include "st_mesa_to_tgsi.h" +#include "st_texture.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "pipe/p_winsys.h" +#include "util/p_tile.h" +#include "util/u_draw_quad.h" +#include "util/u_simple_shaders.h" +#include "shader/prog_instruction.h" +#include "cso_cache/cso_context.h" + + + +/** + * Make fragment program for glBitmap: + * Sample the texture and kill the fragment if the bit is 0. + * This program will be combined with the user's fragment program. + */ +static struct st_fragment_program * +make_bitmap_fragment_program(GLcontext *ctx) +{ + struct st_fragment_program *stfp; + struct gl_program *p; + GLuint ic = 0; + + p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + if (!p) + return NULL; + + p->NumInstructions = 5; + + p->Instructions = _mesa_alloc_instructions(p->NumInstructions); + if (!p->Instructions) { + ctx->Driver.DeleteProgram(ctx, p); + return NULL; + } + _mesa_init_instructions(p->Instructions, p->NumInstructions); + + /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */ + p->Instructions[ic].Opcode = OPCODE_TEX; + p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; + p->Instructions[ic].DstReg.Index = 0; + p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; + p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; + p->Instructions[ic].TexSrcUnit = 0; + p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; + + /* SWZ tmp0.x, tmp0.x, 1111; # tmp0.x = 1.0 */ + p->Instructions[ic].Opcode = OPCODE_SWZ; + p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; + p->Instructions[ic].DstReg.Index = 0; + p->Instructions[ic].DstReg.WriteMask = WRITEMASK_X; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + p->Instructions[ic].SrcReg[0].Swizzle + = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE ); + ic++; + + /* SUB tmp0, tmp0.wwww, tmp0.xxxx; # tmp0.w -= 1 */ + p->Instructions[ic].Opcode = OPCODE_SUB; + p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; + p->Instructions[ic].DstReg.Index = 0; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; + p->Instructions[ic].SrcReg[1].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[1].Index = 0; + p->Instructions[ic].SrcReg[1].Swizzle = SWIZZLE_XXXX; /* 1.0 */ + ic++; + + /* KIL if tmp0 < 0 */ + p->Instructions[ic].Opcode = OPCODE_KIL; + p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + p->Instructions[ic].SrcReg[0].Index = 0; + ic++; + + /* END; */ + p->Instructions[ic++].Opcode = OPCODE_END; + + assert(ic == p->NumInstructions); + + p->InputsRead = FRAG_BIT_TEX0; + p->OutputsWritten = 0x0; + + stfp = (struct st_fragment_program *) p; + stfp->Base.UsesKill = GL_TRUE; + st_translate_fragment_program(ctx->st, stfp, NULL); + + return stfp; +} + + +/** + * Combine basic bitmap fragment program with the user-defined program. + */ +static struct st_fragment_program * +combined_bitmap_fragment_program(GLcontext *ctx) +{ + struct st_context *st = ctx->st; + struct st_fragment_program *stfp; + + if (!st->bitmap.program) { + /* create the basic bitmap fragment program */ + st->bitmap.program = make_bitmap_fragment_program(ctx); + } + + if (st->bitmap.user_prog_sn == st->fp->serialNo) { + /* re-use */ + stfp = st->bitmap.combined_prog; + } + else { + /* Concatenate the bitmap program with the current user-defined program. + */ + stfp = (struct st_fragment_program *) + _mesa_combine_programs(ctx, + &st->bitmap.program->Base.Base, + &st->fp->Base.Base); + +#if 0 + { + struct gl_program *p = &stfp->Base.Base; + printf("Combined bitmap program:\n"); + _mesa_print_program(p); + printf("InputsRead: 0x%x\n", p->InputsRead); + printf("OutputsWritten: 0x%x\n", p->OutputsWritten); + _mesa_print_parameter_list(p->Parameters); + } +#endif + + /* translate to TGSI tokens */ + st_translate_fragment_program(st, stfp, NULL); + + /* save new program, update serial numbers */ + st->bitmap.user_prog_sn = st->fp->serialNo; + st->bitmap.combined_prog = stfp; + } + + /* Ideally we'd have updated the pipe constants during the normal + * st/atom mechanism. But we can't since this is specific to glBitmap. + */ + st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT); + + return stfp; +} + + + +/** + * Create a texture which represents a bitmap image. + */ +static struct pipe_texture * +make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *surface; + uint format = 0, cpp, comp; + ubyte *dest; + struct pipe_texture *pt; + int row, col; + + /* find a texture format we know */ + if (screen->is_format_supported( screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE )) { + format = PIPE_FORMAT_U_I8; + cpp = 1; + comp = 0; + } + else if (screen->is_format_supported( screen, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_TEXTURE )) { + format = PIPE_FORMAT_A8R8G8B8_UNORM; + cpp = 4; + comp = 3; /* alpha channel */ /*XXX little-endian dependency */ + } + else { + /* XXX support more formats */ + assert( 0 ); + } + + /** + * Create a texture. + */ + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, width, height, + 1, 0); + if (!pt) + return NULL; + + if (unpack->BufferObj && unpack->BufferObj->Name) { + /* + pt->region = buffer_object_region(unpack->BufferObj); + */ + printf("st_Bitmap (sourcing from PBO not implemented yet)\n"); + } + + surface = screen->get_tex_surface(screen, pt, 0, 0, 0); + + /* map texture surface */ + dest = pipe_surface_map(surface); + + /* Put image into texture surface. + * Note that the image is actually going to be upside down in + * the texture. We deal with that with texcoords. + */ + + for (row = 0; row < height; row++) { + const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, + bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); + ubyte *destRow = dest + row * surface->pitch * cpp; + + if (unpack->LsbFirst) { + /* Lsb first */ + GLubyte mask = 1U << (unpack->SkipPixels & 0x7); + for (col = 0; col < width; col++) { + + /* set texel to 255 if bit is set */ + destRow[comp] = (*src & mask) ? 255 : 0; + destRow += cpp; + + if (mask == 128U) { + src++; + mask = 1U; + } + else { + mask = mask << 1; + } + } + + /* get ready for next row */ + if (mask != 1) + src++; + } + else { + /* Msb first */ + GLubyte mask = 128U >> (unpack->SkipPixels & 0x7); + for (col = 0; col < width; col++) { + + /* set texel to 255 if bit is set */ + destRow[comp] =(*src & mask) ? 255 : 0; + destRow += cpp; + + if (mask == 1U) { + src++; + mask = 128U; + } + else { + mask = mask >> 1; + } + } + + /* get ready for next row */ + if (mask != 128) + src++; + } + + } /* row */ + + /* Release surface */ + pipe_surface_unmap(surface); + pipe_surface_reference(&surface, NULL); + pipe->texture_update(pipe, pt, 0, 0x1); + + pt->format = format; + + return pt; +} + + +static void +setup_bitmap_vertex_data(struct st_context *st, + int x, int y, int width, int height, + float z, const float color[4]) +{ + struct pipe_context *pipe = st->pipe; + const struct gl_framebuffer *fb = st->ctx->DrawBuffer; + const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP); + const GLfloat x0 = x; + const GLfloat x1 = x + width; + const GLfloat y0 = invert ? (fb->Height - y - height) : y; + const GLfloat y1 = invert ? (y0 + height) : y + height; + const GLfloat bias = st->bitmap_texcoord_bias; + const GLfloat xBias = bias / (x1-x0); + const GLfloat yBias = bias / (y1-y0); + const GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias; + const GLfloat tTop = 1.0 - yBias, tBot = 1.0 - tTop - yBias; + GLuint i; + void *buf; + + if (!st->bitmap.vbuf) { + st->bitmap.vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(st->bitmap.vertices)); + } + + /* positions, texcoords */ + st->bitmap.vertices[0][0][0] = x0; + st->bitmap.vertices[0][0][1] = y0; + st->bitmap.vertices[0][2][0] = sLeft; + st->bitmap.vertices[0][2][1] = tTop; + + st->bitmap.vertices[1][0][0] = x1; + st->bitmap.vertices[1][0][1] = y0; + st->bitmap.vertices[1][2][0] = sRight; + st->bitmap.vertices[1][2][1] = tTop; + + st->bitmap.vertices[2][0][0] = x1; + st->bitmap.vertices[2][0][1] = y1; + st->bitmap.vertices[2][2][0] = sRight; + st->bitmap.vertices[2][2][1] = tBot; + + st->bitmap.vertices[3][0][0] = x0; + st->bitmap.vertices[3][0][1] = y1; + st->bitmap.vertices[3][2][0] = sLeft; + st->bitmap.vertices[3][2][1] = tBot; + + /* same for all verts: */ + for (i = 0; i < 4; i++) { + st->bitmap.vertices[i][0][2] = z; + st->bitmap.vertices[i][0][3] = 1.0; + st->bitmap.vertices[i][1][0] = color[0]; + st->bitmap.vertices[i][1][1] = color[1]; + st->bitmap.vertices[i][1][2] = color[2]; + st->bitmap.vertices[i][1][3] = color[3]; + st->bitmap.vertices[i][2][2] = 0.0; /*R*/ + st->bitmap.vertices[i][2][3] = 1.0; /*Q*/ + } + + /* put vertex data into vbuf */ + buf = pipe->winsys->buffer_map(pipe->winsys, st->bitmap.vbuf, + PIPE_BUFFER_USAGE_CPU_WRITE); + memcpy(buf, st->bitmap.vertices, sizeof(st->bitmap.vertices)); + pipe->winsys->buffer_unmap(pipe->winsys, st->bitmap.vbuf); +} + + + +/** + * Render a glBitmap by drawing a textured quad + */ +static void +draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, + GLsizei width, GLsizei height, + struct pipe_texture *pt, + struct st_fragment_program *stfp) +{ + struct st_context *st = ctx->st; + struct pipe_context *pipe = ctx->st->pipe; + struct cso_context *cso = ctx->st->cso_context; + GLuint maxSize; + + /* limit checks */ + /* XXX if DrawPixels image is larger than max texture size, break + * it up into chunks. + */ + maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1); + assert(width <= maxSize); + assert(height <= maxSize); + + cso_save_rasterizer(cso); + //cso_save_viewport(cso); + + /* rasterizer state: just scissor */ + { + struct pipe_rasterizer_state rasterizer; + memset(&rasterizer, 0, sizeof(rasterizer)); + if (ctx->Scissor.Enabled) + rasterizer.scissor = 1; + rasterizer.bypass_clipping = 1; + + cso_set_rasterizer(cso, &rasterizer); + } + + /* fragment shader state: TEX lookup program */ + pipe->bind_fs_state(pipe, stfp->driver_shader); + + /* vertex shader state: position + texcoord pass-through */ + pipe->bind_vs_state(pipe, st->bitmap.vs); + + /* sampler / texture state */ + { + struct pipe_sampler_state sampler; + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP; + sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.normalized_coords = 1; + + cso_single_sampler(cso, 0, &sampler); + cso_single_sampler_done(cso); + + pipe->set_sampler_textures(pipe, 1, &pt); + } + + /* draw textured quad */ + setup_bitmap_vertex_data(st, x, y, width, height, + ctx->Current.RasterPos[2], + ctx->Current.RasterColor); + + util_draw_vertex_buffer(pipe, st->bitmap.vbuf, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 3); /* attribs/vert */ + + + /* restore state */ + cso_restore_rasterizer(cso); + cso_restore_samplers(cso); + //cso_restore_viewport(cso); + /* shaders don't go through cso yet */ + pipe->bind_fs_state(pipe, st->fp->driver_shader); + pipe->bind_vs_state(pipe, st->vp->driver_shader); + + pipe->set_sampler_textures(pipe, ctx->st->state.num_textures, + ctx->st->state.sampler_texture); +} + + + +static void +st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap ) +{ + struct st_fragment_program *stfp; + struct st_context *st = ctx->st; + struct pipe_texture *pt; + + stfp = combined_bitmap_fragment_program(ctx); + + if (!st->bitmap.vs) { + /* create pass-through vertex shader now */ + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, + TGSI_SEMANTIC_COLOR, + TGSI_SEMANTIC_GENERIC }; + const uint semantic_indexes[] = { 0, 0, 0 }; + st->bitmap.vs = util_make_vertex_passthrough_shader(st->pipe, 3, + semantic_names, + semantic_indexes); + } + + st_validate_state(st); + + pt = make_bitmap_texture(ctx, width, height, unpack, bitmap); + if (pt) { + draw_bitmap_quad(ctx, x, y, ctx->Current.RasterPos[2], + width, height, + pt, stfp); + pipe_texture_reference(&pt, NULL); + } +} + + + +void st_init_bitmap_functions(struct dd_function_table *functions) +{ + functions->Bitmap = st_Bitmap; +} + + +void +st_destroy_bitmap(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + + /* XXX free frag shader state */ + + if (st->bitmap.vs) { + pipe->delete_vs_state(pipe, st->bitmap.vs); + st->bitmap.vs = NULL; + } + if (st->bitmap.vbuf) { + pipe->winsys->buffer_destroy(pipe->winsys, st->bitmap.vbuf); + st->bitmap.vbuf = NULL; + } +} + diff --git a/src/mesa/state_tracker/st_cb_bitmap.h b/src/mesa/state_tracker/st_cb_bitmap.h new file mode 100644 index 0000000000..ac19e0ebb1 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_bitmap.h @@ -0,0 +1,41 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_BITMAP_H +#define ST_CB_BITMAP_H + + +extern void +st_init_bitmap_functions(struct dd_function_table *functions); + + +extern void +st_destroy_bitmap(struct st_context *st); + + +#endif /* ST_CB_BITMAP_H */ -- cgit v1.2.3 From f6cd3778c54c0329c3f497a7368a158087d653d3 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 14:50:17 -0600 Subject: gallium: glBitmap code now separe from glDraw/CopyPixels code Also, glBitmap now re-uses the vertex buffer to avoid frequent allocations/ deallocations. And, use u_simple_shaders utility code. --- src/mesa/sources | 1 + src/mesa/state_tracker/st_cb_drawpixels.c | 295 +----------------------------- src/mesa/state_tracker/st_cb_drawpixels.h | 4 - src/mesa/state_tracker/st_context.c | 7 +- src/mesa/state_tracker/st_context.h | 3 + 5 files changed, 16 insertions(+), 294 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 287af7121a..d109dce5bc 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -172,6 +172,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_atom_texture.c \ state_tracker/st_atom_viewport.c \ state_tracker/st_cb_accum.c \ + state_tracker/st_cb_bitmap.c \ state_tracker/st_cb_blit.c \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_cb_clear.c \ diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 99d5e3e848..b4cd93cd54 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -56,6 +56,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "util/p_tile.h" +#include "util/u_draw_quad.h" #include "shader/prog_instruction.h" #include "cso_cache/cso_context.h" @@ -85,140 +86,6 @@ is_passthrough_program(const struct gl_fragment_program *prog) } -/** - * Make fragment program for glBitmap: - * Sample the texture and kill the fragment if the bit is 0. - * This program will be combined with the user's fragment program. - */ -static struct st_fragment_program * -make_bitmap_fragment_program(GLcontext *ctx) -{ - struct st_fragment_program *stfp; - struct gl_program *p; - GLuint ic = 0; - - p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); - if (!p) - return NULL; - - p->NumInstructions = 5; - - p->Instructions = _mesa_alloc_instructions(p->NumInstructions); - if (!p->Instructions) { - ctx->Driver.DeleteProgram(ctx, p); - return NULL; - } - _mesa_init_instructions(p->Instructions, p->NumInstructions); - - /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */ - p->Instructions[ic].Opcode = OPCODE_TEX; - p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; - p->Instructions[ic].DstReg.Index = 0; - p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; - p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; - p->Instructions[ic].TexSrcUnit = 0; - p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; - ic++; - - /* SWZ tmp0.x, tmp0.x, 1111; # tmp0.x = 1.0 */ - p->Instructions[ic].Opcode = OPCODE_SWZ; - p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; - p->Instructions[ic].DstReg.Index = 0; - p->Instructions[ic].DstReg.WriteMask = WRITEMASK_X; - p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[0].Index = 0; - p->Instructions[ic].SrcReg[0].Swizzle - = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE ); - ic++; - - /* SUB tmp0, tmp0.wwww, tmp0.xxxx; # tmp0.w -= 1 */ - p->Instructions[ic].Opcode = OPCODE_SUB; - p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; - p->Instructions[ic].DstReg.Index = 0; - p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[0].Index = 0; - p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; - p->Instructions[ic].SrcReg[1].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[1].Index = 0; - p->Instructions[ic].SrcReg[1].Swizzle = SWIZZLE_XXXX; /* 1.0 */ - ic++; - - /* KIL if tmp0 < 0 */ - p->Instructions[ic].Opcode = OPCODE_KIL; - p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[0].Index = 0; - ic++; - - /* END; */ - p->Instructions[ic++].Opcode = OPCODE_END; - - assert(ic == p->NumInstructions); - - p->InputsRead = FRAG_BIT_TEX0; - p->OutputsWritten = 0x0; - - stfp = (struct st_fragment_program *) p; - stfp->Base.UsesKill = GL_TRUE; - st_translate_fragment_program(ctx->st, stfp, NULL); - - return stfp; -} - - -/** - * Combine basic bitmap fragment program with the user-defined program. - */ -static struct st_fragment_program * -combined_bitmap_fragment_program(GLcontext *ctx) -{ - struct st_context *st = ctx->st; - struct st_fragment_program *stfp; - - if (!st->bitmap.program) { - /* create the basic bitmap fragment program */ - st->bitmap.program = make_bitmap_fragment_program(ctx); - } - - if (st->bitmap.user_prog_sn == st->fp->serialNo) { - /* re-use */ - stfp = st->bitmap.combined_prog; - } - else { - /* Concatenate the bitmap program with the current user-defined program. - */ - stfp = (struct st_fragment_program *) - _mesa_combine_programs(ctx, - &st->bitmap.program->Base.Base, - &st->fp->Base.Base); - -#if 0 - { - struct gl_program *p = &stfp->Base.Base; - printf("Combined bitmap program:\n"); - _mesa_print_program(p); - printf("InputsRead: 0x%x\n", p->InputsRead); - printf("OutputsWritten: 0x%x\n", p->OutputsWritten); - _mesa_print_parameter_list(p->Parameters); - } -#endif - - /* translate to TGSI tokens */ - st_translate_fragment_program(st, stfp, NULL); - - /* save new program, update serial numbers */ - st->bitmap.user_prog_sn = st->fp->serialNo; - st->bitmap.combined_prog = stfp; - } - - /* Ideally we'd have updated the pipe constants during the normal - * st/atom mechanism. But we can't since this is specific to glBitmap. - */ - st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT); - - return stfp; -} - - /** * Make fragment shader for glDraw/CopyPixels. This shader is made @@ -351,7 +218,7 @@ make_fragment_shader_z(struct st_context *st) * Create a simple vertex shader that just passes through the * vertex position and texcoord (and optionally, color). */ -struct st_vertex_program * +static struct st_vertex_program * st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor) { /* only make programs once and re-use */ @@ -655,14 +522,14 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_save_rasterizer(cso); cso_save_viewport(cso); - /* setup state: just scissor */ + /* rasterizer state: just scissor */ { - struct pipe_rasterizer_state setup; - memset(&setup, 0, sizeof(setup)); + struct pipe_rasterizer_state rasterizer; + memset(&rasterizer, 0, sizeof(rasterizer)); if (ctx->Scissor.Enabled) - setup.scissor = 1; + rasterizer.scissor = 1; - cso_set_rasterizer(cso, &setup); + cso_set_rasterizer(cso, &rasterizer); } /* fragment shader state: TEX lookup program */ @@ -990,153 +857,6 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, -/** - * Create a texture which represents a bitmap image. - */ -static struct pipe_texture * -make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, - const GLubyte *bitmap) -{ - struct pipe_context *pipe = ctx->st->pipe; - struct pipe_screen *screen = pipe->screen; - struct pipe_surface *surface; - uint format = 0, cpp, comp; - ubyte *dest; - struct pipe_texture *pt; - int row, col; - - /* find a texture format we know */ - if (screen->is_format_supported( screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE )) { - format = PIPE_FORMAT_U_I8; - cpp = 1; - comp = 0; - } - else if (screen->is_format_supported( screen, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_TEXTURE )) { - format = PIPE_FORMAT_A8R8G8B8_UNORM; - cpp = 4; - comp = 3; /* alpha channel */ /*XXX little-endian dependency */ - } - else { - /* XXX support more formats */ - assert( 0 ); - } - - /** - * Create a texture. - */ - pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, width, height, - 1, 0); - if (!pt) - return NULL; - - if (unpack->BufferObj && unpack->BufferObj->Name) { - /* - pt->region = buffer_object_region(unpack->BufferObj); - */ - printf("st_Bitmap (sourcing from PBO not implemented yet)\n"); - } - - surface = screen->get_tex_surface(screen, pt, 0, 0, 0); - - /* map texture surface */ - dest = pipe_surface_map(surface); - - /* Put image into texture surface. - * Note that the image is actually going to be upside down in - * the texture. We deal with that with texcoords. - */ - - for (row = 0; row < height; row++) { - const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, - bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); - ubyte *destRow = dest + row * surface->pitch * cpp; - - if (unpack->LsbFirst) { - /* Lsb first */ - GLubyte mask = 1U << (unpack->SkipPixels & 0x7); - for (col = 0; col < width; col++) { - - /* set texel to 255 if bit is set */ - destRow[comp] = (*src & mask) ? 255 : 0; - destRow += cpp; - - if (mask == 128U) { - src++; - mask = 1U; - } - else { - mask = mask << 1; - } - } - - /* get ready for next row */ - if (mask != 1) - src++; - } - else { - /* Msb first */ - GLubyte mask = 128U >> (unpack->SkipPixels & 0x7); - for (col = 0; col < width; col++) { - - /* set texel to 255 if bit is set */ - destRow[comp] =(*src & mask) ? 255 : 0; - destRow += cpp; - - if (mask == 1U) { - src++; - mask = 128U; - } - else { - mask = mask >> 1; - } - } - - /* get ready for next row */ - if (mask != 128) - src++; - } - - } /* row */ - - /* Release surface */ - pipe_surface_unmap(surface); - pipe_surface_reference(&surface, NULL); - pipe->texture_update(pipe, pt, 0, 0x1); - - pt->format = format; - - return pt; -} - - - -static void -st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap ) -{ - struct st_fragment_program *stfp; - struct st_vertex_program *stvp; - struct st_context *st = ctx->st; - struct pipe_texture *pt; - - stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE); - stfp = combined_bitmap_fragment_program(ctx); - - st_validate_state(st); - - pt = make_bitmap_texture(ctx, width, height, unpack, bitmap); - if (pt) { - draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2], - width, height, 1.0, 1.0, - pt, stvp, stfp, - ctx->Current.RasterColor, GL_FALSE); - - pipe_texture_reference(&pt, NULL); - } -} - - static void copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, GLsizei width, GLsizei height, @@ -1337,5 +1057,4 @@ void st_init_drawpixels_functions(struct dd_function_table *functions) { functions->DrawPixels = st_DrawPixels; functions->CopyPixels = st_CopyPixels; - functions->Bitmap = st_Bitmap; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h index b8b906f06b..71ba487020 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.h +++ b/src/mesa/state_tracker/st_cb_drawpixels.h @@ -30,10 +30,6 @@ #define ST_CB_DRAWPIXELS_H -extern struct st_vertex_program * -st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor); - - extern void st_init_drawpixels_functions(struct dd_function_table *functions); diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index b1681be2eb..d9e8722976 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -35,8 +35,9 @@ #include "st_public.h" #include "st_context.h" #include "st_cb_accum.h" -#include "st_cb_bufferobjects.h" +#include "st_cb_bitmap.h" #include "st_cb_blit.h" +#include "st_cb_bufferobjects.h" #include "st_cb_clear.h" #include "st_cb_drawpixels.h" #include "st_cb_fbo.h" @@ -154,6 +155,7 @@ static void st_destroy_context_priv( struct st_context *st ) st_destroy_atoms( st ); st_destroy_draw( st ); st_destroy_generate_mipmap(st); + st_destroy_bitmap(st); st_destroy_blit(st); st_destroy_clear(st); @@ -221,8 +223,9 @@ void st_init_driver_functions(struct dd_function_table *functions) _mesa_init_glsl_driver_functions(functions); st_init_accum_functions(functions); - st_init_bufferobject_functions(functions); + st_init_bitmap_functions(functions); st_init_blit_functions(functions); + st_init_bufferobject_functions(functions); st_init_clear_functions(functions); st_init_drawpixels_functions(functions); st_init_fbo_functions(functions); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 2563c7fed0..2d37086799 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -147,6 +147,9 @@ struct st_context struct st_fragment_program *program; /**< bitmap tex/kil program */ GLuint user_prog_sn; /**< user fragment program serial no. */ struct st_fragment_program *combined_prog; + void *vs; + float vertices[4][3][4]; /**< vertex pos + color + texcoord */ + struct pipe_buffer *vbuf; } bitmap; /** for glClear */ -- cgit v1.2.3 From 80567f9c9f84d9df1cdb2d91a3c0814888cc5d08 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 15:01:44 -0600 Subject: gallium: added width, height to pipe_framebuffer_state --- src/gallium/include/pipe/p_state.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 5791a10119..3e531c4da4 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -220,6 +220,8 @@ struct pipe_blend_color struct pipe_framebuffer_state { + unsigned width, height; + /** multiple colorbuffers for multiple render targets */ unsigned num_cbufs; struct pipe_surface *cbufs[PIPE_MAX_COLOR_BUFS]; -- cgit v1.2.3 From f3d23aa787bb08bad7638540b128fa0e65110246 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 15:02:10 -0600 Subject: gallium: unused var silence warning --- src/mesa/state_tracker/st_debug.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index 8b9b91aa34..23ecfff0aa 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -50,9 +50,10 @@ st_print_current(void) { GET_CURRENT_CONTEXT(ctx); struct st_context *st = ctx->st; - int i; #if 0 + int i; + printf("Vertex Transform Inputs:\n"); for (i = 0; i < st->vp->state.num_inputs; i++) { printf(" Slot %d: VERT_ATTRIB_%d\n", i, st->vp->index_to_input[i]); -- cgit v1.2.3 From 3a4ece89a799a7c699be229a6d965d4820af0196 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 15:02:32 -0600 Subject: gallium: assign framebuffer width, height --- src/mesa/state_tracker/st_atom_framebuffer.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 43259c3ecb..02573af8f0 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -52,6 +52,9 @@ update_framebuffer_state( struct st_context *st ) memset(framebuffer, 0, sizeof(*framebuffer)); + framebuffer->width = fb->Width; + framebuffer->height = fb->Height; + /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state * to determine which surfaces to draw to */ -- cgit v1.2.3 From 2fb30b77ad09016efcf969456de9b0341bc53bac Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 15:02:59 -0600 Subject: cell: use pipe_framebuffer_state.width, height --- src/gallium/drivers/cell/ppu/cell_state_derived.c | 13 ++----------- src/gallium/drivers/cell/ppu/cell_state_emit.c | 4 ++-- 2 files changed, 4 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_derived.c b/src/gallium/drivers/cell/ppu/cell_state_derived.c index 5c240a55c0..5480534ad9 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_derived.c +++ b/src/gallium/drivers/cell/ppu/cell_state_derived.c @@ -141,17 +141,8 @@ calculate_vertex_layout( struct cell_context *cell ) static void compute_cliprect(struct cell_context *sp) { - unsigned surfWidth, surfHeight; - - if (sp->framebuffer.num_cbufs > 0) { - surfWidth = sp->framebuffer.cbufs[0]->width; - surfHeight = sp->framebuffer.cbufs[0]->height; - } - else { - /* no surface? */ - surfWidth = sp->scissor.maxx; - surfHeight = sp->scissor.maxy; - } + uint surfWidth = sp->framebuffer.width; + uint surfHeight = sp->framebuffer.height; if (sp->rasterizer->scissor) { /* clip to scissor rect */ diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 4d589bcdbf..31cc938f07 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -60,8 +60,8 @@ cell_emit_state(struct cell_context *cell) fb->color_format = cbuf->format; fb->depth_start = cell->zsbuf_map; fb->depth_format = zbuf ? zbuf->format : PIPE_FORMAT_NONE; - fb->width = cell->framebuffer.cbufs[0]->width; - fb->height = cell->framebuffer.cbufs[0]->height; + fb->width = cell->framebuffer.width; + fb->height = cell->framebuffer.height; } if (cell->dirty & CELL_NEW_BLEND) { -- cgit v1.2.3 From 09f67990abd4bb9b79349be2fca9a6ae850b6f5f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 15:03:43 -0600 Subject: gallium: use new framebuffer width, height fields --- src/gallium/drivers/softpipe/sp_context.h | 2 -- src/gallium/drivers/softpipe/sp_prim_setup.c | 2 +- src/gallium/drivers/softpipe/sp_quad_stipple.c | 2 +- src/gallium/drivers/softpipe/sp_state_derived.c | 4 ++-- src/gallium/drivers/softpipe/sp_state_surface.c | 19 +++---------------- 5 files changed, 7 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 31d7062dcc..19e6cfaf02 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -84,8 +84,6 @@ struct softpipe_context { unsigned num_samplers; unsigned num_textures; - uint fb_width, fb_height; - /* Counter for occlusion queries. Note this supports overlapping * queries. */ diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index f2d6043e2e..6a81e4d8cc 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -485,7 +485,7 @@ setup_fragcoord_coeff(struct setup_stage *setup, uint slot) /*Y*/ if (setup->softpipe->rasterizer->origin_lower_left) { /* y=0=bottom */ - const int winHeight = setup->softpipe->fb_height; + const int winHeight = setup->softpipe->framebuffer.height; setup->coef[slot].a0[1] = (float) (winHeight - 1); setup->coef[slot].dady[1] = -1.0; } diff --git a/src/gallium/drivers/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c index 9a39249576..f1e9b80e09 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stipple.c +++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c @@ -25,7 +25,7 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad) int y0, y1; uint stipple0, stipple1; if (softpipe->rasterizer->origin_lower_left) { - y0 = softpipe->fb_height - 1 - quad->y0; + y0 = softpipe->framebuffer.height - 1 - quad->y0; y1 = y0 - 1; } else { diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 53b2056b8a..14abb20eeb 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -171,8 +171,8 @@ softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe) static void compute_cliprect(struct softpipe_context *sp) { - uint surfWidth = sp->fb_width; - uint surfHeight = sp->fb_height; + uint surfWidth = sp->framebuffer.width; + uint surfHeight = sp->framebuffer.height; if (sp->rasterizer->scissor) { /* clip to scissor rect */ diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c index 2f55684464..ba8c9eece7 100644 --- a/src/gallium/drivers/softpipe/sp_state_surface.c +++ b/src/gallium/drivers/softpipe/sp_state_surface.c @@ -48,9 +48,6 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, struct softpipe_context *sp = softpipe_context(pipe); uint i; - /* updated below */ - sp->fb_width = sp->fb_height = 0; - for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { /* check if changing cbuf */ if (sp->framebuffer.cbufs[i] != fb->cbufs[i]) { @@ -63,10 +60,6 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* update cache */ sp_tile_cache_set_surface(sp->cbuf_cache[i], fb->cbufs[i]); } - if (fb->cbufs[i]) { - sp->fb_width = fb->cbufs[i]->width; - sp->fb_height = fb->cbufs[i]->height; - } } sp->framebuffer.num_cbufs = fb->num_cbufs; @@ -81,11 +74,6 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, /* update cache */ sp_tile_cache_set_surface(sp->zsbuf_cache, fb->zsbuf); - - if (!sp->fb_width && fb->zsbuf) { - sp->fb_width = fb->zsbuf->width; - sp->fb_height = fb->zsbuf->height; - } } #if 0 @@ -113,9 +101,8 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe, } #endif + sp->framebuffer.width = fb->width; + sp->framebuffer.height = fb->height; + sp->dirty |= SP_NEW_FRAMEBUFFER; } - - - - -- cgit v1.2.3 From 00cf178d93e5bc36a88a9f8ff444f60c493be14d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 15:21:55 -0600 Subject: gallium: enable vp input semantic info --- src/gallium/auxiliary/util/u_simple_shaders.c | 15 ++++++--------- 1 file changed, 6 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index 8cb74e81d2..d230ddfbeb 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -84,16 +84,15 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, /* declare inputs */ for (i = 0; i < num_attribs; i++) { - decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_INPUT; - /* + decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; - decl.Semantic.SemanticIndex = 0; - */ + decl.Semantic.SemanticName = semantic_names[i]; + decl.Semantic.SemanticIndex = semantic_indexes[i]; + decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; + decl.u.DeclarationRange.Last = i; ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, @@ -102,19 +101,17 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, /* declare outputs */ for (i = 0; i < num_attribs; i++) { - decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_OUTPUT; decl.Declaration.Semantic = 1; decl.Semantic.SemanticName = semantic_names[i]; decl.Semantic.SemanticIndex = semantic_indexes[i]; decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; + decl.u.DeclarationRange.Last = i; ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, maxTokens - ti); - } /* emit MOV instructions */ -- cgit v1.2.3 From 0565e6888a332956661f6bc8b5778b058168e5f9 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 20 Mar 2008 15:22:20 -0600 Subject: gallium: set fb.width/height --- src/gallium/auxiliary/util/u_blit.c | 2 ++ src/gallium/auxiliary/util/u_gen_mipmap.c | 2 ++ 2 files changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index bb316554a4..28a404fd01 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -316,6 +316,8 @@ util_blit_pixels(struct blit_state *ctx, /* drawing dest */ memset(&fb, 0, sizeof(fb)); + fb.width = dst->width; + fb.height = dst->height; fb.num_cbufs = 1; fb.cbufs[0] = dst; cso_set_framebuffer(ctx->cso, &fb); diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index b84752e0b4..cf02f00b1b 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -901,6 +901,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, * Setup framebuffer / dest surface */ fb.cbufs[0] = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); + fb.width = pt->width[dstLevel]; + fb.height = pt->height[dstLevel]; cso_set_framebuffer(ctx->cso, &fb); /* -- cgit v1.2.3 From 3b3774b1227743147159676795b542c0eb7c2bdf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 20 Mar 2008 17:08:07 -0600 Subject: gallium: catch some out of memory conditions in the texture image code. st_finalize_texture()'s return code now indicates success/fail instead of presence of texture border (which we discard earlier). --- src/mesa/state_tracker/st_atom_texture.c | 5 ++++- src/mesa/state_tracker/st_cb_texture.c | 13 +++++++++---- 2 files changed, 13 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 2a711e513d..9aef30f456 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -64,7 +64,10 @@ update_textures(struct st_context *st) GLboolean flush, retval; retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); - /* XXX retval indicates whether there's a texture border */ + if (!retval) { + /* out of mem */ + continue; + } st->state.num_textures = unit + 1; } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 306b27c423..a6c1a35355 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -564,7 +564,8 @@ st_TexImage(GLcontext * ctx, if (!stObj->pt) { guess_and_alloc_texture(ctx->st, stObj, stImage); if (!stObj->pt) { - DBG("guess_and_alloc_texture: failed\n"); + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + return; } } @@ -1379,7 +1380,7 @@ copy_image_data_to_texture(struct st_context *st, /** * Called during state validation. When this function is finished, * the texture object should be ready for rendering. - * \return GL_FALSE if a texture border is present, GL_TRUE otherwise + * \return GL_TRUE for success, GL_FALSE for failure (out of mem) */ GLboolean st_finalize_texture(GLcontext *ctx, @@ -1405,6 +1406,7 @@ st_finalize_texture(GLcontext *ctx, calculate_first_last_level(stObj); firstImage = st_texture_image(stObj->base.Image[0][stObj->base.BaseLevel]); +#if 0 /* Fallback case: */ if (firstImage->base.Border) { @@ -1413,7 +1415,7 @@ st_finalize_texture(GLcontext *ctx, } return GL_FALSE; } - +#endif /* If both firstImage and stObj point to a texture which can contain * all active images, favour firstImage. Note that because of the @@ -1466,6 +1468,10 @@ st_finalize_texture(GLcontext *ctx, firstImage->base.Height, firstImage->base.Depth, comp_byte); + if (!stObj->pt) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + return GL_FALSE; + } } /* Pull in any images not in the object's texture: @@ -1486,7 +1492,6 @@ st_finalize_texture(GLcontext *ctx, } } - return GL_TRUE; } -- cgit v1.2.3 From 2b21bde3b1fa6fe357a3a5adc6249e89d6915524 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 19 Mar 2008 17:29:39 -0700 Subject: cell: Use code-gen for alpha blend So far this is only tested when GL_BLEND is disabled. --- src/gallium/drivers/cell/common.h | 10 +++++ src/gallium/drivers/cell/ppu/cell_pipe_state.c | 11 +++-- src/gallium/drivers/cell/ppu/cell_state_emit.c | 17 +++++-- src/gallium/drivers/cell/spu/Makefile | 1 - src/gallium/drivers/cell/spu/spu_blend.c | 62 -------------------------- src/gallium/drivers/cell/spu/spu_blend.h | 37 --------------- src/gallium/drivers/cell/spu/spu_main.c | 50 ++++++++++++++++++--- src/gallium/drivers/cell/spu/spu_main.h | 17 ++++++- src/gallium/drivers/cell/spu/spu_tri.c | 56 +++++++++++++++-------- 9 files changed, 129 insertions(+), 132 deletions(-) delete mode 100644 src/gallium/drivers/cell/spu/spu_blend.c delete mode 100644 src/gallium/drivers/cell/spu/spu_blend.h (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index fe93fd8e1a..d59e4f7036 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -114,6 +114,16 @@ struct cell_command_depth_stencil_alpha_test { }; +/** + * Upload code to perform framebuffer blend operation + */ +struct cell_command_blend { + uint64_t base; /**< Effective address of code start. */ + unsigned size; /**< Size in bytes of test code. */ + unsigned read_fb; /**< Flag: should framebuffer be read? */ +}; + + /** * Tell SPUs about the framebuffer size, location */ diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index c880760e4b..4ca8c153b0 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -54,14 +54,19 @@ cell_create_blend_state(struct pipe_context *pipe, static void -cell_bind_blend_state(struct pipe_context *pipe, void *blend) +cell_bind_blend_state(struct pipe_context *pipe, void *state) { struct cell_context *cell = cell_context(pipe); + struct cell_blend_state *blend = (struct cell_blend_state *) state; + draw_flush(cell->draw); - cell->blend = (const struct cell_blend_state *)blend; + if ((blend != NULL) && (blend->code.store == NULL)) { + cell_generate_depth_stencil_test(blend); + } + cell->blend = blend; cell->dirty |= CELL_NEW_BLEND; } @@ -70,7 +75,7 @@ static void cell_delete_blend_state(struct pipe_context *pipe, void *blend) { struct cell_blend_state *cb = (struct cell_blend_state *) blend; - + spe_release_func(& cb->code); FREE(cb); } diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 31cc938f07..5709b48f12 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -65,9 +65,20 @@ cell_emit_state(struct cell_context *cell) } if (cell->dirty & CELL_NEW_BLEND) { - emit_state_cmd(cell, CELL_CMD_STATE_BLEND, - cell->blend, - sizeof(struct pipe_blend_state)); + struct cell_command_blend blend; + + if (cell->blend != NULL) { + blend.base = (intptr_t) cell->blend->code.store; + blend.size = (char *) cell->blend->code.csr + - (char *) cell->blend->code.store; + blend.read_fb = TRUE; + } else { + blend.base = 0; + blend.size = 0; + blend.read_fb = FALSE; + } + + emit_state_cmd(cell, CELL_CMD_STATE_BLEND, &blend, sizeof(blend)); } if (cell->dirty & CELL_NEW_DEPTH_STENCIL) { diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile index 115ca8cd90..8e83610790 100644 --- a/src/gallium/drivers/cell/spu/Makefile +++ b/src/gallium/drivers/cell/spu/Makefile @@ -17,7 +17,6 @@ PROG_SPU_EMBED_O = $(PROG)_spu-embed.o SOURCES = \ spu_main.c \ - spu_blend.c \ spu_dcache.c \ spu_per_fragment_op.c \ spu_render.c \ diff --git a/src/gallium/drivers/cell/spu/spu_blend.c b/src/gallium/drivers/cell/spu/spu_blend.c deleted file mode 100644 index 23ec0eeb45..0000000000 --- a/src/gallium/drivers/cell/spu/spu_blend.c +++ /dev/null @@ -1,62 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "spu_main.h" -#include "spu_blend.h" -#include "spu_colorpack.h" - - -void -blend_quad(uint itx, uint ity, vector float colors[4]) -{ - /* simple SRC_ALPHA, ONE_MINUS_SRC_ALPHA blending */ - vector float fbc00 = spu_unpack_color(spu.ctile.ui[ity][itx]); - vector float fbc01 = spu_unpack_color(spu.ctile.ui[ity][itx+1]); - vector float fbc10 = spu_unpack_color(spu.ctile.ui[ity+1][itx]); - vector float fbc11 = spu_unpack_color(spu.ctile.ui[ity+1][itx+1]); - - vector float alpha00 = spu_splats(spu_extract(colors[0], 3)); - vector float alpha01 = spu_splats(spu_extract(colors[1], 3)); - vector float alpha10 = spu_splats(spu_extract(colors[2], 3)); - vector float alpha11 = spu_splats(spu_extract(colors[3], 3)); - - vector float one_minus_alpha00 = spu_sub(spu_splats(1.0f), alpha00); - vector float one_minus_alpha01 = spu_sub(spu_splats(1.0f), alpha01); - vector float one_minus_alpha10 = spu_sub(spu_splats(1.0f), alpha10); - vector float one_minus_alpha11 = spu_sub(spu_splats(1.0f), alpha11); - - colors[0] = spu_add(spu_mul(colors[0], alpha00), - spu_mul(fbc00, one_minus_alpha00)); - colors[1] = spu_add(spu_mul(colors[1], alpha01), - spu_mul(fbc01, one_minus_alpha01)); - colors[2] = spu_add(spu_mul(colors[2], alpha10), - spu_mul(fbc10, one_minus_alpha10)); - colors[3] = spu_add(spu_mul(colors[3], alpha11), - spu_mul(fbc11, one_minus_alpha11)); -} - diff --git a/src/gallium/drivers/cell/spu/spu_blend.h b/src/gallium/drivers/cell/spu/spu_blend.h deleted file mode 100644 index 2b594b578b..0000000000 --- a/src/gallium/drivers/cell/spu/spu_blend.h +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef SPU_BLEND_H -#define SPU_BLEND_H - - -extern void -blend_quad(uint itx, uint ity, vector float colors[4]); - - -#endif /* SPU_BLEND_H */ diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 937962285d..41bebf5362 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -61,6 +61,25 @@ static unsigned char attribute_fetch_code_buffer[136 * PIPE_ATTRIB_MAX] static unsigned char depth_stencil_code_buffer[4 * 64] ALIGN16_ATTRIB; +static unsigned char fb_blend_code_buffer[4 * 64] + ALIGN16_ATTRIB; + +static struct spu_blend_results +default_blend(qword frag_r, qword frag_g, qword frag_b, qword frag_a, + qword pixel_r, qword pixel_g, qword pixel_b, qword pixel_a, + qword frag_mask) +{ + struct spu_blend_results result; + + result.r = si_selb(pixel_r, frag_r, frag_mask); + result.g = si_selb(pixel_g, frag_g, frag_mask); + result.b = si_selb(pixel_b, frag_b, frag_mask); + result.a = si_selb(pixel_a, frag_a, frag_mask); + + return result; +} + + /** * Tell the PPU that this SPU has finished copying a buffer to * local store and that it may be reused by the PPU. @@ -246,14 +265,31 @@ cmd_state_framebuffer(const struct cell_command_framebuffer *cmd) static void -cmd_state_blend(const struct pipe_blend_state *state) +cmd_state_blend(const struct cell_command_blend *state) { if (Debug) printf("SPU %u: BLEND: enabled %d\n", spu.init.id, - state->blend_enable); + (state->size != 0)); + + ASSERT_ALIGN16(state->base); - memcpy(&spu.blend, state, sizeof(*state)); + if (state->size != 0) { + mfc_get(fb_blend_code_buffer, + (unsigned int) state->base, /* src */ + ROUNDUP16(state->size), + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); + spu.blend = (blend_func) fb_blend_code_buffer; + spu.read_fb = state->read_fb; + } else { + /* If there is no code, use the default; + */ + spu.blend = default_blend; + spu.read_fb = FALSE; + } } @@ -441,9 +477,8 @@ cmd_batch(uint opcode) pos += 1; break; case CELL_CMD_STATE_BLEND: - cmd_state_blend((struct pipe_blend_state *) - &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct pipe_blend_state)) / 8); + cmd_state_blend((struct cell_command_blend *) &buffer[pos+1]); + pos += (1 + ROUNDUP8(sizeof(struct cell_command_blend)) / 8); break; case CELL_CMD_STATE_DEPTH_STENCIL: cmd_state_depth_stencil((struct cell_command_depth_stencil_alpha_test *) @@ -587,6 +622,9 @@ one_time_init(void) memset(spu.ctile_status, TILE_STATUS_DEFINED, sizeof(spu.ctile_status)); memset(spu.ztile_status, TILE_STATUS_DEFINED, sizeof(spu.ztile_status)); invalidate_tex_cache(); + + spu.blend = default_blend; + spu.read_fb = FALSE; } diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index 444e218645..56d0968676 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -67,6 +67,18 @@ typedef struct spu_frag_test_results (*frag_test_func)(qword frag_mask, qword frag_alpha, qword facing); +struct spu_blend_results { + qword r; + qword g; + qword b; + qword a; +}; + +typedef struct spu_blend_results (*blend_func)( + qword frag_r, qword frag_g, qword frag_b, qword frag_a, + qword pixel_r, qword pixel_g, qword pixel_b, qword pixel_a, + qword frag_mask); + struct spu_framebuffer { void *color_start; /**< addr of color surface in main memory */ void *depth_start; /**< addr of depth surface in main memory */ @@ -93,7 +105,10 @@ struct spu_global boolean read_depth; boolean read_stencil; frag_test_func frag_test; - struct pipe_blend_state blend; + + boolean read_fb; + blend_func blend; + struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct cell_command_texture texture; diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index 81823f2463..c9f8cadcda 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -29,10 +29,10 @@ * Triangle rendering within a tile. */ +#include #include "pipe/p_compiler.h" #include "pipe/p_format.h" #include "pipe/p_util.h" -#include "spu_blend.h" #include "spu_colorpack.h" #include "spu_main.h" #include "spu_texture.h" @@ -326,27 +326,45 @@ emit_quad( int x, int y, mask_t mask ) eval_coeff(1, (float) x, (float) y, colors); } -#if 1 - if (spu.blend.blend_enable) - blend_quad(ix % TILE_SIZE, iy % TILE_SIZE, colors); -#endif - if (spu_extract(mask, 0)) - spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0], shuffle); - if (spu_extract(mask, 1)) - spu.ctile.ui[iy][ix+1] = spu_pack_color_shuffle(colors[1], shuffle); - if (spu_extract(mask, 2)) - spu.ctile.ui[iy+1][ix] = spu_pack_color_shuffle(colors[2], shuffle); - if (spu_extract(mask, 3)) - spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(colors[3], shuffle); + /* Read the current framebuffer values. + * + * Ignore read_fb for now. In the future we can use this to avoid + * reading the framebuffer if read_fb is false and the fragment mask is + * all 0xffffffff. This is the common case, so it is probably worth + * the effort. We'll have to profile to determine whether or not the + * extra conditional branches hurt overall performance. + */ + vec_float4 aos_pix[4] = { + spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+0]), + spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+1]), + spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+0]), + spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+1]), + }; -#if 0 - /* SIMD_Z with swizzled color buffer (someday) */ - vector unsigned int uicolors = *((vector unsigned int *) &colors); - spu.ctile.ui4[iy/2][ix/2] = spu_sel(spu.ctile.ui4[iy/2][ix/2], uicolors, mask); -#endif - } + qword soa_pix[4]; + qword soa_frag[4]; + /* Convert pixel and fragment data from AoS to SoA format. + */ + _transpose_matrix4x4((vec_float4 *) soa_pix, aos_pix); + _transpose_matrix4x4((vec_float4 *) soa_frag, colors); + + const struct spu_blend_results result = + (*spu.blend)(soa_frag[0], soa_frag[1], soa_frag[2], soa_frag[3], + soa_pix[0], soa_pix[1], soa_pix[2], soa_pix[3], + (qword) mask); + + + /* Convert final pixel data from SoA to AoS format. + */ + _transpose_matrix4x4(aos_pix, (const vec_float4 *) &result); + + spu.ctile.ui[iy+0][ix+0] = spu_pack_color_shuffle(aos_pix[0], shuffle); + spu.ctile.ui[iy+0][ix+1] = spu_pack_color_shuffle(aos_pix[1], shuffle); + spu.ctile.ui[iy+1][ix+0] = spu_pack_color_shuffle(aos_pix[2], shuffle); + spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(aos_pix[3], shuffle); + } #endif } -- cgit v1.2.3 From df1d6e2410dbc6af66ca416124587918b9764ee8 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 20 Mar 2008 17:36:05 -0700 Subject: cell: Fix bus error when there is no depth buffer --- src/gallium/drivers/cell/spu/spu_tri.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index c9f8cadcda..c4272d6e93 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -261,6 +261,9 @@ do_depth_test(int x, int y, mask_t quadmask) float4 zvals; mask_t mask; + if (spu.fb.depth_format == PIPE_FORMAT_NONE) + return quadmask; + zvals.v = eval_z((float) x, (float) y); mask = (mask_t) spu_do_depth_stencil(x - setup.cliprect_minx, -- cgit v1.2.3 From 4b9520fc05f169b74835c096c933d67c67c6d8cd Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 20 Mar 2008 17:36:31 -0700 Subject: cell: Call the correct function to generate blending code Cut-and-paste for the lose. :( --- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 4ca8c153b0..86fcdcff1f 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -63,7 +63,7 @@ cell_bind_blend_state(struct pipe_context *pipe, void *state) draw_flush(cell->draw); if ((blend != NULL) && (blend->code.store == NULL)) { - cell_generate_depth_stencil_test(blend); + cell_generate_alpha_blend(blend, &cell->blend_color); } cell->blend = blend; -- cgit v1.2.3 From cab68957c72d6f198546a250b6fe0a74732cb3ec Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Thu, 20 Mar 2008 18:17:48 -0700 Subject: cell: Fix several bugs in blend code-gen - Alpha factor set to a _COLOR mode was mishandled - Cases when either dst factor or src factor was ZERO were mishandled - MIN and MAX cases were backwards - Case when blend was disabled was mishandled - Incorrect comments about number of instructions generated The tests blendminmax and blendsquare run correctly. --- .../drivers/cell/ppu/cell_state_per_fragment.c | 112 +++++++++++++++++---- 1 file changed, 92 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c index 9c47968459..988c251e20 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -669,7 +669,7 @@ emit_alpha_factor_calculation(struct spe_function *f, /** - * \note Emits a maximum of 5 instructions + * \note Emits a maximum of 6 instructions */ static void emit_color_factor_calculation(struct spe_function *f, @@ -864,7 +864,11 @@ emit_blend_calculation(struct spe_function *f, spe_il(f, src, 0); } else if (dF == PIPE_BLENDFACTOR_ONE) { spe_or(f, src, dst, dst); + } else { + spe_fm(f, src, dst, dst_factor); } + } else if (dF == PIPE_BLENDFACTOR_ZERO) { + spe_fm(f, src, src, src_factor); } else { spe_fm(f, tmp, dst, dst_factor); spe_fma(f, src, src, src_factor, tmp); @@ -884,7 +888,11 @@ emit_blend_calculation(struct spe_function *f, } else if (dF == PIPE_BLENDFACTOR_ONE) { spe_il(f, tmp, 0); spe_fs(f, src, tmp, dst); + } else { + spe_fm(f, src, dst, dst_factor); } + } else if (dF == PIPE_BLENDFACTOR_ZERO) { + spe_fm(f, src, src, src_factor); } else { spe_fm(f, tmp, dst, dst_factor); spe_fms(f, src, src, src_factor, tmp); @@ -904,7 +912,11 @@ emit_blend_calculation(struct spe_function *f, spe_il(f, src, 0); } else if (dF == PIPE_BLENDFACTOR_ONE) { spe_or(f, src, dst, dst); + } else { + spe_fm(f, src, dst, dst_factor); } + } else if (dF == PIPE_BLENDFACTOR_ZERO) { + spe_fm(f, src, src, src_factor); } else { spe_fm(f, tmp, src, src_factor); spe_fms(f, src, src, dst_factor, tmp); @@ -913,12 +925,12 @@ emit_blend_calculation(struct spe_function *f, case PIPE_BLEND_MIN: spe_cgt(f, tmp, src, dst); - spe_selb(f, src, dst, src, tmp); + spe_selb(f, src, src, dst, tmp); break; case PIPE_BLEND_MAX: spe_cgt(f, tmp, src, dst); - spe_selb(f, src, src, dst, tmp); + spe_selb(f, src, dst, src, tmp); break; default: @@ -940,9 +952,9 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, struct spe_function *const f = &cb->code; /* This code generates a maximum of 3 (source alpha factor) - * + 3 (destination alpha factor) + (3 * 5) (source color factor) - * + (3 * 5) (destination color factor) + (4 * 2) (blend equation) - * + 4 (fragment mask) + 1 (return) = 49 instlructions. Round up to 64 to + * + 3 (destination alpha factor) + (3 * 6) (source color factor) + * + (3 * 6) (destination color factor) + (4 * 2) (blend equation) + * + 4 (fragment mask) + 1 (return) = 55 instlructions. Round up to 64 to * make it a happy power-of-two. */ spe_init_func(f, 4 * 64); @@ -984,16 +996,57 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, && (b->alpha_func != PIPE_BLEND_MAX); - sF[0] = b->rgb_src_factor; - sF[1] = sF[0]; - sF[2] = sF[0]; - sF[3] = (b->alpha_src_factor == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE) - ? PIPE_BLENDFACTOR_ONE : b->alpha_src_factor; + if (b->blend_enable) { + sF[0] = b->rgb_src_factor; + sF[1] = sF[0]; + sF[2] = sF[0]; + switch (b->alpha_src_factor & 0x0f) { + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + sF[3] = PIPE_BLENDFACTOR_ONE; + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + case PIPE_BLENDFACTOR_DST_COLOR: + case PIPE_BLENDFACTOR_CONST_COLOR: + case PIPE_BLENDFACTOR_SRC1_COLOR: + sF[3] = b->alpha_src_factor + 1; + break; + default: + sF[3] = b->alpha_src_factor; + } - dF[0] = b->rgb_dst_factor; - dF[1] = dF[0]; - dF[2] = dF[0]; - dF[3] = b->rgb_dst_factor; + dF[0] = b->rgb_dst_factor; + dF[1] = dF[0]; + dF[2] = dF[0]; + switch (b->alpha_dst_factor & 0x0f) { + case PIPE_BLENDFACTOR_SRC_COLOR: + case PIPE_BLENDFACTOR_DST_COLOR: + case PIPE_BLENDFACTOR_CONST_COLOR: + case PIPE_BLENDFACTOR_SRC1_COLOR: + dF[3] = b->alpha_dst_factor + 1; + break; + default: + dF[3] = b->alpha_dst_factor; + } + + func[0] = b->rgb_func; + func[1] = func[0]; + func[2] = func[0]; + func[3] = b->alpha_func; + } else { + sF[0] = PIPE_BLENDFACTOR_ONE; + sF[1] = PIPE_BLENDFACTOR_ONE; + sF[2] = PIPE_BLENDFACTOR_ONE; + sF[3] = PIPE_BLENDFACTOR_ONE; + dF[0] = PIPE_BLENDFACTOR_ZERO; + dF[1] = PIPE_BLENDFACTOR_ZERO; + dF[2] = PIPE_BLENDFACTOR_ZERO; + dF[3] = PIPE_BLENDFACTOR_ZERO; + + func[0] = PIPE_BLEND_ADD; + func[1] = PIPE_BLEND_ADD; + func[2] = PIPE_BLEND_ADD; + func[3] = PIPE_BLEND_ADD; + } /* If alpha writing is enabled and the alpha blend mode requires use of @@ -1054,11 +1107,6 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, - func[0] = b->rgb_func; - func[1] = func[0]; - func[2] = func[0]; - func[3] = b->alpha_func; - for (i = 0; i < 4; ++i) { if ((b->colormask & (1U << i)) != 0) { emit_blend_calculation(f, @@ -1072,4 +1120,28 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, } spe_bi(f, 0, 0, 0); + +#if 0 + { + const uint32_t *p = f->store; + + printf("# %u instructions\n", f->csr - f->store); + printf("# blend (%sabled)\n", + (cb->base.blend_enable) ? "en" : "dis"); + printf("# RGB func / sf / df: %u %u %u\n", + cb->base.rgb_func, + cb->base.rgb_src_factor, + cb->base.rgb_dst_factor); + printf("# ALP func / sf / df: %u %u %u\n", + cb->base.alpha_func, + cb->base.alpha_src_factor, + cb->base.alpha_dst_factor); + + printf("\t.text\n"); + for (/* empty */; p < f->csr; p++) { + printf("\t.long\t0x%04x\n", *p); + } + fflush(stdout); + } +#endif } -- cgit v1.2.3 From 23c98f618b4502222933e0f0aa8f43941f2ccfad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 21 Mar 2008 12:14:06 +0000 Subject: gallium: Update scons build for state tracker files added/removed. --- src/mesa/SConscript | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index 678e4ad0e8..b816a21b38 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -142,6 +142,8 @@ STATETRACKER_SOURCES = [ 'state_tracker/st_atom_texture.c', 'state_tracker/st_atom_viewport.c', 'state_tracker/st_cb_accum.c', + 'state_tracker/st_cb_bitmap.c', + 'state_tracker/st_cb_blit.c', 'state_tracker/st_cb_bufferobjects.c', 'state_tracker/st_cb_clear.c', 'state_tracker/st_cb_flush.c', @@ -154,7 +156,6 @@ STATETRACKER_SOURCES = [ 'state_tracker/st_cb_readpixels.c', 'state_tracker/st_cb_strings.c', 'state_tracker/st_cb_texture.c', - 'state_tracker/st_cache.c', 'state_tracker/st_context.c', 'state_tracker/st_debug.c', 'state_tracker/st_draw.c', -- cgit v1.2.3 From 403c7ba91a72f4740ab64823172c8760d837770d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 21 Mar 2008 10:38:56 -0600 Subject: gallium: additional comments, fix typos, etc --- src/gallium/include/pipe/p_format.h | 31 +++++++++++++++++++++++-------- 1 file changed, 23 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index f90087b3c9..d09b6dbefc 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -38,13 +38,14 @@ extern "C" { #endif /** - * The PIPE_FORMAT is a 32-bit wide bitfield that encodes all the information - * needed to uniquely describe a pixel format. + * The pipe_format enum is a 32-bit wide bitfield that encodes all the + * information needed to uniquely describe a pixel format. */ /** - * Possible format layouts -- occupy first 2 bits. The interpretation of - * the remaining 30 bits depends on a particual format layout. + * Possible format layouts are encoded in the first 2 bits. + * The interpretation of the remaining 30 bits depends on a particular + * format layout. */ #define PIPE_FORMAT_LAYOUT_RGBAZS 0 #define PIPE_FORMAT_LAYOUT_YCBCR 1 @@ -286,6 +287,7 @@ enum pipe_format { /** + * Unsigned 8-bit stencil format. * XXX should remove this, but S8_UNORM is a poor name */ #define PIPE_FORMAT_U_S8 PIPE_FORMAT_S8_UNORM @@ -294,11 +296,12 @@ enum pipe_format { /** * Builds pipe format name from format token. */ -static INLINE char *pf_sprint_name( char *str, uint format ) +static INLINE char *pf_sprint_name( char *str, enum pipe_format format ) { strcpy( str, "PIPE_FORMAT_" ); switch (pf_layout( format )) { - case PIPE_FORMAT_LAYOUT_RGBAZS: { + case PIPE_FORMAT_LAYOUT_RGBAZS: + { pipe_format_rgbazs_t rgbazs = (pipe_format_rgbazs_t) format; uint i; uint scale = 1 << (pf_exp8( rgbazs ) * 3); @@ -362,7 +365,8 @@ static INLINE char *pf_sprint_name( char *str, uint format ) } } break; - case PIPE_FORMAT_LAYOUT_YCBCR: { + case PIPE_FORMAT_LAYOUT_YCBCR: + { pipe_format_ycbcr_t ycbcr = (pipe_format_ycbcr_t) format; strcat( str, "YCBCR" ); @@ -375,6 +379,10 @@ static INLINE char *pf_sprint_name( char *str, uint format ) return str; } +/** + * Return bits for a particular component. + * \param comp component index, starting at 0 + */ static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) { uint size; @@ -397,6 +405,9 @@ static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) return size << (pf_exp8(format) * 3); } +/** + * Return total bits needed for the pixel format. + */ static INLINE uint pf_get_bits( enum pipe_format format ) { if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { @@ -417,7 +428,11 @@ static INLINE uint pf_get_bits( enum pipe_format format ) } } -static INLINE uint pf_get_size( enum pipe_format format ) { +/** + * Return bytes per pixel for the given format. + */ +static INLINE uint pf_get_size( enum pipe_format format ) +{ assert(pf_get_bits(format) % 8 == 0); return pf_get_bits(format) / 8; } -- cgit v1.2.3 From ad62644290cfb849db13fddab67bbf8515698d27 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 21 Mar 2008 10:43:56 -0600 Subject: gallium: added sRGB formats --- src/gallium/include/pipe/p_format.h | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index d09b6dbefc..86e9704256 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -80,6 +80,8 @@ static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ #define PIPE_FORMAT_TYPE_SNORM 3 #define PIPE_FORMAT_TYPE_USCALED 4 #define PIPE_FORMAT_TYPE_SSCALED 5 +#define PIPE_FORMAT_TYPE_SRGB 6 + /** * Because the destination vector is assumed to be RGBA FLOAT, we @@ -282,7 +284,12 @@ enum pipe_format { PIPE_FORMAT_R8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R8G8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R8G8B8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8B8A8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ) + PIPE_FORMAT_R8G8B8A8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ), + /* sRGB formats */ + PIPE_FORMAT_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), + PIPE_FORMAT_A8_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), + PIPE_FORMAT_R8G8B8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ), + PIPE_FORMAT_R8G8B8A8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ) }; -- cgit v1.2.3 From a472e3558e96dc4efcfee5db0e0cd21d9318dd30 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 21 Mar 2008 10:45:33 -0600 Subject: gallium: PIPE_FORMAT_TYPE_ comments --- src/gallium/include/pipe/p_format.h | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 86e9704256..5f7b7c74b3 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -75,12 +75,12 @@ static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ * Format types for RGBAZS layout. */ #define PIPE_FORMAT_TYPE_UNKNOWN 0 -#define PIPE_FORMAT_TYPE_FLOAT 1 -#define PIPE_FORMAT_TYPE_UNORM 2 -#define PIPE_FORMAT_TYPE_SNORM 3 -#define PIPE_FORMAT_TYPE_USCALED 4 -#define PIPE_FORMAT_TYPE_SSCALED 5 -#define PIPE_FORMAT_TYPE_SRGB 6 +#define PIPE_FORMAT_TYPE_FLOAT 1 /**< 16/32/64-bit/channel formats */ +#define PIPE_FORMAT_TYPE_UNORM 2 /**< uints, normalized to [0,1] */ +#define PIPE_FORMAT_TYPE_SNORM 3 /**< ints, normalized to [-1,1] */ +#define PIPE_FORMAT_TYPE_USCALED 4 /**< uints, not normalized */ +#define PIPE_FORMAT_TYPE_SSCALED 5 /**< ints, not normalized */ +#define PIPE_FORMAT_TYPE_SRGB 6 /**< sRGB colorspace */ /** -- cgit v1.2.3 From 419248b5287e98494d4460ed418e572a8e556ac6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 21 Mar 2008 10:50:26 -0600 Subject: gallium: document is_format_supported()'s type param --- src/gallium/include/pipe/p_screen.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h index 6be9a82b68..35bd46d691 100644 --- a/src/gallium/include/pipe/p_screen.h +++ b/src/gallium/include/pipe/p_screen.h @@ -69,6 +69,7 @@ struct pipe_screen { float (*get_paramf)( struct pipe_screen *, int param ); + /**< type is one of PIPE_TEXTURE, PIPE_SURFACE */ boolean (*is_format_supported)( struct pipe_screen *, enum pipe_format format, uint type ); -- cgit v1.2.3 From 13540d38d9d4ac4f71533ce239f0d9a2b950e6db Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 21 Mar 2008 10:51:21 -0600 Subject: gallium: use is_format_supported() to determine if GL_EXT_texture_sRGB is supported --- src/mesa/state_tracker/st_extensions.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 0962b5f74c..9926e118c4 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -155,7 +155,6 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.EXT_texture_env_combine = GL_TRUE; ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE; ctx->Extensions.EXT_texture_lod_bias = GL_TRUE; - ctx->Extensions.EXT_texture_sRGB = GL_TRUE; /* XXX temp */ ctx->Extensions.NV_blend_square = GL_TRUE; ctx->Extensions.NV_texgen_reflection = GL_TRUE; @@ -214,4 +213,8 @@ void st_init_extensions(struct st_context *st) /*ctx->Extensions.ARB_shadow_ambient = GL_TRUE;*/ } + if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_SRGB, + PIPE_TEXTURE)) { + ctx->Extensions.EXT_texture_sRGB = GL_TRUE; + } } -- cgit v1.2.3 From 4394736252fd79827a635a9746243689fef76b05 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 21 Mar 2008 11:03:53 -0600 Subject: gallium: added DXT formats (preliminary, will probably change) --- src/gallium/include/pipe/p_format.h | 24 +++++++++++++++++++++++- 1 file changed, 23 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 5f7b7c74b3..3238b152b6 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -49,6 +49,8 @@ extern "C" { */ #define PIPE_FORMAT_LAYOUT_RGBAZS 0 #define PIPE_FORMAT_LAYOUT_YCBCR 1 +#define PIPE_FORMAT_LAYOUT_DXT 2 /**< XXX temporary? */ + static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ { @@ -199,6 +201,20 @@ static INLINE uint pf_rev(pipe_format_ycbcr_t f) return (f >> 2) & 0x1; } + +/** + * Compresssed format layouts (this will probably change) + */ +#define _PIPE_FORMAT_DXT( LEVEL, RSIZE, GSIZE, BSIZE, ASIZE ) \ + ((PIPE_FORMAT_LAYOUT_DXT << 0) | \ + ((LEVEL) << 2) | \ + ((RSIZE) << 5) | \ + ((GSIZE) << 8) | \ + ((BSIZE) << 11) | \ + ((ASIZE) << 14) ) + + + /** * Texture/surface image formats (preliminary) */ @@ -289,7 +305,13 @@ enum pipe_format { PIPE_FORMAT_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), PIPE_FORMAT_A8_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), PIPE_FORMAT_R8G8B8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_R8G8B8A8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ) + PIPE_FORMAT_R8G8B8A8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), + + /* compressed formats */ + PIPE_FORMAT_DXT1_RGB = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 0 ), + PIPE_FORMAT_DXT1_RGBA = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 8 ), + PIPE_FORMAT_DXT3_RGBA = _PIPE_FORMAT_DXT( 3, 8, 8, 8, 8 ), + PIPE_FORMAT_DXT5_RGBA = _PIPE_FORMAT_DXT( 5, 8, 8, 8, 8 ) }; -- cgit v1.2.3 From 3600e7402c2a1824c8e5e2cc6ea189bcdeb4db6b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 21 Mar 2008 11:05:02 -0600 Subject: gallium: use is_format_supported(PIPE_FORMAT_DXT5_RGBA) to check if s3tc is supported The PIPE_CAP_S3TC token will be deprecated. --- src/mesa/state_tracker/st_extensions.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 9926e118c4..7fab4d041d 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -189,10 +189,6 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.ATI_separate_stencil = GL_TRUE; } - if (screen->get_param(screen, PIPE_CAP_S3TC)) { - ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE; - } - if (screen->get_param(screen, PIPE_CAP_ANISOTROPIC_FILTER)) { ctx->Extensions.EXT_texture_filter_anisotropic = GL_TRUE; } @@ -217,4 +213,10 @@ void st_init_extensions(struct st_context *st) PIPE_TEXTURE)) { ctx->Extensions.EXT_texture_sRGB = GL_TRUE; } + + if (screen->is_format_supported(screen, PIPE_FORMAT_DXT5_RGBA, + PIPE_TEXTURE)) { + ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE; + } + } -- cgit v1.2.3 From b70a6babfbc035d64dbe35ac4bf9218e8232b435 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Mar 2008 10:03:23 -0600 Subject: i915: added to-do note about setting the max_lod register to get proper min/mag filter selection --- src/gallium/drivers/i915simple/i915_state_sampler.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index 9dbb1b1b23..84f6529a3a 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -206,6 +206,12 @@ i915_update_texture(struct i915_context *i915, uint unit, | format | MS3_USE_FENCE_REGS); + /* + * XXX sampler->max_lod should be used to program the MAX_LOD field below. + * Also, when min_filter != mag_filter and there's just one mipmap level, + * set max_lod = 1 to make sure i915 chooses between min/mag filtering. + */ + /* MS4 state */ state[1] = ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) -- cgit v1.2.3 From c80a380ebb58c15db87309d466ef57fb006b40ae Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Mar 2008 12:32:48 -0600 Subject: Fix some PBO breakage. In _mesa_Bitmap, can't early return if bitmap ptr is NULL, it may be an offset into a PBO. Similarly for _mesa_GetTexImage. --- src/mesa/main/drawpix.c | 14 ++++++-------- src/mesa/main/teximage.c | 3 --- 2 files changed, 6 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c index ae9c7e29a1..5df55ef0c9 100644 --- a/src/mesa/main/drawpix.c +++ b/src/mesa/main/drawpix.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -343,12 +343,10 @@ _mesa_Bitmap( GLsizei width, GLsizei height, } if (ctx->RenderMode == GL_RENDER) { - if (bitmap && width && height) { - /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */ - GLint x = IFLOOR(ctx->Current.RasterPos[0] - xorig); - GLint y = IFLOOR(ctx->Current.RasterPos[1] - yorig); - ctx->Driver.Bitmap( ctx, x, y, width, height, &ctx->Unpack, bitmap ); - } + /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */ + GLint x = IFLOOR(ctx->Current.RasterPos[0] - xorig); + GLint y = IFLOOR(ctx->Current.RasterPos[1] - yorig); + ctx->Driver.Bitmap( ctx, x, y, width, height, &ctx->Unpack, bitmap ); } #if _HAVE_FULL_GL else if (ctx->RenderMode == GL_FEEDBACK) { diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c index f15e404527..384e145520 100644 --- a/src/mesa/main/teximage.c +++ b/src/mesa/main/teximage.c @@ -2323,9 +2323,6 @@ _mesa_GetTexImage( GLenum target, GLint level, GLenum format, return; } - if (!pixels) - return; - _mesa_lock_texture(ctx, texObj); { texImage = _mesa_select_tex_image(ctx, texObj, target, level); -- cgit v1.2.3 From ff938bf059a41a9bdf4c2c93cebe4a3b8a89c201 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 21 Mar 2008 13:43:07 -0600 Subject: add a number of PBO validate/map/unmap functions Helper functions for (some) drivers, including swrast. cherry-picked from Mesa/master --- src/mesa/main/bufferobj.c | 187 +++++++++++++++++++++++++++++++++++++++++++- src/mesa/main/bufferobj.h | 39 ++++++++- src/mesa/swrast/s_bitmap.c | 32 ++------ src/mesa/swrast/s_drawpix.c | 30 ++----- src/mesa/swrast/s_readpix.c | 30 ++----- 5 files changed, 240 insertions(+), 78 deletions(-) (limited to 'src') diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c index 009055a6ab..71e571353f 100644 --- a/src/mesa/main/bufferobj.c +++ b/src/mesa/main/bufferobj.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.1 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -469,6 +469,189 @@ _mesa_validate_pbo_access(GLuint dimensions, } +/** + * If the source of glBitmap data is a PBO, check that we won't read out + * of buffer bounds, then map the buffer. + * If not sourcing from a PBO, just return the bitmap pointer. + * This is a helper function for (some) drivers. + * Return NULL if error. + * If non-null return, must call validate_and_map_bitmap_pbo() when done. + */ +const GLubyte * +_mesa_validate_and_map_bitmap_pbo(GLcontext *ctx, + GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap) +{ + const GLubyte *buf; + + if (unpack->BufferObj->Name) { + /* unpack from PBO */ + if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, + GL_COLOR_INDEX, GL_BITMAP, + (GLvoid *) bitmap)) { + _mesa_error(ctx, GL_INVALID_OPERATION,"glBitmap(invalid PBO access)"); + return NULL; + } + + if (unpack->BufferObj->Pointer) { + /* buffer is already mapped - that's an error */ + _mesa_error(ctx, GL_INVALID_OPERATION, "glBitmap(PBO is mapped)"); + return NULL; + } + + buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, + GL_READ_ONLY_ARB, + unpack->BufferObj); + if (!buf) + return NULL; + + buf = ADD_POINTERS(buf, bitmap); + } + else { + /* unpack from normal memory */ + buf = bitmap; + } + + return buf; +} + + +/** + * Counterpart to validate_and_map_bitmap_pbo() + * This is a helper function for (some) drivers. + */ +void +_mesa_unmap_bitmap_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack) +{ + if (unpack->BufferObj->Name) { + ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, + unpack->BufferObj); + } +} + + +/** + * \sa _mesa_validate_and_map_bitmap_pbo + */ +const GLvoid * +_mesa_validate_and_map_drawpix_pbo(GLcontext *ctx, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels) +{ + const GLvoid *buf; + + if (unpack->BufferObj->Name) { + /* unpack from PBO */ + + if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, + format, type, pixels)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glDrawPixels(invalid PBO access)"); + return NULL; + } + + if (unpack->BufferObj->Pointer) { + /* buffer is already mapped - that's an error */ + _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels(PBO is mapped)"); + return NULL; + } + + buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, + GL_READ_ONLY_ARB, + unpack->BufferObj); + if (!buf) + return NULL; + + buf = ADD_POINTERS(buf, pixels); + } + else { + /* unpack from normal memory */ + buf = pixels; + } + + return buf; +} + + +/** + * \sa _mesa_unmap_bitmap_pbo + */ +void +_mesa_unmap_drapix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack) +{ + if (unpack->BufferObj->Name) { + ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, + unpack->BufferObj); + } +} + + +/** + * When doing glReadPixels into a PBO, this function will check for errors + * and map the buffer. + * Call _mesa_unmap_readpix_pbo() when finished + * \return NULL if error + */ +void * +_mesa_validate_and_map_readpix_pbo(GLcontext *ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *pack, + GLvoid *dest) +{ + void *buf; + + if (pack->BufferObj->Name) { + /* pack into PBO */ + if (!_mesa_validate_pbo_access(2, pack, width, height, 1, + format, type, dest)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glReadPixels(invalid PBO access)"); + return NULL; + } + + if (pack->BufferObj->Pointer) { + /* buffer is already mapped - that's an error */ + _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(PBO is mapped)"); + return NULL; + } + + buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT, + GL_WRITE_ONLY_ARB, + pack->BufferObj); + if (!buf) + return NULL; + + buf = ADD_POINTERS(buf, dest); + } + else { + /* pack to normal memory */ + buf = dest; + } + + return buf; +} + + +/** + * Counterpart to validate_and_map_readpix_pbo() + */ +void +_mesa_unmap_readpix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *pack) +{ + if (pack->BufferObj->Name) { + ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT, pack->BufferObj); + } +} + + /** * Return the gl_buffer_object for the given ID. * Always return NULL for ID 0. diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h index f54f9e9ff0..efb3b4711e 100644 --- a/src/mesa/main/bufferobj.h +++ b/src/mesa/main/bufferobj.h @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.3 + * Version: 7.1 * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -86,6 +86,41 @@ _mesa_validate_pbo_access(GLuint dimensions, GLsizei width, GLsizei height, GLsizei depth, GLenum format, GLenum type, const GLvoid *ptr); +extern const GLubyte * +_mesa_validate_and_map_bitmap_pbo(GLcontext *ctx, + GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap); + +extern void +_mesa_unmap_bitmap_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack); + +extern const GLvoid * +_mesa_validate_and_map_drawpix_pbo(GLcontext *ctx, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels); + +extern void +_mesa_unmap_drapix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack); + + +extern void * +_mesa_validate_and_map_readpix_pbo(GLcontext *ctx, + GLint x, GLint y, + GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *pack, + GLvoid *dest); + +extern void +_mesa_unmap_readpix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *pack); + + extern void _mesa_unbind_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj ); diff --git a/src/mesa/swrast/s_bitmap.c b/src/mesa/swrast/s_bitmap.c index 1e7f6c18e6..17f639fd55 100644 --- a/src/mesa/swrast/s_bitmap.c +++ b/src/mesa/swrast/s_bitmap.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.2 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -57,24 +57,10 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py, ASSERT(ctx->RenderMode == GL_RENDER); - if (unpack->BufferObj->Name) { - /* unpack from PBO */ - GLubyte *buf; - if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, - GL_COLOR_INDEX, GL_BITMAP, - (GLvoid *) bitmap)) { - _mesa_error(ctx, GL_INVALID_OPERATION,"glBitmap(invalid PBO access)"); - return; - } - buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, - GL_READ_ONLY_ARB, - unpack->BufferObj); - if (!buf) { - /* buffer is already mapped - that's an error */ - _mesa_error(ctx, GL_INVALID_OPERATION, "glBitmap(PBO is mapped)"); - return; - } - bitmap = ADD_POINTERS(buf, bitmap); + bitmap = _mesa_validate_and_map_bitmap_pbo(ctx, width, height, + unpack, bitmap); + if (!bitmap) { + return NULL; } RENDER_START(swrast,ctx); @@ -150,11 +136,7 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py, RENDER_FINISH(swrast,ctx); - if (unpack->BufferObj->Name) { - /* done with PBO so unmap it now */ - ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, - unpack->BufferObj); - } + _mesa_unmap_bitmap_pbo(ctx, unpack); } diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c index 0cf425e1c6..2cf3501274 100644 --- a/src/mesa/swrast/s_drawpix.c +++ b/src/mesa/swrast/s_drawpix.c @@ -812,7 +812,6 @@ draw_depth_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } - /** * Execute software-based glDrawPixels. * By time we get here, all error checking will have been done. @@ -835,25 +834,10 @@ _swrast_DrawPixels( GLcontext *ctx, if (swrast->NewState) _swrast_validate_derived( ctx ); - if (unpack->BufferObj->Name) { - /* unpack from PBO */ - GLubyte *buf; - if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glDrawPixels(invalid PBO access)"); - goto end; - } - buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, - GL_READ_ONLY_ARB, - unpack->BufferObj); - if (!buf) { - /* buffer is already mapped - that's an error */ - _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels(PBO is mapped)"); - goto end; - } - pixels = ADD_POINTERS(buf, pixels); - } + pixels = _mesa_validate_and_map_drawpix_pbo(ctx, x, y, width, height, + format, type, unpack, pixels); + if (!pixels) + return; switch (format) { case GL_STENCIL_INDEX: @@ -894,11 +878,7 @@ end: RENDER_FINISH(swrast,ctx); - if (unpack->BufferObj->Name) { - /* done with PBO so unmap it now */ - ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, - unpack->BufferObj); - } + _mesa_unmap_drapix_pbo(ctx, unpack); } diff --git a/src/mesa/swrast/s_readpix.c b/src/mesa/swrast/s_readpix.c index fe9a70f4ea..f4f882ae84 100644 --- a/src/mesa/swrast/s_readpix.c +++ b/src/mesa/swrast/s_readpix.c @@ -572,25 +572,11 @@ _swrast_ReadPixels( GLcontext *ctx, goto end; } - if (clippedPacking.BufferObj->Name) { - /* pack into PBO */ - GLubyte *buf; - if (!_mesa_validate_pbo_access(2, &clippedPacking, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glReadPixels(invalid PBO access)"); - goto end; - } - buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT, - GL_WRITE_ONLY_ARB, - clippedPacking.BufferObj); - if (!buf) { - /* buffer is already mapped - that's an error */ - _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(PBO is mapped)"); - goto end; - } - pixels = ADD_POINTERS(buf, pixels); - } + pixels = _mesa_validate_and_map_readpix_pbo(ctx, x, y, width, height, + format, type, + &clippedPacking, pixels); + if (!pixels) + return; switch (format) { case GL_COLOR_INDEX: @@ -632,9 +618,5 @@ _swrast_ReadPixels( GLcontext *ctx, end: RENDER_FINISH(swrast, ctx); - if (clippedPacking.BufferObj->Name) { - /* done with PBO so unmap it now */ - ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT, - clippedPacking.BufferObj); - } + _mesa_unmap_readpix_pbo(ctx, &clippedPacking); } -- cgit v1.2.3 From f1626f0bfd2b14ad8ca2afaad2ea8afb539e6491 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 21 Mar 2008 13:44:27 -0600 Subject: gallium: implement PBO operations for glDraw/ReadPixels and glBitmap At this time there are no optimizations for directly blitting between buffer objects and surfaces (always go through mappings). glean pbo test passes now --- src/mesa/state_tracker/st_cb_bitmap.c | 22 +++++++++++++--------- src/mesa/state_tracker/st_cb_drawpixels.c | 22 +++++++++++++--------- src/mesa/state_tracker/st_cb_readpixels.c | 19 +++++++++---------- 3 files changed, 35 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 33256196bb..4e23db0edc 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -32,6 +32,7 @@ #include "main/imports.h" #include "main/image.h" +#include "main/bufferobj.h" #include "main/macros.h" #include "main/texformat.h" #include "shader/program.h" @@ -191,7 +192,6 @@ combined_bitmap_fragment_program(GLcontext *ctx) } - /** * Create a texture which represents a bitmap image. */ @@ -224,19 +224,21 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, assert( 0 ); } + /* PBO source... */ + bitmap = _mesa_validate_and_map_bitmap_pbo(ctx, width, height, + unpack, bitmap); + if (!bitmap) { + return NULL; + } + /** - * Create a texture. + * Create texture to hold bitmap pattern. */ pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, width, height, 1, 0); - if (!pt) + if (!pt) { + _mesa_unmap_bitmap_pbo(ctx, unpack); return NULL; - - if (unpack->BufferObj && unpack->BufferObj->Name) { - /* - pt->region = buffer_object_region(unpack->BufferObj); - */ - printf("st_Bitmap (sourcing from PBO not implemented yet)\n"); } surface = screen->get_tex_surface(screen, pt, 0, 0, 0); @@ -301,6 +303,8 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, } /* row */ + _mesa_unmap_bitmap_pbo(ctx, unpack); + /* Release surface */ pipe_surface_unmap(surface); pipe_surface_reference(&surface, NULL); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index b4cd93cd54..c0f8e5ffdd 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -32,6 +32,7 @@ #include "main/imports.h" #include "main/image.h" +#include "main/bufferobj.h" #include "main/macros.h" #include "main/texformat.h" #include "shader/program.h" @@ -334,16 +335,17 @@ make_texture(struct st_context *st, assert(pipeFormat); cpp = st_sizeof_format(pipeFormat); + pixels = _mesa_validate_and_map_drawpix_pbo(ctx, width, height, + format, type, + unpack, pixels); + if (!pixels) + return NULL; + pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height, 1, 0); - if (!pt) + if (!pt) { + _mesa_unmap_drapix_pbo(ctx, unpack); return NULL; - - if (unpack->BufferObj && unpack->BufferObj->Name) { - /* - pt->region = buffer_object_region(unpack->BufferObj); - */ - printf("st_DrawPixels (sourcing from PBO not implemented yet)\n"); } { @@ -388,6 +390,8 @@ make_texture(struct st_context *st, ctx->_ImageTransferState = imageTransferStateSave; } + _mesa_unmap_drapix_pbo(ctx, unpack); + return pt; } @@ -836,9 +840,9 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, bufferFormat = ps->format; - if (any_fragment_ops(st) || + if (1/*any_fragment_ops(st) || any_pixel_transfer_ops(st) || - !compatible_formats(format, type, ps->format)) { + !compatible_formats(format, type, ps->format)*/) { /* textured quad */ struct pipe_texture *pt = make_texture(ctx->st, width, height, format, type, unpack, pixels); diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index c89c74229e..b22e846a15 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -34,6 +34,7 @@ #include "main/imports.h" +#include "main/bufferobj.h" #include "main/context.h" #include "main/image.h" @@ -126,7 +127,6 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } - /** * Do glReadPixels by getting rows from the framebuffer surface with * get_tile(). Convert to requested format/type with Mesa image routines. @@ -155,18 +155,15 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, return; } + dest = _mesa_validate_and_map_readpix_pbo(ctx, x, y, width, height, + format, type, + &clippedPacking, dest); + if (!dest) + return; + /* make sure rendering has completed */ pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); - if (pack->BufferObj && pack->BufferObj->Name) { - /* reading into a PBO */ - - } - else { - /* reading into user memory/buffer */ - - } - if (format == GL_STENCIL_INDEX) { st_read_stencil_pixels(ctx, x, y, width, height, type, pack, dest); return; @@ -285,6 +282,8 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } } } + + _mesa_unmap_readpix_pbo(ctx, &clippedPacking); } -- cgit v1.2.3 From a39091bc5b68e4d4f5302f1d3f1a138798f54b77 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Mar 2008 14:20:07 -0600 Subject: Refactor PBO validate/map code. We always need to do PBO validation, so do that in core Mesa before calling driv er routine. cherry-picked from Mesa/master. --- src/mesa/main/bufferobj.c | 76 ++++++------------------------- src/mesa/main/bufferobj.h | 24 ++++------ src/mesa/main/drawpix.c | 50 ++++++++++++++++++++ src/mesa/state_tracker/st_cb_bitmap.c | 3 +- src/mesa/state_tracker/st_cb_drawpixels.c | 4 +- src/mesa/state_tracker/st_cb_readpixels.c | 4 +- src/mesa/swrast/s_bitmap.c | 8 ++-- src/mesa/swrast/s_drawpix.c | 3 +- src/mesa/swrast/s_readpix.c | 4 +- 9 files changed, 82 insertions(+), 94 deletions(-) (limited to 'src') diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c index 71e571353f..e762eb3b63 100644 --- a/src/mesa/main/bufferobj.c +++ b/src/mesa/main/bufferobj.c @@ -475,31 +475,17 @@ _mesa_validate_pbo_access(GLuint dimensions, * If not sourcing from a PBO, just return the bitmap pointer. * This is a helper function for (some) drivers. * Return NULL if error. - * If non-null return, must call validate_and_map_bitmap_pbo() when done. + * If non-null return, must call _mesa_unmap_bitmap_pbo() when done. */ const GLubyte * -_mesa_validate_and_map_bitmap_pbo(GLcontext *ctx, - GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, - const GLubyte *bitmap) +_mesa_map_bitmap_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap) { const GLubyte *buf; if (unpack->BufferObj->Name) { /* unpack from PBO */ - if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, - GL_COLOR_INDEX, GL_BITMAP, - (GLvoid *) bitmap)) { - _mesa_error(ctx, GL_INVALID_OPERATION,"glBitmap(invalid PBO access)"); - return NULL; - } - - if (unpack->BufferObj->Pointer) { - /* buffer is already mapped - that's an error */ - _mesa_error(ctx, GL_INVALID_OPERATION, "glBitmap(PBO is mapped)"); - return NULL; - } - buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, GL_READ_ONLY_ARB, unpack->BufferObj); @@ -518,7 +504,7 @@ _mesa_validate_and_map_bitmap_pbo(GLcontext *ctx, /** - * Counterpart to validate_and_map_bitmap_pbo() + * Counterpart to _mesa_map_bitmap_pbo() * This is a helper function for (some) drivers. */ void @@ -533,33 +519,17 @@ _mesa_unmap_bitmap_pbo(GLcontext *ctx, /** - * \sa _mesa_validate_and_map_bitmap_pbo + * \sa _mesa_map_bitmap_pbo */ const GLvoid * -_mesa_validate_and_map_drawpix_pbo(GLcontext *ctx, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels) +_mesa_map_drawpix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels) { const GLvoid *buf; if (unpack->BufferObj->Name) { /* unpack from PBO */ - - if (!_mesa_validate_pbo_access(2, unpack, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glDrawPixels(invalid PBO access)"); - return NULL; - } - - if (unpack->BufferObj->Pointer) { - /* buffer is already mapped - that's an error */ - _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels(PBO is mapped)"); - return NULL; - } - buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, GL_READ_ONLY_ARB, unpack->BufferObj); @@ -592,36 +562,19 @@ _mesa_unmap_drapix_pbo(GLcontext *ctx, /** - * When doing glReadPixels into a PBO, this function will check for errors - * and map the buffer. + * If PBO is bound, map the buffer, return dest pointer in mapped buffer. * Call _mesa_unmap_readpix_pbo() when finished * \return NULL if error */ void * -_mesa_validate_and_map_readpix_pbo(GLcontext *ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - GLvoid *dest) +_mesa_map_readpix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *pack, + GLvoid *dest) { void *buf; if (pack->BufferObj->Name) { /* pack into PBO */ - if (!_mesa_validate_pbo_access(2, pack, width, height, 1, - format, type, dest)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glReadPixels(invalid PBO access)"); - return NULL; - } - - if (pack->BufferObj->Pointer) { - /* buffer is already mapped - that's an error */ - _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(PBO is mapped)"); - return NULL; - } - buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT, GL_WRITE_ONLY_ARB, pack->BufferObj); @@ -640,7 +593,7 @@ _mesa_validate_and_map_readpix_pbo(GLcontext *ctx, /** - * Counterpart to validate_and_map_readpix_pbo() + * Counterpart to _mesa_map_readpix_pbo() */ void _mesa_unmap_readpix_pbo(GLcontext *ctx, @@ -652,6 +605,7 @@ _mesa_unmap_readpix_pbo(GLcontext *ctx, } + /** * Return the gl_buffer_object for the given ID. * Always return NULL for ID 0. diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h index efb3b4711e..46525f08ae 100644 --- a/src/mesa/main/bufferobj.h +++ b/src/mesa/main/bufferobj.h @@ -87,21 +87,18 @@ _mesa_validate_pbo_access(GLuint dimensions, GLenum format, GLenum type, const GLvoid *ptr); extern const GLubyte * -_mesa_validate_and_map_bitmap_pbo(GLcontext *ctx, - GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, - const GLubyte *bitmap); +_mesa_map_bitmap_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap); extern void _mesa_unmap_bitmap_pbo(GLcontext *ctx, const struct gl_pixelstore_attrib *unpack); extern const GLvoid * -_mesa_validate_and_map_drawpix_pbo(GLcontext *ctx, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *unpack, - const GLvoid *pixels); +_mesa_map_drawpix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack, + const GLvoid *pixels); extern void _mesa_unmap_drapix_pbo(GLcontext *ctx, @@ -109,12 +106,9 @@ _mesa_unmap_drapix_pbo(GLcontext *ctx, extern void * -_mesa_validate_and_map_readpix_pbo(GLcontext *ctx, - GLint x, GLint y, - GLsizei width, GLsizei height, - GLenum format, GLenum type, - const struct gl_pixelstore_attrib *pack, - GLvoid *dest); +_mesa_map_readpix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *pack, + GLvoid *dest); extern void _mesa_unmap_readpix_pbo(GLcontext *ctx, diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c index 5df55ef0c9..0f64f1c1c0 100644 --- a/src/mesa/main/drawpix.c +++ b/src/mesa/main/drawpix.c @@ -24,6 +24,7 @@ #include "glheader.h" #include "imports.h" +#include "bufferobj.h" #include "context.h" #include "drawpix.h" #include "feedback.h" @@ -183,6 +184,23 @@ _mesa_DrawPixels( GLsizei width, GLsizei height, /* Round, to satisfy conformance tests (matches SGI's OpenGL) */ GLint x = IROUND(ctx->Current.RasterPos[0]); GLint y = IROUND(ctx->Current.RasterPos[1]); + + if (ctx->Unpack.BufferObj->Name) { + /* unpack from PBO */ + if (!_mesa_validate_pbo_access(2, &ctx->Unpack, width, height, 1, + format, type, pixels)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glDrawPixels(invalid PBO access)"); + return; + } + if (ctx->Unpack.BufferObj->Pointer) { + /* buffer is mapped - that's an error */ + _mesa_error(ctx, GL_INVALID_OPERATION, + "glDrawPixels(PBO is mapped)"); + return; + } + } + ctx->Driver.DrawPixels(ctx, x, y, width, height, format, type, &ctx->Unpack, pixels); } @@ -303,6 +321,21 @@ _mesa_ReadPixels( GLint x, GLint y, GLsizei width, GLsizei height, return; } + if (ctx->Pack.BufferObj->Name) { + if (!_mesa_validate_pbo_access(2, &ctx->Pack, width, height, 1, + format, type, pixels)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glReadPixels(invalid PBO access)"); + return; + } + + if (ctx->Pack.BufferObj->Pointer) { + /* buffer is mapped - that's an error */ + _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(PBO is mapped)"); + return; + } + } + ctx->Driver.ReadPixels(ctx, x, y, width, height, format, type, &ctx->Pack, pixels); } @@ -346,6 +379,23 @@ _mesa_Bitmap( GLsizei width, GLsizei height, /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */ GLint x = IFLOOR(ctx->Current.RasterPos[0] - xorig); GLint y = IFLOOR(ctx->Current.RasterPos[1] - yorig); + + if (ctx->Unpack.BufferObj->Name) { + /* unpack from PBO */ + if (!_mesa_validate_pbo_access(2, &ctx->Unpack, width, height, 1, + GL_COLOR_INDEX, GL_BITMAP, + (GLvoid *) bitmap)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glBitmap(invalid PBO access)"); + return; + } + if (ctx->Unpack.BufferObj->Pointer) { + /* buffer is mapped - that's an error */ + _mesa_error(ctx, GL_INVALID_OPERATION, "glBitmap(PBO is mapped)"); + return; + } + } + ctx->Driver.Bitmap( ctx, x, y, width, height, &ctx->Unpack, bitmap ); } #if _HAVE_FULL_GL diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 4e23db0edc..acc22d4323 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -225,8 +225,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, } /* PBO source... */ - bitmap = _mesa_validate_and_map_bitmap_pbo(ctx, width, height, - unpack, bitmap); + bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap); if (!bitmap) { return NULL; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c0f8e5ffdd..2ebaf8a2c3 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -335,9 +335,7 @@ make_texture(struct st_context *st, assert(pipeFormat); cpp = st_sizeof_format(pipeFormat); - pixels = _mesa_validate_and_map_drawpix_pbo(ctx, width, height, - format, type, - unpack, pixels); + pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels); if (!pixels) return NULL; diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index b22e846a15..4cf9adcd28 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -155,9 +155,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, return; } - dest = _mesa_validate_and_map_readpix_pbo(ctx, x, y, width, height, - format, type, - &clippedPacking, dest); + dest = _mesa_map_readpix_pbo(ctx, &clippedPacking, dest); if (!dest) return; diff --git a/src/mesa/swrast/s_bitmap.c b/src/mesa/swrast/s_bitmap.c index 17f639fd55..f3dda12e25 100644 --- a/src/mesa/swrast/s_bitmap.c +++ b/src/mesa/swrast/s_bitmap.c @@ -57,11 +57,9 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py, ASSERT(ctx->RenderMode == GL_RENDER); - bitmap = _mesa_validate_and_map_bitmap_pbo(ctx, width, height, - unpack, bitmap); - if (!bitmap) { - return NULL; - } + bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap); + if (!bitmap) + return; RENDER_START(swrast,ctx); diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c index 2cf3501274..fb04d9f746 100644 --- a/src/mesa/swrast/s_drawpix.c +++ b/src/mesa/swrast/s_drawpix.c @@ -834,8 +834,7 @@ _swrast_DrawPixels( GLcontext *ctx, if (swrast->NewState) _swrast_validate_derived( ctx ); - pixels = _mesa_validate_and_map_drawpix_pbo(ctx, x, y, width, height, - format, type, unpack, pixels); + pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels); if (!pixels) return; diff --git a/src/mesa/swrast/s_readpix.c b/src/mesa/swrast/s_readpix.c index f4f882ae84..2f155d0b70 100644 --- a/src/mesa/swrast/s_readpix.c +++ b/src/mesa/swrast/s_readpix.c @@ -572,9 +572,7 @@ _swrast_ReadPixels( GLcontext *ctx, goto end; } - pixels = _mesa_validate_and_map_readpix_pbo(ctx, x, y, width, height, - format, type, - &clippedPacking, pixels); + pixels = _mesa_map_readpix_pbo(ctx, &clippedPacking, pixels); if (!pixels) return; -- cgit v1.2.3 From fb1f4e207d9becca4a83d4934ff6aebf8270e51a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 21 Mar 2008 15:54:27 -0600 Subject: gallium: fix const buffer update bug If only glUniform is called between two renderings, the const buffers weren't getting updated. Need to set the _NEW_PROGRAM flag in st_upload_constants() as that's the dirty flag set by glUniform. Fixes glean tapi2 test. --- src/mesa/state_tracker/st_atom_constbuf.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index 21416da2e0..cd7372902b 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -66,7 +66,8 @@ void st_upload_constants( struct st_context *st, /* Update our own dependency flags. This works because this * function will also be called whenever the program changes. */ - st->constants.tracked_state[id].dirty.mesa = params->StateFlags; + st->constants.tracked_state[id].dirty.mesa = + (params->StateFlags | _NEW_PROGRAM); _mesa_load_state_parameters(st->ctx, params); @@ -115,7 +116,7 @@ static void update_vs_constants(struct st_context *st ) const struct st_tracked_state st_update_vs_constants = { .name = "st_update_vs_constants", .dirty = { - .mesa = 0, + .mesa = 0, /* set dynamically above */ .st = ST_NEW_VERTEX_PROGRAM, }, .update = update_vs_constants @@ -134,7 +135,7 @@ static void update_fs_constants(struct st_context *st ) const struct st_tracked_state st_update_fs_constants = { .name = "st_update_fs_constants", .dirty = { - .mesa = 0, + .mesa = 0, /* set dynamically above */ .st = ST_NEW_FRAGMENT_PROGRAM, }, .update = update_fs_constants -- cgit v1.2.3 From 2902c164a22b6bcb6a42d7cd7fa82b608875093b Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 21 Mar 2008 10:23:52 -0700 Subject: cell: Remove unnecessary default_blend work-around I suspect that there was some other bug in the blend code-gen that made this work-around necessary. --- src/gallium/drivers/cell/spu/spu_main.c | 21 --------------------- 1 file changed, 21 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 41bebf5362..0a490ab277 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -64,21 +64,6 @@ static unsigned char depth_stencil_code_buffer[4 * 64] static unsigned char fb_blend_code_buffer[4 * 64] ALIGN16_ATTRIB; -static struct spu_blend_results -default_blend(qword frag_r, qword frag_g, qword frag_b, qword frag_a, - qword pixel_r, qword pixel_g, qword pixel_b, qword pixel_a, - qword frag_mask) -{ - struct spu_blend_results result; - - result.r = si_selb(pixel_r, frag_r, frag_mask); - result.g = si_selb(pixel_g, frag_g, frag_mask); - result.b = si_selb(pixel_b, frag_b, frag_mask); - result.a = si_selb(pixel_a, frag_a, frag_mask); - - return result; -} - /** * Tell the PPU that this SPU has finished copying a buffer to @@ -285,9 +270,6 @@ cmd_state_blend(const struct cell_command_blend *state) spu.blend = (blend_func) fb_blend_code_buffer; spu.read_fb = state->read_fb; } else { - /* If there is no code, use the default; - */ - spu.blend = default_blend; spu.read_fb = FALSE; } } @@ -622,9 +604,6 @@ one_time_init(void) memset(spu.ctile_status, TILE_STATUS_DEFINED, sizeof(spu.ctile_status)); memset(spu.ztile_status, TILE_STATUS_DEFINED, sizeof(spu.ztile_status)); invalidate_tex_cache(); - - spu.blend = default_blend; - spu.read_fb = FALSE; } -- cgit v1.2.3 From f140062b72ee2df05020d86abdc47336262494f9 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 21 Mar 2008 10:25:58 -0700 Subject: Tabs to spaces --- .../drivers/cell/ppu/cell_state_per_fragment.c | 54 +++++++++++----------- 1 file changed, 27 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c index 988c251e20..f353aeab0a 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -72,8 +72,8 @@ emit_alpha_test(struct pipe_depth_stencil_alpha_state *dsa, int tmp_a = spe_allocate_available_register(f); int tmp_b = spe_allocate_available_register(f); union { - float f; - unsigned u; + float f; + unsigned u; } ref_val; boolean complement = FALSE; @@ -84,42 +84,42 @@ emit_alpha_test(struct pipe_depth_stencil_alpha_state *dsa, switch (dsa->alpha.func) { case PIPE_FUNC_NOTEQUAL: - complement = TRUE; - /* FALLTHROUGH */ + complement = TRUE; + /* FALLTHROUGH */ case PIPE_FUNC_EQUAL: - spe_fceq(f, tmp_a, ref, alphas); - break; + spe_fceq(f, tmp_a, ref, alphas); + break; case PIPE_FUNC_LEQUAL: - complement = TRUE; - /* FALLTHROUGH */ + complement = TRUE; + /* FALLTHROUGH */ case PIPE_FUNC_GREATER: - spe_fcgt(f, tmp_a, ref, alphas); - break; + spe_fcgt(f, tmp_a, ref, alphas); + break; case PIPE_FUNC_LESS: - complement = TRUE; - /* FALLTHROUGH */ + complement = TRUE; + /* FALLTHROUGH */ case PIPE_FUNC_GEQUAL: - spe_fcgt(f, tmp_a, ref, alphas); - spe_fceq(f, tmp_b, ref, alphas); - spe_or(f, tmp_a, tmp_b, tmp_a); - break; + spe_fcgt(f, tmp_a, ref, alphas); + spe_fceq(f, tmp_b, ref, alphas); + spe_or(f, tmp_a, tmp_b, tmp_a); + break; case PIPE_FUNC_ALWAYS: case PIPE_FUNC_NEVER: default: - assert(0); - break; + assert(0); + break; } if (complement) { - spe_andc(f, mask, mask, tmp_a); + spe_andc(f, mask, mask, tmp_a); } else { - spe_and(f, mask, mask, tmp_a); + spe_and(f, mask, mask, tmp_a); } spe_release_register(f, ref); @@ -674,7 +674,7 @@ emit_alpha_factor_calculation(struct spe_function *f, static void emit_color_factor_calculation(struct spe_function *f, unsigned sF, unsigned mask, - const struct pipe_blend_color *blend_color, + const struct pipe_blend_color *blend_color, const int *src, const int *dst, int *factor) @@ -754,10 +754,10 @@ emit_color_factor_calculation(struct spe_function *f, /* FALLTHROUGH */ case PIPE_BLENDFACTOR_CONST_COLOR: for (i = 0; i < 3; i++) { - factor[i] = spe_allocate_available_register(f); + factor[i] = spe_allocate_available_register(f); - spe_il(f, factor[i], color.u[i] & 0x0ffff); - spe_ilh(f, factor[i], color.u[i] >> 16); + spe_il(f, factor[i], color.u[i] & 0x0ffff); + spe_ilh(f, factor[i], color.u[i] >> 16); } break; @@ -946,7 +946,7 @@ emit_blend_calculation(struct spe_function *f, */ void cell_generate_alpha_blend(struct cell_blend_state *cb, - const struct pipe_blend_color *blend_color) + const struct pipe_blend_color *blend_color) { struct pipe_blend_state *const b = &cb->base; struct spe_function *const f = &cb->code; @@ -1054,7 +1054,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, */ if (((b->colormask & 8) != 0) && need_alpha_factor) { src_factor[3] = emit_alpha_factor_calculation(f, sF[3], - blend_color->color[3], + blend_color->color[3], frag[3], pixel[3]); /* If the alpha destination blend factor is the same as the alpha source @@ -1063,7 +1063,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, dst_factor[3] = (dF[3] == sF[3]) ? src_factor[3] : emit_alpha_factor_calculation(f, dF[3], - blend_color->color[3], + blend_color->color[3], frag[3], pixel[3]); } -- cgit v1.2.3 From 600499cf888fee9a91ff3106beca939ea0c7b2bd Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 21 Mar 2008 11:15:49 -0700 Subject: cell: Change code-gen for CONST_COLOR blend factor Previously the constant color blend factor was compiled into the generated code. This meant that the code had to be regenerated each time the constant color was changed. This doesn't fit with the model used in Gallium. As-is, the code could be better. The constant color is loaded for every quad processed, even if it is not used. Also, if a lot of (1-x) blend factors are used, 1.0 will be loaded and reloaded into registers many times. --- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 2 +- .../drivers/cell/ppu/cell_state_per_fragment.c | 93 +++++++++++----------- .../drivers/cell/ppu/cell_state_per_fragment.h | 3 +- src/gallium/drivers/cell/spu/spu_main.h | 2 + src/gallium/drivers/cell/spu/spu_tri.c | 2 + 5 files changed, 53 insertions(+), 49 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 86fcdcff1f..d956d6fad4 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -63,7 +63,7 @@ cell_bind_blend_state(struct pipe_context *pipe, void *state) draw_flush(cell->draw); if ((blend != NULL) && (blend->code.store == NULL)) { - cell_generate_alpha_blend(blend, &cell->blend_color); + cell_generate_alpha_blend(blend); } cell->blend = blend; diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c index f353aeab0a..c750b1d89d 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -588,19 +588,13 @@ cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa) */ static int emit_alpha_factor_calculation(struct spe_function *f, - unsigned factor, float const_alpha, - int src_alpha, int dst_alpha) + unsigned factor, + int src_alpha, int dst_alpha, int const_alpha) { - union { - float f; - unsigned u; - } alpha; int factor_reg; int tmp; - alpha.f = const_alpha; - switch (factor) { case PIPE_BLENDFACTOR_ONE: factor_reg = -1; @@ -621,13 +615,17 @@ emit_alpha_factor_calculation(struct spe_function *f, break; case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - const_alpha = 1.0 - const_alpha; - /* FALLTHROUGH */ - case PIPE_BLENDFACTOR_CONST_ALPHA: factor_reg = spe_allocate_available_register(f); - spe_il(f, factor_reg, alpha.u & 0x0ffff); - spe_ilh(f, factor_reg, alpha.u >> 16); + tmp = spe_allocate_available_register(f); + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + spe_fs(f, factor_reg, tmp, const_alpha); + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_CONST_ALPHA: + factor_reg = const_alpha; break; case PIPE_BLENDFACTOR_ZERO: @@ -674,24 +672,15 @@ emit_alpha_factor_calculation(struct spe_function *f, static void emit_color_factor_calculation(struct spe_function *f, unsigned sF, unsigned mask, - const struct pipe_blend_color *blend_color, const int *src, const int *dst, + const int *const_color, int *factor) { - union { - float f[4]; - unsigned u[4]; - } color; int tmp; unsigned i; - color.f[0] = blend_color->color[0]; - color.f[1] = blend_color->color[1]; - color.f[2] = blend_color->color[2]; - color.f[3] = blend_color->color[3]; - factor[0] = -1; factor[1] = -1; factor[2] = -1; @@ -748,29 +737,40 @@ emit_color_factor_calculation(struct spe_function *f, break; case PIPE_BLENDFACTOR_INV_CONST_COLOR: - color.f[0] = 1.0 - color.f[0]; - color.f[1] = 1.0 - color.f[1]; - color.f[2] = 1.0 - color.f[2]; - /* FALLTHROUGH */ - case PIPE_BLENDFACTOR_CONST_COLOR: + tmp = spe_allocate_available_register(f); + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + for (i = 0; i < 3; i++) { factor[i] = spe_allocate_available_register(f); - spe_il(f, factor[i], color.u[i] & 0x0ffff); - spe_ilh(f, factor[i], color.u[i] >> 16); + spe_fs(f, factor[i], tmp, const_color[i]); + } + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_CONST_COLOR: + for (i = 0; i < 3; i++) { + factor[i] = const_color[i]; } break; case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - color.f[3] = 1.0 - color.f[3]; - /* FALLTHROUGH */ - case PIPE_BLENDFACTOR_CONST_ALPHA: factor[0] = spe_allocate_available_register(f); factor[1] = factor[0]; factor[2] = factor[0]; - spe_il(f, factor[0], color.u[3] & 0x0ffff); - spe_ilh(f, factor[0], color.u[3] >> 16); + tmp = spe_allocate_available_register(f); + spe_il(f, tmp, 1); + spe_cuflt(f, tmp, tmp, 0); + spe_fs(f, factor[0], tmp, const_color[3]); + spe_release_register(f, tmp); + break; + + case PIPE_BLENDFACTOR_CONST_ALPHA: + factor[0] = const_color[3]; + factor[1] = factor[0]; + factor[2] = factor[0]; break; case PIPE_BLENDFACTOR_ZERO: @@ -945,8 +945,7 @@ emit_blend_calculation(struct spe_function *f, * Generate code to perform alpha blending on the SPE */ void -cell_generate_alpha_blend(struct cell_blend_state *cb, - const struct pipe_blend_color *blend_color) +cell_generate_alpha_blend(struct cell_blend_state *cb) { struct pipe_blend_state *const b = &cb->base; struct spe_function *const f = &cb->code; @@ -972,7 +971,13 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, spe_allocate_register(f, 9), spe_allocate_register(f, 10), }; - const int mask = spe_allocate_register(f, 11); + const int const_color[4] = { + spe_allocate_register(f, 11), + spe_allocate_register(f, 12), + spe_allocate_register(f, 13), + spe_allocate_register(f, 14), + }; + const int mask = spe_allocate_register(f, 15); unsigned func[4]; unsigned sF[4]; unsigned dF[4]; @@ -1053,8 +1058,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, * the alpha factor, calculate the alpha factor. */ if (((b->colormask & 8) != 0) && need_alpha_factor) { - src_factor[3] = emit_alpha_factor_calculation(f, sF[3], - blend_color->color[3], + src_factor[3] = emit_alpha_factor_calculation(f, sF[3], const_color[3], frag[3], pixel[3]); /* If the alpha destination blend factor is the same as the alpha source @@ -1062,8 +1066,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, */ dst_factor[3] = (dF[3] == sF[3]) ? src_factor[3] - : emit_alpha_factor_calculation(f, dF[3], - blend_color->color[3], + : emit_alpha_factor_calculation(f, dF[3], const_color[3], frag[3], pixel[3]); } @@ -1080,8 +1083,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, emit_color_factor_calculation(f, b->rgb_src_factor, b->colormask, - blend_color, - frag, pixel, src_factor); + frag, pixel, const_color, src_factor); } @@ -1101,8 +1103,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb, emit_color_factor_calculation(f, b->rgb_dst_factor, b->colormask, - blend_color, - frag, pixel, dst_factor); + frag, pixel, const_color, dst_factor); } diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h index 541c3b3be0..f699247f9e 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h @@ -29,7 +29,6 @@ extern void cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa); extern void -cell_generate_alpha_blend(struct cell_blend_state *cb, - const struct pipe_blend_color *blend_color); +cell_generate_alpha_blend(struct cell_blend_state *cb); #endif /* CELL_STATE_PER_FRAGMENT_H */ diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index 56d0968676..49f5d99674 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -77,6 +77,7 @@ struct spu_blend_results { typedef struct spu_blend_results (*blend_func)( qword frag_r, qword frag_g, qword frag_b, qword frag_a, qword pixel_r, qword pixel_g, qword pixel_b, qword pixel_a, + qword const_r, qword const_g, qword const_b, qword const_a, qword frag_mask); struct spu_framebuffer { @@ -108,6 +109,7 @@ struct spu_global boolean read_fb; blend_func blend; + qword const_blend_color[4] ALIGN16_ATTRIB; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct cell_command_texture texture; diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index c4272d6e93..e6a1ce01df 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -356,6 +356,8 @@ emit_quad( int x, int y, mask_t mask ) const struct spu_blend_results result = (*spu.blend)(soa_frag[0], soa_frag[1], soa_frag[2], soa_frag[3], soa_pix[0], soa_pix[1], soa_pix[2], soa_pix[3], + spu.const_blend_color[0], spu.const_blend_color[1], + spu.const_blend_color[2], spu.const_blend_color[3], (qword) mask); -- cgit v1.2.3 From 47531442e9c89c3ca764e9be225cfaec388609a1 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Fri, 21 Mar 2008 11:20:49 -0700 Subject: cell: Generate blend / depth test code when state atom is created Code generation should be performed when the device-specific state atom is created, not when it is bound. --- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 30 ++++++++------------------ 1 file changed, 9 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index d956d6fad4..00f4be7401 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -1,5 +1,5 @@ /************************************************************************** - * + * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ /* Authors: @@ -47,7 +47,7 @@ cell_create_blend_state(struct pipe_context *pipe, struct cell_blend_state *cb = MALLOC(sizeof(struct cell_blend_state)); (void) memcpy(cb, blend, sizeof(*blend)); - cb->code.store = NULL; + cell_generate_alpha_blend(cb); return cb; } @@ -57,16 +57,10 @@ static void cell_bind_blend_state(struct pipe_context *pipe, void *state) { struct cell_context *cell = cell_context(pipe); - struct cell_blend_state *blend = (struct cell_blend_state *) state; - draw_flush(cell->draw); - if ((blend != NULL) && (blend->code.store == NULL)) { - cell_generate_alpha_blend(blend); - } - - cell->blend = blend; + cell->blend = (struct cell_blend_state *) state; cell->dirty |= CELL_NEW_BLEND; } @@ -105,7 +99,7 @@ cell_create_depth_stencil_alpha_state(struct pipe_context *pipe, MALLOC(sizeof(struct cell_depth_stencil_alpha_state)); (void) memcpy(cdsa, depth_stencil, sizeof(*depth_stencil)); - cdsa->code.store = NULL; + cell_generate_depth_stencil_test(cdsa); return cdsa; } @@ -116,16 +110,11 @@ cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe, void *depth_stencil) { struct cell_context *cell = cell_context(pipe); - struct cell_depth_stencil_alpha_state *cdsa = - (struct cell_depth_stencil_alpha_state *) depth_stencil; draw_flush(cell->draw); - if ((cdsa != NULL) && (cdsa->code.store == NULL)) { - cell_generate_depth_stencil_test(cdsa); - } - - cell->depth_stencil = cdsa; + cell->depth_stencil = + (struct cell_depth_stencil_alpha_state *) depth_stencil; cell->dirty |= CELL_NEW_DEPTH_STENCIL; } @@ -362,4 +351,3 @@ cell_init_state_functions(struct cell_context *cell) cell->pipe.set_scissor_state = cell_set_scissor_state; cell->pipe.set_viewport_state = cell_set_viewport_state; } - -- cgit v1.2.3 From 4e977fb35befa60b2f74a21c0c9818854e6a7c85 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 22 Mar 2008 10:05:55 +0000 Subject: gallium: Remove pedantic asserts. Move these to a higher level instead. --- src/gallium/auxiliary/util/u_handle_table.c | 18 +++++++----------- 1 file changed, 7 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index 5a731a6b96..2176a00959 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -170,7 +170,7 @@ handle_table_set(struct handle_table *ht, unsigned index; assert(ht); - assert(handle > 0); + assert(handle); if(!handle) return 0; @@ -184,7 +184,9 @@ handle_table_set(struct handle_table *ht, if(!handle_table_resize(ht, index)) return 0; - assert(!ht->objects[index]); + if(ht->objects[index] && ht->destroy) + ht->destroy(ht->objects[index]); + ht->objects[index] = object; return handle; @@ -198,13 +200,11 @@ handle_table_get(struct handle_table *ht, void *object; assert(ht); - assert(handle > 0); - assert(handle <= ht->size); + assert(handle); if(!handle || handle > ht->size) return NULL; object = ht->objects[handle - 1]; - assert(object); return object; } @@ -218,18 +218,14 @@ handle_table_remove(struct handle_table *ht, unsigned index; assert(ht); - assert(handle > 0); - assert(handle <= ht->size); + assert(handle); if(!handle || handle > ht->size) return; index = handle - 1; object = ht->objects[index]; - if(!object) { - /* XXX: this warning may be noisy for legitimate use -- remove later */ - debug_warning("removing empty handle"); + if(!object) return; - } if(ht->destroy) ht->destroy(object); -- cgit v1.2.3 From f73cfd9e5cb0f47057f5b78b019787726798f238 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 22 Mar 2008 09:12:02 -0600 Subject: delete default programs with ctx->Driver.DeleteProgram() --- src/mesa/main/context.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index a9f9bd9da4..d06644f65d 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -695,10 +695,10 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) _mesa_DeleteHashTable(ss->Programs); #endif #if FEATURE_ARB_vertex_program - _mesa_delete_program(ctx, ss->DefaultVertexProgram); + ctx->Driver.DeleteProgram(ctx, ss->DefaultVertexProgram); #endif #if FEATURE_ARB_fragment_program - _mesa_delete_program(ctx, ss->DefaultFragmentProgram); + ctx->Driver.DeleteProgram(ctx, ss->DefaultFragmentProgram); #endif #if FEATURE_ATI_fragment_shader -- cgit v1.2.3 From 28b3b078959db8a1e60adfb66f35ceb04d4f414d Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 22 Mar 2008 09:12:27 -0600 Subject: gallium: fix mem leak (fee pstipple stage) --- src/gallium/auxiliary/draw/draw_context.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 4cca965ac1..41da93cdf8 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -113,6 +113,8 @@ void draw_destroy( struct draw_context *draw ) draw->pipeline.aaline->destroy( draw->pipeline.aaline ); if (draw->pipeline.aapoint) draw->pipeline.aapoint->destroy( draw->pipeline.aapoint ); + if (draw->pipeline.pstipple) + draw->pipeline.pstipple->destroy( draw->pipeline.pstipple ); if (draw->pipeline.rasterize) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); tgsi_exec_machine_free_data(&draw->machine); -- cgit v1.2.3 From 0b60ade834dd09ba806f8b5dd2dfed25469a8c8e Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 22 Mar 2008 09:12:52 -0600 Subject: gallium: fix a few mem leaks when deleting programs --- src/mesa/state_tracker/st_cb_program.c | 33 ++++++++++++++++++++------------- 1 file changed, 20 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 0a62e20e0c..7cf352d710 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -126,12 +126,17 @@ static void st_delete_program( GLcontext *ctx, struct gl_program *prog ) { struct st_context *st = st_context(ctx); + struct pipe_context *pipe = st->pipe; switch( prog->Target ) { case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; st_remove_vertex_program(st, stvp); + if (stvp->driver_shader) { + pipe->delete_vs_state(pipe, stvp->driver_shader); + stvp->driver_shader = NULL; + } } break; case GL_FRAGMENT_PROGRAM_ARB: @@ -139,6 +144,13 @@ static void st_delete_program( GLcontext *ctx, struct st_fragment_program *stfp = (struct st_fragment_program *) prog; st_remove_fragment_program(st, stfp); + if (stfp->driver_shader) { + pipe->delete_fs_state(pipe, stfp->driver_shader); + stfp->driver_shader = NULL; + } + + assert(!stfp->vertex_programs); + } break; default: @@ -162,19 +174,17 @@ static void st_program_string_notify( GLcontext *ctx, struct gl_program *prog ) { struct st_context *st = st_context(ctx); + struct pipe_context *pipe = st->pipe; if (target == GL_FRAGMENT_PROGRAM_ARB) { struct st_fragment_program *stfp = (struct st_fragment_program *) prog; stfp->serialNo++; -#if 0 - if (stfp->cso) { - /* free the TGSI code */ - // cso_delete(stfp->vs); - stfp->cso = NULL; + if (stfp->driver_shader) { + pipe->delete_fs_state(pipe, stfp->driver_shader); + stfp->driver_shader = NULL; } -#endif stfp->param_state = stfp->Base.Base.Parameters->StateFlags; @@ -191,14 +201,11 @@ static void st_program_string_notify( GLcontext *ctx, stvp->serialNo++; -#if 0 - if (stvp->cso) { - /* free the CSO data */ - st->pipe->delete_vs_state(st->pipe, stvp->cso->data); - FREE((void *) stvp->cso); - stvp->cso = NULL; + if (stvp->driver_shader) { + pipe->delete_vs_state(pipe, stvp->driver_shader); + stvp->driver_shader = NULL; } -#endif + if (stvp->draw_shader) { draw_delete_vertex_shader(st->draw, stvp->draw_shader); stvp->draw_shader = NULL; -- cgit v1.2.3 From 2f8f6c2918e721f6525b2124cde053b84beecafe Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 22 Mar 2008 09:15:23 -0600 Subject: gallium: remove temporary _screen suffix from function names --- src/gallium/drivers/softpipe/sp_texture.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index a98b3b1a4a..256586ec88 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -81,8 +81,8 @@ softpipe_texture_layout(struct softpipe_texture * spt) static struct pipe_texture * -softpipe_texture_create_screen(struct pipe_screen *screen, - const struct pipe_texture *templat) +softpipe_texture_create(struct pipe_screen *screen, + const struct pipe_texture *templat) { struct pipe_winsys *ws = screen->winsys; struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); @@ -110,8 +110,8 @@ softpipe_texture_create_screen(struct pipe_screen *screen, static void -softpipe_texture_release_screen(struct pipe_screen *screen, - struct pipe_texture **pt) +softpipe_texture_release(struct pipe_screen *screen, + struct pipe_texture **pt) { if (!*pt) return; @@ -136,9 +136,9 @@ softpipe_texture_release_screen(struct pipe_screen *screen, static struct pipe_surface * -softpipe_get_tex_surface_screen(struct pipe_screen *screen, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) +softpipe_get_tex_surface(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice) { struct pipe_winsys *ws = screen->winsys; struct softpipe_texture *spt = softpipe_texture(pt); @@ -197,7 +197,7 @@ softpipe_init_texture_funcs( struct softpipe_context *softpipe ) void softpipe_init_screen_texture_funcs(struct pipe_screen *screen) { - screen->texture_create = softpipe_texture_create_screen; - screen->texture_release = softpipe_texture_release_screen; - screen->get_tex_surface = softpipe_get_tex_surface_screen; + screen->texture_create = softpipe_texture_create; + screen->texture_release = softpipe_texture_release; + screen->get_tex_surface = softpipe_get_tex_surface; } -- cgit v1.2.3 From dae7993afc56be1e71e600af60e01e51eab17eda Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 22 Mar 2008 09:42:59 -0600 Subject: gallium: free samplers, textures in destroy() --- src/gallium/auxiliary/draw/draw_aaline.c | 7 +++++++ src/gallium/auxiliary/draw/draw_pstipple.c | 6 ++++++ 2 files changed, 13 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index b4fa6bd967..6742f7f4b9 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -665,7 +665,14 @@ aaline_reset_stipple_counter(struct draw_stage *stage) static void aaline_destroy(struct draw_stage *stage) { + struct aaline_stage *aaline = aaline_stage(stage); + + aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso); + + pipe_texture_release(&aaline->texture); + draw_free_temp_verts( stage ); + FREE( stage ); } diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index 9d154a6838..bd8d3a76ae 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -534,6 +534,12 @@ pstip_reset_stipple_counter(struct draw_stage *stage) static void pstip_destroy(struct draw_stage *stage) { + struct pstip_stage *pstip = pstip_stage(stage); + + pstip->pipe->delete_sampler_state(pstip->pipe, pstip->sampler_cso); + + pipe_texture_release(&pstip->texture); + draw_free_temp_verts( stage ); FREE( stage ); } -- cgit v1.2.3 From f8acc3965e3fa9ec40ca63f3ddd0cb33966e8c95 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 22 Mar 2008 10:27:55 -0600 Subject: use ctx->Driver.DeleteProgram() in a few more places --- src/mesa/shader/program.c | 2 +- src/mesa/shader/shader_api.c | 6 +++--- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 3869e31714..39c4ca1dd5 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -354,7 +354,7 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog) clone->Format = prog->Format; clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions); if (!clone->Instructions) { - _mesa_delete_program(ctx, clone); + ctx->Driver.DeleteProgram(ctx, clone); return NULL; } _mesa_copy_instructions(clone->Instructions, prog->Instructions, diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 4470a87af0..4cb8bb8ed1 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -79,7 +79,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx, /* to prevent a double-free in the next call */ shProg->VertexProgram->Base.Parameters = NULL; } - _mesa_delete_program(ctx, &shProg->VertexProgram->Base); + ctx->Driver.DeleteProgram(ctx, &shProg->VertexProgram->Base); shProg->VertexProgram = NULL; } @@ -88,7 +88,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx, /* to prevent a double-free in the next call */ shProg->FragmentProgram->Base.Parameters = NULL; } - _mesa_delete_program(ctx, &shProg->FragmentProgram->Base); + ctx->Driver.DeleteProgram(ctx, &shProg->FragmentProgram->Base); shProg->FragmentProgram = NULL; } @@ -246,7 +246,7 @@ _mesa_free_shader(GLcontext *ctx, struct gl_shader *sh) _mesa_free(sh->InfoLog); for (i = 0; i < sh->NumPrograms; i++) { assert(sh->Programs[i]); - _mesa_delete_program(ctx, sh->Programs[i]); + ctx->Driver.DeleteProgram(ctx, sh->Programs[i]); } if (sh->Programs) _mesa_free(sh->Programs); -- cgit v1.2.3 From a35c1ca3ad4361fee30d21ef13d8d37ae91aee66 Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 22 Mar 2008 10:29:30 -0600 Subject: gallium: fix a few more shader-related mem leaks --- src/mesa/state_tracker/st_atom_shader.c | 43 ++++++++++++--------------------- src/mesa/state_tracker/st_atom_shader.h | 6 ++--- src/mesa/state_tracker/st_cb_program.c | 28 +++++++++++++-------- src/mesa/state_tracker/st_program.c | 10 ++++++++ 4 files changed, 45 insertions(+), 42 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 0726688493..ec39026eb3 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -77,34 +77,6 @@ struct translated_vertex_program -/** - * Free data hanging off the st vert prog. - */ -void -st_remove_vertex_program(struct st_context *st, struct st_vertex_program *stvp) -{ - /* no-op, for now? */ -} - - -/** - * Free data hanging off the st frag prog. - */ -void -st_remove_fragment_program(struct st_context *st, - struct st_fragment_program *stfp) -{ - struct translated_vertex_program *xvp, *next; - - for (xvp = stfp->vertex_programs; xvp; xvp = next) { - next = xvp->next; - /* XXX free xvp->vs */ - free(xvp); - } -} - - - /** * Given a vertex program output attribute, return the corresponding * fragment program input attribute. @@ -265,6 +237,21 @@ find_translated_vp(struct st_context *st, } +void +st_free_translated_vertex_programs(struct st_context *st, + struct translated_vertex_program *xvp) +{ + struct translated_vertex_program *next; + + while (xvp) { + next = xvp->next; + free(xvp); + xvp = next; + } +} + + + static void update_linkage( struct st_context *st ) { diff --git a/src/mesa/state_tracker/st_atom_shader.h b/src/mesa/state_tracker/st_atom_shader.h index f3124d87c0..8403bc66c9 100644 --- a/src/mesa/state_tracker/st_atom_shader.h +++ b/src/mesa/state_tracker/st_atom_shader.h @@ -31,10 +31,8 @@ extern void -st_remove_vertex_program(struct st_context *, struct st_vertex_program *); - -extern void -st_remove_fragment_program(struct st_context *, struct st_fragment_program *); +st_free_translated_vertex_programs(struct st_context *st, + struct translated_vertex_program *xvp); #endif /* ST_ATOM_SHADER_H */ diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 7cf352d710..a739fcd336 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -132,25 +132,33 @@ static void st_delete_program( GLcontext *ctx, case GL_VERTEX_PROGRAM_ARB: { struct st_vertex_program *stvp = (struct st_vertex_program *) prog; - st_remove_vertex_program(st, stvp); + if (stvp->driver_shader) { pipe->delete_vs_state(pipe, stvp->driver_shader); stvp->driver_shader = NULL; } + + if (stvp->state.tokens) { + FREE((void *) stvp->state.tokens); + stvp->state.tokens = NULL; + } } break; case GL_FRAGMENT_PROGRAM_ARB: { - struct st_fragment_program *stfp - = (struct st_fragment_program *) prog; - st_remove_fragment_program(st, stfp); + struct st_fragment_program *stfp = (struct st_fragment_program *) prog; + if (stfp->driver_shader) { pipe->delete_fs_state(pipe, stfp->driver_shader); stfp->driver_shader = NULL; } + + if (stfp->state.tokens) { + FREE((void *) stfp->state.tokens); + stfp->state.tokens = NULL; + } - assert(!stfp->vertex_programs); - + st_free_translated_vertex_programs(st, stfp->vertex_programs); } break; default: @@ -186,13 +194,13 @@ static void st_program_string_notify( GLcontext *ctx, stfp->driver_shader = NULL; } - stfp->param_state = stfp->Base.Base.Parameters->StateFlags; - if (stfp->state.tokens) { FREE((void *) stfp->state.tokens); stfp->state.tokens = NULL; } + stfp->param_state = stfp->Base.Base.Parameters->StateFlags; + if (st->fp == stfp) st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; } @@ -211,13 +219,13 @@ static void st_program_string_notify( GLcontext *ctx, stvp->draw_shader = NULL; } - stvp->param_state = stvp->Base.Base.Parameters->StateFlags; - if (stvp->state.tokens) { FREE((void *) stvp->state.tokens); stvp->state.tokens = NULL; } + stvp->param_state = stvp->Base.Base.Parameters->StateFlags; + if (st->vp == stvp) st->dirty.st |= ST_NEW_VERTEX_PROGRAM; } diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index d9d11ee0e8..101a6a48af 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -260,6 +260,16 @@ st_translate_vertex_program(struct st_context *st, outputMapping = defaultOutputMapping; } + /* free old shader state, if any */ + if (stvp->state.tokens) { + FREE((void *) stvp->state.tokens); + stvp->state.tokens = NULL; + } + if (stvp->driver_shader) { + pipe->delete_vs_state(pipe, stvp->driver_shader); + stvp->driver_shader = NULL; + } + /* XXX: fix static allocation of tokens: */ num_tokens = tgsi_translate_mesa_program( TGSI_PROCESSOR_VERTEX, -- cgit v1.2.3 From f40357e25c0520ef1d64ffab03501da4c8b93529 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 23 Mar 2008 16:44:59 +0000 Subject: gallium: beginnings of draw module vertex rework Trying to put a structure in place that we can actually optimize. Initially just implementing a passthrough mode, this will fairly soon replace all the vertex_cache/prim_queue/shader_queue stuff that's so hard to understand... Split the vertex processing into a couple of distinct stages: - Frontend - Prepares two lists of elements (fetch and draw) to be processed by the next stage. This stage doesn't fetch or draw vertices, but makes the decision which to draw. Multiple implementations of this will implement different strategies, currently just a vcache implementation. - MiddleEnd - Takes the list of fetch elements, fetches them, runs the vertex shader, cliptest, viewport transform on them to produce a linear array of vertex_header vertices. - Passes that list of vertices, plus the draw_elements (which index into that list) onto the backend - Backend - Either the existing primitive/clipping pipeline, or the vbuf_render hardware backend provided by the driver. Currently, the middle-end is the old passthrough code, and it build hardware vertices, not vertex_header vertices as above. It may be that passthrough is a special case in this respect. --- src/gallium/auxiliary/draw/Makefile | 4 + src/gallium/auxiliary/draw/SConscript | 6 +- src/gallium/auxiliary/draw/draw_context.c | 65 ++++- src/gallium/auxiliary/draw/draw_passthrough.c | 4 +- src/gallium/auxiliary/draw/draw_prim.c | 2 +- src/gallium/auxiliary/draw/draw_private.h | 33 ++- src/gallium/auxiliary/draw/draw_pt.c | 206 +++++++++++++++ src/gallium/auxiliary/draw/draw_pt.h | 116 ++++++++ src/gallium/auxiliary/draw/draw_pt_elts.c | 88 ++++++ src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 326 +++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_pt_vcache.c | 338 ++++++++++++++++++++++++ 11 files changed, 1169 insertions(+), 19 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt.c create mode 100644 src/gallium/auxiliary/draw/draw_pt.h create mode 100644 src/gallium/auxiliary/draw/draw_pt_elts.c create mode 100644 src/gallium/auxiliary/draw/draw_pt_fetch_emit.c create mode 100644 src/gallium/auxiliary/draw/draw_pt_vcache.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 21e9f737b7..0c7ce5da5b 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -16,6 +16,10 @@ C_SOURCES = \ draw_flatshade.c \ draw_offset.c \ draw_passthrough.c \ + draw_pt.c \ + draw_pt_vcache.c \ + draw_pt_fetch_emit.c \ + draw_pt_elts.c \ draw_prim.c \ draw_pstipple.c \ draw_stipple.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index d7fb86d992..9b3e7247c5 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -14,7 +14,11 @@ draw = env.ConvenienceLibrary( 'draw_debug.c', 'draw_flatshade.c', 'draw_offset.c', - 'draw_passthrough.c', + 'draw_passthrough.c', # going away soon + 'draw_pt.c', + 'draw_pt_vcache.c', + 'draw_pt_fetch_emit.c', + 'draw_pt_elts.c', 'draw_prim.c', 'draw_pstipple.c', 'draw_stipple.c', diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 41da93cdf8..903cc26766 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -37,10 +37,11 @@ #include "draw_vbuf.h" - struct draw_context *draw_create( void ) { struct draw_context *draw = CALLOC_STRUCT( draw_context ); + if (draw == NULL) + goto fail; #if defined(__i386__) || defined(__386__) draw->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; @@ -61,6 +62,19 @@ struct draw_context *draw_create( void ) draw->pipeline.validate = draw_validate_stage( draw ); draw->pipeline.first = draw->pipeline.validate; + if (!draw->pipeline.wide_line || + !draw->pipeline.wide_point || + !draw->pipeline.stipple || + !draw->pipeline.unfilled || + !draw->pipeline.twoside || + !draw->pipeline.offset || + !draw->pipeline.clip || + !draw->pipeline.flatshade || + !draw->pipeline.cull || + !draw->pipeline.validate) + goto fail; + + ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); ASSIGN_4V( draw->plane[2], 0, -1, 0, 1 ); @@ -75,6 +89,8 @@ struct draw_context *draw_create( void ) uint i; const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; char *tmp = align_malloc(Elements(draw->vs.queue) * size, 16); + if (!tmp) + goto fail; for (i = 0; i < Elements(draw->vs.queue); i++) draw->vs.queue[i].vertex = (struct vertex_header *)(tmp + i * size); @@ -93,22 +109,42 @@ struct draw_context *draw_create( void ) draw_vertex_cache_invalidate( draw ); draw_set_mapped_element_buffer( draw, 0, NULL ); + if (!draw_pt_init( draw )) + goto fail; + return draw; + +fail: + draw_destroy( draw ); + return NULL; } void draw_destroy( struct draw_context *draw ) { - draw->pipeline.wide_line->destroy( draw->pipeline.wide_line ); - draw->pipeline.wide_point->destroy( draw->pipeline.wide_point ); - draw->pipeline.stipple->destroy( draw->pipeline.stipple ); - draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); - draw->pipeline.twoside->destroy( draw->pipeline.twoside ); - draw->pipeline.offset->destroy( draw->pipeline.offset ); - draw->pipeline.clip->destroy( draw->pipeline.clip ); - draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); - draw->pipeline.cull->destroy( draw->pipeline.cull ); - draw->pipeline.validate->destroy( draw->pipeline.validate ); + if (!draw) + return; + + if (draw->pipeline.wide_line) + draw->pipeline.wide_line->destroy( draw->pipeline.wide_line ); + if (draw->pipeline.wide_point) + draw->pipeline.wide_point->destroy( draw->pipeline.wide_point ); + if (draw->pipeline.stipple) + draw->pipeline.stipple->destroy( draw->pipeline.stipple ); + if (draw->pipeline.unfilled) + draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); + if (draw->pipeline.twoside) + draw->pipeline.twoside->destroy( draw->pipeline.twoside ); + if (draw->pipeline.offset) + draw->pipeline.offset->destroy( draw->pipeline.offset ); + if (draw->pipeline.clip) + draw->pipeline.clip->destroy( draw->pipeline.clip ); + if (draw->pipeline.flatshade) + draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); + if (draw->pipeline.cull) + draw->pipeline.cull->destroy( draw->pipeline.cull ); + if (draw->pipeline.validate) + draw->pipeline.validate->destroy( draw->pipeline.validate ); if (draw->pipeline.aaline) draw->pipeline.aaline->destroy( draw->pipeline.aaline ); if (draw->pipeline.aapoint) @@ -117,8 +153,11 @@ void draw_destroy( struct draw_context *draw ) draw->pipeline.pstipple->destroy( draw->pipeline.pstipple ); if (draw->pipeline.rasterize) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); + tgsi_exec_machine_free_data(&draw->machine); - align_free( draw->vs.queue[0].vertex ); /* Frees all the vertices. */ + + if (draw->vs.queue[0].vertex) + align_free( draw->vs.queue[0].vertex ); /* Frees all the vertices. */ /* Not so fast -- we're just borrowing this at the moment. * @@ -126,6 +165,8 @@ void draw_destroy( struct draw_context *draw ) draw->render->destroy( draw->render ); */ + draw_pt_destroy( draw ); + FREE( draw ); } diff --git a/src/gallium/auxiliary/draw/draw_passthrough.c b/src/gallium/auxiliary/draw/draw_passthrough.c index fdec6a591b..dd00894c5b 100644 --- a/src/gallium/auxiliary/draw/draw_passthrough.c +++ b/src/gallium/auxiliary/draw/draw_passthrough.c @@ -125,9 +125,9 @@ fetch_store_general( struct draw_context *draw, case PIPE_FORMAT_B8G8R8A8_UNORM: { ubyte *ub = (ubyte *) from; - attrib[0] = UBYTE_TO_FLOAT(ub[0]); + attrib[2] = UBYTE_TO_FLOAT(ub[0]); attrib[1] = UBYTE_TO_FLOAT(ub[1]); - attrib[2] = UBYTE_TO_FLOAT(ub[2]); + attrib[0] = UBYTE_TO_FLOAT(ub[2]); attrib[3] = UBYTE_TO_FLOAT(ub[3]); } break; diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index cb0277fb6c..4fe0ddc02a 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -526,7 +526,7 @@ draw_arrays(struct draw_context *draw, unsigned prim, /* drawing done here: */ if (!draw->rasterizer->bypass_vs || - !draw_passthrough_arrays(draw, prim, start, count)) { + !draw_pt_arrays(draw, prim, start, count)) { /* we have to run the whole pipeline */ draw_prim(draw, prim, start, count); } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 1c65c3d1b2..0c5afcacfa 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -53,6 +53,9 @@ struct pipe_context; struct gallivm_prog; struct gallivm_cpu_engine; +struct draw_pt_middle_end; +struct draw_pt_front_end; + /** * Basic vertex info. * Carry some useful information around with the vertices in the prim pipe. @@ -203,8 +206,21 @@ struct draw_context /* Support prototype passthrough path: */ struct { - unsigned prim; - unsigned hw_vertex_size; + unsigned prim; /* XXX: to be removed */ + unsigned hw_vertex_size; /* XXX: to be removed */ + + struct { + struct draw_pt_middle_end *fetch_emit; + struct draw_pt_middle_end *fetch_shade_emit; + struct draw_pt_middle_end *fetch_shade_cliptest_pipeline_or_emit; + } middle; + + struct { + struct draw_pt_front_end *noop; + struct draw_pt_front_end *split_arrays; + struct draw_pt_front_end *vcache; + } front; + } pt; @@ -351,7 +367,18 @@ extern void draw_update_vertex_fetch( struct draw_context *draw ); extern boolean draw_need_pipeline(const struct draw_context *draw); -/* Prototype/hack +/* Passthrough mode (second attempt): + */ +boolean draw_pt_init( struct draw_context *draw ); +void draw_pt_destroy( struct draw_context *draw ); +boolean draw_pt_arrays( struct draw_context *draw, + unsigned prim, + unsigned start, + unsigned count ); + + + +/* Prototype/hack (DEPRECATED) */ boolean draw_passthrough_arrays(struct draw_context *draw, diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c new file mode 100644 index 0000000000..219b563422 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -0,0 +1,206 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_pt.h" + + +static boolean too_many_verts( struct draw_context *draw, + unsigned verts ) +{ + return verts < 1024; +} + +static boolean too_many_elts( struct draw_context *draw, + unsigned elts ) +{ + return elts < (16 * 1024); +} + + +boolean +draw_pt_arrays(struct draw_context *draw, + unsigned prim, + unsigned start, + unsigned count) +{ + const boolean pipeline = draw_need_pipeline(draw); + const boolean cliptest = !draw->rasterizer->bypass_clipping; + const boolean shading = !draw->rasterizer->bypass_vs; + struct draw_pt_front_end *frontend = NULL; + struct draw_pt_middle_end *middle = NULL; + + + /* Overall we do: + * - frontend -- prepare fetch_elts, draw_elts - eg vcache + * - middle -- fetch, shade, cliptest, viewport + * - pipeline -- the prim pipeline: clipping, wide lines, etc + * - backend -- the vbuf_render provided by the driver. + */ + + +#if 0 + if (!cliptest && !pipeline && !shading) { + /* This is the 'passthrough' path: + */ + /* Fetch user verts, emit hw verts: + */ + middle = draw->pt.middle.fetch_emit; + } + else if (!cliptest && !pipeline) { + /* Fetch user verts, run vertex shader, emit hw verts: + */ + middle = draw->pt.middle.fetch_shade_emit; + } + else if (!pipeline) { + /* Even though !pipeline, we have to run it to get clipping. We + * do know that the pipeline is just the clipping operation, but + * that probably doesn't help much. + * + * This is going to be the most important path for a lot of + * swtnl cards. + */ + /* Fetch user verts, + * run vertex shader, + * cliptest and viewport trasform + * if no clipped vertices, + * emit hw verts + * else + * run pipline + */ + middle = draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit; + } + else if (!cliptest) { + /* Fetch user verts, run vertex shader, run pipeline: + */ + middle = draw->pt.middle.fetch_shade_pipeline; + } + else { + /* This is what we're currently always doing: + */ + /* Fetch user verts, run vertex shader, cliptest, run pipeline: + */ + middle = draw->pt.middle.fetch_shade_cliptest_pipeline; + } +#else + if (cliptest /*|| pipeline*/ || shading) + return FALSE; + + middle = draw->pt.middle.fetch_emit; +#endif + + + /* If !pipeline, need to make sure we respect the driver's limited + * capabilites to receive blocks of vertex data and elements. + */ +#if 0 + if (!pipeline) { + unsigned vertex_mode = passthrough; + unsigned nr_verts = count_vertices( draw, start, count ); + unsigned hw_prim = prim; + + if (is_elts(draw)) { + frontend = draw->pt.front.vcache; + hw_prim = reduced_prim(prim); + } + + if (too_many_verts(nr_verts)) { + /* if (is_verts(draw) && can_split(prim)) { + draw = draw_arrays_split; + } + else */ { + frontend = draw->pt.front.vcache; + hw_prim = reduced_prim(prim); + } + } + + if (too_many_elts(count)) { + + /* if (is_elts(draw) && can_split(prim)) { + draw = draw_elts_split; + } + else */ { + frontend = draw->pt.front.vcache; + hw_prim = reduced_prim(prim); + } + } + + if (!good_prim(hw_prim)) { + draw = draw->pt.front.vcache; + } + } +#else + frontend = draw->pt.front.vcache; +#endif + + frontend->prepare( frontend, middle ); + + frontend->run( frontend, + prim, + draw_pt_elt_func( draw ), + draw_pt_elt_ptr( draw, start ), + count ); + + frontend->finish( frontend ); + + return TRUE; +} + + +boolean draw_pt_init( struct draw_context *draw ) +{ + draw->pt.middle.fetch_emit = draw_pt_fetch_emit( draw ); + if (!draw->pt.middle.fetch_emit) + return FALSE; + + draw->pt.front.vcache = draw_pt_vcache(); + if (!draw->pt.front.vcache) + return FALSE; + + return TRUE; +} + + +void draw_pt_destroy( struct draw_context *draw ) +{ + if (draw->pt.middle.fetch_emit) { + draw->pt.middle.fetch_emit->destroy( draw->pt.middle.fetch_emit ); + draw->pt.middle.fetch_emit = NULL; + } + + if (draw->pt.front.vcache) { + draw->pt.front.vcache->destroy( draw->pt.front.vcache ); + draw->pt.front.vcache = NULL; + } +} diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h new file mode 100644 index 0000000000..1b81d196f6 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -0,0 +1,116 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#ifndef DRAW_PT_H +#define DRAW_PT_H + +#include "pipe/p_compiler.h" + +typedef ushort (*pt_elt_func)( const void *elts, ushort idx ); + + +/* The "front end" - prepare sets of fetch, draw elements for the + * middle end. + * + * Currenly one version of this: + * - vcache - catchall implementation, decomposes to TRI/LINE/POINT prims + * Later: + * - varray, varray_split + * - velement, velement_split + * + * Currenly only using the vcache version. + */ +struct draw_pt_front_end { + void (*prepare)( struct draw_pt_front_end *, + struct draw_pt_middle_end * ); + + void (*run)( struct draw_pt_front_end *, + unsigned prim, + pt_elt_func elt_func, + const void *elt_ptr, + unsigned count ); + + void (*finish)( struct draw_pt_front_end * ); + void (*destroy)( struct draw_pt_front_end * ); +}; + + +/* The "middle end" - prepares actual hardware vertices for the + * hardware backend. + * + * Currently two versions of this: + * - fetch, vertex shade, cliptest, prim-pipeline + * - fetch, emit (ie passthrough) + * Later: + * - fetch, vertex shade, cliptest, maybe-pipeline, maybe-emit + * - fetch, vertex shade, emit + * + * Currenly only using the passthrough version. + */ +struct draw_pt_middle_end { + void (*prepare)( struct draw_pt_middle_end * ); + + void (*run)( struct draw_pt_middle_end *, + unsigned prim, + const unsigned *fetch_elts, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ); + + void (*finish)( struct draw_pt_middle_end * ); + void (*destroy)( struct draw_pt_middle_end * ); +}; + + +/* The "back end" - supplied by the driver, defined in draw_vbuf.h. + * + * Not sure whether to wrap the prim pipeline up as an alternate + * backend. Would be a win for everything except pure passthrough + * mode... + */ +struct vbuf_render; + + +/* Helper functions. + */ +pt_elt_func draw_pt_elt_func( struct draw_context *draw ); +const void *draw_pt_elt_ptr( struct draw_context *draw, + unsigned start ); + +/* Implementations: + */ +struct draw_pt_front_end *draw_pt_vcache( void ); +struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); + + + +#endif diff --git a/src/gallium/auxiliary/draw/draw_pt_elts.c b/src/gallium/auxiliary/draw/draw_pt_elts.c new file mode 100644 index 0000000000..585b83fa90 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_elts.c @@ -0,0 +1,88 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "draw/draw_pt.h" +#include "draw/draw_private.h" + +/* Neat get_elt func that also works for varrays drawing by encoding + * the start value into a pointer. + */ + +static ushort elt_uint( const void *elts, ushort idx ) +{ + return *(((const uint *)elts) + idx); +} + +static ushort elt_ushort( const void *elts, ushort idx ) +{ + return *(((const ushort *)elts) + idx); +} + +static ushort elt_ubyte( const void *elts, ushort idx ) +{ + return *(((const ubyte *)elts) + idx); +} + +static ushort elt_vert( const void *elts, ushort idx ) +{ + return (const ubyte *)elts - (const ubyte *)NULL + idx; +} + +pt_elt_func draw_pt_elt_func( struct draw_context *draw ) +{ + switch (draw->user.eltSize) { + case 0: return elt_vert; + case 1: return elt_ubyte; + case 2: return elt_ushort; + case 4: return elt_uint; + default: return NULL; + } +} + +const void *draw_pt_elt_ptr( struct draw_context *draw, + unsigned start ) +{ + const char *elts = draw->user.elts; + + switch (draw->user.eltSize) { + case 0: + return (const void *)(((const ubyte *)NULL) + start); + case 1: + return (const void *)(((const ubyte *)elts) + start); + case 2: + return (const void *)(((const ushort *)elts) + start); + case 4: + return (const void *)(((const uint *)elts) + start); + default: + return NULL; + } +} diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c new file mode 100644 index 0000000000..74a946545a --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -0,0 +1,326 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" + +/* The simplest 'middle end' in the new vertex code. + * + * The responsibilities of a middle end are to: + * - perform vertex fetch using + * - draw vertex element/buffer state + * - a list of fetch indices we received as an input + * - run the vertex shader + * - cliptest, + * - clip coord calculation + * - viewport transformation + * - if necessary, run the primitive pipeline, passing it: + * - a linear array of vertex_header vertices constructed here + * - a set of draw indices we received as an input + * - otherwise, drive the hw backend, + * - allocate space for hardware format vertices + * - translate the vertex-shader output vertices to hw format + * - calling the backend draw functions. + * + * For convenience, we provide a helper function to drive the hardware + * backend given similar inputs to those required to run the pipeline. + * + * In the case of passthrough mode, many of these actions are disabled + * or noops, so we end up doing: + * + * - perform vertex fetch + * - drive the hw backend + * + * IE, basically just vertex fetch to post-vs-format vertices, + * followed by a call to the backend helper function. + */ + + +struct fetch_emit_middle_end { + struct draw_pt_middle_end base; + struct draw_context *draw; + + struct { + const ubyte *ptr; + unsigned pitch; + enum pipe_format src_format; + unsigned dst_format; + } fetch[PIPE_ATTRIB_MAX]; + + unsigned nr_fetch; + unsigned hw_vertex_size; +}; + + +/** + * General-purpose fetch from user's vertex arrays, emit to driver's + * vertex buffer. + * + * XXX this is totally temporary. + */ +static void +fetch_store_general( struct fetch_emit_middle_end *feme, + void *out_ptr, + const unsigned *fetch_elts, + unsigned count ) +{ + float *out = (float *)out_ptr; + struct vbuf_render *render = feme->draw->render; + uint i, j; + + for (i = 0; i < count; i++) { + unsigned elt = fetch_elts[i]; + + for (j = 0; j < feme->nr_fetch; j++) { + float attrib[4]; + const ubyte *from = (feme->fetch[j].ptr + + feme->fetch[j].pitch * elt); + + /* The normal fetch/emit code: + */ + switch (feme->fetch[j].src_format) { + case PIPE_FORMAT_B8G8R8A8_UNORM: + { + ubyte *ub = (ubyte *) from; + attrib[2] = UBYTE_TO_FLOAT(ub[0]); + attrib[1] = UBYTE_TO_FLOAT(ub[1]); + attrib[0] = UBYTE_TO_FLOAT(ub[2]); + attrib[3] = UBYTE_TO_FLOAT(ub[3]); + } + break; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + { + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = f[1]; + attrib[2] = f[2]; + attrib[3] = f[3]; + } + break; + case PIPE_FORMAT_R32G32B32_FLOAT: + { + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = f[1]; + attrib[2] = f[2]; + attrib[3] = 1.0; + } + break; + case PIPE_FORMAT_R32G32_FLOAT: + { + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = f[1]; + attrib[2] = 0.0; + attrib[3] = 1.0; + } + break; + case PIPE_FORMAT_R32_FLOAT: + { + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = 0.0; + attrib[2] = 0.0; + attrib[3] = 1.0; + } + break; + default: + assert(0); + } + + if (0) debug_printf("attrib %d: %f %f %f %f\n", j, + attrib[0], attrib[1], attrib[2], attrib[3]); + + switch (feme->fetch[j].dst_format) { + case EMIT_1F: + out[0] = attrib[0]; + out += 1; + break; + case EMIT_2F: + out[0] = attrib[0]; + out[1] = attrib[1]; + out += 2; + break; + case EMIT_4F: + out[0] = attrib[0]; + out[1] = attrib[1]; + out[2] = attrib[2]; + out[3] = attrib[3]; + out += 4; + break; + default: + assert(0); + } + } + } +} + + + +static void fetch_emit_prepare( struct draw_pt_middle_end *middle ) +{ + struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; + struct draw_context *draw = feme->draw; + const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); + unsigned nr_attrs = vinfo->num_attribs; + unsigned i; + + for (i = 0; i < nr_attrs; i++) { + enum pipe_format src_format; + unsigned dst_format; + const void *src_ptr; + unsigned pitch; + + if (vinfo->emit[i] == EMIT_HEADER) { + static const unsigned zero = 0; + src_ptr = &zero; + src_format = PIPE_FORMAT_R32_FLOAT; + pitch = 0; + dst_format = EMIT_1F; + } + else if (vinfo->emit[i] == EMIT_1F_PSIZE) { + src_ptr = &feme->draw->rasterizer->point_size; + src_format = PIPE_FORMAT_R32_FLOAT; + pitch = 0; + dst_format = EMIT_1F; + } + else { + unsigned src_element = vinfo->src_index[i]; + unsigned src_buffer = draw->vertex_element[src_element].vertex_buffer_index; + + src_ptr = ((const ubyte *)draw->user.vbuffer[src_buffer] + + draw->vertex_buffer[src_buffer].buffer_offset + + draw->vertex_element[src_element].src_offset); + + src_format = draw->vertex_element[src_element].src_format; + pitch = draw->vertex_buffer[src_buffer].pitch; + dst_format = vinfo->emit[i]; + } + + feme->fetch[i].src_format = src_format; + feme->fetch[i].ptr = src_ptr; + feme->fetch[i].pitch = pitch; + feme->fetch[i].dst_format = dst_format; + } + + + feme->nr_fetch = nr_attrs; + feme->hw_vertex_size = vinfo->size * 4; +} + + + + + +static void fetch_emit_run( struct draw_pt_middle_end *middle, + unsigned prim, + const unsigned *fetch_elts, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; + struct draw_context *draw = feme->draw; + void *hw_verts; + boolean ok; + + ok = draw->render->set_primitive( draw->render, + prim ); + if (!ok) { + assert(0); + return; + } + + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)feme->hw_vertex_size, + (ushort)fetch_count ); + if (!hw_verts) { + assert(0); + return; + } + + + /* Single routine to fetch vertices and emit HW verts. + */ + fetch_store_general( feme, + hw_verts, + fetch_elts, + fetch_count ); + + /* XXX: Draw arrays path to avoid re-emitting index list again and + * again. + */ + draw->render->draw( draw->render, + draw_elts, + draw_count ); + + /* Done -- that was easy, wasn't it: + */ + draw->render->release_vertices( draw->render, + hw_verts, + feme->hw_vertex_size, + fetch_count ); + +} + + + +static void fetch_emit_finish( struct draw_pt_middle_end *middle ) +{ + /* nothing to do */ +} + +static void fetch_emit_destroy( struct draw_pt_middle_end *middle ) +{ + FREE(middle); +} + + +struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ) +{ + struct fetch_emit_middle_end *fetch_emit = CALLOC_STRUCT( fetch_emit_middle_end ); + + fetch_emit->base.prepare = fetch_emit_prepare; + fetch_emit->base.run = fetch_emit_run; + fetch_emit->base.finish = fetch_emit_finish; + fetch_emit->base.destroy = fetch_emit_destroy; + + fetch_emit->draw = draw; + + return &fetch_emit->base; +} + diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c new file mode 100644 index 0000000000..07e9f0ae5f --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -0,0 +1,338 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +//#include "draw/draw_vbuf.h" +//#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" + + +#define CACHE_MAX 32 +#define FETCH_MAX 128 +#define DRAW_MAX (16*1024) + +struct vcache_frontend { + struct draw_pt_front_end base; + + ushort in[CACHE_MAX]; + ushort out[CACHE_MAX]; + + ushort draw_elts[DRAW_MAX]; + unsigned fetch_elts[FETCH_MAX]; + + unsigned draw_count; + unsigned fetch_count; + + pt_elt_func elt_func; + const void *elt_ptr; + + struct draw_pt_middle_end *middle; + unsigned output_prim; +}; + +static void vcache_flush( struct vcache_frontend *vcache ) +{ +#if 0 + /* Should always be true if output_prim == input_prim, otherwise + * not so much... + */ + unsigned i; + for (i = 0; i < vcache->draw_count; i++) { + assert( vcache->fetch_elts[vcache->draw_elts[i]] == + vcache->elt_func(vcache->elt_ptr, i) ); + } +#endif + + if (vcache->draw_count) + vcache->middle->run( vcache->middle, + vcache->output_prim, + vcache->fetch_elts, + vcache->fetch_count, + vcache->draw_elts, + vcache->draw_count ); + + memset(vcache->in, ~0, sizeof(vcache->in)); + vcache->fetch_count = 0; + vcache->draw_count = 0; +} + +static void vcache_check_flush( struct vcache_frontend *vcache ) +{ + if ( vcache->draw_count + 6 >= DRAW_MAX || + vcache->fetch_count + 4 >= FETCH_MAX ) + { + vcache_flush( vcache ); + } +} + + +static void vcache_elt( struct vcache_frontend *vcache, + unsigned felt ) +{ + // ushort felt = elt(draw, i); + + ushort idx = felt % CACHE_MAX; + + if (vcache->in[idx] != felt) { + assert(vcache->fetch_count < FETCH_MAX); + + vcache->in[idx] = felt; + vcache->out[idx] = vcache->fetch_count; + vcache->fetch_elts[vcache->fetch_count++] = felt; + } + + vcache->draw_elts[vcache->draw_count++] = vcache->out[idx]; +} + +static void vcache_triangle( struct vcache_frontend *vcache, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + /* TODO: encode edgeflags in draw_elts */ + vcache_elt(vcache, i0); + vcache_elt(vcache, i1); + vcache_elt(vcache, i2); + vcache_check_flush(vcache); +} + +static void vcache_line( struct vcache_frontend *vcache, + boolean reset, + unsigned i0, + unsigned i1 ) +{ + /* TODO: encode reset-line-stipple in draw_elts */ + (void) reset; + vcache_elt(vcache, i0); + vcache_elt(vcache, i1); + vcache_check_flush(vcache); +} + + +static void vcache_point( struct vcache_frontend *vcache, + unsigned i0 ) +{ + vcache_elt(vcache, i0); + vcache_check_flush(vcache); +} + +static void vcache_quad( struct vcache_frontend *vcache, + unsigned i0, + unsigned i1, + unsigned i2, + unsigned i3 ) +{ + vcache_triangle( vcache, i0, i1, i3 ); + vcache_triangle( vcache, i1, i2, i3 ); +} + + +static void vcache_prepare( struct draw_pt_front_end *frontend, + struct draw_pt_middle_end *middle ) +{ + struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; + vcache->middle = middle; + middle->prepare( middle ); +} + +static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES +}; + + +static void vcache_run( struct draw_pt_front_end *frontend, + unsigned prim, + pt_elt_func get_elt, + const void *elts, + unsigned count ) +{ + struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; + unsigned i; + + /* These are for validation only: + */ + vcache->elt_func = get_elt; + vcache->elt_ptr = elts; + vcache->output_prim = reduced_prim[prim]; + + switch (prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i ++) { + vcache_point( vcache, + get_elt(elts, i) ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) { + vcache_line( vcache, + TRUE, + get_elt(elts, i + 0), + get_elt(elts, i + 1)); + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + for (i = 1; i < count; i++) { + vcache_line( vcache, + i == 1, /* XXX: only if vb not split */ + get_elt(elts, i - 1), + get_elt(elts, i) ); + } + + vcache_line( vcache, + 0, + get_elt(elts, count - 1), + get_elt(elts, 0) ); + } + break; + + case PIPE_PRIM_LINE_STRIP: + for (i = 1; i < count; i++) { + vcache_line( vcache, + i == 1, + get_elt(elts, i - 1), + get_elt(elts, i) ); + } + break; + + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + vcache_triangle( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2) ); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + vcache_triangle( vcache, + get_elt(elts, i + 1), + get_elt(elts, i + 0), + get_elt(elts, i + 2) ); + } + else { + vcache_triangle( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2) ); + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + for (i = 0; i+2 < count; i++) { + vcache_triangle( vcache, + get_elt(elts, 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2) ); + } + break; + + + case PIPE_PRIM_QUADS: + for (i = 0; i+3 < count; i += 4) { + vcache_quad( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, i + 3)); + } + break; + + case PIPE_PRIM_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + vcache_quad( vcache, + get_elt(elts, i + 2), + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 3)); + } + break; + + case PIPE_PRIM_POLYGON: + for (i = 0; i+2 < count; i++) { + vcache_triangle( vcache, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0)); + } + break; + + default: + assert(0); + break; + } + + vcache_flush( vcache ); +} + +static void vcache_finish( struct draw_pt_front_end *frontend ) +{ + struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; + vcache->middle->finish( vcache->middle ); + vcache->middle = NULL; +} + +static void vcache_destroy( struct draw_pt_front_end *frontend ) +{ + FREE(frontend); +} + + +struct draw_pt_front_end *draw_pt_vcache( void ) +{ + struct vcache_frontend *vcache = CALLOC_STRUCT( vcache_frontend ); + + vcache->base.prepare = vcache_prepare; + vcache->base.run = vcache_run; + vcache->base.finish = vcache_finish; + vcache->base.destroy = vcache_destroy; + + memset(vcache->in, ~0, sizeof(vcache->in)); + + return &vcache->base; +} -- cgit v1.2.3 From 301b187ca9a811b608894d20bab934af0a10b8ab Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 23 Mar 2008 18:21:00 +0000 Subject: draw: fix some unsigned vs ushort confusion Middle-end elements are ushort, but prior to that have to treat all elements as unsigned to avoid wrapping and/or overruns. --- src/gallium/auxiliary/draw/draw_pt.h | 2 +- src/gallium/auxiliary/draw/draw_pt_elts.c | 8 ++++---- src/gallium/auxiliary/draw/draw_pt_vcache.c | 9 ++++----- 3 files changed, 9 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 1b81d196f6..f56fa8d5cc 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -35,7 +35,7 @@ #include "pipe/p_compiler.h" -typedef ushort (*pt_elt_func)( const void *elts, ushort idx ); +typedef unsigned (*pt_elt_func)( const void *elts, unsigned idx ); /* The "front end" - prepare sets of fetch, draw elements for the diff --git a/src/gallium/auxiliary/draw/draw_pt_elts.c b/src/gallium/auxiliary/draw/draw_pt_elts.c index 585b83fa90..f337f71ea0 100644 --- a/src/gallium/auxiliary/draw/draw_pt_elts.c +++ b/src/gallium/auxiliary/draw/draw_pt_elts.c @@ -37,22 +37,22 @@ * the start value into a pointer. */ -static ushort elt_uint( const void *elts, ushort idx ) +static unsigned elt_uint( const void *elts, unsigned idx ) { return *(((const uint *)elts) + idx); } -static ushort elt_ushort( const void *elts, ushort idx ) +static unsigned elt_ushort( const void *elts, unsigned idx ) { return *(((const ushort *)elts) + idx); } -static ushort elt_ubyte( const void *elts, ushort idx ) +static unsigned elt_ubyte( const void *elts, unsigned idx ) { return *(((const ubyte *)elts) + idx); } -static ushort elt_vert( const void *elts, ushort idx ) +static unsigned elt_vert( const void *elts, unsigned idx ) { return (const ubyte *)elts - (const ubyte *)NULL + idx; } diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 07e9f0ae5f..98c22eb4d4 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -45,7 +45,7 @@ struct vcache_frontend { struct draw_pt_front_end base; - ushort in[CACHE_MAX]; + unsigned in[CACHE_MAX]; ushort out[CACHE_MAX]; ushort draw_elts[DRAW_MAX]; @@ -74,13 +74,14 @@ static void vcache_flush( struct vcache_frontend *vcache ) } #endif - if (vcache->draw_count) + if (vcache->draw_count) { vcache->middle->run( vcache->middle, vcache->output_prim, vcache->fetch_elts, vcache->fetch_count, vcache->draw_elts, vcache->draw_count ); + } memset(vcache->in, ~0, sizeof(vcache->in)); vcache->fetch_count = 0; @@ -100,9 +101,7 @@ static void vcache_check_flush( struct vcache_frontend *vcache ) static void vcache_elt( struct vcache_frontend *vcache, unsigned felt ) { - // ushort felt = elt(draw, i); - - ushort idx = felt % CACHE_MAX; + unsigned idx = felt % CACHE_MAX; if (vcache->in[idx] != felt) { assert(vcache->fetch_count < FETCH_MAX); -- cgit v1.2.3 From 312cbc5a5c7416745976c2281a9bbce6e3332965 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 23 Mar 2008 18:30:53 +0000 Subject: gallium: wrap decls in extern "C" --- src/gallium/drivers/softpipe/sp_winsys.h | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h index fc372dba27..291825dfe2 100644 --- a/src/gallium/drivers/softpipe/sp_winsys.h +++ b/src/gallium/drivers/softpipe/sp_winsys.h @@ -37,6 +37,12 @@ #include "pipe/p_compiler.h" /* for boolean */ + +#ifdef __cplusplus +extern "C" { +#endif + + enum pipe_format; struct softpipe_winsys { @@ -60,4 +66,8 @@ struct pipe_screen * softpipe_create_screen(struct pipe_winsys *); +#ifdef __cplusplus +} +#endif + #endif /* SP_WINSYS_H */ -- cgit v1.2.3 From 732422f6708199d6655185b1a5daec86efe2f1b7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 23 Mar 2008 18:38:10 +0000 Subject: gallium: Memory debugging utilities. There are no known tools for windows kernel memory debugging, so this is a simple set of malloc etc wrappers. Enabled by default on win32 debug builds --- src/gallium/auxiliary/util/SConscript | 1 + src/gallium/auxiliary/util/p_debug_mem.c | 172 +++++++++++++++++++++++++++++++ src/gallium/include/pipe/p_debug.h | 23 +++++ src/gallium/include/pipe/p_util.h | 21 +++- 4 files changed, 215 insertions(+), 2 deletions(-) create mode 100644 src/gallium/auxiliary/util/p_debug_mem.c (limited to 'src') diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index b44f2d5e39..9b3929eb2d 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -4,6 +4,7 @@ util = env.ConvenienceLibrary( target = 'util', source = [ 'p_debug.c', + 'p_debug_mem.c', 'p_tile.c', 'p_util.c', 'u_blit.c', diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c new file mode 100644 index 0000000000..56599dafa6 --- /dev/null +++ b/src/gallium/auxiliary/util/p_debug_mem.c @@ -0,0 +1,172 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Memory debugging. + * + * @author José Fonseca + */ + +#ifdef WIN32 +#include +#include +#else +#include +#include +#endif + +#include "pipe/p_debug.h" +#include "util/u_double_list.h" + + +#define DEBUG_MEMORY_MAGIC 0x6e34090aU + + +#if defined(WIN32) && !defined(WINCE) +#define real_malloc(_size) EngAllocMem(0, _size, 'D3AG') +#define real_free(_ptr) EngFreeMem(_ptr) +#else +#define real_malloc(_size) malloc(_size) +#define real_free(_ptr) free(_ptr) +#endif + + +struct debug_memory_header +{ + struct list_head head; + + unsigned long no; + const char *file; + unsigned line; + const char *function; + size_t size; + unsigned magic; +}; + +static struct list_head list = { &list, &list }; + +static unsigned long start_no = 0; +static unsigned long end_no = 0; + + +void * +debug_malloc(const char *file, unsigned line, const char *function, + size_t size) +{ + struct debug_memory_header *hdr; + + hdr = real_malloc(sizeof(*hdr) + size); + if(!hdr) + return NULL; + + hdr->no = end_no++; + hdr->file = file; + hdr->line = line; + hdr->function = function; + hdr->size = size; + hdr->magic = DEBUG_MEMORY_MAGIC; + + LIST_ADDTAIL(&hdr->head, &list); + + return (void *)((char *)hdr + sizeof(*hdr)); +} + +void +debug_free(const char *file, unsigned line, const char *function, + void *ptr) +{ + struct debug_memory_header *hdr; + + if(!ptr) + return; + + hdr = (struct debug_memory_header *)((char *)ptr - sizeof(*hdr)); + if(hdr->magic != DEBUG_MEMORY_MAGIC) { + debug_printf("%s:%u:%s: freeing bad or corrupted memory %p\n", + file, line, function, + ptr); + debug_assert(0); + return; + } + + LIST_DEL(&hdr->head); + hdr->magic = 0; + + real_free(hdr); +} + +void * +debug_calloc(const char *file, unsigned line, const char *function, + size_t count, size_t size ) +{ + void *ptr = debug_malloc( file, line, function, count * size ); + if( ptr ) + memset( ptr, 0, count * size ); + return ptr; +} + +void * +debug_realloc(const char *file, unsigned line, const char *function, + void *old_ptr, size_t old_size, size_t new_size ) +{ + void *new_ptr = NULL; + + if (new_size != 0) { + new_ptr = debug_malloc( file, line, function, new_size ); + + if( new_ptr && old_ptr ) + memcpy( new_ptr, old_ptr, old_size ); + } + + debug_free( file, line, function, old_ptr ); + return new_ptr; +} + +void +debug_memory_reset(void) +{ + start_no = end_no; +} + +void +debug_memory_report(void) +{ + struct list_head *entry; + + entry = list.prev; + for (; entry != &list; entry = entry->prev) { + struct debug_memory_header *hdr; + void *ptr; + hdr = LIST_ENTRY(struct debug_memory_header, entry, head); + ptr = (void *)((char *)hdr + sizeof(*hdr)); + if(hdr->no >= start_no) + debug_printf("%s:%u:%s: %u byte st %p not freed\n", + hdr->file, hdr->line, hdr->function, + hdr->size, ptr); + } +} diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 14f8056924..c549513b6f 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -225,6 +225,29 @@ debug_dump_flags(const struct debug_named_value *names, unsigned long value); +void * +debug_malloc(const char *file, unsigned line, const char *function, + size_t size); + +void +debug_free(const char *file, unsigned line, const char *function, + void *ptr); + +void * +debug_calloc(const char *file, unsigned line, const char *function, + size_t count, size_t size ); + +void * +debug_realloc(const char *file, unsigned line, const char *function, + void *old_ptr, size_t old_size, size_t new_size ); + +void +debug_memory_reset(void); + +void +debug_memory_report(void); + + #ifdef __cplusplus } #endif diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index c2e0f8c6a5..d3b2fa2950 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -39,6 +39,22 @@ extern "C" { #endif +#if defined(WIN32) && defined(DEBUG) /* memory debugging */ + +#include "p_debug.h" + +#define MALLOC( _size ) \ + debug_malloc( __FILE__, __LINE__, __FUNCTION__, _size ) +#define CALLOC( _count, _size ) \ + debug_calloc(__FILE__, __LINE__, __FUNCTION__, _count, _size ) +#define FREE( _ptr ) \ + debug_free( __FILE__, __LINE__, __FUNCTION__, _ptr ) +#define REALLOC( _ptr, _old_size, _size ) \ + debug_realloc( __FILE__, __LINE__, __FUNCTION__, _ptr, _old_size, _size ) +#define GETENV( X ) NULL + +#else + #ifdef WIN32 void * __stdcall @@ -104,7 +120,7 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define GETENV( X ) NULL -#else /* WIN32 */ +#else /* !WIN32 */ #define MALLOC( SIZE ) malloc( SIZE ) @@ -116,7 +132,8 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define GETENV( X ) getenv( X ) -#endif /* WIN32 */ +#endif /* !WIN32 */ +#endif /* !DEBUG */ #define MALLOC_STRUCT(T) (struct T *) MALLOC(sizeof(struct T)) -- cgit v1.2.3 From d09b92d7e48ecee898ed158353021b7b3b6bae85 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 23 Mar 2008 18:52:37 +0000 Subject: gallium: Fix typo. --- src/gallium/auxiliary/util/p_debug_mem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c index 56599dafa6..c160afe5b7 100644 --- a/src/gallium/auxiliary/util/p_debug_mem.c +++ b/src/gallium/auxiliary/util/p_debug_mem.c @@ -165,7 +165,7 @@ debug_memory_report(void) hdr = LIST_ENTRY(struct debug_memory_header, entry, head); ptr = (void *)((char *)hdr + sizeof(*hdr)); if(hdr->no >= start_no) - debug_printf("%s:%u:%s: %u byte st %p not freed\n", + debug_printf("%s:%u:%s: %u bytes at %p not freed\n", hdr->file, hdr->line, hdr->function, hdr->size, ptr); } -- cgit v1.2.3 From 48ef11d308c395837c1685df6ab701a69507e8b9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 23 Mar 2008 18:57:35 +0000 Subject: gallium: Remove the debug_mask_* stuff. Overcomplex and not much different from using a global variable... --- src/gallium/auxiliary/util/p_debug.c | 34 ------------------- src/gallium/include/pipe/p_debug.h | 64 ------------------------------------ 2 files changed, 98 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index c51e9e6a69..19ad3f4663 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -168,40 +168,6 @@ void debug_assert_fail(const char *expr, const char *file, unsigned line) } -#define DEBUG_MASK_TABLE_SIZE 256 - - -/** - * Mask hash table. - * - * For now we just take the lower bits of the key, and do no attempt to solve - * collisions. Use a proper hash table when we have dozens of drivers. - */ -static uint32_t debug_mask_table[DEBUG_MASK_TABLE_SIZE]; - - -void debug_mask_set(uint32_t uuid, uint32_t mask) -{ - unsigned hash = uuid & (DEBUG_MASK_TABLE_SIZE - 1); - debug_mask_table[hash] = mask; -} - - -uint32_t debug_mask_get(uint32_t uuid) -{ - unsigned hash = uuid & (DEBUG_MASK_TABLE_SIZE - 1); - return debug_mask_table[hash]; -} - - -void debug_mask_vprintf(uint32_t uuid, uint32_t what, const char *format, va_list ap) -{ - uint32_t mask = debug_mask_get(uuid); - if(mask & what) - debug_vprintf(format, ap); -} - - const char * debug_dump_enum(const struct debug_named_value *names, unsigned long value) diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index c549513b6f..15fc2006d5 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -103,70 +103,6 @@ void debug_assert_fail(const char *expr, const char *file, unsigned line); #define assert(expr) debug_assert(expr) -/** - * Set a channel's debug mask. - * - * uuid is just a random 32 bit integer that uniquely identifies the debugging - * channel. - * - * @note Due to current implementation issues, make sure the lower 8 bits of - * UUID are unique. - */ -void debug_mask_set(uint32_t uuid, uint32_t mask); - - -uint32_t debug_mask_get(uint32_t uuid); - - -/** - * Conditional debug output. - * - * This is just a generalization of the debug filtering mechanism used - * throughout Gallium. - * - * You use this function as: - * - * @code - * #define MYDRIVER_UUID 0x12345678 // random 32 bit identifier - * - * static void inline - * mydriver_debug(uint32_t what, const char *format, ...) - * { - * #ifdef DEBUG - * va_list ap; - * va_start(ap, format); - * debug_mask_vprintf(MYDRIVER_UUID, what, format, ap); - * va_end(ap); - * #endif - * } - * - * ... - * - * debug_mask_set(MYDRIVER_UUID, - * MYDRIVER_DEBUG_THIS | - * MYDRIVER_DEBUG_THAT | - * ... ); - * - * ... - * - * mydriver_debug(MYDRIVER_DEBUG_THIS, - * "this and this happened\n"); - * - * mydriver_debug(MYDRIVER_DEBUG_THAT, - * "that = %f\n", that); - * ... - * @endcode - * - * You can also define several variants of mydriver_debug, with hardcoded what. - * Note that although macros with variable number of arguments would accomplish - * more in less code, they are not portable. - */ -void debug_mask_vprintf(uint32_t uuid, - uint32_t what, - const char *format, - va_list ap); - - #ifdef DEBUG #define debug_warning(__msg) \ debug_printf("%s:%i:warning: %s\n", __FILE__, __LINE__, (__msg)) -- cgit v1.2.3 From 38dc0f809d5f72b382676be6f72ea675a3929455 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 23 Mar 2008 19:38:11 +0000 Subject: gallium: Fix memory leak. pipe cso's were being destroyed, but the hash elements themselves not. proper fix is IMHO add a destructor callback to cso_hash. --- src/gallium/auxiliary/cso_cache/cso_cache.c | 41 ++++++++++++++++------------- 1 file changed, 22 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index a2764b4265..18acab0967 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -29,6 +29,7 @@ */ #include "pipe/p_util.h" +#include "pipe/p_debug.h" #include "cso_cache.h" #include "cso_hash.h" @@ -131,75 +132,77 @@ static int _cso_size_for_type(enum cso_cache_type type) static void delete_blend_state(void *state, void *data) { struct cso_blend *cso = (struct cso_blend *)state; - if (cso->delete_state && cso->data != &cso->state) + if (cso->delete_state) cso->delete_state(cso->context, cso->data); + FREE(state); } static void delete_depth_stencil_state(void *state, void *data) { struct cso_depth_stencil_alpha *cso = (struct cso_depth_stencil_alpha *)state; - if (cso->delete_state && cso->data != &cso->state) + if (cso->delete_state) cso->delete_state(cso->context, cso->data); + FREE(state); } static void delete_sampler_state(void *state, void *data) { struct cso_sampler *cso = (struct cso_sampler *)state; - if (cso->delete_state && cso->data != &cso->state) + if (cso->delete_state) cso->delete_state(cso->context, cso->data); + FREE(state); } static void delete_rasterizer_state(void *state, void *data) { struct cso_rasterizer *cso = (struct cso_rasterizer *)state; - if (cso->delete_state && cso->data != &cso->state) + if (cso->delete_state) cso->delete_state(cso->context, cso->data); + FREE(state); } static void delete_fs_state(void *state, void *data) { struct cso_fragment_shader *cso = (struct cso_fragment_shader *)state; - if (cso->delete_state && cso->data != &cso->state) + if (cso->delete_state) cso->delete_state(cso->context, cso->data); + FREE(state); } static void delete_vs_state(void *state, void *data) { struct cso_vertex_shader *cso = (struct cso_vertex_shader *)state; - if (cso->delete_state && cso->data != &cso->state) + if (cso->delete_state) cso->delete_state(cso->context, cso->data); + FREE(state); } static INLINE void delete_cso(void *state, enum cso_cache_type type) { switch (type) { - case CSO_BLEND: { + case CSO_BLEND: delete_blend_state(state, 0); - } break; - case CSO_SAMPLER: { + case CSO_SAMPLER: delete_sampler_state(state, 0); - } break; - case CSO_DEPTH_STENCIL_ALPHA: { + case CSO_DEPTH_STENCIL_ALPHA: delete_depth_stencil_state(state, 0); - } break; - case CSO_RASTERIZER: { + case CSO_RASTERIZER: delete_rasterizer_state(state, 0); - } break; - case CSO_FRAGMENT_SHADER: { + case CSO_FRAGMENT_SHADER: delete_fs_state(state, 0); - } break; - case CSO_VERTEX_SHADER: { + case CSO_VERTEX_SHADER: delete_vs_state(state, 0); - } break; + default: + assert(0); + FREE(state); } - FREE(state); } static INLINE void sanitize_hash(struct cso_hash *hash, enum cso_cache_type type, -- cgit v1.2.3 From 6edaef531806f4ac6c92c4a2934da5a37262e2e3 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 23 Mar 2008 20:39:35 +0100 Subject: gallium: Add util_gen_mipmap_filter(). We need a way to specify the type of minification filter used to downsample mipmap levels. The old util_gen_mipmap() retains its behaviour and uses LINEAR filter. --- src/gallium/auxiliary/util/u_gen_mipmap.c | 27 +++++++++++++++++++++++---- src/gallium/auxiliary/util/u_gen_mipmap.h | 5 +++++ 2 files changed, 28 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index cf02f00b1b..9203a78191 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -719,7 +719,6 @@ util_create_gen_mipmap(struct pipe_context *pipe, ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; - ctx->sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; ctx->sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; ctx->sampler.normalized_coords = 1; @@ -849,11 +848,12 @@ simple_viewport(struct pipe_context *pipe, uint width, uint height) * \param face which cube face to generate mipmaps for (0 for non-cube maps) * \param baseLevel the first mipmap level to use as a src * \param lastLevel the last mipmap level to generate + * \param filter the minification filter used to generate mipmap levels with */ void -util_gen_mipmap(struct gen_mipmap_state *ctx, - struct pipe_texture *pt, - uint face, uint baseLevel, uint lastLevel) +util_gen_mipmap_filter(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel, uint filter) { struct pipe_context *pipe = ctx->pipe; struct pipe_screen *screen = pipe->screen; @@ -914,6 +914,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, */ ctx->sampler.min_lod = ctx->sampler.max_lod = (float) srcLevel; ctx->sampler.lod_bias = (float) srcLevel; + ctx->sampler.min_img_filter = filter; cso_single_sampler(ctx->cso, 0, &ctx->sampler); cso_single_sampler_done(ctx->cso); #if 0 @@ -945,3 +946,21 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_restore_sampler_textures(ctx->cso); cso_restore_framebuffer(ctx->cso); } + + +/** + * Generate mipmap images with a linear minification filter. + * See util_gen_mipmap_filter for more info. + * + * \param pt the texture to generate mipmap levels for + * \param face which cube face to generate mipmaps for (0 for non-cube maps) + * \param baseLevel the first mipmap level to use as a src + * \param lastLevel the last mipmap level to generate + */ +void +util_gen_mipmap(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel) +{ + util_gen_mipmap_filter( ctx, pt, face, baseLevel, lastLevel, PIPE_TEX_FILTER_LINEAR ); +} diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.h b/src/gallium/auxiliary/util/u_gen_mipmap.h index eeabf3bf07..64abdeae98 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.h +++ b/src/gallium/auxiliary/util/u_gen_mipmap.h @@ -52,5 +52,10 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, struct pipe_texture *pt, uint face, uint baseLevel, uint lastLevel); +extern void +util_gen_mipmap_filter(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel, uint filter); + #endif -- cgit v1.2.3 From 899fcde366646a3ab3e42ae819620e790161ac75 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 23 Mar 2008 20:36:00 +0000 Subject: draw: restructure fetch/emit as a pair of function calls --- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 261 ++++++++++++++---------- 1 file changed, 152 insertions(+), 109 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 74a946545a..f43483f883 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -76,8 +76,8 @@ struct fetch_emit_middle_end { struct { const ubyte *ptr; unsigned pitch; - enum pipe_format src_format; - unsigned dst_format; + void (*fetch)( const void *from, float *attrib); + void (*emit)( const float *attrib, float **out ); } fetch[PIPE_ATTRIB_MAX]; unsigned nr_fetch; @@ -85,6 +85,98 @@ struct fetch_emit_middle_end { }; +static void fetch_B8G8R8A8_UNORM( const void *from, + float *attrib ) +{ + ubyte *ub = (ubyte *) from; + attrib[2] = UBYTE_TO_FLOAT(ub[0]); + attrib[1] = UBYTE_TO_FLOAT(ub[1]); + attrib[0] = UBYTE_TO_FLOAT(ub[2]); + attrib[3] = UBYTE_TO_FLOAT(ub[3]); +} + +static void fetch_R32G32B32A32_FLOAT( const void *from, + float *attrib ) +{ + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = f[1]; + attrib[2] = f[2]; + attrib[3] = f[3]; +} + +static void fetch_R32G32B32_FLOAT( const void *from, + float *attrib ) +{ + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = f[1]; + attrib[2] = f[2]; + attrib[3] = 1.0; +} + +static void fetch_R32G32_FLOAT( const void *from, + float *attrib ) +{ + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = f[1]; + attrib[2] = 0.0; + attrib[3] = 1.0; +} + +static void fetch_R32_FLOAT( const void *from, + float *attrib ) +{ + float *f = (float *) from; + attrib[0] = f[0]; + attrib[1] = 0.0; + attrib[2] = 0.0; + attrib[3] = 1.0; +} + +static void fetch_HEADER( const void *from, + float *attrib ) +{ + attrib[0] = 0; +} + + +static void emit_1F( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out) += 1; +} + +static void emit_2F( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out)[1] = attrib[1]; + (*out) += 2; +} + +static void emit_3F( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out)[1] = attrib[1]; + (*out)[2] = attrib[2]; + (*out) += 3; +} + +static void emit_4F( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out)[1] = attrib[1]; + (*out)[2] = attrib[2]; + (*out)[3] = attrib[3]; + (*out) += 4; +} + + /** * General-purpose fetch from user's vertex arrays, emit to driver's * vertex buffer. @@ -109,81 +201,8 @@ fetch_store_general( struct fetch_emit_middle_end *feme, const ubyte *from = (feme->fetch[j].ptr + feme->fetch[j].pitch * elt); - /* The normal fetch/emit code: - */ - switch (feme->fetch[j].src_format) { - case PIPE_FORMAT_B8G8R8A8_UNORM: - { - ubyte *ub = (ubyte *) from; - attrib[2] = UBYTE_TO_FLOAT(ub[0]); - attrib[1] = UBYTE_TO_FLOAT(ub[1]); - attrib[0] = UBYTE_TO_FLOAT(ub[2]); - attrib[3] = UBYTE_TO_FLOAT(ub[3]); - } - break; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - { - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = f[1]; - attrib[2] = f[2]; - attrib[3] = f[3]; - } - break; - case PIPE_FORMAT_R32G32B32_FLOAT: - { - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = f[1]; - attrib[2] = f[2]; - attrib[3] = 1.0; - } - break; - case PIPE_FORMAT_R32G32_FLOAT: - { - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = f[1]; - attrib[2] = 0.0; - attrib[3] = 1.0; - } - break; - case PIPE_FORMAT_R32_FLOAT: - { - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = 0.0; - attrib[2] = 0.0; - attrib[3] = 1.0; - } - break; - default: - assert(0); - } - - if (0) debug_printf("attrib %d: %f %f %f %f\n", j, - attrib[0], attrib[1], attrib[2], attrib[3]); - - switch (feme->fetch[j].dst_format) { - case EMIT_1F: - out[0] = attrib[0]; - out += 1; - break; - case EMIT_2F: - out[0] = attrib[0]; - out[1] = attrib[1]; - out += 2; - break; - case EMIT_4F: - out[0] = attrib[0]; - out[1] = attrib[1]; - out[2] = attrib[2]; - out[3] = attrib[3]; - out += 4; - break; - default: - assert(0); - } + feme->fetch[j].fetch( from, attrib ); + feme->fetch[j].emit( attrib, &out ); } } } @@ -199,44 +218,68 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle ) unsigned i; for (i = 0; i < nr_attrs; i++) { - enum pipe_format src_format; - unsigned dst_format; - const void *src_ptr; - unsigned pitch; + unsigned src_element = vinfo->src_index[i]; + unsigned src_buffer = draw->vertex_element[src_element].vertex_buffer_index; + + feme->fetch[i].ptr = ((const ubyte *)draw->user.vbuffer[src_buffer] + + draw->vertex_buffer[src_buffer].buffer_offset + + draw->vertex_element[src_element].src_offset); - if (vinfo->emit[i] == EMIT_HEADER) { - static const unsigned zero = 0; - src_ptr = &zero; - src_format = PIPE_FORMAT_R32_FLOAT; - pitch = 0; - dst_format = EMIT_1F; - } - else if (vinfo->emit[i] == EMIT_1F_PSIZE) { - src_ptr = &feme->draw->rasterizer->point_size; - src_format = PIPE_FORMAT_R32_FLOAT; - pitch = 0; - dst_format = EMIT_1F; + feme->fetch[i].pitch = draw->vertex_buffer[src_buffer].pitch; + + switch (draw->vertex_element[src_element].src_format) { + case PIPE_FORMAT_B8G8R8A8_UNORM: + feme->fetch[i].fetch = fetch_B8G8R8A8_UNORM; + break; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + feme->fetch[i].fetch = fetch_R32G32B32A32_FLOAT; + break; + case PIPE_FORMAT_R32G32B32_FLOAT: + feme->fetch[i].fetch = fetch_R32G32B32_FLOAT; + break; + case PIPE_FORMAT_R32G32_FLOAT: + feme->fetch[i].fetch = fetch_R32G32_FLOAT; + break; + case PIPE_FORMAT_R32_FLOAT: + feme->fetch[i].fetch = fetch_R32_FLOAT; + break; + default: + assert(0); + feme->fetch[i].fetch = NULL; + break; } - else { - unsigned src_element = vinfo->src_index[i]; - unsigned src_buffer = draw->vertex_element[src_element].vertex_buffer_index; - src_ptr = ((const ubyte *)draw->user.vbuffer[src_buffer] + - draw->vertex_buffer[src_buffer].buffer_offset + - draw->vertex_element[src_element].src_offset); - - src_format = draw->vertex_element[src_element].src_format; - pitch = draw->vertex_buffer[src_buffer].pitch; - dst_format = vinfo->emit[i]; + switch (vinfo->emit[i]) { + case EMIT_4F: + feme->fetch[i].emit = emit_4F; + break; + case EMIT_3F: + feme->fetch[i].emit = emit_3F; + break; + case EMIT_2F: + feme->fetch[i].emit = emit_2F; + break; + case EMIT_1F: + feme->fetch[i].emit = emit_1F; + break; + case EMIT_HEADER: + feme->fetch[i].ptr = NULL; + feme->fetch[i].pitch = 0; + feme->fetch[i].fetch = fetch_HEADER; + feme->fetch[i].emit = emit_1F; + break; + case EMIT_1F_PSIZE: + feme->fetch[i].ptr = (const ubyte *)&feme->draw->rasterizer->point_size; + feme->fetch[i].pitch = 0; + feme->fetch[i].fetch = fetch_R32_FLOAT; + feme->fetch[i].emit = emit_1F; + default: + assert(0); + feme->fetch[i].emit = NULL; + break; } - - feme->fetch[i].src_format = src_format; - feme->fetch[i].ptr = src_ptr; - feme->fetch[i].pitch = pitch; - feme->fetch[i].dst_format = dst_format; } - feme->nr_fetch = nr_attrs; feme->hw_vertex_size = vinfo->size * 4; } -- cgit v1.2.3 From 743e990831d74ff764f4677836b7bd9b044d39af Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 23 Mar 2008 21:31:14 +0000 Subject: draw: rename emit functions to match pipe_format names --- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 39 +++++++++++-------------- 1 file changed, 17 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index f43483f883..64ef83d800 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -135,30 +135,24 @@ static void fetch_R32_FLOAT( const void *from, attrib[3] = 1.0; } -static void fetch_HEADER( const void *from, - float *attrib ) -{ - attrib[0] = 0; -} - -static void emit_1F( const float *attrib, - float **out ) +static void emit_R32_FLOAT( const float *attrib, + float **out ) { (*out)[0] = attrib[0]; (*out) += 1; } -static void emit_2F( const float *attrib, - float **out ) +static void emit_R32G32_FLOAT( const float *attrib, + float **out ) { (*out)[0] = attrib[0]; (*out)[1] = attrib[1]; (*out) += 2; } -static void emit_3F( const float *attrib, - float **out ) +static void emit_R32G32B32_FLOAT( const float *attrib, + float **out ) { (*out)[0] = attrib[0]; (*out)[1] = attrib[1]; @@ -166,8 +160,8 @@ static void emit_3F( const float *attrib, (*out) += 3; } -static void emit_4F( const float *attrib, - float **out ) +static void emit_R32G32B32A32_FLOAT( const float *attrib, + float **out ) { (*out)[0] = attrib[0]; (*out)[1] = attrib[1]; @@ -211,6 +205,7 @@ fetch_store_general( struct fetch_emit_middle_end *feme, static void fetch_emit_prepare( struct draw_pt_middle_end *middle ) { + static const float zero = 0; struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; struct draw_context *draw = feme->draw; const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); @@ -251,28 +246,28 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle ) switch (vinfo->emit[i]) { case EMIT_4F: - feme->fetch[i].emit = emit_4F; + feme->fetch[i].emit = emit_R32G32B32A32_FLOAT; break; case EMIT_3F: - feme->fetch[i].emit = emit_3F; + feme->fetch[i].emit = emit_R32G32B32_FLOAT; break; case EMIT_2F: - feme->fetch[i].emit = emit_2F; + feme->fetch[i].emit = emit_R32G32_FLOAT; break; case EMIT_1F: - feme->fetch[i].emit = emit_1F; + feme->fetch[i].emit = emit_R32_FLOAT; break; case EMIT_HEADER: - feme->fetch[i].ptr = NULL; + feme->fetch[i].ptr = (const ubyte *)&zero; feme->fetch[i].pitch = 0; - feme->fetch[i].fetch = fetch_HEADER; - feme->fetch[i].emit = emit_1F; + feme->fetch[i].fetch = fetch_R32_FLOAT; + feme->fetch[i].emit = emit_R32_FLOAT; break; case EMIT_1F_PSIZE: feme->fetch[i].ptr = (const ubyte *)&feme->draw->rasterizer->point_size; feme->fetch[i].pitch = 0; feme->fetch[i].fetch = fetch_R32_FLOAT; - feme->fetch[i].emit = emit_1F; + feme->fetch[i].emit = emit_R32_FLOAT; default: assert(0); feme->fetch[i].emit = NULL; -- cgit v1.2.3 From e6ea786c003762143fdf4c63a6fc6a08f27e444e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 23 Mar 2008 21:31:23 +0000 Subject: draw: fix crlf --- src/gallium/auxiliary/draw/draw_pt.h | 232 +++++++++++++++--------------- src/gallium/auxiliary/draw/draw_pt_elts.c | 176 +++++++++++------------ 2 files changed, 204 insertions(+), 204 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index f56fa8d5cc..dfc29b2e7d 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -1,116 +1,116 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#ifndef DRAW_PT_H -#define DRAW_PT_H - -#include "pipe/p_compiler.h" - -typedef unsigned (*pt_elt_func)( const void *elts, unsigned idx ); - - -/* The "front end" - prepare sets of fetch, draw elements for the - * middle end. - * - * Currenly one version of this: - * - vcache - catchall implementation, decomposes to TRI/LINE/POINT prims - * Later: - * - varray, varray_split - * - velement, velement_split - * - * Currenly only using the vcache version. - */ -struct draw_pt_front_end { - void (*prepare)( struct draw_pt_front_end *, - struct draw_pt_middle_end * ); - - void (*run)( struct draw_pt_front_end *, - unsigned prim, - pt_elt_func elt_func, - const void *elt_ptr, - unsigned count ); - - void (*finish)( struct draw_pt_front_end * ); - void (*destroy)( struct draw_pt_front_end * ); -}; - - -/* The "middle end" - prepares actual hardware vertices for the - * hardware backend. - * - * Currently two versions of this: - * - fetch, vertex shade, cliptest, prim-pipeline - * - fetch, emit (ie passthrough) - * Later: - * - fetch, vertex shade, cliptest, maybe-pipeline, maybe-emit - * - fetch, vertex shade, emit - * - * Currenly only using the passthrough version. - */ -struct draw_pt_middle_end { - void (*prepare)( struct draw_pt_middle_end * ); - - void (*run)( struct draw_pt_middle_end *, - unsigned prim, - const unsigned *fetch_elts, - unsigned fetch_count, - const ushort *draw_elts, - unsigned draw_count ); - - void (*finish)( struct draw_pt_middle_end * ); - void (*destroy)( struct draw_pt_middle_end * ); -}; - - -/* The "back end" - supplied by the driver, defined in draw_vbuf.h. - * - * Not sure whether to wrap the prim pipeline up as an alternate - * backend. Would be a win for everything except pure passthrough - * mode... - */ -struct vbuf_render; - - -/* Helper functions. - */ -pt_elt_func draw_pt_elt_func( struct draw_context *draw ); -const void *draw_pt_elt_ptr( struct draw_context *draw, - unsigned start ); - -/* Implementations: - */ -struct draw_pt_front_end *draw_pt_vcache( void ); -struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); - - - -#endif +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#ifndef DRAW_PT_H +#define DRAW_PT_H + +#include "pipe/p_compiler.h" + +typedef unsigned (*pt_elt_func)( const void *elts, unsigned idx ); + + +/* The "front end" - prepare sets of fetch, draw elements for the + * middle end. + * + * Currenly one version of this: + * - vcache - catchall implementation, decomposes to TRI/LINE/POINT prims + * Later: + * - varray, varray_split + * - velement, velement_split + * + * Currenly only using the vcache version. + */ +struct draw_pt_front_end { + void (*prepare)( struct draw_pt_front_end *, + struct draw_pt_middle_end * ); + + void (*run)( struct draw_pt_front_end *, + unsigned prim, + pt_elt_func elt_func, + const void *elt_ptr, + unsigned count ); + + void (*finish)( struct draw_pt_front_end * ); + void (*destroy)( struct draw_pt_front_end * ); +}; + + +/* The "middle end" - prepares actual hardware vertices for the + * hardware backend. + * + * Currently two versions of this: + * - fetch, vertex shade, cliptest, prim-pipeline + * - fetch, emit (ie passthrough) + * Later: + * - fetch, vertex shade, cliptest, maybe-pipeline, maybe-emit + * - fetch, vertex shade, emit + * + * Currenly only using the passthrough version. + */ +struct draw_pt_middle_end { + void (*prepare)( struct draw_pt_middle_end * ); + + void (*run)( struct draw_pt_middle_end *, + unsigned prim, + const unsigned *fetch_elts, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ); + + void (*finish)( struct draw_pt_middle_end * ); + void (*destroy)( struct draw_pt_middle_end * ); +}; + + +/* The "back end" - supplied by the driver, defined in draw_vbuf.h. + * + * Not sure whether to wrap the prim pipeline up as an alternate + * backend. Would be a win for everything except pure passthrough + * mode... + */ +struct vbuf_render; + + +/* Helper functions. + */ +pt_elt_func draw_pt_elt_func( struct draw_context *draw ); +const void *draw_pt_elt_ptr( struct draw_context *draw, + unsigned start ); + +/* Implementations: + */ +struct draw_pt_front_end *draw_pt_vcache( void ); +struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); + + + +#endif diff --git a/src/gallium/auxiliary/draw/draw_pt_elts.c b/src/gallium/auxiliary/draw/draw_pt_elts.c index f337f71ea0..d49770e7b2 100644 --- a/src/gallium/auxiliary/draw/draw_pt_elts.c +++ b/src/gallium/auxiliary/draw/draw_pt_elts.c @@ -1,88 +1,88 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "draw/draw_pt.h" -#include "draw/draw_private.h" - -/* Neat get_elt func that also works for varrays drawing by encoding - * the start value into a pointer. - */ - -static unsigned elt_uint( const void *elts, unsigned idx ) -{ - return *(((const uint *)elts) + idx); -} - -static unsigned elt_ushort( const void *elts, unsigned idx ) -{ - return *(((const ushort *)elts) + idx); -} - -static unsigned elt_ubyte( const void *elts, unsigned idx ) -{ - return *(((const ubyte *)elts) + idx); -} - -static unsigned elt_vert( const void *elts, unsigned idx ) -{ - return (const ubyte *)elts - (const ubyte *)NULL + idx; -} - -pt_elt_func draw_pt_elt_func( struct draw_context *draw ) -{ - switch (draw->user.eltSize) { - case 0: return elt_vert; - case 1: return elt_ubyte; - case 2: return elt_ushort; - case 4: return elt_uint; - default: return NULL; - } -} - -const void *draw_pt_elt_ptr( struct draw_context *draw, - unsigned start ) -{ - const char *elts = draw->user.elts; - - switch (draw->user.eltSize) { - case 0: - return (const void *)(((const ubyte *)NULL) + start); - case 1: - return (const void *)(((const ubyte *)elts) + start); - case 2: - return (const void *)(((const ushort *)elts) + start); - case 4: - return (const void *)(((const uint *)elts) + start); - default: - return NULL; - } -} +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "draw/draw_pt.h" +#include "draw/draw_private.h" + +/* Neat get_elt func that also works for varrays drawing by encoding + * the start value into a pointer. + */ + +static unsigned elt_uint( const void *elts, unsigned idx ) +{ + return *(((const uint *)elts) + idx); +} + +static unsigned elt_ushort( const void *elts, unsigned idx ) +{ + return *(((const ushort *)elts) + idx); +} + +static unsigned elt_ubyte( const void *elts, unsigned idx ) +{ + return *(((const ubyte *)elts) + idx); +} + +static unsigned elt_vert( const void *elts, unsigned idx ) +{ + return (const ubyte *)elts - (const ubyte *)NULL + idx; +} + +pt_elt_func draw_pt_elt_func( struct draw_context *draw ) +{ + switch (draw->user.eltSize) { + case 0: return elt_vert; + case 1: return elt_ubyte; + case 2: return elt_ushort; + case 4: return elt_uint; + default: return NULL; + } +} + +const void *draw_pt_elt_ptr( struct draw_context *draw, + unsigned start ) +{ + const char *elts = draw->user.elts; + + switch (draw->user.eltSize) { + case 0: + return (const void *)(((const ubyte *)NULL) + start); + case 1: + return (const void *)(((const ubyte *)elts) + start); + case 2: + return (const void *)(((const ushort *)elts) + start); + case 4: + return (const void *)(((const uint *)elts) + start); + default: + return NULL; + } +} -- cgit v1.2.3 From 3e9b1bc1009ca1a2923f844d00bf478dc9d24644 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 23 Mar 2008 21:54:44 +0000 Subject: draw: tweak the definition of draw_need_pipeline, fix minor bug --- src/gallium/auxiliary/draw/draw_validate.c | 26 +++++++++++++++----------- 1 file changed, 15 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index 33e5559508..70b477ba5c 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -34,20 +34,14 @@ /** - * Check if we need any special pipeline stages, or whether prims/verts - * can go through untouched. + * Check if we need any special pipeline stages, or whether + * prims/verts can go through untouched. Don't test for bypass + * clipping or vs modes, this function is just about the primitive + * pipeline stages. */ boolean draw_need_pipeline(const struct draw_context *draw) { - /* clipping */ - if (!draw->rasterizer->bypass_clipping) - return TRUE; - - /* vertex shader */ - if (!draw->rasterizer->bypass_vs) - return TRUE; - /* line stipple */ if (draw->rasterizer->line_stipple_enable && draw->line_stipple) return TRUE; @@ -72,6 +66,11 @@ draw_need_pipeline(const struct draw_context *draw) if (draw->rasterizer->poly_stipple_enable && draw->pipeline.pstipple) return TRUE; + /* unfilled polygons */ + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) + return TRUE; + /* polygon offset */ if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) return TRUE; @@ -84,9 +83,14 @@ draw_need_pipeline(const struct draw_context *draw) if (draw->rasterizer->light_twoside) return TRUE; - /* polygon cull */ + /* polygon cull - this is difficult - hardware can cull just fine + * most of the time (though sometimes CULL_NEITHER is unsupported. + * + * Generally this isn't a reason to require the pipeline, though. + * if (draw->rasterizer->cull_mode) return TRUE; + */ return FALSE; } -- cgit v1.2.3 From 3b217c7fa77d397540e4b6299a35f586ad53a1a1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 23 Mar 2008 21:55:19 +0000 Subject: draw: check need_pipeline() in passthrough --- src/gallium/auxiliary/draw/draw_pt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 219b563422..d7169f78f4 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -114,7 +114,7 @@ draw_pt_arrays(struct draw_context *draw, middle = draw->pt.middle.fetch_shade_cliptest_pipeline; } #else - if (cliptest /*|| pipeline*/ || shading) + if (cliptest || pipeline || shading) return FALSE; middle = draw->pt.middle.fetch_emit; -- cgit v1.2.3 From a6d17bf671d6bfbb187a62ba14b9ad08fb5dafe1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Mon, 24 Mar 2008 09:56:12 +0000 Subject: draw: Fix #include order to fix Linux build. --- src/gallium/auxiliary/draw/draw_pt_elts.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_elts.c b/src/gallium/auxiliary/draw/draw_pt_elts.c index d49770e7b2..02fb6d9f41 100644 --- a/src/gallium/auxiliary/draw/draw_pt_elts.c +++ b/src/gallium/auxiliary/draw/draw_pt_elts.c @@ -30,8 +30,8 @@ * Keith Whitwell */ -#include "draw/draw_pt.h" #include "draw/draw_private.h" +#include "draw/draw_pt.h" /* Neat get_elt func that also works for varrays drawing by encoding * the start value into a pointer. -- cgit v1.2.3 From f4588c175280cd2b1f5341cd02bed7e9e7d1fd05 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 24 Mar 2008 10:52:20 +0000 Subject: Revert "draw: Fix #include order to fix Linux build." This reverts commit a6d17bf671d6bfbb187a62ba14b9ad08fb5dafe1. --- src/gallium/auxiliary/draw/draw_pt_elts.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_elts.c b/src/gallium/auxiliary/draw/draw_pt_elts.c index 02fb6d9f41..d49770e7b2 100644 --- a/src/gallium/auxiliary/draw/draw_pt_elts.c +++ b/src/gallium/auxiliary/draw/draw_pt_elts.c @@ -30,8 +30,8 @@ * Keith Whitwell */ -#include "draw/draw_private.h" #include "draw/draw_pt.h" +#include "draw/draw_private.h" /* Neat get_elt func that also works for varrays drawing by encoding * the start value into a pointer. -- cgit v1.2.3 From e0a9ce10f4c30b1aec08c46b6a33885782d40dd0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 24 Mar 2008 10:53:26 +0000 Subject: draw: pre-declare referenced structs --- src/gallium/auxiliary/draw/draw_pt.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index dfc29b2e7d..439fa4c881 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -37,6 +37,8 @@ typedef unsigned (*pt_elt_func)( const void *elts, unsigned idx ); +struct draw_pt_middle_end; +struct draw_context; /* The "front end" - prepare sets of fetch, draw elements for the * middle end. -- cgit v1.2.3 From c8c373514a18d31181cfaeefc23cd06eab3fddd0 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 24 Mar 2008 08:09:34 -0600 Subject: gallium: tweak coords in u_gen_mipmap code --- src/gallium/auxiliary/util/u_gen_mipmap.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 9203a78191..13e4531d4a 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -773,23 +773,23 @@ set_vertex_data(struct gen_mipmap_state *ctx, float width, float height) { void *buf; - ctx->vertices[0][0][0] = 0.0f; /*x*/ - ctx->vertices[0][0][1] = 0.0f; /*y*/ + ctx->vertices[0][0][0] = -0.5f; /*x*/ + ctx->vertices[0][0][1] = -0.5f; /*y*/ ctx->vertices[0][1][0] = 0.0f; /*s*/ ctx->vertices[0][1][1] = 0.0f; /*t*/ - ctx->vertices[1][0][0] = width; /*x*/ - ctx->vertices[1][0][1] = 0.0f; /*y*/ + ctx->vertices[1][0][0] = width - 0.5f; /*x*/ + ctx->vertices[1][0][1] = -0.5f; /*y*/ ctx->vertices[1][1][0] = 1.0f; /*s*/ ctx->vertices[1][1][1] = 0.0f; /*t*/ - ctx->vertices[2][0][0] = width; - ctx->vertices[2][0][1] = height; + ctx->vertices[2][0][0] = width - 0.5f; + ctx->vertices[2][0][1] = height - 0.5f; ctx->vertices[2][1][0] = 1.0f; ctx->vertices[2][1][1] = 1.0f; - ctx->vertices[3][0][0] = 0.0f; - ctx->vertices[3][0][1] = height; + ctx->vertices[3][0][0] = -0.5f; + ctx->vertices[3][0][1] = height - 0.5f; ctx->vertices[3][1][0] = 0.0f; ctx->vertices[3][1][1] = 1.0f; -- cgit v1.2.3 From 110b63d00fa0a555a00f5b1560452323517eafe1 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 08:53:16 -0600 Subject: gallium: pass the filter mode to util_gen_mipmap(). Remove util_gen_mipmap_filter() when no longer used. --- src/gallium/auxiliary/util/u_gen_mipmap.c | 25 ++++++++++--------------- src/gallium/auxiliary/util/u_gen_mipmap.h | 2 +- src/mesa/state_tracker/st_gen_mipmap.c | 3 ++- 3 files changed, 13 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 13e4531d4a..26df5f29f8 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -719,7 +719,6 @@ util_create_gen_mipmap(struct pipe_context *pipe, ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; - ctx->sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; ctx->sampler.normalized_coords = 1; @@ -849,11 +848,12 @@ simple_viewport(struct pipe_context *pipe, uint width, uint height) * \param baseLevel the first mipmap level to use as a src * \param lastLevel the last mipmap level to generate * \param filter the minification filter used to generate mipmap levels with + * \param filter one of PIPE_TEX_FILTER_LINEAR, PIPE_TEX_FILTER_NEAREST */ void -util_gen_mipmap_filter(struct gen_mipmap_state *ctx, - struct pipe_texture *pt, - uint face, uint baseLevel, uint lastLevel, uint filter) +util_gen_mipmap(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel, uint filter) { struct pipe_context *pipe = ctx->pipe; struct pipe_screen *screen = pipe->screen; @@ -914,6 +914,7 @@ util_gen_mipmap_filter(struct gen_mipmap_state *ctx, */ ctx->sampler.min_lod = ctx->sampler.max_lod = (float) srcLevel; ctx->sampler.lod_bias = (float) srcLevel; + ctx->sampler.mag_img_filter = filter; ctx->sampler.min_img_filter = filter; cso_single_sampler(ctx->cso, 0, &ctx->sampler); cso_single_sampler_done(ctx->cso); @@ -949,18 +950,12 @@ util_gen_mipmap_filter(struct gen_mipmap_state *ctx, /** - * Generate mipmap images with a linear minification filter. - * See util_gen_mipmap_filter for more info. - * - * \param pt the texture to generate mipmap levels for - * \param face which cube face to generate mipmaps for (0 for non-cube maps) - * \param baseLevel the first mipmap level to use as a src - * \param lastLevel the last mipmap level to generate + * XXX remove this */ void -util_gen_mipmap(struct gen_mipmap_state *ctx, - struct pipe_texture *pt, - uint face, uint baseLevel, uint lastLevel) +util_gen_mipmap_filter(struct gen_mipmap_state *ctx, + struct pipe_texture *pt, + uint face, uint baseLevel, uint lastLevel, uint filter) { - util_gen_mipmap_filter( ctx, pt, face, baseLevel, lastLevel, PIPE_TEX_FILTER_LINEAR ); + util_gen_mipmap_filter( ctx, pt, face, baseLevel, lastLevel, filter ); } diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.h b/src/gallium/auxiliary/util/u_gen_mipmap.h index 64abdeae98..a5df8481bf 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.h +++ b/src/gallium/auxiliary/util/u_gen_mipmap.h @@ -50,7 +50,7 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx); extern void util_gen_mipmap(struct gen_mipmap_state *ctx, struct pipe_texture *pt, - uint face, uint baseLevel, uint lastLevel); + uint face, uint baseLevel, uint lastLevel, uint filter); extern void util_gen_mipmap_filter(struct gen_mipmap_state *ctx, diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 61e1d9621c..a931911227 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -92,7 +92,8 @@ st_render_mipmap(struct st_context *st, return FALSE; } - util_gen_mipmap(st->gen_mipmap, pt, face, baseLevel, lastLevel); + util_gen_mipmap(st->gen_mipmap, pt, face, baseLevel, lastLevel, + PIPE_TEX_FILTER_LINEAR); /* shaders don't go through CSO yet */ if (st->fp) -- cgit v1.2.3 From bf8de6d4dc1f956b7e70be285ff0d983b1a8eb5b Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 24 Mar 2008 16:50:39 +0100 Subject: gallium: Remove util_gen_mipmap_filter(). --- src/gallium/auxiliary/util/u_gen_mipmap.c | 12 ------------ src/gallium/auxiliary/util/u_gen_mipmap.h | 6 ------ 2 files changed, 18 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 26df5f29f8..427b217755 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -947,15 +947,3 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_restore_sampler_textures(ctx->cso); cso_restore_framebuffer(ctx->cso); } - - -/** - * XXX remove this - */ -void -util_gen_mipmap_filter(struct gen_mipmap_state *ctx, - struct pipe_texture *pt, - uint face, uint baseLevel, uint lastLevel, uint filter) -{ - util_gen_mipmap_filter( ctx, pt, face, baseLevel, lastLevel, filter ); -} diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.h b/src/gallium/auxiliary/util/u_gen_mipmap.h index a5df8481bf..bd9af54fb7 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.h +++ b/src/gallium/auxiliary/util/u_gen_mipmap.h @@ -52,10 +52,4 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, struct pipe_texture *pt, uint face, uint baseLevel, uint lastLevel, uint filter); -extern void -util_gen_mipmap_filter(struct gen_mipmap_state *ctx, - struct pipe_texture *pt, - uint face, uint baseLevel, uint lastLevel, uint filter); - - #endif -- cgit v1.2.3 From d246274b9c48c01b61490f0a76c22aebb8e64e1e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 24 Mar 2008 11:46:41 -0600 Subject: gallium: add missing cso_save_samplers() call --- src/mesa/state_tracker/st_cb_bitmap.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index acc22d4323..dbf2f77c2e 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -406,6 +406,7 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, assert(height <= maxSize); cso_save_rasterizer(cso); + cso_save_samplers(cso); //cso_save_viewport(cso); /* rasterizer state: just scissor */ -- cgit v1.2.3 From 22a3d022542fed267d48b48a44e10226d55c8afd Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 24 Mar 2008 11:55:29 -0600 Subject: gallium: free bitmap fragment shaders, misc clean-up --- src/mesa/state_tracker/st_cb_bitmap.c | 12 ++++++++---- src/mesa/state_tracker/st_cb_program.c | 5 +++-- src/mesa/state_tracker/st_cb_program.h | 3 +++ src/mesa/state_tracker/st_context.c | 1 + src/mesa/state_tracker/st_program.h | 4 ---- 5 files changed, 15 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index dbf2f77c2e..6e59439874 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -44,6 +44,7 @@ #include "st_atom_constbuf.h" #include "st_program.h" #include "st_cb_bitmap.h" +#include "st_cb_program.h" #include "st_mesa_to_tgsi.h" #include "st_texture.h" #include "pipe/p_context.h" @@ -407,7 +408,6 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_save_rasterizer(cso); cso_save_samplers(cso); - //cso_save_viewport(cso); /* rasterizer state: just scissor */ { @@ -458,11 +458,9 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* restore state */ cso_restore_rasterizer(cso); cso_restore_samplers(cso); - //cso_restore_viewport(cso); /* shaders don't go through cso yet */ pipe->bind_fs_state(pipe, st->fp->driver_shader); pipe->bind_vs_state(pipe, st->vp->driver_shader); - pipe->set_sampler_textures(pipe, ctx->st->state.num_textures, ctx->st->state.sampler_texture); } @@ -514,7 +512,13 @@ st_destroy_bitmap(struct st_context *st) { struct pipe_context *pipe = st->pipe; - /* XXX free frag shader state */ + if (st->bitmap.combined_prog) { + st_delete_program(st->ctx, &st->bitmap.combined_prog->Base.Base); + } + + if (st->bitmap.program) { + st_delete_program(st->ctx, &st->bitmap.program->Base.Base); + } if (st->bitmap.vs) { pipe->delete_vs_state(pipe, st->bitmap.vs); diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index a739fcd336..003ec0d2b7 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -45,6 +45,7 @@ #include "st_context.h" #include "st_program.h" #include "st_atom_shader.h" +#include "st_cb_program.h" static GLuint SerialNo = 1; @@ -122,8 +123,8 @@ static struct gl_program *st_new_program( GLcontext *ctx, } -static void st_delete_program( GLcontext *ctx, - struct gl_program *prog ) +void +st_delete_program(GLcontext *ctx, struct gl_program *prog) { struct st_context *st = st_context(ctx); struct pipe_context *pipe = st->pipe; diff --git a/src/mesa/state_tracker/st_cb_program.h b/src/mesa/state_tracker/st_cb_program.h index 45dc753dff..0de96f2fd2 100644 --- a/src/mesa/state_tracker/st_cb_program.h +++ b/src/mesa/state_tracker/st_cb_program.h @@ -32,5 +32,8 @@ extern void st_init_program_functions(struct dd_function_table *functions); +extern void +st_delete_program(GLcontext *ctx, struct gl_program *prog); + #endif diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index d9e8722976..7c18386575 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -42,6 +42,7 @@ #include "st_cb_drawpixels.h" #include "st_cb_fbo.h" #include "st_cb_feedback.h" +#include "st_cb_program.h" #include "st_cb_queryobj.h" #include "st_cb_rasterpos.h" #include "st_cb_readpixels.h" diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 9ef2a07eaa..63d6590540 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -97,10 +97,6 @@ struct st_vertex_program }; -extern void -st_init_program_functions(struct dd_function_table *functions); - - static inline struct st_fragment_program * st_fragment_program( struct gl_fragment_program *fp ) { -- cgit v1.2.3 From 48726e129a3727ec01e130835e74d62007093051 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 24 Mar 2008 18:25:31 +0000 Subject: gallium: Protect operator precedence in macro. --- src/gallium/include/pipe/p_util.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index d3b2fa2950..283f7fb8f5 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -201,7 +201,7 @@ mem_dup(const void *src, uint size) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) -#define Elements(x) sizeof(x)/sizeof(*(x)) +#define Elements(x) (sizeof(x)/sizeof((x)[0])) #define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER)) /** -- cgit v1.2.3 From 4e398df6825f4b6d568f055b689e7d7043e2785e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 12:50:28 -0600 Subject: gallium: fix REALLOC() to copy smaller of old/new sizes --- src/gallium/include/pipe/p_util.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 283f7fb8f5..505626c60b 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -107,10 +107,10 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) void *new_ptr = NULL; if (new_size != 0) { + unsigned copy_size = old_size < new_size ? old_size : new_size; new_ptr = MALLOC( new_size ); - - if( new_ptr && old_ptr ) { - memcpy( new_ptr, old_ptr, old_size ); + if (new_ptr && old_ptr && copy_size) { + memcpy( new_ptr, old_ptr, copy_size ); } } -- cgit v1.2.3 From d83e75c75958673e8019475f5ba5c2cff9b8415f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 12:51:50 -0600 Subject: gallium: move filter assignment out of loop --- src/gallium/auxiliary/util/u_gen_mipmap.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 427b217755..e129c062be 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -890,6 +890,10 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, memset(&fb, 0, sizeof(fb)); fb.num_cbufs = 1; + /* set min/mag to same filter for faster sw speed */ + ctx->sampler.mag_img_filter = filter; + ctx->sampler.min_img_filter = filter; + /* * XXX for small mipmap levels, it may be faster to use the software * fallback path... @@ -914,8 +918,6 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, */ ctx->sampler.min_lod = ctx->sampler.max_lod = (float) srcLevel; ctx->sampler.lod_bias = (float) srcLevel; - ctx->sampler.mag_img_filter = filter; - ctx->sampler.min_img_filter = filter; cso_single_sampler(ctx->cso, 0, &ctx->sampler); cso_single_sampler_done(ctx->cso); #if 0 -- cgit v1.2.3 From dd51365acdd515577ee76850ceda01347ceb27c0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 24 Mar 2008 20:18:59 +0000 Subject: gallium: cleanup p_debug Now debug_printf is disabled on release builds. Use debug_error or _debug_printf to output messages on release versions. --- src/gallium/auxiliary/util/p_debug.c | 31 ++++------ src/gallium/include/pipe/p_debug.h | 117 +++++++++++++++++++++++++++++++---- 2 files changed, 118 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 19ad3f4663..bd58e0e3a8 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -58,7 +58,7 @@ int rpl_snprintf(char *str, size_t size, const char *format, ...); #endif -void debug_vprintf(const char *format, va_list ap) +void _debug_vprintf(const char *format, va_list ap) { #ifdef WIN32 #ifndef WINCE @@ -76,15 +76,7 @@ void debug_vprintf(const char *format, va_list ap) } -void debug_printf(const char *format, ...) -{ - va_list ap; - va_start(ap, format); - debug_vprintf(format, ap); - va_end(ap); -} - - +#ifdef DEBUG void debug_print_blob( const char *name, const void *blob, unsigned size ) @@ -99,12 +91,10 @@ void debug_print_blob( const char *name, debug_printf("%d:\t%08x\n", i, ublob[i]); } } +#endif -/* TODO: implement a debug_abort that calls EngBugCheckEx on WIN32 */ - - -static INLINE void debug_break(void) +void _debug_break(void) { #if (defined(__i386__) || defined(__386__)) && defined(__GNUC__) __asm("int3"); @@ -155,15 +145,18 @@ static unsigned abort_en() } #endif -void debug_assert_fail(const char *expr, const char *file, unsigned line) +void _debug_assert_fail(const char *expr, + const char *file, + unsigned line, + const char *function) { - debug_printf("%s:%i: Assertion `%s' failed.\n", file, line, expr); + _debug_printf("%s:%u:%s: Assertion `%s' failed.\n", file, line, function, expr); if (abort_en()) { debug_break(); } else { - debug_printf("continuing...\n"); + _debug_printf("continuing...\n"); } } @@ -180,7 +173,7 @@ debug_dump_enum(const struct debug_named_value *names, ++names; } - snprintf(rest, sizeof(rest), "0x%08x", value); + snprintf(rest, sizeof(rest), "0x%08lx", value); return rest; } @@ -213,7 +206,7 @@ debug_dump_flags(const struct debug_named_value *names, else first = 0; - snprintf(rest, sizeof(rest), "0x%08x", value); + snprintf(rest, sizeof(rest), "0x%08lx", value); strncat(output, rest, sizeof(output)); } diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 15fc2006d5..494cc3bb71 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -60,58 +60,153 @@ extern "C" { #endif +void _debug_vprintf(const char *format, va_list ap); + + +static void INLINE +_debug_printf(const char *format, ...) +{ + va_list ap; + va_start(ap, format); + _debug_vprintf(format, ap); + va_end(ap); +} + + /** * Print debug messages. * - * A debug message will be printed regardless of the DEBUG/NDEBUG macros. - * * The actual channel used to output debug message is platform specific. To * avoid misformating or truncation, follow these rules of thumb: * - output whole lines * - avoid outputing large strings (512 bytes is the current maximum length * that is guaranteed to be printed in all platforms) */ -void debug_printf(const char *format, ...); +static void INLINE +debug_printf(const char *format, ...) +{ +#ifdef DEBUG + va_list ap; + va_start(ap, format); + _debug_vprintf(format, ap); + va_end(ap); +#endif +} + +#ifdef DEBUG +#define debug_vprintf(_format, _ap) _debug_vprintf(_format, _ap) +#else +#define debug_vprintf(_format, _ap) ((void)0) +#endif -/* Dump a blob in hex to the same place that debug_printf sends its - * messages: + +/** + * Dump a blob in hex to the same place that debug_printf sends its + * messages. */ +#ifdef DEBUG void debug_print_blob( const char *name, const void *blob, unsigned size ); +#else +#define debug_print_blob(_name, _blob, _size) ((void)0) +#endif + + +void _debug_break(void); + /** - * @sa debug_printf + * Hard-coded breakpoint. */ -void debug_vprintf(const char *format, va_list ap); +#ifdef DEBUG +#if (defined(__i386__) || defined(__386__)) && defined(__GNUC__) +#define debug_break() __asm("int3") +#elif (defined(__i386__) || defined(__386__)) && defined(__MSC__) +#define debug_break() _asm {int 3} +#else +#define debug_break() _debug_break() +#endif +#else /* !DEBUG */ +#define debug_break() ((void)0) +#endif /* !DEBUG */ + -void debug_assert_fail(const char *expr, const char *file, unsigned line); +void _debug_assert_fail(const char *expr, + const char *file, + unsigned line, + const char *function); -/** Assert macro */ +/** + * Assert macro + * + * Do not expect that the assert call terminates -- errors must be handled + * regardless of assert behavior. + */ #ifdef DEBUG -#define debug_assert(expr) ((expr) ? (void)0 : debug_assert_fail(#expr, __FILE__, __LINE__)) +#define debug_assert(expr) ((expr) ? (void)0 : _debug_assert_fail(#expr, __FILE__, __LINE__, __FUNCTION__)) #else #define debug_assert(expr) ((void)0) #endif +/** Override standard assert macro */ #ifdef assert #undef assert #endif #define assert(expr) debug_assert(expr) +/** + * Output the current function name. + */ +#ifdef DEBUG +#define debug_checkpoint() \ + _debug_printf("%s\n", __FUNCTION__) +#else +#define debug_checkpoint() \ + ((void)0) +#endif + + +/** + * Output the full source code position. + */ +#ifdef DEBUG +#define debug_checkpoint_full() \ + debug_printf("%s:%u:%s", __FILE__, __LINE__, __FUNCTION__) +#else +#define debug_checkpoint_full() \ + ((void)0) +#endif + + +/** + * Output a warning message. Muted on release version. + */ #ifdef DEBUG #define debug_warning(__msg) \ - debug_printf("%s:%i:warning: %s\n", __FILE__, __LINE__, (__msg)) + _debug_printf("%s:%u:%s: warning: %s\n", __FILE__, __LINE__, __FUNCTION__, (__msg)) #else #define debug_warning(__msg) \ ((void)0) #endif +/** + * Output an error message. Not muted on release version. + */ +#ifdef DEBUG +#define debug_error(__msg) \ + _debug_printf("%s:%u:%s: error: %s\n", __FILE__, __LINE__, __FUNCTION__, (__msg)) +#else +#define debug_error(__msg) \ + _debug_printf("error: %s\n", __msg)) +#endif + + /** * Used by debug_dump_enum and debug_dump_flags to describe symbols. */ -- cgit v1.2.3 From e8c6ea4f608296ed976f1597ffd46ac0b42fc84a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 24 Mar 2008 22:30:33 +0000 Subject: gallium: Simple cross platform get-opt system. Uses getenv on Linux, and a memory mapped text file on Windows. It supports boolean options, flags, and plain strings. --- src/gallium/auxiliary/util/p_debug.c | 144 +++++++++++++++++++++++++++++++++++ src/gallium/include/pipe/p_debug.h | 27 +++++++ 2 files changed, 171 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index bd58e0e3a8..351de95c95 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -107,6 +107,150 @@ void _debug_break(void) #endif } + +#ifdef WIN32 +static const char * +find(const char *start, const char *end, char c) +{ + const char *p; + for(p = start; !end || p != end; ++p) { + if(*p == c) + return p; + if(*p < 32) + break; + } + return NULL; +} + +static int +compare(const char *start, const char *end, const char *s) +{ + const char *p, *q; + for(p = start, q = s; p != end && *q != '\0'; ++p, ++q) { + if(*p != *q) + return 0; + } + return p == end && *q == '\0'; +} + +static void +copy(char *dst, const char *start, const char *end, size_t n) +{ + const char *p; + char *q; + for(p = start, q = dst, n = n - 1; p != end && n; ++p, ++q, --n) + *q = *p; + *q = '\0'; +} +#endif + + +const char * +debug_get_option(const char *name, const char *dfault) +{ + const char *result; +#ifdef WIN32 + ULONG_PTR iFile = 0; + const void *pMap = NULL; + const char *sol, *eol, *sep; + static char output[1024]; + + pMap = EngMapFile(L"\\??\\c:\\gallium.cfg", 0, &iFile); + if(!pMap) + result = dfault; + else { + sol = (const char *)pMap; + while(1) { + /* TODO: handle LF line endings */ + eol = find(sol, NULL, '\r'); + if(!eol || eol == sol) + break; + sep = find(sol, eol, '='); + if(!sep) + break; + if(compare(sol, sep, name)) { + copy(output, sep + 1, eol, sizeof(output)); + result = output; + break; + } + sol = eol + 2; + } + EngUnmapFile(iFile); + } +#else + + result = getenv(name); + if(!result) + result = dfault; +#endif + + if(result) + debug_printf("%s: %s = %s\n", __FUNCTION__, name, result); + else + debug_printf("%s: %s = (null)\n", __FUNCTION__, name); + + return result; +} + +boolean +debug_get_bool_option(const char *name, boolean dfault) +{ + const char *str = debug_get_option(name, NULL); + boolean result; + + if(str == NULL) + result = dfault; + else if(!strcmp(str, "no")) + result = FALSE; + else if(!strcmp(str, "0")) + result = FALSE; + else if(!strcmp(str, "f")) + result = FALSE; + else if(!strcmp(str, "false")) + result = FALSE; + else + result = TRUE; + + debug_printf("%s: %s = %s\n", __FUNCTION__, name, result ? "TRUE" : "FALSE"); + + return result; +} + + +long +debug_get_num_option(const char *name, long dfault) +{ + /* FIXME */ + return dfault; +} + + +unsigned long +debug_get_flags_option(const char *name, + const struct debug_named_value *flags, + unsigned long dfault) +{ + unsigned long result; + const char *str; + + str = debug_get_option(name, NULL); + if(!str) + result = dfault; + else { + result = 0; + while( flags->name ) { + if (!strcmp(str, "all") || strstr(str, flags->name )) + result |= flags->value; + ++flags; + } + } + + debug_printf("%s: %s = 0x%lx\n", __FUNCTION__, name, result); + + return result; +} + + #if defined(WIN32) ULONG_PTR debug_config_file = 0; void *mapped_config_file = 0; diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 494cc3bb71..95bdf2d3b9 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -256,6 +256,33 @@ debug_dump_flags(const struct debug_named_value *names, unsigned long value); +/** + * Get option. + * + * It is an alias for getenv on Linux. + * + * On Windows it reads C:\gallium.cfg, which is a text file with CR+LF line + * endings with one option per line as + * + * NAME=value + * + * This file must be terminated with an extra empty line. + */ +const char * +debug_get_option(const char *name, const char *dfault); + +boolean +debug_get_bool_option(const char *name, boolean dfault); + +long +debug_get_unsigned_option(const char *name, long dfault); + +unsigned long +debug_get_flags_option(const char *name, + const struct debug_named_value *flags, + unsigned long dfault); + + void * debug_malloc(const char *file, unsigned line, const char *function, size_t size); -- cgit v1.2.3 From 648e26aa95b519f1f4abc429b5a23abaf4a5195b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 16:24:32 -0600 Subject: gallium: added tgsi_num_tokens() function to return number of tokens in token array. Maybe move to a different file someday. --- src/gallium/auxiliary/tgsi/util/tgsi_parse.c | 13 +++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_parse.h | 4 ++++ 2 files changed, 17 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c index bf6b89ce56..c3526cb71f 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c @@ -317,3 +317,16 @@ tgsi_parse_token( } } + +unsigned +tgsi_num_tokens(const struct tgsi_token *tokens) +{ + struct tgsi_parse_context ctx; + if (tgsi_parse_init(&ctx, tokens) == TGSI_PARSE_OK) { + unsigned len = (ctx.FullHeader.Header.HeaderSize + + ctx.FullHeader.Header.BodySize + + 1); + return len; + } + return 0; +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h index 40083728d6..a98e88e343 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h @@ -113,6 +113,10 @@ void tgsi_parse_token( struct tgsi_parse_context *ctx ); +unsigned +tgsi_num_tokens(const struct tgsi_token *tokens); + + #if defined __cplusplus } #endif -- cgit v1.2.3 From 7f430293772f201a59bcf62edd1ed4f942f8be29 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 16:26:45 -0600 Subject: gallium: use pipe_texture_reference() in a few places (fixes refcounting bugs) --- src/gallium/auxiliary/draw/draw_aaline.c | 9 +++++++-- src/gallium/auxiliary/draw/draw_pstipple.c | 12 ++++++++++-- 2 files changed, 17 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index 6742f7f4b9..cc1873abad 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -621,7 +621,7 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) bind_aaline_fragment_shader(aaline); aaline->state.sampler[num] = aaline->sampler_cso; - aaline->state.texture[num] = aaline->texture; + pipe_texture_reference(&aaline->state.texture[num], aaline->texture); aaline->driver_bind_sampler_states(pipe, num + 1, aaline->state.sampler); aaline->driver_set_sampler_textures(pipe, num + 1, aaline->state.texture); @@ -769,9 +769,14 @@ aaline_set_sampler_textures(struct pipe_context *pipe, unsigned num, struct pipe_texture **texture) { struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + uint i; + /* save current */ - memcpy(aaline->state.texture, texture, num * sizeof(struct pipe_texture *)); + for (i = 0; i < num; i++) { + pipe_texture_reference(&aaline->state.texture[i], texture[i]); + } aaline->num_textures = num; + /* pass-through */ aaline->driver_set_sampler_textures(aaline->pipe, num, texture); } diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index bd8d3a76ae..f00b23959b 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -398,6 +398,7 @@ pstip_create_texture(struct pstip_stage *pstip) texTemp.cpp = 1; pstip->texture = screen->texture_create(screen, &texTemp); + assert(pstip->texture->refcount == 1); //pstip_update_texture(pstip); } @@ -492,7 +493,8 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) /* plug in our sampler, texture */ pstip->state.samplers[pstip->sampler_unit] = pstip->sampler_cso; - pstip->state.textures[pstip->sampler_unit] = pstip->texture; + pipe_texture_reference(&pstip->state.textures[pstip->sampler_unit], + pstip->texture); pstip->driver_bind_sampler_states(pipe, num_samplers, pstip->state.samplers); pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures); @@ -624,6 +626,7 @@ pstip_bind_sampler_states(struct pipe_context *pipe, unsigned num, void **sampler) { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + /* save current */ memcpy(pstip->state.samplers, sampler, num * sizeof(void *)); pstip->num_samplers = num; @@ -637,9 +640,14 @@ pstip_set_sampler_textures(struct pipe_context *pipe, unsigned num, struct pipe_texture **texture) { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + uint i; + /* save current */ - memcpy(pstip->state.textures, texture, num * sizeof(struct pipe_texture *)); + for (i = 0; i < num; i++) { + pipe_texture_reference(&pstip->state.textures[i], texture[i]); + } pstip->num_textures = num; + /* pass-through */ pstip->driver_set_sampler_textures(pstip->pipe, num, texture); } -- cgit v1.2.3 From ae146e4bc86aeade59d018100e39e160f7553994 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 16:31:15 -0600 Subject: gallium: make a copy of the vertex shader's token array. This solves problems when the state tracker frees the token array when the draw module still needs it. --- src/gallium/auxiliary/draw/draw_vs_exec.c | 7 ++++++- src/gallium/auxiliary/draw/draw_vs_llvm.c | 7 ++++++- src/gallium/auxiliary/draw/draw_vs_sse.c | 6 +++++- 3 files changed, 17 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 364693e0b4..4e2fa72707 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -38,6 +38,8 @@ #include "draw_context.h" #include "draw_vs.h" +#include "tgsi/util/tgsi_parse.h" + static INLINE unsigned compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) @@ -187,6 +189,7 @@ vs_exec_run( struct draw_vertex_shader *shader, static void vs_exec_delete( struct draw_vertex_shader *dvs ) { + FREE((void*) dvs->state.tokens); FREE( dvs ); } @@ -196,11 +199,13 @@ draw_create_vs_exec(struct draw_context *draw, const struct pipe_shader_state *state) { struct draw_vertex_shader *vs = CALLOC_STRUCT( draw_vertex_shader ); + uint nt = tgsi_num_tokens(state->tokens); if (vs == NULL) return NULL; - vs->state = *state; + /* we make a private copy of the tokens */ + vs->state.tokens = mem_dup(state->tokens, nt * sizeof(state->tokens[0])); vs->prepare = vs_exec_prepare; vs->run = vs_exec_run; vs->delete = vs_exec_delete; diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 53c260be53..bd983f2ddf 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -38,6 +38,8 @@ #include "draw_context.h" #include "draw_vs.h" +#include "tgsi/util/tgsi_parse.h" + #ifdef MESA_LLVM #include "gallivm/gallivm.h" @@ -186,6 +188,7 @@ vs_llvm_delete( struct draw_vertex_shader *base ) /* Do something to free compiled shader: */ + FREE( (void*) shader->base.state.tokens ); FREE( shader ); } @@ -197,12 +200,14 @@ draw_create_vs_llvm(struct draw_context *draw, const struct pipe_shader_state *templ) { struct draw_llvm_vertex_shader *vs; + uint nt = tgsi_num_tokens(templ->tokens); vs = CALLOC_STRUCT( draw_llvm_vertex_shader ); if (vs == NULL) return NULL; - vs->base.state = templ; + /* we make a private copy of the tokens */ + vs->base.state.tokens = mem_dup(templ->tokens, nt * sizeof(templ->tokens[0])); vs->base.prepare = vs_llvm_prepare; vs->base.run = vs_llvm_run; vs->base.delete = vs_llvm_delete; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 5ee2adb344..a4503c143e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -43,6 +43,7 @@ #include "rtasm/rtasm_x86sse.h" #include "tgsi/exec/tgsi_sse2.h" +#include "tgsi/util/tgsi_parse.h" typedef void (XSTDCALL *codegen_function) ( @@ -204,6 +205,7 @@ vs_sse_delete( struct draw_vertex_shader *base ) x86_release_func( &shader->sse2_program ); + FREE( (void*) shader->base.state.tokens ); FREE( shader ); } @@ -213,6 +215,7 @@ draw_create_vs_sse(struct draw_context *draw, const struct pipe_shader_state *templ) { struct draw_sse_vertex_shader *vs; + uint nt = tgsi_num_tokens(templ->tokens); if (!draw->use_sse) return NULL; @@ -221,7 +224,8 @@ draw_create_vs_sse(struct draw_context *draw, if (vs == NULL) return NULL; - vs->base.state = *templ; + /* we make a private copy of the tokens */ + vs->base.state.tokens = mem_dup(templ->tokens, nt * sizeof(templ->tokens[0])); vs->base.prepare = vs_sse_prepare; vs->base.run = vs_sse_run; vs->base.delete = vs_sse_delete; -- cgit v1.2.3 From 762e96c80454c94981cd0ccb4b342f192ba8f77e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 24 Mar 2008 12:59:54 -0600 Subject: gallium: check SamplersUsed to determine if texture needed, as in st_atom_sampler.c Otherwise, we were sometimes setting texture state but not the corresponding sampler state. --- src/mesa/state_tracker/st_atom_texture.c | 31 +++++++++++++++---------------- 1 file changed, 15 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 9aef30f456..e1528c4d7b 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -56,28 +56,27 @@ update_textures(struct st_context *st) for (unit = 0; unit < st->ctx->Const.MaxTextureCoordUnits; unit++) { const GLuint su = fprog->Base.SamplerUnits[unit]; - struct gl_texture_object *texObj = st->ctx->Texture.Unit[su]._Current; - struct st_texture_object *stObj = st_texture_object(texObj); - struct pipe_texture *pt; + struct pipe_texture *pt = NULL; - if (texObj) { - GLboolean flush, retval; + if (fprog->Base.SamplersUsed & (1 << su)) { + struct gl_texture_object *texObj = st->ctx->Texture.Unit[su]._Current; + struct st_texture_object *stObj = st_texture_object(texObj); - retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); - if (!retval) { - /* out of mem */ - continue; + if (texObj) { + GLboolean flush, retval; + + retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); + if (!retval) { + /* out of mem */ + continue; + } + + st->state.num_textures = unit + 1; } - st->state.num_textures = unit + 1; + pt = st_get_stobj_texture(stObj); } - /* XXX: need to ensure that textures are unbound/removed from - * this table before being deleted, otherwise the pointer - * comparison below could fail. - */ - - pt = st_get_stobj_texture(stObj); pipe_texture_reference(&st->state.sampler_texture[unit], pt); } -- cgit v1.2.3 From 45cc35e77600af8628393475405160e26d56d421 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 24 Mar 2008 17:20:22 -0600 Subject: gallium: fix a few texture border/mipmap glitches found w/ conform --- src/mesa/state_tracker/st_cb_texture.c | 27 +++++++++++++-------------- 1 file changed, 13 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index a6c1a35355..d0bc234f36 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -265,19 +265,17 @@ guess_and_alloc_texture(struct st_context *st, { GLuint firstLevel; GLuint lastLevel; - GLuint width = stImage->base.Width; - GLuint height = stImage->base.Height; - GLuint depth = stImage->base.Depth; + GLuint width = stImage->base.Width2; /* size w/out border */ + GLuint height = stImage->base.Height2; + GLuint depth = stImage->base.Depth2; GLuint i, comp_byte = 0; DBG("%s\n", __FUNCTION__); assert(!stObj->pt); - if (stImage->base.Border) - return; - - if (stImage->level > stObj->base.BaseLevel && + if (stObj->pt && + stImage->level > stObj->base.BaseLevel && (stImage->base.Width == 1 || (stObj->base.Target != GL_TEXTURE_1D && stImage->base.Height == 1) || @@ -297,7 +295,8 @@ guess_and_alloc_texture(struct st_context *st, /* Figure out image dimensions at start level. */ for (i = stImage->level; i > firstLevel; i--) { - width <<= 1; + if (width != 1) + width <<= 1; if (height != 1) height <<= 1; if (depth != 1) @@ -1448,9 +1447,9 @@ st_finalize_texture(GLcontext *ctx, stObj->pt->format != st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat) || stObj->pt->last_level != stObj->lastLevel || - stObj->pt->width[0] != firstImage->base.Width || - stObj->pt->height[0] != firstImage->base.Height || - stObj->pt->depth[0] != firstImage->base.Depth || + stObj->pt->width[0] != firstImage->base.Width2 || + stObj->pt->height[0] != firstImage->base.Height2 || + stObj->pt->depth[0] != firstImage->base.Depth2 || stObj->pt->cpp != cpp || stObj->pt->compressed != firstImage->base.IsCompressed)) { pipe_texture_release(&stObj->pt); @@ -1464,9 +1463,9 @@ st_finalize_texture(GLcontext *ctx, gl_target_to_pipe(stObj->base.Target), st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat), stObj->lastLevel, - firstImage->base.Width, - firstImage->base.Height, - firstImage->base.Depth, + firstImage->base.Width2, + firstImage->base.Height2, + firstImage->base.Depth2, comp_byte); if (!stObj->pt) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); -- cgit v1.2.3 From e27127a61c80c8da0ffc0273688ff79b5166e163 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 18:31:01 -0600 Subject: gallium: save/restore samplers in draw_textured_quad() --- src/mesa/state_tracker/st_cb_drawpixels.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 2ebaf8a2c3..0a2d37c117 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -523,6 +523,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_save_rasterizer(cso); cso_save_viewport(cso); + cso_save_samplers(cso); /* rasterizer state: just scissor */ { @@ -596,13 +597,11 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* restore state */ cso_restore_rasterizer(cso); cso_restore_viewport(cso); + cso_restore_samplers(cso); /* shaders don't go through cso yet */ pipe->bind_fs_state(pipe, st->fp->driver_shader); pipe->bind_vs_state(pipe, st->vp->driver_shader); - cso_set_rasterizer(cso, &st->state.rasterizer); - cso_set_samplers(cso, PIPE_MAX_SAMPLERS, - (const struct pipe_sampler_state **) st->state.sampler_list); pipe->set_sampler_textures(pipe, ctx->st->state.num_textures, ctx->st->state.sampler_texture); } -- cgit v1.2.3 From d6af8fc51d0288f34b92b7249c565c382e83765a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 18:32:48 -0600 Subject: gallium: move sampler_unit field to pstip_fragment_shader since it's per-shader Also, fix another texture refcounting bug. --- src/gallium/auxiliary/draw/draw_pstipple.c | 21 ++++++++++++++++----- 1 file changed, 16 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index f00b23959b..9cec986640 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -56,6 +56,7 @@ struct pstip_fragment_shader struct pipe_shader_state state; void *driver_fs; void *pstip_fs; + uint sampler_unit; }; @@ -67,7 +68,6 @@ struct pstip_stage struct draw_stage stage; void *sampler_cso; - uint sampler_unit; struct pipe_texture *texture; uint num_samplers; uint num_textures; @@ -329,7 +329,7 @@ generate_pstip_fs(struct pstip_stage *pstip) tgsi_dump(pstip_fs.tokens, 0); #endif - pstip->sampler_unit = transform.maxSampler + 1; + pstip->fs->sampler_unit = transform.maxSampler + 1; pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs); } @@ -489,13 +489,15 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) /* how many samplers? */ /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ num_samplers = MAX2(pstip->num_textures, pstip->num_samplers); - num_samplers = MAX2(num_samplers, pstip->sampler_unit + 1); + num_samplers = MAX2(num_samplers, pstip->fs->sampler_unit + 1); /* plug in our sampler, texture */ - pstip->state.samplers[pstip->sampler_unit] = pstip->sampler_cso; - pipe_texture_reference(&pstip->state.textures[pstip->sampler_unit], + pstip->state.samplers[pstip->fs->sampler_unit] = pstip->sampler_cso; + pipe_texture_reference(&pstip->state.textures[pstip->fs->sampler_unit], pstip->texture); + assert(num_samplers <= PIPE_MAX_SAMPLERS); + pstip->driver_bind_sampler_states(pipe, num_samplers, pstip->state.samplers); pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures); @@ -626,9 +628,14 @@ pstip_bind_sampler_states(struct pipe_context *pipe, unsigned num, void **sampler) { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + uint i; /* save current */ memcpy(pstip->state.samplers, sampler, num * sizeof(void *)); + for (i = num; i < PIPE_MAX_SAMPLERS; i++) { + pstip->state.samplers[i] = NULL; + } + pstip->num_samplers = num; /* pass-through */ pstip->driver_bind_sampler_states(pstip->pipe, num, sampler); @@ -646,6 +653,10 @@ pstip_set_sampler_textures(struct pipe_context *pipe, for (i = 0; i < num; i++) { pipe_texture_reference(&pstip->state.textures[i], texture[i]); } + for (; i < PIPE_MAX_SAMPLERS; i++) { + pipe_texture_reference(&pstip->state.textures[i], NULL); + } + pstip->num_textures = num; /* pass-through */ -- cgit v1.2.3 From 4654803e2595ea041ea83baf5e13e6c68890e9a7 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 18:49:56 -0600 Subject: gallium: fix a few bugs, warnings in the p_debug code added missing _ to a _debug_printf() call. --- src/gallium/auxiliary/util/p_debug.c | 4 ++-- src/gallium/include/pipe/p_debug.h | 9 ++++++--- 2 files changed, 8 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 351de95c95..cc036dabf8 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -260,7 +260,7 @@ enum { }; /* Check for aborts enabled. */ -static unsigned abort_en() +static unsigned abort_en(void) { if (!mapped_config_file) { @@ -283,7 +283,7 @@ static unsigned abort_en() return ((((char *)mapped_config_file)[0]) - 0x30) & eAssertAbortEn; } #else /* WIN32 */ -static unsigned abort_en() +static unsigned abort_en(void) { return !GETENV("GALLIUM_ABORT_ON_ASSERT"); } diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 95bdf2d3b9..577cdaebcd 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -63,7 +63,7 @@ extern "C" { void _debug_vprintf(const char *format, va_list ap); -static void INLINE +static INLINE void _debug_printf(const char *format, ...) { va_list ap; @@ -82,7 +82,7 @@ _debug_printf(const char *format, ...) * - avoid outputing large strings (512 bytes is the current maximum length * that is guaranteed to be printed in all platforms) */ -static void INLINE +static INLINE void debug_printf(const char *format, ...) { #ifdef DEBUG @@ -133,6 +133,9 @@ void _debug_break(void); #endif /* !DEBUG */ +long +debug_get_num_option(const char *name, long dfault); + void _debug_assert_fail(const char *expr, const char *file, unsigned line, @@ -176,7 +179,7 @@ void _debug_assert_fail(const char *expr, */ #ifdef DEBUG #define debug_checkpoint_full() \ - debug_printf("%s:%u:%s", __FILE__, __LINE__, __FUNCTION__) + _debug_printf("%s:%u:%s", __FILE__, __LINE__, __FUNCTION__) #else #define debug_checkpoint_full() \ ((void)0) -- cgit v1.2.3 From 6579440ea98e61871fe781c1c9c681645ddcc075 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 24 Mar 2008 19:36:44 -0600 Subject: gallium: be smarter about picking the sampler unit for pstipple, aaalines Also, if the app really uses all available sampler/texture units, don't just die. Just use the last sampler for the pstipple or aaline texture. --- src/gallium/auxiliary/draw/draw_aaline.c | 54 ++++++++++++++++++++++++------ src/gallium/auxiliary/draw/draw_pstipple.c | 41 +++++++++++++++++------ 2 files changed, 74 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index cc1873abad..a999ef1227 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -61,6 +61,7 @@ struct aaline_fragment_shader void *aaline_fs; void *aapoint_fs; /* not yet */ void *sprite_fs; /* not yet */ + uint sampler_unit; }; @@ -117,7 +118,8 @@ struct aa_transform_context { struct tgsi_transform_context base; uint tempsUsed; /**< bitmask */ int colorOutput; /**< which output is the primary color */ - int maxSampler; /**< max sampler index found */ + uint samplersUsed; /**< bitfield of samplers used */ + int freeSampler; /** an available sampler for the pstipple */ int maxInput, maxGeneric; /**< max input index found */ int colorTemp, texTemp; /**< temp registers */ boolean firstInstruction; @@ -140,8 +142,11 @@ aa_transform_decl(struct tgsi_transform_context *ctx, aactx->colorOutput = decl->u.DeclarationRange.First; } else if (decl->Declaration.File == TGSI_FILE_SAMPLER) { - if ((int) decl->u.DeclarationRange.Last > aactx->maxSampler) - aactx->maxSampler = decl->u.DeclarationRange.Last + 1; + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + aactx->samplersUsed |= 1 << i; + } } else if (decl->Declaration.File == TGSI_FILE_INPUT) { if ((int) decl->u.DeclarationRange.Last > aactx->maxInput) @@ -163,6 +168,21 @@ aa_transform_decl(struct tgsi_transform_context *ctx, } +/** + * Find the lowest zero bit in the given word, or -1 if bitfield is all ones. + */ +static int +free_bit(uint bitfield) +{ + int i; + for (i = 0; i < 32; i++) { + if ((bitfield & (1 << i)) == 0) + return i; + } + return -1; +} + + /** * TGSI instruction transform callback. * Replace writes to result.color w/ a temp reg. @@ -180,6 +200,11 @@ aa_transform_inst(struct tgsi_transform_context *ctx, struct tgsi_full_declaration decl; uint i; + /* find free sampler */ + aactx->freeSampler = free_bit(aactx->samplersUsed); + if (aactx->freeSampler >= PIPE_MAX_SAMPLERS) + aactx->freeSampler = PIPE_MAX_SAMPLERS - 1; + /* find two free temp regs */ for (i = 0; i < 32; i++) { if ((aactx->tempsUsed & (1 << i)) == 0) { @@ -212,7 +237,7 @@ aa_transform_inst(struct tgsi_transform_context *ctx, decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_SAMPLER; decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->maxSampler + 1; + decl.u.DeclarationRange.Last = aactx->freeSampler; ctx->emit_declaration(ctx, &decl); /* declare new temp regs */ @@ -246,7 +271,7 @@ aa_transform_inst(struct tgsi_transform_context *ctx, newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->maxInput + 1; newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->maxSampler + 1; + newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->freeSampler; ctx->emit_instruction(ctx, &newInst); @@ -322,7 +347,6 @@ generate_aaline_fs(struct aaline_stage *aaline) memset(&transform, 0, sizeof(transform)); transform.colorOutput = -1; - transform.maxSampler = -1; transform.maxInput = -1; transform.maxGeneric = -1; transform.colorTemp = -1; @@ -340,6 +364,8 @@ generate_aaline_fs(struct aaline_stage *aaline) tgsi_dump(aaline_fs.tokens, 0); #endif + aaline->fs->sampler_unit = transform.freeSampler; + aaline->fs->aaline_fs = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); @@ -602,7 +628,7 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) auto struct aaline_stage *aaline = aaline_stage(stage); struct draw_context *draw = stage->draw; struct pipe_context *pipe = aaline->pipe; - uint num = MAX2(aaline->num_textures, aaline->num_samplers); + uint num_samplers; assert(draw->rasterizer->line_smooth); @@ -620,11 +646,17 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) */ bind_aaline_fragment_shader(aaline); - aaline->state.sampler[num] = aaline->sampler_cso; - pipe_texture_reference(&aaline->state.texture[num], aaline->texture); + /* how many samplers? */ + /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ + num_samplers = MAX2(aaline->num_textures, aaline->num_samplers); + num_samplers = MAX2(num_samplers, aaline->fs->sampler_unit + 1); + + aaline->state.sampler[aaline->fs->sampler_unit] = aaline->sampler_cso; + pipe_texture_reference(&aaline->state.texture[aaline->fs->sampler_unit], + aaline->texture); - aaline->driver_bind_sampler_states(pipe, num + 1, aaline->state.sampler); - aaline->driver_set_sampler_textures(pipe, num + 1, aaline->state.texture); + aaline->driver_bind_sampler_states(pipe, num_samplers, aaline->state.sampler); + aaline->driver_set_sampler_textures(pipe, num_samplers, aaline->state.texture); /* now really draw first line */ stage->line = aaline_line; diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c index 9cec986640..4dddb72906 100644 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pstipple.c @@ -112,7 +112,8 @@ struct pstip_transform_context { uint tempsUsed; /**< bitmask */ int wincoordInput; int maxInput; - int maxSampler; /**< max sampler index found */ + uint samplersUsed; /**< bitfield of samplers used */ + int freeSampler; /** an available sampler for the pstipple */ int texTemp; /**< temp registers */ int numImmed; boolean firstInstruction; @@ -130,8 +131,11 @@ pstip_transform_decl(struct tgsi_transform_context *ctx, struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; if (decl->Declaration.File == TGSI_FILE_SAMPLER) { - if ((int) decl->u.DeclarationRange.Last > pctx->maxSampler) - pctx->maxSampler = (int) decl->u.DeclarationRange.Last; + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + pctx->samplersUsed |= 1 << i; + } } else if (decl->Declaration.File == TGSI_FILE_INPUT) { pctx->maxInput = MAX2(pctx->maxInput, (int) decl->u.DeclarationRange.Last); @@ -159,6 +163,21 @@ pstip_transform_immed(struct tgsi_transform_context *ctx, } +/** + * Find the lowest zero bit in the given word, or -1 if bitfield is all ones. + */ +static int +free_bit(uint bitfield) +{ + int i; + for (i = 0; i < 32; i++) { + if ((bitfield & (1 << i)) == 0) + return i; + } + return -1; +} + + /** * TGSI instruction transform callback. * Replace writes to result.color w/ a temp reg. @@ -177,7 +196,11 @@ pstip_transform_inst(struct tgsi_transform_context *ctx, struct tgsi_full_instruction newInst; uint i; int wincoordInput; - const int sampler = pctx->maxSampler + 1; + + /* find free sampler */ + pctx->freeSampler = free_bit(pctx->samplersUsed); + if (pctx->freeSampler >= PIPE_MAX_SAMPLERS) + pctx->freeSampler = PIPE_MAX_SAMPLERS - 1; if (pctx->wincoordInput < 0) wincoordInput = pctx->maxInput + 1; @@ -214,7 +237,7 @@ pstip_transform_inst(struct tgsi_transform_context *ctx, decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_SAMPLER; decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = sampler; + decl.u.DeclarationRange.Last = pctx->freeSampler; ctx->emit_declaration(ctx, &decl); /* declare new temp regs */ @@ -274,7 +297,7 @@ pstip_transform_inst(struct tgsi_transform_context *ctx, newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp; newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - newInst.FullSrcRegisters[1].SrcRegister.Index = sampler; + newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->freeSampler; ctx->emit_instruction(ctx, &newInst); /* KILP texTemp; # if texTemp < 0, KILL fragment */ @@ -313,7 +336,6 @@ generate_pstip_fs(struct pstip_stage *pstip) memset(&transform, 0, sizeof(transform)); transform.wincoordInput = -1; transform.maxInput = -1; - transform.maxSampler = -1; transform.texTemp = -1; transform.firstInstruction = TRUE; transform.base.transform_instruction = pstip_transform_inst; @@ -329,7 +351,8 @@ generate_pstip_fs(struct pstip_stage *pstip) tgsi_dump(pstip_fs.tokens, 0); #endif - pstip->fs->sampler_unit = transform.maxSampler + 1; + pstip->fs->sampler_unit = transform.freeSampler; + assert(pstip->fs->sampler_unit < PIPE_MAX_SAMPLERS); pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs); } @@ -399,8 +422,6 @@ pstip_create_texture(struct pstip_stage *pstip) pstip->texture = screen->texture_create(screen, &texTemp); assert(pstip->texture->refcount == 1); - - //pstip_update_texture(pstip); } -- cgit v1.2.3 From d453042bc67cc94bb215d229b60e8cd70a80ff8d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 24 Mar 2008 19:59:59 -0600 Subject: gallium: don't enable stencil test if no stencil buffer. Also, if not doing two-sided stencil, set back-face state = front-face state. --- src/mesa/state_tracker/st_atom_depth.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index 827ad3b548..03057e37fa 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -105,7 +105,7 @@ update_depth_stencil_alpha(struct st_context *st) st->ctx->Query.CurrentOcclusionObject->Active) dsa->depth.occlusion_count = 1; - if (st->ctx->Stencil.Enabled) { + if (st->ctx->Stencil.Enabled && st->ctx->Visual.stencilBits > 0) { dsa->stencil[0].enabled = 1; dsa->stencil[0].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[0]); dsa->stencil[0].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[0]); @@ -125,6 +125,9 @@ update_depth_stencil_alpha(struct st_context *st) dsa->stencil[1].value_mask = st->ctx->Stencil.ValueMask[1] & 0xff; dsa->stencil[1].write_mask = st->ctx->Stencil.WriteMask[1] & 0xff; } + else { + dsa->stencil[1] = dsa->stencil[0]; + } } if (st->ctx->Color.AlphaEnabled) { -- cgit v1.2.3 From 6fa0bd067176d9159a741014301dfb5fbbd9c4b5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 25 Mar 2008 11:37:24 +0000 Subject: gallium: Fix default option on Windows. --- src/gallium/auxiliary/util/p_debug.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index cc036dabf8..5447bbb6f5 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -155,10 +155,12 @@ debug_get_option(const char *name, const char *dfault) const char *sol, *eol, *sep; static char output[1024]; + result = dfault; + /* XXX: this creates the file if it does not exists, so it must either be + * disabled on release versions, or put in a less conspicuous place. + */ pMap = EngMapFile(L"\\??\\c:\\gallium.cfg", 0, &iFile); - if(!pMap) - result = dfault; - else { + if(pMap) { sol = (const char *)pMap; while(1) { /* TODO: handle LF line endings */ @@ -184,10 +186,7 @@ debug_get_option(const char *name, const char *dfault) result = dfault; #endif - if(result) - debug_printf("%s: %s = %s\n", __FUNCTION__, name, result); - else - debug_printf("%s: %s = (null)\n", __FUNCTION__, name); + debug_printf("%s: %s = %s\n", __FUNCTION__, name, result ? result : "(null)"); return result; } -- cgit v1.2.3 From aacfc326cc94be180864201cd9377db08985698e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 25 Mar 2008 11:37:54 +0000 Subject: gallium: Use debug_get_option for GETENV --- src/gallium/include/pipe/p_util.h | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 505626c60b..1e7b8181f9 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -51,7 +51,6 @@ extern "C" { debug_free( __FILE__, __LINE__, __FUNCTION__, _ptr ) #define REALLOC( _ptr, _old_size, _size ) \ debug_realloc( __FILE__, __LINE__, __FUNCTION__, _ptr, _old_size, _size ) -#define GETENV( X ) NULL #else @@ -118,8 +117,6 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) return new_ptr; } -#define GETENV( X ) NULL - #else /* !WIN32 */ #define MALLOC( SIZE ) malloc( SIZE ) @@ -130,8 +127,6 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) -#define GETENV( X ) getenv( X ) - #endif /* !WIN32 */ #endif /* !DEBUG */ @@ -139,6 +134,8 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) +#define GETENV( X ) debug_get_option( X, NULL ) + /** * Return memory on given byte alignment -- cgit v1.2.3 From 4505acf3b28f0b88bf97838ed7898f10e9200b93 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 25 Mar 2008 15:17:53 +0000 Subject: draw: take primitive into account when deciding if the pipeline is active --- src/gallium/auxiliary/draw/draw_passthrough.c | 2 +- src/gallium/auxiliary/draw/draw_private.h | 3 +- src/gallium/auxiliary/draw/draw_pt.c | 2 +- src/gallium/auxiliary/draw/draw_validate.c | 99 ++++++++++++++++++--------- 4 files changed, 69 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_passthrough.c b/src/gallium/auxiliary/draw/draw_passthrough.c index dd00894c5b..2198079a88 100644 --- a/src/gallium/auxiliary/draw/draw_passthrough.c +++ b/src/gallium/auxiliary/draw/draw_passthrough.c @@ -424,7 +424,7 @@ draw_passthrough_arrays(struct draw_context *draw, debug_printf("%s %d %d %d\n", __FUNCTION__, prim, start, count); - if (draw_need_pipeline(draw)) + if (draw_need_pipeline(draw, prim)) return FALSE; debug_printf("%s AAA\n", __FUNCTION__); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 0c5afcacfa..8a879809c3 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -364,7 +364,8 @@ extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); extern void draw_update_vertex_fetch( struct draw_context *draw ); -extern boolean draw_need_pipeline(const struct draw_context *draw); +extern boolean draw_need_pipeline(const struct draw_context *draw, + unsigned prim ); /* Passthrough mode (second attempt): diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index d7169f78f4..3ec31ec25f 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -55,7 +55,7 @@ draw_pt_arrays(struct draw_context *draw, unsigned start, unsigned count) { - const boolean pipeline = draw_need_pipeline(draw); + const boolean pipeline = draw_need_pipeline(draw, prim); const boolean cliptest = !draw->rasterizer->bypass_clipping; const boolean shading = !draw->rasterizer->bypass_vs; struct draw_pt_front_end *frontend = NULL; diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index 70b477ba5c..e163e078f0 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -32,6 +32,22 @@ #include "pipe/p_defines.h" #include "draw_private.h" +static boolean points( unsigned prim ) +{ + return (prim == PIPE_PRIM_POINTS); +} + +static boolean lines( unsigned prim ) +{ + return (prim == PIPE_PRIM_LINES || + prim == PIPE_PRIM_LINE_STRIP || + prim == PIPE_PRIM_LINE_LOOP); +} + +static boolean triangles( unsigned prim ) +{ + return prim >= PIPE_PRIM_TRIANGLES; +} /** * Check if we need any special pipeline stages, or whether @@ -40,48 +56,63 @@ * pipeline stages. */ boolean -draw_need_pipeline(const struct draw_context *draw) +draw_need_pipeline(const struct draw_context *draw, + unsigned int prim ) { - /* line stipple */ - if (draw->rasterizer->line_stipple_enable && draw->line_stipple) - return TRUE; - - /* wide lines */ - if (draw->rasterizer->line_width > draw->wide_line_threshold) - return TRUE; + /* Don't have to worry about triangles turning into lines/points + * and triggering the pipeline, because we have to trigger the + * pipeline *anyway* if unfilled mode is active. + */ + if (lines(prim)) + { + /* line stipple */ + if (draw->rasterizer->line_stipple_enable && draw->line_stipple) + return TRUE; - /* large points */ - if (draw->rasterizer->point_size > draw->wide_point_threshold) - return TRUE; + /* wide lines */ + if (draw->rasterizer->line_width > draw->wide_line_threshold) + return TRUE; - /* AA lines */ - if (draw->rasterizer->line_smooth && draw->pipeline.aaline) - return TRUE; - - /* AA points */ - if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) - return TRUE; + /* AA lines */ + if (draw->rasterizer->line_smooth && draw->pipeline.aaline) + return TRUE; + } - /* polygon stipple */ - if (draw->rasterizer->poly_stipple_enable && draw->pipeline.pstipple) - return TRUE; + if (points(prim)) + { + /* large points */ + if (draw->rasterizer->point_size > draw->wide_point_threshold) + return TRUE; - /* unfilled polygons */ - if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) - return TRUE; + /* AA points */ + if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) + return TRUE; - /* polygon offset */ - if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) - return TRUE; + /* point sprites */ + if (draw->rasterizer->point_sprite && draw->point_sprite) + return TRUE; + } - /* point sprites */ - if (draw->rasterizer->point_sprite && draw->point_sprite) - return TRUE; - /* two-side lighting */ - if (draw->rasterizer->light_twoside) - return TRUE; + if (triangles(prim)) + { + /* polygon stipple */ + if (draw->rasterizer->poly_stipple_enable && draw->pipeline.pstipple) + return TRUE; + + /* unfilled polygons */ + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) + return TRUE; + + /* polygon offset */ + if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) + return TRUE; + + /* two-side lighting */ + if (draw->rasterizer->light_twoside) + return TRUE; + } /* polygon cull - this is difficult - hardware can cull just fine * most of the time (though sometimes CULL_NEITHER is unsupported. -- cgit v1.2.3 From 331a56136e96717704788b633c1b2e474b88d8ba Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 25 Mar 2008 17:47:39 +0000 Subject: Fix typo --- src/gallium/include/pipe/p_debug.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 577cdaebcd..097fc7c0b1 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -191,7 +191,7 @@ void _debug_assert_fail(const char *expr, */ #ifdef DEBUG #define debug_warning(__msg) \ - _debug_printf("%s:%u:%s: warning: %s\n", __FILE__, __LINE__, __FUNCTION__, (__msg)) + _debug_printf("%s:%u:%s: warning: %s\n", __FILE__, __LINE__, __FUNCTION__, __msg) #else #define debug_warning(__msg) \ ((void)0) @@ -203,10 +203,10 @@ void _debug_assert_fail(const char *expr, */ #ifdef DEBUG #define debug_error(__msg) \ - _debug_printf("%s:%u:%s: error: %s\n", __FILE__, __LINE__, __FUNCTION__, (__msg)) + _debug_printf("%s:%u:%s: error: %s\n", __FILE__, __LINE__, __FUNCTION__, __msg) #else #define debug_error(__msg) \ - _debug_printf("error: %s\n", __msg)) + _debug_printf("error: %s\n", __msg) #endif -- cgit v1.2.3 From bedb93e989340a2d65ed2fb28af7410983cb484d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 25 Mar 2008 11:22:57 -0600 Subject: mesa: when negating scalar src args, use NEGATE_XYZW, not NEGATE_X This makes things easier on the back-end when generating GPU code. --- src/mesa/shader/arbprogparse.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 9e5169eff1..6782491cc5 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -2578,7 +2578,7 @@ parse_fp_vector_src_reg(GLcontext * ctx, const GLubyte ** inst, GLboolean isRelOffset; /* Grab the sign */ - negate = (parse_sign (inst) == -1) ? 0xf : 0x0; + negate = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; /* And the src reg */ if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset)) @@ -2635,7 +2635,7 @@ parse_fp_scalar_src_reg (GLcontext * ctx, const GLubyte ** inst, GLboolean IsRelOffset; /* Grab the sign */ - Negate = (parse_sign (inst) == -1) ? 0x1 : 0x0; + Negate = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; /* And the src reg */ if (parse_src_reg (ctx, inst, vc_head, Program, &File, &Index, &IsRelOffset)) @@ -3123,7 +3123,7 @@ parse_vp_vector_src_reg(GLcontext * ctx, const GLubyte ** inst, GLboolean isRelOffset; /* Grab the sign */ - negateMask = (parse_sign (inst) == -1) ? 0xf : 0x0; + negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; /* And the src reg */ if (parse_src_reg (ctx, inst, vc_head, program, &file, &index, &isRelOffset)) @@ -3150,12 +3150,12 @@ parse_vp_scalar_src_reg (GLcontext * ctx, const GLubyte ** inst, { enum register_file File; GLint Index; - GLubyte Negate; + GLubyte negateMask; GLubyte Swizzle[4]; GLboolean IsRelOffset; /* Grab the sign */ - Negate = (parse_sign (inst) == -1) ? 0x1 : 0x0; + negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; /* And the src reg */ if (parse_src_reg (ctx, inst, vc_head, Program, &File, &Index, &IsRelOffset)) @@ -3167,7 +3167,7 @@ parse_vp_scalar_src_reg (GLcontext * ctx, const GLubyte ** inst, reg->File = File; reg->Index = Index; reg->Swizzle = (Swizzle[0] << 0); - reg->NegateBase = Negate; + reg->NegateBase = negateMask; reg->RelAddr = IsRelOffset; return 0; } -- cgit v1.2.3 From c0ef16647424452ba60e165c82c59f5ff64f12d4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 25 Mar 2008 11:32:31 -0600 Subject: mesa: append fog instructions after parsing if a fog option is set Drivers don't have to worry about it then. --- src/mesa/shader/arbprogparse.c | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 6782491cc5..6648ed7b71 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -35,6 +35,7 @@ #include "shader/grammar/grammar_mesa.h" #include "arbprogparse.h" #include "program.h" +#include "programopt.h" #include "prog_parameter.h" #include "prog_statevars.h" #include "context.h" @@ -3929,6 +3930,16 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target, _mesa_free_parameter_list(program->Base.Parameters); program->Base.Parameters = ap.Base.Parameters; + /* Append fog instructions now if the program has "OPTION ARB_fog_exp" + * or similar. We used to leave this up to drivers, but it appears + * there's no hardware that wants to do fog in a discrete stage separate + * from the fragment shader. + */ + if (program->FogOption != GL_NONE) { + _mesa_append_fog_code(ctx, program); + program->FogOption = GL_NONE; + } + #if DEBUG_FP _mesa_printf("____________Fragment program %u ________\n", program->Base.ID); _mesa_print_program(&program->Base); -- cgit v1.2.3 From cbec00849186db11d77fd00822145e11e69cb07f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 25 Mar 2008 18:09:25 +0000 Subject: draw: don't use fetch_and_store for bypass_vs mode, it's not quite right --- src/gallium/auxiliary/draw/draw_prim.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index 4fe0ddc02a..75b2f79bab 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -170,10 +170,7 @@ void draw_do_flush( struct draw_context *draw, unsigned flags ) if (flags >= DRAW_FLUSH_SHADER_QUEUE) { if (draw->vs.queue_nr) { - if (draw->rasterizer->bypass_vs) - fetch_and_store(draw); - else - (*draw->shader_queue_flush)(draw); + (*draw->shader_queue_flush)(draw); } if (flags >= DRAW_FLUSH_PRIM_QUEUE) { -- cgit v1.2.3 From 05a4ecdec2b5fc590eb09cc5a6b4208e0f739c5a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 25 Mar 2008 18:15:58 +0000 Subject: draw: vertex fetch can be validated too early leading to an assertion... disable --- src/gallium/auxiliary/draw/draw_vertex_fetch.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vertex_fetch.c b/src/gallium/auxiliary/draw/draw_vertex_fetch.c index b56d85396d..11f99babf6 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_fetch.c +++ b/src/gallium/auxiliary/draw/draw_vertex_fetch.c @@ -314,7 +314,11 @@ static fetch_func get_fetch_func( enum pipe_format format ) return NULL; /* not sure why this is needed */ default: - assert(0); + /* This can get hit because draw-state-validation is too eager, + and can jump in here validating stuff before the state tracker has set + up everything. + */ + /* assert(0); */ return NULL; } } -- cgit v1.2.3 From cf7daba79101a2d2813f133a0f3d394316e66cce Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 25 Mar 2008 12:27:32 -0600 Subject: mesa: fix some issues in _mesa_combine_programs() Use a temporay register to connect outputs of first program to inputs of second program. Also, fix bug in replace_registers(): didn't search/replace DstReg. --- src/mesa/shader/program.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 39c4ca1dd5..09a8494bd3 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -426,6 +426,7 @@ replace_registers(struct prog_instruction *inst, GLuint numInst, { GLuint i, j; for (i = 0; i < numInst; i++) { + /* src regs */ for (j = 0; j < _mesa_num_inst_src_regs(inst->Opcode); j++) { if (inst[i].SrcReg[j].File == oldFile && inst[i].SrcReg[j].Index == oldIndex) { @@ -433,6 +434,11 @@ replace_registers(struct prog_instruction *inst, GLuint numInst, inst[i].SrcReg[j].Index = newIndex; } } + /* dst reg */ + if (inst[i].DstReg.File == oldFile && inst[i].DstReg.Index == oldIndex) { + inst[i].DstReg.File = newFile; + inst[i].DstReg.Index = newIndex; + } } } @@ -504,12 +510,25 @@ _mesa_combine_programs(GLcontext *ctx, newFprog->UsesKill = fprogA->UsesKill || fprogB->UsesKill; - /* connect color outputs/inputs */ + /* Connect color outputs of fprogA to color inputs of fprogB, via a + * new temporary register. + */ if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) && (progB->InputsRead & (1 << FRAG_ATTRIB_COL0))) { + GLint tempReg = _mesa_find_free_register(newProg, PROGRAM_TEMPORARY); + if (!tempReg) { + _mesa_problem(ctx, "No free temp regs found in " + "_mesa_combine_programs(), using 31"); + tempReg = 31; + } + /* replace writes to result.color[0] with tempReg */ + replace_registers(newInst, lenA, + PROGRAM_OUTPUT, FRAG_RESULT_COLR, + PROGRAM_TEMPORARY, tempReg); + /* replace reads from input.color[0] with tempReg */ replace_registers(newInst + lenA, lenB, PROGRAM_INPUT, FRAG_ATTRIB_COL0, - PROGRAM_OUTPUT, FRAG_RESULT_COLR); + PROGRAM_TEMPORARY, tempReg); } inputsB = progB->InputsRead; -- cgit v1.2.3 From 070a7446221e26aee5ab6e6e12988ea9c1688ba6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 25 Mar 2008 12:27:48 -0600 Subject: gallium: added some debug code (disable) --- src/mesa/state_tracker/st_cb_drawpixels.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 0a2d37c117..026f015fd8 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -114,6 +114,12 @@ combined_drawpix_fragment_program(GLcontext *ctx) _mesa_clone_program(ctx, &st->pixel_xfer.program->Base.Base); } else { +#if 0 + printf("Base program:\n"); + _mesa_print_program(&st->fp->Base.Base); + printf("DrawPix program:\n"); + _mesa_print_program(&st->pixel_xfer.program->Base.Base); +#endif stfp = (struct st_fragment_program *) _mesa_combine_programs(ctx, &st->pixel_xfer.program->Base.Base, -- cgit v1.2.3 From 7f56c8ea5fe6909291ba440221b270f5ef37366f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 25 Mar 2008 14:12:55 -0600 Subject: gallium: disable the selection/feedback draw module's options for wide lines, points, etc. Disable paths that would convert points/lines to tris as that upsets selection, feedback, rastpos. --- src/mesa/state_tracker/st_context.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 7c18386575..726e06d7c2 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -95,6 +95,14 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st->draw = draw_create(); /* for selection/feedback */ + /* Disable draw options that might convert points/lines to tris, etc. + * as that would foul-up feedback/selection mode. + */ + draw_wide_line_threshold(st->draw, 1000.0f); + draw_wide_point_threshold(st->draw, 1000.0f); + draw_enable_line_stipple(st->draw, FALSE); + draw_enable_point_sprites(st->draw, FALSE); + st->dirty.mesa = ~0; st->dirty.st = ~0; -- cgit v1.2.3 From dc9757e1a8a0451a198d329880558b805adff42a Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 25 Mar 2008 23:48:30 +0100 Subject: gallium: Introduce flatshade_first field to rasterizer_state. This bit tells us which vertex of the primitive is used to propagate color for the remaining vertices if flatshade mode. --- src/gallium/include/pipe/p_state.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 3e531c4da4..a2bd8c6aaa 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -112,6 +112,7 @@ struct pipe_rasterizer_state unsigned bypass_clipping:1; unsigned bypass_vs:1; /**< vertices are already fully transformed */ unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ + unsigned flatshade_first:1; /**< take color attribute from the first vertex of a primitive */ float line_width; float point_size; /**< used when no per-vertex size */ -- cgit v1.2.3 From e1543fa55c7972e3634f3f7ba297c010337dfb0d Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 25 Mar 2008 23:49:11 +0100 Subject: draw: Take flatshade_first rasterizer bit into account. --- src/gallium/auxiliary/draw/draw_prim.c | 162 ++++++++++++++++++++++++--------- 1 file changed, 117 insertions(+), 45 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index 75b2f79bab..f589d0c017 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -330,6 +330,8 @@ draw_prim( struct draw_context *draw, unsigned i; boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); + boolean flatfirst = + (draw->rasterizer->flatshade & draw->rasterizer->flatshade_first) ? TRUE : FALSE; // debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); @@ -342,11 +344,21 @@ draw_prim( struct draw_context *draw, break; case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) { - do_line( draw, - TRUE, - start + i + 0, - start + i + 1); + if (flatfirst) { + for (i = 0; i+1 < count; i += 2) { + do_line( draw, + TRUE, + start + i + 1, + start + i + 0); + } + } + else { + for (i = 0; i+1 < count; i += 2) { + do_line( draw, + TRUE, + start + i + 0, + start + i + 1); + } } break; @@ -367,60 +379,120 @@ draw_prim( struct draw_context *draw, break; case PIPE_PRIM_LINE_STRIP: - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, - start + i - 1, - start + i ); + if (flatfirst) { + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, + start + i, + start + i - 1 ); + } + } + else { + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, + start + i - 1, + start + i ); + } } break; case PIPE_PRIM_TRIANGLES: - if (unfilled) { - for (i = 0; i+2 < count; i += 3) { - do_ef_triangle( draw, - 1, - ~0, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } + if (flatfirst) { + if (unfilled) { + for (i = 0; i+2 < count; i += 3) { + do_ef_triangle( draw, + 1, + ~0, + start + i + 1, + start + i + 2, + start + i + 0 ); + } + } + else { + for (i = 0; i+2 < count; i += 3) { + do_triangle( draw, + start + i + 1, + start + i + 2, + start + i + 0 ); + } + } + } else { - for (i = 0; i+2 < count; i += 3) { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } + if (unfilled) { + for (i = 0; i+2 < count; i += 3) { + do_ef_triangle( draw, + 1, + ~0, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } + else { + for (i = 0; i+2 < count; i += 3) { + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } } break; case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - do_triangle( draw, - start + i + 1, - start + i + 0, - start + i + 2 ); - } - else { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + if (i & 1) { + do_triangle( draw, + start + i + 2, + start + i + 1, + start + i + 0 ); + } + else { + do_triangle( draw, + start + i + 1, + start + i + 2, + start + i + 0 ); + } + } + } + else { + for (i = 0; i+2 < count; i++) { + if (i & 1) { + do_triangle( draw, + start + i + 1, + start + i + 0, + start + i + 2 ); + } + else { + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); + } + } } break; case PIPE_PRIM_TRIANGLE_FAN: if (count >= 3) { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + 0, - start + i + 1, - start + i + 2 ); - } + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + do_triangle( draw, + start + i + 2, + start + 0, + start + i + 1 ); + } + } + else { + for (i = 0; i+2 < count; i++) { + do_triangle( draw, + start + 0, + start + i + 1, + start + i + 2 ); + } + } } break; -- cgit v1.2.3 From 84d8030735844785c3c97679db2bc1892a9c8c70 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Mon, 24 Mar 2008 12:15:59 -0700 Subject: cell: Float convert-to and convert-from instructions use different shift bias --- src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c | 4 ++-- src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h | 10 +++++----- 2 files changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c index 24be65bff9..7f6bf577b2 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c @@ -267,10 +267,10 @@ void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ emit_RI7(p, _op, rT, rA, imm); \ } -#define EMIT_RI8(_name, _op) \ +#define EMIT_RI8(_name, _op, bias) \ void _name (struct spe_function *p, unsigned rT, unsigned rA, int imm) \ { \ - emit_RI8(p, _op, rT, rA, 155 - imm); \ + emit_RI8(p, _op, rT, rA, bias - imm); \ } #define EMIT_RI10(_name, _op) \ diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h index 5a1eb1ed8d..1cacc717b1 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.h @@ -76,7 +76,7 @@ extern void spe_release_register(struct spe_function *p, int reg); #define EMIT_RI7(_name, _op) \ extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ int imm) -#define EMIT_RI8(_name, _op) \ +#define EMIT_RI8(_name, _op, bias) \ extern void _name (struct spe_function *p, unsigned rT, unsigned rA, \ int imm) #define EMIT_RI10(_name, _op) \ @@ -289,10 +289,10 @@ EMIT_RR (spe_dfnma, 0x35f); EMIT_R (spe_frest, 0x1b8); EMIT_R (spe_frsqest, 0x1b9); EMIT_RR (spe_fi, 0x3d4); -EMIT_RI8 (spe_csflt, 0x1da); -EMIT_RI8 (spe_cflts, 0x1d8); -EMIT_RI8 (spe_cuflt, 0x1db); -EMIT_RI8 (spe_cfltu, 0x1d9); +EMIT_RI8 (spe_csflt, 0x1da, 155); +EMIT_RI8 (spe_cflts, 0x1d8, 173); +EMIT_RI8 (spe_cuflt, 0x1db, 155); +EMIT_RI8 (spe_cfltu, 0x1d9, 173); EMIT_R (spe_frds, 0x3b9); EMIT_R (spe_fesd, 0x3b8); EMIT_RR (spe_dfceq, 0x3c3); -- cgit v1.2.3 From b09de96a17028c3c936f6a196e048c4f224da89f Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Tue, 25 Mar 2008 18:07:56 -0700 Subject: xlib: Add support for MIT-SHM in xlib winsys driver Gives about a 3% performance increase in gears on x86-64 (non-tiled) and about 10% performance increase in gears on Cell (tiled). I actually expected more of a boost. :( --- src/gallium/winsys/xlib/xm_api.c | 16 +- src/gallium/winsys/xlib/xm_winsys.c | 301 ++++++++++++++++++++++++++------ src/gallium/winsys/xlib/xm_winsys_aub.h | 3 +- src/gallium/winsys/xlib/xmesaP.h | 3 + 4 files changed, 267 insertions(+), 56 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c index e5fef1d7a8..6b46327e21 100644 --- a/src/gallium/winsys/xlib/xm_api.c +++ b/src/gallium/winsys/xlib/xm_api.c @@ -104,12 +104,16 @@ static int host_byte_order( void ) * 1 = shared XImage support available * 2 = shared Pixmap support available also */ -static int check_for_xshm( XMesaDisplay *display ) +int xmesa_check_for_xshm( XMesaDisplay *display ) { #if defined(USE_XSHM) && !defined(XFree86Server) int major, minor, ignore; Bool pixmaps; + if (getenv("MESA_NOSHM")) { + return 0; + } + if (XQueryExtension( display, "MIT-SHM", &ignore, &ignore, &ignore )) { if (XShmQueryVersion( display, &major, &minor, &pixmaps )==True) { return (pixmaps==True) ? 2 : 1; @@ -528,7 +532,7 @@ initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, /* Setup for single/double buffering */ if (v->mesa_visual.doubleBufferMode) { /* Double buffered */ - b->shm = check_for_xshm( v->display ); + b->shm = xmesa_check_for_xshm( v->display ); } /* X11 graphics context */ @@ -770,13 +774,16 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) pf = choose_pixel_format(v); assert(pf); + c->xm_visual = v; + c->xm_buffer = NULL; /* set later by XMesaMakeCurrent */ + if (!getenv("XM_AUB")) { xmesa_mode = XMESA_SOFTPIPE; pipe = xmesa_create_pipe_context( c, pf ); } else { xmesa_mode = XMESA_AUB; - pipe = xmesa_create_i965simple( xmesa_get_pipe_winsys_aub() ); + pipe = xmesa_create_i965simple(xmesa_get_pipe_winsys_aub(v)); } c->st = st_create_context(pipe, &v->mesa_visual, @@ -800,9 +807,6 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) #endif /* finish up xmesa context initializations */ - c->xm_visual = v; - c->xm_buffer = NULL; /* set later by XMesaMakeCurrent */ - c->st->haveFramebufferSurfaces = GL_TRUE; return c; diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 018cf01cd6..7baaae295c 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -65,8 +65,19 @@ struct xm_buffer boolean userBuffer; /** Is this a user-space buffer? */ void *data; void *mapped; + + XImage *tempImage; + int shm; +#if defined(USE_XSHM) && !defined(XFree86Server) + XShmSegmentInfo shminfo; +#endif }; +#if defined(USE_XSHM) && !defined(XFree86Server) +# define XSHM_ENABLED(b) ((b)->shm) +#else +# define XSHM_ENABLED(b) 0 +#endif struct xmesa_surface { @@ -88,6 +99,16 @@ struct xmesa_softpipe_winsys }; +struct xmesa_pipe_winsys +{ + struct pipe_winsys base; + struct xmesa_visual *xm_visual; + int shm; +}; + + +static void alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, + unsigned width, unsigned height); /** Cast wrapper */ static INLINE struct xmesa_surface * @@ -136,13 +157,27 @@ xm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) static void xm_buffer_destroy(struct pipe_winsys *pws, - struct pipe_buffer *buf) + struct pipe_buffer *buf) { struct xm_buffer *oldBuf = xm_buffer(buf); if (oldBuf->data) { - if (!oldBuf->userBuffer) - align_free(oldBuf->data); +#if defined(USE_XSHM) && !defined(XFree86Server) + if (oldBuf->shminfo.shmid >= 0) { + shmdt(oldBuf->shminfo.shmaddr); + shmctl(oldBuf->shminfo.shmid, IPC_RMID, 0); + + oldBuf->shminfo.shmid = -1; + oldBuf->shminfo.shmaddr = (char *) -1; + } + else +#endif + { + if (!oldBuf->userBuffer) { + align_free(oldBuf->data); + } + } + oldBuf->data = NULL; } @@ -157,7 +192,7 @@ xm_buffer_destroy(struct pipe_winsys *pws, static void xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) { - XImage *ximage = b->tempImage; + XImage *ximage; struct xm_buffer *xm_buf = xm_buffer(surf->buffer); const uint tilesPerRow = (surf->width + TILE_SIZE - 1) / TILE_SIZE; uint x, y; @@ -166,10 +201,18 @@ xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) assert(ximage->format); assert(ximage->bitmap_unit); - /* update XImage's fields */ - ximage->width = TILE_SIZE; - ximage->height = TILE_SIZE; - ximage->bytes_per_line = TILE_SIZE * 4; + if (XSHM_ENABLED(xm_buf) && (xm_buf->tempImage == NULL)) { + alloc_shm_ximage(xm_buf, b, TILE_SIZE, TILE_SIZE); + } + + ximage = (XSHM_ENABLED(xm_buf)) ? xm_buf->tempImage : b->tempImage; + + if (!XSHM_ENABLED(xm_buf)) { + /* update XImage's fields */ + ximage->width = TILE_SIZE; + ximage->height = TILE_SIZE; + ximage->bytes_per_line = TILE_SIZE * 4; + } for (y = 0; y < surf->height; y += TILE_SIZE) { for (x = 0; x < surf->width; x += TILE_SIZE) { @@ -183,8 +226,15 @@ xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) ximage->data = (char *) xm_buf->data + offset; - XPutImage(b->xm_visual->display, b->drawable, b->gc, - ximage, 0, 0, dx, dy, TILE_SIZE, TILE_SIZE); + if (XSHM_ENABLED(xm_buf)) { +#if defined(USE_XSHM) && !defined(XFree86Server) + XShmPutImage(b->xm_visual->display, b->drawable, b->gc, + ximage, 0, 0, x, y, TILE_SIZE, TILE_SIZE, False); +#endif + } else { + XPutImage(b->xm_visual->display, b->drawable, b->gc, + ximage, 0, 0, dx, dy, TILE_SIZE, TILE_SIZE); + } } } } @@ -197,7 +247,7 @@ xmesa_display_surface_tiled(XMesaBuffer b, const struct pipe_surface *surf) void xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) { - XImage *ximage = b->tempImage; + XImage *ximage; struct xm_buffer *xm_buf = xm_buffer(surf->buffer); const struct xmesa_surface *xm_surf = xmesa_surface((struct pipe_surface *) surf); @@ -207,19 +257,33 @@ xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) return; } - /* check that the XImage has been previously initialized */ - assert(ximage->format); - assert(ximage->bitmap_unit); - /* update XImage's fields */ - ximage->width = surf->width; - ximage->height = surf->height; - ximage->bytes_per_line = surf->pitch * (ximage->bits_per_pixel / 8); + if (XSHM_ENABLED(xm_buf) && (xm_buf->tempImage == NULL)) { + alloc_shm_ximage(xm_buf, b, surf->pitch, surf->height); + } + + ximage = (XSHM_ENABLED(xm_buf)) ? xm_buf->tempImage : b->tempImage; ximage->data = xm_buf->data; /* display image in Window */ - XPutImage(b->xm_visual->display, b->drawable, b->gc, - ximage, 0, 0, 0, 0, surf->width, surf->height); + if (XSHM_ENABLED(xm_buf)) { +#if defined(USE_XSHM) && !defined(XFree86Server) + XShmPutImage(b->xm_visual->display, b->drawable, b->gc, + ximage, 0, 0, 0, 0, surf->width, surf->height, False); +#endif + } else { + /* check that the XImage has been previously initialized */ + assert(ximage->format); + assert(ximage->bitmap_unit); + + /* update XImage's fields */ + ximage->width = surf->width; + ximage->height = surf->height; + ximage->bytes_per_line = surf->pitch * surf->cpp; + + XPutImage(b->xm_visual->display, b->drawable, b->gc, + ximage, 0, 0, 0, 0, surf->width, surf->height); + } } @@ -256,6 +320,119 @@ xm_get_name(struct pipe_winsys *pws) } +#if defined(USE_XSHM) && !defined(XFree86Server) +static volatile int mesaXErrorFlag = 0; + +/** + * Catches potential Xlib errors. + */ +static int +mesaHandleXError(XMesaDisplay *dpy, XErrorEvent *event) +{ + (void) dpy; + (void) event; + mesaXErrorFlag = 1; + return 0; +} + + +static GLboolean alloc_shm(struct xm_buffer *buf, unsigned size) +{ + XShmSegmentInfo *const shminfo = & buf->shminfo; + + shminfo->shmid = shmget(IPC_PRIVATE, size, IPC_CREAT|0777); + if (shminfo->shmid < 0) { + return GL_FALSE; + } + + shminfo->shmaddr = (char *) shmat(shminfo->shmid, 0, 0); + if (shminfo->shmaddr == (char *) -1) { + shmctl(shminfo->shmid, IPC_RMID, 0); + return GL_FALSE; + } + + shminfo->readOnly = False; + return GL_TRUE; +} + + +/** + * Allocate a shared memory XImage back buffer for the given XMesaBuffer. + */ +static void +alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, + unsigned width, unsigned height) +{ + /* + * We have to do a _lot_ of error checking here to be sure we can + * really use the XSHM extension. It seems different servers trigger + * errors at different points if the extension won't work. Therefore + * we have to be very careful... + */ +#if 0 + GC gc; +#endif + int (*old_handler)(XMesaDisplay *, XErrorEvent *); + + b->tempImage = XShmCreateImage(xmb->xm_visual->display, + xmb->xm_visual->visinfo->visual, + xmb->xm_visual->visinfo->depth, + ZPixmap, + NULL, + &b->shminfo, + width, height); + if (b->tempImage == NULL) { + b->shm = 0; + return; + } + + + mesaXErrorFlag = 0; + old_handler = XSetErrorHandler(mesaHandleXError); + /* This may trigger the X protocol error we're ready to catch: */ + XShmAttach(xmb->xm_visual->display, &b->shminfo); + XSync(xmb->xm_visual->display, False); + + if (mesaXErrorFlag) { + /* we are on a remote display, this error is normal, don't print it */ + XFlush(xmb->xm_visual->display); + mesaXErrorFlag = 0; + XDestroyImage(b->tempImage); + b->tempImage = NULL; + b->shm = 0; + (void) XSetErrorHandler(old_handler); + return; + } + + + /* Finally, try an XShmPutImage to be really sure the extension works */ +#if 0 + gc = XCreateGC(xmb->xm_visual->display, xmb->drawable, 0, NULL); + XShmPutImage(xmb->xm_visual->display, xmb->drawable, gc, + b->tempImage, 0, 0, 0, 0, 1, 1 /*one pixel*/, False); + XSync(xmb->xm_visual->display, False); + XFreeGC(xmb->xm_visual->display, gc); + (void) XSetErrorHandler(old_handler); + if (mesaXErrorFlag) { + XFlush(xmb->xm_visual->display); + mesaXErrorFlag = 0; + XDestroyImage(b->tempImage); + b->tempImage = NULL; + b->shm = 0; + return; + } +#endif +} +#else +static void +alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, + unsigned width, unsigned height) +{ + b->shm = 0; +} +#endif + + static struct pipe_buffer * xm_buffer_create(struct pipe_winsys *pws, unsigned alignment, @@ -263,13 +440,35 @@ xm_buffer_create(struct pipe_winsys *pws, unsigned size) { struct xm_buffer *buffer = CALLOC_STRUCT(xm_buffer); +#if defined(USE_XSHM) && !defined(XFree86Server) + struct xmesa_pipe_winsys *xpws = (struct xmesa_pipe_winsys *) pws; +#endif + buffer->base.refcount = 1; buffer->base.alignment = alignment; buffer->base.usage = usage; buffer->base.size = size; - /* align to 16-byte multiple for Cell */ - buffer->data = align_malloc(size, max(alignment, 16)); + +#if defined(USE_XSHM) && !defined(XFree86Server) + buffer->shminfo.shmid = -1; + buffer->shminfo.shmaddr = (char *) -1; + + if (xpws->shm && (usage & PIPE_BUFFER_USAGE_PIXEL) != 0) { + buffer->shm = xpws->shm; + + if (alloc_shm(buffer, size)) { + buffer->data = buffer->shminfo.shmaddr; + } + } +#endif + + if (buffer->data == NULL) { + buffer->shm = 0; + + /* align to 16-byte multiple for Cell */ + buffer->data = align_malloc(size, max(alignment, 16)); + } return &buffer->base; } @@ -286,6 +485,7 @@ xm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) buffer->base.size = bytes; buffer->userBuffer = TRUE; buffer->data = ptr; + buffer->shm = 0; return &buffer->base; } @@ -377,35 +577,38 @@ xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) * Nothing special for the Xlib driver so no subclassing or anything. */ struct pipe_winsys * -xmesa_get_pipe_winsys_aub(void) +xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis) { - static struct pipe_winsys *ws = NULL; + static struct xmesa_pipe_winsys *ws = NULL; if (!ws && getenv("XM_AUB")) { - ws = xmesa_create_pipe_winsys_aub(); + ws = (struct xmesa_pipe_winsys *) xmesa_create_pipe_winsys_aub(); } else if (!ws) { - ws = CALLOC_STRUCT(pipe_winsys); - + ws = CALLOC_STRUCT(xmesa_pipe_winsys); + + ws->xm_visual = xm_vis; + ws->shm = xmesa_check_for_xshm(xm_vis->display); + /* Fill in this struct with callbacks that pipe will need to * communicate with the window system, buffer manager, etc. */ - ws->buffer_create = xm_buffer_create; - ws->user_buffer_create = xm_user_buffer_create; - ws->buffer_map = xm_buffer_map; - ws->buffer_unmap = xm_buffer_unmap; - ws->buffer_destroy = xm_buffer_destroy; - - ws->surface_alloc = xm_surface_alloc; - ws->surface_alloc_storage = xm_surface_alloc_storage; - ws->surface_release = xm_surface_release; - - ws->flush_frontbuffer = xm_flush_frontbuffer; - ws->printf = xm_printf; - ws->get_name = xm_get_name; + ws->base.buffer_create = xm_buffer_create; + ws->base.user_buffer_create = xm_user_buffer_create; + ws->base.buffer_map = xm_buffer_map; + ws->base.buffer_unmap = xm_buffer_unmap; + ws->base.buffer_destroy = xm_buffer_destroy; + + ws->base.surface_alloc = xm_surface_alloc; + ws->base.surface_alloc_storage = xm_surface_alloc_storage; + ws->base.surface_release = xm_surface_release; + + ws->base.flush_frontbuffer = xm_flush_frontbuffer; + ws->base.printf = xm_printf; + ws->base.get_name = xm_get_name; } - return ws; + return &ws->base; } @@ -445,27 +648,27 @@ xmesa_get_softpipe_winsys(uint pixelformat) struct pipe_context * xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) { - struct pipe_winsys *pws = xmesa_get_pipe_winsys_aub(); + struct pipe_winsys *pws = xmesa_get_pipe_winsys_aub(xmesa->xm_visual); struct pipe_context *pipe; #ifdef GALLIUM_CELL if (!getenv("GALLIUM_NOCELL")) { struct cell_winsys *cws = cell_get_winsys(pixelformat); struct pipe_screen *screen = cell_create_screen(pws); + pipe = cell_create_context(screen, cws); - if (pipe) - pipe->priv = xmesa; - return pipe; } else #endif { struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(pixelformat); struct pipe_screen *screen = softpipe_create_screen(pws); - pipe = softpipe_create( screen, pws, spws ); - if (pipe) - pipe->priv = xmesa; - return pipe; + pipe = softpipe_create(screen, pws, spws); } + + if (pipe) + pipe->priv = xmesa; + + return pipe; } diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.h b/src/gallium/winsys/xlib/xm_winsys_aub.h index 7bee199116..cc2a755277 100644 --- a/src/gallium/winsys/xlib/xm_winsys_aub.h +++ b/src/gallium/winsys/xlib/xm_winsys_aub.h @@ -62,6 +62,7 @@ void xmesa_commands_aub(struct pipe_winsys *winsys, void xmesa_display_aub( /* struct pipe_winsys *winsys, */ struct pipe_surface *surface ); -struct pipe_winsys *xmesa_get_pipe_winsys_aub(void); +extern struct pipe_winsys * +xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis); #endif diff --git a/src/gallium/winsys/xlib/xmesaP.h b/src/gallium/winsys/xlib/xmesaP.h index fa8d1f14b9..9b15b2ddf9 100644 --- a/src/gallium/winsys/xlib/xmesaP.h +++ b/src/gallium/winsys/xlib/xmesaP.h @@ -173,4 +173,7 @@ xmesa_buffer_height(XMesaBuffer b) extern void xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf); +extern int +xmesa_check_for_xshm(XMesaDisplay *display); + #endif -- cgit v1.2.3 From 9f7cd571e01569abd0898fdee62d0e0f946bf3a8 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Mar 2008 19:18:56 -0600 Subject: gallium: added fragment emit/write debug counters --- src/gallium/drivers/softpipe/sp_prim_setup.c | 30 ++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 6a81e4d8cc..c7eb12b3bb 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -44,6 +44,7 @@ #include "pipe/p_shader_tokens.h" #define DEBUG_VERTS 0 +#define DEBUG_FRAGS 0 /** * Triangle edge info @@ -92,6 +93,11 @@ struct setup_stage { unsigned y_flags; unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ } span; + +#if DEBUG_FRAGS + uint numFragsEmitted; /**< per primitive */ + uint numFragsWritten; /**< per primitive */ +#endif }; @@ -160,7 +166,20 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) setup->quad.x0 = x; setup->quad.y0 = y; setup->quad.mask = mask; +#if DEBUG_FRAGS + if (mask & 1) setup->numFragsEmitted++; + if (mask & 2) setup->numFragsEmitted++; + if (mask & 4) setup->numFragsEmitted++; + if (mask & 8) setup->numFragsEmitted++; +#endif sp->quad.first->run(sp->quad.first, &setup->quad); +#if DEBUG_FRAGS + mask = setup->quad.mask; + if (mask & 1) setup->numFragsWritten++; + if (mask & 2) setup->numFragsWritten++; + if (mask & 4) setup->numFragsWritten++; + if (mask & 8) setup->numFragsWritten++; +#endif } @@ -674,6 +693,11 @@ static void setup_tri( struct draw_stage *stage, debug_printf("%s\n", __FUNCTION__ ); */ +#if DEBUG_FRAGS + setup->numFragsEmitted = 0; + setup->numFragsWritten = 0; +#endif + setup_sort_vertices( setup, prim ); setup_tri_coefficients( setup ); setup_tri_edges( setup ); @@ -702,6 +726,12 @@ static void setup_tri( struct draw_stage *stage, } flush_spans( setup ); + +#if DEBUG_FRAGS + printf("Tri: %u frags emitted, %u written\n", + setup->numFragsEmitted, + setup->numFragsWritten); +#endif } -- cgit v1.2.3 From e55dccd0bfc41dbcf306f864c01758f8e28fc660 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 25 Mar 2008 19:19:14 -0600 Subject: gallium: the generic attrib we use for computing coverage is per-shader Fixes a very tricky conformance failure. --- src/gallium/auxiliary/draw/draw_aapoint.c | 24 +++++++++++++----------- 1 file changed, 13 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index 67a7a8ebab..fcebe3e7a0 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -68,6 +68,7 @@ struct aapoint_fragment_shader struct pipe_shader_state state; void *driver_fs; /**< the regular shader */ void *aapoint_fs; /**< the aa point-augmented shader */ + int generic_attrib; /**< The generic input attrib/texcoord we'll use */ }; @@ -486,7 +487,6 @@ static void generate_aapoint_fs(struct aapoint_stage *aapoint) { const struct pipe_shader_state *orig_fs = &aapoint->fs->state; - struct draw_context *draw = aapoint->stage.draw; struct pipe_shader_state aapoint_fs; struct aa_transform_context transform; @@ -510,18 +510,16 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) MAX, &transform.base); #if 0 /* DEBUG */ + printf("draw_aapoint, orig shader:\n"); tgsi_dump(orig_fs->tokens, 0); + printf("draw_aapoint, new shader:\n"); tgsi_dump(aapoint_fs.tokens, 0); #endif aapoint->fs->aapoint_fs = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs); - /* advertise the extra post-transform vertex attributes which will have - * the texcoords. - */ - draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; - draw->extra_vp_outputs.semantic_index = transform.maxGeneric + 1; + aapoint->fs->generic_attrib = transform.maxGeneric + 1; } @@ -675,15 +673,19 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header) else aapoint->radius = 0.5f * draw->rasterizer->point_size; - aapoint->tex_slot = draw->num_vs_outputs; - assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */ - draw->extra_vp_outputs.slot = aapoint->tex_slot; - /* - * Bind our fragprog. + * Bind (generate) our fragprog. */ bind_aapoint_fragment_shader(aapoint); + /* update vertex attrib info */ + aapoint->tex_slot = draw->num_vs_outputs; + assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */ + + draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; + draw->extra_vp_outputs.semantic_index = aapoint->fs->generic_attrib; + draw->extra_vp_outputs.slot = aapoint->tex_slot; + /* find psize slot in post-transform vertex */ aapoint->psize_slot = -1; if (draw->rasterizer->point_size_per_vertex) { -- cgit v1.2.3 From 4abe1eb980ed76d2b2d3383eaab520d0aa2ae6f4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 26 Mar 2008 09:36:40 +0000 Subject: gallium: Change pipe->flush() interface to optionally return a fence. The cell driver still uses an internal CELL_FLUSH_WAIT flag, in the long run proper fencing should be implemented for it. --- src/gallium/auxiliary/util/u_gen_mipmap.c | 2 +- src/gallium/drivers/cell/ppu/cell_flush.c | 15 +++++++--- src/gallium/drivers/cell/ppu/cell_flush.h | 5 +++- src/gallium/drivers/cell/ppu/cell_vbuf.c | 2 +- src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 2 +- src/gallium/drivers/failover/fo_context.c | 4 +-- src/gallium/drivers/i915simple/i915_batch.h | 4 +-- src/gallium/drivers/i915simple/i915_blit.c | 4 +-- src/gallium/drivers/i915simple/i915_flush.c | 14 +++------- src/gallium/drivers/i915simple/i915_prim_emit.c | 2 +- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 2 +- src/gallium/drivers/i915simple/i915_state_emit.c | 2 +- src/gallium/drivers/i915simple/i915_winsys.h | 5 ++-- src/gallium/drivers/i965simple/brw_flush.c | 13 ++------- src/gallium/drivers/softpipe/sp_flush.c | 10 +++---- src/gallium/drivers/softpipe/sp_flush.h | 4 ++- src/gallium/include/pipe/p_context.h | 5 ++-- src/gallium/include/pipe/p_defines.h | 3 +- src/gallium/winsys/dri/intel/intel_context.c | 4 +-- src/gallium/winsys/dri/intel/intel_winsys_i915.c | 25 ++++++++++------- src/gallium/winsys/xlib/xm_winsys.c | 31 +++++++++++++++++++++ src/mesa/state_tracker/st_cb_accum.c | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 4 +-- src/mesa/state_tracker/st_cb_fbo.c | 2 +- src/mesa/state_tracker/st_cb_flush.c | 34 +++++++++++++---------- src/mesa/state_tracker/st_cb_readpixels.c | 2 +- src/mesa/state_tracker/st_framebuffer.c | 3 +- src/mesa/state_tracker/st_public.h | 5 +++- 28 files changed, 127 insertions(+), 83 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index e129c062be..ed12768e7f 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -935,7 +935,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, 4, /* verts */ 2); /* attribs/vert */ - pipe->flush(pipe, PIPE_FLUSH_WAIT); + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); /* need to signal that the texture has changed _after_ rendering to it */ pipe->texture_update(pipe, pt, face, (1 << dstLevel)); diff --git a/src/gallium/drivers/cell/ppu/cell_flush.c b/src/gallium/drivers/cell/ppu/cell_flush.c index 66a5627d84..3aaf3de668 100644 --- a/src/gallium/drivers/cell/ppu/cell_flush.c +++ b/src/gallium/drivers/cell/ppu/cell_flush.c @@ -35,12 +35,19 @@ void -cell_flush(struct pipe_context *pipe, unsigned flags) +cell_flush(struct pipe_context *pipe, unsigned flags, + struct pipe_fence_handle **fence) { struct cell_context *cell = cell_context(pipe); + if (fence) { + *fence = NULL; + /* XXX: Implement real fencing */ + flags |= CELL_FLUSH_WAIT; + } + if (flags & PIPE_FLUSH_SWAPBUFFERS) - flags |= PIPE_FLUSH_WAIT; + flags |= CELL_FLUSH_WAIT; draw_flush( cell->draw ); cell_flush_int(pipe, flags); @@ -58,7 +65,7 @@ cell_flush_int(struct pipe_context *pipe, unsigned flags) ASSERT(!flushing); flushing = TRUE; - if (flags & PIPE_FLUSH_WAIT) { + if (flags & CELL_FLUSH_WAIT) { uint64_t *cmd = (uint64_t *) cell_batch_alloc(cell, sizeof(uint64_t)); *cmd = CELL_CMD_FINISH; } @@ -72,7 +79,7 @@ cell_flush_int(struct pipe_context *pipe, unsigned flags) } #endif - if (flags & PIPE_FLUSH_WAIT) { + if (flags & CELL_FLUSH_WAIT) { /* Wait for ack */ for (i = 0; i < cell->num_spus; i++) { uint k = wait_mbox_message(cell_global.spe_contexts[i]); diff --git a/src/gallium/drivers/cell/ppu/cell_flush.h b/src/gallium/drivers/cell/ppu/cell_flush.h index 7f940ae76b..8f0645c429 100644 --- a/src/gallium/drivers/cell/ppu/cell_flush.h +++ b/src/gallium/drivers/cell/ppu/cell_flush.h @@ -29,8 +29,11 @@ #ifndef CELL_FLUSH #define CELL_FLUSH +#define CELL_FLUSH_WAIT 0x80000000 + extern void -cell_flush(struct pipe_context *pipe, unsigned flags); +cell_flush(struct pipe_context *pipe, unsigned flags, + struct pipe_fence_handle **fence); extern void cell_flush_int(struct pipe_context *pipe, unsigned flags); diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.c b/src/gallium/drivers/cell/ppu/cell_vbuf.c index cc727ff4ed..3a181b585c 100644 --- a/src/gallium/drivers/cell/ppu/cell_vbuf.c +++ b/src/gallium/drivers/cell/ppu/cell_vbuf.c @@ -243,7 +243,7 @@ cell_vbuf_draw(struct vbuf_render *vbr, #if 0 /* helpful for debug */ - cell_flush_int(&cell->pipe, PIPE_FLUSH_WAIT); + cell_flush_int(&cell->pipe, CELL_FLUSH_WAIT); #endif } diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index f5c27852c1..b418857ccd 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -133,7 +133,7 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) vs->num_elts = n; send_mbox_message(cell_global.spe_contexts[0], CELL_CMD_VS_EXECUTE); - cell_flush_int(& cell->pipe, PIPE_FLUSH_WAIT); + cell_flush_int(& cell->pipe, CELL_FLUSH_WAIT); } draw->vs.post_nr = draw->vs.queue_nr; diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c index afc0d7eb1e..cb95ba516f 100644 --- a/src/gallium/drivers/failover/fo_context.c +++ b/src/gallium/drivers/failover/fo_context.c @@ -69,7 +69,7 @@ static boolean failover_draw_elements( struct pipe_context *pipe, start, count )) { - failover->hw->flush( failover->hw, ~0 ); + failover->hw->flush( failover->hw, ~0, NULL ); failover->mode = FO_SW; } } @@ -92,7 +92,7 @@ static boolean failover_draw_elements( struct pipe_context *pipe, * intervening flush. Unlikely to be much performance impact to * this: */ - failover->sw->flush( failover->sw, ~0 ); + failover->sw->flush( failover->sw, ~0, NULL ); } return TRUE; diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h index fb88cd6db0..4ea06ce02b 100644 --- a/src/gallium/drivers/i915simple/i915_batch.h +++ b/src/gallium/drivers/i915simple/i915_batch.h @@ -44,9 +44,9 @@ #define ADVANCE_BATCH() -#define FLUSH_BATCH() do { \ +#define FLUSH_BATCH(fence) do { \ if (0) i915_dump_batchbuffer( i915 ); \ - i915->winsys->batch_flush( i915->winsys ); \ + i915->winsys->batch_flush( i915->winsys, fence ); \ i915->batch_start = NULL; \ i915->hardware_dirty = ~0; \ } while (0) diff --git a/src/gallium/drivers/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c index db4671ff55..24449e3fb3 100644 --- a/src/gallium/drivers/i915simple/i915_blit.c +++ b/src/gallium/drivers/i915simple/i915_blit.c @@ -70,7 +70,7 @@ i915_fill_blit(struct i915_context *i915, if (!BEGIN_BATCH(6, 1)) { - FLUSH_BATCH(); + FLUSH_BATCH(NULL); assert(BEGIN_BATCH(6, 1)); } OUT_BATCH(CMD); @@ -145,7 +145,7 @@ i915_copy_blit( struct i915_context *i915, if (!BEGIN_BATCH(8, 2)) { - FLUSH_BATCH(); + FLUSH_BATCH(NULL); assert(BEGIN_BATCH(8, 2)); } OUT_BATCH(CMD); diff --git a/src/gallium/drivers/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c index 96a54281f1..7d23e6b6b9 100644 --- a/src/gallium/drivers/i915simple/i915_flush.c +++ b/src/gallium/drivers/i915simple/i915_flush.c @@ -37,11 +37,9 @@ #include "i915_batch.h" -/** - * In future we may want a fence-like interface instead of finish. - */ static void i915_flush( struct pipe_context *pipe, - unsigned flags ) + unsigned flags, + struct pipe_fence_handle **fence ) { struct i915_context *i915 = i915_context(pipe); @@ -60,7 +58,7 @@ static void i915_flush( struct pipe_context *pipe, flush |= FLUSH_MAP_CACHE; if (!BEGIN_BATCH(1, 0)) { - FLUSH_BATCH(); + FLUSH_BATCH(NULL); assert(BEGIN_BATCH(1, 0)); } OUT_BATCH( flush ); @@ -69,11 +67,7 @@ static void i915_flush( struct pipe_context *pipe, /* If there are no flags, just flush pending commands to hardware: */ - FLUSH_BATCH(); - - if (flags & PIPE_FLUSH_WAIT) { - i915->winsys->batch_finish(i915->winsys); - } + FLUSH_BATCH(fence); } diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c index d8de5178f6..b6fb0a6d88 100644 --- a/src/gallium/drivers/i915simple/i915_prim_emit.c +++ b/src/gallium/drivers/i915simple/i915_prim_emit.c @@ -140,7 +140,7 @@ emit_prim( struct draw_stage *stage, assert(vertex_size >= 12); /* never smaller than 12 bytes */ if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) { - FLUSH_BATCH(); + FLUSH_BATCH(NULL); /* Make sure state is re-emitted after a flush: */ diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index eb64f51943..7fb2adbb53 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -161,7 +161,7 @@ i915_vbuf_render_draw( struct vbuf_render *render, i915_emit_hardware_state( i915 ); if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { - FLUSH_BATCH(); + FLUSH_BATCH(NULL); /* Make sure state is re-emitted after a flush: */ diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index a7498d22b7..6f947d4346 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -115,7 +115,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) #endif if(!BEGIN_BATCH(dwords, relocs)) { - FLUSH_BATCH(); + FLUSH_BATCH(NULL); assert(BEGIN_BATCH(dwords, relocs)); } diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h index aea3003281..5e16543f4e 100644 --- a/src/gallium/drivers/i915simple/i915_winsys.h +++ b/src/gallium/drivers/i915simple/i915_winsys.h @@ -56,6 +56,7 @@ extern "C" { */ struct pipe_buffer; +struct pipe_fence_handle; struct pipe_winsys; struct pipe_screen; @@ -103,8 +104,8 @@ struct i915_winsys { unsigned access_flags, unsigned delta ); - void (*batch_flush)( struct i915_winsys *sws ); - void (*batch_finish)( struct i915_winsys *sws ); + void (*batch_flush)( struct i915_winsys *sws, + struct pipe_fence_handle **fence ); }; #define I915_BUFFER_ACCESS_WRITE 0x1 diff --git a/src/gallium/drivers/i965simple/brw_flush.c b/src/gallium/drivers/i965simple/brw_flush.c index 5216c680cf..e6001c30d9 100644 --- a/src/gallium/drivers/i965simple/brw_flush.c +++ b/src/gallium/drivers/i965simple/brw_flush.c @@ -36,14 +36,11 @@ #include "brw_batch.h" -/** - * In future we may want a fence-like interface instead of finish. - */ static void brw_flush( struct pipe_context *pipe, - unsigned flags ) + unsigned flags, + struct pipe_fence_handle **fence ) { struct brw_context *brw = brw_context(pipe); - struct pipe_fence_handle *fence; /* Do we need to emit an MI_FLUSH command to flush the hardware * caches? @@ -65,11 +62,7 @@ static void brw_flush( struct pipe_context *pipe, /* If there are no flags, just flush pending commands to hardware: */ - FLUSH_BATCH( &fence ); - - if (flags & PIPE_FLUSH_WAIT) { -// brw->winsys->wait_fence(brw->winsys, fence); - } + FLUSH_BATCH( fence ); } diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c index 2cbd0d7cab..0625b69099 100644 --- a/src/gallium/drivers/softpipe/sp_flush.c +++ b/src/gallium/drivers/softpipe/sp_flush.c @@ -40,13 +40,10 @@ #include "sp_winsys.h" -/* There will be actual work to do here. In future we may want a - * fence-like interface instead of finish, and perhaps flush will take - * flags to indicate what type of flush is required. - */ void softpipe_flush( struct pipe_context *pipe, - unsigned flags ) + unsigned flags, + struct pipe_fence_handle **fence ) { struct softpipe_context *softpipe = softpipe_context(pipe); uint i; @@ -72,5 +69,8 @@ softpipe_flush( struct pipe_context *pipe, * to unmap surfaces when flushing. */ softpipe_unmap_surfaces(softpipe); + + if (fence) + *fence = NULL; } diff --git a/src/gallium/drivers/softpipe/sp_flush.h b/src/gallium/drivers/softpipe/sp_flush.h index 34ec617866..68d9b5fa83 100644 --- a/src/gallium/drivers/softpipe/sp_flush.h +++ b/src/gallium/drivers/softpipe/sp_flush.h @@ -29,7 +29,9 @@ #define SP_FLUSH_H struct pipe_context; +struct pipe_fence_handle; -void softpipe_flush(struct pipe_context *pipe, unsigned flags ); +void softpipe_flush(struct pipe_context *pipe, unsigned flags, + struct pipe_fence_handle **fence); #endif diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index a3824601be..b2e49bef4c 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -37,7 +37,7 @@ extern "C" { struct pipe_screen; - +struct pipe_fence_handle; struct pipe_state_cache; /* Opaque driver handles: @@ -202,7 +202,8 @@ struct pipe_context { /* Flush rendering: */ void (*flush)( struct pipe_context *pipe, - unsigned flags ); + unsigned flags, + struct pipe_fence_handle **fence ); }; diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index bc938ba253..586951d956 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -201,8 +201,7 @@ enum pipe_texture_target { */ #define PIPE_FLUSH_RENDER_CACHE 0x1 #define PIPE_FLUSH_TEXTURE_CACHE 0x2 -#define PIPE_FLUSH_WAIT 0x4 -#define PIPE_FLUSH_SWAPBUFFERS 0x8 +#define PIPE_FLUSH_SWAPBUFFERS 0x4 /** diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index 79b320c6bf..8eba33c313 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -228,7 +228,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) assert(intel); /* should never be null */ if (intel) { - st_flush(intel->st, PIPE_FLUSH_WAIT); + st_finish(intel->st); intel_batchbuffer_free(intel->batch); @@ -256,7 +256,7 @@ GLboolean intelUnbindContext(__DRIcontextPrivate * driContextPriv) { struct intel_context *intel = intel_context(driContextPriv); - st_flush(intel->st, 0x0); + st_flush(intel->st, PIPE_FLUSH_RENDER_CACHE, NULL); /* XXX make_current(NULL)? */ return GL_TRUE; } diff --git a/src/gallium/winsys/dri/intel/intel_winsys_i915.c b/src/gallium/winsys/dri/intel/intel_winsys_i915.c index 2def1afc31..4d183db7c3 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_i915.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_i915.c @@ -39,12 +39,14 @@ #include "intel_winsys.h" #include "pipe/p_util.h" +#include "pipe/p_winsys.h" #include "i915simple/i915_winsys.h" #include "i915simple/i915_screen.h" struct intel_i915_winsys { struct i915_winsys winsys; /**< batch buffer funcs */ + struct pipe_winsys *pws; struct intel_context *intel; }; @@ -112,19 +114,22 @@ static void intel_i915_batch_reloc( struct i915_winsys *sws, -static void intel_i915_batch_flush( struct i915_winsys *sws ) +static void intel_i915_batch_flush( struct i915_winsys *sws, + struct pipe_fence_handle **fence ) { - struct intel_context *intel = intel_i915_winsys(sws)->intel; + struct intel_i915_winsys *iws = intel_i915_winsys(sws); + struct intel_context *intel = iws->intel; + union { + struct _DriFenceObject *dri; + struct pipe_fence_handle *pipe; + } fu; - intel_batchbuffer_flush( intel->batch ); -// if (0) intel_i915_batch_wait_idle( sws ); -} + fu.dri = intel_batchbuffer_flush( intel->batch ); + if (fu.dri) + iws->pws->fence_reference(iws->pws, fence, fu.pipe); -static void intel_i915_batch_finish( struct i915_winsys *sws ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - intel_batchbuffer_finish( intel->batch ); +// if (0) intel_i915_batch_wait_idle( sws ); } @@ -145,7 +150,7 @@ intel_create_i915simple( struct intel_context *intel, iws->winsys.batch_dword = intel_i915_batch_dword; iws->winsys.batch_reloc = intel_i915_batch_reloc; iws->winsys.batch_flush = intel_i915_batch_flush; - iws->winsys.batch_finish = intel_i915_batch_finish; + iws->pws = winsys; iws->intel = intel; screen = i915_create_screen(winsys, intel->intelScreen->deviceID); diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 7baaae295c..c930a1d196 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -570,6 +570,33 @@ xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) } +/* + * Fence functions - basically nothing to do, as we don't create any actual + * fence objects. + */ + +static void +xm_fence_reference(struct pipe_winsys *sws, struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) +{ +} + + +static int +xm_fence_signalled(struct pipe_winsys *sws, struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + + +static int +xm_fence_finish(struct pipe_winsys *sws, struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + /** * Return pointer to a pipe_winsys object. @@ -603,6 +630,10 @@ xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis) ws->base.surface_alloc_storage = xm_surface_alloc_storage; ws->base.surface_release = xm_surface_release; + ws->fence_reference = xm_fence_reference; + ws->fence_signalled = xm_fence_signalled; + ws->fence_finish = xm_fence_finish; + ws->base.flush_frontbuffer = xm_flush_frontbuffer; ws->base.printf = xm_printf; ws->base.get_name = xm_get_name; diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index f1fddc4e02..a623d0bcc0 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -214,7 +214,7 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value) const GLint height = ctx->DrawBuffer->_Ymax - ypos; /* make sure color bufs aren't cached */ - pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); switch (op) { case GL_ADD: diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 026f015fd8..43cc21d1fb 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -737,7 +737,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, GLint skipPixels; ubyte *stmap; - pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); /* map the stencil buffer */ stmap = pipe_surface_map(ps); @@ -952,7 +952,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, enum pipe_format srcFormat, texFormat; /* make sure rendering has completed */ - pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); st_validate_state(st); diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 5384252a8e..ec7788923a 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -366,7 +366,7 @@ st_finish_render_texture(GLcontext *ctx, assert(strb); - ctx->st->pipe->flush(ctx->st->pipe, PIPE_FLUSH_RENDER_CACHE); + ctx->st->pipe->flush(ctx->st->pipe, PIPE_FLUSH_RENDER_CACHE, NULL); /* printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface); diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index dbec993f1b..a536a059bd 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -43,19 +43,21 @@ #include "pipe/p_winsys.h" -void st_flush( struct st_context *st, uint pipeFlushFlags ) +void st_flush( struct st_context *st, uint pipeFlushFlags, + struct pipe_fence_handle **fence ) { FLUSH_VERTICES(st->ctx, 0); - st->pipe->flush( st->pipe, pipeFlushFlags ); + st->pipe->flush( st->pipe, pipeFlushFlags, fence ); } -static void st_gl_flush( struct st_context *st, uint pipeFlushFlags ) +static void st_gl_flush( struct st_context *st, uint pipeFlushFlags, + struct pipe_fence_handle **fence ) { GLframebuffer *fb = st->ctx->DrawBuffer; - FLUSH_VERTICES(st->ctx, 0); + st_flush( st, pipeFlushFlags, fence ); if (!fb) return; @@ -80,15 +82,6 @@ static void st_gl_flush( struct st_context *st, uint pipeFlushFlags ) = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); struct pipe_surface *front_surf = strb->surface; - /* If we aren't rendering to the frontbuffer, this is a noop. - * This should be uncontroversial for glFlush, though people may - * feel more strongly about glFinish. - * - * Additionally, need to make sure that the frontbuffer_dirty - * flag really gets set on frontbuffer rendering. - */ - st->pipe->flush( st->pipe, pipeFlushFlags ); - /* Hook for copying "fake" frontbuffer if necessary: */ st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf, @@ -103,7 +96,18 @@ static void st_gl_flush( struct st_context *st, uint pipeFlushFlags ) */ static void st_glFlush(GLcontext *ctx) { - st_gl_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE); + st_gl_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); +} + + +void st_finish( struct st_context *st ) +{ + struct pipe_fence_handle *fence; + + st_gl_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence); + + st->pipe->winsys->fence_finish(st->pipe->winsys, fence, 0); + st->pipe->winsys->fence_reference(st->pipe->winsys, &fence, NULL); } @@ -112,7 +116,7 @@ static void st_glFlush(GLcontext *ctx) */ static void st_glFinish(GLcontext *ctx) { - st_gl_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_WAIT); + st_finish( ctx->st ); } diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 4cf9adcd28..e9fcdf69a1 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -160,7 +160,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, return; /* make sure rendering has completed */ - pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE); + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); if (format == GL_STENCIL_INDEX) { st_read_stencil_pixels(ctx, x, y, width, height, type, pack, dest); diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index d46a9178b1..075e9d1bd6 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -186,7 +186,8 @@ st_notify_swapbuffers(struct st_framebuffer *stfb) if (ctx && ctx->DrawBuffer == &stfb->Base) { st_flush( ctx->st, - PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_SWAPBUFFERS); + PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_SWAPBUFFERS, + NULL ); } } diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 3c397b126a..9d88ce9764 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -45,6 +45,7 @@ struct st_context; struct st_framebuffer; struct pipe_context; +struct pipe_fence_handle; struct pipe_surface; @@ -78,7 +79,9 @@ void st_make_current(struct st_context *st, struct st_framebuffer *draw, struct st_framebuffer *read); -void st_flush( struct st_context *st, uint pipeFlushFlags ); +void st_flush( struct st_context *st, uint pipeFlushFlags, + struct pipe_fence_handle **fence ); +void st_finish( struct st_context *st ); void st_notify_swapbuffers(struct st_framebuffer *stfb); void st_notify_swapbuffers_complete(struct st_framebuffer *stfb); -- cgit v1.2.3 From 8cb85807d3bd42cb0e511970e4b409c542d2716b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 26 Mar 2008 08:21:17 -0600 Subject: gallium: as for aapoints, make the extra texcoord per-shader state --- src/gallium/auxiliary/draw/draw_aaline.c | 24 +++++++++++++----------- 1 file changed, 13 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c index a999ef1227..e8d2a45102 100644 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ b/src/gallium/auxiliary/draw/draw_aaline.c @@ -62,6 +62,7 @@ struct aaline_fragment_shader void *aapoint_fs; /* not yet */ void *sprite_fs; /* not yet */ uint sampler_unit; + int generic_attrib; /**< texcoord/generic used for texture */ }; @@ -336,7 +337,7 @@ static void generate_aaline_fs(struct aaline_stage *aaline) { const struct pipe_shader_state *orig_fs = &aaline->fs->state; - struct draw_context *draw = aaline->stage.draw; + //struct draw_context *draw = aaline->stage.draw; struct pipe_shader_state aaline_fs; struct aa_transform_context transform; @@ -369,11 +370,7 @@ generate_aaline_fs(struct aaline_stage *aaline) aaline->fs->aaline_fs = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); - /* advertise the extra post-transform vertex attributes which will have - * the texcoords. - */ - draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; - draw->extra_vp_outputs.semantic_index = transform.maxGeneric + 1; + aaline->fs->generic_attrib = transform.maxGeneric + 1; } @@ -637,15 +634,20 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) else aaline->half_line_width = 0.5f * draw->rasterizer->line_width; - aaline->tex_slot = draw->num_vs_outputs; - assert(aaline->tex_slot > 0); /* output[0] is vertex pos */ - draw->extra_vp_outputs.slot = aaline->tex_slot; - /* - * Bind our fragprog, sampler and texture + * Bind (generate) our fragprog, sampler and texture */ bind_aaline_fragment_shader(aaline); + /* update vertex attrib info */ + aaline->tex_slot = draw->num_vs_outputs; + assert(aaline->tex_slot > 0); /* output[0] is vertex pos */ + + /* advertise the extra post-transformed vertex attribute */ + draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; + draw->extra_vp_outputs.semantic_index = aaline->fs->generic_attrib; + draw->extra_vp_outputs.slot = aaline->tex_slot; + /* how many samplers? */ /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ num_samplers = MAX2(aaline->num_textures, aaline->num_samplers); -- cgit v1.2.3 From 1ecb2e4a7a5881d5a98679b421d78fd11c729ebc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 26 Mar 2008 09:02:54 -0600 Subject: gallium: need to call st_validate_state() in Bitmap() --- src/mesa/state_tracker/st_cb_bitmap.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 6e59439874..0291b03143 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -475,6 +475,8 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct st_context *st = ctx->st; struct pipe_texture *pt; + st_validate_state(st); + stfp = combined_bitmap_fragment_program(ctx); if (!st->bitmap.vs) { -- cgit v1.2.3 From 92126cea846959bb2152905a7712753d1114bd6b Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 26 Mar 2008 10:45:32 -0700 Subject: cell: Implement code-gen for logic op This also implements code-gen for the float-to-packed color conversion. It's currently hardcoded for A8R8G8B8, but that can easily be fixed as soon as other color depths are supported by the Cell driver. --- src/gallium/drivers/cell/common.h | 11 +- src/gallium/drivers/cell/ppu/cell_context.h | 2 + src/gallium/drivers/cell/ppu/cell_state_emit.c | 17 ++ .../drivers/cell/ppu/cell_state_per_fragment.c | 261 ++++++++++++++++++++- .../drivers/cell/ppu/cell_state_per_fragment.h | 4 + src/gallium/drivers/cell/spu/spu_main.c | 19 ++ src/gallium/drivers/cell/spu/spu_main.h | 9 +- src/gallium/drivers/cell/spu/spu_tri.c | 59 +++-- 8 files changed, 349 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index d59e4f7036..b0928fefd2 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -92,8 +92,9 @@ #define CELL_CMD_STATE_BIND_VS 18 #define CELL_CMD_STATE_BLEND 19 #define CELL_CMD_STATE_ATTRIB_FETCH 20 -#define CELL_CMD_VS_EXECUTE 21 -#define CELL_CMD_FLUSH_BUFFER_RANGE 22 +#define CELL_CMD_STATE_LOGICOP 21 +#define CELL_CMD_VS_EXECUTE 22 +#define CELL_CMD_FLUSH_BUFFER_RANGE 23 #define CELL_NUM_BUFFERS 4 @@ -124,6 +125,12 @@ struct cell_command_blend { }; +struct cell_command_logicop { + uint64_t base; /**< Effective address of code start. */ + unsigned size; /**< Size in bytes of test code. */ +}; + + /** * Tell SPUs about the framebuffer size, location */ diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 9e79db0ace..0442abddc1 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -92,6 +92,8 @@ struct cell_context const struct cell_vertex_shader_state *vs; const struct cell_fragment_shader_state *fs; + struct spe_function logic_op; + struct pipe_blend_color blend_color; struct pipe_clip_state clip; struct pipe_constant_buffer constants[2]; diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 5709b48f12..5c1310d0b0 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -50,6 +50,23 @@ emit_state_cmd(struct cell_context *cell, uint cmd, void cell_emit_state(struct cell_context *cell) { + if (cell->dirty & (CELL_NEW_FRAMEBUFFER | CELL_NEW_BLEND)) { + struct cell_command_logicop logicop; + + if (cell->logic_op.store != NULL) { + spe_release_func(& cell->logic_op); + } + + cell_generate_logic_op(& cell->logic_op, + & cell->blend->base, + cell->framebuffer.cbufs[0]); + + logicop.base = (intptr_t) cell->logic_op.store; + logicop.size = 64 * 4; + emit_state_cmd(cell, CELL_CMD_STATE_LOGICOP, &logicop, + sizeof(logicop)); + } + if (cell->dirty & CELL_NEW_FRAMEBUFFER) { struct pipe_surface *cbuf = cell->framebuffer.cbufs[0]; struct pipe_surface *zbuf = cell->framebuffer.zsbuf; diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c index c750b1d89d..f10025bd7c 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -977,7 +977,6 @@ cell_generate_alpha_blend(struct cell_blend_state *cb) spe_allocate_register(f, 13), spe_allocate_register(f, 14), }; - const int mask = spe_allocate_register(f, 15); unsigned func[4]; unsigned sF[4]; unsigned dF[4]; @@ -1114,9 +1113,6 @@ cell_generate_alpha_blend(struct cell_blend_state *cb) func[i], sF[i], dF[i], frag[i], src_factor[i], pixel[i], dst_factor[i]); - spe_selb(f, frag[i], pixel[i], frag[i], mask); - } else { - spe_or(f, frag[i], pixel[i], pixel[i]); } } @@ -1146,3 +1142,260 @@ cell_generate_alpha_blend(struct cell_blend_state *cb) } #endif } + + +int PC_OFFSET(const struct spe_function *f, const void *d) +{ + const intptr_t pc = (intptr_t) f->csr; + const intptr_t ea = ~0x0f & (intptr_t) d; + + return (ea - pc) >> 2; +} + + +/** + * Generate code to perform color conversion and logic op + * + * \bug + * The code generated by this function should also perform dithering. + * + * \bug + * The code generated by this function should also perform color-write + * masking. + * + * \bug + * This routine is hard-coded to only work with ARGB8 data. + */ +void +cell_generate_logic_op(struct spe_function *f, struct pipe_blend_state *blend, + struct pipe_surface *surf) +{ + const unsigned logic_op = (blend->logicop_enable) + ? blend->logicop_func : PIPE_LOGICOP_COPY; + + /* This code generates a maximum of 37 instructions. An additional 32 + * bytes (equiv. to 8 instructions) are needed for data storage. Round up + * to 64 to make it a happy power-of-two. + */ + spe_init_func(f, 4 * 64); + + + /* Pixel colors in framebuffer format in AoS layout. + */ + const int pixel[4] = { + spe_allocate_register(f, 3), + spe_allocate_register(f, 4), + spe_allocate_register(f, 5), + spe_allocate_register(f, 6), + }; + + /* Fragment colors stored as floats in SoA layout. + */ + const int frag[4] = { + spe_allocate_register(f, 7), + spe_allocate_register(f, 8), + spe_allocate_register(f, 9), + spe_allocate_register(f, 10), + }; + + const int mask = spe_allocate_register(f, 11); + + + /* Short-circuit the noop and invert cases. + */ + if ((logic_op == PIPE_LOGICOP_NOOP) || (blend->colormask == 0)) { + spe_bi(f, 0, 0, 0); + return; + } else if (logic_op == PIPE_LOGICOP_INVERT) { + spe_nor(f, pixel[0], pixel[0], pixel[0]); + spe_nor(f, pixel[1], pixel[1], pixel[1]); + spe_nor(f, pixel[2], pixel[2], pixel[2]); + spe_nor(f, pixel[3], pixel[3], pixel[3]); + spe_bi(f, 0, 0, 0); + return; + } + + + const int tmp[4] = { + spe_allocate_available_register(f), + spe_allocate_available_register(f), + spe_allocate_available_register(f), + spe_allocate_available_register(f), + }; + + const int shuf_xpose_hi = spe_allocate_available_register(f); + const int shuf_xpose_lo = spe_allocate_available_register(f); + const int shuf_color = spe_allocate_available_register(f); + + + /* Pointer to the begining of the function's private data area. + */ + uint32_t *const data = ((uint32_t *) f->store) + (64 - 8); + + + /* Convert fragment colors to framebuffer format in AoS layout. + */ + data[0] = 0x00010203; + data[1] = 0x10111213; + data[2] = 0x04050607; + data[3] = 0x14151617; + + data[4] = 0x0c000408; + data[5] = 0x80808080; + data[6] = 0x80808080; + data[7] = 0x80808080; + + spe_ilh(f, tmp[0], 0x0808); + spe_lqr(f, shuf_xpose_hi, PC_OFFSET(f, data+0)); + spe_lqr(f, shuf_color, PC_OFFSET(f, data+4)); + spe_a(f, shuf_xpose_lo, shuf_xpose_hi, tmp[0]); + + spe_shufb(f, tmp[0], frag[0], frag[2], shuf_xpose_hi); + spe_shufb(f, tmp[1], frag[0], frag[2], shuf_xpose_lo); + spe_shufb(f, tmp[2], frag[1], frag[3], shuf_xpose_hi); + spe_shufb(f, tmp[3], frag[1], frag[3], shuf_xpose_lo); + + spe_shufb(f, frag[0], tmp[0], tmp[2], shuf_xpose_hi); + spe_shufb(f, frag[1], tmp[0], tmp[2], shuf_xpose_lo); + spe_shufb(f, frag[2], tmp[1], tmp[3], shuf_xpose_hi); + spe_shufb(f, frag[3], tmp[1], tmp[3], shuf_xpose_lo); + + spe_cfltu(f, frag[0], frag[0], 32); + spe_cfltu(f, frag[1], frag[1], 32); + spe_cfltu(f, frag[2], frag[2], 32); + spe_cfltu(f, frag[3], frag[3], 32); + + spe_shufb(f, frag[0], frag[0], pixel[0], shuf_color); + spe_shufb(f, frag[1], frag[1], pixel[1], shuf_color); + spe_shufb(f, frag[2], frag[2], pixel[2], shuf_color); + spe_shufb(f, frag[3], frag[3], pixel[3], shuf_color); + + + /* If logic op is enabled, perform the requested logical operation on the + * converted fragment colors and the pixel colors. + */ + switch (logic_op) { + case PIPE_LOGICOP_CLEAR: + spe_il(f, frag[0], 0); + spe_il(f, frag[1], 0); + spe_il(f, frag[2], 0); + spe_il(f, frag[3], 0); + break; + case PIPE_LOGICOP_NOR: + spe_nor(f, frag[0], frag[0], pixel[0]); + spe_nor(f, frag[1], frag[1], pixel[1]); + spe_nor(f, frag[2], frag[2], pixel[2]); + spe_nor(f, frag[3], frag[3], pixel[3]); + break; + case PIPE_LOGICOP_AND_INVERTED: + spe_andc(f, frag[0], pixel[0], frag[0]); + spe_andc(f, frag[1], pixel[1], frag[1]); + spe_andc(f, frag[2], pixel[2], frag[2]); + spe_andc(f, frag[3], pixel[3], frag[3]); + break; + case PIPE_LOGICOP_COPY_INVERTED: + spe_nor(f, frag[0], frag[0], frag[0]); + spe_nor(f, frag[1], frag[1], frag[1]); + spe_nor(f, frag[2], frag[2], frag[2]); + spe_nor(f, frag[3], frag[3], frag[3]); + break; + case PIPE_LOGICOP_AND_REVERSE: + spe_andc(f, frag[0], frag[0], pixel[0]); + spe_andc(f, frag[1], frag[1], pixel[1]); + spe_andc(f, frag[2], frag[2], pixel[2]); + spe_andc(f, frag[3], frag[3], pixel[3]); + break; + case PIPE_LOGICOP_XOR: + spe_xor(f, frag[0], frag[0], pixel[0]); + spe_xor(f, frag[1], frag[1], pixel[1]); + spe_xor(f, frag[2], frag[2], pixel[2]); + spe_xor(f, frag[3], frag[3], pixel[3]); + break; + case PIPE_LOGICOP_NAND: + spe_nand(f, frag[0], frag[0], pixel[0]); + spe_nand(f, frag[1], frag[1], pixel[1]); + spe_nand(f, frag[2], frag[2], pixel[2]); + spe_nand(f, frag[3], frag[3], pixel[3]); + break; + case PIPE_LOGICOP_AND: + spe_and(f, frag[0], frag[0], pixel[0]); + spe_and(f, frag[1], frag[1], pixel[1]); + spe_and(f, frag[2], frag[2], pixel[2]); + spe_and(f, frag[3], frag[3], pixel[3]); + break; + case PIPE_LOGICOP_EQUIV: + spe_eqv(f, frag[0], frag[0], pixel[0]); + spe_eqv(f, frag[1], frag[1], pixel[1]); + spe_eqv(f, frag[2], frag[2], pixel[2]); + spe_eqv(f, frag[3], frag[3], pixel[3]); + break; + case PIPE_LOGICOP_OR_INVERTED: + spe_orc(f, frag[0], pixel[0], frag[0]); + spe_orc(f, frag[1], pixel[1], frag[1]); + spe_orc(f, frag[2], pixel[2], frag[2]); + spe_orc(f, frag[3], pixel[3], frag[3]); + break; + case PIPE_LOGICOP_COPY: + break; + case PIPE_LOGICOP_OR_REVERSE: + spe_orc(f, frag[0], frag[0], pixel[0]); + spe_orc(f, frag[1], frag[1], pixel[1]); + spe_orc(f, frag[2], frag[2], pixel[2]); + spe_orc(f, frag[3], frag[3], pixel[3]); + break; + case PIPE_LOGICOP_OR: + spe_or(f, frag[0], frag[0], pixel[0]); + spe_or(f, frag[1], frag[1], pixel[1]); + spe_or(f, frag[2], frag[2], pixel[2]); + spe_or(f, frag[3], frag[3], pixel[3]); + break; + case PIPE_LOGICOP_SET: + spe_il(f, frag[0], ~0); + spe_il(f, frag[1], ~0); + spe_il(f, frag[2], ~0); + spe_il(f, frag[3], ~0); + break; + + /* These two cases are short-circuited above. + */ + case PIPE_LOGICOP_INVERT: + case PIPE_LOGICOP_NOOP: + default: + assert(0); + } + + + /* Apply fragment mask. + */ + spe_ilh(f, tmp[0], 0x0000); + spe_ilh(f, tmp[1], 0x0404); + spe_ilh(f, tmp[2], 0x0808); + spe_ilh(f, tmp[3], 0x0c0c); + + spe_shufb(f, tmp[0], mask, mask, tmp[0]); + spe_shufb(f, tmp[1], mask, mask, tmp[1]); + spe_shufb(f, tmp[2], mask, mask, tmp[2]); + spe_shufb(f, tmp[3], mask, mask, tmp[3]); + + spe_selb(f, pixel[0], pixel[0], frag[0], tmp[0]); + spe_selb(f, pixel[1], pixel[1], frag[1], tmp[1]); + spe_selb(f, pixel[2], pixel[2], frag[2], tmp[2]); + spe_selb(f, pixel[3], pixel[3], frag[3], tmp[3]); + + spe_bi(f, 0, 0, 0); + +#if 0 + { + const uint32_t *p = f->store; + unsigned i; + + printf("# %u instructions\n", f->csr - f->store); + + printf("\t.text\n"); + for (i = 0; i < 64; i++) { + printf("\t.long\t0x%04x\n", p[i]); + } + fflush(stdout); + } +#endif +} diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h index f699247f9e..ab4de96c69 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h @@ -31,4 +31,8 @@ cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa); extern void cell_generate_alpha_blend(struct cell_blend_state *cb); +extern void +cell_generate_logic_op(struct spe_function *f, struct pipe_blend_state *blend, + struct pipe_surface *surf); + #endif /* CELL_STATE_PER_FRAGMENT_H */ diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 0a490ab277..fccff01e10 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -64,6 +64,9 @@ static unsigned char depth_stencil_code_buffer[4 * 64] static unsigned char fb_blend_code_buffer[4 * 64] ALIGN16_ATTRIB; +static unsigned char logicop_code_buffer[4 * 64] + ALIGN16_ATTRIB; + /** * Tell the PPU that this SPU has finished copying a buffer to @@ -513,6 +516,22 @@ cmd_batch(uint opcode) pos += (1 + ROUNDUP8(sizeof(struct cell_attribute_fetch_code)) / 8); break; } + case CELL_CMD_STATE_LOGICOP: { + struct cell_command_logicop *code = + (struct cell_command_logicop *) &buffer[pos+1]; + + mfc_get(logicop_code_buffer, + (unsigned int) code->base, /* src */ + code->size, + TAG_BATCH_BUFFER, + 0, /* tid */ + 0 /* rid */); + wait_on_mask(1 << TAG_BATCH_BUFFER); + + spu.logicop = (logicop_func) logicop_code_buffer; + pos += (1 + ROUNDUP8(sizeof(struct cell_command_logicop)) / 8); + break; + } case CELL_CMD_FLUSH_BUFFER_RANGE: { struct cell_buffer_range *br = (struct cell_buffer_range *) &buffer[pos+1]; diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index 49f5d99674..c20452931a 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -77,9 +77,14 @@ struct spu_blend_results { typedef struct spu_blend_results (*blend_func)( qword frag_r, qword frag_g, qword frag_b, qword frag_a, qword pixel_r, qword pixel_g, qword pixel_b, qword pixel_a, - qword const_r, qword const_g, qword const_b, qword const_a, + qword const_r, qword const_g, qword const_b, qword const_a); + +typedef struct spu_blend_results (*logicop_func)( + qword pixel_r, qword pixel_g, qword pixel_b, qword pixel_a, + qword frag_r, qword frag_g, qword frag_b, qword frag_a, qword frag_mask); + struct spu_framebuffer { void *color_start; /**< addr of color surface in main memory */ void *depth_start; /**< addr of depth surface in main memory */ @@ -111,6 +116,8 @@ struct spu_global blend_func blend; qword const_blend_color[4] ALIGN16_ATTRIB; + logicop_func logicop; + struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; struct cell_command_texture texture; diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index e6a1ce01df..95c629a8aa 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -305,7 +305,6 @@ emit_quad( int x, int y, mask_t mask ) if (spu_extract(spu_orx(mask), 0)) { const int ix = x - setup.cliprect_minx; const int iy = y - setup.cliprect_miny; - const vector unsigned char shuffle = spu.color_shuffle; vector float colors[4]; spu.cur_ctile_status = TILE_STATUS_DIRTY; @@ -330,45 +329,53 @@ emit_quad( int x, int y, mask_t mask ) } + /* Convert fragment data from AoS to SoA format. + */ + qword soa_frag[4]; + _transpose_matrix4x4((vec_float4 *) soa_frag, colors); + /* Read the current framebuffer values. - * - * Ignore read_fb for now. In the future we can use this to avoid - * reading the framebuffer if read_fb is false and the fragment mask is - * all 0xffffffff. This is the common case, so it is probably worth - * the effort. We'll have to profile to determine whether or not the - * extra conditional branches hurt overall performance. */ - vec_float4 aos_pix[4] = { - spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+0]), - spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+1]), - spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+0]), - spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+1]), + const qword pix[4] = { + (qword) spu_splats(spu.ctile.ui[iy+0][ix+0]), + (qword) spu_splats(spu.ctile.ui[iy+0][ix+1]), + (qword) spu_splats(spu.ctile.ui[iy+1][ix+0]), + (qword) spu_splats(spu.ctile.ui[iy+1][ix+1]), }; qword soa_pix[4]; - qword soa_frag[4]; - /* Convert pixel and fragment data from AoS to SoA format. - */ - _transpose_matrix4x4((vec_float4 *) soa_pix, aos_pix); - _transpose_matrix4x4((vec_float4 *) soa_frag, colors); + if (spu.read_fb) { + /* Convert pixel data from AoS to SoA format. + */ + vec_float4 aos_pix[4] = { + spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+0]), + spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+1]), + spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+0]), + spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+1]), + }; + + _transpose_matrix4x4((vec_float4 *) soa_pix, aos_pix); + } - const struct spu_blend_results result = + + struct spu_blend_results result = (*spu.blend)(soa_frag[0], soa_frag[1], soa_frag[2], soa_frag[3], soa_pix[0], soa_pix[1], soa_pix[2], soa_pix[3], spu.const_blend_color[0], spu.const_blend_color[1], - spu.const_blend_color[2], spu.const_blend_color[3], - (qword) mask); + spu.const_blend_color[2], spu.const_blend_color[3]); /* Convert final pixel data from SoA to AoS format. */ - _transpose_matrix4x4(aos_pix, (const vec_float4 *) &result); - - spu.ctile.ui[iy+0][ix+0] = spu_pack_color_shuffle(aos_pix[0], shuffle); - spu.ctile.ui[iy+0][ix+1] = spu_pack_color_shuffle(aos_pix[1], shuffle); - spu.ctile.ui[iy+1][ix+0] = spu_pack_color_shuffle(aos_pix[2], shuffle); - spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(aos_pix[3], shuffle); + result = (*spu.logicop)(pix[0], pix[1], pix[2], pix[3], + result.r, result.g, result.b, result.a, + (qword) mask); + + spu.ctile.ui[iy+0][ix+0] = spu_extract((vec_uint4) result.r, 0); + spu.ctile.ui[iy+0][ix+1] = spu_extract((vec_uint4) result.g, 0); + spu.ctile.ui[iy+1][ix+0] = spu_extract((vec_uint4) result.b, 0); + spu.ctile.ui[iy+1][ix+1] = spu_extract((vec_uint4) result.a, 0); } #endif } -- cgit v1.2.3 From 70659e8ec8ba0c599daa36d9ed81e86efd99eb52 Mon Sep 17 00:00:00 2001 From: Ian Romanick Date: Wed, 26 Mar 2008 10:52:08 -0700 Subject: xlib: Fix build error from recent fence changes --- src/gallium/winsys/xlib/xm_winsys.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index c930a1d196..7459756279 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -630,9 +630,9 @@ xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis) ws->base.surface_alloc_storage = xm_surface_alloc_storage; ws->base.surface_release = xm_surface_release; - ws->fence_reference = xm_fence_reference; - ws->fence_signalled = xm_fence_signalled; - ws->fence_finish = xm_fence_finish; + ws->base.fence_reference = xm_fence_reference; + ws->base.fence_signalled = xm_fence_signalled; + ws->base.fence_finish = xm_fence_finish; ws->base.flush_frontbuffer = xm_flush_frontbuffer; ws->base.printf = xm_printf; -- cgit v1.2.3 From 979358c47115d8ea50001832372f8043a60a5b80 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:26:22 -0600 Subject: cell: fix unclosed comment --- src/gallium/drivers/cell/spu/spu_exec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 061fbebf61..48edc62f49 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -1453,7 +1453,7 @@ exec_instruction( break; case TGSI_OPCODE_TXP: - /* Texture lookup with projection + /* Texture lookup with projection */ /* src[0] = texcoord (src[0].w = projection) */ /* src[1] = sampler unit */ exec_tex(mach, inst, TRUE, TRUE); -- cgit v1.2.3 From 8b8a947111ad911a986e48a66c9fe31f120de9a2 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:26:54 -0600 Subject: cell: added (uint64_t) cast to silence warning --- src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index b418857ccd..f753960a0f 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -69,7 +69,7 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*cf)); batch[0] = CELL_CMD_STATE_ATTRIB_FETCH; cf = (struct cell_attribute_fetch_code *) (&batch[1]); - cf->base = cell->attrib_fetch.store; + cf->base = (uint64_t) cell->attrib_fetch.store; cf->size = ROUNDUP16((unsigned)((void *) cell->attrib_fetch.csr - (void *) cell->attrib_fetch.store)); -- cgit v1.2.3 From 132df5ebce8529f7b0999cdd574d70ec0f12a4fd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:27:14 -0600 Subject: cell: include cell_state_per_fragment.h to silence warning --- src/gallium/drivers/cell/ppu/cell_state_emit.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 5c1310d0b0..4c75caa025 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -29,6 +29,7 @@ #include "cell_context.h" #include "cell_state.h" #include "cell_state_emit.h" +#include "cell_state_per_fragment.h" #include "cell_batch.h" #include "cell_texture.h" #include "draw/draw_context.h" -- cgit v1.2.3 From d355eee5cacac30e2b4c8ac2f10964e2861a539e Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:27:31 -0600 Subject: gallium: silence unused var warning --- src/gallium/include/pipe/p_debug.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 097fc7c0b1..9e52ad9a37 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -90,6 +90,8 @@ debug_printf(const char *format, ...) va_start(ap, format); _debug_vprintf(format, ap); va_end(ap); +#else + (void) format; /* silence warning */ #endif } -- cgit v1.2.3 From df1744c0433f3f73ebf4b06567fefa946a29c3d8 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:33:47 -0600 Subject: gallium: remove temporary static var --- src/gallium/auxiliary/draw/draw_prim.c | 11 +++++------ src/gallium/auxiliary/draw/draw_private.h | 1 + 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index f589d0c017..ddcde01d9a 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -158,15 +158,15 @@ static INLINE void fetch_and_store(struct draw_context *draw) void draw_do_flush( struct draw_context *draw, unsigned flags ) { - static boolean flushing = FALSE; - if (0) debug_printf("Flushing with %d verts, %d prims\n", draw->vs.queue_nr, draw->pq.queue_nr ); - if (!flushing) { - flushing = TRUE; + if (draw->flushing) + return; + + draw->flushing = TRUE; if (flags >= DRAW_FLUSH_SHADER_QUEUE) { if (draw->vs.queue_nr) { @@ -189,8 +189,7 @@ void draw_do_flush( struct draw_context *draw, unsigned flags ) } } - flushing = FALSE; - } + draw->flushing = FALSE; } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 8a879809c3..7007ee22c4 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -223,6 +223,7 @@ struct draw_context } pt; + boolean flushing; /* pipe state that we need: */ const struct pipe_rasterizer_state *rasterizer; -- cgit v1.2.3 From ba49fa39f3028d1ee44a999d84b29a0cfbfab0bd Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:35:25 -0600 Subject: gallium: fix incorrect types for shaders --- src/gallium/auxiliary/util/u_blit.c | 5 ++--- src/gallium/auxiliary/util/u_gen_mipmap.c | 6 +++--- 2 files changed, 5 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 28a404fd01..9986dc3570 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -58,9 +58,8 @@ struct blit_state struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state sampler; - /*struct pipe_viewport_state viewport;*/ - struct pipe_sampler_state *vs; - struct pipe_sampler_state *fs; + void *vs; + void *fs; struct pipe_buffer *vbuf; /**< quad vertices */ float vertices[4][2][4]; /**< vertex/texcoords for quad */ diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index ed12768e7f..5549652abb 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -61,9 +61,9 @@ struct gen_mipmap_state struct pipe_depth_stencil_alpha_state depthstencil; struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state sampler; - /*struct pipe_viewport_state viewport;*/ - struct pipe_sampler_state *vs; - struct pipe_sampler_state *fs; + + void *vs; + void *fs; struct pipe_buffer *vbuf; /**< quad vertices */ float vertices[4][2][4]; /**< vertex/texcoords for quad */ -- cgit v1.2.3 From 89222ee49de340774279c9c0bf884649e66ad6df Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:41:52 -0600 Subject: gallium: fix Element() macro redefinition warning --- src/mesa/state_tracker/st_texture.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index cbc6f84998..8e3235cc99 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -30,6 +30,8 @@ #include "st_texture.h" #include "enums.h" +#undef Elements /* fix re-defined macro warning */ + #include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -- cgit v1.2.3 From dccbfd8bf0624250a435948029916073d3390191 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:42:09 -0600 Subject: gallium: return pipe_shader_state from the simple shader functions Allows us to fix a mem leak (tokens array). --- src/gallium/auxiliary/util/u_blit.c | 10 +++++++-- src/gallium/auxiliary/util/u_gen_mipmap.c | 10 +++++++-- src/gallium/auxiliary/util/u_simple_shaders.c | 31 ++++++++++++++++----------- src/gallium/auxiliary/util/u_simple_shaders.h | 10 ++++++--- 4 files changed, 42 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 9986dc3570..eec5e600c9 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -58,6 +58,8 @@ struct blit_state struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state sampler; + struct pipe_shader_state vert_shader; + struct pipe_shader_state frag_shader; void *vs; void *fs; @@ -129,11 +131,12 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) TGSI_SEMANTIC_GENERIC }; const uint semantic_indexes[] = { 0, 0 }; ctx->vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names, - semantic_indexes); + semantic_indexes, + &ctx->vert_shader); } /* fragment shader */ - ctx->fs = util_make_fragment_tex_shader(pipe); + ctx->fs = util_make_fragment_tex_shader(pipe, &ctx->frag_shader); ctx->vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, @@ -168,6 +171,9 @@ util_destroy_blit(struct blit_state *ctx) pipe->delete_vs_state(pipe, ctx->vs); pipe->delete_fs_state(pipe, ctx->fs); + FREE((void*) ctx->vert_shader.tokens); + FREE((void*) ctx->frag_shader.tokens); + pipe->winsys->buffer_destroy(pipe->winsys, ctx->vbuf); FREE(ctx); diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 5549652abb..2fd214d22e 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -62,6 +62,8 @@ struct gen_mipmap_state struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state sampler; + struct pipe_shader_state vert_shader; + struct pipe_shader_state frag_shader; void *vs; void *fs; @@ -740,11 +742,12 @@ util_create_gen_mipmap(struct pipe_context *pipe, TGSI_SEMANTIC_GENERIC }; const uint semantic_indexes[] = { 0, 0 }; ctx->vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names, - semantic_indexes); + semantic_indexes, + &ctx->vert_shader); } /* fragment shader */ - ctx->fs = util_make_fragment_tex_shader(pipe); + ctx->fs = util_make_fragment_tex_shader(pipe, &ctx->frag_shader); ctx->vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, @@ -813,6 +816,9 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx) pipe->delete_vs_state(pipe, ctx->vs); pipe->delete_fs_state(pipe, ctx->fs); + FREE((void*) ctx->vert_shader.tokens); + FREE((void*) ctx->frag_shader.tokens); + pipe->winsys->buffer_destroy(pipe->winsys, ctx->vbuf); FREE(ctx); diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index d230ddfbeb..5f8d12191d 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -56,7 +56,9 @@ void * util_make_vertex_passthrough_shader(struct pipe_context *pipe, uint num_attribs, const uint *semantic_names, - const uint *semantic_indexes) + const uint *semantic_indexes, + struct pipe_shader_state *shader) + { uint maxTokens = 100; struct tgsi_token *tokens; @@ -66,7 +68,6 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, struct tgsi_full_instruction inst; const uint procType = TGSI_PROCESSOR_VERTEX; uint ti, i; - struct pipe_shader_state shader; tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0])); @@ -145,8 +146,10 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, tgsi_dump(tokens, 0); #endif - shader.tokens = tokens; - return pipe->create_vs_state(pipe, &shader); + shader->tokens = tokens; + /*shader->num_tokens = ti;*/ + + return pipe->create_vs_state(pipe, shader); } @@ -158,7 +161,8 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, * END; */ void * -util_make_fragment_tex_shader(struct pipe_context *pipe) +util_make_fragment_tex_shader(struct pipe_context *pipe, + struct pipe_shader_state *shader) { uint maxTokens = 100; struct tgsi_token *tokens; @@ -168,7 +172,6 @@ util_make_fragment_tex_shader(struct pipe_context *pipe) struct tgsi_full_instruction inst; const uint procType = TGSI_PROCESSOR_FRAGMENT; uint ti; - struct pipe_shader_state shader; tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0])); @@ -254,8 +257,10 @@ util_make_fragment_tex_shader(struct pipe_context *pipe) tgsi_dump(tokens, 0); #endif - shader.tokens = tokens; - return pipe->create_fs_state(pipe, &shader); + shader->tokens = tokens; + /*shader->num_tokens = ti;*/ + + return pipe->create_fs_state(pipe, shader); } @@ -266,7 +271,8 @@ util_make_fragment_tex_shader(struct pipe_context *pipe) * Make simple fragment color pass-through shader. */ void * -util_make_fragment_passthrough_shader(struct pipe_context *pipe) +util_make_fragment_passthrough_shader(struct pipe_context *pipe, + struct pipe_shader_state *shader) { uint maxTokens = 40; struct tgsi_token *tokens; @@ -276,7 +282,6 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe) struct tgsi_full_instruction inst; const uint procType = TGSI_PROCESSOR_FRAGMENT; uint ti; - struct pipe_shader_state shader; tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0])); @@ -349,7 +354,9 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe) tgsi_dump(tokens, 0); #endif - shader.tokens = tokens; - return pipe->create_fs_state(pipe, &shader); + shader->tokens = tokens; + /*shader->num_tokens = ti;*/ + + return pipe->create_fs_state(pipe, shader); } diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h index ca219a092c..8ca4977d71 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.h +++ b/src/gallium/auxiliary/util/u_simple_shaders.h @@ -34,6 +34,7 @@ struct pipe_context; +struct pipe_shader_state; #ifdef __cplusplus @@ -45,15 +46,18 @@ extern void * util_make_vertex_passthrough_shader(struct pipe_context *pipe, uint num_attribs, const uint *semantic_names, - const uint *semantic_indexes); + const uint *semantic_indexes, + struct pipe_shader_state *shader); extern void * -util_make_fragment_tex_shader(struct pipe_context *pipe); +util_make_fragment_tex_shader(struct pipe_context *pipe, + struct pipe_shader_state *shader); extern void * -util_make_fragment_passthrough_shader(struct pipe_context *pipe); +util_make_fragment_passthrough_shader(struct pipe_context *pipe, + struct pipe_shader_state *shader); #ifdef __cplusplus -- cgit v1.2.3 From 6f8286163c79a8187c2912a9b673a6f11f4f60c6 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:42:52 -0600 Subject: gallium: Update calls to the simple shader functions --- src/mesa/state_tracker/st_cb_bitmap.c | 4 +++- src/mesa/state_tracker/st_cb_clear.c | 5 +++-- src/mesa/state_tracker/st_context.h | 3 +++ 3 files changed, 9 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 0291b03143..64b1882424 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -487,13 +487,15 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, const uint semantic_indexes[] = { 0, 0, 0 }; st->bitmap.vs = util_make_vertex_passthrough_shader(st->pipe, 3, semantic_names, - semantic_indexes); + semantic_indexes, + &st->bitmap.vert_shader); } st_validate_state(st); pt = make_bitmap_texture(ctx, width, height, unpack, bitmap); if (pt) { + assert(pt->target == PIPE_TEXTURE_2D); draw_bitmap_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, pt, stfp); diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 5ca15df602..ec8d3e1022 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -251,7 +251,7 @@ clear_with_quad(GLcontext *ctx, /* fragment shader state: color pass-through program */ if (!st->clear.fs) { - st->clear.fs = util_make_fragment_passthrough_shader(pipe); + st->clear.fs = util_make_fragment_passthrough_shader(pipe, &st->clear.frag_shader); } pipe->bind_fs_state(pipe, st->clear.fs); @@ -264,7 +264,8 @@ clear_with_quad(GLcontext *ctx, const uint semantic_indexes[] = { 0, 0 }; st->clear.vs = util_make_vertex_passthrough_shader(pipe, 2, semantic_names, - semantic_indexes); + semantic_indexes, + &st->clear.vert_shader); } pipe->bind_vs_state(pipe, st->clear.vs); #endif diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 2d37086799..f235c194b7 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -147,6 +147,7 @@ struct st_context struct st_fragment_program *program; /**< bitmap tex/kil program */ GLuint user_prog_sn; /**< user fragment program serial no. */ struct st_fragment_program *combined_prog; + struct pipe_shader_state vert_shader; void *vs; float vertices[4][3][4]; /**< vertex pos + color + texcoord */ struct pipe_buffer *vbuf; @@ -154,6 +155,8 @@ struct st_context /** for glClear */ struct { + struct pipe_shader_state vert_shader; + struct pipe_shader_state frag_shader; void *vs; void *fs; float vertices[4][2][4]; /**< vertex pos + color */ -- cgit v1.2.3 From 7b91d875401e039635718d1bc1e96d37ec66777c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:48:54 -0600 Subject: Fix state.texgen parsing error (bug 12313). Replace *(*inst++) with *(*inst)++ in a few places. Cherry-picked from master. --- src/mesa/shader/arbprogparse.c | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 6648ed7b71..83fc195142 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -1227,10 +1227,10 @@ parse_state_single_item (GLcontext * ctx, const GLubyte ** inst, state_tokens[1] = coord; /* EYE or OBJECT */ - type = *(*inst++); + type = *(*inst)++; /* 0 - s, 1 - t, 2 - r, 3 - q */ - coord = *(*inst++); + coord = *(*inst)++; if (type == TEX_GEN_EYE) { switch (coord) { @@ -1246,6 +1246,9 @@ parse_state_single_item (GLcontext * ctx, const GLubyte ** inst, case COMPONENT_W: state_tokens[2] = STATE_TEXGEN_EYE_Q; break; + default: + _mesa_problem(ctx, "bad texgen component in " + "parse_state_single_item()"); } } else { @@ -1262,6 +1265,9 @@ parse_state_single_item (GLcontext * ctx, const GLubyte ** inst, case COMPONENT_W: state_tokens[2] = STATE_TEXGEN_OBJECT_Q; break; + default: + _mesa_problem(ctx, "bad texgen component in " + "parse_state_single_item()"); } } } @@ -1284,7 +1290,7 @@ parse_state_single_item (GLcontext * ctx, const GLubyte ** inst, break; case STATE_POINT: - switch (*(*inst++)) { + switch (*(*inst)++) { case POINT_SIZE: state_tokens[0] = STATE_POINT_SIZE; break; @@ -2455,8 +2461,9 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst, Program->Position = parse_position (inst); if (!found) { - program_error(ctx, Program->Position, - "2: Undefined variable"); /* src->name */ + char s[1000]; + sprintf(s, "Undefined variable: %s", src->name); + program_error(ctx, Program->Position, s); return 1; } -- cgit v1.2.3 From 50a20d472956d54e3dc66e6501f1e5fb8581c84f Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:51:07 -0600 Subject: added program_error2() function for better error reporting cherry-picked from master --- src/mesa/shader/arbprogparse.c | 107 +++++++++++++++++++++++------------------ 1 file changed, 60 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 83fc195142..38e174f3ad 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -1,6 +1,6 @@ /* * Mesa 3-D graphics library - * Version: 6.5.3 + * Version: 7.1 * * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * @@ -634,6 +634,41 @@ program_error(GLcontext *ctx, GLint position, const char *descrip) } +/** + * As above, but with an extra string parameter for more info. + */ +static void +program_error2(GLcontext *ctx, GLint position, const char *descrip, + const char *var) +{ + if (descrip) { + const char *prefix = "glProgramString(", *suffix = ")"; + char *str = (char *) _mesa_malloc(_mesa_strlen(descrip) + + _mesa_strlen(": ") + + _mesa_strlen(var) + + _mesa_strlen(prefix) + + _mesa_strlen(suffix) + 1); + if (str) { + _mesa_sprintf(str, "%s%s: %s%s", prefix, descrip, var, suffix); + _mesa_error(ctx, GL_INVALID_OPERATION, str); + _mesa_free(str); + } + } + { + char *str = (char *) _mesa_malloc(_mesa_strlen(descrip) + + _mesa_strlen(": ") + + _mesa_strlen(var) + 1); + if (str) { + _mesa_sprintf(str, "%s: %s", descrip, var); + } + _mesa_set_program_error(ctx, position, str); + if (str) { + _mesa_free(str); + } + } +} + + /** * constructs an integer from 4 GLubytes in LE format @@ -1692,18 +1727,14 @@ parse_attrib (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head struct arb_program *Program) { GLuint found; - char *error_msg; struct var_cache *attrib_var; attrib_var = parse_string (inst, vc_head, Program, &found); Program->Position = parse_position (inst); if (found) { - error_msg = (char *) - _mesa_malloc (_mesa_strlen ((char *) attrib_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", - attrib_var->name); - program_error(ctx, Program->Position, error_msg); - _mesa_free (error_msg); + program_error2(ctx, Program->Position, + "Duplicate variable declaration", + (char *) attrib_var->name); return 1; } @@ -1883,12 +1914,9 @@ parse_param (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, Program->Position = parse_position (inst); if (found) { - char *error_msg = (char *) - _mesa_malloc (_mesa_strlen ((char *) param_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", - param_var->name); - program_error (ctx, Program->Position, error_msg); - _mesa_free (error_msg); + program_error2(ctx, Program->Position, + "Duplicate variable declaration", + (char *) param_var->name); return 1; } @@ -1983,12 +2011,9 @@ parse_temp (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, temp_var = parse_string (inst, vc_head, Program, &found); Program->Position = parse_position (inst); if (found) { - char *error_msg = (char *) - _mesa_malloc (_mesa_strlen ((char *) temp_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", - temp_var->name); - program_error(ctx, Program->Position, error_msg); - _mesa_free (error_msg); + program_error2(ctx, Program->Position, + "Duplicate variable declaration", + (char *) temp_var->name); return 1; } @@ -2029,12 +2054,9 @@ parse_output (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head output_var = parse_string (inst, vc_head, Program, &found); Program->Position = parse_position (inst); if (found) { - char *error_msg = (char *) - _mesa_malloc (_mesa_strlen ((char *) output_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", - output_var->name); - program_error (ctx, Program->Position, error_msg); - _mesa_free (error_msg); + program_error2(ctx, Program->Position, + "Duplicate variable declaration", + (char *) output_var->name); return 1; } @@ -2060,12 +2082,9 @@ parse_alias (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, Program->Position = parse_position (inst); if (found) { - char *error_msg = (char *) - _mesa_malloc (_mesa_strlen ((char *) temp_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", - temp_var->name); - program_error(ctx, Program->Position, error_msg); - _mesa_free (error_msg); + program_error2(ctx, Program->Position, + "Duplicate variable declaration", + (char *) temp_var->name); return 1; } @@ -2075,12 +2094,9 @@ parse_alias (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_head, if (!found) { - char *error_msg = (char *) - _mesa_malloc (_mesa_strlen ((char *) temp_var->name) + 40); - _mesa_sprintf (error_msg, "Alias value %s is not defined", - temp_var->alias_binding->name); - program_error (ctx, Program->Position, error_msg); - _mesa_free (error_msg); + program_error2(ctx, Program->Position, + "Undefined alias value", + (char *) temp_var->alias_binding->name); return 1; } @@ -2103,12 +2119,9 @@ parse_address (GLcontext * ctx, const GLubyte ** inst, struct var_cache **vc_hea temp_var = parse_string (inst, vc_head, Program, &found); Program->Position = parse_position (inst); if (found) { - char *error_msg = (char *) - _mesa_malloc (_mesa_strlen ((char *) temp_var->name) + 40); - _mesa_sprintf (error_msg, "Duplicate Variable Declaration: %s", - temp_var->name); - program_error (ctx, Program->Position, error_msg); - _mesa_free (error_msg); + program_error2(ctx, Program->Position, + "Duplicate variable declaration", + (char *) temp_var->name); return 1; } @@ -2461,9 +2474,9 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst, Program->Position = parse_position (inst); if (!found) { - char s[1000]; - sprintf(s, "Undefined variable: %s", src->name); - program_error(ctx, Program->Position, s); + program_error2(ctx, Program->Position, + "Undefined variable", + (char *) src->name); return 1; } -- cgit v1.2.3 From a072094bf01fbdac255fd6ffa3e0100f6fc7cb6d Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:52:00 -0600 Subject: make sure state token values are fully initialized cherry-picked from master --- src/mesa/shader/arbprogparse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 38e174f3ad..2fd356efad 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -1767,7 +1767,7 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst, { GLint idx; GLuint err = 0; - gl_state_index state_tokens[STATE_LENGTH]; + gl_state_index state_tokens[STATE_LENGTH] = {0, 0, 0, 0, 0}; GLfloat const_values[4]; switch (*(*inst)++) { -- cgit v1.2.3 From 4adee848a6ce6cdf9873aa29d284db640c8ab606 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 15:54:44 -0600 Subject: Fix the compile of disabled DEBUG_PARSING code. cherry-picked from master --- src/mesa/shader/prog_parameter.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index bb964274b9..4eb7313bec 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -387,7 +387,7 @@ sizeof_state_reference(const GLint *stateTokens) * PARAM ambient = state.material.front.ambient; * * \param paramList the parameter list - * \param state an array of 6 (STATE_LENGTH) state tokens + * \param stateTokens an array of 5 (STATE_LENGTH) state tokens * \return index of the new parameter. */ GLint -- cgit v1.2.3 From d46ac956db8e4156f67b83b171569cb05d15c263 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 16:04:20 -0600 Subject: raise GL_INVALID_OPERATION if glProgramString compilation fails cherry-picked from master --- src/mesa/shader/arbprogparse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 2fd356efad..ac4da35d12 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -3983,7 +3983,7 @@ _mesa_parse_arb_vertex_program(GLcontext *ctx, GLenum target, ASSERT(target == GL_VERTEX_PROGRAM_ARB); if (!_mesa_parse_arb_program(ctx, target, (const GLubyte*) str, len, &ap)) { - /* Error in the program. Just return. */ + _mesa_error(ctx, GL_INVALID_OPERATION, "glProgramString(bad program)"); return; } -- cgit v1.2.3 From 5052dabb44a419446922d1f0b39eade5ff874134 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 16:05:40 -0600 Subject: mesa: misc sync-up with master --- src/mesa/shader/arbprogparse.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index ac4da35d12..042bddc312 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -3815,11 +3815,11 @@ _mesa_parse_arb_program(GLcontext *ctx, GLenum target, int line, col; char *s; fprintf(stderr, "program: %s\n", (char *) strz); - fprintf(stderr, "Error Pos: %d\n", ctx->program.ErrorPos); - s = (char *) _mesa_find_line_column(strz, strz+ctx->program.ErrorPos, + fprintf(stderr, "Error Pos: %d\n", ctx->Program.ErrorPos); + s = (char *) _mesa_find_line_column(strz, strz+ctx->Program.ErrorPos, &line, &col); fprintf(stderr, "line %d col %d: %s\n", line, col, s); - } while (0) + } while (0); #endif _mesa_free(strz); -- cgit v1.2.3 From bdf279b55543d0fd3e3a54e9a7a32fb6d90f5404 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 16:19:46 -0600 Subject: consolidate some parsing functions that were pretty much identical for vertex/fragment programs cherry-picked from master --- src/mesa/shader/arbprogparse.c | 223 +++++++++++++---------------------------- 1 file changed, 71 insertions(+), 152 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index 042bddc312..b60b9656c6 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -2,7 +2,7 @@ * Mesa 3-D graphics library * Version: 7.1 * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -2583,23 +2583,24 @@ parse_src_reg (GLcontext * ctx, const GLubyte ** inst, return 0; } + /** - * Parse fragment program vector source register. + * Parse vertex/fragment program vector source register. */ static GLuint -parse_fp_vector_src_reg(GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, - struct arb_program *program, - struct prog_src_register *reg) +parse_vector_src_reg(GLcontext *ctx, const GLubyte **inst, + struct var_cache **vc_head, + struct arb_program *program, + struct prog_src_register *reg) { enum register_file file; GLint index; - GLboolean negate; + GLubyte negateMask; GLubyte swizzle[4]; GLboolean isRelOffset; /* Grab the sign */ - negate = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; + negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; /* And the src reg */ if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset)) @@ -2610,66 +2611,66 @@ parse_fp_vector_src_reg(GLcontext * ctx, const GLubyte ** inst, reg->File = file; reg->Index = index; - reg->NegateBase = negate; reg->Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], swizzle[2], swizzle[3]); + reg->NegateBase = negateMask; + reg->RelAddr = isRelOffset; return 0; } /** - * Parse fragment program destination register. - * \return 1 if error, 0 if no error. + * Parse vertex/fragment program scalar source register. */ -static GLuint -parse_fp_dst_reg(GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, struct arb_program *Program, - struct prog_dst_register *reg ) +static GLuint +parse_scalar_src_reg(GLcontext *ctx, const GLubyte **inst, + struct var_cache **vc_head, + struct arb_program *program, + struct prog_src_register *reg) { - GLint mask; - GLuint idx; enum register_file file; + GLint index; + GLubyte negateMask; + GLubyte swizzle[4]; + GLboolean isRelOffset; + + /* Grab the sign */ + negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; - if (parse_masked_dst_reg (ctx, inst, vc_head, Program, &file, &idx, &mask)) + /* And the src reg */ + if (parse_src_reg(ctx, inst, vc_head, program, &file, &index, &isRelOffset)) return 1; + /* finally, the swizzle */ + parse_swizzle_mask(inst, swizzle, 1); + reg->File = file; - reg->Index = idx; - reg->WriteMask = mask; + reg->Index = index; + reg->Swizzle = (swizzle[0] << 0); + reg->NegateBase = negateMask; + reg->RelAddr = isRelOffset; return 0; } /** - * Parse fragment program scalar src register. + * Parse vertex/fragment program destination register. * \return 1 if error, 0 if no error. */ -static GLuint -parse_fp_scalar_src_reg (GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, - struct arb_program *Program, - struct prog_src_register *reg ) +static GLuint +parse_dst_reg(GLcontext * ctx, const GLubyte ** inst, + struct var_cache **vc_head, struct arb_program *program, + struct prog_dst_register *reg ) { - enum register_file File; - GLint Index; - GLubyte Negate; - GLubyte Swizzle[4]; - GLboolean IsRelOffset; - - /* Grab the sign */ - Negate = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; + GLint mask; + GLuint idx; + enum register_file file; - /* And the src reg */ - if (parse_src_reg (ctx, inst, vc_head, Program, &File, &Index, &IsRelOffset)) + if (parse_masked_dst_reg (ctx, inst, vc_head, program, &file, &idx, &mask)) return 1; - /* finally, the swizzle */ - parse_swizzle_mask(inst, Swizzle, 1); - - reg->File = File; - reg->Index = Index; - reg->NegateBase = Negate; - reg->Swizzle = (Swizzle[0] << 0); - + reg->File = file; + reg->Index = idx; + reg->WriteMask = mask; return 0; } @@ -2750,10 +2751,10 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, break; } - if (parse_fp_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) + if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) return 1; - if (parse_fp_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) + if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) return 1; break; @@ -2803,10 +2804,10 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, break; } - if (parse_fp_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) + if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) return 1; - if (parse_fp_scalar_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) + if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) return 1; break; @@ -2819,11 +2820,11 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, break; } - if (parse_fp_dst_reg(ctx, inst, vc_head, Program, &fp->DstReg)) + if (parse_dst_reg(ctx, inst, vc_head, Program, &fp->DstReg)) return 1; for (a = 0; a < 2; a++) { - if (parse_fp_scalar_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a])) + if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a])) return 1; } break; @@ -2904,10 +2905,10 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, break; } - if (parse_fp_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) + if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) return 1; for (a = 0; a < 2; a++) { - if (parse_fp_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a])) + if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a])) return 1; } break; @@ -2933,11 +2934,11 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, break; } - if (parse_fp_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) + if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) return 1; for (a = 0; a < 3; a++) { - if (parse_fp_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a])) + if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[a])) return 1; } break; @@ -2950,7 +2951,7 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, fp->Opcode = OPCODE_SWZ; break; } - if (parse_fp_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) + if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) return 1; { @@ -2993,10 +2994,10 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, break; } - if (parse_fp_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) + if (parse_dst_reg (ctx, inst, vc_head, Program, &fp->DstReg)) return 1; - if (parse_fp_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) + if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) return 1; /* texImageUnit */ @@ -3068,7 +3069,7 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, case OP_TEX_KIL: Program->UsesKill = 1; - if (parse_fp_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) + if (parse_vector_src_reg(ctx, inst, vc_head, Program, &fp->SrcReg[0])) return 1; fp->Opcode = OPCODE_KIL; break; @@ -3080,23 +3081,6 @@ parse_fp_instruction (GLcontext * ctx, const GLubyte ** inst, return 0; } -static GLuint -parse_vp_dst_reg(GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, struct arb_program *Program, - struct prog_dst_register *reg ) -{ - GLint mask; - GLuint idx; - enum register_file file; - - if (parse_masked_dst_reg(ctx, inst, vc_head, Program, &file, &idx, &mask)) - return 1; - - reg->File = file; - reg->Index = idx; - reg->WriteMask = mask; - return 0; -} /** * Handle the parsing out of a masked address register @@ -3128,71 +3112,6 @@ parse_vp_address_reg (GLcontext * ctx, const GLubyte ** inst, return 0; } -/** - * Parse vertex program vector source register. - */ -static GLuint -parse_vp_vector_src_reg(GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, - struct arb_program *program, - struct prog_src_register *reg ) -{ - enum register_file file; - GLint index; - GLubyte negateMask; - GLubyte swizzle[4]; - GLboolean isRelOffset; - - /* Grab the sign */ - negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; - - /* And the src reg */ - if (parse_src_reg (ctx, inst, vc_head, program, &file, &index, &isRelOffset)) - return 1; - - /* finally, the swizzle */ - parse_swizzle_mask(inst, swizzle, 4); - - reg->File = file; - reg->Index = index; - reg->Swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1], - swizzle[2], swizzle[3]); - reg->NegateBase = negateMask; - reg->RelAddr = isRelOffset; - return 0; -} - - -static GLuint -parse_vp_scalar_src_reg (GLcontext * ctx, const GLubyte ** inst, - struct var_cache **vc_head, - struct arb_program *Program, - struct prog_src_register *reg ) -{ - enum register_file File; - GLint Index; - GLubyte negateMask; - GLubyte Swizzle[4]; - GLboolean IsRelOffset; - - /* Grab the sign */ - negateMask = (parse_sign (inst) == -1) ? NEGATE_XYZW : NEGATE_NONE; - - /* And the src reg */ - if (parse_src_reg (ctx, inst, vc_head, Program, &File, &Index, &IsRelOffset)) - return 1; - - /* finally, the swizzle */ - parse_swizzle_mask(inst, Swizzle, 1); - - reg->File = File; - reg->Index = Index; - reg->Swizzle = (Swizzle[0] << 0); - reg->NegateBase = negateMask; - reg->RelAddr = IsRelOffset; - return 0; -} - /** * This is a big mother that handles getting opcodes into the instruction @@ -3230,7 +3149,7 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, vp->DstReg.File = PROGRAM_ADDRESS; /* Get a scalar src register */ - if (parse_vp_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0])) + if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0])) return 1; break; @@ -3254,10 +3173,10 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, break; } - if (parse_vp_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) + if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) return 1; - if (parse_vp_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0])) + if (parse_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0])) return 1; break; @@ -3282,10 +3201,10 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, vp->Opcode = OPCODE_RSQ; break; } - if (parse_vp_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) + if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) return 1; - if (parse_vp_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0])) + if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[0])) return 1; break; @@ -3295,11 +3214,11 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, vp->Opcode = OPCODE_POW; break; } - if (parse_vp_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) + if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) return 1; for (a = 0; a < 2; a++) { - if (parse_vp_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a])) + if (parse_scalar_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a])) return 1; } break; @@ -3343,11 +3262,11 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, vp->Opcode = OPCODE_XPD; break; } - if (parse_vp_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) + if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) return 1; for (a = 0; a < 2; a++) { - if (parse_vp_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a])) + if (parse_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a])) return 1; } break; @@ -3359,11 +3278,11 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, break; } - if (parse_vp_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) + if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) return 1; for (a = 0; a < 3; a++) { - if (parse_vp_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a])) + if (parse_vector_src_reg(ctx, inst, vc_head, Program, &vp->SrcReg[a])) return 1; } break; @@ -3381,7 +3300,7 @@ parse_vp_instruction (GLcontext * ctx, const GLubyte ** inst, enum register_file file; GLint index; - if (parse_vp_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) + if (parse_dst_reg(ctx, inst, vc_head, Program, &vp->DstReg)) return 1; if (parse_src_reg(ctx, inst, vc_head, Program, &file, &index, &relAddr)) -- cgit v1.2.3 From 37da2d685102ab5a706e0634fc55c60229598faa Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 27 Mar 2008 17:18:25 -0600 Subject: gallium: updated/improved comments, minor re-formatting --- src/gallium/include/pipe/p_context.h | 58 +++++++++++++++++------------------- 1 file changed, 27 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index b2e49bef4c..2d063e3f9e 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -39,31 +39,28 @@ extern "C" { struct pipe_screen; struct pipe_fence_handle; struct pipe_state_cache; - -/* Opaque driver handles: - */ struct pipe_query; + /** * Gallium rendering context. Basically: * - state setting functions * - VBO drawing functions * - surface functions - * - device queries */ struct pipe_context { struct pipe_winsys *winsys; struct pipe_screen *screen; - void *priv; /** context private data (for DRI for example) */ - void *draw; /** private, for draw module (temporary? */ + void *priv; /**< context private data (for DRI for example) */ + void *draw; /**< private, for draw module (temporary?) */ void (*destroy)( struct pipe_context * ); - /* - * Drawing. - * Return false on fallbacks (temporary??) + /** + * VBO drawing (return false on fallbacks (temporary??)) */ + /*@{*/ boolean (*draw_arrays)( struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count); @@ -71,11 +68,13 @@ struct pipe_context { struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count); + /*@}*/ /** * Query objects */ + /*@{*/ struct pipe_query *(*create_query)( struct pipe_context *pipe, unsigned query_type ); @@ -89,10 +88,12 @@ struct pipe_context { struct pipe_query *q, boolean wait, uint64 *result); + /*@}*/ - /* - * State functions + /** + * State functions (create/bind/destroy state objects) */ + /*@{*/ void * (*create_blend_state)(struct pipe_context *, const struct pipe_blend_state *); void (*bind_blend_state)(struct pipe_context *, void *); @@ -122,10 +123,12 @@ struct pipe_context { const struct pipe_shader_state *); void (*bind_vs_state)(struct pipe_context *, void *); void (*delete_vs_state)(struct pipe_context *, void *); + /*@}*/ - /* The following look more properties than states. - * maybe combine a few of them into states or pass them - * in the bind calls to the state */ + /** + * Parameter-like state (or properties) + */ + /*@{*/ void (*set_blend_color)( struct pipe_context *, const struct pipe_blend_color * ); @@ -145,19 +148,13 @@ struct pipe_context { void (*set_scissor_state)( struct pipe_context *, const struct pipe_scissor_state * ); + void (*set_viewport_state)( struct pipe_context *, + const struct pipe_viewport_state * ); - /* Currently a sampler is constrained to sample from a single texture: - */ void (*set_sampler_textures)( struct pipe_context *, - unsigned num, + unsigned num_textures, struct pipe_texture ** ); - void (*set_viewport_state)( struct pipe_context *, - const struct pipe_viewport_state * ); - - /* - * Vertex arrays - */ void (*set_vertex_buffer)( struct pipe_context *, unsigned index, const struct pipe_vertex_buffer * ); @@ -165,12 +162,13 @@ struct pipe_context { void (*set_vertex_element)( struct pipe_context *, unsigned index, const struct pipe_vertex_element * ); + /*@}*/ - /* + /** * Surface functions */ - + /*@{*/ void (*surface_copy)(struct pipe_context *pipe, unsigned do_flip, /*<< flip surface contents vertically */ struct pipe_surface *dest, @@ -189,18 +187,16 @@ struct pipe_context { void (*clear)(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue); + /*@}*/ - /** - * Called when texture data is changed. - */ + + /** Called when texture data is changed */ void (*texture_update)(struct pipe_context *pipe, struct pipe_texture *texture, uint face, uint dirtyLevelsMask); - - /* Flush rendering: - */ + /** Flush rendering (flags = bitmask of PIPE_FLUSH_x tokens) */ void (*flush)( struct pipe_context *pipe, unsigned flags, struct pipe_fence_handle **fence ); -- cgit v1.2.3 From 39038c11699bbc9baab744542e96d54e91cb452a Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 17:41:55 -0600 Subject: gallium: replace PIPE_ATTRIB_MAX with PIPE_MAX_ATTRIBS The later follows the naming scheme of other limits. Keep the old definition until all possible usage is updated. --- src/gallium/auxiliary/draw/draw_context.c | 4 ++-- src/gallium/auxiliary/draw/draw_private.h | 12 ++++++------ src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 2 +- src/gallium/auxiliary/draw/draw_vf.c | 4 ++-- src/gallium/auxiliary/draw/draw_vf.h | 2 +- src/gallium/auxiliary/draw/draw_vs_exec.c | 4 ++-- src/gallium/auxiliary/draw/draw_vs_llvm.c | 4 ++-- src/gallium/auxiliary/draw/draw_vs_sse.c | 4 ++-- src/gallium/drivers/cell/ppu/cell_context.h | 6 +++--- src/gallium/drivers/cell/ppu/cell_draw_arrays.c | 4 ++-- src/gallium/drivers/cell/ppu/cell_state_vertex.c | 4 ++-- src/gallium/drivers/cell/ppu/cell_vertex_fetch.c | 2 +- src/gallium/drivers/cell/spu/spu_main.c | 4 ++-- src/gallium/drivers/cell/spu/spu_vertex_shader.c | 4 ++-- src/gallium/drivers/cell/spu/spu_vertex_shader.h | 8 ++++---- src/gallium/drivers/failover/fo_context.h | 4 ++-- src/gallium/drivers/failover/fo_state_emit.c | 4 ++-- src/gallium/drivers/i915simple/i915_context.c | 4 ++-- src/gallium/drivers/i915simple/i915_context.h | 2 +- src/gallium/drivers/i965simple/brw_clip.h | 2 +- src/gallium/drivers/i965simple/brw_context.h | 12 ++++++------ src/gallium/drivers/i965simple/brw_state.c | 2 +- src/gallium/drivers/i965simple/brw_wm.h | 4 ++-- src/gallium/drivers/softpipe/sp_context.h | 6 +++--- src/gallium/drivers/softpipe/sp_draw_arrays.c | 4 ++-- src/gallium/drivers/softpipe/sp_quad_fs.c | 4 ++-- src/gallium/drivers/softpipe/sp_state_vertex.c | 4 ++-- src/gallium/include/pipe/p_state.h | 3 ++- src/mesa/state_tracker/st_draw.c | 2 +- 29 files changed, 63 insertions(+), 62 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 903cc26766..81858e01ca 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -238,7 +238,7 @@ draw_set_vertex_buffer(struct draw_context *draw, const struct pipe_vertex_buffer *buffer) { draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - assert(attr < PIPE_ATTRIB_MAX); + assert(attr < PIPE_MAX_ATTRIBS); draw->vertex_buffer[attr] = *buffer; } @@ -249,7 +249,7 @@ draw_set_vertex_element(struct draw_context *draw, const struct pipe_vertex_element *element) { draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - assert(attr < PIPE_ATTRIB_MAX); + assert(attr < PIPE_MAX_ATTRIBS); draw->vertex_element[attr] = *element; } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 7007ee22c4..8eb2f515cb 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -228,8 +228,8 @@ struct draw_context /* pipe state that we need: */ const struct pipe_rasterizer_state *rasterizer; struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; struct draw_vertex_shader *vertex_shader; uint num_vs_outputs; /**< convenience, from vertex_shader */ @@ -242,7 +242,7 @@ struct draw_context unsigned eltSize; /** vertex arrays */ - const void *vbuffer[PIPE_ATTRIB_MAX]; + const void *vbuffer[PIPE_MAX_ATTRIBS]; /** constant buffer (for vertex shader) */ const void *constants; @@ -275,9 +275,9 @@ struct draw_context /* Vertex fetch internal state */ struct { - const ubyte *src_ptr[PIPE_ATTRIB_MAX]; - unsigned pitch[PIPE_ATTRIB_MAX]; - fetch_func fetch[PIPE_ATTRIB_MAX]; + const ubyte *src_ptr[PIPE_MAX_ATTRIBS]; + unsigned pitch[PIPE_MAX_ATTRIBS]; + fetch_func fetch[PIPE_MAX_ATTRIBS]; unsigned nr_attrs; full_fetch_func fetch_func; pt_fetch_func pt_fetch; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 64ef83d800..9b098bc173 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -78,7 +78,7 @@ struct fetch_emit_middle_end { unsigned pitch; void (*fetch)( const void *from, float *attrib); void (*emit)( const float *attrib, float **out ); - } fetch[PIPE_ATTRIB_MAX]; + } fetch[PIPE_MAX_ATTRIBS]; unsigned nr_fetch; unsigned hw_vertex_size; diff --git a/src/gallium/auxiliary/draw/draw_vf.c b/src/gallium/auxiliary/draw/draw_vf.c index f4e29a6293..7bb34ace7a 100644 --- a/src/gallium/auxiliary/draw/draw_vf.c +++ b/src/gallium/auxiliary/draw/draw_vf.c @@ -158,7 +158,7 @@ draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, unsigned offset = 0; unsigned i, j; - assert(nr < PIPE_ATTRIB_MAX); + assert(nr < PIPE_MAX_ATTRIBS); for (j = 0, i = 0; i < nr; i++) { const unsigned format = map[i].format; @@ -390,7 +390,7 @@ struct draw_vertex_fetch *draw_vf_create( void ) struct draw_vertex_fetch *vf = CALLOC_STRUCT(draw_vertex_fetch); unsigned i; - for (i = 0; i < PIPE_ATTRIB_MAX; i++) + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) vf->attr[i].vf = vf; vf->identity[0] = 0.0; diff --git a/src/gallium/auxiliary/draw/draw_vf.h b/src/gallium/auxiliary/draw/draw_vf.h index 011c8f0ff1..7555d1bd58 100644 --- a/src/gallium/auxiliary/draw/draw_vf.h +++ b/src/gallium/auxiliary/draw/draw_vf.h @@ -169,7 +169,7 @@ struct draw_vf_attr struct draw_vertex_fetch { - struct draw_vf_attr attr[PIPE_ATTRIB_MAX]; + struct draw_vf_attr attr[PIPE_MAX_ATTRIBS]; unsigned attr_count; unsigned vertex_stride; diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 4e2fa72707..487d0ea7f4 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -99,8 +99,8 @@ vs_exec_run( struct draw_vertex_shader *shader, struct tgsi_exec_machine *machine = &draw->machine; unsigned int j; - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index bd983f2ddf..d29cb18efe 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -107,8 +107,8 @@ vs_llvm_run( struct draw_vertex_shader *base, struct tgsi_exec_machine *machine = &draw->machine; unsigned int j; - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index a4503c143e..bc910dc2d0 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -114,8 +114,8 @@ vs_sse_run( struct draw_vertex_shader *base, struct tgsi_exec_machine *machine = &draw->machine; unsigned int j; - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 0442abddc1..7f656a9744 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -103,8 +103,8 @@ struct cell_context struct cell_texture *texture[PIPE_MAX_SAMPLERS]; uint num_textures; struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS]; ubyte *zsbuf_map; @@ -141,7 +141,7 @@ struct cell_context struct spe_function attrib_fetch; - unsigned attrib_fetch_offsets[PIPE_ATTRIB_MAX]; + unsigned attrib_fetch_offsets[PIPE_MAX_ATTRIBS]; }; diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c index c839fb4d12..b896252f81 100644 --- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c @@ -123,7 +123,7 @@ cell_draw_elements(struct pipe_context *pipe, /* * Map vertex buffers */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (sp->vertex_buffer[i].buffer) { void *buf = pipe->winsys->buffer_map(pipe->winsys, sp->vertex_buffer[i].buffer, @@ -151,7 +151,7 @@ cell_draw_elements(struct pipe_context *pipe, /* * unmap vertex/index buffers - will cause draw module to flush */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (sp->vertex_buffer[i].buffer) { draw_set_mapped_vertex_buffer(draw, i, NULL); pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); diff --git a/src/gallium/drivers/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c index 563831b62d..37d25fb357 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_vertex.c +++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c @@ -41,7 +41,7 @@ cell_set_vertex_element(struct pipe_context *pipe, const struct pipe_vertex_element *attrib) { struct cell_context *cell = cell_context(pipe); - assert(index < PIPE_ATTRIB_MAX); + assert(index < PIPE_MAX_ATTRIBS); cell->vertex_element[index] = *attrib; /* struct copy */ cell->dirty |= CELL_NEW_VERTEX; @@ -55,7 +55,7 @@ cell_set_vertex_buffer(struct pipe_context *pipe, const struct pipe_vertex_buffer *buffer) { struct cell_context *cell = cell_context(pipe); - assert(index < PIPE_ATTRIB_MAX); + assert(index < PIPE_MAX_ATTRIBS); cell->vertex_buffer[index] = *buffer; /* struct copy */ cell->dirty |= CELL_NEW_VERTEX; diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c index 4828a8023b..49d5443cde 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c @@ -263,7 +263,7 @@ void cell_update_vertex_fetch(struct draw_context *draw) struct cell_context *const cell = (struct cell_context *) draw->driver_private; struct spe_function *p = &cell->attrib_fetch; - unsigned function_index[PIPE_ATTRIB_MAX]; + unsigned function_index[PIPE_MAX_ATTRIBS]; unsigned unique_attr_formats; int out_ptr; int in_ptr; diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index fccff01e10..d7f46f8024 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -55,7 +55,7 @@ struct spu_global spu; struct spu_vs_context draw; -static unsigned char attribute_fetch_code_buffer[136 * PIPE_ATTRIB_MAX] +static unsigned char attribute_fetch_code_buffer[136 * PIPE_MAX_ATTRIBS] ALIGN16_ATTRIB; static unsigned char depth_stencil_code_buffer[4 * 64] @@ -361,7 +361,7 @@ cmd_state_vs_array_info(const struct cell_array_info *vs_info) { const unsigned attr = vs_info->attr; - ASSERT(attr < PIPE_ATTRIB_MAX); + ASSERT(attr < PIPE_MAX_ATTRIBS); draw.vertex_fetch.src_ptr[attr] = vs_info->base; draw.vertex_fetch.pitch[attr] = vs_info->pitch; draw.vertex_fetch.size[attr] = vs_info->size; diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.c b/src/gallium/drivers/cell/spu/spu_vertex_shader.c index 8363efeeb6..3119a78c06 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_shader.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.c @@ -86,8 +86,8 @@ run_vertex_program(struct spu_vs_context *draw, struct spu_exec_machine *machine = &draw->machine; unsigned int j; - ALIGN16_DECL(struct spu_exec_vector, inputs, PIPE_ATTRIB_MAX); - ALIGN16_DECL(struct spu_exec_vector, outputs, PIPE_ATTRIB_MAX); + ALIGN16_DECL(struct spu_exec_vector, inputs, PIPE_MAX_ATTRIBS); + ALIGN16_DECL(struct spu_exec_vector, outputs, PIPE_MAX_ATTRIBS); const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.h b/src/gallium/drivers/cell/spu/spu_vertex_shader.h index 54a4b8d9b9..4c74f5e74d 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_shader.h +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.h @@ -16,10 +16,10 @@ struct spu_vs_context { struct pipe_viewport_state viewport; struct { - uint64_t src_ptr[PIPE_ATTRIB_MAX]; - unsigned pitch[PIPE_ATTRIB_MAX]; - unsigned size[PIPE_ATTRIB_MAX]; - unsigned code_offset[PIPE_ATTRIB_MAX]; + uint64_t src_ptr[PIPE_MAX_ATTRIBS]; + unsigned pitch[PIPE_MAX_ATTRIBS]; + unsigned size[PIPE_MAX_ATTRIBS]; + unsigned code_offset[PIPE_MAX_ATTRIBS]; unsigned nr_attrs; boolean dirty; diff --git a/src/gallium/drivers/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h index 8f3ad3ee79..4afe10c4b8 100644 --- a/src/gallium/drivers/failover/fo_context.h +++ b/src/gallium/drivers/failover/fo_context.h @@ -84,8 +84,8 @@ struct failover_context { struct pipe_scissor_state scissor; struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; void *sw_sampler_state[PIPE_MAX_SAMPLERS]; void *hw_sampler_state[PIPE_MAX_SAMPLERS]; diff --git a/src/gallium/drivers/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c index 3de931e04e..bb89f925e9 100644 --- a/src/gallium/drivers/failover/fo_state_emit.c +++ b/src/gallium/drivers/failover/fo_state_emit.c @@ -104,7 +104,7 @@ failover_state_emit( struct failover_context *failover ) } if (failover->dirty & FO_NEW_VERTEX_BUFFER) { - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (failover->dirty_vertex_buffer & (1<sw->set_vertex_buffer( failover->sw, i, &failover->vertex_buffer[i] ); @@ -113,7 +113,7 @@ failover_state_emit( struct failover_context *failover ) } if (failover->dirty & FO_NEW_VERTEX_ELEMENT) { - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (failover->dirty_vertex_element & (1<sw->set_vertex_element( failover->sw, i, &failover->vertex_element[i] ); diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 15ff2360b7..fee33d82de 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -65,7 +65,7 @@ i915_draw_elements( struct pipe_context *pipe, /* * Map vertex buffers */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (i915->vertex_buffer[i].buffer) { void *buf = pipe->winsys->buffer_map(pipe->winsys, @@ -96,7 +96,7 @@ i915_draw_elements( struct pipe_context *pipe, /* * unmap vertex/index buffers */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (i915->vertex_buffer[i].buffer) { pipe->winsys->buffer_unmap(pipe->winsys, i915->vertex_buffer[i].buffer); draw_set_mapped_vertex_buffer(draw, i, NULL); diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 746f18ba38..8e707ea574 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -232,7 +232,7 @@ struct i915_context struct pipe_scissor_state scissor; struct i915_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; unsigned dirty; diff --git a/src/gallium/drivers/i965simple/brw_clip.h b/src/gallium/drivers/i965simple/brw_clip.h index a89d08b791..d70fc094ff 100644 --- a/src/gallium/drivers/i965simple/brw_clip.h +++ b/src/gallium/drivers/i965simple/brw_clip.h @@ -116,7 +116,7 @@ struct brw_clip_compile { unsigned last_mrf; unsigned header_position_offset; - unsigned offset[PIPE_ATTRIB_MAX]; + unsigned offset[PIPE_MAX_ATTRIBS]; }; #define ATTR_SIZE (4*4) diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index b83a13c3b6..0c96ba1732 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -433,17 +433,17 @@ struct brw_cached_batch_item { -/* Protect against a future where PIPE_ATTRIB_MAX > 32. Wouldn't life +/* Protect against a future where PIPE_MAX_ATTRIBS > 32. Wouldn't life * be easier if C allowed arrays of packed elements? */ -#define ATTRIB_BIT_DWORDS ((PIPE_ATTRIB_MAX+31)/32) +#define ATTRIB_BIT_DWORDS ((PIPE_MAX_ATTRIBS+31)/32) struct brw_vertex_info { - unsigned varying; /* varying:1[PIPE_ATTRIB_MAX] */ - unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_ATTRIB_MAX] */ + unsigned varying; /* varying:1[PIPE_MAX_ATTRIBS] */ + unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_MAX_ATTRIBS] */ }; @@ -496,9 +496,9 @@ struct brw_context /* Arrays with buffer objects to copy non-bufferobj arrays into * for upload: */ - const struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX]; + const struct pipe_vertex_buffer *vbo_array[PIPE_MAX_ATTRIBS]; - struct brw_vertex_element_state inputs[PIPE_ATTRIB_MAX]; + struct brw_vertex_element_state inputs[PIPE_MAX_ATTRIBS]; #define BRW_NR_UPLOAD_BUFS 17 #define BRW_UPLOAD_INIT_SIZE (128*1024) diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index f5efe9fc06..0d04a8a594 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -292,7 +292,7 @@ static void brw_set_vertex_element(struct pipe_context *pipe, /* flush ? */ struct brw_context *brw = brw_context(pipe); - assert(index < PIPE_ATTRIB_MAX); + assert(index < PIPE_MAX_ATTRIBS); struct brw_vertex_element_state el; memset(&el, 0, sizeof(el)); diff --git a/src/gallium/drivers/i965simple/brw_wm.h b/src/gallium/drivers/i965simple/brw_wm.h index a1ac0f504a..b29c4393f0 100644 --- a/src/gallium/drivers/i965simple/brw_wm.h +++ b/src/gallium/drivers/i965simple/brw_wm.h @@ -76,7 +76,7 @@ struct brw_wm_prog_key { #define PROGRAM_INTERNAL_PARAM #define MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS 1024 /* 72 for GL_ARB_f_p */ -#define BRW_WM_MAX_INSN (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + PIPE_ATTRIB_MAX + 3) +#define BRW_WM_MAX_INSN (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + PIPE_MAX_ATTRIBS + 3) #define BRW_WM_MAX_GRF 128 /* hardware limit */ #define BRW_WM_MAX_VREG (BRW_WM_MAX_INSN * 4) #define BRW_WM_MAX_REF (BRW_WM_MAX_INSN * 12) @@ -84,7 +84,7 @@ struct brw_wm_prog_key { #define BRW_WM_MAX_CONST 256 #define BRW_WM_MAX_KILLS MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS -#define PAYLOAD_DEPTH (PIPE_ATTRIB_MAX) +#define PAYLOAD_DEPTH (PIPE_MAX_ATTRIBS) #define MAX_IFSN 32 #define MAX_LOOP_DEPTH 32 diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 19e6cfaf02..dc9d0e6d5d 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -77,8 +77,8 @@ struct softpipe_context { struct pipe_scissor_state scissor; struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX]; - struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX]; + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; unsigned dirty; unsigned num_samplers; @@ -92,7 +92,7 @@ struct softpipe_context { /* * Mapped vertex buffers */ - ubyte *mapped_vbuffer[PIPE_ATTRIB_MAX]; + ubyte *mapped_vbuffer[PIPE_MAX_ATTRIBS]; /** Mapped constant buffers */ void *mapped_constants[PIPE_SHADER_TYPES]; diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 5b5a0fe573..ab54050d3f 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -125,7 +125,7 @@ softpipe_draw_elements(struct pipe_context *pipe, /* * Map vertex buffers */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (sp->vertex_buffer[i].buffer) { void *buf = pipe->winsys->buffer_map(pipe->winsys, @@ -153,7 +153,7 @@ softpipe_draw_elements(struct pipe_context *pipe, /* * unmap vertex/index buffers - will cause draw module to flush */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (sp->vertex_buffer[i].buffer) { draw_set_mapped_vertex_buffer(draw, i, NULL); pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 861285101f..c10ad80e01 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -185,8 +185,8 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe ) uint i; /* allocate storage for program inputs/outputs, aligned to 16 bytes */ - qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16); - qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16); + qss->inputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->inputs) + 16); + qss->outputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->outputs) + 16); qss->machine.Inputs = align16(qss->inputs); qss->machine.Outputs = align16(qss->outputs); diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c index f01a10de3b..c054e76d9b 100644 --- a/src/gallium/drivers/softpipe/sp_state_vertex.c +++ b/src/gallium/drivers/softpipe/sp_state_vertex.c @@ -42,7 +42,7 @@ softpipe_set_vertex_element(struct pipe_context *pipe, const struct pipe_vertex_element *attrib) { struct softpipe_context *softpipe = softpipe_context(pipe); - assert(index < PIPE_ATTRIB_MAX); + assert(index < PIPE_MAX_ATTRIBS); softpipe->vertex_element[index] = *attrib; /* struct copy */ softpipe->dirty |= SP_NEW_VERTEX; @@ -56,7 +56,7 @@ softpipe_set_vertex_buffer(struct pipe_context *pipe, const struct pipe_vertex_buffer *buffer) { struct softpipe_context *softpipe = softpipe_context(pipe); - assert(index < PIPE_ATTRIB_MAX); + assert(index < PIPE_MAX_ATTRIBS); softpipe->vertex_buffer[index] = *buffer; /* struct copy */ softpipe->dirty |= SP_NEW_VERTEX; diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index a2bd8c6aaa..2490412126 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -54,7 +54,8 @@ extern "C" { #define PIPE_MAX_SAMPLERS 8 #define PIPE_MAX_CLIP_PLANES 6 #define PIPE_MAX_CONSTANT 32 -#define PIPE_ATTRIB_MAX 32 +#define PIPE_MAX_ATTRIBS 32 +#define PIPE_ATTRIB_MAX 32 /* XXX obsolete - remove */ #define PIPE_MAX_COLOR_BUFS 8 #define PIPE_MAX_TEXTURE_LEVELS 16 #define PIPE_MAX_FEEDBACK_ATTRIBS 16 diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 20af90df7d..4aca3311b7 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -606,7 +606,7 @@ st_feedback_draw_vbo(GLcontext *ctx, /* * unmap vertex/index buffers */ - for (i = 0; i < PIPE_ATTRIB_MAX; i++) { + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (draw->vertex_buffer[i].buffer) { pipe->winsys->buffer_unmap(pipe->winsys, draw->vertex_buffer[i].buffer); -- cgit v1.2.3 From 5935b16ff5d8cbe07c97f2990e281a9c5986d0a1 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 27 Mar 2008 21:11:45 -0600 Subject: If visual is 32bpp, set visual->alphaBits = 8. This works since we always use XImages for color buffers. It wouldn't always work for Windows/Pixmaps. --- src/gallium/winsys/xlib/xm_api.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c index 6b46327e21..a82d3c990e 100644 --- a/src/gallium/winsys/xlib/xm_api.c +++ b/src/gallium/winsys/xlib/xm_api.c @@ -511,6 +511,14 @@ initialize_visual_and_buffer(XMesaVisual v, XMesaBuffer b, return GL_FALSE; } v->mesa_visual.indexBits = 0; + + if (v->BitsPerPixel == 32) { + /* We use XImages for all front/back buffers. If an X Window or + * X Pixmap is 32bpp, there's no guarantee that the alpha channel + * will be preserved. For XImages we're in luck. + */ + v->mesa_visual.alphaBits = 8; + } } /* -- cgit v1.2.3 From 4e7bcaa4c20cc358206cf4668316f998d79c85b0 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 28 Mar 2008 12:31:35 +0100 Subject: gallium: Bump PIPE_MAX_SAMPLERS to 16. We need it to fulfil D3D minimum requirements. --- src/gallium/include/pipe/p_state.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 2490412126..b47595deca 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -51,7 +51,7 @@ extern "C" { /** * Implementation limits */ -#define PIPE_MAX_SAMPLERS 8 +#define PIPE_MAX_SAMPLERS 16 #define PIPE_MAX_CLIP_PLANES 6 #define PIPE_MAX_CONSTANT 32 #define PIPE_MAX_ATTRIBS 32 -- cgit v1.2.3 From 5615ab78b03cf1cb5fb19fc04fef52818f91b0be Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 28 Mar 2008 15:43:00 +0100 Subject: gallium: remove redundant compare bit in sampler state --- src/gallium/drivers/i965simple/brw_wm.c | 3 +-- src/gallium/include/pipe/p_state.h | 1 - src/mesa/state_tracker/st_atom_sampler.c | 1 - 3 files changed, 1 insertion(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_wm.c b/src/gallium/drivers/i965simple/brw_wm.c index 1c4b5b5ede..7fc5f59a98 100644 --- a/src/gallium/drivers/i965simple/brw_wm.c +++ b/src/gallium/drivers/i965simple/brw_wm.c @@ -161,8 +161,7 @@ static void brw_wm_populate_key( struct brw_context *brw, if (unit) { - if (unit->compare && - unit->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + if (unit->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { key->shadowtex_mask |= 1<Image[0][t->BaseLevel]->InternalFormat == GL_YCBCR_MESA) diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index b47595deca..0eeee47a9a 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -243,7 +243,6 @@ struct pipe_sampler_state unsigned min_img_filter:2; /**< PIPE_TEX_FILTER_x */ unsigned min_mip_filter:2; /**< PIPE_TEX_MIPFILTER_x */ unsigned mag_img_filter:2; /**< PIPE_TEX_FILTER_x */ - unsigned compare:1; /**< shadow/depth compare enabled? */ unsigned compare_mode:1; /**< PIPE_TEX_COMPARE_x */ unsigned compare_func:3; /**< PIPE_FUNC_x */ unsigned normalized_coords:1; /**< Are coords normalized to [0,1]? */ diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 5787a7492c..5dd242ac66 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -167,7 +167,6 @@ update_samplers(struct st_context *st) /* only care about ARB_shadow, not SGI shadow */ if (texobj->CompareMode == GL_COMPARE_R_TO_TEXTURE) { - sampler->compare = 1; sampler->compare_mode = PIPE_TEX_COMPARE_R_TO_TEXTURE; sampler->compare_func = st_compare_func_to_pipe(texobj->CompareFunc); -- cgit v1.2.3 From de3634fca8f93223647a54f49f066494447906bb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 28 Mar 2008 13:09:19 -0600 Subject: gallium: re-work texture format selection code Use same code for choosing texture format and renderbuffer format. --- src/mesa/state_tracker/st_format.c | 260 +++++++++++++------------------------ 1 file changed, 93 insertions(+), 167 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 9aeda65a5c..cb6fa9a573 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -271,6 +271,10 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) return PIPE_FORMAT_U_I8; case MESA_FORMAT_Z16: return PIPE_FORMAT_Z16_UNORM; + case MESA_FORMAT_Z32: + return PIPE_FORMAT_Z32_UNORM; + case MESA_FORMAT_Z24_S8: + return PIPE_FORMAT_Z24S8_UNORM; default: assert(0); return 0; @@ -281,7 +285,7 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) * Find an RGBA format supported by the context/winsys. */ static GLuint -default_rgba_format(struct pipe_context *pipe, uint type) +default_rgba_format(struct pipe_screen *screen, uint type) { static const enum pipe_format colorFormats[] = { PIPE_FORMAT_A8R8G8B8_UNORM, @@ -289,7 +293,6 @@ default_rgba_format(struct pipe_context *pipe, uint type) PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_R5G6B5_UNORM }; - struct pipe_screen *screen = pipe->screen; uint i; for (i = 0; i < Elements(colorFormats); i++) { if (screen->is_format_supported( screen, colorFormats[i], type )) { @@ -304,9 +307,8 @@ default_rgba_format(struct pipe_context *pipe, uint type) * Search list of formats for first RGBA format with >8 bits/channel. */ static GLuint -default_deep_rgba_format(struct pipe_context *pipe, uint type) +default_deep_rgba_format(struct pipe_screen *screen, uint type) { - struct pipe_screen *screen = pipe->screen; if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SNORM, type)) { return PIPE_FORMAT_R16G16B16A16_SNORM; } @@ -318,7 +320,7 @@ default_deep_rgba_format(struct pipe_context *pipe, uint type) * Find an Z format supported by the context/winsys. */ static GLuint -default_depth_format(struct pipe_context *pipe, uint type) +default_depth_format(struct pipe_screen *screen, uint type) { static const enum pipe_format zFormats[] = { PIPE_FORMAT_Z16_UNORM, @@ -326,7 +328,6 @@ default_depth_format(struct pipe_context *pipe, uint type) PIPE_FORMAT_S8Z24_UNORM, PIPE_FORMAT_Z24S8_UNORM }; - struct pipe_screen *screen = pipe->screen; uint i; for (i = 0; i < Elements(zFormats); i++) { if (screen->is_format_supported( screen, zFormats[i], type )) { @@ -338,15 +339,16 @@ default_depth_format(struct pipe_context *pipe, uint type) /** - * Choose the PIPE_FORMAT_ to use for user-created renderbuffers. - * - * \return PIPE_FORMAT_NONE if error/problem. + * Given an OpenGL internalFormat value for a texture or surface, return + * the best matching PIPE_FORMAT_x, or PIPE_FORMAT_NONE if there's no match. */ -enum pipe_format -st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) +static enum pipe_format +choose_format(struct pipe_context *pipe, GLint internalFormat, uint surfType) { struct pipe_screen *screen = pipe->screen; - uint surfType = PIPE_SURFACE; + + assert(surfType == PIPE_SURFACE || + surfType == PIPE_TEXTURE); switch (internalFormat) { case 4: @@ -358,26 +360,26 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: - return default_rgba_format( pipe, surfType ); + return default_rgba_format( screen, surfType ); case GL_RGBA16: - return default_deep_rgba_format( pipe, surfType ); + return default_deep_rgba_format( screen, surfType ); case GL_RGBA4: case GL_RGBA2: if (screen->is_format_supported( screen, PIPE_FORMAT_A4R4G4B4_UNORM, surfType )) return PIPE_FORMAT_A4R4G4B4_UNORM; - return default_rgba_format( pipe, surfType ); + return default_rgba_format( screen, surfType ); case GL_RGB5_A1: if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) return PIPE_FORMAT_A1R5G5B5_UNORM; - return default_rgba_format( pipe, surfType ); + return default_rgba_format( screen, surfType ); case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return default_rgba_format( pipe, surfType ); + return default_rgba_format( screen, surfType ); case GL_RGB5: case GL_RGB4: @@ -386,7 +388,7 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) return PIPE_FORMAT_A1R5G5B5_UNORM; if (screen->is_format_supported( screen, PIPE_FORMAT_R5G6B5_UNORM, surfType )) return PIPE_FORMAT_R5G6B5_UNORM; - return default_rgba_format( pipe, surfType ); + return default_rgba_format( screen, surfType ); case GL_ALPHA: case GL_ALPHA4: @@ -396,7 +398,7 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_COMPRESSED_ALPHA: if (screen->is_format_supported( screen, PIPE_FORMAT_U_A8, surfType )) return PIPE_FORMAT_U_A8; - return default_rgba_format( pipe, surfType ); + return default_rgba_format( screen, surfType ); case 1: case GL_LUMINANCE: @@ -406,8 +408,8 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: if (screen->is_format_supported( screen, PIPE_FORMAT_U_L8, surfType )) - return PIPE_FORMAT_U_A8; - return default_rgba_format( pipe, surfType ); + return PIPE_FORMAT_U_L8; + return default_rgba_format( screen, surfType ); case 2: case GL_LUMINANCE_ALPHA: @@ -420,7 +422,7 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_COMPRESSED_LUMINANCE_ALPHA: if (screen->is_format_supported( screen, PIPE_FORMAT_U_A8_L8, surfType )) return PIPE_FORMAT_U_A8_L8; - return default_rgba_format( pipe, surfType ); + return default_rgba_format( screen, surfType ); case GL_INTENSITY: case GL_INTENSITY4: @@ -430,32 +432,32 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) case GL_COMPRESSED_INTENSITY: if (screen->is_format_supported( screen, PIPE_FORMAT_U_I8, surfType )) return PIPE_FORMAT_U_I8; - return default_rgba_format( pipe, surfType ); + return default_rgba_format( screen, surfType ); -#if 0 - /* not supported for renderbuffers */ case GL_YCBCR_MESA: return PIPE_FORMAT_NONE; - case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; case GL_RGB_S3TC: case GL_RGB4_S3TC: case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: - return &_mesa_texformat_rgb_dxt1; + return PIPE_FORMAT_DXT1_RGB; case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: - return &_mesa_texformat_rgba_dxt1; + return PIPE_FORMAT_DXT1_RGBA; case GL_RGBA_S3TC: case GL_RGBA4_S3TC: case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: - return &_mesa_texformat_rgba_dxt3; + return PIPE_FORMAT_DXT3_RGBA; case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: - return &_mesa_texformat_rgba_dxt5; + return PIPE_FORMAT_DXT5_RGBA; + +#if 0 + case GL_COMPRESSED_RGB_FXT1_3DFX: + return PIPE_FORMAT_RGB_FXT1; + case GL_COMPRESSED_RGBA_FXT1_3DFX: + return PIPE_FORMAT_RGB_FXT1; #endif case GL_DEPTH_COMPONENT16: @@ -473,7 +475,7 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) return PIPE_FORMAT_Z32_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT: - return default_depth_format( pipe, surfType ); + return default_depth_format( screen, surfType ); case GL_STENCIL_INDEX: case GL_STENCIL_INDEX1_EXT: @@ -502,155 +504,79 @@ st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) } - -/* It works out that this function is fine for all the supported - * hardware. However, there is still a need to map the formats onto - * hardware descriptors. - */ -/* Note that the i915 can actually support many more formats than - * these if we take the step of simply swizzling the colors - * immediately after sampling... +/** + * Called by FBO code to choose a PIPE_FORMAT_ for drawing surfaces. */ -const struct gl_texture_format * -st_ChooseTextureFormat(GLcontext * ctx, GLint internalFormat, - GLenum format, GLenum type) +enum pipe_format +st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) { -#if 0 - struct intel_context *intel = intel_context(ctx); - const GLboolean do32bpt = (intel->intelScreen->front.cpp == 4); -#else - const GLboolean do32bpt = 1; -#endif - - (void) ctx; - - switch (internalFormat) { - case 4: - case GL_RGBA: - case GL_COMPRESSED_RGBA: - if (format == GL_BGRA) { - if (type == GL_UNSIGNED_BYTE || type == GL_UNSIGNED_INT_8_8_8_8_REV) { - return &_mesa_texformat_argb8888; - } - else if (type == GL_UNSIGNED_SHORT_4_4_4_4_REV) { - return &_mesa_texformat_argb4444; - } - else if (type == GL_UNSIGNED_SHORT_1_5_5_5_REV) { - return &_mesa_texformat_argb1555; - } - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; - - case 3: - case GL_RGB: - case GL_COMPRESSED_RGB: - if (format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5) { - return &_mesa_texformat_rgb565; - } - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565; - - case GL_RGBA8: - case GL_RGB10_A2: - case GL_RGBA12: - case GL_RGBA16: - return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444; + return choose_format(pipe, internalFormat, PIPE_SURFACE); +} - case GL_RGBA4: - case GL_RGBA2: - return &_mesa_texformat_argb4444; - case GL_RGB5_A1: - return &_mesa_texformat_argb1555; - - case GL_RGB8: - case GL_RGB10: - case GL_RGB12: - case GL_RGB16: +static const struct gl_texture_format * +translate_gallium_format_to_mesa_format(enum pipe_format format) +{ + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: return &_mesa_texformat_argb8888; - - case GL_RGB5: - case GL_RGB4: - case GL_R3_G3_B2: + case PIPE_FORMAT_A1R5G5B5_UNORM: + return &_mesa_texformat_argb1555; + case PIPE_FORMAT_A4R4G4B4_UNORM: + return &_mesa_texformat_argb4444; + case PIPE_FORMAT_R5G6B5_UNORM: return &_mesa_texformat_rgb565; - - case GL_ALPHA: - case GL_ALPHA4: - case GL_ALPHA8: - case GL_ALPHA12: - case GL_ALPHA16: - case GL_COMPRESSED_ALPHA: + case PIPE_FORMAT_U_A8_L8: + return &_mesa_texformat_al88; + case PIPE_FORMAT_U_A8: return &_mesa_texformat_a8; - - case 1: - case GL_LUMINANCE: - case GL_LUMINANCE4: - case GL_LUMINANCE8: - case GL_LUMINANCE12: - case GL_LUMINANCE16: - case GL_COMPRESSED_LUMINANCE: + case PIPE_FORMAT_U_L8: return &_mesa_texformat_l8; - - case 2: - case GL_LUMINANCE_ALPHA: - case GL_LUMINANCE4_ALPHA4: - case GL_LUMINANCE6_ALPHA2: - case GL_LUMINANCE8_ALPHA8: - case GL_LUMINANCE12_ALPHA4: - case GL_LUMINANCE12_ALPHA12: - case GL_LUMINANCE16_ALPHA16: - case GL_COMPRESSED_LUMINANCE_ALPHA: - return &_mesa_texformat_al88; - - case GL_INTENSITY: - case GL_INTENSITY4: - case GL_INTENSITY8: - case GL_INTENSITY12: - case GL_INTENSITY16: - case GL_COMPRESSED_INTENSITY: + case PIPE_FORMAT_U_I8: return &_mesa_texformat_i8; - - case GL_YCBCR_MESA: - if (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE) - return &_mesa_texformat_ycbcr; - else - return &_mesa_texformat_ycbcr_rev; - - case GL_COMPRESSED_RGB_FXT1_3DFX: - return &_mesa_texformat_rgb_fxt1; - case GL_COMPRESSED_RGBA_FXT1_3DFX: - return &_mesa_texformat_rgba_fxt1; - - case GL_RGB_S3TC: - case GL_RGB4_S3TC: - case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: + case PIPE_FORMAT_Z16_UNORM: + return &_mesa_texformat_z16; + case PIPE_FORMAT_Z32_UNORM: + return &_mesa_texformat_z32; + case PIPE_FORMAT_S8Z24_UNORM: + /* XXX fallthrough OK? */ + case PIPE_FORMAT_Z24S8_UNORM: + return &_mesa_texformat_z24_s8; + case PIPE_FORMAT_YCBCR: + return &_mesa_texformat_ycbcr; + case PIPE_FORMAT_YCBCR_REV: + return &_mesa_texformat_ycbcr_rev; + case PIPE_FORMAT_DXT1_RGB: return &_mesa_texformat_rgb_dxt1; - - case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT: + case PIPE_FORMAT_DXT1_RGBA: return &_mesa_texformat_rgba_dxt1; - - case GL_RGBA_S3TC: - case GL_RGBA4_S3TC: - case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT: + case PIPE_FORMAT_DXT3_RGBA: return &_mesa_texformat_rgba_dxt3; - - case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT: + case PIPE_FORMAT_DXT5_RGBA: return &_mesa_texformat_rgba_dxt5; + /* XXX add additional cases */ + default: + assert(0); + return NULL; + } +} - case GL_DEPTH_COMPONENT: - case GL_DEPTH_COMPONENT16: - case GL_DEPTH_COMPONENT24: - case GL_DEPTH_COMPONENT32: - return &_mesa_texformat_z16; - case GL_DEPTH_STENCIL_EXT: - case GL_DEPTH24_STENCIL8_EXT: - return &_mesa_texformat_z24_s8; +/** + * Called via ctx->Driver.chooseTextureFormat(). + */ +const struct gl_texture_format * +st_ChooseTextureFormat(GLcontext *ctx, GLint internalFormat, + GLenum format, GLenum type) +{ + enum pipe_format pFormat; - default: - fprintf(stderr, "unexpected texture format %s in %s\n", - _mesa_lookup_enum_by_nr(internalFormat), __FUNCTION__); + (void) format; + (void) type; + + pFormat = choose_format(ctx->st->pipe, internalFormat, PIPE_TEXTURE); + if (pFormat == PIPE_FORMAT_NONE) return NULL; - } - return NULL; /* never get here */ + return translate_gallium_format_to_mesa_format(pFormat); } -- cgit v1.2.3 From 13041da714106ae61b4184b79e847c2b382e07ad Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 28 Mar 2008 13:10:16 -0600 Subject: mesa: fix texture/renderbuffer mix-up in test_attachment_completeness() --- src/mesa/main/fbobject.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 6a8cba4d8a..800f6ee9a3 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -307,7 +307,7 @@ test_attachment_completeness(const GLcontext *ctx, GLenum format, /* OK */ } else if (ctx->Extensions.EXT_packed_depth_stencil && - att->Renderbuffer->_BaseFormat == GL_DEPTH_STENCIL_EXT) { + texImage->TexFormat->BaseFormat == GL_DEPTH_STENCIL_EXT) { /* OK */ } else { -- cgit v1.2.3 From c62b197b528293abb56b099503344e3cdd7d6c40 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Mar 2008 14:53:47 -0600 Subject: gallium: implement a glBitmap cache The bitmap cache attempts to accumulate a series of glBitmap calls in a buffer to effectively render a whole bunch of bitmaps at once. The cache can be disabled, if needed, by setting UseBitmapCache=GL_FALSE. --- src/mesa/state_tracker/st_atom.c | 3 + src/mesa/state_tracker/st_cb_bitmap.c | 234 ++++++++++++++++++++++++++++++++-- src/mesa/state_tracker/st_cb_bitmap.h | 5 + src/mesa/state_tracker/st_cb_flush.c | 3 + src/mesa/state_tracker/st_context.c | 1 + src/mesa/state_tracker/st_context.h | 2 + 6 files changed, 236 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 0e22a2fa6e..40e4142631 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -32,6 +32,7 @@ #include "pipe/p_defines.h" #include "st_context.h" #include "st_atom.h" +#include "st_cb_bitmap.h" #include "st_program.h" @@ -147,6 +148,8 @@ void st_validate_state( struct st_context *st ) struct st_state_flags *state = &st->dirty; GLuint i; + st_flush_bitmap_cache(st); + check_program_state( st ); if (state->st == 0) diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 64b1882424..e46dcbf661 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -59,6 +59,31 @@ +/** + * The bitmap cache attempts to accumulate multiple glBitmap calls in a + * buffer which is then rendered en mass upon a flush, state change, etc. + * A wide, short buffer is used to target the common case of a series + * of glBitmap calls being used to draw text. + */ +static GLboolean UseBitmapCache = 0*GL_TRUE; + + +#define BITMAP_CACHE_WIDTH 512 +#define BITMAP_CACHE_HEIGHT 32 + +struct bitmap_cache +{ + /** An I8 texture image: */ + GLubyte buffer[BITMAP_CACHE_HEIGHT][BITMAP_CACHE_WIDTH]; + GLboolean empty; + /** Window pos to render the cached image */ + GLint xpos, ypos; + struct pipe_texture *texture; +}; + + + + /** * Make fragment program for glBitmap: * Sample the texture and kill the fragment if the bit is 0. @@ -390,16 +415,18 @@ setup_bitmap_vertex_data(struct st_context *st, static void draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, GLsizei width, GLsizei height, - struct pipe_texture *pt, - struct st_fragment_program *stfp) + struct pipe_texture *pt) { struct st_context *st = ctx->st; struct pipe_context *pipe = ctx->st->pipe; struct cso_context *cso = ctx->st->cso_context; + struct st_fragment_program *stfp; GLuint maxSize; + stfp = combined_bitmap_fragment_program(ctx); + /* limit checks */ - /* XXX if DrawPixels image is larger than max texture size, break + /* XXX if the bitmap is larger than the max texture size, break * it up into chunks. */ maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1); @@ -467,18 +494,185 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, +static void +init_bitmap_cache(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; + enum pipe_format format; + + st->bitmap.cache = CALLOC_STRUCT(bitmap_cache); + if (!st->bitmap.cache) + return; + + /* find a usable texture format */ + if (screen->is_format_supported(screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE)) { + format = PIPE_FORMAT_U_I8; + } + else { + /* XXX support more formats */ + assert(0); + } + + st->bitmap.cache->texture + = st_texture_create(st, PIPE_TEXTURE_2D, format, 0, + BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, 1, 0); + if (!st->bitmap.cache->texture) { + FREE(st->bitmap.cache); + st->bitmap.cache = NULL; + return; + } + + st->bitmap.cache->empty = GL_TRUE; +} + + +/** + * If there's anything in the bitmap cache, draw/flush it now. + */ +void +st_flush_bitmap_cache(struct st_context *st) +{ + if (!st->bitmap.cache->empty) { + struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *surf; + void *dest; + + /* update the texture map image */ + surf = screen->get_tex_surface(screen, st->bitmap.cache->texture, 0, 0, 0); + dest = pipe_surface_map(surf); + memcpy(dest, st->bitmap.cache->buffer, sizeof(st->bitmap.cache->buffer)); + pipe_surface_unmap(surf); + pipe_surface_reference(&surf, NULL); + + pipe->texture_update(pipe, st->bitmap.cache->texture, 0, 0x1); + + draw_bitmap_quad(st->ctx, + st->bitmap.cache->xpos, + st->bitmap.cache->ypos, + st->ctx->Current.RasterPos[2], + BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, + st->bitmap.cache->texture); + + memset(st->bitmap.cache->buffer, 0, sizeof(st->bitmap.cache->buffer)); + st->bitmap.cache->empty = GL_TRUE; + } +} + + +/** + * Try to accumulate this glBitmap call in the bitmap cache. + * \return GL_TRUE for success, GL_FALSE if bitmap is too large, etc. + */ +static GLboolean +accum_bitmap(struct st_context *st, + GLint x, GLint y, GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap ) +{ + int row, col; + int px = -999, py; + + if (width > BITMAP_CACHE_WIDTH || + height > BITMAP_CACHE_HEIGHT) + return GL_FALSE; /* too big to cache */ + + if (!st->bitmap.cache->empty) { + px = x - st->bitmap.cache->xpos; /* pos in buffer */ + py = y - st->bitmap.cache->ypos; + if (px < 0 || px + width > BITMAP_CACHE_WIDTH || + py < 0 || py + height > BITMAP_CACHE_HEIGHT) { + /* This bitmap would extend beyond cache bounds, + * so flush and continue. + */ + st_flush_bitmap_cache(st); + } + } + + if (st->bitmap.cache->empty) { + /* Initialize. Center bitmap vertically in the buffer. */ + px = 0; + py = (BITMAP_CACHE_HEIGHT - height) / 2; + st->bitmap.cache->xpos = x; + st->bitmap.cache->ypos = y - py; + st->bitmap.cache->empty = GL_FALSE; + } + + assert(px != -999); + + /* XXX try to combine this code with code in make_bitmap_texture() */ +#define SET_PIXEL(COL, ROW) \ + st->bitmap.cache->buffer[py + (ROW)][px + (COL)] = 0xff; + + for (row = 0; row < height; row++) { + const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, + bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); + + if (unpack->LsbFirst) { + /* Lsb first */ + GLubyte mask = 1U << (unpack->SkipPixels & 0x7); + for (col = 0; col < width; col++) { + + if (*src & mask) { + SET_PIXEL(col, row); + } + + if (mask == 128U) { + src++; + mask = 1U; + } + else { + mask = mask << 1; + } + } + + /* get ready for next row */ + if (mask != 1) + src++; + } + else { + /* Msb first */ + GLubyte mask = 128U >> (unpack->SkipPixels & 0x7); + for (col = 0; col < width; col++) { + + if (*src & mask) { + SET_PIXEL(col, row); + } + + if (mask == 1U) { + src++; + mask = 128U; + } + else { + mask = mask >> 1; + } + } + + /* get ready for next row */ + if (mask != 128) + src++; + } + + } /* row */ + + return GL_TRUE; /* accumulated */ +} + + + +/** + * Called via ctx->Driver.Bitmap() + */ static void st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap ) { - struct st_fragment_program *stfp; struct st_context *st = ctx->st; struct pipe_texture *pt; st_validate_state(st); - stfp = combined_bitmap_fragment_program(ctx); - if (!st->bitmap.vs) { /* create pass-through vertex shader now */ const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, @@ -491,26 +685,36 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, &st->bitmap.vert_shader); } - st_validate_state(st); + if (UseBitmapCache && accum_bitmap(st, x, y, width, height, unpack, bitmap)) + return; pt = make_bitmap_texture(ctx, width, height, unpack, bitmap); if (pt) { assert(pt->target == PIPE_TEXTURE_2D); draw_bitmap_quad(ctx, x, y, ctx->Current.RasterPos[2], - width, height, - pt, stfp); + width, height, pt); pipe_texture_reference(&pt, NULL); } } - -void st_init_bitmap_functions(struct dd_function_table *functions) +/** Per-context init */ +void +st_init_bitmap_functions(struct dd_function_table *functions) { functions->Bitmap = st_Bitmap; } +/** Per-context init */ +void +st_init_bitmap(struct st_context *st) +{ + init_bitmap_cache(st); +} + + +/** Per-context tear-down */ void st_destroy_bitmap(struct st_context *st) { @@ -528,9 +732,15 @@ st_destroy_bitmap(struct st_context *st) pipe->delete_vs_state(pipe, st->bitmap.vs); st->bitmap.vs = NULL; } + if (st->bitmap.vbuf) { pipe->winsys->buffer_destroy(pipe->winsys, st->bitmap.vbuf); st->bitmap.vbuf = NULL; } -} + if (st->bitmap.cache) { + pipe_texture_release(&st->bitmap.cache->texture); + FREE(st->bitmap.cache); + st->bitmap.cache = NULL; + } +} diff --git a/src/mesa/state_tracker/st_cb_bitmap.h b/src/mesa/state_tracker/st_cb_bitmap.h index ac19e0ebb1..aae11d34c9 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.h +++ b/src/mesa/state_tracker/st_cb_bitmap.h @@ -33,9 +33,14 @@ extern void st_init_bitmap_functions(struct dd_function_table *functions); +extern void +st_init_bitmap(struct st_context *st); extern void st_destroy_bitmap(struct st_context *st); +extern void +st_flush_bitmap_cache(struct st_context *st); + #endif /* ST_CB_BITMAP_H */ diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index a536a059bd..e321b401e2 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -35,6 +35,7 @@ #include "main/macros.h" #include "main/context.h" #include "st_context.h" +#include "st_cb_bitmap.h" #include "st_cb_flush.h" #include "st_cb_fbo.h" #include "st_public.h" @@ -48,6 +49,8 @@ void st_flush( struct st_context *st, uint pipeFlushFlags, { FLUSH_VERTICES(st->ctx, 0); + st_flush_bitmap_cache(st); + st->pipe->flush( st->pipe, pipeFlushFlags, fence ); } diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 726e06d7c2..a20195f2de 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -109,6 +109,7 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st->cso_context = cso_create_context(pipe); st_init_atoms( st ); + st_init_bitmap(st); st_init_draw( st ); st_init_generate_mipmap(st); st_init_blit(st); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index f235c194b7..85e3d47e1a 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -42,6 +42,7 @@ struct cso_cache; struct cso_blend; struct gen_mipmap_state; struct blit_state; +struct bitmap_cache; #define ST_NEW_MESA 0x1 /* Mesa state has changed */ @@ -151,6 +152,7 @@ struct st_context void *vs; float vertices[4][3][4]; /**< vertex pos + color + texcoord */ struct pipe_buffer *vbuf; + struct bitmap_cache *cache; } bitmap; /** for glClear */ -- cgit v1.2.3 From 7292db2138001b48bba006cc08e9ff7091d16559 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Mar 2008 14:56:05 -0600 Subject: gallium: disable a debug hack --- src/mesa/state_tracker/st_cb_bitmap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index e46dcbf661..464e22d576 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -65,7 +65,7 @@ * A wide, short buffer is used to target the common case of a series * of glBitmap calls being used to draw text. */ -static GLboolean UseBitmapCache = 0*GL_TRUE; +static GLboolean UseBitmapCache = GL_TRUE; #define BITMAP_CACHE_WIDTH 512 -- cgit v1.2.3 From 5a460c7391ef35b1dcf6ad7f5494fb23279b2e45 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Mar 2008 15:17:50 -0600 Subject: gallium: don't call st_flush_bitmap_cache() if the only change is _NEW_PACKUNPACK state --- src/mesa/state_tracker/st_atom.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 40e4142631..18063adc79 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -148,7 +148,12 @@ void st_validate_state( struct st_context *st ) struct st_state_flags *state = &st->dirty; GLuint i; - st_flush_bitmap_cache(st); + /* The bitmap cache is immune to pixel unpack changes. + * Note that GLUT makes several calls to glPixelStore for each + * bitmap char it draws so this is an important check. + */ + if (state->mesa & ~_NEW_PACKUNPACK) + st_flush_bitmap_cache(st); check_program_state( st ); -- cgit v1.2.3 From cbfe6ee5d58e7342012392a8ead7ae373625c00a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Mar 2008 15:22:34 -0600 Subject: gallium: Fix computation of Z values when not using early Z. This fixes the missing bitmaps in the engine and fogcoord demos. --- src/gallium/drivers/softpipe/sp_quad_fs.c | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index c10ad80e01..a73df31383 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -113,15 +113,18 @@ shade_quad( } } else { - /* copy input Z (which was interpolated by the executor) to output Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i]; - /* XXX not sure the above line is always correct. The following - * might be better: - quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i]; - */ - } + /* compute Z values now, as in the quad earlyz stage */ + /* XXX we should really only do this if the earlyz stage is not used */ + const float fx = (float) quad->x0; + const float fy = (float) quad->y0; + const float dzdx = quad->posCoef->dadx[2]; + const float dzdy = quad->posCoef->dady[2]; + const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy; + + quad->outputs.depth[0] = z0; + quad->outputs.depth[1] = z0 + dzdx; + quad->outputs.depth[2] = z0 + dzdy; + quad->outputs.depth[3] = z0 + dzdx + dzdy; } /* shader may cull fragments */ -- cgit v1.2.3 From f6908a766dce645d610ff04bb49eaa8c5ee9e65a Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Mar 2008 18:18:55 -0600 Subject: gallium: added an (int) cast in setup_bitmap_vertex_data() to fix a signed/unsigned arithmetic problem Negative values became very large uints. --- src/mesa/state_tracker/st_cb_bitmap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 464e22d576..b600b92dc6 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -351,7 +351,7 @@ setup_bitmap_vertex_data(struct st_context *st, const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP); const GLfloat x0 = x; const GLfloat x1 = x + width; - const GLfloat y0 = invert ? (fb->Height - y - height) : y; + const GLfloat y0 = invert ? ((int) fb->Height - y - height) : y; const GLfloat y1 = invert ? (y0 + height) : y + height; const GLfloat bias = st->bitmap_texcoord_bias; const GLfloat xBias = bias / (x1-x0); -- cgit v1.2.3 From 737e34aee598f32f8ff078ba823ed149b282ebc8 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 28 Mar 2008 18:41:10 -0600 Subject: gallium: begin some bounding box code for bitmap cache --- src/mesa/state_tracker/st_cb_bitmap.c | 72 +++++++++++++++++++++++++---------- 1 file changed, 52 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index b600b92dc6..ec56b25f7c 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -73,12 +73,14 @@ static GLboolean UseBitmapCache = GL_TRUE; struct bitmap_cache { - /** An I8 texture image: */ - GLubyte buffer[BITMAP_CACHE_HEIGHT][BITMAP_CACHE_WIDTH]; - GLboolean empty; /** Window pos to render the cached image */ GLint xpos, ypos; + /** Bounds of region used in window coords */ + GLint xmin, ymin, xmax, ymax; struct pipe_texture *texture; + GLboolean empty; + /** An I8 texture image: */ + GLubyte buffer[BITMAP_CACHE_HEIGHT][BITMAP_CACHE_WIDTH]; }; @@ -493,6 +495,18 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, } +static void +reset_cache(struct st_context *st) +{ + memset(st->bitmap.cache->buffer, 0, sizeof(st->bitmap.cache->buffer)); + st->bitmap.cache->empty = GL_TRUE; + + st->bitmap.cache->xmin = 1000000; + st->bitmap.cache->xmax = -1000000; + st->bitmap.cache->ymin = 1000000; + st->bitmap.cache->ymax = -1000000; +} + static void init_bitmap_cache(struct st_context *st) @@ -523,7 +537,7 @@ init_bitmap_cache(struct st_context *st) return; } - st->bitmap.cache->empty = GL_TRUE; + reset_cache(st); } @@ -534,29 +548,37 @@ void st_flush_bitmap_cache(struct st_context *st) { if (!st->bitmap.cache->empty) { + struct bitmap_cache *cache = st->bitmap.cache; struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; struct pipe_surface *surf; void *dest; + assert(cache->xmin <= cache->xmax); + /* + printf("flush size %d x %d at %d, %d\n", + cache->xmax - cache->xmin, + cache->ymax - cache->ymin, + cache->xpos, cache->ypos); + */ + /* update the texture map image */ - surf = screen->get_tex_surface(screen, st->bitmap.cache->texture, 0, 0, 0); + surf = screen->get_tex_surface(screen, cache->texture, 0, 0, 0); dest = pipe_surface_map(surf); - memcpy(dest, st->bitmap.cache->buffer, sizeof(st->bitmap.cache->buffer)); + memcpy(dest, cache->buffer, sizeof(cache->buffer)); pipe_surface_unmap(surf); pipe_surface_reference(&surf, NULL); - pipe->texture_update(pipe, st->bitmap.cache->texture, 0, 0x1); + pipe->texture_update(pipe, cache->texture, 0, 0x1); draw_bitmap_quad(st->ctx, - st->bitmap.cache->xpos, - st->bitmap.cache->ypos, + cache->xpos, + cache->ypos, st->ctx->Current.RasterPos[2], BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, - st->bitmap.cache->texture); + cache->texture); - memset(st->bitmap.cache->buffer, 0, sizeof(st->bitmap.cache->buffer)); - st->bitmap.cache->empty = GL_TRUE; + reset_cache(st); } } @@ -571,6 +593,7 @@ accum_bitmap(struct st_context *st, const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap ) { + struct bitmap_cache *cache = st->bitmap.cache; int row, col; int px = -999, py; @@ -578,9 +601,9 @@ accum_bitmap(struct st_context *st, height > BITMAP_CACHE_HEIGHT) return GL_FALSE; /* too big to cache */ - if (!st->bitmap.cache->empty) { - px = x - st->bitmap.cache->xpos; /* pos in buffer */ - py = y - st->bitmap.cache->ypos; + if (!cache->empty) { + px = x - cache->xpos; /* pos in buffer */ + py = y - cache->ypos; if (px < 0 || px + width > BITMAP_CACHE_WIDTH || py < 0 || py + height > BITMAP_CACHE_HEIGHT) { /* This bitmap would extend beyond cache bounds, @@ -590,20 +613,29 @@ accum_bitmap(struct st_context *st, } } - if (st->bitmap.cache->empty) { + if (cache->empty) { /* Initialize. Center bitmap vertically in the buffer. */ px = 0; py = (BITMAP_CACHE_HEIGHT - height) / 2; - st->bitmap.cache->xpos = x; - st->bitmap.cache->ypos = y - py; - st->bitmap.cache->empty = GL_FALSE; + cache->xpos = x; + cache->ypos = y - py; + cache->empty = GL_FALSE; } assert(px != -999); + if (x < cache->xmin) + cache->xmin = x; + if (y < cache->ymin) + cache->ymin = y; + if (x + width > cache->xmax) + cache->xmax = x + width; + if (y + height > cache->ymax) + cache->ymax = y + height; + /* XXX try to combine this code with code in make_bitmap_texture() */ #define SET_PIXEL(COL, ROW) \ - st->bitmap.cache->buffer[py + (ROW)][px + (COL)] = 0xff; + cache->buffer[py + (ROW)][px + (COL)] = 0xff; for (row = 0; row < height; row++) { const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, -- cgit v1.2.3 From f10016b9a0639d7bc814c7b92a30d5b5b2cba5ad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 25 Mar 2008 17:30:34 +0000 Subject: gallium: Fix some MSVC warnings. --- src/gallium/auxiliary/draw/draw_pt_vcache.c | 2 +- src/gallium/auxiliary/util/u_snprintf.c | 8 ++++---- 2 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 98c22eb4d4..da9a3a52ae 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -107,7 +107,7 @@ static void vcache_elt( struct vcache_frontend *vcache, assert(vcache->fetch_count < FETCH_MAX); vcache->in[idx] = felt; - vcache->out[idx] = vcache->fetch_count; + vcache->out[idx] = (ushort)vcache->fetch_count; vcache->fetch_elts[vcache->fetch_count++] = felt; } diff --git a/src/gallium/auxiliary/util/u_snprintf.c b/src/gallium/auxiliary/util/u_snprintf.c index 61c20b48f7..48426abcb7 100644 --- a/src/gallium/auxiliary/util/u_snprintf.c +++ b/src/gallium/auxiliary/util/u_snprintf.c @@ -783,19 +783,19 @@ rpl_vsnprintf(char *str, size_t size, const char *format, va_list args) switch (cflags) { case PRINT_C_CHAR: charptr = va_arg(args, signed char *); - *charptr = len; + *charptr = (signed char)len; break; case PRINT_C_SHORT: shortptr = va_arg(args, short int *); - *shortptr = len; + *shortptr = (short int)len; break; case PRINT_C_LONG: longptr = va_arg(args, long int *); - *longptr = len; + *longptr = (long int)len; break; case PRINT_C_LLONG: llongptr = va_arg(args, LLONG *); - *llongptr = len; + *llongptr = (LLONG)len; break; case PRINT_C_SIZE: /* -- cgit v1.2.3 From a52c0416d1f2105960b4646e2e268aed26814689 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 29 Mar 2008 14:41:03 +0100 Subject: gallium: Set vertex state/buffers en-mass. --- src/gallium/auxiliary/draw/draw_context.c | 24 ++++--- src/gallium/auxiliary/draw/draw_context.h | 12 ++-- src/gallium/auxiliary/util/u_draw_quad.c | 16 +++-- src/gallium/drivers/cell/ppu/cell_context.c | 4 +- src/gallium/drivers/cell/ppu/cell_state.h | 12 ++-- src/gallium/drivers/cell/ppu/cell_state_vertex.c | 30 +++++---- src/gallium/drivers/failover/fo_context.h | 9 +-- src/gallium/drivers/failover/fo_state.c | 35 +++++----- src/gallium/drivers/failover/fo_state_emit.c | 22 ++----- src/gallium/drivers/i915simple/i915_state.c | 23 +++---- src/gallium/drivers/i965simple/brw_state.c | 60 +++++++++-------- src/gallium/drivers/softpipe/sp_context.c | 4 +- src/gallium/drivers/softpipe/sp_state.h | 12 ++-- src/gallium/drivers/softpipe/sp_state_vertex.c | 31 +++++---- src/gallium/include/pipe/p_context.h | 12 ++-- src/mesa/state_tracker/st_draw.c | 82 +++++++++++++----------- 16 files changed, 205 insertions(+), 183 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 81858e01ca..10bf9f54c1 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -233,24 +233,28 @@ void draw_set_viewport_state( struct draw_context *draw, void -draw_set_vertex_buffer(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_buffer *buffer) +draw_set_vertex_buffers(struct draw_context *draw, + unsigned count, + const struct pipe_vertex_buffer *buffers) { + assert(count <= PIPE_MAX_ATTRIBS); + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - assert(attr < PIPE_MAX_ATTRIBS); - draw->vertex_buffer[attr] = *buffer; + + memcpy(draw->vertex_buffer, buffers, count * sizeof(buffers[0])); } void -draw_set_vertex_element(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_element *element) +draw_set_vertex_elements(struct draw_context *draw, + unsigned count, + const struct pipe_vertex_element *elements) { + assert(count <= PIPE_MAX_ATTRIBS); + draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - assert(attr < PIPE_MAX_ATTRIBS); - draw->vertex_element[attr] = *element; + + memcpy(draw->vertex_element, elements, count * sizeof(elements[0])); } diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index dae687e590..84bae3bd78 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -138,13 +138,13 @@ void draw_delete_vertex_shader(struct draw_context *draw, * Vertex data functions */ -void draw_set_vertex_buffer(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_buffer *buffer); +void draw_set_vertex_buffers(struct draw_context *draw, + unsigned count, + const struct pipe_vertex_buffer *buffers); -void draw_set_vertex_element(struct draw_context *draw, - unsigned attr, - const struct pipe_vertex_element *element); +void draw_set_vertex_elements(struct draw_context *draw, + unsigned count, + const struct pipe_vertex_element *elements); void draw_set_mapped_element_buffer( struct draw_context *draw, unsigned eltSize, void *elements ); diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c index 37e8533609..e659edb088 100644 --- a/src/gallium/auxiliary/util/u_draw_quad.c +++ b/src/gallium/auxiliary/util/u_draw_quad.c @@ -45,23 +45,25 @@ util_draw_vertex_buffer(struct pipe_context *pipe, uint num_attribs) { struct pipe_vertex_buffer vbuffer; - struct pipe_vertex_element velement; + struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS]; uint i; + assert(num_attribs <= PIPE_MAX_ATTRIBS); + /* tell pipe about the vertex buffer */ vbuffer.buffer = vbuf; vbuffer.pitch = num_attribs * 4 * sizeof(float); /* vertex size */ vbuffer.buffer_offset = 0; - pipe->set_vertex_buffer(pipe, 0, &vbuffer); + pipe->set_vertex_buffers(pipe, 1, &vbuffer); /* tell pipe about the vertex attributes */ for (i = 0; i < num_attribs; i++) { - velement.src_offset = i * 4 * sizeof(float); - velement.vertex_buffer_index = 0; - velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - velement.nr_components = 4; - pipe->set_vertex_element(pipe, i, &velement); + velements[i].src_offset = i * 4 * sizeof(float); + velements[i].vertex_buffer_index = 0; + velements[i].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + velements[i].nr_components = 4; } + pipe->set_vertex_elements(pipe, num_attribs, velements); /* draw */ pipe->draw_arrays(pipe, prim_type, 0, num_verts); diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index ccbbd1d331..12eb5aa254 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -103,8 +103,8 @@ cell_create_context(struct pipe_screen *screen, cell->pipe.destroy = cell_destroy_context; /* state setters */ - cell->pipe.set_vertex_buffer = cell_set_vertex_buffer; - cell->pipe.set_vertex_element = cell_set_vertex_element; + cell->pipe.set_vertex_buffers = cell_set_vertex_buffers; + cell->pipe.set_vertex_elements = cell_set_vertex_elements; cell->pipe.draw_arrays = cell_draw_arrays; cell->pipe.draw_elements = cell_draw_elements; diff --git a/src/gallium/drivers/cell/ppu/cell_state.h b/src/gallium/drivers/cell/ppu/cell_state.h index 31ce505e21..82580ea35a 100644 --- a/src/gallium/drivers/cell/ppu/cell_state.h +++ b/src/gallium/drivers/cell/ppu/cell_state.h @@ -48,13 +48,13 @@ #define CELL_NEW_VERTEX_INFO 0x8000 -void cell_set_vertex_element(struct pipe_context *, - unsigned index, - const struct pipe_vertex_element *); +void cell_set_vertex_elements(struct pipe_context *, + unsigned count, + const struct pipe_vertex_element *); -void cell_set_vertex_buffer(struct pipe_context *, - unsigned index, - const struct pipe_vertex_buffer *); +void cell_set_vertex_buffers(struct pipe_context *, + unsigned count, + const struct pipe_vertex_buffer *); void cell_update_derived( struct cell_context *softpipe ); diff --git a/src/gallium/drivers/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c index 37d25fb357..6c83b8dc72 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_vertex.c +++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c @@ -36,28 +36,34 @@ void -cell_set_vertex_element(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_element *attrib) +cell_set_vertex_elements(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_element *elements) { struct cell_context *cell = cell_context(pipe); - assert(index < PIPE_MAX_ATTRIBS); - cell->vertex_element[index] = *attrib; /* struct copy */ + + assert(count <= PIPE_MAX_ATTRIBS); + + memcpy(cell->vertex_element, elements, count * sizeof(elements[0])); + cell->dirty |= CELL_NEW_VERTEX; - draw_set_vertex_element(cell->draw, index, attrib); + draw_set_vertex_elements(cell->draw, count, elements); } void -cell_set_vertex_buffer(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer) +cell_set_vertex_buffers(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_buffer *buffers) { struct cell_context *cell = cell_context(pipe); - assert(index < PIPE_MAX_ATTRIBS); - cell->vertex_buffer[index] = *buffer; /* struct copy */ + + assert(count <= PIPE_MAX_ATTRIBS); + + memcpy(cell->vertex_buffer, buffers, count * sizeof(buffers[0])); + cell->dirty |= CELL_NEW_VERTEX; - draw_set_vertex_buffer(cell->draw, index, buffer); + draw_set_vertex_buffers(cell->draw, count, buffers); } diff --git a/src/gallium/drivers/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h index 4afe10c4b8..c6409fe1e1 100644 --- a/src/gallium/drivers/failover/fo_context.h +++ b/src/gallium/drivers/failover/fo_context.h @@ -84,15 +84,16 @@ struct failover_context { struct pipe_scissor_state scissor; struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; - struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS]; + + uint num_vertex_buffers; + uint num_vertex_elements; void *sw_sampler_state[PIPE_MAX_SAMPLERS]; void *hw_sampler_state[PIPE_MAX_SAMPLERS]; unsigned dirty; - unsigned dirty_vertex_buffer; - unsigned dirty_vertex_element; unsigned num_samplers; unsigned num_textures; diff --git a/src/gallium/drivers/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c index 11eec2714e..6a79706632 100644 --- a/src/gallium/drivers/failover/fo_state.c +++ b/src/gallium/drivers/failover/fo_state.c @@ -402,32 +402,35 @@ failover_set_viewport_state( struct pipe_context *pipe, static void -failover_set_vertex_buffer(struct pipe_context *pipe, - unsigned unit, - const struct pipe_vertex_buffer *vertex_buffer) +failover_set_vertex_buffers(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_buffer *vertex_buffers) { struct failover_context *failover = failover_context(pipe); - failover->vertex_buffer[unit] = *vertex_buffer; + memcpy(failover->vertex_buffers, vertex_buffers, + count * sizeof(vertex_buffers[0])); failover->dirty |= FO_NEW_VERTEX_BUFFER; - failover->dirty_vertex_buffer |= (1<sw->set_vertex_buffer( failover->sw, unit, vertex_buffer ); - failover->hw->set_vertex_buffer( failover->hw, unit, vertex_buffer ); + failover->num_vertex_buffers = count; + failover->sw->set_vertex_buffers( failover->sw, count, vertex_buffers ); + failover->hw->set_vertex_buffers( failover->hw, count, vertex_buffers ); } static void -failover_set_vertex_element(struct pipe_context *pipe, - unsigned unit, - const struct pipe_vertex_element *vertex_element) +failover_set_vertex_elements(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_element *vertex_elements) { struct failover_context *failover = failover_context(pipe); - failover->vertex_element[unit] = *vertex_element; + memcpy(failover->vertex_elements, vertex_elements, + count * sizeof(vertex_elements[0])); + failover->dirty |= FO_NEW_VERTEX_ELEMENT; - failover->dirty_vertex_element |= (1<sw->set_vertex_element( failover->sw, unit, vertex_element ); - failover->hw->set_vertex_element( failover->hw, unit, vertex_element ); + failover->num_vertex_elements = count; + failover->sw->set_vertex_elements( failover->sw, count, vertex_elements ); + failover->hw->set_vertex_elements( failover->hw, count, vertex_elements ); } void @@ -474,7 +477,7 @@ failover_init_state_functions( struct failover_context *failover ) failover->pipe.set_scissor_state = failover_set_scissor_state; failover->pipe.set_sampler_textures = failover_set_sampler_textures; failover->pipe.set_viewport_state = failover_set_viewport_state; - failover->pipe.set_vertex_buffer = failover_set_vertex_buffer; - failover->pipe.set_vertex_element = failover_set_vertex_element; + failover->pipe.set_vertex_buffers = failover_set_vertex_buffers; + failover->pipe.set_vertex_elements = failover_set_vertex_elements; failover->pipe.set_constant_buffer = failover_set_constant_buffer; } diff --git a/src/gallium/drivers/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c index bb89f925e9..bd4fce9d20 100644 --- a/src/gallium/drivers/failover/fo_state_emit.c +++ b/src/gallium/drivers/failover/fo_state_emit.c @@ -53,8 +53,6 @@ void failover_state_emit( struct failover_context *failover ) { - unsigned i; - if (failover->dirty & FO_NEW_BLEND) failover->sw->bind_blend_state( failover->sw, failover->blend->sw_state ); @@ -104,24 +102,16 @@ failover_state_emit( struct failover_context *failover ) } if (failover->dirty & FO_NEW_VERTEX_BUFFER) { - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (failover->dirty_vertex_buffer & (1<sw->set_vertex_buffer( failover->sw, i, - &failover->vertex_buffer[i] ); - } - } + failover->sw->set_vertex_buffers( failover->sw, + failover->num_vertex_buffers, + failover->vertex_buffers ); } if (failover->dirty & FO_NEW_VERTEX_ELEMENT) { - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (failover->dirty_vertex_element & (1<sw->set_vertex_element( failover->sw, i, - &failover->vertex_element[i] ); - } - } + failover->sw->set_vertex_elements( failover->sw, + failover->num_vertex_elements, + failover->vertex_elements ); } failover->dirty = 0; - failover->dirty_vertex_element = 0; - failover->dirty_vertex_buffer = 0; } diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 1cec36e206..4404bc4590 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -687,23 +687,24 @@ static void i915_delete_rasterizer_state(struct pipe_context *pipe, FREE(raster); } -static void i915_set_vertex_buffer( struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer ) +static void i915_set_vertex_buffers(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_buffer *buffers) { struct i915_context *i915 = i915_context(pipe); - i915->vertex_buffer[index] = *buffer; + + memcpy(i915->vertex_buffer, buffers, count * sizeof(buffers[0])); /* pass-through to draw module */ - draw_set_vertex_buffer(i915->draw, index, buffer); + draw_set_vertex_buffers(i915->draw, count, buffers); } -static void i915_set_vertex_element( struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_element *element) +static void i915_set_vertex_elements(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_element *elements) { struct i915_context *i915 = i915_context(pipe); /* pass-through to draw module */ - draw_set_vertex_element(i915->draw, index, element); + draw_set_vertex_elements(i915->draw, count, elements); } @@ -742,6 +743,6 @@ i915_init_state_functions( struct i915_context *i915 ) i915->pipe.set_scissor_state = i915_set_scissor_state; i915->pipe.set_sampler_textures = i915_set_sampler_textures; i915->pipe.set_viewport_state = i915_set_viewport_state; - i915->pipe.set_vertex_buffer = i915_set_vertex_buffer; - i915->pipe.set_vertex_element = i915_set_vertex_element; + i915->pipe.set_vertex_buffers = i915_set_vertex_buffers; + i915->pipe.set_vertex_elements = i915_set_vertex_elements; } diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 0d04a8a594..376f1487b2 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -277,45 +277,49 @@ static void brw_set_viewport_state( struct pipe_context *pipe, } -static void brw_set_vertex_buffer( struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer ) +static void brw_set_vertex_buffers(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_buffer *buffers) { struct brw_context *brw = brw_context(pipe); - brw->vb.vbo_array[index] = buffer; + memcpy(brw->vb.vbo_array, buffers, count * sizeof(buffers[0])); } -static void brw_set_vertex_element(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_element *element) +static void brw_set_vertex_elements(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_element *elements) { /* flush ? */ struct brw_context *brw = brw_context(pipe); + uint i; - assert(index < PIPE_MAX_ATTRIBS); - struct brw_vertex_element_state el; - memset(&el, 0, sizeof(el)); + assert(count <= PIPE_MAX_ATTRIBS); - el.ve0.src_offset = element->src_offset; - el.ve0.src_format = brw_translate_surface_format(element->src_format); - el.ve0.valid = 1; - el.ve0.vertex_buffer_index = element->vertex_buffer_index; + for (i = 0; i < count; i++) { + struct brw_vertex_element_state el; + memset(&el, 0, sizeof(el)); - el.ve1.dst_offset = index * 4; + el.ve0.src_offset = elements[i].src_offset; + el.ve0.src_format = brw_translate_surface_format(elements[i].src_format); + el.ve0.valid = 1; + el.ve0.vertex_buffer_index = elements[i].vertex_buffer_index; - el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.dst_offset = i * 4; - switch (element->nr_components) { - case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; - case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; - case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; - break; - } + el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; - brw->vb.inputs[index] = el; + switch (elements[i].nr_components) { + case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; + case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; + case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; + break; + } + + brw->vb.inputs[i] = el; + } } @@ -457,6 +461,6 @@ brw_init_state_functions( struct brw_context *brw ) brw->pipe.set_scissor_state = brw_set_scissor_state; brw->pipe.set_sampler_textures = brw_set_sampler_textures; brw->pipe.set_viewport_state = brw_set_viewport_state; - brw->pipe.set_vertex_buffer = brw_set_vertex_buffer; - brw->pipe.set_vertex_element = brw_set_vertex_element; + brw->pipe.set_vertex_buffers = brw_set_vertex_buffers; + brw->pipe.set_vertex_elements = brw_set_vertex_elements; } diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 16fb06f176..e298ed37c3 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -174,8 +174,8 @@ softpipe_create( struct pipe_screen *screen, softpipe->pipe.set_sampler_textures = softpipe_set_sampler_textures; softpipe->pipe.set_viewport_state = softpipe_set_viewport_state; - softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer; - softpipe->pipe.set_vertex_element = softpipe_set_vertex_element; + softpipe->pipe.set_vertex_buffers = softpipe_set_vertex_buffers; + softpipe->pipe.set_vertex_elements = softpipe_set_vertex_elements; softpipe->pipe.draw_arrays = softpipe_draw_arrays; softpipe->pipe.draw_elements = softpipe_draw_elements; diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 0bb1095aec..6e6501f5bc 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -151,13 +151,13 @@ void softpipe_set_sampler_textures( struct pipe_context *, void softpipe_set_viewport_state( struct pipe_context *, const struct pipe_viewport_state * ); -void softpipe_set_vertex_element(struct pipe_context *, - unsigned index, - const struct pipe_vertex_element *); +void softpipe_set_vertex_elements(struct pipe_context *, + unsigned count, + const struct pipe_vertex_element *); -void softpipe_set_vertex_buffer(struct pipe_context *, - unsigned index, - const struct pipe_vertex_buffer *); +void softpipe_set_vertex_buffers(struct pipe_context *, + unsigned count, + const struct pipe_vertex_buffer *); void softpipe_update_derived( struct softpipe_context *softpipe ); diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c index c054e76d9b..e0230e16a4 100644 --- a/src/gallium/drivers/softpipe/sp_state_vertex.c +++ b/src/gallium/drivers/softpipe/sp_state_vertex.c @@ -37,28 +37,35 @@ void -softpipe_set_vertex_element(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_element *attrib) +softpipe_set_vertex_elements(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_element *attribs) { struct softpipe_context *softpipe = softpipe_context(pipe); - assert(index < PIPE_MAX_ATTRIBS); - softpipe->vertex_element[index] = *attrib; /* struct copy */ + + assert(count <= PIPE_MAX_ATTRIBS); + + memcpy(softpipe->vertex_element, attribs, + count * sizeof(struct pipe_vertex_element)); + softpipe->dirty |= SP_NEW_VERTEX; - draw_set_vertex_element(softpipe->draw, index, attrib); + draw_set_vertex_elements(softpipe->draw, count, attribs); } void -softpipe_set_vertex_buffer(struct pipe_context *pipe, - unsigned index, - const struct pipe_vertex_buffer *buffer) +softpipe_set_vertex_buffers(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_buffer *buffers) { struct softpipe_context *softpipe = softpipe_context(pipe); - assert(index < PIPE_MAX_ATTRIBS); - softpipe->vertex_buffer[index] = *buffer; /* struct copy */ + + assert(count <= PIPE_MAX_ATTRIBS); + + memcpy(softpipe->vertex_buffer, buffers, count * sizeof(buffers[0])); + softpipe->dirty |= SP_NEW_VERTEX; - draw_set_vertex_buffer(softpipe->draw, index, buffer); + draw_set_vertex_buffers(softpipe->draw, count, buffers); } diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 2d063e3f9e..324f70185a 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -155,13 +155,13 @@ struct pipe_context { unsigned num_textures, struct pipe_texture ** ); - void (*set_vertex_buffer)( struct pipe_context *, - unsigned index, - const struct pipe_vertex_buffer * ); + void (*set_vertex_buffers)( struct pipe_context *, + unsigned num_buffers, + const struct pipe_vertex_buffer * ); - void (*set_vertex_element)( struct pipe_context *, - unsigned index, - const struct pipe_vertex_element * ); + void (*set_vertex_elements)( struct pipe_context *, + unsigned num_elements, + const struct pipe_vertex_element * ); /*@}*/ diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 4aca3311b7..f0f62246dd 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -210,6 +210,7 @@ st_draw_vbo(GLcontext *ctx, const struct pipe_shader_state *vs; struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; GLuint attr; + struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS]; /* sanity check for pointer arithmetic below */ assert(sizeof(arrays[0]->Ptr[0]) == 1); @@ -226,7 +227,6 @@ st_draw_vbo(GLcontext *ctx, for (attr = 0; attr < vp->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; - struct pipe_vertex_element velement; if (bufobj && bufobj->Name) { /* Attribute data is in a VBO. @@ -239,8 +239,8 @@ st_draw_vbo(GLcontext *ctx, vbuffer[attr].buffer = NULL; pipe_buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); vbuffer[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ - velement.src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; - assert(velement.src_offset <= 2048); /* 11-bit field */ + velements[attr].src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; + assert(velements[attr].src_offset <= 2048); /* 11-bit field */ } else { /* attribute data is in user-space memory, not a VBO */ @@ -259,24 +259,24 @@ st_draw_vbo(GLcontext *ctx, (void *) arrays[mesaAttr]->Ptr, bytes); vbuffer[attr].buffer_offset = 0; - velement.src_offset = 0; + velements[attr].src_offset = 0; } /* common-case setup */ vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ vbuffer[attr].max_index = max_index; - velement.vertex_buffer_index = attr; - velement.nr_components = arrays[mesaAttr]->Size; - velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, - arrays[mesaAttr]->Size, - arrays[mesaAttr]->Normalized); - assert(velement.src_format); - - /* tell pipe about this attribute */ - pipe->set_vertex_buffer(pipe, attr, &vbuffer[attr]); - pipe->set_vertex_element(pipe, attr, &velement); + velements[attr].vertex_buffer_index = attr; + velements[attr].nr_components = arrays[mesaAttr]->Size; + velements[attr].src_format + = pipe_vertex_format(arrays[mesaAttr]->Type, + arrays[mesaAttr]->Size, + arrays[mesaAttr]->Normalized); + assert(velements[attr].src_format); } + pipe->set_vertex_buffers(pipe, vp->num_inputs, vbuffer); + pipe->set_vertex_elements(pipe, vp->num_inputs, velements); + /* do actual drawing */ if (ib) { @@ -336,8 +336,8 @@ st_draw_vbo(GLcontext *ctx, for (attr = 0; attr < vp->num_inputs; attr++) { pipe_buffer_reference(winsys, &vbuffer[attr].buffer, NULL); assert(!vbuffer[attr].buffer); - pipe->set_vertex_buffer(pipe, attr, &vbuffer[attr]); } + pipe->set_vertex_buffers(pipe, vp->num_inputs, vbuffer); } @@ -362,7 +362,7 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, struct pipe_context *pipe = ctx->st->pipe; struct pipe_buffer *vbuf; struct pipe_vertex_buffer vbuffer; - struct pipe_vertex_element velement; + struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS]; unsigned i; assert(numAttribs > 0); @@ -393,16 +393,16 @@ st_draw_vertices(GLcontext *ctx, unsigned prim, vbuffer.buffer = vbuf; vbuffer.pitch = numAttribs * 4 * sizeof(float); /* vertex size */ vbuffer.buffer_offset = 0; - pipe->set_vertex_buffer(pipe, 0, &vbuffer); + pipe->set_vertex_buffers(pipe, 1, &vbuffer); /* tell pipe about the vertex attributes */ for (i = 0; i < numAttribs; i++) { - velement.src_offset = i * 4 * sizeof(GLfloat); - velement.vertex_buffer_index = 0; - velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - velement.nr_components = 4; - pipe->set_vertex_element(pipe, i, &velement); + velements[i].src_offset = i * 4 * sizeof(GLfloat); + velements[i].vertex_buffer_index = 0; + velements[i].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + velements[i].nr_components = 4; } + pipe->set_vertex_elements(pipe, numAttribs, velements); /* draw */ pipe->draw_arrays(pipe, prim, 0, numVertex); @@ -470,7 +470,8 @@ st_feedback_draw_vbo(GLcontext *ctx, const struct st_vertex_program *vp; const struct pipe_shader_state *vs; struct pipe_buffer *index_buffer_handle = 0; - struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; + struct pipe_vertex_buffer vbuffers[PIPE_MAX_SHADER_INPUTS]; + struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS]; GLuint attr, i; ubyte *mapped_constants; @@ -505,7 +506,6 @@ st_feedback_draw_vbo(GLcontext *ctx, for (attr = 0; attr < vp->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; - struct pipe_vertex_element velement; void *map; if (bufobj && bufobj->Name) { @@ -516,10 +516,10 @@ st_feedback_draw_vbo(GLcontext *ctx, struct st_buffer_object *stobj = st_buffer_object(bufobj); assert(stobj->buffer); - vbuffer[attr].buffer = NULL; - pipe_buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); - vbuffer[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ - velement.src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; + vbuffers[attr].buffer = NULL; + pipe_buffer_reference(winsys, &vbuffers[attr].buffer, stobj->buffer); + vbuffers[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ + velements[attr].src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; } else { /* attribute data is in user-space memory, not a VBO */ @@ -528,35 +528,39 @@ st_feedback_draw_vbo(GLcontext *ctx, * (max_index + 1)); /* wrap user data */ - vbuffer[attr].buffer + vbuffers[attr].buffer = winsys->user_buffer_create(winsys, (void *) arrays[mesaAttr]->Ptr, bytes); - vbuffer[attr].buffer_offset = 0; - velement.src_offset = 0; + vbuffers[attr].buffer_offset = 0; + velements[attr].src_offset = 0; } /* common-case setup */ - vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ - vbuffer[attr].max_index = max_index; - velement.vertex_buffer_index = attr; - velement.nr_components = arrays[mesaAttr]->Size; - velement.src_format = pipe_vertex_format(arrays[mesaAttr]->Type, + vbuffers[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ + vbuffers[attr].max_index = max_index; + velements[attr].vertex_buffer_index = attr; + velements[attr].nr_components = arrays[mesaAttr]->Size; + velements[attr].src_format = pipe_vertex_format(arrays[mesaAttr]->Type, arrays[mesaAttr]->Size, arrays[mesaAttr]->Normalized); - assert(velement.src_format); + assert(velements[attr].src_format); /* tell draw about this attribute */ +#if 0 draw_set_vertex_buffer(draw, attr, &vbuffer[attr]); - draw_set_vertex_element(draw, attr, &velement); +#endif /* map the attrib buffer */ map = pipe->winsys->buffer_map(pipe->winsys, - vbuffer[attr].buffer, + vbuffers[attr].buffer, PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, attr, map); } + draw_set_vertex_buffers(draw, vp->num_inputs, vbuffers); + draw_set_vertex_elements(draw, vp->num_inputs, velements); + if (ib) { unsigned indexSize; struct gl_buffer_object *bufobj = ib->obj; -- cgit v1.2.3 From fc9888014470286d8d651c569aaadf9cd69d8282 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 30 Mar 2008 22:48:49 +0200 Subject: draw: Fix bypass_vs semantic misuse. --- src/gallium/auxiliary/draw/draw_prim.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index ddcde01d9a..404c28d76a 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -594,6 +594,7 @@ draw_arrays(struct draw_context *draw, unsigned prim, /* drawing done here: */ if (!draw->rasterizer->bypass_vs || + (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first) || !draw_pt_arrays(draw, prim, start, count)) { /* we have to run the whole pipeline */ draw_prim(draw, prim, start, count); -- cgit v1.2.3 From 63950b11b6060e4e0d06e0d14548ff132a295067 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 30 Mar 2008 23:21:20 +0200 Subject: draw: Do not run full pipeline when flatshade_first for point primitives. --- src/gallium/auxiliary/draw/draw_prim.c | 1 - src/gallium/auxiliary/draw/draw_pt.c | 3 +++ src/gallium/auxiliary/draw/draw_validate.c | 8 ++++++++ 3 files changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index 404c28d76a..ddcde01d9a 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -594,7 +594,6 @@ draw_arrays(struct draw_context *draw, unsigned prim, /* drawing done here: */ if (!draw->rasterizer->bypass_vs || - (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first) || !draw_pt_arrays(draw, prim, start, count)) { /* we have to run the whole pipeline */ draw_prim(draw, prim, start, count); diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 3ec31ec25f..fc9304197a 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -36,6 +36,9 @@ #include "draw/draw_pt.h" +/* XXX: Shouldn't those two functions below use the '>' operator??? + */ + static boolean too_many_verts( struct draw_context *draw, unsigned verts ) { diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index e163e078f0..ad43f06f73 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -76,6 +76,10 @@ draw_need_pipeline(const struct draw_context *draw, /* AA lines */ if (draw->rasterizer->line_smooth && draw->pipeline.aaline) return TRUE; + + /* first-vertex driven flatshading */ + if (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first) + return TRUE; } if (points(prim)) @@ -112,6 +116,10 @@ draw_need_pipeline(const struct draw_context *draw, /* two-side lighting */ if (draw->rasterizer->light_twoside) return TRUE; + + /* first-vertex driven flatshading */ + if (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first) + return TRUE; } /* polygon cull - this is difficult - hardware can cull just fine -- cgit v1.2.3 From baab98a637d526871fb77ec6f313012f49c0e998 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 31 Mar 2008 09:02:08 +0900 Subject: gallium: Eliminate p_winsys::printf Not convenient and almost not used at all. Better replacements in p_debug.h --- src/gallium/drivers/i915simple/i915_debug.c | 19 +++++++------------ src/gallium/drivers/i915simple/i915_debug.h | 4 +--- src/gallium/drivers/i915simple/i915_debug_fp.c | 4 +--- src/gallium/drivers/i915simple/i915_screen.c | 5 ++--- src/gallium/drivers/i965simple/brw_context.c | 5 ++--- src/gallium/drivers/i965simple/brw_context.h | 6 +++--- src/gallium/include/pipe/p_winsys.h | 4 ---- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 10 ---------- src/gallium/winsys/xlib/xm_winsys.c | 11 ----------- src/gallium/winsys/xlib/xm_winsys_aub.c | 10 ---------- src/mesa/drivers/x11/xm_winsys.c | 10 ---------- 11 files changed, 16 insertions(+), 72 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c index 78102dbac2..9b9111167f 100644 --- a/src/gallium/drivers/i915simple/i915_debug.c +++ b/src/gallium/drivers/i915simple/i915_debug.c @@ -30,6 +30,7 @@ #include "i915_winsys.h" #include "i915_debug.h" #include "pipe/p_winsys.h" +#include "pipe/p_debug.h" static void @@ -39,11 +40,9 @@ PRINTF( ... ) { va_list args; - char buffer[256]; va_start( args, fmt ); - vsprintf( buffer, fmt, args ); - stream->winsys->printf( stream->winsys, buffer ); + debug_vprintf( fmt, args ); va_end( args ); } @@ -200,14 +199,12 @@ BITS( ... ) { va_list args; - char buffer[256]; unsigned himask = ~0UL >> (31 - (hi)); PRINTF(stream, "\t\t "); va_start( args, fmt ); - vsprintf( buffer, fmt, args ); - stream->winsys->printf( stream->winsys, buffer ); + debug_vprintf( fmt, args ); va_end( args ); PRINTF(stream, ": 0x%x\n", ((dw) & himask) >> (lo)); @@ -231,13 +228,11 @@ FLAG( { if (((dw) >> (bit)) & 1) { va_list args; - char buffer[256]; PRINTF(stream, "\t\t "); va_start( args, fmt ); - vsprintf( buffer, fmt, args ); - stream->winsys->printf( stream->winsys, buffer ); + debug_vprintf( fmt, args ); va_end( args ); PRINTF(stream, "\n"); @@ -877,11 +872,11 @@ i915_dump_batchbuffer( struct i915_context *i915 ) stream.winsys = i915->pipe.winsys; if (!start || !end) { - stream.winsys->printf( stream.winsys, "\n\nBATCH: ???\n"); + debug_printf( "\n\nBATCH: ???\n"); return; } - stream.winsys->printf( stream.winsys, "\n\nBATCH: (%d)\n", bytes / 4); + debug_printf( "\n\nBATCH: (%d)\n", bytes / 4); while (!done && stream.offset < bytes) @@ -893,7 +888,7 @@ i915_dump_batchbuffer( struct i915_context *i915 ) stream.offset >= 0); } - stream.winsys->printf( stream.winsys, "END-BATCH\n\n\n"); + debug_printf( "END-BATCH\n\n\n"); } diff --git a/src/gallium/drivers/i915simple/i915_debug.h b/src/gallium/drivers/i915simple/i915_debug.h index 0bcd094233..afb63edabf 100644 --- a/src/gallium/drivers/i915simple/i915_debug.h +++ b/src/gallium/drivers/i915simple/i915_debug.h @@ -83,11 +83,9 @@ I915_DBG( { if ((i915)->debug & FILE_DEBUG_FLAG) { va_list args; - char buffer[256]; va_start( args, fmt ); - vsprintf( buffer, fmt, args ); - i915->pipe.winsys->printf( i915->pipe.winsys, buffer ); + debug_vprintf( fmt, args ); va_end( args ); } } diff --git a/src/gallium/drivers/i915simple/i915_debug_fp.c b/src/gallium/drivers/i915simple/i915_debug_fp.c index ebfdb3d93c..37a3508fe1 100644 --- a/src/gallium/drivers/i915simple/i915_debug_fp.c +++ b/src/gallium/drivers/i915simple/i915_debug_fp.c @@ -39,11 +39,9 @@ PRINTF( ... ) { va_list args; - char buffer[256]; va_start( args, fmt ); - vsprintf( buffer, fmt, args ); - stream->winsys->printf( stream->winsys, buffer ); + debug_vprintf( fmt, args ); va_end( args ); } diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 8d7bf0b33e..839b98c0ce 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -226,9 +226,8 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id) break; default: - winsys->printf(winsys, - "%s: unknown pci id 0x%x, cannot create screen\n", - __FUNCTION__, pci_id); + debug_printf("%s: unknown pci id 0x%x, cannot create screen\n", + __FUNCTION__, pci_id); return NULL; } diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c index 7c908da672..a276cc0535 100644 --- a/src/gallium/drivers/i965simple/brw_context.c +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -77,9 +77,8 @@ struct pipe_context *brw_create(struct pipe_screen *screen, { struct brw_context *brw; - screen->winsys->printf(screen->winsys, - "%s: creating brw_context with pci id 0x%x\n", - __FUNCTION__, pci_id); + debug_printf("%s: creating brw_context with pci id 0x%x\n", + __FUNCTION__, pci_id); brw = CALLOC_STRUCT(brw_context); if (brw == NULL) diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index 0c96ba1732..eeccf36785 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -183,12 +183,12 @@ extern int BRW_DEBUG; #define DEBUG_MIPTREE 0x800000 #define DBG(...) do { \ - if (BRW_DEBUG & FILE_DEBUG_FLAG) \ - brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ + if (BRW_DEBUG & FILE_DEBUG_FLAG) \ + debug_printf(__VA_ARGS__); \ } while(0) #define PRINT(...) do { \ - brw->pipe.winsys->printf(brw->pipe.winsys, __VA_ARGS__); \ + debug_printf(brw->pipe.winsys, __VA_ARGS__); \ } while(0) struct brw_state_flags { diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h index 1383bd0544..8569cdcf12 100644 --- a/src/gallium/include/pipe/p_winsys.h +++ b/src/gallium/include/pipe/p_winsys.h @@ -73,10 +73,6 @@ struct pipe_winsys struct pipe_surface *surf, void *context_private ); - /** Debug output */ - void (*printf)( struct pipe_winsys *sws, - const char *, ... ); - /** allocate a new surface (no context dependency) */ struct pipe_surface *(*surface_alloc)(struct pipe_winsys *ws); diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 789a386500..77dec9488d 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -243,15 +243,6 @@ intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) -static void -intel_printf( struct pipe_winsys *winsys, const char *fmtString, ... ) -{ - va_list args; - va_start( args, fmtString ); - vfprintf(stderr, fmtString, args); - va_end( args ); -} - static const char * intel_get_name( struct pipe_winsys *winsys ) { @@ -277,7 +268,6 @@ intel_create_pipe_winsys( int fd ) iws->winsys.buffer_unmap = intel_buffer_unmap; iws->winsys.buffer_destroy = intel_buffer_destroy; iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; - iws->winsys.printf = intel_printf; iws->winsys.get_name = intel_get_name; iws->winsys.surface_alloc = intel_i915_surface_alloc; iws->winsys.surface_alloc_storage = intel_i915_surface_alloc_storage; diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 7459756279..9a20bdfb69 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -303,16 +303,6 @@ xm_flush_frontbuffer(struct pipe_winsys *pws, -static void -xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) -{ - va_list args; - va_start( args, fmtString ); - vfprintf(stderr, fmtString, args); - va_end( args ); -} - - static const char * xm_get_name(struct pipe_winsys *pws) { @@ -635,7 +625,6 @@ xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis) ws->base.fence_finish = xm_fence_finish; ws->base.flush_frontbuffer = xm_flush_frontbuffer; - ws->base.printf = xm_printf; ws->base.get_name = xm_get_name; } diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c index d55d8c39eb..f42f7fcc5f 100644 --- a/src/gallium/winsys/xlib/xm_winsys_aub.c +++ b/src/gallium/winsys/xlib/xm_winsys_aub.c @@ -311,15 +311,6 @@ aub_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) -static void -aub_printf( struct pipe_winsys *winsys, const char *fmtString, ... ) -{ - va_list args; - va_start( args, fmtString ); - vfprintf(stderr, fmtString, args); - va_end( args ); -} - static const char * aub_get_name( struct pipe_winsys *winsys ) { @@ -344,7 +335,6 @@ xmesa_create_pipe_winsys_aub( void ) iws->winsys.buffer_unmap = aub_buffer_unmap; iws->winsys.buffer_destroy = aub_buffer_destroy; iws->winsys.flush_frontbuffer = aub_flush_frontbuffer; - iws->winsys.printf = aub_printf; iws->winsys.get_name = aub_get_name; iws->winsys.surface_alloc = aub_i915_surface_alloc; diff --git a/src/mesa/drivers/x11/xm_winsys.c b/src/mesa/drivers/x11/xm_winsys.c index 2edc697693..eab9fd3852 100644 --- a/src/mesa/drivers/x11/xm_winsys.c +++ b/src/mesa/drivers/x11/xm_winsys.c @@ -193,15 +193,6 @@ xm_wait_idle(struct pipe_winsys *pws) /* no-op */ } -static void -xm_printf(struct pipe_winsys *pws, const char *fmtString, ...) -{ - va_list args; - va_start( args, fmtString ); - vfprintf(stderr, fmtString, args); - va_end( args ); -} - static const char * xm_get_name(struct pipe_winsys *pws) { @@ -353,7 +344,6 @@ xmesa_create_pipe_winsys( XMesaContext xmesa ) xws->winsys.flush_frontbuffer = xm_flush_frontbuffer; xws->winsys.wait_idle = xm_wait_idle; - xws->winsys.printf = xm_printf; xws->winsys.get_name = xm_get_name; xws->xmesa = xmesa; -- cgit v1.2.3 From e66cb602b79f52bd1f01b1c3f7af039eeff283a6 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 12:18:20 -0600 Subject: gallium: turn on clipping for bitmaps Bitmaps can extend beyond window edges so we need to clip. Also, move some state atom vars to st_context to be a bit more efficient. --- src/mesa/state_tracker/st_cb_bitmap.c | 94 +++++++++++++++++++++-------------- src/mesa/state_tracker/st_context.h | 2 + 2 files changed, 60 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index ec56b25f7c..0cc910a67f 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -350,16 +350,21 @@ setup_bitmap_vertex_data(struct st_context *st, { struct pipe_context *pipe = st->pipe; const struct gl_framebuffer *fb = st->ctx->DrawBuffer; - const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP); + const GLfloat fb_width = fb->Width; + const GLfloat fb_height = fb->Height; const GLfloat x0 = x; const GLfloat x1 = x + width; - const GLfloat y0 = invert ? ((int) fb->Height - y - height) : y; - const GLfloat y1 = invert ? (y0 + height) : y + height; + const GLfloat y0 = y; + const GLfloat y1 = y + height; const GLfloat bias = st->bitmap_texcoord_bias; const GLfloat xBias = bias / (x1-x0); const GLfloat yBias = bias / (y1-y0); const GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias; - const GLfloat tTop = 1.0 - yBias, tBot = 1.0 - tTop - yBias; + const GLfloat tTop = yBias, tBot = 1.0 - tTop - yBias; + const GLfloat clip_x0 = x0 / fb_width * 2.0 - 1.0; + const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; + const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; + const GLfloat clip_y1 = y1 / fb_height * 2.0 - 1.0; GLuint i; void *buf; @@ -369,24 +374,26 @@ setup_bitmap_vertex_data(struct st_context *st, sizeof(st->bitmap.vertices)); } - /* positions, texcoords */ - st->bitmap.vertices[0][0][0] = x0; - st->bitmap.vertices[0][0][1] = y0; + /* Positions are in clip coords since we need to do clipping in case + * the bitmap quad goes beyond the window bounds. + */ + st->bitmap.vertices[0][0][0] = clip_x0; + st->bitmap.vertices[0][0][1] = clip_y0; st->bitmap.vertices[0][2][0] = sLeft; st->bitmap.vertices[0][2][1] = tTop; - st->bitmap.vertices[1][0][0] = x1; - st->bitmap.vertices[1][0][1] = y0; + st->bitmap.vertices[1][0][0] = clip_x1; + st->bitmap.vertices[1][0][1] = clip_y0; st->bitmap.vertices[1][2][0] = sRight; st->bitmap.vertices[1][2][1] = tTop; - st->bitmap.vertices[2][0][0] = x1; - st->bitmap.vertices[2][0][1] = y1; + st->bitmap.vertices[2][0][0] = clip_x1; + st->bitmap.vertices[2][0][1] = clip_y1; st->bitmap.vertices[2][2][0] = sRight; st->bitmap.vertices[2][2][1] = tBot; - st->bitmap.vertices[3][0][0] = x0; - st->bitmap.vertices[3][0][1] = y1; + st->bitmap.vertices[3][0][0] = clip_x0; + st->bitmap.vertices[3][0][1] = clip_y1; st->bitmap.vertices[3][2][0] = sLeft; st->bitmap.vertices[3][2][1] = tBot; @@ -437,17 +444,11 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_save_rasterizer(cso); cso_save_samplers(cso); + cso_save_viewport(cso); /* rasterizer state: just scissor */ - { - struct pipe_rasterizer_state rasterizer; - memset(&rasterizer, 0, sizeof(rasterizer)); - if (ctx->Scissor.Enabled) - rasterizer.scissor = 1; - rasterizer.bypass_clipping = 1; - - cso_set_rasterizer(cso, &rasterizer); - } + st->bitmap.rasterizer.scissor = ctx->Scissor.Enabled; + cso_set_rasterizer(cso, &st->bitmap.rasterizer); /* fragment shader state: TEX lookup program */ pipe->bind_fs_state(pipe, stfp->driver_shader); @@ -456,21 +457,26 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, pipe->bind_vs_state(pipe, st->bitmap.vs); /* sampler / texture state */ + cso_single_sampler(cso, 0, &st->bitmap.sampler); + cso_single_sampler_done(cso); + pipe->set_sampler_textures(pipe, 1, &pt); + + /* viewport state: viewport matching window dims */ { - struct pipe_sampler_state sampler; - memset(&sampler, 0, sizeof(sampler)); - sampler.wrap_s = PIPE_TEX_WRAP_CLAMP; - sampler.wrap_t = PIPE_TEX_WRAP_CLAMP; - sampler.wrap_r = PIPE_TEX_WRAP_CLAMP; - sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; - sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; - sampler.normalized_coords = 1; - - cso_single_sampler(cso, 0, &sampler); - cso_single_sampler_done(cso); - - pipe->set_sampler_textures(pipe, 1, &pt); + const struct gl_framebuffer *fb = st->ctx->DrawBuffer; + const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP); + const float width = fb->Width; + const float height = fb->Height; + struct pipe_viewport_state vp; + vp.scale[0] = 0.5 * width; + vp.scale[1] = height * (invert ? -0.5 : 0.5); + vp.scale[2] = 1.0; + vp.scale[3] = 1.0; + vp.translate[0] = 0.5 * width; + vp.translate[1] = 0.5 * height; + vp.translate[2] = 0.0; + vp.translate[3] = 0.0; + cso_set_viewport(cso, &vp); } /* draw textured quad */ @@ -487,6 +493,7 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* restore state */ cso_restore_rasterizer(cso); cso_restore_samplers(cso); + cso_restore_viewport(cso); /* shaders don't go through cso yet */ pipe->bind_fs_state(pipe, st->fp->driver_shader); pipe->bind_vs_state(pipe, st->vp->driver_shader); @@ -742,6 +749,21 @@ st_init_bitmap_functions(struct dd_function_table *functions) void st_init_bitmap(struct st_context *st) { + struct pipe_sampler_state *sampler = &st->bitmap.sampler; + + /* init sampler state once */ + memset(sampler, 0, sizeof(*sampler)); + sampler->wrap_s = PIPE_TEX_WRAP_CLAMP; + sampler->wrap_t = PIPE_TEX_WRAP_CLAMP; + sampler->wrap_r = PIPE_TEX_WRAP_CLAMP; + sampler->min_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler->min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler->mag_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler->normalized_coords = 1; + + /* init scissor state once */ + memset(&st->bitmap.rasterizer, 0, sizeof(st->bitmap.rasterizer)); + init_bitmap_cache(st); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 85e3d47e1a..44705bc89a 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -148,6 +148,8 @@ struct st_context struct st_fragment_program *program; /**< bitmap tex/kil program */ GLuint user_prog_sn; /**< user fragment program serial no. */ struct st_fragment_program *combined_prog; + struct pipe_rasterizer_state rasterizer; + struct pipe_sampler_state sampler; struct pipe_shader_state vert_shader; void *vs; float vertices[4][3][4]; /**< vertex pos + color + texcoord */ -- cgit v1.2.3 From 499d8aaa476fb67b7355122dc6fbc641e1b44ed0 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 14:06:42 -0600 Subject: gallium: draw_passthrough.c is not used anymore --- src/gallium/auxiliary/draw/Makefile | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 0c7ce5da5b..a0db2e4555 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -15,7 +15,6 @@ C_SOURCES = \ draw_debug.c \ draw_flatshade.c \ draw_offset.c \ - draw_passthrough.c \ draw_pt.c \ draw_pt_vcache.c \ draw_pt_fetch_emit.c \ -- cgit v1.2.3 From 9cbd8400433fb27da03f300b36495baef464cc6b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 14:13:09 -0600 Subject: gallium: draw_passthrough.c is not used anymore --- src/gallium/auxiliary/draw/SConscript | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 9b3e7247c5..981225a8c2 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -14,7 +14,6 @@ draw = env.ConvenienceLibrary( 'draw_debug.c', 'draw_flatshade.c', 'draw_offset.c', - 'draw_passthrough.c', # going away soon 'draw_pt.c', 'draw_pt_vcache.c', 'draw_pt_fetch_emit.c', -- cgit v1.2.3 From 594dab4769533afaeb30a588e1731a6753a93f0d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 14:14:30 -0600 Subject: gallium: move the test for bypass_vs into the vs_XXX_run() functions Also: 1. Added an identity_viewport flag to skip viewport transformation when it has no effect. Might also add an explicit bypass_viewport flag someday. 2. Separate the code for computing clip codes and doing the viewport transform. Predicate them separately. Note: even if bypass_vs is set, we still look at the shader to determine the number of inputs and outputs. --- src/gallium/auxiliary/draw/draw_context.c | 8 ++++++ src/gallium/auxiliary/draw/draw_prim.c | 3 +-- src/gallium/auxiliary/draw/draw_private.h | 2 ++ src/gallium/auxiliary/draw/draw_vs_exec.c | 24 ++++++++++++------ src/gallium/auxiliary/draw/draw_vs_llvm.c | 42 ++++++++++++++++++++----------- src/gallium/auxiliary/draw/draw_vs_sse.c | 32 +++++++++++++++-------- 6 files changed, 76 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 10bf9f54c1..d0d5f66b37 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -228,6 +228,14 @@ void draw_set_viewport_state( struct draw_context *draw, { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); draw->viewport = *viewport; /* struct copy */ + draw->identity_viewport = (viewport->scale[0] == 1.0f && + viewport->scale[1] == 1.0f && + viewport->scale[2] == 1.0f && + viewport->scale[3] == 1.0f && + viewport->translate[0] == 0.0f && + viewport->translate[1] == 0.0f && + viewport->translate[2] == 0.0f && + viewport->translate[3] == 0.0f); } diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index ddcde01d9a..9779aa8440 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -593,8 +593,7 @@ draw_arrays(struct draw_context *draw, unsigned prim, } /* drawing done here: */ - if (!draw->rasterizer->bypass_vs || - !draw_pt_arrays(draw, prim, start, count)) { + if (!draw_pt_arrays(draw, prim, start, count)) { /* we have to run the whole pipeline */ draw_prim(draw, prim, start, count); } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 8eb2f515cb..9a9b25297f 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -232,6 +232,8 @@ struct draw_context struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; struct draw_vertex_shader *vertex_shader; + boolean identity_viewport; + uint num_vs_outputs; /**< convenience, from vertex_shader */ /* user-space vertex data, buffers */ diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 487d0ea7f4..c8ed17c00a 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -110,13 +110,20 @@ vs_exec_run( struct draw_vertex_shader *shader, machine->Consts = (float (*)[4]) draw->user.constants; machine->Inputs = ALIGN16_ASSIGN(inputs); - machine->Outputs = ALIGN16_ASSIGN(outputs); + if (draw->rasterizer->bypass_vs) { + /* outputs are just the inputs */ + machine->Outputs = machine->Inputs; + } + else { + machine->Outputs = ALIGN16_ASSIGN(outputs); + } draw->vertex_fetch.fetch_func( draw, machine, elts, count ); - /* run interpreter */ - tgsi_exec_machine_run( machine ); - + if (!draw->rasterizer->bypass_vs) { + /* run interpreter */ + tgsi_exec_machine_run( machine ); + } /* store machine results */ for (j = 0; j < count; j++) { @@ -136,8 +143,13 @@ vs_exec_run( struct draw_vertex_shader *shader, if (!draw->rasterizer->bypass_clipping) { vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; + } + else { + vOut[j]->clipmask = 0; + } + vOut[j]->edgeflag = 1; + if (!draw->identity_viewport) { /* divide by w */ w = 1.0f / w; x *= w; @@ -151,8 +163,6 @@ vs_exec_run( struct draw_vertex_shader *shader, vOut[j]->data[0][3] = w; } else { - vOut[j]->clipmask = 0; - vOut[j]->edgeflag = 1; vOut[j]->data[0][0] = x; vOut[j]->data[0][1] = y; vOut[j]->data[0][2] = z; diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index d29cb18efe..8aa8a617bb 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -121,31 +121,45 @@ vs_llvm_run( struct draw_vertex_shader *base, machine->Consts = (float (*)[4]) draw->user.constants; machine->Inputs = ALIGN16_ASSIGN(inputs); - machine->Outputs = ALIGN16_ASSIGN(outputs); + if (draw->rasterizer->bypass_vs) { + /* outputs are just the inputs */ + machine->Outputs = machine->Inputs; + } + else { + machine->Outputs = ALIGN16_ASSIGN(outputs); + } + draw->vertex_fetch.fetch_func( draw, machine, elts, count ); - /* run shader */ - gallivm_cpu_vs_exec(shader->llvm_prog, - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps); + if (!draw->rasterizer->bypass_vs) { + /* run shader */ + gallivm_cpu_vs_exec(shader->llvm_prog, + machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps); + } /* store machine results */ for (j = 0; j < count; j++) { unsigned slot; float x, y, z, w; - if (!draw->rasterizer->bypass_clipping) { - x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + if (!draw->rasterizer->bypass_clipping) { vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; + } + else { + vOut[j]->clipmask = 0; + } + vOut[j]->edgeflag = 1; + if (!draw->identity_viewport) { /* divide by w */ w = 1.0f / w; x *= w; @@ -159,8 +173,6 @@ vs_llvm_run( struct draw_vertex_shader *base, vOut[j]->data[0][3] = w; } else { - vOut[j]->clipmask = 0; - vOut[j]->edgeflag = 1; vOut[j]->data[0][0] = x; vOut[j]->data[0][1] = y; vOut[j]->data[0][2] = z; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index bc910dc2d0..701137f908 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -126,7 +126,13 @@ vs_sse_run( struct draw_vertex_shader *base, /* Consts does not require 16 byte alignment. */ machine->Consts = (float (*)[4]) draw->user.constants; machine->Inputs = ALIGN16_ASSIGN(inputs); - machine->Outputs = ALIGN16_ASSIGN(outputs); + if (draw->rasterizer->bypass_vs) { + /* outputs are just the inputs */ + machine->Outputs = machine->Inputs; + } + else { + machine->Outputs = ALIGN16_ASSIGN(outputs); + } /* Fetch vertices. This may at some point be integrated into the @@ -137,13 +143,14 @@ vs_sse_run( struct draw_vertex_shader *base, draw->vertex_fetch.fetch_func( draw, machine, elts, count ); - /* run compiled shader - */ - shader->func( - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps ); + if (!draw->rasterizer->bypass_vs) { + /* run compiled shader + */ + shader->func(machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps ); + } /* XXX: Computing the clipmask and emitting results should be done @@ -161,8 +168,13 @@ vs_sse_run( struct draw_vertex_shader *base, if (!draw->rasterizer->bypass_clipping) { vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - vOut[j]->edgeflag = 1; + } + else { + vOut[j]->clipmask = 0; + } + vOut[j]->edgeflag = 1; + if (!draw->identity_viewport) { /* divide by w */ w = 1.0f / w; x *= w; @@ -176,8 +188,6 @@ vs_sse_run( struct draw_vertex_shader *base, vOut[j]->data[0][3] = w; } else { - vOut[j]->clipmask = 0; - vOut[j]->edgeflag = 1; vOut[j]->data[0][0] = x; vOut[j]->data[0][1] = y; vOut[j]->data[0][2] = z; -- cgit v1.2.3 From 6a382250a10e7022e93b6be6fe4657445d85a568 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 14:18:36 -0600 Subject: gallium: set the bypass_vs flag now The glBitmap vertex shader is a no-op, but we still have to specify it in order to convey the number of inputs/outputs. --- src/mesa/state_tracker/st_cb_bitmap.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 0cc910a67f..3f0d354e94 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -761,8 +761,9 @@ st_init_bitmap(struct st_context *st) sampler->mag_img_filter = PIPE_TEX_FILTER_NEAREST; sampler->normalized_coords = 1; - /* init scissor state once */ + /* init baseline rasterizer state once */ memset(&st->bitmap.rasterizer, 0, sizeof(st->bitmap.rasterizer)); + st->bitmap.rasterizer.bypass_vs = 1; init_bitmap_cache(st); } -- cgit v1.2.3 From 7139b8ef78adb8d08c13e439fc8add31a2d79f36 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 14:20:16 -0600 Subject: gallium: draw_passthrough.c is obsolete - removed --- src/gallium/auxiliary/draw/draw_passthrough.c | 473 -------------------------- 1 file changed, 473 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_passthrough.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_passthrough.c b/src/gallium/auxiliary/draw/draw_passthrough.c deleted file mode 100644 index 2198079a88..0000000000 --- a/src/gallium/auxiliary/draw/draw_passthrough.c +++ /dev/null @@ -1,473 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -/* This code is a prototype of what a passhthrough vertex shader might - * look like. - * - * Probably the best approach for us is to do: - * - vertex fetch - * - vertex shader - * - cliptest / viewport transform - * - * in one step, then examine the clipOrMask & choose between two paths: - * - * Either: - * - build primitive headers - * - clip and the primitive path - * - build clipped vertex buffers, - * - vertex-emit to vbuf buffers - * - * Or, if no clipping: - * - vertex-emit directly to vbuf buffers - * - * But when bypass clipping is enabled, we just take the latter - * choice. If (some new) passthrough-vertex-shader flag is also set, - * the pipeline degenerates to: - * - * - vertex fetch - * - vertex emit to vbuf buffers - * - * Which is what is prototyped here. - */ -#include "pipe/p_util.h" -#include "draw/draw_context.h" -#include "draw/draw_private.h" -#include "draw/draw_vbuf.h" -#include "draw/draw_vertex.h" - - -/** - * General-purpose fetch from user's vertex arrays, emit to driver's - * vertex buffer. - * - * XXX this is totally temporary. - */ -static void -fetch_store_general( struct draw_context *draw, - float *out, - unsigned start, - unsigned count ) -{ - const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); - const unsigned nr_attrs = vinfo->num_attribs; - uint i, j; - - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - - for (i = start; i < start + count; i++) { - for (j = 0; j < nr_attrs; j++) { - /* vinfo->src_index is the output of the vertex shader - * matching this hw-vertex component. - * - * In passthrough, we require a 1:1 mapping between vertex - * shader outputs and inputs, which in turn correspond to - * vertex elements in the state. So, this is the vertex - * element we're interested in... - */ - const uint jj = vinfo->src_index[j]; - const enum pipe_format srcFormat = draw->vertex_element[jj].src_format; - const ubyte *from = src[jj] + i * pitch[jj]; - float attrib[4]; - - /* Except... When we're not. Two cases EMIT_HEADER & - * EMIT_1F_PSIZE don't consume an input. Should have some - * method for indicating this, or change the logic here - * somewhat so it doesn't matter. - * - * Just hack this up now, do something better about it later. - */ - if (vinfo->emit[j] == EMIT_HEADER) { - memset(out, 0, sizeof(struct vertex_header)); - out += sizeof(struct vertex_header) / 4; - continue; - } - else if (vinfo->emit[j] == EMIT_1F_PSIZE) { - out[0] = 1.0; /* xxx */ - out += 1; - continue; - } - - - /* The normal fetch/emit code: - */ - switch (srcFormat) { - case PIPE_FORMAT_B8G8R8A8_UNORM: - { - ubyte *ub = (ubyte *) from; - attrib[2] = UBYTE_TO_FLOAT(ub[0]); - attrib[1] = UBYTE_TO_FLOAT(ub[1]); - attrib[0] = UBYTE_TO_FLOAT(ub[2]); - attrib[3] = UBYTE_TO_FLOAT(ub[3]); - } - break; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - { - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = f[1]; - attrib[2] = f[2]; - attrib[3] = f[3]; - } - break; - case PIPE_FORMAT_R32G32B32_FLOAT: - { - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = f[1]; - attrib[2] = f[2]; - attrib[3] = 1.0; - } - break; - case PIPE_FORMAT_R32G32_FLOAT: - { - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = f[1]; - attrib[2] = 0.0; - attrib[3] = 1.0; - } - break; - case PIPE_FORMAT_R32_FLOAT: - { - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = 0.0; - attrib[2] = 0.0; - attrib[3] = 1.0; - } - break; - default: - assert(0); - } - - debug_printf("attrib %d: %f %f %f %f\n", j, - attrib[0], attrib[1], attrib[2], attrib[3]); - - switch (vinfo->emit[j]) { - case EMIT_1F: - out[0] = attrib[0]; - out += 1; - break; - case EMIT_2F: - out[0] = attrib[0]; - out[1] = attrib[1]; - out += 2; - break; - case EMIT_4F: - out[0] = attrib[0]; - out[1] = attrib[1]; - out[2] = attrib[2]; - out[3] = attrib[3]; - out += 4; - break; - default: - assert(0); - } - } - debug_printf("\n"); - } -} - - - -static boolean update_shader( struct draw_context *draw ) -{ - const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); - - unsigned nr_attrs = vinfo->num_attribs; - unsigned i; - - for (i = 0; i < nr_attrs; i++) { - unsigned buf = draw->vertex_element[i].vertex_buffer_index; - - draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + - draw->vertex_buffer[buf].buffer_offset + - draw->vertex_element[i].src_offset; - - draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; - draw->vertex_fetch.fetch[i] = NULL; - } - - draw->vertex_fetch.nr_attrs = nr_attrs; - draw->vertex_fetch.fetch_func = NULL; - draw->vertex_fetch.pt_fetch = NULL; - - draw->pt.hw_vertex_size = vinfo->size * 4; - - draw->vertex_fetch.pt_fetch = fetch_store_general; - return TRUE; -} - - - - -static boolean split_prim_inplace(unsigned prim, unsigned *first, unsigned *incr) -{ - switch (prim) { - case PIPE_PRIM_POINTS: - *first = 1; - *incr = 1; - return TRUE; - case PIPE_PRIM_LINES: - *first = 2; - *incr = 2; - return TRUE; - case PIPE_PRIM_LINE_STRIP: - *first = 2; - *incr = 1; - return TRUE; - case PIPE_PRIM_TRIANGLES: - *first = 3; - *incr = 3; - return TRUE; - case PIPE_PRIM_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - return TRUE; - case PIPE_PRIM_QUADS: - *first = 4; - *incr = 4; - return TRUE; - case PIPE_PRIM_QUAD_STRIP: - *first = 4; - *incr = 2; - return TRUE; - default: - *first = 0; - *incr = 1; /* set to one so that count % incr works */ - return FALSE; - } -} - - - -static boolean set_prim( struct draw_context *draw, - unsigned prim, - unsigned count ) -{ - assert(!draw->user.elts); - - switch (prim) { - case PIPE_PRIM_LINE_LOOP: - if (count > 1024) - return FALSE; - return draw->render->set_primitive( draw->render, PIPE_PRIM_LINE_STRIP ); - - case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: - if (count > 1024) - return FALSE; - return draw->render->set_primitive( draw->render, prim ); - - case PIPE_PRIM_QUADS: - case PIPE_PRIM_QUAD_STRIP: - return draw->render->set_primitive( draw->render, PIPE_PRIM_TRIANGLES ); - - default: - return draw->render->set_primitive( draw->render, prim ); - break; - } - - return TRUE; -} - - - -#define INDEX(i) (start + (i)) -static void pt_draw_arrays( struct draw_context *draw, - unsigned start, - unsigned length ) -{ - ushort *tmp = NULL; - unsigned i, j; - - switch (draw->pt.prim) { - case PIPE_PRIM_LINE_LOOP: - tmp = MALLOC( sizeof(ushort) * (length + 1) ); - - for (i = 0; i < length; i++) - tmp[i] = INDEX(i); - tmp[length] = 0; - - draw->render->draw( draw->render, - tmp, - length+1 ); - break; - - - case PIPE_PRIM_QUAD_STRIP: - tmp = MALLOC( sizeof(ushort) * (length / 2 * 6) ); - - for (j = i = 0; i + 3 < length; i += 2, j += 6) { - tmp[j+0] = INDEX(i+0); - tmp[j+1] = INDEX(i+1); - tmp[j+2] = INDEX(i+3); - - tmp[j+3] = INDEX(i+2); - tmp[j+4] = INDEX(i+0); - tmp[j+5] = INDEX(i+3); - } - - if (j) - draw->render->draw( draw->render, tmp, j ); - break; - - case PIPE_PRIM_QUADS: - tmp = MALLOC( sizeof(int) * (length / 4 * 6) ); - - for (j = i = 0; i + 3 < length; i += 4, j += 6) { - tmp[j+0] = INDEX(i+0); - tmp[j+1] = INDEX(i+1); - tmp[j+2] = INDEX(i+3); - - tmp[j+3] = INDEX(i+1); - tmp[j+4] = INDEX(i+2); - tmp[j+5] = INDEX(i+3); - } - - if (j) - draw->render->draw( draw->render, tmp, j ); - break; - - default: - draw->render->draw_arrays( draw->render, - start, - length ); - break; - } - - if (tmp) - FREE(tmp); -} - - - -static boolean do_draw( struct draw_context *draw, - unsigned start, unsigned count ) -{ - float *hw_verts = - draw->render->allocate_vertices( draw->render, - (ushort)draw->pt.hw_vertex_size, - (ushort)count ); - - if (!hw_verts) - return FALSE; - - /* Single routine to fetch vertices and emit HW verts. - */ - draw->vertex_fetch.pt_fetch( draw, - hw_verts, - start, count ); - - /* Draw arrays path to avoid re-emitting index list again and - * again. - */ - pt_draw_arrays( draw, - 0, - count ); - - - draw->render->release_vertices( draw->render, - hw_verts, - draw->pt.hw_vertex_size, - count ); - - return TRUE; -} - - -boolean -draw_passthrough_arrays(struct draw_context *draw, - unsigned prim, - unsigned start, - unsigned count) -{ - unsigned i = 0; - unsigned first, incr; - - //debug_printf("%s prim %d start %d count %d\n", __FUNCTION__, prim, start, count); - - split_prim_inplace(prim, &first, &incr); - - count -= (count - first) % incr; - - debug_printf("%s %d %d %d\n", __FUNCTION__, prim, start, count); - - if (draw_need_pipeline(draw, prim)) - return FALSE; - - debug_printf("%s AAA\n", __FUNCTION__); - - if (!set_prim(draw, prim, count)) - return FALSE; - - /* XXX: need a single value that reflects the most recent call to - * driver->set_primitive: - */ - draw->pt.prim = prim; - - debug_printf("%s BBB\n", __FUNCTION__); - - if (!update_shader(draw)) - return FALSE; - - debug_printf("%s CCC\n", __FUNCTION__); - - /* Chop this up into bite-sized pieces that a driver should be able - * to devour -- problem is we don't have a quick way to query the - * driver on the maximum size for this chunk in the current state. - */ - while (i + first <= count) { - int nr = MIN2( count - i, 1024 ); - - /* snap to prim boundary - */ - nr -= (nr - first) % incr; - - if (!do_draw( draw, start + i, nr )) { - assert(0); - return FALSE; - } - - /* increment allowing for repeated vertices - */ - i += nr - (first - incr); - } - - - debug_printf("%s DONE\n", __FUNCTION__); - return TRUE; -} - - -- cgit v1.2.3 From ddc0f91bfc571dcd05c13f094c179b4250e53bcc Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 14:38:27 -0600 Subject: gallium: use cso_save/restore_sampler_textures() functions --- src/mesa/state_tracker/st_cb_bitmap.c | 4 ++-- src/mesa/state_tracker/st_cb_drawpixels.c | 14 +++++--------- 2 files changed, 7 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 3f0d354e94..8feae73fe0 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -444,6 +444,7 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_save_rasterizer(cso); cso_save_samplers(cso); + cso_save_sampler_textures(cso); cso_save_viewport(cso); /* rasterizer state: just scissor */ @@ -493,12 +494,11 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* restore state */ cso_restore_rasterizer(cso); cso_restore_samplers(cso); + cso_restore_sampler_textures(cso); cso_restore_viewport(cso); /* shaders don't go through cso yet */ pipe->bind_fs_state(pipe, st->fp->driver_shader); pipe->bind_vs_state(pipe, st->vp->driver_shader); - pipe->set_sampler_textures(pipe, ctx->st->state.num_textures, - ctx->st->state.sampler_texture); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 43cc21d1fb..c181575f15 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -530,14 +530,13 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_save_rasterizer(cso); cso_save_viewport(cso); cso_save_samplers(cso); + cso_save_sampler_textures(cso); /* rasterizer state: just scissor */ { struct pipe_rasterizer_state rasterizer; memset(&rasterizer, 0, sizeof(rasterizer)); - if (ctx->Scissor.Enabled) - rasterizer.scissor = 1; - + rasterizer.scissor = ctx->Scissor.Enabled; cso_set_rasterizer(cso, &rasterizer); } @@ -581,9 +580,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, } /* texture state: */ - { - pipe->set_sampler_textures(pipe, 1, &pt); - } + pipe->set_sampler_textures(pipe, 1, &pt); /* Compute window coords (y=0=bottom) with pixel zoom. * Recall that these coords are transformed by the current @@ -604,12 +601,11 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_restore_rasterizer(cso); cso_restore_viewport(cso); cso_restore_samplers(cso); + cso_restore_sampler_textures(cso); + /* shaders don't go through cso yet */ pipe->bind_fs_state(pipe, st->fp->driver_shader); pipe->bind_vs_state(pipe, st->vp->driver_shader); - - pipe->set_sampler_textures(pipe, ctx->st->state.num_textures, - ctx->st->state.sampler_texture); } -- cgit v1.2.3 From 9136747a2b232ae7023972e6611cb2c845a86886 Mon Sep 17 00:00:00 2001 From: Markus Amsler Date: Mon, 31 Mar 2008 14:58:11 -0600 Subject: fix parsing bug involving comments at the end of ARB v/f programs --- src/mesa/shader/arbprogram.syn | 6 +++--- src/mesa/shader/arbprogram_syn.h | 6 +++--- 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogram.syn b/src/mesa/shader/arbprogram.syn index 1746a876c3..ba17d638a3 100644 --- a/src/mesa/shader/arbprogram.syn +++ b/src/mesa/shader/arbprogram.syn @@ -2643,14 +2643,14 @@ white_char ' ' .or '\t' .or '\n' .or '\r'; comment_block - '#' .and .loop comment_char .and new_line; + '#' .and .loop comment_char .and optional_new_line; /* All ASCII characters except '\r', '\n' and '\0' */ comment_char '\x0E'-'\xFF' .or '\x01'-'\x09' .or '\x0B'-'\x0C'; -new_line - '\n' .or crlf .or '\0'; +optional_new_line + '\n' .or crlf .or .true; crlf '\r' .and '\n'; diff --git a/src/mesa/shader/arbprogram_syn.h b/src/mesa/shader/arbprogram_syn.h index 5f3f7d6cf4..e0c901ea8c 100644 --- a/src/mesa/shader/arbprogram_syn.h +++ b/src/mesa/shader/arbprogram_syn.h @@ -1242,11 +1242,11 @@ "white_char\n" " ' ' .or '\\t' .or '\\n' .or '\\r';\n" "comment_block\n" -" '#' .and .loop comment_char .and new_line;\n" +" '#' .and .loop comment_char .and optional_new_line;\n" "comment_char\n" " '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n" -"new_line\n" -" '\\n' .or crlf .or '\\0';\n" +"optional_new_line\n" +" '\\n' .or crlf .or .true;\n" "crlf\n" " '\\r' .and '\\n';\n" "semicolon\n" -- cgit v1.2.3 From 23b03c536daa47b53e585fa98476bad96eb73529 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 15:12:01 -0600 Subject: gallium: updated comment for bypass_vs --- src/gallium/include/pipe/p_state.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 0eeee47a9a..a73028814e 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -111,7 +111,8 @@ struct pipe_rasterizer_state unsigned line_stipple_pattern:16; unsigned line_last_pixel:1; unsigned bypass_clipping:1; - unsigned bypass_vs:1; /**< vertices are already fully transformed */ + unsigned bypass_vs:1; /**< Skip the vertex shader. Note that the shader is + still needed though, to indicate inputs/outputs */ unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ unsigned flatshade_first:1; /**< take color attribute from the first vertex of a primitive */ -- cgit v1.2.3 From 6b0be38a6cfecd2babc7175b0877bed522ca9f72 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 15:12:14 -0600 Subject: gallium: used inverted bitmap texture to simplify the fragment shader. "Off" bits get stored as texel=0xff and "on" bits get stored as texel=0x0. Then use KIL -tmp to kill the off bits and keep the on bits. This shortens the fragment program by two instructions. --- src/mesa/state_tracker/st_cb_bitmap.c | 49 ++++++++++++++--------------------- 1 file changed, 20 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 8feae73fe0..975a55d6dc 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -59,6 +59,19 @@ +/** + * glBitmaps are drawn as textured quads. The user's bitmap pattern + * is stored in a texture image. An alpha8 texture format is used. + * The fragment shader samples a bit (texel) from the texture, then + * discards the fragment if the bit is off. + * + * Note that we actually store the inverse image of the bitmap to + * simplify the fragment program. An "on" bit gets stored as texel=0x0 + * and an "off" bit is stored as texel=0xff. Then we kill the + * fragment if the negated texel value is less than zero. + */ + + /** * The bitmap cache attempts to accumulate multiple glBitmap calls in a * buffer which is then rendered en mass upon a flush, state change, etc. @@ -102,7 +115,7 @@ make_bitmap_fragment_program(GLcontext *ctx) if (!p) return NULL; - p->NumInstructions = 5; + p->NumInstructions = 3; p->Instructions = _mesa_alloc_instructions(p->NumInstructions); if (!p->Instructions) { @@ -121,33 +134,11 @@ make_bitmap_fragment_program(GLcontext *ctx) p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; ic++; - /* SWZ tmp0.x, tmp0.x, 1111; # tmp0.x = 1.0 */ - p->Instructions[ic].Opcode = OPCODE_SWZ; - p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; - p->Instructions[ic].DstReg.Index = 0; - p->Instructions[ic].DstReg.WriteMask = WRITEMASK_X; - p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[0].Index = 0; - p->Instructions[ic].SrcReg[0].Swizzle - = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE ); - ic++; - - /* SUB tmp0, tmp0.wwww, tmp0.xxxx; # tmp0.w -= 1 */ - p->Instructions[ic].Opcode = OPCODE_SUB; - p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY; - p->Instructions[ic].DstReg.Index = 0; - p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[0].Index = 0; - p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW; - p->Instructions[ic].SrcReg[1].File = PROGRAM_TEMPORARY; - p->Instructions[ic].SrcReg[1].Index = 0; - p->Instructions[ic].SrcReg[1].Swizzle = SWIZZLE_XXXX; /* 1.0 */ - ic++; - - /* KIL if tmp0 < 0 */ + /* KIL if -tmp0 < 0 # texel=0 -> keep / texel=0 -> discard */ p->Instructions[ic].Opcode = OPCODE_KIL; p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY; p->Instructions[ic].SrcReg[0].Index = 0; + p->Instructions[ic].SrcReg[0].NegateBase = NEGATE_XYZW; ic++; /* END; */ @@ -289,7 +280,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, for (col = 0; col < width; col++) { /* set texel to 255 if bit is set */ - destRow[comp] = (*src & mask) ? 255 : 0; + destRow[comp] = (*src & mask) ? 0x0 : 0xff; destRow += cpp; if (mask == 128U) { @@ -311,7 +302,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, for (col = 0; col < width; col++) { /* set texel to 255 if bit is set */ - destRow[comp] =(*src & mask) ? 255 : 0; + destRow[comp] =(*src & mask) ? 0x0 : 0xff; destRow += cpp; if (mask == 1U) { @@ -505,7 +496,7 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, static void reset_cache(struct st_context *st) { - memset(st->bitmap.cache->buffer, 0, sizeof(st->bitmap.cache->buffer)); + memset(st->bitmap.cache->buffer, 0xff, sizeof(st->bitmap.cache->buffer)); st->bitmap.cache->empty = GL_TRUE; st->bitmap.cache->xmin = 1000000; @@ -642,7 +633,7 @@ accum_bitmap(struct st_context *st, /* XXX try to combine this code with code in make_bitmap_texture() */ #define SET_PIXEL(COL, ROW) \ - cache->buffer[py + (ROW)][px + (COL)] = 0xff; + cache->buffer[py + (ROW)][px + (COL)] = 0x0; for (row = 0; row < height; row++) { const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, -- cgit v1.2.3 From 8a81429fa866cc1e2a14dec7e888fd2c888dc40a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 1 Apr 2008 07:22:10 +0900 Subject: gallium: Compute YCBCR bit depth. --- src/gallium/include/pipe/p_format.h | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 3238b152b6..9e0f91f202 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -439,7 +439,8 @@ static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) */ static INLINE uint pf_get_bits( enum pipe_format format ) { - if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) { + switch (pf_layout(format)) { + case PIPE_FORMAT_LAYOUT_RGBAZS: return pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) + pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) + @@ -447,11 +448,11 @@ static INLINE uint pf_get_bits( enum pipe_format format ) pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) + pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + pf_get_component_bits( format, PIPE_FORMAT_COMP_S ); - } - else { - assert( pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR ); - - /* TODO */ + case PIPE_FORMAT_LAYOUT_YCBCR: + assert( format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV ); + /* return effective bits per pixel */ + return 16; + default: assert( 0 ); return 0; } -- cgit v1.2.3 From ab8bcc4ec626be2d09bcdbaba2d1030b8dac7e25 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 16:35:13 -0600 Subject: cell: implement logicop/output for PIPE_FORMAT_B8G8R8A8_UNORM Remote display to my usual terminal shows the right colors again. Not 100% sure about the shuffle control words, but they seem to work. --- .../drivers/cell/ppu/cell_state_per_fragment.c | 36 ++++++++++++++++------ 1 file changed, 26 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c index f10025bd7c..0a79cccc83 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -1164,7 +1164,7 @@ int PC_OFFSET(const struct spe_function *f, const void *d) * masking. * * \bug - * This routine is hard-coded to only work with ARGB8 data. + * Only two framebuffer formats are supported at this time. */ void cell_generate_logic_op(struct spe_function *f, struct pipe_blend_state *blend, @@ -1235,15 +1235,31 @@ cell_generate_logic_op(struct spe_function *f, struct pipe_blend_state *blend, /* Convert fragment colors to framebuffer format in AoS layout. */ - data[0] = 0x00010203; - data[1] = 0x10111213; - data[2] = 0x04050607; - data[3] = 0x14151617; - - data[4] = 0x0c000408; - data[5] = 0x80808080; - data[6] = 0x80808080; - data[7] = 0x80808080; + switch (surf->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + data[0] = 0x00010203; + data[1] = 0x10111213; + data[2] = 0x04050607; + data[3] = 0x14151617; + data[4] = 0x0c000408; + data[5] = 0x80808080; + data[6] = 0x80808080; + data[7] = 0x80808080; + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + data[0] = 0x03020100; + data[1] = 0x13121110; + data[2] = 0x07060504; + data[3] = 0x17161514; + data[4] = 0x0804000c; + data[5] = 0x80808080; + data[6] = 0x80808080; + data[7] = 0x80808080; + break; + default: + fprintf(stderr, "CELL: Bad pixel format in cell_generate_logic_op()"); + ASSERT(0); + } spe_ilh(f, tmp[0], 0x0808); spe_lqr(f, shuf_xpose_hi, PC_OFFSET(f, data+0)); -- cgit v1.2.3 From 58b6690cf84147f88ea2ba95d2a929089e93b57f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 16:44:56 -0600 Subject: cell: updated comments: s/test/SPE/ --- src/gallium/drivers/cell/common.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index b0928fefd2..c9e873b35c 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -109,7 +109,7 @@ */ struct cell_command_depth_stencil_alpha_test { uint64_t base; /**< Effective address of code start. */ - unsigned size; /**< Size in bytes of test code. */ + unsigned size; /**< Size in bytes of SPE code. */ unsigned read_depth; /**< Flag: should depth be read? */ unsigned read_stencil; /**< Flag: should stencil be read? */ }; @@ -120,14 +120,14 @@ struct cell_command_depth_stencil_alpha_test { */ struct cell_command_blend { uint64_t base; /**< Effective address of code start. */ - unsigned size; /**< Size in bytes of test code. */ + unsigned size; /**< Size in bytes of SPE code. */ unsigned read_fb; /**< Flag: should framebuffer be read? */ }; struct cell_command_logicop { uint64_t base; /**< Effective address of code start. */ - unsigned size; /**< Size in bytes of test code. */ + unsigned size; /**< Size in bytes of SPE code. */ }; -- cgit v1.2.3 From 84c2821d2a3b0252d6ccdfc88c6acd8f72134ebf Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 16:54:31 -0600 Subject: cell: added const qualifier --- src/gallium/drivers/cell/ppu/cell_state_per_fragment.c | 3 ++- src/gallium/drivers/cell/ppu/cell_state_per_fragment.h | 5 +++-- 2 files changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c index 0a79cccc83..53ae3aa50e 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c @@ -1167,7 +1167,8 @@ int PC_OFFSET(const struct spe_function *f, const void *d) * Only two framebuffer formats are supported at this time. */ void -cell_generate_logic_op(struct spe_function *f, struct pipe_blend_state *blend, +cell_generate_logic_op(struct spe_function *f, + const struct pipe_blend_state *blend, struct pipe_surface *surf) { const unsigned logic_op = (blend->logicop_enable) diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h index ab4de96c69..a8267a5133 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h +++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h @@ -32,7 +32,8 @@ extern void cell_generate_alpha_blend(struct cell_blend_state *cb); extern void -cell_generate_logic_op(struct spe_function *f, struct pipe_blend_state *blend, - struct pipe_surface *surf); +cell_generate_logic_op(struct spe_function *f, + const struct pipe_blend_state *blend, + struct pipe_surface *surf); #endif /* CELL_STATE_PER_FRAGMENT_H */ -- cgit v1.2.3 From 14452aee73e16f2ede075cf894e69d62cc539f5e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 17:38:21 -0600 Subject: cell: initial work to support multi-texture --- src/gallium/drivers/cell/common.h | 8 +++++-- src/gallium/drivers/cell/ppu/cell_state_emit.c | 18 +++++++--------- src/gallium/drivers/cell/spu/spu_main.c | 27 ++++++++++++++++-------- src/gallium/drivers/cell/spu/spu_main.h | 8 +++---- src/gallium/drivers/cell/spu/spu_texture.c | 29 ++++++++++++++++---------- src/gallium/drivers/cell/spu/spu_tri.c | 2 +- 6 files changed, 55 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index c9e873b35c..298812fc20 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -66,6 +66,8 @@ #define CELL_MAX_SPUS 6 +#define CELL_MAX_SAMPLERS 4 + #define TILE_SIZE 32 @@ -228,8 +230,10 @@ struct cell_command_release_verts struct cell_command_texture { - void *start; /**< Address in main memory */ - uint width, height; + struct { + void *start; /**< Address in main memory */ + ushort width, height; + } texture[CELL_MAX_SAMPLERS]; }; diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 4c75caa025..4fbe1a21b8 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -129,17 +129,15 @@ cell_emit_state(struct cell_context *cell) if (cell->dirty & CELL_NEW_TEXTURE) { struct cell_command_texture texture; - if (cell->texture[0]) { - texture.start = cell->texture[0]->tiled_data; - texture.width = cell->texture[0]->base.width[0]; - texture.height = cell->texture[0]->base.height[0]; + uint i; + memset(&texture, 0, sizeof(texture)); + for (i = 0;i < CELL_MAX_SAMPLERS; i++) { + if (cell->texture[i]) { + texture.texture[i].start = cell->texture[i]->tiled_data; + texture.texture[i].width = cell->texture[i]->base.width[0]; + texture.texture[i].height = cell->texture[i]->base.height[0]; + } } - else { - texture.start = NULL; - texture.width = 0; - texture.height = 0; - } - emit_state_cmd(cell, CELL_CMD_STATE_TEXTURE, &texture, sizeof(struct cell_command_texture)); } diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index d7f46f8024..80fa5f7859 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -329,17 +329,26 @@ cmd_state_sampler(const struct pipe_sampler_state *state) static void cmd_state_texture(const struct cell_command_texture *texture) { - if (Debug) - printf("SPU %u: TEXTURE at %p size %u x %u\n", - spu.init.id, texture->start, texture->width, texture->height); + uint i; + + if (1||Debug) { + printf("SPU %u: TEXTURE\n", spu.init.id); + for (i = 0; i < CELL_MAX_SAMPLERS; i++) { + printf(" %d: at %p size %u x %u\n", i, texture->texture[i].start, + texture->texture[i].width, texture->texture[i].height); + } + } memcpy(&spu.texture, texture, sizeof(*texture)); - spu.tex_size = (vector float) - { spu.texture.width, spu.texture.height, 0.0, 0.0}; - spu.tex_size_mask = (vector unsigned int) - { spu.texture.width - 1, spu.texture.height - 1, 0, 0 }; - spu.tex_size_x_mask = spu_splats(spu.texture.width - 1); - spu.tex_size_y_mask = spu_splats(spu.texture.height - 1); + for (i = 0; i < CELL_MAX_SAMPLERS; i++) { + const uint width = texture->texture[i].width; + const uint height = texture->texture[i].height; + spu.tex_size[i] = (vector float) { width, height, 0.0, 0.0}; + spu.tex_size_mask[i] = (vector unsigned int) + { width - 1, height - 1, 0, 0 }; + spu.tex_size_x_mask[i] = spu_splats(width - 1); + spu.tex_size_y_mask[i] = spu_splats(height - 1); + } } diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index c20452931a..8a87787537 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -141,10 +141,10 @@ struct spu_global /** for converting RGBA to PIPE_FORMAT_x colors */ vector unsigned char color_shuffle; - vector float tex_size; - vector unsigned int tex_size_mask; /**< == int(size - 1) */ - vector unsigned int tex_size_x_mask; /**< == int(size - 1) */ - vector unsigned int tex_size_y_mask; /**< == int(size - 1) */ + vector float tex_size[CELL_MAX_SAMPLERS]; + vector unsigned int tex_size_mask[CELL_MAX_SAMPLERS]; /**< == int(size - 1) */ + vector unsigned int tex_size_x_mask[CELL_MAX_SAMPLERS]; /**< == int(size - 1) */ + vector unsigned int tex_size_y_mask[CELL_MAX_SAMPLERS]; /**< == int(size - 1) */ vector float (*sample_texture)(vector float texcoord); diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c index 67eb08196a..91a6aec5ec 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.c +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -40,25 +40,29 @@ void invalidate_tex_cache(void) { - spu_dcache_mark_dirty((unsigned) spu.texture.start, - 4 * spu.texture.width * spu.texture.height); + uint unit = 0; + uint bytes = 4 * spu.texture.texture[unit].width + * spu.texture.texture[unit].height; + + spu_dcache_mark_dirty((unsigned) spu.texture.texture[unit].start, bytes); } static uint get_texel(vec_uint4 coordinate) { + const uint unit = 0; vec_uint4 tmp; unsigned x = spu_extract(coordinate, 0); unsigned y = spu_extract(coordinate, 1); - const unsigned tiles_per_row = spu.texture.width / TILE_SIZE; + const unsigned tiles_per_row = spu.texture.texture[unit].width / TILE_SIZE; unsigned tile_offset = sizeof(tile_t) * ((y / TILE_SIZE * tiles_per_row) + (x / TILE_SIZE)); unsigned texel_offset = 4 * (((y % TILE_SIZE) * TILE_SIZE) + (x % TILE_SIZE)); spu_dcache_fetch_unaligned((qword *) & tmp, - spu.texture.start + tile_offset + texel_offset, + spu.texture.texture[unit].start + tile_offset + texel_offset, 4); return spu_extract(tmp, 0); } @@ -67,13 +71,14 @@ get_texel(vec_uint4 coordinate) static void get_four_texels(vec_uint4 x, vec_uint4 y, vec_uint4 *texels) { - const unsigned texture_ea = (uintptr_t) spu.texture.start; + const uint unit = 0; + const unsigned texture_ea = (uintptr_t) spu.texture.texture[unit].start; vec_uint4 tile_x = spu_rlmask(x, -5); vec_uint4 tile_y = spu_rlmask(y, -5); const qword offset_x = si_andi((qword) x, 0x1f); const qword offset_y = si_andi((qword) y, 0x1f); - const qword tiles_per_row = (qword) spu_splats(spu.texture.width / TILE_SIZE); + const qword tiles_per_row = (qword) spu_splats(spu.texture.texture[unit].width / TILE_SIZE); const qword tile_size = (qword) spu_splats(sizeof(tile_t)); qword tile_offset = si_mpya((qword) tile_y, tiles_per_row, (qword) tile_x); @@ -101,9 +106,10 @@ get_four_texels(vec_uint4 x, vec_uint4 y, vec_uint4 *texels) vector float sample_texture_nearest(vector float texcoord) { - vector float tc = spu_mul(texcoord, spu.tex_size); + const uint unit = 0; + vector float tc = spu_mul(texcoord, spu.tex_size[unit]); vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ - itc = spu_and(itc, spu.tex_size_mask); /* mask (GL_REPEAT) */ + itc = spu_and(itc, spu.tex_size_mask[unit]); /* mask (GL_REPEAT) */ uint texel = get_texel(itc); return spu_unpack_A8R8G8B8(texel); } @@ -112,10 +118,11 @@ sample_texture_nearest(vector float texcoord) vector float sample_texture_bilinear(vector float texcoord) { + const uint unit = 0; static const vec_uint4 offset_x = {0, 0, 1, 1}; static const vec_uint4 offset_y = {0, 1, 0, 1}; - vector float tc = spu_mul(texcoord, spu.tex_size); + vector float tc = spu_mul(texcoord, spu.tex_size[unit]); tc = spu_add(tc, spu_splats(-0.5f)); /* half texel bias */ /* integer texcoords S,T: */ @@ -129,8 +136,8 @@ sample_texture_bilinear(vector float texcoord) x = spu_add(x, offset_x); y = spu_add(y, offset_y); - x = spu_and(x, spu.tex_size_x_mask); - y = spu_and(y, spu.tex_size_y_mask); + x = spu_and(x, spu.tex_size_x_mask[unit]); + y = spu_and(y, spu.tex_size_y_mask[unit]); get_four_texels(x, y, texels); diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index 95c629a8aa..9f63317b1f 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -309,7 +309,7 @@ emit_quad( int x, int y, mask_t mask ) spu.cur_ctile_status = TILE_STATUS_DIRTY; - if (spu.texture.start) { + if (spu.texture.texture[0].start) { /* texture mapping */ vector float texcoords[4]; eval_coeff(2, (float) x, (float) y, texcoords); -- cgit v1.2.3 From 686a6c746851b4bc2572aaa1153d570717bdd77f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 1 Apr 2008 10:41:43 +0900 Subject: gallium: Do not assume that buffers are freed in the same order they are fenced. Also free buffers as soon as possible. This short term fix corrects the fenced list behavior but it will impact on performance. The long term fix is probably replace the linked list (legacy from the bufpool code) by a binary tree. --- .../auxiliary/pipebuffer/pb_buffer_fenced.c | 39 ++++++++++++---------- 1 file changed, 21 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 6e217eb2e0..55f32e6816 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -29,8 +29,8 @@ * \file * Implementation of fenced buffers. * - * \author José Fonseca - * \author Thomas Hellström + * \author José Fonseca + * \author Thomas Hellström */ @@ -44,7 +44,7 @@ #include "pb_buffer.h" #include "pb_buffer_fenced.h" -#ifndef __MSC__ +#ifndef WIN32 #include #endif @@ -93,8 +93,6 @@ fenced_buffer(struct pb_buffer *buf) } - - static void _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, int wait) @@ -105,15 +103,6 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, int signaled = -1; list = fenced_list->delayed.next; - - if (fenced_list->numDelayed > 3) { - unsigned i; - - for (i = 0; i < fenced_list->numDelayed; i += 3) { - list = list->next; - } - } - prev = list->prev; for (; list != &fenced_list->delayed; list = prev, prev = list->prev) { @@ -128,11 +117,17 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, } } - if (signaled != 0) + if (signaled != 0) { +#if 0 /* XXX: we are assuming that buffers are freed in the same order they * are fenced which may not always be true... */ break; +#else + signaled = -1; + continue; +#endif + } winsys->fence_reference(winsys, &fenced_buf->fence, NULL); @@ -154,8 +149,16 @@ fenced_buffer_destroy(struct pb_buffer *buf) struct fenced_buffer_list *fenced_list = fenced_buf->list; if (fenced_buf->fence) { - LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); - fenced_list->numDelayed++; + struct pipe_winsys *winsys = fenced_list->winsys; + if(winsys->fence_finish(winsys, fenced_buf->fence, 0) != 0) { + LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); + fenced_list->numDelayed++; + } + else { + winsys->fence_reference(winsys, &fenced_buf->fence, NULL); + pb_reference(&fenced_buf->buffer, NULL); + FREE(fenced_buf); + } } else { pb_reference(&fenced_buf->buffer, NULL); @@ -285,7 +288,7 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) /* Wait on outstanding fences */ while (fenced_list->numDelayed) { _glthread_UNLOCK_MUTEX(fenced_list->mutex); -#ifndef __MSC__ +#ifndef WIN32 sched_yield(); #endif _fenced_buffer_list_check_free(fenced_list, 1); -- cgit v1.2.3 From 5553a3b6757f0baaabbd67dd2f86d834d2f291ca Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 20:36:53 -0600 Subject: cell: set cell->num_textures in cell_set_sampler_textures() --- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 00f4be7401..52c3126050 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -273,6 +273,7 @@ cell_set_sampler_textures(struct pipe_context *pipe, pipe_texture_reference((struct pipe_texture **) &cell->texture[i], tex); } + cell->num_textures = num; cell_update_texture_mapping(cell); -- cgit v1.2.3 From e6c981f22c0b6469ef44e9d7a34113db34647fef Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 21:09:02 -0600 Subject: cell: more work for multi-texture support --- src/gallium/drivers/cell/common.h | 17 ++++++-- src/gallium/drivers/cell/ppu/cell_state_emit.c | 31 ++++++++++----- src/gallium/drivers/cell/spu/spu_main.c | 55 +++++++++++++++----------- src/gallium/drivers/cell/spu/spu_main.h | 18 ++++++--- src/gallium/drivers/cell/spu/spu_texture.c | 24 +++++------ src/gallium/drivers/cell/spu/spu_tri.c | 2 +- 6 files changed, 90 insertions(+), 57 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index 298812fc20..f430e88b9c 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -36,6 +36,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_util.h" #include "pipe/p_format.h" +#include "pipe/p_state.h" /** The standard assert macro doesn't seem to work reliably */ @@ -228,12 +229,20 @@ struct cell_command_release_verts }; +struct cell_command_sampler +{ + uint64_t opcode; /**< CELL_CMD_STATE_SAMPLER */ + uint unit; + struct pipe_sampler_state state; +}; + + struct cell_command_texture { - struct { - void *start; /**< Address in main memory */ - ushort width, height; - } texture[CELL_MAX_SAMPLERS]; + uint64_t opcode; /**< CELL_CMD_STATE_TEXTURE */ + uint unit; + void *start; /**< Address in main memory */ + ushort width, height; }; diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 4fbe1a21b8..9cae67f091 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -121,25 +121,36 @@ cell_emit_state(struct cell_context *cell) } if (cell->dirty & CELL_NEW_SAMPLER) { - if (cell->sampler[0]) { - emit_state_cmd(cell, CELL_CMD_STATE_SAMPLER, - cell->sampler[0], sizeof(struct pipe_sampler_state)); + uint i; + for (i = 0; i < CELL_MAX_SAMPLERS; i++) { + if (cell->sampler[i]) { + struct cell_command_sampler *sampler + = cell_batch_alloc(cell, sizeof(*sampler)); + sampler->opcode = CELL_CMD_STATE_SAMPLER; + sampler->unit = i; + sampler->state = *cell->sampler[i]; + } } } if (cell->dirty & CELL_NEW_TEXTURE) { - struct cell_command_texture texture; uint i; - memset(&texture, 0, sizeof(texture)); for (i = 0;i < CELL_MAX_SAMPLERS; i++) { + struct cell_command_texture *texture + = cell_batch_alloc(cell, sizeof(*texture)); + texture->opcode = CELL_CMD_STATE_TEXTURE; + texture->unit = i; if (cell->texture[i]) { - texture.texture[i].start = cell->texture[i]->tiled_data; - texture.texture[i].width = cell->texture[i]->base.width[0]; - texture.texture[i].height = cell->texture[i]->base.height[0]; + texture->start = cell->texture[i]->tiled_data; + texture->width = cell->texture[i]->base.width[0]; + texture->height = cell->texture[i]->base.height[0]; + } + else { + texture->start = NULL; + texture->width = 1; + texture->height = 1; } } - emit_state_cmd(cell, CELL_CMD_STATE_TEXTURE, - &texture, sizeof(struct cell_command_texture)); } if (cell->dirty & CELL_NEW_VERTEX_INFO) { diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 80fa5f7859..7f0473d198 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -312,13 +312,13 @@ cmd_state_depth_stencil(const struct cell_command_depth_stencil_alpha_test *stat static void -cmd_state_sampler(const struct pipe_sampler_state *state) +cmd_state_sampler(const struct cell_command_sampler *sampler) { if (Debug) - printf("SPU %u: SAMPLER\n", - spu.init.id); + printf("SPU %u: SAMPLER [%u]\n", + spu.init.id, sampler->unit); - memcpy(&spu.sampler[0], state, sizeof(*state)); + spu.sampler[sampler->unit] = sampler->state; if (spu.sampler[0].min_img_filter == PIPE_TEX_FILTER_LINEAR) spu.sample_texture = sample_texture_bilinear; else @@ -329,26 +329,25 @@ cmd_state_sampler(const struct pipe_sampler_state *state) static void cmd_state_texture(const struct cell_command_texture *texture) { - uint i; + const uint unit = texture->unit; + const uint width = texture->width; + const uint height = texture->height; - if (1||Debug) { - printf("SPU %u: TEXTURE\n", spu.init.id); - for (i = 0; i < CELL_MAX_SAMPLERS; i++) { - printf(" %d: at %p size %u x %u\n", i, texture->texture[i].start, - texture->texture[i].width, texture->texture[i].height); - } + if (Debug) { + printf("SPU %u: TEXTURE [%u] at %p size %u x %u\n", spu.init.id, + texture->unit, texture->start, + texture->width, texture->height); } - memcpy(&spu.texture, texture, sizeof(*texture)); - for (i = 0; i < CELL_MAX_SAMPLERS; i++) { - const uint width = texture->texture[i].width; - const uint height = texture->texture[i].height; - spu.tex_size[i] = (vector float) { width, height, 0.0, 0.0}; - spu.tex_size_mask[i] = (vector unsigned int) + spu.texture[unit].start = texture->start; + spu.texture[unit].width = width; + spu.texture[unit].height = height; + + spu.texture[unit].tex_size = (vector float) { width, height, 0.0, 0.0}; + spu.texture[unit].tex_size_mask = (vector unsigned int) { width - 1, height - 1, 0, 0 }; - spu.tex_size_x_mask[i] = spu_splats(width - 1); - spu.tex_size_y_mask[i] = spu_splats(height - 1); - } + spu.texture[unit].tex_size_x_mask = spu_splats(width - 1); + spu.texture[unit].tex_size_y_mask = spu_splats(height - 1); } @@ -480,12 +479,20 @@ cmd_batch(uint opcode) pos += (1 + ROUNDUP8(sizeof(struct cell_command_depth_stencil_alpha_test)) / 8); break; case CELL_CMD_STATE_SAMPLER: - cmd_state_sampler((struct pipe_sampler_state *) &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct pipe_sampler_state)) / 8); + { + struct cell_command_sampler *sampler + = (struct cell_command_sampler *) &buffer[pos]; + cmd_state_sampler(sampler); + pos += sizeof(*sampler) / 8; + } break; case CELL_CMD_STATE_TEXTURE: - cmd_state_texture((struct cell_command_texture *) &buffer[pos+1]); - pos += (1 + ROUNDUP8(sizeof(struct cell_command_texture)) / 8); + { + struct cell_command_texture *texture + = (struct cell_command_texture *) &buffer[pos]; + cmd_state_texture(texture); + pos += sizeof(*texture) / 8; + } break; case CELL_CMD_STATE_VERTEX_INFO: cmd_state_vertex_info((struct vertex_info *) &buffer[pos+1]); diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index 8a87787537..2bfad3535a 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -100,6 +100,17 @@ struct spu_framebuffer { } ALIGN16_ATTRIB; +struct spu_texture +{ + void *start; + uint width, height; + vector float tex_size; + vector unsigned int tex_size_mask; /**< == int(size - 1) */ + vector unsigned int tex_size_x_mask; /**< == int(size - 1) */ + vector unsigned int tex_size_y_mask; /**< == int(size - 1) */ +} ALIGN16_ATTRIB; + + /** * All SPU global/context state will be in singleton object of this type: */ @@ -119,7 +130,7 @@ struct spu_global logicop_func logicop; struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS]; - struct cell_command_texture texture; + struct spu_texture texture[PIPE_MAX_SAMPLERS]; struct vertex_info vertex_info; @@ -141,11 +152,6 @@ struct spu_global /** for converting RGBA to PIPE_FORMAT_x colors */ vector unsigned char color_shuffle; - vector float tex_size[CELL_MAX_SAMPLERS]; - vector unsigned int tex_size_mask[CELL_MAX_SAMPLERS]; /**< == int(size - 1) */ - vector unsigned int tex_size_x_mask[CELL_MAX_SAMPLERS]; /**< == int(size - 1) */ - vector unsigned int tex_size_y_mask[CELL_MAX_SAMPLERS]; /**< == int(size - 1) */ - vector float (*sample_texture)(vector float texcoord); } ALIGN16_ATTRIB; diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c index 91a6aec5ec..4612501eb3 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.c +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -41,10 +41,10 @@ void invalidate_tex_cache(void) { uint unit = 0; - uint bytes = 4 * spu.texture.texture[unit].width - * spu.texture.texture[unit].height; + uint bytes = 4 * spu.texture[unit].width + * spu.texture[unit].height; - spu_dcache_mark_dirty((unsigned) spu.texture.texture[unit].start, bytes); + spu_dcache_mark_dirty((unsigned) spu.texture[unit].start, bytes); } @@ -55,14 +55,14 @@ get_texel(vec_uint4 coordinate) vec_uint4 tmp; unsigned x = spu_extract(coordinate, 0); unsigned y = spu_extract(coordinate, 1); - const unsigned tiles_per_row = spu.texture.texture[unit].width / TILE_SIZE; + const unsigned tiles_per_row = spu.texture[unit].width / TILE_SIZE; unsigned tile_offset = sizeof(tile_t) * ((y / TILE_SIZE * tiles_per_row) + (x / TILE_SIZE)); unsigned texel_offset = 4 * (((y % TILE_SIZE) * TILE_SIZE) + (x % TILE_SIZE)); spu_dcache_fetch_unaligned((qword *) & tmp, - spu.texture.texture[unit].start + tile_offset + texel_offset, + spu.texture[unit].start + tile_offset + texel_offset, 4); return spu_extract(tmp, 0); } @@ -72,13 +72,13 @@ static void get_four_texels(vec_uint4 x, vec_uint4 y, vec_uint4 *texels) { const uint unit = 0; - const unsigned texture_ea = (uintptr_t) spu.texture.texture[unit].start; + const unsigned texture_ea = (uintptr_t) spu.texture[unit].start; vec_uint4 tile_x = spu_rlmask(x, -5); vec_uint4 tile_y = spu_rlmask(y, -5); const qword offset_x = si_andi((qword) x, 0x1f); const qword offset_y = si_andi((qword) y, 0x1f); - const qword tiles_per_row = (qword) spu_splats(spu.texture.texture[unit].width / TILE_SIZE); + const qword tiles_per_row = (qword) spu_splats(spu.texture[unit].width / TILE_SIZE); const qword tile_size = (qword) spu_splats(sizeof(tile_t)); qword tile_offset = si_mpya((qword) tile_y, tiles_per_row, (qword) tile_x); @@ -107,9 +107,9 @@ vector float sample_texture_nearest(vector float texcoord) { const uint unit = 0; - vector float tc = spu_mul(texcoord, spu.tex_size[unit]); + vector float tc = spu_mul(texcoord, spu.texture[unit].tex_size); vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ - itc = spu_and(itc, spu.tex_size_mask[unit]); /* mask (GL_REPEAT) */ + itc = spu_and(itc, spu.texture[unit].tex_size_mask); /* mask (GL_REPEAT) */ uint texel = get_texel(itc); return spu_unpack_A8R8G8B8(texel); } @@ -122,7 +122,7 @@ sample_texture_bilinear(vector float texcoord) static const vec_uint4 offset_x = {0, 0, 1, 1}; static const vec_uint4 offset_y = {0, 1, 0, 1}; - vector float tc = spu_mul(texcoord, spu.tex_size[unit]); + vector float tc = spu_mul(texcoord, spu.texture[unit].tex_size); tc = spu_add(tc, spu_splats(-0.5f)); /* half texel bias */ /* integer texcoords S,T: */ @@ -136,8 +136,8 @@ sample_texture_bilinear(vector float texcoord) x = spu_add(x, offset_x); y = spu_add(y, offset_y); - x = spu_and(x, spu.tex_size_x_mask[unit]); - y = spu_and(y, spu.tex_size_y_mask[unit]); + x = spu_and(x, spu.texture[unit].tex_size_x_mask); + y = spu_and(y, spu.texture[unit].tex_size_y_mask); get_four_texels(x, y, texels); diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index 9f63317b1f..17e337bbdf 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -309,7 +309,7 @@ emit_quad( int x, int y, mask_t mask ) spu.cur_ctile_status = TILE_STATUS_DIRTY; - if (spu.texture.texture[0].start) { + if (spu.texture[0].start) { /* texture mapping */ vector float texcoords[4]; eval_coeff(2, (float) x, (float) y, texcoords); -- cgit v1.2.3 From d83e0c45bec8d08c249088f9e8575505355fe595 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 31 Mar 2008 21:15:57 -0600 Subject: cell: update some of the CAP, texformat queries --- src/gallium/drivers/cell/ppu/cell_screen.c | 20 ++++++++++++-------- 1 file changed, 12 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c index 124670df25..84b48bf4f1 100644 --- a/src/gallium/drivers/cell/ppu/cell_screen.c +++ b/src/gallium/drivers/cell/ppu/cell_screen.c @@ -55,11 +55,11 @@ cell_get_param(struct pipe_screen *screen, int param) { switch (param) { case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return 8; + return PIPE_MAX_SAMPLERS; case PIPE_CAP_NPOT_TEXTURES: - return 1; + return 0; case PIPE_CAP_TWO_SIDED_STENCIL: - return 1; + return 0; case PIPE_CAP_GLSL: return 1; case PIPE_CAP_S3TC: @@ -67,13 +67,13 @@ cell_get_param(struct pipe_screen *screen, int param) case PIPE_CAP_ANISOTROPIC_FILTER: return 0; case PIPE_CAP_POINT_SPRITE: - return 1; + return 0; case PIPE_CAP_MAX_RENDER_TARGETS: return 1; case PIPE_CAP_OCCLUSION_QUERY: - return 1; + return 0; case PIPE_CAP_TEXTURE_SHADOW_MAP: - return 1; + return 0; case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: return 12; /* max 2Kx2K */ case PIPE_CAP_MAX_TEXTURE_3D_LEVELS: @@ -118,8 +118,12 @@ cell_is_format_supported( struct pipe_screen *screen, { switch (type) { case PIPE_TEXTURE: - /* cell supports all texture formats, XXX for now anyway */ - return TRUE; + /* cell supports most texture formats, XXX for now anyway */ + if (format == PIPE_FORMAT_DXT5_RGBA || + format == PIPE_FORMAT_R8G8B8A8_SRGB) + return FALSE; + else + return TRUE; case PIPE_SURFACE: /* cell supports all (off-screen) surface formats, XXX for now */ return TRUE; -- cgit v1.2.3 From 2df9941368c807fb677d3d95a5c1dfffe719c26b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 1 Apr 2008 13:00:51 +0100 Subject: scons: Fix test for building dri/intel winsys. --- src/gallium/winsys/SConscript | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript index f18d3bd2f8..e8a581adb2 100644 --- a/src/gallium/winsys/SConscript +++ b/src/gallium/winsys/SConscript @@ -1,6 +1,6 @@ Import('*') -if dri: +if 'intel' in env['winsys'] and dri: SConscript([ 'dri/SConscript', ]) -- cgit v1.2.3 From cc85573a9f3d976b70669a53403ce7355ad84394 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 1 Apr 2008 13:02:30 +0100 Subject: gallium: Fencing fix. Make sure the struct pipe_fence_handle* we point st_flush() to is initialized to NULL, so winsys->fence_reference() doesn't try to unreference a random struct pipe_fence_handle* pointer. --- src/mesa/state_tracker/st_cb_flush.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index e321b401e2..1de3676bda 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -105,7 +105,7 @@ static void st_glFlush(GLcontext *ctx) void st_finish( struct st_context *st ) { - struct pipe_fence_handle *fence; + struct pipe_fence_handle *fence = NULL; st_gl_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence); -- cgit v1.2.3 From 4b1377b2403bcb34081f91991f1ffde06df17af1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 1 Apr 2008 14:14:06 +0100 Subject: draw: flush between pt/non-pt modes --- src/gallium/auxiliary/draw/draw_pt.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index fc9304197a..df306a5fa4 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -167,6 +167,10 @@ draw_pt_arrays(struct draw_context *draw, frontend = draw->pt.front.vcache; #endif + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + frontend->prepare( frontend, middle ); frontend->run( frontend, -- cgit v1.2.3 From 52f40dcc468039fc9cca45a4de20a5aa11228b67 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 1 Apr 2008 14:14:46 +0100 Subject: draw: associate rhw divide with clipping not viewport flag --- src/gallium/auxiliary/draw/draw_vs_exec.c | 12 ++++++------ src/gallium/auxiliary/draw/draw_vs_llvm.c | 12 ++++++------ src/gallium/auxiliary/draw/draw_vs_sse.c | 12 ++++++------ 3 files changed, 18 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index c8ed17c00a..c6e503686a 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -143,6 +143,12 @@ vs_exec_run( struct draw_vertex_shader *shader, if (!draw->rasterizer->bypass_clipping) { vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; } else { vOut[j]->clipmask = 0; @@ -150,12 +156,6 @@ vs_exec_run( struct draw_vertex_shader *shader, vOut[j]->edgeflag = 1; if (!draw->identity_viewport) { - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - /* Viewport mapping */ vOut[j]->data[0][0] = x * scale[0] + trans[0]; vOut[j]->data[0][1] = y * scale[1] + trans[1]; diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 8aa8a617bb..c8268317ef 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -153,6 +153,12 @@ vs_llvm_run( struct draw_vertex_shader *base, if (!draw->rasterizer->bypass_clipping) { vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; } else { vOut[j]->clipmask = 0; @@ -160,12 +166,6 @@ vs_llvm_run( struct draw_vertex_shader *base, vOut[j]->edgeflag = 1; if (!draw->identity_viewport) { - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - /* Viewport mapping */ vOut[j]->data[0][0] = x * scale[0] + trans[0]; vOut[j]->data[0][1] = y * scale[1] + trans[1]; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 701137f908..f40d65df08 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -168,6 +168,12 @@ vs_sse_run( struct draw_vertex_shader *base, if (!draw->rasterizer->bypass_clipping) { vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; } else { vOut[j]->clipmask = 0; @@ -175,12 +181,6 @@ vs_sse_run( struct draw_vertex_shader *base, vOut[j]->edgeflag = 1; if (!draw->identity_viewport) { - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - /* Viewport mapping */ vOut[j]->data[0][0] = x * scale[0] + trans[0]; vOut[j]->data[0][1] = y * scale[1] + trans[1]; -- cgit v1.2.3 From caa44763f7f7aa26ed0b0d1e5af0c410fba6bfe6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 1 Apr 2008 14:48:59 +0100 Subject: draw: respect flatshade_first in flatshade stage --- src/gallium/auxiliary/draw/draw_flatshade.c | 55 +++++++++++++++++++++++++---- 1 file changed, 49 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_flatshade.c b/src/gallium/auxiliary/draw/draw_flatshade.c index ccad71d695..af2cb05c98 100644 --- a/src/gallium/auxiliary/draw/draw_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_flatshade.c @@ -84,8 +84,25 @@ static INLINE void copy_colors2( struct draw_stage *stage, * Flatshade tri. Required for clipping and when unfilled tris are * active, otherwise handled by hardware. */ -static void flatshade_tri( struct draw_stage *stage, - struct prim_header *header ) +static void flatshade_tri_0( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + tmp.v[0] = header->v[0]; + tmp.v[1] = dup_vert(stage, header->v[1], 0); + tmp.v[2] = dup_vert(stage, header->v[2], 1); + + copy_colors2(stage, tmp.v[1], tmp.v[2], tmp.v[0]); + + stage->next->tri( stage->next, &tmp ); +} + + +static void flatshade_tri_2( struct draw_stage *stage, + struct prim_header *header ) { struct prim_header tmp; @@ -101,11 +118,27 @@ static void flatshade_tri( struct draw_stage *stage, } + + + /** * Flatshade line. Required for clipping. */ -static void flatshade_line( struct draw_stage *stage, - struct prim_header *header ) +static void flatshade_line_0( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.v[0] = header->v[0]; + tmp.v[1] = dup_vert(stage, header->v[1], 0); + + copy_colors(stage, tmp.v[1], tmp.v[0]); + + stage->next->line( stage->next, &tmp ); +} + +static void flatshade_line_1( struct draw_stage *stage, + struct prim_header *header ) { struct prim_header tmp; @@ -118,6 +151,8 @@ static void flatshade_line( struct draw_stage *stage, } +/* Flatshade point -- passthrough. + */ static void flatshade_point( struct draw_stage *stage, struct prim_header *header ) { @@ -140,8 +175,16 @@ static void flatshade_init_state( struct draw_stage *stage ) } } - stage->line = flatshade_line; - stage->tri = flatshade_tri; + /* Choose flatshade routine according to provoking vertex: + */ + if (stage->draw->rasterizer->flatshade_first) { + stage->line = flatshade_line_0; + stage->tri = flatshade_tri_0; + } + else { + stage->line = flatshade_line_1; + stage->tri = flatshade_tri_2; + } } static void flatshade_first_tri( struct draw_stage *stage, -- cgit v1.2.3 From edfa8201a50c47376b7aa0c05d7851e3e1353bde Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 1 Apr 2008 14:49:56 +0100 Subject: draw: more flatshade_first changes - Reduce the number of changes to the normal vertex ordering - Assume that the hardware knows how to do this in the standard case. - Add support to the passthrough vcache path. --- src/gallium/auxiliary/draw/draw_prim.c | 102 +++++++--------------- src/gallium/auxiliary/draw/draw_pt.c | 2 +- src/gallium/auxiliary/draw/draw_pt.h | 2 +- src/gallium/auxiliary/draw/draw_pt_vcache.c | 127 ++++++++++++++++++++++++---- src/gallium/auxiliary/draw/draw_validate.c | 8 -- 5 files changed, 144 insertions(+), 97 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index 9779aa8440..4452376a70 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -343,21 +343,11 @@ draw_prim( struct draw_context *draw, break; case PIPE_PRIM_LINES: - if (flatfirst) { - for (i = 0; i+1 < count; i += 2) { - do_line( draw, - TRUE, - start + i + 1, - start + i + 0); - } - } - else { - for (i = 0; i+1 < count; i += 2) { - do_line( draw, - TRUE, - start + i + 0, - start + i + 1); - } + for (i = 0; i+1 < count; i += 2) { + do_line( draw, + TRUE, + start + i + 0, + start + i + 1); } break; @@ -378,63 +368,31 @@ draw_prim( struct draw_context *draw, break; case PIPE_PRIM_LINE_STRIP: - if (flatfirst) { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, - start + i, - start + i - 1 ); - } - } - else { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, - start + i - 1, - start + i ); - } + for (i = 1; i < count; i++) { + do_line( draw, + i == 1, + start + i - 1, + start + i ); } break; case PIPE_PRIM_TRIANGLES: - if (flatfirst) { - if (unfilled) { - for (i = 0; i+2 < count; i += 3) { - do_ef_triangle( draw, - 1, - ~0, - start + i + 1, - start + i + 2, - start + i + 0 ); - } - } - else { - for (i = 0; i+2 < count; i += 3) { - do_triangle( draw, - start + i + 1, - start + i + 2, - start + i + 0 ); - } + if (unfilled) { + for (i = 0; i+2 < count; i += 3) { + do_ef_triangle( draw, + 1, + ~0, + start + i + 0, + start + i + 1, + start + i + 2 ); } - } + } else { - if (unfilled) { - for (i = 0; i+2 < count; i += 3) { - do_ef_triangle( draw, - 1, - ~0, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - else { - for (i = 0; i+2 < count; i += 3) { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } + for (i = 0; i+2 < count; i += 3) { + do_triangle( draw, + start + i + 0, + start + i + 1, + start + i + 2 ); } } break; @@ -444,15 +402,15 @@ draw_prim( struct draw_context *draw, for (i = 0; i+2 < count; i++) { if (i & 1) { do_triangle( draw, + start + i + 0, start + i + 2, - start + i + 1, - start + i + 0 ); + start + i + 1 ); } else { do_triangle( draw, + start + i + 0, start + i + 1, - start + i + 2, - start + i + 0 ); + start + i + 2 ); } } } @@ -479,9 +437,9 @@ draw_prim( struct draw_context *draw, if (flatfirst) { for (i = 0; i+2 < count; i++) { do_triangle( draw, + start + i + 1, start + i + 2, - start + 0, - start + i + 1 ); + start + 0 ); } } else { diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index df306a5fa4..2ea96c686d 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -191,7 +191,7 @@ boolean draw_pt_init( struct draw_context *draw ) if (!draw->pt.middle.fetch_emit) return FALSE; - draw->pt.front.vcache = draw_pt_vcache(); + draw->pt.front.vcache = draw_pt_vcache( draw ); if (!draw->pt.front.vcache) return FALSE; diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 439fa4c881..f878616079 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -110,7 +110,7 @@ const void *draw_pt_elt_ptr( struct draw_context *draw, /* Implementations: */ -struct draw_pt_front_end *draw_pt_vcache( void ); +struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index da9a3a52ae..16ffedf580 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -44,6 +44,7 @@ struct vcache_frontend { struct draw_pt_front_end base; + struct draw_context *draw; unsigned in[CACHE_MAX]; ushort out[CACHE_MAX]; @@ -157,14 +158,6 @@ static void vcache_quad( struct vcache_frontend *vcache, } -static void vcache_prepare( struct draw_pt_front_end *frontend, - struct draw_pt_middle_end *middle ) -{ - struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; - vcache->middle = middle; - middle->prepare( middle ); -} - static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { PIPE_PRIM_POINTS, PIPE_PRIM_LINES, @@ -179,11 +172,11 @@ static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { }; -static void vcache_run( struct draw_pt_front_end *frontend, - unsigned prim, - pt_elt_func get_elt, - const void *elts, - unsigned count ) +static void vcache_run_pv2( struct draw_pt_front_end *frontend, + unsigned prim, + pt_elt_func get_elt, + const void *elts, + unsigned count ) { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; unsigned i; @@ -309,6 +302,109 @@ static void vcache_run( struct draw_pt_front_end *frontend, vcache_flush( vcache ); } + +static void vcache_run_pv0( struct draw_pt_front_end *frontend, + unsigned prim, + pt_elt_func get_elt, + const void *elts, + unsigned count ) +{ + struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; + unsigned i; + + /* These are for validation only: + */ + vcache->elt_func = get_elt; + vcache->elt_ptr = elts; + vcache->output_prim = reduced_prim[prim]; + + switch (prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i ++) { + vcache_point( vcache, + get_elt(elts, i) ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) { + vcache_line( vcache, + TRUE, + get_elt(elts, i + 0), + get_elt(elts, i + 1)); + } + break; + + case PIPE_PRIM_LINE_STRIP: + for (i = 1; i < count; i++) { + vcache_line( vcache, + i == 1, + get_elt(elts, i - 1), + get_elt(elts, i) ); + } + break; + + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + vcache_triangle( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2) ); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + for (i = 0; i+2 < count; i++) { + if (i & 1) { + vcache_triangle( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 2), + get_elt(elts, i + 1) ); + } + else { + vcache_triangle( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2) ); + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + for (i = 0; i+2 < count; i++) { + vcache_triangle( vcache, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0) ); + } + break; + + + default: + assert(0); + break; + } + + vcache_flush( vcache ); +} + +static void vcache_prepare( struct draw_pt_front_end *frontend, + struct draw_pt_middle_end *middle ) +{ + struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; + + if (vcache->draw->rasterizer->flatshade_first) + vcache->base.run = vcache_run_pv0; + else + vcache->base.run = vcache_run_pv2; + + vcache->middle = middle; + middle->prepare( middle ); +} + + + + static void vcache_finish( struct draw_pt_front_end *frontend ) { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; @@ -322,14 +418,15 @@ static void vcache_destroy( struct draw_pt_front_end *frontend ) } -struct draw_pt_front_end *draw_pt_vcache( void ) +struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ) { struct vcache_frontend *vcache = CALLOC_STRUCT( vcache_frontend ); vcache->base.prepare = vcache_prepare; - vcache->base.run = vcache_run; + vcache->base.run = NULL; vcache->base.finish = vcache_finish; vcache->base.destroy = vcache_destroy; + vcache->draw = draw; memset(vcache->in, ~0, sizeof(vcache->in)); diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c index ad43f06f73..e163e078f0 100644 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ b/src/gallium/auxiliary/draw/draw_validate.c @@ -76,10 +76,6 @@ draw_need_pipeline(const struct draw_context *draw, /* AA lines */ if (draw->rasterizer->line_smooth && draw->pipeline.aaline) return TRUE; - - /* first-vertex driven flatshading */ - if (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first) - return TRUE; } if (points(prim)) @@ -116,10 +112,6 @@ draw_need_pipeline(const struct draw_context *draw, /* two-side lighting */ if (draw->rasterizer->light_twoside) return TRUE; - - /* first-vertex driven flatshading */ - if (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first) - return TRUE; } /* polygon cull - this is difficult - hardware can cull just fine -- cgit v1.2.3 From 8f8b95ae58c6e51f3bf0bc6ed2fa5ac8b9e82f1e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 1 Apr 2008 15:19:30 +0100 Subject: draw: remove dead code --- src/gallium/auxiliary/draw/draw_prim.c | 37 ---------------------------------- 1 file changed, 37 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index 4452376a70..51b6950334 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -119,43 +119,6 @@ static void draw_prim_queue_flush( struct draw_context *draw ) draw_vertex_cache_unreference( draw ); } -static INLINE void fetch_and_store(struct draw_context *draw) -{ - /* run vertex shader on vertex cache entries, four per invokation */ -#if 0 - { - const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); - memcpy(draw->vs.queue[0].vertex, draw->vs.queue[i + j].elt, - count * vinfo->size); - } -#elif 0 - unsigned i; - - draw_update_vertex_fetch(draw); - for (i = 0; i < draw->vs.queue_nr; i += 4) { - struct vertex_header *dests[4]; - unsigned elts[4]; - struct tgsi_exec_machine *machine = &draw->machine; - int j, n = MIN2(4, draw->vs.queue_nr - i); - - for (j = 0; j < n; j++) { - elts[j] = draw->vs.queue[i + j].elt; - dests[j] = draw->vs.queue[i + j].vertex; - } - - for ( ; j < 4; j++) { - elts[j] = elts[0]; - dests[j] = draw->vs.queue[i + j].vertex; - } - //fetch directly into dests - draw->vertex_fetch.fetch_func(draw, machine, dests, count); - } -#endif - - draw->vs.post_nr = draw->vs.queue_nr; - draw->vs.queue_nr = 0; -} - void draw_do_flush( struct draw_context *draw, unsigned flags ) { if (0) -- cgit v1.2.3 From f3a62372bf155d14876f98b1a085841607bb344c Mon Sep 17 00:00:00 2001 From: Ben Skeggs Date: Tue, 1 Apr 2008 08:59:13 -0600 Subject: handle IsPositionInvariant flag --- src/mesa/state_tracker/st_program.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 101a6a48af..4d1ef1bf3d 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -34,6 +34,7 @@ #include "main/imports.h" #include "main/mtypes.h" #include "shader/prog_print.h" +#include "shader/programopt.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -95,6 +96,9 @@ st_translate_vertex_program(struct st_context *st, memset(&vs, 0, sizeof(vs)); + if (stvp->Base.IsPositionInvariant) + _mesa_insert_mvp_code(st->ctx, &stvp->Base); + /* * Determine number of inputs, the mappings between VERT_ATTRIB_x * and TGSI generic input indexes, plus input attrib semantic info. -- cgit v1.2.3 From f8c09464f801e97b24ccdb1ba70444c60d4235bd Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Apr 2008 11:05:32 -0600 Subject: cell: enable #define CACHE_STATS to print a cache report upon exit --- src/gallium/drivers/cell/spu/spu_dcache.c | 18 ++++++++++++++++++ src/gallium/drivers/cell/spu/spu_dcache.h | 3 +++ src/gallium/drivers/cell/spu/spu_main.c | 2 ++ 3 files changed, 23 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_dcache.c b/src/gallium/drivers/cell/spu/spu_dcache.c index a1701d80d1..167404cdc5 100644 --- a/src/gallium/drivers/cell/spu/spu_dcache.c +++ b/src/gallium/drivers/cell/spu/spu_dcache.c @@ -36,6 +36,7 @@ #define CACHE_SET_TAGID(set) (((set) & 0x03) + TAG_DCACHE0) #define CACHE_LOG2NNWAY 2 #define CACHE_LOG2NSETS 6 +/*#define CACHE_STATS 1*/ #include /* Yes folks, this is ugly. @@ -123,3 +124,20 @@ spu_dcache_mark_dirty(unsigned ea, unsigned size) ? (entry & ~CACHELINE_VALID) : entry; } } + + +/** + * Print cache utilization report + */ +void +spu_dcache_report(void) +{ +#ifdef CACHE_STATS + if (spu.init.id == 0) { + printf("SPU 0: Texture cache report:\n"); + cache_pr_stats(data); + } +#endif +} + + diff --git a/src/gallium/drivers/cell/spu/spu_dcache.h b/src/gallium/drivers/cell/spu/spu_dcache.h index 7a06b8c25a..39a19eb31b 100644 --- a/src/gallium/drivers/cell/spu/spu_dcache.h +++ b/src/gallium/drivers/cell/spu/spu_dcache.h @@ -31,4 +31,7 @@ spu_dcache_fetch_unaligned(qword *dst, unsigned ea, unsigned size); extern void spu_dcache_mark_dirty(unsigned ea, unsigned size); +extern void +spu_dcache_report(void); + #endif /* SPU_DCACHE_H */ diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 7f0473d198..5b5a570a3c 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -629,6 +629,8 @@ main_loop(void) if (Debug) printf("SPU %u: Exit main loop\n", spu.init.id); + + spu_dcache_report(); } -- cgit v1.2.3 From 9d287a184111693a1fbb9231b1d7a8afa56e9ae5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Apr 2008 11:28:01 -0600 Subject: gallium: init ctx->Const.MaxTextureUnits --- src/mesa/state_tracker/st_extensions.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 7fab4d041d..20f4652413 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -83,7 +83,8 @@ void st_init_limits(struct st_context *st) c->MaxTextureRectSize = min(1 << (c->MaxTextureLevels - 1), MAX_TEXTURE_RECT_SIZE); - c->MaxTextureImageUnits + c->MaxTextureUnits + = c->MaxTextureImageUnits = c->MaxTextureCoordUnits = min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS), MAX_TEXTURE_IMAGE_UNITS); -- cgit v1.2.3 From 6ddd2df1aec329be494d342dbd88a9f5af5e7b2c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Apr 2008 11:28:27 -0600 Subject: cell: return CELL_MAX_SAMPLERS to indicate number of texture units --- src/gallium/drivers/cell/ppu/cell_screen.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c index 84b48bf4f1..5198b51441 100644 --- a/src/gallium/drivers/cell/ppu/cell_screen.c +++ b/src/gallium/drivers/cell/ppu/cell_screen.c @@ -31,6 +31,7 @@ #include "pipe/p_defines.h" #include "pipe/p_screen.h" +#include "cell/common.h" #include "cell_screen.h" #include "cell_texture.h" #include "cell_winsys.h" @@ -55,7 +56,7 @@ cell_get_param(struct pipe_screen *screen, int param) { switch (param) { case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: - return PIPE_MAX_SAMPLERS; + return CELL_MAX_SAMPLERS; case PIPE_CAP_NPOT_TEXTURES: return 0; case PIPE_CAP_TWO_SIDED_STENCIL: -- cgit v1.2.3 From c14da8f52407529f20f819e31a01356535de0117 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Apr 2008 11:30:17 -0600 Subject: cell: assert num samplers/textures <= CELL_MAX_SAMPLERS --- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 52c3126050..67b87f16d7 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -232,12 +232,12 @@ cell_bind_sampler_states(struct pipe_context *pipe, { struct cell_context *cell = cell_context(pipe); - draw_flush(cell->draw); + assert(num <= CELL_MAX_SAMPLERS); - assert(unit < PIPE_MAX_SAMPLERS); + draw_flush(cell->draw); memcpy(cell->sampler, samplers, num * sizeof(void *)); - memset(&cell->sampler[num], 0, (PIPE_MAX_SAMPLERS - num) * + memset(&cell->sampler[num], 0, (CELL_MAX_SAMPLERS - num) * sizeof(void *)); cell->num_samplers = num; @@ -261,6 +261,8 @@ cell_set_sampler_textures(struct pipe_context *pipe, struct cell_context *cell = cell_context(pipe); uint i; + assert(num <= CELL_MAX_SAMPLERS); + /* Check for no-op */ if (num == cell->num_textures && !memcmp(cell->texture, texture, num * sizeof(struct pipe_texture *))) @@ -268,7 +270,7 @@ cell_set_sampler_textures(struct pipe_context *pipe, draw_flush(cell->draw); - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + for (i = 0; i < CELL_MAX_SAMPLERS; i++) { struct pipe_texture *tex = i < num ? texture[i] : NULL; pipe_texture_reference((struct pipe_texture **) &cell->texture[i], tex); -- cgit v1.2.3 From e7b23d36df1ab3ac5b54ef8e4e56c4fd46db8257 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Apr 2008 11:35:53 -0600 Subject: cell: checkpoint: more multi-texture work --- src/gallium/drivers/cell/ppu/cell_texture.c | 8 +++++-- src/gallium/drivers/cell/spu/spu_main.c | 2 +- src/gallium/drivers/cell/spu/spu_main.h | 2 +- src/gallium/drivers/cell/spu/spu_texture.c | 6 ++--- src/gallium/drivers/cell/spu/spu_texture.h | 4 ++-- src/gallium/drivers/cell/spu/spu_tri.c | 34 +++++++++++++++++++++++++---- 6 files changed, 42 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index 9c694e136d..c59d1f7f23 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -245,9 +245,13 @@ void cell_update_texture_mapping(struct cell_context *cell) { uint face = 0, level = 0, zslice = 0; + uint i; + + for (i = 0; i < CELL_MAX_SAMPLERS; i++) { + if (cell->texture[i]) + cell_tile_texture(cell, cell->texture[i]); + } - if (cell->texture[0]) - cell_tile_texture(cell, cell->texture[0]); #if 0 if (cell->tex_surf && cell->tex_map) { pipe_surface_unmap(cell->tex_surf); diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 5b5a570a3c..a840d01596 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -333,7 +333,7 @@ cmd_state_texture(const struct cell_command_texture *texture) const uint width = texture->width; const uint height = texture->height; - if (Debug) { + if (1||Debug) { printf("SPU %u: TEXTURE [%u] at %p size %u x %u\n", spu.init.id, texture->unit, texture->start, texture->width, texture->height); diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index 2bfad3535a..26e050cfc3 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -152,7 +152,7 @@ struct spu_global /** for converting RGBA to PIPE_FORMAT_x colors */ vector unsigned char color_shuffle; - vector float (*sample_texture)(vector float texcoord); + vector float (*sample_texture)(uint unit, vector float texcoord); } ALIGN16_ATTRIB; diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c index 4612501eb3..58a426cc40 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.c +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -104,9 +104,8 @@ get_four_texels(vec_uint4 x, vec_uint4 y, vec_uint4 *texels) * XXX this is extremely primitive for now. */ vector float -sample_texture_nearest(vector float texcoord) +sample_texture_nearest(uint unit, vector float texcoord) { - const uint unit = 0; vector float tc = spu_mul(texcoord, spu.texture[unit].tex_size); vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ itc = spu_and(itc, spu.texture[unit].tex_size_mask); /* mask (GL_REPEAT) */ @@ -116,9 +115,8 @@ sample_texture_nearest(vector float texcoord) vector float -sample_texture_bilinear(vector float texcoord) +sample_texture_bilinear(uint unit, vector float texcoord) { - const uint unit = 0; static const vec_uint4 offset_x = {0, 0, 1, 1}; static const vec_uint4 offset_y = {0, 1, 0, 1}; diff --git a/src/gallium/drivers/cell/spu/spu_texture.h b/src/gallium/drivers/cell/spu/spu_texture.h index 95eb87080f..f7c9738be8 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.h +++ b/src/gallium/drivers/cell/spu/spu_texture.h @@ -37,11 +37,11 @@ invalidate_tex_cache(void); extern vector float -sample_texture_nearest(vector float texcoord); +sample_texture_nearest(uint unit, vector float texcoord); extern vector float -sample_texture_bilinear(vector float texcoord); +sample_texture_bilinear(uint unit, vector float texcoord); #endif /* SPU_TEXTURE_H */ diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index 17e337bbdf..51abcb1757 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -311,17 +311,43 @@ emit_quad( int x, int y, mask_t mask ) if (spu.texture[0].start) { /* texture mapping */ + const uint unit = 0; vector float texcoords[4]; eval_coeff(2, (float) x, (float) y, texcoords); if (spu_extract(mask, 0)) - colors[0] = spu.sample_texture(texcoords[0]); + colors[0] = spu.sample_texture(unit, texcoords[0]); if (spu_extract(mask, 1)) - colors[1] = spu.sample_texture(texcoords[1]); + colors[1] = spu.sample_texture(unit, texcoords[1]); if (spu_extract(mask, 2)) - colors[2] = spu.sample_texture(texcoords[2]); + colors[2] = spu.sample_texture(unit, texcoords[2]); if (spu_extract(mask, 3)) - colors[3] = spu.sample_texture(texcoords[3]); + colors[3] = spu.sample_texture(unit, texcoords[3]); + + + if (spu.texture[1].start) { + /* multi-texture mapping */ + const uint unit = 1; + vector float colors1[4]; + + eval_coeff(3, (float) x, (float) y, texcoords); + + if (spu_extract(mask, 0)) + colors1[0] = spu.sample_texture(unit, texcoords[0]); + if (spu_extract(mask, 1)) + colors1[1] = spu.sample_texture(unit, texcoords[1]); + if (spu_extract(mask, 2)) + colors1[2] = spu.sample_texture(unit, texcoords[2]); + if (spu_extract(mask, 3)) + colors1[3] = spu.sample_texture(unit, texcoords[3]); + + /* hack: modulate first texture by second */ + colors[0] = spu_mul(colors[0], colors1[0]); + colors[1] = spu_mul(colors[1], colors1[1]); + colors[2] = spu_mul(colors[2], colors1[2]); + colors[3] = spu_mul(colors[3], colors1[3]); + } + } else { /* simple shading */ -- cgit v1.2.3 From 2d02ee85093d8068f11ecb286c5f02f52786cd95 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Apr 2008 14:52:04 -0600 Subject: cell: fix bug in texture tiling function (non-square textures work now) --- src/gallium/drivers/cell/ppu/cell_texture.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index c59d1f7f23..07717da8f6 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -186,15 +186,17 @@ tile_copy_data(uint w, uint h, uint tile_size, uint *dst, const uint *src) uint it, jt; /* tile counters */ uint i, j; /* intra-tile counters */ + /* loop over dest tiles */ for (it = 0; it < h_t; it++) { for (jt = 0; jt < w_t; jt++) { - /* fill in tile (i, j) */ + /* start of dest tile: */ uint *tdst = dst + (it * w_t + jt) * tile_size2; + /* loop over texels in the tile */ for (i = 0; i < tile_size; i++) { for (j = 0; j < tile_size; j++) { const uint srci = it * tile_size + i; const uint srcj = jt * tile_size + j; - *tdst++ = src[srci * h + srcj]; + *tdst++ = src[srci * w + srcj]; } } } -- cgit v1.2.3 From 9e7b730eb25f08065d718dee4a67c67d1d133b6e Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Apr 2008 14:52:25 -0600 Subject: cell: pass tex unit to get_texel() --- src/gallium/drivers/cell/spu/spu_texture.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c index 58a426cc40..ba0d57b959 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.c +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -49,9 +49,8 @@ invalidate_tex_cache(void) static uint -get_texel(vec_uint4 coordinate) +get_texel(uint unit, vec_uint4 coordinate) { - const uint unit = 0; vec_uint4 tmp; unsigned x = spu_extract(coordinate, 0); unsigned y = spu_extract(coordinate, 1); @@ -109,7 +108,7 @@ sample_texture_nearest(uint unit, vector float texcoord) vector float tc = spu_mul(texcoord, spu.texture[unit].tex_size); vector unsigned int itc = spu_convtu(tc, 0); /* convert to int */ itc = spu_and(itc, spu.texture[unit].tex_size_mask); /* mask (GL_REPEAT) */ - uint texel = get_texel(itc); + uint texel = get_texel(unit, itc); return spu_unpack_A8R8G8B8(texel); } -- cgit v1.2.3 From 9d1444092fbfd9f975cfb996695f0533a78410f7 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Apr 2008 14:55:31 -0600 Subject: cell: turn off some debug output --- src/gallium/drivers/cell/spu/spu_main.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index a840d01596..5b5a570a3c 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -333,7 +333,7 @@ cmd_state_texture(const struct cell_command_texture *texture) const uint width = texture->width; const uint height = texture->height; - if (1||Debug) { + if (Debug) { printf("SPU %u: TEXTURE [%u] at %p size %u x %u\n", spu.init.id, texture->unit, texture->start, texture->width, texture->height); -- cgit v1.2.3 From bccd3f138ccc717fad9073110d15e3321b590476 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 1 Apr 2008 15:42:42 -0600 Subject: cell: more multi-texture fixes (mostly working now) --- src/gallium/drivers/cell/spu/spu_main.c | 6 +++--- src/gallium/drivers/cell/spu/spu_main.h | 4 +++- src/gallium/drivers/cell/spu/spu_texture.c | 5 ++--- src/gallium/drivers/cell/spu/spu_tri.c | 18 +++++++++--------- 4 files changed, 17 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 5b5a570a3c..1ab1c40379 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -319,10 +319,10 @@ cmd_state_sampler(const struct cell_command_sampler *sampler) spu.init.id, sampler->unit); spu.sampler[sampler->unit] = sampler->state; - if (spu.sampler[0].min_img_filter == PIPE_TEX_FILTER_LINEAR) - spu.sample_texture = sample_texture_bilinear; + if (spu.sampler[sampler->unit].min_img_filter == PIPE_TEX_FILTER_LINEAR) + spu.sample_texture[sampler->unit] = sample_texture_bilinear; else - spu.sample_texture = sample_texture_nearest; + spu.sample_texture[sampler->unit] = sample_texture_nearest; } diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index 26e050cfc3..e9e39cbeab 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -85,6 +85,8 @@ typedef struct spu_blend_results (*logicop_func)( qword frag_mask); +typedef vector float (*sample_texture_func)(uint unit, vector float texcoord); + struct spu_framebuffer { void *color_start; /**< addr of color surface in main memory */ void *depth_start; /**< addr of depth surface in main memory */ @@ -152,7 +154,7 @@ struct spu_global /** for converting RGBA to PIPE_FORMAT_x colors */ vector unsigned char color_shuffle; - vector float (*sample_texture)(uint unit, vector float texcoord); + sample_texture_func sample_texture[CELL_MAX_SAMPLERS]; } ALIGN16_ATTRIB; diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c index ba0d57b959..ceab246980 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.c +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -68,9 +68,8 @@ get_texel(uint unit, vec_uint4 coordinate) static void -get_four_texels(vec_uint4 x, vec_uint4 y, vec_uint4 *texels) +get_four_texels(uint unit, vec_uint4 x, vec_uint4 y, vec_uint4 *texels) { - const uint unit = 0; const unsigned texture_ea = (uintptr_t) spu.texture[unit].start; vec_uint4 tile_x = spu_rlmask(x, -5); vec_uint4 tile_y = spu_rlmask(y, -5); @@ -136,7 +135,7 @@ sample_texture_bilinear(uint unit, vector float texcoord) x = spu_and(x, spu.texture[unit].tex_size_x_mask); y = spu_and(y, spu.texture[unit].tex_size_y_mask); - get_four_texels(x, y, texels); + get_four_texels(unit, x, y, texels); vector float texel00 = spu_unpack_A8R8G8B8(spu_extract(texels[0], 0)); vector float texel01 = spu_unpack_A8R8G8B8(spu_extract(texels[1], 0)); diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index 51abcb1757..ab4ff8160a 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -316,13 +316,13 @@ emit_quad( int x, int y, mask_t mask ) eval_coeff(2, (float) x, (float) y, texcoords); if (spu_extract(mask, 0)) - colors[0] = spu.sample_texture(unit, texcoords[0]); + colors[0] = spu.sample_texture[unit](unit, texcoords[0]); if (spu_extract(mask, 1)) - colors[1] = spu.sample_texture(unit, texcoords[1]); + colors[1] = spu.sample_texture[unit](unit, texcoords[1]); if (spu_extract(mask, 2)) - colors[2] = spu.sample_texture(unit, texcoords[2]); + colors[2] = spu.sample_texture[unit](unit, texcoords[2]); if (spu_extract(mask, 3)) - colors[3] = spu.sample_texture(unit, texcoords[3]); + colors[3] = spu.sample_texture[unit](unit, texcoords[3]); if (spu.texture[1].start) { @@ -330,16 +330,16 @@ emit_quad( int x, int y, mask_t mask ) const uint unit = 1; vector float colors1[4]; - eval_coeff(3, (float) x, (float) y, texcoords); + eval_coeff(2, (float) x, (float) y, texcoords); if (spu_extract(mask, 0)) - colors1[0] = spu.sample_texture(unit, texcoords[0]); + colors1[0] = spu.sample_texture[unit](unit, texcoords[0]); if (spu_extract(mask, 1)) - colors1[1] = spu.sample_texture(unit, texcoords[1]); + colors1[1] = spu.sample_texture[unit](unit, texcoords[1]); if (spu_extract(mask, 2)) - colors1[2] = spu.sample_texture(unit, texcoords[2]); + colors1[2] = spu.sample_texture[unit](unit, texcoords[2]); if (spu_extract(mask, 3)) - colors1[3] = spu.sample_texture(unit, texcoords[3]); + colors1[3] = spu.sample_texture[unit](unit, texcoords[3]); /* hack: modulate first texture by second */ colors[0] = spu_mul(colors[0], colors1[0]); -- cgit v1.2.3 From bc739440c29c551fcc44e9e12d0d9c170d8d24fb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 2 Apr 2008 10:43:37 +0100 Subject: gallium: add temporary facility for rasterization-time clamping of point sizes --- src/gallium/auxiliary/draw/draw_wide_point.c | 14 +++++++++++++- src/gallium/include/pipe/p_state.h | 2 ++ src/mesa/state_tracker/st_atom_rasterizer.c | 4 ++++ 3 files changed, 19 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_wide_point.c b/src/gallium/auxiliary/draw/draw_wide_point.c index c53f7e6cb3..86281ca3d8 100644 --- a/src/gallium/auxiliary/draw/draw_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_wide_point.c @@ -38,6 +38,8 @@ struct widepoint_stage { struct draw_stage stage; float half_point_size; + float point_size_min; + float point_size_max; uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; @@ -128,7 +130,15 @@ static void widepoint_point( struct draw_stage *stage, /* point size is either per-vertex or fixed size */ if (wide->psize_slot >= 0) { - half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; + half_size = header->v[0]->data[wide->psize_slot][0]; + + /* XXX: temporary -- do this in the vertex shader?? + */ + half_size = CLAMP(half_size, + wide->point_size_min, + wide->point_size_max); + + half_size *= 0.5f; } else { half_size = wide->half_point_size; @@ -182,6 +192,8 @@ static void widepoint_first_point( struct draw_stage *stage, struct draw_context *draw = stage->draw; wide->half_point_size = 0.5f * draw->rasterizer->point_size; + wide->point_size_min = draw->rasterizer->point_size_min; + wide->point_size_max = draw->rasterizer->point_size_max; /* XXX we won't know the real size if it's computed by the vertex shader! */ if ((draw->rasterizer->point_size > draw->wide_point_threshold) || diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index a73028814e..e407e3bc72 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -118,6 +118,8 @@ struct pipe_rasterizer_state float line_width; float point_size; /**< used when no per-vertex size */ + float point_size_min; /* XXX - temporary, will go away */ + float point_size_max; /* XXX - temporary, will go away */ float offset_units; float offset_scale; ubyte sprite_coord_mode[PIPE_MAX_SHADER_OUTPUTS]; /**< PIPE_SPRITE_COORD_ */ diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 17d77f90ae..14c26c16c0 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -198,6 +198,10 @@ static void update_raster_state( struct st_context *st ) /* _NEW_POINT */ raster->point_size = ctx->Point.Size; + + raster->point_size_min = 0; /* temporary, will go away */ + raster->point_size_max = 1000; /* temporary, will go away */ + raster->point_smooth = ctx->Point.SmoothFlag; raster->point_sprite = ctx->Point.PointSprite; for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) { -- cgit v1.2.3 From 8e33194837dd206d920889851d9cf22190100c99 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 2 Apr 2008 11:38:33 +0100 Subject: gallium: add a flag to turn on gl rasterization rules Use this to set up hardware rasterization (if your hardware can do it) or otherwise turn on various tweaks in the draw module. Currently only hooked up to point biasing code. --- src/gallium/auxiliary/draw/draw_wide_point.c | 19 +++++++++++++------ src/gallium/include/pipe/p_state.h | 1 + src/mesa/state_tracker/st_atom_rasterizer.c | 2 ++ 3 files changed, 16 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_wide_point.c b/src/gallium/auxiliary/draw/draw_wide_point.c index 86281ca3d8..6fc7c9fcd7 100644 --- a/src/gallium/auxiliary/draw/draw_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_wide_point.c @@ -41,6 +41,9 @@ struct widepoint_stage { float point_size_min; float point_size_max; + float xbias; + float ybias; + uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; uint num_texcoords; @@ -126,8 +129,6 @@ static void widepoint_point( struct draw_stage *stage, float *pos2 = v2->data[0]; float *pos3 = v3->data[0]; - const float xbias = 0.0, ybias = -0.125; - /* point size is either per-vertex or fixed size */ if (wide->psize_slot >= 0) { half_size = header->v[0]->data[wide->psize_slot][0]; @@ -144,10 +145,10 @@ static void widepoint_point( struct draw_stage *stage, half_size = wide->half_point_size; } - left_adj = -half_size + xbias; - right_adj = half_size + xbias; - bot_adj = half_size + ybias; - top_adj = -half_size + ybias; + left_adj = -half_size + wide->xbias; + right_adj = half_size + wide->xbias; + bot_adj = half_size + wide->ybias; + top_adj = -half_size + wide->ybias; pos0[0] += left_adj; pos0[1] += top_adj; @@ -194,6 +195,12 @@ static void widepoint_first_point( struct draw_stage *stage, wide->half_point_size = 0.5f * draw->rasterizer->point_size; wide->point_size_min = draw->rasterizer->point_size_min; wide->point_size_max = draw->rasterizer->point_size_max; + wide->xbias = 0.0; + wide->ybias = 0.0; + + if (draw->rasterizer->gl_rasterization_rules) { + wide->ybias = -0.125; + } /* XXX we won't know the real size if it's computed by the vertex shader! */ if ((draw->rasterizer->point_size > draw->wide_point_threshold) || diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index e407e3bc72..3593446e1c 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -115,6 +115,7 @@ struct pipe_rasterizer_state still needed though, to indicate inputs/outputs */ unsigned origin_lower_left:1; /**< Is (0,0) the lower-left corner? */ unsigned flatshade_first:1; /**< take color attribute from the first vertex of a primitive */ + unsigned gl_rasterization_rules:1; /**< enable tweaks for GL rasterization? */ float line_width; float point_size; /**< used when no per-vertex size */ diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 14c26c16c0..bb14cf9045 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -261,6 +261,8 @@ static void update_raster_state( struct st_context *st ) if (ctx->Scissor.Enabled) raster->scissor = 1; + raster->gl_rasterization_rules = 1; + cso_set_rasterizer(st->cso_context, raster); } -- cgit v1.2.3 From ae3c91e98ce3355bca22738440f8ba313b3b8b23 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 2 Apr 2008 11:55:03 +0100 Subject: draw: Set the backend prim in the pt 'prepare' operation Leaving it until 'run' is bad as the primitive is pretty much state for some drivers and so needs to get set early. In some drivers this is used to determine things like vertex format, etc -- by the time we get to 'run', it's too late to change this. --- src/gallium/auxiliary/draw/draw_pt.c | 3 +-- src/gallium/auxiliary/draw/draw_pt.h | 6 ++--- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 33 ++++++++++++++----------- src/gallium/auxiliary/draw/draw_pt_vcache.c | 17 +++++++------ 4 files changed, 31 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 2ea96c686d..c3baf5b7da 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -171,10 +171,9 @@ draw_pt_arrays(struct draw_context *draw, */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); - frontend->prepare( frontend, middle ); + frontend->prepare( frontend, prim, middle ); frontend->run( frontend, - prim, draw_pt_elt_func( draw ), draw_pt_elt_ptr( draw, start ), count ); diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index f878616079..428b1e0e6b 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -53,10 +53,10 @@ struct draw_context; */ struct draw_pt_front_end { void (*prepare)( struct draw_pt_front_end *, + unsigned prim, struct draw_pt_middle_end * ); void (*run)( struct draw_pt_front_end *, - unsigned prim, pt_elt_func elt_func, const void *elt_ptr, unsigned count ); @@ -79,10 +79,10 @@ struct draw_pt_front_end { * Currenly only using the passthrough version. */ struct draw_pt_middle_end { - void (*prepare)( struct draw_pt_middle_end * ); + void (*prepare)( struct draw_pt_middle_end *, + unsigned prim ); void (*run)( struct draw_pt_middle_end *, - unsigned prim, const unsigned *fetch_elts, unsigned fetch_count, const ushort *draw_elts, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 9b098bc173..9339cf1f88 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -203,16 +203,29 @@ fetch_store_general( struct fetch_emit_middle_end *feme, -static void fetch_emit_prepare( struct draw_pt_middle_end *middle ) +static void fetch_emit_prepare( struct draw_pt_middle_end *middle, + unsigned prim ) { static const float zero = 0; struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; struct draw_context *draw = feme->draw; - const struct vertex_info *vinfo = draw->render->get_vertex_info(draw->render); - unsigned nr_attrs = vinfo->num_attribs; + const struct vertex_info *vinfo; unsigned i; + boolean ok; + + + ok = draw->render->set_primitive( draw->render, + prim ); + if (!ok) { + assert(0); + return; + } + + /* Must do this after set_primitive() above: + */ + vinfo = draw->render->get_vertex_info(draw->render); - for (i = 0; i < nr_attrs; i++) { + for (i = 0; i < vinfo->num_attribs; i++) { unsigned src_element = vinfo->src_index[i]; unsigned src_buffer = draw->vertex_element[src_element].vertex_buffer_index; @@ -275,7 +288,7 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle ) } } - feme->nr_fetch = nr_attrs; + feme->nr_fetch = vinfo->num_attribs; feme->hw_vertex_size = vinfo->size * 4; } @@ -284,7 +297,6 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle ) static void fetch_emit_run( struct draw_pt_middle_end *middle, - unsigned prim, const unsigned *fetch_elts, unsigned fetch_count, const ushort *draw_elts, @@ -293,16 +305,7 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; struct draw_context *draw = feme->draw; void *hw_verts; - boolean ok; - - ok = draw->render->set_primitive( draw->render, - prim ); - if (!ok) { - assert(0); - return; - } - hw_verts = draw->render->allocate_vertices( draw->render, (ushort)feme->hw_vertex_size, (ushort)fetch_count ); diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 16ffedf580..5a068761df 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -59,6 +59,8 @@ struct vcache_frontend { const void *elt_ptr; struct draw_pt_middle_end *middle; + + unsigned input_prim; unsigned output_prim; }; @@ -77,7 +79,6 @@ static void vcache_flush( struct vcache_frontend *vcache ) if (vcache->draw_count) { vcache->middle->run( vcache->middle, - vcache->output_prim, vcache->fetch_elts, vcache->fetch_count, vcache->draw_elts, @@ -173,7 +174,6 @@ static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { static void vcache_run_pv2( struct draw_pt_front_end *frontend, - unsigned prim, pt_elt_func get_elt, const void *elts, unsigned count ) @@ -185,9 +185,8 @@ static void vcache_run_pv2( struct draw_pt_front_end *frontend, */ vcache->elt_func = get_elt; vcache->elt_ptr = elts; - vcache->output_prim = reduced_prim[prim]; - switch (prim) { + switch (vcache->input_prim) { case PIPE_PRIM_POINTS: for (i = 0; i < count; i ++) { vcache_point( vcache, @@ -304,7 +303,6 @@ static void vcache_run_pv2( struct draw_pt_front_end *frontend, static void vcache_run_pv0( struct draw_pt_front_end *frontend, - unsigned prim, pt_elt_func get_elt, const void *elts, unsigned count ) @@ -316,9 +314,8 @@ static void vcache_run_pv0( struct draw_pt_front_end *frontend, */ vcache->elt_func = get_elt; vcache->elt_ptr = elts; - vcache->output_prim = reduced_prim[prim]; - switch (prim) { + switch (vcache->input_prim) { case PIPE_PRIM_POINTS: for (i = 0; i < count; i ++) { vcache_point( vcache, @@ -389,6 +386,7 @@ static void vcache_run_pv0( struct draw_pt_front_end *frontend, } static void vcache_prepare( struct draw_pt_front_end *frontend, + unsigned prim, struct draw_pt_middle_end *middle ) { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; @@ -398,8 +396,11 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, else vcache->base.run = vcache_run_pv2; + vcache->input_prim = prim; + vcache->output_prim = reduced_prim[prim]; + vcache->middle = middle; - middle->prepare( middle ); + middle->prepare( middle, vcache->output_prim ); } -- cgit v1.2.3 From add46fbc8cc04d3bce303815541a7bc5d0b33953 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 2 Apr 2008 12:05:55 +0100 Subject: draw: add missing break statement --- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 9339cf1f88..f863a46d9c 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -281,6 +281,7 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, feme->fetch[i].pitch = 0; feme->fetch[i].fetch = fetch_R32_FLOAT; feme->fetch[i].emit = emit_R32_FLOAT; + break; default: assert(0); feme->fetch[i].emit = NULL; -- cgit v1.2.3 From d2cb4ba0bb2388c784f145c59f3798f914dc7f39 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 3 Apr 2008 12:21:30 +0100 Subject: draw: add passthrough path to the pipeline This handles the case where bypass_vs is set, but vertices need to go through the pipeline for some reason - eg unfilled polygon mode. Demonstrates how to drive the pipeline from inside one of these things. --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/SConscript | 1 + src/gallium/auxiliary/draw/draw_private.h | 6 + src/gallium/auxiliary/draw/draw_pt.c | 23 +- src/gallium/auxiliary/draw/draw_pt.h | 1 + src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 64 +--- .../auxiliary/draw/draw_pt_fetch_pipeline.c | 391 +++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vertex_fetch.c | 4 +- 8 files changed, 424 insertions(+), 67 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index a0db2e4555..b28e516396 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -18,6 +18,7 @@ C_SOURCES = \ draw_pt.c \ draw_pt_vcache.c \ draw_pt_fetch_emit.c \ + draw_pt_fetch_pipeline.c \ draw_pt_elts.c \ draw_prim.c \ draw_pstipple.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 981225a8c2..9ca4197441 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -17,6 +17,7 @@ draw = env.ConvenienceLibrary( 'draw_pt.c', 'draw_pt_vcache.c', 'draw_pt_fetch_emit.c', + 'draw_pt_fetch_pipeline.c', 'draw_pt_elts.c', 'draw_prim.c', 'draw_pstipple.c', diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 9a9b25297f..0c9f9c2e03 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -160,6 +160,11 @@ struct draw_vertex_shader { /* Internal function for vertex fetch. */ typedef void (*fetch_func)(const void *ptr, float *attrib); + +fetch_func draw_get_fetch_func( enum pipe_format format ); + + + typedef void (*full_fetch_func)( struct draw_context *draw, struct tgsi_exec_machine *machine, const unsigned *elts, @@ -211,6 +216,7 @@ struct draw_context struct { struct draw_pt_middle_end *fetch_emit; + struct draw_pt_middle_end *fetch_pipeline; struct draw_pt_middle_end *fetch_shade_emit; struct draw_pt_middle_end *fetch_shade_cliptest_pipeline_or_emit; } middle; diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index c3baf5b7da..f59fb86f78 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -73,7 +73,6 @@ draw_pt_arrays(struct draw_context *draw, */ -#if 0 if (!cliptest && !pipeline && !shading) { /* This is the 'passthrough' path: */ @@ -81,6 +80,14 @@ draw_pt_arrays(struct draw_context *draw, */ middle = draw->pt.middle.fetch_emit; } + else if (!cliptest && !shading) { + /* This is the 'passthrough' path targetting the pipeline backend. + */ + /* Fetch user verts, emit pipeline verts, run pipeline: + */ + middle = draw->pt.middle.fetch_pipeline; + } +#if 0 else if (!cliptest && !pipeline) { /* Fetch user verts, run vertex shader, emit hw verts: */ @@ -117,10 +124,9 @@ draw_pt_arrays(struct draw_context *draw, middle = draw->pt.middle.fetch_shade_cliptest_pipeline; } #else - if (cliptest || pipeline || shading) + else { return FALSE; - - middle = draw->pt.middle.fetch_emit; + } #endif @@ -190,6 +196,10 @@ boolean draw_pt_init( struct draw_context *draw ) if (!draw->pt.middle.fetch_emit) return FALSE; + draw->pt.middle.fetch_pipeline = draw_pt_fetch_pipeline( draw ); + if (!draw->pt.middle.fetch_pipeline) + return FALSE; + draw->pt.front.vcache = draw_pt_vcache( draw ); if (!draw->pt.front.vcache) return FALSE; @@ -205,6 +215,11 @@ void draw_pt_destroy( struct draw_context *draw ) draw->pt.middle.fetch_emit = NULL; } + if (draw->pt.middle.fetch_pipeline) { + draw->pt.middle.fetch_pipeline->destroy( draw->pt.middle.fetch_pipeline ); + draw->pt.middle.fetch_pipeline = NULL; + } + if (draw->pt.front.vcache) { draw->pt.front.vcache->destroy( draw->pt.front.vcache ); draw->pt.front.vcache = NULL; diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 428b1e0e6b..800072c511 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -112,6 +112,7 @@ const void *draw_pt_elt_ptr( struct draw_context *draw, */ struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); +struct draw_pt_middle_end *draw_pt_fetch_pipeline( struct draw_context *draw ); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index f863a46d9c..39f0b40838 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -85,45 +85,6 @@ struct fetch_emit_middle_end { }; -static void fetch_B8G8R8A8_UNORM( const void *from, - float *attrib ) -{ - ubyte *ub = (ubyte *) from; - attrib[2] = UBYTE_TO_FLOAT(ub[0]); - attrib[1] = UBYTE_TO_FLOAT(ub[1]); - attrib[0] = UBYTE_TO_FLOAT(ub[2]); - attrib[3] = UBYTE_TO_FLOAT(ub[3]); -} - -static void fetch_R32G32B32A32_FLOAT( const void *from, - float *attrib ) -{ - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = f[1]; - attrib[2] = f[2]; - attrib[3] = f[3]; -} - -static void fetch_R32G32B32_FLOAT( const void *from, - float *attrib ) -{ - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = f[1]; - attrib[2] = f[2]; - attrib[3] = 1.0; -} - -static void fetch_R32G32_FLOAT( const void *from, - float *attrib ) -{ - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = f[1]; - attrib[2] = 0.0; - attrib[3] = 1.0; -} static void fetch_R32_FLOAT( const void *from, float *attrib ) @@ -235,28 +196,9 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, feme->fetch[i].pitch = draw->vertex_buffer[src_buffer].pitch; - switch (draw->vertex_element[src_element].src_format) { - case PIPE_FORMAT_B8G8R8A8_UNORM: - feme->fetch[i].fetch = fetch_B8G8R8A8_UNORM; - break; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - feme->fetch[i].fetch = fetch_R32G32B32A32_FLOAT; - break; - case PIPE_FORMAT_R32G32B32_FLOAT: - feme->fetch[i].fetch = fetch_R32G32B32_FLOAT; - break; - case PIPE_FORMAT_R32G32_FLOAT: - feme->fetch[i].fetch = fetch_R32G32_FLOAT; - break; - case PIPE_FORMAT_R32_FLOAT: - feme->fetch[i].fetch = fetch_R32_FLOAT; - break; - default: - assert(0); - feme->fetch[i].fetch = NULL; - break; - } - + feme->fetch[i].fetch = draw_get_fetch_func(draw->vertex_element[src_element].src_format); + + switch (vinfo->emit[i]) { case EMIT_4F: feme->fetch[i].emit = emit_R32G32B32A32_FLOAT; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c new file mode 100644 index 0000000000..94e7d01be4 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c @@ -0,0 +1,391 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" + +/* The simplest 'middle end' in the new vertex code. + * + * The responsibilities of a middle end are to: + * - perform vertex fetch using + * - draw vertex element/buffer state + * - a list of fetch indices we received as an input + * - run the vertex shader + * - cliptest, + * - clip coord calculation + * - viewport transformation + * - if necessary, run the primitive pipeline, passing it: + * - a linear array of vertex_header vertices constructed here + * - a set of draw indices we received as an input + * - otherwise, drive the hw backend, + * - allocate space for hardware format vertices + * - translate the vertex-shader output vertices to hw format + * - calling the backend draw functions. + * + * For convenience, we provide a helper function to drive the hardware + * backend given similar inputs to those required to run the pipeline. + * + * In the case of passthrough mode, many of these actions are disabled + * or noops, so we end up doing: + * + * - perform vertex fetch + * - drive the hw backend + * + * IE, basically just vertex fetch to post-vs-format vertices, + * followed by a call to the backend helper function. + */ + + +struct fetch_pipeline_middle_end { + struct draw_pt_middle_end base; + struct draw_context *draw; + + struct { + const ubyte *ptr; + unsigned pitch; + void (*fetch)( const void *from, float *attrib); + void (*emit)( const float *attrib, float **out ); + } fetch[PIPE_MAX_ATTRIBS]; + + unsigned nr_fetch; + unsigned pipeline_vertex_size; + unsigned prim; +}; + + +static void fetch_NULL( const void *from, + float *attrib ) +{ +} + + + +static void emit_R32_FLOAT( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out) += 1; +} + +static void emit_R32G32_FLOAT( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out)[1] = attrib[1]; + (*out) += 2; +} + +static void emit_R32G32B32_FLOAT( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out)[1] = attrib[1]; + (*out)[2] = attrib[2]; + (*out) += 3; +} + +static void emit_R32G32B32A32_FLOAT( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out)[1] = attrib[1]; + (*out)[2] = attrib[2]; + (*out)[3] = attrib[3]; + (*out) += 4; +} + +static void emit_header( const float *attrib, + float **out ) +{ + (*out)[0] = 0; + (*out)[1] = 0; + (*out)[2] = 0; + (*out)[3] = 0; + (*out)[3] = 1; + (*out) += 5; +} + +/** + * General-purpose fetch from user's vertex arrays, emit to driver's + * vertex buffer. + * + * XXX this is totally temporary. + */ +static void +fetch_store_general( struct fetch_pipeline_middle_end *fpme, + void *out_ptr, + const unsigned *fetch_elts, + unsigned count ) +{ + float *out = (float *)out_ptr; + uint i, j; + + for (i = 0; i < count; i++) { + unsigned elt = fetch_elts[i]; + + for (j = 0; j < fpme->nr_fetch; j++) { + float attrib[4]; + const ubyte *from = (fpme->fetch[j].ptr + + fpme->fetch[j].pitch * elt); + + fpme->fetch[j].fetch( from, attrib ); + fpme->fetch[j].emit( attrib, &out ); + } + } +} + + +/* We aren't running a vertex shader, but are running the pipeline. + * That means the vertices we need to build look like: + * + * dw0: vertex header (zero?) + * dw1: clip coord 0 + * dw2: clip coord 1 + * dw3: clip coord 2 + * dw4: clip coord 4 + * dw5: screen coord 0 + * dw6: screen coord 0 + * dw7: screen coord 0 + * dw8: screen coord 0 + * dw9: other attribs... + * + */ +static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, + unsigned prim ) +{ + static const float zero = 0; + struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; + struct draw_context *draw = fpme->draw; + unsigned i, nr = 0; + + fpme->prim = prim; + + /* Emit the vertex header and empty clipspace coord field: + */ + { + fpme->fetch[nr].ptr = NULL; + fpme->fetch[nr].pitch = 0; + fpme->fetch[nr].fetch = fetch_NULL; + fpme->fetch[nr].emit = emit_header; + nr++; + } + + + /* Need to look at vertex shader inputs (we know it is a + * passthrough shader, so these define the outputs too). If we + * were running a shader, we'd still be looking at the inputs at + * this point. + */ + for (i = 0; i < draw->vertex_shader->info.num_inputs; i++) { + unsigned buf = draw->vertex_element[i].vertex_buffer_index; + enum pipe_format format = draw->vertex_element[i].src_format; + + fpme->fetch[nr].ptr = ((const ubyte *) draw->user.vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[i].src_offset); + + fpme->fetch[nr].pitch = draw->vertex_buffer[buf].pitch; + fpme->fetch[nr].fetch = draw_get_fetch_func( format ); + + /* Always do this -- somewhat redundant... + */ + fpme->fetch[nr].emit = emit_R32G32B32A32_FLOAT; + nr++; + } + + fpme->nr_fetch = nr; + fpme->pipeline_vertex_size = (5 + (nr-1) * 4) * sizeof(float); +} + + + +/** + * Add a point to the primitive queue. + * \param i0 index into user's vertex arrays + */ +static void do_point( struct draw_context *draw, + const char *v0 ) +{ + struct prim_header prim; + + prim.reset_line_stipple = 0; + prim.edgeflags = 1; + prim.pad = 0; + prim.v[0] = (struct vertex_header *)v0; + + draw->pipeline.first->point( draw->pipeline.first, &prim ); +} + + +/** + * Add a line to the primitive queue. + * \param i0 index into user's vertex arrays + * \param i1 index into user's vertex arrays + */ +static void do_line( struct draw_context *draw, + const char *v0, + const char *v1 ) +{ + struct prim_header prim; + + prim.reset_line_stipple = 1; /* fixme */ + prim.edgeflags = 1; + prim.pad = 0; + prim.v[0] = (struct vertex_header *)v0; + prim.v[1] = (struct vertex_header *)v1; + + draw->pipeline.first->line( draw->pipeline.first, &prim ); +} + +/** + * Add a triangle to the primitive queue. + */ +static void do_triangle( struct draw_context *draw, + char *v0, + char *v1, + char *v2 ) +{ + struct prim_header prim; + +// _mesa_printf("tri %d %d %d\n", i0, i1, i2); + prim.reset_line_stipple = 1; + prim.edgeflags = ~0; + prim.pad = 0; + prim.v[0] = (struct vertex_header *)v0; + prim.v[1] = (struct vertex_header *)v1; + prim.v[2] = (struct vertex_header *)v2; + + draw->pipeline.first->tri( draw->pipeline.first, &prim ); +} + + +static void run_pipeline( struct fetch_pipeline_middle_end *fpme, + char *verts, + const ushort *elts, + unsigned count ) +{ + struct draw_context *draw = fpme->draw; + unsigned stride = fpme->pipeline_vertex_size; + unsigned i; + + switch (fpme->prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i++) + do_point( draw, + verts + stride * elts[i] ); + break; + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) + do_line( draw, + verts + stride * elts[i+0], + verts + stride * elts[i+1]); + break; + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) + do_triangle( draw, + verts + stride * elts[i+0], + verts + stride * elts[i+1], + verts + stride * elts[i+2]); + break; + } +} + + + + +static void fetch_pipeline_run( struct draw_pt_middle_end *middle, + const unsigned *fetch_elts, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; + struct draw_context *draw = fpme->draw; + char *pipeline_verts; + + pipeline_verts = MALLOC( fpme->pipeline_vertex_size * + fetch_count ); + if (!pipeline_verts) { + assert(0); + return; + } + + + /* Single routine to fetch vertices and emit pipeline verts. + */ + fetch_store_general( fpme, + pipeline_verts, + fetch_elts, + fetch_count ); + + + run_pipeline( fpme, + pipeline_verts, + draw_elts, + draw_count ); + + + /* Done -- that was easy, wasn't it: + */ + FREE( pipeline_verts ); +} + + + +static void fetch_pipeline_finish( struct draw_pt_middle_end *middle ) +{ + /* nothing to do */ +} + +static void fetch_pipeline_destroy( struct draw_pt_middle_end *middle ) +{ + FREE(middle); +} + + +struct draw_pt_middle_end *draw_pt_fetch_pipeline( struct draw_context *draw ) +{ + struct fetch_pipeline_middle_end *fetch_pipeline = CALLOC_STRUCT( fetch_pipeline_middle_end ); + + fetch_pipeline->base.prepare = fetch_pipeline_prepare; + fetch_pipeline->base.run = fetch_pipeline_run; + fetch_pipeline->base.finish = fetch_pipeline_finish; + fetch_pipeline->base.destroy = fetch_pipeline_destroy; + + fetch_pipeline->draw = draw; + + return &fetch_pipeline->base; +} + diff --git a/src/gallium/auxiliary/draw/draw_vertex_fetch.c b/src/gallium/auxiliary/draw/draw_vertex_fetch.c index 11f99babf6..9041041006 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_fetch.c +++ b/src/gallium/auxiliary/draw/draw_vertex_fetch.c @@ -166,7 +166,7 @@ fetch_B8G8R8A8_UNORM(const void *ptr, float *attrib) } -static fetch_func get_fetch_func( enum pipe_format format ) +fetch_func draw_get_fetch_func( enum pipe_format format ) { #if 0 { @@ -502,7 +502,7 @@ void draw_update_vertex_fetch( struct draw_context *draw ) draw->vertex_element[i].src_offset; draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; - draw->vertex_fetch.fetch[i] = get_fetch_func( format ); + draw->vertex_fetch.fetch[i] = draw_get_fetch_func( format ); } draw->vertex_fetch.nr_attrs = nr_attrs; -- cgit v1.2.3 From 333df9656a257507d5eb5a6a8e8f2d9fc7577f20 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 3 Apr 2008 13:19:38 +0100 Subject: gallium: Only build softpipe driver by default for xlib winsys with scons. --- src/gallium/winsys/xlib/SConscript | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript index c38b5be52c..218b89285f 100644 --- a/src/gallium/winsys/xlib/SConscript +++ b/src/gallium/winsys/xlib/SConscript @@ -22,8 +22,6 @@ sources = [ drivers = [ softpipe, - i915simple, - i965simple, ] # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions -- cgit v1.2.3 From f07fe3c728aec5a715cf615822151e11a00cda44 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Apr 2008 12:01:17 -0600 Subject: gallium: set gl_rasterization_rules --- src/mesa/state_tracker/st_cb_drawpixels.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c181575f15..c7796cfb6a 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -536,6 +536,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, { struct pipe_rasterizer_state rasterizer; memset(&rasterizer, 0, sizeof(rasterizer)); + rasterizer.gl_rasterization_rules = 1; rasterizer.scissor = ctx->Scissor.Enabled; cso_set_rasterizer(cso, &rasterizer); } -- cgit v1.2.3 From 0bc2e2c9cd552e4187dd8a2250e350efcdc9aea1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Apr 2008 12:01:53 -0600 Subject: gallium: call st_flush_bitmap_cache() --- src/mesa/state_tracker/st_cb_readpixels.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index e9fcdf69a1..82df28b29a 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -159,6 +159,8 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (!dest) return; + st_flush_bitmap_cache(ctx->st); + /* make sure rendering has completed */ pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); -- cgit v1.2.3 From 4f67a3f7d3bdeaa8d16d877ce9b277c97bd2f6b4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Apr 2008 12:42:26 -0600 Subject: gallium: use identity viewport fix broken clear_with_quad() path Since bypass_clipping is set and we're specifying quad vertexes in window coords, setup identity viewport. --- src/mesa/state_tracker/st_cb_clear.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index ec8d3e1022..0c6f77bc82 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -271,17 +271,15 @@ clear_with_quad(GLcontext *ctx, #endif #if !TEST_DRAW_PASSTHROUGH - /* viewport state: viewport matching window dims */ + /* viewport state: identity since we're drawing in window coords */ { - const float width = ctx->DrawBuffer->Width; - const float height = ctx->DrawBuffer->Height; struct pipe_viewport_state vp; - vp.scale[0] = 0.5 * width; - vp.scale[1] = -0.5 * height; + vp.scale[0] = 1.0; + vp.scale[1] = 1.0; vp.scale[2] = 1.0; vp.scale[3] = 1.0; - vp.translate[0] = 0.5 * width; - vp.translate[1] = 0.5 * height; + vp.translate[0] = 0.0; + vp.translate[1] = 0.0; vp.translate[2] = 0.0; vp.translate[3] = 0.0; cso_set_viewport(st->cso_context, &vp); -- cgit v1.2.3 From c4f8c8b304a47b2490fe5b1d133e314b045854df Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Apr 2008 12:44:58 -0600 Subject: gallium: remove the temporary/test TEST_DRAW_PASSTHROUGH code --- src/mesa/state_tracker/st_cb_clear.c | 12 ------------ 1 file changed, 12 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 0c6f77bc82..041b9be2cc 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -55,11 +55,6 @@ #include "cso_cache/cso_context.h" -/* XXX for testing draw module vertex passthrough: */ -/* XXX this hack is broken now */ -#define TEST_DRAW_PASSTHROUGH 0 - - void st_destroy_clear(struct st_context *st) { @@ -243,9 +238,6 @@ clear_with_quad(GLcontext *ctx, struct pipe_rasterizer_state raster; memset(&raster, 0, sizeof(raster)); raster.bypass_clipping = 1; -#if TEST_DRAW_PASSTHROUGH - raster.bypass_vs = 1; -#endif cso_set_rasterizer(st->cso_context, &raster); } @@ -256,7 +248,6 @@ clear_with_quad(GLcontext *ctx, pipe->bind_fs_state(pipe, st->clear.fs); -#if !TEST_DRAW_PASSTHROUGH /* vertex shader state: color/position pass-through */ if (!st->clear.vs) { const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, @@ -268,9 +259,7 @@ clear_with_quad(GLcontext *ctx, &st->clear.vert_shader); } pipe->bind_vs_state(pipe, st->clear.vs); -#endif -#if !TEST_DRAW_PASSTHROUGH /* viewport state: identity since we're drawing in window coords */ { struct pipe_viewport_state vp; @@ -284,7 +273,6 @@ clear_with_quad(GLcontext *ctx, vp.translate[3] = 0.0; cso_set_viewport(st->cso_context, &vp); } -#endif /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); -- cgit v1.2.3 From e8823bb7dfad7c6241da185cd0ac94ede42c33e5 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Apr 2008 12:52:55 -0600 Subject: gallium: include st_cb_bitmap.h to silence warning --- src/mesa/state_tracker/st_cb_readpixels.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 82df28b29a..2bcc8c99fb 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -43,6 +43,7 @@ #include "pipe/p_inlines.h" #include "util/p_tile.h" #include "st_context.h" +#include "st_cb_bitmap.h" #include "st_cb_readpixels.h" #include "st_cb_fbo.h" #include "st_format.h" -- cgit v1.2.3 From ce5c867cbb17b2444ebc3db5c6a03cee5e2edb8a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Apr 2008 12:54:32 -0600 Subject: gallium: streamline viewport/raster/shader state for clearing with quads Move init of these items to new st_init_clear(). --- src/mesa/state_tracker/st_cb_clear.c | 75 ++++++++++++++++++------------------ src/mesa/state_tracker/st_cb_clear.h | 4 ++ src/mesa/state_tracker/st_context.c | 1 + src/mesa/state_tracker/st_context.h | 2 + 4 files changed, 45 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 041b9be2cc..b5e737e0d1 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -55,6 +55,42 @@ #include "cso_cache/cso_context.h" +void +st_init_clear(struct st_context *st) +{ + struct pipe_context *pipe = st->pipe; + + /* rasterizer state: bypass clipping */ + memset(&st->clear.raster, 0, sizeof(st->clear.raster)); + st->clear.raster.bypass_clipping = 1; + + /* viewport state: identity since we're drawing in window coords */ + st->clear.viewport.scale[0] = 1.0; + st->clear.viewport.scale[1] = 1.0; + st->clear.viewport.scale[2] = 1.0; + st->clear.viewport.scale[3] = 1.0; + st->clear.viewport.translate[0] = 0.0; + st->clear.viewport.translate[1] = 0.0; + st->clear.viewport.translate[2] = 0.0; + st->clear.viewport.translate[3] = 0.0; + + /* fragment shader state: color pass-through program */ + st->clear.fs = + util_make_fragment_passthrough_shader(pipe, &st->clear.frag_shader); + + /* vertex shader state: color/position pass-through */ + { + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, + TGSI_SEMANTIC_COLOR }; + const uint semantic_indexes[] = { 0, 0 }; + st->clear.vs = util_make_vertex_passthrough_shader(pipe, 2, + semantic_names, + semantic_indexes, + &st->clear.vert_shader); + } +} + + void st_destroy_clear(struct st_context *st) { @@ -233,47 +269,12 @@ clear_with_quad(GLcontext *ctx, cso_set_depth_stencil_alpha(st->cso_context, &depth_stencil); } - /* rasterizer state: bypass clipping */ - { - struct pipe_rasterizer_state raster; - memset(&raster, 0, sizeof(raster)); - raster.bypass_clipping = 1; - cso_set_rasterizer(st->cso_context, &raster); - } + cso_set_rasterizer(st->cso_context, &st->clear.raster); + cso_set_viewport(st->cso_context, &st->clear.viewport); - /* fragment shader state: color pass-through program */ - if (!st->clear.fs) { - st->clear.fs = util_make_fragment_passthrough_shader(pipe, &st->clear.frag_shader); - } pipe->bind_fs_state(pipe, st->clear.fs); - - - /* vertex shader state: color/position pass-through */ - if (!st->clear.vs) { - const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, - TGSI_SEMANTIC_COLOR }; - const uint semantic_indexes[] = { 0, 0 }; - st->clear.vs = util_make_vertex_passthrough_shader(pipe, 2, - semantic_names, - semantic_indexes, - &st->clear.vert_shader); - } pipe->bind_vs_state(pipe, st->clear.vs); - /* viewport state: identity since we're drawing in window coords */ - { - struct pipe_viewport_state vp; - vp.scale[0] = 1.0; - vp.scale[1] = 1.0; - vp.scale[2] = 1.0; - vp.scale[3] = 1.0; - vp.translate[0] = 0.0; - vp.translate[1] = 0.0; - vp.translate[2] = 0.0; - vp.translate[3] = 0.0; - cso_set_viewport(st->cso_context, &vp); - } - /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); diff --git a/src/mesa/state_tracker/st_cb_clear.h b/src/mesa/state_tracker/st_cb_clear.h index dfa4033faa..f49387747d 100644 --- a/src/mesa/state_tracker/st_cb_clear.h +++ b/src/mesa/state_tracker/st_cb_clear.h @@ -30,6 +30,10 @@ #define ST_CB_CLEAR_H +extern void +st_init_clear(struct st_context *st); + + extern void st_destroy_clear(struct st_context *st); diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index a20195f2de..7511c28074 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -110,6 +110,7 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st_init_atoms( st ); st_init_bitmap(st); + st_init_clear(st); st_init_draw( st ); st_init_generate_mipmap(st); st_init_blit(st); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 44705bc89a..bcebbd4943 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -161,6 +161,8 @@ struct st_context struct { struct pipe_shader_state vert_shader; struct pipe_shader_state frag_shader; + struct pipe_rasterizer_state raster; + struct pipe_viewport_state viewport; void *vs; void *fs; float vertices[4][2][4]; /**< vertex pos + color */ -- cgit v1.2.3 From 124e1345c9ba4abe17bb04b8781ec0fe803eda7b Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Apr 2008 13:16:37 -0600 Subject: gallium: set rasterizer.gl_rasterization_rules = 1 in a few more places --- src/mesa/state_tracker/st_cb_bitmap.c | 1 + src/mesa/state_tracker/st_cb_clear.c | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 975a55d6dc..0872f2bd28 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -754,6 +754,7 @@ st_init_bitmap(struct st_context *st) /* init baseline rasterizer state once */ memset(&st->bitmap.rasterizer, 0, sizeof(st->bitmap.rasterizer)); + st->bitmap.rasterizer.gl_rasterization_rules = 1; st->bitmap.rasterizer.bypass_vs = 1; init_bitmap_cache(st); diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index b5e737e0d1..fa9f986f11 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -62,6 +62,7 @@ st_init_clear(struct st_context *st) /* rasterizer state: bypass clipping */ memset(&st->clear.raster, 0, sizeof(st->clear.raster)); + st->clear.raster.gl_rasterization_rules = 1; st->clear.raster.bypass_clipping = 1; /* viewport state: identity since we're drawing in window coords */ -- cgit v1.2.3 From 766f3a545ebb317d2115b9053a8fc13b49ceec12 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Apr 2008 16:36:07 -0600 Subject: gallium: implement ycbcr->rgba tile conversion --- src/gallium/auxiliary/util/p_tile.c | 71 +++++++++++++++++++++++++++++++++++++ 1 file changed, 71 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 287d783981..c9a9c8f4f7 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -560,6 +560,69 @@ z24s8_get_tile_rgba(unsigned *src, } +/*** PIPE_FORMAT_YCBCR / PIPE_FORMAT_YCBCR_REV ***/ + +/** + * Convert YCbCr (or YCrCb) to RGBA. + */ +static void +ycbcr_get_tile_rgba(ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride, + boolean rev) +{ + const float scale = 1.0f / 255.0f; + unsigned i, j; + + /* we're assuming we're being asked for an even number of texels */ + assert((w & 1) == 0); + + for (i = 0; i < h; i++) { + float *pRow = p; + /* do two texels at a time */ + for (j = 0; j < w; j += 2, src += 2) { + const ushort t0 = src[0]; + const ushort t1 = src[1]; + const ubyte y0 = (t0 >> 8) & 0xff; /* luminance */ + const ubyte y1 = (t1 >> 8) & 0xff; /* luminance */ + ubyte cb, cr; + if (rev) { + cb = t1 & 0xff; /* chroma U */ + cr = t0 & 0xff; /* chroma V */ + } + else { + cb = t0 & 0xff; /* chroma U */ + cr = t1 & 0xff; /* chroma V */ + } + float r, g, b; + + /* even pixel: y0,cr,cb */ + r = 1.164 * (y0-16) + 1.596 * (cr-128); + g = 1.164 * (y0-16) - 0.813 * (cr-128) - 0.391 * (cb-128); + b = 1.164 * (y0-16) + 2.018 * (cb-128); + pRow[0] = r * scale; + pRow[1] = g * scale; + pRow[2] = b * scale; + pRow[3] = 1.0f; + pRow += 4; + + /* odd pixel: use y1,cr,cb */ + r = 1.164 * (y1-16) + 1.596 * (cr-128); + g = 1.164 * (y1-16) - 0.813 * (cr-128) - 0.391 * (cb-128); + b = 1.164 * (y1-16) + 2.018 * (cb-128); + pRow[0] = r * scale; + pRow[1] = g * scale; + pRow[2] = b * scale; + pRow[3] = 1.0f; + pRow += 4; + + } + p += dst_stride; + } +} + + void pipe_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, @@ -622,6 +685,14 @@ pipe_get_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_Z24S8_UNORM: z24s8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; + case PIPE_FORMAT_YCBCR: + assert((x & 1) == 0); + ycbcr_get_tile_rgba((ushort *) packed, w, h, p, dst_stride, FALSE); + break; + case PIPE_FORMAT_YCBCR_REV: + assert((x & 1) == 0); + ycbcr_get_tile_rgba((ushort *) packed, w, h, p, dst_stride, TRUE); + break; default: assert(0); } -- cgit v1.2.3 From 1307eebfa07e3440554eb2871e5dbd94e98ed7a8 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 3 Apr 2008 16:37:30 -0600 Subject: gallium: test if PIPE_FORMAT_YCBCR[_REV] is supported and enable GL_MESA_ycbcr_texture Update texture format selection code too. --- src/mesa/state_tracker/st_extensions.c | 5 +++++ src/mesa/state_tracker/st_format.c | 25 +++++++++++++++++-------- 2 files changed, 22 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 20f4652413..47a50d40ca 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -220,4 +220,9 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE; } + if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR, PIPE_TEXTURE) || + screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV, PIPE_TEXTURE)) { + ctx->Extensions.MESA_ycbcr_texture = GL_TRUE; + } + } diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index cb6fa9a573..761645c105 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -189,15 +189,14 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) pinfo->base_format = GL_RGBA; } } + else if (pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR) { + pinfo->base_format = GL_YCBCR_MESA; + pinfo->datatype = GL_UNSIGNED_SHORT; + pinfo->size = 2; /* two bytes per "texel" */ + } else { - pipe_format_ycbcr_t info; - - assert( pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR ); - - info = format; - - /* TODO */ - assert( 0 ); + /* compressed format? */ + assert(0); } #if 0 @@ -275,6 +274,8 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) return PIPE_FORMAT_Z32_UNORM; case MESA_FORMAT_Z24_S8: return PIPE_FORMAT_Z24S8_UNORM; + case MESA_FORMAT_YCBCR: + return PIPE_FORMAT_YCBCR; default: assert(0); return 0; @@ -435,6 +436,14 @@ choose_format(struct pipe_context *pipe, GLint internalFormat, uint surfType) return default_rgba_format( screen, surfType ); case GL_YCBCR_MESA: + if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR, + PIPE_TEXTURE)) { + return PIPE_FORMAT_YCBCR; + } + if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV, + PIPE_TEXTURE)) { + return PIPE_FORMAT_YCBCR_REV; + } return PIPE_FORMAT_NONE; case GL_RGB_S3TC: -- cgit v1.2.3 From 217d37940771dd02ff1aa365105eca2c7a09d623 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 2 Apr 2008 14:01:42 -0600 Subject: cell: minor texture improvements Precompute tiles_per_row. Use ushort multiplies in a few places. New comments. --- src/gallium/drivers/cell/spu/spu_main.c | 2 ++ src/gallium/drivers/cell/spu/spu_main.h | 3 ++- src/gallium/drivers/cell/spu/spu_texture.c | 32 ++++++++++++++++++++---------- 3 files changed, 26 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c index 1ab1c40379..e04ffeb9b1 100644 --- a/src/gallium/drivers/cell/spu/spu_main.c +++ b/src/gallium/drivers/cell/spu/spu_main.c @@ -343,6 +343,8 @@ cmd_state_texture(const struct cell_command_texture *texture) spu.texture[unit].width = width; spu.texture[unit].height = height; + spu.texture[unit].tiles_per_row = width / TILE_SIZE; + spu.texture[unit].tex_size = (vector float) { width, height, 0.0, 0.0}; spu.texture[unit].tex_size_mask = (vector unsigned int) { width - 1, height - 1, 0, 0 }; diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h index e9e39cbeab..e962e1426c 100644 --- a/src/gallium/drivers/cell/spu/spu_main.h +++ b/src/gallium/drivers/cell/spu/spu_main.h @@ -105,7 +105,8 @@ struct spu_framebuffer { struct spu_texture { void *start; - uint width, height; + ushort width, height; + ushort tiles_per_row; vector float tex_size; vector unsigned int tex_size_mask; /**< == int(size - 1) */ vector unsigned int tex_size_x_mask; /**< == int(size - 1) */ diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c index ceab246980..e9a2754e57 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.c +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -51,22 +51,25 @@ invalidate_tex_cache(void) static uint get_texel(uint unit, vec_uint4 coordinate) { + const unsigned texture_ea = (uintptr_t) spu.texture[unit].start; + ushort x = spu_extract(coordinate, 0); + ushort y = spu_extract(coordinate, 1); + unsigned tile_offset = sizeof(tile_t) + * ((y / TILE_SIZE * spu.texture[unit].tiles_per_row) + (x / TILE_SIZE)); + ushort texel_offset = (ushort) 4 + * (ushort) (((ushort) (y % TILE_SIZE) * (ushort) TILE_SIZE) + (x % TILE_SIZE)); vec_uint4 tmp; - unsigned x = spu_extract(coordinate, 0); - unsigned y = spu_extract(coordinate, 1); - const unsigned tiles_per_row = spu.texture[unit].width / TILE_SIZE; - unsigned tile_offset = sizeof(tile_t) * ((y / TILE_SIZE * tiles_per_row) - + (x / TILE_SIZE)); - unsigned texel_offset = 4 * (((y % TILE_SIZE) * TILE_SIZE) - + (x % TILE_SIZE)); spu_dcache_fetch_unaligned((qword *) & tmp, - spu.texture[unit].start + tile_offset + texel_offset, + texture_ea + tile_offset + texel_offset, 4); return spu_extract(tmp, 0); } +/** + * Get four texels from locations (x[0], y[0]), (x[1], y[1]) ... + */ static void get_four_texels(uint unit, vec_uint4 x, vec_uint4 y, vec_uint4 *texels) { @@ -76,7 +79,7 @@ get_four_texels(uint unit, vec_uint4 x, vec_uint4 y, vec_uint4 *texels) const qword offset_x = si_andi((qword) x, 0x1f); const qword offset_y = si_andi((qword) y, 0x1f); - const qword tiles_per_row = (qword) spu_splats(spu.texture[unit].width / TILE_SIZE); + const qword tiles_per_row = (qword) spu_splats(spu.texture[unit].tiles_per_row); const qword tile_size = (qword) spu_splats(sizeof(tile_t)); qword tile_offset = si_mpya((qword) tile_y, tiles_per_row, (qword) tile_x); @@ -97,6 +100,7 @@ get_four_texels(uint unit, vec_uint4 x, vec_uint4 y, vec_uint4 *texels) texture_ea + spu_extract(offset, 3), 4); } + /** * Get texture sample at texcoord. * XXX this is extremely primitive for now. @@ -126,17 +130,25 @@ sample_texture_bilinear(uint unit, vector float texcoord) vec_uint4 texels[4]; + /* setup texcoords for quad: + * +-----+-----+ + * |x0,y0|x1,y1| + * +-----+-----+ + * |x2,y2|x3,y3| + * +-----+-----+ + */ vec_uint4 x = spu_splats(spu_extract(itc, 0)); vec_uint4 y = spu_splats(spu_extract(itc, 1)); - x = spu_add(x, offset_x); y = spu_add(y, offset_y); + /* GL_REPEAT wrap mode: */ x = spu_and(x, spu.texture[unit].tex_size_x_mask); y = spu_and(y, spu.texture[unit].tex_size_y_mask); get_four_texels(unit, x, y, texels); + /* integer A8R8G8B8 to float texel conversion */ vector float texel00 = spu_unpack_A8R8G8B8(spu_extract(texels[0], 0)); vector float texel01 = spu_unpack_A8R8G8B8(spu_extract(texels[1], 0)); vector float texel10 = spu_unpack_A8R8G8B8(spu_extract(texels[2], 0)); -- cgit v1.2.3 From a7504ad587ee8cbfa9958ad23321a691ce0823d3 Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 2 Apr 2008 14:30:28 -0600 Subject: cell: added some comments/ideas about better texture sampling --- src/gallium/drivers/cell/spu/spu_texture.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_texture.c b/src/gallium/drivers/cell/spu/spu_texture.c index e9a2754e57..5051774f00 100644 --- a/src/gallium/drivers/cell/spu/spu_texture.c +++ b/src/gallium/drivers/cell/spu/spu_texture.c @@ -48,9 +48,16 @@ invalidate_tex_cache(void) } +/** + * XXX look into getting texels for all four pixels in a quad at once. + */ static uint get_texel(uint unit, vec_uint4 coordinate) { + /* + * XXX we could do the "/ TILE_SIZE" and "% TILE_SIZE" operations as + * SIMD since X and Y are already in a SIMD register. + */ const unsigned texture_ea = (uintptr_t) spu.texture[unit].start; ushort x = spu_extract(coordinate, 0); ushort y = spu_extract(coordinate, 1); @@ -69,6 +76,16 @@ get_texel(uint unit, vec_uint4 coordinate) /** * Get four texels from locations (x[0], y[0]), (x[1], y[1]) ... + * + * NOTE: in the typical case of bilinear filtering, the four texels + * are in a 2x2 group so we could get by with just two dcache fetches + * (two side-by-side texels per fetch). But when bilinear filtering + * wraps around a texture edge, we'll probably need code like we have + * now. + * FURTHERMORE: since we're rasterizing a quad of 2x2 pixels at a time, + * it's quite likely that the four pixels in a quad will need some of the + * same texels. So look into doing texture fetches for four pixels at + * a time. */ static void get_four_texels(uint unit, vec_uint4 x, vec_uint4 y, vec_uint4 *texels) @@ -103,7 +120,6 @@ get_four_texels(uint unit, vec_uint4 x, vec_uint4 y, vec_uint4 *texels) /** * Get texture sample at texcoord. - * XXX this is extremely primitive for now. */ vector float sample_texture_nearest(uint unit, vector float texcoord) -- cgit v1.2.3 From c07d3f6a9e62bd88e05d4ebf72430139b0531be8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 4 Apr 2008 09:06:13 +0100 Subject: gallium: Always allocate new const buffers instead of modifying existing ones. --- src/mesa/state_tracker/st_atom_constbuf.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index cd7372902b..03093579e1 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -71,13 +71,12 @@ void st_upload_constants( struct st_context *st, _mesa_load_state_parameters(st->ctx, params); - if (cbuf->buffer && cbuf->size != paramBytes) - pipe_buffer_reference( ws, &cbuf->buffer, NULL ); - - if (!cbuf->buffer) { - cbuf->buffer = ws->buffer_create(ws, 1, PIPE_BUFFER_USAGE_CONSTANT, - paramBytes); - } + /* We always need to get a new buffer, to keep the drivers simple and + * avoid gratuitous rendering synchronization. + */ + pipe_buffer_reference( ws, &cbuf->buffer, NULL ); + cbuf->buffer = ws->buffer_create( ws, 1, PIPE_BUFFER_USAGE_CONSTANT, + paramBytes ); if (0) { -- cgit v1.2.3 From 7a7bce7b24ea4f63faa1d5bfe3f71d09b412c838 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 4 Apr 2008 11:13:10 +0100 Subject: gallium: make msvc less unhappy --- src/gallium/auxiliary/util/p_tile.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index c9a9c8f4f7..520da5cecd 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -587,6 +587,8 @@ ycbcr_get_tile_rgba(ushort *src, const ubyte y0 = (t0 >> 8) & 0xff; /* luminance */ const ubyte y1 = (t1 >> 8) & 0xff; /* luminance */ ubyte cb, cr; + float r, g, b; + if (rev) { cb = t1 & 0xff; /* chroma U */ cr = t0 & 0xff; /* chroma V */ @@ -595,12 +597,11 @@ ycbcr_get_tile_rgba(ushort *src, cb = t0 & 0xff; /* chroma U */ cr = t1 & 0xff; /* chroma V */ } - float r, g, b; /* even pixel: y0,cr,cb */ - r = 1.164 * (y0-16) + 1.596 * (cr-128); - g = 1.164 * (y0-16) - 0.813 * (cr-128) - 0.391 * (cb-128); - b = 1.164 * (y0-16) + 2.018 * (cb-128); + r = 1.164f * (y0-16) + 1.596f * (cr-128); + g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128); + b = 1.164f * (y0-16) + 2.018f * (cb-128); pRow[0] = r * scale; pRow[1] = g * scale; pRow[2] = b * scale; @@ -608,9 +609,9 @@ ycbcr_get_tile_rgba(ushort *src, pRow += 4; /* odd pixel: use y1,cr,cb */ - r = 1.164 * (y1-16) + 1.596 * (cr-128); - g = 1.164 * (y1-16) - 0.813 * (cr-128) - 0.391 * (cb-128); - b = 1.164 * (y1-16) + 2.018 * (cb-128); + r = 1.164f * (y1-16) + 1.596f * (cr-128); + g = 1.164f * (y1-16) - 0.813f * (cr-128) - 0.391f * (cb-128); + b = 1.164f * (y1-16) + 2.018f * (cb-128); pRow[0] = r * scale; pRow[1] = g * scale; pRow[2] = b * scale; -- cgit v1.2.3 From cf9b07ea3474cd33e797eeb10b3fd73ad54ae8d5 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 4 Apr 2008 01:59:38 +0200 Subject: gallium: fix two-side stencil handling Previously all drivers were in twosided mode since they checked for stencil.enable[1] flag which was a copy of stencil.enable[0]. Note that drivers should not reference stencil[1] state (other than the enable) if twosided stenciling is disabled (for now the stencil state is still copied but for instance clear_with_quads won't provide useful values in there). Also, use _TestTwoSide instead of TestTwoSide since results would be bogus otherwise if using APIs with implicit two side stencil enable (i.e. core ogl 2.0). --- src/gallium/include/pipe/p_state.h | 2 +- src/mesa/state_tracker/st_atom_depth.c | 3 ++- 2 files changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 3593446e1c..2dc9a92186 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -181,7 +181,7 @@ struct pipe_depth_stencil_alpha_state unsigned occlusion_count:1; /**< do occlusion counting? */ } depth; struct { - unsigned enabled:1; + unsigned enabled:1; /**< stencil[0]: stencil enabled, stencil[1]: two-side enabled */ unsigned func:3; /**< PIPE_FUNC_x */ unsigned fail_op:3; /**< PIPE_STENCIL_OP_x */ unsigned zpass_op:3; /**< PIPE_STENCIL_OP_x */ diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index 03057e37fa..ef467582c0 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -115,7 +115,7 @@ update_depth_stencil_alpha(struct st_context *st) dsa->stencil[0].value_mask = st->ctx->Stencil.ValueMask[0] & 0xff; dsa->stencil[0].write_mask = st->ctx->Stencil.WriteMask[0] & 0xff; - if (st->ctx->Stencil.TestTwoSide) { + if (st->ctx->Stencil._TestTwoSide) { dsa->stencil[1].enabled = 1; dsa->stencil[1].func = st_compare_func_to_pipe(st->ctx->Stencil.Function[1]); dsa->stencil[1].fail_op = gl_stencil_op_to_pipe(st->ctx->Stencil.FailFunc[1]); @@ -127,6 +127,7 @@ update_depth_stencil_alpha(struct st_context *st) } else { dsa->stencil[1] = dsa->stencil[0]; + dsa->stencil[1].enabled = 0; } } -- cgit v1.2.3 From 04097f558325eae35b5da1b72e6da938b433ad0a Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 30 Oct 2007 14:09:17 +0100 Subject: add missing _mesa_StencilFuncSeparateATI function --- src/mesa/main/stencil.c | 75 +++++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/stencil.h | 2 ++ 2 files changed, 77 insertions(+) (limited to 'src') diff --git a/src/mesa/main/stencil.c b/src/mesa/main/stencil.c index e61eb0030c..ca7f6eaf88 100644 --- a/src/mesa/main/stencil.c +++ b/src/mesa/main/stencil.c @@ -82,6 +82,81 @@ _mesa_ClearStencil( GLint s ) } +/** + * Set the function and reference value for stencil testing. + * + * \param frontfunc front test function. + * \param backfunc back test function. + * \param ref front and back reference value. + * \param mask front and back bitmask. + * + * \sa glStencilFunc(). + * + * Verifies the parameters and updates the respective values in + * __GLcontextRec::Stencil. On change flushes the vertices and notifies the + * driver via the dd_function_table::StencilFunc callback. + */ +void GLAPIENTRY +_mesa_StencilFuncSeparateATI( GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask ) +{ + GET_CURRENT_CONTEXT(ctx); + const GLint stencilMax = (1 << ctx->DrawBuffer->Visual.stencilBits) - 1; + ASSERT_OUTSIDE_BEGIN_END(ctx); + + switch (frontfunc) { + case GL_NEVER: + case GL_LESS: + case GL_LEQUAL: + case GL_GREATER: + case GL_GEQUAL: + case GL_EQUAL: + case GL_NOTEQUAL: + case GL_ALWAYS: + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glStencilFuncSeparateATI (0x%04x)", frontfunc ); + return; + } + + switch (backfunc) { + case GL_NEVER: + case GL_LESS: + case GL_LEQUAL: + case GL_GREATER: + case GL_GEQUAL: + case GL_EQUAL: + case GL_NOTEQUAL: + case GL_ALWAYS: + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glStencilFuncSeparateATI (0x%04x)", backfunc ); + return; + } + + ref = CLAMP( ref, 0, stencilMax ); + + /* set both front and back state */ + if (ctx->Stencil.Function[0] == frontfunc && + ctx->Stencil.Function[1] == backfunc && + ctx->Stencil.ValueMask[0] == mask && + ctx->Stencil.ValueMask[1] == mask && + ctx->Stencil.Ref[0] == ref && + ctx->Stencil.Ref[1] == ref) + return; + FLUSH_VERTICES(ctx, _NEW_STENCIL); + ctx->Stencil.Function[0] = frontfunc; + ctx->Stencil.Function[1] = backfunc; + ctx->Stencil.Ref[0] = ctx->Stencil.Ref[1] = ref; + ctx->Stencil.ValueMask[0] = ctx->Stencil.ValueMask[1] = mask; + if (ctx->Driver.StencilFuncSeparate) { + ctx->Driver.StencilFuncSeparate(ctx, GL_FRONT, + frontfunc, ref, mask); + ctx->Driver.StencilFuncSeparate(ctx, GL_BACK, + backfunc, ref, mask); + } +} + + /** * Set the function and reference value for stencil testing. * diff --git a/src/mesa/main/stencil.h b/src/mesa/main/stencil.h index 27c6362023..0be9810005 100644 --- a/src/mesa/main/stencil.h +++ b/src/mesa/main/stencil.h @@ -62,6 +62,8 @@ _mesa_StencilOpSeparate(GLenum face, GLenum fail, GLenum zfail, GLenum zpass); extern void GLAPIENTRY _mesa_StencilFuncSeparate(GLenum face, GLenum func, GLint ref, GLuint mask); +extern void GLAPIENTRY +_mesa_StencilFuncSeparateATI(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); extern void GLAPIENTRY _mesa_StencilMaskSeparate(GLenum face, GLuint mask); -- cgit v1.2.3 From 102f2ef4fc2d45c51926add6bdf51ef6fcb43b35 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 09:13:58 -0600 Subject: Finish up ATI_separate_stencil Add entrypoints to glapi XML file and regenerate files. Implement glStencilOpSeparateATI(). Consolidate some code in stencil.c --- src/mesa/drivers/dri/common/extension_helper.h | 22 + src/mesa/glapi/dispatch.h | 24 +- src/mesa/glapi/gl_API.xml | 29 + src/mesa/glapi/glapioffsets.h | 14 +- src/mesa/glapi/glapitable.h | 10 +- src/mesa/glapi/glapitemp.h | 32 +- src/mesa/glapi/glprocs.h | 564 +++++------ src/mesa/main/enums.c | 1224 ++++++++++++------------ src/mesa/main/state.c | 4 + src/mesa/main/stencil.c | 289 +++--- src/mesa/main/stencil.h | 8 +- src/mesa/sparc/glapi_sparc.S | 4 + src/mesa/x86-64/glapi_x86-64.S | 92 ++ src/mesa/x86/glapi_x86.S | 16 +- 14 files changed, 1258 insertions(+), 1074 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/common/extension_helper.h b/src/mesa/drivers/dri/common/extension_helper.h index bf103a3931..726d9900c3 100644 --- a/src/mesa/drivers/dri/common/extension_helper.h +++ b/src/mesa/drivers/dri/common/extension_helper.h @@ -3839,6 +3839,13 @@ static const char Binormal3svEXT_names[] = ""; #endif +#if defined(need_GL_ATI_separate_stencil) +static const char StencilOpSeparateATI_names[] = + "iiii\0" /* Parameter signature */ + "glStencilOpSeparateATI\0" + ""; +#endif + #if defined(need_GL_EXT_light_texture) static const char ApplyTextureEXT_names[] = "i\0" /* Parameter signature */ @@ -4179,6 +4186,13 @@ static const char ActiveStencilFaceEXT_names[] = ""; #endif +#if defined(need_GL_ATI_separate_stencil) +static const char StencilFuncSeparateATI_names[] = + "iiii\0" /* Parameter signature */ + "glStencilFuncSeparateATI\0" + ""; +#endif + #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_shader_objects) static const char GetShaderSourceARB_names[] = "iipp\0" /* Parameter signature */ @@ -5182,6 +5196,14 @@ static const struct dri_extension_function GL_ATI_fragment_shader_functions[] = }; #endif +#if defined(need_GL_ATI_separate_stencil) +static const struct dri_extension_function GL_ATI_separate_stencil_functions[] = { + { StencilOpSeparateATI_names, StencilOpSeparateATI_remap_index, -1 }, + { StencilFuncSeparateATI_names, StencilFuncSeparateATI_remap_index, -1 }, + { NULL, 0, 0 } +}; +#endif + #if defined(need_GL_EXT_blend_color) static const struct dri_extension_function GL_EXT_blend_color_functions[] = { { BlendColor_names, -1, 336 }, diff --git a/src/mesa/glapi/dispatch.h b/src/mesa/glapi/dispatch.h index 0ce59f7b99..e46f159f94 100644 --- a/src/mesa/glapi/dispatch.h +++ b/src/mesa/glapi/dispatch.h @@ -2365,6 +2365,12 @@ #define CALL_FramebufferTextureLayerEXT(disp, parameters) (*((disp)->FramebufferTextureLayerEXT)) parameters #define GET_FramebufferTextureLayerEXT(disp) ((disp)->FramebufferTextureLayerEXT) #define SET_FramebufferTextureLayerEXT(disp, fn) ((disp)->FramebufferTextureLayerEXT = fn) +#define CALL_StencilFuncSeparateATI(disp, parameters) (*((disp)->StencilFuncSeparateATI)) parameters +#define GET_StencilFuncSeparateATI(disp) ((disp)->StencilFuncSeparateATI) +#define SET_StencilFuncSeparateATI(disp, fn) ((disp)->StencilFuncSeparateATI = fn) +#define CALL_StencilOpSeparateATI(disp, parameters) (*((disp)->StencilOpSeparateATI)) parameters +#define GET_StencilOpSeparateATI(disp) ((disp)->StencilOpSeparateATI) +#define SET_StencilOpSeparateATI(disp, fn) ((disp)->StencilOpSeparateATI = fn) #define CALL_ProgramEnvParameters4fvEXT(disp, parameters) (*((disp)->ProgramEnvParameters4fvEXT)) parameters #define GET_ProgramEnvParameters4fvEXT(disp) ((disp)->ProgramEnvParameters4fvEXT) #define SET_ProgramEnvParameters4fvEXT(disp, fn) ((disp)->ProgramEnvParameters4fvEXT = fn) @@ -2380,7 +2386,7 @@ #else -#define driDispatchRemapTable_size 365 +#define driDispatchRemapTable_size 367 extern int driDispatchRemapTable[ driDispatchRemapTable_size ]; #define AttachShader_remap_index 0 @@ -2744,10 +2750,12 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ]; #define RenderbufferStorageEXT_remap_index 358 #define BlitFramebufferEXT_remap_index 359 #define FramebufferTextureLayerEXT_remap_index 360 -#define ProgramEnvParameters4fvEXT_remap_index 361 -#define ProgramLocalParameters4fvEXT_remap_index 362 -#define GetQueryObjecti64vEXT_remap_index 363 -#define GetQueryObjectui64vEXT_remap_index 364 +#define StencilFuncSeparateATI_remap_index 361 +#define StencilOpSeparateATI_remap_index 362 +#define ProgramEnvParameters4fvEXT_remap_index 363 +#define ProgramLocalParameters4fvEXT_remap_index 364 +#define GetQueryObjecti64vEXT_remap_index 365 +#define GetQueryObjectui64vEXT_remap_index 366 #define CALL_AttachShader(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLuint, GLuint)), driDispatchRemapTable[AttachShader_remap_index], parameters) #define GET_AttachShader(disp) GET_by_offset(disp, driDispatchRemapTable[AttachShader_remap_index]) @@ -3832,6 +3840,12 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ]; #define CALL_FramebufferTextureLayerEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLuint, GLint, GLint)), driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index], parameters) #define GET_FramebufferTextureLayerEXT(disp) GET_by_offset(disp, driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index]) #define SET_FramebufferTextureLayerEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index], fn) +#define CALL_StencilFuncSeparateATI(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLint, GLuint)), driDispatchRemapTable[StencilFuncSeparateATI_remap_index], parameters) +#define GET_StencilFuncSeparateATI(disp) GET_by_offset(disp, driDispatchRemapTable[StencilFuncSeparateATI_remap_index]) +#define SET_StencilFuncSeparateATI(disp, fn) SET_by_offset(disp, driDispatchRemapTable[StencilFuncSeparateATI_remap_index], fn) +#define CALL_StencilOpSeparateATI(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLenum, GLenum)), driDispatchRemapTable[StencilOpSeparateATI_remap_index], parameters) +#define GET_StencilOpSeparateATI(disp) GET_by_offset(disp, driDispatchRemapTable[StencilOpSeparateATI_remap_index]) +#define SET_StencilOpSeparateATI(disp, fn) SET_by_offset(disp, driDispatchRemapTable[StencilOpSeparateATI_remap_index], fn) #define CALL_ProgramEnvParameters4fvEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLuint, GLsizei, const GLfloat *)), driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index], parameters) #define GET_ProgramEnvParameters4fvEXT(disp) GET_by_offset(disp, driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index]) #define SET_ProgramEnvParameters4fvEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index], fn) diff --git a/src/mesa/glapi/gl_API.xml b/src/mesa/glapi/gl_API.xml index 4bd3b2f0fb..e16ffe870f 100644 --- a/src/mesa/glapi/gl_API.xml +++ b/src/mesa/glapi/gl_API.xml @@ -12156,6 +12156,35 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/src/mesa/glapi/glapioffsets.h b/src/mesa/glapi/glapioffsets.h index c207a06ac4..a2db342ffe 100644 --- a/src/mesa/glapi/glapioffsets.h +++ b/src/mesa/glapi/glapioffsets.h @@ -801,11 +801,13 @@ #define _gloffset_RenderbufferStorageEXT 766 #define _gloffset_BlitFramebufferEXT 767 #define _gloffset_FramebufferTextureLayerEXT 768 -#define _gloffset_ProgramEnvParameters4fvEXT 769 -#define _gloffset_ProgramLocalParameters4fvEXT 770 -#define _gloffset_GetQueryObjecti64vEXT 771 -#define _gloffset_GetQueryObjectui64vEXT 772 -#define _gloffset_FIRST_DYNAMIC 773 +#define _gloffset_StencilFuncSeparateATI 769 +#define _gloffset_StencilOpSeparateATI 770 +#define _gloffset_ProgramEnvParameters4fvEXT 771 +#define _gloffset_ProgramLocalParameters4fvEXT 772 +#define _gloffset_GetQueryObjecti64vEXT 773 +#define _gloffset_GetQueryObjectui64vEXT 774 +#define _gloffset_FIRST_DYNAMIC 775 #else @@ -1170,6 +1172,8 @@ #define _gloffset_RenderbufferStorageEXT driDispatchRemapTable[RenderbufferStorageEXT_remap_index] #define _gloffset_BlitFramebufferEXT driDispatchRemapTable[BlitFramebufferEXT_remap_index] #define _gloffset_FramebufferTextureLayerEXT driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index] +#define _gloffset_StencilFuncSeparateATI driDispatchRemapTable[StencilFuncSeparateATI_remap_index] +#define _gloffset_StencilOpSeparateATI driDispatchRemapTable[StencilOpSeparateATI_remap_index] #define _gloffset_ProgramEnvParameters4fvEXT driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index] #define _gloffset_ProgramLocalParameters4fvEXT driDispatchRemapTable[ProgramLocalParameters4fvEXT_remap_index] #define _gloffset_GetQueryObjecti64vEXT driDispatchRemapTable[GetQueryObjecti64vEXT_remap_index] diff --git a/src/mesa/glapi/glapitable.h b/src/mesa/glapi/glapitable.h index 1e4c2949e4..148aa96b88 100644 --- a/src/mesa/glapi/glapitable.h +++ b/src/mesa/glapi/glapitable.h @@ -810,10 +810,12 @@ struct _glapi_table void (GLAPIENTRYP RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height); /* 766 */ void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 767 */ void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 768 */ - void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 769 */ - void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 770 */ - void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 771 */ - void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 772 */ + void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 769 */ + void (GLAPIENTRYP StencilOpSeparateATI)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass); /* 770 */ + void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 771 */ + void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 772 */ + void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 773 */ + void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 774 */ }; #endif /* !defined( _GLAPI_TABLE_H_ ) */ diff --git a/src/mesa/glapi/glapitemp.h b/src/mesa/glapi/glapitemp.h index 10af9b3085..cd420fdee5 100644 --- a/src/mesa/glapi/glapitemp.h +++ b/src/mesa/glapi/glapitemp.h @@ -5446,30 +5446,44 @@ KEYWORD1 void KEYWORD2 NAME(FramebufferTextureLayerEXT)(GLenum target, GLenum at DISPATCH(FramebufferTextureLayerEXT, (target, attachment, texture, level, layer), (F, "glFramebufferTextureLayerEXT(0x%x, 0x%x, %d, %d, %d);\n", target, attachment, texture, level, layer)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask) +{ + DISPATCH(StencilFuncSeparateATI, (frontfunc, backfunc, ref, mask), (F, "glStencilFuncSeparateATI(0x%x, 0x%x, %d, %d);\n", frontfunc, backfunc, ref, mask)); +} + +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass); + +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass) +{ + DISPATCH(StencilOpSeparateATI, (face, sfail, zfail, zpass), (F, "glStencilOpSeparateATI(0x%x, 0x%x, 0x%x, 0x%x);\n", face, sfail, zfail, zpass)); +} + +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); + +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) { DISPATCH(ProgramEnvParameters4fvEXT, (target, index, count, params), (F, "glProgramEnvParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) { DISPATCH(ProgramLocalParameters4fvEXT, (target, index, count, params), (F, "glProgramLocalParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLuint id, GLenum pname, GLint64EXT * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLint64EXT * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLuint id, GLenum pname, GLint64EXT * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLint64EXT * params) { DISPATCH(GetQueryObjecti64vEXT, (id, pname, params), (F, "glGetQueryObjecti64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLuint id, GLenum pname, GLuint64EXT * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_774)(GLuint id, GLenum pname, GLuint64EXT * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLuint id, GLenum pname, GLuint64EXT * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_774)(GLuint id, GLenum pname, GLuint64EXT * params) { DISPATCH(GetQueryObjectui64vEXT, (id, pname, params), (F, "glGetQueryObjectui64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params)); } @@ -6261,6 +6275,8 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = { TABLE_ENTRY(_dispatch_stub_770), TABLE_ENTRY(_dispatch_stub_771), TABLE_ENTRY(_dispatch_stub_772), + TABLE_ENTRY(_dispatch_stub_773), + TABLE_ENTRY(_dispatch_stub_774), /* A whole bunch of no-op functions. These might be called * when someone tries to call a dynamically-registered * extension function without a current rendering context. diff --git a/src/mesa/glapi/glprocs.h b/src/mesa/glapi/glprocs.h index c461333c51..657b74e578 100644 --- a/src/mesa/glapi/glprocs.h +++ b/src/mesa/glapi/glprocs.h @@ -821,6 +821,8 @@ static const char gl_string_table[] = "glRenderbufferStorageEXT\0" "glBlitFramebufferEXT\0" "glFramebufferTextureLayerEXT\0" + "glStencilFuncSeparateATI\0" + "glStencilOpSeparateATI\0" "glProgramEnvParameters4fvEXT\0" "glProgramLocalParameters4fvEXT\0" "glGetQueryObjecti64vEXT\0" @@ -1148,6 +1150,8 @@ static const char gl_string_table[] = #define gl_dispatch_stub_770 mgl_dispatch_stub_770 #define gl_dispatch_stub_771 mgl_dispatch_stub_771 #define gl_dispatch_stub_772 mgl_dispatch_stub_772 +#define gl_dispatch_stub_773 mgl_dispatch_stub_773 +#define gl_dispatch_stub_774 mgl_dispatch_stub_774 #endif /* USE_MGL_NAMESPACE */ @@ -1198,6 +1202,8 @@ extern void gl_dispatch_stub_769(void); extern void gl_dispatch_stub_770(void); extern void gl_dispatch_stub_771(void); extern void gl_dispatch_stub_772(void); +extern void gl_dispatch_stub_773(void); +extern void gl_dispatch_stub_774(void); #endif /* defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING) */ static const glprocs_table_t static_functions[] = { @@ -1970,284 +1976,286 @@ static const glprocs_table_t static_functions[] = { NAME_FUNC_OFFSET(13424, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT), NAME_FUNC_OFFSET(13449, gl_dispatch_stub_767, gl_dispatch_stub_767, NULL, _gloffset_BlitFramebufferEXT), NAME_FUNC_OFFSET(13470, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT), - NAME_FUNC_OFFSET(13499, gl_dispatch_stub_769, gl_dispatch_stub_769, NULL, _gloffset_ProgramEnvParameters4fvEXT), - NAME_FUNC_OFFSET(13528, gl_dispatch_stub_770, gl_dispatch_stub_770, NULL, _gloffset_ProgramLocalParameters4fvEXT), - NAME_FUNC_OFFSET(13559, gl_dispatch_stub_771, gl_dispatch_stub_771, NULL, _gloffset_GetQueryObjecti64vEXT), - NAME_FUNC_OFFSET(13583, gl_dispatch_stub_772, gl_dispatch_stub_772, NULL, _gloffset_GetQueryObjectui64vEXT), - NAME_FUNC_OFFSET(13608, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement), - NAME_FUNC_OFFSET(13626, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture), - NAME_FUNC_OFFSET(13643, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays), - NAME_FUNC_OFFSET(13659, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident), - NAME_FUNC_OFFSET(13684, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D), - NAME_FUNC_OFFSET(13704, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D), - NAME_FUNC_OFFSET(13724, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D), - NAME_FUNC_OFFSET(13747, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D), - NAME_FUNC_OFFSET(13770, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures), - NAME_FUNC_OFFSET(13790, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures), - NAME_FUNC_OFFSET(13807, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv), - NAME_FUNC_OFFSET(13824, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture), - NAME_FUNC_OFFSET(13839, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures), - NAME_FUNC_OFFSET(13863, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D), - NAME_FUNC_OFFSET(13882, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D), - NAME_FUNC_OFFSET(13901, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor), - NAME_FUNC_OFFSET(13917, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation), - NAME_FUNC_OFFSET(13936, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements), - NAME_FUNC_OFFSET(13959, glColorTable, glColorTable, NULL, _gloffset_ColorTable), - NAME_FUNC_OFFSET(13975, glColorTable, glColorTable, NULL, _gloffset_ColorTable), - NAME_FUNC_OFFSET(13991, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv), - NAME_FUNC_OFFSET(14018, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv), - NAME_FUNC_OFFSET(14045, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable), - NAME_FUNC_OFFSET(14065, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), - NAME_FUNC_OFFSET(14084, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), - NAME_FUNC_OFFSET(14103, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), - NAME_FUNC_OFFSET(14133, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), - NAME_FUNC_OFFSET(14163, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), - NAME_FUNC_OFFSET(14193, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), - NAME_FUNC_OFFSET(14223, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable), - NAME_FUNC_OFFSET(14242, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable), - NAME_FUNC_OFFSET(14265, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D), - NAME_FUNC_OFFSET(14290, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D), - NAME_FUNC_OFFSET(14315, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf), - NAME_FUNC_OFFSET(14342, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv), - NAME_FUNC_OFFSET(14370, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri), - NAME_FUNC_OFFSET(14397, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv), - NAME_FUNC_OFFSET(14425, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D), - NAME_FUNC_OFFSET(14454, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D), - NAME_FUNC_OFFSET(14483, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter), - NAME_FUNC_OFFSET(14509, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv), - NAME_FUNC_OFFSET(14540, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv), - NAME_FUNC_OFFSET(14571, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter), - NAME_FUNC_OFFSET(14595, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D), - NAME_FUNC_OFFSET(14618, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram), - NAME_FUNC_OFFSET(14636, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv), - NAME_FUNC_OFFSET(14665, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv), - NAME_FUNC_OFFSET(14694, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax), - NAME_FUNC_OFFSET(14709, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv), - NAME_FUNC_OFFSET(14735, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv), - NAME_FUNC_OFFSET(14761, glHistogram, glHistogram, NULL, _gloffset_Histogram), - NAME_FUNC_OFFSET(14776, glMinmax, glMinmax, NULL, _gloffset_Minmax), - NAME_FUNC_OFFSET(14788, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram), - NAME_FUNC_OFFSET(14808, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax), - NAME_FUNC_OFFSET(14825, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D), - NAME_FUNC_OFFSET(14841, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D), - NAME_FUNC_OFFSET(14860, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D), - NAME_FUNC_OFFSET(14883, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB), - NAME_FUNC_OFFSET(14899, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB), - NAME_FUNC_OFFSET(14921, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB), - NAME_FUNC_OFFSET(14939, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB), - NAME_FUNC_OFFSET(14958, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB), - NAME_FUNC_OFFSET(14976, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB), - NAME_FUNC_OFFSET(14995, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB), - NAME_FUNC_OFFSET(15013, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB), - NAME_FUNC_OFFSET(15032, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB), - NAME_FUNC_OFFSET(15050, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB), - NAME_FUNC_OFFSET(15069, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB), - NAME_FUNC_OFFSET(15087, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB), - NAME_FUNC_OFFSET(15106, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB), - NAME_FUNC_OFFSET(15124, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB), - NAME_FUNC_OFFSET(15143, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB), - NAME_FUNC_OFFSET(15161, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB), - NAME_FUNC_OFFSET(15180, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB), - NAME_FUNC_OFFSET(15198, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB), - NAME_FUNC_OFFSET(15217, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB), - NAME_FUNC_OFFSET(15235, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB), - NAME_FUNC_OFFSET(15254, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB), - NAME_FUNC_OFFSET(15272, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB), - NAME_FUNC_OFFSET(15291, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB), - NAME_FUNC_OFFSET(15309, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB), - NAME_FUNC_OFFSET(15328, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB), - NAME_FUNC_OFFSET(15346, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB), - NAME_FUNC_OFFSET(15365, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB), - NAME_FUNC_OFFSET(15383, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB), - NAME_FUNC_OFFSET(15402, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB), - NAME_FUNC_OFFSET(15420, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB), - NAME_FUNC_OFFSET(15439, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB), - NAME_FUNC_OFFSET(15457, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB), - NAME_FUNC_OFFSET(15476, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB), - NAME_FUNC_OFFSET(15494, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB), - NAME_FUNC_OFFSET(15513, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB), - NAME_FUNC_OFFSET(15536, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB), - NAME_FUNC_OFFSET(15559, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB), - NAME_FUNC_OFFSET(15582, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB), - NAME_FUNC_OFFSET(15605, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB), - NAME_FUNC_OFFSET(15622, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB), - NAME_FUNC_OFFSET(15645, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB), - NAME_FUNC_OFFSET(15668, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB), - NAME_FUNC_OFFSET(15691, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB), - NAME_FUNC_OFFSET(15717, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB), - NAME_FUNC_OFFSET(15743, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB), - NAME_FUNC_OFFSET(15769, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB), - NAME_FUNC_OFFSET(15793, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB), - NAME_FUNC_OFFSET(15820, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB), - NAME_FUNC_OFFSET(15846, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB), - NAME_FUNC_OFFSET(15866, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB), - NAME_FUNC_OFFSET(15886, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB), - NAME_FUNC_OFFSET(15906, glVertexAttrib1dARB, glVertexAttrib1dARB, NULL, _gloffset_VertexAttrib1dARB), - NAME_FUNC_OFFSET(15923, glVertexAttrib1dvARB, glVertexAttrib1dvARB, NULL, _gloffset_VertexAttrib1dvARB), - NAME_FUNC_OFFSET(15941, glVertexAttrib1fARB, glVertexAttrib1fARB, NULL, _gloffset_VertexAttrib1fARB), - NAME_FUNC_OFFSET(15958, glVertexAttrib1fvARB, glVertexAttrib1fvARB, NULL, _gloffset_VertexAttrib1fvARB), - NAME_FUNC_OFFSET(15976, glVertexAttrib1sARB, glVertexAttrib1sARB, NULL, _gloffset_VertexAttrib1sARB), - NAME_FUNC_OFFSET(15993, glVertexAttrib1svARB, glVertexAttrib1svARB, NULL, _gloffset_VertexAttrib1svARB), - NAME_FUNC_OFFSET(16011, glVertexAttrib2dARB, glVertexAttrib2dARB, NULL, _gloffset_VertexAttrib2dARB), - NAME_FUNC_OFFSET(16028, glVertexAttrib2dvARB, glVertexAttrib2dvARB, NULL, _gloffset_VertexAttrib2dvARB), - NAME_FUNC_OFFSET(16046, glVertexAttrib2fARB, glVertexAttrib2fARB, NULL, _gloffset_VertexAttrib2fARB), - NAME_FUNC_OFFSET(16063, glVertexAttrib2fvARB, glVertexAttrib2fvARB, NULL, _gloffset_VertexAttrib2fvARB), - NAME_FUNC_OFFSET(16081, glVertexAttrib2sARB, glVertexAttrib2sARB, NULL, _gloffset_VertexAttrib2sARB), - NAME_FUNC_OFFSET(16098, glVertexAttrib2svARB, glVertexAttrib2svARB, NULL, _gloffset_VertexAttrib2svARB), - NAME_FUNC_OFFSET(16116, glVertexAttrib3dARB, glVertexAttrib3dARB, NULL, _gloffset_VertexAttrib3dARB), - NAME_FUNC_OFFSET(16133, glVertexAttrib3dvARB, glVertexAttrib3dvARB, NULL, _gloffset_VertexAttrib3dvARB), - NAME_FUNC_OFFSET(16151, glVertexAttrib3fARB, glVertexAttrib3fARB, NULL, _gloffset_VertexAttrib3fARB), - NAME_FUNC_OFFSET(16168, glVertexAttrib3fvARB, glVertexAttrib3fvARB, NULL, _gloffset_VertexAttrib3fvARB), - NAME_FUNC_OFFSET(16186, glVertexAttrib3sARB, glVertexAttrib3sARB, NULL, _gloffset_VertexAttrib3sARB), - NAME_FUNC_OFFSET(16203, glVertexAttrib3svARB, glVertexAttrib3svARB, NULL, _gloffset_VertexAttrib3svARB), - NAME_FUNC_OFFSET(16221, glVertexAttrib4NbvARB, glVertexAttrib4NbvARB, NULL, _gloffset_VertexAttrib4NbvARB), - NAME_FUNC_OFFSET(16240, glVertexAttrib4NivARB, glVertexAttrib4NivARB, NULL, _gloffset_VertexAttrib4NivARB), - NAME_FUNC_OFFSET(16259, glVertexAttrib4NsvARB, glVertexAttrib4NsvARB, NULL, _gloffset_VertexAttrib4NsvARB), - NAME_FUNC_OFFSET(16278, glVertexAttrib4NubARB, glVertexAttrib4NubARB, NULL, _gloffset_VertexAttrib4NubARB), - NAME_FUNC_OFFSET(16297, glVertexAttrib4NubvARB, glVertexAttrib4NubvARB, NULL, _gloffset_VertexAttrib4NubvARB), - NAME_FUNC_OFFSET(16317, glVertexAttrib4NuivARB, glVertexAttrib4NuivARB, NULL, _gloffset_VertexAttrib4NuivARB), - NAME_FUNC_OFFSET(16337, glVertexAttrib4NusvARB, glVertexAttrib4NusvARB, NULL, _gloffset_VertexAttrib4NusvARB), - NAME_FUNC_OFFSET(16357, glVertexAttrib4bvARB, glVertexAttrib4bvARB, NULL, _gloffset_VertexAttrib4bvARB), - NAME_FUNC_OFFSET(16375, glVertexAttrib4dARB, glVertexAttrib4dARB, NULL, _gloffset_VertexAttrib4dARB), - NAME_FUNC_OFFSET(16392, glVertexAttrib4dvARB, glVertexAttrib4dvARB, NULL, _gloffset_VertexAttrib4dvARB), - NAME_FUNC_OFFSET(16410, glVertexAttrib4fARB, glVertexAttrib4fARB, NULL, _gloffset_VertexAttrib4fARB), - NAME_FUNC_OFFSET(16427, glVertexAttrib4fvARB, glVertexAttrib4fvARB, NULL, _gloffset_VertexAttrib4fvARB), - NAME_FUNC_OFFSET(16445, glVertexAttrib4ivARB, glVertexAttrib4ivARB, NULL, _gloffset_VertexAttrib4ivARB), - NAME_FUNC_OFFSET(16463, glVertexAttrib4sARB, glVertexAttrib4sARB, NULL, _gloffset_VertexAttrib4sARB), - NAME_FUNC_OFFSET(16480, glVertexAttrib4svARB, glVertexAttrib4svARB, NULL, _gloffset_VertexAttrib4svARB), - NAME_FUNC_OFFSET(16498, glVertexAttrib4ubvARB, glVertexAttrib4ubvARB, NULL, _gloffset_VertexAttrib4ubvARB), - NAME_FUNC_OFFSET(16517, glVertexAttrib4uivARB, glVertexAttrib4uivARB, NULL, _gloffset_VertexAttrib4uivARB), - NAME_FUNC_OFFSET(16536, glVertexAttrib4usvARB, glVertexAttrib4usvARB, NULL, _gloffset_VertexAttrib4usvARB), - NAME_FUNC_OFFSET(16555, glVertexAttribPointerARB, glVertexAttribPointerARB, NULL, _gloffset_VertexAttribPointerARB), - NAME_FUNC_OFFSET(16577, glBindBufferARB, glBindBufferARB, NULL, _gloffset_BindBufferARB), - NAME_FUNC_OFFSET(16590, glBufferDataARB, glBufferDataARB, NULL, _gloffset_BufferDataARB), - NAME_FUNC_OFFSET(16603, glBufferSubDataARB, glBufferSubDataARB, NULL, _gloffset_BufferSubDataARB), - NAME_FUNC_OFFSET(16619, glDeleteBuffersARB, glDeleteBuffersARB, NULL, _gloffset_DeleteBuffersARB), - NAME_FUNC_OFFSET(16635, glGenBuffersARB, glGenBuffersARB, NULL, _gloffset_GenBuffersARB), - NAME_FUNC_OFFSET(16648, glGetBufferParameterivARB, glGetBufferParameterivARB, NULL, _gloffset_GetBufferParameterivARB), - NAME_FUNC_OFFSET(16671, glGetBufferPointervARB, glGetBufferPointervARB, NULL, _gloffset_GetBufferPointervARB), - NAME_FUNC_OFFSET(16691, glGetBufferSubDataARB, glGetBufferSubDataARB, NULL, _gloffset_GetBufferSubDataARB), - NAME_FUNC_OFFSET(16710, glIsBufferARB, glIsBufferARB, NULL, _gloffset_IsBufferARB), - NAME_FUNC_OFFSET(16721, glMapBufferARB, glMapBufferARB, NULL, _gloffset_MapBufferARB), - NAME_FUNC_OFFSET(16733, glUnmapBufferARB, glUnmapBufferARB, NULL, _gloffset_UnmapBufferARB), - NAME_FUNC_OFFSET(16747, glBeginQueryARB, glBeginQueryARB, NULL, _gloffset_BeginQueryARB), - NAME_FUNC_OFFSET(16760, glDeleteQueriesARB, glDeleteQueriesARB, NULL, _gloffset_DeleteQueriesARB), - NAME_FUNC_OFFSET(16776, glEndQueryARB, glEndQueryARB, NULL, _gloffset_EndQueryARB), - NAME_FUNC_OFFSET(16787, glGenQueriesARB, glGenQueriesARB, NULL, _gloffset_GenQueriesARB), - NAME_FUNC_OFFSET(16800, glGetQueryObjectivARB, glGetQueryObjectivARB, NULL, _gloffset_GetQueryObjectivARB), - NAME_FUNC_OFFSET(16819, glGetQueryObjectuivARB, glGetQueryObjectuivARB, NULL, _gloffset_GetQueryObjectuivARB), - NAME_FUNC_OFFSET(16839, glGetQueryivARB, glGetQueryivARB, NULL, _gloffset_GetQueryivARB), - NAME_FUNC_OFFSET(16852, glIsQueryARB, glIsQueryARB, NULL, _gloffset_IsQueryARB), - NAME_FUNC_OFFSET(16862, glCompileShaderARB, glCompileShaderARB, NULL, _gloffset_CompileShaderARB), - NAME_FUNC_OFFSET(16878, glGetActiveUniformARB, glGetActiveUniformARB, NULL, _gloffset_GetActiveUniformARB), - NAME_FUNC_OFFSET(16897, glGetShaderSourceARB, glGetShaderSourceARB, NULL, _gloffset_GetShaderSourceARB), - NAME_FUNC_OFFSET(16915, glGetUniformLocationARB, glGetUniformLocationARB, NULL, _gloffset_GetUniformLocationARB), - NAME_FUNC_OFFSET(16936, glGetUniformfvARB, glGetUniformfvARB, NULL, _gloffset_GetUniformfvARB), - NAME_FUNC_OFFSET(16951, glGetUniformivARB, glGetUniformivARB, NULL, _gloffset_GetUniformivARB), - NAME_FUNC_OFFSET(16966, glLinkProgramARB, glLinkProgramARB, NULL, _gloffset_LinkProgramARB), - NAME_FUNC_OFFSET(16980, glShaderSourceARB, glShaderSourceARB, NULL, _gloffset_ShaderSourceARB), - NAME_FUNC_OFFSET(16995, glUniform1fARB, glUniform1fARB, NULL, _gloffset_Uniform1fARB), - NAME_FUNC_OFFSET(17007, glUniform1fvARB, glUniform1fvARB, NULL, _gloffset_Uniform1fvARB), - NAME_FUNC_OFFSET(17020, glUniform1iARB, glUniform1iARB, NULL, _gloffset_Uniform1iARB), - NAME_FUNC_OFFSET(17032, glUniform1ivARB, glUniform1ivARB, NULL, _gloffset_Uniform1ivARB), - NAME_FUNC_OFFSET(17045, glUniform2fARB, glUniform2fARB, NULL, _gloffset_Uniform2fARB), - NAME_FUNC_OFFSET(17057, glUniform2fvARB, glUniform2fvARB, NULL, _gloffset_Uniform2fvARB), - NAME_FUNC_OFFSET(17070, glUniform2iARB, glUniform2iARB, NULL, _gloffset_Uniform2iARB), - NAME_FUNC_OFFSET(17082, glUniform2ivARB, glUniform2ivARB, NULL, _gloffset_Uniform2ivARB), - NAME_FUNC_OFFSET(17095, glUniform3fARB, glUniform3fARB, NULL, _gloffset_Uniform3fARB), - NAME_FUNC_OFFSET(17107, glUniform3fvARB, glUniform3fvARB, NULL, _gloffset_Uniform3fvARB), - NAME_FUNC_OFFSET(17120, glUniform3iARB, glUniform3iARB, NULL, _gloffset_Uniform3iARB), - NAME_FUNC_OFFSET(17132, glUniform3ivARB, glUniform3ivARB, NULL, _gloffset_Uniform3ivARB), - NAME_FUNC_OFFSET(17145, glUniform4fARB, glUniform4fARB, NULL, _gloffset_Uniform4fARB), - NAME_FUNC_OFFSET(17157, glUniform4fvARB, glUniform4fvARB, NULL, _gloffset_Uniform4fvARB), - NAME_FUNC_OFFSET(17170, glUniform4iARB, glUniform4iARB, NULL, _gloffset_Uniform4iARB), - NAME_FUNC_OFFSET(17182, glUniform4ivARB, glUniform4ivARB, NULL, _gloffset_Uniform4ivARB), - NAME_FUNC_OFFSET(17195, glUniformMatrix2fvARB, glUniformMatrix2fvARB, NULL, _gloffset_UniformMatrix2fvARB), - NAME_FUNC_OFFSET(17214, glUniformMatrix3fvARB, glUniformMatrix3fvARB, NULL, _gloffset_UniformMatrix3fvARB), - NAME_FUNC_OFFSET(17233, glUniformMatrix4fvARB, glUniformMatrix4fvARB, NULL, _gloffset_UniformMatrix4fvARB), - NAME_FUNC_OFFSET(17252, glUseProgramObjectARB, glUseProgramObjectARB, NULL, _gloffset_UseProgramObjectARB), - NAME_FUNC_OFFSET(17265, glValidateProgramARB, glValidateProgramARB, NULL, _gloffset_ValidateProgramARB), - NAME_FUNC_OFFSET(17283, glBindAttribLocationARB, glBindAttribLocationARB, NULL, _gloffset_BindAttribLocationARB), - NAME_FUNC_OFFSET(17304, glGetActiveAttribARB, glGetActiveAttribARB, NULL, _gloffset_GetActiveAttribARB), - NAME_FUNC_OFFSET(17322, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB), - NAME_FUNC_OFFSET(17342, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB), - NAME_FUNC_OFFSET(17356, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB), - NAME_FUNC_OFFSET(17373, gl_dispatch_stub_568, gl_dispatch_stub_568, NULL, _gloffset_SampleMaskSGIS), - NAME_FUNC_OFFSET(17389, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SamplePatternSGIS), - NAME_FUNC_OFFSET(17408, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT), - NAME_FUNC_OFFSET(17426, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT), - NAME_FUNC_OFFSET(17447, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT), - NAME_FUNC_OFFSET(17469, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT), - NAME_FUNC_OFFSET(17488, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT), - NAME_FUNC_OFFSET(17510, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT), - NAME_FUNC_OFFSET(17533, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT), - NAME_FUNC_OFFSET(17552, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT), - NAME_FUNC_OFFSET(17572, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT), - NAME_FUNC_OFFSET(17591, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT), - NAME_FUNC_OFFSET(17611, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT), - NAME_FUNC_OFFSET(17630, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT), - NAME_FUNC_OFFSET(17650, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT), - NAME_FUNC_OFFSET(17669, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT), - NAME_FUNC_OFFSET(17689, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT), - NAME_FUNC_OFFSET(17708, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT), - NAME_FUNC_OFFSET(17728, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT), - NAME_FUNC_OFFSET(17748, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT), - NAME_FUNC_OFFSET(17769, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT), - NAME_FUNC_OFFSET(17789, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT), - NAME_FUNC_OFFSET(17810, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT), - NAME_FUNC_OFFSET(17830, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT), - NAME_FUNC_OFFSET(17851, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT), - NAME_FUNC_OFFSET(17875, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT), - NAME_FUNC_OFFSET(17893, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT), - NAME_FUNC_OFFSET(17913, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT), - NAME_FUNC_OFFSET(17931, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT), - NAME_FUNC_OFFSET(17943, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT), - NAME_FUNC_OFFSET(17956, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT), - NAME_FUNC_OFFSET(17968, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT), - NAME_FUNC_OFFSET(17981, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT), - NAME_FUNC_OFFSET(18001, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT), - NAME_FUNC_OFFSET(18025, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA), - NAME_FUNC_OFFSET(18039, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA), - NAME_FUNC_OFFSET(18056, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA), - NAME_FUNC_OFFSET(18071, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA), - NAME_FUNC_OFFSET(18089, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA), - NAME_FUNC_OFFSET(18103, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA), - NAME_FUNC_OFFSET(18120, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA), - NAME_FUNC_OFFSET(18135, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA), - NAME_FUNC_OFFSET(18153, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA), - NAME_FUNC_OFFSET(18167, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA), - NAME_FUNC_OFFSET(18184, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA), - NAME_FUNC_OFFSET(18199, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA), - NAME_FUNC_OFFSET(18217, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA), - NAME_FUNC_OFFSET(18231, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA), - NAME_FUNC_OFFSET(18248, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA), - NAME_FUNC_OFFSET(18263, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA), - NAME_FUNC_OFFSET(18281, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA), - NAME_FUNC_OFFSET(18295, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA), - NAME_FUNC_OFFSET(18312, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA), - NAME_FUNC_OFFSET(18327, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA), - NAME_FUNC_OFFSET(18345, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA), - NAME_FUNC_OFFSET(18359, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA), - NAME_FUNC_OFFSET(18376, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA), - NAME_FUNC_OFFSET(18391, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA), - NAME_FUNC_OFFSET(18409, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA), - NAME_FUNC_OFFSET(18423, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA), - NAME_FUNC_OFFSET(18440, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA), - NAME_FUNC_OFFSET(18455, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA), - NAME_FUNC_OFFSET(18473, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA), - NAME_FUNC_OFFSET(18487, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA), - NAME_FUNC_OFFSET(18504, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA), - NAME_FUNC_OFFSET(18519, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA), - NAME_FUNC_OFFSET(18537, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV), - NAME_FUNC_OFFSET(18554, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV), - NAME_FUNC_OFFSET(18574, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV), - NAME_FUNC_OFFSET(18591, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV), - NAME_FUNC_OFFSET(18617, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV), - NAME_FUNC_OFFSET(18646, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV), - NAME_FUNC_OFFSET(18661, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV), - NAME_FUNC_OFFSET(18679, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV), - NAME_FUNC_OFFSET(18698, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT), - NAME_FUNC_OFFSET(18722, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT), + NAME_FUNC_OFFSET(13499, gl_dispatch_stub_769, gl_dispatch_stub_769, NULL, _gloffset_StencilFuncSeparateATI), + NAME_FUNC_OFFSET(13524, gl_dispatch_stub_770, gl_dispatch_stub_770, NULL, _gloffset_StencilOpSeparateATI), + NAME_FUNC_OFFSET(13547, gl_dispatch_stub_771, gl_dispatch_stub_771, NULL, _gloffset_ProgramEnvParameters4fvEXT), + NAME_FUNC_OFFSET(13576, gl_dispatch_stub_772, gl_dispatch_stub_772, NULL, _gloffset_ProgramLocalParameters4fvEXT), + NAME_FUNC_OFFSET(13607, gl_dispatch_stub_773, gl_dispatch_stub_773, NULL, _gloffset_GetQueryObjecti64vEXT), + NAME_FUNC_OFFSET(13631, gl_dispatch_stub_774, gl_dispatch_stub_774, NULL, _gloffset_GetQueryObjectui64vEXT), + NAME_FUNC_OFFSET(13656, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement), + NAME_FUNC_OFFSET(13674, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture), + NAME_FUNC_OFFSET(13691, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays), + NAME_FUNC_OFFSET(13707, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident), + NAME_FUNC_OFFSET(13732, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D), + NAME_FUNC_OFFSET(13752, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D), + NAME_FUNC_OFFSET(13772, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D), + NAME_FUNC_OFFSET(13795, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D), + NAME_FUNC_OFFSET(13818, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures), + NAME_FUNC_OFFSET(13838, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures), + NAME_FUNC_OFFSET(13855, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv), + NAME_FUNC_OFFSET(13872, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture), + NAME_FUNC_OFFSET(13887, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures), + NAME_FUNC_OFFSET(13911, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D), + NAME_FUNC_OFFSET(13930, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D), + NAME_FUNC_OFFSET(13949, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor), + NAME_FUNC_OFFSET(13965, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation), + NAME_FUNC_OFFSET(13984, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements), + NAME_FUNC_OFFSET(14007, glColorTable, glColorTable, NULL, _gloffset_ColorTable), + NAME_FUNC_OFFSET(14023, glColorTable, glColorTable, NULL, _gloffset_ColorTable), + NAME_FUNC_OFFSET(14039, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv), + NAME_FUNC_OFFSET(14066, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv), + NAME_FUNC_OFFSET(14093, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable), + NAME_FUNC_OFFSET(14113, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), + NAME_FUNC_OFFSET(14132, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), + NAME_FUNC_OFFSET(14151, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), + NAME_FUNC_OFFSET(14181, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), + NAME_FUNC_OFFSET(14211, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), + NAME_FUNC_OFFSET(14241, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), + NAME_FUNC_OFFSET(14271, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable), + NAME_FUNC_OFFSET(14290, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable), + NAME_FUNC_OFFSET(14313, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D), + NAME_FUNC_OFFSET(14338, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D), + NAME_FUNC_OFFSET(14363, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf), + NAME_FUNC_OFFSET(14390, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv), + NAME_FUNC_OFFSET(14418, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri), + NAME_FUNC_OFFSET(14445, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv), + NAME_FUNC_OFFSET(14473, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D), + NAME_FUNC_OFFSET(14502, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D), + NAME_FUNC_OFFSET(14531, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter), + NAME_FUNC_OFFSET(14557, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv), + NAME_FUNC_OFFSET(14588, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv), + NAME_FUNC_OFFSET(14619, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter), + NAME_FUNC_OFFSET(14643, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D), + NAME_FUNC_OFFSET(14666, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram), + NAME_FUNC_OFFSET(14684, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv), + NAME_FUNC_OFFSET(14713, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv), + NAME_FUNC_OFFSET(14742, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax), + NAME_FUNC_OFFSET(14757, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv), + NAME_FUNC_OFFSET(14783, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv), + NAME_FUNC_OFFSET(14809, glHistogram, glHistogram, NULL, _gloffset_Histogram), + NAME_FUNC_OFFSET(14824, glMinmax, glMinmax, NULL, _gloffset_Minmax), + NAME_FUNC_OFFSET(14836, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram), + NAME_FUNC_OFFSET(14856, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax), + NAME_FUNC_OFFSET(14873, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D), + NAME_FUNC_OFFSET(14889, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D), + NAME_FUNC_OFFSET(14908, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D), + NAME_FUNC_OFFSET(14931, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB), + NAME_FUNC_OFFSET(14947, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB), + NAME_FUNC_OFFSET(14969, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB), + NAME_FUNC_OFFSET(14987, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB), + NAME_FUNC_OFFSET(15006, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB), + NAME_FUNC_OFFSET(15024, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB), + NAME_FUNC_OFFSET(15043, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB), + NAME_FUNC_OFFSET(15061, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB), + NAME_FUNC_OFFSET(15080, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB), + NAME_FUNC_OFFSET(15098, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB), + NAME_FUNC_OFFSET(15117, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB), + NAME_FUNC_OFFSET(15135, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB), + NAME_FUNC_OFFSET(15154, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB), + NAME_FUNC_OFFSET(15172, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB), + NAME_FUNC_OFFSET(15191, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB), + NAME_FUNC_OFFSET(15209, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB), + NAME_FUNC_OFFSET(15228, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB), + NAME_FUNC_OFFSET(15246, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB), + NAME_FUNC_OFFSET(15265, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB), + NAME_FUNC_OFFSET(15283, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB), + NAME_FUNC_OFFSET(15302, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB), + NAME_FUNC_OFFSET(15320, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB), + NAME_FUNC_OFFSET(15339, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB), + NAME_FUNC_OFFSET(15357, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB), + NAME_FUNC_OFFSET(15376, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB), + NAME_FUNC_OFFSET(15394, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB), + NAME_FUNC_OFFSET(15413, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB), + NAME_FUNC_OFFSET(15431, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB), + NAME_FUNC_OFFSET(15450, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB), + NAME_FUNC_OFFSET(15468, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB), + NAME_FUNC_OFFSET(15487, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB), + NAME_FUNC_OFFSET(15505, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB), + NAME_FUNC_OFFSET(15524, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB), + NAME_FUNC_OFFSET(15542, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB), + NAME_FUNC_OFFSET(15561, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB), + NAME_FUNC_OFFSET(15584, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB), + NAME_FUNC_OFFSET(15607, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB), + NAME_FUNC_OFFSET(15630, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB), + NAME_FUNC_OFFSET(15653, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB), + NAME_FUNC_OFFSET(15670, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB), + NAME_FUNC_OFFSET(15693, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB), + NAME_FUNC_OFFSET(15716, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB), + NAME_FUNC_OFFSET(15739, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB), + NAME_FUNC_OFFSET(15765, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB), + NAME_FUNC_OFFSET(15791, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB), + NAME_FUNC_OFFSET(15817, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB), + NAME_FUNC_OFFSET(15841, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB), + NAME_FUNC_OFFSET(15868, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB), + NAME_FUNC_OFFSET(15894, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB), + NAME_FUNC_OFFSET(15914, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB), + NAME_FUNC_OFFSET(15934, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB), + NAME_FUNC_OFFSET(15954, glVertexAttrib1dARB, glVertexAttrib1dARB, NULL, _gloffset_VertexAttrib1dARB), + NAME_FUNC_OFFSET(15971, glVertexAttrib1dvARB, glVertexAttrib1dvARB, NULL, _gloffset_VertexAttrib1dvARB), + NAME_FUNC_OFFSET(15989, glVertexAttrib1fARB, glVertexAttrib1fARB, NULL, _gloffset_VertexAttrib1fARB), + NAME_FUNC_OFFSET(16006, glVertexAttrib1fvARB, glVertexAttrib1fvARB, NULL, _gloffset_VertexAttrib1fvARB), + NAME_FUNC_OFFSET(16024, glVertexAttrib1sARB, glVertexAttrib1sARB, NULL, _gloffset_VertexAttrib1sARB), + NAME_FUNC_OFFSET(16041, glVertexAttrib1svARB, glVertexAttrib1svARB, NULL, _gloffset_VertexAttrib1svARB), + NAME_FUNC_OFFSET(16059, glVertexAttrib2dARB, glVertexAttrib2dARB, NULL, _gloffset_VertexAttrib2dARB), + NAME_FUNC_OFFSET(16076, glVertexAttrib2dvARB, glVertexAttrib2dvARB, NULL, _gloffset_VertexAttrib2dvARB), + NAME_FUNC_OFFSET(16094, glVertexAttrib2fARB, glVertexAttrib2fARB, NULL, _gloffset_VertexAttrib2fARB), + NAME_FUNC_OFFSET(16111, glVertexAttrib2fvARB, glVertexAttrib2fvARB, NULL, _gloffset_VertexAttrib2fvARB), + NAME_FUNC_OFFSET(16129, glVertexAttrib2sARB, glVertexAttrib2sARB, NULL, _gloffset_VertexAttrib2sARB), + NAME_FUNC_OFFSET(16146, glVertexAttrib2svARB, glVertexAttrib2svARB, NULL, _gloffset_VertexAttrib2svARB), + NAME_FUNC_OFFSET(16164, glVertexAttrib3dARB, glVertexAttrib3dARB, NULL, _gloffset_VertexAttrib3dARB), + NAME_FUNC_OFFSET(16181, glVertexAttrib3dvARB, glVertexAttrib3dvARB, NULL, _gloffset_VertexAttrib3dvARB), + NAME_FUNC_OFFSET(16199, glVertexAttrib3fARB, glVertexAttrib3fARB, NULL, _gloffset_VertexAttrib3fARB), + NAME_FUNC_OFFSET(16216, glVertexAttrib3fvARB, glVertexAttrib3fvARB, NULL, _gloffset_VertexAttrib3fvARB), + NAME_FUNC_OFFSET(16234, glVertexAttrib3sARB, glVertexAttrib3sARB, NULL, _gloffset_VertexAttrib3sARB), + NAME_FUNC_OFFSET(16251, glVertexAttrib3svARB, glVertexAttrib3svARB, NULL, _gloffset_VertexAttrib3svARB), + NAME_FUNC_OFFSET(16269, glVertexAttrib4NbvARB, glVertexAttrib4NbvARB, NULL, _gloffset_VertexAttrib4NbvARB), + NAME_FUNC_OFFSET(16288, glVertexAttrib4NivARB, glVertexAttrib4NivARB, NULL, _gloffset_VertexAttrib4NivARB), + NAME_FUNC_OFFSET(16307, glVertexAttrib4NsvARB, glVertexAttrib4NsvARB, NULL, _gloffset_VertexAttrib4NsvARB), + NAME_FUNC_OFFSET(16326, glVertexAttrib4NubARB, glVertexAttrib4NubARB, NULL, _gloffset_VertexAttrib4NubARB), + NAME_FUNC_OFFSET(16345, glVertexAttrib4NubvARB, glVertexAttrib4NubvARB, NULL, _gloffset_VertexAttrib4NubvARB), + NAME_FUNC_OFFSET(16365, glVertexAttrib4NuivARB, glVertexAttrib4NuivARB, NULL, _gloffset_VertexAttrib4NuivARB), + NAME_FUNC_OFFSET(16385, glVertexAttrib4NusvARB, glVertexAttrib4NusvARB, NULL, _gloffset_VertexAttrib4NusvARB), + NAME_FUNC_OFFSET(16405, glVertexAttrib4bvARB, glVertexAttrib4bvARB, NULL, _gloffset_VertexAttrib4bvARB), + NAME_FUNC_OFFSET(16423, glVertexAttrib4dARB, glVertexAttrib4dARB, NULL, _gloffset_VertexAttrib4dARB), + NAME_FUNC_OFFSET(16440, glVertexAttrib4dvARB, glVertexAttrib4dvARB, NULL, _gloffset_VertexAttrib4dvARB), + NAME_FUNC_OFFSET(16458, glVertexAttrib4fARB, glVertexAttrib4fARB, NULL, _gloffset_VertexAttrib4fARB), + NAME_FUNC_OFFSET(16475, glVertexAttrib4fvARB, glVertexAttrib4fvARB, NULL, _gloffset_VertexAttrib4fvARB), + NAME_FUNC_OFFSET(16493, glVertexAttrib4ivARB, glVertexAttrib4ivARB, NULL, _gloffset_VertexAttrib4ivARB), + NAME_FUNC_OFFSET(16511, glVertexAttrib4sARB, glVertexAttrib4sARB, NULL, _gloffset_VertexAttrib4sARB), + NAME_FUNC_OFFSET(16528, glVertexAttrib4svARB, glVertexAttrib4svARB, NULL, _gloffset_VertexAttrib4svARB), + NAME_FUNC_OFFSET(16546, glVertexAttrib4ubvARB, glVertexAttrib4ubvARB, NULL, _gloffset_VertexAttrib4ubvARB), + NAME_FUNC_OFFSET(16565, glVertexAttrib4uivARB, glVertexAttrib4uivARB, NULL, _gloffset_VertexAttrib4uivARB), + NAME_FUNC_OFFSET(16584, glVertexAttrib4usvARB, glVertexAttrib4usvARB, NULL, _gloffset_VertexAttrib4usvARB), + NAME_FUNC_OFFSET(16603, glVertexAttribPointerARB, glVertexAttribPointerARB, NULL, _gloffset_VertexAttribPointerARB), + NAME_FUNC_OFFSET(16625, glBindBufferARB, glBindBufferARB, NULL, _gloffset_BindBufferARB), + NAME_FUNC_OFFSET(16638, glBufferDataARB, glBufferDataARB, NULL, _gloffset_BufferDataARB), + NAME_FUNC_OFFSET(16651, glBufferSubDataARB, glBufferSubDataARB, NULL, _gloffset_BufferSubDataARB), + NAME_FUNC_OFFSET(16667, glDeleteBuffersARB, glDeleteBuffersARB, NULL, _gloffset_DeleteBuffersARB), + NAME_FUNC_OFFSET(16683, glGenBuffersARB, glGenBuffersARB, NULL, _gloffset_GenBuffersARB), + NAME_FUNC_OFFSET(16696, glGetBufferParameterivARB, glGetBufferParameterivARB, NULL, _gloffset_GetBufferParameterivARB), + NAME_FUNC_OFFSET(16719, glGetBufferPointervARB, glGetBufferPointervARB, NULL, _gloffset_GetBufferPointervARB), + NAME_FUNC_OFFSET(16739, glGetBufferSubDataARB, glGetBufferSubDataARB, NULL, _gloffset_GetBufferSubDataARB), + NAME_FUNC_OFFSET(16758, glIsBufferARB, glIsBufferARB, NULL, _gloffset_IsBufferARB), + NAME_FUNC_OFFSET(16769, glMapBufferARB, glMapBufferARB, NULL, _gloffset_MapBufferARB), + NAME_FUNC_OFFSET(16781, glUnmapBufferARB, glUnmapBufferARB, NULL, _gloffset_UnmapBufferARB), + NAME_FUNC_OFFSET(16795, glBeginQueryARB, glBeginQueryARB, NULL, _gloffset_BeginQueryARB), + NAME_FUNC_OFFSET(16808, glDeleteQueriesARB, glDeleteQueriesARB, NULL, _gloffset_DeleteQueriesARB), + NAME_FUNC_OFFSET(16824, glEndQueryARB, glEndQueryARB, NULL, _gloffset_EndQueryARB), + NAME_FUNC_OFFSET(16835, glGenQueriesARB, glGenQueriesARB, NULL, _gloffset_GenQueriesARB), + NAME_FUNC_OFFSET(16848, glGetQueryObjectivARB, glGetQueryObjectivARB, NULL, _gloffset_GetQueryObjectivARB), + NAME_FUNC_OFFSET(16867, glGetQueryObjectuivARB, glGetQueryObjectuivARB, NULL, _gloffset_GetQueryObjectuivARB), + NAME_FUNC_OFFSET(16887, glGetQueryivARB, glGetQueryivARB, NULL, _gloffset_GetQueryivARB), + NAME_FUNC_OFFSET(16900, glIsQueryARB, glIsQueryARB, NULL, _gloffset_IsQueryARB), + NAME_FUNC_OFFSET(16910, glCompileShaderARB, glCompileShaderARB, NULL, _gloffset_CompileShaderARB), + NAME_FUNC_OFFSET(16926, glGetActiveUniformARB, glGetActiveUniformARB, NULL, _gloffset_GetActiveUniformARB), + NAME_FUNC_OFFSET(16945, glGetShaderSourceARB, glGetShaderSourceARB, NULL, _gloffset_GetShaderSourceARB), + NAME_FUNC_OFFSET(16963, glGetUniformLocationARB, glGetUniformLocationARB, NULL, _gloffset_GetUniformLocationARB), + NAME_FUNC_OFFSET(16984, glGetUniformfvARB, glGetUniformfvARB, NULL, _gloffset_GetUniformfvARB), + NAME_FUNC_OFFSET(16999, glGetUniformivARB, glGetUniformivARB, NULL, _gloffset_GetUniformivARB), + NAME_FUNC_OFFSET(17014, glLinkProgramARB, glLinkProgramARB, NULL, _gloffset_LinkProgramARB), + NAME_FUNC_OFFSET(17028, glShaderSourceARB, glShaderSourceARB, NULL, _gloffset_ShaderSourceARB), + NAME_FUNC_OFFSET(17043, glUniform1fARB, glUniform1fARB, NULL, _gloffset_Uniform1fARB), + NAME_FUNC_OFFSET(17055, glUniform1fvARB, glUniform1fvARB, NULL, _gloffset_Uniform1fvARB), + NAME_FUNC_OFFSET(17068, glUniform1iARB, glUniform1iARB, NULL, _gloffset_Uniform1iARB), + NAME_FUNC_OFFSET(17080, glUniform1ivARB, glUniform1ivARB, NULL, _gloffset_Uniform1ivARB), + NAME_FUNC_OFFSET(17093, glUniform2fARB, glUniform2fARB, NULL, _gloffset_Uniform2fARB), + NAME_FUNC_OFFSET(17105, glUniform2fvARB, glUniform2fvARB, NULL, _gloffset_Uniform2fvARB), + NAME_FUNC_OFFSET(17118, glUniform2iARB, glUniform2iARB, NULL, _gloffset_Uniform2iARB), + NAME_FUNC_OFFSET(17130, glUniform2ivARB, glUniform2ivARB, NULL, _gloffset_Uniform2ivARB), + NAME_FUNC_OFFSET(17143, glUniform3fARB, glUniform3fARB, NULL, _gloffset_Uniform3fARB), + NAME_FUNC_OFFSET(17155, glUniform3fvARB, glUniform3fvARB, NULL, _gloffset_Uniform3fvARB), + NAME_FUNC_OFFSET(17168, glUniform3iARB, glUniform3iARB, NULL, _gloffset_Uniform3iARB), + NAME_FUNC_OFFSET(17180, glUniform3ivARB, glUniform3ivARB, NULL, _gloffset_Uniform3ivARB), + NAME_FUNC_OFFSET(17193, glUniform4fARB, glUniform4fARB, NULL, _gloffset_Uniform4fARB), + NAME_FUNC_OFFSET(17205, glUniform4fvARB, glUniform4fvARB, NULL, _gloffset_Uniform4fvARB), + NAME_FUNC_OFFSET(17218, glUniform4iARB, glUniform4iARB, NULL, _gloffset_Uniform4iARB), + NAME_FUNC_OFFSET(17230, glUniform4ivARB, glUniform4ivARB, NULL, _gloffset_Uniform4ivARB), + NAME_FUNC_OFFSET(17243, glUniformMatrix2fvARB, glUniformMatrix2fvARB, NULL, _gloffset_UniformMatrix2fvARB), + NAME_FUNC_OFFSET(17262, glUniformMatrix3fvARB, glUniformMatrix3fvARB, NULL, _gloffset_UniformMatrix3fvARB), + NAME_FUNC_OFFSET(17281, glUniformMatrix4fvARB, glUniformMatrix4fvARB, NULL, _gloffset_UniformMatrix4fvARB), + NAME_FUNC_OFFSET(17300, glUseProgramObjectARB, glUseProgramObjectARB, NULL, _gloffset_UseProgramObjectARB), + NAME_FUNC_OFFSET(17313, glValidateProgramARB, glValidateProgramARB, NULL, _gloffset_ValidateProgramARB), + NAME_FUNC_OFFSET(17331, glBindAttribLocationARB, glBindAttribLocationARB, NULL, _gloffset_BindAttribLocationARB), + NAME_FUNC_OFFSET(17352, glGetActiveAttribARB, glGetActiveAttribARB, NULL, _gloffset_GetActiveAttribARB), + NAME_FUNC_OFFSET(17370, glGetAttribLocationARB, glGetAttribLocationARB, NULL, _gloffset_GetAttribLocationARB), + NAME_FUNC_OFFSET(17390, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB), + NAME_FUNC_OFFSET(17404, glDrawBuffersARB, glDrawBuffersARB, NULL, _gloffset_DrawBuffersARB), + NAME_FUNC_OFFSET(17421, gl_dispatch_stub_568, gl_dispatch_stub_568, NULL, _gloffset_SampleMaskSGIS), + NAME_FUNC_OFFSET(17437, gl_dispatch_stub_569, gl_dispatch_stub_569, NULL, _gloffset_SamplePatternSGIS), + NAME_FUNC_OFFSET(17456, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT), + NAME_FUNC_OFFSET(17474, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT), + NAME_FUNC_OFFSET(17495, glPointParameterfEXT, glPointParameterfEXT, NULL, _gloffset_PointParameterfEXT), + NAME_FUNC_OFFSET(17517, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT), + NAME_FUNC_OFFSET(17536, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT), + NAME_FUNC_OFFSET(17558, glPointParameterfvEXT, glPointParameterfvEXT, NULL, _gloffset_PointParameterfvEXT), + NAME_FUNC_OFFSET(17581, glSecondaryColor3bEXT, glSecondaryColor3bEXT, NULL, _gloffset_SecondaryColor3bEXT), + NAME_FUNC_OFFSET(17600, glSecondaryColor3bvEXT, glSecondaryColor3bvEXT, NULL, _gloffset_SecondaryColor3bvEXT), + NAME_FUNC_OFFSET(17620, glSecondaryColor3dEXT, glSecondaryColor3dEXT, NULL, _gloffset_SecondaryColor3dEXT), + NAME_FUNC_OFFSET(17639, glSecondaryColor3dvEXT, glSecondaryColor3dvEXT, NULL, _gloffset_SecondaryColor3dvEXT), + NAME_FUNC_OFFSET(17659, glSecondaryColor3fEXT, glSecondaryColor3fEXT, NULL, _gloffset_SecondaryColor3fEXT), + NAME_FUNC_OFFSET(17678, glSecondaryColor3fvEXT, glSecondaryColor3fvEXT, NULL, _gloffset_SecondaryColor3fvEXT), + NAME_FUNC_OFFSET(17698, glSecondaryColor3iEXT, glSecondaryColor3iEXT, NULL, _gloffset_SecondaryColor3iEXT), + NAME_FUNC_OFFSET(17717, glSecondaryColor3ivEXT, glSecondaryColor3ivEXT, NULL, _gloffset_SecondaryColor3ivEXT), + NAME_FUNC_OFFSET(17737, glSecondaryColor3sEXT, glSecondaryColor3sEXT, NULL, _gloffset_SecondaryColor3sEXT), + NAME_FUNC_OFFSET(17756, glSecondaryColor3svEXT, glSecondaryColor3svEXT, NULL, _gloffset_SecondaryColor3svEXT), + NAME_FUNC_OFFSET(17776, glSecondaryColor3ubEXT, glSecondaryColor3ubEXT, NULL, _gloffset_SecondaryColor3ubEXT), + NAME_FUNC_OFFSET(17796, glSecondaryColor3ubvEXT, glSecondaryColor3ubvEXT, NULL, _gloffset_SecondaryColor3ubvEXT), + NAME_FUNC_OFFSET(17817, glSecondaryColor3uiEXT, glSecondaryColor3uiEXT, NULL, _gloffset_SecondaryColor3uiEXT), + NAME_FUNC_OFFSET(17837, glSecondaryColor3uivEXT, glSecondaryColor3uivEXT, NULL, _gloffset_SecondaryColor3uivEXT), + NAME_FUNC_OFFSET(17858, glSecondaryColor3usEXT, glSecondaryColor3usEXT, NULL, _gloffset_SecondaryColor3usEXT), + NAME_FUNC_OFFSET(17878, glSecondaryColor3usvEXT, glSecondaryColor3usvEXT, NULL, _gloffset_SecondaryColor3usvEXT), + NAME_FUNC_OFFSET(17899, glSecondaryColorPointerEXT, glSecondaryColorPointerEXT, NULL, _gloffset_SecondaryColorPointerEXT), + NAME_FUNC_OFFSET(17923, glMultiDrawArraysEXT, glMultiDrawArraysEXT, NULL, _gloffset_MultiDrawArraysEXT), + NAME_FUNC_OFFSET(17941, glMultiDrawElementsEXT, glMultiDrawElementsEXT, NULL, _gloffset_MultiDrawElementsEXT), + NAME_FUNC_OFFSET(17961, glFogCoordPointerEXT, glFogCoordPointerEXT, NULL, _gloffset_FogCoordPointerEXT), + NAME_FUNC_OFFSET(17979, glFogCoorddEXT, glFogCoorddEXT, NULL, _gloffset_FogCoorddEXT), + NAME_FUNC_OFFSET(17991, glFogCoorddvEXT, glFogCoorddvEXT, NULL, _gloffset_FogCoorddvEXT), + NAME_FUNC_OFFSET(18004, glFogCoordfEXT, glFogCoordfEXT, NULL, _gloffset_FogCoordfEXT), + NAME_FUNC_OFFSET(18016, glFogCoordfvEXT, glFogCoordfvEXT, NULL, _gloffset_FogCoordfvEXT), + NAME_FUNC_OFFSET(18029, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT), + NAME_FUNC_OFFSET(18049, glBlendFuncSeparateEXT, glBlendFuncSeparateEXT, NULL, _gloffset_BlendFuncSeparateEXT), + NAME_FUNC_OFFSET(18073, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA), + NAME_FUNC_OFFSET(18087, glWindowPos2dMESA, glWindowPos2dMESA, NULL, _gloffset_WindowPos2dMESA), + NAME_FUNC_OFFSET(18104, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA), + NAME_FUNC_OFFSET(18119, glWindowPos2dvMESA, glWindowPos2dvMESA, NULL, _gloffset_WindowPos2dvMESA), + NAME_FUNC_OFFSET(18137, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA), + NAME_FUNC_OFFSET(18151, glWindowPos2fMESA, glWindowPos2fMESA, NULL, _gloffset_WindowPos2fMESA), + NAME_FUNC_OFFSET(18168, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA), + NAME_FUNC_OFFSET(18183, glWindowPos2fvMESA, glWindowPos2fvMESA, NULL, _gloffset_WindowPos2fvMESA), + NAME_FUNC_OFFSET(18201, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA), + NAME_FUNC_OFFSET(18215, glWindowPos2iMESA, glWindowPos2iMESA, NULL, _gloffset_WindowPos2iMESA), + NAME_FUNC_OFFSET(18232, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA), + NAME_FUNC_OFFSET(18247, glWindowPos2ivMESA, glWindowPos2ivMESA, NULL, _gloffset_WindowPos2ivMESA), + NAME_FUNC_OFFSET(18265, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA), + NAME_FUNC_OFFSET(18279, glWindowPos2sMESA, glWindowPos2sMESA, NULL, _gloffset_WindowPos2sMESA), + NAME_FUNC_OFFSET(18296, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA), + NAME_FUNC_OFFSET(18311, glWindowPos2svMESA, glWindowPos2svMESA, NULL, _gloffset_WindowPos2svMESA), + NAME_FUNC_OFFSET(18329, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA), + NAME_FUNC_OFFSET(18343, glWindowPos3dMESA, glWindowPos3dMESA, NULL, _gloffset_WindowPos3dMESA), + NAME_FUNC_OFFSET(18360, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA), + NAME_FUNC_OFFSET(18375, glWindowPos3dvMESA, glWindowPos3dvMESA, NULL, _gloffset_WindowPos3dvMESA), + NAME_FUNC_OFFSET(18393, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA), + NAME_FUNC_OFFSET(18407, glWindowPos3fMESA, glWindowPos3fMESA, NULL, _gloffset_WindowPos3fMESA), + NAME_FUNC_OFFSET(18424, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA), + NAME_FUNC_OFFSET(18439, glWindowPos3fvMESA, glWindowPos3fvMESA, NULL, _gloffset_WindowPos3fvMESA), + NAME_FUNC_OFFSET(18457, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA), + NAME_FUNC_OFFSET(18471, glWindowPos3iMESA, glWindowPos3iMESA, NULL, _gloffset_WindowPos3iMESA), + NAME_FUNC_OFFSET(18488, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA), + NAME_FUNC_OFFSET(18503, glWindowPos3ivMESA, glWindowPos3ivMESA, NULL, _gloffset_WindowPos3ivMESA), + NAME_FUNC_OFFSET(18521, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA), + NAME_FUNC_OFFSET(18535, glWindowPos3sMESA, glWindowPos3sMESA, NULL, _gloffset_WindowPos3sMESA), + NAME_FUNC_OFFSET(18552, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA), + NAME_FUNC_OFFSET(18567, glWindowPos3svMESA, glWindowPos3svMESA, NULL, _gloffset_WindowPos3svMESA), + NAME_FUNC_OFFSET(18585, glBindProgramNV, glBindProgramNV, NULL, _gloffset_BindProgramNV), + NAME_FUNC_OFFSET(18602, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV), + NAME_FUNC_OFFSET(18622, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV), + NAME_FUNC_OFFSET(18639, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV), + NAME_FUNC_OFFSET(18665, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV), + NAME_FUNC_OFFSET(18694, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV), + NAME_FUNC_OFFSET(18709, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV), + NAME_FUNC_OFFSET(18727, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV), + NAME_FUNC_OFFSET(18746, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT), + NAME_FUNC_OFFSET(18770, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT), NAME_FUNC_OFFSET(-1, NULL, NULL, NULL, 0) }; diff --git a/src/mesa/main/enums.c b/src/mesa/main/enums.c index d5019ae045..6aeb18fa27 100644 --- a/src/mesa/main/enums.c +++ b/src/mesa/main/enums.c @@ -1433,9 +1433,13 @@ LONGSTRING static const char enum_string_table[] = "GL_STENCIL\0" "GL_STENCIL_ATTACHMENT_EXT\0" "GL_STENCIL_BACK_FAIL\0" + "GL_STENCIL_BACK_FAIL_ATI\0" "GL_STENCIL_BACK_FUNC\0" + "GL_STENCIL_BACK_FUNC_ATI\0" "GL_STENCIL_BACK_PASS_DEPTH_FAIL\0" + "GL_STENCIL_BACK_PASS_DEPTH_FAIL_ATI\0" "GL_STENCIL_BACK_PASS_DEPTH_PASS\0" + "GL_STENCIL_BACK_PASS_DEPTH_PASS_ATI\0" "GL_STENCIL_BACK_REF\0" "GL_STENCIL_BACK_VALUE_MASK\0" "GL_STENCIL_BACK_WRITEMASK\0" @@ -1783,7 +1787,7 @@ LONGSTRING static const char enum_string_table[] = "GL_ZOOM_Y\0" ; -static const enum_elt all_enums[1746] = +static const enum_elt all_enums[1750] = { { 0, 0x00000600 }, /* GL_2D */ { 6, 0x00001407 }, /* GL_2_BYTES */ @@ -3183,354 +3187,358 @@ static const enum_elt all_enums[1746] = { 29675, 0x00001802 }, /* GL_STENCIL */ { 29686, 0x00008D20 }, /* GL_STENCIL_ATTACHMENT_EXT */ { 29712, 0x00008801 }, /* GL_STENCIL_BACK_FAIL */ - { 29733, 0x00008800 }, /* GL_STENCIL_BACK_FUNC */ - { 29754, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */ - { 29786, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */ - { 29818, 0x00008CA3 }, /* GL_STENCIL_BACK_REF */ - { 29838, 0x00008CA4 }, /* GL_STENCIL_BACK_VALUE_MASK */ - { 29865, 0x00008CA5 }, /* GL_STENCIL_BACK_WRITEMASK */ - { 29891, 0x00000D57 }, /* GL_STENCIL_BITS */ - { 29907, 0x00000400 }, /* GL_STENCIL_BUFFER_BIT */ - { 29929, 0x00000B91 }, /* GL_STENCIL_CLEAR_VALUE */ - { 29952, 0x00000B94 }, /* GL_STENCIL_FAIL */ - { 29968, 0x00000B92 }, /* GL_STENCIL_FUNC */ - { 29984, 0x00001901 }, /* GL_STENCIL_INDEX */ - { 30001, 0x00008D49 }, /* GL_STENCIL_INDEX16_EXT */ - { 30024, 0x00008D46 }, /* GL_STENCIL_INDEX1_EXT */ - { 30046, 0x00008D47 }, /* GL_STENCIL_INDEX4_EXT */ - { 30068, 0x00008D48 }, /* GL_STENCIL_INDEX8_EXT */ - { 30090, 0x00008D45 }, /* GL_STENCIL_INDEX_EXT */ - { 30111, 0x00000B95 }, /* GL_STENCIL_PASS_DEPTH_FAIL */ - { 30138, 0x00000B96 }, /* GL_STENCIL_PASS_DEPTH_PASS */ - { 30165, 0x00000B97 }, /* GL_STENCIL_REF */ - { 30180, 0x00000B90 }, /* GL_STENCIL_TEST */ - { 30196, 0x00008910 }, /* GL_STENCIL_TEST_TWO_SIDE_EXT */ - { 30225, 0x00000B93 }, /* GL_STENCIL_VALUE_MASK */ - { 30247, 0x00000B98 }, /* GL_STENCIL_WRITEMASK */ - { 30268, 0x00000C33 }, /* GL_STEREO */ - { 30278, 0x000088E2 }, /* GL_STREAM_COPY */ - { 30293, 0x000088E2 }, /* GL_STREAM_COPY_ARB */ - { 30312, 0x000088E0 }, /* GL_STREAM_DRAW */ - { 30327, 0x000088E0 }, /* GL_STREAM_DRAW_ARB */ - { 30346, 0x000088E1 }, /* GL_STREAM_READ */ - { 30361, 0x000088E1 }, /* GL_STREAM_READ_ARB */ - { 30380, 0x00000D50 }, /* GL_SUBPIXEL_BITS */ - { 30397, 0x000084E7 }, /* GL_SUBTRACT */ - { 30409, 0x000084E7 }, /* GL_SUBTRACT_ARB */ - { 30425, 0x00002001 }, /* GL_T */ - { 30430, 0x00002A2A }, /* GL_T2F_C3F_V3F */ - { 30445, 0x00002A2C }, /* GL_T2F_C4F_N3F_V3F */ - { 30464, 0x00002A29 }, /* GL_T2F_C4UB_V3F */ - { 30480, 0x00002A2B }, /* GL_T2F_N3F_V3F */ - { 30495, 0x00002A27 }, /* GL_T2F_V3F */ - { 30506, 0x00002A2D }, /* GL_T4F_C4F_N3F_V4F */ - { 30525, 0x00002A28 }, /* GL_T4F_V4F */ - { 30536, 0x00008031 }, /* GL_TABLE_TOO_LARGE_EXT */ - { 30559, 0x00001702 }, /* GL_TEXTURE */ - { 30570, 0x000084C0 }, /* GL_TEXTURE0 */ - { 30582, 0x000084C0 }, /* GL_TEXTURE0_ARB */ - { 30598, 0x000084C1 }, /* GL_TEXTURE1 */ - { 30610, 0x000084CA }, /* GL_TEXTURE10 */ - { 30623, 0x000084CA }, /* GL_TEXTURE10_ARB */ - { 30640, 0x000084CB }, /* GL_TEXTURE11 */ - { 30653, 0x000084CB }, /* GL_TEXTURE11_ARB */ - { 30670, 0x000084CC }, /* GL_TEXTURE12 */ - { 30683, 0x000084CC }, /* GL_TEXTURE12_ARB */ - { 30700, 0x000084CD }, /* GL_TEXTURE13 */ - { 30713, 0x000084CD }, /* GL_TEXTURE13_ARB */ - { 30730, 0x000084CE }, /* GL_TEXTURE14 */ - { 30743, 0x000084CE }, /* GL_TEXTURE14_ARB */ - { 30760, 0x000084CF }, /* GL_TEXTURE15 */ - { 30773, 0x000084CF }, /* GL_TEXTURE15_ARB */ - { 30790, 0x000084D0 }, /* GL_TEXTURE16 */ - { 30803, 0x000084D0 }, /* GL_TEXTURE16_ARB */ - { 30820, 0x000084D1 }, /* GL_TEXTURE17 */ - { 30833, 0x000084D1 }, /* GL_TEXTURE17_ARB */ - { 30850, 0x000084D2 }, /* GL_TEXTURE18 */ - { 30863, 0x000084D2 }, /* GL_TEXTURE18_ARB */ - { 30880, 0x000084D3 }, /* GL_TEXTURE19 */ - { 30893, 0x000084D3 }, /* GL_TEXTURE19_ARB */ - { 30910, 0x000084C1 }, /* GL_TEXTURE1_ARB */ - { 30926, 0x000084C2 }, /* GL_TEXTURE2 */ - { 30938, 0x000084D4 }, /* GL_TEXTURE20 */ - { 30951, 0x000084D4 }, /* GL_TEXTURE20_ARB */ - { 30968, 0x000084D5 }, /* GL_TEXTURE21 */ - { 30981, 0x000084D5 }, /* GL_TEXTURE21_ARB */ - { 30998, 0x000084D6 }, /* GL_TEXTURE22 */ - { 31011, 0x000084D6 }, /* GL_TEXTURE22_ARB */ - { 31028, 0x000084D7 }, /* GL_TEXTURE23 */ - { 31041, 0x000084D7 }, /* GL_TEXTURE23_ARB */ - { 31058, 0x000084D8 }, /* GL_TEXTURE24 */ - { 31071, 0x000084D8 }, /* GL_TEXTURE24_ARB */ - { 31088, 0x000084D9 }, /* GL_TEXTURE25 */ - { 31101, 0x000084D9 }, /* GL_TEXTURE25_ARB */ - { 31118, 0x000084DA }, /* GL_TEXTURE26 */ - { 31131, 0x000084DA }, /* GL_TEXTURE26_ARB */ - { 31148, 0x000084DB }, /* GL_TEXTURE27 */ - { 31161, 0x000084DB }, /* GL_TEXTURE27_ARB */ - { 31178, 0x000084DC }, /* GL_TEXTURE28 */ - { 31191, 0x000084DC }, /* GL_TEXTURE28_ARB */ - { 31208, 0x000084DD }, /* GL_TEXTURE29 */ - { 31221, 0x000084DD }, /* GL_TEXTURE29_ARB */ - { 31238, 0x000084C2 }, /* GL_TEXTURE2_ARB */ - { 31254, 0x000084C3 }, /* GL_TEXTURE3 */ - { 31266, 0x000084DE }, /* GL_TEXTURE30 */ - { 31279, 0x000084DE }, /* GL_TEXTURE30_ARB */ - { 31296, 0x000084DF }, /* GL_TEXTURE31 */ - { 31309, 0x000084DF }, /* GL_TEXTURE31_ARB */ - { 31326, 0x000084C3 }, /* GL_TEXTURE3_ARB */ - { 31342, 0x000084C4 }, /* GL_TEXTURE4 */ - { 31354, 0x000084C4 }, /* GL_TEXTURE4_ARB */ - { 31370, 0x000084C5 }, /* GL_TEXTURE5 */ - { 31382, 0x000084C5 }, /* GL_TEXTURE5_ARB */ - { 31398, 0x000084C6 }, /* GL_TEXTURE6 */ - { 31410, 0x000084C6 }, /* GL_TEXTURE6_ARB */ - { 31426, 0x000084C7 }, /* GL_TEXTURE7 */ - { 31438, 0x000084C7 }, /* GL_TEXTURE7_ARB */ - { 31454, 0x000084C8 }, /* GL_TEXTURE8 */ - { 31466, 0x000084C8 }, /* GL_TEXTURE8_ARB */ - { 31482, 0x000084C9 }, /* GL_TEXTURE9 */ - { 31494, 0x000084C9 }, /* GL_TEXTURE9_ARB */ - { 31510, 0x00000DE0 }, /* GL_TEXTURE_1D */ - { 31524, 0x00008C18 }, /* GL_TEXTURE_1D_ARRAY_EXT */ - { 31548, 0x00000DE1 }, /* GL_TEXTURE_2D */ - { 31562, 0x00008C1A }, /* GL_TEXTURE_2D_ARRAY_EXT */ - { 31586, 0x0000806F }, /* GL_TEXTURE_3D */ - { 31600, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE */ - { 31622, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE_EXT */ - { 31648, 0x0000813C }, /* GL_TEXTURE_BASE_LEVEL */ - { 31670, 0x00008068 }, /* GL_TEXTURE_BINDING_1D */ - { 31692, 0x00008C1C }, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */ - { 31724, 0x00008069 }, /* GL_TEXTURE_BINDING_2D */ - { 31746, 0x00008C1D }, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */ - { 31778, 0x0000806A }, /* GL_TEXTURE_BINDING_3D */ - { 31800, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP */ - { 31828, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP_ARB */ - { 31860, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */ - { 31893, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_NV */ - { 31925, 0x00040000 }, /* GL_TEXTURE_BIT */ - { 31940, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE */ - { 31961, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE_EXT */ - { 31986, 0x00001005 }, /* GL_TEXTURE_BORDER */ - { 32004, 0x00001004 }, /* GL_TEXTURE_BORDER_COLOR */ - { 32028, 0x00008171 }, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */ - { 32059, 0x00008176 }, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */ - { 32089, 0x00008172 }, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */ - { 32119, 0x00008175 }, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */ - { 32154, 0x00008173 }, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */ - { 32185, 0x00008174 }, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */ - { 32223, 0x000080BC }, /* GL_TEXTURE_COLOR_TABLE_SGI */ - { 32250, 0x000081EF }, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */ - { 32282, 0x000080BF }, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ - { 32316, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC */ - { 32340, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC_ARB */ - { 32368, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE */ - { 32392, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE_ARB */ - { 32420, 0x0000819B }, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */ - { 32453, 0x0000819A }, /* GL_TEXTURE_COMPARE_SGIX */ - { 32477, 0x00001003 }, /* GL_TEXTURE_COMPONENTS */ - { 32499, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED */ - { 32521, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED_ARB */ - { 32547, 0x000086A3 }, /* GL_TEXTURE_COMPRESSED_FORMATS_ARB */ - { 32581, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */ - { 32614, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE_ARB */ - { 32651, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT */ - { 32679, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT_ARB */ - { 32711, 0x00008078 }, /* GL_TEXTURE_COORD_ARRAY */ - { 32734, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */ - { 32772, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB */ - { 32814, 0x00008092 }, /* GL_TEXTURE_COORD_ARRAY_POINTER */ - { 32845, 0x00008088 }, /* GL_TEXTURE_COORD_ARRAY_SIZE */ - { 32873, 0x0000808A }, /* GL_TEXTURE_COORD_ARRAY_STRIDE */ - { 32903, 0x00008089 }, /* GL_TEXTURE_COORD_ARRAY_TYPE */ - { 32931, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP */ - { 32951, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP_ARB */ - { 32975, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */ - { 33006, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB */ - { 33041, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */ - { 33072, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB */ - { 33107, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */ - { 33138, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB */ - { 33173, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */ - { 33204, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB */ - { 33239, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */ - { 33270, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB */ - { 33305, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */ - { 33336, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB */ - { 33371, 0x00008071 }, /* GL_TEXTURE_DEPTH */ - { 33388, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE */ - { 33410, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE_ARB */ - { 33436, 0x00002300 }, /* GL_TEXTURE_ENV */ - { 33451, 0x00002201 }, /* GL_TEXTURE_ENV_COLOR */ - { 33472, 0x00002200 }, /* GL_TEXTURE_ENV_MODE */ - { 33492, 0x00008500 }, /* GL_TEXTURE_FILTER_CONTROL */ - { 33518, 0x00002500 }, /* GL_TEXTURE_GEN_MODE */ - { 33538, 0x00000C63 }, /* GL_TEXTURE_GEN_Q */ - { 33555, 0x00000C62 }, /* GL_TEXTURE_GEN_R */ - { 33572, 0x00000C60 }, /* GL_TEXTURE_GEN_S */ - { 33589, 0x00000C61 }, /* GL_TEXTURE_GEN_T */ - { 33606, 0x0000819D }, /* GL_TEXTURE_GEQUAL_R_SGIX */ - { 33631, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE */ - { 33653, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE_EXT */ - { 33679, 0x00001001 }, /* GL_TEXTURE_HEIGHT */ - { 33697, 0x000080ED }, /* GL_TEXTURE_INDEX_SIZE_EXT */ - { 33723, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE */ - { 33749, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE_EXT */ - { 33779, 0x00001003 }, /* GL_TEXTURE_INTERNAL_FORMAT */ - { 33806, 0x0000819C }, /* GL_TEXTURE_LEQUAL_R_SGIX */ - { 33831, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS */ - { 33851, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS_EXT */ - { 33875, 0x00008190 }, /* GL_TEXTURE_LOD_BIAS_R_SGIX */ - { 33902, 0x0000818E }, /* GL_TEXTURE_LOD_BIAS_S_SGIX */ - { 33929, 0x0000818F }, /* GL_TEXTURE_LOD_BIAS_T_SGIX */ - { 33956, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE */ - { 33982, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE_EXT */ - { 34012, 0x00002800 }, /* GL_TEXTURE_MAG_FILTER */ - { 34034, 0x00000BA8 }, /* GL_TEXTURE_MATRIX */ - { 34052, 0x000084FE }, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */ - { 34082, 0x0000836B }, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */ - { 34110, 0x00008369 }, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */ - { 34138, 0x0000836A }, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */ - { 34166, 0x0000813D }, /* GL_TEXTURE_MAX_LEVEL */ - { 34187, 0x0000813B }, /* GL_TEXTURE_MAX_LOD */ - { 34206, 0x00002801 }, /* GL_TEXTURE_MIN_FILTER */ - { 34228, 0x0000813A }, /* GL_TEXTURE_MIN_LOD */ - { 34247, 0x00008066 }, /* GL_TEXTURE_PRIORITY */ - { 34267, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_ARB */ - { 34292, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_NV */ - { 34316, 0x0000805C }, /* GL_TEXTURE_RED_SIZE */ - { 34336, 0x0000805C }, /* GL_TEXTURE_RED_SIZE_EXT */ - { 34360, 0x00008067 }, /* GL_TEXTURE_RESIDENT */ - { 34380, 0x00000BA5 }, /* GL_TEXTURE_STACK_DEPTH */ - { 34403, 0x00008065 }, /* GL_TEXTURE_TOO_LARGE_EXT */ - { 34428, 0x0000888F }, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */ - { 34462, 0x00001000 }, /* GL_TEXTURE_WIDTH */ - { 34479, 0x00008072 }, /* GL_TEXTURE_WRAP_R */ - { 34497, 0x00002802 }, /* GL_TEXTURE_WRAP_S */ - { 34515, 0x00002803 }, /* GL_TEXTURE_WRAP_T */ - { 34533, 0x000088BF }, /* GL_TIME_ELAPSED_EXT */ - { 34553, 0x00008648 }, /* GL_TRACK_MATRIX_NV */ - { 34572, 0x00008649 }, /* GL_TRACK_MATRIX_TRANSFORM_NV */ - { 34601, 0x00001000 }, /* GL_TRANSFORM_BIT */ - { 34618, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX */ - { 34644, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX_ARB */ - { 34674, 0x000088B7 }, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */ - { 34706, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX */ - { 34736, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX_ARB */ - { 34770, 0x0000862C }, /* GL_TRANSPOSE_NV */ - { 34786, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX */ - { 34817, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX_ARB */ - { 34852, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX */ - { 34880, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX_ARB */ - { 34912, 0x00000004 }, /* GL_TRIANGLES */ - { 34925, 0x00000006 }, /* GL_TRIANGLE_FAN */ - { 34941, 0x00008615 }, /* GL_TRIANGLE_MESH_SUN */ - { 34962, 0x00000005 }, /* GL_TRIANGLE_STRIP */ - { 34980, 0x00000001 }, /* GL_TRUE */ - { 34988, 0x00000CF5 }, /* GL_UNPACK_ALIGNMENT */ - { 35008, 0x0000806E }, /* GL_UNPACK_IMAGE_HEIGHT */ - { 35031, 0x00000CF1 }, /* GL_UNPACK_LSB_FIRST */ - { 35051, 0x00000CF2 }, /* GL_UNPACK_ROW_LENGTH */ - { 35072, 0x0000806D }, /* GL_UNPACK_SKIP_IMAGES */ - { 35094, 0x00000CF4 }, /* GL_UNPACK_SKIP_PIXELS */ - { 35116, 0x00000CF3 }, /* GL_UNPACK_SKIP_ROWS */ - { 35136, 0x00000CF0 }, /* GL_UNPACK_SWAP_BYTES */ - { 35157, 0x00001401 }, /* GL_UNSIGNED_BYTE */ - { 35174, 0x00008362 }, /* GL_UNSIGNED_BYTE_2_3_3_REV */ - { 35201, 0x00008032 }, /* GL_UNSIGNED_BYTE_3_3_2 */ - { 35224, 0x00001405 }, /* GL_UNSIGNED_INT */ - { 35240, 0x00008036 }, /* GL_UNSIGNED_INT_10_10_10_2 */ - { 35267, 0x000084FA }, /* GL_UNSIGNED_INT_24_8_NV */ - { 35291, 0x00008368 }, /* GL_UNSIGNED_INT_2_10_10_10_REV */ - { 35322, 0x00008035 }, /* GL_UNSIGNED_INT_8_8_8_8 */ - { 35346, 0x00008367 }, /* GL_UNSIGNED_INT_8_8_8_8_REV */ - { 35374, 0x00001403 }, /* GL_UNSIGNED_SHORT */ - { 35392, 0x00008366 }, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */ - { 35422, 0x00008033 }, /* GL_UNSIGNED_SHORT_4_4_4_4 */ - { 35448, 0x00008365 }, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */ - { 35478, 0x00008034 }, /* GL_UNSIGNED_SHORT_5_5_5_1 */ - { 35504, 0x00008363 }, /* GL_UNSIGNED_SHORT_5_6_5 */ - { 35528, 0x00008364 }, /* GL_UNSIGNED_SHORT_5_6_5_REV */ - { 35556, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_APPLE */ - { 35584, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_MESA */ - { 35611, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */ - { 35643, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_MESA */ - { 35674, 0x00008CA2 }, /* GL_UPPER_LEFT */ - { 35688, 0x00002A20 }, /* GL_V2F */ - { 35695, 0x00002A21 }, /* GL_V3F */ - { 35702, 0x00008B83 }, /* GL_VALIDATE_STATUS */ - { 35721, 0x00001F00 }, /* GL_VENDOR */ - { 35731, 0x00001F02 }, /* GL_VERSION */ - { 35742, 0x00008074 }, /* GL_VERTEX_ARRAY */ - { 35758, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING_APPLE */ - { 35788, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING */ - { 35819, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING_ARB */ - { 35854, 0x0000808E }, /* GL_VERTEX_ARRAY_POINTER */ - { 35878, 0x0000807A }, /* GL_VERTEX_ARRAY_SIZE */ - { 35899, 0x0000807C }, /* GL_VERTEX_ARRAY_STRIDE */ - { 35922, 0x0000807B }, /* GL_VERTEX_ARRAY_TYPE */ - { 35943, 0x00008650 }, /* GL_VERTEX_ATTRIB_ARRAY0_NV */ - { 35970, 0x0000865A }, /* GL_VERTEX_ATTRIB_ARRAY10_NV */ - { 35998, 0x0000865B }, /* GL_VERTEX_ATTRIB_ARRAY11_NV */ - { 36026, 0x0000865C }, /* GL_VERTEX_ATTRIB_ARRAY12_NV */ - { 36054, 0x0000865D }, /* GL_VERTEX_ATTRIB_ARRAY13_NV */ - { 36082, 0x0000865E }, /* GL_VERTEX_ATTRIB_ARRAY14_NV */ - { 36110, 0x0000865F }, /* GL_VERTEX_ATTRIB_ARRAY15_NV */ - { 36138, 0x00008651 }, /* GL_VERTEX_ATTRIB_ARRAY1_NV */ - { 36165, 0x00008652 }, /* GL_VERTEX_ATTRIB_ARRAY2_NV */ - { 36192, 0x00008653 }, /* GL_VERTEX_ATTRIB_ARRAY3_NV */ - { 36219, 0x00008654 }, /* GL_VERTEX_ATTRIB_ARRAY4_NV */ - { 36246, 0x00008655 }, /* GL_VERTEX_ATTRIB_ARRAY5_NV */ - { 36273, 0x00008656 }, /* GL_VERTEX_ATTRIB_ARRAY6_NV */ - { 36300, 0x00008657 }, /* GL_VERTEX_ATTRIB_ARRAY7_NV */ - { 36327, 0x00008658 }, /* GL_VERTEX_ATTRIB_ARRAY8_NV */ - { 36354, 0x00008659 }, /* GL_VERTEX_ATTRIB_ARRAY9_NV */ - { 36381, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */ - { 36419, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB */ - { 36461, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */ - { 36492, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB */ - { 36527, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */ - { 36561, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB */ - { 36599, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */ - { 36630, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB */ - { 36665, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */ - { 36693, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB */ - { 36725, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */ - { 36755, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB */ - { 36789, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */ - { 36817, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB */ - { 36849, 0x000086A7 }, /* GL_VERTEX_BLEND_ARB */ - { 36869, 0x00008620 }, /* GL_VERTEX_PROGRAM_ARB */ - { 36891, 0x0000864A }, /* GL_VERTEX_PROGRAM_BINDING_NV */ - { 36920, 0x00008620 }, /* GL_VERTEX_PROGRAM_NV */ - { 36941, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE */ - { 36970, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_ARB */ - { 37003, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_NV */ - { 37035, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE */ - { 37062, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_ARB */ - { 37093, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_NV */ - { 37123, 0x00008B31 }, /* GL_VERTEX_SHADER */ - { 37140, 0x00008B31 }, /* GL_VERTEX_SHADER_ARB */ - { 37161, 0x00008621 }, /* GL_VERTEX_STATE_PROGRAM_NV */ - { 37188, 0x00000BA2 }, /* GL_VIEWPORT */ - { 37200, 0x00000800 }, /* GL_VIEWPORT_BIT */ - { 37216, 0x000086AD }, /* GL_WEIGHT_ARRAY_ARB */ - { 37236, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */ - { 37267, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB */ - { 37302, 0x000086AC }, /* GL_WEIGHT_ARRAY_POINTER_ARB */ - { 37330, 0x000086AB }, /* GL_WEIGHT_ARRAY_SIZE_ARB */ - { 37355, 0x000086AA }, /* GL_WEIGHT_ARRAY_STRIDE_ARB */ - { 37382, 0x000086A9 }, /* GL_WEIGHT_ARRAY_TYPE_ARB */ - { 37407, 0x000086A6 }, /* GL_WEIGHT_SUM_UNITY_ARB */ - { 37431, 0x000081D4 }, /* GL_WRAP_BORDER_SUN */ - { 37450, 0x000088B9 }, /* GL_WRITE_ONLY */ - { 37464, 0x000088B9 }, /* GL_WRITE_ONLY_ARB */ - { 37482, 0x00001506 }, /* GL_XOR */ - { 37489, 0x000085B9 }, /* GL_YCBCR_422_APPLE */ - { 37508, 0x00008757 }, /* GL_YCBCR_MESA */ - { 37522, 0x00000000 }, /* GL_ZERO */ - { 37530, 0x00000D16 }, /* GL_ZOOM_X */ - { 37540, 0x00000D17 }, /* GL_ZOOM_Y */ + { 29733, 0x00008801 }, /* GL_STENCIL_BACK_FAIL_ATI */ + { 29758, 0x00008800 }, /* GL_STENCIL_BACK_FUNC */ + { 29779, 0x00008800 }, /* GL_STENCIL_BACK_FUNC_ATI */ + { 29804, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */ + { 29836, 0x00008802 }, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL_ATI */ + { 29872, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */ + { 29904, 0x00008803 }, /* GL_STENCIL_BACK_PASS_DEPTH_PASS_ATI */ + { 29940, 0x00008CA3 }, /* GL_STENCIL_BACK_REF */ + { 29960, 0x00008CA4 }, /* GL_STENCIL_BACK_VALUE_MASK */ + { 29987, 0x00008CA5 }, /* GL_STENCIL_BACK_WRITEMASK */ + { 30013, 0x00000D57 }, /* GL_STENCIL_BITS */ + { 30029, 0x00000400 }, /* GL_STENCIL_BUFFER_BIT */ + { 30051, 0x00000B91 }, /* GL_STENCIL_CLEAR_VALUE */ + { 30074, 0x00000B94 }, /* GL_STENCIL_FAIL */ + { 30090, 0x00000B92 }, /* GL_STENCIL_FUNC */ + { 30106, 0x00001901 }, /* GL_STENCIL_INDEX */ + { 30123, 0x00008D49 }, /* GL_STENCIL_INDEX16_EXT */ + { 30146, 0x00008D46 }, /* GL_STENCIL_INDEX1_EXT */ + { 30168, 0x00008D47 }, /* GL_STENCIL_INDEX4_EXT */ + { 30190, 0x00008D48 }, /* GL_STENCIL_INDEX8_EXT */ + { 30212, 0x00008D45 }, /* GL_STENCIL_INDEX_EXT */ + { 30233, 0x00000B95 }, /* GL_STENCIL_PASS_DEPTH_FAIL */ + { 30260, 0x00000B96 }, /* GL_STENCIL_PASS_DEPTH_PASS */ + { 30287, 0x00000B97 }, /* GL_STENCIL_REF */ + { 30302, 0x00000B90 }, /* GL_STENCIL_TEST */ + { 30318, 0x00008910 }, /* GL_STENCIL_TEST_TWO_SIDE_EXT */ + { 30347, 0x00000B93 }, /* GL_STENCIL_VALUE_MASK */ + { 30369, 0x00000B98 }, /* GL_STENCIL_WRITEMASK */ + { 30390, 0x00000C33 }, /* GL_STEREO */ + { 30400, 0x000088E2 }, /* GL_STREAM_COPY */ + { 30415, 0x000088E2 }, /* GL_STREAM_COPY_ARB */ + { 30434, 0x000088E0 }, /* GL_STREAM_DRAW */ + { 30449, 0x000088E0 }, /* GL_STREAM_DRAW_ARB */ + { 30468, 0x000088E1 }, /* GL_STREAM_READ */ + { 30483, 0x000088E1 }, /* GL_STREAM_READ_ARB */ + { 30502, 0x00000D50 }, /* GL_SUBPIXEL_BITS */ + { 30519, 0x000084E7 }, /* GL_SUBTRACT */ + { 30531, 0x000084E7 }, /* GL_SUBTRACT_ARB */ + { 30547, 0x00002001 }, /* GL_T */ + { 30552, 0x00002A2A }, /* GL_T2F_C3F_V3F */ + { 30567, 0x00002A2C }, /* GL_T2F_C4F_N3F_V3F */ + { 30586, 0x00002A29 }, /* GL_T2F_C4UB_V3F */ + { 30602, 0x00002A2B }, /* GL_T2F_N3F_V3F */ + { 30617, 0x00002A27 }, /* GL_T2F_V3F */ + { 30628, 0x00002A2D }, /* GL_T4F_C4F_N3F_V4F */ + { 30647, 0x00002A28 }, /* GL_T4F_V4F */ + { 30658, 0x00008031 }, /* GL_TABLE_TOO_LARGE_EXT */ + { 30681, 0x00001702 }, /* GL_TEXTURE */ + { 30692, 0x000084C0 }, /* GL_TEXTURE0 */ + { 30704, 0x000084C0 }, /* GL_TEXTURE0_ARB */ + { 30720, 0x000084C1 }, /* GL_TEXTURE1 */ + { 30732, 0x000084CA }, /* GL_TEXTURE10 */ + { 30745, 0x000084CA }, /* GL_TEXTURE10_ARB */ + { 30762, 0x000084CB }, /* GL_TEXTURE11 */ + { 30775, 0x000084CB }, /* GL_TEXTURE11_ARB */ + { 30792, 0x000084CC }, /* GL_TEXTURE12 */ + { 30805, 0x000084CC }, /* GL_TEXTURE12_ARB */ + { 30822, 0x000084CD }, /* GL_TEXTURE13 */ + { 30835, 0x000084CD }, /* GL_TEXTURE13_ARB */ + { 30852, 0x000084CE }, /* GL_TEXTURE14 */ + { 30865, 0x000084CE }, /* GL_TEXTURE14_ARB */ + { 30882, 0x000084CF }, /* GL_TEXTURE15 */ + { 30895, 0x000084CF }, /* GL_TEXTURE15_ARB */ + { 30912, 0x000084D0 }, /* GL_TEXTURE16 */ + { 30925, 0x000084D0 }, /* GL_TEXTURE16_ARB */ + { 30942, 0x000084D1 }, /* GL_TEXTURE17 */ + { 30955, 0x000084D1 }, /* GL_TEXTURE17_ARB */ + { 30972, 0x000084D2 }, /* GL_TEXTURE18 */ + { 30985, 0x000084D2 }, /* GL_TEXTURE18_ARB */ + { 31002, 0x000084D3 }, /* GL_TEXTURE19 */ + { 31015, 0x000084D3 }, /* GL_TEXTURE19_ARB */ + { 31032, 0x000084C1 }, /* GL_TEXTURE1_ARB */ + { 31048, 0x000084C2 }, /* GL_TEXTURE2 */ + { 31060, 0x000084D4 }, /* GL_TEXTURE20 */ + { 31073, 0x000084D4 }, /* GL_TEXTURE20_ARB */ + { 31090, 0x000084D5 }, /* GL_TEXTURE21 */ + { 31103, 0x000084D5 }, /* GL_TEXTURE21_ARB */ + { 31120, 0x000084D6 }, /* GL_TEXTURE22 */ + { 31133, 0x000084D6 }, /* GL_TEXTURE22_ARB */ + { 31150, 0x000084D7 }, /* GL_TEXTURE23 */ + { 31163, 0x000084D7 }, /* GL_TEXTURE23_ARB */ + { 31180, 0x000084D8 }, /* GL_TEXTURE24 */ + { 31193, 0x000084D8 }, /* GL_TEXTURE24_ARB */ + { 31210, 0x000084D9 }, /* GL_TEXTURE25 */ + { 31223, 0x000084D9 }, /* GL_TEXTURE25_ARB */ + { 31240, 0x000084DA }, /* GL_TEXTURE26 */ + { 31253, 0x000084DA }, /* GL_TEXTURE26_ARB */ + { 31270, 0x000084DB }, /* GL_TEXTURE27 */ + { 31283, 0x000084DB }, /* GL_TEXTURE27_ARB */ + { 31300, 0x000084DC }, /* GL_TEXTURE28 */ + { 31313, 0x000084DC }, /* GL_TEXTURE28_ARB */ + { 31330, 0x000084DD }, /* GL_TEXTURE29 */ + { 31343, 0x000084DD }, /* GL_TEXTURE29_ARB */ + { 31360, 0x000084C2 }, /* GL_TEXTURE2_ARB */ + { 31376, 0x000084C3 }, /* GL_TEXTURE3 */ + { 31388, 0x000084DE }, /* GL_TEXTURE30 */ + { 31401, 0x000084DE }, /* GL_TEXTURE30_ARB */ + { 31418, 0x000084DF }, /* GL_TEXTURE31 */ + { 31431, 0x000084DF }, /* GL_TEXTURE31_ARB */ + { 31448, 0x000084C3 }, /* GL_TEXTURE3_ARB */ + { 31464, 0x000084C4 }, /* GL_TEXTURE4 */ + { 31476, 0x000084C4 }, /* GL_TEXTURE4_ARB */ + { 31492, 0x000084C5 }, /* GL_TEXTURE5 */ + { 31504, 0x000084C5 }, /* GL_TEXTURE5_ARB */ + { 31520, 0x000084C6 }, /* GL_TEXTURE6 */ + { 31532, 0x000084C6 }, /* GL_TEXTURE6_ARB */ + { 31548, 0x000084C7 }, /* GL_TEXTURE7 */ + { 31560, 0x000084C7 }, /* GL_TEXTURE7_ARB */ + { 31576, 0x000084C8 }, /* GL_TEXTURE8 */ + { 31588, 0x000084C8 }, /* GL_TEXTURE8_ARB */ + { 31604, 0x000084C9 }, /* GL_TEXTURE9 */ + { 31616, 0x000084C9 }, /* GL_TEXTURE9_ARB */ + { 31632, 0x00000DE0 }, /* GL_TEXTURE_1D */ + { 31646, 0x00008C18 }, /* GL_TEXTURE_1D_ARRAY_EXT */ + { 31670, 0x00000DE1 }, /* GL_TEXTURE_2D */ + { 31684, 0x00008C1A }, /* GL_TEXTURE_2D_ARRAY_EXT */ + { 31708, 0x0000806F }, /* GL_TEXTURE_3D */ + { 31722, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE */ + { 31744, 0x0000805F }, /* GL_TEXTURE_ALPHA_SIZE_EXT */ + { 31770, 0x0000813C }, /* GL_TEXTURE_BASE_LEVEL */ + { 31792, 0x00008068 }, /* GL_TEXTURE_BINDING_1D */ + { 31814, 0x00008C1C }, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */ + { 31846, 0x00008069 }, /* GL_TEXTURE_BINDING_2D */ + { 31868, 0x00008C1D }, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */ + { 31900, 0x0000806A }, /* GL_TEXTURE_BINDING_3D */ + { 31922, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP */ + { 31950, 0x00008514 }, /* GL_TEXTURE_BINDING_CUBE_MAP_ARB */ + { 31982, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */ + { 32015, 0x000084F6 }, /* GL_TEXTURE_BINDING_RECTANGLE_NV */ + { 32047, 0x00040000 }, /* GL_TEXTURE_BIT */ + { 32062, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE */ + { 32083, 0x0000805E }, /* GL_TEXTURE_BLUE_SIZE_EXT */ + { 32108, 0x00001005 }, /* GL_TEXTURE_BORDER */ + { 32126, 0x00001004 }, /* GL_TEXTURE_BORDER_COLOR */ + { 32150, 0x00008171 }, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */ + { 32181, 0x00008176 }, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */ + { 32211, 0x00008172 }, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */ + { 32241, 0x00008175 }, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */ + { 32276, 0x00008173 }, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */ + { 32307, 0x00008174 }, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */ + { 32345, 0x000080BC }, /* GL_TEXTURE_COLOR_TABLE_SGI */ + { 32372, 0x000081EF }, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */ + { 32404, 0x000080BF }, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ + { 32438, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC */ + { 32462, 0x0000884D }, /* GL_TEXTURE_COMPARE_FUNC_ARB */ + { 32490, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE */ + { 32514, 0x0000884C }, /* GL_TEXTURE_COMPARE_MODE_ARB */ + { 32542, 0x0000819B }, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */ + { 32575, 0x0000819A }, /* GL_TEXTURE_COMPARE_SGIX */ + { 32599, 0x00001003 }, /* GL_TEXTURE_COMPONENTS */ + { 32621, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED */ + { 32643, 0x000086A1 }, /* GL_TEXTURE_COMPRESSED_ARB */ + { 32669, 0x000086A3 }, /* GL_TEXTURE_COMPRESSED_FORMATS_ARB */ + { 32703, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */ + { 32736, 0x000086A0 }, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE_ARB */ + { 32773, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT */ + { 32801, 0x000084EF }, /* GL_TEXTURE_COMPRESSION_HINT_ARB */ + { 32833, 0x00008078 }, /* GL_TEXTURE_COORD_ARRAY */ + { 32856, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */ + { 32894, 0x0000889A }, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING_ARB */ + { 32936, 0x00008092 }, /* GL_TEXTURE_COORD_ARRAY_POINTER */ + { 32967, 0x00008088 }, /* GL_TEXTURE_COORD_ARRAY_SIZE */ + { 32995, 0x0000808A }, /* GL_TEXTURE_COORD_ARRAY_STRIDE */ + { 33025, 0x00008089 }, /* GL_TEXTURE_COORD_ARRAY_TYPE */ + { 33053, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP */ + { 33073, 0x00008513 }, /* GL_TEXTURE_CUBE_MAP_ARB */ + { 33097, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */ + { 33128, 0x00008516 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB */ + { 33163, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */ + { 33194, 0x00008518 }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y_ARB */ + { 33229, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */ + { 33260, 0x0000851A }, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z_ARB */ + { 33295, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */ + { 33326, 0x00008515 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB */ + { 33361, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */ + { 33392, 0x00008517 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB */ + { 33427, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */ + { 33458, 0x00008519 }, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z_ARB */ + { 33493, 0x00008071 }, /* GL_TEXTURE_DEPTH */ + { 33510, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE */ + { 33532, 0x0000884A }, /* GL_TEXTURE_DEPTH_SIZE_ARB */ + { 33558, 0x00002300 }, /* GL_TEXTURE_ENV */ + { 33573, 0x00002201 }, /* GL_TEXTURE_ENV_COLOR */ + { 33594, 0x00002200 }, /* GL_TEXTURE_ENV_MODE */ + { 33614, 0x00008500 }, /* GL_TEXTURE_FILTER_CONTROL */ + { 33640, 0x00002500 }, /* GL_TEXTURE_GEN_MODE */ + { 33660, 0x00000C63 }, /* GL_TEXTURE_GEN_Q */ + { 33677, 0x00000C62 }, /* GL_TEXTURE_GEN_R */ + { 33694, 0x00000C60 }, /* GL_TEXTURE_GEN_S */ + { 33711, 0x00000C61 }, /* GL_TEXTURE_GEN_T */ + { 33728, 0x0000819D }, /* GL_TEXTURE_GEQUAL_R_SGIX */ + { 33753, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE */ + { 33775, 0x0000805D }, /* GL_TEXTURE_GREEN_SIZE_EXT */ + { 33801, 0x00001001 }, /* GL_TEXTURE_HEIGHT */ + { 33819, 0x000080ED }, /* GL_TEXTURE_INDEX_SIZE_EXT */ + { 33845, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE */ + { 33871, 0x00008061 }, /* GL_TEXTURE_INTENSITY_SIZE_EXT */ + { 33901, 0x00001003 }, /* GL_TEXTURE_INTERNAL_FORMAT */ + { 33928, 0x0000819C }, /* GL_TEXTURE_LEQUAL_R_SGIX */ + { 33953, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS */ + { 33973, 0x00008501 }, /* GL_TEXTURE_LOD_BIAS_EXT */ + { 33997, 0x00008190 }, /* GL_TEXTURE_LOD_BIAS_R_SGIX */ + { 34024, 0x0000818E }, /* GL_TEXTURE_LOD_BIAS_S_SGIX */ + { 34051, 0x0000818F }, /* GL_TEXTURE_LOD_BIAS_T_SGIX */ + { 34078, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE */ + { 34104, 0x00008060 }, /* GL_TEXTURE_LUMINANCE_SIZE_EXT */ + { 34134, 0x00002800 }, /* GL_TEXTURE_MAG_FILTER */ + { 34156, 0x00000BA8 }, /* GL_TEXTURE_MATRIX */ + { 34174, 0x000084FE }, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */ + { 34204, 0x0000836B }, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */ + { 34232, 0x00008369 }, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */ + { 34260, 0x0000836A }, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */ + { 34288, 0x0000813D }, /* GL_TEXTURE_MAX_LEVEL */ + { 34309, 0x0000813B }, /* GL_TEXTURE_MAX_LOD */ + { 34328, 0x00002801 }, /* GL_TEXTURE_MIN_FILTER */ + { 34350, 0x0000813A }, /* GL_TEXTURE_MIN_LOD */ + { 34369, 0x00008066 }, /* GL_TEXTURE_PRIORITY */ + { 34389, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_ARB */ + { 34414, 0x000084F5 }, /* GL_TEXTURE_RECTANGLE_NV */ + { 34438, 0x0000805C }, /* GL_TEXTURE_RED_SIZE */ + { 34458, 0x0000805C }, /* GL_TEXTURE_RED_SIZE_EXT */ + { 34482, 0x00008067 }, /* GL_TEXTURE_RESIDENT */ + { 34502, 0x00000BA5 }, /* GL_TEXTURE_STACK_DEPTH */ + { 34525, 0x00008065 }, /* GL_TEXTURE_TOO_LARGE_EXT */ + { 34550, 0x0000888F }, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */ + { 34584, 0x00001000 }, /* GL_TEXTURE_WIDTH */ + { 34601, 0x00008072 }, /* GL_TEXTURE_WRAP_R */ + { 34619, 0x00002802 }, /* GL_TEXTURE_WRAP_S */ + { 34637, 0x00002803 }, /* GL_TEXTURE_WRAP_T */ + { 34655, 0x000088BF }, /* GL_TIME_ELAPSED_EXT */ + { 34675, 0x00008648 }, /* GL_TRACK_MATRIX_NV */ + { 34694, 0x00008649 }, /* GL_TRACK_MATRIX_TRANSFORM_NV */ + { 34723, 0x00001000 }, /* GL_TRANSFORM_BIT */ + { 34740, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX */ + { 34766, 0x000084E6 }, /* GL_TRANSPOSE_COLOR_MATRIX_ARB */ + { 34796, 0x000088B7 }, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */ + { 34828, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX */ + { 34858, 0x000084E3 }, /* GL_TRANSPOSE_MODELVIEW_MATRIX_ARB */ + { 34892, 0x0000862C }, /* GL_TRANSPOSE_NV */ + { 34908, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX */ + { 34939, 0x000084E4 }, /* GL_TRANSPOSE_PROJECTION_MATRIX_ARB */ + { 34974, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX */ + { 35002, 0x000084E5 }, /* GL_TRANSPOSE_TEXTURE_MATRIX_ARB */ + { 35034, 0x00000004 }, /* GL_TRIANGLES */ + { 35047, 0x00000006 }, /* GL_TRIANGLE_FAN */ + { 35063, 0x00008615 }, /* GL_TRIANGLE_MESH_SUN */ + { 35084, 0x00000005 }, /* GL_TRIANGLE_STRIP */ + { 35102, 0x00000001 }, /* GL_TRUE */ + { 35110, 0x00000CF5 }, /* GL_UNPACK_ALIGNMENT */ + { 35130, 0x0000806E }, /* GL_UNPACK_IMAGE_HEIGHT */ + { 35153, 0x00000CF1 }, /* GL_UNPACK_LSB_FIRST */ + { 35173, 0x00000CF2 }, /* GL_UNPACK_ROW_LENGTH */ + { 35194, 0x0000806D }, /* GL_UNPACK_SKIP_IMAGES */ + { 35216, 0x00000CF4 }, /* GL_UNPACK_SKIP_PIXELS */ + { 35238, 0x00000CF3 }, /* GL_UNPACK_SKIP_ROWS */ + { 35258, 0x00000CF0 }, /* GL_UNPACK_SWAP_BYTES */ + { 35279, 0x00001401 }, /* GL_UNSIGNED_BYTE */ + { 35296, 0x00008362 }, /* GL_UNSIGNED_BYTE_2_3_3_REV */ + { 35323, 0x00008032 }, /* GL_UNSIGNED_BYTE_3_3_2 */ + { 35346, 0x00001405 }, /* GL_UNSIGNED_INT */ + { 35362, 0x00008036 }, /* GL_UNSIGNED_INT_10_10_10_2 */ + { 35389, 0x000084FA }, /* GL_UNSIGNED_INT_24_8_NV */ + { 35413, 0x00008368 }, /* GL_UNSIGNED_INT_2_10_10_10_REV */ + { 35444, 0x00008035 }, /* GL_UNSIGNED_INT_8_8_8_8 */ + { 35468, 0x00008367 }, /* GL_UNSIGNED_INT_8_8_8_8_REV */ + { 35496, 0x00001403 }, /* GL_UNSIGNED_SHORT */ + { 35514, 0x00008366 }, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */ + { 35544, 0x00008033 }, /* GL_UNSIGNED_SHORT_4_4_4_4 */ + { 35570, 0x00008365 }, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */ + { 35600, 0x00008034 }, /* GL_UNSIGNED_SHORT_5_5_5_1 */ + { 35626, 0x00008363 }, /* GL_UNSIGNED_SHORT_5_6_5 */ + { 35650, 0x00008364 }, /* GL_UNSIGNED_SHORT_5_6_5_REV */ + { 35678, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_APPLE */ + { 35706, 0x000085BA }, /* GL_UNSIGNED_SHORT_8_8_MESA */ + { 35733, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */ + { 35765, 0x000085BB }, /* GL_UNSIGNED_SHORT_8_8_REV_MESA */ + { 35796, 0x00008CA2 }, /* GL_UPPER_LEFT */ + { 35810, 0x00002A20 }, /* GL_V2F */ + { 35817, 0x00002A21 }, /* GL_V3F */ + { 35824, 0x00008B83 }, /* GL_VALIDATE_STATUS */ + { 35843, 0x00001F00 }, /* GL_VENDOR */ + { 35853, 0x00001F02 }, /* GL_VERSION */ + { 35864, 0x00008074 }, /* GL_VERTEX_ARRAY */ + { 35880, 0x000085B5 }, /* GL_VERTEX_ARRAY_BINDING_APPLE */ + { 35910, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING */ + { 35941, 0x00008896 }, /* GL_VERTEX_ARRAY_BUFFER_BINDING_ARB */ + { 35976, 0x0000808E }, /* GL_VERTEX_ARRAY_POINTER */ + { 36000, 0x0000807A }, /* GL_VERTEX_ARRAY_SIZE */ + { 36021, 0x0000807C }, /* GL_VERTEX_ARRAY_STRIDE */ + { 36044, 0x0000807B }, /* GL_VERTEX_ARRAY_TYPE */ + { 36065, 0x00008650 }, /* GL_VERTEX_ATTRIB_ARRAY0_NV */ + { 36092, 0x0000865A }, /* GL_VERTEX_ATTRIB_ARRAY10_NV */ + { 36120, 0x0000865B }, /* GL_VERTEX_ATTRIB_ARRAY11_NV */ + { 36148, 0x0000865C }, /* GL_VERTEX_ATTRIB_ARRAY12_NV */ + { 36176, 0x0000865D }, /* GL_VERTEX_ATTRIB_ARRAY13_NV */ + { 36204, 0x0000865E }, /* GL_VERTEX_ATTRIB_ARRAY14_NV */ + { 36232, 0x0000865F }, /* GL_VERTEX_ATTRIB_ARRAY15_NV */ + { 36260, 0x00008651 }, /* GL_VERTEX_ATTRIB_ARRAY1_NV */ + { 36287, 0x00008652 }, /* GL_VERTEX_ATTRIB_ARRAY2_NV */ + { 36314, 0x00008653 }, /* GL_VERTEX_ATTRIB_ARRAY3_NV */ + { 36341, 0x00008654 }, /* GL_VERTEX_ATTRIB_ARRAY4_NV */ + { 36368, 0x00008655 }, /* GL_VERTEX_ATTRIB_ARRAY5_NV */ + { 36395, 0x00008656 }, /* GL_VERTEX_ATTRIB_ARRAY6_NV */ + { 36422, 0x00008657 }, /* GL_VERTEX_ATTRIB_ARRAY7_NV */ + { 36449, 0x00008658 }, /* GL_VERTEX_ATTRIB_ARRAY8_NV */ + { 36476, 0x00008659 }, /* GL_VERTEX_ATTRIB_ARRAY9_NV */ + { 36503, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */ + { 36541, 0x0000889F }, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING_ARB */ + { 36583, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */ + { 36614, 0x00008622 }, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED_ARB */ + { 36649, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */ + { 36683, 0x0000886A }, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED_ARB */ + { 36721, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */ + { 36752, 0x00008645 }, /* GL_VERTEX_ATTRIB_ARRAY_POINTER_ARB */ + { 36787, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */ + { 36815, 0x00008623 }, /* GL_VERTEX_ATTRIB_ARRAY_SIZE_ARB */ + { 36847, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */ + { 36877, 0x00008624 }, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE_ARB */ + { 36911, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */ + { 36939, 0x00008625 }, /* GL_VERTEX_ATTRIB_ARRAY_TYPE_ARB */ + { 36971, 0x000086A7 }, /* GL_VERTEX_BLEND_ARB */ + { 36991, 0x00008620 }, /* GL_VERTEX_PROGRAM_ARB */ + { 37013, 0x0000864A }, /* GL_VERTEX_PROGRAM_BINDING_NV */ + { 37042, 0x00008620 }, /* GL_VERTEX_PROGRAM_NV */ + { 37063, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE */ + { 37092, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_ARB */ + { 37125, 0x00008642 }, /* GL_VERTEX_PROGRAM_POINT_SIZE_NV */ + { 37157, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE */ + { 37184, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_ARB */ + { 37215, 0x00008643 }, /* GL_VERTEX_PROGRAM_TWO_SIDE_NV */ + { 37245, 0x00008B31 }, /* GL_VERTEX_SHADER */ + { 37262, 0x00008B31 }, /* GL_VERTEX_SHADER_ARB */ + { 37283, 0x00008621 }, /* GL_VERTEX_STATE_PROGRAM_NV */ + { 37310, 0x00000BA2 }, /* GL_VIEWPORT */ + { 37322, 0x00000800 }, /* GL_VIEWPORT_BIT */ + { 37338, 0x000086AD }, /* GL_WEIGHT_ARRAY_ARB */ + { 37358, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */ + { 37389, 0x0000889E }, /* GL_WEIGHT_ARRAY_BUFFER_BINDING_ARB */ + { 37424, 0x000086AC }, /* GL_WEIGHT_ARRAY_POINTER_ARB */ + { 37452, 0x000086AB }, /* GL_WEIGHT_ARRAY_SIZE_ARB */ + { 37477, 0x000086AA }, /* GL_WEIGHT_ARRAY_STRIDE_ARB */ + { 37504, 0x000086A9 }, /* GL_WEIGHT_ARRAY_TYPE_ARB */ + { 37529, 0x000086A6 }, /* GL_WEIGHT_SUM_UNITY_ARB */ + { 37553, 0x000081D4 }, /* GL_WRAP_BORDER_SUN */ + { 37572, 0x000088B9 }, /* GL_WRITE_ONLY */ + { 37586, 0x000088B9 }, /* GL_WRITE_ONLY_ARB */ + { 37604, 0x00001506 }, /* GL_XOR */ + { 37611, 0x000085B9 }, /* GL_YCBCR_422_APPLE */ + { 37630, 0x00008757 }, /* GL_YCBCR_MESA */ + { 37644, 0x00000000 }, /* GL_ZERO */ + { 37652, 0x00000D16 }, /* GL_ZOOM_X */ + { 37662, 0x00000D17 }, /* GL_ZOOM_Y */ }; static const unsigned reduced_enums[1284] = @@ -3539,9 +3547,9 @@ static const unsigned reduced_enums[1284] = 645, /* GL_LINES */ 647, /* GL_LINE_LOOP */ 654, /* GL_LINE_STRIP */ - 1637, /* GL_TRIANGLES */ - 1640, /* GL_TRIANGLE_STRIP */ - 1638, /* GL_TRIANGLE_FAN */ + 1641, /* GL_TRIANGLES */ + 1644, /* GL_TRIANGLE_STRIP */ + 1642, /* GL_TRIANGLE_FAN */ 1211, /* GL_QUADS */ 1213, /* GL_QUAD_STRIP */ 1099, /* GL_POLYGON */ @@ -3664,24 +3672,24 @@ static const unsigned reduced_enums[1284] = 321, /* GL_DEPTH_CLEAR_VALUE */ 332, /* GL_DEPTH_FUNC */ 12, /* GL_ACCUM_CLEAR_VALUE */ - 1418, /* GL_STENCIL_TEST */ - 1406, /* GL_STENCIL_CLEAR_VALUE */ - 1408, /* GL_STENCIL_FUNC */ - 1420, /* GL_STENCIL_VALUE_MASK */ - 1407, /* GL_STENCIL_FAIL */ - 1415, /* GL_STENCIL_PASS_DEPTH_FAIL */ - 1416, /* GL_STENCIL_PASS_DEPTH_PASS */ - 1417, /* GL_STENCIL_REF */ - 1421, /* GL_STENCIL_WRITEMASK */ + 1422, /* GL_STENCIL_TEST */ + 1410, /* GL_STENCIL_CLEAR_VALUE */ + 1412, /* GL_STENCIL_FUNC */ + 1424, /* GL_STENCIL_VALUE_MASK */ + 1411, /* GL_STENCIL_FAIL */ + 1419, /* GL_STENCIL_PASS_DEPTH_FAIL */ + 1420, /* GL_STENCIL_PASS_DEPTH_PASS */ + 1421, /* GL_STENCIL_REF */ + 1425, /* GL_STENCIL_WRITEMASK */ 791, /* GL_MATRIX_MODE */ 958, /* GL_NORMALIZE */ - 1727, /* GL_VIEWPORT */ + 1731, /* GL_VIEWPORT */ 932, /* GL_MODELVIEW_STACK_DEPTH */ 1191, /* GL_PROJECTION_STACK_DEPTH */ - 1616, /* GL_TEXTURE_STACK_DEPTH */ + 1620, /* GL_TEXTURE_STACK_DEPTH */ 930, /* GL_MODELVIEW_MATRIX */ 1190, /* GL_PROJECTION_MATRIX */ - 1601, /* GL_TEXTURE_MATRIX */ + 1605, /* GL_TEXTURE_MATRIX */ 61, /* GL_ATTRIB_STACK_DEPTH */ 127, /* GL_CLIENT_ATTRIB_STACK_DEPTH */ 43, /* GL_ALPHA_TEST */ @@ -3706,17 +3714,17 @@ static const unsigned reduced_enums[1284] = 587, /* GL_INDEX_MODE */ 1287, /* GL_RGBA_MODE */ 353, /* GL_DOUBLEBUFFER */ - 1422, /* GL_STEREO */ + 1426, /* GL_STEREO */ 1246, /* GL_RENDER_MODE */ 1043, /* GL_PERSPECTIVE_CORRECTION_HINT */ 1092, /* GL_POINT_SMOOTH_HINT */ 650, /* GL_LINE_SMOOTH_HINT */ 1109, /* GL_POLYGON_SMOOTH_HINT */ 478, /* GL_FOG_HINT */ - 1582, /* GL_TEXTURE_GEN_S */ - 1583, /* GL_TEXTURE_GEN_T */ - 1581, /* GL_TEXTURE_GEN_R */ - 1580, /* GL_TEXTURE_GEN_Q */ + 1586, /* GL_TEXTURE_GEN_S */ + 1587, /* GL_TEXTURE_GEN_T */ + 1585, /* GL_TEXTURE_GEN_R */ + 1584, /* GL_TEXTURE_GEN_Q */ 1056, /* GL_PIXEL_MAP_I_TO_I */ 1062, /* GL_PIXEL_MAP_S_TO_S */ 1058, /* GL_PIXEL_MAP_I_TO_R */ @@ -3737,12 +3745,12 @@ static const unsigned reduced_enums[1284] = 1049, /* GL_PIXEL_MAP_G_TO_G_SIZE */ 1047, /* GL_PIXEL_MAP_B_TO_B_SIZE */ 1045, /* GL_PIXEL_MAP_A_TO_A_SIZE */ - 1649, /* GL_UNPACK_SWAP_BYTES */ - 1644, /* GL_UNPACK_LSB_FIRST */ - 1645, /* GL_UNPACK_ROW_LENGTH */ - 1648, /* GL_UNPACK_SKIP_ROWS */ - 1647, /* GL_UNPACK_SKIP_PIXELS */ - 1642, /* GL_UNPACK_ALIGNMENT */ + 1653, /* GL_UNPACK_SWAP_BYTES */ + 1648, /* GL_UNPACK_LSB_FIRST */ + 1649, /* GL_UNPACK_ROW_LENGTH */ + 1652, /* GL_UNPACK_SKIP_ROWS */ + 1651, /* GL_UNPACK_SKIP_PIXELS */ + 1646, /* GL_UNPACK_ALIGNMENT */ 1031, /* GL_PACK_SWAP_BYTES */ 1026, /* GL_PACK_LSB_FIRST */ 1027, /* GL_PACK_ROW_LENGTH */ @@ -3755,8 +3763,8 @@ static const unsigned reduced_enums[1284] = 588, /* GL_INDEX_OFFSET */ 1235, /* GL_RED_SCALE */ 1233, /* GL_RED_BIAS */ - 1744, /* GL_ZOOM_X */ - 1745, /* GL_ZOOM_Y */ + 1748, /* GL_ZOOM_X */ + 1749, /* GL_ZOOM_Y */ 551, /* GL_GREEN_SCALE */ 549, /* GL_GREEN_BIAS */ 92, /* GL_BLUE_SCALE */ @@ -3777,14 +3785,14 @@ static const unsigned reduced_enums[1284] = 866, /* GL_MAX_TEXTURE_STACK_DEPTH */ 880, /* GL_MAX_VIEWPORT_DIMS */ 797, /* GL_MAX_CLIENT_ATTRIB_STACK_DEPTH */ - 1429, /* GL_SUBPIXEL_BITS */ + 1433, /* GL_SUBPIXEL_BITS */ 584, /* GL_INDEX_BITS */ 1234, /* GL_RED_BITS */ 550, /* GL_GREEN_BITS */ 91, /* GL_BLUE_BITS */ 41, /* GL_ALPHA_BITS */ 316, /* GL_DEPTH_BITS */ - 1404, /* GL_STENCIL_BITS */ + 1408, /* GL_STENCIL_BITS */ 14, /* GL_ACCUM_RED_BITS */ 13, /* GL_ACCUM_GREEN_BITS */ 10, /* GL_ACCUM_BLUE_BITS */ @@ -3813,18 +3821,18 @@ static const unsigned reduced_enums[1284] = 692, /* GL_MAP1_GRID_SEGMENTS */ 718, /* GL_MAP2_GRID_DOMAIN */ 719, /* GL_MAP2_GRID_SEGMENTS */ - 1506, /* GL_TEXTURE_1D */ - 1508, /* GL_TEXTURE_2D */ + 1510, /* GL_TEXTURE_1D */ + 1512, /* GL_TEXTURE_2D */ 439, /* GL_FEEDBACK_BUFFER_POINTER */ 440, /* GL_FEEDBACK_BUFFER_SIZE */ 441, /* GL_FEEDBACK_BUFFER_TYPE */ 1330, /* GL_SELECTION_BUFFER_POINTER */ 1331, /* GL_SELECTION_BUFFER_SIZE */ - 1619, /* GL_TEXTURE_WIDTH */ - 1587, /* GL_TEXTURE_HEIGHT */ - 1543, /* GL_TEXTURE_COMPONENTS */ - 1527, /* GL_TEXTURE_BORDER_COLOR */ - 1526, /* GL_TEXTURE_BORDER */ + 1623, /* GL_TEXTURE_WIDTH */ + 1591, /* GL_TEXTURE_HEIGHT */ + 1547, /* GL_TEXTURE_COMPONENTS */ + 1531, /* GL_TEXTURE_BORDER_COLOR */ + 1530, /* GL_TEXTURE_BORDER */ 345, /* GL_DONT_CARE */ 437, /* GL_FASTEST */ 954, /* GL_NICEST */ @@ -3841,11 +3849,11 @@ static const unsigned reduced_enums[1284] = 219, /* GL_COMPILE */ 220, /* GL_COMPILE_AND_EXECUTE */ 111, /* GL_BYTE */ - 1650, /* GL_UNSIGNED_BYTE */ + 1654, /* GL_UNSIGNED_BYTE */ 1344, /* GL_SHORT */ - 1659, /* GL_UNSIGNED_SHORT */ + 1663, /* GL_UNSIGNED_SHORT */ 592, /* GL_INT */ - 1653, /* GL_UNSIGNED_INT */ + 1657, /* GL_UNSIGNED_INT */ 444, /* GL_FLOAT */ 1, /* GL_2_BYTES */ 5, /* GL_3_BYTES */ @@ -3857,7 +3865,7 @@ static const unsigned reduced_enums[1284] = 269, /* GL_COPY */ 50, /* GL_AND_INVERTED */ 956, /* GL_NOOP */ - 1740, /* GL_XOR */ + 1744, /* GL_XOR */ 1018, /* GL_OR */ 957, /* GL_NOR */ 427, /* GL_EQUIV */ @@ -3873,12 +3881,12 @@ static const unsigned reduced_enums[1284] = 165, /* GL_COLOR_INDEXES */ 897, /* GL_MODELVIEW */ 1189, /* GL_PROJECTION */ - 1441, /* GL_TEXTURE */ + 1445, /* GL_TEXTURE */ 138, /* GL_COLOR */ 313, /* GL_DEPTH */ 1395, /* GL_STENCIL */ 164, /* GL_COLOR_INDEX */ - 1409, /* GL_STENCIL_INDEX */ + 1413, /* GL_STENCIL_INDEX */ 322, /* GL_DEPTH_COMPONENT */ 1230, /* GL_RED */ 548, /* GL_GREEN */ @@ -3901,23 +3909,23 @@ static const unsigned reduced_enums[1284] = 1248, /* GL_REPLACE */ 575, /* GL_INCR */ 309, /* GL_DECR */ - 1674, /* GL_VENDOR */ + 1678, /* GL_VENDOR */ 1245, /* GL_RENDERER */ - 1675, /* GL_VERSION */ + 1679, /* GL_VERSION */ 431, /* GL_EXTENSIONS */ 1294, /* GL_S */ - 1432, /* GL_T */ + 1436, /* GL_T */ 1220, /* GL_R */ 1209, /* GL_Q */ 933, /* GL_MODULATE */ 308, /* GL_DECAL */ - 1577, /* GL_TEXTURE_ENV_MODE */ - 1576, /* GL_TEXTURE_ENV_COLOR */ - 1575, /* GL_TEXTURE_ENV */ + 1581, /* GL_TEXTURE_ENV_MODE */ + 1580, /* GL_TEXTURE_ENV_COLOR */ + 1579, /* GL_TEXTURE_ENV */ 432, /* GL_EYE_LINEAR */ 980, /* GL_OBJECT_LINEAR */ 1374, /* GL_SPHERE_MAP */ - 1579, /* GL_TEXTURE_GEN_MODE */ + 1583, /* GL_TEXTURE_GEN_MODE */ 982, /* GL_OBJECT_PLANE */ 433, /* GL_EYE_PLANE */ 948, /* GL_NEAREST */ @@ -3926,30 +3934,30 @@ static const unsigned reduced_enums[1284] = 644, /* GL_LINEAR_MIPMAP_NEAREST */ 951, /* GL_NEAREST_MIPMAP_LINEAR */ 643, /* GL_LINEAR_MIPMAP_LINEAR */ - 1600, /* GL_TEXTURE_MAG_FILTER */ - 1608, /* GL_TEXTURE_MIN_FILTER */ - 1621, /* GL_TEXTURE_WRAP_S */ - 1622, /* GL_TEXTURE_WRAP_T */ + 1604, /* GL_TEXTURE_MAG_FILTER */ + 1612, /* GL_TEXTURE_MIN_FILTER */ + 1625, /* GL_TEXTURE_WRAP_S */ + 1626, /* GL_TEXTURE_WRAP_T */ 117, /* GL_CLAMP */ 1247, /* GL_REPEAT */ 1107, /* GL_POLYGON_OFFSET_UNITS */ 1106, /* GL_POLYGON_OFFSET_POINT */ 1105, /* GL_POLYGON_OFFSET_LINE */ 1221, /* GL_R3_G3_B2 */ - 1671, /* GL_V2F */ - 1672, /* GL_V3F */ + 1675, /* GL_V2F */ + 1676, /* GL_V3F */ 114, /* GL_C4UB_V2F */ 115, /* GL_C4UB_V3F */ 112, /* GL_C3F_V3F */ 945, /* GL_N3F_V3F */ 113, /* GL_C4F_N3F_V3F */ - 1437, /* GL_T2F_V3F */ - 1439, /* GL_T4F_V4F */ - 1435, /* GL_T2F_C4UB_V3F */ - 1433, /* GL_T2F_C3F_V3F */ - 1436, /* GL_T2F_N3F_V3F */ - 1434, /* GL_T2F_C4F_N3F_V3F */ - 1438, /* GL_T4F_C4F_N3F_V4F */ + 1441, /* GL_T2F_V3F */ + 1443, /* GL_T4F_V4F */ + 1439, /* GL_T2F_C4UB_V3F */ + 1437, /* GL_T2F_C3F_V3F */ + 1440, /* GL_T2F_N3F_V3F */ + 1438, /* GL_T2F_C4F_N3F_V3F */ + 1442, /* GL_T4F_C4F_N3F_V4F */ 130, /* GL_CLIP_PLANE0 */ 131, /* GL_CLIP_PLANE1 */ 132, /* GL_CLIP_PLANE2 */ @@ -4009,12 +4017,12 @@ static const unsigned reduced_enums[1284] = 882, /* GL_MINMAX */ 884, /* GL_MINMAX_FORMAT */ 886, /* GL_MINMAX_SINK */ - 1440, /* GL_TABLE_TOO_LARGE_EXT */ - 1652, /* GL_UNSIGNED_BYTE_3_3_2 */ - 1661, /* GL_UNSIGNED_SHORT_4_4_4_4 */ - 1663, /* GL_UNSIGNED_SHORT_5_5_5_1 */ - 1657, /* GL_UNSIGNED_INT_8_8_8_8 */ - 1654, /* GL_UNSIGNED_INT_10_10_10_2 */ + 1444, /* GL_TABLE_TOO_LARGE_EXT */ + 1656, /* GL_UNSIGNED_BYTE_3_3_2 */ + 1665, /* GL_UNSIGNED_SHORT_4_4_4_4 */ + 1667, /* GL_UNSIGNED_SHORT_5_5_5_1 */ + 1661, /* GL_UNSIGNED_INT_8_8_8_8 */ + 1658, /* GL_UNSIGNED_INT_10_10_10_2 */ 1104, /* GL_POLYGON_OFFSET_FILL */ 1103, /* GL_POLYGON_OFFSET_FACTOR */ 1102, /* GL_POLYGON_OFFSET_BIAS */ @@ -4052,39 +4060,39 @@ static const unsigned reduced_enums[1284] = 1256, /* GL_RGB10_A2 */ 1274, /* GL_RGBA12 */ 1276, /* GL_RGBA16 */ - 1613, /* GL_TEXTURE_RED_SIZE */ - 1585, /* GL_TEXTURE_GREEN_SIZE */ - 1524, /* GL_TEXTURE_BLUE_SIZE */ - 1511, /* GL_TEXTURE_ALPHA_SIZE */ - 1598, /* GL_TEXTURE_LUMINANCE_SIZE */ - 1589, /* GL_TEXTURE_INTENSITY_SIZE */ + 1617, /* GL_TEXTURE_RED_SIZE */ + 1589, /* GL_TEXTURE_GREEN_SIZE */ + 1528, /* GL_TEXTURE_BLUE_SIZE */ + 1515, /* GL_TEXTURE_ALPHA_SIZE */ + 1602, /* GL_TEXTURE_LUMINANCE_SIZE */ + 1593, /* GL_TEXTURE_INTENSITY_SIZE */ 1249, /* GL_REPLACE_EXT */ 1197, /* GL_PROXY_TEXTURE_1D */ 1200, /* GL_PROXY_TEXTURE_2D */ - 1617, /* GL_TEXTURE_TOO_LARGE_EXT */ - 1610, /* GL_TEXTURE_PRIORITY */ - 1615, /* GL_TEXTURE_RESIDENT */ - 1514, /* GL_TEXTURE_BINDING_1D */ - 1516, /* GL_TEXTURE_BINDING_2D */ - 1518, /* GL_TEXTURE_BINDING_3D */ + 1621, /* GL_TEXTURE_TOO_LARGE_EXT */ + 1614, /* GL_TEXTURE_PRIORITY */ + 1619, /* GL_TEXTURE_RESIDENT */ + 1518, /* GL_TEXTURE_BINDING_1D */ + 1520, /* GL_TEXTURE_BINDING_2D */ + 1522, /* GL_TEXTURE_BINDING_3D */ 1028, /* GL_PACK_SKIP_IMAGES */ 1024, /* GL_PACK_IMAGE_HEIGHT */ - 1646, /* GL_UNPACK_SKIP_IMAGES */ - 1643, /* GL_UNPACK_IMAGE_HEIGHT */ - 1510, /* GL_TEXTURE_3D */ + 1650, /* GL_UNPACK_SKIP_IMAGES */ + 1647, /* GL_UNPACK_IMAGE_HEIGHT */ + 1514, /* GL_TEXTURE_3D */ 1203, /* GL_PROXY_TEXTURE_3D */ - 1572, /* GL_TEXTURE_DEPTH */ - 1620, /* GL_TEXTURE_WRAP_R */ + 1576, /* GL_TEXTURE_DEPTH */ + 1624, /* GL_TEXTURE_WRAP_R */ 794, /* GL_MAX_3D_TEXTURE_SIZE */ - 1676, /* GL_VERTEX_ARRAY */ + 1680, /* GL_VERTEX_ARRAY */ 959, /* GL_NORMAL_ARRAY */ 139, /* GL_COLOR_ARRAY */ 578, /* GL_INDEX_ARRAY */ - 1551, /* GL_TEXTURE_COORD_ARRAY */ + 1555, /* GL_TEXTURE_COORD_ARRAY */ 415, /* GL_EDGE_FLAG_ARRAY */ - 1681, /* GL_VERTEX_ARRAY_SIZE */ - 1683, /* GL_VERTEX_ARRAY_TYPE */ - 1682, /* GL_VERTEX_ARRAY_STRIDE */ + 1685, /* GL_VERTEX_ARRAY_SIZE */ + 1687, /* GL_VERTEX_ARRAY_TYPE */ + 1686, /* GL_VERTEX_ARRAY_STRIDE */ 964, /* GL_NORMAL_ARRAY_TYPE */ 963, /* GL_NORMAL_ARRAY_STRIDE */ 143, /* GL_COLOR_ARRAY_SIZE */ @@ -4092,15 +4100,15 @@ static const unsigned reduced_enums[1284] = 144, /* GL_COLOR_ARRAY_STRIDE */ 583, /* GL_INDEX_ARRAY_TYPE */ 582, /* GL_INDEX_ARRAY_STRIDE */ - 1555, /* GL_TEXTURE_COORD_ARRAY_SIZE */ - 1557, /* GL_TEXTURE_COORD_ARRAY_TYPE */ - 1556, /* GL_TEXTURE_COORD_ARRAY_STRIDE */ + 1559, /* GL_TEXTURE_COORD_ARRAY_SIZE */ + 1561, /* GL_TEXTURE_COORD_ARRAY_TYPE */ + 1560, /* GL_TEXTURE_COORD_ARRAY_STRIDE */ 419, /* GL_EDGE_FLAG_ARRAY_STRIDE */ - 1680, /* GL_VERTEX_ARRAY_POINTER */ + 1684, /* GL_VERTEX_ARRAY_POINTER */ 962, /* GL_NORMAL_ARRAY_POINTER */ 142, /* GL_COLOR_ARRAY_POINTER */ 581, /* GL_INDEX_ARRAY_POINTER */ - 1554, /* GL_TEXTURE_COORD_ARRAY_POINTER */ + 1558, /* GL_TEXTURE_COORD_ARRAY_POINTER */ 418, /* GL_EDGE_FLAG_ARRAY_POINTER */ 938, /* GL_MULTISAMPLE */ 1306, /* GL_SAMPLE_ALPHA_TO_COVERAGE */ @@ -4121,9 +4129,9 @@ static const unsigned reduced_enums[1284] = 1123, /* GL_POST_COLOR_MATRIX_GREEN_BIAS */ 1118, /* GL_POST_COLOR_MATRIX_BLUE_BIAS */ 1114, /* GL_POST_COLOR_MATRIX_ALPHA_BIAS */ - 1534, /* GL_TEXTURE_COLOR_TABLE_SGI */ + 1538, /* GL_TEXTURE_COLOR_TABLE_SGI */ 1204, /* GL_PROXY_TEXTURE_COLOR_TABLE_SGI */ - 1536, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ + 1540, /* GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ 80, /* GL_BLEND_DST_RGB */ 88, /* GL_BLEND_SRC_RGB */ 79, /* GL_BLEND_DST_ALPHA */ @@ -4148,7 +4156,7 @@ static const unsigned reduced_enums[1284] = 72, /* GL_BGRA */ 816, /* GL_MAX_ELEMENTS_VERTICES */ 815, /* GL_MAX_ELEMENTS_INDICES */ - 1588, /* GL_TEXTURE_INDEX_SIZE_EXT */ + 1592, /* GL_TEXTURE_INDEX_SIZE_EXT */ 136, /* GL_CLIP_VOLUME_CLIPPING_HINT_EXT */ 1086, /* GL_POINT_SIZE_MIN */ 1082, /* GL_POINT_SIZE_MAX */ @@ -4156,10 +4164,10 @@ static const unsigned reduced_enums[1284] = 1072, /* GL_POINT_DISTANCE_ATTENUATION */ 118, /* GL_CLAMP_TO_BORDER */ 121, /* GL_CLAMP_TO_EDGE */ - 1609, /* GL_TEXTURE_MIN_LOD */ - 1607, /* GL_TEXTURE_MAX_LOD */ - 1513, /* GL_TEXTURE_BASE_LEVEL */ - 1606, /* GL_TEXTURE_MAX_LEVEL */ + 1613, /* GL_TEXTURE_MIN_LOD */ + 1611, /* GL_TEXTURE_MAX_LOD */ + 1517, /* GL_TEXTURE_BASE_LEVEL */ + 1610, /* GL_TEXTURE_MAX_LEVEL */ 572, /* GL_IGNORE_BORDER_HP */ 246, /* GL_CONSTANT_BORDER_HP */ 1250, /* GL_REPLICATE_BORDER_HP */ @@ -4167,51 +4175,51 @@ static const unsigned reduced_enums[1284] = 987, /* GL_OCCLUSION_TEST_HP */ 988, /* GL_OCCLUSION_TEST_RESULT_HP */ 641, /* GL_LINEAR_CLIPMAP_LINEAR_SGIX */ - 1528, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */ - 1530, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */ - 1532, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */ - 1533, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */ - 1531, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */ - 1529, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */ + 1532, /* GL_TEXTURE_CLIPMAP_CENTER_SGIX */ + 1534, /* GL_TEXTURE_CLIPMAP_FRAME_SGIX */ + 1536, /* GL_TEXTURE_CLIPMAP_OFFSET_SGIX */ + 1537, /* GL_TEXTURE_CLIPMAP_VIRTUAL_DEPTH_SGIX */ + 1535, /* GL_TEXTURE_CLIPMAP_LOD_OFFSET_SGIX */ + 1533, /* GL_TEXTURE_CLIPMAP_DEPTH_SGIX */ 798, /* GL_MAX_CLIPMAP_DEPTH_SGIX */ 799, /* GL_MAX_CLIPMAP_VIRTUAL_DEPTH_SGIX */ 1149, /* GL_POST_TEXTURE_FILTER_BIAS_SGIX */ 1151, /* GL_POST_TEXTURE_FILTER_SCALE_SGIX */ 1148, /* GL_POST_TEXTURE_FILTER_BIAS_RANGE_SGIX */ 1150, /* GL_POST_TEXTURE_FILTER_SCALE_RANGE_SGIX */ - 1596, /* GL_TEXTURE_LOD_BIAS_S_SGIX */ - 1597, /* GL_TEXTURE_LOD_BIAS_T_SGIX */ - 1595, /* GL_TEXTURE_LOD_BIAS_R_SGIX */ + 1600, /* GL_TEXTURE_LOD_BIAS_S_SGIX */ + 1601, /* GL_TEXTURE_LOD_BIAS_T_SGIX */ + 1599, /* GL_TEXTURE_LOD_BIAS_R_SGIX */ 518, /* GL_GENERATE_MIPMAP */ 519, /* GL_GENERATE_MIPMAP_HINT */ 481, /* GL_FOG_OFFSET_SGIX */ 482, /* GL_FOG_OFFSET_VALUE_SGIX */ - 1542, /* GL_TEXTURE_COMPARE_SGIX */ - 1541, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */ - 1592, /* GL_TEXTURE_LEQUAL_R_SGIX */ - 1584, /* GL_TEXTURE_GEQUAL_R_SGIX */ + 1546, /* GL_TEXTURE_COMPARE_SGIX */ + 1545, /* GL_TEXTURE_COMPARE_OPERATOR_SGIX */ + 1596, /* GL_TEXTURE_LEQUAL_R_SGIX */ + 1588, /* GL_TEXTURE_GEQUAL_R_SGIX */ 323, /* GL_DEPTH_COMPONENT16 */ 326, /* GL_DEPTH_COMPONENT24 */ 329, /* GL_DEPTH_COMPONENT32 */ 274, /* GL_CULL_VERTEX_EXT */ 276, /* GL_CULL_VERTEX_OBJECT_POSITION_EXT */ 275, /* GL_CULL_VERTEX_EYE_POSITION_EXT */ - 1737, /* GL_WRAP_BORDER_SUN */ - 1535, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */ + 1741, /* GL_WRAP_BORDER_SUN */ + 1539, /* GL_TEXTURE_COLOR_WRITEMASK_SGIS */ 634, /* GL_LIGHT_MODEL_COLOR_CONTROL */ 1345, /* GL_SINGLE_COLOR */ 1333, /* GL_SEPARATE_SPECULAR_COLOR */ 1342, /* GL_SHARED_TEXTURE_PALETTE_EXT */ - 1651, /* GL_UNSIGNED_BYTE_2_3_3_REV */ - 1664, /* GL_UNSIGNED_SHORT_5_6_5 */ - 1665, /* GL_UNSIGNED_SHORT_5_6_5_REV */ - 1662, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */ - 1660, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */ - 1658, /* GL_UNSIGNED_INT_8_8_8_8_REV */ - 1656, /* GL_UNSIGNED_INT_2_10_10_10_REV */ - 1604, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */ - 1605, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */ - 1603, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */ + 1655, /* GL_UNSIGNED_BYTE_2_3_3_REV */ + 1668, /* GL_UNSIGNED_SHORT_5_6_5 */ + 1669, /* GL_UNSIGNED_SHORT_5_6_5_REV */ + 1666, /* GL_UNSIGNED_SHORT_4_4_4_4_REV */ + 1664, /* GL_UNSIGNED_SHORT_1_5_5_5_REV */ + 1662, /* GL_UNSIGNED_INT_8_8_8_8_REV */ + 1660, /* GL_UNSIGNED_INT_2_10_10_10_REV */ + 1608, /* GL_TEXTURE_MAX_CLAMP_S_SGIX */ + 1609, /* GL_TEXTURE_MAX_CLAMP_T_SGIX */ + 1607, /* GL_TEXTURE_MAX_CLAMP_R_SGIX */ 889, /* GL_MIRRORED_REPEAT */ 1289, /* GL_RGB_S3TC */ 1266, /* GL_RGB4_S3TC */ @@ -4244,46 +4252,46 @@ static const unsigned reduced_enums[1284] = 528, /* GL_GL_CURRENT_RASTER_SECONDARY_COLOR */ 28, /* GL_ALIASED_POINT_SIZE_RANGE */ 27, /* GL_ALIASED_LINE_WIDTH_RANGE */ - 1442, /* GL_TEXTURE0 */ - 1444, /* GL_TEXTURE1 */ - 1466, /* GL_TEXTURE2 */ - 1488, /* GL_TEXTURE3 */ - 1494, /* GL_TEXTURE4 */ - 1496, /* GL_TEXTURE5 */ - 1498, /* GL_TEXTURE6 */ - 1500, /* GL_TEXTURE7 */ - 1502, /* GL_TEXTURE8 */ - 1504, /* GL_TEXTURE9 */ - 1445, /* GL_TEXTURE10 */ - 1447, /* GL_TEXTURE11 */ - 1449, /* GL_TEXTURE12 */ - 1451, /* GL_TEXTURE13 */ - 1453, /* GL_TEXTURE14 */ - 1455, /* GL_TEXTURE15 */ - 1457, /* GL_TEXTURE16 */ - 1459, /* GL_TEXTURE17 */ - 1461, /* GL_TEXTURE18 */ - 1463, /* GL_TEXTURE19 */ - 1467, /* GL_TEXTURE20 */ - 1469, /* GL_TEXTURE21 */ - 1471, /* GL_TEXTURE22 */ - 1473, /* GL_TEXTURE23 */ - 1475, /* GL_TEXTURE24 */ - 1477, /* GL_TEXTURE25 */ - 1479, /* GL_TEXTURE26 */ - 1481, /* GL_TEXTURE27 */ - 1483, /* GL_TEXTURE28 */ - 1485, /* GL_TEXTURE29 */ - 1489, /* GL_TEXTURE30 */ - 1491, /* GL_TEXTURE31 */ + 1446, /* GL_TEXTURE0 */ + 1448, /* GL_TEXTURE1 */ + 1470, /* GL_TEXTURE2 */ + 1492, /* GL_TEXTURE3 */ + 1498, /* GL_TEXTURE4 */ + 1500, /* GL_TEXTURE5 */ + 1502, /* GL_TEXTURE6 */ + 1504, /* GL_TEXTURE7 */ + 1506, /* GL_TEXTURE8 */ + 1508, /* GL_TEXTURE9 */ + 1449, /* GL_TEXTURE10 */ + 1451, /* GL_TEXTURE11 */ + 1453, /* GL_TEXTURE12 */ + 1455, /* GL_TEXTURE13 */ + 1457, /* GL_TEXTURE14 */ + 1459, /* GL_TEXTURE15 */ + 1461, /* GL_TEXTURE16 */ + 1463, /* GL_TEXTURE17 */ + 1465, /* GL_TEXTURE18 */ + 1467, /* GL_TEXTURE19 */ + 1471, /* GL_TEXTURE20 */ + 1473, /* GL_TEXTURE21 */ + 1475, /* GL_TEXTURE22 */ + 1477, /* GL_TEXTURE23 */ + 1479, /* GL_TEXTURE24 */ + 1481, /* GL_TEXTURE25 */ + 1483, /* GL_TEXTURE26 */ + 1485, /* GL_TEXTURE27 */ + 1487, /* GL_TEXTURE28 */ + 1489, /* GL_TEXTURE29 */ + 1493, /* GL_TEXTURE30 */ + 1495, /* GL_TEXTURE31 */ 18, /* GL_ACTIVE_TEXTURE */ 124, /* GL_CLIENT_ACTIVE_TEXTURE */ 867, /* GL_MAX_TEXTURE_UNITS */ - 1630, /* GL_TRANSPOSE_MODELVIEW_MATRIX */ - 1633, /* GL_TRANSPOSE_PROJECTION_MATRIX */ - 1635, /* GL_TRANSPOSE_TEXTURE_MATRIX */ - 1627, /* GL_TRANSPOSE_COLOR_MATRIX */ - 1430, /* GL_SUBTRACT */ + 1634, /* GL_TRANSPOSE_MODELVIEW_MATRIX */ + 1637, /* GL_TRANSPOSE_PROJECTION_MATRIX */ + 1639, /* GL_TRANSPOSE_TEXTURE_MATRIX */ + 1631, /* GL_TRANSPOSE_COLOR_MATRIX */ + 1434, /* GL_SUBTRACT */ 856, /* GL_MAX_RENDERBUFFER_SIZE_EXT */ 222, /* GL_COMPRESSED_ALPHA */ 226, /* GL_COMPRESSED_LUMINANCE */ @@ -4291,18 +4299,18 @@ static const unsigned reduced_enums[1284] = 224, /* GL_COMPRESSED_INTENSITY */ 230, /* GL_COMPRESSED_RGB */ 231, /* GL_COMPRESSED_RGBA */ - 1549, /* GL_TEXTURE_COMPRESSION_HINT */ - 1611, /* GL_TEXTURE_RECTANGLE_ARB */ - 1521, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */ + 1553, /* GL_TEXTURE_COMPRESSION_HINT */ + 1615, /* GL_TEXTURE_RECTANGLE_ARB */ + 1525, /* GL_TEXTURE_BINDING_RECTANGLE_ARB */ 1207, /* GL_PROXY_TEXTURE_RECTANGLE_ARB */ 854, /* GL_MAX_RECTANGLE_TEXTURE_SIZE_ARB */ 335, /* GL_DEPTH_STENCIL_NV */ - 1655, /* GL_UNSIGNED_INT_24_8_NV */ + 1659, /* GL_UNSIGNED_INT_24_8_NV */ 863, /* GL_MAX_TEXTURE_LOD_BIAS */ - 1602, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */ + 1606, /* GL_TEXTURE_MAX_ANISOTROPY_EXT */ 864, /* GL_MAX_TEXTURE_MAX_ANISOTROPY_EXT */ - 1578, /* GL_TEXTURE_FILTER_CONTROL */ - 1593, /* GL_TEXTURE_LOD_BIAS */ + 1582, /* GL_TEXTURE_FILTER_CONTROL */ + 1597, /* GL_TEXTURE_LOD_BIAS */ 207, /* GL_COMBINE4 */ 857, /* GL_MAX_SHININESS_NV */ 858, /* GL_MAX_SPOT_EXPONENT_NV */ @@ -4311,14 +4319,14 @@ static const unsigned reduced_enums[1284] = 909, /* GL_MODELVIEW1_ARB */ 965, /* GL_NORMAL_MAP */ 1236, /* GL_REFLECTION_MAP */ - 1558, /* GL_TEXTURE_CUBE_MAP */ - 1519, /* GL_TEXTURE_BINDING_CUBE_MAP */ - 1566, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */ - 1560, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */ - 1568, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */ - 1562, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */ - 1570, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */ - 1564, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */ + 1562, /* GL_TEXTURE_CUBE_MAP */ + 1523, /* GL_TEXTURE_BINDING_CUBE_MAP */ + 1570, /* GL_TEXTURE_CUBE_MAP_POSITIVE_X */ + 1564, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_X */ + 1572, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Y */ + 1566, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Y */ + 1574, /* GL_TEXTURE_CUBE_MAP_POSITIVE_Z */ + 1568, /* GL_TEXTURE_CUBE_MAP_NEGATIVE_Z */ 1205, /* GL_PROXY_TEXTURE_CUBE_MAP */ 810, /* GL_MAX_CUBE_MAP_TEXTURE_SIZE */ 944, /* GL_MULTISAMPLE_FILTER_HINT_NV */ @@ -4350,26 +4358,26 @@ static const unsigned reduced_enums[1284] = 1004, /* GL_OPERAND1_ALPHA */ 1010, /* GL_OPERAND2_ALPHA */ 1016, /* GL_OPERAND3_ALPHA_NV */ - 1677, /* GL_VERTEX_ARRAY_BINDING_APPLE */ - 1741, /* GL_YCBCR_422_APPLE */ - 1666, /* GL_UNSIGNED_SHORT_8_8_APPLE */ - 1668, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */ + 1681, /* GL_VERTEX_ARRAY_BINDING_APPLE */ + 1745, /* GL_YCBCR_422_APPLE */ + 1670, /* GL_UNSIGNED_SHORT_8_8_APPLE */ + 1672, /* GL_UNSIGNED_SHORT_8_8_REV_APPLE */ 1347, /* GL_SLICE_ACCUM_SUN */ 1212, /* GL_QUAD_MESH_SUN */ - 1639, /* GL_TRIANGLE_MESH_SUN */ - 1715, /* GL_VERTEX_PROGRAM_ARB */ - 1726, /* GL_VERTEX_STATE_PROGRAM_NV */ - 1702, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */ - 1708, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */ - 1710, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */ - 1712, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */ + 1643, /* GL_TRIANGLE_MESH_SUN */ + 1719, /* GL_VERTEX_PROGRAM_ARB */ + 1730, /* GL_VERTEX_STATE_PROGRAM_NV */ + 1706, /* GL_VERTEX_ATTRIB_ARRAY_ENABLED */ + 1712, /* GL_VERTEX_ATTRIB_ARRAY_SIZE */ + 1714, /* GL_VERTEX_ATTRIB_ARRAY_STRIDE */ + 1716, /* GL_VERTEX_ATTRIB_ARRAY_TYPE */ 301, /* GL_CURRENT_VERTEX_ATTRIB */ 1168, /* GL_PROGRAM_LENGTH_ARB */ 1182, /* GL_PROGRAM_STRING_ARB */ 931, /* GL_MODELVIEW_PROJECTION_NV */ 571, /* GL_IDENTITY_NV */ 616, /* GL_INVERSE_NV */ - 1632, /* GL_TRANSPOSE_NV */ + 1636, /* GL_TRANSPOSE_NV */ 617, /* GL_INVERSE_TRANSPOSE_NV */ 840, /* GL_MAX_PROGRAM_MATRIX_STACK_DEPTH_ARB */ 839, /* GL_MAX_PROGRAM_MATRICES_ARB */ @@ -4383,33 +4391,33 @@ static const unsigned reduced_enums[1284] = 783, /* GL_MATRIX7_NV */ 286, /* GL_CURRENT_MATRIX_STACK_DEPTH_ARB */ 283, /* GL_CURRENT_MATRIX_ARB */ - 1718, /* GL_VERTEX_PROGRAM_POINT_SIZE */ - 1721, /* GL_VERTEX_PROGRAM_TWO_SIDE */ + 1722, /* GL_VERTEX_PROGRAM_POINT_SIZE */ + 1725, /* GL_VERTEX_PROGRAM_TWO_SIDE */ 1180, /* GL_PROGRAM_PARAMETER_NV */ - 1706, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */ + 1710, /* GL_VERTEX_ATTRIB_ARRAY_POINTER */ 1184, /* GL_PROGRAM_TARGET_NV */ 1181, /* GL_PROGRAM_RESIDENT_NV */ - 1624, /* GL_TRACK_MATRIX_NV */ - 1625, /* GL_TRACK_MATRIX_TRANSFORM_NV */ - 1716, /* GL_VERTEX_PROGRAM_BINDING_NV */ + 1628, /* GL_TRACK_MATRIX_NV */ + 1629, /* GL_TRACK_MATRIX_TRANSFORM_NV */ + 1720, /* GL_VERTEX_PROGRAM_BINDING_NV */ 1162, /* GL_PROGRAM_ERROR_POSITION_ARB */ 320, /* GL_DEPTH_CLAMP_NV */ - 1684, /* GL_VERTEX_ATTRIB_ARRAY0_NV */ - 1691, /* GL_VERTEX_ATTRIB_ARRAY1_NV */ - 1692, /* GL_VERTEX_ATTRIB_ARRAY2_NV */ - 1693, /* GL_VERTEX_ATTRIB_ARRAY3_NV */ - 1694, /* GL_VERTEX_ATTRIB_ARRAY4_NV */ - 1695, /* GL_VERTEX_ATTRIB_ARRAY5_NV */ - 1696, /* GL_VERTEX_ATTRIB_ARRAY6_NV */ - 1697, /* GL_VERTEX_ATTRIB_ARRAY7_NV */ - 1698, /* GL_VERTEX_ATTRIB_ARRAY8_NV */ - 1699, /* GL_VERTEX_ATTRIB_ARRAY9_NV */ - 1685, /* GL_VERTEX_ATTRIB_ARRAY10_NV */ - 1686, /* GL_VERTEX_ATTRIB_ARRAY11_NV */ - 1687, /* GL_VERTEX_ATTRIB_ARRAY12_NV */ - 1688, /* GL_VERTEX_ATTRIB_ARRAY13_NV */ - 1689, /* GL_VERTEX_ATTRIB_ARRAY14_NV */ - 1690, /* GL_VERTEX_ATTRIB_ARRAY15_NV */ + 1688, /* GL_VERTEX_ATTRIB_ARRAY0_NV */ + 1695, /* GL_VERTEX_ATTRIB_ARRAY1_NV */ + 1696, /* GL_VERTEX_ATTRIB_ARRAY2_NV */ + 1697, /* GL_VERTEX_ATTRIB_ARRAY3_NV */ + 1698, /* GL_VERTEX_ATTRIB_ARRAY4_NV */ + 1699, /* GL_VERTEX_ATTRIB_ARRAY5_NV */ + 1700, /* GL_VERTEX_ATTRIB_ARRAY6_NV */ + 1701, /* GL_VERTEX_ATTRIB_ARRAY7_NV */ + 1702, /* GL_VERTEX_ATTRIB_ARRAY8_NV */ + 1703, /* GL_VERTEX_ATTRIB_ARRAY9_NV */ + 1689, /* GL_VERTEX_ATTRIB_ARRAY10_NV */ + 1690, /* GL_VERTEX_ATTRIB_ARRAY11_NV */ + 1691, /* GL_VERTEX_ATTRIB_ARRAY12_NV */ + 1692, /* GL_VERTEX_ATTRIB_ARRAY13_NV */ + 1693, /* GL_VERTEX_ATTRIB_ARRAY14_NV */ + 1694, /* GL_VERTEX_ATTRIB_ARRAY15_NV */ 701, /* GL_MAP1_VERTEX_ATTRIB0_4_NV */ 708, /* GL_MAP1_VERTEX_ATTRIB1_4_NV */ 709, /* GL_MAP1_VERTEX_ATTRIB2_4_NV */ @@ -4442,20 +4450,20 @@ static const unsigned reduced_enums[1284] = 732, /* GL_MAP2_VERTEX_ATTRIB13_4_NV */ 733, /* GL_MAP2_VERTEX_ATTRIB14_4_NV */ 734, /* GL_MAP2_VERTEX_ATTRIB15_4_NV */ - 1547, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */ - 1544, /* GL_TEXTURE_COMPRESSED */ + 1551, /* GL_TEXTURE_COMPRESSED_IMAGE_SIZE */ + 1548, /* GL_TEXTURE_COMPRESSED */ 970, /* GL_NUM_COMPRESSED_TEXTURE_FORMATS */ 240, /* GL_COMPRESSED_TEXTURE_FORMATS */ 879, /* GL_MAX_VERTEX_UNITS_ARB */ 22, /* GL_ACTIVE_VERTEX_UNITS_ARB */ - 1736, /* GL_WEIGHT_SUM_UNITY_ARB */ - 1714, /* GL_VERTEX_BLEND_ARB */ + 1740, /* GL_WEIGHT_SUM_UNITY_ARB */ + 1718, /* GL_VERTEX_BLEND_ARB */ 303, /* GL_CURRENT_WEIGHT_ARB */ - 1735, /* GL_WEIGHT_ARRAY_TYPE_ARB */ - 1734, /* GL_WEIGHT_ARRAY_STRIDE_ARB */ - 1733, /* GL_WEIGHT_ARRAY_SIZE_ARB */ - 1732, /* GL_WEIGHT_ARRAY_POINTER_ARB */ - 1729, /* GL_WEIGHT_ARRAY_ARB */ + 1739, /* GL_WEIGHT_ARRAY_TYPE_ARB */ + 1738, /* GL_WEIGHT_ARRAY_STRIDE_ARB */ + 1737, /* GL_WEIGHT_ARRAY_SIZE_ARB */ + 1736, /* GL_WEIGHT_ARRAY_POINTER_ARB */ + 1733, /* GL_WEIGHT_ARRAY_ARB */ 346, /* GL_DOT3_RGB */ 347, /* GL_DOT3_RGBA */ 238, /* GL_COMPRESSED_RGB_FXT1_3DFX */ @@ -4500,17 +4508,17 @@ static const unsigned reduced_enums[1284] = 934, /* GL_MODULATE_ADD_ATI */ 935, /* GL_MODULATE_SIGNED_ADD_ATI */ 936, /* GL_MODULATE_SUBTRACT_ATI */ - 1742, /* GL_YCBCR_MESA */ + 1746, /* GL_YCBCR_MESA */ 1025, /* GL_PACK_INVERT_MESA */ 306, /* GL_DEBUG_OBJECT_MESA */ 307, /* GL_DEBUG_PRINT_MESA */ 305, /* GL_DEBUG_ASSERT_MESA */ 107, /* GL_BUFFER_SIZE */ 109, /* GL_BUFFER_USAGE */ - 1398, /* GL_STENCIL_BACK_FUNC */ + 1399, /* GL_STENCIL_BACK_FUNC */ 1397, /* GL_STENCIL_BACK_FAIL */ - 1399, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */ - 1400, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */ + 1401, /* GL_STENCIL_BACK_PASS_DEPTH_FAIL */ + 1403, /* GL_STENCIL_BACK_PASS_DEPTH_PASS */ 485, /* GL_FRAGMENT_PROGRAM_ARB */ 1159, /* GL_PROGRAM_ALU_INSTRUCTIONS_ARB */ 1187, /* GL_PROGRAM_TEX_INSTRUCTIONS_ARB */ @@ -4552,10 +4560,10 @@ static const unsigned reduced_enums[1284] = 790, /* GL_MATRIX_INDEX_ARRAY_TYPE_ARB */ 789, /* GL_MATRIX_INDEX_ARRAY_STRIDE_ARB */ 787, /* GL_MATRIX_INDEX_ARRAY_POINTER_ARB */ - 1573, /* GL_TEXTURE_DEPTH_SIZE */ + 1577, /* GL_TEXTURE_DEPTH_SIZE */ 339, /* GL_DEPTH_TEXTURE_MODE */ - 1539, /* GL_TEXTURE_COMPARE_MODE */ - 1537, /* GL_TEXTURE_COMPARE_FUNC */ + 1543, /* GL_TEXTURE_COMPARE_MODE */ + 1541, /* GL_TEXTURE_COMPARE_FUNC */ 217, /* GL_COMPARE_R_TO_TEXTURE */ 1093, /* GL_POINT_SPRITE */ 266, /* GL_COORD_REPLACE */ @@ -4565,7 +4573,7 @@ static const unsigned reduced_enums[1284] = 1216, /* GL_QUERY_RESULT */ 1218, /* GL_QUERY_RESULT_AVAILABLE */ 873, /* GL_MAX_VERTEX_ATTRIBS */ - 1704, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */ + 1708, /* GL_VERTEX_ATTRIB_ARRAY_NORMALIZED */ 337, /* GL_DEPTH_STENCIL_TO_RGBA_NV */ 336, /* GL_DEPTH_STENCIL_TO_BGRA_NV */ 859, /* GL_MAX_TEXTURE_COORDS */ @@ -4573,23 +4581,23 @@ static const unsigned reduced_enums[1284] = 1164, /* GL_PROGRAM_ERROR_STRING_ARB */ 1166, /* GL_PROGRAM_FORMAT_ASCII_ARB */ 1165, /* GL_PROGRAM_FORMAT_ARB */ - 1618, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */ + 1622, /* GL_TEXTURE_UNSIGNED_REMAP_MODE_NV */ 318, /* GL_DEPTH_BOUNDS_TEST_EXT */ 317, /* GL_DEPTH_BOUNDS_EXT */ 52, /* GL_ARRAY_BUFFER */ 420, /* GL_ELEMENT_ARRAY_BUFFER */ 54, /* GL_ARRAY_BUFFER_BINDING */ 422, /* GL_ELEMENT_ARRAY_BUFFER_BINDING */ - 1678, /* GL_VERTEX_ARRAY_BUFFER_BINDING */ + 1682, /* GL_VERTEX_ARRAY_BUFFER_BINDING */ 960, /* GL_NORMAL_ARRAY_BUFFER_BINDING */ 140, /* GL_COLOR_ARRAY_BUFFER_BINDING */ 579, /* GL_INDEX_ARRAY_BUFFER_BINDING */ - 1552, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */ + 1556, /* GL_TEXTURE_COORD_ARRAY_BUFFER_BINDING */ 416, /* GL_EDGE_FLAG_ARRAY_BUFFER_BINDING */ 1323, /* GL_SECONDARY_COLOR_ARRAY_BUFFER_BINDING */ 463, /* GL_FOG_COORDINATE_ARRAY_BUFFER_BINDING */ - 1730, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */ - 1700, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */ + 1734, /* GL_WEIGHT_ARRAY_BUFFER_BINDING */ + 1704, /* GL_VERTEX_ATTRIB_ARRAY_BUFFER_BINDING */ 1167, /* GL_PROGRAM_INSTRUCTIONS_ARB */ 835, /* GL_MAX_PROGRAM_INSTRUCTIONS_ARB */ 1173, /* GL_PROGRAM_NATIVE_INSTRUCTIONS_ARB */ @@ -4613,14 +4621,14 @@ static const unsigned reduced_enums[1284] = 836, /* GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB */ 832, /* GL_MAX_PROGRAM_ENV_PARAMETERS_ARB */ 1188, /* GL_PROGRAM_UNDER_NATIVE_LIMITS_ARB */ - 1629, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */ + 1633, /* GL_TRANSPOSE_CURRENT_MATRIX_ARB */ 1226, /* GL_READ_ONLY */ - 1738, /* GL_WRITE_ONLY */ + 1742, /* GL_WRITE_ONLY */ 1228, /* GL_READ_WRITE */ 101, /* GL_BUFFER_ACCESS */ 103, /* GL_BUFFER_MAPPED */ 105, /* GL_BUFFER_MAP_POINTER */ - 1623, /* GL_TIME_ELAPSED_EXT */ + 1627, /* GL_TIME_ELAPSED_EXT */ 746, /* GL_MATRIX0_ARB */ 758, /* GL_MATRIX1_ARB */ 770, /* GL_MATRIX2_ARB */ @@ -4653,9 +4661,9 @@ static const unsigned reduced_enums[1284] = 769, /* GL_MATRIX29_ARB */ 772, /* GL_MATRIX30_ARB */ 773, /* GL_MATRIX31_ARB */ - 1425, /* GL_STREAM_DRAW */ - 1427, /* GL_STREAM_READ */ - 1423, /* GL_STREAM_COPY */ + 1429, /* GL_STREAM_DRAW */ + 1431, /* GL_STREAM_READ */ + 1427, /* GL_STREAM_COPY */ 1391, /* GL_STATIC_DRAW */ 1393, /* GL_STATIC_READ */ 1389, /* GL_STATIC_COPY */ @@ -4672,12 +4680,12 @@ static const unsigned reduced_enums[1284] = 838, /* GL_MAX_PROGRAM_LOOP_DEPTH_NV */ 837, /* GL_MAX_PROGRAM_LOOP_COUNT_NV */ 795, /* GL_MAX_ARRAY_TEXTURE_LAYERS_EXT */ - 1419, /* GL_STENCIL_TEST_TWO_SIDE_EXT */ + 1423, /* GL_STENCIL_TEST_TWO_SIDE_EXT */ 17, /* GL_ACTIVE_STENCIL_FACE_EXT */ 894, /* GL_MIRROR_CLAMP_TO_BORDER_EXT */ 1304, /* GL_SAMPLES_PASSED */ 486, /* GL_FRAGMENT_SHADER */ - 1724, /* GL_VERTEX_SHADER */ + 1728, /* GL_VERTEX_SHADER */ 1178, /* GL_PROGRAM_OBJECT_ARB */ 1336, /* GL_SHADER_OBJECT_ARB */ 819, /* GL_MAX_FRAGMENT_UNIFORM_COMPONENTS */ @@ -4715,7 +4723,7 @@ static const unsigned reduced_enums[1284] = 312, /* GL_DELETE_STATUS */ 221, /* GL_COMPILE_STATUS */ 659, /* GL_LINK_STATUS */ - 1673, /* GL_VALIDATE_STATUS */ + 1677, /* GL_VALIDATE_STATUS */ 591, /* GL_INFO_LOG_LENGTH */ 56, /* GL_ATTACHED_SHADERS */ 20, /* GL_ACTIVE_UNIFORMS */ @@ -4738,12 +4746,12 @@ static const unsigned reduced_enums[1284] = 1038, /* GL_PALETTE8_RGB5_A1_OES */ 574, /* GL_IMPLEMENTATION_COLOR_READ_TYPE_OES */ 573, /* GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES */ - 1507, /* GL_TEXTURE_1D_ARRAY_EXT */ + 1511, /* GL_TEXTURE_1D_ARRAY_EXT */ 1198, /* GL_PROXY_TEXTURE_1D_ARRAY_EXT */ - 1509, /* GL_TEXTURE_2D_ARRAY_EXT */ + 1513, /* GL_TEXTURE_2D_ARRAY_EXT */ 1201, /* GL_PROXY_TEXTURE_2D_ARRAY_EXT */ - 1515, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */ - 1517, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */ + 1519, /* GL_TEXTURE_BINDING_1D_ARRAY_EXT */ + 1521, /* GL_TEXTURE_BINDING_2D_ARRAY_EXT */ 543, /* GL_GL_SRGB */ 544, /* GL_GL_SRGB8 */ 546, /* GL_GL_SRGB_ALPHA */ @@ -4758,10 +4766,10 @@ static const unsigned reduced_enums[1284] = 525, /* GL_GL_COMPRESSED_SLUMINANCE_ALPHA */ 1095, /* GL_POINT_SPRITE_COORD_ORIGIN */ 667, /* GL_LOWER_LEFT */ - 1670, /* GL_UPPER_LEFT */ - 1401, /* GL_STENCIL_BACK_REF */ - 1402, /* GL_STENCIL_BACK_VALUE_MASK */ - 1403, /* GL_STENCIL_BACK_WRITEMASK */ + 1674, /* GL_UPPER_LEFT */ + 1405, /* GL_STENCIL_BACK_REF */ + 1406, /* GL_STENCIL_BACK_VALUE_MASK */ + 1407, /* GL_STENCIL_BACK_WRITEMASK */ 403, /* GL_DRAW_FRAMEBUFFER_BINDING_EXT */ 1240, /* GL_RENDERBUFFER_BINDING_EXT */ 1225, /* GL_READ_FRAMEBUFFER_EXT */ @@ -4806,15 +4814,15 @@ static const unsigned reduced_enums[1284] = 1244, /* GL_RENDERBUFFER_WIDTH_EXT */ 1242, /* GL_RENDERBUFFER_HEIGHT_EXT */ 1243, /* GL_RENDERBUFFER_INTERNAL_FORMAT_EXT */ - 1414, /* GL_STENCIL_INDEX_EXT */ - 1411, /* GL_STENCIL_INDEX1_EXT */ - 1412, /* GL_STENCIL_INDEX4_EXT */ - 1413, /* GL_STENCIL_INDEX8_EXT */ - 1410, /* GL_STENCIL_INDEX16_EXT */ + 1418, /* GL_STENCIL_INDEX_EXT */ + 1415, /* GL_STENCIL_INDEX1_EXT */ + 1416, /* GL_STENCIL_INDEX4_EXT */ + 1417, /* GL_STENCIL_INDEX8_EXT */ + 1414, /* GL_STENCIL_INDEX16_EXT */ 428, /* GL_EVAL_BIT */ 1222, /* GL_RASTER_POSITION_UNCLIPPED_IBM */ 661, /* GL_LIST_BIT */ - 1523, /* GL_TEXTURE_BIT */ + 1527, /* GL_TEXTURE_BIT */ 1319, /* GL_SCISSOR_BIT */ 29, /* GL_ALL_ATTRIB_BITS */ 941, /* GL_MULTISAMPLE_BIT */ diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 2f88ec615a..34f279f1f4 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -818,6 +818,10 @@ _mesa_init_exec_table(struct _glapi_table *exec) #if FEATURE_EXT_framebuffer_object SET_FramebufferTextureLayerEXT(exec, _mesa_FramebufferTextureLayerEXT); #endif + + /* GL_ATI_separate_stencil */ + SET_StencilFuncSeparateATI(exec, _mesa_StencilFuncSeparateATI); + SET_StencilOpSeparateATI(exec, _mesa_StencilOpSeparateATI); } diff --git a/src/mesa/main/stencil.c b/src/mesa/main/stencil.c index ca7f6eaf88..17464310b1 100644 --- a/src/mesa/main/stencil.c +++ b/src/mesa/main/stencil.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.2 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -53,6 +53,48 @@ #include "mtypes.h" +static GLboolean +validate_stencil_op(GLcontext *ctx, GLenum op) +{ + switch (op) { + case GL_KEEP: + case GL_ZERO: + case GL_REPLACE: + case GL_INCR: + case GL_DECR: + case GL_INVERT: + return GL_TRUE; + case GL_INCR_WRAP_EXT: + case GL_DECR_WRAP_EXT: + if (ctx->Extensions.EXT_stencil_wrap) { + return GL_TRUE; + } + /* FALL-THROUGH */ + default: + return GL_FALSE; + } +} + + +static GLboolean +validate_stencil_func(GLcontext *ctx, GLenum func) +{ + switch (func) { + case GL_NEVER: + case GL_LESS: + case GL_LEQUAL: + case GL_GREATER: + case GL_GEQUAL: + case GL_EQUAL: + case GL_NOTEQUAL: + case GL_ALWAYS: + return GL_TRUE; + default: + return GL_FALSE; + } +} + + /** * Set the clear value for the stencil buffer. * @@ -103,34 +145,15 @@ _mesa_StencilFuncSeparateATI( GLenum frontfunc, GLenum backfunc, GLint ref, GLui const GLint stencilMax = (1 << ctx->DrawBuffer->Visual.stencilBits) - 1; ASSERT_OUTSIDE_BEGIN_END(ctx); - switch (frontfunc) { - case GL_NEVER: - case GL_LESS: - case GL_LEQUAL: - case GL_GREATER: - case GL_GEQUAL: - case GL_EQUAL: - case GL_NOTEQUAL: - case GL_ALWAYS: - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glStencilFuncSeparateATI (0x%04x)", frontfunc ); - return; + if (!validate_stencil_func(ctx, frontfunc)) { + _mesa_error(ctx, GL_INVALID_ENUM, + "glStencilFuncSeparateATI(frontfunc)"); + return; } - - switch (backfunc) { - case GL_NEVER: - case GL_LESS: - case GL_LEQUAL: - case GL_GREATER: - case GL_GEQUAL: - case GL_EQUAL: - case GL_NOTEQUAL: - case GL_ALWAYS: - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glStencilFuncSeparateATI (0x%04x)", backfunc ); - return; + if (!validate_stencil_func(ctx, backfunc)) { + _mesa_error(ctx, GL_INVALID_ENUM, + "glStencilFuncSeparateATI(backfunc)"); + return; } ref = CLAMP( ref, 0, stencilMax ); @@ -157,6 +180,61 @@ _mesa_StencilFuncSeparateATI( GLenum frontfunc, GLenum backfunc, GLint ref, GLui } +void APIENTRY +_mesa_StencilOpSeparateATI(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass) +{ + GLboolean set = GL_FALSE; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (!validate_stencil_op(ctx, sfail)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparateATI(sfail)"); + return; + } + if (!validate_stencil_op(ctx, zfail)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparateATI(zfail)"); + return; + } + if (!validate_stencil_op(ctx, zpass)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparateATI(zpass)"); + return; + } + if (face != GL_FRONT && face != GL_BACK && face != GL_FRONT_AND_BACK) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparateATI(face)"); + return; + } + + if (face != GL_BACK) { + /* set front */ + if (ctx->Stencil.ZFailFunc[0] != zfail || + ctx->Stencil.ZPassFunc[0] != zpass || + ctx->Stencil.FailFunc[0] != sfail){ + FLUSH_VERTICES(ctx, _NEW_STENCIL); + ctx->Stencil.ZFailFunc[0] = zfail; + ctx->Stencil.ZPassFunc[0] = zpass; + ctx->Stencil.FailFunc[0] = sfail; + set = GL_TRUE; + } + } + if (face != GL_FRONT) { + /* set back */ + if (ctx->Stencil.ZFailFunc[1] != zfail || + ctx->Stencil.ZPassFunc[1] != zpass || + ctx->Stencil.FailFunc[1] != sfail) { + FLUSH_VERTICES(ctx, _NEW_STENCIL); + ctx->Stencil.ZFailFunc[1] = zfail; + ctx->Stencil.ZPassFunc[1] = zpass; + ctx->Stencil.FailFunc[1] = sfail; + set = GL_TRUE; + } + } + if (set && ctx->Driver.StencilOpSeparate) { + ctx->Driver.StencilOpSeparate(ctx, face, sfail, zfail, zpass); + } +} + + + /** * Set the function and reference value for stencil testing. * @@ -177,19 +255,9 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask ) const GLint stencilMax = (1 << ctx->DrawBuffer->Visual.stencilBits) - 1; ASSERT_OUTSIDE_BEGIN_END(ctx); - switch (func) { - case GL_NEVER: - case GL_LESS: - case GL_LEQUAL: - case GL_GREATER: - case GL_GEQUAL: - case GL_EQUAL: - case GL_NOTEQUAL: - case GL_ALWAYS: - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glStencilFunc (0x%04x)", func ); - return; + if (!validate_stencil_func(ctx, func)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilFunc(func)"); + return; } ref = CLAMP( ref, 0, stencilMax ); @@ -293,59 +361,17 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass) GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - switch (fail) { - case GL_KEEP: - case GL_ZERO: - case GL_REPLACE: - case GL_INCR: - case GL_DECR: - case GL_INVERT: - break; - case GL_INCR_WRAP_EXT: - case GL_DECR_WRAP_EXT: - if (ctx->Extensions.EXT_stencil_wrap) { - break; - } - /* FALL-THROUGH */ - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOp"); - return; + if (!validate_stencil_op(ctx, fail)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOp(sfail)"); + return; } - switch (zfail) { - case GL_KEEP: - case GL_ZERO: - case GL_REPLACE: - case GL_INCR: - case GL_DECR: - case GL_INVERT: - break; - case GL_INCR_WRAP_EXT: - case GL_DECR_WRAP_EXT: - if (ctx->Extensions.EXT_stencil_wrap) { - break; - } - /* FALL-THROUGH */ - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOp"); - return; + if (!validate_stencil_op(ctx, zfail)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOp(zfail)"); + return; } - switch (zpass) { - case GL_KEEP: - case GL_ZERO: - case GL_REPLACE: - case GL_INCR: - case GL_DECR: - case GL_INVERT: - break; - case GL_INCR_WRAP_EXT: - case GL_DECR_WRAP_EXT: - if (ctx->Extensions.EXT_stencil_wrap) { - break; - } - /* FALL-THROUGH */ - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOp"); - return; + if (!validate_stencil_op(ctx, zpass)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOp(zpass)"); + return; } if (ctx->Extensions.ATI_separate_stencil) { @@ -428,59 +454,17 @@ _mesa_StencilOpSeparate(GLenum face, GLenum fail, GLenum zfail, GLenum zpass) return; } - switch (fail) { - case GL_KEEP: - case GL_ZERO: - case GL_REPLACE: - case GL_INCR: - case GL_DECR: - case GL_INVERT: - break; - case GL_INCR_WRAP_EXT: - case GL_DECR_WRAP_EXT: - if (ctx->Extensions.EXT_stencil_wrap) { - break; - } - /* FALL-THROUGH */ - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(fail)"); - return; + if (!validate_stencil_op(ctx, fail)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(sfail)"); + return; } - switch (zfail) { - case GL_KEEP: - case GL_ZERO: - case GL_REPLACE: - case GL_INCR: - case GL_DECR: - case GL_INVERT: - break; - case GL_INCR_WRAP_EXT: - case GL_DECR_WRAP_EXT: - if (ctx->Extensions.EXT_stencil_wrap) { - break; - } - /* FALL-THROUGH */ - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(zfail)"); - return; + if (!validate_stencil_op(ctx, zfail)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(zfail)"); + return; } - switch (zpass) { - case GL_KEEP: - case GL_ZERO: - case GL_REPLACE: - case GL_INCR: - case GL_DECR: - case GL_INVERT: - break; - case GL_INCR_WRAP_EXT: - case GL_DECR_WRAP_EXT: - if (ctx->Extensions.EXT_stencil_wrap) { - break; - } - /* FALL-THROUGH */ - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(zpass)"); - return; + if (!validate_stencil_op(ctx, zpass)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(zpass)"); + return; } FLUSH_VERTICES(ctx, _NEW_STENCIL); @@ -513,20 +497,9 @@ _mesa_StencilFuncSeparate(GLenum face, GLenum func, GLint ref, GLuint mask) _mesa_error(ctx, GL_INVALID_ENUM, "glStencilFuncSeparate(face)"); return; } - - switch (func) { - case GL_NEVER: - case GL_LESS: - case GL_LEQUAL: - case GL_GREATER: - case GL_GEQUAL: - case GL_EQUAL: - case GL_NOTEQUAL: - case GL_ALWAYS: - break; - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilFuncSeparate(func)"); - return; + if (!validate_stencil_func(ctx, func)) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilFuncSeparate(func)"); + return; } ref = CLAMP(ref, 0, stencilMax); diff --git a/src/mesa/main/stencil.h b/src/mesa/main/stencil.h index 0be9810005..15a65dbcac 100644 --- a/src/mesa/main/stencil.h +++ b/src/mesa/main/stencil.h @@ -5,9 +5,9 @@ /* * Mesa 3-D graphics library - * Version: 6.5 + * Version: 7.1 * - * Copyright (C) 1999-2005 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -62,6 +62,10 @@ _mesa_StencilOpSeparate(GLenum face, GLenum fail, GLenum zfail, GLenum zpass); extern void GLAPIENTRY _mesa_StencilFuncSeparate(GLenum face, GLenum func, GLint ref, GLuint mask); + +extern void APIENTRY +_mesa_StencilOpSeparateATI(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass); + extern void GLAPIENTRY _mesa_StencilFuncSeparateATI(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); diff --git a/src/mesa/sparc/glapi_sparc.S b/src/mesa/sparc/glapi_sparc.S index 8725c7ee4d..91dc8c1792 100644 --- a/src/mesa/sparc/glapi_sparc.S +++ b/src/mesa/sparc/glapi_sparc.S @@ -838,6 +838,8 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */ .globl gl_dispatch_stub_770 ; .type gl_dispatch_stub_770,#function .globl gl_dispatch_stub_771 ; .type gl_dispatch_stub_771,#function .globl gl_dispatch_stub_772 ; .type gl_dispatch_stub_772,#function + .globl gl_dispatch_stub_773 ; .type gl_dispatch_stub_773,#function + .globl gl_dispatch_stub_774 ; .type gl_dispatch_stub_774,#function .globl _mesa_sparc_glapi_begin ; .type _mesa_sparc_glapi_begin,#function _mesa_sparc_glapi_begin: @@ -1614,6 +1616,8 @@ _mesa_sparc_glapi_begin: GL_STUB(gl_dispatch_stub_770, _gloffset__dispatch_stub_770) GL_STUB(gl_dispatch_stub_771, _gloffset__dispatch_stub_771) GL_STUB(gl_dispatch_stub_772, _gloffset__dispatch_stub_772) + GL_STUB(gl_dispatch_stub_773, _gloffset__dispatch_stub_773) + GL_STUB(gl_dispatch_stub_774, _gloffset__dispatch_stub_774) .globl _mesa_sparc_glapi_end ; .type _mesa_sparc_glapi_end,#function _mesa_sparc_glapi_end: diff --git a/src/mesa/x86-64/glapi_x86-64.S b/src/mesa/x86-64/glapi_x86-64.S index 171e95d1b8..041ced0268 100644 --- a/src/mesa/x86-64/glapi_x86-64.S +++ b/src/mesa/x86-64/glapi_x86-64.S @@ -29220,7 +29220,11 @@ GL_PREFIX(_dispatch_stub_771): pushq %rdi pushq %rsi pushq %rdx + pushq %rcx + pushq %rbp call _x86_64_get_dispatch@PLT + popq %rbp + popq %rcx popq %rdx popq %rsi popq %rdi @@ -29236,7 +29240,11 @@ GL_PREFIX(_dispatch_stub_771): pushq %rdi pushq %rsi pushq %rdx + pushq %rcx + pushq %rbp call _glapi_get_dispatch + popq %rbp + popq %rcx popq %rdx popq %rsi popq %rdi @@ -29258,7 +29266,11 @@ GL_PREFIX(_dispatch_stub_772): pushq %rdi pushq %rsi pushq %rdx + pushq %rcx + pushq %rbp call _x86_64_get_dispatch@PLT + popq %rbp + popq %rcx popq %rdx popq %rsi popq %rdi @@ -29274,7 +29286,11 @@ GL_PREFIX(_dispatch_stub_772): pushq %rdi pushq %rsi pushq %rdx + pushq %rcx + pushq %rbp call _glapi_get_dispatch + popq %rbp + popq %rcx popq %rdx popq %rsi popq %rdi @@ -29283,6 +29299,82 @@ GL_PREFIX(_dispatch_stub_772): #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(_dispatch_stub_772), .-GL_PREFIX(_dispatch_stub_772) + .p2align 4,,15 + .globl GL_PREFIX(_dispatch_stub_773) + .type GL_PREFIX(_dispatch_stub_773), @function + HIDDEN(GL_PREFIX(_dispatch_stub_773)) +GL_PREFIX(_dispatch_stub_773): +#if defined(GLX_USE_TLS) + call _x86_64_get_dispatch@PLT + movq 6184(%rax), %r11 + jmp *%r11 +#elif defined(PTHREADS) + pushq %rdi + pushq %rsi + pushq %rdx + call _x86_64_get_dispatch@PLT + popq %rdx + popq %rsi + popq %rdi + movq 6184(%rax), %r11 + jmp *%r11 +#else + movq _glapi_Dispatch(%rip), %rax + testq %rax, %rax + je 1f + movq 6184(%rax), %r11 + jmp *%r11 +1: + pushq %rdi + pushq %rsi + pushq %rdx + call _glapi_get_dispatch + popq %rdx + popq %rsi + popq %rdi + movq 6184(%rax), %r11 + jmp *%r11 +#endif /* defined(GLX_USE_TLS) */ + .size GL_PREFIX(_dispatch_stub_773), .-GL_PREFIX(_dispatch_stub_773) + + .p2align 4,,15 + .globl GL_PREFIX(_dispatch_stub_774) + .type GL_PREFIX(_dispatch_stub_774), @function + HIDDEN(GL_PREFIX(_dispatch_stub_774)) +GL_PREFIX(_dispatch_stub_774): +#if defined(GLX_USE_TLS) + call _x86_64_get_dispatch@PLT + movq 6192(%rax), %r11 + jmp *%r11 +#elif defined(PTHREADS) + pushq %rdi + pushq %rsi + pushq %rdx + call _x86_64_get_dispatch@PLT + popq %rdx + popq %rsi + popq %rdi + movq 6192(%rax), %r11 + jmp *%r11 +#else + movq _glapi_Dispatch(%rip), %rax + testq %rax, %rax + je 1f + movq 6192(%rax), %r11 + jmp *%r11 +1: + pushq %rdi + pushq %rsi + pushq %rdx + call _glapi_get_dispatch + popq %rdx + popq %rsi + popq %rdi + movq 6192(%rax), %r11 + jmp *%r11 +#endif /* defined(GLX_USE_TLS) */ + .size GL_PREFIX(_dispatch_stub_774), .-GL_PREFIX(_dispatch_stub_774) + .globl GL_PREFIX(ArrayElementEXT) ; .set GL_PREFIX(ArrayElementEXT), GL_PREFIX(ArrayElement) .globl GL_PREFIX(BindTextureEXT) ; .set GL_PREFIX(BindTextureEXT), GL_PREFIX(BindTexture) .globl GL_PREFIX(DrawArraysEXT) ; .set GL_PREFIX(DrawArraysEXT), GL_PREFIX(DrawArrays) diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S index 8b7204cdb7..0afb054a14 100644 --- a/src/mesa/x86/glapi_x86.S +++ b/src/mesa/x86/glapi_x86.S @@ -948,14 +948,18 @@ GLNAME(gl_dispatch_functions_start): GL_STUB(_dispatch_stub_767, _gloffset_BlitFramebufferEXT, _dispatch_stub_767@40) HIDDEN(GL_PREFIX(_dispatch_stub_767, _dispatch_stub_767@40)) GL_STUB(FramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT, FramebufferTextureLayerEXT@20) - GL_STUB(_dispatch_stub_769, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_769@16) + GL_STUB(_dispatch_stub_769, _gloffset_StencilFuncSeparateATI, _dispatch_stub_769@16) HIDDEN(GL_PREFIX(_dispatch_stub_769, _dispatch_stub_769@16)) - GL_STUB(_dispatch_stub_770, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_770@16) + GL_STUB(_dispatch_stub_770, _gloffset_StencilOpSeparateATI, _dispatch_stub_770@16) HIDDEN(GL_PREFIX(_dispatch_stub_770, _dispatch_stub_770@16)) - GL_STUB(_dispatch_stub_771, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_771@12) - HIDDEN(GL_PREFIX(_dispatch_stub_771, _dispatch_stub_771@12)) - GL_STUB(_dispatch_stub_772, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_772@12) - HIDDEN(GL_PREFIX(_dispatch_stub_772, _dispatch_stub_772@12)) + GL_STUB(_dispatch_stub_771, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_771@16) + HIDDEN(GL_PREFIX(_dispatch_stub_771, _dispatch_stub_771@16)) + GL_STUB(_dispatch_stub_772, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_772@16) + HIDDEN(GL_PREFIX(_dispatch_stub_772, _dispatch_stub_772@16)) + GL_STUB(_dispatch_stub_773, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_773@12) + HIDDEN(GL_PREFIX(_dispatch_stub_773, _dispatch_stub_773@12)) + GL_STUB(_dispatch_stub_774, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_774@12) + HIDDEN(GL_PREFIX(_dispatch_stub_774, _dispatch_stub_774@12)) GL_STUB_ALIAS(ArrayElementEXT, _gloffset_ArrayElement, ArrayElementEXT@4, ArrayElement, ArrayElement@4) GL_STUB_ALIAS(BindTextureEXT, _gloffset_BindTexture, BindTextureEXT@8, BindTexture, BindTexture@8) GL_STUB_ALIAS(DrawArraysEXT, _gloffset_DrawArrays, DrawArraysEXT@12, DrawArrays, DrawArrays@12) -- cgit v1.2.3 From 9edac96d69b6f9942c170c573c9aba4c35550639 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 30 Oct 2007 10:24:34 -0600 Subject: Alias glStencilOpSeparateATI with glStencilOpSeparate. --- src/mesa/drivers/dri/common/extension_helper.h | 12 +- src/mesa/glapi/dispatch.h | 17 +-- src/mesa/glapi/gl_API.xml | 4 +- src/mesa/glapi/glapioffsets.h | 12 +- src/mesa/glapi/glapitable.h | 11 +- src/mesa/glapi/glapitemp.h | 33 ++--- src/mesa/glapi/glprocs.h | 196 ++++++++++++------------- src/mesa/main/state.c | 1 - src/mesa/main/stencil.c | 101 ++++--------- src/mesa/main/stencil.h | 3 - src/mesa/sparc/glapi_sparc.S | 2 - src/mesa/x86-64/glapi_x86-64.S | 46 ------ src/mesa/x86/glapi_x86.S | 13 +- 13 files changed, 168 insertions(+), 283 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/common/extension_helper.h b/src/mesa/drivers/dri/common/extension_helper.h index 726d9900c3..065c5d8dae 100644 --- a/src/mesa/drivers/dri/common/extension_helper.h +++ b/src/mesa/drivers/dri/common/extension_helper.h @@ -1105,10 +1105,11 @@ static const char IsRenderbufferEXT_names[] = ""; #endif -#if defined(need_GL_VERSION_2_0) +#if defined(need_GL_VERSION_2_0) || defined(need_GL_ATI_separate_stencil) static const char StencilOpSeparate_names[] = "iiii\0" /* Parameter signature */ "glStencilOpSeparate\0" + "glStencilOpSeparateATI\0" ""; #endif @@ -3839,13 +3840,6 @@ static const char Binormal3svEXT_names[] = ""; #endif -#if defined(need_GL_ATI_separate_stencil) -static const char StencilOpSeparateATI_names[] = - "iiii\0" /* Parameter signature */ - "glStencilOpSeparateATI\0" - ""; -#endif - #if defined(need_GL_EXT_light_texture) static const char ApplyTextureEXT_names[] = "i\0" /* Parameter signature */ @@ -5198,7 +5192,7 @@ static const struct dri_extension_function GL_ATI_fragment_shader_functions[] = #if defined(need_GL_ATI_separate_stencil) static const struct dri_extension_function GL_ATI_separate_stencil_functions[] = { - { StencilOpSeparateATI_names, StencilOpSeparateATI_remap_index, -1 }, + { StencilOpSeparate_names, StencilOpSeparate_remap_index, -1 }, { StencilFuncSeparateATI_names, StencilFuncSeparateATI_remap_index, -1 }, { NULL, 0, 0 } }; diff --git a/src/mesa/glapi/dispatch.h b/src/mesa/glapi/dispatch.h index e46f159f94..db23e44396 100644 --- a/src/mesa/glapi/dispatch.h +++ b/src/mesa/glapi/dispatch.h @@ -2368,9 +2368,6 @@ #define CALL_StencilFuncSeparateATI(disp, parameters) (*((disp)->StencilFuncSeparateATI)) parameters #define GET_StencilFuncSeparateATI(disp) ((disp)->StencilFuncSeparateATI) #define SET_StencilFuncSeparateATI(disp, fn) ((disp)->StencilFuncSeparateATI = fn) -#define CALL_StencilOpSeparateATI(disp, parameters) (*((disp)->StencilOpSeparateATI)) parameters -#define GET_StencilOpSeparateATI(disp) ((disp)->StencilOpSeparateATI) -#define SET_StencilOpSeparateATI(disp, fn) ((disp)->StencilOpSeparateATI = fn) #define CALL_ProgramEnvParameters4fvEXT(disp, parameters) (*((disp)->ProgramEnvParameters4fvEXT)) parameters #define GET_ProgramEnvParameters4fvEXT(disp) ((disp)->ProgramEnvParameters4fvEXT) #define SET_ProgramEnvParameters4fvEXT(disp, fn) ((disp)->ProgramEnvParameters4fvEXT = fn) @@ -2386,7 +2383,7 @@ #else -#define driDispatchRemapTable_size 367 +#define driDispatchRemapTable_size 366 extern int driDispatchRemapTable[ driDispatchRemapTable_size ]; #define AttachShader_remap_index 0 @@ -2751,11 +2748,10 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ]; #define BlitFramebufferEXT_remap_index 359 #define FramebufferTextureLayerEXT_remap_index 360 #define StencilFuncSeparateATI_remap_index 361 -#define StencilOpSeparateATI_remap_index 362 -#define ProgramEnvParameters4fvEXT_remap_index 363 -#define ProgramLocalParameters4fvEXT_remap_index 364 -#define GetQueryObjecti64vEXT_remap_index 365 -#define GetQueryObjectui64vEXT_remap_index 366 +#define ProgramEnvParameters4fvEXT_remap_index 362 +#define ProgramLocalParameters4fvEXT_remap_index 363 +#define GetQueryObjecti64vEXT_remap_index 364 +#define GetQueryObjectui64vEXT_remap_index 365 #define CALL_AttachShader(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLuint, GLuint)), driDispatchRemapTable[AttachShader_remap_index], parameters) #define GET_AttachShader(disp) GET_by_offset(disp, driDispatchRemapTable[AttachShader_remap_index]) @@ -3843,9 +3839,6 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ]; #define CALL_StencilFuncSeparateATI(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLint, GLuint)), driDispatchRemapTable[StencilFuncSeparateATI_remap_index], parameters) #define GET_StencilFuncSeparateATI(disp) GET_by_offset(disp, driDispatchRemapTable[StencilFuncSeparateATI_remap_index]) #define SET_StencilFuncSeparateATI(disp, fn) SET_by_offset(disp, driDispatchRemapTable[StencilFuncSeparateATI_remap_index], fn) -#define CALL_StencilOpSeparateATI(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLenum, GLenum, GLenum)), driDispatchRemapTable[StencilOpSeparateATI_remap_index], parameters) -#define GET_StencilOpSeparateATI(disp) GET_by_offset(disp, driDispatchRemapTable[StencilOpSeparateATI_remap_index]) -#define SET_StencilOpSeparateATI(disp, fn) SET_by_offset(disp, driDispatchRemapTable[StencilOpSeparateATI_remap_index], fn) #define CALL_ProgramEnvParameters4fvEXT(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLuint, GLsizei, const GLfloat *)), driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index], parameters) #define GET_ProgramEnvParameters4fvEXT(disp) GET_by_offset(disp, driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index]) #define SET_ProgramEnvParameters4fvEXT(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index], fn) diff --git a/src/mesa/glapi/gl_API.xml b/src/mesa/glapi/gl_API.xml index e16ffe870f..3d47e6f2ce 100644 --- a/src/mesa/glapi/gl_API.xml +++ b/src/mesa/glapi/gl_API.xml @@ -5138,7 +5138,7 @@ - + @@ -12170,7 +12170,7 @@ - + diff --git a/src/mesa/glapi/glapioffsets.h b/src/mesa/glapi/glapioffsets.h index a2db342ffe..4799fd3076 100644 --- a/src/mesa/glapi/glapioffsets.h +++ b/src/mesa/glapi/glapioffsets.h @@ -802,12 +802,11 @@ #define _gloffset_BlitFramebufferEXT 767 #define _gloffset_FramebufferTextureLayerEXT 768 #define _gloffset_StencilFuncSeparateATI 769 -#define _gloffset_StencilOpSeparateATI 770 -#define _gloffset_ProgramEnvParameters4fvEXT 771 -#define _gloffset_ProgramLocalParameters4fvEXT 772 -#define _gloffset_GetQueryObjecti64vEXT 773 -#define _gloffset_GetQueryObjectui64vEXT 774 -#define _gloffset_FIRST_DYNAMIC 775 +#define _gloffset_ProgramEnvParameters4fvEXT 770 +#define _gloffset_ProgramLocalParameters4fvEXT 771 +#define _gloffset_GetQueryObjecti64vEXT 772 +#define _gloffset_GetQueryObjectui64vEXT 773 +#define _gloffset_FIRST_DYNAMIC 774 #else @@ -1173,7 +1172,6 @@ #define _gloffset_BlitFramebufferEXT driDispatchRemapTable[BlitFramebufferEXT_remap_index] #define _gloffset_FramebufferTextureLayerEXT driDispatchRemapTable[FramebufferTextureLayerEXT_remap_index] #define _gloffset_StencilFuncSeparateATI driDispatchRemapTable[StencilFuncSeparateATI_remap_index] -#define _gloffset_StencilOpSeparateATI driDispatchRemapTable[StencilOpSeparateATI_remap_index] #define _gloffset_ProgramEnvParameters4fvEXT driDispatchRemapTable[ProgramEnvParameters4fvEXT_remap_index] #define _gloffset_ProgramLocalParameters4fvEXT driDispatchRemapTable[ProgramLocalParameters4fvEXT_remap_index] #define _gloffset_GetQueryObjecti64vEXT driDispatchRemapTable[GetQueryObjecti64vEXT_remap_index] diff --git a/src/mesa/glapi/glapitable.h b/src/mesa/glapi/glapitable.h index 148aa96b88..22c2dc2f69 100644 --- a/src/mesa/glapi/glapitable.h +++ b/src/mesa/glapi/glapitable.h @@ -464,7 +464,7 @@ struct _glapi_table GLboolean (GLAPIENTRYP IsShader)(GLuint shader); /* 420 */ void (GLAPIENTRYP StencilFuncSeparate)(GLenum face, GLenum func, GLint ref, GLuint mask); /* 421 */ void (GLAPIENTRYP StencilMaskSeparate)(GLenum face, GLuint mask); /* 422 */ - void (GLAPIENTRYP StencilOpSeparate)(GLenum face, GLenum fail, GLenum zfail, GLenum zpass); /* 423 */ + void (GLAPIENTRYP StencilOpSeparate)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass); /* 423 */ void (GLAPIENTRYP UniformMatrix2x3fv)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value); /* 424 */ void (GLAPIENTRYP UniformMatrix2x4fv)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value); /* 425 */ void (GLAPIENTRYP UniformMatrix3x2fv)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value); /* 426 */ @@ -811,11 +811,10 @@ struct _glapi_table void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 767 */ void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 768 */ void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 769 */ - void (GLAPIENTRYP StencilOpSeparateATI)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass); /* 770 */ - void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 771 */ - void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 772 */ - void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 773 */ - void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 774 */ + void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 770 */ + void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 771 */ + void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 772 */ + void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 773 */ }; #endif /* !defined( _GLAPI_TABLE_H_ ) */ diff --git a/src/mesa/glapi/glapitemp.h b/src/mesa/glapi/glapitemp.h index cd420fdee5..6ded362c0e 100644 --- a/src/mesa/glapi/glapitemp.h +++ b/src/mesa/glapi/glapitemp.h @@ -2754,9 +2754,16 @@ KEYWORD1 void KEYWORD2 NAME(StencilMaskSeparate)(GLenum face, GLuint mask) DISPATCH(StencilMaskSeparate, (face, mask), (F, "glStencilMaskSeparate(0x%x, %d);\n", face, mask)); } -KEYWORD1 void KEYWORD2 NAME(StencilOpSeparate)(GLenum face, GLenum fail, GLenum zfail, GLenum zpass) +KEYWORD1 void KEYWORD2 NAME(StencilOpSeparate)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass) { - DISPATCH(StencilOpSeparate, (face, fail, zfail, zpass), (F, "glStencilOpSeparate(0x%x, 0x%x, 0x%x, 0x%x);\n", face, fail, zfail, zpass)); + DISPATCH(StencilOpSeparate, (face, sfail, zfail, zpass), (F, "glStencilOpSeparate(0x%x, 0x%x, 0x%x, 0x%x);\n", face, sfail, zfail, zpass)); +} + +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_423)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass); + +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_423)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass) +{ + DISPATCH(StencilOpSeparate, (face, sfail, zfail, zpass), (F, "glStencilOpSeparateATI(0x%x, 0x%x, 0x%x, 0x%x);\n", face, sfail, zfail, zpass)); } KEYWORD1 void KEYWORD2 NAME(UniformMatrix2x3fv)(GLint location, GLsizei count, GLboolean transpose, const GLfloat * value) @@ -5453,37 +5460,30 @@ KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLenum frontfunc, GLenum bac DISPATCH(StencilFuncSeparateATI, (frontfunc, backfunc, ref, mask), (F, "glStencilFuncSeparateATI(0x%x, 0x%x, %d, %d);\n", frontfunc, backfunc, ref, mask)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) { - DISPATCH(StencilOpSeparateATI, (face, sfail, zfail, zpass), (F, "glStencilOpSeparateATI(0x%x, 0x%x, 0x%x, 0x%x);\n", face, sfail, zfail, zpass)); + DISPATCH(ProgramEnvParameters4fvEXT, (target, index, count, params), (F, "glProgramEnvParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params)); } KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) -{ - DISPATCH(ProgramEnvParameters4fvEXT, (target, index, count, params), (F, "glProgramEnvParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params)); -} - -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); - -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) { DISPATCH(ProgramLocalParameters4fvEXT, (target, index, count, params), (F, "glProgramLocalParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLint64EXT * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLuint id, GLenum pname, GLint64EXT * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLint64EXT * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLuint id, GLenum pname, GLint64EXT * params) { DISPATCH(GetQueryObjecti64vEXT, (id, pname, params), (F, "glGetQueryObjecti64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_774)(GLuint id, GLenum pname, GLuint64EXT * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLuint64EXT * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_774)(GLuint id, GLenum pname, GLuint64EXT * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLuint64EXT * params) { DISPATCH(GetQueryObjectui64vEXT, (id, pname, params), (F, "glGetQueryObjectui64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params)); } @@ -6276,7 +6276,6 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = { TABLE_ENTRY(_dispatch_stub_771), TABLE_ENTRY(_dispatch_stub_772), TABLE_ENTRY(_dispatch_stub_773), - TABLE_ENTRY(_dispatch_stub_774), /* A whole bunch of no-op functions. These might be called * when someone tries to call a dynamically-registered * extension function without a current rendering context. diff --git a/src/mesa/glapi/glprocs.h b/src/mesa/glapi/glprocs.h index 657b74e578..abc22a9a04 100644 --- a/src/mesa/glapi/glprocs.h +++ b/src/mesa/glapi/glprocs.h @@ -822,7 +822,6 @@ static const char gl_string_table[] = "glBlitFramebufferEXT\0" "glFramebufferTextureLayerEXT\0" "glStencilFuncSeparateATI\0" - "glStencilOpSeparateATI\0" "glProgramEnvParameters4fvEXT\0" "glProgramLocalParameters4fvEXT\0" "glGetQueryObjecti64vEXT\0" @@ -918,6 +917,7 @@ static const char gl_string_table[] = "glMultiTexCoord4iv\0" "glMultiTexCoord4s\0" "glMultiTexCoord4sv\0" + "glStencilOpSeparateATI\0" "glLoadTransposeMatrixd\0" "glLoadTransposeMatrixf\0" "glMultTransposeMatrixd\0" @@ -1151,7 +1151,6 @@ static const char gl_string_table[] = #define gl_dispatch_stub_771 mgl_dispatch_stub_771 #define gl_dispatch_stub_772 mgl_dispatch_stub_772 #define gl_dispatch_stub_773 mgl_dispatch_stub_773 -#define gl_dispatch_stub_774 mgl_dispatch_stub_774 #endif /* USE_MGL_NAMESPACE */ @@ -1203,7 +1202,6 @@ extern void gl_dispatch_stub_770(void); extern void gl_dispatch_stub_771(void); extern void gl_dispatch_stub_772(void); extern void gl_dispatch_stub_773(void); -extern void gl_dispatch_stub_774(void); #endif /* defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING) */ static const glprocs_table_t static_functions[] = { @@ -1977,102 +1975,102 @@ static const glprocs_table_t static_functions[] = { NAME_FUNC_OFFSET(13449, gl_dispatch_stub_767, gl_dispatch_stub_767, NULL, _gloffset_BlitFramebufferEXT), NAME_FUNC_OFFSET(13470, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT), NAME_FUNC_OFFSET(13499, gl_dispatch_stub_769, gl_dispatch_stub_769, NULL, _gloffset_StencilFuncSeparateATI), - NAME_FUNC_OFFSET(13524, gl_dispatch_stub_770, gl_dispatch_stub_770, NULL, _gloffset_StencilOpSeparateATI), - NAME_FUNC_OFFSET(13547, gl_dispatch_stub_771, gl_dispatch_stub_771, NULL, _gloffset_ProgramEnvParameters4fvEXT), - NAME_FUNC_OFFSET(13576, gl_dispatch_stub_772, gl_dispatch_stub_772, NULL, _gloffset_ProgramLocalParameters4fvEXT), - NAME_FUNC_OFFSET(13607, gl_dispatch_stub_773, gl_dispatch_stub_773, NULL, _gloffset_GetQueryObjecti64vEXT), - NAME_FUNC_OFFSET(13631, gl_dispatch_stub_774, gl_dispatch_stub_774, NULL, _gloffset_GetQueryObjectui64vEXT), - NAME_FUNC_OFFSET(13656, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement), - NAME_FUNC_OFFSET(13674, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture), - NAME_FUNC_OFFSET(13691, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays), - NAME_FUNC_OFFSET(13707, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident), - NAME_FUNC_OFFSET(13732, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D), - NAME_FUNC_OFFSET(13752, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D), - NAME_FUNC_OFFSET(13772, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D), - NAME_FUNC_OFFSET(13795, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D), - NAME_FUNC_OFFSET(13818, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures), - NAME_FUNC_OFFSET(13838, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures), - NAME_FUNC_OFFSET(13855, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv), - NAME_FUNC_OFFSET(13872, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture), - NAME_FUNC_OFFSET(13887, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures), - NAME_FUNC_OFFSET(13911, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D), - NAME_FUNC_OFFSET(13930, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D), - NAME_FUNC_OFFSET(13949, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor), - NAME_FUNC_OFFSET(13965, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation), - NAME_FUNC_OFFSET(13984, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements), - NAME_FUNC_OFFSET(14007, glColorTable, glColorTable, NULL, _gloffset_ColorTable), - NAME_FUNC_OFFSET(14023, glColorTable, glColorTable, NULL, _gloffset_ColorTable), - NAME_FUNC_OFFSET(14039, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv), - NAME_FUNC_OFFSET(14066, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv), - NAME_FUNC_OFFSET(14093, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable), - NAME_FUNC_OFFSET(14113, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), - NAME_FUNC_OFFSET(14132, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), - NAME_FUNC_OFFSET(14151, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), - NAME_FUNC_OFFSET(14181, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), - NAME_FUNC_OFFSET(14211, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), - NAME_FUNC_OFFSET(14241, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), - NAME_FUNC_OFFSET(14271, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable), - NAME_FUNC_OFFSET(14290, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable), - NAME_FUNC_OFFSET(14313, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D), - NAME_FUNC_OFFSET(14338, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D), - NAME_FUNC_OFFSET(14363, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf), - NAME_FUNC_OFFSET(14390, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv), - NAME_FUNC_OFFSET(14418, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri), - NAME_FUNC_OFFSET(14445, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv), - NAME_FUNC_OFFSET(14473, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D), - NAME_FUNC_OFFSET(14502, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D), - NAME_FUNC_OFFSET(14531, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter), - NAME_FUNC_OFFSET(14557, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv), - NAME_FUNC_OFFSET(14588, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv), - NAME_FUNC_OFFSET(14619, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter), - NAME_FUNC_OFFSET(14643, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D), - NAME_FUNC_OFFSET(14666, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram), - NAME_FUNC_OFFSET(14684, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv), - NAME_FUNC_OFFSET(14713, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv), - NAME_FUNC_OFFSET(14742, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax), - NAME_FUNC_OFFSET(14757, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv), - NAME_FUNC_OFFSET(14783, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv), - NAME_FUNC_OFFSET(14809, glHistogram, glHistogram, NULL, _gloffset_Histogram), - NAME_FUNC_OFFSET(14824, glMinmax, glMinmax, NULL, _gloffset_Minmax), - NAME_FUNC_OFFSET(14836, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram), - NAME_FUNC_OFFSET(14856, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax), - NAME_FUNC_OFFSET(14873, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D), - NAME_FUNC_OFFSET(14889, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D), - NAME_FUNC_OFFSET(14908, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D), - NAME_FUNC_OFFSET(14931, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB), - NAME_FUNC_OFFSET(14947, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB), - NAME_FUNC_OFFSET(14969, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB), - NAME_FUNC_OFFSET(14987, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB), - NAME_FUNC_OFFSET(15006, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB), - NAME_FUNC_OFFSET(15024, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB), - NAME_FUNC_OFFSET(15043, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB), - NAME_FUNC_OFFSET(15061, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB), - NAME_FUNC_OFFSET(15080, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB), - NAME_FUNC_OFFSET(15098, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB), - NAME_FUNC_OFFSET(15117, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB), - NAME_FUNC_OFFSET(15135, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB), - NAME_FUNC_OFFSET(15154, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB), - NAME_FUNC_OFFSET(15172, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB), - NAME_FUNC_OFFSET(15191, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB), - NAME_FUNC_OFFSET(15209, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB), - NAME_FUNC_OFFSET(15228, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB), - NAME_FUNC_OFFSET(15246, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB), - NAME_FUNC_OFFSET(15265, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB), - NAME_FUNC_OFFSET(15283, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB), - NAME_FUNC_OFFSET(15302, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB), - NAME_FUNC_OFFSET(15320, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB), - NAME_FUNC_OFFSET(15339, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB), - NAME_FUNC_OFFSET(15357, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB), - NAME_FUNC_OFFSET(15376, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB), - NAME_FUNC_OFFSET(15394, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB), - NAME_FUNC_OFFSET(15413, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB), - NAME_FUNC_OFFSET(15431, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB), - NAME_FUNC_OFFSET(15450, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB), - NAME_FUNC_OFFSET(15468, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB), - NAME_FUNC_OFFSET(15487, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB), - NAME_FUNC_OFFSET(15505, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB), - NAME_FUNC_OFFSET(15524, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB), - NAME_FUNC_OFFSET(15542, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB), + NAME_FUNC_OFFSET(13524, gl_dispatch_stub_770, gl_dispatch_stub_770, NULL, _gloffset_ProgramEnvParameters4fvEXT), + NAME_FUNC_OFFSET(13553, gl_dispatch_stub_771, gl_dispatch_stub_771, NULL, _gloffset_ProgramLocalParameters4fvEXT), + NAME_FUNC_OFFSET(13584, gl_dispatch_stub_772, gl_dispatch_stub_772, NULL, _gloffset_GetQueryObjecti64vEXT), + NAME_FUNC_OFFSET(13608, gl_dispatch_stub_773, gl_dispatch_stub_773, NULL, _gloffset_GetQueryObjectui64vEXT), + NAME_FUNC_OFFSET(13633, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement), + NAME_FUNC_OFFSET(13651, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture), + NAME_FUNC_OFFSET(13668, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays), + NAME_FUNC_OFFSET(13684, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident), + NAME_FUNC_OFFSET(13709, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D), + NAME_FUNC_OFFSET(13729, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D), + NAME_FUNC_OFFSET(13749, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D), + NAME_FUNC_OFFSET(13772, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D), + NAME_FUNC_OFFSET(13795, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures), + NAME_FUNC_OFFSET(13815, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures), + NAME_FUNC_OFFSET(13832, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv), + NAME_FUNC_OFFSET(13849, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture), + NAME_FUNC_OFFSET(13864, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures), + NAME_FUNC_OFFSET(13888, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D), + NAME_FUNC_OFFSET(13907, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D), + NAME_FUNC_OFFSET(13926, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor), + NAME_FUNC_OFFSET(13942, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation), + NAME_FUNC_OFFSET(13961, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements), + NAME_FUNC_OFFSET(13984, glColorTable, glColorTable, NULL, _gloffset_ColorTable), + NAME_FUNC_OFFSET(14000, glColorTable, glColorTable, NULL, _gloffset_ColorTable), + NAME_FUNC_OFFSET(14016, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv), + NAME_FUNC_OFFSET(14043, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv), + NAME_FUNC_OFFSET(14070, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable), + NAME_FUNC_OFFSET(14090, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), + NAME_FUNC_OFFSET(14109, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), + NAME_FUNC_OFFSET(14128, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), + NAME_FUNC_OFFSET(14158, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), + NAME_FUNC_OFFSET(14188, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), + NAME_FUNC_OFFSET(14218, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), + NAME_FUNC_OFFSET(14248, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable), + NAME_FUNC_OFFSET(14267, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable), + NAME_FUNC_OFFSET(14290, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D), + NAME_FUNC_OFFSET(14315, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D), + NAME_FUNC_OFFSET(14340, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf), + NAME_FUNC_OFFSET(14367, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv), + NAME_FUNC_OFFSET(14395, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri), + NAME_FUNC_OFFSET(14422, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv), + NAME_FUNC_OFFSET(14450, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D), + NAME_FUNC_OFFSET(14479, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D), + NAME_FUNC_OFFSET(14508, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter), + NAME_FUNC_OFFSET(14534, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv), + NAME_FUNC_OFFSET(14565, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv), + NAME_FUNC_OFFSET(14596, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter), + NAME_FUNC_OFFSET(14620, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D), + NAME_FUNC_OFFSET(14643, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram), + NAME_FUNC_OFFSET(14661, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv), + NAME_FUNC_OFFSET(14690, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv), + NAME_FUNC_OFFSET(14719, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax), + NAME_FUNC_OFFSET(14734, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv), + NAME_FUNC_OFFSET(14760, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv), + NAME_FUNC_OFFSET(14786, glHistogram, glHistogram, NULL, _gloffset_Histogram), + NAME_FUNC_OFFSET(14801, glMinmax, glMinmax, NULL, _gloffset_Minmax), + NAME_FUNC_OFFSET(14813, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram), + NAME_FUNC_OFFSET(14833, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax), + NAME_FUNC_OFFSET(14850, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D), + NAME_FUNC_OFFSET(14866, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D), + NAME_FUNC_OFFSET(14885, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D), + NAME_FUNC_OFFSET(14908, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB), + NAME_FUNC_OFFSET(14924, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB), + NAME_FUNC_OFFSET(14946, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB), + NAME_FUNC_OFFSET(14964, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB), + NAME_FUNC_OFFSET(14983, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB), + NAME_FUNC_OFFSET(15001, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB), + NAME_FUNC_OFFSET(15020, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB), + NAME_FUNC_OFFSET(15038, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB), + NAME_FUNC_OFFSET(15057, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB), + NAME_FUNC_OFFSET(15075, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB), + NAME_FUNC_OFFSET(15094, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB), + NAME_FUNC_OFFSET(15112, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB), + NAME_FUNC_OFFSET(15131, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB), + NAME_FUNC_OFFSET(15149, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB), + NAME_FUNC_OFFSET(15168, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB), + NAME_FUNC_OFFSET(15186, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB), + NAME_FUNC_OFFSET(15205, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB), + NAME_FUNC_OFFSET(15223, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB), + NAME_FUNC_OFFSET(15242, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB), + NAME_FUNC_OFFSET(15260, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB), + NAME_FUNC_OFFSET(15279, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB), + NAME_FUNC_OFFSET(15297, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB), + NAME_FUNC_OFFSET(15316, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB), + NAME_FUNC_OFFSET(15334, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB), + NAME_FUNC_OFFSET(15353, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB), + NAME_FUNC_OFFSET(15371, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB), + NAME_FUNC_OFFSET(15390, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB), + NAME_FUNC_OFFSET(15408, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB), + NAME_FUNC_OFFSET(15427, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB), + NAME_FUNC_OFFSET(15445, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB), + NAME_FUNC_OFFSET(15464, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB), + NAME_FUNC_OFFSET(15482, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB), + NAME_FUNC_OFFSET(15501, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB), + NAME_FUNC_OFFSET(15519, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB), + NAME_FUNC_OFFSET(15538, glStencilOpSeparate, glStencilOpSeparate, NULL, _gloffset_StencilOpSeparate), NAME_FUNC_OFFSET(15561, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB), NAME_FUNC_OFFSET(15584, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB), NAME_FUNC_OFFSET(15607, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB), diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 34f279f1f4..0b1c56fdd5 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -821,7 +821,6 @@ _mesa_init_exec_table(struct _glapi_table *exec) /* GL_ATI_separate_stencil */ SET_StencilFuncSeparateATI(exec, _mesa_StencilFuncSeparateATI); - SET_StencilOpSeparateATI(exec, _mesa_StencilOpSeparateATI); } diff --git a/src/mesa/main/stencil.c b/src/mesa/main/stencil.c index 17464310b1..c1906197de 100644 --- a/src/mesa/main/stencil.c +++ b/src/mesa/main/stencil.c @@ -180,61 +180,6 @@ _mesa_StencilFuncSeparateATI( GLenum frontfunc, GLenum backfunc, GLint ref, GLui } -void APIENTRY -_mesa_StencilOpSeparateATI(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass) -{ - GLboolean set = GL_FALSE; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (!validate_stencil_op(ctx, sfail)) { - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparateATI(sfail)"); - return; - } - if (!validate_stencil_op(ctx, zfail)) { - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparateATI(zfail)"); - return; - } - if (!validate_stencil_op(ctx, zpass)) { - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparateATI(zpass)"); - return; - } - if (face != GL_FRONT && face != GL_BACK && face != GL_FRONT_AND_BACK) { - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparateATI(face)"); - return; - } - - if (face != GL_BACK) { - /* set front */ - if (ctx->Stencil.ZFailFunc[0] != zfail || - ctx->Stencil.ZPassFunc[0] != zpass || - ctx->Stencil.FailFunc[0] != sfail){ - FLUSH_VERTICES(ctx, _NEW_STENCIL); - ctx->Stencil.ZFailFunc[0] = zfail; - ctx->Stencil.ZPassFunc[0] = zpass; - ctx->Stencil.FailFunc[0] = sfail; - set = GL_TRUE; - } - } - if (face != GL_FRONT) { - /* set back */ - if (ctx->Stencil.ZFailFunc[1] != zfail || - ctx->Stencil.ZPassFunc[1] != zpass || - ctx->Stencil.FailFunc[1] != sfail) { - FLUSH_VERTICES(ctx, _NEW_STENCIL); - ctx->Stencil.ZFailFunc[1] = zfail; - ctx->Stencil.ZPassFunc[1] = zpass; - ctx->Stencil.FailFunc[1] = sfail; - set = GL_TRUE; - } - } - if (set && ctx->Driver.StencilOpSeparate) { - ctx->Driver.StencilOpSeparate(ctx, face, sfail, zfail, zpass); - } -} - - - /** * Set the function and reference value for stencil testing. * @@ -444,17 +389,13 @@ _mesa_ActiveStencilFaceEXT(GLenum face) * instead. */ void GLAPIENTRY -_mesa_StencilOpSeparate(GLenum face, GLenum fail, GLenum zfail, GLenum zpass) +_mesa_StencilOpSeparate(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass) { + GLboolean set = GL_FALSE; GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - if (face != GL_FRONT && face != GL_BACK && face != GL_FRONT_AND_BACK) { - _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(face)"); - return; - } - - if (!validate_stencil_op(ctx, fail)) { + if (!validate_stencil_op(ctx, sfail)) { _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(sfail)"); return; } @@ -466,21 +407,37 @@ _mesa_StencilOpSeparate(GLenum face, GLenum fail, GLenum zfail, GLenum zpass) _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(zpass)"); return; } - - FLUSH_VERTICES(ctx, _NEW_STENCIL); + if (face != GL_FRONT && face != GL_BACK && face != GL_FRONT_AND_BACK) { + _mesa_error(ctx, GL_INVALID_ENUM, "glStencilOpSeparate(face)"); + return; + } if (face != GL_BACK) { - ctx->Stencil.FailFunc[0] = fail; - ctx->Stencil.ZFailFunc[0] = zfail; - ctx->Stencil.ZPassFunc[0] = zpass; + /* set front */ + if (ctx->Stencil.ZFailFunc[0] != zfail || + ctx->Stencil.ZPassFunc[0] != zpass || + ctx->Stencil.FailFunc[0] != sfail){ + FLUSH_VERTICES(ctx, _NEW_STENCIL); + ctx->Stencil.ZFailFunc[0] = zfail; + ctx->Stencil.ZPassFunc[0] = zpass; + ctx->Stencil.FailFunc[0] = sfail; + set = GL_TRUE; + } } if (face != GL_FRONT) { - ctx->Stencil.FailFunc[1] = fail; - ctx->Stencil.ZFailFunc[1] = zfail; - ctx->Stencil.ZPassFunc[1] = zpass; + /* set back */ + if (ctx->Stencil.ZFailFunc[1] != zfail || + ctx->Stencil.ZPassFunc[1] != zpass || + ctx->Stencil.FailFunc[1] != sfail) { + FLUSH_VERTICES(ctx, _NEW_STENCIL); + ctx->Stencil.ZFailFunc[1] = zfail; + ctx->Stencil.ZPassFunc[1] = zpass; + ctx->Stencil.FailFunc[1] = sfail; + set = GL_TRUE; + } } - if (ctx->Driver.StencilOpSeparate) { - ctx->Driver.StencilOpSeparate(ctx, face, fail, zfail, zpass); + if (set && ctx->Driver.StencilOpSeparate) { + ctx->Driver.StencilOpSeparate(ctx, face, sfail, zfail, zpass); } } diff --git a/src/mesa/main/stencil.h b/src/mesa/main/stencil.h index 15a65dbcac..252ac932ac 100644 --- a/src/mesa/main/stencil.h +++ b/src/mesa/main/stencil.h @@ -63,9 +63,6 @@ extern void GLAPIENTRY _mesa_StencilFuncSeparate(GLenum face, GLenum func, GLint ref, GLuint mask); -extern void APIENTRY -_mesa_StencilOpSeparateATI(GLenum face, GLenum sfail, GLenum zfail, GLenum zpass); - extern void GLAPIENTRY _mesa_StencilFuncSeparateATI(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); diff --git a/src/mesa/sparc/glapi_sparc.S b/src/mesa/sparc/glapi_sparc.S index 91dc8c1792..b214bef54d 100644 --- a/src/mesa/sparc/glapi_sparc.S +++ b/src/mesa/sparc/glapi_sparc.S @@ -839,7 +839,6 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */ .globl gl_dispatch_stub_771 ; .type gl_dispatch_stub_771,#function .globl gl_dispatch_stub_772 ; .type gl_dispatch_stub_772,#function .globl gl_dispatch_stub_773 ; .type gl_dispatch_stub_773,#function - .globl gl_dispatch_stub_774 ; .type gl_dispatch_stub_774,#function .globl _mesa_sparc_glapi_begin ; .type _mesa_sparc_glapi_begin,#function _mesa_sparc_glapi_begin: @@ -1617,7 +1616,6 @@ _mesa_sparc_glapi_begin: GL_STUB(gl_dispatch_stub_771, _gloffset__dispatch_stub_771) GL_STUB(gl_dispatch_stub_772, _gloffset__dispatch_stub_772) GL_STUB(gl_dispatch_stub_773, _gloffset__dispatch_stub_773) - GL_STUB(gl_dispatch_stub_774, _gloffset__dispatch_stub_774) .globl _mesa_sparc_glapi_end ; .type _mesa_sparc_glapi_end,#function _mesa_sparc_glapi_end: diff --git a/src/mesa/x86-64/glapi_x86-64.S b/src/mesa/x86-64/glapi_x86-64.S index 041ced0268..a239a6749d 100644 --- a/src/mesa/x86-64/glapi_x86-64.S +++ b/src/mesa/x86-64/glapi_x86-64.S @@ -29266,11 +29266,7 @@ GL_PREFIX(_dispatch_stub_772): pushq %rdi pushq %rsi pushq %rdx - pushq %rcx - pushq %rbp call _x86_64_get_dispatch@PLT - popq %rbp - popq %rcx popq %rdx popq %rsi popq %rdi @@ -29286,11 +29282,7 @@ GL_PREFIX(_dispatch_stub_772): pushq %rdi pushq %rsi pushq %rdx - pushq %rcx - pushq %rbp call _glapi_get_dispatch - popq %rbp - popq %rcx popq %rdx popq %rsi popq %rdi @@ -29337,44 +29329,6 @@ GL_PREFIX(_dispatch_stub_773): #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(_dispatch_stub_773), .-GL_PREFIX(_dispatch_stub_773) - .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_774) - .type GL_PREFIX(_dispatch_stub_774), @function - HIDDEN(GL_PREFIX(_dispatch_stub_774)) -GL_PREFIX(_dispatch_stub_774): -#if defined(GLX_USE_TLS) - call _x86_64_get_dispatch@PLT - movq 6192(%rax), %r11 - jmp *%r11 -#elif defined(PTHREADS) - pushq %rdi - pushq %rsi - pushq %rdx - call _x86_64_get_dispatch@PLT - popq %rdx - popq %rsi - popq %rdi - movq 6192(%rax), %r11 - jmp *%r11 -#else - movq _glapi_Dispatch(%rip), %rax - testq %rax, %rax - je 1f - movq 6192(%rax), %r11 - jmp *%r11 -1: - pushq %rdi - pushq %rsi - pushq %rdx - call _glapi_get_dispatch - popq %rdx - popq %rsi - popq %rdi - movq 6192(%rax), %r11 - jmp *%r11 -#endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_774), .-GL_PREFIX(_dispatch_stub_774) - .globl GL_PREFIX(ArrayElementEXT) ; .set GL_PREFIX(ArrayElementEXT), GL_PREFIX(ArrayElement) .globl GL_PREFIX(BindTextureEXT) ; .set GL_PREFIX(BindTextureEXT), GL_PREFIX(BindTexture) .globl GL_PREFIX(DrawArraysEXT) ; .set GL_PREFIX(DrawArraysEXT), GL_PREFIX(DrawArrays) diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S index 0afb054a14..40ecc20753 100644 --- a/src/mesa/x86/glapi_x86.S +++ b/src/mesa/x86/glapi_x86.S @@ -950,16 +950,14 @@ GLNAME(gl_dispatch_functions_start): GL_STUB(FramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT, FramebufferTextureLayerEXT@20) GL_STUB(_dispatch_stub_769, _gloffset_StencilFuncSeparateATI, _dispatch_stub_769@16) HIDDEN(GL_PREFIX(_dispatch_stub_769, _dispatch_stub_769@16)) - GL_STUB(_dispatch_stub_770, _gloffset_StencilOpSeparateATI, _dispatch_stub_770@16) + GL_STUB(_dispatch_stub_770, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_770@16) HIDDEN(GL_PREFIX(_dispatch_stub_770, _dispatch_stub_770@16)) - GL_STUB(_dispatch_stub_771, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_771@16) + GL_STUB(_dispatch_stub_771, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_771@16) HIDDEN(GL_PREFIX(_dispatch_stub_771, _dispatch_stub_771@16)) - GL_STUB(_dispatch_stub_772, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_772@16) - HIDDEN(GL_PREFIX(_dispatch_stub_772, _dispatch_stub_772@16)) - GL_STUB(_dispatch_stub_773, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_773@12) + GL_STUB(_dispatch_stub_772, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_772@12) + HIDDEN(GL_PREFIX(_dispatch_stub_772, _dispatch_stub_772@12)) + GL_STUB(_dispatch_stub_773, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_773@12) HIDDEN(GL_PREFIX(_dispatch_stub_773, _dispatch_stub_773@12)) - GL_STUB(_dispatch_stub_774, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_774@12) - HIDDEN(GL_PREFIX(_dispatch_stub_774, _dispatch_stub_774@12)) GL_STUB_ALIAS(ArrayElementEXT, _gloffset_ArrayElement, ArrayElementEXT@4, ArrayElement, ArrayElement@4) GL_STUB_ALIAS(BindTextureEXT, _gloffset_BindTexture, BindTextureEXT@8, BindTexture, BindTexture@8) GL_STUB_ALIAS(DrawArraysEXT, _gloffset_DrawArrays, DrawArraysEXT@12, DrawArrays, DrawArrays@12) @@ -1091,6 +1089,7 @@ GLNAME(gl_dispatch_functions_start): GL_STUB_ALIAS(MultiTexCoord4iv, _gloffset_MultiTexCoord4ivARB, MultiTexCoord4iv@8, MultiTexCoord4ivARB, MultiTexCoord4ivARB@8) GL_STUB_ALIAS(MultiTexCoord4s, _gloffset_MultiTexCoord4sARB, MultiTexCoord4s@20, MultiTexCoord4sARB, MultiTexCoord4sARB@20) GL_STUB_ALIAS(MultiTexCoord4sv, _gloffset_MultiTexCoord4svARB, MultiTexCoord4sv@8, MultiTexCoord4svARB, MultiTexCoord4svARB@8) + GL_STUB_ALIAS(StencilOpSeparateATI, _gloffset_StencilOpSeparate, StencilOpSeparateATI@16, StencilOpSeparate, StencilOpSeparate@16) GL_STUB_ALIAS(LoadTransposeMatrixd, _gloffset_LoadTransposeMatrixdARB, LoadTransposeMatrixd@4, LoadTransposeMatrixdARB, LoadTransposeMatrixdARB@4) GL_STUB_ALIAS(LoadTransposeMatrixf, _gloffset_LoadTransposeMatrixfARB, LoadTransposeMatrixf@4, LoadTransposeMatrixfARB, LoadTransposeMatrixfARB@4) GL_STUB_ALIAS(MultTransposeMatrixd, _gloffset_MultTransposeMatrixdARB, MultTransposeMatrixd@4, MultTransposeMatrixdARB, MultTransposeMatrixdARB@4) -- cgit v1.2.3 From 0b20d1b9b5e0514a68ab460d748753d29df2e70b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 4 Apr 2008 13:18:09 +0100 Subject: draw: move code to run pipeline from pt to new file Add facility for draw_vbuf.c to reset these vertex ids on flushes. Pre-initialize vertex ids correctly. --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/SConscript | 1 + src/gallium/auxiliary/draw/draw_context.c | 3 +- src/gallium/auxiliary/draw/draw_private.h | 14 ++ .../auxiliary/draw/draw_pt_fetch_pipeline.c | 114 ++------------- src/gallium/auxiliary/draw/draw_pt_pipeline.c | 162 +++++++++++++++++++++ 6 files changed, 196 insertions(+), 99 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_pipeline.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index b28e516396..28262a92c6 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -19,6 +19,7 @@ C_SOURCES = \ draw_pt_vcache.c \ draw_pt_fetch_emit.c \ draw_pt_fetch_pipeline.c \ + draw_pt_pipeline.c \ draw_pt_elts.c \ draw_prim.c \ draw_pstipple.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 9ca4197441..52107912f5 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -18,6 +18,7 @@ draw = env.ConvenienceLibrary( 'draw_pt_vcache.c', 'draw_pt_fetch_emit.c', 'draw_pt_fetch_pipeline.c', + 'draw_pt_pipeline.c', 'draw_pt_elts.c', 'draw_prim.c', 'draw_pstipple.c', diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index d0d5f66b37..470c1c571b 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -434,7 +434,8 @@ void draw_reset_vertex_ids(struct draw_context *draw) stage = stage->next; } - draw_vertex_cache_reset_vertex_ids(draw); + draw_vertex_cache_reset_vertex_ids(draw); /* going away soon */ + draw_pt_reset_vertex_ids(draw); } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 0c9f9c2e03..48545af9e2 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -227,6 +227,12 @@ struct draw_context struct draw_pt_front_end *vcache; } front; + struct { + char *verts; + unsigned vertex_stride; + unsigned vertex_count; + } pipeline; + } pt; boolean flushing; @@ -386,6 +392,14 @@ boolean draw_pt_arrays( struct draw_context *draw, unsigned start, unsigned count ); +void draw_pt_reset_vertex_ids( struct draw_context *draw ); +void draw_pt_run_pipeline( struct draw_context *draw, + unsigned prim, + char *verts, + unsigned vertex_stride, + unsigned vertex_count, + const ushort *elts, + unsigned count ); /* Prototype/hack (DEPRECATED) diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c index 94e7d01be4..0ddb400f7e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c @@ -129,7 +129,13 @@ static void emit_R32G32B32A32_FLOAT( const float *attrib, static void emit_header( const float *attrib, float **out ) { - (*out)[0] = 0; + struct vertex_header *header = (struct vertex_header *) (*out); + + header->clipmask = 0; + header->edgeflag = 1; + header->pad = 0; + header->vertex_id = UNDEFINED_VERTEX_ID; + (*out)[1] = 0; (*out)[2] = 0; (*out)[3] = 0; @@ -231,99 +237,6 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, -/** - * Add a point to the primitive queue. - * \param i0 index into user's vertex arrays - */ -static void do_point( struct draw_context *draw, - const char *v0 ) -{ - struct prim_header prim; - - prim.reset_line_stipple = 0; - prim.edgeflags = 1; - prim.pad = 0; - prim.v[0] = (struct vertex_header *)v0; - - draw->pipeline.first->point( draw->pipeline.first, &prim ); -} - - -/** - * Add a line to the primitive queue. - * \param i0 index into user's vertex arrays - * \param i1 index into user's vertex arrays - */ -static void do_line( struct draw_context *draw, - const char *v0, - const char *v1 ) -{ - struct prim_header prim; - - prim.reset_line_stipple = 1; /* fixme */ - prim.edgeflags = 1; - prim.pad = 0; - prim.v[0] = (struct vertex_header *)v0; - prim.v[1] = (struct vertex_header *)v1; - - draw->pipeline.first->line( draw->pipeline.first, &prim ); -} - -/** - * Add a triangle to the primitive queue. - */ -static void do_triangle( struct draw_context *draw, - char *v0, - char *v1, - char *v2 ) -{ - struct prim_header prim; - -// _mesa_printf("tri %d %d %d\n", i0, i1, i2); - prim.reset_line_stipple = 1; - prim.edgeflags = ~0; - prim.pad = 0; - prim.v[0] = (struct vertex_header *)v0; - prim.v[1] = (struct vertex_header *)v1; - prim.v[2] = (struct vertex_header *)v2; - - draw->pipeline.first->tri( draw->pipeline.first, &prim ); -} - - -static void run_pipeline( struct fetch_pipeline_middle_end *fpme, - char *verts, - const ushort *elts, - unsigned count ) -{ - struct draw_context *draw = fpme->draw; - unsigned stride = fpme->pipeline_vertex_size; - unsigned i; - - switch (fpme->prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i++) - do_point( draw, - verts + stride * elts[i] ); - break; - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) - do_line( draw, - verts + stride * elts[i+0], - verts + stride * elts[i+1]); - break; - case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) - do_triangle( draw, - verts + stride * elts[i+0], - verts + stride * elts[i+1], - verts + stride * elts[i+2]); - break; - } -} - - - static void fetch_pipeline_run( struct draw_pt_middle_end *middle, const unsigned *fetch_elts, @@ -351,10 +264,15 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, fetch_count ); - run_pipeline( fpme, - pipeline_verts, - draw_elts, - draw_count ); + /* Run the pipeline + */ + draw_pt_run_pipeline( fpme->draw, + fpme->prim, + pipeline_verts, + fpme->pipeline_vertex_size, + fetch_count, + draw_elts, + draw_count ); /* Done -- that was easy, wasn't it: diff --git a/src/gallium/auxiliary/draw/draw_pt_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_pipeline.c new file mode 100644 index 0000000000..6e46d3925f --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_pipeline.c @@ -0,0 +1,162 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" + + +/** + * Add a point to the primitive queue. + * \param i0 index into user's vertex arrays + */ +static void do_point( struct draw_context *draw, + const char *v0 ) +{ + struct prim_header prim; + + prim.reset_line_stipple = 0; + prim.edgeflags = 1; + prim.pad = 0; + prim.v[0] = (struct vertex_header *)v0; + + draw->pipeline.first->point( draw->pipeline.first, &prim ); +} + + +/** + * Add a line to the primitive queue. + * \param i0 index into user's vertex arrays + * \param i1 index into user's vertex arrays + */ +static void do_line( struct draw_context *draw, + const char *v0, + const char *v1 ) +{ + struct prim_header prim; + + prim.reset_line_stipple = 1; /* fixme */ + prim.edgeflags = 1; + prim.pad = 0; + prim.v[0] = (struct vertex_header *)v0; + prim.v[1] = (struct vertex_header *)v1; + + draw->pipeline.first->line( draw->pipeline.first, &prim ); +} + +/** + * Add a triangle to the primitive queue. + */ +static void do_triangle( struct draw_context *draw, + char *v0, + char *v1, + char *v2 ) +{ + struct prim_header prim; + +// _mesa_printf("tri %d %d %d\n", i0, i1, i2); + prim.reset_line_stipple = 1; + prim.edgeflags = ~0; + prim.pad = 0; + prim.v[0] = (struct vertex_header *)v0; + prim.v[1] = (struct vertex_header *)v1; + prim.v[2] = (struct vertex_header *)v2; + + draw->pipeline.first->tri( draw->pipeline.first, &prim ); +} + + + +void draw_pt_reset_vertex_ids( struct draw_context *draw ) +{ + unsigned i; + char *verts = draw->pt.pipeline.verts; + unsigned stride = draw->pt.pipeline.vertex_stride; + + for (i = 0; i < draw->pt.pipeline.vertex_count; i++) { + ((struct vertex_header *)verts)->vertex_id = UNDEFINED_VERTEX_ID; + verts += stride; + } +} + + +/* Code to run the pipeline on a fairly arbitary collection of vertices. + * + * Vertex headers must be pre-initialized with the + * UNDEFINED_VERTEX_ID, this code will cause that id to become + * overwritten, so it may have to be reset if there is the intention + * to reuse the vertices. + * + * This code provides a callback to reset the vertex id's which the + * draw_vbuf.c code uses when it has to perform a flush. + */ +void draw_pt_run_pipeline( struct draw_context *draw, + unsigned prim, + char *verts, + unsigned stride, + unsigned vertex_count, + const ushort *elts, + unsigned count ) +{ + unsigned i; + + draw->pt.pipeline.verts = verts; + draw->pt.pipeline.vertex_stride = stride; + draw->pt.pipeline.vertex_count = vertex_count; + + switch (prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i++) + do_point( draw, + verts + stride * elts[i] ); + break; + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) + do_line( draw, + verts + stride * elts[i+0], + verts + stride * elts[i+1]); + break; + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) + do_triangle( draw, + verts + stride * elts[i+0], + verts + stride * elts[i+1], + verts + stride * elts[i+2]); + break; + } + + draw->pt.pipeline.verts = NULL; + draw->pt.pipeline.vertex_count = 0; +} + -- cgit v1.2.3 From 7b5303adbf5965510061a21f40d40c931add73eb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 4 Apr 2008 16:17:33 +0100 Subject: gallium: Call st_flush() instead of st_gl_flush() in st_finish(). This is enough for the current purpose of st_finish(), which is to wait for things to settle down before context teardown. --- src/mesa/state_tracker/st_cb_flush.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 1de3676bda..44bf77ae71 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -107,7 +107,7 @@ void st_finish( struct st_context *st ) { struct pipe_fence_handle *fence = NULL; - st_gl_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence); + st_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence); st->pipe->winsys->fence_finish(st->pipe->winsys, fence, 0); st->pipe->winsys->fence_reference(st->pipe->winsys, &fence, NULL); -- cgit v1.2.3 From 1d6877b32642d718fb7b29eca647f4d1dd0f99bb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 4 Apr 2008 16:21:42 +0100 Subject: gallium: Revert st_gl_flush() changes from when the **fence argument was added. As st_gl_flush() isn't used by st_finish() anymore, it doesn't have to make sure pipe->flush() always gets called. --- src/mesa/state_tracker/st_cb_flush.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 44bf77ae71..c90bedd3e5 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -60,7 +60,7 @@ static void st_gl_flush( struct st_context *st, uint pipeFlushFlags, { GLframebuffer *fb = st->ctx->DrawBuffer; - st_flush( st, pipeFlushFlags, fence ); + FLUSH_VERTICES(st->ctx, 0); if (!fb) return; @@ -85,6 +85,15 @@ static void st_gl_flush( struct st_context *st, uint pipeFlushFlags, = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); struct pipe_surface *front_surf = strb->surface; + /* If we aren't rendering to the frontbuffer, this is a noop. + * This should be uncontroversial for glFlush, though people may + * feel more strongly about glFinish. + * + * Additionally, need to make sure that the frontbuffer_dirty + * flag really gets set on frontbuffer rendering. + */ + st->pipe->flush( st->pipe, pipeFlushFlags, fence ); + /* Hook for copying "fake" frontbuffer if necessary: */ st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf, -- cgit v1.2.3 From 84501e68f6294370d6f2f6aec4e7eab57bcc0e72 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 4 Apr 2008 17:02:20 +0100 Subject: gallium: Handle client-supplied edgeflags. Also, implement support in the draw module. We were hardwiring these to one for quite a long time... Currently using a draw_set_edgeflags() function, may be better to push the argument into the draw_arrays() function. TBD. --- src/gallium/auxiliary/draw/draw_context.c | 16 +++++++ src/gallium/auxiliary/draw/draw_context.h | 3 ++ src/gallium/auxiliary/draw/draw_private.h | 13 ++---- .../auxiliary/draw/draw_pt_fetch_pipeline.c | 49 +++++++++++++++------- src/gallium/auxiliary/draw/draw_pt_pipeline.c | 8 ++-- src/gallium/auxiliary/draw/draw_vertex_cache.c | 2 +- src/gallium/include/pipe/p_context.h | 8 ++++ 7 files changed, 71 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 470c1c571b..b3c65c90d6 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -444,3 +444,19 @@ void draw_set_render( struct draw_context *draw, { draw->render = render; } + +void draw_set_edgeflags( struct draw_context *draw, + const unsigned *edgeflag ) +{ + draw->user.edgeflag = edgeflag; +} + + +boolean draw_get_edgeflag( struct draw_context *draw, + unsigned idx ) +{ + if (draw->user.edgeflag) + return (draw->user.edgeflag[idx/32] & (1 << (idx%32))) != 0; + else + return 1; +} diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index 84bae3bd78..c7ac32b452 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -155,6 +155,9 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw, void draw_set_mapped_constant_buffer(struct draw_context *draw, const void *buffer); +void draw_set_edgeflags( struct draw_context *draw, + const unsigned *edgeflag ); + /*********************************************************************** * draw_prim.c diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 48545af9e2..4d056f6dba 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -250,6 +250,8 @@ struct draw_context /* user-space vertex data, buffers */ struct { + const unsigned *edgeflag; + /** vertex element/index buffer (ex: glDrawElements) */ const void *elts; /** bytes per index (0, 1, 2 or 4) */ @@ -402,15 +404,6 @@ void draw_pt_run_pipeline( struct draw_context *draw, unsigned count ); -/* Prototype/hack (DEPRECATED) - */ -boolean -draw_passthrough_arrays(struct draw_context *draw, - unsigned prim, - unsigned start, - unsigned count); - - #define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ #define DRAW_FLUSH_PRIM_QUEUE 0x2 #define DRAW_FLUSH_VERTEX_CACHE 0x4 @@ -420,6 +413,8 @@ draw_passthrough_arrays(struct draw_context *draw, void draw_do_flush( struct draw_context *draw, unsigned flags ); +boolean draw_get_edgeflag( struct draw_context *draw, + unsigned idx ); /** diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c index 0ddb400f7e..ba95420b72 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c @@ -72,6 +72,8 @@ struct fetch_pipeline_middle_end { struct draw_pt_middle_end base; struct draw_context *draw; + void (*header)( const unsigned *edgeflag, unsigned count, float **out); + struct { const ubyte *ptr; unsigned pitch; @@ -85,12 +87,6 @@ struct fetch_pipeline_middle_end { }; -static void fetch_NULL( const void *from, - float *attrib ) -{ -} - - static void emit_R32_FLOAT( const float *attrib, float **out ) @@ -126,8 +122,9 @@ static void emit_R32G32B32A32_FLOAT( const float *attrib, (*out) += 4; } -static void emit_header( const float *attrib, - float **out ) +static void header( const unsigned *edgeflag, + unsigned idx, + float **out ) { struct vertex_header *header = (struct vertex_header *) (*out); @@ -143,6 +140,26 @@ static void emit_header( const float *attrib, (*out) += 5; } + +static void header_ef( const unsigned *edgeflag, + unsigned idx, + float **out ) +{ + struct vertex_header *header = (struct vertex_header *) (*out); + + header->clipmask = 0; + header->edgeflag = (edgeflag[idx/32] & (1 << (idx%32))) != 0; + header->pad = 0; + header->vertex_id = UNDEFINED_VERTEX_ID; + + (*out)[1] = 0; + (*out)[2] = 0; + (*out)[3] = 0; + (*out)[3] = 1; + (*out) += 5; +} + + /** * General-purpose fetch from user's vertex arrays, emit to driver's * vertex buffer. @@ -155,12 +172,15 @@ fetch_store_general( struct fetch_pipeline_middle_end *fpme, const unsigned *fetch_elts, unsigned count ) { + const unsigned *edgeflag = fpme->draw->user.edgeflag; float *out = (float *)out_ptr; uint i, j; for (i = 0; i < count; i++) { unsigned elt = fetch_elts[i]; + fpme->header( edgeflag, i, &out ); + for (j = 0; j < fpme->nr_fetch; j++) { float attrib[4]; const ubyte *from = (fpme->fetch[j].ptr + @@ -200,12 +220,11 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, /* Emit the vertex header and empty clipspace coord field: */ - { - fpme->fetch[nr].ptr = NULL; - fpme->fetch[nr].pitch = 0; - fpme->fetch[nr].fetch = fetch_NULL; - fpme->fetch[nr].emit = emit_header; - nr++; + if (draw->user.edgeflag) { + fpme->header = header_ef; + } + else { + fpme->header = header; } @@ -232,7 +251,7 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, } fpme->nr_fetch = nr; - fpme->pipeline_vertex_size = (5 + (nr-1) * 4) * sizeof(float); + fpme->pipeline_vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); } diff --git a/src/gallium/auxiliary/draw/draw_pt_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_pipeline.c index 6e46d3925f..942df518eb 100644 --- a/src/gallium/auxiliary/draw/draw_pt_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_pipeline.c @@ -86,12 +86,14 @@ static void do_triangle( struct draw_context *draw, struct prim_header prim; // _mesa_printf("tri %d %d %d\n", i0, i1, i2); - prim.reset_line_stipple = 1; - prim.edgeflags = ~0; - prim.pad = 0; prim.v[0] = (struct vertex_header *)v0; prim.v[1] = (struct vertex_header *)v1; prim.v[2] = (struct vertex_header *)v2; + prim.reset_line_stipple = 1; + prim.edgeflags = ((prim.v[0]->edgeflag) | + (prim.v[1]->edgeflag << 1) | + (prim.v[2]->edgeflag << 2)); + prim.pad = 0; draw->pipeline.first->tri( draw->pipeline.first, &prim ); } diff --git a/src/gallium/auxiliary/draw/draw_vertex_cache.c b/src/gallium/auxiliary/draw/draw_vertex_cache.c index 161b247d4e..c0248979e2 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_cache.c +++ b/src/gallium/auxiliary/draw/draw_vertex_cache.c @@ -101,7 +101,7 @@ static struct vertex_header *get_vertex( struct draw_context *draw, draw->vs.queue[out].elt = i; draw->vs.queue[out].vertex->clipmask = 0; - draw->vs.queue[out].vertex->edgeflag = 1; /*XXX use user's edge flag! */ + draw->vs.queue[out].vertex->edgeflag = draw_get_edgeflag(draw, i); draw->vs.queue[out].vertex->pad = 0; draw->vs.queue[out].vertex->vertex_id = UNDEFINED_VERTEX_ID; diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 324f70185a..f3a9c2cd8b 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -57,6 +57,14 @@ struct pipe_context { void (*destroy)( struct pipe_context * ); + + /* Possible interface for setting edgeflags. These aren't really + * vertex elements, so don't fit there. + */ + void (*set_edgeflags)( struct pipe_context *, + const unsigned *bitfield ); + + /** * VBO drawing (return false on fallbacks (temporary??)) */ -- cgit v1.2.3 From 72f2c55069f167a46560005931382e3b472f92ed Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Apr 2008 11:20:44 -0600 Subject: gallium: make sure to set the SamplersUsed field for bitmap/drawpixels shaders Also, make sure that field is copied/updated in the program clone and combine functions. Without this we weren't getting SAMP declarations in the TGSI shaders. --- src/mesa/shader/program.c | 2 ++ src/mesa/state_tracker/st_atom_pixeltransfer.c | 1 + src/mesa/state_tracker/st_cb_bitmap.c | 1 + src/mesa/state_tracker/st_cb_drawpixels.c | 1 + 4 files changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 09a8494bd3..3069b04836 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -361,6 +361,7 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog) prog->NumInstructions); clone->InputsRead = prog->InputsRead; clone->OutputsWritten = prog->OutputsWritten; + clone->SamplersUsed = prog->SamplersUsed; memcpy(clone->TexturesUsed, prog->TexturesUsed, sizeof(prog->TexturesUsed)); if (prog->Parameters) @@ -537,6 +538,7 @@ _mesa_combine_programs(GLcontext *ctx, } newProg->InputsRead = progA->InputsRead | inputsB; newProg->OutputsWritten = progB->OutputsWritten; + newProg->SamplersUsed = progA->SamplersUsed | progB->SamplersUsed; } else { /* vertex program */ diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index 1e3effd549..6410e7cb24 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -130,6 +130,7 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) ic++; fp->Base.InputsRead = (1 << FRAG_ATTRIB_TEX0); fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR); + fp->Base.SamplersUsed = 0x1; /* sampler 0 (bit 0) is used */ /* MAD colorTemp, colorTemp, scale, bias; */ if (key->scaleAndBias) { diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 0872f2bd28..59f3c385df 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -148,6 +148,7 @@ make_bitmap_fragment_program(GLcontext *ctx) p->InputsRead = FRAG_BIT_TEX0; p->OutputsWritten = 0x0; + p->SamplersUsed = 0x1; /* sampler 0 (bit 0) is used */ stfp = (struct st_fragment_program *) p; stfp->Base.UsesKill = GL_TRUE; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c7796cfb6a..c57e05312a 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -212,6 +212,7 @@ make_fragment_shader_z(struct st_context *st) p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0; p->OutputsWritten = (1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR); + p->SamplersUsed = 0x1; /* sampler 0 (bit 0) is used */ stfp = (struct st_fragment_program *) p; st_translate_fragment_program(st, stfp, NULL); -- cgit v1.2.3 From 5ffc5cce1507fd407399911abefeea988a69394e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Apr 2008 12:24:01 -0600 Subject: gallium: new debug code, disabled --- src/gallium/auxiliary/draw/draw_clip.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_clip.c b/src/gallium/auxiliary/draw/draw_clip.c index 200152ecab..fd1c71152e 100644 --- a/src/gallium/auxiliary/draw/draw_clip.c +++ b/src/gallium/auxiliary/draw/draw_clip.c @@ -181,6 +181,21 @@ static void emit_poly( struct draw_stage *stage, (header.v[1]->edgeflag << 1) | (header.v[2]->edgeflag << 2)); + if (0) { + const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + uint j, k; + printf("Clipped tri:\n"); + for (j = 0; j < 3; j++) { + for (k = 0; k < vs->info.num_outputs; k++) { + printf(" Vert %d: Attr %d: %f %f %f %f\n", j, k, + header.v[j]->data[k][0], + header.v[j]->data[k][1], + header.v[j]->data[k][2], + header.v[j]->data[k][3]); + } + } + } + stage->next->tri( stage->next, &header ); header.v[1]->edgeflag = tmp1; -- cgit v1.2.3 From ac2ca39f8f6024b13b0ddbfe767f9dbc9a528e6c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Apr 2008 12:24:25 -0600 Subject: gallium: add new call to st_flush_bitmap_cache() to fix recent regression --- src/mesa/state_tracker/st_cb_flush.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index c90bedd3e5..c7efa40e38 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -60,6 +60,8 @@ static void st_gl_flush( struct st_context *st, uint pipeFlushFlags, { GLframebuffer *fb = st->ctx->DrawBuffer; + st_flush_bitmap_cache(st); + FLUSH_VERTICES(st->ctx, 0); if (!fb) -- cgit v1.2.3 From f32c51277498887b348133ebcd947dbc8acce756 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 4 Apr 2008 21:10:07 +0200 Subject: gallium: state tracker fixes for compressed textures --- src/mesa/state_tracker/st_cb_texture.c | 1 + src/mesa/state_tracker/st_format.c | 8 ++++++++ 2 files changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index d0bc234f36..43b43ec14c 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1519,6 +1519,7 @@ st_init_texture_functions(struct dd_function_table *functions) /* compressed texture functions */ functions->CompressedTexImage2D = st_CompressedTexImage2D; functions->GetCompressedTexImage = st_GetCompressedTexImage; + functions->CompressedTextureSize = _mesa_compressed_texture_size; functions->NewTextureObject = st_NewTextureObject; functions->NewTextureImage = st_NewTextureImage; diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 761645c105..20defd5e16 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -276,6 +276,14 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) return PIPE_FORMAT_Z24S8_UNORM; case MESA_FORMAT_YCBCR: return PIPE_FORMAT_YCBCR; + case MESA_FORMAT_RGB_DXT1: + return PIPE_FORMAT_DXT1_RGB; + case MESA_FORMAT_RGBA_DXT1: + return PIPE_FORMAT_DXT1_RGBA; + case MESA_FORMAT_RGBA_DXT3: + return PIPE_FORMAT_DXT3_RGBA; + case MESA_FORMAT_RGBA_DXT5: + return PIPE_FORMAT_DXT5_RGBA; default: assert(0); return 0; -- cgit v1.2.3 From 74a6a46b749668032c17f8fb2312e927f214768c Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 4 Apr 2008 16:44:09 -0600 Subject: gallium: adjust the code in update_textures() to look more like update_samplers() --- src/mesa/state_tracker/st_atom_texture.c | 15 +++++++-------- 1 file changed, 7 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index e1528c4d7b..01c07dc26d 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -47,19 +47,18 @@ static void update_textures(struct st_context *st) { - /* ST_NEW_FRAGMENT_PROGRAM - */ struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current; - GLuint unit; + GLuint su; st->state.num_textures = 0; - for (unit = 0; unit < st->ctx->Const.MaxTextureCoordUnits; unit++) { - const GLuint su = fprog->Base.SamplerUnits[unit]; + for (su = 0; su < st->ctx->Const.MaxTextureCoordUnits; su++) { struct pipe_texture *pt = NULL; if (fprog->Base.SamplersUsed & (1 << su)) { - struct gl_texture_object *texObj = st->ctx->Texture.Unit[su]._Current; + const GLuint texUnit = fprog->Base.SamplerUnits[su]; + struct gl_texture_object *texObj + = st->ctx->Texture.Unit[texUnit]._Current; struct st_texture_object *stObj = st_texture_object(texObj); if (texObj) { @@ -71,13 +70,13 @@ update_textures(struct st_context *st) continue; } - st->state.num_textures = unit + 1; + st->state.num_textures = su + 1; } pt = st_get_stobj_texture(stObj); } - pipe_texture_reference(&st->state.sampler_texture[unit], pt); + pipe_texture_reference(&st->state.sampler_texture[su], pt); } st->pipe->set_sampler_textures(st->pipe, st->state.num_textures, -- cgit v1.2.3 From b53d6ae4be955c701695e093ea4457ab1c9c0ed7 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 4 Apr 2008 18:28:14 -0600 Subject: gallium: always pass size=4 to make_immediate() Mesa always packs 4 immediates into each parameter/const buffer slot. I think we were just getting lucky with this as it was. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index f3cfda0bfb..578fd2ecb0 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -833,8 +833,7 @@ tgsi_translate_mesa_program( i++) { if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { struct tgsi_full_immediate fullimm - = make_immediate(program->Parameters->ParameterValues[i], - program->Parameters->Parameters[i].Size); + = make_immediate(program->Parameters->ParameterValues[i], 4); ti += tgsi_build_full_immediate(&fullimm, &tokens[ti], header, -- cgit v1.2.3 From b8cc9e88e067a5cd6a1acbae6d6a314e9165652f Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 4 Apr 2008 18:57:40 -0600 Subject: mesa: new functions for managing list/index of uniforms --- src/mesa/shader/prog_uniform.c | 157 +++++++++++++++++++++++++++++++++++++++++ src/mesa/shader/prog_uniform.h | 91 ++++++++++++++++++++++++ 2 files changed, 248 insertions(+) create mode 100644 src/mesa/shader/prog_uniform.c create mode 100644 src/mesa/shader/prog_uniform.h (limited to 'src') diff --git a/src/mesa/shader/prog_uniform.c b/src/mesa/shader/prog_uniform.c new file mode 100644 index 0000000000..20e004b350 --- /dev/null +++ b/src/mesa/shader/prog_uniform.c @@ -0,0 +1,157 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file prog_uniform.c + * Shader uniform functions. + * \author Brian Paul + */ + +#include "main/imports.h" +#include "main/mtypes.h" +#include "prog_uniform.h" + + +struct gl_uniform_list * +_mesa_new_uniform_list(void) +{ + return CALLOC_STRUCT(gl_uniform_list); +} + + +void +_mesa_free_uniform_list(struct gl_uniform_list *list) +{ + GLuint i; + for (i = 0; i < list->NumUniforms; i++) { + _mesa_free((void *) list->Uniforms[i].Name); + } + _mesa_free(list->Uniforms); + _mesa_free(list); +} + + +GLboolean +_mesa_append_uniform(struct gl_uniform_list *list, + const char *name, GLenum target, GLuint progPos) +{ + const GLuint oldNum = list->NumUniforms; + GLint index; + + assert(target == GL_VERTEX_PROGRAM_ARB || + target == GL_FRAGMENT_PROGRAM_ARB); + + index = _mesa_lookup_uniform(list, name); + if (index < 0) { + /* not found - append to list */ + + if (oldNum + 1 > list->Size) { + /* Need to grow the list array (alloc some extra) */ + list->Size += 4; + + /* realloc arrays */ + list->Uniforms = (struct gl_uniform *) + _mesa_realloc(list->Uniforms, + oldNum * sizeof(struct gl_uniform), + list->Size * sizeof(struct gl_uniform)); + } + + if (!list->Uniforms) { + /* out of memory */ + list->NumUniforms = 0; + list->Size = 0; + return GL_FALSE; + } + + list->Uniforms[oldNum].Name = _mesa_strdup(name); + list->Uniforms[oldNum].VertPos = -1; + list->Uniforms[oldNum].FragPos = -1; + index = oldNum; + list->NumUniforms++; + } + + /* update position for the vertex or fragment program */ + if (target == GL_VERTEX_PROGRAM_ARB) { + if (list->Uniforms[index].VertPos != -1) { + /* this uniform is already in the list - that shouldn't happen */ + return GL_FALSE; + } + list->Uniforms[index].VertPos = progPos; + } + else { + if (list->Uniforms[index].FragPos != -1) { + /* this uniform is already in the list - that shouldn't happen */ + return GL_FALSE; + } + list->Uniforms[index].FragPos = progPos; + } + + return GL_TRUE; +} + + +/** + * Return the location/index of the named uniform in the uniform list, + * or -1 if not found. + */ +GLint +_mesa_lookup_uniform(const struct gl_uniform_list *list, const char *name) +{ + GLuint i; + for (i = 0; i < list->NumUniforms; i++) { + if (!_mesa_strcmp(list->Uniforms[i].Name, name)) { + return i; + } + } + return -1; +} + + +GLint +_mesa_longest_uniform_name(const struct gl_uniform_list *list) +{ + GLint max = 0; + GLuint i; + for (i = 0; i < list->NumUniforms; i++) { + GLuint len = _mesa_strlen(list->Uniforms[i].Name); + if (len > max) + max = len; + } + return max; +} + + +void +_mesa_print_uniforms(const struct gl_uniform_list *list) +{ + GLuint i; + printf("Uniform list %p:\n", (void *) list); + for (i = 0; i < list->NumUniforms; i++) { + printf("%d: %s %d %d\n", + i, + list->Uniforms[i].Name, + list->Uniforms[i].VertPos, + list->Uniforms[i].FragPos); + } +} diff --git a/src/mesa/shader/prog_uniform.h b/src/mesa/shader/prog_uniform.h new file mode 100644 index 0000000000..735de28705 --- /dev/null +++ b/src/mesa/shader/prog_uniform.h @@ -0,0 +1,91 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file prog_uniform.c + * Shader uniform functions. + * \author Brian Paul + */ + +#ifndef PROG_UNIFORM_H +#define PROG_UNIFORM_H + +#include "main/mtypes.h" +#include "prog_statevars.h" + + +/** + * Shader program uniform variable. + * The glGetUniformLocation() and glUniform() commands will use this + * information. + * Note that a uniform such as "binormal" might be used in both the + * vertex shader and the fragment shader. When glUniform() is called to + * set the uniform's value, it must be updated in both the vertex and + * fragment shaders. The uniform may be in different locations in the + * two shaders so we keep track of that here. + */ +struct gl_uniform +{ + const char *Name; /**< Null-terminated string */ + GLint VertPos; + GLint FragPos; +#if 0 + GLenum DataType; /**< GL_FLOAT, GL_FLOAT_VEC2, etc */ + GLuint Size; /**< Number of components (1..4) */ +#endif +}; + + +/** + * List of gl_uniforms + */ +struct gl_uniform_list +{ + GLuint Size; /**< allocated size of Uniforms array */ + GLuint NumUniforms; /**< number of uniforms in the array */ + struct gl_uniform *Uniforms; /**< Array [Size] */ +}; + + +extern struct gl_uniform_list * +_mesa_new_uniform_list(void); + +extern void +_mesa_free_uniform_list(struct gl_uniform_list *list); + +extern GLboolean +_mesa_append_uniform(struct gl_uniform_list *list, + const char *name, GLenum target, GLuint progPos); + +extern GLint +_mesa_lookup_uniform(const struct gl_uniform_list *list, const char *name); + +extern GLint +_mesa_longest_uniform_name(const struct gl_uniform_list *list); + +extern void +_mesa_print_uniforms(const struct gl_uniform_list *list); + + +#endif /* PROG_UNIFORM_H */ -- cgit v1.2.3 From bc029247d9d886f4546a4c3a36737d09c488b7f9 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 4 Apr 2008 18:59:21 -0600 Subject: mesa: no longer combine vertex/fragment shader parameters/uniforms GLSL Vertex and fragment shaders now have independent parameter buffers. A new gl_uniform_list is used to keep track of program uniforms and where each uniform is located in each shader's parameter buffer. This makes better use of the space in each buffer and simplifies shader linking. --- src/mesa/main/mtypes.h | 8 +- src/mesa/shader/shader_api.c | 308 ++++++++++++++++++++++--------------- src/mesa/shader/slang/slang_link.c | 192 +++++++---------------- src/mesa/sources | 1 + 4 files changed, 239 insertions(+), 270 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 8e49431a8f..50b22d25bf 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -1868,6 +1868,7 @@ enum register_file /** Vertex and fragment instructions */ struct prog_instruction; struct gl_program_parameter_list; +struct gl_uniform_list; /** @@ -2104,7 +2105,7 @@ struct gl_query_state /** - * A GLSL shader object. + * A GLSL vertex or fragment shader object. */ struct gl_shader { @@ -2122,7 +2123,8 @@ struct gl_shader /** - * A GLSL program object. Basically a linked collection of "shaders". + * A GLSL program object. + * Basically a linked collection of vertex and fragment shaders. */ struct gl_shader_program { @@ -2137,7 +2139,7 @@ struct gl_shader_program /* post-link info: */ struct gl_vertex_program *VertexProgram; /**< Linked vertex program */ struct gl_fragment_program *FragmentProgram; /**< Linked fragment prog */ - struct gl_program_parameter_list *Uniforms; /**< Plus constants, etc */ + struct gl_uniform_list *Uniforms; struct gl_program_parameter_list *Varying; struct gl_program_parameter_list *Attributes; /**< Vertex attributes */ GLboolean LinkStatus; /**< GL_LINK_STATUS */ diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 4cb8bb8ed1..9c419c9903 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -43,6 +43,7 @@ #include "prog_parameter.h" #include "prog_print.h" #include "prog_statevars.h" +#include "prog_uniform.h" #include "shader/shader_api.h" #include "shader/slang/slang_compile.h" #include "shader/slang/slang_link.h" @@ -75,25 +76,25 @@ _mesa_clear_shader_program_data(GLcontext *ctx, struct gl_shader_program *shProg) { if (shProg->VertexProgram) { - if (shProg->VertexProgram->Base.Parameters == shProg->Uniforms) { - /* to prevent a double-free in the next call */ - shProg->VertexProgram->Base.Parameters = NULL; - } + /* Set ptr to NULL since the param list is shared with the + * original/unlinked program. + */ + shProg->VertexProgram->Base.Parameters = NULL; ctx->Driver.DeleteProgram(ctx, &shProg->VertexProgram->Base); shProg->VertexProgram = NULL; } if (shProg->FragmentProgram) { - if (shProg->FragmentProgram->Base.Parameters == shProg->Uniforms) { - /* to prevent a double-free in the next call */ - shProg->FragmentProgram->Base.Parameters = NULL; - } + /* Set ptr to NULL since the param list is shared with the + * original/unlinked program. + */ + shProg->FragmentProgram->Base.Parameters = NULL; ctx->Driver.DeleteProgram(ctx, &shProg->FragmentProgram->Base); shProg->FragmentProgram = NULL; } if (shProg->Uniforms) { - _mesa_free_parameter_list(shProg->Uniforms); + _mesa_free_uniform_list(shProg->Uniforms); shProg->Uniforms = NULL; } @@ -673,39 +674,43 @@ _mesa_get_active_uniform(GLcontext *ctx, GLuint program, GLuint index, GLsizei maxLength, GLsizei *length, GLint *size, GLenum *type, GLchar *nameOut) { - struct gl_shader_program *shProg + const struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, program); - GLuint ind, j; + const struct gl_program *prog; + GLint progPos; if (!shProg) { _mesa_error(ctx, GL_INVALID_VALUE, "glGetActiveUniform"); return; } - if (!shProg->Uniforms || index >= shProg->Uniforms->NumParameters) { + if (!shProg->Uniforms || index >= shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glGetActiveUniform(index)"); return; } - ind = 0; - for (j = 0; j < shProg->Uniforms->NumParameters; j++) { - if (shProg->Uniforms->Parameters[j].Type == PROGRAM_UNIFORM || - shProg->Uniforms->Parameters[j].Type == PROGRAM_SAMPLER) { - if (ind == index) { - /* found it */ - copy_string(nameOut, maxLength, length, - shProg->Uniforms->Parameters[j].Name); - if (size) - *size = shProg->Uniforms->Parameters[j].Size; - if (type) - *type = shProg->Uniforms->Parameters[j].DataType; - return; - } - ind++; + progPos = shProg->Uniforms->Uniforms[index].VertPos; + if (progPos >= 0) { + prog = &shProg->VertexProgram->Base; + } + else { + progPos = shProg->Uniforms->Uniforms[index].FragPos; + if (progPos >= 0) { + prog = &shProg->FragmentProgram->Base; } } - _mesa_error(ctx, GL_INVALID_VALUE, "glGetActiveUniform(index)"); + if (!prog || progPos < 0) + return; /* should never happen */ + + if (nameOut) + copy_string(nameOut, maxLength, length, + prog->Parameters->Parameters[progPos].Name); + if (size) + *size = prog->Parameters->Parameters[progPos].Size; + + if (type) + *type = prog->Parameters->Parameters[progPos].DataType; } @@ -792,14 +797,10 @@ _mesa_get_programiv(GLcontext *ctx, GLuint program, PROGRAM_INPUT) + 1; break; case GL_ACTIVE_UNIFORMS: - *params - = _mesa_num_parameters_of_type(shProg->Uniforms, PROGRAM_UNIFORM) - + _mesa_num_parameters_of_type(shProg->Uniforms, PROGRAM_SAMPLER); + *params = shProg->Uniforms ? shProg->Uniforms->NumUniforms : 0; break; case GL_ACTIVE_UNIFORM_MAX_LENGTH: - *params = MAX2( - _mesa_longest_parameter_name(shProg->Uniforms, PROGRAM_UNIFORM), - _mesa_longest_parameter_name(shProg->Uniforms, PROGRAM_SAMPLER)); + *params = _mesa_longest_uniform_name(shProg->Uniforms); if (*params > 0) (*params)++; /* add one for terminating zero */ break; @@ -896,10 +897,23 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, program); if (shProg) { - GLint i; - if (location >= 0 && location < shProg->Uniforms->NumParameters) { - for (i = 0; i < shProg->Uniforms->Parameters[location].Size; i++) { - params[i] = shProg->Uniforms->ParameterValues[location][i]; + if (location < shProg->Uniforms->NumUniforms) { + GLint progPos, i; + const struct gl_program *prog; + + progPos = shProg->Uniforms->Uniforms[location].VertPos; + if (progPos >= 0) { + prog = &shProg->VertexProgram->Base; + } + else { + progPos = shProg->Uniforms->Uniforms[location].FragPos; + if (progPos >= 0) { + prog = &shProg->FragmentProgram->Base; + } + } + + for (i = 0; i < prog->Parameters->Parameters[progPos].Size; i++) { + params[i] = prog->Parameters->ParameterValues[progPos][i]; } } else { @@ -920,23 +934,10 @@ _mesa_get_uniform_location(GLcontext *ctx, GLuint program, const GLchar *name) { struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, program); - if (shProg) { - GLuint loc; - for (loc = 0; loc < shProg->Uniforms->NumParameters; loc++) { - const struct gl_program_parameter *u - = shProg->Uniforms->Parameters + loc; - /* XXX this is a temporary simplification / short-cut. - * We need to handle things like "e.c[0].b" as seen in the - * GLSL orange book, page 189. - */ - if ((u->Type == PROGRAM_UNIFORM || - u->Type == PROGRAM_SAMPLER) && !strcmp(u->Name, name)) { - return loc; - } - } - } - return -1; + if (!shProg) + return -1; + return _mesa_lookup_uniform(shProg->Uniforms, name); } @@ -1067,6 +1068,78 @@ update_textures_used(struct gl_program *prog) } +/** + * Set the value of a program's uniform variable. + * \param program the program whose uniform to update + * \param location the location/index of the uniform + * \param type the datatype of the uniform + * \param count the number of uniforms to set + * \param elems number of elements per uniform + * \param values the new values + */ +static void +set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, + GLenum type, GLint count, GLint elems, const void *values) +{ + if (program->Parameters->Parameters[location].Type == PROGRAM_SAMPLER) { + /* This controls which texture unit which is used by a sampler */ + GLuint texUnit, sampler; + + /* data type for setting samplers must be int */ + if (type != GL_INT || count != 1) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glUniform(only glUniform1i can be used " + "to set sampler uniforms)"); + return; + } + + sampler = (GLuint) program->Parameters->ParameterValues[location][0]; + texUnit = ((GLuint *) values)[0]; + + /* check that the sampler (tex unit index) is legal */ + if (texUnit >= ctx->Const.MaxTextureImageUnits) { + _mesa_error(ctx, GL_INVALID_VALUE, + "glUniform1(invalid sampler/tex unit index)"); + return; + } + + /* This maps a sampler to a texture unit: */ + program->SamplerUnits[sampler] = texUnit; + update_textures_used(program); + + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + } + else { + /* ordinary uniform variable */ + GLint k, i; + + if (count * elems > program->Parameters->Parameters[location].Size) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(count too large)"); + return; + } + + for (k = 0; k < count; k++) { + GLfloat *uniformVal = program->Parameters->ParameterValues[location + k]; + if (type == GL_INT || + type == GL_INT_VEC2 || + type == GL_INT_VEC3 || + type == GL_INT_VEC4) { + const GLint *iValues = ((const GLint *) values) + k * elems; + for (i = 0; i < elems; i++) { + uniformVal[i] = (GLfloat) iValues[i]; + } + } + else { + const GLfloat *fValues = ((const GLfloat *) values) + k * elems; + for (i = 0; i < elems; i++) { + uniformVal[i] = fValues[i]; + } + } + } + } +} + + /** * Called via ctx->Driver.Uniform(). */ @@ -1075,20 +1148,18 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, const GLvoid *values, GLenum type) { struct gl_shader_program *shProg = ctx->Shader.CurrentProgram; - GLint elems, i, k; + GLint elems; if (!shProg || !shProg->LinkStatus) { _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(program not linked)"); return; } - if (location < 0 || location >= (GLint) shProg->Uniforms->NumParameters) { + if (location < 0 || location >= (GLint) shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniform(location)"); return; } - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - if (count < 0) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniform(count < 0)"); return; @@ -1116,63 +1187,54 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, return; } - if (count * elems > shProg->Uniforms->Parameters[location].Size) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(count too large)"); - return; - } - - if (shProg->Uniforms->Parameters[location].Type == PROGRAM_SAMPLER) { - /* This controls which texture unit which is used by a sampler */ - GLuint texUnit, sampler; - - /* data type for setting samplers must be int */ - if (type != GL_INT || count != 1) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glUniform(only glUniform1i can be used " - "to set sampler uniforms)"); - return; - } - - sampler = (GLuint) shProg->Uniforms->ParameterValues[location][0]; - texUnit = ((GLuint *) values)[0]; + FLUSH_VERTICES(ctx, _NEW_PROGRAM); - /* check that the sampler (tex unit index) is legal */ - if (texUnit >= ctx->Const.MaxTextureImageUnits) { - _mesa_error(ctx, GL_INVALID_VALUE, - "glUniform1(invalid sampler/tex unit index)"); - return; + /* A uniform var may be used by both a vertex shader and a fragment + * shader. We may need to update one or both shader's uniform here: + */ + if (shProg->VertexProgram) { + GLint loc = shProg->Uniforms->Uniforms[location].VertPos; + if (loc >= 0) { + set_program_uniform(ctx, &shProg->VertexProgram->Base, + loc, type, count, elems, values); } + } - if (shProg->VertexProgram) { - shProg->VertexProgram->Base.SamplerUnits[sampler] = texUnit; - update_textures_used(&shProg->VertexProgram->Base); - } - if (shProg->FragmentProgram) { - shProg->FragmentProgram->Base.SamplerUnits[sampler] = texUnit; - update_textures_used(&shProg->FragmentProgram->Base); + if (shProg->FragmentProgram) { + GLint loc = shProg->Uniforms->Uniforms[location].FragPos; + if (loc >= 0) { + set_program_uniform(ctx, &shProg->FragmentProgram->Base, + loc, type, count, elems, values); } + } +} - FLUSH_VERTICES(ctx, _NEW_TEXTURE); +static void +set_program_uniform_matrix(GLcontext *ctx, struct gl_program *program, + GLuint location, GLuint rows, GLuint cols, + GLboolean transpose, const GLfloat *values) +{ + /* + * Note: the _columns_ of a matrix are stored in program registers, not + * the rows. + */ + /* XXXX need to test 3x3 and 2x2 matrices... */ + if (transpose) { + GLuint row, col; + for (col = 0; col < cols; col++) { + GLfloat *v = program->Parameters->ParameterValues[location + col]; + for (row = 0; row < rows; row++) { + v[row] = values[row * cols + col]; + } + } } else { - /* ordinary uniform variable */ - for (k = 0; k < count; k++) { - GLfloat *uniformVal = shProg->Uniforms->ParameterValues[location + k]; - if (type == GL_INT || - type == GL_INT_VEC2 || - type == GL_INT_VEC3 || - type == GL_INT_VEC4) { - const GLint *iValues = ((const GLint *) values) + k * elems; - for (i = 0; i < elems; i++) { - uniformVal[i] = (GLfloat) iValues[i]; - } - } - else { - const GLfloat *fValues = ((const GLfloat *) values) + k * elems; - for (i = 0; i < elems; i++) { - uniformVal[i] = fValues[i]; - } + GLuint row, col; + for (col = 0; col < cols; col++) { + GLfloat *v = program->Parameters->ParameterValues[location + col]; + for (row = 0; row < rows; row++) { + v[row] = values[col * rows + row]; } } } @@ -1193,7 +1255,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, "glUniformMatrix(program not linked)"); return; } - if (location < 0 || location >= shProg->Uniforms->NumParameters) { + if (location < 0 || location >= shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniformMatrix(location)"); return; } @@ -1204,27 +1266,19 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, FLUSH_VERTICES(ctx, _NEW_PROGRAM); - /* - * Note: the _columns_ of a matrix are stored in program registers, not - * the rows. - */ - /* XXXX need to test 3x3 and 2x2 matrices... */ - if (transpose) { - GLuint row, col; - for (col = 0; col < cols; col++) { - GLfloat *v = shProg->Uniforms->ParameterValues[location + col]; - for (row = 0; row < rows; row++) { - v[row] = values[row * cols + col]; - } + if (shProg->VertexProgram) { + GLint loc = shProg->Uniforms->Uniforms[location].VertPos; + if (loc >= 0) { + set_program_uniform_matrix(ctx, &shProg->VertexProgram->Base, + loc, rows, cols, transpose, values); } } - else { - GLuint row, col; - for (col = 0; col < cols; col++) { - GLfloat *v = shProg->Uniforms->ParameterValues[location + col]; - for (row = 0; row < rows; row++) { - v[row] = values[col * rows + row]; - } + + if (shProg->FragmentProgram) { + GLint loc = shProg->Uniforms->Uniforms[location].FragPos; + if (loc >= 0) { + set_program_uniform_matrix(ctx, &shProg->FragmentProgram->Base, + loc, rows, cols, transpose, values); } } } diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 390ae56aa5..addff20421 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -37,12 +37,17 @@ #include "shader/prog_parameter.h" #include "shader/prog_print.h" #include "shader/prog_statevars.h" +#include "shader/prog_uniform.h" #include "shader/shader_api.h" #include "slang_link.h" - +/** + * Linking varying vars involves rearranging varying vars so that the + * vertex program's output varyings matches the order of the fragment + * program's input varyings. + */ static GLboolean link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog) { @@ -132,148 +137,53 @@ link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog) } -static GLboolean -is_uniform(GLuint file) -{ - return (file == PROGRAM_ENV_PARAM || - file == PROGRAM_STATE_VAR || - file == PROGRAM_NAMED_PARAM || - file == PROGRAM_CONSTANT || - file == PROGRAM_SAMPLER || - file == PROGRAM_UNIFORM); -} - - -static GLuint shProg_NumSamplers = 0; /** XXX temporary */ - - -static GLboolean -link_uniform_vars(struct gl_shader_program *shProg, struct gl_program *prog) +/** + * Build the shProg->Uniforms list. + * This is basically a list/index of all uniforms found in either/both of + * the vertex and fragment shaders. + */ +static void +link_uniform_vars(struct gl_shader_program *shProg, + struct gl_program *prog, + GLuint *numSamplers) { - GLuint *map, i; GLuint samplerMap[MAX_SAMPLERS]; + GLuint i; -#if 0 - printf("================ pre link uniforms ===============\n"); - _mesa_print_parameter_list(shProg->Uniforms); -#endif - - map = (GLuint *) malloc(prog->Parameters->NumParameters * sizeof(GLuint)); - if (!map) - return GL_FALSE; - - for (i = 0; i < prog->Parameters->NumParameters; /* incr below*/) { - /* see if this uniform is in the linked uniform list */ + for (i = 0; i < prog->Parameters->NumParameters; i++) { const struct gl_program_parameter *p = prog->Parameters->Parameters + i; - const GLfloat *pVals = prog->Parameters->ParameterValues[i]; - GLint j; - GLint size; - /* sanity check */ - assert(is_uniform(p->Type)); - - /* See if this uniform is already in the linked program's list */ - if (p->Name) { - /* this is a named uniform */ - j = _mesa_lookup_parameter_index(shProg->Uniforms, -1, p->Name); - } - else { - /* this is an unnamed constant */ - /*GLuint swizzle;*/ - ASSERT(p->Type == PROGRAM_CONSTANT); - if (_mesa_lookup_parameter_constant(shProg->Uniforms, pVals, - p->Size, &j, NULL)) { - assert(j >= 0); - } - else { - j = -1; - } - } - - if (j >= 0) { - /* already in linked program's list */ - /* check size XXX check this */ -#if 0 - assert(p->Size == shProg->Uniforms->Parameters[j].Size); -#endif - } - else { - /* not already in linked list */ - switch (p->Type) { - case PROGRAM_ENV_PARAM: - j = _mesa_add_named_parameter(shProg->Uniforms, p->Name, pVals); - break; - case PROGRAM_CONSTANT: - j = _mesa_add_named_constant(shProg->Uniforms, p->Name, pVals, p->Size); - break; - case PROGRAM_STATE_VAR: - j = _mesa_add_state_reference(shProg->Uniforms, p->StateIndexes); - break; - case PROGRAM_UNIFORM: - j = _mesa_add_uniform(shProg->Uniforms, p->Name, p->Size, p->DataType); - break; - case PROGRAM_SAMPLER: - { - GLuint sampNum = shProg_NumSamplers++; - GLuint oldSampNum; - j = _mesa_add_sampler(shProg->Uniforms, p->Name, - p->DataType, sampNum); - oldSampNum = (GLuint) prog->Parameters->ParameterValues[i][0]; - assert(oldSampNum < MAX_SAMPLERS); - samplerMap[oldSampNum] = sampNum; - } - break; - default: - _mesa_problem(NULL, "bad parameter type in link_uniform_vars()"); - return GL_FALSE; - } + /* + * XXX FIX NEEDED HERE + * We should also be adding a uniform if p->Type == PROGRAM_STATE_VAR. + * For example, modelview matrix, light pos, etc. + * Also, we need to update the state-var name-generator code to + * generate GLSL-style names, like "gl_LightSource[0].position". + * Furthermore, we'll need to fix the state-var's size/datatype info. + */ + + if (p->Type == PROGRAM_UNIFORM || + p->Type == PROGRAM_SAMPLER) { + _mesa_append_uniform(shProg->Uniforms, p->Name, prog->Target, i); } - ASSERT(j >= 0); - - size = p->Size; - while (size > 0) { - map[i] = j; - i++; - j++; - size -= 4; + if (p->Type == PROGRAM_SAMPLER) { + /* Allocate a new sampler index */ + GLuint sampNum = *numSamplers; + GLuint oldSampNum = (GLuint) prog->Parameters->ParameterValues[i][0]; + assert(oldSampNum < MAX_SAMPLERS); + samplerMap[oldSampNum] = sampNum; + (*numSamplers)++; } - } -#if 0 - printf("================ post link uniforms ===============\n"); - _mesa_print_parameter_list(shProg->Uniforms); -#endif - -#if 0 - { - GLuint i; - for (i = 0; i < prog->Parameters->NumParameters; i++) { - printf("map[%d] = %d\n", i, map[i]); - } - _mesa_print_parameter_list(shProg->Uniforms); - } -#endif - /* OK, now scan the program/shader instructions looking for uniform vars, + /* OK, now scan the program/shader instructions looking for sampler vars, * replacing the old index with the new index. */ prog->SamplersUsed = 0x0; for (i = 0; i < prog->NumInstructions; i++) { struct prog_instruction *inst = prog->Instructions + i; - GLuint j; - - if (is_uniform(inst->DstReg.File)) { - inst->DstReg.Index = map[ inst->DstReg.Index ]; - } - - for (j = 0; j < 3; j++) { - if (is_uniform(inst->SrcReg[j].File)) { - inst->SrcReg[j].Index = map[ inst->SrcReg[j].Index ]; - } - } - if (_mesa_is_tex_instruction(inst->Opcode)) { /* printf("====== remap sampler from %d to %d\n", @@ -286,9 +196,6 @@ link_uniform_vars(struct gl_shader_program *shProg, struct gl_program *prog) } } - free(map); - - return GL_TRUE; } @@ -476,13 +383,12 @@ _slang_link(GLcontext *ctx, { const struct gl_vertex_program *vertProg; const struct gl_fragment_program *fragProg; + GLuint numSamplers = 0; GLuint i; - shProg_NumSamplers = 0; /** XXX temporary */ - _mesa_clear_shader_program_data(ctx, shProg); - shProg->Uniforms = _mesa_new_parameter_list(); + shProg->Uniforms = _mesa_new_uniform_list(); shProg->Varying = _mesa_new_parameter_list(); /** @@ -519,24 +425,30 @@ _slang_link(GLcontext *ctx, shProg->FragmentProgram = NULL; } + /* link varying vars */ if (shProg->VertexProgram) link_varying_vars(shProg, &shProg->VertexProgram->Base); if (shProg->FragmentProgram) link_varying_vars(shProg, &shProg->FragmentProgram->Base); + /* link uniform vars */ if (shProg->VertexProgram) - link_uniform_vars(shProg, &shProg->VertexProgram->Base); + link_uniform_vars(shProg, &shProg->VertexProgram->Base, &numSamplers); if (shProg->FragmentProgram) - link_uniform_vars(shProg, &shProg->FragmentProgram->Base); + link_uniform_vars(shProg, &shProg->FragmentProgram->Base, &numSamplers); + + /*_mesa_print_uniforms(shProg->Uniforms);*/ - /* The vertex and fragment programs share a common set of uniforms now */ if (shProg->VertexProgram) { - _mesa_free_parameter_list(shProg->VertexProgram->Base.Parameters); - shProg->VertexProgram->Base.Parameters = shProg->Uniforms; + /* Rather than cloning the parameter list here, just share it. + * We need to be careful _mesa_clear_shader_program_data() in + * to avoid double-freeing. + */ + shProg->VertexProgram->Base.Parameters = vertProg->Base.Parameters; } if (shProg->FragmentProgram) { - _mesa_free_parameter_list(shProg->FragmentProgram->Base.Parameters); - shProg->FragmentProgram->Base.Parameters = shProg->Uniforms; + /* see comment just above */ + shProg->FragmentProgram->Base.Parameters = fragProg->Base.Parameters; } if (shProg->VertexProgram) { diff --git a/src/mesa/sources b/src/mesa/sources index d109dce5bc..a4e2026e4b 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -213,6 +213,7 @@ SHADER_SOURCES = \ shader/prog_parameter.c \ shader/prog_print.c \ shader/prog_statevars.c \ + shader/prog_uniform.c \ shader/programopt.c \ shader/shader_api.c \ -- cgit v1.2.3 From c1d26d3dccafed808349c47dc12b94081f956560 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 2 Apr 2008 10:21:24 +0900 Subject: gallium: Use the custom snprintf implementation everywhere (for Win32). Because winddk's implemenation does not handle floats. --- src/gallium/auxiliary/util/p_debug.c | 11 +++-------- src/gallium/include/pipe/p_util.h | 9 +++++++++ 2 files changed, 12 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 5447bbb6f5..090e3b7794 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -43,18 +43,13 @@ #ifdef WIN32 static INLINE void -rpl_EngDebugPrint(const char *format, ...) +_EngDebugPrint(const char *format, ...) { va_list ap; va_start(ap, format); EngDebugPrint("", (PCHAR)format, ap); va_end(ap); } - -int rpl_vsnprintf(char *, size_t, const char *, va_list); -int rpl_snprintf(char *str, size_t size, const char *format, ...); -#define vsnprintf rpl_vsnprintf -#define snprintf rpl_snprintf #endif @@ -65,8 +60,8 @@ void _debug_vprintf(const char *format, va_list ap) /* EngDebugPrint does not handle float point arguments, so we need to use * our own vsnprintf implementation */ char buf[512 + 1]; - rpl_vsnprintf(buf, sizeof(buf), format, ap); - rpl_EngDebugPrint("%s", buf); + vsnprintf(buf, sizeof(buf), format, ap); + _EngDebugPrint("%s", buf); #else /* TODO: Implement debug print for WINCE */ #endif diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 1e7b8181f9..8e3aaee496 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -32,6 +32,7 @@ #include "p_debug.h" #include "p_pointer.h" #include +#include #ifdef __cplusplus @@ -137,6 +138,14 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define GETENV( X ) debug_get_option( X, NULL ) +#ifdef WIN32 +int rpl_vsnprintf(char *, size_t, const char *, va_list); +int rpl_snprintf(char *str, size_t size, const char *format, ...); +#define vsnprintf rpl_vsnprintf +#define snprintf rpl_snprintf +#endif + + /** * Return memory on given byte alignment */ -- cgit v1.2.3 From f1efef809caddff442ed45a59645b3f39498f521 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 5 Apr 2008 09:49:50 +0900 Subject: gallium: Fix typo. --- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 55f32e6816..1ecc7d42ee 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -150,7 +150,7 @@ fenced_buffer_destroy(struct pb_buffer *buf) if (fenced_buf->fence) { struct pipe_winsys *winsys = fenced_list->winsys; - if(winsys->fence_finish(winsys, fenced_buf->fence, 0) != 0) { + if(winsys->fence_signalled(winsys, fenced_buf->fence, 0) != 0) { LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); fenced_list->numDelayed++; } -- cgit v1.2.3 From fdff063343ddfbfb1b2fa921e2efcc2fae35d0ad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 5 Apr 2008 11:29:26 +0900 Subject: gallium: Keep fenced buffers list ordered. This allows to keep the list small without the overhead of full walks. --- .../auxiliary/pipebuffer/pb_buffer_fenced.c | 152 ++++++++++++++------- 1 file changed, 99 insertions(+), 53 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 1ecc7d42ee..24ba61a0b7 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -62,7 +62,6 @@ struct fenced_buffer_list struct pipe_winsys *winsys; size_t numDelayed; - size_t checkDelayed; struct list_head delayed; }; @@ -93,51 +92,93 @@ fenced_buffer(struct pb_buffer *buf) } +static INLINE void +_fenced_buffer_add(struct fenced_buffer *fenced_buf) +{ + struct fenced_buffer_list *fenced_list = fenced_buf->list; + + assert(fenced_buf->fence); + assert(!fenced_buf->head.prev); + assert(!fenced_buf->head.next); + LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); + ++fenced_list->numDelayed; +} + + +/** + * Actually destroy the buffer. + */ +static INLINE void +_fenced_buffer_destroy(struct fenced_buffer *fenced_buf) +{ + struct fenced_buffer_list *fenced_list = fenced_buf->list; + + assert(!fenced_buf->base.base.refcount); + assert(!fenced_buf->fence); + pb_reference(&fenced_buf->buffer, NULL); + FREE(fenced_buf); +} + + +static INLINE void +_fenced_buffer_remove(struct fenced_buffer *fenced_buf) +{ + struct fenced_buffer_list *fenced_list = fenced_buf->list; + struct pipe_winsys *winsys = fenced_list->winsys; + + assert(fenced_buf->fence); + + assert(winsys->fence_signalled(winsys, fenced_buf->fence, 0) == 0); + winsys->fence_reference(winsys, &fenced_buf->fence, NULL); + + assert(fenced_buf->head.prev); + assert(fenced_buf->head.next); + LIST_DEL(&fenced_buf->head); +#ifdef DEBUG + fenced_buf->head.prev = NULL; + fenced_buf->head.next = NULL; +#endif + + assert(fenced_list->numDelayed); + --fenced_list->numDelayed; + + if(!fenced_buf->base.base.refcount) + _fenced_buffer_destroy(fenced_buf); +} + + +/** + * Free as many fenced buffers from the list head as possible. + */ static void _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, int wait) { struct pipe_winsys *winsys = fenced_list->winsys; - struct fenced_buffer *fenced_buf; - struct list_head *list, *prev; - int signaled = -1; - - list = fenced_list->delayed.next; - prev = list->prev; - for (; list != &fenced_list->delayed; list = prev, prev = list->prev) { - - fenced_buf = LIST_ENTRY(struct fenced_buffer, list, head); - - if (signaled != 0) { - if (wait) { - signaled = winsys->fence_finish(winsys, fenced_buf->fence, 0); - } - else { - signaled = winsys->fence_signalled(winsys, fenced_buf->fence, 0); - } + struct list_head *curr, *next; + struct fenced_buffer *fenced_buf; + struct pipe_fence_handle *prev_fence = NULL; + + curr = fenced_list->delayed.next; + next = curr->next; + while(curr != &fenced_list->delayed) { + fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); + + if(fenced_buf->fence != prev_fence) { + int signaled; + if (wait) + signaled = winsys->fence_finish(winsys, fenced_buf->fence, 0); + else + signaled = winsys->fence_signalled(winsys, fenced_buf->fence, 0); + if (signaled != 0) + break; + prev_fence = fenced_buf->fence; } - if (signaled != 0) { -#if 0 - /* XXX: we are assuming that buffers are freed in the same order they - * are fenced which may not always be true... - */ - break; -#else - signaled = -1; - continue; -#endif - } + _fenced_buffer_remove(fenced_buf); - winsys->fence_reference(winsys, &fenced_buf->fence, NULL); - - LIST_DEL(list); - fenced_list->numDelayed--; - - /* Do the delayed destroy: - */ - pb_reference(&fenced_buf->buffer, NULL); - FREE(fenced_buf); + curr = next; + next = curr->next; } } @@ -148,25 +189,29 @@ fenced_buffer_destroy(struct pb_buffer *buf) struct fenced_buffer *fenced_buf = fenced_buffer(buf); struct fenced_buffer_list *fenced_list = fenced_buf->list; + _glthread_LOCK_MUTEX(fenced_list->mutex); + assert(fenced_buf->base.base.refcount == 0); if (fenced_buf->fence) { struct pipe_winsys *winsys = fenced_list->winsys; - if(winsys->fence_signalled(winsys, fenced_buf->fence, 0) != 0) { - LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); - fenced_list->numDelayed++; - } + if(winsys->fence_signalled(winsys, fenced_buf->fence, 0) == 0) { + struct list_head *curr, *prev; + curr = &fenced_buf->head; + prev = curr->prev; + do { + fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); + _fenced_buffer_remove(fenced_buf); + curr = prev; + prev = curr->prev; + } while (curr != &fenced_list->delayed); + } else { - winsys->fence_reference(winsys, &fenced_buf->fence, NULL); - pb_reference(&fenced_buf->buffer, NULL); - FREE(fenced_buf); + /* delay destruction */ } } else { - pb_reference(&fenced_buf->buffer, NULL); - FREE(fenced_buf); + _fenced_buffer_destroy(fenced_buf); } - - if ((fenced_list->numDelayed % fenced_list->checkDelayed) == 0) - _fenced_buffer_list_check_free(fenced_list, 0); + _glthread_UNLOCK_MUTEX(fenced_list->mutex); } @@ -241,7 +286,11 @@ buffer_fence(struct pb_buffer *buf, struct pipe_winsys *winsys = fenced_list->winsys; _glthread_LOCK_MUTEX(fenced_list->mutex); + if (fenced_buf->fence) + _fenced_buffer_remove(fenced_buf); winsys->fence_reference(winsys, &fenced_buf->fence, fence); + if (fenced_buf->fence) + _fenced_buffer_add(fenced_buf); _glthread_UNLOCK_MUTEX(fenced_list->mutex); } @@ -261,9 +310,6 @@ fenced_buffer_list_create(struct pipe_winsys *winsys) fenced_list->numDelayed = 0; - /* TODO: don't hard code this */ - fenced_list->checkDelayed = 5; - _glthread_INIT_MUTEX(fenced_list->mutex); return fenced_list; -- cgit v1.2.3 From a8ca54955322b34c77a7459246e5639d3f8610cd Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 5 Apr 2008 10:22:47 +0200 Subject: draw: Use debug_printf(). --- src/gallium/auxiliary/draw/draw_clip.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_clip.c b/src/gallium/auxiliary/draw/draw_clip.c index fd1c71152e..e24c5d8032 100644 --- a/src/gallium/auxiliary/draw/draw_clip.c +++ b/src/gallium/auxiliary/draw/draw_clip.c @@ -184,10 +184,10 @@ static void emit_poly( struct draw_stage *stage, if (0) { const struct draw_vertex_shader *vs = stage->draw->vertex_shader; uint j, k; - printf("Clipped tri:\n"); + debug_printf("Clipped tri:\n"); for (j = 0; j < 3; j++) { for (k = 0; k < vs->info.num_outputs; k++) { - printf(" Vert %d: Attr %d: %f %f %f %f\n", j, k, + debug_printf(" Vert %d: Attr %d: %f %f %f %f\n", j, k, header.v[j]->data[k][0], header.v[j]->data[k][1], header.v[j]->data[k][2], -- cgit v1.2.3 From 5c19e47362c2d193850e98bd43a2bc2b783b0b5c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 6 Apr 2008 14:29:30 +0100 Subject: draw: fix edgeflag handling on the pt paths Encode edgeflags (and reset_stipple info) into the top two bits of the fetch elements. This info could be moved elsewhere, but for now we can live with a 1<<30 maximum element size... Also use the primitive decomposition code from draw_prim.c verbatim, as it includes all this stuff and is known to work. --- src/gallium/auxiliary/draw/draw_pt.h | 10 + .../auxiliary/draw/draw_pt_fetch_pipeline.c | 16 +- src/gallium/auxiliary/draw/draw_pt_pipeline.c | 6 +- src/gallium/auxiliary/draw/draw_pt_vcache.c | 396 ++++++++++++--------- 4 files changed, 253 insertions(+), 175 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 800072c511..590823fd33 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -40,6 +40,16 @@ typedef unsigned (*pt_elt_func)( const void *elts, unsigned idx ); struct draw_pt_middle_end; struct draw_context; +/* We use the top couple of bits in the vertex fetch index to convey a + * little API information. This limits the number of vertices we can + * address to only 1 billion -- if that becomes a problem, these could + * be moved out & passed separately. + */ +#define DRAW_PT_EDGEFLAG (1<<30) +#define DRAW_PT_RESET_STIPPLE (1<<31) +#define DRAW_PT_FLAG_MASK (3<<30) + + /* The "front end" - prepare sets of fetch, draw elements for the * middle end. * diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c index ba95420b72..4c2a281b29 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c @@ -72,7 +72,7 @@ struct fetch_pipeline_middle_end { struct draw_pt_middle_end base; struct draw_context *draw; - void (*header)( const unsigned *edgeflag, unsigned count, float **out); + void (*header)( unsigned idx, float **out); struct { const ubyte *ptr; @@ -122,8 +122,7 @@ static void emit_R32G32B32A32_FLOAT( const float *attrib, (*out) += 4; } -static void header( const unsigned *edgeflag, - unsigned idx, +static void header( unsigned idx, float **out ) { struct vertex_header *header = (struct vertex_header *) (*out); @@ -141,14 +140,15 @@ static void header( const unsigned *edgeflag, } -static void header_ef( const unsigned *edgeflag, - unsigned idx, +static void header_ef( unsigned idx, float **out ) { struct vertex_header *header = (struct vertex_header *) (*out); + /* XXX: need a reset_stipple flag in the vertex header too? + */ header->clipmask = 0; - header->edgeflag = (edgeflag[idx/32] & (1 << (idx%32))) != 0; + header->edgeflag = (idx & DRAW_PT_EDGEFLAG) != 0; header->pad = 0; header->vertex_id = UNDEFINED_VERTEX_ID; @@ -172,14 +172,14 @@ fetch_store_general( struct fetch_pipeline_middle_end *fpme, const unsigned *fetch_elts, unsigned count ) { - const unsigned *edgeflag = fpme->draw->user.edgeflag; float *out = (float *)out_ptr; uint i, j; for (i = 0; i < count; i++) { unsigned elt = fetch_elts[i]; - fpme->header( edgeflag, i, &out ); + fpme->header( elt, &out ); + elt &= ~DRAW_PT_FLAG_MASK; for (j = 0; j < fpme->nr_fetch; j++) { float attrib[4]; diff --git a/src/gallium/auxiliary/draw/draw_pt_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_pipeline.c index 942df518eb..e70e63d08f 100644 --- a/src/gallium/auxiliary/draw/draw_pt_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_pipeline.c @@ -85,7 +85,6 @@ static void do_triangle( struct draw_context *draw, { struct prim_header prim; -// _mesa_printf("tri %d %d %d\n", i0, i1, i2); prim.v[0] = (struct vertex_header *)v0; prim.v[1] = (struct vertex_header *)v1; prim.v[2] = (struct vertex_header *)v2; @@ -95,6 +94,11 @@ static void do_triangle( struct draw_context *draw, (prim.v[2]->edgeflag << 2)); prim.pad = 0; + if (0) debug_printf("tri ef: %d %d %d\n", + prim.v[0]->edgeflag, + prim.v[1]->edgeflag, + prim.v[2]->edgeflag); + draw->pipeline.first->tri( draw->pipeline.first, &prim ); } diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 5a068761df..107dcfc269 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -55,9 +55,6 @@ struct vcache_frontend { unsigned draw_count; unsigned fetch_count; - pt_elt_func elt_func; - const void *elt_ptr; - struct draw_pt_middle_end *middle; unsigned input_prim; @@ -66,17 +63,6 @@ struct vcache_frontend { static void vcache_flush( struct vcache_frontend *vcache ) { -#if 0 - /* Should always be true if output_prim == input_prim, otherwise - * not so much... - */ - unsigned i; - for (i = 0; i < vcache->draw_count; i++) { - assert( vcache->fetch_elts[vcache->draw_elts[i]] == - vcache->elt_func(vcache->elt_ptr, i) ); - } -#endif - if (vcache->draw_count) { vcache->middle->run( vcache->middle, vcache->fetch_elts, @@ -115,26 +101,73 @@ static void vcache_elt( struct vcache_frontend *vcache, vcache->draw_elts[vcache->draw_count++] = vcache->out[idx]; } + +static unsigned add_edgeflag( struct vcache_frontend *vcache, + unsigned idx, + unsigned mask ) +{ + if (mask && draw_get_edgeflag(vcache->draw, idx)) + return idx | DRAW_PT_EDGEFLAG; + else + return idx; +} + + +static unsigned add_reset_stipple( unsigned idx, + unsigned reset ) +{ + if (reset) + return idx | DRAW_PT_RESET_STIPPLE; + else + return idx; +} + static void vcache_triangle( struct vcache_frontend *vcache, unsigned i0, unsigned i1, unsigned i2 ) { - /* TODO: encode edgeflags in draw_elts */ + vcache_elt(vcache, i0 | DRAW_PT_EDGEFLAG | DRAW_PT_RESET_STIPPLE); + vcache_elt(vcache, i1 | DRAW_PT_EDGEFLAG); + vcache_elt(vcache, i2 | DRAW_PT_EDGEFLAG); + vcache_check_flush(vcache); +} + + +static void vcache_ef_triangle( struct vcache_frontend *vcache, + boolean reset_stipple, + unsigned ef_mask, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + i0 = add_edgeflag( vcache, i0, (ef_mask >> 0) & 1 ); + i1 = add_edgeflag( vcache, i1, (ef_mask >> 1) & 1 ); + i2 = add_edgeflag( vcache, i2, (ef_mask >> 2) & 1 ); + + i0 = add_reset_stipple( i0, reset_stipple ); + vcache_elt(vcache, i0); vcache_elt(vcache, i1); vcache_elt(vcache, i2); vcache_check_flush(vcache); + + if (0) debug_printf("emit tri ef: %d %d %d\n", + !!(i0 & DRAW_PT_EDGEFLAG), + !!(i1 & DRAW_PT_EDGEFLAG), + !!(i2 & DRAW_PT_EDGEFLAG)); + } + static void vcache_line( struct vcache_frontend *vcache, - boolean reset, + boolean reset_stipple, unsigned i0, unsigned i1 ) { - /* TODO: encode reset-line-stipple in draw_elts */ - (void) reset; + i0 = add_reset_stipple( i0, reset_stipple ); + vcache_elt(vcache, i0); vcache_elt(vcache, i1); vcache_check_flush(vcache); @@ -158,39 +191,43 @@ static void vcache_quad( struct vcache_frontend *vcache, vcache_triangle( vcache, i1, i2, i3 ); } +static void vcache_ef_quad( struct vcache_frontend *vcache, + unsigned i0, + unsigned i1, + unsigned i2, + unsigned i3 ) +{ + const unsigned omitEdge2 = ~(1 << 1); + const unsigned omitEdge3 = ~(1 << 2); + vcache_ef_triangle( vcache, 1, omitEdge2, i0, i1, i3 ); + vcache_ef_triangle( vcache, 0, omitEdge3, i1, i2, i3 ); +} + -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { - PIPE_PRIM_POINTS, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES -}; -static void vcache_run_pv2( struct draw_pt_front_end *frontend, - pt_elt_func get_elt, - const void *elts, - unsigned count ) +static void vcache_run( struct draw_pt_front_end *frontend, + pt_elt_func get_elt, + const void *elts, + unsigned count ) { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; + struct draw_context *draw = vcache->draw; + + boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); + + boolean flatfirst = (draw->rasterizer->flatshade && + draw->rasterizer->flatshade_first); unsigned i; - - /* These are for validation only: - */ - vcache->elt_func = get_elt; - vcache->elt_ptr = elts; + +// debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); switch (vcache->input_prim) { case PIPE_PRIM_POINTS: for (i = 0; i < count; i ++) { - vcache_point( vcache, - get_elt(elts, i) ); + vcache_point( vcache, + get_elt(elts, i + 0) ); } break; @@ -205,17 +242,17 @@ static void vcache_run_pv2( struct draw_pt_front_end *frontend, case PIPE_PRIM_LINE_LOOP: if (count >= 2) { - for (i = 1; i < count; i++) { - vcache_line( vcache, + for (i = 1; i < count; i++) { + vcache_line( vcache, i == 1, /* XXX: only if vb not split */ get_elt(elts, i - 1), - get_elt(elts, i) ); - } + get_elt(elts, i )); + } - vcache_line( vcache, + vcache_line( vcache, 0, get_elt(elts, count - 1), - get_elt(elts, 0) ); + get_elt(elts, 0 )); } break; @@ -224,72 +261,163 @@ static void vcache_run_pv2( struct draw_pt_front_end *frontend, vcache_line( vcache, i == 1, get_elt(elts, i - 1), - get_elt(elts, i) ); + get_elt(elts, i )); } break; case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - vcache_triangle( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2) ); + if (unfilled) { + for (i = 0; i+2 < count; i += 3) { + vcache_ef_triangle( vcache, + 1, + ~0, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2 )); + } + } + else { + for (i = 0; i+2 < count; i += 3) { + vcache_triangle( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2 )); + } } break; case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - vcache_triangle( vcache, - get_elt(elts, i + 1), - get_elt(elts, i + 0), - get_elt(elts, i + 2) ); + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + if (i & 1) { + vcache_triangle( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 2), + get_elt(elts, i + 1 )); + } + else { + vcache_triangle( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2 )); + } } - else { - vcache_triangle( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2) ); + } + else { + for (i = 0; i+2 < count; i++) { + if (i & 1) { + vcache_triangle( vcache, + get_elt(elts, i + 1), + get_elt(elts, i + 0), + get_elt(elts, i + 2 )); + } + else { + vcache_triangle( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2 )); + } } } break; case PIPE_PRIM_TRIANGLE_FAN: - for (i = 0; i+2 < count; i++) { - vcache_triangle( vcache, - get_elt(elts, 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2) ); + if (count >= 3) { + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + vcache_triangle( vcache, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0 )); + } + } + else { + for (i = 0; i+2 < count; i++) { + vcache_triangle( vcache, + get_elt(elts, 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2 )); + } + } } break; case PIPE_PRIM_QUADS: - for (i = 0; i+3 < count; i += 4) { - vcache_quad( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, i + 3)); + if (unfilled) { + for (i = 0; i+3 < count; i += 4) { + vcache_ef_quad( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, i + 3)); + } + } + else { + for (i = 0; i+3 < count; i += 4) { + vcache_quad( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, i + 3)); + } } break; case PIPE_PRIM_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - vcache_quad( vcache, - get_elt(elts, i + 2), - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 3)); + if (unfilled) { + for (i = 0; i+3 < count; i += 2) { + vcache_ef_quad( vcache, + get_elt(elts, i + 2), + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 3)); + } + } + else { + for (i = 0; i+3 < count; i += 2) { + vcache_quad( vcache, + get_elt(elts, i + 2), + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 3)); + } } break; case PIPE_PRIM_POLYGON: - for (i = 0; i+2 < count; i++) { - vcache_triangle( vcache, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0)); + if (unfilled) { + /* These bitflags look a little odd because we submit the + * vertices as (1,2,0) to satisfy flatshade requirements. + */ + const unsigned edge_first = (1<<2); + const unsigned edge_middle = (1<<0); + const unsigned edge_last = (1<<1); + + for (i = 0; i+2 < count; i++) { + unsigned ef_mask = edge_middle; + + if (i == 0) + ef_mask |= edge_first; + + if (i + 3 == count) + ef_mask |= edge_last; + + vcache_ef_triangle( vcache, + i == 0, + ef_mask, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0)); + } + } + else { + for (i = 0; i+2 < count; i++) { + vcache_triangle( vcache, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0)); + } } break; @@ -302,88 +430,21 @@ static void vcache_run_pv2( struct draw_pt_front_end *frontend, } -static void vcache_run_pv0( struct draw_pt_front_end *frontend, - pt_elt_func get_elt, - const void *elts, - unsigned count ) -{ - struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; - unsigned i; - - /* These are for validation only: - */ - vcache->elt_func = get_elt; - vcache->elt_ptr = elts; - - switch (vcache->input_prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i ++) { - vcache_point( vcache, - get_elt(elts, i) ); - } - break; - - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) { - vcache_line( vcache, - TRUE, - get_elt(elts, i + 0), - get_elt(elts, i + 1)); - } - break; - - case PIPE_PRIM_LINE_STRIP: - for (i = 1; i < count; i++) { - vcache_line( vcache, - i == 1, - get_elt(elts, i - 1), - get_elt(elts, i) ); - } - break; - - case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - vcache_triangle( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2) ); - } - break; - - case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - if (i & 1) { - vcache_triangle( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 2), - get_elt(elts, i + 1) ); - } - else { - vcache_triangle( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2) ); - } - } - break; - case PIPE_PRIM_TRIANGLE_FAN: - for (i = 0; i+2 < count; i++) { - vcache_triangle( vcache, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0) ); - } - break; +static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES +}; - default: - assert(0); - break; - } - - vcache_flush( vcache ); -} static void vcache_prepare( struct draw_pt_front_end *frontend, unsigned prim, @@ -391,11 +452,14 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; +/* if (vcache->draw->rasterizer->flatshade_first) vcache->base.run = vcache_run_pv0; else vcache->base.run = vcache_run_pv2; - +*/ + + vcache->base.run = vcache_run; vcache->input_prim = prim; vcache->output_prim = reduced_prim[prim]; -- cgit v1.2.3 From a8a5376406cabf5aa6a44f7d37f5f8abbb4adf56 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 7 Apr 2008 12:28:31 +0100 Subject: draw: strip edgeflags out of fetch-emit path --- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 39f0b40838..0806076956 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -149,7 +149,7 @@ fetch_store_general( struct fetch_emit_middle_end *feme, uint i, j; for (i = 0; i < count; i++) { - unsigned elt = fetch_elts[i]; + unsigned elt = fetch_elts[i] & ~DRAW_PT_FLAG_MASK; for (j = 0; j < feme->nr_fetch; j++) { float attrib[4]; -- cgit v1.2.3 From f3bd7bf5c913d2a58d424e995b4d441e402bd62b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 7 Apr 2008 11:15:23 -0600 Subject: mesa: added _mesa_free_instructions() --- src/mesa/shader/prog_instruction.c | 17 +++++++++++++++++ src/mesa/shader/prog_instruction.h | 3 +++ 2 files changed, 20 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/prog_instruction.c b/src/mesa/shader/prog_instruction.c index 066129037a..bea5d0551e 100644 --- a/src/mesa/shader/prog_instruction.c +++ b/src/mesa/shader/prog_instruction.c @@ -118,6 +118,23 @@ _mesa_copy_instructions(struct prog_instruction *dest, } +/** + * Free an array of instructions + */ +void +_mesa_free_instructions(struct prog_instruction *inst, GLuint count) +{ + GLuint i; + for (i = 0; i < count; i++) { + if (inst[i].Data) + _mesa_free(inst[i].Data); + if (inst[i].Comment) + _mesa_free((char *) inst[i].Comment); + } + _mesa_free(inst); +} + + /** * Basic info about each instruction */ diff --git a/src/mesa/shader/prog_instruction.h b/src/mesa/shader/prog_instruction.h index e8a2407ea8..711166f9dd 100644 --- a/src/mesa/shader/prog_instruction.h +++ b/src/mesa/shader/prog_instruction.h @@ -439,6 +439,9 @@ extern struct prog_instruction * _mesa_copy_instructions(struct prog_instruction *dest, const struct prog_instruction *src, GLuint n); +extern void +_mesa_free_instructions(struct prog_instruction *inst, GLuint count); + extern GLuint _mesa_num_inst_src_regs(gl_inst_opcode opcode); -- cgit v1.2.3 From 5d1e73028aabfa1470bfed02c705a2696706f857 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 7 Apr 2008 11:16:18 -0600 Subject: mesa: added _mesa_insert_instructions() Also, use new _mesa_free_instructions() in a few places. --- src/mesa/shader/program.c | 60 +++++++++++++++++++++++++++++++++++++++-------- src/mesa/shader/program.h | 2 ++ 2 files changed, 52 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 3069b04836..0ed7f833d2 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -287,16 +287,7 @@ _mesa_delete_program(GLcontext *ctx, struct gl_program *prog) if (prog->String) _mesa_free(prog->String); - if (prog->Instructions) { - GLuint i; - for (i = 0; i < prog->NumInstructions; i++) { - if (prog->Instructions[i].Data) - _mesa_free(prog->Instructions[i].Data); - if (prog->Instructions[i].Comment) - _mesa_free((char *) prog->Instructions[i].Comment); - } - _mesa_free(prog->Instructions); - } + _mesa_free_instructions(prog->Instructions, prog->NumInstructions); if (prog->Parameters) { _mesa_free_parameter_list(prog->Parameters); @@ -415,6 +406,55 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog) } +/** + * Insert 'count' NOP instructions at 'start' in the given program. + * Adjust branch targets accordingly. + */ +GLboolean +_mesa_insert_instructions(struct gl_program *prog, GLuint start, GLuint count) +{ + const GLuint origLen = prog->NumInstructions; + const GLuint newLen = origLen + count; + struct prog_instruction *newInst; + GLuint i; + + /* adjust branches */ + for (i = 0; i < prog->NumInstructions; i++) { + struct prog_instruction *inst = prog->Instructions + i; + if (inst->BranchTarget > 0) { + if (inst->BranchTarget >= start) { + inst->BranchTarget += count; + } + } + } + + /* Alloc storage for new instructions */ + newInst = _mesa_alloc_instructions(newLen); + if (!newInst) { + return GL_FALSE; + } + + /* Copy 'start' instructions into new instruction buffer */ + _mesa_copy_instructions(newInst, prog->Instructions, start); + + /* init the new instructions */ + _mesa_init_instructions(newInst + start, count); + + /* Copy the remaining/tail instructions to new inst buffer */ + _mesa_copy_instructions(newInst + start + count, + prog->Instructions + start, + origLen - start); + + /* free old instructions */ + _mesa_free_instructions(prog->Instructions, origLen); + + /* install new instructions */ + prog->Instructions = newInst; + prog->NumInstructions = newLen; + + return GL_TRUE; +} + /** * Search instructions for registers that match (oldFile, oldIndex), diff --git a/src/mesa/shader/program.h b/src/mesa/shader/program.h index 4b7297e4c6..414a57d39c 100644 --- a/src/mesa/shader/program.h +++ b/src/mesa/shader/program.h @@ -87,6 +87,8 @@ _mesa_lookup_program(GLcontext *ctx, GLuint id); extern struct gl_program * _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog); +extern GLboolean +_mesa_insert_instructions(struct gl_program *prog, GLuint start, GLuint count); extern struct gl_program * _mesa_combine_programs(GLcontext *ctx, -- cgit v1.2.3 From 48a25bdd3693ec4a2556efb3c387cc3eb8151cb5 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 7 Apr 2008 11:20:21 -0600 Subject: mesa: new _mesa_remove_varying_reads() function We'll apply this function to GLSL vertex programs. In GLSL it's legal to read and write varying (output) vars in a vertex shader. But reading from an output register isn't supported by all hardware. This routine examines the vertex program for that condition and rewrites it to use temporary registers where needed. --- src/mesa/shader/programopt.c | 96 +++++++++++++++++++++++++++++++++++++++++++- src/mesa/shader/programopt.h | 2 + 2 files changed, 96 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index 9eeb71db1b..7d560c74a5 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -35,6 +35,7 @@ #include "context.h" #include "prog_parameter.h" #include "prog_statevars.h" +#include "program.h" #include "programopt.h" #include "prog_instruction.h" @@ -102,7 +103,7 @@ _mesa_insert_mvp_code(GLcontext *ctx, struct gl_vertex_program *vprog) _mesa_copy_instructions (newInst + 4, vprog->Base.Instructions, origLen); /* free old instructions */ - _mesa_free(vprog->Base.Instructions); + _mesa_free_instructions(vprog->Base.Instructions, origLen); /* install new instructions */ vprog->Base.Instructions = newInst; @@ -274,7 +275,7 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog) inst++; /* free old instructions */ - _mesa_free(fprog->Base.Instructions); + _mesa_free_instructions(fprog->Base.Instructions, origLen); /* install new instructions */ fprog->Base.Instructions = newInst; @@ -364,3 +365,94 @@ _mesa_count_texture_instructions(struct gl_program *prog) } } + +/** + * Scan/rewrite program to remove reads of varying (output) registers. + * In GLSL vertex shaders, varying vars can be read and written. + * Normally, vertex varying vars are implemented as output registers. + * On some hardware, trying to read an output register causes trouble. + * So, rewrite the program to use a temporary register in this case. + */ +void +_mesa_remove_varying_reads(struct gl_program *prog) +{ + GLuint i; + GLint outputMap[VERT_RESULT_MAX]; + GLuint numVaryingReads = 0; + + assert(prog->Target == GL_VERTEX_PROGRAM_ARB); + + for (i = 0; i < VERT_RESULT_MAX; i++) + outputMap[i] = -1; + + /* look for instructions which read from varying vars */ + for (i = 0; i < prog->NumInstructions; i++) { + struct prog_instruction *inst = prog->Instructions + i; + const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode); + GLuint j; + for (j = 0; j < numSrc; j++) { + if (inst->SrcReg[j].File == PROGRAM_VARYING) { + /* replace the read with a temp reg */ + const GLuint var = inst->SrcReg[j].Index; + if (outputMap[var] == -1) { + numVaryingReads++; + outputMap[var] = _mesa_find_free_register(prog, + PROGRAM_TEMPORARY); + } + inst->SrcReg[j].File = PROGRAM_TEMPORARY; + inst->SrcReg[j].Index = outputMap[var]; + } + } + } + + if (numVaryingReads == 0) + return; /* nothing to be done */ + + /* look for instructions which write to the varying vars identified above */ + for (i = 0; i < prog->NumInstructions; i++) { + struct prog_instruction *inst = prog->Instructions + i; + const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode); + GLuint j; + for (j = 0; j < numSrc; j++) { + if (inst->DstReg.File == PROGRAM_VARYING && + outputMap[inst->DstReg.Index] >= 0) { + /* change inst to write to the temp reg, instead of the varying */ + inst->DstReg.File = PROGRAM_TEMPORARY; + inst->DstReg.Index = outputMap[inst->DstReg.Index]; + } + } + } + + /* insert new instructions to copy the temp vars to the varying vars */ + { + struct prog_instruction *inst; + GLint endPos, var; + + /* Look for END instruction and insert the new varying writes */ + endPos = -1; + for (i = 0; i < prog->NumInstructions; i++) { + struct prog_instruction *inst = prog->Instructions + i; + if (inst->Opcode == OPCODE_END) { + endPos = i; + _mesa_insert_instructions(prog, i, numVaryingReads); + break; + } + } + + assert(endPos >= 0); + + /* insert new MOV instructions here */ + inst = prog->Instructions + endPos; + for (var = 0; var < VERT_RESULT_MAX; var++) { + if (outputMap[var] >= 0) { + /* MOV VAR[var], TEMP[tmp]; */ + inst->Opcode = OPCODE_MOV; + inst->DstReg.File = PROGRAM_VARYING; + inst->DstReg.Index = var; + inst->SrcReg[0].File = PROGRAM_TEMPORARY; + inst->SrcReg[0].Index = outputMap[var]; + inst++; + } + } + } +} diff --git a/src/mesa/shader/programopt.h b/src/mesa/shader/programopt.h index ce63644bbf..47ff2f0c7b 100644 --- a/src/mesa/shader/programopt.h +++ b/src/mesa/shader/programopt.h @@ -39,5 +39,7 @@ _mesa_count_texture_indirections(struct gl_program *prog); extern void _mesa_count_texture_instructions(struct gl_program *prog); +extern void +_mesa_remove_varying_reads(struct gl_program *prog); #endif /* PROGRAMOPT_H */ -- cgit v1.2.3 From 42a9218daa8f564ea1b992a48cc3c375fd1b815e Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 7 Apr 2008 11:22:41 -0600 Subject: mesa: call _mesa_remove_varying_reads() after compiling vertex shaders --- src/mesa/shader/slang/slang_compile.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index bfb9ca4db6..0df673085b 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -31,6 +31,8 @@ #include "main/imports.h" #include "main/context.h" #include "shader/program.h" +#include "shader/programopt.h" +#include "shader/prog_print.h" #include "shader/prog_parameter.h" #include "shader/grammar/grammar_mesa.h" #include "slang_codegen.h" @@ -2186,6 +2188,19 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) _slang_delete_mempool((slang_mempool *) ctx->Shader.MemPool); ctx->Shader.MemPool = NULL; + if (shader->Type == GL_VERTEX_SHADER) { + /* remove any reads of varying (output) registers */ +#if 0 + printf("Pre-remove output reads:\n"); + _mesa_print_program(shader->Programs[0]); +#endif + _mesa_remove_varying_reads(shader->Programs[0]); +#if 0 + printf("Post-remove output reads:\n"); + _mesa_print_program(shader->Programs[0]); +#endif + } + return success; } -- cgit v1.2.3 From 24fc93ebdc119ee01d6ce47ea4fe33a2a8067bd2 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 7 Apr 2008 19:39:28 +0200 Subject: gallium: Stop warnings --- src/gallium/include/pipe/p_util.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 8e3aaee496..1bc00ac7a0 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -207,7 +207,10 @@ mem_dup(const void *src, uint size) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) +/* JB stop warnings */ +#ifndef Elements #define Elements(x) (sizeof(x)/sizeof((x)[0])) +#endif #define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER)) /** -- cgit v1.2.3 From d945e8ddd7cc8f2882ea0a2336bc2e534a1bca78 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 7 Apr 2008 19:40:01 +0200 Subject: mesa: Pulled in glthread from i915tex --- src/mesa/glapi/glthread.h | 48 +++++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 44 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/glapi/glthread.h b/src/mesa/glapi/glthread.h index afb04f7bf2..b8e9d5f59b 100644 --- a/src/mesa/glapi/glthread.h +++ b/src/mesa/glapi/glthread.h @@ -116,9 +116,49 @@ typedef pthread_mutex_t _glthread_Mutex; #define _glthread_UNLOCK_MUTEX(name) \ (void) pthread_mutex_unlock(&(name)) -#endif /* PTHREADS */ +typedef pthread_cond_t _glthread_Cond; +#define _glthread_DECLARE_STATIC_COND(name) \ + static _glthread_Cond name = PTHREAD_COND_INITIALIZER +#define _glthread_INIT_COND(cond) \ + pthread_cond_init(&(cond), NULL) + +#define _glthread_DESTROY_COND(name) \ + pthread_cond_destroy(&(name)) + +#define _glthread_COND_WAIT(cond, mutex) \ + pthread_cond_wait(&(cond), &(mutex)) + +#define _glthread_COND_SIGNAL(cond) \ + pthread_cond_signal(&(cond)) + +#define _glthread_COND_BROADCAST(cond) \ + pthread_cond_broadcast(&(cond)) + + +#else /* PTHREADS */ + +typedef unsigned int _glthread_Cond; +#define _glthread_DECLARE_STATIC_COND(name) \ +// #warning Condition variables not implemented. + +#define _glthread_INIT_COND(cond) \ + abort(); + +#define _glthread_DESTROY_COND(name) \ + abort(); + +#define _glthread_COND_WAIT(cond, mutex) \ + abort(); + +#define _glthread_COND_SIGNAL(cond) \ + abort(); + +#define _glthread_COND_BROADCAST(cond) \ + abort(); + +#endif /* @@ -259,11 +299,11 @@ typedef benaphore _glthread_Mutex; * THREADS not defined */ -typedef unsigned _glthread_TSD; +typedef GLuint _glthread_TSD; -typedef unsigned _glthread_Thread; +typedef GLuint _glthread_Thread; -typedef unsigned _glthread_Mutex; +typedef GLuint _glthread_Mutex; #define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 -- cgit v1.2.3 From d5d93a31580cf58744b4d186a4ab8fac09743a33 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 7 Apr 2008 19:40:43 +0200 Subject: i915: Pulled in changes from i915tex --- src/gallium/winsys/dri/intel/Makefile | 7 +- src/gallium/winsys/dri/intel/intel_batchbuffer.c | 488 ++++++----- src/gallium/winsys/dri/intel/intel_batchbuffer.h | 108 ++- src/gallium/winsys/dri/intel/intel_batchpool.c | 427 ---------- src/gallium/winsys/dri/intel/intel_batchpool.h | 37 - src/gallium/winsys/dri/intel/intel_context.c | 5 +- src/gallium/winsys/dri/intel/intel_context.h | 2 +- src/gallium/winsys/dri/intel/intel_ioctl.c | 138 ++++ src/gallium/winsys/dri/intel/intel_ioctl.h | 40 + src/gallium/winsys/dri/intel/intel_screen.c | 53 +- src/gallium/winsys/dri/intel/intel_screen.h | 9 +- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 4 +- src/gallium/winsys/dri/intel/intel_winsys.h | 2 +- src/gallium/winsys/dri/intel/intel_winsys_i915.c | 30 +- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 55 +- src/gallium/winsys/dri/intel/server/i830_common.h | 46 +- src/gallium/winsys/dri/intel/ws_dri_bufmgr.c | 945 ++++++++++++++++++++++ src/gallium/winsys/dri/intel/ws_dri_bufmgr.h | 138 ++++ src/gallium/winsys/dri/intel/ws_dri_bufpool.h | 102 +++ src/gallium/winsys/dri/intel/ws_dri_drmpool.c | 268 ++++++ src/gallium/winsys/dri/intel/ws_dri_fencemgr.c | 372 +++++++++ src/gallium/winsys/dri/intel/ws_dri_fencemgr.h | 115 +++ src/gallium/winsys/dri/intel/ws_dri_mallocpool.c | 162 ++++ src/gallium/winsys/dri/intel/ws_dri_slabpool.c | 945 ++++++++++++++++++++++ 24 files changed, 3742 insertions(+), 756 deletions(-) delete mode 100644 src/gallium/winsys/dri/intel/intel_batchpool.c delete mode 100644 src/gallium/winsys/dri/intel/intel_batchpool.h create mode 100644 src/gallium/winsys/dri/intel/intel_ioctl.c create mode 100644 src/gallium/winsys/dri/intel/intel_ioctl.h create mode 100644 src/gallium/winsys/dri/intel/ws_dri_bufmgr.c create mode 100644 src/gallium/winsys/dri/intel/ws_dri_bufmgr.h create mode 100644 src/gallium/winsys/dri/intel/ws_dri_bufpool.h create mode 100644 src/gallium/winsys/dri/intel/ws_dri_drmpool.c create mode 100644 src/gallium/winsys/dri/intel/ws_dri_fencemgr.c create mode 100644 src/gallium/winsys/dri/intel/ws_dri_fencemgr.h create mode 100644 src/gallium/winsys/dri/intel/ws_dri_mallocpool.c create mode 100644 src/gallium/winsys/dri/intel/ws_dri_slabpool.c (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/Makefile b/src/gallium/winsys/dri/intel/Makefile index 40654bb2ac..f140939248 100644 --- a/src/gallium/winsys/dri/intel/Makefile +++ b/src/gallium/winsys/dri/intel/Makefile @@ -19,11 +19,14 @@ DRIVER_SOURCES = \ intel_context.c \ intel_lock.c \ intel_screen.c \ - intel_batchpool.c + ws_dri_bufmgr.c \ + ws_dri_drmpool.c \ + ws_dri_fencemgr.c \ + ws_dri_mallocpool.c \ + ws_dri_slabpool.c C_SOURCES = \ $(COMMON_GALLIUM_SOURCES) \ - $(COMMON_BM_SOURCES) \ $(DRIVER_SOURCES) ASM_SOURCES = diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c index 5830b88b37..05223bf700 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.c +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.c @@ -25,108 +25,95 @@ * **************************************************************************/ -#include -#include #include "intel_batchbuffer.h" -#include "intel_context.h" -#include "intel_screen.h" -#include "intel_reg.h" -#include "drm.h" - -/* Relocations in kernel space: - * - pass dma buffer seperately - * - memory manager knows how to patch - * - pass list of dependent buffers - * - pass relocation list - * - * Either: - * - get back an offset for buffer to fire - * - memory manager knows how to fire buffer - * - * Really want the buffer to be AGP and pinned. - * - */ - -/* Cliprect fence: The highest fence protecting a dma buffer - * containing explicit cliprect information. Like the old drawable - * lock but irq-driven. X server must wait for this fence to expire - * before changing cliprects [and then doing sw rendering?]. For - * other dma buffers, the scheduler will grab current cliprect info - * and mix into buffer. X server must hold the lock while changing - * cliprects??? Make per-drawable. Need cliprects in shared memory - * -- beats storing them with every cmd buffer in the queue. - * - * ==> X server must wait for this fence to expire before touching the - * framebuffer with new cliprects. - * - * ==> Cliprect-dependent buffers associated with a - * cliprect-timestamp. All of the buffers associated with a timestamp - * must go to hardware before any buffer with a newer timestamp. - * - * ==> Dma should be queued per-drawable for correct X/GL - * synchronization. Or can fences be used for this? - * - * Applies to: Blit operations, metaops, X server operations -- X - * server automatically waits on its own dma to complete before - * modifying cliprects ??? - */ +#include "intel_ioctl.h" +#include +#if 0 static void -intel_dump_batchbuffer(uint offset, uint * ptr, uint count) +intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) { int i; - printf("\n\n\nSTART BATCH (%d dwords):\n", count / 4); - for (i = 0; i < count / 4; i += 1) - printf("\t0x%08x\n", ptr[i]); - printf("END BATCH\n\n\n"); + fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); + for (i = 0; i < count / 4; i += 4) + fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", + offset + i * 4, ptr[i], ptr[i + 1], ptr[i + 2], ptr[i + 3]); + fprintf(stderr, "END BATCH\n\n\n"); +} +#endif + +static void +intel_realloc_relocs(struct intel_batchbuffer *batch, int num_relocs) +{ + unsigned long size = num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER; + + size *= sizeof(uint32_t); + batch->reloc = realloc(batch->reloc, size); + batch->reloc_size = num_relocs; } void intel_batchbuffer_reset(struct intel_batchbuffer *batch) { - int i; - - if (batch->map) { - driBOUnmap(batch->buffer); - batch->map = NULL; - } - /* * Get a new, free batchbuffer. */ - batch->size = BATCH_SZ; - driBOData(batch->buffer, batch->size, NULL, 0); + drmBO *bo; + struct drm_bo_info_req *req; + + driBOUnrefUserList(batch->list); + driBOResetList(batch->list); - driBOResetList(&batch->list); + batch->size = 4096; // ZZZ JB batch->intel->intelScreen->maxBatchSize; + driBOData(batch->buffer, batch->size, NULL, NULL, 0); /* - * Unreference buffers previously on the relocation list. + * Add the batchbuffer to the validate list. */ - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - driBOUnReference(r->buf); - } - batch->list_count = 0; - batch->nr_relocs = 0; - batch->flags = 0; + driBOAddListItem(batch->list, batch->buffer, + DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_EXE | DRM_BO_MASK_MEM, + &batch->dest_location, &batch->node); + + req = &batch->node->bo_arg.d.req.bo_req; /* - * We don't refcount the batchbuffer itself since we can't destroy it - * while it's on the list. + * Set up information needed for us to make relocations + * relative to the underlying drm buffer objects. */ - driBOAddListItem(&batch->list, batch->buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_EXE); + driReadLockKernelBO(); + bo = driBOKernel(batch->buffer); + req->presumed_offset = (uint64_t) bo->offset; + req->hint = DRM_BO_HINT_PRESUMED_OFFSET; + batch->drmBOVirtual = (uint8_t *) bo->virtual; + driReadUnlockKernelBO(); + /* + * Adjust the relocation buffer size. + */ + + if (batch->reloc_size > INTEL_MAX_RELOCS || + batch->reloc == NULL) + intel_realloc_relocs(batch, INTEL_DEFAULT_RELOCS); + + assert(batch->reloc != NULL); + batch->reloc[0] = 0; /* No relocs yet. */ + batch->reloc[1] = 1; /* Reloc type 1 */ + batch->reloc[2] = 0; /* Only a single relocation list. */ + batch->reloc[3] = 0; /* Only a single relocation list. */ batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); + batch->poolOffset = driBOPoolOffset(batch->buffer); batch->ptr = batch->map; + batch->dirty_state = ~0; + batch->nr_relocs = 0; + batch->flags = 0; + batch->id = 0;//batch->intel->intelScreen->batch_id++; } - /*====================================================================== * Public functions */ @@ -141,121 +128,253 @@ intel_batchbuffer_alloc(struct intel_context *intel) &batch->buffer, 4096, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); batch->last_fence = NULL; - driBOCreateList(20, &batch->list); + batch->list = driBOCreateList(20); + batch->reloc = NULL; intel_batchbuffer_reset(batch); return batch; } - void intel_batchbuffer_free(struct intel_batchbuffer *batch) { if (batch->last_fence) { driFenceFinish(batch->last_fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, - GL_FALSE); - driFenceUnReference(batch->last_fence); - batch->last_fence = NULL; + DRM_FENCE_TYPE_EXE, GL_FALSE); + driFenceUnReference(&batch->last_fence); } if (batch->map) { driBOUnmap(batch->buffer); batch->map = NULL; } driBOUnReference(batch->buffer); + driBOFreeList(batch->list); + if (batch->reloc) + free(batch->reloc); batch->buffer = NULL; free(batch); } +void +intel_offset_relocation(struct intel_batchbuffer *batch, + unsigned pre_add, + struct _DriBufferObject *driBO, + uint64_t val_flags, + uint64_t val_mask) +{ + int itemLoc; + struct _drmBONode *node; + uint32_t *reloc; + struct drm_bo_info_req *req; + + driBOAddListItem(batch->list, driBO, val_flags, val_mask, + &itemLoc, &node); + req = &node->bo_arg.d.req.bo_req; + + if (!(req->hint & DRM_BO_HINT_PRESUMED_OFFSET)) { + + /* + * Stop other threads from tampering with the underlying + * drmBO while we're reading its offset. + */ + + driReadLockKernelBO(); + req->presumed_offset = (uint64_t) driBOKernel(driBO)->offset; + driReadUnlockKernelBO(); + req->hint = DRM_BO_HINT_PRESUMED_OFFSET; + } + + pre_add += driBOPoolOffset(driBO); + + if (batch->nr_relocs == batch->reloc_size) + intel_realloc_relocs(batch, batch->reloc_size * 2); + + reloc = batch->reloc + + (I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE); + + reloc[0] = ((uint8_t *)batch->ptr - batch->drmBOVirtual); + intel_batchbuffer_emit_dword(batch, req->presumed_offset + pre_add); + reloc[1] = pre_add; + reloc[2] = itemLoc; + reloc[3] = batch->dest_location; + batch->nr_relocs++; +} static void -intel_batch_ioctl(struct intel_context *intel, - uint start_offset, uint used, boolean allow_unlock) +i915_drm_copy_reply(const struct drm_bo_info_rep * rep, drmBO * buf) { - drmI830BatchBuffer batch; - - batch.start = start_offset; - batch.used = used; - batch.cliprects = NULL; /* unused */ - batch.num_cliprects = 0; - batch.DR1 = 0; - batch.DR4 = 0; /* still need this ? */ - - DBG(IOCTL, "%s: 0x%x..0x%x DR4: %x cliprects: %d\n", - __FUNCTION__, - batch.start, - batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); - - if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, - sizeof(batch))) { - printf("DRM_I830_BATCHBUFFER: %d\n", -errno); - UNLOCK_HARDWARE(intel); - exit(1); - } + buf->handle = rep->handle; + buf->flags = rep->flags; + buf->size = rep->size; + buf->offset = rep->offset; + buf->mapHandle = rep->arg_handle; + buf->proposedFlags = rep->proposed_flags; + buf->start = rep->buffer_start; + buf->fenceFlags = rep->fence_flags; + buf->replyFlags = rep->rep_flags; + buf->pageAlignment = rep->page_alignment; } +static int +i915_execbuf(struct intel_batchbuffer *batch, + GLuint used, + GLboolean ignore_cliprects, + drmBOList *list, + struct drm_i915_execbuffer *ea) +{ + struct intel_context *intel = batch->intel; + drmBONode *node; + drmMMListHead *l; + struct drm_i915_op_arg *arg, *first; + struct drm_bo_op_req *req; + struct drm_bo_info_rep *rep; + uint64_t *prevNext = NULL; + drmBO *buf; + int ret = 0; + uint32_t count = 0; + + first = NULL; + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + + arg = &node->bo_arg; + req = &arg->d.req; + + if (!first) + first = arg; + + if (prevNext) + *prevNext = (unsigned long)arg; + + prevNext = &arg->next; + req->bo_req.handle = node->buf->handle; + req->op = drm_bo_validate; + req->bo_req.flags = node->arg0; + req->bo_req.mask = node->arg1; + req->bo_req.hint |= 0; + count++; + } + + memset(ea, 0, sizeof(*ea)); + ea->num_buffers = count; + ea->batch.start = batch->poolOffset; + ea->batch.used = used; +#if 0 /* ZZZ JB: no cliprects used */ + ea->batch.cliprects = intel->pClipRects; + ea->batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + ea->batch.DR1 = 0; + ea->batch.DR4 = 0;((((GLuint) intel->drawX) & 0xffff) | + (((GLuint) intel->drawY) << 16)); +#else + ea->batch.cliprects = NULL; + ea->batch.num_cliprects = 0; + ea->batch.DR1 = 0; + ea->batch.DR4 = 0; +#endif + ea->fence_arg.flags = DRM_I915_FENCE_FLAG_FLUSHED; + ea->ops_list = (unsigned long) first; + first->reloc_ptr = (unsigned long) batch->reloc; + batch->reloc[0] = batch->nr_relocs; + + //return -EFAULT; + do { + ret = drmCommandWriteRead(intel->driFd, DRM_I915_EXECBUFFER, ea, + sizeof(*ea)); + } while (ret == -EAGAIN); + + if (ret != 0) + return ret; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + arg = &node->bo_arg; + rep = &arg->d.rep.bo_info; + + if (!arg->handled) { + return -EFAULT; + } + if (arg->d.rep.ret) + return arg->d.rep.ret; + + buf = node->buf; + i915_drm_copy_reply(rep, buf); + } + return 0; +} /* TODO: Push this whole function into bufmgr. */ -static void +static struct _DriFenceObject * do_flush_locked(struct intel_batchbuffer *batch, - uint used, boolean allow_unlock) + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) { - uint *ptr; - uint i, fenceFlags; + struct intel_context *intel = batch->intel; struct _DriFenceObject *fo; + drmFence fence; + drmBOList *boList; + struct drm_i915_execbuffer ea; + int ret = 0; + + driBOValidateUserList(batch->list); + boList = driGetdrmBOList(batch->list); + +#if 0 /* ZZZ JB Allways run */ + if (!(intel->numClipRects == 0 && !ignore_cliprects)) { +#else + if (1) { +#endif + ret = i915_execbuf(batch, used, ignore_cliprects, boList, &ea); + } else { + driPutdrmBOList(batch->list); + fo = NULL; + goto out; + } + driPutdrmBOList(batch->list); + if (ret) + abort(); - driBOValidateList(batch->intel->driFd, &batch->list); - - /* Apply the relocations. This nasty map indicates to me that the - * whole task should be done internally by the memory manager, and - * that dma buffers probably need to be pinned within agp space. - */ - ptr = (uint *) driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, - DRM_BO_HINT_ALLOW_UNFENCED_MAP); + if (ea.fence_arg.error != 0) { - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; + /* + * The hardware has been idled by the kernel. + * Don't fence the driBOs. + */ - ptr[r->offset / 4] = driBOOffset(r->buf) + r->delta; + if (batch->last_fence) + driFenceUnReference(&batch->last_fence); +#if 0 /* ZZZ JB: no _mesa_* funcs in gallium */ + _mesa_printf("fence error\n"); +#endif + batch->last_fence = NULL; + fo = NULL; + goto out; } - if (0) - intel_dump_batchbuffer(0, ptr, used); + fence.handle = ea.fence_arg.handle; + fence.fence_class = ea.fence_arg.fence_class; + fence.type = ea.fence_arg.type; + fence.flags = ea.fence_arg.flags; + fence.signaled = ea.fence_arg.signaled; - driBOUnmap(batch->buffer); - batch->map = NULL; - - intel_batch_ioctl(batch->intel, - driBOOffset(batch->buffer), - used, allow_unlock); - - /* - * Kernel fencing. The flags tells the kernel that we've - * programmed an MI_FLUSH. - */ - fenceFlags = DRM_I915_FENCE_FLAG_FLUSHED; - fo = driFenceBuffers(batch->intel->driFd, "Batch fence", fenceFlags); + fo = driBOFenceUserList(batch->intel->intelScreen->mgr, batch->list, + "SuperFence", &fence); + if (driFenceType(fo) & DRM_I915_FENCE_TYPE_RW) { + if (batch->last_fence) + driFenceUnReference(&batch->last_fence); /* - * User space fencing. - */ - driBOFence(batch->buffer, fo); - - if (driFenceType(fo) == DRM_FENCE_TYPE_EXE) { - /* - * Oops. We only validated a batch buffer. This means we - * didn't do any proper rendering. Discard this fence object. - */ - driFenceUnReference(fo); - } - else { - driFenceUnReference(batch->last_fence); - batch->last_fence = fo; - for (i = 0; i < batch->nr_relocs; i++) { - struct buffer_reloc *r = &batch->reloc[i]; - driBOFence(r->buf, fo); - } - } + * FIXME: Context last fence?? + */ + batch->last_fence = fo; + driFenceReference(fo); + } + out: +#if 0 /* ZZZ JB: fix this */ + intel->vtbl.lost_hardware(intel); +#else + (void)intel; +#endif + return fo; } @@ -263,29 +382,41 @@ struct _DriFenceObject * intel_batchbuffer_flush(struct intel_batchbuffer *batch) { struct intel_context *intel = batch->intel; - uint used = batch->ptr - batch->map; - const boolean was_locked = intel->locked; + GLuint used = batch->ptr - batch->map; + GLboolean was_locked = intel->locked; + struct _DriFenceObject *fence; if (used == 0) return batch->last_fence; -#define MI_FLUSH ((0 << 29) | (4 << 23)) - /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a * performance drain that we would like to avoid. */ +#if 0 /* ZZZ JB: what should we do here? */ if (used & 4) { - ((int *) batch->ptr)[0] = MI_FLUSH; + ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); ((int *) batch->ptr)[1] = 0; ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; used += 12; } else { - ((int *) batch->ptr)[0] = MI_FLUSH; + ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; used += 8; } - +#else + if (used & 4) { + ((int *) batch->ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->ptr)[1] = 0; + ((int *) batch->ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; + used += 8; + } +#endif driBOUnmap(batch->buffer); batch->ptr = NULL; batch->map = NULL; @@ -296,7 +427,8 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch) if (!was_locked) LOCK_HARDWARE(intel); - do_flush_locked(batch, used, GL_FALSE); + fence = do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), + GL_FALSE); if (!was_locked) UNLOCK_HARDWARE(intel); @@ -304,52 +436,20 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch) /* Reset the buffer: */ intel_batchbuffer_reset(batch); - return batch->last_fence; + return fence; } - void intel_batchbuffer_finish(struct intel_batchbuffer *batch) { struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); - if (fence) { - driFenceReference(fence); - driFenceFinish(fence, - DRM_FENCE_TYPE_EXE | DRM_I915_FENCE_TYPE_RW, - GL_FALSE); - driFenceUnReference(fence); - } + driFenceFinish(fence, driFenceType(fence), GL_FALSE); + driFenceUnReference(&fence); } - -/* This is the only way buffers get added to the validate list. - */ -boolean -intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, - struct _DriBufferObject *buffer, - uint flags, uint mask, uint delta) -{ - assert(batch->nr_relocs < MAX_RELOCS); - - driBOAddListItem(&batch->list, buffer, flags, mask); - - { - struct buffer_reloc *r = &batch->reloc[batch->nr_relocs++]; - driBOReference(buffer); - r->buf = buffer; - r->offset = batch->ptr - batch->map; - r->delta = delta; - *(uint *) batch->ptr = 0x12345678; - } - - batch->ptr += 4; - return GL_TRUE; -} - - void intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, uint bytes, uint flags) + const void *data, GLuint bytes, GLuint flags) { assert((bytes & 3) == 0); intel_batchbuffer_require_space(batch, bytes, flags); diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.h b/src/gallium/winsys/dri/intel/intel_batchbuffer.h index caf6870a3c..9e4b8043bf 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.h +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.h @@ -1,54 +1,20 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - #ifndef INTEL_BATCHBUFFER_H #define INTEL_BATCHBUFFER_H -#include "pipe/p_debug.h" -#include "pipe/p_compiler.h" -#include "dri_bufmgr.h" +#include "mtypes.h" +#include "ws_dri_bufmgr.h" struct intel_context; #define BATCH_SZ 16384 #define BATCH_RESERVED 16 -#define MAX_RELOCS 4096 +#define INTEL_DEFAULT_RELOCS 100 +#define INTEL_MAX_RELOCS 400 #define INTEL_BATCH_NO_CLIPRECTS 0x1 #define INTEL_BATCH_CLIPRECTS 0x2 -struct buffer_reloc -{ - struct _DriBufferObject *buf; - uint offset; - uint delta; /* not needed? */ -}; - struct intel_batchbuffer { struct bufmgr *bm; @@ -56,19 +22,30 @@ struct intel_batchbuffer struct _DriBufferObject *buffer; struct _DriFenceObject *last_fence; - uint flags; + GLuint flags; - drmBOList list; - uint list_count; - ubyte *map; - ubyte *ptr; + struct _DriBufferList *list; + GLuint list_count; + GLubyte *map; + GLubyte *ptr; - struct buffer_reloc reloc[MAX_RELOCS]; - uint nr_relocs; - uint size; + uint32_t *reloc; + GLuint reloc_size; + GLuint nr_relocs; + + GLuint size; + + GLuint dirty_state; + GLuint id; + + uint32_t poolOffset; + uint8_t *drmBOVirtual; + struct _drmBONode *node; /* Validation list node for this buffer */ + int dest_location; /* Validation list sequence for this buffer */ }; -struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context *intel); +struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context + *intel); void intel_batchbuffer_free(struct intel_batchbuffer *batch); @@ -82,26 +59,28 @@ void intel_batchbuffer_reset(struct intel_batchbuffer *batch); /* Unlike bmBufferData, this currently requires the buffer be mapped. - * Consider it a convenience function wrapping multiple + * Consider it a convenience function wrapping multple * intel_buffer_dword() calls. */ void intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, uint bytes, uint flags); + const void *data, GLuint bytes, GLuint flags); void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, - uint bytes); + GLuint bytes); -boolean intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch, - struct _DriBufferObject *buffer, - uint flags, - uint mask, uint offset); +void +intel_offset_relocation(struct intel_batchbuffer *batch, + unsigned pre_add, + struct _DriBufferObject *driBO, + uint64_t val_flags, + uint64_t val_mask); /* Inline functions - might actually be better off with these * non-inlined. Certainly better off switching all command packets to * be passed as structs rather than dwords, but that's a little bit of * work... */ -static INLINE uint +static INLINE GLuint intel_batchbuffer_space(struct intel_batchbuffer *batch) { return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); @@ -109,22 +88,26 @@ intel_batchbuffer_space(struct intel_batchbuffer *batch) static INLINE void -intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, uint dword) +intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) { assert(batch->map); assert(intel_batchbuffer_space(batch) >= 4); - *(uint *) (batch->ptr) = dword; + *(GLuint *) (batch->ptr) = dword; batch->ptr += 4; } static INLINE void intel_batchbuffer_require_space(struct intel_batchbuffer *batch, - uint sz, uint flags) + GLuint sz, GLuint flags) { + struct _DriFenceObject *fence; + assert(sz < batch->size - 8); if (intel_batchbuffer_space(batch) < sz || - (batch->flags != 0 && flags != 0 && batch->flags != flags)) - intel_batchbuffer_flush(batch); + (batch->flags != 0 && flags != 0 && batch->flags != flags)) { + fence = intel_batchbuffer_flush(batch); + driFenceUnReference(&fence); + } batch->flags |= flags; } @@ -134,14 +117,15 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch, #define BATCH_LOCALS #define BEGIN_BATCH(n, flags) do { \ + assert(!intel->prim.flush); \ intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ } while (0) #define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) #define OUT_RELOC(buf,flags,mask,delta) do { \ - assert((delta) >= 0); \ - intel_batchbuffer_emit_reloc(intel->batch, buf, flags, mask, delta); \ + assert((delta) >= 0); \ + intel_offset_relocation(intel->batch, delta, buf, flags, mask); \ } while (0) #define ADVANCE_BATCH() do { } while(0) diff --git a/src/gallium/winsys/dri/intel/intel_batchpool.c b/src/gallium/winsys/dri/intel/intel_batchpool.c deleted file mode 100644 index ce154c7b88..0000000000 --- a/src/gallium/winsys/dri/intel/intel_batchpool.c +++ /dev/null @@ -1,427 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -/** - * XXX NOTE: there are no intel dependencies in this file. - * Rename to dri_batchpool.c? - */ - -#include -#include -#include -#include - -#include "pipe/p_compiler.h" -#include "pipe/p_thread.h" - -#include "dri_bufpool.h" -#include "dri_bufmgr.h" -#include "intel_batchpool.h" - - -typedef struct -{ - drmMMListHead head; - struct _BPool *parent; - struct _DriFenceObject *fence; - unsigned long start; - int unfenced; - int mapped; -} BBuf; - -typedef struct _BPool -{ - _glthread_Mutex mutex; - unsigned long bufSize; - unsigned poolSize; - unsigned numFree; - unsigned numTot; - unsigned numDelayed; - unsigned checkDelayed; - drmMMListHead free; - drmMMListHead delayed; - drmMMListHead head; - drmBO kernelBO; - void *virtual; - BBuf *bufs; -} BPool; - - -static BPool * -createBPool(int fd, unsigned long bufSize, unsigned numBufs, unsigned flags, - unsigned checkDelayed) -{ - BPool *p = (BPool *) malloc(sizeof(*p)); - BBuf *buf; - int i; - - if (!p) - return NULL; - - p->bufs = (BBuf *) malloc(numBufs * sizeof(*p->bufs)); - if (!p->bufs) { - free(p); - return NULL; - } - - DRMINITLISTHEAD(&p->free); - DRMINITLISTHEAD(&p->head); - DRMINITLISTHEAD(&p->delayed); - - p->numTot = numBufs; - p->numFree = numBufs; - p->bufSize = bufSize; - p->numDelayed = 0; - p->checkDelayed = checkDelayed; - - _glthread_INIT_MUTEX(p->mutex); - - if (drmBOCreate(fd, 0, numBufs * bufSize, 0, NULL, drm_bo_type_dc, - flags, DRM_BO_HINT_DONT_FENCE, &p->kernelBO)) { - free(p->bufs); - free(p); - return NULL; - } - if (drmBOMap(fd, &p->kernelBO, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, - &p->virtual)) { - drmBODestroy(fd, &p->kernelBO); - free(p->bufs); - free(p); - return NULL; - } - - /* - * We unmap the buffer so that we can validate it later. Note that this is - * just a synchronizing operation. The buffer will have a virtual mapping - * until it is destroyed. - */ - - drmBOUnmap(fd, &p->kernelBO); - - buf = p->bufs; - for (i = 0; i < numBufs; ++i) { - buf->parent = p; - buf->fence = NULL; - buf->start = i * bufSize; - buf->mapped = 0; - buf->unfenced = 0; - DRMLISTADDTAIL(&buf->head, &p->free); - buf++; - } - - return p; -} - - -static void -pool_checkFree(BPool * p, int wait) -{ - drmMMListHead *list, *prev; - BBuf *buf; - int signaled = 0; - int i; - - list = p->delayed.next; - - if (p->numDelayed > 3) { - for (i = 0; i < p->numDelayed; i += 3) { - list = list->next; - } - } - - prev = list->prev; - for (; list != &p->delayed; list = prev, prev = list->prev) { - - buf = DRMLISTENTRY(BBuf, list, head); - - if (!signaled) { - if (wait) { - driFenceFinish(buf->fence, DRM_FENCE_TYPE_EXE, 1); - signaled = 1; - } - else { - signaled = driFenceSignaled(buf->fence, DRM_FENCE_TYPE_EXE); - } - } - - if (!signaled) - break; - - driFenceUnReference(buf->fence); - buf->fence = NULL; - DRMLISTDEL(list); - p->numDelayed--; - DRMLISTADD(list, &p->free); - p->numFree++; - } -} - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, unsigned flags, unsigned hint, - unsigned alignment) -{ - BPool *p = (BPool *) pool->data; - - drmMMListHead *item; - - if (alignment && (alignment != 4096)) - return NULL; - - _glthread_LOCK_MUTEX(p->mutex); - - if (p->numFree == 0) - pool_checkFree(p, TRUE); - - if (p->numFree == 0) { - fprintf(stderr, "Out of fixed size buffer objects\n"); - BM_CKFATAL(-ENOMEM); - } - - item = p->free.next; - - if (item == &p->free) { - fprintf(stderr, "Fixed size buffer pool corruption\n"); - } - - DRMLISTDEL(item); - --p->numFree; - - _glthread_UNLOCK_MUTEX(p->mutex); - return (void *) DRMLISTENTRY(BBuf, item, head); -} - - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - - if (buf->fence) { - DRMLISTADDTAIL(&buf->head, &p->delayed); - p->numDelayed++; - } - else { - buf->unfenced = 0; - DRMLISTADD(&buf->head, &p->free); - p->numFree++; - } - - if ((p->numDelayed % p->checkDelayed) == 0) - pool_checkFree(p, 0); - - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, void **virtual) -{ - - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - - /* - * Currently Mesa doesn't have any condition variables to resolve this - * cleanly in a multithreading environment. - * We bail out instead. - */ - - if (buf->mapped) { - fprintf(stderr, "Trying to map already mapped buffer object\n"); - BM_CKFATAL(-EINVAL); - } - -#if 0 - if (buf->unfenced && !(hint & DRM_BO_HINT_ALLOW_UNFENCED_MAP)) { - fprintf(stderr, "Trying to map an unfenced buffer object 0x%08x" - " 0x%08x %d\n", hint, flags, buf->start); - BM_CKFATAL(-EINVAL); - } - -#endif - - if (buf->fence) { - _glthread_UNLOCK_MUTEX(p->mutex); - return -EBUSY; - } - - buf->mapped = TRUE; - *virtual = (unsigned char *) p->virtual + buf->start; - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, int lazy) -{ - BBuf *buf = (BBuf *) private; - driFenceFinish(buf->fence, 0x0, lazy); - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - - buf->mapped = 0; - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - return p->kernelBO.offset + buf->start; -} - -static unsigned -pool_flags(struct _DriBufferPool *pool, void *private) -{ - BPool *p = (BPool *) pool->data; - - return p->kernelBO.flags; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - BPool *p = (BPool *) pool->data; - - return p->bufSize; -} - - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - _glthread_LOCK_MUTEX(p->mutex); - if (buf->fence) { - driFenceUnReference(buf->fence); - } - buf->fence = fence; - buf->unfenced = 0; - driFenceReference(buf->fence); - _glthread_UNLOCK_MUTEX(p->mutex); - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - - return &p->kernelBO; -} - -static int -pool_validate(struct _DriBufferPool *pool, void *private) -{ - BBuf *buf = (BBuf *) private; - BPool *p = buf->parent; - _glthread_LOCK_MUTEX(p->mutex); - buf->unfenced = TRUE; - _glthread_UNLOCK_MUTEX(p->mutex); - return 0; -} - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - BPool *p = (BPool *) pool->data; - - /* - * Wait on outstanding fences. - */ - - _glthread_LOCK_MUTEX(p->mutex); - while ((p->numFree < p->numTot) && p->numDelayed) { - _glthread_UNLOCK_MUTEX(p->mutex); - sched_yield(); - pool_checkFree(p, TRUE); - _glthread_LOCK_MUTEX(p->mutex); - } - - drmBODestroy(pool->fd, &p->kernelBO); - free(p->bufs); - _glthread_UNLOCK_MUTEX(p->mutex); - free(p); - free(pool); -} - - -struct _DriBufferPool * -driBatchPoolInit(int fd, unsigned flags, - unsigned long bufSize, - unsigned numBufs, unsigned checkDelayed) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - if (!pool) - return NULL; - - pool->data = createBPool(fd, bufSize, numBufs, flags, checkDelayed); - if (!pool->data) - return NULL; - - pool->fd = fd; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = &pool_validate; - pool->waitIdle = &pool_waitIdle; - pool->setstatic = NULL; - pool->takeDown = &pool_takedown; - return pool; -} diff --git a/src/gallium/winsys/dri/intel/intel_batchpool.h b/src/gallium/winsys/dri/intel/intel_batchpool.h deleted file mode 100644 index f6a95723bc..0000000000 --- a/src/gallium/winsys/dri/intel/intel_batchpool.h +++ /dev/null @@ -1,37 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_BATCHPOOL_H -#define INTEL_BATCHPOOL_H - -extern struct _DriBufferPool *driBatchPoolInit(int fd, unsigned flags, - unsigned long bufSize, - unsigned numBufs, - unsigned checkDelayed); - - -#endif /* INTEL_BATCHPOOL_H */ diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index 8eba33c313..cc2fc11c5e 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -162,6 +162,7 @@ intelCreateContext(const __GLcontextModes * visual, * memory pools */ DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + // ZZZ JB should be per screen and not be done per context havePools = intelCreatePools(sPriv); DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); if (!havePools) @@ -234,12 +235,12 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) if (intel->last_swap_fence) { driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->last_swap_fence); + driFenceUnReference(&intel->last_swap_fence); intel->last_swap_fence = NULL; } if (intel->first_swap_fence) { driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(intel->first_swap_fence); + driFenceUnReference(&intel->first_swap_fence); intel->first_swap_fence = NULL; } diff --git a/src/gallium/winsys/dri/intel/intel_context.h b/src/gallium/winsys/dri/intel/intel_context.h index 45430984d8..597dc13830 100644 --- a/src/gallium/winsys/dri/intel/intel_context.h +++ b/src/gallium/winsys/dri/intel/intel_context.h @@ -28,7 +28,7 @@ #ifndef INTEL_CONTEXT_H #define INTEL_CONTEXT_H - +#include #include "drm.h" #include "pipe/p_debug.h" diff --git a/src/gallium/winsys/dri/intel/intel_ioctl.c b/src/gallium/winsys/dri/intel/intel_ioctl.c new file mode 100644 index 0000000000..3250c6b3a9 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_ioctl.c @@ -0,0 +1,138 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include +#include +#include +#include + +#include "mtypes.h" +#include "context.h" +#include "swrast/swrast.h" + +#include "intel_context.h" +#include "intel_ioctl.h" +#include "intel_batchbuffer.h" +#include "intel_blit.h" +#include "intel_regions.h" +#include "drm.h" + +#define FILE_DEBUG_FLAG DEBUG_IOCTL + +int +intelEmitIrqLocked(struct intel_context *intel) +{ + drmI830IrqEmit ie; + int ret, seq; + + assert(((*(int *) intel->driHwLock) & ~DRM_LOCK_CONT) == + (DRM_LOCK_HELD | intel->hHWContext)); + + ie.irq_seq = &seq; + + ret = drmCommandWriteRead(intel->driFd, DRM_I830_IRQ_EMIT, + &ie, sizeof(ie)); + if (ret) { + fprintf(stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret); + exit(1); + } + + DBG("%s --> %d\n", __FUNCTION__, seq); + + return seq; +} + +void +intelWaitIrq(struct intel_context *intel, int seq) +{ + int ret; + + DBG("%s %d\n", __FUNCTION__, seq); + + intel->iw.irq_seq = seq; + + do { + ret = + drmCommandWrite(intel->driFd, DRM_I830_IRQ_WAIT, &intel->iw, + sizeof(intel->iw)); + } while (ret == -EAGAIN || ret == -EINTR); + + if (ret) { + fprintf(stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret); + exit(1); + } +} + + +void +intel_batch_ioctl(struct intel_context *intel, + GLuint start_offset, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) +{ + drmI830BatchBuffer batch; + + assert(intel->locked); + assert(used); + + DBG("%s used %d offset %x..%x ignore_cliprects %d\n", + __FUNCTION__, + used, start_offset, start_offset + used, ignore_cliprects); + + /* Throw away non-effective packets. Won't work once we have + * hardware contexts which would preserve statechanges beyond a + * single buffer. + */ + + + + batch.start = start_offset; + batch.used = used; + batch.cliprects = intel->pClipRects; + batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + batch.DR1 = 0; + batch.DR4 = ((((GLuint) intel->drawX) & 0xffff) | + (((GLuint) intel->drawY) << 16)); + + DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n", + __FUNCTION__, + batch.start, + batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); + + if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, + sizeof(batch))) { + fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n", -errno); + UNLOCK_HARDWARE(intel); + exit(1); + } + + /* FIXME: use hardware contexts to avoid 'losing' hardware after + * each buffer flush. + */ + intel->vtbl.lost_hardware(intel); +} diff --git a/src/gallium/winsys/dri/intel/intel_ioctl.h b/src/gallium/winsys/dri/intel/intel_ioctl.h new file mode 100644 index 0000000000..e8d07de893 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_ioctl.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_IOCTL_H +#define INTEL_IOCTL_H + +#include "intel_context.h" + +void intelWaitIrq(struct intel_context *intel, int seq); +int intelEmitIrqLocked(struct intel_context *intel); + +void intel_batch_ioctl(struct intel_context *intel, + GLuint start_offset, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock); +#endif diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 9e31c013a9..a3fc6d7344 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -32,12 +32,12 @@ #include "intel_context.h" #include "intel_screen.h" #include "intel_batchbuffer.h" -#include "intel_batchpool.h" +//#include "intel_batchpool.h" #include "intel_swapbuffers.h" #include "intel_winsys.h" #include "i830_dri.h" -#include "dri_bufpool.h" +#include "ws_dri_bufpool.h" #include "pipe/p_context.h" #include "state_tracker/st_public.h" @@ -132,6 +132,7 @@ intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) assert( sarea->front_size >= intelScreen->front.pitch * intelScreen->front.height ); +#if 0 /* JB not important */ if (!sarea->front_handle) return; @@ -142,30 +143,41 @@ intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) fprintf(stderr, "drmMap(frontbuffer) failed!\n"); return; } +#endif +#if 0 /* JB */ if (intelScreen->staticPool) { driGenBuffers(intelScreen->staticPool, "static region", 1, &intelScreen->front.buffer, 64, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - + driBOSetStatic(intelScreen->front.buffer, intelScreen->front.offset, intelScreen->front.pitch * intelScreen->front.height, intelScreen->front.map, 0); } +#else + if (intelScreen->staticPool) { + if (intelScreen->front.buffer) + driBOUnReference(intelScreen->front.buffer); + driGenBuffers(intelScreen->staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); + driBOSetReferenced(intelScreen->front.buffer, sarea->front_bo_handle); + } +#endif } boolean intelCreatePools(__DRIscreenPrivate * sPriv) { - unsigned batchPoolSize = 1024*1024; + //unsigned batchPoolSize = 1024*1024; struct intel_screen *intelScreen = intel_screen(sPriv); if (intelScreen->havePools) return GL_TRUE; +#if 0 /* ZZZ JB fix this */ intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); if (!intelScreen->staticPool) return GL_FALSE; @@ -181,7 +193,17 @@ intelCreatePools(__DRIscreenPrivate * sPriv) fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); return GL_FALSE; } - +#else + intelScreen->staticPool = driDRMPoolInit(sPriv->fd); + intelScreen->batchPool = driSlabPoolInit(sPriv->fd, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + 4096, //intelScreen->maxBatchSize, + 1, 40, 16*16384, 0, + intelScreen->fMan); +#endif intelScreen->havePools = GL_TRUE; intelUpdateScreenRotation(sPriv, intelScreen->sarea); @@ -240,7 +262,26 @@ intelInitDriver(__DRIscreenPrivate * sPriv) (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); } - intelScreen->winsys = intel_create_pipe_winsys(sPriv->fd); + + +#if 1 // ZZZ JB + intelScreen->mgr = driFenceMgrTTMInit(sPriv->fd); + if (!intelScreen->mgr) { + fprintf(stderr, "Failed to create fence manager.\n"); + return GL_FALSE; + } + + intelScreen->fMan = driInitFreeSlabManager(10, 10); + if (!intelScreen->fMan) { + fprintf(stderr, "Failed to create free slab manager.\n"); + return GL_FALSE; + } + + if (!intelCreatePools(sPriv)) + return GL_FALSE; +#endif + + intelScreen->winsys = intel_create_pipe_winsys(sPriv->fd, intelScreen->fMan); return GL_TRUE; } diff --git a/src/gallium/winsys/dri/intel/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h index 3396f9e564..c17f0202e4 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.h +++ b/src/gallium/winsys/dri/intel/intel_screen.h @@ -31,7 +31,7 @@ #include "dri_util.h" #include "i830_common.h" #include "xmlconfig.h" -#include "dri_bufpool.h" +#include "ws_dri_bufpool.h" #include "pipe/p_compiler.h" @@ -74,6 +74,13 @@ struct intel_screen */ struct intel_context *dummyContext; + /* + * New stuff form the i915tex integration + */ + struct _DriFenceMgr *mgr; + struct _DriFreeSlabManager *fMan; + unsigned batch_id; + struct pipe_winsys *winsys; }; diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index 56b86d6a63..1a42389169 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -63,7 +63,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, if (intel->last_swap_fence) { driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); - driFenceUnReference(intel->last_swap_fence); + driFenceUnReference(&intel->last_swap_fence); intel->last_swap_fence = NULL; } intel->last_swap_fence = intel->first_swap_fence; @@ -178,7 +178,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, } if (intel->first_swap_fence) - driFenceUnReference(intel->first_swap_fence); + driFenceUnReference(&intel->first_swap_fence); intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); driFenceReference(intel->first_swap_fence); } diff --git a/src/gallium/winsys/dri/intel/intel_winsys.h b/src/gallium/winsys/dri/intel/intel_winsys.h index ffc40782be..d0a319f9a4 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys.h +++ b/src/gallium/winsys/dri/intel/intel_winsys.h @@ -37,7 +37,7 @@ struct pipe_buffer; struct _DriBufferObject; struct pipe_winsys * -intel_create_pipe_winsys( int fd ); +intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ); void intel_destroy_pipe_winsys( struct pipe_winsys *winsys ); diff --git a/src/gallium/winsys/dri/intel/intel_winsys_i915.c b/src/gallium/winsys/dri/intel/intel_winsys_i915.c index 4d183db7c3..ae3e9654d3 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_i915.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_i915.c @@ -31,8 +31,8 @@ #include #include -#include "dri_bufpool.h" -#include "dri_bufmgr.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" #include "intel_context.h" #include "intel_batchbuffer.h" @@ -106,10 +106,18 @@ static void intel_i915_batch_reloc( struct i915_winsys *sws, mask |= DRM_BO_FLAG_READ; } - intel_batchbuffer_emit_reloc( intel->batch, +#if 0 /* JB old */ + intel_batchbuffer_emit_reloc( intel->batch, dri_bo( buf ), - flags, mask, + flags, mask, delta ); +#else /* new */ + intel_offset_relocation( intel->batch, + delta, + dri_bo( buf ), + flags, + mask ); +#endif } @@ -126,8 +134,18 @@ static void intel_i915_batch_flush( struct i915_winsys *sws, fu.dri = intel_batchbuffer_flush( intel->batch ); - if (fu.dri) - iws->pws->fence_reference(iws->pws, fence, fu.pipe); + if (!fu.dri && fence) { + *fence = NULL; + return; + } + + if (fu.dri) { + if (fence) + *fence = fu.pipe; + else + iws->pws->fence_reference(iws->pws, &fu.dri, NULL); + } + // if (0) intel_i915_batch_wait_idle( sws ); } diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 77dec9488d..59729a0245 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -31,8 +31,8 @@ #include #include -#include "dri_bufpool.h" -#include "dri_bufmgr.h" +//#include "dri_bufpool.h" +//#include "dri_bufmgr.h" #include "intel_context.h" #include "intel_winsys.h" @@ -50,6 +50,7 @@ struct intel_pipe_winsys { struct pipe_winsys winsys; struct _DriBufferPool *regionPool; + struct _DriFreeSlabManager *fMan; }; @@ -142,7 +143,7 @@ intel_buffer_create(struct pipe_winsys *winsys, driGenBuffers( iws->regionPool, "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); - driBOData( buffer->driBO, size, NULL, 0 ); + driBOData( buffer->driBO, size, NULL, iws->regionPool, 0 ); return &buffer->base; } @@ -155,7 +156,7 @@ intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); driGenUserBuffer( iws->regionPool, - "pipe user buffer", &buffer->driBO, ptr, bytes); + "pipe user buffer", &buffer->driBO, ptr, bytes ); return &buffer->base; } @@ -209,7 +210,7 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, unsigned flags) { const unsigned alignment = 64; - int ret; + //int ret; surf->width = width; surf->height = height; @@ -249,9 +250,37 @@ intel_get_name( struct pipe_winsys *winsys ) return "Intel/DRI/ttm"; } +static void +intel_fence_reference( struct pipe_winsys *sws, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence ) +{ + if (*ptr) + driFenceUnReference((struct _DriFenceObject **)ptr); + + if (fence) + *ptr = (struct pipe_fence_handle *)driFenceReference((struct _DriFenceObject *)fence); +} + +static int +intel_fence_signalled( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ) +{ + return driFenceSignaled((struct _DriFenceObject *)fence, flag); +} + +static int +intel_fence_finish( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ) +{ + /* JB: Lets allways lazy wait */ + return driFenceFinish((struct _DriFenceObject *)fence, flag, 1); +} struct pipe_winsys * -intel_create_pipe_winsys( int fd ) +intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) { struct intel_pipe_winsys *iws = CALLOC_STRUCT( intel_pipe_winsys ); @@ -273,8 +302,20 @@ intel_create_pipe_winsys( int fd ) iws->winsys.surface_alloc_storage = intel_i915_surface_alloc_storage; iws->winsys.surface_release = intel_i915_surface_release; + iws->winsys.fence_reference = intel_fence_reference; + iws->winsys.fence_signalled = intel_fence_signalled; + iws->winsys.fence_finish = intel_fence_finish; + if (fd) - iws->regionPool = driDRMPoolInit(fd); + iws->regionPool = driSlabPoolInit(fd, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + 64, 6, 16, 4096, 0, + fMan); return &iws->winsys; } diff --git a/src/gallium/winsys/dri/intel/server/i830_common.h b/src/gallium/winsys/dri/intel/server/i830_common.h index d4d58886ce..f1fd3939ab 100644 --- a/src/gallium/winsys/dri/intel/server/i830_common.h +++ b/src/gallium/winsys/dri/intel/server/i830_common.h @@ -54,6 +54,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define DRM_I830_DESTROY_HEAP 0x0c #define DRM_I830_SET_VBLANK_PIPE 0x0d #define DRM_I830_GET_VBLANK_PIPE 0x0e +#define DRM_I830_MMIO 0x10 typedef struct { enum { @@ -85,6 +86,7 @@ typedef struct { int last_enqueue; /* last time a buffer was enqueued */ int last_dispatch; /* age of the most recently dispatched buffer */ int ctxOwner; /* last context to upload state */ + /** Last context that used the buffer manager. */ int texAge; int pf_enabled; /* is pageflipping allowed? */ int pf_active; @@ -121,20 +123,29 @@ typedef struct { unsigned int rotated_tiled; unsigned int rotated2_tiled; - int pipeA_x; - int pipeA_y; - int pipeA_w; - int pipeA_h; - int pipeB_x; - int pipeB_y; - int pipeB_w; - int pipeB_h; + int planeA_x; + int planeA_y; + int planeA_w; + int planeA_h; + int planeB_x; + int planeB_y; + int planeB_w; + int planeB_h; /* Triple buffering */ drm_handle_t third_handle; int third_offset; int third_size; unsigned int third_tiled; + + /* buffer object handles for the static buffers. May change + * over the lifetime of the client, though it doesn't in our current + * implementation. + */ + unsigned int front_bo_handle; + unsigned int back_bo_handle; + unsigned int third_bo_handle; + unsigned int depth_bo_handle; } drmI830Sarea; /* Flags for perf_boxes @@ -223,4 +234,23 @@ typedef struct { int pipe; } drmI830VBlankPipe; +#define MMIO_READ 0 +#define MMIO_WRITE 1 + +#define MMIO_REGS_IA_PRIMATIVES_COUNT 0 +#define MMIO_REGS_IA_VERTICES_COUNT 1 +#define MMIO_REGS_VS_INVOCATION_COUNT 2 +#define MMIO_REGS_GS_PRIMITIVES_COUNT 3 +#define MMIO_REGS_GS_INVOCATION_COUNT 4 +#define MMIO_REGS_CL_PRIMITIVES_COUNT 5 +#define MMIO_REGS_CL_INVOCATION_COUNT 6 +#define MMIO_REGS_PS_INVOCATION_COUNT 7 +#define MMIO_REGS_PS_DEPTH_COUNT 8 + +typedef struct { + unsigned int read_write:1; + unsigned int reg:31; + void __user *data; +} drmI830MMIO; + #endif /* _I830_DRM_H_ */ diff --git a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c new file mode 100644 index 0000000000..eb5216cb37 --- /dev/null +++ b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c @@ -0,0 +1,945 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + * Keith Whitwell + */ + +#include +#include +#include "glthread.h" +#include "errno.h" +#include "ws_dri_bufmgr.h" +#include "string.h" +#include "imports.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" + +/* + * This lock is here to protect drmBO structs changing underneath us during a + * validate list call, since validatelist cannot take individiual locks for + * each drmBO. Validatelist takes this lock in write mode. Any access to an + * individual drmBO should take this lock in read mode, since in that case, the + * driBufferObject mutex will protect the access. Locking order is + * driBufferObject mutex - > this rw lock. + */ + +_glthread_DECLARE_STATIC_MUTEX(bmMutex); +_glthread_DECLARE_STATIC_COND(bmCond); + +static int kernelReaders = 0; + +static drmBO *drmBOListBuf(void *iterator) +{ + drmBONode *node; + drmMMListHead *l = (drmMMListHead *) iterator; + node = DRMLISTENTRY(drmBONode, l, head); + return node->buf; +} + +static void *drmBOListIterator(drmBOList *list) +{ + void *ret = list->list.next; + + if (ret == &list->list) + return NULL; + return ret; +} + +static void *drmBOListNext(drmBOList *list, void *iterator) +{ + void *ret; + + drmMMListHead *l = (drmMMListHead *) iterator; + ret = l->next; + if (ret == &list->list) + return NULL; + return ret; +} + +static drmBONode *drmAddListItem(drmBOList *list, drmBO *item, + uint64_t arg0, + uint64_t arg1) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->free.next; + if (l == &list->free) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + return NULL; + } + list->numCurrent++; + } + else { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + } + node->buf = item; + node->arg0 = arg0; + node->arg1 = arg1; + DRMLISTADD(&node->head, &list->list); + list->numOnList++; + return node; +} + +static int drmAddValidateItem(drmBOList *list, drmBO *buf, uint64_t flags, + uint64_t mask, int *newItem) +{ + drmBONode *node, *cur; + drmMMListHead *l; + + *newItem = 0; + cur = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + if (node->buf == buf) { + cur = node; + break; + } + } + if (!cur) { + cur = drmAddListItem(list, buf, flags, mask); + if (!cur) { + return -ENOMEM; + } + *newItem = 1; + cur->arg0 = flags; + cur->arg1 = mask; + } + else { + uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; + uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; + + if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { + return -EINVAL; + } + + cur->arg1 |= mask; + cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); + + if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && + (cur->arg0 & DRM_BO_MASK_MEM) == 0) { + return -EINVAL; + } + } + return 0; +} + +static void drmBOFreeList(drmBOList *list) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->list.next; + while(l != &list->list) { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + l = list->list.next; + list->numCurrent--; + list->numOnList--; + } + + l = list->free.next; + while(l != &list->free) { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + l = list->free.next; + list->numCurrent--; + } +} + +static int drmAdjustListNodes(drmBOList *list) +{ + drmBONode *node; + drmMMListHead *l; + int ret = 0; + + while(list->numCurrent < list->numTarget) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + ret = -ENOMEM; + break; + } + list->numCurrent++; + DRMLISTADD(&node->head, &list->free); + } + + while(list->numCurrent > list->numTarget) { + l = list->free.next; + if (l == &list->free) + break; + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + list->numCurrent--; + } + return ret; +} + +static int drmBOCreateList(int numTarget, drmBOList *list) +{ + DRMINITLISTHEAD(&list->list); + DRMINITLISTHEAD(&list->free); + list->numTarget = numTarget; + list->numCurrent = 0; + list->numOnList = 0; + return drmAdjustListNodes(list); +} + +static int drmBOResetList(drmBOList *list) +{ + drmMMListHead *l; + int ret; + + ret = drmAdjustListNodes(list); + if (ret) + return ret; + + l = list->list.next; + while (l != &list->list) { + DRMLISTDEL(l); + DRMLISTADD(l, &list->free); + list->numOnList--; + l = list->list.next; + } + return drmAdjustListNodes(list); +} + +void driWriteLockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + while(kernelReaders != 0) + _glthread_COND_WAIT(bmCond, bmMutex); +} + +void driWriteUnlockKernelBO(void) +{ + _glthread_UNLOCK_MUTEX(bmMutex); +} + +void driReadLockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + kernelReaders++; + _glthread_UNLOCK_MUTEX(bmMutex); +} + +void driReadUnlockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + if (--kernelReaders == 0) + _glthread_COND_BROADCAST(bmCond); + _glthread_UNLOCK_MUTEX(bmMutex); +} + + + + +/* + * TODO: Introduce fence pools in the same way as + * buffer object pools. + */ + +typedef struct _DriBufferObject +{ + DriBufferPool *pool; + _glthread_Mutex mutex; + int refCount; + const char *name; + uint64_t flags; + unsigned hint; + unsigned alignment; + unsigned createdByReference; + void *private; + /* user-space buffer: */ + unsigned userBuffer; + void *userData; + unsigned userSize; +} DriBufferObject; + +typedef struct _DriBufferList { + drmBOList drmBuffers; /* List of kernel buffers needing validation */ + drmBOList driBuffers; /* List of user-space buffers needing validation */ +} DriBufferList; + + +void +bmError(int val, const char *file, const char *function, int line) +{ + _mesa_printf("Fatal video memory manager error \"%s\".\n" + "Check kernel logs or set the LIBGL_DEBUG\n" + "environment variable to \"verbose\" for more info.\n" + "Detected in file %s, line %d, function %s.\n", + strerror(-val), file, line, function); +#ifndef NDEBUG + abort(); +#else + abort(); +#endif +} + +extern drmBO * +driBOKernel(struct _DriBufferObject *buf) +{ + drmBO *ret; + + driReadLockKernelBO(); + _glthread_LOCK_MUTEX(buf->mutex); + assert(buf->private != NULL); + ret = buf->pool->kernel(buf->pool, buf->private); + if (!ret) + BM_CKFATAL(-EINVAL); + _glthread_UNLOCK_MUTEX(buf->mutex); + driReadUnlockKernelBO(); + + return ret; +} + +void +driBOWaitIdle(struct _DriBufferObject *buf, int lazy) +{ + + /* + * This function may block. Is it sane to keep the mutex held during + * that time?? + */ + + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy)); + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void * +driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint) +{ + void *virtual; + int retval; + + if (buf->userBuffer) { + return buf->userData; + } + + _glthread_LOCK_MUTEX(buf->mutex); + assert(buf->private != NULL); + retval = buf->pool->map(buf->pool, buf->private, flags, hint, + &buf->mutex, &virtual); + _glthread_UNLOCK_MUTEX(buf->mutex); + + return retval == 0 ? virtual : NULL; +} + +void +driBOUnmap(struct _DriBufferObject *buf) +{ + if (buf->userBuffer) + return; + + assert(buf->private != NULL); + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +unsigned long +driBOOffset(struct _DriBufferObject *buf) +{ + unsigned long ret; + + assert(buf->private != NULL); + + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->offset(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + return ret; +} + +unsigned long +driBOPoolOffset(struct _DriBufferObject *buf) +{ + unsigned long ret; + + assert(buf->private != NULL); + + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->poolOffset(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + return ret; +} + +uint64_t +driBOFlags(struct _DriBufferObject *buf) +{ + uint64_t ret; + + assert(buf->private != NULL); + + driReadLockKernelBO(); + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->flags(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + driReadUnlockKernelBO(); + return ret; +} + +struct _DriBufferObject * +driBOReference(struct _DriBufferObject *buf) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (++buf->refCount == 1) { + _glthread_UNLOCK_MUTEX(buf->mutex); + BM_CKFATAL(-EINVAL); + } + _glthread_UNLOCK_MUTEX(buf->mutex); + return buf; +} + +void +driBOUnReference(struct _DriBufferObject *buf) +{ + int tmp; + + if (!buf) + return; + + + _glthread_LOCK_MUTEX(buf->mutex); + tmp = --buf->refCount; + if (!tmp) { + _glthread_UNLOCK_MUTEX(buf->mutex); + if (buf->private) { + if (buf->createdByReference) + buf->pool->unreference(buf->pool, buf->private); + else + buf->pool->destroy(buf->pool, buf->private); + } + free(buf); + } else + _glthread_UNLOCK_MUTEX(buf->mutex); + +} + + +int +driBOData(struct _DriBufferObject *buf, + unsigned size, const void *data, + DriBufferPool *newPool, + uint64_t flags) +{ + void *virtual = NULL; + int newBuffer; + int retval = 0; + struct _DriBufferPool *pool; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + pool = buf->pool; + + if (pool == NULL && newPool != NULL) { + buf->pool = newPool; + pool = newPool; + } + if (newPool == NULL) + newPool = pool; + + if (!pool->create) { + _mesa_error(NULL, GL_INVALID_OPERATION, + "driBOData called on invalid buffer\n"); + BM_CKFATAL(-EINVAL); + } + + newBuffer = (!buf->private || pool != newPool || + pool->size(pool, buf->private) < size); + + if (!flags) + flags = buf->flags; + + if (newBuffer) { + + if (buf->createdByReference) { + _mesa_error(NULL, GL_INVALID_OPERATION, + "driBOData requiring resizing called on " + "shared buffer.\n"); + BM_CKFATAL(-EINVAL); + } + + if (buf->private) + buf->pool->destroy(buf->pool, buf->private); + + pool = newPool; + buf->pool = newPool; + buf->private = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, + buf->alignment); + if (!buf->private) + retval = -ENOMEM; + + if (retval == 0) + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, + DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual); + } else if (pool->map(pool, buf->private, DRM_BO_FLAG_WRITE, + DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual)) { + /* + * Buffer is busy. need to create a new one. + */ + + void *newBuf; + + newBuf = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, + buf->alignment); + if (newBuf) { + buf->pool->destroy(buf->pool, buf->private); + buf->private = newBuf; + } + + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); + } else { + uint64_t flag_diff = flags ^ buf->flags; + + /* + * We might need to change buffer flags. + */ + + if (flag_diff){ + assert(pool->setStatus != NULL); + BM_CKFATAL(pool->unmap(pool, buf->private)); + BM_CKFATAL(pool->setStatus(pool, buf->private, flag_diff, + buf->flags)); + if (!data) + goto out; + + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); + } + } + + if (retval == 0) { + if (data) + memcpy(virtual, data, size); + + BM_CKFATAL(pool->unmap(pool, buf->private)); + } + + out: + _glthread_UNLOCK_MUTEX(buf->mutex); + + return retval; +} + +void +driBOSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, const void *data) +{ + void *virtual; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + if (size && data) { + BM_CKFATAL(buf->pool->map(buf->pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, + &virtual)); + memcpy((unsigned char *) virtual + offset, data, size); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + } + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void +driBOGetSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, void *data) +{ + void *virtual; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + if (size && data) { + BM_CKFATAL(buf->pool->map(buf->pool, buf->private, + DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual)); + memcpy(data, (unsigned char *) virtual + offset, size); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + } + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void +driBOSetReferenced(struct _DriBufferObject *buf, + unsigned long handle) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->private != NULL) { + _mesa_error(NULL, GL_INVALID_OPERATION, + "Invalid buffer for setReferenced\n"); + BM_CKFATAL(-EINVAL); + + } + if (buf->pool->reference == NULL) { + _mesa_error(NULL, GL_INVALID_OPERATION, + "Invalid buffer pool for setReferenced\n"); + BM_CKFATAL(-EINVAL); + } + buf->private = buf->pool->reference(buf->pool, handle); + if (!buf->private) { + _mesa_error(NULL, GL_OUT_OF_MEMORY, + "Invalid buffer pool for setStatic\n"); + BM_CKFATAL(-ENOMEM); + } + buf->createdByReference = GL_TRUE; + buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags; + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +int +driGenBuffers(struct _DriBufferPool *pool, + const char *name, + unsigned n, + struct _DriBufferObject *buffers[], + unsigned alignment, uint64_t flags, unsigned hint) +{ + struct _DriBufferObject *buf; + int i; + + flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM | + DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + + + assert(pool); + + for (i = 0; i < n; ++i) { + buf = (struct _DriBufferObject *) calloc(1, sizeof(*buf)); + if (!buf) + return -ENOMEM; + + _glthread_INIT_MUTEX(buf->mutex); + _glthread_LOCK_MUTEX(buf->mutex); + buf->refCount = 1; + buf->flags = flags; + buf->hint = hint; + buf->name = name; + buf->alignment = alignment; + buf->pool = pool; + buf->createdByReference = 0; + _glthread_UNLOCK_MUTEX(buf->mutex); + buffers[i] = buf; + } + return 0; +} + +void +driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject **buffers, + void *ptr, unsigned bytes) +{ + const unsigned alignment = 1, flags = 0, hint = 0; + + driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); + + (*buffers)->userBuffer = 1; + (*buffers)->userData = ptr; + (*buffers)->userSize = bytes; +} + +void +driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]) +{ + int i; + + for (i = 0; i < n; ++i) { + driBOUnReference(buffers[i]); + } +} + + +void +driInitBufMgr(int fd) +{ + ; +} + +/* + * Note that lists are per-context and don't need mutex protection. + */ + +struct _DriBufferList * +driBOCreateList(int target) +{ + struct _DriBufferList *list = calloc(sizeof(*list), 1); + + BM_CKFATAL(drmBOCreateList(target, &list->drmBuffers)); + BM_CKFATAL(drmBOCreateList(target, &list->driBuffers)); + return list; +} + +int +driBOResetList(struct _DriBufferList * list) +{ + int ret; + ret = drmBOResetList(&list->drmBuffers); + if (ret) + return ret; + ret = drmBOResetList(&list->driBuffers); + return ret; +} + +void +driBOFreeList(struct _DriBufferList * list) +{ + drmBOFreeList(&list->drmBuffers); + drmBOFreeList(&list->driBuffers); + free(list); +} + + +/* + * Copied from libdrm, because it is needed by driAddValidateItem. + */ + +static drmBONode * +driAddListItem(drmBOList * list, drmBO * item, + uint64_t arg0, uint64_t arg1) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->free.next; + if (l == &list->free) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + return NULL; + } + list->numCurrent++; + } else { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + } + memset(&node->bo_arg, 0, sizeof(node->bo_arg)); + node->buf = item; + node->arg0 = arg0; + node->arg1 = arg1; + DRMLISTADDTAIL(&node->head, &list->list); + list->numOnList++; + return node; +} + +/* + * Slightly modified version compared to the libdrm version. + * This one returns the list index of the buffer put on the list. + */ + +static int +driAddValidateItem(drmBOList * list, drmBO * buf, uint64_t flags, + uint64_t mask, int *itemLoc, + struct _drmBONode **pnode) +{ + drmBONode *node, *cur; + drmMMListHead *l; + int count = 0; + + cur = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + if (node->buf == buf) { + cur = node; + break; + } + count++; + } + if (!cur) { + cur = driAddListItem(list, buf, flags, mask); + if (!cur) + return -ENOMEM; + + cur->arg0 = flags; + cur->arg1 = mask; + } else { + uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; + uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; + + if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { + return -EINVAL; + } + + cur->arg1 |= mask; + cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); + + if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && + (cur->arg0 & DRM_BO_MASK_MEM) == 0) { + return -EINVAL; + } + } + *itemLoc = count; + *pnode = cur; + return 0; +} + + +void +driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf, + uint64_t flags, uint64_t mask, int *itemLoc, + struct _drmBONode **node) +{ + int newItem; + + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(driAddValidateItem(&list->drmBuffers, + buf->pool->kernel(buf->pool, buf->private), + flags, mask, itemLoc, node)); + BM_CKFATAL(drmAddValidateItem(&list->driBuffers, (drmBO *) buf, + flags, mask, &newItem)); + if (newItem) + buf->refCount++; + + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +drmBOList *driGetdrmBOList(struct _DriBufferList *list) +{ + driWriteLockKernelBO(); + return &list->drmBuffers; +} + +void driPutdrmBOList(struct _DriBufferList *list) +{ + driWriteUnlockKernelBO(); +} + + +void +driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->pool->fence) + BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence)); + _glthread_UNLOCK_MUTEX(buf->mutex); + +} + +void +driBOUnrefUserList(struct _DriBufferList *list) +{ + struct _DriBufferObject *buf; + void *curBuf; + + curBuf = drmBOListIterator(&list->driBuffers); + while (curBuf) { + buf = (struct _DriBufferObject *)drmBOListBuf(curBuf); + driBOUnReference(buf); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } +} + +struct _DriFenceObject * +driBOFenceUserList(struct _DriFenceMgr *mgr, + struct _DriBufferList *list, const char *name, + drmFence *kFence) +{ + struct _DriFenceObject *fence; + struct _DriBufferObject *buf; + void *curBuf; + + fence = driFenceCreate(mgr, kFence->fence_class, kFence->type, + kFence, sizeof(*kFence)); + curBuf = drmBOListIterator(&list->driBuffers); + + /* + * User-space fencing callbacks. + */ + + while (curBuf) { + buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); + driBOFence(buf, fence); + driBOUnReference(buf); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } + + driBOResetList(list); + return fence; +} + +void +driBOValidateUserList(struct _DriBufferList * list) +{ + void *curBuf; + struct _DriBufferObject *buf; + + curBuf = drmBOListIterator(&list->driBuffers); + + /* + * User-space validation callbacks. + */ + + while (curBuf) { + buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->pool->validate) + BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex)); + _glthread_UNLOCK_MUTEX(buf->mutex); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } +} + + +void +driPoolTakeDown(struct _DriBufferPool *pool) +{ + pool->takeDown(pool); + +} + +unsigned long +driBOSize(struct _DriBufferObject *buf) +{ + unsigned long size; + + _glthread_LOCK_MUTEX(buf->mutex); + size = buf->pool->size(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + + return size; + +} + +drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list) +{ + return &list->drmBuffers; +} + +drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list) +{ + return &list->driBuffers; +} + diff --git a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.h b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.h new file mode 100644 index 0000000000..fdaf5ee93a --- /dev/null +++ b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.h @@ -0,0 +1,138 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + * Keith Whitwell + */ + +#ifndef _PSB_BUFMGR_H_ +#define _PSB_BUFMGR_H_ +#include +#include "i915_drm.h" +#include "ws_dri_fencemgr.h" + +typedef struct _drmBONode +{ + drmMMListHead head; + drmBO *buf; + struct drm_i915_op_arg bo_arg; + uint64_t arg0; + uint64_t arg1; +} drmBONode; + +typedef struct _drmBOList { + unsigned numTarget; + unsigned numCurrent; + unsigned numOnList; + drmMMListHead list; + drmMMListHead free; +} drmBOList; + + +struct _DriFenceObject; +struct _DriBufferObject; +struct _DriBufferPool; +struct _DriBufferList; + +/* + * Return a pointer to the libdrm buffer object this DriBufferObject + * uses. + */ + +extern drmBO *driBOKernel(struct _DriBufferObject *buf); +extern void *driBOMap(struct _DriBufferObject *buf, unsigned flags, + unsigned hint); +extern void driBOUnmap(struct _DriBufferObject *buf); +extern unsigned long driBOOffset(struct _DriBufferObject *buf); +extern unsigned long driBOPoolOffset(struct _DriBufferObject *buf); + +extern uint64_t driBOFlags(struct _DriBufferObject *buf); +extern struct _DriBufferObject *driBOReference(struct _DriBufferObject *buf); +extern void driBOUnReference(struct _DriBufferObject *buf); + +extern int driBOData(struct _DriBufferObject *r_buf, + unsigned size, const void *data, + struct _DriBufferPool *pool, uint64_t flags); + +extern void driBOSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, + const void *data); +extern void driBOGetSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, + void *data); +extern int driGenBuffers(struct _DriBufferPool *pool, + const char *name, + unsigned n, + struct _DriBufferObject *buffers[], + unsigned alignment, uint64_t flags, unsigned hint); +extern void driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject *buffers[], + void *ptr, unsigned bytes); +extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]); +extern void driInitBufMgr(int fd); +extern struct _DriBufferList *driBOCreateList(int target); +extern int driBOResetList(struct _DriBufferList * list); +extern void driBOAddListItem(struct _DriBufferList * list, + struct _DriBufferObject *buf, + uint64_t flags, uint64_t mask, int *itemLoc, + struct _drmBONode **node); + +extern void driBOValidateList(int fd, struct _DriBufferList * list); +extern void driBOFreeList(struct _DriBufferList * list); +extern struct _DriFenceObject *driBOFenceUserList(struct _DriFenceMgr *mgr, + struct _DriBufferList *list, + const char *name, + drmFence *kFence); +extern void driBOUnrefUserList(struct _DriBufferList *list); +extern void driBOValidateUserList(struct _DriBufferList * list); +extern drmBOList *driGetdrmBOList(struct _DriBufferList *list); +extern void driPutdrmBOList(struct _DriBufferList *list); + +extern void driBOFence(struct _DriBufferObject *buf, + struct _DriFenceObject *fence); + +extern void driPoolTakeDown(struct _DriBufferPool *pool); +extern void driBOSetReferenced(struct _DriBufferObject *buf, + unsigned long handle); +unsigned long driBOSize(struct _DriBufferObject *buf); +extern void driBOWaitIdle(struct _DriBufferObject *buf, int lazy); +extern void driPoolTakeDown(struct _DriBufferPool *pool); + +extern void driReadLockKernelBO(void); +extern void driReadUnlockKernelBO(void); +extern void driWriteLockKernelBO(void); +extern void driWriteUnlockKernelBO(void); + +/* + * For debugging purposes. + */ + +extern drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list); +extern drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list); +#endif diff --git a/src/gallium/winsys/dri/intel/ws_dri_bufpool.h b/src/gallium/winsys/dri/intel/ws_dri_bufpool.h new file mode 100644 index 0000000000..3a302e13d3 --- /dev/null +++ b/src/gallium/winsys/dri/intel/ws_dri_bufpool.h @@ -0,0 +1,102 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#ifndef _PSB_BUFPOOL_H_ +#define _PSB_BUFPOOL_H_ + +#include +#include +struct _DriFenceObject; + +typedef struct _DriBufferPool +{ + int fd; + int (*map) (struct _DriBufferPool * pool, void *private, + unsigned flags, int hint, _glthread_Mutex *mutex, + void **virtual); + int (*unmap) (struct _DriBufferPool * pool, void *private); + int (*destroy) (struct _DriBufferPool * pool, void *private); + unsigned long (*offset) (struct _DriBufferPool * pool, void *private); + unsigned long (*poolOffset) (struct _DriBufferPool * pool, void *private); + uint64_t (*flags) (struct _DriBufferPool * pool, void *private); + unsigned long (*size) (struct _DriBufferPool * pool, void *private); + void *(*create) (struct _DriBufferPool * pool, unsigned long size, + uint64_t flags, unsigned hint, unsigned alignment); + void *(*reference) (struct _DriBufferPool * pool, unsigned handle); + int (*unreference) (struct _DriBufferPool * pool, void *private); + int (*fence) (struct _DriBufferPool * pool, void *private, + struct _DriFenceObject * fence); + drmBO *(*kernel) (struct _DriBufferPool * pool, void *private); + int (*validate) (struct _DriBufferPool * pool, void *private, _glthread_Mutex *mutex); + int (*waitIdle) (struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, + int lazy); + int (*setStatus) (struct _DriBufferPool *pool, void *private, + uint64_t flag_diff, uint64_t old_flags); + void (*takeDown) (struct _DriBufferPool * pool); + void *data; +} DriBufferPool; + +extern void bmError(int val, const char *file, const char *function, + int line); +#define BM_CKFATAL(val) \ + do{ \ + int tstVal = (val); \ + if (tstVal) \ + bmError(tstVal, __FILE__, __FUNCTION__, __LINE__); \ + } while(0); + + +/* + * Builtin pools. + */ + +/* + * Kernel buffer objects. Size in multiples of page size. Page size aligned. + */ + +extern struct _DriBufferPool *driDRMPoolInit(int fd); +extern struct _DriBufferPool *driMallocPoolInit(void); + +struct _DriFreeSlabManager; +extern struct _DriBufferPool * driSlabPoolInit(int fd, uint64_t flags, + uint64_t validMask, + uint32_t smallestSize, + uint32_t numSizes, + uint32_t desiredNumBuffers, + uint32_t maxSlabSize, + uint32_t pageAlignment, + struct _DriFreeSlabManager *fMan); +extern void driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan); +extern struct _DriFreeSlabManager * +driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec); + + +#endif diff --git a/src/gallium/winsys/dri/intel/ws_dri_drmpool.c b/src/gallium/winsys/dri/intel/ws_dri_drmpool.c new file mode 100644 index 0000000000..7c55dbc674 --- /dev/null +++ b/src/gallium/winsys/dri/intel/ws_dri_drmpool.c @@ -0,0 +1,268 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" +#include "assert.h" + +/* + * Buffer pool implementation using DRM buffer objects as DRI buffer objects. + */ + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, uint64_t flags, unsigned hint, + unsigned alignment) +{ + drmBO *buf = (drmBO *) malloc(sizeof(*buf)); + int ret; + unsigned pageSize = getpagesize(); + + if (!buf) + return NULL; + + if ((alignment > pageSize) && (alignment % pageSize)) { + free(buf); + return NULL; + } + + ret = drmBOCreate(pool->fd, size, alignment / pageSize, + NULL, + flags, hint, buf); + if (ret) { + free(buf); + return NULL; + } + + return (void *) buf; +} + +static void * +pool_reference(struct _DriBufferPool *pool, unsigned handle) +{ + drmBO *buf = (drmBO *) malloc(sizeof(*buf)); + int ret; + + if (!buf) + return NULL; + + ret = drmBOReference(pool->fd, handle, buf); + + if (ret) { + free(buf); + return NULL; + } + + return (void *) buf; +} + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + int ret; + drmBO *buf = (drmBO *) private; + driReadLockKernelBO(); + ret = drmBOUnreference(pool->fd, buf); + free(buf); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_unreference(struct _DriBufferPool *pool, void *private) +{ + int ret; + drmBO *buf = (drmBO *) private; + driReadLockKernelBO(); + ret = drmBOUnreference(pool->fd, buf); + free(buf); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOMap(pool->fd, buf, flags, hint, virtual); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOUnmap(pool->fd, buf); + driReadUnlockKernelBO(); + + return ret; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + unsigned long offset; + + driReadLockKernelBO(); + assert(buf->flags & DRM_BO_FLAG_NO_MOVE); + offset = buf->offset; + driReadUnlockKernelBO(); + + return buf->offset; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + return 0; +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + uint64_t flags; + + driReadLockKernelBO(); + flags = buf->flags; + driReadUnlockKernelBO(); + + return flags; +} + + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + unsigned long size; + + driReadLockKernelBO(); + size = buf->size; + driReadUnlockKernelBO(); + + return buf->size; +} + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + /* + * Noop. The kernel handles all fencing. + */ + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + return (drmBO *) private; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, + int lazy) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOWaitIdle(pool->fd, buf, (lazy) ? DRM_BO_HINT_WAIT_LAZY:0); + driReadUnlockKernelBO(); + + return ret; +} + + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + free(pool); +} + +/*static int +pool_setStatus(struct _DriBufferPool *pool, void *private, + uint64_t flag_diff, uint64_t old_flags) +{ + drmBO *buf = (drmBO *) private; + uint64_t new_flags = old_flags ^ flag_diff; + int ret; + + driReadLockKernelBO(); + ret = drmBOSetStatus(pool->fd, buf, new_flags, flag_diff, + 0, 0, 0); + driReadUnlockKernelBO(); + return ret; +}*/ + +struct _DriBufferPool * +driDRMPoolInit(int fd) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + + if (!pool) + return NULL; + + pool->fd = fd; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->poolOffset = &pool_poolOffset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = NULL; + pool->waitIdle = &pool_waitIdle; + pool->takeDown = &pool_takedown; + pool->reference = &pool_reference; + pool->unreference = &pool_unreference; + pool->data = NULL; + return pool; +} diff --git a/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c b/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c new file mode 100644 index 0000000000..8aaef1c620 --- /dev/null +++ b/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c @@ -0,0 +1,372 @@ +#include "ws_dri_fencemgr.h" +#include "glthread.h" +#include +#include +#include + +/* + * Note: Locking order is + * _DriFenceObject::mutex + * _DriFenceMgr::mutex + */ + +struct _DriFenceMgr { + /* + * Constant members. Need no mutex protection. + */ + struct _DriFenceMgrCreateInfo info; + void *private; + + /* + * These members are protected by this->mutex + */ + _glthread_Mutex mutex; + int refCount; + drmMMListHead *heads; +}; + +struct _DriFenceObject { + + /* + * These members are constant and need no mutex protection. + */ + struct _DriFenceMgr *mgr; + uint32_t fence_class; + uint32_t fence_type; + + /* + * These members are protected by mgr->mutex. + */ + drmMMListHead head; + int refCount; + + /* + * These members are protected by this->mutex. + */ + _glthread_Mutex mutex; + uint32_t signaled_type; + void *private; +}; + +uint32_t +driFenceType(struct _DriFenceObject *fence) +{ + return fence->fence_type; +} + +struct _DriFenceMgr * +driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info) +{ + struct _DriFenceMgr *tmp; + uint32_t i; + + tmp = calloc(1, sizeof(*tmp)); + if (!tmp) + return NULL; + + _glthread_INIT_MUTEX(tmp->mutex); + _glthread_LOCK_MUTEX(tmp->mutex); + tmp->refCount = 1; + tmp->info = *info; + tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads)); + if (!tmp->heads) + goto out_err; + + for (i=0; iinfo.num_classes; ++i) { + DRMINITLISTHEAD(&tmp->heads[i]); + } + _glthread_UNLOCK_MUTEX(tmp->mutex); + return tmp; + + out_err: + if (tmp) + free(tmp); + return NULL; +} + +static void +driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr) +{ + struct _DriFenceMgr *mgr = *pMgr; + + *pMgr = NULL; + if (--mgr->refCount == 0) + free(mgr); + else + _glthread_UNLOCK_MUTEX(mgr->mutex); +} + +void +driFenceMgrUnReference(struct _DriFenceMgr **pMgr) +{ + _glthread_LOCK_MUTEX((*pMgr)->mutex); + driFenceMgrUnrefUnlock(pMgr); +} + +static void +driFenceUnReferenceLocked(struct _DriFenceObject **pFence) +{ + struct _DriFenceObject *fence = *pFence; + struct _DriFenceMgr *mgr = fence->mgr; + + *pFence = NULL; + if (--fence->refCount == 0) { + DRMLISTDELINIT(&fence->head); + if (fence->private) + mgr->info.unreference(mgr, &fence->private); + fence->mgr = NULL; + --mgr->refCount; + free(fence); + } +} + + +static void +driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr, + drmMMListHead *list, + uint32_t fence_class, + uint32_t fence_type) +{ + struct _DriFenceObject *entry; + drmMMListHead *prev; + + while(list != &mgr->heads[fence_class]) { + entry = DRMLISTENTRY(struct _DriFenceObject, list, head); + + /* + * Up refcount so that entry doesn't disappear from under us + * when we unlock-relock mgr to get the correct locking order. + */ + + ++entry->refCount; + _glthread_UNLOCK_MUTEX(mgr->mutex); + _glthread_LOCK_MUTEX(entry->mutex); + _glthread_LOCK_MUTEX(mgr->mutex); + + prev = list->prev; + + + + if (list->prev == list) { + + /* + * Somebody else removed the entry from the list. + */ + + _glthread_UNLOCK_MUTEX(entry->mutex); + driFenceUnReferenceLocked(&entry); + return; + } + + entry->signaled_type |= (fence_type & entry->fence_type); + if (entry->signaled_type == entry->fence_type) { + DRMLISTDELINIT(list); + mgr->info.unreference(mgr, &entry->private); + } + _glthread_UNLOCK_MUTEX(entry->mutex); + driFenceUnReferenceLocked(&entry); + list = prev; + } +} + + +int +driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, + int lazy_hint) +{ + struct _DriFenceMgr *mgr = fence->mgr; + int ret = 0; + + _glthread_LOCK_MUTEX(fence->mutex); + + if ((fence->signaled_type & fence_type) == fence_type) + goto out0; + + ret = mgr->info.finish(mgr, fence->private, fence_type, lazy_hint); + if (ret) + goto out0; + + _glthread_LOCK_MUTEX(mgr->mutex); + _glthread_UNLOCK_MUTEX(fence->mutex); + + driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, + fence_type); + _glthread_UNLOCK_MUTEX(mgr->mutex); + return 0; + + out0: + _glthread_UNLOCK_MUTEX(fence->mutex); + return ret; +} + +uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence) +{ + uint32_t ret; + + _glthread_LOCK_MUTEX(fence->mutex); + ret = fence->signaled_type; + _glthread_UNLOCK_MUTEX(fence->mutex); + + return ret; +} + +int +driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type, + uint32_t *signaled) +{ + int ret = 0; + struct _DriFenceMgr *mgr; + + _glthread_LOCK_MUTEX(fence->mutex); + mgr = fence->mgr; + *signaled = fence->signaled_type; + if ((fence->signaled_type & flush_type) == flush_type) + goto out0; + + ret = mgr->info.signaled(mgr, fence->private, flush_type, signaled); + if (ret) { + *signaled = fence->signaled_type; + goto out0; + } + + if ((fence->signaled_type | *signaled) == fence->signaled_type) + goto out0; + + _glthread_LOCK_MUTEX(mgr->mutex); + _glthread_UNLOCK_MUTEX(fence->mutex); + + driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, + *signaled); + + _glthread_UNLOCK_MUTEX(mgr->mutex); + return 0; + out0: + _glthread_UNLOCK_MUTEX(fence->mutex); + return ret; +} + +struct _DriFenceObject * +driFenceReference(struct _DriFenceObject *fence) +{ + _glthread_LOCK_MUTEX(fence->mgr->mutex); + ++fence->refCount; + _glthread_UNLOCK_MUTEX(fence->mgr->mutex); + return fence; +} + +void +driFenceUnReference(struct _DriFenceObject **pFence) +{ + struct _DriFenceMgr *mgr; + + if (*pFence == NULL) + return; + + mgr = (*pFence)->mgr; + _glthread_LOCK_MUTEX(mgr->mutex); + ++mgr->refCount; + driFenceUnReferenceLocked(pFence); + driFenceMgrUnrefUnlock(&mgr); +} + +struct _DriFenceObject +*driFenceCreate(struct _DriFenceMgr *mgr, uint32_t fence_class, + uint32_t fence_type, void *private, size_t private_size) +{ + struct _DriFenceObject *fence; + size_t fence_size = sizeof(*fence); + + if (private_size) + fence_size = ((fence_size + 15) & ~15); + + fence = calloc(1, fence_size + private_size); + + if (!fence) { + int ret = mgr->info.finish(mgr, private, fence_type, 0); + + if (ret) + usleep(10000000); + + return NULL; + } + + _glthread_INIT_MUTEX(fence->mutex); + _glthread_LOCK_MUTEX(fence->mutex); + _glthread_LOCK_MUTEX(mgr->mutex); + fence->refCount = 1; + DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]); + fence->mgr = mgr; + ++mgr->refCount; + _glthread_UNLOCK_MUTEX(mgr->mutex); + fence->fence_class = fence_class; + fence->fence_type = fence_type; + fence->signaled_type = 0; + fence->private = private; + if (private_size) { + fence->private = (void *)(((uint8_t *) fence) + fence_size); + memcpy(fence->private, private, private_size); + } + + _glthread_UNLOCK_MUTEX(fence->mutex); + return fence; +} + + +static int +tSignaled(struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, + uint32_t *signaled_type) +{ + long fd = (long) mgr->private; + int dummy; + drmFence *fence = (drmFence *) private; + int ret; + + *signaled_type = 0; + ret = drmFenceSignaled((int) fd, fence, flush_type, &dummy); + if (ret) + return ret; + + *signaled_type = fence->signaled; + + return 0; +} + +static int +tFinish(struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, + int lazy_hint) +{ + long fd = (long) mgr->private; + unsigned flags = lazy_hint ? DRM_FENCE_FLAG_WAIT_LAZY : 0; + + return drmFenceWait((int)fd, flags, (drmFence *) private, fence_type); +} + +static int +tUnref(struct _DriFenceMgr *mgr, void **private) +{ + long fd = (long) mgr->private; + drmFence *fence = (drmFence *) *private; + *private = NULL; + + return drmFenceUnreference(fd, fence); +} + +struct _DriFenceMgr *driFenceMgrTTMInit(int fd) +{ + struct _DriFenceMgrCreateInfo info; + struct _DriFenceMgr *mgr; + + info.flags = DRI_FENCE_CLASS_ORDERED; + info.num_classes = 4; + info.signaled = tSignaled; + info.finish = tFinish; + info.unreference = tUnref; + + mgr = driFenceMgrCreate(&info); + if (mgr == NULL) + return NULL; + + mgr->private = (void *) (long) fd; + return mgr; +} + diff --git a/src/gallium/winsys/dri/intel/ws_dri_fencemgr.h b/src/gallium/winsys/dri/intel/ws_dri_fencemgr.h new file mode 100644 index 0000000000..4ea58dfe18 --- /dev/null +++ b/src/gallium/winsys/dri/intel/ws_dri_fencemgr.h @@ -0,0 +1,115 @@ +#ifndef DRI_FENCEMGR_H +#define DRI_FENCEMGR_H + +#include +#include + +struct _DriFenceObject; +struct _DriFenceMgr; + +/* + * Do a quick check to see if the fence manager has registered the fence + * object as signaled. Note that this function may return a false negative + * answer. + */ +extern uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence); + +/* + * Check if the fence object is signaled. This function can be substantially + * more expensive to call than the above function, but will not return a false + * negative answer. The argument "flush_type" sets the types that the + * underlying mechanism must make sure will eventually signal. + */ +extern int driFenceSignaledType(struct _DriFenceObject *fence, + uint32_t flush_type, uint32_t *signaled); + +/* + * Convenience functions. + */ + +static inline int driFenceSignaled(struct _DriFenceObject *fence, + uint32_t flush_type) +{ + uint32_t signaled_types; + int ret = driFenceSignaledType(fence, flush_type, &signaled_types); + if (ret) + return 0; + return ((signaled_types & flush_type) == flush_type); +} + +static inline int driFenceSignaledCached(struct _DriFenceObject *fence, + uint32_t flush_type) +{ + uint32_t signaled_types = + driFenceSignaledTypeCached(fence); + + return ((signaled_types & flush_type) == flush_type); +} + +/* + * Reference a fence object. + */ +extern struct _DriFenceObject *driFenceReference(struct _DriFenceObject *fence); + +/* + * Unreference a fence object. The fence object pointer will be reset to NULL. + */ + +extern void driFenceUnReference(struct _DriFenceObject **pFence); + + +/* + * Wait for a fence to signal the indicated fence_type. + * If "lazy_hint" is true, it indicates that the wait may sleep to avoid + * busy-wait polling. + */ +extern int driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, + int lazy_hint); + +/* + * Create a DriFenceObject for manager "mgr". + * + * "private" is a pointer that should be used for the callbacks in + * struct _DriFenceMgrCreateInfo. + * + * if private_size is nonzero, then the info stored at *private, with size + * private size will be copied and the fence manager will instead use a + * pointer to the copied data for the callbacks in + * struct _DriFenceMgrCreateInfo. In that case, the object pointed to by + * "private" may be destroyed after the call to driFenceCreate. + */ +extern struct _DriFenceObject *driFenceCreate(struct _DriFenceMgr *mgr, + uint32_t fence_class, + uint32_t fence_type, + void *private, + size_t private_size); + +extern uint32_t driFenceType(struct _DriFenceObject *fence); + +/* + * Fence creations are ordered. If a fence signals a fence_type, + * it is safe to assume that all fences of the same class that was + * created before that fence has signaled the same type. + */ + +#define DRI_FENCE_CLASS_ORDERED (1 << 0) + +struct _DriFenceMgrCreateInfo { + uint32_t flags; + uint32_t num_classes; + int (*signaled) (struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, + uint32_t *signaled_type); + int (*finish) (struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, int lazy_hint); + int (*unreference) (struct _DriFenceMgr *mgr, void **private); +}; + +extern struct _DriFenceMgr * +driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info); + +void +driFenceMgrUnReference(struct _DriFenceMgr **pMgr); + +extern struct _DriFenceMgr * +driFenceMgrTTMInit(int fd); + +#endif diff --git a/src/gallium/winsys/dri/intel/ws_dri_mallocpool.c b/src/gallium/winsys/dri/intel/ws_dri_mallocpool.c new file mode 100644 index 0000000000..bf97d7e440 --- /dev/null +++ b/src/gallium/winsys/dri/intel/ws_dri_mallocpool.c @@ -0,0 +1,162 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "imports.h" +#include "glthread.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" +#include "intel_screen.h" + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, uint64_t flags, unsigned hint, + unsigned alignment) +{ + unsigned long *private = malloc(size + 2*sizeof(unsigned long)); + if ((flags & DRM_BO_MASK_MEM) != DRM_BO_FLAG_MEM_LOCAL) + abort(); + + *private = size; + return (void *)private; +} + + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + free(private); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, + _glthread_Mutex *mutex, int lazy) +{ + return 0; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + *virtual = (void *)((unsigned long *)private + 2); + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + /* + * BUG + */ + abort(); + return 0UL; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + /* + * BUG + */ + abort(); +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + return DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + return *(unsigned long *) private; +} + + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + abort(); + return 0UL; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + abort(); + return NULL; +} + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + free(pool); +} + + +struct _DriBufferPool * +driMallocPoolInit(void) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + if (!pool) + return NULL; + + pool->data = NULL; + pool->fd = -1; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->poolOffset = &pool_poolOffset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = NULL; + pool->waitIdle = &pool_waitIdle; + pool->takeDown = &pool_takedown; + return pool; +} diff --git a/src/gallium/winsys/dri/intel/ws_dri_slabpool.c b/src/gallium/winsys/dri/intel/ws_dri_slabpool.c new file mode 100644 index 0000000000..235f0ac972 --- /dev/null +++ b/src/gallium/winsys/dri/intel/ws_dri_slabpool.c @@ -0,0 +1,945 @@ +/************************************************************************** + * + * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellstrom + */ + +#include +#include +#include +#include +#include +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" +#include "ws_dri_bufmgr.h" +#include "glthread.h" + +#define DRI_SLABPOOL_ALLOC_RETRIES 100 + +struct _DriSlab; + +struct _DriSlabBuffer { + int isSlabBuffer; + drmBO *bo; + struct _DriFenceObject *fence; + struct _DriSlab *parent; + drmMMListHead head; + uint32_t mapCount; + uint32_t start; + uint32_t fenceType; + int unFenced; + _glthread_Cond event; +}; + +struct _DriKernelBO { + int fd; + drmBO bo; + drmMMListHead timeoutHead; + drmMMListHead head; + struct timeval timeFreed; + uint32_t pageAlignment; + void *virtual; +}; + +struct _DriSlab{ + drmMMListHead head; + drmMMListHead freeBuffers; + uint32_t numBuffers; + uint32_t numFree; + struct _DriSlabBuffer *buffers; + struct _DriSlabSizeHeader *header; + struct _DriKernelBO *kbo; +}; + + +struct _DriSlabSizeHeader { + drmMMListHead slabs; + drmMMListHead freeSlabs; + drmMMListHead delayedBuffers; + uint32_t numDelayed; + struct _DriSlabPool *slabPool; + uint32_t bufSize; + _glthread_Mutex mutex; +}; + +struct _DriFreeSlabManager { + struct timeval slabTimeout; + struct timeval checkInterval; + struct timeval nextCheck; + drmMMListHead timeoutList; + drmMMListHead unCached; + drmMMListHead cached; + _glthread_Mutex mutex; +}; + + +struct _DriSlabPool { + + /* + * The data of this structure remains constant after + * initialization and thus needs no mutex protection. + */ + + struct _DriFreeSlabManager *fMan; + uint64_t proposedFlags; + uint64_t validMask; + uint32_t *bucketSizes; + uint32_t numBuckets; + uint32_t pageSize; + int fd; + int pageAlignment; + int maxSlabSize; + int desiredNumBuffers; + struct _DriSlabSizeHeader *headers; +}; + +/* + * FIXME: Perhaps arrange timeout slabs in size buckets for fast + * retreival?? + */ + + +static inline int +driTimeAfterEq(struct timeval *arg1, struct timeval *arg2) +{ + return ((arg1->tv_sec > arg2->tv_sec) || + ((arg1->tv_sec == arg2->tv_sec) && + (arg1->tv_usec > arg2->tv_usec))); +} + +static inline void +driTimeAdd(struct timeval *arg, struct timeval *add) +{ + unsigned int sec; + + arg->tv_sec += add->tv_sec; + arg->tv_usec += add->tv_usec; + sec = arg->tv_usec / 1000000; + arg->tv_sec += sec; + arg->tv_usec -= sec*1000000; +} + +static void +driFreeKernelBO(struct _DriKernelBO *kbo) +{ + if (!kbo) + return; + + (void) drmBOUnreference(kbo->fd, &kbo->bo); + free(kbo); +} + + +static void +driFreeTimeoutKBOsLocked(struct _DriFreeSlabManager *fMan, + struct timeval *time) +{ + drmMMListHead *list, *next; + struct _DriKernelBO *kbo; + + if (!driTimeAfterEq(time, &fMan->nextCheck)) + return; + + for (list = fMan->timeoutList.next, next = list->next; + list != &fMan->timeoutList; + list = next, next = list->next) { + + kbo = DRMLISTENTRY(struct _DriKernelBO, list, timeoutHead); + + if (!driTimeAfterEq(time, &kbo->timeFreed)) + break; + + DRMLISTDELINIT(&kbo->timeoutHead); + DRMLISTDELINIT(&kbo->head); + driFreeKernelBO(kbo); + } + + fMan->nextCheck = *time; + driTimeAdd(&fMan->nextCheck, &fMan->checkInterval); +} + + +/* + * Add a _DriKernelBO to the free slab manager. + * This means that it is available for reuse, but if it's not + * reused in a while, it will be freed. + */ + +static void +driSetKernelBOFree(struct _DriFreeSlabManager *fMan, + struct _DriKernelBO *kbo) +{ + struct timeval time; + + _glthread_LOCK_MUTEX(fMan->mutex); + gettimeofday(&time, NULL); + driTimeAdd(&time, &fMan->slabTimeout); + + kbo->timeFreed = time; + + if (kbo->bo.flags & DRM_BO_FLAG_CACHED) + DRMLISTADD(&kbo->head, &fMan->cached); + else + DRMLISTADD(&kbo->head, &fMan->unCached); + + DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList); + driFreeTimeoutKBOsLocked(fMan, &time); + + _glthread_UNLOCK_MUTEX(fMan->mutex); +} + +/* + * Get a _DriKernelBO for us to use as storage for a slab. + * + */ + +static struct _DriKernelBO * +driAllocKernelBO(struct _DriSlabSizeHeader *header) + +{ + struct _DriSlabPool *slabPool = header->slabPool; + struct _DriFreeSlabManager *fMan = slabPool->fMan; + drmMMListHead *list, *next, *head; + uint32_t size = header->bufSize * slabPool->desiredNumBuffers; + struct _DriKernelBO *kbo; + int ret; + + /* + * FIXME: We should perhaps allow some variation in slabsize in order + * to efficiently reuse slabs. + */ + + size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize; + size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1); + _glthread_LOCK_MUTEX(fMan->mutex); + + kbo = NULL; + + retry: + head = (slabPool->proposedFlags & DRM_BO_FLAG_CACHED) ? + &fMan->cached : &fMan->unCached; + + for (list = head->next, next = list->next; + list != head; + list = next, next = list->next) { + + kbo = DRMLISTENTRY(struct _DriKernelBO, list, head); + + if ((kbo->bo.size == size) && + (slabPool->pageAlignment == 0 || + (kbo->pageAlignment % slabPool->pageAlignment) == 0)) { + + DRMLISTDELINIT(&kbo->head); + DRMLISTDELINIT(&kbo->timeoutHead); + break; + } + + kbo = NULL; + } + + _glthread_UNLOCK_MUTEX(fMan->mutex); + + if (kbo) { + ret = 0; + ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, + (slabPool->proposedFlags ^ kbo->bo.flags), + DRM_BO_HINT_DONT_FENCE, 0, 0); + if (ret == 0) + return kbo; + + driFreeKernelBO(kbo); + kbo = NULL; + goto retry; + } + + kbo = calloc(1, sizeof(struct _DriKernelBO)); + if (!kbo) + return NULL; + + kbo->fd = slabPool->fd; + DRMINITLISTHEAD(&kbo->head); + DRMINITLISTHEAD(&kbo->timeoutHead); + + ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL, + slabPool->proposedFlags, + DRM_BO_HINT_DONT_FENCE, &kbo->bo); + if (ret) + goto out_err0; + + ret = drmBOMap(kbo->fd, &kbo->bo, + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, &kbo->virtual); + + if (ret) + goto out_err1; + + ret = drmBOUnmap(kbo->fd, &kbo->bo); + if (ret) + goto out_err1; + + return kbo; + + out_err1: + drmBOUnreference(kbo->fd, &kbo->bo); + out_err0: + free(kbo); + return NULL; +} + + +static int +driAllocSlab(struct _DriSlabSizeHeader *header) +{ + struct _DriSlab *slab; + struct _DriSlabBuffer *buf; + uint32_t numBuffers; + int ret; + int i; + + slab = calloc(1, sizeof(*slab)); + if (!slab) + return -ENOMEM; + + slab->kbo = driAllocKernelBO(header); + if (!slab->kbo) { + ret = -ENOMEM; + goto out_err0; + } + + numBuffers = slab->kbo->bo.size / header->bufSize; + + slab->buffers = calloc(numBuffers, sizeof(*slab->buffers)); + if (!slab->buffers) { + ret = -ENOMEM; + goto out_err1; + } + + DRMINITLISTHEAD(&slab->head); + DRMINITLISTHEAD(&slab->freeBuffers); + slab->numBuffers = numBuffers; + slab->numFree = 0; + slab->header = header; + + buf = slab->buffers; + for (i=0; i < numBuffers; ++i) { + buf->parent = slab; + buf->start = i* header->bufSize; + buf->mapCount = 0; + buf->isSlabBuffer = 1; + _glthread_INIT_COND(buf->event); + DRMLISTADDTAIL(&buf->head, &slab->freeBuffers); + slab->numFree++; + buf++; + } + + DRMLISTADDTAIL(&slab->head, &header->slabs); + + return 0; + + out_err1: + driSetKernelBOFree(header->slabPool->fMan, slab->kbo); + free(slab->buffers); + out_err0: + free(slab); + return ret; +} + +/* + * Delete a buffer from the slab header delayed list and put + * it on the slab free list. + */ + +static void +driSlabFreeBufferLocked(struct _DriSlabBuffer *buf) +{ + struct _DriSlab *slab = buf->parent; + struct _DriSlabSizeHeader *header = slab->header; + drmMMListHead *list = &buf->head; + + DRMLISTDEL(list); + DRMLISTADDTAIL(list, &slab->freeBuffers); + slab->numFree++; + + if (slab->head.next == &slab->head) + DRMLISTADDTAIL(&slab->head, &header->slabs); + + if (slab->numFree == slab->numBuffers) { + list = &slab->head; + DRMLISTDEL(list); + DRMLISTADDTAIL(list, &header->freeSlabs); + } + + if (header->slabs.next == &header->slabs || + slab->numFree != slab->numBuffers) { + + drmMMListHead *next; + struct _DriFreeSlabManager *fMan = header->slabPool->fMan; + + for (list = header->freeSlabs.next, next = list->next; + list != &header->freeSlabs; + list = next, next = list->next) { + + slab = DRMLISTENTRY(struct _DriSlab, list, head); + + DRMLISTDELINIT(list); + driSetKernelBOFree(fMan, slab->kbo); + free(slab->buffers); + free(slab); + } + } +} + +static void +driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait) +{ + drmMMListHead *list, *prev; + struct _DriSlabBuffer *buf; + struct _DriSlab *slab; + + int signaled = 0; + int i; + int ret; + + list = header->delayedBuffers.next; + + /* Only examine the oldest 1/3 of delayed buffers: + */ + if (header->numDelayed > 3) { + for (i = 0; i < header->numDelayed; i += 3) { + list = list->next; + } + } + + prev = list->prev; + for (; list != &header->delayedBuffers; list = prev, prev = list->prev) { + buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); + slab = buf->parent; + + if (!signaled) { + if (wait) { + ret = driFenceFinish(buf->fence, buf->fenceType, 0); + if (ret) + break; + signaled = 1; + } else { + signaled = driFenceSignaled(buf->fence, buf->fenceType); + } + if (signaled) { + driFenceUnReference(&buf->fence); + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { + driFenceUnReference(&buf->fence); + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + } +} + + +static struct _DriSlabBuffer * +driSlabAllocBuffer(struct _DriSlabSizeHeader *header) +{ + static struct _DriSlabBuffer *buf; + struct _DriSlab *slab; + drmMMListHead *list; + int count = DRI_SLABPOOL_ALLOC_RETRIES; + + _glthread_LOCK_MUTEX(header->mutex); + while(header->slabs.next == &header->slabs && count > 0) { + driSlabCheckFreeLocked(header, 0); + if (header->slabs.next != &header->slabs) + break; + + _glthread_UNLOCK_MUTEX(header->mutex); + if (count != DRI_SLABPOOL_ALLOC_RETRIES) + usleep(1); + (void) driAllocSlab(header); + _glthread_LOCK_MUTEX(header->mutex); + count--; + } + + list = header->slabs.next; + if (list == &header->slabs) { + _glthread_UNLOCK_MUTEX(header->mutex); + //assert(0, "no buffers in slab"); + return NULL; + } + slab = DRMLISTENTRY(struct _DriSlab, list, head); + if (--slab->numFree == 0) + DRMLISTDELINIT(list); + + list = slab->freeBuffers.next; + DRMLISTDELINIT(list); + + _glthread_UNLOCK_MUTEX(header->mutex); + buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); + return buf; +} + +static void * +pool_create(struct _DriBufferPool *driPool, unsigned long size, + uint64_t flags, unsigned hint, unsigned alignment) +{ + struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; + struct _DriSlabSizeHeader *header; + struct _DriSlabBuffer *buf; + void *dummy; + int i; + int ret; + + + /* + * FIXME: Check for compatibility. + */ + + header = pool->headers; + for (i=0; inumBuckets; ++i) { + if (header->bufSize >= size) + break; + header++; + } + + if (i < pool->numBuckets) + return driSlabAllocBuffer(header); + + + /* + * Fall back to allocate a buffer object directly from DRM. + * and wrap it in a driBO structure. + */ + + + buf = calloc(1, sizeof(*buf)); + + if (!buf) + return NULL; + + buf->bo = calloc(1, sizeof(*buf->bo)); + if (!buf->bo) + goto out_err0; + + if (alignment) { + if ((alignment < pool->pageSize) && (pool->pageSize % alignment)) + goto out_err1; + if ((alignment > pool->pageSize) && (alignment % pool->pageSize)) + goto out_err1; + } + + ret = drmBOCreate(pool->fd, size, alignment / pool->pageSize, NULL, + flags, hint, buf->bo); + if (ret) + goto out_err1; + + ret = drmBOMap(pool->fd, buf->bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, &dummy); + if (ret) + goto out_err2; + + ret = drmBOUnmap(pool->fd, buf->bo); + if (ret) + goto out_err2; + + return buf; + out_err2: + drmBOUnreference(pool->fd, buf->bo); + out_err1: + free(buf->bo); + out_err0: + free(buf); + return NULL; +} + +static int +pool_destroy(struct _DriBufferPool *driPool, void *private) +{ + struct _DriSlabBuffer *buf = + (struct _DriSlabBuffer *) private; + struct _DriSlab *slab; + struct _DriSlabSizeHeader *header; + + if (!buf->isSlabBuffer) { + struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; + int ret; + + ret = drmBOUnreference(pool->fd, buf->bo); + free(buf->bo); + free(buf); + return ret; + } + + slab = buf->parent; + header = slab->header; + + _glthread_LOCK_MUTEX(header->mutex); + buf->unFenced = 0; + buf->mapCount = 0; + + if (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)) { + DRMLISTADDTAIL(&buf->head, &header->delayedBuffers); + header->numDelayed++; + } else { + driSlabFreeBufferLocked(buf); + } + + _glthread_UNLOCK_MUTEX(header->mutex); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *driPool, void *private, + _glthread_Mutex *mutex, int lazy) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + while(buf->unFenced) + _glthread_COND_WAIT(buf->event, *mutex); + + if (!buf->fence) + return 0; + + driFenceFinish(buf->fence, buf->fenceType, lazy); + driFenceUnReference(&buf->fence); + + return 0; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + int busy; + + if (buf->isSlabBuffer) + busy = buf->unFenced || (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)); + else + busy = buf->fence && !driFenceSignaled(buf->fence, buf->fenceType); + + + if (busy) { + if (hint & DRM_BO_HINT_DONT_BLOCK) + return -EBUSY; + else { + (void) pool_waitIdle(pool, private, mutex, 0); + } + } + + ++buf->mapCount; + *virtual = (buf->isSlabBuffer) ? + (void *) ((uint8_t *) buf->parent->kbo->virtual + buf->start) : + (void *) buf->bo->virtual; + + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + --buf->mapCount; + if (buf->mapCount == 0 && buf->isSlabBuffer) + _glthread_COND_BROADCAST(buf->event); + + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + struct _DriSlab *slab; + struct _DriSlabSizeHeader *header; + + if (!buf->isSlabBuffer) { + assert(buf->bo->proposedFlags & DRM_BO_FLAG_NO_MOVE); + return buf->bo->offset; + } + + slab = buf->parent; + header = slab->header; + + (void) header; + assert(header->slabPool->proposedFlags & DRM_BO_FLAG_NO_MOVE); + return slab->kbo->bo.offset + buf->start; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + return buf->start; +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + if (!buf->isSlabBuffer) + return buf->bo->flags; + + return buf->parent->kbo->bo.flags; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + if (!buf->isSlabBuffer) + return buf->bo->size; + + return buf->parent->header->bufSize; +} + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + drmBO *bo; + + if (buf->fence) + driFenceUnReference(&buf->fence); + + buf->fence = driFenceReference(fence); + bo = (buf->isSlabBuffer) ? + &buf->parent->kbo->bo: + buf->bo; + buf->fenceType = bo->fenceFlags; + + buf->unFenced = 0; + _glthread_COND_BROADCAST(buf->event); + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + return (buf->isSlabBuffer) ? &buf->parent->kbo->bo : buf->bo; +} + +static int +pool_validate(struct _DriBufferPool *pool, void *private, + _glthread_Mutex *mutex) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + if (!buf->isSlabBuffer) + return 0; + + while(buf->mapCount != 0) + _glthread_COND_WAIT(buf->event, *mutex); + + buf->unFenced = 1; + return 0; +} + + +struct _DriFreeSlabManager * +driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec) +{ + struct _DriFreeSlabManager *tmp; + + tmp = calloc(1, sizeof(*tmp)); + if (!tmp) + return NULL; + + _glthread_INIT_MUTEX(tmp->mutex); + _glthread_LOCK_MUTEX(tmp->mutex); + tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000; + tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000; + tmp->slabTimeout.tv_usec -= tmp->slabTimeout.tv_sec*1000000; + + tmp->checkInterval.tv_usec = checkIntervalMsec*1000; + tmp->checkInterval.tv_sec = tmp->checkInterval.tv_usec / 1000000; + tmp->checkInterval.tv_usec -= tmp->checkInterval.tv_sec*1000000; + + gettimeofday(&tmp->nextCheck, NULL); + driTimeAdd(&tmp->nextCheck, &tmp->checkInterval); + DRMINITLISTHEAD(&tmp->timeoutList); + DRMINITLISTHEAD(&tmp->unCached); + DRMINITLISTHEAD(&tmp->cached); + _glthread_UNLOCK_MUTEX(tmp->mutex); + + return tmp; +} + +void +driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan) +{ + struct timeval time; + + time = fMan->nextCheck; + driTimeAdd(&time, &fMan->checkInterval); + + _glthread_LOCK_MUTEX(fMan->mutex); + driFreeTimeoutKBOsLocked(fMan, &time); + _glthread_UNLOCK_MUTEX(fMan->mutex); + + assert(fMan->timeoutList.next == &fMan->timeoutList); + assert(fMan->unCached.next == &fMan->unCached); + assert(fMan->cached.next == &fMan->cached); + + free(fMan); +} + +static void +driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size, + struct _DriSlabSizeHeader *header) +{ + _glthread_INIT_MUTEX(header->mutex); + _glthread_LOCK_MUTEX(header->mutex); + + DRMINITLISTHEAD(&header->slabs); + DRMINITLISTHEAD(&header->freeSlabs); + DRMINITLISTHEAD(&header->delayedBuffers); + + header->numDelayed = 0; + header->slabPool = pool; + header->bufSize = size; + + _glthread_UNLOCK_MUTEX(header->mutex); +} + +static void +driFinishSizeHeader(struct _DriSlabSizeHeader *header) +{ + drmMMListHead *list, *next; + struct _DriSlabBuffer *buf; + + _glthread_LOCK_MUTEX(header->mutex); + for (list = header->delayedBuffers.next, next = list->next; + list != &header->delayedBuffers; + list = next, next = list->next) { + + buf = DRMLISTENTRY(struct _DriSlabBuffer, list , head); + if (buf->fence) { + (void) driFenceFinish(buf->fence, buf->fenceType, 0); + driFenceUnReference(&buf->fence); + } + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + _glthread_UNLOCK_MUTEX(header->mutex); +} + +static void +pool_takedown(struct _DriBufferPool *driPool) +{ + struct _DriSlabPool *pool = driPool->data; + int i; + + for (i=0; inumBuckets; ++i) { + driFinishSizeHeader(&pool->headers[i]); + } + + free(pool->headers); + free(pool->bucketSizes); + free(pool); + free(driPool); +} + +struct _DriBufferPool * +driSlabPoolInit(int fd, uint64_t flags, + uint64_t validMask, + uint32_t smallestSize, + uint32_t numSizes, + uint32_t desiredNumBuffers, + uint32_t maxSlabSize, + uint32_t pageAlignment, + struct _DriFreeSlabManager *fMan) +{ + struct _DriBufferPool *driPool; + struct _DriSlabPool *pool; + uint32_t i; + + driPool = calloc(1, sizeof(*driPool)); + if (!driPool) + return NULL; + + pool = calloc(1, sizeof(*pool)); + if (!pool) + goto out_err0; + + pool->bucketSizes = calloc(numSizes, sizeof(*pool->bucketSizes)); + if (!pool->bucketSizes) + goto out_err1; + + pool->headers = calloc(numSizes, sizeof(*pool->headers)); + if (!pool->headers) + goto out_err2; + + pool->fMan = fMan; + pool->proposedFlags = flags; + pool->validMask = validMask; + pool->numBuckets = numSizes; + pool->pageSize = getpagesize(); + pool->fd = fd; + pool->pageAlignment = pageAlignment; + pool->maxSlabSize = maxSlabSize; + pool->desiredNumBuffers = desiredNumBuffers; + + for (i=0; inumBuckets; ++i) { + pool->bucketSizes[i] = (smallestSize << i); + driInitSizeHeader(pool, pool->bucketSizes[i], + &pool->headers[i]); + } + + driPool->data = (void *) pool; + driPool->map = &pool_map; + driPool->unmap = &pool_unmap; + driPool->destroy = &pool_destroy; + driPool->offset = &pool_offset; + driPool->poolOffset = &pool_poolOffset; + driPool->flags = &pool_flags; + driPool->size = &pool_size; + driPool->create = &pool_create; + driPool->fence = &pool_fence; + driPool->kernel = &pool_kernel; + driPool->validate = &pool_validate; + driPool->waitIdle = &pool_waitIdle; + driPool->takeDown = &pool_takedown; + + return driPool; + + out_err2: + free(pool->bucketSizes); + out_err1: + free(pool); + out_err0: + free(driPool); + + return NULL; +} -- cgit v1.2.3 From 2b2d0e05842691e715782a64845aeca12a428427 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Apr 2008 13:50:02 -0600 Subject: gallium: clean-up in st_renderbuffer_alloc_storage() --- src/mesa/state_tracker/st_cb_fbo.c | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index ec7788923a..e55281a430 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -80,6 +80,8 @@ init_renderbuffer_bits(struct st_renderbuffer *strb, /** * gl_renderbuffer::AllocStorage() + * This is called to allocate the original drawing surface, and + * during window resize. */ static GLboolean st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, @@ -90,8 +92,10 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, struct st_renderbuffer *strb = st_renderbuffer(rb); enum pipe_format pipeFormat; GLbitfield flags = 0x0; /* XXX needed? */ + int ret; if (!strb->surface) { + /* first time surface creation */ strb->surface = pipe->winsys->surface_alloc(pipe->winsys); assert(strb->surface); assert(strb->surface->refcount); @@ -99,10 +103,10 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, if (!strb->surface) return GL_FALSE; } - - if (strb->surface->buffer) - pipe_buffer_reference(pipe->winsys, &strb->surface->buffer, - NULL); + else if (strb->surface->buffer) { + /* release/discard the old surface buffer */ + pipe_buffer_reference(pipe->winsys, &strb->surface->buffer, NULL); + } /* Determine surface format here */ if (strb->format != PIPE_FORMAT_NONE) { @@ -116,14 +120,15 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, init_renderbuffer_bits(strb, pipeFormat); - pipe->winsys->surface_alloc_storage(pipe->winsys, - strb->surface, - width, - height, - pipeFormat, - flags); - if (!strb->surface->buffer) + ret = pipe->winsys->surface_alloc_storage(pipe->winsys, + strb->surface, + width, + height, + pipeFormat, + flags); + if (ret || !strb->surface->buffer) { return GL_FALSE; /* out of memory, try s/w buffer? */ + } ASSERT(strb->surface->buffer); ASSERT(strb->surface->format); -- cgit v1.2.3 From 29b65a709cae779debc4bcdbfaa66cfe72f7074f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Apr 2008 14:17:32 -0600 Subject: gallium: remove stray comment --- src/mesa/state_tracker/st_cb_queryobj.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_queryobj.c b/src/mesa/state_tracker/st_cb_queryobj.c index c1d0d086b4..21c2c7dd9f 100644 --- a/src/mesa/state_tracker/st_cb_queryobj.c +++ b/src/mesa/state_tracker/st_cb_queryobj.c @@ -90,11 +90,7 @@ st_DeleteQuery(GLcontext *ctx, struct gl_query_object *q) FREE(stq); } -/** - * Do glReadPixels by getting rows from the framebuffer surface with - * get_tile(). Convert to requested format/type with Mesa image routines. - * Image transfer ops are done in software too. - */ + static void st_BeginQuery(GLcontext *ctx, struct gl_query_object *q) { -- cgit v1.2.3 From 0dd596fbc7f88b88467529a7f176aca70d70f731 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Apr 2008 14:53:49 -0600 Subject: gallium: accum buffer fixes If the driver can't create a PIPE_FORMAT_R16G16B16A16_SNORM surface, create an accum surface using a shallower format and taller height. Since only the accum buffer code accesses the surface the actual format doesn't really matter, just that there's enough memory. --- src/mesa/state_tracker/st_cb_accum.c | 61 +++++++++++++++++++++++++++++---- src/mesa/state_tracker/st_cb_fbo.c | 29 ++++++++++++++-- src/mesa/state_tracker/st_cb_fbo.h | 3 ++ src/mesa/state_tracker/st_framebuffer.c | 2 +- 4 files changed, 84 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index a623d0bcc0..35352351d5 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -56,6 +56,51 @@ */ +/** + * Wrapper for pipe_get_tile_rgba(). Do format/cpp override to make the + * tile util function think the surface is 16bit/channel, even if it's not. + * See also: st_renderbuffer_alloc_storage() + */ +static void +acc_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps, + uint x, uint y, uint w, uint h, float *p) +{ + const enum pipe_format f = acc_ps->format; + const int cpp = acc_ps->cpp; + + acc_ps->format = DEFAULT_ACCUM_PIPE_FORMAT; + acc_ps->cpp = 8; + + pipe_get_tile_rgba(pipe, acc_ps, x, y, w, h, p); + + acc_ps->format = f; + acc_ps->cpp = cpp; +} + + +/** + * Wrapper for pipe_put_tile_rgba(). Do format/cpp override to make the + * tile util function think the surface is 16bit/channel, even if it's not. + * See also: st_renderbuffer_alloc_storage() + */ +static void +acc_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps, + uint x, uint y, uint w, uint h, const float *p) +{ + enum pipe_format f = acc_ps->format; + const int cpp = acc_ps->cpp; + + acc_ps->format = DEFAULT_ACCUM_PIPE_FORMAT; + acc_ps->cpp = 8; + + pipe_put_tile_rgba(pipe, acc_ps, x, y, w, h, p); + + acc_ps->format = f; + acc_ps->cpp = cpp; +} + + + void st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { @@ -80,7 +125,9 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) accBuf[i*4+3] = a; } - pipe_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + + free(accBuf); } @@ -95,13 +142,13 @@ accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + acc_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] * scale + bias; } - pipe_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(accBuf); } @@ -120,13 +167,13 @@ accum_accum(struct pipe_context *pipe, GLfloat value, accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, colorBuf); - pipe_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + acc_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] + colorBuf[i] * value; } - pipe_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); free(colorBuf); free(accBuf); @@ -150,7 +197,7 @@ accum_load(struct pipe_context *pipe, GLfloat value, buf[i] = buf[i] * value; } - pipe_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, buf); + acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, buf); free(buf); } @@ -169,7 +216,7 @@ accum_return(GLcontext *ctx, GLfloat value, abuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); + acc_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) { cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index e55281a430..b1a56f3ca6 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -127,14 +127,37 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, pipeFormat, flags); if (ret || !strb->surface->buffer) { - return GL_FALSE; /* out of memory, try s/w buffer? */ + if (pipeFormat == DEFAULT_ACCUM_PIPE_FORMAT) { + /* Accum buffer. Try a different surface format. Since accum + * buffers are s/w only for now, the surface pixel format doesn't + * really matter, only that the buffer is large enough. + */ + int sz, mult; + enum pipe_format accum_format; + + /* allocate a buffer of (typically) double height to get 64bpp */ + accum_format = st_choose_renderbuffer_format(pipe, GL_RGBA); + sz = pf_get_size(accum_format); + mult = pf_get_size(DEFAULT_ACCUM_PIPE_FORMAT) / sz; + + ret = pipe->winsys->surface_alloc_storage(pipe->winsys, + strb->surface, + width, height * mult, + accum_format, flags); + if (ret) + return GL_FALSE; /* we've _really_ failed */ + + } + else { + return GL_FALSE; /* out of memory, try s/w buffer? */ + } } ASSERT(strb->surface->buffer); ASSERT(strb->surface->format); ASSERT(strb->surface->cpp); ASSERT(strb->surface->width == width); - ASSERT(strb->surface->height == height); + /*ASSERT(strb->surface->height == height);*/ ASSERT(strb->surface->pitch); strb->Base.Width = width; @@ -252,7 +275,7 @@ st_new_renderbuffer_fb(enum pipe_format format) strb->Base.InternalFormat = GL_STENCIL_INDEX8_EXT; strb->Base._BaseFormat = GL_STENCIL_INDEX; break; - case PIPE_FORMAT_R16G16B16A16_SNORM: + case DEFAULT_ACCUM_PIPE_FORMAT: /*PIPE_FORMAT_R16G16B16A16_SNORM*/ strb->Base.InternalFormat = GL_RGBA16; strb->Base._BaseFormat = GL_RGBA; break; diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index 21e531d1d0..c1aa14f9b2 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -30,6 +30,9 @@ #define ST_CB_FBO_H +#define DEFAULT_ACCUM_PIPE_FORMAT PIPE_FORMAT_R16G16B16A16_SNORM + + /** * Derived renderbuffer class. Just need to add a pointer to the diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 075e9d1bd6..ea09d9234c 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -107,7 +107,7 @@ st_create_framebuffer( const __GLcontextModes *visual, if (visual->accumRedBits > 0) { /* 16-bit/channel accum */ struct gl_renderbuffer *accumRb - = st_new_renderbuffer_fb(PIPE_FORMAT_R16G16B16A16_SNORM); + = st_new_renderbuffer_fb(DEFAULT_ACCUM_PIPE_FORMAT); _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb); } -- cgit v1.2.3 From 2b8b2420d48b1b9ce0ec4010ce3968a91a57a4c9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Apr 2008 15:05:28 -0600 Subject: gallium: slightly improved accum clear/mad operations Instead of get/put_tile(), map the buffer and operate on values directly. --- src/mesa/state_tracker/st_cb_accum.c | 88 ++++++++++++++++++++++-------------- 1 file changed, 55 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 35352351d5..4f9a22161c 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -104,53 +104,75 @@ acc_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps, void st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { - struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *acc_strb = st_renderbuffer(rb); struct pipe_surface *acc_ps = acc_strb->surface; const GLint xpos = ctx->DrawBuffer->_Xmin; const GLint ypos = ctx->DrawBuffer->_Ymin; const GLint width = ctx->DrawBuffer->_Xmax - xpos; const GLint height = ctx->DrawBuffer->_Ymax - ypos; - const GLfloat r = ctx->Accum.ClearColor[0]; - const GLfloat g = ctx->Accum.ClearColor[1]; - const GLfloat b = ctx->Accum.ClearColor[2]; - const GLfloat a = ctx->Accum.ClearColor[3]; - GLfloat *accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - int i; - - for (i = 0; i < width * height; i++) { - accBuf[i*4+0] = r; - accBuf[i*4+1] = g; - accBuf[i*4+2] = b; - accBuf[i*4+3] = a; + GLvoid *map; + + map = pipe_surface_map(acc_ps); + + /* note acc_strb->format might not equal acc_ps->format */ + switch (acc_strb->format) { + case PIPE_FORMAT_R16G16B16A16_SNORM: + { + GLshort r = FLOAT_TO_SHORT(ctx->Accum.ClearColor[0]); + GLshort g = FLOAT_TO_SHORT(ctx->Accum.ClearColor[1]); + GLshort b = FLOAT_TO_SHORT(ctx->Accum.ClearColor[2]); + GLshort a = FLOAT_TO_SHORT(ctx->Accum.ClearColor[3]); + int i, j; + for (i = 0; i < height; i++) { + GLshort *dst = ((GLshort *) map + + ((ypos + i) * acc_ps->pitch + xpos) * 4); + for (j = 0; j < width; j++) { + dst[0] = r; + dst[1] = g; + dst[2] = b; + dst[3] = a; + dst += 4; + } + } + } + break; + default: + _mesa_problem(ctx, "unexpected format in st_clear_accum_buffer()"); } - acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); - - free(accBuf); + pipe_surface_unmap(acc_ps); } /** For ADD/MULT */ static void -accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, +accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias, GLint xpos, GLint ypos, GLint width, GLint height, - struct pipe_surface *acc_ps) + struct st_renderbuffer *acc_strb) { - GLfloat *accBuf; - GLint i; - - accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - - acc_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); - - for (i = 0; i < 4 * width * height; i++) { - accBuf[i] = accBuf[i] * scale + bias; + struct pipe_surface *acc_ps = acc_strb->surface; + GLvoid *map; + + map = pipe_surface_map(acc_ps); + + /* note acc_strb->format might not equal acc_ps->format */ + switch (acc_strb->format) { + case PIPE_FORMAT_R16G16B16A16_SNORM: + { + int i, j; + for (i = 0; i < height; i++) { + GLshort *acc = ((GLshort *) map + + ((ypos + i) * acc_ps->pitch + xpos) * 4); + for (j = 0; j < width * 4; j++) { + float val = SHORT_TO_FLOAT(acc[j]) * scale + bias; + acc[j] = FLOAT_TO_SHORT(val); + } + } + } + break; + default: + _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()"); } - - acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); - - free(accBuf); } @@ -266,12 +288,12 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value) switch (op) { case GL_ADD: if (value != 0.0F) { - accum_mad(pipe, 1.0, value, xpos, ypos, width, height, acc_ps); + accum_mad(ctx, 1.0, value, xpos, ypos, width, height, acc_strb); } break; case GL_MULT: if (value != 1.0F) { - accum_mad(pipe, value, 0.0, xpos, ypos, width, height, acc_ps); + accum_mad(ctx, value, 0.0, xpos, ypos, width, height, acc_strb); } break; case GL_ACCUM: -- cgit v1.2.3 From d28a2004b8a69d66595f2c6b5398085b73873533 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Apr 2008 15:53:49 -0600 Subject: gallium: fix the texture case in default_deep_rgba_format() Fixes glean pixelFormat test --- src/mesa/state_tracker/st_format.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 20defd5e16..cd41dc4684 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -321,7 +321,10 @@ default_deep_rgba_format(struct pipe_screen *screen, uint type) if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SNORM, type)) { return PIPE_FORMAT_R16G16B16A16_SNORM; } - return PIPE_FORMAT_NONE; + if (type == PIPE_TEXTURE) + return default_rgba_format(screen, type); + else + return PIPE_FORMAT_NONE; } -- cgit v1.2.3 From 985134211d9fbfe9b7885beeeb85fdcf2ab78f87 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Apr 2008 16:45:17 -0600 Subject: gallium: check ctx->DrawBuffer before flushing bitmap cache. During context unbind, we may not have a draw buffer. This fixes demos/tunnel2.c --- src/mesa/state_tracker/st_cb_bitmap.c | 60 ++++++++++++++++++----------------- 1 file changed, 31 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 59f3c385df..505a13cc2b 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -547,36 +547,38 @@ void st_flush_bitmap_cache(struct st_context *st) { if (!st->bitmap.cache->empty) { - struct bitmap_cache *cache = st->bitmap.cache; - struct pipe_context *pipe = st->pipe; - struct pipe_screen *screen = pipe->screen; - struct pipe_surface *surf; - void *dest; - - assert(cache->xmin <= cache->xmax); - /* - printf("flush size %d x %d at %d, %d\n", - cache->xmax - cache->xmin, - cache->ymax - cache->ymin, - cache->xpos, cache->ypos); - */ - - /* update the texture map image */ - surf = screen->get_tex_surface(screen, cache->texture, 0, 0, 0); - dest = pipe_surface_map(surf); - memcpy(dest, cache->buffer, sizeof(cache->buffer)); - pipe_surface_unmap(surf); - pipe_surface_reference(&surf, NULL); - - pipe->texture_update(pipe, cache->texture, 0, 0x1); - - draw_bitmap_quad(st->ctx, - cache->xpos, - cache->ypos, - st->ctx->Current.RasterPos[2], - BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, - cache->texture); + if (st->ctx->DrawBuffer) { + struct bitmap_cache *cache = st->bitmap.cache; + struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *surf; + void *dest; + + assert(cache->xmin <= cache->xmax); + /* + printf("flush size %d x %d at %d, %d\n", + cache->xmax - cache->xmin, + cache->ymax - cache->ymin, + cache->xpos, cache->ypos); + */ + + /* update the texture map image */ + surf = screen->get_tex_surface(screen, cache->texture, 0, 0, 0); + dest = pipe_surface_map(surf); + memcpy(dest, cache->buffer, sizeof(cache->buffer)); + pipe_surface_unmap(surf); + pipe_surface_reference(&surf, NULL); + + pipe->texture_update(pipe, cache->texture, 0, 0x1); + + draw_bitmap_quad(st->ctx, + cache->xpos, + cache->ypos, + st->ctx->Current.RasterPos[2], + BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, + cache->texture); + } reset_cache(st); } } -- cgit v1.2.3 From 4e2127b0e5cb6411123e16dd562626cd70814a9a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 8 Apr 2008 11:30:36 +0900 Subject: gallium: Allow to debug memory leaks in nested scopes. --- src/gallium/auxiliary/util/p_debug_mem.c | 16 ++++++++-------- src/gallium/include/pipe/p_debug.h | 6 +++--- 2 files changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c index c160afe5b7..97ec9c5b39 100644 --- a/src/gallium/auxiliary/util/p_debug_mem.c +++ b/src/gallium/auxiliary/util/p_debug_mem.c @@ -70,8 +70,7 @@ struct debug_memory_header static struct list_head list = { &list, &list }; -static unsigned long start_no = 0; -static unsigned long end_no = 0; +static unsigned long last_no = 0; void * @@ -84,7 +83,7 @@ debug_malloc(const char *file, unsigned line, const char *function, if(!hdr) return NULL; - hdr->no = end_no++; + hdr->no = last_no++; hdr->file = file; hdr->line = line; hdr->function = function; @@ -147,14 +146,14 @@ debug_realloc(const char *file, unsigned line, const char *function, return new_ptr; } -void -debug_memory_reset(void) +unsigned long +debug_memory_begin(void) { - start_no = end_no; + return last_no; } void -debug_memory_report(void) +debug_memory_end(unsigned long start_no) { struct list_head *entry; @@ -164,7 +163,8 @@ debug_memory_report(void) void *ptr; hdr = LIST_ENTRY(struct debug_memory_header, entry, head); ptr = (void *)((char *)hdr + sizeof(*hdr)); - if(hdr->no >= start_no) + if(start_no <= hdr->no && hdr->no < last_no || + last_no < start_no && (hdr->no < last_no || start_no <= hdr->no)) debug_printf("%s:%u:%s: %u bytes at %p not freed\n", hdr->file, hdr->line, hdr->function, hdr->size, ptr); diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 9e52ad9a37..235c47abac 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -304,11 +304,11 @@ void * debug_realloc(const char *file, unsigned line, const char *function, void *old_ptr, size_t old_size, size_t new_size ); -void -debug_memory_reset(void); +unsigned long +debug_memory_begin(void); void -debug_memory_report(void); +debug_memory_end(unsigned long beginning); #ifdef __cplusplus -- cgit v1.2.3 From c7daa68ca312cc98abe351be2fef8d8246929627 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 7 Apr 2008 21:59:12 -0600 Subject: gallium: begin reworking quad stages for multiple color outputs --- src/gallium/drivers/softpipe/sp_context.h | 2 - src/gallium/drivers/softpipe/sp_fs_llvm.c | 16 +- src/gallium/drivers/softpipe/sp_headers.h | 4 +- src/gallium/drivers/softpipe/sp_quad.c | 9 - src/gallium/drivers/softpipe/sp_quad_alpha_test.c | 5 +- src/gallium/drivers/softpipe/sp_quad_blend.c | 1192 +++++++++++---------- src/gallium/drivers/softpipe/sp_quad_bufloop.c | 4 +- src/gallium/drivers/softpipe/sp_quad_colormask.c | 59 +- src/gallium/drivers/softpipe/sp_quad_coverage.c | 15 +- src/gallium/drivers/softpipe/sp_quad_fs.c | 60 +- src/gallium/drivers/softpipe/sp_quad_output.c | 39 +- 11 files changed, 734 insertions(+), 671 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index dc9d0e6d5d..9724d185d1 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -138,8 +138,6 @@ struct softpipe_context { struct draw_stage *vbuf; struct softpipe_vbuf_render *vbuf_render; - uint current_cbuf; /**< current color buffer being written to */ - struct softpipe_tile_cache *cbuf_cache[PIPE_MAX_COLOR_BUFS]; struct softpipe_tile_cache *zsbuf_cache; diff --git a/src/gallium/drivers/softpipe/sp_fs_llvm.c b/src/gallium/drivers/softpipe/sp_fs_llvm.c index 07d058155f..6e1d9280bb 100644 --- a/src/gallium/drivers/softpipe/sp_fs_llvm.c +++ b/src/gallium/drivers/softpipe/sp_fs_llvm.c @@ -99,19 +99,19 @@ shade_quad_llvm(struct quad_stage *qs, allvmrt(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] == TGSI_SEMANTIC_COLOR); for (i = 0; i < QUAD_SIZE; ++i) { - quad->outputs.color[0][i] = dests[i][qss->colorOutSlot][0]; - quad->outputs.color[1][i] = dests[i][qss->colorOutSlot][1]; - quad->outputs.color[2][i] = dests[i][qss->colorOutSlot][2]; - quad->outputs.color[3][i] = dests[i][qss->colorOutSlot][3]; + quad->outputs.color[0][0][i] = dests[i][qss->colorOutSlot][0]; + quad->outputs.color[0][1][i] = dests[i][qss->colorOutSlot][1]; + quad->outputs.color[0][2][i] = dests[i][qss->colorOutSlot][2]; + quad->outputs.color[0][3][i] = dests[i][qss->colorOutSlot][3]; } } #if DLLVM for (int i = 0; i < QUAD_SIZE; ++i) { debug_printf("QLLVM%d(%d) [%f, %f, %f, %f]\n", i, qss->colorOutSlot, - quad->outputs.color[0][i], - quad->outputs.color[1][i], - quad->outputs.color[2][i], - quad->outputs.color[3][i]); + quad->outputs.color[0][0][i], + quad->outputs.color[0][1][i], + quad->outputs.color[0][2][i], + quad->outputs.color[0][3][i]); } #endif diff --git a/src/gallium/drivers/softpipe/sp_headers.h b/src/gallium/drivers/softpipe/sp_headers.h index 9cf8222133..3d9ede69bb 100644 --- a/src/gallium/drivers/softpipe/sp_headers.h +++ b/src/gallium/drivers/softpipe/sp_headers.h @@ -31,6 +31,7 @@ #ifndef SP_HEADERS_H #define SP_HEADERS_H +#include "pipe/p_state.h" #include "tgsi/exec/tgsi_exec.h" #define PRIM_POINT 1 @@ -66,7 +67,8 @@ struct quad_header { unsigned prim:2; /**< PRIM_POINT, LINE, TRI */ struct { - float color[NUM_CHANNELS][QUAD_SIZE]; /* rrrr, gggg, bbbb, aaaa */ + /** colors in SOA format (rrrr, gggg, bbbb, aaaa) */ + float color[PIPE_MAX_COLOR_BUFS][NUM_CHANNELS][QUAD_SIZE]; float depth[QUAD_SIZE]; } outputs; diff --git a/src/gallium/drivers/softpipe/sp_quad.c b/src/gallium/drivers/softpipe/sp_quad.c index 8603c1a367..bc83d78ea1 100644 --- a/src/gallium/drivers/softpipe/sp_quad.c +++ b/src/gallium/drivers/softpipe/sp_quad.c @@ -76,15 +76,6 @@ sp_build_quad_pipeline(struct softpipe_context *sp) sp_push_quad_first( sp, sp->quad.blend ); } - if (sp->framebuffer.num_cbufs == 1) { - /* the usual case: write to exactly one colorbuf */ - sp->current_cbuf = 0; - } - else { - /* insert bufloop stage */ - sp_push_quad_first( sp, sp->quad.bufloop ); - } - if (sp->depth_stencil->depth.occlusion_count) { sp_push_quad_first( sp, sp->quad.occlusion ); } diff --git a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c index 318916fe30..7a42b08ef5 100644 --- a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c @@ -16,7 +16,8 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) struct softpipe_context *softpipe = qs->softpipe; const float ref = softpipe->depth_stencil->alpha.ref; unsigned passMask = 0x0, j; - const float *aaaa = quad->outputs.color[3]; + const uint cbuf = 0; /* only output[0].alpha is tested */ + const float *aaaa = quad->outputs.color[cbuf][3]; switch (softpipe->depth_stencil->alpha.func) { case PIPE_FUNC_NEVER: @@ -25,7 +26,7 @@ alpha_test_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_FUNC_LESS: /* * If mask were an array [4] we could do this SIMD-style: - * passMask = (quad->outputs.color[3] <= vec4(ref)); + * passMask = (quad->outputs.color[0][3] <= vec4(ref)); */ for (j = 0; j < QUAD_SIZE; j++) { if (aaaa[j] < ref) { diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index 0b79cfa1ed..8be8025f40 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -101,114 +101,119 @@ static void logicop_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - float dest[4][QUAD_SIZE]; - ubyte src[4][4], dst[4][4], res[4][4]; - uint *src4 = (uint *) src; - uint *dst4 = (uint *) dst; - uint *res4 = (uint *) res; - struct softpipe_cached_tile * - tile = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - float (*quadColor)[4] = quad->outputs.color; - uint i, j; - - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); - int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); - for (i = 0; i < 4; i++) { - dest[i][j] = tile->data.color[y][x][i]; + uint cbuf; + + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + float dest[4][QUAD_SIZE]; + ubyte src[4][4], dst[4][4], res[4][4]; + uint *src4 = (uint *) src; + uint *dst4 = (uint *) dst; + uint *res4 = (uint *) res; + struct softpipe_cached_tile * + tile = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color[cbuf]; + uint i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } } - } - /* convert to ubyte */ - for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][0], dest[j][0]); /* P0 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][1], dest[j][1]); /* P1 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][2], dest[j][2]); /* P2 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][3], dest[j][3]); /* P3 */ - - UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quadColor[j][0]); /* P0 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quadColor[j][1]); /* P1 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quadColor[j][2]); /* P2 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quadColor[j][3]); /* P3 */ - } + /* convert to ubyte */ + for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][0], dest[j][0]); /* P0 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][1], dest[j][1]); /* P1 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][2], dest[j][2]); /* P2 */ + UNCLAMPED_FLOAT_TO_UBYTE(dst[j][3], dest[j][3]); /* P3 */ + + UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quadColor[j][0]); /* P0 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quadColor[j][1]); /* P1 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quadColor[j][2]); /* P2 */ + UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quadColor[j][3]); /* P3 */ + } - switch (softpipe->blend->logicop_func) { - case PIPE_LOGICOP_CLEAR: - for (j = 0; j < 4; j++) - res4[j] = 0; - break; - case PIPE_LOGICOP_NOR: - for (j = 0; j < 4; j++) - res4[j] = ~(src4[j] | dst4[j]); - break; - case PIPE_LOGICOP_AND_INVERTED: - for (j = 0; j < 4; j++) - res4[j] = ~src4[j] & dst4[j]; - break; - case PIPE_LOGICOP_COPY_INVERTED: - for (j = 0; j < 4; j++) - res4[j] = ~src4[j]; - break; - case PIPE_LOGICOP_AND_REVERSE: - for (j = 0; j < 4; j++) - res4[j] = src4[j] & ~dst4[j]; - break; - case PIPE_LOGICOP_INVERT: - for (j = 0; j < 4; j++) - res4[j] = ~dst4[j]; - break; - case PIPE_LOGICOP_XOR: - for (j = 0; j < 4; j++) - res4[j] = dst4[j] ^ src4[j]; - break; - case PIPE_LOGICOP_NAND: - for (j = 0; j < 4; j++) - res4[j] = ~(src4[j] & dst4[j]); - break; - case PIPE_LOGICOP_AND: - for (j = 0; j < 4; j++) - res4[j] = src4[j] & dst4[j]; - break; - case PIPE_LOGICOP_EQUIV: - for (j = 0; j < 4; j++) - res4[j] = ~(src4[j] ^ dst4[j]); - break; - case PIPE_LOGICOP_NOOP: - for (j = 0; j < 4; j++) - res4[j] = dst4[j]; - break; - case PIPE_LOGICOP_OR_INVERTED: - for (j = 0; j < 4; j++) - res4[j] = ~src4[j] | dst4[j]; - break; - case PIPE_LOGICOP_COPY: - for (j = 0; j < 4; j++) - res4[j] = src4[j]; - break; - case PIPE_LOGICOP_OR_REVERSE: - for (j = 0; j < 4; j++) - res4[j] = src4[j] | ~dst4[j]; - break; - case PIPE_LOGICOP_OR: - for (j = 0; j < 4; j++) - res4[j] = src4[j] | dst4[j]; - break; - case PIPE_LOGICOP_SET: - for (j = 0; j < 4; j++) - res4[j] = ~0; - break; - default: - assert(0); - } + switch (softpipe->blend->logicop_func) { + case PIPE_LOGICOP_CLEAR: + for (j = 0; j < 4; j++) + res4[j] = 0; + break; + case PIPE_LOGICOP_NOR: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] | dst4[j]); + break; + case PIPE_LOGICOP_AND_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j] & dst4[j]; + break; + case PIPE_LOGICOP_COPY_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j]; + break; + case PIPE_LOGICOP_AND_REVERSE: + for (j = 0; j < 4; j++) + res4[j] = src4[j] & ~dst4[j]; + break; + case PIPE_LOGICOP_INVERT: + for (j = 0; j < 4; j++) + res4[j] = ~dst4[j]; + break; + case PIPE_LOGICOP_XOR: + for (j = 0; j < 4; j++) + res4[j] = dst4[j] ^ src4[j]; + break; + case PIPE_LOGICOP_NAND: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] & dst4[j]); + break; + case PIPE_LOGICOP_AND: + for (j = 0; j < 4; j++) + res4[j] = src4[j] & dst4[j]; + break; + case PIPE_LOGICOP_EQUIV: + for (j = 0; j < 4; j++) + res4[j] = ~(src4[j] ^ dst4[j]); + break; + case PIPE_LOGICOP_NOOP: + for (j = 0; j < 4; j++) + res4[j] = dst4[j]; + break; + case PIPE_LOGICOP_OR_INVERTED: + for (j = 0; j < 4; j++) + res4[j] = ~src4[j] | dst4[j]; + break; + case PIPE_LOGICOP_COPY: + for (j = 0; j < 4; j++) + res4[j] = src4[j]; + break; + case PIPE_LOGICOP_OR_REVERSE: + for (j = 0; j < 4; j++) + res4[j] = src4[j] | ~dst4[j]; + break; + case PIPE_LOGICOP_OR: + for (j = 0; j < 4; j++) + res4[j] = src4[j] | dst4[j]; + break; + case PIPE_LOGICOP_SET: + for (j = 0; j < 4; j++) + res4[j] = ~0; + break; + default: + assert(0); + } - for (j = 0; j < 4; j++) { - quadColor[j][0] = UBYTE_TO_FLOAT(res[j][0]); - quadColor[j][1] = UBYTE_TO_FLOAT(res[j][1]); - quadColor[j][2] = UBYTE_TO_FLOAT(res[j][2]); - quadColor[j][3] = UBYTE_TO_FLOAT(res[j][3]); + for (j = 0; j < 4; j++) { + quadColor[j][0] = UBYTE_TO_FLOAT(res[j][0]); + quadColor[j][1] = UBYTE_TO_FLOAT(res[j][1]); + quadColor[j][2] = UBYTE_TO_FLOAT(res[j][2]); + quadColor[j][3] = UBYTE_TO_FLOAT(res[j][3]); + } } /* pass quad to next stage */ @@ -221,504 +226,511 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) static void blend_quad(struct quad_stage *qs, struct quad_header *quad) { - struct softpipe_context *softpipe = qs->softpipe; static const float zero[4] = { 0, 0, 0, 0 }; static const float one[4] = { 1, 1, 1, 1 }; - float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - float (*quadColor)[4] = quad->outputs.color; - uint i, j; - - if (softpipe->blend->logicop_enable) { - logicop_quad(qs, quad); - return; - } - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); - int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); - for (i = 0; i < 4; i++) { - dest[i][j] = tile->data.color[y][x][i]; + struct softpipe_context *softpipe = qs->softpipe; + uint cbuf; + + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color[cbuf]; + uint i, j; + + if (softpipe->blend->logicop_enable) { + logicop_quad(qs, quad); + return; } - } - /* - * Compute src/first term RGB - */ - switch (softpipe->blend->rgb_src_factor) { - case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[0], quadColor[0]); /* R */ - VEC4_COPY(source[1], quadColor[1]); /* G */ - VEC4_COPY(source[2], quadColor[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - VEC4_MUL(source[0], quadColor[0], quadColor[0]); /* R */ - VEC4_MUL(source[1], quadColor[1], quadColor[1]); /* G */ - VEC4_MUL(source[2], quadColor[2], quadColor[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - { - const float *alpha = quadColor[3]; - VEC4_MUL(source[0], quadColor[0], alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_DST_COLOR: - VEC4_MUL(source[0], quadColor[0], dest[0]); /* R */ - VEC4_MUL(source[1], quadColor[1], dest[1]); /* G */ - VEC4_MUL(source[2], quadColor[2], dest[2]); /* B */ - break; - case PIPE_BLENDFACTOR_DST_ALPHA: - { - const float *alpha = dest[3]; - VEC4_MUL(source[0], quadColor[0], alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - { - const float *alpha = quadColor[3]; - float diff[4]; - VEC4_SUB(diff, one, dest[3]); - VEC4_MIN(source[0], alpha, diff); /* R */ - VEC4_MIN(source[1], alpha, diff); /* G */ - VEC4_MIN(source[2], alpha, diff); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ - VEC4_MUL(source[0], quadColor[0], comp); /* R */ - VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ - VEC4_MUL(source[1], quadColor[1], comp); /* G */ - VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ - VEC4_MUL(source[2], quadColor[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float alpha[4]; - VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quadColor[0], alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_SRC1_COLOR: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_SRC1_ALPHA: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(source[0], zero); /* R */ - VEC4_COPY(source[1], zero); /* G */ - VEC4_COPY(source[2], zero); /* B */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ - VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ - VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ - VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, quadColor[3]); - VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, dest[3]); - VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[0]); /* R */ - VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, dest[1]); /* G */ - VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, dest[2]); /* B */ - VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - { - float inv_comp[4]; - /* R */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); - VEC4_MUL(source[0], quadColor[0], inv_comp); - /* G */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); - VEC4_MUL(source[1], quadColor[1], inv_comp); - /* B */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); - VEC4_MUL(source[2], quadColor[2], inv_comp); - } - break; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_alpha[4]; - VEC4_SCALAR(inv_alpha, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ - VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ - VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - assert(0); /* to do */ - break; - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: - assert(0); /* to do */ - break; - default: - assert(0); - } - - /* - * Compute src/first term A - */ - switch (softpipe->blend->alpha_src_factor) { - case PIPE_BLENDFACTOR_ONE: - VEC4_COPY(source[3], quadColor[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_SRC_ALPHA: - { - const float *alpha = quadColor[3]; - VEC4_MUL(source[3], quadColor[3], alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - { - const float *alpha = quadColor[3]; - float diff[4]; - VEC4_SUB(diff, one, dest[3]); - VEC4_MIN(source[3], alpha, diff); /* A */ - } - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(source[3], quadColor[3], comp); /* A */ + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(source[3], zero); /* A */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, quadColor[3]); - VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_alpha[4]; - VEC4_SUB(inv_alpha, one, dest[3]); - VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_comp[4]; - /* A */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(source[3], quadColor[3], inv_comp); - } - break; - default: - assert(0); - } - - - /* - * Compute dest/second term RGB - */ - switch (softpipe->blend->rgb_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - /* dest = dest * 1 NO-OP, leave dest as-is */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - VEC4_MUL(dest[0], dest[0], quadColor[0]); /* R */ - VEC4_MUL(dest[1], dest[1], quadColor[1]); /* G */ - VEC4_MUL(dest[2], dest[2], quadColor[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA: - VEC4_MUL(dest[0], dest[0], quadColor[3]); /* R * A */ - VEC4_MUL(dest[1], dest[1], quadColor[3]); /* G * A */ - VEC4_MUL(dest[2], dest[2], quadColor[3]); /* B * A */ - break; - case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */ - VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */ - VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */ - break; - case PIPE_BLENDFACTOR_DST_COLOR: - VEC4_MUL(dest[0], dest[0], dest[0]); /* R */ - VEC4_MUL(dest[1], dest[1], dest[1]); /* G */ - VEC4_MUL(dest[2], dest[2], dest[2]); /* B */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - assert(0); /* illegal */ - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ - VEC4_MUL(dest[0], dest[0], comp); /* R */ - VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ - VEC4_MUL(dest[1], dest[1], comp); /* G */ - VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ - VEC4_MUL(dest[2], dest[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(dest[0], dest[0], comp); /* R */ - VEC4_MUL(dest[1], dest[1], comp); /* G */ - VEC4_MUL(dest[2], dest[2], comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[0], zero); /* R */ - VEC4_COPY(dest[1], zero); /* G */ - VEC4_COPY(dest[2], zero); /* B */ - break; - case PIPE_BLENDFACTOR_SRC1_COLOR: - case PIPE_BLENDFACTOR_SRC1_ALPHA: - /* XXX what are these? */ - assert(0); - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ - VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ - VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ - VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ - VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ - VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quadColor[3]); - VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ - VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ - VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[3]); /* A */ - VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ - VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ - VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[0]); /* R */ - VEC4_MUL(dest[0], dest[0], inv_comp); /* R */ - VEC4_SUB(inv_comp, one, dest[1]); /* G */ - VEC4_MUL(dest[1], dest[1], inv_comp); /* G */ - VEC4_SUB(inv_comp, one, dest[2]); /* B */ - VEC4_MUL(dest[2], dest[2], inv_comp); /* B */ - } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - { - float inv_comp[4]; - /* R */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); - VEC4_MUL(dest[0], dest[0], inv_comp); - /* G */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); - VEC4_MUL(dest[1], dest[1], inv_comp); - /* B */ - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); - VEC4_MUL(dest[2], dest[2], inv_comp); + + /* + * Compute src/first term RGB + */ + switch (softpipe->blend->rgb_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[0], quadColor[0]); /* R */ + VEC4_COPY(source[1], quadColor[1]); /* G */ + VEC4_COPY(source[2], quadColor[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + VEC4_MUL(source[0], quadColor[0], quadColor[0]); /* R */ + VEC4_MUL(source[1], quadColor[1], quadColor[1]); /* G */ + VEC4_MUL(source[2], quadColor[2], quadColor[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const float *alpha = quadColor[3]; + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_DST_COLOR: + VEC4_MUL(source[0], quadColor[0], dest[0]); /* R */ + VEC4_MUL(source[1], quadColor[1], dest[1]); /* G */ + VEC4_MUL(source[2], quadColor[2], dest[2]); /* B */ + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + { + const float *alpha = dest[3]; + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + { + const float *alpha = quadColor[3]; + float diff[4]; + VEC4_SUB(diff, one, dest[3]); + VEC4_MIN(source[0], alpha, diff); /* R */ + VEC4_MIN(source[1], alpha, diff); /* G */ + VEC4_MIN(source[2], alpha, diff); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], comp); /* R */ + VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], comp); /* G */ + VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float alpha[4]; + VEC4_SCALAR(alpha, softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quadColor[0], alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_SRC1_ALPHA: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[0], zero); /* R */ + VEC4_COPY(source[1], zero); /* G */ + VEC4_COPY(source[2], zero); /* B */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, quadColor[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, dest[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[0]); /* R */ + VEC4_MUL(source[0], quadColor[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, dest[1]); /* G */ + VEC4_MUL(source[1], quadColor[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, dest[2]); /* B */ + VEC4_MUL(source[2], quadColor[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + { + float inv_comp[4]; + /* R */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); + VEC4_MUL(source[0], quadColor[0], inv_comp); + /* G */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); + VEC4_MUL(source[1], quadColor[1], inv_comp); + /* B */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); + VEC4_MUL(source[2], quadColor[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_alpha[4]; + VEC4_SCALAR(inv_alpha, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */ + VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */ + VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + assert(0); /* to do */ + break; + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + assert(0); /* to do */ + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_comp[4]; - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(dest[0], dest[0], inv_comp); - VEC4_MUL(dest[1], dest[1], inv_comp); - VEC4_MUL(dest[2], dest[2], inv_comp); + + /* + * Compute src/first term A + */ + switch (softpipe->blend->alpha_src_factor) { + case PIPE_BLENDFACTOR_ONE: + VEC4_COPY(source[3], quadColor[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_SRC_ALPHA: + { + const float *alpha = quadColor[3]; + VEC4_MUL(source[3], quadColor[3], alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + { + const float *alpha = quadColor[3]; + float diff[4]; + VEC4_SUB(diff, one, dest[3]); + VEC4_MIN(source[3], alpha, diff); /* A */ + } + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(source[3], quadColor[3], comp); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(source[3], zero); /* A */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, quadColor[3]); + VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_alpha[4]; + VEC4_SUB(inv_alpha, one, dest[3]); + VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + /* A */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(source[3], quadColor[3], inv_comp); + } + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_INV_SRC1_COLOR: - case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: - /* XXX what are these? */ - assert(0); - break; - default: - assert(0); - } - - /* - * Compute dest/second term A - */ - switch (softpipe->blend->alpha_dst_factor) { - case PIPE_BLENDFACTOR_ONE: - /* dest = dest * 1 NO-OP, leave dest as-is */ - break; - case PIPE_BLENDFACTOR_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_SRC_ALPHA: - VEC4_MUL(dest[3], dest[3], quadColor[3]); /* A * A */ - break; - case PIPE_BLENDFACTOR_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_DST_ALPHA: - VEC4_MUL(dest[3], dest[3], dest[3]); /* A */ - break; - case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - assert(0); /* illegal */ - break; - case PIPE_BLENDFACTOR_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_CONST_ALPHA: - { - float comp[4]; - VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ - VEC4_MUL(dest[3], dest[3], comp); /* A */ + + + /* + * Compute dest/second term RGB + */ + switch (softpipe->blend->rgb_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + VEC4_MUL(dest[0], dest[0], quadColor[0]); /* R */ + VEC4_MUL(dest[1], dest[1], quadColor[1]); /* G */ + VEC4_MUL(dest[2], dest[2], quadColor[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA: + VEC4_MUL(dest[0], dest[0], quadColor[3]); /* R * A */ + VEC4_MUL(dest[1], dest[1], quadColor[3]); /* G * A */ + VEC4_MUL(dest[2], dest[2], quadColor[3]); /* B * A */ + break; + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */ + VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */ + VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */ + break; + case PIPE_BLENDFACTOR_DST_COLOR: + VEC4_MUL(dest[0], dest[0], dest[0]); /* R */ + VEC4_MUL(dest[1], dest[1], dest[1]); /* G */ + VEC4_MUL(dest[2], dest[2], dest[2]); /* B */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* illegal */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[0]); /* R */ + VEC4_MUL(dest[0], dest[0], comp); /* R */ + VEC4_SCALAR(comp, softpipe->blend_color.color[1]); /* G */ + VEC4_MUL(dest[1], dest[1], comp); /* G */ + VEC4_SCALAR(comp, softpipe->blend_color.color[2]); /* B */ + VEC4_MUL(dest[2], dest[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(dest[0], dest[0], comp); /* R */ + VEC4_MUL(dest[1], dest[1], comp); /* G */ + VEC4_MUL(dest[2], dest[2], comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[0], zero); /* R */ + VEC4_COPY(dest[1], zero); /* G */ + VEC4_COPY(dest[2], zero); /* B */ + break; + case PIPE_BLENDFACTOR_SRC1_COLOR: + case PIPE_BLENDFACTOR_SRC1_ALPHA: + /* XXX what are these? */ + assert(0); + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quadColor[0]); /* R */ + VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ + VEC4_SUB(inv_comp, one, quadColor[1]); /* G */ + VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ + VEC4_SUB(inv_comp, one, quadColor[2]); /* B */ + VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quadColor[3]); + VEC4_MUL(dest[0], dest[0], one_minus_alpha); /* R */ + VEC4_MUL(dest[1], dest[1], one_minus_alpha); /* G */ + VEC4_MUL(dest[2], dest[2], one_minus_alpha); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, quadColor[3]); /* A */ + VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ + VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ + VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[0]); /* R */ + VEC4_MUL(dest[0], dest[0], inv_comp); /* R */ + VEC4_SUB(inv_comp, one, dest[1]); /* G */ + VEC4_MUL(dest[1], dest[1], inv_comp); /* G */ + VEC4_SUB(inv_comp, one, dest[2]); /* B */ + VEC4_MUL(dest[2], dest[2], inv_comp); /* B */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + { + float inv_comp[4]; + /* R */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[0]); + VEC4_MUL(dest[0], dest[0], inv_comp); + /* G */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[1]); + VEC4_MUL(dest[1], dest[1], inv_comp); + /* B */ + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[2]); + VEC4_MUL(dest[2], dest[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(dest[0], dest[0], inv_comp); + VEC4_MUL(dest[1], dest[1], inv_comp); + VEC4_MUL(dest[2], dest[2], inv_comp); + } + break; + case PIPE_BLENDFACTOR_INV_SRC1_COLOR: + case PIPE_BLENDFACTOR_INV_SRC1_ALPHA: + /* XXX what are these? */ + assert(0); + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_ZERO: - VEC4_COPY(dest[3], zero); /* A */ - break; - case PIPE_BLENDFACTOR_INV_SRC_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_SRC_ALPHA: - { - float one_minus_alpha[QUAD_SIZE]; - VEC4_SUB(one_minus_alpha, one, quadColor[3]); - VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ + + /* + * Compute dest/second term A + */ + switch (softpipe->blend->alpha_dst_factor) { + case PIPE_BLENDFACTOR_ONE: + /* dest = dest * 1 NO-OP, leave dest as-is */ + break; + case PIPE_BLENDFACTOR_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_SRC_ALPHA: + VEC4_MUL(dest[3], dest[3], quadColor[3]); /* A * A */ + break; + case PIPE_BLENDFACTOR_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_DST_ALPHA: + VEC4_MUL(dest[3], dest[3], dest[3]); /* A */ + break; + case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: + assert(0); /* illegal */ + break; + case PIPE_BLENDFACTOR_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_CONST_ALPHA: + { + float comp[4]; + VEC4_SCALAR(comp, softpipe->blend_color.color[3]); /* A */ + VEC4_MUL(dest[3], dest[3], comp); /* A */ + } + break; + case PIPE_BLENDFACTOR_ZERO: + VEC4_COPY(dest[3], zero); /* A */ + break; + case PIPE_BLENDFACTOR_INV_SRC_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_SRC_ALPHA: + { + float one_minus_alpha[QUAD_SIZE]; + VEC4_SUB(one_minus_alpha, one, quadColor[3]); + VEC4_MUL(dest[3], dest[3], one_minus_alpha); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_DST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_DST_ALPHA: + { + float inv_comp[4]; + VEC4_SUB(inv_comp, one, dest[3]); /* A */ + VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */ + } + break; + case PIPE_BLENDFACTOR_INV_CONST_COLOR: + /* fall-through */ + case PIPE_BLENDFACTOR_INV_CONST_ALPHA: + { + float inv_comp[4]; + VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); + VEC4_MUL(dest[3], dest[3], inv_comp); + } + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_INV_DST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_DST_ALPHA: - { - float inv_comp[4]; - VEC4_SUB(inv_comp, one, dest[3]); /* A */ - VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */ + + /* + * Combine RGB terms + */ + switch (softpipe->blend->rgb_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quadColor[0], source[0], dest[0]); /* R */ + VEC4_ADD(quadColor[1], source[1], dest[1]); /* G */ + VEC4_ADD(quadColor[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quadColor[0], source[0], dest[0]); /* R */ + VEC4_SUB(quadColor[1], source[1], dest[1]); /* G */ + VEC4_SUB(quadColor[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quadColor[0], dest[0], source[0]); /* R */ + VEC4_SUB(quadColor[1], dest[1], source[1]); /* G */ + VEC4_SUB(quadColor[2], dest[2], source[2]); /* B */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quadColor[0], source[0], dest[0]); /* R */ + VEC4_MIN(quadColor[1], source[1], dest[1]); /* G */ + VEC4_MIN(quadColor[2], source[2], dest[2]); /* B */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quadColor[0], source[0], dest[0]); /* R */ + VEC4_MAX(quadColor[1], source[1], dest[1]); /* G */ + VEC4_MAX(quadColor[2], source[2], dest[2]); /* B */ + break; + default: + assert(0); } - break; - case PIPE_BLENDFACTOR_INV_CONST_COLOR: - /* fall-through */ - case PIPE_BLENDFACTOR_INV_CONST_ALPHA: - { - float inv_comp[4]; - VEC4_SCALAR(inv_comp, 1.0f - softpipe->blend_color.color[3]); - VEC4_MUL(dest[3], dest[3], inv_comp); + + /* + * Combine A terms + */ + switch (softpipe->blend->alpha_func) { + case PIPE_BLEND_ADD: + VEC4_ADD(quadColor[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_SUBTRACT: + VEC4_SUB(quadColor[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_REVERSE_SUBTRACT: + VEC4_SUB(quadColor[3], dest[3], source[3]); /* A */ + break; + case PIPE_BLEND_MIN: + VEC4_MIN(quadColor[3], source[3], dest[3]); /* A */ + break; + case PIPE_BLEND_MAX: + VEC4_MAX(quadColor[3], source[3], dest[3]); /* A */ + break; + default: + assert(0); } - break; - default: - assert(0); - } - - /* - * Combine RGB terms - */ - switch (softpipe->blend->rgb_func) { - case PIPE_BLEND_ADD: - VEC4_ADD(quadColor[0], source[0], dest[0]); /* R */ - VEC4_ADD(quadColor[1], source[1], dest[1]); /* G */ - VEC4_ADD(quadColor[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quadColor[0], source[0], dest[0]); /* R */ - VEC4_SUB(quadColor[1], source[1], dest[1]); /* G */ - VEC4_SUB(quadColor[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quadColor[0], dest[0], source[0]); /* R */ - VEC4_SUB(quadColor[1], dest[1], source[1]); /* G */ - VEC4_SUB(quadColor[2], dest[2], source[2]); /* B */ - break; - case PIPE_BLEND_MIN: - VEC4_MIN(quadColor[0], source[0], dest[0]); /* R */ - VEC4_MIN(quadColor[1], source[1], dest[1]); /* G */ - VEC4_MIN(quadColor[2], source[2], dest[2]); /* B */ - break; - case PIPE_BLEND_MAX: - VEC4_MAX(quadColor[0], source[0], dest[0]); /* R */ - VEC4_MAX(quadColor[1], source[1], dest[1]); /* G */ - VEC4_MAX(quadColor[2], source[2], dest[2]); /* B */ - break; - default: - assert(0); - } - - /* - * Combine A terms - */ - switch (softpipe->blend->alpha_func) { - case PIPE_BLEND_ADD: - VEC4_ADD(quadColor[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_SUBTRACT: - VEC4_SUB(quadColor[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_REVERSE_SUBTRACT: - VEC4_SUB(quadColor[3], dest[3], source[3]); /* A */ - break; - case PIPE_BLEND_MIN: - VEC4_MIN(quadColor[3], source[3], dest[3]); /* A */ - break; - case PIPE_BLEND_MAX: - VEC4_MAX(quadColor[3], source[3], dest[3]); /* A */ - break; - default: - assert(0); - } - + + } /* cbuf loop */ + /* pass blended quad to next stage */ qs->next->run(qs->next, quad); } diff --git a/src/gallium/drivers/softpipe/sp_quad_bufloop.c b/src/gallium/drivers/softpipe/sp_quad_bufloop.c index 2ae4e22a7d..b3db428ef1 100644 --- a/src/gallium/drivers/softpipe/sp_quad_bufloop.c +++ b/src/gallium/drivers/softpipe/sp_quad_bufloop.c @@ -13,7 +13,7 @@ static void cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - float tmp[4][QUAD_SIZE]; + float tmp[PIPE_MAX_COLOR_BUFS][4][QUAD_SIZE]; unsigned i; assert(sizeof(quad->outputs.color) == sizeof(tmp)); @@ -30,7 +30,9 @@ cbuf_loop_quad(struct quad_stage *qs, struct quad_header *quad) for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) { /* set current cbuffer */ +#if 0 /* obsolete & going away */ softpipe->current_cbuf = i; +#endif /* pass blended quad to next stage */ qs->next->run(qs->next, quad); diff --git a/src/gallium/drivers/softpipe/sp_quad_colormask.c b/src/gallium/drivers/softpipe/sp_quad_colormask.c index 1f09d900ca..7fe080990b 100644 --- a/src/gallium/drivers/softpipe/sp_quad_colormask.c +++ b/src/gallium/drivers/softpipe/sp_quad_colormask.c @@ -47,38 +47,43 @@ static void colormask_quad(struct quad_stage *qs, struct quad_header *quad) { struct softpipe_context *softpipe = qs->softpipe; - float dest[4][QUAD_SIZE]; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); - float (*quadColor)[4] = quad->outputs.color; - uint i, j; - - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); - int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); - for (i = 0; i < 4; i++) { - dest[i][j] = tile->data.color[y][x][i]; + uint cbuf; + + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + float dest[4][QUAD_SIZE]; + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color[cbuf]; + uint i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); + int y = (quad->y0 & (TILE_SIZE-1)) + (j >> 1); + for (i = 0; i < 4; i++) { + dest[i][j] = tile->data.color[y][x][i]; + } } - } - /* R */ - if (!(softpipe->blend->colormask & PIPE_MASK_R)) - COPY_4V(quadColor[0], dest[0]); + /* R */ + if (!(softpipe->blend->colormask & PIPE_MASK_R)) + COPY_4V(quadColor[0], dest[0]); - /* G */ - if (!(softpipe->blend->colormask & PIPE_MASK_G)) - COPY_4V(quadColor[1], dest[1]); + /* G */ + if (!(softpipe->blend->colormask & PIPE_MASK_G)) + COPY_4V(quadColor[1], dest[1]); - /* B */ - if (!(softpipe->blend->colormask & PIPE_MASK_B)) - COPY_4V(quadColor[2], dest[2]); + /* B */ + if (!(softpipe->blend->colormask & PIPE_MASK_B)) + COPY_4V(quadColor[2], dest[2]); - /* A */ - if (!(softpipe->blend->colormask & PIPE_MASK_A)) - COPY_4V(quadColor[3], dest[3]); + /* A */ + if (!(softpipe->blend->colormask & PIPE_MASK_A)) + COPY_4V(quadColor[3], dest[3]); + } /* pass quad to next stage */ qs->next->run(qs->next, quad); diff --git a/src/gallium/drivers/softpipe/sp_quad_coverage.c b/src/gallium/drivers/softpipe/sp_quad_coverage.c index b3d3fae22f..dd5ebb2296 100644 --- a/src/gallium/drivers/softpipe/sp_quad_coverage.c +++ b/src/gallium/drivers/softpipe/sp_quad_coverage.c @@ -50,12 +50,17 @@ coverage_quad(struct quad_stage *qs, struct quad_header *quad) if ((softpipe->rasterizer->poly_smooth && quad->prim == PRIM_TRI) || (softpipe->rasterizer->line_smooth && quad->prim == PRIM_LINE) || (softpipe->rasterizer->point_smooth && quad->prim == PRIM_POINT)) { - float (*quadColor)[4] = quad->outputs.color; - unsigned j; - for (j = 0; j < QUAD_SIZE; j++) { - assert(quad->coverage[j] >= 0.0); - assert(quad->coverage[j] <= 1.0); + uint cbuf; + + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + float (*quadColor)[4] = quad->outputs.color[cbuf]; + unsigned j; + for (j = 0; j < QUAD_SIZE; j++) { + assert(quad->coverage[j] >= 0.0); + assert(quad->coverage[j] <= 1.0); quadColor[3][j] *= quad->coverage[j]; + } } } diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index a73df31383..9a20c056a2 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -88,22 +88,64 @@ shade_quad( &qss->machine, quad ); - /* store result color */ +#if 0 /* XXX multi color outputs - untested */ + /* store outputs */ + boolean z_written = FALSE; + { + const ubyte *sem_name = softpipe->fs->info.output_semantic_name; + const ubyte *sem_index = softpipe->fs->info.output_semantic_index; + const uint n = qss->stage.softpipe->fs->info.num_outputs; + uint i; + for (i = 0; i < n; i++) { + switch (sem_name[i]) { + case TGSI_SEMANTIC_COLOR: + { + uint cbuf = sem_index[i]; + memcpy(quad->outputs.color[cbuf], + &machine->Outputs[i].xyzw[0].f[0], + sizeof(quad->outputs.color[0]) ); + } + break; + case TGSI_SEMANTIC_POSITION: + { + uint j; + for (j = 0; j < 4; j++) { + quad->outputs.depth[j] = machine->Outputs[0].xyzw[2].f[j]; + } + z_written = TRUE; + } + break; + } + } + } + + if (!z_written) { + /* compute Z values now, as in the quad earlyz stage */ + /* XXX we should really only do this if the earlyz stage is not used */ + const float fx = (float) quad->x0; + const float fy = (float) quad->y0; + const float dzdx = quad->posCoef->dadx[2]; + const float dzdy = quad->posCoef->dady[2]; + const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy; + + quad->outputs.depth[0] = z0; + quad->outputs.depth[1] = z0 + dzdx; + quad->outputs.depth[2] = z0 + dzdy; + quad->outputs.depth[3] = z0 + dzdx + dzdy; + } +#endif + + /* store result color(s) */ if (qss->colorOutSlot >= 0) { /* XXX need to handle multiple color outputs someday */ - assert(qss->stage.softpipe->fs->info.output_semantic_name[qss->colorOutSlot] + assert(softpipe->fs->info.output_semantic_name[qss->colorOutSlot] == TGSI_SEMANTIC_COLOR); memcpy( - quad->outputs.color, + quad->outputs.color[0], &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], - sizeof( quad->outputs.color ) ); + sizeof( quad->outputs.color[0] ) ); } - /* - * XXX the following code for updating quad->outputs.depth - * isn't really needed if we did early z testing. - */ - /* store result Z */ if (qss->depthOutSlot >= 0) { /* output[slot] is new Z */ diff --git a/src/gallium/drivers/softpipe/sp_quad_output.c b/src/gallium/drivers/softpipe/sp_quad_output.c index cfe8f11808..40083138a4 100644 --- a/src/gallium/drivers/softpipe/sp_quad_output.c +++ b/src/gallium/drivers/softpipe/sp_quad_output.c @@ -34,31 +34,36 @@ /** - * Write quad to framebuffer, taking mask into account. - * - * Note that surfaces support only full quad reads and writes. + * Last step of quad processing: write quad colors to the framebuffer, + * taking mask into account. */ static void output_quad(struct quad_stage *qs, struct quad_header *quad) { - struct softpipe_context *softpipe = qs->softpipe; - struct softpipe_cached_tile *tile - = sp_get_cached_tile(softpipe, - softpipe->cbuf_cache[softpipe->current_cbuf], - quad->x0, quad->y0); /* in-tile pos: */ const int itx = quad->x0 % TILE_SIZE; const int ity = quad->y0 % TILE_SIZE; - float (*quadColor)[4] = quad->outputs.color; - int i, j; - /* get/swizzle dest colors */ - for (j = 0; j < QUAD_SIZE; j++) { - if (quad->mask & (1 << j)) { - int x = itx + (j & 1); - int y = ity + (j >> 1); - for (i = 0; i < 4; i++) { /* loop over color chans */ - tile->data.color[y][x][i] = quadColor[i][j]; + struct softpipe_context *softpipe = qs->softpipe; + uint cbuf; + + /* loop over colorbuffer outputs */ + for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { + struct softpipe_cached_tile *tile + = sp_get_cached_tile(softpipe, + softpipe->cbuf_cache[cbuf], + quad->x0, quad->y0); + float (*quadColor)[4] = quad->outputs.color[cbuf]; + int i, j; + + /* get/swizzle dest colors */ + for (j = 0; j < QUAD_SIZE; j++) { + if (quad->mask & (1 << j)) { + int x = itx + (j & 1); + int y = ity + (j >> 1); + for (i = 0; i < 4; i++) { /* loop over color chans */ + tile->data.color[y][x][i] = quadColor[i][j]; + } } } } -- cgit v1.2.3 From 17f640990350823f530257ee1b1fa7fa1f83493f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 7 Apr 2008 22:00:41 -0600 Subject: gallium: get rid of bufloop quad stage --- src/gallium/drivers/softpipe/Makefile | 1 - src/gallium/drivers/softpipe/SConscript | 1 - src/gallium/drivers/softpipe/sp_context.c | 2 -- src/gallium/drivers/softpipe/sp_context.h | 1 - src/gallium/drivers/softpipe/sp_quad.h | 1 - 5 files changed, 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile index f32db35d58..19dfd8c1a7 100644 --- a/src/gallium/drivers/softpipe/Makefile +++ b/src/gallium/drivers/softpipe/Makefile @@ -17,7 +17,6 @@ C_SOURCES = \ sp_quad.c \ sp_quad_alpha_test.c \ sp_quad_blend.c \ - sp_quad_bufloop.c \ sp_quad_colormask.c \ sp_quad_coverage.c \ sp_quad_depth_test.c \ diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript index 88c21ee3b0..06931fa8d8 100644 --- a/src/gallium/drivers/softpipe/SConscript +++ b/src/gallium/drivers/softpipe/SConscript @@ -16,7 +16,6 @@ softpipe = env.ConvenienceLibrary( 'sp_prim_vbuf.c', 'sp_quad_alpha_test.c', 'sp_quad_blend.c', - 'sp_quad_bufloop.c', 'sp_quad.c', 'sp_quad_colormask.c', 'sp_quad_coverage.c', diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index e298ed37c3..8c84ddbe19 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -98,7 +98,6 @@ static void softpipe_destroy( struct pipe_context *pipe ) softpipe->quad.stencil_test->destroy( softpipe->quad.stencil_test ); softpipe->quad.occlusion->destroy( softpipe->quad.occlusion ); softpipe->quad.coverage->destroy( softpipe->quad.coverage ); - softpipe->quad.bufloop->destroy( softpipe->quad.bufloop ); softpipe->quad.blend->destroy( softpipe->quad.blend ); softpipe->quad.colormask->destroy( softpipe->quad.colormask ); softpipe->quad.output->destroy( softpipe->quad.output ); @@ -207,7 +206,6 @@ softpipe_create( struct pipe_screen *screen, softpipe->quad.stencil_test = sp_quad_stencil_test_stage(softpipe); softpipe->quad.occlusion = sp_quad_occlusion_stage(softpipe); softpipe->quad.coverage = sp_quad_coverage_stage(softpipe); - softpipe->quad.bufloop = sp_quad_bufloop_stage(softpipe); softpipe->quad.blend = sp_quad_blend_stage(softpipe); softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 9724d185d1..6f0b234d89 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -124,7 +124,6 @@ struct softpipe_context { struct quad_stage *depth_test; struct quad_stage *occlusion; struct quad_stage *coverage; - struct quad_stage *bufloop; struct quad_stage *blend; struct quad_stage *colormask; struct quad_stage *output; diff --git a/src/gallium/drivers/softpipe/sp_quad.h b/src/gallium/drivers/softpipe/sp_quad.h index f1e0281764..08513cb95f 100644 --- a/src/gallium/drivers/softpipe/sp_quad.h +++ b/src/gallium/drivers/softpipe/sp_quad.h @@ -58,7 +58,6 @@ struct quad_stage *sp_quad_stencil_test_stage( struct softpipe_context *softpipe struct quad_stage *sp_quad_depth_test_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_occlusion_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_coverage_stage( struct softpipe_context *softpipe ); -struct quad_stage *sp_quad_bufloop_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_blend_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_colormask_stage( struct softpipe_context *softpipe ); struct quad_stage *sp_quad_output_stage( struct softpipe_context *softpipe ); -- cgit v1.2.3 From 4c0f72432393e324d608474d24d20ebb465cb2ca Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 8 Apr 2008 11:26:52 +0200 Subject: Fixed user_buffer memory leak --- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 4 +++- src/gallium/winsys/dri/intel/ws_dri_bufmgr.c | 1 - 2 files changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 59729a0245..40902197e1 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -156,7 +156,9 @@ intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); driGenUserBuffer( iws->regionPool, - "pipe user buffer", &buffer->driBO, ptr, bytes ); + "pipe user buffer", &buffer->driBO, ptr, bytes ); + + buffer->base.refcount = 1; return &buffer->base; } diff --git a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c index eb5216cb37..bf387f672b 100644 --- a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c +++ b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c @@ -431,7 +431,6 @@ driBOUnReference(struct _DriBufferObject *buf) if (!buf) return; - _glthread_LOCK_MUTEX(buf->mutex); tmp = --buf->refCount; if (!tmp) { -- cgit v1.2.3 From caa5b1736dd4c4ba85966fa7710c52da406b1dce Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 8 Apr 2008 11:48:44 +0200 Subject: Added fence counter to fence manager --- src/gallium/winsys/dri/intel/ws_dri_fencemgr.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c b/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c index 8aaef1c620..1f893b47ce 100644 --- a/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c +++ b/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c @@ -23,6 +23,7 @@ struct _DriFenceMgr { _glthread_Mutex mutex; int refCount; drmMMListHead *heads; + int num_fences; }; struct _DriFenceObject { @@ -68,6 +69,7 @@ driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info) _glthread_LOCK_MUTEX(tmp->mutex); tmp->refCount = 1; tmp->info = *info; + tmp->num_fences = 0; tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads)); if (!tmp->heads) goto out_err; @@ -114,9 +116,11 @@ driFenceUnReferenceLocked(struct _DriFenceObject **pFence) DRMLISTDELINIT(&fence->head); if (fence->private) mgr->info.unreference(mgr, &fence->private); + --mgr->num_fences; fence->mgr = NULL; --mgr->refCount; free(fence); + } } @@ -297,6 +301,7 @@ struct _DriFenceObject DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]); fence->mgr = mgr; ++mgr->refCount; + ++mgr->num_fences; _glthread_UNLOCK_MUTEX(mgr->mutex); fence->fence_class = fence_class; fence->fence_type = fence_type; -- cgit v1.2.3 From f97c56e057f954cbc477df10ad767636f3fe83ce Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 8 Apr 2008 10:53:37 +0100 Subject: gallium: Fix up scons build. --- src/mesa/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index b816a21b38..a10c34c8c5 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -184,6 +184,7 @@ SHADER_SOURCES = [ 'shader/prog_parameter.c', 'shader/prog_print.c', 'shader/prog_statevars.c', + 'shader/prog_uniform.c', 'shader/programopt.c', 'shader/shader_api.c', ] -- cgit v1.2.3 From 4382b0c9cba3efa8a60252f6ddf2f0653352f7d8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 8 Apr 2008 20:42:06 +0900 Subject: gallium: Fix overzealous assert. --- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 24ba61a0b7..b1f7d93057 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -97,6 +97,7 @@ _fenced_buffer_add(struct fenced_buffer *fenced_buf) { struct fenced_buffer_list *fenced_list = fenced_buf->list; + assert(fenced_buf->base.base.refcount); assert(fenced_buf->fence); assert(!fenced_buf->head.prev); assert(!fenced_buf->head.next); @@ -128,7 +129,6 @@ _fenced_buffer_remove(struct fenced_buffer *fenced_buf) assert(fenced_buf->fence); - assert(winsys->fence_signalled(winsys, fenced_buf->fence, 0) == 0); winsys->fence_reference(winsys, &fenced_buf->fence, NULL); assert(fenced_buf->head.prev); @@ -174,6 +174,9 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, break; prev_fence = fenced_buf->fence; } + else { + assert(winsys->fence_signalled(winsys, fenced_buf->fence, 0) == 0); + } _fenced_buffer_remove(fenced_buf); @@ -199,6 +202,7 @@ fenced_buffer_destroy(struct pb_buffer *buf) prev = curr->prev; do { fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); + assert(winsys->fence_signalled(winsys, fenced_buf->fence, 0) == 0); _fenced_buffer_remove(fenced_buf); curr = prev; prev = curr->prev; -- cgit v1.2.3 From 2dd6022e65972b0706011734b68da4e9177b05c3 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 8 Apr 2008 15:01:16 +0200 Subject: i915: Fixed fence leak in intel_swapbuffers.c --- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index 1a42389169..923b542771 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -180,7 +180,6 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, if (intel->first_swap_fence) driFenceUnReference(&intel->first_swap_fence); intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); - driFenceReference(intel->first_swap_fence); } UNLOCK_HARDWARE(intel); -- cgit v1.2.3 From 1d43a8b25513866d67b5b16ae1945d6e9d95743d Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 8 Apr 2008 17:24:33 +0200 Subject: i915: Fixed fence related problems --- src/gallium/winsys/dri/intel/intel_batchbuffer.c | 4 +++- src/gallium/winsys/dri/intel/intel_winsys_i915.c | 18 ++++++++++-------- 2 files changed, 13 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c index 05223bf700..3a8ff5a4eb 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.c +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.c @@ -386,8 +386,10 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch) GLboolean was_locked = intel->locked; struct _DriFenceObject *fence; - if (used == 0) + if (used == 0) { + driFenceReference(batch->last_fence); return batch->last_fence; + } /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a * performance drain that we would like to avoid. diff --git a/src/gallium/winsys/dri/intel/intel_winsys_i915.c b/src/gallium/winsys/dri/intel/intel_winsys_i915.c index ae3e9654d3..a35825d36a 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_i915.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_i915.c @@ -113,10 +113,10 @@ static void intel_i915_batch_reloc( struct i915_winsys *sws, delta ); #else /* new */ intel_offset_relocation( intel->batch, - delta, - dri_bo( buf ), - flags, - mask ); + delta, + dri_bo( buf ), + flags, + mask ); #endif } @@ -132,9 +132,13 @@ static void intel_i915_batch_flush( struct i915_winsys *sws, struct pipe_fence_handle *pipe; } fu; + if (fence) + assert(!*fence); + fu.dri = intel_batchbuffer_flush( intel->batch ); - if (!fu.dri && fence) { + if (!fu.dri) { + assert(0); *fence = NULL; return; } @@ -143,11 +147,9 @@ static void intel_i915_batch_flush( struct i915_winsys *sws, if (fence) *fence = fu.pipe; else - iws->pws->fence_reference(iws->pws, &fu.dri, NULL); + driFenceUnReference(&fu.dri); } - -// if (0) intel_i915_batch_wait_idle( sws ); } -- cgit v1.2.3 From bd56b21adf5eafebe6df90eac12f0f9eaa9e225f Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 8 Apr 2008 18:40:36 +0200 Subject: i915: Added debug counter to bufmgr --- src/gallium/winsys/dri/intel/ws_dri_bufmgr.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c index bf387f672b..1bc1089352 100644 --- a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c +++ b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c @@ -53,6 +53,8 @@ _glthread_DECLARE_STATIC_MUTEX(bmMutex); _glthread_DECLARE_STATIC_COND(bmCond); static int kernelReaders = 0; +static int num_buffers = 0; +static int num_user_buffers = 0; static drmBO *drmBOListBuf(void *iterator) { @@ -441,6 +443,10 @@ driBOUnReference(struct _DriBufferObject *buf) else buf->pool->destroy(buf->pool, buf->private); } + if (buf->userBuffer) + num_user_buffers--; + else + num_buffers--; free(buf); } else _glthread_UNLOCK_MUTEX(buf->mutex); @@ -633,6 +639,7 @@ driGenBuffers(struct _DriBufferPool *pool, flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + ++num_buffers; assert(pool); @@ -664,7 +671,9 @@ driGenUserBuffer(struct _DriBufferPool *pool, { const unsigned alignment = 1, flags = 0, hint = 0; + --num_buffers; /* JB: is inced in GenBuffes */ driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); + ++num_user_buffers; (*buffers)->userBuffer = 1; (*buffers)->userData = ptr; -- cgit v1.2.3 From 28cf8c8fdcbd2817b93f27cad2bba3b4fb8aecf5 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 8 Apr 2008 21:07:14 -0600 Subject: gallium: keep track of num_vertex_attribs/buffers for shorter loops --- src/gallium/drivers/softpipe/sp_context.h | 2 ++ src/gallium/drivers/softpipe/sp_draw_arrays.c | 22 +++++++++------------- src/gallium/drivers/softpipe/sp_state_vertex.c | 2 ++ 3 files changed, 13 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 6f0b234d89..2442bd1eb0 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -83,6 +83,8 @@ struct softpipe_context { unsigned num_samplers; unsigned num_textures; + unsigned num_vertex_elements; + unsigned num_vertex_buffers; /* Counter for occlusion queries. Note this supports overlapping * queries. diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index ab54050d3f..61bcf51899 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -125,14 +125,12 @@ softpipe_draw_elements(struct pipe_context *pipe, /* * Map vertex buffers */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (sp->vertex_buffer[i].buffer) { - void *buf - = pipe->winsys->buffer_map(pipe->winsys, - sp->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_vertex_buffer(draw, i, buf); - } + for (i = 0; i < sp->num_vertex_buffers; i++) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + sp->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); } /* Map index buffer, if present */ if (indexBuffer) { @@ -153,11 +151,9 @@ softpipe_draw_elements(struct pipe_context *pipe, /* * unmap vertex/index buffers - will cause draw module to flush */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (sp->vertex_buffer[i].buffer) { - draw_set_mapped_vertex_buffer(draw, i, NULL); - pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); - } + for (i = 0; i < sp->num_vertex_buffers; i++) { + draw_set_mapped_vertex_buffer(draw, i, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); } if (indexBuffer) { draw_set_mapped_element_buffer(draw, 0, NULL); diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c index e0230e16a4..46b6991195 100644 --- a/src/gallium/drivers/softpipe/sp_state_vertex.c +++ b/src/gallium/drivers/softpipe/sp_state_vertex.c @@ -47,6 +47,7 @@ softpipe_set_vertex_elements(struct pipe_context *pipe, memcpy(softpipe->vertex_element, attribs, count * sizeof(struct pipe_vertex_element)); + softpipe->num_vertex_elements = count; softpipe->dirty |= SP_NEW_VERTEX; @@ -64,6 +65,7 @@ softpipe_set_vertex_buffers(struct pipe_context *pipe, assert(count <= PIPE_MAX_ATTRIBS); memcpy(softpipe->vertex_buffer, buffers, count * sizeof(buffers[0])); + softpipe->num_vertex_buffers = count; softpipe->dirty |= SP_NEW_VERTEX; -- cgit v1.2.3 From 1f888abf16ce4de00231505b8d1bc68426b04e8f Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 8 Apr 2008 21:14:58 -0600 Subject: i915: keep track of num_vertex_attribs/buffers for shorter loops --- src/gallium/drivers/i915simple/i915_context.c | 22 +++++++++------------- src/gallium/drivers/i915simple/i915_context.h | 2 ++ src/gallium/drivers/i915simple/i915_state.c | 3 +++ 3 files changed, 14 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index fee33d82de..58a5854f0d 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -65,14 +65,12 @@ i915_draw_elements( struct pipe_context *pipe, /* * Map vertex buffers */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (i915->vertex_buffer[i].buffer) { - void *buf - = pipe->winsys->buffer_map(pipe->winsys, - i915->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_vertex_buffer(draw, i, buf); - } + for (i = 0; i < i915->num_vertex_buffers; i++) { + void *buf + = pipe->winsys->buffer_map(pipe->winsys, + i915->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_vertex_buffer(draw, i, buf); } /* Map index buffer, if present */ if (indexBuffer) { @@ -96,11 +94,9 @@ i915_draw_elements( struct pipe_context *pipe, /* * unmap vertex/index buffers */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (i915->vertex_buffer[i].buffer) { - pipe->winsys->buffer_unmap(pipe->winsys, i915->vertex_buffer[i].buffer); - draw_set_mapped_vertex_buffer(draw, i, NULL); - } + for (i = 0; i < i915->num_vertex_buffers; i++) { + pipe->winsys->buffer_unmap(pipe->winsys, i915->vertex_buffer[i].buffer); + draw_set_mapped_vertex_buffer(draw, i, NULL); } if (indexBuffer) { pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 8e707ea574..38e6a34884 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -238,6 +238,8 @@ struct i915_context unsigned num_samplers; unsigned num_textures; + unsigned num_vertex_elements; + unsigned num_vertex_buffers; unsigned *batch_start; diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 4404bc4590..24d31ad740 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -694,6 +694,8 @@ static void i915_set_vertex_buffers(struct pipe_context *pipe, struct i915_context *i915 = i915_context(pipe); memcpy(i915->vertex_buffer, buffers, count * sizeof(buffers[0])); + i915->num_vertex_buffers = count; + /* pass-through to draw module */ draw_set_vertex_buffers(i915->draw, count, buffers); } @@ -703,6 +705,7 @@ static void i915_set_vertex_elements(struct pipe_context *pipe, const struct pipe_vertex_element *elements) { struct i915_context *i915 = i915_context(pipe); + i915->num_vertex_elements = count; /* pass-through to draw module */ draw_set_vertex_elements(i915->draw, count, elements); } -- cgit v1.2.3 From 7e57a9e8bba322b2ba8a02eec4b79c90e7052738 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 8 Apr 2008 21:18:21 -0600 Subject: cell: keep track of num_vertex_attribs/buffers for shorter loops --- src/gallium/drivers/cell/ppu/cell_context.h | 2 ++ src/gallium/drivers/cell/ppu/cell_draw_arrays.c | 22 +++++++++------------- src/gallium/drivers/cell/ppu/cell_state_vertex.c | 2 ++ 3 files changed, 13 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 7f656a9744..8df41c1d4c 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -104,7 +104,9 @@ struct cell_context uint num_textures; struct pipe_viewport_state viewport; struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + uint num_vertex_buffers; struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; + uint num_vertex_elements; ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS]; ubyte *zsbuf_map; diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c index b896252f81..36af5be5f0 100644 --- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c @@ -123,14 +123,12 @@ cell_draw_elements(struct pipe_context *pipe, /* * Map vertex buffers */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (sp->vertex_buffer[i].buffer) { - void *buf = pipe->winsys->buffer_map(pipe->winsys, - sp->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - cell_flush_buffer_range(sp, buf, sp->vertex_buffer[i].buffer->size); - draw_set_mapped_vertex_buffer(draw, i, buf); - } + for (i = 0; i < sp->num_vertex_buffers; i++) { + void *buf = pipe->winsys->buffer_map(pipe->winsys, + sp->vertex_buffer[i].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + cell_flush_buffer_range(sp, buf, sp->vertex_buffer[i].buffer->size); + draw_set_mapped_vertex_buffer(draw, i, buf); } /* Map index buffer, if present */ if (indexBuffer) { @@ -151,11 +149,9 @@ cell_draw_elements(struct pipe_context *pipe, /* * unmap vertex/index buffers - will cause draw module to flush */ - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (sp->vertex_buffer[i].buffer) { - draw_set_mapped_vertex_buffer(draw, i, NULL); - pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); - } + for (i = 0; i < sp->num_vertex_buffers; i++) { + draw_set_mapped_vertex_buffer(draw, i, NULL); + pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer); } if (indexBuffer) { draw_set_mapped_element_buffer(draw, 0, NULL); diff --git a/src/gallium/drivers/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c index 6c83b8dc72..114684c2a3 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_vertex.c +++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c @@ -45,6 +45,7 @@ cell_set_vertex_elements(struct pipe_context *pipe, assert(count <= PIPE_MAX_ATTRIBS); memcpy(cell->vertex_element, elements, count * sizeof(elements[0])); + cell->num_vertex_elements = count; cell->dirty |= CELL_NEW_VERTEX; @@ -62,6 +63,7 @@ cell_set_vertex_buffers(struct pipe_context *pipe, assert(count <= PIPE_MAX_ATTRIBS); memcpy(cell->vertex_buffer, buffers, count * sizeof(buffers[0])); + cell->num_vertex_buffers = count; cell->dirty |= CELL_NEW_VERTEX; -- cgit v1.2.3 From be37e8350f9b95fb50fbce3a2bd84098ad27e462 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 8 Apr 2008 21:38:44 -0600 Subject: gallium: remove obsolete/unused PIPE_ATTRIB_MAX --- src/gallium/include/pipe/p_state.h | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 2dc9a92186..79efc4f53d 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -55,7 +55,6 @@ extern "C" { #define PIPE_MAX_CLIP_PLANES 6 #define PIPE_MAX_CONSTANT 32 #define PIPE_MAX_ATTRIBS 32 -#define PIPE_ATTRIB_MAX 32 /* XXX obsolete - remove */ #define PIPE_MAX_COLOR_BUFS 8 #define PIPE_MAX_TEXTURE_LEVELS 16 #define PIPE_MAX_FEEDBACK_ATTRIBS 16 -- cgit v1.2.3 From da8934034b33adef5dc41395cfde4bdd26ba207c Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 8 Apr 2008 21:43:36 -0600 Subject: gallium: re-order, clean-up PIPE_MAX_* definitions --- src/gallium/include/pipe/p_state.h | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 79efc4f53d..912d84e7b9 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -51,15 +51,14 @@ extern "C" { /** * Implementation limits */ -#define PIPE_MAX_SAMPLERS 16 -#define PIPE_MAX_CLIP_PLANES 6 -#define PIPE_MAX_CONSTANT 32 -#define PIPE_MAX_ATTRIBS 32 -#define PIPE_MAX_COLOR_BUFS 8 -#define PIPE_MAX_TEXTURE_LEVELS 16 -#define PIPE_MAX_FEEDBACK_ATTRIBS 16 -#define PIPE_MAX_SHADER_INPUTS 16 -#define PIPE_MAX_SHADER_OUTPUTS 16 +#define PIPE_MAX_ATTRIBS 32 +#define PIPE_MAX_CLIP_PLANES 6 +#define PIPE_MAX_COLOR_BUFS 8 +#define PIPE_MAX_CONSTANT 32 +#define PIPE_MAX_SAMPLERS 16 +#define PIPE_MAX_SHADER_INPUTS 16 +#define PIPE_MAX_SHADER_OUTPUTS 16 +#define PIPE_MAX_TEXTURE_LEVELS 16 /* fwd decls */ -- cgit v1.2.3 From f7e475280a0b98484a8c1a98f18b2733532486b4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Apr 2008 14:17:48 -0600 Subject: gallium: fix bug in PIPE_BLENDFACTOR_INV_DST_ALPHA case --- src/gallium/drivers/softpipe/sp_quad_blend.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index 8be8025f40..802472df45 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -561,7 +561,7 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_INV_DST_ALPHA: { float inv_comp[4]; - VEC4_SUB(inv_comp, one, quadColor[3]); /* A */ + VEC4_SUB(inv_comp, one, dest[3]); /* A */ VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */ VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */ VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */ -- cgit v1.2.3 From a52faa9325db178601811f4bdad6d9747de5f238 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Apr 2008 16:09:46 -0600 Subject: gallium: remove unneeded st->haveFramebufferSurfaces field. --- src/gallium/winsys/xlib/xm_api.c | 3 --- src/mesa/state_tracker/st_cb_drawpixels.c | 6 +----- src/mesa/state_tracker/st_context.c | 2 -- src/mesa/state_tracker/st_context.h | 6 ------ 4 files changed, 1 insertion(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c index a82d3c990e..0c248344b1 100644 --- a/src/gallium/winsys/xlib/xm_api.c +++ b/src/gallium/winsys/xlib/xm_api.c @@ -814,9 +814,6 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) mesaCtx->Const.CheckArrayBounds = GL_TRUE; #endif - /* finish up xmesa context initializations */ - c->st->haveFramebufferSurfaces = GL_TRUE; - return c; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c57e05312a..b1affc4a91 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1017,11 +1017,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, srcy = ctx->DrawBuffer->Height - srcy - height; } - /* For some drivers (like Xlib) it's not possible to treat the - * front/back color buffers as surfaces (they're XImages and Pixmaps). - * So, this var tells us if we can use surface_copy here... - */ - if (st->haveFramebufferSurfaces && srcFormat == texFormat) { + if (srcFormat == texFormat) { /* copy source framebuffer surface into mipmap/texture */ pipe->surface_copy(pipe, FALSE, diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 7511c28074..154327239d 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -128,8 +128,6 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st->ctx->VertexProgram._MaintainTnlProgram = GL_TRUE; - st->haveFramebufferSurfaces = GL_TRUE; - st->pixel_xfer.cache = _mesa_new_program_cache(); /* GL limits and extensions */ diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index bcebbd4943..212687cf4a 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -111,12 +111,6 @@ struct st_context char vendor[100]; char renderer[100]; - /** Can we access the front/back color buffers as pipe_surfaces? - * We can't with the Xlib driver... - * This is a hack that should be fixed someday. - */ - GLboolean haveFramebufferSurfaces; - /* State to be validated: */ struct st_tracked_state **atoms; -- cgit v1.2.3 From 9c86c0e88b09370584f767747c52b7f352844ac5 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Apr 2008 16:26:56 -0600 Subject: gallium: refactor code, new flush_front_buffer() function --- src/mesa/state_tracker/st_cb_flush.c | 42 +++++++++++++++++++++--------------- 1 file changed, 25 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index c7efa40e38..1fc3a20d05 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -44,26 +44,12 @@ #include "pipe/p_winsys.h" -void st_flush( struct st_context *st, uint pipeFlushFlags, - struct pipe_fence_handle **fence ) -{ - FLUSH_VERTICES(st->ctx, 0); - - st_flush_bitmap_cache(st); - - st->pipe->flush( st->pipe, pipeFlushFlags, fence ); -} - - -static void st_gl_flush( struct st_context *st, uint pipeFlushFlags, - struct pipe_fence_handle **fence ) +static void +flush_front_buffer(struct st_context *st, uint pipeFlushFlags, + struct pipe_fence_handle **fence) { GLframebuffer *fb = st->ctx->DrawBuffer; - st_flush_bitmap_cache(st); - - FLUSH_VERTICES(st->ctx, 0); - if (!fb) return; @@ -105,6 +91,28 @@ static void st_gl_flush( struct st_context *st, uint pipeFlushFlags, } +void st_flush( struct st_context *st, uint pipeFlushFlags, + struct pipe_fence_handle **fence ) +{ + FLUSH_VERTICES(st->ctx, 0); + + st_flush_bitmap_cache(st); + + st->pipe->flush( st->pipe, pipeFlushFlags, fence ); +} + + +static void st_gl_flush( struct st_context *st, uint pipeFlushFlags, + struct pipe_fence_handle **fence ) +{ + st_flush_bitmap_cache(st); + + FLUSH_VERTICES(st->ctx, 0); + + flush_front_buffer(st, pipeFlushFlags, fence); +} + + /** * Called via ctx->Driver.Flush() */ -- cgit v1.2.3 From 574f964667c5ec35f4832c839a9dcc24f92e2aab Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Apr 2008 16:30:28 -0600 Subject: gallium: fold st_gl_flush() into st_glFlush() --- src/mesa/state_tracker/st_cb_flush.c | 17 +++++------------ 1 file changed, 5 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 1fc3a20d05..d60ad7b82a 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -102,23 +102,16 @@ void st_flush( struct st_context *st, uint pipeFlushFlags, } -static void st_gl_flush( struct st_context *st, uint pipeFlushFlags, - struct pipe_fence_handle **fence ) -{ - st_flush_bitmap_cache(st); - - FLUSH_VERTICES(st->ctx, 0); - - flush_front_buffer(st, pipeFlushFlags, fence); -} - - /** * Called via ctx->Driver.Flush() */ static void st_glFlush(GLcontext *ctx) { - st_gl_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + st_flush_bitmap_cache(ctx->st); + + FLUSH_VERTICES(ctx, 0); + + flush_front_buffer(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); } -- cgit v1.2.3 From aade2f41b0d5cf0fb44e094c0b10cfaf1f621aec Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Apr 2008 16:32:22 -0600 Subject: gallium: call_flush_front_buffer() from st_glFinish() --- src/mesa/state_tracker/st_cb_flush.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index d60ad7b82a..167c7ccbe2 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -132,6 +132,8 @@ void st_finish( struct st_context *st ) static void st_glFinish(GLcontext *ctx) { st_finish( ctx->st ); + + flush_front_buffer(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); } -- cgit v1.2.3 From 4ecbd5a70fba3e7d5f8b56ede34867ea5964afc6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Apr 2008 16:32:46 -0600 Subject: gallium: reorder funcs --- src/mesa/state_tracker/st_cb_flush.c | 23 ++++++++++++----------- 1 file changed, 12 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 167c7ccbe2..6bda877bb0 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -102,6 +102,18 @@ void st_flush( struct st_context *st, uint pipeFlushFlags, } +void st_finish( struct st_context *st ) +{ + struct pipe_fence_handle *fence = NULL; + + st_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence); + + st->pipe->winsys->fence_finish(st->pipe->winsys, fence, 0); + st->pipe->winsys->fence_reference(st->pipe->winsys, &fence, NULL); +} + + + /** * Called via ctx->Driver.Flush() */ @@ -115,17 +127,6 @@ static void st_glFlush(GLcontext *ctx) } -void st_finish( struct st_context *st ) -{ - struct pipe_fence_handle *fence = NULL; - - st_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence); - - st->pipe->winsys->fence_finish(st->pipe->winsys, fence, 0); - st->pipe->winsys->fence_reference(st->pipe->winsys, &fence, NULL); -} - - /** * Called via ctx->Driver.Finish() */ -- cgit v1.2.3 From 54d7c399a888283711bdc00f93cb54a3ce0b30eb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Apr 2008 16:59:14 -0600 Subject: gallium: more flush/finish changes New, separate is_front_buffer_dirty() function. --- src/mesa/state_tracker/st_cb_flush.c | 70 ++++++++++++++++++++---------------- 1 file changed, 40 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 6bda877bb0..1bcd1b4cd9 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -44,14 +44,13 @@ #include "pipe/p_winsys.h" -static void -flush_front_buffer(struct st_context *st, uint pipeFlushFlags, - struct pipe_fence_handle **fence) +static GLboolean +is_front_buffer_dirty(struct st_context *st) { GLframebuffer *fb = st->ctx->DrawBuffer; if (!fb) - return; + return GL_FALSE; /* XXX: temporary hack. This flag should only be set if we do any * rendering to the front buffer. @@ -68,26 +67,26 @@ flush_front_buffer(struct st_context *st, uint pipeFlushFlags, st->flags.frontbuffer_dirty = (fb->_ColorDrawBufferMask[0] & BUFFER_BIT_FRONT_LEFT); - if (st->flags.frontbuffer_dirty) { - struct st_renderbuffer *strb - = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); - struct pipe_surface *front_surf = strb->surface; - - /* If we aren't rendering to the frontbuffer, this is a noop. - * This should be uncontroversial for glFlush, though people may - * feel more strongly about glFinish. - * - * Additionally, need to make sure that the frontbuffer_dirty - * flag really gets set on frontbuffer rendering. - */ - st->pipe->flush( st->pipe, pipeFlushFlags, fence ); - - /* Hook for copying "fake" frontbuffer if necessary: - */ - st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf, - st->pipe->priv ); - st->flags.frontbuffer_dirty = 0; - } + return st->flags.frontbuffer_dirty; +} + + +/** + * Tell the winsys to display the front color buffer on-screen. + */ +static void +display_front_buffer(struct st_context *st) +{ + GLframebuffer *fb = st->ctx->DrawBuffer; + struct st_renderbuffer *strb + = st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + struct pipe_surface *front_surf = strb->surface; + + /* Hook for copying "fake" frontbuffer if necessary: + */ + st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf, + st->pipe->priv ); + st->flags.frontbuffer_dirty = 0; } @@ -102,6 +101,9 @@ void st_flush( struct st_context *st, uint pipeFlushFlags, } +/** + * Flush, and wait for completion. + */ void st_finish( struct st_context *st ) { struct pipe_fence_handle *fence = NULL; @@ -119,11 +121,15 @@ void st_finish( struct st_context *st ) */ static void st_glFlush(GLcontext *ctx) { - st_flush_bitmap_cache(ctx->st); - - FLUSH_VERTICES(ctx, 0); + struct st_context *st = ctx->st; - flush_front_buffer(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + if (is_front_buffer_dirty(st)) { + st_finish(st); + display_front_buffer(st); + } + else { + st_flush(st, PIPE_FLUSH_RENDER_CACHE, NULL); + } } @@ -132,9 +138,13 @@ static void st_glFlush(GLcontext *ctx) */ static void st_glFinish(GLcontext *ctx) { - st_finish( ctx->st ); + struct st_context *st = ctx->st; - flush_front_buffer(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + st_finish(st); + + if (is_front_buffer_dirty(st)) { + display_front_buffer(st); + } } -- cgit v1.2.3 From 311e40268414649f047ee177ba22a17a2d437843 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Apr 2008 18:39:51 -0600 Subject: gallium: more elaborate tracking of front color buffer state This fixes the case where the app calls SwapBuffers then calls glReadPixels to read the front color buffer. We now keep track of when the front buffer is a _logically_ copy of the back buffer (after SwapBuffers) and read from the back color buffer instead of the front. --- src/mesa/state_tracker/st_atom_framebuffer.c | 8 +++++++ src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_cb_flush.c | 27 ++++------------------- src/mesa/state_tracker/st_cb_readpixels.c | 33 ++++++++++++++++++++++++++-- src/mesa/state_tracker/st_cb_readpixels.h | 3 +++ src/mesa/state_tracker/st_context.h | 9 +++++--- src/mesa/state_tracker/st_framebuffer.c | 1 + 7 files changed, 54 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 02573af8f0..8a95096ec9 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -81,6 +81,14 @@ update_framebuffer_state( struct st_context *st ) } cso_set_framebuffer(st->cso_context, framebuffer); + + if (fb->_ColorDrawBufferMask[0] & BUFFER_BIT_FRONT_LEFT) { + if (st->frontbuffer_status == FRONT_STATUS_COPY_OF_BACK) { + /* XXX copy back buf to front? */ + } + /* we're assuming we'll really draw to the front buffer */ + st->frontbuffer_status = FRONT_STATUS_DIRTY; + } } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index b1affc4a91..5f8c90cbba 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -961,7 +961,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, } if (type == GL_COLOR) { - rbRead = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + rbRead = st_get_color_read_renderbuffer(ctx); color = NULL; stfp = combined_drawpix_fragment_program(ctx); stvp = st_make_passthrough_vertex_shader(ctx->st, GL_FALSE); diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 1bcd1b4cd9..1b3402cee2 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -44,30 +44,10 @@ #include "pipe/p_winsys.h" -static GLboolean +static INLINE GLboolean is_front_buffer_dirty(struct st_context *st) { - GLframebuffer *fb = st->ctx->DrawBuffer; - - if (!fb) - return GL_FALSE; - - /* XXX: temporary hack. This flag should only be set if we do any - * rendering to the front buffer. - * - * Further more, the scissor rectangle could be tracked to - * construct a dirty region of the front buffer, to avoid - * situations where it must be copied repeatedly. - * - * In the extreme case, some kind of timer could be set up to allow - * coalescing of multiple flushes to the frontbuffer, which can be - * quite a performance drain if there are a sufficient number of - * them. - */ - st->flags.frontbuffer_dirty - = (fb->_ColorDrawBufferMask[0] & BUFFER_BIT_FRONT_LEFT); - - return st->flags.frontbuffer_dirty; + return st->frontbuffer_status == FRONT_STATUS_DIRTY; } @@ -86,7 +66,8 @@ display_front_buffer(struct st_context *st) */ st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf, st->pipe->priv ); - st->flags.frontbuffer_dirty = 0; + + st->frontbuffer_status = FRONT_STATUS_UNDEFINED; } diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 2bcc8c99fb..29a06ffaa9 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -128,6 +128,34 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } +/** + * Return renderbuffer to use for reading color pixels for glRead/CopyPixel + * commands. + * Special care is needed for the front buffer. + */ +struct st_renderbuffer * +st_get_color_read_renderbuffer(GLcontext *ctx) +{ + struct gl_framebuffer *fb = ctx->ReadBuffer; + struct st_renderbuffer *strb = + st_renderbuffer(fb->_ColorReadBuffer); + struct st_renderbuffer *front = + st_renderbuffer(fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer); + + if (strb == front + && ctx->st->frontbuffer_status == FRONT_STATUS_COPY_OF_BACK) { + /* reading from front color buffer, which is a logical copy of the + * back color buffer. + */ + struct st_renderbuffer *back = + st_renderbuffer(fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer); + strb = back; + } + + return strb; +} + + /** * Do glReadPixels by getting rows from the framebuffer surface with * get_tile(). Convert to requested format/type with Mesa image routines. @@ -173,12 +201,13 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, strb = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer); } else { - strb = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + /* Read color buffer */ + strb = st_get_color_read_renderbuffer(ctx); } + if (!strb) return; - if (format == GL_RGBA && type == GL_FLOAT) { /* write tile(row) directly into user's buffer */ df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width, diff --git a/src/mesa/state_tracker/st_cb_readpixels.h b/src/mesa/state_tracker/st_cb_readpixels.h index 79acdad88e..9e151be51f 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.h +++ b/src/mesa/state_tracker/st_cb_readpixels.h @@ -29,6 +29,9 @@ #ifndef ST_CB_READPIXELS_H #define ST_CB_READPIXELS_H +extern struct st_renderbuffer * +st_get_color_read_renderbuffer(GLcontext *ctx); + extern void st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLenum type, diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 212687cf4a..d89e54c43c 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -45,6 +45,11 @@ struct blit_state; struct bitmap_cache; +#define FRONT_STATUS_UNDEFINED 0 +#define FRONT_STATUS_DIRTY 1 +#define FRONT_STATUS_COPY_OF_BACK 2 + + #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 #define ST_NEW_VERTEX_PROGRAM 0x4 @@ -104,9 +109,7 @@ struct st_context struct gl_fragment_program *fragment_program; } cb; - struct { - GLuint frontbuffer_dirty:1; - } flags; + GLuint frontbuffer_status; /**< one of FRONT_STATUS_ */ char vendor[100]; char renderer[100]; diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index ea09d9234c..47466c97d8 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -188,6 +188,7 @@ st_notify_swapbuffers(struct st_framebuffer *stfb) st_flush( ctx->st, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_SWAPBUFFERS, NULL ); + ctx->st->frontbuffer_status = FRONT_STATUS_COPY_OF_BACK; } } -- cgit v1.2.3 From 7a8ad75c89b45520043693a37d9f0c7e0b24fc5d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 10 Apr 2008 10:30:46 -0600 Subject: gallium: fix readback of z16 values --- src/mesa/state_tracker/st_cb_readpixels.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 29a06ffaa9..1fef55b844 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -269,7 +269,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } else if (strb->surface->format == PIPE_FORMAT_Z16_UNORM) { for (i = 0; i < height; i++) { - GLshort ztemp[MAX_WIDTH], j; + GLushort ztemp[MAX_WIDTH], j; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / 0xffff; pipe_get_tile_raw(pipe, strb->surface, x, y, width, 1, ztemp, 0); -- cgit v1.2.3 From c95dcc49629b72b95826e87e067d7a48753605fb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 8 Apr 2008 17:59:28 +0100 Subject: remove usage of vertex_header --- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 6 - src/gallium/auxiliary/draw/draw_vbuf.c | 20 - src/gallium/auxiliary/draw/draw_vertex.c | 5 - src/gallium/auxiliary/draw/draw_vertex.h | 2 - src/gallium/auxiliary/draw/draw_vf.c | 56 +- src/gallium/drivers/softpipe/Makefile | 1 + src/gallium/drivers/softpipe/SConscript | 1 + src/gallium/drivers/softpipe/sp_prim_setup.c | 1189 +-------------------- src/gallium/drivers/softpipe/sp_prim_setup.h | 6 + src/gallium/drivers/softpipe/sp_prim_vbuf.c | 223 ++-- src/gallium/drivers/softpipe/sp_setup.c | 1249 +++++++++++++++++++++++ src/gallium/drivers/softpipe/sp_setup.h | 53 + src/gallium/drivers/softpipe/sp_state_derived.c | 13 +- 13 files changed, 1480 insertions(+), 1344 deletions(-) create mode 100644 src/gallium/drivers/softpipe/sp_setup.c create mode 100644 src/gallium/drivers/softpipe/sp_setup.h (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 0806076956..e4e144ef19 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -212,12 +212,6 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, case EMIT_1F: feme->fetch[i].emit = emit_R32_FLOAT; break; - case EMIT_HEADER: - feme->fetch[i].ptr = (const ubyte *)&zero; - feme->fetch[i].pitch = 0; - feme->fetch[i].fetch = fetch_R32_FLOAT; - feme->fetch[i].emit = emit_R32_FLOAT; - break; case EMIT_1F_PSIZE: feme->fetch[i].ptr = (const ubyte *)&feme->draw->rasterizer->point_size; feme->fetch[i].pitch = 0; diff --git a/src/gallium/auxiliary/draw/draw_vbuf.c b/src/gallium/auxiliary/draw/draw_vbuf.c index f83b441e93..e3216ff711 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_vbuf.c @@ -126,13 +126,6 @@ dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) case EMIT_OMIT: debug_printf("EMIT_OMIT:"); break; - case EMIT_ALL: - assert(i == 0); - assert(j == 0); - debug_printf("EMIT_ALL:\t"); - for(k = 0; k < vinfo->size*4; ++k) - debug_printf("%02x ", *data++); - break; case EMIT_1F: debug_printf("EMIT_1F:\t"); debug_printf("%f ", *(float *)data); data += sizeof(float); @@ -217,19 +210,6 @@ emit_vertex( struct vbuf_stage *vbuf, case EMIT_OMIT: /* no-op */ break; - case EMIT_ALL: - /* just copy the whole vertex as-is to the vbuf */ - assert(i == 0); - assert(j == 0); - memcpy(vbuf->vertex_ptr, vertex, vinfo->size * 4); - vbuf->vertex_ptr += vinfo->size; - count += vinfo->size; - break; - case EMIT_HEADER: - memcpy(vbuf->vertex_ptr, vertex, sizeof(*vertex)); - *vbuf->vertex_ptr += sizeof(*vertex) / 4; - count += sizeof(*vertex) / 4; - break; case EMIT_1F: *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); count++; diff --git a/src/gallium/auxiliary/draw/draw_vertex.c b/src/gallium/auxiliary/draw/draw_vertex.c index 970adc95e7..168036eee8 100644 --- a/src/gallium/auxiliary/draw/draw_vertex.c +++ b/src/gallium/auxiliary/draw/draw_vertex.c @@ -52,9 +52,6 @@ draw_compute_vertex_size(struct vertex_info *vinfo) switch (vinfo->emit[i]) { case EMIT_OMIT: break; - case EMIT_HEADER: - vinfo->size += sizeof(struct vertex_header) / 4; - break; case EMIT_4UB: /* fall-through */ case EMIT_1F_PSIZE: @@ -71,8 +68,6 @@ draw_compute_vertex_size(struct vertex_info *vinfo) case EMIT_4F: vinfo->size += 4; break; - case EMIT_ALL: - /* fall-through */ default: assert(0); } diff --git a/src/gallium/auxiliary/draw/draw_vertex.h b/src/gallium/auxiliary/draw/draw_vertex.h index abd2017ed3..65818463ca 100644 --- a/src/gallium/auxiliary/draw/draw_vertex.h +++ b/src/gallium/auxiliary/draw/draw_vertex.h @@ -47,8 +47,6 @@ */ enum attrib_emit { EMIT_OMIT, /**< don't emit the attribute */ - EMIT_ALL, /**< emit whole post-xform vertex, w/ header */ - EMIT_HEADER, /**< emit vertex_header struct (XXX temp?) */ EMIT_1F, EMIT_1F_PSIZE, /**< insert constant point size */ EMIT_2F, diff --git a/src/gallium/auxiliary/draw/draw_vf.c b/src/gallium/auxiliary/draw/draw_vf.c index 7bb34ace7a..9d0154c50d 100644 --- a/src/gallium/auxiliary/draw/draw_vf.c +++ b/src/gallium/auxiliary/draw/draw_vf.c @@ -205,7 +205,7 @@ void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, const struct vertex_info *vinfo, float point_size ) { - unsigned i, j, k; + unsigned i, j; struct draw_vf_attr *a = vf->attr; struct draw_vf_attr_map attrs[PIPE_MAX_SHADER_INPUTS]; unsigned count = 0; /* for debug/sanity */ @@ -217,60 +217,6 @@ void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, case EMIT_OMIT: /* no-op */ break; - case EMIT_ALL: { - /* just copy the whole vertex as-is to the vbuf */ - unsigned s = vinfo->size; - assert(i == 0); - assert(j == 0); - /* copy the vertex header */ - /* XXX: we actually don't copy the header, just pad it */ - attrs[nr_attrs].attrib = 0; - attrs[nr_attrs].format = DRAW_EMIT_PAD; - attrs[nr_attrs].offset = offsetof(struct vertex_header, data); - s -= offsetof(struct vertex_header, data)/4; - count += offsetof(struct vertex_header, data)/4; - nr_attrs++; - /* copy the vertex data */ - for(k = 0; k < (s & ~0x3); k += 4) { - attrs[nr_attrs].attrib = k/4; - attrs[nr_attrs].format = DRAW_EMIT_4F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 4; - } - /* tail */ - /* XXX: actually, this shouldn't be needed */ - attrs[nr_attrs].attrib = k/4; - attrs[nr_attrs].offset = 0; - switch(s & 0x3) { - case 0: - break; - case 1: - attrs[nr_attrs].format = DRAW_EMIT_1F; - nr_attrs++; - count += 1; - break; - case 2: - attrs[nr_attrs].format = DRAW_EMIT_2F; - nr_attrs++; - count += 2; - break; - case 3: - attrs[nr_attrs].format = DRAW_EMIT_3F; - nr_attrs++; - count += 3; - break; - } - break; - } - case EMIT_HEADER: - /* XXX emit new DRAW_EMIT_HEADER attribute??? */ - attrs[nr_attrs].attrib = 0; - attrs[nr_attrs].format = DRAW_EMIT_PAD; - attrs[nr_attrs].offset = offsetof(struct vertex_header, data); - count += offsetof(struct vertex_header, data)/4; - nr_attrs++; - break; case EMIT_1F: attrs[nr_attrs].attrib = j; attrs[nr_attrs].format = DRAW_EMIT_1F; diff --git a/src/gallium/drivers/softpipe/Makefile b/src/gallium/drivers/softpipe/Makefile index 19dfd8c1a7..120bdfd9dd 100644 --- a/src/gallium/drivers/softpipe/Makefile +++ b/src/gallium/drivers/softpipe/Makefile @@ -27,6 +27,7 @@ C_SOURCES = \ sp_quad_stencil.c \ sp_quad_stipple.c \ sp_screen.c \ + sp_setup.c \ sp_state_blend.c \ sp_state_clip.c \ sp_state_derived.c \ diff --git a/src/gallium/drivers/softpipe/SConscript b/src/gallium/drivers/softpipe/SConscript index 06931fa8d8..c1f7daa8ab 100644 --- a/src/gallium/drivers/softpipe/SConscript +++ b/src/gallium/drivers/softpipe/SConscript @@ -14,6 +14,7 @@ softpipe = env.ConvenienceLibrary( 'sp_flush.c', 'sp_prim_setup.c', 'sp_prim_vbuf.c', + 'sp_setup.c', 'sp_quad_alpha_test.c', 'sp_quad_blend.c', 'sp_quad.c', diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index c7eb12b3bb..6fe463b74c 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -26,7 +26,9 @@ **************************************************************************/ /** - * \brief Primitive rasterization/rendering (points, lines, triangles) + * \brief A draw stage that drives our triangle setup routines from + * within the draw pipeline. One of two ways to drive setup, the + * other being in sp_prim_vbuf.c. * * \author Keith Whitwell * \author Brian Paul @@ -34,29 +36,12 @@ #include "sp_context.h" -#include "sp_headers.h" -#include "sp_quad.h" +#include "sp_setup.h" #include "sp_state.h" #include "sp_prim_setup.h" #include "draw/draw_private.h" #include "draw/draw_vertex.h" #include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" - -#define DEBUG_VERTS 0 -#define DEBUG_FRAGS 0 - -/** - * Triangle edge info - */ -struct edge { - float dx; /**< X(v1) - X(v0), used only during setup */ - float dy; /**< Y(v1) - Y(v0), used only during setup */ - float dxdy; /**< dx/dy */ - float sx, sy; /**< first sample point coord */ - int lines; /**< number of lines on this edge */ -}; - /** * Triangle setup info (derived from draw_stage). @@ -65,39 +50,7 @@ struct edge { struct setup_stage { struct draw_stage stage; /**< This must be first (base class) */ - struct softpipe_context *softpipe; - - /* Vertices are just an array of floats making up each attribute in - * turn. Currently fixed at 4 floats, but should change in time. - * Codegen will help cope with this. - */ - const struct vertex_header *vmax; - const struct vertex_header *vmid; - const struct vertex_header *vmin; - const struct vertex_header *vprovoke; - - struct edge ebot; - struct edge etop; - struct edge emaj; - - float oneoverarea; - - struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; - struct tgsi_interp_coef posCoef; /* For Z, W */ - struct quad_header quad; - - struct { - int left[2]; /**< [0] = row0, [1] = row1 */ - int right[2]; - int y; - unsigned y_flags; - unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ - } span; - -#if DEBUG_FRAGS - uint numFragsEmitted; /**< per primitive */ - uint numFragsWritten; /**< per primitive */ -#endif + struct setup_context *setup; }; @@ -111,1112 +64,50 @@ static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) } -/** - * Clip setup->quad against the scissor/surface bounds. - */ -static INLINE void -quad_clip(struct setup_stage *setup) -{ - const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; - - if (setup->quad.x0 >= maxx || - setup->quad.y0 >= maxy || - setup->quad.x0 + 1 < minx || - setup->quad.y0 + 1 < miny) { - /* totally clipped */ - setup->quad.mask = 0x0; - return; - } - if (setup->quad.x0 < minx) - setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - if (setup->quad.y0 < miny) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - if (setup->quad.x0 == maxx - 1) - setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - if (setup->quad.y0 == maxy - 1) - setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); -} - - -/** - * Emit a quad (pass to next stage) with clipping. - */ -static INLINE void -clip_emit_quad(struct setup_stage *setup) -{ - quad_clip(setup); - if (setup->quad.mask) { - struct softpipe_context *sp = setup->softpipe; - sp->quad.first->run(sp->quad.first, &setup->quad); - } -} - - -/** - * Emit a quad (pass to next stage). No clipping is done. - */ -static INLINE void -emit_quad( struct setup_stage *setup, int x, int y, unsigned mask ) -{ - struct softpipe_context *sp = setup->softpipe; - setup->quad.x0 = x; - setup->quad.y0 = y; - setup->quad.mask = mask; -#if DEBUG_FRAGS - if (mask & 1) setup->numFragsEmitted++; - if (mask & 2) setup->numFragsEmitted++; - if (mask & 4) setup->numFragsEmitted++; - if (mask & 8) setup->numFragsEmitted++; -#endif - sp->quad.first->run(sp->quad.first, &setup->quad); -#if DEBUG_FRAGS - mask = setup->quad.mask; - if (mask & 1) setup->numFragsWritten++; - if (mask & 2) setup->numFragsWritten++; - if (mask & 4) setup->numFragsWritten++; - if (mask & 8) setup->numFragsWritten++; -#endif -} - - -/** - * Given an X or Y coordinate, return the block/quad coordinate that it - * belongs to. - */ -static INLINE int block( int x ) -{ - return x & ~1; -} - - -/** - * Compute mask which indicates which pixels in the 2x2 quad are actually inside - * the triangle's bounds. - * - * this is pretty nasty... may need to rework flush_spans again to - * fix it, if possible. - */ -static unsigned calculate_mask( struct setup_stage *setup, int x ) -{ - unsigned mask = 0x0; - - if (x >= setup->span.left[0] && x < setup->span.right[0]) - mask |= MASK_TOP_LEFT; - - if (x >= setup->span.left[1] && x < setup->span.right[1]) - mask |= MASK_BOTTOM_LEFT; - - if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) - mask |= MASK_TOP_RIGHT; - - if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) - mask |= MASK_BOTTOM_RIGHT; - - return mask; -} - - -/** - * Render a horizontal span of quads - */ -static void flush_spans( struct setup_stage *setup ) -{ - int minleft, maxright; - int x; - - switch (setup->span.y_flags) { - case 0x3: - /* both odd and even lines written (both quad rows) */ - minleft = MIN2(setup->span.left[0], setup->span.left[1]); - maxright = MAX2(setup->span.right[0], setup->span.right[1]); - break; - - case 0x1: - /* only even line written (quad top row) */ - minleft = setup->span.left[0]; - maxright = setup->span.right[0]; - break; - - case 0x2: - /* only odd line written (quad bottom row) */ - minleft = setup->span.left[1]; - maxright = setup->span.right[1]; - break; - - default: - return; - } - - /* XXX this loop could be moved into the above switch cases and - * calculate_mask() could be simplified a bit... - */ - for (x = block(minleft); x <= block(maxright); x += 2) { - emit_quad( setup, x, setup->span.y, - calculate_mask( setup, x ) ); - } - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; -} - -#if DEBUG_VERTS -static void print_vertex(const struct setup_stage *setup, - const struct vertex_header *v) -{ - int i; - debug_printf("Vertex: (%p)\n", v); - for (i = 0; i < setup->quad.nr_attrs; i++) { - debug_printf(" %d: %f %f %f %f\n", i, - v->data[i][0], v->data[i][1], v->data[i][2], v->data[i][3]); - } -} -#endif - -static boolean setup_sort_vertices( struct setup_stage *setup, - const struct prim_header *prim ) -{ - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; - const struct vertex_header *v2 = prim->v[2]; - -#if DEBUG_VERTS - debug_printf("Triangle:\n"); - print_vertex(setup, v0); - print_vertex(setup, v1); - print_vertex(setup, v2); -#endif - - setup->vprovoke = v2; - - /* determine bottom to top order of vertices */ - { - float y0 = v0->data[0][1]; - float y1 = v1->data[0][1]; - float y2 = v2->data[0][1]; - if (y0 <= y1) { - if (y1 <= y2) { - /* y0<=y1<=y2 */ - setup->vmin = v0; - setup->vmid = v1; - setup->vmax = v2; - } - else if (y2 <= y0) { - /* y2<=y0<=y1 */ - setup->vmin = v2; - setup->vmid = v0; - setup->vmax = v1; - } - else { - /* y0<=y2<=y1 */ - setup->vmin = v0; - setup->vmid = v2; - setup->vmax = v1; - } - } - else { - if (y0 <= y2) { - /* y1<=y0<=y2 */ - setup->vmin = v1; - setup->vmid = v0; - setup->vmax = v2; - } - else if (y2 <= y1) { - /* y2<=y1<=y0 */ - setup->vmin = v2; - setup->vmid = v1; - setup->vmax = v0; - } - else { - /* y1<=y2<=y0 */ - setup->vmin = v1; - setup->vmid = v2; - setup->vmax = v0; - } - } - } - - setup->ebot.dx = setup->vmid->data[0][0] - setup->vmin->data[0][0]; - setup->ebot.dy = setup->vmid->data[0][1] - setup->vmin->data[0][1]; - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - setup->etop.dx = setup->vmax->data[0][0] - setup->vmid->data[0][0]; - setup->etop.dy = setup->vmax->data[0][1] - setup->vmid->data[0][1]; - - /* - * Compute triangle's area. Use 1/area to compute partial - * derivatives of attributes later. - * - * The area will be the same as prim->det, but the sign may be - * different depending on how the vertices get sorted above. - * - * To determine whether the primitive is front or back facing we - * use the prim->det value because its sign is correct. - */ - { - const float area = (setup->emaj.dx * setup->ebot.dy - - setup->ebot.dx * setup->emaj.dy); - - setup->oneoverarea = 1.0f / area; - /* - debug_printf("%s one-over-area %f area %f det %f\n", - __FUNCTION__, setup->oneoverarea, area, prim->det ); - */ - } - - /* We need to know if this is a front or back-facing triangle for: - * - the GLSL gl_FrontFacing fragment attribute (bool) - * - two-sided stencil test - */ - setup->quad.facing = (prim->det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); - - return TRUE; -} - - -/** - * Compute a0 for a constant-valued coefficient (GL_FLAT shading). - * The value value comes from vertex->data[slot][i]. - * The result will be put into setup->coef[slot].a0[i]. - * \param slot which attribute slot - * \param i which component of the slot (0..3) - */ -static void const_coeff( struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - assert(i <= 3); - - coef->dadx[i] = 0; - coef->dady[i] = 0; - /* need provoking vertex info! - */ - coef->a0[i] = setup->vprovoke->data[vertSlot][i]; -} - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a triangle. - */ -static void tri_linear_coeff( struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - float botda = setup->vmid->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; - float majda = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - float dadx = a * setup->oneoverarea; - float dady = b * setup->oneoverarea; - - assert(i <= 3); - - coef->dadx[i] = dadx; - coef->dady[i] = dady; - - /* calculate a0 as the value which would be sampled for the - * fragment at (0,0), taking into account that we want to sample at - * pixel centers, in other words (0.5, 0.5). - * - * this is neat but unfortunately not a good way to do things for - * triangles with very large values of dadx or dady as it will - * result in the subtraction and re-addition from a0 of a very - * large number, which means we'll end up loosing a lot of the - * fractional bits and precision from a0. the way to fix this is - * to define a0 as the sample at a pixel center somewhere near vmin - * instead - i'll switch to this later. - */ - coef->a0[i] = (setup->vmin->data[vertSlot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); - - /* - debug_printf("attr[%d].%c: %f dx:%f dy:%f\n", - slot, "xyzw"[i], - setup->coef[slot].a0[i], - setup->coef[slot].dadx[i], - setup->coef[slot].dady[i]); - */ -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a triangle. - * We basically multiply the vertex value by 1/w before computing - * the plane coefficients (a0, dadx, dady). - * Later, when we compute the value at a particular fragment position we'll - * divide the interpolated value by the interpolated W at that fragment. - */ -static void tri_persp_coeff( struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - /* premultiply by 1/w (v->data[0][3] is always W): - */ - float mina = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; - float mida = setup->vmid->data[vertSlot][i] * setup->vmid->data[0][3]; - float maxa = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; - float botda = mida - mina; - float majda = maxa - mina; - float a = setup->ebot.dy * majda - botda * setup->emaj.dy; - float b = setup->emaj.dx * botda - majda * setup->ebot.dx; - float dadx = a * setup->oneoverarea; - float dady = b * setup->oneoverarea; - - /* - debug_printf("tri persp %d,%d: %f %f %f\n", vertSlot, i, - setup->vmin->data[vertSlot][i], - setup->vmid->data[vertSlot][i], - setup->vmax->data[vertSlot][i] - ); - */ - assert(i <= 3); - - coef->dadx[i] = dadx; - coef->dady[i] = dady; - coef->a0[i] = (mina - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); -} - - -/** - * Special coefficient setup for gl_FragCoord. - * X and Y are trivial, though Y has to be inverted for OpenGL. - * Z and W are copied from posCoef which should have already been computed. - * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask. - */ static void -setup_fragcoord_coeff(struct setup_stage *setup, uint slot) -{ - /*X*/ - setup->coef[slot].a0[0] = 0; - setup->coef[slot].dadx[0] = 1.0; - setup->coef[slot].dady[0] = 0.0; - /*Y*/ - if (setup->softpipe->rasterizer->origin_lower_left) { - /* y=0=bottom */ - const int winHeight = setup->softpipe->framebuffer.height; - setup->coef[slot].a0[1] = (float) (winHeight - 1); - setup->coef[slot].dady[1] = -1.0; - } - else { - /* y=0=top */ - setup->coef[slot].a0[1] = 0.0; - setup->coef[slot].dady[1] = 1.0; - } - setup->coef[slot].dadx[1] = 0.0; - /*Z*/ - setup->coef[slot].a0[2] = setup->posCoef.a0[2]; - setup->coef[slot].dadx[2] = setup->posCoef.dadx[2]; - setup->coef[slot].dady[2] = setup->posCoef.dady[2]; - /*W*/ - setup->coef[slot].a0[3] = setup->posCoef.a0[3]; - setup->coef[slot].dadx[3] = setup->posCoef.dadx[3]; - setup->coef[slot].dady[3] = setup->posCoef.dady[3]; -} - - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. - */ -static void setup_tri_coefficients( struct setup_stage *setup ) -{ - struct softpipe_context *softpipe = setup->softpipe; - const struct sp_fragment_shader *spfs = softpipe->fs; - const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); - uint fragSlot; - - /* z and w are done by linear interpolation: - */ - tri_linear_coeff(setup, &setup->posCoef, 0, 2); - tri_linear_coeff(setup, &setup->posCoef, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { - const uint vertSlot = vinfo->src_index[fragSlot]; - uint j; - - switch (vinfo->interp_mode[fragSlot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_POS: - setup_fragcoord_coeff(setup, fragSlot); - break; - default: - assert(0); - } - - if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - } -} - - - -static void setup_tri_edges( struct setup_stage *setup ) -{ - float vmin_x = setup->vmin->data[0][0] + 0.5f; - float vmid_x = setup->vmid->data[0][0] + 0.5f; - - float vmin_y = setup->vmin->data[0][1] - 0.5f; - float vmid_y = setup->vmid->data[0][1] - 0.5f; - float vmax_y = setup->vmax->data[0][1] - 0.5f; - - setup->emaj.sy = CEILF(vmin_y); - setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); - setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; - setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; - - setup->etop.sy = CEILF(vmid_y); - setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); - setup->etop.dxdy = setup->etop.dx / setup->etop.dy; - setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; - - setup->ebot.sy = CEILF(vmin_y); - setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); - setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; - setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; -} - - -/** - * Render the upper or lower half of a triangle. - * Scissoring/cliprect is applied here too. - */ -static void subtriangle( struct setup_stage *setup, - struct edge *eleft, - struct edge *eright, - unsigned lines ) -{ - const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; - const int minx = (int) cliprect->minx; - const int maxx = (int) cliprect->maxx; - const int miny = (int) cliprect->miny; - const int maxy = (int) cliprect->maxy; - int y, start_y, finish_y; - int sy = (int)eleft->sy; - - assert((int)eleft->sy == (int) eright->sy); - - /* clip top/bottom */ - start_y = sy; - finish_y = sy + lines; - - if (start_y < miny) - start_y = miny; - - if (finish_y > maxy) - finish_y = maxy; - - start_y -= sy; - finish_y -= sy; - - /* - debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); - */ - - for (y = start_y; y < finish_y; y++) { - - /* avoid accumulating adds as floats don't have the precision to - * accurately iterate large triangle edges that way. luckily we - * can just multiply these days. - * - * this is all drowned out by the attribute interpolation anyway. - */ - int left = (int)(eleft->sx + y * eleft->dxdy); - int right = (int)(eright->sx + y * eright->dxdy); - - /* clip left/right */ - if (left < minx) - left = minx; - if (right > maxx) - right = maxx; - - if (left < right) { - int _y = sy + y; - if (block(_y) != setup->span.y) { - flush_spans(setup); - setup->span.y = block(_y); - } - - setup->span.left[_y&1] = left; - setup->span.right[_y&1] = right; - setup->span.y_flags |= 1<<(_y&1); - } - } - - - /* save the values so that emaj can be restarted: - */ - eleft->sx += lines * eleft->dxdy; - eright->sx += lines * eright->dxdy; - eleft->sy += lines; - eright->sy += lines; -} - - -/** - * Do setup for triangle rasterization, then render the triangle. - */ -static void setup_tri( struct draw_stage *stage, - struct prim_header *prim ) +do_tri(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - /* - debug_printf("%s\n", __FUNCTION__ ); - */ - -#if DEBUG_FRAGS - setup->numFragsEmitted = 0; - setup->numFragsWritten = 0; -#endif - - setup_sort_vertices( setup, prim ); - setup_tri_coefficients( setup ); - setup_tri_edges( setup ); - - setup->quad.prim = PRIM_TRI; - - setup->span.y = 0; - setup->span.y_flags = 0; - setup->span.right[0] = 0; - setup->span.right[1] = 0; - /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ - - /* init_constant_attribs( setup ); */ - - if (setup->oneoverarea < 0.0) { - /* emaj on left: - */ - subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); - subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); - } - else { - /* emaj on right: - */ - subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); - subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); - } - - flush_spans( setup ); - -#if DEBUG_FRAGS - printf("Tri: %u frags emitted, %u written\n", - setup->numFragsEmitted, - setup->numFragsWritten); -#endif -} - - - -/** - * Compute a0, dadx and dady for a linearly interpolated coefficient, - * for a line. - */ -static void -line_linear_coeff(struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - const float da = setup->vmax->data[vertSlot][i] - setup->vmin->data[vertSlot][i]; - const float dadx = da * setup->emaj.dx * setup->oneoverarea; - const float dady = da * setup->emaj.dy * setup->oneoverarea; - coef->dadx[i] = dadx; - coef->dady[i] = dady; - coef->a0[i] = (setup->vmin->data[vertSlot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); -} - - -/** - * Compute a0, dadx and dady for a perspective-corrected interpolant, - * for a line. - */ -static void -line_persp_coeff(struct setup_stage *setup, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - /* XXX double-check/verify this arithmetic */ - const float a0 = setup->vmin->data[vertSlot][i] * setup->vmin->data[0][3]; - const float a1 = setup->vmax->data[vertSlot][i] * setup->vmax->data[0][3]; - const float da = a1 - a0; - const float dadx = da * setup->emaj.dx * setup->oneoverarea; - const float dady = da * setup->emaj.dy * setup->oneoverarea; - coef->dadx[i] = dadx; - coef->dady[i] = dady; - coef->a0[i] = (setup->vmin->data[vertSlot][i] - - (dadx * (setup->vmin->data[0][0] - 0.5f) + - dady * (setup->vmin->data[0][1] - 0.5f))); -} - - -/** - * Compute the setup->coef[] array dadx, dady, a0 values. - * Must be called after setup->vmin,vmax are initialized. - */ -static INLINE void -setup_line_coefficients(struct setup_stage *setup, struct prim_header *prim) -{ - struct softpipe_context *softpipe = setup->softpipe; - const struct sp_fragment_shader *spfs = softpipe->fs; - const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); - uint fragSlot; - - /* use setup->vmin, vmax to point to vertices */ - setup->vprovoke = prim->v[1]; - setup->vmin = prim->v[0]; - setup->vmax = prim->v[1]; - - setup->emaj.dx = setup->vmax->data[0][0] - setup->vmin->data[0][0]; - setup->emaj.dy = setup->vmax->data[0][1] - setup->vmin->data[0][1]; - /* NOTE: this is not really 1/area */ - setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx + - setup->emaj.dy * setup->emaj.dy); - - /* z and w are done by linear interpolation: - */ - line_linear_coeff(setup, &setup->posCoef, 0, 2); - line_linear_coeff(setup, &setup->posCoef, 0, 3); - - /* setup interpolation for all the remaining attributes: - */ - for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { - const uint vertSlot = vinfo->src_index[fragSlot]; - uint j; - - switch (vinfo->interp_mode[fragSlot]) { - case INTERP_CONSTANT: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_POS: - setup_fragcoord_coeff(setup, fragSlot); - break; - default: - assert(0); - } - - if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - } + setup_tri( setup->setup, + prim->det, + prim->v[0]->data, + prim->v[1]->data, + prim->v[2]->data ); } - -/** - * Plot a pixel in a line segment. - */ -static INLINE void -plot(struct setup_stage *setup, int x, int y) -{ - const int iy = y & 1; - const int ix = x & 1; - const int quadX = x - ix; - const int quadY = y - iy; - const int mask = (1 << ix) << (2 * iy); - - if (quadX != setup->quad.x0 || - quadY != setup->quad.y0) - { - /* flush prev quad, start new quad */ - - if (setup->quad.x0 != -1) - clip_emit_quad(setup); - - setup->quad.x0 = quadX; - setup->quad.y0 = quadY; - setup->quad.mask = 0x0; - } - - setup->quad.mask |= mask; -} - - -/** - * Do setup for line rasterization, then render the line. - * Single-pixel width, no stipple, etc. We rely on the 'draw' module - * to handle stippling and wide lines. - */ static void -setup_line(struct draw_stage *stage, struct prim_header *prim) +do_line(struct draw_stage *stage, struct prim_header *prim) { - const struct vertex_header *v0 = prim->v[0]; - const struct vertex_header *v1 = prim->v[1]; struct setup_stage *setup = setup_stage( stage ); - int x0 = (int) v0->data[0][0]; - int x1 = (int) v1->data[0][0]; - int y0 = (int) v0->data[0][1]; - int y1 = (int) v1->data[0][1]; - int dx = x1 - x0; - int dy = y1 - y0; - int xstep, ystep; - if (dx == 0 && dy == 0) - return; - - setup_line_coefficients(setup, prim); - - if (dx < 0) { - dx = -dx; /* make positive */ - xstep = -1; - } - else { - xstep = 1; - } - - if (dy < 0) { - dy = -dy; /* make positive */ - ystep = -1; - } - else { - ystep = 1; - } - - assert(dx >= 0); - assert(dy >= 0); - - setup->quad.x0 = setup->quad.y0 = -1; - setup->quad.mask = 0x0; - setup->quad.prim = PRIM_LINE; - /* XXX temporary: set coverage to 1.0 so the line appears - * if AA mode happens to be enabled. - */ - setup->quad.coverage[0] = - setup->quad.coverage[1] = - setup->quad.coverage[2] = - setup->quad.coverage[3] = 1.0; - - if (dx > dy) { - /*** X-major line ***/ - int i; - const int errorInc = dy + dy; - int error = errorInc - dx; - const int errorDec = error - dx; - - for (i = 0; i < dx; i++) { - plot(setup, x0, y0); - - x0 += xstep; - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - y0 += ystep; - } - } - } - else { - /*** Y-major line ***/ - int i; - const int errorInc = dx + dx; - int error = errorInc - dy; - const int errorDec = error - dy; - - for (i = 0; i < dy; i++) { - plot(setup, x0, y0); - - y0 += ystep; - if (error < 0) { - error += errorInc; - } - else { - error += errorDec; - x0 += xstep; - } - } - } - - /* draw final quad */ - if (setup->quad.mask) { - clip_emit_quad(setup); - } + setup_line( setup->setup, + prim->v[0]->data, + prim->v[1]->data ); } - static void -point_persp_coeff(struct setup_stage *setup, - const struct vertex_header *vert, - struct tgsi_interp_coef *coef, - uint vertSlot, uint i) -{ - assert(i <= 3); - coef->dadx[i] = 0.0F; - coef->dady[i] = 0.0F; - coef->a0[i] = vert->data[vertSlot][i] * vert->data[0][3]; -} - - -/** - * Do setup for point rasterization, then render the point. - * Round or square points... - * XXX could optimize a lot for 1-pixel points. - */ -static void -setup_point(struct draw_stage *stage, struct prim_header *prim) +do_point(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - struct softpipe_context *softpipe = setup->softpipe; - const struct sp_fragment_shader *spfs = softpipe->fs; - const struct vertex_header *v0 = prim->v[0]; - const int sizeAttr = setup->softpipe->psize_slot; - const float size - = sizeAttr > 0 ? v0->data[sizeAttr][0] - : setup->softpipe->rasterizer->point_size; - const float halfSize = 0.5F * size; - const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth; - const float x = v0->data[0][0]; /* Note: data[0] is always position */ - const float y = v0->data[0][1]; - const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); - uint fragSlot; - - /* For points, all interpolants are constant-valued. - * However, for point sprites, we'll need to setup texcoords appropriately. - * XXX: which coefficients are the texcoords??? - * We may do point sprites as textured quads... - * - * KW: We don't know which coefficients are texcoords - ultimately - * the choice of what interpolation mode to use for each attribute - * should be determined by the fragment program, using - * per-attribute declaration statements that include interpolation - * mode as a parameter. So either the fragment program will have - * to be adjusted for pointsprite vs normal point behaviour, or - * otherwise a special interpolation mode will have to be defined - * which matches the required behaviour for point sprites. But - - * the latter is not a feature of normal hardware, and as such - * probably should be ruled out on that basis. - */ - setup->vprovoke = prim->v[0]; - - /* setup Z, W */ - const_coeff(setup, &setup->posCoef, 0, 2); - const_coeff(setup, &setup->posCoef, 0, 3); - - for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { - const uint vertSlot = vinfo->src_index[fragSlot]; - uint j; - - switch (vinfo->interp_mode[fragSlot]) { - case INTERP_CONSTANT: - /* fall-through */ - case INTERP_LINEAR: - for (j = 0; j < NUM_CHANNELS; j++) - const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_PERSPECTIVE: - for (j = 0; j < NUM_CHANNELS; j++) - point_persp_coeff(setup, setup->vprovoke, - &setup->coef[fragSlot], vertSlot, j); - break; - case INTERP_POS: - setup_fragcoord_coeff(setup, fragSlot); - break; - default: - assert(0); - } - - if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { - /* FOG.y = front/back facing XXX fix this */ - setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; - setup->coef[fragSlot].dadx[1] = 0.0; - setup->coef[fragSlot].dady[1] = 0.0; - } - } - - setup->quad.prim = PRIM_POINT; - - if (halfSize <= 0.5 && !round) { - /* special case for 1-pixel points */ - const int ix = ((int) x) & 1; - const int iy = ((int) y) & 1; - setup->quad.x0 = (int) x - ix; - setup->quad.y0 = (int) y - iy; - setup->quad.mask = (1 << ix) << (2 * iy); - clip_emit_quad(setup); - } - else { - if (round) { - /* rounded points */ - const int ixmin = block((int) (x - halfSize)); - const int ixmax = block((int) (x + halfSize)); - const int iymin = block((int) (y - halfSize)); - const int iymax = block((int) (y + halfSize)); - const float rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ - const float rmax = halfSize + 0.7071F; - const float rmin2 = MAX2(0.0F, rmin * rmin); - const float rmax2 = rmax * rmax; - const float cscale = 1.0F / (rmax2 - rmin2); - int ix, iy; - - for (iy = iymin; iy <= iymax; iy += 2) { - for (ix = ixmin; ix <= ixmax; ix += 2) { - float dx, dy, dist2, cover; - - setup->quad.mask = 0x0; - - dx = (ix + 0.5f) - x; - dy = (iy + 0.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_TOP_LEFT; - } - - dx = (ix + 1.5f) - x; - dy = (iy + 0.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_TOP_RIGHT; - } - dx = (ix + 0.5f) - x; - dy = (iy + 1.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_BOTTOM_LEFT; - } - - dx = (ix + 1.5f) - x; - dy = (iy + 1.5f) - y; - dist2 = dx * dx + dy * dy; - if (dist2 <= rmax2) { - cover = 1.0F - (dist2 - rmin2) * cscale; - setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f); - setup->quad.mask |= MASK_BOTTOM_RIGHT; - } - - if (setup->quad.mask) { - setup->quad.x0 = ix; - setup->quad.y0 = iy; - clip_emit_quad(setup); - } - } - } - } - else { - /* square points */ - const int xmin = (int) (x + 0.75 - halfSize); - const int ymin = (int) (y + 0.25 - halfSize); - const int xmax = xmin + (int) size; - const int ymax = ymin + (int) size; - /* XXX could apply scissor to xmin,ymin,xmax,ymax now */ - const int ixmin = block(xmin); - const int ixmax = block(xmax - 1); - const int iymin = block(ymin); - const int iymax = block(ymax - 1); - int ix, iy; - - /* - debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax); - */ - for (iy = iymin; iy <= iymax; iy += 2) { - uint rowMask = 0xf; - if (iy < ymin) { - /* above the top edge */ - rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); - } - if (iy + 1 >= ymax) { - /* below the bottom edge */ - rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); - } - - for (ix = ixmin; ix <= ixmax; ix += 2) { - uint mask = rowMask; - - if (ix < xmin) { - /* fragment is past left edge of point, turn off left bits */ - mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); - } - if (ix + 1 >= xmax) { - /* past the right edge */ - mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); - } - - setup->quad.mask = mask; - setup->quad.x0 = ix; - setup->quad.y0 = iy; - clip_emit_quad(setup); - } - } - } - } + setup_point( setup->setup, + prim->v[0]->data ); } + static void setup_begin( struct draw_stage *stage ) { struct setup_stage *setup = setup_stage(stage); - struct softpipe_context *sp = setup->softpipe; - const struct sp_fragment_shader *fs = setup->softpipe->fs; - uint i; - - if (sp->dirty) { - softpipe_update_derived(sp); - } - /* Mark surfaces as defined now */ - for (i = 0; i < sp->framebuffer.num_cbufs; i++){ - if (sp->framebuffer.cbufs[i]) { - sp->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED; - } - } - if (sp->framebuffer.zsbuf) { - sp->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_DEFINED; - } + setup_prepare( setup->setup ); - setup->quad.nr_attrs = fs->info.num_inputs; - - sp->quad.first->begin(sp->quad.first); - - stage->point = setup_point; - stage->line = setup_line; - stage->tri = setup_tri; + stage->point = do_point; + stage->line = do_line; + stage->tri = do_tri; } @@ -1269,19 +160,29 @@ static void render_destroy( struct draw_stage *stage ) */ struct draw_stage *sp_draw_render_stage( struct softpipe_context *softpipe ) { - struct setup_stage *setup = CALLOC_STRUCT(setup_stage); + struct setup_stage *sstage = CALLOC_STRUCT(setup_stage); - setup->softpipe = softpipe; - setup->stage.draw = softpipe->draw; - setup->stage.point = setup_first_point; - setup->stage.line = setup_first_line; - setup->stage.tri = setup_first_tri; - setup->stage.flush = setup_flush; - setup->stage.reset_stipple_counter = reset_stipple_counter; - setup->stage.destroy = render_destroy; + sstage->setup = setup_create_context(softpipe); + sstage->stage.draw = softpipe->draw; + sstage->stage.point = setup_first_point; + sstage->stage.line = setup_first_line; + sstage->stage.tri = setup_first_tri; + sstage->stage.flush = setup_flush; + sstage->stage.reset_stipple_counter = reset_stipple_counter; + sstage->stage.destroy = render_destroy; - setup->quad.coef = setup->coef; - setup->quad.posCoef = &setup->posCoef; + return (struct draw_stage *)sstage; +} - return &setup->stage; +struct setup_context * +sp_draw_setup_context( struct draw_stage *stage ) +{ + struct setup_stage *ssetup = setup_stage(stage); + return ssetup->setup; +} + +void +sp_draw_flush( struct draw_stage *stage ) +{ + stage->flush( stage, 0 ); } diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.h b/src/gallium/drivers/softpipe/sp_prim_setup.h index f3e8a79dd9..49bdd98ed8 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.h +++ b/src/gallium/drivers/softpipe/sp_prim_setup.h @@ -69,6 +69,12 @@ typedef void (*vbuf_draw_func)( struct pipe_context *pipe, extern struct draw_stage * sp_draw_render_stage( struct softpipe_context *softpipe ); +extern struct setup_context * +sp_draw_setup_context( struct draw_stage * ); + +extern void +sp_draw_flush( struct draw_stage * ); + extern struct draw_stage * sp_draw_vbuf_stage( struct draw_context *draw_context, diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index d940718ed2..025a0113bf 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ /** @@ -30,7 +30,7 @@ * softpipe rendering pipeline. * Probably not desired in general, but useful for testing/debuggin. * Enabled/Disabled with SP_VBUF env var. - * + * * Authors * Brian Paul */ @@ -39,14 +39,16 @@ #include "sp_context.h" #include "sp_state.h" #include "sp_prim_vbuf.h" +#include "sp_prim_setup.h" +#include "sp_setup.h" #include "draw/draw_context.h" -#include "draw/draw_private.h" #include "draw/draw_vbuf.h" #define SP_MAX_VBUF_INDEXES 1024 #define SP_MAX_VBUF_SIZE 4096 +typedef const float (*cptrf4)[4]; /** * Subclass of vbuf_render. @@ -69,7 +71,6 @@ softpipe_vbuf_render(struct vbuf_render *vbr) } - static const struct vertex_info * sp_vbuf_get_vertex_info(struct vbuf_render *vbr) { @@ -91,7 +92,7 @@ sp_vbuf_allocate_vertices(struct vbuf_render *vbr, static void -sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, +sp_vbuf_release_vertices(struct vbuf_render *vbr, void *vertices, unsigned vertex_size, unsigned vertices_used) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); @@ -114,7 +115,7 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) else { return FALSE; } - + } @@ -122,22 +123,19 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) * Recalculate prim's determinant. * XXX is this needed? */ -static void -calc_det(struct prim_header *header) +static float +calc_det( const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ) { - /* Window coords: */ - const float *v0 = header->v[0]->data[0]; - const float *v1 = header->v[1]->data[0]; - const float *v2 = header->v[2]->data[0]; - /* edge vectors e = v0 - v2, f = v1 - v2 */ - const float ex = v0[0] - v2[0]; - const float ey = v0[1] - v2[1]; - const float fx = v1[0] - v2[0]; - const float fy = v1[1] - v2[1]; - + const float ex = v0[0][0] - v2[0][0]; + const float ey = v0[0][1] - v2[0][1]; + const float fx = v1[0][0] - v2[0][0]; + const float fy = v1[0][1] - v2[0][1]; + /* det = cross(e,f).z */ - header->det = ex * fy - ey * fx; + return ex * fy - ey * fx; } @@ -147,48 +145,51 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; struct draw_stage *setup = softpipe->setup; - struct prim_header prim; unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); unsigned i, j; void *vertex_buffer = cvbr->vertex_buffer; - - prim.det = 0; - prim.reset_line_stipple = 0; - prim.edgeflags = 0; - prim.pad = 0; + cptrf4 v[3]; + struct setup_context *setup_ctx = sp_draw_setup_context(setup); switch (cvbr->prim) { case PIPE_PRIM_TRIANGLES: for (i = 0; i < nr_indices; i += 3) { - for (j = 0; j < 3; j++) - prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + - indices[i+j] * vertex_size); - - calc_det(&prim); - setup->tri( setup, &prim ); + for (j = 0; j < 3; j++) + v[j] = (cptrf4)((char *)vertex_buffer + + indices[i+j] * vertex_size); + + setup_tri( setup_ctx, + calc_det(v[0], v[1], v[2]), + v[0], + v[1], + v[2]); } break; case PIPE_PRIM_LINES: for (i = 0; i < nr_indices; i += 2) { - for (j = 0; j < 2; j++) - prim.v[j] = (struct vertex_header *)((char *)vertex_buffer + - indices[i+j] * vertex_size); - - setup->line( setup, &prim ); + for (j = 0; j < 2; j++) + v[j] = (cptrf4)((char *)vertex_buffer + + indices[i+j] * vertex_size); + + setup_line( setup_ctx, + v[0], + v[1] ); } break; case PIPE_PRIM_POINTS: for (i = 0; i < nr_indices; i++) { - prim.v[0] = (struct vertex_header *)((char *)vertex_buffer + - indices[i] * vertex_size); - setup->point( setup, &prim ); + v[0] = (cptrf4)((char *)vertex_buffer + + indices[i] * vertex_size); + + setup_point( setup_ctx, + v[0] ); } break; } - setup->flush( setup, 0 ); + sp_draw_flush( setup ); } @@ -202,111 +203,131 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; struct draw_stage *setup = softpipe->setup; - struct prim_header prim; const void *vertex_buffer = cvbr->vertex_buffer; const unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); unsigned i; - - prim.det = 0; - prim.reset_line_stipple = 0; - prim.edgeflags = 0; - prim.pad = 0; + struct setup_context *setup_ctx = sp_draw_setup_context(setup); + cptrf4 v[3]; #define VERTEX(I) \ - (struct vertex_header *) ((char *) vertex_buffer + (I) * vertex_size) + (cptrf4) ((char *) vertex_buffer + (I) * vertex_size) switch (cvbr->prim) { case PIPE_PRIM_POINTS: for (i = 0; i < nr; i++) { - prim.v[0] = VERTEX(i); - setup->point( setup, &prim ); + v[0] = VERTEX(i); + setup_point( setup_ctx, v[0] ); } break; case PIPE_PRIM_LINES: assert(nr % 2 == 0); for (i = 0; i < nr; i += 2) { - prim.v[0] = VERTEX(i); - prim.v[1] = VERTEX(i + 1); - setup->line( setup, &prim ); + v[0] = VERTEX(i); + v[1] = VERTEX(i + 1); + setup_line( setup_ctx, v[0], v[1] ); } break; case PIPE_PRIM_LINE_STRIP: for (i = 1; i < nr; i++) { - prim.v[0] = VERTEX(i - 1); - prim.v[1] = VERTEX(i); - setup->line( setup, &prim ); + v[0] = VERTEX(i - 1); + v[1] = VERTEX(i); + setup_line( setup_ctx, v[0], v[1] ); } break; case PIPE_PRIM_TRIANGLES: assert(nr % 3 == 0); for (i = 0; i < nr; i += 3) { - prim.v[0] = VERTEX(i + 0); - prim.v[1] = VERTEX(i + 1); - prim.v[2] = VERTEX(i + 2); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(i + 0); + v[1] = VERTEX(i + 1); + v[2] = VERTEX(i + 2); + setup_tri( setup_ctx, + calc_det(v[0], v[1], v[2]), + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_TRIANGLE_STRIP: assert(nr >= 3); for (i = 2; i < nr; i++) { - prim.v[0] = VERTEX(i - 2); - prim.v[1] = VERTEX(i - 1); - prim.v[2] = VERTEX(i); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(i - 2); + v[1] = VERTEX(i - 1); + v[2] = VERTEX(i); + setup_tri( setup_ctx, + calc_det(v[0], v[1], v[2]), + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_TRIANGLE_FAN: assert(nr >= 3); for (i = 2; i < nr; i++) { - prim.v[0] = VERTEX(0); - prim.v[1] = VERTEX(i - 1); - prim.v[2] = VERTEX(i); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(0); + v[1] = VERTEX(i - 1); + v[2] = VERTEX(i); + setup_tri( setup_ctx, + calc_det(v[0], v[1], v[2]), + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_QUADS: assert(nr % 4 == 0); for (i = 0; i < nr; i += 4) { - prim.v[0] = VERTEX(i + 0); - prim.v[1] = VERTEX(i + 1); - prim.v[2] = VERTEX(i + 2); - calc_det(&prim); - setup->tri( setup, &prim ); - - prim.v[0] = VERTEX(i + 0); - prim.v[1] = VERTEX(i + 2); - prim.v[2] = VERTEX(i + 3); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(i + 0); + v[1] = VERTEX(i + 1); + v[2] = VERTEX(i + 2); + setup_tri( setup_ctx, + calc_det(v[0], v[1], v[2]), + v[0], + v[1], + v[2] ); + + v[0] = VERTEX(i + 0); + v[1] = VERTEX(i + 2); + v[2] = VERTEX(i + 3); + setup_tri( setup_ctx, + calc_det(v[0], v[1], v[2]), + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_QUAD_STRIP: assert(nr >= 4); for (i = 2; i < nr; i += 2) { - prim.v[0] = VERTEX(i - 2); - prim.v[1] = VERTEX(i); - prim.v[2] = VERTEX(i + 1); - calc_det(&prim); - setup->tri( setup, &prim ); - - prim.v[0] = VERTEX(i - 2); - prim.v[1] = VERTEX(i + 1); - prim.v[2] = VERTEX(i - 1); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(i - 2); + v[1] = VERTEX(i); + v[2] = VERTEX(i + 1); + setup_tri( setup_ctx, + calc_det(v[0], v[1], v[2]), + v[0], + v[1], + v[2] ); + + v[0] = VERTEX(i - 2); + v[1] = VERTEX(i + 1); + v[2] = VERTEX(i - 1); + setup_tri( setup_ctx, + calc_det(v[0], v[1], v[2]), + v[0], + v[1], + v[2] ); } break; case PIPE_PRIM_POLYGON: /* draw as tri fan */ for (i = 2; i < nr; i++) { - prim.v[0] = VERTEX(0); - prim.v[1] = VERTEX(i - 1); - prim.v[2] = VERTEX(i); - calc_det(&prim); - setup->tri( setup, &prim ); + v[0] = VERTEX(0); + v[1] = VERTEX(i - 1); + v[2] = VERTEX(i); + setup_tri( setup_ctx, + calc_det(v[0], v[1], v[2]), + v[0], + v[1], + v[2] ); } break; default: diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c new file mode 100644 index 0000000000..48617a66ed --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -0,0 +1,1249 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Primitive rasterization/rendering (points, lines, triangles) + * + * \author Keith Whitwell + * \author Brian Paul + */ + +#include "sp_setup.h" + +#include "sp_context.h" +#include "sp_headers.h" +#include "sp_quad.h" +#include "sp_state.h" +#include "sp_prim_setup.h" +#include "draw/draw_private.h" +#include "draw/draw_vertex.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + +#define DEBUG_VERTS 0 +#define DEBUG_FRAGS 0 + +/** + * Triangle edge info + */ +struct edge { + float dx; /**< X(v1) - X(v0), used only during setup */ + float dy; /**< Y(v1) - Y(v0), used only during setup */ + float dxdy; /**< dx/dy */ + float sx, sy; /**< first sample point coord */ + int lines; /**< number of lines on this edge */ +}; + + +/** + * Triangle setup info (derived from draw_stage). + * Also used for line drawing (taking some liberties). + */ +struct setup_context { + struct softpipe_context *softpipe; + + /* Vertices are just an array of floats making up each attribute in + * turn. Currently fixed at 4 floats, but should change in time. + * Codegen will help cope with this. + */ + const float (*vmax)[4]; + const float (*vmid)[4]; + const float (*vmin)[4]; + const float (*vprovoke)[4]; + + struct edge ebot; + struct edge etop; + struct edge emaj; + + float oneoverarea; + + struct tgsi_interp_coef coef[PIPE_MAX_SHADER_INPUTS]; + struct tgsi_interp_coef posCoef; /* For Z, W */ + struct quad_header quad; + + struct { + int left[2]; /**< [0] = row0, [1] = row1 */ + int right[2]; + int y; + unsigned y_flags; + unsigned mask; /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */ + } span; + +#if DEBUG_FRAGS + uint numFragsEmitted; /**< per primitive */ + uint numFragsWritten; /**< per primitive */ +#endif +}; + + + + + +/** + * Recalculate prim's determinant. + * XXX is this needed? + */ +static INLINE float +calc_det( const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ) +{ + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0][0] - v2[0][0]; + const float ey = v0[0][1] - v2[0][1]; + const float fx = v1[0][0] - v2[0][0]; + const float fy = v1[0][1] - v2[0][1]; + + /* det = cross(e,f).z */ + return ex * fy - ey * fx; +} + +/** + * Clip setup->quad against the scissor/surface bounds. + */ +static INLINE void +quad_clip(struct setup_context *setup) +{ + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + + if (setup->quad.x0 >= maxx || + setup->quad.y0 >= maxy || + setup->quad.x0 + 1 < minx || + setup->quad.y0 + 1 < miny) { + /* totally clipped */ + setup->quad.mask = 0x0; + return; + } + if (setup->quad.x0 < minx) + setup->quad.mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + if (setup->quad.y0 < miny) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + if (setup->quad.x0 == maxx - 1) + setup->quad.mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + if (setup->quad.y0 == maxy - 1) + setup->quad.mask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); +} + + +/** + * Emit a quad (pass to next stage) with clipping. + */ +static INLINE void +clip_emit_quad(struct setup_context *setup) +{ + quad_clip(setup); + if (setup->quad.mask) { + struct softpipe_context *sp = setup->softpipe; + sp->quad.first->run(sp->quad.first, &setup->quad); + } +} + + +/** + * Emit a quad (pass to next stage). No clipping is done. + */ +static INLINE void +emit_quad( struct setup_context *setup, int x, int y, unsigned mask ) +{ + struct softpipe_context *sp = setup->softpipe; + setup->quad.x0 = x; + setup->quad.y0 = y; + setup->quad.mask = mask; +#if DEBUG_FRAGS + if (mask & 1) setup->numFragsEmitted++; + if (mask & 2) setup->numFragsEmitted++; + if (mask & 4) setup->numFragsEmitted++; + if (mask & 8) setup->numFragsEmitted++; +#endif + sp->quad.first->run(sp->quad.first, &setup->quad); +#if DEBUG_FRAGS + mask = setup->quad.mask; + if (mask & 1) setup->numFragsWritten++; + if (mask & 2) setup->numFragsWritten++; + if (mask & 4) setup->numFragsWritten++; + if (mask & 8) setup->numFragsWritten++; +#endif +} + + +/** + * Given an X or Y coordinate, return the block/quad coordinate that it + * belongs to. + */ +static INLINE int block( int x ) +{ + return x & ~1; +} + + +/** + * Compute mask which indicates which pixels in the 2x2 quad are actually inside + * the triangle's bounds. + * + * this is pretty nasty... may need to rework flush_spans again to + * fix it, if possible. + */ +static unsigned calculate_mask( struct setup_context *setup, int x ) +{ + unsigned mask = 0x0; + + if (x >= setup->span.left[0] && x < setup->span.right[0]) + mask |= MASK_TOP_LEFT; + + if (x >= setup->span.left[1] && x < setup->span.right[1]) + mask |= MASK_BOTTOM_LEFT; + + if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) + mask |= MASK_TOP_RIGHT; + + if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) + mask |= MASK_BOTTOM_RIGHT; + + return mask; +} + + +/** + * Render a horizontal span of quads + */ +static void flush_spans( struct setup_context *setup ) +{ + int minleft, maxright; + int x; + + switch (setup->span.y_flags) { + case 0x3: + /* both odd and even lines written (both quad rows) */ + minleft = MIN2(setup->span.left[0], setup->span.left[1]); + maxright = MAX2(setup->span.right[0], setup->span.right[1]); + break; + + case 0x1: + /* only even line written (quad top row) */ + minleft = setup->span.left[0]; + maxright = setup->span.right[0]; + break; + + case 0x2: + /* only odd line written (quad bottom row) */ + minleft = setup->span.left[1]; + maxright = setup->span.right[1]; + break; + + default: + return; + } + + /* XXX this loop could be moved into the above switch cases and + * calculate_mask() could be simplified a bit... + */ + for (x = block(minleft); x <= block(maxright); x += 2) { + emit_quad( setup, x, setup->span.y, + calculate_mask( setup, x ) ); + } + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; +} + +#if DEBUG_VERTS +static void print_vertex(const struct setup_context *setup, + const float (*v)[4]) +{ + int i; + debug_printf("Vertex: (%p)\n", v); + for (i = 0; i < setup->quad.nr_attrs; i++) { + debug_printf(" %d: %f %f %f %f\n", i, + v[i][0], v[i][1], v[i][2], v[i][3]); + } +} +#endif + +static boolean setup_sort_vertices( struct setup_context *setup, + float det, + const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ) +{ +#if DEBUG_VERTS + debug_printf("Triangle:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); + print_vertex(setup, v2); +#endif + + setup->vprovoke = v2; + + /* determine bottom to top order of vertices */ + { + float y0 = v0[0][1]; + float y1 = v1[0][1]; + float y2 = v2[0][1]; + if (y0 <= y1) { + if (y1 <= y2) { + /* y0<=y1<=y2 */ + setup->vmin = v0; + setup->vmid = v1; + setup->vmax = v2; + } + else if (y2 <= y0) { + /* y2<=y0<=y1 */ + setup->vmin = v2; + setup->vmid = v0; + setup->vmax = v1; + } + else { + /* y0<=y2<=y1 */ + setup->vmin = v0; + setup->vmid = v2; + setup->vmax = v1; + } + } + else { + if (y0 <= y2) { + /* y1<=y0<=y2 */ + setup->vmin = v1; + setup->vmid = v0; + setup->vmax = v2; + } + else if (y2 <= y1) { + /* y2<=y1<=y0 */ + setup->vmin = v2; + setup->vmid = v1; + setup->vmax = v0; + } + else { + /* y1<=y2<=y0 */ + setup->vmin = v1; + setup->vmid = v2; + setup->vmax = v0; + } + } + } + + setup->ebot.dx = setup->vmid[0][0] - setup->vmin[0][0]; + setup->ebot.dy = setup->vmid[0][1] - setup->vmin[0][1]; + setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0]; + setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1]; + setup->etop.dx = setup->vmax[0][0] - setup->vmid[0][0]; + setup->etop.dy = setup->vmax[0][1] - setup->vmid[0][1]; + + /* + * Compute triangle's area. Use 1/area to compute partial + * derivatives of attributes later. + * + * The area will be the same as prim->det, but the sign may be + * different depending on how the vertices get sorted above. + * + * To determine whether the primitive is front or back facing we + * use the prim->det value because its sign is correct. + */ + { + const float area = (setup->emaj.dx * setup->ebot.dy - + setup->ebot.dx * setup->emaj.dy); + + setup->oneoverarea = 1.0f / area; + /* + debug_printf("%s one-over-area %f area %f det %f\n", + __FUNCTION__, setup->oneoverarea, area, det ); + */ + } + + /* We need to know if this is a front or back-facing triangle for: + * - the GLSL gl_FrontFacing fragment attribute (bool) + * - two-sided stencil test + */ + setup->quad.facing = (det > 0.0) ^ (setup->softpipe->rasterizer->front_winding == PIPE_WINDING_CW); + + return TRUE; +} + + +/** + * Compute a0 for a constant-valued coefficient (GL_FLAT shading). + * The value value comes from vertex[slot][i]. + * The result will be put into setup->coef[slot].a0[i]. + * \param slot which attribute slot + * \param i which component of the slot (0..3) + */ +static void const_coeff( struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + assert(i <= 3); + + coef->dadx[i] = 0; + coef->dady[i] = 0; + + /* need provoking vertex info! + */ + coef->a0[i] = setup->vprovoke[vertSlot][i]; +} + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a triangle. + */ +static void tri_linear_coeff( struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + float botda = setup->vmid[vertSlot][i] - setup->vmin[vertSlot][i]; + float majda = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i]; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float dadx = a * setup->oneoverarea; + float dady = b * setup->oneoverarea; + + assert(i <= 3); + + coef->dadx[i] = dadx; + coef->dady[i] = dady; + + /* calculate a0 as the value which would be sampled for the + * fragment at (0,0), taking into account that we want to sample at + * pixel centers, in other words (0.5, 0.5). + * + * this is neat but unfortunately not a good way to do things for + * triangles with very large values of dadx or dady as it will + * result in the subtraction and re-addition from a0 of a very + * large number, which means we'll end up loosing a lot of the + * fractional bits and precision from a0. the way to fix this is + * to define a0 as the sample at a pixel center somewhere near vmin + * instead - i'll switch to this later. + */ + coef->a0[i] = (setup->vmin[vertSlot][i] - + (dadx * (setup->vmin[0][0] - 0.5f) + + dady * (setup->vmin[0][1] - 0.5f))); + + /* + debug_printf("attr[%d].%c: %f dx:%f dy:%f\n", + slot, "xyzw"[i], + setup->coef[slot].a0[i], + setup->coef[slot].dadx[i], + setup->coef[slot].dady[i]); + */ +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a triangle. + * We basically multiply the vertex value by 1/w before computing + * the plane coefficients (a0, dadx, dady). + * Later, when we compute the value at a particular fragment position we'll + * divide the interpolated value by the interpolated W at that fragment. + */ +static void tri_persp_coeff( struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + /* premultiply by 1/w (v[0][3] is always W): + */ + float mina = setup->vmin[vertSlot][i] * setup->vmin[0][3]; + float mida = setup->vmid[vertSlot][i] * setup->vmid[0][3]; + float maxa = setup->vmax[vertSlot][i] * setup->vmax[0][3]; + float botda = mida - mina; + float majda = maxa - mina; + float a = setup->ebot.dy * majda - botda * setup->emaj.dy; + float b = setup->emaj.dx * botda - majda * setup->ebot.dx; + float dadx = a * setup->oneoverarea; + float dady = b * setup->oneoverarea; + + /* + debug_printf("tri persp %d,%d: %f %f %f\n", vertSlot, i, + setup->vmin[vertSlot][i], + setup->vmid[vertSlot][i], + setup->vmax[vertSlot][i] + ); + */ + assert(i <= 3); + + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (mina - + (dadx * (setup->vmin[0][0] - 0.5f) + + dady * (setup->vmin[0][1] - 0.5f))); +} + + +/** + * Special coefficient setup for gl_FragCoord. + * X and Y are trivial, though Y has to be inverted for OpenGL. + * Z and W are copied from posCoef which should have already been computed. + * We could do a bit less work if we'd examine gl_FragCoord's swizzle mask. + */ +static void +setup_fragcoord_coeff(struct setup_context *setup, uint slot) +{ + /*X*/ + setup->coef[slot].a0[0] = 0; + setup->coef[slot].dadx[0] = 1.0; + setup->coef[slot].dady[0] = 0.0; + /*Y*/ + if (setup->softpipe->rasterizer->origin_lower_left) { + /* y=0=bottom */ + const int winHeight = setup->softpipe->framebuffer.height; + setup->coef[slot].a0[1] = (float) (winHeight - 1); + setup->coef[slot].dady[1] = -1.0; + } + else { + /* y=0=top */ + setup->coef[slot].a0[1] = 0.0; + setup->coef[slot].dady[1] = 1.0; + } + setup->coef[slot].dadx[1] = 0.0; + /*Z*/ + setup->coef[slot].a0[2] = setup->posCoef.a0[2]; + setup->coef[slot].dadx[2] = setup->posCoef.dadx[2]; + setup->coef[slot].dady[2] = setup->posCoef.dady[2]; + /*W*/ + setup->coef[slot].a0[3] = setup->posCoef.a0[3]; + setup->coef[slot].dadx[3] = setup->posCoef.dadx[3]; + setup->coef[slot].dady[3] = setup->posCoef.dady[3]; +} + + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmid,vmax,vprovoke are initialized. + */ +static void setup_tri_coefficients( struct setup_context *setup ) +{ + struct softpipe_context *softpipe = setup->softpipe; + const struct sp_fragment_shader *spfs = softpipe->fs; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); + uint fragSlot; + + /* z and w are done by linear interpolation: + */ + tri_linear_coeff(setup, &setup->posCoef, 0, 2); + tri_linear_coeff(setup, &setup->posCoef, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { + const uint vertSlot = vinfo->src_index[fragSlot]; + uint j; + + switch (vinfo->interp_mode[fragSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + tri_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + tri_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + setup_fragcoord_coeff(setup, fragSlot); + break; + default: + assert(0); + } + + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } + } +} + + + +static void setup_tri_edges( struct setup_context *setup ) +{ + float vmin_x = setup->vmin[0][0] + 0.5f; + float vmid_x = setup->vmid[0][0] + 0.5f; + + float vmin_y = setup->vmin[0][1] - 0.5f; + float vmid_y = setup->vmid[0][1] - 0.5f; + float vmax_y = setup->vmax[0][1] - 0.5f; + + setup->emaj.sy = CEILF(vmin_y); + setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); + setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; + setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; + + setup->etop.sy = CEILF(vmid_y); + setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); + setup->etop.dxdy = setup->etop.dx / setup->etop.dy; + setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; + + setup->ebot.sy = CEILF(vmin_y); + setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); + setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; + setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; +} + + +/** + * Render the upper or lower half of a triangle. + * Scissoring/cliprect is applied here too. + */ +static void subtriangle( struct setup_context *setup, + struct edge *eleft, + struct edge *eright, + unsigned lines ) +{ + const struct pipe_scissor_state *cliprect = &setup->softpipe->cliprect; + const int minx = (int) cliprect->minx; + const int maxx = (int) cliprect->maxx; + const int miny = (int) cliprect->miny; + const int maxy = (int) cliprect->maxy; + int y, start_y, finish_y; + int sy = (int)eleft->sy; + + assert((int)eleft->sy == (int) eright->sy); + + /* clip top/bottom */ + start_y = sy; + finish_y = sy + lines; + + if (start_y < miny) + start_y = miny; + + if (finish_y > maxy) + finish_y = maxy; + + start_y -= sy; + finish_y -= sy; + + /* + debug_printf("%s %d %d\n", __FUNCTION__, start_y, finish_y); + */ + + for (y = start_y; y < finish_y; y++) { + + /* avoid accumulating adds as floats don't have the precision to + * accurately iterate large triangle edges that way. luckily we + * can just multiply these days. + * + * this is all drowned out by the attribute interpolation anyway. + */ + int left = (int)(eleft->sx + y * eleft->dxdy); + int right = (int)(eright->sx + y * eright->dxdy); + + /* clip left/right */ + if (left < minx) + left = minx; + if (right > maxx) + right = maxx; + + if (left < right) { + int _y = sy + y; + if (block(_y) != setup->span.y) { + flush_spans(setup); + setup->span.y = block(_y); + } + + setup->span.left[_y&1] = left; + setup->span.right[_y&1] = right; + setup->span.y_flags |= 1<<(_y&1); + } + } + + + /* save the values so that emaj can be restarted: + */ + eleft->sx += lines * eleft->dxdy; + eright->sx += lines * eright->dxdy; + eleft->sy += lines; + eright->sy += lines; +} + + +/** + * Do setup for triangle rasterization, then render the triangle. + */ +void setup_tri( struct setup_context *setup, + float det, + const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ) +{ + /* + debug_printf("%s\n", __FUNCTION__ ); + */ + +#if DEBUG_FRAGS + setup->numFragsEmitted = 0; + setup->numFragsWritten = 0; +#endif + + setup_sort_vertices( setup, calc_det(v0, v1, v2), + v0, v1, v2 ); + setup_tri_coefficients( setup ); + setup_tri_edges( setup ); + + setup->quad.prim = PRIM_TRI; + + setup->span.y = 0; + setup->span.y_flags = 0; + setup->span.right[0] = 0; + setup->span.right[1] = 0; + /* setup->span.z_mode = tri_z_mode( setup->ctx ); */ + + /* init_constant_attribs( setup ); */ + + if (setup->oneoverarea < 0.0) { + /* emaj on left: + */ + subtriangle( setup, &setup->emaj, &setup->ebot, setup->ebot.lines ); + subtriangle( setup, &setup->emaj, &setup->etop, setup->etop.lines ); + } + else { + /* emaj on right: + */ + subtriangle( setup, &setup->ebot, &setup->emaj, setup->ebot.lines ); + subtriangle( setup, &setup->etop, &setup->emaj, setup->etop.lines ); + } + + flush_spans( setup ); + +#if DEBUG_FRAGS + printf("Tri: %u frags emitted, %u written\n", + setup->numFragsEmitted, + setup->numFragsWritten); +#endif +} + + + +/** + * Compute a0, dadx and dady for a linearly interpolated coefficient, + * for a line. + */ +static void +line_linear_coeff(struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + const float da = setup->vmax[vertSlot][i] - setup->vmin[vertSlot][i]; + const float dadx = da * setup->emaj.dx * setup->oneoverarea; + const float dady = da * setup->emaj.dy * setup->oneoverarea; + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (setup->vmin[vertSlot][i] - + (dadx * (setup->vmin[0][0] - 0.5f) + + dady * (setup->vmin[0][1] - 0.5f))); +} + + +/** + * Compute a0, dadx and dady for a perspective-corrected interpolant, + * for a line. + */ +static void +line_persp_coeff(struct setup_context *setup, + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + /* XXX double-check/verify this arithmetic */ + const float a0 = setup->vmin[vertSlot][i] * setup->vmin[0][3]; + const float a1 = setup->vmax[vertSlot][i] * setup->vmax[0][3]; + const float da = a1 - a0; + const float dadx = da * setup->emaj.dx * setup->oneoverarea; + const float dady = da * setup->emaj.dy * setup->oneoverarea; + coef->dadx[i] = dadx; + coef->dady[i] = dady; + coef->a0[i] = (setup->vmin[vertSlot][i] - + (dadx * (setup->vmin[0][0] - 0.5f) + + dady * (setup->vmin[0][1] - 0.5f))); +} + + +/** + * Compute the setup->coef[] array dadx, dady, a0 values. + * Must be called after setup->vmin,vmax are initialized. + */ +static INLINE void +setup_line_coefficients(struct setup_context *setup, + const float (*v0)[4], + const float (*v1)[4]) +{ + struct softpipe_context *softpipe = setup->softpipe; + const struct sp_fragment_shader *spfs = softpipe->fs; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); + uint fragSlot; + + /* use setup->vmin, vmax to point to vertices */ + setup->vprovoke = v1; + setup->vmin = v0; + setup->vmax = v1; + + setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0]; + setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1]; + /* NOTE: this is not really 1/area */ + setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx + + setup->emaj.dy * setup->emaj.dy); + + /* z and w are done by linear interpolation: + */ + line_linear_coeff(setup, &setup->posCoef, 0, 2); + line_linear_coeff(setup, &setup->posCoef, 0, 3); + + /* setup interpolation for all the remaining attributes: + */ + for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { + const uint vertSlot = vinfo->src_index[fragSlot]; + uint j; + + switch (vinfo->interp_mode[fragSlot]) { + case INTERP_CONSTANT: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + line_linear_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + line_persp_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + setup_fragcoord_coeff(setup, fragSlot); + break; + default: + assert(0); + } + + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } + } +} + + +/** + * Plot a pixel in a line segment. + */ +static INLINE void +plot(struct setup_context *setup, int x, int y) +{ + const int iy = y & 1; + const int ix = x & 1; + const int quadX = x - ix; + const int quadY = y - iy; + const int mask = (1 << ix) << (2 * iy); + + if (quadX != setup->quad.x0 || + quadY != setup->quad.y0) + { + /* flush prev quad, start new quad */ + + if (setup->quad.x0 != -1) + clip_emit_quad(setup); + + setup->quad.x0 = quadX; + setup->quad.y0 = quadY; + setup->quad.mask = 0x0; + } + + setup->quad.mask |= mask; +} + + +/** + * Do setup for line rasterization, then render the line. + * Single-pixel width, no stipple, etc. We rely on the 'draw' module + * to handle stippling and wide lines. + */ +void +setup_line(struct setup_context *setup, + const float (*v0)[4], + const float (*v1)[4]) +{ + int x0 = (int) v0[0][0]; + int x1 = (int) v1[0][0]; + int y0 = (int) v0[0][1]; + int y1 = (int) v1[0][1]; + int dx = x1 - x0; + int dy = y1 - y0; + int xstep, ystep; + + if (dx == 0 && dy == 0) + return; + + setup_line_coefficients(setup, v0, v1); + + if (dx < 0) { + dx = -dx; /* make positive */ + xstep = -1; + } + else { + xstep = 1; + } + + if (dy < 0) { + dy = -dy; /* make positive */ + ystep = -1; + } + else { + ystep = 1; + } + + assert(dx >= 0); + assert(dy >= 0); + + setup->quad.x0 = setup->quad.y0 = -1; + setup->quad.mask = 0x0; + setup->quad.prim = PRIM_LINE; + /* XXX temporary: set coverage to 1.0 so the line appears + * if AA mode happens to be enabled. + */ + setup->quad.coverage[0] = + setup->quad.coverage[1] = + setup->quad.coverage[2] = + setup->quad.coverage[3] = 1.0; + + if (dx > dy) { + /*** X-major line ***/ + int i; + const int errorInc = dy + dy; + int error = errorInc - dx; + const int errorDec = error - dx; + + for (i = 0; i < dx; i++) { + plot(setup, x0, y0); + + x0 += xstep; + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + y0 += ystep; + } + } + } + else { + /*** Y-major line ***/ + int i; + const int errorInc = dx + dx; + int error = errorInc - dy; + const int errorDec = error - dy; + + for (i = 0; i < dy; i++) { + plot(setup, x0, y0); + + y0 += ystep; + if (error < 0) { + error += errorInc; + } + else { + error += errorDec; + x0 += xstep; + } + } + } + + /* draw final quad */ + if (setup->quad.mask) { + clip_emit_quad(setup); + } +} + + +static void +point_persp_coeff(struct setup_context *setup, + const float (*vert)[4], + struct tgsi_interp_coef *coef, + uint vertSlot, uint i) +{ + assert(i <= 3); + coef->dadx[i] = 0.0F; + coef->dady[i] = 0.0F; + coef->a0[i] = vert[vertSlot][i] * vert[0][3]; +} + + +/** + * Do setup for point rasterization, then render the point. + * Round or square points... + * XXX could optimize a lot for 1-pixel points. + */ +void +setup_point( struct setup_context *setup, + const float (*v0)[4] ) +{ + struct softpipe_context *softpipe = setup->softpipe; + const struct sp_fragment_shader *spfs = softpipe->fs; + const int sizeAttr = setup->softpipe->psize_slot; + const float size + = sizeAttr > 0 ? v0[sizeAttr][0] + : setup->softpipe->rasterizer->point_size; + const float halfSize = 0.5F * size; + const boolean round = (boolean) setup->softpipe->rasterizer->point_smooth; + const float x = v0[0][0]; /* Note: data[0] is always position */ + const float y = v0[0][1]; + const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); + uint fragSlot; + + /* For points, all interpolants are constant-valued. + * However, for point sprites, we'll need to setup texcoords appropriately. + * XXX: which coefficients are the texcoords??? + * We may do point sprites as textured quads... + * + * KW: We don't know which coefficients are texcoords - ultimately + * the choice of what interpolation mode to use for each attribute + * should be determined by the fragment program, using + * per-attribute declaration statements that include interpolation + * mode as a parameter. So either the fragment program will have + * to be adjusted for pointsprite vs normal point behaviour, or + * otherwise a special interpolation mode will have to be defined + * which matches the required behaviour for point sprites. But - + * the latter is not a feature of normal hardware, and as such + * probably should be ruled out on that basis. + */ + setup->vprovoke = v0; + + /* setup Z, W */ + const_coeff(setup, &setup->posCoef, 0, 2); + const_coeff(setup, &setup->posCoef, 0, 3); + + for (fragSlot = 0; fragSlot < spfs->info.num_inputs; fragSlot++) { + const uint vertSlot = vinfo->src_index[fragSlot]; + uint j; + + switch (vinfo->interp_mode[fragSlot]) { + case INTERP_CONSTANT: + /* fall-through */ + case INTERP_LINEAR: + for (j = 0; j < NUM_CHANNELS; j++) + const_coeff(setup, &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_PERSPECTIVE: + for (j = 0; j < NUM_CHANNELS; j++) + point_persp_coeff(setup, setup->vprovoke, + &setup->coef[fragSlot], vertSlot, j); + break; + case INTERP_POS: + setup_fragcoord_coeff(setup, fragSlot); + break; + default: + assert(0); + } + + if (spfs->info.input_semantic_name[fragSlot] == TGSI_SEMANTIC_FOG) { + /* FOG.y = front/back facing XXX fix this */ + setup->coef[fragSlot].a0[1] = 1.0f - setup->quad.facing; + setup->coef[fragSlot].dadx[1] = 0.0; + setup->coef[fragSlot].dady[1] = 0.0; + } + } + + setup->quad.prim = PRIM_POINT; + + if (halfSize <= 0.5 && !round) { + /* special case for 1-pixel points */ + const int ix = ((int) x) & 1; + const int iy = ((int) y) & 1; + setup->quad.x0 = (int) x - ix; + setup->quad.y0 = (int) y - iy; + setup->quad.mask = (1 << ix) << (2 * iy); + clip_emit_quad(setup); + } + else { + if (round) { + /* rounded points */ + const int ixmin = block((int) (x - halfSize)); + const int ixmax = block((int) (x + halfSize)); + const int iymin = block((int) (y - halfSize)); + const int iymax = block((int) (y + halfSize)); + const float rmin = halfSize - 0.7071F; /* 0.7071 = sqrt(2)/2 */ + const float rmax = halfSize + 0.7071F; + const float rmin2 = MAX2(0.0F, rmin * rmin); + const float rmax2 = rmax * rmax; + const float cscale = 1.0F / (rmax2 - rmin2); + int ix, iy; + + for (iy = iymin; iy <= iymax; iy += 2) { + for (ix = ixmin; ix <= ixmax; ix += 2) { + float dx, dy, dist2, cover; + + setup->quad.mask = 0x0; + + dx = (ix + 0.5f) - x; + dy = (iy + 0.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_TOP_LEFT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_TOP_LEFT; + } + + dx = (ix + 1.5f) - x; + dy = (iy + 0.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_TOP_RIGHT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_TOP_RIGHT; + } + + dx = (ix + 0.5f) - x; + dy = (iy + 1.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_BOTTOM_LEFT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_BOTTOM_LEFT; + } + + dx = (ix + 1.5f) - x; + dy = (iy + 1.5f) - y; + dist2 = dx * dx + dy * dy; + if (dist2 <= rmax2) { + cover = 1.0F - (dist2 - rmin2) * cscale; + setup->quad.coverage[QUAD_BOTTOM_RIGHT] = MIN2(cover, 1.0f); + setup->quad.mask |= MASK_BOTTOM_RIGHT; + } + + if (setup->quad.mask) { + setup->quad.x0 = ix; + setup->quad.y0 = iy; + clip_emit_quad(setup); + } + } + } + } + else { + /* square points */ + const int xmin = (int) (x + 0.75 - halfSize); + const int ymin = (int) (y + 0.25 - halfSize); + const int xmax = xmin + (int) size; + const int ymax = ymin + (int) size; + /* XXX could apply scissor to xmin,ymin,xmax,ymax now */ + const int ixmin = block(xmin); + const int ixmax = block(xmax - 1); + const int iymin = block(ymin); + const int iymax = block(ymax - 1); + int ix, iy; + + /* + debug_printf("(%f, %f) -> X:%d..%d Y:%d..%d\n", x, y, xmin, xmax,ymin,ymax); + */ + for (iy = iymin; iy <= iymax; iy += 2) { + uint rowMask = 0xf; + if (iy < ymin) { + /* above the top edge */ + rowMask &= (MASK_BOTTOM_LEFT | MASK_BOTTOM_RIGHT); + } + if (iy + 1 >= ymax) { + /* below the bottom edge */ + rowMask &= (MASK_TOP_LEFT | MASK_TOP_RIGHT); + } + + for (ix = ixmin; ix <= ixmax; ix += 2) { + uint mask = rowMask; + + if (ix < xmin) { + /* fragment is past left edge of point, turn off left bits */ + mask &= (MASK_BOTTOM_RIGHT | MASK_TOP_RIGHT); + } + if (ix + 1 >= xmax) { + /* past the right edge */ + mask &= (MASK_BOTTOM_LEFT | MASK_TOP_LEFT); + } + + setup->quad.mask = mask; + setup->quad.x0 = ix; + setup->quad.y0 = iy; + clip_emit_quad(setup); + } + } + } + } +} + +void setup_prepare( struct setup_context *setup ) +{ + struct softpipe_context *sp = setup->softpipe; + unsigned i; + + if (sp->dirty) { + softpipe_update_derived(sp); + } + + /* Mark surfaces as defined now */ + for (i = 0; i < sp->framebuffer.num_cbufs; i++){ + if (sp->framebuffer.cbufs[i]) { + sp->framebuffer.cbufs[i]->status = PIPE_SURFACE_STATUS_DEFINED; + } + } + if (sp->framebuffer.zsbuf) { + sp->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_DEFINED; + } + + { + const struct sp_fragment_shader *fs = setup->softpipe->fs; + setup->quad.nr_attrs = fs->info.num_inputs; + sp->quad.first->begin(sp->quad.first); + } +} + + + +void setup_destroy_context( struct setup_context *setup ) +{ + FREE( setup ); +} + + +/** + * Create a new primitive setup/render stage. + */ +struct setup_context *setup_create_context( struct softpipe_context *softpipe ) +{ + struct setup_context *setup = CALLOC_STRUCT(setup_context); + + setup->softpipe = softpipe; + + setup->quad.coef = setup->coef; + setup->quad.posCoef = &setup->posCoef; + + return setup; +} diff --git a/src/gallium/drivers/softpipe/sp_setup.h b/src/gallium/drivers/softpipe/sp_setup.h new file mode 100644 index 0000000000..3133fc2a3d --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_setup.h @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ +#ifndef SP_SETUP_H +#define SP_SETUP_H + +struct setup_context; +struct softpipe_context; + +void setup_tri( struct setup_context *setup, + float det, + const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ); + +void +setup_line(struct setup_context *setup, + const float (*v0)[4], + const float (*v1)[4]); + +void +setup_point( struct setup_context *setup, + const float (*v0)[4] ); + + +struct setup_context *setup_create_context( struct softpipe_context *softpipe ); +void setup_prepare( struct setup_context *setup ); +void setup_destroy_context( struct setup_context *setup ); + +#endif diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index 14abb20eeb..f10a1fa471 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -71,25 +71,16 @@ softpipe_get_vertex_info(struct softpipe_context *softpipe) * simply emit the whole post-xform vertex as-is: */ struct vertex_info *vinfo_vbuf = &softpipe->vertex_info_vbuf; -#if 0 - vinfo_vbuf->num_attribs = 0; - /* special-case to allow memcpy of whole vertex */ - draw_emit_vertex_attr(vinfo_vbuf, EMIT_ALL, INTERP_NONE, 0); - /* size in dwords or floats */ - vinfo_vbuf->size = 4 * draw_num_vs_outputs(softpipe->draw) - + sizeof(struct vertex_header) / 4; -#else - /* for pass-through mode, we need a more explicit list of attribs */ const uint num = draw_num_vs_outputs(softpipe->draw); uint i; + /* No longer any need to try and emit draw vertex_header info. + */ vinfo_vbuf->num_attribs = 0; - draw_emit_vertex_attr(vinfo_vbuf, EMIT_HEADER, INTERP_NONE, 0); for (i = 0; i < num; i++) { draw_emit_vertex_attr(vinfo_vbuf, EMIT_4F, INTERP_PERSPECTIVE, i); } draw_compute_vertex_size(vinfo_vbuf); -#endif } /* -- cgit v1.2.3 From bc56e87ce180ddca63bcb9774366fc380214a2ef Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 10 Apr 2008 22:57:21 +0900 Subject: gallium: Attribute realloc leaks to the first malloc call. --- src/gallium/auxiliary/util/p_debug_mem.c | 71 +++++++++++++++++++++++++----- src/gallium/auxiliary/util/u_double_list.h | 8 ++++ 2 files changed, 68 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c index 97ec9c5b39..aba69c0294 100644 --- a/src/gallium/auxiliary/util/p_debug_mem.c +++ b/src/gallium/auxiliary/util/p_debug_mem.c @@ -73,6 +73,25 @@ static struct list_head list = { &list, &list }; static unsigned long last_no = 0; +static INLINE struct debug_memory_header * +header_from_data(void *data) +{ + if(data) + return (struct debug_memory_header *)((char *)data - sizeof(struct debug_memory_header)); + else + return NULL; +} + +static INLINE void * +data_from_header(struct debug_memory_header *hdr) +{ + if(hdr) + return (void *)((char *)hdr + sizeof(struct debug_memory_header)); + else + return NULL; +} + + void * debug_malloc(const char *file, unsigned line, const char *function, size_t size) @@ -92,7 +111,7 @@ debug_malloc(const char *file, unsigned line, const char *function, LIST_ADDTAIL(&hdr->head, &list); - return (void *)((char *)hdr + sizeof(*hdr)); + return data_from_header(hdr); } void @@ -104,7 +123,7 @@ debug_free(const char *file, unsigned line, const char *function, if(!ptr) return; - hdr = (struct debug_memory_header *)((char *)ptr - sizeof(*hdr)); + hdr = header_from_data(ptr); if(hdr->magic != DEBUG_MEMORY_MAGIC) { debug_printf("%s:%u:%s: freeing bad or corrupted memory %p\n", file, line, function, @@ -133,16 +152,46 @@ void * debug_realloc(const char *file, unsigned line, const char *function, void *old_ptr, size_t old_size, size_t new_size ) { - void *new_ptr = NULL; - - if (new_size != 0) { - new_ptr = debug_malloc( file, line, function, new_size ); - - if( new_ptr && old_ptr ) - memcpy( new_ptr, old_ptr, old_size ); + struct debug_memory_header *old_hdr, *new_hdr; + void *new_ptr; + + if(!old_ptr) + return debug_malloc( file, line, function, new_size ); + + if(!new_size) { + debug_free( file, line, function, old_ptr ); + return NULL; } + + old_hdr = header_from_data(old_ptr); + if(old_hdr->magic != DEBUG_MEMORY_MAGIC) { + debug_printf("%s:%u:%s: reallocating bad or corrupted memory %p\n", + file, line, function, + old_ptr); + debug_assert(0); + return NULL; + } + + /* alloc new */ + new_hdr = real_malloc(sizeof(*new_hdr) + new_size); + if(!new_hdr) + return NULL; + new_hdr->no = old_hdr->no; + new_hdr->file = old_hdr->file; + new_hdr->line = old_hdr->line; + new_hdr->function = old_hdr->function; + new_hdr->size = new_size; + new_hdr->magic = DEBUG_MEMORY_MAGIC; + LIST_REPLACE(&old_hdr->head, &new_hdr->head); + + /* copy data */ + new_ptr = data_from_header(new_hdr); + memcpy( new_ptr, old_ptr, old_size < new_size ? old_size : new_size ); + + /* free old */ + old_hdr->magic = 0; + real_free(old_hdr); - debug_free( file, line, function, old_ptr ); return new_ptr; } @@ -162,7 +211,7 @@ debug_memory_end(unsigned long start_no) struct debug_memory_header *hdr; void *ptr; hdr = LIST_ENTRY(struct debug_memory_header, entry, head); - ptr = (void *)((char *)hdr + sizeof(*hdr)); + ptr = data_from_header(hdr); if(start_no <= hdr->no && hdr->no < last_no || last_no < start_no && (hdr->no < last_no || start_no <= hdr->no)) debug_printf("%s:%u:%s: %u bytes at %p not freed\n", diff --git a/src/gallium/auxiliary/util/u_double_list.h b/src/gallium/auxiliary/util/u_double_list.h index 8cb10c9820..d108d92e52 100644 --- a/src/gallium/auxiliary/util/u_double_list.h +++ b/src/gallium/auxiliary/util/u_double_list.h @@ -70,6 +70,14 @@ struct list_head (__list)->prev = (__item); \ } while(0) +#define LIST_REPLACE(__from, __to) \ + do { \ + (__to)->prev = (__from)->prev; \ + (__to)->next = (__from)->next; \ + (__from)->next->prev = (__to); \ + (__from)->prev->next = (__to); \ + } while (0) + #define LIST_DEL(__item) \ do { \ (__item)->prev->next = (__item)->next; \ -- cgit v1.2.3 From 87b0b8e7bc3b4ce4be90a271b9fced1c97df6c95 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Apr 2008 16:53:41 -0600 Subject: gallium: flags param to surface_alloc_storage() is unused and deprecated --- src/gallium/include/pipe/p_winsys.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h index 8569cdcf12..bf5aad7351 100644 --- a/src/gallium/include/pipe/p_winsys.h +++ b/src/gallium/include/pipe/p_winsys.h @@ -80,6 +80,7 @@ struct pipe_winsys /** * Allocate storage for a pipe_surface. * Returns 0 if succeeds. + * XXX: flags is unused and will be removed someday. */ int (*surface_alloc_storage)(struct pipe_winsys *ws, struct pipe_surface *surf, -- cgit v1.2.3 From 03a3373bdf929b6b7c43600c9eebb0c4ee3ff38f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 11 Apr 2008 12:11:14 +0900 Subject: gallium: Thread condition variables. --- src/gallium/include/pipe/p_thread.h | 34 ++++++++++++++++++++++++++++++++++ 1 file changed, 34 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h index 4325abc951..af9150aafb 100644 --- a/src/gallium/include/pipe/p_thread.h +++ b/src/gallium/include/pipe/p_thread.h @@ -114,6 +114,26 @@ typedef pthread_mutex_t _glthread_Mutex; #define _glthread_UNLOCK_MUTEX(name) \ (void) pthread_mutex_unlock(&(name)) +typedef pthread_cond_t _glthread_Cond; + +#define _glthread_DECLARE_STATIC_COND(name) \ + static _glthread_Cond name = PTHREAD_COND_INITIALIZER + +#define _glthread_INIT_COND(cond) \ + pthread_cond_init(&(cond), NULL) + +#define _glthread_DESTROY_COND(name) \ + pthread_cond_destroy(&(name)) + +#define _glthread_COND_WAIT(cond, mutex) \ + pthread_cond_wait(&(cond), &(mutex)) + +#define _glthread_COND_SIGNAL(cond) \ + pthread_cond_signal(&(cond)) + +#define _glthread_COND_BROADCAST(cond) \ + pthread_cond_broadcast(&(cond)) + #endif /* PTHREADS */ @@ -273,6 +293,20 @@ typedef unsigned _glthread_Mutex; #define _glthread_UNLOCK_MUTEX(name) (void) name +typedef unsigned _glthread_Cond; + +#define _glthread_DECLARE_STATIC_COND(name) static _glthread_Cond name = 0 + +#define _glthread_INIT_COND(name) (void) name + +#define _glthread_DESTROY_COND(name) (void) name + +#define _glthread_COND_WAIT(name, mutex) (void) name + +#define _glthread_COND_SIGNAL(name) (void) name + +#define _glthread_COND_BROADCAST(name) (void) name + #endif /* THREADS */ -- cgit v1.2.3 From 2483062f10e93fbbc5e3f629627b9e8addcc3f84 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 7 Apr 2008 11:44:34 +0200 Subject: tgsi: Dump semantics before interpolator. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 7d292778ad..ad9770e474 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -746,17 +746,19 @@ dump_declaration_short( } } - if( decl->Declaration.Interpolate ) { + if (decl->Declaration.Semantic) { TXT( ", " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); + if (decl->Semantic.SemanticIndex != 0) { + CHR( '[' ); + UID( decl->Semantic.SemanticIndex ); + CHR( ']' ); + } } - if( decl->Declaration.Semantic ) { + if (decl->Declaration.Interpolate) { TXT( ", " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); - CHR( '[' ); - UID( decl->Semantic.SemanticIndex ); - CHR( ']' ); + ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); } } -- cgit v1.2.3 From f41cc50ef063e9a56dc2d0fc6564a0c0f4344b83 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 7 Apr 2008 11:46:57 +0200 Subject: tgsi: Dump processor type and version as a single token. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index ad9770e474..1d6e8b155d 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -1402,7 +1402,6 @@ dump_gen( CHR( '\n' ); ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); - CHR( ' ' ); UID( parse.FullVersion.Version.MajorVersion ); CHR( '.' ); UID( parse.FullVersion.Version.MinorVersion ); -- cgit v1.2.3 From d758479b9fbff803bdac15f3f39d32ef9064db71 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 11 Apr 2008 10:14:17 -0600 Subject: mesa: Fix glBegin-time test for invalid programs/shaders. Cherry-picked from master. --- src/mesa/vbo/vbo_exec.h | 3 +++ src/mesa/vbo/vbo_exec_api.c | 31 +++++++++++++++++++++++-------- src/mesa/vbo/vbo_exec_array.c | 15 +++++++++++++++ 3 files changed, 41 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo_exec.h b/src/mesa/vbo/vbo_exec.h index b7e8c9fe79..ddbcbe1181 100644 --- a/src/mesa/vbo/vbo_exec.h +++ b/src/mesa/vbo/vbo_exec.h @@ -162,4 +162,7 @@ void vbo_exec_do_EvalCoord2f( struct vbo_exec_context *exec, void vbo_exec_do_EvalCoord1f( struct vbo_exec_context *exec, GLfloat u); +extern GLboolean +vbo_validate_shaders(GLcontext *ctx); + #endif diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index b7f4d8a307..98580170e3 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -477,6 +477,23 @@ static void GLAPIENTRY vbo_exec_EvalPoint2( GLint i, GLint j ) } +/** + * Check if programs/shaders are enabled and valid at glBegin time. + */ +GLboolean +vbo_validate_shaders(GLcontext *ctx) +{ + if ((ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) || + (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled)) { + return GL_FALSE; + } + if (ctx->Shader.CurrentProgram && !ctx->Shader.CurrentProgram->LinkStatus) { + return GL_FALSE; + } + return GL_TRUE; +} + + /* Build a list of primitives on the fly. Keep * ctx->Driver.CurrentExecPrimitive uptodate as well. */ @@ -491,18 +508,16 @@ static void GLAPIENTRY vbo_exec_Begin( GLenum mode ) if (ctx->NewState) { _mesa_update_state( ctx ); - /* XXX also need to check if shader enabled, but invalid */ - if ((ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) || - (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glBegin (invalid vertex/fragment program)"); - return; - } - CALL_Begin(ctx->Exec, (mode)); return; } + if (!vbo_validate_shaders(ctx)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glBegin (invalid vertex/fragment program)"); + return; + } + /* Heuristic: attempt to isolate attributes occuring outside * begin/end pairs. */ diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c index 77f3cf1455..a52521db64 100644 --- a/src/mesa/vbo/vbo_exec_array.c +++ b/src/mesa/vbo/vbo_exec_array.c @@ -245,6 +245,11 @@ vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count) if (ctx->NewState) _mesa_update_state( ctx ); + if (!vbo_validate_shaders(ctx)) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawArrays(bad shader)"); + return; + } + bind_arrays( ctx ); prim[0].begin = 1; @@ -280,6 +285,11 @@ vbo_exec_DrawRangeElements(GLenum mode, if (ctx->NewState) _mesa_update_state( ctx ); + if (!vbo_validate_shaders(ctx)) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawRangeElements(bad shader)"); + return; + } + bind_arrays( ctx ); ib.count = count; @@ -340,6 +350,11 @@ vbo_exec_DrawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *ind if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices )) return; + if (!vbo_validate_shaders(ctx)) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawElements(bad shader)"); + return; + } + if (ctx->Array.ElementArrayBufferObj->Name) { const GLvoid *map = ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER_ARB, -- cgit v1.2.3 From 02250c855fbec5299a2d6118fefa0523ec73654c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 11 Apr 2008 17:45:41 +0100 Subject: gallium: Flush render cache at the beginning of fallback_copy_texsubimage(). It may get stale bits otherwise. --- src/mesa/state_tracker/st_cb_texture.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 43b43ec14c..3db2790267 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -41,6 +41,7 @@ #include "state_tracker/st_cb_fbo.h" #include "state_tracker/st_cb_texture.h" #include "state_tracker/st_format.h" +#include "state_tracker/st_public.h" #include "state_tracker/st_texture.h" #include "state_tracker/st_gen_mipmap.h" @@ -1033,6 +1034,8 @@ fallback_copy_texsubimage(GLcontext *ctx, GLfloat *data; GLint row, yStep; + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); + /* determine bottom-to-top vs. top-to-bottom order */ if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { destY = height - 1 - destY; -- cgit v1.2.3 From 4d184cc33131b440f9aafbcdd2d657050411db49 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 11 Apr 2008 13:20:52 -0600 Subject: gallium: fix broken x86_call() --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 4d33950e99..aea8b28e58 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -317,7 +317,7 @@ void x86_call( struct x86_function *p, void (*label)()) void x86_call( struct x86_function *p, struct x86_reg reg) { emit_1ub(p, 0xff); - emit_modrm(p, reg, reg); + emit_modrm_noreg(p, 2, reg); } #endif -- cgit v1.2.3 From 16900515214912557dfd35e3b333e0e312b8bc61 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 11 Apr 2008 13:21:22 -0600 Subject: mesa: fix broken x86_call() --- src/mesa/x86/rtasm/x86sse.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index 39c0e9b851..7a91364ed8 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -1,3 +1,4 @@ +#ifdef USE_X86_ASM #if defined(__i386__) || defined(__386__) #include "imports.h" @@ -290,7 +291,7 @@ void x86_call( struct x86_function *p, void (*label)()) void x86_call( struct x86_function *p, struct x86_reg reg) { emit_1ub(p, 0xff); - emit_modrm(p, reg, reg); + emit_modrm_noreg(p, 2, reg); } #endif @@ -1191,3 +1192,9 @@ void x86sse_dummy( void ) } #endif + +#else /* USE_X86_ASM */ + +int x86sse_c_dummy_var; /* silence warning */ + +#endif /* USE_X86_ASM */ -- cgit v1.2.3 From 9e8a85ef670dfb7d356d6066bad4710683a07fd9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 11 Apr 2008 13:21:39 -0600 Subject: gallium: fix SCS codegen (sin scalar src comes from X, not Y) --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 4e80597b3f..35c63feccc 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -2020,7 +2020,7 @@ emit_instruction( STORE( func, *inst, 0, 0, CHAN_X ); } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); + FETCH( func, *inst, 0, 0, CHAN_X ); emit_sin( func, 0 ); STORE( func, *inst, 0, 0, CHAN_Y ); } -- cgit v1.2.3 From 097301395d33d57d19bc942f236b4a8c912cc0cc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 11 Apr 2008 13:44:47 -0600 Subject: gallium: comments --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 31 +++++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 35c63feccc..93813f36cb 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -548,6 +548,12 @@ emit_xorps( * Data fetch helpers. */ +/** + * Copy a shader constant to xmm register + * \param xmm the destination xmm register + * \param vec the src const buffer index + * \param chan src channel to fetch (X, Y, Z or W) + */ static void emit_const( struct x86_function *func, @@ -566,6 +572,12 @@ emit_const( SHUF( 0, 0, 0, 0 ) ); } +/** + * Copy a shader input to xmm register + * \param xmm the destination xmm register + * \param vec the src input attrib + * \param chan src channel to fetch (X, Y, Z or W) + */ static void emit_inputf( struct x86_function *func, @@ -579,6 +591,12 @@ emit_inputf( get_input( vec, chan ) ); } +/** + * Store an xmm register to a shader output + * \param xmm the source xmm register + * \param vec the dest output attrib + * \param chan src dest channel to store (X, Y, Z or W) + */ static void emit_output( struct x86_function *func, @@ -592,6 +610,12 @@ emit_output( make_xmm( xmm ) ); } +/** + * Copy a shader temporary to xmm register + * \param xmm the destination xmm register + * \param vec the src temp register + * \param chan src channel to fetch (X, Y, Z or W) + */ static void emit_tempf( struct x86_function *func, @@ -605,6 +629,13 @@ emit_tempf( get_temp( vec, chan ) ); } +/** + * Load an xmm register with an input attrib coefficient (a0, dadx or dady) + * \param xmm the destination xmm register + * \param vec the src input/attribute coefficient index + * \param chan src channel to fetch (X, Y, Z or W) + * \param member 0=a0, 1=dadx, 2=dady + */ static void emit_coef( struct x86_function *func, -- cgit v1.2.3 From e3cf0cd6a9f3f072594e5712763b98ce7e579bcf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 11 Apr 2008 14:18:07 -0600 Subject: gallium: implement immediates (aka literals) for SSE fragment shaders --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 82 +++++++++++++++++++++++++++-- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h | 4 +- src/gallium/drivers/softpipe/sp_fs_sse.c | 10 ++-- 3 files changed, 88 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 93813f36cb..748c53dbfc 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -225,6 +225,15 @@ get_coef_base( void ) return get_output_base(); } +static struct x86_reg +get_immediate_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_DI ); +} + + /** * Data access helpers. */ @@ -238,6 +247,16 @@ get_argument( (index + 1) * 4 ); } +static struct x86_reg +get_immediate( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_immediate_base(), + (vec * 4 + chan) * 4 ); +} + static struct x86_reg get_const( unsigned vec, @@ -572,6 +591,25 @@ emit_const( SHUF( 0, 0, 0, 0 ) ); } +static void +emit_immediate( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_movss( + func, + make_xmm( xmm ), + get_immediate( vec, chan ) ); + emit_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + + /** * Copy a shader input to xmm register * \param xmm the destination xmm register @@ -1191,6 +1229,14 @@ emit_fetch( swizzle ); break; + case TGSI_FILE_IMMEDIATE: + emit_immediate( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + case TGSI_FILE_INPUT: emit_inputf( func, @@ -2343,15 +2389,21 @@ tgsi_emit_sse2( * DECLARATION and INSTRUCTION phase. * GP register holding the output argument is aliased with the coeff argument, * as outputs are not needed in the DECLARATION phase. + * + * \param tokens the TGSI input shader + * \param func the output SSE code/function + * \param immediates buffer to place immediates, later passed to SSE func */ unsigned tgsi_emit_sse2_fs( struct tgsi_token *tokens, - struct x86_function *func ) + struct x86_function *func, + float (*immediates)[4]) { struct tgsi_parse_context parse; boolean instruction_phase = FALSE; unsigned ok = 1; + uint num_immediates = 0; DUMP_START(); @@ -2362,6 +2414,7 @@ tgsi_emit_sse2_fs( func, get_input_base(), get_argument( 0 ) ); + /* skipping outputs argument here */ emit_mov( func, get_const_base(), @@ -2374,6 +2427,10 @@ tgsi_emit_sse2_fs( func, get_coef_base(), get_argument( 4 ) ); + emit_mov( + func, + get_immediate_base(), + get_argument( 5 ) ); tgsi_parse_init( &parse, tokens ); @@ -2407,9 +2464,26 @@ tgsi_emit_sse2_fs( break; case TGSI_TOKEN_TYPE_IMMEDIATE: - /* XXX implement this */ - ok = 0; - debug_printf("failed to emit immediate value to SSE\n"); + /* simply copy the immediate values into the next immediates[] slot */ + { + const uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; + uint i; + assert(size <= 4); + assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES); + for( i = 0; i < size; i++ ) { + immediates[num_immediates][i] = + parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } +#if 0 + debug_printf("SSE FS immediate[%d] = %f %f %f %f\n", + num_immediates, + immediates[num_immediates][0], + immediates[num_immediates][1], + immediates[num_immediates][2], + immediates[num_immediates][3]); +#endif + num_immediates++; + } break; default: diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h index 63b8ef3911..fde06047fe 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h @@ -16,7 +16,9 @@ tgsi_emit_sse2( unsigned tgsi_emit_sse2_fs( struct tgsi_token *tokens, - struct x86_function *function ); + struct x86_function *function, + float (*immediates)[4] + ); #if defined __cplusplus } diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 8095d662ee..9c7948264f 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -51,7 +51,8 @@ typedef void (XSTDCALL *codegen_function)( struct tgsi_exec_vector *output, float (*constant)[4], struct tgsi_exec_vector *temporary, - const struct tgsi_interp_coef *coef + const struct tgsi_interp_coef *coef, + float (*immediates)[4] //, const struct tgsi_exec_vector *quadPos ); @@ -60,6 +61,7 @@ struct sp_sse_fragment_shader { struct sp_fragment_shader base; struct x86_function sse2_program; codegen_function func; + float immediates[TGSI_EXEC_NUM_IMMEDIATES][4]; }; @@ -96,7 +98,8 @@ fs_sse_run( struct sp_fragment_shader *base, machine->Outputs, machine->Consts, machine->Temps, - machine->InterpCoefs + machine->InterpCoefs, + shader->immediates // , &machine->QuadPos ); @@ -129,7 +132,8 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe, x86_init_func( &shader->sse2_program ); - if (!tgsi_emit_sse2_fs( templ->tokens, &shader->sse2_program )) { + if (!tgsi_emit_sse2_fs( templ->tokens, &shader->sse2_program, + shader->immediates)) { FREE(shader); return NULL; } -- cgit v1.2.3 From 7c2416f06e518bc1491fe13e145dcc9487d75449 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 11 Apr 2008 15:02:21 -0600 Subject: gallium: handle TGSI immediates in SSE code for vertex shaders --- src/gallium/auxiliary/draw/draw_vs_sse.c | 9 ++++++--- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 23 +++++++++++++++++++---- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h | 4 +++- 3 files changed, 28 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index f40d65df08..13394129bc 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -50,13 +50,15 @@ typedef void (XSTDCALL *codegen_function) ( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, float (*constant)[4], - struct tgsi_exec_vector *temporary ); + struct tgsi_exec_vector *temporary, + float (*immediates)[4] ); struct draw_sse_vertex_shader { struct draw_vertex_shader base; struct x86_function sse2_program; codegen_function func; + float immediates[TGSI_EXEC_NUM_IMMEDIATES][4]; }; @@ -149,7 +151,8 @@ vs_sse_run( struct draw_vertex_shader *base, shader->func(machine->Inputs, machine->Outputs, machine->Consts, - machine->Temps ); + machine->Temps, + shader->immediates); } @@ -243,7 +246,7 @@ draw_create_vs_sse(struct draw_context *draw, x86_init_func( &vs->sse2_program ); if (!tgsi_emit_sse2( (struct tgsi_token *) vs->base.state.tokens, - &vs->sse2_program )) + &vs->sse2_program, vs->immediates )) goto fail; vs->func = (codegen_function) x86_get_func( &vs->sse2_program ); diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 748c53dbfc..e55fae6047 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -2316,10 +2316,12 @@ emit_declaration( unsigned tgsi_emit_sse2( struct tgsi_token *tokens, - struct x86_function *func ) + struct x86_function *func, + float (*immediates)[4] ) { struct tgsi_parse_context parse; unsigned ok = 1; + uint num_immediates = 0; DUMP_START(); @@ -2341,6 +2343,10 @@ tgsi_emit_sse2( func, get_temp_base(), get_argument( 3 ) ); + emit_mov( + func, + get_immediate_base(), + get_argument( 4 ) ); tgsi_parse_init( &parse, tokens ); @@ -2363,9 +2369,18 @@ tgsi_emit_sse2( break; case TGSI_TOKEN_TYPE_IMMEDIATE: - /* XXX implement this */ - ok = 0; - debug_printf("failed to emit immediate value to SSE\n"); + /* simply copy the immediate values into the next immediates[] slot */ + { + const uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; + uint i; + assert(size <= 4); + assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES); + for( i = 0; i < size; i++ ) { + immediates[num_immediates][i] = + parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } + num_immediates++; + } break; default: diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h index fde06047fe..d1190727d0 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h @@ -11,7 +11,9 @@ struct x86_function; unsigned tgsi_emit_sse2( struct tgsi_token *tokens, - struct x86_function *function ); + struct x86_function *function, + float (*immediates)[4] + ); unsigned tgsi_emit_sse2_fs( -- cgit v1.2.3 From 593cf5a6b55eb9b490a2aee2c3850d2d493fc4df Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 11 Apr 2008 15:27:00 -0600 Subject: gallium: merge the tgsi_emit_sse2() and tgsi_emit_sse2_fs() functions. The two functions were mostly the same. We can look at the shader header info to determine if it's a vertex or fragment shader. --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 201 +++++++++++----------------- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h | 7 - src/gallium/drivers/softpipe/sp_fs_sse.c | 4 +- 3 files changed, 80 insertions(+), 132 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index e55fae6047..c37e201b2b 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -2313,104 +2313,25 @@ emit_declaration( } } -unsigned -tgsi_emit_sse2( - struct tgsi_token *tokens, - struct x86_function *func, - float (*immediates)[4] ) -{ - struct tgsi_parse_context parse; - unsigned ok = 1; - uint num_immediates = 0; - - DUMP_START(); - - func->csr = func->store; - - emit_mov( - func, - get_input_base(), - get_argument( 0 ) ); - emit_mov( - func, - get_output_base(), - get_argument( 1 ) ); - emit_mov( - func, - get_const_base(), - get_argument( 2 ) ); - emit_mov( - func, - get_temp_base(), - get_argument( 3 ) ); - emit_mov( - func, - get_immediate_base(), - get_argument( 4 ) ); - - tgsi_parse_init( &parse, tokens ); - - while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - ok = emit_instruction( - func, - &parse.FullToken.FullInstruction ); - - if (!ok) { - debug_printf("failed to translate tgsi opcode %d to SSE\n", - parse.FullToken.FullInstruction.Instruction.Opcode ); - } - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - /* simply copy the immediate values into the next immediates[] slot */ - { - const uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; - uint i; - assert(size <= 4); - assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES); - for( i = 0; i < size; i++ ) { - immediates[num_immediates][i] = - parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; - } - num_immediates++; - } - break; - - default: - assert( 0 ); - ok = 0; - break; - } - } - - tgsi_parse_free( &parse ); - - DUMP_END(); - - return ok; -} /** - * Fragment shaders are responsible for interpolating shader inputs. Because on - * x86 we have only 4 GP registers, and here we have 5 shader arguments (input, - * output, const, temp and coef), the code is split into two phases -- - * DECLARATION and INSTRUCTION phase. - * GP register holding the output argument is aliased with the coeff argument, - * as outputs are not needed in the DECLARATION phase. + * Translate a TGSI vertex/fragment shader to SSE2 code. + * Slightly different things are done for vertex vs. fragment shaders. + * + * Note that fragment shaders are responsible for interpolating shader + * inputs. Because on x86 we have only 4 GP registers, and here we + * have 5 shader arguments (input, output, const, temp and coef), the + * code is split into two phases -- DECLARATION and INSTRUCTION phase. + * GP register holding the output argument is aliased with the coeff + * argument, as outputs are not needed in the DECLARATION phase. * * \param tokens the TGSI input shader * \param func the output SSE code/function * \param immediates buffer to place immediates, later passed to SSE func + * \param return 1 for success, 0 if translation failed */ unsigned -tgsi_emit_sse2_fs( +tgsi_emit_sse2( struct tgsi_token *tokens, struct x86_function *func, float (*immediates)[4]) @@ -2424,50 +2345,84 @@ tgsi_emit_sse2_fs( func->csr = func->store; - /* DECLARATION phase, do not load output argument. */ - emit_mov( - func, - get_input_base(), - get_argument( 0 ) ); - /* skipping outputs argument here */ - emit_mov( - func, - get_const_base(), - get_argument( 2 ) ); - emit_mov( - func, - get_temp_base(), - get_argument( 3 ) ); - emit_mov( - func, - get_coef_base(), - get_argument( 4 ) ); - emit_mov( - func, - get_immediate_base(), - get_argument( 5 ) ); - tgsi_parse_init( &parse, tokens ); + /* + * Different function args for vertex/fragment shaders: + */ + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { + /* DECLARATION phase, do not load output argument. */ + emit_mov( + func, + get_input_base(), + get_argument( 0 ) ); + /* skipping outputs argument here */ + emit_mov( + func, + get_const_base(), + get_argument( 2 ) ); + emit_mov( + func, + get_temp_base(), + get_argument( 3 ) ); + emit_mov( + func, + get_coef_base(), + get_argument( 4 ) ); + emit_mov( + func, + get_immediate_base(), + get_argument( 5 ) ); + } + else { + assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX); + + emit_mov( + func, + get_input_base(), + get_argument( 0 ) ); + emit_mov( + func, + get_output_base(), + get_argument( 1 ) ); + emit_mov( + func, + get_const_base(), + get_argument( 2 ) ); + emit_mov( + func, + get_temp_base(), + get_argument( 3 ) ); + emit_mov( + func, + get_immediate_base(), + get_argument( 4 ) ); + } + while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { tgsi_parse_token( &parse ); switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: - emit_declaration( - func, - &parse.FullToken.FullDeclaration ); + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { + emit_declaration( + func, + &parse.FullToken.FullDeclaration ); + } break; case TGSI_TOKEN_TYPE_INSTRUCTION: - if( !instruction_phase ) { - /* INSTRUCTION phase, overwrite coeff with output. */ - instruction_phase = TRUE; - emit_mov( - func, - get_output_base(), - get_argument( 1 ) ); + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { + if( !instruction_phase ) { + /* INSTRUCTION phase, overwrite coeff with output. */ + instruction_phase = TRUE; + emit_mov( + func, + get_output_base(), + get_argument( 1 ) ); + } } + ok = emit_instruction( func, &parse.FullToken.FullInstruction ); diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h index d1190727d0..d56bf7f98a 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h @@ -15,13 +15,6 @@ tgsi_emit_sse2( float (*immediates)[4] ); -unsigned -tgsi_emit_sse2_fs( - struct tgsi_token *tokens, - struct x86_function *function, - float (*immediates)[4] - ); - #if defined __cplusplus } #endif diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 9c7948264f..5ef02a7142 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -132,8 +132,8 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe, x86_init_func( &shader->sse2_program ); - if (!tgsi_emit_sse2_fs( templ->tokens, &shader->sse2_program, - shader->immediates)) { + if (!tgsi_emit_sse2( templ->tokens, &shader->sse2_program, + shader->immediates)) { FREE(shader); return NULL; } -- cgit v1.2.3 From 5dcfc1f1b637aefbed631cc1265ce3808f2ff249 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Sat, 12 Apr 2008 14:34:59 +0200 Subject: Fixed memory leak --- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 40902197e1..44baa6d9fa 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -93,6 +93,7 @@ intel_buffer_destroy(struct pipe_winsys *winsys, struct pipe_buffer *buf) { driBOUnReference( dri_bo(buf) ); + FREE(buf); } -- cgit v1.2.3 From 2ebc99fcbc0c8fc6f6ce50e2ee674312e214ea2f Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 12 Apr 2008 11:03:56 -0600 Subject: gallium: move duplicated compute_clipmask() code to draw_vs.h --- src/gallium/auxiliary/draw/draw_vs.h | 33 +++++++++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs_exec.c | 26 ------------------------ src/gallium/auxiliary/draw/draw_vs_llvm.c | 27 ------------------------- src/gallium/auxiliary/draw/draw_vs_sse.c | 28 -------------------------- 4 files changed, 33 insertions(+), 81 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 4ee7e705e9..33ce1e335e 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -31,6 +31,10 @@ #ifndef DRAW_VS_H #define DRAW_VS_H +#include "draw_context.h" +#include "draw_private.h" + + struct draw_vertex_shader; struct draw_context; struct pipe_shader_state; @@ -47,4 +51,33 @@ struct draw_vertex_shader * draw_create_vs_llvm(struct draw_context *draw, const struct pipe_shader_state *templ); + +/* Should be part of the generated shader: + */ +static INLINE unsigned +compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) +{ + unsigned mask = 0x0; + unsigned i; + + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1< Date: Sat, 12 Apr 2008 22:59:17 +0200 Subject: tgsi: Fix source register short dump code. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 62 +++++++++++++++-------------- 1 file changed, 32 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 1d6e8b155d..cb3573ceb6 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -968,18 +968,18 @@ dump_instruction_short( } CHR( ' ' ); - if( src->SrcRegisterExtMod.Complement ) { - TXT( "(1 - " ); - } - if( src->SrcRegisterExtMod.Negate ) { - CHR( '-' ); - } - if( src->SrcRegisterExtMod.Absolute ) { + if (src->SrcRegisterExtMod.Negate) + TXT( "-(" ); + if (src->SrcRegisterExtMod.Absolute) CHR( '|' ); - } - if( src->SrcRegister.Negate ) { + if (src->SrcRegisterExtMod.Scale2X) + TXT( "2*(" ); + if (src->SrcRegisterExtMod.Bias) + CHR( '(' ); + if (src->SrcRegisterExtMod.Complement) + TXT( "1-(" ); + if (src->SrcRegister.Negate) CHR( '-' ); - } ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); @@ -987,35 +987,37 @@ dump_instruction_short( SID( src->SrcRegister.Index ); CHR( ']' ); - if (src->SrcRegister.Extended) { - if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || - src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || - src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || - src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { - CHR( '.' ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); - } - } - else if( src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || - src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || - src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || - src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W ) { + if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) { CHR( '.' ); ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); } + if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { + CHR( '.' ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); + } - if( src->SrcRegisterExtMod.Absolute ) { + if (src->SrcRegisterExtMod.Complement) + CHR( ')' ); + if (src->SrcRegisterExtMod.Bias) + TXT( ")-.5" ); + if (src->SrcRegisterExtMod.Scale2X) + CHR( ')' ); + if (src->SrcRegisterExtMod.Absolute) CHR( '|' ); - } - if( src->SrcRegisterExtMod.Complement ) { + if (src->SrcRegisterExtMod.Negate) CHR( ')' ); - } first_reg = FALSE; } -- cgit v1.2.3 From cd5931240688cb8bd12834e3ba23f858f26dbf8c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 13 Apr 2008 12:23:26 +0900 Subject: gallium: OS independent time-manipulation functions. --- src/gallium/auxiliary/util/Makefile | 3 +- src/gallium/auxiliary/util/SConscript | 1 + src/gallium/auxiliary/util/u_time.c | 152 ++++++++++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_time.h | 99 ++++++++++++++++++++++ 4 files changed, 254 insertions(+), 1 deletion(-) create mode 100644 src/gallium/auxiliary/util/u_time.c create mode 100644 src/gallium/auxiliary/util/u_time.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 9b6c2708b6..fe82fa6378 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -14,7 +14,8 @@ C_SOURCES = \ u_hash_table.c \ u_mm.c \ u_simple_shaders.c \ - u_snprintf.c + u_snprintf.c \ + u_mm.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 9b3929eb2d..d55d2c7081 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -15,6 +15,7 @@ util = env.ConvenienceLibrary( 'u_mm.c', 'u_simple_shaders.c', 'u_snprintf.c', + 'u_time.c', ]) auxiliaries.insert(0, util) diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c new file mode 100644 index 0000000000..e6c0b19ff6 --- /dev/null +++ b/src/gallium/auxiliary/util/u_time.c @@ -0,0 +1,152 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * OS independent time-manipulation functions. + * + * @author Jose Fonseca + */ + + +#ifndef WIN32 +#include +#else +#include +#include +#endif + +#include "util/u_time.h" + + +#ifdef WIN32 +static LONGLONG frequency = 0; +#endif + + +void +util_time_get(struct util_time *t) +{ +#ifndef WIN32 + gettimeofday(&t->tv, NULL); +#else + EngQueryPerformanceCounter(&t->counter); +#endif +} + + +void +util_time_add(const struct util_time *t1, + int64_t usecs, + struct util_time *t2) +{ +#ifndef WIN32 + t2->tv.tv_sec = t1->tv.tv_sec + usecs / 1000000; + t2->tv.tv_usec = t1->tv.tv_usec + usecs % 1000000; +#else + if(!frequency) + EngQueryPerformanceFrequency(&frequency); + t2->counter = t1->counter + (usecs * frequency + 999999LL)/1000000LL; +#endif +} + + +int64_t +util_time_diff(const struct util_time *t1, + const struct util_time *t2) +{ +#ifndef WIN32 + return (t2->tv.tv_usec - t1->tv.tv_usec) + + (t2->tv.tv_sec - t1->tv.tv_sec)*1000000; +#else + return (t2->counter - t1->counter)*1000000LL/frequency; +#endif +} + + +/** + * Compare two time values. + * + * Not publicly available because it does not take in account wrap-arounds. + * Use util_time_timeout instead. + */ +static INLINE int +util_time_compare(const struct util_time *t1, + const struct util_time *t2) +{ +#ifndef WIN32 + if (t1->tv.tv_sec < t2->tv.tv_sec) + return -1; + else if(t1->tv.tv_sec > t2->tv.tv_sec) + return 1; + else if (t1->tv.tv_usec < t2->tv.tv_usec) + return -1; + else if(t1->tv.tv_usec > t2->tv.tv_usec) + return 1; + else + return 0; +#else + if (t1->counter < t2->counter) + return -1; + else if(t1->counter > t2->counter) + return 1; + else + return 0; +#endif +} + + +int +util_time_timeout(const struct util_time *start, + const struct util_time *end, + const struct util_time *curr) +{ + if(util_time_compare(start, end) <= 0) + return util_time_compare(start, curr) <= 0 && util_time_compare(curr, end) < 0; + else + return util_time_compare(start, curr) <= 0 || util_time_compare(curr, end) < 0; +} + + +#ifdef WIN32 +void util_time_usleep(unsigned usecs) +{ + LONGLONG start, curr, end; + + EngQueryPerformanceCounter(&start); + + if(!frequency) + EngQueryPerformanceFrequency(&frequency); + + end = start + (usecs * frequency + 999999LL)/1000000LL; + + do { + EngQueryPerformanceCounter(&curr); + } while(start <= curr && curr < end || + end < start && (curr < end || start <= curr)); +} +#endif diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h new file mode 100644 index 0000000000..32035cceb5 --- /dev/null +++ b/src/gallium/auxiliary/util/u_time.h @@ -0,0 +1,99 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * OS independent time-manipulation functions. + * + * @author Jose Fonseca + */ + +#ifndef U_TIME_H_ +#define U_TIME_H_ + + +#ifndef WIN32 +#include /* timeval */ +#include /* usleep */ +#endif + +#include "pipe/p_compiler.h" + + +#ifdef __cplusplus +extern "C" { +#endif + + +/** + * Time abstraction. + * + * Do not access this structure directly. Use the provided function instead. + */ +struct util_time +{ +#ifndef WIN32 + struct timeval tv; +#else + long long counter; +#endif +}; + + +void +util_time_get(struct util_time *t); + +void +util_time_add(const struct util_time *t1, + int64_t usecs, + struct util_time *t2); + +int64_t +util_time_diff(const struct util_time *t1, + const struct util_time *t2); + +/** + * Returns zero when the timeout expires, non zero otherwise. + */ +int +util_time_timeout(const struct util_time *start, + const struct util_time *end, + const struct util_time *curr); + +#ifndef WIN32 +#define util_time_sleep usleep +#else +int +util_time_sleep(unsigned usecs); +#endif + + +#ifdef __cplusplus +} +#endif + +#endif /* U_TIME_H_ */ -- cgit v1.2.3 From 21c302b0ec39480a7eaab7827cce5b609d196606 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 13 Apr 2008 14:10:46 +0900 Subject: gallium: Initial port of Thomas slab suballocator to pipebuffer. Not tested yet -- just compiles. This includes only the slab algorithm. Fencing is already implemented in pb_bufmgr_fence and time-based caching will be commited in a separate module shortly. --- src/gallium/auxiliary/pipebuffer/Makefile | 1 + src/gallium/auxiliary/pipebuffer/SConscript | 1 + src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 9 + src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c | 429 ++++++++++++++++++++++ 4 files changed, 440 insertions(+) create mode 100644 src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile index a9fa518c67..0770cad45a 100644 --- a/src/gallium/auxiliary/pipebuffer/Makefile +++ b/src/gallium/auxiliary/pipebuffer/Makefile @@ -9,6 +9,7 @@ C_SOURCES = \ pb_bufmgr_fenced.c \ pb_bufmgr_mm.c \ pb_bufmgr_pool.c \ + pb_bufmgr_slab.c \ pb_winsys.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript index 3d41fd84a6..14c3c77e37 100644 --- a/src/gallium/auxiliary/pipebuffer/SConscript +++ b/src/gallium/auxiliary/pipebuffer/SConscript @@ -8,6 +8,7 @@ pipebuffer = env.ConvenienceLibrary( 'pb_bufmgr_fenced.c', 'pb_bufmgr_mm.c', 'pb_bufmgr_pool.c', + 'pb_bufmgr_slab.c', 'pb_winsys.c', ]) diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index 0cf8e92e37..a2377f70e2 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -114,6 +114,15 @@ mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, size_t size, size_t align2); +struct pb_manager * +pb_slab_manager_create(struct pb_manager *provider, + const struct pb_desc *desc, + size_t smallestSize, + size_t numSizes, + size_t desiredNumBuffers, + size_t maxSlabSize, + size_t pageAlignment); + /** * Fenced buffer manager. * diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c new file mode 100644 index 0000000000..676e8e29b9 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c @@ -0,0 +1,429 @@ +/************************************************************************** + * + * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, FREE of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * @file + * S-lab pool implementation. + * + * @author Thomas Hellstrom + * @author Jose Fonseca + */ + +#include "pipe/p_compiler.h" +#include "pipe/p_error.h" +#include "pipe/p_debug.h" +#include "pipe/p_thread.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "util/u_double_list.h" +#include "util/u_time.h" + +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +#define DRI_SLABPOOL_ALLOC_RETRIES 100 + + +struct pb_slab; + +struct pb_slab_buffer +{ + struct pb_buffer base; + + struct pb_slab *slab; + struct list_head head; + unsigned mapCount; + size_t start; + _glthread_Cond event; +}; + +struct pb_slab +{ + struct list_head head; + struct list_head freeBuffers; + size_t numBuffers; + size_t numFree; + struct pb_slab_buffer *buffers; + struct pb_slab_size_header *header; + + struct pb_buffer *bo; + size_t pageAlignment; + void *virtual; +}; + +struct pb_slab_size_header +{ + struct list_head slabs; + struct list_head freeSlabs; + struct pb_slab_manager *pool; + size_t bufSize; + _glthread_Mutex mutex; +}; + +/** + * The data of this structure remains constant after + * initialization and thus needs no mutex protection. + */ +struct pb_slab_manager +{ + struct pb_manager base; + + struct pb_desc desc; + size_t *bucketSizes; + size_t numBuckets; + size_t pageSize; + struct pb_manager *provider; + unsigned pageAlignment; + unsigned maxSlabSize; + unsigned desiredNumBuffers; + struct pb_slab_size_header *headers; +}; + + +static INLINE struct pb_slab_buffer * +pb_slab_buffer(struct pb_buffer *buf) +{ + assert(buf); + return (struct pb_slab_buffer *)buf; +} + + +static INLINE struct pb_slab_manager * +pb_slab_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct pb_slab_manager *)mgr; +} + + +/** + * Delete a buffer from the slab header delayed list and put + * it on the slab FREE list. + */ +static void +pb_slab_buffer_destroy(struct pb_buffer *_buf) +{ + struct pb_slab_buffer *buf = pb_slab_buffer(_buf); + struct pb_slab *slab = buf->slab; + struct pb_slab_size_header *header = slab->header; + struct list_head *list = &buf->head; + + _glthread_LOCK_MUTEX(header->mutex); + + assert(buf->base.base.refcount == 0); + + buf->mapCount = 0; + + LIST_DEL(list); + LIST_ADDTAIL(list, &slab->freeBuffers); + slab->numFree++; + + if (slab->head.next == &slab->head) + LIST_ADDTAIL(&slab->head, &header->slabs); + + if (slab->numFree == slab->numBuffers) { + list = &slab->head; + LIST_DEL(list); + LIST_ADDTAIL(list, &header->freeSlabs); + } + + if (header->slabs.next == &header->slabs || slab->numFree + != slab->numBuffers) { + + struct list_head *next; + + for (list = header->freeSlabs.next, next = list->next; list + != &header->freeSlabs; list = next, next = list->next) { + + slab = LIST_ENTRY(struct pb_slab, list, head); + + LIST_DELINIT(list); + pb_reference(&slab->bo, NULL); + FREE(slab->buffers); + FREE(slab); + } + } + + _glthread_UNLOCK_MUTEX(header->mutex); +} + + +static void * +pb_slab_buffer_map(struct pb_buffer *_buf, + unsigned flags) +{ + struct pb_slab_buffer *buf = pb_slab_buffer(_buf); + + ++buf->mapCount; + return (void *) ((uint8_t *) buf->slab->virtual + buf->start); +} + + +static void +pb_slab_buffer_unmap(struct pb_buffer *_buf) +{ + struct pb_slab_buffer *buf = pb_slab_buffer(_buf); + + --buf->mapCount; + if (buf->mapCount == 0) + _glthread_COND_BROADCAST(buf->event); +} + + +static void +pb_slab_buffer_get_base_buffer(struct pb_buffer *_buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + struct pb_slab_buffer *buf = pb_slab_buffer(_buf); + pb_get_base_buffer(buf->slab->bo, base_buf, offset); + *offset += buf->start; +} + + +static const struct pb_vtbl +pb_slab_buffer_vtbl = { + pb_slab_buffer_destroy, + pb_slab_buffer_map, + pb_slab_buffer_unmap, + pb_slab_buffer_get_base_buffer +}; + + +static enum pipe_error +pb_slab_create(struct pb_slab_size_header *header) +{ + struct pb_slab_manager *pool = header->pool; + size_t size = header->bufSize * pool->desiredNumBuffers; + struct pb_slab *slab; + struct pb_slab_buffer *buf; + size_t numBuffers; + int ret; + unsigned i; + + slab = CALLOC_STRUCT(pb_slab); + if (!slab) + return PIPE_ERROR_OUT_OF_MEMORY; + + /* + * FIXME: We should perhaps allow some variation in slabsize in order + * to efficiently reuse slabs. + */ + + size = (size <= pool->maxSlabSize) ? size : pool->maxSlabSize; + size = (size + pool->pageSize - 1) & ~(pool->pageSize - 1); + + slab->bo = pool->provider->create_buffer(pool->provider, size, &pool->desc); + if(!slab->bo) + goto out_err0; + + slab->virtual = pb_map(slab->bo, + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); + if(!slab->virtual) + goto out_err1; + + pb_unmap(slab->bo); + + numBuffers = slab->bo->base.size / header->bufSize; + + slab->buffers = CALLOC(numBuffers, sizeof(*slab->buffers)); + if (!slab->buffers) { + ret = PIPE_ERROR_OUT_OF_MEMORY; + goto out_err1; + } + + LIST_INITHEAD(&slab->head); + LIST_INITHEAD(&slab->freeBuffers); + slab->numBuffers = numBuffers; + slab->numFree = 0; + slab->header = header; + + buf = slab->buffers; + for (i=0; i < numBuffers; ++i) { + buf->base.base.refcount = 0; + buf->base.base.size = header->bufSize; + buf->base.base.alignment = 0; + buf->base.base.usage = 0; + buf->base.vtbl = &pb_slab_buffer_vtbl; + buf->slab = slab; + buf->start = i* header->bufSize; + buf->mapCount = 0; + _glthread_INIT_COND(buf->event); + LIST_ADDTAIL(&buf->head, &slab->freeBuffers); + slab->numFree++; + buf++; + } + + LIST_ADDTAIL(&slab->head, &header->slabs); + + return PIPE_OK; + +out_err1: + pb_reference(&slab->bo, NULL); +out_err0: + FREE(slab); + return ret; +} + + +static struct pb_buffer * +pb_slab_manager_create_buffer(struct pb_manager *_pool, + size_t size, + const struct pb_desc *desc) +{ + struct pb_slab_manager *pool = pb_slab_manager(_pool); + struct pb_slab_size_header *header; + unsigned i; + static struct pb_slab_buffer *buf; + struct pb_slab *slab; + struct list_head *list; + int count = DRI_SLABPOOL_ALLOC_RETRIES; + + /* + * FIXME: Check for compatibility. + */ + + header = pool->headers; + for (i=0; inumBuckets; ++i) { + if (header->bufSize >= size) + break; + header++; + } + + if (i >= pool->numBuckets) + /* Fall back to allocate a buffer object directly from the provider. */ + return pool->provider->create_buffer(pool->provider, size, desc); + + + _glthread_LOCK_MUTEX(header->mutex); + while (header->slabs.next == &header->slabs && count > 0) { + if (header->slabs.next != &header->slabs) + break; + + _glthread_UNLOCK_MUTEX(header->mutex); + if (count != DRI_SLABPOOL_ALLOC_RETRIES) + util_time_sleep(1); + _glthread_LOCK_MUTEX(header->mutex); + (void) pb_slab_create(header); + count--; + } + + list = header->slabs.next; + if (list == &header->slabs) { + _glthread_UNLOCK_MUTEX(header->mutex); + return NULL; + } + slab = LIST_ENTRY(struct pb_slab, list, head); + if (--slab->numFree == 0) + LIST_DELINIT(list); + + list = slab->freeBuffers.next; + LIST_DELINIT(list); + + _glthread_UNLOCK_MUTEX(header->mutex); + buf = LIST_ENTRY(struct pb_slab_buffer, list, head); + ++buf->base.base.refcount; + return &buf->base; +} + + +static void +pb_slab_manager_destroy(struct pb_manager *_pool) +{ + struct pb_slab_manager *pool = pb_slab_manager(_pool); + + FREE(pool->headers); + FREE(pool->bucketSizes); + FREE(pool); +} + + +struct pb_manager * +pb_slab_manager_create(struct pb_manager *provider, + const struct pb_desc *desc, + size_t smallestSize, + size_t numSizes, + size_t desiredNumBuffers, + size_t maxSlabSize, + size_t pageAlignment) +{ + struct pb_slab_manager *pool; + size_t i; + + pool = CALLOC_STRUCT(pb_slab_manager); + if (!pool) + goto out_err0; + + pool->bucketSizes = CALLOC(numSizes, sizeof(*pool->bucketSizes)); + if (!pool->bucketSizes) + goto out_err1; + + pool->headers = CALLOC(numSizes, sizeof(*pool->headers)); + if (!pool->headers) + goto out_err2; + + pool->desc = *desc; + pool->numBuckets = numSizes; +#ifdef WIN32 + pool->pageSize = 4096; +#else + pool->pageSize = getpagesize(); +#endif + pool->provider = provider; + pool->pageAlignment = pageAlignment; + pool->maxSlabSize = maxSlabSize; + pool->desiredNumBuffers = desiredNumBuffers; + + for (i=0; inumBuckets; ++i) { + struct pb_slab_size_header *header = &pool->headers[i]; + + pool->bucketSizes[i] = (smallestSize << i); + + _glthread_INIT_MUTEX(header->mutex); + + LIST_INITHEAD(&header->slabs); + LIST_INITHEAD(&header->freeSlabs); + + header->pool = pool; + header->bufSize = (smallestSize << i); + } + + pool->base.destroy = pb_slab_manager_destroy; + pool->base.create_buffer = pb_slab_manager_create_buffer; + + return &pool->base; + +out_err2: + FREE(pool->bucketSizes); +out_err1: + FREE(pool); +out_err0: + return NULL; +} -- cgit v1.2.3 From fb2b5f7a4ac411a5bb5cde12ba15265b30c032e8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 13 Apr 2008 15:10:01 +0900 Subject: gallium: Buffer cache. --- src/gallium/auxiliary/pipebuffer/Makefile | 1 + src/gallium/auxiliary/pipebuffer/SConscript | 1 + src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 20 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c | 299 +++++++++++++++++++++ 4 files changed, 318 insertions(+), 3 deletions(-) create mode 100644 src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile index 0770cad45a..d654dbcc91 100644 --- a/src/gallium/auxiliary/pipebuffer/Makefile +++ b/src/gallium/auxiliary/pipebuffer/Makefile @@ -6,6 +6,7 @@ LIBNAME = pipebuffer C_SOURCES = \ pb_buffer_fenced.c \ pb_buffer_malloc.c \ + pb_bufmgr_cache.c \ pb_bufmgr_fenced.c \ pb_bufmgr_mm.c \ pb_bufmgr_pool.c \ diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript index 14c3c77e37..604a217982 100644 --- a/src/gallium/auxiliary/pipebuffer/SConscript +++ b/src/gallium/auxiliary/pipebuffer/SConscript @@ -5,6 +5,7 @@ pipebuffer = env.ConvenienceLibrary( source = [ 'pb_buffer_fenced.c', 'pb_buffer_malloc.c', + 'pb_bufmgr_cache.c', 'pb_bufmgr_fenced.c', 'pb_bufmgr_mm.c', 'pb_bufmgr_pool.c', diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index a2377f70e2..b2d2520b67 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -80,7 +80,7 @@ struct pb_manager /** - * Static buffer pool manager. + * Static buffer pool sub-allocator. * * Manages the allocation of equally sized buffers. It does so by allocating * a single big buffer and divide it equally sized buffers. @@ -94,7 +94,7 @@ pool_bufmgr_create(struct pb_manager *provider, /** - * Wraper around the old memory manager. + * Static sub-allocator based the old memory manager. * * It managers buffers of different sizes. It does so by allocating a buffer * with the size of the heap, and then using the old mm memory manager to manage @@ -114,6 +114,9 @@ mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, size_t size, size_t align2); +/** + * Slab sub-allocator. + */ struct pb_manager * pb_slab_manager_create(struct pb_manager *provider, const struct pb_desc *desc, @@ -122,7 +125,18 @@ pb_slab_manager_create(struct pb_manager *provider, size_t desiredNumBuffers, size_t maxSlabSize, size_t pageAlignment); - + + +/** + * Time-based buffer cache. + * + * This manager keeps a cache of destroyed buffers during a time interval. + */ +struct pb_manager * +pb_cache_manager_create(struct pb_manager *provider, + unsigned usecs); + + /** * Fenced buffer manager. * diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c new file mode 100644 index 0000000000..06de0bb6c3 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c @@ -0,0 +1,299 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Buffer cache. + * + * \author José Fonseca + * \author Thomas Hellström + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_winsys.h" +#include "pipe/p_thread.h" +#include "pipe/p_util.h" +#include "util/u_double_list.h" +#include "util/u_time.h" + +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct pb_cache_manager; + + +/** + * Wrapper around a pipe buffer which adds delayed destruction. + */ +struct pb_cache_buffer +{ + struct pb_buffer base; + + struct pb_buffer *buffer; + struct pb_cache_manager *mgr; + + /** Caching time interval */ + struct util_time start, end; + + struct list_head head; +}; + + +struct pb_cache_manager +{ + struct pb_manager base; + + struct pb_manager *provider; + unsigned usecs; + + _glthread_Mutex mutex; + + struct list_head delayed; + size_t numDelayed; +}; + + +static INLINE struct pb_cache_buffer * +pb_cache_buffer(struct pb_buffer *buf) +{ + assert(buf); + return (struct pb_cache_buffer *)buf; +} + + +static INLINE struct pb_cache_manager * +pb_cache_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct pb_cache_manager *)mgr; +} + + +/** + * Actually destroy the buffer. + */ +static INLINE void +_pb_cache_buffer_destroy(struct pb_cache_buffer *buf) +{ + struct pb_cache_manager *mgr = buf->mgr; + + LIST_DEL(&buf->head); + assert(mgr->numDelayed); + --mgr->numDelayed; + assert(!buf->base.base.refcount); + pb_reference(&buf->buffer, NULL); + FREE(buf); +} + + +/** + * Free as many cache buffers from the list head as possible. + */ +static void +_pb_cache_buffer_list_check_free(struct pb_cache_manager *mgr) +{ + struct list_head *curr, *next; + struct pb_cache_buffer *buf; + struct util_time now; + + util_time_get(&now); + + curr = mgr->delayed.next; + next = curr->next; + while(curr != &mgr->delayed) { + buf = LIST_ENTRY(struct pb_cache_buffer, curr, head); + + if(util_time_timeout(&buf->start, &buf->end, &now) != 0) + break; + + _pb_cache_buffer_destroy(buf); + + curr = next; + next = curr->next; + } +} + + +static void +pb_cache_buffer_destroy(struct pb_buffer *_buf) +{ + struct pb_cache_buffer *buf = pb_cache_buffer(_buf); + struct pb_cache_manager *mgr = buf->mgr; + + _glthread_LOCK_MUTEX(mgr->mutex); + assert(buf->base.base.refcount == 0); + + _pb_cache_buffer_list_check_free(mgr); + + util_time_get(&buf->start); + util_time_add(&buf->start, mgr->usecs, &buf->end); + LIST_ADDTAIL(&buf->head, &mgr->delayed); + ++mgr->numDelayed; + _glthread_UNLOCK_MUTEX(mgr->mutex); +} + + +static void * +pb_cache_buffer_map(struct pb_buffer *_buf, + unsigned flags) +{ + struct pb_cache_buffer *buf = pb_cache_buffer(_buf); + return pb_map(buf->buffer, flags); +} + + +static void +pb_cache_buffer_unmap(struct pb_buffer *_buf) +{ + struct pb_cache_buffer *buf = pb_cache_buffer(_buf); + pb_unmap(buf->buffer); +} + + +static void +pb_cache_buffer_get_base_buffer(struct pb_buffer *_buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + struct pb_cache_buffer *buf = pb_cache_buffer(_buf); + pb_get_base_buffer(buf->buffer, base_buf, offset); +} + + +const struct pb_vtbl +pb_cache_buffer_vtbl = { + pb_cache_buffer_destroy, + pb_cache_buffer_map, + pb_cache_buffer_unmap, + pb_cache_buffer_get_base_buffer +}; + + +static struct pb_buffer * +pb_cache_manager_create_buffer(struct pb_manager *_mgr, + size_t size, + const struct pb_desc *desc) +{ + struct pb_cache_manager *mgr = pb_cache_manager(_mgr); + struct pb_cache_buffer *buf; + struct list_head *curr, *next; + struct util_time now; + + util_time_get(&now); + curr = mgr->delayed.next; + next = curr->next; + while(curr != &mgr->delayed) { + buf = LIST_ENTRY(struct pb_cache_buffer, curr, head); + + if(buf->base.base.size == size && + buf->base.base.alignment >= desc->alignment && + (buf->base.base.alignment % desc->alignment) == 0 && + /* buf->base.base.usage == usage */ 1) { + ++buf->base.base.refcount; + return &buf->base; + } + + if(util_time_timeout(&buf->start, &buf->end, &now) != 0) + _pb_cache_buffer_destroy(buf); + + curr = next; + next = curr->next; + } + + buf = CALLOC_STRUCT(pb_cache_buffer); + if(!buf) + return NULL; + + buf->buffer = mgr->provider->create_buffer(mgr->provider, size, desc); + if(!buf->buffer) { + FREE(buf); + return NULL; + } + + buf->base.base.refcount = 1; + buf->base.base.alignment = buf->buffer->base.alignment; + buf->base.base.usage = buf->buffer->base.usage; + buf->base.base.size = buf->buffer->base.size; + + buf->base.vtbl = &pb_cache_buffer_vtbl; + buf->mgr = mgr; + + return &buf->base; +} + + +static void +pb_cache_manager_destroy(struct pb_manager *_mgr) +{ + struct pb_cache_manager *mgr = pb_cache_manager(_mgr); + struct list_head *curr, *next; + struct pb_cache_buffer *buf; + + _glthread_LOCK_MUTEX(mgr->mutex); + curr = mgr->delayed.next; + next = curr->next; + while(curr != &mgr->delayed) { + buf = LIST_ENTRY(struct pb_cache_buffer, curr, head); + _pb_cache_buffer_destroy(buf); + curr = next; + next = curr->next; + } + _glthread_UNLOCK_MUTEX(mgr->mutex); + + FREE(mgr); +} + + +struct pb_manager * +pb_cache_manager_create(struct pb_manager *provider, + unsigned usecs) +{ + struct pb_cache_manager *mgr; + + mgr = (struct pb_cache_manager *)CALLOC(1, sizeof(*mgr)); + if (!mgr) + return NULL; + + mgr->base.destroy = pb_cache_manager_destroy; + mgr->base.create_buffer = pb_cache_manager_create_buffer; + mgr->provider = provider; + mgr->usecs = usecs; + LIST_INITHEAD(&mgr->delayed); + mgr->numDelayed = 0; + _glthread_INIT_MUTEX(mgr->mutex); + + return &mgr->base; +} -- cgit v1.2.3 From 84994693f53668d5ca72d57765a0729fe343593b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 13 Apr 2008 18:52:54 +0900 Subject: gallium: Add u_time.c --- src/gallium/auxiliary/util/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index fe82fa6378..05bc43131a 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -15,6 +15,7 @@ C_SOURCES = \ u_mm.c \ u_simple_shaders.c \ u_snprintf.c \ + u_time.c \ u_mm.c include ../../Makefile.template -- cgit v1.2.3 From 4f550ab821f9aef9f19d9f1e10785f8c1f511ad4 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 7 Apr 2008 20:38:39 -0400 Subject: introduce a define to maxout the processed vertices --- src/gallium/auxiliary/draw/draw_vertex_shader.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index 133418baca..d5f37bca21 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -37,7 +37,7 @@ #include "draw_context.h" #include "draw_vs.h" - +#define MAX_SHADER_VERTICES 4 /** * Run the vertex shader on all vertices in the vertex queue. @@ -58,23 +58,23 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) // fprintf(stderr, "%s %d\n", __FUNCTION__, draw->vs.queue_nr ); /* run vertex shader on vertex cache entries, four per invokation */ - for (i = 0; i < draw->vs.queue_nr; i += 4) { - struct vertex_header *dests[4]; - unsigned elts[4]; - int j, n = MIN2(4, draw->vs.queue_nr - i); + for (i = 0; i < draw->vs.queue_nr; i += MAX_SHADER_VERTICES) { + struct vertex_header *dests[MAX_SHADER_VERTICES]; + unsigned elts[MAX_SHADER_VERTICES]; + int j, n = MIN2(MAX_SHADER_VERTICES, - i); for (j = 0; j < n; j++) { elts[j] = draw->vs.queue[i + j].elt; dests[j] = draw->vs.queue[i + j].vertex; } - for ( ; j < 4; j++) { + for ( ; j < MAX_SHADER_VERTICES; j++) { elts[j] = elts[0]; dests[j] = draw->vs.queue[i + j].vertex; } assert(n > 0); - assert(n <= 4); + assert(n <= MAX_SHADER_VERTICES); shader->run(shader, draw, elts, n, dests); } -- cgit v1.2.3 From aadbb1d7fbbaada6e378cb60194e5861cadf98d1 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sat, 12 Apr 2008 15:45:28 -0400 Subject: pass arbitrary number of vertices to the shader execution cycle --- src/gallium/auxiliary/draw/draw_private.h | 2 + src/gallium/auxiliary/draw/draw_vertex_shader.c | 4 +- src/gallium/auxiliary/draw/draw_vs_exec.c | 144 ++++++++++++------------ src/gallium/auxiliary/draw/draw_vs_sse.c | 139 +++++++++++------------ 4 files changed, 146 insertions(+), 143 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 4d056f6dba..f9aea9f355 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -56,6 +56,8 @@ struct gallivm_cpu_engine; struct draw_pt_middle_end; struct draw_pt_front_end; +#define MAX_SHADER_VERTICES 128 + /** * Basic vertex info. * Carry some useful information around with the vertices in the prim pipe. diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index d5f37bca21..726921d77b 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -37,8 +37,6 @@ #include "draw_context.h" #include "draw_vs.h" -#define MAX_SHADER_VERTICES 4 - /** * Run the vertex shader on all vertices in the vertex queue. * Called by the draw module when the vertx cache needs to be flushed. @@ -61,7 +59,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) for (i = 0; i < draw->vs.queue_nr; i += MAX_SHADER_VERTICES) { struct vertex_header *dests[MAX_SHADER_VERTICES]; unsigned elts[MAX_SHADER_VERTICES]; - int j, n = MIN2(MAX_SHADER_VERTICES, - i); + int j, n = MIN2(MAX_SHADER_VERTICES, draw->vs.queue_nr - i); for (j = 0; j < n; j++) { elts[j] = draw->vs.queue[i + j].elt; diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 9629410abb..df0051d693 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -40,6 +40,7 @@ #include "tgsi/util/tgsi_parse.h" +#define MAX_TGSI_VERTICES 4 static void vs_exec_prepare( struct draw_vertex_shader *shader, @@ -71,14 +72,13 @@ vs_exec_run( struct draw_vertex_shader *shader, struct vertex_header *vOut[] ) { struct tgsi_exec_machine *machine = &draw->machine; - unsigned int j; + unsigned int i, j; ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; - assert(count <= 4); assert(draw->vertex_shader->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); @@ -92,80 +92,82 @@ vs_exec_run( struct draw_vertex_shader *shader, machine->Outputs = ALIGN16_ASSIGN(outputs); } - draw->vertex_fetch.fetch_func( draw, machine, elts, count ); + for (i = 0; i < count; i += MAX_TGSI_VERTICES) { + unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); + draw->vertex_fetch.fetch_func( draw, machine, &elts[i], max_vertices ); - if (!draw->rasterizer->bypass_vs) { - /* run interpreter */ - tgsi_exec_machine_run( machine ); - } - - /* store machine results */ - for (j = 0; j < count; j++) { - unsigned slot; - float x, y, z, w; - - /* Handle attr[0] (position) specially: - * - * XXX: Computing the clipmask should be done in the vertex - * program as a set of DP4 instructions appended to the - * user-provided code. - */ - x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - - if (!draw->rasterizer->bypass_clipping) { - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - } - else { - vOut[j]->clipmask = 0; - } - vOut[j]->edgeflag = 1; - - if (!draw->identity_viewport) { - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; - } - else { - vOut[j]->data[0][0] = x; - vOut[j]->data[0][1] = y; - vOut[j]->data[0][2] = z; - vOut[j]->data[0][3] = w; + if (!draw->rasterizer->bypass_vs) { + /* run interpreter */ + tgsi_exec_machine_run( machine ); } - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; - } + /* store machine results */ + for (j = 0; j < max_vertices; j++) { + unsigned slot; + float x, y, z, w; + + /* Handle attr[0] (position) specially: + * + * XXX: Computing the clipmask should be done in the vertex + * program as a set of DP4 instructions appended to the + * user-provided code. + */ + x = vOut[i + j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[i + j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[i + j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[i + j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + if (!draw->rasterizer->bypass_clipping) { + vOut[i + j]->clipmask = compute_clipmask(vOut[i + j]->clip, draw->plane, + draw->nr_planes); + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + } + else { + vOut[i + j]->clipmask = 0; + } + vOut[i + j]->edgeflag = 1; + + if (!draw->identity_viewport) { + /* Viewport mapping */ + vOut[i + j]->data[0][0] = x * scale[0] + trans[0]; + vOut[i + j]->data[0][1] = y * scale[1] + trans[1]; + vOut[i + j]->data[0][2] = z * scale[2] + trans[2]; + vOut[i + j]->data[0][3] = w; + } + else { + vOut[i + j]->data[0][0] = x; + vOut[i + j]->data[0][1] = y; + vOut[i + j]->data[0][2] = z; + vOut[i + j]->data[0][3] = w; + } + + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + vOut[i + j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + vOut[i + j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + vOut[i + j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + vOut[i + j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } #if 0 /*DEBUG*/ - printf("Post xform vert:\n"); - for (slot = 0; slot < draw->num_vs_outputs; slot++) { - printf("%d: %f %f %f %f\n", slot, - vOut[j]->data[slot][0], - vOut[j]->data[slot][1], - vOut[j]->data[slot][2], - vOut[j]->data[slot][3]); - } -#endif - - - } /* loop over vertices */ + printf("%d) Post xform vert:\n", i + j); + for (slot = 0; slot < draw->num_vs_outputs; slot++) { + printf("\t%d: %f %f %f %f\n", slot, + vOut[i + j]->data[slot][0], + vOut[i + j]->data[slot][1], + vOut[i + j]->data[slot][2], + vOut[i + j]->data[slot][3]); + } +#endif + } /* loop over vertices */ + } } diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 0ee991d764..bfec89254e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -45,6 +45,7 @@ #include "tgsi/exec/tgsi_sse2.h" #include "tgsi/util/tgsi_parse.h" +#define SSE_MAX_VERTICES 4 typedef void (XSTDCALL *codegen_function) ( const struct tgsi_exec_vector *input, @@ -86,14 +87,13 @@ vs_sse_run( struct draw_vertex_shader *base, { struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; struct tgsi_exec_machine *machine = &draw->machine; - unsigned int j; + unsigned int i, j; ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; - assert(count <= 4); assert(draw->vertex_shader->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); @@ -108,77 +108,78 @@ vs_sse_run( struct draw_vertex_shader *base, machine->Outputs = ALIGN16_ASSIGN(outputs); } - - /* Fetch vertices. This may at some point be integrated into the - * compiled shader -- that would require a reorganization where - * multiple versions of the compiled shader might exist, - * specialized for each fetch state. - */ - draw->vertex_fetch.fetch_func( draw, machine, elts, count ); - - - if (!draw->rasterizer->bypass_vs) { - /* run compiled shader - */ - shader->func(machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps, - shader->immediates); - } - - - /* XXX: Computing the clipmask and emitting results should be done - * in the vertex program as a set of instructions appended to - * the user-provided code. - */ - for (j = 0; j < count; j++) { - unsigned slot; - float x, y, z, w; - - x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; - - if (!draw->rasterizer->bypass_clipping) { - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - } - else { - vOut[j]->clipmask = 0; - } - vOut[j]->edgeflag = 1; - - if (!draw->identity_viewport) { - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; - } - else { - vOut[j]->data[0][0] = x; - vOut[j]->data[0][1] = y; - vOut[j]->data[0][2] = z; - vOut[j]->data[0][3] = w; + for (i = 0; i < count; i += SSE_MAX_VERTICES) { + unsigned int max_vertices = MIN2(SSE_MAX_VERTICES, count - i); + /* Fetch vertices. This may at some point be integrated into the + * compiled shader -- that would require a reorganization where + * multiple versions of the compiled shader might exist, + * specialized for each fetch state. + */ + draw->vertex_fetch.fetch_func(draw, machine, &elts[i], max_vertices); + + if (!draw->rasterizer->bypass_vs) { + /* run compiled shader + */ + shader->func(machine->Inputs, + machine->Outputs, + machine->Consts, + machine->Temps, + shader->immediates); } - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. + /* XXX: Computing the clipmask and emitting results should be done + * in the vertex program as a set of instructions appended to + * the user-provided code. */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + for (j = 0; j < max_vertices; j++) { + unsigned slot; + float x, y, z, w; + + x = vOut[i + j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = vOut[i + j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = vOut[i + j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = vOut[i + j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + + if (!draw->rasterizer->bypass_clipping) { + vOut[i + j]->clipmask = compute_clipmask(vOut[i + j]->clip, draw->plane, + draw->nr_planes); + + /* divide by w */ + w = 1.0f / w; + x *= w; + y *= w; + z *= w; + } + else { + vOut[i + j]->clipmask = 0; + } + vOut[j]->edgeflag = 1; + + if (!draw->identity_viewport) { + /* Viewport mapping */ + vOut[i + j]->data[0][0] = x * scale[0] + trans[0]; + vOut[i + j]->data[0][1] = y * scale[1] + trans[1]; + vOut[i + j]->data[0][2] = z * scale[2] + trans[2]; + vOut[i + j]->data[0][3] = w; + } + else { + vOut[i + j]->data[0][0] = x; + vOut[i + j]->data[0][1] = y; + vOut[i + j]->data[0][2] = z; + vOut[i + j]->data[0][3] = w; + } + + /* Remaining attributes are packed into sequential post-transform + * vertex attrib slots. + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + vOut[i + j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + vOut[i + j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + vOut[i + j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + vOut[i + j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } } - } + } } -- cgit v1.2.3 From 808f968f3ad0cb32e86f517753d5715d00e9ec2c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 11 Apr 2008 19:58:22 -0400 Subject: return true if one of the vertices has been clipped --- src/gallium/auxiliary/draw/draw_private.h | 12 ++++++------ src/gallium/auxiliary/draw/draw_vs_exec.c | 5 ++++- src/gallium/auxiliary/draw/draw_vs_llvm.c | 18 +++++++++++------- src/gallium/auxiliary/draw/draw_vs_sse.c | 5 ++++- 4 files changed, 25 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index f9aea9f355..5c710667fc 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -148,12 +148,12 @@ struct draw_vertex_shader { /* Run the shader - this interface will get cleaned up in the * future: */ - void (*run)( struct draw_vertex_shader *shader, - struct draw_context *draw, - const unsigned *elts, - unsigned count, - struct vertex_header *vOut[] ); - + boolean (*run)( struct draw_vertex_shader *shader, + struct draw_context *draw, + const unsigned *elts, + unsigned count, + struct vertex_header *vOut[] ); + void (*delete)( struct draw_vertex_shader * ); }; diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index df0051d693..09e0d0eaab 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -64,7 +64,7 @@ vs_exec_prepare( struct draw_vertex_shader *shader, * \param count number of vertices to shade [1..4] * \param vOut array of pointers to four output vertices */ -static void +static boolean vs_exec_run( struct draw_vertex_shader *shader, struct draw_context *draw, const unsigned *elts, @@ -73,6 +73,7 @@ vs_exec_run( struct draw_vertex_shader *shader, { struct tgsi_exec_machine *machine = &draw->machine; unsigned int i, j; + unsigned int clipped = 0; ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); @@ -120,6 +121,7 @@ vs_exec_run( struct draw_vertex_shader *shader, if (!draw->rasterizer->bypass_clipping) { vOut[i + j]->clipmask = compute_clipmask(vOut[i + j]->clip, draw->plane, draw->nr_planes); + clipped += vOut[i + j]->clipmask; /* divide by w */ w = 1.0f / w; @@ -168,6 +170,7 @@ vs_exec_run( struct draw_vertex_shader *shader, #endif } /* loop over vertices */ } + return clipped != 0; } diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 7e48428cdd..6db1e65a2d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -67,18 +67,19 @@ vs_llvm_prepare( struct draw_vertex_shader *base, * \param count number of vertices to shade [1..4] * \param vOut array of pointers to four output vertices */ -static void +static boolean vs_llvm_run( struct draw_vertex_shader *base, - struct draw_context *draw, - const unsigned *elts, + struct draw_context *draw, + const unsigned *elts, unsigned count, struct vertex_header *vOut[] ) { - struct draw_llvm_vertex_shader *shader = + struct draw_llvm_vertex_shader *shader = (struct draw_llvm_vertex_shader *)base; struct tgsi_exec_machine *machine = &draw->machine; unsigned int j; + unsigned int clipped = 0; ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); @@ -125,19 +126,21 @@ vs_llvm_run( struct draw_vertex_shader *base, w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; if (!draw->rasterizer->bypass_clipping) { - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, draw->nr_planes); + vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, + draw->nr_planes); + clipped += vOut[j]->clipmask; /* divide by w */ w = 1.0f / w; x *= w; y *= w; - z *= w; + z *= w; } else { vOut[j]->clipmask = 0; } vOut[j]->edgeflag = 1; - + if (!draw->identity_viewport) { /* Viewport mapping */ vOut[j]->data[0][0] = x * scale[0] + trans[0]; @@ -162,6 +165,7 @@ vs_llvm_run( struct draw_vertex_shader *base, vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; } } /* loop over vertices */ + return clipped != 0; } static void diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index bfec89254e..6e8d2021f5 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -78,7 +78,7 @@ vs_sse_prepare( struct draw_vertex_shader *base, * \param count number of vertices to shade [1..4] * \param vOut array of pointers to four output vertices */ -static void +static boolean vs_sse_run( struct draw_vertex_shader *base, struct draw_context *draw, const unsigned *elts, @@ -88,6 +88,7 @@ vs_sse_run( struct draw_vertex_shader *base, struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; struct tgsi_exec_machine *machine = &draw->machine; unsigned int i, j; + unsigned int clipped = 0; ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); @@ -143,6 +144,7 @@ vs_sse_run( struct draw_vertex_shader *base, if (!draw->rasterizer->bypass_clipping) { vOut[i + j]->clipmask = compute_clipmask(vOut[i + j]->clip, draw->plane, draw->nr_planes); + clipped += vOut[i + j]->clipmask; /* divide by w */ w = 1.0f / w; @@ -180,6 +182,7 @@ vs_sse_run( struct draw_vertex_shader *base, } } } + return clipped != 0; } -- cgit v1.2.3 From 3f7a3dd58c0ce2719af83ff1d89a26185d08c04c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sat, 12 Apr 2008 21:52:46 -0400 Subject: Make shaders operate on a block of memory instead of arrays of vertex_header's --- src/gallium/auxiliary/draw/draw_context.c | 10 ++--- src/gallium/auxiliary/draw/draw_private.h | 15 ++++--- src/gallium/auxiliary/draw/draw_vertex_cache.c | 33 +++++++++------ src/gallium/auxiliary/draw/draw_vertex_shader.c | 7 ++-- src/gallium/auxiliary/draw/draw_vs_exec.c | 54 ++++++++++++------------ src/gallium/auxiliary/draw/draw_vs_llvm.c | 2 +- src/gallium/auxiliary/draw/draw_vs_sse.c | 56 +++++++++++++++---------- 7 files changed, 100 insertions(+), 77 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index b3c65c90d6..cdca556fbd 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -86,14 +86,12 @@ struct draw_context *draw_create( void ) /* Statically allocate maximum sized vertices for the cache - could be cleverer... */ { - uint i; const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; - char *tmp = align_malloc(Elements(draw->vs.queue) * size, 16); + char *tmp = align_malloc(VS_QUEUE_LENGTH * size, 16); if (!tmp) goto fail; - for (i = 0; i < Elements(draw->vs.queue); i++) - draw->vs.queue[i].vertex = (struct vertex_header *)(tmp + i * size); + draw->vs.vertex_cache = tmp; } draw->shader_queue_flush = draw_vertex_shader_queue_flush; @@ -156,8 +154,8 @@ void draw_destroy( struct draw_context *draw ) tgsi_exec_machine_free_data(&draw->machine); - if (draw->vs.queue[0].vertex) - align_free( draw->vs.queue[0].vertex ); /* Frees all the vertices. */ + if (draw->vs.vertex_cache) + align_free( draw->vs.vertex_cache ); /* Frees all the vertices. */ /* Not so fast -- we're just borrowing this at the moment. * diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 5c710667fc..1115166192 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -152,7 +152,7 @@ struct draw_vertex_shader { struct draw_context *draw, const unsigned *elts, unsigned count, - struct vertex_header *vOut[] ); + void *out ); void (*delete)( struct draw_vertex_shader * ); @@ -321,10 +321,8 @@ struct draw_context /* Vertex shader queue: */ struct { - struct { - unsigned elt; /**< index into the user's vertex arrays */ - struct vertex_header *vertex; - } queue[VS_QUEUE_LENGTH]; + unsigned elts[VS_QUEUE_LENGTH]; /**< index into the user's vertex arrays */ + char *vertex_cache; unsigned queue_nr; unsigned post_nr; } vs; @@ -450,4 +448,11 @@ dot4(const float *a, const float *b) return result; } +static INLINE struct vertex_header * +draw_header_from_block(char *block, int num) +{ + static const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; + return (struct vertex_header*)(block + num * size); +} + #endif /* DRAW_PRIVATE_H */ diff --git a/src/gallium/auxiliary/draw/draw_vertex_cache.c b/src/gallium/auxiliary/draw/draw_vertex_cache.c index c0248979e2..9256dec4ea 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_cache.c +++ b/src/gallium/auxiliary/draw/draw_vertex_cache.c @@ -71,25 +71,27 @@ static struct vertex_header *get_vertex( struct draw_context *draw, /* Cache hit? */ if (draw->vcache.idx[slot].in == i) { -// _mesa_printf("HIT %d %d\n", slot, i); + /*debug_printf("HIT %d %d\n", slot, i);*/ assert(draw->vcache.idx[slot].out < draw->vs.queue_nr); - return draw->vs.queue[draw->vcache.idx[slot].out].vertex; + return draw_header_from_block(draw->vs.vertex_cache, + draw->vcache.idx[slot].out); } /* Otherwise a collision */ slot = VCACHE_SIZE + draw->vcache.overflow++; -// _mesa_printf("XXX %d --> %d\n", i, slot); + /*debug_printf("XXX %d --> %d\n", i, slot);*/ } /* Deal with the cache miss: */ { unsigned out; - + struct vertex_header *header; + assert(slot < Elements(draw->vcache.idx)); -// _mesa_printf("NEW %d %d\n", slot, i); + /*debug_printf("NEW %d %d\n", slot, i);*/ draw->vcache.idx[slot].in = i; draw->vcache.idx[slot].out = out = draw->vs.queue_nr++; draw->vcache.referenced |= (1 << slot); @@ -99,17 +101,19 @@ static struct vertex_header *get_vertex( struct draw_context *draw, */ assert(draw->vs.queue_nr < VS_QUEUE_LENGTH); - draw->vs.queue[out].elt = i; - draw->vs.queue[out].vertex->clipmask = 0; - draw->vs.queue[out].vertex->edgeflag = draw_get_edgeflag(draw, i); - draw->vs.queue[out].vertex->pad = 0; - draw->vs.queue[out].vertex->vertex_id = UNDEFINED_VERTEX_ID; + header = draw_header_from_block(draw->vs.vertex_cache, out); + draw->vs.elts[out] = i; + header->clipmask = 0; + header->edgeflag = draw_get_edgeflag(draw, i); + header->pad = 0; + header->vertex_id = UNDEFINED_VERTEX_ID; /* Need to set the vertex's edge flag here. If we're being called * by do_ef_triangle(), that function needs edge flag info! */ - return draw->vs.queue[draw->vcache.idx[slot].out].vertex; + return draw_header_from_block(draw->vs.vertex_cache, + draw->vcache.idx[slot].out); } } @@ -142,8 +146,11 @@ void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) { unsigned i; - for (i = 0; i < draw->vs.post_nr; i++) - draw->vs.queue[i].vertex->vertex_id = UNDEFINED_VERTEX_ID; + for (i = 0; i < draw->vs.post_nr; i++) { + struct vertex_header * header = + draw_header_from_block(draw->vs.vertex_cache, i); + header->vertex_id = UNDEFINED_VERTEX_ID; + } } diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index 726921d77b..452d0175c3 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -57,18 +57,17 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) /* run vertex shader on vertex cache entries, four per invokation */ for (i = 0; i < draw->vs.queue_nr; i += MAX_SHADER_VERTICES) { - struct vertex_header *dests[MAX_SHADER_VERTICES]; unsigned elts[MAX_SHADER_VERTICES]; int j, n = MIN2(MAX_SHADER_VERTICES, draw->vs.queue_nr - i); + struct vertex_header *dests = + draw_header_from_block(draw->vs.vertex_cache, i); for (j = 0; j < n; j++) { - elts[j] = draw->vs.queue[i + j].elt; - dests[j] = draw->vs.queue[i + j].vertex; + elts[j] = draw->vs.elts[i + j]; } for ( ; j < MAX_SHADER_VERTICES; j++) { elts[j] = elts[0]; - dests[j] = draw->vs.queue[i + j].vertex; } assert(n > 0); diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 09e0d0eaab..6fe4e554d5 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -69,7 +69,7 @@ vs_exec_run( struct draw_vertex_shader *shader, struct draw_context *draw, const unsigned *elts, unsigned count, - struct vertex_header *vOut[] ) + void *vOut ) { struct tgsi_exec_machine *machine = &draw->machine; unsigned int i, j; @@ -106,6 +106,8 @@ vs_exec_run( struct draw_vertex_shader *shader, for (j = 0; j < max_vertices; j++) { unsigned slot; float x, y, z, w; + struct vertex_header *out = + draw_header_from_block(vOut, i + j); /* Handle attr[0] (position) specially: * @@ -113,15 +115,15 @@ vs_exec_run( struct draw_vertex_shader *shader, * program as a set of DP4 instructions appended to the * user-provided code. */ - x = vOut[i + j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[i + j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[i + j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[i + j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + x = out->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = out->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = out->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = out->clip[3] = machine->Outputs[0].xyzw[3].f[j]; if (!draw->rasterizer->bypass_clipping) { - vOut[i + j]->clipmask = compute_clipmask(vOut[i + j]->clip, draw->plane, - draw->nr_planes); - clipped += vOut[i + j]->clipmask; + out->clipmask = compute_clipmask(out->clip, draw->plane, + draw->nr_planes); + clipped += out->clipmask; /* divide by w */ w = 1.0f / w; @@ -130,42 +132,42 @@ vs_exec_run( struct draw_vertex_shader *shader, z *= w; } else { - vOut[i + j]->clipmask = 0; + out->clipmask = 0; } - vOut[i + j]->edgeflag = 1; + out->edgeflag = 1; if (!draw->identity_viewport) { /* Viewport mapping */ - vOut[i + j]->data[0][0] = x * scale[0] + trans[0]; - vOut[i + j]->data[0][1] = y * scale[1] + trans[1]; - vOut[i + j]->data[0][2] = z * scale[2] + trans[2]; - vOut[i + j]->data[0][3] = w; + out->data[0][0] = x * scale[0] + trans[0]; + out->data[0][1] = y * scale[1] + trans[1]; + out->data[0][2] = z * scale[2] + trans[2]; + out->data[0][3] = w; } else { - vOut[i + j]->data[0][0] = x; - vOut[i + j]->data[0][1] = y; - vOut[i + j]->data[0][2] = z; - vOut[i + j]->data[0][3] = w; + out->data[0][0] = x; + out->data[0][1] = y; + out->data[0][2] = z; + out->data[0][3] = w; } /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. */ for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[i + j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - vOut[i + j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - vOut[i + j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - vOut[i + j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + out->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + out->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + out->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + out->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; } #if 0 /*DEBUG*/ printf("%d) Post xform vert:\n", i + j); for (slot = 0; slot < draw->num_vs_outputs; slot++) { printf("\t%d: %f %f %f %f\n", slot, - vOut[i + j]->data[slot][0], - vOut[i + j]->data[slot][1], - vOut[i + j]->data[slot][2], - vOut[i + j]->data[slot][3]); + out->data[slot][0], + out->data[slot][1], + out->data[slot][2], + out->data[slot][3]); } #endif } /* loop over vertices */ diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 6db1e65a2d..72317c67ae 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -72,7 +72,7 @@ vs_llvm_run( struct draw_vertex_shader *base, struct draw_context *draw, const unsigned *elts, unsigned count, - struct vertex_header *vOut[] ) + void *vOut ) { struct draw_llvm_vertex_shader *shader = (struct draw_llvm_vertex_shader *)base; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 6e8d2021f5..c877f5ee3a 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -83,7 +83,7 @@ vs_sse_run( struct draw_vertex_shader *base, struct draw_context *draw, const unsigned *elts, unsigned count, - struct vertex_header *vOut[] ) + void *vOut ) { struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; struct tgsi_exec_machine *machine = &draw->machine; @@ -135,16 +135,18 @@ vs_sse_run( struct draw_vertex_shader *base, for (j = 0; j < max_vertices; j++) { unsigned slot; float x, y, z, w; + struct vertex_header *out = + draw_header_from_block(vOut, i + j); - x = vOut[i + j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[i + j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[i + j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[i + j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + x = out->clip[0] = machine->Outputs[0].xyzw[0].f[j]; + y = out->clip[1] = machine->Outputs[0].xyzw[1].f[j]; + z = out->clip[2] = machine->Outputs[0].xyzw[2].f[j]; + w = out->clip[3] = machine->Outputs[0].xyzw[3].f[j]; if (!draw->rasterizer->bypass_clipping) { - vOut[i + j]->clipmask = compute_clipmask(vOut[i + j]->clip, draw->plane, - draw->nr_planes); - clipped += vOut[i + j]->clipmask; + out->clipmask = compute_clipmask(out->clip, draw->plane, + draw->nr_planes); + clipped += out->clipmask; /* divide by w */ w = 1.0f / w; @@ -153,33 +155,43 @@ vs_sse_run( struct draw_vertex_shader *base, z *= w; } else { - vOut[i + j]->clipmask = 0; + out->clipmask = 0; } - vOut[j]->edgeflag = 1; + out->edgeflag = 1; if (!draw->identity_viewport) { /* Viewport mapping */ - vOut[i + j]->data[0][0] = x * scale[0] + trans[0]; - vOut[i + j]->data[0][1] = y * scale[1] + trans[1]; - vOut[i + j]->data[0][2] = z * scale[2] + trans[2]; - vOut[i + j]->data[0][3] = w; + out->data[0][0] = x * scale[0] + trans[0]; + out->data[0][1] = y * scale[1] + trans[1]; + out->data[0][2] = z * scale[2] + trans[2]; + out->data[0][3] = w; } else { - vOut[i + j]->data[0][0] = x; - vOut[i + j]->data[0][1] = y; - vOut[i + j]->data[0][2] = z; - vOut[i + j]->data[0][3] = w; + out->data[0][0] = x; + out->data[0][1] = y; + out->data[0][2] = z; + out->data[0][3] = w; } /* Remaining attributes are packed into sequential post-transform * vertex attrib slots. */ for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[i + j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - vOut[i + j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - vOut[i + j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - vOut[i + j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + out->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + out->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + out->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + out->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; } +#if 0 /*DEBUG*/ + printf("%d) Post xform vert:\n", i + j); + for (slot = 0; slot < draw->num_vs_outputs; slot++) { + printf("\t%d: %f %f %f %f\n", slot, + out->data[slot][0], + out->data[slot][1], + out->data[slot][2], + out->data[slot][3]); + } +#endif } } return clipped != 0; -- cgit v1.2.3 From 0c1cb54923f3ab31caa2821e095685277174dd2f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sun, 13 Apr 2008 01:47:07 -0400 Subject: Implement fetch/shade/pipeline or emit vertex passthrough. --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/draw_pt.c | 54 +++-- src/gallium/auxiliary/draw/draw_pt.h | 2 +- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 253 +++++++++++++++++++++ 4 files changed, 287 insertions(+), 23 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 28262a92c6..5ab3cfe5ce 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -19,6 +19,7 @@ C_SOURCES = \ draw_pt_vcache.c \ draw_pt_fetch_emit.c \ draw_pt_fetch_pipeline.c \ + draw_pt_fetch_shade_pipeline.c \ draw_pt_pipeline.c \ draw_pt_elts.c \ draw_prim.c \ diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index f59fb86f78..c8663c0e84 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -38,19 +38,23 @@ /* XXX: Shouldn't those two functions below use the '>' operator??? */ - -static boolean too_many_verts( struct draw_context *draw, - unsigned verts ) +static boolean too_many_elts( struct draw_context *draw, + unsigned elts ) { - return verts < 1024; + return elts > (8 * 1024); } -static boolean too_many_elts( struct draw_context *draw, - unsigned elts ) +static INLINE unsigned reduced_prim(unsigned prim) { - return elts < (16 * 1024); + /*FIXME*/ + return prim; } +static INLINE boolean good_prim(unsigned prim) +{ + /*FIXME*/ + return FALSE; +} boolean draw_pt_arrays(struct draw_context *draw, @@ -64,6 +68,9 @@ draw_pt_arrays(struct draw_context *draw, struct draw_pt_front_end *frontend = NULL; struct draw_pt_middle_end *middle = NULL; + if (!draw->render) + return FALSE; + /*debug_printf("XXXXXXXXXX needs_pipeline = %d\n", pipeline);*/ /* Overall we do: * - frontend -- prepare fetch_elts, draw_elts - eg vcache @@ -87,7 +94,6 @@ draw_pt_arrays(struct draw_context *draw, */ middle = draw->pt.middle.fetch_pipeline; } -#if 0 else if (!cliptest && !pipeline) { /* Fetch user verts, run vertex shader, emit hw verts: */ @@ -111,23 +117,15 @@ draw_pt_arrays(struct draw_context *draw, */ middle = draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit; } - else if (!cliptest) { - /* Fetch user verts, run vertex shader, run pipeline: - */ - middle = draw->pt.middle.fetch_shade_pipeline; - } else { /* This is what we're currently always doing: */ - /* Fetch user verts, run vertex shader, cliptest, run pipeline: + /* Fetch user verts, run vertex shader, cliptest, run pipeline + * or + * Fetch user verts, run vertex shader, run pipeline */ - middle = draw->pt.middle.fetch_shade_cliptest_pipeline; - } -#else - else { - return FALSE; + middle = draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit; } -#endif /* If !pipeline, need to make sure we respect the driver's limited @@ -143,7 +141,7 @@ draw_pt_arrays(struct draw_context *draw, frontend = draw->pt.front.vcache; hw_prim = reduced_prim(prim); } - +#if 0 if (too_many_verts(nr_verts)) { /* if (is_verts(draw) && can_split(prim)) { draw = draw_arrays_split; @@ -153,6 +151,7 @@ draw_pt_arrays(struct draw_context *draw, hw_prim = reduced_prim(prim); } } +#endif if (too_many_elts(count)) { @@ -166,7 +165,7 @@ draw_pt_arrays(struct draw_context *draw, } if (!good_prim(hw_prim)) { - draw = draw->pt.front.vcache; + frontend = draw->pt.front.vcache; } } #else @@ -200,6 +199,11 @@ boolean draw_pt_init( struct draw_context *draw ) if (!draw->pt.middle.fetch_pipeline) return FALSE; + draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit = + draw_pt_fetch_pipeline_or_emit( draw ); + if (!draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit) + return FALSE; + draw->pt.front.vcache = draw_pt_vcache( draw ); if (!draw->pt.front.vcache) return FALSE; @@ -220,6 +224,12 @@ void draw_pt_destroy( struct draw_context *draw ) draw->pt.middle.fetch_pipeline = NULL; } + if (draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit) { + draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit->destroy( + draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit ); + draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit = NULL; + } + if (draw->pt.front.vcache) { draw->pt.front.vcache->destroy( draw->pt.front.vcache ); draw->pt.front.vcache = NULL; diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 590823fd33..48413b648a 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -123,7 +123,7 @@ const void *draw_pt_elt_ptr( struct draw_context *draw, struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_pipeline( struct draw_context *draw ); - +struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw); #endif diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c new file mode 100644 index 0000000000..e7550fe328 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -0,0 +1,253 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" + +struct fetch_pipeline_middle_end { + struct draw_pt_middle_end base; + struct draw_context *draw; + + struct { + const ubyte *ptr; + unsigned pitch; + void (*fetch)( const void *from, float *attrib); + void (*emit)( const float *attrib, float **out ); + } fetch[PIPE_MAX_ATTRIBS]; + + unsigned nr_fetch; + unsigned pipeline_vertex_size; + unsigned hw_vertex_size; + unsigned prim; +}; + +static void emit_R32_FLOAT( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out) += 1; +} + +static void emit_R32G32_FLOAT( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out)[1] = attrib[1]; + (*out) += 2; +} + +static void emit_R32G32B32_FLOAT( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out)[1] = attrib[1]; + (*out)[2] = attrib[2]; + (*out) += 3; +} + +static void emit_R32G32B32A32_FLOAT( const float *attrib, + float **out ) +{ + (*out)[0] = attrib[0]; + (*out)[1] = attrib[1]; + (*out)[2] = attrib[2]; + (*out)[3] = attrib[3]; + (*out) += 4; +} + +static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, + unsigned prim ) +{ + struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; + struct draw_context *draw = fpme->draw; + unsigned i, nr = 0; + boolean ok; + const struct vertex_info *vinfo; + + fpme->prim = prim; + + ok = draw->render->set_primitive(draw->render, prim); + if (!ok) { + assert(0); + return; + } + /* Must do this after set_primitive() above: + */ + vinfo = draw->render->get_vertex_info(draw->render); + + /* Need to look at vertex shader inputs (we know it is a + * passthrough shader, so these define the outputs too). If we + * were running a shader, we'd still be looking at the inputs at + * this point. + */ + for (i = 0; i < draw->vertex_shader->info.num_inputs; i++) { + unsigned buf = draw->vertex_element[i].vertex_buffer_index; + enum pipe_format format = draw->vertex_element[i].src_format; + + fpme->fetch[nr].ptr = ((const ubyte *) draw->user.vbuffer[buf] + + draw->vertex_buffer[buf].buffer_offset + + draw->vertex_element[i].src_offset); + + fpme->fetch[nr].pitch = draw->vertex_buffer[buf].pitch; + fpme->fetch[nr].fetch = draw_get_fetch_func( format ); + + /* Always do this -- somewhat redundant... + */ + fpme->fetch[nr].emit = emit_R32G32B32A32_FLOAT; + nr++; + } + + fpme->nr_fetch = nr; + //fpme->pipeline_vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); + fpme->pipeline_vertex_size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; + fpme->hw_vertex_size = vinfo->size * 4; +} + + + + +static void fetch_pipeline_run( struct draw_pt_middle_end *middle, + const unsigned *fetch_elts, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; + struct draw_context *draw = fpme->draw; + struct draw_vertex_shader *shader = draw->vertex_shader; + char *pipeline_verts; + unsigned i; + + //debug_printf("fc = %d, VS = %d\n", fetch_count, VS_QUEUE_LENGTH); + if (fetch_count < VS_QUEUE_LENGTH) { + pipeline_verts = draw->vs.vertex_cache; + } else { + pipeline_verts = MALLOC(fpme->pipeline_vertex_size * + fetch_count); + } + + if (!pipeline_verts) { + assert(0); + return; + } + + /*FIXME: this init phase should go away */ + for (i = 0; i < fetch_count; ++i) { + struct vertex_header *header = + (struct vertex_header*)(pipeline_verts + (fpme->pipeline_vertex_size * i)); + header->clipmask = 0; + header->edgeflag = draw_get_edgeflag(draw, i); + header->pad = 0; + header->vertex_id = UNDEFINED_VERTEX_ID; + } + + /* Shade + */ + shader->prepare(shader, draw); + if (shader->run(shader, draw, fetch_elts, fetch_count, pipeline_verts)) { + /* Run the pipeline */ + draw_pt_run_pipeline( fpme->draw, + fpme->prim, + pipeline_verts, + fpme->pipeline_vertex_size, + fetch_count, + draw_elts, + draw_count ); + } else { + unsigned i, j; + void *hw_verts; + float *out; + + hw_verts = draw->render->allocate_vertices(draw->render, + (ushort)fpme->hw_vertex_size, + (ushort)fetch_count); + if (!hw_verts) { + assert(0); + return; + } + + out = (float *)hw_verts; + for (i = 0; i < fetch_count; i++) { + struct vertex_header *header = + (struct vertex_header*)(pipeline_verts + (fpme->pipeline_vertex_size * i)); + + for (j = 0; j < fpme->nr_fetch; j++) { + float *attrib = header->data[j]; + /*debug_printf("emiting [%f, %f, %f, %f]\n", + attrib[0], attrib[1], + attrib[2], attrib[3]);*/ + fpme->fetch[j].emit(attrib, &out); + } + } + /* XXX: Draw arrays path to avoid re-emitting index list again and + * again. + */ + draw->render->draw(draw->render, + draw_elts, + draw_count); + + draw->render->release_vertices(draw->render, + hw_verts, + fpme->hw_vertex_size, + fetch_count); + } + + + if (pipeline_verts != draw->vs.vertex_cache) + FREE(pipeline_verts); +} + + + +static void fetch_pipeline_finish( struct draw_pt_middle_end *middle ) +{ + /* nothing to do */ +} + +static void fetch_pipeline_destroy( struct draw_pt_middle_end *middle ) +{ + FREE(middle); +} + + +struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit( struct draw_context *draw ) +{ + struct fetch_pipeline_middle_end *fetch_pipeline = CALLOC_STRUCT( fetch_pipeline_middle_end ); + + fetch_pipeline->base.prepare = fetch_pipeline_prepare; + fetch_pipeline->base.run = fetch_pipeline_run; + fetch_pipeline->base.finish = fetch_pipeline_finish; + fetch_pipeline->base.destroy = fetch_pipeline_destroy; + + fetch_pipeline->draw = draw; + + return &fetch_pipeline->base; +} -- cgit v1.2.3 From 8e7326832a7420154fc0d526ac682494db1be160 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 14 Apr 2008 11:32:50 +0100 Subject: softpipe: do our own culling, don't rely on the draw module. May not always happen due to passthrough modes, etc. --- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 21 +++++++++++--- src/gallium/drivers/softpipe/sp_setup.c | 44 +++++++++++++++++------------ 2 files changed, 43 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 025a0113bf..74cd675908 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -120,8 +120,9 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) /** - * Recalculate prim's determinant. - * XXX is this needed? + * Recalculate prim's determinant. This is needed as we don't have + * get this information through the vbuf_render interface & we must + * calculate it here. */ static float calc_det( const float (*v0)[4], @@ -144,12 +145,21 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; - struct draw_stage *setup = softpipe->setup; unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); unsigned i, j; void *vertex_buffer = cvbr->vertex_buffer; cptrf4 v[3]; - struct setup_context *setup_ctx = sp_draw_setup_context(setup); + + /* XXX: break this dependency - make setup_context live under + * softpipe, rename the old "setup" draw stage to something else. + */ + struct draw_stage *setup = softpipe->setup; + struct setup_context *setup_ctx = sp_draw_setup_context(softpipe->setup); + + /* XXX: call this from allocate_vertices: + */ + setup_prepare( setup_ctx ); + switch (cvbr->prim) { case PIPE_PRIM_TRIANGLES: @@ -189,6 +199,9 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) break; } + /* XXX: why are we calling this??? If we had to call something, it + * would be a function in sp_setup.c: + */ sp_draw_flush( setup ); } diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 48617a66ed..5a30788850 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -97,31 +97,35 @@ struct setup_context { uint numFragsEmitted; /**< per primitive */ uint numFragsWritten; /**< per primitive */ #endif + + unsigned winding; /* which winding to cull */ }; -/** - * Recalculate prim's determinant. - * XXX is this needed? - */ -static INLINE float -calc_det( const float (*v0)[4], - const float (*v1)[4], - const float (*v2)[4] ) +static boolean cull_tri( struct setup_context *setup, + float det ) { - /* edge vectors e = v0 - v2, f = v1 - v2 */ - const float ex = v0[0][0] - v2[0][0]; - const float ey = v0[0][1] - v2[0][1]; - const float fx = v1[0][0] - v2[0][0]; - const float fy = v1[0][1] - v2[0][1]; - - /* det = cross(e,f).z */ - return ex * fy - ey * fx; + if (det != 0) + { + /* if (det < 0 then Z points toward camera and triangle is + * counter-clockwise winding. + */ + unsigned winding = (det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; + + if ((winding & setup->winding) == 0) + return FALSE; + } + + /* Culled: + */ + return TRUE; } + + /** * Clip setup->quad against the scissor/surface bounds. */ @@ -709,8 +713,10 @@ void setup_tri( struct setup_context *setup, setup->numFragsWritten = 0; #endif - setup_sort_vertices( setup, calc_det(v0, v1, v2), - v0, v1, v2 ); + if (cull_tri( setup, det )) + return; + + setup_sort_vertices( setup, det, v0, v1, v2 ); setup_tri_coefficients( setup ); setup_tri_edges( setup ); @@ -1223,6 +1229,8 @@ void setup_prepare( struct setup_context *setup ) setup->quad.nr_attrs = fs->info.num_inputs; sp->quad.first->begin(sp->quad.first); } + + setup->winding = sp->rasterizer->cull_mode; } -- cgit v1.2.3 From caf293343fd236e97ce399533ac0ada3c7afee7a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 14 Apr 2008 12:08:46 +0100 Subject: draw: hide passthrough shading paths behind an environment variable --- src/gallium/auxiliary/draw/draw_context.c | 2 ++ src/gallium/auxiliary/draw/draw_private.h | 1 + src/gallium/auxiliary/draw/draw_pt.c | 3 +++ 3 files changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index cdca556fbd..1e70a77523 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -49,6 +49,8 @@ struct draw_context *draw_create( void ) draw->use_sse = FALSE; #endif + draw->use_pt_shaders = GETENV( "GALLIUM_PT_SHADERS" ) != NULL; + /* create pipeline stages */ draw->pipeline.wide_line = draw_wide_line_stage( draw ); draw->pipeline.wide_point = draw_wide_point_stage( draw ); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 1115166192..3042d26847 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -276,6 +276,7 @@ struct draw_context boolean line_stipple; /**< do line stipple? */ boolean point_sprite; /**< convert points to quads for sprites? */ boolean use_sse; + boolean use_pt_shaders; /* temporary flag to switch on pt shader paths */ /* If a prim stage introduces new vertex attributes, they'll be stored here */ diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index c8663c0e84..c67a217982 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -79,6 +79,9 @@ draw_pt_arrays(struct draw_context *draw, * - backend -- the vbuf_render provided by the driver. */ + if (shading && !draw->use_pt_shaders) + return FALSE; + if (!cliptest && !pipeline && !shading) { /* This is the 'passthrough' path: -- cgit v1.2.3 From e106b2d3d65585b0aaa0e60afd541da020d9e220 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 14 Apr 2008 12:27:25 +0100 Subject: draw: move vertex header init out of fetch_shade_pipeline.c --- src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c | 2 -- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 9 --------- src/gallium/auxiliary/draw/draw_vs_exec.c | 1 + src/gallium/auxiliary/draw/draw_vs_llvm.c | 1 + src/gallium/auxiliary/draw/draw_vs_sse.c | 1 + 5 files changed, 3 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c index 4c2a281b29..0914a90440 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c @@ -211,7 +211,6 @@ fetch_store_general( struct fetch_pipeline_middle_end *fpme, static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, unsigned prim ) { - static const float zero = 0; struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; unsigned i, nr = 0; @@ -264,7 +263,6 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, unsigned draw_count ) { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; - struct draw_context *draw = fpme->draw; char *pipeline_verts; pipeline_verts = MALLOC( fpme->pipeline_vertex_size * diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index e7550fe328..fb9b2da5c0 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -159,15 +159,6 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, return; } - /*FIXME: this init phase should go away */ - for (i = 0; i < fetch_count; ++i) { - struct vertex_header *header = - (struct vertex_header*)(pipeline_verts + (fpme->pipeline_vertex_size * i)); - header->clipmask = 0; - header->edgeflag = draw_get_edgeflag(draw, i); - header->pad = 0; - header->vertex_id = UNDEFINED_VERTEX_ID; - } /* Shade */ diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 6fe4e554d5..27cf060cc9 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -135,6 +135,7 @@ vs_exec_run( struct draw_vertex_shader *shader, out->clipmask = 0; } out->edgeflag = 1; + out->vertex_id = UNDEFINED_VERTEX_ID; if (!draw->identity_viewport) { /* Viewport mapping */ diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 72317c67ae..73076d2467 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -140,6 +140,7 @@ vs_llvm_run( struct draw_vertex_shader *base, vOut[j]->clipmask = 0; } vOut[j]->edgeflag = 1; + vOut[j]->vertex_id = UNDEFINED_VERTEX_ID; if (!draw->identity_viewport) { /* Viewport mapping */ diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index c877f5ee3a..92b9947e9f 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -158,6 +158,7 @@ vs_sse_run( struct draw_vertex_shader *base, out->clipmask = 0; } out->edgeflag = 1; + out->vertex_id = UNDEFINED_VERTEX_ID; if (!draw->identity_viewport) { /* Viewport mapping */ -- cgit v1.2.3 From a82e4996a13ef3cae1497fef95c2fca7631cd889 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 14 Apr 2008 12:32:53 +0100 Subject: draw: flush pipeline before trying to allocate more hw vertices --- src/gallium/auxiliary/draw/draw_pt.c | 4 ---- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 4 ++++ src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 4 ++++ 3 files changed, 8 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index c67a217982..5c16165c15 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -175,10 +175,6 @@ draw_pt_arrays(struct draw_context *draw, frontend = draw->pt.front.vcache; #endif - /* XXX: need to flush to get prim_vbuf.c to release its allocation?? - */ - draw_do_flush( draw, DRAW_FLUSH_BACKEND ); - frontend->prepare( frontend, prim, middle ); frontend->run( frontend, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index e4e144ef19..2f8b08db79 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -243,6 +243,10 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, struct draw_context *draw = feme->draw; void *hw_verts; + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + hw_verts = draw->render->allocate_vertices( draw->render, (ushort)feme->hw_vertex_size, (ushort)fetch_count ); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index fb9b2da5c0..b5d2f81a14 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -177,6 +177,10 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, void *hw_verts; float *out; + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + hw_verts = draw->render->allocate_vertices(draw->render, (ushort)fpme->hw_vertex_size, (ushort)fetch_count); -- cgit v1.2.3 From 36bacf97a6b10f7274f0d3fcf37bf7ebf9388161 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 14 Apr 2008 12:46:47 +0100 Subject: draw: always malloc verts for fetch_shade_pipeline --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 13 +++---------- 1 file changed, 3 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index b5d2f81a14..427128f335 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -144,15 +144,9 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, struct draw_context *draw = fpme->draw; struct draw_vertex_shader *shader = draw->vertex_shader; char *pipeline_verts; - unsigned i; - //debug_printf("fc = %d, VS = %d\n", fetch_count, VS_QUEUE_LENGTH); - if (fetch_count < VS_QUEUE_LENGTH) { - pipeline_verts = draw->vs.vertex_cache; - } else { - pipeline_verts = MALLOC(fpme->pipeline_vertex_size * - fetch_count); - } + pipeline_verts = MALLOC(fpme->pipeline_vertex_size * + fetch_count); if (!pipeline_verts) { assert(0); @@ -216,8 +210,7 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, } - if (pipeline_verts != draw->vs.vertex_cache) - FREE(pipeline_verts); + FREE(pipeline_verts); } -- cgit v1.2.3 From ca750dd045893ca5a9e1c33bd31528ade7bb1009 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 14 Apr 2008 17:02:02 +0200 Subject: Removed intel_ioctl it wasn't needed --- src/gallium/winsys/dri/intel/intel_batchbuffer.c | 2 +- src/gallium/winsys/dri/intel/intel_ioctl.c | 138 ----------------------- src/gallium/winsys/dri/intel/intel_ioctl.h | 40 ------- 3 files changed, 1 insertion(+), 179 deletions(-) delete mode 100644 src/gallium/winsys/dri/intel/intel_ioctl.c delete mode 100644 src/gallium/winsys/dri/intel/intel_ioctl.h (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c index 3a8ff5a4eb..3a8591b580 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.c +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.c @@ -26,7 +26,7 @@ **************************************************************************/ #include "intel_batchbuffer.h" -#include "intel_ioctl.h" +#include "intel_context.h" #include #if 0 diff --git a/src/gallium/winsys/dri/intel/intel_ioctl.c b/src/gallium/winsys/dri/intel/intel_ioctl.c deleted file mode 100644 index 3250c6b3a9..0000000000 --- a/src/gallium/winsys/dri/intel/intel_ioctl.c +++ /dev/null @@ -1,138 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include -#include -#include -#include - -#include "mtypes.h" -#include "context.h" -#include "swrast/swrast.h" - -#include "intel_context.h" -#include "intel_ioctl.h" -#include "intel_batchbuffer.h" -#include "intel_blit.h" -#include "intel_regions.h" -#include "drm.h" - -#define FILE_DEBUG_FLAG DEBUG_IOCTL - -int -intelEmitIrqLocked(struct intel_context *intel) -{ - drmI830IrqEmit ie; - int ret, seq; - - assert(((*(int *) intel->driHwLock) & ~DRM_LOCK_CONT) == - (DRM_LOCK_HELD | intel->hHWContext)); - - ie.irq_seq = &seq; - - ret = drmCommandWriteRead(intel->driFd, DRM_I830_IRQ_EMIT, - &ie, sizeof(ie)); - if (ret) { - fprintf(stderr, "%s: drmI830IrqEmit: %d\n", __FUNCTION__, ret); - exit(1); - } - - DBG("%s --> %d\n", __FUNCTION__, seq); - - return seq; -} - -void -intelWaitIrq(struct intel_context *intel, int seq) -{ - int ret; - - DBG("%s %d\n", __FUNCTION__, seq); - - intel->iw.irq_seq = seq; - - do { - ret = - drmCommandWrite(intel->driFd, DRM_I830_IRQ_WAIT, &intel->iw, - sizeof(intel->iw)); - } while (ret == -EAGAIN || ret == -EINTR); - - if (ret) { - fprintf(stderr, "%s: drmI830IrqWait: %d\n", __FUNCTION__, ret); - exit(1); - } -} - - -void -intel_batch_ioctl(struct intel_context *intel, - GLuint start_offset, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) -{ - drmI830BatchBuffer batch; - - assert(intel->locked); - assert(used); - - DBG("%s used %d offset %x..%x ignore_cliprects %d\n", - __FUNCTION__, - used, start_offset, start_offset + used, ignore_cliprects); - - /* Throw away non-effective packets. Won't work once we have - * hardware contexts which would preserve statechanges beyond a - * single buffer. - */ - - - - batch.start = start_offset; - batch.used = used; - batch.cliprects = intel->pClipRects; - batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; - batch.DR1 = 0; - batch.DR4 = ((((GLuint) intel->drawX) & 0xffff) | - (((GLuint) intel->drawY) << 16)); - - DBG("%s: 0x%x..0x%x DR4: %x cliprects: %d\n", - __FUNCTION__, - batch.start, - batch.start + batch.used * 4, batch.DR4, batch.num_cliprects); - - if (drmCommandWrite(intel->driFd, DRM_I830_BATCHBUFFER, &batch, - sizeof(batch))) { - fprintf(stderr, "DRM_I830_BATCHBUFFER: %d\n", -errno); - UNLOCK_HARDWARE(intel); - exit(1); - } - - /* FIXME: use hardware contexts to avoid 'losing' hardware after - * each buffer flush. - */ - intel->vtbl.lost_hardware(intel); -} diff --git a/src/gallium/winsys/dri/intel/intel_ioctl.h b/src/gallium/winsys/dri/intel/intel_ioctl.h deleted file mode 100644 index e8d07de893..0000000000 --- a/src/gallium/winsys/dri/intel/intel_ioctl.h +++ /dev/null @@ -1,40 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_IOCTL_H -#define INTEL_IOCTL_H - -#include "intel_context.h" - -void intelWaitIrq(struct intel_context *intel, int seq); -int intelEmitIrqLocked(struct intel_context *intel); - -void intel_batch_ioctl(struct intel_context *intel, - GLuint start_offset, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock); -#endif -- cgit v1.2.3 From 8cbda9f1088718e5dbb97b9a6ddcc43737f94351 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 14 Apr 2008 16:15:39 +0100 Subject: draw: remove dead code --- src/gallium/auxiliary/draw/draw_vf.h | 4 ---- 1 file changed, 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vf.h b/src/gallium/auxiliary/draw/draw_vf.h index 7555d1bd58..0ef98d6257 100644 --- a/src/gallium/auxiliary/draw/draw_vf.h +++ b/src/gallium/auxiliary/draw/draw_vf.h @@ -128,9 +128,6 @@ draw_vf_destroy( struct draw_vertex_fetch *vf ); struct draw_vf_attr; -typedef void (*draw_vf_extract_func)( const struct draw_vf_attr *a, - float *out, - const uint8_t *v ); typedef void (*draw_vf_insert_func)( const struct draw_vf_attr *a, uint8_t *v, @@ -164,7 +161,6 @@ struct draw_vf_attr uint8_t *inputptr; const draw_vf_insert_func *insert; draw_vf_insert_func do_insert; - draw_vf_extract_func extract; }; struct draw_vertex_fetch -- cgit v1.2.3 From 871d39ec8c168fa58d8758013e99da63fa58111d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 14 Apr 2008 16:18:00 +0100 Subject: softpipe: calculate determinant for all triangles, don't rely on draw module to do it --- src/gallium/drivers/softpipe/sp_prim_setup.c | 1 - src/gallium/drivers/softpipe/sp_prim_vbuf.c | 28 ---------------------------- src/gallium/drivers/softpipe/sp_setup.c | 26 +++++++++++++++++++++++++- src/gallium/drivers/softpipe/sp_setup.h | 10 +++++----- 4 files changed, 30 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 6fe463b74c..0ddb06764a 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -71,7 +71,6 @@ do_tri(struct draw_stage *stage, struct prim_header *prim) struct setup_stage *setup = setup_stage( stage ); setup_tri( setup->setup, - prim->det, prim->v[0]->data, prim->v[1]->data, prim->v[2]->data ); diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 74cd675908..4fed19ecb6 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -119,25 +119,6 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) } -/** - * Recalculate prim's determinant. This is needed as we don't have - * get this information through the vbuf_render interface & we must - * calculate it here. - */ -static float -calc_det( const float (*v0)[4], - const float (*v1)[4], - const float (*v2)[4] ) -{ - /* edge vectors e = v0 - v2, f = v1 - v2 */ - const float ex = v0[0][0] - v2[0][0]; - const float ey = v0[0][1] - v2[0][1]; - const float fx = v1[0][0] - v2[0][0]; - const float fy = v1[0][1] - v2[0][1]; - - /* det = cross(e,f).z */ - return ex * fy - ey * fx; -} static void @@ -169,7 +150,6 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) indices[i+j] * vertex_size); setup_tri( setup_ctx, - calc_det(v[0], v[1], v[2]), v[0], v[1], v[2]); @@ -254,7 +234,6 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) v[1] = VERTEX(i + 1); v[2] = VERTEX(i + 2); setup_tri( setup_ctx, - calc_det(v[0], v[1], v[2]), v[0], v[1], v[2] ); @@ -267,7 +246,6 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) v[1] = VERTEX(i - 1); v[2] = VERTEX(i); setup_tri( setup_ctx, - calc_det(v[0], v[1], v[2]), v[0], v[1], v[2] ); @@ -280,7 +258,6 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) v[1] = VERTEX(i - 1); v[2] = VERTEX(i); setup_tri( setup_ctx, - calc_det(v[0], v[1], v[2]), v[0], v[1], v[2] ); @@ -293,7 +270,6 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) v[1] = VERTEX(i + 1); v[2] = VERTEX(i + 2); setup_tri( setup_ctx, - calc_det(v[0], v[1], v[2]), v[0], v[1], v[2] ); @@ -302,7 +278,6 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) v[1] = VERTEX(i + 2); v[2] = VERTEX(i + 3); setup_tri( setup_ctx, - calc_det(v[0], v[1], v[2]), v[0], v[1], v[2] ); @@ -315,7 +290,6 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) v[1] = VERTEX(i); v[2] = VERTEX(i + 1); setup_tri( setup_ctx, - calc_det(v[0], v[1], v[2]), v[0], v[1], v[2] ); @@ -324,7 +298,6 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) v[1] = VERTEX(i + 1); v[2] = VERTEX(i - 1); setup_tri( setup_ctx, - calc_det(v[0], v[1], v[2]), v[0], v[1], v[2] ); @@ -337,7 +310,6 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) v[1] = VERTEX(i - 1); v[2] = VERTEX(i); setup_tri( setup_ctx, - calc_det(v[0], v[1], v[2]), v[0], v[1], v[2] ); diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 5a30788850..0164d0588d 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -695,15 +695,37 @@ static void subtriangle( struct setup_context *setup, } +/** + * Recalculate prim's determinant. This is needed as we don't have + * get this information through the vbuf_render interface & we must + * calculate it here. + */ +static float +calc_det( const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ) +{ + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0][0] - v2[0][0]; + const float ey = v0[0][1] - v2[0][1]; + const float fx = v1[0][0] - v2[0][0]; + const float fy = v1[0][1] - v2[0][1]; + + /* det = cross(e,f).z */ + return ex * fy - ey * fx; +} + + /** * Do setup for triangle rasterization, then render the triangle. */ void setup_tri( struct setup_context *setup, - float det, const float (*v0)[4], const float (*v1)[4], const float (*v2)[4] ) { + float det = calc_det(v0, v1, v2); + /* debug_printf("%s\n", __FUNCTION__ ); */ @@ -713,6 +735,8 @@ void setup_tri( struct setup_context *setup, setup->numFragsWritten = 0; #endif + + if (cull_tri( setup, det )) return; diff --git a/src/gallium/drivers/softpipe/sp_setup.h b/src/gallium/drivers/softpipe/sp_setup.h index 3133fc2a3d..d54f334428 100644 --- a/src/gallium/drivers/softpipe/sp_setup.h +++ b/src/gallium/drivers/softpipe/sp_setup.h @@ -30,11 +30,11 @@ struct setup_context; struct softpipe_context; -void setup_tri( struct setup_context *setup, - float det, - const float (*v0)[4], - const float (*v1)[4], - const float (*v2)[4] ); +void +setup_tri( struct setup_context *setup, + const float (*v0)[4], + const float (*v1)[4], + const float (*v2)[4] ); void setup_line(struct setup_context *setup, -- cgit v1.2.3 From e3309197855b5caf7c4c167d1e7beedf33ed2fdd Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 14 Apr 2008 12:27:24 -0400 Subject: pass vertex size to shaders so that callee can decide on the size of the vertices and not always have to use the maximum vertex allocation size for them --- src/gallium/auxiliary/draw/draw_private.h | 7 ++++--- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 3 ++- src/gallium/auxiliary/draw/draw_vertex_cache.c | 8 ++++++-- src/gallium/auxiliary/draw/draw_vertex_shader.c | 5 +++-- src/gallium/auxiliary/draw/draw_vs_exec.c | 5 +++-- src/gallium/auxiliary/draw/draw_vs_sse.c | 5 +++-- 6 files changed, 21 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 3042d26847..c8cb96c8ba 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -78,6 +78,7 @@ struct vertex_header { /* XXX This is too large */ #define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) +#define MAX_VERTEX_ALLOCATION ((MAX_VERTEX_SIZE + 0x0f) & ~0x0f) @@ -152,7 +153,8 @@ struct draw_vertex_shader { struct draw_context *draw, const unsigned *elts, unsigned count, - void *out ); + void *out, + unsigned vertex_size); void (*delete)( struct draw_vertex_shader * ); @@ -450,9 +452,8 @@ dot4(const float *a, const float *b) } static INLINE struct vertex_header * -draw_header_from_block(char *block, int num) +draw_header_from_block(char *block, int size, int num) { - static const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; return (struct vertex_header*)(block + num * size); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 427128f335..3fa6dcc5e7 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -157,7 +157,8 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, /* Shade */ shader->prepare(shader, draw); - if (shader->run(shader, draw, fetch_elts, fetch_count, pipeline_verts)) { + if (shader->run(shader, draw, fetch_elts, fetch_count, pipeline_verts, + fpme->pipeline_vertex_size)) { /* Run the pipeline */ draw_pt_run_pipeline( fpme->draw, fpme->prim, diff --git a/src/gallium/auxiliary/draw/draw_vertex_cache.c b/src/gallium/auxiliary/draw/draw_vertex_cache.c index 9256dec4ea..730c18bcb3 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_cache.c +++ b/src/gallium/auxiliary/draw/draw_vertex_cache.c @@ -74,6 +74,7 @@ static struct vertex_header *get_vertex( struct draw_context *draw, /*debug_printf("HIT %d %d\n", slot, i);*/ assert(draw->vcache.idx[slot].out < draw->vs.queue_nr); return draw_header_from_block(draw->vs.vertex_cache, + MAX_VERTEX_ALLOCATION, draw->vcache.idx[slot].out); } @@ -101,7 +102,8 @@ static struct vertex_header *get_vertex( struct draw_context *draw, */ assert(draw->vs.queue_nr < VS_QUEUE_LENGTH); - header = draw_header_from_block(draw->vs.vertex_cache, out); + header = draw_header_from_block(draw->vs.vertex_cache, MAX_VERTEX_ALLOCATION, + out); draw->vs.elts[out] = i; header->clipmask = 0; header->edgeflag = draw_get_edgeflag(draw, i); @@ -113,6 +115,7 @@ static struct vertex_header *get_vertex( struct draw_context *draw, */ return draw_header_from_block(draw->vs.vertex_cache, + MAX_VERTEX_ALLOCATION, draw->vcache.idx[slot].out); } } @@ -148,7 +151,8 @@ void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) for (i = 0; i < draw->vs.post_nr; i++) { struct vertex_header * header = - draw_header_from_block(draw->vs.vertex_cache, i); + draw_header_from_block(draw->vs.vertex_cache, + MAX_VERTEX_ALLOCATION, i); header->vertex_id = UNDEFINED_VERTEX_ID; } } diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index 452d0175c3..8572a6d40c 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -60,7 +60,8 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) unsigned elts[MAX_SHADER_VERTICES]; int j, n = MIN2(MAX_SHADER_VERTICES, draw->vs.queue_nr - i); struct vertex_header *dests = - draw_header_from_block(draw->vs.vertex_cache, i); + draw_header_from_block(draw->vs.vertex_cache, + MAX_VERTEX_ALLOCATION, i); for (j = 0; j < n; j++) { elts[j] = draw->vs.elts[i + j]; @@ -73,7 +74,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) assert(n > 0); assert(n <= MAX_SHADER_VERTICES); - shader->run(shader, draw, elts, n, dests); + shader->run(shader, draw, elts, n, dests, MAX_VERTEX_ALLOCATION); } draw->vs.post_nr = draw->vs.queue_nr; diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 27cf060cc9..5c88c2e24e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -69,7 +69,8 @@ vs_exec_run( struct draw_vertex_shader *shader, struct draw_context *draw, const unsigned *elts, unsigned count, - void *vOut ) + void *vOut, + unsigned vertex_size) { struct tgsi_exec_machine *machine = &draw->machine; unsigned int i, j; @@ -107,7 +108,7 @@ vs_exec_run( struct draw_vertex_shader *shader, unsigned slot; float x, y, z, w; struct vertex_header *out = - draw_header_from_block(vOut, i + j); + draw_header_from_block(vOut, vertex_size, i + j); /* Handle attr[0] (position) specially: * diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 92b9947e9f..ee0a3105b9 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -83,7 +83,8 @@ vs_sse_run( struct draw_vertex_shader *base, struct draw_context *draw, const unsigned *elts, unsigned count, - void *vOut ) + void *vOut, + unsigned vertex_size ) { struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; struct tgsi_exec_machine *machine = &draw->machine; @@ -136,7 +137,7 @@ vs_sse_run( struct draw_vertex_shader *base, unsigned slot; float x, y, z, w; struct vertex_header *out = - draw_header_from_block(vOut, i + j); + draw_header_from_block(vOut, vertex_size, i + j); x = out->clip[0] = machine->Outputs[0].xyzw[0].f[j]; y = out->clip[1] = machine->Outputs[0].xyzw[1].f[j]; -- cgit v1.2.3 From 2ba6e1fa71be07a2d75abe2d085d485046c0932b Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 14 Apr 2008 12:29:23 -0400 Subject: silence some warnings --- src/gallium/auxiliary/draw/draw_pt.c | 4 ++-- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 2 -- src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c | 4 ++-- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 3 ++- 4 files changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 5c16165c15..3d2e7bf7b8 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -36,13 +36,13 @@ #include "draw/draw_pt.h" -/* XXX: Shouldn't those two functions below use the '>' operator??? - */ +#if 0 static boolean too_many_elts( struct draw_context *draw, unsigned elts ) { return elts > (8 * 1024); } +#endif static INLINE unsigned reduced_prim(unsigned prim) { diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 2f8b08db79..3a26a5d712 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -145,7 +145,6 @@ fetch_store_general( struct fetch_emit_middle_end *feme, unsigned count ) { float *out = (float *)out_ptr; - struct vbuf_render *render = feme->draw->render; uint i, j; for (i = 0; i < count; i++) { @@ -167,7 +166,6 @@ fetch_store_general( struct fetch_emit_middle_end *feme, static void fetch_emit_prepare( struct draw_pt_middle_end *middle, unsigned prim ) { - static const float zero = 0; struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; struct draw_context *draw = feme->draw; const struct vertex_info *vinfo; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c index 0914a90440..a70d129c93 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c @@ -87,7 +87,7 @@ struct fetch_pipeline_middle_end { }; - +#if 0 static void emit_R32_FLOAT( const float *attrib, float **out ) { @@ -111,7 +111,7 @@ static void emit_R32G32B32_FLOAT( const float *attrib, (*out)[2] = attrib[2]; (*out) += 3; } - +#endif static void emit_R32G32B32A32_FLOAT( const float *attrib, float **out ) { diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 3fa6dcc5e7..b49c9efa65 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -49,6 +49,7 @@ struct fetch_pipeline_middle_end { unsigned prim; }; +#if 0 static void emit_R32_FLOAT( const float *attrib, float **out ) { @@ -72,7 +73,7 @@ static void emit_R32G32B32_FLOAT( const float *attrib, (*out)[2] = attrib[2]; (*out) += 3; } - +#endif static void emit_R32G32B32A32_FLOAT( const float *attrib, float **out ) { -- cgit v1.2.3 From 983b6a73e1842b436d158dc1d018bd483a1c9929 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 14 Apr 2008 12:32:25 -0400 Subject: use the new macro --- src/gallium/auxiliary/draw/draw_context.c | 3 +-- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 2 +- 2 files changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 1e70a77523..0c314f6e1d 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -88,8 +88,7 @@ struct draw_context *draw_create( void ) /* Statically allocate maximum sized vertices for the cache - could be cleverer... */ { - const unsigned size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; - char *tmp = align_malloc(VS_QUEUE_LENGTH * size, 16); + char *tmp = align_malloc(VS_QUEUE_LENGTH * MAX_VERTEX_ALLOCATION, 16); if (!tmp) goto fail; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index b49c9efa65..04b3d2c4cf 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -128,7 +128,7 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, fpme->nr_fetch = nr; //fpme->pipeline_vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); - fpme->pipeline_vertex_size = (MAX_VERTEX_SIZE + 0x0f) & ~0x0f; + fpme->pipeline_vertex_size = MAX_VERTEX_ALLOCATION; fpme->hw_vertex_size = vinfo->size * 4; } -- cgit v1.2.3 From f58ab8e75c082a0aea36ed63fe7e21fb5fac14b6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 14 Apr 2008 10:56:56 -0600 Subject: gallium: take reduced prim, fill modes into account when culling --- src/gallium/drivers/softpipe/sp_context.h | 2 ++ src/gallium/drivers/softpipe/sp_draw_arrays.c | 15 +++++++++++++++ src/gallium/drivers/softpipe/sp_setup.c | 11 ++++++++++- 3 files changed, 27 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 2442bd1eb0..0e1d5e561d 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -105,6 +105,8 @@ struct softpipe_context { int psize_slot; + unsigned reduced_api_prim; /**< PIPE_PRIM_POINTS, _LINES or _TRIANGLES */ + #if 0 /* Stipple derived state: */ diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 61bcf51899..421509495a 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -78,6 +78,20 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp) } +static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES +}; + + boolean softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, unsigned start, unsigned count) @@ -115,6 +129,7 @@ softpipe_draw_elements(struct pipe_context *pipe, assert(0); #endif + sp->reduced_api_prim = reduced_prim[mode]; if (sp->dirty) softpipe_update_derived( sp ); diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 0164d0588d..813d703108 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -1254,7 +1254,16 @@ void setup_prepare( struct setup_context *setup ) sp->quad.first->begin(sp->quad.first); } - setup->winding = sp->rasterizer->cull_mode; + if (sp->reduced_api_prim == PIPE_PRIM_TRIANGLES && + sp->rasterizer->fill_cw == PIPE_POLYGON_MODE_FILL && + sp->rasterizer->fill_ccw == PIPE_POLYGON_MODE_FILL) { + /* we'll do culling */ + setup->winding = sp->rasterizer->cull_mode; + } + else { + /* 'draw' will do culling */ + setup->winding = PIPE_WINDING_NONE; + } } -- cgit v1.2.3 From 78852986e6379a615a5742951f0fc6470e7c9d12 Mon Sep 17 00:00:00 2001 From: David Flynn Date: Mon, 14 Apr 2008 12:56:10 -0600 Subject: mesa: define #extension GL_ARB_texture_rectangle --- src/mesa/shader/slang/slang_preprocess.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_preprocess.c b/src/mesa/shader/slang/slang_preprocess.c index 076e982f8f..cf8cd5f87a 100644 --- a/src/mesa/shader/slang/slang_preprocess.c +++ b/src/mesa/shader/slang/slang_preprocess.c @@ -483,6 +483,7 @@ pp_cond_stack_reevaluate (pp_cond_stack *self) typedef struct { GLboolean MESA_shader_debug; /* GL_MESA_shader_debug enable */ + GLboolean GL_ARB_texture_rectangle; /* GL_ARB_texture_rectangle enable */ } pp_ext; /* @@ -498,6 +499,7 @@ static GLvoid pp_ext_init (pp_ext *self) { pp_ext_disable_all (self); + self->GL_ARB_texture_rectangle = GL_TRUE; /* Other initialization code goes here. */ } @@ -506,6 +508,8 @@ pp_ext_set (pp_ext *self, const char *name, GLboolean enable) { if (_mesa_strcmp (name, "MESA_shader_debug") == 0) self->MESA_shader_debug = enable; + else if (_mesa_strcmp (name, "GL_ARB_texture_rectangle") == 0) + self->GL_ARB_texture_rectangle = enable; /* Next extension name tests go here. */ else return GL_FALSE; -- cgit v1.2.3 From 5807c0242c38742d96c7ada3503f3198070aa208 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 14 Apr 2008 14:38:11 -0600 Subject: fix GL_ARB_texture_rectangle breakage --- src/mesa/shader/slang/slang_preprocess.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_preprocess.c b/src/mesa/shader/slang/slang_preprocess.c index cf8cd5f87a..1645afcc18 100644 --- a/src/mesa/shader/slang/slang_preprocess.c +++ b/src/mesa/shader/slang/slang_preprocess.c @@ -483,7 +483,7 @@ pp_cond_stack_reevaluate (pp_cond_stack *self) typedef struct { GLboolean MESA_shader_debug; /* GL_MESA_shader_debug enable */ - GLboolean GL_ARB_texture_rectangle; /* GL_ARB_texture_rectangle enable */ + GLboolean ARB_texture_rectangle; /* GL_ARB_texture_rectangle enable */ } pp_ext; /* @@ -499,7 +499,7 @@ static GLvoid pp_ext_init (pp_ext *self) { pp_ext_disable_all (self); - self->GL_ARB_texture_rectangle = GL_TRUE; + self->ARB_texture_rectangle = GL_TRUE; /* Other initialization code goes here. */ } @@ -509,7 +509,7 @@ pp_ext_set (pp_ext *self, const char *name, GLboolean enable) if (_mesa_strcmp (name, "MESA_shader_debug") == 0) self->MESA_shader_debug = enable; else if (_mesa_strcmp (name, "GL_ARB_texture_rectangle") == 0) - self->GL_ARB_texture_rectangle = enable; + self->ARB_texture_rectangle = enable; /* Next extension name tests go here. */ else return GL_FALSE; -- cgit v1.2.3 From 21ae3d2721326d56c76370fd8bfcc1536203925d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 14 Apr 2008 22:39:33 +0900 Subject: gallium: Serialize buffers writes. Allow concurrent reads from buffers by the CPU/GPU, but serialize all writes. --- .../auxiliary/pipebuffer/pb_buffer_fenced.c | 81 ++++++++++++++++++++-- 1 file changed, 76 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index b1f7d93057..65b6584003 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -35,6 +35,7 @@ #include "pipe/p_compiler.h" +#include "pipe/p_error.h" #include "pipe/p_debug.h" #include "pipe/p_winsys.h" #include "pipe/p_thread.h" @@ -54,6 +55,13 @@ */ #define SUPER(__derived) (&(__derived)->base) +#define PIPE_BUFFER_USAGE_CPU_READ_WRITE \ + ( PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE ) +#define PIPE_BUFFER_USAGE_GPU_READ_WRITE \ + ( PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE ) +#define PIPE_BUFFER_USAGE_WRITE \ + ( PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_GPU_WRITE ) + struct fenced_buffer_list { @@ -76,6 +84,15 @@ struct fenced_buffer struct pb_buffer *buffer; + /* FIXME: protect access with mutex */ + + /** + * A bitmask of PIPE_BUFFER_USAGE_CPU/GPU_READ/WRITE describing the current + * buffer usage. + */ + unsigned flags; + + unsigned mapcount; struct pipe_fence_handle *fence; struct list_head head; @@ -98,7 +115,9 @@ _fenced_buffer_add(struct fenced_buffer *fenced_buf) struct fenced_buffer_list *fenced_list = fenced_buf->list; assert(fenced_buf->base.base.refcount); + assert(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE); assert(fenced_buf->fence); + assert(!fenced_buf->head.prev); assert(!fenced_buf->head.next); LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed); @@ -130,6 +149,7 @@ _fenced_buffer_remove(struct fenced_buffer *fenced_buf) assert(fenced_buf->fence); winsys->fence_reference(winsys, &fenced_buf->fence, NULL); + fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE; assert(fenced_buf->head.prev); assert(fenced_buf->head.next); @@ -186,6 +206,34 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, } +/** + * Serialize writes, but allow concurrent reads. + */ +static INLINE enum pipe_error +fenced_buffer_serialize(struct fenced_buffer *fenced_buf, unsigned flags) +{ + struct fenced_buffer_list *fenced_list = fenced_buf->list; + struct pipe_winsys *winsys = fenced_list->winsys; + + if(((fenced_buf->flags | flags) & PIPE_BUFFER_USAGE_WRITE) == 0) + return PIPE_OK; + + if(fenced_buf->mapcount) { + /* FIXME */ + debug_warning("attemp to write concurrently to buffer"); + return PIPE_ERROR_RETRY; + } + + if(fenced_buf->fence) { + if(winsys->fence_finish(winsys, fenced_buf->fence, 0) != 0) + return PIPE_ERROR_RETRY; + _fenced_buffer_remove(fenced_buf); + } + + return PIPE_OK; +} + + static void fenced_buffer_destroy(struct pb_buffer *buf) { @@ -223,16 +271,30 @@ static void * fenced_buffer_map(struct pb_buffer *buf, unsigned flags) { - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - return pb_map(fenced_buf->buffer, flags); + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + void *map; + assert((flags & ~PIPE_BUFFER_USAGE_CPU_READ_WRITE) == 0); + + if(fenced_buffer_serialize(fenced_buf, flags) != PIPE_OK) + return NULL; + + map = pb_map(fenced_buf->buffer, flags); + if(map) + ++fenced_buf->mapcount; + fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE; + return map; } static void fenced_buffer_unmap(struct pb_buffer *buf) { - struct fenced_buffer *fenced_buf = fenced_buffer(buf); + struct fenced_buffer *fenced_buf = fenced_buffer(buf); + assert(fenced_buf->mapcount); pb_unmap(fenced_buf->buffer); + --fenced_buf->mapcount; + if(!fenced_buf->mapcount) + fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE; } @@ -288,13 +350,22 @@ buffer_fence(struct pb_buffer *buf, struct fenced_buffer *fenced_buf = fenced_buffer(buf); struct fenced_buffer_list *fenced_list = fenced_buf->list; struct pipe_winsys *winsys = fenced_list->winsys; + /* FIXME: receive this as a parameter */ + unsigned flags = fence ? PIPE_BUFFER_USAGE_GPU_READ_WRITE : 0; + + if(fenced_buffer_serialize(fenced_buf, flags) != PIPE_OK) { + /* FIXME: propagate error */ + (void)0; + } _glthread_LOCK_MUTEX(fenced_list->mutex); if (fenced_buf->fence) _fenced_buffer_remove(fenced_buf); - winsys->fence_reference(winsys, &fenced_buf->fence, fence); - if (fenced_buf->fence) + if (fence) { + winsys->fence_reference(winsys, &fenced_buf->fence, fence); + fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE; _fenced_buffer_add(fenced_buf); + } _glthread_UNLOCK_MUTEX(fenced_list->mutex); } -- cgit v1.2.3 From 5b8fa518476868530d748ce6d03674e9cca3d89f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 14 Apr 2008 23:55:36 +0900 Subject: gallium: Don't assume snprintf are always available. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 9 ++-- src/gallium/auxiliary/util/p_debug.c | 7 +-- src/gallium/auxiliary/util/u_snprintf.c | 16 +++--- src/gallium/auxiliary/util/u_string.h | 63 ++++++++++++++++++++++ .../drivers/i915simple/i915_fpc_translate.c | 3 +- src/gallium/drivers/i915simple/i915_screen.c | 3 +- src/gallium/drivers/i965simple/brw_screen.c | 3 +- src/gallium/include/pipe/p_format.h | 4 +- src/gallium/include/pipe/p_util.h | 8 --- 9 files changed, 88 insertions(+), 28 deletions(-) create mode 100644 src/gallium/auxiliary/util/u_string.h (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index cb3573ceb6..ff6a2c4194 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -30,6 +30,7 @@ #include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" +#include "util/u_string.h" #include "tgsi_dump.h" #include "tgsi_parse.h" #include "tgsi_build.h" @@ -147,7 +148,7 @@ gen_dump_uix( { char str[36]; - sprintf( str, "0x%x", ui ); + util_snprintf( str, sizeof(str), "0x%x", ui ); gen_dump_str( dump, str ); } @@ -158,7 +159,7 @@ gen_dump_uid( { char str[16]; - sprintf( str, "%u", ui ); + util_snprintf( str, sizeof(str), "%u", ui ); gen_dump_str( dump, str ); } @@ -169,7 +170,7 @@ gen_dump_sid( { char str[16]; - sprintf( str, "%d", si ); + util_snprintf( str, sizeof(str), "%d", si ); gen_dump_str( dump, str ); } @@ -180,7 +181,7 @@ gen_dump_flt( { char str[48]; - sprintf( str, "%10.4f", flt ); + util_snprintf( str, sizeof(str), "%10.4f", flt ); gen_dump_str( dump, str ); } diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 090e3b7794..f9366467cd 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -39,6 +39,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_util.h" #include "pipe/p_debug.h" +#include "util/u_string.h" #ifdef WIN32 @@ -60,7 +61,7 @@ void _debug_vprintf(const char *format, va_list ap) /* EngDebugPrint does not handle float point arguments, so we need to use * our own vsnprintf implementation */ char buf[512 + 1]; - vsnprintf(buf, sizeof(buf), format, ap); + util_vsnprintf(buf, sizeof(buf), format, ap); _EngDebugPrint("%s", buf); #else /* TODO: Implement debug print for WINCE */ @@ -311,7 +312,7 @@ debug_dump_enum(const struct debug_named_value *names, ++names; } - snprintf(rest, sizeof(rest), "0x%08lx", value); + util_snprintf(rest, sizeof(rest), "0x%08lx", value); return rest; } @@ -344,7 +345,7 @@ debug_dump_flags(const struct debug_named_value *names, else first = 0; - snprintf(rest, sizeof(rest), "0x%08lx", value); + util_snprintf(rest, sizeof(rest), "0x%08lx", value); strncat(output, rest, sizeof(output)); } diff --git a/src/gallium/auxiliary/util/u_snprintf.c b/src/gallium/auxiliary/util/u_snprintf.c index 48426abcb7..c4f4bbd30c 100644 --- a/src/gallium/auxiliary/util/u_snprintf.c +++ b/src/gallium/auxiliary/util/u_snprintf.c @@ -166,8 +166,8 @@ #include #else #ifdef WIN32 -#define vsnprintf rpl_vsnprintf -#define snprintf rpl_snprintf +#define vsnprintf util_vsnprintf +#define snprintf util_snprintf #define HAVE_VSNPRINTF 0 #define HAVE_SNPRINTF 0 #define HAVE_VASPRINTF 1 /* not needed */ @@ -445,7 +445,7 @@ static UINTMAX_T myround(LDOUBLE); static LDOUBLE mypow10(int); int -rpl_vsnprintf(char *str, size_t size, const char *format, va_list args) +util_vsnprintf(char *str, size_t size, const char *format, va_list args) { LDOUBLE fvalue; INTMAX_T value; @@ -1404,7 +1404,7 @@ mymemcpy(void *dst, void *src, size_t len) #endif /* NEED_MYMEMCPY */ int -rpl_vasprintf(char **ret, const char *format, va_list ap) +util_vasprintf(char **ret, const char *format, va_list ap) { size_t size; int len; @@ -1422,10 +1422,10 @@ rpl_vasprintf(char **ret, const char *format, va_list ap) #if !HAVE_SNPRINTF #if HAVE_STDARG_H int -rpl_snprintf(char *str, size_t size, const char *format, ...) +util_snprintf(char *str, size_t size, const char *format, ...) #else int -rpl_snprintf(va_alist) va_dcl +util_snprintf(va_alist) va_dcl #endif /* HAVE_STDARG_H */ { #if !HAVE_STDARG_H @@ -1449,10 +1449,10 @@ rpl_snprintf(va_alist) va_dcl #if !HAVE_ASPRINTF #if HAVE_STDARG_H int -rpl_asprintf(char **ret, const char *format, ...) +util_asprintf(char **ret, const char *format, ...) #else int -rpl_asprintf(va_alist) va_dcl +util_asprintf(va_alist) va_dcl #endif /* HAVE_STDARG_H */ { #if !HAVE_STDARG_H diff --git a/src/gallium/auxiliary/util/u_string.h b/src/gallium/auxiliary/util/u_string.h new file mode 100644 index 0000000000..b99d4e8021 --- /dev/null +++ b/src/gallium/auxiliary/util/u_string.h @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Platform independent functions for string manipulation. + * + * @author Jose Fonseca + */ + +#ifndef U_STRING_H_ +#define U_STRING_H_ + +#ifndef WIN32 +#include +#endif +#include +#include + + +#ifdef __cplusplus +extern "C" { +#endif + + +#ifdef WIN32 +int util_vsnprintf(char *, size_t, const char *, va_list); +int util_snprintf(char *str, size_t size, const char *format, ...); +#else +#define util_vsnprintf vsnprintf +#define util_snprintf snprintf +#endif + + +#ifdef __cplusplus +} +#endif + +#endif /* U_STRING_H_ */ diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 7b4fca5db1..3ccf74c72c 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -33,6 +33,7 @@ #include "i915_fpc.h" #include "pipe/p_shader_tokens.h" +#include "util/u_string.h" #include "tgsi/util/tgsi_parse.h" #include "tgsi/util/tgsi_dump.h" @@ -122,7 +123,7 @@ i915_program_error(struct i915_fp_compile *p, const char *msg, ...) debug_printf("i915_program_error: "); va_start( args, msg ); - vsprintf( buffer, msg, args ); + util_vsnprintf( buffer, sizeof(buffer), msg, args ); va_end( args ); debug_printf(buffer); debug_printf("\n"); diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 839b98c0ce..9ae594ce54 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -28,6 +28,7 @@ #include "pipe/p_util.h" #include "pipe/p_winsys.h" +#include "util/u_string.h" #include "i915_reg.h" #include "i915_context.h" @@ -78,7 +79,7 @@ i915_get_name( struct pipe_screen *pscreen ) break; } - sprintf(buffer, "i915 (chipset: %s)", chipset); + util_snprintf(buffer, sizeof(buffer), "i915 (chipset: %s)", chipset); return buffer; } diff --git a/src/gallium/drivers/i965simple/brw_screen.c b/src/gallium/drivers/i965simple/brw_screen.c index 5be369fe52..6845c7abde 100644 --- a/src/gallium/drivers/i965simple/brw_screen.c +++ b/src/gallium/drivers/i965simple/brw_screen.c @@ -28,6 +28,7 @@ #include "pipe/p_util.h" #include "pipe/p_winsys.h" +#include "util/u_string.h" #include "brw_context.h" #include "brw_screen.h" @@ -66,7 +67,7 @@ brw_get_name( struct pipe_screen *screen ) break; } - sprintf(buffer, "i965 (chipset: %s)", chipset); + util_snprintf(buffer, sizeof(buffer), "i965 (chipset: %s)", chipset); return buffer; } diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 9e0f91f202..ef9e3a3d6c 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -28,7 +28,7 @@ #ifndef PIPE_FORMAT_H #define PIPE_FORMAT_H -#include /* for sprintf */ +#include "util/u_string.h" #include "p_compiler.h" #include "p_debug.h" @@ -367,7 +367,7 @@ static INLINE char *pf_sprint_name( char *str, enum pipe_format format ) strcat( str, "S" ); break; } - sprintf( &str[strlen( str )], "%u", size * scale ); + util_snprintf( &str[strlen( str )], 32, "%u", size * scale ); } if (i != 0) { strcat( str, "_" ); diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 8e3aaee496..dbca080a4b 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -138,14 +138,6 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define GETENV( X ) debug_get_option( X, NULL ) -#ifdef WIN32 -int rpl_vsnprintf(char *, size_t, const char *, va_list); -int rpl_snprintf(char *str, size_t size, const char *format, ...); -#define vsnprintf rpl_vsnprintf -#define snprintf rpl_snprintf -#endif - - /** * Return memory on given byte alignment */ -- cgit v1.2.3 From 01c7dd2629d161bf87af679a3045e1e2d54259fc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Apr 2008 10:38:05 +0900 Subject: gallium: Add draw_pt_fetch_shade_pipeline.c to scons build. --- src/gallium/auxiliary/draw/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 52107912f5..a7fb5dbd61 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -18,6 +18,7 @@ draw = env.ConvenienceLibrary( 'draw_pt_vcache.c', 'draw_pt_fetch_emit.c', 'draw_pt_fetch_pipeline.c', + 'draw_pt_fetch_shade_pipeline.c', 'draw_pt_pipeline.c', 'draw_pt_elts.c', 'draw_prim.c', -- cgit v1.2.3 From fdd794dcfa33482bdabe7c04ec9df655e0c69bfc Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Apr 2008 20:55:14 -0600 Subject: gallium: fix PIPE_CAP_MAX_RENDER_TARGETS query --- src/gallium/drivers/softpipe/sp_screen.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index 1850a1ced3..7dacb1c461 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -68,7 +68,7 @@ softpipe_get_param(struct pipe_screen *screen, int param) case PIPE_CAP_POINT_SPRITE: return 1; case PIPE_CAP_MAX_RENDER_TARGETS: - return 1; + return PIPE_MAX_COLOR_BUFS; case PIPE_CAP_OCCLUSION_QUERY: return 1; case PIPE_CAP_TEXTURE_SHADOW_MAP: -- cgit v1.2.3 From b54225ccd6d3bc1b678e27c2f00ebddf5bf1046d Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Apr 2008 20:56:08 -0600 Subject: gallium: set ctx->Const.MaxDrawBuffers --- src/mesa/state_tracker/st_extensions.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 47a50d40ca..2f7ac074da 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -111,6 +111,10 @@ void st_init_limits(struct st_context *st) st->bitmap_texcoord_bias = screen->get_paramf(screen, PIPE_CAP_BITMAP_TEXCOORD_BIAS); + + c->MaxDrawBuffers + = CLAMP(screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS), + 1, MAX_DRAW_BUFFERS); } -- cgit v1.2.3 From e4b3c13d7a7dbd716bdf4b4d2dda8c6e579bd2d1 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Apr 2008 20:57:15 -0600 Subject: gallium: fix multi drawbuffer fb state --- src/mesa/state_tracker/st_atom_framebuffer.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 8a95096ec9..14eeb58cc1 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -48,7 +48,7 @@ update_framebuffer_state( struct st_context *st ) struct pipe_framebuffer_state *framebuffer = &st->state.framebuffer; struct gl_framebuffer *fb = st->ctx->DrawBuffer; struct st_renderbuffer *strb; - GLuint i; + GLuint i, j; memset(framebuffer, 0, sizeof(*framebuffer)); @@ -58,11 +58,14 @@ update_framebuffer_state( struct st_context *st ) /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state * to determine which surfaces to draw to */ - framebuffer->num_cbufs = fb->_NumColorDrawBuffers[0]; - for (i = 0; i < framebuffer->num_cbufs; i++) { - strb = st_renderbuffer(fb->_ColorDrawBuffers[0][i]); - assert(strb->surface); - framebuffer->cbufs[i] = strb->surface; + framebuffer->num_cbufs = 0; + for (j = 0; j < MAX_DRAW_BUFFERS; j++) { + for (i = 0; i < fb->_NumColorDrawBuffers[j]; i++) { + strb = st_renderbuffer(fb->_ColorDrawBuffers[j][i]); + assert(strb->surface); + framebuffer->cbufs[framebuffer->num_cbufs] = strb->surface; + framebuffer->num_cbufs++; + } } strb = st_renderbuffer(fb->Attachment[BUFFER_DEPTH].Renderbuffer); -- cgit v1.2.3 From 90b9a11a6d69f1cf6c837def0e8a9b598079ef1b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Apr 2008 20:58:05 -0600 Subject: gallium: fix semantic indexes for outputs --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 578fd2ecb0..524b5af50b 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -767,12 +767,14 @@ tgsi_translate_mesa_program( switch (outputSemanticName[i]) { case TGSI_SEMANTIC_POSITION: fulldecl = make_output_decl(i, - TGSI_SEMANTIC_POSITION, 0, /* Z / Depth */ + TGSI_SEMANTIC_POSITION, /* Z / Depth */ + outputSemanticIndex[i], TGSI_WRITEMASK_Z ); break; case TGSI_SEMANTIC_COLOR: fulldecl = make_output_decl(i, - TGSI_SEMANTIC_COLOR, 0, + TGSI_SEMANTIC_COLOR, + outputSemanticIndex[i], TGSI_WRITEMASK_XYZW ); break; default: -- cgit v1.2.3 From d3878b070b7b5084526b65499737cc686a6039b6 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 14 Apr 2008 21:01:40 -0600 Subject: gallium: enable new quad output code, remove old code --- src/gallium/drivers/softpipe/sp_quad_fs.c | 36 ------------------------------- 1 file changed, 36 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 9a20c056a2..8dbdbe5764 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -88,7 +88,6 @@ shade_quad( &qss->machine, quad ); -#if 0 /* XXX multi color outputs - untested */ /* store outputs */ boolean z_written = FALSE; { @@ -133,41 +132,6 @@ shade_quad( quad->outputs.depth[2] = z0 + dzdy; quad->outputs.depth[3] = z0 + dzdx + dzdy; } -#endif - - /* store result color(s) */ - if (qss->colorOutSlot >= 0) { - /* XXX need to handle multiple color outputs someday */ - assert(softpipe->fs->info.output_semantic_name[qss->colorOutSlot] - == TGSI_SEMANTIC_COLOR); - memcpy( - quad->outputs.color[0], - &machine->Outputs[qss->colorOutSlot].xyzw[0].f[0], - sizeof( quad->outputs.color[0] ) ); - } - - /* store result Z */ - if (qss->depthOutSlot >= 0) { - /* output[slot] is new Z */ - uint i; - for (i = 0; i < 4; i++) { - quad->outputs.depth[i] = machine->Outputs[0].xyzw[2].f[i]; - } - } - else { - /* compute Z values now, as in the quad earlyz stage */ - /* XXX we should really only do this if the earlyz stage is not used */ - const float fx = (float) quad->x0; - const float fy = (float) quad->y0; - const float dzdx = quad->posCoef->dadx[2]; - const float dzdy = quad->posCoef->dady[2]; - const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy; - - quad->outputs.depth[0] = z0; - quad->outputs.depth[1] = z0 + dzdx; - quad->outputs.depth[2] = z0 + dzdy; - quad->outputs.depth[3] = z0 + dzdx + dzdy; - } /* shader may cull fragments */ if( quad->mask ) { -- cgit v1.2.3 From a175e15f20b2a231cc9d09099e7b6d8aea6c624e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Apr 2008 12:34:33 +0900 Subject: gallium: Allow to use a single slab. We often want to use a pool of equally sized buffers, so this makes the slab suballocator a drop-in replacement. --- src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 22 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c | 315 +++++++++++++--------- 2 files changed, 209 insertions(+), 128 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index b2d2520b67..96f9af3825 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -118,13 +118,21 @@ mm_bufmgr_create_from_buffer(struct pb_buffer *buffer, * Slab sub-allocator. */ struct pb_manager * -pb_slab_manager_create(struct pb_manager *provider, - const struct pb_desc *desc, - size_t smallestSize, - size_t numSizes, - size_t desiredNumBuffers, - size_t maxSlabSize, - size_t pageAlignment); +pb_slab_manager_create(struct pb_manager *provider, + size_t bufSize, + size_t slabSize, + const struct pb_desc *desc); + +/** + * Allow a range of buffer size, by aggregating multiple slabs sub-allocators + * with different bucket sizes. + */ +struct pb_manager * +pb_slab_range_manager_create(struct pb_manager *provider, + size_t minBufSize, + size_t maxBufSize, + size_t slabSize, + const struct pb_desc *desc); /** diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c index 676e8e29b9..9506ac9ae1 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c @@ -70,19 +70,24 @@ struct pb_slab size_t numBuffers; size_t numFree; struct pb_slab_buffer *buffers; - struct pb_slab_size_header *header; + struct pb_slab_manager *mgr; struct pb_buffer *bo; - size_t pageAlignment; void *virtual; }; -struct pb_slab_size_header +struct pb_slab_manager { + struct pb_manager base; + + struct pb_manager *provider; + size_t bufSize; + size_t slabSize; + struct pb_desc desc; + struct list_head slabs; struct list_head freeSlabs; - struct pb_slab_manager *pool; - size_t bufSize; + _glthread_Mutex mutex; }; @@ -90,19 +95,18 @@ struct pb_slab_size_header * The data of this structure remains constant after * initialization and thus needs no mutex protection. */ -struct pb_slab_manager +struct pb_slab_range_manager { struct pb_manager base; + struct pb_manager *provider; + size_t minBufSize; + size_t maxBufSize; struct pb_desc desc; + + unsigned numBuckets; size_t *bucketSizes; - size_t numBuckets; - size_t pageSize; - struct pb_manager *provider; - unsigned pageAlignment; - unsigned maxSlabSize; - unsigned desiredNumBuffers; - struct pb_slab_size_header *headers; + struct pb_manager **buckets; }; @@ -122,8 +126,16 @@ pb_slab_manager(struct pb_manager *mgr) } +static INLINE struct pb_slab_range_manager * +pb_slab_range_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct pb_slab_range_manager *)mgr; +} + + /** - * Delete a buffer from the slab header delayed list and put + * Delete a buffer from the slab delayed list and put * it on the slab FREE list. */ static void @@ -131,10 +143,10 @@ pb_slab_buffer_destroy(struct pb_buffer *_buf) { struct pb_slab_buffer *buf = pb_slab_buffer(_buf); struct pb_slab *slab = buf->slab; - struct pb_slab_size_header *header = slab->header; + struct pb_slab_manager *mgr = slab->mgr; struct list_head *list = &buf->head; - _glthread_LOCK_MUTEX(header->mutex); + _glthread_LOCK_MUTEX(mgr->mutex); assert(buf->base.base.refcount == 0); @@ -145,21 +157,21 @@ pb_slab_buffer_destroy(struct pb_buffer *_buf) slab->numFree++; if (slab->head.next == &slab->head) - LIST_ADDTAIL(&slab->head, &header->slabs); + LIST_ADDTAIL(&slab->head, &mgr->slabs); if (slab->numFree == slab->numBuffers) { list = &slab->head; LIST_DEL(list); - LIST_ADDTAIL(list, &header->freeSlabs); + LIST_ADDTAIL(list, &mgr->freeSlabs); } - if (header->slabs.next == &header->slabs || slab->numFree + if (mgr->slabs.next == &mgr->slabs || slab->numFree != slab->numBuffers) { struct list_head *next; - for (list = header->freeSlabs.next, next = list->next; list - != &header->freeSlabs; list = next, next = list->next) { + for (list = mgr->freeSlabs.next, next = list->next; list + != &mgr->freeSlabs; list = next, next = list->next) { slab = LIST_ENTRY(struct pb_slab, list, head); @@ -170,7 +182,7 @@ pb_slab_buffer_destroy(struct pb_buffer *_buf) } } - _glthread_UNLOCK_MUTEX(header->mutex); + _glthread_UNLOCK_MUTEX(mgr->mutex); } @@ -217,15 +229,13 @@ pb_slab_buffer_vtbl = { static enum pipe_error -pb_slab_create(struct pb_slab_size_header *header) +pb_slab_create(struct pb_slab_manager *mgr) { - struct pb_slab_manager *pool = header->pool; - size_t size = header->bufSize * pool->desiredNumBuffers; struct pb_slab *slab; struct pb_slab_buffer *buf; - size_t numBuffers; - int ret; + unsigned numBuffers; unsigned i; + enum pipe_error ret; slab = CALLOC_STRUCT(pb_slab); if (!slab) @@ -236,22 +246,23 @@ pb_slab_create(struct pb_slab_size_header *header) * to efficiently reuse slabs. */ - size = (size <= pool->maxSlabSize) ? size : pool->maxSlabSize; - size = (size + pool->pageSize - 1) & ~(pool->pageSize - 1); - - slab->bo = pool->provider->create_buffer(pool->provider, size, &pool->desc); - if(!slab->bo) + slab->bo = mgr->provider->create_buffer(mgr->provider, mgr->slabSize, &mgr->desc); + if(!slab->bo) { + ret = PIPE_ERROR_OUT_OF_MEMORY; goto out_err0; + } slab->virtual = pb_map(slab->bo, - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE); - if(!slab->virtual) + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); + if(!slab->virtual) { + ret = PIPE_ERROR_OUT_OF_MEMORY; goto out_err1; + } pb_unmap(slab->bo); - numBuffers = slab->bo->base.size / header->bufSize; + numBuffers = slab->bo->base.size / mgr->bufSize; slab->buffers = CALLOC(numBuffers, sizeof(*slab->buffers)); if (!slab->buffers) { @@ -263,17 +274,17 @@ pb_slab_create(struct pb_slab_size_header *header) LIST_INITHEAD(&slab->freeBuffers); slab->numBuffers = numBuffers; slab->numFree = 0; - slab->header = header; + slab->mgr = mgr; buf = slab->buffers; for (i=0; i < numBuffers; ++i) { buf->base.base.refcount = 0; - buf->base.base.size = header->bufSize; + buf->base.base.size = mgr->bufSize; buf->base.base.alignment = 0; buf->base.base.usage = 0; buf->base.vtbl = &pb_slab_buffer_vtbl; buf->slab = slab; - buf->start = i* header->bufSize; + buf->start = i* mgr->bufSize; buf->mapCount = 0; _glthread_INIT_COND(buf->event); LIST_ADDTAIL(&buf->head, &slab->freeBuffers); @@ -281,7 +292,7 @@ pb_slab_create(struct pb_slab_size_header *header) buf++; } - LIST_ADDTAIL(&slab->head, &header->slabs); + LIST_ADDTAIL(&slab->head, &mgr->slabs); return PIPE_OK; @@ -293,51 +304,55 @@ out_err0: } +static int +check_alignment(size_t requested, size_t provided) +{ + return requested <= provided && (provided % requested) == 0; +} + + static struct pb_buffer * -pb_slab_manager_create_buffer(struct pb_manager *_pool, +pb_slab_manager_create_buffer(struct pb_manager *_mgr, size_t size, const struct pb_desc *desc) { - struct pb_slab_manager *pool = pb_slab_manager(_pool); - struct pb_slab_size_header *header; - unsigned i; + struct pb_slab_manager *mgr = pb_slab_manager(_mgr); static struct pb_slab_buffer *buf; struct pb_slab *slab; struct list_head *list; int count = DRI_SLABPOOL_ALLOC_RETRIES; - /* - * FIXME: Check for compatibility. - */ - - header = pool->headers; - for (i=0; inumBuckets; ++i) { - if (header->bufSize >= size) - break; - header++; - } - - if (i >= pool->numBuckets) - /* Fall back to allocate a buffer object directly from the provider. */ - return pool->provider->create_buffer(pool->provider, size, desc); - + /* check size */ + assert(size == mgr->bufSize); + if(size != mgr->bufSize) + return NULL; + + /* check if we can provide the requested alignment */ + assert(check_alignment(desc->alignment, mgr->desc.alignment)); + if(!check_alignment(desc->alignment, mgr->desc.alignment)) + return NULL; + assert(check_alignment(desc->alignment, mgr->bufSize)); + if(!check_alignment(desc->alignment, mgr->bufSize)) + return NULL; - _glthread_LOCK_MUTEX(header->mutex); - while (header->slabs.next == &header->slabs && count > 0) { - if (header->slabs.next != &header->slabs) + /* XXX: check for compatible buffer usage too? */ + + _glthread_LOCK_MUTEX(mgr->mutex); + while (mgr->slabs.next == &mgr->slabs && count > 0) { + if (mgr->slabs.next != &mgr->slabs) break; - _glthread_UNLOCK_MUTEX(header->mutex); + _glthread_UNLOCK_MUTEX(mgr->mutex); if (count != DRI_SLABPOOL_ALLOC_RETRIES) util_time_sleep(1); - _glthread_LOCK_MUTEX(header->mutex); - (void) pb_slab_create(header); + _glthread_LOCK_MUTEX(mgr->mutex); + (void) pb_slab_create(mgr); count--; } - list = header->slabs.next; - if (list == &header->slabs) { - _glthread_UNLOCK_MUTEX(header->mutex); + list = mgr->slabs.next; + if (list == &mgr->slabs) { + _glthread_UNLOCK_MUTEX(mgr->mutex); return NULL; } slab = LIST_ENTRY(struct pb_slab, list, head); @@ -347,83 +362,141 @@ pb_slab_manager_create_buffer(struct pb_manager *_pool, list = slab->freeBuffers.next; LIST_DELINIT(list); - _glthread_UNLOCK_MUTEX(header->mutex); + _glthread_UNLOCK_MUTEX(mgr->mutex); buf = LIST_ENTRY(struct pb_slab_buffer, list, head); + ++buf->base.base.refcount; + buf->base.base.alignment = desc->alignment; + buf->base.base.usage = desc->usage; + return &buf->base; } static void -pb_slab_manager_destroy(struct pb_manager *_pool) +pb_slab_manager_destroy(struct pb_manager *_mgr) { - struct pb_slab_manager *pool = pb_slab_manager(_pool); + struct pb_slab_manager *mgr = pb_slab_manager(_mgr); - FREE(pool->headers); - FREE(pool->bucketSizes); - FREE(pool); + /* TODO: cleanup all allocated buffers */ + FREE(mgr); } struct pb_manager * -pb_slab_manager_create(struct pb_manager *provider, - const struct pb_desc *desc, - size_t smallestSize, - size_t numSizes, - size_t desiredNumBuffers, - size_t maxSlabSize, - size_t pageAlignment) +pb_slab_manager_create(struct pb_manager *provider, + size_t bufSize, + size_t slabSize, + const struct pb_desc *desc) +{ + struct pb_slab_manager *mgr; + + mgr = CALLOC_STRUCT(pb_slab_manager); + if (!mgr) + return NULL; + + mgr->base.destroy = pb_slab_manager_destroy; + mgr->base.create_buffer = pb_slab_manager_create_buffer; + + mgr->provider = provider; + mgr->bufSize = bufSize; + mgr->slabSize = slabSize; + mgr->desc = *desc; + + LIST_INITHEAD(&mgr->slabs); + LIST_INITHEAD(&mgr->freeSlabs); + + _glthread_INIT_MUTEX(mgr->mutex); + + return &mgr->base; +} + + +static struct pb_buffer * +pb_slab_range_manager_create_buffer(struct pb_manager *_mgr, + size_t size, + const struct pb_desc *desc) { - struct pb_slab_manager *pool; - size_t i; + struct pb_slab_range_manager *mgr = pb_slab_range_manager(_mgr); + size_t bufSize; + unsigned i; - pool = CALLOC_STRUCT(pb_slab_manager); - if (!pool) + bufSize = mgr->minBufSize; + for (i = 0; i < mgr->numBuckets; ++i) { + if(bufSize >= size) + return mgr->buckets[i]->create_buffer(mgr->buckets[i], size, desc); + bufSize *= 2; + } + + /* Fall back to allocate a buffer object directly from the provider. */ + return mgr->provider->create_buffer(mgr->provider, size, desc); +} + + +static void +pb_slab_range_manager_destroy(struct pb_manager *_mgr) +{ + struct pb_slab_range_manager *mgr = pb_slab_range_manager(_mgr); + unsigned i; + + for (i = 0; i < mgr->numBuckets; ++i) + mgr->buckets[i]->destroy(mgr->buckets[i]); + FREE(mgr->buckets); + FREE(mgr->bucketSizes); + FREE(mgr); +} + + +struct pb_manager * +pb_slab_range_manager_create(struct pb_manager *provider, + size_t minBufSize, + size_t maxBufSize, + size_t slabSize, + const struct pb_desc *desc) +{ + struct pb_slab_range_manager *mgr; + size_t bufSize; + unsigned i; + + mgr = CALLOC_STRUCT(pb_slab_range_manager); + if (!mgr) goto out_err0; - pool->bucketSizes = CALLOC(numSizes, sizeof(*pool->bucketSizes)); - if (!pool->bucketSizes) - goto out_err1; + mgr->base.destroy = pb_slab_range_manager_destroy; + mgr->base.create_buffer = pb_slab_range_manager_create_buffer; - pool->headers = CALLOC(numSizes, sizeof(*pool->headers)); - if (!pool->headers) - goto out_err2; - - pool->desc = *desc; - pool->numBuckets = numSizes; -#ifdef WIN32 - pool->pageSize = 4096; -#else - pool->pageSize = getpagesize(); -#endif - pool->provider = provider; - pool->pageAlignment = pageAlignment; - pool->maxSlabSize = maxSlabSize; - pool->desiredNumBuffers = desiredNumBuffers; - - for (i=0; inumBuckets; ++i) { - struct pb_slab_size_header *header = &pool->headers[i]; - - pool->bucketSizes[i] = (smallestSize << i); - - _glthread_INIT_MUTEX(header->mutex); - - LIST_INITHEAD(&header->slabs); - LIST_INITHEAD(&header->freeSlabs); - - header->pool = pool; - header->bufSize = (smallestSize << i); + mgr->provider = provider; + mgr->minBufSize = minBufSize; + mgr->maxBufSize = maxBufSize; + + mgr->numBuckets = 1; + bufSize = minBufSize; + while(bufSize < maxBufSize) { + bufSize *= 2; + ++mgr->numBuckets; } + + mgr->buckets = CALLOC(mgr->numBuckets, sizeof(*mgr->buckets)); + if (!mgr->buckets) + goto out_err1; - pool->base.destroy = pb_slab_manager_destroy; - pool->base.create_buffer = pb_slab_manager_create_buffer; + bufSize = minBufSize; + for (i = 0; i < mgr->numBuckets; ++i) { + mgr->buckets[i] = pb_slab_manager_create(provider, bufSize, slabSize, desc); + if(!mgr->buckets[i]) + goto out_err2; + bufSize *= 2; + } - return &pool->base; + return &mgr->base; out_err2: - FREE(pool->bucketSizes); + for (i = 0; i < mgr->numBuckets; ++i) + if(mgr->buckets[i]) + mgr->buckets[i]->destroy(mgr->buckets[i]); + FREE(mgr->buckets); out_err1: - FREE(pool); + FREE(mgr); out_err0: return NULL; } -- cgit v1.2.3 From 0b995b44e5a02dd4a3abdb6b2a0821a32e597e7c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Apr 2008 12:35:00 +0900 Subject: gallium: Fix mismatching prototypes. --- src/gallium/auxiliary/util/u_time.c | 2 +- src/gallium/auxiliary/util/u_time.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index e6c0b19ff6..04f85c42ef 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -133,7 +133,7 @@ util_time_timeout(const struct util_time *start, #ifdef WIN32 -void util_time_usleep(unsigned usecs) +void util_time_sleep(unsigned usecs) { LONGLONG start, curr, end; diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h index 32035cceb5..2133958446 100644 --- a/src/gallium/auxiliary/util/u_time.h +++ b/src/gallium/auxiliary/util/u_time.h @@ -87,7 +87,7 @@ util_time_timeout(const struct util_time *start, #ifndef WIN32 #define util_time_sleep usleep #else -int +void util_time_sleep(unsigned usecs); #endif -- cgit v1.2.3 From 50bbbbe581edd6b8d4fe9f8ba7f134e17dc80a0b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Apr 2008 14:58:32 +0900 Subject: gallium: Remove middle of scope declarations. --- src/gallium/drivers/softpipe/sp_quad_fs.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 8dbdbe5764..625d0f9b48 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -77,7 +77,8 @@ shade_quad( struct quad_shade_stage *qss = quad_shade_stage( qs ); struct softpipe_context *softpipe = qs->softpipe; struct tgsi_exec_machine *machine = &qss->machine; - + boolean z_written; + /* Consts do not require 16 byte alignment. */ machine->Consts = softpipe->mapped_constants[PIPE_SHADER_FRAGMENT]; @@ -89,7 +90,7 @@ shade_quad( quad ); /* store outputs */ - boolean z_written = FALSE; + z_written = FALSE; { const ubyte *sem_name = softpipe->fs->info.output_semantic_name; const ubyte *sem_index = softpipe->fs->info.output_semantic_index; -- cgit v1.2.3 From d005befcb9e191ae90619fbdd3c37e262ae3b03e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Apr 2008 15:40:35 +0900 Subject: gallium: Less confusing interface for timeouts. --- src/gallium/auxiliary/util/u_time.c | 6 +++--- src/gallium/auxiliary/util/u_time.h | 4 ++-- 2 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index 04f85c42ef..01112ebe5a 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -120,15 +120,15 @@ util_time_compare(const struct util_time *t1, } -int +boolean util_time_timeout(const struct util_time *start, const struct util_time *end, const struct util_time *curr) { if(util_time_compare(start, end) <= 0) - return util_time_compare(start, curr) <= 0 && util_time_compare(curr, end) < 0; + return !(util_time_compare(start, curr) <= 0 && util_time_compare(curr, end) < 0); else - return util_time_compare(start, curr) <= 0 || util_time_compare(curr, end) < 0; + return !(util_time_compare(start, curr) <= 0 || util_time_compare(curr, end) < 0); } diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h index 2133958446..c8836c137f 100644 --- a/src/gallium/auxiliary/util/u_time.h +++ b/src/gallium/auxiliary/util/u_time.h @@ -77,9 +77,9 @@ util_time_diff(const struct util_time *t1, const struct util_time *t2); /** - * Returns zero when the timeout expires, non zero otherwise. + * Returns non-zero when the timeout expires. */ -int +boolean util_time_timeout(const struct util_time *start, const struct util_time *end, const struct util_time *curr); -- cgit v1.2.3 From 95aeeb6d746e57473116ef4d72c05330902f68a5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Apr 2008 15:41:08 +0900 Subject: gallium: Several fixes to buffer caching. --- src/gallium/auxiliary/pipebuffer/pb_buffer.h | 11 ++++ src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c | 67 +++++++++++++++++----- src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c | 15 ++--- 3 files changed, 68 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h index 4b09c80b2a..49705cb862 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h @@ -192,6 +192,17 @@ pb_reference(struct pb_buffer **dst, } +/** + * Utility function to check whether a requested alignment is consistent with + * the provided alignment or not. + */ +static INLINE int +pb_check_alignment(size_t requested, size_t provided) +{ + return requested <= provided && (provided % requested) == 0; +} + + /** * Malloc-based buffer to store data that can't be used by the graphics * hardware. diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c index 06de0bb6c3..543fd51253 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c @@ -136,7 +136,7 @@ _pb_cache_buffer_list_check_free(struct pb_cache_manager *mgr) while(curr != &mgr->delayed) { buf = LIST_ENTRY(struct pb_cache_buffer, curr, head); - if(util_time_timeout(&buf->start, &buf->end, &now) != 0) + if(!util_time_timeout(&buf->start, &buf->end, &now)) break; _pb_cache_buffer_destroy(buf); @@ -202,6 +202,24 @@ pb_cache_buffer_vtbl = { }; +static INLINE boolean +pb_cache_is_buffer_compat(struct pb_cache_buffer *buf, + size_t size, + const struct pb_desc *desc) +{ + /* TODO: be more lenient with size */ + if(buf->base.base.size != size) + return FALSE; + + if(!pb_check_alignment(desc->alignment, buf->base.base.alignment)) + return FALSE; + + /* XXX: check usage too? */ + + return TRUE; +} + + static struct pb_buffer * pb_cache_manager_create_buffer(struct pb_manager *_mgr, size_t size, @@ -209,29 +227,45 @@ pb_cache_manager_create_buffer(struct pb_manager *_mgr, { struct pb_cache_manager *mgr = pb_cache_manager(_mgr); struct pb_cache_buffer *buf; + struct pb_cache_buffer *curr_buf; struct list_head *curr, *next; struct util_time now; - util_time_get(&now); + _glthread_LOCK_MUTEX(mgr->mutex); + + buf = NULL; curr = mgr->delayed.next; next = curr->next; + + /* search in the expired buffers, freeing them in the process */ + util_time_get(&now); while(curr != &mgr->delayed) { - buf = LIST_ENTRY(struct pb_cache_buffer, curr, head); - - if(buf->base.base.size == size && - buf->base.base.alignment >= desc->alignment && - (buf->base.base.alignment % desc->alignment) == 0 && - /* buf->base.base.usage == usage */ 1) { - ++buf->base.base.refcount; - return &buf->base; - } - - if(util_time_timeout(&buf->start, &buf->end, &now) != 0) - _pb_cache_buffer_destroy(buf); + curr_buf = LIST_ENTRY(struct pb_cache_buffer, curr, head); + if(!buf && pb_cache_is_buffer_compat(curr_buf, size, desc)) + buf = curr_buf; + else if(util_time_timeout(&curr_buf->start, &curr_buf->end, &now)) + _pb_cache_buffer_destroy(curr_buf); + curr = next; + next = curr->next; + } + /* keep searching in the hot buffers */ + while(!buf && curr != &mgr->delayed) { + curr_buf = LIST_ENTRY(struct pb_cache_buffer, curr, head); + if(pb_cache_is_buffer_compat(curr_buf, size, desc)) + buf = curr_buf; curr = next; next = curr->next; } + + if(buf) { + LIST_DEL(&buf->head); + _glthread_UNLOCK_MUTEX(mgr->mutex); + ++buf->base.base.refcount; + return &buf->base; + } + + _glthread_UNLOCK_MUTEX(mgr->mutex); buf = CALLOC_STRUCT(pb_cache_buffer); if(!buf) @@ -243,6 +277,11 @@ pb_cache_manager_create_buffer(struct pb_manager *_mgr, return NULL; } + assert(buf->buffer->base.refcount >= 1); + assert(pb_check_alignment(desc->alignment, buf->buffer->base.alignment)); + assert((buf->buffer->base.usage & desc->usage) == desc->usage); + assert(buf->buffer->base.size >= size); + buf->base.base.refcount = 1; buf->base.base.alignment = buf->buffer->base.alignment; buf->base.base.usage = buf->buffer->base.usage; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c index 9506ac9ae1..b931455056 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c @@ -304,13 +304,6 @@ out_err0: } -static int -check_alignment(size_t requested, size_t provided) -{ - return requested <= provided && (provided % requested) == 0; -} - - static struct pb_buffer * pb_slab_manager_create_buffer(struct pb_manager *_mgr, size_t size, @@ -328,11 +321,11 @@ pb_slab_manager_create_buffer(struct pb_manager *_mgr, return NULL; /* check if we can provide the requested alignment */ - assert(check_alignment(desc->alignment, mgr->desc.alignment)); - if(!check_alignment(desc->alignment, mgr->desc.alignment)) + assert(pb_check_alignment(desc->alignment, mgr->desc.alignment)); + if(!pb_check_alignment(desc->alignment, mgr->desc.alignment)) return NULL; - assert(check_alignment(desc->alignment, mgr->bufSize)); - if(!check_alignment(desc->alignment, mgr->bufSize)) + assert(pb_check_alignment(desc->alignment, mgr->bufSize)); + if(!pb_check_alignment(desc->alignment, mgr->bufSize)) return NULL; /* XXX: check for compatible buffer usage too? */ -- cgit v1.2.3 From 7619240cc0c24d3ad4d2424e65110c0326a12dad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Apr 2008 15:58:14 +0900 Subject: gallium: Fix seg fault (James Vogt). --- src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c index bffca5b244..9d809e2f9b 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c @@ -30,7 +30,7 @@ * \file * A buffer manager that wraps buffers in fenced buffers. * - * \author José Fonseca + * \author José Fonseca */ @@ -101,7 +101,8 @@ fenced_bufmgr_destroy(struct pb_manager *mgr) fenced_buffer_list_destroy(fenced_mgr->fenced_list); - fenced_mgr->provider->destroy(fenced_mgr->provider); + if(fenced_mgr->provider) + fenced_mgr->provider->destroy(fenced_mgr->provider); FREE(fenced_mgr); } @@ -113,6 +114,9 @@ fenced_bufmgr_create(struct pb_manager *provider, { struct fenced_pb_manager *fenced_mgr; + if(!provider) + return NULL; + fenced_mgr = (struct fenced_pb_manager *)CALLOC(1, sizeof(*fenced_mgr)); if (!fenced_mgr) return NULL; -- cgit v1.2.3 From 3c4f1ba5a2edefd69b2c47abaf534fb3af3f259d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Apr 2008 17:08:10 +0900 Subject: gallium: Eliminate stdio file usage. Remove unused stuff. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 334 +++------------------------- src/gallium/auxiliary/tgsi/util/tgsi_dump.h | 10 - 2 files changed, 34 insertions(+), 310 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index ff6a2c4194..26bfc2051f 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -25,8 +25,6 @@ * **************************************************************************/ -#include - #include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" @@ -35,196 +33,28 @@ #include "tgsi_parse.h" #include "tgsi_build.h" -struct gen_dump -{ - unsigned tabs; - void (* write)( - struct gen_dump *dump, - const void *data, - unsigned size ); -}; - -struct text_dump -{ - struct gen_dump base; - char *text; - unsigned length; - unsigned capacity; -}; - -static void -_text_dump_write( - struct gen_dump *dump, - const void *data, - unsigned size ) -{ - struct text_dump *td = (struct text_dump *) dump; - unsigned new_length = td->length + size; - - if( new_length >= td->capacity ) { - unsigned new_capacity = td->capacity; - - do { - if( new_capacity == 0 ) { - new_capacity = 256; - } - else { - new_capacity *= 2; - } - } while( new_length >= new_capacity ); - td->text = (char *) REALLOC( - td->text, - td->capacity, - new_capacity ); - td->capacity = new_capacity; - } - memcpy( - &td->text[td->length], - data, - size ); - td->length = new_length; - td->text[td->length] = '\0'; -} - -struct file_dump -{ - struct gen_dump base; - FILE *file; -}; - -static void -_file_dump_write( - struct gen_dump *dump, - const void *data, - unsigned size ) -{ - struct file_dump *fd = (struct file_dump *) dump; - -#if 0 - fwrite( data, 1, size, fd->file ); -#else - { - unsigned i; - - for (i = 0; i < size; i++ ) { - fprintf( fd->file, "%c", ((const char *) data)[i] ); - } - } -#endif -} - -static void -gen_dump_str( - struct gen_dump *dump, - const char *str ) -{ - unsigned i; - size_t len = strlen( str ); - - for (i = 0; i < len; i++) { - dump->write( dump, &str[i], 1 ); - if (str[i] == '\n') { - unsigned i; - - for (i = 0; i < dump->tabs; i++) { - dump->write( dump, " ", 4 ); - } - } - } -} - -static void -gen_dump_chr( - struct gen_dump *dump, - const char chr ) -{ - dump->write( dump, &chr, 1 ); -} - -static void -gen_dump_uix( - struct gen_dump *dump, - const unsigned ui ) -{ - char str[36]; - - util_snprintf( str, sizeof(str), "0x%x", ui ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_uid( - struct gen_dump *dump, - const unsigned ui ) -{ - char str[16]; - - util_snprintf( str, sizeof(str), "%u", ui ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_sid( - struct gen_dump *dump, - const int si ) -{ - char str[16]; - - util_snprintf( str, sizeof(str), "%d", si ); - gen_dump_str( dump, str ); -} - static void -gen_dump_flt( - struct gen_dump *dump, - const float flt ) -{ - char str[48]; - - util_snprintf( str, sizeof(str), "%10.4f", flt ); - gen_dump_str( dump, str ); -} - -static void -gen_dump_enum( - struct gen_dump *dump, +dump_enum( const unsigned e, const char **enums, const unsigned enums_count ) { if (e >= enums_count) { - gen_dump_uid( dump, e ); + debug_printf( "%u", e ); } else { - gen_dump_str( dump, enums[e] ); + debug_printf( "%s", enums[e] ); } } -static void -gen_dump_tab( - struct gen_dump *dump ) -{ - ++dump->tabs; -} - -static void -gen_dump_untab( - struct gen_dump *dump ) -{ - assert( dump->tabs > 0 ); - - --dump->tabs; -} - -#define TXT(S) gen_dump_str( dump, S ) -#define CHR(C) gen_dump_chr( dump, C ) -#define UIX(I) gen_dump_uix( dump, I ) -#define UID(I) gen_dump_uid( dump, I ) -#define SID(I) gen_dump_sid( dump, I ) -#define FLT(F) gen_dump_flt( dump, F ) -#define TAB() gen_dump_tab( dump ) -#define UNT() gen_dump_untab( dump ) -#define ENM(E,ENUMS) gen_dump_enum( dump, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) +#define EOL() debug_printf( "\n" ) +#define TXT(S) debug_printf( "%s", S ) +#define CHR(C) debug_printf( "%c", C ) +#define UIX(I) debug_printf( "0x%x", I ) +#define UID(I) debug_printf( "%u", I ) +#define SID(I) debug_printf( "%d", I ) +#define FLT(F) debug_printf( "%10.4f", F ) +#define ENM(E,ENUMS) dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) static const char *TGSI_PROCESSOR_TYPES[] = { @@ -711,7 +541,6 @@ static const char *TGSI_MODULATES[] = static void dump_declaration_short( - struct gen_dump *dump, struct tgsi_full_declaration *decl ) { TXT( "\nDCL " ); @@ -765,7 +594,6 @@ dump_declaration_short( static void dump_declaration_verbose( - struct gen_dump *dump, struct tgsi_full_declaration *decl, unsigned ignored, unsigned deflt, @@ -803,7 +631,7 @@ dump_declaration_verbose( UIX( decl->Declaration.Padding ); } - CHR( '\n' ); + EOL(); switch( decl->Declaration.Declare ) { case TGSI_DECLARE_RANGE: TXT( "\nFirst: " ); @@ -822,7 +650,7 @@ dump_declaration_verbose( } if( decl->Declaration.Interpolate ) { - CHR( '\n' ); + EOL(); TXT( "\nInterpolate: " ); ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); if( ignored ) { @@ -832,7 +660,7 @@ dump_declaration_verbose( } if( decl->Declaration.Semantic ) { - CHR( '\n' ); + EOL(); TXT( "\nSemanticName : " ); ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); TXT( "\nSemanticIndex: " ); @@ -846,7 +674,6 @@ dump_declaration_verbose( static void dump_immediate_short( - struct gen_dump *dump, struct tgsi_full_immediate *imm ) { unsigned i; @@ -874,7 +701,6 @@ dump_immediate_short( static void dump_immediate_verbose( - struct gen_dump *dump, struct tgsi_full_immediate *imm, unsigned ignored ) { @@ -888,7 +714,7 @@ dump_immediate_verbose( } for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - CHR( '\n' ); + EOL(); switch( imm->Immediate.DataType ) { case TGSI_IMM_FLOAT32: TXT( "\nFloat: " ); @@ -903,14 +729,13 @@ dump_immediate_verbose( static void dump_instruction_short( - struct gen_dump *dump, struct tgsi_full_instruction *inst, unsigned instno ) { unsigned i; boolean first_reg = TRUE; - CHR( '\n' ); + EOL(); UID( instno ); CHR( ':' ); ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); @@ -1042,7 +867,6 @@ dump_instruction_short( static void dump_instruction_verbose( - struct gen_dump *dump, struct tgsi_full_instruction *inst, unsigned ignored, unsigned deflt, @@ -1070,7 +894,7 @@ dump_instruction_verbose( } if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { - CHR( '\n' ); + EOL(); TXT( "\nType : " ); ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { @@ -1124,7 +948,7 @@ dump_instruction_verbose( } if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { - CHR( '\n' ); + EOL(); TXT( "\nType : " ); ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { @@ -1142,7 +966,7 @@ dump_instruction_verbose( } if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { - CHR( '\n' ); + EOL(); TXT( "\nType : " ); ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { @@ -1163,7 +987,7 @@ dump_instruction_verbose( struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; - CHR( '\n' ); + EOL(); TXT( "\nFile : " ); ENM( dst->DstRegister.File, TGSI_FILES ); if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { @@ -1194,7 +1018,7 @@ dump_instruction_verbose( } if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { - CHR( '\n' ); + EOL(); TXT( "\nType : " ); ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { @@ -1232,7 +1056,7 @@ dump_instruction_verbose( } if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { - CHR( '\n' ); + EOL(); TXT( "\nType : " ); ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { @@ -1254,7 +1078,7 @@ dump_instruction_verbose( struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; - CHR( '\n' ); + EOL(); TXT( "\nFile : "); ENM( src->SrcRegister.File, TGSI_FILES ); if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { @@ -1299,7 +1123,7 @@ dump_instruction_verbose( } if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { - CHR( '\n' ); + EOL(); TXT( "\nType : " ); ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { @@ -1345,7 +1169,7 @@ dump_instruction_verbose( } if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { - CHR( '\n' ); + EOL(); TXT( "\nType : " ); ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { @@ -1380,9 +1204,8 @@ dump_instruction_verbose( } } -static void -dump_gen( - struct gen_dump *dump, +void +tgsi_dump( const struct tgsi_token *tokens, unsigned flags ) { @@ -1394,16 +1217,16 @@ dump_gen( unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); unsigned instno = 0; - dump->tabs = 0; - - /* sanity check */ + /* sanity checks */ assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); + assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "END") == 0); tgsi_parse_init( &parse, tokens ); TXT( "tgsi-dump begin -----------------" ); - CHR( '\n' ); + EOL(); ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); UID( parse.FullVersion.Version.MajorVersion ); CHR( '.' ); @@ -1414,7 +1237,7 @@ dump_gen( UID( parse.FullVersion.Version.MajorVersion ); TXT( "\nMinorVersion: " ); UID( parse.FullVersion.Version.MinorVersion ); - CHR( '\n' ); + EOL(); TXT( "\nHeaderSize: " ); UID( parse.FullHeader.Header.HeaderSize ); @@ -1422,7 +1245,7 @@ dump_gen( UID( parse.FullHeader.Header.BodySize ); TXT( "\nProcessor : " ); ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); - CHR( '\n' ); + EOL(); } fi = tgsi_default_full_instruction(); @@ -1434,19 +1257,16 @@ dump_gen( switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: dump_declaration_short( - dump, &parse.FullToken.FullDeclaration ); break; case TGSI_TOKEN_TYPE_IMMEDIATE: dump_immediate_short( - dump, &parse.FullToken.FullImmediate ); break; case TGSI_TOKEN_TYPE_INSTRUCTION: dump_instruction_short( - dump, &parse.FullToken.FullInstruction, instno ); instno++; @@ -1471,7 +1291,6 @@ dump_gen( switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: dump_declaration_verbose( - dump, &parse.FullToken.FullDeclaration, ignored, deflt, @@ -1480,14 +1299,12 @@ dump_gen( case TGSI_TOKEN_TYPE_IMMEDIATE: dump_immediate_verbose( - dump, &parse.FullToken.FullImmediate, ignored ); break; case TGSI_TOKEN_TYPE_INSTRUCTION: dump_instruction_verbose( - dump, &parse.FullToken.FullInstruction, ignored, deflt, @@ -1498,7 +1315,7 @@ dump_gen( assert( 0 ); } - CHR( '\n' ); + EOL(); } } @@ -1506,86 +1323,3 @@ dump_gen( tgsi_parse_free( &parse ); } - - -static void -sanity_checks(void) -{ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); - assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "END") == 0); -} - - -void -tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ) -{ - struct file_dump dump; - - sanity_checks(); - - dump.base.write = _file_dump_write; -#if 0 - { - static unsigned counter = 0; - char buffer[64]; - sprintf( buffer, "tgsi-dump-%.4u.txt", counter++ ); - dump.file = fopen( buffer, "wt" ); - } -#else - dump.file = stderr; -#endif - - dump_gen( - &dump.base, - tokens, - flags ); - -#if 0 - fclose( dump.file ); -#endif -} - -void -tgsi_dump_str( - char **str, - const struct tgsi_token *tokens, - unsigned flags ) -{ - struct text_dump dump; - - dump.base.write = _text_dump_write; - dump.text = NULL; - dump.length = 0; - dump.capacity = 0; - - dump_gen( - &dump.base, - tokens, - flags ); - - *str = dump.text; -} - - -void tgsi_debug_dump( struct tgsi_token *tokens ) -{ - char *str, *p; - - tgsi_dump_str( &str, tokens, 0 ); - - p = str; - while (p != NULL) - { - char *end = strchr( p, '\n' ); - if (end != NULL) - { - *end++ = '\0'; - } - debug_printf( "%s\n", p ); - p = end; - } - - FREE( str ); -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h index 51d79a0362..beb0155d56 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h @@ -14,16 +14,6 @@ tgsi_dump( const struct tgsi_token *tokens, unsigned flags ); -void -tgsi_dump_str( - char **str, - const struct tgsi_token *tokens, - unsigned flags ); - -/* Dump to debug_printf() - */ -void tgsi_debug_dump( struct tgsi_token *tokens ); - #if defined __cplusplus } #endif -- cgit v1.2.3 From a68f664124592829a3b715388e6cfa43f82900c8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Apr 2008 18:11:47 +0900 Subject: gallium: Cache one line worth of debug output on windows. The windbg connection seems synchronous, so this speeds up when printing little text at a time (e.g., tgsi output). --- src/gallium/auxiliary/util/p_debug.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index f9366467cd..c195f61820 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -59,10 +59,15 @@ void _debug_vprintf(const char *format, va_list ap) #ifdef WIN32 #ifndef WINCE /* EngDebugPrint does not handle float point arguments, so we need to use - * our own vsnprintf implementation */ - char buf[512 + 1]; - util_vsnprintf(buf, sizeof(buf), format, ap); - _EngDebugPrint("%s", buf); + * our own vsnprintf implementation. It is also very slow, so buffer until + * we find a newline. */ + static char buf[512 + 1] = {'\0'}; + size_t len = strlen(buf); + int ret = util_vsnprintf(buf + len, sizeof(buf) - len, format, ap); + if(ret > (int)(sizeof(buf) - len - 1) || strchr(buf + len, '\n')) { + _EngDebugPrint("%s", buf); + buf[0] = '\0'; + } #else /* TODO: Implement debug print for WINCE */ #endif -- cgit v1.2.3 From f2ee51e7d9a9bde8fd7de29b382d11fe0d58226d Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 15 Apr 2008 12:08:36 +0200 Subject: i915: Changed name to i915_dri.so --- src/gallium/winsys/dri/intel/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/Makefile b/src/gallium/winsys/dri/intel/Makefile index f140939248..c0ce2f927b 100644 --- a/src/gallium/winsys/dri/intel/Makefile +++ b/src/gallium/winsys/dri/intel/Makefile @@ -2,7 +2,7 @@ TOP = ../../../../.. include $(TOP)/configs/current -LIBNAME = i915tex_dri.so +LIBNAME = i915_dri.so MINIGLX_SOURCES = server/intel_dri.c -- cgit v1.2.3 From 6a26a9c58cc38ff636ee88ce01fed40eea500fc0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 15 Apr 2008 14:28:41 +0100 Subject: draw: fetch_shade_pipeline needs to translate to hw vertex format (from get_vertex_info) --- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 158 +++++++++++++-------- 1 file changed, 102 insertions(+), 56 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 04b3d2c4cf..557cd43f5b 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -36,52 +36,60 @@ struct fetch_pipeline_middle_end { struct draw_pt_middle_end base; struct draw_context *draw; + const ubyte *input_buf[2]; + struct { - const ubyte *ptr; - unsigned pitch; - void (*fetch)( const void *from, float *attrib); - void (*emit)( const float *attrib, float **out ); - } fetch[PIPE_MAX_ATTRIBS]; + const ubyte **input_buf; + unsigned input_offset; + unsigned output_offset; + + void (*emit)( const float *attrib, void *ptr ); + } translate[PIPE_MAX_ATTRIBS]; + unsigned nr_translate; - unsigned nr_fetch; unsigned pipeline_vertex_size; unsigned hw_vertex_size; unsigned prim; }; -#if 0 + +static void emit_NULL( const float *attrib, + void *ptr ) +{ +} + static void emit_R32_FLOAT( const float *attrib, - float **out ) + void *ptr ) { - (*out)[0] = attrib[0]; - (*out) += 1; + float *out = (float *)ptr; + out[0] = attrib[0]; } static void emit_R32G32_FLOAT( const float *attrib, - float **out ) + void *ptr ) { - (*out)[0] = attrib[0]; - (*out)[1] = attrib[1]; - (*out) += 2; + float *out = (float *)ptr; + out[0] = attrib[0]; + out[1] = attrib[1]; } static void emit_R32G32B32_FLOAT( const float *attrib, - float **out ) + void *ptr ) { - (*out)[0] = attrib[0]; - (*out)[1] = attrib[1]; - (*out)[2] = attrib[2]; - (*out) += 3; + float *out = (float *)ptr; + out[0] = attrib[0]; + out[1] = attrib[1]; + out[2] = attrib[2]; } -#endif + static void emit_R32G32B32A32_FLOAT( const float *attrib, - float **out ) + void *ptr ) { - (*out)[0] = attrib[0]; - (*out)[1] = attrib[1]; - (*out)[2] = attrib[2]; - (*out)[3] = attrib[3]; - (*out) += 4; + float *out = (float *)ptr; + out[0] = attrib[0]; + out[1] = attrib[1]; + out[2] = attrib[2]; + out[3] = attrib[3]; } static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, @@ -89,9 +97,10 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; - unsigned i, nr = 0; + unsigned i; boolean ok; const struct vertex_info *vinfo; + unsigned dst_offset; fpme->prim = prim; @@ -100,33 +109,63 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, assert(0); return; } + /* Must do this after set_primitive() above: */ vinfo = draw->render->get_vertex_info(draw->render); - /* Need to look at vertex shader inputs (we know it is a - * passthrough shader, so these define the outputs too). If we - * were running a shader, we'd still be looking at the inputs at - * this point. - */ - for (i = 0; i < draw->vertex_shader->info.num_inputs; i++) { - unsigned buf = draw->vertex_element[i].vertex_buffer_index; - enum pipe_format format = draw->vertex_element[i].src_format; - - fpme->fetch[nr].ptr = ((const ubyte *) draw->user.vbuffer[buf] + - draw->vertex_buffer[buf].buffer_offset + - draw->vertex_element[i].src_offset); - fpme->fetch[nr].pitch = draw->vertex_buffer[buf].pitch; - fpme->fetch[nr].fetch = draw_get_fetch_func( format ); + /* In passthrough mode, need to translate from vertex shader + * outputs to hw vertices. + */ + dst_offset = 0; + for (i = 0; i < vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned src_buffer = 0; + unsigned src_offset = (sizeof(struct vertex_header) + + vinfo->src_index[i] * 4 * sizeof(float) ); + + + + switch (vinfo->emit[i]) { + case EMIT_4F: + fpme->translate[i].emit = emit_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + fpme->translate[i].emit = emit_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + fpme->translate[i].emit = emit_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + fpme->translate[i].emit = emit_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + fpme->translate[i].emit = emit_R32_FLOAT; + emit_sz = 1 * sizeof(float); + src_buffer = 1; + src_offset = 0; + break; + default: + assert(0); + fpme->translate[i].emit = emit_NULL; + emit_sz = 0; + break; + } - /* Always do this -- somewhat redundant... - */ - fpme->fetch[nr].emit = emit_R32G32B32A32_FLOAT; - nr++; + fpme->translate[i].input_buf = &fpme->input_buf[src_buffer]; + fpme->translate[i].input_offset = src_offset; + fpme->translate[i].output_offset = dst_offset; + dst_offset += emit_sz; } - fpme->nr_fetch = nr; + fpme->nr_translate = vinfo->num_attribs; + fpme->hw_vertex_size = vinfo->size * 4; + //fpme->pipeline_vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); fpme->pipeline_vertex_size = MAX_VERTEX_ALLOCATION; fpme->hw_vertex_size = vinfo->size * 4; @@ -171,7 +210,7 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, } else { unsigned i, j; void *hw_verts; - float *out; + char *out_buf; /* XXX: need to flush to get prim_vbuf.c to release its allocation?? */ @@ -185,22 +224,29 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, return; } - out = (float *)hw_verts; + out_buf = (char *)hw_verts; + fpme->input_buf[0] = (const ubyte *)pipeline_verts; + fpme->input_buf[1] = (const ubyte *)&fpme->draw->rasterizer->point_size; + for (i = 0; i < fetch_count; i++) { - struct vertex_header *header = - (struct vertex_header*)(pipeline_verts + (fpme->pipeline_vertex_size * i)); - for (j = 0; j < fpme->nr_fetch; j++) { - float *attrib = header->data[j]; + for (j = 0; j < fpme->nr_translate; j++) { + + const float *attrib = (const float *)( (*fpme->translate[i].input_buf) + + fpme->translate[i].input_offset ); + + char *dest = out_buf + fpme->translate[i].output_offset; + /*debug_printf("emiting [%f, %f, %f, %f]\n", attrib[0], attrib[1], attrib[2], attrib[3]);*/ - fpme->fetch[j].emit(attrib, &out); + + fpme->translate[j].emit(attrib, dest); } + + fpme->input_buf[0] += fpme->pipeline_vertex_size; } - /* XXX: Draw arrays path to avoid re-emitting index list again and - * again. - */ + draw->render->draw(draw->render, draw_elts, draw_count); -- cgit v1.2.3 From c81bbab6f6c0413996799800cac6fb49a698e765 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 15 Apr 2008 14:35:29 +0100 Subject: gallium: add a generic vertex (or other) buffer translation module --- src/gallium/auxiliary/translate/Makefile | 12 + src/gallium/auxiliary/translate/SConscript | 9 + src/gallium/auxiliary/translate/translate.h | 82 +++ .../auxiliary/translate/translate_generic.c | 630 +++++++++++++++++++++ 4 files changed, 733 insertions(+) create mode 100644 src/gallium/auxiliary/translate/Makefile create mode 100644 src/gallium/auxiliary/translate/SConscript create mode 100644 src/gallium/auxiliary/translate/translate.h create mode 100644 src/gallium/auxiliary/translate/translate_generic.c (limited to 'src') diff --git a/src/gallium/auxiliary/translate/Makefile b/src/gallium/auxiliary/translate/Makefile new file mode 100644 index 0000000000..051987bb7e --- /dev/null +++ b/src/gallium/auxiliary/translate/Makefile @@ -0,0 +1,12 @@ +TOP = ../../../.. +include $(TOP)/configs/current + +LIBNAME = translate + +C_SOURCES = \ + translate_generic.c + +include ../../Makefile.template + +symlinks: + diff --git a/src/gallium/auxiliary/translate/SConscript b/src/gallium/auxiliary/translate/SConscript new file mode 100644 index 0000000000..947b9438cb --- /dev/null +++ b/src/gallium/auxiliary/translate/SConscript @@ -0,0 +1,9 @@ +Import('*') + +cso_cache = env.ConvenienceLibrary( + target = 'translate', + source = [ + 'translate_generic.c', + ]) + +auxiliaries.insert(0, translate) diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h new file mode 100644 index 0000000000..ac6bc7088b --- /dev/null +++ b/src/gallium/auxiliary/translate/translate.h @@ -0,0 +1,82 @@ +/* + * Copyright 2008 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * Vertex fetch/store/convert code. This functionality is used in two places: + * 1. Vertex fetch/convert - to grab vertex data from incoming vertex + * arrays and convert to format needed by vertex shaders. + * 2. Vertex store/emit - to convert simple float[][4] vertex attributes + * (which is the organization used throughout the draw/prim pipeline) to + * hardware-specific formats and emit into hardware vertex buffers. + * + * + * Authors: + * Keith Whitwell + */ + +#ifndef _TRANSLATE_H +#define _TRANSLATE_H + + +#include "pipe/p_compiler.h" +#include "pipe/p_format.h" + +struct translate_element +{ + enum pipe_format input_format; + unsigned input_buffer; + unsigned input_offset; + + enum pipe_format output_format; + unsigned output_offset; +}; + + +struct translate { + void (*destroy)( struct translate * ); + + void (*set_buffer)( struct translate *, + unsigned i, + const void *ptr, + unsigned stride ); + + void (*run_elts)( struct translate *, + const unsigned *elts, + unsigned count, + void *output_buffer); +}; + + + +struct translate *translate_sse2_create( unsigned output_stride, + const struct translate_element *elements, + unsigned nr_elements ); + +struct translate *translate_generic_create( unsigned output_stride, + const struct translate_element *elements, + unsigned nr_elements ); + + +#endif diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c new file mode 100644 index 0000000000..643f6430c5 --- /dev/null +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -0,0 +1,630 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_state.h" +#include "translate.h" + + +#define DRAW_DBG 0 + +typedef void (*fetch_func)(const void *ptr, float *attrib); +typedef void (*emit_func)(const float *attrib, void *ptr); + + + +struct translate_generic { + struct translate translate; + + struct { + fetch_func fetch; + unsigned buffer; + unsigned input_offset; + + emit_func emit; + unsigned output_offset; + + char *input_ptr; + unsigned input_stride; + + } attrib[PIPE_MAX_ATTRIBS]; + + unsigned nr_attrib; + unsigned output_stride; +}; + + +static struct translate_generic *translate_generic( struct translate *translate ) +{ + return (struct translate_generic *)translate; +} + +/** + * Fetch a float[4] vertex attribute from memory, doing format/type + * conversion as needed. + * + * This is probably needed/dupliocated elsewhere, eg format + * conversion, texture sampling etc. + */ +#define ATTRIB( NAME, SZ, TYPE, FROM, TO ) \ +static void \ +fetch_##NAME(const void *ptr, float *attrib) \ +{ \ + const float defaults[4] = { 0,0,0,1 }; \ + int i; \ + \ + for (i = 0; i < SZ; i++) { \ + attrib[i] = FROM(i); \ + } \ + \ + for (; i < 4; i++) { \ + attrib[i] = defaults[i]; \ + } \ +} \ + \ +static void \ +emit_##NAME(const float *attrib, void *ptr) \ +{ \ + unsigned i; \ + TYPE *out = (TYPE *)ptr; \ + \ + for (i = 0; i < SZ; i++) { \ + out[i] = TO(attrib[i]); \ + } \ +} + + +#define FROM_64_FLOAT(i) ((float) ((double *) ptr)[i]) +#define FROM_32_FLOAT(i) (((float *) ptr)[i]) + +#define FROM_8_USCALED(i) ((float) ((unsigned char *) ptr)[i]) +#define FROM_16_USCALED(i) ((float) ((unsigned short *) ptr)[i]) +#define FROM_32_USCALED(i) ((float) ((unsigned int *) ptr)[i]) + +#define FROM_8_SSCALED(i) ((float) ((char *) ptr)[i]) +#define FROM_16_SSCALED(i) ((float) ((short *) ptr)[i]) +#define FROM_32_SSCALED(i) ((float) ((int *) ptr)[i]) + +#define FROM_8_UNORM(i) ((float) ((unsigned char *) ptr)[i] / 255.0f) +#define FROM_16_UNORM(i) ((float) ((unsigned short *) ptr)[i] / 65535.0f) +#define FROM_32_UNORM(i) ((float) ((unsigned int *) ptr)[i] / 4294967295.0f) + +#define FROM_8_SNORM(i) ((float) ((char *) ptr)[i] / 127.0f) +#define FROM_16_SNORM(i) ((float) ((short *) ptr)[i] / 32767.0f) +#define FROM_32_SNORM(i) ((float) ((int *) ptr)[i] / 2147483647.0f) + +#define TO_64_FLOAT(f) ((double) f) +#define TO_32_FLOAT(f) (f) + +#define TO_8_USCALED(f) ((unsigned char) f) +#define TO_16_USCALED(f) ((unsigned short) f) +#define TO_32_USCALED(f) ((unsigned int) f) + +#define TO_8_SSCALED(f) ((char) f) +#define TO_16_SSCALED(f) ((short) f) +#define TO_32_SSCALED(f) ((int) f) + +#define TO_8_UNORM(f) ((unsigned char) (f * 255.0f)) +#define TO_16_UNORM(f) ((unsigned short) (f * 65535.0f)) +#define TO_32_UNORM(f) ((unsigned int) (f * 4294967295.0f)) + +#define TO_8_SNORM(f) ((char) (f * 127.0f)) +#define TO_16_SNORM(f) ((short) (f * 32767.0f)) +#define TO_32_SNORM(f) ((int) (f * 2147483647.0f)) + + + +ATTRIB( R64G64B64A64_FLOAT, 4, double, FROM_64_FLOAT, TO_64_FLOAT ) +ATTRIB( R64G64B64_FLOAT, 3, double, FROM_64_FLOAT, TO_64_FLOAT ) +ATTRIB( R64G64_FLOAT, 2, double, FROM_64_FLOAT, TO_64_FLOAT ) +ATTRIB( R64_FLOAT, 1, double, FROM_64_FLOAT, TO_64_FLOAT ) + +ATTRIB( R32G32B32A32_FLOAT, 4, float, FROM_32_FLOAT, TO_32_FLOAT ) +ATTRIB( R32G32B32_FLOAT, 3, float, FROM_32_FLOAT, TO_32_FLOAT ) +ATTRIB( R32G32_FLOAT, 2, float, FROM_32_FLOAT, TO_32_FLOAT ) +ATTRIB( R32_FLOAT, 1, float, FROM_32_FLOAT, TO_32_FLOAT ) + +ATTRIB( R32G32B32A32_USCALED, 4, unsigned, FROM_32_USCALED, TO_32_USCALED ) +ATTRIB( R32G32B32_USCALED, 3, unsigned, FROM_32_USCALED, TO_32_USCALED ) +ATTRIB( R32G32_USCALED, 2, unsigned, FROM_32_USCALED, TO_32_USCALED ) +ATTRIB( R32_USCALED, 1, unsigned, FROM_32_USCALED, TO_32_USCALED ) + +ATTRIB( R32G32B32A32_SSCALED, 4, int, FROM_32_SSCALED, TO_32_SSCALED ) +ATTRIB( R32G32B32_SSCALED, 3, int, FROM_32_SSCALED, TO_32_SSCALED ) +ATTRIB( R32G32_SSCALED, 2, int, FROM_32_SSCALED, TO_32_SSCALED ) +ATTRIB( R32_SSCALED, 1, int, FROM_32_SSCALED, TO_32_SSCALED ) + +ATTRIB( R32G32B32A32_UNORM, 4, unsigned, FROM_32_UNORM, TO_32_UNORM ) +ATTRIB( R32G32B32_UNORM, 3, unsigned, FROM_32_UNORM, TO_32_UNORM ) +ATTRIB( R32G32_UNORM, 2, unsigned, FROM_32_UNORM, TO_32_UNORM ) +ATTRIB( R32_UNORM, 1, unsigned, FROM_32_UNORM, TO_32_UNORM ) + +ATTRIB( R32G32B32A32_SNORM, 4, int, FROM_32_SNORM, TO_32_SNORM ) +ATTRIB( R32G32B32_SNORM, 3, int, FROM_32_SNORM, TO_32_SNORM ) +ATTRIB( R32G32_SNORM, 2, int, FROM_32_SNORM, TO_32_SNORM ) +ATTRIB( R32_SNORM, 1, int, FROM_32_SNORM, TO_32_SNORM ) + +ATTRIB( R16G16B16A16_USCALED, 4, ushort, FROM_16_USCALED, TO_16_USCALED ) +ATTRIB( R16G16B16_USCALED, 3, ushort, FROM_16_USCALED, TO_16_USCALED ) +ATTRIB( R16G16_USCALED, 2, ushort, FROM_16_USCALED, TO_16_USCALED ) +ATTRIB( R16_USCALED, 1, ushort, FROM_16_USCALED, TO_16_USCALED ) + +ATTRIB( R16G16B16A16_SSCALED, 4, short, FROM_16_SSCALED, TO_16_SSCALED ) +ATTRIB( R16G16B16_SSCALED, 3, short, FROM_16_SSCALED, TO_16_SSCALED ) +ATTRIB( R16G16_SSCALED, 2, short, FROM_16_SSCALED, TO_16_SSCALED ) +ATTRIB( R16_SSCALED, 1, short, FROM_16_SSCALED, TO_16_SSCALED ) + +ATTRIB( R16G16B16A16_UNORM, 4, ushort, FROM_16_UNORM, TO_16_UNORM ) +ATTRIB( R16G16B16_UNORM, 3, ushort, FROM_16_UNORM, TO_16_UNORM ) +ATTRIB( R16G16_UNORM, 2, ushort, FROM_16_UNORM, TO_16_UNORM ) +ATTRIB( R16_UNORM, 1, ushort, FROM_16_UNORM, TO_16_UNORM ) + +ATTRIB( R16G16B16A16_SNORM, 4, short, FROM_16_SNORM, TO_16_SNORM ) +ATTRIB( R16G16B16_SNORM, 3, short, FROM_16_SNORM, TO_16_SNORM ) +ATTRIB( R16G16_SNORM, 2, short, FROM_16_SNORM, TO_16_SNORM ) +ATTRIB( R16_SNORM, 1, short, FROM_16_SNORM, TO_16_SNORM ) + +ATTRIB( R8G8B8A8_USCALED, 4, ubyte, FROM_8_USCALED, TO_8_USCALED ) +ATTRIB( R8G8B8_USCALED, 3, ubyte, FROM_8_USCALED, TO_8_USCALED ) +ATTRIB( R8G8_USCALED, 2, ubyte, FROM_8_USCALED, TO_8_USCALED ) +ATTRIB( R8_USCALED, 1, ubyte, FROM_8_USCALED, TO_8_USCALED ) + +ATTRIB( R8G8B8A8_SSCALED, 4, char, FROM_8_SSCALED, TO_8_SSCALED ) +ATTRIB( R8G8B8_SSCALED, 3, char, FROM_8_SSCALED, TO_8_SSCALED ) +ATTRIB( R8G8_SSCALED, 2, char, FROM_8_SSCALED, TO_8_SSCALED ) +ATTRIB( R8_SSCALED, 1, char, FROM_8_SSCALED, TO_8_SSCALED ) + +ATTRIB( R8G8B8A8_UNORM, 4, ubyte, FROM_8_UNORM, TO_8_UNORM ) +ATTRIB( R8G8B8_UNORM, 3, ubyte, FROM_8_UNORM, TO_8_UNORM ) +ATTRIB( R8G8_UNORM, 2, ubyte, FROM_8_UNORM, TO_8_UNORM ) +ATTRIB( R8_UNORM, 1, ubyte, FROM_8_UNORM, TO_8_UNORM ) + +ATTRIB( R8G8B8A8_SNORM, 4, char, FROM_8_SNORM, TO_8_SNORM ) +ATTRIB( R8G8B8_SNORM, 3, char, FROM_8_SNORM, TO_8_SNORM ) +ATTRIB( R8G8_SNORM, 2, char, FROM_8_SNORM, TO_8_SNORM ) +ATTRIB( R8_SNORM, 1, char, FROM_8_SNORM, TO_8_SNORM ) + +ATTRIB( A8R8G8B8_UNORM, 4, ubyte, FROM_8_UNORM, TO_8_UNORM ) +//ATTRIB( R8G8B8A8_UNORM, 4, ubyte, FROM_8_UNORM, TO_8_UNORM ) + + + +static void +fetch_B8G8R8A8_UNORM(const void *ptr, float *attrib) +{ + attrib[2] = FROM_8_UNORM(0); + attrib[1] = FROM_8_UNORM(1); + attrib[0] = FROM_8_UNORM(2); + attrib[3] = FROM_8_UNORM(3); +} + +static void +emit_B8G8R8A8_UNORM( const float *attrib, void *ptr) +{ + ubyte *out = (ubyte *)ptr; + out[2] = TO_8_UNORM(out[0]); + out[1] = TO_8_UNORM(out[1]); + out[0] = TO_8_UNORM(out[2]); + out[3] = TO_8_UNORM(out[3]); +} + +static void +fetch_NULL( const void *ptr, float *attrib ) +{ + attrib[0] = 0; + attrib[1] = 0; + attrib[2] = 0; + attrib[3] = 1; +} + +static void +emit_NULL( const float *attrib, void *ptr ) +{ + /* do nothing is the only sensible option */ +} + +static fetch_func get_fetch_func( enum pipe_format format ) +{ + switch (format) { + case PIPE_FORMAT_R64_FLOAT: + return fetch_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return fetch_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return fetch_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return fetch_R64G64B64A64_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + return fetch_R32_FLOAT; + case PIPE_FORMAT_R32G32_FLOAT: + return fetch_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return fetch_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return fetch_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return fetch_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return fetch_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return fetch_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return fetch_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return fetch_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return fetch_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return fetch_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return fetch_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return fetch_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return fetch_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return fetch_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return fetch_R32G32B32A32_SNORM; + + case PIPE_FORMAT_R32_SSCALED: + return fetch_R32_SSCALED; + case PIPE_FORMAT_R32G32_SSCALED: + return fetch_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return fetch_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return fetch_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return fetch_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return fetch_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return fetch_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return fetch_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return fetch_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return fetch_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return fetch_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return fetch_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return fetch_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return fetch_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return fetch_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return fetch_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return fetch_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return fetch_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return fetch_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return fetch_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return fetch_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return fetch_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return fetch_R8G8B8_UNORM; + case PIPE_FORMAT_R8G8B8A8_UNORM: + return fetch_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return fetch_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return fetch_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return fetch_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return fetch_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return fetch_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return fetch_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return fetch_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return fetch_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return fetch_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return fetch_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return fetch_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return fetch_R8G8B8A8_SSCALED; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + return fetch_A8R8G8B8_UNORM; + + case PIPE_FORMAT_B8G8R8A8_UNORM: + return fetch_B8G8R8A8_UNORM; + + default: + assert(0); + return fetch_NULL; + } +} + + + + +static emit_func get_emit_func( enum pipe_format format ) +{ + switch (format) { + case PIPE_FORMAT_R64_FLOAT: + return emit_R64_FLOAT; + case PIPE_FORMAT_R64G64_FLOAT: + return emit_R64G64_FLOAT; + case PIPE_FORMAT_R64G64B64_FLOAT: + return emit_R64G64B64_FLOAT; + case PIPE_FORMAT_R64G64B64A64_FLOAT: + return emit_R64G64B64A64_FLOAT; + + case PIPE_FORMAT_R32_FLOAT: + return emit_R32_FLOAT; + case PIPE_FORMAT_R32G32_FLOAT: + return emit_R32G32_FLOAT; + case PIPE_FORMAT_R32G32B32_FLOAT: + return emit_R32G32B32_FLOAT; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + return emit_R32G32B32A32_FLOAT; + + case PIPE_FORMAT_R32_UNORM: + return emit_R32_UNORM; + case PIPE_FORMAT_R32G32_UNORM: + return emit_R32G32_UNORM; + case PIPE_FORMAT_R32G32B32_UNORM: + return emit_R32G32B32_UNORM; + case PIPE_FORMAT_R32G32B32A32_UNORM: + return emit_R32G32B32A32_UNORM; + + case PIPE_FORMAT_R32_USCALED: + return emit_R32_USCALED; + case PIPE_FORMAT_R32G32_USCALED: + return emit_R32G32_USCALED; + case PIPE_FORMAT_R32G32B32_USCALED: + return emit_R32G32B32_USCALED; + case PIPE_FORMAT_R32G32B32A32_USCALED: + return emit_R32G32B32A32_USCALED; + + case PIPE_FORMAT_R32_SNORM: + return emit_R32_SNORM; + case PIPE_FORMAT_R32G32_SNORM: + return emit_R32G32_SNORM; + case PIPE_FORMAT_R32G32B32_SNORM: + return emit_R32G32B32_SNORM; + case PIPE_FORMAT_R32G32B32A32_SNORM: + return emit_R32G32B32A32_SNORM; + + case PIPE_FORMAT_R32_SSCALED: + return emit_R32_SSCALED; + case PIPE_FORMAT_R32G32_SSCALED: + return emit_R32G32_SSCALED; + case PIPE_FORMAT_R32G32B32_SSCALED: + return emit_R32G32B32_SSCALED; + case PIPE_FORMAT_R32G32B32A32_SSCALED: + return emit_R32G32B32A32_SSCALED; + + case PIPE_FORMAT_R16_UNORM: + return emit_R16_UNORM; + case PIPE_FORMAT_R16G16_UNORM: + return emit_R16G16_UNORM; + case PIPE_FORMAT_R16G16B16_UNORM: + return emit_R16G16B16_UNORM; + case PIPE_FORMAT_R16G16B16A16_UNORM: + return emit_R16G16B16A16_UNORM; + + case PIPE_FORMAT_R16_USCALED: + return emit_R16_USCALED; + case PIPE_FORMAT_R16G16_USCALED: + return emit_R16G16_USCALED; + case PIPE_FORMAT_R16G16B16_USCALED: + return emit_R16G16B16_USCALED; + case PIPE_FORMAT_R16G16B16A16_USCALED: + return emit_R16G16B16A16_USCALED; + + case PIPE_FORMAT_R16_SNORM: + return emit_R16_SNORM; + case PIPE_FORMAT_R16G16_SNORM: + return emit_R16G16_SNORM; + case PIPE_FORMAT_R16G16B16_SNORM: + return emit_R16G16B16_SNORM; + case PIPE_FORMAT_R16G16B16A16_SNORM: + return emit_R16G16B16A16_SNORM; + + case PIPE_FORMAT_R16_SSCALED: + return emit_R16_SSCALED; + case PIPE_FORMAT_R16G16_SSCALED: + return emit_R16G16_SSCALED; + case PIPE_FORMAT_R16G16B16_SSCALED: + return emit_R16G16B16_SSCALED; + case PIPE_FORMAT_R16G16B16A16_SSCALED: + return emit_R16G16B16A16_SSCALED; + + case PIPE_FORMAT_R8_UNORM: + return emit_R8_UNORM; + case PIPE_FORMAT_R8G8_UNORM: + return emit_R8G8_UNORM; + case PIPE_FORMAT_R8G8B8_UNORM: + return emit_R8G8B8_UNORM; + case PIPE_FORMAT_R8G8B8A8_UNORM: + return emit_R8G8B8A8_UNORM; + + case PIPE_FORMAT_R8_USCALED: + return emit_R8_USCALED; + case PIPE_FORMAT_R8G8_USCALED: + return emit_R8G8_USCALED; + case PIPE_FORMAT_R8G8B8_USCALED: + return emit_R8G8B8_USCALED; + case PIPE_FORMAT_R8G8B8A8_USCALED: + return emit_R8G8B8A8_USCALED; + + case PIPE_FORMAT_R8_SNORM: + return emit_R8_SNORM; + case PIPE_FORMAT_R8G8_SNORM: + return emit_R8G8_SNORM; + case PIPE_FORMAT_R8G8B8_SNORM: + return emit_R8G8B8_SNORM; + case PIPE_FORMAT_R8G8B8A8_SNORM: + return emit_R8G8B8A8_SNORM; + + case PIPE_FORMAT_R8_SSCALED: + return emit_R8_SSCALED; + case PIPE_FORMAT_R8G8_SSCALED: + return emit_R8G8_SSCALED; + case PIPE_FORMAT_R8G8B8_SSCALED: + return emit_R8G8B8_SSCALED; + case PIPE_FORMAT_R8G8B8A8_SSCALED: + return emit_R8G8B8A8_SSCALED; + + case PIPE_FORMAT_A8R8G8B8_UNORM: + return emit_A8R8G8B8_UNORM; + + case PIPE_FORMAT_B8G8R8A8_UNORM: + return emit_B8G8R8A8_UNORM; + + default: + assert(0); + return emit_NULL; + } +} + + + +/** + * Fetch vertex attributes for 'count' vertices. + */ +static void generic_run_elts( struct translate *translate, + const unsigned *elts, + unsigned count, + void *output_buffer ) +{ + struct translate_generic *tg = translate_generic(translate); + char *vert = output_buffer; + unsigned nr_attrs = tg->nr_attrib; + unsigned attr; + unsigned i; + + /* loop over vertex attributes (vertex shader inputs) + */ + for (i = 0; i < count; i++) { + unsigned elt = *elts++; + + for (attr = 0; attr < nr_attrs; attr++) { + float data[4]; + + const char *src = (tg->attrib[attr].input_ptr + + tg->attrib[attr].input_stride * elt); + + char *dst = (vert + + tg->attrib[attr].output_offset); + + tg->attrib[attr].fetch( src, data ); + tg->attrib[attr].emit( data, dst ); + } + + vert += tg->output_stride; + } +} + + + +static void generic_set_buffer( struct translate *translate, + unsigned buf, + const void *ptr, + unsigned stride ) +{ + struct translate_generic *tg = translate_generic(translate); + unsigned i; + + for (i = 0; i < tg->nr_attrib; i++) { + if (tg->attrib[i].buffer == buf) { + tg->attrib[i].input_ptr = ((char *)ptr + + tg->attrib[i].input_offset); + tg->attrib[i].input_stride = stride; + } + } +} + + +static void generic_destroy( struct translate *translate ) +{ + FREE(translate); +} + +struct translate *translate_generic_create( unsigned output_stride, + const struct translate_element *elements, + unsigned nr_elements ) +{ + struct translate_generic *tg = CALLOC_STRUCT(translate_generic); + unsigned i; + + if (tg == NULL) + return NULL; + + tg->translate.destroy = generic_destroy; + tg->translate.set_buffer = generic_set_buffer; + tg->translate.run_elts = generic_run_elts; + + for (i = 0; i < nr_elements; i++) { + tg->attrib[i].fetch = get_fetch_func(elements[i].input_format); + tg->attrib[i].buffer = elements[i].input_buffer; + tg->attrib[i].input_offset = elements[i].input_offset; + + tg->attrib[i].emit = get_emit_func(elements[i].output_format); + tg->attrib[i].output_offset = elements[i].output_offset; + } + + tg->nr_attrib = nr_elements; + + + return &tg->translate; +} -- cgit v1.2.3 From 59f68f36c4ad37496e3ed5d7d3142e3ae35dd351 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 15 Apr 2008 14:40:16 +0100 Subject: translate: typo in emit_B8G8R8A8_UNORM --- src/gallium/auxiliary/translate/translate_generic.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index 643f6430c5..7e75ba8365 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -231,10 +231,10 @@ static void emit_B8G8R8A8_UNORM( const float *attrib, void *ptr) { ubyte *out = (ubyte *)ptr; - out[2] = TO_8_UNORM(out[0]); - out[1] = TO_8_UNORM(out[1]); - out[0] = TO_8_UNORM(out[2]); - out[3] = TO_8_UNORM(out[3]); + out[2] = TO_8_UNORM(attrib[0]); + out[1] = TO_8_UNORM(attrib[1]); + out[0] = TO_8_UNORM(attrib[2]); + out[3] = TO_8_UNORM(attrib[3]); } static void -- cgit v1.2.3 From 7eb6f130a1dfb8179ff371eb4e75b47d6ee45d2a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 15 Apr 2008 14:44:39 +0100 Subject: draw: fix first glitch in vertex emit --- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 19 ++++++++++++++++--- 1 file changed, 16 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 557cd43f5b..889cab700c 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -92,6 +92,16 @@ static void emit_R32G32B32A32_FLOAT( const float *attrib, out[3] = attrib[3]; } +static void +emit_B8G8R8A8_UNORM( const float *attrib, void *ptr) +{ + ubyte *out = (ubyte *)ptr; + out[2] = float_to_ubyte(attrib[0]); + out[1] = float_to_ubyte(attrib[1]); + out[0] = float_to_ubyte(attrib[2]); + out[3] = float_to_ubyte(attrib[3]); +} + static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, unsigned prim ) { @@ -150,6 +160,9 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, src_buffer = 1; src_offset = 0; break; + case EMIT_4UB: + fpme->translate[i].emit = emit_B8G8R8A8_UNORM; + emit_sz = 4 * sizeof(ubyte); default: assert(0); fpme->translate[i].emit = emit_NULL; @@ -232,10 +245,10 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, for (j = 0; j < fpme->nr_translate; j++) { - const float *attrib = (const float *)( (*fpme->translate[i].input_buf) + - fpme->translate[i].input_offset ); + const float *attrib = (const float *)( (*fpme->translate[j].input_buf) + + fpme->translate[j].input_offset ); - char *dest = out_buf + fpme->translate[i].output_offset; + char *dest = out_buf + fpme->translate[j].output_offset; /*debug_printf("emiting [%f, %f, %f, %f]\n", attrib[0], attrib[1], -- cgit v1.2.3 From 8cac6f3fcf9d6c08959efc20f8fce9eddbdcd0ef Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 15 Apr 2008 14:52:56 +0100 Subject: draw: increment output vertex properly --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 889cab700c..75bbcc8d41 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -250,14 +250,16 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, char *dest = out_buf + fpme->translate[j].output_offset; - /*debug_printf("emiting [%f, %f, %f, %f]\n", - attrib[0], attrib[1], - attrib[2], attrib[3]);*/ + if (0) + debug_printf("emiting [%f, %f, %f, %f]\n", + attrib[0], attrib[1], + attrib[2], attrib[3]); fpme->translate[j].emit(attrib, dest); } fpme->input_buf[0] += fpme->pipeline_vertex_size; + out_buf += fpme->hw_vertex_size; } draw->render->draw(draw->render, -- cgit v1.2.3 From a5a7dc24ba2dcf9bbdd73709c4c182e324bdc3a5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 15 Apr 2008 19:14:31 +0100 Subject: gallium: Switch one vertex path over to new translate module Will eventually do this for all instances where we are converting vertices from one format to another. --- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 146 +++++++-------------- src/gallium/auxiliary/translate/translate.h | 36 ++++- .../auxiliary/translate/translate_generic.c | 66 ++++++++-- 3 files changed, 128 insertions(+), 120 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 75bbcc8d41..2d83e11940 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -31,77 +31,20 @@ #include "draw/draw_vbuf.h" #include "draw/draw_vertex.h" #include "draw/draw_pt.h" +#include "translate/translate.h" + struct fetch_pipeline_middle_end { struct draw_pt_middle_end base; struct draw_context *draw; - const ubyte *input_buf[2]; - - struct { - const ubyte **input_buf; - unsigned input_offset; - unsigned output_offset; - - void (*emit)( const float *attrib, void *ptr ); - } translate[PIPE_MAX_ATTRIBS]; - unsigned nr_translate; + struct translate *translate; unsigned pipeline_vertex_size; - unsigned hw_vertex_size; unsigned prim; }; -static void emit_NULL( const float *attrib, - void *ptr ) -{ -} - -static void emit_R32_FLOAT( const float *attrib, - void *ptr ) -{ - float *out = (float *)ptr; - out[0] = attrib[0]; -} - -static void emit_R32G32_FLOAT( const float *attrib, - void *ptr ) -{ - float *out = (float *)ptr; - out[0] = attrib[0]; - out[1] = attrib[1]; -} - -static void emit_R32G32B32_FLOAT( const float *attrib, - void *ptr ) -{ - float *out = (float *)ptr; - out[0] = attrib[0]; - out[1] = attrib[1]; - out[2] = attrib[2]; -} - -static void emit_R32G32B32A32_FLOAT( const float *attrib, - void *ptr ) -{ - float *out = (float *)ptr; - out[0] = attrib[0]; - out[1] = attrib[1]; - out[2] = attrib[2]; - out[3] = attrib[3]; -} - -static void -emit_B8G8R8A8_UNORM( const float *attrib, void *ptr) -{ - ubyte *out = (ubyte *)ptr; - out[2] = float_to_ubyte(attrib[0]); - out[1] = float_to_ubyte(attrib[1]); - out[0] = float_to_ubyte(attrib[2]); - out[3] = float_to_ubyte(attrib[3]); -} - static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, unsigned prim ) { @@ -111,6 +54,7 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, boolean ok; const struct vertex_info *vinfo; unsigned dst_offset; + struct translate_key hw_key; fpme->prim = prim; @@ -132,6 +76,7 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, for (i = 0; i < vinfo->num_attribs; i++) { unsigned emit_sz = 0; unsigned src_buffer = 0; + unsigned output_format; unsigned src_offset = (sizeof(struct vertex_header) + vinfo->src_index[i] * 4 * sizeof(float) ); @@ -139,49 +84,64 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, switch (vinfo->emit[i]) { case EMIT_4F: - fpme->translate[i].emit = emit_R32G32B32A32_FLOAT; + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; emit_sz = 4 * sizeof(float); break; case EMIT_3F: - fpme->translate[i].emit = emit_R32G32B32_FLOAT; + output_format = PIPE_FORMAT_R32G32B32_FLOAT; emit_sz = 3 * sizeof(float); break; case EMIT_2F: - fpme->translate[i].emit = emit_R32G32_FLOAT; + output_format = PIPE_FORMAT_R32G32_FLOAT; emit_sz = 2 * sizeof(float); break; case EMIT_1F: - fpme->translate[i].emit = emit_R32_FLOAT; + output_format = PIPE_FORMAT_R32_FLOAT; emit_sz = 1 * sizeof(float); break; case EMIT_1F_PSIZE: - fpme->translate[i].emit = emit_R32_FLOAT; + output_format = PIPE_FORMAT_R32_FLOAT; emit_sz = 1 * sizeof(float); src_buffer = 1; src_offset = 0; break; case EMIT_4UB: - fpme->translate[i].emit = emit_B8G8R8A8_UNORM; + output_format = PIPE_FORMAT_B8G8R8A8_UNORM; emit_sz = 4 * sizeof(ubyte); default: assert(0); - fpme->translate[i].emit = emit_NULL; + output_format = PIPE_FORMAT_NONE; emit_sz = 0; break; } + + hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + hw_key.element[i].input_buffer = src_buffer; + hw_key.element[i].input_offset = src_offset; + hw_key.element[i].output_format = output_format; + hw_key.element[i].output_offset = dst_offset; - fpme->translate[i].input_buf = &fpme->input_buf[src_buffer]; - fpme->translate[i].input_offset = src_offset; - fpme->translate[i].output_offset = dst_offset; dst_offset += emit_sz; } - fpme->nr_translate = vinfo->num_attribs; - fpme->hw_vertex_size = vinfo->size * 4; + hw_key.nr_elements = vinfo->num_attribs; + hw_key.output_stride = vinfo->size * 4; + + /* Don't bother with caching at this stage: + */ + if (!fpme->translate || + memcmp(&fpme->translate->key, &hw_key, sizeof(hw_key)) != 0) + { + if (fpme->translate) + fpme->translate->release(fpme->translate); + + fpme->translate = translate_generic_create( &hw_key ); + } + + //fpme->pipeline_vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); fpme->pipeline_vertex_size = MAX_VERTEX_ALLOCATION; - fpme->hw_vertex_size = vinfo->size * 4; } @@ -221,46 +181,34 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, draw_elts, draw_count ); } else { - unsigned i, j; + struct translate *translate = fpme->translate; void *hw_verts; - char *out_buf; /* XXX: need to flush to get prim_vbuf.c to release its allocation?? */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); hw_verts = draw->render->allocate_vertices(draw->render, - (ushort)fpme->hw_vertex_size, + (ushort)fpme->translate->key.output_stride, (ushort)fetch_count); if (!hw_verts) { assert(0); return; } - out_buf = (char *)hw_verts; - fpme->input_buf[0] = (const ubyte *)pipeline_verts; - fpme->input_buf[1] = (const ubyte *)&fpme->draw->rasterizer->point_size; - - for (i = 0; i < fetch_count; i++) { - - for (j = 0; j < fpme->nr_translate; j++) { + translate->set_buffer(translate, + 0, + pipeline_verts, + fpme->pipeline_vertex_size ); - const float *attrib = (const float *)( (*fpme->translate[j].input_buf) + - fpme->translate[j].input_offset ); + translate->set_buffer(translate, + 1, + &fpme->draw->rasterizer->point_size, + 0); - char *dest = out_buf + fpme->translate[j].output_offset; - - if (0) - debug_printf("emiting [%f, %f, %f, %f]\n", - attrib[0], attrib[1], - attrib[2], attrib[3]); - - fpme->translate[j].emit(attrib, dest); - } - - fpme->input_buf[0] += fpme->pipeline_vertex_size; - out_buf += fpme->hw_vertex_size; - } + translate->run( translate, + 0, fetch_count, + hw_verts ); draw->render->draw(draw->render, draw_elts, @@ -268,7 +216,7 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, draw->render->release_vertices(draw->render, hw_verts, - fpme->hw_vertex_size, + fpme->translate->key.output_stride, fetch_count); } diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h index ac6bc7088b..4f9f40e51a 100644 --- a/src/gallium/auxiliary/translate/translate.h +++ b/src/gallium/auxiliary/translate/translate.h @@ -54,8 +54,17 @@ struct translate_element }; +struct translate_key { + unsigned output_stride; + unsigned nr_elements; + struct translate_element element[PIPE_MAX_ATTRIBS]; +}; + + struct translate { - void (*destroy)( struct translate * ); + struct translate_key key; + + void (*release)( struct translate * ); void (*set_buffer)( struct translate *, unsigned i, @@ -66,17 +75,30 @@ struct translate { const unsigned *elts, unsigned count, void *output_buffer); + + void (*run)( struct translate *, + unsigned start, + unsigned count, + void *output_buffer); }; -struct translate *translate_sse2_create( unsigned output_stride, - const struct translate_element *elements, - unsigned nr_elements ); +#if 0 +struct translate_context *translate_context_create( void ); +void translate_context_destroy( struct translate_context * ); + +struct translate *translate_lookup_or_create( struct translate_context *tctx, + const struct translate_key *key ); +#endif + + +/******************************************************************************* + * Private: + */ +struct translate *translate_sse2_create( const struct translate_key *key ); -struct translate *translate_generic_create( unsigned output_stride, - const struct translate_element *elements, - unsigned nr_elements ); +struct translate *translate_generic_create( const struct translate_key *key ); #endif diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index 7e75ba8365..fc9060900b 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -59,7 +59,6 @@ struct translate_generic { } attrib[PIPE_MAX_ATTRIBS]; unsigned nr_attrib; - unsigned output_stride; }; @@ -571,7 +570,42 @@ static void generic_run_elts( struct translate *translate, tg->attrib[attr].emit( data, dst ); } - vert += tg->output_stride; + vert += tg->translate.key.output_stride; + } +} + + + +static void generic_run( struct translate *translate, + unsigned start, + unsigned count, + void *output_buffer ) +{ + struct translate_generic *tg = translate_generic(translate); + char *vert = output_buffer; + unsigned nr_attrs = tg->nr_attrib; + unsigned attr; + unsigned i; + + /* loop over vertex attributes (vertex shader inputs) + */ + for (i = 0; i < count; i++) { + unsigned elt = start + i; + + for (attr = 0; attr < nr_attrs; attr++) { + float data[4]; + + const char *src = (tg->attrib[attr].input_ptr + + tg->attrib[attr].input_stride * elt); + + char *dst = (vert + + tg->attrib[attr].output_offset); + + tg->attrib[attr].fetch( src, data ); + tg->attrib[attr].emit( data, dst ); + } + + vert += tg->translate.key.output_stride; } } @@ -595,14 +629,14 @@ static void generic_set_buffer( struct translate *translate, } -static void generic_destroy( struct translate *translate ) +static void generic_release( struct translate *translate ) { + /* Refcount? + */ FREE(translate); } -struct translate *translate_generic_create( unsigned output_stride, - const struct translate_element *elements, - unsigned nr_elements ) +struct translate *translate_generic_create( const struct translate_key *key ) { struct translate_generic *tg = CALLOC_STRUCT(translate_generic); unsigned i; @@ -610,20 +644,24 @@ struct translate *translate_generic_create( unsigned output_stride, if (tg == NULL) return NULL; - tg->translate.destroy = generic_destroy; + tg->translate.key = *key; + tg->translate.release = generic_release; tg->translate.set_buffer = generic_set_buffer; tg->translate.run_elts = generic_run_elts; + tg->translate.run = generic_run; + + for (i = 0; i < key->nr_elements; i++) { + + tg->attrib[i].fetch = get_fetch_func(key->element[i].input_format); + tg->attrib[i].buffer = key->element[i].input_buffer; + tg->attrib[i].input_offset = key->element[i].input_offset; - for (i = 0; i < nr_elements; i++) { - tg->attrib[i].fetch = get_fetch_func(elements[i].input_format); - tg->attrib[i].buffer = elements[i].input_buffer; - tg->attrib[i].input_offset = elements[i].input_offset; + tg->attrib[i].emit = get_emit_func(key->element[i].output_format); + tg->attrib[i].output_offset = key->element[i].output_offset; - tg->attrib[i].emit = get_emit_func(elements[i].output_format); - tg->attrib[i].output_offset = elements[i].output_offset; } - tg->nr_attrib = nr_elements; + tg->nr_attrib = key->nr_elements; return &tg->translate; -- cgit v1.2.3 From e2269e94c0ecaa97c7153815489e59c9ca64c243 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Apr 2008 11:39:22 +0900 Subject: gallium: Build and link the translate module. --- src/gallium/SConscript | 1 + src/gallium/auxiliary/translate/SConscript | 2 +- 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/SConscript b/src/gallium/SConscript index f09778ce99..2653f91bd2 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -18,6 +18,7 @@ SConscript([ 'auxiliary/rtasm/SConscript', 'auxiliary/tgsi/SConscript', 'auxiliary/cso_cache/SConscript', + 'auxiliary/translate/SConscript', 'auxiliary/draw/SConscript', 'auxiliary/pipebuffer/SConscript', ]) diff --git a/src/gallium/auxiliary/translate/SConscript b/src/gallium/auxiliary/translate/SConscript index 947b9438cb..42513ba3b0 100644 --- a/src/gallium/auxiliary/translate/SConscript +++ b/src/gallium/auxiliary/translate/SConscript @@ -1,6 +1,6 @@ Import('*') -cso_cache = env.ConvenienceLibrary( +translate = env.ConvenienceLibrary( target = 'translate', source = [ 'translate_generic.c', -- cgit v1.2.3 From fd6acabd2f62fe006b078ae7640a944c7f65903c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Apr 2008 11:42:08 +0900 Subject: gallium: Get the translate module to build on msvc. Appearently MSVC c-preprocessor parses "255.0f" as two tokens: "255.0" and "f", and performs variable substitution on "f". --- .../auxiliary/translate/translate_generic.c | 32 +++++++++++----------- 1 file changed, 16 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index fc9060900b..e7fb1dcb2d 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -78,8 +78,8 @@ static struct translate_generic *translate_generic( struct translate *translate static void \ fetch_##NAME(const void *ptr, float *attrib) \ { \ - const float defaults[4] = { 0,0,0,1 }; \ - int i; \ + const float defaults[4] = { 0.0f,0.0f,0.0f,1.0f }; \ + unsigned i; \ \ for (i = 0; i < SZ; i++) { \ attrib[i] = FROM(i); \ @@ -121,24 +121,24 @@ emit_##NAME(const float *attrib, void *ptr) \ #define FROM_16_SNORM(i) ((float) ((short *) ptr)[i] / 32767.0f) #define FROM_32_SNORM(i) ((float) ((int *) ptr)[i] / 2147483647.0f) -#define TO_64_FLOAT(f) ((double) f) -#define TO_32_FLOAT(f) (f) +#define TO_64_FLOAT(x) ((double) x) +#define TO_32_FLOAT(x) (x) -#define TO_8_USCALED(f) ((unsigned char) f) -#define TO_16_USCALED(f) ((unsigned short) f) -#define TO_32_USCALED(f) ((unsigned int) f) +#define TO_8_USCALED(x) ((unsigned char) x) +#define TO_16_USCALED(x) ((unsigned short) x) +#define TO_32_USCALED(x) ((unsigned int) x) -#define TO_8_SSCALED(f) ((char) f) -#define TO_16_SSCALED(f) ((short) f) -#define TO_32_SSCALED(f) ((int) f) +#define TO_8_SSCALED(x) ((char) x) +#define TO_16_SSCALED(x) ((short) x) +#define TO_32_SSCALED(x) ((int) x) -#define TO_8_UNORM(f) ((unsigned char) (f * 255.0f)) -#define TO_16_UNORM(f) ((unsigned short) (f * 65535.0f)) -#define TO_32_UNORM(f) ((unsigned int) (f * 4294967295.0f)) +#define TO_8_UNORM(x) ((unsigned char) (x * 255.0f)) +#define TO_16_UNORM(x) ((unsigned short) (x * 65535.0f)) +#define TO_32_UNORM(x) ((unsigned int) (x * 4294967295.0f)) -#define TO_8_SNORM(f) ((char) (f * 127.0f)) -#define TO_16_SNORM(f) ((short) (f * 32767.0f)) -#define TO_32_SNORM(f) ((int) (f * 2147483647.0f)) +#define TO_8_SNORM(x) ((char) (x * 127.0f)) +#define TO_16_SNORM(x) ((short) (x * 32767.0f)) +#define TO_32_SNORM(x) ((int) (x * 2147483647.0f)) -- cgit v1.2.3 From a8582efaca35d09c8ca18918a243a9284583356d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 16 Apr 2008 10:03:18 +0100 Subject: draw: make pt run pipeline when need_pipeline is true, not just when clipped --- src/gallium/auxiliary/draw/draw_private.h | 13 +- src/gallium/auxiliary/draw/draw_pt.c | 182 +++++---------------- src/gallium/auxiliary/draw/draw_pt.h | 6 +- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 3 +- .../auxiliary/draw/draw_pt_fetch_pipeline.c | 3 +- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 18 +- src/gallium/auxiliary/draw/draw_pt_vcache.c | 5 +- 7 files changed, 75 insertions(+), 155 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index c8cb96c8ba..8ac061cc9f 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -182,6 +182,12 @@ typedef void (*pt_fetch_func)( struct draw_context *draw, struct vbuf_render; + +#define PT_SHADE 0x1 +#define PT_CLIPTEST 0x2 +#define PT_PIPELINE 0x4 +#define PT_MAX_MIDDLE 0x8 + /** * Private context for the drawing module. */ @@ -219,15 +225,10 @@ struct draw_context unsigned hw_vertex_size; /* XXX: to be removed */ struct { - struct draw_pt_middle_end *fetch_emit; - struct draw_pt_middle_end *fetch_pipeline; - struct draw_pt_middle_end *fetch_shade_emit; - struct draw_pt_middle_end *fetch_shade_cliptest_pipeline_or_emit; + struct draw_pt_middle_end *opt[PT_MAX_MIDDLE]; } middle; struct { - struct draw_pt_front_end *noop; - struct draw_pt_front_end *split_arrays; struct draw_pt_front_end *vcache; } front; diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 3d2e7bf7b8..895224c952 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -36,146 +36,58 @@ #include "draw/draw_pt.h" -#if 0 -static boolean too_many_elts( struct draw_context *draw, - unsigned elts ) -{ - return elts > (8 * 1024); -} -#endif - -static INLINE unsigned reduced_prim(unsigned prim) -{ - /*FIXME*/ - return prim; -} -static INLINE boolean good_prim(unsigned prim) -{ - /*FIXME*/ - return FALSE; -} +/* Overall we split things into: + * - frontend -- prepare fetch_elts, draw_elts - eg vcache + * - middle -- fetch, shade, cliptest, viewport + * - pipeline -- the prim pipeline: clipping, wide lines, etc + * - backend -- the vbuf_render provided by the driver. + */ boolean draw_pt_arrays(struct draw_context *draw, unsigned prim, unsigned start, unsigned count) { - const boolean pipeline = draw_need_pipeline(draw, prim); - const boolean cliptest = !draw->rasterizer->bypass_clipping; - const boolean shading = !draw->rasterizer->bypass_vs; struct draw_pt_front_end *frontend = NULL; struct draw_pt_middle_end *middle = NULL; + unsigned opt = 0; if (!draw->render) return FALSE; /*debug_printf("XXXXXXXXXX needs_pipeline = %d\n", pipeline);*/ - /* Overall we do: - * - frontend -- prepare fetch_elts, draw_elts - eg vcache - * - middle -- fetch, shade, cliptest, viewport - * - pipeline -- the prim pipeline: clipping, wide lines, etc - * - backend -- the vbuf_render provided by the driver. - */ - - if (shading && !draw->use_pt_shaders) - return FALSE; - - if (!cliptest && !pipeline && !shading) { - /* This is the 'passthrough' path: - */ - /* Fetch user verts, emit hw verts: - */ - middle = draw->pt.middle.fetch_emit; - } - else if (!cliptest && !shading) { - /* This is the 'passthrough' path targetting the pipeline backend. - */ - /* Fetch user verts, emit pipeline verts, run pipeline: - */ - middle = draw->pt.middle.fetch_pipeline; - } - else if (!cliptest && !pipeline) { - /* Fetch user verts, run vertex shader, emit hw verts: - */ - middle = draw->pt.middle.fetch_shade_emit; - } - else if (!pipeline) { - /* Even though !pipeline, we have to run it to get clipping. We - * do know that the pipeline is just the clipping operation, but - * that probably doesn't help much. - * - * This is going to be the most important path for a lot of - * swtnl cards. - */ - /* Fetch user verts, - * run vertex shader, - * cliptest and viewport trasform - * if no clipped vertices, - * emit hw verts - * else - * run pipline - */ - middle = draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit; - } - else { - /* This is what we're currently always doing: - */ - /* Fetch user verts, run vertex shader, cliptest, run pipeline - * or - * Fetch user verts, run vertex shader, run pipeline - */ - middle = draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit; + if (draw_need_pipeline(draw, prim)) { + opt |= PT_PIPELINE; } + if (!draw->rasterizer->bypass_clipping) { + opt |= PT_CLIPTEST; + } - /* If !pipeline, need to make sure we respect the driver's limited - * capabilites to receive blocks of vertex data and elements. - */ -#if 0 - if (!pipeline) { - unsigned vertex_mode = passthrough; - unsigned nr_verts = count_vertices( draw, start, count ); - unsigned hw_prim = prim; - - if (is_elts(draw)) { - frontend = draw->pt.front.vcache; - hw_prim = reduced_prim(prim); - } -#if 0 - if (too_many_verts(nr_verts)) { - /* if (is_verts(draw) && can_split(prim)) { - draw = draw_arrays_split; - } - else */ { - frontend = draw->pt.front.vcache; - hw_prim = reduced_prim(prim); - } - } -#endif + if (!draw->rasterizer->bypass_vs) { + opt |= PT_SHADE; - if (too_many_elts(count)) { + if (!draw->use_pt_shaders) + return FALSE; + } - /* if (is_elts(draw) && can_split(prim)) { - draw = draw_elts_split; - } - else */ { - frontend = draw->pt.front.vcache; - hw_prim = reduced_prim(prim); - } - } - if (!good_prim(hw_prim)) { - frontend = draw->pt.front.vcache; - } + if (draw->pt.middle.opt[opt] == NULL) { + opt = PT_PIPELINE | PT_CLIPTEST | PT_SHADE; } -#else + + middle = draw->pt.middle.opt[opt]; + assert(middle); + + /* May create a short-circuited version of this for small primitives: + */ frontend = draw->pt.front.vcache; -#endif - frontend->prepare( frontend, prim, middle ); + + frontend->prepare( frontend, prim, middle, opt ); frontend->run( frontend, draw_pt_elt_func( draw ), @@ -190,21 +102,19 @@ draw_pt_arrays(struct draw_context *draw, boolean draw_pt_init( struct draw_context *draw ) { - draw->pt.middle.fetch_emit = draw_pt_fetch_emit( draw ); - if (!draw->pt.middle.fetch_emit) - return FALSE; - - draw->pt.middle.fetch_pipeline = draw_pt_fetch_pipeline( draw ); - if (!draw->pt.middle.fetch_pipeline) + draw->pt.front.vcache = draw_pt_vcache( draw ); + if (!draw->pt.front.vcache) return FALSE; - draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit = + draw->pt.middle.opt[0] = draw_pt_fetch_emit( draw ); + draw->pt.middle.opt[PT_PIPELINE] = draw_pt_fetch_pipeline( draw ); +// draw->pt.middle.opt[PT_SHADE] = draw_pt_shade_emit( draw ); +// draw->pt.middle.opt[PT_SHADE | PT_PIPELINE] = draw_pt_shade_pipeline( draw ); +// draw->pt.middle.opt[PT_SHADE | PT_CLIPTEST] = draw_pt_shade_clip_either( draw ); + draw->pt.middle.opt[PT_SHADE | PT_CLIPTEST | PT_PIPELINE] = draw_pt_fetch_pipeline_or_emit( draw ); - if (!draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit) - return FALSE; - draw->pt.front.vcache = draw_pt_vcache( draw ); - if (!draw->pt.front.vcache) + if (!draw->pt.middle.opt[PT_SHADE | PT_CLIPTEST | PT_PIPELINE]) return FALSE; return TRUE; @@ -213,21 +123,13 @@ boolean draw_pt_init( struct draw_context *draw ) void draw_pt_destroy( struct draw_context *draw ) { - if (draw->pt.middle.fetch_emit) { - draw->pt.middle.fetch_emit->destroy( draw->pt.middle.fetch_emit ); - draw->pt.middle.fetch_emit = NULL; - } + int i; - if (draw->pt.middle.fetch_pipeline) { - draw->pt.middle.fetch_pipeline->destroy( draw->pt.middle.fetch_pipeline ); - draw->pt.middle.fetch_pipeline = NULL; - } - - if (draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit) { - draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit->destroy( - draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit ); - draw->pt.middle.fetch_shade_cliptest_pipeline_or_emit = NULL; - } + for (i = 0; i < PT_MAX_MIDDLE; i++) + if (draw->pt.middle.opt[i]) { + draw->pt.middle.opt[i]->destroy( draw->pt.middle.opt[i] ); + draw->pt.middle.opt[i] = NULL; + } if (draw->pt.front.vcache) { draw->pt.front.vcache->destroy( draw->pt.front.vcache ); diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 48413b648a..08afb60645 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -64,7 +64,8 @@ struct draw_context; struct draw_pt_front_end { void (*prepare)( struct draw_pt_front_end *, unsigned prim, - struct draw_pt_middle_end * ); + struct draw_pt_middle_end *, + unsigned opt ); void (*run)( struct draw_pt_front_end *, pt_elt_func elt_func, @@ -90,7 +91,8 @@ struct draw_pt_front_end { */ struct draw_pt_middle_end { void (*prepare)( struct draw_pt_middle_end *, - unsigned prim ); + unsigned prim, + unsigned opt ); void (*run)( struct draw_pt_middle_end *, const unsigned *fetch_elts, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 3a26a5d712..1a9d331a06 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -164,7 +164,8 @@ fetch_store_general( struct fetch_emit_middle_end *feme, static void fetch_emit_prepare( struct draw_pt_middle_end *middle, - unsigned prim ) + unsigned prim, + unsigned opt ) { struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; struct draw_context *draw = feme->draw; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c index a70d129c93..db6d6c76ea 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c @@ -209,7 +209,8 @@ fetch_store_general( struct fetch_pipeline_middle_end *fpme, * */ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, - unsigned prim ) + unsigned prim, + unsigned opt ) { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 2d83e11940..52be4a64aa 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -42,11 +42,13 @@ struct fetch_pipeline_middle_end { unsigned pipeline_vertex_size; unsigned prim; + unsigned opt; }; static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, - unsigned prim ) + unsigned prim, + unsigned opt ) { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; @@ -57,6 +59,7 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, struct translate_key hw_key; fpme->prim = prim; + fpme->opt = opt; ok = draw->render->set_primitive(draw->render, prim); if (!ok) { @@ -157,6 +160,7 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, struct draw_context *draw = fpme->draw; struct draw_vertex_shader *shader = draw->vertex_shader; char *pipeline_verts; + unsigned pipeline = PT_PIPELINE; pipeline_verts = MALLOC(fpme->pipeline_vertex_size * fetch_count); @@ -170,9 +174,17 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, /* Shade */ shader->prepare(shader, draw); + if (shader->run(shader, draw, fetch_elts, fetch_count, pipeline_verts, - fpme->pipeline_vertex_size)) { - /* Run the pipeline */ + fpme->pipeline_vertex_size)) + { + pipeline |= PT_CLIPTEST; + } + + + /* Do we need to run the pipeline? + */ + if (fpme->opt & pipeline) { draw_pt_run_pipeline( fpme->draw, fpme->prim, pipeline_verts, diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 107dcfc269..5561f2b6fb 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -448,7 +448,8 @@ static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { static void vcache_prepare( struct draw_pt_front_end *frontend, unsigned prim, - struct draw_pt_middle_end *middle ) + struct draw_pt_middle_end *middle, + unsigned opt ) { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; @@ -464,7 +465,7 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, vcache->output_prim = reduced_prim[prim]; vcache->middle = middle; - middle->prepare( middle, vcache->output_prim ); + middle->prepare( middle, vcache->output_prim, opt ); } -- cgit v1.2.3 From 909e8ce543a6c1e97d55791d2069cbdc56ea9db6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 16 Apr 2008 10:12:12 +0100 Subject: draw: update debug code --- src/gallium/auxiliary/draw/draw_vbuf.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vbuf.c b/src/gallium/auxiliary/draw/draw_vbuf.c index e3216ff711..eeb194afaa 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_vbuf.c @@ -113,12 +113,12 @@ check_space( struct vbuf_stage *vbuf, unsigned nr ) } -#if 0 +#if 1 static INLINE void dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) { - assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); - unsigned i, j, k; +// assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); + unsigned i, j; for (i = 0; i < vinfo->num_attribs; i++) { j = vinfo->src_index[i]; @@ -264,6 +264,8 @@ emit_vertex( struct vbuf_stage *vbuf, } else { draw_vf_emit_vertex(vbuf->vf, vertex, vbuf->vertex_ptr); + + if (0) dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); vbuf->vertex_ptr += vbuf->vertex_size/4; } -- cgit v1.2.3 From 8b607f42d094e61432c5718b8baa0a68268ec150 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 16 Apr 2008 12:20:39 +0100 Subject: softpipe: call setup_prepare earlier so that vertex info is correct when queried --- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 4fed19ecb6..e063fe82ef 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -106,6 +106,16 @@ static boolean sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); + + /* XXX: break this dependency - make setup_context live under + * softpipe, rename the old "setup" draw stage to something else. + */ + struct setup_context *setup_ctx = sp_draw_setup_context(cvbr->softpipe->setup); + + setup_prepare( setup_ctx ); + + + if (prim == PIPE_PRIM_TRIANGLES || prim == PIPE_PRIM_LINES || prim == PIPE_PRIM_POINTS) { @@ -136,10 +146,6 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) */ struct draw_stage *setup = softpipe->setup; struct setup_context *setup_ctx = sp_draw_setup_context(softpipe->setup); - - /* XXX: call this from allocate_vertices: - */ - setup_prepare( setup_ctx ); switch (cvbr->prim) { -- cgit v1.2.3 From d8c389171872b69af3c94ebab02ad5f4bcd2d4df Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 16 Apr 2008 12:31:19 +0100 Subject: draw: remove vbuf non-vf debug path, fix some failure modes in draw_vbuf_create() --- src/gallium/auxiliary/draw/draw_vbuf.c | 106 ++++++--------------------------- 1 file changed, 19 insertions(+), 87 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vbuf.c b/src/gallium/auxiliary/draw/draw_vbuf.c index eeb194afaa..59b63f5bfa 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_vbuf.c @@ -113,7 +113,6 @@ check_space( struct vbuf_stage *vbuf, unsigned nr ) } -#if 1 static INLINE void dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) { @@ -167,7 +166,6 @@ dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) } debug_printf("\n"); } -#endif /** @@ -181,94 +179,22 @@ static INLINE void emit_vertex( struct vbuf_stage *vbuf, struct vertex_header *vertex ) { -#if 0 - debug_printf("emit vertex %d to %p\n", - vbuf->nr_vertices, vbuf->vertex_ptr); -#endif - if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { if(vertex->vertex_id < vbuf->nr_vertices) return; else debug_printf("Bad vertex id 0x%04x (>= 0x%04x)\n", - vertex->vertex_id, vbuf->nr_vertices); + vertex->vertex_id, vbuf->nr_vertices); return; } vertex->vertex_id = vbuf->nr_vertices++; - if(!vbuf->vf) { - const struct vertex_info *vinfo = vbuf->vinfo; - uint i; - uint count = 0; /* for debug/sanity */ - - assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); - - for (i = 0; i < vinfo->num_attribs; i++) { - uint j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; - case EMIT_1F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - count++; - break; - case EMIT_1F_PSIZE: - *vbuf->vertex_ptr++ = fui(vbuf->stage.draw->rasterizer->point_size); - count++; - break; - case EMIT_2F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - count += 2; - break; - case EMIT_3F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); - count += 3; - break; - case EMIT_4F: - *vbuf->vertex_ptr++ = fui(vertex->data[j][0]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][1]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][2]); - *vbuf->vertex_ptr++ = fui(vertex->data[j][3]); - count += 4; - break; - case EMIT_4UB: - *vbuf->vertex_ptr++ = pack_ub4(float_to_ubyte( vertex->data[j][2] ), - float_to_ubyte( vertex->data[j][1] ), - float_to_ubyte( vertex->data[j][0] ), - float_to_ubyte( vertex->data[j][3] )); - count += 1; - break; - default: - assert(0); - } - } - assert(count == vinfo->size); -#if 0 - { - static float data[256]; - draw_vf_emit_vertex(vbuf->vf, vertex, data); - if(memcmp((uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size, data, vbuf->vertex_size)) { - debug_printf("With VF:\n"); - dump_emitted_vertex(vbuf->vinfo, (uint8_t *)data); - debug_printf("Without VF:\n"); - dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr - vbuf->vertex_size); - assert(0); - } - } -#endif - } - else { - draw_vf_emit_vertex(vbuf->vf, vertex, vbuf->vertex_ptr); + draw_vf_emit_vertex(vbuf->vf, vertex, vbuf->vertex_ptr); - if (0) dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); + if (0) dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); - vbuf->vertex_ptr += vbuf->vertex_size/4; - } + vbuf->vertex_ptr += vbuf->vertex_size/4; } @@ -526,7 +452,7 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); if(!vbuf) - return NULL; + goto fail; vbuf->stage.draw = draw; vbuf->stage.point = vbuf_first_point; @@ -537,21 +463,27 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, vbuf->stage.destroy = vbuf_destroy; vbuf->render = render; + vbuf->max_indices = MAX2(render->max_indices, UNDEFINED_VERTEX_ID-1); - assert(render->max_indices < UNDEFINED_VERTEX_ID); - vbuf->max_indices = render->max_indices; - vbuf->indices = (ushort *) - align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); + vbuf->indices = (ushort *) align_malloc( vbuf->max_indices * + sizeof(vbuf->indices[0]), + 16 ); if(!vbuf->indices) - vbuf_destroy(&vbuf->stage); + goto fail; vbuf->vertices = NULL; vbuf->vertex_ptr = vbuf->vertices; vbuf->prim = ~0; - - if(!GETENV("GALLIUM_NOVF")) - vbuf->vf = draw_vf_create(); + vbuf->vf = draw_vf_create(); + if (!vbuf->vf) + goto fail; return &vbuf->stage; + + fail: + if (vbuf) + vbuf_destroy(&vbuf->stage); + + return NULL; } -- cgit v1.2.3 From 3be453bf7fcffd945ee0238e66ff4247a4491f1c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 16 Apr 2008 12:55:47 +0100 Subject: draw: allow pt paths to run without a vbuf render stage --- src/gallium/auxiliary/draw/draw_pt.c | 7 +- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 137 +++++++++++---------- 2 files changed, 72 insertions(+), 72 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 895224c952..c0125c906f 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -54,10 +54,9 @@ draw_pt_arrays(struct draw_context *draw, struct draw_pt_middle_end *middle = NULL; unsigned opt = 0; - if (!draw->render) - return FALSE; - /*debug_printf("XXXXXXXXXX needs_pipeline = %d\n", pipeline);*/ - + if (!draw->render) { + opt |= PT_PIPELINE; + } if (draw_need_pipeline(draw, prim)) { opt |= PT_PIPELINE; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 52be4a64aa..a66e8b0ba2 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -61,88 +61,89 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, fpme->prim = prim; fpme->opt = opt; - ok = draw->render->set_primitive(draw->render, prim); - if (!ok) { - assert(0); - return; - } + if (!(opt & PT_PIPELINE)) { + ok = draw->render->set_primitive(draw->render, prim); + if (!ok) { + assert(0); + return; + } - /* Must do this after set_primitive() above: - */ - vinfo = draw->render->get_vertex_info(draw->render); + /* Must do this after set_primitive() above: + */ + vinfo = draw->render->get_vertex_info(draw->render); - /* In passthrough mode, need to translate from vertex shader - * outputs to hw vertices. - */ - dst_offset = 0; - for (i = 0; i < vinfo->num_attribs; i++) { - unsigned emit_sz = 0; - unsigned src_buffer = 0; - unsigned output_format; - unsigned src_offset = (sizeof(struct vertex_header) + - vinfo->src_index[i] * 4 * sizeof(float) ); + /* In passthrough mode, need to translate from vertex shader + * outputs to hw vertices. + */ + dst_offset = 0; + for (i = 0; i < vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned src_buffer = 0; + unsigned output_format; + unsigned src_offset = (sizeof(struct vertex_header) + + vinfo->src_index[i] * 4 * sizeof(float) ); - switch (vinfo->emit[i]) { - case EMIT_4F: - output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - emit_sz = 4 * sizeof(float); - break; - case EMIT_3F: - output_format = PIPE_FORMAT_R32G32B32_FLOAT; - emit_sz = 3 * sizeof(float); - break; - case EMIT_2F: - output_format = PIPE_FORMAT_R32G32_FLOAT; - emit_sz = 2 * sizeof(float); - break; - case EMIT_1F: - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - break; - case EMIT_1F_PSIZE: - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - src_buffer = 1; - src_offset = 0; - break; - case EMIT_4UB: - output_format = PIPE_FORMAT_B8G8R8A8_UNORM; - emit_sz = 4 * sizeof(ubyte); - default: - assert(0); - output_format = PIPE_FORMAT_NONE; - emit_sz = 0; - break; - } + switch (vinfo->emit[i]) { + case EMIT_4F: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + src_buffer = 1; + src_offset = 0; + break; + case EMIT_4UB: + output_format = PIPE_FORMAT_B8G8R8A8_UNORM; + emit_sz = 4 * sizeof(ubyte); + default: + assert(0); + output_format = PIPE_FORMAT_NONE; + emit_sz = 0; + break; + } - hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - hw_key.element[i].input_buffer = src_buffer; - hw_key.element[i].input_offset = src_offset; - hw_key.element[i].output_format = output_format; - hw_key.element[i].output_offset = dst_offset; + hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + hw_key.element[i].input_buffer = src_buffer; + hw_key.element[i].input_offset = src_offset; + hw_key.element[i].output_format = output_format; + hw_key.element[i].output_offset = dst_offset; - dst_offset += emit_sz; - } + dst_offset += emit_sz; + } - hw_key.nr_elements = vinfo->num_attribs; - hw_key.output_stride = vinfo->size * 4; + hw_key.nr_elements = vinfo->num_attribs; + hw_key.output_stride = vinfo->size * 4; - /* Don't bother with caching at this stage: - */ - if (!fpme->translate || - memcmp(&fpme->translate->key, &hw_key, sizeof(hw_key)) != 0) - { - if (fpme->translate) - fpme->translate->release(fpme->translate); + /* Don't bother with caching at this stage: + */ + if (!fpme->translate || + memcmp(&fpme->translate->key, &hw_key, sizeof(hw_key)) != 0) + { + if (fpme->translate) + fpme->translate->release(fpme->translate); - fpme->translate = translate_generic_create( &hw_key ); + fpme->translate = translate_generic_create( &hw_key ); + } } - //fpme->pipeline_vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); fpme->pipeline_vertex_size = MAX_VERTEX_ALLOCATION; } -- cgit v1.2.3 From 3a765bbe00940c4dffce72b659f625d97ea17971 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Apr 2008 22:32:10 +0900 Subject: gallium: Use debug_get_bool_option for controlling assert failure behaviour. Add GALLIUM_ABORT_ON_ASSERT=no to C:\gallium.cfg instead. --- src/gallium/auxiliary/util/p_debug.c | 47 +++--------------------------------- 1 file changed, 4 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index c195f61820..25b132b40c 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -200,6 +200,8 @@ debug_get_bool_option(const char *name, boolean dfault) if(str == NULL) result = dfault; + else if(!strcmp(str, "n")) + result = FALSE; else if(!strcmp(str, "no")) result = FALSE; else if(!strcmp(str, "0")) @@ -251,57 +253,16 @@ debug_get_flags_option(const char *name, } -#if defined(WIN32) -ULONG_PTR debug_config_file = 0; -void *mapped_config_file = 0; - -enum { - eAssertAbortEn = 0x1, -}; - -/* Check for aborts enabled. */ -static unsigned abort_en(void) -{ - if (!mapped_config_file) - { - /* Open an 8 byte file for configuration data. */ - mapped_config_file = EngMapFile(L"\\??\\c:\\gaDebug.cfg", 8, &debug_config_file); - } - - /* A value of "0" (ascii) in the configuration file will clear the - * first 8 bits in the test byte. - * - * A value of "1" (ascii) in the configuration file will set the - * first bit in the test byte. - * - * A value of "2" (ascii) in the configuration file will set the - * second bit in the test byte. - * - * Currently the only interesting values are 0 and 1, which clear - * and set abort-on-assert behaviour respectively. - */ - return ((((char *)mapped_config_file)[0]) - 0x30) & eAssertAbortEn; -} -#else /* WIN32 */ -static unsigned abort_en(void) -{ - return !GETENV("GALLIUM_ABORT_ON_ASSERT"); -} -#endif - void _debug_assert_fail(const char *expr, const char *file, unsigned line, const char *function) { _debug_printf("%s:%u:%s: Assertion `%s' failed.\n", file, line, function, expr); - if (abort_en()) - { + if (debug_get_bool_option("GALLIUM_ABORT_ON_ASSERT", TRUE)) debug_break(); - } else - { + else _debug_printf("continuing...\n"); - } } -- cgit v1.2.3 From 302daeb2ec62c59bd28d0c53cdc0fb07dc9d0f58 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Apr 2008 09:04:39 -0600 Subject: mesa: added internal post color matrix scale/bias vars --- src/mesa/shader/prog_statevars.c | 12 ++++++++++++ src/mesa/shader/prog_statevars.h | 2 ++ 2 files changed, 14 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 6cc490bc77..45352d30b5 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -450,6 +450,12 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], value[2] = ctx->Pixel.BlueBias; value[3] = ctx->Pixel.AlphaBias; break; + case STATE_PCM_SCALE: + COPY_4V(value, ctx->Pixel.PostColorMatrixScale); + break; + case STATE_PCM_BIAS: + COPY_4V(value, ctx->Pixel.PostColorMatrixBias); + break; default: /* unknown state indexes are silently ignored * should be handled by the driver. @@ -697,6 +703,12 @@ append_token(char *dst, gl_state_index k) case STATE_PT_BIAS: append(dst, "PTbias"); break; + case STATE_PCM_SCALE: + append(dst, "PCMscale"); + break; + case STATE_PCM_BIAS: + append(dst, "PCMbias"); + break; default: ; } diff --git a/src/mesa/shader/prog_statevars.h b/src/mesa/shader/prog_statevars.h index 9fedcfab8a..d12142055f 100644 --- a/src/mesa/shader/prog_statevars.h +++ b/src/mesa/shader/prog_statevars.h @@ -111,6 +111,8 @@ typedef enum gl_state_index_ { STATE_SPOT_DIR_NORMALIZED, /* pre-normalized spot dir */ STATE_PT_SCALE, /**< Pixel transfer RGBA scale */ STATE_PT_BIAS, /**< Pixel transfer RGBA bias */ + STATE_PCM_SCALE, /**< Post color matrix RGBA scale */ + STATE_PCM_BIAS, /**< Post color matrix RGBA bias */ STATE_INTERNAL_DRIVER /* first available state index for drivers (must be last) */ } gl_state_index; -- cgit v1.2.3 From e38f677e8f5596d92a6756e13f41f6523de737c2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Apr 2008 09:09:08 -0600 Subject: gallium: finish-up and fix support for GL_COLOR matrix on pixel xfer path --- src/mesa/state_tracker/st_atom_pixeltransfer.c | 75 ++++++++++++++++++-------- src/mesa/state_tracker/st_extensions.c | 1 + 2 files changed, 54 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index 6410e7cb24..efb92b735e 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -50,6 +50,7 @@ struct state_key { GLuint scaleAndBias:1; GLuint colorMatrix:1; + GLuint colorMatrixPostScaleBias:1; #if 0 GLfloat Maps[3][256][4]; @@ -80,6 +81,9 @@ is_identity(const GLfloat m[16]) static void make_state_key(GLcontext *ctx, struct state_key *key) { + static const GLfloat zero[4] = { 0.0f, 0.0f, 0.0f, 0.0f }; + static const GLfloat one[4] = { 1.0f, 1.0f, 1.0f, 1.0f }; + memset(key, 0, sizeof(*key)); if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 || @@ -92,8 +96,12 @@ make_state_key(GLcontext *ctx, struct state_key *key) if (!is_identity(ctx->ColorMatrixStack.Top->m)) { key->colorMatrix = 1; } -} + if (!TEST_EQ_4V(ctx->Pixel.PostColorMatrixScale, one) || + !TEST_EQ_4V(ctx->Pixel.PostColorMatrixBias, zero)) { + key->colorMatrixPostScaleBias = 1; + } +} @@ -182,64 +190,87 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) GLint row3_p = _mesa_add_state_reference(params, row3_state); const GLuint temp = 1; - /* MOV temp, colorTemp; */ - _mesa_init_instructions(inst + ic, 1); - inst[ic].Opcode = OPCODE_MOV; - inst[ic].DstReg.File = PROGRAM_TEMPORARY; - inst[ic].DstReg.Index = temp; - inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - inst[ic].SrcReg[0].Index = colorTemp; - ic++; - /* XXX reimplement in terms of MUL/MAD (see t_vp_build.c) */ - /* DP4 colorTemp.x, temp, matrow0; */ + /* DP4 temp.x, colorTemp, matrow0; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; inst[ic].DstReg.File = PROGRAM_TEMPORARY; - inst[ic].DstReg.Index = colorTemp; + inst[ic].DstReg.Index = temp; inst[ic].DstReg.WriteMask = WRITEMASK_X; inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - inst[ic].SrcReg[0].Index = temp; + inst[ic].SrcReg[0].Index = colorTemp; inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = row0_p; ic++; - /* DP4 colorTemp.y, temp, matrow1; */ + /* DP4 temp.y, colorTemp, matrow1; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; inst[ic].DstReg.File = PROGRAM_TEMPORARY; - inst[ic].DstReg.Index = colorTemp; + inst[ic].DstReg.Index = temp; inst[ic].DstReg.WriteMask = WRITEMASK_Y; inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - inst[ic].SrcReg[0].Index = temp; + inst[ic].SrcReg[0].Index = colorTemp; inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = row1_p; ic++; - /* DP4 colorTemp.z, temp, matrow2; */ + /* DP4 temp.z, colorTemp, matrow2; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; inst[ic].DstReg.File = PROGRAM_TEMPORARY; - inst[ic].DstReg.Index = colorTemp; + inst[ic].DstReg.Index = temp; inst[ic].DstReg.WriteMask = WRITEMASK_Z; inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - inst[ic].SrcReg[0].Index = temp; + inst[ic].SrcReg[0].Index = colorTemp; inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = row2_p; ic++; - /* DP4 colorTemp.w, temp, matrow3; */ + /* DP4 temp.w, colorTemp, matrow3; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; inst[ic].DstReg.File = PROGRAM_TEMPORARY; - inst[ic].DstReg.Index =colorTemp; + inst[ic].DstReg.Index = temp; inst[ic].DstReg.WriteMask = WRITEMASK_W; inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; - inst[ic].SrcReg[0].Index = temp; + inst[ic].SrcReg[0].Index = colorTemp; inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; inst[ic].SrcReg[1].Index = row3_p; ic++; + + /* MOV colorTemp, temp; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_MOV; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = colorTemp; + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = temp; + ic++; + } + + if (key->colorMatrixPostScaleBias) { + static const gl_state_index scale_state[STATE_LENGTH] = + { STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 }; + static const gl_state_index bias_state[STATE_LENGTH] = + { STATE_INTERNAL, STATE_PT_BIAS, 0, 0, 0 }; + GLint scale_param, bias_param; + + scale_param = _mesa_add_state_reference(params, scale_state); + bias_param = _mesa_add_state_reference(params, bias_state); + + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_MAD; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = colorTemp; + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = colorTemp; + inst[ic].SrcReg[1].File = PROGRAM_STATE_VAR; + inst[ic].SrcReg[1].Index = scale_param; + inst[ic].SrcReg[2].File = PROGRAM_STATE_VAR; + inst[ic].SrcReg[2].Index = bias_param; + ic++; } /* Modify last instruction's dst reg to write to result.color */ diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 2f7ac074da..f2d40e84b3 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -164,6 +164,7 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.NV_blend_square = GL_TRUE; ctx->Extensions.NV_texgen_reflection = GL_TRUE; + ctx->Extensions.SGI_color_matrix = GL_TRUE; ctx->Extensions.SGIS_generate_mipmap = GL_TRUE; /* XXX temp */ /* -- cgit v1.2.3 From dc5a853c85d2daa8b7b0e6d32e00c4976746d704 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Apr 2008 09:45:56 -0600 Subject: gallium: make choose_format() non-static: st_choose_format() --- src/mesa/state_tracker/st_format.c | 9 +++++---- src/mesa/state_tracker/st_format.h | 4 ++++ 2 files changed, 9 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index cd41dc4684..3d15a03cab 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -354,8 +354,9 @@ default_depth_format(struct pipe_screen *screen, uint type) * Given an OpenGL internalFormat value for a texture or surface, return * the best matching PIPE_FORMAT_x, or PIPE_FORMAT_NONE if there's no match. */ -static enum pipe_format -choose_format(struct pipe_context *pipe, GLint internalFormat, uint surfType) +enum pipe_format +st_choose_format(struct pipe_context *pipe, GLint internalFormat, + uint surfType) { struct pipe_screen *screen = pipe->screen; @@ -530,7 +531,7 @@ choose_format(struct pipe_context *pipe, GLint internalFormat, uint surfType) enum pipe_format st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) { - return choose_format(pipe, internalFormat, PIPE_SURFACE); + return st_choose_format(pipe, internalFormat, PIPE_SURFACE); } @@ -594,7 +595,7 @@ st_ChooseTextureFormat(GLcontext *ctx, GLint internalFormat, (void) format; (void) type; - pFormat = choose_format(ctx->st->pipe, internalFormat, PIPE_TEXTURE); + pFormat = st_choose_format(ctx->st->pipe, internalFormat, PIPE_TEXTURE); if (pFormat == PIPE_FORMAT_NONE) return NULL; diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index c9a11de504..ff0fd042db 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -63,6 +63,10 @@ extern enum pipe_format st_mesa_format_to_pipe_format(GLuint mesaFormat); +extern enum pipe_format +st_choose_format(struct pipe_context *pipe, GLint internalFormat, + uint surfType); + extern enum pipe_format st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat); -- cgit v1.2.3 From 52544aa23bebe68d6fc9b13dc55f6bb8c03430ff Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Apr 2008 09:46:38 -0600 Subject: gallium: added util_pack_color_ub() --- src/gallium/auxiliary/util/u_pack_color.h | 39 +++++++++++++++++++++++++++++++ 1 file changed, 39 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index cd13823985..1f6604c554 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -39,6 +39,45 @@ #include "pipe/p_format.h" +/** + * Pack ubyte R,G,B,A into dest pixel. + */ +static INLINE void +util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, + enum pipe_format format, void *dest) +{ + switch (format) { + case PIPE_FORMAT_R8G8B8A8_UNORM: + { + uint *d = (uint *) dest; + *d = (r << 24) | (g << 16) | (b << 8) | a; + } + return; + case PIPE_FORMAT_A8R8G8B8_UNORM: + { + uint *d = (uint *) dest; + *d = (a << 24) | (r << 16) | (g << 8) | b; + } + return; + case PIPE_FORMAT_B8G8R8A8_UNORM: + { + uint *d = (uint *) dest; + *d = (b << 24) | (g << 16) | (r << 8) | a; + } + return; + case PIPE_FORMAT_R5G6B5_UNORM: + { + ushort *d = (ushort *) dest; + *d = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); + } + return; + /* XXX lots more cases to add */ + default: + debug_printf("gallium: unhandled format in util_pack_color_ub()"); + } +} + + /** * Note rgba outside [0,1] will be clamped for int pixel formats. */ -- cgit v1.2.3 From 4b822a101680532ce6df52904af91194b78a16ba Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Apr 2008 16:52:12 -0600 Subject: gallium: implement RGBA pixel maps in the pixel transfer fragment program --- src/mesa/state_tracker/st_atom_pixeltransfer.c | 136 ++++++++++++++++++++++++- src/mesa/state_tracker/st_cb_drawpixels.c | 13 ++- src/mesa/state_tracker/st_context.h | 2 + 3 files changed, 146 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index efb92b735e..76356bbad7 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -42,8 +42,14 @@ #include "shader/prog_print.h" #include "st_context.h" +#include "st_format.h" #include "st_program.h" +#include "st_texture.h" +#include "pipe/p_screen.h" +#include "pipe/p_context.h" +#include "pipe/p_inlines.h" +#include "util/u_pack_color.h" struct state_key @@ -51,6 +57,7 @@ struct state_key GLuint scaleAndBias:1; GLuint colorMatrix:1; GLuint colorMatrixPostScaleBias:1; + GLuint pixelMaps:1; #if 0 GLfloat Maps[3][256][4]; @@ -101,6 +108,69 @@ make_state_key(GLcontext *ctx, struct state_key *key) !TEST_EQ_4V(ctx->Pixel.PostColorMatrixBias, zero)) { key->colorMatrixPostScaleBias = 1; } + + key->pixelMaps = ctx->Pixel.MapColorFlag; +} + + +static struct pipe_texture * +create_color_map_texture(GLcontext *ctx) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct pipe_texture *pt; + enum pipe_format format; + const uint texSize = 256; /* simple, and usually perfect */ + + /* find an RGBA texture format */ + format = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE); + + /* create texture for color map/table */ + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, + texSize, texSize, 1, 0); + return pt; +} + + +/** + * Update the pixelmap texture with the contents of the R/G/B/A pixel maps. + */ +static void +load_color_map_texture(GLcontext *ctx, struct pipe_texture *pt) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *surface; + const GLuint rSize = ctx->PixelMaps.RtoR.Size; + const GLuint gSize = ctx->PixelMaps.GtoG.Size; + const GLuint bSize = ctx->PixelMaps.BtoB.Size; + const GLuint aSize = ctx->PixelMaps.AtoA.Size; + const uint texSize = pt->width[0]; + uint *dest; + uint i, j; + + surface = screen->get_tex_surface(screen, pt, 0, 0, 0); + dest = (uint *) pipe_surface_map(surface); + + /* Pack four 1D maps into a 2D texture: + * R map is placed horizontally, indexed by S, in channel 0 + * G map is placed vertically, indexed by T, in channel 1 + * B map is placed horizontally, indexed by S, in channel 2 + * A map is placed vertically, indexed by T, in channel 3 + */ + for (i = 0; i < texSize; i++) { + for (j = 0; j < texSize; j++) { + int k = (i * texSize + j); + ubyte r = ctx->PixelMaps.RtoR.Map8[j * rSize / texSize]; + ubyte g = ctx->PixelMaps.GtoG.Map8[i * gSize / texSize]; + ubyte b = ctx->PixelMaps.BtoB.Map8[j * bSize / texSize]; + ubyte a = ctx->PixelMaps.AtoA.Map8[i * aSize / texSize]; + util_pack_color_ub(r, g, b, a, pt->format, dest + k); + } + } + + pipe_surface_unmap(surface); + pipe_surface_reference(&surface, NULL); + pipe->texture_update(pipe, pt, 0, 0x1); } @@ -113,6 +183,7 @@ make_state_key(GLcontext *ctx, struct state_key *key) static struct gl_fragment_program * get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) { + struct st_context *st = ctx->st; struct prog_instruction inst[MAX_INST]; struct gl_program_parameter_list *params; struct gl_fragment_program *fp; @@ -126,7 +197,10 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) params = _mesa_new_parameter_list(); - /* TEX colorTemp, fragment.texcoord[0], texture[0], 2D; */ + /* + * Get initial pixel color from the texture. + * TEX colorTemp, fragment.texcoord[0], texture[0], 2D; + */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_TEX; inst[ic].DstReg.File = PROGRAM_TEMPORARY; @@ -140,7 +214,6 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) fp->Base.OutputsWritten = (1 << FRAG_RESULT_COLR); fp->Base.SamplersUsed = 0x1; /* sampler 0 (bit 0) is used */ - /* MAD colorTemp, colorTemp, scale, bias; */ if (key->scaleAndBias) { static const gl_state_index scale_state[STATE_LENGTH] = { STATE_INTERNAL, STATE_PT_SCALE, 0, 0, 0 }; @@ -161,6 +234,7 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) scale_p = _mesa_add_state_reference(params, scale_state); bias_p = _mesa_add_state_reference(params, bias_state); + /* MAD colorTemp, colorTemp, scale, bias; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_MAD; inst[ic].DstReg.File = PROGRAM_TEMPORARY; @@ -174,6 +248,56 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) ic++; } + if (key->pixelMaps) { + const GLuint temp = 1; + + /* create the colormap/texture now if not already done */ + if (!st->pixel_xfer.pixelmap_texture) { + st->pixel_xfer.pixelmap_texture = create_color_map_texture(ctx); + } + + /* with a little effort, we can do four pixel map look-ups with + * two TEX instructions: + */ + + /* TEX temp.rg, colorTemp.rgba, texture[1], 2D; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_TEX; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = temp; + inst[ic].DstReg.WriteMask = WRITEMASK_XY; /* write R,G */ + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = colorTemp; + inst[ic].TexSrcUnit = 1; + inst[ic].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; + + /* TEX temp.ba, colorTemp.baba, texture[1], 2D; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_TEX; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = temp; + inst[ic].DstReg.WriteMask = WRITEMASK_ZW; /* write B,A */ + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = colorTemp; + inst[ic].SrcReg[0].Swizzle = MAKE_SWIZZLE4(SWIZZLE_Z, SWIZZLE_W, + SWIZZLE_Z, SWIZZLE_W); + inst[ic].TexSrcUnit = 1; + inst[ic].TexSrcTarget = TEXTURE_2D_INDEX; + ic++; + + /* MOV colorTemp, temp; */ + _mesa_init_instructions(inst + ic, 1); + inst[ic].Opcode = OPCODE_MOV; + inst[ic].DstReg.File = PROGRAM_TEMPORARY; + inst[ic].DstReg.Index = colorTemp; + inst[ic].SrcReg[0].File = PROGRAM_TEMPORARY; + inst[ic].SrcReg[0].Index = temp; + ic++; + + fp->Base.SamplersUsed |= (1 << 1); /* sampler 1 is used */ + } + if (key->colorMatrix) { static const gl_state_index row0_state[STATE_LENGTH] = { STATE_COLOR_MATRIX, 0, 0, 0, 0 }; @@ -190,8 +314,6 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) GLint row3_p = _mesa_add_state_reference(params, row3_state); const GLuint temp = 1; - /* XXX reimplement in terms of MUL/MAD (see t_vp_build.c) */ - /* DP4 temp.x, colorTemp, matrow0; */ _mesa_init_instructions(inst + ic, 1); inst[ic].Opcode = OPCODE_DP4; @@ -316,6 +438,7 @@ get_pixel_transfer_program(GLcontext *ctx, const struct state_key *key) static void update_pixel_transfer(struct st_context *st) { + GLcontext *ctx = st->ctx; struct state_key key; struct gl_fragment_program *fp; @@ -329,6 +452,11 @@ update_pixel_transfer(struct st_context *st) &key, sizeof(key), &fp->Base); } + if (ctx->Pixel.MapColorFlag) { + load_color_map_texture(ctx, st->pixel_xfer.pixelmap_texture); + } + st->pixel_xfer.pixelmap_enabled = ctx->Pixel.MapColorFlag; + st->pixel_xfer.program = (struct st_fragment_program *) fp; } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 5f8c90cbba..67f468c689 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -562,6 +562,9 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, sampler.normalized_coords = 1; cso_single_sampler(cso, 0, &sampler); + if (st->pixel_xfer.pixelmap_enabled) { + cso_single_sampler(cso, 1, &sampler); + } cso_single_sampler_done(cso); } @@ -582,7 +585,15 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, } /* texture state: */ - pipe->set_sampler_textures(pipe, 1, &pt); + if (st->pixel_xfer.pixelmap_enabled) { + struct pipe_texture *textures[2]; + textures[0] = pt; + textures[1] = st->pixel_xfer.pixelmap_texture; + pipe->set_sampler_textures(pipe, 2, textures); + } + else { + pipe->set_sampler_textures(pipe, 1, &pt); + } /* Compute window coords (y=0=bottom) with pixel zoom. * Recall that these coords are transformed by the current diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index d89e54c43c..ae1ba41d4f 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -138,6 +138,8 @@ struct st_context GLuint user_prog_sn; /**< user fragment program serial no. */ struct st_fragment_program *combined_prog; GLuint combined_prog_sn; + struct pipe_texture *pixelmap_texture; + boolean pixelmap_enabled; /**< use the pixelmap texture? */ } pixel_xfer; /** for glBitmap */ -- cgit v1.2.3 From 8abfcea690c66f75d61905f0ec5497b8fd7950d4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 17 Apr 2008 10:37:53 +0100 Subject: draw: remove misleading comments --- src/gallium/auxiliary/draw/draw_pt_pipeline.c | 14 +------------- 1 file changed, 1 insertion(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_pipeline.c index e70e63d08f..17ce6febec 100644 --- a/src/gallium/auxiliary/draw/draw_pt_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_pipeline.c @@ -36,11 +36,6 @@ #include "draw/draw_vertex.h" #include "draw/draw_pt.h" - -/** - * Add a point to the primitive queue. - * \param i0 index into user's vertex arrays - */ static void do_point( struct draw_context *draw, const char *v0 ) { @@ -55,11 +50,6 @@ static void do_point( struct draw_context *draw, } -/** - * Add a line to the primitive queue. - * \param i0 index into user's vertex arrays - * \param i1 index into user's vertex arrays - */ static void do_line( struct draw_context *draw, const char *v0, const char *v1 ) @@ -75,9 +65,7 @@ static void do_line( struct draw_context *draw, draw->pipeline.first->line( draw->pipeline.first, &prim ); } -/** - * Add a triangle to the primitive queue. - */ + static void do_triangle( struct draw_context *draw, char *v0, char *v1, -- cgit v1.2.3 From c96d565643de271c6bda066e892b25d0a97ea4d0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 17 Apr 2008 13:14:22 +0100 Subject: draw: keep record of number of active vertex buffers --- src/gallium/auxiliary/draw/draw_context.c | 1 + src/gallium/auxiliary/draw/draw_private.h | 3 +++ 2 files changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 0c314f6e1d..cdcbfdc323 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -249,6 +249,7 @@ draw_set_vertex_buffers(struct draw_context *draw, draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); memcpy(draw->vertex_buffer, buffers, count * sizeof(buffers[0])); + draw->nr_vertex_buffers = count; } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 8ac061cc9f..359a8b8a3d 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -245,7 +245,10 @@ struct draw_context /* pipe state that we need: */ const struct pipe_rasterizer_state *rasterizer; struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + unsigned nr_vertex_buffers; + struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; struct draw_vertex_shader *vertex_shader; -- cgit v1.2.3 From 49becd2d7c751e563ce6be9051dd8e6dad88d1f7 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 17 Apr 2008 13:14:55 +0100 Subject: draw: add comment --- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 1a9d331a06..d696afa1aa 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -185,6 +185,11 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, */ vinfo = draw->render->get_vertex_info(draw->render); + + /* This is unique as it transforms straight from API vertex + * information to HW vertices. All other cases go through our + * intermediate float[4] format. + */ for (i = 0; i < vinfo->num_attribs; i++) { unsigned src_element = vinfo->src_index[i]; unsigned src_buffer = draw->vertex_element[src_element].vertex_buffer_index; -- cgit v1.2.3 From 938ec19210c5b4e19dcb2b606c9ade415f2c1f84 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 17 Apr 2008 14:19:03 +0100 Subject: tsgi: make Consts const --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h index 45c49dd007..92e2e5e985 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h @@ -166,7 +166,7 @@ struct tgsi_exec_machine float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; unsigned ImmLimit; - float (*Consts)[4]; + const float (*Consts)[4]; struct tgsi_exec_vector *Inputs; struct tgsi_exec_vector *Outputs; const struct tgsi_token *Tokens; -- cgit v1.2.3 From 280bcff1fa200b790d8712946a4ffbaa47a67433 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 17 Apr 2008 14:20:00 +0100 Subject: draw: add vertex shader run_linear function --- src/gallium/auxiliary/draw/draw_aapoint.c | 2 +- src/gallium/auxiliary/draw/draw_clip.c | 2 +- src/gallium/auxiliary/draw/draw_context.c | 17 +++- src/gallium/auxiliary/draw/draw_flatshade.c | 2 +- src/gallium/auxiliary/draw/draw_private.h | 28 +------ .../auxiliary/draw/draw_pt_fetch_pipeline.c | 2 +- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 2 +- src/gallium/auxiliary/draw/draw_twoside.c | 2 +- src/gallium/auxiliary/draw/draw_vertex_fetch.c | 2 +- src/gallium/auxiliary/draw/draw_vertex_shader.c | 7 +- src/gallium/auxiliary/draw/draw_vs.h | 40 ++++++++- src/gallium/auxiliary/draw/draw_vs_exec.c | 95 +++++++++++++++++++--- src/gallium/auxiliary/draw/draw_vs_llvm.c | 66 ++++++++++++++- src/gallium/auxiliary/draw/draw_vs_sse.c | 76 ++++++++++++++++- src/gallium/auxiliary/draw/draw_wide_point.c | 2 +- 15 files changed, 286 insertions(+), 59 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c index fcebe3e7a0..e84d380e50 100644 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_aapoint.c @@ -48,7 +48,7 @@ #include "tgsi/util/tgsi_dump.h" #include "draw_context.h" -#include "draw_private.h" +#include "draw_vs.h" /* diff --git a/src/gallium/auxiliary/draw/draw_clip.c b/src/gallium/auxiliary/draw/draw_clip.c index e24c5d8032..0ac3a240e5 100644 --- a/src/gallium/auxiliary/draw/draw_clip.c +++ b/src/gallium/auxiliary/draw/draw_clip.c @@ -36,7 +36,7 @@ #include "pipe/p_shader_tokens.h" #include "draw_context.h" -#include "draw_private.h" +#include "draw_vs.h" #ifndef IS_NEGATIVE diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index cdcbfdc323..9b2dcc0b57 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -33,8 +33,8 @@ #include "pipe/p_util.h" #include "draw_context.h" -#include "draw_private.h" #include "draw_vbuf.h" +#include "draw_vs.h" struct draw_context *draw_create( void ) @@ -108,6 +108,8 @@ struct draw_context *draw_create( void ) draw_vertex_cache_invalidate( draw ); draw_set_mapped_element_buffer( draw, 0, NULL ); + tgsi_exec_machine_init(&draw->machine); + if (!draw_pt_init( draw )) goto fail; @@ -460,3 +462,16 @@ boolean draw_get_edgeflag( struct draw_context *draw, else return 1; } + + +#if 0 +/* Crufty init function. Fix me. + */ +boolean draw_init_machine( struct draw_context *draw ) +{ + ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); + ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); + machine->Inputs = ALIGN16_ASSIGN(inputs); + machine->Outputs = ALIGN16_ASSIGN(outputs); +} +#endif diff --git a/src/gallium/auxiliary/draw/draw_flatshade.c b/src/gallium/auxiliary/draw/draw_flatshade.c index af2cb05c98..54baa1fbc9 100644 --- a/src/gallium/auxiliary/draw/draw_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_flatshade.c @@ -30,7 +30,7 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "draw_private.h" +#include "draw_vs.h" /** subclass of draw_stage */ diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 359a8b8a3d..7953dbb7d9 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -55,6 +55,7 @@ struct gallivm_cpu_engine; struct draw_pt_middle_end; struct draw_pt_front_end; +struct draw_vertex_shader; #define MAX_SHADER_VERTICES 128 @@ -132,33 +133,6 @@ struct draw_stage #define VCACHE_OVERFLOW 4 #define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ -/** - * Private version of the compiled vertex_shader - */ -struct draw_vertex_shader { - - /* This member will disappear shortly: - */ - struct pipe_shader_state state; - - struct tgsi_shader_info info; - - void (*prepare)( struct draw_vertex_shader *shader, - struct draw_context *draw ); - - /* Run the shader - this interface will get cleaned up in the - * future: - */ - boolean (*run)( struct draw_vertex_shader *shader, - struct draw_context *draw, - const unsigned *elts, - unsigned count, - void *out, - unsigned vertex_size); - - - void (*delete)( struct draw_vertex_shader * ); -}; /* Internal function for vertex fetch. diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c index db6d6c76ea..79548d4156 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c @@ -32,9 +32,9 @@ #include "pipe/p_util.h" #include "draw/draw_context.h" -#include "draw/draw_private.h" #include "draw/draw_vertex.h" #include "draw/draw_pt.h" +#include "draw/draw_vs.h" /* The simplest 'middle end' in the new vertex code. * diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index a66e8b0ba2..cf37493eb6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -27,10 +27,10 @@ #include "pipe/p_util.h" #include "draw/draw_context.h" -#include "draw/draw_private.h" #include "draw/draw_vbuf.h" #include "draw/draw_vertex.h" #include "draw/draw_pt.h" +#include "draw/draw_vs.h" #include "translate/translate.h" diff --git a/src/gallium/auxiliary/draw/draw_twoside.c b/src/gallium/auxiliary/draw/draw_twoside.c index 3debaac282..01d905c153 100644 --- a/src/gallium/auxiliary/draw/draw_twoside.c +++ b/src/gallium/auxiliary/draw/draw_twoside.c @@ -31,7 +31,7 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" -#include "draw_private.h" +#include "draw_vs.h" struct twoside_stage { diff --git a/src/gallium/auxiliary/draw/draw_vertex_fetch.c b/src/gallium/auxiliary/draw/draw_vertex_fetch.c index 9041041006..0bc2fcb424 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_fetch.c +++ b/src/gallium/auxiliary/draw/draw_vertex_fetch.c @@ -32,7 +32,7 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "draw_private.h" +#include "draw_vs.h" #include "draw_context.h" diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index 8572a6d40c..c0c17c116e 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -95,10 +95,8 @@ draw_create_vertex_shader(struct draw_context *draw, vs = draw_create_vs_exec( draw, shader ); } } - assert(vs); - - tgsi_scan_shader(shader->tokens, &vs->info); + assert(vs); return vs; } @@ -113,9 +111,6 @@ draw_bind_vertex_shader(struct draw_context *draw, { draw->vertex_shader = dvs; draw->num_vs_outputs = dvs->info.num_outputs; - - tgsi_exec_machine_init(&draw->machine); - dvs->prepare( dvs, draw ); } else { diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 33ce1e335e..e88d00e247 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -35,10 +35,46 @@ #include "draw_private.h" -struct draw_vertex_shader; struct draw_context; struct pipe_shader_state; +/** + * Private version of the compiled vertex_shader + */ +struct draw_vertex_shader { + + /* This member will disappear shortly: + */ + struct pipe_shader_state state; + + struct tgsi_shader_info info; + + void (*prepare)( struct draw_vertex_shader *shader, + struct draw_context *draw ); + + /* Run the shader - this interface will get cleaned up in the + * future: + */ + boolean (*run)( struct draw_vertex_shader *shader, + struct draw_context *draw, + const unsigned *elts, + unsigned count, + void *out, + unsigned vertex_size); + + void (*run_linear)( struct draw_vertex_shader *shader, + const float (*input)[4], + float (*output)[4], + const float (*constants)[4], + unsigned count, + unsigned input_stride, + unsigned output_stride ); + + + void (*delete)( struct draw_vertex_shader * ); +}; + + struct draw_vertex_shader * draw_create_vs_exec(struct draw_context *draw, const struct pipe_shader_state *templ); @@ -79,5 +115,7 @@ compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) return mask; } +#define MAX_TGSI_VERTICES 4 + #endif diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 5c88c2e24e..0e05b79715 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -40,8 +40,20 @@ #include "tgsi/util/tgsi_parse.h" -#define MAX_TGSI_VERTICES 4 +struct exec_vertex_shader { + struct draw_vertex_shader base; + struct tgsi_exec_machine *machine; +}; + +static struct exec_vertex_shader *exec_vertex_shader( struct draw_vertex_shader *vs ) +{ + return (struct exec_vertex_shader *)vs; +} + + +/* Not required for run_linear. + */ static void vs_exec_prepare( struct draw_vertex_shader *shader, struct draw_context *draw ) @@ -81,10 +93,9 @@ vs_exec_run( struct draw_vertex_shader *shader, const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; - assert(draw->vertex_shader->info.output_semantic_name[0] - == TGSI_SEMANTIC_POSITION); + assert(shader->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); - machine->Consts = (float (*)[4]) draw->user.constants; + machine->Consts = (const float (*)[4]) draw->user.constants; machine->Inputs = ALIGN16_ASSIGN(inputs); if (draw->rasterizer->bypass_vs) { /* outputs are just the inputs */ @@ -179,6 +190,64 @@ vs_exec_run( struct draw_vertex_shader *shader, +/* Simplified vertex shader interface for the pt paths. Given the + * complexity of code-generating all the above operations together, + * it's time to try doing all the other stuff separately. + */ +static void +vs_exec_run_linear( struct draw_vertex_shader *shader, + const float (*input)[4], + float (*output)[4], + const float (*constants)[4], + unsigned count, + unsigned input_stride, + unsigned output_stride ) +{ + struct exec_vertex_shader *evs = exec_vertex_shader(shader); + struct tgsi_exec_machine *machine = evs->machine; + unsigned int i, j; + unsigned slot; + + machine->Consts = constants; + + for (i = 0; i < count; i += MAX_TGSI_VERTICES) { + unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); + + /* Swizzle inputs. + */ + for (j = 0; j < max_vertices; j++) { + for (slot = 0; slot < shader->info.num_inputs; slot++) { + machine->Inputs[slot].xyzw[0].f[j] = input[slot][0]; + machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; + machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; + machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; + } + } + + /* run interpreter */ + tgsi_exec_machine_run( machine ); + + /* Unswizzle all output results. + */ + for (j = 0; j < max_vertices; j++) { + for (slot = 0; slot < shader->info.num_outputs; slot++) { + output[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + } + + /* Advance input, output pointers: + */ + input = (const float (*)[4])((const char *)input + input_stride); + output = (float (*)[4])((char *)output + output_stride); + } +} + + + + static void vs_exec_delete( struct draw_vertex_shader *dvs ) { @@ -191,17 +260,23 @@ struct draw_vertex_shader * draw_create_vs_exec(struct draw_context *draw, const struct pipe_shader_state *state) { - struct draw_vertex_shader *vs = CALLOC_STRUCT( draw_vertex_shader ); + struct exec_vertex_shader *vs = CALLOC_STRUCT( exec_vertex_shader ); uint nt = tgsi_num_tokens(state->tokens); if (vs == NULL) return NULL; /* we make a private copy of the tokens */ - vs->state.tokens = mem_dup(state->tokens, nt * sizeof(state->tokens[0])); - vs->prepare = vs_exec_prepare; - vs->run = vs_exec_run; - vs->delete = vs_exec_delete; + vs->base.state.tokens = mem_dup(state->tokens, nt * sizeof(state->tokens[0])); + tgsi_scan_shader(state->tokens, &vs->base.info); + + + vs->base.prepare = vs_exec_prepare; + vs->base.run = vs_exec_run; + vs->base.run_linear = vs_exec_run_linear; + vs->base.delete = vs_exec_delete; + vs->machine = &draw->machine; + - return vs; + return &vs->base; } diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 73076d2467..d0baca715e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -88,8 +88,7 @@ vs_llvm_run( struct draw_vertex_shader *base, assert(count <= 4); - assert(draw->vertex_shader->state->output_semantic_name[0] - == TGSI_SEMANTIC_POSITION); + assert(base->state->output_semantic_name[0] == TGSI_SEMANTIC_POSITION); /* Consts does not require 16 byte alignment. */ machine->Consts = (float (*)[4]) draw->user.constants; @@ -169,6 +168,65 @@ vs_llvm_run( struct draw_vertex_shader *base, return clipped != 0; } + + + +static void +vs_llvm_run_linear( struct draw_vertex_shader *base, + const float (*input)[4], + float (*output)[4], + const float (*constants)[4], + unsigned count, + unsigned input_stride, + unsigned output_stride ) +{ + struct draw_llvm_vertex_shader *shader = + (struct draw_llvm_vertex_shader *)base; + + struct tgsi_exec_machine *machine = &draw->machine; + unsigned int j; + + + for (i = 0; i < count; i += MAX_TGSI_VERTICES) { + unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); + + /* Swizzle inputs. + */ + for (j = 0; j < max_vertices; j++) { + for (slot = 0; slot < draw->num_vs_inputs; slot++) { + machine->Inputs[slot].xyzw[0].f[j] = input[slot][0]; + machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; + machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; + machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; + } + } + + /* run shader */ + gallivm_cpu_vs_exec(shader->llvm_prog, + machine->Inputs, + machine->Outputs, + (float (*)[4]) constants, + machine->Temps); + + + /* Unswizzle all output results + */ + for (slot = 1; slot < draw->num_vs_outputs; slot++) { + output[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + + /* Advance input, output pointers: + */ + input = (const float (*)[4])((const char *)input + input_stride); + output = (float (*)[4])((char *)output + output_stride); + } +} + + + static void vs_llvm_delete( struct draw_vertex_shader *base ) { @@ -198,8 +256,12 @@ draw_create_vs_llvm(struct draw_context *draw, /* we make a private copy of the tokens */ vs->base.state.tokens = mem_dup(templ->tokens, nt * sizeof(templ->tokens[0])); + + tgsi_scan_shader(shader->tokens, &vs->base.info); + vs->base.prepare = vs_llvm_prepare; vs->base.run = vs_llvm_run; + vs->base.run_linear = vs_llvm_run_linear; vs->base.delete = vs_llvm_delete; { diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index ee0a3105b9..873ecfdc5d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -58,7 +58,11 @@ typedef void (XSTDCALL *codegen_function) ( struct draw_sse_vertex_shader { struct draw_vertex_shader base; struct x86_function sse2_program; + codegen_function func; + + struct tgsi_exec_machine *machine; + float immediates[TGSI_EXEC_NUM_IMMEDIATES][4]; }; @@ -96,11 +100,10 @@ vs_sse_run( struct draw_vertex_shader *base, const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; - assert(draw->vertex_shader->info.output_semantic_name[0] - == TGSI_SEMANTIC_POSITION); + assert(base->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); /* Consts does not require 16 byte alignment. */ - machine->Consts = (float (*)[4]) draw->user.constants; + machine->Consts = (const float (*)[4]) draw->user.constants; machine->Inputs = ALIGN16_ASSIGN(inputs); if (draw->rasterizer->bypass_vs) { /* outputs are just the inputs */ @@ -124,7 +127,7 @@ vs_sse_run( struct draw_vertex_shader *base, */ shader->func(machine->Inputs, machine->Outputs, - machine->Consts, + (float (*)[4])machine->Consts, machine->Temps, shader->immediates); } @@ -200,6 +203,67 @@ vs_sse_run( struct draw_vertex_shader *base, } +/* Simplified vertex shader interface for the pt paths. Given the + * complexity of code-generating all the above operations together, + * it's time to try doing all the other stuff separately. + */ +static void +vs_sse_run_linear( struct draw_vertex_shader *base, + const float (*input)[4], + float (*output)[4], + const float (*constants)[4], + unsigned count, + unsigned input_stride, + unsigned output_stride ) +{ + struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; + struct tgsi_exec_machine *machine = shader->machine; + unsigned int i, j; + unsigned slot; + + for (i = 0; i < count; i += MAX_TGSI_VERTICES) { + unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); + + /* Swizzle inputs. + */ + for (j = 0; j < max_vertices; j++) { + for (slot = 0; slot < base->info.num_inputs; slot++) { + machine->Inputs[slot].xyzw[0].f[j] = input[slot][0]; + machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; + machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; + machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; + } + } + + /* run compiled shader + */ + shader->func(machine->Inputs, + machine->Outputs, + (float (*)[4])constants, + machine->Temps, + shader->immediates); + + + /* Unswizzle all output results. + */ + for (j = 0; j < max_vertices; j++) { + for (slot = 0; slot < base->info.num_outputs; slot++) { + output[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + } + + /* Advance input, output pointers: + */ + input = (const float (*)[4])((const char *)input + input_stride); + output = (float (*)[4])((char *)output + output_stride); + } +} + + + static void vs_sse_delete( struct draw_vertex_shader *base ) @@ -229,8 +293,12 @@ draw_create_vs_sse(struct draw_context *draw, /* we make a private copy of the tokens */ vs->base.state.tokens = mem_dup(templ->tokens, nt * sizeof(templ->tokens[0])); + + tgsi_scan_shader(templ->tokens, &vs->base.info); + vs->base.prepare = vs_sse_prepare; vs->base.run = vs_sse_run; + vs->base.run_linear = vs_sse_run_linear; vs->base.delete = vs_sse_delete; x86_init_func( &vs->sse2_program ); diff --git a/src/gallium/auxiliary/draw/draw_wide_point.c b/src/gallium/auxiliary/draw/draw_wide_point.c index 6fc7c9fcd7..0fa7c9313c 100644 --- a/src/gallium/auxiliary/draw/draw_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_wide_point.c @@ -31,7 +31,7 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" -#include "draw_private.h" +#include "draw_vs.h" struct widepoint_stage { -- cgit v1.2.3 From fe8af141246bdfc0e98beec0089acdb92407bcfa Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 17 Apr 2008 14:42:05 +0100 Subject: draw: don't always run pipeline if clipping --- src/gallium/auxiliary/draw/draw_pt.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index c0125c906f..60a47f3911 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -74,11 +74,11 @@ draw_pt_arrays(struct draw_context *draw, } - if (draw->pt.middle.opt[opt] == NULL) { - opt = PT_PIPELINE | PT_CLIPTEST | PT_SHADE; + middle = draw->pt.middle.opt[opt]; + if (middle == NULL) { + middle = draw->pt.middle.opt[PT_PIPELINE | PT_CLIPTEST | PT_SHADE]; } - middle = draw->pt.middle.opt[opt]; assert(middle); /* May create a short-circuited version of this for small primitives: -- cgit v1.2.3 From c503e55d74cf84f87f82b3dab3cb4d38b201d47a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 17 Apr 2008 14:43:40 +0100 Subject: draw: move hw vertex emit to a new module --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/draw_private.h | 8 - src/gallium/auxiliary/draw/draw_pt.h | 33 ++++ src/gallium/auxiliary/draw/draw_pt_emit.c | 208 +++++++++++++++++++++ .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 161 +++------------- 5 files changed, 270 insertions(+), 141 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_emit.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 5ab3cfe5ce..836e98f086 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -21,6 +21,7 @@ C_SOURCES = \ draw_pt_fetch_pipeline.c \ draw_pt_fetch_shade_pipeline.c \ draw_pt_pipeline.c \ + draw_pt_emit.c \ draw_pt_elts.c \ draw_prim.c \ draw_pstipple.c \ diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 7953dbb7d9..9407217fd3 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -376,14 +376,6 @@ boolean draw_pt_arrays( struct draw_context *draw, unsigned count ); void draw_pt_reset_vertex_ids( struct draw_context *draw ); -void draw_pt_run_pipeline( struct draw_context *draw, - unsigned prim, - char *verts, - unsigned vertex_stride, - unsigned vertex_count, - const ushort *elts, - unsigned count ); - #define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ #define DRAW_FLUSH_PRIM_QUEUE 0x2 diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 08afb60645..31d18ec62b 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -128,4 +128,37 @@ struct draw_pt_middle_end *draw_pt_fetch_pipeline( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw); +/* More helpers: + */ +void draw_pt_run_pipeline( struct draw_context *draw, + unsigned prim, + char *verts, + unsigned vertex_stride, + unsigned vertex_count, + const ushort *elts, + unsigned count ); + + +/* HW vertex emit: + */ +struct pt_emit; + +void draw_pt_emit_prepare( struct pt_emit *emit, + unsigned prim, + unsigned opt ); + +void draw_pt_emit( struct pt_emit *emit, + char *verts, + unsigned stride, + unsigned vertex_count, + const ushort *elts, + unsigned count ); + +void draw_pt_emit_destroy( struct pt_emit *emit ); + +struct pt_emit *draw_pt_emit_create( struct draw_context *draw ); + + + + #endif diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c new file mode 100644 index 0000000000..e9ed29450a --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -0,0 +1,208 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" +#include "translate/translate.h" + + +struct pt_emit { + struct draw_context *draw; + + struct translate *translate; + + unsigned pipeline_vertex_size; + unsigned prim; + unsigned opt; +}; + + +void draw_pt_emit_prepare( struct pt_emit *emit, + unsigned prim, + unsigned opt ) +{ + struct draw_context *draw = emit->draw; + const struct vertex_info *vinfo; + unsigned dst_offset; + struct translate_key hw_key; + unsigned i; + boolean ok; + + ok = draw->render->set_primitive(draw->render, prim); + if (!ok) { + assert(0); + return; + } + + /* Must do this after set_primitive() above: + */ + vinfo = draw->render->get_vertex_info(draw->render); + + + /* In passthrough mode, need to translate from vertex shader + * outputs to hw vertices. + */ + dst_offset = 0; + for (i = 0; i < vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned src_buffer = 0; + unsigned output_format; + unsigned src_offset = (sizeof(struct vertex_header) + + vinfo->src_index[i] * 4 * sizeof(float) ); + + + + switch (vinfo->emit[i]) { + case EMIT_4F: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + src_buffer = 1; + src_offset = 0; + break; + case EMIT_4UB: + output_format = PIPE_FORMAT_B8G8R8A8_UNORM; + emit_sz = 4 * sizeof(ubyte); + default: + assert(0); + output_format = PIPE_FORMAT_NONE; + emit_sz = 0; + break; + } + + hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + hw_key.element[i].input_buffer = src_buffer; + hw_key.element[i].input_offset = src_offset; + hw_key.element[i].output_format = output_format; + hw_key.element[i].output_offset = dst_offset; + + dst_offset += emit_sz; + } + + hw_key.nr_elements = vinfo->num_attribs; + hw_key.output_stride = vinfo->size * 4; + + /* Don't bother with caching at this stage: + */ + if (!emit->translate || + memcmp(&emit->translate->key, &hw_key, sizeof(hw_key)) != 0) + { + if (emit->translate) + emit->translate->release(emit->translate); + + emit->translate = translate_generic_create( &hw_key ); + } +} + + +void draw_pt_emit( struct pt_emit *emit, + char *verts, + unsigned stride, + unsigned vertex_count, + const ushort *elts, + unsigned count ) +{ + struct draw_context *draw = emit->draw; + struct translate *translate = emit->translate; + struct vbuf_render *render = draw->render; + void *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = render->allocate_vertices(render, + (ushort)translate->key.output_stride, + (ushort)count); + if (!hw_verts) { + assert(0); + return; + } + + translate->set_buffer(translate, + 0, + verts, + stride ); + + translate->set_buffer(translate, + 1, + &draw->rasterizer->point_size, + 0); + + translate->run( translate, + 0, + vertex_count, + hw_verts ); + + render->draw(render, + elts, + count); + + render->release_vertices(render, + hw_verts, + translate->key.output_stride, + vertex_count); +} + + +struct pt_emit *draw_pt_emit_create( struct draw_context *draw ) +{ + struct pt_emit *emit = CALLOC_STRUCT(pt_emit); + if (!emit) + return NULL; + + emit->draw = draw; + + return emit; +} + +void draw_pt_emit_destroy( struct pt_emit *emit ) +{ + if (emit->translate) + emit->translate->release( emit->translate ); + + FREE(emit); +} diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index cf37493eb6..8db9e31e2d 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -38,7 +38,7 @@ struct fetch_pipeline_middle_end { struct draw_pt_middle_end base; struct draw_context *draw; - struct translate *translate; + struct pt_emit *emit; unsigned pipeline_vertex_size; unsigned prim; @@ -51,98 +51,12 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, unsigned opt ) { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; - struct draw_context *draw = fpme->draw; - unsigned i; - boolean ok; - const struct vertex_info *vinfo; - unsigned dst_offset; - struct translate_key hw_key; fpme->prim = prim; fpme->opt = opt; - if (!(opt & PT_PIPELINE)) { - ok = draw->render->set_primitive(draw->render, prim); - if (!ok) { - assert(0); - return; - } - - /* Must do this after set_primitive() above: - */ - vinfo = draw->render->get_vertex_info(draw->render); - - - /* In passthrough mode, need to translate from vertex shader - * outputs to hw vertices. - */ - dst_offset = 0; - for (i = 0; i < vinfo->num_attribs; i++) { - unsigned emit_sz = 0; - unsigned src_buffer = 0; - unsigned output_format; - unsigned src_offset = (sizeof(struct vertex_header) + - vinfo->src_index[i] * 4 * sizeof(float) ); - - - - switch (vinfo->emit[i]) { - case EMIT_4F: - output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - emit_sz = 4 * sizeof(float); - break; - case EMIT_3F: - output_format = PIPE_FORMAT_R32G32B32_FLOAT; - emit_sz = 3 * sizeof(float); - break; - case EMIT_2F: - output_format = PIPE_FORMAT_R32G32_FLOAT; - emit_sz = 2 * sizeof(float); - break; - case EMIT_1F: - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - break; - case EMIT_1F_PSIZE: - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - src_buffer = 1; - src_offset = 0; - break; - case EMIT_4UB: - output_format = PIPE_FORMAT_B8G8R8A8_UNORM; - emit_sz = 4 * sizeof(ubyte); - default: - assert(0); - output_format = PIPE_FORMAT_NONE; - emit_sz = 0; - break; - } - - hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - hw_key.element[i].input_buffer = src_buffer; - hw_key.element[i].input_offset = src_offset; - hw_key.element[i].output_format = output_format; - hw_key.element[i].output_offset = dst_offset; - - dst_offset += emit_sz; - } - - hw_key.nr_elements = vinfo->num_attribs; - hw_key.output_stride = vinfo->size * 4; - - /* Don't bother with caching at this stage: - */ - if (!fpme->translate || - memcmp(&fpme->translate->key, &hw_key, sizeof(hw_key)) != 0) - { - if (fpme->translate) - fpme->translate->release(fpme->translate); - - fpme->translate = translate_generic_create( &hw_key ); - } - } - + if (!(opt & PT_PIPELINE)) + draw_pt_emit_prepare( fpme->emit, prim, opt ); //fpme->pipeline_vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); fpme->pipeline_vertex_size = MAX_VERTEX_ALLOCATION; @@ -194,43 +108,12 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, draw_elts, draw_count ); } else { - struct translate *translate = fpme->translate; - void *hw_verts; - - /* XXX: need to flush to get prim_vbuf.c to release its allocation?? - */ - draw_do_flush( draw, DRAW_FLUSH_BACKEND ); - - hw_verts = draw->render->allocate_vertices(draw->render, - (ushort)fpme->translate->key.output_stride, - (ushort)fetch_count); - if (!hw_verts) { - assert(0); - return; - } - - translate->set_buffer(translate, - 0, - pipeline_verts, - fpme->pipeline_vertex_size ); - - translate->set_buffer(translate, - 1, - &fpme->draw->rasterizer->point_size, - 0); - - translate->run( translate, - 0, fetch_count, - hw_verts ); - - draw->render->draw(draw->render, - draw_elts, - draw_count); - - draw->render->release_vertices(draw->render, - hw_verts, - fpme->translate->key.output_stride, - fetch_count); + draw_pt_emit( fpme->emit, + pipeline_verts, + fpme->pipeline_vertex_size, + fetch_count, + draw_elts, + draw_count ); } @@ -252,14 +135,26 @@ static void fetch_pipeline_destroy( struct draw_pt_middle_end *middle ) struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit( struct draw_context *draw ) { - struct fetch_pipeline_middle_end *fetch_pipeline = CALLOC_STRUCT( fetch_pipeline_middle_end ); + struct fetch_pipeline_middle_end *fpme = CALLOC_STRUCT( fetch_pipeline_middle_end ); + if (!fpme) + goto fail; + + fpme->base.prepare = fetch_pipeline_prepare; + fpme->base.run = fetch_pipeline_run; + fpme->base.finish = fetch_pipeline_finish; + fpme->base.destroy = fetch_pipeline_destroy; + + fpme->draw = draw; + + fpme->emit = draw_pt_emit_create( draw ); + if (!fpme->emit) + goto fail; - fetch_pipeline->base.prepare = fetch_pipeline_prepare; - fetch_pipeline->base.run = fetch_pipeline_run; - fetch_pipeline->base.finish = fetch_pipeline_finish; - fetch_pipeline->base.destroy = fetch_pipeline_destroy; + return &fpme->base; - fetch_pipeline->draw = draw; + fail: + if (fpme) + fetch_pipeline_destroy( &fpme->base ); - return &fetch_pipeline->base; + return NULL; } -- cgit v1.2.3 From 4dcb09d48981f07a82d7f609fe492453c8fdbcf4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 17 Apr 2008 17:53:33 +0100 Subject: gallium: Add new file to scons build. --- src/gallium/auxiliary/draw/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index a7fb5dbd61..62e0cd9e50 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -16,6 +16,7 @@ draw = env.ConvenienceLibrary( 'draw_offset.c', 'draw_pt.c', 'draw_pt_vcache.c', + 'draw_pt_emit.c', 'draw_pt_fetch_emit.c', 'draw_pt_fetch_pipeline.c', 'draw_pt_fetch_shade_pipeline.c', -- cgit v1.2.3 From f2b3f6cbdfc4799a0f742c06d2ce0755582c50c4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 17 Apr 2008 11:12:44 -0600 Subject: gallium: reorder some of the destroy context code --- src/mesa/state_tracker/st_context.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 154327239d..dadc524b51 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -170,10 +170,6 @@ static void st_destroy_context_priv( struct st_context *st ) _vbo_DestroyContext(st->ctx); - cso_destroy_context(st->cso_context); - - _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache); - for (i = 0; i < Elements(st->state.constants); i++) { if (st->state.constants[i].buffer) { pipe_buffer_reference(ws, &st->state.constants[i].buffer, NULL); @@ -188,6 +184,12 @@ static void st_destroy_context_priv( struct st_context *st ) void st_destroy_context( struct st_context *st ) { GLcontext *ctx = st->ctx; + + /* need to unbind and destroy CSO objects before anything else */ + cso_destroy_context(st->cso_context); + + _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache); + _mesa_free_context_data(ctx); st_destroy_context_priv(st); free(ctx); -- cgit v1.2.3 From ce454d2192918ae4b2535d0e76c68ebde3c4653f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 17 Apr 2008 11:13:05 -0600 Subject: gallium: assert that we're not deleting the currently bound shader --- src/gallium/drivers/softpipe/sp_state_fs.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 4eefd1d61f..2921066ce3 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -82,7 +82,10 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) void softpipe_delete_fs_state(struct pipe_context *pipe, void *fs) { + struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_fragment_shader *state = fs; + + assert(fs != softpipe->fs); state->delete( state ); } -- cgit v1.2.3 From 211170c192dfac5c2b39a6f34056255712fdfed6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 17 Apr 2008 11:31:14 -0600 Subject: gallium: plug in a select_destroy() function (fixes segfault on exit) --- src/mesa/state_tracker/st_cb_feedback.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 5315294c07..605bfee743 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -238,6 +238,12 @@ select_reset_stipple_counter( struct draw_stage *stage ) /* no-op */ } +static void +select_destroy( struct draw_stage *stage ) +{ + /* no-op */ +} + /** * Create GL selection mode drawing stage. @@ -254,6 +260,7 @@ draw_glselect_stage(GLcontext *ctx, struct draw_context *draw) fs->stage.tri = select_tri; fs->stage.flush = select_flush; fs->stage.reset_stipple_counter = select_reset_stipple_counter; + fs->stage.destroy = select_destroy; fs->ctx = ctx; return &fs->stage; -- cgit v1.2.3 From 5af9a690b2afc4c8f0f1118dad891b5ab53db597 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 17 Apr 2008 13:21:26 -0600 Subject: gallium: add missing pipe_surface_unmap() call --- src/mesa/state_tracker/st_cb_accum.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 4f9a22161c..1636bed91a 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -173,6 +173,8 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias, default: _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()"); } + + pipe_surface_unmap(acc_ps); } -- cgit v1.2.3 From 01b6354e72a84f8c3c22be1f77eab8d9c05920a3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 17 Apr 2008 15:26:37 -0600 Subject: gallium: tweak x/ybias values --- src/gallium/auxiliary/draw/draw_wide_point.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_wide_point.c b/src/gallium/auxiliary/draw/draw_wide_point.c index 0fa7c9313c..3d0add0c1a 100644 --- a/src/gallium/auxiliary/draw/draw_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_wide_point.c @@ -199,7 +199,7 @@ static void widepoint_first_point( struct draw_stage *stage, wide->ybias = 0.0; if (draw->rasterizer->gl_rasterization_rules) { - wide->ybias = -0.125; + wide->xbias = 0.125; } /* XXX we won't know the real size if it's computed by the vertex shader! */ -- cgit v1.2.3 From a773f06e969a3992451dd7fe6fd55ea96b2774fa Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 17 Apr 2008 23:44:32 +0100 Subject: draw: split off all the extra functionality in the vertex shader This will at least allow us to make the initial gains to get decent vertex performance much more quickly & with higher confidence of getting it right. At some later point can look again at code-generating all the fetch/cliptest/viewport extras in the same block as the vertex shader. For now, just need to get some decent baseline performance. --- src/gallium/auxiliary/draw/Makefile | 4 +- src/gallium/auxiliary/draw/draw_context.c | 26 +-- src/gallium/auxiliary/draw/draw_private.h | 2 + src/gallium/auxiliary/draw/draw_pt.h | 51 +++++- src/gallium/auxiliary/draw/draw_pt_emit.c | 16 +- src/gallium/auxiliary/draw/draw_pt_fetch.c | 175 ++++++++++++++++++ .../auxiliary/draw/draw_pt_fetch_pipeline.c | 4 +- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 109 +++++++++-- src/gallium/auxiliary/draw/draw_pt_pipeline.c | 5 +- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 202 +++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs_exec.c | 82 ++++++--- src/gallium/auxiliary/draw/draw_vs_llvm.c | 39 ++-- src/gallium/auxiliary/draw/draw_vs_sse.c | 38 ++-- 13 files changed, 635 insertions(+), 118 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_fetch.c create mode 100644 src/gallium/auxiliary/draw/draw_pt_post_vs.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 836e98f086..154c8a99b5 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -20,8 +20,10 @@ C_SOURCES = \ draw_pt_fetch_emit.c \ draw_pt_fetch_pipeline.c \ draw_pt_fetch_shade_pipeline.c \ - draw_pt_pipeline.c \ + draw_pt_fetch.c \ + draw_pt_post_vs.c \ draw_pt_emit.c \ + draw_pt_pipeline.c \ draw_pt_elts.c \ draw_prim.c \ draw_pstipple.c \ diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 9b2dcc0b57..4838b68ed1 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -110,6 +110,12 @@ struct draw_context *draw_create( void ) tgsi_exec_machine_init(&draw->machine); + /* FIXME: give this machine thing a proper constructor: + */ + draw->machine.Inputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); + draw->machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); + + if (!draw_pt_init( draw )) goto fail; @@ -155,8 +161,13 @@ void draw_destroy( struct draw_context *draw ) if (draw->pipeline.rasterize) draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); + if (draw->machine.Inputs) + align_free(draw->machine.Inputs); + if (draw->machine.Outputs) + align_free(draw->machine.Outputs); tgsi_exec_machine_free_data(&draw->machine); - + + if (draw->vs.vertex_cache) align_free( draw->vs.vertex_cache ); /* Frees all the vertices. */ @@ -265,6 +276,7 @@ draw_set_vertex_elements(struct draw_context *draw, draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); memcpy(draw->vertex_element, elements, count * sizeof(elements[0])); + draw->nr_vertex_elements = count; } @@ -463,15 +475,3 @@ boolean draw_get_edgeflag( struct draw_context *draw, return 1; } - -#if 0 -/* Crufty init function. Fix me. - */ -boolean draw_init_machine( struct draw_context *draw ) -{ - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); - machine->Inputs = ALIGN16_ASSIGN(inputs); - machine->Outputs = ALIGN16_ASSIGN(outputs); -} -#endif diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 9407217fd3..da94e69781 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -224,6 +224,8 @@ struct draw_context unsigned nr_vertex_buffers; struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; + unsigned nr_vertex_elements; + struct draw_vertex_shader *vertex_shader; boolean identity_viewport; diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 31d18ec62b..316289969b 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -112,6 +112,7 @@ struct draw_pt_middle_end { * mode... */ struct vbuf_render; +struct vertex_header; /* Helper functions. @@ -132,25 +133,25 @@ struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *d */ void draw_pt_run_pipeline( struct draw_context *draw, unsigned prim, - char *verts, - unsigned vertex_stride, + struct vertex_header *verts, unsigned vertex_count, + unsigned vertex_stride, const ushort *elts, unsigned count ); -/* HW vertex emit: +/******************************************************************************* + * HW vertex emit: */ struct pt_emit; void draw_pt_emit_prepare( struct pt_emit *emit, - unsigned prim, - unsigned opt ); + unsigned prim ); void draw_pt_emit( struct pt_emit *emit, - char *verts, - unsigned stride, + const float (*vertex_data)[4], unsigned vertex_count, + unsigned stride, const ushort *elts, unsigned count ); @@ -159,6 +160,42 @@ void draw_pt_emit_destroy( struct pt_emit *emit ); struct pt_emit *draw_pt_emit_create( struct draw_context *draw ); +/******************************************************************************* + * API vertex fetch: + */ + +struct pt_fetch; +void draw_pt_fetch_prepare( struct pt_fetch *fetch, + boolean emit_header, + unsigned vertex_size ); + +void draw_pt_fetch_run( struct pt_fetch *fetch, + const unsigned *elts, + unsigned count, + char *verts ); + +void draw_pt_fetch_destroy( struct pt_fetch *fetch ); + +struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ); + +/******************************************************************************* + * Post-VS: cliptest, rhw, viewport + */ +struct pt_post_vs; + +boolean draw_pt_post_vs_run( struct pt_post_vs *pvs, + struct vertex_header *pipeline_verts, + unsigned stride, + unsigned count ); + +void draw_pt_post_vs_prepare( struct pt_post_vs *pvs, + boolean bypass_clipping, + boolean identity_viewport, + boolean opengl ); + +struct pt_post_vs *draw_pt_post_vs_create( struct draw_context *draw ); + +void draw_pt_post_vs_destroy( struct pt_post_vs *pvs ); #endif diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index e9ed29450a..ef9db70a02 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -38,16 +38,11 @@ struct pt_emit { struct draw_context *draw; struct translate *translate; - - unsigned pipeline_vertex_size; - unsigned prim; - unsigned opt; }; void draw_pt_emit_prepare( struct pt_emit *emit, - unsigned prim, - unsigned opt ) + unsigned prim ) { struct draw_context *draw = emit->draw; const struct vertex_info *vinfo; @@ -75,8 +70,7 @@ void draw_pt_emit_prepare( struct pt_emit *emit, unsigned emit_sz = 0; unsigned src_buffer = 0; unsigned output_format; - unsigned src_offset = (sizeof(struct vertex_header) + - vinfo->src_index[i] * 4 * sizeof(float) ); + unsigned src_offset = (vinfo->src_index[i] * 4 * sizeof(float) ); @@ -139,9 +133,9 @@ void draw_pt_emit_prepare( struct pt_emit *emit, void draw_pt_emit( struct pt_emit *emit, - char *verts, - unsigned stride, + const float (*vertex_data)[4], unsigned vertex_count, + unsigned stride, const ushort *elts, unsigned count ) { @@ -164,7 +158,7 @@ void draw_pt_emit( struct pt_emit *emit, translate->set_buffer(translate, 0, - verts, + vertex_data, stride ); translate->set_buffer(translate, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c new file mode 100644 index 0000000000..a7553023b8 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -0,0 +1,175 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" +#include "translate/translate.h" + + +struct pt_fetch { + struct draw_context *draw; + + struct translate *translate; + + unsigned vertex_size; +}; + + + +/* Perform the fetch from API vertex elements & vertex buffers, to a + * contiguous set of float[4] attributes as required for the + * vertex_shader->run_linear() method. + * + * This is used in all cases except pure passthrough + * (draw_pt_fetch_emit.c) which has its own version to translate + * directly to hw vertices. + * + */ +void draw_pt_fetch_prepare( struct pt_fetch *fetch, + boolean emit_header, + unsigned vertex_size ) +{ + struct draw_context *draw = fetch->draw; + unsigned i, nr = 0; + unsigned dst_offset = 0; + struct translate_key key; + + fetch->vertex_size = vertex_size; + + memset(&key, 0, sizeof(key)); + + /* If PT_SHADE is not set, then we are creating post-shader + * vertices, meaning that we need to emit/leave space for a vertex + * header. + * + * It's worth considering whether the vertex headers should contain + * a pointer to the 'data', rather than having it inline. + * Something to look at after we've fully switched over to the pt + * paths. + */ + if (emit_header) + { + /* Need to set header->vertex_id = 0xffff somehow. + */ + key.element[nr].input_format = PIPE_FORMAT_R32_FLOAT; + key.element[nr].input_buffer = draw->nr_vertex_buffers; + key.element[nr].input_offset = 0; + key.element[nr].output_format = PIPE_FORMAT_R32_FLOAT; + key.element[nr].output_offset = dst_offset; + dst_offset += 1 * sizeof(float); + nr++; + + + /* Just leave the clip[] array untouched. + */ + dst_offset += 4 * sizeof(float); + } + + + for (i = 0; i < draw->nr_vertex_elements; i++) { + key.element[nr].input_format = draw->vertex_element[i].src_format; + key.element[nr].input_buffer = draw->vertex_element[i].vertex_buffer_index; + key.element[nr].input_offset = draw->vertex_element[i].src_offset; + key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + key.element[nr].output_offset = dst_offset; + + dst_offset += 4 * sizeof(float); + nr++; + } + + assert(dst_offset <= vertex_size); + + key.nr_elements = nr; + key.output_stride = vertex_size; + + + /* Don't bother with caching at this stage: + */ + if (!fetch->translate || + memcmp(&fetch->translate->key, &key, sizeof(key)) != 0) + { + if (fetch->translate) + fetch->translate->release(fetch->translate); + + fetch->translate = translate_generic_create( &key ); + + if (emit_header) { + static struct vertex_header vh = { 0, 0, 0, 0xffff }; + fetch->translate->set_buffer(fetch->translate, + draw->nr_vertex_buffers, + &vh, + 0); + } + } +} + + + + +void draw_pt_fetch_run( struct pt_fetch *fetch, + const unsigned *elts, + unsigned count, + char *verts ) +{ + struct draw_context *draw = fetch->draw; + struct translate *translate = fetch->translate; + unsigned i; + + for (i = 0; i < draw->nr_vertex_buffers; i++) { + translate->set_buffer(translate, + i, + ((char *)draw->user.vbuffer[i] + + draw->vertex_buffer[i].buffer_offset), + draw->vertex_buffer[i].pitch ); + } + + translate->run_elts( translate, + elts, + count, + verts ); +} + + +struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) +{ + struct pt_fetch *fetch = CALLOC_STRUCT(pt_fetch); + if (!fetch) + return NULL; + + fetch->draw = draw; + return fetch; +} + +void draw_pt_fetch_destroy( struct pt_fetch *fetch ) +{ + FREE(fetch); +} + diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c index 79548d4156..26d0b37286 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c @@ -286,9 +286,9 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, */ draw_pt_run_pipeline( fpme->draw, fpme->prim, - pipeline_verts, - fpme->pipeline_vertex_size, + (struct vertex_header *)pipeline_verts, fetch_count, + fpme->pipeline_vertex_size, draw_elts, draw_count ); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 8db9e31e2d..0b9e8d15ba 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -39,8 +39,11 @@ struct fetch_pipeline_middle_end { struct draw_context *draw; struct pt_emit *emit; + struct pt_fetch *fetch; + struct pt_post_vs *post_vs; - unsigned pipeline_vertex_size; + unsigned vertex_data_offset; + unsigned vertex_size; unsigned prim; unsigned opt; }; @@ -51,15 +54,43 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, unsigned opt ) { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; + struct draw_context *draw = fpme->draw; + struct draw_vertex_shader *vs = draw->vertex_shader; + unsigned nr = MAX2( vs->info.num_inputs, + vs->info.num_outputs ); fpme->prim = prim; fpme->opt = opt; + /* Always leave room for the vertex header whether we need it or + * not. It's hard to get rid of it in particular because of the + * viewport code in draw_pt_post_vs.c. + */ + fpme->vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); + + + + draw_pt_fetch_prepare( fpme->fetch, + (opt & (PT_CLIPTEST | PT_PIPELINE)) != 0, + fpme->vertex_size ); + + /* XXX: it's not really gl rasterization rules we care about here, + * but gl vs dx9 clip spaces. + */ + draw_pt_post_vs_prepare( fpme->post_vs, + draw->rasterizer->bypass_clipping, + draw->identity_viewport, + draw->rasterizer->gl_rasterization_rules ); + + if (!(opt & PT_PIPELINE)) - draw_pt_emit_prepare( fpme->emit, prim, opt ); + draw_pt_emit_prepare( fpme->emit, + prim ); + + /* No need to prepare the shader. + */ + vs->prepare(vs, draw); - //fpme->pipeline_vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); - fpme->pipeline_vertex_size = MAX_VERTEX_ALLOCATION; } @@ -74,44 +105,63 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; struct draw_vertex_shader *shader = draw->vertex_shader; - char *pipeline_verts; - unsigned pipeline = PT_PIPELINE; + unsigned opt = fpme->opt; - pipeline_verts = MALLOC(fpme->pipeline_vertex_size * - fetch_count); + struct vertex_header *pipeline_verts = + (struct vertex_header *)MALLOC(fpme->vertex_size * fetch_count); if (!pipeline_verts) { assert(0); return; } - - /* Shade + /* Fetch into our vertex buffer */ - shader->prepare(shader, draw); - - if (shader->run(shader, draw, fetch_elts, fetch_count, pipeline_verts, - fpme->pipeline_vertex_size)) + draw_pt_fetch_run( fpme->fetch, + fetch_elts, + fetch_count, + (char *)pipeline_verts ); + + /* Run the shader, note that this overwrites the data[] parts of + * the pipeline verts. If there is no shader, ie a bypass shader, + * then the inputs == outputs, and are already in the correct + * place. + */ + if (opt & PT_SHADE) { - pipeline |= PT_CLIPTEST; + shader->run_linear(shader, + (const float (*)[4])pipeline_verts->data, + ( float (*)[4])pipeline_verts->data, + (const float (*)[4])draw->user.constants, + fetch_count, + fpme->vertex_size, + fpme->vertex_size); } + if (draw_pt_post_vs_run( fpme->post_vs, + pipeline_verts, + fetch_count, + fpme->vertex_size )) + { + opt |= PT_PIPELINE; + } /* Do we need to run the pipeline? */ - if (fpme->opt & pipeline) { + if (opt & PT_PIPELINE) { draw_pt_run_pipeline( fpme->draw, fpme->prim, pipeline_verts, - fpme->pipeline_vertex_size, fetch_count, + fpme->vertex_size, draw_elts, draw_count ); - } else { + } + else { draw_pt_emit( fpme->emit, - pipeline_verts, - fpme->pipeline_vertex_size, + (const float (*)[4])pipeline_verts->data, fetch_count, + fpme->vertex_size, draw_elts, draw_count ); } @@ -129,6 +179,17 @@ static void fetch_pipeline_finish( struct draw_pt_middle_end *middle ) static void fetch_pipeline_destroy( struct draw_pt_middle_end *middle ) { + struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; + + if (fpme->fetch) + draw_pt_fetch_destroy( fpme->fetch ); + + if (fpme->emit) + draw_pt_emit_destroy( fpme->emit ); + + if (fpme->post_vs) + draw_pt_post_vs_destroy( fpme->post_vs ); + FREE(middle); } @@ -146,6 +207,14 @@ struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit( struct draw_context * fpme->draw = draw; + fpme->fetch = draw_pt_fetch_create( draw ); + if (!fpme->fetch) + goto fail; + + fpme->post_vs = draw_pt_post_vs_create( draw ); + if (!fpme->post_vs) + goto fail; + fpme->emit = draw_pt_emit_create( draw ); if (!fpme->emit) goto fail; diff --git a/src/gallium/auxiliary/draw/draw_pt_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_pipeline.c index 17ce6febec..1a9a3adb03 100644 --- a/src/gallium/auxiliary/draw/draw_pt_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_pipeline.c @@ -117,12 +117,13 @@ void draw_pt_reset_vertex_ids( struct draw_context *draw ) */ void draw_pt_run_pipeline( struct draw_context *draw, unsigned prim, - char *verts, - unsigned stride, + struct vertex_header *pipeline_verts, unsigned vertex_count, + unsigned stride, const ushort *elts, unsigned count ) { + char *verts = (char *)pipeline_verts; unsigned i; draw->pt.pipeline.verts = verts; diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c new file mode 100644 index 0000000000..315b02f4ee --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -0,0 +1,202 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_context.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" + +struct pt_post_vs { + struct draw_context *draw; + + boolean (*run)( struct pt_post_vs *pvs, + struct vertex_header *vertices, + unsigned count, + unsigned stride ); +}; + + + +static INLINE unsigned +compute_clipmask_gl(const float *clip, /*const*/ float plane[][4], unsigned nr) +{ + unsigned mask = 0x0; + unsigned i; + + /* Do the hardwired planes first: + */ + if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; + if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; + if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; + if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; + if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; + if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + + /* Followed by any remaining ones: + */ + for (i = 6; i < nr; i++) { + if (dot4(clip, plane[i]) < 0) + mask |= (1<draw->viewport.scale; + const float *trans = pvs->draw->viewport.translate; + unsigned j; + unsigned clipped = 0; + + for (j = 0; j < count; j++) { + out->clip[0] = out->data[0][0]; + out->clip[1] = out->data[0][1]; + out->clip[2] = out->data[0][2]; + out->clip[3] = out->data[0][3]; + + out->vertex_id = 0xffff; + out->edgeflag = 1; + out->clipmask = compute_clipmask_gl(out->clip, + pvs->draw->plane, + pvs->draw->nr_planes); + clipped += out->clipmask; + + if (out->clipmask == 0) + { + /* divide by w */ + float w = 1.0f / out->data[0][3]; + + /* Viewport mapping */ + out->data[0][0] = out->data[0][0] * w * scale[0] + trans[0]; + out->data[0][1] = out->data[0][1] * w * scale[1] + trans[1]; + out->data[0][2] = out->data[0][2] * w * scale[2] + trans[2]; + out->data[0][3] = w; + } + + out = (struct vertex_header *)( (char *)out + stride ); + } + + return clipped != 0; +} + + + +/* If bypass_clipping is set, skip cliptest and rhw divide. + */ +static boolean post_vs_viewport( struct pt_post_vs *pvs, + struct vertex_header *vertices, + unsigned count, + unsigned stride ) +{ + struct vertex_header *out = vertices; + const float *scale = pvs->draw->viewport.scale; + const float *trans = pvs->draw->viewport.translate; + unsigned j; + + debug_printf("%s\n", __FUNCTION__); + for (j = 0; j < count; j++) { + /* Viewport mapping only, no cliptest/rhw divide + */ + out->data[0][0] = out->data[0][0] * scale[0] + trans[0]; + out->data[0][1] = out->data[0][1] * scale[1] + trans[1]; + out->data[0][2] = out->data[0][2] * scale[2] + trans[2]; + + out = (struct vertex_header *)((char *)out + stride); + } + + return FALSE; +} + + +/* If bypass_clipping is set and we have an identity viewport, nothing + * to do. + */ +static boolean post_vs_none( struct pt_post_vs *pvs, + struct vertex_header *vertices, + unsigned count, + unsigned stride ) +{ + debug_printf("%s\n", __FUNCTION__); + return FALSE; +} + +boolean draw_pt_post_vs_run( struct pt_post_vs *pvs, + struct vertex_header *pipeline_verts, + unsigned count, + unsigned stride ) +{ + return pvs->run( pvs, pipeline_verts, count, stride ); +} + + +void draw_pt_post_vs_prepare( struct pt_post_vs *pvs, + boolean bypass_clipping, + boolean identity_viewport, + boolean opengl ) +{ + if (bypass_clipping) { + if (identity_viewport) + pvs->run = post_vs_none; + else + pvs->run = post_vs_viewport; + } + else { + //if (opengl) + pvs->run = post_vs_cliptest_viewport_gl; + } +} + + +struct pt_post_vs *draw_pt_post_vs_create( struct draw_context *draw ) +{ + struct pt_post_vs *pvs = CALLOC_STRUCT( pt_post_vs ); + if (!pvs) + return NULL; + + pvs->draw = draw; + + return pvs; +} + +void draw_pt_post_vs_destroy( struct pt_post_vs *pvs ) +{ + FREE(pvs); +} diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 0e05b79715..184151b9b1 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -58,8 +58,10 @@ static void vs_exec_prepare( struct draw_vertex_shader *shader, struct draw_context *draw ) { + struct exec_vertex_shader *evs = exec_vertex_shader(shader); + /* specify the vertex program to interpret/execute */ - tgsi_exec_machine_bind_shader(&draw->machine, + tgsi_exec_machine_bind_shader(evs->machine, shader->state.tokens, PIPE_MAX_SAMPLERS, NULL /*samplers*/ ); @@ -84,31 +86,45 @@ vs_exec_run( struct draw_vertex_shader *shader, void *vOut, unsigned vertex_size) { - struct tgsi_exec_machine *machine = &draw->machine; + struct exec_vertex_shader *evs = exec_vertex_shader(shader); + struct tgsi_exec_machine *machine = evs->machine; unsigned int i, j; unsigned int clipped = 0; - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); + struct tgsi_exec_vector *outputs = 0; const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; assert(shader->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); machine->Consts = (const float (*)[4]) draw->user.constants; - machine->Inputs = ALIGN16_ASSIGN(inputs); + if (draw->rasterizer->bypass_vs) { /* outputs are just the inputs */ - machine->Outputs = machine->Inputs; + outputs = machine->Inputs; } else { - machine->Outputs = ALIGN16_ASSIGN(outputs); + outputs = machine->Outputs; } for (i = 0; i < count; i += MAX_TGSI_VERTICES) { unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); draw->vertex_fetch.fetch_func( draw, machine, &elts[i], max_vertices ); +#if 0 + for (j = 0; j < max_vertices; j++) { + unsigned slot; + debug_printf("%d) Input vert:\n", i + j); + for (slot = 0; slot < shader->info.num_inputs; slot++) { + debug_printf("\t%d: %f %f %f %f\n", slot, + machine->Inputs[slot].xyzw[0].f[j], + machine->Inputs[slot].xyzw[1].f[j], + machine->Inputs[slot].xyzw[2].f[j], + machine->Inputs[slot].xyzw[3].f[j]); + } + } +#endif + + if (!draw->rasterizer->bypass_vs) { /* run interpreter */ tgsi_exec_machine_run( machine ); @@ -127,10 +143,10 @@ vs_exec_run( struct draw_vertex_shader *shader, * program as a set of DP4 instructions appended to the * user-provided code. */ - x = out->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = out->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = out->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = out->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + x = out->clip[0] = outputs[0].xyzw[0].f[j]; + y = out->clip[1] = outputs[0].xyzw[1].f[j]; + z = out->clip[2] = outputs[0].xyzw[2].f[j]; + w = out->clip[3] = outputs[0].xyzw[3].f[j]; if (!draw->rasterizer->bypass_clipping) { out->clipmask = compute_clipmask(out->clip, draw->plane, @@ -156,7 +172,8 @@ vs_exec_run( struct draw_vertex_shader *shader, out->data[0][2] = z * scale[2] + trans[2]; out->data[0][3] = w; } - else { + else + { out->data[0][0] = x; out->data[0][1] = y; out->data[0][2] = z; @@ -167,10 +184,10 @@ vs_exec_run( struct draw_vertex_shader *shader, * vertex attrib slots. */ for (slot = 1; slot < draw->num_vs_outputs; slot++) { - out->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - out->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - out->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - out->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + out->data[slot][0] = outputs[slot].xyzw[0].f[j]; + out->data[slot][1] = outputs[slot].xyzw[1].f[j]; + out->data[slot][2] = outputs[slot].xyzw[2].f[j]; + out->data[slot][3] = outputs[slot].xyzw[3].f[j]; } #if 0 /*DEBUG*/ @@ -216,12 +233,25 @@ vs_exec_run_linear( struct draw_vertex_shader *shader, /* Swizzle inputs. */ for (j = 0; j < max_vertices; j++) { +#if 0 + debug_printf("%d) Input vert:\n", i + j); + for (slot = 0; slot < shader->info.num_inputs; slot++) { + debug_printf("\t%d: %f %f %f %f\n", slot, + input[slot][0], + input[slot][1], + input[slot][2], + input[slot][3]); + } +#endif + for (slot = 0; slot < shader->info.num_inputs; slot++) { machine->Inputs[slot].xyzw[0].f[j] = input[slot][0]; machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; } + + input = (const float (*)[4])((const char *)input + input_stride); } /* run interpreter */ @@ -235,13 +265,23 @@ vs_exec_run_linear( struct draw_vertex_shader *shader, output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + +#if 0 + debug_printf("%d) Post xform vert:\n", i + j); + for (slot = 0; slot < shader->info.num_outputs; slot++) { + debug_printf("\t%d: %f %f %f %f\n", slot, + output[slot][0], + output[slot][1], + output[slot][2], + output[slot][3]); + } +#endif + + output = (float (*)[4])((char *)output + output_stride); } - /* Advance input, output pointers: - */ - input = (const float (*)[4])((const char *)input + input_stride); - output = (float (*)[4])((char *)output + output_stride); } } diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index d0baca715e..4dbfa955a4 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -47,6 +47,7 @@ struct draw_llvm_vertex_shader { struct draw_vertex_shader base; struct gallivm_prog *llvm_prog; + struct tgsi_exec_machine *machine; }; @@ -77,12 +78,9 @@ vs_llvm_run( struct draw_vertex_shader *base, struct draw_llvm_vertex_shader *shader = (struct draw_llvm_vertex_shader *)base; - struct tgsi_exec_machine *machine = &draw->machine; + struct tgsi_exec_machine *machine = shader->machine; unsigned int j; unsigned int clipped = 0; - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; @@ -93,13 +91,12 @@ vs_llvm_run( struct draw_vertex_shader *base, /* Consts does not require 16 byte alignment. */ machine->Consts = (float (*)[4]) draw->user.constants; - machine->Inputs = ALIGN16_ASSIGN(inputs); if (draw->rasterizer->bypass_vs) { /* outputs are just the inputs */ - machine->Outputs = machine->Inputs; + outputs = machine->Inputs; } else { - machine->Outputs = ALIGN16_ASSIGN(outputs); + outputs = machine->Outputs; } @@ -119,10 +116,10 @@ vs_llvm_run( struct draw_vertex_shader *base, unsigned slot; float x, y, z, w; - x = vOut[j]->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + x = vOut[j]->clip[0] = outputs[0].xyzw[0].f[j]; + y = vOut[j]->clip[1] = outputs[0].xyzw[1].f[j]; + z = vOut[j]->clip[2] = outputs[0].xyzw[2].f[j]; + w = vOut[j]->clip[3] = outputs[0].xyzw[3].f[j]; if (!draw->rasterizer->bypass_clipping) { vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, @@ -159,10 +156,10 @@ vs_llvm_run( struct draw_vertex_shader *base, * vertex attrib slots. */ for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[j]->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + vOut[j]->data[slot][0] = outputs[slot].xyzw[0].f[j]; + vOut[j]->data[slot][1] = outputs[slot].xyzw[1].f[j]; + vOut[j]->data[slot][2] = outputs[slot].xyzw[2].f[j]; + vOut[j]->data[slot][3] = outputs[slot].xyzw[3].f[j]; } } /* loop over vertices */ return clipped != 0; @@ -183,7 +180,7 @@ vs_llvm_run_linear( struct draw_vertex_shader *base, struct draw_llvm_vertex_shader *shader = (struct draw_llvm_vertex_shader *)base; - struct tgsi_exec_machine *machine = &draw->machine; + struct tgsi_exec_machine *machine = shader->machine; unsigned int j; @@ -199,6 +196,8 @@ vs_llvm_run_linear( struct draw_vertex_shader *base, machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; } + + input = (const float (*)[4])((const char *)input + input_stride); } /* run shader */ @@ -216,12 +215,9 @@ vs_llvm_run_linear( struct draw_vertex_shader *base, output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; - } - /* Advance input, output pointers: - */ - input = (const float (*)[4])((const char *)input + input_stride); - output = (float (*)[4])((char *)output + output_stride); + output = (float (*)[4])((char *)output + output_stride); + } } } @@ -263,6 +259,7 @@ draw_create_vs_llvm(struct draw_context *draw, vs->base.run = vs_llvm_run; vs->base.run_linear = vs_llvm_run_linear; vs->base.delete = vs_llvm_delete; + vs->machine = &draw->machine; { struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 873ecfdc5d..a763f3845c 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -91,12 +91,10 @@ vs_sse_run( struct draw_vertex_shader *base, unsigned vertex_size ) { struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; - struct tgsi_exec_machine *machine = &draw->machine; + struct tgsi_exec_machine *machine = shader->machine; unsigned int i, j; unsigned int clipped = 0; - - ALIGN16_DECL(struct tgsi_exec_vector, inputs, PIPE_MAX_ATTRIBS); - ALIGN16_DECL(struct tgsi_exec_vector, outputs, PIPE_MAX_ATTRIBS); + struct tgsi_exec_vector *outputs = 0; const float *scale = draw->viewport.scale; const float *trans = draw->viewport.translate; @@ -104,13 +102,13 @@ vs_sse_run( struct draw_vertex_shader *base, /* Consts does not require 16 byte alignment. */ machine->Consts = (const float (*)[4]) draw->user.constants; - machine->Inputs = ALIGN16_ASSIGN(inputs); + if (draw->rasterizer->bypass_vs) { /* outputs are just the inputs */ - machine->Outputs = machine->Inputs; + outputs = machine->Inputs; } else { - machine->Outputs = ALIGN16_ASSIGN(outputs); + outputs = machine->Outputs; } for (i = 0; i < count; i += SSE_MAX_VERTICES) { @@ -142,10 +140,10 @@ vs_sse_run( struct draw_vertex_shader *base, struct vertex_header *out = draw_header_from_block(vOut, vertex_size, i + j); - x = out->clip[0] = machine->Outputs[0].xyzw[0].f[j]; - y = out->clip[1] = machine->Outputs[0].xyzw[1].f[j]; - z = out->clip[2] = machine->Outputs[0].xyzw[2].f[j]; - w = out->clip[3] = machine->Outputs[0].xyzw[3].f[j]; + x = out->clip[0] = outputs[0].xyzw[0].f[j]; + y = out->clip[1] = outputs[0].xyzw[1].f[j]; + z = out->clip[2] = outputs[0].xyzw[2].f[j]; + w = out->clip[3] = outputs[0].xyzw[3].f[j]; if (!draw->rasterizer->bypass_clipping) { out->clipmask = compute_clipmask(out->clip, draw->plane, @@ -182,10 +180,10 @@ vs_sse_run( struct draw_vertex_shader *base, * vertex attrib slots. */ for (slot = 1; slot < draw->num_vs_outputs; slot++) { - out->data[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - out->data[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - out->data[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - out->data[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + out->data[slot][0] = outputs[slot].xyzw[0].f[j]; + out->data[slot][1] = outputs[slot].xyzw[1].f[j]; + out->data[slot][2] = outputs[slot].xyzw[2].f[j]; + out->data[slot][3] = outputs[slot].xyzw[3].f[j]; } #if 0 /*DEBUG*/ printf("%d) Post xform vert:\n", i + j); @@ -233,6 +231,8 @@ vs_sse_run_linear( struct draw_vertex_shader *base, machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; } + + input = (const float (*)[4])((const char *)input + input_stride); } /* run compiled shader @@ -253,12 +253,9 @@ vs_sse_run_linear( struct draw_vertex_shader *base, output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; } - } - /* Advance input, output pointers: - */ - input = (const float (*)[4])((const char *)input + input_stride); - output = (float (*)[4])((char *)output + output_stride); + output = (float (*)[4])((char *)output + output_stride); + } } } @@ -300,6 +297,7 @@ draw_create_vs_sse(struct draw_context *draw, vs->base.run = vs_sse_run; vs->base.run_linear = vs_sse_run_linear; vs->base.delete = vs_sse_delete; + vs->machine = &draw->machine; x86_init_func( &vs->sse2_program ); -- cgit v1.2.3 From da9079b936684f88da79425a810d7902e4d6e7ad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 18 Apr 2008 20:59:28 +0900 Subject: gallium: Add missing files to scons. --- src/gallium/auxiliary/draw/SConscript | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 62e0cd9e50..0274d48e7f 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -6,24 +6,23 @@ draw = env.ConvenienceLibrary( 'draw_aaline.c', 'draw_aapoint.c', 'draw_clip.c', - 'draw_vs_exec.c', - 'draw_vs_sse.c', - 'draw_vs_llvm.c', 'draw_context.c', 'draw_cull.c', 'draw_debug.c', 'draw_flatshade.c', 'draw_offset.c', + 'draw_prim.c', + 'draw_pstipple.c', 'draw_pt.c', - 'draw_pt_vcache.c', + 'draw_pt_elts.c', 'draw_pt_emit.c', + 'draw_pt_fetch.c', 'draw_pt_fetch_emit.c', 'draw_pt_fetch_pipeline.c', 'draw_pt_fetch_shade_pipeline.c', 'draw_pt_pipeline.c', - 'draw_pt_elts.c', - 'draw_prim.c', - 'draw_pstipple.c', + 'draw_pt_post_vs.c', + 'draw_pt_vcache.c', 'draw_stipple.c', 'draw_twoside.c', 'draw_unfilled.c', @@ -36,8 +35,11 @@ draw = env.ConvenienceLibrary( 'draw_vf.c', 'draw_vf_generic.c', 'draw_vf_sse.c', + 'draw_vs_exec.c', + 'draw_vs_llvm.c', + 'draw_vs_sse.c', + 'draw_wide_line.c', 'draw_wide_point.c', - 'draw_wide_line.c' ]) auxiliaries.insert(0, draw) -- cgit v1.2.3 From 073d9a28c2dc955956c940be6fcc4b3ab354cc6c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 18 Apr 2008 13:38:06 +0100 Subject: gallium: Always pass colour clear value to driver as A8R8G8B8. --- src/mesa/state_tracker/st_cb_clear.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index fa9f986f11..6469511c6f 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -406,7 +406,7 @@ clear_color_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) /* clear whole buffer w/out masking */ struct st_renderbuffer *strb = st_renderbuffer(rb); uint clearValue; - util_pack_color(ctx->Color.ClearColor, strb->surface->format, &clearValue); + util_pack_color(ctx->Color.ClearColor, PIPE_FORMAT_A8R8G8B8_UNORM, &clearValue); ctx->st->pipe->clear(ctx->st->pipe, strb->surface, clearValue); } } -- cgit v1.2.3 From 5b97c762ed9882dd922f48c2fbf13b14ad86a96e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 17:32:39 +0100 Subject: rtasm: add a couple more insns, clean up x86_mul --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 34 ++++++++++++++++++++++++------ src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 2 ++ 2 files changed, 29 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index aea8b28e58..5c25fa155d 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -33,11 +33,6 @@ #define DISASSEM 0 #define X86_TWOB 0x0f -static unsigned char *cptr( void (*label)() ) -{ - return (unsigned char *) label; -} - static void do_realloc( struct x86_function *p ) { @@ -304,6 +299,11 @@ void x86_jmp( struct x86_function *p, unsigned char *label) } #if 0 +static unsigned char *cptr( void (*label)() ) +{ + return (unsigned char *) label; +} + /* This doesn't work once we start reallocating & copying the * generated code on buffer fills, because the call is relative to the * current pc. @@ -417,11 +417,14 @@ void x86_add( struct x86_function *p, emit_op_modrm(p, 0x03, 0x01, dst, src ); } +/* Calculate EAX * src, results in EDX:EAX. + */ void x86_mul( struct x86_function *p, struct x86_reg src ) { - assert (src.file == file_REG32 && src.mod == mod_REG); - emit_op_modrm(p, 0xf7, 0, x86_make_reg (file_REG32, reg_SP), src ); +// assert (src.file == file_REG32 && src.mod == mod_REG); + emit_1ub(p, 0xf7); + emit_modrm_noreg(p, 4, src ); } void x86_sub( struct x86_function *p, @@ -646,6 +649,14 @@ void sse_cvtps2pi( struct x86_function *p, emit_modrm( p, dst, src ); } +void sse2_cvtdq2ps( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0x5b); + emit_modrm( p, dst, src ); +} + /* Shufps can also be used to implement a reduced swizzle when dest == * arg0. @@ -735,6 +746,15 @@ void sse2_packuswb( struct x86_function *p, emit_modrm( p, dst, src ); } +void sse2_punpcklbw( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_3ub(p, 0x66, X86_TWOB, 0x60); + emit_modrm( p, dst, src ); +} + + void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index 606b41eb35..dfde661f46 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -165,6 +165,7 @@ void mmx_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg sr void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_cvttps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse2_cvtdq2ps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -202,6 +203,7 @@ void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, unsigned char shuf ); void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); +void sse2_punpcklbw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -- cgit v1.2.3 From f631bebe1a05c8af863e514546763433343b7c53 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 17:34:55 +0100 Subject: softpipe: fix const-related compiler warnings --- src/gallium/drivers/softpipe/sp_fs_sse.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 5ef02a7142..f857d26143 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -49,7 +49,7 @@ typedef void (XSTDCALL *codegen_function)( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, - float (*constant)[4], + const float (*constant)[4], struct tgsi_exec_vector *temporary, const struct tgsi_interp_coef *coef, float (*immediates)[4] @@ -67,9 +67,9 @@ struct sp_sse_fragment_shader { static void -fs_sse_prepare( struct sp_fragment_shader *base, - struct tgsi_exec_machine *machine, - struct tgsi_sampler *samplers ) +fs_sse_prepare( const struct sp_fragment_shader *base, + struct tgsi_exec_machine *machine, + struct tgsi_sampler *samplers ) { } @@ -80,9 +80,9 @@ fs_sse_prepare( struct sp_fragment_shader *base, * TODO: process >1 quad at a time */ static unsigned -fs_sse_run( struct sp_fragment_shader *base, - struct tgsi_exec_machine *machine, - struct quad_header *quad ) +fs_sse_run( const struct sp_fragment_shader *base, + struct tgsi_exec_machine *machine, + struct quad_header *quad ) { struct sp_sse_fragment_shader *shader = (struct sp_sse_fragment_shader *) base; -- cgit v1.2.3 From c5f0158a9179463593d63b33cf3b5490167faac9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 17:35:32 +0100 Subject: tgsi: add const qualifier to tokens on sse emit --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2 +- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index c37e201b2b..0a3a7559ca 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -2332,7 +2332,7 @@ emit_declaration( */ unsigned tgsi_emit_sse2( - struct tgsi_token *tokens, + const struct tgsi_token *tokens, struct x86_function *func, float (*immediates)[4]) { diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h index d56bf7f98a..063287dc5e 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h @@ -10,7 +10,7 @@ struct x86_function; unsigned tgsi_emit_sse2( - struct tgsi_token *tokens, + const struct tgsi_token *tokens, struct x86_function *function, float (*immediates)[4] ); -- cgit v1.2.3 From 363f7abf2000c1cf5993ae8f83ba81b2054bf6e0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 18:30:41 +0100 Subject: rtasm: add x86_imul --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 11 ++++++++++- src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 1 + 2 files changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 5c25fa155d..7f8cc23d15 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -422,11 +422,20 @@ void x86_add( struct x86_function *p, void x86_mul( struct x86_function *p, struct x86_reg src ) { -// assert (src.file == file_REG32 && src.mod == mod_REG); emit_1ub(p, 0xf7); emit_modrm_noreg(p, 4, src ); } + +void x86_imul( struct x86_function *p, + struct x86_reg dst, + struct x86_reg src ) +{ + emit_2ub(p, X86_TWOB, 0xAF); + emit_modrm(p, dst, src); +} + + void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index dfde661f46..5e99ceea70 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -213,6 +213,7 @@ void x86_inc( struct x86_function *p, struct x86_reg reg ); void x86_lea( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_mov( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_mul( struct x86_function *p, struct x86_reg src ); +void x86_imul( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_or( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_pop( struct x86_function *p, struct x86_reg reg ); void x86_push( struct x86_function *p, struct x86_reg reg ); -- cgit v1.2.3 From 7400bc4b6fb0c20a935cd108afa92814eeafec6d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 18:31:45 +0100 Subject: translate: add sse version based on old draw_vf_sse.c --- src/gallium/auxiliary/draw/draw_pt_emit.c | 2 +- src/gallium/auxiliary/draw/draw_pt_fetch.c | 2 +- src/gallium/auxiliary/translate/Makefile | 4 +- src/gallium/auxiliary/translate/SConscript | 2 + src/gallium/auxiliary/translate/translate.h | 4 + src/gallium/auxiliary/translate/translate_sse.c | 615 ++++++++++++++++++++++++ 6 files changed, 626 insertions(+), 3 deletions(-) create mode 100644 src/gallium/auxiliary/translate/translate_sse.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index ef9db70a02..a6a3ff6cac 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -127,7 +127,7 @@ void draw_pt_emit_prepare( struct pt_emit *emit, if (emit->translate) emit->translate->release(emit->translate); - emit->translate = translate_generic_create( &hw_key ); + emit->translate = translate_create( &hw_key ); } } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index a7553023b8..1aed251c85 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -119,7 +119,7 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, if (fetch->translate) fetch->translate->release(fetch->translate); - fetch->translate = translate_generic_create( &key ); + fetch->translate = translate_create( &key ); if (emit_header) { static struct vertex_header vh = { 0, 0, 0, 0xffff }; diff --git a/src/gallium/auxiliary/translate/Makefile b/src/gallium/auxiliary/translate/Makefile index 051987bb7e..39dfb0de30 100644 --- a/src/gallium/auxiliary/translate/Makefile +++ b/src/gallium/auxiliary/translate/Makefile @@ -4,7 +4,9 @@ include $(TOP)/configs/current LIBNAME = translate C_SOURCES = \ - translate_generic.c + translate_generic.c \ + translate_sse.c \ + translate.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/translate/SConscript b/src/gallium/auxiliary/translate/SConscript index 42513ba3b0..7608908915 100644 --- a/src/gallium/auxiliary/translate/SConscript +++ b/src/gallium/auxiliary/translate/SConscript @@ -4,6 +4,8 @@ translate = env.ConvenienceLibrary( target = 'translate', source = [ 'translate_generic.c', + 'translate_sse.c', + 'translate.c', ]) auxiliaries.insert(0, translate) diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h index 4f9f40e51a..d95d1ac4f3 100644 --- a/src/gallium/auxiliary/translate/translate.h +++ b/src/gallium/auxiliary/translate/translate.h @@ -42,6 +42,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_format.h" +#include "pipe/p_state.h" struct translate_element { @@ -93,6 +94,9 @@ struct translate *translate_lookup_or_create( struct translate_context *tctx, #endif +struct translate *translate_create( const struct translate_key *key ); + + /******************************************************************************* * Private: */ diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c new file mode 100644 index 0000000000..cb8815c173 --- /dev/null +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -0,0 +1,615 @@ +/* + * Copyright 2003 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: + * Keith Whitwell + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_util.h" +#include "util/u_simple_list.h" + +#include "translate.h" + + +#if defined(__i386__) || defined(__386__) || defined(i386) + +#include "rtasm/rtasm_cpu.h" +#include "rtasm/rtasm_x86sse.h" + + +#define X 0 +#define Y 1 +#define Z 2 +#define W 3 + + +typedef void (*run_func)( struct translate *translate, + unsigned start, + unsigned count, + void *output_buffer ); + +typedef void (*run_elts_func)( struct translate *translate, + const unsigned *elts, + unsigned count, + void *output_buffer ); + + + +struct translate_sse { + struct translate translate; + + struct x86_function linear_func; + struct x86_function elt_func; + struct x86_function *func; + + boolean loaded_identity; + boolean loaded_255; + boolean loaded_inv_255; + + float identity[4]; + float float_255[4]; + float inv_255[4]; + + struct { + char *input_ptr; + unsigned input_stride; + } attrib[PIPE_MAX_ATTRIBS]; + + run_func gen_run; + run_elts_func gen_run_elts; + +}; + +static int get_offset( const void *a, const void *b ) +{ + return (const char *)b - (const char *)a; +} + + + +static struct x86_reg get_identity( struct translate_sse *p ) +{ + struct x86_reg reg = x86_make_reg(file_XMM, 6); + + if (!p->loaded_identity) { + /* Nasty: + */ + struct x86_reg translateESI = x86_make_reg(file_REG32, reg_SI); + + p->loaded_identity = TRUE; + p->identity[0] = 0; + p->identity[1] = 0; + p->identity[2] = 0; + p->identity[3] = 1; + + sse_movups(p->func, reg, + x86_make_disp(translateESI, + get_offset(p, &p->identity[0]))); + } + + return reg; +} + +static struct x86_reg get_255( struct translate_sse *p ) +{ + struct x86_reg reg = x86_make_reg(file_XMM, 6); + + if (!p->loaded_255) { + struct x86_reg translateESI = x86_make_reg(file_REG32, reg_SI); + + p->loaded_255 = TRUE; + p->float_255[0] = + p->float_255[1] = + p->float_255[2] = + p->float_255[3] = 255.0f; + + sse_movups(p->func, reg, + x86_make_disp(translateESI, + get_offset(p, &p->float_255[0]))); + } + + return reg; + return x86_make_reg(file_XMM, 7); +} + +static struct x86_reg get_inv_255( struct translate_sse *p ) +{ + struct x86_reg reg = x86_make_reg(file_XMM, 5); + + if (!p->loaded_inv_255) { + struct x86_reg translateESI = x86_make_reg(file_REG32, reg_SI); + + p->loaded_inv_255 = TRUE; + p->inv_255[0] = + p->inv_255[1] = + p->inv_255[2] = + p->inv_255[3] = 1.0 / 255.0f; + + sse_movups(p->func, reg, + x86_make_disp(translateESI, + get_offset(p, &p->inv_255[0]))); + } + + return reg; +} + + +static void emit_load_R32G32B32A32( struct translate_sse *p, + struct x86_reg data, + struct x86_reg arg0 ) +{ + sse_movups(p->func, data, arg0); +} + +static void emit_load_R32G32B32( struct translate_sse *p, + struct x86_reg data, + struct x86_reg arg0 ) +{ + /* Have to jump through some hoops: + * + * c 0 0 0 + * c 0 0 1 + * 0 0 c 1 + * a b c 1 + */ + sse_movss(p->func, data, x86_make_disp(arg0, 8)); + sse_shufps(p->func, data, get_identity(p), SHUF(X,Y,Z,W) ); + sse_shufps(p->func, data, data, SHUF(Y,Z,X,W) ); + sse_movlps(p->func, data, arg0); +} + +static void emit_load_R32G32( struct translate_sse *p, + struct x86_reg data, + struct x86_reg arg0 ) +{ + /* 0 0 0 1 + * a b 0 1 + */ + sse_movups(p->func, data, get_identity(p) ); + sse_movlps(p->func, data, arg0); +} + + +static void emit_load_R32( struct translate_sse *p, + struct x86_reg data, + struct x86_reg arg0 ) +{ + /* a 0 0 0 + * a 0 0 1 + */ + sse_movss(p->func, data, arg0); + sse_orps(p->func, data, get_identity(p) ); +} + + +static void emit_load_R8G8B8A8_UNORM( struct translate_sse *p, + struct x86_reg data, + struct x86_reg src ) +{ + + /* Load and unpack twice: + */ + sse_movss(p->func, data, src); + sse2_punpcklbw(p->func, src, get_identity(p)); + sse2_punpcklbw(p->func, src, get_identity(p)); + + /* Convert to float: + */ + sse2_cvtdq2ps(p->func, src, src); + + + /* Scale by 1/255.0 + */ + sse_mulps(p->func, src, get_inv_255(p)); +} + + + + +static void emit_store_R32G32B32A32( struct translate_sse *p, + struct x86_reg dest, + struct x86_reg dataXMM ) +{ + sse_movups(p->func, dest, dataXMM); +} + +static void emit_store_R32G32B32( struct translate_sse *p, + struct x86_reg dest, + struct x86_reg dataXMM ) +{ + /* Emit two, shuffle, emit one. + */ + sse_movlps(p->func, dest, dataXMM); + sse_shufps(p->func, dataXMM, dataXMM, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */ + sse_movss(p->func, x86_make_disp(dest,8), dataXMM); +} + +static void emit_store_R32G32( struct translate_sse *p, + struct x86_reg dest, + struct x86_reg dataXMM ) +{ + sse_movlps(p->func, dest, dataXMM); +} + +static void emit_store_R32( struct translate_sse *p, + struct x86_reg dest, + struct x86_reg dataXMM ) +{ + sse_movss(p->func, dest, dataXMM); +} + + + +static void emit_store_R8G8B8A8_UNORM( struct translate_sse *p, + struct x86_reg dest, + struct x86_reg dataXMM ) +{ + /* Scale by 255.0 + */ + sse_mulps(p->func, dataXMM, get_255(p)); + + /* Pack and emit: + */ + sse2_cvtps2dq(p->func, dataXMM, dataXMM); + sse2_packssdw(p->func, dataXMM, dataXMM); + sse2_packuswb(p->func, dataXMM, dataXMM); + sse_movss(p->func, dest, dataXMM); +} + + + + + +static void get_src_ptr( struct translate_sse *p, + struct x86_reg srcEAX, + struct x86_reg translateREG, + struct x86_reg eltREG, + unsigned a ) +{ + struct x86_reg input_ptr = + x86_make_disp(translateREG, + get_offset(p, &p->attrib[a].input_ptr)); + + struct x86_reg input_stride = + x86_make_disp(translateREG, + get_offset(p, &p->attrib[a].input_stride)); + + /* Calculate pointer to current attrib: + */ + x86_mov(p->func, srcEAX, input_stride); + x86_imul(p->func, srcEAX, eltREG); + x86_add(p->func, srcEAX, input_ptr); +} + + +/* Extended swizzles? Maybe later. + */ +static void emit_swizzle( struct translate_sse *p, + struct x86_reg dest, + struct x86_reg src, + unsigned shuffle ) +{ + sse_shufps(p->func, dest, src, shuffle); +} + + +static boolean translate_attr( struct translate_sse *p, + const struct translate_element *a, + struct x86_reg srcECX, + struct x86_reg dstEAX) +{ + struct x86_reg dataXMM = x86_make_reg(file_XMM, 0); + + switch (a->input_format) { + case PIPE_FORMAT_R32_FLOAT: + emit_load_R32(p, dataXMM, srcECX); + break; + case PIPE_FORMAT_R32G32_FLOAT: + emit_load_R32G32(p, dataXMM, srcECX); + break; + case PIPE_FORMAT_R32G32B32_FLOAT: + emit_load_R32G32B32(p, dataXMM, srcECX); + break; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + emit_load_R32G32B32A32(p, dataXMM, srcECX); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + emit_load_R8G8B8A8_UNORM(p, dataXMM, srcECX); + emit_swizzle(p, dataXMM, dataXMM, SHUF(Z,Y,X,W)); + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + emit_load_R8G8B8A8_UNORM(p, dataXMM, srcECX); + break; + default: + return FALSE; + } + + switch (a->output_format) { + case PIPE_FORMAT_R32_FLOAT: + emit_store_R32(p, dstEAX, dataXMM); + break; + case PIPE_FORMAT_R32G32_FLOAT: + emit_store_R32G32(p, dstEAX, dataXMM); + break; + case PIPE_FORMAT_R32G32B32_FLOAT: + emit_store_R32G32B32(p, dstEAX, dataXMM); + break; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + emit_store_R32G32B32A32(p, dstEAX, dataXMM); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + emit_swizzle(p, dataXMM, dataXMM, SHUF(Z,Y,X,W)); + emit_store_R8G8B8A8_UNORM(p, dstEAX, dataXMM); + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + emit_store_R8G8B8A8_UNORM(p, dstEAX, dataXMM); + break; + default: + return FALSE; + } + + return TRUE; +} + +/* Build run( struct translate *translate, + * unsigned start, + * unsigned count, + * void *output_buffer ) + * or + * run_elts( struct translate *translate, + * unsigned *elts, + * unsigned count, + * void *output_buffer ) + * + * Lots of hardcoding + * + * EAX -- pointer to current output vertex + * ECX -- pointer to current attribute + * + */ +static boolean build_vertex_emit( struct translate_sse *p, + struct x86_function *func, + boolean linear ) +{ + struct x86_reg vertexECX = x86_make_reg(file_REG32, reg_AX); + struct x86_reg idxEBX = x86_make_reg(file_REG32, reg_BX); + struct x86_reg srcEAX = x86_make_reg(file_REG32, reg_CX); + struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); + struct x86_reg translateESI = x86_make_reg(file_REG32, reg_SI); + uint8_t *fixup, *label; + unsigned j; + + p->func = func; + p->loaded_inv_255 = FALSE; + p->loaded_255 = FALSE; + p->loaded_identity = FALSE; + + x86_init_func(p->func); + + /* Push a few regs? + */ + x86_push(p->func, countEBP); + x86_push(p->func, translateESI); + x86_push(p->func, idxEBX); + + /* Get vertex count, compare to zero + */ + x86_xor(p->func, idxEBX, idxEBX); + x86_mov(p->func, countEBP, x86_fn_arg(p->func, 3)); + x86_cmp(p->func, countEBP, idxEBX); + fixup = x86_jcc_forward(p->func, cc_E); + + /* If linear, idx is the current element, otherwise it is a pointer + * to the current element. + */ + x86_mov(p->func, idxEBX, x86_fn_arg(p->func, 2)); + + /* Initialize destination register. + */ + x86_mov(p->func, vertexECX, x86_fn_arg(p->func, 4)); + + /* Move argument 1 (translate_sse pointer) into a reg: + */ + x86_mov(p->func, translateESI, x86_fn_arg(p->func, 1)); + + + /* always load, needed or not: + */ + + /* Note address for loop jump */ + label = x86_get_label(p->func); + + + for (j = 0; j < p->translate.key.nr_elements; j++) { + const struct translate_element *a = &p->translate.key.element[j]; + + struct x86_reg destEAX = x86_make_disp(vertexECX, + a->output_offset); + + /* Figure out source pointer address: + */ + if (linear) { + get_src_ptr(p, srcEAX, translateESI, idxEBX, j); + } + else { + get_src_ptr(p, srcEAX, translateESI, x86_deref(idxEBX), j); + } + + if (!translate_attr( p, a, x86_deref(srcEAX), destEAX )) + return FALSE; + } + + /* Next vertex: + */ + x86_lea(p->func, vertexECX, x86_make_disp(vertexECX, p->translate.key.output_stride)); + + /* Incr index + */ /* Emit code for each of the attributes. Currently routes + * everything through SSE registers, even when it might be more + * efficient to stick with regular old x86. No optimization or + * other tricks - enough new ground to cover here just getting + * things working. + */ + + if (linear) { + x86_inc(p->func, idxEBX); + } + else { + x86_lea(p->func, idxEBX, x86_make_disp(idxEBX, 4)); + } + + /* decr count, loop if not zero + */ + x86_dec(p->func, countEBP); + x86_test(p->func, countEBP, countEBP); + x86_jcc(p->func, cc_NZ, label); + + /* Exit mmx state? + */ + if (p->func->need_emms) + mmx_emms(p->func); + + /* Land forward jump here: + */ + x86_fixup_fwd_jump(p->func, fixup); + + /* Pop regs and return + */ + + x86_pop(p->func, idxEBX); + x86_pop(p->func, translateESI); + x86_pop(p->func, countEBP); + x86_ret(p->func); + + return TRUE; +} + + + + + + + +static void translate_sse_set_buffer( struct translate *translate, + unsigned buf, + const void *ptr, + unsigned stride ) +{ + struct translate_sse *p = (struct translate_sse *)translate; + unsigned i; + + for (i = 0; i < p->translate.key.nr_elements; i++) { + if (p->translate.key.element[i].input_buffer == buf) { + p->attrib[i].input_ptr = ((char *)ptr + + p->translate.key.element[i].input_offset); + p->attrib[i].input_stride = stride; + } + } +} + + +static void translate_sse_release( struct translate *translate ) +{ + struct translate_sse *p = (struct translate_sse *)translate; + + x86_release_func( &p->linear_func ); + x86_release_func( &p->elt_func ); + + FREE(p); +} + +static void translate_sse_run_elts( struct translate *translate, + const unsigned *elts, + unsigned count, + void *output_buffer ) +{ + struct translate_sse *p = (struct translate_sse *)translate; + + p->gen_run_elts( translate, + elts, + count, + output_buffer ); + +} + +static void translate_sse_run( struct translate *translate, + unsigned start, + unsigned count, + void *output_buffer ) +{ + struct translate_sse *p = (struct translate_sse *)translate; + + p->gen_run( translate, + start, + count, + output_buffer ); +} + + +struct translate *translate_sse2_create( const struct translate_key *key ) +{ + struct translate_sse *p = CALLOC_STRUCT( translate_sse ); + + if (p == NULL) + goto fail; + + if (!rtasm_cpu_has_sse() || !rtasm_cpu_has_sse2()) + goto fail; + + + p->translate.key = *key; + p->translate.release = translate_sse_release; + p->translate.set_buffer = translate_sse_set_buffer; + p->translate.run_elts = translate_sse_run_elts; + p->translate.run = translate_sse_run; + + if (!build_vertex_emit(p, &p->linear_func, TRUE)) + goto fail; + + if (!build_vertex_emit(p, &p->elt_func, FALSE)) + goto fail; + + p->gen_run = (run_func)x86_get_func(&p->linear_func); + p->gen_run_elts = (run_elts_func)x86_get_func(&p->elt_func); + + return &p->translate; + + fail: + if (p) + p->translate.release( &p->translate ); + + return NULL; +} + + + +#else + +void translate_create_sse( const struct translate_key *key ) +{ + return NULL; +} + +#endif -- cgit v1.2.3 From 26c27f6636069ca849a740c3969c577d841484e2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 18:42:41 +0100 Subject: draw: remove fetch_pipeline middle end -- just use the general path --- src/gallium/auxiliary/draw/Makefile | 1 - src/gallium/auxiliary/draw/SConscript | 1 - src/gallium/auxiliary/draw/draw_pt.c | 4 - src/gallium/auxiliary/draw/draw_pt.h | 26 +- .../auxiliary/draw/draw_pt_fetch_pipeline.c | 327 --------------------- 5 files changed, 15 insertions(+), 344 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 154c8a99b5..916e134719 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -18,7 +18,6 @@ C_SOURCES = \ draw_pt.c \ draw_pt_vcache.c \ draw_pt_fetch_emit.c \ - draw_pt_fetch_pipeline.c \ draw_pt_fetch_shade_pipeline.c \ draw_pt_fetch.c \ draw_pt_post_vs.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 0274d48e7f..6dc0195d93 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -18,7 +18,6 @@ draw = env.ConvenienceLibrary( 'draw_pt_emit.c', 'draw_pt_fetch.c', 'draw_pt_fetch_emit.c', - 'draw_pt_fetch_pipeline.c', 'draw_pt_fetch_shade_pipeline.c', 'draw_pt_pipeline.c', 'draw_pt_post_vs.c', diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 60a47f3911..b48d0cae9f 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -106,10 +106,6 @@ boolean draw_pt_init( struct draw_context *draw ) return FALSE; draw->pt.middle.opt[0] = draw_pt_fetch_emit( draw ); - draw->pt.middle.opt[PT_PIPELINE] = draw_pt_fetch_pipeline( draw ); -// draw->pt.middle.opt[PT_SHADE] = draw_pt_shade_emit( draw ); -// draw->pt.middle.opt[PT_SHADE | PT_PIPELINE] = draw_pt_shade_pipeline( draw ); -// draw->pt.middle.opt[PT_SHADE | PT_CLIPTEST] = draw_pt_shade_clip_either( draw ); draw->pt.middle.opt[PT_SHADE | PT_CLIPTEST | PT_PIPELINE] = draw_pt_fetch_pipeline_or_emit( draw ); diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 316289969b..eb23ee1c1e 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -83,11 +83,6 @@ struct draw_pt_front_end { * Currently two versions of this: * - fetch, vertex shade, cliptest, prim-pipeline * - fetch, emit (ie passthrough) - * Later: - * - fetch, vertex shade, cliptest, maybe-pipeline, maybe-emit - * - fetch, vertex shade, emit - * - * Currenly only using the passthrough version. */ struct draw_pt_middle_end { void (*prepare)( struct draw_pt_middle_end *, @@ -106,10 +101,6 @@ struct draw_pt_middle_end { /* The "back end" - supplied by the driver, defined in draw_vbuf.h. - * - * Not sure whether to wrap the prim pipeline up as an alternate - * backend. Would be a win for everything except pure passthrough - * mode... */ struct vbuf_render; struct vertex_header; @@ -121,11 +112,24 @@ pt_elt_func draw_pt_elt_func( struct draw_context *draw ); const void *draw_pt_elt_ptr( struct draw_context *draw, unsigned start ); -/* Implementations: +/* Frontends: + * + * Currently only the general-purpose vcache implementation, could add + * a special case for tiny vertex buffers. */ struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ); + +/* Middle-ends: + * + * Currently one general-purpose case which can do all possibilities, + * at the slight expense of creating a vertex_header in some cases + * unecessarily. + * + * The special case fetch_emit code avoids pipeline vertices + * altogether and builds hardware vertices directly from API + * vertex_elements. + */ struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); -struct draw_pt_middle_end *draw_pt_fetch_pipeline( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c deleted file mode 100644 index 26d0b37286..0000000000 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_pipeline.c +++ /dev/null @@ -1,327 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "draw/draw_context.h" -#include "draw/draw_vertex.h" -#include "draw/draw_pt.h" -#include "draw/draw_vs.h" - -/* The simplest 'middle end' in the new vertex code. - * - * The responsibilities of a middle end are to: - * - perform vertex fetch using - * - draw vertex element/buffer state - * - a list of fetch indices we received as an input - * - run the vertex shader - * - cliptest, - * - clip coord calculation - * - viewport transformation - * - if necessary, run the primitive pipeline, passing it: - * - a linear array of vertex_header vertices constructed here - * - a set of draw indices we received as an input - * - otherwise, drive the hw backend, - * - allocate space for hardware format vertices - * - translate the vertex-shader output vertices to hw format - * - calling the backend draw functions. - * - * For convenience, we provide a helper function to drive the hardware - * backend given similar inputs to those required to run the pipeline. - * - * In the case of passthrough mode, many of these actions are disabled - * or noops, so we end up doing: - * - * - perform vertex fetch - * - drive the hw backend - * - * IE, basically just vertex fetch to post-vs-format vertices, - * followed by a call to the backend helper function. - */ - - -struct fetch_pipeline_middle_end { - struct draw_pt_middle_end base; - struct draw_context *draw; - - void (*header)( unsigned idx, float **out); - - struct { - const ubyte *ptr; - unsigned pitch; - void (*fetch)( const void *from, float *attrib); - void (*emit)( const float *attrib, float **out ); - } fetch[PIPE_MAX_ATTRIBS]; - - unsigned nr_fetch; - unsigned pipeline_vertex_size; - unsigned prim; -}; - - -#if 0 -static void emit_R32_FLOAT( const float *attrib, - float **out ) -{ - (*out)[0] = attrib[0]; - (*out) += 1; -} - -static void emit_R32G32_FLOAT( const float *attrib, - float **out ) -{ - (*out)[0] = attrib[0]; - (*out)[1] = attrib[1]; - (*out) += 2; -} - -static void emit_R32G32B32_FLOAT( const float *attrib, - float **out ) -{ - (*out)[0] = attrib[0]; - (*out)[1] = attrib[1]; - (*out)[2] = attrib[2]; - (*out) += 3; -} -#endif -static void emit_R32G32B32A32_FLOAT( const float *attrib, - float **out ) -{ - (*out)[0] = attrib[0]; - (*out)[1] = attrib[1]; - (*out)[2] = attrib[2]; - (*out)[3] = attrib[3]; - (*out) += 4; -} - -static void header( unsigned idx, - float **out ) -{ - struct vertex_header *header = (struct vertex_header *) (*out); - - header->clipmask = 0; - header->edgeflag = 1; - header->pad = 0; - header->vertex_id = UNDEFINED_VERTEX_ID; - - (*out)[1] = 0; - (*out)[2] = 0; - (*out)[3] = 0; - (*out)[3] = 1; - (*out) += 5; -} - - -static void header_ef( unsigned idx, - float **out ) -{ - struct vertex_header *header = (struct vertex_header *) (*out); - - /* XXX: need a reset_stipple flag in the vertex header too? - */ - header->clipmask = 0; - header->edgeflag = (idx & DRAW_PT_EDGEFLAG) != 0; - header->pad = 0; - header->vertex_id = UNDEFINED_VERTEX_ID; - - (*out)[1] = 0; - (*out)[2] = 0; - (*out)[3] = 0; - (*out)[3] = 1; - (*out) += 5; -} - - -/** - * General-purpose fetch from user's vertex arrays, emit to driver's - * vertex buffer. - * - * XXX this is totally temporary. - */ -static void -fetch_store_general( struct fetch_pipeline_middle_end *fpme, - void *out_ptr, - const unsigned *fetch_elts, - unsigned count ) -{ - float *out = (float *)out_ptr; - uint i, j; - - for (i = 0; i < count; i++) { - unsigned elt = fetch_elts[i]; - - fpme->header( elt, &out ); - elt &= ~DRAW_PT_FLAG_MASK; - - for (j = 0; j < fpme->nr_fetch; j++) { - float attrib[4]; - const ubyte *from = (fpme->fetch[j].ptr + - fpme->fetch[j].pitch * elt); - - fpme->fetch[j].fetch( from, attrib ); - fpme->fetch[j].emit( attrib, &out ); - } - } -} - - -/* We aren't running a vertex shader, but are running the pipeline. - * That means the vertices we need to build look like: - * - * dw0: vertex header (zero?) - * dw1: clip coord 0 - * dw2: clip coord 1 - * dw3: clip coord 2 - * dw4: clip coord 4 - * dw5: screen coord 0 - * dw6: screen coord 0 - * dw7: screen coord 0 - * dw8: screen coord 0 - * dw9: other attribs... - * - */ -static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, - unsigned prim, - unsigned opt ) -{ - struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; - struct draw_context *draw = fpme->draw; - unsigned i, nr = 0; - - fpme->prim = prim; - - /* Emit the vertex header and empty clipspace coord field: - */ - if (draw->user.edgeflag) { - fpme->header = header_ef; - } - else { - fpme->header = header; - } - - - /* Need to look at vertex shader inputs (we know it is a - * passthrough shader, so these define the outputs too). If we - * were running a shader, we'd still be looking at the inputs at - * this point. - */ - for (i = 0; i < draw->vertex_shader->info.num_inputs; i++) { - unsigned buf = draw->vertex_element[i].vertex_buffer_index; - enum pipe_format format = draw->vertex_element[i].src_format; - - fpme->fetch[nr].ptr = ((const ubyte *) draw->user.vbuffer[buf] + - draw->vertex_buffer[buf].buffer_offset + - draw->vertex_element[i].src_offset); - - fpme->fetch[nr].pitch = draw->vertex_buffer[buf].pitch; - fpme->fetch[nr].fetch = draw_get_fetch_func( format ); - - /* Always do this -- somewhat redundant... - */ - fpme->fetch[nr].emit = emit_R32G32B32A32_FLOAT; - nr++; - } - - fpme->nr_fetch = nr; - fpme->pipeline_vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float); -} - - - - -static void fetch_pipeline_run( struct draw_pt_middle_end *middle, - const unsigned *fetch_elts, - unsigned fetch_count, - const ushort *draw_elts, - unsigned draw_count ) -{ - struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; - char *pipeline_verts; - - pipeline_verts = MALLOC( fpme->pipeline_vertex_size * - fetch_count ); - if (!pipeline_verts) { - assert(0); - return; - } - - - /* Single routine to fetch vertices and emit pipeline verts. - */ - fetch_store_general( fpme, - pipeline_verts, - fetch_elts, - fetch_count ); - - - /* Run the pipeline - */ - draw_pt_run_pipeline( fpme->draw, - fpme->prim, - (struct vertex_header *)pipeline_verts, - fetch_count, - fpme->pipeline_vertex_size, - draw_elts, - draw_count ); - - - /* Done -- that was easy, wasn't it: - */ - FREE( pipeline_verts ); -} - - - -static void fetch_pipeline_finish( struct draw_pt_middle_end *middle ) -{ - /* nothing to do */ -} - -static void fetch_pipeline_destroy( struct draw_pt_middle_end *middle ) -{ - FREE(middle); -} - - -struct draw_pt_middle_end *draw_pt_fetch_pipeline( struct draw_context *draw ) -{ - struct fetch_pipeline_middle_end *fetch_pipeline = CALLOC_STRUCT( fetch_pipeline_middle_end ); - - fetch_pipeline->base.prepare = fetch_pipeline_prepare; - fetch_pipeline->base.run = fetch_pipeline_run; - fetch_pipeline->base.finish = fetch_pipeline_finish; - fetch_pipeline->base.destroy = fetch_pipeline_destroy; - - fetch_pipeline->draw = draw; - - return &fetch_pipeline->base; -} - -- cgit v1.2.3 From 19218e2195f3dffc9403f16a742ba8c63edbf8b4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 18 Apr 2008 11:15:18 -0600 Subject: gallium: implement recip sqrt() with C code for now. Some conformance lighting tests fail with the SSE rsqrt instruction. --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 30 +++++++++++++++++++++++++++++ 1 file changed, 30 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 0a3a7559ca..6f785be3f5 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -36,6 +36,8 @@ #if defined(__i386__) || defined(__386__) +#define HIGH_PRECISION 1 /* for 1/sqrt() */ + #define DUMP_SSE 0 #if DUMP_SSE @@ -1137,16 +1139,44 @@ emit_rcp ( make_xmm( xmm_src ) ); } +#if HIGH_PRECISION +static void XSTDCALL +rsqrt4f( + float *store ) +{ +#ifdef WIN32 + store[0] = 1.0F / (float) sqrt( (double) store[0] ); + store[1] = 1.0F / (float) sqrt( (double) store[1] ); + store[2] = 1.0F / (float) sqrt( (double) store[2] ); + store[3] = 1.0F / (float) sqrt( (double) store[3] ); +#else + const unsigned X = TEMP_R0 * 16; + store[X + 0] = 1.0F / sqrt( store[X + 0] ); + store[X + 1] = 1.0F / sqrt( store[X + 1] ); + store[X + 2] = 1.0F / sqrt( store[X + 2] ); + store[X + 3] = 1.0F / sqrt( store[X + 3] ); +#endif +} +#endif + static void emit_rsqrt( struct x86_function *func, unsigned xmm_dst, unsigned xmm_src ) { +#if HIGH_PRECISION + emit_func_call_dst_src( + func, + xmm_dst, + xmm_src, + rsqrt4f ); +#else emit_rsqrtps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); +#endif } static void -- cgit v1.2.3 From e430d885e0d819172068805b1492cb6f10eb5d7f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 18 Apr 2008 11:15:53 -0600 Subject: gallium: a few comments --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 6f785be3f5..d47935e982 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -1598,20 +1598,25 @@ emit_instruction( STORE( func, *inst, 0, 0, CHAN_Y ); } if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + /* XMM[1] = SrcReg[0].yyyy */ FETCH( func, *inst, 1, 0, CHAN_Y ); + /* XMM[1] = max(XMM[1], 0) */ emit_maxps( func, make_xmm( 1 ), get_temp( TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ) ); + /* XMM[2] = SrcReg[0].wwww */ FETCH( func, *inst, 2, 0, CHAN_W ); + /* XMM[2] = min(XMM[2], 128.0) */ emit_minps( func, make_xmm( 2 ), get_temp( TGSI_EXEC_TEMP_128_I, TGSI_EXEC_TEMP_128_C ) ); + /* XMM[2] = max(XMM[2], -128.0) */ emit_maxps( func, make_xmm( 2 ), -- cgit v1.2.3 From 39c06c4336ee44ad2c0ff12705a384dede050bee Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 18 Apr 2008 11:33:42 -0600 Subject: gallium: get rid of unnecessary surface mapping --- src/mesa/state_tracker/st_cb_drawpixels.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 67f468c689..5fb27857a7 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -882,7 +882,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer); struct pipe_surface *psRead = rbRead->surface; struct pipe_surface *psDraw = rbDraw->surface; - ubyte *readMap, *drawMap; + ubyte *drawMap; ubyte *buffer; int i; @@ -892,14 +892,13 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, return; } - /* map the stencil buffers */ - readMap = pipe_surface_map(psRead); - drawMap = pipe_surface_map(psDraw); - /* this will do stencil pixel transfer ops */ st_read_stencil_pixels(ctx, srcx, srcy, width, height, GL_UNSIGNED_BYTE, &ctx->DefaultPacking, buffer); + /* map the stencil buffer */ + drawMap = pipe_surface_map(psDraw); + /* draw */ /* XXX PixelZoom not handled yet */ for (i = 0; i < height; i++) { @@ -937,8 +936,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, free(buffer); - /* unmap the stencil buffers */ - pipe_surface_unmap(psRead); + /* unmap the stencil buffer */ pipe_surface_unmap(psDraw); } -- cgit v1.2.3 From 7b34a43d1ab74974bd157a2339d7a491aed9c9b4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 18:51:43 +0100 Subject: translate: missing file --- src/gallium/auxiliary/translate/translate.c | 46 +++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+) create mode 100644 src/gallium/auxiliary/translate/translate.c (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate.c b/src/gallium/auxiliary/translate/translate.c new file mode 100644 index 0000000000..86ebeae2a7 --- /dev/null +++ b/src/gallium/auxiliary/translate/translate.c @@ -0,0 +1,46 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_state.h" +#include "translate.h" + +struct translate *translate_create( const struct translate_key *key ) +{ + struct translate *translate = NULL; + + translate = translate_sse2_create( key ); + if (translate) + return translate; + + return translate_generic_create( key ); +} -- cgit v1.2.3 From cb9f0a589623397c3437911aeef39f189213527a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 20:05:36 +0100 Subject: draw: remove draw_vf code, use translate instead --- src/gallium/auxiliary/draw/Makefile | 3 - src/gallium/auxiliary/draw/SConscript | 3 - src/gallium/auxiliary/draw/draw_pt_emit.c | 3 +- src/gallium/auxiliary/draw/draw_vbuf.c | 188 ++++---- src/gallium/auxiliary/draw/draw_vf.c | 378 ----------------- src/gallium/auxiliary/draw/draw_vf.h | 232 ---------- src/gallium/auxiliary/draw/draw_vf_generic.c | 585 ------------------------- src/gallium/auxiliary/draw/draw_vf_sse.c | 613 --------------------------- 8 files changed, 115 insertions(+), 1890 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_vf.c delete mode 100644 src/gallium/auxiliary/draw/draw_vf.h delete mode 100644 src/gallium/auxiliary/draw/draw_vf_generic.c delete mode 100644 src/gallium/auxiliary/draw/draw_vf_sse.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 916e134719..27464e5c86 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -35,9 +35,6 @@ C_SOURCES = \ draw_vertex_cache.c \ draw_vertex_fetch.c \ draw_vertex_shader.c \ - draw_vf.c \ - draw_vf_generic.c \ - draw_vf_sse.c \ draw_wide_line.c \ draw_wide_point.c diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 6dc0195d93..7af65c3c05 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -31,9 +31,6 @@ draw = env.ConvenienceLibrary( 'draw_vertex_cache.c', 'draw_vertex_fetch.c', 'draw_vertex_shader.c', - 'draw_vf.c', - 'draw_vf_generic.c', - 'draw_vf_sse.c', 'draw_vs_exec.c', 'draw_vs_llvm.c', 'draw_vs_sse.c', diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index a6a3ff6cac..490da4cca3 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -62,8 +62,7 @@ void draw_pt_emit_prepare( struct pt_emit *emit, vinfo = draw->render->get_vertex_info(draw->render); - /* In passthrough mode, need to translate from vertex shader - * outputs to hw vertices. + /* Translate from pipeline vertices to hw vertices. */ dst_offset = 0; for (i = 0; i < vinfo->num_attribs; i++) { diff --git a/src/gallium/auxiliary/draw/draw_vbuf.c b/src/gallium/auxiliary/draw/draw_vbuf.c index 59b63f5bfa..dbbcf05c3c 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_vbuf.c @@ -40,7 +40,7 @@ #include "draw_vbuf.h" #include "draw_private.h" #include "draw_vertex.h" -#include "draw_vf.h" +#include "translate/translate.h" /** @@ -56,7 +56,7 @@ struct vbuf_stage { /** Vertex size in bytes */ unsigned vertex_size; - struct draw_vertex_fetch *vf; + struct translate *translate; /* FIXME: we have no guarantee that 'unsigned' is 32bit */ @@ -71,8 +71,9 @@ struct vbuf_stage { unsigned max_indices; unsigned nr_indices; - /** Pipe primitive */ - unsigned prim; + /* Cache point size somewhere it's address won't change: + */ + float point_size; }; @@ -175,26 +176,25 @@ dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) * have a couple of slots at the beginning (1-dword header, 4-dword * clip pos) that we ignore here. We only use the vertex->data[] fields. */ -static INLINE void +static INLINE ushort emit_vertex( struct vbuf_stage *vbuf, struct vertex_header *vertex ) { - if(vertex->vertex_id != UNDEFINED_VERTEX_ID) { - if(vertex->vertex_id < vbuf->nr_vertices) - return; - else - debug_printf("Bad vertex id 0x%04x (>= 0x%04x)\n", - vertex->vertex_id, vbuf->nr_vertices); - return; - } + if(vertex->vertex_id == UNDEFINED_VERTEX_ID) { + /* Hmm - vertices are emitted one at a time - better make sure + * set_buffer is efficient. Consider a special one-shot mode for + * translate. + */ + vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0); + vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr); + + if (0) dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); - vertex->vertex_id = vbuf->nr_vertices++; - - draw_vf_emit_vertex(vbuf->vf, vertex, vbuf->vertex_ptr); + vbuf->vertex_ptr += vbuf->vertex_size/4; + vertex->vertex_id = vbuf->nr_vertices++; + } - if (0) dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); - - vbuf->vertex_ptr += vbuf->vertex_size/4; + return vertex->vertex_id; } @@ -208,9 +208,7 @@ vbuf_tri( struct draw_stage *stage, check_space( vbuf, 3 ); for (i = 0; i < 3; i++) { - emit_vertex( vbuf, prim->v[i] ); - - vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] ); } } @@ -225,9 +223,7 @@ vbuf_line( struct draw_stage *stage, check_space( vbuf, 2 ); for (i = 0; i < 2; i++) { - emit_vertex( vbuf, prim->v[i] ); - - vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[i]->vertex_id; + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] ); } } @@ -240,43 +236,112 @@ vbuf_point( struct draw_stage *stage, check_space( vbuf, 1 ); - emit_vertex( vbuf, prim->v[0] ); - - vbuf->indices[vbuf->nr_indices++] = (ushort) prim->v[0]->vertex_id; + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[0] ); } + + /** * Set the prim type for subsequent vertices. * This may result in a new vertex size. The existing vbuffer (if any) * will be flushed if needed and a new one allocated. */ static void -vbuf_set_prim( struct vbuf_stage *vbuf, uint newprim ) +vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) { - const struct vertex_info *vinfo; - unsigned vertex_size; - - assert(newprim == PIPE_PRIM_POINTS || - newprim == PIPE_PRIM_LINES || - newprim == PIPE_PRIM_TRIANGLES); + struct translate_key hw_key; + unsigned dst_offset; + unsigned i; - vbuf->prim = newprim; - vbuf->render->set_primitive(vbuf->render, newprim); + vbuf->render->set_primitive(vbuf->render, prim); - vinfo = vbuf->render->get_vertex_info(vbuf->render); - vertex_size = vinfo->size * sizeof(float); + /* Must do this after set_primitive() above: + * + * XXX: need some state managment to track when this needs to be + * recalculated. The driver should tell us whether there was a + * state change. + */ + vbuf->vinfo = vbuf->render->get_vertex_info(vbuf->render); - if (vertex_size != vbuf->vertex_size) + if (vbuf->vertex_size != vbuf->vinfo->size * sizeof(float)) { vbuf_flush_vertices(vbuf); + vbuf->vertex_size = vbuf->vinfo->size * sizeof(float); + } - vbuf->vinfo = vinfo; - vbuf->vertex_size = vertex_size; - if(vbuf->vf) - draw_vf_set_vertex_info(vbuf->vf, - vbuf->vinfo, - vbuf->stage.draw->rasterizer->point_size); - + /* Translate from pipeline vertices to hw vertices. + */ + dst_offset = 0; + memset(&hw_key, 0, sizeof(hw_key)); + + for (i = 0; i < vbuf->vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned src_buffer = 0; + unsigned output_format; + unsigned src_offset = (vbuf->vinfo->src_index[i] * 4 * sizeof(float) ); + + switch (vbuf->vinfo->emit[i]) { + case EMIT_4F: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + src_buffer = 1; + src_offset = 0; + break; + case EMIT_4UB: + output_format = PIPE_FORMAT_B8G8R8A8_UNORM; + emit_sz = 4 * sizeof(ubyte); + default: + assert(0); + output_format = PIPE_FORMAT_NONE; + emit_sz = 0; + break; + } + + hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + hw_key.element[i].input_buffer = src_buffer; + hw_key.element[i].input_offset = src_offset; + hw_key.element[i].output_format = output_format; + hw_key.element[i].output_offset = dst_offset; + + dst_offset += emit_sz; + } + + hw_key.nr_elements = vbuf->vinfo->num_attribs; + hw_key.output_stride = vbuf->vinfo->size * 4; + + /* Don't bother with caching at this stage: + */ + if (!vbuf->translate || + memcmp(&vbuf->translate->key, &hw_key, sizeof(hw_key)) != 0) + { + if (vbuf->translate) + vbuf->translate->release(vbuf->translate); + + vbuf->translate = translate_create( &hw_key ); + + vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0); + } + + vbuf->point_size = vbuf->stage.draw->rasterizer->point_size; + + /* Allocate new buffer? + */ if (!vbuf->vertices) vbuf_alloc_vertices(vbuf); } @@ -330,29 +395,9 @@ vbuf_flush_indices( struct vbuf_stage *vbuf ) assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); - switch(vbuf->prim) { - case PIPE_PRIM_POINTS: - break; - case PIPE_PRIM_LINES: - assert(vbuf->nr_indices % 2 == 0); - break; - case PIPE_PRIM_TRIANGLES: - assert(vbuf->nr_indices % 3 == 0); - break; - default: - assert(0); - } - vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); vbuf->nr_indices = 0; - - /* don't need to reset point/line/tri functions */ -#if 0 - stage->point = vbuf_first_point; - stage->line = vbuf_first_line; - stage->tri = vbuf_first_tri; -#endif } @@ -433,8 +478,8 @@ static void vbuf_destroy( struct draw_stage *stage ) if(vbuf->indices) align_free( vbuf->indices ); - if(vbuf->vf) - draw_vf_destroy( vbuf->vf ); + if(vbuf->translate) + vbuf->translate->release( vbuf->translate ); if (vbuf->render) vbuf->render->destroy( vbuf->render ); @@ -473,11 +518,6 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, vbuf->vertices = NULL; vbuf->vertex_ptr = vbuf->vertices; - - vbuf->prim = ~0; - vbuf->vf = draw_vf_create(); - if (!vbuf->vf) - goto fail; return &vbuf->stage; diff --git a/src/gallium/auxiliary/draw/draw_vf.c b/src/gallium/auxiliary/draw/draw_vf.c deleted file mode 100644 index 9d0154c50d..0000000000 --- a/src/gallium/auxiliary/draw/draw_vf.c +++ /dev/null @@ -1,378 +0,0 @@ -/* - * Copyright 2003 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: - * Keith Whitwell - */ - - -#include - -#include "pipe/p_compiler.h" -#include "pipe/p_util.h" -#include "rtasm/rtasm_execmem.h" - -#include "draw_vf.h" - - -#define DRAW_VF_DBG 0 - - -static boolean match_fastpath( struct draw_vertex_fetch *vf, - const struct draw_vf_fastpath *fp) -{ - unsigned j; - - if (vf->attr_count != fp->attr_count) - return FALSE; - - for (j = 0; j < vf->attr_count; j++) - if (vf->attr[j].format != fp->attr[j].format || - vf->attr[j].inputsize != fp->attr[j].size || - vf->attr[j].vertoffset != fp->attr[j].offset) - return FALSE; - - if (fp->match_strides) { - if (vf->vertex_stride != fp->vertex_stride) - return FALSE; - - for (j = 0; j < vf->attr_count; j++) - if (vf->attr[j].inputstride != fp->attr[j].stride) - return FALSE; - } - - return TRUE; -} - -static boolean search_fastpath_emit( struct draw_vertex_fetch *vf ) -{ - struct draw_vf_fastpath *fp = vf->fastpath; - - for ( ; fp ; fp = fp->next) { - if (match_fastpath(vf, fp)) { - vf->emit = fp->func; - return TRUE; - } - } - - return FALSE; -} - -void draw_vf_register_fastpath( struct draw_vertex_fetch *vf, - boolean match_strides ) -{ - struct draw_vf_fastpath *fastpath = CALLOC_STRUCT(draw_vf_fastpath); - unsigned i; - - fastpath->vertex_stride = vf->vertex_stride; - fastpath->attr_count = vf->attr_count; - fastpath->match_strides = match_strides; - fastpath->func = vf->emit; - fastpath->attr = (struct draw_vf_attr_type *) - MALLOC(vf->attr_count * sizeof(fastpath->attr[0])); - - for (i = 0; i < vf->attr_count; i++) { - fastpath->attr[i].format = vf->attr[i].format; - fastpath->attr[i].stride = vf->attr[i].inputstride; - fastpath->attr[i].size = vf->attr[i].inputsize; - fastpath->attr[i].offset = vf->attr[i].vertoffset; - } - - fastpath->next = vf->fastpath; - vf->fastpath = fastpath; -} - - - - -/*********************************************************************** - * Build codegen functions or return generic ones: - */ -static void choose_emit_func( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *dest) -{ - vf->emit = NULL; - - /* Does this match an existing (hardwired, codegen or known-bad) - * fastpath? - */ - if (search_fastpath_emit(vf)) { - /* Use this result. If it is null, then it is already known - * that the current state will fail for codegen and there is no - * point trying again. - */ - } - else if (vf->codegen_emit) { - vf->codegen_emit( vf ); - } - - if (!vf->emit) { - draw_vf_generate_hardwired_emit(vf); - } - - /* Otherwise use the generic version: - */ - if (!vf->emit) - vf->emit = draw_vf_generic_emit; - - vf->emit( vf, count, dest ); -} - - - - - -/*********************************************************************** - * Public entrypoints, mostly dispatch to the above: - */ - - - -static unsigned -draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, - const struct draw_vf_attr_map *map, - unsigned nr, - unsigned vertex_stride ) -{ - unsigned offset = 0; - unsigned i, j; - - assert(nr < PIPE_MAX_ATTRIBS); - - for (j = 0, i = 0; i < nr; i++) { - const unsigned format = map[i].format; - if (format == DRAW_EMIT_PAD) { -#if (DRAW_VF_DBG) - debug_printf("%d: pad %d, offset %d\n", i, - map[i].offset, offset); -#endif - - offset += map[i].offset; - - } - else { - vf->attr[j].attrib = map[i].attrib; - vf->attr[j].format = format; - vf->attr[j].insert = draw_vf_format_info[format].insert; - vf->attr[j].vertattrsize = draw_vf_format_info[format].attrsize; - vf->attr[j].vertoffset = offset; - vf->attr[j].isconst = draw_vf_format_info[format].isconst; - if(vf->attr[j].isconst) - memcpy(vf->attr[j].data, &map[i].data, vf->attr[j].vertattrsize); - -#if (DRAW_VF_DBG) - debug_printf("%d: %s, offset %d\n", i, - draw_vf_format_info[format].name, - vf->attr[j].vertoffset); -#endif - - offset += draw_vf_format_info[format].attrsize; - j++; - } - } - - vf->attr_count = j; - vf->vertex_stride = vertex_stride ? vertex_stride : offset; - vf->emit = choose_emit_func; - - assert(vf->vertex_stride >= offset); - return vf->vertex_stride; -} - - -void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, - const struct vertex_info *vinfo, - float point_size ) -{ - unsigned i, j; - struct draw_vf_attr *a = vf->attr; - struct draw_vf_attr_map attrs[PIPE_MAX_SHADER_INPUTS]; - unsigned count = 0; /* for debug/sanity */ - unsigned nr_attrs = 0; - - for (i = 0; i < vinfo->num_attribs; i++) { - j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; - case EMIT_1F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_1F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count++; - break; - case EMIT_1F_PSIZE: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_1F_CONST; - attrs[nr_attrs].offset = 0; - attrs[nr_attrs].data.f[0] = point_size; - nr_attrs++; - count++; - break; - case EMIT_2F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_2F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 2; - break; - case EMIT_3F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_3F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 3; - break; - case EMIT_4F: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_4F; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 4; - break; - case EMIT_4UB: - attrs[nr_attrs].attrib = j; - attrs[nr_attrs].format = DRAW_EMIT_4UB_4F_BGRA; - attrs[nr_attrs].offset = 0; - nr_attrs++; - count += 1; - break; - default: - assert(0); - } - } - - assert(count == vinfo->size); - - draw_vf_set_vertex_attributes(vf, - attrs, - nr_attrs, - vinfo->size * sizeof(float) ); - - for (j = 0; j < vf->attr_count; j++) { - a[j].inputsize = 4; - a[j].do_insert = a[j].insert[4 - 1]; - if(a[j].isconst) { - a[j].inputptr = a[j].data; - a[j].inputstride = 0; - } - } -} - - -#if 0 -/* Set attribute pointers, adjusted for start position: - */ -void draw_vf_set_sources( struct draw_vertex_fetch *vf, - GLvector4f * const sources[], - unsigned start ) -{ - struct draw_vf_attr *a = vf->attr; - unsigned j; - - for (j = 0; j < vf->attr_count; j++) { - const GLvector4f *vptr = sources[a[j].attrib]; - - if ((a[j].inputstride != vptr->stride) || - (a[j].inputsize != vptr->size)) - vf->emit = choose_emit_func; - - a[j].inputstride = vptr->stride; - a[j].inputsize = vptr->size; - a[j].do_insert = a[j].insert[vptr->size - 1]; - a[j].inputptr = ((uint8_t *)vptr->data) + start * vptr->stride; - } -} -#endif - - -/** - * Emit a vertex to dest. - */ -void draw_vf_emit_vertex( struct draw_vertex_fetch *vf, - struct vertex_header *vertex, - void *dest ) -{ - struct draw_vf_attr *a = vf->attr; - unsigned j; - - for (j = 0; j < vf->attr_count; j++) { - if (!a[j].isconst) { - a[j].inputptr = (uint8_t *)&vertex->data[a[j].attrib][0]; - a[j].inputstride = 0; /* XXX: one-vertex-max ATM */ - } - } - - vf->emit( vf, 1, (uint8_t*) dest ); -} - - - -struct draw_vertex_fetch *draw_vf_create( void ) -{ - struct draw_vertex_fetch *vf = CALLOC_STRUCT(draw_vertex_fetch); - unsigned i; - - for (i = 0; i < PIPE_MAX_ATTRIBS; i++) - vf->attr[i].vf = vf; - - vf->identity[0] = 0.0; - vf->identity[1] = 0.0; - vf->identity[2] = 0.0; - vf->identity[3] = 1.0; - - vf->codegen_emit = NULL; - -#ifdef USE_SSE_ASM - if (!GETENV("GALLIUM_NO_CODEGEN")) - vf->codegen_emit = draw_vf_generate_sse_emit; -#endif - - return vf; -} - - -void draw_vf_destroy( struct draw_vertex_fetch *vf ) -{ - struct draw_vf_fastpath *fp, *tmp; - - for (fp = vf->fastpath ; fp ; fp = tmp) { - tmp = fp->next; - FREE(fp->attr); - - /* KW: At the moment, fp->func is constrained to be allocated by - * rtasm_exec_alloc(), as the hardwired fastpaths in - * t_vertex_generic.c are handled specially. It would be nice - * to unify them, but this probably won't change until this - * module gets another overhaul. - */ - //rtasm_exec_free((void *) fp->func); - FREE(fp); - } - - vf->fastpath = NULL; - FREE(vf); -} diff --git a/src/gallium/auxiliary/draw/draw_vf.h b/src/gallium/auxiliary/draw/draw_vf.h deleted file mode 100644 index 0ef98d6257..0000000000 --- a/src/gallium/auxiliary/draw/draw_vf.h +++ /dev/null @@ -1,232 +0,0 @@ -/* - * Copyright 2008 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/** - * Vertex fetch/store/convert code. This functionality is used in two places: - * 1. Vertex fetch/convert - to grab vertex data from incoming vertex - * arrays and convert to format needed by vertex shaders. - * 2. Vertex store/emit - to convert simple float[][4] vertex attributes - * (which is the organization used throughout the draw/prim pipeline) to - * hardware-specific formats and emit into hardware vertex buffers. - * - * - * Authors: - * Keith Whitwell - */ - -#ifndef DRAW_VF_H -#define DRAW_VF_H - - -#include "pipe/p_compiler.h" -#include "pipe/p_state.h" - -#include "draw_vertex.h" -#include "draw_private.h" /* for vertex_header */ - - -enum draw_vf_attr_format { - DRAW_EMIT_1F, - DRAW_EMIT_2F, - DRAW_EMIT_3F, - DRAW_EMIT_4F, - DRAW_EMIT_3F_XYW, /**< for projective texture */ - DRAW_EMIT_1UB_1F, /**< for fog coordinate */ - DRAW_EMIT_3UB_3F_RGB, /**< for specular color */ - DRAW_EMIT_3UB_3F_BGR, /**< for specular color */ - DRAW_EMIT_4UB_4F_RGBA, /**< for color */ - DRAW_EMIT_4UB_4F_BGRA, /**< for color */ - DRAW_EMIT_4UB_4F_ARGB, /**< for color */ - DRAW_EMIT_4UB_4F_ABGR, /**< for color */ - DRAW_EMIT_1F_CONST, - DRAW_EMIT_2F_CONST, - DRAW_EMIT_3F_CONST, - DRAW_EMIT_4F_CONST, - DRAW_EMIT_PAD, /**< leave a hole of 'offset' bytes */ - DRAW_EMIT_MAX -}; - -struct draw_vf_attr_map -{ - /** Input attribute number */ - unsigned attrib; - - enum draw_vf_attr_format format; - - unsigned offset; - - /** - * Constant data for DRAW_EMIT_*_CONST - */ - union { - uint8_t ub[4]; - float f[4]; - } data; -}; - -struct draw_vertex_fetch; - - - -#if 0 -unsigned -draw_vf_set_vertex_attributes( struct draw_vertex_fetch *vf, - const struct draw_vf_attr_map *map, - unsigned nr, - unsigned vertex_stride ); -#endif - -void draw_vf_set_vertex_info( struct draw_vertex_fetch *vf, - const struct vertex_info *vinfo, - float point_size ); - -#if 0 -void -draw_vf_set_sources( struct draw_vertex_fetch *vf, - GLvector4f * const attrib[], - unsigned start ); -#endif - -void -draw_vf_emit_vertex( struct draw_vertex_fetch *vf, - struct vertex_header *vertex, - void *dest ); - -struct draw_vertex_fetch * -draw_vf_create( void ); - -void -draw_vf_destroy( struct draw_vertex_fetch *vf ); - - - -/*********************************************************************** - * Internal functions and structs: - */ - -struct draw_vf_attr; - - -typedef void (*draw_vf_insert_func)( const struct draw_vf_attr *a, - uint8_t *v, - const float *in ); - -typedef void (*draw_vf_emit_func)( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *dest ); - - - -/** - * Describes how to convert/move a vertex attribute from a vertex - * array to a vertex structure. - */ -struct draw_vf_attr -{ - struct draw_vertex_fetch *vf; - - unsigned format; - unsigned inputsize; - unsigned inputstride; - unsigned vertoffset; /**< position of the attrib in the vertex struct */ - - boolean isconst; /**< read from const data below */ - uint8_t data[16]; - - unsigned attrib; /**< which vertex attrib (0=position, etc) */ - unsigned vertattrsize; /**< size of the attribute in bytes */ - - uint8_t *inputptr; - const draw_vf_insert_func *insert; - draw_vf_insert_func do_insert; -}; - -struct draw_vertex_fetch -{ - struct draw_vf_attr attr[PIPE_MAX_ATTRIBS]; - unsigned attr_count; - unsigned vertex_stride; - - draw_vf_emit_func emit; - - /* Parameters and constants for codegen: - */ - float identity[4]; - - struct draw_vf_fastpath *fastpath; - - void (*codegen_emit)( struct draw_vertex_fetch *vf ); -}; - - -struct draw_vf_attr_type { - unsigned format; - unsigned size; - unsigned stride; - unsigned offset; -}; - -/** XXX this could be moved into draw_vf.c */ -struct draw_vf_fastpath { - unsigned vertex_stride; - unsigned attr_count; - boolean match_strides; - - struct draw_vf_attr_type *attr; - - draw_vf_emit_func func; - struct draw_vf_fastpath *next; -}; - - -void -draw_vf_register_fastpath( struct draw_vertex_fetch *vtx, - boolean match_strides ); - -void -draw_vf_generic_emit( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *v ); - -void -draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ); - -void -draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ); - - -/** XXX this type and function could probably be moved into draw_vf.c */ -struct draw_vf_format_info { - const char *name; - draw_vf_insert_func insert[4]; - const unsigned attrsize; - const boolean isconst; -}; - -extern const struct draw_vf_format_info -draw_vf_format_info[DRAW_EMIT_MAX]; - - -#endif diff --git a/src/gallium/auxiliary/draw/draw_vf_generic.c b/src/gallium/auxiliary/draw/draw_vf_generic.c deleted file mode 100644 index 7a60a9db9c..0000000000 --- a/src/gallium/auxiliary/draw/draw_vf_generic.c +++ /dev/null @@ -1,585 +0,0 @@ - -/* - * Copyright 2003 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: - * Keith Whitwell - */ - - -#include "pipe/p_compiler.h" -#include "pipe/p_debug.h" -#include "pipe/p_util.h" - -#include "draw_vf.h" - - - -static INLINE void insert_4f_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; - out[3] = in[3]; -} - -static INLINE void insert_4f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; - out[3] = 1; -} - -static INLINE void insert_4f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = 0; - out[3] = 1; -} - -static INLINE void insert_4f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = 0; - out[2] = 0; - out[3] = 1; -} - -static INLINE void insert_3f_xyw_4( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[3]; -} - -static INLINE void insert_3f_xyw_err( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - (void) a; (void) v; (void) in; - assert(0); -} - -static INLINE void insert_3f_3( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = in[2]; -} - -static INLINE void insert_3f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; - out[2] = 0; -} - -static INLINE void insert_3f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = 0; - out[2] = 0; -} - - -static INLINE void insert_2f_2( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = in[1]; -} - -static INLINE void insert_2f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; - out[1] = 0; -} - -static INLINE void insert_1f_1( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - float *out = (float *)(v); - (void) a; - - out[0] = in[0]; -} - -static INLINE void insert_null( const struct draw_vf_attr *a, uint8_t *v, const float *in ) -{ - (void) a; (void) v; (void) in; -} - -static INLINE void insert_4ub_4f_rgba_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); -} - -static INLINE void insert_4ub_4f_rgba_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_rgba_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[2] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_rgba_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - v[1] = 0; - v[2] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_bgra_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]); -} - -static INLINE void insert_4ub_4f_bgra_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_bgra_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[0] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_bgra_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - v[1] = 0; - v[0] = 0; - v[3] = 0xff; -} - -static INLINE void insert_4ub_4f_argb_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); -} - -static INLINE void insert_4ub_4f_argb_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]); - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_argb_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - v[3] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_argb_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]); - v[2] = 0x00; - v[3] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_abgr_4( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]); -} - -static INLINE void insert_4ub_4f_abgr_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]); - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_abgr_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]); - v[1] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_4ub_4f_abgr_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]); - v[2] = 0x00; - v[1] = 0x00; - v[0] = 0xff; -} - -static INLINE void insert_3ub_3f_rgb_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]); -} - -static INLINE void insert_3ub_3f_rgb_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[2] = 0; -} - -static INLINE void insert_3ub_3f_rgb_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); - v[1] = 0; - v[2] = 0; -} - -static INLINE void insert_3ub_3f_bgr_3( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]); -} - -static INLINE void insert_3ub_3f_bgr_2( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]); - v[0] = 0; -} - -static INLINE void insert_3ub_3f_bgr_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]); - v[1] = 0; - v[0] = 0; -} - - -static INLINE void insert_1ub_1f_1( const struct draw_vf_attr *a, uint8_t *v, - const float *in ) -{ - (void) a; - UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]); -} - - -const struct draw_vf_format_info draw_vf_format_info[DRAW_EMIT_MAX] = -{ - { "1f", - { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, - sizeof(float), FALSE }, - - { "2f", - { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, - 2 * sizeof(float), FALSE }, - - { "3f", - { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, - 3 * sizeof(float), FALSE }, - - { "4f", - { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, - 4 * sizeof(float), FALSE }, - - { "3f_xyw", - { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, - insert_3f_xyw_4 }, - 3 * sizeof(float), FALSE }, - - { "1ub_1f", - { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 }, - sizeof(uint8_t), FALSE }, - - { "3ub_3f_rgb", - { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3, - insert_3ub_3f_rgb_3 }, - 3 * sizeof(uint8_t), FALSE }, - - { "3ub_3f_bgr", - { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3, - insert_3ub_3f_bgr_3 }, - 3 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_rgba", - { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, - insert_4ub_4f_rgba_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_bgra", - { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3, - insert_4ub_4f_bgra_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_argb", - { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3, - insert_4ub_4f_argb_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "4ub_4f_abgr", - { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3, - insert_4ub_4f_abgr_4 }, - 4 * sizeof(uint8_t), FALSE }, - - { "1f_const", - { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 }, - sizeof(float), TRUE }, - - { "2f_const", - { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 }, - 2 * sizeof(float), TRUE }, - - { "3f_const", - { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 }, - 3 * sizeof(float), TRUE }, - - { "4f_const", - { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 }, - 4 * sizeof(float), TRUE }, - - { "pad", - { NULL, NULL, NULL, NULL }, - 0, FALSE }, - -}; - - - - -/*********************************************************************** - * Hardwired fastpaths for emitting whole vertices or groups of - * vertices - */ -#define EMIT5(NR, F0, F1, F2, F3, F4, NAME) \ -static void NAME( struct draw_vertex_fetch *vf, \ - unsigned count, \ - uint8_t *v ) \ -{ \ - struct draw_vf_attr *a = vf->attr; \ - unsigned i; \ - \ - for (i = 0 ; i < count ; i++, v += vf->vertex_stride) { \ - if (NR > 0) { \ - F0( &a[0], v + a[0].vertoffset, (float *)a[0].inputptr ); \ - a[0].inputptr += a[0].inputstride; \ - } \ - \ - if (NR > 1) { \ - F1( &a[1], v + a[1].vertoffset, (float *)a[1].inputptr ); \ - a[1].inputptr += a[1].inputstride; \ - } \ - \ - if (NR > 2) { \ - F2( &a[2], v + a[2].vertoffset, (float *)a[2].inputptr ); \ - a[2].inputptr += a[2].inputstride; \ - } \ - \ - if (NR > 3) { \ - F3( &a[3], v + a[3].vertoffset, (float *)a[3].inputptr ); \ - a[3].inputptr += a[3].inputstride; \ - } \ - \ - if (NR > 4) { \ - F4( &a[4], v + a[4].vertoffset, (float *)a[4].inputptr ); \ - a[4].inputptr += a[4].inputstride; \ - } \ - } \ -} - - -#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \ - insert_null, insert_null, NAME) - -#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \ - insert_null, NAME) - -#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \ - insert_null, NAME) - - -EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4) - -EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2) - -EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2) - - -/* Use the codegen paths to select one of a number of hardwired - * fastpaths. - */ -void draw_vf_generate_hardwired_emit( struct draw_vertex_fetch *vf ) -{ - draw_vf_emit_func func = NULL; - - /* Does it fit a hardwired fastpath? Help! this is growing out of - * control! - */ - switch (vf->attr_count) { - case 2: - if (vf->attr[0].do_insert == insert_3f_3 && - vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - func = emit_xyz3_rgba4; - } - break; - case 3: - if (vf->attr[2].do_insert == insert_2f_2) { - if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - if (vf->attr[0].do_insert == insert_4f_4) - func = emit_xyzw4_rgba4_st2; - } - } - break; - case 4: - if (vf->attr[2].do_insert == insert_2f_2 && - vf->attr[3].do_insert == insert_2f_2) { - if (vf->attr[1].do_insert == insert_4ub_4f_rgba_4) { - if (vf->attr[0].do_insert == insert_4f_4) - func = emit_xyzw4_rgba4_st2_st2; - } - } - break; - } - - vf->emit = func; -} - -/*********************************************************************** - * Generic (non-codegen) functions for whole vertices or groups of - * vertices - */ - -void draw_vf_generic_emit( struct draw_vertex_fetch *vf, - unsigned count, - uint8_t *v ) -{ - struct draw_vf_attr *a = vf->attr; - const unsigned attr_count = vf->attr_count; - const unsigned stride = vf->vertex_stride; - unsigned i, j; - - for (i = 0 ; i < count ; i++, v += stride) { - for (j = 0; j < attr_count; j++) { - float *in = (float *)a[j].inputptr; - a[j].inputptr += a[j].inputstride; - a[j].do_insert( &a[j], v + a[j].vertoffset, in ); - } - } -} - - diff --git a/src/gallium/auxiliary/draw/draw_vf_sse.c b/src/gallium/auxiliary/draw/draw_vf_sse.c deleted file mode 100644 index aff4ffd985..0000000000 --- a/src/gallium/auxiliary/draw/draw_vf_sse.c +++ /dev/null @@ -1,613 +0,0 @@ -/* - * Copyright 2003 Tungsten Graphics, inc. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: - * Keith Whitwell - */ - - -#include "pipe/p_compiler.h" -#include "util/u_simple_list.h" - -#include "draw_vf.h" - - -#if defined(USE_SSE_ASM) - -#include "rtasm/rtasm_cpu.h" -#include "rtasm/rtasm_x86sse.h" - - -#define X 0 -#define Y 1 -#define Z 2 -#define W 3 - - -struct x86_program { - struct x86_function func; - - struct draw_vertex_fetch *vf; - boolean inputs_safe; - boolean outputs_safe; - boolean have_sse2; - - struct x86_reg identity; - struct x86_reg chan0; -}; - - -static struct x86_reg get_identity( struct x86_program *p ) -{ - return p->identity; -} - -static void emit_load4f_4( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movups(&p->func, dest, arg0); -} - -static void emit_load4f_3( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Have to jump through some hoops: - * - * c 0 0 0 - * c 0 0 1 - * 0 0 c 1 - * a b c 1 - */ - sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); - sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); - sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) ); - sse_movlps(&p->func, dest, arg0); -} - -static void emit_load4f_2( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Initialize from identity, then pull in low two words: - */ - sse_movups(&p->func, dest, get_identity(p)); - sse_movlps(&p->func, dest, arg0); -} - -static void emit_load4f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Pull in low word, then swizzle in identity */ - sse_movss(&p->func, dest, arg0); - sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) ); -} - - - -static void emit_load3f_3( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - /* Over-reads by 1 dword - potential SEGV if input is a vertex - * array. - */ - if (p->inputs_safe) { - sse_movups(&p->func, dest, arg0); - } - else { - /* c 0 0 0 - * c c c c - * a b c c - */ - sse_movss(&p->func, dest, x86_make_disp(arg0, 8)); - sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X)); - sse_movlps(&p->func, dest, arg0); - } -} - -static void emit_load3f_2( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - emit_load4f_2(p, dest, arg0); -} - -static void emit_load3f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - emit_load4f_1(p, dest, arg0); -} - -static void emit_load2f_2( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movlps(&p->func, dest, arg0); -} - -static void emit_load2f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - emit_load4f_1(p, dest, arg0); -} - -static void emit_load1f_1( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movss(&p->func, dest, arg0); -} - -static void (*load[4][4])( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) = { - { emit_load1f_1, - emit_load1f_1, - emit_load1f_1, - emit_load1f_1 }, - - { emit_load2f_1, - emit_load2f_2, - emit_load2f_2, - emit_load2f_2 }, - - { emit_load3f_1, - emit_load3f_2, - emit_load3f_3, - emit_load3f_3 }, - - { emit_load4f_1, - emit_load4f_2, - emit_load4f_3, - emit_load4f_4 } -}; - -static void emit_load( struct x86_program *p, - struct x86_reg dest, - unsigned sz, - struct x86_reg src, - unsigned src_sz) -{ - load[sz-1][src_sz-1](p, dest, src); -} - -static void emit_store4f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movups(&p->func, dest, arg0); -} - -static void emit_store3f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - if (p->outputs_safe) { - /* Emit the extra dword anyway. This may hurt writecombining, - * may cause other problems. - */ - sse_movups(&p->func, dest, arg0); - } - else { - /* Alternate strategy - emit two, shuffle, emit one. - */ - sse_movlps(&p->func, dest, arg0); - sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */ - sse_movss(&p->func, x86_make_disp(dest,8), arg0); - } -} - -static void emit_store2f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movlps(&p->func, dest, arg0); -} - -static void emit_store1f( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) -{ - sse_movss(&p->func, dest, arg0); -} - - -static void (*store[4])( struct x86_program *p, - struct x86_reg dest, - struct x86_reg arg0 ) = -{ - emit_store1f, - emit_store2f, - emit_store3f, - emit_store4f -}; - -static void emit_store( struct x86_program *p, - struct x86_reg dest, - unsigned sz, - struct x86_reg temp ) - -{ - store[sz-1](p, dest, temp); -} - -static void emit_pack_store_4ub( struct x86_program *p, - struct x86_reg dest, - struct x86_reg temp ) -{ - /* Scale by 255.0 - */ - sse_mulps(&p->func, temp, p->chan0); - - if (p->have_sse2) { - sse2_cvtps2dq(&p->func, temp, temp); - sse2_packssdw(&p->func, temp, temp); - sse2_packuswb(&p->func, temp, temp); - sse_movss(&p->func, dest, temp); - } - else { - struct x86_reg mmx0 = x86_make_reg(file_MMX, 0); - struct x86_reg mmx1 = x86_make_reg(file_MMX, 1); - sse_cvtps2pi(&p->func, mmx0, temp); - sse_movhlps(&p->func, temp, temp); - sse_cvtps2pi(&p->func, mmx1, temp); - mmx_packssdw(&p->func, mmx0, mmx1); - mmx_packuswb(&p->func, mmx0, mmx0); - mmx_movd(&p->func, dest, mmx0); - } -} - -static int get_offset( const void *a, const void *b ) -{ - return (const char *)b - (const char *)a; -} - -/* Not much happens here. Eventually use this function to try and - * avoid saving/reloading the source pointers each vertex (if some of - * them can fit in registers). - */ -static void get_src_ptr( struct x86_program *p, - struct x86_reg srcREG, - struct x86_reg vfREG, - struct draw_vf_attr *a ) -{ - struct draw_vertex_fetch *vf = p->vf; - struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); - - /* Load current a[j].inputptr - */ - x86_mov(&p->func, srcREG, ptr_to_src); -} - -static void update_src_ptr( struct x86_program *p, - struct x86_reg srcREG, - struct x86_reg vfREG, - struct draw_vf_attr *a ) -{ - if (a->inputstride) { - struct draw_vertex_fetch *vf = p->vf; - struct x86_reg ptr_to_src = x86_make_disp(vfREG, get_offset(vf, &a->inputptr)); - - /* add a[j].inputstride (hardcoded value - could just as easily - * pull the stride value from memory each time). - */ - x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride)); - - /* save new value of a[j].inputptr - */ - x86_mov(&p->func, ptr_to_src, srcREG); - } -} - - -/* Lots of hardcoding - * - * EAX -- pointer to current output vertex - * ECX -- pointer to current attribute - * - */ -static boolean build_vertex_emit( struct x86_program *p ) -{ - struct draw_vertex_fetch *vf = p->vf; - unsigned j = 0; - - struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX); - struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX); - struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); - struct x86_reg vfESI = x86_make_reg(file_REG32, reg_SI); - struct x86_reg temp = x86_make_reg(file_XMM, 0); - uint8_t *fixup, *label; - - /* Push a few regs? - */ - x86_push(&p->func, countEBP); - x86_push(&p->func, vfESI); - - - /* Get vertex count, compare to zero - */ - x86_xor(&p->func, srcECX, srcECX); - x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2)); - x86_cmp(&p->func, countEBP, srcECX); - fixup = x86_jcc_forward(&p->func, cc_E); - - /* Initialize destination register. - */ - x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3)); - - /* Move argument 1 (vf) into a reg: - */ - x86_mov(&p->func, vfESI, x86_fn_arg(&p->func, 1)); - - - /* always load, needed or not: - */ - sse_movups(&p->func, p->identity, x86_make_disp(vfESI, get_offset(vf, &vf->identity[0]))); - - /* Note address for loop jump */ - label = x86_get_label(&p->func); - - /* Emit code for each of the attributes. Currently routes - * everything through SSE registers, even when it might be more - * efficient to stick with regular old x86. No optimization or - * other tricks - enough new ground to cover here just getting - * things working. - */ - while (j < vf->attr_count) { - struct draw_vf_attr *a = &vf->attr[j]; - struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset); - - /* Now, load an XMM reg from src, perhaps transform, then save. - * Could be shortcircuited in specific cases: - */ - switch (a->format) { - case DRAW_EMIT_1F: - case DRAW_EMIT_1F_CONST: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 1, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_2F: - case DRAW_EMIT_2F_CONST: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 2, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_3F: - case DRAW_EMIT_3F_CONST: - /* Potentially the worst case - hardcode 2+1 copying: - */ - if (0) { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 3, temp); - update_src_ptr(p, srcECX, vfESI, a); - } - else { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 2, temp); - if (a->inputsize > 2) { - emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1); - emit_store(p, x86_make_disp(dest,8), 1, temp); - } - else { - sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p)); - } - update_src_ptr(p, srcECX, vfESI, a); - } - break; - case DRAW_EMIT_4F: - case DRAW_EMIT_4F_CONST: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - emit_store(p, dest, 4, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_3F_XYW: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z)); - emit_store(p, dest, 3, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - - case DRAW_EMIT_1UB_1F: - /* Test for PAD3 + 1UB: - */ - if (j > 0 && - a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3) - { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X)); - emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */ - update_src_ptr(p, srcECX, vfESI, a); - } - else { - debug_printf("Can't emit 1ub %x %x %d\n", - a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize ); - return FALSE; - } - break; - case DRAW_EMIT_3UB_3F_RGB: - case DRAW_EMIT_3UB_3F_BGR: - /* Test for 3UB + PAD1: - */ - if (j == vf->attr_count - 1 || - a[1].vertoffset >= a->vertoffset + 4) { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - if (a->format == DRAW_EMIT_3UB_3F_BGR) - sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - } - /* Test for 3UB + 1UB: - */ - else if (j < vf->attr_count - 1 && - a[1].format == DRAW_EMIT_1UB_1F && - a[1].vertoffset == a->vertoffset + 3) { - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize); - update_src_ptr(p, srcECX, vfESI, a); - - /* Make room for incoming value: - */ - sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); - - get_src_ptr(p, srcECX, vfESI, &a[1]); - emit_load(p, temp, 1, x86_deref(srcECX), a[1].inputsize); - update_src_ptr(p, srcECX, vfESI, &a[1]); - - /* Rearrange and possibly do BGR conversion: - */ - if (a->format == DRAW_EMIT_3UB_3F_BGR) - sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); - else - sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X)); - - emit_pack_store_4ub(p, dest, temp); - j++; /* NOTE: two attrs consumed */ - } - else { - debug_printf("Can't emit 3ub\n"); - } - return FALSE; /* add this later */ - break; - - case DRAW_EMIT_4UB_4F_RGBA: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4UB_4F_BGRA: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4UB_4F_ARGB: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - case DRAW_EMIT_4UB_4F_ABGR: - get_src_ptr(p, srcECX, vfESI, a); - emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize); - sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X)); - emit_pack_store_4ub(p, dest, temp); - update_src_ptr(p, srcECX, vfESI, a); - break; - default: - debug_printf("unknown a[%d].format %d\n", j, a->format); - return FALSE; /* catch any new opcodes */ - } - - /* Increment j by at least 1 - may have been incremented above also: - */ - j++; - } - - /* Next vertex: - */ - x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vf->vertex_stride)); - - /* decr count, loop if not zero - */ - x86_dec(&p->func, countEBP); - x86_test(&p->func, countEBP, countEBP); - x86_jcc(&p->func, cc_NZ, label); - - /* Exit mmx state? - */ - if (p->func.need_emms) - mmx_emms(&p->func); - - /* Land forward jump here: - */ - x86_fixup_fwd_jump(&p->func, fixup); - - /* Pop regs and return - */ - x86_pop(&p->func, x86_get_base_reg(vfESI)); - x86_pop(&p->func, countEBP); - x86_ret(&p->func); - - vf->emit = (draw_vf_emit_func)x86_get_func(&p->func); - return TRUE; -} - - - -void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) -{ - struct x86_program p; - - if (!rtasm_cpu_has_sse()) { - vf->codegen_emit = NULL; - return; - } - - memset(&p, 0, sizeof(p)); - - p.vf = vf; - p.inputs_safe = 0; /* for now */ - p.outputs_safe = 1; /* for now */ - p.have_sse2 = rtasm_cpu_has_sse2(); - p.identity = x86_make_reg(file_XMM, 6); - p.chan0 = x86_make_reg(file_XMM, 7); - - x86_init_func(&p.func); - - if (build_vertex_emit(&p)) { - draw_vf_register_fastpath( vf, TRUE ); - } - else { - /* Note the failure so that we don't keep trying to codegen an - * impossible state: - */ - draw_vf_register_fastpath( vf, FALSE ); - x86_release_func(&p.func); - } -} - -#else - -void draw_vf_generate_sse_emit( struct draw_vertex_fetch *vf ) -{ - /* Dummy version for when USE_SSE_ASM not defined */ -} - -#endif -- cgit v1.2.3 From a41c05b20a36d2160aa232d08ed57d3095438025 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 20:11:16 +0100 Subject: draw: switch over to draw_pt paths, will remove old code shortly --- src/gallium/auxiliary/draw/draw_context.c | 2 -- src/gallium/auxiliary/draw/draw_private.h | 1 - src/gallium/auxiliary/draw/draw_pt.c | 3 --- 3 files changed, 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 4838b68ed1..5dc3358bd1 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -49,8 +49,6 @@ struct draw_context *draw_create( void ) draw->use_sse = FALSE; #endif - draw->use_pt_shaders = GETENV( "GALLIUM_PT_SHADERS" ) != NULL; - /* create pipeline stages */ draw->pipeline.wide_line = draw_wide_line_stage( draw ); draw->pipeline.wide_point = draw_wide_point_stage( draw ); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index da94e69781..18ce6c0ec5 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -258,7 +258,6 @@ struct draw_context boolean line_stipple; /**< do line stipple? */ boolean point_sprite; /**< convert points to quads for sprites? */ boolean use_sse; - boolean use_pt_shaders; /* temporary flag to switch on pt shader paths */ /* If a prim stage introduces new vertex attributes, they'll be stored here */ diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index b48d0cae9f..941f6efbe4 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -68,9 +68,6 @@ draw_pt_arrays(struct draw_context *draw, if (!draw->rasterizer->bypass_vs) { opt |= PT_SHADE; - - if (!draw->use_pt_shaders) - return FALSE; } -- cgit v1.2.3 From b11d89dc6d230f7f945f9eb420d39921c648ec20 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 20:36:38 +0100 Subject: draw: remove draw_prim.c --- src/gallium/auxiliary/draw/Makefile | 1 - src/gallium/auxiliary/draw/SConscript | 1 - src/gallium/auxiliary/draw/draw_prim.c | 23 -------------- src/gallium/auxiliary/draw/draw_pt.c | 55 ++++++++++++++++++++++++++++++++++ 4 files changed, 55 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 27464e5c86..080311a667 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -24,7 +24,6 @@ C_SOURCES = \ draw_pt_emit.c \ draw_pt_pipeline.c \ draw_pt_elts.c \ - draw_prim.c \ draw_pstipple.c \ draw_stipple.c \ draw_twoside.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 7af65c3c05..238e3f7d28 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -11,7 +11,6 @@ draw = env.ConvenienceLibrary( 'draw_debug.c', 'draw_flatshade.c', 'draw_offset.c', - 'draw_prim.c', 'draw_pstipple.c', 'draw_pt.c', 'draw_pt_elts.c', diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c index 51b6950334..d61cd25243 100644 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ b/src/gallium/auxiliary/draw/draw_prim.c @@ -497,27 +497,4 @@ draw_prim( struct draw_context *draw, -/** - * Draw vertex arrays - * This is the main entrypoint into the drawing module. - * \param prim one of PIPE_PRIM_x - * \param start index of first vertex to draw - * \param count number of vertices to draw - */ -void -draw_arrays(struct draw_context *draw, unsigned prim, - unsigned start, unsigned count) -{ - if (reduced_prim[prim] != draw->reduced_prim) { - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->reduced_prim = reduced_prim[prim]; - } - - /* drawing done here: */ - if (!draw_pt_arrays(draw, prim, start, count)) { - /* we have to run the whole pipeline */ - draw_prim(draw, prim, start, count); - } -} - diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 941f6efbe4..ecaed84070 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -128,3 +128,58 @@ void draw_pt_destroy( struct draw_context *draw ) draw->pt.front.vcache = NULL; } } + + + +static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES +}; + + +/** + * Draw vertex arrays + * This is the main entrypoint into the drawing module. + * \param prim one of PIPE_PRIM_x + * \param start index of first vertex to draw + * \param count number of vertices to draw + */ +void +draw_arrays(struct draw_context *draw, unsigned prim, + unsigned start, unsigned count) +{ + if (reduced_prim[prim] != draw->reduced_prim) { + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + draw->reduced_prim = reduced_prim[prim]; + } + + /* drawing done here: */ + draw_pt_arrays(draw, prim, start, count); +} + + +/* Revamp me please: + */ +void draw_do_flush( struct draw_context *draw, unsigned flags ) +{ + if (!draw->flushing) + { + draw->flushing = TRUE; + + if (flags >= DRAW_FLUSH_STATE_CHANGE) { + draw->pipeline.first->flush( draw->pipeline.first, flags ); + draw->pipeline.first = draw->pipeline.validate; + draw->reduced_prim = ~0; + } + + draw->flushing = FALSE; + } +} -- cgit v1.2.3 From 66891826421d5b774e081f7a2a85580cd0523fab Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 20:39:13 +0100 Subject: draw: remove draw_vertex_cache.c --- src/gallium/auxiliary/draw/Makefile | 1 - src/gallium/auxiliary/draw/draw_context.c | 20 +- src/gallium/auxiliary/draw/draw_prim.c | 500 ------------------------- src/gallium/auxiliary/draw/draw_vertex_cache.c | 219 ----------- 4 files changed, 18 insertions(+), 722 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_prim.c delete mode 100644 src/gallium/auxiliary/draw/draw_vertex_cache.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 080311a667..03210f6a4a 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -31,7 +31,6 @@ C_SOURCES = \ draw_validate.c \ draw_vbuf.c \ draw_vertex.c \ - draw_vertex_cache.c \ draw_vertex_fetch.c \ draw_vertex_shader.c \ draw_wide_line.c \ diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 5dc3358bd1..c611300841 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -103,7 +103,6 @@ struct draw_context *draw_create( void ) draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ - draw_vertex_cache_invalidate( draw ); draw_set_mapped_element_buffer( draw, 0, NULL ); tgsi_exec_machine_init(&draw->machine); @@ -446,7 +445,6 @@ void draw_reset_vertex_ids(struct draw_context *draw) stage = stage->next; } - draw_vertex_cache_reset_vertex_ids(draw); /* going away soon */ draw_pt_reset_vertex_ids(draw); } @@ -473,3 +471,21 @@ boolean draw_get_edgeflag( struct draw_context *draw, return 1; } + +/** + * Tell the drawing context about the index/element buffer to use + * (ala glDrawElements) + * If no element buffer is to be used (i.e. glDrawArrays) then this + * should be called with eltSize=0 and elements=NULL. + * + * \param draw the drawing context + * \param eltSize size of each element (1, 2 or 4 bytes) + * \param elements the element buffer ptr + */ +void +draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, void *elements ) +{ + draw->user.elts = elements; + draw->user.eltSize = eltSize; +} diff --git a/src/gallium/auxiliary/draw/draw_prim.c b/src/gallium/auxiliary/draw/draw_prim.c deleted file mode 100644 index d61cd25243..0000000000 --- a/src/gallium/auxiliary/draw/draw_prim.c +++ /dev/null @@ -1,500 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" - -#include "draw_private.h" -#include "draw_context.h" - - - -#define RP_NONE 0 -#define RP_POINT 1 -#define RP_LINE 2 -#define RP_TRI 3 - - -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { - RP_POINT, - RP_LINE, - RP_LINE, - RP_LINE, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI, - RP_TRI -}; - - -static void draw_prim_queue_flush( struct draw_context *draw ) -{ - unsigned i; - - if (0) - debug_printf("Flushing with %d prims, %d verts\n", - draw->pq.queue_nr, draw->vs.queue_nr); - - assert (draw->pq.queue_nr != 0); - - /* NOTE: we cannot save draw->pipeline->first in a local var because - * draw->pipeline->first is often changed by the first call to tri(), - * line(), etc. - */ - if (draw->rasterizer->line_stipple_enable) { - switch (draw->reduced_prim) { - case RP_TRI: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - - draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); - } - break; - case RP_LINE: - for (i = 0; i < draw->pq.queue_nr; i++) { - if (draw->pq.queue[i].reset_line_stipple) - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - - draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); - } - break; - case RP_POINT: - draw->pipeline.first->reset_stipple_counter( draw->pipeline.first ); - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); - break; - } - } - else { - switch (draw->reduced_prim) { - case RP_TRI: - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->tri( draw->pipeline.first, &draw->pq.queue[i] ); - break; - case RP_LINE: - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->line( draw->pipeline.first, &draw->pq.queue[i] ); - break; - case RP_POINT: - for (i = 0; i < draw->pq.queue_nr; i++) - draw->pipeline.first->point( draw->pipeline.first, &draw->pq.queue[i] ); - break; - } - } - - draw->pq.queue_nr = 0; - draw->vs.post_nr = 0; - draw_vertex_cache_unreference( draw ); -} - -void draw_do_flush( struct draw_context *draw, unsigned flags ) -{ - if (0) - debug_printf("Flushing with %d verts, %d prims\n", - draw->vs.queue_nr, - draw->pq.queue_nr ); - - if (draw->flushing) - return; - - draw->flushing = TRUE; - - if (flags >= DRAW_FLUSH_SHADER_QUEUE) { - if (draw->vs.queue_nr) { - (*draw->shader_queue_flush)(draw); - } - - if (flags >= DRAW_FLUSH_PRIM_QUEUE) { - if (draw->pq.queue_nr) - draw_prim_queue_flush(draw); - - if (flags >= DRAW_FLUSH_VERTEX_CACHE) { - draw_vertex_cache_invalidate(draw); - - if (flags >= DRAW_FLUSH_STATE_CHANGE) { - draw->pipeline.first->flush( draw->pipeline.first, flags ); - draw->pipeline.first = draw->pipeline.validate; - draw->reduced_prim = ~0; - } - } - } - } - - draw->flushing = FALSE; -} - - - -/* Return a pointer to a freshly queued primitive header. Ensure that - * there is room in the vertex cache for a maximum of "nr_verts" new - * vertices. Flush primitive and/or vertex queues if necessary to - * make space. - */ -static struct prim_header *get_queued_prim( struct draw_context *draw, - unsigned nr_verts ) -{ - if (!draw_vertex_cache_check_space( draw, nr_verts )) { -// debug_printf("v"); - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE ); - } - else if (draw->pq.queue_nr == PRIM_QUEUE_LENGTH) { -// debug_printf("p"); - draw_do_flush( draw, DRAW_FLUSH_PRIM_QUEUE ); - } - - assert(draw->pq.queue_nr < PRIM_QUEUE_LENGTH); - - return &draw->pq.queue[draw->pq.queue_nr++]; -} - - - -/** - * Add a point to the primitive queue. - * \param i0 index into user's vertex arrays - */ -static void do_point( struct draw_context *draw, - unsigned i0 ) -{ - struct prim_header *prim = get_queued_prim( draw, 1 ); - - prim->reset_line_stipple = 0; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->vcache.get_vertex( draw, i0 ); -} - - -/** - * Add a line to the primitive queue. - * \param i0 index into user's vertex arrays - * \param i1 index into user's vertex arrays - */ -static void do_line( struct draw_context *draw, - boolean reset_stipple, - unsigned i0, - unsigned i1 ) -{ - struct prim_header *prim = get_queued_prim( draw, 2 ); - - prim->reset_line_stipple = reset_stipple; - prim->edgeflags = 1; - prim->pad = 0; - prim->v[0] = draw->vcache.get_vertex( draw, i0 ); - prim->v[1] = draw->vcache.get_vertex( draw, i1 ); -} - -/** - * Add a triangle to the primitive queue. - */ -static void do_triangle( struct draw_context *draw, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - -// _mesa_printf("tri %d %d %d\n", i0, i1, i2); - prim->reset_line_stipple = 1; - prim->edgeflags = ~0; - prim->pad = 0; - prim->v[0] = draw->vcache.get_vertex( draw, i0 ); - prim->v[1] = draw->vcache.get_vertex( draw, i1 ); - prim->v[2] = draw->vcache.get_vertex( draw, i2 ); -} - -static void do_ef_triangle( struct draw_context *draw, - boolean reset_stipple, - unsigned ef_mask, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - struct prim_header *prim = get_queued_prim( draw, 3 ); - struct vertex_header *v0 = draw->vcache.get_vertex( draw, i0 ); - struct vertex_header *v1 = draw->vcache.get_vertex( draw, i1 ); - struct vertex_header *v2 = draw->vcache.get_vertex( draw, i2 ); - - prim->reset_line_stipple = reset_stipple; - - prim->edgeflags = ef_mask & ((v0->edgeflag << 0) | - (v1->edgeflag << 1) | - (v2->edgeflag << 2)); - prim->pad = 0; - prim->v[0] = v0; - prim->v[1] = v1; - prim->v[2] = v2; -} - - -static void do_ef_quad( struct draw_context *draw, - unsigned v0, - unsigned v1, - unsigned v2, - unsigned v3 ) -{ - const unsigned omitEdge2 = ~(1 << 1); - const unsigned omitEdge3 = ~(1 << 2); - do_ef_triangle( draw, 1, omitEdge2, v0, v1, v3 ); - do_ef_triangle( draw, 0, omitEdge3, v1, v2, v3 ); -} - -static void do_quad( struct draw_context *draw, - unsigned v0, - unsigned v1, - unsigned v2, - unsigned v3 ) -{ - do_triangle( draw, v0, v1, v3 ); - do_triangle( draw, v1, v2, v3 ); -} - - -/** - * Main entrypoint to draw some number of points/lines/triangles - */ -static void -draw_prim( struct draw_context *draw, - unsigned prim, unsigned start, unsigned count ) -{ - unsigned i; - boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); - boolean flatfirst = - (draw->rasterizer->flatshade & draw->rasterizer->flatshade_first) ? TRUE : FALSE; - -// debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - - switch (prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i ++) { - do_point( draw, - start + i ); - } - break; - - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) { - do_line( draw, - TRUE, - start + i + 0, - start + i + 1); - } - break; - - case PIPE_PRIM_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, /* XXX: only if vb not split */ - start + i - 1, - start + i ); - } - - do_line( draw, - 0, - start + count - 1, - start + 0 ); - } - break; - - case PIPE_PRIM_LINE_STRIP: - for (i = 1; i < count; i++) { - do_line( draw, - i == 1, - start + i - 1, - start + i ); - } - break; - - case PIPE_PRIM_TRIANGLES: - if (unfilled) { - for (i = 0; i+2 < count; i += 3) { - do_ef_triangle( draw, - 1, - ~0, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - else { - for (i = 0; i+2 < count; i += 3) { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - break; - - case PIPE_PRIM_TRIANGLE_STRIP: - if (flatfirst) { - for (i = 0; i+2 < count; i++) { - if (i & 1) { - do_triangle( draw, - start + i + 0, - start + i + 2, - start + i + 1 ); - } - else { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - } - else { - for (i = 0; i+2 < count; i++) { - if (i & 1) { - do_triangle( draw, - start + i + 1, - start + i + 0, - start + i + 2 ); - } - else { - do_triangle( draw, - start + i + 0, - start + i + 1, - start + i + 2 ); - } - } - } - break; - - case PIPE_PRIM_TRIANGLE_FAN: - if (count >= 3) { - if (flatfirst) { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + i + 1, - start + i + 2, - start + 0 ); - } - } - else { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + 0, - start + i + 1, - start + i + 2 ); - } - } - } - break; - - - case PIPE_PRIM_QUADS: - if (unfilled) { - for (i = 0; i+3 < count; i += 4) { - do_ef_quad( draw, - start + i + 0, - start + i + 1, - start + i + 2, - start + i + 3); - } - } - else { - for (i = 0; i+3 < count; i += 4) { - do_quad( draw, - start + i + 0, - start + i + 1, - start + i + 2, - start + i + 3); - } - } - break; - - case PIPE_PRIM_QUAD_STRIP: - if (unfilled) { - for (i = 0; i+3 < count; i += 2) { - do_ef_quad( draw, - start + i + 2, - start + i + 0, - start + i + 1, - start + i + 3); - } - } - else { - for (i = 0; i+3 < count; i += 2) { - do_quad( draw, - start + i + 2, - start + i + 0, - start + i + 1, - start + i + 3); - } - } - break; - - case PIPE_PRIM_POLYGON: - if (unfilled) { - unsigned ef_mask = (1<<2) | (1<<0); - - for (i = 0; i+2 < count; i++) { - - if (i + 3 >= count) - ef_mask |= (1<<1); - - do_ef_triangle( draw, - i == 0, - ef_mask, - start + i + 1, - start + i + 2, - start + 0); - - ef_mask &= ~(1<<2); - } - } - else { - for (i = 0; i+2 < count; i++) { - do_triangle( draw, - start + i + 1, - start + i + 2, - start + 0); - } - } - break; - - default: - assert(0); - break; - } -} - - - - - diff --git a/src/gallium/auxiliary/draw/draw_vertex_cache.c b/src/gallium/auxiliary/draw/draw_vertex_cache.c deleted file mode 100644 index 730c18bcb3..0000000000 --- a/src/gallium/auxiliary/draw/draw_vertex_cache.c +++ /dev/null @@ -1,219 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "draw_private.h" -#include "draw_context.h" - - -void draw_vertex_cache_invalidate( struct draw_context *draw ) -{ - assert(draw->pq.queue_nr == 0); - assert(draw->vs.queue_nr == 0); - assert(draw->vcache.referenced == 0); - - /* There's an error somewhere in the vcache code that requires this - * memset. The bug is exposed in q3demo demo001, but probably - * elsewhere as well. Will track it down later. - */ - memset(draw->vcache.idx, ~0, sizeof(draw->vcache.idx)); -} - - -/** - * Check if vertex is in cache, otherwise add it. It won't go through - * VS yet, not until there is a flush operation or the VS queue fills up. - * - * Note that cache entries are basically just two pointers: the first - * an index into the user's vertex arrays, the second a location in - * the vertex shader cache for the post-transformed vertex. - * - * \return pointer to location of (post-transformed) vertex header in the cache - */ -static struct vertex_header *get_vertex( struct draw_context *draw, - unsigned i ) -{ - unsigned slot = (i + (i>>5)) % VCACHE_SIZE; - - assert(slot < 32); /* so we don't exceed the bitfield size below */ - - if (draw->vcache.referenced & (1<vcache.idx[slot].in == i) { - /*debug_printf("HIT %d %d\n", slot, i);*/ - assert(draw->vcache.idx[slot].out < draw->vs.queue_nr); - return draw_header_from_block(draw->vs.vertex_cache, - MAX_VERTEX_ALLOCATION, - draw->vcache.idx[slot].out); - } - - /* Otherwise a collision - */ - slot = VCACHE_SIZE + draw->vcache.overflow++; - /*debug_printf("XXX %d --> %d\n", i, slot);*/ - } - - /* Deal with the cache miss: - */ - { - unsigned out; - struct vertex_header *header; - - assert(slot < Elements(draw->vcache.idx)); - - /*debug_printf("NEW %d %d\n", slot, i);*/ - draw->vcache.idx[slot].in = i; - draw->vcache.idx[slot].out = out = draw->vs.queue_nr++; - draw->vcache.referenced |= (1 << slot); - - - /* Add to vertex shader queue: - */ - assert(draw->vs.queue_nr < VS_QUEUE_LENGTH); - - header = draw_header_from_block(draw->vs.vertex_cache, MAX_VERTEX_ALLOCATION, - out); - draw->vs.elts[out] = i; - header->clipmask = 0; - header->edgeflag = draw_get_edgeflag(draw, i); - header->pad = 0; - header->vertex_id = UNDEFINED_VERTEX_ID; - - /* Need to set the vertex's edge flag here. If we're being called - * by do_ef_triangle(), that function needs edge flag info! - */ - - return draw_header_from_block(draw->vs.vertex_cache, - MAX_VERTEX_ALLOCATION, - draw->vcache.idx[slot].out); - } -} - - -static struct vertex_header *get_uint_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const unsigned *elts = (const unsigned *) draw->user.elts; - return get_vertex( draw, elts[i] ); -} - - -static struct vertex_header *get_ushort_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const ushort *elts = (const ushort *) draw->user.elts; - return get_vertex( draw, elts[i] ); -} - - -static struct vertex_header *get_ubyte_elt_vertex( struct draw_context *draw, - unsigned i ) -{ - const ubyte *elts = (const ubyte *) draw->user.elts; - return get_vertex( draw, elts[i] ); -} - - -void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ) -{ - unsigned i; - - for (i = 0; i < draw->vs.post_nr; i++) { - struct vertex_header * header = - draw_header_from_block(draw->vs.vertex_cache, - MAX_VERTEX_ALLOCATION, i); - header->vertex_id = UNDEFINED_VERTEX_ID; - } -} - - -void draw_vertex_cache_unreference( struct draw_context *draw ) -{ - draw->vcache.referenced = 0; - draw->vcache.overflow = 0; -} - - -int draw_vertex_cache_check_space( struct draw_context *draw, - unsigned nr_verts ) -{ - if (draw->vcache.overflow + nr_verts < VCACHE_OVERFLOW) { - /* The vs queue is sized so that this can never happen: - */ - assert(draw->vs.queue_nr + nr_verts < VS_QUEUE_LENGTH); - return TRUE; - } - else - return FALSE; -} - - - -/** - * Tell the drawing context about the index/element buffer to use - * (ala glDrawElements) - * If no element buffer is to be used (i.e. glDrawArrays) then this - * should be called with eltSize=0 and elements=NULL. - * - * \param draw the drawing context - * \param eltSize size of each element (1, 2 or 4 bytes) - * \param elements the element buffer ptr - */ -void -draw_set_mapped_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ) -{ -// draw_statechange( draw ); - - /* choose the get_vertex() function to use */ - switch (eltSize) { - case 0: - draw->vcache.get_vertex = get_vertex; - break; - case 1: - draw->vcache.get_vertex = get_ubyte_elt_vertex; - break; - case 2: - draw->vcache.get_vertex = get_ushort_elt_vertex; - break; - case 4: - draw->vcache.get_vertex = get_uint_elt_vertex; - break; - default: - assert(0); - } - draw->user.elts = elements; - draw->user.eltSize = eltSize; -} - -- cgit v1.2.3 From dd903d83b3ff8dd19f75ac7542b96bc8f1387fe6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 20:41:39 +0100 Subject: draw: remove old vertex_shader->run() functions --- src/gallium/auxiliary/draw/draw_vertex_shader.c | 2 +- src/gallium/auxiliary/draw/draw_vs.h | 7 -- src/gallium/auxiliary/draw/draw_vs_exec.c | 137 ------------------------ src/gallium/auxiliary/draw/draw_vs_llvm.c | 108 ------------------- src/gallium/auxiliary/draw/draw_vs_sse.c | 126 ---------------------- 5 files changed, 1 insertion(+), 379 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index c0c17c116e..118664d619 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -74,7 +74,7 @@ draw_vertex_shader_queue_flush(struct draw_context *draw) assert(n > 0); assert(n <= MAX_SHADER_VERTICES); - shader->run(shader, draw, elts, n, dests, MAX_VERTEX_ALLOCATION); +// shader->run(shader, draw, elts, n, dests, MAX_VERTEX_ALLOCATION); } draw->vs.post_nr = draw->vs.queue_nr; diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index e88d00e247..b6f3bbdaea 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -55,13 +55,6 @@ struct draw_vertex_shader { /* Run the shader - this interface will get cleaned up in the * future: */ - boolean (*run)( struct draw_vertex_shader *shader, - struct draw_context *draw, - const unsigned *elts, - unsigned count, - void *out, - unsigned vertex_size); - void (*run_linear)( struct draw_vertex_shader *shader, const float (*input)[4], float (*output)[4], diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 184151b9b1..54a2b2ab04 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -66,145 +66,9 @@ vs_exec_prepare( struct draw_vertex_shader *shader, PIPE_MAX_SAMPLERS, NULL /*samplers*/ ); - draw_update_vertex_fetch( draw ); } -/** - * Transform vertices with the current vertex program/shader - * Up to four vertices can be shaded at a time. - * \param vbuffer the input vertex data - * \param elts indexes of four input vertices - * \param count number of vertices to shade [1..4] - * \param vOut array of pointers to four output vertices - */ -static boolean -vs_exec_run( struct draw_vertex_shader *shader, - struct draw_context *draw, - const unsigned *elts, - unsigned count, - void *vOut, - unsigned vertex_size) -{ - struct exec_vertex_shader *evs = exec_vertex_shader(shader); - struct tgsi_exec_machine *machine = evs->machine; - unsigned int i, j; - unsigned int clipped = 0; - struct tgsi_exec_vector *outputs = 0; - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - - assert(shader->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); - - machine->Consts = (const float (*)[4]) draw->user.constants; - - if (draw->rasterizer->bypass_vs) { - /* outputs are just the inputs */ - outputs = machine->Inputs; - } - else { - outputs = machine->Outputs; - } - - for (i = 0; i < count; i += MAX_TGSI_VERTICES) { - unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); - draw->vertex_fetch.fetch_func( draw, machine, &elts[i], max_vertices ); - -#if 0 - for (j = 0; j < max_vertices; j++) { - unsigned slot; - debug_printf("%d) Input vert:\n", i + j); - for (slot = 0; slot < shader->info.num_inputs; slot++) { - debug_printf("\t%d: %f %f %f %f\n", slot, - machine->Inputs[slot].xyzw[0].f[j], - machine->Inputs[slot].xyzw[1].f[j], - machine->Inputs[slot].xyzw[2].f[j], - machine->Inputs[slot].xyzw[3].f[j]); - } - } -#endif - - - if (!draw->rasterizer->bypass_vs) { - /* run interpreter */ - tgsi_exec_machine_run( machine ); - } - - /* store machine results */ - for (j = 0; j < max_vertices; j++) { - unsigned slot; - float x, y, z, w; - struct vertex_header *out = - draw_header_from_block(vOut, vertex_size, i + j); - - /* Handle attr[0] (position) specially: - * - * XXX: Computing the clipmask should be done in the vertex - * program as a set of DP4 instructions appended to the - * user-provided code. - */ - x = out->clip[0] = outputs[0].xyzw[0].f[j]; - y = out->clip[1] = outputs[0].xyzw[1].f[j]; - z = out->clip[2] = outputs[0].xyzw[2].f[j]; - w = out->clip[3] = outputs[0].xyzw[3].f[j]; - - if (!draw->rasterizer->bypass_clipping) { - out->clipmask = compute_clipmask(out->clip, draw->plane, - draw->nr_planes); - clipped += out->clipmask; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - } - else { - out->clipmask = 0; - } - out->edgeflag = 1; - out->vertex_id = UNDEFINED_VERTEX_ID; - - if (!draw->identity_viewport) { - /* Viewport mapping */ - out->data[0][0] = x * scale[0] + trans[0]; - out->data[0][1] = y * scale[1] + trans[1]; - out->data[0][2] = z * scale[2] + trans[2]; - out->data[0][3] = w; - } - else - { - out->data[0][0] = x; - out->data[0][1] = y; - out->data[0][2] = z; - out->data[0][3] = w; - } - - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - out->data[slot][0] = outputs[slot].xyzw[0].f[j]; - out->data[slot][1] = outputs[slot].xyzw[1].f[j]; - out->data[slot][2] = outputs[slot].xyzw[2].f[j]; - out->data[slot][3] = outputs[slot].xyzw[3].f[j]; - } - -#if 0 /*DEBUG*/ - printf("%d) Post xform vert:\n", i + j); - for (slot = 0; slot < draw->num_vs_outputs; slot++) { - printf("\t%d: %f %f %f %f\n", slot, - out->data[slot][0], - out->data[slot][1], - out->data[slot][2], - out->data[slot][3]); - } -#endif - } /* loop over vertices */ - } - return clipped != 0; -} - /* Simplified vertex shader interface for the pt paths. Given the @@ -312,7 +176,6 @@ draw_create_vs_exec(struct draw_context *draw, vs->base.prepare = vs_exec_prepare; - vs->base.run = vs_exec_run; vs->base.run_linear = vs_exec_run_linear; vs->base.delete = vs_exec_delete; vs->machine = &draw->machine; diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 4dbfa955a4..5e27bc9ff0 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -60,113 +60,6 @@ vs_llvm_prepare( struct draw_vertex_shader *base, -/** - * Transform vertices with the current vertex program/shader - * Up to four vertices can be shaded at a time. - * \param vbuffer the input vertex data - * \param elts indexes of four input vertices - * \param count number of vertices to shade [1..4] - * \param vOut array of pointers to four output vertices - */ -static boolean -vs_llvm_run( struct draw_vertex_shader *base, - struct draw_context *draw, - const unsigned *elts, - unsigned count, - void *vOut ) -{ - struct draw_llvm_vertex_shader *shader = - (struct draw_llvm_vertex_shader *)base; - - struct tgsi_exec_machine *machine = shader->machine; - unsigned int j; - unsigned int clipped = 0; - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - - - assert(count <= 4); - assert(base->state->output_semantic_name[0] == TGSI_SEMANTIC_POSITION); - - /* Consts does not require 16 byte alignment. */ - machine->Consts = (float (*)[4]) draw->user.constants; - - if (draw->rasterizer->bypass_vs) { - /* outputs are just the inputs */ - outputs = machine->Inputs; - } - else { - outputs = machine->Outputs; - } - - - draw->vertex_fetch.fetch_func( draw, machine, elts, count ); - - if (!draw->rasterizer->bypass_vs) { - /* run shader */ - gallivm_cpu_vs_exec(shader->llvm_prog, - machine->Inputs, - machine->Outputs, - machine->Consts, - machine->Temps); - } - - /* store machine results */ - for (j = 0; j < count; j++) { - unsigned slot; - float x, y, z, w; - - x = vOut[j]->clip[0] = outputs[0].xyzw[0].f[j]; - y = vOut[j]->clip[1] = outputs[0].xyzw[1].f[j]; - z = vOut[j]->clip[2] = outputs[0].xyzw[2].f[j]; - w = vOut[j]->clip[3] = outputs[0].xyzw[3].f[j]; - - if (!draw->rasterizer->bypass_clipping) { - vOut[j]->clipmask = compute_clipmask(vOut[j]->clip, draw->plane, - draw->nr_planes); - clipped += vOut[j]->clipmask; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - } - else { - vOut[j]->clipmask = 0; - } - vOut[j]->edgeflag = 1; - vOut[j]->vertex_id = UNDEFINED_VERTEX_ID; - - if (!draw->identity_viewport) { - /* Viewport mapping */ - vOut[j]->data[0][0] = x * scale[0] + trans[0]; - vOut[j]->data[0][1] = y * scale[1] + trans[1]; - vOut[j]->data[0][2] = z * scale[2] + trans[2]; - vOut[j]->data[0][3] = w; - } - else { - vOut[j]->data[0][0] = x; - vOut[j]->data[0][1] = y; - vOut[j]->data[0][2] = z; - vOut[j]->data[0][3] = w; - } - - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - vOut[j]->data[slot][0] = outputs[slot].xyzw[0].f[j]; - vOut[j]->data[slot][1] = outputs[slot].xyzw[1].f[j]; - vOut[j]->data[slot][2] = outputs[slot].xyzw[2].f[j]; - vOut[j]->data[slot][3] = outputs[slot].xyzw[3].f[j]; - } - } /* loop over vertices */ - return clipped != 0; -} - - - static void vs_llvm_run_linear( struct draw_vertex_shader *base, @@ -256,7 +149,6 @@ draw_create_vs_llvm(struct draw_context *draw, tgsi_scan_shader(shader->tokens, &vs->base.info); vs->base.prepare = vs_llvm_prepare; - vs->base.run = vs_llvm_run; vs->base.run_linear = vs_llvm_run_linear; vs->base.delete = vs_llvm_delete; vs->machine = &draw->machine; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index a763f3845c..60f60a5b53 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -74,131 +74,6 @@ vs_sse_prepare( struct draw_vertex_shader *base, draw_update_vertex_fetch( draw ); } -/** - * Transform vertices with the current vertex program/shader - * Up to four vertices can be shaded at a time. - * \param vbuffer the input vertex data - * \param elts indexes of four input vertices - * \param count number of vertices to shade [1..4] - * \param vOut array of pointers to four output vertices - */ -static boolean -vs_sse_run( struct draw_vertex_shader *base, - struct draw_context *draw, - const unsigned *elts, - unsigned count, - void *vOut, - unsigned vertex_size ) -{ - struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; - struct tgsi_exec_machine *machine = shader->machine; - unsigned int i, j; - unsigned int clipped = 0; - struct tgsi_exec_vector *outputs = 0; - const float *scale = draw->viewport.scale; - const float *trans = draw->viewport.translate; - - assert(base->info.output_semantic_name[0] == TGSI_SEMANTIC_POSITION); - - /* Consts does not require 16 byte alignment. */ - machine->Consts = (const float (*)[4]) draw->user.constants; - - if (draw->rasterizer->bypass_vs) { - /* outputs are just the inputs */ - outputs = machine->Inputs; - } - else { - outputs = machine->Outputs; - } - - for (i = 0; i < count; i += SSE_MAX_VERTICES) { - unsigned int max_vertices = MIN2(SSE_MAX_VERTICES, count - i); - /* Fetch vertices. This may at some point be integrated into the - * compiled shader -- that would require a reorganization where - * multiple versions of the compiled shader might exist, - * specialized for each fetch state. - */ - draw->vertex_fetch.fetch_func(draw, machine, &elts[i], max_vertices); - - if (!draw->rasterizer->bypass_vs) { - /* run compiled shader - */ - shader->func(machine->Inputs, - machine->Outputs, - (float (*)[4])machine->Consts, - machine->Temps, - shader->immediates); - } - - /* XXX: Computing the clipmask and emitting results should be done - * in the vertex program as a set of instructions appended to - * the user-provided code. - */ - for (j = 0; j < max_vertices; j++) { - unsigned slot; - float x, y, z, w; - struct vertex_header *out = - draw_header_from_block(vOut, vertex_size, i + j); - - x = out->clip[0] = outputs[0].xyzw[0].f[j]; - y = out->clip[1] = outputs[0].xyzw[1].f[j]; - z = out->clip[2] = outputs[0].xyzw[2].f[j]; - w = out->clip[3] = outputs[0].xyzw[3].f[j]; - - if (!draw->rasterizer->bypass_clipping) { - out->clipmask = compute_clipmask(out->clip, draw->plane, - draw->nr_planes); - clipped += out->clipmask; - - /* divide by w */ - w = 1.0f / w; - x *= w; - y *= w; - z *= w; - } - else { - out->clipmask = 0; - } - out->edgeflag = 1; - out->vertex_id = UNDEFINED_VERTEX_ID; - - if (!draw->identity_viewport) { - /* Viewport mapping */ - out->data[0][0] = x * scale[0] + trans[0]; - out->data[0][1] = y * scale[1] + trans[1]; - out->data[0][2] = z * scale[2] + trans[2]; - out->data[0][3] = w; - } - else { - out->data[0][0] = x; - out->data[0][1] = y; - out->data[0][2] = z; - out->data[0][3] = w; - } - - /* Remaining attributes are packed into sequential post-transform - * vertex attrib slots. - */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { - out->data[slot][0] = outputs[slot].xyzw[0].f[j]; - out->data[slot][1] = outputs[slot].xyzw[1].f[j]; - out->data[slot][2] = outputs[slot].xyzw[2].f[j]; - out->data[slot][3] = outputs[slot].xyzw[3].f[j]; - } -#if 0 /*DEBUG*/ - printf("%d) Post xform vert:\n", i + j); - for (slot = 0; slot < draw->num_vs_outputs; slot++) { - printf("\t%d: %f %f %f %f\n", slot, - out->data[slot][0], - out->data[slot][1], - out->data[slot][2], - out->data[slot][3]); - } -#endif - } - } - return clipped != 0; -} /* Simplified vertex shader interface for the pt paths. Given the @@ -294,7 +169,6 @@ draw_create_vs_sse(struct draw_context *draw, tgsi_scan_shader(templ->tokens, &vs->base.info); vs->base.prepare = vs_sse_prepare; - vs->base.run = vs_sse_run; vs->base.run_linear = vs_sse_run_linear; vs->base.delete = vs_sse_delete; vs->machine = &draw->machine; -- cgit v1.2.3 From 709e33cf0bfd552220e46f44e8cfa2063c3cef69 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 20:44:13 +0100 Subject: draw: remove old draw_vertex_shader_queue_flush function --- src/gallium/auxiliary/draw/draw_context.c | 2 -- src/gallium/auxiliary/draw/draw_private.h | 6 ---- src/gallium/auxiliary/draw/draw_vertex_shader.c | 43 ------------------------- 3 files changed, 51 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index c611300841..9c3ae9bc0d 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -93,8 +93,6 @@ struct draw_context *draw_create( void ) draw->vs.vertex_cache = tmp; } - draw->shader_queue_flush = draw_vertex_shader_queue_flush; - /* these defaults are oriented toward the needs of softpipe */ draw->wide_point_threshold = 1000000.0; /* infinity */ draw->wide_line_threshold = 1.0; diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 18ce6c0ec5..93add257a3 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -309,11 +309,6 @@ struct draw_context unsigned post_nr; } vs; - /** - * Run the vertex shader on all vertices in the vertex queue. - */ - void (*shader_queue_flush)(struct draw_context *draw); - /* Prim pipeline queue: */ struct { @@ -359,7 +354,6 @@ extern void draw_vertex_cache_invalidate( struct draw_context *draw ); extern void draw_vertex_cache_unreference( struct draw_context *draw ); extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); -extern void draw_vertex_shader_queue_flush( struct draw_context *draw ); extern void draw_update_vertex_fetch( struct draw_context *draw ); diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c index 118664d619..03fe00a951 100644 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ b/src/gallium/auxiliary/draw/draw_vertex_shader.c @@ -37,49 +37,6 @@ #include "draw_context.h" #include "draw_vs.h" -/** - * Run the vertex shader on all vertices in the vertex queue. - * Called by the draw module when the vertx cache needs to be flushed. - */ -void -draw_vertex_shader_queue_flush(struct draw_context *draw) -{ - struct draw_vertex_shader *shader = draw->vertex_shader; - unsigned i; - - assert(draw->vs.queue_nr != 0); - - /* XXX: do this on statechange: - */ - shader->prepare( shader, draw ); - -// fprintf(stderr, "%s %d\n", __FUNCTION__, draw->vs.queue_nr ); - - /* run vertex shader on vertex cache entries, four per invokation */ - for (i = 0; i < draw->vs.queue_nr; i += MAX_SHADER_VERTICES) { - unsigned elts[MAX_SHADER_VERTICES]; - int j, n = MIN2(MAX_SHADER_VERTICES, draw->vs.queue_nr - i); - struct vertex_header *dests = - draw_header_from_block(draw->vs.vertex_cache, - MAX_VERTEX_ALLOCATION, i); - - for (j = 0; j < n; j++) { - elts[j] = draw->vs.elts[i + j]; - } - - for ( ; j < MAX_SHADER_VERTICES; j++) { - elts[j] = elts[0]; - } - - assert(n > 0); - assert(n <= MAX_SHADER_VERTICES); - -// shader->run(shader, draw, elts, n, dests, MAX_VERTEX_ALLOCATION); - } - - draw->vs.post_nr = draw->vs.queue_nr; - draw->vs.queue_nr = 0; -} struct draw_vertex_shader * -- cgit v1.2.3 From 415e8e039ba38716336e8de3d7b3cdc23b9a9d8e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 18 Apr 2008 20:46:06 +0100 Subject: draw: remove draw_vertex_fetch.c --- src/gallium/auxiliary/draw/Makefile | 1 - src/gallium/auxiliary/draw/SConscript | 2 - src/gallium/auxiliary/draw/draw_vertex_fetch.c | 528 ------------------------- src/gallium/auxiliary/draw/draw_vs_llvm.c | 1 - src/gallium/auxiliary/draw/draw_vs_sse.c | 1 - 5 files changed, 533 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_vertex_fetch.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 03210f6a4a..60bb9dfe88 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -31,7 +31,6 @@ C_SOURCES = \ draw_validate.c \ draw_vbuf.c \ draw_vertex.c \ - draw_vertex_fetch.c \ draw_vertex_shader.c \ draw_wide_line.c \ draw_wide_point.c diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 238e3f7d28..246083a962 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -27,8 +27,6 @@ draw = env.ConvenienceLibrary( 'draw_validate.c', 'draw_vbuf.c', 'draw_vertex.c', - 'draw_vertex_cache.c', - 'draw_vertex_fetch.c', 'draw_vertex_shader.c', 'draw_vs_exec.c', 'draw_vs_llvm.c', diff --git a/src/gallium/auxiliary/draw/draw_vertex_fetch.c b/src/gallium/auxiliary/draw/draw_vertex_fetch.c deleted file mode 100644 index 0bc2fcb424..0000000000 --- a/src/gallium/auxiliary/draw/draw_vertex_fetch.c +++ /dev/null @@ -1,528 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "draw_vs.h" -#include "draw_context.h" - - -#define DRAW_DBG 0 - - -/** - * Fetch a float[4] vertex attribute from memory, doing format/type - * conversion as needed. - * - * This is probably needed/dupliocated elsewhere, eg format - * conversion, texture sampling etc. - */ -#define FETCH_ATTRIB( NAME, SZ, CVT ) \ -static void \ -fetch_##NAME(const void *ptr, float *attrib) \ -{ \ - static const float defaults[4] = { 0,0,0,1 }; \ - int i; \ - \ - for (i = 0; i < SZ; i++) { \ - attrib[i] = CVT(i); \ - } \ - \ - for (; i < 4; i++) { \ - attrib[i] = defaults[i]; \ - } \ -} - -#define CVT_64_FLOAT(i) (float) ((double *) ptr)[i] -#define CVT_32_FLOAT(i) ((float *) ptr)[i] - -#define CVT_8_USCALED(i) (float) ((unsigned char *) ptr)[i] -#define CVT_16_USCALED(i) (float) ((unsigned short *) ptr)[i] -#define CVT_32_USCALED(i) (float) ((unsigned int *) ptr)[i] - -#define CVT_8_SSCALED(i) (float) ((char *) ptr)[i] -#define CVT_16_SSCALED(i) (float) ((short *) ptr)[i] -#define CVT_32_SSCALED(i) (float) ((int *) ptr)[i] - -#define CVT_8_UNORM(i) (float) ((unsigned char *) ptr)[i] / 255.0f -#define CVT_16_UNORM(i) (float) ((unsigned short *) ptr)[i] / 65535.0f -#define CVT_32_UNORM(i) (float) ((unsigned int *) ptr)[i] / 4294967295.0f - -#define CVT_8_SNORM(i) (float) ((char *) ptr)[i] / 127.0f -#define CVT_16_SNORM(i) (float) ((short *) ptr)[i] / 32767.0f -#define CVT_32_SNORM(i) (float) ((int *) ptr)[i] / 2147483647.0f - -FETCH_ATTRIB( R64G64B64A64_FLOAT, 4, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64B64_FLOAT, 3, CVT_64_FLOAT ) -FETCH_ATTRIB( R64G64_FLOAT, 2, CVT_64_FLOAT ) -FETCH_ATTRIB( R64_FLOAT, 1, CVT_64_FLOAT ) - -FETCH_ATTRIB( R32G32B32A32_FLOAT, 4, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32B32_FLOAT, 3, CVT_32_FLOAT ) -FETCH_ATTRIB( R32G32_FLOAT, 2, CVT_32_FLOAT ) -FETCH_ATTRIB( R32_FLOAT, 1, CVT_32_FLOAT ) - -FETCH_ATTRIB( R32G32B32A32_USCALED, 4, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32B32_USCALED, 3, CVT_32_USCALED ) -FETCH_ATTRIB( R32G32_USCALED, 2, CVT_32_USCALED ) -FETCH_ATTRIB( R32_USCALED, 1, CVT_32_USCALED ) - -FETCH_ATTRIB( R32G32B32A32_SSCALED, 4, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32B32_SSCALED, 3, CVT_32_SSCALED ) -FETCH_ATTRIB( R32G32_SSCALED, 2, CVT_32_SSCALED ) -FETCH_ATTRIB( R32_SSCALED, 1, CVT_32_SSCALED ) - -FETCH_ATTRIB( R32G32B32A32_UNORM, 4, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32B32_UNORM, 3, CVT_32_UNORM ) -FETCH_ATTRIB( R32G32_UNORM, 2, CVT_32_UNORM ) -FETCH_ATTRIB( R32_UNORM, 1, CVT_32_UNORM ) - -FETCH_ATTRIB( R32G32B32A32_SNORM, 4, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32B32_SNORM, 3, CVT_32_SNORM ) -FETCH_ATTRIB( R32G32_SNORM, 2, CVT_32_SNORM ) -FETCH_ATTRIB( R32_SNORM, 1, CVT_32_SNORM ) - -FETCH_ATTRIB( R16G16B16A16_USCALED, 4, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16B16_USCALED, 3, CVT_16_USCALED ) -FETCH_ATTRIB( R16G16_USCALED, 2, CVT_16_USCALED ) -FETCH_ATTRIB( R16_USCALED, 1, CVT_16_USCALED ) - -FETCH_ATTRIB( R16G16B16A16_SSCALED, 4, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16B16_SSCALED, 3, CVT_16_SSCALED ) -FETCH_ATTRIB( R16G16_SSCALED, 2, CVT_16_SSCALED ) -FETCH_ATTRIB( R16_SSCALED, 1, CVT_16_SSCALED ) - -FETCH_ATTRIB( R16G16B16A16_UNORM, 4, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16B16_UNORM, 3, CVT_16_UNORM ) -FETCH_ATTRIB( R16G16_UNORM, 2, CVT_16_UNORM ) -FETCH_ATTRIB( R16_UNORM, 1, CVT_16_UNORM ) - -FETCH_ATTRIB( R16G16B16A16_SNORM, 4, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16B16_SNORM, 3, CVT_16_SNORM ) -FETCH_ATTRIB( R16G16_SNORM, 2, CVT_16_SNORM ) -FETCH_ATTRIB( R16_SNORM, 1, CVT_16_SNORM ) - -FETCH_ATTRIB( R8G8B8A8_USCALED, 4, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8B8_USCALED, 3, CVT_8_USCALED ) -FETCH_ATTRIB( R8G8_USCALED, 2, CVT_8_USCALED ) -FETCH_ATTRIB( R8_USCALED, 1, CVT_8_USCALED ) - -FETCH_ATTRIB( R8G8B8A8_SSCALED, 4, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8B8_SSCALED, 3, CVT_8_SSCALED ) -FETCH_ATTRIB( R8G8_SSCALED, 2, CVT_8_SSCALED ) -FETCH_ATTRIB( R8_SSCALED, 1, CVT_8_SSCALED ) - -FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8B8_UNORM, 3, CVT_8_UNORM ) -FETCH_ATTRIB( R8G8_UNORM, 2, CVT_8_UNORM ) -FETCH_ATTRIB( R8_UNORM, 1, CVT_8_UNORM ) - -FETCH_ATTRIB( R8G8B8A8_SNORM, 4, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8B8_SNORM, 3, CVT_8_SNORM ) -FETCH_ATTRIB( R8G8_SNORM, 2, CVT_8_SNORM ) -FETCH_ATTRIB( R8_SNORM, 1, CVT_8_SNORM ) - -FETCH_ATTRIB( A8R8G8B8_UNORM, 4, CVT_8_UNORM ) -//FETCH_ATTRIB( R8G8B8A8_UNORM, 4, CVT_8_UNORM ) - - - -static void -fetch_B8G8R8A8_UNORM(const void *ptr, float *attrib) -{ - attrib[2] = CVT_8_UNORM(0); - attrib[1] = CVT_8_UNORM(1); - attrib[0] = CVT_8_UNORM(2); - attrib[3] = CVT_8_UNORM(3); -} - - -fetch_func draw_get_fetch_func( enum pipe_format format ) -{ -#if 0 - { - char tmp[80]; - pf_sprint_name(tmp, format); - debug_printf("%s: %s\n", __FUNCTION__, tmp); - } -#endif - - switch (format) { - case PIPE_FORMAT_R64_FLOAT: - return fetch_R64_FLOAT; - case PIPE_FORMAT_R64G64_FLOAT: - return fetch_R64G64_FLOAT; - case PIPE_FORMAT_R64G64B64_FLOAT: - return fetch_R64G64B64_FLOAT; - case PIPE_FORMAT_R64G64B64A64_FLOAT: - return fetch_R64G64B64A64_FLOAT; - - case PIPE_FORMAT_R32_FLOAT: - return fetch_R32_FLOAT; - case PIPE_FORMAT_R32G32_FLOAT: - return fetch_R32G32_FLOAT; - case PIPE_FORMAT_R32G32B32_FLOAT: - return fetch_R32G32B32_FLOAT; - case PIPE_FORMAT_R32G32B32A32_FLOAT: - return fetch_R32G32B32A32_FLOAT; - - case PIPE_FORMAT_R32_UNORM: - return fetch_R32_UNORM; - case PIPE_FORMAT_R32G32_UNORM: - return fetch_R32G32_UNORM; - case PIPE_FORMAT_R32G32B32_UNORM: - return fetch_R32G32B32_UNORM; - case PIPE_FORMAT_R32G32B32A32_UNORM: - return fetch_R32G32B32A32_UNORM; - - case PIPE_FORMAT_R32_USCALED: - return fetch_R32_USCALED; - case PIPE_FORMAT_R32G32_USCALED: - return fetch_R32G32_USCALED; - case PIPE_FORMAT_R32G32B32_USCALED: - return fetch_R32G32B32_USCALED; - case PIPE_FORMAT_R32G32B32A32_USCALED: - return fetch_R32G32B32A32_USCALED; - - case PIPE_FORMAT_R32_SNORM: - return fetch_R32_SNORM; - case PIPE_FORMAT_R32G32_SNORM: - return fetch_R32G32_SNORM; - case PIPE_FORMAT_R32G32B32_SNORM: - return fetch_R32G32B32_SNORM; - case PIPE_FORMAT_R32G32B32A32_SNORM: - return fetch_R32G32B32A32_SNORM; - - case PIPE_FORMAT_R32_SSCALED: - return fetch_R32_SSCALED; - case PIPE_FORMAT_R32G32_SSCALED: - return fetch_R32G32_SSCALED; - case PIPE_FORMAT_R32G32B32_SSCALED: - return fetch_R32G32B32_SSCALED; - case PIPE_FORMAT_R32G32B32A32_SSCALED: - return fetch_R32G32B32A32_SSCALED; - - case PIPE_FORMAT_R16_UNORM: - return fetch_R16_UNORM; - case PIPE_FORMAT_R16G16_UNORM: - return fetch_R16G16_UNORM; - case PIPE_FORMAT_R16G16B16_UNORM: - return fetch_R16G16B16_UNORM; - case PIPE_FORMAT_R16G16B16A16_UNORM: - return fetch_R16G16B16A16_UNORM; - - case PIPE_FORMAT_R16_USCALED: - return fetch_R16_USCALED; - case PIPE_FORMAT_R16G16_USCALED: - return fetch_R16G16_USCALED; - case PIPE_FORMAT_R16G16B16_USCALED: - return fetch_R16G16B16_USCALED; - case PIPE_FORMAT_R16G16B16A16_USCALED: - return fetch_R16G16B16A16_USCALED; - - case PIPE_FORMAT_R16_SNORM: - return fetch_R16_SNORM; - case PIPE_FORMAT_R16G16_SNORM: - return fetch_R16G16_SNORM; - case PIPE_FORMAT_R16G16B16_SNORM: - return fetch_R16G16B16_SNORM; - case PIPE_FORMAT_R16G16B16A16_SNORM: - return fetch_R16G16B16A16_SNORM; - - case PIPE_FORMAT_R16_SSCALED: - return fetch_R16_SSCALED; - case PIPE_FORMAT_R16G16_SSCALED: - return fetch_R16G16_SSCALED; - case PIPE_FORMAT_R16G16B16_SSCALED: - return fetch_R16G16B16_SSCALED; - case PIPE_FORMAT_R16G16B16A16_SSCALED: - return fetch_R16G16B16A16_SSCALED; - - case PIPE_FORMAT_R8_UNORM: - return fetch_R8_UNORM; - case PIPE_FORMAT_R8G8_UNORM: - return fetch_R8G8_UNORM; - case PIPE_FORMAT_R8G8B8_UNORM: - return fetch_R8G8B8_UNORM; - case PIPE_FORMAT_R8G8B8A8_UNORM: - return fetch_R8G8B8A8_UNORM; - - case PIPE_FORMAT_R8_USCALED: - return fetch_R8_USCALED; - case PIPE_FORMAT_R8G8_USCALED: - return fetch_R8G8_USCALED; - case PIPE_FORMAT_R8G8B8_USCALED: - return fetch_R8G8B8_USCALED; - case PIPE_FORMAT_R8G8B8A8_USCALED: - return fetch_R8G8B8A8_USCALED; - - case PIPE_FORMAT_R8_SNORM: - return fetch_R8_SNORM; - case PIPE_FORMAT_R8G8_SNORM: - return fetch_R8G8_SNORM; - case PIPE_FORMAT_R8G8B8_SNORM: - return fetch_R8G8B8_SNORM; - case PIPE_FORMAT_R8G8B8A8_SNORM: - return fetch_R8G8B8A8_SNORM; - - case PIPE_FORMAT_R8_SSCALED: - return fetch_R8_SSCALED; - case PIPE_FORMAT_R8G8_SSCALED: - return fetch_R8G8_SSCALED; - case PIPE_FORMAT_R8G8B8_SSCALED: - return fetch_R8G8B8_SSCALED; - case PIPE_FORMAT_R8G8B8A8_SSCALED: - return fetch_R8G8B8A8_SSCALED; - - case PIPE_FORMAT_A8R8G8B8_UNORM: - return fetch_A8R8G8B8_UNORM; - - - case PIPE_FORMAT_B8G8R8A8_UNORM: - return fetch_B8G8R8A8_UNORM; - - case 0: - return NULL; /* not sure why this is needed */ - - default: - /* This can get hit because draw-state-validation is too eager, - and can jump in here validating stuff before the state tracker has set - up everything. - */ - /* assert(0); */ - return NULL; - } -} - - -static void -transpose_4x4( float *out, const float *in ) -{ - /* This can be achieved in 12 sse instructions, plus the final - * stores I guess. This is probably a bit more than that - maybe - * 32 or so? - */ - out[0] = in[0]; out[1] = in[4]; out[2] = in[8]; out[3] = in[12]; - out[4] = in[1]; out[5] = in[5]; out[6] = in[9]; out[7] = in[13]; - out[8] = in[2]; out[9] = in[6]; out[10] = in[10]; out[11] = in[14]; - out[12] = in[3]; out[13] = in[7]; out[14] = in[11]; out[15] = in[15]; -} - - - -static void fetch_xyz_rgb( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - - assert(count <= 4); - -// debug_printf("%s\n", __FUNCTION__); - - /* loop over vertex attributes (vertex shader inputs) - */ - - for (i = 0; i < 4; i++) { - { - const float *in = (const float *)(src[0] + elts[i] * pitch[0]); - float *out = &machine->Inputs[0].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - - { - const float *in = (const float *)(src[1] + elts[i] * pitch[1]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - } -} - - - - -static void fetch_xyz_rgb_st( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - const unsigned *pitch = draw->vertex_fetch.pitch; - const ubyte **src = draw->vertex_fetch.src_ptr; - int i; - - assert(count <= 4); - - /* loop over vertex attributes (vertex shader inputs) - */ - - for (i = 0; i < 4; i++) { - { - const float *in = (const float *)(src[0] + elts[i] * pitch[0]); - float *out = &machine->Inputs[0].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - - { - const float *in = (const float *)(src[1] + elts[i] * pitch[1]); - float *out = &machine->Inputs[1].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = in[2]; - out[12] = 1.0f; - } - - { - const float *in = (const float *)(src[2] + elts[i] * pitch[2]); - float *out = &machine->Inputs[2].xyzw[0].f[i]; - out[0] = in[0]; - out[4] = in[1]; - out[8] = 0.0f; - out[12] = 1.0f; - } - } -} - - - - -/** - * Fetch vertex attributes for 'count' vertices. - */ -static void generic_vertex_fetch( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ) -{ - unsigned nr_attrs = draw->vertex_fetch.nr_attrs; - unsigned attr; - - assert(count <= 4); - -// debug_printf("%s %d\n", __FUNCTION__, count); - - /* loop over vertex attributes (vertex shader inputs) - */ - for (attr = 0; attr < nr_attrs; attr++) { - - const unsigned pitch = draw->vertex_fetch.pitch[attr]; - const ubyte *src = draw->vertex_fetch.src_ptr[attr]; - const fetch_func fetch = draw->vertex_fetch.fetch[attr]; - unsigned i; - float p[4][4]; - - - /* Fetch four attributes for four vertices. - * - * Could fetch directly into AOS format, but this is meant to be - * a prototype for an sse implementation, which would have - * difficulties doing that. - */ - for (i = 0; i < count; i++) - fetch( src + elts[i] * pitch, p[i] ); - - /* Be nice and zero out any missing vertices: - */ - for ( ; i < 4; i++) - p[i][0] = p[i][1] = p[i][2] = p[i][3] = 0; - - /* Transpose/swizzle into sse-friendly format. Currently - * assuming that all vertex shader inputs are float[4], but this - * isn't true -- if the vertex shader only wants tex0.xy, we - * could optimize for that. - * - * To do so fully without codegen would probably require an - * excessive number of fetch functions, but we could at least - * minimize the transpose step: - */ - transpose_4x4( (float *)&machine->Inputs[attr].xyzw[0].f[0], (float *)p ); - } -} - - - -void draw_update_vertex_fetch( struct draw_context *draw ) -{ - unsigned nr_attrs, i; - -// debug_printf("%s\n", __FUNCTION__); - - /* this may happend during context init */ - if (!draw->vertex_shader) - return; - - nr_attrs = draw->vertex_shader->info.num_inputs; - - for (i = 0; i < nr_attrs; i++) { - unsigned buf = draw->vertex_element[i].vertex_buffer_index; - enum pipe_format format = draw->vertex_element[i].src_format; - - draw->vertex_fetch.src_ptr[i] = (const ubyte *) draw->user.vbuffer[buf] + - draw->vertex_buffer[buf].buffer_offset + - draw->vertex_element[i].src_offset; - - draw->vertex_fetch.pitch[i] = draw->vertex_buffer[buf].pitch; - draw->vertex_fetch.fetch[i] = draw_get_fetch_func( format ); - } - - draw->vertex_fetch.nr_attrs = nr_attrs; - - draw->vertex_fetch.fetch_func = generic_vertex_fetch; - - switch (nr_attrs) { - case 2: - if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT) - draw->vertex_fetch.fetch_func = fetch_xyz_rgb; - break; - case 3: - if (draw->vertex_element[0].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[1].src_format == PIPE_FORMAT_R32G32B32_FLOAT && - draw->vertex_element[2].src_format == PIPE_FORMAT_R32G32_FLOAT) - draw->vertex_fetch.fetch_func = fetch_xyz_rgb_st; - break; - default: - break; - } - -} diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 5e27bc9ff0..c219a91156 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -55,7 +55,6 @@ static void vs_llvm_prepare( struct draw_vertex_shader *base, struct draw_context *draw ) { - draw_update_vertex_fetch( draw ); } diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 60f60a5b53..8e2d381f14 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -71,7 +71,6 @@ static void vs_sse_prepare( struct draw_vertex_shader *base, struct draw_context *draw ) { - draw_update_vertex_fetch( draw ); } -- cgit v1.2.3 From 6094e79f4e3350d123c7532b1c73faa60834a62d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 00:27:12 +0100 Subject: draw: remove dead data structures --- src/gallium/auxiliary/draw/draw_context.c | 20 ----------- src/gallium/auxiliary/draw/draw_private.h | 59 ------------------------------- 2 files changed, 79 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 9c3ae9bc0d..6012bc155e 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -83,16 +83,6 @@ struct draw_context *draw_create( void ) ASSIGN_4V( draw->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ draw->nr_planes = 6; - /* Statically allocate maximum sized vertices for the cache - could be cleverer... - */ - { - char *tmp = align_malloc(VS_QUEUE_LENGTH * MAX_VERTEX_ALLOCATION, 16); - if (!tmp) - goto fail; - - draw->vs.vertex_cache = tmp; - } - /* these defaults are oriented toward the needs of softpipe */ draw->wide_point_threshold = 1000000.0; /* infinity */ draw->wide_line_threshold = 1.0; @@ -162,10 +152,6 @@ void draw_destroy( struct draw_context *draw ) align_free(draw->machine.Outputs); tgsi_exec_machine_free_data(&draw->machine); - - if (draw->vs.vertex_cache) - align_free( draw->vs.vertex_cache ); /* Frees all the vertices. */ - /* Not so fast -- we're just borrowing this at the moment. * if (draw->render) @@ -254,8 +240,6 @@ draw_set_vertex_buffers(struct draw_context *draw, { assert(count <= PIPE_MAX_ATTRIBS); - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - memcpy(draw->vertex_buffer, buffers, count * sizeof(buffers[0])); draw->nr_vertex_buffers = count; } @@ -268,8 +252,6 @@ draw_set_vertex_elements(struct draw_context *draw, { assert(count <= PIPE_MAX_ATTRIBS); - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); - memcpy(draw->vertex_element, elements, count * sizeof(elements[0])); draw->nr_vertex_elements = count; } @@ -282,7 +264,6 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw, unsigned attr, const void *buffer) { - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); draw->user.vbuffer[attr] = buffer; } @@ -291,7 +272,6 @@ void draw_set_mapped_constant_buffer(struct draw_context *draw, const void *buffer) { - draw_do_flush( draw, DRAW_FLUSH_VERTEX_CACHE/*STATE_CHANGE*/ ); draw->user.constants = buffer; } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 93add257a3..78265c64a6 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -195,9 +195,6 @@ struct draw_context /* Support prototype passthrough path: */ struct { - unsigned prim; /* XXX: to be removed */ - unsigned hw_vertex_size; /* XXX: to be removed */ - struct { struct draw_pt_middle_end *opt[PT_MAX_MIDDLE]; } middle; @@ -272,54 +269,6 @@ struct draw_context /** TGSI program interpreter runtime state */ struct tgsi_exec_machine machine; - /* Vertex fetch internal state - */ - struct { - const ubyte *src_ptr[PIPE_MAX_ATTRIBS]; - unsigned pitch[PIPE_MAX_ATTRIBS]; - fetch_func fetch[PIPE_MAX_ATTRIBS]; - unsigned nr_attrs; - full_fetch_func fetch_func; - pt_fetch_func pt_fetch; - } vertex_fetch; - - /* Post-tnl vertex cache: - */ - struct { - unsigned referenced; /**< bitfield */ - - struct { - unsigned in; /* client array element */ - unsigned out; /* index in vs queue/array */ - } idx[VCACHE_SIZE + VCACHE_OVERFLOW]; - - unsigned overflow; - - /** To find space in the vertex cache: */ - struct vertex_header *(*get_vertex)( struct draw_context *draw, - unsigned i ); - } vcache; - - /* Vertex shader queue: - */ - struct { - unsigned elts[VS_QUEUE_LENGTH]; /**< index into the user's vertex arrays */ - char *vertex_cache; - unsigned queue_nr; - unsigned post_nr; - } vs; - - /* Prim pipeline queue: - */ - struct { - /* Need to queue up primitives until their vertices have been - * transformed by a vs queue flush. - */ - struct prim_header queue[PRIM_QUEUE_LENGTH]; - unsigned queue_nr; - } pq; - - /* This (and the tgsi_exec_machine struct) probably need to be moved somewhere private. */ struct gallivm_cpu_engine *engine; @@ -372,9 +321,6 @@ boolean draw_pt_arrays( struct draw_context *draw, void draw_pt_reset_vertex_ids( struct draw_context *draw ); -#define DRAW_FLUSH_SHADER_QUEUE 0x1 /* sized not to overflow, never raised */ -#define DRAW_FLUSH_PRIM_QUEUE 0x2 -#define DRAW_FLUSH_VERTEX_CACHE 0x4 #define DRAW_FLUSH_STATE_CHANGE 0x8 #define DRAW_FLUSH_BACKEND 0x10 @@ -416,10 +362,5 @@ dot4(const float *a, const float *b) return result; } -static INLINE struct vertex_header * -draw_header_from_block(char *block, int size, int num) -{ - return (struct vertex_header*)(block + num * size); -} #endif /* DRAW_PRIVATE_H */ -- cgit v1.2.3 From 251ebcc175d479dda8d0d5b64fc42f44e747197e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 00:29:25 +0100 Subject: draw: remove more dead data structures --- src/gallium/auxiliary/draw/draw_private.h | 25 ------------------------- 1 file changed, 25 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 78265c64a6..476b1184d4 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -128,31 +128,6 @@ struct draw_stage }; -#define PRIM_QUEUE_LENGTH 32 -#define VCACHE_SIZE 32 -#define VCACHE_OVERFLOW 4 -#define VS_QUEUE_LENGTH (VCACHE_SIZE + VCACHE_OVERFLOW + 1) /* can never fill up */ - - - -/* Internal function for vertex fetch. - */ -typedef void (*fetch_func)(const void *ptr, float *attrib); - -fetch_func draw_get_fetch_func( enum pipe_format format ); - - - -typedef void (*full_fetch_func)( struct draw_context *draw, - struct tgsi_exec_machine *machine, - const unsigned *elts, - unsigned count ); - -typedef void (*pt_fetch_func)( struct draw_context *draw, - float *out, - unsigned start, - unsigned count ); - struct vbuf_render; -- cgit v1.2.3 From 6d9132de04fc190fea56978849dfc427e5359912 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 00:31:06 +0100 Subject: draw: make draw_pt_fetch_emit use translate facility --- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 186 +++++++++--------------- src/gallium/auxiliary/draw/draw_vbuf.c | 4 +- 2 files changed, 72 insertions(+), 118 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index d696afa1aa..6e4fea460b 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -36,6 +36,7 @@ #include "draw/draw_vbuf.h" #include "draw/draw_vertex.h" #include "draw/draw_pt.h" +#include "translate/translate.h" /* The simplest 'middle end' in the new vertex code. * @@ -72,94 +73,15 @@ struct fetch_emit_middle_end { struct draw_pt_middle_end base; struct draw_context *draw; - - struct { - const ubyte *ptr; - unsigned pitch; - void (*fetch)( const void *from, float *attrib); - void (*emit)( const float *attrib, float **out ); - } fetch[PIPE_MAX_ATTRIBS]; - unsigned nr_fetch; - unsigned hw_vertex_size; -}; - - - -static void fetch_R32_FLOAT( const void *from, - float *attrib ) -{ - float *f = (float *) from; - attrib[0] = f[0]; - attrib[1] = 0.0; - attrib[2] = 0.0; - attrib[3] = 1.0; -} - - -static void emit_R32_FLOAT( const float *attrib, - float **out ) -{ - (*out)[0] = attrib[0]; - (*out) += 1; -} - -static void emit_R32G32_FLOAT( const float *attrib, - float **out ) -{ - (*out)[0] = attrib[0]; - (*out)[1] = attrib[1]; - (*out) += 2; -} - -static void emit_R32G32B32_FLOAT( const float *attrib, - float **out ) -{ - (*out)[0] = attrib[0]; - (*out)[1] = attrib[1]; - (*out)[2] = attrib[2]; - (*out) += 3; -} - -static void emit_R32G32B32A32_FLOAT( const float *attrib, - float **out ) -{ - (*out)[0] = attrib[0]; - (*out)[1] = attrib[1]; - (*out)[2] = attrib[2]; - (*out)[3] = attrib[3]; - (*out) += 4; -} + struct translate *translate; + /* Cache point size somewhere it's address won't change: + */ + float point_size; -/** - * General-purpose fetch from user's vertex arrays, emit to driver's - * vertex buffer. - * - * XXX this is totally temporary. - */ -static void -fetch_store_general( struct fetch_emit_middle_end *feme, - void *out_ptr, - const unsigned *fetch_elts, - unsigned count ) -{ - float *out = (float *)out_ptr; - uint i, j; +}; - for (i = 0; i < count; i++) { - unsigned elt = fetch_elts[i] & ~DRAW_PT_FLAG_MASK; - - for (j = 0; j < feme->nr_fetch; j++) { - float attrib[4]; - const ubyte *from = (feme->fetch[j].ptr + - feme->fetch[j].pitch * elt); - - feme->fetch[j].fetch( from, attrib ); - feme->fetch[j].emit( attrib, &out ); - } - } -} @@ -170,8 +92,9 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; struct draw_context *draw = feme->draw; const struct vertex_info *vinfo; - unsigned i; + unsigned i, dst_offset; boolean ok; + struct translate_key key; ok = draw->render->set_primitive( draw->render, @@ -186,51 +109,82 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, vinfo = draw->render->get_vertex_info(draw->render); - /* This is unique as it transforms straight from API vertex - * information to HW vertices. All other cases go through our - * intermediate float[4] format. + + /* Transform from API vertices to HW vertices, skipping the + * pipeline_vertex intermediate step. */ - for (i = 0; i < vinfo->num_attribs; i++) { - unsigned src_element = vinfo->src_index[i]; - unsigned src_buffer = draw->vertex_element[src_element].vertex_buffer_index; - - feme->fetch[i].ptr = ((const ubyte *)draw->user.vbuffer[src_buffer] + - draw->vertex_buffer[src_buffer].buffer_offset + - draw->vertex_element[src_element].src_offset); + dst_offset = 0; + memset(&key, 0, sizeof(key)); - feme->fetch[i].pitch = draw->vertex_buffer[src_buffer].pitch; - - feme->fetch[i].fetch = draw_get_fetch_func(draw->vertex_element[src_element].src_format); + for (i = 0; i < vinfo->num_attribs; i++) { + const struct pipe_vertex_element *src = &draw->vertex_element[vinfo->src_index[i]]; + unsigned emit_sz = 0; + unsigned input_format = src->src_format; + unsigned input_buffer = src->vertex_buffer_index; + unsigned input_offset = src->src_offset; + unsigned output_format; switch (vinfo->emit[i]) { case EMIT_4F: - feme->fetch[i].emit = emit_R32G32B32A32_FLOAT; + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); break; case EMIT_3F: - feme->fetch[i].emit = emit_R32G32B32_FLOAT; + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); break; case EMIT_2F: - feme->fetch[i].emit = emit_R32G32_FLOAT; + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); break; case EMIT_1F: - feme->fetch[i].emit = emit_R32_FLOAT; + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); break; case EMIT_1F_PSIZE: - feme->fetch[i].ptr = (const ubyte *)&feme->draw->rasterizer->point_size; - feme->fetch[i].pitch = 0; - feme->fetch[i].fetch = fetch_R32_FLOAT; - feme->fetch[i].emit = emit_R32_FLOAT; + input_format = PIPE_FORMAT_R32_FLOAT; + input_buffer = draw->nr_vertex_buffers; + input_offset = 0; + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); break; default: assert(0); - feme->fetch[i].emit = NULL; - break; + output_format = PIPE_FORMAT_NONE; + emit_sz = 0; + continue; } + + key.element[i].input_format = input_format; + key.element[i].input_buffer = input_buffer; + key.element[i].input_offset = input_offset; + key.element[i].output_format = output_format; + key.element[i].output_offset = dst_offset; + + dst_offset += emit_sz; } - feme->nr_fetch = vinfo->num_attribs; - feme->hw_vertex_size = vinfo->size * 4; + key.nr_elements = vinfo->num_attribs; + key.output_stride = vinfo->size * 4; + + /* Don't bother with caching at this stage: + */ + if (!feme->translate || + memcmp(&feme->translate->key, &key, sizeof(key)) != 0) + { + if (feme->translate) + feme->translate->release(feme->translate); + + feme->translate = translate_create( &key ); + + feme->translate->set_buffer(feme->translate, + draw->nr_vertex_buffers, + &feme->point_size, + 0); + } + + feme->point_size = draw->rasterizer->point_size; } @@ -252,7 +206,7 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, draw_do_flush( draw, DRAW_FLUSH_BACKEND ); hw_verts = draw->render->allocate_vertices( draw->render, - (ushort)feme->hw_vertex_size, + (ushort)feme->translate->key.output_stride, (ushort)fetch_count ); if (!hw_verts) { assert(0); @@ -262,10 +216,10 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, /* Single routine to fetch vertices and emit HW verts. */ - fetch_store_general( feme, - hw_verts, - fetch_elts, - fetch_count ); + feme->translate->run_elts( feme->translate, + fetch_elts, + fetch_count, + hw_verts ); /* XXX: Draw arrays path to avoid re-emitting index list again and * again. @@ -278,7 +232,7 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, */ draw->render->release_vertices( draw->render, hw_verts, - feme->hw_vertex_size, + feme->translate->key.output_stride, fetch_count ); } diff --git a/src/gallium/auxiliary/draw/draw_vbuf.c b/src/gallium/auxiliary/draw/draw_vbuf.c index dbbcf05c3c..30dceeb43d 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_vbuf.c @@ -439,8 +439,8 @@ vbuf_alloc_vertices( struct vbuf_stage *vbuf ) /* Allocate a new vertex buffer */ vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, - (ushort) vbuf->vertex_size, - (ushort) vbuf->max_vertices); + (ushort) vbuf->vertex_size, + (ushort) vbuf->max_vertices); vbuf->vertex_ptr = vbuf->vertices; } -- cgit v1.2.3 From 6494946db66b62d280e34a0486b83cca15f5b457 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 00:36:35 +0100 Subject: draw: remove draw_debug.c --- src/gallium/auxiliary/draw/Makefile | 1 - src/gallium/auxiliary/draw/SConscript | 1 - src/gallium/auxiliary/draw/draw_debug.c | 113 -------------------------------- 3 files changed, 115 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_debug.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 60bb9dfe88..45def9153c 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -12,7 +12,6 @@ C_SOURCES = \ draw_vs_llvm.c \ draw_context.c\ draw_cull.c \ - draw_debug.c \ draw_flatshade.c \ draw_offset.c \ draw_pt.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 246083a962..bff32d2c8b 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -8,7 +8,6 @@ draw = env.ConvenienceLibrary( 'draw_clip.c', 'draw_context.c', 'draw_cull.c', - 'draw_debug.c', 'draw_flatshade.c', 'draw_offset.c', 'draw_pstipple.c', diff --git a/src/gallium/auxiliary/draw/draw_debug.c b/src/gallium/auxiliary/draw/draw_debug.c deleted file mode 100644 index d6220b5f62..0000000000 --- a/src/gallium/auxiliary/draw/draw_debug.c +++ /dev/null @@ -1,113 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "draw_private.h" -#include "draw_context.h" - - - -static void -draw_prim_info(unsigned prim, unsigned *first, unsigned *incr) -{ - assert(prim >= PIPE_PRIM_POINTS); - assert(prim <= PIPE_PRIM_POLYGON); - - switch (prim) { - case PIPE_PRIM_POINTS: - *first = 1; - *incr = 1; - break; - case PIPE_PRIM_LINES: - *first = 2; - *incr = 2; - break; - case PIPE_PRIM_LINE_STRIP: - *first = 2; - *incr = 1; - break; - case PIPE_PRIM_LINE_LOOP: - *first = 2; - *incr = 1; - break; - case PIPE_PRIM_TRIANGLES: - *first = 3; - *incr = 3; - break; - case PIPE_PRIM_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - break; - case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: - *first = 3; - *incr = 1; - break; - case PIPE_PRIM_QUADS: - *first = 4; - *incr = 4; - break; - case PIPE_PRIM_QUAD_STRIP: - *first = 4; - *incr = 2; - break; - default: - assert(0); - *first = 1; - *incr = 1; - break; - } -} - - -unsigned -draw_trim_prim( unsigned mode, unsigned count ) -{ - unsigned length, first, incr; - - draw_prim_info( mode, &first, &incr ); - - if (count < first) - length = 0; - else - length = count - (count - first) % incr; - - return length; -} - - -boolean -draw_validate_prim( unsigned mode, unsigned count ) -{ - return (count > 0 && - count == draw_trim_prim( mode, count )); -} - -- cgit v1.2.3 From 6c38c600ff1212699e2e8e0f2928cd9e69559ac5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 00:37:51 +0100 Subject: draw: rename draw_vertex_shader.c -> draw_vs.c --- src/gallium/auxiliary/draw/Makefile | 2 +- src/gallium/auxiliary/draw/SConscript | 2 +- src/gallium/auxiliary/draw/draw_vertex_shader.c | 85 ------------------------- src/gallium/auxiliary/draw/draw_vs.c | 85 +++++++++++++++++++++++++ 4 files changed, 87 insertions(+), 87 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_vertex_shader.c create mode 100644 src/gallium/auxiliary/draw/draw_vs.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 45def9153c..9c41649883 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -7,6 +7,7 @@ C_SOURCES = \ draw_aaline.c \ draw_aapoint.c \ draw_clip.c \ + draw_vs.c \ draw_vs_exec.c \ draw_vs_sse.c \ draw_vs_llvm.c \ @@ -30,7 +31,6 @@ C_SOURCES = \ draw_validate.c \ draw_vbuf.c \ draw_vertex.c \ - draw_vertex_shader.c \ draw_wide_line.c \ draw_wide_point.c diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index bff32d2c8b..388d7879dd 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -26,7 +26,7 @@ draw = env.ConvenienceLibrary( 'draw_validate.c', 'draw_vbuf.c', 'draw_vertex.c', - 'draw_vertex_shader.c', + 'draw_vs.c', 'draw_vs_exec.c', 'draw_vs_llvm.c', 'draw_vs_sse.c', diff --git a/src/gallium/auxiliary/draw/draw_vertex_shader.c b/src/gallium/auxiliary/draw/draw_vertex_shader.c deleted file mode 100644 index 03fe00a951..0000000000 --- a/src/gallium/auxiliary/draw/draw_vertex_shader.c +++ /dev/null @@ -1,85 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" -#include "draw_context.h" -#include "draw_vs.h" - - - -struct draw_vertex_shader * -draw_create_vertex_shader(struct draw_context *draw, - const struct pipe_shader_state *shader) -{ - struct draw_vertex_shader *vs; - - vs = draw_create_vs_llvm( draw, shader ); - if (!vs) { - vs = draw_create_vs_sse( draw, shader ); - if (!vs) { - vs = draw_create_vs_exec( draw, shader ); - } - } - - assert(vs); - return vs; -} - - -void -draw_bind_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs) -{ - draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - - if (dvs) - { - draw->vertex_shader = dvs; - draw->num_vs_outputs = dvs->info.num_outputs; - dvs->prepare( dvs, draw ); - } - else { - draw->vertex_shader = NULL; - draw->num_vs_outputs = 0; - } -} - - -void -draw_delete_vertex_shader(struct draw_context *draw, - struct draw_vertex_shader *dvs) -{ - dvs->delete( dvs ); -} diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c new file mode 100644 index 0000000000..03fe00a951 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -0,0 +1,85 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + * Brian Paul + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" +#include "draw_context.h" +#include "draw_vs.h" + + + +struct draw_vertex_shader * +draw_create_vertex_shader(struct draw_context *draw, + const struct pipe_shader_state *shader) +{ + struct draw_vertex_shader *vs; + + vs = draw_create_vs_llvm( draw, shader ); + if (!vs) { + vs = draw_create_vs_sse( draw, shader ); + if (!vs) { + vs = draw_create_vs_exec( draw, shader ); + } + } + + assert(vs); + return vs; +} + + +void +draw_bind_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + if (dvs) + { + draw->vertex_shader = dvs; + draw->num_vs_outputs = dvs->info.num_outputs; + dvs->prepare( dvs, draw ); + } + else { + draw->vertex_shader = NULL; + draw->num_vs_outputs = 0; + } +} + + +void +draw_delete_vertex_shader(struct draw_context *draw, + struct draw_vertex_shader *dvs) +{ + dvs->delete( dvs ); +} -- cgit v1.2.3 From 1b411f894369f6a55c6f11cf650511eaa18a8510 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 00:39:38 +0100 Subject: draw: sort makefile sources --- src/gallium/auxiliary/draw/Makefile | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 9c41649883..0b0578a179 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -7,30 +7,30 @@ C_SOURCES = \ draw_aaline.c \ draw_aapoint.c \ draw_clip.c \ - draw_vs.c \ - draw_vs_exec.c \ - draw_vs_sse.c \ - draw_vs_llvm.c \ draw_context.c\ draw_cull.c \ draw_flatshade.c \ draw_offset.c \ + draw_pstipple.c \ draw_pt.c \ - draw_pt_vcache.c \ + draw_pt_elts.c \ + draw_pt_emit.c \ + draw_pt_fetch.c \ draw_pt_fetch_emit.c \ draw_pt_fetch_shade_pipeline.c \ - draw_pt_fetch.c \ - draw_pt_post_vs.c \ - draw_pt_emit.c \ draw_pt_pipeline.c \ - draw_pt_elts.c \ - draw_pstipple.c \ + draw_pt_post_vs.c \ + draw_pt_vcache.c \ draw_stipple.c \ draw_twoside.c \ draw_unfilled.c \ draw_validate.c \ draw_vbuf.c \ draw_vertex.c \ + draw_vs.c \ + draw_vs_exec.c \ + draw_vs_llvm.c \ + draw_vs_sse.c \ draw_wide_line.c \ draw_wide_point.c -- cgit v1.2.3 From c717f1fbe25f16ce6e607b0f7319ce74f9ae99b4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 00:43:58 +0100 Subject: softpipe: don't need to trim prims --- src/gallium/drivers/softpipe/sp_draw_arrays.c | 11 ----------- 1 file changed, 11 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 421509495a..778291dded 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -118,17 +118,6 @@ softpipe_draw_elements(struct pipe_context *pipe, struct draw_context *draw = sp->draw; unsigned i; - /* first, check that the primitive is not malformed. It is the - * state tracker's responsibility to do send only correctly formed - * primitives down. It currently isn't doing that though... - */ -#if 1 - count = draw_trim_prim( mode, count ); -#else - if (!draw_validate_prim( mode, count )) - assert(0); -#endif - sp->reduced_api_prim = reduced_prim[mode]; if (sp->dirty) -- cgit v1.2.3 From 43452886e2e33e33bdc57abe7e0b4af0abbbd2b1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 00:44:09 +0100 Subject: cell: don't need to trim prims --- src/gallium/drivers/cell/ppu/cell_draw_arrays.c | 11 ----------- 1 file changed, 11 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c index 36af5be5f0..6e08cf6fe8 100644 --- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c @@ -101,17 +101,6 @@ cell_draw_elements(struct pipe_context *pipe, struct draw_context *draw = sp->draw; unsigned i; - /* first, check that the primitive is not malformed. It is the - * state tracker's responsibility to do send only correctly formed - * primitives down. It currently isn't doing that though... - */ -#if 1 - count = draw_trim_prim( mode, count ); -#else - if (!draw_validate_prim( mode, count )) - assert(0); -#endif - if (sp->dirty) cell_update_derived( sp ); -- cgit v1.2.3 From 26831bdac594a11e51b6c4b09df78bb11444f5dd Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 00:45:05 +0100 Subject: draw: rename pipeline files to draw_pipe_* --- src/gallium/auxiliary/draw/Makefile | 32 +- src/gallium/auxiliary/draw/SConscript | 30 +- src/gallium/auxiliary/draw/draw_aaline.c | 859 ---------------------- src/gallium/auxiliary/draw/draw_aapoint.c | 846 --------------------- src/gallium/auxiliary/draw/draw_clip.c | 503 ------------- src/gallium/auxiliary/draw/draw_cull.c | 150 ---- src/gallium/auxiliary/draw/draw_flatshade.c | 248 ------- src/gallium/auxiliary/draw/draw_offset.c | 186 ----- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 859 ++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 846 +++++++++++++++++++++ src/gallium/auxiliary/draw/draw_pipe_clip.c | 503 +++++++++++++ src/gallium/auxiliary/draw/draw_pipe_cull.c | 150 ++++ src/gallium/auxiliary/draw/draw_pipe_flatshade.c | 248 +++++++ src/gallium/auxiliary/draw/draw_pipe_offset.c | 186 +++++ src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 746 +++++++++++++++++++ src/gallium/auxiliary/draw/draw_pipe_stipple.c | 239 ++++++ src/gallium/auxiliary/draw/draw_pipe_twoside.c | 203 +++++ src/gallium/auxiliary/draw/draw_pipe_unfilled.c | 206 ++++++ src/gallium/auxiliary/draw/draw_pipe_validate.c | 312 ++++++++ src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 529 +++++++++++++ src/gallium/auxiliary/draw/draw_pipe_wide_line.c | 190 +++++ src/gallium/auxiliary/draw/draw_pipe_wide_point.c | 281 +++++++ src/gallium/auxiliary/draw/draw_pipe_wide_prims.c | 366 +++++++++ src/gallium/auxiliary/draw/draw_pstipple.c | 746 ------------------- src/gallium/auxiliary/draw/draw_stipple.c | 239 ------ src/gallium/auxiliary/draw/draw_twoside.c | 203 ----- src/gallium/auxiliary/draw/draw_unfilled.c | 206 ------ src/gallium/auxiliary/draw/draw_validate.c | 312 -------- src/gallium/auxiliary/draw/draw_vbuf.c | 529 ------------- src/gallium/auxiliary/draw/draw_wide_line.c | 190 ----- src/gallium/auxiliary/draw/draw_wide_point.c | 281 ------- src/gallium/auxiliary/draw/draw_wide_prims.c | 366 --------- 32 files changed, 5895 insertions(+), 5895 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_aaline.c delete mode 100644 src/gallium/auxiliary/draw/draw_aapoint.c delete mode 100644 src/gallium/auxiliary/draw/draw_clip.c delete mode 100644 src/gallium/auxiliary/draw/draw_cull.c delete mode 100644 src/gallium/auxiliary/draw/draw_flatshade.c delete mode 100644 src/gallium/auxiliary/draw/draw_offset.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_aaline.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_aapoint.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_clip.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_cull.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_flatshade.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_offset.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_pstipple.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_stipple.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_twoside.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_unfilled.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_validate.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_vbuf.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_wide_line.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_wide_point.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe_wide_prims.c delete mode 100644 src/gallium/auxiliary/draw/draw_pstipple.c delete mode 100644 src/gallium/auxiliary/draw/draw_stipple.c delete mode 100644 src/gallium/auxiliary/draw/draw_twoside.c delete mode 100644 src/gallium/auxiliary/draw/draw_unfilled.c delete mode 100644 src/gallium/auxiliary/draw/draw_validate.c delete mode 100644 src/gallium/auxiliary/draw/draw_vbuf.c delete mode 100644 src/gallium/auxiliary/draw/draw_wide_line.c delete mode 100644 src/gallium/auxiliary/draw/draw_wide_point.c delete mode 100644 src/gallium/auxiliary/draw/draw_wide_prims.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 0b0578a179..62f46c6db1 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -4,14 +4,21 @@ include $(TOP)/configs/current LIBNAME = draw C_SOURCES = \ - draw_aaline.c \ - draw_aapoint.c \ - draw_clip.c \ - draw_context.c\ - draw_cull.c \ - draw_flatshade.c \ - draw_offset.c \ - draw_pstipple.c \ + draw_context.c \ + draw_pipe_aaline.c \ + draw_pipe_aapoint.c \ + draw_pipe_clip.c \ + draw_pipe_cull.c \ + draw_pipe_flatshade.c \ + draw_pipe_offset.c \ + draw_pipe_pstipple.c \ + draw_pipe_stipple.c \ + draw_pipe_twoside.c \ + draw_pipe_unfilled.c \ + draw_pipe_validate.c \ + draw_pipe_vbuf.c \ + draw_pipe_wide_line.c \ + draw_pipe_wide_point.c \ draw_pt.c \ draw_pt_elts.c \ draw_pt_emit.c \ @@ -21,18 +28,11 @@ C_SOURCES = \ draw_pt_pipeline.c \ draw_pt_post_vs.c \ draw_pt_vcache.c \ - draw_stipple.c \ - draw_twoside.c \ - draw_unfilled.c \ - draw_validate.c \ - draw_vbuf.c \ draw_vertex.c \ draw_vs.c \ draw_vs_exec.c \ draw_vs_llvm.c \ - draw_vs_sse.c \ - draw_wide_line.c \ - draw_wide_point.c + draw_vs_sse.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 388d7879dd..5fa35d3005 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -3,14 +3,22 @@ Import('*') draw = env.ConvenienceLibrary( target = 'draw', source = [ - 'draw_aaline.c', - 'draw_aapoint.c', - 'draw_clip.c', 'draw_context.c', - 'draw_cull.c', - 'draw_flatshade.c', - 'draw_offset.c', - 'draw_pstipple.c', + 'draw_pipe_aaline.c', + 'draw_pipe_aapoint.c', + 'draw_pipe_clip.c', + 'draw_pipe_cull.c', + 'draw_pipe_flatshade.c', + 'draw_pipe_offset.c', + 'draw_pipe_pstipple.c', + 'draw_pipe_stipple.c', + 'draw_pipe_twoside.c', + 'draw_pipe_unfilled.c', + 'draw_pipe_validate.c', + 'draw_pipe_vbuf.c', + 'draw_pipe_vertex.c', + 'draw_pipe_wide_line.c', + 'draw_pipe_wide_point.c', 'draw_pt.c', 'draw_pt_elts.c', 'draw_pt_emit.c', @@ -20,18 +28,10 @@ draw = env.ConvenienceLibrary( 'draw_pt_pipeline.c', 'draw_pt_post_vs.c', 'draw_pt_vcache.c', - 'draw_stipple.c', - 'draw_twoside.c', - 'draw_unfilled.c', - 'draw_validate.c', - 'draw_vbuf.c', - 'draw_vertex.c', 'draw_vs.c', 'draw_vs_exec.c', 'draw_vs_llvm.c', 'draw_vs_sse.c', - 'draw_wide_line.c', - 'draw_wide_point.c', ]) auxiliaries.insert(0, draw) diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c deleted file mode 100644 index e8d2a45102..0000000000 --- a/src/gallium/auxiliary/draw/draw_aaline.c +++ /dev/null @@ -1,859 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * AA line stage: AA lines are converted to texture mapped triangles. - * - * Authors: Brian Paul - */ - - -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" - -#include "tgsi/util/tgsi_transform.h" -#include "tgsi/util/tgsi_dump.h" - -#include "draw_context.h" -#include "draw_private.h" - - -/** - * Max texture level for the alpha texture used for antialiasing - */ -#define MAX_TEXTURE_LEVEL 5 /* 32 x 32 */ - - -/** - * Subclass of pipe_shader_state to carry extra fragment shader info. - */ -struct aaline_fragment_shader -{ - struct pipe_shader_state state; - void *driver_fs; - void *aaline_fs; - void *aapoint_fs; /* not yet */ - void *sprite_fs; /* not yet */ - uint sampler_unit; - int generic_attrib; /**< texcoord/generic used for texture */ -}; - - -/** - * Subclass of draw_stage - */ -struct aaline_stage -{ - struct draw_stage stage; - - float half_line_width; - - /** For AA lines, this is the vertex attrib slot for the new texcoords */ - uint tex_slot; - - void *sampler_cso; - struct pipe_texture *texture; - uint num_samplers; - uint num_textures; - - - /* - * Currently bound state - */ - struct aaline_fragment_shader *fs; - struct { - void *sampler[PIPE_MAX_SAMPLERS]; - struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; - } state; - - /* - * Driver interface/override functions - */ - void * (*driver_create_fs_state)(struct pipe_context *, - const struct pipe_shader_state *); - void (*driver_bind_fs_state)(struct pipe_context *, void *); - void (*driver_delete_fs_state)(struct pipe_context *, void *); - - void (*driver_bind_sampler_states)(struct pipe_context *, unsigned, - void **); - void (*driver_set_sampler_textures)(struct pipe_context *, unsigned, - struct pipe_texture **); - - struct pipe_context *pipe; -}; - - - -/** - * Subclass of tgsi_transform_context, used for transforming the - * user's fragment shader to add the special AA instructions. - */ -struct aa_transform_context { - struct tgsi_transform_context base; - uint tempsUsed; /**< bitmask */ - int colorOutput; /**< which output is the primary color */ - uint samplersUsed; /**< bitfield of samplers used */ - int freeSampler; /** an available sampler for the pstipple */ - int maxInput, maxGeneric; /**< max input index found */ - int colorTemp, texTemp; /**< temp registers */ - boolean firstInstruction; -}; - - -/** - * TGSI declaration transform callback. - * Look for a free sampler, a free input attrib, and two free temp regs. - */ -static void -aa_transform_decl(struct tgsi_transform_context *ctx, - struct tgsi_full_declaration *decl) -{ - struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; - - if (decl->Declaration.File == TGSI_FILE_OUTPUT && - decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR && - decl->Semantic.SemanticIndex == 0) { - aactx->colorOutput = decl->u.DeclarationRange.First; - } - else if (decl->Declaration.File == TGSI_FILE_SAMPLER) { - uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { - aactx->samplersUsed |= 1 << i; - } - } - else if (decl->Declaration.File == TGSI_FILE_INPUT) { - if ((int) decl->u.DeclarationRange.Last > aactx->maxInput) - aactx->maxInput = decl->u.DeclarationRange.Last; - if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && - (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) { - aactx->maxGeneric = decl->Semantic.SemanticIndex; - } - } - else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { - uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { - aactx->tempsUsed |= (1 << i); - } - } - - ctx->emit_declaration(ctx, decl); -} - - -/** - * Find the lowest zero bit in the given word, or -1 if bitfield is all ones. - */ -static int -free_bit(uint bitfield) -{ - int i; - for (i = 0; i < 32; i++) { - if ((bitfield & (1 << i)) == 0) - return i; - } - return -1; -} - - -/** - * TGSI instruction transform callback. - * Replace writes to result.color w/ a temp reg. - * Upon END instruction, insert texture sampling code for antialiasing. - */ -static void -aa_transform_inst(struct tgsi_transform_context *ctx, - struct tgsi_full_instruction *inst) -{ - struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; - - if (aactx->firstInstruction) { - /* emit our new declarations before the first instruction */ - - struct tgsi_full_declaration decl; - uint i; - - /* find free sampler */ - aactx->freeSampler = free_bit(aactx->samplersUsed); - if (aactx->freeSampler >= PIPE_MAX_SAMPLERS) - aactx->freeSampler = PIPE_MAX_SAMPLERS - 1; - - /* find two free temp regs */ - for (i = 0; i < 32; i++) { - if ((aactx->tempsUsed & (1 << i)) == 0) { - /* found a free temp */ - if (aactx->colorTemp < 0) - aactx->colorTemp = i; - else if (aactx->texTemp < 0) - aactx->texTemp = i; - else - break; - } - } - assert(aactx->colorTemp >= 0); - assert(aactx->texTemp >= 0); - - /* declare new generic input/texcoord */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; - decl.Semantic.SemanticIndex = aactx->maxGeneric + 1; - decl.Declaration.Interpolate = 1; - /* XXX this could be linear... */ - decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->maxInput + 1; - ctx->emit_declaration(ctx, &decl); - - /* declare new sampler */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_SAMPLER; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->freeSampler; - ctx->emit_declaration(ctx, &decl); - - /* declare new temp regs */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->texTemp; - ctx->emit_declaration(ctx, &decl); - - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->colorTemp; - ctx->emit_declaration(ctx, &decl); - - aactx->firstInstruction = FALSE; - } - - if (inst->Instruction.Opcode == TGSI_OPCODE_END && - aactx->colorOutput != -1) { - struct tgsi_full_instruction newInst; - - /* TEX */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_TEX; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = aactx->texTemp; - newInst.Instruction.NumSrcRegs = 2; - newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->maxInput + 1; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->freeSampler; - - ctx->emit_instruction(ctx, &newInst); - - /* MOV rgb */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_MOV; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ; - newInst.Instruction.NumSrcRegs = 1; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; - ctx->emit_instruction(ctx, &newInst); - - /* MUL alpha */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_MUL; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->texTemp; - ctx->emit_instruction(ctx, &newInst); - - /* END */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_END; - newInst.Instruction.NumDstRegs = 0; - newInst.Instruction.NumSrcRegs = 0; - ctx->emit_instruction(ctx, &newInst); - } - else { - /* Not an END instruction. - * Look for writes to result.color and replace with colorTemp reg. - */ - uint i; - - for (i = 0; i < inst->Instruction.NumDstRegs; i++) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - if (dst->DstRegister.File == TGSI_FILE_OUTPUT && - dst->DstRegister.Index == aactx->colorOutput) { - dst->DstRegister.File = TGSI_FILE_TEMPORARY; - dst->DstRegister.Index = aactx->colorTemp; - } - } - - ctx->emit_instruction(ctx, inst); - } -} - - -/** - * Generate the frag shader we'll use for drawing AA lines. - * This will be the user's shader plus some texture/modulate instructions. - */ -static void -generate_aaline_fs(struct aaline_stage *aaline) -{ - const struct pipe_shader_state *orig_fs = &aaline->fs->state; - //struct draw_context *draw = aaline->stage.draw; - struct pipe_shader_state aaline_fs; - struct aa_transform_context transform; - -#define MAX 1000 - - aaline_fs = *orig_fs; /* copy to init */ - aaline_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); - - memset(&transform, 0, sizeof(transform)); - transform.colorOutput = -1; - transform.maxInput = -1; - transform.maxGeneric = -1; - transform.colorTemp = -1; - transform.texTemp = -1; - transform.firstInstruction = TRUE; - transform.base.transform_instruction = aa_transform_inst; - transform.base.transform_declaration = aa_transform_decl; - - tgsi_transform_shader(orig_fs->tokens, - (struct tgsi_token *) aaline_fs.tokens, - MAX, &transform.base); - -#if 0 /* DEBUG */ - tgsi_dump(orig_fs->tokens, 0); - tgsi_dump(aaline_fs.tokens, 0); -#endif - - aaline->fs->sampler_unit = transform.freeSampler; - - aaline->fs->aaline_fs - = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); - - aaline->fs->generic_attrib = transform.maxGeneric + 1; -} - - -/** - * Create the texture map we'll use for antialiasing the lines. - */ -static void -aaline_create_texture(struct aaline_stage *aaline) -{ - struct pipe_context *pipe = aaline->pipe; - struct pipe_screen *screen = pipe->screen; - struct pipe_texture texTemp; - uint level; - - memset(&texTemp, 0, sizeof(texTemp)); - texTemp.target = PIPE_TEXTURE_2D; - texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */ - texTemp.last_level = MAX_TEXTURE_LEVEL; - texTemp.width[0] = 1 << MAX_TEXTURE_LEVEL; - texTemp.height[0] = 1 << MAX_TEXTURE_LEVEL; - texTemp.depth[0] = 1; - texTemp.cpp = 1; - - aaline->texture = screen->texture_create(screen, &texTemp); - - /* Fill in mipmap images. - * Basically each level is solid opaque, except for the outermost - * texels which are zero. Special case the 1x1 and 2x2 levels. - */ - for (level = 0; level <= MAX_TEXTURE_LEVEL; level++) { - struct pipe_surface *surface; - const uint size = aaline->texture->width[level]; - ubyte *data; - uint i, j; - - assert(aaline->texture->width[level] == aaline->texture->height[level]); - - surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0); - data = pipe_surface_map(surface); - - for (i = 0; i < size; i++) { - for (j = 0; j < size; j++) { - ubyte d; - if (size == 1) { - d = 255; - } - else if (size == 2) { - d = 200; /* tuneable */ - } - else if (i == 0 || j == 0 || i == size - 1 || j == size - 1) { - d = 0; - } - else { - d = 255; - } - data[i * surface->pitch + j] = d; - } - } - - /* unmap */ - pipe_surface_unmap(surface); - pipe_surface_reference(&surface, NULL); - pipe->texture_update(pipe, aaline->texture, 0, (1 << level)); - } -} - - -/** - * Create the sampler CSO that'll be used for antialiasing. - * By using a mipmapped texture, we don't have to generate a different - * texture image for each line size. - */ -static void -aaline_create_sampler(struct aaline_stage *aaline) -{ - struct pipe_sampler_state sampler; - struct pipe_context *pipe = aaline->pipe; - - memset(&sampler, 0, sizeof(sampler)); - sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_LINEAR; - sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; - sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; - sampler.normalized_coords = 1; - sampler.min_lod = 0.0f; - sampler.max_lod = MAX_TEXTURE_LEVEL; - - aaline->sampler_cso = pipe->create_sampler_state(pipe, &sampler); -} - - -/** - * When we're about to draw our first AA line in a batch, this function is - * called to tell the driver to bind our modified fragment shader. - */ -static void -bind_aaline_fragment_shader(struct aaline_stage *aaline) -{ - if (!aaline->fs->aaline_fs) { - generate_aaline_fs(aaline); - } - aaline->driver_bind_fs_state(aaline->pipe, aaline->fs->aaline_fs); -} - - - -static INLINE struct aaline_stage * -aaline_stage( struct draw_stage *stage ) -{ - return (struct aaline_stage *) stage; -} - - -static void -passthrough_point(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->point(stage->next, header); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} - - -/** - * Draw a wide line by drawing a quad, using geometry which will - * fullfill GL's antialiased line requirements. - */ -static void -aaline_line(struct draw_stage *stage, struct prim_header *header) -{ - const struct aaline_stage *aaline = aaline_stage(stage); - const float half_width = aaline->half_line_width; - struct prim_header tri; - struct vertex_header *v[8]; - uint texPos = aaline->tex_slot; - float *pos, *tex; - float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; - float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; - double a = atan2(dy, dx); - float c_a = (float) cos(a), s_a = (float) sin(a); - uint i; - - /* XXX the ends of lines aren't quite perfect yet, but probably passable */ - dx = 0.5F * half_width; - dy = half_width; - - /* allocate/dup new verts */ - for (i = 0; i < 8; i++) { - v[i] = dup_vert(stage, header->v[i/4], i); - } - - /* - * Quad strip for line from v0 to v1 (*=endpoints): - * - * 1 3 5 7 - * +---+---------------------+---+ - * | | - * | *v0 v1* | - * | | - * +---+---------------------+---+ - * 0 2 4 6 - */ - - /* new verts */ - pos = v[0]->data[0]; - pos[0] += (-dx * c_a - dy * s_a); - pos[1] += (-dx * s_a + dy * c_a); - - pos = v[1]->data[0]; - pos[0] += (-dx * c_a - -dy * s_a); - pos[1] += (-dx * s_a + -dy * c_a); - - pos = v[2]->data[0]; - pos[0] += ( dx * c_a - dy * s_a); - pos[1] += ( dx * s_a + dy * c_a); - - pos = v[3]->data[0]; - pos[0] += ( dx * c_a - -dy * s_a); - pos[1] += ( dx * s_a + -dy * c_a); - - pos = v[4]->data[0]; - pos[0] += (-dx * c_a - dy * s_a); - pos[1] += (-dx * s_a + dy * c_a); - - pos = v[5]->data[0]; - pos[0] += (-dx * c_a - -dy * s_a); - pos[1] += (-dx * s_a + -dy * c_a); - - pos = v[6]->data[0]; - pos[0] += ( dx * c_a - dy * s_a); - pos[1] += ( dx * s_a + dy * c_a); - - pos = v[7]->data[0]; - pos[0] += ( dx * c_a - -dy * s_a); - pos[1] += ( dx * s_a + -dy * c_a); - - /* new texcoords */ - tex = v[0]->data[texPos]; - ASSIGN_4V(tex, 0, 0, 0, 1); - - tex = v[1]->data[texPos]; - ASSIGN_4V(tex, 0, 1, 0, 1); - - tex = v[2]->data[texPos]; - ASSIGN_4V(tex, .5, 0, 0, 1); - - tex = v[3]->data[texPos]; - ASSIGN_4V(tex, .5, 1, 0, 1); - - tex = v[4]->data[texPos]; - ASSIGN_4V(tex, .5, 0, 0, 1); - - tex = v[5]->data[texPos]; - ASSIGN_4V(tex, .5, 1, 0, 1); - - tex = v[6]->data[texPos]; - ASSIGN_4V(tex, 1, 0, 0, 1); - - tex = v[7]->data[texPos]; - ASSIGN_4V(tex, 1, 1, 0, 1); - - /* emit 6 tris for the quad strip */ - tri.v[0] = v[2]; tri.v[1] = v[1]; tri.v[2] = v[0]; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v[3]; tri.v[1] = v[1]; tri.v[2] = v[2]; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v[4]; tri.v[1] = v[3]; tri.v[2] = v[2]; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v[5]; tri.v[1] = v[3]; tri.v[2] = v[4]; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v[6]; tri.v[1] = v[5]; tri.v[2] = v[4]; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v[7]; tri.v[1] = v[5]; tri.v[2] = v[6]; - stage->next->tri( stage->next, &tri ); -} - - -static void -aaline_first_line(struct draw_stage *stage, struct prim_header *header) -{ - auto struct aaline_stage *aaline = aaline_stage(stage); - struct draw_context *draw = stage->draw; - struct pipe_context *pipe = aaline->pipe; - uint num_samplers; - - assert(draw->rasterizer->line_smooth); - - if (draw->rasterizer->line_width <= 3.0) - aaline->half_line_width = 1.5f; - else - aaline->half_line_width = 0.5f * draw->rasterizer->line_width; - - /* - * Bind (generate) our fragprog, sampler and texture - */ - bind_aaline_fragment_shader(aaline); - - /* update vertex attrib info */ - aaline->tex_slot = draw->num_vs_outputs; - assert(aaline->tex_slot > 0); /* output[0] is vertex pos */ - - /* advertise the extra post-transformed vertex attribute */ - draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; - draw->extra_vp_outputs.semantic_index = aaline->fs->generic_attrib; - draw->extra_vp_outputs.slot = aaline->tex_slot; - - /* how many samplers? */ - /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ - num_samplers = MAX2(aaline->num_textures, aaline->num_samplers); - num_samplers = MAX2(num_samplers, aaline->fs->sampler_unit + 1); - - aaline->state.sampler[aaline->fs->sampler_unit] = aaline->sampler_cso; - pipe_texture_reference(&aaline->state.texture[aaline->fs->sampler_unit], - aaline->texture); - - aaline->driver_bind_sampler_states(pipe, num_samplers, aaline->state.sampler); - aaline->driver_set_sampler_textures(pipe, num_samplers, aaline->state.texture); - - /* now really draw first line */ - stage->line = aaline_line; - stage->line(stage, header); -} - - -static void -aaline_flush(struct draw_stage *stage, unsigned flags) -{ - struct draw_context *draw = stage->draw; - struct aaline_stage *aaline = aaline_stage(stage); - struct pipe_context *pipe = aaline->pipe; - - stage->line = aaline_first_line; - stage->next->flush( stage->next, flags ); - - /* restore original frag shader */ - aaline->driver_bind_fs_state(pipe, aaline->fs->driver_fs); - - /* XXX restore original texture, sampler state */ - aaline->driver_bind_sampler_states(pipe, aaline->num_samplers, - aaline->state.sampler); - aaline->driver_set_sampler_textures(pipe, aaline->num_textures, - aaline->state.texture); - - draw->extra_vp_outputs.slot = 0; -} - - -static void -aaline_reset_stipple_counter(struct draw_stage *stage) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void -aaline_destroy(struct draw_stage *stage) -{ - struct aaline_stage *aaline = aaline_stage(stage); - - aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso); - - pipe_texture_release(&aaline->texture); - - draw_free_temp_verts( stage ); - - FREE( stage ); -} - - -static struct aaline_stage * -draw_aaline_stage(struct draw_context *draw) -{ - struct aaline_stage *aaline = CALLOC_STRUCT(aaline_stage); - - draw_alloc_temp_verts( &aaline->stage, 8 ); - - aaline->stage.draw = draw; - aaline->stage.next = NULL; - aaline->stage.point = passthrough_point; - aaline->stage.line = aaline_first_line; - aaline->stage.tri = passthrough_tri; - aaline->stage.flush = aaline_flush; - aaline->stage.reset_stipple_counter = aaline_reset_stipple_counter; - aaline->stage.destroy = aaline_destroy; - - return aaline; -} - - -static struct aaline_stage * -aaline_stage_from_pipe(struct pipe_context *pipe) -{ - struct draw_context *draw = (struct draw_context *) pipe->draw; - return aaline_stage(draw->pipeline.aaline); -} - - -/** - * This function overrides the driver's create_fs_state() function and - * will typically be called by the state tracker. - */ -static void * -aaline_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *fs) -{ - struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); - struct aaline_fragment_shader *aafs = CALLOC_STRUCT(aaline_fragment_shader); - - if (aafs) { - aafs->state = *fs; - - /* pass-through */ - aafs->driver_fs = aaline->driver_create_fs_state(aaline->pipe, fs); - } - - return aafs; -} - - -static void -aaline_bind_fs_state(struct pipe_context *pipe, void *fs) -{ - struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); - struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs; - /* save current */ - aaline->fs = aafs; - /* pass-through */ - aaline->driver_bind_fs_state(aaline->pipe, - (aafs ? aafs->driver_fs : NULL)); -} - - -static void -aaline_delete_fs_state(struct pipe_context *pipe, void *fs) -{ - struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); - struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs; - /* pass-through */ - aaline->driver_delete_fs_state(aaline->pipe, aafs->driver_fs); - FREE(aafs); -} - - -static void -aaline_bind_sampler_states(struct pipe_context *pipe, - unsigned num, void **sampler) -{ - struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); - /* save current */ - memcpy(aaline->state.sampler, sampler, num * sizeof(void *)); - aaline->num_samplers = num; - /* pass-through */ - aaline->driver_bind_sampler_states(aaline->pipe, num, sampler); -} - - -static void -aaline_set_sampler_textures(struct pipe_context *pipe, - unsigned num, struct pipe_texture **texture) -{ - struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); - uint i; - - /* save current */ - for (i = 0; i < num; i++) { - pipe_texture_reference(&aaline->state.texture[i], texture[i]); - } - aaline->num_textures = num; - - /* pass-through */ - aaline->driver_set_sampler_textures(aaline->pipe, num, texture); -} - - -/** - * Called by drivers that want to install this AA line prim stage - * into the draw module's pipeline. This will not be used if the - * hardware has native support for AA lines. - */ -void -draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) -{ - struct aaline_stage *aaline; - - pipe->draw = (void *) draw; - - /* - * Create / install AA line drawing / prim stage - */ - aaline = draw_aaline_stage( draw ); - assert(aaline); - draw->pipeline.aaline = &aaline->stage; - - aaline->pipe = pipe; - - /* create special texture, sampler state */ - aaline_create_texture(aaline); - aaline_create_sampler(aaline); - - /* save original driver functions */ - aaline->driver_create_fs_state = pipe->create_fs_state; - aaline->driver_bind_fs_state = pipe->bind_fs_state; - aaline->driver_delete_fs_state = pipe->delete_fs_state; - - aaline->driver_bind_sampler_states = pipe->bind_sampler_states; - aaline->driver_set_sampler_textures = pipe->set_sampler_textures; - - /* override the driver's functions */ - pipe->create_fs_state = aaline_create_fs_state; - pipe->bind_fs_state = aaline_bind_fs_state; - pipe->delete_fs_state = aaline_delete_fs_state; - - pipe->bind_sampler_states = aaline_bind_sampler_states; - pipe->set_sampler_textures = aaline_set_sampler_textures; -} diff --git a/src/gallium/auxiliary/draw/draw_aapoint.c b/src/gallium/auxiliary/draw/draw_aapoint.c deleted file mode 100644 index e84d380e50..0000000000 --- a/src/gallium/auxiliary/draw/draw_aapoint.c +++ /dev/null @@ -1,846 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * AA point stage: AA points are converted to quads and rendered with a - * special fragment shader. Another approach would be to use a texture - * map image of a point, but experiments indicate the quality isn't nearly - * as good as this approach. - * - * Note: this looks a lot like draw_aaline.c but there's actually little - * if any code that can be shared. - * - * Authors: Brian Paul - */ - - -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" - -#include "tgsi/util/tgsi_transform.h" -#include "tgsi/util/tgsi_dump.h" - -#include "draw_context.h" -#include "draw_vs.h" - - -/* - * Enabling NORMALIZE might give _slightly_ better results. - * Basically, it controls whether we compute distance as d=sqrt(x*x+y*y) or - * d=x*x+y*y. Since we're working with a unit circle, the later seems - * close enough and saves some costly instructions. - */ -#define NORMALIZE 0 - - -/** - * Subclass of pipe_shader_state to carry extra fragment shader info. - */ -struct aapoint_fragment_shader -{ - struct pipe_shader_state state; - void *driver_fs; /**< the regular shader */ - void *aapoint_fs; /**< the aa point-augmented shader */ - int generic_attrib; /**< The generic input attrib/texcoord we'll use */ -}; - - -/** - * Subclass of draw_stage - */ -struct aapoint_stage -{ - struct draw_stage stage; - - int psize_slot; - float radius; - - /** this is the vertex attrib slot for the new texcoords */ - uint tex_slot; - - /* - * Currently bound state - */ - struct aapoint_fragment_shader *fs; - - /* - * Driver interface/override functions - */ - void * (*driver_create_fs_state)(struct pipe_context *, - const struct pipe_shader_state *); - void (*driver_bind_fs_state)(struct pipe_context *, void *); - void (*driver_delete_fs_state)(struct pipe_context *, void *); - - struct pipe_context *pipe; -}; - - - -/** - * Subclass of tgsi_transform_context, used for transforming the - * user's fragment shader to add the special AA instructions. - */ -struct aa_transform_context { - struct tgsi_transform_context base; - uint tempsUsed; /**< bitmask */ - int colorOutput; /**< which output is the primary color */ - int maxInput, maxGeneric; /**< max input index found */ - int tmp0, colorTemp; /**< temp registers */ - boolean firstInstruction; -}; - - -/** - * TGSI declaration transform callback. - * Look for two free temp regs and available input reg for new texcoords. - */ -static void -aa_transform_decl(struct tgsi_transform_context *ctx, - struct tgsi_full_declaration *decl) -{ - struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; - - if (decl->Declaration.File == TGSI_FILE_OUTPUT && - decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR && - decl->Semantic.SemanticIndex == 0) { - aactx->colorOutput = decl->u.DeclarationRange.First; - } - else if (decl->Declaration.File == TGSI_FILE_INPUT) { - if ((int) decl->u.DeclarationRange.Last > aactx->maxInput) - aactx->maxInput = decl->u.DeclarationRange.Last; - if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && - (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) { - aactx->maxGeneric = decl->Semantic.SemanticIndex; - } - } - else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { - uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { - aactx->tempsUsed |= (1 << i); - } - } - - ctx->emit_declaration(ctx, decl); -} - - -/** - * TGSI instruction transform callback. - * Replace writes to result.color w/ a temp reg. - * Upon END instruction, insert texture sampling code for antialiasing. - */ -static void -aa_transform_inst(struct tgsi_transform_context *ctx, - struct tgsi_full_instruction *inst) -{ - struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; - struct tgsi_full_instruction newInst; - - if (aactx->firstInstruction) { - /* emit our new declarations before the first instruction */ - - struct tgsi_full_declaration decl; - const int texInput = aactx->maxInput + 1; - int tmp0; - uint i; - - /* find two free temp regs */ - for (i = 0; i < 32; i++) { - if ((aactx->tempsUsed & (1 << i)) == 0) { - /* found a free temp */ - if (aactx->tmp0 < 0) - aactx->tmp0 = i; - else if (aactx->colorTemp < 0) - aactx->colorTemp = i; - else - break; - } - } - - assert(aactx->colorTemp != aactx->tmp0); - - tmp0 = aactx->tmp0; - - /* declare new generic input/texcoord */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; - decl.Semantic.SemanticIndex = aactx->maxGeneric + 1; - decl.Declaration.Interpolate = 1; - /* XXX this could be linear... */ - decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = texInput; - ctx->emit_declaration(ctx, &decl); - - /* declare new temp regs */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = tmp0; - ctx->emit_declaration(ctx, &decl); - - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->colorTemp; - ctx->emit_declaration(ctx, &decl); - - aactx->firstInstruction = FALSE; - - - /* - * Emit code to compute fragment coverage, kill if outside point radius - * - * Temp reg0 usage: - * t0.x = distance of fragment from center point - * t0.y = boolean, is t0.x > 1.0, also misc temp usage - * t0.z = temporary for computing 1/(1-k) value - * t0.w = final coverage value - */ - - /* MUL t0.xy, tex, tex; # compute x^2, y^2 */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_MUL; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XY; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; - ctx->emit_instruction(ctx, &newInst); - - /* ADD t0.x, t0.x, t0.y; # x^2 + y^2 */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_ADD; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; - ctx->emit_instruction(ctx, &newInst); - -#if NORMALIZE /* OPTIONAL normalization of length */ - /* RSQ t0.x, t0.x; */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_RSQ; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; - newInst.Instruction.NumSrcRegs = 1; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - ctx->emit_instruction(ctx, &newInst); - - /* RCP t0.x, t0.x; */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_RCP; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; - newInst.Instruction.NumSrcRegs = 1; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - ctx->emit_instruction(ctx, &newInst); -#endif - - /* SGT t0.y, t0.xxxx, t0.wwww; # bool b = d > 1 (NOTE t0.w == 1) */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_SGT; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; - ctx->emit_instruction(ctx, &newInst); - - /* KILP -t0.yyyy; # if b, KILL */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_KILP; - newInst.Instruction.NumDstRegs = 0; - newInst.Instruction.NumSrcRegs = 1; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; - newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; - ctx->emit_instruction(ctx, &newInst); - - - /* compute coverage factor = (1-d)/(1-k) */ - - /* SUB t0.z, tex.w, tex.z; # m = 1 - k */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_SUB; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z; - ctx->emit_instruction(ctx, &newInst); - - /* RCP t0.z, t0.z; # t0.z = 1 / m */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_RCP; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; - newInst.Instruction.NumSrcRegs = 1; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Z; - ctx->emit_instruction(ctx, &newInst); - - /* SUB t0.y, 1, t0.x; # d = 1 - d */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_SUB; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; - ctx->emit_instruction(ctx, &newInst); - - /* MUL t0.w, t0.y, t0.z; # coverage = d * m */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_MUL; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Z; - ctx->emit_instruction(ctx, &newInst); - - /* SLE t0.y, t0.x, tex.z; # bool b = distance <= k */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_SLE; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Z; - ctx->emit_instruction(ctx, &newInst); - - /* CMP t0.w, -t0.y, tex.w, t0.w; - * # if -t0.y < 0 then - * t0.w = 1 - * else - * t0.w = t0.w - */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_CMP; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = tmp0; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; - newInst.Instruction.NumSrcRegs = 3; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; - newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; - newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[2].SrcRegister.Index = tmp0; - newInst.FullSrcRegisters[2].SrcRegister.SwizzleX = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[2].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[2].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; - newInst.FullSrcRegisters[2].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; - ctx->emit_instruction(ctx, &newInst); - - } - - if (inst->Instruction.Opcode == TGSI_OPCODE_END) { - /* add alpha modulation code at tail of program */ - - /* MOV result.color.xyz, colorTemp; */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_MOV; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ; - newInst.Instruction.NumSrcRegs = 1; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; - ctx->emit_instruction(ctx, &newInst); - - /* MUL result.color.w, colorTemp, tmp0.w; */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_MUL; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; - newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; - newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->tmp0; - ctx->emit_instruction(ctx, &newInst); - } - else { - /* Not an END instruction. - * Look for writes to result.color and replace with colorTemp reg. - */ - uint i; - - for (i = 0; i < inst->Instruction.NumDstRegs; i++) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - if (dst->DstRegister.File == TGSI_FILE_OUTPUT && - dst->DstRegister.Index == aactx->colorOutput) { - dst->DstRegister.File = TGSI_FILE_TEMPORARY; - dst->DstRegister.Index = aactx->colorTemp; - } - } - } - - ctx->emit_instruction(ctx, inst); -} - - -/** - * Generate the frag shader we'll use for drawing AA lines. - * This will be the user's shader plus some texture/modulate instructions. - */ -static void -generate_aapoint_fs(struct aapoint_stage *aapoint) -{ - const struct pipe_shader_state *orig_fs = &aapoint->fs->state; - struct pipe_shader_state aapoint_fs; - struct aa_transform_context transform; - -#define MAX 1000 - - aapoint_fs = *orig_fs; /* copy to init */ - aapoint_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); - - memset(&transform, 0, sizeof(transform)); - transform.colorOutput = -1; - transform.maxInput = -1; - transform.maxGeneric = -1; - transform.colorTemp = -1; - transform.tmp0 = -1; - transform.firstInstruction = TRUE; - transform.base.transform_instruction = aa_transform_inst; - transform.base.transform_declaration = aa_transform_decl; - - tgsi_transform_shader(orig_fs->tokens, - (struct tgsi_token *) aapoint_fs.tokens, - MAX, &transform.base); - -#if 0 /* DEBUG */ - printf("draw_aapoint, orig shader:\n"); - tgsi_dump(orig_fs->tokens, 0); - printf("draw_aapoint, new shader:\n"); - tgsi_dump(aapoint_fs.tokens, 0); -#endif - - aapoint->fs->aapoint_fs - = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs); - - aapoint->fs->generic_attrib = transform.maxGeneric + 1; -} - - -/** - * When we're about to draw our first AA line in a batch, this function is - * called to tell the driver to bind our modified fragment shader. - */ -static void -bind_aapoint_fragment_shader(struct aapoint_stage *aapoint) -{ - if (!aapoint->fs->aapoint_fs) { - generate_aapoint_fs(aapoint); - } - aapoint->driver_bind_fs_state(aapoint->pipe, aapoint->fs->aapoint_fs); -} - - - -static INLINE struct aapoint_stage * -aapoint_stage( struct draw_stage *stage ) -{ - return (struct aapoint_stage *) stage; -} - - -static void -passthrough_line(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->line(stage->next, header); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} - - -/** - * Draw an AA point by drawing a quad. - */ -static void -aapoint_point(struct draw_stage *stage, struct prim_header *header) -{ - const struct aapoint_stage *aapoint = aapoint_stage(stage); - struct prim_header tri; - struct vertex_header *v[4]; - uint texPos = aapoint->tex_slot; - float radius, *pos, *tex; - uint i; - float k; - - if (aapoint->psize_slot >= 0) { - radius = 0.5f * header->v[0]->data[aapoint->psize_slot][0]; - } - else { - radius = aapoint->radius; - } - - /* - * Note: the texcoords (generic attrib, really) we use are special: - * The S and T components simply vary from -1 to +1. - * The R component is k, below. - * The Q component is 1.0 and will used as a handy constant in the - * fragment shader. - */ - - /* - * k is the threshold distance from the point's center at which - * we begin alpha attenuation (the coverage value). - * Operating within a unit circle, we'll compute the fragment's - * distance 'd' from the center point using the texcoords. - * IF d > 1.0 THEN - * KILL fragment - * ELSE IF d > k THEN - * compute coverage in [0,1] proportional to d in [k, 1]. - * ELSE - * coverage = 1.0; // full coverage - * ENDIF - * - * Note: the ELSEIF and ELSE clauses are actually implemented with CMP to - * avoid using IF/ELSE/ENDIF TGSI opcodes. - */ - -#if !NORMALIZE - k = 1.0f / radius; - k = 1.0f - 2.0f * k + k * k; -#else - k = 1.0f - 1.0f / radius; -#endif - - /* allocate/dup new verts */ - for (i = 0; i < 4; i++) { - v[i] = dup_vert(stage, header->v[0], i); - } - - /* new verts */ - pos = v[0]->data[0]; - pos[0] -= radius; - pos[1] -= radius; - - pos = v[1]->data[0]; - pos[0] += radius; - pos[1] -= radius; - - pos = v[2]->data[0]; - pos[0] += radius; - pos[1] += radius; - - pos = v[3]->data[0]; - pos[0] -= radius; - pos[1] += radius; - - /* new texcoords */ - tex = v[0]->data[texPos]; - ASSIGN_4V(tex, -1, -1, k, 1); - - tex = v[1]->data[texPos]; - ASSIGN_4V(tex, 1, -1, k, 1); - - tex = v[2]->data[texPos]; - ASSIGN_4V(tex, 1, 1, k, 1); - - tex = v[3]->data[texPos]; - ASSIGN_4V(tex, -1, 1, k, 1); - - /* emit 2 tris for the quad strip */ - tri.v[0] = v[0]; - tri.v[1] = v[1]; - tri.v[2] = v[2]; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v[0]; - tri.v[1] = v[2]; - tri.v[2] = v[3]; - stage->next->tri( stage->next, &tri ); -} - - -static void -aapoint_first_point(struct draw_stage *stage, struct prim_header *header) -{ - auto struct aapoint_stage *aapoint = aapoint_stage(stage); - struct draw_context *draw = stage->draw; - - assert(draw->rasterizer->point_smooth); - - if (draw->rasterizer->point_size <= 2.0) - aapoint->radius = 1.0; - else - aapoint->radius = 0.5f * draw->rasterizer->point_size; - - /* - * Bind (generate) our fragprog. - */ - bind_aapoint_fragment_shader(aapoint); - - /* update vertex attrib info */ - aapoint->tex_slot = draw->num_vs_outputs; - assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */ - - draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; - draw->extra_vp_outputs.semantic_index = aapoint->fs->generic_attrib; - draw->extra_vp_outputs.slot = aapoint->tex_slot; - - /* find psize slot in post-transform vertex */ - aapoint->psize_slot = -1; - if (draw->rasterizer->point_size_per_vertex) { - /* find PSIZ vertex output */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i; - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { - aapoint->psize_slot = i; - break; - } - } - } - - /* now really draw first line */ - stage->point = aapoint_point; - stage->point(stage, header); -} - - -static void -aapoint_flush(struct draw_stage *stage, unsigned flags) -{ - struct draw_context *draw = stage->draw; - struct aapoint_stage *aapoint = aapoint_stage(stage); - struct pipe_context *pipe = aapoint->pipe; - - stage->point = aapoint_first_point; - stage->next->flush( stage->next, flags ); - - /* restore original frag shader */ - aapoint->driver_bind_fs_state(pipe, aapoint->fs->driver_fs); - - draw->extra_vp_outputs.slot = 0; -} - - -static void -aapoint_reset_stipple_counter(struct draw_stage *stage) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void -aapoint_destroy(struct draw_stage *stage) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -static struct aapoint_stage * -draw_aapoint_stage(struct draw_context *draw) -{ - struct aapoint_stage *aapoint = CALLOC_STRUCT(aapoint_stage); - - draw_alloc_temp_verts( &aapoint->stage, 4 ); - - aapoint->stage.draw = draw; - aapoint->stage.next = NULL; - aapoint->stage.point = aapoint_first_point; - aapoint->stage.line = passthrough_line; - aapoint->stage.tri = passthrough_tri; - aapoint->stage.flush = aapoint_flush; - aapoint->stage.reset_stipple_counter = aapoint_reset_stipple_counter; - aapoint->stage.destroy = aapoint_destroy; - - return aapoint; -} - - -static struct aapoint_stage * -aapoint_stage_from_pipe(struct pipe_context *pipe) -{ - struct draw_context *draw = (struct draw_context *) pipe->draw; - return aapoint_stage(draw->pipeline.aapoint); -} - - -/** - * This function overrides the driver's create_fs_state() function and - * will typically be called by the state tracker. - */ -static void * -aapoint_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *fs) -{ - struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); - struct aapoint_fragment_shader *aafs = CALLOC_STRUCT(aapoint_fragment_shader); - - if (aafs) { - aafs->state = *fs; - - /* pass-through */ - aafs->driver_fs = aapoint->driver_create_fs_state(aapoint->pipe, fs); - } - - return aafs; -} - - -static void -aapoint_bind_fs_state(struct pipe_context *pipe, void *fs) -{ - struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); - struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs; - /* save current */ - aapoint->fs = aafs; - /* pass-through */ - aapoint->driver_bind_fs_state(aapoint->pipe, - (aafs ? aafs->driver_fs : NULL)); -} - - -static void -aapoint_delete_fs_state(struct pipe_context *pipe, void *fs) -{ - struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); - struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs; - /* pass-through */ - aapoint->driver_delete_fs_state(aapoint->pipe, aafs->driver_fs); - FREE(aafs); -} - - -/** - * Called by drivers that want to install this AA point prim stage - * into the draw module's pipeline. This will not be used if the - * hardware has native support for AA points. - */ -void -draw_install_aapoint_stage(struct draw_context *draw, - struct pipe_context *pipe) -{ - struct aapoint_stage *aapoint; - - pipe->draw = (void *) draw; - - /* - * Create / install AA point drawing / prim stage - */ - aapoint = draw_aapoint_stage( draw ); - assert(aapoint); - draw->pipeline.aapoint = &aapoint->stage; - - aapoint->pipe = pipe; - - /* save original driver functions */ - aapoint->driver_create_fs_state = pipe->create_fs_state; - aapoint->driver_bind_fs_state = pipe->bind_fs_state; - aapoint->driver_delete_fs_state = pipe->delete_fs_state; - - /* override the driver's functions */ - pipe->create_fs_state = aapoint_create_fs_state; - pipe->bind_fs_state = aapoint_bind_fs_state; - pipe->delete_fs_state = aapoint_delete_fs_state; -} diff --git a/src/gallium/auxiliary/draw/draw_clip.c b/src/gallium/auxiliary/draw/draw_clip.c deleted file mode 100644 index 0ac3a240e5..0000000000 --- a/src/gallium/auxiliary/draw/draw_clip.c +++ /dev/null @@ -1,503 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Clipping stage - * - * \author Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" - -#include "draw_context.h" -#include "draw_vs.h" - - -#ifndef IS_NEGATIVE -#define IS_NEGATIVE(X) ((X) < 0.0) -#endif - -#ifndef DIFFERENT_SIGNS -#define DIFFERENT_SIGNS(x, y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F) -#endif - -#ifndef MAX_CLIPPED_VERTICES -#define MAX_CLIPPED_VERTICES ((2 * (6 + PIPE_MAX_CLIP_PLANES))+1) -#endif - - - -struct clipper { - struct draw_stage stage; /**< base class */ - - /* Basically duplicate some of the flatshading logic here: - */ - boolean flat; - uint num_color_attribs; - uint color_attribs[4]; /* front/back primary/secondary colors */ - - float (*plane)[4]; -}; - - -/* This is a bit confusing: - */ -static INLINE struct clipper *clipper_stage( struct draw_stage *stage ) -{ - return (struct clipper *)stage; -} - - -#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) - - -/* All attributes are float[4], so this is easy: - */ -static void interp_attr( float *fdst, - float t, - const float *fin, - const float *fout ) -{ - fdst[0] = LINTERP( t, fout[0], fin[0] ); - fdst[1] = LINTERP( t, fout[1], fin[1] ); - fdst[2] = LINTERP( t, fout[2], fin[2] ); - fdst[3] = LINTERP( t, fout[3], fin[3] ); -} - -static void copy_colors( struct draw_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - const struct clipper *clipper = clipper_stage(stage); - uint i; - for (i = 0; i < clipper->num_color_attribs; i++) { - const uint attr = clipper->color_attribs[i]; - COPY_4FV(dst->data[attr], src->data[attr]); - } -} - - - -/* Interpolate between two vertices to produce a third. - */ -static void interp( const struct clipper *clip, - struct vertex_header *dst, - float t, - const struct vertex_header *out, - const struct vertex_header *in ) -{ - const unsigned nr_attrs = clip->stage.draw->num_vs_outputs; - unsigned j; - - /* Vertex header. - */ - { - dst->clipmask = 0; - dst->edgeflag = 0; - dst->pad = 0; - dst->vertex_id = UNDEFINED_VERTEX_ID; - } - - /* Clip coordinates: interpolate normally - */ - { - interp_attr(dst->clip, t, in->clip, out->clip); - } - - /* Do the projective divide and insert window coordinates: - */ - { - const float *pos = dst->clip; - const float *scale = clip->stage.draw->viewport.scale; - const float *trans = clip->stage.draw->viewport.translate; - const float oow = 1.0f / pos[3]; - - dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; - dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; - dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; - dst->data[0][3] = oow; - } - - /* Other attributes - * Note: start at 1 to skip winpos (data[0]) since we just computed - * it above. - */ - for (j = 1; j < nr_attrs; j++) { - interp_attr(dst->data[j], t, in->data[j], out->data[j]); - } -} - - -static void emit_poly( struct draw_stage *stage, - struct vertex_header **inlist, - unsigned n, - const struct prim_header *origPrim) -{ - struct prim_header header; - unsigned i; - - /* later stages may need the determinant, but only the sign matters */ - header.det = origPrim->det; - - for (i = 2; i < n; i++) { - header.v[0] = inlist[i-1]; - header.v[1] = inlist[i]; - header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ - - { - unsigned tmp1 = header.v[1]->edgeflag; - unsigned tmp2 = header.v[2]->edgeflag; - - if (i != n-1) header.v[1]->edgeflag = 0; - if (i != 2) header.v[2]->edgeflag = 0; - - header.edgeflags = ((header.v[0]->edgeflag << 0) | - (header.v[1]->edgeflag << 1) | - (header.v[2]->edgeflag << 2)); - - if (0) { - const struct draw_vertex_shader *vs = stage->draw->vertex_shader; - uint j, k; - debug_printf("Clipped tri:\n"); - for (j = 0; j < 3; j++) { - for (k = 0; k < vs->info.num_outputs; k++) { - debug_printf(" Vert %d: Attr %d: %f %f %f %f\n", j, k, - header.v[j]->data[k][0], - header.v[j]->data[k][1], - header.v[j]->data[k][2], - header.v[j]->data[k][3]); - } - } - } - - stage->next->tri( stage->next, &header ); - - header.v[1]->edgeflag = tmp1; - header.v[2]->edgeflag = tmp2; - } - } -} - - - - -/* Clip a triangle against the viewport and user clip planes. - */ -static void -do_clip_tri( struct draw_stage *stage, - struct prim_header *header, - unsigned clipmask ) -{ - struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *a[MAX_CLIPPED_VERTICES]; - struct vertex_header *b[MAX_CLIPPED_VERTICES]; - struct vertex_header **inlist = a; - struct vertex_header **outlist = b; - unsigned tmpnr = 0; - unsigned n = 3; - unsigned i; - - inlist[0] = header->v[0]; - inlist[1] = header->v[1]; - inlist[2] = header->v[2]; - - while (clipmask && n >= 3) { - const unsigned plane_idx = ffs(clipmask)-1; - const float *plane = clipper->plane[plane_idx]; - struct vertex_header *vert_prev = inlist[0]; - float dp_prev = dot4( vert_prev->clip, plane ); - unsigned outcount = 0; - - clipmask &= ~(1<clip, plane ); - - if (!IS_NEGATIVE(dp_prev)) { - outlist[outcount++] = vert_prev; - } - - if (DIFFERENT_SIGNS(dp, dp_prev)) { - struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; - outlist[outcount++] = new_vert; - - if (IS_NEGATIVE(dp)) { - /* Going out of bounds. Avoid division by zero as we - * know dp != dp_prev from DIFFERENT_SIGNS, above. - */ - float t = dp / (dp - dp_prev); - interp( clipper, new_vert, t, vert, vert_prev ); - - /* Force edgeflag true in this case: - */ - new_vert->edgeflag = 1; - } else { - /* Coming back in. - */ - float t = dp_prev / (dp_prev - dp); - interp( clipper, new_vert, t, vert_prev, vert ); - - /* Copy starting vert's edgeflag: - */ - new_vert->edgeflag = vert_prev->edgeflag; - } - } - - vert_prev = vert; - dp_prev = dp; - } - - { - struct vertex_header **tmp = inlist; - inlist = outlist; - outlist = tmp; - n = outcount; - } - } - - /* If flat-shading, copy color to new provoking vertex. - */ - if (clipper->flat && inlist[0] != header->v[2]) { - if (1) { - inlist[0] = dup_vert(stage, inlist[0], tmpnr++); - } - - copy_colors(stage, inlist[0], header->v[2]); - } - - - - /* Emit the polygon as triangles to the setup stage: - */ - if (n >= 3) - emit_poly( stage, inlist, n, header ); -} - - -/* Clip a line against the viewport and user clip planes. - */ -static void -do_clip_line( struct draw_stage *stage, - struct prim_header *header, - unsigned clipmask ) -{ - const struct clipper *clipper = clipper_stage( stage ); - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - const float *pos0 = v0->clip; - const float *pos1 = v1->clip; - float t0 = 0.0F; - float t1 = 0.0F; - struct prim_header newprim; - - while (clipmask) { - const unsigned plane_idx = ffs(clipmask)-1; - const float *plane = clipper->plane[plane_idx]; - const float dp0 = dot4( pos0, plane ); - const float dp1 = dot4( pos1, plane ); - - if (dp1 < 0.0F) { - float t = dp1 / (dp1 - dp0); - t1 = MAX2(t1, t); - } - - if (dp0 < 0.0F) { - float t = dp0 / (dp0 - dp1); - t0 = MAX2(t0, t); - } - - if (t0 + t1 >= 1.0F) - return; /* discard */ - - clipmask &= ~(1 << plane_idx); /* turn off this plane's bit */ - } - - if (v0->clipmask) { - interp( clipper, stage->tmp[0], t0, v0, v1 ); - - if (clipper->flat) - copy_colors(stage, stage->tmp[0], v0); - - newprim.v[0] = stage->tmp[0]; - } - else { - newprim.v[0] = v0; - } - - if (v1->clipmask) { - interp( clipper, stage->tmp[1], t1, v1, v0 ); - newprim.v[1] = stage->tmp[1]; - } - else { - newprim.v[1] = v1; - } - - stage->next->line( stage->next, &newprim ); -} - - -static void -clip_point( struct draw_stage *stage, - struct prim_header *header ) -{ - if (header->v[0]->clipmask == 0) - stage->next->point( stage->next, header ); -} - - -static void -clip_line( struct draw_stage *stage, - struct prim_header *header ) -{ - unsigned clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask); - - if (clipmask == 0) { - /* no clipping needed */ - stage->next->line( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask) == 0) { - do_clip_line(stage, header, clipmask); - } - /* else, totally clipped */ -} - - -static void -clip_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - unsigned clipmask = (header->v[0]->clipmask | - header->v[1]->clipmask | - header->v[2]->clipmask); - - if (clipmask == 0) { - /* no clipping needed */ - stage->next->tri( stage->next, header ); - } - else if ((header->v[0]->clipmask & - header->v[1]->clipmask & - header->v[2]->clipmask) == 0) { - do_clip_tri(stage, header, clipmask); - } -} - -/* Update state. Could further delay this until we hit the first - * primitive that really requires clipping. - */ -static void -clip_init_state( struct draw_stage *stage ) -{ - struct clipper *clipper = clipper_stage( stage ); - - clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; - - if (clipper->flat) { - const struct draw_vertex_shader *vs = stage->draw->vertex_shader; - uint i; - - clipper->num_color_attribs = 0; - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR || - vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - clipper->color_attribs[clipper->num_color_attribs++] = i; - } - } - } - - stage->tri = clip_tri; - stage->line = clip_line; -} - - - -static void clip_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - clip_init_state( stage ); - stage->tri( stage, header ); -} - -static void clip_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - clip_init_state( stage ); - stage->line( stage, header ); -} - - -static void clip_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->tri = clip_first_tri; - stage->line = clip_first_line; - stage->next->flush( stage->next, flags ); -} - - -static void clip_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void clip_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Allocate a new clipper stage. - * \return pointer to new stage object - */ -struct draw_stage *draw_clip_stage( struct draw_context *draw ) -{ - struct clipper *clipper = CALLOC_STRUCT(clipper); - - draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); - - clipper->stage.draw = draw; - clipper->stage.point = clip_point; - clipper->stage.line = clip_first_line; - clipper->stage.tri = clip_first_tri; - clipper->stage.flush = clip_flush; - clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; - clipper->stage.destroy = clip_destroy; - - clipper->plane = draw->plane; - - return &clipper->stage; -} diff --git a/src/gallium/auxiliary/draw/draw_cull.c b/src/gallium/auxiliary/draw/draw_cull.c deleted file mode 100644 index 8177b0ac86..0000000000 --- a/src/gallium/auxiliary/draw/draw_cull.c +++ /dev/null @@ -1,150 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Drawing stage for polygon culling - */ - -/* Authors: Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "draw_private.h" - - -struct cull_stage { - struct draw_stage stage; - unsigned winding; /**< which winding(s) to cull (one of PIPE_WINDING_x) */ -}; - - -static INLINE struct cull_stage *cull_stage( struct draw_stage *stage ) -{ - return (struct cull_stage *)stage; -} - - - - -static void cull_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - /* Window coords: */ - const float *v0 = header->v[0]->data[0]; - const float *v1 = header->v[1]->data[0]; - const float *v2 = header->v[2]->data[0]; - - /* edge vectors e = v0 - v2, f = v1 - v2 */ - const float ex = v0[0] - v2[0]; - const float ey = v0[1] - v2[1]; - const float fx = v1[0] - v2[0]; - const float fy = v1[1] - v2[1]; - - /* det = cross(e,f).z */ - header->det = ex * fy - ey * fx; - - if (header->det != 0) { - /* if (det < 0 then Z points toward camera and triangle is - * counter-clockwise winding. - */ - unsigned winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; - - if ((winding & cull_stage(stage)->winding) == 0) { - /* triangle is not culled, pass to next stage */ - stage->next->tri( stage->next, header ); - } - } -} - -static void cull_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct cull_stage *cull = cull_stage(stage); - - cull->winding = stage->draw->rasterizer->cull_mode; - - stage->tri = cull_tri; - stage->tri( stage, header ); -} - - - -static void cull_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void cull_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void cull_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->tri = cull_first_tri; - stage->next->flush( stage->next, flags ); -} - -static void cull_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void cull_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create a new polygon culling stage. - */ -struct draw_stage *draw_cull_stage( struct draw_context *draw ) -{ - struct cull_stage *cull = CALLOC_STRUCT(cull_stage); - - draw_alloc_temp_verts( &cull->stage, 0 ); - - cull->stage.draw = draw; - cull->stage.next = NULL; - cull->stage.point = cull_point; - cull->stage.line = cull_line; - cull->stage.tri = cull_first_tri; - cull->stage.flush = cull_flush; - cull->stage.reset_stipple_counter = cull_reset_stipple_counter; - cull->stage.destroy = cull_destroy; - - return &cull->stage; -} diff --git a/src/gallium/auxiliary/draw/draw_flatshade.c b/src/gallium/auxiliary/draw/draw_flatshade.c deleted file mode 100644 index 54baa1fbc9..0000000000 --- a/src/gallium/auxiliary/draw/draw_flatshade.c +++ /dev/null @@ -1,248 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "draw_vs.h" - - -/** subclass of draw_stage */ -struct flat_stage -{ - struct draw_stage stage; - - uint num_color_attribs; - uint color_attribs[4]; /* front/back primary/secondary colors */ -}; - - -static INLINE struct flat_stage * -flat_stage(struct draw_stage *stage) -{ - return (struct flat_stage *) stage; -} - - -/** Copy all the color attributes from 'src' vertex to 'dst' vertex */ -static INLINE void copy_colors( struct draw_stage *stage, - struct vertex_header *dst, - const struct vertex_header *src ) -{ - const struct flat_stage *flat = flat_stage(stage); - uint i; - for (i = 0; i < flat->num_color_attribs; i++) { - const uint attr = flat->color_attribs[i]; - COPY_4FV(dst->data[attr], src->data[attr]); - } -} - - -/** Copy all the color attributes from src vertex to dst0 & dst1 vertices */ -static INLINE void copy_colors2( struct draw_stage *stage, - struct vertex_header *dst0, - struct vertex_header *dst1, - const struct vertex_header *src ) -{ - const struct flat_stage *flat = flat_stage(stage); - uint i; - for (i = 0; i < flat->num_color_attribs; i++) { - const uint attr = flat->color_attribs[i]; - COPY_4FV(dst0->data[attr], src->data[attr]); - COPY_4FV(dst1->data[attr], src->data[attr]); - } -} - - -/** - * Flatshade tri. Required for clipping and when unfilled tris are - * active, otherwise handled by hardware. - */ -static void flatshade_tri_0( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - tmp.v[0] = header->v[0]; - tmp.v[1] = dup_vert(stage, header->v[1], 0); - tmp.v[2] = dup_vert(stage, header->v[2], 1); - - copy_colors2(stage, tmp.v[1], tmp.v[2], tmp.v[0]); - - stage->next->tri( stage->next, &tmp ); -} - - -static void flatshade_tri_2( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = header->v[2]; - - copy_colors2(stage, tmp.v[0], tmp.v[1], tmp.v[2]); - - stage->next->tri( stage->next, &tmp ); -} - - - - - -/** - * Flatshade line. Required for clipping. - */ -static void flatshade_line_0( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.v[0] = header->v[0]; - tmp.v[1] = dup_vert(stage, header->v[1], 0); - - copy_colors(stage, tmp.v[1], tmp.v[0]); - - stage->next->line( stage->next, &tmp ); -} - -static void flatshade_line_1( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = header->v[1]; - - copy_colors(stage, tmp.v[0], tmp.v[1]); - - stage->next->line( stage->next, &tmp ); -} - - -/* Flatshade point -- passthrough. - */ -static void flatshade_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void flatshade_init_state( struct draw_stage *stage ) -{ - struct flat_stage *flat = flat_stage(stage); - const struct draw_vertex_shader *vs = stage->draw->vertex_shader; - uint i; - - /* Find which vertex shader outputs are colors, make a list */ - flat->num_color_attribs = 0; - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR || - vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - flat->color_attribs[flat->num_color_attribs++] = i; - } - } - - /* Choose flatshade routine according to provoking vertex: - */ - if (stage->draw->rasterizer->flatshade_first) { - stage->line = flatshade_line_0; - stage->tri = flatshade_tri_0; - } - else { - stage->line = flatshade_line_1; - stage->tri = flatshade_tri_2; - } -} - -static void flatshade_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - flatshade_init_state( stage ); - stage->tri( stage, header ); -} - -static void flatshade_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - flatshade_init_state( stage ); - stage->line( stage, header ); -} - - -static void flatshade_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->tri = flatshade_first_tri; - stage->line = flatshade_first_line; - stage->next->flush( stage->next, flags ); -} - - -static void flatshade_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void flatshade_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create flatshading drawing stage. - */ -struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) -{ - struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage); - - draw_alloc_temp_verts( &flatshade->stage, 2 ); - - flatshade->stage.draw = draw; - flatshade->stage.next = NULL; - flatshade->stage.point = flatshade_point; - flatshade->stage.line = flatshade_first_line; - flatshade->stage.tri = flatshade_first_tri; - flatshade->stage.flush = flatshade_flush; - flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; - flatshade->stage.destroy = flatshade_destroy; - - return &flatshade->stage; -} - - diff --git a/src/gallium/auxiliary/draw/draw_offset.c b/src/gallium/auxiliary/draw/draw_offset.c deleted file mode 100644 index dbc676deae..0000000000 --- a/src/gallium/auxiliary/draw/draw_offset.c +++ /dev/null @@ -1,186 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief polygon offset state - * - * \author Keith Whitwell - * \author Brian Paul - */ - -#include "pipe/p_util.h" -#include "draw_private.h" - - - -struct offset_stage { - struct draw_stage stage; - - float scale; - float units; -}; - - - -static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) -{ - return (struct offset_stage *) stage; -} - - - - - -/** - * Offset tri Z. Some hardware can handle this, but not usually when - * doing unfilled rendering. - */ -static void do_offset_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct offset_stage *offset = offset_stage(stage); - float inv_det = 1.0f / header->det; - - /* Window coords: - */ - float *v0 = header->v[0]->data[0]; - float *v1 = header->v[1]->data[0]; - float *v2 = header->v[2]->data[0]; - - /* edge vectors e = v0 - v2, f = v1 - v2 */ - float ex = v0[0] - v2[0]; - float ey = v0[1] - v2[1]; - float ez = v0[2] - v2[2]; - float fx = v1[0] - v2[0]; - float fy = v1[1] - v2[1]; - float fz = v1[2] - v2[2]; - - /* (a,b) = cross(e,f).xy */ - float a = ey*fz - ez*fy; - float b = ez*fx - ex*fz; - - float dzdx = FABSF(a * inv_det); - float dzdy = FABSF(b * inv_det); - - float zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; - - /* - * Note: we're applying the offset and clamping per-vertex. - * Ideally, the offset is applied per-fragment prior to fragment shading. - */ - v0[2] = CLAMP(v0[2] + zoffset, 0.0f, 1.0f); - v1[2] = CLAMP(v1[2] + zoffset, 0.0f, 1.0f); - v2[2] = CLAMP(v2[2] + zoffset, 0.0f, 1.0f); - - stage->next->tri( stage->next, header ); -} - - -static void offset_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - tmp.v[0] = dup_vert(stage, header->v[0], 0); - tmp.v[1] = dup_vert(stage, header->v[1], 1); - tmp.v[2] = dup_vert(stage, header->v[2], 2); - - do_offset_tri( stage, &tmp ); -} - - -static void offset_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct offset_stage *offset = offset_stage(stage); - float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ - - offset->units = stage->draw->rasterizer->offset_units * mrd; - offset->scale = stage->draw->rasterizer->offset_scale; - - stage->tri = offset_tri; - stage->tri( stage, header ); -} - - -static void offset_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void offset_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void offset_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->tri = offset_first_tri; - stage->next->flush( stage->next, flags ); -} - - -static void offset_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void offset_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create polygon offset drawing stage. - */ -struct draw_stage *draw_offset_stage( struct draw_context *draw ) -{ - struct offset_stage *offset = CALLOC_STRUCT(offset_stage); - - draw_alloc_temp_verts( &offset->stage, 3 ); - - offset->stage.draw = draw; - offset->stage.next = NULL; - offset->stage.point = offset_point; - offset->stage.line = offset_line; - offset->stage.tri = offset_first_tri; - offset->stage.flush = offset_flush; - offset->stage.reset_stipple_counter = offset_reset_stipple_counter; - offset->stage.destroy = offset_destroy; - - return &offset->stage; -} diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c new file mode 100644 index 0000000000..e8d2a45102 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -0,0 +1,859 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * AA line stage: AA lines are converted to texture mapped triangles. + * + * Authors: Brian Paul + */ + + +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/util/tgsi_transform.h" +#include "tgsi/util/tgsi_dump.h" + +#include "draw_context.h" +#include "draw_private.h" + + +/** + * Max texture level for the alpha texture used for antialiasing + */ +#define MAX_TEXTURE_LEVEL 5 /* 32 x 32 */ + + +/** + * Subclass of pipe_shader_state to carry extra fragment shader info. + */ +struct aaline_fragment_shader +{ + struct pipe_shader_state state; + void *driver_fs; + void *aaline_fs; + void *aapoint_fs; /* not yet */ + void *sprite_fs; /* not yet */ + uint sampler_unit; + int generic_attrib; /**< texcoord/generic used for texture */ +}; + + +/** + * Subclass of draw_stage + */ +struct aaline_stage +{ + struct draw_stage stage; + + float half_line_width; + + /** For AA lines, this is the vertex attrib slot for the new texcoords */ + uint tex_slot; + + void *sampler_cso; + struct pipe_texture *texture; + uint num_samplers; + uint num_textures; + + + /* + * Currently bound state + */ + struct aaline_fragment_shader *fs; + struct { + void *sampler[PIPE_MAX_SAMPLERS]; + struct pipe_texture *texture[PIPE_MAX_SAMPLERS]; + } state; + + /* + * Driver interface/override functions + */ + void * (*driver_create_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*driver_bind_fs_state)(struct pipe_context *, void *); + void (*driver_delete_fs_state)(struct pipe_context *, void *); + + void (*driver_bind_sampler_states)(struct pipe_context *, unsigned, + void **); + void (*driver_set_sampler_textures)(struct pipe_context *, unsigned, + struct pipe_texture **); + + struct pipe_context *pipe; +}; + + + +/** + * Subclass of tgsi_transform_context, used for transforming the + * user's fragment shader to add the special AA instructions. + */ +struct aa_transform_context { + struct tgsi_transform_context base; + uint tempsUsed; /**< bitmask */ + int colorOutput; /**< which output is the primary color */ + uint samplersUsed; /**< bitfield of samplers used */ + int freeSampler; /** an available sampler for the pstipple */ + int maxInput, maxGeneric; /**< max input index found */ + int colorTemp, texTemp; /**< temp registers */ + boolean firstInstruction; +}; + + +/** + * TGSI declaration transform callback. + * Look for a free sampler, a free input attrib, and two free temp regs. + */ +static void +aa_transform_decl(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl) +{ + struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; + + if (decl->Declaration.File == TGSI_FILE_OUTPUT && + decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR && + decl->Semantic.SemanticIndex == 0) { + aactx->colorOutput = decl->u.DeclarationRange.First; + } + else if (decl->Declaration.File == TGSI_FILE_SAMPLER) { + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + aactx->samplersUsed |= 1 << i; + } + } + else if (decl->Declaration.File == TGSI_FILE_INPUT) { + if ((int) decl->u.DeclarationRange.Last > aactx->maxInput) + aactx->maxInput = decl->u.DeclarationRange.Last; + if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && + (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) { + aactx->maxGeneric = decl->Semantic.SemanticIndex; + } + } + else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + aactx->tempsUsed |= (1 << i); + } + } + + ctx->emit_declaration(ctx, decl); +} + + +/** + * Find the lowest zero bit in the given word, or -1 if bitfield is all ones. + */ +static int +free_bit(uint bitfield) +{ + int i; + for (i = 0; i < 32; i++) { + if ((bitfield & (1 << i)) == 0) + return i; + } + return -1; +} + + +/** + * TGSI instruction transform callback. + * Replace writes to result.color w/ a temp reg. + * Upon END instruction, insert texture sampling code for antialiasing. + */ +static void +aa_transform_inst(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst) +{ + struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; + + if (aactx->firstInstruction) { + /* emit our new declarations before the first instruction */ + + struct tgsi_full_declaration decl; + uint i; + + /* find free sampler */ + aactx->freeSampler = free_bit(aactx->samplersUsed); + if (aactx->freeSampler >= PIPE_MAX_SAMPLERS) + aactx->freeSampler = PIPE_MAX_SAMPLERS - 1; + + /* find two free temp regs */ + for (i = 0; i < 32; i++) { + if ((aactx->tempsUsed & (1 << i)) == 0) { + /* found a free temp */ + if (aactx->colorTemp < 0) + aactx->colorTemp = i; + else if (aactx->texTemp < 0) + aactx->texTemp = i; + else + break; + } + } + assert(aactx->colorTemp >= 0); + assert(aactx->texTemp >= 0); + + /* declare new generic input/texcoord */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; + decl.Semantic.SemanticIndex = aactx->maxGeneric + 1; + decl.Declaration.Interpolate = 1; + /* XXX this could be linear... */ + decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->maxInput + 1; + ctx->emit_declaration(ctx, &decl); + + /* declare new sampler */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_SAMPLER; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->freeSampler; + ctx->emit_declaration(ctx, &decl); + + /* declare new temp regs */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->texTemp; + ctx->emit_declaration(ctx, &decl); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->colorTemp; + ctx->emit_declaration(ctx, &decl); + + aactx->firstInstruction = FALSE; + } + + if (inst->Instruction.Opcode == TGSI_OPCODE_END && + aactx->colorOutput != -1) { + struct tgsi_full_instruction newInst; + + /* TEX */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_TEX; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->texTemp; + newInst.Instruction.NumSrcRegs = 2; + newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->maxInput + 1; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->freeSampler; + + ctx->emit_instruction(ctx, &newInst); + + /* MOV rgb */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MOV; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; + ctx->emit_instruction(ctx, &newInst); + + /* MUL alpha */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->texTemp; + ctx->emit_instruction(ctx, &newInst); + + /* END */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_END; + newInst.Instruction.NumDstRegs = 0; + newInst.Instruction.NumSrcRegs = 0; + ctx->emit_instruction(ctx, &newInst); + } + else { + /* Not an END instruction. + * Look for writes to result.color and replace with colorTemp reg. + */ + uint i; + + for (i = 0; i < inst->Instruction.NumDstRegs; i++) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + if (dst->DstRegister.File == TGSI_FILE_OUTPUT && + dst->DstRegister.Index == aactx->colorOutput) { + dst->DstRegister.File = TGSI_FILE_TEMPORARY; + dst->DstRegister.Index = aactx->colorTemp; + } + } + + ctx->emit_instruction(ctx, inst); + } +} + + +/** + * Generate the frag shader we'll use for drawing AA lines. + * This will be the user's shader plus some texture/modulate instructions. + */ +static void +generate_aaline_fs(struct aaline_stage *aaline) +{ + const struct pipe_shader_state *orig_fs = &aaline->fs->state; + //struct draw_context *draw = aaline->stage.draw; + struct pipe_shader_state aaline_fs; + struct aa_transform_context transform; + +#define MAX 1000 + + aaline_fs = *orig_fs; /* copy to init */ + aaline_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); + + memset(&transform, 0, sizeof(transform)); + transform.colorOutput = -1; + transform.maxInput = -1; + transform.maxGeneric = -1; + transform.colorTemp = -1; + transform.texTemp = -1; + transform.firstInstruction = TRUE; + transform.base.transform_instruction = aa_transform_inst; + transform.base.transform_declaration = aa_transform_decl; + + tgsi_transform_shader(orig_fs->tokens, + (struct tgsi_token *) aaline_fs.tokens, + MAX, &transform.base); + +#if 0 /* DEBUG */ + tgsi_dump(orig_fs->tokens, 0); + tgsi_dump(aaline_fs.tokens, 0); +#endif + + aaline->fs->sampler_unit = transform.freeSampler; + + aaline->fs->aaline_fs + = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); + + aaline->fs->generic_attrib = transform.maxGeneric + 1; +} + + +/** + * Create the texture map we'll use for antialiasing the lines. + */ +static void +aaline_create_texture(struct aaline_stage *aaline) +{ + struct pipe_context *pipe = aaline->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_texture texTemp; + uint level; + + memset(&texTemp, 0, sizeof(texTemp)); + texTemp.target = PIPE_TEXTURE_2D; + texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */ + texTemp.last_level = MAX_TEXTURE_LEVEL; + texTemp.width[0] = 1 << MAX_TEXTURE_LEVEL; + texTemp.height[0] = 1 << MAX_TEXTURE_LEVEL; + texTemp.depth[0] = 1; + texTemp.cpp = 1; + + aaline->texture = screen->texture_create(screen, &texTemp); + + /* Fill in mipmap images. + * Basically each level is solid opaque, except for the outermost + * texels which are zero. Special case the 1x1 and 2x2 levels. + */ + for (level = 0; level <= MAX_TEXTURE_LEVEL; level++) { + struct pipe_surface *surface; + const uint size = aaline->texture->width[level]; + ubyte *data; + uint i, j; + + assert(aaline->texture->width[level] == aaline->texture->height[level]); + + surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0); + data = pipe_surface_map(surface); + + for (i = 0; i < size; i++) { + for (j = 0; j < size; j++) { + ubyte d; + if (size == 1) { + d = 255; + } + else if (size == 2) { + d = 200; /* tuneable */ + } + else if (i == 0 || j == 0 || i == size - 1 || j == size - 1) { + d = 0; + } + else { + d = 255; + } + data[i * surface->pitch + j] = d; + } + } + + /* unmap */ + pipe_surface_unmap(surface); + pipe_surface_reference(&surface, NULL); + pipe->texture_update(pipe, aaline->texture, 0, (1 << level)); + } +} + + +/** + * Create the sampler CSO that'll be used for antialiasing. + * By using a mipmapped texture, we don't have to generate a different + * texture image for each line size. + */ +static void +aaline_create_sampler(struct aaline_stage *aaline) +{ + struct pipe_sampler_state sampler; + struct pipe_context *pipe = aaline->pipe; + + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_LINEAR; + sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR; + sampler.normalized_coords = 1; + sampler.min_lod = 0.0f; + sampler.max_lod = MAX_TEXTURE_LEVEL; + + aaline->sampler_cso = pipe->create_sampler_state(pipe, &sampler); +} + + +/** + * When we're about to draw our first AA line in a batch, this function is + * called to tell the driver to bind our modified fragment shader. + */ +static void +bind_aaline_fragment_shader(struct aaline_stage *aaline) +{ + if (!aaline->fs->aaline_fs) { + generate_aaline_fs(aaline); + } + aaline->driver_bind_fs_state(aaline->pipe, aaline->fs->aaline_fs); +} + + + +static INLINE struct aaline_stage * +aaline_stage( struct draw_stage *stage ) +{ + return (struct aaline_stage *) stage; +} + + +static void +passthrough_point(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->point(stage->next, header); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Draw a wide line by drawing a quad, using geometry which will + * fullfill GL's antialiased line requirements. + */ +static void +aaline_line(struct draw_stage *stage, struct prim_header *header) +{ + const struct aaline_stage *aaline = aaline_stage(stage); + const float half_width = aaline->half_line_width; + struct prim_header tri; + struct vertex_header *v[8]; + uint texPos = aaline->tex_slot; + float *pos, *tex; + float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; + float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; + double a = atan2(dy, dx); + float c_a = (float) cos(a), s_a = (float) sin(a); + uint i; + + /* XXX the ends of lines aren't quite perfect yet, but probably passable */ + dx = 0.5F * half_width; + dy = half_width; + + /* allocate/dup new verts */ + for (i = 0; i < 8; i++) { + v[i] = dup_vert(stage, header->v[i/4], i); + } + + /* + * Quad strip for line from v0 to v1 (*=endpoints): + * + * 1 3 5 7 + * +---+---------------------+---+ + * | | + * | *v0 v1* | + * | | + * +---+---------------------+---+ + * 0 2 4 6 + */ + + /* new verts */ + pos = v[0]->data[0]; + pos[0] += (-dx * c_a - dy * s_a); + pos[1] += (-dx * s_a + dy * c_a); + + pos = v[1]->data[0]; + pos[0] += (-dx * c_a - -dy * s_a); + pos[1] += (-dx * s_a + -dy * c_a); + + pos = v[2]->data[0]; + pos[0] += ( dx * c_a - dy * s_a); + pos[1] += ( dx * s_a + dy * c_a); + + pos = v[3]->data[0]; + pos[0] += ( dx * c_a - -dy * s_a); + pos[1] += ( dx * s_a + -dy * c_a); + + pos = v[4]->data[0]; + pos[0] += (-dx * c_a - dy * s_a); + pos[1] += (-dx * s_a + dy * c_a); + + pos = v[5]->data[0]; + pos[0] += (-dx * c_a - -dy * s_a); + pos[1] += (-dx * s_a + -dy * c_a); + + pos = v[6]->data[0]; + pos[0] += ( dx * c_a - dy * s_a); + pos[1] += ( dx * s_a + dy * c_a); + + pos = v[7]->data[0]; + pos[0] += ( dx * c_a - -dy * s_a); + pos[1] += ( dx * s_a + -dy * c_a); + + /* new texcoords */ + tex = v[0]->data[texPos]; + ASSIGN_4V(tex, 0, 0, 0, 1); + + tex = v[1]->data[texPos]; + ASSIGN_4V(tex, 0, 1, 0, 1); + + tex = v[2]->data[texPos]; + ASSIGN_4V(tex, .5, 0, 0, 1); + + tex = v[3]->data[texPos]; + ASSIGN_4V(tex, .5, 1, 0, 1); + + tex = v[4]->data[texPos]; + ASSIGN_4V(tex, .5, 0, 0, 1); + + tex = v[5]->data[texPos]; + ASSIGN_4V(tex, .5, 1, 0, 1); + + tex = v[6]->data[texPos]; + ASSIGN_4V(tex, 1, 0, 0, 1); + + tex = v[7]->data[texPos]; + ASSIGN_4V(tex, 1, 1, 0, 1); + + /* emit 6 tris for the quad strip */ + tri.v[0] = v[2]; tri.v[1] = v[1]; tri.v[2] = v[0]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[3]; tri.v[1] = v[1]; tri.v[2] = v[2]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[4]; tri.v[1] = v[3]; tri.v[2] = v[2]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[5]; tri.v[1] = v[3]; tri.v[2] = v[4]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[6]; tri.v[1] = v[5]; tri.v[2] = v[4]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[7]; tri.v[1] = v[5]; tri.v[2] = v[6]; + stage->next->tri( stage->next, &tri ); +} + + +static void +aaline_first_line(struct draw_stage *stage, struct prim_header *header) +{ + auto struct aaline_stage *aaline = aaline_stage(stage); + struct draw_context *draw = stage->draw; + struct pipe_context *pipe = aaline->pipe; + uint num_samplers; + + assert(draw->rasterizer->line_smooth); + + if (draw->rasterizer->line_width <= 3.0) + aaline->half_line_width = 1.5f; + else + aaline->half_line_width = 0.5f * draw->rasterizer->line_width; + + /* + * Bind (generate) our fragprog, sampler and texture + */ + bind_aaline_fragment_shader(aaline); + + /* update vertex attrib info */ + aaline->tex_slot = draw->num_vs_outputs; + assert(aaline->tex_slot > 0); /* output[0] is vertex pos */ + + /* advertise the extra post-transformed vertex attribute */ + draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; + draw->extra_vp_outputs.semantic_index = aaline->fs->generic_attrib; + draw->extra_vp_outputs.slot = aaline->tex_slot; + + /* how many samplers? */ + /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ + num_samplers = MAX2(aaline->num_textures, aaline->num_samplers); + num_samplers = MAX2(num_samplers, aaline->fs->sampler_unit + 1); + + aaline->state.sampler[aaline->fs->sampler_unit] = aaline->sampler_cso; + pipe_texture_reference(&aaline->state.texture[aaline->fs->sampler_unit], + aaline->texture); + + aaline->driver_bind_sampler_states(pipe, num_samplers, aaline->state.sampler); + aaline->driver_set_sampler_textures(pipe, num_samplers, aaline->state.texture); + + /* now really draw first line */ + stage->line = aaline_line; + stage->line(stage, header); +} + + +static void +aaline_flush(struct draw_stage *stage, unsigned flags) +{ + struct draw_context *draw = stage->draw; + struct aaline_stage *aaline = aaline_stage(stage); + struct pipe_context *pipe = aaline->pipe; + + stage->line = aaline_first_line; + stage->next->flush( stage->next, flags ); + + /* restore original frag shader */ + aaline->driver_bind_fs_state(pipe, aaline->fs->driver_fs); + + /* XXX restore original texture, sampler state */ + aaline->driver_bind_sampler_states(pipe, aaline->num_samplers, + aaline->state.sampler); + aaline->driver_set_sampler_textures(pipe, aaline->num_textures, + aaline->state.texture); + + draw->extra_vp_outputs.slot = 0; +} + + +static void +aaline_reset_stipple_counter(struct draw_stage *stage) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +aaline_destroy(struct draw_stage *stage) +{ + struct aaline_stage *aaline = aaline_stage(stage); + + aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso); + + pipe_texture_release(&aaline->texture); + + draw_free_temp_verts( stage ); + + FREE( stage ); +} + + +static struct aaline_stage * +draw_aaline_stage(struct draw_context *draw) +{ + struct aaline_stage *aaline = CALLOC_STRUCT(aaline_stage); + + draw_alloc_temp_verts( &aaline->stage, 8 ); + + aaline->stage.draw = draw; + aaline->stage.next = NULL; + aaline->stage.point = passthrough_point; + aaline->stage.line = aaline_first_line; + aaline->stage.tri = passthrough_tri; + aaline->stage.flush = aaline_flush; + aaline->stage.reset_stipple_counter = aaline_reset_stipple_counter; + aaline->stage.destroy = aaline_destroy; + + return aaline; +} + + +static struct aaline_stage * +aaline_stage_from_pipe(struct pipe_context *pipe) +{ + struct draw_context *draw = (struct draw_context *) pipe->draw; + return aaline_stage(draw->pipeline.aaline); +} + + +/** + * This function overrides the driver's create_fs_state() function and + * will typically be called by the state tracker. + */ +static void * +aaline_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *fs) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + struct aaline_fragment_shader *aafs = CALLOC_STRUCT(aaline_fragment_shader); + + if (aafs) { + aafs->state = *fs; + + /* pass-through */ + aafs->driver_fs = aaline->driver_create_fs_state(aaline->pipe, fs); + } + + return aafs; +} + + +static void +aaline_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs; + /* save current */ + aaline->fs = aafs; + /* pass-through */ + aaline->driver_bind_fs_state(aaline->pipe, + (aafs ? aafs->driver_fs : NULL)); +} + + +static void +aaline_delete_fs_state(struct pipe_context *pipe, void *fs) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs; + /* pass-through */ + aaline->driver_delete_fs_state(aaline->pipe, aafs->driver_fs); + FREE(aafs); +} + + +static void +aaline_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **sampler) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + /* save current */ + memcpy(aaline->state.sampler, sampler, num * sizeof(void *)); + aaline->num_samplers = num; + /* pass-through */ + aaline->driver_bind_sampler_states(aaline->pipe, num, sampler); +} + + +static void +aaline_set_sampler_textures(struct pipe_context *pipe, + unsigned num, struct pipe_texture **texture) +{ + struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + uint i; + + /* save current */ + for (i = 0; i < num; i++) { + pipe_texture_reference(&aaline->state.texture[i], texture[i]); + } + aaline->num_textures = num; + + /* pass-through */ + aaline->driver_set_sampler_textures(aaline->pipe, num, texture); +} + + +/** + * Called by drivers that want to install this AA line prim stage + * into the draw module's pipeline. This will not be used if the + * hardware has native support for AA lines. + */ +void +draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) +{ + struct aaline_stage *aaline; + + pipe->draw = (void *) draw; + + /* + * Create / install AA line drawing / prim stage + */ + aaline = draw_aaline_stage( draw ); + assert(aaline); + draw->pipeline.aaline = &aaline->stage; + + aaline->pipe = pipe; + + /* create special texture, sampler state */ + aaline_create_texture(aaline); + aaline_create_sampler(aaline); + + /* save original driver functions */ + aaline->driver_create_fs_state = pipe->create_fs_state; + aaline->driver_bind_fs_state = pipe->bind_fs_state; + aaline->driver_delete_fs_state = pipe->delete_fs_state; + + aaline->driver_bind_sampler_states = pipe->bind_sampler_states; + aaline->driver_set_sampler_textures = pipe->set_sampler_textures; + + /* override the driver's functions */ + pipe->create_fs_state = aaline_create_fs_state; + pipe->bind_fs_state = aaline_bind_fs_state; + pipe->delete_fs_state = aaline_delete_fs_state; + + pipe->bind_sampler_states = aaline_bind_sampler_states; + pipe->set_sampler_textures = aaline_set_sampler_textures; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c new file mode 100644 index 0000000000..e84d380e50 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -0,0 +1,846 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * AA point stage: AA points are converted to quads and rendered with a + * special fragment shader. Another approach would be to use a texture + * map image of a point, but experiments indicate the quality isn't nearly + * as good as this approach. + * + * Note: this looks a lot like draw_aaline.c but there's actually little + * if any code that can be shared. + * + * Authors: Brian Paul + */ + + +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/util/tgsi_transform.h" +#include "tgsi/util/tgsi_dump.h" + +#include "draw_context.h" +#include "draw_vs.h" + + +/* + * Enabling NORMALIZE might give _slightly_ better results. + * Basically, it controls whether we compute distance as d=sqrt(x*x+y*y) or + * d=x*x+y*y. Since we're working with a unit circle, the later seems + * close enough and saves some costly instructions. + */ +#define NORMALIZE 0 + + +/** + * Subclass of pipe_shader_state to carry extra fragment shader info. + */ +struct aapoint_fragment_shader +{ + struct pipe_shader_state state; + void *driver_fs; /**< the regular shader */ + void *aapoint_fs; /**< the aa point-augmented shader */ + int generic_attrib; /**< The generic input attrib/texcoord we'll use */ +}; + + +/** + * Subclass of draw_stage + */ +struct aapoint_stage +{ + struct draw_stage stage; + + int psize_slot; + float radius; + + /** this is the vertex attrib slot for the new texcoords */ + uint tex_slot; + + /* + * Currently bound state + */ + struct aapoint_fragment_shader *fs; + + /* + * Driver interface/override functions + */ + void * (*driver_create_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*driver_bind_fs_state)(struct pipe_context *, void *); + void (*driver_delete_fs_state)(struct pipe_context *, void *); + + struct pipe_context *pipe; +}; + + + +/** + * Subclass of tgsi_transform_context, used for transforming the + * user's fragment shader to add the special AA instructions. + */ +struct aa_transform_context { + struct tgsi_transform_context base; + uint tempsUsed; /**< bitmask */ + int colorOutput; /**< which output is the primary color */ + int maxInput, maxGeneric; /**< max input index found */ + int tmp0, colorTemp; /**< temp registers */ + boolean firstInstruction; +}; + + +/** + * TGSI declaration transform callback. + * Look for two free temp regs and available input reg for new texcoords. + */ +static void +aa_transform_decl(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl) +{ + struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; + + if (decl->Declaration.File == TGSI_FILE_OUTPUT && + decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR && + decl->Semantic.SemanticIndex == 0) { + aactx->colorOutput = decl->u.DeclarationRange.First; + } + else if (decl->Declaration.File == TGSI_FILE_INPUT) { + if ((int) decl->u.DeclarationRange.Last > aactx->maxInput) + aactx->maxInput = decl->u.DeclarationRange.Last; + if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && + (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) { + aactx->maxGeneric = decl->Semantic.SemanticIndex; + } + } + else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + aactx->tempsUsed |= (1 << i); + } + } + + ctx->emit_declaration(ctx, decl); +} + + +/** + * TGSI instruction transform callback. + * Replace writes to result.color w/ a temp reg. + * Upon END instruction, insert texture sampling code for antialiasing. + */ +static void +aa_transform_inst(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst) +{ + struct aa_transform_context *aactx = (struct aa_transform_context *) ctx; + struct tgsi_full_instruction newInst; + + if (aactx->firstInstruction) { + /* emit our new declarations before the first instruction */ + + struct tgsi_full_declaration decl; + const int texInput = aactx->maxInput + 1; + int tmp0; + uint i; + + /* find two free temp regs */ + for (i = 0; i < 32; i++) { + if ((aactx->tempsUsed & (1 << i)) == 0) { + /* found a free temp */ + if (aactx->tmp0 < 0) + aactx->tmp0 = i; + else if (aactx->colorTemp < 0) + aactx->colorTemp = i; + else + break; + } + } + + assert(aactx->colorTemp != aactx->tmp0); + + tmp0 = aactx->tmp0; + + /* declare new generic input/texcoord */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; + decl.Semantic.SemanticIndex = aactx->maxGeneric + 1; + decl.Declaration.Interpolate = 1; + /* XXX this could be linear... */ + decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = texInput; + ctx->emit_declaration(ctx, &decl); + + /* declare new temp regs */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = tmp0; + ctx->emit_declaration(ctx, &decl); + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = aactx->colorTemp; + ctx->emit_declaration(ctx, &decl); + + aactx->firstInstruction = FALSE; + + + /* + * Emit code to compute fragment coverage, kill if outside point radius + * + * Temp reg0 usage: + * t0.x = distance of fragment from center point + * t0.y = boolean, is t0.x > 1.0, also misc temp usage + * t0.z = temporary for computing 1/(1-k) value + * t0.w = final coverage value + */ + + /* MUL t0.xy, tex, tex; # compute x^2, y^2 */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XY; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + ctx->emit_instruction(ctx, &newInst); + + /* ADD t0.x, t0.x, t0.y; # x^2 + y^2 */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_ADD; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + ctx->emit_instruction(ctx, &newInst); + +#if NORMALIZE /* OPTIONAL normalization of length */ + /* RSQ t0.x, t0.x; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_RSQ; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + ctx->emit_instruction(ctx, &newInst); + + /* RCP t0.x, t0.x; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_RCP; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + ctx->emit_instruction(ctx, &newInst); +#endif + + /* SGT t0.y, t0.xxxx, t0.wwww; # bool b = d > 1 (NOTE t0.w == 1) */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SGT; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; + ctx->emit_instruction(ctx, &newInst); + + /* KILP -t0.yyyy; # if b, KILL */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_KILP; + newInst.Instruction.NumDstRegs = 0; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; + ctx->emit_instruction(ctx, &newInst); + + + /* compute coverage factor = (1-d)/(1-k) */ + + /* SUB t0.z, tex.w, tex.z; # m = 1 - k */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SUB; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* RCP t0.z, t0.z; # t0.z = 1 / m */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_RCP; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Z; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* SUB t0.y, 1, t0.x; # d = 1 - d */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SUB; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + ctx->emit_instruction(ctx, &newInst); + + /* MUL t0.w, t0.y, t0.z; # coverage = d * m */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* SLE t0.y, t0.x, tex.z; # bool b = distance <= k */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_SLE; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_Y; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Z; + ctx->emit_instruction(ctx, &newInst); + + /* CMP t0.w, -t0.y, tex.w, t0.w; + * # if -t0.y < 0 then + * t0.w = 1 + * else + * t0.w = t0.w + */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_CMP; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = tmp0; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 3; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y; + newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[1].SrcRegister.Index = texInput; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[2].SrcRegister.Index = tmp0; + newInst.FullSrcRegisters[2].SrcRegister.SwizzleX = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[2].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[2].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W; + newInst.FullSrcRegisters[2].SrcRegister.SwizzleW = TGSI_SWIZZLE_W; + ctx->emit_instruction(ctx, &newInst); + + } + + if (inst->Instruction.Opcode == TGSI_OPCODE_END) { + /* add alpha modulation code at tail of program */ + + /* MOV result.color.xyz, colorTemp; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MOV; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_XYZ; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; + ctx->emit_instruction(ctx, &newInst); + + /* MUL result.color.w, colorTemp, tmp0.w; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT; + newInst.FullDstRegisters[0].DstRegister.Index = aactx->colorOutput; + newInst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_W; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = aactx->colorTemp; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[1].SrcRegister.Index = aactx->tmp0; + ctx->emit_instruction(ctx, &newInst); + } + else { + /* Not an END instruction. + * Look for writes to result.color and replace with colorTemp reg. + */ + uint i; + + for (i = 0; i < inst->Instruction.NumDstRegs; i++) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + if (dst->DstRegister.File == TGSI_FILE_OUTPUT && + dst->DstRegister.Index == aactx->colorOutput) { + dst->DstRegister.File = TGSI_FILE_TEMPORARY; + dst->DstRegister.Index = aactx->colorTemp; + } + } + } + + ctx->emit_instruction(ctx, inst); +} + + +/** + * Generate the frag shader we'll use for drawing AA lines. + * This will be the user's shader plus some texture/modulate instructions. + */ +static void +generate_aapoint_fs(struct aapoint_stage *aapoint) +{ + const struct pipe_shader_state *orig_fs = &aapoint->fs->state; + struct pipe_shader_state aapoint_fs; + struct aa_transform_context transform; + +#define MAX 1000 + + aapoint_fs = *orig_fs; /* copy to init */ + aapoint_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); + + memset(&transform, 0, sizeof(transform)); + transform.colorOutput = -1; + transform.maxInput = -1; + transform.maxGeneric = -1; + transform.colorTemp = -1; + transform.tmp0 = -1; + transform.firstInstruction = TRUE; + transform.base.transform_instruction = aa_transform_inst; + transform.base.transform_declaration = aa_transform_decl; + + tgsi_transform_shader(orig_fs->tokens, + (struct tgsi_token *) aapoint_fs.tokens, + MAX, &transform.base); + +#if 0 /* DEBUG */ + printf("draw_aapoint, orig shader:\n"); + tgsi_dump(orig_fs->tokens, 0); + printf("draw_aapoint, new shader:\n"); + tgsi_dump(aapoint_fs.tokens, 0); +#endif + + aapoint->fs->aapoint_fs + = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs); + + aapoint->fs->generic_attrib = transform.maxGeneric + 1; +} + + +/** + * When we're about to draw our first AA line in a batch, this function is + * called to tell the driver to bind our modified fragment shader. + */ +static void +bind_aapoint_fragment_shader(struct aapoint_stage *aapoint) +{ + if (!aapoint->fs->aapoint_fs) { + generate_aapoint_fs(aapoint); + } + aapoint->driver_bind_fs_state(aapoint->pipe, aapoint->fs->aapoint_fs); +} + + + +static INLINE struct aapoint_stage * +aapoint_stage( struct draw_stage *stage ) +{ + return (struct aapoint_stage *) stage; +} + + +static void +passthrough_line(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->line(stage->next, header); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Draw an AA point by drawing a quad. + */ +static void +aapoint_point(struct draw_stage *stage, struct prim_header *header) +{ + const struct aapoint_stage *aapoint = aapoint_stage(stage); + struct prim_header tri; + struct vertex_header *v[4]; + uint texPos = aapoint->tex_slot; + float radius, *pos, *tex; + uint i; + float k; + + if (aapoint->psize_slot >= 0) { + radius = 0.5f * header->v[0]->data[aapoint->psize_slot][0]; + } + else { + radius = aapoint->radius; + } + + /* + * Note: the texcoords (generic attrib, really) we use are special: + * The S and T components simply vary from -1 to +1. + * The R component is k, below. + * The Q component is 1.0 and will used as a handy constant in the + * fragment shader. + */ + + /* + * k is the threshold distance from the point's center at which + * we begin alpha attenuation (the coverage value). + * Operating within a unit circle, we'll compute the fragment's + * distance 'd' from the center point using the texcoords. + * IF d > 1.0 THEN + * KILL fragment + * ELSE IF d > k THEN + * compute coverage in [0,1] proportional to d in [k, 1]. + * ELSE + * coverage = 1.0; // full coverage + * ENDIF + * + * Note: the ELSEIF and ELSE clauses are actually implemented with CMP to + * avoid using IF/ELSE/ENDIF TGSI opcodes. + */ + +#if !NORMALIZE + k = 1.0f / radius; + k = 1.0f - 2.0f * k + k * k; +#else + k = 1.0f - 1.0f / radius; +#endif + + /* allocate/dup new verts */ + for (i = 0; i < 4; i++) { + v[i] = dup_vert(stage, header->v[0], i); + } + + /* new verts */ + pos = v[0]->data[0]; + pos[0] -= radius; + pos[1] -= radius; + + pos = v[1]->data[0]; + pos[0] += radius; + pos[1] -= radius; + + pos = v[2]->data[0]; + pos[0] += radius; + pos[1] += radius; + + pos = v[3]->data[0]; + pos[0] -= radius; + pos[1] += radius; + + /* new texcoords */ + tex = v[0]->data[texPos]; + ASSIGN_4V(tex, -1, -1, k, 1); + + tex = v[1]->data[texPos]; + ASSIGN_4V(tex, 1, -1, k, 1); + + tex = v[2]->data[texPos]; + ASSIGN_4V(tex, 1, 1, k, 1); + + tex = v[3]->data[texPos]; + ASSIGN_4V(tex, -1, 1, k, 1); + + /* emit 2 tris for the quad strip */ + tri.v[0] = v[0]; + tri.v[1] = v[1]; + tri.v[2] = v[2]; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v[0]; + tri.v[1] = v[2]; + tri.v[2] = v[3]; + stage->next->tri( stage->next, &tri ); +} + + +static void +aapoint_first_point(struct draw_stage *stage, struct prim_header *header) +{ + auto struct aapoint_stage *aapoint = aapoint_stage(stage); + struct draw_context *draw = stage->draw; + + assert(draw->rasterizer->point_smooth); + + if (draw->rasterizer->point_size <= 2.0) + aapoint->radius = 1.0; + else + aapoint->radius = 0.5f * draw->rasterizer->point_size; + + /* + * Bind (generate) our fragprog. + */ + bind_aapoint_fragment_shader(aapoint); + + /* update vertex attrib info */ + aapoint->tex_slot = draw->num_vs_outputs; + assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */ + + draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; + draw->extra_vp_outputs.semantic_index = aapoint->fs->generic_attrib; + draw->extra_vp_outputs.slot = aapoint->tex_slot; + + /* find psize slot in post-transform vertex */ + aapoint->psize_slot = -1; + if (draw->rasterizer->point_size_per_vertex) { + /* find PSIZ vertex output */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + aapoint->psize_slot = i; + break; + } + } + } + + /* now really draw first line */ + stage->point = aapoint_point; + stage->point(stage, header); +} + + +static void +aapoint_flush(struct draw_stage *stage, unsigned flags) +{ + struct draw_context *draw = stage->draw; + struct aapoint_stage *aapoint = aapoint_stage(stage); + struct pipe_context *pipe = aapoint->pipe; + + stage->point = aapoint_first_point; + stage->next->flush( stage->next, flags ); + + /* restore original frag shader */ + aapoint->driver_bind_fs_state(pipe, aapoint->fs->driver_fs); + + draw->extra_vp_outputs.slot = 0; +} + + +static void +aapoint_reset_stipple_counter(struct draw_stage *stage) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +aapoint_destroy(struct draw_stage *stage) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +static struct aapoint_stage * +draw_aapoint_stage(struct draw_context *draw) +{ + struct aapoint_stage *aapoint = CALLOC_STRUCT(aapoint_stage); + + draw_alloc_temp_verts( &aapoint->stage, 4 ); + + aapoint->stage.draw = draw; + aapoint->stage.next = NULL; + aapoint->stage.point = aapoint_first_point; + aapoint->stage.line = passthrough_line; + aapoint->stage.tri = passthrough_tri; + aapoint->stage.flush = aapoint_flush; + aapoint->stage.reset_stipple_counter = aapoint_reset_stipple_counter; + aapoint->stage.destroy = aapoint_destroy; + + return aapoint; +} + + +static struct aapoint_stage * +aapoint_stage_from_pipe(struct pipe_context *pipe) +{ + struct draw_context *draw = (struct draw_context *) pipe->draw; + return aapoint_stage(draw->pipeline.aapoint); +} + + +/** + * This function overrides the driver's create_fs_state() function and + * will typically be called by the state tracker. + */ +static void * +aapoint_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *fs) +{ + struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); + struct aapoint_fragment_shader *aafs = CALLOC_STRUCT(aapoint_fragment_shader); + + if (aafs) { + aafs->state = *fs; + + /* pass-through */ + aafs->driver_fs = aapoint->driver_create_fs_state(aapoint->pipe, fs); + } + + return aafs; +} + + +static void +aapoint_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); + struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs; + /* save current */ + aapoint->fs = aafs; + /* pass-through */ + aapoint->driver_bind_fs_state(aapoint->pipe, + (aafs ? aafs->driver_fs : NULL)); +} + + +static void +aapoint_delete_fs_state(struct pipe_context *pipe, void *fs) +{ + struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); + struct aapoint_fragment_shader *aafs = (struct aapoint_fragment_shader *) fs; + /* pass-through */ + aapoint->driver_delete_fs_state(aapoint->pipe, aafs->driver_fs); + FREE(aafs); +} + + +/** + * Called by drivers that want to install this AA point prim stage + * into the draw module's pipeline. This will not be used if the + * hardware has native support for AA points. + */ +void +draw_install_aapoint_stage(struct draw_context *draw, + struct pipe_context *pipe) +{ + struct aapoint_stage *aapoint; + + pipe->draw = (void *) draw; + + /* + * Create / install AA point drawing / prim stage + */ + aapoint = draw_aapoint_stage( draw ); + assert(aapoint); + draw->pipeline.aapoint = &aapoint->stage; + + aapoint->pipe = pipe; + + /* save original driver functions */ + aapoint->driver_create_fs_state = pipe->create_fs_state; + aapoint->driver_bind_fs_state = pipe->bind_fs_state; + aapoint->driver_delete_fs_state = pipe->delete_fs_state; + + /* override the driver's functions */ + pipe->create_fs_state = aapoint_create_fs_state; + pipe->bind_fs_state = aapoint_bind_fs_state; + pipe->delete_fs_state = aapoint_delete_fs_state; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c new file mode 100644 index 0000000000..0ac3a240e5 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -0,0 +1,503 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Clipping stage + * + * \author Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" + +#include "draw_context.h" +#include "draw_vs.h" + + +#ifndef IS_NEGATIVE +#define IS_NEGATIVE(X) ((X) < 0.0) +#endif + +#ifndef DIFFERENT_SIGNS +#define DIFFERENT_SIGNS(x, y) ((x) * (y) <= 0.0F && (x) - (y) != 0.0F) +#endif + +#ifndef MAX_CLIPPED_VERTICES +#define MAX_CLIPPED_VERTICES ((2 * (6 + PIPE_MAX_CLIP_PLANES))+1) +#endif + + + +struct clipper { + struct draw_stage stage; /**< base class */ + + /* Basically duplicate some of the flatshading logic here: + */ + boolean flat; + uint num_color_attribs; + uint color_attribs[4]; /* front/back primary/secondary colors */ + + float (*plane)[4]; +}; + + +/* This is a bit confusing: + */ +static INLINE struct clipper *clipper_stage( struct draw_stage *stage ) +{ + return (struct clipper *)stage; +} + + +#define LINTERP(T, OUT, IN) ((OUT) + (T) * ((IN) - (OUT))) + + +/* All attributes are float[4], so this is easy: + */ +static void interp_attr( float *fdst, + float t, + const float *fin, + const float *fout ) +{ + fdst[0] = LINTERP( t, fout[0], fin[0] ); + fdst[1] = LINTERP( t, fout[1], fin[1] ); + fdst[2] = LINTERP( t, fout[2], fin[2] ); + fdst[3] = LINTERP( t, fout[3], fin[3] ); +} + +static void copy_colors( struct draw_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + const struct clipper *clipper = clipper_stage(stage); + uint i; + for (i = 0; i < clipper->num_color_attribs; i++) { + const uint attr = clipper->color_attribs[i]; + COPY_4FV(dst->data[attr], src->data[attr]); + } +} + + + +/* Interpolate between two vertices to produce a third. + */ +static void interp( const struct clipper *clip, + struct vertex_header *dst, + float t, + const struct vertex_header *out, + const struct vertex_header *in ) +{ + const unsigned nr_attrs = clip->stage.draw->num_vs_outputs; + unsigned j; + + /* Vertex header. + */ + { + dst->clipmask = 0; + dst->edgeflag = 0; + dst->pad = 0; + dst->vertex_id = UNDEFINED_VERTEX_ID; + } + + /* Clip coordinates: interpolate normally + */ + { + interp_attr(dst->clip, t, in->clip, out->clip); + } + + /* Do the projective divide and insert window coordinates: + */ + { + const float *pos = dst->clip; + const float *scale = clip->stage.draw->viewport.scale; + const float *trans = clip->stage.draw->viewport.translate; + const float oow = 1.0f / pos[3]; + + dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; + dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; + dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; + dst->data[0][3] = oow; + } + + /* Other attributes + * Note: start at 1 to skip winpos (data[0]) since we just computed + * it above. + */ + for (j = 1; j < nr_attrs; j++) { + interp_attr(dst->data[j], t, in->data[j], out->data[j]); + } +} + + +static void emit_poly( struct draw_stage *stage, + struct vertex_header **inlist, + unsigned n, + const struct prim_header *origPrim) +{ + struct prim_header header; + unsigned i; + + /* later stages may need the determinant, but only the sign matters */ + header.det = origPrim->det; + + for (i = 2; i < n; i++) { + header.v[0] = inlist[i-1]; + header.v[1] = inlist[i]; + header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ + + { + unsigned tmp1 = header.v[1]->edgeflag; + unsigned tmp2 = header.v[2]->edgeflag; + + if (i != n-1) header.v[1]->edgeflag = 0; + if (i != 2) header.v[2]->edgeflag = 0; + + header.edgeflags = ((header.v[0]->edgeflag << 0) | + (header.v[1]->edgeflag << 1) | + (header.v[2]->edgeflag << 2)); + + if (0) { + const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + uint j, k; + debug_printf("Clipped tri:\n"); + for (j = 0; j < 3; j++) { + for (k = 0; k < vs->info.num_outputs; k++) { + debug_printf(" Vert %d: Attr %d: %f %f %f %f\n", j, k, + header.v[j]->data[k][0], + header.v[j]->data[k][1], + header.v[j]->data[k][2], + header.v[j]->data[k][3]); + } + } + } + + stage->next->tri( stage->next, &header ); + + header.v[1]->edgeflag = tmp1; + header.v[2]->edgeflag = tmp2; + } + } +} + + + + +/* Clip a triangle against the viewport and user clip planes. + */ +static void +do_clip_tri( struct draw_stage *stage, + struct prim_header *header, + unsigned clipmask ) +{ + struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *a[MAX_CLIPPED_VERTICES]; + struct vertex_header *b[MAX_CLIPPED_VERTICES]; + struct vertex_header **inlist = a; + struct vertex_header **outlist = b; + unsigned tmpnr = 0; + unsigned n = 3; + unsigned i; + + inlist[0] = header->v[0]; + inlist[1] = header->v[1]; + inlist[2] = header->v[2]; + + while (clipmask && n >= 3) { + const unsigned plane_idx = ffs(clipmask)-1; + const float *plane = clipper->plane[plane_idx]; + struct vertex_header *vert_prev = inlist[0]; + float dp_prev = dot4( vert_prev->clip, plane ); + unsigned outcount = 0; + + clipmask &= ~(1<clip, plane ); + + if (!IS_NEGATIVE(dp_prev)) { + outlist[outcount++] = vert_prev; + } + + if (DIFFERENT_SIGNS(dp, dp_prev)) { + struct vertex_header *new_vert = clipper->stage.tmp[tmpnr++]; + outlist[outcount++] = new_vert; + + if (IS_NEGATIVE(dp)) { + /* Going out of bounds. Avoid division by zero as we + * know dp != dp_prev from DIFFERENT_SIGNS, above. + */ + float t = dp / (dp - dp_prev); + interp( clipper, new_vert, t, vert, vert_prev ); + + /* Force edgeflag true in this case: + */ + new_vert->edgeflag = 1; + } else { + /* Coming back in. + */ + float t = dp_prev / (dp_prev - dp); + interp( clipper, new_vert, t, vert_prev, vert ); + + /* Copy starting vert's edgeflag: + */ + new_vert->edgeflag = vert_prev->edgeflag; + } + } + + vert_prev = vert; + dp_prev = dp; + } + + { + struct vertex_header **tmp = inlist; + inlist = outlist; + outlist = tmp; + n = outcount; + } + } + + /* If flat-shading, copy color to new provoking vertex. + */ + if (clipper->flat && inlist[0] != header->v[2]) { + if (1) { + inlist[0] = dup_vert(stage, inlist[0], tmpnr++); + } + + copy_colors(stage, inlist[0], header->v[2]); + } + + + + /* Emit the polygon as triangles to the setup stage: + */ + if (n >= 3) + emit_poly( stage, inlist, n, header ); +} + + +/* Clip a line against the viewport and user clip planes. + */ +static void +do_clip_line( struct draw_stage *stage, + struct prim_header *header, + unsigned clipmask ) +{ + const struct clipper *clipper = clipper_stage( stage ); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const float *pos0 = v0->clip; + const float *pos1 = v1->clip; + float t0 = 0.0F; + float t1 = 0.0F; + struct prim_header newprim; + + while (clipmask) { + const unsigned plane_idx = ffs(clipmask)-1; + const float *plane = clipper->plane[plane_idx]; + const float dp0 = dot4( pos0, plane ); + const float dp1 = dot4( pos1, plane ); + + if (dp1 < 0.0F) { + float t = dp1 / (dp1 - dp0); + t1 = MAX2(t1, t); + } + + if (dp0 < 0.0F) { + float t = dp0 / (dp0 - dp1); + t0 = MAX2(t0, t); + } + + if (t0 + t1 >= 1.0F) + return; /* discard */ + + clipmask &= ~(1 << plane_idx); /* turn off this plane's bit */ + } + + if (v0->clipmask) { + interp( clipper, stage->tmp[0], t0, v0, v1 ); + + if (clipper->flat) + copy_colors(stage, stage->tmp[0], v0); + + newprim.v[0] = stage->tmp[0]; + } + else { + newprim.v[0] = v0; + } + + if (v1->clipmask) { + interp( clipper, stage->tmp[1], t1, v1, v0 ); + newprim.v[1] = stage->tmp[1]; + } + else { + newprim.v[1] = v1; + } + + stage->next->line( stage->next, &newprim ); +} + + +static void +clip_point( struct draw_stage *stage, + struct prim_header *header ) +{ + if (header->v[0]->clipmask == 0) + stage->next->point( stage->next, header ); +} + + +static void +clip_line( struct draw_stage *stage, + struct prim_header *header ) +{ + unsigned clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask); + + if (clipmask == 0) { + /* no clipping needed */ + stage->next->line( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask) == 0) { + do_clip_line(stage, header, clipmask); + } + /* else, totally clipped */ +} + + +static void +clip_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + unsigned clipmask = (header->v[0]->clipmask | + header->v[1]->clipmask | + header->v[2]->clipmask); + + if (clipmask == 0) { + /* no clipping needed */ + stage->next->tri( stage->next, header ); + } + else if ((header->v[0]->clipmask & + header->v[1]->clipmask & + header->v[2]->clipmask) == 0) { + do_clip_tri(stage, header, clipmask); + } +} + +/* Update state. Could further delay this until we hit the first + * primitive that really requires clipping. + */ +static void +clip_init_state( struct draw_stage *stage ) +{ + struct clipper *clipper = clipper_stage( stage ); + + clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; + + if (clipper->flat) { + const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + uint i; + + clipper->num_color_attribs = 0; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + clipper->color_attribs[clipper->num_color_attribs++] = i; + } + } + } + + stage->tri = clip_tri; + stage->line = clip_line; +} + + + +static void clip_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + clip_init_state( stage ); + stage->tri( stage, header ); +} + +static void clip_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + clip_init_state( stage ); + stage->line( stage, header ); +} + + +static void clip_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->tri = clip_first_tri; + stage->line = clip_first_line; + stage->next->flush( stage->next, flags ); +} + + +static void clip_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void clip_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Allocate a new clipper stage. + * \return pointer to new stage object + */ +struct draw_stage *draw_clip_stage( struct draw_context *draw ) +{ + struct clipper *clipper = CALLOC_STRUCT(clipper); + + draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); + + clipper->stage.draw = draw; + clipper->stage.point = clip_point; + clipper->stage.line = clip_first_line; + clipper->stage.tri = clip_first_tri; + clipper->stage.flush = clip_flush; + clipper->stage.reset_stipple_counter = clip_reset_stipple_counter; + clipper->stage.destroy = clip_destroy; + + clipper->plane = draw->plane; + + return &clipper->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c new file mode 100644 index 0000000000..8177b0ac86 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c @@ -0,0 +1,150 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Drawing stage for polygon culling + */ + +/* Authors: Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + +struct cull_stage { + struct draw_stage stage; + unsigned winding; /**< which winding(s) to cull (one of PIPE_WINDING_x) */ +}; + + +static INLINE struct cull_stage *cull_stage( struct draw_stage *stage ) +{ + return (struct cull_stage *)stage; +} + + + + +static void cull_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + /* Window coords: */ + const float *v0 = header->v[0]->data[0]; + const float *v1 = header->v[1]->data[0]; + const float *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0] - v2[0]; + const float ey = v0[1] - v2[1]; + const float fx = v1[0] - v2[0]; + const float fy = v1[1] - v2[1]; + + /* det = cross(e,f).z */ + header->det = ex * fy - ey * fx; + + if (header->det != 0) { + /* if (det < 0 then Z points toward camera and triangle is + * counter-clockwise winding. + */ + unsigned winding = (header->det < 0) ? PIPE_WINDING_CCW : PIPE_WINDING_CW; + + if ((winding & cull_stage(stage)->winding) == 0) { + /* triangle is not culled, pass to next stage */ + stage->next->tri( stage->next, header ); + } + } +} + +static void cull_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct cull_stage *cull = cull_stage(stage); + + cull->winding = stage->draw->rasterizer->cull_mode; + + stage->tri = cull_tri; + stage->tri( stage, header ); +} + + + +static void cull_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void cull_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void cull_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->tri = cull_first_tri; + stage->next->flush( stage->next, flags ); +} + +static void cull_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void cull_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create a new polygon culling stage. + */ +struct draw_stage *draw_cull_stage( struct draw_context *draw ) +{ + struct cull_stage *cull = CALLOC_STRUCT(cull_stage); + + draw_alloc_temp_verts( &cull->stage, 0 ); + + cull->stage.draw = draw; + cull->stage.next = NULL; + cull->stage.point = cull_point; + cull->stage.line = cull_line; + cull->stage.tri = cull_first_tri; + cull->stage.flush = cull_flush; + cull->stage.reset_stipple_counter = cull_reset_stipple_counter; + cull->stage.destroy = cull_destroy; + + return &cull->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c new file mode 100644 index 0000000000..54baa1fbc9 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c @@ -0,0 +1,248 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "draw_vs.h" + + +/** subclass of draw_stage */ +struct flat_stage +{ + struct draw_stage stage; + + uint num_color_attribs; + uint color_attribs[4]; /* front/back primary/secondary colors */ +}; + + +static INLINE struct flat_stage * +flat_stage(struct draw_stage *stage) +{ + return (struct flat_stage *) stage; +} + + +/** Copy all the color attributes from 'src' vertex to 'dst' vertex */ +static INLINE void copy_colors( struct draw_stage *stage, + struct vertex_header *dst, + const struct vertex_header *src ) +{ + const struct flat_stage *flat = flat_stage(stage); + uint i; + for (i = 0; i < flat->num_color_attribs; i++) { + const uint attr = flat->color_attribs[i]; + COPY_4FV(dst->data[attr], src->data[attr]); + } +} + + +/** Copy all the color attributes from src vertex to dst0 & dst1 vertices */ +static INLINE void copy_colors2( struct draw_stage *stage, + struct vertex_header *dst0, + struct vertex_header *dst1, + const struct vertex_header *src ) +{ + const struct flat_stage *flat = flat_stage(stage); + uint i; + for (i = 0; i < flat->num_color_attribs; i++) { + const uint attr = flat->color_attribs[i]; + COPY_4FV(dst0->data[attr], src->data[attr]); + COPY_4FV(dst1->data[attr], src->data[attr]); + } +} + + +/** + * Flatshade tri. Required for clipping and when unfilled tris are + * active, otherwise handled by hardware. + */ +static void flatshade_tri_0( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + tmp.v[0] = header->v[0]; + tmp.v[1] = dup_vert(stage, header->v[1], 0); + tmp.v[2] = dup_vert(stage, header->v[2], 1); + + copy_colors2(stage, tmp.v[1], tmp.v[2], tmp.v[0]); + + stage->next->tri( stage->next, &tmp ); +} + + +static void flatshade_tri_2( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = header->v[2]; + + copy_colors2(stage, tmp.v[0], tmp.v[1], tmp.v[2]); + + stage->next->tri( stage->next, &tmp ); +} + + + + + +/** + * Flatshade line. Required for clipping. + */ +static void flatshade_line_0( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.v[0] = header->v[0]; + tmp.v[1] = dup_vert(stage, header->v[1], 0); + + copy_colors(stage, tmp.v[1], tmp.v[0]); + + stage->next->line( stage->next, &tmp ); +} + +static void flatshade_line_1( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = header->v[1]; + + copy_colors(stage, tmp.v[0], tmp.v[1]); + + stage->next->line( stage->next, &tmp ); +} + + +/* Flatshade point -- passthrough. + */ +static void flatshade_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void flatshade_init_state( struct draw_stage *stage ) +{ + struct flat_stage *flat = flat_stage(stage); + const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + uint i; + + /* Find which vertex shader outputs are colors, make a list */ + flat->num_color_attribs = 0; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR || + vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + flat->color_attribs[flat->num_color_attribs++] = i; + } + } + + /* Choose flatshade routine according to provoking vertex: + */ + if (stage->draw->rasterizer->flatshade_first) { + stage->line = flatshade_line_0; + stage->tri = flatshade_tri_0; + } + else { + stage->line = flatshade_line_1; + stage->tri = flatshade_tri_2; + } +} + +static void flatshade_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + flatshade_init_state( stage ); + stage->tri( stage, header ); +} + +static void flatshade_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + flatshade_init_state( stage ); + stage->line( stage, header ); +} + + +static void flatshade_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->tri = flatshade_first_tri; + stage->line = flatshade_first_line; + stage->next->flush( stage->next, flags ); +} + + +static void flatshade_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void flatshade_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create flatshading drawing stage. + */ +struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) +{ + struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage); + + draw_alloc_temp_verts( &flatshade->stage, 2 ); + + flatshade->stage.draw = draw; + flatshade->stage.next = NULL; + flatshade->stage.point = flatshade_point; + flatshade->stage.line = flatshade_first_line; + flatshade->stage.tri = flatshade_first_tri; + flatshade->stage.flush = flatshade_flush; + flatshade->stage.reset_stipple_counter = flatshade_reset_stipple_counter; + flatshade->stage.destroy = flatshade_destroy; + + return &flatshade->stage; +} + + diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c new file mode 100644 index 0000000000..dbc676deae --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c @@ -0,0 +1,186 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief polygon offset state + * + * \author Keith Whitwell + * \author Brian Paul + */ + +#include "pipe/p_util.h" +#include "draw_private.h" + + + +struct offset_stage { + struct draw_stage stage; + + float scale; + float units; +}; + + + +static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) +{ + return (struct offset_stage *) stage; +} + + + + + +/** + * Offset tri Z. Some hardware can handle this, but not usually when + * doing unfilled rendering. + */ +static void do_offset_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + float inv_det = 1.0f / header->det; + + /* Window coords: + */ + float *v0 = header->v[0]->data[0]; + float *v1 = header->v[1]->data[0]; + float *v2 = header->v[2]->data[0]; + + /* edge vectors e = v0 - v2, f = v1 - v2 */ + float ex = v0[0] - v2[0]; + float ey = v0[1] - v2[1]; + float ez = v0[2] - v2[2]; + float fx = v1[0] - v2[0]; + float fy = v1[1] - v2[1]; + float fz = v1[2] - v2[2]; + + /* (a,b) = cross(e,f).xy */ + float a = ey*fz - ez*fy; + float b = ez*fx - ex*fz; + + float dzdx = FABSF(a * inv_det); + float dzdy = FABSF(b * inv_det); + + float zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; + + /* + * Note: we're applying the offset and clamping per-vertex. + * Ideally, the offset is applied per-fragment prior to fragment shading. + */ + v0[2] = CLAMP(v0[2] + zoffset, 0.0f, 1.0f); + v1[2] = CLAMP(v1[2] + zoffset, 0.0f, 1.0f); + v2[2] = CLAMP(v2[2] + zoffset, 0.0f, 1.0f); + + stage->next->tri( stage->next, header ); +} + + +static void offset_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + tmp.v[0] = dup_vert(stage, header->v[0], 0); + tmp.v[1] = dup_vert(stage, header->v[1], 1); + tmp.v[2] = dup_vert(stage, header->v[2], 2); + + do_offset_tri( stage, &tmp ); +} + + +static void offset_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct offset_stage *offset = offset_stage(stage); + float mrd = 1.0f / 65535.0f; /* XXX this depends on depthbuffer bits! */ + + offset->units = stage->draw->rasterizer->offset_units * mrd; + offset->scale = stage->draw->rasterizer->offset_scale; + + stage->tri = offset_tri; + stage->tri( stage, header ); +} + + +static void offset_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void offset_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void offset_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->tri = offset_first_tri; + stage->next->flush( stage->next, flags ); +} + + +static void offset_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void offset_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create polygon offset drawing stage. + */ +struct draw_stage *draw_offset_stage( struct draw_context *draw ) +{ + struct offset_stage *offset = CALLOC_STRUCT(offset_stage); + + draw_alloc_temp_verts( &offset->stage, 3 ); + + offset->stage.draw = draw; + offset->stage.next = NULL; + offset->stage.point = offset_point; + offset->stage.line = offset_line; + offset->stage.tri = offset_first_tri; + offset->stage.flush = offset_flush; + offset->stage.reset_stipple_counter = offset_reset_stipple_counter; + offset->stage.destroy = offset_destroy; + + return &offset->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c new file mode 100644 index 0000000000..4dddb72906 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -0,0 +1,746 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Polygon stipple stage: implement polygon stipple with texture map and + * fragment program. The fragment program samples the texture and does + * a fragment kill for the stipple-failing fragments. + * + * Authors: Brian Paul + */ + + +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" + +#include "tgsi/util/tgsi_transform.h" +#include "tgsi/util/tgsi_dump.h" + +#include "draw_context.h" +#include "draw_private.h" + + + +/** + * Subclass of pipe_shader_state to carry extra fragment shader info. + */ +struct pstip_fragment_shader +{ + struct pipe_shader_state state; + void *driver_fs; + void *pstip_fs; + uint sampler_unit; +}; + + +/** + * Subclass of draw_stage + */ +struct pstip_stage +{ + struct draw_stage stage; + + void *sampler_cso; + struct pipe_texture *texture; + uint num_samplers; + uint num_textures; + + /* + * Currently bound state + */ + struct pstip_fragment_shader *fs; + struct { + void *samplers[PIPE_MAX_SAMPLERS]; + struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; + const struct pipe_poly_stipple *stipple; + } state; + + /* + * Driver interface/override functions + */ + void * (*driver_create_fs_state)(struct pipe_context *, + const struct pipe_shader_state *); + void (*driver_bind_fs_state)(struct pipe_context *, void *); + void (*driver_delete_fs_state)(struct pipe_context *, void *); + + void (*driver_bind_sampler_states)(struct pipe_context *, unsigned, void **); + + void (*driver_set_sampler_textures)(struct pipe_context *, unsigned, + struct pipe_texture **); + + void (*driver_set_polygon_stipple)(struct pipe_context *, + const struct pipe_poly_stipple *); + + struct pipe_context *pipe; +}; + + + +/** + * Subclass of tgsi_transform_context, used for transforming the + * user's fragment shader to add the special AA instructions. + */ +struct pstip_transform_context { + struct tgsi_transform_context base; + uint tempsUsed; /**< bitmask */ + int wincoordInput; + int maxInput; + uint samplersUsed; /**< bitfield of samplers used */ + int freeSampler; /** an available sampler for the pstipple */ + int texTemp; /**< temp registers */ + int numImmed; + boolean firstInstruction; +}; + + +/** + * TGSI declaration transform callback. + * Look for a free sampler, a free input attrib, and two free temp regs. + */ +static void +pstip_transform_decl(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl) +{ + struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; + + if (decl->Declaration.File == TGSI_FILE_SAMPLER) { + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + pctx->samplersUsed |= 1 << i; + } + } + else if (decl->Declaration.File == TGSI_FILE_INPUT) { + pctx->maxInput = MAX2(pctx->maxInput, (int) decl->u.DeclarationRange.Last); + if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) + pctx->wincoordInput = (int) decl->u.DeclarationRange.First; + } + else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { + uint i; + for (i = decl->u.DeclarationRange.First; + i <= decl->u.DeclarationRange.Last; i++) { + pctx->tempsUsed |= (1 << i); + } + } + + ctx->emit_declaration(ctx, decl); +} + + +static void +pstip_transform_immed(struct tgsi_transform_context *ctx, + struct tgsi_full_immediate *immed) +{ + struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; + pctx->numImmed++; +} + + +/** + * Find the lowest zero bit in the given word, or -1 if bitfield is all ones. + */ +static int +free_bit(uint bitfield) +{ + int i; + for (i = 0; i < 32; i++) { + if ((bitfield & (1 << i)) == 0) + return i; + } + return -1; +} + + +/** + * TGSI instruction transform callback. + * Replace writes to result.color w/ a temp reg. + * Upon END instruction, insert texture sampling code for antialiasing. + */ +static void +pstip_transform_inst(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst) +{ + struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; + + if (pctx->firstInstruction) { + /* emit our new declarations before the first instruction */ + + struct tgsi_full_declaration decl; + struct tgsi_full_instruction newInst; + uint i; + int wincoordInput; + + /* find free sampler */ + pctx->freeSampler = free_bit(pctx->samplersUsed); + if (pctx->freeSampler >= PIPE_MAX_SAMPLERS) + pctx->freeSampler = PIPE_MAX_SAMPLERS - 1; + + if (pctx->wincoordInput < 0) + wincoordInput = pctx->maxInput + 1; + else + wincoordInput = pctx->wincoordInput; + + /* find one free temp reg */ + for (i = 0; i < 32; i++) { + if ((pctx->tempsUsed & (1 << i)) == 0) { + /* found a free temp */ + if (pctx->texTemp < 0) + pctx->texTemp = i; + else + break; + } + } + assert(pctx->texTemp >= 0); + + if (pctx->wincoordInput < 0) { + /* declare new position input reg */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; + decl.Semantic.SemanticIndex = 0; + decl.Declaration.Interpolate = 1; + decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR; /* XXX? */ + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = wincoordInput; + ctx->emit_declaration(ctx, &decl); + } + + /* declare new sampler */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_SAMPLER; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = pctx->freeSampler; + ctx->emit_declaration(ctx, &decl); + + /* declare new temp regs */ + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_TEMPORARY; + decl.u.DeclarationRange.First = + decl.u.DeclarationRange.Last = pctx->texTemp; + ctx->emit_declaration(ctx, &decl); + + /* emit immediate = {1/32, 1/32, 1, 1} + * The index/position of this immediate will be pctx->numImmed + */ + { + static const float value[4] = { 1.0/32, 1.0/32, 1.0, 1.0 }; + struct tgsi_full_immediate immed; + uint size = 4; + immed = tgsi_default_full_immediate(); + immed.Immediate.Size = 1 + size; /* one for the token itself */ + immed.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; + ctx->emit_immediate(ctx, &immed); + } + + pctx->firstInstruction = FALSE; + + + /* + * Insert new MUL/TEX/KILP instructions at start of program + * Take gl_FragCoord, divide by 32 (stipple size), sample the + * texture and kill fragment if needed. + * + * We'd like to use non-normalized texcoords to index into a RECT + * texture, but we can only use GL_REPEAT wrap mode with normalized + * texcoords. Darn. + */ + + /* MUL texTemp, INPUT[wincoord], 1/32; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_MUL; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp; + newInst.Instruction.NumSrcRegs = 2; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; + newInst.FullSrcRegisters[0].SrcRegister.Index = wincoordInput; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_IMMEDIATE; + newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->numImmed; + ctx->emit_instruction(ctx, &newInst); + + /* TEX texTemp, texTemp, sampler; */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_TEX; + newInst.Instruction.NumDstRegs = 1; + newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp; + newInst.Instruction.NumSrcRegs = 2; + newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp; + newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; + newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->freeSampler; + ctx->emit_instruction(ctx, &newInst); + + /* KILP texTemp; # if texTemp < 0, KILL fragment */ + newInst = tgsi_default_full_instruction(); + newInst.Instruction.Opcode = TGSI_OPCODE_KILP; + newInst.Instruction.NumDstRegs = 0; + newInst.Instruction.NumSrcRegs = 1; + newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; + newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp; + newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; + ctx->emit_instruction(ctx, &newInst); + } + + /* emit this instruction */ + ctx->emit_instruction(ctx, inst); +} + + +/** + * Generate the frag shader we'll use for doing polygon stipple. + * This will be the user's shader prefixed with a TEX and KIL instruction. + */ +static void +generate_pstip_fs(struct pstip_stage *pstip) +{ + const struct pipe_shader_state *orig_fs = &pstip->fs->state; + /*struct draw_context *draw = pstip->stage.draw;*/ + struct pipe_shader_state pstip_fs; + struct pstip_transform_context transform; + +#define MAX 1000 + + pstip_fs = *orig_fs; /* copy to init */ + pstip_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); + + memset(&transform, 0, sizeof(transform)); + transform.wincoordInput = -1; + transform.maxInput = -1; + transform.texTemp = -1; + transform.firstInstruction = TRUE; + transform.base.transform_instruction = pstip_transform_inst; + transform.base.transform_declaration = pstip_transform_decl; + transform.base.transform_immediate = pstip_transform_immed; + + tgsi_transform_shader(orig_fs->tokens, + (struct tgsi_token *) pstip_fs.tokens, + MAX, &transform.base); + +#if 0 /* DEBUG */ + tgsi_dump(orig_fs->tokens, 0); + tgsi_dump(pstip_fs.tokens, 0); +#endif + + pstip->fs->sampler_unit = transform.freeSampler; + assert(pstip->fs->sampler_unit < PIPE_MAX_SAMPLERS); + + pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs); +} + + +/** + * Load texture image with current stipple pattern. + */ +static void +pstip_update_texture(struct pstip_stage *pstip) +{ + static const uint bit31 = 1 << 31; + struct pipe_context *pipe = pstip->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *surface; + const uint *stipple = pstip->state.stipple->stipple; + uint i, j; + ubyte *data; + + surface = screen->get_tex_surface(screen, pstip->texture, 0, 0, 0); + data = pipe_surface_map(surface); + + /* + * Load alpha texture. + * Note: 0 means keep the fragment, 255 means kill it. + * We'll negate the texel value and use KILP which kills if value + * is negative. + */ + for (i = 0; i < 32; i++) { + for (j = 0; j < 32; j++) { + if (stipple[i] & (bit31 >> j)) { + /* fragment "on" */ + data[i * surface->pitch + j] = 0; + } + else { + /* fragment "off" */ + data[i * surface->pitch + j] = 255; + } + } + } + + /* unmap */ + pipe_surface_unmap(surface); + pipe_surface_reference(&surface, NULL); + pipe->texture_update(pipe, pstip->texture, 0, 0x1); +} + + +/** + * Create the texture map we'll use for stippling. + */ +static void +pstip_create_texture(struct pstip_stage *pstip) +{ + struct pipe_context *pipe = pstip->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_texture texTemp; + + memset(&texTemp, 0, sizeof(texTemp)); + texTemp.target = PIPE_TEXTURE_2D; + texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */ + texTemp.last_level = 0; + texTemp.width[0] = 32; + texTemp.height[0] = 32; + texTemp.depth[0] = 1; + texTemp.cpp = 1; + + pstip->texture = screen->texture_create(screen, &texTemp); + assert(pstip->texture->refcount == 1); +} + + +/** + * Create the sampler CSO that'll be used for antialiasing. + * By using a mipmapped texture, we don't have to generate a different + * texture image for each line size. + */ +static void +pstip_create_sampler(struct pstip_stage *pstip) +{ + struct pipe_sampler_state sampler; + struct pipe_context *pipe = pstip->pipe; + + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_REPEAT; + sampler.wrap_t = PIPE_TEX_WRAP_REPEAT; + sampler.wrap_r = PIPE_TEX_WRAP_REPEAT; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; + sampler.normalized_coords = 1; + sampler.min_lod = 0.0f; + sampler.max_lod = 0.0f; + + pstip->sampler_cso = pipe->create_sampler_state(pipe, &sampler); +} + + +/** + * When we're about to draw our first AA line in a batch, this function is + * called to tell the driver to bind our modified fragment shader. + */ +static void +bind_pstip_fragment_shader(struct pstip_stage *pstip) +{ + if (!pstip->fs->pstip_fs) { + generate_pstip_fs(pstip); + } + pstip->driver_bind_fs_state(pstip->pipe, pstip->fs->pstip_fs); +} + + + +static INLINE struct pstip_stage * +pstip_stage( struct draw_stage *stage ) +{ + return (struct pstip_stage *) stage; +} + + +static void +passthrough_point(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->point(stage->next, header); +} + + +static void +passthrough_line(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->line(stage->next, header); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + + +static void +pstip_first_tri(struct draw_stage *stage, struct prim_header *header) +{ + struct pstip_stage *pstip = pstip_stage(stage); + struct pipe_context *pipe = pstip->pipe; + uint num_samplers; + + assert(stage->draw->rasterizer->poly_stipple_enable); + + /* bind our fragprog */ + bind_pstip_fragment_shader(pstip); + + /* how many samplers? */ + /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ + num_samplers = MAX2(pstip->num_textures, pstip->num_samplers); + num_samplers = MAX2(num_samplers, pstip->fs->sampler_unit + 1); + + /* plug in our sampler, texture */ + pstip->state.samplers[pstip->fs->sampler_unit] = pstip->sampler_cso; + pipe_texture_reference(&pstip->state.textures[pstip->fs->sampler_unit], + pstip->texture); + + assert(num_samplers <= PIPE_MAX_SAMPLERS); + + pstip->driver_bind_sampler_states(pipe, num_samplers, pstip->state.samplers); + pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures); + + /* now really draw first line */ + stage->tri = passthrough_tri; + stage->tri(stage, header); +} + + +static void +pstip_flush(struct draw_stage *stage, unsigned flags) +{ + /*struct draw_context *draw = stage->draw;*/ + struct pstip_stage *pstip = pstip_stage(stage); + struct pipe_context *pipe = pstip->pipe; + + stage->tri = pstip_first_tri; + stage->next->flush( stage->next, flags ); + + /* restore original frag shader */ + pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs); + + /* XXX restore original texture, sampler state */ + pstip->driver_bind_sampler_states(pipe, pstip->num_samplers, + pstip->state.samplers); + pstip->driver_set_sampler_textures(pipe, pstip->num_textures, + pstip->state.textures); +} + + +static void +pstip_reset_stipple_counter(struct draw_stage *stage) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +pstip_destroy(struct draw_stage *stage) +{ + struct pstip_stage *pstip = pstip_stage(stage); + + pstip->pipe->delete_sampler_state(pstip->pipe, pstip->sampler_cso); + + pipe_texture_release(&pstip->texture); + + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +static struct pstip_stage * +draw_pstip_stage(struct draw_context *draw) +{ + struct pstip_stage *pstip = CALLOC_STRUCT(pstip_stage); + + draw_alloc_temp_verts( &pstip->stage, 8 ); + + pstip->stage.draw = draw; + pstip->stage.next = NULL; + pstip->stage.point = passthrough_point; + pstip->stage.line = passthrough_line; + pstip->stage.tri = pstip_first_tri; + pstip->stage.flush = pstip_flush; + pstip->stage.reset_stipple_counter = pstip_reset_stipple_counter; + pstip->stage.destroy = pstip_destroy; + + return pstip; +} + + +static struct pstip_stage * +pstip_stage_from_pipe(struct pipe_context *pipe) +{ + struct draw_context *draw = (struct draw_context *) pipe->draw; + return pstip_stage(draw->pipeline.pstipple); +} + + +/** + * This function overrides the driver's create_fs_state() function and + * will typically be called by the state tracker. + */ +static void * +pstip_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *fs) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + struct pstip_fragment_shader *aafs = CALLOC_STRUCT(pstip_fragment_shader); + + if (aafs) { + aafs->state = *fs; + + /* pass-through */ + aafs->driver_fs = pstip->driver_create_fs_state(pstip->pipe, fs); + } + + return aafs; +} + + +static void +pstip_bind_fs_state(struct pipe_context *pipe, void *fs) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs; + /* save current */ + pstip->fs = aafs; + /* pass-through */ + pstip->driver_bind_fs_state(pstip->pipe, + (aafs ? aafs->driver_fs : NULL)); +} + + +static void +pstip_delete_fs_state(struct pipe_context *pipe, void *fs) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs; + /* pass-through */ + pstip->driver_delete_fs_state(pstip->pipe, aafs->driver_fs); + FREE(aafs); +} + + +static void +pstip_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **sampler) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + uint i; + + /* save current */ + memcpy(pstip->state.samplers, sampler, num * sizeof(void *)); + for (i = num; i < PIPE_MAX_SAMPLERS; i++) { + pstip->state.samplers[i] = NULL; + } + + pstip->num_samplers = num; + /* pass-through */ + pstip->driver_bind_sampler_states(pstip->pipe, num, sampler); +} + + +static void +pstip_set_sampler_textures(struct pipe_context *pipe, + unsigned num, struct pipe_texture **texture) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + uint i; + + /* save current */ + for (i = 0; i < num; i++) { + pipe_texture_reference(&pstip->state.textures[i], texture[i]); + } + for (; i < PIPE_MAX_SAMPLERS; i++) { + pipe_texture_reference(&pstip->state.textures[i], NULL); + } + + pstip->num_textures = num; + + /* pass-through */ + pstip->driver_set_sampler_textures(pstip->pipe, num, texture); +} + + +static void +pstip_set_polygon_stipple(struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple) +{ + struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + /* save current */ + pstip->state.stipple = stipple; + /* pass-through */ + pstip->driver_set_polygon_stipple(pstip->pipe, stipple); + + pstip_update_texture(pstip); +} + + + +/** + * Called by drivers that want to install this AA line prim stage + * into the draw module's pipeline. This will not be used if the + * hardware has native support for AA lines. + */ +void +draw_install_pstipple_stage(struct draw_context *draw, + struct pipe_context *pipe) +{ + struct pstip_stage *pstip; + + pipe->draw = (void *) draw; + + /* + * Create / install AA line drawing / prim stage + */ + pstip = draw_pstip_stage( draw ); + assert(pstip); + draw->pipeline.pstipple = &pstip->stage; + + pstip->pipe = pipe; + + /* create special texture, sampler state */ + pstip_create_texture(pstip); + pstip_create_sampler(pstip); + + /* save original driver functions */ + pstip->driver_create_fs_state = pipe->create_fs_state; + pstip->driver_bind_fs_state = pipe->bind_fs_state; + pstip->driver_delete_fs_state = pipe->delete_fs_state; + + pstip->driver_bind_sampler_states = pipe->bind_sampler_states; + pstip->driver_set_sampler_textures = pipe->set_sampler_textures; + pstip->driver_set_polygon_stipple = pipe->set_polygon_stipple; + + /* override the driver's functions */ + pipe->create_fs_state = pstip_create_fs_state; + pipe->bind_fs_state = pstip_bind_fs_state; + pipe->delete_fs_state = pstip_delete_fs_state; + + pipe->bind_sampler_states = pstip_bind_sampler_states; + pipe->set_sampler_textures = pstip_set_sampler_textures; + pipe->set_polygon_stipple = pstip_set_polygon_stipple; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c new file mode 100644 index 0000000000..506f33512c --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c @@ -0,0 +1,239 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +/* Implement line stipple by cutting lines up into smaller lines. + * There are hundreds of ways to implement line stipple, this is one + * choice that should work in all situations, requires no state + * manipulations, but with a penalty in terms of large amounts of + * generated geometry. + */ + + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +/** Subclass of draw_stage */ +struct stipple_stage { + struct draw_stage stage; + float counter; + uint pattern; + uint factor; +}; + + +static INLINE struct stipple_stage * +stipple_stage(struct draw_stage *stage) +{ + return (struct stipple_stage *) stage; +} + + +/** + * Compute interpolated vertex attributes for 'dst' at position 't' + * between 'v0' and 'v1'. + * XXX using linear interpolation for all attribs at this time. + */ +static void +screen_interp( struct draw_context *draw, + struct vertex_header *dst, + float t, + const struct vertex_header *v0, + const struct vertex_header *v1 ) +{ + uint attr; + for (attr = 0; attr < draw->num_vs_outputs; attr++) { + const float *val0 = v0->data[attr]; + const float *val1 = v1->data[attr]; + float *newv = dst->data[attr]; + uint i; + for (i = 0; i < 4; i++) { + newv[i] = val0[i] + t * (val1[i] - val0[i]); + } + } +} + + +static void +emit_segment(struct draw_stage *stage, struct prim_header *header, + float t0, float t1) +{ + struct vertex_header *v0new = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1new = dup_vert(stage, header->v[1], 1); + struct prim_header newprim = *header; + + if (t0 > 0.0) { + screen_interp( stage->draw, v0new, t0, header->v[0], header->v[1] ); + newprim.v[0] = v0new; + } + + if (t1 < 1.0) { + screen_interp( stage->draw, v1new, t1, header->v[0], header->v[1] ); + newprim.v[1] = v1new; + } + + stage->next->line( stage->next, &newprim ); +} + + +static INLINE unsigned +stipple_test(int counter, ushort pattern, int factor) +{ + int b = (counter / factor) & 0xf; + return (1 << b) & pattern; +} + + +static void +stipple_line(struct draw_stage *stage, struct prim_header *header) +{ + struct stipple_stage *stipple = stipple_stage(stage); + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + const float *pos0 = v0->data[0]; + const float *pos1 = v1->data[0]; + float start = 0; + int state = 0; + + float x0 = pos0[0]; + float x1 = pos1[0]; + float y0 = pos0[1]; + float y1 = pos1[1]; + + float dx = x0 > x1 ? x0 - x1 : x1 - x0; + float dy = y0 > y1 ? y0 - y1 : y1 - y0; + + float length = MAX2(dx, dy); + int i; + + /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table. + */ + for (i = 0; i < length; i++) { + int result = stipple_test( (int) stipple->counter+i, + (ushort) stipple->pattern, stipple->factor ); + if (result != state) { + /* changing from "off" to "on" or vice versa */ + if (state) { + if (start != i) { + /* finishing an "on" segment */ + emit_segment( stage, header, start / length, i / length ); + } + } + else { + /* starting an "on" segment */ + start = (float) i; + } + state = result; + } + } + + if (state && start < length) + emit_segment( stage, header, start / length, 1.0 ); + + stipple->counter += length; +} + + +static void +reset_stipple_counter(struct draw_stage *stage) +{ + struct stipple_stage *stipple = stipple_stage(stage); + stipple->counter = 0; + stage->next->reset_stipple_counter( stage->next ); +} + + +static void +stipple_first_line(struct draw_stage *stage, + struct prim_header *header) +{ + struct stipple_stage *stipple = stipple_stage(stage); + struct draw_context *draw = stage->draw; + + stipple->pattern = draw->rasterizer->line_stipple_pattern; + stipple->factor = draw->rasterizer->line_stipple_factor + 1; + + stage->line = stipple_line; + stage->line( stage, header ); +} + + +static void +stipple_flush(struct draw_stage *stage, unsigned flags) +{ + stage->line = stipple_first_line; + stage->next->flush( stage->next, flags ); +} + + +static void +passthrough_point(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->point( stage->next, header ); +} + + +static void +passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + +static void +stipple_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create line stippler stage + */ +struct draw_stage *draw_stipple_stage( struct draw_context *draw ) +{ + struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); + + draw_alloc_temp_verts( &stipple->stage, 2 ); + + stipple->stage.draw = draw; + stipple->stage.next = NULL; + stipple->stage.point = passthrough_point; + stipple->stage.line = stipple_first_line; + stipple->stage.tri = passthrough_tri; + stipple->stage.reset_stipple_counter = reset_stipple_counter; + stipple->stage.flush = stipple_flush; + stipple->stage.destroy = stipple_destroy; + + return &stipple->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c new file mode 100644 index 0000000000..01d905c153 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c @@ -0,0 +1,203 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_vs.h" + + +struct twoside_stage { + struct draw_stage stage; + float sign; /**< +1 or -1 */ + uint attrib_front0, attrib_back0; + uint attrib_front1, attrib_back1; +}; + + +static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage ) +{ + return (struct twoside_stage *)stage; +} + + + + +/** + * Copy back color(s) to front color(s). + */ +static INLINE struct vertex_header * +copy_bfc( struct twoside_stage *twoside, + const struct vertex_header *v, + unsigned idx ) +{ + struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); + + if (twoside->attrib_back0) { + COPY_4FV(tmp->data[twoside->attrib_front0], + tmp->data[twoside->attrib_back0]); + } + if (twoside->attrib_back1) { + COPY_4FV(tmp->data[twoside->attrib_front1], + tmp->data[twoside->attrib_back1]); + } + + return tmp; +} + + +/* Twoside tri: + */ +static void twoside_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + + if (header->det * twoside->sign < 0.0) { + /* this is a back-facing triangle */ + struct prim_header tmp; + + tmp.det = header->det; + tmp.edgeflags = header->edgeflags; + /* copy back attribs to front attribs */ + tmp.v[0] = copy_bfc(twoside, header->v[0], 0); + tmp.v[1] = copy_bfc(twoside, header->v[1], 1); + tmp.v[2] = copy_bfc(twoside, header->v[2], 2); + + stage->next->tri( stage->next, &tmp ); + } + else { + stage->next->tri( stage->next, header ); + } +} + + +static void twoside_line( struct draw_stage *stage, + struct prim_header *header ) +{ + /* pass-through */ + stage->next->line( stage->next, header ); +} + + +static void twoside_point( struct draw_stage *stage, + struct prim_header *header ) +{ + /* pass-through */ + stage->next->point( stage->next, header ); +} + + +static void twoside_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct twoside_stage *twoside = twoside_stage(stage); + const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + uint i; + + twoside->attrib_front0 = 0; + twoside->attrib_front1 = 0; + twoside->attrib_back0 = 0; + twoside->attrib_back1 = 0; + + /* Find which vertex shader outputs are front/back colors */ + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { + if (vs->info.output_semantic_index[i] == 0) + twoside->attrib_front0 = i; + else + twoside->attrib_front1 = i; + } + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { + if (vs->info.output_semantic_index[i] == 0) + twoside->attrib_back0 = i; + else + twoside->attrib_back1 = i; + } + } + + if (!twoside->attrib_back0) + twoside->attrib_front0 = 0; + + if (!twoside->attrib_back1) + twoside->attrib_front1 = 0; + + /* + * We'll multiply the primitive's determinant by this sign to determine + * if the triangle is back-facing (negative). + * sign = -1 for CCW, +1 for CW + */ + twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; + + stage->tri = twoside_tri; + stage->tri( stage, header ); +} + + +static void twoside_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->tri = twoside_first_tri; + stage->next->flush( stage->next, flags ); +} + + +static void twoside_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void twoside_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create twoside pipeline stage. + */ +struct draw_stage *draw_twoside_stage( struct draw_context *draw ) +{ + struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); + + draw_alloc_temp_verts( &twoside->stage, 3 ); + + twoside->stage.draw = draw; + twoside->stage.next = NULL; + twoside->stage.point = twoside_point; + twoside->stage.line = twoside_line; + twoside->stage.tri = twoside_first_tri; + twoside->stage.flush = twoside_flush; + twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; + twoside->stage.destroy = twoside_destroy; + + return &twoside->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c new file mode 100644 index 0000000000..b07860cd9e --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c @@ -0,0 +1,206 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \brief Drawing stage for handling glPolygonMode(line/point). + * Convert triangles to points or lines as needed. + */ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + + +struct unfilled_stage { + struct draw_stage stage; + + /** [0] = front face, [1] = back face. + * legal values: PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE, + * and PIPE_POLYGON_MODE_POINT, + */ + unsigned mode[2]; +}; + + +static INLINE struct unfilled_stage *unfilled_stage( struct draw_stage *stage ) +{ + return (struct unfilled_stage *)stage; +} + + + +static void point( struct draw_stage *stage, + struct vertex_header *v0 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + stage->next->point( stage->next, &tmp ); +} + +static void line( struct draw_stage *stage, + struct vertex_header *v0, + struct vertex_header *v1 ) +{ + struct prim_header tmp; + tmp.v[0] = v0; + tmp.v[1] = v1; + stage->next->line( stage->next, &tmp ); +} + + +static void points( struct draw_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + + if (header->edgeflags & 0x1) point( stage, v0 ); + if (header->edgeflags & 0x2) point( stage, v1 ); + if (header->edgeflags & 0x4) point( stage, v2 ); +} + + +static void lines( struct draw_stage *stage, + struct prim_header *header ) +{ + struct vertex_header *v0 = header->v[0]; + struct vertex_header *v1 = header->v[1]; + struct vertex_header *v2 = header->v[2]; + +#if 0 + assert(((header->edgeflags & 0x1) >> 0) == header->v[0]->edgeflag); + assert(((header->edgeflags & 0x2) >> 1) == header->v[1]->edgeflag); + assert(((header->edgeflags & 0x4) >> 2) == header->v[2]->edgeflag); +#endif + + if (header->edgeflags & 0x4) line( stage, v2, v0 ); + if (header->edgeflags & 0x1) line( stage, v0, v1 ); + if (header->edgeflags & 0x2) line( stage, v1, v2 ); +} + + +/* Unfilled tri: + * + * Note edgeflags in the vertex struct is not sufficient as we will + * need to manipulate them when decomposing primitives??? + */ +static void unfilled_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + unsigned mode = unfilled->mode[header->det >= 0.0]; + + switch (mode) { + case PIPE_POLYGON_MODE_FILL: + stage->next->tri( stage->next, header ); + break; + case PIPE_POLYGON_MODE_LINE: + lines( stage, header ); + break; + case PIPE_POLYGON_MODE_POINT: + points( stage, header ); + break; + default: + assert(0); + } +} + + +static void unfilled_first_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct unfilled_stage *unfilled = unfilled_stage(stage); + + unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */ + unfilled->mode[1] = stage->draw->rasterizer->fill_cw; /* back */ + + stage->tri = unfilled_tri; + stage->tri( stage, header ); +} + + +static void unfilled_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line( stage->next, header ); +} + + +static void unfilled_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void unfilled_flush( struct draw_stage *stage, + unsigned flags ) +{ + stage->next->flush( stage->next, flags ); + + stage->tri = unfilled_first_tri; +} + + +static void unfilled_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void unfilled_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +/** + * Create unfilled triangle stage. + */ +struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) +{ + struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); + + draw_alloc_temp_verts( &unfilled->stage, 0 ); + + unfilled->stage.draw = draw; + unfilled->stage.next = NULL; + unfilled->stage.tmp = NULL; + unfilled->stage.point = unfilled_point; + unfilled->stage.line = unfilled_line; + unfilled->stage.tri = unfilled_first_tri; + unfilled->stage.flush = unfilled_flush; + unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; + unfilled->stage.destroy = unfilled_destroy; + + return &unfilled->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c new file mode 100644 index 0000000000..e163e078f0 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c @@ -0,0 +1,312 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "draw_private.h" + +static boolean points( unsigned prim ) +{ + return (prim == PIPE_PRIM_POINTS); +} + +static boolean lines( unsigned prim ) +{ + return (prim == PIPE_PRIM_LINES || + prim == PIPE_PRIM_LINE_STRIP || + prim == PIPE_PRIM_LINE_LOOP); +} + +static boolean triangles( unsigned prim ) +{ + return prim >= PIPE_PRIM_TRIANGLES; +} + +/** + * Check if we need any special pipeline stages, or whether + * prims/verts can go through untouched. Don't test for bypass + * clipping or vs modes, this function is just about the primitive + * pipeline stages. + */ +boolean +draw_need_pipeline(const struct draw_context *draw, + unsigned int prim ) +{ + /* Don't have to worry about triangles turning into lines/points + * and triggering the pipeline, because we have to trigger the + * pipeline *anyway* if unfilled mode is active. + */ + if (lines(prim)) + { + /* line stipple */ + if (draw->rasterizer->line_stipple_enable && draw->line_stipple) + return TRUE; + + /* wide lines */ + if (draw->rasterizer->line_width > draw->wide_line_threshold) + return TRUE; + + /* AA lines */ + if (draw->rasterizer->line_smooth && draw->pipeline.aaline) + return TRUE; + } + + if (points(prim)) + { + /* large points */ + if (draw->rasterizer->point_size > draw->wide_point_threshold) + return TRUE; + + /* AA points */ + if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) + return TRUE; + + /* point sprites */ + if (draw->rasterizer->point_sprite && draw->point_sprite) + return TRUE; + } + + + if (triangles(prim)) + { + /* polygon stipple */ + if (draw->rasterizer->poly_stipple_enable && draw->pipeline.pstipple) + return TRUE; + + /* unfilled polygons */ + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) + return TRUE; + + /* polygon offset */ + if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) + return TRUE; + + /* two-side lighting */ + if (draw->rasterizer->light_twoside) + return TRUE; + } + + /* polygon cull - this is difficult - hardware can cull just fine + * most of the time (though sometimes CULL_NEITHER is unsupported. + * + * Generally this isn't a reason to require the pipeline, though. + * + if (draw->rasterizer->cull_mode) + return TRUE; + */ + + return FALSE; +} + + + +/** + * Rebuild the rendering pipeline. + */ +static struct draw_stage *validate_pipeline( struct draw_stage *stage ) +{ + struct draw_context *draw = stage->draw; + struct draw_stage *next = draw->pipeline.rasterize; + int need_det = 0; + int precalc_flat = 0; + boolean wide_lines, wide_points; + + /* Set the validate's next stage to the rasterize stage, so that it + * can be found later if needed for flushing. + */ + stage->next = next; + + /* drawing wide lines? */ + wide_lines = (draw->rasterizer->line_width > draw->wide_line_threshold + && !draw->rasterizer->line_smooth); + + /* drawing large points? */ + if (draw->rasterizer->point_sprite && draw->point_sprite) + wide_points = TRUE; + else if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) + wide_points = FALSE; + else if (draw->rasterizer->point_size > draw->wide_point_threshold) + wide_points = TRUE; + else + wide_points = FALSE; + + /* + * NOTE: we build up the pipeline in end-to-start order. + * + * TODO: make the current primitive part of the state and build + * shorter pipelines for lines & points. + */ + + if (draw->rasterizer->line_smooth && draw->pipeline.aaline) { + draw->pipeline.aaline->next = next; + next = draw->pipeline.aaline; + } + + if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) { + draw->pipeline.aapoint->next = next; + next = draw->pipeline.aapoint; + } + + if (wide_lines) { + draw->pipeline.wide_line->next = next; + next = draw->pipeline.wide_line; + precalc_flat = 1; + } + + if (wide_points || draw->rasterizer->point_sprite) { + draw->pipeline.wide_point->next = next; + next = draw->pipeline.wide_point; + } + + if (draw->rasterizer->line_stipple_enable && draw->line_stipple) { + draw->pipeline.stipple->next = next; + next = draw->pipeline.stipple; + precalc_flat = 1; /* only needed for lines really */ + } + + if (draw->rasterizer->poly_stipple_enable + && draw->pipeline.pstipple) { + draw->pipeline.pstipple->next = next; + next = draw->pipeline.pstipple; + } + + if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { + draw->pipeline.unfilled->next = next; + next = draw->pipeline.unfilled; + precalc_flat = 1; /* only needed for triangles really */ + need_det = 1; + } + + if (draw->rasterizer->flatshade && precalc_flat) { + draw->pipeline.flatshade->next = next; + next = draw->pipeline.flatshade; + } + + if (draw->rasterizer->offset_cw || + draw->rasterizer->offset_ccw) { + draw->pipeline.offset->next = next; + next = draw->pipeline.offset; + need_det = 1; + } + + if (draw->rasterizer->light_twoside) { + draw->pipeline.twoside->next = next; + next = draw->pipeline.twoside; + need_det = 1; + } + + /* Always run the cull stage as we calculate determinant there + * also. + * + * This can actually be a win as culling out the triangles can lead + * to less work emitting vertices, smaller vertex buffers, etc. + * It's difficult to say whether this will be true in general. + */ + if (need_det || draw->rasterizer->cull_mode) { + draw->pipeline.cull->next = next; + next = draw->pipeline.cull; + } + + /* Clip stage + */ + if (!draw->rasterizer->bypass_clipping) + { + draw->pipeline.clip->next = next; + next = draw->pipeline.clip; + } + + + draw->pipeline.first = next; + return next; +} + +static void validate_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->tri( pipeline, header ); +} + +static void validate_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->line( pipeline, header ); +} + +static void validate_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->point( pipeline, header ); +} + +static void validate_reset_stipple_counter( struct draw_stage *stage ) +{ + struct draw_stage *pipeline = validate_pipeline( stage ); + pipeline->reset_stipple_counter( pipeline ); +} + +static void validate_flush( struct draw_stage *stage, + unsigned flags ) +{ + /* May need to pass a backend flush on to the rasterize stage. + */ + if (stage->next) + stage->next->flush( stage->next, flags ); +} + + +static void validate_destroy( struct draw_stage *stage ) +{ + FREE( stage ); +} + + +/** + * Create validate pipeline stage. + */ +struct draw_stage *draw_validate_stage( struct draw_context *draw ) +{ + struct draw_stage *stage = CALLOC_STRUCT(draw_stage); + + stage->draw = draw; + stage->next = NULL; + stage->point = validate_point; + stage->line = validate_line; + stage->tri = validate_tri; + stage->flush = validate_flush; + stage->reset_stipple_counter = validate_reset_stipple_counter; + stage->destroy = validate_destroy; + + return stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c new file mode 100644 index 0000000000..30dceeb43d --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -0,0 +1,529 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Vertex buffer drawing stage. + * + * \author José Fonseca + * \author Keith Whitwell + */ + + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" + +#include "draw_vbuf.h" +#include "draw_private.h" +#include "draw_vertex.h" +#include "translate/translate.h" + + +/** + * Vertex buffer emit stage. + */ +struct vbuf_stage { + struct draw_stage stage; /**< This must be first (base class) */ + + struct vbuf_render *render; + + const struct vertex_info *vinfo; + + /** Vertex size in bytes */ + unsigned vertex_size; + + struct translate *translate; + + /* FIXME: we have no guarantee that 'unsigned' is 32bit */ + + /** Vertices in hardware format */ + unsigned *vertices; + unsigned *vertex_ptr; + unsigned max_vertices; + unsigned nr_vertices; + + /** Indices */ + ushort *indices; + unsigned max_indices; + unsigned nr_indices; + + /* Cache point size somewhere it's address won't change: + */ + float point_size; +}; + + +/** + * Basically a cast wrapper. + */ +static INLINE struct vbuf_stage * +vbuf_stage( struct draw_stage *stage ) +{ + assert(stage); + return (struct vbuf_stage *)stage; +} + + +static void vbuf_flush_indices( struct vbuf_stage *vbuf ); +static void vbuf_flush_vertices( struct vbuf_stage *vbuf ); +static void vbuf_alloc_vertices( struct vbuf_stage *vbuf ); + + +static INLINE boolean +overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) +{ + unsigned long used = (unsigned long) ((char *)ptr - (char *)map); + return (used + bytes) > bufsz; +} + + +static INLINE void +check_space( struct vbuf_stage *vbuf, unsigned nr ) +{ + if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { + vbuf_flush_vertices(vbuf); + vbuf_alloc_vertices(vbuf); + } + + if (vbuf->nr_indices + nr > vbuf->max_indices ) + vbuf_flush_indices(vbuf); +} + + +static INLINE void +dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) +{ +// assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); + unsigned i, j; + + for (i = 0; i < vinfo->num_attribs; i++) { + j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + debug_printf("EMIT_OMIT:"); + break; + case EMIT_1F: + debug_printf("EMIT_1F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_1F_PSIZE: + debug_printf("EMIT_1F_PSIZE:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_2F: + debug_printf("EMIT_2F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_3F: + debug_printf("EMIT_3F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + data += sizeof(float); + break; + case EMIT_4F: + debug_printf("EMIT_4F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_4UB: + debug_printf("EMIT_4UB:\t"); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + break; + default: + assert(0); + } + debug_printf("\n"); + } + debug_printf("\n"); +} + + +/** + * Extract the needed fields from post-transformed vertex and emit + * a hardware(driver) vertex. + * Recall that the vertices are constructed by the 'draw' module and + * have a couple of slots at the beginning (1-dword header, 4-dword + * clip pos) that we ignore here. We only use the vertex->data[] fields. + */ +static INLINE ushort +emit_vertex( struct vbuf_stage *vbuf, + struct vertex_header *vertex ) +{ + if(vertex->vertex_id == UNDEFINED_VERTEX_ID) { + /* Hmm - vertices are emitted one at a time - better make sure + * set_buffer is efficient. Consider a special one-shot mode for + * translate. + */ + vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0); + vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr); + + if (0) dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); + + vbuf->vertex_ptr += vbuf->vertex_size/4; + vertex->vertex_id = vbuf->nr_vertices++; + } + + return vertex->vertex_id; +} + + +static void +vbuf_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + check_space( vbuf, 3 ); + + for (i = 0; i < 3; i++) { + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] ); + } +} + + +static void +vbuf_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + unsigned i; + + check_space( vbuf, 2 ); + + for (i = 0; i < 2; i++) { + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] ); + } +} + + +static void +vbuf_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + check_space( vbuf, 1 ); + + vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[0] ); +} + + + + +/** + * Set the prim type for subsequent vertices. + * This may result in a new vertex size. The existing vbuffer (if any) + * will be flushed if needed and a new one allocated. + */ +static void +vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) +{ + struct translate_key hw_key; + unsigned dst_offset; + unsigned i; + + vbuf->render->set_primitive(vbuf->render, prim); + + /* Must do this after set_primitive() above: + * + * XXX: need some state managment to track when this needs to be + * recalculated. The driver should tell us whether there was a + * state change. + */ + vbuf->vinfo = vbuf->render->get_vertex_info(vbuf->render); + + if (vbuf->vertex_size != vbuf->vinfo->size * sizeof(float)) { + vbuf_flush_vertices(vbuf); + vbuf->vertex_size = vbuf->vinfo->size * sizeof(float); + } + + /* Translate from pipeline vertices to hw vertices. + */ + dst_offset = 0; + memset(&hw_key, 0, sizeof(hw_key)); + + for (i = 0; i < vbuf->vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned src_buffer = 0; + unsigned output_format; + unsigned src_offset = (vbuf->vinfo->src_index[i] * 4 * sizeof(float) ); + + switch (vbuf->vinfo->emit[i]) { + case EMIT_4F: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + src_buffer = 1; + src_offset = 0; + break; + case EMIT_4UB: + output_format = PIPE_FORMAT_B8G8R8A8_UNORM; + emit_sz = 4 * sizeof(ubyte); + default: + assert(0); + output_format = PIPE_FORMAT_NONE; + emit_sz = 0; + break; + } + + hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + hw_key.element[i].input_buffer = src_buffer; + hw_key.element[i].input_offset = src_offset; + hw_key.element[i].output_format = output_format; + hw_key.element[i].output_offset = dst_offset; + + dst_offset += emit_sz; + } + + hw_key.nr_elements = vbuf->vinfo->num_attribs; + hw_key.output_stride = vbuf->vinfo->size * 4; + + /* Don't bother with caching at this stage: + */ + if (!vbuf->translate || + memcmp(&vbuf->translate->key, &hw_key, sizeof(hw_key)) != 0) + { + if (vbuf->translate) + vbuf->translate->release(vbuf->translate); + + vbuf->translate = translate_create( &hw_key ); + + vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0); + } + + vbuf->point_size = vbuf->stage.draw->rasterizer->point_size; + + /* Allocate new buffer? + */ + if (!vbuf->vertices) + vbuf_alloc_vertices(vbuf); +} + + +static void +vbuf_first_tri( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + stage->tri = vbuf_tri; + vbuf_set_prim(vbuf, PIPE_PRIM_TRIANGLES); + stage->tri( stage, prim ); +} + + +static void +vbuf_first_line( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + stage->line = vbuf_line; + vbuf_set_prim(vbuf, PIPE_PRIM_LINES); + stage->line( stage, prim ); +} + + +static void +vbuf_first_point( struct draw_stage *stage, + struct prim_header *prim ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + stage->point = vbuf_point; + vbuf_set_prim(vbuf, PIPE_PRIM_POINTS); + stage->point( stage, prim ); +} + + +static void +vbuf_flush_indices( struct vbuf_stage *vbuf ) +{ + if(!vbuf->nr_indices) + return; + + assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == + vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); + + vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); + + vbuf->nr_indices = 0; +} + + +/** + * Flush existing vertex buffer and allocate a new one. + * + * XXX: We separate flush-on-index-full and flush-on-vb-full, but may + * raise issues uploading vertices if the hardware wants to flush when + * we flush. + */ +static void +vbuf_flush_vertices( struct vbuf_stage *vbuf ) +{ + if(vbuf->vertices) { + vbuf_flush_indices(vbuf); + + /* Reset temporary vertices ids */ + if(vbuf->nr_vertices) + draw_reset_vertex_ids( vbuf->stage.draw ); + + /* Free the vertex buffer */ + vbuf->render->release_vertices(vbuf->render, + vbuf->vertices, + vbuf->vertex_size, + vbuf->nr_vertices); + vbuf->max_vertices = vbuf->nr_vertices = 0; + vbuf->vertex_ptr = vbuf->vertices = NULL; + + } +} + + +static void +vbuf_alloc_vertices( struct vbuf_stage *vbuf ) +{ + assert(!vbuf->nr_indices); + assert(!vbuf->vertices); + + /* Allocate a new vertex buffer */ + vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; + vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, + (ushort) vbuf->vertex_size, + (ushort) vbuf->max_vertices); + vbuf->vertex_ptr = vbuf->vertices; +} + + + +static void +vbuf_flush( struct draw_stage *stage, unsigned flags ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + vbuf_flush_indices( vbuf ); + + stage->point = vbuf_first_point; + stage->line = vbuf_first_line; + stage->tri = vbuf_first_tri; + + if (flags & DRAW_FLUSH_BACKEND) + vbuf_flush_vertices( vbuf ); +} + + +static void +vbuf_reset_stipple_counter( struct draw_stage *stage ) +{ + /* XXX: Need to do something here for hardware with linestipple. + */ + (void) stage; +} + + +static void vbuf_destroy( struct draw_stage *stage ) +{ + struct vbuf_stage *vbuf = vbuf_stage( stage ); + + if(vbuf->indices) + align_free( vbuf->indices ); + + if(vbuf->translate) + vbuf->translate->release( vbuf->translate ); + + if (vbuf->render) + vbuf->render->destroy( vbuf->render ); + + FREE( stage ); +} + + +/** + * Create a new primitive vbuf/render stage. + */ +struct draw_stage *draw_vbuf_stage( struct draw_context *draw, + struct vbuf_render *render ) +{ + struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); + + if(!vbuf) + goto fail; + + vbuf->stage.draw = draw; + vbuf->stage.point = vbuf_first_point; + vbuf->stage.line = vbuf_first_line; + vbuf->stage.tri = vbuf_first_tri; + vbuf->stage.flush = vbuf_flush; + vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter; + vbuf->stage.destroy = vbuf_destroy; + + vbuf->render = render; + vbuf->max_indices = MAX2(render->max_indices, UNDEFINED_VERTEX_ID-1); + + vbuf->indices = (ushort *) align_malloc( vbuf->max_indices * + sizeof(vbuf->indices[0]), + 16 ); + if(!vbuf->indices) + goto fail; + + vbuf->vertices = NULL; + vbuf->vertex_ptr = vbuf->vertices; + + return &vbuf->stage; + + fail: + if (vbuf) + vbuf_destroy(&vbuf->stage); + + return NULL; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c new file mode 100644 index 0000000000..9a168ce8bd --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c @@ -0,0 +1,190 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +struct wideline_stage { + struct draw_stage stage; + + float half_line_width; +}; + + + +static INLINE struct wideline_stage *wideline_stage( struct draw_stage *stage ) +{ + return (struct wideline_stage *)stage; +} + + +static void wideline_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + + +static void wideline_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Draw a wide line by drawing a quad (two triangles). + * XXX need to disable polygon stipple. + */ +static void wideline_line( struct draw_stage *stage, + struct prim_header *header ) +{ + /*const struct wideline_stage *wide = wideline_stage(stage);*/ + const float half_width = 0.5f * stage->draw->rasterizer->line_width; + + struct prim_header tri; + + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + const float dx = FABSF(pos0[0] - pos2[0]); + const float dy = FABSF(pos0[1] - pos2[1]); + + /* small tweak to meet GL specification */ + const float bias = 0.125f; + + /* + * Draw wide line as a quad (two tris) by "stretching" the line along + * X or Y. + * We need to tweak coords in several ways to be conformant here. + */ + + if (dx > dy) { + /* x-major line */ + pos0[1] = pos0[1] - half_width - bias; + pos1[1] = pos1[1] + half_width - bias; + pos2[1] = pos2[1] - half_width - bias; + pos3[1] = pos3[1] + half_width - bias; + if (pos0[0] < pos2[0]) { + /* left to right line */ + pos0[0] -= 0.5f; + pos1[0] -= 0.5f; + pos2[0] -= 0.5f; + pos3[0] -= 0.5f; + } + else { + /* right to left line */ + pos0[0] += 0.5f; + pos1[0] += 0.5f; + pos2[0] += 0.5f; + pos3[0] += 0.5f; + } + } + else { + /* y-major line */ + pos0[0] = pos0[0] - half_width + bias; + pos1[0] = pos1[0] + half_width + bias; + pos2[0] = pos2[0] - half_width + bias; + pos3[0] = pos3[0] + half_width + bias; + if (pos0[1] < pos2[1]) { + /* top to bottom line */ + pos0[1] -= 0.5f; + pos1[1] -= 0.5f; + pos2[1] -= 0.5f; + pos3[1] -= 0.5f; + } + else { + /* bottom to top line */ + pos0[1] += 0.5f; + pos1[1] += 0.5f; + pos2[1] += 0.5f; + pos3[1] += 0.5f; + } + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +static void wideline_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->next->flush( stage->next, flags ); +} + + +static void wideline_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void wideline_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +struct draw_stage *draw_wide_line_stage( struct draw_context *draw ) +{ + struct wideline_stage *wide = CALLOC_STRUCT(wideline_stage); + + draw_alloc_temp_verts( &wide->stage, 4 ); + + wide->stage.draw = draw; + wide->stage.next = NULL; + wide->stage.point = wideline_point; + wide->stage.line = wideline_line; + wide->stage.tri = wideline_tri; + wide->stage.flush = wideline_flush; + wide->stage.reset_stipple_counter = wideline_reset_stipple_counter; + wide->stage.destroy = wideline_destroy; + + return &wide->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c new file mode 100644 index 0000000000..3d0add0c1a --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -0,0 +1,281 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_vs.h" + + +struct widepoint_stage { + struct draw_stage stage; + + float half_point_size; + float point_size_min; + float point_size_max; + + float xbias; + float ybias; + + uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; + uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; + uint num_texcoords; + + int psize_slot; +}; + + + +static INLINE struct widepoint_stage * +widepoint_stage( struct draw_stage *stage ) +{ + return (struct widepoint_stage *)stage; +} + + +static void passthrough_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void widepoint_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line(stage->next, header); +} + +static void widepoint_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Set the vertex texcoords for sprite mode. + * Coords may be left untouched or set to a right-side-up or upside-down + * orientation. + */ +static void set_texcoords(const struct widepoint_stage *wide, + struct vertex_header *v, const float tc[4]) +{ + uint i; + for (i = 0; i < wide->num_texcoords; i++) { + if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { + uint j = wide->texcoord_slot[i]; + v->data[j][0] = tc[0]; + if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) + v->data[j][1] = 1.0f - tc[1]; + else + v->data[j][1] = tc[1]; + v->data[j][2] = tc[2]; + v->data[j][3] = tc[3]; + } + } +} + + +/* If there are lots of sprite points (and why wouldn't there be?) it + * would probably be more sensible to change hardware setup to + * optimize this rather than doing the whole thing in software like + * this. + */ +static void widepoint_point( struct draw_stage *stage, + struct prim_header *header ) +{ + const struct widepoint_stage *wide = widepoint_stage(stage); + const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; + float half_size; + float left_adj, right_adj, bot_adj, top_adj; + + struct prim_header tri; + + /* four dups of original vertex */ + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + /* point size is either per-vertex or fixed size */ + if (wide->psize_slot >= 0) { + half_size = header->v[0]->data[wide->psize_slot][0]; + + /* XXX: temporary -- do this in the vertex shader?? + */ + half_size = CLAMP(half_size, + wide->point_size_min, + wide->point_size_max); + + half_size *= 0.5f; + } + else { + half_size = wide->half_point_size; + } + + left_adj = -half_size + wide->xbias; + right_adj = half_size + wide->xbias; + bot_adj = half_size + wide->ybias; + top_adj = -half_size + wide->ybias; + + pos0[0] += left_adj; + pos0[1] += top_adj; + + pos1[0] += left_adj; + pos1[1] += bot_adj; + + pos2[0] += right_adj; + pos2[1] += top_adj; + + pos3[0] += right_adj; + pos3[1] += bot_adj; + + if (sprite) { + static const float tex00[4] = { 0, 0, 0, 1 }; + static const float tex01[4] = { 0, 1, 0, 1 }; + static const float tex11[4] = { 1, 1, 0, 1 }; + static const float tex10[4] = { 1, 0, 0, 1 }; + set_texcoords( wide, v0, tex00 ); + set_texcoords( wide, v1, tex01 ); + set_texcoords( wide, v2, tex10 ); + set_texcoords( wide, v3, tex11 ); + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +static void widepoint_first_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct widepoint_stage *wide = widepoint_stage(stage); + struct draw_context *draw = stage->draw; + + wide->half_point_size = 0.5f * draw->rasterizer->point_size; + wide->point_size_min = draw->rasterizer->point_size_min; + wide->point_size_max = draw->rasterizer->point_size_max; + wide->xbias = 0.0; + wide->ybias = 0.0; + + if (draw->rasterizer->gl_rasterization_rules) { + wide->xbias = 0.125; + } + + /* XXX we won't know the real size if it's computed by the vertex shader! */ + if ((draw->rasterizer->point_size > draw->wide_point_threshold) || + (draw->rasterizer->point_sprite && draw->point_sprite)) { + stage->point = widepoint_point; + } + else { + stage->point = passthrough_point; + } + + if (draw->rasterizer->point_sprite) { + /* find vertex shader texcoord outputs */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i, j = 0; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { + wide->texcoord_slot[j] = i; + wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; + j++; + } + } + wide->num_texcoords = j; + } + + wide->psize_slot = -1; + if (draw->rasterizer->point_size_per_vertex) { + /* find PSIZ vertex output */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + wide->psize_slot = i; + break; + } + } + } + + stage->point( stage, header ); +} + + +static void widepoint_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->point = widepoint_first_point; + stage->next->flush( stage->next, flags ); +} + + +static void widepoint_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void widepoint_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +struct draw_stage *draw_wide_point_stage( struct draw_context *draw ) +{ + struct widepoint_stage *wide = CALLOC_STRUCT(widepoint_stage); + + draw_alloc_temp_verts( &wide->stage, 4 ); + + wide->stage.draw = draw; + wide->stage.next = NULL; + wide->stage.point = widepoint_first_point; + wide->stage.line = widepoint_line; + wide->stage.tri = widepoint_tri; + wide->stage.flush = widepoint_flush; + wide->stage.reset_stipple_counter = widepoint_reset_stipple_counter; + wide->stage.destroy = widepoint_destroy; + + return &wide->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_prims.c b/src/gallium/auxiliary/draw/draw_pipe_wide_prims.c new file mode 100644 index 0000000000..d6bff110b4 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_prims.c @@ -0,0 +1,366 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "pipe/p_defines.h" +#include "pipe/p_shader_tokens.h" +#include "draw_private.h" + + +struct wide_stage { + struct draw_stage stage; + + float half_line_width; + float half_point_size; + + uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; + uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; + uint num_texcoords; + + int psize_slot; +}; + + + +static INLINE struct wide_stage *wide_stage( struct draw_stage *stage ) +{ + return (struct wide_stage *)stage; +} + + +static void passthrough_point( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->point( stage->next, header ); +} + +static void passthrough_line( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->line(stage->next, header); +} + +static void passthrough_tri( struct draw_stage *stage, + struct prim_header *header ) +{ + stage->next->tri(stage->next, header); +} + + +/** + * Draw a wide line by drawing a quad (two triangles). + * XXX need to disable polygon stipple. + */ +static void wide_line( struct draw_stage *stage, + struct prim_header *header ) +{ + const struct wide_stage *wide = wide_stage(stage); + const float half_width = wide->half_line_width; + + struct prim_header tri; + + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + const float dx = FABSF(pos0[0] - pos2[0]); + const float dy = FABSF(pos0[1] - pos2[1]); + + /* + * Draw wide line as a quad (two tris) by "stretching" the line along + * X or Y. + * We need to tweak coords in several ways to be conformant here. + */ + + if (dx > dy) { + /* x-major line */ + pos0[1] = pos0[1] - half_width - 0.25f; + pos1[1] = pos1[1] + half_width - 0.25f; + pos2[1] = pos2[1] - half_width - 0.25f; + pos3[1] = pos3[1] + half_width - 0.25f; + if (pos0[0] < pos2[0]) { + /* left to right line */ + pos0[0] -= 0.5f; + pos1[0] -= 0.5f; + pos2[0] -= 0.5f; + pos3[0] -= 0.5f; + } + else { + /* right to left line */ + pos0[0] += 0.5f; + pos1[0] += 0.5f; + pos2[0] += 0.5f; + pos3[0] += 0.5f; + } + } + else { + /* y-major line */ + pos0[0] = pos0[0] - half_width + 0.25f; + pos1[0] = pos1[0] + half_width + 0.25f; + pos2[0] = pos2[0] - half_width + 0.25f; + pos3[0] = pos3[0] + half_width + 0.25f; + if (pos0[1] < pos2[1]) { + /* top to bottom line */ + pos0[1] -= 0.5f; + pos1[1] -= 0.5f; + pos2[1] -= 0.5f; + pos3[1] -= 0.5f; + } + else { + /* bottom to top line */ + pos0[1] += 0.5f; + pos1[1] += 0.5f; + pos2[1] += 0.5f; + pos3[1] += 0.5f; + } + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +/** + * Set the vertex texcoords for sprite mode. + * Coords may be left untouched or set to a right-side-up or upside-down + * orientation. + */ +static void set_texcoords(const struct wide_stage *wide, + struct vertex_header *v, const float tc[4]) +{ + uint i; + for (i = 0; i < wide->num_texcoords; i++) { + if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { + uint j = wide->texcoord_slot[i]; + v->data[j][0] = tc[0]; + if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) + v->data[j][1] = 1.0f - tc[1]; + else + v->data[j][1] = tc[1]; + v->data[j][2] = tc[2]; + v->data[j][3] = tc[3]; + } + } +} + + +/* If there are lots of sprite points (and why wouldn't there be?) it + * would probably be more sensible to change hardware setup to + * optimize this rather than doing the whole thing in software like + * this. + */ +static void wide_point( struct draw_stage *stage, + struct prim_header *header ) +{ + const struct wide_stage *wide = wide_stage(stage); + const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; + float half_size; + float left_adj, right_adj; + + struct prim_header tri; + + /* four dups of original vertex */ + struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); + struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); + struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); + struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); + + float *pos0 = v0->data[0]; + float *pos1 = v1->data[0]; + float *pos2 = v2->data[0]; + float *pos3 = v3->data[0]; + + /* point size is either per-vertex or fixed size */ + if (wide->psize_slot >= 0) { + half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; + } + else { + half_size = wide->half_point_size; + } + + left_adj = -half_size; /* + 0.25f;*/ + right_adj = half_size; /* + 0.25f;*/ + + pos0[0] += left_adj; + pos0[1] -= half_size; + + pos1[0] += left_adj; + pos1[1] += half_size; + + pos2[0] += right_adj; + pos2[1] -= half_size; + + pos3[0] += right_adj; + pos3[1] += half_size; + + if (sprite) { + static const float tex00[4] = { 0, 0, 0, 1 }; + static const float tex01[4] = { 0, 1, 0, 1 }; + static const float tex11[4] = { 1, 1, 0, 1 }; + static const float tex10[4] = { 1, 0, 0, 1 }; + set_texcoords( wide, v0, tex00 ); + set_texcoords( wide, v1, tex01 ); + set_texcoords( wide, v2, tex10 ); + set_texcoords( wide, v3, tex11 ); + } + + tri.det = header->det; /* only the sign matters */ + tri.v[0] = v0; + tri.v[1] = v2; + tri.v[2] = v3; + stage->next->tri( stage->next, &tri ); + + tri.v[0] = v0; + tri.v[1] = v3; + tri.v[2] = v1; + stage->next->tri( stage->next, &tri ); +} + + +static void wide_first_point( struct draw_stage *stage, + struct prim_header *header ) +{ + struct wide_stage *wide = wide_stage(stage); + struct draw_context *draw = stage->draw; + + wide->half_point_size = 0.5f * draw->rasterizer->point_size; + + /* XXX we won't know the real size if it's computed by the vertex shader! */ + if (draw->rasterizer->point_size > draw->wide_point_threshold) { + stage->point = wide_point; + } + else { + stage->point = passthrough_point; + } + + if (draw->rasterizer->point_sprite) { + /* find vertex shader texcoord outputs */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i, j = 0; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { + wide->texcoord_slot[j] = i; + wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; + j++; + } + } + wide->num_texcoords = j; + } + + wide->psize_slot = -1; + + if (draw->rasterizer->point_size_per_vertex) { + /* find PSIZ vertex output */ + const struct draw_vertex_shader *vs = draw->vertex_shader; + uint i; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { + wide->psize_slot = i; + break; + } + } + } + + stage->point( stage, header ); +} + + + +static void wide_first_line( struct draw_stage *stage, + struct prim_header *header ) +{ + struct wide_stage *wide = wide_stage(stage); + struct draw_context *draw = stage->draw; + + wide->half_line_width = 0.5f * draw->rasterizer->line_width; + + if (draw->rasterizer->line_width != 1.0) { + wide->stage.line = wide_line; + } + else { + wide->stage.line = passthrough_line; + } + + stage->line( stage, header ); +} + + +static void wide_flush( struct draw_stage *stage, unsigned flags ) +{ + stage->line = wide_first_line; + stage->point = wide_first_point; + stage->next->flush( stage->next, flags ); +} + + +static void wide_reset_stipple_counter( struct draw_stage *stage ) +{ + stage->next->reset_stipple_counter( stage->next ); +} + + +static void wide_destroy( struct draw_stage *stage ) +{ + draw_free_temp_verts( stage ); + FREE( stage ); +} + + +struct draw_stage *draw_wide_stage( struct draw_context *draw ) +{ + struct wide_stage *wide = CALLOC_STRUCT(wide_stage); + + draw_alloc_temp_verts( &wide->stage, 4 ); + + wide->stage.draw = draw; + wide->stage.next = NULL; + wide->stage.point = wide_first_point; + wide->stage.line = wide_first_line; + wide->stage.tri = passthrough_tri; + wide->stage.flush = wide_flush; + wide->stage.reset_stipple_counter = wide_reset_stipple_counter; + wide->stage.destroy = wide_destroy; + + return &wide->stage; +} diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c deleted file mode 100644 index 4dddb72906..0000000000 --- a/src/gallium/auxiliary/draw/draw_pstipple.c +++ /dev/null @@ -1,746 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Polygon stipple stage: implement polygon stipple with texture map and - * fragment program. The fragment program samples the texture and does - * a fragment kill for the stipple-failing fragments. - * - * Authors: Brian Paul - */ - - -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" - -#include "tgsi/util/tgsi_transform.h" -#include "tgsi/util/tgsi_dump.h" - -#include "draw_context.h" -#include "draw_private.h" - - - -/** - * Subclass of pipe_shader_state to carry extra fragment shader info. - */ -struct pstip_fragment_shader -{ - struct pipe_shader_state state; - void *driver_fs; - void *pstip_fs; - uint sampler_unit; -}; - - -/** - * Subclass of draw_stage - */ -struct pstip_stage -{ - struct draw_stage stage; - - void *sampler_cso; - struct pipe_texture *texture; - uint num_samplers; - uint num_textures; - - /* - * Currently bound state - */ - struct pstip_fragment_shader *fs; - struct { - void *samplers[PIPE_MAX_SAMPLERS]; - struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; - const struct pipe_poly_stipple *stipple; - } state; - - /* - * Driver interface/override functions - */ - void * (*driver_create_fs_state)(struct pipe_context *, - const struct pipe_shader_state *); - void (*driver_bind_fs_state)(struct pipe_context *, void *); - void (*driver_delete_fs_state)(struct pipe_context *, void *); - - void (*driver_bind_sampler_states)(struct pipe_context *, unsigned, void **); - - void (*driver_set_sampler_textures)(struct pipe_context *, unsigned, - struct pipe_texture **); - - void (*driver_set_polygon_stipple)(struct pipe_context *, - const struct pipe_poly_stipple *); - - struct pipe_context *pipe; -}; - - - -/** - * Subclass of tgsi_transform_context, used for transforming the - * user's fragment shader to add the special AA instructions. - */ -struct pstip_transform_context { - struct tgsi_transform_context base; - uint tempsUsed; /**< bitmask */ - int wincoordInput; - int maxInput; - uint samplersUsed; /**< bitfield of samplers used */ - int freeSampler; /** an available sampler for the pstipple */ - int texTemp; /**< temp registers */ - int numImmed; - boolean firstInstruction; -}; - - -/** - * TGSI declaration transform callback. - * Look for a free sampler, a free input attrib, and two free temp regs. - */ -static void -pstip_transform_decl(struct tgsi_transform_context *ctx, - struct tgsi_full_declaration *decl) -{ - struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; - - if (decl->Declaration.File == TGSI_FILE_SAMPLER) { - uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { - pctx->samplersUsed |= 1 << i; - } - } - else if (decl->Declaration.File == TGSI_FILE_INPUT) { - pctx->maxInput = MAX2(pctx->maxInput, (int) decl->u.DeclarationRange.Last); - if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) - pctx->wincoordInput = (int) decl->u.DeclarationRange.First; - } - else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { - uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { - pctx->tempsUsed |= (1 << i); - } - } - - ctx->emit_declaration(ctx, decl); -} - - -static void -pstip_transform_immed(struct tgsi_transform_context *ctx, - struct tgsi_full_immediate *immed) -{ - struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; - pctx->numImmed++; -} - - -/** - * Find the lowest zero bit in the given word, or -1 if bitfield is all ones. - */ -static int -free_bit(uint bitfield) -{ - int i; - for (i = 0; i < 32; i++) { - if ((bitfield & (1 << i)) == 0) - return i; - } - return -1; -} - - -/** - * TGSI instruction transform callback. - * Replace writes to result.color w/ a temp reg. - * Upon END instruction, insert texture sampling code for antialiasing. - */ -static void -pstip_transform_inst(struct tgsi_transform_context *ctx, - struct tgsi_full_instruction *inst) -{ - struct pstip_transform_context *pctx = (struct pstip_transform_context *) ctx; - - if (pctx->firstInstruction) { - /* emit our new declarations before the first instruction */ - - struct tgsi_full_declaration decl; - struct tgsi_full_instruction newInst; - uint i; - int wincoordInput; - - /* find free sampler */ - pctx->freeSampler = free_bit(pctx->samplersUsed); - if (pctx->freeSampler >= PIPE_MAX_SAMPLERS) - pctx->freeSampler = PIPE_MAX_SAMPLERS - 1; - - if (pctx->wincoordInput < 0) - wincoordInput = pctx->maxInput + 1; - else - wincoordInput = pctx->wincoordInput; - - /* find one free temp reg */ - for (i = 0; i < 32; i++) { - if ((pctx->tempsUsed & (1 << i)) == 0) { - /* found a free temp */ - if (pctx->texTemp < 0) - pctx->texTemp = i; - else - break; - } - } - assert(pctx->texTemp >= 0); - - if (pctx->wincoordInput < 0) { - /* declare new position input reg */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_INPUT; - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; - decl.Semantic.SemanticIndex = 0; - decl.Declaration.Interpolate = 1; - decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR; /* XXX? */ - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = wincoordInput; - ctx->emit_declaration(ctx, &decl); - } - - /* declare new sampler */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_SAMPLER; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = pctx->freeSampler; - ctx->emit_declaration(ctx, &decl); - - /* declare new temp regs */ - decl = tgsi_default_full_declaration(); - decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = pctx->texTemp; - ctx->emit_declaration(ctx, &decl); - - /* emit immediate = {1/32, 1/32, 1, 1} - * The index/position of this immediate will be pctx->numImmed - */ - { - static const float value[4] = { 1.0/32, 1.0/32, 1.0, 1.0 }; - struct tgsi_full_immediate immed; - uint size = 4; - immed = tgsi_default_full_immediate(); - immed.Immediate.Size = 1 + size; /* one for the token itself */ - immed.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; - ctx->emit_immediate(ctx, &immed); - } - - pctx->firstInstruction = FALSE; - - - /* - * Insert new MUL/TEX/KILP instructions at start of program - * Take gl_FragCoord, divide by 32 (stipple size), sample the - * texture and kill fragment if needed. - * - * We'd like to use non-normalized texcoords to index into a RECT - * texture, but we can only use GL_REPEAT wrap mode with normalized - * texcoords. Darn. - */ - - /* MUL texTemp, INPUT[wincoord], 1/32; */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_MUL; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp; - newInst.Instruction.NumSrcRegs = 2; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT; - newInst.FullSrcRegisters[0].SrcRegister.Index = wincoordInput; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_IMMEDIATE; - newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->numImmed; - ctx->emit_instruction(ctx, &newInst); - - /* TEX texTemp, texTemp, sampler; */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_TEX; - newInst.Instruction.NumDstRegs = 1; - newInst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullDstRegisters[0].DstRegister.Index = pctx->texTemp; - newInst.Instruction.NumSrcRegs = 2; - newInst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp; - newInst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER; - newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->freeSampler; - ctx->emit_instruction(ctx, &newInst); - - /* KILP texTemp; # if texTemp < 0, KILL fragment */ - newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_KILP; - newInst.Instruction.NumDstRegs = 0; - newInst.Instruction.NumSrcRegs = 1; - newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; - newInst.FullSrcRegisters[0].SrcRegister.Index = pctx->texTemp; - newInst.FullSrcRegisters[0].SrcRegister.Negate = 1; - ctx->emit_instruction(ctx, &newInst); - } - - /* emit this instruction */ - ctx->emit_instruction(ctx, inst); -} - - -/** - * Generate the frag shader we'll use for doing polygon stipple. - * This will be the user's shader prefixed with a TEX and KIL instruction. - */ -static void -generate_pstip_fs(struct pstip_stage *pstip) -{ - const struct pipe_shader_state *orig_fs = &pstip->fs->state; - /*struct draw_context *draw = pstip->stage.draw;*/ - struct pipe_shader_state pstip_fs; - struct pstip_transform_context transform; - -#define MAX 1000 - - pstip_fs = *orig_fs; /* copy to init */ - pstip_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); - - memset(&transform, 0, sizeof(transform)); - transform.wincoordInput = -1; - transform.maxInput = -1; - transform.texTemp = -1; - transform.firstInstruction = TRUE; - transform.base.transform_instruction = pstip_transform_inst; - transform.base.transform_declaration = pstip_transform_decl; - transform.base.transform_immediate = pstip_transform_immed; - - tgsi_transform_shader(orig_fs->tokens, - (struct tgsi_token *) pstip_fs.tokens, - MAX, &transform.base); - -#if 0 /* DEBUG */ - tgsi_dump(orig_fs->tokens, 0); - tgsi_dump(pstip_fs.tokens, 0); -#endif - - pstip->fs->sampler_unit = transform.freeSampler; - assert(pstip->fs->sampler_unit < PIPE_MAX_SAMPLERS); - - pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs); -} - - -/** - * Load texture image with current stipple pattern. - */ -static void -pstip_update_texture(struct pstip_stage *pstip) -{ - static const uint bit31 = 1 << 31; - struct pipe_context *pipe = pstip->pipe; - struct pipe_screen *screen = pipe->screen; - struct pipe_surface *surface; - const uint *stipple = pstip->state.stipple->stipple; - uint i, j; - ubyte *data; - - surface = screen->get_tex_surface(screen, pstip->texture, 0, 0, 0); - data = pipe_surface_map(surface); - - /* - * Load alpha texture. - * Note: 0 means keep the fragment, 255 means kill it. - * We'll negate the texel value and use KILP which kills if value - * is negative. - */ - for (i = 0; i < 32; i++) { - for (j = 0; j < 32; j++) { - if (stipple[i] & (bit31 >> j)) { - /* fragment "on" */ - data[i * surface->pitch + j] = 0; - } - else { - /* fragment "off" */ - data[i * surface->pitch + j] = 255; - } - } - } - - /* unmap */ - pipe_surface_unmap(surface); - pipe_surface_reference(&surface, NULL); - pipe->texture_update(pipe, pstip->texture, 0, 0x1); -} - - -/** - * Create the texture map we'll use for stippling. - */ -static void -pstip_create_texture(struct pstip_stage *pstip) -{ - struct pipe_context *pipe = pstip->pipe; - struct pipe_screen *screen = pipe->screen; - struct pipe_texture texTemp; - - memset(&texTemp, 0, sizeof(texTemp)); - texTemp.target = PIPE_TEXTURE_2D; - texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */ - texTemp.last_level = 0; - texTemp.width[0] = 32; - texTemp.height[0] = 32; - texTemp.depth[0] = 1; - texTemp.cpp = 1; - - pstip->texture = screen->texture_create(screen, &texTemp); - assert(pstip->texture->refcount == 1); -} - - -/** - * Create the sampler CSO that'll be used for antialiasing. - * By using a mipmapped texture, we don't have to generate a different - * texture image for each line size. - */ -static void -pstip_create_sampler(struct pstip_stage *pstip) -{ - struct pipe_sampler_state sampler; - struct pipe_context *pipe = pstip->pipe; - - memset(&sampler, 0, sizeof(sampler)); - sampler.wrap_s = PIPE_TEX_WRAP_REPEAT; - sampler.wrap_t = PIPE_TEX_WRAP_REPEAT; - sampler.wrap_r = PIPE_TEX_WRAP_REPEAT; - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; - sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST; - sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST; - sampler.normalized_coords = 1; - sampler.min_lod = 0.0f; - sampler.max_lod = 0.0f; - - pstip->sampler_cso = pipe->create_sampler_state(pipe, &sampler); -} - - -/** - * When we're about to draw our first AA line in a batch, this function is - * called to tell the driver to bind our modified fragment shader. - */ -static void -bind_pstip_fragment_shader(struct pstip_stage *pstip) -{ - if (!pstip->fs->pstip_fs) { - generate_pstip_fs(pstip); - } - pstip->driver_bind_fs_state(pstip->pipe, pstip->fs->pstip_fs); -} - - - -static INLINE struct pstip_stage * -pstip_stage( struct draw_stage *stage ) -{ - return (struct pstip_stage *) stage; -} - - -static void -passthrough_point(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->point(stage->next, header); -} - - -static void -passthrough_line(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->line(stage->next, header); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} - - - -static void -pstip_first_tri(struct draw_stage *stage, struct prim_header *header) -{ - struct pstip_stage *pstip = pstip_stage(stage); - struct pipe_context *pipe = pstip->pipe; - uint num_samplers; - - assert(stage->draw->rasterizer->poly_stipple_enable); - - /* bind our fragprog */ - bind_pstip_fragment_shader(pstip); - - /* how many samplers? */ - /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ - num_samplers = MAX2(pstip->num_textures, pstip->num_samplers); - num_samplers = MAX2(num_samplers, pstip->fs->sampler_unit + 1); - - /* plug in our sampler, texture */ - pstip->state.samplers[pstip->fs->sampler_unit] = pstip->sampler_cso; - pipe_texture_reference(&pstip->state.textures[pstip->fs->sampler_unit], - pstip->texture); - - assert(num_samplers <= PIPE_MAX_SAMPLERS); - - pstip->driver_bind_sampler_states(pipe, num_samplers, pstip->state.samplers); - pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures); - - /* now really draw first line */ - stage->tri = passthrough_tri; - stage->tri(stage, header); -} - - -static void -pstip_flush(struct draw_stage *stage, unsigned flags) -{ - /*struct draw_context *draw = stage->draw;*/ - struct pstip_stage *pstip = pstip_stage(stage); - struct pipe_context *pipe = pstip->pipe; - - stage->tri = pstip_first_tri; - stage->next->flush( stage->next, flags ); - - /* restore original frag shader */ - pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs); - - /* XXX restore original texture, sampler state */ - pstip->driver_bind_sampler_states(pipe, pstip->num_samplers, - pstip->state.samplers); - pstip->driver_set_sampler_textures(pipe, pstip->num_textures, - pstip->state.textures); -} - - -static void -pstip_reset_stipple_counter(struct draw_stage *stage) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void -pstip_destroy(struct draw_stage *stage) -{ - struct pstip_stage *pstip = pstip_stage(stage); - - pstip->pipe->delete_sampler_state(pstip->pipe, pstip->sampler_cso); - - pipe_texture_release(&pstip->texture); - - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -static struct pstip_stage * -draw_pstip_stage(struct draw_context *draw) -{ - struct pstip_stage *pstip = CALLOC_STRUCT(pstip_stage); - - draw_alloc_temp_verts( &pstip->stage, 8 ); - - pstip->stage.draw = draw; - pstip->stage.next = NULL; - pstip->stage.point = passthrough_point; - pstip->stage.line = passthrough_line; - pstip->stage.tri = pstip_first_tri; - pstip->stage.flush = pstip_flush; - pstip->stage.reset_stipple_counter = pstip_reset_stipple_counter; - pstip->stage.destroy = pstip_destroy; - - return pstip; -} - - -static struct pstip_stage * -pstip_stage_from_pipe(struct pipe_context *pipe) -{ - struct draw_context *draw = (struct draw_context *) pipe->draw; - return pstip_stage(draw->pipeline.pstipple); -} - - -/** - * This function overrides the driver's create_fs_state() function and - * will typically be called by the state tracker. - */ -static void * -pstip_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *fs) -{ - struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); - struct pstip_fragment_shader *aafs = CALLOC_STRUCT(pstip_fragment_shader); - - if (aafs) { - aafs->state = *fs; - - /* pass-through */ - aafs->driver_fs = pstip->driver_create_fs_state(pstip->pipe, fs); - } - - return aafs; -} - - -static void -pstip_bind_fs_state(struct pipe_context *pipe, void *fs) -{ - struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); - struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs; - /* save current */ - pstip->fs = aafs; - /* pass-through */ - pstip->driver_bind_fs_state(pstip->pipe, - (aafs ? aafs->driver_fs : NULL)); -} - - -static void -pstip_delete_fs_state(struct pipe_context *pipe, void *fs) -{ - struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); - struct pstip_fragment_shader *aafs = (struct pstip_fragment_shader *) fs; - /* pass-through */ - pstip->driver_delete_fs_state(pstip->pipe, aafs->driver_fs); - FREE(aafs); -} - - -static void -pstip_bind_sampler_states(struct pipe_context *pipe, - unsigned num, void **sampler) -{ - struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); - uint i; - - /* save current */ - memcpy(pstip->state.samplers, sampler, num * sizeof(void *)); - for (i = num; i < PIPE_MAX_SAMPLERS; i++) { - pstip->state.samplers[i] = NULL; - } - - pstip->num_samplers = num; - /* pass-through */ - pstip->driver_bind_sampler_states(pstip->pipe, num, sampler); -} - - -static void -pstip_set_sampler_textures(struct pipe_context *pipe, - unsigned num, struct pipe_texture **texture) -{ - struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); - uint i; - - /* save current */ - for (i = 0; i < num; i++) { - pipe_texture_reference(&pstip->state.textures[i], texture[i]); - } - for (; i < PIPE_MAX_SAMPLERS; i++) { - pipe_texture_reference(&pstip->state.textures[i], NULL); - } - - pstip->num_textures = num; - - /* pass-through */ - pstip->driver_set_sampler_textures(pstip->pipe, num, texture); -} - - -static void -pstip_set_polygon_stipple(struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple) -{ - struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); - /* save current */ - pstip->state.stipple = stipple; - /* pass-through */ - pstip->driver_set_polygon_stipple(pstip->pipe, stipple); - - pstip_update_texture(pstip); -} - - - -/** - * Called by drivers that want to install this AA line prim stage - * into the draw module's pipeline. This will not be used if the - * hardware has native support for AA lines. - */ -void -draw_install_pstipple_stage(struct draw_context *draw, - struct pipe_context *pipe) -{ - struct pstip_stage *pstip; - - pipe->draw = (void *) draw; - - /* - * Create / install AA line drawing / prim stage - */ - pstip = draw_pstip_stage( draw ); - assert(pstip); - draw->pipeline.pstipple = &pstip->stage; - - pstip->pipe = pipe; - - /* create special texture, sampler state */ - pstip_create_texture(pstip); - pstip_create_sampler(pstip); - - /* save original driver functions */ - pstip->driver_create_fs_state = pipe->create_fs_state; - pstip->driver_bind_fs_state = pipe->bind_fs_state; - pstip->driver_delete_fs_state = pipe->delete_fs_state; - - pstip->driver_bind_sampler_states = pipe->bind_sampler_states; - pstip->driver_set_sampler_textures = pipe->set_sampler_textures; - pstip->driver_set_polygon_stipple = pipe->set_polygon_stipple; - - /* override the driver's functions */ - pipe->create_fs_state = pstip_create_fs_state; - pipe->bind_fs_state = pstip_bind_fs_state; - pipe->delete_fs_state = pstip_delete_fs_state; - - pipe->bind_sampler_states = pstip_bind_sampler_states; - pipe->set_sampler_textures = pstip_set_sampler_textures; - pipe->set_polygon_stipple = pstip_set_polygon_stipple; -} diff --git a/src/gallium/auxiliary/draw/draw_stipple.c b/src/gallium/auxiliary/draw/draw_stipple.c deleted file mode 100644 index 506f33512c..0000000000 --- a/src/gallium/auxiliary/draw/draw_stipple.c +++ /dev/null @@ -1,239 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -/* Implement line stipple by cutting lines up into smaller lines. - * There are hundreds of ways to implement line stipple, this is one - * choice that should work in all situations, requires no state - * manipulations, but with a penalty in terms of large amounts of - * generated geometry. - */ - - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -/** Subclass of draw_stage */ -struct stipple_stage { - struct draw_stage stage; - float counter; - uint pattern; - uint factor; -}; - - -static INLINE struct stipple_stage * -stipple_stage(struct draw_stage *stage) -{ - return (struct stipple_stage *) stage; -} - - -/** - * Compute interpolated vertex attributes for 'dst' at position 't' - * between 'v0' and 'v1'. - * XXX using linear interpolation for all attribs at this time. - */ -static void -screen_interp( struct draw_context *draw, - struct vertex_header *dst, - float t, - const struct vertex_header *v0, - const struct vertex_header *v1 ) -{ - uint attr; - for (attr = 0; attr < draw->num_vs_outputs; attr++) { - const float *val0 = v0->data[attr]; - const float *val1 = v1->data[attr]; - float *newv = dst->data[attr]; - uint i; - for (i = 0; i < 4; i++) { - newv[i] = val0[i] + t * (val1[i] - val0[i]); - } - } -} - - -static void -emit_segment(struct draw_stage *stage, struct prim_header *header, - float t0, float t1) -{ - struct vertex_header *v0new = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1new = dup_vert(stage, header->v[1], 1); - struct prim_header newprim = *header; - - if (t0 > 0.0) { - screen_interp( stage->draw, v0new, t0, header->v[0], header->v[1] ); - newprim.v[0] = v0new; - } - - if (t1 < 1.0) { - screen_interp( stage->draw, v1new, t1, header->v[0], header->v[1] ); - newprim.v[1] = v1new; - } - - stage->next->line( stage->next, &newprim ); -} - - -static INLINE unsigned -stipple_test(int counter, ushort pattern, int factor) -{ - int b = (counter / factor) & 0xf; - return (1 << b) & pattern; -} - - -static void -stipple_line(struct draw_stage *stage, struct prim_header *header) -{ - struct stipple_stage *stipple = stipple_stage(stage); - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - const float *pos0 = v0->data[0]; - const float *pos1 = v1->data[0]; - float start = 0; - int state = 0; - - float x0 = pos0[0]; - float x1 = pos1[0]; - float y0 = pos0[1]; - float y1 = pos1[1]; - - float dx = x0 > x1 ? x0 - x1 : x1 - x0; - float dy = y0 > y1 ? y0 - y1 : y1 - y0; - - float length = MAX2(dx, dy); - int i; - - /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table. - */ - for (i = 0; i < length; i++) { - int result = stipple_test( (int) stipple->counter+i, - (ushort) stipple->pattern, stipple->factor ); - if (result != state) { - /* changing from "off" to "on" or vice versa */ - if (state) { - if (start != i) { - /* finishing an "on" segment */ - emit_segment( stage, header, start / length, i / length ); - } - } - else { - /* starting an "on" segment */ - start = (float) i; - } - state = result; - } - } - - if (state && start < length) - emit_segment( stage, header, start / length, 1.0 ); - - stipple->counter += length; -} - - -static void -reset_stipple_counter(struct draw_stage *stage) -{ - struct stipple_stage *stipple = stipple_stage(stage); - stipple->counter = 0; - stage->next->reset_stipple_counter( stage->next ); -} - - -static void -stipple_first_line(struct draw_stage *stage, - struct prim_header *header) -{ - struct stipple_stage *stipple = stipple_stage(stage); - struct draw_context *draw = stage->draw; - - stipple->pattern = draw->rasterizer->line_stipple_pattern; - stipple->factor = draw->rasterizer->line_stipple_factor + 1; - - stage->line = stipple_line; - stage->line( stage, header ); -} - - -static void -stipple_flush(struct draw_stage *stage, unsigned flags) -{ - stage->line = stipple_first_line; - stage->next->flush( stage->next, flags ); -} - - -static void -passthrough_point(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->point( stage->next, header ); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} - - -static void -stipple_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create line stippler stage - */ -struct draw_stage *draw_stipple_stage( struct draw_context *draw ) -{ - struct stipple_stage *stipple = CALLOC_STRUCT(stipple_stage); - - draw_alloc_temp_verts( &stipple->stage, 2 ); - - stipple->stage.draw = draw; - stipple->stage.next = NULL; - stipple->stage.point = passthrough_point; - stipple->stage.line = stipple_first_line; - stipple->stage.tri = passthrough_tri; - stipple->stage.reset_stipple_counter = reset_stipple_counter; - stipple->stage.flush = stipple_flush; - stipple->stage.destroy = stipple_destroy; - - return &stipple->stage; -} diff --git a/src/gallium/auxiliary/draw/draw_twoside.c b/src/gallium/auxiliary/draw/draw_twoside.c deleted file mode 100644 index 01d905c153..0000000000 --- a/src/gallium/auxiliary/draw/draw_twoside.c +++ /dev/null @@ -1,203 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_vs.h" - - -struct twoside_stage { - struct draw_stage stage; - float sign; /**< +1 or -1 */ - uint attrib_front0, attrib_back0; - uint attrib_front1, attrib_back1; -}; - - -static INLINE struct twoside_stage *twoside_stage( struct draw_stage *stage ) -{ - return (struct twoside_stage *)stage; -} - - - - -/** - * Copy back color(s) to front color(s). - */ -static INLINE struct vertex_header * -copy_bfc( struct twoside_stage *twoside, - const struct vertex_header *v, - unsigned idx ) -{ - struct vertex_header *tmp = dup_vert( &twoside->stage, v, idx ); - - if (twoside->attrib_back0) { - COPY_4FV(tmp->data[twoside->attrib_front0], - tmp->data[twoside->attrib_back0]); - } - if (twoside->attrib_back1) { - COPY_4FV(tmp->data[twoside->attrib_front1], - tmp->data[twoside->attrib_back1]); - } - - return tmp; -} - - -/* Twoside tri: - */ -static void twoside_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - - if (header->det * twoside->sign < 0.0) { - /* this is a back-facing triangle */ - struct prim_header tmp; - - tmp.det = header->det; - tmp.edgeflags = header->edgeflags; - /* copy back attribs to front attribs */ - tmp.v[0] = copy_bfc(twoside, header->v[0], 0); - tmp.v[1] = copy_bfc(twoside, header->v[1], 1); - tmp.v[2] = copy_bfc(twoside, header->v[2], 2); - - stage->next->tri( stage->next, &tmp ); - } - else { - stage->next->tri( stage->next, header ); - } -} - - -static void twoside_line( struct draw_stage *stage, - struct prim_header *header ) -{ - /* pass-through */ - stage->next->line( stage->next, header ); -} - - -static void twoside_point( struct draw_stage *stage, - struct prim_header *header ) -{ - /* pass-through */ - stage->next->point( stage->next, header ); -} - - -static void twoside_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct twoside_stage *twoside = twoside_stage(stage); - const struct draw_vertex_shader *vs = stage->draw->vertex_shader; - uint i; - - twoside->attrib_front0 = 0; - twoside->attrib_front1 = 0; - twoside->attrib_back0 = 0; - twoside->attrib_back1 = 0; - - /* Find which vertex shader outputs are front/back colors */ - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR) { - if (vs->info.output_semantic_index[i] == 0) - twoside->attrib_front0 = i; - else - twoside->attrib_front1 = i; - } - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - if (vs->info.output_semantic_index[i] == 0) - twoside->attrib_back0 = i; - else - twoside->attrib_back1 = i; - } - } - - if (!twoside->attrib_back0) - twoside->attrib_front0 = 0; - - if (!twoside->attrib_back1) - twoside->attrib_front1 = 0; - - /* - * We'll multiply the primitive's determinant by this sign to determine - * if the triangle is back-facing (negative). - * sign = -1 for CCW, +1 for CW - */ - twoside->sign = (stage->draw->rasterizer->front_winding == PIPE_WINDING_CCW) ? -1.0f : 1.0f; - - stage->tri = twoside_tri; - stage->tri( stage, header ); -} - - -static void twoside_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->tri = twoside_first_tri; - stage->next->flush( stage->next, flags ); -} - - -static void twoside_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void twoside_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create twoside pipeline stage. - */ -struct draw_stage *draw_twoside_stage( struct draw_context *draw ) -{ - struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); - - draw_alloc_temp_verts( &twoside->stage, 3 ); - - twoside->stage.draw = draw; - twoside->stage.next = NULL; - twoside->stage.point = twoside_point; - twoside->stage.line = twoside_line; - twoside->stage.tri = twoside_first_tri; - twoside->stage.flush = twoside_flush; - twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; - twoside->stage.destroy = twoside_destroy; - - return &twoside->stage; -} diff --git a/src/gallium/auxiliary/draw/draw_unfilled.c b/src/gallium/auxiliary/draw/draw_unfilled.c deleted file mode 100644 index b07860cd9e..0000000000 --- a/src/gallium/auxiliary/draw/draw_unfilled.c +++ /dev/null @@ -1,206 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \brief Drawing stage for handling glPolygonMode(line/point). - * Convert triangles to points or lines as needed. - */ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "draw_private.h" - - -struct unfilled_stage { - struct draw_stage stage; - - /** [0] = front face, [1] = back face. - * legal values: PIPE_POLYGON_MODE_FILL, PIPE_POLYGON_MODE_LINE, - * and PIPE_POLYGON_MODE_POINT, - */ - unsigned mode[2]; -}; - - -static INLINE struct unfilled_stage *unfilled_stage( struct draw_stage *stage ) -{ - return (struct unfilled_stage *)stage; -} - - - -static void point( struct draw_stage *stage, - struct vertex_header *v0 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - stage->next->point( stage->next, &tmp ); -} - -static void line( struct draw_stage *stage, - struct vertex_header *v0, - struct vertex_header *v1 ) -{ - struct prim_header tmp; - tmp.v[0] = v0; - tmp.v[1] = v1; - stage->next->line( stage->next, &tmp ); -} - - -static void points( struct draw_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - - if (header->edgeflags & 0x1) point( stage, v0 ); - if (header->edgeflags & 0x2) point( stage, v1 ); - if (header->edgeflags & 0x4) point( stage, v2 ); -} - - -static void lines( struct draw_stage *stage, - struct prim_header *header ) -{ - struct vertex_header *v0 = header->v[0]; - struct vertex_header *v1 = header->v[1]; - struct vertex_header *v2 = header->v[2]; - -#if 0 - assert(((header->edgeflags & 0x1) >> 0) == header->v[0]->edgeflag); - assert(((header->edgeflags & 0x2) >> 1) == header->v[1]->edgeflag); - assert(((header->edgeflags & 0x4) >> 2) == header->v[2]->edgeflag); -#endif - - if (header->edgeflags & 0x4) line( stage, v2, v0 ); - if (header->edgeflags & 0x1) line( stage, v0, v1 ); - if (header->edgeflags & 0x2) line( stage, v1, v2 ); -} - - -/* Unfilled tri: - * - * Note edgeflags in the vertex struct is not sufficient as we will - * need to manipulate them when decomposing primitives??? - */ -static void unfilled_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - unsigned mode = unfilled->mode[header->det >= 0.0]; - - switch (mode) { - case PIPE_POLYGON_MODE_FILL: - stage->next->tri( stage->next, header ); - break; - case PIPE_POLYGON_MODE_LINE: - lines( stage, header ); - break; - case PIPE_POLYGON_MODE_POINT: - points( stage, header ); - break; - default: - assert(0); - } -} - - -static void unfilled_first_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct unfilled_stage *unfilled = unfilled_stage(stage); - - unfilled->mode[0] = stage->draw->rasterizer->fill_ccw; /* front */ - unfilled->mode[1] = stage->draw->rasterizer->fill_cw; /* back */ - - stage->tri = unfilled_tri; - stage->tri( stage, header ); -} - - -static void unfilled_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void unfilled_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void unfilled_flush( struct draw_stage *stage, - unsigned flags ) -{ - stage->next->flush( stage->next, flags ); - - stage->tri = unfilled_first_tri; -} - - -static void unfilled_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void unfilled_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -/** - * Create unfilled triangle stage. - */ -struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) -{ - struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); - - draw_alloc_temp_verts( &unfilled->stage, 0 ); - - unfilled->stage.draw = draw; - unfilled->stage.next = NULL; - unfilled->stage.tmp = NULL; - unfilled->stage.point = unfilled_point; - unfilled->stage.line = unfilled_line; - unfilled->stage.tri = unfilled_first_tri; - unfilled->stage.flush = unfilled_flush; - unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; - unfilled->stage.destroy = unfilled_destroy; - - return &unfilled->stage; -} diff --git a/src/gallium/auxiliary/draw/draw_validate.c b/src/gallium/auxiliary/draw/draw_validate.c deleted file mode 100644 index e163e078f0..0000000000 --- a/src/gallium/auxiliary/draw/draw_validate.c +++ /dev/null @@ -1,312 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "draw_private.h" - -static boolean points( unsigned prim ) -{ - return (prim == PIPE_PRIM_POINTS); -} - -static boolean lines( unsigned prim ) -{ - return (prim == PIPE_PRIM_LINES || - prim == PIPE_PRIM_LINE_STRIP || - prim == PIPE_PRIM_LINE_LOOP); -} - -static boolean triangles( unsigned prim ) -{ - return prim >= PIPE_PRIM_TRIANGLES; -} - -/** - * Check if we need any special pipeline stages, or whether - * prims/verts can go through untouched. Don't test for bypass - * clipping or vs modes, this function is just about the primitive - * pipeline stages. - */ -boolean -draw_need_pipeline(const struct draw_context *draw, - unsigned int prim ) -{ - /* Don't have to worry about triangles turning into lines/points - * and triggering the pipeline, because we have to trigger the - * pipeline *anyway* if unfilled mode is active. - */ - if (lines(prim)) - { - /* line stipple */ - if (draw->rasterizer->line_stipple_enable && draw->line_stipple) - return TRUE; - - /* wide lines */ - if (draw->rasterizer->line_width > draw->wide_line_threshold) - return TRUE; - - /* AA lines */ - if (draw->rasterizer->line_smooth && draw->pipeline.aaline) - return TRUE; - } - - if (points(prim)) - { - /* large points */ - if (draw->rasterizer->point_size > draw->wide_point_threshold) - return TRUE; - - /* AA points */ - if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) - return TRUE; - - /* point sprites */ - if (draw->rasterizer->point_sprite && draw->point_sprite) - return TRUE; - } - - - if (triangles(prim)) - { - /* polygon stipple */ - if (draw->rasterizer->poly_stipple_enable && draw->pipeline.pstipple) - return TRUE; - - /* unfilled polygons */ - if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) - return TRUE; - - /* polygon offset */ - if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) - return TRUE; - - /* two-side lighting */ - if (draw->rasterizer->light_twoside) - return TRUE; - } - - /* polygon cull - this is difficult - hardware can cull just fine - * most of the time (though sometimes CULL_NEITHER is unsupported. - * - * Generally this isn't a reason to require the pipeline, though. - * - if (draw->rasterizer->cull_mode) - return TRUE; - */ - - return FALSE; -} - - - -/** - * Rebuild the rendering pipeline. - */ -static struct draw_stage *validate_pipeline( struct draw_stage *stage ) -{ - struct draw_context *draw = stage->draw; - struct draw_stage *next = draw->pipeline.rasterize; - int need_det = 0; - int precalc_flat = 0; - boolean wide_lines, wide_points; - - /* Set the validate's next stage to the rasterize stage, so that it - * can be found later if needed for flushing. - */ - stage->next = next; - - /* drawing wide lines? */ - wide_lines = (draw->rasterizer->line_width > draw->wide_line_threshold - && !draw->rasterizer->line_smooth); - - /* drawing large points? */ - if (draw->rasterizer->point_sprite && draw->point_sprite) - wide_points = TRUE; - else if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) - wide_points = FALSE; - else if (draw->rasterizer->point_size > draw->wide_point_threshold) - wide_points = TRUE; - else - wide_points = FALSE; - - /* - * NOTE: we build up the pipeline in end-to-start order. - * - * TODO: make the current primitive part of the state and build - * shorter pipelines for lines & points. - */ - - if (draw->rasterizer->line_smooth && draw->pipeline.aaline) { - draw->pipeline.aaline->next = next; - next = draw->pipeline.aaline; - } - - if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) { - draw->pipeline.aapoint->next = next; - next = draw->pipeline.aapoint; - } - - if (wide_lines) { - draw->pipeline.wide_line->next = next; - next = draw->pipeline.wide_line; - precalc_flat = 1; - } - - if (wide_points || draw->rasterizer->point_sprite) { - draw->pipeline.wide_point->next = next; - next = draw->pipeline.wide_point; - } - - if (draw->rasterizer->line_stipple_enable && draw->line_stipple) { - draw->pipeline.stipple->next = next; - next = draw->pipeline.stipple; - precalc_flat = 1; /* only needed for lines really */ - } - - if (draw->rasterizer->poly_stipple_enable - && draw->pipeline.pstipple) { - draw->pipeline.pstipple->next = next; - next = draw->pipeline.pstipple; - } - - if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) { - draw->pipeline.unfilled->next = next; - next = draw->pipeline.unfilled; - precalc_flat = 1; /* only needed for triangles really */ - need_det = 1; - } - - if (draw->rasterizer->flatshade && precalc_flat) { - draw->pipeline.flatshade->next = next; - next = draw->pipeline.flatshade; - } - - if (draw->rasterizer->offset_cw || - draw->rasterizer->offset_ccw) { - draw->pipeline.offset->next = next; - next = draw->pipeline.offset; - need_det = 1; - } - - if (draw->rasterizer->light_twoside) { - draw->pipeline.twoside->next = next; - next = draw->pipeline.twoside; - need_det = 1; - } - - /* Always run the cull stage as we calculate determinant there - * also. - * - * This can actually be a win as culling out the triangles can lead - * to less work emitting vertices, smaller vertex buffers, etc. - * It's difficult to say whether this will be true in general. - */ - if (need_det || draw->rasterizer->cull_mode) { - draw->pipeline.cull->next = next; - next = draw->pipeline.cull; - } - - /* Clip stage - */ - if (!draw->rasterizer->bypass_clipping) - { - draw->pipeline.clip->next = next; - next = draw->pipeline.clip; - } - - - draw->pipeline.first = next; - return next; -} - -static void validate_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->tri( pipeline, header ); -} - -static void validate_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->line( pipeline, header ); -} - -static void validate_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->point( pipeline, header ); -} - -static void validate_reset_stipple_counter( struct draw_stage *stage ) -{ - struct draw_stage *pipeline = validate_pipeline( stage ); - pipeline->reset_stipple_counter( pipeline ); -} - -static void validate_flush( struct draw_stage *stage, - unsigned flags ) -{ - /* May need to pass a backend flush on to the rasterize stage. - */ - if (stage->next) - stage->next->flush( stage->next, flags ); -} - - -static void validate_destroy( struct draw_stage *stage ) -{ - FREE( stage ); -} - - -/** - * Create validate pipeline stage. - */ -struct draw_stage *draw_validate_stage( struct draw_context *draw ) -{ - struct draw_stage *stage = CALLOC_STRUCT(draw_stage); - - stage->draw = draw; - stage->next = NULL; - stage->point = validate_point; - stage->line = validate_line; - stage->tri = validate_tri; - stage->flush = validate_flush; - stage->reset_stipple_counter = validate_reset_stipple_counter; - stage->destroy = validate_destroy; - - return stage; -} diff --git a/src/gallium/auxiliary/draw/draw_vbuf.c b/src/gallium/auxiliary/draw/draw_vbuf.c deleted file mode 100644 index 30dceeb43d..0000000000 --- a/src/gallium/auxiliary/draw/draw_vbuf.c +++ /dev/null @@ -1,529 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * \file - * Vertex buffer drawing stage. - * - * \author José Fonseca - * \author Keith Whitwell - */ - - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" - -#include "draw_vbuf.h" -#include "draw_private.h" -#include "draw_vertex.h" -#include "translate/translate.h" - - -/** - * Vertex buffer emit stage. - */ -struct vbuf_stage { - struct draw_stage stage; /**< This must be first (base class) */ - - struct vbuf_render *render; - - const struct vertex_info *vinfo; - - /** Vertex size in bytes */ - unsigned vertex_size; - - struct translate *translate; - - /* FIXME: we have no guarantee that 'unsigned' is 32bit */ - - /** Vertices in hardware format */ - unsigned *vertices; - unsigned *vertex_ptr; - unsigned max_vertices; - unsigned nr_vertices; - - /** Indices */ - ushort *indices; - unsigned max_indices; - unsigned nr_indices; - - /* Cache point size somewhere it's address won't change: - */ - float point_size; -}; - - -/** - * Basically a cast wrapper. - */ -static INLINE struct vbuf_stage * -vbuf_stage( struct draw_stage *stage ) -{ - assert(stage); - return (struct vbuf_stage *)stage; -} - - -static void vbuf_flush_indices( struct vbuf_stage *vbuf ); -static void vbuf_flush_vertices( struct vbuf_stage *vbuf ); -static void vbuf_alloc_vertices( struct vbuf_stage *vbuf ); - - -static INLINE boolean -overflow( void *map, void *ptr, unsigned bytes, unsigned bufsz ) -{ - unsigned long used = (unsigned long) ((char *)ptr - (char *)map); - return (used + bytes) > bufsz; -} - - -static INLINE void -check_space( struct vbuf_stage *vbuf, unsigned nr ) -{ - if (vbuf->nr_vertices + nr > vbuf->max_vertices ) { - vbuf_flush_vertices(vbuf); - vbuf_alloc_vertices(vbuf); - } - - if (vbuf->nr_indices + nr > vbuf->max_indices ) - vbuf_flush_indices(vbuf); -} - - -static INLINE void -dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) -{ -// assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); - unsigned i, j; - - for (i = 0; i < vinfo->num_attribs; i++) { - j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - debug_printf("EMIT_OMIT:"); - break; - case EMIT_1F: - debug_printf("EMIT_1F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_1F_PSIZE: - debug_printf("EMIT_1F_PSIZE:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_2F: - debug_printf("EMIT_2F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_3F: - debug_printf("EMIT_3F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - data += sizeof(float); - break; - case EMIT_4F: - debug_printf("EMIT_4F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_4UB: - debug_printf("EMIT_4UB:\t"); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - break; - default: - assert(0); - } - debug_printf("\n"); - } - debug_printf("\n"); -} - - -/** - * Extract the needed fields from post-transformed vertex and emit - * a hardware(driver) vertex. - * Recall that the vertices are constructed by the 'draw' module and - * have a couple of slots at the beginning (1-dword header, 4-dword - * clip pos) that we ignore here. We only use the vertex->data[] fields. - */ -static INLINE ushort -emit_vertex( struct vbuf_stage *vbuf, - struct vertex_header *vertex ) -{ - if(vertex->vertex_id == UNDEFINED_VERTEX_ID) { - /* Hmm - vertices are emitted one at a time - better make sure - * set_buffer is efficient. Consider a special one-shot mode for - * translate. - */ - vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0); - vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr); - - if (0) dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); - - vbuf->vertex_ptr += vbuf->vertex_size/4; - vertex->vertex_id = vbuf->nr_vertices++; - } - - return vertex->vertex_id; -} - - -static void -vbuf_tri( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - check_space( vbuf, 3 ); - - for (i = 0; i < 3; i++) { - vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] ); - } -} - - -static void -vbuf_line( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - unsigned i; - - check_space( vbuf, 2 ); - - for (i = 0; i < 2; i++) { - vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[i] ); - } -} - - -static void -vbuf_point( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - check_space( vbuf, 1 ); - - vbuf->indices[vbuf->nr_indices++] = emit_vertex( vbuf, prim->v[0] ); -} - - - - -/** - * Set the prim type for subsequent vertices. - * This may result in a new vertex size. The existing vbuffer (if any) - * will be flushed if needed and a new one allocated. - */ -static void -vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) -{ - struct translate_key hw_key; - unsigned dst_offset; - unsigned i; - - vbuf->render->set_primitive(vbuf->render, prim); - - /* Must do this after set_primitive() above: - * - * XXX: need some state managment to track when this needs to be - * recalculated. The driver should tell us whether there was a - * state change. - */ - vbuf->vinfo = vbuf->render->get_vertex_info(vbuf->render); - - if (vbuf->vertex_size != vbuf->vinfo->size * sizeof(float)) { - vbuf_flush_vertices(vbuf); - vbuf->vertex_size = vbuf->vinfo->size * sizeof(float); - } - - /* Translate from pipeline vertices to hw vertices. - */ - dst_offset = 0; - memset(&hw_key, 0, sizeof(hw_key)); - - for (i = 0; i < vbuf->vinfo->num_attribs; i++) { - unsigned emit_sz = 0; - unsigned src_buffer = 0; - unsigned output_format; - unsigned src_offset = (vbuf->vinfo->src_index[i] * 4 * sizeof(float) ); - - switch (vbuf->vinfo->emit[i]) { - case EMIT_4F: - output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - emit_sz = 4 * sizeof(float); - break; - case EMIT_3F: - output_format = PIPE_FORMAT_R32G32B32_FLOAT; - emit_sz = 3 * sizeof(float); - break; - case EMIT_2F: - output_format = PIPE_FORMAT_R32G32_FLOAT; - emit_sz = 2 * sizeof(float); - break; - case EMIT_1F: - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - break; - case EMIT_1F_PSIZE: - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - src_buffer = 1; - src_offset = 0; - break; - case EMIT_4UB: - output_format = PIPE_FORMAT_B8G8R8A8_UNORM; - emit_sz = 4 * sizeof(ubyte); - default: - assert(0); - output_format = PIPE_FORMAT_NONE; - emit_sz = 0; - break; - } - - hw_key.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - hw_key.element[i].input_buffer = src_buffer; - hw_key.element[i].input_offset = src_offset; - hw_key.element[i].output_format = output_format; - hw_key.element[i].output_offset = dst_offset; - - dst_offset += emit_sz; - } - - hw_key.nr_elements = vbuf->vinfo->num_attribs; - hw_key.output_stride = vbuf->vinfo->size * 4; - - /* Don't bother with caching at this stage: - */ - if (!vbuf->translate || - memcmp(&vbuf->translate->key, &hw_key, sizeof(hw_key)) != 0) - { - if (vbuf->translate) - vbuf->translate->release(vbuf->translate); - - vbuf->translate = translate_create( &hw_key ); - - vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0); - } - - vbuf->point_size = vbuf->stage.draw->rasterizer->point_size; - - /* Allocate new buffer? - */ - if (!vbuf->vertices) - vbuf_alloc_vertices(vbuf); -} - - -static void -vbuf_first_tri( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - stage->tri = vbuf_tri; - vbuf_set_prim(vbuf, PIPE_PRIM_TRIANGLES); - stage->tri( stage, prim ); -} - - -static void -vbuf_first_line( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - stage->line = vbuf_line; - vbuf_set_prim(vbuf, PIPE_PRIM_LINES); - stage->line( stage, prim ); -} - - -static void -vbuf_first_point( struct draw_stage *stage, - struct prim_header *prim ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - stage->point = vbuf_point; - vbuf_set_prim(vbuf, PIPE_PRIM_POINTS); - stage->point( stage, prim ); -} - - -static void -vbuf_flush_indices( struct vbuf_stage *vbuf ) -{ - if(!vbuf->nr_indices) - return; - - assert((uint) (vbuf->vertex_ptr - vbuf->vertices) == - vbuf->nr_vertices * vbuf->vertex_size / sizeof(unsigned)); - - vbuf->render->draw(vbuf->render, vbuf->indices, vbuf->nr_indices); - - vbuf->nr_indices = 0; -} - - -/** - * Flush existing vertex buffer and allocate a new one. - * - * XXX: We separate flush-on-index-full and flush-on-vb-full, but may - * raise issues uploading vertices if the hardware wants to flush when - * we flush. - */ -static void -vbuf_flush_vertices( struct vbuf_stage *vbuf ) -{ - if(vbuf->vertices) { - vbuf_flush_indices(vbuf); - - /* Reset temporary vertices ids */ - if(vbuf->nr_vertices) - draw_reset_vertex_ids( vbuf->stage.draw ); - - /* Free the vertex buffer */ - vbuf->render->release_vertices(vbuf->render, - vbuf->vertices, - vbuf->vertex_size, - vbuf->nr_vertices); - vbuf->max_vertices = vbuf->nr_vertices = 0; - vbuf->vertex_ptr = vbuf->vertices = NULL; - - } -} - - -static void -vbuf_alloc_vertices( struct vbuf_stage *vbuf ) -{ - assert(!vbuf->nr_indices); - assert(!vbuf->vertices); - - /* Allocate a new vertex buffer */ - vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; - vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, - (ushort) vbuf->vertex_size, - (ushort) vbuf->max_vertices); - vbuf->vertex_ptr = vbuf->vertices; -} - - - -static void -vbuf_flush( struct draw_stage *stage, unsigned flags ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - vbuf_flush_indices( vbuf ); - - stage->point = vbuf_first_point; - stage->line = vbuf_first_line; - stage->tri = vbuf_first_tri; - - if (flags & DRAW_FLUSH_BACKEND) - vbuf_flush_vertices( vbuf ); -} - - -static void -vbuf_reset_stipple_counter( struct draw_stage *stage ) -{ - /* XXX: Need to do something here for hardware with linestipple. - */ - (void) stage; -} - - -static void vbuf_destroy( struct draw_stage *stage ) -{ - struct vbuf_stage *vbuf = vbuf_stage( stage ); - - if(vbuf->indices) - align_free( vbuf->indices ); - - if(vbuf->translate) - vbuf->translate->release( vbuf->translate ); - - if (vbuf->render) - vbuf->render->destroy( vbuf->render ); - - FREE( stage ); -} - - -/** - * Create a new primitive vbuf/render stage. - */ -struct draw_stage *draw_vbuf_stage( struct draw_context *draw, - struct vbuf_render *render ) -{ - struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); - - if(!vbuf) - goto fail; - - vbuf->stage.draw = draw; - vbuf->stage.point = vbuf_first_point; - vbuf->stage.line = vbuf_first_line; - vbuf->stage.tri = vbuf_first_tri; - vbuf->stage.flush = vbuf_flush; - vbuf->stage.reset_stipple_counter = vbuf_reset_stipple_counter; - vbuf->stage.destroy = vbuf_destroy; - - vbuf->render = render; - vbuf->max_indices = MAX2(render->max_indices, UNDEFINED_VERTEX_ID-1); - - vbuf->indices = (ushort *) align_malloc( vbuf->max_indices * - sizeof(vbuf->indices[0]), - 16 ); - if(!vbuf->indices) - goto fail; - - vbuf->vertices = NULL; - vbuf->vertex_ptr = vbuf->vertices; - - return &vbuf->stage; - - fail: - if (vbuf) - vbuf_destroy(&vbuf->stage); - - return NULL; -} diff --git a/src/gallium/auxiliary/draw/draw_wide_line.c b/src/gallium/auxiliary/draw/draw_wide_line.c deleted file mode 100644 index 9a168ce8bd..0000000000 --- a/src/gallium/auxiliary/draw/draw_wide_line.c +++ /dev/null @@ -1,190 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -struct wideline_stage { - struct draw_stage stage; - - float half_line_width; -}; - - - -static INLINE struct wideline_stage *wideline_stage( struct draw_stage *stage ) -{ - return (struct wideline_stage *)stage; -} - - -static void wideline_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void wideline_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->tri(stage->next, header); -} - - -/** - * Draw a wide line by drawing a quad (two triangles). - * XXX need to disable polygon stipple. - */ -static void wideline_line( struct draw_stage *stage, - struct prim_header *header ) -{ - /*const struct wideline_stage *wide = wideline_stage(stage);*/ - const float half_width = 0.5f * stage->draw->rasterizer->line_width; - - struct prim_header tri; - - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - const float dx = FABSF(pos0[0] - pos2[0]); - const float dy = FABSF(pos0[1] - pos2[1]); - - /* small tweak to meet GL specification */ - const float bias = 0.125f; - - /* - * Draw wide line as a quad (two tris) by "stretching" the line along - * X or Y. - * We need to tweak coords in several ways to be conformant here. - */ - - if (dx > dy) { - /* x-major line */ - pos0[1] = pos0[1] - half_width - bias; - pos1[1] = pos1[1] + half_width - bias; - pos2[1] = pos2[1] - half_width - bias; - pos3[1] = pos3[1] + half_width - bias; - if (pos0[0] < pos2[0]) { - /* left to right line */ - pos0[0] -= 0.5f; - pos1[0] -= 0.5f; - pos2[0] -= 0.5f; - pos3[0] -= 0.5f; - } - else { - /* right to left line */ - pos0[0] += 0.5f; - pos1[0] += 0.5f; - pos2[0] += 0.5f; - pos3[0] += 0.5f; - } - } - else { - /* y-major line */ - pos0[0] = pos0[0] - half_width + bias; - pos1[0] = pos1[0] + half_width + bias; - pos2[0] = pos2[0] - half_width + bias; - pos3[0] = pos3[0] + half_width + bias; - if (pos0[1] < pos2[1]) { - /* top to bottom line */ - pos0[1] -= 0.5f; - pos1[1] -= 0.5f; - pos2[1] -= 0.5f; - pos3[1] -= 0.5f; - } - else { - /* bottom to top line */ - pos0[1] += 0.5f; - pos1[1] += 0.5f; - pos2[1] += 0.5f; - pos3[1] += 0.5f; - } - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -static void wideline_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->next->flush( stage->next, flags ); -} - - -static void wideline_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void wideline_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -struct draw_stage *draw_wide_line_stage( struct draw_context *draw ) -{ - struct wideline_stage *wide = CALLOC_STRUCT(wideline_stage); - - draw_alloc_temp_verts( &wide->stage, 4 ); - - wide->stage.draw = draw; - wide->stage.next = NULL; - wide->stage.point = wideline_point; - wide->stage.line = wideline_line; - wide->stage.tri = wideline_tri; - wide->stage.flush = wideline_flush; - wide->stage.reset_stipple_counter = wideline_reset_stipple_counter; - wide->stage.destroy = wideline_destroy; - - return &wide->stage; -} diff --git a/src/gallium/auxiliary/draw/draw_wide_point.c b/src/gallium/auxiliary/draw/draw_wide_point.c deleted file mode 100644 index 3d0add0c1a..0000000000 --- a/src/gallium/auxiliary/draw/draw_wide_point.c +++ /dev/null @@ -1,281 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_vs.h" - - -struct widepoint_stage { - struct draw_stage stage; - - float half_point_size; - float point_size_min; - float point_size_max; - - float xbias; - float ybias; - - uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; - uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; - uint num_texcoords; - - int psize_slot; -}; - - - -static INLINE struct widepoint_stage * -widepoint_stage( struct draw_stage *stage ) -{ - return (struct widepoint_stage *)stage; -} - - -static void passthrough_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void widepoint_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line(stage->next, header); -} - -static void widepoint_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->tri(stage->next, header); -} - - -/** - * Set the vertex texcoords for sprite mode. - * Coords may be left untouched or set to a right-side-up or upside-down - * orientation. - */ -static void set_texcoords(const struct widepoint_stage *wide, - struct vertex_header *v, const float tc[4]) -{ - uint i; - for (i = 0; i < wide->num_texcoords; i++) { - if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { - uint j = wide->texcoord_slot[i]; - v->data[j][0] = tc[0]; - if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) - v->data[j][1] = 1.0f - tc[1]; - else - v->data[j][1] = tc[1]; - v->data[j][2] = tc[2]; - v->data[j][3] = tc[3]; - } - } -} - - -/* If there are lots of sprite points (and why wouldn't there be?) it - * would probably be more sensible to change hardware setup to - * optimize this rather than doing the whole thing in software like - * this. - */ -static void widepoint_point( struct draw_stage *stage, - struct prim_header *header ) -{ - const struct widepoint_stage *wide = widepoint_stage(stage); - const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; - float half_size; - float left_adj, right_adj, bot_adj, top_adj; - - struct prim_header tri; - - /* four dups of original vertex */ - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - /* point size is either per-vertex or fixed size */ - if (wide->psize_slot >= 0) { - half_size = header->v[0]->data[wide->psize_slot][0]; - - /* XXX: temporary -- do this in the vertex shader?? - */ - half_size = CLAMP(half_size, - wide->point_size_min, - wide->point_size_max); - - half_size *= 0.5f; - } - else { - half_size = wide->half_point_size; - } - - left_adj = -half_size + wide->xbias; - right_adj = half_size + wide->xbias; - bot_adj = half_size + wide->ybias; - top_adj = -half_size + wide->ybias; - - pos0[0] += left_adj; - pos0[1] += top_adj; - - pos1[0] += left_adj; - pos1[1] += bot_adj; - - pos2[0] += right_adj; - pos2[1] += top_adj; - - pos3[0] += right_adj; - pos3[1] += bot_adj; - - if (sprite) { - static const float tex00[4] = { 0, 0, 0, 1 }; - static const float tex01[4] = { 0, 1, 0, 1 }; - static const float tex11[4] = { 1, 1, 0, 1 }; - static const float tex10[4] = { 1, 0, 0, 1 }; - set_texcoords( wide, v0, tex00 ); - set_texcoords( wide, v1, tex01 ); - set_texcoords( wide, v2, tex10 ); - set_texcoords( wide, v3, tex11 ); - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -static void widepoint_first_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct widepoint_stage *wide = widepoint_stage(stage); - struct draw_context *draw = stage->draw; - - wide->half_point_size = 0.5f * draw->rasterizer->point_size; - wide->point_size_min = draw->rasterizer->point_size_min; - wide->point_size_max = draw->rasterizer->point_size_max; - wide->xbias = 0.0; - wide->ybias = 0.0; - - if (draw->rasterizer->gl_rasterization_rules) { - wide->xbias = 0.125; - } - - /* XXX we won't know the real size if it's computed by the vertex shader! */ - if ((draw->rasterizer->point_size > draw->wide_point_threshold) || - (draw->rasterizer->point_sprite && draw->point_sprite)) { - stage->point = widepoint_point; - } - else { - stage->point = passthrough_point; - } - - if (draw->rasterizer->point_sprite) { - /* find vertex shader texcoord outputs */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i, j = 0; - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { - wide->texcoord_slot[j] = i; - wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; - j++; - } - } - wide->num_texcoords = j; - } - - wide->psize_slot = -1; - if (draw->rasterizer->point_size_per_vertex) { - /* find PSIZ vertex output */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i; - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { - wide->psize_slot = i; - break; - } - } - } - - stage->point( stage, header ); -} - - -static void widepoint_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->point = widepoint_first_point; - stage->next->flush( stage->next, flags ); -} - - -static void widepoint_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void widepoint_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -struct draw_stage *draw_wide_point_stage( struct draw_context *draw ) -{ - struct widepoint_stage *wide = CALLOC_STRUCT(widepoint_stage); - - draw_alloc_temp_verts( &wide->stage, 4 ); - - wide->stage.draw = draw; - wide->stage.next = NULL; - wide->stage.point = widepoint_first_point; - wide->stage.line = widepoint_line; - wide->stage.tri = widepoint_tri; - wide->stage.flush = widepoint_flush; - wide->stage.reset_stipple_counter = widepoint_reset_stipple_counter; - wide->stage.destroy = widepoint_destroy; - - return &wide->stage; -} diff --git a/src/gallium/auxiliary/draw/draw_wide_prims.c b/src/gallium/auxiliary/draw/draw_wide_prims.c deleted file mode 100644 index d6bff110b4..0000000000 --- a/src/gallium/auxiliary/draw/draw_wide_prims.c +++ /dev/null @@ -1,366 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -struct wide_stage { - struct draw_stage stage; - - float half_line_width; - float half_point_size; - - uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; - uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; - uint num_texcoords; - - int psize_slot; -}; - - - -static INLINE struct wide_stage *wide_stage( struct draw_stage *stage ) -{ - return (struct wide_stage *)stage; -} - - -static void passthrough_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void passthrough_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line(stage->next, header); -} - -static void passthrough_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->tri(stage->next, header); -} - - -/** - * Draw a wide line by drawing a quad (two triangles). - * XXX need to disable polygon stipple. - */ -static void wide_line( struct draw_stage *stage, - struct prim_header *header ) -{ - const struct wide_stage *wide = wide_stage(stage); - const float half_width = wide->half_line_width; - - struct prim_header tri; - - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - const float dx = FABSF(pos0[0] - pos2[0]); - const float dy = FABSF(pos0[1] - pos2[1]); - - /* - * Draw wide line as a quad (two tris) by "stretching" the line along - * X or Y. - * We need to tweak coords in several ways to be conformant here. - */ - - if (dx > dy) { - /* x-major line */ - pos0[1] = pos0[1] - half_width - 0.25f; - pos1[1] = pos1[1] + half_width - 0.25f; - pos2[1] = pos2[1] - half_width - 0.25f; - pos3[1] = pos3[1] + half_width - 0.25f; - if (pos0[0] < pos2[0]) { - /* left to right line */ - pos0[0] -= 0.5f; - pos1[0] -= 0.5f; - pos2[0] -= 0.5f; - pos3[0] -= 0.5f; - } - else { - /* right to left line */ - pos0[0] += 0.5f; - pos1[0] += 0.5f; - pos2[0] += 0.5f; - pos3[0] += 0.5f; - } - } - else { - /* y-major line */ - pos0[0] = pos0[0] - half_width + 0.25f; - pos1[0] = pos1[0] + half_width + 0.25f; - pos2[0] = pos2[0] - half_width + 0.25f; - pos3[0] = pos3[0] + half_width + 0.25f; - if (pos0[1] < pos2[1]) { - /* top to bottom line */ - pos0[1] -= 0.5f; - pos1[1] -= 0.5f; - pos2[1] -= 0.5f; - pos3[1] -= 0.5f; - } - else { - /* bottom to top line */ - pos0[1] += 0.5f; - pos1[1] += 0.5f; - pos2[1] += 0.5f; - pos3[1] += 0.5f; - } - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -/** - * Set the vertex texcoords for sprite mode. - * Coords may be left untouched or set to a right-side-up or upside-down - * orientation. - */ -static void set_texcoords(const struct wide_stage *wide, - struct vertex_header *v, const float tc[4]) -{ - uint i; - for (i = 0; i < wide->num_texcoords; i++) { - if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { - uint j = wide->texcoord_slot[i]; - v->data[j][0] = tc[0]; - if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) - v->data[j][1] = 1.0f - tc[1]; - else - v->data[j][1] = tc[1]; - v->data[j][2] = tc[2]; - v->data[j][3] = tc[3]; - } - } -} - - -/* If there are lots of sprite points (and why wouldn't there be?) it - * would probably be more sensible to change hardware setup to - * optimize this rather than doing the whole thing in software like - * this. - */ -static void wide_point( struct draw_stage *stage, - struct prim_header *header ) -{ - const struct wide_stage *wide = wide_stage(stage); - const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; - float half_size; - float left_adj, right_adj; - - struct prim_header tri; - - /* four dups of original vertex */ - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - /* point size is either per-vertex or fixed size */ - if (wide->psize_slot >= 0) { - half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; - } - else { - half_size = wide->half_point_size; - } - - left_adj = -half_size; /* + 0.25f;*/ - right_adj = half_size; /* + 0.25f;*/ - - pos0[0] += left_adj; - pos0[1] -= half_size; - - pos1[0] += left_adj; - pos1[1] += half_size; - - pos2[0] += right_adj; - pos2[1] -= half_size; - - pos3[0] += right_adj; - pos3[1] += half_size; - - if (sprite) { - static const float tex00[4] = { 0, 0, 0, 1 }; - static const float tex01[4] = { 0, 1, 0, 1 }; - static const float tex11[4] = { 1, 1, 0, 1 }; - static const float tex10[4] = { 1, 0, 0, 1 }; - set_texcoords( wide, v0, tex00 ); - set_texcoords( wide, v1, tex01 ); - set_texcoords( wide, v2, tex10 ); - set_texcoords( wide, v3, tex11 ); - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -static void wide_first_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct wide_stage *wide = wide_stage(stage); - struct draw_context *draw = stage->draw; - - wide->half_point_size = 0.5f * draw->rasterizer->point_size; - - /* XXX we won't know the real size if it's computed by the vertex shader! */ - if (draw->rasterizer->point_size > draw->wide_point_threshold) { - stage->point = wide_point; - } - else { - stage->point = passthrough_point; - } - - if (draw->rasterizer->point_sprite) { - /* find vertex shader texcoord outputs */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i, j = 0; - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { - wide->texcoord_slot[j] = i; - wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; - j++; - } - } - wide->num_texcoords = j; - } - - wide->psize_slot = -1; - - if (draw->rasterizer->point_size_per_vertex) { - /* find PSIZ vertex output */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i; - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { - wide->psize_slot = i; - break; - } - } - } - - stage->point( stage, header ); -} - - - -static void wide_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct wide_stage *wide = wide_stage(stage); - struct draw_context *draw = stage->draw; - - wide->half_line_width = 0.5f * draw->rasterizer->line_width; - - if (draw->rasterizer->line_width != 1.0) { - wide->stage.line = wide_line; - } - else { - wide->stage.line = passthrough_line; - } - - stage->line( stage, header ); -} - - -static void wide_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->line = wide_first_line; - stage->point = wide_first_point; - stage->next->flush( stage->next, flags ); -} - - -static void wide_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void wide_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -struct draw_stage *draw_wide_stage( struct draw_context *draw ) -{ - struct wide_stage *wide = CALLOC_STRUCT(wide_stage); - - draw_alloc_temp_verts( &wide->stage, 4 ); - - wide->stage.draw = draw; - wide->stage.next = NULL; - wide->stage.point = wide_first_point; - wide->stage.line = wide_first_line; - wide->stage.tri = passthrough_tri; - wide->stage.flush = wide_flush; - wide->stage.reset_stipple_counter = wide_reset_stipple_counter; - wide->stage.destroy = wide_destroy; - - return &wide->stage; -} -- cgit v1.2.3 From 882e5d84dcb19c24b7b56cfe6049810023f3a17e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 13:12:07 +0100 Subject: draw: remove dead code --- src/gallium/auxiliary/draw/draw_vs.h | 27 --------------------------- 1 file changed, 27 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index b6f3bbdaea..f9772b83b8 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -81,33 +81,6 @@ draw_create_vs_llvm(struct draw_context *draw, const struct pipe_shader_state *templ); -/* Should be part of the generated shader: - */ -static INLINE unsigned -compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) -{ - unsigned mask = 0x0; - unsigned i; - - /* Do the hardwired planes first: - */ - if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; - if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; - if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; - if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; - if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; - if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; - - /* Followed by any remaining ones: - */ - for (i = 6; i < nr; i++) { - if (dot4(clip, plane[i]) < 0) - mask |= (1< Date: Sat, 19 Apr 2008 13:20:26 +0100 Subject: draw: remove named clipmask flags, tidy up pt middle ends --- src/gallium/auxiliary/draw/draw_context.h | 37 +++------------------------- src/gallium/auxiliary/draw/draw_private.h | 9 ++----- src/gallium/auxiliary/draw/draw_pt.c | 36 +++++++++++++-------------- src/gallium/auxiliary/draw/draw_pt.h | 6 +++++ src/gallium/auxiliary/draw/draw_pt_post_vs.c | 12 ++++----- 5 files changed, 35 insertions(+), 65 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index c7ac32b452..a0ac980c89 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -42,37 +42,11 @@ struct pipe_context; -struct vertex_buffer; -struct vertex_info; struct draw_context; struct draw_stage; struct draw_vertex_shader; -/** - * Clipmask flags - */ -/*@{*/ -#define CLIP_RIGHT_BIT 0x01 -#define CLIP_LEFT_BIT 0x02 -#define CLIP_TOP_BIT 0x04 -#define CLIP_BOTTOM_BIT 0x08 -#define CLIP_NEAR_BIT 0x10 -#define CLIP_FAR_BIT 0x20 -/*@}*/ - -/** - * Bitshift for each clip flag - */ -/*@{*/ -#define CLIP_RIGHT_SHIFT 0 -#define CLIP_LEFT_SHIFT 1 -#define CLIP_TOP_SHIFT 2 -#define CLIP_BOTTOM_SHIFT 3 -#define CLIP_NEAR_SHIFT 4 -#define CLIP_FAR_SHIFT 5 -/*@}*/ - struct draw_context *draw_create( void ); @@ -168,15 +142,10 @@ void draw_arrays(struct draw_context *draw, unsigned prim, void draw_flush(struct draw_context *draw); -/*********************************************************************** - * draw_debug.c - */ -boolean draw_validate_prim( unsigned prim, unsigned length ); -unsigned draw_trim_prim( unsigned mode, unsigned count ); - - - +/******************************************************************************* + * Driver backend interface + */ struct vbuf_render; void draw_set_render( struct draw_context *draw, struct vbuf_render *render ); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 476b1184d4..37ffdbf902 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -44,7 +44,6 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "rtasm/rtasm_x86sse.h" #include "tgsi/exec/tgsi_exec.h" #include "tgsi/util/tgsi_scan.h" @@ -171,7 +170,8 @@ struct draw_context */ struct { struct { - struct draw_pt_middle_end *opt[PT_MAX_MIDDLE]; + struct draw_pt_middle_end *fetch_emit; + struct draw_pt_middle_end *general; } middle; struct { @@ -289,11 +289,6 @@ extern boolean draw_need_pipeline(const struct draw_context *draw, */ boolean draw_pt_init( struct draw_context *draw ); void draw_pt_destroy( struct draw_context *draw ); -boolean draw_pt_arrays( struct draw_context *draw, - unsigned prim, - unsigned start, - unsigned count ); - void draw_pt_reset_vertex_ids( struct draw_context *draw ); #define DRAW_FLUSH_STATE_CHANGE 0x8 diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index ecaed84070..f153a3ee2c 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -44,7 +44,7 @@ * - pipeline -- the prim pipeline: clipping, wide lines, etc * - backend -- the vbuf_render provided by the driver. */ -boolean +static boolean draw_pt_arrays(struct draw_context *draw, unsigned prim, unsigned start, @@ -70,19 +70,16 @@ draw_pt_arrays(struct draw_context *draw, opt |= PT_SHADE; } + if (opt) + middle = draw->pt.middle.general; + else + middle = draw->pt.middle.fetch_emit; - middle = draw->pt.middle.opt[opt]; - if (middle == NULL) { - middle = draw->pt.middle.opt[PT_PIPELINE | PT_CLIPTEST | PT_SHADE]; - } - - assert(middle); /* May create a short-circuited version of this for small primitives: */ frontend = draw->pt.front.vcache; - frontend->prepare( frontend, prim, middle, opt ); frontend->run( frontend, @@ -102,11 +99,12 @@ boolean draw_pt_init( struct draw_context *draw ) if (!draw->pt.front.vcache) return FALSE; - draw->pt.middle.opt[0] = draw_pt_fetch_emit( draw ); - draw->pt.middle.opt[PT_SHADE | PT_CLIPTEST | PT_PIPELINE] = - draw_pt_fetch_pipeline_or_emit( draw ); + draw->pt.middle.fetch_emit = draw_pt_fetch_emit( draw ); + if (!draw->pt.middle.fetch_emit) + return FALSE; - if (!draw->pt.middle.opt[PT_SHADE | PT_CLIPTEST | PT_PIPELINE]) + draw->pt.middle.general = draw_pt_fetch_pipeline_or_emit( draw ); + if (!draw->pt.middle.general) return FALSE; return TRUE; @@ -115,13 +113,15 @@ boolean draw_pt_init( struct draw_context *draw ) void draw_pt_destroy( struct draw_context *draw ) { - int i; + if (draw->pt.middle.general) { + draw->pt.middle.general->destroy( draw->pt.middle.general ); + draw->pt.middle.general = NULL; + } - for (i = 0; i < PT_MAX_MIDDLE; i++) - if (draw->pt.middle.opt[i]) { - draw->pt.middle.opt[i]->destroy( draw->pt.middle.opt[i] ); - draw->pt.middle.opt[i] = NULL; - } + if (draw->pt.middle.fetch_emit) { + draw->pt.middle.fetch_emit->destroy( draw->pt.middle.fetch_emit ); + draw->pt.middle.fetch_emit = NULL; + } if (draw->pt.front.vcache) { draw->pt.front.vcache->destroy( draw->pt.front.vcache ); diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index eb23ee1c1e..1119e9c6b8 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -50,6 +50,12 @@ struct draw_context; #define DRAW_PT_FLAG_MASK (3<<30) +#define PT_SHADE 0x1 +#define PT_CLIPTEST 0x2 +#define PT_PIPELINE 0x4 +#define PT_MAX_MIDDLE 0x8 + + /* The "front end" - prepare sets of fetch, draw elements for the * middle end. * diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index 315b02f4ee..b3ecec6ece 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -52,12 +52,12 @@ compute_clipmask_gl(const float *clip, /*const*/ float plane[][4], unsigned nr) /* Do the hardwired planes first: */ - if (-clip[0] + clip[3] < 0) mask |= CLIP_RIGHT_BIT; - if ( clip[0] + clip[3] < 0) mask |= CLIP_LEFT_BIT; - if (-clip[1] + clip[3] < 0) mask |= CLIP_TOP_BIT; - if ( clip[1] + clip[3] < 0) mask |= CLIP_BOTTOM_BIT; - if (-clip[2] + clip[3] < 0) mask |= CLIP_FAR_BIT; - if ( clip[2] + clip[3] < 0) mask |= CLIP_NEAR_BIT; + if (-clip[0] + clip[3] < 0) mask |= (1<<0); + if ( clip[0] + clip[3] < 0) mask |= (1<<1); + if (-clip[1] + clip[3] < 0) mask |= (1<<2); + if ( clip[1] + clip[3] < 0) mask |= (1<<3); + if ( clip[2] + clip[3] < 0) mask |= (1<<4); /* match mesa clipplane numbering - for now */ + if (-clip[2] + clip[3] < 0) mask |= (1<<5); /* match mesa clipplane numbering - for now */ /* Followed by any remaining ones: */ -- cgit v1.2.3 From 0959f909ba585968f0408e78961e1c0ffc69a9f8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 13:30:19 +0100 Subject: draw: remove dead file pipe_wide_prims.c --- src/gallium/auxiliary/draw/draw_pipe_wide_prims.c | 366 ---------------------- 1 file changed, 366 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_pipe_wide_prims.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_prims.c b/src/gallium/auxiliary/draw/draw_pipe_wide_prims.c deleted file mode 100644 index d6bff110b4..0000000000 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_prims.c +++ /dev/null @@ -1,366 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "pipe/p_defines.h" -#include "pipe/p_shader_tokens.h" -#include "draw_private.h" - - -struct wide_stage { - struct draw_stage stage; - - float half_line_width; - float half_point_size; - - uint texcoord_slot[PIPE_MAX_SHADER_OUTPUTS]; - uint texcoord_mode[PIPE_MAX_SHADER_OUTPUTS]; - uint num_texcoords; - - int psize_slot; -}; - - - -static INLINE struct wide_stage *wide_stage( struct draw_stage *stage ) -{ - return (struct wide_stage *)stage; -} - - -static void passthrough_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void passthrough_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line(stage->next, header); -} - -static void passthrough_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->tri(stage->next, header); -} - - -/** - * Draw a wide line by drawing a quad (two triangles). - * XXX need to disable polygon stipple. - */ -static void wide_line( struct draw_stage *stage, - struct prim_header *header ) -{ - const struct wide_stage *wide = wide_stage(stage); - const float half_width = wide->half_line_width; - - struct prim_header tri; - - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - const float dx = FABSF(pos0[0] - pos2[0]); - const float dy = FABSF(pos0[1] - pos2[1]); - - /* - * Draw wide line as a quad (two tris) by "stretching" the line along - * X or Y. - * We need to tweak coords in several ways to be conformant here. - */ - - if (dx > dy) { - /* x-major line */ - pos0[1] = pos0[1] - half_width - 0.25f; - pos1[1] = pos1[1] + half_width - 0.25f; - pos2[1] = pos2[1] - half_width - 0.25f; - pos3[1] = pos3[1] + half_width - 0.25f; - if (pos0[0] < pos2[0]) { - /* left to right line */ - pos0[0] -= 0.5f; - pos1[0] -= 0.5f; - pos2[0] -= 0.5f; - pos3[0] -= 0.5f; - } - else { - /* right to left line */ - pos0[0] += 0.5f; - pos1[0] += 0.5f; - pos2[0] += 0.5f; - pos3[0] += 0.5f; - } - } - else { - /* y-major line */ - pos0[0] = pos0[0] - half_width + 0.25f; - pos1[0] = pos1[0] + half_width + 0.25f; - pos2[0] = pos2[0] - half_width + 0.25f; - pos3[0] = pos3[0] + half_width + 0.25f; - if (pos0[1] < pos2[1]) { - /* top to bottom line */ - pos0[1] -= 0.5f; - pos1[1] -= 0.5f; - pos2[1] -= 0.5f; - pos3[1] -= 0.5f; - } - else { - /* bottom to top line */ - pos0[1] += 0.5f; - pos1[1] += 0.5f; - pos2[1] += 0.5f; - pos3[1] += 0.5f; - } - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -/** - * Set the vertex texcoords for sprite mode. - * Coords may be left untouched or set to a right-side-up or upside-down - * orientation. - */ -static void set_texcoords(const struct wide_stage *wide, - struct vertex_header *v, const float tc[4]) -{ - uint i; - for (i = 0; i < wide->num_texcoords; i++) { - if (wide->texcoord_mode[i] != PIPE_SPRITE_COORD_NONE) { - uint j = wide->texcoord_slot[i]; - v->data[j][0] = tc[0]; - if (wide->texcoord_mode[i] == PIPE_SPRITE_COORD_LOWER_LEFT) - v->data[j][1] = 1.0f - tc[1]; - else - v->data[j][1] = tc[1]; - v->data[j][2] = tc[2]; - v->data[j][3] = tc[3]; - } - } -} - - -/* If there are lots of sprite points (and why wouldn't there be?) it - * would probably be more sensible to change hardware setup to - * optimize this rather than doing the whole thing in software like - * this. - */ -static void wide_point( struct draw_stage *stage, - struct prim_header *header ) -{ - const struct wide_stage *wide = wide_stage(stage); - const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; - float half_size; - float left_adj, right_adj; - - struct prim_header tri; - - /* four dups of original vertex */ - struct vertex_header *v0 = dup_vert(stage, header->v[0], 0); - struct vertex_header *v1 = dup_vert(stage, header->v[0], 1); - struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); - struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); - - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; - - /* point size is either per-vertex or fixed size */ - if (wide->psize_slot >= 0) { - half_size = 0.5f * header->v[0]->data[wide->psize_slot][0]; - } - else { - half_size = wide->half_point_size; - } - - left_adj = -half_size; /* + 0.25f;*/ - right_adj = half_size; /* + 0.25f;*/ - - pos0[0] += left_adj; - pos0[1] -= half_size; - - pos1[0] += left_adj; - pos1[1] += half_size; - - pos2[0] += right_adj; - pos2[1] -= half_size; - - pos3[0] += right_adj; - pos3[1] += half_size; - - if (sprite) { - static const float tex00[4] = { 0, 0, 0, 1 }; - static const float tex01[4] = { 0, 1, 0, 1 }; - static const float tex11[4] = { 1, 1, 0, 1 }; - static const float tex10[4] = { 1, 0, 0, 1 }; - set_texcoords( wide, v0, tex00 ); - set_texcoords( wide, v1, tex01 ); - set_texcoords( wide, v2, tex10 ); - set_texcoords( wide, v3, tex11 ); - } - - tri.det = header->det; /* only the sign matters */ - tri.v[0] = v0; - tri.v[1] = v2; - tri.v[2] = v3; - stage->next->tri( stage->next, &tri ); - - tri.v[0] = v0; - tri.v[1] = v3; - tri.v[2] = v1; - stage->next->tri( stage->next, &tri ); -} - - -static void wide_first_point( struct draw_stage *stage, - struct prim_header *header ) -{ - struct wide_stage *wide = wide_stage(stage); - struct draw_context *draw = stage->draw; - - wide->half_point_size = 0.5f * draw->rasterizer->point_size; - - /* XXX we won't know the real size if it's computed by the vertex shader! */ - if (draw->rasterizer->point_size > draw->wide_point_threshold) { - stage->point = wide_point; - } - else { - stage->point = passthrough_point; - } - - if (draw->rasterizer->point_sprite) { - /* find vertex shader texcoord outputs */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i, j = 0; - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { - wide->texcoord_slot[j] = i; - wide->texcoord_mode[j] = draw->rasterizer->sprite_coord_mode[j]; - j++; - } - } - wide->num_texcoords = j; - } - - wide->psize_slot = -1; - - if (draw->rasterizer->point_size_per_vertex) { - /* find PSIZ vertex output */ - const struct draw_vertex_shader *vs = draw->vertex_shader; - uint i; - for (i = 0; i < vs->info.num_outputs; i++) { - if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { - wide->psize_slot = i; - break; - } - } - } - - stage->point( stage, header ); -} - - - -static void wide_first_line( struct draw_stage *stage, - struct prim_header *header ) -{ - struct wide_stage *wide = wide_stage(stage); - struct draw_context *draw = stage->draw; - - wide->half_line_width = 0.5f * draw->rasterizer->line_width; - - if (draw->rasterizer->line_width != 1.0) { - wide->stage.line = wide_line; - } - else { - wide->stage.line = passthrough_line; - } - - stage->line( stage, header ); -} - - -static void wide_flush( struct draw_stage *stage, unsigned flags ) -{ - stage->line = wide_first_line; - stage->point = wide_first_point; - stage->next->flush( stage->next, flags ); -} - - -static void wide_reset_stipple_counter( struct draw_stage *stage ) -{ - stage->next->reset_stipple_counter( stage->next ); -} - - -static void wide_destroy( struct draw_stage *stage ) -{ - draw_free_temp_verts( stage ); - FREE( stage ); -} - - -struct draw_stage *draw_wide_stage( struct draw_context *draw ) -{ - struct wide_stage *wide = CALLOC_STRUCT(wide_stage); - - draw_alloc_temp_verts( &wide->stage, 4 ); - - wide->stage.draw = draw; - wide->stage.next = NULL; - wide->stage.point = wide_first_point; - wide->stage.line = wide_first_line; - wide->stage.tri = passthrough_tri; - wide->stage.flush = wide_flush; - wide->stage.reset_stipple_counter = wide_reset_stipple_counter; - wide->stage.destroy = wide_destroy; - - return &wide->stage; -} -- cgit v1.2.3 From 507fbe2d327efb8d608ce8e07436b97321560808 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 15:29:27 +0100 Subject: draw: move some pipeline-specific code & state to draw_pipe.[ch] --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/draw_context.c | 141 +++++-------------- src/gallium/auxiliary/draw/draw_pipe.c | 161 ++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_pipe.h | 110 +++++++++++++++ src/gallium/auxiliary/draw/draw_pipe_aaline.c | 1 + src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 1 + src/gallium/auxiliary/draw/draw_pipe_clip.c | 11 +- src/gallium/auxiliary/draw/draw_pipe_cull.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_flatshade.c | 1 + src/gallium/auxiliary/draw/draw_pipe_offset.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_stipple.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_twoside.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_unfilled.c | 1 + src/gallium/auxiliary/draw/draw_pipe_validate.c | 17 +-- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 1 + src/gallium/auxiliary/draw/draw_pipe_wide_line.c | 1 + src/gallium/auxiliary/draw/draw_pipe_wide_point.c | 5 +- src/gallium/auxiliary/draw/draw_private.h | 132 ++++-------------- src/gallium/auxiliary/draw/draw_pt.c | 18 --- src/gallium/auxiliary/draw/draw_pt_pipeline.c | 1 + src/gallium/auxiliary/draw/draw_pt_post_vs.c | 11 ++ src/gallium/auxiliary/draw/draw_vertex.c | 2 - src/gallium/drivers/i915simple/i915_context.c | 2 +- src/gallium/drivers/i915simple/i915_prim_emit.c | 5 +- src/gallium/drivers/softpipe/sp_prim_setup.c | 2 +- src/mesa/state_tracker/st_cb_feedback.c | 2 +- src/mesa/state_tracker/st_cb_rasterpos.c | 2 +- 28 files changed, 377 insertions(+), 262 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pipe.c create mode 100644 src/gallium/auxiliary/draw/draw_pipe.h (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 62f46c6db1..4fffd11b77 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -5,6 +5,7 @@ LIBNAME = draw C_SOURCES = \ draw_context.c \ + draw_pipe.c \ draw_pipe_aaline.c \ draw_pipe_aapoint.c \ draw_pipe_clip.c \ diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 6012bc155e..fa6791fa0b 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -35,6 +35,8 @@ #include "draw_context.h" #include "draw_vbuf.h" #include "draw_vs.h" +#include "draw_pt.h" +#include "draw_pipe.h" struct draw_context *draw_create( void ) @@ -49,32 +51,6 @@ struct draw_context *draw_create( void ) draw->use_sse = FALSE; #endif - /* create pipeline stages */ - draw->pipeline.wide_line = draw_wide_line_stage( draw ); - draw->pipeline.wide_point = draw_wide_point_stage( draw ); - draw->pipeline.stipple = draw_stipple_stage( draw ); - draw->pipeline.unfilled = draw_unfilled_stage( draw ); - draw->pipeline.twoside = draw_twoside_stage( draw ); - draw->pipeline.offset = draw_offset_stage( draw ); - draw->pipeline.clip = draw_clip_stage( draw ); - draw->pipeline.flatshade = draw_flatshade_stage( draw ); - draw->pipeline.cull = draw_cull_stage( draw ); - draw->pipeline.validate = draw_validate_stage( draw ); - draw->pipeline.first = draw->pipeline.validate; - - if (!draw->pipeline.wide_line || - !draw->pipeline.wide_point || - !draw->pipeline.stipple || - !draw->pipeline.unfilled || - !draw->pipeline.twoside || - !draw->pipeline.offset || - !draw->pipeline.clip || - !draw->pipeline.flatshade || - !draw->pipeline.cull || - !draw->pipeline.validate) - goto fail; - - ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); ASSIGN_4V( draw->plane[2], 0, -1, 0, 1 ); @@ -83,11 +59,6 @@ struct draw_context *draw_create( void ) ASSIGN_4V( draw->plane[5], 0, 0, -1, 1 ); /* mesa's a bit wonky */ draw->nr_planes = 6; - /* these defaults are oriented toward the needs of softpipe */ - draw->wide_point_threshold = 1000000.0; /* infinity */ - draw->wide_line_threshold = 1.0; - draw->line_stipple = TRUE; - draw->point_sprite = TRUE; draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ @@ -100,6 +71,8 @@ struct draw_context *draw_create( void ) draw->machine.Inputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); draw->machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); + if (!draw_pipeline_init( draw )) + goto fail; if (!draw_pt_init( draw )) goto fail; @@ -117,39 +90,13 @@ void draw_destroy( struct draw_context *draw ) if (!draw) return; - if (draw->pipeline.wide_line) - draw->pipeline.wide_line->destroy( draw->pipeline.wide_line ); - if (draw->pipeline.wide_point) - draw->pipeline.wide_point->destroy( draw->pipeline.wide_point ); - if (draw->pipeline.stipple) - draw->pipeline.stipple->destroy( draw->pipeline.stipple ); - if (draw->pipeline.unfilled) - draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); - if (draw->pipeline.twoside) - draw->pipeline.twoside->destroy( draw->pipeline.twoside ); - if (draw->pipeline.offset) - draw->pipeline.offset->destroy( draw->pipeline.offset ); - if (draw->pipeline.clip) - draw->pipeline.clip->destroy( draw->pipeline.clip ); - if (draw->pipeline.flatshade) - draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); - if (draw->pipeline.cull) - draw->pipeline.cull->destroy( draw->pipeline.cull ); - if (draw->pipeline.validate) - draw->pipeline.validate->destroy( draw->pipeline.validate ); - if (draw->pipeline.aaline) - draw->pipeline.aaline->destroy( draw->pipeline.aaline ); - if (draw->pipeline.aapoint) - draw->pipeline.aapoint->destroy( draw->pipeline.aapoint ); - if (draw->pipeline.pstipple) - draw->pipeline.pstipple->destroy( draw->pipeline.pstipple ); - if (draw->pipeline.rasterize) - draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); if (draw->machine.Inputs) align_free(draw->machine.Inputs); + if (draw->machine.Outputs) align_free(draw->machine.Outputs); + tgsi_exec_machine_free_data(&draw->machine); /* Not so fast -- we're just borrowing this at the moment. @@ -158,6 +105,7 @@ void draw_destroy( struct draw_context *draw ) draw->render->destroy( draw->render ); */ + draw_pipeline_destroy( draw ); draw_pt_destroy( draw ); FREE( draw ); @@ -284,7 +232,7 @@ void draw_wide_point_threshold(struct draw_context *draw, float threshold) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->wide_point_threshold = threshold; + draw->pipeline.wide_point_threshold = threshold; } @@ -296,7 +244,7 @@ void draw_wide_line_threshold(struct draw_context *draw, float threshold) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->wide_line_threshold = threshold; + draw->pipeline.wide_line_threshold = threshold; } @@ -307,7 +255,7 @@ void draw_enable_line_stipple(struct draw_context *draw, boolean enable) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->line_stipple = enable; + draw->pipeline.line_stipple = enable; } @@ -318,7 +266,7 @@ void draw_enable_point_sprites(struct draw_context *draw, boolean enable) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->point_sprite = enable; + draw->pipeline.point_sprite = enable; } @@ -373,36 +321,6 @@ draw_num_vs_outputs(struct draw_context *draw) } -/** - * Allocate space for temporary post-transform vertices, such as for clipping. - */ -void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) -{ - assert(!stage->tmp); - - stage->nr_tmps = nr; - - if (nr) { - ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); - unsigned i; - - stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); - } -} - - -void draw_free_temp_verts( struct draw_stage *stage ) -{ - if (stage->tmp) { - FREE( stage->tmp[0] ); - FREE( stage->tmp ); - stage->tmp = NULL; - } -} - boolean draw_use_sse(struct draw_context *draw) { @@ -410,23 +328,6 @@ boolean draw_use_sse(struct draw_context *draw) } -void draw_reset_vertex_ids(struct draw_context *draw) -{ - struct draw_stage *stage = draw->pipeline.first; - - while (stage) { - unsigned i; - - for (i = 0; i < stage->nr_tmps; i++) - stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; - - stage = stage->next; - } - - draw_pt_reset_vertex_ids(draw); -} - - void draw_set_render( struct draw_context *draw, struct vbuf_render *render ) { @@ -467,3 +368,23 @@ draw_set_mapped_element_buffer( struct draw_context *draw, draw->user.elts = elements; draw->user.eltSize = eltSize; } + + + +/* Revamp me please: + */ +void draw_do_flush( struct draw_context *draw, unsigned flags ) +{ + if (!draw->flushing) + { + draw->flushing = TRUE; + + if (flags >= DRAW_FLUSH_STATE_CHANGE) { + draw->pipeline.first->flush( draw->pipeline.first, flags ); + draw->pipeline.first = draw->pipeline.validate; + draw->reduced_prim = ~0; /* is reduced_prim needed any more? */ + } + + draw->flushing = FALSE; + } +} diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c new file mode 100644 index 0000000000..9d62cb2c65 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -0,0 +1,161 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw/draw_private.h" +#include "draw/draw_pipe.h" + + +boolean draw_pipeline_init( struct draw_context *draw ) +{ + /* create pipeline stages */ + draw->pipeline.wide_line = draw_wide_line_stage( draw ); + draw->pipeline.wide_point = draw_wide_point_stage( draw ); + draw->pipeline.stipple = draw_stipple_stage( draw ); + draw->pipeline.unfilled = draw_unfilled_stage( draw ); + draw->pipeline.twoside = draw_twoside_stage( draw ); + draw->pipeline.offset = draw_offset_stage( draw ); + draw->pipeline.clip = draw_clip_stage( draw ); + draw->pipeline.flatshade = draw_flatshade_stage( draw ); + draw->pipeline.cull = draw_cull_stage( draw ); + draw->pipeline.validate = draw_validate_stage( draw ); + draw->pipeline.first = draw->pipeline.validate; + + if (!draw->pipeline.wide_line || + !draw->pipeline.wide_point || + !draw->pipeline.stipple || + !draw->pipeline.unfilled || + !draw->pipeline.twoside || + !draw->pipeline.offset || + !draw->pipeline.clip || + !draw->pipeline.flatshade || + !draw->pipeline.cull || + !draw->pipeline.validate) + return FALSE; + + /* these defaults are oriented toward the needs of softpipe */ + draw->pipeline.wide_point_threshold = 1000000.0; /* infinity */ + draw->pipeline.wide_line_threshold = 1.0; + draw->pipeline.line_stipple = TRUE; + draw->pipeline.point_sprite = TRUE; + + return TRUE; +} + + +void draw_pipeline_destroy( struct draw_context *draw ) +{ + if (draw->pipeline.wide_line) + draw->pipeline.wide_line->destroy( draw->pipeline.wide_line ); + if (draw->pipeline.wide_point) + draw->pipeline.wide_point->destroy( draw->pipeline.wide_point ); + if (draw->pipeline.stipple) + draw->pipeline.stipple->destroy( draw->pipeline.stipple ); + if (draw->pipeline.unfilled) + draw->pipeline.unfilled->destroy( draw->pipeline.unfilled ); + if (draw->pipeline.twoside) + draw->pipeline.twoside->destroy( draw->pipeline.twoside ); + if (draw->pipeline.offset) + draw->pipeline.offset->destroy( draw->pipeline.offset ); + if (draw->pipeline.clip) + draw->pipeline.clip->destroy( draw->pipeline.clip ); + if (draw->pipeline.flatshade) + draw->pipeline.flatshade->destroy( draw->pipeline.flatshade ); + if (draw->pipeline.cull) + draw->pipeline.cull->destroy( draw->pipeline.cull ); + if (draw->pipeline.validate) + draw->pipeline.validate->destroy( draw->pipeline.validate ); + if (draw->pipeline.aaline) + draw->pipeline.aaline->destroy( draw->pipeline.aaline ); + if (draw->pipeline.aapoint) + draw->pipeline.aapoint->destroy( draw->pipeline.aapoint ); + if (draw->pipeline.pstipple) + draw->pipeline.pstipple->destroy( draw->pipeline.pstipple ); + if (draw->pipeline.rasterize) + draw->pipeline.rasterize->destroy( draw->pipeline.rasterize ); +} + + + + +/* This is only used for temporary verts. + */ +#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) + + +/** + * Allocate space for temporary post-transform vertices, such as for clipping. + */ +void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) +{ + assert(!stage->tmp); + + stage->nr_tmps = nr; + + if (nr) { + ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); + unsigned i; + + stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } +} + + +void draw_free_temp_verts( struct draw_stage *stage ) +{ + if (stage->tmp) { + FREE( stage->tmp[0] ); + FREE( stage->tmp ); + stage->tmp = NULL; + } +} + +void draw_reset_vertex_ids(struct draw_context *draw) +{ + struct draw_stage *stage = draw->pipeline.first; + + while (stage) { + unsigned i; + + for (i = 0; i < stage->nr_tmps; i++) + stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; + + stage = stage->next; + } + + draw_pt_reset_vertex_ids(draw); +} + + diff --git a/src/gallium/auxiliary/draw/draw_pipe.h b/src/gallium/auxiliary/draw/draw_pipe.h new file mode 100644 index 0000000000..5b97ca5c8c --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe.h @@ -0,0 +1,110 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#ifndef DRAW_PIPE_H +#define DRAW_PIPE_H + +#include "pipe/p_compiler.h" +#include "draw_private.h" /* for sizeof(vertex_header) */ + + + +/** + * Base class for all primitive drawing stages. + */ +struct draw_stage +{ + struct draw_context *draw; /**< parent context */ + + struct draw_stage *next; /**< next stage in pipeline */ + + struct vertex_header **tmp; /**< temp vert storage, such as for clipping */ + unsigned nr_tmps; + + void (*point)( struct draw_stage *, + struct prim_header * ); + + void (*line)( struct draw_stage *, + struct prim_header * ); + + void (*tri)( struct draw_stage *, + struct prim_header * ); + + void (*flush)( struct draw_stage *, + unsigned flags ); + + void (*reset_stipple_counter)( struct draw_stage * ); + + void (*destroy)( struct draw_stage * ); +}; + + +extern struct draw_stage *draw_unfilled_stage( struct draw_context *context ); +extern struct draw_stage *draw_twoside_stage( struct draw_context *context ); +extern struct draw_stage *draw_offset_stage( struct draw_context *context ); +extern struct draw_stage *draw_clip_stage( struct draw_context *context ); +extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); +extern struct draw_stage *draw_cull_stage( struct draw_context *context ); +extern struct draw_stage *draw_stipple_stage( struct draw_context *context ); +extern struct draw_stage *draw_wide_line_stage( struct draw_context *context ); +extern struct draw_stage *draw_wide_point_stage( struct draw_context *context ); +extern struct draw_stage *draw_validate_stage( struct draw_context *context ); + + +extern void draw_free_temp_verts( struct draw_stage *stage ); + +extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); + + + + +/** + * Get a writeable copy of a vertex. + * \param stage drawing stage info + * \param vert the vertex to copy (source) + * \param idx index into stage's tmp[] array to put the copy (dest) + * \return pointer to the copied vertex + */ +static INLINE struct vertex_header * +dup_vert( struct draw_stage *stage, + const struct vertex_header *vert, + unsigned idx ) +{ + struct vertex_header *tmp = stage->tmp[idx]; + const uint vsize = sizeof(struct vertex_header) + + stage->draw->num_vs_outputs * 4 * sizeof(float); + memcpy(tmp, vert, vsize); + tmp->vertex_id = UNDEFINED_VERTEX_ID; + return tmp; +} + +#endif diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index e8d2a45102..24bc87d4f8 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -43,6 +43,7 @@ #include "draw_context.h" #include "draw_private.h" +#include "draw_pipe.h" /** diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index e84d380e50..9f878f6c02 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -49,6 +49,7 @@ #include "draw_context.h" #include "draw_vs.h" +#include "draw_pipe.h" /* diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c index 0ac3a240e5..6780f275d9 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_clip.c +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -35,8 +35,8 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "draw_context.h" #include "draw_vs.h" +#include "draw_pipe.h" #ifndef IS_NEGATIVE @@ -204,7 +204,14 @@ static void emit_poly( struct draw_stage *stage, } } - +static INLINE float +dot4(const float *a, const float *b) +{ + return (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); +} /* Clip a triangle against the viewport and user clip planes. diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c index 8177b0ac86..c406f89d05 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_cull.c +++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c @@ -35,7 +35,7 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" -#include "draw_private.h" +#include "draw_pipe.h" struct cull_stage { diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c index 54baa1fbc9..bdb8b49dc4 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c @@ -31,6 +31,7 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "draw_vs.h" +#include "draw_pipe.h" /** subclass of draw_stage */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c index dbc676deae..dbdece45bb 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_offset.c +++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c @@ -33,7 +33,7 @@ */ #include "pipe/p_util.h" -#include "draw_private.h" +#include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 4dddb72906..4903ba2133 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -44,7 +44,7 @@ #include "tgsi/util/tgsi_dump.h" #include "draw_context.h" -#include "draw_private.h" +#include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c index 506f33512c..49429ee9e1 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c @@ -39,7 +39,7 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" -#include "draw_private.h" +#include "draw_pipe.h" /** Subclass of draw_stage */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c index 01d905c153..09a9d23d57 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_twoside.c +++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c @@ -32,7 +32,7 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" #include "draw_vs.h" - +#include "draw_pipe.h" struct twoside_stage { struct draw_stage stage; diff --git a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c index b07860cd9e..31e24f6a14 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c +++ b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c @@ -36,6 +36,7 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" #include "draw_private.h" +#include "draw_pipe.h" struct unfilled_stage { diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c index e163e078f0..ffddc2f62c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_validate.c +++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c @@ -31,6 +31,7 @@ #include "pipe/p_util.h" #include "pipe/p_defines.h" #include "draw_private.h" +#include "draw_pipe.h" static boolean points( unsigned prim ) { @@ -66,11 +67,11 @@ draw_need_pipeline(const struct draw_context *draw, if (lines(prim)) { /* line stipple */ - if (draw->rasterizer->line_stipple_enable && draw->line_stipple) + if (draw->rasterizer->line_stipple_enable && draw->pipeline.line_stipple) return TRUE; /* wide lines */ - if (draw->rasterizer->line_width > draw->wide_line_threshold) + if (draw->rasterizer->line_width > draw->pipeline.wide_line_threshold) return TRUE; /* AA lines */ @@ -81,7 +82,7 @@ draw_need_pipeline(const struct draw_context *draw, if (points(prim)) { /* large points */ - if (draw->rasterizer->point_size > draw->wide_point_threshold) + if (draw->rasterizer->point_size > draw->pipeline.wide_point_threshold) return TRUE; /* AA points */ @@ -89,7 +90,7 @@ draw_need_pipeline(const struct draw_context *draw, return TRUE; /* point sprites */ - if (draw->rasterizer->point_sprite && draw->point_sprite) + if (draw->rasterizer->point_sprite && draw->pipeline.point_sprite) return TRUE; } @@ -145,15 +146,15 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) stage->next = next; /* drawing wide lines? */ - wide_lines = (draw->rasterizer->line_width > draw->wide_line_threshold + wide_lines = (draw->rasterizer->line_width > draw->pipeline.wide_line_threshold && !draw->rasterizer->line_smooth); /* drawing large points? */ - if (draw->rasterizer->point_sprite && draw->point_sprite) + if (draw->rasterizer->point_sprite && draw->pipeline.point_sprite) wide_points = TRUE; else if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) wide_points = FALSE; - else if (draw->rasterizer->point_size > draw->wide_point_threshold) + else if (draw->rasterizer->point_size > draw->pipeline.wide_point_threshold) wide_points = TRUE; else wide_points = FALSE; @@ -186,7 +187,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) next = draw->pipeline.wide_point; } - if (draw->rasterizer->line_stipple_enable && draw->line_stipple) { + if (draw->rasterizer->line_stipple_enable && draw->pipeline.line_stipple) { draw->pipeline.stipple->next = next; next = draw->pipeline.stipple; precalc_flat = 1; /* only needed for lines really */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 30dceeb43d..c835727e32 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -40,6 +40,7 @@ #include "draw_vbuf.h" #include "draw_private.h" #include "draw_vertex.h" +#include "draw_pipe.h" #include "translate/translate.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c index 9a168ce8bd..329b5d0fb0 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c @@ -32,6 +32,7 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" #include "draw_private.h" +#include "draw_pipe.h" struct wideline_stage { diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index 3d0add0c1a..7a439178a0 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -32,6 +32,7 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" #include "draw_vs.h" +#include "draw_pipe.h" struct widepoint_stage { @@ -203,8 +204,8 @@ static void widepoint_first_point( struct draw_stage *stage, } /* XXX we won't know the real size if it's computed by the vertex shader! */ - if ((draw->rasterizer->point_size > draw->wide_point_threshold) || - (draw->rasterizer->point_sprite && draw->point_sprite)) { + if ((draw->rasterizer->point_size > draw->pipeline.wide_point_threshold) || + (draw->rasterizer->point_sprite && draw->pipeline.point_sprite)) { stage->point = widepoint_point; } else { diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 37ffdbf902..b2b2f82b8f 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -51,12 +51,11 @@ struct pipe_context; struct gallivm_prog; struct gallivm_cpu_engine; - -struct draw_pt_middle_end; -struct draw_pt_front_end; struct draw_vertex_shader; +struct draw_context; +struct draw_stage; +struct vbuf_render; -#define MAX_SHADER_VERTICES 128 /** * Basic vertex info. @@ -70,17 +69,14 @@ struct vertex_header { float clip[4]; - float data[][4]; /* Note variable size */ + /* This will probably become float (*data)[4] soon: + */ + float data[][4]; }; /* NOTE: It should match vertex_id size above */ #define UNDEFINED_VERTEX_ID 0xffff -/* XXX This is too large */ -#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) -#define MAX_VERTEX_ALLOCATION ((MAX_VERTEX_SIZE + 0x0f) & ~0x0f) - - /** * Basic info for a point/line/triangle primitive. @@ -95,41 +91,6 @@ struct prim_header { -struct draw_context; - -/** - * Base class for all primitive drawing stages. - */ -struct draw_stage -{ - struct draw_context *draw; /**< parent context */ - - struct draw_stage *next; /**< next stage in pipeline */ - - struct vertex_header **tmp; /**< temp vert storage, such as for clipping */ - unsigned nr_tmps; - - void (*point)( struct draw_stage *, - struct prim_header * ); - - void (*line)( struct draw_stage *, - struct prim_header * ); - - void (*tri)( struct draw_stage *, - struct prim_header * ); - - void (*flush)( struct draw_stage *, - unsigned flags ); - - void (*reset_stipple_counter)( struct draw_stage * ); - - void (*destroy)( struct draw_stage * ); -}; - - - -struct vbuf_render; - #define PT_SHADE 0x1 #define PT_CLIPTEST 0x2 @@ -161,6 +122,12 @@ struct draw_context struct draw_stage *wide_line; struct draw_stage *wide_point; struct draw_stage *rasterize; + + float wide_point_threshold; /**< convert pnts to tris if larger than this */ + float wide_line_threshold; /**< convert lines to tris if wider than this */ + boolean line_stipple; /**< do line stipple? */ + boolean point_sprite; /**< convert points to quads for sprites? */ + } pipeline; @@ -225,10 +192,6 @@ struct draw_context float plane[12][4]; unsigned nr_planes; - float wide_point_threshold; /**< convert pnts to tris if larger than this */ - float wide_line_threshold; /**< convert lines to tris if wider than this */ - boolean line_stipple; /**< do line stipple? */ - boolean point_sprite; /**< convert points to quads for sprites? */ boolean use_sse; /* If a prim stage introduces new vertex attributes, they'll be stored here @@ -252,44 +215,29 @@ struct draw_context -extern struct draw_stage *draw_unfilled_stage( struct draw_context *context ); -extern struct draw_stage *draw_twoside_stage( struct draw_context *context ); -extern struct draw_stage *draw_offset_stage( struct draw_context *context ); -extern struct draw_stage *draw_clip_stage( struct draw_context *context ); -extern struct draw_stage *draw_flatshade_stage( struct draw_context *context ); -extern struct draw_stage *draw_cull_stage( struct draw_context *context ); -extern struct draw_stage *draw_stipple_stage( struct draw_context *context ); -extern struct draw_stage *draw_wide_line_stage( struct draw_context *context ); -extern struct draw_stage *draw_wide_point_stage( struct draw_context *context ); -extern struct draw_stage *draw_validate_stage( struct draw_context *context ); - - -extern void draw_free_temp_verts( struct draw_stage *stage ); - -extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); extern void draw_reset_vertex_ids( struct draw_context *draw ); -extern int draw_vertex_cache_check_space( struct draw_context *draw, - unsigned nr_verts ); -extern void draw_vertex_cache_invalidate( struct draw_context *draw ); -extern void draw_vertex_cache_unreference( struct draw_context *draw ); -extern void draw_vertex_cache_reset_vertex_ids( struct draw_context *draw ); +/******************************************************************************* + * Vertex processing (was passthrough) code: + */ +boolean draw_pt_init( struct draw_context *draw ); +void draw_pt_destroy( struct draw_context *draw ); +void draw_pt_reset_vertex_ids( struct draw_context *draw ); -extern void draw_update_vertex_fetch( struct draw_context *draw ); +/******************************************************************************* + * Primitive processing (pipelnie) code: + */ -extern boolean draw_need_pipeline(const struct draw_context *draw, - unsigned prim ); +boolean draw_pipeline_init( struct draw_context *draw ); +void draw_pipeline_destroy( struct draw_context *draw ); +boolean draw_need_pipeline(const struct draw_context *draw, + unsigned prim ); -/* Passthrough mode (second attempt): - */ -boolean draw_pt_init( struct draw_context *draw ); -void draw_pt_destroy( struct draw_context *draw ); -void draw_pt_reset_vertex_ids( struct draw_context *draw ); #define DRAW_FLUSH_STATE_CHANGE 0x8 #define DRAW_FLUSH_BACKEND 0x10 @@ -301,36 +249,6 @@ boolean draw_get_edgeflag( struct draw_context *draw, unsigned idx ); -/** - * Get a writeable copy of a vertex. - * \param stage drawing stage info - * \param vert the vertex to copy (source) - * \param idx index into stage's tmp[] array to put the copy (dest) - * \return pointer to the copied vertex - */ -static INLINE struct vertex_header * -dup_vert( struct draw_stage *stage, - const struct vertex_header *vert, - unsigned idx ) -{ - struct vertex_header *tmp = stage->tmp[idx]; - const uint vsize = sizeof(struct vertex_header) - + stage->draw->num_vs_outputs * 4 * sizeof(float); - memcpy(tmp, vert, vsize); - tmp->vertex_id = UNDEFINED_VERTEX_ID; - return tmp; -} - -static INLINE float -dot4(const float *a, const float *b) -{ - float result = (a[0]*b[0] + - a[1]*b[1] + - a[2]*b[2] + - a[3]*b[3]); - - return result; -} #endif /* DRAW_PRIVATE_H */ diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index f153a3ee2c..965269251f 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -165,21 +165,3 @@ draw_arrays(struct draw_context *draw, unsigned prim, draw_pt_arrays(draw, prim, start, count); } - -/* Revamp me please: - */ -void draw_do_flush( struct draw_context *draw, unsigned flags ) -{ - if (!draw->flushing) - { - draw->flushing = TRUE; - - if (flags >= DRAW_FLUSH_STATE_CHANGE) { - draw->pipeline.first->flush( draw->pipeline.first, flags ); - draw->pipeline.first = draw->pipeline.validate; - draw->reduced_prim = ~0; - } - - draw->flushing = FALSE; - } -} diff --git a/src/gallium/auxiliary/draw/draw_pt_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_pipeline.c index 1a9a3adb03..922344e448 100644 --- a/src/gallium/auxiliary/draw/draw_pt_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_pipeline.c @@ -35,6 +35,7 @@ #include "draw/draw_private.h" #include "draw/draw_vertex.h" #include "draw/draw_pt.h" +#include "draw/draw_pipe.h" static void do_point( struct draw_context *draw, const char *v0 ) diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index b3ecec6ece..581f044dae 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -44,6 +44,17 @@ struct pt_post_vs { +static INLINE float +dot4(const float *a, const float *b) +{ + return (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); +} + + + static INLINE unsigned compute_clipmask_gl(const float *clip, /*const*/ float plane[][4], unsigned nr) { diff --git a/src/gallium/auxiliary/draw/draw_vertex.c b/src/gallium/auxiliary/draw/draw_vertex.c index 168036eee8..a42adaafb1 100644 --- a/src/gallium/auxiliary/draw/draw_vertex.c +++ b/src/gallium/auxiliary/draw/draw_vertex.c @@ -72,6 +72,4 @@ draw_compute_vertex_size(struct vertex_info *vinfo) assert(0); } } - - assert(vinfo->size * 4 <= MAX_VERTEX_SIZE); } diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 58a5854f0d..4bef21619c 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -142,7 +142,7 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, */ i915->draw = draw_create(); assert(i915->draw); - if (GETENV("I915_VBUF")) { + if (!GETENV("I915_NO_VBUF")) { draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915)); } else { diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c index b6fb0a6d88..9ffa460138 100644 --- a/src/gallium/drivers/i915simple/i915_prim_emit.c +++ b/src/gallium/drivers/i915simple/i915_prim_emit.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "draw/draw_private.h" +#include "draw/draw_pipe.h" #include "pipe/p_util.h" #include "i915_context.h" @@ -78,9 +78,6 @@ emit_hw_vertex( struct i915_context *i915, const uint j = vinfo->src_index[i]; const float *attrib = vertex->data[j]; switch (vinfo->emit[i]) { - case EMIT_OMIT: - /* no-op */ - break; case EMIT_1F: OUT_BATCH( fui(attrib[0]) ); count++; diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 0ddb06764a..feb35d492a 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -39,7 +39,7 @@ #include "sp_setup.h" #include "sp_state.h" #include "sp_prim_setup.h" -#include "draw/draw_private.h" +#include "draw/draw_pipe.h" #include "draw/draw_vertex.h" #include "pipe/p_util.h" diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 605bfee743..1b50792bd1 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -56,7 +56,7 @@ #include "cso_cache/cso_cache.h" #include "draw/draw_context.h" -#include "draw/draw_private.h" +#include "draw/draw_pipe.h" /** diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 2ed228778e..3cb7b68bea 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -46,7 +46,7 @@ #include "st_cb_rasterpos.h" #include "st_draw.h" #include "draw/draw_context.h" -#include "draw/draw_private.h" +#include "draw/draw_pipe.h" #include "shader/prog_instruction.h" #include "vbo/vbo.h" -- cgit v1.2.3 From c898eae27221bd23b11327553c215a94369eeb99 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 15:35:22 +0100 Subject: draw: always emit header in draw_pt_fetch.c --- src/gallium/auxiliary/draw/draw_pt.h | 1 - src/gallium/auxiliary/draw/draw_pt_fetch.c | 8 ++------ src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 1 - 3 files changed, 2 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 1119e9c6b8..eaf8e0374a 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -176,7 +176,6 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw ); struct pt_fetch; void draw_pt_fetch_prepare( struct pt_fetch *fetch, - boolean emit_header, unsigned vertex_size ); void draw_pt_fetch_run( struct pt_fetch *fetch, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 1aed251c85..037e3765da 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -54,7 +54,6 @@ struct pt_fetch { * */ void draw_pt_fetch_prepare( struct pt_fetch *fetch, - boolean emit_header, unsigned vertex_size ) { struct draw_context *draw = fetch->draw; @@ -66,16 +65,13 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, memset(&key, 0, sizeof(key)); - /* If PT_SHADE is not set, then we are creating post-shader - * vertices, meaning that we need to emit/leave space for a vertex - * header. + /* Always emit/leave space for a vertex header. * * It's worth considering whether the vertex headers should contain * a pointer to the 'data', rather than having it inline. * Something to look at after we've fully switched over to the pt * paths. */ - if (emit_header) { /* Need to set header->vertex_id = 0xffff somehow. */ @@ -121,7 +117,7 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, fetch->translate = translate_create( &key ); - if (emit_header) { + { static struct vertex_header vh = { 0, 0, 0, 0xffff }; fetch->translate->set_buffer(fetch->translate, draw->nr_vertex_buffers, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 0b9e8d15ba..560f1bcd92 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -71,7 +71,6 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, draw_pt_fetch_prepare( fpme->fetch, - (opt & (PT_CLIPTEST | PT_PIPELINE)) != 0, fpme->vertex_size ); /* XXX: it's not really gl rasterization rules we care about here, -- cgit v1.2.3 From bee1d31641674c67676de86fbb4b35ca5bf7f33f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 16:39:24 +0100 Subject: draw: move pt_pipeline code to draw_pipe.c This is now the drawing interface to the pipeline. No more calling into pipeline.first->tri(), etc. --- src/gallium/auxiliary/draw/Makefile | 1 - src/gallium/auxiliary/draw/SConscript | 1 - src/gallium/auxiliary/draw/draw_pipe.c | 117 ++++++++++++++- src/gallium/auxiliary/draw/draw_private.h | 24 +++- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 14 +- src/gallium/auxiliary/draw/draw_pt_pipeline.c | 158 --------------------- 6 files changed, 140 insertions(+), 175 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_pt_pipeline.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 4fffd11b77..5289f2660a 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -26,7 +26,6 @@ C_SOURCES = \ draw_pt_fetch.c \ draw_pt_fetch_emit.c \ draw_pt_fetch_shade_pipeline.c \ - draw_pt_pipeline.c \ draw_pt_post_vs.c \ draw_pt_vcache.c \ draw_vertex.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 5fa35d3005..6f3ca4fa49 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -25,7 +25,6 @@ draw = env.ConvenienceLibrary( 'draw_pt_fetch.c', 'draw_pt_fetch_emit.c', 'draw_pt_fetch_shade_pipeline.c', - 'draw_pt_pipeline.c', 'draw_pt_post_vs.c', 'draw_pt_vcache.c', 'draw_vs.c', diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 9d62cb2c65..b18b747d9c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -35,6 +35,7 @@ #include "draw/draw_pipe.h" + boolean draw_pipeline_init( struct draw_context *draw ) { /* create pipeline stages */ @@ -142,6 +143,60 @@ void draw_free_temp_verts( struct draw_stage *stage ) } } + + +static void do_point( struct draw_context *draw, + const char *v0 ) +{ + struct prim_header prim; + + prim.reset_line_stipple = 0; + prim.edgeflags = 1; + prim.pad = 0; + prim.v[0] = (struct vertex_header *)v0; + + draw->pipeline.first->point( draw->pipeline.first, &prim ); +} + + +static void do_line( struct draw_context *draw, + const char *v0, + const char *v1 ) +{ + struct prim_header prim; + + prim.reset_line_stipple = 1; /* fixme */ + prim.edgeflags = 1; + prim.pad = 0; + prim.v[0] = (struct vertex_header *)v0; + prim.v[1] = (struct vertex_header *)v1; + + draw->pipeline.first->line( draw->pipeline.first, &prim ); +} + + +static void do_triangle( struct draw_context *draw, + char *v0, + char *v1, + char *v2 ) +{ + struct prim_header prim; + + prim.v[0] = (struct vertex_header *)v0; + prim.v[1] = (struct vertex_header *)v1; + prim.v[2] = (struct vertex_header *)v2; + prim.reset_line_stipple = 1; + prim.edgeflags = ((prim.v[0]->edgeflag) | + (prim.v[1]->edgeflag << 1) | + (prim.v[2]->edgeflag << 2)); + prim.pad = 0; + + draw->pipeline.first->tri( draw->pipeline.first, &prim ); +} + + +/* Reset vertex ids. This is basically a type of flush. + */ void draw_reset_vertex_ids(struct draw_context *draw) { struct draw_stage *stage = draw->pipeline.first; @@ -155,7 +210,67 @@ void draw_reset_vertex_ids(struct draw_context *draw) stage = stage->next; } - draw_pt_reset_vertex_ids(draw); + if (draw->pipeline.verts) + { + unsigned i; + char *verts = draw->pipeline.verts; + unsigned stride = draw->pipeline.vertex_stride; + + for (i = 0; i < draw->pipeline.vertex_count; i++) { + ((struct vertex_header *)verts)->vertex_id = UNDEFINED_VERTEX_ID; + verts += stride; + } + } } +/* Code to run the pipeline on a fairly arbitary collection of vertices. + * + * Vertex headers must be pre-initialized with the + * UNDEFINED_VERTEX_ID, this code will cause that id to become + * overwritten, so it may have to be reset if there is the intention + * to reuse the vertices. + * + * This code provides a callback to reset the vertex id's which the + * draw_vbuf.c code uses when it has to perform a flush. + */ +void draw_pipeline_run( struct draw_context *draw, + unsigned prim, + struct vertex_header *vertices, + unsigned vertex_count, + unsigned stride, + const ushort *elts, + unsigned count ) +{ + char *verts = (char *)vertices; + unsigned i; + + draw->pipeline.verts = verts; + draw->pipeline.vertex_stride = stride; + draw->pipeline.vertex_count = vertex_count; + + switch (prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i++) + do_point( draw, + verts + stride * elts[i] ); + break; + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) + do_line( draw, + verts + stride * elts[i+0], + verts + stride * elts[i+1]); + break; + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) + do_triangle( draw, + verts + stride * elts[i+0], + verts + stride * elts[i+1], + verts + stride * elts[i+2]); + break; + } + + draw->pipeline.verts = NULL; + draw->pipeline.vertex_count = 0; +} + diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index b2b2f82b8f..4d123cf8d9 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -128,6 +128,11 @@ struct draw_context boolean line_stipple; /**< do line stipple? */ boolean point_sprite; /**< convert points to quads for sprites? */ + /* Temporary storage while the pipeline is being run: + */ + char *verts; + unsigned vertex_stride; + unsigned vertex_count; } pipeline; @@ -144,13 +149,6 @@ struct draw_context struct { struct draw_pt_front_end *vcache; } front; - - struct { - char *verts; - unsigned vertex_stride; - unsigned vertex_count; - } pipeline; - } pt; boolean flushing; @@ -235,10 +233,22 @@ void draw_pt_reset_vertex_ids( struct draw_context *draw ); boolean draw_pipeline_init( struct draw_context *draw ); void draw_pipeline_destroy( struct draw_context *draw ); +void draw_pipeline_run( struct draw_context *draw, + unsigned prim, + struct vertex_header *vertices, + unsigned vertex_count, + unsigned stride, + const ushort *elts, + unsigned count ); + boolean draw_need_pipeline(const struct draw_context *draw, unsigned prim ); +/******************************************************************************* + * Flushing + */ + #define DRAW_FLUSH_STATE_CHANGE 0x8 #define DRAW_FLUSH_BACKEND 0x10 diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 560f1bcd92..a47693ba20 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -148,13 +148,13 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, /* Do we need to run the pipeline? */ if (opt & PT_PIPELINE) { - draw_pt_run_pipeline( fpme->draw, - fpme->prim, - pipeline_verts, - fetch_count, - fpme->vertex_size, - draw_elts, - draw_count ); + draw_pipeline_run( fpme->draw, + fpme->prim, + pipeline_verts, + fetch_count, + fpme->vertex_size, + draw_elts, + draw_count ); } else { draw_pt_emit( fpme->emit, diff --git a/src/gallium/auxiliary/draw/draw_pt_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_pipeline.c deleted file mode 100644 index 922344e448..0000000000 --- a/src/gallium/auxiliary/draw/draw_pt_pipeline.c +++ /dev/null @@ -1,158 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - -#include "pipe/p_util.h" -#include "draw/draw_context.h" -#include "draw/draw_private.h" -#include "draw/draw_vertex.h" -#include "draw/draw_pt.h" -#include "draw/draw_pipe.h" - -static void do_point( struct draw_context *draw, - const char *v0 ) -{ - struct prim_header prim; - - prim.reset_line_stipple = 0; - prim.edgeflags = 1; - prim.pad = 0; - prim.v[0] = (struct vertex_header *)v0; - - draw->pipeline.first->point( draw->pipeline.first, &prim ); -} - - -static void do_line( struct draw_context *draw, - const char *v0, - const char *v1 ) -{ - struct prim_header prim; - - prim.reset_line_stipple = 1; /* fixme */ - prim.edgeflags = 1; - prim.pad = 0; - prim.v[0] = (struct vertex_header *)v0; - prim.v[1] = (struct vertex_header *)v1; - - draw->pipeline.first->line( draw->pipeline.first, &prim ); -} - - -static void do_triangle( struct draw_context *draw, - char *v0, - char *v1, - char *v2 ) -{ - struct prim_header prim; - - prim.v[0] = (struct vertex_header *)v0; - prim.v[1] = (struct vertex_header *)v1; - prim.v[2] = (struct vertex_header *)v2; - prim.reset_line_stipple = 1; - prim.edgeflags = ((prim.v[0]->edgeflag) | - (prim.v[1]->edgeflag << 1) | - (prim.v[2]->edgeflag << 2)); - prim.pad = 0; - - if (0) debug_printf("tri ef: %d %d %d\n", - prim.v[0]->edgeflag, - prim.v[1]->edgeflag, - prim.v[2]->edgeflag); - - draw->pipeline.first->tri( draw->pipeline.first, &prim ); -} - - - -void draw_pt_reset_vertex_ids( struct draw_context *draw ) -{ - unsigned i; - char *verts = draw->pt.pipeline.verts; - unsigned stride = draw->pt.pipeline.vertex_stride; - - for (i = 0; i < draw->pt.pipeline.vertex_count; i++) { - ((struct vertex_header *)verts)->vertex_id = UNDEFINED_VERTEX_ID; - verts += stride; - } -} - - -/* Code to run the pipeline on a fairly arbitary collection of vertices. - * - * Vertex headers must be pre-initialized with the - * UNDEFINED_VERTEX_ID, this code will cause that id to become - * overwritten, so it may have to be reset if there is the intention - * to reuse the vertices. - * - * This code provides a callback to reset the vertex id's which the - * draw_vbuf.c code uses when it has to perform a flush. - */ -void draw_pt_run_pipeline( struct draw_context *draw, - unsigned prim, - struct vertex_header *pipeline_verts, - unsigned vertex_count, - unsigned stride, - const ushort *elts, - unsigned count ) -{ - char *verts = (char *)pipeline_verts; - unsigned i; - - draw->pt.pipeline.verts = verts; - draw->pt.pipeline.vertex_stride = stride; - draw->pt.pipeline.vertex_count = vertex_count; - - switch (prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i++) - do_point( draw, - verts + stride * elts[i] ); - break; - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) - do_line( draw, - verts + stride * elts[i+0], - verts + stride * elts[i+1]); - break; - case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) - do_triangle( draw, - verts + stride * elts[i+0], - verts + stride * elts[i+1], - verts + stride * elts[i+2]); - break; - } - - draw->pt.pipeline.verts = NULL; - draw->pt.pipeline.vertex_count = 0; -} - -- cgit v1.2.3 From e7bac4276634ea1ee81ac71f6f6869f87e689872 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 16:43:15 +0100 Subject: draw: put pipeline flushing behind a new interface --- src/gallium/auxiliary/draw/draw_context.c | 8 +++----- src/gallium/auxiliary/draw/draw_pipe.c | 8 ++++++++ src/gallium/auxiliary/draw/draw_private.h | 5 ++++- 3 files changed, 15 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index fa6791fa0b..3e69867d11 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -379,11 +379,9 @@ void draw_do_flush( struct draw_context *draw, unsigned flags ) { draw->flushing = TRUE; - if (flags >= DRAW_FLUSH_STATE_CHANGE) { - draw->pipeline.first->flush( draw->pipeline.first, flags ); - draw->pipeline.first = draw->pipeline.validate; - draw->reduced_prim = ~0; /* is reduced_prim needed any more? */ - } + draw_pipeline_flush( draw, flags ); + + draw->reduced_prim = ~0; /* is reduced_prim needed any more? */ draw->flushing = FALSE; } diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index b18b747d9c..3fae908514 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -274,3 +274,11 @@ void draw_pipeline_run( struct draw_context *draw, draw->pipeline.vertex_count = 0; } + + +void draw_pipeline_flush( struct draw_context *draw, + unsigned flags ) +{ + draw->pipeline.first->flush( draw->pipeline.first, flags ); + draw->pipeline.first = draw->pipeline.validate; +} diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 4d123cf8d9..59635d643b 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -227,7 +227,7 @@ void draw_pt_reset_vertex_ids( struct draw_context *draw ); /******************************************************************************* - * Primitive processing (pipelnie) code: + * Primitive processing (pipeline) code: */ boolean draw_pipeline_init( struct draw_context *draw ); @@ -241,6 +241,9 @@ void draw_pipeline_run( struct draw_context *draw, const ushort *elts, unsigned count ); +void draw_pipeline_flush( struct draw_context *draw, + unsigned flags ); + boolean draw_need_pipeline(const struct draw_context *draw, unsigned prim ); -- cgit v1.2.3 From 2dae208fb19e79c7446a29ee5dee53e50283b57c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 17:16:23 +0100 Subject: draw: make room for extra_vs_outputs --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index a47693ba20..e1df594035 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -56,8 +56,12 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; struct draw_vertex_shader *vs = draw->vertex_shader; + + /* Add one to num_outputs because the pipeline occasionally tags on + * an additional texcoord, eg for AA lines. + */ unsigned nr = MAX2( vs->info.num_inputs, - vs->info.num_outputs ); + vs->info.num_outputs + 1 ); fpme->prim = prim; fpme->opt = opt; -- cgit v1.2.3 From dcf6f776ce32b89b7ff784bb38030bd29698e005 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 17:16:41 +0100 Subject: draw: make draw_reset_vertex_ids private to the draw_pipe_* code --- src/gallium/auxiliary/draw/draw_pipe.c | 2 ++ src/gallium/auxiliary/draw/draw_pipe.h | 2 +- src/gallium/auxiliary/draw/draw_private.h | 1 - 3 files changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 3fae908514..0c0840af0c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -196,6 +196,8 @@ static void do_triangle( struct draw_context *draw, /* Reset vertex ids. This is basically a type of flush. + * + * Called only from draw_pipe_vbuf.c */ void draw_reset_vertex_ids(struct draw_context *draw) { diff --git a/src/gallium/auxiliary/draw/draw_pipe.h b/src/gallium/auxiliary/draw/draw_pipe.h index 5b97ca5c8c..f2749b34bb 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.h +++ b/src/gallium/auxiliary/draw/draw_pipe.h @@ -81,9 +81,9 @@ extern struct draw_stage *draw_validate_stage( struct draw_context *context ); extern void draw_free_temp_verts( struct draw_stage *stage ); - extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); +extern void draw_reset_vertex_ids( struct draw_context *draw ); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 59635d643b..6cf3e54e28 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -214,7 +214,6 @@ struct draw_context -extern void draw_reset_vertex_ids( struct draw_context *draw ); -- cgit v1.2.3 From 7d72607e142c0412b88183b849fd701e698b8f79 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 17:27:52 +0100 Subject: draw: move incoming vertex state into draw->pt This state is effectively private to the vertex processing part of the draw module. --- src/gallium/auxiliary/draw/draw_context.c | 28 +++++-------- src/gallium/auxiliary/draw/draw_private.h | 47 +++++++++++----------- src/gallium/auxiliary/draw/draw_pt.c | 8 ++++ src/gallium/auxiliary/draw/draw_pt.h | 9 +---- src/gallium/auxiliary/draw/draw_pt_elts.c | 6 +-- src/gallium/auxiliary/draw/draw_pt_fetch.c | 20 ++++----- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 6 +-- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 2 +- src/gallium/auxiliary/draw/draw_pt_vcache.c | 2 +- src/mesa/state_tracker/st_draw.c | 6 +-- 10 files changed, 63 insertions(+), 71 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 3e69867d11..4988d67faa 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -62,8 +62,6 @@ struct draw_context *draw_create( void ) draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ - draw_set_mapped_element_buffer( draw, 0, NULL ); - tgsi_exec_machine_init(&draw->machine); /* FIXME: give this machine thing a proper constructor: @@ -188,8 +186,8 @@ draw_set_vertex_buffers(struct draw_context *draw, { assert(count <= PIPE_MAX_ATTRIBS); - memcpy(draw->vertex_buffer, buffers, count * sizeof(buffers[0])); - draw->nr_vertex_buffers = count; + memcpy(draw->pt.vertex_buffer, buffers, count * sizeof(buffers[0])); + draw->pt.nr_vertex_buffers = count; } @@ -200,8 +198,8 @@ draw_set_vertex_elements(struct draw_context *draw, { assert(count <= PIPE_MAX_ATTRIBS); - memcpy(draw->vertex_element, elements, count * sizeof(elements[0])); - draw->nr_vertex_elements = count; + memcpy(draw->pt.vertex_element, elements, count * sizeof(elements[0])); + draw->pt.nr_vertex_elements = count; } @@ -212,7 +210,7 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw, unsigned attr, const void *buffer) { - draw->user.vbuffer[attr] = buffer; + draw->pt.user.vbuffer[attr] = buffer; } @@ -220,7 +218,7 @@ void draw_set_mapped_constant_buffer(struct draw_context *draw, const void *buffer) { - draw->user.constants = buffer; + draw->pt.user.constants = buffer; } @@ -337,18 +335,10 @@ void draw_set_render( struct draw_context *draw, void draw_set_edgeflags( struct draw_context *draw, const unsigned *edgeflag ) { - draw->user.edgeflag = edgeflag; + draw->pt.user.edgeflag = edgeflag; } -boolean draw_get_edgeflag( struct draw_context *draw, - unsigned idx ) -{ - if (draw->user.edgeflag) - return (draw->user.edgeflag[idx/32] & (1 << (idx%32))) != 0; - else - return 1; -} /** @@ -365,8 +355,8 @@ void draw_set_mapped_element_buffer( struct draw_context *draw, unsigned eltSize, void *elements ) { - draw->user.elts = elements; - draw->user.eltSize = eltSize; + draw->pt.user.elts = elements; + draw->pt.user.eltSize = eltSize; } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 6cf3e54e28..27f61c2f40 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -149,6 +149,29 @@ struct draw_context struct { struct draw_pt_front_end *vcache; } front; + + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + unsigned nr_vertex_buffers; + + struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; + unsigned nr_vertex_elements; + + /* user-space vertex data, buffers */ + struct { + const unsigned *edgeflag; + + /** vertex element/index buffer (ex: glDrawElements) */ + const void *elts; + /** bytes per index (0, 1, 2 or 4) */ + unsigned eltSize; + + /** vertex arrays */ + const void *vbuffer[PIPE_MAX_ATTRIBS]; + + /** constant buffer (for vertex shader) */ + const void *constants; + } user; + } pt; boolean flushing; @@ -157,33 +180,12 @@ struct draw_context const struct pipe_rasterizer_state *rasterizer; struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; - unsigned nr_vertex_buffers; - - struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; - unsigned nr_vertex_elements; - struct draw_vertex_shader *vertex_shader; boolean identity_viewport; uint num_vs_outputs; /**< convenience, from vertex_shader */ - /* user-space vertex data, buffers */ - struct { - const unsigned *edgeflag; - - /** vertex element/index buffer (ex: glDrawElements) */ - const void *elts; - /** bytes per index (0, 1, 2 or 4) */ - unsigned eltSize; - - /** vertex arrays */ - const void *vbuffer[PIPE_MAX_ATTRIBS]; - - /** constant buffer (for vertex shader) */ - const void *constants; - } user; /* Clip derived state: */ @@ -257,9 +259,6 @@ boolean draw_need_pipeline(const struct draw_context *draw, void draw_do_flush( struct draw_context *draw, unsigned flags ); -boolean draw_get_edgeflag( struct draw_context *draw, - unsigned idx ); - diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 965269251f..9f8e8d3d62 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -165,3 +165,11 @@ draw_arrays(struct draw_context *draw, unsigned prim, draw_pt_arrays(draw, prim, start, count); } +boolean draw_pt_get_edgeflag( struct draw_context *draw, + unsigned idx ) +{ + if (draw->pt.user.edgeflag) + return (draw->pt.user.edgeflag[idx/32] & (1 << (idx%32))) != 0; + else + return 1; +} diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index eaf8e0374a..fd0d158fcf 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -141,13 +141,8 @@ struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *d /* More helpers: */ -void draw_pt_run_pipeline( struct draw_context *draw, - unsigned prim, - struct vertex_header *verts, - unsigned vertex_count, - unsigned vertex_stride, - const ushort *elts, - unsigned count ); +boolean draw_pt_get_edgeflag( struct draw_context *draw, + unsigned idx ); /******************************************************************************* diff --git a/src/gallium/auxiliary/draw/draw_pt_elts.c b/src/gallium/auxiliary/draw/draw_pt_elts.c index d49770e7b2..2094c081ed 100644 --- a/src/gallium/auxiliary/draw/draw_pt_elts.c +++ b/src/gallium/auxiliary/draw/draw_pt_elts.c @@ -59,7 +59,7 @@ static unsigned elt_vert( const void *elts, unsigned idx ) pt_elt_func draw_pt_elt_func( struct draw_context *draw ) { - switch (draw->user.eltSize) { + switch (draw->pt.user.eltSize) { case 0: return elt_vert; case 1: return elt_ubyte; case 2: return elt_ushort; @@ -71,9 +71,9 @@ pt_elt_func draw_pt_elt_func( struct draw_context *draw ) const void *draw_pt_elt_ptr( struct draw_context *draw, unsigned start ) { - const char *elts = draw->user.elts; + const char *elts = draw->pt.user.elts; - switch (draw->user.eltSize) { + switch (draw->pt.user.eltSize) { case 0: return (const void *)(((const ubyte *)NULL) + start); case 1: diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 037e3765da..c588710b75 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -76,7 +76,7 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, /* Need to set header->vertex_id = 0xffff somehow. */ key.element[nr].input_format = PIPE_FORMAT_R32_FLOAT; - key.element[nr].input_buffer = draw->nr_vertex_buffers; + key.element[nr].input_buffer = draw->pt.nr_vertex_buffers; key.element[nr].input_offset = 0; key.element[nr].output_format = PIPE_FORMAT_R32_FLOAT; key.element[nr].output_offset = dst_offset; @@ -90,10 +90,10 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, } - for (i = 0; i < draw->nr_vertex_elements; i++) { - key.element[nr].input_format = draw->vertex_element[i].src_format; - key.element[nr].input_buffer = draw->vertex_element[i].vertex_buffer_index; - key.element[nr].input_offset = draw->vertex_element[i].src_offset; + for (i = 0; i < draw->pt.nr_vertex_elements; i++) { + key.element[nr].input_format = draw->pt.vertex_element[i].src_format; + key.element[nr].input_buffer = draw->pt.vertex_element[i].vertex_buffer_index; + key.element[nr].input_offset = draw->pt.vertex_element[i].src_offset; key.element[nr].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; key.element[nr].output_offset = dst_offset; @@ -120,7 +120,7 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, { static struct vertex_header vh = { 0, 0, 0, 0xffff }; fetch->translate->set_buffer(fetch->translate, - draw->nr_vertex_buffers, + draw->pt.nr_vertex_buffers, &vh, 0); } @@ -139,12 +139,12 @@ void draw_pt_fetch_run( struct pt_fetch *fetch, struct translate *translate = fetch->translate; unsigned i; - for (i = 0; i < draw->nr_vertex_buffers; i++) { + for (i = 0; i < draw->pt.nr_vertex_buffers; i++) { translate->set_buffer(translate, i, - ((char *)draw->user.vbuffer[i] + - draw->vertex_buffer[i].buffer_offset), - draw->vertex_buffer[i].pitch ); + ((char *)draw->pt.user.vbuffer[i] + + draw->pt.vertex_buffer[i].buffer_offset), + draw->pt.vertex_buffer[i].pitch ); } translate->run_elts( translate, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 6e4fea460b..1b9b3bfaa6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -117,7 +117,7 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, memset(&key, 0, sizeof(key)); for (i = 0; i < vinfo->num_attribs; i++) { - const struct pipe_vertex_element *src = &draw->vertex_element[vinfo->src_index[i]]; + const struct pipe_vertex_element *src = &draw->pt.vertex_element[vinfo->src_index[i]]; unsigned emit_sz = 0; unsigned input_format = src->src_format; @@ -144,7 +144,7 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, break; case EMIT_1F_PSIZE: input_format = PIPE_FORMAT_R32_FLOAT; - input_buffer = draw->nr_vertex_buffers; + input_buffer = draw->pt.nr_vertex_buffers; input_offset = 0; output_format = PIPE_FORMAT_R32_FLOAT; emit_sz = 1 * sizeof(float); @@ -179,7 +179,7 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, feme->translate = translate_create( &key ); feme->translate->set_buffer(feme->translate, - draw->nr_vertex_buffers, + draw->pt.nr_vertex_buffers, &feme->point_size, 0); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index e1df594035..881e47d59d 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -135,7 +135,7 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, shader->run_linear(shader, (const float (*)[4])pipeline_verts->data, ( float (*)[4])pipeline_verts->data, - (const float (*)[4])draw->user.constants, + (const float (*)[4])draw->pt.user.constants, fetch_count, fpme->vertex_size, fpme->vertex_size); diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 5561f2b6fb..b61bb50664 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -106,7 +106,7 @@ static unsigned add_edgeflag( struct vcache_frontend *vcache, unsigned idx, unsigned mask ) { - if (mask && draw_get_edgeflag(vcache->draw, idx)) + if (mask && draw_pt_get_edgeflag(vcache->draw, idx)) return idx | DRAW_PT_EDGEFLAG; else return idx; diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index f0f62246dd..befcb96bd8 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -611,10 +611,10 @@ st_feedback_draw_vbo(GLcontext *ctx, * unmap vertex/index buffers */ for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { - if (draw->vertex_buffer[i].buffer) { + if (draw->pt.vertex_buffer[i].buffer) { pipe->winsys->buffer_unmap(pipe->winsys, - draw->vertex_buffer[i].buffer); - pipe_buffer_reference(winsys, &draw->vertex_buffer[i].buffer, NULL); + draw->pt.vertex_buffer[i].buffer); + pipe_buffer_reference(winsys, &draw->pt.vertex_buffer[i].buffer, NULL); draw_set_mapped_vertex_buffer(draw, i, NULL); } } -- cgit v1.2.3 From d3cb62b8b3ea03bfb9800bf4b738d9814ef3c516 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 18:02:42 +0100 Subject: draw: fix scons build --- src/gallium/auxiliary/draw/SConscript | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 6f3ca4fa49..91e6a35c5e 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -4,6 +4,7 @@ draw = env.ConvenienceLibrary( target = 'draw', source = [ 'draw_context.c', + 'draw_pipe.c', 'draw_pipe_aaline.c', 'draw_pipe_aapoint.c', 'draw_pipe_clip.c', @@ -16,7 +17,6 @@ draw = env.ConvenienceLibrary( 'draw_pipe_unfilled.c', 'draw_pipe_validate.c', 'draw_pipe_vbuf.c', - 'draw_pipe_vertex.c', 'draw_pipe_wide_line.c', 'draw_pipe_wide_point.c', 'draw_pt.c', @@ -27,6 +27,7 @@ draw = env.ConvenienceLibrary( 'draw_pt_fetch_shade_pipeline.c', 'draw_pt_post_vs.c', 'draw_pt_vcache.c', + 'draw_vertex.c', 'draw_vs.c', 'draw_vs_exec.c', 'draw_vs_llvm.c', -- cgit v1.2.3 From af523a5bd7828fd554669cf83f18992af967a075 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 18:25:33 +0100 Subject: rtasm: include yet another i386 define varient --- src/gallium/auxiliary/rtasm/rtasm_cpu.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.c b/src/gallium/auxiliary/rtasm/rtasm_cpu.c index eb3359750b..d577ff5b42 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_cpu.c +++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.c @@ -32,7 +32,7 @@ int rtasm_cpu_has_sse(void) { /* FIXME: actually detect this at run-time */ -#if defined(__i386__) || defined(__386__) +#if defined(__i386__) || defined(__386__) || defined(i386) return 1; #else return 0; @@ -42,7 +42,7 @@ int rtasm_cpu_has_sse(void) int rtasm_cpu_has_sse2(void) { /* FIXME: actually detect this at run-time */ -#if defined(__i386__) || defined(__386__) +#if defined(__i386__) || defined(__386__) || defined(i386) return 1; #else return 0; -- cgit v1.2.3 From b1158a5e0031aa33a71baa7bc14ca2c0fe0dabc4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 18:26:39 +0100 Subject: translate: don't crash on failure to create sse version --- src/gallium/auxiliary/translate/translate_generic.c | 11 +++++++++++ src/gallium/auxiliary/translate/translate_sse.c | 13 ++++++------- 2 files changed, 17 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index e7fb1dcb2d..9de007c767 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -567,8 +567,14 @@ static void generic_run_elts( struct translate *translate, tg->attrib[attr].output_offset); tg->attrib[attr].fetch( src, data ); + + debug_printf("vert %d/%d attr %d: %f %f %f %f\n", + i, elt, attr, data[0], data[1], data[2], data[3]); + + tg->attrib[attr].emit( data, dst ); } + debug_printf("\n"); vert += tg->translate.key.output_stride; } @@ -602,8 +608,13 @@ static void generic_run( struct translate *translate, tg->attrib[attr].output_offset); tg->attrib[attr].fetch( src, data ); + + debug_printf("vert %d attr %d: %f %f %f %f\n", + i, attr, data[0], data[1], data[2], data[3]); + tg->attrib[attr].emit( data, dst ); } + debug_printf("\n"); vert += tg->translate.key.output_stride; } diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index cb8815c173..575852d222 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -145,7 +145,7 @@ static struct x86_reg get_inv_255( struct translate_sse *p ) p->inv_255[0] = p->inv_255[1] = p->inv_255[2] = - p->inv_255[3] = 1.0 / 255.0f; + p->inv_255[3] = 1.0f / 255.0f; sse_movups(p->func, reg, x86_make_disp(translateESI, @@ -575,22 +575,21 @@ struct translate *translate_sse2_create( const struct translate_key *key ) if (p == NULL) goto fail; - if (!rtasm_cpu_has_sse() || !rtasm_cpu_has_sse2()) - goto fail; - - - p->translate.key = *key; p->translate.release = translate_sse_release; p->translate.set_buffer = translate_sse_set_buffer; p->translate.run_elts = translate_sse_run_elts; p->translate.run = translate_sse_run; + if (!rtasm_cpu_has_sse() || !rtasm_cpu_has_sse2()) + goto fail; + if (!build_vertex_emit(p, &p->linear_func, TRUE)) goto fail; if (!build_vertex_emit(p, &p->elt_func, FALSE)) goto fail; + p->translate.key = *key; p->gen_run = (run_func)x86_get_func(&p->linear_func); p->gen_run_elts = (run_elts_func)x86_get_func(&p->elt_func); @@ -598,7 +597,7 @@ struct translate *translate_sse2_create( const struct translate_key *key ) fail: if (p) - p->translate.release( &p->translate ); + translate_sse_release( &p->translate ); return NULL; } -- cgit v1.2.3 From bfd179776f5ded75c2134a54f0a57a1579118cd0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 18:41:53 +0100 Subject: draw: add missing translate->set_buffer for fetch emit path --- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 1b9b3bfaa6..70d136dce6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -185,6 +185,14 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, } feme->point_size = draw->rasterizer->point_size; + + for (i = 0; i < draw->pt.nr_vertex_buffers; i++) { + feme->translate->set_buffer(feme->translate, + i, + ((char *)draw->pt.user.vbuffer[i] + + draw->pt.vertex_buffer[i].buffer_offset), + draw->pt.vertex_buffer[i].pitch ); + } } -- cgit v1.2.3 From d0a4bf08b1a80d62f81301c5b37723dfca436b62 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 20:18:40 +0100 Subject: translate: fix several bugs - specify cdecl calling convention on WIN32 - fix load bgra8 function - fix previous don't crash fix. --- .../auxiliary/translate/translate_generic.c | 11 +++--- src/gallium/auxiliary/translate/translate_sse.c | 42 ++++++++++++---------- 2 files changed, 28 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index 9de007c767..402780ee53 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -568,13 +568,11 @@ static void generic_run_elts( struct translate *translate, tg->attrib[attr].fetch( src, data ); - debug_printf("vert %d/%d attr %d: %f %f %f %f\n", - i, elt, attr, data[0], data[1], data[2], data[3]); - + if (0) debug_printf("vert %d/%d attr %d: %f %f %f %f\n", + i, elt, attr, data[0], data[1], data[2], data[3]); tg->attrib[attr].emit( data, dst ); } - debug_printf("\n"); vert += tg->translate.key.output_stride; } @@ -609,12 +607,11 @@ static void generic_run( struct translate *translate, tg->attrib[attr].fetch( src, data ); - debug_printf("vert %d attr %d: %f %f %f %f\n", - i, attr, data[0], data[1], data[2], data[3]); + if (0) debug_printf("vert %d attr %d: %f %f %f %f\n", + i, attr, data[0], data[1], data[2], data[3]); tg->attrib[attr].emit( data, dst ); } - debug_printf("\n"); vert += tg->translate.key.output_stride; } diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index 575852d222..e587e5afec 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -45,15 +45,21 @@ #define W 3 -typedef void (*run_func)( struct translate *translate, - unsigned start, - unsigned count, - void *output_buffer ); +#ifdef WIN32 +#define RTASM __cdecl +#else +#define RTASM +#endif + +typedef void (RTASM *run_func)( struct translate *translate, + unsigned start, + unsigned count, + void *output_buffer ); -typedef void (*run_elts_func)( struct translate *translate, - const unsigned *elts, - unsigned count, - void *output_buffer ); +typedef void (RTASM *run_elts_func)( struct translate *translate, + const unsigned *elts, + unsigned count, + void *output_buffer ); @@ -212,17 +218,17 @@ static void emit_load_R8G8B8A8_UNORM( struct translate_sse *p, /* Load and unpack twice: */ sse_movss(p->func, data, src); - sse2_punpcklbw(p->func, src, get_identity(p)); - sse2_punpcklbw(p->func, src, get_identity(p)); + sse2_punpcklbw(p->func, data, get_identity(p)); + sse2_punpcklbw(p->func, data, get_identity(p)); /* Convert to float: */ - sse2_cvtdq2ps(p->func, src, src); + sse2_cvtdq2ps(p->func, data, data); /* Scale by 1/255.0 */ - sse_mulps(p->func, src, get_inv_255(p)); + sse_mulps(p->func, data, get_inv_255(p)); } @@ -551,7 +557,6 @@ static void translate_sse_run_elts( struct translate *translate, elts, count, output_buffer ); - } static void translate_sse_run( struct translate *translate, @@ -570,26 +575,27 @@ static void translate_sse_run( struct translate *translate, struct translate *translate_sse2_create( const struct translate_key *key ) { - struct translate_sse *p = CALLOC_STRUCT( translate_sse ); + struct translate_sse *p = NULL; + + if (!rtasm_cpu_has_sse() || !rtasm_cpu_has_sse2()) + goto fail; + p = CALLOC_STRUCT( translate_sse ); if (p == NULL) goto fail; + p->translate.key = *key; p->translate.release = translate_sse_release; p->translate.set_buffer = translate_sse_set_buffer; p->translate.run_elts = translate_sse_run_elts; p->translate.run = translate_sse_run; - if (!rtasm_cpu_has_sse() || !rtasm_cpu_has_sse2()) - goto fail; - if (!build_vertex_emit(p, &p->linear_func, TRUE)) goto fail; if (!build_vertex_emit(p, &p->elt_func, FALSE)) goto fail; - p->translate.key = *key; p->gen_run = (run_func)x86_get_func(&p->linear_func); p->gen_run_elts = (run_elts_func)x86_get_func(&p->elt_func); -- cgit v1.2.3 From 68a7cb21fa14eac9e38bf398623739a892cc0d52 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 19 Apr 2008 20:20:40 +0100 Subject: draw: rearrange debug code --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 55 +------------------------ src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 16 +++++-- src/gallium/auxiliary/draw/draw_vertex.c | 54 ++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vertex.h | 2 + 4 files changed, 70 insertions(+), 57 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index c835727e32..c5810febe0 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -115,59 +115,6 @@ check_space( struct vbuf_stage *vbuf, unsigned nr ) } -static INLINE void -dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) -{ -// assert(vinfo == vbuf->render->get_vertex_info(vbuf->render)); - unsigned i, j; - - for (i = 0; i < vinfo->num_attribs; i++) { - j = vinfo->src_index[i]; - switch (vinfo->emit[i]) { - case EMIT_OMIT: - debug_printf("EMIT_OMIT:"); - break; - case EMIT_1F: - debug_printf("EMIT_1F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_1F_PSIZE: - debug_printf("EMIT_1F_PSIZE:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_2F: - debug_printf("EMIT_2F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_3F: - debug_printf("EMIT_3F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - data += sizeof(float); - break; - case EMIT_4F: - debug_printf("EMIT_4F:\t"); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - debug_printf("%f ", *(float *)data); data += sizeof(float); - break; - case EMIT_4UB: - debug_printf("EMIT_4UB:\t"); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - debug_printf("%u ", *data++); - break; - default: - assert(0); - } - debug_printf("\n"); - } - debug_printf("\n"); -} /** @@ -189,7 +136,7 @@ emit_vertex( struct vbuf_stage *vbuf, vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0); vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr); - if (0) dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); + if (0) draw_dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); vbuf->vertex_ptr += vbuf->vertex_size/4; vertex->vertex_id = vbuf->nr_vertices++; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 70d136dce6..002ced2186 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -75,6 +75,7 @@ struct fetch_emit_middle_end { struct draw_context *draw; struct translate *translate; + const struct vertex_info *vinfo; /* Cache point size somewhere it's address won't change: */ @@ -106,9 +107,9 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, /* Must do this after set_primitive() above: */ - vinfo = draw->render->get_vertex_info(draw->render); - - + vinfo = feme->vinfo = draw->render->get_vertex_info(draw->render); + + /* Transform from API vertices to HW vertices, skipping the * pipeline_vertex intermediate step. @@ -229,6 +230,15 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, fetch_count, hw_verts ); + if (0) { + unsigned i; + for (i = 0; i < fetch_count; i++) { + debug_printf("\n\nvertex %d:\n", i); + draw_dump_emitted_vertex( feme->vinfo, + (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i ); + } + } + /* XXX: Draw arrays path to avoid re-emitting index list again and * again. */ diff --git a/src/gallium/auxiliary/draw/draw_vertex.c b/src/gallium/auxiliary/draw/draw_vertex.c index a42adaafb1..1446f785c5 100644 --- a/src/gallium/auxiliary/draw/draw_vertex.c +++ b/src/gallium/auxiliary/draw/draw_vertex.c @@ -73,3 +73,57 @@ draw_compute_vertex_size(struct vertex_info *vinfo) } } } + + +void +draw_dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data) +{ + unsigned i, j; + + for (i = 0; i < vinfo->num_attribs; i++) { + j = vinfo->src_index[i]; + switch (vinfo->emit[i]) { + case EMIT_OMIT: + debug_printf("EMIT_OMIT:"); + break; + case EMIT_1F: + debug_printf("EMIT_1F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_1F_PSIZE: + debug_printf("EMIT_1F_PSIZE:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_2F: + debug_printf("EMIT_2F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_3F: + debug_printf("EMIT_3F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + data += sizeof(float); + break; + case EMIT_4F: + debug_printf("EMIT_4F:\t"); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + debug_printf("%f ", *(float *)data); data += sizeof(float); + break; + case EMIT_4UB: + debug_printf("EMIT_4UB:\t"); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + debug_printf("%u ", *data++); + break; + default: + assert(0); + } + debug_printf("\n"); + } + debug_printf("\n"); +} diff --git a/src/gallium/auxiliary/draw/draw_vertex.h b/src/gallium/auxiliary/draw/draw_vertex.h index 65818463ca..6d8bac5138 100644 --- a/src/gallium/auxiliary/draw/draw_vertex.h +++ b/src/gallium/auxiliary/draw/draw_vertex.h @@ -106,5 +106,7 @@ draw_emit_vertex_attr(struct vertex_info *vinfo, extern void draw_compute_vertex_size(struct vertex_info *vinfo); +void draw_dump_emitted_vertex(const struct vertex_info *vinfo, + const uint8_t *data); #endif /* DRAW_VERTEX_H */ -- cgit v1.2.3 From 29858e1b553cee1fd7e3380ea62c69d2a6b91b95 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 20 Apr 2008 14:41:02 +0900 Subject: gallium: Refcount textures. Pipe driver does refcount textures. If cso_context does not, dangling pointers appear. --- src/gallium/auxiliary/cso_cache/cso_context.c | 37 ++++++++++++++++++++------- 1 file changed, 28 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 4a1a6cb79c..746b176185 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -25,16 +25,19 @@ * **************************************************************************/ - /* Wrap the cso cache & hash mechanisms in a simplified + /** + * @file + * + * Wrap the cso cache & hash mechanisms in a simplified * pipe-driver-specific interface. * - * Authors: - * Zack Rusin - * Keith Whitwell + * @author Zack Rusin + * @author Keith Whitwell */ #include "pipe/p_state.h" #include "pipe/p_util.h" +#include "pipe/p_inlines.h" #include "cso_cache/cso_context.h" #include "cso_cache/cso_cache.h" @@ -277,23 +280,39 @@ void cso_set_sampler_textures( struct cso_context *ctx, ctx->nr_textures = count; for (i = 0; i < count; i++) - ctx->textures[i] = textures[i]; + pipe_texture_reference(&ctx->textures[i], textures[i]); for ( ; i < PIPE_MAX_SAMPLERS; i++) - ctx->textures[i] = NULL; + pipe_texture_reference(&ctx->textures[i], NULL); ctx->pipe->set_sampler_textures(ctx->pipe, count, textures); } void cso_save_sampler_textures( struct cso_context *ctx ) { + uint i; + ctx->nr_textures_saved = ctx->nr_textures; - memcpy(ctx->textures_saved, ctx->textures, sizeof(ctx->textures)); + for (i = 0; i < ctx->nr_textures; i++) { + assert(!ctx->textures_saved[i]); + pipe_texture_reference(&ctx->textures_saved[i], ctx->textures[i]); + } } void cso_restore_sampler_textures( struct cso_context *ctx ) { - cso_set_sampler_textures(ctx, ctx->nr_textures_saved, ctx->textures_saved); - ctx->nr_textures_saved = 0; + uint i; + + ctx->nr_textures = ctx->nr_textures_saved; + + for (i = 0; i < ctx->nr_textures; i++) { + pipe_texture_reference(&ctx->textures[i], NULL); + ctx->textures[i] = ctx->textures_saved[i]; + ctx->textures_saved[i] = NULL; + } + for ( ; i < PIPE_MAX_SAMPLERS; i++) + pipe_texture_reference(&ctx->textures[i], NULL); + + ctx->pipe->set_sampler_textures(ctx->pipe, ctx->nr_textures, ctx->textures); } -- cgit v1.2.3 From 40e0439db448a7d93ddb18faac7f14b47b1343c0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 21 Apr 2008 13:02:59 +0900 Subject: gallium: Centralize SSE usage logic. --- src/gallium/auxiliary/draw/draw_context.c | 12 ------------ src/gallium/auxiliary/draw/draw_context.h | 2 -- src/gallium/auxiliary/draw/draw_private.h | 2 -- src/gallium/auxiliary/draw/draw_vs_sse.c | 3 ++- src/gallium/auxiliary/rtasm/rtasm_cpu.c | 10 ++++++++-- 5 files changed, 10 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 4988d67faa..b4dbdccd61 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -45,12 +45,6 @@ struct draw_context *draw_create( void ) if (draw == NULL) goto fail; -#if defined(__i386__) || defined(__386__) - draw->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; -#else - draw->use_sse = FALSE; -#endif - ASSIGN_4V( draw->plane[0], -1, 0, 0, 1 ); ASSIGN_4V( draw->plane[1], 1, 0, 0, 1 ); ASSIGN_4V( draw->plane[2], 0, -1, 0, 1 ); @@ -320,12 +314,6 @@ draw_num_vs_outputs(struct draw_context *draw) -boolean draw_use_sse(struct draw_context *draw) -{ - return (boolean) draw->use_sse; -} - - void draw_set_render( struct draw_context *draw, struct vbuf_render *render ) { diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index a0ac980c89..68e2efb865 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -73,8 +73,6 @@ void draw_enable_line_stipple(struct draw_context *draw, boolean enable); void draw_enable_point_sprites(struct draw_context *draw, boolean enable); -boolean draw_use_sse(struct draw_context *draw); - void draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 27f61c2f40..da973e868b 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -192,8 +192,6 @@ struct draw_context float plane[12][4]; unsigned nr_planes; - boolean use_sse; - /* If a prim stage introduces new vertex attributes, they'll be stored here */ struct { diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 8e2d381f14..b1e9f67114 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -41,6 +41,7 @@ #include "draw_private.h" #include "draw_context.h" +#include "rtasm/rtasm_cpu.h" #include "rtasm/rtasm_x86sse.h" #include "tgsi/exec/tgsi_sse2.h" #include "tgsi/util/tgsi_parse.h" @@ -155,7 +156,7 @@ draw_create_vs_sse(struct draw_context *draw, struct draw_sse_vertex_shader *vs; uint nt = tgsi_num_tokens(templ->tokens); - if (!draw->use_sse) + if (!rtasm_cpu_has_sse2()) return NULL; vs = CALLOC_STRUCT( draw_sse_vertex_shader ); diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.c b/src/gallium/auxiliary/rtasm/rtasm_cpu.c index d577ff5b42..175245a9f6 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_cpu.c +++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.c @@ -26,14 +26,20 @@ **************************************************************************/ +#include "pipe/p_debug.h" #include "rtasm_cpu.h" +static boolean rtasm_sse_enabled(void) +{ + return !debug_get_bool_option("GALLIUM_NOSSE", FALSE); +} + int rtasm_cpu_has_sse(void) { /* FIXME: actually detect this at run-time */ #if defined(__i386__) || defined(__386__) || defined(i386) - return 1; + return rtasm_sse_enabled(); #else return 0; #endif @@ -43,7 +49,7 @@ int rtasm_cpu_has_sse2(void) { /* FIXME: actually detect this at run-time */ #if defined(__i386__) || defined(__386__) || defined(i386) - return 1; + return rtasm_sse_enabled(); #else return 0; #endif -- cgit v1.2.3 From 201ac414d4df00745e487a6ffbc9979a2e70f0c6 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 21 Apr 2008 00:10:39 -0400 Subject: make llvm draw paths compile with the latest changes switch the method of distribution of builtins (to get rid of the llvm2cpp dependency) --- src/gallium/auxiliary/draw/draw_vs_llvm.c | 13 +- src/gallium/auxiliary/gallivm/Makefile | 8 +- src/gallium/auxiliary/gallivm/gallivm_builtins.cpp | 708 ++++----------------- src/gallium/auxiliary/gallivm/instructions.cpp | 11 +- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 37 +- src/gallium/auxiliary/gallivm/llvm_builtins.c | 2 +- src/gallium/auxiliary/gallivm/soabuiltins.c | 2 +- 7 files changed, 199 insertions(+), 582 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index c219a91156..2ebf05526a 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -73,16 +73,17 @@ vs_llvm_run_linear( struct draw_vertex_shader *base, (struct draw_llvm_vertex_shader *)base; struct tgsi_exec_machine *machine = shader->machine; - unsigned int j; + unsigned int i, j; + unsigned slot; for (i = 0; i < count; i += MAX_TGSI_VERTICES) { unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); - /* Swizzle inputs. + /* Swizzle inputs. */ for (j = 0; j < max_vertices; j++) { - for (slot = 0; slot < draw->num_vs_inputs; slot++) { + for (slot = 0; slot < base->info.num_inputs; slot++) { machine->Inputs[slot].xyzw[0].f[j] = input[slot][0]; machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; @@ -102,7 +103,7 @@ vs_llvm_run_linear( struct draw_vertex_shader *base, /* Unswizzle all output results */ - for (slot = 1; slot < draw->num_vs_outputs; slot++) { + for (slot = 1; slot < base->info.num_inputs; slot++) { output[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; @@ -145,7 +146,7 @@ draw_create_vs_llvm(struct draw_context *draw, /* we make a private copy of the tokens */ vs->base.state.tokens = mem_dup(templ->tokens, nt * sizeof(templ->tokens[0])); - tgsi_scan_shader(shader->tokens, &vs->base.info); + tgsi_scan_shader(vs->base.state.tokens, &vs->base.info); vs->base.prepare = vs_llvm_prepare; vs->base.run_linear = vs_llvm_run_linear; @@ -156,7 +157,7 @@ draw_create_vs_llvm(struct draw_context *draw, struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); gallivm_ir_set_layout(ir, GALLIVM_SOA); gallivm_ir_set_components(ir, 4); - gallivm_ir_fill_from_tgsi(ir, vs->base.state->tokens); + gallivm_ir_fill_from_tgsi(ir, vs->base.state.tokens); vs->llvm_prog = gallivm_ir_compile(ir); gallivm_ir_delete(ir); } diff --git a/src/gallium/auxiliary/gallivm/Makefile b/src/gallium/auxiliary/gallivm/Makefile index c24e19e062..6bfd187fb3 100644 --- a/src/gallium/auxiliary/gallivm/Makefile +++ b/src/gallium/auxiliary/gallivm/Makefile @@ -65,10 +65,14 @@ depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) gallivm_builtins.cpp: llvm_builtins.c - clang --emit-llvm < $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=$@ -f -for=shader -funcname=createGallivmBuiltins + clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp.bin + (echo "static const unsigned char llvm_builtins_data[] = {"; od -txC temp.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/};/") >$@ + rm temp.bin gallivmsoabuiltins.cpp: soabuiltins.c - clang --emit-llvm < $< |llvm-as|opt -std-compile-opts|llvm2cpp -gen-module -o=$@ -f -for=shader -funcname=createSoaBuiltins + clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp.bin + (echo "static const unsigned char soabuiltins_data[] = {"; od -txC temp.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/};/") >$@ + rm temp.bin # Emacs tags tags: diff --git a/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp b/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp index 1796f0a177..a6f8cd043b 100644 --- a/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp +++ b/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp @@ -1,567 +1,141 @@ -// Generated by llvm2cpp - DO NOT MODIFY! - - -Module* createGallivmBuiltins(Module *mod) { - -mod->setModuleIdentifier("shader"); - -// Type Definitions -ArrayType* ArrayTy_0 = ArrayType::get(IntegerType::get(8), 25); - -PointerType* PointerTy_1 = PointerType::get(ArrayTy_0, 0); - -std::vectorFuncTy_2_args; -FuncTy_2_args.push_back(Type::FloatTy); -FuncTy_2_args.push_back(Type::FloatTy); -FunctionType* FuncTy_2 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_2_args, - /*isVarArg=*/false); - -PointerType* PointerTy_3 = PointerType::get(FuncTy_2, 0); - -VectorType* VectorTy_4 = VectorType::get(Type::FloatTy, 4); - -std::vectorFuncTy_5_args; -FuncTy_5_args.push_back(VectorTy_4); -FunctionType* FuncTy_5 = FunctionType::get( - /*Result=*/VectorTy_4, - /*Params=*/FuncTy_5_args, - /*isVarArg=*/false); - -std::vectorFuncTy_6_args; -FuncTy_6_args.push_back(VectorTy_4); -FuncTy_6_args.push_back(VectorTy_4); -FuncTy_6_args.push_back(VectorTy_4); -FunctionType* FuncTy_6 = FunctionType::get( - /*Result=*/VectorTy_4, - /*Params=*/FuncTy_6_args, - /*isVarArg=*/false); - -VectorType* VectorTy_7 = VectorType::get(IntegerType::get(32), 4); - -std::vectorFuncTy_9_args; -FunctionType* FuncTy_9 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_9_args, - /*isVarArg=*/true); - -PointerType* PointerTy_8 = PointerType::get(FuncTy_9, 0); - -PointerType* PointerTy_10 = PointerType::get(IntegerType::get(8), 0); - -std::vectorFuncTy_12_args; -FuncTy_12_args.push_back(Type::FloatTy); -FunctionType* FuncTy_12 = FunctionType::get( - /*Result=*/Type::FloatTy, - /*Params=*/FuncTy_12_args, - /*isVarArg=*/false); - -PointerType* PointerTy_11 = PointerType::get(FuncTy_12, 0); - -std::vectorFuncTy_13_args; -FuncTy_13_args.push_back(VectorTy_4); -FunctionType* FuncTy_13 = FunctionType::get( - /*Result=*/IntegerType::get(32), - /*Params=*/FuncTy_13_args, - /*isVarArg=*/false); - - -// Function Declarations - -Function* func_approx = new Function( - /*Type=*/FuncTy_2, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"approx", mod); -func_approx->setCallingConv(CallingConv::C); -const ParamAttrsList *func_approx_PAL = 0; -func_approx->setParamAttrs(func_approx_PAL); - -Function* func_powf = new Function( - /*Type=*/FuncTy_2, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"powf", mod); // (external, no body) -func_powf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_powf_PAL = 0; -func_powf->setParamAttrs(func_powf_PAL); - -Function* func_lit = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"lit", mod); -func_lit->setCallingConv(CallingConv::C); -const ParamAttrsList *func_lit_PAL = 0; -func_lit->setParamAttrs(func_lit_PAL); - -Function* func_cmp = new Function( - /*Type=*/FuncTy_6, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"cmp", mod); -func_cmp->setCallingConv(CallingConv::C); -const ParamAttrsList *func_cmp_PAL = 0; -{ - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_cmp_PAL = ParamAttrsList::get(Attrs); - -} -func_cmp->setParamAttrs(func_cmp_PAL); - -Function* func_vcos = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"vcos", mod); -func_vcos->setCallingConv(CallingConv::C); -const ParamAttrsList *func_vcos_PAL = 0; -func_vcos->setParamAttrs(func_vcos_PAL); - -Function* func_printf = new Function( - /*Type=*/FuncTy_9, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"printf", mod); // (external, no body) -func_printf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_printf_PAL = 0; -func_printf->setParamAttrs(func_printf_PAL); - -Function* func_cosf = new Function( - /*Type=*/FuncTy_12, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"cosf", mod); // (external, no body) -func_cosf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_cosf_PAL = 0; -func_cosf->setParamAttrs(func_cosf_PAL); - -Function* func_scs = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"scs", mod); -func_scs->setCallingConv(CallingConv::C); -const ParamAttrsList *func_scs_PAL = 0; -func_scs->setParamAttrs(func_scs_PAL); - -Function* func_sinf = new Function( - /*Type=*/FuncTy_12, - /*Linkage=*/GlobalValue::ExternalLinkage, - /*Name=*/"sinf", mod); // (external, no body) -func_sinf->setCallingConv(CallingConv::C); -const ParamAttrsList *func_sinf_PAL = 0; -func_sinf->setParamAttrs(func_sinf_PAL); - -Function* func_vsin = new Function( - /*Type=*/FuncTy_5, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"vsin", mod); -func_vsin->setCallingConv(CallingConv::C); -const ParamAttrsList *func_vsin_PAL = 0; -func_vsin->setParamAttrs(func_vsin_PAL); - -Function* func_kilp = new Function( - /*Type=*/FuncTy_13, - /*Linkage=*/GlobalValue::WeakLinkage, - /*Name=*/"kilp", mod); -func_kilp->setCallingConv(CallingConv::C); -const ParamAttrsList *func_kilp_PAL = 0; -{ - ParamAttrsVector Attrs; - ParamAttrsWithIndex PAWI; - PAWI.index = 0; PAWI.attrs = 0 | ParamAttr::NoUnwind; - Attrs.push_back(PAWI); - func_kilp_PAL = ParamAttrsList::get(Attrs); - -} -func_kilp->setParamAttrs(func_kilp_PAL); - -// Global Variable Declarations - - -GlobalVariable* gvar_array__str = new GlobalVariable( -/*Type=*/ArrayTy_0, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str", -mod); - -GlobalVariable* gvar_array__str1 = new GlobalVariable( -/*Type=*/ArrayTy_0, -/*isConstant=*/true, -/*Linkage=*/GlobalValue::InternalLinkage, -/*Initializer=*/0, // has initializer, specified below -/*Name=*/".str1", -mod); - -// Constant Definitions -Constant* const_array_14 = ConstantArray::get("VEC IN is %f %f %f %f\x0A", true); -Constant* const_array_15 = ConstantArray::get("VEC OUT is %f %f %f %f\x0A", true); -ConstantFP* const_float_16 = ConstantFP::get(Type::FloatTy, APFloat(-1.280000e+02f)); -ConstantFP* const_float_17 = ConstantFP::get(Type::FloatTy, APFloat(1.280000e+02f)); -Constant* const_float_18 = Constant::getNullValue(Type::FloatTy); -Constant* const_int32_19 = Constant::getNullValue(IntegerType::get(32)); -std::vector const_packed_20_elems; -ConstantFP* const_float_21 = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); -const_packed_20_elems.push_back(const_float_21); -UndefValue* const_float_22 = UndefValue::get(Type::FloatTy); -const_packed_20_elems.push_back(const_float_22); -const_packed_20_elems.push_back(const_float_22); -const_packed_20_elems.push_back(const_float_21); -Constant* const_packed_20 = ConstantVector::get(VectorTy_4, const_packed_20_elems); -ConstantInt* const_int32_23 = ConstantInt::get(APInt(32, "1", 10)); -ConstantInt* const_int32_24 = ConstantInt::get(APInt(32, "3", 10)); -ConstantInt* const_int32_25 = ConstantInt::get(APInt(32, "2", 10)); -std::vector const_packed_26_elems; -const_packed_26_elems.push_back(const_float_21); -const_packed_26_elems.push_back(const_float_18); -const_packed_26_elems.push_back(const_float_18); -const_packed_26_elems.push_back(const_float_21); -Constant* const_packed_26 = ConstantVector::get(VectorTy_4, const_packed_26_elems); -Constant* const_double_27 = Constant::getNullValue(Type::DoubleTy); -std::vector const_packed_28_elems; -const_packed_28_elems.push_back(const_int32_19); -ConstantInt* const_int32_29 = ConstantInt::get(APInt(32, "5", 10)); -const_packed_28_elems.push_back(const_int32_29); -const_packed_28_elems.push_back(const_int32_25); -const_packed_28_elems.push_back(const_int32_24); -Constant* const_packed_28 = ConstantVector::get(VectorTy_7, const_packed_28_elems); -std::vector const_packed_30_elems; -const_packed_30_elems.push_back(const_int32_19); -const_packed_30_elems.push_back(const_int32_23); -ConstantInt* const_int32_31 = ConstantInt::get(APInt(32, "6", 10)); -const_packed_30_elems.push_back(const_int32_31); -const_packed_30_elems.push_back(const_int32_24); -Constant* const_packed_30 = ConstantVector::get(VectorTy_7, const_packed_30_elems); -std::vector const_packed_32_elems; -const_packed_32_elems.push_back(const_int32_19); -const_packed_32_elems.push_back(const_int32_23); -const_packed_32_elems.push_back(const_int32_25); -ConstantInt* const_int32_33 = ConstantInt::get(APInt(32, "7", 10)); -const_packed_32_elems.push_back(const_int32_33); -Constant* const_packed_32 = ConstantVector::get(VectorTy_7, const_packed_32_elems); -std::vector const_ptr_34_indices; -const_ptr_34_indices.push_back(const_int32_19); -const_ptr_34_indices.push_back(const_int32_19); -Constant* const_ptr_34 = ConstantExpr::getGetElementPtr(gvar_array__str, &const_ptr_34_indices[0], const_ptr_34_indices.size() ); -UndefValue* const_packed_35 = UndefValue::get(VectorTy_4); -std::vector const_ptr_36_indices; -const_ptr_36_indices.push_back(const_int32_19); -const_ptr_36_indices.push_back(const_int32_19); -Constant* const_ptr_36 = ConstantExpr::getGetElementPtr(gvar_array__str1, &const_ptr_36_indices[0], const_ptr_36_indices.size() ); - -// Global Variable Definitions -gvar_array__str->setInitializer(const_array_14); -gvar_array__str1->setInitializer(const_array_15); - -// Function Definitions - -// Function: approx (func_approx) -{ - Function::arg_iterator args = func_approx->arg_begin(); - Value* float_a = args++; - float_a->setName("a"); - Value* float_b = args++; - float_b->setName("b"); - - BasicBlock* label_entry = new BasicBlock("entry",func_approx,0); - - // Block entry (label_entry) - FCmpInst* int1_cmp = new FCmpInst(FCmpInst::FCMP_OLT, float_b, const_float_16, "cmp", label_entry); - SelectInst* float_b_addr_0 = new SelectInst(int1_cmp, const_float_16, float_b, "b.addr.0", label_entry); - FCmpInst* int1_cmp3 = new FCmpInst(FCmpInst::FCMP_OGT, float_b_addr_0, const_float_17, "cmp3", label_entry); - SelectInst* float_b_addr_1 = new SelectInst(int1_cmp3, const_float_17, float_b_addr_0, "b.addr.1", label_entry); - FCmpInst* int1_cmp7 = new FCmpInst(FCmpInst::FCMP_OLT, float_a, const_float_18, "cmp7", label_entry); - SelectInst* float_a_addr_0 = new SelectInst(int1_cmp7, const_float_18, float_a, "a.addr.0", label_entry); - std::vector float_call_params; - float_call_params.push_back(float_a_addr_0); - float_call_params.push_back(float_b_addr_1); - CallInst* float_call = new CallInst(func_powf, float_call_params.begin(), float_call_params.end(), "call", label_entry); - float_call->setCallingConv(CallingConv::C); - float_call->setTailCall(true);const ParamAttrsList *float_call_PAL = 0; - float_call->setParamAttrs(float_call_PAL); - - new ReturnInst(float_call, label_entry); - -} - -// Function: lit (func_lit) -{ - Function::arg_iterator args = func_lit->arg_begin(); - Value* packed_tmp = args++; - packed_tmp->setName("tmp"); - - BasicBlock* label_entry_38 = new BasicBlock("entry",func_lit,0); - BasicBlock* label_ifthen = new BasicBlock("ifthen",func_lit,0); - BasicBlock* label_UnifiedReturnBlock = new BasicBlock("UnifiedReturnBlock",func_lit,0); - - // Block entry (label_entry_38) - ExtractElementInst* float_tmp6 = new ExtractElementInst(packed_tmp, const_int32_19, "tmp6", label_entry_38); - FCmpInst* int1_cmp_39 = new FCmpInst(FCmpInst::FCMP_OGT, float_tmp6, const_float_18, "cmp", label_entry_38); - new BranchInst(label_ifthen, label_UnifiedReturnBlock, int1_cmp_39, label_entry_38); - - // Block ifthen (label_ifthen) - InsertElementInst* packed_tmp10 = new InsertElementInst(const_packed_20, float_tmp6, const_int32_23, "tmp10", label_ifthen); - ExtractElementInst* float_tmp12 = new ExtractElementInst(packed_tmp, const_int32_23, "tmp12", label_ifthen); - ExtractElementInst* float_tmp14 = new ExtractElementInst(packed_tmp, const_int32_24, "tmp14", label_ifthen); - std::vector float_call_41_params; - float_call_41_params.push_back(float_tmp12); - float_call_41_params.push_back(float_tmp14); - CallInst* float_call_41 = new CallInst(func_approx, float_call_41_params.begin(), float_call_41_params.end(), "call", label_ifthen); - float_call_41->setCallingConv(CallingConv::C); - float_call_41->setTailCall(true);const ParamAttrsList *float_call_41_PAL = 0; - float_call_41->setParamAttrs(float_call_41_PAL); - - InsertElementInst* packed_tmp16 = new InsertElementInst(packed_tmp10, float_call_41, const_int32_25, "tmp16", label_ifthen); - new ReturnInst(packed_tmp16, label_ifthen); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock) - new ReturnInst(const_packed_26, label_UnifiedReturnBlock); - -} - -// Function: cmp (func_cmp) -{ - Function::arg_iterator args = func_cmp->arg_begin(); - Value* packed_tmp0 = args++; - packed_tmp0->setName("tmp0"); - Value* packed_tmp1 = args++; - packed_tmp1->setName("tmp1"); - Value* packed_tmp2 = args++; - packed_tmp2->setName("tmp2"); - - BasicBlock* label_entry_44 = new BasicBlock("entry",func_cmp,0); - BasicBlock* label_cond__14 = new BasicBlock("cond.?14",func_cmp,0); - BasicBlock* label_cond_cont20 = new BasicBlock("cond.cont20",func_cmp,0); - BasicBlock* label_cond__28 = new BasicBlock("cond.?28",func_cmp,0); - BasicBlock* label_cond_cont34 = new BasicBlock("cond.cont34",func_cmp,0); - BasicBlock* label_cond__42 = new BasicBlock("cond.?42",func_cmp,0); - BasicBlock* label_cond_cont48 = new BasicBlock("cond.cont48",func_cmp,0); - - // Block entry (label_entry_44) - ExtractElementInst* float_tmp3 = new ExtractElementInst(packed_tmp0, const_int32_19, "tmp3", label_entry_44); - CastInst* double_conv = new FPExtInst(float_tmp3, Type::DoubleTy, "conv", label_entry_44); - FCmpInst* int1_cmp_45 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv, const_double_27, "cmp", label_entry_44); - ExtractElementInst* float_tmp11 = new ExtractElementInst(packed_tmp0, const_int32_23, "tmp11", label_entry_44); - CastInst* double_conv12 = new FPExtInst(float_tmp11, Type::DoubleTy, "conv12", label_entry_44); - FCmpInst* int1_cmp13 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv12, const_double_27, "cmp13", label_entry_44); - SelectInst* packed_tmp1_tmp2 = new SelectInst(int1_cmp_45, packed_tmp1, packed_tmp2, "tmp1.tmp2", label_entry_44); - new BranchInst(label_cond__14, label_cond_cont20, int1_cmp13, label_entry_44); - - // Block cond.?14 (label_cond__14) - ShuffleVectorInst* packed_tmp233 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp1, const_packed_28, "tmp233", label_cond__14); - ExtractElementInst* float_tmp254 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp254", label_cond__14); - CastInst* double_conv265 = new FPExtInst(float_tmp254, Type::DoubleTy, "conv265", label_cond__14); - FCmpInst* int1_cmp276 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv265, const_double_27, "cmp276", label_cond__14); - new BranchInst(label_cond__28, label_cond_cont34, int1_cmp276, label_cond__14); - - // Block cond.cont20 (label_cond_cont20) - ShuffleVectorInst* packed_tmp23 = new ShuffleVectorInst(packed_tmp1_tmp2, packed_tmp2, const_packed_28, "tmp23", label_cond_cont20); - ExtractElementInst* float_tmp25 = new ExtractElementInst(packed_tmp0, const_int32_25, "tmp25", label_cond_cont20); - CastInst* double_conv26 = new FPExtInst(float_tmp25, Type::DoubleTy, "conv26", label_cond_cont20); - FCmpInst* int1_cmp27 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv26, const_double_27, "cmp27", label_cond_cont20); - new BranchInst(label_cond__28, label_cond_cont34, int1_cmp27, label_cond_cont20); - - // Block cond.?28 (label_cond__28) - PHINode* packed_tmp23_reg2mem_0 = new PHINode(VectorTy_4, "tmp23.reg2mem.0", label_cond__28); - packed_tmp23_reg2mem_0->reserveOperandSpace(2); - packed_tmp23_reg2mem_0->addIncoming(packed_tmp233, label_cond__14); - packed_tmp23_reg2mem_0->addIncoming(packed_tmp23, label_cond_cont20); - - ShuffleVectorInst* packed_tmp378 = new ShuffleVectorInst(packed_tmp23_reg2mem_0, packed_tmp1, const_packed_30, "tmp378", label_cond__28); - ExtractElementInst* float_tmp399 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp399", label_cond__28); - CastInst* double_conv4010 = new FPExtInst(float_tmp399, Type::DoubleTy, "conv4010", label_cond__28); - FCmpInst* int1_cmp4111 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv4010, const_double_27, "cmp4111", label_cond__28); - new BranchInst(label_cond__42, label_cond_cont48, int1_cmp4111, label_cond__28); - - // Block cond.cont34 (label_cond_cont34) - PHINode* packed_tmp23_reg2mem_1 = new PHINode(VectorTy_4, "tmp23.reg2mem.1", label_cond_cont34); - packed_tmp23_reg2mem_1->reserveOperandSpace(2); - packed_tmp23_reg2mem_1->addIncoming(packed_tmp233, label_cond__14); - packed_tmp23_reg2mem_1->addIncoming(packed_tmp23, label_cond_cont20); - - ShuffleVectorInst* packed_tmp37 = new ShuffleVectorInst(packed_tmp23_reg2mem_1, packed_tmp2, const_packed_30, "tmp37", label_cond_cont34); - ExtractElementInst* float_tmp39 = new ExtractElementInst(packed_tmp0, const_int32_24, "tmp39", label_cond_cont34); - CastInst* double_conv40 = new FPExtInst(float_tmp39, Type::DoubleTy, "conv40", label_cond_cont34); - FCmpInst* int1_cmp41 = new FCmpInst(FCmpInst::FCMP_OLT, double_conv40, const_double_27, "cmp41", label_cond_cont34); - new BranchInst(label_cond__42, label_cond_cont48, int1_cmp41, label_cond_cont34); - - // Block cond.?42 (label_cond__42) - PHINode* packed_tmp37_reg2mem_0 = new PHINode(VectorTy_4, "tmp37.reg2mem.0", label_cond__42); - packed_tmp37_reg2mem_0->reserveOperandSpace(2); - packed_tmp37_reg2mem_0->addIncoming(packed_tmp378, label_cond__28); - packed_tmp37_reg2mem_0->addIncoming(packed_tmp37, label_cond_cont34); - - ShuffleVectorInst* packed_tmp5113 = new ShuffleVectorInst(packed_tmp37_reg2mem_0, packed_tmp1, const_packed_32, "tmp5113", label_cond__42); - new ReturnInst(packed_tmp5113, label_cond__42); - - // Block cond.cont48 (label_cond_cont48) - PHINode* packed_tmp37_reg2mem_1 = new PHINode(VectorTy_4, "tmp37.reg2mem.1", label_cond_cont48); - packed_tmp37_reg2mem_1->reserveOperandSpace(2); - packed_tmp37_reg2mem_1->addIncoming(packed_tmp378, label_cond__28); - packed_tmp37_reg2mem_1->addIncoming(packed_tmp37, label_cond_cont34); - - ShuffleVectorInst* packed_tmp51 = new ShuffleVectorInst(packed_tmp37_reg2mem_1, packed_tmp2, const_packed_32, "tmp51", label_cond_cont48); - new ReturnInst(packed_tmp51, label_cond_cont48); - -} - -// Function: vcos (func_vcos) -{ - Function::arg_iterator args = func_vcos->arg_begin(); - Value* packed_val = args++; - packed_val->setName("val"); - - BasicBlock* label_entry_53 = new BasicBlock("entry",func_vcos,0); - - // Block entry (label_entry_53) - ExtractElementInst* float_tmp1 = new ExtractElementInst(packed_val, const_int32_19, "tmp1", label_entry_53); - CastInst* double_conv_54 = new FPExtInst(float_tmp1, Type::DoubleTy, "conv", label_entry_53); - ExtractElementInst* float_tmp3_55 = new ExtractElementInst(packed_val, const_int32_23, "tmp3", label_entry_53); - CastInst* double_conv4 = new FPExtInst(float_tmp3_55, Type::DoubleTy, "conv4", label_entry_53); - ExtractElementInst* float_tmp6_56 = new ExtractElementInst(packed_val, const_int32_25, "tmp6", label_entry_53); - CastInst* double_conv7 = new FPExtInst(float_tmp6_56, Type::DoubleTy, "conv7", label_entry_53); - ExtractElementInst* float_tmp9 = new ExtractElementInst(packed_val, const_int32_24, "tmp9", label_entry_53); - CastInst* double_conv10 = new FPExtInst(float_tmp9, Type::DoubleTy, "conv10", label_entry_53); - std::vector int32_call_params; - int32_call_params.push_back(const_ptr_34); - int32_call_params.push_back(double_conv_54); - int32_call_params.push_back(double_conv4); - int32_call_params.push_back(double_conv7); - int32_call_params.push_back(double_conv10); - CallInst* int32_call = new CallInst(func_printf, int32_call_params.begin(), int32_call_params.end(), "call", label_entry_53); - int32_call->setCallingConv(CallingConv::C); - int32_call->setTailCall(true);const ParamAttrsList *int32_call_PAL = 0; - int32_call->setParamAttrs(int32_call_PAL); - - CallInst* float_call13 = new CallInst(func_cosf, float_tmp1, "call13", label_entry_53); - float_call13->setCallingConv(CallingConv::C); - float_call13->setTailCall(true);const ParamAttrsList *float_call13_PAL = 0; - float_call13->setParamAttrs(float_call13_PAL); - - InsertElementInst* packed_tmp15 = new InsertElementInst(const_packed_35, float_call13, const_int32_19, "tmp15", label_entry_53); - CallInst* float_call18 = new CallInst(func_cosf, float_tmp1, "call18", label_entry_53); - float_call18->setCallingConv(CallingConv::C); - float_call18->setTailCall(true);const ParamAttrsList *float_call18_PAL = 0; - float_call18->setParamAttrs(float_call18_PAL); - - InsertElementInst* packed_tmp20 = new InsertElementInst(packed_tmp15, float_call18, const_int32_23, "tmp20", label_entry_53); - CallInst* float_call23 = new CallInst(func_cosf, float_tmp1, "call23", label_entry_53); - float_call23->setCallingConv(CallingConv::C); - float_call23->setTailCall(true);const ParamAttrsList *float_call23_PAL = 0; - float_call23->setParamAttrs(float_call23_PAL); - - InsertElementInst* packed_tmp25 = new InsertElementInst(packed_tmp20, float_call23, const_int32_25, "tmp25", label_entry_53); - CallInst* float_call28 = new CallInst(func_cosf, float_tmp1, "call28", label_entry_53); - float_call28->setCallingConv(CallingConv::C); - float_call28->setTailCall(true);const ParamAttrsList *float_call28_PAL = 0; - float_call28->setParamAttrs(float_call28_PAL); - - InsertElementInst* packed_tmp30 = new InsertElementInst(packed_tmp25, float_call28, const_int32_24, "tmp30", label_entry_53); - CastInst* double_conv33 = new FPExtInst(float_call13, Type::DoubleTy, "conv33", label_entry_53); - CastInst* double_conv36 = new FPExtInst(float_call18, Type::DoubleTy, "conv36", label_entry_53); - CastInst* double_conv39 = new FPExtInst(float_call23, Type::DoubleTy, "conv39", label_entry_53); - CastInst* double_conv42 = new FPExtInst(float_call28, Type::DoubleTy, "conv42", label_entry_53); - std::vector int32_call43_params; - int32_call43_params.push_back(const_ptr_36); - int32_call43_params.push_back(double_conv33); - int32_call43_params.push_back(double_conv36); - int32_call43_params.push_back(double_conv39); - int32_call43_params.push_back(double_conv42); - CallInst* int32_call43 = new CallInst(func_printf, int32_call43_params.begin(), int32_call43_params.end(), "call43", label_entry_53); - int32_call43->setCallingConv(CallingConv::C); - int32_call43->setTailCall(true);const ParamAttrsList *int32_call43_PAL = 0; - int32_call43->setParamAttrs(int32_call43_PAL); - - new ReturnInst(packed_tmp30, label_entry_53); - -} - -// Function: scs (func_scs) -{ - Function::arg_iterator args = func_scs->arg_begin(); - Value* packed_val_58 = args++; - packed_val_58->setName("val"); - - BasicBlock* label_entry_59 = new BasicBlock("entry",func_scs,0); - - // Block entry (label_entry_59) - ExtractElementInst* float_tmp2 = new ExtractElementInst(packed_val_58, const_int32_19, "tmp2", label_entry_59); - CallInst* float_call_60 = new CallInst(func_cosf, float_tmp2, "call", label_entry_59); - float_call_60->setCallingConv(CallingConv::C); - float_call_60->setTailCall(true);const ParamAttrsList *float_call_60_PAL = 0; - float_call_60->setParamAttrs(float_call_60_PAL); - - InsertElementInst* packed_tmp5 = new InsertElementInst(const_packed_35, float_call_60, const_int32_19, "tmp5", label_entry_59); - CallInst* float_call7 = new CallInst(func_sinf, float_tmp2, "call7", label_entry_59); - float_call7->setCallingConv(CallingConv::C); - float_call7->setTailCall(true);const ParamAttrsList *float_call7_PAL = 0; - float_call7->setParamAttrs(float_call7_PAL); - - InsertElementInst* packed_tmp9 = new InsertElementInst(packed_tmp5, float_call7, const_int32_23, "tmp9", label_entry_59); - new ReturnInst(packed_tmp9, label_entry_59); - -} - -// Function: vsin (func_vsin) -{ - Function::arg_iterator args = func_vsin->arg_begin(); - Value* packed_val_62 = args++; - packed_val_62->setName("val"); - - BasicBlock* label_entry_63 = new BasicBlock("entry",func_vsin,0); - - // Block entry (label_entry_63) - ExtractElementInst* float_tmp2_64 = new ExtractElementInst(packed_val_62, const_int32_19, "tmp2", label_entry_63); - CallInst* float_call_65 = new CallInst(func_sinf, float_tmp2_64, "call", label_entry_63); - float_call_65->setCallingConv(CallingConv::C); - float_call_65->setTailCall(true);const ParamAttrsList *float_call_65_PAL = 0; - float_call_65->setParamAttrs(float_call_65_PAL); - - InsertElementInst* packed_tmp6 = new InsertElementInst(const_packed_35, float_call_65, const_int32_19, "tmp6", label_entry_63); - InsertElementInst* packed_tmp9_66 = new InsertElementInst(packed_tmp6, float_call_65, const_int32_23, "tmp9", label_entry_63); - InsertElementInst* packed_tmp12 = new InsertElementInst(packed_tmp9_66, float_call_65, const_int32_25, "tmp12", label_entry_63); - InsertElementInst* packed_tmp15_67 = new InsertElementInst(packed_tmp12, float_call_65, const_int32_24, "tmp15", label_entry_63); - new ReturnInst(packed_tmp15_67, label_entry_63); - -} - -// Function: kilp (func_kilp) -{ - Function::arg_iterator args = func_kilp->arg_begin(); - Value* packed_val_69 = args++; - packed_val_69->setName("val"); - - BasicBlock* label_entry_70 = new BasicBlock("entry",func_kilp,0); - BasicBlock* label_lor_rhs = new BasicBlock("lor_rhs",func_kilp,0); - BasicBlock* label_lor_rhs5 = new BasicBlock("lor_rhs5",func_kilp,0); - BasicBlock* label_lor_rhs11 = new BasicBlock("lor_rhs11",func_kilp,0); - BasicBlock* label_UnifiedReturnBlock_71 = new BasicBlock("UnifiedReturnBlock",func_kilp,0); - - // Block entry (label_entry_70) - ExtractElementInst* float_tmp1_72 = new ExtractElementInst(packed_val_69, const_int32_19, "tmp1", label_entry_70); - FCmpInst* int1_cmp_73 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp1_72, const_float_18, "cmp", label_entry_70); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs, int1_cmp_73, label_entry_70); - - // Block lor_rhs (label_lor_rhs) - ExtractElementInst* float_tmp3_75 = new ExtractElementInst(packed_val_69, const_int32_23, "tmp3", label_lor_rhs); - FCmpInst* int1_cmp4 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp3_75, const_float_18, "cmp4", label_lor_rhs); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs5, int1_cmp4, label_lor_rhs); - - // Block lor_rhs5 (label_lor_rhs5) - ExtractElementInst* float_tmp7 = new ExtractElementInst(packed_val_69, const_int32_25, "tmp7", label_lor_rhs5); - FCmpInst* int1_cmp8 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp7, const_float_18, "cmp8", label_lor_rhs5); - new BranchInst(label_UnifiedReturnBlock_71, label_lor_rhs11, int1_cmp8, label_lor_rhs5); - - // Block lor_rhs11 (label_lor_rhs11) - ExtractElementInst* float_tmp13 = new ExtractElementInst(packed_val_69, const_int32_24, "tmp13", label_lor_rhs11); - FCmpInst* int1_cmp14 = new FCmpInst(FCmpInst::FCMP_OLT, float_tmp13, const_float_18, "cmp14", label_lor_rhs11); - CastInst* int32_retval = new ZExtInst(int1_cmp14, IntegerType::get(32), "retval", label_lor_rhs11); - new ReturnInst(int32_retval, label_lor_rhs11); - - // Block UnifiedReturnBlock (label_UnifiedReturnBlock_71) - new ReturnInst(const_int32_23, label_UnifiedReturnBlock_71); - -} - -return mod; - -} +static const unsigned char llvm_builtins_data[] = { +0x42,0x43,0xc0,0xde,0x21,0x0c,0x00,0x00,0x2b,0x02,0x00,0x00,0x01,0x10,0x00,0x00, +0x10,0x00,0x00,0x00,0x07,0x81,0x23,0x91,0x41,0xc8,0x04,0x49,0x06,0x10,0x32,0x39, +0x92,0x01,0x84,0x0c,0x25,0x05,0x08,0x19,0x1e,0x04,0x8b,0x62,0x80,0x14,0x45,0x02, +0x42,0x92,0x0b,0x42,0xa4,0x10,0x32,0x14,0x38,0x08,0x18,0x49,0x0a,0x32,0x44,0x24, +0x48,0x0a,0x90,0x21,0x23,0x44,0x72,0x80,0x8c,0x14,0x21,0x86,0x0a,0x8a,0x0a,0x64, +0x0c,0x1f,0x00,0x00,0x49,0x18,0x00,0x00,0x02,0x00,0x00,0x00,0x0b,0x04,0x00,0x0c, +0x00,0x00,0x00,0x00,0x51,0x20,0x00,0x00,0x13,0x00,0x00,0x00,0x32,0x22,0x48,0x09, +0x20,0x65,0x82,0x84,0x00,0x26,0x45,0x48,0x05,0x09,0x26,0x45,0xc6,0x05,0x42,0x52, +0x26,0x08,0xb0,0x19,0x80,0x61,0x04,0x02,0x98,0x23,0x00,0x83,0x21,0x80,0x39,0x82, +0x60,0x0a,0x80,0x2e,0xd5,0x61,0x04,0x42,0x20,0x49,0x90,0x22,0x4d,0xa2,0x73,0x04, +0x08,0xb9,0x32,0x00,0x00,0x8a,0x10,0xc2,0x65,0xb8,0x42,0x84,0x10,0x42,0x0d,0x44, +0x11,0x00,0x18,0x01,0x28,0x82,0x08,0x00,0x13,0xa2,0x74,0xb0,0x03,0x3c,0xb0,0x83, +0x36,0x80,0x87,0x71,0x68,0x03,0x76,0x48,0x07,0x77,0xa8,0x07,0x7c,0x68,0x83,0x73, +0x70,0x87,0x7a,0xd8,0x70,0x0f,0xe5,0xd0,0x06,0xf0,0xa0,0x07,0x73,0x20,0x07,0x7a, +0x30,0x07,0x72,0xa0,0x07,0x73,0x20,0x07,0x6d,0x90,0x0e,0x71,0xa0,0x07,0x78,0xa0, +0x07,0x78,0xd0,0x06,0xe9,0x80,0x07,0x7a,0x80,0x07,0x7a,0x80,0x07,0x6d,0x90,0x0e, +0x71,0x60,0x07,0x7a,0x10,0x07,0x76,0xa0,0x07,0x71,0x60,0x07,0x6d,0x90,0x0e,0x73, +0x20,0x07,0x7a,0x30,0x07,0x72,0xa0,0x07,0x73,0x20,0x07,0x6d,0x90,0x0e,0x76,0x40, +0x07,0x7a,0x30,0x07,0x72,0xa0,0x07,0x76,0x40,0x07,0x6d,0x60,0x0e,0x73,0x20,0x07, +0x7a,0x30,0x07,0x72,0xa0,0x07,0x73,0x20,0x07,0x6d,0x60,0x0e,0x76,0x40,0x07,0x7a, +0x30,0x07,0x72,0xa0,0x07,0x76,0x40,0x07,0x6d,0x60,0x0f,0x76,0x40,0x07,0x7a,0x60, +0x07,0x74,0xa0,0x07,0x76,0x40,0x07,0x6d,0x60,0x0f,0x71,0x20,0x07,0x78,0xa0,0x07, +0x71,0x20,0x07,0x78,0xa0,0x07,0x71,0x20,0x07,0x78,0xd0,0x06,0xe1,0x00,0x07,0x7a, +0x00,0x07,0x7a,0x60,0x07,0x74,0xd0,0x06,0xe6,0x80,0x07,0x70,0xa0,0x07,0x71,0x20, +0x07,0x78,0xa0,0x07,0x71,0x20,0x07,0x78,0xa0,0xf3,0x40,0x88,0x04,0x32,0x32,0x02, +0x04,0x60,0x76,0xc6,0xfc,0x6c,0x48,0xa2,0x00,0x40,0x00,0x00,0x00,0x00,0x0c,0x49, +0x14,0x20,0x00,0x00,0x00,0x00,0x80,0x21,0xc9,0x02,0x00,0x01,0x00,0x00,0x00,0x30, +0x24,0x61,0x00,0x20,0x08,0x00,0x00,0x00,0x86,0x24,0x0b,0x00,0x04,0x00,0x00,0x00, +0xc0,0x90,0xa4,0x01,0x02,0x00,0x00,0x00,0x00,0x18,0x92,0x1c,0x40,0x00,0x00,0x00, +0x00,0x00,0x43,0x92,0x05,0x00,0x02,0x00,0x00,0x00,0x60,0x48,0x72,0x00,0x01,0x00, +0x00,0x00,0x00,0x0c,0x49,0x16,0x00,0x08,0x00,0x00,0x00,0x80,0x21,0x89,0x01,0x00, +0x41,0x00,0x00,0x00,0x90,0x05,0x02,0x00,0x10,0x00,0x00,0x00,0x32,0x1e,0x98,0x10, +0x19,0x11,0x4c,0x90,0x8c,0x09,0x26,0x47,0xc6,0x04,0x43,0x92,0x8a,0x59,0x8b,0x43, +0x50,0xd2,0x09,0x02,0x81,0xd2,0x73,0x50,0xc9,0x0c,0x2a,0x99,0x41,0x25,0x33,0xa8, +0x64,0x56,0x28,0x66,0x2d,0x0e,0x41,0xcf,0x2a,0x15,0x04,0x4a,0xcf,0x41,0x25,0x33, +0xa8,0x64,0x06,0x95,0xcc,0xa0,0x92,0x59,0x01,0x00,0x00,0x00,0x53,0x82,0x26,0x0c, +0x04,0x00,0x00,0x00,0x22,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00, +0x05,0x00,0x00,0x00,0x04,0xc6,0x08,0x40,0x10,0x04,0xe1,0x70,0x18,0x23,0x00,0x41, +0x10,0x84,0xc3,0x60,0x04,0x00,0x00,0x00,0x93,0x0c,0xce,0x43,0x4c,0x31,0x3c,0x8e, +0x34,0xc9,0x30,0x41,0xc2,0x14,0x03,0x34,0x51,0x93,0x0c,0x4d,0x44,0x4c,0x31,0x44, +0x8d,0x35,0x56,0x01,0x04,0xc3,0x55,0x21,0x16,0x0e,0x04,0x00,0x0f,0x00,0x00,0x00, +0x46,0x41,0x08,0xcc,0x73,0x9b,0x05,0x21,0x30,0xcf,0x6e,0x18,0x84,0x00,0x2c,0x8b, +0x35,0x04,0x80,0x39,0x04,0x81,0x5d,0x20,0x80,0x0f,0x0c,0x43,0xe4,0xd3,0x36,0x81, +0x04,0x3e,0x30,0x0c,0x91,0x4f,0x5b,0x05,0x12,0xf8,0xc0,0x30,0x44,0x7e,0x7d,0x00, +0x05,0xd1,0x4c,0x11,0x66,0x12,0x83,0xc0,0x3c,0x00,0x00,0x00,0x00,0x00,0x00,0x00, +0x61,0x20,0x00,0x00,0x2a,0x00,0x00,0x00,0x13,0x04,0x43,0x2c,0x10,0x00,0x00,0x00, +0x08,0x00,0x00,0x00,0x24,0x8a,0xa0,0x0c,0x46,0x00,0x4a,0x80,0xc2,0x1c,0x84,0x55, +0x55,0xd6,0x1c,0x84,0x45,0x51,0x16,0x81,0x19,0x80,0x11,0x80,0x31,0x02,0x10,0x04, +0x41,0xfc,0x03,0x00,0x63,0x08,0x0d,0x34,0xc9,0x70,0x55,0xc2,0x2c,0x43,0x20,0x60, +0x73,0x0c,0xd3,0x15,0x8d,0x21,0x34,0xd1,0x18,0x42,0xf3,0x8c,0x55,0x00,0x81,0xa0, +0x6d,0x73,0x0c,0x19,0xe7,0x60,0x87,0x52,0x38,0x10,0x00,0x00,0x13,0x00,0x00,0x00, +0x17,0x60,0x20,0xc5,0x74,0x10,0x8d,0x65,0x14,0x13,0xf3,0xd4,0xb4,0x6d,0x14,0x13, +0xf3,0xd4,0xb8,0x69,0x14,0x13,0xf3,0xd4,0xb6,0x75,0x14,0x13,0xf3,0xd4,0xba,0x35, +0x0c,0x13,0xf3,0x9c,0x80,0xe4,0x36,0x48,0x81,0x10,0xc3,0x4a,0x4c,0x54,0xd4,0x6c, +0x8b,0x23,0x28,0x76,0x41,0x4c,0xcc,0xa3,0x1b,0x07,0x21,0x00,0xcb,0x72,0x00,0x05, +0xd1,0x4c,0x11,0x66,0x18,0x83,0xc0,0x3c,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, +0x61,0x20,0x00,0x00,0x82,0x00,0x00,0x00,0x13,0x04,0x47,0x2c,0x10,0x00,0x00,0x00, +0x08,0x00,0x00,0x00,0x24,0x46,0x00,0x8a,0xa0,0x0c,0x4a,0xa0,0x14,0x8a,0xa1,0x1c, +0x68,0x8c,0x00,0x10,0x9a,0x83,0x80,0xa8,0x48,0x9a,0x83,0x80,0xa6,0x4a,0x9a,0x83, +0x80,0xa6,0xc8,0x02,0x63,0x08,0x0d,0x64,0xdb,0xc0,0x49,0x06,0xee,0x22,0xc6,0x10, +0x9a,0xc9,0xbc,0x81,0x93,0x0c,0xdf,0x45,0x4c,0x31,0x38,0x4f,0x37,0xcb,0x10,0x08, +0x60,0x30,0xc8,0x10,0x06,0x0e,0x36,0x86,0xd0,0x44,0x36,0x06,0x03,0x27,0x19,0xc8, +0xe0,0x22,0x66,0x19,0x06,0xa2,0x0c,0x06,0x19,0xc2,0xe0,0xc1,0xc6,0x10,0x9a,0xc8, +0xce,0x60,0xe0,0x24,0x03,0x1a,0x5c,0xc4,0x2c,0xc3,0x40,0xa4,0xc1,0x40,0x45,0x20, +0x06,0x81,0x19,0x08,0x83,0x0c,0x6a,0xe0,0x64,0x63,0x08,0x8d,0x64,0x6c,0x30,0x70, +0x92,0xa1,0x0d,0x2e,0x62,0x96,0xa1,0x30,0xdc,0x60,0xa0,0x22,0x10,0x83,0xc0,0x0c, +0x84,0x41,0x86,0x37,0x78,0xb2,0x31,0x84,0x46,0xb2,0x38,0x18,0x38,0xc9,0x20,0x07, +0x17,0x31,0xcb,0x50,0x18,0x73,0x30,0x50,0x11,0xac,0xc1,0x00,0x07,0xc4,0x20,0x03, +0x1d,0x38,0x1a,0xd6,0xc1,0x40,0x45,0xb0,0x06,0x03,0x1c,0x10,0x83,0x0c,0x76,0xf0, +0x68,0x78,0x07,0xe1,0x40,0x00,0x00,0x00,0x4c,0x00,0x00,0x00,0x56,0x62,0x08,0xcc, +0x63,0xef,0x3a,0xa9,0x00,0x19,0x7b,0x73,0x23,0x73,0xf9,0xa1,0x91,0x31,0x98,0x62, +0x62,0x9e,0x7b,0xb7,0x06,0x62,0x62,0x1e,0xda,0x1c,0x88,0x89,0x79,0x6a,0x7b,0x20, +0x26,0xe6,0xb1,0x6d,0x83,0x98,0x98,0xe7,0x36,0x92,0x43,0x70,0x9a,0xca,0xd6,0x73, +0xa3,0x79,0x26,0xe6,0xb9,0x77,0x3f,0x22,0x0c,0x9b,0x21,0x18,0x9f,0xb6,0x90,0x64, +0x62,0x9e,0xda,0x9f,0x98,0xc7,0x36,0x9b,0x67,0x62,0x9e,0x7b,0xf7,0x23,0xc2,0xb0, +0x19,0x82,0xf1,0x6b,0x53,0x79,0x26,0xe6,0xb1,0x6f,0x3f,0x22,0x0c,0x9b,0x21,0x18, +0x9f,0xb6,0x98,0x62,0x62,0x9e,0xbb,0xb7,0x97,0x67,0x62,0x1e,0xfb,0xf6,0x23,0xc2, +0xb0,0x19,0x82,0xf1,0x6b,0x43,0x31,0x04,0xa7,0xa9,0x6c,0xdd,0x66,0x0a,0x81,0x79, +0xf0,0xfa,0x08,0x16,0xc1,0x69,0x06,0x5f,0x70,0x9a,0xe9,0xc6,0x49,0x01,0xc8,0xd8, +0x9b,0x1b,0x99,0xcb,0x4f,0x0c,0x8d,0xad,0x18,0x13,0xf3,0xdc,0x3b,0x6f,0x35,0xc7, +0xc4,0x3c,0x79,0x5d,0xdf,0x06,0x52,0x08,0xcc,0x53,0xdf,0x26,0x62,0x4c,0xcc,0x63, +0xdf,0xb7,0xb9,0x1c,0x02,0xf3,0xe0,0x75,0x5d,0x5b,0xc7,0x20,0x30,0x8f,0x79,0x14, +0x13,0xf3,0xd4,0xf5,0x19,0x2c,0x82,0xd3,0x0c,0xbe,0xe0,0x34,0x13,0xce,0x5b,0x0b, +0x22,0x38,0x4d,0x85,0xd3,0x35,0x6d,0x37,0xc5,0xc4,0x3c,0x79,0x4d,0x1a,0x40,0xc6, +0xde,0xdc,0xc8,0x5c,0x7e,0x64,0x70,0x8c,0x83,0x10,0x9c,0xa6,0xb2,0x94,0x42,0x60, +0x1e,0x7b,0x37,0x19,0x43,0x70,0x9a,0x0a,0xa7,0xcd,0xa4,0x98,0x98,0xc7,0xbe,0x8d, +0xc5,0x98,0x98,0xe7,0xee,0x7b,0x3b,0x29,0x26,0xe6,0xb1,0xf3,0x13,0x58,0x04,0xa7, +0x19,0x7c,0xc1,0x69,0x26,0x9b,0xb6,0x0f,0x43,0x70,0x9a,0xaa,0xb6,0x6d,0xc4,0x98, +0x98,0xc7,0xce,0xf1,0x03,0x28,0x88,0x66,0x8a,0x30,0x00,0x00,0x00,0x00,0x00,0x00, +0x61,0x20,0x00,0x00,0x4a,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00, +0x07,0x00,0x00,0x00,0x24,0xca,0x60,0x04,0xa0,0x04,0x8a,0x80,0xc2,0x0c,0x00,0xbd, +0x61,0x8c,0x04,0x10,0x1e,0xe1,0x19,0xc6,0x48,0x02,0xe1,0x11,0x1e,0x00,0x00,0x00, +0x63,0x08,0xcd,0x63,0xd5,0xc0,0x31,0x84,0x06,0xb2,0x6b,0xe0,0x18,0x42,0x13,0x59, +0x36,0x70,0x0c,0xa1,0x71,0x6c,0x1b,0x38,0x16,0x02,0x04,0xc7,0x64,0x61,0x1a,0x37, +0x16,0x01,0x04,0x48,0x35,0xc7,0x20,0x79,0xcf,0x58,0x04,0x10,0x20,0xd5,0x1c,0xc3, +0x07,0x06,0xd0,0x58,0x04,0x10,0x20,0xd5,0x1c,0x43,0x18,0x88,0x41,0x34,0x16,0x01, +0x04,0x48,0x35,0xc7,0x30,0x06,0x64,0xe0,0x98,0x37,0xd0,0xc0,0x60,0xa0,0x89,0xc1, +0x40,0x23,0x83,0x81,0x63,0x21,0x40,0x70,0x50,0x66,0x70,0x06,0x68,0x90,0x06,0x58, +0x06,0xe1,0x40,0x00,0x25,0x00,0x00,0x00,0x56,0x52,0x4c,0xcc,0x73,0xd3,0x56,0x41, +0x4c,0xcc,0x53,0xdb,0x05,0x31,0x31,0xcf,0x6d,0x19,0xc4,0xc4,0x3c,0xba,0x6d,0x10, +0x13,0xf3,0xf4,0xd6,0x41,0x08,0xc0,0xb2,0x18,0x46,0x21,0x38,0x4d,0x85,0x9b,0x46, +0x21,0x38,0x4d,0xb5,0x9b,0x8a,0x21,0x00,0xcb,0x82,0xdf,0x66,0x62,0x08,0x4e,0x53, +0xdd,0xb7,0x9d,0x18,0x82,0xd3,0x54,0xb7,0x6e,0x28,0x86,0xe0,0x34,0xd5,0xdd,0xdb, +0x47,0x31,0x31,0x4f,0x9d,0x9b,0x87,0x21,0x00,0xcb,0x52,0xdf,0x06,0x62,0x08,0xc0, +0xb2,0xd4,0xbc,0x59,0x10,0x82,0xd3,0x54,0x96,0x62,0x08,0x4e,0x53,0xe1,0xb6,0x85, +0x14,0x13,0xf3,0xd8,0xb4,0x8d,0x14,0x13,0xf3,0xd8,0xb9,0x89,0x18,0x02,0xb0,0x2c, +0xf6,0x6d,0x24,0x86,0x00,0x2c,0x8b,0xcd,0x1b,0x87,0x21,0x38,0x4d,0x55,0xd3,0xd6, +0x30,0x54,0xc0,0x72,0x00,0x05,0xd1,0x4c,0x11,0x06,0x00,0x00,0x00,0x00,0x00,0x00, +0x61,0x20,0x00,0x00,0x19,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00, +0x03,0x00,0x00,0x00,0x24,0x4a,0x60,0x04,0x80,0xc2,0x0c,0x00,0x00,0x00,0x00,0x00, +0x63,0x08,0xcd,0x33,0x16,0x01,0x04,0x48,0x34,0xc7,0x00,0x49,0xcf,0x58,0x04,0x10, +0x28,0xd1,0x1c,0xc3,0x44,0x39,0x58,0x85,0x03,0x01,0x00,0x00,0x0a,0x00,0x00,0x00, +0x16,0x41,0x4c,0xcc,0x63,0xdb,0x04,0x31,0x31,0x4f,0x6e,0x0d,0x43,0x05,0x2c,0x07, +0x50,0x10,0xcd,0x14,0x61,0x56,0x41,0x4c,0xcc,0xd3,0x1b,0x45,0x21,0x00,0xcb,0xb2, +0x9b,0x04,0x21,0x00,0xcb,0x02,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00, +0x1b,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00,0x03,0x00,0x00,0x00, +0x24,0xca,0x60,0x04,0xa0,0x04,0x8a,0x80,0xc2,0x0c,0x00,0x00,0x63,0x08,0xcd,0x33, +0x16,0x01,0x04,0xca,0x34,0xc7,0x20,0x51,0xcf,0x1c,0x43,0x45,0x41,0x73,0x0c,0x16, +0x15,0xcd,0x31,0x5c,0x94,0x83,0x58,0x38,0x10,0x00,0x00,0x00,0x0b,0x00,0x00,0x00, +0x86,0x51,0x4c,0xcc,0x53,0xe7,0x76,0x51,0x4c,0xcc,0x53,0xdb,0x36,0x41,0x4c,0xcc, +0x63,0x5b,0x05,0x31,0x31,0x8f,0x6e,0x0d,0x43,0x05,0x2c,0x66,0x41,0x4c,0xcc,0xd3, +0x1f,0x40,0x41,0x34,0x53,0x84,0x19,0x05,0x21,0x00,0xcb,0x02,0x00,0x00,0x00,0x00, +0x61,0x20,0x00,0x00,0x2f,0x00,0x00,0x00,0x13,0x04,0x45,0x2c,0x10,0x00,0x00,0x00, +0x03,0x00,0x00,0x00,0x24,0xca,0xa0,0x04,0x46,0x00,0x8a,0x80,0xc0,0x08,0x00,0x00, +0x63,0x08,0x0d,0x34,0xc9,0x30,0x49,0xc4,0x2c,0x03,0x11,0x50,0x63,0x08,0xcd,0x33, +0xc9,0x50,0x49,0xc4,0x2c,0x03,0x21,0x58,0x63,0x08,0x4d,0x34,0xc9,0x70,0x49,0xc4, +0x2c,0x03,0x31,0x60,0x63,0x08,0x8d,0x33,0xc9,0x90,0x49,0x84,0x69,0x22,0x70,0xc3, +0x27,0x1c,0x08,0x00,0x1a,0x00,0x00,0x00,0x96,0x51,0x4c,0xcc,0x53,0xdf,0x66,0x41, +0x08,0xcc,0x83,0xdb,0x04,0x31,0x31,0x4f,0x6d,0x15,0xc4,0xc4,0x3c,0xb7,0x61,0x10, +0x02,0xf3,0xf0,0x47,0x20,0xb9,0x0d,0x52,0x20,0xc4,0xb0,0x12,0x13,0x15,0x35,0xdb, +0xe2,0x08,0x8a,0x5d,0x10,0x13,0xf3,0xec,0x37,0x90,0x2c,0x4e,0xf4,0x47,0x87,0x54, +0xd7,0x17,0x70,0x2c,0x4e,0xf4,0x47,0x87,0x74,0x02,0xc8,0xe2,0x44,0x7f,0x74,0x48, +0xb9,0x69,0x14,0x02,0xf3,0xd4,0xb8,0x6d,0x18,0x11,0x31,0x55,0xc0,0x62,0x0d,0x43, +0x05,0x2c,0x07,0x50,0x10,0xcd,0x14,0x61,0x46,0x31,0x08,0xcc,0x03,0x00,0x00,0x00, +0x00,0x00,0x00,0x00,0x71,0x20,0x00,0x00,0x12,0x00,0x00,0x00,0x66,0x40,0x54,0x82, +0x23,0x59,0xc2,0x20,0x09,0x92,0x1d,0x18,0x4f,0x84,0x34,0x53,0x61,0x03,0xc4,0xe3, +0x58,0x85,0x05,0x14,0xbe,0x34,0x45,0xb5,0x21,0x10,0x82,0x23,0x15,0x46,0x30,0x2c, +0xc8,0x64,0x02,0x06,0xf0,0x3c,0x91,0x73,0x19,0x00,0xe1,0x4b,0x53,0x64,0x0a,0x84, +0x84,0x34,0x85,0x31,0x10,0x0a,0xb2,0x3c,0x56,0x30,0x08,0xcc,0x63,0x0b,0x44,0x25, +0x21,0x0d,0x00,0x00,0x00,0x00,0x00,0x00}; diff --git a/src/gallium/auxiliary/gallivm/instructions.cpp b/src/gallium/auxiliary/gallivm/instructions.cpp index 8919491792..b35d6790f7 100644 --- a/src/gallium/auxiliary/gallivm/instructions.cpp +++ b/src/gallium/auxiliary/gallivm/instructions.cpp @@ -35,6 +35,8 @@ #include "storage.h" +#include "pipe/p_util.h" + #include #include #include @@ -42,7 +44,8 @@ #include #include #include -#include +#include +#include #include #include @@ -53,7 +56,6 @@ using namespace llvm; #include "gallivm_builtins.cpp" #if 0 - llvm::Value *arrayFromChannels(std::vector &vals) { VectorType *vectorType = VectorType::get(Type::FloatTy, 4); @@ -84,7 +86,10 @@ Instructions::Instructions(llvm::Module *mod, llvm::Function *func, llvm::BasicB m_llvmLit = 0; m_fmtPtr = 0; - createGallivmBuiltins(m_mod); + MemoryBuffer *buffer = MemoryBuffer::getMemBuffer( + (const char*)&llvm_builtins_data[0], + (const char*)&llvm_builtins_data[Elements(llvm_builtins_data)-1]); + m_mod = ParseBitcodeFile(buffer); } llvm::Value * Instructions::add(llvm::Value *in1, llvm::Value *in2) diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 6f83b56a72..3c16a3692d 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -1,8 +1,35 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ #include "instructionssoa.h" #include "storagesoa.h" #include "pipe/p_shader_tokens.h" +#include "pipe/p_util.h" #include #include @@ -10,7 +37,10 @@ #include #include #include -#include +#include +#include +#include + #include @@ -183,7 +213,10 @@ llvm::Module * InstructionsSoa::currentModule() const void InstructionsSoa::createBuiltins() { - m_builtins = createSoaBuiltins(); + MemoryBuffer *buffer = MemoryBuffer::getMemBuffer( + (const char*)&soabuiltins_data[0], + (const char*)&soabuiltins_data[Elements(soabuiltins_data)-1]); + m_builtins = ParseBitcodeFile(buffer); createDependencies(); } diff --git a/src/gallium/auxiliary/gallivm/llvm_builtins.c b/src/gallium/auxiliary/gallivm/llvm_builtins.c index 4f98d754ba..64b5d499a8 100644 --- a/src/gallium/auxiliary/gallivm/llvm_builtins.c +++ b/src/gallium/auxiliary/gallivm/llvm_builtins.c @@ -30,7 +30,7 @@ * Authors: * Zack Rusin zack@tungstengraphics.com */ -typedef __attribute__(( ocu_vector_type(4) )) float float4; +typedef __attribute__(( ext_vector_type(4) )) float float4; extern float powf(float a, float b); diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index 4d658be520..40addebd8c 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -31,7 +31,7 @@ * Authors: * Zack Rusin zack@tungstengraphics.com */ -typedef __attribute__(( ocu_vector_type(4) )) float float4; +typedef __attribute__(( ext_vector_type(4) )) float float4; void dp3(float4 *res, float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, -- cgit v1.2.3 From b6c9d2ef2cfadbbe3e7aa94f21fd0da36d089952 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 12:37:41 +0100 Subject: rtasm: add dump facility for x86 (from tgsi_sse2.c) --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 267 ++++++++++++++++++++++++++--- 1 file changed, 243 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 7f8cc23d15..f2c08c96a6 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -34,6 +34,116 @@ #define X86_TWOB 0x0f +#define DUMP_SSE 0 + +#if DUMP_SSE + +static void +_print_reg( + struct x86_reg reg ) +{ + if (reg.mod != mod_REG) + debug_printf( "[" ); + + switch( reg.file ) { + case file_REG32: + switch( reg.idx ) { + case reg_AX: debug_printf( "EAX" ); break; + case reg_CX: debug_printf( "ECX" ); break; + case reg_DX: debug_printf( "EDX" ); break; + case reg_BX: debug_printf( "EBX" ); break; + case reg_SP: debug_printf( "ESP" ); break; + case reg_BP: debug_printf( "EBP" ); break; + case reg_SI: debug_printf( "ESI" ); break; + case reg_DI: debug_printf( "EDI" ); break; + } + break; + case file_MMX: + debug_printf( "MMX%u", reg.idx ); + break; + case file_XMM: + debug_printf( "XMM%u", reg.idx ); + break; + case file_x87: + debug_printf( "fp%u", reg.idx ); + break; + } + + if (reg.mod == mod_DISP8 || + reg.mod == mod_DISP32) + debug_printf("+%d", reg.disp); + + if (reg.mod != mod_REG) + debug_printf( "]" ); +} + +static void +_fill( + const char *op ) +{ + unsigned count = 10 - strlen( op ); + + while( count-- ) { + debug_printf( " " ); + } +} + +#define DUMP_START() debug_printf( "\nsse-dump start ----------------" ) +#define DUMP_END() debug_printf( "\nsse-dump end ----------------\n" ) +#define DUMP( OP ) debug_printf( "\n%s", OP ) + +#define DUMP_I( OP, I ) do { \ + debug_printf( "\n%s", OP ); \ + _fill( OP ); \ + debug_printf( "%u", I ); \ +} while( 0 ) + +#define DUMP_R( OP, R0 ) do { \ + debug_printf( "\n%s", OP ); \ + _fill( OP ); \ + _print_reg( R0 ); \ +} while( 0 ) + +#define DUMP_RR( OP, R0, R1 ) do { \ + debug_printf( "\n%s", OP ); \ + _fill( OP ); \ + _print_reg( R0 ); \ + debug_printf( ", " ); \ + _print_reg( R1 ); \ +} while( 0 ) + +#define DUMP_RI( OP, R0, I ) do { \ + debug_printf( "\n%s", OP ); \ + _fill( OP ); \ + _print_reg( R0 ); \ + debug_printf( ", " ); \ + debug_printf( "%u", I ); \ +} while( 0 ) + +#define DUMP_RRI( OP, R0, R1, I ) do { \ + debug_printf( "\n%s", OP ); \ + _fill( OP ); \ + _print_reg( R0 ); \ + debug_printf( ", " ); \ + _print_reg( R1 ); \ + debug_printf( ", " ); \ + debug_printf( "%u", I ); \ +} while( 0 ) + +#else + +#define DUMP_START() +#define DUMP_END() +#define DUMP( OP ) +#define DUMP_I( OP, I ) +#define DUMP_R( OP, R0 ) +#define DUMP_RR( OP, R0, R1 ) +#define DUMP_RI( OP, R0, I ) +#define DUMP_RRI( OP, R0, R1, I ) + +#endif + + static void do_realloc( struct x86_function *p ) { if (p->size == 0) { @@ -272,6 +382,7 @@ unsigned char *x86_jcc_forward( struct x86_function *p, unsigned char *x86_jmp_forward( struct x86_function *p) { + DUMP( __FUNCTION__ ); emit_1ub(p, 0xe9); emit_1i(p, 0); return x86_get_label(p); @@ -279,6 +390,8 @@ unsigned char *x86_jmp_forward( struct x86_function *p) unsigned char *x86_call_forward( struct x86_function *p) { + DUMP( __FUNCTION__ ); + emit_1ub(p, 0xe8); emit_1i(p, 0); return x86_get_label(p); @@ -294,6 +407,7 @@ void x86_fixup_fwd_jump( struct x86_function *p, void x86_jmp( struct x86_function *p, unsigned char *label) { + DUMP_I( __FUNCTION__, label ); emit_1ub(p, 0xe9); emit_1i(p, pointer_to_intptr( label ) - pointer_to_intptr( x86_get_label(p) ) - 4); } @@ -310,12 +424,14 @@ static unsigned char *cptr( void (*label)() ) */ void x86_call( struct x86_function *p, void (*label)()) { + DUMP_I( __FUNCTION__, label ); emit_1ub(p, 0xe8); emit_1i(p, cptr(label) - x86_get_label(p) - 4); } #else void x86_call( struct x86_function *p, struct x86_reg reg) { + DUMP_R( __FUNCTION__, reg ); emit_1ub(p, 0xff); emit_modrm_noreg(p, 2, reg); } @@ -328,6 +444,7 @@ void x86_call( struct x86_function *p, struct x86_reg reg) */ void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) { + DUMP_RI( __FUNCTION__, dst, imm ); assert(dst.mod == mod_REG); emit_1ub(p, 0xb8 + dst.idx); emit_1i(p, imm); @@ -336,6 +453,7 @@ void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) void x86_push( struct x86_function *p, struct x86_reg reg ) { + DUMP_R( __FUNCTION__, reg ); assert(reg.mod == mod_REG); emit_1ub(p, 0x50 + reg.idx); p->stack_offset += 4; @@ -344,6 +462,7 @@ void x86_push( struct x86_function *p, void x86_pop( struct x86_function *p, struct x86_reg reg ) { + DUMP_R( __FUNCTION__, reg ); assert(reg.mod == mod_REG); emit_1ub(p, 0x58 + reg.idx); p->stack_offset -= 4; @@ -352,6 +471,7 @@ void x86_pop( struct x86_function *p, void x86_inc( struct x86_function *p, struct x86_reg reg ) { + DUMP_R( __FUNCTION__, reg ); assert(reg.mod == mod_REG); emit_1ub(p, 0x40 + reg.idx); } @@ -359,17 +479,20 @@ void x86_inc( struct x86_function *p, void x86_dec( struct x86_function *p, struct x86_reg reg ) { + DUMP_R( __FUNCTION__, reg ); assert(reg.mod == mod_REG); emit_1ub(p, 0x48 + reg.idx); } void x86_ret( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_1ub(p, 0xc3); } void x86_sahf( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_1ub(p, 0x9e); } @@ -377,6 +500,7 @@ void x86_mov( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_op_modrm( p, 0x8b, 0x89, dst, src ); } @@ -384,6 +508,7 @@ void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_op_modrm( p, 0x33, 0x31, dst, src ); } @@ -391,6 +516,7 @@ void x86_cmp( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_op_modrm( p, 0x3b, 0x39, dst, src ); } @@ -398,6 +524,7 @@ void x86_lea( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_1ub(p, 0x8d); emit_modrm( p, dst, src ); } @@ -406,6 +533,7 @@ void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_1ub(p, 0x85); emit_modrm( p, dst, src ); } @@ -414,6 +542,7 @@ void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_op_modrm(p, 0x03, 0x01, dst, src ); } @@ -422,6 +551,7 @@ void x86_add( struct x86_function *p, void x86_mul( struct x86_function *p, struct x86_reg src ) { + DUMP_R( __FUNCTION__, src ); emit_1ub(p, 0xf7); emit_modrm_noreg(p, 4, src ); } @@ -431,6 +561,7 @@ void x86_imul( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0xAF); emit_modrm(p, dst, src); } @@ -440,6 +571,7 @@ void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_op_modrm(p, 0x2b, 0x29, dst, src ); } @@ -447,6 +579,7 @@ void x86_or( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_op_modrm( p, 0x0b, 0x09, dst, src ); } @@ -454,6 +587,7 @@ void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_op_modrm( p, 0x23, 0x21, dst, src ); } @@ -468,6 +602,7 @@ void sse_movss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, 0xF3, X86_TWOB); emit_op_modrm( p, 0x10, 0x11, dst, src ); } @@ -476,6 +611,7 @@ void sse_movaps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x28, 0x29, dst, src ); } @@ -484,6 +620,7 @@ void sse_movups( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x10, 0x11, dst, src ); } @@ -492,6 +629,7 @@ void sse_movhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); assert(dst.mod != mod_REG || src.mod != mod_REG); emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x16, 0x17, dst, src ); /* cf movlhps */ @@ -501,6 +639,7 @@ void sse_movlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); assert(dst.mod != mod_REG || src.mod != mod_REG); emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x12, 0x13, dst, src ); /* cf movhlps */ @@ -510,6 +649,7 @@ void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x5F); emit_modrm( p, dst, src ); } @@ -518,6 +658,7 @@ void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x5F); emit_modrm( p, dst, src ); } @@ -526,6 +667,7 @@ void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x5E); emit_modrm( p, dst, src ); } @@ -534,6 +676,7 @@ void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x5D); emit_modrm( p, dst, src ); } @@ -542,6 +685,7 @@ void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x5C); emit_modrm( p, dst, src ); } @@ -550,6 +694,7 @@ void sse_mulps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x59); emit_modrm( p, dst, src ); } @@ -558,6 +703,7 @@ void sse_mulss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x59); emit_modrm( p, dst, src ); } @@ -566,6 +712,7 @@ void sse_addps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x58); emit_modrm( p, dst, src ); } @@ -574,6 +721,7 @@ void sse_addss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x58); emit_modrm( p, dst, src ); } @@ -582,6 +730,7 @@ void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x55); emit_modrm( p, dst, src ); } @@ -590,6 +739,7 @@ void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x54); emit_modrm( p, dst, src ); } @@ -598,6 +748,7 @@ void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x52); emit_modrm( p, dst, src ); } @@ -606,6 +757,7 @@ void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x52); emit_modrm( p, dst, src ); @@ -615,6 +767,7 @@ void sse_movhlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); assert(dst.mod == mod_REG && src.mod == mod_REG); emit_2ub(p, X86_TWOB, 0x12); emit_modrm( p, dst, src ); @@ -624,6 +777,7 @@ void sse_movlhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); assert(dst.mod == mod_REG && src.mod == mod_REG); emit_2ub(p, X86_TWOB, 0x16); emit_modrm( p, dst, src ); @@ -633,6 +787,7 @@ void sse_orps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x56); emit_modrm( p, dst, src ); } @@ -641,6 +796,7 @@ void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x57); emit_modrm( p, dst, src ); } @@ -649,6 +805,7 @@ void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); assert(dst.file == file_MMX && (src.file == file_XMM || src.mod != mod_REG)); @@ -662,6 +819,7 @@ void sse2_cvtdq2ps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x5b); emit_modrm( p, dst, src ); } @@ -671,31 +829,34 @@ void sse2_cvtdq2ps( struct x86_function *p, * arg0. */ void sse_shufps( struct x86_function *p, - struct x86_reg dest, - struct x86_reg arg0, + struct x86_reg dst, + struct x86_reg src, unsigned char shuf) { + DUMP_RRI( __FUNCTION__, dst, src, shuf ); emit_2ub(p, X86_TWOB, 0xC6); - emit_modrm(p, dest, arg0); + emit_modrm(p, dst, src); emit_1ub(p, shuf); } void sse_cmpps( struct x86_function *p, - struct x86_reg dest, - struct x86_reg arg0, + struct x86_reg dst, + struct x86_reg src, unsigned char cc) { + DUMP_RRI( "CMPPS", dst, src, cc ); emit_2ub(p, X86_TWOB, 0xC2); - emit_modrm(p, dest, arg0); + emit_modrm(p, dst, src); emit_1ub(p, cc); } void sse_pmovmskb( struct x86_function *p, - struct x86_reg dest, + struct x86_reg dst, struct x86_reg src) { - emit_3ub(p, 0x66, X86_TWOB, 0xD7); - emit_modrm(p, dest, src); + DUMP_RR( __FUNCTION__, dst, src ); + emit_3ub(p, 0x66, X86_TWOB, 0xD7); + emit_modrm(p, dst, src); } /*********************************************************************** @@ -706,12 +867,13 @@ void sse_pmovmskb( struct x86_function *p, * Perform a reduced swizzle: */ void sse2_pshufd( struct x86_function *p, - struct x86_reg dest, - struct x86_reg arg0, + struct x86_reg dst, + struct x86_reg src, unsigned char shuf) { + DUMP_RRI( __FUNCTION__, dst, src, shuf ); emit_3ub(p, 0x66, X86_TWOB, 0x70); - emit_modrm(p, dest, arg0); + emit_modrm(p, dst, src); emit_1ub(p, shuf); } @@ -719,6 +881,7 @@ void sse2_cvttps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub( p, 0xF3, X86_TWOB, 0x5B ); emit_modrm( p, dst, src ); } @@ -727,6 +890,7 @@ void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x5B); emit_modrm( p, dst, src ); } @@ -735,6 +899,7 @@ void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x6B); emit_modrm( p, dst, src ); } @@ -743,6 +908,7 @@ void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x63); emit_modrm( p, dst, src ); } @@ -751,6 +917,7 @@ void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x67); emit_modrm( p, dst, src ); } @@ -759,6 +926,7 @@ void sse2_punpcklbw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x60); emit_modrm( p, dst, src ); } @@ -768,6 +936,7 @@ void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, X86_TWOB, 0x53); emit_modrm( p, dst, src ); } @@ -776,6 +945,7 @@ void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x53); emit_modrm( p, dst, src ); } @@ -784,6 +954,7 @@ void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); emit_2ub(p, 0x66, X86_TWOB); emit_op_modrm( p, 0x6e, 0x7e, dst, src ); } @@ -796,30 +967,35 @@ void sse2_movd( struct x86_function *p, */ void x87_fist( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); emit_1ub(p, 0xdb); emit_modrm_noreg(p, 2, dst); } void x87_fistp( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); emit_1ub(p, 0xdb); emit_modrm_noreg(p, 3, dst); } void x87_fild( struct x86_function *p, struct x86_reg arg ) { + DUMP_R( __FUNCTION__, arg ); emit_1ub(p, 0xdf); emit_modrm_noreg(p, 0, arg); } void x87_fldz( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xee); } void x87_fldcw( struct x86_function *p, struct x86_reg arg ) { + DUMP_R( __FUNCTION__, arg ); assert(arg.file == file_REG32); assert(arg.mod != mod_REG); emit_1ub(p, 0xd9); @@ -828,26 +1004,31 @@ void x87_fldcw( struct x86_function *p, struct x86_reg arg ) void x87_fld1( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xe8); } void x87_fldl2e( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xea); } void x87_fldln2( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xed); } void x87_fwait( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_1ub(p, 0x9b); } void x87_fnclex( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xdb, 0xe2); } @@ -884,49 +1065,55 @@ static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86 assert(0); } -void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - x87_arith_op(p, dst, arg, + DUMP_RR( __FUNCTION__, dst, src ); + x87_arith_op(p, dst, src, 0xd8, 0xc8, 0xdc, 0xc8, 4); } -void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - x87_arith_op(p, dst, arg, + DUMP_RR( __FUNCTION__, dst, src ); + x87_arith_op(p, dst, src, 0xd8, 0xe0, 0xdc, 0xe8, 4); } -void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - x87_arith_op(p, dst, arg, + DUMP_RR( __FUNCTION__, dst, src ); + x87_arith_op(p, dst, src, 0xd8, 0xe8, 0xdc, 0xe0, 5); } -void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - x87_arith_op(p, dst, arg, + DUMP_RR( __FUNCTION__, dst, src ); + x87_arith_op(p, dst, src, 0xd8, 0xc0, 0xdc, 0xc0, 0); } -void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - x87_arith_op(p, dst, arg, + DUMP_RR( __FUNCTION__, dst, src ); + x87_arith_op(p, dst, src, 0xd8, 0xf0, 0xdc, 0xf8, 6); } -void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) +void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - x87_arith_op(p, dst, arg, + DUMP_RR( __FUNCTION__, dst, src ); + x87_arith_op(p, dst, src, 0xd8, 0xf8, 0xdc, 0xf0, 7); @@ -934,6 +1121,7 @@ void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ) void x87_fmulp( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xc8+dst.idx); @@ -941,6 +1129,7 @@ void x87_fmulp( struct x86_function *p, struct x86_reg dst ) void x87_fsubp( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xe8+dst.idx); @@ -948,6 +1137,7 @@ void x87_fsubp( struct x86_function *p, struct x86_reg dst ) void x87_fsubrp( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xe0+dst.idx); @@ -955,6 +1145,7 @@ void x87_fsubrp( struct x86_function *p, struct x86_reg dst ) void x87_faddp( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xc0+dst.idx); @@ -962,6 +1153,7 @@ void x87_faddp( struct x86_function *p, struct x86_reg dst ) void x87_fdivp( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xf8+dst.idx); @@ -969,6 +1161,7 @@ void x87_fdivp( struct x86_function *p, struct x86_reg dst ) void x87_fdivrp( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xf0+dst.idx); @@ -976,70 +1169,83 @@ void x87_fdivrp( struct x86_function *p, struct x86_reg dst ) void x87_fucom( struct x86_function *p, struct x86_reg arg ) { + DUMP_R( __FUNCTION__, arg ); assert(arg.file == file_x87); emit_2ub(p, 0xdd, 0xe0+arg.idx); } void x87_fucomp( struct x86_function *p, struct x86_reg arg ) { + DUMP_R( __FUNCTION__, arg ); assert(arg.file == file_x87); emit_2ub(p, 0xdd, 0xe8+arg.idx); } void x87_fucompp( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xda, 0xe9); } void x87_fxch( struct x86_function *p, struct x86_reg arg ) { + DUMP_R( __FUNCTION__, arg ); assert(arg.file == file_x87); emit_2ub(p, 0xd9, 0xc8+arg.idx); } void x87_fabs( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xe1); } void x87_fchs( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xe0); } void x87_fcos( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xff); } void x87_fprndint( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xfc); } void x87_fscale( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xfd); } void x87_fsin( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xfe); } void x87_fsincos( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xfb); } void x87_fsqrt( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xfa); } void x87_fxtract( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xf4); } @@ -1049,6 +1255,7 @@ void x87_fxtract( struct x86_function *p ) */ void x87_f2xm1( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xf0); } @@ -1057,6 +1264,7 @@ void x87_f2xm1( struct x86_function *p ) */ void x87_fyl2x( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xf1); } @@ -1067,12 +1275,14 @@ void x87_fyl2x( struct x86_function *p ) */ void x87_fyl2xp1( struct x86_function *p ) { + DUMP( __FUNCTION__ ); emit_2ub(p, 0xd9, 0xf9); } void x87_fld( struct x86_function *p, struct x86_reg arg ) { + DUMP_R( __FUNCTION__, arg ); if (arg.file == file_x87) emit_2ub(p, 0xd9, 0xc0 + arg.idx); else { @@ -1083,6 +1293,7 @@ void x87_fld( struct x86_function *p, struct x86_reg arg ) void x87_fst( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); if (dst.file == file_x87) emit_2ub(p, 0xdd, 0xd0 + dst.idx); else { @@ -1093,6 +1304,7 @@ void x87_fst( struct x86_function *p, struct x86_reg dst ) void x87_fstp( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); if (dst.file == file_x87) emit_2ub(p, 0xdd, 0xd8 + dst.idx); else { @@ -1103,6 +1315,7 @@ void x87_fstp( struct x86_function *p, struct x86_reg dst ) void x87_fcom( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); if (dst.file == file_x87) emit_2ub(p, 0xd8, 0xd0 + dst.idx); else { @@ -1113,6 +1326,7 @@ void x87_fcom( struct x86_function *p, struct x86_reg dst ) void x87_fcomp( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); if (dst.file == file_x87) emit_2ub(p, 0xd8, 0xd8 + dst.idx); else { @@ -1124,6 +1338,7 @@ void x87_fcomp( struct x86_function *p, struct x86_reg dst ) void x87_fnstsw( struct x86_function *p, struct x86_reg dst ) { + DUMP_R( __FUNCTION__, dst ); assert(dst.file == file_REG32); if (dst.idx == reg_AX && @@ -1153,6 +1368,7 @@ void mmx_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); assert(dst.file == file_MMX && (src.file == file_MMX || src.mod != mod_REG)); @@ -1166,6 +1382,7 @@ void mmx_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); assert(dst.file == file_MMX && (src.file == file_MMX || src.mod != mod_REG)); @@ -1179,6 +1396,7 @@ void mmx_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); p->need_emms = 1; emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x6e, 0x7e, dst, src ); @@ -1188,6 +1406,7 @@ void mmx_movq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { + DUMP_RR( __FUNCTION__, dst, src ); p->need_emms = 1; emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x6f, 0x7f, dst, src ); -- cgit v1.2.3 From f30f3206121c56e5c50ccafb1467a8301044d411 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 12:38:14 +0100 Subject: util: add wrappers for float math functions on windows --- src/gallium/include/pipe/p_util.h | 46 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index dbca080a4b..77783b11a8 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -402,6 +402,52 @@ extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, int src_pitch, unsigned src_x, int src_y); + +#ifdef WIN32 + +static INLINE float cosf( float f ) +{ + return (float) cos( (double) f ); +} + +static INLINE float sinf( float f ) +{ + return (float) sin( (double) f ); +} + +static INLINE float ceilf( float f ) +{ + return (float) ceil( (double) f ); +} + +static INLINE float floorf( float f ) +{ + return (float) floor( (double) f ); +} + +static INLINE float powf( float f, float g ) +{ + return (float) pow( (double) f, (double) g ); +} + +static INLINE float sqrtf( float f ) +{ + return (float) sqrt( (double) f ); +} + +static INLINE float fabsf( float f ) +{ + return (float) fabs( (double) f ); +} + +static INLINE float logf( float f ) +{ + return (float) cos( (double) f ); +} + +#endif + + #ifdef __cplusplus } #endif -- cgit v1.2.3 From 615cdd3a535bb71754baa8b37e79b85af01854dd Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 12:39:59 +0100 Subject: tgsi: use new float math funcs, drop local disassembly code --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 88 ++--- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 591 +++++++--------------------- 2 files changed, 178 insertions(+), 501 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 78e7dec569..29e104bbd1 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -287,10 +287,10 @@ micro_abs( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (float) fabs( (double) src->f[0] ); - dst->f[1] = (float) fabs( (double) src->f[1] ); - dst->f[2] = (float) fabs( (double) src->f[2] ); - dst->f[3] = (float) fabs( (double) src->f[3] ); + dst->f[0] = fabsf( src->f[0] ); + dst->f[1] = fabsf( src->f[1] ); + dst->f[2] = fabsf( src->f[2] ); + dst->f[3] = fabsf( src->f[3] ); } static void @@ -334,10 +334,10 @@ micro_ceil( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (float) ceil( (double) src->f[0] ); - dst->f[1] = (float) ceil( (double) src->f[1] ); - dst->f[2] = (float) ceil( (double) src->f[2] ); - dst->f[3] = (float) ceil( (double) src->f[3] ); + dst->f[0] = ceilf( src->f[0] ); + dst->f[1] = ceilf( src->f[1] ); + dst->f[2] = ceilf( src->f[2] ); + dst->f[3] = ceilf( src->f[3] ); } static void @@ -345,10 +345,10 @@ micro_cos( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (float) cos( (double) src->f[0] ); - dst->f[1] = (float) cos( (double) src->f[1] ); - dst->f[2] = (float) cos( (double) src->f[2] ); - dst->f[3] = (float) cos( (double) src->f[3] ); + dst->f[0] = cosf( src->f[0] ); + dst->f[1] = cosf( src->f[1] ); + dst->f[2] = cosf( src->f[2] ); + dst->f[3] = cosf( src->f[3] ); } static void @@ -430,10 +430,10 @@ micro_exp2( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src) { - dst->f[0] = (float) pow( 2.0, (double) src->f[0] ); - dst->f[1] = (float) pow( 2.0, (double) src->f[1] ); - dst->f[2] = (float) pow( 2.0, (double) src->f[2] ); - dst->f[3] = (float) pow( 2.0, (double) src->f[3] ); + dst->f[0] = powf( 2.0f, src->f[0] ); + dst->f[1] = powf( 2.0f, src->f[1] ); + dst->f[2] = powf( 2.0f, src->f[2] ); + dst->f[3] = powf( 2.0f, src->f[3] ); } static void @@ -463,10 +463,10 @@ micro_flr( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (float) floor( (double) src->f[0] ); - dst->f[1] = (float) floor( (double) src->f[1] ); - dst->f[2] = (float) floor( (double) src->f[2] ); - dst->f[3] = (float) floor( (double) src->f[3] ); + dst->f[0] = floorf( src->f[0] ); + dst->f[1] = floorf( src->f[1] ); + dst->f[2] = floorf( src->f[2] ); + dst->f[3] = floorf( src->f[3] ); } static void @@ -474,10 +474,10 @@ micro_frc( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = src->f[0] - (float) floor( (double) src->f[0] ); - dst->f[1] = src->f[1] - (float) floor( (double) src->f[1] ); - dst->f[2] = src->f[2] - (float) floor( (double) src->f[2] ); - dst->f[3] = src->f[3] - (float) floor( (double) src->f[3] ); + dst->f[0] = src->f[0] - floorf( src->f[0] ); + dst->f[1] = src->f[1] - floorf( src->f[1] ); + dst->f[2] = src->f[2] - floorf( src->f[2] ); + dst->f[3] = src->f[3] - floorf( src->f[3] ); } static void @@ -510,10 +510,10 @@ micro_lg2( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (float) log( (double) src->f[0] ) * 1.442695f; - dst->f[1] = (float) log( (double) src->f[1] ) * 1.442695f; - dst->f[2] = (float) log( (double) src->f[2] ) * 1.442695f; - dst->f[3] = (float) log( (double) src->f[3] ) * 1.442695f; + dst->f[0] = logf( src->f[0] ) * 1.442695f; + dst->f[1] = logf( src->f[1] ) * 1.442695f; + dst->f[2] = logf( src->f[2] ) * 1.442695f; + dst->f[3] = logf( src->f[3] ) * 1.442695f; } static void @@ -764,10 +764,10 @@ micro_pow( const union tgsi_exec_channel *src0, const union tgsi_exec_channel *src1 ) { - dst->f[0] = (float) pow( (double) src0->f[0], (double) src1->f[0] ); - dst->f[1] = (float) pow( (double) src0->f[1], (double) src1->f[1] ); - dst->f[2] = (float) pow( (double) src0->f[2], (double) src1->f[2] ); - dst->f[3] = (float) pow( (double) src0->f[3], (double) src1->f[3] ); + dst->f[0] = powf( src0->f[0], src1->f[0] ); + dst->f[1] = powf( src0->f[1], src1->f[1] ); + dst->f[2] = powf( src0->f[2], src1->f[2] ); + dst->f[3] = powf( src0->f[3], src1->f[3] ); } static void @@ -775,10 +775,10 @@ micro_rnd( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (float) floor( (double) (src->f[0] + 0.5f) ); - dst->f[1] = (float) floor( (double) (src->f[1] + 0.5f) ); - dst->f[2] = (float) floor( (double) (src->f[2] + 0.5f) ); - dst->f[3] = (float) floor( (double) (src->f[3] + 0.5f) ); + dst->f[0] = floorf( src->f[0] + 0.5f ); + dst->f[1] = floorf( src->f[1] + 0.5f ); + dst->f[2] = floorf( src->f[2] + 0.5f ); + dst->f[3] = floorf( src->f[3] + 0.5f ); } static void @@ -833,20 +833,20 @@ micro_sin( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (float) sin( (double) src->f[0] ); - dst->f[1] = (float) sin( (double) src->f[1] ); - dst->f[2] = (float) sin( (double) src->f[2] ); - dst->f[3] = (float) sin( (double) src->f[3] ); + dst->f[0] = sinf( src->f[0] ); + dst->f[1] = sinf( src->f[1] ); + dst->f[2] = sinf( src->f[2] ); + dst->f[3] = sinf( src->f[3] ); } static void micro_sqrt( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { - dst->f[0] = (float) sqrt( (double) src->f[0] ); - dst->f[1] = (float) sqrt( (double) src->f[1] ); - dst->f[2] = (float) sqrt( (double) src->f[2] ); - dst->f[3] = (float) sqrt( (double) src->f[3] ); + dst->f[0] = sqrtf( src->f[0] ); + dst->f[1] = sqrtf( src->f[1] ); + dst->f[2] = sqrtf( src->f[2] ); + dst->f[3] = sqrtf( src->f[3] ); } static void diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index d47935e982..c3295a27ff 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -38,113 +38,6 @@ #define HIGH_PRECISION 1 /* for 1/sqrt() */ -#define DUMP_SSE 0 - -#if DUMP_SSE - -static void -_print_reg( - struct x86_reg reg ) -{ - if (reg.mod != mod_REG) - debug_printf( "[" ); - - switch( reg.file ) { - case file_REG32: - switch( reg.idx ) { - case reg_AX: - debug_printf( "EAX" ); - break; - case reg_CX: - debug_printf( "ECX" ); - break; - case reg_DX: - debug_printf( "EDX" ); - break; - case reg_BX: - debug_printf( "EBX" ); - break; - case reg_SP: - debug_printf( "ESP" ); - break; - case reg_BP: - debug_printf( "EBP" ); - break; - case reg_SI: - debug_printf( "ESI" ); - break; - case reg_DI: - debug_printf( "EDI" ); - break; - } - break; - case file_MMX: - assert( 0 ); - break; - case file_XMM: - debug_printf( "XMM%u", reg.idx ); - break; - case file_x87: - assert( 0 ); - break; - } - - if (reg.mod == mod_DISP8 || - reg.mod == mod_DISP32) - debug_printf("+%d", reg.disp); - - if (reg.mod != mod_REG) - debug_printf( "]" ); -} - -static void -_fill( - const char *op ) -{ - unsigned count = 10 - strlen( op ); - - while( count-- ) { - debug_printf( " " ); - } -} - -#define DUMP_START() debug_printf( "\nsse-dump start ----------------" ) -#define DUMP_END() debug_printf( "\nsse-dump end ----------------\n" ) -#define DUMP( OP ) debug_printf( "\n%s", OP ) -#define DUMP_I( OP, I ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - debug_printf( "%u", I ); } while( 0 ) -#define DUMP_R( OP, R0 ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - _print_reg( R0 ); } while( 0 ) -#define DUMP_RR( OP, R0, R1 ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - _print_reg( R0 );\ - debug_printf( ", " );\ - _print_reg( R1 ); } while( 0 ) -#define DUMP_RRI( OP, R0, R1, I ) do {\ - debug_printf( "\n%s", OP );\ - _fill( OP );\ - _print_reg( R0 );\ - debug_printf( ", " );\ - _print_reg( R1 );\ - debug_printf( ", " );\ - debug_printf( "%u", I ); } while( 0 ) - -#else - -#define DUMP_START() -#define DUMP_END() -#define DUMP( OP ) -#define DUMP_I( OP, I ) -#define DUMP_R( OP, R0 ) -#define DUMP_RR( OP, R0, R1 ) -#define DUMP_RRI( OP, R0, R1, I ) - -#endif #define FOR_EACH_CHANNEL( CHAN )\ for( CHAN = 0; CHAN < 4; CHAN++ ) @@ -310,200 +203,6 @@ get_coef( ((vec * 3 + member) * 4 + chan) * 4 ); } -/** - * X86 rtasm wrappers. - */ - -static void -emit_addps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ADDPS", dst, src ); - sse_addps( func, dst, src ); -} - -static void -emit_andnps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ANDNPS", dst, src ); - sse_andnps( func, dst, src ); -} - -static void -emit_andps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ANDPS", dst, src ); - sse_andps( func, dst, src ); -} - -static void -emit_call( - struct x86_function *func, - void (* addr)() ) -{ - struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); - - DUMP_I( "CALL", addr ); - x86_mov_reg_imm( func, ecx, (unsigned long) addr ); - x86_call( func, ecx ); -} - -static void -emit_cmpps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src, - enum sse_cc cc ) -{ - DUMP_RRI( "CMPPS", dst, src, cc ); - sse_cmpps( func, dst, src, cc ); -} - -static void -emit_cvttps2dq( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "CVTTPS2DQ", dst, src ); - sse2_cvttps2dq( func, dst, src ); -} - -static void -emit_maxps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MAXPS", dst, src ); - sse_maxps( func, dst, src ); -} - -static void -emit_minps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MINPS", dst, src ); - sse_minps( func, dst, src ); -} - -static void -emit_mov( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOV", dst, src ); - x86_mov( func, dst, src ); -} - -static void -emit_movaps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOVAPS", dst, src ); - sse_movaps( func, dst, src ); -} - -static void -emit_movss( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOVSS", dst, src ); - sse_movss( func, dst, src ); -} - -static void -emit_movups( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MOVUPS", dst, src ); - sse_movups( func, dst, src ); -} - -static void -emit_mulps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "MULPS", dst, src ); - sse_mulps( func, dst, src ); -} - -static void -emit_or( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "OR", dst, src ); - x86_or( func, dst, src ); -} - -static void -emit_orps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "ORPS", dst, src ); - sse_orps( func, dst, src ); -} - -static void -emit_pmovmskb( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "PMOVMSKB", dst, src ); - sse_pmovmskb( func, dst, src ); -} - -static void -emit_pop( - struct x86_function *func, - struct x86_reg dst ) -{ - DUMP_R( "POP", dst ); - x86_pop( func, dst ); -} - -static void -emit_push( - struct x86_function *func, - struct x86_reg dst ) -{ - DUMP_R( "PUSH", dst ); - x86_push( func, dst ); -} - -static void -emit_rcpps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "RCPPS", dst, src ); - sse2_rcpps( func, dst, src ); -} #ifdef WIN32 static void @@ -511,7 +210,6 @@ emit_retw( struct x86_function *func, unsigned size ) { - DUMP_I( "RET", size ); x86_retw( func, size ); } #else @@ -519,51 +217,10 @@ static void emit_ret( struct x86_function *func ) { - DUMP( "RET" ); x86_ret( func ); } #endif -static void -emit_rsqrtps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "RSQRTPS", dst, src ); - sse_rsqrtps( func, dst, src ); -} - -static void -emit_shufps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src, - unsigned char shuf ) -{ - DUMP_RRI( "SHUFPS", dst, src, shuf ); - sse_shufps( func, dst, src, shuf ); -} - -static void -emit_subps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "SUBPS", dst, src ); - sse_subps( func, dst, src ); -} - -static void -emit_xorps( - struct x86_function *func, - struct x86_reg dst, - struct x86_reg src ) -{ - DUMP_RR( "XORPS", dst, src ); - sse_xorps( func, dst, src ); -} /** * Data fetch helpers. @@ -582,11 +239,11 @@ emit_const( unsigned vec, unsigned chan ) { - emit_movss( + sse_movss( func, make_xmm( xmm ), get_const( vec, chan ) ); - emit_shufps( + sse_shufps( func, make_xmm( xmm ), make_xmm( xmm ), @@ -600,11 +257,11 @@ emit_immediate( unsigned vec, unsigned chan ) { - emit_movss( + sse_movss( func, make_xmm( xmm ), get_immediate( vec, chan ) ); - emit_shufps( + sse_shufps( func, make_xmm( xmm ), make_xmm( xmm ), @@ -625,7 +282,7 @@ emit_inputf( unsigned vec, unsigned chan ) { - emit_movups( + sse_movups( func, make_xmm( xmm ), get_input( vec, chan ) ); @@ -644,7 +301,7 @@ emit_output( unsigned vec, unsigned chan ) { - emit_movups( + sse_movups( func, get_output( vec, chan ), make_xmm( xmm ) ); @@ -663,7 +320,7 @@ emit_tempf( unsigned vec, unsigned chan ) { - emit_movaps( + sse_movaps( func, make_xmm( xmm ), get_temp( vec, chan ) ); @@ -684,11 +341,11 @@ emit_coef( unsigned chan, unsigned member ) { - emit_movss( + sse_movss( func, make_xmm( xmm ), get_coef( vec, chan, member ) ); - emit_shufps( + sse_shufps( func, make_xmm( xmm ), make_xmm( xmm ), @@ -706,7 +363,7 @@ emit_inputs( unsigned vec, unsigned chan ) { - emit_movups( + sse_movups( func, get_input( vec, chan ), make_xmm( xmm ) ); @@ -719,7 +376,7 @@ emit_temps( unsigned vec, unsigned chan ) { - emit_movaps( + sse_movaps( func, get_temp( vec, chan ), make_xmm( xmm ) ); @@ -796,39 +453,39 @@ static void emit_push_gp( struct x86_function *func ) { - emit_push( + x86_push( func, get_const_base() ); - emit_push( + x86_push( func, get_input_base() ); - emit_push( + x86_push( func, get_output_base() ); /* It is important on non-win32 platforms that temp base is pushed last. */ - emit_push( + x86_push( func, get_temp_base() ); } static void -emit_pop_gp( +x86_pop_gp( struct x86_function *func ) { /* Restore GP registers in a reverse order. */ - emit_pop( + x86_pop( func, get_temp_base() ); - emit_pop( + x86_pop( func, get_output_base() ); - emit_pop( + x86_pop( func, get_input_base() ); - emit_pop( + x86_pop( func, get_const_base() ); } @@ -839,7 +496,7 @@ emit_func_call_dst( unsigned xmm_dst, void (*code)() ) { - emit_movaps( + sse_movaps( func, get_temp( TEMP_R0, 0 ), make_xmm( xmm_dst ) ); @@ -848,19 +505,22 @@ emit_func_call_dst( func ); #ifdef WIN32 - emit_push( + x86_push( func, get_temp( TEMP_R0, 0 ) ); #endif - emit_call( - func, - code ); + { + struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); + + x86_mov_reg_imm( func, ecx, (unsigned long) code ); + x86_call( func, ecx ); + } - emit_pop_gp( + x86_pop_gp( func ); - emit_movaps( + sse_movaps( func, make_xmm( xmm_dst ), get_temp( TEMP_R0, 0 ) ); @@ -873,7 +533,7 @@ emit_func_call_dst_src( unsigned xmm_src, void (*code)() ) { - emit_movaps( + sse_movaps( func, get_temp( TEMP_R0, 1 ), make_xmm( xmm_src ) ); @@ -893,7 +553,7 @@ emit_abs( struct x86_function *func, unsigned xmm ) { - emit_andps( + sse_andps( func, make_xmm( xmm ), get_temp( @@ -907,7 +567,7 @@ emit_add( unsigned xmm_dst, unsigned xmm_src ) { - emit_addps( + sse_addps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -918,17 +578,15 @@ cos4f( float *store ) { #ifdef WIN32 - store[0] = (float) cos( (double) store[0] ); - store[1] = (float) cos( (double) store[1] ); - store[2] = (float) cos( (double) store[2] ); - store[3] = (float) cos( (double) store[3] ); + const unsigned X = 0; #else const unsigned X = TEMP_R0 * 16; +#endif + store[X + 0] = cosf( store[X + 0] ); store[X + 1] = cosf( store[X + 1] ); store[X + 2] = cosf( store[X + 2] ); store[X + 3] = cosf( store[X + 3] ); -#endif } static void @@ -947,17 +605,14 @@ ex24f( float *store ) { #ifdef WIN32 - store[0] = (float) pow( 2.0, (double) store[0] ); - store[1] = (float) pow( 2.0, (double) store[1] ); - store[2] = (float) pow( 2.0, (double) store[2] ); - store[3] = (float) pow( 2.0, (double) store[3] ); + const unsigned X = 0; #else const unsigned X = TEMP_R0 * 16; +#endif store[X + 0] = powf( 2.0f, store[X + 0] ); store[X + 1] = powf( 2.0f, store[X + 1] ); store[X + 2] = powf( 2.0f, store[X + 2] ); store[X + 3] = powf( 2.0f, store[X + 3] ); -#endif } static void @@ -976,7 +631,7 @@ emit_f2it( struct x86_function *func, unsigned xmm ) { - emit_cvttps2dq( + sse2_cvttps2dq( func, make_xmm( xmm ), make_xmm( xmm ) ); @@ -991,10 +646,10 @@ flr4f( #else const unsigned X = TEMP_R0 * 16; #endif - store[X + 0] = (float) floor( (double) store[X + 0] ); - store[X + 1] = (float) floor( (double) store[X + 1] ); - store[X + 2] = (float) floor( (double) store[X + 2] ); - store[X + 3] = (float) floor( (double) store[X + 3] ); + store[X + 0] = floorf( store[X + 0] ); + store[X + 1] = floorf( store[X + 1] ); + store[X + 2] = floorf( store[X + 2] ); + store[X + 3] = floorf( store[X + 3] ); } static void @@ -1017,10 +672,10 @@ frc4f( #else const unsigned X = TEMP_R0 * 16; #endif - store[X + 0] -= (float) floor( (double) store[X + 0] ); - store[X + 1] -= (float) floor( (double) store[X + 1] ); - store[X + 2] -= (float) floor( (double) store[X + 2] ); - store[X + 3] -= (float) floor( (double) store[X + 3] ); + store[X + 0] -= floorf( store[X + 0] ); + store[X + 1] -= floorf( store[X + 1] ); + store[X + 2] -= floorf( store[X + 2] ); + store[X + 3] -= floorf( store[X + 3] ); } static void @@ -1066,7 +721,7 @@ emit_MOV( unsigned xmm_dst, unsigned xmm_src ) { - emit_movups( + sse_movups( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -1077,7 +732,7 @@ emit_mul (struct x86_function *func, unsigned xmm_dst, unsigned xmm_src) { - emit_mulps( + sse_mulps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -1088,7 +743,7 @@ emit_neg( struct x86_function *func, unsigned xmm ) { - emit_xorps( + sse_xorps( func, make_xmm( xmm ), get_temp( @@ -1101,17 +756,14 @@ pow4f( float *store ) { #ifdef WIN32 - store[0] = (float) pow( (double) store[0], (double) store[4] ); - store[1] = (float) pow( (double) store[1], (double) store[5] ); - store[2] = (float) pow( (double) store[2], (double) store[6] ); - store[3] = (float) pow( (double) store[3], (double) store[7] ); + const unsigned X = 0; #else const unsigned X = TEMP_R0 * 16; +#endif store[X + 0] = powf( store[X + 0], store[X + 4] ); store[X + 1] = powf( store[X + 1], store[X + 5] ); store[X + 2] = powf( store[X + 2], store[X + 6] ); store[X + 3] = powf( store[X + 3], store[X + 7] ); -#endif } static void @@ -1133,7 +785,11 @@ emit_rcp ( unsigned xmm_dst, unsigned xmm_src ) { - emit_rcpps( + /* On Intel CPUs at least, this is only accurate to 12 bits -- not + * good enough. Need to either emit a proper divide or use the + * iterative technique described below in emit_rsqrt(). + */ + sse2_rcpps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -1145,17 +801,14 @@ rsqrt4f( float *store ) { #ifdef WIN32 - store[0] = 1.0F / (float) sqrt( (double) store[0] ); - store[1] = 1.0F / (float) sqrt( (double) store[1] ); - store[2] = 1.0F / (float) sqrt( (double) store[2] ); - store[3] = 1.0F / (float) sqrt( (double) store[3] ); + const unsigned X = 0; #else const unsigned X = TEMP_R0 * 16; - store[X + 0] = 1.0F / sqrt( store[X + 0] ); - store[X + 1] = 1.0F / sqrt( store[X + 1] ); - store[X + 2] = 1.0F / sqrt( store[X + 2] ); - store[X + 3] = 1.0F / sqrt( store[X + 3] ); #endif + store[X + 0] = 1.0F / sqrtf( store[X + 0] ); + store[X + 1] = 1.0F / sqrtf( store[X + 1] ); + store[X + 2] = 1.0F / sqrtf( store[X + 2] ); + store[X + 3] = 1.0F / sqrtf( store[X + 3] ); } #endif @@ -1166,12 +819,41 @@ emit_rsqrt( unsigned xmm_src ) { #if HIGH_PRECISION +#if 1 emit_func_call_dst_src( func, xmm_dst, xmm_src, rsqrt4f ); #else + /* Although rsqrtps() and rcpps() are low precision on some/all SSE + * implementations, it is possible to improve its precision at + * fairly low cost, using a newton/raphson step, as below: + * + * x1 = 2 * rcpps(a) - a * rcpps(a) * rcpps(a) + * x1 = 0.5 * rsqrtps(a) * [3.0 - (a * rsqrtps(a))* rsqrtps(a)] + * + * See: http://softwarecommunity.intel.com/articles/eng/1818.htm + */ + /* This is some code that woudl do the above for a scalar 'a'. We + * obviously are interested in a vector version: + * + * movss xmm3, a; + * movss xmm1, half; + * movss xmm2, three; + * rsqrtss xmm0, xmm3; + * mulss xmm3, xmm0; + * mulss xmm1, xmm0; + * mulss xmm3, xmm0; + * subss xmm2, xmm3; + * mulss xmm1, xmm2; + * movss x, xmm1; + */ +#endif +#else + /* On Intel CPUs at least, this is only accurate to 12 bits -- not + * good enough. + */ emit_rsqrtps( func, make_xmm( xmm_dst ), @@ -1184,7 +866,7 @@ emit_setsign( struct x86_function *func, unsigned xmm ) { - emit_orps( + sse_orps( func, make_xmm( xmm ), get_temp( @@ -1197,17 +879,14 @@ sin4f( float *store ) { #ifdef WIN32 - store[0] = (float) sin( (double) store[0] ); - store[1] = (float) sin( (double) store[1] ); - store[2] = (float) sin( (double) store[2] ); - store[3] = (float) sin( (double) store[3] ); + const unsigned X = 0; #else const unsigned X = TEMP_R0 * 16; +#endif store[X + 0] = sinf( store[X + 0] ); store[X + 1] = sinf( store[X + 1] ); store[X + 2] = sinf( store[X + 2] ); store[X + 3] = sinf( store[X + 3] ); -#endif } static void @@ -1226,7 +905,7 @@ emit_sub( unsigned xmm_dst, unsigned xmm_src ) { - emit_subps( + sse_subps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); @@ -1435,16 +1114,16 @@ emit_kil( } } - emit_push( + x86_push( func, x86_make_reg( file_REG32, reg_AX ) ); - emit_push( + x86_push( func, x86_make_reg( file_REG32, reg_DX ) ); FOR_EACH_CHANNEL( chan_index ) { if( uniquemask & (1 << chan_index) ) { - emit_cmpps( + sse_cmpps( func, make_xmm( registers[chan_index] ), get_temp( @@ -1453,17 +1132,17 @@ emit_kil( cc_LessThan ); if( chan_index == firstchan ) { - emit_pmovmskb( + sse_pmovmskb( func, x86_make_reg( file_REG32, reg_AX ), make_xmm( registers[chan_index] ) ); } else { - emit_pmovmskb( + sse_pmovmskb( func, x86_make_reg( file_REG32, reg_DX ), make_xmm( registers[chan_index] ) ); - emit_or( + x86_or( func, x86_make_reg( file_REG32, reg_AX ), x86_make_reg( file_REG32, reg_DX ) ); @@ -1471,17 +1150,17 @@ emit_kil( } } - emit_or( + x86_or( func, get_temp( TGSI_EXEC_TEMP_KILMASK_I, TGSI_EXEC_TEMP_KILMASK_C ), x86_make_reg( file_REG32, reg_AX ) ); - emit_pop( + x86_pop( func, x86_make_reg( file_REG32, reg_DX ) ); - emit_pop( + x86_pop( func, x86_make_reg( file_REG32, reg_AX ) ); } @@ -1497,12 +1176,12 @@ emit_setcc( FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); FETCH( func, *inst, 1, 1, chan_index ); - emit_cmpps( + sse_cmpps( func, make_xmm( 0 ), make_xmm( 1 ), cc ); - emit_andps( + sse_andps( func, make_xmm( 0 ), get_temp( @@ -1523,22 +1202,22 @@ emit_cmp( FETCH( func, *inst, 0, 0, chan_index ); FETCH( func, *inst, 1, 1, chan_index ); FETCH( func, *inst, 2, 2, chan_index ); - emit_cmpps( + sse_cmpps( func, make_xmm( 0 ), get_temp( TGSI_EXEC_TEMP_00000000_I, TGSI_EXEC_TEMP_00000000_C ), cc_LessThan ); - emit_andps( + sse_andps( func, make_xmm( 1 ), make_xmm( 0 ) ); - emit_andnps( + sse_andnps( func, make_xmm( 0 ), make_xmm( 2 ) ); - emit_orps( + sse_orps( func, make_xmm( 0 ), make_xmm( 1 ) ); @@ -1589,7 +1268,7 @@ emit_instruction( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { FETCH( func, *inst, 0, 0, CHAN_X ); - emit_maxps( + sse_maxps( func, make_xmm( 0 ), get_temp( @@ -1601,7 +1280,7 @@ emit_instruction( /* XMM[1] = SrcReg[0].yyyy */ FETCH( func, *inst, 1, 0, CHAN_Y ); /* XMM[1] = max(XMM[1], 0) */ - emit_maxps( + sse_maxps( func, make_xmm( 1 ), get_temp( @@ -1610,14 +1289,14 @@ emit_instruction( /* XMM[2] = SrcReg[0].wwww */ FETCH( func, *inst, 2, 0, CHAN_W ); /* XMM[2] = min(XMM[2], 128.0) */ - emit_minps( + sse_minps( func, make_xmm( 2 ), get_temp( TGSI_EXEC_TEMP_128_I, TGSI_EXEC_TEMP_128_C ) ); /* XMM[2] = max(XMM[2], -128.0) */ - emit_maxps( + sse_maxps( func, make_xmm( 2 ), get_temp( @@ -1625,16 +1304,16 @@ emit_instruction( TGSI_EXEC_TEMP_MINUS_128_C ) ); emit_pow( func, 1, 2 ); FETCH( func, *inst, 0, 0, CHAN_X ); - emit_xorps( + sse_xorps( func, make_xmm( 2 ), make_xmm( 2 ) ); - emit_cmpps( + sse_cmpps( func, make_xmm( 2 ), make_xmm( 0 ), cc_LessThanEqual ); - emit_andps( + sse_andps( func, make_xmm( 2 ), make_xmm( 1 ) ); @@ -1756,7 +1435,7 @@ emit_instruction( FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); FETCH( func, *inst, 1, 1, chan_index ); - emit_minps( + sse_minps( func, make_xmm( 0 ), make_xmm( 1 ) ); @@ -1768,7 +1447,7 @@ emit_instruction( FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); FETCH( func, *inst, 1, 1, chan_index ); - emit_maxps( + sse_maxps( func, make_xmm( 0 ), make_xmm( 1 ) ); @@ -2376,8 +2055,6 @@ tgsi_emit_sse2( unsigned ok = 1; uint num_immediates = 0; - DUMP_START(); - func->csr = func->store; tgsi_parse_init( &parse, tokens ); @@ -2387,24 +2064,24 @@ tgsi_emit_sse2( */ if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { /* DECLARATION phase, do not load output argument. */ - emit_mov( + x86_mov( func, get_input_base(), get_argument( 0 ) ); /* skipping outputs argument here */ - emit_mov( + x86_mov( func, get_const_base(), get_argument( 2 ) ); - emit_mov( + x86_mov( func, get_temp_base(), get_argument( 3 ) ); - emit_mov( + x86_mov( func, get_coef_base(), get_argument( 4 ) ); - emit_mov( + x86_mov( func, get_immediate_base(), get_argument( 5 ) ); @@ -2412,23 +2089,23 @@ tgsi_emit_sse2( else { assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX); - emit_mov( + x86_mov( func, get_input_base(), get_argument( 0 ) ); - emit_mov( + x86_mov( func, get_output_base(), get_argument( 1 ) ); - emit_mov( + x86_mov( func, get_const_base(), get_argument( 2 ) ); - emit_mov( + x86_mov( func, get_temp_base(), get_argument( 3 ) ); - emit_mov( + x86_mov( func, get_immediate_base(), get_argument( 4 ) ); @@ -2451,7 +2128,7 @@ tgsi_emit_sse2( if( !instruction_phase ) { /* INSTRUCTION phase, overwrite coeff with output. */ instruction_phase = TRUE; - emit_mov( + x86_mov( func, get_output_base(), get_argument( 1 ) ); @@ -2463,8 +2140,10 @@ tgsi_emit_sse2( &parse.FullToken.FullInstruction ); if (!ok) { - debug_printf("failed to translate tgsi opcode %d to SSE\n", - parse.FullToken.FullInstruction.Instruction.Opcode ); + debug_printf("failed to translate tgsi opcode %d to SSE (%s)\n", + parse.FullToken.FullInstruction.Instruction.Opcode, + parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ? + "vertex shader" : "fragment shader"); } break; @@ -2499,8 +2178,6 @@ tgsi_emit_sse2( tgsi_parse_free( &parse ); - DUMP_END(); - return ok; } -- cgit v1.2.3 From 76a3590046d74c923d9837f82ece1587224b9a41 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 14:19:15 +0100 Subject: util: work around freaky win32 math.h --- src/gallium/include/pipe/p_util.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 77783b11a8..43d94ec4ba 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -405,6 +405,8 @@ extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, #ifdef WIN32 +#if !defined(_INC_MATH) || !defined(__cplusplus) + static INLINE float cosf( float f ) { return (float) cos( (double) f ); @@ -444,7 +446,7 @@ static INLINE float logf( float f ) { return (float) cos( (double) f ); } - +#endif /* _INC_MATH */ #endif -- cgit v1.2.3 From 08717d94619802f7816420be155c0c92fa727109 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 21 Apr 2008 20:44:45 +0900 Subject: gallium: Do not mistake pipe state objects for state tracker state objects. --- src/gallium/auxiliary/cso_cache/cso_context.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 746b176185..4541be8a5c 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -266,8 +266,9 @@ void cso_save_samplers(struct cso_context *ctx) void cso_restore_samplers(struct cso_context *ctx) { - cso_set_samplers(ctx, ctx->nr_samplers_saved, - (const struct pipe_sampler_state **) ctx->samplers_saved); + ctx->nr_samplers = ctx->nr_samplers_saved; + memcpy(ctx->samplers, ctx->samplers_saved, sizeof(ctx->samplers)); + cso_single_sampler_done( ctx ); } @@ -313,6 +314,8 @@ void cso_restore_sampler_textures( struct cso_context *ctx ) pipe_texture_reference(&ctx->textures[i], NULL); ctx->pipe->set_sampler_textures(ctx->pipe, ctx->nr_textures, ctx->textures); + + ctx->nr_textures_saved = 0; } -- cgit v1.2.3 From 13d8b1b211a803f44ffe325e7eed887cce4abaca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 21 Apr 2008 22:26:33 +0900 Subject: gallium: Set all state via cso_context in blit/gen_mipmap utils. cso_restore_* functions are implemented on top of cso_set_*, therefore they require full knowledge of the current pipe state to work correctly. Directly calling pipe's set_*_state functions will lead to undefined state. Also save and restore shaders. --- src/gallium/auxiliary/util/u_blit.c | 10 +++++++--- src/gallium/auxiliary/util/u_gen_mipmap.c | 12 ++++++++---- 2 files changed, 15 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index eec5e600c9..be5e83e834 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -300,6 +300,8 @@ util_blit_pixels(struct blit_state *ctx, cso_save_samplers(ctx->cso); cso_save_sampler_textures(ctx->cso); cso_save_framebuffer(ctx->cso); + cso_save_fragment_shader(ctx->cso); + cso_save_vertex_shader(ctx->cso); /* set misc state we care about */ cso_set_blend(ctx->cso, &ctx->blend); @@ -313,11 +315,11 @@ util_blit_pixels(struct blit_state *ctx, cso_single_sampler_done(ctx->cso); /* texture */ - pipe->set_sampler_textures(pipe, 1, &tex); + cso_set_sampler_textures(ctx->cso, 1, &tex); /* shaders */ - pipe->bind_fs_state(pipe, ctx->fs); - pipe->bind_vs_state(pipe, ctx->vs); + cso_set_fragment_shader(ctx->cso, ctx->fs); + cso_set_vertex_shader(ctx->cso, ctx->vs); /* drawing dest */ memset(&fb, 0, sizeof(fb)); @@ -344,6 +346,8 @@ util_blit_pixels(struct blit_state *ctx, cso_restore_samplers(ctx->cso); cso_restore_sampler_textures(ctx->cso); cso_restore_framebuffer(ctx->cso); + cso_restore_fragment_shader(ctx->cso); + cso_restore_vertex_shader(ctx->cso); /* free the texture */ pipe_surface_reference(&texSurf, NULL); diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 2fd214d22e..f0c4063b28 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -880,16 +880,18 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_save_samplers(ctx->cso); cso_save_sampler_textures(ctx->cso); cso_save_framebuffer(ctx->cso); + cso_save_fragment_shader(ctx->cso); + cso_save_vertex_shader(ctx->cso); /* bind our state */ cso_set_blend(ctx->cso, &ctx->blend); cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); cso_set_rasterizer(ctx->cso, &ctx->rasterizer); - pipe->bind_vs_state(pipe, ctx->vs); - pipe->bind_fs_state(pipe, ctx->fs); + cso_set_fragment_shader(ctx->cso, ctx->fs); + cso_set_vertex_shader(ctx->cso, ctx->vs); #if 0 - pipe->set_viewport_state(pipe, &ctx->viewport); + cso_set_viewport(ctx->cso, &ctx->viewport); #endif /* init framebuffer state */ @@ -930,7 +932,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]); #endif - pipe->set_sampler_textures(pipe, 1, &pt); + cso_set_sampler_textures(ctx->cso, 1, &pt); /* quad coords in window coords (bypassing clipping, viewport mapping) */ set_vertex_data(ctx, @@ -954,4 +956,6 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_restore_samplers(ctx->cso); cso_restore_sampler_textures(ctx->cso); cso_restore_framebuffer(ctx->cso); + cso_restore_fragment_shader(ctx->cso); + cso_restore_vertex_shader(ctx->cso); } -- cgit v1.2.3 From e29583afcb238cf7a70089cfdf50a69ca277c53a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 22 Apr 2008 00:14:08 +0900 Subject: gallium: Include dependent header. --- src/gallium/auxiliary/tgsi/util/tgsi_parse.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h index a98e88e343..da0121c482 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h @@ -1,6 +1,8 @@ #if !defined TGSI_PARSE_H #define TGSI_PARSE_H +#include "pipe/p_shader_tokens.h" + #if defined __cplusplus extern "C" { #endif -- cgit v1.2.3 From d3045ebb0642b09b4d353be6d4a258e6766061e6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 22 Apr 2008 00:16:04 +0900 Subject: gallium: Hash the fragment shader tokens, instead of pipe_shader_state. PS: pipe_shader_state should probably go away now that it is reduced to a single pointer. --- src/gallium/auxiliary/cso_cache/cso_context.c | 24 ++++++++++++++++++------ 1 file changed, 18 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 4541be8a5c..87995c80c3 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -38,6 +38,7 @@ #include "pipe/p_state.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "tgsi/util/tgsi_parse.h" #include "cso_cache/cso_context.h" #include "cso_cache/cso_cache.h" @@ -148,6 +149,7 @@ void cso_set_blend(struct cso_context *ctx, void *handle; if (cso_hash_iter_is_null(iter)) { + /* FIXME: handle OOM */ struct cso_blend *cso = MALLOC(sizeof(struct cso_blend)); cso->state = *templ; @@ -198,6 +200,7 @@ void cso_single_sampler(struct cso_context *ctx, (void*)templ); if (cso_hash_iter_is_null(iter)) { + /* FIXME: handle OOM */ struct cso_sampler *cso = MALLOC(sizeof(struct cso_sampler)); cso->state = *templ; @@ -333,6 +336,7 @@ void cso_set_depth_stencil_alpha(struct cso_context *ctx, void *handle; if (cso_hash_iter_is_null(iter)) { + /* FIXME: handle OOM */ struct cso_depth_stencil_alpha *cso = MALLOC(sizeof(struct cso_depth_stencil_alpha)); cso->state = *templ; @@ -381,6 +385,7 @@ void cso_set_rasterizer(struct cso_context *ctx, void *handle = NULL; if (cso_hash_iter_is_null(iter)) { + /* FIXME: handle OOM */ struct cso_rasterizer *cso = MALLOC(sizeof(struct cso_rasterizer)); cso->state = *templ; @@ -420,17 +425,23 @@ void cso_restore_rasterizer(struct cso_context *ctx) void cso_set_fragment_shader(struct cso_context *ctx, const struct pipe_shader_state *templ) { - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_shader_state)); + const struct tgsi_token *tokens = templ->tokens; + unsigned num_tokens = tgsi_num_tokens(tokens); + size_t tokens_size = num_tokens*sizeof(struct tgsi_token); + unsigned hash_key = cso_construct_key((void*)tokens, tokens_size); struct cso_hash_iter iter = cso_find_state_template(ctx->cache, - hash_key, CSO_FRAGMENT_SHADER, - (void*)templ); + hash_key, + CSO_FRAGMENT_SHADER, + (void*)tokens); void *handle = NULL; if (cso_hash_iter_is_null(iter)) { - struct cso_fragment_shader *cso = MALLOC(sizeof(struct cso_fragment_shader)); + /* FIXME: handle OOM */ + struct cso_fragment_shader *cso = MALLOC(sizeof(struct cso_fragment_shader) + tokens_size); + struct tgsi_token *cso_tokens = (struct tgsi_token *)((char *)cso + sizeof(*cso)); - cso->state = *templ; + memcpy(cso_tokens, tokens, tokens_size); + cso->state.tokens = cso_tokens; cso->data = ctx->pipe->create_fs_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_fs_state; cso->context = ctx->pipe; @@ -477,6 +488,7 @@ void cso_set_vertex_shader(struct cso_context *ctx, void *handle = NULL; if (cso_hash_iter_is_null(iter)) { + /* FIXME: handle OOM */ struct cso_vertex_shader *cso = MALLOC(sizeof(struct cso_vertex_shader)); cso->state = *templ; -- cgit v1.2.3 From a918a9c744f656c8bf2e3fd2841732e01a5ccefc Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 16:10:13 +0100 Subject: draw: consolidate all the passthrough line/tri/point funcs --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/SConscript | 1 + src/gallium/auxiliary/draw/draw_pipe.c | 63 ---------- src/gallium/auxiliary/draw/draw_pipe.h | 6 +- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 38 +++---- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 16 +-- src/gallium/auxiliary/draw/draw_pipe_cull.c | 18 +-- src/gallium/auxiliary/draw/draw_pipe_flatshade.c | 9 +- src/gallium/auxiliary/draw/draw_pipe_offset.c | 16 +-- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 25 +--- src/gallium/auxiliary/draw/draw_pipe_stipple.c | 16 +-- src/gallium/auxiliary/draw/draw_pipe_twoside.c | 19 +--- src/gallium/auxiliary/draw/draw_pipe_unfilled.c | 17 +-- src/gallium/auxiliary/draw/draw_pipe_util.c | 133 ++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_pipe_wide_line.c | 17 +-- src/gallium/auxiliary/draw/draw_pipe_wide_point.c | 23 +--- 16 files changed, 177 insertions(+), 241 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pipe_util.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 5289f2660a..bc6acfe458 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -16,6 +16,7 @@ C_SOURCES = \ draw_pipe_stipple.c \ draw_pipe_twoside.c \ draw_pipe_unfilled.c \ + draw_pipe_util.c \ draw_pipe_validate.c \ draw_pipe_vbuf.c \ draw_pipe_wide_line.c \ diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 91e6a35c5e..0b9852f633 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -15,6 +15,7 @@ draw = env.ConvenienceLibrary( 'draw_pipe_stipple.c', 'draw_pipe_twoside.c', 'draw_pipe_unfilled.c', + 'draw_pipe_util.c', 'draw_pipe_validate.c', 'draw_pipe_vbuf.c', 'draw_pipe_wide_line.c', diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 0c0840af0c..d0890203a5 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -108,40 +108,6 @@ void draw_pipeline_destroy( struct draw_context *draw ) -/* This is only used for temporary verts. - */ -#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) - - -/** - * Allocate space for temporary post-transform vertices, such as for clipping. - */ -void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) -{ - assert(!stage->tmp); - - stage->nr_tmps = nr; - - if (nr) { - ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); - unsigned i; - - stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); - } -} - - -void draw_free_temp_verts( struct draw_stage *stage ) -{ - if (stage->tmp) { - FREE( stage->tmp[0] ); - FREE( stage->tmp ); - stage->tmp = NULL; - } -} @@ -195,35 +161,6 @@ static void do_triangle( struct draw_context *draw, } -/* Reset vertex ids. This is basically a type of flush. - * - * Called only from draw_pipe_vbuf.c - */ -void draw_reset_vertex_ids(struct draw_context *draw) -{ - struct draw_stage *stage = draw->pipeline.first; - - while (stage) { - unsigned i; - - for (i = 0; i < stage->nr_tmps; i++) - stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; - - stage = stage->next; - } - - if (draw->pipeline.verts) - { - unsigned i; - char *verts = draw->pipeline.verts; - unsigned stride = draw->pipeline.vertex_stride; - - for (i = 0; i < draw->pipeline.vertex_count; i++) { - ((struct vertex_header *)verts)->vertex_id = UNDEFINED_VERTEX_ID; - verts += stride; - } - } -} /* Code to run the pipeline on a fairly arbitary collection of vertices. diff --git a/src/gallium/auxiliary/draw/draw_pipe.h b/src/gallium/auxiliary/draw/draw_pipe.h index f2749b34bb..2476abb2b2 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.h +++ b/src/gallium/auxiliary/draw/draw_pipe.h @@ -81,10 +81,14 @@ extern struct draw_stage *draw_validate_stage( struct draw_context *context ); extern void draw_free_temp_verts( struct draw_stage *stage ); -extern void draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); +extern boolean draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ); extern void draw_reset_vertex_ids( struct draw_context *draw ); +void draw_pipe_passthrough_tri(struct draw_stage *stage, struct prim_header *header); +void draw_pipe_passthrough_line(struct draw_stage *stage, struct prim_header *header); +void draw_pipe_passthrough_point(struct draw_stage *stage, struct prim_header *header); + /** diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index 24bc87d4f8..b27360170a 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -334,7 +334,7 @@ aa_transform_inst(struct tgsi_transform_context *ctx, * Generate the frag shader we'll use for drawing AA lines. * This will be the user's shader plus some texture/modulate instructions. */ -static void +static boolean generate_aaline_fs(struct aaline_stage *aaline) { const struct pipe_shader_state *orig_fs = &aaline->fs->state; @@ -346,6 +346,8 @@ generate_aaline_fs(struct aaline_stage *aaline) aaline_fs = *orig_fs; /* copy to init */ aaline_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); + if (aaline_fs.tokens == NULL) + return FALSE; memset(&transform, 0, sizeof(transform)); transform.colorOutput = -1; @@ -372,6 +374,7 @@ generate_aaline_fs(struct aaline_stage *aaline) = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); aaline->fs->generic_attrib = transform.maxGeneric + 1; + return TRUE; } @@ -469,13 +472,15 @@ aaline_create_sampler(struct aaline_stage *aaline) * When we're about to draw our first AA line in a batch, this function is * called to tell the driver to bind our modified fragment shader. */ -static void +static boolean bind_aaline_fragment_shader(struct aaline_stage *aaline) { - if (!aaline->fs->aaline_fs) { - generate_aaline_fs(aaline); - } + if (!aaline->fs->aaline_fs && + !generate_aaline_fs(aaline)) + return FALSE; + aaline->driver_bind_fs_state(aaline->pipe, aaline->fs->aaline_fs); + return TRUE; } @@ -487,20 +492,6 @@ aaline_stage( struct draw_stage *stage ) } -static void -passthrough_point(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->point(stage->next, header); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} - - /** * Draw a wide line by drawing a quad, using geometry which will * fullfill GL's antialiased line requirements. @@ -638,7 +629,10 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) /* * Bind (generate) our fragprog, sampler and texture */ - bind_aaline_fragment_shader(aaline); + if (!bind_aaline_fragment_shader(aaline)) { + stage->line = draw_pipe_passthrough_line; + return; + } /* update vertex attrib info */ aaline->tex_slot = draw->num_vs_outputs; @@ -721,9 +715,9 @@ draw_aaline_stage(struct draw_context *draw) aaline->stage.draw = draw; aaline->stage.next = NULL; - aaline->stage.point = passthrough_point; + aaline->stage.point = draw_pipe_passthrough_point; aaline->stage.line = aaline_first_line; - aaline->stage.tri = passthrough_tri; + aaline->stage.tri = draw_pipe_passthrough_tri; aaline->stage.flush = aaline_flush; aaline->stage.reset_stipple_counter = aaline_reset_stipple_counter; aaline->stage.destroy = aaline_destroy; diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 9f878f6c02..6d22d7ac48 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -546,18 +546,6 @@ aapoint_stage( struct draw_stage *stage ) } -static void -passthrough_line(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->line(stage->next, header); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} /** @@ -749,8 +737,8 @@ draw_aapoint_stage(struct draw_context *draw) aapoint->stage.draw = draw; aapoint->stage.next = NULL; aapoint->stage.point = aapoint_first_point; - aapoint->stage.line = passthrough_line; - aapoint->stage.tri = passthrough_tri; + aapoint->stage.line = draw_pipe_passthrough_line; + aapoint->stage.tri = draw_pipe_passthrough_tri; aapoint->stage.flush = aapoint_flush; aapoint->stage.reset_stipple_counter = aapoint_reset_stipple_counter; aapoint->stage.destroy = aapoint_destroy; diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c index c406f89d05..8c13f40b55 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_cull.c +++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c @@ -95,20 +95,6 @@ static void cull_first_tri( struct draw_stage *stage, -static void cull_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void cull_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - static void cull_flush( struct draw_stage *stage, unsigned flags ) { stage->tri = cull_first_tri; @@ -139,8 +125,8 @@ struct draw_stage *draw_cull_stage( struct draw_context *draw ) cull->stage.draw = draw; cull->stage.next = NULL; - cull->stage.point = cull_point; - cull->stage.line = cull_line; + cull->stage.point = draw_pipe_passthrough_point; + cull->stage.line = draw_pipe_passthrough_line; cull->stage.tri = cull_first_tri; cull->stage.flush = cull_flush; cull->stage.reset_stipple_counter = cull_reset_stipple_counter; diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c index bdb8b49dc4..2aeb309554 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c @@ -152,13 +152,6 @@ static void flatshade_line_1( struct draw_stage *stage, } -/* Flatshade point -- passthrough. - */ -static void flatshade_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} static void flatshade_init_state( struct draw_stage *stage ) @@ -236,7 +229,7 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) flatshade->stage.draw = draw; flatshade->stage.next = NULL; - flatshade->stage.point = flatshade_point; + flatshade->stage.point = draw_pipe_passthrough_point; flatshade->stage.line = flatshade_first_line; flatshade->stage.tri = flatshade_first_tri; flatshade->stage.flush = flatshade_flush; diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c index dbdece45bb..c1dc21cd32 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_offset.c +++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c @@ -129,18 +129,6 @@ static void offset_first_tri( struct draw_stage *stage, } -static void offset_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void offset_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} static void offset_flush( struct draw_stage *stage, @@ -175,8 +163,8 @@ struct draw_stage *draw_offset_stage( struct draw_context *draw ) offset->stage.draw = draw; offset->stage.next = NULL; - offset->stage.point = offset_point; - offset->stage.line = offset_line; + offset->stage.point = draw_pipe_passthrough_point; + offset->stage.line = draw_pipe_passthrough_line; offset->stage.tri = offset_first_tri; offset->stage.flush = offset_flush; offset->stage.reset_stipple_counter = offset_reset_stipple_counter; diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 4903ba2133..5686729cd3 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -473,25 +473,6 @@ pstip_stage( struct draw_stage *stage ) } -static void -passthrough_point(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->point(stage->next, header); -} - - -static void -passthrough_line(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->line(stage->next, header); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} @@ -523,7 +504,7 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures); /* now really draw first line */ - stage->tri = passthrough_tri; + stage->tri = draw_pipe_passthrough_tri; stage->tri(stage, header); } @@ -579,8 +560,8 @@ draw_pstip_stage(struct draw_context *draw) pstip->stage.draw = draw; pstip->stage.next = NULL; - pstip->stage.point = passthrough_point; - pstip->stage.line = passthrough_line; + pstip->stage.point = draw_pipe_passthrough_point; + pstip->stage.line = draw_pipe_passthrough_line; pstip->stage.tri = pstip_first_tri; pstip->stage.flush = pstip_flush; pstip->stage.reset_stipple_counter = pstip_reset_stipple_counter; diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c index 49429ee9e1..9cf5840cce 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c @@ -195,18 +195,6 @@ stipple_flush(struct draw_stage *stage, unsigned flags) } -static void -passthrough_point(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->point( stage->next, header ); -} - - -static void -passthrough_tri(struct draw_stage *stage, struct prim_header *header) -{ - stage->next->tri(stage->next, header); -} static void @@ -228,9 +216,9 @@ struct draw_stage *draw_stipple_stage( struct draw_context *draw ) stipple->stage.draw = draw; stipple->stage.next = NULL; - stipple->stage.point = passthrough_point; + stipple->stage.point = draw_pipe_passthrough_point; stipple->stage.line = stipple_first_line; - stipple->stage.tri = passthrough_tri; + stipple->stage.tri = draw_pipe_passthrough_tri; stipple->stage.reset_stipple_counter = reset_stipple_counter; stipple->stage.flush = stipple_flush; stipple->stage.destroy = stipple_destroy; diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c index 09a9d23d57..453fd3ac71 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_twoside.c +++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c @@ -99,21 +99,6 @@ static void twoside_tri( struct draw_stage *stage, } -static void twoside_line( struct draw_stage *stage, - struct prim_header *header ) -{ - /* pass-through */ - stage->next->line( stage->next, header ); -} - - -static void twoside_point( struct draw_stage *stage, - struct prim_header *header ) -{ - /* pass-through */ - stage->next->point( stage->next, header ); -} - static void twoside_first_tri( struct draw_stage *stage, struct prim_header *header ) @@ -192,8 +177,8 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) twoside->stage.draw = draw; twoside->stage.next = NULL; - twoside->stage.point = twoside_point; - twoside->stage.line = twoside_line; + twoside->stage.point = draw_pipe_passthrough_point; + twoside->stage.line = draw_pipe_passthrough_line; twoside->stage.tri = twoside_first_tri; twoside->stage.flush = twoside_flush; twoside->stage.reset_stipple_counter = twoside_reset_stipple_counter; diff --git a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c index 31e24f6a14..d4ddfec1b3 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c +++ b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c @@ -148,19 +148,6 @@ static void unfilled_first_tri( struct draw_stage *stage, } -static void unfilled_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line( stage->next, header ); -} - - -static void unfilled_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - static void unfilled_flush( struct draw_stage *stage, unsigned flags ) @@ -196,8 +183,8 @@ struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) unfilled->stage.draw = draw; unfilled->stage.next = NULL; unfilled->stage.tmp = NULL; - unfilled->stage.point = unfilled_point; - unfilled->stage.line = unfilled_line; + unfilled->stage.point = draw_pipe_passthrough_point; + unfilled->stage.line = draw_pipe_passthrough_line; unfilled->stage.tri = unfilled_first_tri; unfilled->stage.flush = unfilled_flush; unfilled->stage.reset_stipple_counter = unfilled_reset_stipple_counter; diff --git a/src/gallium/auxiliary/draw/draw_pipe_util.c b/src/gallium/auxiliary/draw/draw_pipe_util.c new file mode 100644 index 0000000000..e9821de976 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pipe_util.c @@ -0,0 +1,133 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw/draw_private.h" +#include "draw/draw_pipe.h" + + + +void +draw_pipe_passthrough_point(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->point(stage->next, header); +} + +void +draw_pipe_passthrough_line(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->line(stage->next, header); +} + +void +draw_pipe_passthrough_tri(struct draw_stage *stage, struct prim_header *header) +{ + stage->next->tri(stage->next, header); +} + + + + + +/* This is only used for temporary verts. + */ +#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) + + +/** + * Allocate space for temporary post-transform vertices, such as for clipping. + */ +boolean draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) +{ + unsigned i; + ubyte *store; + + assert(!stage->tmp); + + stage->tmp = NULL; + stage->nr_tmps = nr; + if (nr == 0) + return FALSE; + + store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); + if (store == NULL) + return FALSE; + + stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + + return TRUE; +} + + +void draw_free_temp_verts( struct draw_stage *stage ) +{ + if (stage->tmp) { + FREE( stage->tmp[0] ); + FREE( stage->tmp ); + stage->tmp = NULL; + } +} + + +/* Reset vertex ids. This is basically a type of flush. + * + * Called only from draw_pipe_vbuf.c + */ +void draw_reset_vertex_ids(struct draw_context *draw) +{ + struct draw_stage *stage = draw->pipeline.first; + + while (stage) { + unsigned i; + + for (i = 0; i < stage->nr_tmps; i++) + stage->tmp[i]->vertex_id = UNDEFINED_VERTEX_ID; + + stage = stage->next; + } + + if (draw->pipeline.verts) + { + unsigned i; + char *verts = draw->pipeline.verts; + unsigned stride = draw->pipeline.vertex_stride; + + for (i = 0; i < draw->pipeline.vertex_count; i++) { + ((struct vertex_header *)verts)->vertex_id = UNDEFINED_VERTEX_ID; + verts += stride; + } + } +} + diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c index 329b5d0fb0..452732e662 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c @@ -49,19 +49,6 @@ static INLINE struct wideline_stage *wideline_stage( struct draw_stage *stage ) } -static void wideline_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - - -static void wideline_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->tri(stage->next, header); -} - /** * Draw a wide line by drawing a quad (two triangles). @@ -180,9 +167,9 @@ struct draw_stage *draw_wide_line_stage( struct draw_context *draw ) wide->stage.draw = draw; wide->stage.next = NULL; - wide->stage.point = wideline_point; + wide->stage.point = draw_pipe_passthrough_point; wide->stage.line = wideline_line; - wide->stage.tri = wideline_tri; + wide->stage.tri = draw_pipe_passthrough_point; wide->stage.flush = wideline_flush; wide->stage.reset_stipple_counter = wideline_reset_stipple_counter; wide->stage.destroy = wideline_destroy; diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index 7a439178a0..8101340680 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -61,23 +61,6 @@ widepoint_stage( struct draw_stage *stage ) } -static void passthrough_point( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->point( stage->next, header ); -} - -static void widepoint_line( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->line(stage->next, header); -} - -static void widepoint_tri( struct draw_stage *stage, - struct prim_header *header ) -{ - stage->next->tri(stage->next, header); -} /** @@ -209,7 +192,7 @@ static void widepoint_first_point( struct draw_stage *stage, stage->point = widepoint_point; } else { - stage->point = passthrough_point; + stage->point = draw_pipe_passthrough_point; } if (draw->rasterizer->point_sprite) { @@ -272,8 +255,8 @@ struct draw_stage *draw_wide_point_stage( struct draw_context *draw ) wide->stage.draw = draw; wide->stage.next = NULL; wide->stage.point = widepoint_first_point; - wide->stage.line = widepoint_line; - wide->stage.tri = widepoint_tri; + wide->stage.line = draw_pipe_passthrough_line; + wide->stage.tri = draw_pipe_passthrough_tri; wide->stage.flush = widepoint_flush; wide->stage.reset_stipple_counter = widepoint_reset_stipple_counter; wide->stage.destroy = widepoint_destroy; -- cgit v1.2.3 From 69ecc2a577dc45451d56cee3e41cb6e7e542b097 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 16:38:09 +0100 Subject: draw: propogate errors out of aaline stage --- src/gallium/auxiliary/draw/draw_context.h | 2 +- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 58 ++++++++++++++++++++++----- 2 files changed, 48 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index 68e2efb865..6171c75cb4 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -73,7 +73,7 @@ void draw_enable_line_stipple(struct draw_context *draw, boolean enable); void draw_enable_point_sprites(struct draw_context *draw, boolean enable); -void +boolean draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe); void diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index b27360170a..ae1abf91c2 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -338,7 +338,6 @@ static boolean generate_aaline_fs(struct aaline_stage *aaline) { const struct pipe_shader_state *orig_fs = &aaline->fs->state; - //struct draw_context *draw = aaline->stage.draw; struct pipe_shader_state aaline_fs; struct aa_transform_context transform; @@ -372,6 +371,8 @@ generate_aaline_fs(struct aaline_stage *aaline) aaline->fs->aaline_fs = aaline->driver_create_fs_state(aaline->pipe, &aaline_fs); + if (aaline->fs->aaline_fs == NULL) + return FALSE; aaline->fs->generic_attrib = transform.maxGeneric + 1; return TRUE; @@ -381,7 +382,7 @@ generate_aaline_fs(struct aaline_stage *aaline) /** * Create the texture map we'll use for antialiasing the lines. */ -static void +static boolean aaline_create_texture(struct aaline_stage *aaline) { struct pipe_context *pipe = aaline->pipe; @@ -399,6 +400,8 @@ aaline_create_texture(struct aaline_stage *aaline) texTemp.cpp = 1; aaline->texture = screen->texture_create(screen, &texTemp); + if (!aaline->texture) + return FALSE; /* Fill in mipmap images. * Basically each level is solid opaque, except for the outermost @@ -414,6 +417,8 @@ aaline_create_texture(struct aaline_stage *aaline) surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0); data = pipe_surface_map(surface); + if (data == NULL) + return FALSE; for (i = 0; i < size; i++) { for (j = 0; j < size; j++) { @@ -439,6 +444,7 @@ aaline_create_texture(struct aaline_stage *aaline) pipe_surface_reference(&surface, NULL); pipe->texture_update(pipe, aaline->texture, 0, (1 << level)); } + return TRUE; } @@ -447,7 +453,7 @@ aaline_create_texture(struct aaline_stage *aaline) * By using a mipmapped texture, we don't have to generate a different * texture image for each line size. */ -static void +static boolean aaline_create_sampler(struct aaline_stage *aaline) { struct pipe_sampler_state sampler; @@ -465,6 +471,10 @@ aaline_create_sampler(struct aaline_stage *aaline) sampler.max_lod = MAX_TEXTURE_LEVEL; aaline->sampler_cso = pipe->create_sampler_state(pipe, &sampler); + if (aaline->sampler_cso == NULL) + return FALSE; + + return TRUE; } @@ -696,9 +706,11 @@ aaline_destroy(struct draw_stage *stage) { struct aaline_stage *aaline = aaline_stage(stage); - aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso); + if (aaline->sampler_cso) + aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso); - pipe_texture_release(&aaline->texture); + if (aaline->texture) + pipe_texture_release(&aaline->texture); draw_free_temp_verts( stage ); @@ -710,8 +722,11 @@ static struct aaline_stage * draw_aaline_stage(struct draw_context *draw) { struct aaline_stage *aaline = CALLOC_STRUCT(aaline_stage); + if (aaline == NULL) + return NULL; - draw_alloc_temp_verts( &aaline->stage, 8 ); + if (draw_alloc_temp_verts( &aaline->stage, 8 )) + goto fail; aaline->stage.draw = draw; aaline->stage.next = NULL; @@ -723,6 +738,12 @@ draw_aaline_stage(struct draw_context *draw) aaline->stage.destroy = aaline_destroy; return aaline; + + fail: + if (aaline) + aaline_destroy(&aaline->stage); + + return NULL; } @@ -816,7 +837,7 @@ aaline_set_sampler_textures(struct pipe_context *pipe, * into the draw module's pipeline. This will not be used if the * hardware has native support for AA lines. */ -void +boolean draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) { struct aaline_stage *aaline; @@ -827,14 +848,17 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) * Create / install AA line drawing / prim stage */ aaline = draw_aaline_stage( draw ); - assert(aaline); - draw->pipeline.aaline = &aaline->stage; + if (!aaline) + goto fail; aaline->pipe = pipe; /* create special texture, sampler state */ - aaline_create_texture(aaline); - aaline_create_sampler(aaline); + if (!aaline_create_texture(aaline)) + goto fail; + + if (!aaline_create_sampler(aaline)) + goto fail; /* save original driver functions */ aaline->driver_create_fs_state = pipe->create_fs_state; @@ -851,4 +875,16 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe) pipe->bind_sampler_states = aaline_bind_sampler_states; pipe->set_sampler_textures = aaline_set_sampler_textures; + + /* Install once everything is known to be OK: + */ + draw->pipeline.aaline = &aaline->stage; + + return TRUE; + + fail: + if (aaline) + aaline->stage.destroy( &aaline->stage ); + + return FALSE; } -- cgit v1.2.3 From 0cd90a917d289363a3edb5cfa40c391eb07aa97c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 16:45:41 +0100 Subject: draw: propogate errors out of aapoint stage --- src/gallium/auxiliary/draw/draw_context.h | 2 +- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 46 ++++++++++++++++++++------ 3 files changed, 38 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index 6171c75cb4..197235b2d8 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -76,7 +76,7 @@ void draw_enable_point_sprites(struct draw_context *draw, boolean enable); boolean draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe); -void +boolean draw_install_aapoint_stage(struct draw_context *draw, struct pipe_context *pipe); void diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index ae1abf91c2..bc514d40f0 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -725,7 +725,7 @@ draw_aaline_stage(struct draw_context *draw) if (aaline == NULL) return NULL; - if (draw_alloc_temp_verts( &aaline->stage, 8 )) + if (!draw_alloc_temp_verts( &aaline->stage, 8 )) goto fail; aaline->stage.draw = draw; diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 6d22d7ac48..6d689c36de 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -484,7 +484,7 @@ aa_transform_inst(struct tgsi_transform_context *ctx, * Generate the frag shader we'll use for drawing AA lines. * This will be the user's shader plus some texture/modulate instructions. */ -static void +static boolean generate_aapoint_fs(struct aapoint_stage *aapoint) { const struct pipe_shader_state *orig_fs = &aapoint->fs->state; @@ -495,6 +495,8 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) aapoint_fs = *orig_fs; /* copy to init */ aapoint_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); + if (aapoint_fs.tokens == NULL) + return FALSE; memset(&transform, 0, sizeof(transform)); transform.colorOutput = -1; @@ -519,8 +521,12 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) aapoint->fs->aapoint_fs = aapoint->driver_create_fs_state(aapoint->pipe, &aapoint_fs); + if (aapoint->fs->aapoint_fs == NULL) + return FALSE; aapoint->fs->generic_attrib = transform.maxGeneric + 1; + + return TRUE; } @@ -528,13 +534,15 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) * When we're about to draw our first AA line in a batch, this function is * called to tell the driver to bind our modified fragment shader. */ -static void +static boolean bind_aapoint_fragment_shader(struct aapoint_stage *aapoint) { - if (!aapoint->fs->aapoint_fs) { - generate_aapoint_fs(aapoint); - } + if (!aapoint->fs->aapoint_fs && + !generate_aapoint_fs(aapoint)) + return FALSE; + aapoint->driver_bind_fs_state(aapoint->pipe, aapoint->fs->aapoint_fs); + return TRUE; } @@ -731,8 +739,11 @@ static struct aapoint_stage * draw_aapoint_stage(struct draw_context *draw) { struct aapoint_stage *aapoint = CALLOC_STRUCT(aapoint_stage); + if (aapoint == NULL) + goto fail; - draw_alloc_temp_verts( &aapoint->stage, 4 ); + if (!draw_alloc_temp_verts( &aapoint->stage, 4 )) + goto fail; aapoint->stage.draw = draw; aapoint->stage.next = NULL; @@ -744,6 +755,13 @@ draw_aapoint_stage(struct draw_context *draw) aapoint->stage.destroy = aapoint_destroy; return aapoint; + + fail: + if (aapoint) + aapoint_destroy(&aapoint->stage); + + return NULL; + } @@ -806,20 +824,19 @@ aapoint_delete_fs_state(struct pipe_context *pipe, void *fs) * into the draw module's pipeline. This will not be used if the * hardware has native support for AA points. */ -void +boolean draw_install_aapoint_stage(struct draw_context *draw, struct pipe_context *pipe) { struct aapoint_stage *aapoint; - pipe->draw = (void *) draw; /* * Create / install AA point drawing / prim stage */ aapoint = draw_aapoint_stage( draw ); - assert(aapoint); - draw->pipeline.aapoint = &aapoint->stage; + if (aapoint == NULL) + goto fail; aapoint->pipe = pipe; @@ -832,4 +849,13 @@ draw_install_aapoint_stage(struct draw_context *draw, pipe->create_fs_state = aapoint_create_fs_state; pipe->bind_fs_state = aapoint_bind_fs_state; pipe->delete_fs_state = aapoint_delete_fs_state; + + pipe->draw = (void *) draw; + draw->pipeline.aapoint = &aapoint->stage; + + fail: + if (aapoint) + aapoint->stage.destroy( &aapoint->stage ); + + return FALSE; } -- cgit v1.2.3 From 0d4ece4c5a243dc4b684331bad49f220311e5520 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 17:03:37 +0100 Subject: draw: propogate lots of errors --- src/gallium/auxiliary/draw/draw_context.h | 2 +- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 10 ++++---- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 14 ++++++----- src/gallium/auxiliary/draw/draw_pipe_clip.c | 11 ++++++++- src/gallium/auxiliary/draw/draw_pipe_cull.c | 11 ++++++++- src/gallium/auxiliary/draw/draw_pipe_flatshade.c | 11 ++++++++- src/gallium/auxiliary/draw/draw_pipe_offset.c | 8 +++++++ src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 14 +++++++++-- src/gallium/auxiliary/draw/draw_pipe_twoside.c | 11 ++++++++- src/gallium/auxiliary/draw/draw_pipe_unfilled.c | 11 ++++++++- src/gallium/auxiliary/draw/draw_pipe_util.c | 28 ++++++++++++---------- src/gallium/auxiliary/draw/draw_pipe_validate.c | 2 ++ src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 5 ++-- src/gallium/auxiliary/draw/draw_pipe_wide_point.c | 11 ++++++++- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 2 ++ .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 2 ++ src/gallium/auxiliary/draw/draw_pt_vcache.c | 2 ++ 17 files changed, 120 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index 197235b2d8..2742001698 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -79,7 +79,7 @@ draw_install_aaline_stage(struct draw_context *draw, struct pipe_context *pipe); boolean draw_install_aapoint_stage(struct draw_context *draw, struct pipe_context *pipe); -void +boolean draw_install_pstipple_stage(struct draw_context *draw, struct pipe_context *pipe); diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index bc514d40f0..fbb5b45b40 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -765,13 +765,13 @@ aaline_create_fs_state(struct pipe_context *pipe, { struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); struct aaline_fragment_shader *aafs = CALLOC_STRUCT(aaline_fragment_shader); + if (aafs == NULL) + return NULL; - if (aafs) { - aafs->state = *fs; + aafs->state = *fs; - /* pass-through */ - aafs->driver_fs = aaline->driver_create_fs_state(aaline->pipe, fs); - } + /* pass-through */ + aafs->driver_fs = aaline->driver_create_fs_state(aaline->pipe, fs); return aafs; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 6d689c36de..ac0aa4cd7c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -783,13 +783,13 @@ aapoint_create_fs_state(struct pipe_context *pipe, { struct aapoint_stage *aapoint = aapoint_stage_from_pipe(pipe); struct aapoint_fragment_shader *aafs = CALLOC_STRUCT(aapoint_fragment_shader); + if (aafs == NULL) + return NULL; - if (aafs) { - aafs->state = *fs; + aafs->state = *fs; - /* pass-through */ - aafs->driver_fs = aapoint->driver_create_fs_state(aapoint->pipe, fs); - } + /* pass-through */ + aafs->driver_fs = aapoint->driver_create_fs_state(aapoint->pipe, fs); return aafs; } @@ -830,6 +830,7 @@ draw_install_aapoint_stage(struct draw_context *draw, { struct aapoint_stage *aapoint; + pipe->draw = (void *) draw; /* * Create / install AA point drawing / prim stage @@ -850,9 +851,10 @@ draw_install_aapoint_stage(struct draw_context *draw, pipe->bind_fs_state = aapoint_bind_fs_state; pipe->delete_fs_state = aapoint_delete_fs_state; - pipe->draw = (void *) draw; draw->pipeline.aapoint = &aapoint->stage; + return TRUE; + fail: if (aapoint) aapoint->stage.destroy( &aapoint->stage ); diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c index 6780f275d9..21216addea 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_clip.c +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -493,8 +493,11 @@ static void clip_destroy( struct draw_stage *stage ) struct draw_stage *draw_clip_stage( struct draw_context *draw ) { struct clipper *clipper = CALLOC_STRUCT(clipper); + if (clipper == NULL) + goto fail; - draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 ); + if (!draw_alloc_temp_verts( &clipper->stage, MAX_CLIPPED_VERTICES+1 )) + goto fail; clipper->stage.draw = draw; clipper->stage.point = clip_point; @@ -507,4 +510,10 @@ struct draw_stage *draw_clip_stage( struct draw_context *draw ) clipper->plane = draw->plane; return &clipper->stage; + + fail: + if (clipper) + clipper->stage.destroy( &clipper->stage ); + + return NULL; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c index 8c13f40b55..87aaf1f85b 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_cull.c +++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c @@ -120,8 +120,11 @@ static void cull_destroy( struct draw_stage *stage ) struct draw_stage *draw_cull_stage( struct draw_context *draw ) { struct cull_stage *cull = CALLOC_STRUCT(cull_stage); + if (cull == NULL) + goto fail; - draw_alloc_temp_verts( &cull->stage, 0 ); + if (!draw_alloc_temp_verts( &cull->stage, 0 )) + goto fail; cull->stage.draw = draw; cull->stage.next = NULL; @@ -133,4 +136,10 @@ struct draw_stage *draw_cull_stage( struct draw_context *draw ) cull->stage.destroy = cull_destroy; return &cull->stage; + + fail: + if (cull) + cull->stage.destroy( &cull->stage ); + + return NULL; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c index 2aeb309554..205000cbea 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c @@ -224,8 +224,11 @@ static void flatshade_destroy( struct draw_stage *stage ) struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) { struct flat_stage *flatshade = CALLOC_STRUCT(flat_stage); + if (flatshade == NULL) + goto fail; - draw_alloc_temp_verts( &flatshade->stage, 2 ); + if (!draw_alloc_temp_verts( &flatshade->stage, 2 )) + goto fail; flatshade->stage.draw = draw; flatshade->stage.next = NULL; @@ -237,6 +240,12 @@ struct draw_stage *draw_flatshade_stage( struct draw_context *draw ) flatshade->stage.destroy = flatshade_destroy; return &flatshade->stage; + + fail: + if (flatshade) + flatshade->stage.destroy( &flatshade->stage ); + + return NULL; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c index c1dc21cd32..ffec85ccdd 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_offset.c +++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c @@ -158,6 +158,8 @@ static void offset_destroy( struct draw_stage *stage ) struct draw_stage *draw_offset_stage( struct draw_context *draw ) { struct offset_stage *offset = CALLOC_STRUCT(offset_stage); + if (offset == NULL) + goto fail; draw_alloc_temp_verts( &offset->stage, 3 ); @@ -171,4 +173,10 @@ struct draw_stage *draw_offset_stage( struct draw_context *draw ) offset->stage.destroy = offset_destroy; return &offset->stage; + + fail: + if (offset) + offset->stage.destroy( &offset->stage ); + + return NULL; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 5686729cd3..ac08aa10ce 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -686,7 +686,7 @@ pstip_set_polygon_stipple(struct pipe_context *pipe, * into the draw module's pipeline. This will not be used if the * hardware has native support for AA lines. */ -void +boolean draw_install_pstipple_stage(struct draw_context *draw, struct pipe_context *pipe) { @@ -698,7 +698,9 @@ draw_install_pstipple_stage(struct draw_context *draw, * Create / install AA line drawing / prim stage */ pstip = draw_pstip_stage( draw ); - assert(pstip); + if (pstip == NULL) + goto fail; + draw->pipeline.pstipple = &pstip->stage; pstip->pipe = pipe; @@ -724,4 +726,12 @@ draw_install_pstipple_stage(struct draw_context *draw, pipe->bind_sampler_states = pstip_bind_sampler_states; pipe->set_sampler_textures = pstip_set_sampler_textures; pipe->set_polygon_stipple = pstip_set_polygon_stipple; + + return TRUE; + + fail: + if (pstip) + pstip->stage.destroy( &pstip->stage ); + + return FALSE; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c index 453fd3ac71..5910dccc43 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_twoside.c +++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c @@ -172,8 +172,11 @@ static void twoside_destroy( struct draw_stage *stage ) struct draw_stage *draw_twoside_stage( struct draw_context *draw ) { struct twoside_stage *twoside = CALLOC_STRUCT(twoside_stage); + if (twoside == NULL) + goto fail; - draw_alloc_temp_verts( &twoside->stage, 3 ); + if (!draw_alloc_temp_verts( &twoside->stage, 3 )) + goto fail; twoside->stage.draw = draw; twoside->stage.next = NULL; @@ -185,4 +188,10 @@ struct draw_stage *draw_twoside_stage( struct draw_context *draw ) twoside->stage.destroy = twoside_destroy; return &twoside->stage; + + fail: + if (twoside) + twoside->stage.destroy( &twoside->stage ); + + return NULL; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c index d4ddfec1b3..eeb2bc43f9 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c +++ b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c @@ -177,8 +177,11 @@ static void unfilled_destroy( struct draw_stage *stage ) struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) { struct unfilled_stage *unfilled = CALLOC_STRUCT(unfilled_stage); + if (unfilled == NULL) + goto fail; - draw_alloc_temp_verts( &unfilled->stage, 0 ); + if (!draw_alloc_temp_verts( &unfilled->stage, 0 )) + goto fail; unfilled->stage.draw = draw; unfilled->stage.next = NULL; @@ -191,4 +194,10 @@ struct draw_stage *draw_unfilled_stage( struct draw_context *draw ) unfilled->stage.destroy = unfilled_destroy; return &unfilled->stage; + + fail: + if (unfilled) + unfilled->stage.destroy( &unfilled->stage ); + + return NULL; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_util.c b/src/gallium/auxiliary/draw/draw_pipe_util.c index e9821de976..04438f4dd0 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_util.c +++ b/src/gallium/auxiliary/draw/draw_pipe_util.c @@ -68,24 +68,28 @@ draw_pipe_passthrough_tri(struct draw_stage *stage, struct prim_header *header) */ boolean draw_alloc_temp_verts( struct draw_stage *stage, unsigned nr ) { - unsigned i; - ubyte *store; - assert(!stage->tmp); stage->tmp = NULL; stage->nr_tmps = nr; - if (nr == 0) - return FALSE; - store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); - if (store == NULL) - return FALSE; + if (nr != 0) + { + unsigned i; + ubyte *store = (ubyte *) MALLOC( MAX_VERTEX_SIZE * nr ); + + if (store == NULL) + return FALSE; - stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); - - for (i = 0; i < nr; i++) - stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + stage->tmp = (struct vertex_header **) MALLOC( sizeof(struct vertex_header *) * nr ); + if (stage->tmp == NULL) { + FREE(store); + return FALSE; + } + + for (i = 0; i < nr; i++) + stage->tmp[i] = (struct vertex_header *)(store + i * MAX_VERTEX_SIZE); + } return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c index ffddc2f62c..a2e0812c89 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_validate.c +++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c @@ -299,6 +299,8 @@ static void validate_destroy( struct draw_stage *stage ) struct draw_stage *draw_validate_stage( struct draw_context *draw ) { struct draw_stage *stage = CALLOC_STRUCT(draw_stage); + if (stage == NULL) + return NULL; stage->draw = draw; stage->next = NULL; diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index c5810febe0..afd5f5544d 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -443,8 +443,7 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, struct vbuf_render *render ) { struct vbuf_stage *vbuf = CALLOC_STRUCT(vbuf_stage); - - if(!vbuf) + if (vbuf == NULL) goto fail; vbuf->stage.draw = draw; @@ -461,7 +460,7 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, vbuf->indices = (ushort *) align_malloc( vbuf->max_indices * sizeof(vbuf->indices[0]), 16 ); - if(!vbuf->indices) + if (!vbuf->indices) goto fail; vbuf->vertices = NULL; diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index 8101340680..ed08573382 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -249,8 +249,11 @@ static void widepoint_destroy( struct draw_stage *stage ) struct draw_stage *draw_wide_point_stage( struct draw_context *draw ) { struct widepoint_stage *wide = CALLOC_STRUCT(widepoint_stage); + if (wide == NULL) + goto fail; - draw_alloc_temp_verts( &wide->stage, 4 ); + if (!draw_alloc_temp_verts( &wide->stage, 4 )) + goto fail; wide->stage.draw = draw; wide->stage.next = NULL; @@ -262,4 +265,10 @@ struct draw_stage *draw_wide_point_stage( struct draw_context *draw ) wide->stage.destroy = widepoint_destroy; return &wide->stage; + + fail: + if (wide) + wide->stage.destroy( &wide->stage ); + + return NULL; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 002ced2186..7735173042 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -271,6 +271,8 @@ static void fetch_emit_destroy( struct draw_pt_middle_end *middle ) struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ) { struct fetch_emit_middle_end *fetch_emit = CALLOC_STRUCT( fetch_emit_middle_end ); + if (fetch_emit == NULL) + return NULL; fetch_emit->base.prepare = fetch_emit_prepare; fetch_emit->base.run = fetch_emit_run; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 881e47d59d..98a2cb45e4 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -114,6 +114,8 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, (struct vertex_header *)MALLOC(fpme->vertex_size * fetch_count); if (!pipeline_verts) { + /* Not much we can do here - just skip the rendering. + */ assert(0); return; } diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index b61bb50664..3cc2941f96 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -487,6 +487,8 @@ static void vcache_destroy( struct draw_pt_front_end *frontend ) struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ) { struct vcache_frontend *vcache = CALLOC_STRUCT( vcache_frontend ); + if (vcache == NULL) + return NULL; vcache->base.prepare = vcache_prepare; vcache->base.run = NULL; -- cgit v1.2.3 From 30b4dc29091347252bc61d3be9370db0a45c16c3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 17:17:27 +0100 Subject: draw: more propogation -- pstipple stage. --- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 1 + src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 43 ++++++++++++++++++------- 2 files changed, 33 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index fbb5b45b40..7e5f8bd281 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -641,6 +641,7 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) */ if (!bind_aaline_fragment_shader(aaline)) { stage->line = draw_pipe_passthrough_line; + stage->line(stage, header); return; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index ac08aa10ce..61bdd29aa0 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -320,7 +320,7 @@ pstip_transform_inst(struct tgsi_transform_context *ctx, * Generate the frag shader we'll use for doing polygon stipple. * This will be the user's shader prefixed with a TEX and KIL instruction. */ -static void +static boolean generate_pstip_fs(struct pstip_stage *pstip) { const struct pipe_shader_state *orig_fs = &pstip->fs->state; @@ -332,6 +332,8 @@ generate_pstip_fs(struct pstip_stage *pstip) pstip_fs = *orig_fs; /* copy to init */ pstip_fs.tokens = MALLOC(sizeof(struct tgsi_token) * MAX); + if (pstip_fs.tokens == NULL) + return FALSE; memset(&transform, 0, sizeof(transform)); transform.wincoordInput = -1; @@ -355,6 +357,8 @@ generate_pstip_fs(struct pstip_stage *pstip) assert(pstip->fs->sampler_unit < PIPE_MAX_SAMPLERS); pstip->fs->pstip_fs = pstip->driver_create_fs_state(pstip->pipe, &pstip_fs); + + return TRUE; } @@ -404,7 +408,7 @@ pstip_update_texture(struct pstip_stage *pstip) /** * Create the texture map we'll use for stippling. */ -static void +static boolean pstip_create_texture(struct pstip_stage *pstip) { struct pipe_context *pipe = pstip->pipe; @@ -421,7 +425,10 @@ pstip_create_texture(struct pstip_stage *pstip) texTemp.cpp = 1; pstip->texture = screen->texture_create(screen, &texTemp); - assert(pstip->texture->refcount == 1); + if (pstip->texture == NULL) + return FALSE; + + return TRUE; } @@ -430,7 +437,7 @@ pstip_create_texture(struct pstip_stage *pstip) * By using a mipmapped texture, we don't have to generate a different * texture image for each line size. */ -static void +static boolean pstip_create_sampler(struct pstip_stage *pstip) { struct pipe_sampler_state sampler; @@ -448,6 +455,10 @@ pstip_create_sampler(struct pstip_stage *pstip) sampler.max_lod = 0.0f; pstip->sampler_cso = pipe->create_sampler_state(pipe, &sampler); + if (pstip->sampler_cso == NULL) + return FALSE; + + return TRUE; } @@ -455,13 +466,15 @@ pstip_create_sampler(struct pstip_stage *pstip) * When we're about to draw our first AA line in a batch, this function is * called to tell the driver to bind our modified fragment shader. */ -static void +static boolean bind_pstip_fragment_shader(struct pstip_stage *pstip) { - if (!pstip->fs->pstip_fs) { - generate_pstip_fs(pstip); - } + if (!pstip->fs->pstip_fs || + !generate_pstip_fs(pstip)) + return FALSE; + pstip->driver_bind_fs_state(pstip->pipe, pstip->fs->pstip_fs); + return TRUE; } @@ -486,7 +499,12 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) assert(stage->draw->rasterizer->poly_stipple_enable); /* bind our fragprog */ - bind_pstip_fragment_shader(pstip); + if (!bind_pstip_fragment_shader(pstip)) { + stage->tri = draw_pipe_passthrough_tri; + stage->tri(stage, header); + return; + } + /* how many samplers? */ /* we'll use sampler/texture[pstip->sampler_unit] for the stipple */ @@ -706,8 +724,11 @@ draw_install_pstipple_stage(struct draw_context *draw, pstip->pipe = pipe; /* create special texture, sampler state */ - pstip_create_texture(pstip); - pstip_create_sampler(pstip); + if (!pstip_create_texture(pstip)) + goto fail; + + if (!pstip_create_sampler(pstip)) + goto fail; /* save original driver functions */ pstip->driver_create_fs_state = pipe->create_fs_state; -- cgit v1.2.3 From 0824fb1d6afc651c0ab814e96f08326c706de216 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 21 Apr 2008 12:42:37 -0400 Subject: actually write the results --- src/gallium/auxiliary/draw/draw_vs_llvm.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 2ebf05526a..dcada66514 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -103,15 +103,16 @@ vs_llvm_run_linear( struct draw_vertex_shader *base, /* Unswizzle all output results */ - for (slot = 1; slot < base->info.num_inputs; slot++) { - output[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; - - output = (float (*)[4])((char *)output + output_stride); + for (j = 0; j < max_vertices; j++) { + for (slot = 0; slot < base->info.num_outputs; slot++) { + output[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + output = (float (*)[4])((char *)output + output_stride); } - } + } } -- cgit v1.2.3 From 65efe807b9067aa07b382e3c4d9cea6222c5fc6b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Mon, 21 Apr 2008 17:51:39 +0100 Subject: gallium: Use CSO cache for shaders. --- src/mesa/state_tracker/st_atom_shader.c | 4 ++-- src/mesa/state_tracker/st_cb_bitmap.c | 11 ++++++----- src/mesa/state_tracker/st_cb_blit.c | 5 ----- src/mesa/state_tracker/st_cb_clear.c | 12 ++++++------ src/mesa/state_tracker/st_cb_drawpixels.c | 12 ++++++------ src/mesa/state_tracker/st_gen_mipmap.c | 6 ------ 6 files changed, 20 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index ec39026eb3..4a641a4a73 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -273,8 +273,8 @@ update_linkage( struct st_context *st ) st->vp = stvp; st->fp = stfp; - st->pipe->bind_vs_state(st->pipe, stvp->driver_shader); - st->pipe->bind_fs_state(st->pipe, stfp->driver_shader); + cso_set_vertex_shader(st->cso_context, stvp->driver_shader); + cso_set_fragment_shader(st->cso_context, stfp->driver_shader); st->vertex_result_to_slot = xvp->output_to_slot; } diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 505a13cc2b..1b863143e0 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -438,16 +438,18 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_save_samplers(cso); cso_save_sampler_textures(cso); cso_save_viewport(cso); + cso_save_fragment_shader(cso); + cso_save_vertex_shader(cso); /* rasterizer state: just scissor */ st->bitmap.rasterizer.scissor = ctx->Scissor.Enabled; cso_set_rasterizer(cso, &st->bitmap.rasterizer); /* fragment shader state: TEX lookup program */ - pipe->bind_fs_state(pipe, stfp->driver_shader); + cso_set_fragment_shader(cso, stfp->driver_shader); /* vertex shader state: position + texcoord pass-through */ - pipe->bind_vs_state(pipe, st->bitmap.vs); + cso_set_vertex_shader(cso, st->bitmap.vs); /* sampler / texture state */ cso_single_sampler(cso, 0, &st->bitmap.sampler); @@ -488,9 +490,8 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_restore_samplers(cso); cso_restore_sampler_textures(cso); cso_restore_viewport(cso); - /* shaders don't go through cso yet */ - pipe->bind_fs_state(pipe, st->fp->driver_shader); - pipe->bind_vs_state(pipe, st->vp->driver_shader); + cso_save_fragment_shader(cso); + cso_save_vertex_shader(cso); } diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c index 63211d8b66..0533d9460f 100644 --- a/src/mesa/state_tracker/st_cb_blit.c +++ b/src/mesa/state_tracker/st_cb_blit.c @@ -71,7 +71,6 @@ st_BlitFramebuffer(GLcontext *ctx, GLbitfield mask, GLenum filter) { struct st_context *st = ctx->st; - struct pipe_context *pipe = st->pipe; const uint pFilter = ((filter == GL_NEAREST) ? PIPE_TEX_MIPFILTER_NEAREST @@ -100,10 +99,6 @@ st_BlitFramebuffer(GLcontext *ctx, 0.0, pFilter); } - - /* shaders don't go through CSO yet */ - pipe->bind_fs_state(pipe, st->fp->driver_shader); - pipe->bind_vs_state(pipe, st->vp->driver_shader); } diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 6469511c6f..bb27faad21 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -195,7 +195,6 @@ clear_with_quad(GLcontext *ctx, GLboolean color, GLboolean depth, GLboolean stencil) { struct st_context *st = ctx->st; - struct pipe_context *pipe = st->pipe; const GLfloat x0 = ctx->DrawBuffer->_Xmin; const GLfloat x1 = ctx->DrawBuffer->_Xmax; GLfloat y0, y1; @@ -222,6 +221,8 @@ clear_with_quad(GLcontext *ctx, cso_save_depth_stencil_alpha(st->cso_context); cso_save_rasterizer(st->cso_context); cso_save_viewport(st->cso_context); + cso_save_fragment_shader(st->cso_context); + cso_save_vertex_shader(st->cso_context); /* blend state: RGBA masking */ { @@ -273,8 +274,8 @@ clear_with_quad(GLcontext *ctx, cso_set_rasterizer(st->cso_context, &st->clear.raster); cso_set_viewport(st->cso_context, &st->clear.viewport); - pipe->bind_fs_state(pipe, st->clear.fs); - pipe->bind_vs_state(pipe, st->clear.vs); + cso_set_fragment_shader(st->cso_context, st->clear.fs); + cso_set_vertex_shader(st->cso_context, st->clear.vs); /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); @@ -284,9 +285,8 @@ clear_with_quad(GLcontext *ctx, cso_restore_depth_stencil_alpha(st->cso_context); cso_restore_rasterizer(st->cso_context); cso_restore_viewport(st->cso_context); - /* these don't go through cso yet */ - pipe->bind_fs_state(pipe, st->fp->driver_shader); - pipe->bind_vs_state(pipe, st->vp->driver_shader); + cso_restore_fragment_shader(st->cso_context); + cso_restore_vertex_shader(st->cso_context); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 5fb27857a7..75261c3350 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -532,6 +532,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_save_viewport(cso); cso_save_samplers(cso); cso_save_sampler_textures(cso); + cso_save_fragment_shader(cso); + cso_save_vertex_shader(cso); /* rasterizer state: just scissor */ { @@ -543,10 +545,10 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, } /* fragment shader state: TEX lookup program */ - pipe->bind_fs_state(pipe, stfp->driver_shader); + cso_set_fragment_shader(cso, stfp->driver_shader); /* vertex shader state: position + texcoord pass-through */ - pipe->bind_vs_state(pipe, stvp->driver_shader); + cso_set_vertex_shader(cso, stvp->driver_shader); /* texture sampling state: */ @@ -615,10 +617,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_restore_viewport(cso); cso_restore_samplers(cso); cso_restore_sampler_textures(cso); - - /* shaders don't go through cso yet */ - pipe->bind_fs_state(pipe, st->fp->driver_shader); - pipe->bind_vs_state(pipe, st->vp->driver_shader); + cso_restore_fragment_shader(cso); + cso_restore_vertex_shader(cso); } diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index a931911227..da9ec12a4d 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -95,12 +95,6 @@ st_render_mipmap(struct st_context *st, util_gen_mipmap(st->gen_mipmap, pt, face, baseLevel, lastLevel, PIPE_TEX_FILTER_LINEAR); - /* shaders don't go through CSO yet */ - if (st->fp) - pipe->bind_fs_state(pipe, st->fp->driver_shader); - if (st->vp) - pipe->bind_vs_state(pipe, st->vp->driver_shader); - return TRUE; } -- cgit v1.2.3 From b17e123a8f20239e8e1fc6816ccf115d9ec57471 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 19:09:38 +0100 Subject: rtasm: propogate errors in x86 emit --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 34 +++++++++++++++++++++++++----- src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 1 + 2 files changed, 30 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index f2c08c96a6..c2fe0e40f5 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -146,7 +146,10 @@ _fill( static void do_realloc( struct x86_function *p ) { - if (p->size == 0) { + if (p->store == p->error_overflow) { + p->csr = p->store; + } + else if (p->size == 0) { p->size = 1024; p->store = rtasm_exec_malloc(p->size); p->csr = p->store; @@ -156,10 +159,22 @@ static void do_realloc( struct x86_function *p ) unsigned char *tmp = p->store; p->size *= 2; p->store = rtasm_exec_malloc(p->size); - memcpy(p->store, tmp, used); - p->csr = p->store + used; + + if (p->store) { + memcpy(p->store, tmp, used); + p->csr = p->store + used; + } + else { + p->csr = p->store; + } + rtasm_exec_free(tmp); } + + if (p->store == NULL) { + p->store = p->csr = p->error_overflow; + p->size = 4; + } } /* Emit bytes to the instruction stream: @@ -1440,12 +1455,17 @@ void x86_init_func_size( struct x86_function *p, unsigned code_size ) { p->size = code_size; p->store = rtasm_exec_malloc(code_size); + if (p->store == NULL) { + p->store = p->error_overflow; + } p->csr = p->store; } void x86_release_func( struct x86_function *p ) { - rtasm_exec_free(p->store); + if (p->store && p->store != p->error_overflow) + rtasm_exec_free(p->store); + p->store = NULL; p->csr = NULL; p->size = 0; @@ -1456,7 +1476,11 @@ void (*x86_get_func( struct x86_function *p ))(void) { if (DISASSEM && p->store) debug_printf("disassemble %p %p\n", p->store, p->csr); - return (void (*)(void)) p->store; + + if (p->store == p->error_overflow) + return (void (*)(void)) NULL; + else + return (void (*)(void)) p->store; } #else diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index 5e99ceea70..695a1cef4e 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -43,6 +43,7 @@ struct x86_function { unsigned char *csr; unsigned stack_offset; int need_emms; + unsigned char error_overflow[4]; const char *fn; }; -- cgit v1.2.3 From d3db46eb8257c1b0cf823f1805ca00457be9aff3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 19:09:54 +0100 Subject: translate: fail on x86 rtasm fail --- src/gallium/auxiliary/translate/translate_sse.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index e587e5afec..f590d48b78 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -597,7 +597,12 @@ struct translate *translate_sse2_create( const struct translate_key *key ) goto fail; p->gen_run = (run_func)x86_get_func(&p->linear_func); + if (p->gen_run == NULL) + goto fail; + p->gen_run_elts = (run_elts_func)x86_get_func(&p->elt_func); + if (p->gen_run_elts == NULL) + goto fail; return &p->translate; -- cgit v1.2.3 From 785831fc6fc56815d9637c7dd2acbcee6dfbbb0a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 19:11:58 +0100 Subject: cso: propogate one easy error - many more --- src/gallium/auxiliary/cso_cache/cso_cache.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 18acab0967..f607528fdc 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -290,6 +290,8 @@ void * cso_take_state(struct cso_cache *sc, struct cso_cache *cso_cache_create(void) { struct cso_cache *sc = MALLOC_STRUCT(cso_cache); + if (sc == NULL) + return NULL; sc->max_size = 4096; sc->blend_hash = cso_hash_create(); -- cgit v1.2.3 From 73c2711bb186692b866720058a09f5eb05950213 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 19:43:53 +0100 Subject: rtasm: clean up debug dumping a little --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 286 ++++++++++++++--------------- 1 file changed, 140 insertions(+), 146 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index c2fe0e40f5..10796c540d 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -77,69 +77,60 @@ _print_reg( debug_printf( "]" ); } -static void -_fill( - const char *op ) -{ - unsigned count = 10 - strlen( op ); - while( count-- ) { - debug_printf( " " ); - } -} +#define DUMP_START() debug_printf( "\n" ) +#define DUMP_END() debug_printf( "\n" ) -#define DUMP_START() debug_printf( "\nsse-dump start ----------------" ) -#define DUMP_END() debug_printf( "\nsse-dump end ----------------\n" ) -#define DUMP( OP ) debug_printf( "\n%s", OP ) +#define DUMP() do { \ + const char *foo = __FUNCTION__; \ + while (*foo && *foo != '_') \ + foo++; \ + if (*foo) \ + foo++; \ + debug_printf( "\n% 15s ", foo ); \ +} while (0) -#define DUMP_I( OP, I ) do { \ - debug_printf( "\n%s", OP ); \ - _fill( OP ); \ +#define DUMP_I( I ) do { \ + DUMP(); \ debug_printf( "%u", I ); \ } while( 0 ) -#define DUMP_R( OP, R0 ) do { \ - debug_printf( "\n%s", OP ); \ - _fill( OP ); \ +#define DUMP_R( R0 ) do { \ + DUMP(); \ _print_reg( R0 ); \ } while( 0 ) -#define DUMP_RR( OP, R0, R1 ) do { \ - debug_printf( "\n%s", OP ); \ - _fill( OP ); \ +#define DUMP_RR( R0, R1 ) do { \ + DUMP(); \ _print_reg( R0 ); \ debug_printf( ", " ); \ _print_reg( R1 ); \ } while( 0 ) -#define DUMP_RI( OP, R0, I ) do { \ - debug_printf( "\n%s", OP ); \ - _fill( OP ); \ +#define DUMP_RI( R0, I ) do { \ + DUMP(); \ _print_reg( R0 ); \ - debug_printf( ", " ); \ - debug_printf( "%u", I ); \ + debug_printf( ", %u", I ); \ } while( 0 ) -#define DUMP_RRI( OP, R0, R1, I ) do { \ - debug_printf( "\n%s", OP ); \ - _fill( OP ); \ +#define DUMP_RRI( R0, R1, I ) do { \ + DUMP(); \ _print_reg( R0 ); \ debug_printf( ", " ); \ _print_reg( R1 ); \ - debug_printf( ", " ); \ - debug_printf( "%u", I ); \ + debug_printf( ", %u", I ); \ } while( 0 ) #else #define DUMP_START() #define DUMP_END() -#define DUMP( OP ) -#define DUMP_I( OP, I ) -#define DUMP_R( OP, R0 ) -#define DUMP_RR( OP, R0, R1 ) -#define DUMP_RI( OP, R0, I ) -#define DUMP_RRI( OP, R0, R1, I ) +#define DUMP( ) +#define DUMP_I( I ) +#define DUMP_R( R0 ) +#define DUMP_RR( R0, R1 ) +#define DUMP_RI( R0, I ) +#define DUMP_RRI( R0, R1, I ) #endif @@ -173,7 +164,7 @@ static void do_realloc( struct x86_function *p ) if (p->store == NULL) { p->store = p->csr = p->error_overflow; - p->size = 4; + p->size = sizeof(p->error_overflow); } } @@ -397,7 +388,7 @@ unsigned char *x86_jcc_forward( struct x86_function *p, unsigned char *x86_jmp_forward( struct x86_function *p) { - DUMP( __FUNCTION__ ); + DUMP(); emit_1ub(p, 0xe9); emit_1i(p, 0); return x86_get_label(p); @@ -405,7 +396,7 @@ unsigned char *x86_jmp_forward( struct x86_function *p) unsigned char *x86_call_forward( struct x86_function *p) { - DUMP( __FUNCTION__ ); + DUMP(); emit_1ub(p, 0xe8); emit_1i(p, 0); @@ -422,7 +413,7 @@ void x86_fixup_fwd_jump( struct x86_function *p, void x86_jmp( struct x86_function *p, unsigned char *label) { - DUMP_I( __FUNCTION__, label ); + DUMP_I( label ); emit_1ub(p, 0xe9); emit_1i(p, pointer_to_intptr( label ) - pointer_to_intptr( x86_get_label(p) ) - 4); } @@ -439,14 +430,14 @@ static unsigned char *cptr( void (*label)() ) */ void x86_call( struct x86_function *p, void (*label)()) { - DUMP_I( __FUNCTION__, label ); + DUMP_I( label ); emit_1ub(p, 0xe8); emit_1i(p, cptr(label) - x86_get_label(p) - 4); } #else void x86_call( struct x86_function *p, struct x86_reg reg) { - DUMP_R( __FUNCTION__, reg ); + DUMP_R( reg ); emit_1ub(p, 0xff); emit_modrm_noreg(p, 2, reg); } @@ -459,7 +450,7 @@ void x86_call( struct x86_function *p, struct x86_reg reg) */ void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) { - DUMP_RI( __FUNCTION__, dst, imm ); + DUMP_RI( dst, imm ); assert(dst.mod == mod_REG); emit_1ub(p, 0xb8 + dst.idx); emit_1i(p, imm); @@ -468,7 +459,7 @@ void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) void x86_push( struct x86_function *p, struct x86_reg reg ) { - DUMP_R( __FUNCTION__, reg ); + DUMP_R( reg ); assert(reg.mod == mod_REG); emit_1ub(p, 0x50 + reg.idx); p->stack_offset += 4; @@ -477,7 +468,7 @@ void x86_push( struct x86_function *p, void x86_pop( struct x86_function *p, struct x86_reg reg ) { - DUMP_R( __FUNCTION__, reg ); + DUMP_R( reg ); assert(reg.mod == mod_REG); emit_1ub(p, 0x58 + reg.idx); p->stack_offset -= 4; @@ -486,7 +477,7 @@ void x86_pop( struct x86_function *p, void x86_inc( struct x86_function *p, struct x86_reg reg ) { - DUMP_R( __FUNCTION__, reg ); + DUMP_R( reg ); assert(reg.mod == mod_REG); emit_1ub(p, 0x40 + reg.idx); } @@ -494,20 +485,20 @@ void x86_inc( struct x86_function *p, void x86_dec( struct x86_function *p, struct x86_reg reg ) { - DUMP_R( __FUNCTION__, reg ); + DUMP_R( reg ); assert(reg.mod == mod_REG); emit_1ub(p, 0x48 + reg.idx); } void x86_ret( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_1ub(p, 0xc3); } void x86_sahf( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_1ub(p, 0x9e); } @@ -515,7 +506,7 @@ void x86_mov( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_op_modrm( p, 0x8b, 0x89, dst, src ); } @@ -523,7 +514,7 @@ void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_op_modrm( p, 0x33, 0x31, dst, src ); } @@ -531,7 +522,7 @@ void x86_cmp( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_op_modrm( p, 0x3b, 0x39, dst, src ); } @@ -539,7 +530,7 @@ void x86_lea( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_1ub(p, 0x8d); emit_modrm( p, dst, src ); } @@ -548,7 +539,7 @@ void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_1ub(p, 0x85); emit_modrm( p, dst, src ); } @@ -557,7 +548,7 @@ void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_op_modrm(p, 0x03, 0x01, dst, src ); } @@ -566,7 +557,7 @@ void x86_add( struct x86_function *p, void x86_mul( struct x86_function *p, struct x86_reg src ) { - DUMP_R( __FUNCTION__, src ); + DUMP_R( src ); emit_1ub(p, 0xf7); emit_modrm_noreg(p, 4, src ); } @@ -576,7 +567,7 @@ void x86_imul( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0xAF); emit_modrm(p, dst, src); } @@ -586,7 +577,7 @@ void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_op_modrm(p, 0x2b, 0x29, dst, src ); } @@ -594,7 +585,7 @@ void x86_or( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_op_modrm( p, 0x0b, 0x09, dst, src ); } @@ -602,7 +593,7 @@ void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_op_modrm( p, 0x23, 0x21, dst, src ); } @@ -617,7 +608,7 @@ void sse_movss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, 0xF3, X86_TWOB); emit_op_modrm( p, 0x10, 0x11, dst, src ); } @@ -626,7 +617,7 @@ void sse_movaps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x28, 0x29, dst, src ); } @@ -635,7 +626,7 @@ void sse_movups( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x10, 0x11, dst, src ); } @@ -644,7 +635,7 @@ void sse_movhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); assert(dst.mod != mod_REG || src.mod != mod_REG); emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x16, 0x17, dst, src ); /* cf movlhps */ @@ -654,7 +645,7 @@ void sse_movlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); assert(dst.mod != mod_REG || src.mod != mod_REG); emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x12, 0x13, dst, src ); /* cf movhlps */ @@ -664,7 +655,7 @@ void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x5F); emit_modrm( p, dst, src ); } @@ -673,7 +664,7 @@ void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x5F); emit_modrm( p, dst, src ); } @@ -682,7 +673,7 @@ void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x5E); emit_modrm( p, dst, src ); } @@ -691,7 +682,7 @@ void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x5D); emit_modrm( p, dst, src ); } @@ -700,7 +691,7 @@ void sse_subps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x5C); emit_modrm( p, dst, src ); } @@ -709,7 +700,7 @@ void sse_mulps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x59); emit_modrm( p, dst, src ); } @@ -718,7 +709,7 @@ void sse_mulss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x59); emit_modrm( p, dst, src ); } @@ -727,7 +718,7 @@ void sse_addps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x58); emit_modrm( p, dst, src ); } @@ -736,7 +727,7 @@ void sse_addss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x58); emit_modrm( p, dst, src ); } @@ -745,7 +736,7 @@ void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x55); emit_modrm( p, dst, src ); } @@ -754,7 +745,7 @@ void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x54); emit_modrm( p, dst, src ); } @@ -763,7 +754,7 @@ void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x52); emit_modrm( p, dst, src ); } @@ -772,7 +763,7 @@ void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x52); emit_modrm( p, dst, src ); @@ -782,7 +773,7 @@ void sse_movhlps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); assert(dst.mod == mod_REG && src.mod == mod_REG); emit_2ub(p, X86_TWOB, 0x12); emit_modrm( p, dst, src ); @@ -792,7 +783,7 @@ void sse_movlhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); assert(dst.mod == mod_REG && src.mod == mod_REG); emit_2ub(p, X86_TWOB, 0x16); emit_modrm( p, dst, src ); @@ -802,7 +793,7 @@ void sse_orps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x56); emit_modrm( p, dst, src ); } @@ -811,7 +802,7 @@ void sse_xorps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x57); emit_modrm( p, dst, src ); } @@ -820,7 +811,7 @@ void sse_cvtps2pi( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); assert(dst.file == file_MMX && (src.file == file_XMM || src.mod != mod_REG)); @@ -834,7 +825,7 @@ void sse2_cvtdq2ps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x5b); emit_modrm( p, dst, src ); } @@ -848,7 +839,7 @@ void sse_shufps( struct x86_function *p, struct x86_reg src, unsigned char shuf) { - DUMP_RRI( __FUNCTION__, dst, src, shuf ); + DUMP_RRI( dst, src, shuf ); emit_2ub(p, X86_TWOB, 0xC6); emit_modrm(p, dst, src); emit_1ub(p, shuf); @@ -859,7 +850,7 @@ void sse_cmpps( struct x86_function *p, struct x86_reg src, unsigned char cc) { - DUMP_RRI( "CMPPS", dst, src, cc ); + DUMP_RRI( dst, src, cc ); emit_2ub(p, X86_TWOB, 0xC2); emit_modrm(p, dst, src); emit_1ub(p, cc); @@ -869,7 +860,7 @@ void sse_pmovmskb( struct x86_function *p, struct x86_reg dst, struct x86_reg src) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0xD7); emit_modrm(p, dst, src); } @@ -886,7 +877,7 @@ void sse2_pshufd( struct x86_function *p, struct x86_reg src, unsigned char shuf) { - DUMP_RRI( __FUNCTION__, dst, src, shuf ); + DUMP_RRI( dst, src, shuf ); emit_3ub(p, 0x66, X86_TWOB, 0x70); emit_modrm(p, dst, src); emit_1ub(p, shuf); @@ -896,7 +887,7 @@ void sse2_cvttps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub( p, 0xF3, X86_TWOB, 0x5B ); emit_modrm( p, dst, src ); } @@ -905,7 +896,7 @@ void sse2_cvtps2dq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x5B); emit_modrm( p, dst, src ); } @@ -914,7 +905,7 @@ void sse2_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x6B); emit_modrm( p, dst, src ); } @@ -923,7 +914,7 @@ void sse2_packsswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x63); emit_modrm( p, dst, src ); } @@ -932,7 +923,7 @@ void sse2_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x67); emit_modrm( p, dst, src ); } @@ -941,7 +932,7 @@ void sse2_punpcklbw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0x66, X86_TWOB, 0x60); emit_modrm( p, dst, src ); } @@ -951,7 +942,7 @@ void sse2_rcpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, X86_TWOB, 0x53); emit_modrm( p, dst, src ); } @@ -960,7 +951,7 @@ void sse2_rcpss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_3ub(p, 0xF3, X86_TWOB, 0x53); emit_modrm( p, dst, src ); } @@ -969,7 +960,7 @@ void sse2_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); emit_2ub(p, 0x66, X86_TWOB); emit_op_modrm( p, 0x6e, 0x7e, dst, src ); } @@ -982,35 +973,35 @@ void sse2_movd( struct x86_function *p, */ void x87_fist( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); emit_1ub(p, 0xdb); emit_modrm_noreg(p, 2, dst); } void x87_fistp( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); emit_1ub(p, 0xdb); emit_modrm_noreg(p, 3, dst); } void x87_fild( struct x86_function *p, struct x86_reg arg ) { - DUMP_R( __FUNCTION__, arg ); + DUMP_R( arg ); emit_1ub(p, 0xdf); emit_modrm_noreg(p, 0, arg); } void x87_fldz( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xee); } void x87_fldcw( struct x86_function *p, struct x86_reg arg ) { - DUMP_R( __FUNCTION__, arg ); + DUMP_R( arg ); assert(arg.file == file_REG32); assert(arg.mod != mod_REG); emit_1ub(p, 0xd9); @@ -1019,31 +1010,31 @@ void x87_fldcw( struct x86_function *p, struct x86_reg arg ) void x87_fld1( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xe8); } void x87_fldl2e( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xea); } void x87_fldln2( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xed); } void x87_fwait( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_1ub(p, 0x9b); } void x87_fnclex( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xdb, 0xe2); } @@ -1082,7 +1073,7 @@ static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86 void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); x87_arith_op(p, dst, src, 0xd8, 0xc8, 0xdc, 0xc8, @@ -1091,7 +1082,7 @@ void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); x87_arith_op(p, dst, src, 0xd8, 0xe0, 0xdc, 0xe8, @@ -1100,7 +1091,7 @@ void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); x87_arith_op(p, dst, src, 0xd8, 0xe8, 0xdc, 0xe0, @@ -1109,7 +1100,7 @@ void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); x87_arith_op(p, dst, src, 0xd8, 0xc0, 0xdc, 0xc0, @@ -1118,7 +1109,7 @@ void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); x87_arith_op(p, dst, src, 0xd8, 0xf0, 0xdc, 0xf8, @@ -1127,7 +1118,7 @@ void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); x87_arith_op(p, dst, src, 0xd8, 0xf8, 0xdc, 0xf0, @@ -1136,7 +1127,7 @@ void x87_fdivr( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void x87_fmulp( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xc8+dst.idx); @@ -1144,7 +1135,7 @@ void x87_fmulp( struct x86_function *p, struct x86_reg dst ) void x87_fsubp( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xe8+dst.idx); @@ -1152,7 +1143,7 @@ void x87_fsubp( struct x86_function *p, struct x86_reg dst ) void x87_fsubrp( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xe0+dst.idx); @@ -1160,7 +1151,7 @@ void x87_fsubrp( struct x86_function *p, struct x86_reg dst ) void x87_faddp( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xc0+dst.idx); @@ -1168,7 +1159,7 @@ void x87_faddp( struct x86_function *p, struct x86_reg dst ) void x87_fdivp( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xf8+dst.idx); @@ -1176,7 +1167,7 @@ void x87_fdivp( struct x86_function *p, struct x86_reg dst ) void x87_fdivrp( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xf0+dst.idx); @@ -1184,83 +1175,83 @@ void x87_fdivrp( struct x86_function *p, struct x86_reg dst ) void x87_fucom( struct x86_function *p, struct x86_reg arg ) { - DUMP_R( __FUNCTION__, arg ); + DUMP_R( arg ); assert(arg.file == file_x87); emit_2ub(p, 0xdd, 0xe0+arg.idx); } void x87_fucomp( struct x86_function *p, struct x86_reg arg ) { - DUMP_R( __FUNCTION__, arg ); + DUMP_R( arg ); assert(arg.file == file_x87); emit_2ub(p, 0xdd, 0xe8+arg.idx); } void x87_fucompp( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xda, 0xe9); } void x87_fxch( struct x86_function *p, struct x86_reg arg ) { - DUMP_R( __FUNCTION__, arg ); + DUMP_R( arg ); assert(arg.file == file_x87); emit_2ub(p, 0xd9, 0xc8+arg.idx); } void x87_fabs( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xe1); } void x87_fchs( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xe0); } void x87_fcos( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xff); } void x87_fprndint( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xfc); } void x87_fscale( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xfd); } void x87_fsin( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xfe); } void x87_fsincos( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xfb); } void x87_fsqrt( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xfa); } void x87_fxtract( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xf4); } @@ -1270,7 +1261,7 @@ void x87_fxtract( struct x86_function *p ) */ void x87_f2xm1( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xf0); } @@ -1279,7 +1270,7 @@ void x87_f2xm1( struct x86_function *p ) */ void x87_fyl2x( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xf1); } @@ -1290,14 +1281,14 @@ void x87_fyl2x( struct x86_function *p ) */ void x87_fyl2xp1( struct x86_function *p ) { - DUMP( __FUNCTION__ ); + DUMP(); emit_2ub(p, 0xd9, 0xf9); } void x87_fld( struct x86_function *p, struct x86_reg arg ) { - DUMP_R( __FUNCTION__, arg ); + DUMP_R( arg ); if (arg.file == file_x87) emit_2ub(p, 0xd9, 0xc0 + arg.idx); else { @@ -1308,7 +1299,7 @@ void x87_fld( struct x86_function *p, struct x86_reg arg ) void x87_fst( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); if (dst.file == file_x87) emit_2ub(p, 0xdd, 0xd0 + dst.idx); else { @@ -1319,7 +1310,7 @@ void x87_fst( struct x86_function *p, struct x86_reg dst ) void x87_fstp( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); if (dst.file == file_x87) emit_2ub(p, 0xdd, 0xd8 + dst.idx); else { @@ -1330,7 +1321,7 @@ void x87_fstp( struct x86_function *p, struct x86_reg dst ) void x87_fcom( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); if (dst.file == file_x87) emit_2ub(p, 0xd8, 0xd0 + dst.idx); else { @@ -1341,7 +1332,7 @@ void x87_fcom( struct x86_function *p, struct x86_reg dst ) void x87_fcomp( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); if (dst.file == file_x87) emit_2ub(p, 0xd8, 0xd8 + dst.idx); else { @@ -1353,7 +1344,7 @@ void x87_fcomp( struct x86_function *p, struct x86_reg dst ) void x87_fnstsw( struct x86_function *p, struct x86_reg dst ) { - DUMP_R( __FUNCTION__, dst ); + DUMP_R( dst ); assert(dst.file == file_REG32); if (dst.idx == reg_AX && @@ -1383,7 +1374,7 @@ void mmx_packssdw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); assert(dst.file == file_MMX && (src.file == file_MMX || src.mod != mod_REG)); @@ -1397,7 +1388,7 @@ void mmx_packuswb( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); assert(dst.file == file_MMX && (src.file == file_MMX || src.mod != mod_REG)); @@ -1411,7 +1402,7 @@ void mmx_movd( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); p->need_emms = 1; emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x6e, 0x7e, dst, src ); @@ -1421,7 +1412,7 @@ void mmx_movq( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) { - DUMP_RR( __FUNCTION__, dst, src ); + DUMP_RR( dst, src ); p->need_emms = 1; emit_1ub(p, X86_TWOB); emit_op_modrm( p, 0x6f, 0x7f, dst, src ); @@ -1449,6 +1440,7 @@ void x86_init_func( struct x86_function *p ) p->size = 0; p->store = NULL; p->csr = p->store; + DUMP_START(); } void x86_init_func_size( struct x86_function *p, unsigned code_size ) @@ -1459,6 +1451,7 @@ void x86_init_func_size( struct x86_function *p, unsigned code_size ) p->store = p->error_overflow; } p->csr = p->store; + DUMP_START(); } void x86_release_func( struct x86_function *p ) @@ -1474,6 +1467,7 @@ void x86_release_func( struct x86_function *p ) void (*x86_get_func( struct x86_function *p ))(void) { + DUMP_END(); if (DISASSEM && p->store) debug_printf("disassemble %p %p\n", p->store, p->csr); -- cgit v1.2.3 From 73706deef59c35472d2410411403f30c9603f22f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 19:48:08 +0100 Subject: rtasm: quieten sse_enabled debug --- src/gallium/auxiliary/rtasm/rtasm_cpu.c | 11 ++++++++++- 1 file changed, 10 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.c b/src/gallium/auxiliary/rtasm/rtasm_cpu.c index 175245a9f6..f01e12faa0 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_cpu.c +++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.c @@ -32,7 +32,16 @@ static boolean rtasm_sse_enabled(void) { - return !debug_get_bool_option("GALLIUM_NOSSE", FALSE); + static boolean firsttime = 1; + static boolean enabled; + + /* This gets called quite often at the moment: + */ + if (firsttime) { + enabled = !debug_get_bool_option("GALLIUM_NOSSE", FALSE); + firsttime = FALSE; + } + return enabled; } int rtasm_cpu_has_sse(void) -- cgit v1.2.3 From a945420ae6f96f0d7024f97e37ffd31329865a84 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 19:48:21 +0100 Subject: rtasm: debug some missing funcs --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 10796c540d..3cd45d7dd9 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -364,6 +364,7 @@ void x86_jcc( struct x86_function *p, unsigned char *label ) { intptr_t offset = pointer_to_intptr( label ) - (pointer_to_intptr( x86_get_label(p) ) + 2); + DUMP_I(cc); if (offset <= 127 && offset >= -128) { emit_1ub(p, 0x70 + cc); @@ -381,6 +382,7 @@ void x86_jcc( struct x86_function *p, unsigned char *x86_jcc_forward( struct x86_function *p, enum x86_cc cc ) { + DUMP_I(cc); emit_2ub(p, 0x0f, 0x80 + cc); emit_1i(p, 0); return x86_get_label(p); @@ -1365,6 +1367,7 @@ void x87_fnstsw( struct x86_function *p, struct x86_reg dst ) void mmx_emms( struct x86_function *p ) { + DUMP(); assert(p->need_emms); emit_2ub(p, 0x0f, 0x77); p->need_emms = 0; -- cgit v1.2.3 From fb1c09305ea8cb727514c53a0346d90e78eeb05f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 21 Apr 2008 15:15:31 -0400 Subject: Use llvm 2.3 (2.2 won't work because of a lot of problems, e.g. lack of constant vectors handling in execution engine) --- src/gallium/auxiliary/gallivm/Makefile | 12 +++--- src/gallium/auxiliary/gallivm/instructions.cpp | 42 +++++++++---------- src/gallium/auxiliary/gallivm/instructions.h | 4 +- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 49 ++++++++++++----------- src/gallium/auxiliary/gallivm/instructionssoa.h | 4 +- src/gallium/auxiliary/gallivm/storage.cpp | 32 +++++++-------- src/gallium/auxiliary/gallivm/storagesoa.cpp | 10 ++--- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 4 +- 8 files changed, 79 insertions(+), 78 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/Makefile b/src/gallium/auxiliary/gallivm/Makefile index 6bfd187fb3..c3f7bfba93 100644 --- a/src/gallium/auxiliary/gallivm/Makefile +++ b/src/gallium/auxiliary/gallivm/Makefile @@ -65,14 +65,14 @@ depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(INC_SOURCES) gallivm_builtins.cpp: llvm_builtins.c - clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp.bin - (echo "static const unsigned char llvm_builtins_data[] = {"; od -txC temp.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/};/") >$@ - rm temp.bin + clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp1.bin + (echo "static const unsigned char llvm_builtins_data[] = {"; od -txC temp1.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/};/") >$@ + rm temp1.bin gallivmsoabuiltins.cpp: soabuiltins.c - clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp.bin - (echo "static const unsigned char soabuiltins_data[] = {"; od -txC temp.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/};/") >$@ - rm temp.bin + clang --emit-llvm < $< |llvm-as|opt -std-compile-opts > temp2.bin + (echo "static const unsigned char soabuiltins_data[] = {"; od -txC temp2.bin | sed -e "s/^[0-9]*//" -e s"/ \([0-9a-f][0-9a-f]\)/0x\1,/g" -e"\$$d" | sed -e"\$$s/,$$/};/") >$@ + rm temp2.bin # Emacs tags tags: diff --git a/src/gallium/auxiliary/gallivm/instructions.cpp b/src/gallium/auxiliary/gallivm/instructions.cpp index b35d6790f7..95a670edaf 100644 --- a/src/gallium/auxiliary/gallivm/instructions.cpp +++ b/src/gallium/auxiliary/gallivm/instructions.cpp @@ -139,12 +139,12 @@ llvm::Value *Instructions::callFSqrt(llvm::Value *val) // predeclare the intrinsic std::vector fsqrtArgs; fsqrtArgs.push_back(Type::FloatTy); - ParamAttrsList *fsqrtPal = 0; + PAListPtr fsqrtPal; FunctionType* fsqrtType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/fsqrtArgs, /*isVarArg=*/false); - m_llvmFSqrt = new Function( + m_llvmFSqrt = Function::Create( /*Type=*/fsqrtType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"llvm.sqrt.f32", m_mod); @@ -196,12 +196,12 @@ llvm::Value *Instructions::callFAbs(llvm::Value *val) // predeclare the intrinsic std::vector fabsArgs; fabsArgs.push_back(Type::FloatTy); - ParamAttrsList *fabsPal = 0; + PAListPtr fabsPal; FunctionType* fabsType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/fabsArgs, /*isVarArg=*/false); - m_llvmFAbs = new Function( + m_llvmFAbs = Function::Create( /*Type=*/fabsType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"fabs", m_mod); @@ -239,12 +239,12 @@ llvm::Value * Instructions::callPow(llvm::Value *val1, llvm::Value *val2) std::vector powArgs; powArgs.push_back(Type::FloatTy); powArgs.push_back(Type::FloatTy); - ParamAttrsList *powPal = 0; + PAListPtr powPal; FunctionType* powType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/powArgs, /*isVarArg=*/false); - m_llvmPow = new Function( + m_llvmPow = Function::Create( /*Type=*/powType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"llvm.pow.f32", m_mod); @@ -338,12 +338,12 @@ llvm::Value * Instructions::callFloor(llvm::Value *val) // predeclare the intrinsic std::vector floorArgs; floorArgs.push_back(Type::FloatTy); - ParamAttrsList *floorPal = 0; + PAListPtr floorPal; FunctionType* floorType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/floorArgs, /*isVarArg=*/false); - m_llvmFloor = new Function( + m_llvmFloor = Function::Create( /*Type=*/floorType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"floorf", m_mod); @@ -381,12 +381,12 @@ llvm::Value * Instructions::callFLog(llvm::Value *val) // predeclare the intrinsic std::vector flogArgs; flogArgs.push_back(Type::FloatTy); - ParamAttrsList *flogPal = 0; + PAListPtr flogPal; FunctionType* flogType = FunctionType::get( /*Result=*/Type::FloatTy, /*Params=*/flogArgs, /*isVarArg=*/false); - m_llvmFlog = new Function( + m_llvmFlog = Function::Create( /*Type=*/flogType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"logf", m_mod); @@ -509,12 +509,12 @@ void Instructions::printVector(llvm::Value *val) llvm::Function * Instructions::declarePrintf() { std::vector args; - ParamAttrsList *params = 0; + PAListPtr params; FunctionType* funcTy = FunctionType::get( /*Result=*/IntegerType::get(32), /*Params=*/args, /*isVarArg=*/true); - Function* func_printf = new Function( + Function* func_printf = Function::Create( /*Type=*/funcTy, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/"printf", m_mod); @@ -638,8 +638,8 @@ llvm::Value * Instructions::abs(llvm::Value *in) void Instructions::ifop(llvm::Value *in) { - BasicBlock *ifthen = new BasicBlock(name("ifthen"), m_func,0); - BasicBlock *ifend = new BasicBlock(name("ifthenend"), m_func,0); + BasicBlock *ifthen = BasicBlock::Create(name("ifthen"), m_func,0); + BasicBlock *ifend = BasicBlock::Create(name("ifthenend"), m_func,0); //BasicBlock *yblock = new BasicBlock(name("yblock"), m_func,0); //BasicBlock *zblock = new BasicBlock(name("zblock"), m_func,0); @@ -665,7 +665,7 @@ llvm::BasicBlock * Instructions::currentBlock() const void Instructions::elseop() { assert(!m_ifStack.empty()); - BasicBlock *ifend = new BasicBlock(name("ifend"), m_func,0); + BasicBlock *ifend = BasicBlock::Create(name("ifend"), m_func,0); m_builder.CreateBr(ifend); m_builder.SetInsertPoint(m_ifStack.top()); currentBlock()->setName(name("ifelse")); @@ -692,8 +692,8 @@ llvm::Value * Instructions::lerp(llvm::Value *in1, llvm::Value *in2, void Instructions::beginLoop() { - BasicBlock *begin = new BasicBlock(name("loop"), m_func,0); - BasicBlock *end = new BasicBlock(name("endloop"), m_func,0); + BasicBlock *begin = BasicBlock::Create(name("loop"), m_func,0); + BasicBlock *end = BasicBlock::Create(name("endloop"), m_func,0); m_builder.CreateBr(begin); Loop loop; @@ -716,7 +716,7 @@ void Instructions::endLoop() void Instructions::brk() { assert(!m_loopStack.empty()); - BasicBlock *unr = new BasicBlock(name("unreachable"), m_func,0); + BasicBlock *unr = BasicBlock::Create(name("unreachable"), m_func,0); m_builder.CreateBr(m_loopStack.top().end); m_builder.SetInsertPoint(unr); } @@ -765,13 +765,13 @@ llvm::Function * Instructions::declareFunc(int label) args.push_back(vecPtr); args.push_back(vecPtr); args.push_back(vecPtr); - ParamAttrsList *params = 0; + PAListPtr params; FunctionType *funcType = FunctionType::get( /*Result=*/Type::VoidTy, /*Params=*/args, /*isVarArg=*/false); std::string name = createFuncName(label); - Function *func = new Function( + Function *func = Function::Create( /*Type=*/funcType, /*Linkage=*/GlobalValue::ExternalLinkage, /*Name=*/name.c_str(), m_mod); @@ -789,7 +789,7 @@ void Instructions::bgnSub(unsigned label) ptr_INPUT->setName("INPUT"); m_storage->pushArguments(ptr_INPUT); - llvm::BasicBlock *entry = new BasicBlock("entry", func, 0); + llvm::BasicBlock *entry = BasicBlock::Create("entry", func, 0); m_func = func; m_builder.SetInsertPoint(entry); diff --git a/src/gallium/auxiliary/gallivm/instructions.h b/src/gallium/auxiliary/gallivm/instructions.h index 9ebc17dd8e..19ca84ddc6 100644 --- a/src/gallium/auxiliary/gallivm/instructions.h +++ b/src/gallium/auxiliary/gallivm/instructions.h @@ -36,7 +36,7 @@ #include #include #include -#include +#include #include #include @@ -125,7 +125,7 @@ private: llvm::Module *m_mod; llvm::Function *m_func; char m_name[32]; - llvm::LLVMFoldingBuilder m_builder; + llvm::IRBuilder m_builder; int m_idx; llvm::VectorType *m_floatVecType; diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 3c16a3692d..f0122802db 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -38,6 +38,7 @@ #include #include #include +//#include #include #include @@ -237,32 +238,32 @@ llvm::Value * InstructionsSoa::allocaTemp() std::vector indices; indices.push_back(m_storage->constantInt(0)); indices.push_back(m_storage->constantInt(0)); - GetElementPtrInst *getElem = new GetElementPtrInst(alloca, - indices.begin(), - indices.end(), - name("allocaPtr"), - m_builder.GetInsertBlock()); + GetElementPtrInst *getElem = GetElementPtrInst::Create(alloca, + indices.begin(), + indices.end(), + name("allocaPtr"), + m_builder.GetInsertBlock()); return getElem; } std::vector InstructionsSoa::allocaToResult(llvm::Value *allocaPtr) { - GetElementPtrInst *xElemPtr = new GetElementPtrInst(allocaPtr, - m_storage->constantInt(0), - name("xPtr"), - m_builder.GetInsertBlock()); - GetElementPtrInst *yElemPtr = new GetElementPtrInst(allocaPtr, - m_storage->constantInt(1), - name("yPtr"), - m_builder.GetInsertBlock()); - GetElementPtrInst *zElemPtr = new GetElementPtrInst(allocaPtr, - m_storage->constantInt(2), - name("zPtr"), - m_builder.GetInsertBlock()); - GetElementPtrInst *wElemPtr = new GetElementPtrInst(allocaPtr, - m_storage->constantInt(3), - name("wPtr"), - m_builder.GetInsertBlock()); + GetElementPtrInst *xElemPtr = GetElementPtrInst::Create(allocaPtr, + m_storage->constantInt(0), + name("xPtr"), + m_builder.GetInsertBlock()); + GetElementPtrInst *yElemPtr = GetElementPtrInst::Create(allocaPtr, + m_storage->constantInt(1), + name("yPtr"), + m_builder.GetInsertBlock()); + GetElementPtrInst *zElemPtr = GetElementPtrInst::Create(allocaPtr, + m_storage->constantInt(2), + name("zPtr"), + m_builder.GetInsertBlock()); + GetElementPtrInst *wElemPtr = GetElementPtrInst::Create(allocaPtr, + m_storage->constantInt(3), + name("wPtr"), + m_builder.GetInsertBlock()); std::vector res(4); res[0] = new LoadInst(xElemPtr, name("xRes"), false, m_builder.GetInsertBlock()); @@ -388,10 +389,10 @@ void InstructionsSoa::injectFunction(llvm::Function *originalFunc, int op) llvm::Function *func = 0; if (originalFunc->isDeclaration()) { std::cout << "function decleration" <getFunctionType(), GlobalValue::ExternalLinkage, - originalFunc->getName(), currentModule()); + func = Function::Create(originalFunc->getFunctionType(), GlobalValue::ExternalLinkage, + originalFunc->getName(), currentModule()); func->setCallingConv(CallingConv::C); - const ParamAttrsList *pal = 0; + const PAListPtr pal; func->setParamAttrs(pal); currentModule()->dump(); } else { diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h index b9104ea286..060ee72f2e 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.h +++ b/src/gallium/auxiliary/gallivm/instructionssoa.h @@ -29,7 +29,7 @@ #define INSTRUCTIONSSOA_H #include -#include +#include #include #include @@ -87,7 +87,7 @@ private: const std::vector in3); void injectFunction(llvm::Function *originalFunc, int op = TGSI_OPCODE_LAST); private: - llvm::LLVMFoldingBuilder m_builder; + llvm::IRBuilder m_builder; StorageSoa *m_storage; std::map m_functionsMap; diff --git a/src/gallium/auxiliary/gallivm/storage.cpp b/src/gallium/auxiliary/gallivm/storage.cpp index c4326de8c5..9d9fd12360 100644 --- a/src/gallium/auxiliary/gallivm/storage.cpp +++ b/src/gallium/auxiliary/gallivm/storage.cpp @@ -186,26 +186,26 @@ llvm::Value *Storage::maskWrite(llvm::Value *src, int mask, llvm::Value *templ) if ((mask & TGSI_WRITEMASK_X)) { llvm::Value *x = new ExtractElementInst(src, unsigned(0), name("x"), m_block); - dst = new InsertElementInst(dst, x, unsigned(0), - name("dstx"), m_block); + dst = InsertElementInst::Create(dst, x, unsigned(0), + name("dstx"), m_block); } if ((mask & TGSI_WRITEMASK_Y)) { llvm::Value *y = new ExtractElementInst(src, unsigned(1), name("y"), m_block); - dst = new InsertElementInst(dst, y, unsigned(1), - name("dsty"), m_block); + dst = InsertElementInst::Create(dst, y, unsigned(1), + name("dsty"), m_block); } if ((mask & TGSI_WRITEMASK_Z)) { llvm::Value *z = new ExtractElementInst(src, unsigned(2), name("z"), m_block); - dst = new InsertElementInst(dst, z, unsigned(2), - name("dstz"), m_block); + dst = InsertElementInst::Create(dst, z, unsigned(2), + name("dstz"), m_block); } if ((mask & TGSI_WRITEMASK_W)) { llvm::Value *w = new ExtractElementInst(src, unsigned(3), name("w"), m_block); - dst = new InsertElementInst(dst, w, unsigned(3), - name("dstw"), m_block); + dst = InsertElementInst::Create(dst, w, unsigned(3), + name("dstw"), m_block); } return dst; } @@ -308,11 +308,11 @@ llvm::Value * Storage::elemPtr(Args arg) std::vector indices; indices.push_back(constantInt(0)); indices.push_back(constantInt(static_cast(arg))); - GetElementPtrInst *getElem = new GetElementPtrInst(m_INPUT, - indices.begin(), - indices.end(), - name("input_ptr"), - m_block); + GetElementPtrInst *getElem = GetElementPtrInst::Create(m_INPUT, + indices.begin(), + indices.end(), + name("input_ptr"), + m_block); return new LoadInst(getElem, name("input_field"), false, m_block); } @@ -322,7 +322,7 @@ llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx, GetElementPtrInst *getElem = 0; if (indIdx) { - getElem = new GetElementPtrInst(ptr, + getElem = GetElementPtrInst::Create(ptr, BinaryOperator::create(Instruction::Add, indIdx, constantInt(idx), @@ -331,7 +331,7 @@ llvm::Value * Storage::elemIdx(llvm::Value *ptr, int idx, name("field"), m_block); } else { - getElem = new GetElementPtrInst(ptr, + getElem = GetElementPtrInst::Create(ptr, constantInt(idx), name("field"), m_block); @@ -350,7 +350,7 @@ void Storage::setKilElement(llvm::Value *val) std::vector indices; indices.push_back(constantInt(0)); indices.push_back(constantInt(static_cast(KilArg))); - GetElementPtrInst *elem = new GetElementPtrInst(m_INPUT, + GetElementPtrInst *elem = GetElementPtrInst::Create(m_INPUT, indices.begin(), indices.end(), name("kil_ptr"), diff --git a/src/gallium/auxiliary/gallivm/storagesoa.cpp b/src/gallium/auxiliary/gallivm/storagesoa.cpp index bb6fe3d7e1..0e6e68c9d7 100644 --- a/src/gallium/auxiliary/gallivm/storagesoa.cpp +++ b/src/gallium/auxiliary/gallivm/storagesoa.cpp @@ -207,11 +207,11 @@ llvm::Value * StorageSoa::elementPointer(llvm::Value *ptr, llvm::Value *index, indices.push_back(index); indices.push_back(constantInt(channel)); - GetElementPtrInst *getElem = new GetElementPtrInst(ptr, - indices.begin(), - indices.end(), - name("ptr"), - m_block); + GetElementPtrInst *getElem = GetElementPtrInst::Create(ptr, + indices.begin(), + indices.end(), + name("ptr"), + m_block); return getElem; } diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index ab9e7a06fb..ab8c851f14 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -1014,7 +1014,7 @@ tgsi_to_llvm(struct gallivm_ir *ir, const struct tgsi_token *tokens) Value *ptr_INPUT = args++; ptr_INPUT->setName("input"); - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + BasicBlock *label_entry = BasicBlock::Create("entry", shader, 0); tgsi_parse_init(&parse, tokens); @@ -1085,7 +1085,7 @@ llvm::Module * tgsi_to_llvmir(struct gallivm_ir *ir, Value *temps = args++; temps->setName("temps"); - BasicBlock *label_entry = new BasicBlock("entry", shader, 0); + BasicBlock *label_entry = BasicBlock::Create("entry", shader, 0); tgsi_parse_init(&parse, tokens); -- cgit v1.2.3 From c2afa182f31ed4a8be01078c2b8bedcd881cd157 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 20:25:06 +0100 Subject: mesa: remove st_cache.c --- src/mesa/state_tracker/st_cache.c | 181 -------------------------------------- 1 file changed, 181 deletions(-) delete mode 100644 src/mesa/state_tracker/st_cache.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_cache.c b/src/mesa/state_tracker/st_cache.c deleted file mode 100644 index 7ee4fadc37..0000000000 --- a/src/mesa/state_tracker/st_cache.c +++ /dev/null @@ -1,181 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Zack Rusin - */ - -#include "st_cache.h" - -#include "st_context.h" - -#include "pipe/p_state.h" - -#include "cso_cache/cso_cache.h" -#include "cso_cache/cso_hash.h" - - -/* Those function will either find the state of the given template - * in the cache or they will create a new state from the given - * template, insert it in the cache and return it. - */ - -/* - * If the driver returns 0 from the create method then they will assign - * the data member of the cso to be the template itself. - */ - -const struct cso_blend * st_cached_blend_state(struct st_context *st, - const struct pipe_blend_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_blend_state)); - struct cso_hash_iter iter = cso_find_state_template(st->cache, - hash_key, CSO_BLEND, - (void*)templ); - if (cso_hash_iter_is_null(iter)) { - struct cso_blend *cso = malloc(sizeof(struct cso_blend)); - memcpy(&cso->state, templ, sizeof(struct pipe_blend_state)); - cso->data = st->pipe->create_blend_state(st->pipe, &cso->state); - if (!cso->data) - cso->data = &cso->state; - cso->delete_state = (cso_state_callback)st->pipe->delete_blend_state; - cso->context = st->pipe; - iter = cso_insert_state(st->cache, hash_key, CSO_BLEND, cso); - } - return ((struct cso_blend *)cso_hash_iter_data(iter)); -} - -const struct cso_sampler * -st_cached_sampler_state(struct st_context *st, - const struct pipe_sampler_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_sampler_state)); - struct cso_hash_iter iter = cso_find_state_template(st->cache, - hash_key, CSO_SAMPLER, - (void*)templ); - if (cso_hash_iter_is_null(iter)) { - struct cso_sampler *cso = malloc(sizeof(struct cso_sampler)); - memcpy(&cso->state, templ, sizeof(struct pipe_sampler_state)); - cso->data = st->pipe->create_sampler_state(st->pipe, &cso->state); - if (!cso->data) - cso->data = &cso->state; - cso->delete_state = (cso_state_callback)st->pipe->delete_sampler_state; - cso->context = st->pipe; - iter = cso_insert_state(st->cache, hash_key, CSO_SAMPLER, cso); - } - return (struct cso_sampler*)(cso_hash_iter_data(iter)); -} - -const struct cso_depth_stencil_alpha * -st_cached_depth_stencil_alpha_state(struct st_context *st, - const struct pipe_depth_stencil_alpha_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_depth_stencil_alpha_state)); - struct cso_hash_iter iter = cso_find_state_template(st->cache, - hash_key, - CSO_DEPTH_STENCIL_ALPHA, - (void*)templ); - if (cso_hash_iter_is_null(iter)) { - struct cso_depth_stencil_alpha *cso = malloc(sizeof(struct cso_depth_stencil_alpha)); - memcpy(&cso->state, templ, sizeof(struct pipe_depth_stencil_alpha_state)); - cso->data = st->pipe->create_depth_stencil_alpha_state(st->pipe, &cso->state); - if (!cso->data) - cso->data = &cso->state; - cso->delete_state = (cso_state_callback)st->pipe->delete_depth_stencil_alpha_state; - cso->context = st->pipe; - iter = cso_insert_state(st->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso); - } - return (struct cso_depth_stencil_alpha*)(cso_hash_iter_data(iter)); -} - -const struct cso_rasterizer* st_cached_rasterizer_state( - struct st_context *st, - const struct pipe_rasterizer_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_rasterizer_state)); - struct cso_hash_iter iter = cso_find_state_template(st->cache, - hash_key, CSO_RASTERIZER, - (void*)templ); - if (cso_hash_iter_is_null(iter)) { - struct cso_rasterizer *cso = malloc(sizeof(struct cso_rasterizer)); - memcpy(&cso->state, templ, sizeof(struct pipe_rasterizer_state)); - cso->data = st->pipe->create_rasterizer_state(st->pipe, &cso->state); - if (!cso->data) - cso->data = &cso->state; - cso->delete_state = (cso_state_callback)st->pipe->delete_rasterizer_state; - cso->context = st->pipe; - iter = cso_insert_state(st->cache, hash_key, CSO_RASTERIZER, cso); - } - return (struct cso_rasterizer*)(cso_hash_iter_data(iter)); -} - -const struct cso_fragment_shader * -st_cached_fs_state(struct st_context *st, - const struct pipe_shader_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_shader_state)); - struct cso_hash_iter iter = cso_find_state_template(st->cache, - hash_key, CSO_FRAGMENT_SHADER, - (void*)templ); - if (cso_hash_iter_is_null(iter)) { - struct cso_fragment_shader *cso = malloc(sizeof(struct cso_fragment_shader)); - memcpy(&cso->state, templ, sizeof(struct pipe_shader_state)); - cso->data = st->pipe->create_fs_state(st->pipe, &cso->state); - if (!cso->data) - cso->data = &cso->state; - cso->delete_state = (cso_state_callback)st->pipe->delete_fs_state; - cso->context = st->pipe; - iter = cso_insert_state(st->cache, hash_key, CSO_FRAGMENT_SHADER, cso); - } - return (struct cso_fragment_shader*)(cso_hash_iter_data(iter)); -} - -const struct cso_vertex_shader * -st_cached_vs_state(struct st_context *st, - const struct pipe_shader_state *templ) -{ - unsigned hash_key = cso_construct_key((void*)templ, - sizeof(struct pipe_shader_state)); - struct cso_hash_iter iter = cso_find_state_template(st->cache, - hash_key, CSO_VERTEX_SHADER, - (void*)templ); - if (cso_hash_iter_is_null(iter)) { - struct cso_vertex_shader *cso = malloc(sizeof(struct cso_vertex_shader)); - memcpy(&cso->state, templ, sizeof(struct pipe_shader_state)); - cso->data = st->pipe->create_vs_state(st->pipe, &cso->state); - if (!cso->data) - cso->data = &cso->state; - cso->delete_state = (cso_state_callback)st->pipe->delete_vs_state; - cso->context = st->pipe; - iter = cso_insert_state(st->cache, hash_key, CSO_VERTEX_SHADER, cso); - } - return (struct cso_vertex_shader*)(cso_hash_iter_data(iter)); -} -- cgit v1.2.3 From 1dc5e56f3e48b629daa18c2d8631c96bda638eb6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 20:25:47 +0100 Subject: cso: provide functions to bind fs/vs handles directly --- src/gallium/auxiliary/cso_cache/cso_context.c | 32 ++++++++++++++++++++------- src/gallium/auxiliary/cso_cache/cso_context.h | 4 ++++ 2 files changed, 28 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 87995c80c3..e6366e7b14 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -422,6 +422,16 @@ void cso_restore_rasterizer(struct cso_context *ctx) } +void cso_set_fragment_shader_handle(struct cso_context *ctx, + void *handle ) +{ + if (ctx->fragment_shader != handle) { + ctx->fragment_shader = handle; + ctx->pipe->bind_fs_state(ctx->pipe, handle); + } +} + + void cso_set_fragment_shader(struct cso_context *ctx, const struct pipe_shader_state *templ) { @@ -453,10 +463,7 @@ void cso_set_fragment_shader(struct cso_context *ctx, handle = ((struct cso_fragment_shader *)cso_hash_iter_data(iter))->data; } - if (ctx->fragment_shader != handle) { - ctx->fragment_shader = handle; - ctx->pipe->bind_fs_state(ctx->pipe, handle); - } + cso_set_fragment_shader_handle( ctx, handle ); } void cso_save_fragment_shader(struct cso_context *ctx) @@ -476,6 +483,16 @@ void cso_restore_fragment_shader(struct cso_context *ctx) } +void cso_set_vertex_shader_handle(struct cso_context *ctx, + void *handle ) +{ + if (ctx->vertex_shader != handle) { + ctx->vertex_shader = handle; + ctx->pipe->bind_vs_state(ctx->pipe, handle); + } +} + + void cso_set_vertex_shader(struct cso_context *ctx, const struct pipe_shader_state *templ) @@ -503,12 +520,11 @@ void cso_set_vertex_shader(struct cso_context *ctx, handle = ((struct cso_vertex_shader *)cso_hash_iter_data(iter))->data; } - if (ctx->vertex_shader != handle) { - ctx->vertex_shader = handle; - ctx->pipe->bind_vs_state(ctx->pipe, handle); - } + cso_set_vertex_shader_handle( ctx, handle ); } + + void cso_save_vertex_shader(struct cso_context *ctx) { assert(!ctx->vertex_shader_saved); diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h index 665e8d9911..945f4881a8 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.h +++ b/src/gallium/auxiliary/cso_cache/cso_context.h @@ -96,6 +96,8 @@ void cso_restore_sampler_textures( struct cso_context *cso ); * (eg mesa's internall-generated texenv programs), it will be up to * the state tracker to implement their own specialized caching. */ +void cso_set_fragment_shader_handle(struct cso_context *ctx, + void *handle ); void cso_set_fragment_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); void cso_save_fragment_shader(struct cso_context *cso); @@ -103,6 +105,8 @@ void cso_restore_fragment_shader(struct cso_context *cso); +void cso_set_vertex_shader_handle(struct cso_context *ctx, + void *handle ); void cso_set_vertex_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); void cso_save_vertex_shader(struct cso_context *cso); -- cgit v1.2.3 From 01dfa6cde157321f565bab949f23f367ed20fa0e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 20:26:19 +0100 Subject: use cso fs/vs handle functions --- src/gallium/auxiliary/util/u_blit.c | 4 ++-- src/gallium/auxiliary/util/u_gen_mipmap.c | 4 ++-- src/mesa/state_tracker/st_atom_shader.c | 4 ++-- src/mesa/state_tracker/st_cb_bitmap.c | 4 ++-- src/mesa/state_tracker/st_cb_clear.c | 4 ++-- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- src/mesa/state_tracker/st_program.h | 4 ++-- 7 files changed, 14 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index be5e83e834..0938b03820 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -318,8 +318,8 @@ util_blit_pixels(struct blit_state *ctx, cso_set_sampler_textures(ctx->cso, 1, &tex); /* shaders */ - cso_set_fragment_shader(ctx->cso, ctx->fs); - cso_set_vertex_shader(ctx->cso, ctx->vs); + cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + cso_set_vertex_shader_handle(ctx->cso, ctx->vs); /* drawing dest */ memset(&fb, 0, sizeof(fb)); diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index f0c4063b28..9822a25ca6 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -888,8 +888,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); cso_set_rasterizer(ctx->cso, &ctx->rasterizer); - cso_set_fragment_shader(ctx->cso, ctx->fs); - cso_set_vertex_shader(ctx->cso, ctx->vs); + cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + cso_set_vertex_shader_handle(ctx->cso, ctx->vs); #if 0 cso_set_viewport(ctx->cso, &ctx->viewport); #endif diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 4a641a4a73..3f5ec71112 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -273,8 +273,8 @@ update_linkage( struct st_context *st ) st->vp = stvp; st->fp = stfp; - cso_set_vertex_shader(st->cso_context, stvp->driver_shader); - cso_set_fragment_shader(st->cso_context, stfp->driver_shader); + cso_set_vertex_shader_handle(st->cso_context, stvp->driver_shader); + cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader); st->vertex_result_to_slot = xvp->output_to_slot; } diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 1b863143e0..71a848ea54 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -446,10 +446,10 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_set_rasterizer(cso, &st->bitmap.rasterizer); /* fragment shader state: TEX lookup program */ - cso_set_fragment_shader(cso, stfp->driver_shader); + cso_set_fragment_shader_handle(cso, stfp->driver_shader); /* vertex shader state: position + texcoord pass-through */ - cso_set_vertex_shader(cso, st->bitmap.vs); + cso_set_vertex_shader_handle(cso, st->bitmap.vs); /* sampler / texture state */ cso_single_sampler(cso, 0, &st->bitmap.sampler); diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index bb27faad21..dac346a06c 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -274,8 +274,8 @@ clear_with_quad(GLcontext *ctx, cso_set_rasterizer(st->cso_context, &st->clear.raster); cso_set_viewport(st->cso_context, &st->clear.viewport); - cso_set_fragment_shader(st->cso_context, st->clear.fs); - cso_set_vertex_shader(st->cso_context, st->clear.vs); + cso_set_fragment_shader_handle(st->cso_context, st->clear.fs); + cso_set_vertex_shader_handle(st->cso_context, st->clear.vs); /* draw quad matching scissor rect (XXX verify coord round-off) */ draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 75261c3350..3921500659 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -545,10 +545,10 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, } /* fragment shader state: TEX lookup program */ - cso_set_fragment_shader(cso, stfp->driver_shader); + cso_set_fragment_shader_handle(cso, stfp->driver_shader); /* vertex shader state: position + texcoord pass-through */ - cso_set_vertex_shader(cso, stvp->driver_shader); + cso_set_vertex_shader_handle(cso, stvp->driver_shader); /* texture sampling state: */ diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 63d6590540..dced31e88e 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -61,7 +61,7 @@ struct st_fragment_program GLuint input_map[PIPE_MAX_SHADER_INPUTS]; struct pipe_shader_state state; - struct pipe_shader_state *driver_shader; + void *driver_shader; GLuint param_state; @@ -88,7 +88,7 @@ struct st_vertex_program GLuint num_inputs; struct pipe_shader_state state; - struct pipe_shader_state *driver_shader; + void *driver_shader; /** For using our private draw module (glRasterPos) */ struct draw_vertex_shader *draw_shader; -- cgit v1.2.3 From e406ad5912985920a0d796f1ada58b40316590ed Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 21:01:38 +0100 Subject: draw: squash a couple of memory leaks --- src/gallium/auxiliary/draw/draw_pt_fetch.c | 3 +++ src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 5 +++++ 2 files changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index c588710b75..f98bce6eac 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -166,6 +166,9 @@ struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) void draw_pt_fetch_destroy( struct pt_fetch *fetch ) { + if (fetch->translate) + fetch->translate->release( fetch->translate ); + FREE(fetch); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 7735173042..68b2c5b1e3 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -264,6 +264,11 @@ static void fetch_emit_finish( struct draw_pt_middle_end *middle ) static void fetch_emit_destroy( struct draw_pt_middle_end *middle ) { + struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; + + if (feme->translate) + feme->translate->release( feme->translate ); + FREE(middle); } -- cgit v1.2.3 From 9a0e6860d3b4602515f39593ed9af00cee7bb106 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 21 Apr 2008 21:05:11 +0100 Subject: cso: disable not-really-working cso_set_*_shader() funcs --- src/gallium/auxiliary/cso_cache/cso_context.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index e6366e7b14..3840dd2239 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -432,6 +432,9 @@ void cso_set_fragment_shader_handle(struct cso_context *ctx, } +/* Not really working: + */ +#if 0 void cso_set_fragment_shader(struct cso_context *ctx, const struct pipe_shader_state *templ) { @@ -465,6 +468,7 @@ void cso_set_fragment_shader(struct cso_context *ctx, cso_set_fragment_shader_handle( ctx, handle ); } +#endif void cso_save_fragment_shader(struct cso_context *ctx) { @@ -493,7 +497,9 @@ void cso_set_vertex_shader_handle(struct cso_context *ctx, } - +/* Not really working: + */ +#if 0 void cso_set_vertex_shader(struct cso_context *ctx, const struct pipe_shader_state *templ) { @@ -522,6 +528,7 @@ void cso_set_vertex_shader(struct cso_context *ctx, cso_set_vertex_shader_handle( ctx, handle ); } +#endif -- cgit v1.2.3 From 9fe63929011cd9c4d86ab6525555a3e53423c854 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 21 Apr 2008 15:44:51 -0400 Subject: initial stab at error handling in cso --- src/gallium/auxiliary/cso_cache/cso_context.c | 98 +++++++++++++++++---------- src/gallium/auxiliary/cso_cache/cso_context.h | 34 +++++----- 2 files changed, 80 insertions(+), 52 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 3840dd2239..6040522a46 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -139,8 +139,8 @@ void cso_destroy_context( struct cso_context *ctx ) * the data member of the cso to be the template itself. */ -void cso_set_blend(struct cso_context *ctx, - const struct pipe_blend_state *templ) +enum pipe_error cso_set_blend(struct cso_context *ctx, + const struct pipe_blend_state *templ) { unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_blend_state)); struct cso_hash_iter iter = cso_find_state_template(ctx->cache, @@ -149,8 +149,9 @@ void cso_set_blend(struct cso_context *ctx, void *handle; if (cso_hash_iter_is_null(iter)) { - /* FIXME: handle OOM */ struct cso_blend *cso = MALLOC(sizeof(struct cso_blend)); + if (!cso) + return PIPE_ERROR_OUT_OF_MEMORY; cso->state = *templ; cso->data = ctx->pipe->create_blend_state(ctx->pipe, &cso->state); @@ -168,6 +169,7 @@ void cso_set_blend(struct cso_context *ctx, ctx->blend = handle; ctx->pipe->bind_blend_state(ctx->pipe, handle); } + return PIPE_OK; } void cso_save_blend(struct cso_context *ctx) @@ -187,12 +189,12 @@ void cso_restore_blend(struct cso_context *ctx) -void cso_single_sampler(struct cso_context *ctx, - unsigned idx, - const struct pipe_sampler_state *templ) +enum pipe_error cso_single_sampler(struct cso_context *ctx, + unsigned idx, + const struct pipe_sampler_state *templ) { void *handle = NULL; - + if (templ != NULL) { unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_sampler_state)); struct cso_hash_iter iter = cso_find_state_template(ctx->cache, @@ -200,9 +202,10 @@ void cso_single_sampler(struct cso_context *ctx, (void*)templ); if (cso_hash_iter_is_null(iter)) { - /* FIXME: handle OOM */ struct cso_sampler *cso = MALLOC(sizeof(struct cso_sampler)); - + if (!cso) + return PIPE_ERROR_OUT_OF_MEMORY; + cso->state = *templ; cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state; @@ -217,11 +220,12 @@ void cso_single_sampler(struct cso_context *ctx, } ctx->samplers[idx] = handle; + return PIPE_OK; } void cso_single_sampler_done( struct cso_context *ctx ) { - unsigned i; + unsigned i; /* find highest non-null sampler */ for (i = PIPE_MAX_SAMPLERS; i > 0; i--) { @@ -232,8 +236,8 @@ void cso_single_sampler_done( struct cso_context *ctx ) ctx->nr_samplers = i; if (ctx->hw.nr_samplers != ctx->nr_samplers || - memcmp(ctx->hw.samplers, - ctx->samplers, + memcmp(ctx->hw.samplers, + ctx->samplers, ctx->nr_samplers * sizeof(void *)) != 0) { memcpy(ctx->hw.samplers, ctx->samplers, ctx->nr_samplers * sizeof(void *)); @@ -243,22 +247,36 @@ void cso_single_sampler_done( struct cso_context *ctx ) } } -void cso_set_samplers( struct cso_context *ctx, - unsigned nr, - const struct pipe_sampler_state **templates ) +/* + * If the function encouters any errors it will return the + * last one. Done to always try to set as many samplers + * as possible. + */ +enum pipe_error cso_set_samplers( struct cso_context *ctx, + unsigned nr, + const struct pipe_sampler_state **templates ) { unsigned i; - + enum pipe_error temp, error = PIPE_OK; + /* TODO: fastpath */ - for (i = 0; i < nr; i++) - cso_single_sampler( ctx, i, templates[i] ); + for (i = 0; i < nr; i++) { + temp = cso_single_sampler( ctx, i, templates[i] ); + if (temp != PIPE_OK) + error = temp; + } + + for ( ; i < ctx->nr_samplers; i++) { + temp = cso_single_sampler( ctx, i, NULL ); + if (temp != PIPE_OK) + error = temp; + } - for ( ; i < ctx->nr_samplers; i++) - cso_single_sampler( ctx, i, NULL ); - cso_single_sampler_done( ctx ); + + return error; } void cso_save_samplers(struct cso_context *ctx) @@ -294,7 +312,7 @@ void cso_set_sampler_textures( struct cso_context *ctx, void cso_save_sampler_textures( struct cso_context *ctx ) { uint i; - + ctx->nr_textures_saved = ctx->nr_textures; for (i = 0; i < ctx->nr_textures; i++) { assert(!ctx->textures_saved[i]); @@ -323,9 +341,8 @@ void cso_restore_sampler_textures( struct cso_context *ctx ) - -void cso_set_depth_stencil_alpha(struct cso_context *ctx, - const struct pipe_depth_stencil_alpha_state *templ) +enum pipe_error cso_set_depth_stencil_alpha(struct cso_context *ctx, + const struct pipe_depth_stencil_alpha_state *templ) { unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_depth_stencil_alpha_state)); @@ -336,8 +353,9 @@ void cso_set_depth_stencil_alpha(struct cso_context *ctx, void *handle; if (cso_hash_iter_is_null(iter)) { - /* FIXME: handle OOM */ struct cso_depth_stencil_alpha *cso = MALLOC(sizeof(struct cso_depth_stencil_alpha)); + if (!cso) + return PIPE_ERROR_OUT_OF_MEMORY; cso->state = *templ; cso->data = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &cso->state); @@ -355,6 +373,7 @@ void cso_set_depth_stencil_alpha(struct cso_context *ctx, ctx->depth_stencil = handle; ctx->pipe->bind_depth_stencil_alpha_state(ctx->pipe, handle); } + return PIPE_OK; } void cso_save_depth_stencil_alpha(struct cso_context *ctx) @@ -374,8 +393,8 @@ void cso_restore_depth_stencil_alpha(struct cso_context *ctx) -void cso_set_rasterizer(struct cso_context *ctx, - const struct pipe_rasterizer_state *templ) +enum pipe_error cso_set_rasterizer(struct cso_context *ctx, + const struct pipe_rasterizer_state *templ) { unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_rasterizer_state)); @@ -385,8 +404,9 @@ void cso_set_rasterizer(struct cso_context *ctx, void *handle = NULL; if (cso_hash_iter_is_null(iter)) { - /* FIXME: handle OOM */ struct cso_rasterizer *cso = MALLOC(sizeof(struct cso_rasterizer)); + if (!cso) + return PIPE_ERROR_OUT_OF_MEMORY; cso->state = *templ; cso->data = ctx->pipe->create_rasterizer_state(ctx->pipe, &cso->state); @@ -404,6 +424,7 @@ void cso_set_rasterizer(struct cso_context *ctx, ctx->rasterizer = handle; ctx->pipe->bind_rasterizer_state(ctx->pipe, handle); } + return PIPE_OK; } void cso_save_rasterizer(struct cso_context *ctx) @@ -422,6 +443,7 @@ void cso_restore_rasterizer(struct cso_context *ctx) } +<<<<<<< HEAD:src/gallium/auxiliary/cso_cache/cso_context.c void cso_set_fragment_shader_handle(struct cso_context *ctx, void *handle ) { @@ -435,8 +457,8 @@ void cso_set_fragment_shader_handle(struct cso_context *ctx, /* Not really working: */ #if 0 -void cso_set_fragment_shader(struct cso_context *ctx, - const struct pipe_shader_state *templ) +enum pipe_error cso_set_fragment_shader(struct cso_context *ctx, + const struct pipe_shader_state *templ) { const struct tgsi_token *tokens = templ->tokens; unsigned num_tokens = tgsi_num_tokens(tokens); @@ -449,10 +471,12 @@ void cso_set_fragment_shader(struct cso_context *ctx, void *handle = NULL; if (cso_hash_iter_is_null(iter)) { - /* FIXME: handle OOM */ struct cso_fragment_shader *cso = MALLOC(sizeof(struct cso_fragment_shader) + tokens_size); struct tgsi_token *cso_tokens = (struct tgsi_token *)((char *)cso + sizeof(*cso)); + if (!cso) + return PIPE_ERROR_OUT_OF_MEMORY; + memcpy(cso_tokens, tokens, tokens_size); cso->state.tokens = cso_tokens; cso->data = ctx->pipe->create_fs_state(ctx->pipe, &cso->state); @@ -467,6 +491,7 @@ void cso_set_fragment_shader(struct cso_context *ctx, } cso_set_fragment_shader_handle( ctx, handle ); + return PIPE_OK; } #endif @@ -500,8 +525,8 @@ void cso_set_vertex_shader_handle(struct cso_context *ctx, /* Not really working: */ #if 0 -void cso_set_vertex_shader(struct cso_context *ctx, - const struct pipe_shader_state *templ) +enum pipe_error cso_set_vertex_shader(struct cso_context *ctx, + const struct pipe_shader_state *templ) { unsigned hash_key = cso_construct_key((void*)templ, sizeof(struct pipe_shader_state)); @@ -511,9 +536,11 @@ void cso_set_vertex_shader(struct cso_context *ctx, void *handle = NULL; if (cso_hash_iter_is_null(iter)) { - /* FIXME: handle OOM */ struct cso_vertex_shader *cso = MALLOC(sizeof(struct cso_vertex_shader)); + if (!cso) + return PIPE_ERROR_OUT_OF_MEMORY; + cso->state = *templ; cso->data = ctx->pipe->create_vs_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_vs_state; @@ -527,6 +554,7 @@ void cso_set_vertex_shader(struct cso_context *ctx, } cso_set_vertex_shader_handle( ctx, handle ); + return PIPE_OK; } #endif diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h index 945f4881a8..ab46b93d5a 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.h +++ b/src/gallium/auxiliary/cso_cache/cso_context.h @@ -31,6 +31,7 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" +#include "pipe/p_error.h" #ifdef __cplusplus @@ -45,39 +46,39 @@ void cso_destroy_context( struct cso_context *cso ); -void cso_set_blend( struct cso_context *cso, - const struct pipe_blend_state *blend ); +enum pipe_error cso_set_blend( struct cso_context *cso, + const struct pipe_blend_state *blend ); void cso_save_blend(struct cso_context *cso); void cso_restore_blend(struct cso_context *cso); -void cso_set_depth_stencil_alpha( struct cso_context *cso, - const struct pipe_depth_stencil_alpha_state *dsa ); +enum pipe_error cso_set_depth_stencil_alpha( struct cso_context *cso, + const struct pipe_depth_stencil_alpha_state *dsa ); void cso_save_depth_stencil_alpha(struct cso_context *cso); void cso_restore_depth_stencil_alpha(struct cso_context *cso); -void cso_set_rasterizer( struct cso_context *cso, - const struct pipe_rasterizer_state *rasterizer ); +enum pipe_error cso_set_rasterizer( struct cso_context *cso, + const struct pipe_rasterizer_state *rasterizer ); void cso_save_rasterizer(struct cso_context *cso); void cso_restore_rasterizer(struct cso_context *cso); -void cso_set_samplers( struct cso_context *cso, - unsigned count, - const struct pipe_sampler_state **states ); +enum pipe_error cso_set_samplers( struct cso_context *cso, + unsigned count, + const struct pipe_sampler_state **states ); void cso_save_samplers(struct cso_context *cso); void cso_restore_samplers(struct cso_context *cso); /* Alternate interface to support state trackers that like to modify * samplers one at a time: */ -void cso_single_sampler( struct cso_context *cso, - unsigned nr, - const struct pipe_sampler_state *states ); +enum pipe_error cso_single_sampler( struct cso_context *cso, + unsigned nr, + const struct pipe_sampler_state *states ); void cso_single_sampler_done( struct cso_context *cso ); @@ -98,17 +99,16 @@ void cso_restore_sampler_textures( struct cso_context *cso ); */ void cso_set_fragment_shader_handle(struct cso_context *ctx, void *handle ); -void cso_set_fragment_shader( struct cso_context *cso, - const struct pipe_shader_state *shader ); +enum pipe_error cso_set_fragment_shader( struct cso_context *cso, + const struct pipe_shader_state *shader ); void cso_save_fragment_shader(struct cso_context *cso); void cso_restore_fragment_shader(struct cso_context *cso); - void cso_set_vertex_shader_handle(struct cso_context *ctx, void *handle ); -void cso_set_vertex_shader( struct cso_context *cso, - const struct pipe_shader_state *shader ); +enum pipe_error cso_set_vertex_shader( struct cso_context *cso, + const struct pipe_shader_state *shader ); void cso_save_vertex_shader(struct cso_context *cso); void cso_restore_vertex_shader(struct cso_context *cso); -- cgit v1.2.3 From 0879237725eca893318137b795d4234300a37e9a Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 21 Apr 2008 16:04:27 -0400 Subject: handle some of the possible allocation failures within the hash itself --- src/gallium/auxiliary/cso_cache/cso_context.c | 36 ++++++++++++++++++++++++--- src/gallium/auxiliary/cso_cache/cso_hash.c | 17 +++++++++++++ 2 files changed, 49 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 6040522a46..364a413580 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -159,6 +159,11 @@ enum pipe_error cso_set_blend(struct cso_context *ctx, cso->context = ctx->pipe; iter = cso_insert_state(ctx->cache, hash_key, CSO_BLEND, cso); + if (cso_hash_iter_is_null(iter)) { + FREE(cso); + return PIPE_ERROR_OUT_OF_MEMORY; + } + handle = cso->data; } else { @@ -212,6 +217,11 @@ enum pipe_error cso_single_sampler(struct cso_context *ctx, cso->context = ctx->pipe; iter = cso_insert_state(ctx->cache, hash_key, CSO_SAMPLER, cso); + if (cso_hash_iter_is_null(iter)) { + FREE(cso); + return PIPE_ERROR_OUT_OF_MEMORY; + } + handle = cso->data; } else { @@ -362,7 +372,12 @@ enum pipe_error cso_set_depth_stencil_alpha(struct cso_context *ctx, cso->delete_state = (cso_state_callback)ctx->pipe->delete_depth_stencil_alpha_state; cso->context = ctx->pipe; - cso_insert_state(ctx->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso); + iter = cso_insert_state(ctx->cache, hash_key, CSO_DEPTH_STENCIL_ALPHA, cso); + if (cso_hash_iter_is_null(iter)) { + FREE(cso); + return PIPE_ERROR_OUT_OF_MEMORY; + } + handle = cso->data; } else { @@ -413,7 +428,12 @@ enum pipe_error cso_set_rasterizer(struct cso_context *ctx, cso->delete_state = (cso_state_callback)ctx->pipe->delete_rasterizer_state; cso->context = ctx->pipe; - cso_insert_state(ctx->cache, hash_key, CSO_RASTERIZER, cso); + iter = cso_insert_state(ctx->cache, hash_key, CSO_RASTERIZER, cso); + if (cso_hash_iter_is_null(iter)) { + FREE(cso); + return PIPE_ERROR_OUT_OF_MEMORY; + } + handle = cso->data; } else { @@ -442,8 +462,6 @@ void cso_restore_rasterizer(struct cso_context *ctx) ctx->rasterizer_saved = NULL; } - -<<<<<<< HEAD:src/gallium/auxiliary/cso_cache/cso_context.c void cso_set_fragment_shader_handle(struct cso_context *ctx, void *handle ) { @@ -484,6 +502,11 @@ enum pipe_error cso_set_fragment_shader(struct cso_context *ctx, cso->context = ctx->pipe; iter = cso_insert_state(ctx->cache, hash_key, CSO_FRAGMENT_SHADER, cso); + if (cso_hash_iter_is_null(iter)) { + FREE(cso); + return PIPE_ERROR_OUT_OF_MEMORY; + } + handle = cso->data; } else { @@ -547,6 +570,11 @@ enum pipe_error cso_set_vertex_shader(struct cso_context *ctx, cso->context = ctx->pipe; iter = cso_insert_state(ctx->cache, hash_key, CSO_VERTEX_SHADER, cso); + if (cso_hash_iter_is_null(iter)) { + FREE(cso); + return PIPE_ERROR_OUT_OF_MEMORY; + } + handle = cso->data; } else { diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index ddce3822f7..8d867f86d2 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -110,6 +110,10 @@ cso_hash_create_node(struct cso_hash *hash, struct cso_node **anextNode) { struct cso_node *node = cso_data_allocate_node(hash->data.d); + + if (!node) + return NULL; + node->key = akey; node->value = avalue; @@ -219,6 +223,11 @@ struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, { struct cso_node **nextNode = cso_hash_find_node(hash, key); struct cso_node *node = cso_hash_create_node(hash, key, data, nextNode); + if (!node) { + struct cso_hash_iter null_iter = {hash, 0}; + return null_iter; + } + struct cso_hash_iter iter = {hash, node}; return iter; } @@ -227,7 +236,15 @@ struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, struct cso_hash * cso_hash_create(void) { struct cso_hash *hash = MALLOC_STRUCT(cso_hash); + if (!hash) + return NULL; + hash->data.d = MALLOC_STRUCT(cso_hash_data); + if (!hash->data.d) { + FREE(hash); + return NULL; + } + hash->data.d->fakeNext = 0; hash->data.d->buckets = 0; hash->data.d->size = 0; -- cgit v1.2.3 From ed187d39a6e0fd921b2a45a143d88ac4b66eee91 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 21 Apr 2008 16:40:27 -0400 Subject: make the api consistent (all set functions return pipe_error status) --- src/gallium/auxiliary/cso_cache/cso_context.c | 39 +++++++++++++++------------ src/gallium/auxiliary/cso_cache/cso_context.h | 26 +++++++++--------- 2 files changed, 35 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 364a413580..2d98a34edf 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -303,9 +303,9 @@ void cso_restore_samplers(struct cso_context *ctx) } -void cso_set_sampler_textures( struct cso_context *ctx, - uint count, - struct pipe_texture **textures ) +enum pipe_error cso_set_sampler_textures( struct cso_context *ctx, + uint count, + struct pipe_texture **textures ) { uint i; @@ -317,6 +317,8 @@ void cso_set_sampler_textures( struct cso_context *ctx, pipe_texture_reference(&ctx->textures[i], NULL); ctx->pipe->set_sampler_textures(ctx->pipe, count, textures); + + return PIPE_OK; } void cso_save_sampler_textures( struct cso_context *ctx ) @@ -462,13 +464,14 @@ void cso_restore_rasterizer(struct cso_context *ctx) ctx->rasterizer_saved = NULL; } -void cso_set_fragment_shader_handle(struct cso_context *ctx, - void *handle ) +enum pipe_error cso_set_fragment_shader_handle(struct cso_context *ctx, + void *handle ) { if (ctx->fragment_shader != handle) { ctx->fragment_shader = handle; ctx->pipe->bind_fs_state(ctx->pipe, handle); } + return PIPE_OK; } @@ -513,8 +516,7 @@ enum pipe_error cso_set_fragment_shader(struct cso_context *ctx, handle = ((struct cso_fragment_shader *)cso_hash_iter_data(iter))->data; } - cso_set_fragment_shader_handle( ctx, handle ); - return PIPE_OK; + return cso_set_fragment_shader_handle( ctx, handle ); } #endif @@ -535,13 +537,14 @@ void cso_restore_fragment_shader(struct cso_context *ctx) } -void cso_set_vertex_shader_handle(struct cso_context *ctx, - void *handle ) +enum pipe_error cso_set_vertex_shader_handle(struct cso_context *ctx, + void *handle ) { if (ctx->vertex_shader != handle) { ctx->vertex_shader = handle; ctx->pipe->bind_vs_state(ctx->pipe, handle); } + return PIPE_OK; } @@ -581,8 +584,7 @@ enum pipe_error cso_set_vertex_shader(struct cso_context *ctx, handle = ((struct cso_vertex_shader *)cso_hash_iter_data(iter))->data; } - cso_set_vertex_shader_handle( ctx, handle ); - return PIPE_OK; + return cso_set_vertex_shader_handle( ctx, handle ); } #endif @@ -606,14 +608,15 @@ void cso_restore_vertex_shader(struct cso_context *ctx) -void cso_set_framebuffer(struct cso_context *ctx, - const struct pipe_framebuffer_state *fb) +enum pipe_error cso_set_framebuffer(struct cso_context *ctx, + const struct pipe_framebuffer_state *fb) { /* XXX this memcmp() fails to detect buffer size changes */ if (1/*memcmp(&ctx->fb, fb, sizeof(*fb))*/) { ctx->fb = *fb; ctx->pipe->set_framebuffer_state(ctx->pipe, fb); } + return PIPE_OK; } void cso_save_framebuffer(struct cso_context *ctx) @@ -630,13 +633,14 @@ void cso_restore_framebuffer(struct cso_context *ctx) } -void cso_set_viewport(struct cso_context *ctx, - const struct pipe_viewport_state *vp) +enum pipe_error cso_set_viewport(struct cso_context *ctx, + const struct pipe_viewport_state *vp) { if (memcmp(&ctx->vp, vp, sizeof(*vp))) { ctx->vp = *vp; ctx->pipe->set_viewport_state(ctx->pipe, vp); } + return PIPE_OK; } void cso_save_viewport(struct cso_context *ctx) @@ -656,11 +660,12 @@ void cso_restore_viewport(struct cso_context *ctx) -void cso_set_blend_color(struct cso_context *ctx, - const struct pipe_blend_color *bc) +enum pipe_error cso_set_blend_color(struct cso_context *ctx, + const struct pipe_blend_color *bc) { if (memcmp(&ctx->blend_color, bc, sizeof(ctx->blend_color))) { ctx->blend_color = *bc; ctx->pipe->set_blend_color(ctx->pipe, bc); } + return PIPE_OK; } diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h index ab46b93d5a..0405944132 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.h +++ b/src/gallium/auxiliary/cso_cache/cso_context.h @@ -84,9 +84,9 @@ void cso_single_sampler_done( struct cso_context *cso ); -void cso_set_sampler_textures( struct cso_context *cso, - uint count, - struct pipe_texture **textures ); +enum pipe_error cso_set_sampler_textures( struct cso_context *cso, + uint count, + struct pipe_texture **textures ); void cso_save_sampler_textures( struct cso_context *cso ); void cso_restore_sampler_textures( struct cso_context *cso ); @@ -97,16 +97,16 @@ void cso_restore_sampler_textures( struct cso_context *cso ); * (eg mesa's internall-generated texenv programs), it will be up to * the state tracker to implement their own specialized caching. */ -void cso_set_fragment_shader_handle(struct cso_context *ctx, - void *handle ); +enum pipe_error cso_set_fragment_shader_handle(struct cso_context *ctx, + void *handle ); enum pipe_error cso_set_fragment_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); void cso_save_fragment_shader(struct cso_context *cso); void cso_restore_fragment_shader(struct cso_context *cso); -void cso_set_vertex_shader_handle(struct cso_context *ctx, - void *handle ); +enum pipe_error cso_set_vertex_shader_handle(struct cso_context *ctx, + void *handle ); enum pipe_error cso_set_vertex_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); void cso_save_vertex_shader(struct cso_context *cso); @@ -114,20 +114,20 @@ void cso_restore_vertex_shader(struct cso_context *cso); -void cso_set_framebuffer(struct cso_context *cso, - const struct pipe_framebuffer_state *fb); +enum pipe_error cso_set_framebuffer(struct cso_context *cso, + const struct pipe_framebuffer_state *fb); void cso_save_framebuffer(struct cso_context *cso); void cso_restore_framebuffer(struct cso_context *cso); -void cso_set_viewport(struct cso_context *cso, - const struct pipe_viewport_state *vp); +enum pipe_error cso_set_viewport(struct cso_context *cso, + const struct pipe_viewport_state *vp); void cso_save_viewport(struct cso_context *cso); void cso_restore_viewport(struct cso_context *cso); -void cso_set_blend_color(struct cso_context *cso, - const struct pipe_blend_color *bc); +enum pipe_error cso_set_blend_color(struct cso_context *cso, + const struct pipe_blend_color *bc); #ifdef __cplusplus -- cgit v1.2.3 From 83fec372b45eb0af9e2d83549b3d92afb17c38af Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 22 Apr 2008 11:26:26 +0200 Subject: cso: Fix build on Win32. --- src/gallium/auxiliary/cso_cache/cso_hash.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index 8d867f86d2..0646efd952 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -228,8 +228,10 @@ struct cso_hash_iter cso_hash_insert(struct cso_hash *hash, return null_iter; } - struct cso_hash_iter iter = {hash, node}; - return iter; + { + struct cso_hash_iter iter = {hash, node}; + return iter; + } } } -- cgit v1.2.3 From 09b668615852eb28cb6289baf84faaf3b6ccc3c2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 10:57:06 +0100 Subject: softpipe: make NUM_ENTRIES 32 so that div/mods are easier --- src/gallium/drivers/softpipe/sp_tile_cache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 19f71887e7..edafd93d8b 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -39,7 +39,7 @@ #include "sp_surface.h" #include "sp_tile_cache.h" -#define NUM_ENTRIES 30 +#define NUM_ENTRIES 32 /** XXX move these */ -- cgit v1.2.3 From a5b87f249ef79b1a8d8b9dbe72879b7ac9eb133c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 11:01:41 +0100 Subject: softpipe: implement SP_NO_RAST env var --- src/gallium/drivers/softpipe/sp_clear.c | 3 +++ src/gallium/drivers/softpipe/sp_context.c | 3 +++ src/gallium/drivers/softpipe/sp_context.h | 2 ++ src/gallium/drivers/softpipe/sp_setup.c | 10 ++++++++++ 4 files changed, 18 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_clear.c b/src/gallium/drivers/softpipe/sp_clear.c index 39aed151c7..1236706891 100644 --- a/src/gallium/drivers/softpipe/sp_clear.c +++ b/src/gallium/drivers/softpipe/sp_clear.c @@ -49,6 +49,9 @@ softpipe_clear(struct pipe_context *pipe, struct pipe_surface *ps, struct softpipe_context *softpipe = softpipe_context(pipe); uint i; + if (softpipe->no_rast) + return; + #if 0 softpipe_update_derived(softpipe); /* not needed?? */ #endif diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 8c84ddbe19..200fb415ac 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -219,6 +219,9 @@ softpipe_create( struct pipe_screen *screen, assert(softpipe->draw); softpipe->setup = sp_draw_render_stage(softpipe); + if (GETENV( "SP_NO_RAST" ) != NULL) + softpipe->no_rast = TRUE; + if (GETENV( "SP_VBUF" ) != NULL) { sp_init_vbuf(softpipe); } diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 0e1d5e561d..b3e2b2e435 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -86,6 +86,8 @@ struct softpipe_context { unsigned num_vertex_elements; unsigned num_vertex_buffers; + boolean no_rast; + /* Counter for occlusion queries. Note this supports overlapping * queries. */ diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 813d703108..0bf2b2dc3e 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -726,6 +726,9 @@ void setup_tri( struct setup_context *setup, { float det = calc_det(v0, v1, v2); + if (setup->softpipe->no_rast) + return; + /* debug_printf("%s\n", __FUNCTION__ ); */ @@ -934,6 +937,9 @@ setup_line(struct setup_context *setup, int dy = y1 - y0; int xstep, ystep; + if (setup->softpipe->no_rast) + return; + if (dx == 0 && dy == 0) return; @@ -1052,6 +1058,10 @@ setup_point( struct setup_context *setup, const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); uint fragSlot; + + if (softpipe->no_rast) + return; + /* For points, all interpolants are constant-valued. * However, for point sprites, we'll need to setup texcoords appropriately. * XXX: which coefficients are the texcoords??? -- cgit v1.2.3 From d07ed9216cc7033fecb5bce661bbaf79189bb391 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 11:01:56 +0100 Subject: xlib: implement SP_NO_RAST env var --- src/gallium/winsys/xlib/xm_winsys.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 9a20bdfb69..5a424d0ac7 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -84,6 +84,7 @@ struct xmesa_surface struct pipe_surface surface; int tileSize; + boolean no_swap; }; @@ -252,6 +253,9 @@ xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) const struct xmesa_surface *xm_surf = xmesa_surface((struct pipe_surface *) surf); + if (xm_surf->no_swap) + return; + if (xm_surf->tileSize) { xmesa_display_surface_tiled(b, surf); return; @@ -529,6 +533,13 @@ static struct pipe_surface * xm_surface_alloc(struct pipe_winsys *ws) { struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); + static boolean no_swap = 0; + static boolean firsttime = 1; + + if (firsttime) { + no_swap = getenv("SP_NO_RAST") != NULL; + firsttime = 0; + } assert(ws); @@ -540,7 +551,9 @@ xm_surface_alloc(struct pipe_winsys *ws) xms->tileSize = 32; /** probably temporary */ } #endif - + + xms->no_swap = no_swap; + return &xms->surface; } -- cgit v1.2.3 From 6a9f6625b38c3669769568d122958993f4a8d5b3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 11:15:51 +0100 Subject: xlib: shortcircuit no-op makecurrent --- src/gallium/winsys/xlib/fakeglx.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/xlib/fakeglx.c b/src/gallium/winsys/xlib/fakeglx.c index 902a755075..6e04cb4117 100644 --- a/src/gallium/winsys/xlib/fakeglx.c +++ b/src/gallium/winsys/xlib/fakeglx.c @@ -1504,6 +1504,13 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, #endif } + if (MakeCurrent_PrevContext == ctx && + MakeCurrent_PrevDrawable == draw && + MakeCurrent_PrevReadable == read && + MakeCurrent_PrevDrawBuffer == drawBuffer && + MakeCurrent_PrevReadBuffer == readBuffer) + return True; + MakeCurrent_PrevContext = ctx; MakeCurrent_PrevDrawable = draw; MakeCurrent_PrevReadable = read; -- cgit v1.2.3 From 57987ea67320e79e4c2d7e66388806148ece09b5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 11:26:07 +0100 Subject: draw: disable broken edgeflag code - didn't work & was killing performance --- src/gallium/auxiliary/draw/draw_pt_vcache.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 3cc2941f96..153055417d 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -106,7 +106,7 @@ static unsigned add_edgeflag( struct vcache_frontend *vcache, unsigned idx, unsigned mask ) { - if (mask && draw_pt_get_edgeflag(vcache->draw, idx)) + if (0 && mask && draw_pt_get_edgeflag(vcache->draw, idx)) return idx | DRAW_PT_EDGEFLAG; else return idx; @@ -116,7 +116,7 @@ static unsigned add_edgeflag( struct vcache_frontend *vcache, static unsigned add_reset_stipple( unsigned idx, unsigned reset ) { - if (reset) + if (0 && reset) return idx | DRAW_PT_RESET_STIPPLE; else return idx; @@ -128,9 +128,9 @@ static void vcache_triangle( struct vcache_frontend *vcache, unsigned i1, unsigned i2 ) { - vcache_elt(vcache, i0 | DRAW_PT_EDGEFLAG | DRAW_PT_RESET_STIPPLE); - vcache_elt(vcache, i1 | DRAW_PT_EDGEFLAG); - vcache_elt(vcache, i2 | DRAW_PT_EDGEFLAG); + vcache_elt(vcache, i0 /* | DRAW_PT_EDGEFLAG | DRAW_PT_RESET_STIPPLE */ ); + vcache_elt(vcache, i1 /* | DRAW_PT_EDGEFLAG */); + vcache_elt(vcache, i2 /* | DRAW_PT_EDGEFLAG */); vcache_check_flush(vcache); } @@ -142,11 +142,12 @@ static void vcache_ef_triangle( struct vcache_frontend *vcache, unsigned i1, unsigned i2 ) { +/* i0 = add_edgeflag( vcache, i0, (ef_mask >> 0) & 1 ); i1 = add_edgeflag( vcache, i1, (ef_mask >> 1) & 1 ); i2 = add_edgeflag( vcache, i2, (ef_mask >> 2) & 1 ); - i0 = add_reset_stipple( i0, reset_stipple ); +*/ vcache_elt(vcache, i0); vcache_elt(vcache, i1); -- cgit v1.2.3 From b4b3a73bdf68adc1d9fbadac913aa6cf60d648d5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 22 Apr 2008 19:47:02 +0900 Subject: pipebuffer: Temporarily allow simultaneous CPU writes. Also, fast path for re-fencing the same buffer multiple times with the same fence. --- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 18 +++++++++++++++++- 1 file changed, 17 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 65b6584003..27032b0c4c 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -215,15 +215,21 @@ fenced_buffer_serialize(struct fenced_buffer *fenced_buf, unsigned flags) struct fenced_buffer_list *fenced_list = fenced_buf->list; struct pipe_winsys *winsys = fenced_list->winsys; + /* Allow concurrent reads */ if(((fenced_buf->flags | flags) & PIPE_BUFFER_USAGE_WRITE) == 0) return PIPE_OK; + /* Wait for the CPU to finish */ if(fenced_buf->mapcount) { - /* FIXME */ + /* FIXME: Use thread conditions variables to signal when mapcount + * reaches zero */ debug_warning("attemp to write concurrently to buffer"); + /* XXX: we must not fail here in order to support texture mipmap generation return PIPE_ERROR_RETRY; + */ } + /* Wait for the GPU to finish */ if(fenced_buf->fence) { if(winsys->fence_finish(winsys, fenced_buf->fence, 0) != 0) return PIPE_ERROR_RETRY; @@ -353,6 +359,16 @@ buffer_fence(struct pb_buffer *buf, /* FIXME: receive this as a parameter */ unsigned flags = fence ? PIPE_BUFFER_USAGE_GPU_READ_WRITE : 0; + if(fence == fenced_buf->fence) { + /* Handle the same fence case specially, not only because it is a fast + * path, but mostly to avoid serializing two writes with the same fence, + * as that would bring the hardware down to synchronous operation without + * any benefit. + */ + fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE; + return; + } + if(fenced_buffer_serialize(fenced_buf, flags) != PIPE_OK) { /* FIXME: propagate error */ (void)0; -- cgit v1.2.3 From 57b85e197b33d39063d431500016dffcce46ab7c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 22 Apr 2008 12:18:31 +0100 Subject: gallium: Don't link x86sse.c into libmesa.a. Otherwise we get a symbol clash with the copy in src/gallium/util . --- src/mesa/SConscript | 1 - src/mesa/sources | 1 - 2 files changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index a10c34c8c5..80fbdb41f0 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -222,7 +222,6 @@ ASM_C_SOURCES = [ 'x86/x86.c', 'x86/3dnow.c', 'x86/sse.c', - 'x86/rtasm/x86sse.c', 'sparc/sparc.c', 'ppc/common_ppc.c', 'x86-64/x86-64.c', diff --git a/src/mesa/sources b/src/mesa/sources index a4e2026e4b..5c7ff50cc6 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -245,7 +245,6 @@ ASM_C_SOURCES = \ x86/x86.c \ x86/3dnow.c \ x86/sse.c \ - x86/rtasm/x86sse.c \ sparc/sparc.c \ ppc/common_ppc.c \ x86-64/x86-64.c -- cgit v1.2.3 From fd4acd6fdaaad4871327d081f5501680ed9da2ed Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 22 Apr 2008 14:22:11 +0100 Subject: gallium: How about restoring shaders instead of saving them again... --- src/mesa/state_tracker/st_cb_bitmap.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 71a848ea54..701d1e397e 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -490,8 +490,8 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, cso_restore_samplers(cso); cso_restore_sampler_textures(cso); cso_restore_viewport(cso); - cso_save_fragment_shader(cso); - cso_save_vertex_shader(cso); + cso_restore_fragment_shader(cso); + cso_restore_vertex_shader(cso); } -- cgit v1.2.3 From f9b1d47d652778012fd35552ffc51717ac0b6f79 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 15:15:21 +0100 Subject: softpipe: do something sensible on an error path, squash warning --- src/gallium/drivers/softpipe/sp_tex_sample.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 34da6356d7..5b63f97997 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -651,7 +651,9 @@ shadow_compare(uint compare_func, k = 0; break; default: + k = 0; assert(0); + break; } rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k; -- cgit v1.2.3 From 88f8eed3c4430505b1436b6c5b0114d34c33f822 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 15:26:38 +0100 Subject: cso: fix vs/fs confusion --- src/gallium/auxiliary/cso_cache/cso_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 2d98a34edf..23b1f5a032 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -600,7 +600,7 @@ void cso_restore_vertex_shader(struct cso_context *ctx) { assert(ctx->vertex_shader_saved); if (ctx->vertex_shader_saved != ctx->vertex_shader) { - ctx->pipe->bind_fs_state(ctx->pipe, ctx->vertex_shader_saved); + ctx->pipe->bind_vs_state(ctx->pipe, ctx->vertex_shader_saved); ctx->vertex_shader = ctx->vertex_shader_saved; } ctx->vertex_shader_saved = NULL; -- cgit v1.2.3 From 1f0f029ba6f22ef4ada01fcdc153da91571a7963 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 20:32:22 +0100 Subject: softpipe: squash warning --- src/gallium/drivers/softpipe/sp_state_fs.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 2921066ce3..525cea9eed 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -82,7 +82,6 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) void softpipe_delete_fs_state(struct pipe_context *pipe, void *fs) { - struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_fragment_shader *state = fs; assert(fs != softpipe->fs); -- cgit v1.2.3 From 0588858702d1a5c9c08573ea6817e2e149473cf6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 20:50:20 +0100 Subject: draw: allow drivers to query pipeline state more easily Also, provide a separate flag to say whether the driver can handle clipping/rhw tasks, in addition to the API flag which indicates they have already been done. --- src/gallium/auxiliary/draw/draw_context.c | 13 ++++++++++ src/gallium/auxiliary/draw/draw_context.h | 12 +++++++++ src/gallium/auxiliary/draw/draw_pipe_validate.c | 30 ++++++++++++---------- src/gallium/auxiliary/draw/draw_private.h | 7 +++-- src/gallium/auxiliary/draw/draw_pt.c | 6 +++-- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 2 +- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 11 +++++--- 7 files changed, 58 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index b4dbdccd61..b916b27877 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -122,6 +122,19 @@ void draw_set_rasterizer_state( struct draw_context *draw, draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); draw->rasterizer = raster; + draw->bypass_clipping = (draw->rasterizer->bypass_clipping || + draw->driver.bypass_clipping); +} + + +void draw_set_driver_clipping( struct draw_context *draw, + boolean bypass_clipping ) +{ + draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); + + draw->driver.bypass_clipping = bypass_clipping; + draw->bypass_clipping = (draw->rasterizer->bypass_clipping || + draw->driver.bypass_clipping); } diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index 2742001698..c5c3d3b09e 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -148,4 +148,16 @@ struct vbuf_render; void draw_set_render( struct draw_context *draw, struct vbuf_render *render ); +void draw_set_driver_clipping( struct draw_context *draw, + boolean bypass_clipping ); + +/******************************************************************************* + * Draw pipeline + */ +boolean draw_need_pipeline(const struct draw_context *draw, + const struct pipe_rasterizer_state *rasterizer, + unsigned prim ); + + + #endif /* DRAW_CONTEXT_H */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c index a2e0812c89..6be1d369c3 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_validate.c +++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c @@ -32,6 +32,7 @@ #include "pipe/p_defines.h" #include "draw_private.h" #include "draw_pipe.h" +#include "draw_context.h" static boolean points( unsigned prim ) { @@ -57,7 +58,8 @@ static boolean triangles( unsigned prim ) * pipeline stages. */ boolean -draw_need_pipeline(const struct draw_context *draw, +draw_need_pipeline(const struct draw_context *draw, + const struct pipe_rasterizer_state *rasterizer, unsigned int prim ) { /* Don't have to worry about triangles turning into lines/points @@ -67,30 +69,30 @@ draw_need_pipeline(const struct draw_context *draw, if (lines(prim)) { /* line stipple */ - if (draw->rasterizer->line_stipple_enable && draw->pipeline.line_stipple) + if (rasterizer->line_stipple_enable && draw->pipeline.line_stipple) return TRUE; /* wide lines */ - if (draw->rasterizer->line_width > draw->pipeline.wide_line_threshold) + if (rasterizer->line_width > draw->pipeline.wide_line_threshold) return TRUE; /* AA lines */ - if (draw->rasterizer->line_smooth && draw->pipeline.aaline) + if (rasterizer->line_smooth && draw->pipeline.aaline) return TRUE; } if (points(prim)) { /* large points */ - if (draw->rasterizer->point_size > draw->pipeline.wide_point_threshold) + if (rasterizer->point_size > draw->pipeline.wide_point_threshold) return TRUE; /* AA points */ - if (draw->rasterizer->point_smooth && draw->pipeline.aapoint) + if (rasterizer->point_smooth && draw->pipeline.aapoint) return TRUE; /* point sprites */ - if (draw->rasterizer->point_sprite && draw->pipeline.point_sprite) + if (rasterizer->point_sprite && draw->pipeline.point_sprite) return TRUE; } @@ -98,20 +100,20 @@ draw_need_pipeline(const struct draw_context *draw, if (triangles(prim)) { /* polygon stipple */ - if (draw->rasterizer->poly_stipple_enable && draw->pipeline.pstipple) + if (rasterizer->poly_stipple_enable && draw->pipeline.pstipple) return TRUE; /* unfilled polygons */ - if (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) + if (rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) return TRUE; /* polygon offset */ - if (draw->rasterizer->offset_cw || draw->rasterizer->offset_ccw) + if (rasterizer->offset_cw || rasterizer->offset_ccw) return TRUE; /* two-side lighting */ - if (draw->rasterizer->light_twoside) + if (rasterizer->light_twoside) return TRUE; } @@ -120,7 +122,7 @@ draw_need_pipeline(const struct draw_context *draw, * * Generally this isn't a reason to require the pipeline, though. * - if (draw->rasterizer->cull_mode) + if (rasterizer->cull_mode) return TRUE; */ @@ -239,7 +241,7 @@ static struct draw_stage *validate_pipeline( struct draw_stage *stage ) /* Clip stage */ - if (!draw->rasterizer->bypass_clipping) + if (!draw->bypass_clipping) { draw->pipeline.clip->next = next; next = draw->pipeline.clip; diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index da973e868b..83b81f66a0 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -174,7 +174,12 @@ struct draw_context } pt; + struct { + boolean bypass_clipping; + } driver; + boolean flushing; + boolean bypass_clipping; /* set if either api or driver bypass_clipping true */ /* pipe state that we need: */ const struct pipe_rasterizer_state *rasterizer; @@ -243,8 +248,6 @@ void draw_pipeline_run( struct draw_context *draw, void draw_pipeline_flush( struct draw_context *draw, unsigned flags ); -boolean draw_need_pipeline(const struct draw_context *draw, - unsigned prim ); /******************************************************************************* diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 9f8e8d3d62..f5a3bf390e 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -58,11 +58,13 @@ draw_pt_arrays(struct draw_context *draw, opt |= PT_PIPELINE; } - if (draw_need_pipeline(draw, prim)) { + if (draw_need_pipeline(draw, + draw->rasterizer, + prim)) { opt |= PT_PIPELINE; } - if (!draw->rasterizer->bypass_clipping) { + if (!draw->bypass_clipping) { opt |= PT_CLIPTEST; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 98a2cb45e4..f0763dad8d 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -81,7 +81,7 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, * but gl vs dx9 clip spaces. */ draw_pt_post_vs_prepare( fpme->post_vs, - draw->rasterizer->bypass_clipping, + draw->bypass_clipping, draw->identity_viewport, draw->rasterizer->gl_rasterization_rules ); diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index 581f044dae..e5e38fa264 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -94,8 +94,10 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs, struct vertex_header *out = vertices; const float *scale = pvs->draw->viewport.scale; const float *trans = pvs->draw->viewport.translate; + boolean clipped = FALSE; unsigned j; - unsigned clipped = 0; + + if (0) debug_printf("%s\n"); for (j = 0; j < count; j++) { out->clip[0] = out->data[0][0]; @@ -108,10 +110,11 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs, out->clipmask = compute_clipmask_gl(out->clip, pvs->draw->plane, pvs->draw->nr_planes); - clipped += out->clipmask; if (out->clipmask == 0) { + clipped = TRUE; + /* divide by w */ float w = 1.0f / out->data[0][3]; @@ -142,7 +145,7 @@ static boolean post_vs_viewport( struct pt_post_vs *pvs, const float *trans = pvs->draw->viewport.translate; unsigned j; - debug_printf("%s\n", __FUNCTION__); + if (0) debug_printf("%s\n", __FUNCTION__); for (j = 0; j < count; j++) { /* Viewport mapping only, no cliptest/rhw divide */ @@ -165,7 +168,7 @@ static boolean post_vs_none( struct pt_post_vs *pvs, unsigned count, unsigned stride ) { - debug_printf("%s\n", __FUNCTION__); + if (0) debug_printf("%s\n", __FUNCTION__); return FALSE; } -- cgit v1.2.3 From 4ebede8c7f43a83adfc73dcca783de2e9efcd9ba Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 22 Apr 2008 21:17:21 +0100 Subject: Revert "softpipe: squash warning" This reverts commit 1f0f029ba6f22ef4ada01fcdc153da91571a7963. Which broke the debug build. --- src/gallium/drivers/softpipe/sp_state_fs.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 525cea9eed..2921066ce3 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -82,6 +82,7 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) void softpipe_delete_fs_state(struct pipe_context *pipe, void *fs) { + struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_fragment_shader *state = fs; assert(fs != softpipe->fs); -- cgit v1.2.3 From 6b4798b9068b58a3c4581a268727ce01680d08cd Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 22 Apr 2008 22:23:22 +0200 Subject: i915: Fixed weird fence leak when I915_VBUF=true --- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index 923b542771..f029dd6a90 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -62,7 +62,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, } if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); + driFenceFinish(intel->last_swap_fence, driFenceType(intel->last_swap_fence), GL_FALSE); driFenceUnReference(&intel->last_swap_fence); intel->last_swap_fence = NULL; } -- cgit v1.2.3 From b456f1374fc958e53efc80ee38ef343f580c1d6b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 14:41:32 -0600 Subject: gallium: st_translate_fragment_program() is void now The return value was never used. --- src/mesa/state_tracker/st_program.c | 5 +---- src/mesa/state_tracker/st_program.h | 2 +- 2 files changed, 2 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 4d1ef1bf3d..3648ded8a1 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -315,7 +315,7 @@ st_translate_vertex_program(struct st_context *st, * \param tokensOut destination for TGSI tokens * \return pointer to cached pipe_shader object. */ -const struct cso_fragment_shader * +void st_translate_fragment_program(struct st_context *st, struct st_fragment_program *stfp, const GLuint inputMapping[]) @@ -325,7 +325,6 @@ st_translate_fragment_program(struct st_context *st, GLuint outputMapping[FRAG_RESULT_MAX]; GLuint defaultInputMapping[FRAG_ATTRIB_MAX]; struct pipe_shader_state fs; - const struct cso_fragment_shader *cso; GLuint interpMode[16]; /* XXX size? */ GLuint attr; const GLbitfield inputsRead = stfp->Base.Base.InputsRead; @@ -475,7 +474,5 @@ st_translate_fragment_program(struct st_context *st, if (TGSI_DEBUG) tgsi_dump( fs.tokens, 0/*TGSI_DUMP_VERBOSE*/ ); - - return cso; } diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index dced31e88e..e58b8a1851 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -111,7 +111,7 @@ st_vertex_program( struct gl_vertex_program *vp ) } -extern const struct cso_fragment_shader * +extern void st_translate_fragment_program(struct st_context *st, struct st_fragment_program *fp, const GLuint inputMapping[]); -- cgit v1.2.3 From 8f4f89c04383b2100f6d856270cad62dfe8a6354 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 14:42:36 -0600 Subject: gallium: remove assertions It's possible the current vs/fs is null when cso_save_vertex/fragment_shader() is called. --- src/gallium/auxiliary/cso_cache/cso_context.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 23b1f5a032..0523cb1949 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -528,7 +528,6 @@ void cso_save_fragment_shader(struct cso_context *ctx) void cso_restore_fragment_shader(struct cso_context *ctx) { - assert(ctx->fragment_shader_saved); if (ctx->fragment_shader_saved != ctx->fragment_shader) { ctx->pipe->bind_fs_state(ctx->pipe, ctx->fragment_shader_saved); ctx->fragment_shader = ctx->fragment_shader_saved; @@ -598,7 +597,6 @@ void cso_save_vertex_shader(struct cso_context *ctx) void cso_restore_vertex_shader(struct cso_context *ctx) { - assert(ctx->vertex_shader_saved); if (ctx->vertex_shader_saved != ctx->vertex_shader) { ctx->pipe->bind_vs_state(ctx->pipe, ctx->vertex_shader_saved); ctx->vertex_shader = ctx->vertex_shader_saved; -- cgit v1.2.3 From 465bc9473a0122d8f66ac1b4a69459e9bd889799 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 16:15:49 -0600 Subject: gallium: move the vertex print/debug code --- src/gallium/drivers/softpipe/sp_setup.c | 14 ++++++-------- 1 file changed, 6 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 0bf2b2dc3e..7df8fc5f67 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -299,13 +299,6 @@ static boolean setup_sort_vertices( struct setup_context *setup, const float (*v1)[4], const float (*v2)[4] ) { -#if DEBUG_VERTS - debug_printf("Triangle:\n"); - print_vertex(setup, v0); - print_vertex(setup, v1); - print_vertex(setup, v2); -#endif - setup->vprovoke = v2; /* determine bottom to top order of vertices */ @@ -738,7 +731,12 @@ void setup_tri( struct setup_context *setup, setup->numFragsWritten = 0; #endif - +#if DEBUG_VERTS + debug_printf("Triangle:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); + print_vertex(setup, v2); +#endif if (cull_tri( setup, det )) return; -- cgit v1.2.3 From aaa43218f34fc9897b280d6cf9bc1a31bbb7dafc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 16:22:21 -0600 Subject: gallium: setup an identity viewport --- src/gallium/auxiliary/util/u_gen_mipmap.c | 35 +++++-------------------------- 1 file changed, 5 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 9822a25ca6..ff5881b979 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -61,6 +61,7 @@ struct gen_mipmap_state struct pipe_depth_stencil_alpha_state depthstencil; struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state sampler; + struct pipe_viewport_state viewport; struct pipe_shader_state vert_shader; struct pipe_shader_state frag_shader; @@ -723,9 +724,7 @@ util_create_gen_mipmap(struct pipe_context *pipe, ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; ctx->sampler.normalized_coords = 1; - -#if 0 - /* viewport */ + /* viewport state (identity, verts are in wincoords) */ ctx->viewport.scale[0] = 1.0; ctx->viewport.scale[1] = 1.0; ctx->viewport.scale[2] = 1.0; @@ -734,7 +733,6 @@ util_create_gen_mipmap(struct pipe_context *pipe, ctx->viewport.translate[1] = 0.0; ctx->viewport.translate[2] = 0.0; ctx->viewport.translate[3] = 0.0; -#endif /* vertex shader */ { @@ -825,26 +823,6 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx) } -#if 0 -static void -simple_viewport(struct pipe_context *pipe, uint width, uint height) -{ - struct pipe_viewport_state vp; - - vp.scale[0] = 0.5 * width; - vp.scale[1] = -0.5 * height; - vp.scale[2] = 1.0; - vp.scale[3] = 1.0; - vp.translate[0] = 0.5 * width; - vp.translate[1] = 0.5 * height; - vp.translate[2] = 0.0; - vp.translate[3] = 0.0; - - pipe->set_viewport_state(pipe, &vp); -} -#endif - - /** * Generate mipmap images. It's assumed all needed texture memory is * already allocated. @@ -882,17 +860,16 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_save_framebuffer(ctx->cso); cso_save_fragment_shader(ctx->cso); cso_save_vertex_shader(ctx->cso); + cso_save_viewport(ctx->cso); /* bind our state */ cso_set_blend(ctx->cso, &ctx->blend); cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); cso_set_rasterizer(ctx->cso, &ctx->rasterizer); + cso_set_viewport(ctx->cso, &ctx->viewport); cso_set_fragment_shader_handle(ctx->cso, ctx->fs); cso_set_vertex_shader_handle(ctx->cso, ctx->vs); -#if 0 - cso_set_viewport(ctx->cso, &ctx->viewport); -#endif /* init framebuffer state */ memset(&fb, 0, sizeof(fb)); @@ -928,9 +905,6 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, ctx->sampler.lod_bias = (float) srcLevel; cso_single_sampler(ctx->cso, 0, &ctx->sampler); cso_single_sampler_done(ctx->cso); -#if 0 - simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]); -#endif cso_set_sampler_textures(ctx->cso, 1, &pt); @@ -958,4 +932,5 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, cso_restore_framebuffer(ctx->cso); cso_restore_fragment_shader(ctx->cso); cso_restore_vertex_shader(ctx->cso); + cso_restore_viewport(ctx->cso); } -- cgit v1.2.3 From 0939a986a8a7dbb8d30c505fcaad8d2718a6527d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 16:25:55 -0600 Subject: gallium: update comment about bypass clipping/viewport --- src/gallium/auxiliary/util/u_gen_mipmap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index ff5881b979..dfdb5f16fe 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -713,7 +713,7 @@ util_create_gen_mipmap(struct pipe_context *pipe, memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer)); ctx->rasterizer.front_winding = PIPE_WINDING_CW; ctx->rasterizer.cull_mode = PIPE_WINDING_NONE; - ctx->rasterizer.bypass_clipping = 1; /* bypasses viewport too */ + ctx->rasterizer.bypass_clipping = 1; /*ctx->rasterizer.bypass_vs = 1;*/ /* sampler state */ -- cgit v1.2.3 From 22cbf6a70437dfa12c10600e5a496ea5771cfc56 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 16:29:48 -0600 Subject: gallium: setup an identity viewport This fixes broken blits. --- src/gallium/auxiliary/util/u_blit.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 0938b03820..1105066cb8 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -57,6 +57,7 @@ struct blit_state struct pipe_depth_stencil_alpha_state depthstencil; struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state sampler; + struct pipe_viewport_state viewport; struct pipe_shader_state vert_shader; struct pipe_shader_state frag_shader; @@ -100,7 +101,7 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer)); ctx->rasterizer.front_winding = PIPE_WINDING_CW; ctx->rasterizer.cull_mode = PIPE_WINDING_NONE; - ctx->rasterizer.bypass_clipping = 1; /* bypasses viewport too */ + ctx->rasterizer.bypass_clipping = 1; /*ctx->rasterizer.bypass_vs = 1;*/ /* samplers */ @@ -113,8 +114,7 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) ctx->sampler.mag_img_filter = 0; /* set later */ ctx->sampler.normalized_coords = 1; -#if 0 - /* viewport */ + /* viewport (identity, we setup vertices in wincoords) */ ctx->viewport.scale[0] = 1.0; ctx->viewport.scale[1] = 1.0; ctx->viewport.scale[2] = 1.0; @@ -123,7 +123,6 @@ util_create_blit(struct pipe_context *pipe, struct cso_context *cso) ctx->viewport.translate[1] = 0.0; ctx->viewport.translate[2] = 0.0; ctx->viewport.translate[3] = 0.0; -#endif /* vertex shader */ { @@ -302,11 +301,13 @@ util_blit_pixels(struct blit_state *ctx, cso_save_framebuffer(ctx->cso); cso_save_fragment_shader(ctx->cso); cso_save_vertex_shader(ctx->cso); + cso_save_viewport(ctx->cso); /* set misc state we care about */ cso_set_blend(ctx->cso, &ctx->blend); cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); cso_set_rasterizer(ctx->cso, &ctx->rasterizer); + cso_set_viewport(ctx->cso, &ctx->viewport); /* sampler */ ctx->sampler.min_img_filter = filter; @@ -348,6 +349,7 @@ util_blit_pixels(struct blit_state *ctx, cso_restore_framebuffer(ctx->cso); cso_restore_fragment_shader(ctx->cso); cso_restore_vertex_shader(ctx->cso); + cso_restore_viewport(ctx->cso); /* free the texture */ pipe_surface_reference(&texSurf, NULL); -- cgit v1.2.3 From 500c41b8ba3ad025c69e63a2c74da90674a8037d Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 22 Apr 2008 18:11:58 -0400 Subject: Fix reporting of clipped vertices. --- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index e5e38fa264..f98e130ed6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -94,7 +94,7 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs, struct vertex_header *out = vertices; const float *scale = pvs->draw->viewport.scale; const float *trans = pvs->draw->viewport.translate; - boolean clipped = FALSE; + unsigned clipped = 0; unsigned j; if (0) debug_printf("%s\n"); @@ -110,11 +110,10 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs, out->clipmask = compute_clipmask_gl(out->clip, pvs->draw->plane, pvs->draw->nr_planes); + clipped += out->clipmask; if (out->clipmask == 0) { - clipped = TRUE; - /* divide by w */ float w = 1.0f / out->data[0][3]; -- cgit v1.2.3 From f088b53769aacbee20135d912c33d688b6002011 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 22 Apr 2008 18:20:34 -0400 Subject: Cache translate's structs for emits and fetches. Results in a fair speed improvement. --- src/gallium/auxiliary/draw/draw_pt_emit.c | 63 ++++++++++++++++++++++++++--- src/gallium/auxiliary/draw/draw_pt_fetch.c | 65 ++++++++++++++++++++++++++---- 2 files changed, 114 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 490da4cca3..d35329aba0 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -33,13 +33,66 @@ #include "draw/draw_pt.h" #include "translate/translate.h" +#include "cso_cache/cso_cache.h" +#include "cso_cache/cso_hash.h" struct pt_emit { struct draw_context *draw; struct translate *translate; + + struct cso_hash *hash; }; +static INLINE unsigned translate_hash_key_size(struct translate_key *key) +{ + unsigned size = sizeof(struct translate_key) - + sizeof(struct translate_element) * (PIPE_MAX_ATTRIBS - key->nr_elements); + return size; +} + +static INLINE unsigned create_key(struct translate_key *key) +{ + unsigned hash_key; + unsigned size = translate_hash_key_size(key); + /*debug_printf("key size = %d, (els = %d)\n", + size, key->nr_elements);*/ + hash_key = cso_construct_key(key, size); + return hash_key; +} + +static struct translate *cached_translate(struct pt_emit *emit, + struct translate_key *key) +{ + unsigned hash_key = create_key(key); + struct cso_hash_iter iter = cso_hash_find(emit->hash, hash_key); + struct translate *translate = 0; + + if (cso_hash_iter_is_null(iter)) { + translate = translate_create(key); + cso_hash_insert(emit->hash, hash_key, translate); + /*debug_printf("\tCREATED with %d\n", hash_key);*/ + } else { + translate = cso_hash_iter_data(iter); + /*debug_printf("\tOK with %d\n", hash_key);*/ + } + + return translate; +} + + +static INLINE void delete_translates(struct pt_emit *emit) +{ + struct cso_hash *hash = emit->hash; + struct cso_hash_iter iter = cso_hash_first_node(hash); + while (!cso_hash_iter_is_null(iter)) { + struct translate *state = (struct translate*)cso_hash_iter_data(iter); + iter = cso_hash_iter_next(iter); + if (state) { + state->release(state); + } + } +} void draw_pt_emit_prepare( struct pt_emit *emit, unsigned prim ) @@ -123,10 +176,7 @@ void draw_pt_emit_prepare( struct pt_emit *emit, if (!emit->translate || memcmp(&emit->translate->key, &hw_key, sizeof(hw_key)) != 0) { - if (emit->translate) - emit->translate->release(emit->translate); - - emit->translate = translate_create( &hw_key ); + emit->translate = cached_translate(emit, &hw_key); } } @@ -188,14 +238,15 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw ) return NULL; emit->draw = draw; + emit->hash = cso_hash_create(); return emit; } void draw_pt_emit_destroy( struct pt_emit *emit ) { - if (emit->translate) - emit->translate->release( emit->translate ); + delete_translates(emit); + cso_hash_delete(emit->hash); FREE(emit); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index f98bce6eac..93da811ed8 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -33,16 +33,67 @@ #include "draw/draw_pt.h" #include "translate/translate.h" +#include "cso_cache/cso_cache.h" +#include "cso_cache/cso_hash.h" struct pt_fetch { struct draw_context *draw; struct translate *translate; - + unsigned vertex_size; + + struct cso_hash *hash; }; +static INLINE unsigned translate_hash_key_size(struct translate_key *key) +{ + unsigned size = sizeof(struct translate_key) - + sizeof(struct translate_element) * (PIPE_MAX_ATTRIBS - key->nr_elements); + return size; +} + +static INLINE unsigned create_key(struct translate_key *key) +{ + unsigned hash_key; + unsigned size = translate_hash_key_size(key); + /*debug_printf("key size = %d, (els = %d)\n", + size, key->nr_elements);*/ + hash_key = cso_construct_key(key, size); + return hash_key; +} + +static struct translate *cached_translate(struct pt_fetch *fetch, + struct translate_key *key) +{ + unsigned hash_key = create_key(key); + struct cso_hash_iter iter = cso_hash_find(fetch->hash, hash_key); + struct translate *translate = 0; + + if (cso_hash_iter_is_null(iter)) { + translate = translate_create(key); + cso_hash_insert(fetch->hash, hash_key, translate); + /*debug_printf("\tCREATED with %d\n", hash_key);*/ + } else { + translate = cso_hash_iter_data(iter); + /*debug_printf("\tOK with %d\n", hash_key);*/ + } + + return translate; +} +static INLINE void delete_translates(struct pt_fetch *fetch) +{ + struct cso_hash *hash = fetch->hash; + struct cso_hash_iter iter = cso_hash_first_node(hash); + while (!cso_hash_iter_is_null(iter)) { + struct translate *state = (struct translate*)cso_hash_iter_data(iter); + iter = cso_hash_iter_next(iter); + if (state) { + state->release(state); + } + } +} /* Perform the fetch from API vertex elements & vertex buffers, to a * contiguous set of float[4] attributes as required for the @@ -112,10 +163,7 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, if (!fetch->translate || memcmp(&fetch->translate->key, &key, sizeof(key)) != 0) { - if (fetch->translate) - fetch->translate->release(fetch->translate); - - fetch->translate = translate_create( &key ); + fetch->translate = cached_translate(fetch, &key); { static struct vertex_header vh = { 0, 0, 0, 0xffff }; @@ -159,15 +207,16 @@ struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) struct pt_fetch *fetch = CALLOC_STRUCT(pt_fetch); if (!fetch) return NULL; - + fetch->draw = draw; + fetch->hash = cso_hash_create(); return fetch; } void draw_pt_fetch_destroy( struct pt_fetch *fetch ) { - if (fetch->translate) - fetch->translate->release( fetch->translate ); + delete_translates(fetch); + cso_hash_delete(fetch->hash); FREE(fetch); } -- cgit v1.2.3 From d8f2e400cfe6e32e82d1656d3483905343124b97 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 22 Apr 2008 18:30:31 -0400 Subject: Fix a crash. Rasterizer can be null --- src/gallium/auxiliary/draw/draw_context.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index b916b27877..db92a53ed2 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -122,8 +122,9 @@ void draw_set_rasterizer_state( struct draw_context *draw, draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); draw->rasterizer = raster; - draw->bypass_clipping = (draw->rasterizer->bypass_clipping || - draw->driver.bypass_clipping); + draw->bypass_clipping = + ((draw->rasterizer && draw->rasterizer->bypass_clipping) || + draw->driver.bypass_clipping); } -- cgit v1.2.3 From 36feb5eacf16467d06d5cd9f63d19f17f933f1ef Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 22 Apr 2008 18:32:29 -0400 Subject: In case the 'func' is deleting the state move the iterator before calling it. --- src/gallium/auxiliary/cso_cache/cso_cache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index f607528fdc..63464e0705 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -334,10 +334,10 @@ void cso_for_each_state(struct cso_cache *sc, enum cso_cache_type type, iter = cso_hash_first_node(hash); while (!cso_hash_iter_is_null(iter)) { void *state = cso_hash_iter_data(iter); + iter = cso_hash_iter_next(iter); if (state) { func(state, user_data); } - iter = cso_hash_iter_next(iter); } } -- cgit v1.2.3 From 1977fbff6010af0a5bd0bba7c0367b7713185a92 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 18:38:13 -0600 Subject: gallium: don't set buffer status to undefined in display_front_buffer() --- src/mesa/state_tracker/st_cb_flush.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 1b3402cee2..5e866b0d41 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -67,7 +67,9 @@ display_front_buffer(struct st_context *st) st->pipe->winsys->flush_frontbuffer( st->pipe->winsys, front_surf, st->pipe->priv ); + /* st->frontbuffer_status = FRONT_STATUS_UNDEFINED; + */ } -- cgit v1.2.3 From 33f3938d2d6340b31d758c96bd35f858c6c8267d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 18:38:37 -0600 Subject: gallium: fix some bitmap frag shader issues If texturing happens to be enabled when glBitmap() is called, need to be careful about choosing a sampler unit, etc. --- src/mesa/state_tracker/st_cb_bitmap.c | 85 +++++++++++++++++++++------------- src/mesa/state_tracker/st_cb_program.c | 4 ++ src/mesa/state_tracker/st_context.h | 3 -- src/mesa/state_tracker/st_program.h | 4 ++ 4 files changed, 62 insertions(+), 34 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 701d1e397e..7752b40e8b 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -105,7 +105,7 @@ struct bitmap_cache * This program will be combined with the user's fragment program. */ static struct st_fragment_program * -make_bitmap_fragment_program(GLcontext *ctx) +make_bitmap_fragment_program(GLcontext *ctx, GLuint samplerIndex) { struct st_fragment_program *stfp; struct gl_program *p; @@ -130,7 +130,7 @@ make_bitmap_fragment_program(GLcontext *ctx) p->Instructions[ic].DstReg.Index = 0; p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT; p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0; - p->Instructions[ic].TexSrcUnit = 0; + p->Instructions[ic].TexSrcUnit = samplerIndex; p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX; ic++; @@ -148,7 +148,7 @@ make_bitmap_fragment_program(GLcontext *ctx) p->InputsRead = FRAG_BIT_TEX0; p->OutputsWritten = 0x0; - p->SamplersUsed = 0x1; /* sampler 0 (bit 0) is used */ + p->SamplersUsed = (1 << samplerIndex); stfp = (struct st_fragment_program *) p; stfp->Base.UsesKill = GL_TRUE; @@ -158,6 +158,19 @@ make_bitmap_fragment_program(GLcontext *ctx) } +static int +find_free_bit(uint bitfield) +{ + int i; + for (i = 0; i < 32; i++) { + if ((bitfield & (1 << i)) == 0) { + return i; + } + } + return -1; +} + + /** * Combine basic bitmap fragment program with the user-defined program. */ @@ -165,28 +178,26 @@ static struct st_fragment_program * combined_bitmap_fragment_program(GLcontext *ctx) { struct st_context *st = ctx->st; - struct st_fragment_program *stfp; - - if (!st->bitmap.program) { - /* create the basic bitmap fragment program */ - st->bitmap.program = make_bitmap_fragment_program(ctx); - } + struct st_fragment_program *stfp = st->fp; - if (st->bitmap.user_prog_sn == st->fp->serialNo) { - /* re-use */ - stfp = st->bitmap.combined_prog; - } - else { - /* Concatenate the bitmap program with the current user-defined program. + if (!stfp->bitmap_program) { + /* + * Generate new program which is the user-defined program prefixed + * with the bitmap sampler/kill instructions. */ - stfp = (struct st_fragment_program *) - _mesa_combine_programs(ctx, - &st->bitmap.program->Base.Base, - &st->fp->Base.Base); + struct st_fragment_program *bitmap_prog; + uint sampler; + sampler = find_free_bit(st->fp->Base.Base.SamplersUsed); + bitmap_prog = make_bitmap_fragment_program(ctx, sampler); + + stfp->bitmap_program = (struct st_fragment_program *) + _mesa_combine_programs(ctx, + &bitmap_prog->Base.Base, &stfp->Base.Base); + stfp->bitmap_program->bitmap_sampler = sampler; #if 0 { - struct gl_program *p = &stfp->Base.Base; + struct gl_program *p = &stfp->bitmap_program->Base.Base; printf("Combined bitmap program:\n"); _mesa_print_program(p); printf("InputsRead: 0x%x\n", p->InputsRead); @@ -196,11 +207,7 @@ combined_bitmap_fragment_program(GLcontext *ctx) #endif /* translate to TGSI tokens */ - st_translate_fragment_program(st, stfp, NULL); - - /* save new program, update serial numbers */ - st->bitmap.user_prog_sn = st->fp->serialNo; - st->bitmap.combined_prog = stfp; + st_translate_fragment_program(st, stfp->bitmap_program, NULL); } /* Ideally we'd have updated the pipe constants during the normal @@ -208,7 +215,7 @@ combined_bitmap_fragment_program(GLcontext *ctx) */ st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT); - return stfp; + return stfp->bitmap_program; } @@ -451,10 +458,25 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* vertex shader state: position + texcoord pass-through */ cso_set_vertex_shader_handle(cso, st->bitmap.vs); - /* sampler / texture state */ - cso_single_sampler(cso, 0, &st->bitmap.sampler); - cso_single_sampler_done(cso); - pipe->set_sampler_textures(pipe, 1, &pt); + /* user samplers, plus our bitmap sampler */ + { + struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS]; + uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_samplers); + uint i; + for (i = 0; i < st->state.num_samplers; i++) { + samplers[i] = &st->state.samplers[i]; + } + samplers[stfp->bitmap_sampler] = &st->bitmap.sampler; + cso_set_samplers(cso, num, (const struct pipe_sampler_state **) samplers); } + + /* user textures, plus the bitmap texture */ + { + struct pipe_texture *textures[PIPE_MAX_SAMPLERS]; + uint num = MAX2(stfp->bitmap_sampler + 1, st->state.num_textures); + memcpy(textures, st->state.sampler_texture, sizeof(textures)); + textures[stfp->bitmap_sampler] = pt; + cso_set_sampler_textures(cso, num, textures); + } /* viewport state: viewport matching window dims */ { @@ -771,6 +793,7 @@ st_destroy_bitmap(struct st_context *st) { struct pipe_context *pipe = st->pipe; +#if 0 if (st->bitmap.combined_prog) { st_delete_program(st->ctx, &st->bitmap.combined_prog->Base.Base); } @@ -778,7 +801,7 @@ st_destroy_bitmap(struct st_context *st) if (st->bitmap.program) { st_delete_program(st->ctx, &st->bitmap.program->Base.Base); } - +#endif if (st->bitmap.vs) { pipe->delete_vs_state(pipe, st->bitmap.vs); st->bitmap.vs = NULL; diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 003ec0d2b7..67f8b1f8eb 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -159,6 +159,10 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) stfp->state.tokens = NULL; } + if (stfp->bitmap_program) { + st_delete_program(ctx, &stfp->bitmap_program->Base.Base); + } + st_free_translated_vertex_programs(st, stfp->vertex_programs); } break; diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index ae1ba41d4f..2851770d4c 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -144,9 +144,6 @@ struct st_context /** for glBitmap */ struct { - struct st_fragment_program *program; /**< bitmap tex/kil program */ - GLuint user_prog_sn; /**< user fragment program serial no. */ - struct st_fragment_program *combined_prog; struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state sampler; struct pipe_shader_state vert_shader; diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index e58b8a1851..1a2062131d 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -69,6 +69,10 @@ struct st_fragment_program * outputs match this fragment program's inputs. */ struct translated_vertex_program *vertex_programs; + + /** Program prefixed with glBitmap prologue */ + struct st_fragment_program *bitmap_program; + uint bitmap_sampler; }; -- cgit v1.2.3 From 91e37b71404a83b5e4258e129a2753f7c8fd0706 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 19:10:52 -0600 Subject: gallium: fix bad logic in bind_pstip_fragment_shader(): use &&, not || --- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 61bdd29aa0..aec485a6e7 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -469,7 +469,7 @@ pstip_create_sampler(struct pstip_stage *pstip) static boolean bind_pstip_fragment_shader(struct pstip_stage *pstip) { - if (!pstip->fs->pstip_fs || + if (!pstip->fs->pstip_fs && !generate_pstip_fs(pstip)) return FALSE; -- cgit v1.2.3 From 72fd5b9c5a78792ad8c1fe7c8713a3583008c50a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Apr 2008 19:11:59 -0600 Subject: gallium: added a flushing_vcache flag, test in draw_do_flush() Fixes broken polygon stipple, aaline, aapoint stages --- src/gallium/auxiliary/draw/draw_context.c | 2 +- src/gallium/auxiliary/draw/draw_private.h | 1 + src/gallium/auxiliary/draw/draw_pt_vcache.c | 2 ++ 3 files changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index db92a53ed2..f90187816b 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -367,7 +367,7 @@ draw_set_mapped_element_buffer( struct draw_context *draw, */ void draw_do_flush( struct draw_context *draw, unsigned flags ) { - if (!draw->flushing) + if (!draw->flushing && !draw->vcache_flushing) { draw->flushing = TRUE; diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 83b81f66a0..39aa81b43c 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -179,6 +179,7 @@ struct draw_context } driver; boolean flushing; + boolean vcache_flushing; boolean bypass_clipping; /* set if either api or driver bypass_clipping true */ /* pipe state that we need: */ diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 153055417d..afcff41043 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -63,6 +63,7 @@ struct vcache_frontend { static void vcache_flush( struct vcache_frontend *vcache ) { + vcache->draw->vcache_flushing = TRUE; if (vcache->draw_count) { vcache->middle->run( vcache->middle, vcache->fetch_elts, @@ -74,6 +75,7 @@ static void vcache_flush( struct vcache_frontend *vcache ) memset(vcache->in, ~0, sizeof(vcache->in)); vcache->fetch_count = 0; vcache->draw_count = 0; + vcache->draw->vcache_flushing = FALSE; } static void vcache_check_flush( struct vcache_frontend *vcache ) -- cgit v1.2.3 From 6fc530ccda2971a5d99a955ad90ae9762238040f Mon Sep 17 00:00:00 2001 From: Ben Skeggs Date: Wed, 23 Apr 2008 12:38:37 +1000 Subject: fix non-i386 builds --- src/gallium/auxiliary/translate/translate.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate.c b/src/gallium/auxiliary/translate/translate.c index 86ebeae2a7..b04bc6eefd 100644 --- a/src/gallium/auxiliary/translate/translate.c +++ b/src/gallium/auxiliary/translate/translate.c @@ -38,9 +38,11 @@ struct translate *translate_create( const struct translate_key *key ) { struct translate *translate = NULL; +#if defined(__i386__) || defined(__386__) || defined(i386) translate = translate_sse2_create( key ); if (translate) return translate; +#endif return translate_generic_create( key ); } -- cgit v1.2.3 From b0a58d526624485b98665b103fb5b507253b851b Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 23 Apr 2008 11:56:35 +0200 Subject: i915: Real fix for weird fence leak. Thanks Thomas --- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 2 +- src/gallium/winsys/dri/intel/ws_dri_slabpool.c | 109 ++++++++++++++--------- 2 files changed, 67 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index f029dd6a90..923b542771 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -62,7 +62,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, } if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, driFenceType(intel->last_swap_fence), GL_FALSE); + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); driFenceUnReference(&intel->last_swap_fence); intel->last_swap_fence = NULL; } diff --git a/src/gallium/winsys/dri/intel/ws_dri_slabpool.c b/src/gallium/winsys/dri/intel/ws_dri_slabpool.c index 235f0ac972..62d82bbd94 100644 --- a/src/gallium/winsys/dri/intel/ws_dri_slabpool.c +++ b/src/gallium/winsys/dri/intel/ws_dri_slabpool.c @@ -227,6 +227,7 @@ driAllocKernelBO(struct _DriSlabSizeHeader *header) drmMMListHead *list, *next, *head; uint32_t size = header->bufSize * slabPool->desiredNumBuffers; struct _DriKernelBO *kbo; + struct _DriKernelBO *kboTmp; int ret; /* @@ -248,27 +249,36 @@ driAllocKernelBO(struct _DriSlabSizeHeader *header) list != head; list = next, next = list->next) { - kbo = DRMLISTENTRY(struct _DriKernelBO, list, head); + kboTmp = DRMLISTENTRY(struct _DriKernelBO, list, head); - if ((kbo->bo.size == size) && + if ((kboTmp->bo.size == size) && (slabPool->pageAlignment == 0 || - (kbo->pageAlignment % slabPool->pageAlignment) == 0)) { + (kboTmp->pageAlignment % slabPool->pageAlignment) == 0)) { + + if (!kbo) + kbo = kboTmp; + + if ((kbo->bo.proposedFlags ^ slabPool->proposedFlags) == 0) + break; - DRMLISTDELINIT(&kbo->head); - DRMLISTDELINIT(&kbo->timeoutHead); - break; } + } - kbo = NULL; + if (kbo) { + DRMLISTDELINIT(&kbo->head); + DRMLISTDELINIT(&kbo->timeoutHead); } _glthread_UNLOCK_MUTEX(fMan->mutex); if (kbo) { + uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags; + ret = 0; - ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, - (slabPool->proposedFlags ^ kbo->bo.flags), - DRM_BO_HINT_DONT_FENCE, 0, 0); + if (new_mask) { + ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, + new_mask, DRM_BO_HINT_DONT_FENCE, 0, 0); + } if (ret == 0) return kbo; @@ -284,7 +294,6 @@ driAllocKernelBO(struct _DriSlabSizeHeader *header) kbo->fd = slabPool->fd; DRMINITLISTHEAD(&kbo->head); DRMINITLISTHEAD(&kbo->timeoutHead); - ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL, slabPool->proposedFlags, DRM_BO_HINT_DONT_FENCE, &kbo->bo); @@ -417,47 +426,61 @@ driSlabFreeBufferLocked(struct _DriSlabBuffer *buf) static void driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait) { - drmMMListHead *list, *prev; + drmMMListHead *list, *prev, *first; struct _DriSlabBuffer *buf; struct _DriSlab *slab; - - int signaled = 0; + int firstWasSignaled = 1; + int signaled; int i; int ret; - list = header->delayedBuffers.next; - - /* Only examine the oldest 1/3 of delayed buffers: + /* + * Rerun the freeing test if the youngest tested buffer + * was signaled, since there might be more idle buffers + * in the delay list. */ - if (header->numDelayed > 3) { - for (i = 0; i < header->numDelayed; i += 3) { - list = list->next; - } - } - prev = list->prev; - for (; list != &header->delayedBuffers; list = prev, prev = list->prev) { - buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); - slab = buf->parent; - - if (!signaled) { - if (wait) { - ret = driFenceFinish(buf->fence, buf->fenceType, 0); - if (ret) - break; - signaled = 1; - } else { - signaled = driFenceSignaled(buf->fence, buf->fenceType); + while (firstWasSignaled) { + firstWasSignaled = 0; + signaled = 0; + first = header->delayedBuffers.next; + + /* Only examine the oldest 1/3 of delayed buffers: + */ + if (header->numDelayed > 3) { + for (i = 0; i < header->numDelayed; i += 3) { + first = first->next; } - if (signaled) { + } + + for (list = first, prev = list->prev; + list != &header->delayedBuffers; + list = prev, prev = list->prev) { + buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); + slab = buf->parent; + + if (!signaled) { + if (wait) { + ret = driFenceFinish(buf->fence, buf->fenceType, 0); + if (ret) + break; + signaled = 1; + wait = 0; + } else { + signaled = driFenceSignaled(buf->fence, buf->fenceType); + } + if (signaled) { + if (list == first) + firstWasSignaled = 1; + driFenceUnReference(&buf->fence); + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { driFenceUnReference(&buf->fence); header->numDelayed--; driSlabFreeBufferLocked(buf); } - } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { - driFenceUnReference(&buf->fence); - header->numDelayed--; - driSlabFreeBufferLocked(buf); } } } @@ -480,15 +503,14 @@ driSlabAllocBuffer(struct _DriSlabSizeHeader *header) _glthread_UNLOCK_MUTEX(header->mutex); if (count != DRI_SLABPOOL_ALLOC_RETRIES) usleep(1); - (void) driAllocSlab(header); _glthread_LOCK_MUTEX(header->mutex); + (void) driAllocSlab(header); count--; } list = header->slabs.next; if (list == &header->slabs) { _glthread_UNLOCK_MUTEX(header->mutex); - //assert(0, "no buffers in slab"); return NULL; } slab = DRMLISTENTRY(struct _DriSlab, list, head); @@ -514,7 +536,6 @@ pool_create(struct _DriBufferPool *driPool, unsigned long size, int i; int ret; - /* * FIXME: Check for compatibility. */ @@ -605,6 +626,8 @@ pool_destroy(struct _DriBufferPool *driPool, void *private) DRMLISTADDTAIL(&buf->head, &header->delayedBuffers); header->numDelayed++; } else { + if (buf->fence) + driFenceUnReference(&buf->fence); driSlabFreeBufferLocked(buf); } -- cgit v1.2.3 From d6a965972b48fec95a2bcb778cf05d0468ba2573 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 23 Apr 2008 13:51:50 +0100 Subject: draw: add missing break (Jakob) --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index afd5f5544d..1eb7c4bc38 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -136,7 +136,7 @@ emit_vertex( struct vbuf_stage *vbuf, vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0); vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr); - if (0) draw_dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); + if (1) draw_dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); vbuf->vertex_ptr += vbuf->vertex_size/4; vertex->vertex_id = vbuf->nr_vertices++; @@ -254,6 +254,7 @@ vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) case EMIT_4UB: output_format = PIPE_FORMAT_B8G8R8A8_UNORM; emit_sz = 4 * sizeof(ubyte); + break; default: assert(0); output_format = PIPE_FORMAT_NONE; -- cgit v1.2.3 From 43be7a4819ad342e1cb3f8e3fb966a8a78dc2c1b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 23 Apr 2008 13:56:05 +0100 Subject: draw: remove stupid debug (Keith) --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 1eb7c4bc38..2a19e6916a 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -136,7 +136,7 @@ emit_vertex( struct vbuf_stage *vbuf, vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0); vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr); - if (1) draw_dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); + if (0) draw_dump_emitted_vertex(vbuf->vinfo, (uint8_t *)vbuf->vertex_ptr); vbuf->vertex_ptr += vbuf->vertex_size/4; vertex->vertex_id = vbuf->nr_vertices++; -- cgit v1.2.3 From 333976c90aafa602816defef3e4cc4a418601a51 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 10:29:52 -0600 Subject: gallium: fix broken hashing for vertex translation It seems we get hash collisions fairly easily and the code as it was didn't deal with that properly. I think we need a simpler hashing interface... --- src/gallium/auxiliary/draw/draw_pt_fetch.c | 18 +++++++++++++----- 1 file changed, 13 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 93da811ed8..a8d4be5b64 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -70,13 +70,21 @@ static struct translate *cached_translate(struct pt_fetch *fetch, struct cso_hash_iter iter = cso_hash_find(fetch->hash, hash_key); struct translate *translate = 0; - if (cso_hash_iter_is_null(iter)) { + while (!cso_hash_iter_is_null(iter)) { + void *iter_data = cso_hash_iter_data(iter); + if (memcmp(iter_data, key, sizeof(*key)) == 0) { + /* found */ + translate = cso_hash_iter_data(iter); + break; + } + iter = cso_hash_iter_next(iter); + /*debug_printf("\tOK with %d\n", hash_key);*/ + } + + if (!translate) { + /* create/insert */ translate = translate_create(key); cso_hash_insert(fetch->hash, hash_key, translate); - /*debug_printf("\tCREATED with %d\n", hash_key);*/ - } else { - translate = cso_hash_iter_data(iter); - /*debug_printf("\tOK with %d\n", hash_key);*/ } return translate; -- cgit v1.2.3 From 53cf833af9776d47bd4c14906784aa8f2027dc4b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 24 Apr 2008 00:10:51 +0900 Subject: gallium: Add extern "C" to the headers. --- src/gallium/auxiliary/util/u_blit.h | 8 ++++++++ src/gallium/auxiliary/util/u_gen_mipmap.h | 10 ++++++++++ 2 files changed, 18 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h index 61f1d9bb32..0ce9732e62 100644 --- a/src/gallium/auxiliary/util/u_blit.h +++ b/src/gallium/auxiliary/util/u_blit.h @@ -30,7 +30,11 @@ #define U_BLIT_H +#ifdef __cplusplus +extern "C" { +#endif + struct pipe_context; struct pipe_surface; struct cso_context; @@ -58,4 +62,8 @@ util_blit_pixels(struct blit_state *ctx, float z, uint filter); +#ifdef __cplusplus +} +#endif + #endif diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.h b/src/gallium/auxiliary/util/u_gen_mipmap.h index bd9af54fb7..3277024f07 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.h +++ b/src/gallium/auxiliary/util/u_gen_mipmap.h @@ -31,6 +31,11 @@ #include "pipe/p_state.h" +#ifdef __cplusplus +extern "C" { +#endif + + struct pipe_context; struct pipe_texture; struct cso_context; @@ -52,4 +57,9 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, struct pipe_texture *pt, uint face, uint baseLevel, uint lastLevel, uint filter); + +#ifdef __cplusplus +} +#endif + #endif -- cgit v1.2.3 From a75a3df851339c782e045e01c2b21ffadb1e09f5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 24 Apr 2008 01:59:57 +0900 Subject: pipebuffer: New function to flush the buffer cache. --- src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 3 +++ src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c | 12 +++++++++--- 2 files changed, 12 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index 96f9af3825..8de286e3f9 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -144,6 +144,9 @@ struct pb_manager * pb_cache_manager_create(struct pb_manager *provider, unsigned usecs); +void +pb_cache_flush(struct pb_manager *mgr); + /** * Fenced buffer manager. diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c index 543fd51253..4bd3f94a6c 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c @@ -294,8 +294,8 @@ pb_cache_manager_create_buffer(struct pb_manager *_mgr, } -static void -pb_cache_manager_destroy(struct pb_manager *_mgr) +void +pb_cache_flush(struct pb_manager *_mgr) { struct pb_cache_manager *mgr = pb_cache_manager(_mgr); struct list_head *curr, *next; @@ -311,7 +311,13 @@ pb_cache_manager_destroy(struct pb_manager *_mgr) next = curr->next; } _glthread_UNLOCK_MUTEX(mgr->mutex); - +} + + +static void +pb_cache_manager_destroy(struct pb_manager *mgr) +{ + pb_cache_flush(mgr); FREE(mgr); } -- cgit v1.2.3 From 4f93a3a680560940630c44be0066b72a86ff008a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 24 Apr 2008 02:13:31 +0900 Subject: gallium: Fix texture refcount leak. --- src/gallium/auxiliary/cso_cache/cso_context.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 0523cb1949..d246dff433 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -104,6 +104,8 @@ out: static void cso_release_all( struct cso_context *ctx ) { + unsigned i; + if (ctx->pipe) { ctx->pipe->bind_blend_state( ctx->pipe, NULL ); ctx->pipe->bind_rasterizer_state( ctx->pipe, NULL ); @@ -113,6 +115,11 @@ static void cso_release_all( struct cso_context *ctx ) ctx->pipe->bind_vs_state( ctx->pipe, NULL ); } + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + pipe_texture_reference(&ctx->textures[i], NULL); + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + pipe_texture_reference(&ctx->textures_saved[i], NULL); + if (ctx->cache) { cso_cache_delete( ctx->cache ); ctx->cache = NULL; -- cgit v1.2.3 From e1180c2d694851ed12e86027aa406ee20546e6d3 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 23 Apr 2008 13:32:41 -0400 Subject: fix the simple hash finding function and use it --- src/gallium/auxiliary/cso_cache/cso_cache.c | 4 ++-- src/gallium/auxiliary/draw/draw_pt_emit.c | 14 ++++++-------- src/gallium/auxiliary/draw/draw_pt_fetch.c | 17 ++++------------- 3 files changed, 12 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 63464e0705..096875807b 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -255,9 +255,9 @@ void *cso_hash_find_data_from_template( struct cso_hash *hash, while (!cso_hash_iter_is_null(iter)) { void *iter_data = cso_hash_iter_data(iter); if (!memcmp(iter_data, templ, size)) { - /* Return the payload: + /* We found a match */ - return (unsigned char *)iter_data + size; + return iter_data; } iter = cso_hash_iter_next(iter); } diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index d35329aba0..92ad8f7049 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -65,16 +65,14 @@ static struct translate *cached_translate(struct pt_emit *emit, struct translate_key *key) { unsigned hash_key = create_key(key); - struct cso_hash_iter iter = cso_hash_find(emit->hash, hash_key); - struct translate *translate = 0; - - if (cso_hash_iter_is_null(iter)) { + struct translate *translate = (struct translate*) + cso_hash_find_data_from_template(emit->hash, + hash_key, + key, sizeof(*key)); + if (!translate) { + /* create/insert */ translate = translate_create(key); cso_hash_insert(emit->hash, hash_key, translate); - /*debug_printf("\tCREATED with %d\n", hash_key);*/ - } else { - translate = cso_hash_iter_data(iter); - /*debug_printf("\tOK with %d\n", hash_key);*/ } return translate; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index a8d4be5b64..013d16e1bc 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -67,19 +67,10 @@ static struct translate *cached_translate(struct pt_fetch *fetch, struct translate_key *key) { unsigned hash_key = create_key(key); - struct cso_hash_iter iter = cso_hash_find(fetch->hash, hash_key); - struct translate *translate = 0; - - while (!cso_hash_iter_is_null(iter)) { - void *iter_data = cso_hash_iter_data(iter); - if (memcmp(iter_data, key, sizeof(*key)) == 0) { - /* found */ - translate = cso_hash_iter_data(iter); - break; - } - iter = cso_hash_iter_next(iter); - /*debug_printf("\tOK with %d\n", hash_key);*/ - } + struct translate *translate = (struct translate*) + cso_hash_find_data_from_template(fetch->hash, + hash_key, + key, sizeof(*key)); if (!translate) { /* create/insert */ -- cgit v1.2.3 From 76d39f0c19ee0673b65d6ad09ab338c8b750251a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 11:22:18 -0600 Subject: gallium: more vertex debug code --- src/gallium/drivers/softpipe/sp_setup.c | 24 +++++++++++++++++------- 1 file changed, 17 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 7df8fc5f67..df7be01fcd 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -719,6 +719,13 @@ void setup_tri( struct setup_context *setup, { float det = calc_det(v0, v1, v2); +#if DEBUG_VERTS + debug_printf("Setup triangle:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); + print_vertex(setup, v2); +#endif + if (setup->softpipe->no_rast) return; @@ -731,13 +738,6 @@ void setup_tri( struct setup_context *setup, setup->numFragsWritten = 0; #endif -#if DEBUG_VERTS - debug_printf("Triangle:\n"); - print_vertex(setup, v0); - print_vertex(setup, v1); - print_vertex(setup, v2); -#endif - if (cull_tri( setup, det )) return; @@ -935,6 +935,12 @@ setup_line(struct setup_context *setup, int dy = y1 - y0; int xstep, ystep; +#if DEBUG_VERTS + debug_printf("Setup line:\n"); + print_vertex(setup, v0); + print_vertex(setup, v1); +#endif + if (setup->softpipe->no_rast) return; @@ -1056,6 +1062,10 @@ setup_point( struct setup_context *setup, const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); uint fragSlot; +#if DEBUG_VERTS + debug_printf("Setup point:\n"); + print_vertex(setup, v0); +#endif if (softpipe->no_rast) return; -- cgit v1.2.3 From 2221cb9f74ceee826efb09840188711f408e5428 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 11:23:12 -0600 Subject: gallium: fix broken PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE --- src/gallium/drivers/softpipe/sp_quad_blend.c | 17 +++++++---------- 1 file changed, 7 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index 802472df45..ca266ec8e6 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -294,11 +294,12 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: { const float *alpha = quadColor[3]; - float diff[4]; + float diff[4], temp[4]; VEC4_SUB(diff, one, dest[3]); - VEC4_MIN(source[0], alpha, diff); /* R */ - VEC4_MIN(source[1], alpha, diff); /* G */ - VEC4_MIN(source[2], alpha, diff); /* B */ + VEC4_MIN(temp, alpha, diff); + VEC4_MUL(source[0], quadColor[0], temp); /* R */ + VEC4_MUL(source[1], quadColor[1], temp); /* G */ + VEC4_MUL(source[2], quadColor[2], temp); /* B */ } break; case PIPE_BLENDFACTOR_CONST_COLOR: @@ -426,12 +427,8 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */ break; case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE: - { - const float *alpha = quadColor[3]; - float diff[4]; - VEC4_SUB(diff, one, dest[3]); - VEC4_MIN(source[3], alpha, diff); /* A */ - } + /* multiply alpha by 1.0 */ + VEC4_COPY(source[3], quadColor[3]); /* A */ break; case PIPE_BLENDFACTOR_CONST_COLOR: /* fall-through */ -- cgit v1.2.3 From 809bc8f9ad3667c297afa2652c6688f98d6451b0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 11:24:42 -0600 Subject: gallium: move logicop test outside of loop --- src/gallium/drivers/softpipe/sp_quad_blend.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index ca266ec8e6..74c6bff84a 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -232,6 +232,11 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) struct softpipe_context *softpipe = qs->softpipe; uint cbuf; + if (softpipe->blend->logicop_enable) { + logicop_quad(qs, quad); + return; + } + /* loop over colorbuffer outputs */ for (cbuf = 0; cbuf < softpipe->framebuffer.num_cbufs; cbuf++) { float source[4][QUAD_SIZE], dest[4][QUAD_SIZE]; @@ -242,11 +247,6 @@ blend_quad(struct quad_stage *qs, struct quad_header *quad) float (*quadColor)[4] = quad->outputs.color[cbuf]; uint i, j; - if (softpipe->blend->logicop_enable) { - logicop_quad(qs, quad); - return; - } - /* get/swizzle dest colors */ for (j = 0; j < QUAD_SIZE; j++) { int x = (quad->x0 & (TILE_SIZE-1)) + (j & 1); -- cgit v1.2.3 From 5d873c87186fd3a59b46b4e1e0c987120aa961a7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 11:36:47 -0600 Subject: gallium: additional debug code --- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 13 +++++++++++++ 1 file changed, 13 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index f98e130ed6..65cb8ba701 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -61,6 +61,12 @@ compute_clipmask_gl(const float *clip, /*const*/ float plane[][4], unsigned nr) unsigned mask = 0x0; unsigned i; +#if 0 + debug_printf("compute clipmask %f %f %f %f\n", + clip[0], clip[1], clip[2], clip[3]); + assert(clip[3] != 0.0); +#endif + /* Do the hardwired planes first: */ if (-clip[0] + clip[3] < 0) mask |= (1<<0); @@ -122,6 +128,13 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs, out->data[0][1] = out->data[0][1] * w * scale[1] + trans[1]; out->data[0][2] = out->data[0][2] * w * scale[2] + trans[2]; out->data[0][3] = w; +#if 0 + debug_printf("post viewport: %f %f %f %f\n", + out->data[0][0], + out->data[0][1], + out->data[0][2], + out->data[0][3]); +#endif } out = (struct vertex_header *)( (char *)out + stride ); -- cgit v1.2.3 From ba47aabc9868b410cdfe3bc8b6d25a44a598cba2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 11:37:49 -0600 Subject: gallium: fix potential divide by zero in fog computation It's legal for Fog.Start == Fog.End and conformance testing does so. --- src/mesa/shader/prog_statevars.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 45352d30b5..eee4d2d925 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -253,7 +253,8 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], value[0] = ctx->Fog.Density; value[1] = ctx->Fog.Start; value[2] = ctx->Fog.End; - value[3] = 1.0F / (ctx->Fog.End - ctx->Fog.Start); + value[3] = (ctx->Fog.End == ctx->Fog.Start) + ? 1.0 : 1.0F / (ctx->Fog.End - ctx->Fog.Start); return; case STATE_CLIPPLANE: { @@ -424,8 +425,9 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], * exp: 2^-(density/ln(2) * fogcoord) * exp2: 2^-((density/(ln(2)^2) * fogcoord)^2) */ - value[0] = -1.0F / (ctx->Fog.End - ctx->Fog.Start); - value[1] = ctx->Fog.End / (ctx->Fog.End - ctx->Fog.Start); + value[0] = (ctx->Fog.End == ctx->Fog.Start) + ? 1.0 : -1.0F / (ctx->Fog.End - ctx->Fog.Start); + value[1] = ctx->Fog.End * -value[0]; value[2] = ctx->Fog.Density * ONE_DIV_LN2; value[3] = ctx->Fog.Density * ONE_DIV_SQRT_LN2; return; -- cgit v1.2.3 From 5fcd84ab39318a371253b1a7285bc657fb82efed Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 23 Apr 2008 14:00:13 -0400 Subject: Create a sharable translate_cache and use it. --- src/gallium/auxiliary/cso_cache/cso_hash.h | 4 +- src/gallium/auxiliary/draw/draw_pt_emit.c | 65 ++------------ src/gallium/auxiliary/draw/draw_pt_fetch.c | 68 ++------------- src/gallium/auxiliary/translate/Makefile | 3 +- src/gallium/auxiliary/translate/SConscript | 1 + src/gallium/auxiliary/translate/translate.h | 1 - src/gallium/auxiliary/translate/translate_cache.c | 100 ++++++++++++++++++++++ src/gallium/auxiliary/translate/translate_cache.h | 54 ++++++++++++ 8 files changed, 173 insertions(+), 123 deletions(-) create mode 100644 src/gallium/auxiliary/translate/translate_cache.c create mode 100644 src/gallium/auxiliary/translate/translate_cache.h (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.h b/src/gallium/auxiliary/cso_cache/cso_hash.h index 73c4742006..85f3e276c6 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.h +++ b/src/gallium/auxiliary/cso_cache/cso_hash.h @@ -106,12 +106,12 @@ struct cso_hash_iter cso_hash_iter_prev(struct cso_hash_iter iter); /** - * Convenience routine to iterate over the collision list while doing a memory + * Convenience routine to iterate over the collision list while doing a memory * comparison to see which entry in the list is a direct copy of our template * and returns that entry. */ void *cso_hash_find_data_from_template( struct cso_hash *hash, - unsigned hash_key, + unsigned hash_key, void *templ, int size ); diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 92ad8f7049..c6d9537530 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -32,66 +32,16 @@ #include "draw/draw_vertex.h" #include "draw/draw_pt.h" #include "translate/translate.h" - -#include "cso_cache/cso_cache.h" -#include "cso_cache/cso_hash.h" +#include "translate/translate_cache.h" struct pt_emit { struct draw_context *draw; struct translate *translate; - struct cso_hash *hash; + struct translate_cache *cache; }; -static INLINE unsigned translate_hash_key_size(struct translate_key *key) -{ - unsigned size = sizeof(struct translate_key) - - sizeof(struct translate_element) * (PIPE_MAX_ATTRIBS - key->nr_elements); - return size; -} - -static INLINE unsigned create_key(struct translate_key *key) -{ - unsigned hash_key; - unsigned size = translate_hash_key_size(key); - /*debug_printf("key size = %d, (els = %d)\n", - size, key->nr_elements);*/ - hash_key = cso_construct_key(key, size); - return hash_key; -} - -static struct translate *cached_translate(struct pt_emit *emit, - struct translate_key *key) -{ - unsigned hash_key = create_key(key); - struct translate *translate = (struct translate*) - cso_hash_find_data_from_template(emit->hash, - hash_key, - key, sizeof(*key)); - if (!translate) { - /* create/insert */ - translate = translate_create(key); - cso_hash_insert(emit->hash, hash_key, translate); - } - - return translate; -} - - -static INLINE void delete_translates(struct pt_emit *emit) -{ - struct cso_hash *hash = emit->hash; - struct cso_hash_iter iter = cso_hash_first_node(hash); - while (!cso_hash_iter_is_null(iter)) { - struct translate *state = (struct translate*)cso_hash_iter_data(iter); - iter = cso_hash_iter_next(iter); - if (state) { - state->release(state); - } - } -} - void draw_pt_emit_prepare( struct pt_emit *emit, unsigned prim ) { @@ -169,12 +119,10 @@ void draw_pt_emit_prepare( struct pt_emit *emit, hw_key.nr_elements = vinfo->num_attribs; hw_key.output_stride = vinfo->size * 4; - /* Don't bother with caching at this stage: - */ if (!emit->translate || - memcmp(&emit->translate->key, &hw_key, sizeof(hw_key)) != 0) + memcmp(&emit->translate->key, &hw_key, sizeof(hw_key)) != 0) { - emit->translate = cached_translate(emit, &hw_key); + emit->translate = translate_cache_find(emit->cache, &hw_key); } } @@ -236,15 +184,14 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw ) return NULL; emit->draw = draw; - emit->hash = cso_hash_create(); + emit->cache = translate_cache_create(); return emit; } void draw_pt_emit_destroy( struct pt_emit *emit ) { - delete_translates(emit); - cso_hash_delete(emit->hash); + translate_cache_destroy(emit->cache); FREE(emit); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 013d16e1bc..8183c51676 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -32,9 +32,8 @@ #include "draw/draw_vertex.h" #include "draw/draw_pt.h" #include "translate/translate.h" +#include "translate/translate_cache.h" -#include "cso_cache/cso_cache.h" -#include "cso_cache/cso_hash.h" struct pt_fetch { struct draw_context *draw; @@ -43,57 +42,9 @@ struct pt_fetch { unsigned vertex_size; - struct cso_hash *hash; + struct translate_cache *cache; }; -static INLINE unsigned translate_hash_key_size(struct translate_key *key) -{ - unsigned size = sizeof(struct translate_key) - - sizeof(struct translate_element) * (PIPE_MAX_ATTRIBS - key->nr_elements); - return size; -} - -static INLINE unsigned create_key(struct translate_key *key) -{ - unsigned hash_key; - unsigned size = translate_hash_key_size(key); - /*debug_printf("key size = %d, (els = %d)\n", - size, key->nr_elements);*/ - hash_key = cso_construct_key(key, size); - return hash_key; -} - -static struct translate *cached_translate(struct pt_fetch *fetch, - struct translate_key *key) -{ - unsigned hash_key = create_key(key); - struct translate *translate = (struct translate*) - cso_hash_find_data_from_template(fetch->hash, - hash_key, - key, sizeof(*key)); - - if (!translate) { - /* create/insert */ - translate = translate_create(key); - cso_hash_insert(fetch->hash, hash_key, translate); - } - - return translate; -} - -static INLINE void delete_translates(struct pt_fetch *fetch) -{ - struct cso_hash *hash = fetch->hash; - struct cso_hash_iter iter = cso_hash_first_node(hash); - while (!cso_hash_iter_is_null(iter)) { - struct translate *state = (struct translate*)cso_hash_iter_data(iter); - iter = cso_hash_iter_next(iter); - if (state) { - state->release(state); - } - } -} - /* Perform the fetch from API vertex elements & vertex buffers, to a * contiguous set of float[4] attributes as required for the * vertex_shader->run_linear() method. @@ -157,17 +108,15 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, key.output_stride = vertex_size; - /* Don't bother with caching at this stage: - */ if (!fetch->translate || - memcmp(&fetch->translate->key, &key, sizeof(key)) != 0) + memcmp(&fetch->translate->key, &key, sizeof(key)) != 0) { - fetch->translate = cached_translate(fetch, &key); + fetch->translate = translate_cache_find(fetch->cache, &key); { static struct vertex_header vh = { 0, 0, 0, 0xffff }; - fetch->translate->set_buffer(fetch->translate, - draw->pt.nr_vertex_buffers, + fetch->translate->set_buffer(fetch->translate, + draw->pt.nr_vertex_buffers, &vh, 0); } @@ -208,14 +157,13 @@ struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) return NULL; fetch->draw = draw; - fetch->hash = cso_hash_create(); + fetch->cache = translate_cache_create(); return fetch; } void draw_pt_fetch_destroy( struct pt_fetch *fetch ) { - delete_translates(fetch); - cso_hash_delete(fetch->hash); + translate_cache_destroy(fetch->cache); FREE(fetch); } diff --git a/src/gallium/auxiliary/translate/Makefile b/src/gallium/auxiliary/translate/Makefile index 39dfb0de30..ad2a5b705e 100644 --- a/src/gallium/auxiliary/translate/Makefile +++ b/src/gallium/auxiliary/translate/Makefile @@ -6,7 +6,8 @@ LIBNAME = translate C_SOURCES = \ translate_generic.c \ translate_sse.c \ - translate.c + translate.c \ + translate_cache.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/translate/SConscript b/src/gallium/auxiliary/translate/SConscript index 7608908915..9553a67537 100644 --- a/src/gallium/auxiliary/translate/SConscript +++ b/src/gallium/auxiliary/translate/SConscript @@ -6,6 +6,7 @@ translate = env.ConvenienceLibrary( 'translate_generic.c', 'translate_sse.c', 'translate.c', + 'translate_cache.c', ]) auxiliaries.insert(0, translate) diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h index d95d1ac4f3..6c15d7e4dc 100644 --- a/src/gallium/auxiliary/translate/translate.h +++ b/src/gallium/auxiliary/translate/translate.h @@ -96,7 +96,6 @@ struct translate *translate_lookup_or_create( struct translate_context *tctx, struct translate *translate_create( const struct translate_key *key ); - /******************************************************************************* * Private: */ diff --git a/src/gallium/auxiliary/translate/translate_cache.c b/src/gallium/auxiliary/translate/translate_cache.c new file mode 100644 index 0000000000..c14f37c42f --- /dev/null +++ b/src/gallium/auxiliary/translate/translate_cache.c @@ -0,0 +1,100 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_state.h" +#include "translate.h" + +#include "cso_cache/cso_cache.h" +#include "cso_cache/cso_hash.h" + +struct translate_cache { + struct cso_hash *hash; +}; + +struct translate_cache * translate_cache_create() +{ + struct translate_cache *cache = MALLOC_STRUCT(translate_cache); + cache->hash = cso_hash_create(); +} + + +static INLINE void delete_translates(struct translate_cache *cache) +{ + struct cso_hash *hash = cache->hash; + struct cso_hash_iter iter = cso_hash_first_node(hash); + while (!cso_hash_iter_is_null(iter)) { + struct translate *state = (struct translate*)cso_hash_iter_data(iter); + iter = cso_hash_iter_next(iter); + if (state) { + state->release(state); + } + } +} + +void translate_cache_destroy(struct translate_cache *cache) +{ + delete_translates(cache); + cso_hash_delete(cache->hash); + FREE(cache); +} + + +static INLINE unsigned translate_hash_key_size(struct translate_key *key) +{ + unsigned size = sizeof(struct translate_key) - + sizeof(struct translate_element) * (PIPE_MAX_ATTRIBS - key->nr_elements); + return size; +} + +static INLINE unsigned create_key(struct translate_key *key) +{ + unsigned hash_key; + unsigned size = translate_hash_key_size(key); + /*debug_printf("key size = %d, (els = %d)\n", + size, key->nr_elements);*/ + hash_key = cso_construct_key(key, size); + return hash_key; +} + +struct translate * translate_cache_find(struct translate_cache *cache, + struct translate_key *key) +{ + unsigned hash_key = create_key(key); + struct translate *translate = (struct translate*) + cso_hash_find_data_from_template(cache->hash, + hash_key, + key, sizeof(*key)); + + if (!translate) { + /* create/insert */ + translate = translate_create(key); + cso_hash_insert(cache->hash, hash_key, translate); + } + + return translate; +} diff --git a/src/gallium/auxiliary/translate/translate_cache.h b/src/gallium/auxiliary/translate/translate_cache.h new file mode 100644 index 0000000000..63fc57b7ea --- /dev/null +++ b/src/gallium/auxiliary/translate/translate_cache.h @@ -0,0 +1,54 @@ +/* + * Copyright 2008 Tungsten Graphics, inc. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _TRANSLATE_CACHE_H +#define _TRANSLATE_CACHE_H + + +/******************************************************************************* + * Translate cache. + * Simply used to cache created translates. Avoids unecessary creation of + * translate's if one suitable for a given translate_key has already been + * created. + * + * Note: this functionality depends and requires the CSO module. + */ +struct translate_cache; + +struct translate_key; +struct translate; + +struct translate_cache *translate_cache_create(); +void translate_cache_destroy(struct translate_cache *cache); + +/** + * Will try to find a translate structure matched by the given key. + * If such a structure doesn't exist in the cache the function + * will automatically create it, insert it in the cache and + * return the created version. + * + */ +struct translate *translate_cache_find(struct translate_cache *cache, + struct translate_key *key); + +#endif -- cgit v1.2.3 From 95f8f8863a80ce1e584160d4d085213a9bbaef12 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 23 Apr 2008 14:07:53 -0400 Subject: Add translate cache to fetch_emit stage and add out of memory checks to code creating the cache. --- src/gallium/auxiliary/draw/draw_pt_emit.c | 4 ++++ src/gallium/auxiliary/draw/draw_pt_fetch.c | 5 +++++ src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 20 +++++++++++++------- 3 files changed, 22 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index c6d9537530..35707af8a8 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -185,6 +185,10 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw ) emit->draw = draw; emit->cache = translate_cache_create(); + if (!emit->cache) { + FREE(emit); + return NULL; + } return emit; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 8183c51676..cc61b113fd 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -158,6 +158,11 @@ struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) fetch->draw = draw; fetch->cache = translate_cache_create(); + if (!fetch->cache) { + FREE(fetch); + return NULL; + } + return fetch; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 68b2c5b1e3..a4de341df8 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -37,6 +37,7 @@ #include "draw/draw_vertex.h" #include "draw/draw_pt.h" #include "translate/translate.h" +#include "translate/translate_cache.h" /* The simplest 'middle end' in the new vertex code. * @@ -81,6 +82,7 @@ struct fetch_emit_middle_end { */ float point_size; + struct translate_cache *cache; }; @@ -174,10 +176,9 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, if (!feme->translate || memcmp(&feme->translate->key, &key, sizeof(key)) != 0) { - if (feme->translate) - feme->translate->release(feme->translate); + feme->translate = translate_cache_find(feme->cache, + &key); - feme->translate = translate_create( &key ); feme->translate->set_buffer(feme->translate, draw->pt.nr_vertex_buffers, @@ -266,9 +267,8 @@ static void fetch_emit_destroy( struct draw_pt_middle_end *middle ) { struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; - if (feme->translate) - feme->translate->release( feme->translate ); - + translate_cache_destroy(feme->cache); + FREE(middle); } @@ -278,7 +278,13 @@ struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ) struct fetch_emit_middle_end *fetch_emit = CALLOC_STRUCT( fetch_emit_middle_end ); if (fetch_emit == NULL) return NULL; - + + fetch_emit->cache = translate_cache_create(); + if (!fetch_emit->cache) { + FREE(fetch_emit); + return NULL; + } + fetch_emit->base.prepare = fetch_emit_prepare; fetch_emit->base.run = fetch_emit_run; fetch_emit->base.finish = fetch_emit_finish; -- cgit v1.2.3 From 3f6242d3e4d3ee61884a91a3eef4be8dfaadee3c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 12:11:00 -0600 Subject: gallium: passthrough tri, not point --- src/gallium/auxiliary/draw/draw_pipe_wide_line.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c index 452732e662..878c9c7169 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c @@ -169,7 +169,7 @@ struct draw_stage *draw_wide_line_stage( struct draw_context *draw ) wide->stage.next = NULL; wide->stage.point = draw_pipe_passthrough_point; wide->stage.line = wideline_line; - wide->stage.tri = draw_pipe_passthrough_point; + wide->stage.tri = draw_pipe_passthrough_tri; wide->stage.flush = wideline_flush; wide->stage.reset_stipple_counter = wideline_reset_stipple_counter; wide->stage.destroy = wideline_destroy; -- cgit v1.2.3 From bff371c431b962c62d74800c543e09d258e67551 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 17:40:39 -0600 Subject: gallium: fix comments, whitespace changes --- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 21 +++++++-------------- 1 file changed, 7 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index aec485a6e7..848d3be55d 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -433,9 +433,7 @@ pstip_create_texture(struct pstip_stage *pstip) /** - * Create the sampler CSO that'll be used for antialiasing. - * By using a mipmapped texture, we don't have to generate a different - * texture image for each line size. + * Create the sampler CSO that'll be used for stippling. */ static boolean pstip_create_sampler(struct pstip_stage *pstip) @@ -463,8 +461,8 @@ pstip_create_sampler(struct pstip_stage *pstip) /** - * When we're about to draw our first AA line in a batch, this function is - * called to tell the driver to bind our modified fragment shader. + * When we're about to draw our first stipple polygon in a batch, this function + * is called to tell the driver to bind our modified fragment shader. */ static boolean bind_pstip_fragment_shader(struct pstip_stage *pstip) @@ -478,7 +476,6 @@ bind_pstip_fragment_shader(struct pstip_stage *pstip) } - static INLINE struct pstip_stage * pstip_stage( struct draw_stage *stage ) { @@ -486,9 +483,6 @@ pstip_stage( struct draw_stage *stage ) } - - - static void pstip_first_tri(struct draw_stage *stage, struct prim_header *header) { @@ -521,7 +515,7 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) pstip->driver_bind_sampler_states(pipe, num_samplers, pstip->state.samplers); pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures); - /* now really draw first line */ + /* now really draw first triangle */ stage->tri = draw_pipe_passthrough_tri; stage->tri(stage, header); } @@ -698,11 +692,10 @@ pstip_set_polygon_stipple(struct pipe_context *pipe, } - /** - * Called by drivers that want to install this AA line prim stage + * Called by drivers that want to install this polygon stipple stage * into the draw module's pipeline. This will not be used if the - * hardware has native support for AA lines. + * hardware has native support for polygon stipple. */ boolean draw_install_pstipple_stage(struct draw_context *draw, @@ -713,7 +706,7 @@ draw_install_pstipple_stage(struct draw_context *draw, pipe->draw = (void *) draw; /* - * Create / install AA line drawing / prim stage + * Create / install pgon stipple drawing / prim stage */ pstip = draw_pstip_stage( draw ); if (pstip == NULL) -- cgit v1.2.3 From 8437f5c763c1a1ac364d71426109c2b095bbcc72 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 17:41:30 -0600 Subject: gallium: fix comments --- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index ac0aa4cd7c..8c38c15d86 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -481,7 +481,7 @@ aa_transform_inst(struct tgsi_transform_context *ctx, /** - * Generate the frag shader we'll use for drawing AA lines. + * Generate the frag shader we'll use for drawing AA points. * This will be the user's shader plus some texture/modulate instructions. */ static boolean @@ -531,7 +531,7 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) /** - * When we're about to draw our first AA line in a batch, this function is + * When we're about to draw our first AA point in a batch, this function is * called to tell the driver to bind our modified fragment shader. */ static boolean @@ -697,7 +697,7 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header) } } - /* now really draw first line */ + /* now really draw first point */ stage->point = aapoint_point; stage->point(stage, header); } -- cgit v1.2.3 From 14d1ca8d867d6e44c756cb759f92421107118b2e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 18:08:20 -0600 Subject: gallium: fix issues in recursive flushing When flushing/rendering, some stages (like AA line/point) need to set pipe/driver state. Those driver functions often call draw_flush(). That leads to recursion. Use new draw->suspend_flush flag to explicitly prevent that in the key places. Remove the draw->vcache_flushing field. Reuse draw->flushing as a debug/assertion var. --- src/gallium/auxiliary/draw/draw_context.c | 4 +++- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 18 +++++++++++++++--- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 7 +++++++ src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 16 ++++++++++++++-- src/gallium/auxiliary/draw/draw_private.h | 6 +++--- src/gallium/auxiliary/draw/draw_pt_vcache.c | 2 -- 6 files changed, 42 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index f90187816b..98e23fa830 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -367,8 +367,10 @@ draw_set_mapped_element_buffer( struct draw_context *draw, */ void draw_do_flush( struct draw_context *draw, unsigned flags ) { - if (!draw->flushing && !draw->vcache_flushing) + if (!draw->suspend_flushing) { + assert(!draw->flushing); /* catch inadvertant recursion */ + draw->flushing = TRUE; draw_pipeline_flush( draw, flags ); diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index 7e5f8bd281..0545dbc171 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -485,11 +485,16 @@ aaline_create_sampler(struct aaline_stage *aaline) static boolean bind_aaline_fragment_shader(struct aaline_stage *aaline) { + struct draw_context *draw = aaline->stage.draw; + if (!aaline->fs->aaline_fs && !generate_aaline_fs(aaline)) return FALSE; + draw->suspend_flushing = TRUE; aaline->driver_bind_fs_state(aaline->pipe, aaline->fs->aaline_fs); + draw->suspend_flushing = FALSE; + return TRUE; } @@ -663,8 +668,10 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) pipe_texture_reference(&aaline->state.texture[aaline->fs->sampler_unit], aaline->texture); + draw->suspend_flushing = TRUE; aaline->driver_bind_sampler_states(pipe, num_samplers, aaline->state.sampler); aaline->driver_set_sampler_textures(pipe, num_samplers, aaline->state.texture); + draw->suspend_flushing = FALSE; /* now really draw first line */ stage->line = aaline_line; @@ -682,14 +689,14 @@ aaline_flush(struct draw_stage *stage, unsigned flags) stage->line = aaline_first_line; stage->next->flush( stage->next, flags ); - /* restore original frag shader */ + /* restore original frag shader, texture, sampler state */ + draw->suspend_flushing = TRUE; aaline->driver_bind_fs_state(pipe, aaline->fs->driver_fs); - - /* XXX restore original texture, sampler state */ aaline->driver_bind_sampler_states(pipe, aaline->num_samplers, aaline->state.sampler); aaline->driver_set_sampler_textures(pipe, aaline->num_textures, aaline->state.texture); + draw->suspend_flushing = FALSE; draw->extra_vp_outputs.slot = 0; } @@ -783,6 +790,7 @@ aaline_bind_fs_state(struct pipe_context *pipe, void *fs) { struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); struct aaline_fragment_shader *aafs = (struct aaline_fragment_shader *) fs; + /* save current */ aaline->fs = aafs; /* pass-through */ @@ -807,9 +815,12 @@ aaline_bind_sampler_states(struct pipe_context *pipe, unsigned num, void **sampler) { struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + struct draw_context *draw = aaline->stage.draw; + /* save current */ memcpy(aaline->state.sampler, sampler, num * sizeof(void *)); aaline->num_samplers = num; + /* pass-through */ aaline->driver_bind_sampler_states(aaline->pipe, num, sampler); } @@ -820,6 +831,7 @@ aaline_set_sampler_textures(struct pipe_context *pipe, unsigned num, struct pipe_texture **texture) { struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); + struct draw_context *draw = aaline->stage.draw; uint i; /* save current */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 8c38c15d86..122a48660a 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -537,11 +537,16 @@ generate_aapoint_fs(struct aapoint_stage *aapoint) static boolean bind_aapoint_fragment_shader(struct aapoint_stage *aapoint) { + struct draw_context *draw = aapoint->stage.draw; + if (!aapoint->fs->aapoint_fs && !generate_aapoint_fs(aapoint)) return FALSE; + draw->suspend_flushing = TRUE; aapoint->driver_bind_fs_state(aapoint->pipe, aapoint->fs->aapoint_fs); + draw->suspend_flushing = FALSE; + return TRUE; } @@ -714,7 +719,9 @@ aapoint_flush(struct draw_stage *stage, unsigned flags) stage->next->flush( stage->next, flags ); /* restore original frag shader */ + draw->suspend_flushing = TRUE; aapoint->driver_bind_fs_state(pipe, aapoint->fs->driver_fs); + draw->suspend_flushing = FALSE; draw->extra_vp_outputs.slot = 0; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 848d3be55d..f8156fe731 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -467,11 +467,14 @@ pstip_create_sampler(struct pstip_stage *pstip) static boolean bind_pstip_fragment_shader(struct pstip_stage *pstip) { + struct draw_context *draw = pstip->stage.draw; if (!pstip->fs->pstip_fs && !generate_pstip_fs(pstip)) return FALSE; + draw->suspend_flushing = TRUE; pstip->driver_bind_fs_state(pstip->pipe, pstip->fs->pstip_fs); + draw->suspend_flushing = FALSE; return TRUE; } @@ -488,6 +491,7 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) { struct pstip_stage *pstip = pstip_stage(stage); struct pipe_context *pipe = pstip->pipe; + struct draw_context *draw = stage->draw; uint num_samplers; assert(stage->draw->rasterizer->poly_stipple_enable); @@ -512,8 +516,10 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) assert(num_samplers <= PIPE_MAX_SAMPLERS); + draw->suspend_flushing = TRUE; pstip->driver_bind_sampler_states(pipe, num_samplers, pstip->state.samplers); pstip->driver_set_sampler_textures(pipe, num_samplers, pstip->state.textures); + draw->suspend_flushing = FALSE; /* now really draw first triangle */ stage->tri = draw_pipe_passthrough_tri; @@ -524,7 +530,7 @@ pstip_first_tri(struct draw_stage *stage, struct prim_header *header) static void pstip_flush(struct draw_stage *stage, unsigned flags) { - /*struct draw_context *draw = stage->draw;*/ + struct draw_context *draw = stage->draw; struct pstip_stage *pstip = pstip_stage(stage); struct pipe_context *pipe = pstip->pipe; @@ -534,11 +540,13 @@ pstip_flush(struct draw_stage *stage, unsigned flags) /* restore original frag shader */ pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs); - /* XXX restore original texture, sampler state */ + /* restore original texture, sampler state */ + draw->suspend_flushing = TRUE; pstip->driver_bind_sampler_states(pipe, pstip->num_samplers, pstip->state.samplers); pstip->driver_set_sampler_textures(pipe, pstip->num_textures, pstip->state.textures); + draw->suspend_flushing = FALSE; } @@ -661,6 +669,7 @@ pstip_set_sampler_textures(struct pipe_context *pipe, unsigned num, struct pipe_texture **texture) { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + struct draw_context *draw = pstip->stage.draw; uint i; /* save current */ @@ -683,8 +692,11 @@ pstip_set_polygon_stipple(struct pipe_context *pipe, const struct pipe_poly_stipple *stipple) { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); + struct draw_context *draw = (struct draw_context *) pipe->draw; + /* save current */ pstip->state.stipple = stipple; + /* pass-through */ pstip->driver_set_polygon_stipple(pstip->pipe, stipple); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 39aa81b43c..ca5021c5c6 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -178,9 +178,9 @@ struct draw_context boolean bypass_clipping; } driver; - boolean flushing; - boolean vcache_flushing; - boolean bypass_clipping; /* set if either api or driver bypass_clipping true */ + boolean flushing; /**< debugging/sanity */ + boolean suspend_flushing; /**< internally set */ + boolean bypass_clipping; /**< set if either api or driver bypass_clipping true */ /* pipe state that we need: */ const struct pipe_rasterizer_state *rasterizer; diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index afcff41043..153055417d 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -63,7 +63,6 @@ struct vcache_frontend { static void vcache_flush( struct vcache_frontend *vcache ) { - vcache->draw->vcache_flushing = TRUE; if (vcache->draw_count) { vcache->middle->run( vcache->middle, vcache->fetch_elts, @@ -75,7 +74,6 @@ static void vcache_flush( struct vcache_frontend *vcache ) memset(vcache->in, ~0, sizeof(vcache->in)); vcache->fetch_count = 0; vcache->draw_count = 0; - vcache->draw->vcache_flushing = FALSE; } static void vcache_check_flush( struct vcache_frontend *vcache ) -- cgit v1.2.3 From bb4f8ae1f93d17c57fd8f62bea24b48131e02037 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Apr 2008 18:09:20 -0600 Subject: gallium: reorder code to fix a recursive flush --- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index f8156fe731..da303a13ad 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -537,11 +537,9 @@ pstip_flush(struct draw_stage *stage, unsigned flags) stage->tri = pstip_first_tri; stage->next->flush( stage->next, flags ); - /* restore original frag shader */ - pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs); - - /* restore original texture, sampler state */ + /* restore original frag shader, texture, sampler state */ draw->suspend_flushing = TRUE; + pstip->driver_bind_fs_state(pipe, pstip->fs->driver_fs); pstip->driver_bind_sampler_states(pipe, pstip->num_samplers, pstip->state.samplers); pstip->driver_set_sampler_textures(pipe, pstip->num_textures, -- cgit v1.2.3 From 7333578d2a5fa18f7f0101fc3fd3b03cf2f356bc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 24 Apr 2008 10:12:07 +0100 Subject: gallium: Initial support for pixel formats with unused storage components. Also clarify that RGB formats with no (used) alpha component are treated as having alpha = 1.0. --- src/gallium/auxiliary/util/p_tile.c | 52 ++++++++++++++ src/gallium/auxiliary/util/u_gen_mipmap.c | 2 + src/gallium/auxiliary/util/u_pack_color.h | 38 +++++++++++ src/gallium/include/pipe/p_format.h | 108 +++++++++++++++++------------- src/mesa/state_tracker/st_cb_fbo.c | 4 ++ 5 files changed, 157 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 520da5cecd..91283a9ab0 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -169,6 +169,52 @@ a8r8g8b8_put_tile_rgba(unsigned *dst, } +/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ + +static void +x8r8g8b8_get_tile_rgba(unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; + pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); + pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); + pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); + pRow[3] = UBYTE_TO_FLOAT(0xff); + } + p += dst_stride; + } +} + + +static void +x8r8g8b8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + *dst++ = (0xff << 24) | (r << 16) | (g << 8) | b; + } + p += src_stride; + } +} + + /*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ static void @@ -647,6 +693,9 @@ pipe_get_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_A8R8G8B8_UNORM: a8r8g8b8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; + case PIPE_FORMAT_X8R8G8B8_UNORM: + x8r8g8b8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + break; case PIPE_FORMAT_B8G8R8A8_UNORM: b8g8r8a8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; @@ -723,6 +772,9 @@ pipe_put_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_A8R8G8B8_UNORM: a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); break; + case PIPE_FORMAT_X8R8G8B8_UNORM: + x8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); + break; case PIPE_FORMAT_B8G8R8A8_UNORM: b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); break; diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index dfdb5f16fe..b8dc6c66c0 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -475,7 +475,9 @@ format_to_type_comps(enum pipe_format pformat, { switch (pformat) { case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_X8R8G8B8_UNORM: case PIPE_FORMAT_B8G8R8A8_UNORM: + case PIPE_FORMAT_B8G8R8X8_UNORM: *datatype = UBYTE; *comps = 4; return; diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 1f6604c554..50ef44992b 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -53,18 +53,36 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, *d = (r << 24) | (g << 16) | (b << 8) | a; } return; + case PIPE_FORMAT_R8G8B8X8_UNORM: + { + uint *d = (uint *) dest; + *d = (r << 24) | (g << 16) | (b << 8) | 0xff; + } + return; case PIPE_FORMAT_A8R8G8B8_UNORM: { uint *d = (uint *) dest; *d = (a << 24) | (r << 16) | (g << 8) | b; } return; + case PIPE_FORMAT_X8R8G8B8_UNORM: + { + uint *d = (uint *) dest; + *d = (0xff << 24) | (r << 16) | (g << 8) | b; + } + return; case PIPE_FORMAT_B8G8R8A8_UNORM: { uint *d = (uint *) dest; *d = (b << 24) | (g << 16) | (r << 8) | a; } return; + case PIPE_FORMAT_B8G8R8X8_UNORM: + { + uint *d = (uint *) dest; + *d = (b << 24) | (g << 16) | (r << 8) | 0xff; + } + return; case PIPE_FORMAT_R5G6B5_UNORM: { ushort *d = (ushort *) dest; @@ -101,18 +119,36 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) *d = (r << 24) | (g << 16) | (b << 8) | a; } return; + case PIPE_FORMAT_R8G8B8X8_UNORM: + { + uint *d = (uint *) dest; + *d = (r << 24) | (g << 16) | (b << 8) | 0xff; + } + return; case PIPE_FORMAT_A8R8G8B8_UNORM: { uint *d = (uint *) dest; *d = (a << 24) | (r << 16) | (g << 8) | b; } return; + case PIPE_FORMAT_X8R8G8B8_UNORM: + { + uint *d = (uint *) dest; + *d = (0xff << 24) | (r << 16) | (g << 8) | b; + } + return; case PIPE_FORMAT_B8G8R8A8_UNORM: { uint *d = (uint *) dest; *d = (b << 24) | (g << 16) | (r << 8) | a; } return; + case PIPE_FORMAT_B8G8R8X8_UNORM: + { + uint *d = (uint *) dest; + *d = (b << 24) | (g << 16) | (r << 8) | 0xff; + } + return; case PIPE_FORMAT_R5G6B5_UNORM: { ushort *d = (ushort *) dest; @@ -159,8 +195,10 @@ util_pack_z(enum pipe_format format, double z) else return (uint) (z * 0xffffffff); case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: return (uint) (z * 0xffffff); case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: return ((uint) (z * 0xffffff)) << 8; default: debug_printf("gallium: unhandled fomrat in util_pack_z()"); diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index ef9e3a3d6c..2ee6f100bf 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -163,18 +163,21 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) /** * Shorthand macro for common format swizzles. */ -#define _PIPE_FORMAT_R000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_RG00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) -#define _PIPE_FORMAT_RGB0 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_R001 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_1 ) +#define _PIPE_FORMAT_RG01 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_1 ) +#define _PIPE_FORMAT_RGB1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_1 ) #define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) #define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) #define _PIPE_FORMAT_BGRA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_A ) +#define _PIPE_FORMAT_1RGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_1, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) +#define _PIPE_FORMAT_BGR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) #define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R ) #define _PIPE_FORMAT_RRR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) #define _PIPE_FORMAT_RRRR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R ) #define _PIPE_FORMAT_RRRG _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G ) #define _PIPE_FORMAT_Z000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) +#define _PIPE_FORMAT_0Z00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_SZ00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_ZS00 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_Z, PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_S000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_S, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) @@ -229,10 +232,12 @@ static INLINE uint pf_rev(pipe_format_ycbcr_t f) enum pipe_format { PIPE_FORMAT_NONE = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_0000, 0, 0, 0, 0, PIPE_FORMAT_TYPE_UNKNOWN ), PIPE_FORMAT_A8R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ARGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_X8R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_1RGB, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_B8G8R8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_B8G8R8X8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_A1R5G5B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_A4R4G4B4_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB0, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB1, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_U_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ PIPE_FORMAT_U_A8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ PIPE_FORMAT_U_I8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ @@ -244,68 +249,75 @@ enum pipe_format { PIPE_FORMAT_Z32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), PIPE_FORMAT_S8Z24_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_SZ00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_Z24S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_ZS00, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_X8Z24_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_0Z00, 1, 3, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_Z24X8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 3, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_S8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_S000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte stencil */ - PIPE_FORMAT_R64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R64G64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R64G64B64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R64G64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R64G64B64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_FLOAT ), PIPE_FORMAT_R64G64B64A64_FLOAT = _PIPE_FORMAT_RGBAZS_64( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32G32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32G32B32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32G32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FLOAT ), + PIPE_FORMAT_R32G32B32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FLOAT ), PIPE_FORMAT_R32G32B32A32_FLOAT = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FLOAT ), - PIPE_FORMAT_R32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32G32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32G32B32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32G32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R32G32B32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_R32G32B32A32_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32G32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32G32B32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32G32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R32G32B32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_USCALED ), PIPE_FORMAT_R32G32B32A32_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32G32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32G32B32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32G32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R32G32B32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SNORM ), PIPE_FORMAT_R32G32B32A32_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R32G32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R32G32B32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32G32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32G32B32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R32G32B32A32_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16G16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16G16B16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16G16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R16G16B16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_R16G16B16A16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16G16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16G16B16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16G16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R16G16B16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_USCALED ), PIPE_FORMAT_R16G16B16A16_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16G16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16G16B16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16G16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R16G16B16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SNORM ), PIPE_FORMAT_R16G16B16A16_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16G16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R16G16B16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 2, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16G16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 2, 2, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R16G16B16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R16G16B16A16_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 2, 2, 2, 2, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8G8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8G8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8G8B8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_R8G8B8A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_R8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8G8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8G8B8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8G8B8X8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_R8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8G8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8G8B8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_USCALED ), PIPE_FORMAT_R8G8B8A8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ), - PIPE_FORMAT_R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8G8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8G8B8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8G8B8X8_USCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_USCALED ), + PIPE_FORMAT_R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8G8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8G8B8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ), PIPE_FORMAT_R8G8B8A8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), - PIPE_FORMAT_R8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), - PIPE_FORMAT_R8G8B8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8B8X8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_R8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8B8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R8G8B8A8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R8G8B8X8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ), /* sRGB formats */ PIPE_FORMAT_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), PIPE_FORMAT_A8_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), - PIPE_FORMAT_R8G8B8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ), + PIPE_FORMAT_R8G8B8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ), PIPE_FORMAT_R8G8B8A8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), + PIPE_FORMAT_R8G8B8X8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), /* compressed formats */ PIPE_FORMAT_DXT1_RGB = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 0 ), @@ -442,6 +454,8 @@ static INLINE uint pf_get_bits( enum pipe_format format ) switch (pf_layout(format)) { case PIPE_FORMAT_LAYOUT_RGBAZS: return + pf_get_component_bits( format, PIPE_FORMAT_COMP_0 ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_1 ) + pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) + pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) + pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) + diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index b1a56f3ca6..b73867bb72 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -252,6 +252,8 @@ st_new_renderbuffer_fb(enum pipe_format format) switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: case PIPE_FORMAT_B8G8R8A8_UNORM: + case PIPE_FORMAT_X8R8G8B8_UNORM: + case PIPE_FORMAT_B8G8R8X8_UNORM: case PIPE_FORMAT_A1R5G5B5_UNORM: case PIPE_FORMAT_A4R4G4B4_UNORM: case PIPE_FORMAT_R5G6B5_UNORM: @@ -268,6 +270,8 @@ st_new_renderbuffer_fb(enum pipe_format format) break; case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_Z24S8_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: + case PIPE_FORMAT_Z24X8_UNORM: strb->Base.InternalFormat = GL_DEPTH24_STENCIL8_EXT; strb->Base._BaseFormat = GL_DEPTH_STENCIL_EXT; break; -- cgit v1.2.3 From dddedd915afb58ab6b87492e850baeadc007fe47 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 24 Apr 2008 12:36:04 +0100 Subject: xlib: only shortcircuit makecurrent if NO_RAST set, otherwise window size updates fail --- src/gallium/winsys/xlib/fakeglx.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/fakeglx.c b/src/gallium/winsys/xlib/fakeglx.c index 6e04cb4117..ec77e81fed 100644 --- a/src/gallium/winsys/xlib/fakeglx.c +++ b/src/gallium/winsys/xlib/fakeglx.c @@ -1459,6 +1459,13 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, GLXDrawable read, GLXContext ctx ) { struct fake_glx_context *glxCtx = (struct fake_glx_context *) ctx; + static boolean firsttime = 1, no_rast = 0; + + if (firsttime) { + no_rast = getenv("SP_NO_RAST") != NULL; + firsttime = 0; + } + if (ctx && draw && read) { XMesaBuffer drawBuffer, readBuffer; @@ -1504,7 +1511,8 @@ Fake_glXMakeContextCurrent( Display *dpy, GLXDrawable draw, #endif } - if (MakeCurrent_PrevContext == ctx && + if (no_rast && + MakeCurrent_PrevContext == ctx && MakeCurrent_PrevDrawable == draw && MakeCurrent_PrevReadable == read && MakeCurrent_PrevDrawBuffer == drawBuffer && -- cgit v1.2.3 From f93332da5655a31b6c44a1079629a15360ff999b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 24 Apr 2008 12:38:15 +0100 Subject: draw: handle edgeflags and reset-line-stipple again --- src/gallium/auxiliary/draw/draw_pipe.c | 19 +- src/gallium/auxiliary/draw/draw_pipe.h | 11 + src/gallium/auxiliary/draw/draw_pipe_clip.c | 56 ++-- src/gallium/auxiliary/draw/draw_pipe_flatshade.c | 6 +- src/gallium/auxiliary/draw/draw_pipe_offset.c | 3 +- src/gallium/auxiliary/draw/draw_pipe_stipple.c | 4 + src/gallium/auxiliary/draw/draw_pipe_twoside.c | 3 +- src/gallium/auxiliary/draw/draw_pipe_unfilled.c | 24 +- src/gallium/auxiliary/draw/draw_private.h | 40 +-- src/gallium/auxiliary/draw/draw_pt.h | 9 - src/gallium/auxiliary/draw/draw_pt_fetch.c | 17 ++ src/gallium/auxiliary/draw/draw_pt_vcache.c | 356 +++++------------------ src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h | 174 +++++++++++ 13 files changed, 348 insertions(+), 374 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index d0890203a5..1c9d8650e2 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -116,8 +116,7 @@ static void do_point( struct draw_context *draw, { struct prim_header prim; - prim.reset_line_stipple = 0; - prim.edgeflags = 1; + prim.flags = 0; prim.pad = 0; prim.v[0] = (struct vertex_header *)v0; @@ -126,13 +125,13 @@ static void do_point( struct draw_context *draw, static void do_line( struct draw_context *draw, + ushort flags, const char *v0, const char *v1 ) { struct prim_header prim; - prim.reset_line_stipple = 1; /* fixme */ - prim.edgeflags = 1; + prim.flags = flags; prim.pad = 0; prim.v[0] = (struct vertex_header *)v0; prim.v[1] = (struct vertex_header *)v1; @@ -142,6 +141,7 @@ static void do_line( struct draw_context *draw, static void do_triangle( struct draw_context *draw, + ushort flags, char *v0, char *v1, char *v2 ) @@ -151,10 +151,7 @@ static void do_triangle( struct draw_context *draw, prim.v[0] = (struct vertex_header *)v0; prim.v[1] = (struct vertex_header *)v1; prim.v[2] = (struct vertex_header *)v2; - prim.reset_line_stipple = 1; - prim.edgeflags = ((prim.v[0]->edgeflag) | - (prim.v[1]->edgeflag << 1) | - (prim.v[2]->edgeflag << 2)); + prim.flags = flags; prim.pad = 0; draw->pipeline.first->tri( draw->pipeline.first, &prim ); @@ -197,13 +194,15 @@ void draw_pipeline_run( struct draw_context *draw, case PIPE_PRIM_LINES: for (i = 0; i+1 < count; i += 2) do_line( draw, - verts + stride * elts[i+0], + elts[i+0], + verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK), verts + stride * elts[i+1]); break; case PIPE_PRIM_TRIANGLES: for (i = 0; i+2 < count; i += 3) do_triangle( draw, - verts + stride * elts[i+0], + elts[i+0], + verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK), verts + stride * elts[i+1], verts + stride * elts[i+2]); break; diff --git a/src/gallium/auxiliary/draw/draw_pipe.h b/src/gallium/auxiliary/draw/draw_pipe.h index 2476abb2b2..f1cb0891ca 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.h +++ b/src/gallium/auxiliary/draw/draw_pipe.h @@ -37,6 +37,17 @@ #include "draw_private.h" /* for sizeof(vertex_header) */ +/** + * Basic info for a point/line/triangle primitive. + */ +struct prim_header { + float det; /**< front/back face determinant */ + ushort flags; + ushort pad; + struct vertex_header *v[3]; /**< 1 to 3 vertex pointers */ +}; + + /** * Base class for all primitive drawing stages. diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c index 21216addea..ce80c94163 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_clip.c +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -119,7 +119,7 @@ static void interp( const struct clipper *clip, */ { dst->clipmask = 0; - dst->edgeflag = 0; + dst->edgeflag = 0; /* will get overwritten later */ dst->pad = 0; dst->vertex_id = UNDEFINED_VERTEX_ID; } @@ -162,45 +162,39 @@ static void emit_poly( struct draw_stage *stage, struct prim_header header; unsigned i; + const ushort edge_first = DRAW_PIPE_EDGE_FLAG_2; + const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const ushort edge_last = DRAW_PIPE_EDGE_FLAG_1; + /* later stages may need the determinant, but only the sign matters */ header.det = origPrim->det; + header.flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + header.pad = 0; - for (i = 2; i < n; i++) { + for (i = 2; i < n; i++, header.flags = 0) { header.v[0] = inlist[i-1]; header.v[1] = inlist[i]; header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ - - { - unsigned tmp1 = header.v[1]->edgeflag; - unsigned tmp2 = header.v[2]->edgeflag; - - if (i != n-1) header.v[1]->edgeflag = 0; - if (i != 2) header.v[2]->edgeflag = 0; - - header.edgeflags = ((header.v[0]->edgeflag << 0) | - (header.v[1]->edgeflag << 1) | - (header.v[2]->edgeflag << 2)); - - if (0) { - const struct draw_vertex_shader *vs = stage->draw->vertex_shader; - uint j, k; - debug_printf("Clipped tri:\n"); - for (j = 0; j < 3; j++) { - for (k = 0; k < vs->info.num_outputs; k++) { - debug_printf(" Vert %d: Attr %d: %f %f %f %f\n", j, k, - header.v[j]->data[k][0], - header.v[j]->data[k][1], - header.v[j]->data[k][2], - header.v[j]->data[k][3]); - } + + if (i == n-1) + header.flags |= edge_last; + + if (0) { + const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + uint j, k; + debug_printf("Clipped tri:\n"); + for (j = 0; j < 3; j++) { + for (k = 0; k < vs->info.num_outputs; k++) { + debug_printf(" Vert %d: Attr %d: %f %f %f %f\n", j, k, + header.v[j]->data[k][0], + header.v[j]->data[k][1], + header.v[j]->data[k][2], + header.v[j]->data[k][3]); } } - - stage->next->tri( stage->next, &header ); - - header.v[1]->edgeflag = tmp1; - header.v[2]->edgeflag = tmp2; } + + stage->next->tri( stage->next, &header ); } } diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c index 205000cbea..09b68c4559 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c @@ -91,7 +91,8 @@ static void flatshade_tri_0( struct draw_stage *stage, struct prim_header tmp; tmp.det = header->det; - tmp.edgeflags = header->edgeflags; + tmp.flags = header->flags; + tmp.pad = header->pad; tmp.v[0] = header->v[0]; tmp.v[1] = dup_vert(stage, header->v[1], 0); tmp.v[2] = dup_vert(stage, header->v[2], 1); @@ -108,7 +109,8 @@ static void flatshade_tri_2( struct draw_stage *stage, struct prim_header tmp; tmp.det = header->det; - tmp.edgeflags = header->edgeflags; + tmp.flags = header->flags; + tmp.pad = header->pad; tmp.v[0] = dup_vert(stage, header->v[0], 0); tmp.v[1] = dup_vert(stage, header->v[1], 1); tmp.v[2] = header->v[2]; diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c index ffec85ccdd..ea6de8c571 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_offset.c +++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c @@ -106,7 +106,8 @@ static void offset_tri( struct draw_stage *stage, struct prim_header tmp; tmp.det = header->det; - tmp.edgeflags = header->edgeflags; + tmp.flags = header->flags; + tmp.pad = header->pad; tmp.v[0] = dup_vert(stage, header->v[0], 0); tmp.v[1] = dup_vert(stage, header->v[1], 1); tmp.v[2] = dup_vert(stage, header->v[2], 2); diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c index 9cf5840cce..3cbced362e 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c @@ -135,6 +135,10 @@ stipple_line(struct draw_stage *stage, struct prim_header *header) float length = MAX2(dx, dy); int i; + if (header->flags & DRAW_PIPE_RESET_STIPPLE) + stipple->counter = 0; + + /* XXX ToDo: intead of iterating pixel-by-pixel, use a look-up table. */ for (i = 0; i < length; i++) { diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c index 5910dccc43..50872fdbe9 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_twoside.c +++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c @@ -85,7 +85,8 @@ static void twoside_tri( struct draw_stage *stage, struct prim_header tmp; tmp.det = header->det; - tmp.edgeflags = header->edgeflags; + tmp.flags = header->flags; + tmp.pad = header->pad; /* copy back attribs to front attribs */ tmp.v[0] = copy_bfc(twoside, header->v[0], 0); tmp.v[1] = copy_bfc(twoside, header->v[1], 1); diff --git a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c index eeb2bc43f9..8f97fdedaa 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c +++ b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c @@ -83,9 +83,9 @@ static void points( struct draw_stage *stage, struct vertex_header *v1 = header->v[1]; struct vertex_header *v2 = header->v[2]; - if (header->edgeflags & 0x1) point( stage, v0 ); - if (header->edgeflags & 0x2) point( stage, v1 ); - if (header->edgeflags & 0x4) point( stage, v2 ); + if ((header->flags & DRAW_PIPE_EDGE_FLAG_0) && v0->edgeflag) point( stage, v0 ); + if ((header->flags & DRAW_PIPE_EDGE_FLAG_1) && v1->edgeflag) point( stage, v1 ); + if ((header->flags & DRAW_PIPE_EDGE_FLAG_2) && v2->edgeflag) point( stage, v2 ); } @@ -96,22 +96,22 @@ static void lines( struct draw_stage *stage, struct vertex_header *v1 = header->v[1]; struct vertex_header *v2 = header->v[2]; -#if 0 - assert(((header->edgeflags & 0x1) >> 0) == header->v[0]->edgeflag); - assert(((header->edgeflags & 0x2) >> 1) == header->v[1]->edgeflag); - assert(((header->edgeflags & 0x4) >> 2) == header->v[2]->edgeflag); -#endif + if (header->flags & DRAW_PIPE_RESET_STIPPLE) + stage->next->reset_stipple_counter( stage->next ); - if (header->edgeflags & 0x4) line( stage, v2, v0 ); - if (header->edgeflags & 0x1) line( stage, v0, v1 ); - if (header->edgeflags & 0x2) line( stage, v1, v2 ); + if ((header->flags & DRAW_PIPE_EDGE_FLAG_2) && v2->edgeflag) line( stage, v2, v0 ); + if ((header->flags & DRAW_PIPE_EDGE_FLAG_0) && v0->edgeflag) line( stage, v0, v1 ); + if ((header->flags & DRAW_PIPE_EDGE_FLAG_1) && v1->edgeflag) line( stage, v1, v2 ); } /* Unfilled tri: * * Note edgeflags in the vertex struct is not sufficient as we will - * need to manipulate them when decomposing primitives??? + * need to manipulate them when decomposing primitives. + * + * We currently keep the vertex edgeflag and primitive edgeflag mask + * separate until the last possible moment. */ static void unfilled_tri( struct draw_stage *stage, struct prim_header *header ) diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index ca5021c5c6..49593c8fad 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -78,25 +78,6 @@ struct vertex_header { #define UNDEFINED_VERTEX_ID 0xffff -/** - * Basic info for a point/line/triangle primitive. - */ -struct prim_header { - float det; /**< front/back face determinant */ - unsigned reset_line_stipple:1; - unsigned edgeflags:3; - unsigned pad:28; - struct vertex_header *v[3]; /**< 1 to 3 vertex pointers */ -}; - - - - -#define PT_SHADE 0x1 -#define PT_CLIPTEST 0x2 -#define PT_PIPELINE 0x4 -#define PT_MAX_MIDDLE 0x8 - /** * Private context for the drawing module. */ @@ -238,6 +219,25 @@ void draw_pt_reset_vertex_ids( struct draw_context *draw ); boolean draw_pipeline_init( struct draw_context *draw ); void draw_pipeline_destroy( struct draw_context *draw ); + + + + +/* We use the top few bits in the elts[] parameter to convey a little + * API information. This limits the number of vertices we can address + * to only 4096 -- if that becomes a problem, we can switch to 32-bit + * draw indices. + * + * These flags expected at first vertex of lines & triangles when + * unfilled and/or line stipple modes are operational. + */ +#define DRAW_PIPE_EDGE_FLAG_0 (0x1<<12) +#define DRAW_PIPE_EDGE_FLAG_1 (0x2<<12) +#define DRAW_PIPE_EDGE_FLAG_2 (0x4<<12) +#define DRAW_PIPE_EDGE_FLAG_ALL (0x7<<12) +#define DRAW_PIPE_RESET_STIPPLE (0x8<<12) +#define DRAW_PIPE_FLAG_MASK (0xf<<12) + void draw_pipeline_run( struct draw_context *draw, unsigned prim, struct vertex_header *vertices, @@ -246,6 +246,8 @@ void draw_pipeline_run( struct draw_context *draw, const ushort *elts, unsigned count ); + + void draw_pipeline_flush( struct draw_context *draw, unsigned flags ); diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index fd0d158fcf..df5c29fc36 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -40,15 +40,6 @@ typedef unsigned (*pt_elt_func)( const void *elts, unsigned idx ); struct draw_pt_middle_end; struct draw_context; -/* We use the top couple of bits in the vertex fetch index to convey a - * little API information. This limits the number of vertices we can - * address to only 1 billion -- if that becomes a problem, these could - * be moved out & passed separately. - */ -#define DRAW_PT_EDGEFLAG (1<<30) -#define DRAW_PT_RESET_STIPPLE (1<<31) -#define DRAW_PT_FLAG_MASK (3<<30) - #define PT_SHADE 0x1 #define PT_CLIPTEST 0x2 diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index cc61b113fd..9f74806e11 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -41,6 +41,7 @@ struct pt_fetch { struct translate *translate; unsigned vertex_size; + boolean need_edgeflags; struct translate_cache *cache; }; @@ -121,6 +122,10 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, 0); } } + + fetch->need_edgeflags = ((draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL) && + draw->pt.user.edgeflag); } @@ -147,6 +152,18 @@ void draw_pt_fetch_run( struct pt_fetch *fetch, elts, count, verts ); + + /* Edgeflags are hard to fit into a translate program, populate + * them separately if required. In the setup above they are + * defaulted to one, so only need this if there is reason to change + * that default: + */ + if (fetch->need_edgeflags) { + for (i = 0; i < count; i++) { + struct vertex_header *vh = (struct vertex_header *)(verts + i * fetch->vertex_size); + vh->edgeflag = draw_pt_get_edgeflag( draw, elts[i] ); + } + } } diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 153055417d..b3133359e0 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -33,8 +33,6 @@ #include "pipe/p_util.h" #include "draw/draw_context.h" #include "draw/draw_private.h" -//#include "draw/draw_vbuf.h" -//#include "draw/draw_vertex.h" #include "draw/draw_pt.h" @@ -86,8 +84,9 @@ static void vcache_check_flush( struct vcache_frontend *vcache ) } -static void vcache_elt( struct vcache_frontend *vcache, - unsigned felt ) +static INLINE void vcache_elt( struct vcache_frontend *vcache, + unsigned felt, + ushort flags ) { unsigned idx = felt % CACHE_MAX; @@ -99,28 +98,9 @@ static void vcache_elt( struct vcache_frontend *vcache, vcache->fetch_elts[vcache->fetch_count++] = felt; } - vcache->draw_elts[vcache->draw_count++] = vcache->out[idx]; + vcache->draw_elts[vcache->draw_count++] = vcache->out[idx] | flags; } -static unsigned add_edgeflag( struct vcache_frontend *vcache, - unsigned idx, - unsigned mask ) -{ - if (0 && mask && draw_pt_get_edgeflag(vcache->draw, idx)) - return idx | DRAW_PT_EDGEFLAG; - else - return idx; -} - - -static unsigned add_reset_stipple( unsigned idx, - unsigned reset ) -{ - if (0 && reset) - return idx | DRAW_PT_RESET_STIPPLE; - else - return idx; -} static void vcache_triangle( struct vcache_frontend *vcache, @@ -128,49 +108,42 @@ static void vcache_triangle( struct vcache_frontend *vcache, unsigned i1, unsigned i2 ) { - vcache_elt(vcache, i0 /* | DRAW_PT_EDGEFLAG | DRAW_PT_RESET_STIPPLE */ ); - vcache_elt(vcache, i1 /* | DRAW_PT_EDGEFLAG */); - vcache_elt(vcache, i2 /* | DRAW_PT_EDGEFLAG */); + vcache_elt(vcache, i0, 0); + vcache_elt(vcache, i1, 0); + vcache_elt(vcache, i2, 0); vcache_check_flush(vcache); } -static void vcache_ef_triangle( struct vcache_frontend *vcache, - boolean reset_stipple, - unsigned ef_mask, - unsigned i0, - unsigned i1, - unsigned i2 ) +static void vcache_triangle_flags( struct vcache_frontend *vcache, + ushort flags, + unsigned i0, + unsigned i1, + unsigned i2 ) { -/* - i0 = add_edgeflag( vcache, i0, (ef_mask >> 0) & 1 ); - i1 = add_edgeflag( vcache, i1, (ef_mask >> 1) & 1 ); - i2 = add_edgeflag( vcache, i2, (ef_mask >> 2) & 1 ); - i0 = add_reset_stipple( i0, reset_stipple ); -*/ - - vcache_elt(vcache, i0); - vcache_elt(vcache, i1); - vcache_elt(vcache, i2); + vcache_elt(vcache, i0, flags); + vcache_elt(vcache, i1, 0); + vcache_elt(vcache, i2, 0); vcache_check_flush(vcache); - - if (0) debug_printf("emit tri ef: %d %d %d\n", - !!(i0 & DRAW_PT_EDGEFLAG), - !!(i1 & DRAW_PT_EDGEFLAG), - !!(i2 & DRAW_PT_EDGEFLAG)); - } - static void vcache_line( struct vcache_frontend *vcache, - boolean reset_stipple, unsigned i0, unsigned i1 ) { - i0 = add_reset_stipple( i0, reset_stipple ); + vcache_elt(vcache, i0, 0); + vcache_elt(vcache, i1, 0); + vcache_check_flush(vcache); +} + - vcache_elt(vcache, i0); - vcache_elt(vcache, i1); +static void vcache_line_flags( struct vcache_frontend *vcache, + ushort flags, + unsigned i0, + unsigned i1 ) +{ + vcache_elt(vcache, i0, flags); + vcache_elt(vcache, i1, 0); vcache_check_flush(vcache); } @@ -178,7 +151,7 @@ static void vcache_line( struct vcache_frontend *vcache, static void vcache_point( struct vcache_frontend *vcache, unsigned i0 ) { - vcache_elt(vcache, i0); + vcache_elt(vcache, i0, 0); vcache_check_flush(vcache); } @@ -198,237 +171,36 @@ static void vcache_ef_quad( struct vcache_frontend *vcache, unsigned i2, unsigned i3 ) { - const unsigned omitEdge2 = ~(1 << 1); - const unsigned omitEdge3 = ~(1 << 2); - vcache_ef_triangle( vcache, 1, omitEdge2, i0, i1, i3 ); - vcache_ef_triangle( vcache, 0, omitEdge3, i1, i2, i3 ); -} + const unsigned omitEdge1 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_2; + const unsigned omitEdge2 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_1; + vcache_triangle_flags( vcache, + DRAW_PIPE_RESET_STIPPLE | omitEdge1, + i0, i1, i3 ); + vcache_triangle_flags( vcache, + omitEdge2, + i1, i2, i3 ); +} +/* At least for now, we're back to using a template include file for + * this. The two paths aren't too different though - it may be + * possible to reunify them. + */ +#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle_flags(vc,flags,i0,i1,i2) +#define QUAD(vc,i0,i1,i2,i3) vcache_ef_quad(vc,i0,i1,i2,i3) +#define LINE(vc,flags,i0,i1) vcache_line_flags(vc,flags,i0,i1) +#define POINT(vc,i0) vcache_point(vc,i0) +#define FUNC vcache_run_extras +#include "draw_pt_vcache_tmp.h" + +#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle(vc,i0,i1,i2) +#define QUAD(vc,i0,i1,i2,i3) vcache_quad(vc,i0,i1,i2,i3) +#define LINE(vc,flags,i0,i1) vcache_line(vc,i0,i1) +#define POINT(vc,i0) vcache_point(vc,i0) +#define FUNC vcache_run +#include "draw_pt_vcache_tmp.h" -static void vcache_run( struct draw_pt_front_end *frontend, - pt_elt_func get_elt, - const void *elts, - unsigned count ) -{ - struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; - struct draw_context *draw = vcache->draw; - - boolean unfilled = (draw->rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - draw->rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL); - - boolean flatfirst = (draw->rasterizer->flatshade && - draw->rasterizer->flatshade_first); - unsigned i; - -// debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count ); - - switch (vcache->input_prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i ++) { - vcache_point( vcache, - get_elt(elts, i + 0) ); - } - break; - - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) { - vcache_line( vcache, - TRUE, - get_elt(elts, i + 0), - get_elt(elts, i + 1)); - } - break; - - case PIPE_PRIM_LINE_LOOP: - if (count >= 2) { - for (i = 1; i < count; i++) { - vcache_line( vcache, - i == 1, /* XXX: only if vb not split */ - get_elt(elts, i - 1), - get_elt(elts, i )); - } - - vcache_line( vcache, - 0, - get_elt(elts, count - 1), - get_elt(elts, 0 )); - } - break; - - case PIPE_PRIM_LINE_STRIP: - for (i = 1; i < count; i++) { - vcache_line( vcache, - i == 1, - get_elt(elts, i - 1), - get_elt(elts, i )); - } - break; - - case PIPE_PRIM_TRIANGLES: - if (unfilled) { - for (i = 0; i+2 < count; i += 3) { - vcache_ef_triangle( vcache, - 1, - ~0, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2 )); - } - } - else { - for (i = 0; i+2 < count; i += 3) { - vcache_triangle( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2 )); - } - } - break; - - case PIPE_PRIM_TRIANGLE_STRIP: - if (flatfirst) { - for (i = 0; i+2 < count; i++) { - if (i & 1) { - vcache_triangle( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 2), - get_elt(elts, i + 1 )); - } - else { - vcache_triangle( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2 )); - } - } - } - else { - for (i = 0; i+2 < count; i++) { - if (i & 1) { - vcache_triangle( vcache, - get_elt(elts, i + 1), - get_elt(elts, i + 0), - get_elt(elts, i + 2 )); - } - else { - vcache_triangle( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2 )); - } - } - } - break; - - case PIPE_PRIM_TRIANGLE_FAN: - if (count >= 3) { - if (flatfirst) { - for (i = 0; i+2 < count; i++) { - vcache_triangle( vcache, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0 )); - } - } - else { - for (i = 0; i+2 < count; i++) { - vcache_triangle( vcache, - get_elt(elts, 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2 )); - } - } - } - break; - - - case PIPE_PRIM_QUADS: - if (unfilled) { - for (i = 0; i+3 < count; i += 4) { - vcache_ef_quad( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, i + 3)); - } - } - else { - for (i = 0; i+3 < count; i += 4) { - vcache_quad( vcache, - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, i + 3)); - } - } - break; - - case PIPE_PRIM_QUAD_STRIP: - if (unfilled) { - for (i = 0; i+3 < count; i += 2) { - vcache_ef_quad( vcache, - get_elt(elts, i + 2), - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 3)); - } - } - else { - for (i = 0; i+3 < count; i += 2) { - vcache_quad( vcache, - get_elt(elts, i + 2), - get_elt(elts, i + 0), - get_elt(elts, i + 1), - get_elt(elts, i + 3)); - } - } - break; - - case PIPE_PRIM_POLYGON: - if (unfilled) { - /* These bitflags look a little odd because we submit the - * vertices as (1,2,0) to satisfy flatshade requirements. - */ - const unsigned edge_first = (1<<2); - const unsigned edge_middle = (1<<0); - const unsigned edge_last = (1<<1); - - for (i = 0; i+2 < count; i++) { - unsigned ef_mask = edge_middle; - - if (i == 0) - ef_mask |= edge_first; - - if (i + 3 == count) - ef_mask |= edge_last; - - vcache_ef_triangle( vcache, - i == 0, - ef_mask, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0)); - } - } - else { - for (i = 0; i+2 < count; i++) { - vcache_triangle( vcache, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0)); - } - } - break; - - default: - assert(0); - break; - } - - vcache_flush( vcache ); -} @@ -453,15 +225,21 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, unsigned opt ) { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; + const struct pipe_rasterizer_state *rasterizer = vcache->draw->rasterizer; + -/* - if (vcache->draw->rasterizer->flatshade_first) - vcache->base.run = vcache_run_pv0; - else - vcache->base.run = vcache_run_pv2; -*/ + if (rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL || + rasterizer->line_stipple_enable) + { + assert(opt & PT_PIPELINE); + vcache->base.run = vcache_run_extras; + } + else + { + vcache->base.run = vcache_run; + } - vcache->base.run = vcache_run; vcache->input_prim = prim; vcache->output_prim = reduced_prim[prim]; diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h new file mode 100644 index 0000000000..cf9f394aa3 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h @@ -0,0 +1,174 @@ + + +static void FUNC( struct draw_pt_front_end *frontend, + pt_elt_func get_elt, + const void *elts, + unsigned count ) +{ + struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; + struct draw_context *draw = vcache->draw; + + boolean flatfirst = (draw->rasterizer->flatshade && + draw->rasterizer->flatshade_first); + unsigned i, flags; + + + switch (vcache->input_prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i ++) { + POINT( vcache, + get_elt(elts, i + 0) ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) { + LINE( vcache, + DRAW_PIPE_RESET_STIPPLE, + get_elt(elts, i + 0), + get_elt(elts, i + 1)); + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + flags = DRAW_PIPE_RESET_STIPPLE; + + for (i = 1; i < count; i++, flags = 0) { + LINE( vcache, + flags, + get_elt(elts, i - 1), + get_elt(elts, i )); + } + + LINE( vcache, + flags, + get_elt(elts, i - 1), + get_elt(elts, 0 )); + } + break; + + case PIPE_PRIM_LINE_STRIP: + flags = DRAW_PIPE_RESET_STIPPLE; + for (i = 1; i < count; i++, flags = 0) { + LINE( vcache, + flags, + get_elt(elts, i - 1), + get_elt(elts, i )); + } + break; + + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2 )); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, i + 0), + get_elt(elts, i + 1 + (i&1)), + get_elt(elts, i + 2 - (i&1))); + } + } + else { + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, i + 0 + (i&1)), + get_elt(elts, i + 1 - (i&1)), + get_elt(elts, i + 2 )); + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + if (count >= 3) { + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0 )); + } + } + else { + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2 )); + } + } + } + break; + + + case PIPE_PRIM_QUADS: + for (i = 0; i+3 < count; i += 4) { + QUAD( vcache, + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, i + 3)); + } + break; + + case PIPE_PRIM_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + QUAD( vcache, + get_elt(elts, i + 2), + get_elt(elts, i + 0), + get_elt(elts, i + 1), + get_elt(elts, i + 3)); + } + break; + + case PIPE_PRIM_POLYGON: + { + /* These bitflags look a little odd because we submit the + * vertices as (1,2,0) to satisfy flatshade requirements. + */ + const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; + const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + + for (i = 0; i+2 < count; i++, flags = edge_middle) { + + if (i + 3 == count) + flags |= edge_last; + + TRIANGLE( vcache, + flags, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0)); + } + } + break; + + default: + assert(0); + break; + } + + vcache_flush( vcache ); +} + + +#undef TRIANGLE +#undef QUAD +#undef POINT +#undef LINE +#undef FUNC -- cgit v1.2.3 From d712eea074303812b1be0e79b302d7b9f49a09a8 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 24 Apr 2008 14:05:39 +0200 Subject: translate: Actually return a value from translate_cache_create(). --- src/gallium/auxiliary/translate/translate_cache.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_cache.c b/src/gallium/auxiliary/translate/translate_cache.c index c14f37c42f..e91d648eba 100644 --- a/src/gallium/auxiliary/translate/translate_cache.c +++ b/src/gallium/auxiliary/translate/translate_cache.c @@ -40,6 +40,7 @@ struct translate_cache * translate_cache_create() { struct translate_cache *cache = MALLOC_STRUCT(translate_cache); cache->hash = cso_hash_create(); + return cache; } -- cgit v1.2.3 From cefa367b5d909f26d943101efb042562de778c5f Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 24 Apr 2008 17:07:45 +0200 Subject: i915: Fix for wrong texture in texobj with VBUF --- src/gallium/drivers/i915simple/i915_state.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 24d31ad740..3d94b52366 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -549,6 +549,9 @@ static void i915_set_sampler_textures(struct pipe_context *pipe, !memcmp(i915->texture, texture, num * sizeof(struct pipe_texture *))) return; + /* Fixes wrong texture in texobj with VBUF */ + draw_flush(i915->draw); + for (i = 0; i < num; i++) pipe_texture_reference((struct pipe_texture **) &i915->texture[i], texture[i]); -- cgit v1.2.3 From 35dc003c6a0845a96934f9c222bd57bb62e0a62f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 24 Apr 2008 18:39:25 +0100 Subject: gallium: Make sure the size of non-existent storage components is set to 0. Before adding support for formats with unused storage components, the size of components swizzled to 0 or 1 was ignored, so this didn't matter. --- src/gallium/include/pipe/p_format.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 2ee6f100bf..94c2eeb62b 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -238,8 +238,8 @@ enum pipe_format { PIPE_FORMAT_A1R5G5B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_A4R4G4B4_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB1, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_U_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ - PIPE_FORMAT_U_A8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ + PIPE_FORMAT_U_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ + PIPE_FORMAT_U_A8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 0, 0, 0, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ PIPE_FORMAT_U_I8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ PIPE_FORMAT_U_A8_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha, luminance */ PIPE_FORMAT_YCBCR = _PIPE_FORMAT_YCBCR( 0 ), @@ -313,7 +313,7 @@ enum pipe_format { PIPE_FORMAT_R8G8B8A8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R8G8B8X8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ), /* sRGB formats */ - PIPE_FORMAT_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), + PIPE_FORMAT_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ), PIPE_FORMAT_A8_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), PIPE_FORMAT_R8G8B8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ), PIPE_FORMAT_R8G8B8A8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), -- cgit v1.2.3 From f2c31257167f85df276322be1b8523064e8b66a9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 11:52:37 -0600 Subject: gallium: added cso_delete_vertex_fragment_shader() functions The state tracker now uses these functions to free shaders, rather than the pipe->delete_vs/fs-state() functions. Before, we could get in a situation where we free() a shader and happen to alloc() a new one at the same address. The cso_set_vertex/fragment_shader() function would no-op the state change since the pointers were the same. This led to problems elsewhere, of course. The new delete functions null-out the CSO's current shader pointers. --- src/gallium/auxiliary/cso_cache/cso_context.c | 17 +++++++++++++++++ src/gallium/auxiliary/cso_cache/cso_context.h | 6 ++++++ 2 files changed, 23 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index d246dff433..cfb91d31de 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -471,6 +471,8 @@ void cso_restore_rasterizer(struct cso_context *ctx) ctx->rasterizer_saved = NULL; } + + enum pipe_error cso_set_fragment_shader_handle(struct cso_context *ctx, void *handle ) { @@ -481,6 +483,13 @@ enum pipe_error cso_set_fragment_shader_handle(struct cso_context *ctx, return PIPE_OK; } +void cso_delete_fragment_shader(struct cso_context *ctx, void *handle ) +{ + if (handle == ctx->fragment_shader) + ctx->pipe->bind_fs_state(ctx->pipe, NULL); + ctx->pipe->delete_fs_state(ctx->pipe, handle); + ctx->fragment_shader = NULL; +} /* Not really working: */ @@ -553,6 +562,14 @@ enum pipe_error cso_set_vertex_shader_handle(struct cso_context *ctx, return PIPE_OK; } +void cso_delete_vertex_shader(struct cso_context *ctx, void *handle ) +{ + if (handle == ctx->vertex_shader) + ctx->pipe->bind_vs_state(ctx->pipe, NULL); + ctx->pipe->delete_vs_state(ctx->pipe, handle); + ctx->vertex_shader = NULL; +} + /* Not really working: */ diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h index 0405944132..cb46f71d51 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.h +++ b/src/gallium/auxiliary/cso_cache/cso_context.h @@ -99,16 +99,22 @@ void cso_restore_sampler_textures( struct cso_context *cso ); */ enum pipe_error cso_set_fragment_shader_handle(struct cso_context *ctx, void *handle ); +void cso_delete_fragment_shader(struct cso_context *ctx, void *handle ); +/* enum pipe_error cso_set_fragment_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); +*/ void cso_save_fragment_shader(struct cso_context *cso); void cso_restore_fragment_shader(struct cso_context *cso); enum pipe_error cso_set_vertex_shader_handle(struct cso_context *ctx, void *handle ); +void cso_delete_vertex_shader(struct cso_context *ctx, void *handle ); +/* enum pipe_error cso_set_vertex_shader( struct cso_context *cso, const struct pipe_shader_state *shader ); +*/ void cso_save_vertex_shader(struct cso_context *cso); void cso_restore_vertex_shader(struct cso_context *cso); -- cgit v1.2.3 From ac79532a15a7109bf0fbd0e40a1ba8e65ed8c435 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 11:56:06 -0600 Subject: gallium: tweak the new shader delete funcs --- src/gallium/auxiliary/cso_cache/cso_context.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index cfb91d31de..febecbbbde 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -485,10 +485,12 @@ enum pipe_error cso_set_fragment_shader_handle(struct cso_context *ctx, void cso_delete_fragment_shader(struct cso_context *ctx, void *handle ) { - if (handle == ctx->fragment_shader) + if (handle == ctx->fragment_shader) { + /* unbind before deleting */ ctx->pipe->bind_fs_state(ctx->pipe, NULL); + ctx->fragment_shader = NULL; + } ctx->pipe->delete_fs_state(ctx->pipe, handle); - ctx->fragment_shader = NULL; } /* Not really working: @@ -564,10 +566,12 @@ enum pipe_error cso_set_vertex_shader_handle(struct cso_context *ctx, void cso_delete_vertex_shader(struct cso_context *ctx, void *handle ) { - if (handle == ctx->vertex_shader) + if (handle == ctx->vertex_shader) { + /* unbind before deleting */ ctx->pipe->bind_vs_state(ctx->pipe, NULL); + ctx->vertex_shader = NULL; + } ctx->pipe->delete_vs_state(ctx->pipe, handle); - ctx->vertex_shader = NULL; } -- cgit v1.2.3 From a41804909d5799cddfbf48a46524f78c736408d4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 11:59:22 -0600 Subject: gallium: minor clean-ups, comments --- src/gallium/auxiliary/cso_cache/cso_context.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index febecbbbde..8cf2313b7f 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -102,6 +102,10 @@ out: return NULL; } + +/** + * Prior to context destruction, this function unbinds all state objects. + */ static void cso_release_all( struct cso_context *ctx ) { unsigned i; @@ -115,10 +119,10 @@ static void cso_release_all( struct cso_context *ctx ) ctx->pipe->bind_vs_state( ctx->pipe, NULL ); } - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { pipe_texture_reference(&ctx->textures[i], NULL); - for (i = 0; i < PIPE_MAX_SAMPLERS; i++) pipe_texture_reference(&ctx->textures_saved[i], NULL); + } if (ctx->cache) { cso_cache_delete( ctx->cache ); @@ -129,10 +133,10 @@ static void cso_release_all( struct cso_context *ctx ) void cso_destroy_context( struct cso_context *ctx ) { - if (ctx) + if (ctx) { cso_release_all( ctx ); - - FREE( ctx ); + FREE( ctx ); + } } -- cgit v1.2.3 From 386102c62a3315182ffbc6319351cb883234511a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 12:10:14 -0600 Subject: gallium: make cso_release_all() public --- src/gallium/auxiliary/cso_cache/cso_context.c | 4 ++-- src/gallium/auxiliary/cso_cache/cso_context.h | 2 ++ 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 8cf2313b7f..b4609e999b 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -106,7 +106,7 @@ out: /** * Prior to context destruction, this function unbinds all state objects. */ -static void cso_release_all( struct cso_context *ctx ) +void cso_release_all( struct cso_context *ctx ) { unsigned i; @@ -134,7 +134,7 @@ static void cso_release_all( struct cso_context *ctx ) void cso_destroy_context( struct cso_context *ctx ) { if (ctx) { - cso_release_all( ctx ); + //cso_release_all( ctx ); FREE( ctx ); } } diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h index cb46f71d51..b04e98bfa1 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.h +++ b/src/gallium/auxiliary/cso_cache/cso_context.h @@ -42,6 +42,8 @@ struct cso_context; struct cso_context *cso_create_context( struct pipe_context *pipe ); +void cso_release_all( struct cso_context *ctx ); + void cso_destroy_context( struct cso_context *cso ); -- cgit v1.2.3 From a770d40c3d4977e2c134661b5d8facaea446b6ea Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 12:11:24 -0600 Subject: gallium: use cso_destroy_vertex/fragment_shader() functions Also, rearrange the st_destroy_context() code a bit to prevent some invalid/NULL ptr derefs during tear-down. --- src/mesa/state_tracker/st_cb_bitmap.c | 2 +- src/mesa/state_tracker/st_cb_clear.c | 4 ++-- src/mesa/state_tracker/st_cb_program.c | 12 +++++------- src/mesa/state_tracker/st_context.c | 11 +++++++++-- src/mesa/state_tracker/st_program.c | 2 +- 5 files changed, 18 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 7752b40e8b..e7f6ff66b5 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -803,7 +803,7 @@ st_destroy_bitmap(struct st_context *st) } #endif if (st->bitmap.vs) { - pipe->delete_vs_state(pipe, st->bitmap.vs); + cso_delete_vertex_shader(st->cso_context, st->bitmap.vs); st->bitmap.vs = NULL; } diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index dac346a06c..95a5fb8db4 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -98,11 +98,11 @@ st_destroy_clear(struct st_context *st) struct pipe_context *pipe = st->pipe; if (st->clear.fs) { - pipe->delete_fs_state(pipe, st->clear.fs); + cso_delete_fragment_shader(st->cso_context, st->clear.fs); st->clear.fs = NULL; } if (st->clear.vs) { - pipe->delete_vs_state(pipe, st->clear.vs); + cso_delete_vertex_shader(st->cso_context, st->clear.vs); st->clear.vs = NULL; } if (st->clear.vbuf) { diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 67f8b1f8eb..a293ec3f0b 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -39,7 +39,7 @@ #include "shader/programopt.h" #include "shader/shader_api.h" -#include "cso_cache/cso_cache.h" +#include "cso_cache/cso_context.h" #include "draw/draw_context.h" #include "st_context.h" @@ -127,7 +127,6 @@ void st_delete_program(GLcontext *ctx, struct gl_program *prog) { struct st_context *st = st_context(ctx); - struct pipe_context *pipe = st->pipe; switch( prog->Target ) { case GL_VERTEX_PROGRAM_ARB: @@ -135,7 +134,7 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) struct st_vertex_program *stvp = (struct st_vertex_program *) prog; if (stvp->driver_shader) { - pipe->delete_vs_state(pipe, stvp->driver_shader); + cso_delete_vertex_shader(st->cso_context, stvp->driver_shader); stvp->driver_shader = NULL; } @@ -150,7 +149,7 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) struct st_fragment_program *stfp = (struct st_fragment_program *) prog; if (stfp->driver_shader) { - pipe->delete_fs_state(pipe, stfp->driver_shader); + cso_delete_fragment_shader(st->cso_context, stfp->driver_shader); stfp->driver_shader = NULL; } @@ -187,7 +186,6 @@ static void st_program_string_notify( GLcontext *ctx, struct gl_program *prog ) { struct st_context *st = st_context(ctx); - struct pipe_context *pipe = st->pipe; if (target == GL_FRAGMENT_PROGRAM_ARB) { struct st_fragment_program *stfp = (struct st_fragment_program *) prog; @@ -195,7 +193,7 @@ static void st_program_string_notify( GLcontext *ctx, stfp->serialNo++; if (stfp->driver_shader) { - pipe->delete_fs_state(pipe, stfp->driver_shader); + cso_delete_fragment_shader(st->cso_context, stfp->driver_shader); stfp->driver_shader = NULL; } @@ -215,7 +213,7 @@ static void st_program_string_notify( GLcontext *ctx, stvp->serialNo++; if (stvp->driver_shader) { - pipe->delete_vs_state(pipe, stvp->driver_shader); + cso_delete_vertex_shader(st->cso_context, stvp->driver_shader); stvp->driver_shader = NULL; } diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index dadc524b51..8a30871fa0 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -176,22 +176,29 @@ static void st_destroy_context_priv( struct st_context *st ) } } - st->pipe->destroy( st->pipe ); free( st ); } void st_destroy_context( struct st_context *st ) { + struct pipe_context *pipe = st->pipe; + struct cso_context *cso = st->cso_context; GLcontext *ctx = st->ctx; /* need to unbind and destroy CSO objects before anything else */ - cso_destroy_context(st->cso_context); + cso_release_all(st->cso_context); _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache); _mesa_free_context_data(ctx); + st_destroy_context_priv(st); + + cso_destroy_context(cso); + + pipe->destroy( pipe ); + free(ctx); } diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 3648ded8a1..d450c30694 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -270,7 +270,7 @@ st_translate_vertex_program(struct st_context *st, stvp->state.tokens = NULL; } if (stvp->driver_shader) { - pipe->delete_vs_state(pipe, stvp->driver_shader); + cso_delete_vertex_shader(st->cso_context, stvp->driver_shader); stvp->driver_shader = NULL; } -- cgit v1.2.3 From 72c8d2f2449d54005eb721fe3853a6009e9b8d17 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 13:36:26 -0600 Subject: mesa: adjust glBitmap coords by a small epsilon Fixes problem with bitmaps jumping around by one pixel depending on window size. The rasterpos is often X.9999 instead of X+1. Run progs/redbook/drawf and resize window to check. --- src/mesa/main/drawpix.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c index 0f64f1c1c0..fa422bb3c7 100644 --- a/src/mesa/main/drawpix.c +++ b/src/mesa/main/drawpix.c @@ -377,8 +377,9 @@ _mesa_Bitmap( GLsizei width, GLsizei height, if (ctx->RenderMode == GL_RENDER) { /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */ - GLint x = IFLOOR(ctx->Current.RasterPos[0] - xorig); - GLint y = IFLOOR(ctx->Current.RasterPos[1] - yorig); + const GLfloat epsilon = 0.0001; + GLint x = IFLOOR(ctx->Current.RasterPos[0] + epsilon - xorig); + GLint y = IFLOOR(ctx->Current.RasterPos[1] + epsilon - yorig); if (ctx->Unpack.BufferObj->Name) { /* unpack from PBO */ -- cgit v1.2.3 From bceebffc178af89154b3b78f3afd97f0d93ca2f8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 13:44:40 -0600 Subject: gallium: disable the bitmap/texcoord bias Doesn't seem to be needed after fixing the float->int conversion problem. --- src/mesa/state_tracker/st_cb_bitmap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index e7f6ff66b5..507ace27c7 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -355,7 +355,7 @@ setup_bitmap_vertex_data(struct st_context *st, const GLfloat x1 = x + width; const GLfloat y0 = y; const GLfloat y1 = y + height; - const GLfloat bias = st->bitmap_texcoord_bias; + const GLfloat bias = 0.0 * st->bitmap_texcoord_bias; /* XXX remove */ const GLfloat xBias = bias / (x1-x0); const GLfloat yBias = bias / (y1-y0); const GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias; -- cgit v1.2.3 From da8312a1cf73d0777d51c63148ee090a9acace8b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 24 Apr 2008 21:13:56 +0100 Subject: draw: default edgeflag should be one --- src/gallium/auxiliary/draw/draw_pt_fetch.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 9f74806e11..1f765b73ad 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -115,7 +115,7 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, fetch->translate = translate_cache_find(fetch->cache, &key); { - static struct vertex_header vh = { 0, 0, 0, 0xffff }; + static struct vertex_header vh = { 0, 1, 0, 0xffff }; fetch->translate->set_buffer(fetch->translate, draw->pt.nr_vertex_buffers, &vh, -- cgit v1.2.3 From 909894e34ca5e575ce21005e38dc0b5e98e4bcd6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 15:58:46 -0600 Subject: gallium: comments --- src/gallium/auxiliary/draw/draw_pipe.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 1c9d8650e2..46afb0f41f 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -194,14 +194,14 @@ void draw_pipeline_run( struct draw_context *draw, case PIPE_PRIM_LINES: for (i = 0; i+1 < count; i += 2) do_line( draw, - elts[i+0], + elts[i+0], /* flags */ verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK), verts + stride * elts[i+1]); break; case PIPE_PRIM_TRIANGLES: for (i = 0; i+2 < count; i += 3) do_triangle( draw, - elts[i+0], + elts[i+0], /* flags */ verts + stride * (elts[i+0] & ~DRAW_PIPE_FLAG_MASK), verts + stride * elts[i+1], verts + stride * elts[i+2]); -- cgit v1.2.3 From 98165318621cd3e01919b2b9ff140ce7a4e12beb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 15:59:08 -0600 Subject: gallium: plug in softpipe_set_edgeflags() function --- src/gallium/drivers/softpipe/sp_context.c | 2 ++ src/gallium/drivers/softpipe/sp_draw_arrays.c | 9 +++++++++ src/gallium/drivers/softpipe/sp_state.h | 3 +++ 3 files changed, 14 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 200fb415ac..f2ce0fdc6e 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -178,6 +178,8 @@ softpipe_create( struct pipe_screen *screen, softpipe->pipe.draw_arrays = softpipe_draw_arrays; softpipe->pipe.draw_elements = softpipe_draw_elements; + softpipe->pipe.set_edgeflags = softpipe_set_edgeflags; + softpipe->pipe.clear = softpipe_clear; softpipe->pipe.flush = softpipe_flush; diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 778291dded..6c58f9909d 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -170,3 +170,12 @@ softpipe_draw_elements(struct pipe_context *pipe, return TRUE; } + + +void +softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags) +{ + struct softpipe_context *sp = softpipe_context(pipe); + draw_set_edgeflags(sp->draw, edgeflags); +} + diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 6e6501f5bc..45056502b8 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -171,6 +171,9 @@ boolean softpipe_draw_elements(struct pipe_context *pipe, unsigned indexSize, unsigned mode, unsigned start, unsigned count); +void +softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags); + void softpipe_map_surfaces(struct softpipe_context *sp); -- cgit v1.2.3 From ef683014dd72612e6eb245e89e754b033acb3a49 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 16:01:18 -0600 Subject: gallium: initial edgeflags code --- src/mesa/state_tracker/st_draw.c | 60 +++++++++++++++++++++++++++++++++++++++- 1 file changed, 59 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index befcb96bd8..097166f444 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -190,6 +190,57 @@ pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized) } +/* + * If edge flags are needed, setup an bitvector of flags and call + * pipe->set_edgeflags(). + * XXX memleak: need to free the returned pointer at some point + */ +static void * +setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count, + const struct gl_client_array *array) +{ + struct pipe_context *pipe = ctx->st->pipe; + + if ((primMode == GL_TRIANGLES || + primMode == GL_QUADS || + primMode == GL_POLYGON) && + (ctx->Polygon.FrontMode != GL_FILL || + ctx->Polygon.BackMode != GL_FILL)) { + /* need edge flags */ + GLuint i; + unsigned *vec = (unsigned *) calloc(sizeof(unsigned), (count + 31) / 32); + struct st_buffer_object *stobj = st_buffer_object(array->BufferObj); + ubyte *map; + + if (!vec) + return NULL; + + map = pipe->winsys->buffer_map(pipe->winsys, stobj->buffer, + PIPE_BUFFER_USAGE_CPU_READ); + map = ADD_POINTERS(map, array->Ptr); + + for (i = 0; i < count; i++) { + if (*((float *) map)) + vec[i/32] |= 1 << (i % 32); + + map += array->StrideB; + } + + pipe->winsys->buffer_unmap(pipe->winsys, stobj->buffer); + + pipe->set_edgeflags(pipe, vec); + + return vec; + } + else { + /* edge flags not needed */ + pipe->set_edgeflags(pipe, NULL); + return NULL; + } +} + + + /** * This function gets plugged into the VBO module and is called when * we have something to render. @@ -277,7 +328,6 @@ st_draw_vbo(GLcontext *ctx, pipe->set_vertex_buffers(pipe, vp->num_inputs, vbuffer); pipe->set_vertex_elements(pipe, vp->num_inputs, velements); - /* do actual drawing */ if (ib) { /* indexed primitive */ @@ -317,6 +367,10 @@ st_draw_vbo(GLcontext *ctx, /* draw */ for (i = 0; i < nr_prims; i++) { + setup_edgeflags(ctx, prims[i].mode, + prims[i].start + indexOffset, prims[i].count, + arrays[VERT_ATTRIB_EDGEFLAG]); + pipe->draw_elements(pipe, indexBuf, indexSize, prims[i].mode, prims[i].start + indexOffset, prims[i].count); @@ -328,6 +382,10 @@ st_draw_vbo(GLcontext *ctx, /* non-indexed */ GLuint i; for (i = 0; i < nr_prims; i++) { + setup_edgeflags(ctx, prims[i].mode, + prims[i].start, prims[i].count, + arrays[VERT_ATTRIB_EDGEFLAG]); + pipe->draw_arrays(pipe, prims[i].mode, prims[i].start, prims[i].count); } } -- cgit v1.2.3 From 2926e59e4ad604dedcb639b2961937841afcf005 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 24 Apr 2008 23:31:35 +0100 Subject: draw: remove old assignment of edgeflag value --- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index 65cb8ba701..c4a67c8289 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -112,7 +112,6 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs, out->clip[3] = out->data[0][3]; out->vertex_id = 0xffff; - out->edgeflag = 1; out->clipmask = compute_clipmask_gl(out->clip, pvs->draw->plane, pvs->draw->nr_planes); -- cgit v1.2.3 From b4c7a48d5c9ed2f9535a17d6c05cd55178c7880a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Apr 2008 16:40:08 -0600 Subject: gallium: fix an edgeflags crash --- src/mesa/state_tracker/st_draw.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 097166f444..e40eeddff5 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -208,10 +208,14 @@ setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count, ctx->Polygon.BackMode != GL_FILL)) { /* need edge flags */ GLuint i; - unsigned *vec = (unsigned *) calloc(sizeof(unsigned), (count + 31) / 32); + unsigned *vec; struct st_buffer_object *stobj = st_buffer_object(array->BufferObj); ubyte *map; + if (!stobj) + return NULL; + + vec = (unsigned *) calloc(sizeof(unsigned), (count + 31) / 32); if (!vec) return NULL; -- cgit v1.2.3 From dacfef158943665fc0d11035867d14ff1f5db332 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 25 Apr 2008 18:18:48 +0900 Subject: gallium: New configuration header. To abstract all those weird #ifdef (__???__) checks. It should typically be the first included header. --- src/gallium/include/pipe/p_compiler.h | 2 + src/gallium/include/pipe/p_config.h | 112 ++++++++++++++++++++++++++++++++++ 2 files changed, 114 insertions(+) create mode 100644 src/gallium/include/pipe/p_config.h (limited to 'src') diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index 91f3d2ac2d..a4b772bc4f 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -29,6 +29,8 @@ #define P_COMPILER_H +#include "p_config.h" + #include #include diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h new file mode 100644 index 0000000000..e44fafeae9 --- /dev/null +++ b/src/gallium/include/pipe/p_config.h @@ -0,0 +1,112 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Gallium configuration defines. + * + * This header file sets several defines based on the compiler, processor + * architecture, and operating system being used. These defines should be used + * throughout the code to facilitate porting to new platforms. It is likely that + * this file is auto-generated by an autoconf-like tool at some point, as some + * things cannot be determined by existing defines alone. + * + * @author José Fonseca + */ + +#ifndef P_CONFIG_H_ +#define P_CONFIG_H_ + + +/* + * Compiler + */ + +#if defined(__GNUC__) +#define PIPE_CC_GCC +#endif + +#if defined(_MSC_VER) || defined(__MSC__) +#define PIPE_CC_MSVC +#endif + +#if defined(__ICL) +#define PIPE_CC_ICL +#endif + + +/* + * Processor architecture + */ + +#if defined(_X86_) || defined(__i386__) || defined(__386__) || defined(i386) +#define PIPE_ARCH_X86 +#endif + +#if 0 /* FIXME */ +#define PIPE_ARCH_X86_64 +#endif + +#if 0 /* FIXME */ +#define PIPE_ARCH_PPC +#endif + + +/* + * Operating system + */ + +#if defined(__linux__) +#define PIPE_OS_LINUX +#endif + +#if defined(_WIN32) || defined(WIN32) +#define PIPE_OS_WINDOWS +#endif + + +/* + * Subsystem + * + * XXX: There is no way to autodetect this. + */ + +#if defined(PIPE_OS_LINUX) +#define PIPE_SUBSYSTEM_DRI +#endif + +#if defined(PIPE_OS_WINDOWS) +#if !defined(PIPE_SUBSYSTEM_USER) && !defined(PIPE_SUBSYSTEM_KERNEL) +#error Neither PIPE_SUBSYSTEM_USER or PIPE_SUBSYSTEM_KERNEL defined. +#endif +#if defined(PIPE_SUBSYSTEM_KERNEL) +#define PIPE_SUBSYSTEM_WINDOWS_DISPLAY +#endif +#endif + + +#endif /* P_CONFIG_H_ */ -- cgit v1.2.3 From b06cd4debfc4fb4162b4b45e61ea91948de0a279 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 25 Apr 2008 18:19:51 +0900 Subject: gallium: Windows user mode portability fixes. --- .../auxiliary/pipebuffer/pb_buffer_fenced.c | 11 ++++--- src/gallium/auxiliary/util/p_debug.c | 14 ++++---- src/gallium/auxiliary/util/p_debug_mem.c | 6 ++-- src/gallium/auxiliary/util/u_time.c | 38 ++++++++++++++-------- src/gallium/auxiliary/util/u_time.h | 8 +++-- src/gallium/include/pipe/p_util.h | 9 ++--- 6 files changed, 53 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 27032b0c4c..d3c1ec4fbe 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -34,6 +34,12 @@ */ +#include "pipe/p_config.h" + +#if defined(PIPE_OS_LINUX) +#include +#endif + #include "pipe/p_compiler.h" #include "pipe/p_error.h" #include "pipe/p_debug.h" @@ -45,9 +51,6 @@ #include "pb_buffer.h" #include "pb_buffer_fenced.h" -#ifndef WIN32 -#include -#endif /** @@ -425,7 +428,7 @@ fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list) /* Wait on outstanding fences */ while (fenced_list->numDelayed) { _glthread_UNLOCK_MUTEX(fenced_list->mutex); -#ifndef WIN32 +#if defined(PIPE_OS_LINUX) sched_yield(); #endif _fenced_buffer_list_check_free(fenced_list, 1); diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 25b132b40c..cd612e23b3 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -26,9 +26,11 @@ **************************************************************************/ +#include "pipe/p_config.h" + #include -#ifdef WIN32 +#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY #include #include #else @@ -42,7 +44,7 @@ #include "util/u_string.h" -#ifdef WIN32 +#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY static INLINE void _EngDebugPrint(const char *format, ...) { @@ -56,7 +58,7 @@ _EngDebugPrint(const char *format, ...) void _debug_vprintf(const char *format, va_list ap) { -#ifdef WIN32 +#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY #ifndef WINCE /* EngDebugPrint does not handle float point arguments, so we need to use * our own vsnprintf implementation. It is also very slow, so buffer until @@ -101,7 +103,7 @@ void _debug_break(void) __asm("int3"); #elif (defined(__i386__) || defined(__386__)) && defined(__MSC__) _asm {int 3}; -#elif defined(WIN32) && !defined(WINCE) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && !defined(WINCE) EngDebugBreak(); #else abort(); @@ -109,7 +111,7 @@ void _debug_break(void) } -#ifdef WIN32 +#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY static const char * find(const char *start, const char *end, char c) { @@ -150,7 +152,7 @@ const char * debug_get_option(const char *name, const char *dfault) { const char *result; -#ifdef WIN32 +#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY ULONG_PTR iFile = 0; const void *pMap = NULL; const char *sol, *eol, *sep; diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c index aba69c0294..9321cf71bb 100644 --- a/src/gallium/auxiliary/util/p_debug_mem.c +++ b/src/gallium/auxiliary/util/p_debug_mem.c @@ -32,7 +32,9 @@ * @author José Fonseca */ -#ifdef WIN32 +#include "pipe/p_config.h" + +#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY #include #include #else @@ -47,7 +49,7 @@ #define DEBUG_MEMORY_MAGIC 0x6e34090aU -#if defined(WIN32) && !defined(WINCE) +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && !defined(WINCE) #define real_malloc(_size) EngAllocMem(0, _size, 'D3AG') #define real_free(_ptr) EngFreeMem(_ptr) #else diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index 01112ebe5a..dd28ff4134 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -33,27 +33,35 @@ */ -#ifndef WIN32 +#include "util/u_time.h" + +#if defined(PIPE_OS_LINUX) #include -#else +#elif defined(PIPE_OS_WINDOWS) #include +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) #include #endif - -#include "util/u_time.h" +#else +#error Unsupported OS +#endif -#ifdef WIN32 +#if defined(PIPE_OS_WINDOWS) static LONGLONG frequency = 0; +#if !defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) +#define EngQueryPerformanceFrequency(p) QueryPerformanceFrequency((LARGE_INTEGER*)(p)) +#define EngQueryPerformanceCounter(p) QueryPerformanceCounter((LARGE_INTEGER*)(p)) +#endif #endif void util_time_get(struct util_time *t) { -#ifndef WIN32 +#if defined(PIPE_OS_LINUX) gettimeofday(&t->tv, NULL); -#else +#elif defined(PIPE_OS_WINDOWS) EngQueryPerformanceCounter(&t->counter); #endif } @@ -64,10 +72,10 @@ util_time_add(const struct util_time *t1, int64_t usecs, struct util_time *t2) { -#ifndef WIN32 +#if defined(PIPE_OS_LINUX) t2->tv.tv_sec = t1->tv.tv_sec + usecs / 1000000; t2->tv.tv_usec = t1->tv.tv_usec + usecs % 1000000; -#else +#elif defined(PIPE_OS_WINDOWS) if(!frequency) EngQueryPerformanceFrequency(&frequency); t2->counter = t1->counter + (usecs * frequency + 999999LL)/1000000LL; @@ -79,10 +87,12 @@ int64_t util_time_diff(const struct util_time *t1, const struct util_time *t2) { -#ifndef WIN32 +#if defined(PIPE_OS_LINUX) return (t2->tv.tv_usec - t1->tv.tv_usec) + (t2->tv.tv_sec - t1->tv.tv_sec)*1000000; -#else +#elif defined(PIPE_OS_WINDOWS) + if(!frequency) + EngQueryPerformanceFrequency(&frequency); return (t2->counter - t1->counter)*1000000LL/frequency; #endif } @@ -98,7 +108,7 @@ static INLINE int util_time_compare(const struct util_time *t1, const struct util_time *t2) { -#ifndef WIN32 +#if defined(PIPE_OS_LINUX) if (t1->tv.tv_sec < t2->tv.tv_sec) return -1; else if(t1->tv.tv_sec > t2->tv.tv_sec) @@ -109,7 +119,7 @@ util_time_compare(const struct util_time *t1, return 1; else return 0; -#else +#elif defined(PIPE_OS_WINDOWS) if (t1->counter < t2->counter) return -1; else if(t1->counter > t2->counter) @@ -132,7 +142,7 @@ util_time_timeout(const struct util_time *start, } -#ifdef WIN32 +#if defined(PIPE_OS_WINDOWS) void util_time_sleep(unsigned usecs) { LONGLONG start, curr, end; diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h index c8836c137f..48ec7a4a96 100644 --- a/src/gallium/auxiliary/util/u_time.h +++ b/src/gallium/auxiliary/util/u_time.h @@ -36,7 +36,9 @@ #define U_TIME_H_ -#ifndef WIN32 +#include "pipe/p_config.h" + +#if defined(PIPE_OS_LINUX) #include /* timeval */ #include /* usleep */ #endif @@ -56,7 +58,7 @@ extern "C" { */ struct util_time { -#ifndef WIN32 +#if defined(PIPE_OS_LINUX) struct timeval tv; #else long long counter; @@ -84,7 +86,7 @@ util_time_timeout(const struct util_time *start, const struct util_time *end, const struct util_time *curr); -#ifndef WIN32 +#if defined(PIPE_OS_LINUX) #define util_time_sleep usleep #else void diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 43d94ec4ba..63301ae3aa 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -28,6 +28,7 @@ #ifndef P_UTIL_H #define P_UTIL_H +#include "p_config.h" #include "p_compiler.h" #include "p_debug.h" #include "p_pointer.h" @@ -40,7 +41,7 @@ extern "C" { #endif -#if defined(WIN32) && defined(DEBUG) /* memory debugging */ +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && defined(DEBUG) /* memory debugging */ #include "p_debug.h" @@ -55,7 +56,7 @@ extern "C" { #else -#ifdef WIN32 +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) void * __stdcall EngAllocMem( @@ -118,7 +119,7 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) return new_ptr; } -#else /* !WIN32 */ +#else /* !PIPE_SUBSYSTEM_WINDOWS_DISPLAY */ #define MALLOC( SIZE ) malloc( SIZE ) @@ -128,7 +129,7 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) -#endif /* !WIN32 */ +#endif /* !PIPE_SUBSYSTEM_WINDOWS_DISPLAY */ #endif /* !DEBUG */ #define MALLOC_STRUCT(T) (struct T *) MALLOC(sizeof(struct T)) -- cgit v1.2.3 From d7b523b46b833fd08c70850d7a6cc7b6fd714a8e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 25 Apr 2008 10:34:20 +0100 Subject: gallium: Tell the driver the texture is updated when we finish rendering to it. --- src/mesa/state_tracker/st_cb_fbo.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index b73867bb72..69dde56e55 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -400,6 +400,10 @@ st_finish_render_texture(GLcontext *ctx, ctx->st->pipe->flush(ctx->st->pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + ctx->st->pipe->texture_update(ctx->st->pipe, + st_get_texobj_texture(att->Texture), + att->CubeMapFace, 1 << att->TextureLevel); + /* printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface); */ -- cgit v1.2.3 From 992d0b997f8f7e965e56852b81e01c290f8c13de Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 24 Apr 2008 16:22:47 -0400 Subject: frontend for rendering without elts --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/draw_private.h | 1 + src/gallium/auxiliary/draw/draw_pt.c | 30 ++- src/gallium/auxiliary/draw/draw_pt.h | 1 + src/gallium/auxiliary/draw/draw_pt_varray.c | 251 ++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 144 ++++++++++++++ 6 files changed, 421 insertions(+), 7 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_varray.c create mode 100644 src/gallium/auxiliary/draw/draw_pt_varray_tmp.h (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index bc6acfe458..da7eded21f 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -28,6 +28,7 @@ C_SOURCES = \ draw_pt_fetch_emit.c \ draw_pt_fetch_shade_pipeline.c \ draw_pt_post_vs.c \ + draw_pt_varray.c \ draw_pt_vcache.c \ draw_vertex.c \ draw_vs.c \ diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 49593c8fad..cee58bbf73 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -129,6 +129,7 @@ struct draw_context struct { struct draw_pt_front_end *vcache; + struct draw_pt_front_end *varray; } front; struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index f5a3bf390e..d6585bbfd0 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -53,12 +53,14 @@ draw_pt_arrays(struct draw_context *draw, struct draw_pt_front_end *frontend = NULL; struct draw_pt_middle_end *middle = NULL; unsigned opt = 0; + pt_elt_func get_elt = 0; + void *elts = 0; if (!draw->render) { opt |= PT_PIPELINE; } - if (draw_need_pipeline(draw, + if (draw_need_pipeline(draw, draw->rasterizer, prim)) { opt |= PT_PIPELINE; @@ -78,16 +80,21 @@ draw_pt_arrays(struct draw_context *draw, middle = draw->pt.middle.fetch_emit; - /* May create a short-circuited version of this for small primitives: + /* Pick the right frontend */ - frontend = draw->pt.front.vcache; + if (draw->pt.user.elts || + count >= 256) { + frontend = draw->pt.front.vcache; + get_elt = draw_pt_elt_func(draw); + elts = draw_pt_elt_ptr(draw, start); + } else { + frontend = draw->pt.front.varray; + elts = start; + } frontend->prepare( frontend, prim, middle, opt ); - frontend->run( frontend, - draw_pt_elt_func( draw ), - draw_pt_elt_ptr( draw, start ), - count ); + frontend->run(frontend, get_elt, elts, count); frontend->finish( frontend ); @@ -101,6 +108,10 @@ boolean draw_pt_init( struct draw_context *draw ) if (!draw->pt.front.vcache) return FALSE; + draw->pt.front.varray = draw_pt_varray(draw); + if (!draw->pt.front.varray) + return FALSE; + draw->pt.middle.fetch_emit = draw_pt_fetch_emit( draw ); if (!draw->pt.middle.fetch_emit) return FALSE; @@ -129,6 +140,11 @@ void draw_pt_destroy( struct draw_context *draw ) draw->pt.front.vcache->destroy( draw->pt.front.vcache ); draw->pt.front.vcache = NULL; } + + if (draw->pt.front.varray) { + draw->pt.front.varray->destroy( draw->pt.front.varray ); + draw->pt.front.varray = NULL; + } } diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index df5c29fc36..2dec376cee 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -115,6 +115,7 @@ const void *draw_pt_elt_ptr( struct draw_context *draw, * a special case for tiny vertex buffers. */ struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ); +struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw); /* Middle-ends: * diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c new file mode 100644 index 0000000000..022098a314 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -0,0 +1,251 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_pt.h" + +#define FETCH_MAX 256 +#define DRAW_MAX (16*FETCH_MAX) + +struct varray_frontend { + struct draw_pt_front_end base; + struct draw_context *draw; + + ushort draw_elts[DRAW_MAX]; + unsigned fetch_elts[FETCH_MAX]; + + unsigned draw_count; + unsigned fetch_count; + + struct draw_pt_middle_end *middle; + + unsigned input_prim; + unsigned output_prim; +}; + +static void varray_flush(struct varray_frontend *varray) +{ + if (varray->draw_count) { +#if 0 + debug_printf("FLUSH fc = %d, dc = %d\n", + varray->fetch_count, + varray->draw_count); +#endif + varray->middle->run(varray->middle, + varray->fetch_elts, + varray->fetch_count, + varray->draw_elts, + varray->draw_count); + } + + varray->fetch_count = 0; + varray->draw_count = 0; +} + +#if 0 +static void varray_check_flush(struct varray_frontend *varray) +{ + if (varray->draw_count + 6 >= DRAW_MAX/* || + varray->fetch_count + 4 >= FETCH_MAX*/) { + varray_flush(varray); + } +} +#endif + +static INLINE void add_draw_el(struct varray_frontend *varray, + int idx, ushort flags) +{ + varray->draw_elts[varray->draw_count++] = idx | flags; +} + + +static INLINE void varray_triangle( struct varray_frontend *varray, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + add_draw_el(varray, i0, 0); + add_draw_el(varray, i1, 0); + add_draw_el(varray, i2, 0); +} + +static INLINE void varray_triangle_flags( struct varray_frontend *varray, + ushort flags, + unsigned i0, + unsigned i1, + unsigned i2 ) +{ + add_draw_el(varray, i0, flags); + add_draw_el(varray, i1, 0); + add_draw_el(varray, i2, 0); +} + +static INLINE void varray_line( struct varray_frontend *varray, + unsigned i0, + unsigned i1 ) +{ + add_draw_el(varray, i0, 0); + add_draw_el(varray, i1, 0); +} + + +static INLINE void varray_line_flags( struct varray_frontend *varray, + ushort flags, + unsigned i0, + unsigned i1 ) +{ + add_draw_el(varray, i0, flags); + add_draw_el(varray, i1, 0); +} + + +static INLINE void varray_point( struct varray_frontend *varray, + unsigned i0 ) +{ + add_draw_el(varray, i0, 0); +} + +static INLINE void varray_quad( struct varray_frontend *varray, + unsigned i0, + unsigned i1, + unsigned i2, + unsigned i3 ) +{ + varray_triangle( varray, i0, i1, i3 ); + varray_triangle( varray, i1, i2, i3 ); +} + +static INLINE void varray_ef_quad( struct varray_frontend *varray, + unsigned i0, + unsigned i1, + unsigned i2, + unsigned i3 ) +{ + const unsigned omitEdge1 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_2; + const unsigned omitEdge2 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_1; + + varray_triangle_flags( varray, + DRAW_PIPE_RESET_STIPPLE | omitEdge1, + i0, i1, i3 ); + + varray_triangle_flags( varray, + omitEdge2, + i1, i2, i3 ); +} + +/* At least for now, we're back to using a template include file for + * this. The two paths aren't too different though - it may be + * possible to reunify them. + */ +#define TRIANGLE(vc,flags,i0,i1,i2) varray_triangle_flags(vc,flags,i0,i1,i2) +#define QUAD(vc,i0,i1,i2,i3) varray_ef_quad(vc,i0,i1,i2,i3) +#define LINE(vc,flags,i0,i1) varray_line_flags(vc,flags,i0,i1) +#define POINT(vc,i0) varray_point(vc,i0) +#define FUNC varray_run_extras +#include "draw_pt_varray_tmp.h" + +#define TRIANGLE(vc,flags,i0,i1,i2) varray_triangle(vc,i0,i1,i2) +#define QUAD(vc,i0,i1,i2,i3) varray_quad(vc,i0,i1,i2,i3) +#define LINE(vc,flags,i0,i1) varray_line(vc,i0,i1) +#define POINT(vc,i0) varray_point(vc,i0) +#define FUNC varray_run +#include "draw_pt_varray_tmp.h" + + + +static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { + PIPE_PRIM_POINTS, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_LINES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES, + PIPE_PRIM_TRIANGLES +}; + + + +static void varray_prepare(struct draw_pt_front_end *frontend, + unsigned prim, + struct draw_pt_middle_end *middle, + unsigned opt) +{ + struct varray_frontend *varray = (struct varray_frontend *)frontend; + const struct pipe_rasterizer_state *rasterizer = varray->draw->rasterizer; + + if (rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || + rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL || + rasterizer->line_stipple_enable) + { + assert(opt & PT_PIPELINE); + varray->base.run = varray_run_extras; + } else { + varray->base.run = varray_run; + } + + varray->input_prim = prim; + varray->output_prim = reduced_prim[prim]; + + varray->middle = middle; + middle->prepare(middle, varray->output_prim, opt); +} + + + + +static void varray_finish(struct draw_pt_front_end *frontend) +{ + struct varray_frontend *varray = (struct varray_frontend *)frontend; + varray->middle->finish(varray->middle); + varray->middle = NULL; +} + +static void varray_destroy(struct draw_pt_front_end *frontend) +{ + FREE(frontend); +} + + +struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw) +{ + struct varray_frontend *varray = CALLOC_STRUCT(varray_frontend); + if (varray == NULL) + return NULL; + + varray->base.prepare = varray_prepare; + varray->base.run = NULL; + varray->base.finish = varray_finish; + varray->base.destroy = varray_destroy; + varray->draw = draw; + + return &varray->base; +} diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h new file mode 100644 index 0000000000..b9a319b253 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -0,0 +1,144 @@ + +static void FUNC(struct draw_pt_front_end *frontend, + pt_elt_func get_elt, + const void *elts, + unsigned count) +{ + struct varray_frontend *varray = (struct varray_frontend *)frontend; + struct draw_context *draw = varray->draw; + unsigned start = (unsigned)elts; + + boolean flatfirst = (draw->rasterizer->flatshade && + draw->rasterizer->flatshade_first); + unsigned i, flags; + +#if 0 + debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count); +#endif +#if 0 + debug_printf("INPUT PRIM = %d (start = %d, count = %d)\n", varray->input_prim, + start, count); +#endif + + for (i = 0; i < count; ++i) { + varray->fetch_elts[i] = start + i; + } + varray->fetch_count = count; + + switch (varray->input_prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i ++) { + POINT(varray, i + 0); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) { + LINE(varray, DRAW_PIPE_RESET_STIPPLE, + i + 0, i + 1); + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + flags = DRAW_PIPE_RESET_STIPPLE; + + for (i = 1; i < count; i++, flags = 0) { + LINE(varray, flags, i - 1, i); + } + LINE(varray, flags, i - 1, 0); + } + break; + + case PIPE_PRIM_LINE_STRIP: + flags = DRAW_PIPE_RESET_STIPPLE; + for (i = 1; i < count; i++, flags = 0) { + LINE(varray, flags, i - 1, i); + } + break; + + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0, i + 1, i + 2); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0, i + 1 + (i&1), i + 2 - (i&1)); + } + } + else { + for (i = 0; i+2 < count; i++) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0 + (i&1), i + 1 - (i&1), i + 2); + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + if (count >= 3) { + if (flatfirst) { + flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; + for (i = 0; i+2 < count; i++) { + TRIANGLE(varray, flags, i + 1, i + 2, 0); + } + } + else { + flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; + for (i = 0; i+2 < count; i++) { + TRIANGLE(varray, flags, 0, i + 1, i + 2); + } + } + } + break; + + case PIPE_PRIM_QUADS: + for (i = 0; i+3 < count; i += 4) { + QUAD(varray, i + 0, i + 1, i + 2, i + 3); + } + break; + + case PIPE_PRIM_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + QUAD(varray, i + 2, i + 0, i + 1, i + 3); + } + break; + + case PIPE_PRIM_POLYGON: + { + /* These bitflags look a little odd because we submit the + * vertices as (1,2,0) to satisfy flatshade requirements. + */ + const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; + const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + + for (i = 0; i+2 < count; i++, flags = edge_middle) { + + if (i + 3 == count) + flags |= edge_last; + + TRIANGLE(varray, flags, i + 1, i + 2, 0); + } + } + break; + + default: + assert(0); + break; + } + + varray_flush(varray); +} + +#undef TRIANGLE +#undef QUAD +#undef POINT +#undef LINE +#undef FUNC -- cgit v1.2.3 From 4e46e6f52b88ca7df40a52cf994e6fe1e4b9870e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 25 Apr 2008 14:10:02 +0100 Subject: draw: remove unused vars --- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 2 -- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 2 -- 2 files changed, 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index 0545dbc171..f05641dee6 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -815,7 +815,6 @@ aaline_bind_sampler_states(struct pipe_context *pipe, unsigned num, void **sampler) { struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); - struct draw_context *draw = aaline->stage.draw; /* save current */ memcpy(aaline->state.sampler, sampler, num * sizeof(void *)); @@ -831,7 +830,6 @@ aaline_set_sampler_textures(struct pipe_context *pipe, unsigned num, struct pipe_texture **texture) { struct aaline_stage *aaline = aaline_stage_from_pipe(pipe); - struct draw_context *draw = aaline->stage.draw; uint i; /* save current */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index da303a13ad..d1d63d73be 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -667,7 +667,6 @@ pstip_set_sampler_textures(struct pipe_context *pipe, unsigned num, struct pipe_texture **texture) { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); - struct draw_context *draw = pstip->stage.draw; uint i; /* save current */ @@ -690,7 +689,6 @@ pstip_set_polygon_stipple(struct pipe_context *pipe, const struct pipe_poly_stipple *stipple) { struct pstip_stage *pstip = pstip_stage_from_pipe(pipe); - struct draw_context *draw = (struct draw_context *) pipe->draw; /* save current */ pstip->state.stipple = stipple; -- cgit v1.2.3 From 14cfcb18204233d3d4848ea8e579465983df3d75 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 25 Apr 2008 14:10:32 +0100 Subject: draw: no need to special-case elts/get_elts for varray frontend --- src/gallium/auxiliary/draw/draw_pt.c | 12 +++++------- 1 file changed, 5 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index d6585bbfd0..c9c5d18313 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -53,8 +53,6 @@ draw_pt_arrays(struct draw_context *draw, struct draw_pt_front_end *frontend = NULL; struct draw_pt_middle_end *middle = NULL; unsigned opt = 0; - pt_elt_func get_elt = 0; - void *elts = 0; if (!draw->render) { opt |= PT_PIPELINE; @@ -83,18 +81,18 @@ draw_pt_arrays(struct draw_context *draw, /* Pick the right frontend */ if (draw->pt.user.elts || - count >= 256) { + count >= 256) { frontend = draw->pt.front.vcache; - get_elt = draw_pt_elt_func(draw); - elts = draw_pt_elt_ptr(draw, start); } else { frontend = draw->pt.front.varray; - elts = start; } frontend->prepare( frontend, prim, middle, opt ); - frontend->run(frontend, get_elt, elts, count); + frontend->run(frontend, + draw_pt_elt_func(draw), + draw_pt_elt_ptr(draw, start), + count); frontend->finish( frontend ); -- cgit v1.2.3 From 2325d1959783aaf57178a86d8a0b28f168761e13 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 25 Apr 2008 14:20:02 +0100 Subject: tgsi: fix compile when HIGH_PRECISION not defined --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index c3295a27ff..559370e613 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -854,7 +854,7 @@ emit_rsqrt( /* On Intel CPUs at least, this is only accurate to 12 bits -- not * good enough. */ - emit_rsqrtps( + sse_rsqrtps( func, make_xmm( xmm_dst ), make_xmm( xmm_src ) ); -- cgit v1.2.3 From 96cfd804f6dcc0ec9f5e887ff1b402a55886fb0b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 25 Apr 2008 14:57:50 +0100 Subject: gallium: Add draw_pt_varray.c to scons build. --- src/gallium/auxiliary/draw/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 0b9852f633..3b5d5ed492 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -27,6 +27,7 @@ draw = env.ConvenienceLibrary( 'draw_pt_fetch_emit.c', 'draw_pt_fetch_shade_pipeline.c', 'draw_pt_post_vs.c', + 'draw_pt_varray.c', 'draw_pt_vcache.c', 'draw_vertex.c', 'draw_vs.c', -- cgit v1.2.3 From 3c0dc8242b64518d5635263ba65b39afa919dd86 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 25 Apr 2008 15:01:09 +0100 Subject: gallium: Use util_blit_pixels() for glCopyTexSubImage when appropriate. --- src/mesa/state_tracker/st_cb_texture.c | 37 +++++++++++++++++++++------------- 1 file changed, 23 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 3db2790267..a6c4351fff 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -49,6 +49,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "util/p_tile.h" +#include "util/u_blit.h" #define DBG if (0) printf @@ -1134,11 +1135,9 @@ do_copy_texsubimage(GLcontext *ctx, dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, stImage->level, destZ); - if (src_format == dest_format && - ctx->_ImageTransferState == 0x0 && + if (ctx->_ImageTransferState == 0x0 && strb->surface->buffer && - dest_surface->buffer && - strb->surface->cpp == stImage->pt->cpp) { + dest_surface->buffer) { /* do blit-style copy */ /* XXX may need to invert image depending on window @@ -1162,16 +1161,26 @@ do_copy_texsubimage(GLcontext *ctx, GL_COPY); /* ? */ #else - pipe->surface_copy(pipe, - do_flip, - /* dest */ - dest_surface, - destX, destY, - /* src */ - strb->surface, - srcX, srcY, - /* size */ - width, height); + if (src_format == dest_format) { + pipe->surface_copy(pipe, + do_flip, + /* dest */ + dest_surface, + destX, destY, + /* src */ + strb->surface, + srcX, srcY, + /* size */ + width, height); + } else { + util_blit_pixels(ctx->st->blit, + strb->surface, + srcX, do_flip ? srcY + height : srcY, + srcX + width, do_flip ? srcY : srcY + height, + dest_surface, + destX, destY, destX + width, destY + height, + 0.0, PIPE_TEX_MIPFILTER_NEAREST); + } #endif } else { -- cgit v1.2.3 From 1aedbf9efe4d1cf45be3c27fc3a0eb4a69a1b1b9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 25 Apr 2008 16:16:06 +0100 Subject: draw: emit extra flags whenever pipeline is active The assert was in fact over-sensitive, but emitting the extra flags is pretty trivial & we may as well just do it whenever we know the pipeline is running. --- src/gallium/auxiliary/draw/draw_pt_varray.c | 9 ++++----- src/gallium/auxiliary/draw/draw_pt_vcache.c | 6 +----- 2 files changed, 5 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 022098a314..c85d8ded50 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -202,13 +202,12 @@ static void varray_prepare(struct draw_pt_front_end *frontend, struct varray_frontend *varray = (struct varray_frontend *)frontend; const struct pipe_rasterizer_state *rasterizer = varray->draw->rasterizer; - if (rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL || - rasterizer->line_stipple_enable) + if (opt & PT_PIPELINE) { - assert(opt & PT_PIPELINE); varray->base.run = varray_run_extras; - } else { + } + else + { varray->base.run = varray_run; } diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index b3133359e0..2f9775814f 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -227,12 +227,8 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; const struct pipe_rasterizer_state *rasterizer = vcache->draw->rasterizer; - - if (rasterizer->fill_cw != PIPE_POLYGON_MODE_FILL || - rasterizer->fill_ccw != PIPE_POLYGON_MODE_FILL || - rasterizer->line_stipple_enable) + if (opt & PT_PIPELINE) { - assert(opt & PT_PIPELINE); vcache->base.run = vcache_run_extras; } else -- cgit v1.2.3 From 658f5e721316f1983ae39b6b94b992baca0e2bc2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 26 Apr 2008 01:14:32 +0900 Subject: scons: A few fixes for building mesa on windows. --- src/mesa/SConscript | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index 80fbdb41f0..6b99dacf2a 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -12,9 +12,19 @@ env.Append(CPPPATH = [ '#/src/mesa/main', ]) -env.Append(CFLAGS = [ - '-std=c99', -]) +if gcc: + env.Append(CFLAGS = [ + '-std=c99', + ]) + +# x86 assembly +if x86 and gcc: + env.Append(CPPDEFINES = [ + 'USE_X86_ASM', + 'USE_MMX_ASM', + 'USE_3DNOW_ASM', + 'USE_SSE_ASM', + ]) ####################################################################### @@ -269,7 +279,7 @@ SPARC_API = [ 'sparc/glapi_sparc.S', ] -if x86: +if x86 and gcc: ASM_SOURCES = ASM_C_SOURCES + X86_SOURCES API_SOURCES = X86_API else: -- cgit v1.2.3 From 3faea292e537b8fe8f86b03ae9147c73f17efc3b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 10:33:48 -0600 Subject: gallium: fix the compressed texture hack in st_texture_image_copy() Actually, the hack is still there and needs to be revisited, but I get a bit further with compressed textures now. --- src/mesa/state_tracker/st_texture.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 8e3235cc99..66d81e2b95 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -290,8 +290,7 @@ st_texture_image_copy(struct pipe_context *pipe, GLuint i; /* XXX this is a hack */ - if (dst->compressed) - height /= 4; + const GLuint copyHeight = dst->compressed ? height / 4 : height; for (i = 0; i < depth; i++) { GLuint srcLevel; @@ -315,7 +314,7 @@ st_texture_image_copy(struct pipe_context *pipe, 0, 0, /* destX, Y */ src_surface, 0, 0, /* srcX, Y */ - width, height); + width, copyHeight); pipe_surface_reference(&dst_surface, NULL); pipe_surface_reference(&src_surface, NULL); -- cgit v1.2.3 From 7dcbbdf353fe14401ba23e5b486658d2b46c550e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 13:59:25 -0600 Subject: gallium: use util_draw_vertex_buffer() instead of st_draw_vertices() --- src/mesa/state_tracker/st_cb_drawpixels.c | 94 +++++++++++++++++++++++-------- 1 file changed, 70 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 3921500659..8d5769f960 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -409,32 +409,43 @@ static void draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, GLfloat x1, GLfloat y1, GLboolean invertTex) { + struct st_context *st = ctx->st; + struct pipe_context *pipe = ctx->st->pipe; + const struct gl_framebuffer *fb = st->ctx->DrawBuffer; + const GLfloat fb_width = fb->Width; + const GLfloat fb_height = fb->Height; + const GLfloat clip_x0 = x0 / fb_width * 2.0 - 1.0; + const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; + const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; + const GLfloat clip_y1 = y1 / fb_height * 2.0 - 1.0; + const GLfloat sLeft = 0.0, sRight = 1.0; + const GLfloat tTop = invertTex, tBot = 1.0 - tTop; GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ + struct pipe_buffer *buf; + ubyte *map; GLuint i; - GLfloat sLeft = 0.0, sRight = 1.0; - GLfloat tTop = invertTex, tBot = 1.0 - tTop; /* upper-left */ - verts[0][0][0] = x0; /* attr[0].x */ - verts[0][0][1] = y0; /* attr[0].y */ - verts[0][1][0] = sLeft; /* attr[1].s */ - verts[0][1][1] = tTop; /* attr[1].t */ + verts[0][0][0] = clip_x0; /* attr[0].x */ + verts[0][0][1] = clip_y0; /* attr[0].y */ + verts[0][1][0] = sLeft; /* attr[1].s */ + verts[0][1][1] = tTop; /* attr[1].t */ /* upper-right */ - verts[1][0][0] = x1; - verts[1][0][1] = y0; + verts[1][0][0] = clip_x1; + verts[1][0][1] = clip_y0; verts[1][1][0] = sRight; verts[1][1][1] = tTop; /* lower-right */ - verts[2][0][0] = x1; - verts[2][0][1] = y1; + verts[2][0][0] = clip_x1; + verts[2][0][1] = clip_y1; verts[2][1][0] = sRight; verts[2][1][1] = tBot; /* lower-left */ - verts[3][0][0] = x0; - verts[3][0][1] = y1; + verts[3][0][0] = clip_x0; + verts[3][0][1] = clip_y1; verts[3][1][0] = sLeft; verts[3][1][1] = tBot; @@ -446,7 +457,21 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, verts[i][1][3] = 1.0; /*Q*/ } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 2, GL_FALSE); + buf = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(verts)); + /* put vertex data into buffer */ + map = pipe->winsys->buffer_map(pipe->winsys, buf, + PIPE_BUFFER_USAGE_CPU_WRITE); + memcpy(map, verts, sizeof(verts)); + pipe->winsys->buffer_unmap(pipe->winsys, buf); + + util_draw_vertex_buffer(pipe, buf, + PIPE_PRIM_QUADS, + 4, /* verts */ + 2); /* attribs/vert */ + + pipe->winsys->buffer_destroy(pipe->winsys, buf); } @@ -455,6 +480,15 @@ draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, GLfloat x1, GLfloat y1, const GLfloat *color, GLboolean invertTex) { + struct st_context *st = ctx->st; + struct pipe_context *pipe = ctx->st->pipe; + const struct gl_framebuffer *fb = st->ctx->DrawBuffer; + const GLfloat fb_width = fb->Width; + const GLfloat fb_height = fb->Height; + const GLfloat clip_x0 = x0 / fb_width * 2.0 - 1.0; + const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; + const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; + const GLfloat clip_y1 = y1 / fb_height * 2.0 - 1.0; GLfloat bias = ctx->st->bitmap_texcoord_bias; GLfloat verts[4][3][4]; /* four verts, three attribs, XYZW */ GLuint i; @@ -462,28 +496,30 @@ draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, GLfloat yBias = bias / (y1-y0); GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias; GLfloat tTop = invertTex - yBias, tBot = 1.0 - tTop - yBias; + struct pipe_buffer *buf; + ubyte *map; /* upper-left */ - verts[0][0][0] = x0; /* attr[0].x */ - verts[0][0][1] = y0; /* attr[0].y */ + verts[0][0][0] = clip_x0; /* attr[0].x */ + verts[0][0][1] = clip_y0; /* attr[0].y */ verts[0][2][0] = sLeft; /* attr[2].s */ verts[0][2][1] = tTop; /* attr[2].t */ /* upper-right */ - verts[1][0][0] = x1; - verts[1][0][1] = y0; + verts[1][0][0] = clip_x1; + verts[1][0][1] = clip_y0; verts[1][2][0] = sRight; verts[1][2][1] = tTop; /* lower-right */ - verts[2][0][0] = x1; - verts[2][0][1] = y1; + verts[2][0][0] = clip_x1; + verts[2][0][1] = clip_y1; verts[2][2][0] = sRight; verts[2][2][1] = tBot; /* lower-left */ - verts[3][0][0] = x0; - verts[3][0][1] = y1; + verts[3][0][0] = clip_x0; + verts[3][0][1] = clip_y1; verts[3][2][0] = sLeft; verts[3][2][1] = tBot; @@ -499,7 +535,19 @@ draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, verts[i][2][3] = 1.0; /*Q*/ } - st_draw_vertices(ctx, PIPE_PRIM_QUADS, 4, (float *) verts, 3, GL_FALSE); + buf = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(verts)); + /* put vertex data into buffer */ + map = pipe->winsys->buffer_map(pipe->winsys, buf, + PIPE_BUFFER_USAGE_CPU_WRITE); + memcpy(map, verts, sizeof(verts)); + pipe->winsys->buffer_unmap(pipe->winsys, buf); + + util_draw_vertex_buffer(pipe, buf, + PIPE_PRIM_QUADS, + 4, /* verts */ + 3); /* attribs/vert */ } @@ -878,9 +926,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, GLsizei width, GLsizei height, GLint dstx, GLint dsty) { - struct st_renderbuffer *rbRead = st_renderbuffer(ctx->ReadBuffer->_StencilBuffer); struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer); - struct pipe_surface *psRead = rbRead->surface; struct pipe_surface *psDraw = rbDraw->surface; ubyte *drawMap; ubyte *buffer; -- cgit v1.2.3 From b6d8b21cc8e36eb4f6fa72a067561f3fa8bd6ebf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 14:00:40 -0600 Subject: gallium: remove unused st_draw_vertices() --- src/mesa/state_tracker/st_draw.c | 70 ---------------------------------------- src/mesa/state_tracker/st_draw.h | 7 ---- 2 files changed, 77 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index e40eeddff5..6c20120ac7 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -404,76 +404,6 @@ st_draw_vbo(GLcontext *ctx, -/** - * Utility function for drawing simple primitives (such as quads for - * glClear and glDrawPixels). Coordinates are in screen space. - * \param mode one of PIPE_PRIM_x - * \param numVertex number of vertices - * \param verts vertex data (all attributes are float[4]) - * \param numAttribs number of attributes per vertex - */ -void -st_draw_vertices(GLcontext *ctx, unsigned prim, - unsigned numVertex, float *verts, - unsigned numAttribs, - GLboolean inClipCoords) -{ - const float width = ctx->DrawBuffer->Width; - const float height = ctx->DrawBuffer->Height; - const unsigned vertex_bytes = numVertex * numAttribs * 4 * sizeof(float); - struct pipe_context *pipe = ctx->st->pipe; - struct pipe_buffer *vbuf; - struct pipe_vertex_buffer vbuffer; - struct pipe_vertex_element velements[PIPE_MAX_ATTRIBS]; - unsigned i; - - assert(numAttribs > 0); - - if (!inClipCoords) { - /* convert to clip coords */ - for (i = 0; i < numVertex; i++) { - float x = verts[i * numAttribs * 4 + 0]; - float y = verts[i * numAttribs * 4 + 1]; - x = x / width * 2.0 - 1.0; - y = y / height * 2.0 - 1.0; - verts[i * numAttribs * 4 + 0] = x; - verts[i * numAttribs * 4 + 1] = y; - } - } - - /* XXX create one-time */ - vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_VERTEX, vertex_bytes); - assert(vbuf); - - memcpy(pipe->winsys->buffer_map(pipe->winsys, vbuf, - PIPE_BUFFER_USAGE_CPU_WRITE), - verts, vertex_bytes); - pipe->winsys->buffer_unmap(pipe->winsys, vbuf); - - /* tell pipe about the vertex buffer */ - vbuffer.buffer = vbuf; - vbuffer.pitch = numAttribs * 4 * sizeof(float); /* vertex size */ - vbuffer.buffer_offset = 0; - pipe->set_vertex_buffers(pipe, 1, &vbuffer); - - /* tell pipe about the vertex attributes */ - for (i = 0; i < numAttribs; i++) { - velements[i].src_offset = i * 4 * sizeof(GLfloat); - velements[i].vertex_buffer_index = 0; - velements[i].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - velements[i].nr_components = 4; - } - pipe->set_vertex_elements(pipe, numAttribs, velements); - - /* draw */ - pipe->draw_arrays(pipe, prim, 0, numVertex); - - /* XXX: do one-time */ - pipe_buffer_reference(pipe->winsys, &vbuf, NULL); -} - - /** * Set the (private) draw module's post-transformed vertex format when in * GL_SELECT or GL_FEEDBACK mode or for glRasterPos. diff --git a/src/mesa/state_tracker/st_draw.h b/src/mesa/state_tracker/st_draw.h index 171bde57e5..c81f2b25da 100644 --- a/src/mesa/state_tracker/st_draw.h +++ b/src/mesa/state_tracker/st_draw.h @@ -59,11 +59,4 @@ st_feedback_draw_vbo(GLcontext *ctx, GLuint min_index, GLuint max_index); -void -st_draw_vertices(GLcontext *ctx, unsigned prim, - unsigned numVertex, float *verts, - unsigned numAttribs, - GLboolean inClipCoords); - - #endif -- cgit v1.2.3 From 149a4175fafdb4dfcf31b28377f83092c9edf335 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 14:14:09 -0600 Subject: gallium: consolidate quad drawing code --- src/mesa/state_tracker/st_cb_drawpixels.c | 232 ++++++++++++------------------ 1 file changed, 90 insertions(+), 142 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 8d5769f960..7c87402f27 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -402,152 +402,105 @@ make_texture(struct st_context *st, /** - * Draw textured quad. + * Draw quad with texcoords and optional color. * Coords are window coords with y=0=bottom. + * \param color may be null + * \param invertTex if true, flip texcoords vertically */ static void draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, - GLfloat x1, GLfloat y1, GLboolean invertTex) + GLfloat x1, GLfloat y1, const GLfloat *color, + GLboolean invertTex) { struct st_context *st = ctx->st; struct pipe_context *pipe = ctx->st->pipe; - const struct gl_framebuffer *fb = st->ctx->DrawBuffer; - const GLfloat fb_width = fb->Width; - const GLfloat fb_height = fb->Height; - const GLfloat clip_x0 = x0 / fb_width * 2.0 - 1.0; - const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; - const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; - const GLfloat clip_y1 = y1 / fb_height * 2.0 - 1.0; - const GLfloat sLeft = 0.0, sRight = 1.0; - const GLfloat tTop = invertTex, tBot = 1.0 - tTop; - GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */ - struct pipe_buffer *buf; - ubyte *map; - GLuint i; - - /* upper-left */ - verts[0][0][0] = clip_x0; /* attr[0].x */ - verts[0][0][1] = clip_y0; /* attr[0].y */ - verts[0][1][0] = sLeft; /* attr[1].s */ - verts[0][1][1] = tTop; /* attr[1].t */ - - /* upper-right */ - verts[1][0][0] = clip_x1; - verts[1][0][1] = clip_y0; - verts[1][1][0] = sRight; - verts[1][1][1] = tTop; - - /* lower-right */ - verts[2][0][0] = clip_x1; - verts[2][0][1] = clip_y1; - verts[2][1][0] = sRight; - verts[2][1][1] = tBot; - - /* lower-left */ - verts[3][0][0] = clip_x0; - verts[3][0][1] = clip_y1; - verts[3][1][0] = sLeft; - verts[3][1][1] = tBot; - - /* same for all verts: */ - for (i = 0; i < 4; i++) { - verts[i][0][2] = z; /*Z*/ - verts[i][0][3] = 1.0; /*W*/ - verts[i][1][2] = 0.0; /*R*/ - verts[i][1][3] = 1.0; /*Q*/ - } - - buf = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_VERTEX, - sizeof(verts)); - /* put vertex data into buffer */ - map = pipe->winsys->buffer_map(pipe->winsys, buf, - PIPE_BUFFER_USAGE_CPU_WRITE); - memcpy(map, verts, sizeof(verts)); - pipe->winsys->buffer_unmap(pipe->winsys, buf); - - util_draw_vertex_buffer(pipe, buf, - PIPE_PRIM_QUADS, - 4, /* verts */ - 2); /* attribs/vert */ - - pipe->winsys->buffer_destroy(pipe->winsys, buf); -} - - -static void -draw_quad_colored(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, - GLfloat x1, GLfloat y1, const GLfloat *color, - GLboolean invertTex) -{ - struct st_context *st = ctx->st; - struct pipe_context *pipe = ctx->st->pipe; - const struct gl_framebuffer *fb = st->ctx->DrawBuffer; - const GLfloat fb_width = fb->Width; - const GLfloat fb_height = fb->Height; - const GLfloat clip_x0 = x0 / fb_width * 2.0 - 1.0; - const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; - const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; - const GLfloat clip_y1 = y1 / fb_height * 2.0 - 1.0; - GLfloat bias = ctx->st->bitmap_texcoord_bias; GLfloat verts[4][3][4]; /* four verts, three attribs, XYZW */ - GLuint i; - GLfloat xBias = bias / (x1-x0); - GLfloat yBias = bias / (y1-y0); - GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias; - GLfloat tTop = invertTex - yBias, tBot = 1.0 - tTop - yBias; - struct pipe_buffer *buf; - ubyte *map; - - /* upper-left */ - verts[0][0][0] = clip_x0; /* attr[0].x */ - verts[0][0][1] = clip_y0; /* attr[0].y */ - verts[0][2][0] = sLeft; /* attr[2].s */ - verts[0][2][1] = tTop; /* attr[2].t */ - - /* upper-right */ - verts[1][0][0] = clip_x1; - verts[1][0][1] = clip_y0; - verts[1][2][0] = sRight; - verts[1][2][1] = tTop; - - /* lower-right */ - verts[2][0][0] = clip_x1; - verts[2][0][1] = clip_y1; - verts[2][2][0] = sRight; - verts[2][2][1] = tBot; - - /* lower-left */ - verts[3][0][0] = clip_x0; - verts[3][0][1] = clip_y1; - verts[3][2][0] = sLeft; - verts[3][2][1] = tBot; - - /* same for all verts: */ - for (i = 0; i < 4; i++) { - verts[i][0][2] = z; /*Z*/ - verts[i][0][3] = 1.0; /*W*/ - verts[i][1][0] = color[0]; - verts[i][1][1] = color[1]; - verts[i][1][2] = color[2]; - verts[i][1][3] = color[3]; - verts[i][2][2] = 0.0; /*R*/ - verts[i][2][3] = 1.0; /*Q*/ + + /* setup vertex data */ + { + const struct gl_framebuffer *fb = st->ctx->DrawBuffer; + const GLfloat fb_width = fb->Width; + const GLfloat fb_height = fb->Height; + const GLfloat clip_x0 = x0 / fb_width * 2.0 - 1.0; + const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; + const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; + const GLfloat clip_y1 = y1 / fb_height * 2.0 - 1.0; + GLfloat bias = ctx->st->bitmap_texcoord_bias; + GLfloat xBias = 0*bias / (x1-x0); + GLfloat yBias = 0*bias / (y1-y0); + GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias; + GLfloat tTop = invertTex - yBias, tBot = 1.0 - tTop - yBias; + GLuint tex, i; + + /* upper-left */ + verts[0][0][0] = clip_x0; /* v[0].attr[0].x */ + verts[0][0][1] = clip_y0; /* v[0].attr[0].y */ + + /* upper-right */ + verts[1][0][0] = clip_x1; + verts[1][0][1] = clip_y0; + + /* lower-right */ + verts[2][0][0] = clip_x1; + verts[2][0][1] = clip_y1; + + /* lower-left */ + verts[3][0][0] = clip_x0; + verts[3][0][1] = clip_y1; + + tex = color ? 2 : 1; + verts[0][tex][0] = sLeft; /* v[0].attr[tex].s */ + verts[0][tex][1] = tTop; /* v[0].attr[tex].t */ + verts[1][tex][0] = sRight; + verts[1][tex][1] = tTop; + verts[2][tex][0] = sRight; + verts[2][tex][1] = tBot; + verts[3][tex][0] = sLeft; + verts[3][tex][1] = tBot; + + /* same for all verts: */ + if (color) { + for (i = 0; i < 4; i++) { + verts[i][0][2] = z; /*Z*/ + verts[i][0][3] = 1.0; /*W*/ + verts[i][1][0] = color[0]; + verts[i][1][1] = color[1]; + verts[i][1][2] = color[2]; + verts[i][1][3] = color[3]; + verts[i][2][2] = 0.0; /*R*/ + verts[i][2][3] = 1.0; /*Q*/ + } + } + else { + for (i = 0; i < 4; i++) { + verts[i][0][2] = z; /*Z*/ + verts[i][0][3] = 1.0; /*W*/ + verts[i][1][2] = 0.0; /*R*/ + verts[i][1][3] = 1.0; /*Q*/ + } + } } - buf = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_VERTEX, - sizeof(verts)); - /* put vertex data into buffer */ - map = pipe->winsys->buffer_map(pipe->winsys, buf, - PIPE_BUFFER_USAGE_CPU_WRITE); - memcpy(map, verts, sizeof(verts)); - pipe->winsys->buffer_unmap(pipe->winsys, buf); - - util_draw_vertex_buffer(pipe, buf, - PIPE_PRIM_QUADS, - 4, /* verts */ - 3); /* attribs/vert */ + { + struct pipe_buffer *buf; + ubyte *map; + + /* allocate/load buffer object with vertex data */ + buf = pipe->winsys->buffer_create(pipe->winsys, 32, + PIPE_BUFFER_USAGE_VERTEX, + sizeof(verts)); + map = pipe->winsys->buffer_map(pipe->winsys, buf, + PIPE_BUFFER_USAGE_CPU_WRITE); + memcpy(map, verts, sizeof(verts)); + pipe->winsys->buffer_unmap(pipe->winsys, buf); + + util_draw_vertex_buffer(pipe, buf, + PIPE_PRIM_QUADS, + 4, /* verts */ + 3); /* attribs/vert */ + + pipe->winsys->buffer_destroy(pipe->winsys, buf); + } } @@ -653,12 +606,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, x1 = x + width * ctx->Pixel.ZoomX; y0 = y; y1 = y + height * ctx->Pixel.ZoomY; - - /* draw textured quad */ - if (color) - draw_quad_colored(ctx, x0, y0, z, x1, y1, color, invertTex); - else - draw_quad(ctx, x0, y0, z, x1, y1, invertTex); + draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex); /* restore state */ cso_restore_rasterizer(cso); -- cgit v1.2.3 From 1437b41d9068017dbe981a784285d5773c1d1ead Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 14:15:42 -0600 Subject: gallium: fix typo s/_mesa_unmap_drapix_pbo/_mesa_unmap_drawpix_pbo/ --- src/mesa/main/bufferobj.c | 4 ++-- src/mesa/main/bufferobj.h | 4 ++-- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- src/mesa/swrast/s_drawpix.c | 2 +- 4 files changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c index e762eb3b63..dc0307feb5 100644 --- a/src/mesa/main/bufferobj.c +++ b/src/mesa/main/bufferobj.c @@ -551,8 +551,8 @@ _mesa_map_drawpix_pbo(GLcontext *ctx, * \sa _mesa_unmap_bitmap_pbo */ void -_mesa_unmap_drapix_pbo(GLcontext *ctx, - const struct gl_pixelstore_attrib *unpack) +_mesa_unmap_drawpix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack) { if (unpack->BufferObj->Name) { ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT, diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h index 46525f08ae..024e5a8c3c 100644 --- a/src/mesa/main/bufferobj.h +++ b/src/mesa/main/bufferobj.h @@ -101,8 +101,8 @@ _mesa_map_drawpix_pbo(GLcontext *ctx, const GLvoid *pixels); extern void -_mesa_unmap_drapix_pbo(GLcontext *ctx, - const struct gl_pixelstore_attrib *unpack); +_mesa_unmap_drawpix_pbo(GLcontext *ctx, + const struct gl_pixelstore_attrib *unpack); extern void * diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 7c87402f27..cdfa041d15 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -349,7 +349,7 @@ make_texture(struct st_context *st, pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height, 1, 0); if (!pt) { - _mesa_unmap_drapix_pbo(ctx, unpack); + _mesa_unmap_drawpix_pbo(ctx, unpack); return NULL; } @@ -395,7 +395,7 @@ make_texture(struct st_context *st, ctx->_ImageTransferState = imageTransferStateSave; } - _mesa_unmap_drapix_pbo(ctx, unpack); + _mesa_unmap_drawpix_pbo(ctx, unpack); return pt; } diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c index fb04d9f746..cbf6617058 100644 --- a/src/mesa/swrast/s_drawpix.c +++ b/src/mesa/swrast/s_drawpix.c @@ -877,7 +877,7 @@ end: RENDER_FINISH(swrast,ctx); - _mesa_unmap_drapix_pbo(ctx, unpack); + _mesa_unmap_drawpix_pbo(ctx, unpack); } -- cgit v1.2.3 From 9b44f5a7550d8ede2eda763770e927df4c15cc8e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 12:00:18 -0600 Subject: gallium: added a null ptr check --- src/mesa/state_tracker/st_cb_texture.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index a6c4351fff..9fd9fd70ea 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -896,6 +896,11 @@ st_TexSubimage(GLcontext * ctx, dstRowStride = stImage->surface->pitch * stImage->surface->cpp; } + if (!texImage->Data) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); + return; + } + src = (const GLubyte *) pixels; for (i = 0; i++ < depth;) { @@ -907,10 +912,11 @@ st_TexSubimage(GLcontext * ctx, texImage->ImageOffsets, width, height, 1, format, type, src, packing)) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "st_TexSubImage"); + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); } if (stImage->pt && i < depth) { + /* map next slice of 3D texture */ st_texture_image_unmap(stImage); texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i); src += srcImageStride; -- cgit v1.2.3 From 4c50969334844bc9aa622176c3ebcbc733394f78 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 15:21:00 -0600 Subject: gallium: added null ptr check --- src/mesa/state_tracker/st_cb_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 9fd9fd70ea..005bb2e54d 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1373,7 +1373,7 @@ copy_image_data_to_texture(struct st_context *st, pipe_texture_release(&stImage->pt); } - else { + else if (stImage->base.Data) { assert(stImage->base.Data != NULL); /* More straightforward upload. -- cgit v1.2.3 From 332b77b852905224741084c5a4f5d2f4625dd119 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 15:23:21 -0600 Subject: gallium: use simple color pass-through fragment shader if textures don't exist If we run out of texture memory we may not have the texture needed by the fragment shader. If this happens, plug in a color passthrough shader. So instead of crashing, we just don't see the texture. GL_OUT_OF_MEMORY is raised, of course. --- src/mesa/state_tracker/st_atom_texture.c | 30 ++++++++++++++++++++++++++++-- src/mesa/state_tracker/st_context.h | 2 ++ 2 files changed, 30 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 01c07dc26d..d15da5895a 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -38,6 +38,23 @@ #include "st_cb_texture.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "cso_cache/cso_context.h" +#include "util/u_simple_shaders.h" + + +static void * +get_passthrough_fs(struct st_context *st) +{ + struct pipe_shader_state shader; + + if (!st->passthrough_fs) { + st->passthrough_fs = + util_make_fragment_passthrough_shader(st->pipe, &shader); + free((void *) shader.tokens); + } + + return st->passthrough_fs; +} /** @@ -49,6 +66,7 @@ update_textures(struct st_context *st) { struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current; GLuint su; + GLboolean missing_textures = GL_FALSE; st->state.num_textures = 0; @@ -67,6 +85,7 @@ update_textures(struct st_context *st) retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); if (!retval) { /* out of mem */ + missing_textures = GL_TRUE; continue; } @@ -79,8 +98,15 @@ update_textures(struct st_context *st) pipe_texture_reference(&st->state.sampler_texture[su], pt); } - st->pipe->set_sampler_textures(st->pipe, st->state.num_textures, - st->state.sampler_texture); + cso_set_sampler_textures(st->cso_context, + st->state.num_textures, + st->state.sampler_texture); + + if (missing_textures) { + /* use a pass-through frag shader that uses no textures */ + void *fs = get_passthrough_fs(st); + cso_set_fragment_shader_handle(st->cso_context, fs); + } } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 2851770d4c..7e933a2b10 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -165,6 +165,8 @@ struct st_context struct pipe_buffer *vbuf; } clear; + void *passthrough_fs; /**< simple pass-through frag shader */ + struct gen_mipmap_state *gen_mipmap; struct blit_state *blit; -- cgit v1.2.3 From 1cf164142768d8338527ee8cab8dee83af82af0b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 16:12:11 -0600 Subject: gallium: remove unneeded st->bitmap_texcoord_bias --- src/mesa/state_tracker/st_cb_bitmap.c | 7 ++----- src/mesa/state_tracker/st_cb_drawpixels.c | 7 ++----- src/mesa/state_tracker/st_context.h | 2 -- src/mesa/state_tracker/st_extensions.c | 6 ++---- 4 files changed, 6 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 507ace27c7..6c2d9a4b89 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -355,11 +355,8 @@ setup_bitmap_vertex_data(struct st_context *st, const GLfloat x1 = x + width; const GLfloat y0 = y; const GLfloat y1 = y + height; - const GLfloat bias = 0.0 * st->bitmap_texcoord_bias; /* XXX remove */ - const GLfloat xBias = bias / (x1-x0); - const GLfloat yBias = bias / (y1-y0); - const GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias; - const GLfloat tTop = yBias, tBot = 1.0 - tTop - yBias; + const GLfloat sLeft = 0.0F, sRight = 1.0F; + const GLfloat tTop = 0.0, tBot = 1.0 - tTop; const GLfloat clip_x0 = x0 / fb_width * 2.0 - 1.0; const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index cdfa041d15..047ea3816b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -425,11 +425,8 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; const GLfloat clip_y1 = y1 / fb_height * 2.0 - 1.0; - GLfloat bias = ctx->st->bitmap_texcoord_bias; - GLfloat xBias = 0*bias / (x1-x0); - GLfloat yBias = 0*bias / (y1-y0); - GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias; - GLfloat tTop = invertTex - yBias, tBot = 1.0 - tTop - yBias; + const GLfloat sLeft = 0.0F, sRight = 1.0F; + const GLfloat tTop = invertTex, tBot = 1.0 - tTop; GLuint tex, i; /* upper-left */ diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 7e933a2b10..80a71d1766 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -123,8 +123,6 @@ struct st_context GLfloat polygon_offset_scale; /* ?? */ - GLfloat bitmap_texcoord_bias; - /** Mapping from VERT_RESULT_x to post-transformed vertex slot */ const GLuint *vertex_result_to_slot; diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index f2d40e84b3..260a2efe88 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -109,9 +109,6 @@ void st_init_limits(struct st_context *st) c->MaxTextureLodBias = screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_LOD_BIAS); - st->bitmap_texcoord_bias - = screen->get_paramf(screen, PIPE_CAP_BITMAP_TEXCOORD_BIAS); - c->MaxDrawBuffers = CLAMP(screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS), 1, MAX_DRAW_BUFFERS); @@ -220,11 +217,12 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.EXT_texture_sRGB = GL_TRUE; } +#if 01 if (screen->is_format_supported(screen, PIPE_FORMAT_DXT5_RGBA, PIPE_TEXTURE)) { ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE; } - +#endif if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR, PIPE_TEXTURE) || screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV, PIPE_TEXTURE)) { ctx->Extensions.MESA_ycbcr_texture = GL_TRUE; -- cgit v1.2.3 From 5e3b0d227c1b009c88a1b79df4a9e8c632e356f9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 15:55:47 -0600 Subject: gallium: tweak comments, minor var renaming --- src/gallium/include/pipe/p_winsys.h | 37 ++++++++++++++++--------------------- 1 file changed, 16 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h index bf5aad7351..3005ec2d94 100644 --- a/src/gallium/include/pipe/p_winsys.h +++ b/src/gallium/include/pipe/p_winsys.h @@ -63,13 +63,13 @@ struct pipe_surface; struct pipe_winsys { /** Returns name of this winsys interface */ - const char *(*get_name)( struct pipe_winsys *sws ); + const char *(*get_name)( struct pipe_winsys *ws ); /** * Do any special operations to ensure frontbuffer contents are * displayed, eg copy fake frontbuffer. */ - void (*flush_frontbuffer)( struct pipe_winsys *sws, + void (*flush_frontbuffer)( struct pipe_winsys *ws, struct pipe_surface *surf, void *context_private ); @@ -79,8 +79,8 @@ struct pipe_winsys /** * Allocate storage for a pipe_surface. - * Returns 0 if succeeds. - * XXX: flags is unused and will be removed someday. + * \param flags XXX unused, remove someday + * \return 0 if succeeds. */ int (*surface_alloc_storage)(struct pipe_winsys *ws, struct pipe_surface *surf, @@ -105,7 +105,7 @@ struct pipe_winsys * alignment indicates the client's alignment requirements, eg for * SSE instructions. */ - struct pipe_buffer *(*buffer_create)( struct pipe_winsys *sws, + struct pipe_buffer *(*buffer_create)( struct pipe_winsys *ws, unsigned alignment, unsigned usage, unsigned size ); @@ -131,7 +131,7 @@ struct pipe_winsys * Note that ptr may be accessed at any time upto the time when the * buffer is destroyed, so the data must not be freed before then. */ - struct pipe_buffer *(*user_buffer_create)(struct pipe_winsys *sws, + struct pipe_buffer *(*user_buffer_create)(struct pipe_winsys *ws, void *ptr, unsigned bytes); @@ -139,45 +139,40 @@ struct pipe_winsys * Map the entire data store of a buffer object into the client's address. * flags is bitmask of PIPE_BUFFER_USAGE_CPU_READ/WRITE flags. */ - void *(*buffer_map)( struct pipe_winsys *sws, + void *(*buffer_map)( struct pipe_winsys *ws, struct pipe_buffer *buf, unsigned usage ); - void (*buffer_unmap)( struct pipe_winsys *sws, + void (*buffer_unmap)( struct pipe_winsys *ws, struct pipe_buffer *buf ); - void (*buffer_destroy)( struct pipe_winsys *sws, + void (*buffer_destroy)( struct pipe_winsys *ws, struct pipe_buffer *buf ); /** Set ptr = fence, with reference counting */ - void (*fence_reference)( struct pipe_winsys *sws, + void (*fence_reference)( struct pipe_winsys *ws, struct pipe_fence_handle **ptr, struct pipe_fence_handle *fence ); /** * Checks whether the fence has been signalled. - * - * The meaning of flag is pipe-driver specific. - * - * Returns zero if it has. + * \param flags driver-specific meaning + * \return zero on success. */ - int (*fence_signalled)( struct pipe_winsys *sws, + int (*fence_signalled)( struct pipe_winsys *ws, struct pipe_fence_handle *fence, unsigned flag ); /** * Wait for the fence to finish. - * - * The meaning of flag is pipe-driver specific. - * - * Returns zero on success. + * \param flags driver-specific meaning + * \return zero on success. */ - int (*fence_finish)( struct pipe_winsys *sws, + int (*fence_finish)( struct pipe_winsys *ws, struct pipe_fence_handle *fence, unsigned flag ); - }; -- cgit v1.2.3 From aad9dd14d8ef44e55f70b639765dae89e9976345 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 15:59:55 -0600 Subject: gallium: tweak comments, minor var renaming --- src/gallium/include/pipe/p_screen.h | 22 +++++++++++++++------- 1 file changed, 15 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h index 35bd46d691..26ac99d287 100644 --- a/src/gallium/include/pipe/p_screen.h +++ b/src/gallium/include/pipe/p_screen.h @@ -58,25 +58,33 @@ struct pipe_screen { void (*destroy)( struct pipe_screen * ); - /* - * Capability queries - */ const char *(*get_name)( struct pipe_screen * ); const char *(*get_vendor)( struct pipe_screen * ); + /** + * Query an integer-valued capability/parameter/limit + * \param param one of PIPE_CAP_x + */ int (*get_param)( struct pipe_screen *, int param ); + /** + * Query a float-valued capability/parameter/limit + * \param param one of PIPE_CAP_x + */ float (*get_paramf)( struct pipe_screen *, int param ); - /**< type is one of PIPE_TEXTURE, PIPE_SURFACE */ + /** + * Check if the given pipe_format is supported as a texture or + * drawing surface. + * \param type one of PIPE_TEXTURE, PIPE_SURFACE + */ boolean (*is_format_supported)( struct pipe_screen *, enum pipe_format format, uint type ); - - /* - * Texture functions + /** + * Create a new texture object, using the given template info. */ struct pipe_texture * (*texture_create)(struct pipe_screen *, const struct pipe_texture *templat); -- cgit v1.2.3 From a94aad297d9804688f888a5112326104a5d00e07 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Apr 2008 18:46:29 -0600 Subject: use PIPE_ARCH_X86 --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 559370e613..9061e00b63 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -34,7 +34,7 @@ #include "rtasm/rtasm_x86sse.h" -#if defined(__i386__) || defined(__386__) +#ifdef PIPE_ARCH_X86 #define HIGH_PRECISION 1 /* for 1/sqrt() */ @@ -2181,4 +2181,4 @@ tgsi_emit_sse2( return ok; } -#endif /* i386 */ +#endif /* PIPE_ARCH_X86 */ -- cgit v1.2.3 From a8e39b6f5a1fedf2f8719e1adb8802ebbfc09688 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 25 Apr 2008 19:25:26 -0600 Subject: gallium: fix broken SGT, SLE --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 18 ++++++++++++++++-- 1 file changed, 16 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 29e104bbd1..d55f907c0d 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -516,6 +516,20 @@ micro_lg2( dst->f[3] = logf( src->f[3] ) * 1.442695f; } +static void +micro_le( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] <= src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] <= src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] <= src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] <= src1->f[3] ? src2->f[3] : src3->f[3]; +} + static void micro_lt( union tgsi_exec_channel *dst, @@ -1975,7 +1989,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); STORE( &r[0], 0, chan_index ); } break; @@ -1992,7 +2006,7 @@ exec_instruction( FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); STORE( &r[0], 0, chan_index ); } break; -- cgit v1.2.3 From e3c415995706d2dda7c34a227e2e24d0745763ec Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 27 Apr 2008 21:09:45 +0900 Subject: rtasm: Implement x86_retw. --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 6 ++++++ src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 1 + 2 files changed, 7 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 3cd45d7dd9..e6cbe9967f 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -498,6 +498,12 @@ void x86_ret( struct x86_function *p ) emit_1ub(p, 0xc3); } +void x86_retw( struct x86_function *p, unsigned short imm ) +{ + DUMP(); + emit_3ub(p, 0xc2, imm & 0xff, (imm >> 8) & 0xff); +} + void x86_sahf( struct x86_function *p ) { DUMP(); diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index 695a1cef4e..1962b07bc5 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -219,6 +219,7 @@ void x86_or( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_pop( struct x86_function *p, struct x86_reg reg ); void x86_push( struct x86_function *p, struct x86_reg reg ); void x86_ret( struct x86_function *p ); +void x86_retw( struct x86_function *p, unsigned short imm ); void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -- cgit v1.2.3 From 083008d808c02226a3dfead6000a84efd5e6a9fa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 27 Apr 2008 21:19:34 +0900 Subject: pipebuffer: Be extra cautious with the incoming buffers. A common mistake is trying to fence user or malloc buffers. So don't let the crash happen inside pipebuffer lib. --- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index d3c1ec4fbe..a92daf91dd 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -356,11 +356,25 @@ void buffer_fence(struct pb_buffer *buf, struct pipe_fence_handle *fence) { - struct fenced_buffer *fenced_buf = fenced_buffer(buf); - struct fenced_buffer_list *fenced_list = fenced_buf->list; - struct pipe_winsys *winsys = fenced_list->winsys; + struct fenced_buffer *fenced_buf; + struct fenced_buffer_list *fenced_list; + struct pipe_winsys *winsys; /* FIXME: receive this as a parameter */ unsigned flags = fence ? PIPE_BUFFER_USAGE_GPU_READ_WRITE : 0; + + /* This is a public function, so be extra cautious with the buffer passed, + * as happens frequently to receive null buffers, or pointer to buffers + * other than fenced buffers. */ + assert(buf); + if(!buf) + return; + assert(buf->vtbl == &fenced_buffer_vtbl); + if(buf->vtbl != &fenced_buffer_vtbl) + return; + + fenced_buf = fenced_buffer(buf); + fenced_list = fenced_buf->list; + winsys = fenced_list->winsys; if(fence == fenced_buf->fence) { /* Handle the same fence case specially, not only because it is a fast -- cgit v1.2.3 From 0d179ffe9774487f37838aa634c59ff8ebc1111d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 28 Apr 2008 09:33:20 -0600 Subject: gallium: add cases for PIPE_FORMAT_X8Z24_UNORM --- src/gallium/auxiliary/util/p_tile.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 91283a9ab0..13175ca46e 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -730,6 +730,7 @@ pipe_get_tile_rgba(struct pipe_context *pipe, z32_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: s8z24_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_Z24S8_UNORM: @@ -808,6 +809,7 @@ pipe_put_tile_rgba(struct pipe_context *pipe, /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_Z24S8_UNORM: @@ -852,6 +854,7 @@ pipe_get_tile_z(struct pipe_context *pipe, } break; case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: { const uint *pSrc = (const uint *) pipe_surface_map(ps) + (y * ps->pitch + x); @@ -912,6 +915,7 @@ pipe_put_tile_z(struct pipe_context *pipe, } break; case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: { uint *pDest = (uint *) pipe_surface_map(ps) + (y * ps->pitch + x); for (i = 0; i < h; i++) { -- cgit v1.2.3 From ee4434121c5ce556af8af899297d4249435c6d71 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 28 Apr 2008 09:38:18 -0600 Subject: gallium: return enum pipe_format, not uint, for a few functions --- src/mesa/state_tracker/st_format.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 3d15a03cab..8b5f84cd56 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -293,7 +293,7 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) /** * Find an RGBA format supported by the context/winsys. */ -static GLuint +static enum pipe_format default_rgba_format(struct pipe_screen *screen, uint type) { static const enum pipe_format colorFormats[] = { @@ -315,7 +315,7 @@ default_rgba_format(struct pipe_screen *screen, uint type) /** * Search list of formats for first RGBA format with >8 bits/channel. */ -static GLuint +static enum pipe_format default_deep_rgba_format(struct pipe_screen *screen, uint type) { if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SNORM, type)) { @@ -331,7 +331,7 @@ default_deep_rgba_format(struct pipe_screen *screen, uint type) /** * Find an Z format supported by the context/winsys. */ -static GLuint +static enum pipe_format default_depth_format(struct pipe_screen *screen, uint type) { static const enum pipe_format zFormats[] = { -- cgit v1.2.3 From 44c79f88b8abbbcafe4e6e462d7f0d8505265e2b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 28 Apr 2008 09:39:48 -0600 Subject: gallium: replace some code with a call to st_choose_format() --- src/mesa/state_tracker/st_cb_drawpixels.c | 18 ++---------------- 1 file changed, 2 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 047ea3816b..7597ea323c 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -983,22 +983,8 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, else { /* srcFormat can't be used as a texture format */ if (type == GL_DEPTH) { - static const enum pipe_format zFormats[] = { - PIPE_FORMAT_Z16_UNORM, - PIPE_FORMAT_Z32_UNORM, - PIPE_FORMAT_S8Z24_UNORM, - PIPE_FORMAT_Z24S8_UNORM - }; - uint i; - texFormat = 0; - for (i = 0; i < Elements(zFormats); i++) { - if (screen->is_format_supported(screen, zFormats[i], - PIPE_TEXTURE)) { - texFormat = zFormats[i]; - break; - } - } - assert(texFormat); /* XXX no depth texture formats??? */ + texFormat = st_choose_format(pipe, GL_DEPTH_COMPONENT, PIPE_TEXTURE); + assert(texFormat != PIPE_FORMAT_NONE); /* XXX no depth texture formats??? */ } else { /* todo */ -- cgit v1.2.3 From 7f5e9d3f07f6fbfa699bef4ffff85fe0b557584a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 28 Apr 2008 09:40:26 -0600 Subject: gallium: check for PIPE_FORMAT_X8Z24_UNORM format --- src/mesa/state_tracker/st_cb_readpixels.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 1fef55b844..ddbe36106c 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -240,7 +240,8 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, const GLint dstStride = _mesa_image_row_stride(&clippedPacking, width, format, type); - if (strb->surface->format == PIPE_FORMAT_S8Z24_UNORM) { + if (strb->surface->format == PIPE_FORMAT_S8Z24_UNORM || + strb->surface->format == PIPE_FORMAT_X8Z24_UNORM) { if (format == GL_DEPTH_COMPONENT) { for (i = 0; i < height; i++) { GLuint ztemp[MAX_WIDTH], j; -- cgit v1.2.3 From 58d3dff0d3115ddd5397b7f77b5bcf4f9ca616b6 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 28 Apr 2008 18:50:27 +0200 Subject: gallium: Generate SSE code to swizzle and unswizzle vs inputs and outputs. Change SSE_SWIZZLES #define to 0 to disable it. --- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 3 +- src/gallium/auxiliary/draw/draw_vs_sse.c | 52 ++++++-- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 14 ++ src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 2 + src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 142 ++++++++++++++++++++- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h | 4 +- src/gallium/drivers/softpipe/sp_fs_sse.c | 2 +- 7 files changed, 204 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index f0763dad8d..4ec20493c4 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -109,9 +109,10 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, struct draw_context *draw = fpme->draw; struct draw_vertex_shader *shader = draw->vertex_shader; unsigned opt = fpme->opt; + unsigned alloc_count = align_int( fetch_count, 4 ); struct vertex_header *pipeline_verts = - (struct vertex_header *)MALLOC(fpme->vertex_size * fetch_count); + (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count); if (!pipeline_verts) { /* Not much we can do here - just skip the rendering. diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index b1e9f67114..07f85bc448 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -47,14 +47,29 @@ #include "tgsi/util/tgsi_parse.h" #define SSE_MAX_VERTICES 4 +#define SSE_SWIZZLES 1 +#if SSE_SWIZZLES +typedef void (XSTDCALL *codegen_function) ( + const struct tgsi_exec_vector *input, + struct tgsi_exec_vector *output, + float (*constant)[4], + struct tgsi_exec_vector *temporary, + float (*immediates)[4], + const float (*aos_input)[4], + uint num_inputs, + uint input_stride, + float (*aos_output)[4], + uint num_outputs, + uint output_stride ); +#else typedef void (XSTDCALL *codegen_function) ( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, float (*constant)[4], struct tgsi_exec_vector *temporary, float (*immediates)[4] ); - +#endif struct draw_sse_vertex_shader { struct draw_vertex_shader base; @@ -91,12 +106,31 @@ vs_sse_run_linear( struct draw_vertex_shader *base, { struct draw_sse_vertex_shader *shader = (struct draw_sse_vertex_shader *)base; struct tgsi_exec_machine *machine = shader->machine; - unsigned int i, j; - unsigned slot; + unsigned int i; for (i = 0; i < count; i += MAX_TGSI_VERTICES) { unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); +#if SSE_SWIZZLES + /* run compiled shader + */ + shader->func(machine->Inputs, + machine->Outputs, + (float (*)[4])constants, + machine->Temps, + shader->immediates, + input, + base->info.num_inputs, + input_stride, + output, + base->info.num_outputs, + output_stride ); + + input = (const float (*)[4])((const char *)input + input_stride * max_vertices); + output = (float (*)[4])((char *)output + output_stride * max_vertices); +#else + unsigned int j, slot; + /* Swizzle inputs. */ for (j = 0; j < max_vertices; j++) { @@ -105,10 +139,10 @@ vs_sse_run_linear( struct draw_vertex_shader *base, machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; - } + } input = (const float (*)[4])((const char *)input + input_stride); - } + } /* run compiled shader */ @@ -118,7 +152,6 @@ vs_sse_run_linear( struct draw_vertex_shader *base, machine->Temps, shader->immediates); - /* Unswizzle all output results. */ for (j = 0; j < max_vertices; j++) { @@ -127,10 +160,11 @@ vs_sse_run_linear( struct draw_vertex_shader *base, output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; - } + } output = (float (*)[4])((char *)output + output_stride); - } + } +#endif } } @@ -176,7 +210,7 @@ draw_create_vs_sse(struct draw_context *draw, x86_init_func( &vs->sse2_program ); if (!tgsi_emit_sse2( (struct tgsi_token *) vs->base.state.tokens, - &vs->sse2_program, vs->immediates )) + &vs->sse2_program, vs->immediates, SSE_SWIZZLES )) goto fail; vs->func = (codegen_function) x86_get_func( &vs->sse2_program ); diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index e6cbe9967f..d7e2230557 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -853,6 +853,20 @@ void sse_shufps( struct x86_function *p, emit_1ub(p, shuf); } +void sse_unpckhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) +{ + DUMP_RR( dst, src ); + emit_2ub( p, X86_TWOB, 0x15 ); + emit_modrm( p, dst, src ); +} + +void sse_unpcklps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) +{ + DUMP_RR( dst, src ); + emit_2ub( p, X86_TWOB, 0x14 ); + emit_modrm( p, dst, src ); +} + void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index 1962b07bc5..ad79b1facf 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -203,6 +203,8 @@ void sse_rsqrtps( struct x86_function *p, struct x86_reg dst, struct x86_reg src void sse_rsqrtss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_shufps( struct x86_function *p, struct x86_reg dest, struct x86_reg arg0, unsigned char shuf ); +void sse_unpckhps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); +void sse_unpcklps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_pmovmskb( struct x86_function *p, struct x86_reg dest, struct x86_reg src ); void sse2_punpcklbw( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 9061e00b63..86ca16c246 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -1788,7 +1788,6 @@ emit_instruction( break; case TGSI_OPCODE_RET: - case TGSI_OPCODE_END: #ifdef WIN32 emit_retw( func, 16 ); #else @@ -1796,6 +1795,9 @@ emit_instruction( #endif break; + case TGSI_OPCODE_END: + break; + case TGSI_OPCODE_SSG: return 0; break; @@ -2027,6 +2029,127 @@ emit_declaration( } } +static void aos_to_soa( struct x86_function *func, uint aos, uint soa, uint num, uint stride ) +{ + struct x86_reg soa_input; + struct x86_reg aos_input; + struct x86_reg num_inputs; + struct x86_reg temp; + unsigned char *inner_loop; + + soa_input = x86_make_reg( file_REG32, reg_AX ); + aos_input = x86_make_reg( file_REG32, reg_BX ); + num_inputs = x86_make_reg( file_REG32, reg_CX ); + temp = x86_make_reg( file_REG32, reg_DX ); + + /* Save EBX */ + x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); + + x86_mov( func, soa_input, get_argument( soa + 1 ) ); + x86_mov( func, aos_input, get_argument( aos + 1 ) ); + x86_mov( func, num_inputs, get_argument( num + 1 ) ); + + inner_loop = x86_get_label( func ); + + x86_mov( func, temp, get_argument( stride + 1 ) ); + x86_push( func, aos_input ); + sse_movlps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); + sse_movlps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); + x86_add( func, aos_input, temp ); + sse_movhps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); + sse_movhps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); + x86_add( func, aos_input, temp ); + sse_movlps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); + sse_movlps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); + x86_add( func, aos_input, temp ); + sse_movhps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); + sse_movhps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); + x86_pop( func, aos_input ); + + sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); + sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); + sse_shufps( func, make_xmm( 0 ), make_xmm( 1 ), 0x88 ); + sse_shufps( func, make_xmm( 2 ), make_xmm( 1 ), 0xdd ); + sse_shufps( func, make_xmm( 3 ), make_xmm( 4 ), 0x88 ); + sse_shufps( func, make_xmm( 5 ), make_xmm( 4 ), 0xdd ); + + sse_movups( func, x86_make_disp( soa_input, 0 ), make_xmm( 0 ) ); + sse_movups( func, x86_make_disp( soa_input, 16 ), make_xmm( 2 ) ); + sse_movups( func, x86_make_disp( soa_input, 32 ), make_xmm( 3 ) ); + sse_movups( func, x86_make_disp( soa_input, 48 ), make_xmm( 5 ) ); + + /* Advance to next input */ + x86_mov_reg_imm( func, temp, 16 ); + x86_add( func, aos_input, temp ); + x86_mov_reg_imm( func, temp, 64 ); + x86_add( func, soa_input, temp ); + x86_dec( func, num_inputs ); + x86_jcc( func, cc_NE, inner_loop ); + + /* Restore EBX */ + x86_pop( func, x86_make_reg( file_REG32, reg_BX ) ); +} + +static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, uint stride ) +{ + struct x86_reg soa_output; + struct x86_reg aos_output; + struct x86_reg num_outputs; + struct x86_reg temp; + unsigned char *inner_loop; + + soa_output = x86_make_reg( file_REG32, reg_AX ); + aos_output = x86_make_reg( file_REG32, reg_BX ); + num_outputs = x86_make_reg( file_REG32, reg_CX ); + temp = x86_make_reg( file_REG32, reg_DX ); + + /* Save EBX */ + x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); + + x86_mov( func, soa_output, get_argument( soa + 1 ) ); + x86_mov( func, aos_output, get_argument( aos + 1 ) ); + x86_mov( func, num_outputs, get_argument( num + 1 ) ); + + inner_loop = x86_get_label( func ); + + sse_movups( func, make_xmm( 0 ), x86_make_disp( soa_output, 0 ) ); + sse_movups( func, make_xmm( 1 ), x86_make_disp( soa_output, 16 ) ); + sse_movups( func, make_xmm( 3 ), x86_make_disp( soa_output, 32 ) ); + sse_movups( func, make_xmm( 4 ), x86_make_disp( soa_output, 48 ) ); + + sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); + sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); + sse_unpcklps( func, make_xmm( 0 ), make_xmm( 1 ) ); + sse_unpckhps( func, make_xmm( 2 ), make_xmm( 1 ) ); + sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) ); + sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) ); + + x86_mov( func, temp, get_argument( stride + 1 ) ); + x86_push( func, aos_output ); + sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); + sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); + x86_add( func, aos_output, temp ); + sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); + sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); + x86_add( func, aos_output, temp ); + sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); + sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); + x86_add( func, aos_output, temp ); + sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); + sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); + x86_pop( func, aos_output ); + + /* Advance to next output */ + x86_mov_reg_imm( func, temp, 16 ); + x86_add( func, aos_output, temp ); + x86_mov_reg_imm( func, temp, 64 ); + x86_add( func, soa_output, temp ); + x86_dec( func, num_outputs ); + x86_jcc( func, cc_NE, inner_loop ); + + /* Restore EBX */ + x86_pop( func, x86_make_reg( file_REG32, reg_BX ) ); +} /** * Translate a TGSI vertex/fragment shader to SSE2 code. @@ -2048,7 +2171,8 @@ unsigned tgsi_emit_sse2( const struct tgsi_token *tokens, struct x86_function *func, - float (*immediates)[4]) + float (*immediates)[4], + boolean do_swizzles ) { struct tgsi_parse_context parse; boolean instruction_phase = FALSE; @@ -2089,6 +2213,9 @@ tgsi_emit_sse2( else { assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX); + if (do_swizzles) + aos_to_soa( func, 5, 0, 6, 7 ); + x86_mov( func, get_input_base(), @@ -2176,6 +2303,17 @@ tgsi_emit_sse2( } } + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) { + if (do_swizzles) + soa_to_aos( func, 8, 1, 9, 10 ); + } + +#ifdef WIN32 + emit_retw( func, 16 ); +#else + emit_ret( func ); +#endif + tgsi_parse_free( &parse ); return ok; diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h index 063287dc5e..e66d115283 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h @@ -12,8 +12,8 @@ unsigned tgsi_emit_sse2( const struct tgsi_token *tokens, struct x86_function *function, - float (*immediates)[4] - ); + float (*immediates)[4], + boolean do_swizzles ); #if defined __cplusplus } diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index f857d26143..4d569e1e22 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -133,7 +133,7 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe, x86_init_func( &shader->sse2_program ); if (!tgsi_emit_sse2( templ->tokens, &shader->sse2_program, - shader->immediates)) { + shader->immediates, FALSE )) { FREE(shader); return NULL; } -- cgit v1.2.3 From b6d9666a420bd7c31a6696f94ba1025e5204d458 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 28 Apr 2008 19:03:20 +0200 Subject: gallium: Set SSE_SWIZZLES to 0 by default. --- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 07f85bc448..90926aec85 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -47,7 +47,7 @@ #include "tgsi/util/tgsi_parse.h" #define SSE_MAX_VERTICES 4 -#define SSE_SWIZZLES 1 +#define SSE_SWIZZLES 0 #if SSE_SWIZZLES typedef void (XSTDCALL *codegen_function) ( -- cgit v1.2.3 From a41b78d107264227f3338446e04dcfda32634f52 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 25 Apr 2008 15:27:33 +0100 Subject: pb: remove unused variable, squash warning --- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index a92daf91dd..2fa0842971 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -134,8 +134,6 @@ _fenced_buffer_add(struct fenced_buffer *fenced_buf) static INLINE void _fenced_buffer_destroy(struct fenced_buffer *fenced_buf) { - struct fenced_buffer_list *fenced_list = fenced_buf->list; - assert(!fenced_buf->base.base.refcount); assert(!fenced_buf->fence); pb_reference(&fenced_buf->buffer, NULL); -- cgit v1.2.3 From 546ab045d8a18758ffc44da9dc76ad1335553cf6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 25 Apr 2008 15:28:54 +0100 Subject: translate: squash warnings --- src/gallium/auxiliary/translate/translate_cache.c | 3 ++- src/gallium/auxiliary/translate/translate_cache.h | 2 +- 2 files changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_cache.c b/src/gallium/auxiliary/translate/translate_cache.c index e91d648eba..115dc9287e 100644 --- a/src/gallium/auxiliary/translate/translate_cache.c +++ b/src/gallium/auxiliary/translate/translate_cache.c @@ -28,6 +28,7 @@ #include "pipe/p_util.h" #include "pipe/p_state.h" #include "translate.h" +#include "translate_cache.h" #include "cso_cache/cso_cache.h" #include "cso_cache/cso_hash.h" @@ -36,7 +37,7 @@ struct translate_cache { struct cso_hash *hash; }; -struct translate_cache * translate_cache_create() +struct translate_cache * translate_cache_create( void ) { struct translate_cache *cache = MALLOC_STRUCT(translate_cache); cache->hash = cso_hash_create(); diff --git a/src/gallium/auxiliary/translate/translate_cache.h b/src/gallium/auxiliary/translate/translate_cache.h index 63fc57b7ea..7dba871e57 100644 --- a/src/gallium/auxiliary/translate/translate_cache.h +++ b/src/gallium/auxiliary/translate/translate_cache.h @@ -38,7 +38,7 @@ struct translate_cache; struct translate_key; struct translate; -struct translate_cache *translate_cache_create(); +struct translate_cache *translate_cache_create( void ); void translate_cache_destroy(struct translate_cache *cache); /** -- cgit v1.2.3 From 9fb444f191015b44498a5c83d762519ccc98ed55 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Apr 2008 18:43:27 +0100 Subject: tsgi: add a makefile --- src/gallium/auxiliary/tgsi/exec/Makefile | 2 ++ 1 file changed, 2 insertions(+) create mode 100644 src/gallium/auxiliary/tgsi/exec/Makefile (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/Makefile b/src/gallium/auxiliary/tgsi/exec/Makefile new file mode 100644 index 0000000000..451911a354 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/exec/Makefile @@ -0,0 +1,2 @@ +default: + cd .. ; make -- cgit v1.2.3 From c4917c62311522df902003d77b146fc677c09a4e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 28 Apr 2008 18:50:31 +0100 Subject: tgsi: make loop structure clearer, use x86_lea for increments --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 134 ++++++++++++++-------------- 1 file changed, 68 insertions(+), 66 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 86ca16c246..1138f59997 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -2049,40 +2049,41 @@ static void aos_to_soa( struct x86_function *func, uint aos, uint soa, uint num, x86_mov( func, aos_input, get_argument( aos + 1 ) ); x86_mov( func, num_inputs, get_argument( num + 1 ) ); + /* do */ inner_loop = x86_get_label( func ); - - x86_mov( func, temp, get_argument( stride + 1 ) ); - x86_push( func, aos_input ); - sse_movlps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); - sse_movlps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); - x86_add( func, aos_input, temp ); - sse_movhps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); - sse_movhps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); - x86_add( func, aos_input, temp ); - sse_movlps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); - sse_movlps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); - x86_add( func, aos_input, temp ); - sse_movhps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); - sse_movhps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); - x86_pop( func, aos_input ); - - sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); - sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); - sse_shufps( func, make_xmm( 0 ), make_xmm( 1 ), 0x88 ); - sse_shufps( func, make_xmm( 2 ), make_xmm( 1 ), 0xdd ); - sse_shufps( func, make_xmm( 3 ), make_xmm( 4 ), 0x88 ); - sse_shufps( func, make_xmm( 5 ), make_xmm( 4 ), 0xdd ); - - sse_movups( func, x86_make_disp( soa_input, 0 ), make_xmm( 0 ) ); - sse_movups( func, x86_make_disp( soa_input, 16 ), make_xmm( 2 ) ); - sse_movups( func, x86_make_disp( soa_input, 32 ), make_xmm( 3 ) ); - sse_movups( func, x86_make_disp( soa_input, 48 ), make_xmm( 5 ) ); - - /* Advance to next input */ - x86_mov_reg_imm( func, temp, 16 ); - x86_add( func, aos_input, temp ); - x86_mov_reg_imm( func, temp, 64 ); - x86_add( func, soa_input, temp ); + { + x86_mov( func, temp, get_argument( stride + 1 ) ); + x86_push( func, aos_input ); + sse_movlps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); + sse_movlps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); + x86_add( func, aos_input, temp ); + sse_movhps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); + sse_movhps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); + x86_add( func, aos_input, temp ); + sse_movlps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); + sse_movlps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); + x86_add( func, aos_input, temp ); + sse_movhps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); + sse_movhps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); + x86_pop( func, aos_input ); + + sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); + sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); + sse_shufps( func, make_xmm( 0 ), make_xmm( 1 ), 0x88 ); + sse_shufps( func, make_xmm( 2 ), make_xmm( 1 ), 0xdd ); + sse_shufps( func, make_xmm( 3 ), make_xmm( 4 ), 0x88 ); + sse_shufps( func, make_xmm( 5 ), make_xmm( 4 ), 0xdd ); + + sse_movups( func, x86_make_disp( soa_input, 0 ), make_xmm( 0 ) ); + sse_movups( func, x86_make_disp( soa_input, 16 ), make_xmm( 2 ) ); + sse_movups( func, x86_make_disp( soa_input, 32 ), make_xmm( 3 ) ); + sse_movups( func, x86_make_disp( soa_input, 48 ), make_xmm( 5 ) ); + + /* Advance to next input */ + x86_lea( func, aos_input, x86_make_disp(aos_input, 16) ); + x86_lea( func, soa_input, x86_make_disp(soa_input, 64) ); + } + /* while --num_inputs */ x86_dec( func, num_inputs ); x86_jcc( func, cc_NE, inner_loop ); @@ -2110,40 +2111,41 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, x86_mov( func, aos_output, get_argument( aos + 1 ) ); x86_mov( func, num_outputs, get_argument( num + 1 ) ); + /* do */ inner_loop = x86_get_label( func ); - - sse_movups( func, make_xmm( 0 ), x86_make_disp( soa_output, 0 ) ); - sse_movups( func, make_xmm( 1 ), x86_make_disp( soa_output, 16 ) ); - sse_movups( func, make_xmm( 3 ), x86_make_disp( soa_output, 32 ) ); - sse_movups( func, make_xmm( 4 ), x86_make_disp( soa_output, 48 ) ); - - sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); - sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); - sse_unpcklps( func, make_xmm( 0 ), make_xmm( 1 ) ); - sse_unpckhps( func, make_xmm( 2 ), make_xmm( 1 ) ); - sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) ); - sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) ); - - x86_mov( func, temp, get_argument( stride + 1 ) ); - x86_push( func, aos_output ); - sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); - sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); - x86_add( func, aos_output, temp ); - sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); - sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); - x86_add( func, aos_output, temp ); - sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); - sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); - x86_add( func, aos_output, temp ); - sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); - sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); - x86_pop( func, aos_output ); - - /* Advance to next output */ - x86_mov_reg_imm( func, temp, 16 ); - x86_add( func, aos_output, temp ); - x86_mov_reg_imm( func, temp, 64 ); - x86_add( func, soa_output, temp ); + { + sse_movups( func, make_xmm( 0 ), x86_make_disp( soa_output, 0 ) ); + sse_movups( func, make_xmm( 1 ), x86_make_disp( soa_output, 16 ) ); + sse_movups( func, make_xmm( 3 ), x86_make_disp( soa_output, 32 ) ); + sse_movups( func, make_xmm( 4 ), x86_make_disp( soa_output, 48 ) ); + + sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); + sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); + sse_unpcklps( func, make_xmm( 0 ), make_xmm( 1 ) ); + sse_unpckhps( func, make_xmm( 2 ), make_xmm( 1 ) ); + sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) ); + sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) ); + + x86_mov( func, temp, get_argument( stride + 1 ) ); + x86_push( func, aos_output ); + sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); + sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); + x86_add( func, aos_output, temp ); + sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); + sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); + x86_add( func, aos_output, temp ); + sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); + sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); + x86_add( func, aos_output, temp ); + sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); + sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); + x86_pop( func, aos_output ); + + /* Advance to next output */ + x86_lea( func, aos_output, x86_make_disp(aos_output, 16) ); + x86_lea( func, soa_output, x86_make_disp(soa_output, 64) ); + } + /* while --num_outputs */ x86_dec( func, num_outputs ); x86_jcc( func, cc_NE, inner_loop ); -- cgit v1.2.3 From c2497879f127251994064a0e0c24901782adae9e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 28 Apr 2008 14:04:19 -0600 Subject: gallium: fix incorrect level param in copy_image_data_to_texture() This fixes a weird conform test failure that depended on something that happened several tests earlier. This took waaaaaay too long to find/fix. --- src/mesa/state_tracker/st_cb_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 005bb2e54d..3454c3e8e8 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1381,7 +1381,7 @@ copy_image_data_to_texture(struct st_context *st, st_texture_image_data(st->pipe, stObj->pt, stImage->face, - stImage->level, + dstLevel, stImage->base.Data, stImage->base.RowStride, stImage->base.RowStride * -- cgit v1.2.3 From 216603348c399b58fadf39b972d3f708e82f253b Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 29 Apr 2008 17:54:38 +0200 Subject: gallium: set border color --- src/mesa/state_tracker/st_atom_sampler.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 5dd242ac66..e1ddb53b80 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -159,6 +159,11 @@ update_samplers(struct st_context *st) sampler->max_lod = MIN2(texobj->MaxLevel, texobj->MaxLod); #endif + sampler->border_color[0] = texobj->BorderColor[RCOMP]; + sampler->border_color[1] = texobj->BorderColor[GCOMP]; + sampler->border_color[2] = texobj->BorderColor[BCOMP]; + sampler->border_color[3] = texobj->BorderColor[ACOMP]; + sampler->max_anisotropy = texobj->MaxAnisotropy; if (sampler->max_anisotropy > 1.0) { sampler->min_img_filter = PIPE_TEX_FILTER_ANISO; -- cgit v1.2.3 From 5e8d7899a8114918054f07c807aef07cbbfcb6b4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 28 Apr 2008 16:32:42 -0600 Subject: gallium: fix error msg typo --- src/gallium/auxiliary/util/u_pack_color.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 50ef44992b..89e3b5dc45 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -201,7 +201,7 @@ util_pack_z(enum pipe_format format, double z) case PIPE_FORMAT_Z24X8_UNORM: return ((uint) (z * 0xffffff)) << 8; default: - debug_printf("gallium: unhandled fomrat in util_pack_z()"); + debug_printf("gallium: unhandled format in util_pack_z()\n"); return 0; } } -- cgit v1.2.3 From 1cff4992b389ad884a663c93bdd7b7c6be6c79d2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 28 Apr 2008 16:33:15 -0600 Subject: gallium: add \n to error strings --- src/gallium/auxiliary/util/u_pack_color.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 89e3b5dc45..8296f1291c 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -91,7 +91,7 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, return; /* XXX lots more cases to add */ default: - debug_printf("gallium: unhandled format in util_pack_color_ub()"); + debug_printf("gallium: unhandled format in util_pack_color_ub()\n"); } } @@ -174,7 +174,7 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) return; /* XXX lots more cases to add */ default: - debug_printf("gallium: unhandled format in util_pack_color()"); + debug_printf("gallium: unhandled format in util_pack_color()\n"); } } -- cgit v1.2.3 From 5fb774ab31b11f3a55d9dc47cee5eeaf5abc5981 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 29 Apr 2008 12:51:06 -0600 Subject: mesa: added _mesa_scale_and_bias_depth_uint() --- src/mesa/main/pixel.c | 15 +++++++++++++++ src/mesa/main/pixel.h | 3 +++ 2 files changed, 18 insertions(+) (limited to 'src') diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c index eb4fd6e7c9..0e9915dd38 100644 --- a/src/mesa/main/pixel.c +++ b/src/mesa/main/pixel.c @@ -1341,6 +1341,21 @@ _mesa_scale_and_bias_depth(const GLcontext *ctx, GLuint n, } +void +_mesa_scale_and_bias_depth_uint(const GLcontext *ctx, GLuint n, + GLuint depthValues[]) +{ + const GLdouble max = (double) 0xffffffff; + const GLdouble scale = ctx->Pixel.DepthScale; + const GLdouble bias = ctx->Pixel.DepthBias * max; + GLuint i; + for (i = 0; i < n; i++) { + GLdouble d = (GLdouble) depthValues[i] * scale + bias; + d = CLAMP(d, 0.0, max); + depthValues[i] = (GLuint) d; + } +} + /**********************************************************************/ /***** State Management *****/ diff --git a/src/mesa/main/pixel.h b/src/mesa/main/pixel.h index 09155cfd70..3ba5b6689a 100644 --- a/src/mesa/main/pixel.h +++ b/src/mesa/main/pixel.h @@ -116,6 +116,9 @@ extern void _mesa_scale_and_bias_depth(const GLcontext *ctx, GLuint n, GLfloat depthValues[]); +extern void +_mesa_scale_and_bias_depth_uint(const GLcontext *ctx, GLuint n, + GLuint depthValues[]); extern void _mesa_update_pixel( GLcontext *ctx, GLuint newstate ); -- cgit v1.2.3 From cc94863429d8a6cef8dbf861e78701a458d30f5a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 29 Apr 2008 12:53:37 -0600 Subject: gallium: fix broken GL_DEPTH_COMPONENT case in fallback_copy_texsubimage() Also, make sure surfaces are suitable for blitting before calling util_blit_pixels(). --- src/mesa/state_tracker/st_cb_texture.c | 59 +++++++++++++++++++++++----------- 1 file changed, 40 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 3454c3e8e8..3d9c550d8c 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -31,6 +31,7 @@ #include "main/image.h" #include "main/macros.h" #include "main/mipmap.h" +#include "main/pixel.h" #include "main/texcompress.h" #include "main/texformat.h" #include "main/teximage.h" @@ -1038,7 +1039,6 @@ fallback_copy_texsubimage(GLcontext *ctx, const uint face = texture_face(target); struct pipe_texture *pt = stImage->pt; struct pipe_surface *src_surf, *dest_surf; - GLfloat *data; GLint row, yStep; st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); @@ -1056,25 +1056,38 @@ fallback_copy_texsubimage(GLcontext *ctx, dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ); - /* buffer for one row */ - data = (GLfloat *) malloc(width * 4 * sizeof(GLfloat)); + assert(width <= MAX_WIDTH); - /* do copy row by row */ - for (row = 0; row < height; row++) { - pipe_get_tile_rgba(pipe, src_surf, srcX, srcY + row, width, 1, data); - - /* XXX we're ignoring convolution for now */ - if (ctx->_ImageTransferState) { - _mesa_apply_rgba_transfer_ops(ctx, - ctx->_ImageTransferState & ~IMAGE_CONVOLUTION_BIT, - width, (GLfloat (*)[4])data); + /* + * To avoid a large temp memory allocation, do copy row by row. + */ + if (baseFormat == GL_DEPTH_COMPONENT) { + const GLboolean scaleOrBias = (ctx->Pixel.DepthScale != 1.0F || + ctx->Pixel.DepthBias != 0.0F); + + for (row = 0; row < height; row++, srcY++, destY += yStep) { + uint data[MAX_WIDTH]; + pipe_get_tile_z(pipe, src_surf, srcX, srcY, width, 1, data); + if (scaleOrBias) { + _mesa_scale_and_bias_depth_uint(ctx, width, data); + } + pipe_put_tile_z(pipe, dest_surf, destX, destY, width, 1, data); + } + } + else { + /* RGBA format */ + for (row = 0; row < height; row++, srcY++, destY += yStep) { + float data[4 * MAX_WIDTH]; + pipe_get_tile_rgba(pipe, src_surf, srcX, srcY, width, 1, data); + /* XXX we're ignoring convolution for now */ + if (ctx->_ImageTransferState) { + _mesa_apply_rgba_transfer_ops(ctx, + ctx->_ImageTransferState & ~IMAGE_CONVOLUTION_BIT, + width, (GLfloat (*)[4]) data); + } + pipe_put_tile_rgba(pipe, dest_surf, destX, destY, width, 1, data); } - - pipe_put_tile_rgba(pipe, dest_surf, destX, destY, width, 1, data); - destY += yStep; } - - free(data); } @@ -1111,6 +1124,7 @@ do_copy_texsubimage(GLcontext *ctx, struct pipe_surface *dest_surface; uint dest_format, src_format; uint do_flip = FALSE; + GLboolean use_fallback = GL_TRUE; (void) texImage; @@ -1178,7 +1192,12 @@ do_copy_texsubimage(GLcontext *ctx, srcX, srcY, /* size */ width, height); - } else { + use_fallback = GL_FALSE; + } + else if (screen->is_format_supported(screen, strb->surface->format, + PIPE_TEXTURE) && + screen->is_format_supported(screen, dest_surface->format, + PIPE_SURFACE)) { util_blit_pixels(ctx->st->blit, strb->surface, srcX, do_flip ? srcY + height : srcY, @@ -1186,10 +1205,12 @@ do_copy_texsubimage(GLcontext *ctx, dest_surface, destX, destY, destX + width, destY + height, 0.0, PIPE_TEX_MIPFILTER_NEAREST); + use_fallback = GL_FALSE; } #endif } - else { + + if (use_fallback) { fallback_copy_texsubimage(ctx, target, level, strb, stImage, baseFormat, destX, destY, destZ, -- cgit v1.2.3 From dd9dc7df80d208b884b4c090e4408c9a12aa6095 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 29 Apr 2008 12:54:52 -0600 Subject: gallium: added some assertions to st_render_texture() to check surface format Make sure we can really render to the texture surface given its format. --- src/mesa/state_tracker/st_cb_fbo.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 69dde56e55..2d741d9390 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -370,6 +370,8 @@ st_render_texture(GLcontext *ctx, att->TextureLevel, att->Zoffset); assert(strb->surface); + assert(screen->is_format_supported(screen, strb->surface->format, PIPE_TEXTURE)); + assert(screen->is_format_supported(screen, strb->surface->format, PIPE_SURFACE)); init_renderbuffer_bits(strb, pt->format); -- cgit v1.2.3 From 733bc4df1a53bb1899933685e06c52109d096bee Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 29 Apr 2008 12:55:41 -0600 Subject: gallium: added some assertions to be sure the blit's surface formats are suitable --- src/gallium/auxiliary/util/u_blit.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 1105066cb8..9e9912c6e4 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -264,6 +264,9 @@ util_blit_pixels(struct blit_state *ctx, dstY1 = tmp; } + assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE)); + assert(screen->is_format_supported(screen, dst->format, PIPE_SURFACE)); + /* * XXX for now we're always creating a temporary texture. * Strictly speaking that's not always needed. -- cgit v1.2.3 From 9bfe1a3d505733489f7583fe603b7d192f38fa8c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 29 Apr 2008 20:33:37 +0100 Subject: gallium: add debug_print_format() make it easier to print format error messages --- src/gallium/auxiliary/util/p_debug.c | 97 +++++++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_pack_color.h | 9 ++- src/gallium/include/pipe/p_debug.h | 11 ++-- src/gallium/include/pipe/p_format.h | 83 +------------------------- 4 files changed, 111 insertions(+), 89 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index cd612e23b3..eaf2c9bd98 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -41,6 +41,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_util.h" #include "pipe/p_debug.h" +#include "pipe/p_format.h" #include "util/u_string.h" @@ -324,3 +325,99 @@ debug_dump_flags(const struct debug_named_value *names, } + + + + +char *pf_sprint_name( char *str, enum pipe_format format ) +{ + strcpy( str, "PIPE_FORMAT_" ); + switch (pf_layout( format )) { + case PIPE_FORMAT_LAYOUT_RGBAZS: + { + pipe_format_rgbazs_t rgbazs = (pipe_format_rgbazs_t) format; + uint i; + uint scale = 1 << (pf_exp8( rgbazs ) * 3); + + for (i = 0; i < 4; i++) { + uint size = pf_size_xyzw( rgbazs, i ); + + if (size == 0) { + break; + } + switch (pf_swizzle_xyzw( rgbazs, i )) { + case PIPE_FORMAT_COMP_R: + strcat( str, "R" ); + break; + case PIPE_FORMAT_COMP_G: + strcat( str, "G" ); + break; + case PIPE_FORMAT_COMP_B: + strcat( str, "B" ); + break; + case PIPE_FORMAT_COMP_A: + strcat( str, "A" ); + break; + case PIPE_FORMAT_COMP_0: + strcat( str, "0" ); + break; + case PIPE_FORMAT_COMP_1: + strcat( str, "1" ); + break; + case PIPE_FORMAT_COMP_Z: + strcat( str, "Z" ); + break; + case PIPE_FORMAT_COMP_S: + strcat( str, "S" ); + break; + } + util_snprintf( &str[strlen( str )], 32, "%u", size * scale ); + } + if (i != 0) { + strcat( str, "_" ); + } + switch (pf_type( rgbazs )) { + case PIPE_FORMAT_TYPE_UNKNOWN: + strcat( str, "NONE" ); + break; + case PIPE_FORMAT_TYPE_FLOAT: + strcat( str, "FLOAT" ); + break; + case PIPE_FORMAT_TYPE_UNORM: + strcat( str, "UNORM" ); + break; + case PIPE_FORMAT_TYPE_SNORM: + strcat( str, "SNORM" ); + break; + case PIPE_FORMAT_TYPE_USCALED: + strcat( str, "USCALED" ); + break; + case PIPE_FORMAT_TYPE_SSCALED: + strcat( str, "SSCALED" ); + break; + } + } + break; + case PIPE_FORMAT_LAYOUT_YCBCR: + { + pipe_format_ycbcr_t ycbcr = (pipe_format_ycbcr_t) format; + + strcat( str, "YCBCR" ); + if (pf_rev( ycbcr )) { + strcat( str, "_REV" ); + } + } + break; + } + return str; +} + + +void debug_print_format(const char *msg, enum pipe_format fmt ) +{ + char fmtstr[80]; + + pf_sprint_name(fmtstr, fmt); + + debug_printf("%s: %s\n", msg, fmtstr); +} diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 8296f1291c..7e26b1642a 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -91,7 +91,8 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, return; /* XXX lots more cases to add */ default: - debug_printf("gallium: unhandled format in util_pack_color_ub()\n"); + debug_print_format("gallium: unhandled format in util_pack_color_ub()", format); + assert(0); } } @@ -174,7 +175,8 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) return; /* XXX lots more cases to add */ default: - debug_printf("gallium: unhandled format in util_pack_color()\n"); + debug_print_format("gallium: unhandled format in util_pack_color()", format); + assert(0); } } @@ -201,7 +203,8 @@ util_pack_z(enum pipe_format format, double z) case PIPE_FORMAT_Z24X8_UNORM: return ((uint) (z * 0xffffff)) << 8; default: - debug_printf("gallium: unhandled format in util_pack_z()\n"); + debug_print_format("gallium: unhandled format in util_pack_z()", format); + assert(0); return 0; } } diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 235c47abac..f87247994b 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -103,16 +103,19 @@ debug_printf(const char *format, ...) #endif +#ifdef DEBUG /** * Dump a blob in hex to the same place that debug_printf sends its * messages. */ -#ifdef DEBUG -void debug_print_blob( const char *name, - const void *blob, - unsigned size ); +void debug_print_blob( const char *name, const void *blob, unsigned size ); + +/* Print a message along with a prettified format string + */ +void debug_print_format(const char *msg, enum pipe_format fmt ); #else #define debug_print_blob(_name, _blob, _size) ((void)0) +#define debug_print_format(_msg, _fmt) ((void)0) #endif diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 94c2eeb62b..bc23fe142e 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -337,88 +337,7 @@ enum pipe_format { /** * Builds pipe format name from format token. */ -static INLINE char *pf_sprint_name( char *str, enum pipe_format format ) -{ - strcpy( str, "PIPE_FORMAT_" ); - switch (pf_layout( format )) { - case PIPE_FORMAT_LAYOUT_RGBAZS: - { - pipe_format_rgbazs_t rgbazs = (pipe_format_rgbazs_t) format; - uint i; - uint scale = 1 << (pf_exp8( rgbazs ) * 3); - - for (i = 0; i < 4; i++) { - uint size = pf_size_xyzw( rgbazs, i ); - - if (size == 0) { - break; - } - switch (pf_swizzle_xyzw( rgbazs, i )) { - case PIPE_FORMAT_COMP_R: - strcat( str, "R" ); - break; - case PIPE_FORMAT_COMP_G: - strcat( str, "G" ); - break; - case PIPE_FORMAT_COMP_B: - strcat( str, "B" ); - break; - case PIPE_FORMAT_COMP_A: - strcat( str, "A" ); - break; - case PIPE_FORMAT_COMP_0: - strcat( str, "0" ); - break; - case PIPE_FORMAT_COMP_1: - strcat( str, "1" ); - break; - case PIPE_FORMAT_COMP_Z: - strcat( str, "Z" ); - break; - case PIPE_FORMAT_COMP_S: - strcat( str, "S" ); - break; - } - util_snprintf( &str[strlen( str )], 32, "%u", size * scale ); - } - if (i != 0) { - strcat( str, "_" ); - } - switch (pf_type( rgbazs )) { - case PIPE_FORMAT_TYPE_UNKNOWN: - strcat( str, "NONE" ); - break; - case PIPE_FORMAT_TYPE_FLOAT: - strcat( str, "FLOAT" ); - break; - case PIPE_FORMAT_TYPE_UNORM: - strcat( str, "UNORM" ); - break; - case PIPE_FORMAT_TYPE_SNORM: - strcat( str, "SNORM" ); - break; - case PIPE_FORMAT_TYPE_USCALED: - strcat( str, "USCALED" ); - break; - case PIPE_FORMAT_TYPE_SSCALED: - strcat( str, "SSCALED" ); - break; - } - } - break; - case PIPE_FORMAT_LAYOUT_YCBCR: - { - pipe_format_ycbcr_t ycbcr = (pipe_format_ycbcr_t) format; - - strcat( str, "YCBCR" ); - if (pf_rev( ycbcr )) { - strcat( str, "_REV" ); - } - } - break; - } - return str; -} +extern char *pf_sprint_name( char *str, enum pipe_format format ); /** * Return bits for a particular component. -- cgit v1.2.3 From 7441f2e3ea77404064d65b604e1e525cacc453e9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 29 Apr 2008 20:42:35 +0100 Subject: gallium: add pack for A1R5B5G5 --- src/gallium/auxiliary/util/u_pack_color.h | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 7e26b1642a..7d4024e82c 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -89,6 +89,12 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, *d = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); } return; + case PIPE_FORMAT_A1R5G5B5_UNORM: + { + ushort *d = (ushort *) dest; + *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 3) | (b >> 3); + } + return; /* XXX lots more cases to add */ default: debug_print_format("gallium: unhandled format in util_pack_color_ub()", format); @@ -156,6 +162,12 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) *d = ((r & 0xf8) << 8) | ((g & 0xfc) << 3) | (b >> 3); } return; + case PIPE_FORMAT_A1R5G5B5_UNORM: + { + ushort *d = (ushort *) dest; + *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 3) | (b >> 3); + } + return; case PIPE_FORMAT_R32G32B32A32_FLOAT: { float *d = (float *) dest; -- cgit v1.2.3 From fbddc8097ce3a9d38a061105542875dbb9f909f7 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 29 Apr 2008 21:05:48 +0100 Subject: gallium: add pack for A4R4G4B4 --- src/gallium/auxiliary/util/u_pack_color.h | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 7d4024e82c..f641329d18 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -95,6 +95,12 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 3) | (b >> 3); } return; + case PIPE_FORMAT_A4R4G4B4_UNORM: + { + ushort *d = (ushort *) dest; + *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4); + } + return; /* XXX lots more cases to add */ default: debug_print_format("gallium: unhandled format in util_pack_color_ub()", format); @@ -168,6 +174,12 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 3) | (b >> 3); } return; + case PIPE_FORMAT_A4R4G4B4_UNORM: + { + ushort *d = (ushort *) dest; + *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4); + } + return; case PIPE_FORMAT_R32G32B32A32_FLOAT: { float *d = (float *) dest; -- cgit v1.2.3 From bbafa8aa2fc8009fb8e32f996d4972c56e6b46e6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 29 Apr 2008 21:30:04 +0100 Subject: gallium: fix pack for A1R5B5G5 --- src/gallium/auxiliary/util/u_pack_color.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index f641329d18..bb2c34e5f6 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -92,7 +92,7 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, case PIPE_FORMAT_A1R5G5B5_UNORM: { ushort *d = (ushort *) dest; - *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 3) | (b >> 3); + *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 2) | (b >> 3); } return; case PIPE_FORMAT_A4R4G4B4_UNORM: @@ -171,7 +171,7 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) case PIPE_FORMAT_A1R5G5B5_UNORM: { ushort *d = (ushort *) dest; - *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 3) | (b >> 3); + *d = ((a & 0x80) << 8) | ((r & 0xf8) << 7) | ((g & 0xf8) << 2) | (b >> 3); } return; case PIPE_FORMAT_A4R4G4B4_UNORM: -- cgit v1.2.3 From 1e97ab685689ef06181a5f22fae9a3a82c83142c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 29 Apr 2008 21:30:31 +0100 Subject: gallium: add pack for _FLOAT formats to pack_ub --- src/gallium/auxiliary/util/u_pack_color.h | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index bb2c34e5f6..0b917c005f 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -101,6 +101,24 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4); } return; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + { + float *d = (float *) dest; + d[0] = (float)r / 255.0f; + d[1] = (float)g / 255.0f; + d[2] = (float)b / 255.0f; + d[3] = (float)a / 255.0f; + } + return; + case PIPE_FORMAT_R32G32B32_FLOAT: + { + float *d = (float *) dest; + d[0] = (float)r / 255.0f; + d[1] = (float)g / 255.0f; + d[2] = (float)b / 255.0f; + } + return; + /* XXX lots more cases to add */ default: debug_print_format("gallium: unhandled format in util_pack_color_ub()", format); -- cgit v1.2.3 From fce5951b56a84304d0cb0dce4785237d90a71eb2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 29 Apr 2008 14:39:42 -0600 Subject: gallium: declare pipe_format enum to silence warnings --- src/gallium/include/pipe/p_debug.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index f87247994b..ed47c0e14c 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -60,6 +60,8 @@ extern "C" { #endif +enum pipe_format; + void _debug_vprintf(const char *format, va_list ap); -- cgit v1.2.3 From 0d80f407f128f1a324e9dc0db2d0910bf32ba736 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 29 Apr 2008 17:21:10 -0400 Subject: =?UTF-8?q?silence=20p=5Fdebug.h:63:=20warning:=20ISO=20C=20forbid?= =?UTF-8?q?s=20forward=20references=20to=20=E2=80=98enum=E2=80=99=20types?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- src/gallium/auxiliary/util/p_debug.c | 4 ++-- src/gallium/include/pipe/p_debug.h | 5 +---- 2 files changed, 3 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index eaf2c9bd98..8ef2880191 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -413,11 +413,11 @@ char *pf_sprint_name( char *str, enum pipe_format format ) } -void debug_print_format(const char *msg, enum pipe_format fmt ) +void debug_print_format(const char *msg, unsigned fmt ) { char fmtstr[80]; - pf_sprint_name(fmtstr, fmt); + pf_sprint_name(fmtstr, (enum pipe_format)fmt); debug_printf("%s: %s\n", msg, fmtstr); } diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index ed47c0e14c..7a7312ea12 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -59,9 +59,6 @@ extern "C" { #endif #endif - -enum pipe_format; - void _debug_vprintf(const char *format, va_list ap); @@ -114,7 +111,7 @@ void debug_print_blob( const char *name, const void *blob, unsigned size ); /* Print a message along with a prettified format string */ -void debug_print_format(const char *msg, enum pipe_format fmt ); +void debug_print_format(const char *msg, unsigned fmt ); #else #define debug_print_blob(_name, _blob, _size) ((void)0) #define debug_print_format(_msg, _fmt) ((void)0) -- cgit v1.2.3 From 6fb40092261b891b78e504f453fddf041f9efac6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 29 Apr 2008 17:29:33 -0600 Subject: gallium: flush pipe before updating bitmap texture Fixes duplicated text bug in progs/tests/texwrap.c --- src/mesa/state_tracker/st_cb_bitmap.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 6c2d9a4b89..836758a336 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -589,6 +589,11 @@ st_flush_bitmap_cache(struct st_context *st) pipe_surface_unmap(surf); pipe_surface_reference(&surf, NULL); + /* flush in case the previous texture contents haven't been + * used yet. XXX this is not ideal! Revisit. + */ + st->pipe->flush( st->pipe, 0x0, NULL ); + pipe->texture_update(pipe, cache->texture, 0, 0x1); draw_bitmap_quad(st->ctx, @@ -597,7 +602,6 @@ st_flush_bitmap_cache(struct st_context *st) st->ctx->Current.RasterPos[2], BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, cache->texture); - } reset_cache(st); } -- cgit v1.2.3 From 62a29412b90008a247fd3b61f1b882df2e5e81c1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 16:50:56 -0600 Subject: gallium: test for new PIPE_ARCH_X86 --- src/gallium/drivers/softpipe/sp_context.c | 2 +- src/gallium/drivers/softpipe/sp_fs_sse.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index f2ce0fdc6e..edf91ecafa 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -127,7 +127,7 @@ softpipe_create( struct pipe_screen *screen, struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); uint i; -#if defined(__i386__) || defined(__386__) +#ifdef PIPE_ARCH_X86 softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; #else softpipe->use_sse = FALSE; diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 4d569e1e22..25fdfea491 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -40,7 +40,7 @@ #include "tgsi/exec/tgsi_sse2.h" -#if defined(__i386__) || defined(__386__) +#ifdef PIPE_ARCH_X86 #include "rtasm/rtasm_x86sse.h" -- cgit v1.2.3 From d0b2561c57763db196b45ab052a01a12fc37b06f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Apr 2008 17:50:48 -0600 Subject: gallium: updated comment in xm_flush_frontbuffer() --- src/gallium/winsys/xlib/xm_winsys.c | 7 +++---- 1 file changed, 3 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 5a424d0ac7..8a89278cde 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -296,10 +296,9 @@ xm_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surf, void *context_private) { - /* The Xlib driver's front color surfaces are actually X Windows so - * this flush is a no-op. - * If we instead did front buffer rendering to a temporary XImage, - * this would be the place to copy the Ximage to the on-screen Window. + /* + * The front color buffer is actually just another XImage buffer. + * This function copies that XImage to the actual X Window. */ XMesaContext xmctx = (XMesaContext) context_private; xmesa_display_surface(xmctx->xm_buffer, surf); -- cgit v1.2.3 From 42fb48492e71016c5a2888cd3d2507a89dbd91f3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 29 Apr 2008 18:34:14 -0600 Subject: gallium: init hw_key with memset() to silence valgrind warnings We shouldn't be hashing with keys that have uninitialized memory. --- src/gallium/auxiliary/draw/draw_pt_emit.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 35707af8a8..f9ac16786e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -58,6 +58,8 @@ void draw_pt_emit_prepare( struct pt_emit *emit, return; } + memset(&hw_key, 0, sizeof(hw_key)); + /* Must do this after set_primitive() above: */ vinfo = draw->render->get_vertex_info(draw->render); -- cgit v1.2.3 From eefb9d3b51e40867ac3fee43b6496a62b1888b3c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 09:41:55 -0600 Subject: gallium: simplify texture format selection --- src/mesa/state_tracker/st_cb_bitmap.c | 57 +++++++++++------------------------ src/mesa/state_tracker/st_context.h | 1 + 2 files changed, 18 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 836758a336..45842fe231 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -230,27 +230,10 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, struct pipe_context *pipe = ctx->st->pipe; struct pipe_screen *screen = pipe->screen; struct pipe_surface *surface; - uint format = 0, cpp, comp; ubyte *dest; struct pipe_texture *pt; int row, col; - /* find a texture format we know */ - if (screen->is_format_supported( screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE )) { - format = PIPE_FORMAT_U_I8; - cpp = 1; - comp = 0; - } - else if (screen->is_format_supported( screen, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_TEXTURE )) { - format = PIPE_FORMAT_A8R8G8B8_UNORM; - cpp = 4; - comp = 3; /* alpha channel */ /*XXX little-endian dependency */ - } - else { - /* XXX support more formats */ - assert( 0 ); - } - /* PBO source... */ bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap); if (!bitmap) { @@ -260,8 +243,8 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, /** * Create texture to hold bitmap pattern. */ - pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, width, height, - 1, 0); + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, ctx->st->bitmap.tex_format, + 0, width, height, 1, 0); if (!pt) { _mesa_unmap_bitmap_pbo(ctx, unpack); return NULL; @@ -280,7 +263,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, for (row = 0; row < height; row++) { const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); - ubyte *destRow = dest + row * surface->pitch * cpp; + ubyte *destRow = dest + row * surface->pitch; if (unpack->LsbFirst) { /* Lsb first */ @@ -288,8 +271,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, for (col = 0; col < width; col++) { /* set texel to 255 if bit is set */ - destRow[comp] = (*src & mask) ? 0x0 : 0xff; - destRow += cpp; + destRow[col] = (*src & mask) ? 0x0 : 0xff; if (mask == 128U) { src++; @@ -310,8 +292,7 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, for (col = 0; col < width; col++) { /* set texel to 255 if bit is set */ - destRow[comp] =(*src & mask) ? 0x0 : 0xff; - destRow += cpp; + destRow[col] =(*src & mask) ? 0x0 : 0xff; if (mask == 1U) { src++; @@ -336,8 +317,6 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, pipe_surface_reference(&surface, NULL); pipe->texture_update(pipe, pt, 0, 0x1); - pt->format = format; - return pt; } @@ -530,25 +509,12 @@ reset_cache(struct st_context *st) static void init_bitmap_cache(struct st_context *st) { - struct pipe_context *pipe = st->pipe; - struct pipe_screen *screen = pipe->screen; - enum pipe_format format; - st->bitmap.cache = CALLOC_STRUCT(bitmap_cache); if (!st->bitmap.cache) return; - /* find a usable texture format */ - if (screen->is_format_supported(screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE)) { - format = PIPE_FORMAT_U_I8; - } - else { - /* XXX support more formats */ - assert(0); - } - st->bitmap.cache->texture - = st_texture_create(st, PIPE_TEXTURE_2D, format, 0, + = st_texture_create(st, PIPE_TEXTURE_2D, st->bitmap.tex_format, 0, BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, 1, 0); if (!st->bitmap.cache->texture) { FREE(st->bitmap.cache); @@ -768,6 +734,8 @@ void st_init_bitmap(struct st_context *st) { struct pipe_sampler_state *sampler = &st->bitmap.sampler; + struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; /* init sampler state once */ memset(sampler, 0, sizeof(*sampler)); @@ -784,6 +752,15 @@ st_init_bitmap(struct st_context *st) st->bitmap.rasterizer.gl_rasterization_rules = 1; st->bitmap.rasterizer.bypass_vs = 1; + /* find a usable texture format */ + if (screen->is_format_supported(screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE)) { + st->bitmap.tex_format = PIPE_FORMAT_U_I8; + } + else { + /* XXX support more formats */ + assert(0); + } + init_bitmap_cache(st); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 80a71d1766..1ca779d0a9 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -145,6 +145,7 @@ struct st_context struct pipe_rasterizer_state rasterizer; struct pipe_sampler_state sampler; struct pipe_shader_state vert_shader; + enum pipe_format tex_format; void *vs; float vertices[4][3][4]; /**< vertex pos + color + texcoord */ struct pipe_buffer *vbuf; -- cgit v1.2.3 From d3b98330d800d347e3fccf80a40fa87453659c6c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 10:00:44 -0600 Subject: gallium: create a new texture each time we init/flush the bitmap cache --- src/mesa/state_tracker/st_cb_bitmap.c | 80 ++++++++++++++++++----------------- 1 file changed, 42 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 45842fe231..e9d375dc58 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -90,10 +90,14 @@ struct bitmap_cache GLint xpos, ypos; /** Bounds of region used in window coords */ GLint xmin, ymin, xmax, ymax; + struct pipe_texture *texture; + struct pipe_surface *surf; + GLboolean empty; + /** An I8 texture image: */ - GLubyte buffer[BITMAP_CACHE_HEIGHT][BITMAP_CACHE_WIDTH]; + ubyte *buffer; }; @@ -496,33 +500,34 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, static void reset_cache(struct st_context *st) { - memset(st->bitmap.cache->buffer, 0xff, sizeof(st->bitmap.cache->buffer)); - st->bitmap.cache->empty = GL_TRUE; + struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; + struct bitmap_cache *cache = st->bitmap.cache; - st->bitmap.cache->xmin = 1000000; - st->bitmap.cache->xmax = -1000000; - st->bitmap.cache->ymin = 1000000; - st->bitmap.cache->ymax = -1000000; -} + //memset(cache->buffer, 0xff, sizeof(cache->buffer)); + cache->empty = GL_TRUE; + cache->xmin = 1000000; + cache->xmax = -1000000; + cache->ymin = 1000000; + cache->ymax = -1000000; -static void -init_bitmap_cache(struct st_context *st) -{ - st->bitmap.cache = CALLOC_STRUCT(bitmap_cache); - if (!st->bitmap.cache) - return; + assert(!cache->texture); - st->bitmap.cache->texture - = st_texture_create(st, PIPE_TEXTURE_2D, st->bitmap.tex_format, 0, - BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, 1, 0); - if (!st->bitmap.cache->texture) { - FREE(st->bitmap.cache); - st->bitmap.cache = NULL; - return; - } + /* allocate a new texture */ + cache->texture = st_texture_create(st, PIPE_TEXTURE_2D, + st->bitmap.tex_format, 0, + BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, + 1, 0); - reset_cache(st); + /* Map the texture surface. + * Subsequent glBitmap calls will write into the texture image. + */ + cache->surf = screen->get_tex_surface(screen, cache->texture, 0, 0, 0); + cache->buffer = pipe_surface_map(cache->surf); + + /* init image to all 0xff */ + memset(cache->buffer, 0xff, BITMAP_CACHE_WIDTH * BITMAP_CACHE_HEIGHT); } @@ -536,9 +541,6 @@ st_flush_bitmap_cache(struct st_context *st) if (st->ctx->DrawBuffer) { struct bitmap_cache *cache = st->bitmap.cache; struct pipe_context *pipe = st->pipe; - struct pipe_screen *screen = pipe->screen; - struct pipe_surface *surf; - void *dest; assert(cache->xmin <= cache->xmax); /* @@ -548,18 +550,13 @@ st_flush_bitmap_cache(struct st_context *st) cache->xpos, cache->ypos); */ - /* update the texture map image */ - surf = screen->get_tex_surface(screen, cache->texture, 0, 0, 0); - dest = pipe_surface_map(surf); - memcpy(dest, cache->buffer, sizeof(cache->buffer)); - pipe_surface_unmap(surf); - pipe_surface_reference(&surf, NULL); - - /* flush in case the previous texture contents haven't been - * used yet. XXX this is not ideal! Revisit. + /* The texture surface has been mapped until now. + * So unmap and release the texture surface before drawing. */ - st->pipe->flush( st->pipe, 0x0, NULL ); + pipe_surface_unmap(cache->surf); + pipe_surface_reference(&cache->surf, NULL); + /* XXX is this needed? */ pipe->texture_update(pipe, cache->texture, 0, 0x1); draw_bitmap_quad(st->ctx, @@ -568,6 +565,9 @@ st_flush_bitmap_cache(struct st_context *st) st->ctx->Current.RasterPos[2], BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, cache->texture); + + /* release/free the texture */ + pipe_texture_reference(&cache->texture, NULL); } reset_cache(st); } @@ -626,7 +626,7 @@ accum_bitmap(struct st_context *st, /* XXX try to combine this code with code in make_bitmap_texture() */ #define SET_PIXEL(COL, ROW) \ - cache->buffer[py + (ROW)][px + (COL)] = 0x0; + cache->buffer[(py + (ROW)) * BITMAP_CACHE_WIDTH + px + (COL)] = 0x0; for (row = 0; row < height; row++) { const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, @@ -716,6 +716,7 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, assert(pt->target == PIPE_TEXTURE_2D); draw_bitmap_quad(ctx, x, y, ctx->Current.RasterPos[2], width, height, pt); + /* release/free the texture */ pipe_texture_reference(&pt, NULL); } } @@ -761,7 +762,10 @@ st_init_bitmap(struct st_context *st) assert(0); } - init_bitmap_cache(st); + /* alloc bitmap cache object */ + st->bitmap.cache = CALLOC_STRUCT(bitmap_cache); + + reset_cache(st); } -- cgit v1.2.3 From 7146a1a29d3897fc0bd46dd56f3b36f2351d0f88 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 10:10:44 -0600 Subject: gallium: consolidate the bitmap->texel conversion code --- src/mesa/state_tracker/st_cb_bitmap.c | 163 ++++++++++++++-------------------- 1 file changed, 67 insertions(+), 96 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index e9d375dc58..d73cd4abfa 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -224,58 +224,37 @@ combined_bitmap_fragment_program(GLcontext *ctx) /** - * Create a texture which represents a bitmap image. + * Copy user-provide bitmap bits into texture buffer, expanding + * bits into texels. + * "On" bits will set texels to 0xff. + * "Off" bits will not modify texels. + * Note that the image is actually going to be upside down in + * the texture. We deal with that with texcoords. */ -static struct pipe_texture * -make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, - const struct gl_pixelstore_attrib *unpack, - const GLubyte *bitmap) +static void +unpack_bitmap(struct st_context *st, + GLint px, GLint py, GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap, + ubyte *destBuffer, uint destStride) { - struct pipe_context *pipe = ctx->st->pipe; - struct pipe_screen *screen = pipe->screen; - struct pipe_surface *surface; - ubyte *dest; - struct pipe_texture *pt; - int row, col; - - /* PBO source... */ - bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap); - if (!bitmap) { - return NULL; - } + GLint row, col; - /** - * Create texture to hold bitmap pattern. - */ - pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, ctx->st->bitmap.tex_format, - 0, width, height, 1, 0); - if (!pt) { - _mesa_unmap_bitmap_pbo(ctx, unpack); - return NULL; - } - - surface = screen->get_tex_surface(screen, pt, 0, 0, 0); - - /* map texture surface */ - dest = pipe_surface_map(surface); - - /* Put image into texture surface. - * Note that the image is actually going to be upside down in - * the texture. We deal with that with texcoords. - */ +#define SET_PIXEL(COL, ROW) \ + destBuffer[(py + (ROW)) * destStride + px + (COL)] = 0x0; for (row = 0; row < height; row++) { const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); - ubyte *destRow = dest + row * surface->pitch; if (unpack->LsbFirst) { /* Lsb first */ GLubyte mask = 1U << (unpack->SkipPixels & 0x7); for (col = 0; col < width; col++) { - /* set texel to 255 if bit is set */ - destRow[col] = (*src & mask) ? 0x0 : 0xff; + if (*src & mask) { + SET_PIXEL(col, row); + } if (mask == 128U) { src++; @@ -295,8 +274,9 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, GLubyte mask = 128U >> (unpack->SkipPixels & 0x7); for (col = 0; col < width; col++) { - /* set texel to 255 if bit is set */ - destRow[col] =(*src & mask) ? 0x0 : 0xff; + if (*src & mask) { + SET_PIXEL(col, row); + } if (mask == 1U) { src++; @@ -314,6 +294,50 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, } /* row */ +#undef SET_PIXEL +} + + +/** + * Create a texture which represents a bitmap image. + */ +static struct pipe_texture * +make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, + const struct gl_pixelstore_attrib *unpack, + const GLubyte *bitmap) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *surface; + ubyte *dest; + struct pipe_texture *pt; + + /* PBO source... */ + bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap); + if (!bitmap) { + return NULL; + } + + /** + * Create texture to hold bitmap pattern. + */ + pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, ctx->st->bitmap.tex_format, + 0, width, height, 1, 0); + if (!pt) { + _mesa_unmap_bitmap_pbo(ctx, unpack); + return NULL; + } + + surface = screen->get_tex_surface(screen, pt, 0, 0, 0); + + /* map texture surface */ + dest = pipe_surface_map(surface); + + /* Put image into texture surface */ + memset(dest, 0xff, height * surface->pitch); + unpack_bitmap(ctx->st, 0, 0, width, height, unpack, bitmap, + dest, surface->pitch); + _mesa_unmap_bitmap_pbo(ctx, unpack); /* Release surface */ @@ -585,7 +609,6 @@ accum_bitmap(struct st_context *st, const GLubyte *bitmap ) { struct bitmap_cache *cache = st->bitmap.cache; - int row, col; int px = -999, py; if (width > BITMAP_CACHE_WIDTH || @@ -624,60 +647,8 @@ accum_bitmap(struct st_context *st, if (y + height > cache->ymax) cache->ymax = y + height; - /* XXX try to combine this code with code in make_bitmap_texture() */ -#define SET_PIXEL(COL, ROW) \ - cache->buffer[(py + (ROW)) * BITMAP_CACHE_WIDTH + px + (COL)] = 0x0; - - for (row = 0; row < height; row++) { - const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack, - bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0); - - if (unpack->LsbFirst) { - /* Lsb first */ - GLubyte mask = 1U << (unpack->SkipPixels & 0x7); - for (col = 0; col < width; col++) { - - if (*src & mask) { - SET_PIXEL(col, row); - } - - if (mask == 128U) { - src++; - mask = 1U; - } - else { - mask = mask << 1; - } - } - - /* get ready for next row */ - if (mask != 1) - src++; - } - else { - /* Msb first */ - GLubyte mask = 128U >> (unpack->SkipPixels & 0x7); - for (col = 0; col < width; col++) { - - if (*src & mask) { - SET_PIXEL(col, row); - } - - if (mask == 1U) { - src++; - mask = 128U; - } - else { - mask = mask >> 1; - } - } - - /* get ready for next row */ - if (mask != 128) - src++; - } - - } /* row */ + unpack_bitmap(st, px, py, width, height, unpack, bitmap, + cache->buffer, BITMAP_CACHE_WIDTH); return GL_TRUE; /* accumulated */ } -- cgit v1.2.3 From 15318c8d8eb97cec8c8528cc91aaeab8858f33c6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 10:43:09 -0600 Subject: gallium: new pipe_buffer alloc/map/unmap/ref wrappers --- src/gallium/include/pipe/p_inlines.h | 50 ++++++++++++++++++++++++++++++++++++ 1 file changed, 50 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h index 274f76a383..8eb604e73f 100644 --- a/src/gallium/include/pipe/p_inlines.h +++ b/src/gallium/include/pipe/p_inlines.h @@ -129,6 +129,56 @@ pipe_texture_release(struct pipe_texture **ptr) } +/** + * Convenience wrappers for winsys buffer functions. + */ + +static INLINE struct pipe_buffer * +pipe_buffer_create( struct pipe_context *pipe, + unsigned alignment, unsigned usage, unsigned size ) +{ + return pipe->winsys->buffer_create(pipe->winsys, alignment, usage, size); +} + +static INLINE struct pipe_buffer * +pipe_user_buffer_create( struct pipe_context *pipe, void *ptr, unsigned size ) +{ + return pipe->winsys->user_buffer_create(pipe->winsys, ptr, size); +} + +static INLINE void +pipe_buffer_destroy( struct pipe_context *pipe, struct pipe_buffer *buf ) +{ + pipe->winsys->buffer_destroy(pipe->winsys, buf); +} + +static INLINE void * +pipe_buffer_map(struct pipe_context *pipe, + struct pipe_buffer *buf, + unsigned usage) +{ + return pipe->winsys->buffer_map(pipe->winsys, buf, usage); +} + +static INLINE void +pipe_buffer_unmap(struct pipe_context *pipe, + struct pipe_buffer *buf) +{ + pipe->winsys->buffer_unmap(pipe->winsys, buf); +} + +/* XXX when we're using this everywhere, get rid of + * pipe_buffer_reference() above. + */ +static INLINE void +pipe_reference_buffer(struct pipe_context *pipe, + struct pipe_buffer **ptr, + struct pipe_buffer *buf) +{ + pipe_buffer_reference(pipe->winsys, ptr, buf); +} + + #ifdef __cplusplus } #endif -- cgit v1.2.3 From 99fba5466bfd14c4e052041c0571821be529e762 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 10:43:59 -0600 Subject: gallium: use new buffer wrapper functions in p_inlines.h This allows us to remove most of the direct references to winsys in the state tracker. --- src/mesa/state_tracker/st_atom_constbuf.c | 14 ++++---- src/mesa/state_tracker/st_cb_bitmap.c | 13 +++---- src/mesa/state_tracker/st_cb_bufferobjects.c | 22 +++++------- src/mesa/state_tracker/st_cb_clear.c | 14 ++++---- src/mesa/state_tracker/st_cb_drawpixels.c | 13 +++---- src/mesa/state_tracker/st_cb_fbo.c | 2 +- src/mesa/state_tracker/st_context.c | 4 +-- src/mesa/state_tracker/st_draw.c | 54 ++++++++++++---------------- src/mesa/state_tracker/st_gen_mipmap.c | 14 ++++---- src/mesa/state_tracker/st_texture.c | 1 - 10 files changed, 62 insertions(+), 89 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index 03093579e1..2b659aebbc 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -36,7 +36,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/p_winsys.h" #include "pipe/p_inlines.h" #include "st_context.h" @@ -54,7 +53,7 @@ void st_upload_constants( struct st_context *st, struct gl_program_parameter_list *params, unsigned id) { - struct pipe_winsys *ws = st->pipe->winsys; + struct pipe_context *pipe = st->pipe; struct pipe_constant_buffer *cbuf = &st->state.constants[id]; assert(id == PIPE_SHADER_VERTEX || id == PIPE_SHADER_FRAGMENT); @@ -74,8 +73,8 @@ void st_upload_constants( struct st_context *st, /* We always need to get a new buffer, to keep the drivers simple and * avoid gratuitous rendering synchronization. */ - pipe_buffer_reference( ws, &cbuf->buffer, NULL ); - cbuf->buffer = ws->buffer_create( ws, 1, PIPE_BUFFER_USAGE_CONSTANT, + pipe_reference_buffer(pipe, &cbuf->buffer, NULL ); + cbuf->buffer = pipe_buffer_create(pipe, 1, PIPE_BUFFER_USAGE_CONSTANT, paramBytes ); if (0) @@ -87,9 +86,10 @@ void st_upload_constants( struct st_context *st, /* load Mesa constants into the constant buffer */ if (cbuf->buffer) { - memcpy(ws->buffer_map(ws, cbuf->buffer, PIPE_BUFFER_USAGE_CPU_WRITE), - params->ParameterValues, paramBytes); - ws->buffer_unmap(ws, cbuf->buffer); + void *map = pipe_buffer_map(pipe, cbuf->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE); + memcpy(map, params->ParameterValues, paramBytes); + pipe_buffer_unmap(pipe, cbuf->buffer); } cbuf->size = paramBytes; diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index d73cd4abfa..f0a3b75357 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -50,7 +50,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" #include "util/p_tile.h" #include "util/u_draw_quad.h" #include "util/u_simple_shaders.h" @@ -372,9 +371,8 @@ setup_bitmap_vertex_data(struct st_context *st, void *buf; if (!st->bitmap.vbuf) { - st->bitmap.vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_VERTEX, - sizeof(st->bitmap.vertices)); + st->bitmap.vbuf = pipe_buffer_create(pipe, 32, PIPE_BUFFER_USAGE_VERTEX, + sizeof(st->bitmap.vertices)); } /* Positions are in clip coords since we need to do clipping in case @@ -413,10 +411,9 @@ setup_bitmap_vertex_data(struct st_context *st, } /* put vertex data into vbuf */ - buf = pipe->winsys->buffer_map(pipe->winsys, st->bitmap.vbuf, - PIPE_BUFFER_USAGE_CPU_WRITE); + buf = pipe_buffer_map(pipe, st->bitmap.vbuf, PIPE_BUFFER_USAGE_CPU_WRITE); memcpy(buf, st->bitmap.vertices, sizeof(st->bitmap.vertices)); - pipe->winsys->buffer_unmap(pipe->winsys, st->bitmap.vbuf); + pipe_buffer_unmap(pipe, st->bitmap.vbuf); } @@ -761,7 +758,7 @@ st_destroy_bitmap(struct st_context *st) } if (st->bitmap.vbuf) { - pipe->winsys->buffer_destroy(pipe->winsys, st->bitmap.vbuf); + pipe_buffer_destroy(pipe, st->bitmap.vbuf); st->bitmap.vbuf = NULL; } diff --git a/src/mesa/state_tracker/st_cb_bufferobjects.c b/src/mesa/state_tracker/st_cb_bufferobjects.c index fa1254ff7c..af79aefa96 100644 --- a/src/mesa/state_tracker/st_cb_bufferobjects.c +++ b/src/mesa/state_tracker/st_cb_bufferobjects.c @@ -35,7 +35,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/p_winsys.h" #include "pipe/p_inlines.h" @@ -79,7 +78,7 @@ st_bufferobj_free(GLcontext *ctx, struct gl_buffer_object *obj) struct st_buffer_object *st_obj = st_buffer_object(obj); if (st_obj->buffer) - pipe_buffer_reference(pipe->winsys, &st_obj->buffer, NULL); + pipe_reference_buffer(pipe, &st_obj->buffer, NULL); free(st_obj); } @@ -106,10 +105,9 @@ st_bufferobj_subdata(GLcontext *ctx, if (offset >= st_obj->size || size > (st_obj->size - offset)) return; - map = pipe->winsys->buffer_map(pipe->winsys, st_obj->buffer, - PIPE_BUFFER_USAGE_CPU_WRITE); + map = pipe_buffer_map(pipe, st_obj->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); memcpy(map + offset, data, size); - pipe->winsys->buffer_unmap(pipe->winsys, st_obj->buffer); + pipe_buffer_unmap(pipe, st_obj->buffer); } @@ -130,10 +128,9 @@ st_bufferobj_get_subdata(GLcontext *ctx, if (offset >= st_obj->size || size > (st_obj->size - offset)) return; - map = pipe->winsys->buffer_map(pipe->winsys, st_obj->buffer, - PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pipe, st_obj->buffer, PIPE_BUFFER_USAGE_CPU_READ); memcpy(data, map + offset, size); - pipe->winsys->buffer_unmap(pipe->winsys, st_obj->buffer); + pipe_buffer_unmap(pipe, st_obj->buffer); } @@ -174,10 +171,9 @@ st_bufferobj_data(GLcontext *ctx, buffer_usage = 0; } - pipe_buffer_reference( pipe->winsys, &st_obj->buffer, NULL ); + pipe_reference_buffer( pipe, &st_obj->buffer, NULL ); - st_obj->buffer = pipe->winsys->buffer_create( pipe->winsys, 32, buffer_usage, - size ); + st_obj->buffer = pipe_buffer_create( pipe, 32, buffer_usage, size ); st_obj->size = size; @@ -211,7 +207,7 @@ st_bufferobj_map(GLcontext *ctx, GLenum target, GLenum access, break; } - obj->Pointer = pipe->winsys->buffer_map(pipe->winsys, st_obj->buffer, flags); + obj->Pointer = pipe_buffer_map(pipe, st_obj->buffer, flags); return obj->Pointer; } @@ -225,7 +221,7 @@ st_bufferobj_unmap(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj) struct pipe_context *pipe = st_context(ctx)->pipe; struct st_buffer_object *st_obj = st_buffer_object(obj); - pipe->winsys->buffer_unmap(pipe->winsys, st_obj->buffer); + pipe_buffer_unmap(pipe, st_obj->buffer); obj->Pointer = NULL; return GL_TRUE; } diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index 95a5fb8db4..fe979f10bd 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -45,9 +45,9 @@ #include "st_mesa_to_tgsi.h" #include "pipe/p_context.h" +#include "pipe/p_inlines.h" #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "pipe/p_winsys.h" #include "util/u_pack_color.h" #include "util/u_simple_shaders.h" #include "util/u_draw_quad.h" @@ -106,7 +106,7 @@ st_destroy_clear(struct st_context *st) st->clear.vs = NULL; } if (st->clear.vbuf) { - pipe->winsys->buffer_destroy(pipe->winsys, st->clear.vbuf); + pipe_buffer_destroy(pipe, st->clear.vbuf); st->clear.vbuf = NULL; } } @@ -142,9 +142,8 @@ draw_quad(GLcontext *ctx, void *buf; if (!st->clear.vbuf) { - st->clear.vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_VERTEX, - sizeof(st->clear.vertices)); + st->clear.vbuf = pipe_buffer_create(pipe, 32, PIPE_BUFFER_USAGE_VERTEX, + sizeof(st->clear.vertices)); } /* positions */ @@ -171,10 +170,9 @@ draw_quad(GLcontext *ctx, } /* put vertex data into vbuf */ - buf = pipe->winsys->buffer_map(pipe->winsys, st->clear.vbuf, - PIPE_BUFFER_USAGE_CPU_WRITE); + buf = pipe_buffer_map(pipe, st->clear.vbuf, PIPE_BUFFER_USAGE_CPU_WRITE); memcpy(buf, st->clear.vertices, sizeof(st->clear.vertices)); - pipe->winsys->buffer_unmap(pipe->winsys, st->clear.vbuf); + pipe_buffer_unmap(pipe, st->clear.vbuf); /* draw */ util_draw_vertex_buffer(pipe, st->clear.vbuf, diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 7597ea323c..65bfd6cfcc 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -55,7 +55,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" #include "util/p_tile.h" #include "util/u_draw_quad.h" #include "shader/prog_instruction.h" @@ -483,20 +482,18 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, ubyte *map; /* allocate/load buffer object with vertex data */ - buf = pipe->winsys->buffer_create(pipe->winsys, 32, - PIPE_BUFFER_USAGE_VERTEX, - sizeof(verts)); - map = pipe->winsys->buffer_map(pipe->winsys, buf, - PIPE_BUFFER_USAGE_CPU_WRITE); + buf = pipe_buffer_create(pipe,32, PIPE_BUFFER_USAGE_VERTEX, + sizeof(verts)); + map = pipe_buffer_map(pipe, buf, PIPE_BUFFER_USAGE_CPU_WRITE); memcpy(map, verts, sizeof(verts)); - pipe->winsys->buffer_unmap(pipe->winsys, buf); + pipe_buffer_unmap(pipe, buf); util_draw_vertex_buffer(pipe, buf, PIPE_PRIM_QUADS, 4, /* verts */ 3); /* attribs/vert */ - pipe->winsys->buffer_destroy(pipe->winsys, buf); + pipe_buffer_destroy(pipe, buf); } } diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 2d741d9390..fc8a5ea7f6 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -105,7 +105,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, } else if (strb->surface->buffer) { /* release/discard the old surface buffer */ - pipe_buffer_reference(pipe->winsys, &strb->surface->buffer, NULL); + pipe_reference_buffer(pipe, &strb->surface->buffer, NULL); } /* Determine surface format here */ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 8a30871fa0..c900064f2b 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -55,7 +55,6 @@ #include "st_gen_mipmap.h" #include "st_program.h" #include "pipe/p_context.h" -#include "pipe/p_winsys.h" #include "pipe/p_inlines.h" #include "draw/draw_context.h" #include "cso_cache/cso_cache.h" @@ -157,7 +156,6 @@ struct st_context *st_create_context(struct pipe_context *pipe, static void st_destroy_context_priv( struct st_context *st ) { - struct pipe_winsys *ws = st->pipe->winsys; uint i; draw_destroy(st->draw); @@ -172,7 +170,7 @@ static void st_destroy_context_priv( struct st_context *st ) for (i = 0; i < Elements(st->state.constants); i++) { if (st->state.constants[i].buffer) { - pipe_buffer_reference(ws, &st->state.constants[i].buffer, NULL); + pipe_reference_buffer(st->pipe, &st->state.constants[i].buffer, NULL); } } diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 6c20120ac7..0fe4d198bd 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -43,7 +43,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/p_winsys.h" #include "pipe/p_inlines.h" #include "draw/draw_private.h" @@ -219,8 +218,7 @@ setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count, if (!vec) return NULL; - map = pipe->winsys->buffer_map(pipe->winsys, stobj->buffer, - PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pipe, stobj->buffer, PIPE_BUFFER_USAGE_CPU_READ); map = ADD_POINTERS(map, array->Ptr); for (i = 0; i < count; i++) { @@ -230,7 +228,7 @@ setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count, map += array->StrideB; } - pipe->winsys->buffer_unmap(pipe->winsys, stobj->buffer); + pipe_buffer_unmap(pipe, stobj->buffer); pipe->set_edgeflags(pipe, vec); @@ -260,7 +258,6 @@ st_draw_vbo(GLcontext *ctx, GLuint max_index) { struct pipe_context *pipe = ctx->st->pipe; - struct pipe_winsys *winsys = pipe->winsys; const struct st_vertex_program *vp; const struct pipe_shader_state *vs; struct pipe_vertex_buffer vbuffer[PIPE_MAX_SHADER_INPUTS]; @@ -292,7 +289,7 @@ st_draw_vbo(GLcontext *ctx, assert(stobj->buffer); vbuffer[attr].buffer = NULL; - pipe_buffer_reference(winsys, &vbuffer[attr].buffer, stobj->buffer); + pipe_reference_buffer(pipe, &vbuffer[attr].buffer, stobj->buffer); vbuffer[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ velements[attr].src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; assert(velements[attr].src_offset <= 2048); /* 11-bit field */ @@ -310,9 +307,8 @@ st_draw_vbo(GLcontext *ctx, /* wrap user data */ vbuffer[attr].buffer - = winsys->user_buffer_create(winsys, - (void *) arrays[mesaAttr]->Ptr, - bytes); + = pipe_user_buffer_create(pipe, (void *) arrays[mesaAttr]->Ptr, + bytes); vbuffer[attr].buffer_offset = 0; velements[attr].src_offset = 0; } @@ -358,14 +354,13 @@ st_draw_vbo(GLcontext *ctx, if (bufobj && bufobj->Name) { /* elements/indexes are in a real VBO */ struct st_buffer_object *stobj = st_buffer_object(bufobj); - pipe_buffer_reference(winsys, &indexBuf, stobj->buffer); + pipe_reference_buffer(pipe, &indexBuf, stobj->buffer); indexOffset = (unsigned) ib->ptr / indexSize; } else { /* element/indicies are in user space memory */ - indexBuf = winsys->user_buffer_create(winsys, - (void *) ib->ptr, - ib->count * indexSize); + indexBuf = pipe_user_buffer_create(pipe, (void *) ib->ptr, + ib->count * indexSize); indexOffset = 0; } @@ -380,7 +375,7 @@ st_draw_vbo(GLcontext *ctx, prims[i].start + indexOffset, prims[i].count); } - pipe_buffer_reference(winsys, &indexBuf, NULL); + pipe_reference_buffer(pipe, &indexBuf, NULL); } else { /* non-indexed */ @@ -396,7 +391,7 @@ st_draw_vbo(GLcontext *ctx, /* unreference buffers (frees wrapped user-space buffer objects) */ for (attr = 0; attr < vp->num_inputs; attr++) { - pipe_buffer_reference(winsys, &vbuffer[attr].buffer, NULL); + pipe_reference_buffer(pipe, &vbuffer[attr].buffer, NULL); assert(!vbuffer[attr].buffer); } pipe->set_vertex_buffers(pipe, vp->num_inputs, vbuffer); @@ -458,7 +453,6 @@ st_feedback_draw_vbo(GLcontext *ctx, struct st_context *st = ctx->st; struct pipe_context *pipe = st->pipe; struct draw_context *draw = st->draw; - struct pipe_winsys *winsys = pipe->winsys; const struct st_vertex_program *vp; const struct pipe_shader_state *vs; struct pipe_buffer *index_buffer_handle = 0; @@ -509,7 +503,7 @@ st_feedback_draw_vbo(GLcontext *ctx, assert(stobj->buffer); vbuffers[attr].buffer = NULL; - pipe_buffer_reference(winsys, &vbuffers[attr].buffer, stobj->buffer); + pipe_reference_buffer(pipe, &vbuffers[attr].buffer, stobj->buffer); vbuffers[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ velements[attr].src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; } @@ -521,9 +515,8 @@ st_feedback_draw_vbo(GLcontext *ctx, /* wrap user data */ vbuffers[attr].buffer - = winsys->user_buffer_create(winsys, - (void *) arrays[mesaAttr]->Ptr, - bytes); + = pipe_user_buffer_create(pipe, (void *) arrays[mesaAttr]->Ptr, + bytes); vbuffers[attr].buffer_offset = 0; velements[attr].src_offset = 0; } @@ -544,9 +537,8 @@ st_feedback_draw_vbo(GLcontext *ctx, #endif /* map the attrib buffer */ - map = pipe->winsys->buffer_map(pipe->winsys, - vbuffers[attr].buffer, - PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pipe, vbuffers[attr].buffer, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, attr, map); } @@ -572,9 +564,8 @@ st_feedback_draw_vbo(GLcontext *ctx, return; } - map = pipe->winsys->buffer_map(pipe->winsys, - index_buffer_handle, - PIPE_BUFFER_USAGE_CPU_READ); + map = pipe_buffer_map(pipe, index_buffer_handle, + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(draw, indexSize, map); } else { @@ -584,7 +575,7 @@ st_feedback_draw_vbo(GLcontext *ctx, /* map constant buffers */ - mapped_constants = winsys->buffer_map(winsys, + mapped_constants = pipe_buffer_map(pipe, st->state.constants[PIPE_SHADER_VERTEX].buffer, PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_constant_buffer(st->draw, mapped_constants); @@ -597,21 +588,20 @@ st_feedback_draw_vbo(GLcontext *ctx, /* unmap constant buffers */ - winsys->buffer_unmap(winsys, st->state.constants[PIPE_SHADER_VERTEX].buffer); + pipe_buffer_unmap(pipe, st->state.constants[PIPE_SHADER_VERTEX].buffer); /* * unmap vertex/index buffers */ for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { if (draw->pt.vertex_buffer[i].buffer) { - pipe->winsys->buffer_unmap(pipe->winsys, - draw->pt.vertex_buffer[i].buffer); - pipe_buffer_reference(winsys, &draw->pt.vertex_buffer[i].buffer, NULL); + pipe_buffer_unmap(pipe, draw->pt.vertex_buffer[i].buffer); + pipe_reference_buffer(pipe, &draw->pt.vertex_buffer[i].buffer, NULL); draw_set_mapped_vertex_buffer(draw, i, NULL); } } if (ib) { - pipe->winsys->buffer_unmap(pipe->winsys, index_buffer_handle); + pipe_buffer_unmap(pipe, index_buffer_handle); draw_set_mapped_element_buffer(draw, 0, NULL); } } diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index da9ec12a4d..1a0e19c2f9 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -36,7 +36,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" #include "util/u_gen_mipmap.h" #include "cso_cache/cso_cache.h" @@ -105,7 +104,6 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, { struct pipe_context *pipe = ctx->st->pipe; struct pipe_screen *screen = pipe->screen; - struct pipe_winsys *ws = pipe->winsys; struct pipe_texture *pt = st_get_texobj_texture(texObj); const uint baseLevel = texObj->BaseLevel; const uint lastLevel = pt->last_level; @@ -128,11 +126,11 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice); dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); - srcData = (ubyte *) ws->buffer_map(ws, srcSurf->buffer, - PIPE_BUFFER_USAGE_CPU_READ) + srcData = (ubyte *) pipe_buffer_map(pipe, srcSurf->buffer, + PIPE_BUFFER_USAGE_CPU_READ) + srcSurf->offset; - dstData = (ubyte *) ws->buffer_map(ws, dstSurf->buffer, - PIPE_BUFFER_USAGE_CPU_WRITE) + dstData = (ubyte *) pipe_buffer_map(pipe, dstSurf->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE) + dstSurf->offset; _mesa_generate_mipmap_level(target, datatype, comps, @@ -144,8 +142,8 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, dstSurf->pitch * dstSurf->cpp, /* stride in bytes */ dstData); - ws->buffer_unmap(ws, srcSurf->buffer); - ws->buffer_unmap(ws, dstSurf->buffer); + pipe_buffer_unmap(pipe, srcSurf->buffer); + pipe_buffer_unmap(pipe, dstSurf->buffer); pipe_surface_reference(&srcSurf, NULL); pipe_surface_reference(&dstSurf, NULL); diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 66d81e2b95..f68bef1207 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -38,7 +38,6 @@ #include "pipe/p_inlines.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "pipe/p_winsys.h" #define DBG if(0) printf -- cgit v1.2.3 From cafb545721e6c9479c07d3a7a891236e006d3376 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 1 May 2008 00:59:12 +0900 Subject: d3d: Windows miniport driver portability fixes. --- src/gallium/auxiliary/util/p_debug.c | 2 +- src/gallium/auxiliary/util/p_debug_mem.c | 7 +++- src/gallium/auxiliary/util/u_time.c | 6 +-- src/gallium/include/pipe/p_config.h | 9 +++++ src/gallium/include/pipe/p_util.h | 64 +++++++++++++++++--------------- 5 files changed, 53 insertions(+), 35 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 8ef2880191..56eecaab62 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -104,7 +104,7 @@ void _debug_break(void) __asm("int3"); #elif (defined(__i386__) || defined(__386__)) && defined(__MSC__) _asm {int 3}; -#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && !defined(WINCE) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) EngDebugBreak(); #else abort(); diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c index 9321cf71bb..3b5e4fbaee 100644 --- a/src/gallium/auxiliary/util/p_debug_mem.c +++ b/src/gallium/auxiliary/util/p_debug_mem.c @@ -34,9 +34,11 @@ #include "pipe/p_config.h" -#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) #include #include +#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) +#include #else #include #include @@ -52,6 +54,9 @@ #if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && !defined(WINCE) #define real_malloc(_size) EngAllocMem(0, _size, 'D3AG') #define real_free(_ptr) EngFreeMem(_ptr) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) +#define real_malloc(_size) ExAllocatePool(0, _size) +#define real_free(_ptr) ExFreePool(_ptr) #else #define real_malloc(_size) malloc(_size) #define real_free(_ptr) free(_ptr) diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index dd28ff4134..9b97050d51 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -37,11 +37,11 @@ #if defined(PIPE_OS_LINUX) #include -#elif defined(PIPE_OS_WINDOWS) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) #include -#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) #include -#endif +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) +#include #else #error Unsupported OS #endif diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h index e44fafeae9..5c030bdfff 100644 --- a/src/gallium/include/pipe/p_config.h +++ b/src/gallium/include/pipe/p_config.h @@ -106,7 +106,16 @@ #if defined(PIPE_SUBSYSTEM_KERNEL) #define PIPE_SUBSYSTEM_WINDOWS_DISPLAY #endif +#if 0 /* FIXME */ +#define PIPE_SUBSYSTEM_WINDOWS_MINIPORT +#endif +#if 0 /* FIXME */ +#define PIPE_SUBSYSTEM_WINDOWS_CE +#endif +#if defined(PIPE_SUBSYSTEM_USER) +#define PIPE_SUBSYSTEM_WINDOWS_USER #endif +#endif /* PIPE_OS_WINDOWS */ #endif /* P_CONFIG_H_ */ diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 63301ae3aa..0e7e246666 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -41,7 +41,9 @@ extern "C" { #endif -#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && defined(DEBUG) /* memory debugging */ +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && defined(DEBUG) + +/* memory debugging */ #include "p_debug.h" @@ -54,9 +56,7 @@ extern "C" { #define REALLOC( _ptr, _old_size, _size ) \ debug_realloc( __FILE__, __LINE__, __FUNCTION__, _ptr, _old_size, _size ) -#else - -#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) void * __stdcall EngAllocMem( @@ -68,17 +68,33 @@ void __stdcall EngFreeMem( void *Mem ); -static INLINE void * -MALLOC( unsigned size ) -{ -#ifdef WINCE - /* TODO: Need to abstract this */ - return malloc( size ); +#define MALLOC( _size ) EngAllocMem( 0, _size, 'D3AG' ) +#define _FREE( _ptr ) EngFreeMem( _ptr ) + +#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) + +void * +ExAllocatePool( + unsigned long PoolType, + size_t NumberOfBytes); + +void +ExFreePool(void *P); + +#define MALLOC(_size) ExAllocatePool(0, _size) +#define _FREE(_ptr) ExFreePool(_ptr) + #else - return EngAllocMem( 0, size, 'D3AG' ); + +#define MALLOC( SIZE ) malloc( SIZE ) +#define CALLOC( COUNT, SIZE ) calloc( COUNT, SIZE ) +#define FREE( PTR ) free( PTR ) +#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) + #endif -} + +#ifndef CALLOC static INLINE void * CALLOC( unsigned count, unsigned size ) { @@ -88,20 +104,19 @@ CALLOC( unsigned count, unsigned size ) } return ptr; } +#endif /* !CALLOC */ +#ifndef FREE static INLINE void FREE( void *ptr ) { if( ptr ) { -#ifdef WINCE - /* TODO: Need to abstract this */ - free( ptr ); -#else - EngFreeMem( ptr ); -#endif + _FREE( ptr ); } } +#endif /* !FREE */ +#ifndef REALLOC static INLINE void * REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) { @@ -118,19 +133,8 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) FREE( old_ptr ); return new_ptr; } +#endif /* !REALLOC */ -#else /* !PIPE_SUBSYSTEM_WINDOWS_DISPLAY */ - -#define MALLOC( SIZE ) malloc( SIZE ) - -#define CALLOC( COUNT, SIZE ) calloc( COUNT, SIZE ) - -#define FREE( PTR ) free( PTR ) - -#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) - -#endif /* !PIPE_SUBSYSTEM_WINDOWS_DISPLAY */ -#endif /* !DEBUG */ #define MALLOC_STRUCT(T) (struct T *) MALLOC(sizeof(struct T)) -- cgit v1.2.3 From dadb11f5d96354975f1c9f7c8134bb4727129249 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 1 May 2008 03:43:19 +0900 Subject: gallium: Fix release build. --- src/gallium/auxiliary/util/p_debug.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 56eecaab62..f1fb07bf5b 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -413,6 +413,7 @@ char *pf_sprint_name( char *str, enum pipe_format format ) } +#ifdef DEBUG void debug_print_format(const char *msg, unsigned fmt ) { char fmtstr[80]; @@ -421,3 +422,4 @@ void debug_print_format(const char *msg, unsigned fmt ) debug_printf("%s: %s\n", msg, fmtstr); } +#endif \ No newline at end of file -- cgit v1.2.3 From 026e31a068981724fb0c98f6d1fc87d086fd2da6 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 30 Apr 2008 02:04:28 -0400 Subject: try to fix the deleting of the bound cso during cleanup of the hash --- src/gallium/auxiliary/cso_cache/cso_cache.c | 40 ++++++-- src/gallium/auxiliary/cso_cache/cso_cache.h | 28 ++++-- src/gallium/auxiliary/cso_cache/cso_context.c | 128 ++++++++++++++++++++++++++ 3 files changed, 178 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 096875807b..36dc46ff80 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -43,6 +43,9 @@ struct cso_cache { struct cso_hash *rasterizer_hash; struct cso_hash *sampler_hash; int max_size; + + cso_sanitize_callback sanitize_cb; + void *sanitize_data; }; #if 1 @@ -205,8 +208,19 @@ static INLINE void delete_cso(void *state, enum cso_cache_type type) } } -static INLINE void sanitize_hash(struct cso_hash *hash, enum cso_cache_type type, + +static INLINE void sanitize_hash(struct cso_cache *sc, + struct cso_hash *hash, + enum cso_cache_type type, int max_size) +{ + if (sc->sanitize_cb) + sc->sanitize_cb(hash, type, max_size, sc->sanitize_data); +} + + +static INLINE void sanitize_cb(struct cso_hash *hash, enum cso_cache_type type, + int max_size, void *user_data) { /* if we're approach the maximum size, remove fourth of the entries * otherwise every subsequent call will go through the same */ @@ -231,7 +245,7 @@ cso_insert_state(struct cso_cache *sc, void *state) { struct cso_hash *hash = _cso_hash_for_type(sc, type); - sanitize_hash(hash, type, sc->max_size); + sanitize_hash(sc, hash, type, sc->max_size); return cso_hash_insert(hash, hash_key, state); } @@ -300,6 +314,8 @@ struct cso_cache *cso_cache_create(void) sc->rasterizer_hash = cso_hash_create(); sc->fs_hash = cso_hash_create(); sc->vs_hash = cso_hash_create(); + sc->sanitize_cb = sanitize_cb; + sc->sanitize_data = 0; return sc; } @@ -365,13 +381,13 @@ void cso_set_maximum_cache_size(struct cso_cache *sc, int number) { sc->max_size = number; - sanitize_hash(sc->blend_hash, CSO_BLEND, sc->max_size); - sanitize_hash(sc->depth_stencil_hash, CSO_DEPTH_STENCIL_ALPHA, + sanitize_hash(sc, sc->blend_hash, CSO_BLEND, sc->max_size); + sanitize_hash(sc, sc->depth_stencil_hash, CSO_DEPTH_STENCIL_ALPHA, sc->max_size); - sanitize_hash(sc->fs_hash, CSO_FRAGMENT_SHADER, sc->max_size); - sanitize_hash(sc->vs_hash, CSO_VERTEX_SHADER, sc->max_size); - sanitize_hash(sc->rasterizer_hash, CSO_RASTERIZER, sc->max_size); - sanitize_hash(sc->sampler_hash, CSO_SAMPLER, sc->max_size); + sanitize_hash(sc, sc->fs_hash, CSO_FRAGMENT_SHADER, sc->max_size); + sanitize_hash(sc, sc->vs_hash, CSO_VERTEX_SHADER, sc->max_size); + sanitize_hash(sc, sc->rasterizer_hash, CSO_RASTERIZER, sc->max_size); + sanitize_hash(sc, sc->sampler_hash, CSO_SAMPLER, sc->max_size); } int cso_maximum_cache_size(const struct cso_cache *sc) @@ -379,3 +395,11 @@ int cso_maximum_cache_size(const struct cso_cache *sc) return sc->max_size; } +void cso_cache_set_sanitize_callback(struct cso_cache *sc, + cso_sanitize_callback cb, + void *user_data) +{ + sc->sanitize_cb = cb; + sc->sanitize_data = user_data; +} + diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.h b/src/gallium/auxiliary/cso_cache/cso_cache.h index e5edbbb556..6b5c230e8f 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.h +++ b/src/gallium/auxiliary/cso_cache/cso_cache.h @@ -84,8 +84,22 @@ extern "C" { #endif +enum cso_cache_type { + CSO_BLEND, + CSO_SAMPLER, + CSO_DEPTH_STENCIL_ALPHA, + CSO_RASTERIZER, + CSO_FRAGMENT_SHADER, + CSO_VERTEX_SHADER +}; + typedef void (*cso_state_callback)(void *ctx, void *obj); +typedef void (*cso_sanitize_callback)(struct cso_hash *hash, + enum cso_cache_type type, + int max_size, + void *user_data); + struct cso_cache; struct cso_blend { @@ -130,21 +144,15 @@ struct cso_sampler { struct pipe_context *context; }; - -enum cso_cache_type { - CSO_BLEND, - CSO_SAMPLER, - CSO_DEPTH_STENCIL_ALPHA, - CSO_RASTERIZER, - CSO_FRAGMENT_SHADER, - CSO_VERTEX_SHADER -}; - unsigned cso_construct_key(void *item, int item_size); struct cso_cache *cso_cache_create(void); void cso_cache_delete(struct cso_cache *sc); +void cso_cache_set_sanitize_callback(struct cso_cache *sc, + cso_sanitize_callback cb, + void *user_data); + struct cso_hash_iter cso_insert_state(struct cso_cache *sc, unsigned hash_key, enum cso_cache_type type, void *state); diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index b4609e999b..eef898f486 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -80,6 +80,131 @@ struct cso_context { }; +static boolean delete_blend_state(struct cso_context *ctx, void *state) +{ + struct cso_blend *cso = (struct cso_blend *)state; + + if (ctx->blend == state) + return FALSE; + + if (cso->delete_state) + cso->delete_state(cso->context, cso->data); + FREE(state); + return TRUE; +} + +static boolean delete_depth_stencil_state(struct cso_context *ctx, void *state) +{ + struct cso_depth_stencil_alpha *cso = (struct cso_depth_stencil_alpha *)state; + + if (ctx->depth_stencil == cso->data) + return FALSE; + + if (cso->delete_state) + cso->delete_state(cso->context, cso->data); + FREE(state); + + return TRUE; +} + +static boolean delete_sampler_state(struct cso_context *ctx, void *state) +{ + struct cso_sampler *cso = (struct cso_sampler *)state; + if (cso->delete_state) + cso->delete_state(cso->context, cso->data); + FREE(state); + return TRUE; +} + +static boolean delete_rasterizer_state(struct cso_context *ctx, void *state) +{ + struct cso_rasterizer *cso = (struct cso_rasterizer *)state; + + if (ctx->rasterizer == cso->data) + return FALSE; + if (cso->delete_state) + cso->delete_state(cso->context, cso->data); + FREE(state); + return TRUE; +} + +static boolean delete_fs_state(struct cso_context *ctx, void *state) +{ + struct cso_fragment_shader *cso = (struct cso_fragment_shader *)state; + if (ctx->fragment_shader == cso->data) + return FALSE; + if (cso->delete_state) + cso->delete_state(cso->context, cso->data); + FREE(state); + return TRUE; +} + +static boolean delete_vs_state(struct cso_context *ctx, void *state) +{ + struct cso_vertex_shader *cso = (struct cso_vertex_shader *)state; + if (ctx->vertex_shader == cso->data) + return TRUE; + if (cso->delete_state) + cso->delete_state(cso->context, cso->data); + FREE(state); + return FALSE; +} + + +static INLINE boolean delete_cso(struct cso_context *ctx, + void *state, enum cso_cache_type type) +{ + switch (type) { + case CSO_BLEND: + return delete_blend_state(ctx, state); + break; + case CSO_SAMPLER: + return delete_sampler_state(ctx, state); + break; + case CSO_DEPTH_STENCIL_ALPHA: + return delete_depth_stencil_state(ctx, state); + break; + case CSO_RASTERIZER: + return delete_rasterizer_state(ctx, state); + break; + case CSO_FRAGMENT_SHADER: + return delete_fs_state(ctx, state); + break; + case CSO_VERTEX_SHADER: + return delete_vs_state(ctx, state); + break; + default: + assert(0); + FREE(state); + } + return FALSE; +} + +static INLINE void sanitize_hash(struct cso_hash *hash, enum cso_cache_type type, + int max_size, void *user_data) +{ + struct cso_context *ctx = (struct cso_context *)user_data; + /* if we're approach the maximum size, remove fourth of the entries + * otherwise every subsequent call will go through the same */ + int hash_size = cso_hash_size(hash); + int max_entries = (max_size > hash_size) ? max_size : hash_size; + int to_remove = (max_size < max_entries) * max_entries/4; + struct cso_hash_iter iter = cso_hash_first_node(hash); + if (hash_size > max_size) + to_remove += hash_size - max_size; + while (to_remove) { + /*remove elements until we're good */ + /*fixme: currently we pick the nodes to remove at random*/ + void *cso = cso_hash_iter_data(iter); + if (delete_cso(ctx, cso, type)) { + iter = cso_hash_erase(hash, iter); + --to_remove; + } else + iter = cso_hash_iter_next(iter); + } +} + + struct cso_context *cso_create_context( struct pipe_context *pipe ) { struct cso_context *ctx = CALLOC_STRUCT(cso_context); @@ -89,6 +214,9 @@ struct cso_context *cso_create_context( struct pipe_context *pipe ) ctx->cache = cso_cache_create(); if (ctx->cache == NULL) goto out; + cso_cache_set_sanitize_callback(ctx->cache, + sanitize_hash, + ctx); ctx->pipe = pipe; -- cgit v1.2.3 From 653da2d0698d18a8d3dcad1b1590437dee7bb403 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 30 Apr 2008 14:05:49 -0400 Subject: plug a memleak, destroy setup context --- src/gallium/drivers/softpipe/sp_prim_setup.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index feb35d492a..1cf9ffa632 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -150,6 +150,8 @@ static void reset_stipple_counter( struct draw_stage *stage ) static void render_destroy( struct draw_stage *stage ) { + struct setup_stage *ssetup = setup_stage(stage); + setup_destroy_context(ssetup->setup); FREE( stage ); } -- cgit v1.2.3 From bd5b99d3890f162ec0928a589b6afbc8e67580b7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 13:24:35 -0600 Subject: gallium: fix typo: s/custam/custom/ --- src/gallium/include/pipe/p_defines.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index 586951d956..d5d341d689 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -192,7 +192,7 @@ enum pipe_texture_target { #define PIPE_BUFFER_USAGE_VERTEX (1 << 5) #define PIPE_BUFFER_USAGE_INDEX (1 << 6) #define PIPE_BUFFER_USAGE_CONSTANT (1 << 7) -/** Pipe driver custam usage flags should be greater or equal to this value */ +/** Pipe driver custom usage flags should be greater or equal to this value */ #define PIPE_BUFFER_USAGE_CUSTOM (1 << 16) -- cgit v1.2.3 From c32477742facd06c22befcd300e9fdfeafb6995b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 16:05:01 -0600 Subject: Add support for GL_REPLACE_EXT texture env mode. GL_REPLACE_EXT comes from the ancient GL_EXT_texture extension. Found an old demo that actually uses it. The values of the GL_REPLACE and GL_REPLACE_EXT tokens is different, unfortunately. --- src/mesa/main/texstate.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c index cb7da39b51..84acfbd92c 100644 --- a/src/mesa/main/texstate.c +++ b/src/mesa/main/texstate.c @@ -238,6 +238,9 @@ calculate_derived_texenv( struct gl_tex_env_combine_state *state, return; } + if (mode == GL_REPLACE_EXT) + mode = GL_REPLACE; + switch (mode) { case GL_REPLACE: case GL_MODULATE: @@ -340,7 +343,9 @@ _mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param ) switch (pname) { case GL_TEXTURE_ENV_MODE: { - const GLenum mode = (GLenum) (GLint) *param; + GLenum mode = (GLenum) (GLint) *param; + if (mode == GL_REPLACE_EXT) + mode = GL_REPLACE; if (texUnit->EnvMode == mode) return; if (mode == GL_MODULATE || -- cgit v1.2.3 From 18f4d962653cdbb76f08e8498fbbec6e1759f21e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 16:41:53 -0600 Subject: gallium: replace old PIPE_FORMAT_U_S8 with PIPE_FORMAT_S8_UNORM --- src/gallium/drivers/softpipe/sp_quad_stencil.c | 4 ++-- src/gallium/drivers/softpipe/sp_tile_cache.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_stencil.c b/src/gallium/drivers/softpipe/sp_quad_stencil.c index 92a0da0083..b4c7e942fa 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stencil.c +++ b/src/gallium/drivers/softpipe/sp_quad_stencil.c @@ -243,7 +243,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) stencilVals[j] = tile->data.depth32[y][x] & 0xff; } break; - case PIPE_FORMAT_U_S8: + case PIPE_FORMAT_S8_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); @@ -311,7 +311,7 @@ stencil_test_quad(struct quad_stage *qs, struct quad_header *quad) tile->data.depth32[y][x] = z24s8; } break; - case PIPE_FORMAT_U_S8: + case PIPE_FORMAT_S8_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index edafd93d8b..a88aad5d09 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -168,7 +168,7 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || ps->format == PIPE_FORMAT_Z16_UNORM || ps->format == PIPE_FORMAT_Z32_UNORM || - ps->format == PIPE_FORMAT_U_S8); + ps->format == PIPE_FORMAT_S8_UNORM); } } -- cgit v1.2.3 From 8d45576ec5110d39fe48c2f65b04db16ce845968 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 16:49:38 -0600 Subject: gallium: rename old PIPE_FORMAT_U_I8 with PIPE_FORMAT_I8_UNORM, etc. Now all the packed color types are consistantly named. Added temporary #defines for the old names until all drivers are updated. --- src/gallium/include/pipe/p_format.h | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index bc23fe142e..e32212d36c 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -238,10 +238,10 @@ enum pipe_format { PIPE_FORMAT_A1R5G5B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_A4R4G4B4_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB1, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_U_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ - PIPE_FORMAT_U_A8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 0, 0, 0, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ - PIPE_FORMAT_U_I8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ - PIPE_FORMAT_U_A8_L8 = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha, luminance */ + PIPE_FORMAT_L8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ + PIPE_FORMAT_A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 0, 0, 0, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ + PIPE_FORMAT_I8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ + PIPE_FORMAT_A8L8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha, luminance */ PIPE_FORMAT_YCBCR = _PIPE_FORMAT_YCBCR( 0 ), PIPE_FORMAT_YCBCR_REV = _PIPE_FORMAT_YCBCR( 1 ), PIPE_FORMAT_Z16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), @@ -327,11 +327,12 @@ enum pipe_format { }; -/** - * Unsigned 8-bit stencil format. - * XXX should remove this, but S8_UNORM is a poor name - */ -#define PIPE_FORMAT_U_S8 PIPE_FORMAT_S8_UNORM +/** XXX remove these deprecated names */ +#define PIPE_FORMAT_U_L8 PIPE_FORMAT_L8_UNORM +#define PIPE_FORMAT_U_A8 PIPE_FORMAT_A8_UNORM +#define PIPE_FORMAT_U_I8 PIPE_FORMAT_I8_UNORM +#define PIPE_FORMAT_U_A8_L8 PIPE_FORMAT_A8L8_UNORM +#define PIPE_FORMAT_U_S8 PIPE_FORMAT_S8_UNORM /** -- cgit v1.2.3 From 54f94a790e4488445347abcff9a636a9c440d7f9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 16:50:17 -0600 Subject: gallium: use the newer PIPE_FORMAT_x_UNORM format names --- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 2 +- src/gallium/auxiliary/util/p_tile.c | 24 ++++++++-------- src/gallium/auxiliary/util/u_gen_mipmap.c | 8 +++--- src/gallium/drivers/i915simple/i915_screen.c | 8 +++--- .../drivers/i915simple/i915_state_sampler.c | 8 +++--- src/gallium/drivers/i965simple/brw_screen.c | 24 ++++++++-------- .../drivers/i965simple/brw_wm_surface_state.c | 8 +++--- src/mesa/state_tracker/st_cb_bitmap.c | 4 +-- src/mesa/state_tracker/st_format.c | 32 +++++++++++----------- 10 files changed, 60 insertions(+), 60 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index f05641dee6..f501b2aed4 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -392,7 +392,7 @@ aaline_create_texture(struct aaline_stage *aaline) memset(&texTemp, 0, sizeof(texTemp)); texTemp.target = PIPE_TEXTURE_2D; - texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */ + texTemp.format = PIPE_FORMAT_A8_UNORM; /* XXX verify supported by driver! */ texTemp.last_level = MAX_TEXTURE_LEVEL; texTemp.width[0] = 1 << MAX_TEXTURE_LEVEL; texTemp.height[0] = 1 << MAX_TEXTURE_LEVEL; diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index d1d63d73be..c4de9d2698 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -417,7 +417,7 @@ pstip_create_texture(struct pstip_stage *pstip) memset(&texTemp, 0, sizeof(texTemp)); texTemp.target = PIPE_TEXTURE_2D; - texTemp.format = PIPE_FORMAT_U_A8; /* XXX verify supported by driver! */ + texTemp.format = PIPE_FORMAT_A8_UNORM; /* XXX verify supported by driver! */ texTemp.last_level = 0; texTemp.width[0] = 32; texTemp.height[0] = 32; diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 13175ca46e..63e1cc6013 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -385,7 +385,7 @@ z16_get_tile_rgba(ushort *src, -/*** PIPE_FORMAT_U_L8 ***/ +/*** PIPE_FORMAT_L8_UNORM ***/ static void l8_get_tile_rgba(ubyte *src, @@ -408,7 +408,7 @@ l8_get_tile_rgba(ubyte *src, } -/*** PIPE_FORMAT_U_A8 ***/ +/*** PIPE_FORMAT_A8_UNORM ***/ static void a8_get_tile_rgba(ubyte *src, @@ -476,7 +476,7 @@ r16g16b16a16_put_tile_rgba(short *dst, -/*** PIPE_FORMAT_U_I8 ***/ +/*** PIPE_FORMAT_I8_UNORM ***/ static void i8_get_tile_rgba(ubyte *src, @@ -499,7 +499,7 @@ i8_get_tile_rgba(ubyte *src, } -/*** PIPE_FORMAT_U_A8_L8 ***/ +/*** PIPE_FORMAT_A8L8_UNORM ***/ static void a8_l8_get_tile_rgba(ushort *src, @@ -708,16 +708,16 @@ pipe_get_tile_rgba(struct pipe_context *pipe, case PIPE_FORMAT_R5G6B5_UNORM: r5g6b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); break; - case PIPE_FORMAT_U_L8: + case PIPE_FORMAT_L8_UNORM: l8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); break; - case PIPE_FORMAT_U_A8: + case PIPE_FORMAT_A8_UNORM: a8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); break; - case PIPE_FORMAT_U_I8: + case PIPE_FORMAT_I8_UNORM: i8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); break; - case PIPE_FORMAT_U_A8_L8: + case PIPE_FORMAT_A8L8_UNORM: a8_l8_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); break; case PIPE_FORMAT_R16G16B16A16_SNORM: @@ -787,16 +787,16 @@ pipe_put_tile_rgba(struct pipe_context *pipe, break; case PIPE_FORMAT_R8G8B8A8_UNORM: break; - case PIPE_FORMAT_U_L8: + case PIPE_FORMAT_L8_UNORM: /*l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ break; - case PIPE_FORMAT_U_A8: + case PIPE_FORMAT_A8_UNORM: /*a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ break; - case PIPE_FORMAT_U_I8: + case PIPE_FORMAT_I8_UNORM: /*i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ break; - case PIPE_FORMAT_U_A8_L8: + case PIPE_FORMAT_A8L8_UNORM: /*a8_l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_R16G16B16A16_SNORM: diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index b8dc6c66c0..0348629ab8 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -493,13 +493,13 @@ format_to_type_comps(enum pipe_format pformat, *datatype = USHORT_5_6_5; *comps = 3; return; - case PIPE_FORMAT_U_L8: - case PIPE_FORMAT_U_A8: - case PIPE_FORMAT_U_I8: + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_A8_UNORM: + case PIPE_FORMAT_I8_UNORM: *datatype = UBYTE; *comps = 1; return; - case PIPE_FORMAT_U_A8_L8: + case PIPE_FORMAT_A8L8_UNORM: *datatype = UBYTE; *comps = 2; return; diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 9ae594ce54..631642e1b6 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -154,10 +154,10 @@ i915_is_format_supported( struct pipe_screen *screen, PIPE_FORMAT_R8G8B8A8_UNORM, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_R5G6B5_UNORM, - PIPE_FORMAT_U_L8, - PIPE_FORMAT_U_A8, - PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_A8_L8, + PIPE_FORMAT_L8_UNORM, + PIPE_FORMAT_A8_UNORM, + PIPE_FORMAT_I8_UNORM, + PIPE_FORMAT_A8L8_UNORM, PIPE_FORMAT_YCBCR, PIPE_FORMAT_YCBCR_REV, PIPE_FORMAT_S8Z24_UNORM, diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index 84f6529a3a..982eec4a1b 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -131,13 +131,13 @@ static uint translate_texture_format(enum pipe_format pipeFormat) { switch (pipeFormat) { - case PIPE_FORMAT_U_L8: + case PIPE_FORMAT_L8_UNORM: return MAPSURF_8BIT | MT_8BIT_L8; - case PIPE_FORMAT_U_I8: + case PIPE_FORMAT_I8_UNORM: return MAPSURF_8BIT | MT_8BIT_I8; - case PIPE_FORMAT_U_A8: + case PIPE_FORMAT_A8_UNORM: return MAPSURF_8BIT | MT_8BIT_A8; - case PIPE_FORMAT_U_A8_L8: + case PIPE_FORMAT_A8L8_UNORM: return MAPSURF_16BIT | MT_16BIT_AY88; case PIPE_FORMAT_R5G6B5_UNORM: return MAPSURF_16BIT | MT_16BIT_RGB565; diff --git a/src/gallium/drivers/i965simple/brw_screen.c b/src/gallium/drivers/i965simple/brw_screen.c index 6845c7abde..b700f7e4f5 100644 --- a/src/gallium/drivers/i965simple/brw_screen.c +++ b/src/gallium/drivers/i965simple/brw_screen.c @@ -141,13 +141,13 @@ brw_is_format_supported( struct pipe_screen *screen, #if 0 /* XXX: This is broken -- rewrite if still needed. */ static const unsigned tex_supported[] = { - PIPE_FORMAT_U_R8_G8_B8_A8, - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, - PIPE_FORMAT_U_L8, - PIPE_FORMAT_U_A8, - PIPE_FORMAT_U_I8, - PIPE_FORMAT_U_L8_A8, + PIPE_FORMAT_R8G8B8A8_UNORM, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_L8_UNORM, + PIPE_FORMAT_A8_UNORM, + PIPE_FORMAT_I8_UNORM, + PIPE_FORMAT_L8A8_UNORM, PIPE_FORMAT_YCBCR, PIPE_FORMAT_YCBCR_REV, PIPE_FORMAT_S8_Z24, @@ -157,16 +157,16 @@ brw_is_format_supported( struct pipe_screen *screen, /* Actually a lot more than this - add later: */ static const unsigned render_supported[] = { - PIPE_FORMAT_U_A8_R8_G8_B8, - PIPE_FORMAT_U_R5_G6_B5, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, }; /* */ static const unsigned z_stencil_supported[] = { - PIPE_FORMAT_U_Z16, - PIPE_FORMAT_U_Z32, - PIPE_FORMAT_S8_Z24, + PIPE_FORMAT_Z16_UNORM, + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_S8Z24_UNORM, }; switch (type) { diff --git a/src/gallium/drivers/i965simple/brw_wm_surface_state.c b/src/gallium/drivers/i965simple/brw_wm_surface_state.c index 853c743ccf..69e56dc8bd 100644 --- a/src/gallium/drivers/i965simple/brw_wm_surface_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_surface_state.c @@ -57,16 +57,16 @@ static unsigned translate_tex_target( enum pipe_texture_target target ) static unsigned translate_tex_format( enum pipe_format pipe_format ) { switch( pipe_format ) { - case PIPE_FORMAT_U_L8: + case PIPE_FORMAT_L8_UNORM: return BRW_SURFACEFORMAT_L8_UNORM; - case PIPE_FORMAT_U_I8: + case PIPE_FORMAT_I8_UNORM: return BRW_SURFACEFORMAT_I8_UNORM; - case PIPE_FORMAT_U_A8: + case PIPE_FORMAT_A8_UNORM: return BRW_SURFACEFORMAT_A8_UNORM; - case PIPE_FORMAT_U_A8_L8: + case PIPE_FORMAT_A8L8_UNORM: return BRW_SURFACEFORMAT_L8A8_UNORM; case PIPE_FORMAT_R8G8B8_UNORM: diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index f0a3b75357..ce8fefe703 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -722,8 +722,8 @@ st_init_bitmap(struct st_context *st) st->bitmap.rasterizer.bypass_vs = 1; /* find a usable texture format */ - if (screen->is_format_supported(screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE)) { - st->bitmap.tex_format = PIPE_FORMAT_U_I8; + if (screen->is_format_supported(screen, PIPE_FORMAT_I8_UNORM, PIPE_TEXTURE)) { + st->bitmap.tex_format = PIPE_FORMAT_I8_UNORM; } else { /* XXX support more formats */ diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 8b5f84cd56..9a385a0457 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -261,13 +261,13 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) case MESA_FORMAT_RGB565: return PIPE_FORMAT_R5G6B5_UNORM; case MESA_FORMAT_AL88: - return PIPE_FORMAT_U_A8_L8; + return PIPE_FORMAT_A8L8_UNORM; case MESA_FORMAT_A8: - return PIPE_FORMAT_U_A8; + return PIPE_FORMAT_A8_UNORM; case MESA_FORMAT_L8: - return PIPE_FORMAT_U_L8; + return PIPE_FORMAT_L8_UNORM; case MESA_FORMAT_I8: - return PIPE_FORMAT_U_I8; + return PIPE_FORMAT_I8_UNORM; case MESA_FORMAT_Z16: return PIPE_FORMAT_Z16_UNORM; case MESA_FORMAT_Z32: @@ -409,8 +409,8 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - if (screen->is_format_supported( screen, PIPE_FORMAT_U_A8, surfType )) - return PIPE_FORMAT_U_A8; + if (screen->is_format_supported( screen, PIPE_FORMAT_A8_UNORM, surfType )) + return PIPE_FORMAT_A8_UNORM; return default_rgba_format( screen, surfType ); case 1: @@ -420,8 +420,8 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - if (screen->is_format_supported( screen, PIPE_FORMAT_U_L8, surfType )) - return PIPE_FORMAT_U_L8; + if (screen->is_format_supported( screen, PIPE_FORMAT_L8_UNORM, surfType )) + return PIPE_FORMAT_L8_UNORM; return default_rgba_format( screen, surfType ); case 2: @@ -433,8 +433,8 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - if (screen->is_format_supported( screen, PIPE_FORMAT_U_A8_L8, surfType )) - return PIPE_FORMAT_U_A8_L8; + if (screen->is_format_supported( screen, PIPE_FORMAT_A8L8_UNORM, surfType )) + return PIPE_FORMAT_A8L8_UNORM; return default_rgba_format( screen, surfType ); case GL_INTENSITY: @@ -443,8 +443,8 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - if (screen->is_format_supported( screen, PIPE_FORMAT_U_I8, surfType )) - return PIPE_FORMAT_U_I8; + if (screen->is_format_supported( screen, PIPE_FORMAT_I8_UNORM, surfType )) + return PIPE_FORMAT_I8_UNORM; return default_rgba_format( screen, surfType ); case GL_YCBCR_MESA: @@ -547,13 +547,13 @@ translate_gallium_format_to_mesa_format(enum pipe_format format) return &_mesa_texformat_argb4444; case PIPE_FORMAT_R5G6B5_UNORM: return &_mesa_texformat_rgb565; - case PIPE_FORMAT_U_A8_L8: + case PIPE_FORMAT_A8L8_UNORM: return &_mesa_texformat_al88; - case PIPE_FORMAT_U_A8: + case PIPE_FORMAT_A8_UNORM: return &_mesa_texformat_a8; - case PIPE_FORMAT_U_L8: + case PIPE_FORMAT_L8_UNORM: return &_mesa_texformat_l8; - case PIPE_FORMAT_U_I8: + case PIPE_FORMAT_I8_UNORM: return &_mesa_texformat_i8; case PIPE_FORMAT_Z16_UNORM: return &_mesa_texformat_z16; -- cgit v1.2.3 From 21989edd55409d72ee55187f4f9062496ca3fbf8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 18:21:10 -0600 Subject: gallium: fix texture border removal bug --- src/mesa/state_tracker/st_cb_texture.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 3d9c550d8c..981246221b 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -494,6 +494,10 @@ st_TexImage(GLcontext * ctx, strip_texture_border(border, &width, &height, &depth, unpack, &unpackNB); unpack = &unpackNB; + texImage->Width = width; + texImage->Height = height; + texImage->Depth = depth; + texImage->Border = 0; border = 0; } -- cgit v1.2.3 From 7584bcf3f746573fc379c7748acc0be96a3db7de Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 30 Apr 2008 18:23:17 -0600 Subject: gallium: remove old files --- src/mesa/state_tracker/st_texobj.c | 102 ------------------------------------- src/mesa/state_tracker/st_texobj.h | 41 --------------- 2 files changed, 143 deletions(-) delete mode 100644 src/mesa/state_tracker/st_texobj.c delete mode 100644 src/mesa/state_tracker/st_texobj.h (limited to 'src') diff --git a/src/mesa/state_tracker/st_texobj.c b/src/mesa/state_tracker/st_texobj.c deleted file mode 100644 index eb5bdb2d08..0000000000 --- a/src/mesa/state_tracker/st_texobj.c +++ /dev/null @@ -1,102 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* - * Authors: - * Brian Paul - */ - -#include "imports.h" -#include "texformat.h" - -#include "st_context.h" -#include "st_texobj.h" -#include "pipe/p_defines.h" - - -/** - * Create a pipe_texture_object from a Mesa texture object. - * Eventually, gl_texture_object may be derived from this... - */ -struct pipe_texture_object * -create_texture_object(struct gl_texture_object *texObj) -{ - struct pipe_texture_object *pto; - const struct gl_texture_image *texImage; - - pto = calloc(1, sizeof(*pto)); - if (!pto) - return NULL; - - /* XXX: Member not defined. Comment-out to get it compile. */ - /*assert(texObj->Complete);*/ - - switch (texObj->Target) { - case GL_TEXTURE_1D: - pto->type = PIPE_TEXTURE_1D; - break; - case GL_TEXTURE_2D: - pto->type = PIPE_TEXTURE_2D; - break; - case GL_TEXTURE_3D: - pto->type = PIPE_TEXTURE_3D; - break; - case GL_TEXTURE_CUBE_MAP: - pto->type = PIPE_TEXTURE_CUBE; - break; - default: - assert(0); - return NULL; - } - - texImage = texObj->Image[0][texObj->BaseLevel]; - assert(texImage); - - switch (texImage->TexFormat->MesaFormat) { - case MESA_FORMAT_RGBA8888: - pto->format = PIPE_FORMAT_U_R8_G8_B8_A8; - break; - case MESA_FORMAT_RGB565: - pto->format = PIPE_FORMAT_U_R5_G6_B5; - break; - - /* XXX fill in more formats */ - - default: - assert(0); - return NULL; - } - - pto->width = texImage->Width; - pto->height = texImage->Height; - pto->depth = texImage->Depth; - - /* XXX verify this */ - pto->mipmapped = texObj->Image[0][texObj->BaseLevel + 1] != NULL; - - return pto; -} diff --git a/src/mesa/state_tracker/st_texobj.h b/src/mesa/state_tracker/st_texobj.h deleted file mode 100644 index 3c66031018..0000000000 --- a/src/mesa/state_tracker/st_texobj.h +++ /dev/null @@ -1,41 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* - * Authors: - * Brian Paul - */ - -#ifndef ST_TEXOBJ_H -#define ST_TEXOBJ_H 1 - - -extern struct pipe_texture_object * -create_texture_object(struct gl_texture_object *texObj); - - -#endif /* ST_TEXOBJ_H */ -- cgit v1.2.3 From c9ed86a96483063f3d6789ed16645a3dca77d726 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 11:07:21 +0100 Subject: gallium: tex surface checkpoint --- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 12 ++-- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 15 +++-- src/gallium/auxiliary/util/p_tile.c | 50 ++++++++++---- src/gallium/auxiliary/util/u_blit.c | 3 +- src/gallium/auxiliary/util/u_gen_mipmap.c | 21 ++++-- src/gallium/drivers/failover/fo_context.c | 2 +- src/gallium/drivers/i915simple/i915_screen.c | 31 +++++++++ src/gallium/drivers/i915simple/i915_surface.c | 23 +++++-- src/gallium/drivers/i915simple/i915_texture.c | 13 ++-- src/gallium/drivers/i965simple/brw_surface.c | 51 ++++++--------- src/gallium/drivers/i965simple/brw_tex_layout.c | 2 +- src/gallium/drivers/softpipe/sp_context.c | 6 +- src/gallium/drivers/softpipe/sp_draw_arrays.c | 6 +- src/gallium/drivers/softpipe/sp_flush.c | 35 +++++----- src/gallium/drivers/softpipe/sp_screen.c | 22 ++++--- src/gallium/drivers/softpipe/sp_surface.c | 23 +++++-- src/gallium/drivers/softpipe/sp_texture.c | 87 ++++++++++++++++++------- src/gallium/drivers/softpipe/sp_texture.h | 1 - src/gallium/drivers/softpipe/sp_tile_cache.c | 36 ++++++---- src/gallium/drivers/softpipe/sp_tile_cache.h | 2 +- src/gallium/include/pipe/p_context.h | 6 -- src/gallium/include/pipe/p_inlines.h | 31 ++++----- src/gallium/include/pipe/p_screen.h | 17 ++++- src/gallium/include/pipe/p_state.h | 4 ++ src/gallium/include/pipe/p_util.h | 3 + src/gallium/include/pipe/p_winsys.h | 2 +- src/gallium/winsys/xlib/xm_winsys.c | 2 + src/mesa/state_tracker/st_atom_pixeltransfer.c | 9 +-- src/mesa/state_tracker/st_cb_accum.c | 12 ++-- src/mesa/state_tracker/st_cb_bitmap.c | 23 +++++-- src/mesa/state_tracker/st_cb_drawpixels.c | 56 +++++++++------- src/mesa/state_tracker/st_cb_fbo.c | 27 +++++--- src/mesa/state_tracker/st_cb_readpixels.c | 5 +- src/mesa/state_tracker/st_cb_texture.c | 48 +++++++------- src/mesa/state_tracker/st_gen_mipmap.c | 6 +- src/mesa/state_tracker/st_texture.c | 34 ++++++---- src/mesa/state_tracker/st_texture.h | 6 +- 37 files changed, 465 insertions(+), 267 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index f501b2aed4..6dc20f2c90 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -415,8 +415,11 @@ aaline_create_texture(struct aaline_stage *aaline) assert(aaline->texture->width[level] == aaline->texture->height[level]); - surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0); - data = pipe_surface_map(surface); + /* This texture is new, no need to flush. + */ + surface = screen->get_tex_surface(screen, aaline->texture, 0, level, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); + data = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_WRITE); if (data == NULL) return FALSE; @@ -440,9 +443,8 @@ aaline_create_texture(struct aaline_stage *aaline) } /* unmap */ - pipe_surface_unmap(surface); - pipe_surface_reference(&surface, NULL); - pipe->texture_update(pipe, aaline->texture, 0, (1 << level)); + screen->surface_unmap(screen, surface); + screen->tex_surface_release(screen, &surface); } return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index c4de9d2698..3aa326acc7 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -376,8 +376,14 @@ pstip_update_texture(struct pstip_stage *pstip) uint i, j; ubyte *data; - surface = screen->get_tex_surface(screen, pstip->texture, 0, 0, 0); - data = pipe_surface_map(surface); + /* XXX: want to avoid flushing just because we use stipple: + */ + pipe->flush( pipe, PIPE_FLUSH_TEXTURE_CACHE, NULL ); + + surface = screen->get_tex_surface(screen, pstip->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); + data = screen->surface_map(screen, surface, + PIPE_BUFFER_USAGE_CPU_WRITE); /* * Load alpha texture. @@ -399,9 +405,8 @@ pstip_update_texture(struct pstip_stage *pstip) } /* unmap */ - pipe_surface_unmap(surface); - pipe_surface_reference(&surface, NULL); - pipe->texture_update(pipe, pstip->texture, 0, 0x1); + screen->surface_unmap(screen, surface); + screen->tex_surface_release(screen, &surface); } diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 63e1cc6013..5728757d2f 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -50,6 +50,7 @@ pipe_get_tile_raw(struct pipe_context *pipe, uint x, uint y, uint w, uint h, void *p, int dst_stride) { + struct pipe_screen *screen = pipe->screen; const uint cpp = ps->cpp; const ubyte *pSrc; const uint src_stride = ps->pitch * cpp; @@ -63,7 +64,11 @@ pipe_get_tile_raw(struct pipe_context *pipe, if (pipe_clip_tile(x, y, &w, &h, ps)) return; - pSrc = (const ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; + pSrc = (const ubyte *) screen->surface_map(screen, ps, + PIPE_BUFFER_USAGE_CPU_READ); + assert(pSrc); /* XXX: proper error handling! */ + + pSrc += (y * ps->pitch + x) * cpp; pDest = (ubyte *) p; for (i = 0; i < h; i++) { @@ -72,7 +77,7 @@ pipe_get_tile_raw(struct pipe_context *pipe, pSrc += src_stride; } - pipe_surface_unmap(ps); + screen->surface_unmap(screen, ps); } @@ -86,6 +91,7 @@ pipe_put_tile_raw(struct pipe_context *pipe, uint x, uint y, uint w, uint h, const void *p, int src_stride) { + struct pipe_screen *screen = pipe->screen; const uint cpp = ps->cpp; const ubyte *pSrc; const uint dst_stride = ps->pitch * cpp; @@ -100,7 +106,11 @@ pipe_put_tile_raw(struct pipe_context *pipe, return; pSrc = (const ubyte *) p; - pDest = (ubyte *) pipe_surface_map(ps) + (y * ps->pitch + x) * cpp; + + pDest = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE); + assert(pDest); /* XXX: proper error handling */ + + pDest += (y * ps->pitch + x) * cpp; for (i = 0; i < h; i++) { memcpy(pDest, pSrc, w * cpp); @@ -108,7 +118,7 @@ pipe_put_tile_raw(struct pipe_context *pipe, pSrc += src_stride; } - pipe_surface_unmap(ps); + screen->surface_unmap(screen, ps); } @@ -834,18 +844,26 @@ pipe_get_tile_z(struct pipe_context *pipe, uint x, uint y, uint w, uint h, uint *z) { + struct pipe_screen *screen = pipe->screen; const uint dstStride = w; + void *map; uint *pDest = z; uint i, j; if (pipe_clip_tile(x, y, &w, &h, ps)) return; + map = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ); + if (!map) { + assert(0); + return; + } + switch (ps->format) { case PIPE_FORMAT_Z32_UNORM: { const uint *pSrc - = (const uint *) pipe_surface_map(ps) + (y * ps->pitch + x); + = (const uint *)map + (y * ps->pitch + x); for (i = 0; i < h; i++) { memcpy(pDest, pSrc, 4 * w); pDest += dstStride; @@ -857,7 +875,7 @@ pipe_get_tile_z(struct pipe_context *pipe, case PIPE_FORMAT_X8Z24_UNORM: { const uint *pSrc - = (const uint *) pipe_surface_map(ps) + (y * ps->pitch + x); + = (const uint *)map + (y * ps->pitch + x); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 24-bit Z to 32-bit Z */ @@ -871,7 +889,7 @@ pipe_get_tile_z(struct pipe_context *pipe, case PIPE_FORMAT_Z16_UNORM: { const ushort *pSrc - = (const ushort *) pipe_surface_map(ps) + (y * ps->pitch + x); + = (const ushort *)map + (y * ps->pitch + x); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 16-bit Z to 32-bit Z */ @@ -886,7 +904,7 @@ pipe_get_tile_z(struct pipe_context *pipe, assert(0); } - pipe_surface_unmap(ps); + screen->surface_unmap(screen, ps); } @@ -896,17 +914,25 @@ pipe_put_tile_z(struct pipe_context *pipe, uint x, uint y, uint w, uint h, const uint *zSrc) { + struct pipe_screen *screen = pipe->screen; const uint srcStride = w; const uint *pSrc = zSrc; + void *map; uint i, j; if (pipe_clip_tile(x, y, &w, &h, ps)) return; + map = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE); + if (!map) { + assert(0); + return; + } + switch (ps->format) { case PIPE_FORMAT_Z32_UNORM: { - uint *pDest = (uint *) pipe_surface_map(ps) + (y * ps->pitch + x); + uint *pDest = (uint *) map + (y * ps->pitch + x); for (i = 0; i < h; i++) { memcpy(pDest, pSrc, 4 * w); pDest += ps->pitch; @@ -917,7 +943,7 @@ pipe_put_tile_z(struct pipe_context *pipe, case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_X8Z24_UNORM: { - uint *pDest = (uint *) pipe_surface_map(ps) + (y * ps->pitch + x); + uint *pDest = (uint *) map + (y * ps->pitch + x); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 32-bit Z to 24-bit Z (0 stencil) */ @@ -930,7 +956,7 @@ pipe_put_tile_z(struct pipe_context *pipe, break; case PIPE_FORMAT_Z16_UNORM: { - ushort *pDest = (ushort *) pipe_surface_map(ps) + (y * ps->pitch + x); + ushort *pDest = (ushort *) map + (y * ps->pitch + x); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 32-bit Z to 16-bit Z */ @@ -945,7 +971,7 @@ pipe_put_tile_z(struct pipe_context *pipe, assert(0); } - pipe_surface_unmap(ps); + screen->surface_unmap(screen, ps); } diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 9e9912c6e4..257473ab26 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -287,7 +287,8 @@ util_blit_pixels(struct blit_state *ctx, if (!tex) return; - texSurf = screen->get_tex_surface(screen, tex, 0, 0, 0); + texSurf = screen->get_tex_surface(screen, tex, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE); /* load temp texture */ pipe->surface_copy(pipe, FALSE, diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 0348629ab8..6ed5503c9a 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -586,8 +586,11 @@ make_1d_mipmap(struct gen_mipmap_state *ctx, struct pipe_surface *srcSurf, *dstSurf; void *srcMap, *dstMap; - srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice); - dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); + srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice, + PIPE_BUFFER_USAGE_CPU_READ); + + dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice, + PIPE_BUFFER_USAGE_CPU_WRITE); srcMap = ((ubyte *) winsys->buffer_map(winsys, srcSurf->buffer, PIPE_BUFFER_USAGE_CPU_READ) @@ -626,8 +629,10 @@ make_2d_mipmap(struct gen_mipmap_state *ctx, struct pipe_surface *srcSurf, *dstSurf; ubyte *srcMap, *dstMap; - srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice); - dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); + srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice, + PIPE_BUFFER_USAGE_CPU_READ); + dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice, + PIPE_BUFFER_USAGE_CPU_WRITE); srcMap = ((ubyte *) winsys->buffer_map(winsys, srcSurf->buffer, PIPE_BUFFER_USAGE_CPU_READ) @@ -888,10 +893,14 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; + struct pipe_surface *surf = + screen->get_tex_surface(screen, pt, face, dstLevel, zslice, + PIPE_BUFFER_USAGE_GPU_WRITE); + /* * Setup framebuffer / dest surface */ - fb.cbufs[0] = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); + fb.cbufs[0] = surf; fb.width = pt->width[dstLevel]; fb.height = pt->height[dstLevel]; cso_set_framebuffer(ctx->cso, &fb); @@ -922,7 +931,7 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); /* need to signal that the texture has changed _after_ rendering to it */ - pipe->texture_update(pipe, pt, face, (1 << dstLevel)); + pipe_surface_reference( &surf, NULL ); } /* restore state we changed */ diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c index cb95ba516f..014a3e31d5 100644 --- a/src/gallium/drivers/failover/fo_context.c +++ b/src/gallium/drivers/failover/fo_context.c @@ -147,8 +147,8 @@ struct pipe_context *failover_create( struct pipe_context *hw, failover->pipe.texture_create = hw->texture_create; failover->pipe.texture_release = hw->texture_release; failover->pipe.get_tex_surface = hw->get_tex_surface; -#endif failover->pipe.texture_update = hw->texture_update; +#endif failover->pipe.flush = hw->flush; diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 631642e1b6..dcd349e478 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -200,6 +200,35 @@ i915_destroy_screen( struct pipe_screen *screen ) } +static void * +i915_surface_map( struct pipe_screen *screen, + struct pipe_surface *surface, + unsigned flags ) +{ + char *map = screen->winsys->buffer_map( screen->winsys, surface->buffer, flags ); + if (map == NULL) + return NULL; + + if (surface->texture && + (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) + { + /* Do something to notify contexts of a texture change. + */ + /* i915_screen(screen)->timestamp++; */ + } + + return map + surface->offset; +} + +static void +i915_surface_unmap(struct pipe_screen *screen, + struct pipe_surface *surface) +{ + screen->winsys->buffer_unmap( screen->winsys, surface->buffer ); +} + + + /** * Create a new i915_screen object */ @@ -243,6 +272,8 @@ i915_create_screen(struct pipe_winsys *winsys, uint pci_id) i915screen->screen.get_param = i915_get_param; i915screen->screen.get_paramf = i915_get_paramf; i915screen->screen.is_format_supported = i915_is_format_supported; + i915screen->screen.surface_map = i915_surface_map; + i915screen->screen.surface_unmap = i915_surface_unmap; i915_init_screen_texture_functions(&i915screen->screen); diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index f4fbedbe9b..98367ac073 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -51,17 +51,25 @@ i915_surface_copy(struct pipe_context *pipe, assert( dst->cpp == src->cpp ); if (0) { - pipe_copy_rect(pipe_surface_map(dst), + void *dst_map = pipe->screen->surface_map( pipe->screen, + dst, + PIPE_BUFFER_USAGE_CPU_WRITE ); + + const void *src_map = pipe->screen->surface_map( pipe->screen, + src, + PIPE_BUFFER_USAGE_CPU_READ ); + + pipe_copy_rect(dst_map, dst->cpp, dst->pitch, dstx, dsty, width, height, - pipe_surface_map(src), + src_map, do_flip ? -(int) src->pitch : src->pitch, srcx, do_flip ? 1 - srcy - height : srcy); - pipe_surface_unmap(src); - pipe_surface_unmap(dst); + pipe->screen->surface_unmap(pipe->screen, src); + pipe->screen->surface_unmap(pipe->screen, dst); } else { i915_copy_blit( i915_context(pipe), @@ -92,7 +100,10 @@ i915_surface_fill(struct pipe_context *pipe, { if (0) { unsigned i, j; - void *dst_map = pipe_surface_map(dst); + void *dst_map = pipe->screen->surface_map( pipe->screen, + dst, + PIPE_BUFFER_USAGE_CPU_WRITE ); + switch (dst->cpp) { case 1: { @@ -126,7 +137,7 @@ i915_surface_fill(struct pipe_context *pipe, break; } - pipe_surface_unmap( dst ); + pipe->screen->surface_unmap(pipe->screen, dst); } else { i915_fill_blit( i915_context(pipe), diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index c39e747705..7b9359a0fe 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -541,13 +541,6 @@ i915_texture_release_screen(struct pipe_screen *screen, } -static void -i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture, - uint face, uint levelsMask) -{ - /* no-op? */ -} - /* * XXX note: same as code in sp_surface.c @@ -555,7 +548,8 @@ i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture, static struct pipe_surface * i915_get_tex_surface_screen(struct pipe_screen *screen, struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) + unsigned face, unsigned level, unsigned zslice, + unsigned flags) { struct i915_texture *tex = (struct i915_texture *)pt; struct pipe_winsys *ws = screen->winsys; @@ -586,6 +580,7 @@ i915_get_tex_surface_screen(struct pipe_screen *screen, ps->height = pt->height[level]; ps->pitch = tex->pitch; ps->offset = offset; + ps->usage = flags; } return ps; } @@ -594,7 +589,7 @@ i915_get_tex_surface_screen(struct pipe_screen *screen, void i915_init_texture_functions(struct i915_context *i915) { - i915->pipe.texture_update = i915_texture_update; +// i915->pipe.texture_update = i915_texture_update; } diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index c99a91dcf7..3e3736b280 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -35,27 +35,6 @@ #include "util/p_tile.h" -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -static void -brw_surface_data(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - const void *src, unsigned src_pitch, - unsigned srcx, unsigned srcy, unsigned width, unsigned height) -{ - pipe_copy_rect(pipe_surface_map(dst) + dst->offset, - dst->cpp, dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - pipe_surface_unmap(dst); -} - /* Assumes all values are within bounds -- no checking at this level - * do it higher up if required. @@ -72,17 +51,25 @@ brw_surface_copy(struct pipe_context *pipe, assert(dst->cpp == src->cpp); if (0) { - pipe_copy_rect(pipe_surface_map(dst) + dst->offset, + void *dst_map = pipe->screen->surface_map( pipe->screen, + dst, + PIPE_BUFFER_USAGE_CPU_WRITE ); + + const void *src_map = pipe->screen->surface_map( pipe->screen, + src, + PIPE_BUFFER_USAGE_CPU_READ ); + + pipe_copy_rect(dst_map, dst->cpp, dst->pitch, - dstx, dsty, - width, height, - pipe_surface_map(src) + src->offset, - do_flip ? -src->pitch : src->pitch, + dstx, dsty, + width, height, + src_map, + do_flip ? -(int) src->pitch : src->pitch, srcx, do_flip ? 1 - srcy - height : srcy); - pipe_surface_unmap(src); - pipe_surface_unmap(dst); + pipe->screen->surface_unmap(pipe->screen, src); + pipe->screen->surface_unmap(pipe->screen, dst); } else { brw_copy_blit(brw_context(pipe), @@ -113,7 +100,10 @@ brw_surface_fill(struct pipe_context *pipe, { if (0) { unsigned i, j; - void *dst_map = pipe_surface_map(dst); + void *dst_map = pipe->screen->surface_map( pipe->screen, + dst, + PIPE_BUFFER_USAGE_CPU_WRITE ); + switch (dst->cpp) { case 1: { @@ -147,7 +137,7 @@ brw_surface_fill(struct pipe_context *pipe, break; } - pipe_surface_unmap( dst ); + pipe->screen->surface_unmap(pipe->screen, dst); } else { brw_fill_blit(brw_context(pipe), @@ -164,7 +154,6 @@ brw_surface_fill(struct pipe_context *pipe, void brw_init_surface_functions(struct brw_context *brw) { - (void) brw_surface_data; /* silence warning */ brw->pipe.surface_copy = brw_surface_copy; brw->pipe.surface_fill = brw_surface_fill; } diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index b580f98204..ba4c4a7bcf 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -407,7 +407,7 @@ brw_get_tex_surface_screen(struct pipe_screen *screen, void brw_init_texture_functions(struct brw_context *brw) { - brw->pipe.texture_update = brw_texture_update; +// brw->pipe.texture_update = brw_texture_update; } diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index edf91ecafa..ee74826763 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -192,11 +192,11 @@ softpipe_create( struct pipe_screen *screen, * Must be before quad stage setup! */ for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) - softpipe->cbuf_cache[i] = sp_create_tile_cache(); - softpipe->zsbuf_cache = sp_create_tile_cache(); + softpipe->cbuf_cache[i] = sp_create_tile_cache( screen ); + softpipe->zsbuf_cache = sp_create_tile_cache( screen ); for (i = 0; i < PIPE_MAX_SAMPLERS; i++) - softpipe->tex_cache[i] = sp_create_tile_cache(); + softpipe->tex_cache[i] = sp_create_tile_cache( screen ); /* setup quad rendering stages */ diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 6c58f9909d..355c120d18 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -50,7 +50,7 @@ softpipe_map_constant_buffers(struct softpipe_context *sp) for (i = 0; i < 2; i++) { if (sp->constants[i].size) sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); + PIPE_BUFFER_USAGE_GPU_READ); } draw_set_mapped_constant_buffer(sp->draw, @@ -133,14 +133,14 @@ softpipe_draw_elements(struct pipe_context *pipe, void *buf = pipe->winsys->buffer_map(pipe->winsys, sp->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_CPU_READ); + PIPE_BUFFER_USAGE_GPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } /* Map index buffer, if present */ if (indexBuffer) { void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_USAGE_CPU_READ); + PIPE_BUFFER_USAGE_GPU_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c index 0625b69099..e03994b63b 100644 --- a/src/gallium/drivers/softpipe/sp_flush.c +++ b/src/gallium/drivers/softpipe/sp_flush.c @@ -50,25 +50,28 @@ softpipe_flush( struct pipe_context *pipe, draw_flush(softpipe->draw); - /* - flush the quad pipeline - * - flush the texture cache - * - flush the render cache - */ + if (flags & PIPE_FLUSH_TEXTURE_CACHE) { + for (i = 0; i < softpipe->num_textures; i++) { + sp_flush_tile_cache(softpipe, softpipe->tex_cache[i]); + } + } - for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) - if (softpipe->cbuf_cache[i]) - sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]); + if (flags & PIPE_FLUSH_RENDER_CACHE) { + for (i = 0; i < softpipe->framebuffer.num_cbufs; i++) + if (softpipe->cbuf_cache[i]) + sp_flush_tile_cache(softpipe, softpipe->cbuf_cache[i]); - if (softpipe->zsbuf_cache) - sp_flush_tile_cache(softpipe, softpipe->zsbuf_cache); + if (softpipe->zsbuf_cache) + sp_flush_tile_cache(softpipe, softpipe->zsbuf_cache); - /* Need this call for hardware buffers before swapbuffers. - * - * there should probably be another/different flush-type function - * that's called before swapbuffers because we don't always want - * to unmap surfaces when flushing. - */ - softpipe_unmap_surfaces(softpipe); + /* Need this call for hardware buffers before swapbuffers. + * + * there should probably be another/different flush-type function + * that's called before swapbuffers because we don't always want + * to unmap surfaces when flushing. + */ + softpipe_unmap_surfaces(softpipe); + } if (fence) *fence = NULL; diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index 7dacb1c461..e9926bf41f 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -33,6 +33,7 @@ #include "sp_texture.h" #include "sp_winsys.h" +#include "sp_screen.h" static const char * @@ -137,6 +138,7 @@ softpipe_destroy_screen( struct pipe_screen *screen ) } + /** * Create a new pipe_screen object * Note: we're not presently subclassing pipe_screen (no softpipe_screen). @@ -144,22 +146,22 @@ softpipe_destroy_screen( struct pipe_screen *screen ) struct pipe_screen * softpipe_create_screen(struct pipe_winsys *winsys) { - struct pipe_screen *screen = CALLOC_STRUCT(pipe_screen); + struct softpipe_screen *screen = CALLOC_STRUCT(softpipe_screen); if (!screen) return NULL; - screen->winsys = winsys; + screen->base.winsys = winsys; - screen->destroy = softpipe_destroy_screen; + screen->base.destroy = softpipe_destroy_screen; - screen->get_name = softpipe_get_name; - screen->get_vendor = softpipe_get_vendor; - screen->get_param = softpipe_get_param; - screen->get_paramf = softpipe_get_paramf; - screen->is_format_supported = softpipe_is_format_supported; + screen->base.get_name = softpipe_get_name; + screen->base.get_vendor = softpipe_get_vendor; + screen->base.get_param = softpipe_get_param; + screen->base.get_paramf = softpipe_get_paramf; + screen->base.is_format_supported = softpipe_is_format_supported; - softpipe_init_screen_texture_funcs(screen); + softpipe_init_screen_texture_funcs(&screen->base); - return screen; + return &screen->base; } diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index 653449c4f1..b5cc053548 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -47,18 +47,27 @@ sp_surface_copy(struct pipe_context *pipe, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { assert( dst->cpp == src->cpp ); + void *dst_map = pipe->screen->surface_map( pipe->screen, + dst, + PIPE_BUFFER_USAGE_GPU_WRITE ); - pipe_copy_rect(pipe_surface_map(dst), + const void *src_map = pipe->screen->surface_map( pipe->screen, + src, + PIPE_BUFFER_USAGE_GPU_READ ); + + assert(src_map && dst_map); + + pipe_copy_rect(dst_map, dst->cpp, dst->pitch, dstx, dsty, width, height, - pipe_surface_map(src), + src_map, do_flip ? -(int) src->pitch : src->pitch, srcx, do_flip ? 1 - srcy - height : srcy); - pipe_surface_unmap(src); - pipe_surface_unmap(dst); + pipe->screen->surface_unmap(pipe->screen, src); + pipe->screen->surface_unmap(pipe->screen, dst); } @@ -83,7 +92,9 @@ sp_surface_fill(struct pipe_context *pipe, unsigned width, unsigned height, unsigned value) { unsigned i, j; - void *dst_map = pipe_surface_map(dst); + void *dst_map = pipe->screen->surface_map( pipe->screen, + dst, + PIPE_BUFFER_USAGE_GPU_WRITE ); assert(dst->pitch > 0); assert(width <= dst->pitch); @@ -147,7 +158,7 @@ sp_surface_fill(struct pipe_context *pipe, break; } - pipe_surface_unmap( dst ); + pipe->screen->surface_unmap(pipe->screen, dst); } diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 256586ec88..ee3fa994f9 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -40,6 +40,7 @@ #include "sp_state.h" #include "sp_texture.h" #include "sp_tile_cache.h" +#include "sp_screen.h" /* Simple, maximally packed layout. @@ -116,19 +117,10 @@ softpipe_texture_release(struct pipe_screen *screen, if (!*pt) return; - /* - DBG("%s %p refcount will be %d\n", - __FUNCTION__, (void *) *pt, (*pt)->refcount - 1); - */ if (--(*pt)->refcount <= 0) { struct softpipe_texture *spt = softpipe_texture(*pt); - /* - DBG("%s deleting %p\n", __FUNCTION__, (void *) spt); - */ - pipe_buffer_reference(screen->winsys, &spt->buffer, NULL); - FREE(spt); } *pt = NULL; @@ -138,7 +130,8 @@ softpipe_texture_release(struct pipe_screen *screen, static struct pipe_surface * softpipe_get_tex_surface(struct pipe_screen *screen, struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) + unsigned face, unsigned level, unsigned zslice, + unsigned usage) { struct pipe_winsys *ws = screen->winsys; struct softpipe_texture *spt = softpipe_texture(pt); @@ -157,6 +150,7 @@ softpipe_get_tex_surface(struct pipe_screen *screen, ps->height = pt->height[level]; ps->pitch = ps->width; ps->offset = spt->level_offset[level]; + ps->usage = usage; if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * @@ -167,30 +161,74 @@ softpipe_get_tex_surface(struct pipe_screen *screen, assert(face == 0); assert(zslice == 0); } + + if (usage & (PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_GPU_WRITE)) { + /* XXX if writing to the texture, invalidate the texcache entries!!! */ + assert(0); + } } return ps; } -static void -softpipe_texture_update(struct pipe_context *pipe, - struct pipe_texture *texture, - uint face, uint levelsMask) +static void +softpipe_tex_surface_release(struct pipe_screen *screen, + struct pipe_surface **s) { - struct softpipe_context *softpipe = softpipe_context(pipe); - uint unit; - for (unit = 0; unit < softpipe->num_textures; unit++) { - if (softpipe->texture[unit] == texture) { - sp_flush_tile_cache(softpipe, softpipe->tex_cache[unit]); - } + /* Effectively do the texture_update work here - if texture images + * needed post-processing to put them into hardware layout, this is + * where it would happen. For softpipe, nothing to do. + */ + assert ((*s)->texture); + + screen->winsys->surface_release(screen->winsys, s); +} + + +static void * +softpipe_surface_map( struct pipe_screen *screen, + struct pipe_surface *surface, + unsigned flags ) +{ + ubyte *map; + + if (flags & ~surface->usage) { + assert(0); + return NULL; + } + + map = screen->winsys->buffer_map( screen->winsys, surface->buffer, flags ); + if (map == NULL) + return NULL; + + /* May want to different things here depending on read/write nature + * of the map: + */ + if (surface->texture && + (flags & PIPE_BUFFER_USAGE_GPU_WRITE)) + { + /* Do something to notify sharing contexts of a texture change. + * In softpipe, that would mean flushing the texture cache. + */ + softpipe_screen(screen)->timestamp++; } + + return map + surface->offset; +} + + +static void +softpipe_surface_unmap(struct pipe_screen *screen, + struct pipe_surface *surface) +{ + screen->winsys->buffer_unmap( screen->winsys, surface->buffer ); } void -softpipe_init_texture_funcs( struct softpipe_context *softpipe ) +softpipe_init_texture_funcs(struct softpipe_context *sp) { - softpipe->pipe.texture_update = softpipe_texture_update; } @@ -199,5 +237,10 @@ softpipe_init_screen_texture_funcs(struct pipe_screen *screen) { screen->texture_create = softpipe_texture_create; screen->texture_release = softpipe_texture_release; + screen->get_tex_surface = softpipe_get_tex_surface; + screen->tex_surface_release = softpipe_tex_surface_release; + + screen->surface_map = softpipe_surface_map; + screen->surface_unmap = softpipe_surface_unmap; } diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h index a7322144e6..2ba093320d 100644 --- a/src/gallium/drivers/softpipe/sp_texture.h +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -61,7 +61,6 @@ softpipe_texture(struct pipe_texture *pt) extern void softpipe_init_texture_funcs( struct softpipe_context *softpipe ); - extern void softpipe_init_screen_texture_funcs(struct pipe_screen *screen); diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index a88aad5d09..a3fd375a2d 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -49,6 +49,7 @@ struct softpipe_tile_cache { + struct pipe_screen *screen; struct pipe_surface *surface; /**< the surface we're caching */ void *surface_map; struct pipe_texture *texture; /**< if caching a texture */ @@ -109,13 +110,14 @@ clear_clear_flag(uint *bitvec, int x, int y) struct softpipe_tile_cache * -sp_create_tile_cache(void) +sp_create_tile_cache( struct pipe_screen *screen ) { struct softpipe_tile_cache *tc; uint pos; tc = CALLOC_STRUCT( softpipe_tile_cache ); if (tc) { + tc->screen = screen; for (pos = 0; pos < NUM_ENTRIES; pos++) { tc->entries[pos].x = tc->entries[pos].y = -1; @@ -154,16 +156,17 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, assert(!tc->texture); if (tc->surface_map) { - /*assert(tc->surface != ps);*/ - pipe_surface_unmap(tc->surface); + tc->screen->surface_unmap(tc->screen, tc->surface); tc->surface_map = NULL; } pipe_surface_reference(&tc->surface, ps); - if (ps) { - if (tc->surface_map) - tc->surface_map = pipe_surface_map(ps); + if (tc->surface) { + if (tc->surface_map) /* XXX: this is always NULL!? */ + tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || ps->format == PIPE_FORMAT_Z16_UNORM || @@ -187,10 +190,13 @@ void sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc) { if (tc->surface && !tc->surface_map) - tc->surface_map = pipe_surface_map(tc->surface); + tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface, + PIPE_BUFFER_USAGE_GPU_WRITE | + PIPE_BUFFER_USAGE_GPU_READ); if (tc->tex_surf && !tc->tex_surf_map) - tc->tex_surf_map = pipe_surface_map(tc->tex_surf); + tc->tex_surf_map = tc->screen->surface_map(tc->screen, tc->tex_surf, + PIPE_BUFFER_USAGE_GPU_READ); } @@ -198,12 +204,12 @@ void sp_tile_cache_unmap_surfaces(struct softpipe_tile_cache *tc) { if (tc->surface_map) { - pipe_surface_unmap(tc->surface); + tc->screen->surface_unmap(tc->screen, tc->surface); tc->surface_map = NULL; } if (tc->tex_surf_map) { - pipe_surface_unmap(tc->tex_surf); + tc->screen->surface_unmap(tc->screen, tc->tex_surf); tc->tex_surf_map = NULL; } } @@ -224,7 +230,7 @@ sp_tile_cache_set_texture(struct pipe_context *pipe, pipe_texture_reference(&tc->texture, texture); if (tc->tex_surf_map) { - pipe_surface_unmap(tc->tex_surf); + tc->screen->surface_unmap(tc->screen, tc->tex_surf); tc->tex_surf_map = NULL; } pipe_surface_reference(&tc->tex_surf, NULL); @@ -514,10 +520,12 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, /* get new surface (view into texture) */ if (tc->tex_surf_map) - pipe_surface_unmap(tc->tex_surf); + tc->screen->surface_unmap(tc->screen, tc->tex_surf); - tc->tex_surf = screen->get_tex_surface(screen, tc->texture, face, level, z); - tc->tex_surf_map = pipe_surface_map(tc->tex_surf); + tc->tex_surf = screen->get_tex_surface(screen, tc->texture, face, level, z, + PIPE_BUFFER_USAGE_GPU_READ); + tc->tex_surf_map = screen->surface_map(screen, tc->tex_surf, + PIPE_BUFFER_USAGE_GPU_READ); tc->tex_face = face; tc->tex_level = level; diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.h b/src/gallium/drivers/softpipe/sp_tile_cache.h index 2631e29a3a..bc96c941f6 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.h +++ b/src/gallium/drivers/softpipe/sp_tile_cache.h @@ -61,7 +61,7 @@ struct softpipe_cached_tile extern struct softpipe_tile_cache * -sp_create_tile_cache(void); +sp_create_tile_cache( struct pipe_screen *screen ); extern void sp_destroy_tile_cache(struct softpipe_tile_cache *tc); diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index f3a9c2cd8b..0f68f592f7 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -198,12 +198,6 @@ struct pipe_context { /*@}*/ - /** Called when texture data is changed */ - void (*texture_update)(struct pipe_context *pipe, - struct pipe_texture *texture, - uint face, uint dirtyLevelsMask); - - /** Flush rendering (flags = bitmask of PIPE_FLUSH_x tokens) */ void (*flush)( struct pipe_context *pipe, unsigned flags, diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h index 8eb604e73f..592c3c87c2 100644 --- a/src/gallium/include/pipe/p_inlines.h +++ b/src/gallium/include/pipe/p_inlines.h @@ -39,20 +39,6 @@ extern "C" { #endif -static INLINE void * -pipe_surface_map(struct pipe_surface *surface) -{ - return (char *)surface->winsys->buffer_map( surface->winsys, surface->buffer, - PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_CPU_READ ) - + surface->offset; -} - -static INLINE void -pipe_surface_unmap(struct pipe_surface *surface) -{ - surface->winsys->buffer_unmap( surface->winsys, surface->buffer ); -} /** * Set 'ptr' to point to 'surf' and update reference counting. @@ -66,9 +52,20 @@ pipe_surface_reference(struct pipe_surface **ptr, struct pipe_surface *surf) if (surf) surf->refcount++; - if (*ptr /* && --(*ptr)->refcount == 0 */) { - struct pipe_winsys *winsys = (*ptr)->winsys; - winsys->surface_release(winsys, ptr); + if (*ptr) { + + /* There are currently two sorts of surfaces... This needs to be + * fixed so that all surfaces are views into a texture. + */ + if ((*ptr)->texture) { + struct pipe_screen *screen = (*ptr)->texture->screen; + screen->tex_surface_release( screen, ptr ); + } + else { + struct pipe_winsys *winsys = (*ptr)->winsys; + winsys->surface_release(winsys, ptr); + } + assert(!*ptr); } diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h index 26ac99d287..c080579c26 100644 --- a/src/gallium/include/pipe/p_screen.h +++ b/src/gallium/include/pipe/p_screen.h @@ -96,7 +96,22 @@ struct pipe_screen { struct pipe_surface *(*get_tex_surface)(struct pipe_screen *, struct pipe_texture *texture, unsigned face, unsigned level, - unsigned zslice); + unsigned zslice, + unsigned usage ); + + /* Surfaces allocated by the above must be released here: + */ + void (*tex_surface_release)( struct pipe_screen *, + struct pipe_surface ** ); + + + void *(*surface_map)( struct pipe_screen *, + struct pipe_surface *surface, + unsigned flags ); + + void (*surface_unmap)( struct pipe_screen *, + struct pipe_surface *surface ); + }; diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 912d84e7b9..62b05a403b 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -273,7 +273,11 @@ struct pipe_surface unsigned pitch; /**< in pixels */ unsigned offset; /**< offset from start of buffer, in bytes */ unsigned refcount; + unsigned usage; /**< PIPE_BUFFER_USAGE_* */ + struct pipe_winsys *winsys; /**< winsys which owns/created the surface */ + + struct pipe_texture *texture; /**< optional texture into which this is a view */ }; diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 0e7e246666..0d8ed167b2 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -204,7 +204,10 @@ mem_dup(const void *src, uint size) #define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) #define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) +#ifndef Elements #define Elements(x) (sizeof(x)/sizeof((x)[0])) +#endif + #define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER)) /** diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h index 3005ec2d94..87a66b66d7 100644 --- a/src/gallium/include/pipe/p_winsys.h +++ b/src/gallium/include/pipe/p_winsys.h @@ -90,7 +90,7 @@ struct pipe_winsys void (*surface_release)(struct pipe_winsys *ws, struct pipe_surface **s); - + /** * Buffer management. Buffer attributes are mostly fixed over its lifetime. * diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 8a89278cde..fd2f56eff2 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -508,6 +508,7 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys, surf->format = format; surf->cpp = pf_get_size(format); surf->pitch = round_up(width, alignment / surf->cpp); + surf->usage = flags; #ifdef GALLIUM_CELL /* XXX a bit of a hack */ height = round_up(height, TILE_SIZE); @@ -562,6 +563,7 @@ static void xm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) { struct pipe_surface *surf = *s; + assert(!surf->texture); surf->refcount--; if (surf->refcount == 0) { if (surf->buffer) diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index 76356bbad7..e7186a85da 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -148,8 +148,10 @@ load_color_map_texture(GLcontext *ctx, struct pipe_texture *pt) uint *dest; uint i, j; - surface = screen->get_tex_surface(screen, pt, 0, 0, 0); - dest = (uint *) pipe_surface_map(surface); + surface = screen->get_tex_surface(screen, pt, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); + dest = (uint *) screen->surface_map(screen, surface, + PIPE_BUFFER_USAGE_CPU_WRITE); /* Pack four 1D maps into a 2D texture: * R map is placed horizontally, indexed by S, in channel 0 @@ -168,9 +170,8 @@ load_color_map_texture(GLcontext *ctx, struct pipe_texture *pt) } } - pipe_surface_unmap(surface); + screen->surface_unmap(screen, surface); pipe_surface_reference(&surface, NULL); - pipe->texture_update(pipe, pt, 0, 0x1); } diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 1636bed91a..e4ef3e16b7 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -106,13 +106,15 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { struct st_renderbuffer *acc_strb = st_renderbuffer(rb); struct pipe_surface *acc_ps = acc_strb->surface; + struct pipe_screen *screen = ctx->st->pipe->screen; const GLint xpos = ctx->DrawBuffer->_Xmin; const GLint ypos = ctx->DrawBuffer->_Ymin; const GLint width = ctx->DrawBuffer->_Xmax - xpos; const GLint height = ctx->DrawBuffer->_Ymax - ypos; GLvoid *map; - map = pipe_surface_map(acc_ps); + map = screen->surface_map(screen, acc_ps, + PIPE_BUFFER_USAGE_CPU_WRITE); /* note acc_strb->format might not equal acc_ps->format */ switch (acc_strb->format) { @@ -140,7 +142,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) _mesa_problem(ctx, "unexpected format in st_clear_accum_buffer()"); } - pipe_surface_unmap(acc_ps); + screen->surface_unmap(screen, acc_ps); } @@ -150,10 +152,12 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias, GLint xpos, GLint ypos, GLint width, GLint height, struct st_renderbuffer *acc_strb) { + struct pipe_screen *screen = ctx->st->pipe->screen; struct pipe_surface *acc_ps = acc_strb->surface; GLvoid *map; - map = pipe_surface_map(acc_ps); + map = screen->surface_map(screen, acc_ps, + PIPE_BUFFER_USAGE_CPU_WRITE); /* note acc_strb->format might not equal acc_ps->format */ switch (acc_strb->format) { @@ -174,7 +178,7 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias, _mesa_problem(NULL, "unexpected format in st_clear_accum_buffer()"); } - pipe_surface_unmap(acc_ps); + screen->surface_unmap(screen, acc_ps); } diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index ce8fefe703..873b765c2c 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -327,10 +327,11 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, return NULL; } - surface = screen->get_tex_surface(screen, pt, 0, 0, 0); + surface = screen->get_tex_surface(screen, pt, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); /* map texture surface */ - dest = pipe_surface_map(surface); + dest = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_WRITE); /* Put image into texture surface */ memset(dest, 0xff, height * surface->pitch); @@ -340,9 +341,8 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, _mesa_unmap_bitmap_pbo(ctx, unpack); /* Release surface */ - pipe_surface_unmap(surface); + screen->surface_unmap(screen, surface); pipe_surface_reference(&surface, NULL); - pipe->texture_update(pipe, pt, 0, 0x1); return pt; } @@ -544,8 +544,10 @@ reset_cache(struct st_context *st) /* Map the texture surface. * Subsequent glBitmap calls will write into the texture image. */ - cache->surf = screen->get_tex_surface(screen, cache->texture, 0, 0, 0); - cache->buffer = pipe_surface_map(cache->surf); + cache->surf = screen->get_tex_surface(screen, cache->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); + cache->buffer = screen->surface_map(screen, cache->surf, + PIPE_BUFFER_USAGE_CPU_WRITE); /* init image to all 0xff */ memset(cache->buffer, 0xff, BITMAP_CACHE_WIDTH * BITMAP_CACHE_HEIGHT); @@ -562,6 +564,7 @@ st_flush_bitmap_cache(struct st_context *st) if (st->ctx->DrawBuffer) { struct bitmap_cache *cache = st->bitmap.cache; struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; assert(cache->xmin <= cache->xmax); /* @@ -574,12 +577,18 @@ st_flush_bitmap_cache(struct st_context *st) /* The texture surface has been mapped until now. * So unmap and release the texture surface before drawing. */ +#if 0 pipe_surface_unmap(cache->surf); pipe_surface_reference(&cache->surf, NULL); +#else + screen->surface_unmap(screen, cache->surf); + screen->tex_surface_release(screen, &cache->surf); +#endif +#if 0 /* XXX is this needed? */ pipe->texture_update(pipe, cache->texture, 0, 0x1); - +#endif draw_bitmap_quad(st->ctx, cache->xpos, cache->ypos, diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 65bfd6cfcc..9ae53c95f8 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -362,10 +362,12 @@ make_texture(struct st_context *st, /* we'll do pixel transfer in a fragment shader */ ctx->_ImageTransferState = 0x0; - surface = screen->get_tex_surface(screen, pt, 0, 0, 0); + surface = screen->get_tex_surface(screen, pt, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); /* map texture surface */ - dest = pipe_surface_map(surface); + dest = screen->surface_map(screen, surface, + PIPE_BUFFER_USAGE_CPU_WRITE); /* Put image into texture surface. * Note that the image is actually going to be upside down in @@ -384,9 +386,8 @@ make_texture(struct st_context *st, unpack); /* unmap */ - pipe_surface_unmap(surface); + screen->surface_unmap(screen, surface); pipe_surface_reference(&surface, NULL); - pipe->texture_update(pipe, pt, 0, 0x1); assert(success); @@ -731,6 +732,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, { struct st_context *st = ctx->st; struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; struct pipe_surface *ps = st->state.framebuffer.zsbuf; const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0; GLint skipPixels; @@ -739,7 +741,8 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); /* map the stencil buffer */ - stmap = pipe_surface_map(ps); + stmap = screen->surface_map(screen, ps, + PIPE_BUFFER_USAGE_CPU_WRITE); /* if width > MAX_WIDTH, have to process image in chunks */ skipPixels = 0; @@ -796,7 +799,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } /* unmap the stencil buffer */ - pipe_surface_unmap(ps); + screen->surface_unmap(screen, ps); } @@ -869,6 +872,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, GLint dstx, GLint dsty) { struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer); + struct pipe_screen *screen = ctx->st->pipe->screen; struct pipe_surface *psDraw = rbDraw->surface; ubyte *drawMap; ubyte *buffer; @@ -885,7 +889,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, &ctx->DefaultPacking, buffer); /* map the stencil buffer */ - drawMap = pipe_surface_map(psDraw); + drawMap = screen->surface_map(screen, psDraw, PIPE_BUFFER_USAGE_CPU_WRITE); /* draw */ /* XXX PixelZoom not handled yet */ @@ -925,7 +929,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, free(buffer); /* unmap the stencil buffer */ - pipe_surface_unmap(psDraw); + screen->surface_unmap(screen, psDraw); } @@ -994,13 +998,14 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, if (!pt) return; - psTex = screen->get_tex_surface(screen, pt, 0, 0, 0); - if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { srcy = ctx->DrawBuffer->Height - srcy - height; } if (srcFormat == texFormat) { + psTex = screen->get_tex_surface(screen, pt, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_WRITE ); + /* copy source framebuffer surface into mipmap/texture */ pipe->surface_copy(pipe, FALSE, @@ -1009,21 +1014,26 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, psRead, srcx, srcy, width, height); } - else if (type == GL_COLOR) { - /* alternate path using get/put_tile() */ - GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + else { + psTex = screen->get_tex_surface(screen, pt, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE ); - pipe_get_tile_rgba(pipe, psRead, srcx, srcy, width, height, buf); - pipe_put_tile_rgba(pipe, psTex, 0, 0, width, height, buf); + if (type == GL_COLOR) { + /* alternate path using get/put_tile() */ + GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - free(buf); - } - else { - /* GL_DEPTH */ - GLuint *buf = (GLuint *) malloc(width * height * sizeof(GLuint)); - pipe_get_tile_z(pipe, psRead, srcx, srcy, width, height, buf); - pipe_put_tile_z(pipe, psTex, 0, 0, width, height, buf); - free(buf); + pipe_get_tile_rgba(pipe, psRead, srcx, srcy, width, height, buf); + pipe_put_tile_rgba(pipe, psTex, 0, 0, width, height, buf); + + free(buf); + } + else { + /* GL_DEPTH */ + GLuint *buf = (GLuint *) malloc(width * height * sizeof(GLuint)); + pipe_get_tile_z(pipe, psRead, srcx, srcy, width, height, buf); + pipe_put_tile_z(pipe, psTex, 0, 0, width, height, buf); + free(buf); + } } /* draw textured quad */ diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index fc8a5ea7f6..7fdc0bddd6 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -91,9 +91,14 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); enum pipe_format pipeFormat; - GLbitfield flags = 0x0; /* XXX needed? */ + unsigned flags = (PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE | + PIPE_BUFFER_USAGE_GPU_READ); int ret; + pipe_surface_reference( &strb->surface, NULL ); + if (!strb->surface) { /* first time surface creation */ strb->surface = pipe->winsys->surface_alloc(pipe->winsys); @@ -103,11 +108,16 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, if (!strb->surface) return GL_FALSE; } +#if 0 else if (strb->surface->buffer) { /* release/discard the old surface buffer */ pipe_reference_buffer(pipe, &strb->surface->buffer, NULL); } - +#else + else { + assert(0); + } +#endif /* Determine surface format here */ if (strb->format != PIPE_FORMAT_NONE) { assert(strb->format != 0); @@ -368,7 +378,11 @@ st_render_texture(GLcontext *ctx, strb->surface = screen->get_tex_surface(screen, pt, att->CubeMapFace, att->TextureLevel, - att->Zoffset); + att->Zoffset, + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); assert(strb->surface); assert(screen->is_format_supported(screen, strb->surface->format, PIPE_TEXTURE)); assert(screen->is_format_supported(screen, strb->surface->format, PIPE_SURFACE)); @@ -396,22 +410,19 @@ static void st_finish_render_texture(GLcontext *ctx, struct gl_renderbuffer_attachment *att) { + struct pipe_screen *screen = ctx->st->pipe->screen; struct st_renderbuffer *strb = st_renderbuffer(att->Renderbuffer); assert(strb); ctx->st->pipe->flush(ctx->st->pipe, PIPE_FLUSH_RENDER_CACHE, NULL); - ctx->st->pipe->texture_update(ctx->st->pipe, - st_get_texobj_texture(att->Texture), - att->CubeMapFace, 1 << att->TextureLevel); + screen->tex_surface_release( screen, &strb->surface ); /* printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface); */ - pipe_surface_reference(&strb->surface, NULL); - _mesa_reference_renderbuffer(&att->Renderbuffer, NULL); /* restore previous framebuffer state */ diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index ddbe36106c..e242195e7a 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -61,13 +61,14 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, GLvoid *pixels) { struct gl_framebuffer *fb = ctx->ReadBuffer; + struct pipe_screen *screen = ctx->st->pipe->screen; struct st_renderbuffer *strb = st_renderbuffer(fb->_StencilBuffer); struct pipe_surface *ps = strb->surface; ubyte *stmap; GLint j; /* map the stencil buffer */ - stmap = pipe_surface_map(ps); + stmap = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ); /* width should never be > MAX_WIDTH since we did clipping earlier */ ASSERT(width <= MAX_WIDTH); @@ -124,7 +125,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, /* unmap the stencil buffer */ - pipe_surface_unmap(ps); + screen->surface_unmap(screen, ps); } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 981246221b..05e0339e0e 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -478,7 +478,6 @@ st_TexImage(GLcontext * ctx, struct gl_texture_image *texImage, GLsizei imageSize, int compressed) { - struct pipe_context *pipe = ctx->st->pipe; struct st_texture_object *stObj = st_texture_object(texObj); struct st_texture_image *stImage = st_texture_image(texImage); GLint postConvWidth, postConvHeight; @@ -635,7 +634,8 @@ st_TexImage(GLcontext * ctx, return; if (stImage->pt) { - texImage->Data = st_texture_image_map(ctx->st, stImage, 0); + texImage->Data = st_texture_image_map(ctx->st, stImage, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); dstRowStride = stImage->surface->pitch * stImage->surface->cpp; } else { @@ -684,8 +684,9 @@ st_TexImage(GLcontext * ctx, } if (stImage->pt && i < depth) { - st_texture_image_unmap(stImage); - texImage->Data = st_texture_image_map(ctx->st, stImage, i); + st_texture_image_unmap(ctx->st, stImage); + texImage->Data = st_texture_image_map(ctx->st, stImage, i, + PIPE_BUFFER_USAGE_CPU_WRITE); src += srcImageStride; } } @@ -694,13 +695,10 @@ st_TexImage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, unpack); if (stImage->pt) { - st_texture_image_unmap(stImage); + st_texture_image_unmap(ctx->st, stImage); texImage->Data = NULL; } - if (stObj->pt) - pipe->texture_update(pipe, stObj->pt, stImage->face, (1 << level)); - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); } @@ -793,7 +791,8 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, /* Image is stored in hardware format in a buffer managed by the * kernel. Need to explicitly map and unmap it. */ - texImage->Data = st_texture_image_map(ctx->st, stImage, 0); + texImage->Data = st_texture_image_map(ctx->st, stImage, 0, + PIPE_BUFFER_USAGE_CPU_READ); texImage->RowStride = stImage->surface->pitch; } else { @@ -823,8 +822,9 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, } if (stImage->pt && i < depth) { - st_texture_image_unmap(stImage); - texImage->Data = st_texture_image_map(ctx->st, stImage, i); + st_texture_image_unmap(ctx->st, stImage); + texImage->Data = st_texture_image_map(ctx->st, stImage, i, + PIPE_BUFFER_USAGE_CPU_READ); dest += dstImageStride; } } @@ -833,7 +833,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, /* Unmap */ if (stImage->pt) { - st_texture_image_unmap(stImage); + st_texture_image_unmap(ctx->st, stImage); texImage->Data = NULL; } } @@ -874,8 +874,6 @@ st_TexSubimage(GLcontext * ctx, struct gl_texture_object *texObj, struct gl_texture_image *texImage) { - struct pipe_context *pipe = ctx->st->pipe; - struct st_texture_object *stObj = st_texture_object(texObj); struct st_texture_image *stImage = st_texture_image(texImage); GLuint dstRowStride; GLuint srcImageStride = _mesa_image_image_stride(packing, width, height, @@ -897,7 +895,8 @@ st_TexSubimage(GLcontext * ctx, * from uploading the buffer under us. */ if (stImage->pt) { - texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset); + texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset, + PIPE_BUFFER_USAGE_CPU_WRITE); dstRowStride = stImage->surface->pitch * stImage->surface->cpp; } @@ -922,8 +921,9 @@ st_TexSubimage(GLcontext * ctx, if (stImage->pt && i < depth) { /* map next slice of 3D texture */ - st_texture_image_unmap(stImage); - texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i); + st_texture_image_unmap(ctx->st, stImage); + texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset + i, + PIPE_BUFFER_USAGE_CPU_WRITE); src += srcImageStride; } } @@ -935,11 +935,9 @@ st_TexSubimage(GLcontext * ctx, _mesa_unmap_teximage_pbo(ctx, packing); if (stImage->pt) { - st_texture_image_unmap(stImage); + st_texture_image_unmap(ctx->st, stImage); texImage->Data = NULL; } - - pipe->texture_update(pipe, stObj->pt, stImage->face, (1 << level)); } @@ -1058,7 +1056,8 @@ fallback_copy_texsubimage(GLcontext *ctx, src_surf = strb->surface; - dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ); + dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ, + PIPE_BUFFER_USAGE_CPU_WRITE); assert(width <= MAX_WIDTH); @@ -1119,7 +1118,6 @@ do_copy_texsubimage(GLcontext *ctx, struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); struct st_texture_image *stImage = st_texture_image(texImage); - struct st_texture_object *stObj = st_texture_object(texObj); GLenum baseFormat = texImage->InternalFormat; struct gl_framebuffer *fb = ctx->ReadBuffer; struct st_renderbuffer *strb; @@ -1157,7 +1155,8 @@ do_copy_texsubimage(GLcontext *ctx, dest_format = stImage->pt->format; dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, - stImage->level, destZ); + stImage->level, destZ, + PIPE_BUFFER_USAGE_CPU_WRITE); if (ctx->_ImageTransferState == 0x0 && strb->surface->buffer && @@ -1223,8 +1222,6 @@ do_copy_texsubimage(GLcontext *ctx, pipe_surface_reference(&dest_surface, NULL); - pipe->texture_update(pipe, stObj->pt, stImage->face, (1 << level)); - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); } @@ -1529,7 +1526,6 @@ st_finalize_texture(GLcontext *ctx, if (stImage && stObj->pt != stImage->pt) { copy_image_data_to_texture(ctx->st, stObj, level, stImage); *needFlush = GL_TRUE; - pipe->texture_update(pipe, stObj->pt, face, (1 << level)); } } } diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 1a0e19c2f9..cfacfdd04c 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -123,8 +123,10 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, const ubyte *srcData; ubyte *dstData; - srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice); - dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice); + srcSurf = screen->get_tex_surface(screen, pt, face, srcLevel, zslice, + PIPE_BUFFER_USAGE_CPU_READ); + dstSurf = screen->get_tex_surface(screen, pt, face, dstLevel, zslice, + PIPE_BUFFER_USAGE_CPU_WRITE); srcData = (ubyte *) pipe_buffer_map(pipe, srcSurf->buffer, PIPE_BUFFER_USAGE_CPU_READ) diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index f68bef1207..482a054f64 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -184,25 +184,30 @@ st_texture_image_offset(const struct pipe_texture * pt, */ GLubyte * st_texture_image_map(struct st_context *st, struct st_texture_image *stImage, - GLuint zoffset) + GLuint zoffset, + GLuint flags ) { struct pipe_screen *screen = st->pipe->screen; struct pipe_texture *pt = stImage->pt; DBG("%s \n", __FUNCTION__); stImage->surface = screen->get_tex_surface(screen, pt, stImage->face, - stImage->level, zoffset); + stImage->level, zoffset, + flags); - return pipe_surface_map(stImage->surface); + return screen->surface_map(screen, stImage->surface, flags); } void -st_texture_image_unmap(struct st_texture_image *stImage) +st_texture_image_unmap(struct st_context *st, + struct st_texture_image *stImage) { + struct pipe_screen *screen = st->pipe->screen; + DBG("%s\n", __FUNCTION__); - pipe_surface_unmap(stImage->surface); + screen->surface_unmap(screen, stImage->surface); pipe_surface_reference(&stImage->surface, NULL); } @@ -224,12 +229,15 @@ st_surface_data(struct pipe_context *pipe, const void *src, unsigned src_pitch, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - pipe_copy_rect(pipe_surface_map(dst), + struct pipe_screen *screen = pipe->screen; + void *map = screen->surface_map(screen, dst, PIPE_BUFFER_USAGE_CPU_WRITE); + + pipe_copy_rect(map, dst->cpp, dst->pitch, dstx, dsty, width, height, src, src_pitch, srcx, srcy); - pipe_surface_unmap(dst); + screen->surface_unmap(screen, dst); } @@ -256,7 +264,8 @@ st_texture_image_data(struct pipe_context *pipe, if(dst->compressed) height /= 4; - dst_surface = screen->get_tex_surface(screen, dst, face, level, i); + dst_surface = screen->get_tex_surface(screen, dst, face, level, i, + PIPE_BUFFER_USAGE_CPU_WRITE); st_surface_data(pipe, dst_surface, 0, 0, /* dstx, dsty */ @@ -265,7 +274,7 @@ st_texture_image_data(struct pipe_context *pipe, 0, 0, /* source x, y */ dst->width[level], height); /* width, height */ - pipe_surface_reference(&dst_surface, NULL); + screen->tex_surface_release(screen, &dst_surface); srcUB += src_image_pitch * dst->cpp; } @@ -304,8 +313,11 @@ st_texture_image_copy(struct pipe_context *pipe, assert(src->width[srcLevel] == width); assert(src->height[srcLevel] == height); - dst_surface = screen->get_tex_surface(screen, dst, face, dstLevel, i); - src_surface = screen->get_tex_surface(screen, src, face, srcLevel, i); + dst_surface = screen->get_tex_surface(screen, dst, face, dstLevel, i, + PIPE_BUFFER_USAGE_GPU_WRITE); + + src_surface = screen->get_tex_surface(screen, src, face, srcLevel, i, + PIPE_BUFFER_USAGE_GPU_READ); pipe->surface_copy(pipe, FALSE, diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 7abccb3a69..f6d5733e21 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -121,10 +121,12 @@ st_texture_match_image(const struct pipe_texture *pt, extern GLubyte * st_texture_image_map(struct st_context *st, struct st_texture_image *stImage, - GLuint zoffset); + GLuint zoffset, + GLuint flags); extern void -st_texture_image_unmap(struct st_texture_image *stImage); +st_texture_image_unmap(struct st_context *st, + struct st_texture_image *stImage); /* Return pointers to each 2d slice within an image. Indexed by depth -- cgit v1.2.3 From 27e46611f04108765fa99890822a474820d5c563 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 11:28:47 +0100 Subject: softpipe: use CPU flags for mapping But when creating surfaces, adjust incoming flags from GPU->CPU usage. --- src/gallium/drivers/softpipe/sp_draw_arrays.c | 6 ++--- src/gallium/drivers/softpipe/sp_surface.c | 6 ++--- src/gallium/drivers/softpipe/sp_texture.c | 34 ++++++++++++++++++++++++--- src/gallium/drivers/softpipe/sp_tile_cache.c | 14 +++++------ 4 files changed, 44 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 355c120d18..6c58f9909d 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -50,7 +50,7 @@ softpipe_map_constant_buffers(struct softpipe_context *sp) for (i = 0; i < 2; i++) { if (sp->constants[i].size) sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, - PIPE_BUFFER_USAGE_GPU_READ); + PIPE_BUFFER_USAGE_CPU_READ); } draw_set_mapped_constant_buffer(sp->draw, @@ -133,14 +133,14 @@ softpipe_draw_elements(struct pipe_context *pipe, void *buf = pipe->winsys->buffer_map(pipe->winsys, sp->vertex_buffer[i].buffer, - PIPE_BUFFER_USAGE_GPU_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_vertex_buffer(draw, i, buf); } /* Map index buffer, if present */ if (indexBuffer) { void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, - PIPE_BUFFER_USAGE_GPU_READ); + PIPE_BUFFER_USAGE_CPU_READ); draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); } else { diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index b5cc053548..b82b1a8f37 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -49,11 +49,11 @@ sp_surface_copy(struct pipe_context *pipe, assert( dst->cpp == src->cpp ); void *dst_map = pipe->screen->surface_map( pipe->screen, dst, - PIPE_BUFFER_USAGE_GPU_WRITE ); + PIPE_BUFFER_USAGE_CPU_WRITE ); const void *src_map = pipe->screen->surface_map( pipe->screen, src, - PIPE_BUFFER_USAGE_GPU_READ ); + PIPE_BUFFER_USAGE_CPU_READ ); assert(src_map && dst_map); @@ -94,7 +94,7 @@ sp_surface_fill(struct pipe_context *pipe, unsigned i, j; void *dst_map = pipe->screen->surface_map( pipe->screen, dst, - PIPE_BUFFER_USAGE_GPU_WRITE ); + PIPE_BUFFER_USAGE_CPU_WRITE ); assert(dst->pitch > 0); assert(width <= dst->pitch); diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index ee3fa994f9..2b31cd4f25 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -151,6 +151,23 @@ softpipe_get_tex_surface(struct pipe_screen *screen, ps->pitch = ps->width; ps->offset = spt->level_offset[level]; ps->usage = usage; + + /* Because we are softpipe, anything that the state tracker + * thought was going to be done with the GPU will actually get + * done with the CPU. Let's adjust the flags to take that into + * account. + */ + if (ps->usage & PIPE_BUFFER_USAGE_GPU_WRITE) + ps->usage |= PIPE_BUFFER_USAGE_CPU_WRITE; + + if (ps->usage & PIPE_BUFFER_USAGE_GPU_READ) + ps->usage |= PIPE_BUFFER_USAGE_CPU_READ; + + + pipe_texture_reference(&ps->texture, pt); + ps->face = face; + ps->level = level; + ps->zslice = zslice; if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * @@ -164,8 +181,18 @@ softpipe_get_tex_surface(struct pipe_screen *screen, if (usage & (PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_GPU_WRITE)) { - /* XXX if writing to the texture, invalidate the texcache entries!!! */ - assert(0); + /* XXX if writing to the texture, invalidate the texcache entries!!! + * + * Actually, no. Flushing dependent contexts is still done + * explicitly and separately. Hardware drivers won't insert + * FLUSH commands into a command stream at this point, + * neither should softpipe try to flush caches. + * + * Those contexts could be living in separate threads & doing + * all sorts of unrelated stuff... Context<->texture + * dependency tracking needs to happen elsewhere. + */ + /* assert(0); */ } } return ps; @@ -181,6 +208,7 @@ softpipe_tex_surface_release(struct pipe_screen *screen, * where it would happen. For softpipe, nothing to do. */ assert ((*s)->texture); + pipe_texture_reference(&(*s)->texture, NULL); screen->winsys->surface_release(screen->winsys, s); } @@ -206,7 +234,7 @@ softpipe_surface_map( struct pipe_screen *screen, * of the map: */ if (surface->texture && - (flags & PIPE_BUFFER_USAGE_GPU_WRITE)) + (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) { /* Do something to notify sharing contexts of a texture change. * In softpipe, that would mean flushing the texture cache. diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index a3fd375a2d..142faf5074 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -165,8 +165,8 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, if (tc->surface) { if (tc->surface_map) /* XXX: this is always NULL!? */ tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface, - PIPE_BUFFER_USAGE_GPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE); + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || ps->format == PIPE_FORMAT_Z16_UNORM || @@ -191,12 +191,12 @@ sp_tile_cache_map_surfaces(struct softpipe_tile_cache *tc) { if (tc->surface && !tc->surface_map) tc->surface_map = tc->screen->surface_map(tc->screen, tc->surface, - PIPE_BUFFER_USAGE_GPU_WRITE | - PIPE_BUFFER_USAGE_GPU_READ); + PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_CPU_READ); if (tc->tex_surf && !tc->tex_surf_map) tc->tex_surf_map = tc->screen->surface_map(tc->screen, tc->tex_surf, - PIPE_BUFFER_USAGE_GPU_READ); + PIPE_BUFFER_USAGE_CPU_READ); } @@ -523,9 +523,9 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, tc->screen->surface_unmap(tc->screen, tc->tex_surf); tc->tex_surf = screen->get_tex_surface(screen, tc->texture, face, level, z, - PIPE_BUFFER_USAGE_GPU_READ); + PIPE_BUFFER_USAGE_CPU_READ); tc->tex_surf_map = screen->surface_map(screen, tc->tex_surf, - PIPE_BUFFER_USAGE_GPU_READ); + PIPE_BUFFER_USAGE_CPU_READ); tc->tex_face = face; tc->tex_level = level; -- cgit v1.2.3 From 228aaa6cab9ebb32eb23b85fc8a5f05c1dbe975a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 12:21:48 +0100 Subject: softpipe: missing file --- src/gallium/drivers/softpipe/sp_screen.h | 58 ++++++++++++++++++++++++++++++++ 1 file changed, 58 insertions(+) create mode 100644 src/gallium/drivers/softpipe/sp_screen.h (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_screen.h b/src/gallium/drivers/softpipe/sp_screen.h new file mode 100644 index 0000000000..3d4bfd3e84 --- /dev/null +++ b/src/gallium/drivers/softpipe/sp_screen.h @@ -0,0 +1,58 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef SP_SCREEN_H +#define SP_SCREEN_H + +#include "pipe/p_screen.h" +#include "pipe/p_defines.h" + + + +struct softpipe_screen { + struct pipe_screen base; + + /* Increments whenever textures are modified. Contexts can track + * this. + */ + unsigned timestamp; +}; + + + + +static INLINE struct softpipe_screen * +softpipe_screen( struct pipe_screen *pipe ) +{ + return (struct softpipe_screen *)pipe; +} + + +#endif /* SP_SCREEN_H */ -- cgit v1.2.3 From b59f9c95c1f7aca60c5d6889d4dc388d19fbc3e1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 12:28:50 +0100 Subject: gallium: add information to surface to identify which texture image it is pointing at --- src/gallium/include/pipe/p_state.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 62b05a403b..47e57e2957 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -278,6 +278,9 @@ struct pipe_surface struct pipe_winsys *winsys; /**< winsys which owns/created the surface */ struct pipe_texture *texture; /**< optional texture into which this is a view */ + unsigned face; + unsigned level; + unsigned zslice; }; -- cgit v1.2.3 From 4584c0efbd547559d30ba9a5e76549fc1b679619 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 13:47:09 +0100 Subject: draw: turn on SSE swizzle code --- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 90926aec85..07f85bc448 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -47,7 +47,7 @@ #include "tgsi/util/tgsi_parse.h" #define SSE_MAX_VERTICES 4 -#define SSE_SWIZZLES 0 +#define SSE_SWIZZLES 1 #if SSE_SWIZZLES typedef void (XSTDCALL *codegen_function) ( -- cgit v1.2.3 From be7e1297f4d5a2c2949968bea428b4c0e7c75d63 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 13:47:27 +0100 Subject: draw: squash warnings --- src/gallium/auxiliary/draw/draw_pt_varray.c | 1 - src/gallium/auxiliary/draw/draw_pt_vcache.c | 1 - 2 files changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index c85d8ded50..355093f945 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -200,7 +200,6 @@ static void varray_prepare(struct draw_pt_front_end *frontend, unsigned opt) { struct varray_frontend *varray = (struct varray_frontend *)frontend; - const struct pipe_rasterizer_state *rasterizer = varray->draw->rasterizer; if (opt & PT_PIPELINE) { diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 2f9775814f..6b3fb1406b 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -225,7 +225,6 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, unsigned opt ) { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; - const struct pipe_rasterizer_state *rasterizer = vcache->draw->rasterizer; if (opt & PT_PIPELINE) { -- cgit v1.2.3 From c6fadd9fade40e3397f56fdbfa2325861799e49c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 1 May 2008 10:20:31 +0900 Subject: gallium: Add newline to eof. --- src/gallium/auxiliary/util/p_debug.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index f1fb07bf5b..4ec1746662 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -422,4 +422,4 @@ void debug_print_format(const char *msg, unsigned fmt ) debug_printf("%s: %s\n", msg, fmtstr); } -#endif \ No newline at end of file +#endif -- cgit v1.2.3 From 073bb94a4150d7b3c78ea5420383cb7ed8f97566 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 1 May 2008 21:52:05 +0900 Subject: Add Brian's explanation for inheritance in C. --- src/gallium/README.portability | 65 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 65 insertions(+) (limited to 'src') diff --git a/src/gallium/README.portability b/src/gallium/README.portability index ab0c197847..18a97f449b 100644 --- a/src/gallium/README.portability +++ b/src/gallium/README.portability @@ -24,6 +24,7 @@ headers in general, should stricly follow these guidelines to ensure * Don't use variable number of macro arguments. Use static inline functions instead. +* Don't use C99 features. = Standard Library = @@ -42,3 +43,67 @@ portable way. * Use the functions/macros in p_debug.h. * Don't include assert.h, call abort, printf, etc. + + += Code Style = + +== Inherantice in C == + +The main thing we do is mimic inheritance by structure containment. + +Here's a silly made-up example: + +/* base class */ +struct buffer +{ + int size; + void (*validate)(struct buffer *buf); +}; + +/* sub-class of bufffer */ +struct texture_buffer +{ + struct buffer base; /* the base class, MUST COME FIRST! */ + int format; + int width, height; +}; + + +Then, we'll typically have cast-wrapper functions to convert base-class +pointers to sub-class pointers where needed: + +static inline struct vertex_buffer *vertex_buffer(struct buffer *buf) +{ + return (struct vertex_buffer *) buf; +} + + +To create/init a sub-classed object: + +struct buffer *create_texture_buffer(int w, int h, int format) +{ + struct texture_buffer *t = malloc(sizeof(*t)); + t->format = format; + t->width = w; + t->height = h; + t->base.size = w * h; + t->base.validate = tex_validate; + return &t->base; +} + +Example sub-class method: + +void tex_validate(struct buffer *buf) +{ + struct texture_buffer *tb = texture_buffer(buf); + assert(tb->format); + assert(tb->width); + assert(tb->height); +} + + +Note that we typically do not use typedefs to make "class names"; we use +'struct whatever' everywhere. + +Gallium's pipe_context and the subclassed psb_context, etc are prime examples +of this. There's also many examples in Mesa and the Mesa state tracker. -- cgit v1.2.3 From e97bedb302701e2e50cef664690b83a1fe6c95ed Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Thu, 1 May 2008 14:54:56 +0100 Subject: Fix build problem with MSVC --- src/mesa/main/context.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) mode change 100644 => 100755 src/mesa/main/context.c (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c old mode 100644 new mode 100755 index d06644f65d..7b8d934170 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1504,15 +1504,19 @@ _mesa_make_current( GLcontext *newCtx, GLframebuffer *drawBuffer, * or not bound to a user-created FBO. */ if (!newCtx->DrawBuffer || newCtx->DrawBuffer->Name == 0) { - _mesa_reference_framebuffer(&newCtx->DrawBuffer, drawBuffer); /* fix up the fb fields - these will end up wrong otherwise - if the DRIdrawable changes, and everything relies on them. - This is a bit messy (same as needed in _mesa_BindFramebufferEXT) */ + * if the DRIdrawable changes, and everything relies on them. + * This is a bit messy (same as needed in _mesa_BindFramebufferEXT) + */ int i; GLenum buffers[MAX_DRAW_BUFFERS]; + + _mesa_reference_framebuffer(&newCtx->DrawBuffer, drawBuffer); + for(i = 0; i < newCtx->Const.MaxDrawBuffers; i++) { buffers[i] = newCtx->Color.DrawBuffer[i]; } + _mesa_drawbuffers(newCtx, newCtx->Const.MaxDrawBuffers, buffers, NULL); } if (!newCtx->ReadBuffer || newCtx->ReadBuffer->Name == 0) { -- cgit v1.2.3 From 35b0efb8c6afd319ae36e99aa578ac6c75faf2f5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 14:19:25 +0100 Subject: gallium: do something sensible on the error path to try to avoid crashing in release builds --- src/gallium/auxiliary/util/u_gen_mipmap.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 0348629ab8..2e8417ee13 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -505,6 +505,9 @@ format_to_type_comps(enum pipe_format pformat, return; default: assert(0); + *datatype = UBYTE; + *comps = 0; + break; } } -- cgit v1.2.3 From 1e4217e1b857c6a3c5da7d1eceb74683bf0b9a00 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 14:51:18 +0100 Subject: tgsi: use ESI instead of EBX on non-win32 platforms --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 1138f59997..4587b0930a 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -36,7 +36,7 @@ #ifdef PIPE_ARCH_X86 -#define HIGH_PRECISION 1 /* for 1/sqrt() */ +#define HIGH_PRECISION 0 /* for 1/sqrt() */ #define FOR_EACH_CHANNEL( CHAN )\ @@ -2038,12 +2038,12 @@ static void aos_to_soa( struct x86_function *func, uint aos, uint soa, uint num, unsigned char *inner_loop; soa_input = x86_make_reg( file_REG32, reg_AX ); - aos_input = x86_make_reg( file_REG32, reg_BX ); + aos_input = get_temp_base(); /* BX or SI */ num_inputs = x86_make_reg( file_REG32, reg_CX ); temp = x86_make_reg( file_REG32, reg_DX ); /* Save EBX */ - x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); + x86_push( func, aos_input ); x86_mov( func, soa_input, get_argument( soa + 1 ) ); x86_mov( func, aos_input, get_argument( aos + 1 ) ); @@ -2088,7 +2088,7 @@ static void aos_to_soa( struct x86_function *func, uint aos, uint soa, uint num, x86_jcc( func, cc_NE, inner_loop ); /* Restore EBX */ - x86_pop( func, x86_make_reg( file_REG32, reg_BX ) ); + x86_pop( func, aos_input ); } static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, uint stride ) @@ -2100,12 +2100,12 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, unsigned char *inner_loop; soa_output = x86_make_reg( file_REG32, reg_AX ); - aos_output = x86_make_reg( file_REG32, reg_BX ); + aos_output = get_temp_base(); /* BX or SI */ num_outputs = x86_make_reg( file_REG32, reg_CX ); temp = x86_make_reg( file_REG32, reg_DX ); /* Save EBX */ - x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); + x86_push( func, aos_output ); x86_mov( func, soa_output, get_argument( soa + 1 ) ); x86_mov( func, aos_output, get_argument( aos + 1 ) ); @@ -2150,7 +2150,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, x86_jcc( func, cc_NE, inner_loop ); /* Restore EBX */ - x86_pop( func, x86_make_reg( file_REG32, reg_BX ) ); + x86_pop( func, aos_output ); } /** -- cgit v1.2.3 From 7810e7f623a47978cdd1a167cc9e6b743d56d949 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 15:13:46 +0100 Subject: tgsi: use x86_fn_arg instead of get_argument() -- it knows about push/pops to the stack --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 63 ++++++++++++++++------------- 1 file changed, 34 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 4587b0930a..5e133a9663 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -133,14 +133,6 @@ get_immediate_base( void ) * Data access helpers. */ -static struct x86_reg -get_argument( - unsigned index ) -{ - return x86_make_disp( - x86_make_reg( file_REG32, reg_SP ), - (index + 1) * 4 ); -} static struct x86_reg get_immediate( @@ -2045,14 +2037,14 @@ static void aos_to_soa( struct x86_function *func, uint aos, uint soa, uint num, /* Save EBX */ x86_push( func, aos_input ); - x86_mov( func, soa_input, get_argument( soa + 1 ) ); - x86_mov( func, aos_input, get_argument( aos + 1 ) ); - x86_mov( func, num_inputs, get_argument( num + 1 ) ); + x86_mov( func, soa_input, x86_fn_arg( func, soa ) ); + x86_mov( func, aos_input, x86_fn_arg( func, aos ) ); + x86_mov( func, num_inputs, x86_fn_arg( func, num ) ); /* do */ inner_loop = x86_get_label( func ); { - x86_mov( func, temp, get_argument( stride + 1 ) ); + x86_mov( func, temp, x86_fn_arg( func, stride ) ); x86_push( func, aos_input ); sse_movlps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); sse_movlps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); @@ -2107,9 +2099,9 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, /* Save EBX */ x86_push( func, aos_output ); - x86_mov( func, soa_output, get_argument( soa + 1 ) ); - x86_mov( func, aos_output, get_argument( aos + 1 ) ); - x86_mov( func, num_outputs, get_argument( num + 1 ) ); + x86_mov( func, soa_output, x86_fn_arg( func, soa ) ); + x86_mov( func, aos_output, x86_fn_arg( func, aos ) ); + x86_mov( func, num_outputs, x86_fn_arg( func, num ) ); /* do */ inner_loop = x86_get_label( func ); @@ -2126,7 +2118,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) ); sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) ); - x86_mov( func, temp, get_argument( stride + 1 ) ); + x86_mov( func, temp, x86_fn_arg( func, stride ) ); x86_push( func, aos_output ); sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); @@ -2185,6 +2177,13 @@ tgsi_emit_sse2( tgsi_parse_init( &parse, tokens ); + /* Can't just use EDI without save/restoring it: + */ + x86_push( + func, + get_immediate_base() ); + + /* * Different function args for vertex/fragment shaders: */ @@ -2193,51 +2192,51 @@ tgsi_emit_sse2( x86_mov( func, get_input_base(), - get_argument( 0 ) ); + x86_fn_arg( func, 1 ) ); /* skipping outputs argument here */ x86_mov( func, get_const_base(), - get_argument( 2 ) ); + x86_fn_arg( func, 3 ) ); x86_mov( func, get_temp_base(), - get_argument( 3 ) ); + x86_fn_arg( func, 4 ) ); x86_mov( func, get_coef_base(), - get_argument( 4 ) ); + x86_fn_arg( func, 5 ) ); x86_mov( func, get_immediate_base(), - get_argument( 5 ) ); + x86_fn_arg( func, 6 ) ); } else { assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX); if (do_swizzles) - aos_to_soa( func, 5, 0, 6, 7 ); + aos_to_soa( func, 6, 1, 7, 8 ); x86_mov( func, get_input_base(), - get_argument( 0 ) ); + x86_fn_arg( func, 1 ) ); x86_mov( func, get_output_base(), - get_argument( 1 ) ); + x86_fn_arg( func, 2 ) ); x86_mov( func, get_const_base(), - get_argument( 2 ) ); + x86_fn_arg( func, 3 ) ); x86_mov( func, get_temp_base(), - get_argument( 3 ) ); + x86_fn_arg( func, 4 ) ); x86_mov( func, get_immediate_base(), - get_argument( 4 ) ); + x86_fn_arg( func, 5 ) ); } while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { @@ -2260,7 +2259,7 @@ tgsi_emit_sse2( x86_mov( func, get_output_base(), - get_argument( 1 ) ); + x86_fn_arg( func, 2 ) ); } } @@ -2307,9 +2306,15 @@ tgsi_emit_sse2( if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) { if (do_swizzles) - soa_to_aos( func, 8, 1, 9, 10 ); + soa_to_aos( func, 9, 2, 10, 11 ); } + /* Can't just use EDI without save/restoring it: + */ + x86_pop( + func, + get_immediate_base() ); + #ifdef WIN32 emit_retw( func, 16 ); #else -- cgit v1.2.3 From 47aa416821b69d3afa33c79ec8cb8499688a0e8e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 15:27:53 +0100 Subject: tgsi: use EBX everywhere, be sure to push/pop it --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 18 ++++++++++-------- 1 file changed, 10 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 5e133a9663..06df3dbb05 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -103,15 +103,9 @@ get_output_base( void ) static struct x86_reg get_temp_base( void ) { -#ifdef WIN32 return x86_make_reg( file_REG32, reg_BX ); -#else - return x86_make_reg( - file_REG32, - reg_SI ); -#endif } static struct x86_reg @@ -2177,12 +2171,16 @@ tgsi_emit_sse2( tgsi_parse_init( &parse, tokens ); - /* Can't just use EDI without save/restoring it: + /* Can't just use EDI, EBX without save/restoring them: */ x86_push( func, get_immediate_base() ); + x86_push( + func, + get_temp_base() ); + /* * Different function args for vertex/fragment shaders: @@ -2309,8 +2307,12 @@ tgsi_emit_sse2( soa_to_aos( func, 9, 2, 10, 11 ); } - /* Can't just use EDI without save/restoring it: + /* Can't just use EBX, EDI without save/restoring them: */ + x86_pop( + func, + get_temp_base() ); + x86_pop( func, get_immediate_base() ); -- cgit v1.2.3 From 6980823da9120d8d8533f7a78eac2d63dece430f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 15:30:50 +0100 Subject: draw: avoid aliasing warning --- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index c4de9d2698..54fdcc1a3b 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -256,7 +256,7 @@ pstip_transform_inst(struct tgsi_transform_context *ctx, uint size = 4; immed = tgsi_default_full_immediate(); immed.Immediate.Size = 1 + size; /* one for the token itself */ - immed.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; + immed.u.Pointer = value; ctx->emit_immediate(ctx, &immed); } -- cgit v1.2.3 From 727257f32002544658219d2e0163993c1cbc5644 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 15:31:17 +0100 Subject: rtasm: assert stack is fully popped in return --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index d7e2230557..40f6f973d6 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -495,6 +495,7 @@ void x86_dec( struct x86_function *p, void x86_ret( struct x86_function *p ) { DUMP(); + assert(p->stack_offset == 0); emit_1ub(p, 0xc3); } -- cgit v1.2.3 From bf1e120b318ba8cbda7316fb8862acd44a3b459f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 15:32:03 +0100 Subject: mesa: squash harmless warning --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 524b5af50b..d2be450bdb 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -805,7 +805,7 @@ tgsi_translate_mesa_program( { GLboolean tempsUsed[MAX_PROGRAM_TEMPS + 1]; GLboolean inside_range = GL_FALSE; - GLuint start_range; + GLuint start_range = 0; find_temporaries(program, tempsUsed); tempsUsed[MAX_PROGRAM_TEMPS] = GL_FALSE; -- cgit v1.2.3 From afe67db8038855d9f7b4ce46b610701c55736c1f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 15:36:14 +0100 Subject: tgsi: add some const qualifiers to immediate pointers --- src/gallium/auxiliary/tgsi/util/tgsi_parse.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h index da0121c482..15e76feb7c 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h @@ -52,8 +52,8 @@ struct tgsi_full_immediate struct tgsi_immediate Immediate; union { - void *Pointer; - struct tgsi_immediate_float32 *ImmediateFloat32; + const void *Pointer; + const struct tgsi_immediate_float32 *ImmediateFloat32; } u; }; -- cgit v1.2.3 From 419f3c447520d1dc95c529afa693ffe3fffe5560 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 15:45:51 +0100 Subject: tgsi: restore HIGH_PRECISION setting --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 06df3dbb05..45453c34ce 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -36,7 +36,7 @@ #ifdef PIPE_ARCH_X86 -#define HIGH_PRECISION 0 /* for 1/sqrt() */ +#define HIGH_PRECISION 1 /* for 1/sqrt() */ #define FOR_EACH_CHANNEL( CHAN )\ -- cgit v1.2.3 From f1f52a8be98efa26c7c9bc480a2483fc2106d654 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 1 May 2008 17:30:17 +0100 Subject: gallium: Notify driver of texture updates in util_blit_pixels(). --- src/gallium/auxiliary/util/u_blit.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 9e9912c6e4..568d62ced1 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -295,6 +295,8 @@ util_blit_pixels(struct blit_state *ctx, src, srcLeft, srcTop, /* src */ srcW, srcH); /* size */ + pipe->texture_update(pipe, tex, 0, 1 << 0); + /* save state (restored below) */ cso_save_blend(ctx->cso); cso_save_depth_stencil_alpha(ctx->cso); -- cgit v1.2.3 From de7277f81a0f4330f11ec429bf46e93d0a748dfe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 1 May 2008 17:31:39 +0100 Subject: gallium: Set sampler->min_lod instead of always reallocating pipe_textures. --- src/mesa/state_tracker/st_atom_sampler.c | 10 +--------- src/mesa/state_tracker/st_cb_texture.c | 9 ++------- 2 files changed, 3 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index e1ddb53b80..0237da3693 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -37,6 +37,7 @@ #include "st_program.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" +#include "pipe/p_util.h" #include "cso_cache/cso_context.h" @@ -147,17 +148,8 @@ update_samplers(struct st_context *st) sampler->normalized_coords = 1; sampler->lod_bias = st->ctx->Texture.Unit[su].LodBias; -#if 1 - sampler->min_lod = (texobj->MinLod) < 0.0 ? 0.0 : texobj->MinLod; - sampler->max_lod = texobj->MaxLod; -#else - /* min/max lod should really be as follows (untested). - * Also, calculate_first_last_level() needs to be overhauled - * since today's hardware had real support for LOD clamping. - */ sampler->min_lod = MAX2(texobj->BaseLevel, texobj->MinLod); sampler->max_lod = MIN2(texobj->MaxLevel, texobj->MaxLod); -#endif sampler->border_color[0] = texobj->BorderColor[RCOMP]; sampler->border_color[1] = texobj->BorderColor[GCOMP]; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 981246221b..5fd192a8b8 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1342,8 +1342,6 @@ static void calculate_first_last_level(struct st_texture_object *stObj) { struct gl_texture_object *tObj = &stObj->base; - const struct gl_texture_image *const baseImage = - tObj->Image[0][tObj->BaseLevel]; /* These must be signed values. MinLod and MaxLod can be negative numbers, * and having firstLevel and lastLevel as signed prevents the need for @@ -1366,7 +1364,7 @@ calculate_first_last_level(struct st_texture_object *stObj) } else { firstLevel = 0; - lastLevel = MIN2(tObj->MaxLevel - tObj->BaseLevel, baseImage->MaxLog2); + lastLevel = MIN2(tObj->MaxLevel, tObj->Image[0][0]->WidthLog2); } break; case GL_TEXTURE_RECTANGLE_NV: @@ -1489,10 +1487,7 @@ st_finalize_texture(GLcontext *ctx, (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || stObj->pt->format != st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat) || - stObj->pt->last_level != stObj->lastLevel || - stObj->pt->width[0] != firstImage->base.Width2 || - stObj->pt->height[0] != firstImage->base.Height2 || - stObj->pt->depth[0] != firstImage->base.Depth2 || + stObj->pt->last_level < stObj->lastLevel || stObj->pt->cpp != cpp || stObj->pt->compressed != firstImage->base.IsCompressed)) { pipe_texture_release(&stObj->pt); -- cgit v1.2.3 From 0000792a2006a2c8fde1b54d070490a625fb8435 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 18:48:28 +0100 Subject: sct: fix bug in remove_context_from_surface --- src/gallium/auxiliary/sct/sct.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/sct/sct.c b/src/gallium/auxiliary/sct/sct.c index 97ee5882a1..5e4126e014 100644 --- a/src/gallium/auxiliary/sct/sct.c +++ b/src/gallium/auxiliary/sct/sct.c @@ -209,6 +209,7 @@ remove_context_from_surface(struct sct_surface *si, } else { prev = curr; + next = curr->next; } } } -- cgit v1.2.3 From f30285e99c1e158971855b12331df3da38555004 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 18:49:07 +0100 Subject: softpipe: fix warning --- src/gallium/drivers/softpipe/sp_state_fs.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 2921066ce3..9e77b7e91b 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -82,10 +82,9 @@ softpipe_bind_fs_state(struct pipe_context *pipe, void *fs) void softpipe_delete_fs_state(struct pipe_context *pipe, void *fs) { - struct softpipe_context *softpipe = softpipe_context(pipe); struct sp_fragment_shader *state = fs; - assert(fs != softpipe->fs); + assert(fs != softpipe_context(pipe)->fs); state->delete( state ); } -- cgit v1.2.3 From 26bcef898af4e6dfd578783ed33818a2bd38b06d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 18:49:52 +0100 Subject: i915: fix warning --- src/gallium/drivers/i915simple/i915_debug_fp.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_debug_fp.c b/src/gallium/drivers/i915simple/i915_debug_fp.c index 37a3508fe1..c024a051a5 100644 --- a/src/gallium/drivers/i915simple/i915_debug_fp.c +++ b/src/gallium/drivers/i915simple/i915_debug_fp.c @@ -333,12 +333,11 @@ void i915_disassemble_program(struct debug_stream *stream, const unsigned * program, unsigned sz) { - unsigned size = program[0] & 0x1ff; unsigned i; PRINTF(stream, "\t\tBEGIN\n"); - assert(size + 2 == sz); + assert((program[0] & 0x1ff) + 2 == sz); program++; for (i = 1; i < sz; i += 3, program += 3) { -- cgit v1.2.3 From b8936ca1c22de7b0cb695ee3b392e4473fd17aa0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 18:50:33 +0100 Subject: i915: avoid crashing on bad parameter --- src/gallium/drivers/i915simple/i915_screen.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 631642e1b6..646cfd921d 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -182,6 +182,7 @@ i915_is_format_supported( struct pipe_screen *screen, break; default: assert(0); + return FALSE; } for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) { -- cgit v1.2.3 From fb3623b235f5caa9d76e656b1e5eda797c7c73eb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 20:41:03 +0100 Subject: rtasm: fix labels after (not so) recent change to allow dynamic fn growth Using char * for labels doesn't work if you realloc the function during assembly and free the old storage... --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 49 +++++++++---------------- src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 14 +++---- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 4 +- src/gallium/auxiliary/translate/translate_sse.c | 2 +- 4 files changed, 28 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 40f6f973d6..e69251f072 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -347,9 +347,9 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg ) return x86_make_reg( reg.file, reg.idx ); } -unsigned char *x86_get_label( struct x86_function *p ) +int x86_get_label( struct x86_function *p ) { - return p->csr; + return p->csr - p->store; } @@ -361,17 +361,22 @@ unsigned char *x86_get_label( struct x86_function *p ) void x86_jcc( struct x86_function *p, enum x86_cc cc, - unsigned char *label ) + int label ) { - intptr_t offset = pointer_to_intptr( label ) - (pointer_to_intptr( x86_get_label(p) ) + 2); + int offset = label - (x86_get_label(p) + 2); DUMP_I(cc); + if (offset < 0) { + int amt = p->csr - p->store; + assert(amt > -offset); + } + if (offset <= 127 && offset >= -128) { emit_1ub(p, 0x70 + cc); emit_1b(p, (char) offset); } else { - offset = pointer_to_intptr( label ) - (pointer_to_intptr( x86_get_label(p) ) + 6); + offset = label - (x86_get_label(p) + 6); emit_2ub(p, 0x0f, 0x80 + cc); emit_1i(p, offset); } @@ -379,8 +384,8 @@ void x86_jcc( struct x86_function *p, /* Always use a 32bit offset for forward jumps: */ -unsigned char *x86_jcc_forward( struct x86_function *p, - enum x86_cc cc ) +int x86_jcc_forward( struct x86_function *p, + enum x86_cc cc ) { DUMP_I(cc); emit_2ub(p, 0x0f, 0x80 + cc); @@ -388,7 +393,7 @@ unsigned char *x86_jcc_forward( struct x86_function *p, return x86_get_label(p); } -unsigned char *x86_jmp_forward( struct x86_function *p) +int x86_jmp_forward( struct x86_function *p) { DUMP(); emit_1ub(p, 0xe9); @@ -396,7 +401,7 @@ unsigned char *x86_jmp_forward( struct x86_function *p) return x86_get_label(p); } -unsigned char *x86_call_forward( struct x86_function *p) +int x86_call_forward( struct x86_function *p) { DUMP(); @@ -408,42 +413,24 @@ unsigned char *x86_call_forward( struct x86_function *p) /* Fixup offset from forward jump: */ void x86_fixup_fwd_jump( struct x86_function *p, - unsigned char *fixup ) + int fixup ) { - *(int *)(fixup - 4) = pointer_to_intptr( x86_get_label(p) ) - pointer_to_intptr( fixup ); + *(int *)(p->store + fixup - 4) = x86_get_label(p) - fixup; } -void x86_jmp( struct x86_function *p, unsigned char *label) +void x86_jmp( struct x86_function *p, int label) { DUMP_I( label ); emit_1ub(p, 0xe9); - emit_1i(p, pointer_to_intptr( label ) - pointer_to_intptr( x86_get_label(p) ) - 4); -} - -#if 0 -static unsigned char *cptr( void (*label)() ) -{ - return (unsigned char *) label; + emit_1i(p, label - x86_get_label(p) - 4); } -/* This doesn't work once we start reallocating & copying the - * generated code on buffer fills, because the call is relative to the - * current pc. - */ -void x86_call( struct x86_function *p, void (*label)()) -{ - DUMP_I( label ); - emit_1ub(p, 0xe8); - emit_1i(p, cptr(label) - x86_get_label(p) - 4); -} -#else void x86_call( struct x86_function *p, struct x86_reg reg) { DUMP_R( reg ); emit_1ub(p, 0xff); emit_modrm_noreg(p, 2, reg); } -#endif /* michal: diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index ad79b1facf..eacaeeaf6f 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -124,23 +124,23 @@ struct x86_reg x86_get_base_reg( struct x86_reg reg ); /* Labels, jumps and fixup: */ -unsigned char *x86_get_label( struct x86_function *p ); +int x86_get_label( struct x86_function *p ); void x86_jcc( struct x86_function *p, enum x86_cc cc, - unsigned char *label ); + int label ); -unsigned char *x86_jcc_forward( struct x86_function *p, +int x86_jcc_forward( struct x86_function *p, enum x86_cc cc ); -unsigned char *x86_jmp_forward( struct x86_function *p); +int x86_jmp_forward( struct x86_function *p); -unsigned char *x86_call_forward( struct x86_function *p); +int x86_call_forward( struct x86_function *p); void x86_fixup_fwd_jump( struct x86_function *p, - unsigned char *fixup ); + int fixup ); -void x86_jmp( struct x86_function *p, unsigned char *label ); +void x86_jmp( struct x86_function *p, int label ); /* void x86_call( struct x86_function *p, void (*label)() ); */ void x86_call( struct x86_function *p, struct x86_reg reg); diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 45453c34ce..07db3292b4 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -2021,7 +2021,7 @@ static void aos_to_soa( struct x86_function *func, uint aos, uint soa, uint num, struct x86_reg aos_input; struct x86_reg num_inputs; struct x86_reg temp; - unsigned char *inner_loop; + int inner_loop; soa_input = x86_make_reg( file_REG32, reg_AX ); aos_input = get_temp_base(); /* BX or SI */ @@ -2083,7 +2083,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, struct x86_reg aos_output; struct x86_reg num_outputs; struct x86_reg temp; - unsigned char *inner_loop; + int inner_loop; soa_output = x86_make_reg( file_REG32, reg_AX ); aos_output = get_temp_base(); /* BX or SI */ diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index f590d48b78..a54ac5a82f 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -404,7 +404,7 @@ static boolean build_vertex_emit( struct translate_sse *p, struct x86_reg srcEAX = x86_make_reg(file_REG32, reg_CX); struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP); struct x86_reg translateESI = x86_make_reg(file_REG32, reg_SI); - uint8_t *fixup, *label; + int fixup, label; unsigned j; p->func = func; -- cgit v1.2.3 From 2004b8a769110456e66d040398eacf25c8592710 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 20:42:44 +0100 Subject: draw: label fn args -- shouldn't this be defined where the fn is created? --- src/gallium/auxiliary/draw/draw_vs_sse.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 07f85bc448..a57c938fbf 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -51,17 +51,17 @@ #if SSE_SWIZZLES typedef void (XSTDCALL *codegen_function) ( - const struct tgsi_exec_vector *input, - struct tgsi_exec_vector *output, - float (*constant)[4], - struct tgsi_exec_vector *temporary, - float (*immediates)[4], - const float (*aos_input)[4], - uint num_inputs, - uint input_stride, - float (*aos_output)[4], - uint num_outputs, - uint output_stride ); + const struct tgsi_exec_vector *input, /* 1 */ + struct tgsi_exec_vector *output, /* 2 */ + float (*constant)[4], /* 3 */ + struct tgsi_exec_vector *temporary, /* 4 */ + float (*immediates)[4], /* 5 */ + const float (*aos_input)[4], /* 6 */ + uint num_inputs, /* 7 */ + uint input_stride, /* 8 */ + float (*aos_output)[4], /* 9 */ + uint num_outputs, /* 10 */ + uint output_stride ); /* 11 */ #else typedef void (XSTDCALL *codegen_function) ( const struct tgsi_exec_vector *input, -- cgit v1.2.3 From 2c89b75e36fd35d5a003107d1d2f97b537321f95 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 20:44:41 +0100 Subject: rtasm: learn another version of push --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index e69251f072..4e036d9032 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -449,8 +449,15 @@ void x86_push( struct x86_function *p, struct x86_reg reg ) { DUMP_R( reg ); - assert(reg.mod == mod_REG); - emit_1ub(p, 0x50 + reg.idx); + if (reg.mod == mod_REG) + emit_1ub(p, 0x50 + reg.idx); + else + { + emit_1ub(p, 0xff); + emit_modrm_noreg(p, 6, reg); + } + + p->stack_offset += 4; } -- cgit v1.2.3 From f067c6c452bdd5f5cc6b0f6b2f79fb3fc1162822 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 20:45:15 +0100 Subject: tgsi: remove some bogus win vs. linux crud Pass arguments properly in linux now. Still need to change this to use a single calling convention on both platforms. --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 118 +++++++++++----------------- 1 file changed, 46 insertions(+), 72 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 07db3292b4..a67bc8567b 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -441,19 +441,13 @@ emit_push_gp( { x86_push( func, - get_const_base() ); + x86_make_reg( file_REG32, reg_AX) ); x86_push( func, - get_input_base() ); + x86_make_reg( file_REG32, reg_CX) ); x86_push( func, - get_output_base() ); - - /* It is important on non-win32 platforms that temp base is pushed last. - */ - x86_push( - func, - get_temp_base() ); + x86_make_reg( file_REG32, reg_DX) ); } static void @@ -464,16 +458,13 @@ x86_pop_gp( */ x86_pop( func, - get_temp_base() ); + x86_make_reg( file_REG32, reg_DX) ); x86_pop( func, - get_output_base() ); + x86_make_reg( file_REG32, reg_CX) ); x86_pop( func, - get_input_base() ); - x86_pop( - func, - get_const_base() ); + x86_make_reg( file_REG32, reg_AX) ); } static void @@ -490,19 +481,23 @@ emit_func_call_dst( emit_push_gp( func ); -#ifdef WIN32 - x86_push( - func, - get_temp( TEMP_R0, 0 ) ); -#endif - { struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); + x86_lea( + func, + ecx, + get_temp( TEMP_R0, 0 ) ); + + x86_push( func, ecx ); x86_mov_reg_imm( func, ecx, (unsigned long) code ); x86_call( func, ecx ); +#ifndef WIN32 + x86_pop(func, ecx ); +#endif } + x86_pop_gp( func ); @@ -563,11 +558,7 @@ static void XSTDCALL cos4f( float *store ) { -#ifdef WIN32 const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif store[X + 0] = cosf( store[X + 0] ); store[X + 1] = cosf( store[X + 1] ); @@ -590,11 +581,8 @@ static void XSTDCALL ex24f( float *store ) { -#ifdef WIN32 const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif + store[X + 0] = powf( 2.0f, store[X + 0] ); store[X + 1] = powf( 2.0f, store[X + 1] ); store[X + 2] = powf( 2.0f, store[X + 2] ); @@ -627,11 +615,8 @@ static void XSTDCALL flr4f( float *store ) { -#ifdef WIN32 const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif + store[X + 0] = floorf( store[X + 0] ); store[X + 1] = floorf( store[X + 1] ); store[X + 2] = floorf( store[X + 2] ); @@ -653,11 +638,8 @@ static void XSTDCALL frc4f( float *store ) { -#ifdef WIN32 const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif + store[X + 0] -= floorf( store[X + 0] ); store[X + 1] -= floorf( store[X + 1] ); store[X + 2] -= floorf( store[X + 2] ); @@ -679,11 +661,8 @@ static void XSTDCALL lg24f( float *store ) { -#ifdef WIN32 const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif + store[X + 0] = LOG2( store[X + 0] ); store[X + 1] = LOG2( store[X + 1] ); store[X + 2] = LOG2( store[X + 2] ); @@ -741,11 +720,8 @@ static void XSTDCALL pow4f( float *store ) { -#ifdef WIN32 const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif + store[X + 0] = powf( store[X + 0], store[X + 4] ); store[X + 1] = powf( store[X + 1], store[X + 5] ); store[X + 2] = powf( store[X + 2], store[X + 6] ); @@ -786,11 +762,8 @@ static void XSTDCALL rsqrt4f( float *store ) { -#ifdef WIN32 const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif + store[X + 0] = 1.0F / sqrtf( store[X + 0] ); store[X + 1] = 1.0F / sqrtf( store[X + 1] ); store[X + 2] = 1.0F / sqrtf( store[X + 2] ); @@ -864,11 +837,8 @@ static void XSTDCALL sin4f( float *store ) { -#ifdef WIN32 const unsigned X = 0; -#else - const unsigned X = TEMP_R0 * 16; -#endif + store[X + 0] = sinf( store[X + 0] ); store[X + 1] = sinf( store[X + 1] ); store[X + 2] = sinf( store[X + 2] ); @@ -2015,40 +1985,40 @@ emit_declaration( } } -static void aos_to_soa( struct x86_function *func, uint aos, uint soa, uint num, uint stride ) +static void aos_to_soa( struct x86_function *func, + uint arg_aos, + uint arg_soa, + uint arg_num, + uint arg_stride ) { - struct x86_reg soa_input; - struct x86_reg aos_input; - struct x86_reg num_inputs; - struct x86_reg temp; + struct x86_reg soa_input = x86_make_reg( file_REG32, reg_AX ); + struct x86_reg aos_input = x86_make_reg( file_REG32, reg_BX ); + struct x86_reg num_inputs = x86_make_reg( file_REG32, reg_CX ); + struct x86_reg stride = x86_make_reg( file_REG32, reg_DX ); int inner_loop; - soa_input = x86_make_reg( file_REG32, reg_AX ); - aos_input = get_temp_base(); /* BX or SI */ - num_inputs = x86_make_reg( file_REG32, reg_CX ); - temp = x86_make_reg( file_REG32, reg_DX ); /* Save EBX */ - x86_push( func, aos_input ); + x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); - x86_mov( func, soa_input, x86_fn_arg( func, soa ) ); - x86_mov( func, aos_input, x86_fn_arg( func, aos ) ); - x86_mov( func, num_inputs, x86_fn_arg( func, num ) ); + x86_mov( func, aos_input, x86_fn_arg( func, arg_aos ) ); + x86_mov( func, soa_input, x86_fn_arg( func, arg_soa ) ); + x86_mov( func, num_inputs, x86_fn_arg( func, arg_num ) ); + x86_mov( func, stride, x86_fn_arg( func, arg_stride ) ); /* do */ inner_loop = x86_get_label( func ); { - x86_mov( func, temp, x86_fn_arg( func, stride ) ); x86_push( func, aos_input ); sse_movlps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); sse_movlps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); - x86_add( func, aos_input, temp ); + x86_add( func, aos_input, stride ); sse_movhps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); sse_movhps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); - x86_add( func, aos_input, temp ); + x86_add( func, aos_input, stride ); sse_movlps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); sse_movlps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); - x86_add( func, aos_input, temp ); + x86_add( func, aos_input, stride ); sse_movhps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); sse_movhps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); x86_pop( func, aos_input ); @@ -2086,7 +2056,7 @@ static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, int inner_loop; soa_output = x86_make_reg( file_REG32, reg_AX ); - aos_output = get_temp_base(); /* BX or SI */ + aos_output = x86_make_reg( file_REG32, reg_BX ); num_outputs = x86_make_reg( file_REG32, reg_CX ); temp = x86_make_reg( file_REG32, reg_DX ); @@ -2213,7 +2183,11 @@ tgsi_emit_sse2( assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX); if (do_swizzles) - aos_to_soa( func, 6, 1, 7, 8 ); + aos_to_soa( func, + 6, /* aos_input */ + 1, /* machine->input */ + 7, /* num_inputs */ + 8 ); /* input_stride */ x86_mov( func, -- cgit v1.2.3 From 4687272b20f92184a838fe2187857162a0a90a06 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 1 May 2008 15:08:05 -0600 Subject: gallium: use screen->tex_surface_release() --- src/mesa/state_tracker/st_texture.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 482a054f64..d6268fc80c 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -327,7 +327,7 @@ st_texture_image_copy(struct pipe_context *pipe, 0, 0, /* srcX, Y */ width, copyHeight); - pipe_surface_reference(&dst_surface, NULL); - pipe_surface_reference(&src_surface, NULL); + screen->tex_surface_release(screen, &src_surface); + screen->tex_surface_release(screen, &dst_surface); } } -- cgit v1.2.3 From c1abd758c51247ebaf3d4808a77513d7814205cd Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 1 May 2008 15:19:00 -0600 Subject: gallium: remove the unused softpipe_winsys code The struct is still there though until all winsys layers are updated --- src/gallium/drivers/softpipe/sp_context.c | 4 +- src/gallium/drivers/softpipe/sp_context.h | 2 - src/gallium/drivers/softpipe/sp_winsys.h | 2 +- src/gallium/winsys/xlib/xm_winsys.c | 81 +++++++++---------------------- 4 files changed, 25 insertions(+), 64 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index edf91ecafa..fe9cd8375e 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -122,7 +122,7 @@ static void softpipe_destroy( struct pipe_context *pipe ) struct pipe_context * softpipe_create( struct pipe_screen *screen, struct pipe_winsys *pipe_winsys, - struct softpipe_winsys *softpipe_winsys ) + void *unused ) { struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); uint i; @@ -212,8 +212,6 @@ softpipe_create( struct pipe_screen *screen, softpipe->quad.colormask = sp_quad_colormask_stage(softpipe); softpipe->quad.output = sp_quad_output_stage(softpipe); - softpipe->winsys = softpipe_winsys; - /* * Create drawing context and plug our rendering stage into it. */ diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index b3e2b2e435..62eabfb30e 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -57,8 +57,6 @@ struct sp_vertex_shader; struct softpipe_context { struct pipe_context pipe; /**< base class */ - struct softpipe_winsys *winsys; /**< window system interface */ - /* The most recent drawing state as set by the driver: */ diff --git a/src/gallium/drivers/softpipe/sp_winsys.h b/src/gallium/drivers/softpipe/sp_winsys.h index 291825dfe2..4ab666486c 100644 --- a/src/gallium/drivers/softpipe/sp_winsys.h +++ b/src/gallium/drivers/softpipe/sp_winsys.h @@ -59,7 +59,7 @@ struct pipe_context; struct pipe_context *softpipe_create( struct pipe_screen *, struct pipe_winsys *, - struct softpipe_winsys * ); + void *unused ); struct pipe_screen * diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 8a89278cde..71ba076bff 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -37,6 +37,7 @@ #include "xmesaP.h" #undef ASSERT +#undef Elements #include "pipe/p_winsys.h" #include "pipe/p_format.h" @@ -88,18 +89,6 @@ struct xmesa_surface }; -/** - * Derived from softpipe_winsys. - * We just need one extra field which indicates the pixel format to use for - * drawing surfaces so that we're compatible with the XVisual/window format. - */ -struct xmesa_softpipe_winsys -{ - struct softpipe_winsys spws; - enum pipe_format pixelformat; -}; - - struct xmesa_pipe_winsys { struct pipe_winsys base; @@ -119,12 +108,7 @@ xmesa_surface(struct pipe_surface *ps) return (struct xmesa_surface *) ps; } -/** cast wrapper */ -static INLINE struct xmesa_softpipe_winsys * -xmesa_softpipe_winsys(struct softpipe_winsys *spws) -{ - return (struct xmesa_softpipe_winsys *) spws; -} + /** * Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque @@ -525,8 +509,7 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys, /** - * Called via pipe->surface_alloc() to create new surfaces (textures, - * renderbuffers, etc. + * Called via winsys->surface_alloc() to create new surfaces. */ static struct pipe_surface * xm_surface_alloc(struct pipe_winsys *ws) @@ -610,10 +593,19 @@ xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis) { static struct xmesa_pipe_winsys *ws = NULL; - if (!ws && getenv("XM_AUB")) { + if (!ws) { ws = (struct xmesa_pipe_winsys *) xmesa_create_pipe_winsys_aub(); } - else if (!ws) { + return &ws->base; +} + + +struct pipe_winsys * +xmesa_get_pipe_winsys(struct xmesa_visual *xm_vis) +{ + static struct xmesa_pipe_winsys *ws = NULL; + + if (!ws) { ws = CALLOC_STRUCT(xmesa_pipe_winsys); ws->xm_visual = xm_vis; @@ -644,45 +636,19 @@ xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis) } -/** - * Called via softpipe_winsys->is_format_supported(). - * This function is only called to test formats for front/back color surfaces. - * The winsys being queried will have been created at glXCreateContext - * time, with a pixel format corresponding to the context's visual. - */ -static boolean -xmesa_is_format_supported(struct softpipe_winsys *sws, - enum pipe_format format) -{ - struct xmesa_softpipe_winsys *xmws = xmesa_softpipe_winsys(sws); - return (format == xmws->pixelformat); -} - - -/** - * Return pointer to a softpipe_winsys object. - */ -static struct softpipe_winsys * -xmesa_get_softpipe_winsys(uint pixelformat) -{ - struct xmesa_softpipe_winsys *xmws - = CALLOC_STRUCT(xmesa_softpipe_winsys); - if (!xmws) - return NULL; - - xmws->spws.is_format_supported = xmesa_is_format_supported; - xmws->pixelformat = pixelformat; - - return &xmws->spws; -} - - struct pipe_context * xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) { - struct pipe_winsys *pws = xmesa_get_pipe_winsys_aub(xmesa->xm_visual); + struct pipe_winsys *pws; struct pipe_context *pipe; + if (getenv("XM_AUB")) { + pws = xmesa_get_pipe_winsys_aub(xmesa->xm_visual); + } + else { + pws = xmesa_get_pipe_winsys(xmesa->xm_visual); + } + #ifdef GALLIUM_CELL if (!getenv("GALLIUM_NOCELL")) { struct cell_winsys *cws = cell_get_winsys(pixelformat); @@ -693,10 +659,9 @@ xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) else #endif { - struct softpipe_winsys *spws = xmesa_get_softpipe_winsys(pixelformat); struct pipe_screen *screen = softpipe_create_screen(pws); - pipe = softpipe_create(screen, pws, spws); + pipe = softpipe_create(screen, pws, NULL); } if (pipe) -- cgit v1.2.3 From bc4952987419d77fabbf4fa43913f6e488bdb211 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 1 May 2008 15:21:40 -0600 Subject: added cast for MSVC --- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 54fdcc1a3b..73ee419858 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -256,7 +256,7 @@ pstip_transform_inst(struct tgsi_transform_context *ctx, uint size = 4; immed = tgsi_default_full_immediate(); immed.Immediate.Size = 1 + size; /* one for the token itself */ - immed.u.Pointer = value; + immed.u.Pointer = (void *) value; ctx->emit_immediate(ctx, &immed); } -- cgit v1.2.3 From 7a4313b63bcd06318437d384875472e7139070a1 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 May 2008 18:42:01 -0600 Subject: gallium: implement TGSI_OPCODE_LOG --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 30 +++++++++++++++-------------- 1 file changed, 16 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index d55f907c0d..4b33d7e8bd 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -1549,22 +1549,24 @@ exec_instruction( break; case TGSI_OPCODE_LOG: - debug_printf("TGSI: LOG opcode not implemented\n"); - /* from ARB_v_p: - tmp = fabs(ScalarLoad(op0)); - result.x = floor(log2(tmp)); - result.y = tmp / 2^(floor(log2(tmp))); - result.z = RoughApproxLog2(tmp); - result.w = 1.0; - */ -#if 0 - /* something like this: */ FETCH( &r[0], 0, CHAN_X ); - micro_lg2( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + micro_abs( &r[2], &r[0] ); /* r2 = abs(r0) */ + micro_lg2( &r[1], &r[2] ); /* r1 = lg2(r2) */ + micro_flr( &r[0], &r[1] ); /* r0 = floor(r1) */ + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &r[0], 0, CHAN_X ); + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_exp2( &r[0], &r[0] ); /* r0 = 2 ^ r0 */ + micro_div( &r[0], &r[2], &r[0] ); /* r0 = r2 / r0 */ + STORE( &r[0], 0, CHAN_Y ); + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( &r[1], 0, CHAN_Z ); + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); } -#endif break; case TGSI_OPCODE_MUL: -- cgit v1.2.3 From 3b63bc8ac6db7af4077f12cfd44876a9d43cc6ec Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 May 2008 18:49:20 -0600 Subject: gallium: implement TGSI_OPCODE_EXP --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 29 +++++++++++++++-------------- 1 file changed, 15 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 4b33d7e8bd..5d5125f7cb 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -1530,22 +1530,23 @@ exec_instruction( break; case TGSI_OPCODE_EXP: - debug_printf("TGSI: EXP opcode not implemented\n"); - /* from ARB_v_p: - tmp = ScalarLoad(op0); - result.x = 2^floor(tmp); - result.y = tmp - floor(tmp); - result.z = RoughApprox2ToX(tmp); - result.w = 1.0; - */ -#if 0 - /* something like this: */ FETCH( &r[0], 0, CHAN_X ); - micro_exp2( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); + micro_flr( &r[1], &r[0] ); /* r1 = floor(r0) */ + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + micro_exp2( &r[2], &r[1] ); /* r2 = 2 ^ r1 */ + STORE( &r[2], 0, CHAN_X ); /* store r2 */ + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_sub( &r[2], &r[0], &r[1] ); /* r2 = r0 - r1 */ + STORE( &r[2], 0, CHAN_Y ); /* store r2 */ + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + micro_exp2( &r[2], &r[0] ); /* r2 = 2 ^ r0 */ + STORE( &r[2], 0, CHAN_Z ); /* store r2 */ + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); } -#endif break; case TGSI_OPCODE_LOG: -- cgit v1.2.3 From 869b0836c1c4339de91c9918ae07926c846a004c Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 May 2008 18:56:20 -0600 Subject: gallium: temporarily disable broken SSE2 code for ARL opcode --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index a67bc8567b..2fd76a3072 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -1190,11 +1190,16 @@ emit_instruction( switch( inst->Instruction.Opcode ) { case TGSI_OPCODE_ARL: +#if 0 + /* XXX this isn't working properly (see glean vertProg1 test) */ FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); emit_f2it( func, 0 ); STORE( func, *inst, 0, 0, chan_index ); } +#else + return 0; +#endif break; case TGSI_OPCODE_MOV: -- cgit v1.2.3 From 9d151a2517de3f83d676624a21b4f73d5accecbe Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 1 May 2008 16:39:54 +0200 Subject: tgsi: Dump destination register modulate modifier. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 26bfc2051f..4c65ffd780 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -767,6 +767,31 @@ dump_instruction_short( SID( dst->DstRegister.Index ); CHR( ']' ); + switch (dst->DstRegisterExtModulate.Modulate) { + case TGSI_MODULATE_1X: + break; + case TGSI_MODULATE_2X: + TXT( "_2X" ); + break; + case TGSI_MODULATE_4X: + TXT( "_4X" ); + break; + case TGSI_MODULATE_8X: + TXT( "_8X" ); + break; + case TGSI_MODULATE_HALF: + TXT( "_D2" ); + break; + case TGSI_MODULATE_QUARTER: + TXT( "_D4" ); + break; + case TGSI_MODULATE_EIGHTH: + TXT( "_D8" ); + break; + default: + assert( 0 ); + } + if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { CHR( '.' ); if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { -- cgit v1.2.3 From 36f93c5e5159ebd99a5a4504efccdf6c5bf40716 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 2 May 2008 10:20:53 +0200 Subject: tgsi: Fix build on Win32. --- src/gallium/auxiliary/tgsi/util/tgsi_parse.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c index c3526cb71f..5bea773840 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c @@ -43,7 +43,7 @@ tgsi_full_token_free( union tgsi_full_token *full_token ) { if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { - FREE( full_token->FullImmediate.u.Pointer ); + FREE( (void *) full_token->FullImmediate.u.Pointer ); } } @@ -156,7 +156,7 @@ tgsi_parse_token( imm->u.Pointer = MALLOC( sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - next_token( ctx, &imm->u.ImmediateFloat32[i] ); + next_token( ctx, (struct tgsi_immediate_float32 *) &imm->u.ImmediateFloat32[i] ); } break; -- cgit v1.2.3 From a1cb0c2b915532e934b5d37bd0c550b1bfcc77ba Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 2 May 2008 11:13:58 +0200 Subject: tgsi: Do not assume IN and OUT registers are declared sequentially. --- src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 15 +++++++-------- 1 file changed, 7 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c index ea4a72967d..65650ed22a 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -103,18 +103,14 @@ tgsi_scan_shader(const struct tgsi_token *tokens, info->file_max[file] = MAX2(info->file_max[file], (int)i); if (file == TGSI_FILE_INPUT) { - info->input_semantic_name[info->num_inputs] - = (ubyte)fulldecl->Semantic.SemanticName; - info->input_semantic_index[info->num_inputs] - = (ubyte)fulldecl->Semantic.SemanticIndex; + info->input_semantic_name[i] = (ubyte)fulldecl->Semantic.SemanticName; + info->input_semantic_index[i] = (ubyte)fulldecl->Semantic.SemanticIndex; info->num_inputs++; } if (file == TGSI_FILE_OUTPUT) { - info->output_semantic_name[info->num_outputs] - = (ubyte)fulldecl->Semantic.SemanticName; - info->output_semantic_index[info->num_outputs] - = (ubyte)fulldecl->Semantic.SemanticIndex; + info->output_semantic_name[i] = (ubyte)fulldecl->Semantic.SemanticName; + info->output_semantic_index[i] = (ubyte)fulldecl->Semantic.SemanticIndex; info->num_outputs++; } @@ -137,6 +133,9 @@ tgsi_scan_shader(const struct tgsi_token *tokens, } } + assert( info->file_max[TGSI_FILE_INPUT] + 1 == info->num_inputs ); + assert( info->file_max[TGSI_FILE_OUTPUT] + 1 == info->num_outputs ); + info->uses_kill = (info->opcode_count[TGSI_OPCODE_KIL] || info->opcode_count[TGSI_OPCODE_KILP]); -- cgit v1.2.3 From 251db95945c6b484a093336e7bf12aed6091de54 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 17:55:52 +0100 Subject: cso: can memcmp-compare pipe_framebuffer_state now it includes fb dimensions --- src/gallium/auxiliary/cso_cache/cso_context.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index eef898f486..5d626b7cdc 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -769,8 +769,7 @@ void cso_restore_vertex_shader(struct cso_context *ctx) enum pipe_error cso_set_framebuffer(struct cso_context *ctx, const struct pipe_framebuffer_state *fb) { - /* XXX this memcmp() fails to detect buffer size changes */ - if (1/*memcmp(&ctx->fb, fb, sizeof(*fb))*/) { + if (memcmp(&ctx->fb, fb, sizeof(*fb)) != 0) { ctx->fb = *fb; ctx->pipe->set_framebuffer_state(ctx->pipe, fb); } -- cgit v1.2.3 From 731e7b961cd081ac6a64b636937716ce3a623c2c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 1 May 2008 18:13:46 +0100 Subject: re-add pipe_surface map/unmap inlines --- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- src/gallium/drivers/softpipe/sp_surface.c | 2 +- src/gallium/include/pipe/p_inlines.h | 33 +++++++++++++++++++++++++++++++ 3 files changed, 35 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 90926aec85..07f85bc448 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -47,7 +47,7 @@ #include "tgsi/util/tgsi_parse.h" #define SSE_MAX_VERTICES 4 -#define SSE_SWIZZLES 0 +#define SSE_SWIZZLES 1 #if SSE_SWIZZLES typedef void (XSTDCALL *codegen_function) ( diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index b82b1a8f37..29a1e92416 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -46,7 +46,6 @@ sp_surface_copy(struct pipe_context *pipe, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - assert( dst->cpp == src->cpp ); void *dst_map = pipe->screen->surface_map( pipe->screen, dst, PIPE_BUFFER_USAGE_CPU_WRITE ); @@ -55,6 +54,7 @@ sp_surface_copy(struct pipe_context *pipe, src, PIPE_BUFFER_USAGE_CPU_READ ); + assert( dst->cpp == src->cpp ); assert(src_map && dst_map); pipe_copy_rect(dst_map, diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h index 592c3c87c2..1e4b98edb4 100644 --- a/src/gallium/include/pipe/p_inlines.h +++ b/src/gallium/include/pipe/p_inlines.h @@ -39,6 +39,39 @@ extern "C" { #endif +/* XXX: these are a kludge. will fix when all surfaces are views into + * textures, and free-floating winsys surfaces go away. + */ +static INLINE void * +pipe_surface_map( struct pipe_surface *surf, unsigned flags ) +{ + if (surf->texture) { + struct pipe_screen *screen = surf->texture->screen; + return surf->texture->screen->surface_map( screen, surf, flags ); + } + else { + struct pipe_winsys *winsys = surf->winsys; + char *map = (char *)winsys->buffer_map( winsys, surf->buffer, flags ); + if (map == NULL) + return NULL; + return (void *)(map + surf->offset); + } +} + +static INLINE void +pipe_surface_unmap( struct pipe_surface *surf ) +{ + if (surf->texture) { + struct pipe_screen *screen = surf->texture->screen; + surf->texture->screen->surface_unmap( screen, surf ); + } + else { + struct pipe_winsys *winsys = surf->winsys; + winsys->buffer_unmap( winsys, surf->buffer ); + } +} + + /** * Set 'ptr' to point to 'surf' and update reference counting. -- cgit v1.2.3 From 54507125e735ffa595e252282eaabf38095c21e1 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Fri, 2 May 2008 10:08:03 +0000 Subject: Some changed for non-C99 compilers --- src/mesa/state_tracker/st_atom_blend.c | 15 +++++--------- src/mesa/state_tracker/st_atom_clip.c | 15 +++++--------- src/mesa/state_tracker/st_atom_constbuf.c | 20 +++++++++--------- src/mesa/state_tracker/st_atom_depth.c | 10 ++++----- src/mesa/state_tracker/st_atom_fixedfunction.c | 10 ++++----- src/mesa/state_tracker/st_atom_framebuffer.c | 10 ++++----- src/mesa/state_tracker/st_atom_pixeltransfer.c | 10 ++++----- src/mesa/state_tracker/st_atom_rasterizer.c | 12 +++++------ src/mesa/state_tracker/st_atom_sampler.c | 15 +++++--------- src/mesa/state_tracker/st_atom_scissor.c | 15 +++++--------- src/mesa/state_tracker/st_atom_shader.c | 10 ++++----- src/mesa/state_tracker/st_atom_stipple.c | 10 ++++----- src/mesa/state_tracker/st_atom_texture.c | 15 +++++--------- src/mesa/state_tracker/st_atom_viewport.c | 10 ++++----- src/mesa/state_tracker/st_cb_clear.c | 2 +- src/mesa/state_tracker/st_draw.c | 5 +++-- src/mesa/state_tracker/st_extensions.c | 28 +++++++++++++------------- 17 files changed, 94 insertions(+), 118 deletions(-) mode change 100644 => 100755 src/mesa/state_tracker/st_atom_blend.c mode change 100644 => 100755 src/mesa/state_tracker/st_atom_clip.c mode change 100644 => 100755 src/mesa/state_tracker/st_atom_constbuf.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c old mode 100644 new mode 100755 index 2a0e92245c..35c09c3e08 --- a/src/mesa/state_tracker/st_atom_blend.c +++ b/src/mesa/state_tracker/st_atom_blend.c @@ -223,15 +223,10 @@ update_blend( struct st_context *st ) const struct st_tracked_state st_update_blend = { - .name = "st_update_blend", - .dirty = { - .mesa = (_NEW_COLOR), /* XXX _NEW_BLEND someday? */ - .st = 0, + "st_update_blend", /* name */ + { /* dirty */ + (_NEW_COLOR), /* XXX _NEW_BLEND someday? */ /* mesa */ + 0, /* st */ }, - .update = update_blend + update_blend, /* update */ }; - - - - - diff --git a/src/mesa/state_tracker/st_atom_clip.c b/src/mesa/state_tracker/st_atom_clip.c old mode 100644 new mode 100755 index a6f0568660..23d709b814 --- a/src/mesa/state_tracker/st_atom_clip.c +++ b/src/mesa/state_tracker/st_atom_clip.c @@ -62,15 +62,10 @@ static void update_clip( struct st_context *st ) const struct st_tracked_state st_update_clip = { - .name = "st_update_clip", - .dirty = { - .mesa = (_NEW_TRANSFORM), - .st = 0, + "st_update_clip", /* name */ + { /* dirty */ + (_NEW_TRANSFORM), /* mesa */ + 0, /* st */ }, - .update = update_clip + update_clip /* update */ }; - - - - - diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c old mode 100644 new mode 100755 index 2b659aebbc..2856e0f0e0 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -113,12 +113,12 @@ static void update_vs_constants(struct st_context *st ) } const struct st_tracked_state st_update_vs_constants = { - .name = "st_update_vs_constants", - .dirty = { - .mesa = 0, /* set dynamically above */ - .st = ST_NEW_VERTEX_PROGRAM, + "st_update_vs_constants", /* name */ + { /* dirty */ + 0, /* set dynamically above */ /* mesa */ + ST_NEW_VERTEX_PROGRAM, /* st */ }, - .update = update_vs_constants + update_vs_constants /* update */ }; /* Fragment shader: @@ -132,11 +132,11 @@ static void update_fs_constants(struct st_context *st ) } const struct st_tracked_state st_update_fs_constants = { - .name = "st_update_fs_constants", - .dirty = { - .mesa = 0, /* set dynamically above */ - .st = ST_NEW_FRAGMENT_PROGRAM, + "st_update_fs_constants", /* name */ + { /* dirty */ + 0, /* set dynamically above */ /* mesa */ + ST_NEW_FRAGMENT_PROGRAM, /* st */ }, - .update = update_fs_constants + update_fs_constants /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_depth.c b/src/mesa/state_tracker/st_atom_depth.c index ef467582c0..0e791ceb20 100644 --- a/src/mesa/state_tracker/st_atom_depth.c +++ b/src/mesa/state_tracker/st_atom_depth.c @@ -142,10 +142,10 @@ update_depth_stencil_alpha(struct st_context *st) const struct st_tracked_state st_update_depth_stencil_alpha = { - .name = "st_update_depth_stencil", - .dirty = { - .mesa = (_NEW_DEPTH|_NEW_STENCIL|_NEW_COLOR), - .st = 0, + "st_update_depth_stencil", /* name */ + { /* dirty */ + (_NEW_DEPTH|_NEW_STENCIL|_NEW_COLOR), /* mesa */ + 0, /* st */ }, - .update = update_depth_stencil_alpha + update_depth_stencil_alpha /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_fixedfunction.c b/src/mesa/state_tracker/st_atom_fixedfunction.c index 3f137e1633..165567af70 100644 --- a/src/mesa/state_tracker/st_atom_fixedfunction.c +++ b/src/mesa/state_tracker/st_atom_fixedfunction.c @@ -55,12 +55,12 @@ static void update_tnl( struct st_context *st ) const struct st_tracked_state st_update_tnl = { - .name = "st_update_tnl", - .dirty = { - .mesa = TNL_FIXED_FUNCTION_STATE_FLAGS, - .st = 0 + "st_update_tnl", /* name */ + { /* dirty */ + TNL_FIXED_FUNCTION_STATE_FLAGS, /* mesa */ + 0 /* st */ }, - .update = update_tnl + update_tnl /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 14eeb58cc1..0a6974d8a7 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -96,11 +96,11 @@ update_framebuffer_state( struct st_context *st ) const struct st_tracked_state st_update_framebuffer = { - .name = "st_update_framebuffer", - .dirty = { - .mesa = _NEW_BUFFERS, - .st = 0, + "st_update_framebuffer", /* name */ + { /* dirty */ + _NEW_BUFFERS, /* mesa */ + 0, /* st */ }, - .update = update_framebuffer_state + update_framebuffer_state /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index 76356bbad7..999c148449 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -463,10 +463,10 @@ update_pixel_transfer(struct st_context *st) const struct st_tracked_state st_update_pixel_transfer = { - .name = "st_update_pixel_transfer", - .dirty = { - .mesa = _NEW_PIXEL | _NEW_COLOR_MATRIX, - .st = 0, + "st_update_pixel_transfer", /* name */ + { /* dirty */ + _NEW_PIXEL | _NEW_COLOR_MATRIX, /* mesa */ + 0, /* st */ }, - .update = update_pixel_transfer + update_pixel_transfer /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index bb14cf9045..87a91d56d0 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -267,11 +267,11 @@ static void update_raster_state( struct st_context *st ) } const struct st_tracked_state st_update_rasterizer = { - .name = "st_update_rasterizer", - .dirty = { - .mesa = (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | _NEW_SCISSOR | - _NEW_POINT | _NEW_BUFFERS | _NEW_MULTISAMPLE), - .st = 0, + "st_update_rasterizer", /* name */ + { /* dirty */ + (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | _NEW_SCISSOR | /* mesa */ + _NEW_POINT | _NEW_BUFFERS | _NEW_MULTISAMPLE), + 0, /* st */ }, - .update = update_raster_state + update_raster_state /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 0237da3693..7515bb30cc 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -185,15 +185,10 @@ update_samplers(struct st_context *st) const struct st_tracked_state st_update_sampler = { - .name = "st_update_sampler", - .dirty = { - .mesa = _NEW_TEXTURE, - .st = 0, + "st_update_sampler", /* name */ + { /* dirty */ + _NEW_TEXTURE, /* mesa */ + 0, /* st */ }, - .update = update_samplers + update_samplers /* update */ }; - - - - - diff --git a/src/mesa/state_tracker/st_atom_scissor.c b/src/mesa/state_tracker/st_atom_scissor.c index 59601e91a1..f5db492403 100644 --- a/src/mesa/state_tracker/st_atom_scissor.c +++ b/src/mesa/state_tracker/st_atom_scissor.c @@ -83,15 +83,10 @@ update_scissor( struct st_context *st ) const struct st_tracked_state st_update_scissor = { - .name = "st_update_scissor", - .dirty = { - .mesa = (_NEW_SCISSOR | _NEW_BUFFERS), - .st = 0, + "st_update_scissor", /* name */ + { /* dirty */ + (_NEW_SCISSOR | _NEW_BUFFERS), /* mesa */ + 0, /* st */ }, - .update = update_scissor + update_scissor /* update */ }; - - - - - diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 3f5ec71112..652500f52a 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -281,10 +281,10 @@ update_linkage( struct st_context *st ) const struct st_tracked_state st_update_shader = { - .name = "st_update_shader", - .dirty = { - .mesa = 0, - .st = ST_NEW_VERTEX_PROGRAM | ST_NEW_FRAGMENT_PROGRAM + "st_update_shader", /* name */ + { /* dirty */ + 0, /* mesa */ + ST_NEW_VERTEX_PROGRAM | ST_NEW_FRAGMENT_PROGRAM /* st */ }, - .update = update_linkage + update_linkage /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_stipple.c b/src/mesa/state_tracker/st_atom_stipple.c index c91214059a..f395930ab4 100644 --- a/src/mesa/state_tracker/st_atom_stipple.c +++ b/src/mesa/state_tracker/st_atom_stipple.c @@ -54,10 +54,10 @@ update_stipple( struct st_context *st ) const struct st_tracked_state st_update_polygon_stipple = { - .name = "st_update_polygon_stipple", - .dirty = { - .mesa = (_NEW_POLYGONSTIPPLE), - .st = 0, + "st_update_polygon_stipple", /* name */ + { /* dirty */ + (_NEW_POLYGONSTIPPLE), /* mesa */ + 0, /* st */ }, - .update = update_stipple + update_stipple /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index d15da5895a..f42b2f8d66 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -111,15 +111,10 @@ update_textures(struct st_context *st) const struct st_tracked_state st_update_texture = { - .name = "st_update_texture", - .dirty = { - .mesa = _NEW_TEXTURE, - .st = ST_NEW_FRAGMENT_PROGRAM, + "st_update_texture", /* name */ + { /* dirty */ + _NEW_TEXTURE, /* mesa */ + ST_NEW_FRAGMENT_PROGRAM, /* st */ }, - .update = update_textures + update_textures /* update */ }; - - - - - diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index eb3f62cfbe..4b51521470 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -82,10 +82,10 @@ update_viewport( struct st_context *st ) const struct st_tracked_state st_update_viewport = { - .name = "st_update_viewport", - .dirty = { - .mesa = _NEW_BUFFERS | _NEW_VIEWPORT, - .st = 0, + "st_update_viewport", /* name */ + { /* dirty */ + _NEW_BUFFERS | _NEW_VIEWPORT, /* mesa */ + 0, /* st */ }, - .update = update_viewport + update_viewport /* update */ }; diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index fe979f10bd..b7d7204633 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -34,8 +34,8 @@ #include "main/glheader.h" #include "main/macros.h" #include "shader/prog_instruction.h" -#include "st_atom.h" #include "st_context.h" +#include "st_atom.h" #include "st_cb_accum.h" #include "st_cb_clear.h" #include "st_cb_fbo.h" diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 0fe4d198bd..a3bffbfc95 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -35,8 +35,8 @@ #include "vbo/vbo.h" -#include "st_atom.h" #include "st_context.h" +#include "st_atom.h" #include "st_cb_bufferobjects.h" #include "st_draw.h" #include "st_program.h" @@ -549,9 +549,10 @@ st_feedback_draw_vbo(GLcontext *ctx, unsigned indexSize; struct gl_buffer_object *bufobj = ib->obj; struct st_buffer_object *stobj = st_buffer_object(bufobj); - index_buffer_handle = stobj->buffer; void *map; + index_buffer_handle = stobj->buffer; + switch (ib->type) { case GL_UNSIGNED_INT: indexSize = 4; diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 260a2efe88..6f94ba39ae 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -38,17 +38,17 @@ #include "st_extensions.h" -static int min(int a, int b) +static int _min(int a, int b) { return (a < b) ? a : b; } -static int max(int a, int b) +static int _max(int a, int b) { return (a > b) ? a : b; } -static int clamp(int a, int min, int max) +static int _clamp(int a, int min, int max) { if (a < min) return min; @@ -69,42 +69,42 @@ void st_init_limits(struct st_context *st) struct gl_constants *c = &st->ctx->Const; c->MaxTextureLevels - = min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS), + = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS), MAX_TEXTURE_LEVELS); c->Max3DTextureLevels - = min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_3D_LEVELS), + = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_3D_LEVELS), MAX_3D_TEXTURE_LEVELS); c->MaxCubeTextureLevels - = min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS), + = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS), MAX_CUBE_TEXTURE_LEVELS); c->MaxTextureRectSize - = min(1 << (c->MaxTextureLevels - 1), MAX_TEXTURE_RECT_SIZE); + = _min(1 << (c->MaxTextureLevels - 1), MAX_TEXTURE_RECT_SIZE); c->MaxTextureUnits = c->MaxTextureImageUnits = c->MaxTextureCoordUnits - = min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS), + = _min(screen->get_param(screen, PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS), MAX_TEXTURE_IMAGE_UNITS); c->MaxDrawBuffers - = clamp(screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS), + = _clamp(screen->get_param(screen, PIPE_CAP_MAX_RENDER_TARGETS), 1, MAX_DRAW_BUFFERS); c->MaxLineWidth - = max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH)); + = _max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH)); c->MaxLineWidthAA - = max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH_AA)); + = _max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH_AA)); c->MaxPointSize - = max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH)); + = _max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH)); c->MaxPointSizeAA - = max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH_AA)); + = _max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH_AA)); c->MaxTextureMaxAnisotropy - = max(2.0, screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_ANISOTROPY)); + = _max(2.0, screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_ANISOTROPY)); c->MaxTextureLodBias = screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_LOD_BIAS); -- cgit v1.2.3 From 7e4bc84dfc6c05cea3d6fc52173708ddeb33cf45 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Fri, 2 May 2008 10:31:46 +0000 Subject: inline -> INLINE --- src/mesa/state_tracker/st_program.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 1a2062131d..d8f26da2ee 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -101,14 +101,14 @@ struct st_vertex_program }; -static inline struct st_fragment_program * +static INLINE struct st_fragment_program * st_fragment_program( struct gl_fragment_program *fp ) { return (struct st_fragment_program *)fp; } -static inline struct st_vertex_program * +static INLINE struct st_vertex_program * st_vertex_program( struct gl_vertex_program *vp ) { return (struct st_vertex_program *)vp; -- cgit v1.2.3 From 5e49037caa4cf9062efd0bbebf67b467684b633b Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Fri, 2 May 2008 10:34:58 +0000 Subject: revert mode change back to 644 --- src/mesa/state_tracker/st_atom_blend.c | 0 src/mesa/state_tracker/st_atom_clip.c | 0 src/mesa/state_tracker/st_atom_constbuf.c | 0 3 files changed, 0 insertions(+), 0 deletions(-) mode change 100755 => 100644 src/mesa/state_tracker/st_atom_blend.c mode change 100755 => 100644 src/mesa/state_tracker/st_atom_clip.c mode change 100755 => 100644 src/mesa/state_tracker/st_atom_constbuf.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c old mode 100755 new mode 100644 diff --git a/src/mesa/state_tracker/st_atom_clip.c b/src/mesa/state_tracker/st_atom_clip.c old mode 100755 new mode 100644 diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c old mode 100755 new mode 100644 -- cgit v1.2.3 From 25d60838b5dfdbde54f19f26b41977fc25011474 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 2 May 2008 12:17:11 +0100 Subject: gallium: fix build after merge --- src/gallium/auxiliary/util/u_blit.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index b70bcbfa66..999a3e5099 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -296,7 +296,9 @@ util_blit_pixels(struct blit_state *ctx, src, srcLeft, srcTop, /* src */ srcW, srcH); /* size */ - pipe->texture_update(pipe, tex, 0, 1 << 0); + /* free the surface, update the texture if necessary. + */ + screen->tex_surface_release(screen, &texSurf); /* save state (restored below) */ cso_save_blend(ctx->cso); @@ -357,8 +359,6 @@ util_blit_pixels(struct blit_state *ctx, cso_restore_vertex_shader(ctx->cso); cso_restore_viewport(ctx->cso); - /* free the texture */ - pipe_surface_reference(&texSurf, NULL); screen->texture_release(screen, &tex); } -- cgit v1.2.3 From 4a159132082429d5492f5298c2ccb0df551c9f65 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 2 May 2008 14:27:10 +0100 Subject: gallium: remove usage of winsys->surface_alloc_storage from state tracker Allocate a texture containing storage instead. Also clean up ACCUM buffer allocation slightly -- drivers will need some changes to texture allocation logic to accomodate the concept of a texture that will only as image storage by the CPU, but it's cleaner than it was. --- src/gallium/include/pipe/p_state.h | 2 + src/mesa/state_tracker/st_cb_fbo.c | 138 +++++++++++++++++-------------------- 2 files changed, 64 insertions(+), 76 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 47e57e2957..277ee4b319 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -299,6 +299,8 @@ struct pipe_texture unsigned cpp:8; unsigned last_level:8; /**< Index of last mipmap level present/defined */ unsigned compressed:1; + + unsigned usage; /* These are also refcounted: */ diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 7fdc0bddd6..b174714171 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -90,90 +90,79 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); - enum pipe_format pipeFormat; - unsigned flags = (PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE | - PIPE_BUFFER_USAGE_GPU_READ); - int ret; + struct pipe_texture template, *texture; + + /* Free the old surface (and texture if we hold the last + * reference): + */ pipe_surface_reference( &strb->surface, NULL ); - if (!strb->surface) { - /* first time surface creation */ - strb->surface = pipe->winsys->surface_alloc(pipe->winsys); - assert(strb->surface); - assert(strb->surface->refcount); - assert(strb->surface->winsys); - if (!strb->surface) - return GL_FALSE; - } -#if 0 - else if (strb->surface->buffer) { - /* release/discard the old surface buffer */ - pipe_reference_buffer(pipe, &strb->surface->buffer, NULL); - } -#else - else { - assert(0); - } -#endif - /* Determine surface format here */ + memset(&template, 0, sizeof(template)); + if (strb->format != PIPE_FORMAT_NONE) { - assert(strb->format != 0); - /* we'll hit this for front/back color bufs */ - pipeFormat = strb->format; + template.format = strb->format; } else { - pipeFormat = st_choose_renderbuffer_format(pipe, internalFormat); + template.format = st_choose_renderbuffer_format(pipe, internalFormat); } - init_renderbuffer_bits(strb, pipeFormat); - - ret = pipe->winsys->surface_alloc_storage(pipe->winsys, - strb->surface, - width, - height, - pipeFormat, - flags); - if (ret || !strb->surface->buffer) { - if (pipeFormat == DEFAULT_ACCUM_PIPE_FORMAT) { - /* Accum buffer. Try a different surface format. Since accum - * buffers are s/w only for now, the surface pixel format doesn't - * really matter, only that the buffer is large enough. - */ - int sz, mult; - enum pipe_format accum_format; - - /* allocate a buffer of (typically) double height to get 64bpp */ - accum_format = st_choose_renderbuffer_format(pipe, GL_RGBA); - sz = pf_get_size(accum_format); - mult = pf_get_size(DEFAULT_ACCUM_PIPE_FORMAT) / sz; - - ret = pipe->winsys->surface_alloc_storage(pipe->winsys, - strb->surface, - width, height * mult, - accum_format, flags); - if (ret) - return GL_FALSE; /* we've _really_ failed */ - - } - else { - return GL_FALSE; /* out of memory, try s/w buffer? */ - } + strb->Base.Width = width; + strb->Base.Height = height; + init_renderbuffer_bits(strb, template.format); + + template.compressed = 0; + template.cpp = pf_get_size(template.format); + template.width[0] = width; + template.height[0] = height; + template.depth[0] = 1; + template.last_level = 0; + template.usage = (PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE | + PIPE_BUFFER_USAGE_GPU_READ); + + texture = pipe->screen->texture_create( pipe->screen, + &template ); + + /* Special path for accum buffers. + * + * Try a different surface format. Since accum buffers are s/w + * only for now, the surface pixel format doesn't really matter, + * only that the buffer is large enough. + */ + if (!texture && template.format == DEFAULT_ACCUM_PIPE_FORMAT) + { + /* Actually, just setting this usage value should be sufficient + * to tell the driver to go ahead and allocate the buffer, even + * if HW doesn't support the format. + */ + template.usage = (PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); + + texture = pipe->screen->texture_create( pipe->screen, + &template ); } - ASSERT(strb->surface->buffer); - ASSERT(strb->surface->format); - ASSERT(strb->surface->cpp); - ASSERT(strb->surface->width == width); - /*ASSERT(strb->surface->height == height);*/ - ASSERT(strb->surface->pitch); + if (!texture) + return FALSE; - strb->Base.Width = width; - strb->Base.Height = height; + strb->surface = pipe->screen->get_tex_surface( pipe->screen, + texture, + 0, 0, 0, + template.usage ); + + pipe_texture_reference( &texture, NULL ); - return GL_TRUE; + assert(strb->surface->buffer); + assert(strb->surface->format); + assert(strb->surface->cpp); + assert(strb->surface->width == width); + assert(strb->surface->height == height); + assert(strb->surface->pitch); + + + return strb->surface != NULL; } @@ -185,10 +174,7 @@ st_renderbuffer_delete(struct gl_renderbuffer *rb) { struct st_renderbuffer *strb = st_renderbuffer(rb); ASSERT(strb); - if (strb->surface) { - struct pipe_winsys *ws = strb->surface->winsys; - ws->surface_release(ws, &strb->surface); - } + pipe_surface_reference(&strb->surface, NULL); free(strb); } -- cgit v1.2.3 From 17058e07469f2dc5b47b4f820bd5a31b7ed9177c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 2 May 2008 16:02:18 +0200 Subject: tgsi: Implement fast rsqrtf. Not tested, inactive. --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 6 ++++ src/gallium/auxiliary/tgsi/exec/tgsi_exec.h | 10 +++++-- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 43 +++++++++++++++++------------ 3 files changed, 40 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 5d5125f7cb..826b432f09 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -88,6 +88,10 @@ #define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C #define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I #define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C +#define TEMP_3_I TGSI_EXEC_TEMP_THREE_I +#define TEMP_3_C TGSI_EXEC_TEMP_THREE_C +#define TEMP_HALF_I TGSI_EXEC_TEMP_HALF_I +#define TEMP_HALF_C TGSI_EXEC_TEMP_HALF_C #define TEMP_R0 TGSI_EXEC_TEMP_R0 #define FOR_EACH_CHANNEL(CHAN)\ @@ -262,6 +266,8 @@ tgsi_exec_machine_init( mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; + mach->Temps[TEMP_3_I].xyzw[TEMP_3_C].f[i] = 3.0f; + mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C].f[i] = 0.5f; } } diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h index 92e2e5e985..19bd78df3d 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h @@ -133,9 +133,15 @@ struct tgsi_exec_labels #define TGSI_EXEC_TEMP_PRIMITIVE_I 34 #define TGSI_EXEC_TEMP_PRIMITIVE_C 2 -#define TGSI_EXEC_TEMP_R0 35 +#define TGSI_EXEC_TEMP_THREE_I 34 +#define TGSI_EXEC_TEMP_THREE_C 3 -#define TGSI_EXEC_NUM_TEMPS (32 + 4) +#define TGSI_EXEC_TEMP_HALF_I 35 +#define TGSI_EXEC_TEMP_HALF_C 0 + +#define TGSI_EXEC_TEMP_R0 36 + +#define TGSI_EXEC_NUM_TEMPS (32 + 5) #define TGSI_EXEC_NUM_ADDRS 1 #define TGSI_EXEC_NUM_IMMEDIATES 256 diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 2fd76a3072..dbf002130b 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -36,7 +36,11 @@ #ifdef PIPE_ARCH_X86 -#define HIGH_PRECISION 1 /* for 1/sqrt() */ +/* for 1/sqrt() + * + * This costs about 100fps (close to 10%) in gears: + */ +#define HIGH_PRECISION 1 #define FOR_EACH_CHANNEL( CHAN )\ @@ -794,20 +798,25 @@ emit_rsqrt( * * See: http://softwarecommunity.intel.com/articles/eng/1818.htm */ - /* This is some code that woudl do the above for a scalar 'a'. We - * obviously are interested in a vector version: - * - * movss xmm3, a; - * movss xmm1, half; - * movss xmm2, three; - * rsqrtss xmm0, xmm3; - * mulss xmm3, xmm0; - * mulss xmm1, xmm0; - * mulss xmm3, xmm0; - * subss xmm2, xmm3; - * mulss xmm1, xmm2; - * movss x, xmm1; - */ + { + struct x86_reg dst = make_xmm( xmm_dst ); + struct x86_reg src = make_xmm( xmm_src ); + struct x86_reg tmp0 = make_xmm( 2 ); + struct x86_reg tmp1 = make_xmm( 3 ); + + assert( xmm_dst != xmm_src ); + assert( xmm_dst != 2 && xmm_dst != 3 ); + assert( xmm_src != 2 && xmm_src != 3 ); + + sse_movaps( func, dst, get_temp( TGSI_EXEC_TEMP_HALF_I, TGSI_EXEC_TEMP_HALF_C ) ); + sse_movaps( func, tmp0, get_temp( TGSI_EXEC_TEMP_THREE_I, TGSI_EXEC_TEMP_THREE_C ) ); + sse_rsqrtps( func, tmp1, src ); + sse_mulps( func, src, tmp1 ); + sse_mulps( func, dst, tmp1 ); + sse_mulps( func, src, tmp1 ); + sse_subps( func, tmp0, src ); + sse_mulps( func, dst, tmp0 ); + } #endif #else /* On Intel CPUs at least, this is only accurate to 12 bits -- not @@ -1295,9 +1304,9 @@ emit_instruction( case TGSI_OPCODE_RSQ: /* TGSI_OPCODE_RECIPSQRT */ FETCH( func, *inst, 0, 0, CHAN_X ); - emit_rsqrt( func, 0, 0 ); + emit_rsqrt( func, 1, 0 ); FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); + STORE( func, *inst, 1, 0, chan_index ); } break; -- cgit v1.2.3 From 6c15a70b75b1625b69790f98f2f44e9ae4435f6a Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 2 May 2008 16:12:55 +0200 Subject: tgsi: Enable fast high precision rsqrt. --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 22 ---------------------- 1 file changed, 22 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index dbf002130b..b6b05944be 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -761,20 +761,6 @@ emit_rcp ( make_xmm( xmm_src ) ); } -#if HIGH_PRECISION -static void XSTDCALL -rsqrt4f( - float *store ) -{ - const unsigned X = 0; - - store[X + 0] = 1.0F / sqrtf( store[X + 0] ); - store[X + 1] = 1.0F / sqrtf( store[X + 1] ); - store[X + 2] = 1.0F / sqrtf( store[X + 2] ); - store[X + 3] = 1.0F / sqrtf( store[X + 3] ); -} -#endif - static void emit_rsqrt( struct x86_function *func, @@ -782,13 +768,6 @@ emit_rsqrt( unsigned xmm_src ) { #if HIGH_PRECISION -#if 1 - emit_func_call_dst_src( - func, - xmm_dst, - xmm_src, - rsqrt4f ); -#else /* Although rsqrtps() and rcpps() are low precision on some/all SSE * implementations, it is possible to improve its precision at * fairly low cost, using a newton/raphson step, as below: @@ -817,7 +796,6 @@ emit_rsqrt( sse_subps( func, tmp0, src ); sse_mulps( func, dst, tmp0 ); } -#endif #else /* On Intel CPUs at least, this is only accurate to 12 bits -- not * good enough. -- cgit v1.2.3 From be8a42b4fcad0e0c7803e63f9c4b488c2f618cef Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Fri, 2 May 2008 15:25:37 +0000 Subject: snprintf -> util_snprintf --- src/mesa/state_tracker/st_cb_strings.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_strings.c b/src/mesa/state_tracker/st_cb_strings.c index 247519ab3d..066124f8f3 100644 --- a/src/mesa/state_tracker/st_cb_strings.c +++ b/src/mesa/state_tracker/st_cb_strings.c @@ -53,21 +53,21 @@ st_get_string(GLcontext * ctx, GLenum name) const char *vendor = screen->get_vendor( screen ); const char *tungsten = "Tungsten Graphics, Inc."; - /* Tungsten developed the state_tracker module (and much of - * Mesa), but the driver itself may come from elsewhere. The - * additional string allows "and XyzCorp" to reflect this. + /* Tungsten Graphics, Inc. developed the state_tracker module + * (and much of Mesa), but the driver itself may come from elsewhere. + * The additional string allows "and XyzCorp" to reflect this. */ if (vendor && strcmp(vendor, tungsten) != 0) - snprintf(st->vendor, sizeof(st->vendor), + util_snprintf(st->vendor, sizeof(st->vendor), "%s and %s", tungsten, vendor); else - snprintf(st->vendor, sizeof(st->vendor), "%s", tungsten); + util_snprintf(st->vendor, sizeof(st->vendor), "%s", tungsten); return (GLubyte *) st->vendor; } case GL_RENDERER: - snprintf(st->renderer, sizeof(st->renderer), "Gallium %s, %s on %s", + util_snprintf(st->renderer, sizeof(st->renderer), "Gallium %s, %s on %s", ST_VERSION_STRING, screen->get_name( screen ), screen->winsys->get_name( screen->winsys )); -- cgit v1.2.3 From d4074c509b5d28be0a2ec51d40329e1aed7047ef Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 2 May 2008 16:22:20 +0100 Subject: gallium: Reallocate pipe_texture in st_TexImage if the texture object was used. Fixes problems with interleaved glTexImage and rendering calls. --- src/mesa/state_tracker/st_atom_texture.c | 2 ++ src/mesa/state_tracker/st_cb_texture.c | 12 +++++++----- src/mesa/state_tracker/st_texture.h | 4 +--- 3 files changed, 10 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index f42b2f8d66..767654f3d0 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -90,6 +90,8 @@ update_textures(struct st_context *st) } st->state.num_textures = su + 1; + + stObj->teximage_realloc = TRUE; } pt = st_get_stobj_texture(stObj); diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 5fd192a8b8..80122546fc 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -556,15 +556,17 @@ st_TexImage(GLcontext * ctx, * waiting on any outstanding fences. */ if (stObj->pt && - /*stObj->pt->first_level == level &&*/ - stObj->pt->last_level == level && - stObj->pt->target != PIPE_TEXTURE_CUBE && - !st_texture_match_image(stObj->pt, &stImage->base, - stImage->face, stImage->level)) { + (stObj->teximage_realloc || + (/*stObj->pt->first_level == level &&*/ + stObj->pt->last_level == level && + stObj->pt->target != PIPE_TEXTURE_CUBE && + !st_texture_match_image(stObj->pt, &stImage->base, + stImage->face, stImage->level)))) { DBG("release it\n"); pipe_texture_release(&stObj->pt); assert(!stObj->pt); + stObj->teximage_realloc = FALSE; } if (!stObj->pt) { diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 7abccb3a69..b05762113c 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -68,9 +68,7 @@ struct st_texture_object */ struct pipe_texture *pt; - GLboolean imageOverride; - GLint depthOverride; - GLuint pitchOverride; + GLboolean teximage_realloc; }; -- cgit v1.2.3 From 8a5bf456dcb84b315152b87e33c8736248ad0a46 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 1 May 2008 15:29:25 -0600 Subject: gallium: comments, clean-ups --- src/gallium/winsys/xlib/xm_winsys.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 71ba076bff..51b247ffc3 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -58,6 +58,7 @@ /** + * Subclass of pipe_buffer for Xlib winsys. * Low-level OS/window system memory buffer */ struct xm_buffer @@ -80,6 +81,10 @@ struct xm_buffer # define XSHM_ENABLED(b) 0 #endif + +/** + * Subclass of pipe_surface for Xlib winsys + */ struct xmesa_surface { struct pipe_surface surface; @@ -89,6 +94,9 @@ struct xmesa_surface }; +/** + * Subclass of pipe_winsys for Xlib winsys + */ struct xmesa_pipe_winsys { struct pipe_winsys base; @@ -100,11 +108,11 @@ struct xmesa_pipe_winsys static void alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, unsigned width, unsigned height); + /** Cast wrapper */ static INLINE struct xmesa_surface * xmesa_surface(struct pipe_surface *ps) { -// assert(0); return (struct xmesa_surface *) ps; } -- cgit v1.2.3 From e9513bd403c0ba0b004d2414fda6736b38cecfd5 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 2 May 2008 07:54:11 -0600 Subject: gallium: code movement (XSHM code) --- src/gallium/winsys/xlib/xm_winsys.c | 252 ++++++++++++++++++------------------ 1 file changed, 125 insertions(+), 127 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 51b247ffc3..641ec11652 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -75,12 +75,6 @@ struct xm_buffer #endif }; -#if defined(USE_XSHM) && !defined(XFree86Server) -# define XSHM_ENABLED(b) ((b)->shm) -#else -# define XSHM_ENABLED(b) 0 -#endif - /** * Subclass of pipe_surface for Xlib winsys @@ -105,9 +99,6 @@ struct xmesa_pipe_winsys }; -static void alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, - unsigned width, unsigned height); - /** Cast wrapper */ static INLINE struct xmesa_surface * @@ -117,11 +108,7 @@ xmesa_surface(struct pipe_surface *ps) } - -/** - * Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque - * buffer pointer... - */ +/** Cast wrapper */ static INLINE struct xm_buffer * xm_buffer( struct pipe_buffer *buf ) { @@ -129,6 +116,130 @@ xm_buffer( struct pipe_buffer *buf ) } +/** + * X Shared Memory Image extension code + */ +#if defined(USE_XSHM) && !defined(XFree86Server) + +#define XSHM_ENABLED(b) ((b)->shm) + +static volatile int mesaXErrorFlag = 0; + +/** + * Catches potential Xlib errors. + */ +static int +mesaHandleXError(XMesaDisplay *dpy, XErrorEvent *event) +{ + (void) dpy; + (void) event; + mesaXErrorFlag = 1; + return 0; +} + + +static GLboolean alloc_shm(struct xm_buffer *buf, unsigned size) +{ + XShmSegmentInfo *const shminfo = & buf->shminfo; + + shminfo->shmid = shmget(IPC_PRIVATE, size, IPC_CREAT|0777); + if (shminfo->shmid < 0) { + return GL_FALSE; + } + + shminfo->shmaddr = (char *) shmat(shminfo->shmid, 0, 0); + if (shminfo->shmaddr == (char *) -1) { + shmctl(shminfo->shmid, IPC_RMID, 0); + return GL_FALSE; + } + + shminfo->readOnly = False; + return GL_TRUE; +} + + +/** + * Allocate a shared memory XImage back buffer for the given XMesaBuffer. + */ +static void +alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, + unsigned width, unsigned height) +{ + /* + * We have to do a _lot_ of error checking here to be sure we can + * really use the XSHM extension. It seems different servers trigger + * errors at different points if the extension won't work. Therefore + * we have to be very careful... + */ +#if 0 + GC gc; +#endif + int (*old_handler)(XMesaDisplay *, XErrorEvent *); + + b->tempImage = XShmCreateImage(xmb->xm_visual->display, + xmb->xm_visual->visinfo->visual, + xmb->xm_visual->visinfo->depth, + ZPixmap, + NULL, + &b->shminfo, + width, height); + if (b->tempImage == NULL) { + b->shm = 0; + return; + } + + + mesaXErrorFlag = 0; + old_handler = XSetErrorHandler(mesaHandleXError); + /* This may trigger the X protocol error we're ready to catch: */ + XShmAttach(xmb->xm_visual->display, &b->shminfo); + XSync(xmb->xm_visual->display, False); + + if (mesaXErrorFlag) { + /* we are on a remote display, this error is normal, don't print it */ + XFlush(xmb->xm_visual->display); + mesaXErrorFlag = 0; + XDestroyImage(b->tempImage); + b->tempImage = NULL; + b->shm = 0; + (void) XSetErrorHandler(old_handler); + return; + } + + + /* Finally, try an XShmPutImage to be really sure the extension works */ +#if 0 + gc = XCreateGC(xmb->xm_visual->display, xmb->drawable, 0, NULL); + XShmPutImage(xmb->xm_visual->display, xmb->drawable, gc, + b->tempImage, 0, 0, 0, 0, 1, 1 /*one pixel*/, False); + XSync(xmb->xm_visual->display, False); + XFreeGC(xmb->xm_visual->display, gc); + (void) XSetErrorHandler(old_handler); + if (mesaXErrorFlag) { + XFlush(xmb->xm_visual->display); + mesaXErrorFlag = 0; + XDestroyImage(b->tempImage); + b->tempImage = NULL; + b->shm = 0; + return; + } +#endif +} + +#else + +#define XSHM_ENABLED(b) 0 + +static void +alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, + unsigned width, unsigned height) +{ + b->shm = 0; +} +#endif /* USE_XSHM */ + + + /* Most callbacks map direcly onto dri_bufmgr operations: */ @@ -305,119 +416,6 @@ xm_get_name(struct pipe_winsys *pws) } -#if defined(USE_XSHM) && !defined(XFree86Server) -static volatile int mesaXErrorFlag = 0; - -/** - * Catches potential Xlib errors. - */ -static int -mesaHandleXError(XMesaDisplay *dpy, XErrorEvent *event) -{ - (void) dpy; - (void) event; - mesaXErrorFlag = 1; - return 0; -} - - -static GLboolean alloc_shm(struct xm_buffer *buf, unsigned size) -{ - XShmSegmentInfo *const shminfo = & buf->shminfo; - - shminfo->shmid = shmget(IPC_PRIVATE, size, IPC_CREAT|0777); - if (shminfo->shmid < 0) { - return GL_FALSE; - } - - shminfo->shmaddr = (char *) shmat(shminfo->shmid, 0, 0); - if (shminfo->shmaddr == (char *) -1) { - shmctl(shminfo->shmid, IPC_RMID, 0); - return GL_FALSE; - } - - shminfo->readOnly = False; - return GL_TRUE; -} - - -/** - * Allocate a shared memory XImage back buffer for the given XMesaBuffer. - */ -static void -alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, - unsigned width, unsigned height) -{ - /* - * We have to do a _lot_ of error checking here to be sure we can - * really use the XSHM extension. It seems different servers trigger - * errors at different points if the extension won't work. Therefore - * we have to be very careful... - */ -#if 0 - GC gc; -#endif - int (*old_handler)(XMesaDisplay *, XErrorEvent *); - - b->tempImage = XShmCreateImage(xmb->xm_visual->display, - xmb->xm_visual->visinfo->visual, - xmb->xm_visual->visinfo->depth, - ZPixmap, - NULL, - &b->shminfo, - width, height); - if (b->tempImage == NULL) { - b->shm = 0; - return; - } - - - mesaXErrorFlag = 0; - old_handler = XSetErrorHandler(mesaHandleXError); - /* This may trigger the X protocol error we're ready to catch: */ - XShmAttach(xmb->xm_visual->display, &b->shminfo); - XSync(xmb->xm_visual->display, False); - - if (mesaXErrorFlag) { - /* we are on a remote display, this error is normal, don't print it */ - XFlush(xmb->xm_visual->display); - mesaXErrorFlag = 0; - XDestroyImage(b->tempImage); - b->tempImage = NULL; - b->shm = 0; - (void) XSetErrorHandler(old_handler); - return; - } - - - /* Finally, try an XShmPutImage to be really sure the extension works */ -#if 0 - gc = XCreateGC(xmb->xm_visual->display, xmb->drawable, 0, NULL); - XShmPutImage(xmb->xm_visual->display, xmb->drawable, gc, - b->tempImage, 0, 0, 0, 0, 1, 1 /*one pixel*/, False); - XSync(xmb->xm_visual->display, False); - XFreeGC(xmb->xm_visual->display, gc); - (void) XSetErrorHandler(old_handler); - if (mesaXErrorFlag) { - XFlush(xmb->xm_visual->display); - mesaXErrorFlag = 0; - XDestroyImage(b->tempImage); - b->tempImage = NULL; - b->shm = 0; - return; - } -#endif -} -#else -static void -alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb, - unsigned width, unsigned height) -{ - b->shm = 0; -} -#endif - - static struct pipe_buffer * xm_buffer_create(struct pipe_winsys *pws, unsigned alignment, -- cgit v1.2.3 From cff8d3bdcbf78b57b52a2f60c54e5a3cae286137 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 2 May 2008 08:22:25 -0600 Subject: gallium: remove ^M (CR) chars --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 46 ++++++++++++++--------------- 1 file changed, 23 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index b6b05944be..8018bd7fa4 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -36,11 +36,11 @@ #ifdef PIPE_ARCH_X86 -/* for 1/sqrt() - * - * This costs about 100fps (close to 10%) in gears: - */ -#define HIGH_PRECISION 1 +/* for 1/sqrt() + * + * This costs about 100fps (close to 10%) in gears: + */ +#define HIGH_PRECISION 1 #define FOR_EACH_CHANNEL( CHAN )\ @@ -777,24 +777,24 @@ emit_rsqrt( * * See: http://softwarecommunity.intel.com/articles/eng/1818.htm */ - { - struct x86_reg dst = make_xmm( xmm_dst ); - struct x86_reg src = make_xmm( xmm_src ); - struct x86_reg tmp0 = make_xmm( 2 ); - struct x86_reg tmp1 = make_xmm( 3 ); - - assert( xmm_dst != xmm_src ); - assert( xmm_dst != 2 && xmm_dst != 3 ); - assert( xmm_src != 2 && xmm_src != 3 ); - - sse_movaps( func, dst, get_temp( TGSI_EXEC_TEMP_HALF_I, TGSI_EXEC_TEMP_HALF_C ) ); - sse_movaps( func, tmp0, get_temp( TGSI_EXEC_TEMP_THREE_I, TGSI_EXEC_TEMP_THREE_C ) ); - sse_rsqrtps( func, tmp1, src ); - sse_mulps( func, src, tmp1 ); - sse_mulps( func, dst, tmp1 ); - sse_mulps( func, src, tmp1 ); - sse_subps( func, tmp0, src ); - sse_mulps( func, dst, tmp0 ); + { + struct x86_reg dst = make_xmm( xmm_dst ); + struct x86_reg src = make_xmm( xmm_src ); + struct x86_reg tmp0 = make_xmm( 2 ); + struct x86_reg tmp1 = make_xmm( 3 ); + + assert( xmm_dst != xmm_src ); + assert( xmm_dst != 2 && xmm_dst != 3 ); + assert( xmm_src != 2 && xmm_src != 3 ); + + sse_movaps( func, dst, get_temp( TGSI_EXEC_TEMP_HALF_I, TGSI_EXEC_TEMP_HALF_C ) ); + sse_movaps( func, tmp0, get_temp( TGSI_EXEC_TEMP_THREE_I, TGSI_EXEC_TEMP_THREE_C ) ); + sse_rsqrtps( func, tmp1, src ); + sse_mulps( func, src, tmp1 ); + sse_mulps( func, dst, tmp1 ); + sse_mulps( func, src, tmp1 ); + sse_subps( func, tmp0, src ); + sse_mulps( func, dst, tmp0 ); } #else /* On Intel CPUs at least, this is only accurate to 12 bits -- not -- cgit v1.2.3 From cc2af38f2afa0e6003c8338d51c4f5fbabde40e1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 2 May 2008 09:26:17 -0600 Subject: gallium: fix typos, comments, whitespace --- src/gallium/drivers/softpipe/sp_tile_cache.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index a88aad5d09..1117c0ad4c 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -131,7 +131,7 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc) uint pos; for (pos = 0; pos < NUM_ENTRIES; pos++) { - //assert(tc->entries[pos].x < 0); + /*assert(tc->entries[pos].x < 0);*/ } if (tc->surface) { pipe_surface_reference(&tc->surface, NULL); @@ -332,8 +332,8 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, for (x = 0; x < w; x += TILE_SIZE) { if (is_clear_flag_set(tc->clear_flags, x, y)) { pipe_put_tile_raw(pipe, ps, - x, y, TILE_SIZE, TILE_SIZE, - tc->tile.data.color32, 0/*STRIDE*/); + x, y, TILE_SIZE, TILE_SIZE, + tc->tile.data.color32, 0/*STRIDE*/); /* do this? */ clear_clear_flag(tc->clear_flags, x, y); @@ -367,8 +367,8 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, if (tile->x >= 0) { if (tc->depth_stencil) { pipe_put_tile_raw(pipe, ps, - tile->x, tile->y, TILE_SIZE, TILE_SIZE, - tile->data.depth32, 0/*STRIDE*/); + tile->x, tile->y, TILE_SIZE, TILE_SIZE, + tile->data.depth32, 0/*STRIDE*/); } else { pipe_put_tile_rgba(pipe, ps, @@ -385,7 +385,7 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, #endif } else if (tc->texture) { - /* caching a texture, mark all entries as embpy */ + /* caching a texture, mark all entries as empty */ for (pos = 0; pos < NUM_ENTRIES; pos++) { tc->entries[pos].x = -1; } -- cgit v1.2.3 From a73ae3d5eb8419feab5aea26573aa41b72f941eb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 2 May 2008 16:46:31 +0100 Subject: gallium: Add texture usage flags, special-case allocation of display targets For many envirionments it's necessary to allocate display targets in a window-system friendly manner. Add facilities so that a driver can tell if a texture is likely to be used to generate a display surface and if use special allocation paths if necessary. Hook up softpipe to call into the winsys->surface_alloc_storage() routine in this case, though we probably want to change that interface slightly also. --- src/gallium/drivers/softpipe/sp_texture.c | 101 ++++++++++++++++--------- src/gallium/drivers/softpipe/sp_texture.h | 2 +- src/gallium/include/pipe/p_state.h | 6 +- src/mesa/state_tracker/st_atom_pixeltransfer.c | 3 +- src/mesa/state_tracker/st_cb_bitmap.c | 6 +- src/mesa/state_tracker/st_cb_drawpixels.c | 6 +- src/mesa/state_tracker/st_cb_fbo.c | 23 ++++-- src/mesa/state_tracker/st_cb_texture.c | 10 ++- src/mesa/state_tracker/st_texture.c | 4 +- src/mesa/state_tracker/st_texture.h | 3 +- 10 files changed, 111 insertions(+), 53 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 2b31cd4f25..599ff2ac45 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -52,40 +52,87 @@ static unsigned minify( unsigned d ) } -static void -softpipe_texture_layout(struct softpipe_texture * spt) +/* Conventional allocation path for non-display textures: + */ +static boolean +softpipe_texture_layout(struct pipe_screen *screen, + struct softpipe_texture * spt) { + struct pipe_winsys *ws = screen->winsys; struct pipe_texture *pt = &spt->base; unsigned level; unsigned width = pt->width[0]; unsigned height = pt->height[0]; unsigned depth = pt->depth[0]; - spt->buffer_size = 0; + unsigned buffer_size = 0; for (level = 0; level <= pt->last_level; level++) { pt->width[level] = width; pt->height[level] = height; pt->depth[level] = depth; + spt->pitch[level] = width; - spt->level_offset[level] = spt->buffer_size; + spt->level_offset[level] = buffer_size; - spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * - ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * - width * pt->cpp; + buffer_size += (((pt->compressed) ? MAX2(1, height/4) : height) * + ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * + width * pt->cpp); width = minify(width); height = minify(height); depth = minify(depth); } + + spt->buffer = ws->buffer_create(ws, 32, + PIPE_BUFFER_USAGE_PIXEL, + buffer_size); + + return spt->buffer != NULL; } + +/* Hack it up to use the old winsys->surface_alloc_storage() + * method for now: + */ +static boolean +softpipe_displaytarget_layout(struct pipe_screen *screen, + struct softpipe_texture * spt) +{ + struct pipe_winsys *ws = screen->winsys; + struct pipe_surface surf; + unsigned flags = (PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + + memset(&surf, 0, sizeof(surf)); + + ws->surface_alloc_storage( ws, + &surf, + spt->base.width[0], + spt->base.height[0], + spt->base.format, + flags); + + /* Now extract the goodies: + */ + spt->buffer = surf.buffer; + spt->pitch[0] = surf.pitch; + + return spt->buffer != NULL; +} + + + + + static struct pipe_texture * softpipe_texture_create(struct pipe_screen *screen, const struct pipe_texture *templat) { - struct pipe_winsys *ws = screen->winsys; struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture); if (!spt) return NULL; @@ -94,19 +141,21 @@ softpipe_texture_create(struct pipe_screen *screen, spt->base.refcount = 1; spt->base.screen = screen; - softpipe_texture_layout(spt); - - spt->buffer = ws->buffer_create(ws, 32, - PIPE_BUFFER_USAGE_PIXEL, - spt->buffer_size); - if (!spt->buffer) { - FREE(spt); - return NULL; + if (spt->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { + if (!softpipe_displaytarget_layout(screen, spt)) + goto fail; } - + else { + if (!softpipe_texture_layout(screen, spt)) + goto fail; + } + assert(spt->base.refcount == 1); - return &spt->base; + + fail: + FREE(spt); + return NULL; } @@ -178,22 +227,6 @@ softpipe_get_tex_surface(struct pipe_screen *screen, assert(face == 0); assert(zslice == 0); } - - if (usage & (PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_GPU_WRITE)) { - /* XXX if writing to the texture, invalidate the texcache entries!!! - * - * Actually, no. Flushing dependent contexts is still done - * explicitly and separately. Hardware drivers won't insert - * FLUSH commands into a command stream at this point, - * neither should softpipe try to flush caches. - * - * Those contexts could be living in separate threads & doing - * all sorts of unrelated stuff... Context<->texture - * dependency tracking needs to happen elsewhere. - */ - /* assert(0); */ - } } return ps; } diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h index 2ba093320d..779a9d8fc9 100644 --- a/src/gallium/drivers/softpipe/sp_texture.h +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -42,11 +42,11 @@ struct softpipe_texture struct pipe_texture base; unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; + unsigned long pitch[PIPE_MAX_TEXTURE_LEVELS]; /* The data is held here: */ struct pipe_buffer *buffer; - unsigned long buffer_size; }; diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 277ee4b319..d7565dff96 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -284,6 +284,10 @@ struct pipe_surface }; +#define PIPE_TEXTURE_USAGE_RENDER_TARGET 0x1 +#define PIPE_TEXTURE_USAGE_DISPLAY_TARGET 0x2 /* ie a backbuffer */ +#define PIPE_TEXTURE_USAGE_SAMPLER 0x4 + /** * Texture object. */ @@ -300,7 +304,7 @@ struct pipe_texture unsigned last_level:8; /**< Index of last mipmap level present/defined */ unsigned compressed:1; - unsigned usage; + unsigned tex_usage; /* PIPE_TEXTURE_USAGE_* */ /* These are also refcounted: */ diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index 0c32d53c4a..e500ac8684 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -126,7 +126,8 @@ create_color_map_texture(GLcontext *ctx) /* create texture for color map/table */ pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, - texSize, texSize, 1, 0); + texSize, texSize, 1, 0, + PIPE_TEXTURE_USAGE_SAMPLER); return pt; } diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 873b765c2c..f816e59104 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -321,7 +321,8 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, * Create texture to hold bitmap pattern. */ pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, ctx->st->bitmap.tex_format, - 0, width, height, 1, 0); + 0, width, height, 1, 0, + PIPE_TEXTURE_USAGE_SAMPLER); if (!pt) { _mesa_unmap_bitmap_pbo(ctx, unpack); return NULL; @@ -539,7 +540,8 @@ reset_cache(struct st_context *st) cache->texture = st_texture_create(st, PIPE_TEXTURE_2D, st->bitmap.tex_format, 0, BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, - 1, 0); + 1, 0, + PIPE_TEXTURE_USAGE_SAMPLER); /* Map the texture surface. * Subsequent glBitmap calls will write into the texture image. diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 9ae53c95f8..8c775ad886 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -346,7 +346,8 @@ make_texture(struct st_context *st, return NULL; pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height, - 1, 0); + 1, 0, + PIPE_TEXTURE_USAGE_SAMPLER); if (!pt) { _mesa_unmap_drawpix_pbo(ctx, unpack); return NULL; @@ -994,7 +995,8 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, } pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, texFormat, 0, - width, height, 1, 0); + width, height, 1, 0, + PIPE_TEXTURE_USAGE_SAMPLER); if (!pt) return; diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index b174714171..21d61e2163 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -90,8 +90,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); - struct pipe_texture template, *texture; + unsigned surface_usage; /* Free the old surface (and texture if we hold the last * reference): @@ -117,10 +117,15 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, template.height[0] = height; template.depth[0] = 1; template.last_level = 0; - template.usage = (PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE | - PIPE_BUFFER_USAGE_GPU_READ); + template.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_RENDER_TARGET); + + /* Probably need dedicated flags for surface usage too: + */ + surface_usage = (PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE | + PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); texture = pipe->screen->texture_create( pipe->screen, &template ); @@ -137,11 +142,13 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, * to tell the driver to go ahead and allocate the buffer, even * if HW doesn't support the format. */ - template.usage = (PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE); + template.tex_usage = 0; + surface_usage = (PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE); texture = pipe->screen->texture_create( pipe->screen, &template ); + } if (!texture) @@ -150,7 +157,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, strb->surface = pipe->screen->get_tex_surface( pipe->screen, texture, 0, 0, 0, - template.usage ); + surface_usage ); pipe_texture_reference( &texture, NULL ); diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 4cca3364c1..06caa06e77 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -333,7 +333,9 @@ guess_and_alloc_texture(struct st_context *st, width, height, depth, - comp_byte); + comp_byte, + ( PIPE_TEXTURE_USAGE_RENDER_TARGET | + PIPE_TEXTURE_USAGE_SAMPLER )); DBG("%s - success\n", __FUNCTION__); } @@ -1501,7 +1503,11 @@ st_finalize_texture(GLcontext *ctx, firstImage->base.Width2, firstImage->base.Height2, firstImage->base.Depth2, - comp_byte); + comp_byte, + + ( PIPE_TEXTURE_USAGE_RENDER_TARGET | + PIPE_TEXTURE_USAGE_SAMPLER )); + if (!stObj->pt) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); return GL_FALSE; diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index d6268fc80c..2b3742d4e5 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -75,7 +75,8 @@ st_texture_create(struct st_context *st, GLuint width0, GLuint height0, GLuint depth0, - GLuint compress_byte) + GLuint compress_byte, + GLuint usage ) { struct pipe_texture pt, *newtex; struct pipe_screen *screen = st->pipe->screen; @@ -98,6 +99,7 @@ st_texture_create(struct st_context *st, pt.depth[0] = depth0; pt.compressed = compress_byte ? 1 : 0; pt.cpp = pt.compressed ? compress_byte : st_sizeof_format(format); + pt.tex_usage = usage; newtex = screen->texture_create(screen, &pt); diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index f6d5733e21..6a9f08ec6b 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -105,7 +105,8 @@ st_texture_create(struct st_context *st, GLuint width0, GLuint height0, GLuint depth0, - GLuint compress_byte); + GLuint compress_byte, + GLuint tex_usage ); /* Check if an image fits into an existing texture object. -- cgit v1.2.3 From 5cb29dae06a4d97dc40ac7573e7ae7211e329b3c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 2 May 2008 16:56:06 +0100 Subject: i915: update to new display target allocation --- src/gallium/drivers/i915simple/i915_texture.c | 91 ++++++++++++++++++++++----- 1 file changed, 75 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 7b9359a0fe..f668e2e7d7 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -105,6 +105,50 @@ i915_miptree_set_image_offset(struct i915_texture *tex, } +/* Hack it up to use the old winsys->surface_alloc_storage() + * method for now: + */ +static boolean +i915_displaytarget_layout(struct pipe_screen *screen, + struct i915_texture *tex) +{ + struct pipe_winsys *ws = screen->winsys; + struct pipe_surface surf; + unsigned flags = (PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); + + + memset(&surf, 0, sizeof(surf)); + + ws->surface_alloc_storage( ws, + &surf, + tex->base.width[0], + tex->base.height[0], + tex->base.format, + flags); + + /* Now extract the goodies: + */ + i915_miptree_set_image_offset( tex, 0, 0, 0, 0 ); + i915_miptree_set_level_info( tex, 0, 0, 0, 0, + tex->base.width[0], + tex->base.height[0], + 1 ); + + tex->buffer = surf.buffer; + tex->pitch = surf.pitch; + tex->total_height = 0; + + + return tex->buffer != NULL; +} + + + + + static void i945_miptree_layout_2d( struct i915_texture *tex ) { @@ -483,30 +527,45 @@ static struct pipe_texture * i915_texture_create_screen(struct pipe_screen *screen, const struct pipe_texture *templat) { + struct i915_screen *i915screen = i915_screen(screen); + struct pipe_winsys *ws = screen->winsys; struct i915_texture *tex = CALLOC_STRUCT(i915_texture); - if (tex) { - struct i915_screen *i915screen = i915_screen(screen); - struct pipe_winsys *ws = screen->winsys; - - tex->base = *templat; - tex->base.refcount = 1; - tex->base.screen = screen; + if (!tex) + return NULL; - if (i915screen->is_i945 ? i945_miptree_layout(tex) : - i915_miptree_layout(tex)) - tex->buffer = ws->buffer_create(ws, 64, - PIPE_BUFFER_USAGE_PIXEL, - tex->pitch * tex->base.cpp * - tex->total_height); + tex->base = *templat; + tex->base.refcount = 1; + tex->base.screen = screen; - if (!tex->buffer) { - FREE(tex); - return NULL; + if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { + if (!i915_displaytarget_layout(screen, tex)) + goto fail; + } + else { + if (i915screen->is_i945) { + if (!i945_miptree_layout(tex)) + goto fail; } + else { + if (!i915_miptree_layout(tex)) + goto fail; + } + + tex->buffer = ws->buffer_create(ws, 64, + PIPE_BUFFER_USAGE_PIXEL, + tex->pitch * tex->base.cpp * + tex->total_height); + + if (!tex->buffer) + goto fail; } return &tex->base; + + fail: + FREE(tex); + return NULL; } -- cgit v1.2.3 From 3d53d38d5e35386de4793162b9dd32e171927059 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 2 May 2008 10:37:20 -0600 Subject: gallium: new debug code (disabled) --- src/gallium/drivers/softpipe/sp_tex_sample.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 5b63f97997..be0b57d9fa 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -1051,5 +1051,19 @@ sp_get_samples(struct tgsi_sampler *sampler, default: assert(0); } + +#if 0 /* DEBUG */ + { + int i; + printf("Sampled at %f, %f, %f:\n", s[0], t[0], p[0]); + for (i = 0; i < 4; i++) { + printf("Frag %d: %f %f %f %f\n", i, + rgba[0][i], + rgba[1][i], + rgba[2][i], + rgba[3][i]); + } + } +#endif } -- cgit v1.2.3 From 479d19f5a2bca79104f3b7f94147e94a52a27fea Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 2 May 2008 10:38:39 -0600 Subject: gallium: in st_finalize_texture() check texture dimensions Check dimensions in addition to target, format, etc. Fixes a bug where we failed to detect a change in texture image sizes and wound up using the old texture data. --- src/mesa/state_tracker/st_cb_texture.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 80122546fc..1eabef6d92 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1483,13 +1483,16 @@ st_finalize_texture(GLcontext *ctx, } /* Check texture can hold all active levels. Check texture matches - * target, imageFormat, etc. + * target, imageFormat, dimensions, etc. */ if (stObj->pt && (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || stObj->pt->format != st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat) || stObj->pt->last_level < stObj->lastLevel || + stObj->pt->width[0] != firstImage->base.Width2 || + stObj->pt->height[0] != firstImage->base.Height2 || + stObj->pt->depth[0] != firstImage->base.Depth2 || stObj->pt->cpp != cpp || stObj->pt->compressed != firstImage->base.IsCompressed)) { pipe_texture_release(&stObj->pt); -- cgit v1.2.3 From ce509401738e7073de94a2b7fc41461c52a73da0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 2 May 2008 10:43:29 -0600 Subject: gallium: minor code, comments clean-up --- src/mesa/state_tracker/st_cb_texture.c | 42 +++++++++++++--------------------- 1 file changed, 16 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 1eabef6d92..02ef961e4d 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1448,17 +1448,6 @@ st_finalize_texture(GLcontext *ctx, calculate_first_last_level(stObj); firstImage = st_texture_image(stObj->base.Image[0][stObj->base.BaseLevel]); -#if 0 - /* Fallback case: - */ - if (firstImage->base.Border) { - if (stObj->pt) { - pipe_texture_release(&stObj->pt); - } - return GL_FALSE; - } -#endif - /* If both firstImage and stObj point to a texture which can contain * all active images, favour firstImage. Note that because of the * completeness requirement, we know that the image dimensions @@ -1482,24 +1471,25 @@ st_finalize_texture(GLcontext *ctx, cpp = firstImage->base.TexFormat->TexelBytes; } - /* Check texture can hold all active levels. Check texture matches - * target, imageFormat, dimensions, etc. + /* If we already have a gallium texture, check that it matches the texture + * object's format, target, size, num_levels, etc. */ - if (stObj->pt && - (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || - stObj->pt->format != - st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat) || - stObj->pt->last_level < stObj->lastLevel || - stObj->pt->width[0] != firstImage->base.Width2 || - stObj->pt->height[0] != firstImage->base.Height2 || - stObj->pt->depth[0] != firstImage->base.Depth2 || - stObj->pt->cpp != cpp || - stObj->pt->compressed != firstImage->base.IsCompressed)) { - pipe_texture_release(&stObj->pt); + if (stObj->pt) { + const enum pipe_format fmt = + st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat); + if (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || + stObj->pt->format != fmt || + stObj->pt->last_level < stObj->lastLevel || + stObj->pt->width[0] != firstImage->base.Width2 || + stObj->pt->height[0] != firstImage->base.Height2 || + stObj->pt->depth[0] != firstImage->base.Depth2 || + stObj->pt->cpp != cpp || + stObj->pt->compressed != firstImage->base.IsCompressed) { + pipe_texture_release(&stObj->pt); + } } - - /* May need to create a new texture: + /* May need to create a new gallium texture: */ if (!stObj->pt) { stObj->pt = st_texture_create(ctx->st, -- cgit v1.2.3 From 7849ccb2a7dba3b9d751acaac9dd9aec3abe3b59 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 2 May 2008 17:55:22 +0100 Subject: brw: remove dead code --- src/gallium/drivers/i965simple/brw_tex_layout.c | 8 -------- 1 file changed, 8 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index ba4c4a7bcf..78ae0b1223 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -357,14 +357,6 @@ brw_texture_release_screen(struct pipe_screen *screen, } -static void -brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture, - uint face, uint levelsMask) -{ - /* no-op? */ -} - - static struct pipe_surface * brw_get_tex_surface_screen(struct pipe_screen *screen, struct pipe_texture *pt, -- cgit v1.2.3 From b2021e7c06a9ec13b82eeeb352ad2408fe060518 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 2 May 2008 17:56:01 +0100 Subject: gallium: identify depth-stencil textures And don't use the display-target path to allocate them. --- src/gallium/include/pipe/p_state.h | 3 ++- src/mesa/state_tracker/st_cb_fbo.c | 16 ++++++++++++++-- 2 files changed, 16 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index d7565dff96..4936b7f507 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -286,7 +286,8 @@ struct pipe_surface #define PIPE_TEXTURE_USAGE_RENDER_TARGET 0x1 #define PIPE_TEXTURE_USAGE_DISPLAY_TARGET 0x2 /* ie a backbuffer */ -#define PIPE_TEXTURE_USAGE_SAMPLER 0x4 +#define PIPE_TEXTURE_USAGE_DEPTH_STENCIL 0x4 +#define PIPE_TEXTURE_USAGE_SAMPLER 0x8 /** * Texture object. diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 21d61e2163..cb5736f7c3 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -77,6 +77,11 @@ init_renderbuffer_bits(struct st_renderbuffer *strb, return info.size; } +static INLINE GLboolean pf_is_depth_stencil( enum pipe_format format ) +{ + return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_S )) != 0; +} /** * gl_renderbuffer::AllocStorage() @@ -117,8 +122,15 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, template.height[0] = height; template.depth[0] = 1; template.last_level = 0; - template.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET | - PIPE_TEXTURE_USAGE_RENDER_TARGET); + + if (pf_is_depth_stencil(template.format)) { + template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL; + } + else { + template.tex_usage = (PIPE_TEXTURE_USAGE_DISPLAY_TARGET | + PIPE_TEXTURE_USAGE_RENDER_TARGET); + } + /* Probably need dedicated flags for surface usage too: */ -- cgit v1.2.3 From 99df379b2c5b8e4e2ee7e5f2af864daf0a9eb1f7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 2 May 2008 11:12:15 -0600 Subject: gallium: set template.target = PIPE_TEXTURE_2D --- src/mesa/state_tracker/st_cb_fbo.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index cb5736f7c3..1fa82b07e6 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -116,6 +116,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, strb->Base.Height = height; init_renderbuffer_bits(strb, template.format); + template.target = PIPE_TEXTURE_2D; template.compressed = 0; template.cpp = pf_get_size(template.format); template.width[0] = width; -- cgit v1.2.3 From 07aaf3a7a730d65052ec34a82cce9b373673b556 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 2 May 2008 14:00:08 -0600 Subject: gallium: remove obsolete PIPE_CAP_BITMAP_TEXCOORD_BIAS --- src/gallium/include/pipe/p_defines.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index d5d341d689..46a355daa2 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -268,7 +268,7 @@ enum pipe_texture_target { #define PIPE_CAP_MAX_POINT_WIDTH_AA 17 #define PIPE_CAP_MAX_TEXTURE_ANISOTROPY 18 #define PIPE_CAP_MAX_TEXTURE_LOD_BIAS 19 -#define PIPE_CAP_BITMAP_TEXCOORD_BIAS 20 + #ifdef __cplusplus } -- cgit v1.2.3 From 6e004e973bcc5b789ee3f70b70f5d728c8b8ea71 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 2 May 2008 14:00:35 -0600 Subject: gallium: remove 0.5 vertex biases in set_vertex_data() These should not be needed and were causing garbage to appear along the edges of the mipmap images. --- src/gallium/auxiliary/util/u_gen_mipmap.c | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 2e8417ee13..c53c512268 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -778,23 +778,23 @@ set_vertex_data(struct gen_mipmap_state *ctx, float width, float height) { void *buf; - ctx->vertices[0][0][0] = -0.5f; /*x*/ - ctx->vertices[0][0][1] = -0.5f; /*y*/ + ctx->vertices[0][0][0] = 0.0f; /*x*/ + ctx->vertices[0][0][1] = 0.0f; /*y*/ ctx->vertices[0][1][0] = 0.0f; /*s*/ ctx->vertices[0][1][1] = 0.0f; /*t*/ - ctx->vertices[1][0][0] = width - 0.5f; /*x*/ - ctx->vertices[1][0][1] = -0.5f; /*y*/ - ctx->vertices[1][1][0] = 1.0f; /*s*/ - ctx->vertices[1][1][1] = 0.0f; /*t*/ + ctx->vertices[1][0][0] = width; + ctx->vertices[1][0][1] = 0.0f; + ctx->vertices[1][1][0] = 1.0f; + ctx->vertices[1][1][1] = 0.0f; - ctx->vertices[2][0][0] = width - 0.5f; - ctx->vertices[2][0][1] = height - 0.5f; + ctx->vertices[2][0][0] = width; + ctx->vertices[2][0][1] = height; ctx->vertices[2][1][0] = 1.0f; ctx->vertices[2][1][1] = 1.0f; - ctx->vertices[3][0][0] = -0.5f; - ctx->vertices[3][0][1] = height - 0.5f; + ctx->vertices[3][0][0] = 0.0f; + ctx->vertices[3][0][1] = height; ctx->vertices[3][1][0] = 0.0f; ctx->vertices[3][1][1] = 1.0f; -- cgit v1.2.3 From 8e6a3807981bb96bb6747683dbc9f92ae7117dd2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 3 May 2008 15:41:05 +0100 Subject: gallium: add pipe surface layout value (Roland Scheidegger) --- src/gallium/include/pipe/p_defines.h | 6 ++++++ src/gallium/include/pipe/p_state.h | 1 + 2 files changed, 7 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index d5d341d689..2e77fb42a8 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -173,6 +173,12 @@ enum pipe_texture_target { #define PIPE_SURFACE 2 /**< user-created surfaces */ +/** + * Surface layout + */ +#define PIPE_SURFACE_LAYOUT_LINEAR 0 + + /** * Surface status */ diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 4936b7f507..571ea8c6e1 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -271,6 +271,7 @@ struct pipe_surface unsigned width; unsigned height; unsigned pitch; /**< in pixels */ + unsigned layout; /**< PIPE_SURFACE_LAYOUT_x */ unsigned offset; /**< offset from start of buffer, in bytes */ unsigned refcount; unsigned usage; /**< PIPE_BUFFER_USAGE_* */ -- cgit v1.2.3 From 3668d20a9f382a174e002dc9b226710bdd983992 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 3 May 2008 08:58:44 -0600 Subject: gallium: fix computation of sampler->min_lod The texture BaseLevel is accounted for in texture layout so it doesn't factor in here. May also need to adjust max_lod... --- src/mesa/state_tracker/st_atom_sampler.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 7515bb30cc..b020d43505 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -148,7 +148,7 @@ update_samplers(struct st_context *st) sampler->normalized_coords = 1; sampler->lod_bias = st->ctx->Texture.Unit[su].LodBias; - sampler->min_lod = MAX2(texobj->BaseLevel, texobj->MinLod); + sampler->min_lod = MAX2(0, texobj->MinLod); sampler->max_lod = MIN2(texobj->MaxLevel, texobj->MaxLod); sampler->border_color[0] = texobj->BorderColor[RCOMP]; -- cgit v1.2.3 From 3837d401cc665eccf079eba5822b1a3eec565b81 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 3 May 2008 09:07:11 -0600 Subject: gallium: fix sampler->max_lod computation --- src/mesa/state_tracker/st_atom_sampler.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index b020d43505..379b41c654 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -148,8 +148,9 @@ update_samplers(struct st_context *st) sampler->normalized_coords = 1; sampler->lod_bias = st->ctx->Texture.Unit[su].LodBias; - sampler->min_lod = MAX2(0, texobj->MinLod); - sampler->max_lod = MIN2(texobj->MaxLevel, texobj->MaxLod); + sampler->min_lod = MAX2(0.0f, texobj->MinLod); + sampler->max_lod = MIN2(texobj->MaxLevel - texobj->BaseLevel, + texobj->MaxLod); sampler->border_color[0] = texobj->BorderColor[RCOMP]; sampler->border_color[1] = texobj->BorderColor[GCOMP]; -- cgit v1.2.3 From 736f535b4f1c5e6912b5b2fe9415a3b44a678844 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 3 May 2008 09:26:25 -0600 Subject: gallium: fix warnings --- src/gallium/winsys/xlib/xm_winsys.c | 2 +- src/mesa/state_tracker/st_atom_sampler.c | 4 +++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 641ec11652..14c3892559 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -606,7 +606,7 @@ xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis) } -struct pipe_winsys * +static struct pipe_winsys * xmesa_get_pipe_winsys(struct xmesa_visual *xm_vis) { static struct xmesa_pipe_winsys *ws = NULL; diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 379b41c654..10283d31a1 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -32,12 +32,14 @@ */ +#include "main/macros.h" + #include "st_context.h" #include "st_atom.h" #include "st_program.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" + #include "cso_cache/cso_context.h" -- cgit v1.2.3 From ffde4e03cf178719c06c300939f8f469d7fc5e9c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 3 May 2008 17:01:20 -0600 Subject: gallium: fix some BaseLevel, lastLevel bugs --- src/mesa/state_tracker/st_atom_sampler.c | 5 +++-- src/mesa/state_tracker/st_cb_texture.c | 5 ++++- 2 files changed, 7 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 7515bb30cc..0c22e03883 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -148,8 +148,9 @@ update_samplers(struct st_context *st) sampler->normalized_coords = 1; sampler->lod_bias = st->ctx->Texture.Unit[su].LodBias; - sampler->min_lod = MAX2(texobj->BaseLevel, texobj->MinLod); - sampler->max_lod = MIN2(texobj->MaxLevel, texobj->MaxLod); + sampler->min_lod = MAX2(0.0f, texobj->MinLod - texobj->BaseLevel); + sampler->max_lod = MIN2(texobj->MaxLevel - texobj->BaseLevel, + texobj->MaxLod); sampler->border_color[0] = texobj->BorderColor[RCOMP]; sampler->border_color[1] = texobj->BorderColor[GCOMP]; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 06caa06e77..21c0141ac3 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1363,7 +1363,7 @@ calculate_first_last_level(struct st_texture_object *stObj) } else { firstLevel = 0; - lastLevel = MIN2(tObj->MaxLevel, tObj->Image[0][0]->WidthLog2); + lastLevel = MIN2(tObj->MaxLevel, tObj->Image[0][tObj->BaseLevel]->WidthLog2); } break; case GL_TEXTURE_RECTANGLE_NV: @@ -1488,6 +1488,9 @@ st_finalize_texture(GLcontext *ctx, st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat) || stObj->pt->last_level < stObj->lastLevel || stObj->pt->cpp != cpp || + stObj->pt->width[0] != firstImage->base.Width2 || + stObj->pt->height[0] != firstImage->base.Height2 || + stObj->pt->depth[0] != firstImage->base.Depth2 || stObj->pt->compressed != firstImage->base.IsCompressed)) { pipe_texture_release(&stObj->pt); } -- cgit v1.2.3 From 131a1fbc91725c11e4822b82e58b94ec3a711476 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 5 May 2008 23:58:37 +0900 Subject: util: Alternative implementation for standard c library string functions. --- src/gallium/auxiliary/util/u_string.h | 112 ++++++++++++++++++++++++++++++++++ 1 file changed, 112 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_string.h b/src/gallium/auxiliary/util/u_string.h index b99d4e8021..2fede229d9 100644 --- a/src/gallium/auxiliary/util/u_string.h +++ b/src/gallium/auxiliary/util/u_string.h @@ -41,6 +41,8 @@ #include #include +#include "pipe/p_compiler.h" + #ifdef __cplusplus extern "C" { @@ -48,11 +50,121 @@ extern "C" { #ifdef WIN32 + int util_vsnprintf(char *, size_t, const char *, va_list); int util_snprintf(char *str, size_t size, const char *format, ...); + +static INLINE void +util_sprintf(char *str, const char *format, ...) +{ + va_list ap; + va_start(ap, format); + util_vsnprintf(str, (size_t)-1, format, ap); + va_end(ap); +} + +static INLINE char * +util_strchr(const char *s, char c) +{ + while(*s) { + if(*s == c) + return (char *)s; + ++s; + } + return NULL; +} + +static INLINE char* +util_strncat(char *dst, const char *src, size_t n) +{ + char *p = dst + strlen(dst); + const char *q = src; + size_t i; + + for (i = 0; i < n && *q != '\0'; ++i) + *p++ = *q++; + *p = '\0'; + + return dst; +} + +static INLINE int +util_strcmp(const char *s1, const char *s2) +{ + unsigned char u1, u2; + + while (1) { + u1 = (unsigned char) *s1++; + u2 = (unsigned char) *s2++; + if (u1 != u2) + return u1 - u2; + if (u1 == '\0') + return 0; + } + return 0; +} + +static INLINE int +util_strncmp(const char *s1, const char *s2, size_t n) +{ + unsigned char u1, u2; + + while (n-- > 0) { + u1 = (unsigned char) *s1++; + u2 = (unsigned char) *s2++; + if (u1 != u2) + return u1 - u2; + if (u1 == '\0') + return 0; + } + return 0; +} + +static INLINE char * +util_strstr(const char *haystack, const char *needle) +{ + const char *p = haystack; + int len = strlen(needle); + + for (; (p = util_strchr(p, *needle)) != 0; p++) { + if (util_strncmp(p, needle, len) == 0) { + return (char *)p; + } + } + return NULL; +} + +static INLINE void * +util_memmove(void *dest, const void *src, size_t n) +{ + char *p = (char *)dest; + const char *q = (const char *)src; + if (dest < src) { + while (n--) + *p++ = *q++; + } + else + { + p += n; + q += n; + while (n--) + *--p = *--q; + } + return dest; +} + + #else + #define util_vsnprintf vsnprintf #define util_snprintf snprintf +#define util_strchr strchr +#define util_strcmp strcmp +#define util_strncmp strncmp +#define util_strncat strncat +#define util_strstr strstr +#define util_memmove memmove + #endif -- cgit v1.2.3 From 736374c1052be647bd7c377344acf8db0af4ddfc Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Mon, 5 May 2008 23:08:59 +0100 Subject: declare atoms as extern --- src/mesa/state_tracker/st_atom.h | 28 ++++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index 3a63e2dec0..c6c6eba812 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -44,20 +44,20 @@ void st_destroy_atoms( struct st_context *st ); void st_validate_state( struct st_context *st ); -const struct st_tracked_state st_update_framebuffer; -const struct st_tracked_state st_update_clip; -const struct st_tracked_state st_update_depth_stencil_alpha; -const struct st_tracked_state st_update_shader; -const struct st_tracked_state st_update_rasterizer; -const struct st_tracked_state st_update_polygon_stipple; -const struct st_tracked_state st_update_viewport; -const struct st_tracked_state st_update_scissor; -const struct st_tracked_state st_update_blend; -const struct st_tracked_state st_update_sampler; -const struct st_tracked_state st_update_texture; -const struct st_tracked_state st_update_fs_constants; -const struct st_tracked_state st_update_vs_constants; -const struct st_tracked_state st_update_pixel_transfer; +extern const struct st_tracked_state st_update_framebuffer; +extern const struct st_tracked_state st_update_clip; +extern const struct st_tracked_state st_update_depth_stencil_alpha; +extern const struct st_tracked_state st_update_shader; +extern const struct st_tracked_state st_update_rasterizer; +extern const struct st_tracked_state st_update_polygon_stipple; +extern const struct st_tracked_state st_update_viewport; +extern const struct st_tracked_state st_update_scissor; +extern const struct st_tracked_state st_update_blend; +extern const struct st_tracked_state st_update_sampler; +extern const struct st_tracked_state st_update_texture; +extern const struct st_tracked_state st_update_fs_constants; +extern const struct st_tracked_state st_update_vs_constants; +extern const struct st_tracked_state st_update_pixel_transfer; uint st_compare_func_to_pipe(GLenum func); -- cgit v1.2.3 From f77442fbd3b539aa3da927630c12c3a1a377f6da Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Mon, 5 May 2008 23:09:38 +0100 Subject: fix _mesa_ffs for alternative compilers --- src/mesa/main/imports.c | 36 +++++++++++++++++------------------- 1 file changed, 17 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c index d8d35af15e..d798f80e25 100644 --- a/src/mesa/main/imports.c +++ b/src/mesa/main/imports.c @@ -542,26 +542,24 @@ int _mesa_ffs(int i) { #if (defined(_WIN32) && !defined(__MINGW32__) ) || defined(__IBMC__) || defined(__IBMCPP__) - register int bit = 0; - if (i != 0) { - if ((i & 0xffff) == 0) { - bit += 16; - i >>= 16; - } - if ((i & 0xff) == 0) { - bit += 8; - i >>= 8; - } - if ((i & 0xf) == 0) { - bit += 4; - i >>= 4; - } - while ((i & 1) == 0) { - bit++; - i >>= 1; - } + register int bit = 1; + if ((i & 0xffff) == 0) { + bit += 16; + i >>= 16; + } + if ((i & 0xff) == 0) { + bit += 8; + i >>= 8; + } + if ((i & 0xf) == 0) { + bit += 4; + i >>= 4; + } + if ((i & 0x3) == 0) { + bit += 2; + i >>= 2; } - return bit; + return (i) ? (bit + ((i + 1) & 0x01)) : 0; #else return ffs(i); #endif -- cgit v1.2.3 From 6361d6f48d13ce481253faf106fba5c6a41488ed Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 6 May 2008 14:57:18 +0900 Subject: gallium: New pipe_screen interface to overlay a texture on existing memory. --- src/gallium/include/pipe/p_screen.h | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h index c080579c26..cc8430dae1 100644 --- a/src/gallium/include/pipe/p_screen.h +++ b/src/gallium/include/pipe/p_screen.h @@ -89,6 +89,18 @@ struct pipe_screen { struct pipe_texture * (*texture_create)(struct pipe_screen *, const struct pipe_texture *templat); + /** + * Create a new texture object, using the given template info, but on top of + * existing memory. + * + * It is assumed that the buffer data is layed out according to the expected + * by the hardware. NULL will be returned if any inconsistency is found. + */ + struct pipe_texture * (*texture_blanket)(struct pipe_screen *, + const struct pipe_texture *templat, + const unsigned *pitch, + struct pipe_buffer *buffer); + void (*texture_release)(struct pipe_screen *, struct pipe_texture **pt); -- cgit v1.2.3 From 4c6c073f993e13da0b68f897a4221e6bb7875fbe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 6 May 2008 09:07:11 +0100 Subject: gallium: Define util_sprintf for non-WIN32. --- src/gallium/auxiliary/util/u_string.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_string.h b/src/gallium/auxiliary/util/u_string.h index 2fede229d9..d7d925ee9b 100644 --- a/src/gallium/auxiliary/util/u_string.h +++ b/src/gallium/auxiliary/util/u_string.h @@ -158,6 +158,7 @@ util_memmove(void *dest, const void *src, size_t n) #define util_vsnprintf vsnprintf #define util_snprintf snprintf +#define util_sprintf sprintf #define util_strchr strchr #define util_strcmp strcmp #define util_strncmp strncmp -- cgit v1.2.3 From 650c57f19398800dfdcf357b6e9ec7b68bfa34f1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 6 May 2008 12:23:16 +0100 Subject: gallium: glCopyTexSubImage improvement. Only get a texture surface for attempting an accelerated copy, and mark it for GPU use only. --- src/mesa/state_tracker/st_cb_texture.c | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index c877e4382d..2b587a0848 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1127,7 +1127,6 @@ do_copy_texsubimage(GLcontext *ctx, struct st_renderbuffer *strb; struct pipe_context *pipe = ctx->st->pipe; struct pipe_screen *screen = pipe->screen; - struct pipe_surface *dest_surface; uint dest_format, src_format; uint do_flip = FALSE; GLboolean use_fallback = GL_TRUE; @@ -1158,13 +1157,7 @@ do_copy_texsubimage(GLcontext *ctx, src_format = strb->surface->format; dest_format = stImage->pt->format; - dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, - stImage->level, destZ, - PIPE_BUFFER_USAGE_CPU_WRITE); - - if (ctx->_ImageTransferState == 0x0 && - strb->surface->buffer && - dest_surface->buffer) { + if (ctx->_ImageTransferState == 0x0) { /* do blit-style copy */ /* XXX may need to invert image depending on window @@ -1187,6 +1180,14 @@ do_copy_texsubimage(GLcontext *ctx, x, y + height, dstx, dsty, width, height, GL_COPY); /* ? */ #else + struct pipe_surface *dest_surface; + + dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, + stImage->level, destZ, + PIPE_BUFFER_USAGE_GPU_WRITE); + + assert(strb->surface->buffer); + assert(dest_surface->buffer); if (src_format == dest_format) { pipe->surface_copy(pipe, @@ -1214,6 +1215,8 @@ do_copy_texsubimage(GLcontext *ctx, 0.0, PIPE_TEX_MIPFILTER_NEAREST); use_fallback = GL_FALSE; } + + pipe_surface_reference(&dest_surface, NULL); #endif } @@ -1224,8 +1227,6 @@ do_copy_texsubimage(GLcontext *ctx, srcX, srcY, width, height); } - pipe_surface_reference(&dest_surface, NULL); - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); } -- cgit v1.2.3 From d0279fc4b38c72356a341173317bcd45d9093f45 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 6 May 2008 12:24:04 +0100 Subject: gallium: Make sure to release texture surfaces (at the right time). --- src/mesa/state_tracker/st_cb_drawpixels.c | 3 ++- src/mesa/state_tracker/st_cb_texture.c | 2 ++ 2 files changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 8c775ad886..6ec3c343cd 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1038,12 +1038,13 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, } } + pipe_surface_reference(&psTex, NULL); + /* draw textured quad */ draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2], width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY, pt, stvp, stfp, color, GL_TRUE); - pipe_surface_reference(&psTex, NULL); pipe_texture_reference(&pt, NULL); } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 2b587a0848..1a1ab99f23 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1095,6 +1095,8 @@ fallback_copy_texsubimage(GLcontext *ctx, pipe_put_tile_rgba(pipe, dest_surf, destX, destY, width, 1, data); } } + + screen->tex_surface_release(screen, &dest_surf); } -- cgit v1.2.3 From 809dd9089bae70cf35cea6a75258e700e7455738 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 6 May 2008 08:37:28 -0600 Subject: gallium: sync up texture/sampler changes with master --- src/mesa/state_tracker/st_atom_sampler.c | 2 +- src/mesa/state_tracker/st_cb_texture.c | 24 +++++++++++++----------- 2 files changed, 14 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 4ce7c41e90..10283d31a1 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -150,7 +150,7 @@ update_samplers(struct st_context *st) sampler->normalized_coords = 1; sampler->lod_bias = st->ctx->Texture.Unit[su].LodBias; - sampler->min_lod = MAX2(0.0f, texobj->MinLod - texobj->BaseLevel); + sampler->min_lod = MAX2(0.0f, texobj->MinLod); sampler->max_lod = MIN2(texobj->MaxLevel - texobj->BaseLevel, texobj->MaxLod); diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 1a1ab99f23..a77b16e0ab 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1476,17 +1476,19 @@ st_finalize_texture(GLcontext *ctx, /* If we already have a gallium texture, check that it matches the texture * object's format, target, size, num_levels, etc. */ - if (stObj->pt && - (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || - stObj->pt->format != - st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat) || - stObj->pt->last_level < stObj->lastLevel || - stObj->pt->cpp != cpp || - stObj->pt->width[0] != firstImage->base.Width2 || - stObj->pt->height[0] != firstImage->base.Height2 || - stObj->pt->depth[0] != firstImage->base.Depth2 || - stObj->pt->compressed != firstImage->base.IsCompressed)) { - pipe_texture_release(&stObj->pt); + if (stObj->pt) { + const enum pipe_format fmt = + st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat); + if (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) || + stObj->pt->format != fmt || + stObj->pt->last_level < stObj->lastLevel || + stObj->pt->width[0] != firstImage->base.Width2 || + stObj->pt->height[0] != firstImage->base.Height2 || + stObj->pt->depth[0] != firstImage->base.Depth2 || + stObj->pt->cpp != cpp || + stObj->pt->compressed != firstImage->base.IsCompressed) { + pipe_texture_release(&stObj->pt); + } } /* May need to create a new gallium texture: -- cgit v1.2.3 From 66f703dca938f7749edc717fd8f690aba2d6e936 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 6 May 2008 08:40:41 -0600 Subject: gallium: change calculate_first_last_level() to match gallium-tex-surfaces branch --- src/mesa/state_tracker/st_cb_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 02ef961e4d..866683c23e 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1366,7 +1366,7 @@ calculate_first_last_level(struct st_texture_object *stObj) } else { firstLevel = 0; - lastLevel = MIN2(tObj->MaxLevel, tObj->Image[0][0]->WidthLog2); + lastLevel = MIN2(tObj->MaxLevel, tObj->Image[0][tObj->BaseLevel]->WidthLog2); } break; case GL_TEXTURE_RECTANGLE_NV: -- cgit v1.2.3 From a6ad4927740e5699f1a047f4c78f069f6a91c6ea Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 May 2008 02:51:49 +0900 Subject: gallium: Simple facility to dump and view images for debugging. --- src/gallium/auxiliary/util/p_debug.c | 47 ++++++++++++++++++++++++++++++++++++ src/gallium/include/pipe/p_debug.h | 11 +++++++++ 2 files changed, 58 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 4ec1746662..f971ee03ba 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -423,3 +423,50 @@ void debug_print_format(const char *msg, unsigned fmt ) debug_printf("%s: %s\n", msg, fmtstr); } #endif + + +#ifdef DEBUG +void debug_dump_image(const char *prefix, + unsigned format, unsigned cpp, + unsigned width, unsigned height, + unsigned pitch, + const void *data) +{ +#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY + static unsigned no = 0; + char filename[256]; + WCHAR wfilename[sizeof(filename)]; + ULONG_PTR iFile = 0; + struct { + unsigned format; + unsigned cpp; + unsigned width; + unsigned height; + } header; + unsigned char *pMap = NULL; + unsigned i; + + util_snprintf(filename, sizeof(filename), "\\??\\c:\\%03u%s.raw", ++no, prefix); + for(i = 0; i < sizeof(filename); ++i) + wfilename[i] = (WCHAR)filename[i]; + + pMap = (unsigned char *)EngMapFile(wfilename, sizeof(header) + cpp*width*height, &iFile); + if(!pMap) + return; + + header.format = format; + header.cpp = cpp; + header.width = width; + header.height = height; + memcpy(pMap, &header, sizeof(header)); + pMap += sizeof(header); + + for(i = 0; i < height; ++i) { + memcpy(pMap, (unsigned char *)data + cpp*pitch*i, cpp*width); + pMap += cpp*width; + } + + EngUnmapFile(iFile); +#endif +} +#endif \ No newline at end of file diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 7a7312ea12..1263d0da61 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -313,6 +313,17 @@ void debug_memory_end(unsigned long beginning); +#ifdef DEBUG +void debug_dump_image(const char *prefix, + unsigned format, unsigned cpp, + unsigned width, unsigned height, + unsigned pitch, + const void *data); +#else +#define debug_dump_image(prefix, format, cpp, width, height, pitch, data) ((void)0) +#endif + + #ifdef __cplusplus } #endif -- cgit v1.2.3 From f7dbd18371f9cb6686b6a97642b3ca5577e83472 Mon Sep 17 00:00:00 2001 From: Michal Danzer Date: Tue, 6 May 2008 12:34:34 -0600 Subject: gallium: move setup of dest_surface in do_copy_texsubimage() --- src/mesa/state_tracker/st_cb_texture.c | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index a77b16e0ab..f6f833a0db 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1129,6 +1129,7 @@ do_copy_texsubimage(GLcontext *ctx, struct st_renderbuffer *strb; struct pipe_context *pipe = ctx->st->pipe; struct pipe_screen *screen = pipe->screen; + struct pipe_surface *dest_surface; uint dest_format, src_format; uint do_flip = FALSE; GLboolean use_fallback = GL_TRUE; @@ -1159,7 +1160,13 @@ do_copy_texsubimage(GLcontext *ctx, src_format = strb->surface->format; dest_format = stImage->pt->format; - if (ctx->_ImageTransferState == 0x0) { + dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, + stImage->level, destZ, + PIPE_BUFFER_USAGE_CPU_WRITE); + + if (ctx->_ImageTransferState == 0x0 && + strb->surface->buffer && + dest_surface->buffer) { /* do blit-style copy */ /* XXX may need to invert image depending on window @@ -1182,14 +1189,6 @@ do_copy_texsubimage(GLcontext *ctx, x, y + height, dstx, dsty, width, height, GL_COPY); /* ? */ #else - struct pipe_surface *dest_surface; - - dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, - stImage->level, destZ, - PIPE_BUFFER_USAGE_GPU_WRITE); - - assert(strb->surface->buffer); - assert(dest_surface->buffer); if (src_format == dest_format) { pipe->surface_copy(pipe, @@ -1217,8 +1216,6 @@ do_copy_texsubimage(GLcontext *ctx, 0.0, PIPE_TEX_MIPFILTER_NEAREST); use_fallback = GL_FALSE; } - - pipe_surface_reference(&dest_surface, NULL); #endif } @@ -1229,6 +1226,8 @@ do_copy_texsubimage(GLcontext *ctx, srcX, srcY, width, height); } + pipe_surface_reference(&dest_surface, NULL); + if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); } -- cgit v1.2.3 From 973d0c014dba87308e358291de0730d38d50a733 Mon Sep 17 00:00:00 2001 From: Michal Danzer Date: Tue, 6 May 2008 12:35:25 -0600 Subject: gallium: create renderbuffer surface w/out CPU_READ/WRITE flags --- src/mesa/state_tracker/st_cb_fbo.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 1fa82b07e6..e0578f0b4d 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -385,8 +385,6 @@ st_render_texture(GLcontext *ctx, att->CubeMapFace, att->TextureLevel, att->Zoffset, - PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE); assert(strb->surface); -- cgit v1.2.3 From 296378b6c8b205048244746e260739448c4ee590 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 6 May 2008 13:47:41 -0600 Subject: gallium: create drawing surfaces as GPU_READ/WRITE only Create different temporary surfaces for CPU_READ/WRITE when needed (such as for glReadPixels, glAccum, some glCopy/DrawPixels, glCopyTexSubImage, etc). --- src/mesa/state_tracker/st_cb_accum.c | 73 ++++++++++++++++++++++--------- src/mesa/state_tracker/st_cb_drawpixels.c | 33 +++++++++++--- src/mesa/state_tracker/st_cb_fbo.c | 25 ++++++----- src/mesa/state_tracker/st_cb_fbo.h | 3 +- src/mesa/state_tracker/st_cb_readpixels.c | 34 +++++++++----- src/mesa/state_tracker/st_cb_texture.c | 10 ++++- 6 files changed, 129 insertions(+), 49 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index e4ef3e16b7..8098d75e18 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -105,7 +105,7 @@ void st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) { struct st_renderbuffer *acc_strb = st_renderbuffer(rb); - struct pipe_surface *acc_ps = acc_strb->surface; + struct pipe_surface *acc_ps; struct pipe_screen *screen = ctx->st->pipe->screen; const GLint xpos = ctx->DrawBuffer->_Xmin; const GLint ypos = ctx->DrawBuffer->_Ymin; @@ -113,6 +113,8 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) const GLint height = ctx->DrawBuffer->_Ymax - ypos; GLvoid *map; + acc_ps = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); map = screen->surface_map(screen, acc_ps, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -143,6 +145,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) } screen->surface_unmap(screen, acc_ps); + pipe_surface_reference(&acc_ps, NULL); } @@ -185,70 +188,100 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias, static void accum_accum(struct pipe_context *pipe, GLfloat value, GLint xpos, GLint ypos, GLint width, GLint height, - struct pipe_surface *acc_ps, - struct pipe_surface *color_ps) + struct st_renderbuffer *acc_strb, + struct st_renderbuffer *color_strb) { + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *acc_surf, *color_surf; GLfloat *colorBuf, *accBuf; GLint i; + acc_surf = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0, + (PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_CPU_READ)); + + color_surf = screen->get_tex_surface(screen, color_strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_READ); + colorBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, colorBuf); - acc_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + pipe_get_tile_rgba(pipe, color_surf, xpos, ypos, width, height, colorBuf); + acc_get_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { accBuf[i] = accBuf[i] + colorBuf[i] * value; } - acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + acc_put_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, accBuf); free(colorBuf); free(accBuf); + pipe_surface_reference(&acc_surf, NULL); + pipe_surface_reference(&color_surf, NULL); } static void accum_load(struct pipe_context *pipe, GLfloat value, GLint xpos, GLint ypos, GLint width, GLint height, - struct pipe_surface *acc_ps, - struct pipe_surface *color_ps) + struct st_renderbuffer *acc_strb, + struct st_renderbuffer *color_strb) { + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *acc_surf, *color_surf; GLfloat *buf; GLint i; + acc_surf = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); + + color_surf = screen->get_tex_surface(screen, color_strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_READ); + buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, buf); + pipe_get_tile_rgba(pipe, color_surf, xpos, ypos, width, height, buf); for (i = 0; i < 4 * width * height; i++) { buf[i] = buf[i] * value; } - acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, buf); + acc_put_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, buf); free(buf); + pipe_surface_reference(&acc_surf, NULL); + pipe_surface_reference(&color_surf, NULL); } static void accum_return(GLcontext *ctx, GLfloat value, GLint xpos, GLint ypos, GLint width, GLint height, - struct pipe_surface *acc_ps, - struct pipe_surface *color_ps) + struct st_renderbuffer *acc_strb, + struct st_renderbuffer *color_strb) { struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; const GLubyte *colormask = ctx->Color.ColorMask; + struct pipe_surface *acc_surf, *color_surf; GLfloat *abuf, *cbuf = NULL; GLint i, ch; abuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - acc_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); + acc_surf = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_READ); + + color_surf = screen->get_tex_surface(screen, color_strb->texture, 0, 0, 0, + (PIPE_BUFFER_USAGE_CPU_READ | + PIPE_BUFFER_USAGE_CPU_WRITE)); + + acc_get_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, abuf); if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) { cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, cbuf); + pipe_get_tile_rgba(pipe, color_surf, xpos, ypos, width, height, cbuf); } for (i = 0; i < width * height; i++) { @@ -263,11 +296,13 @@ accum_return(GLcontext *ctx, GLfloat value, } } - pipe_put_tile_rgba(pipe, color_ps, xpos, ypos, width, height, abuf); + pipe_put_tile_rgba(pipe, color_surf, xpos, ypos, width, height, abuf); free(abuf); if (cbuf) free(cbuf); + pipe_surface_reference(&acc_surf, NULL); + pipe_surface_reference(&color_surf, NULL); } @@ -280,8 +315,6 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value) = st_renderbuffer(ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer); struct st_renderbuffer *color_strb = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); - struct pipe_surface *acc_ps = acc_strb->surface; - struct pipe_surface *color_ps = color_strb->surface; const GLint xpos = ctx->DrawBuffer->_Xmin; const GLint ypos = ctx->DrawBuffer->_Ymin; @@ -304,14 +337,14 @@ st_Accum(GLcontext *ctx, GLenum op, GLfloat value) break; case GL_ACCUM: if (value != 0.0F) { - accum_accum(pipe, value, xpos, ypos, width, height, acc_ps, color_ps); + accum_accum(pipe, value, xpos, ypos, width, height, acc_strb, color_strb); } break; case GL_LOAD: - accum_load(pipe, value, xpos, ypos, width, height, acc_ps, color_ps); + accum_load(pipe, value, xpos, ypos, width, height, acc_strb, color_strb); break; case GL_RETURN: - accum_return(ctx, value, xpos, ypos, width, height, acc_ps, color_ps); + accum_return(ctx, value, xpos, ypos, width, height, acc_strb, color_strb); break; default: assert(0); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 6ec3c343cd..c967c989de 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -734,13 +734,19 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, struct st_context *st = ctx->st; struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; - struct pipe_surface *ps = st->state.framebuffer.zsbuf; + struct st_renderbuffer *strb; + struct pipe_surface *ps; const GLboolean zoom = ctx->Pixel.ZoomX != 1.0 || ctx->Pixel.ZoomY != 1.0; GLint skipPixels; ubyte *stmap; pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + strb = st_renderbuffer(ctx->DrawBuffer-> + Attachment[BUFFER_STENCIL].Renderbuffer); + ps = screen->get_tex_surface(screen, strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); + /* map the stencil buffer */ stmap = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -801,6 +807,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, /* unmap the stencil buffer */ screen->surface_unmap(screen, ps); + pipe_surface_reference(&ps, NULL); } @@ -874,7 +881,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, { struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer); struct pipe_screen *screen = ctx->st->pipe->screen; - struct pipe_surface *psDraw = rbDraw->surface; + struct pipe_surface *psDraw; ubyte *drawMap; ubyte *buffer; int i; @@ -889,6 +896,9 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, st_read_stencil_pixels(ctx, srcx, srcy, width, height, GL_UNSIGNED_BYTE, &ctx->DefaultPacking, buffer); + psDraw = screen->get_tex_surface(screen, rbDraw->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE); + /* map the stencil buffer */ drawMap = screen->surface_map(screen, psDraw, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -931,6 +941,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, /* unmap the stencil buffer */ screen->surface_unmap(screen, psDraw); + pipe_surface_reference(&psDraw, NULL); } @@ -945,7 +956,6 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, struct st_renderbuffer *rbRead; struct st_vertex_program *stvp; struct st_fragment_program *stfp; - struct pipe_surface *psRead; struct pipe_surface *psTex; struct pipe_texture *pt; GLfloat *color; @@ -976,8 +986,12 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE); } +#if 0 psRead = rbRead->surface; srcFormat = psRead->format; +#else + srcFormat = rbRead->texture->format; +#endif if (screen->is_format_supported(screen, srcFormat, PIPE_TEXTURE)) { texFormat = srcFormat; @@ -1005,18 +1019,26 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, } if (srcFormat == texFormat) { + /* copy source framebuffer surface into mipmap/texture */ + struct pipe_surface *psRead = screen->get_tex_surface(screen, + rbRead->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_GPU_READ); psTex = screen->get_tex_surface(screen, pt, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_WRITE ); - - /* copy source framebuffer surface into mipmap/texture */ pipe->surface_copy(pipe, FALSE, psTex, /* dest */ 0, 0, /* destx/y */ psRead, srcx, srcy, width, height); + pipe_surface_reference(&psRead, NULL); } else { + /* CPU-based fallback/conversion */ + struct pipe_surface *psRead = screen->get_tex_surface(screen, + rbRead->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_READ); + psTex = screen->get_tex_surface(screen, pt, 0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE ); @@ -1036,6 +1058,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, pipe_put_tile_z(pipe, psTex, 0, 0, width, height, buf); free(buf); } + pipe_surface_reference(&psRead, NULL); } pipe_surface_reference(&psTex, NULL); diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index e0578f0b4d..747d4905e6 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -95,7 +95,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, { struct pipe_context *pipe = ctx->st->pipe; struct st_renderbuffer *strb = st_renderbuffer(rb); - struct pipe_texture template, *texture; + struct pipe_texture template; unsigned surface_usage; /* Free the old surface (and texture if we hold the last @@ -136,12 +136,14 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, /* Probably need dedicated flags for surface usage too: */ surface_usage = (PIPE_BUFFER_USAGE_GPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE | + PIPE_BUFFER_USAGE_GPU_WRITE); +#if 0 PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE); +#endif - texture = pipe->screen->texture_create( pipe->screen, - &template ); + strb->texture = pipe->screen->texture_create( pipe->screen, + &template ); /* Special path for accum buffers. * @@ -149,7 +151,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, * only for now, the surface pixel format doesn't really matter, * only that the buffer is large enough. */ - if (!texture && template.format == DEFAULT_ACCUM_PIPE_FORMAT) + if (!strb->texture && template.format == DEFAULT_ACCUM_PIPE_FORMAT) { /* Actually, just setting this usage value should be sufficient * to tell the driver to go ahead and allocate the buffer, even @@ -159,21 +161,19 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, surface_usage = (PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE); - texture = pipe->screen->texture_create( pipe->screen, - &template ); + strb->texture = pipe->screen->texture_create( pipe->screen, + &template ); } - if (!texture) + if (!strb->texture) return FALSE; strb->surface = pipe->screen->get_tex_surface( pipe->screen, - texture, + strb->texture, 0, 0, 0, surface_usage ); - pipe_texture_reference( &texture, NULL ); - assert(strb->surface->buffer); assert(strb->surface->format); assert(strb->surface->cpp); @@ -195,6 +195,7 @@ st_renderbuffer_delete(struct gl_renderbuffer *rb) struct st_renderbuffer *strb = st_renderbuffer(rb); ASSERT(strb); pipe_surface_reference(&strb->surface, NULL); + pipe_texture_reference(&strb->texture, NULL); free(strb); } @@ -380,6 +381,8 @@ st_render_texture(GLcontext *ctx, rb->Width = pt->width[att->TextureLevel]; rb->Height = pt->height[att->TextureLevel]; + pipe_texture_reference( &strb->texture, pt ); + /* the renderbuffer's surface is inside the texture */ strb->surface = screen->get_tex_surface(screen, pt, att->CubeMapFace, diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index c1aa14f9b2..f9cec91314 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -41,7 +41,8 @@ struct st_renderbuffer { struct gl_renderbuffer Base; - struct pipe_surface *surface; + struct pipe_texture *texture; + struct pipe_surface *surface; /* temporary view into texture */ enum pipe_format format; /** preferred format, or PIPE_FORMAT_NONE */ }; diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index e242195e7a..0b2b9d544d 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -63,10 +63,14 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, struct gl_framebuffer *fb = ctx->ReadBuffer; struct pipe_screen *screen = ctx->st->pipe->screen; struct st_renderbuffer *strb = st_renderbuffer(fb->_StencilBuffer); - struct pipe_surface *ps = strb->surface; + struct pipe_surface *ps; ubyte *stmap; GLint j; + /* Create a CPU-READ surface/view into the renderbuffer's texture */ + ps = screen->get_tex_surface(screen, strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_READ); + /* map the stencil buffer */ stmap = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ); @@ -126,6 +130,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, /* unmap the stencil buffer */ screen->surface_unmap(screen, ps); + pipe_surface_reference(&ps, NULL); } @@ -169,12 +174,14 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLvoid *dest) { struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; GLfloat temp[MAX_WIDTH][4]; const GLbitfield transferOps = ctx->_ImageTransferState; GLint i, yStep, dfStride; GLfloat *df; struct st_renderbuffer *strb; struct gl_pixelstore_attrib clippedPacking = *pack; + struct pipe_surface *surf; /* XXX convolution not done yet */ assert((transferOps & IMAGE_CONVOLUTION_BIT) == 0); @@ -230,6 +237,10 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, yStep = 1; } + /* Create a CPU-READ surface/view into the renderbuffer's texture */ + surf = screen->get_tex_surface(screen, strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_READ); + /* * Copy pixels from pipe_surface to user memory */ @@ -241,15 +252,14 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, const GLint dstStride = _mesa_image_row_stride(&clippedPacking, width, format, type); - if (strb->surface->format == PIPE_FORMAT_S8Z24_UNORM || - strb->surface->format == PIPE_FORMAT_X8Z24_UNORM) { + if (surf->format == PIPE_FORMAT_S8Z24_UNORM || + surf->format == PIPE_FORMAT_X8Z24_UNORM) { if (format == GL_DEPTH_COMPONENT) { for (i = 0; i < height; i++) { GLuint ztemp[MAX_WIDTH], j; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / ((1 << 24) - 1); - pipe_get_tile_raw(pipe, strb->surface, x, y, - width, 1, ztemp, 0); + pipe_get_tile_raw(pipe, surf, x, y, width, 1, ztemp, 0); y += yStep; for (j = 0; j < width; j++) { zfloat[j] = (float) (scale * (ztemp[j] & 0xffffff)); @@ -263,18 +273,18 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, /* untested, but simple: */ assert(format == GL_DEPTH_STENCIL_EXT); for (i = 0; i < height; i++) { - pipe_get_tile_raw(pipe, strb->surface, x, y, width, 1, dst, 0); + pipe_get_tile_raw(pipe, surf, x, y, width, 1, dst, 0); y += yStep; dst += dstStride; } } } - else if (strb->surface->format == PIPE_FORMAT_Z16_UNORM) { + else if (surf->format == PIPE_FORMAT_Z16_UNORM) { for (i = 0; i < height; i++) { GLushort ztemp[MAX_WIDTH], j; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / 0xffff; - pipe_get_tile_raw(pipe, strb->surface, x, y, width, 1, ztemp, 0); + pipe_get_tile_raw(pipe, surf, x, y, width, 1, ztemp, 0); y += yStep; for (j = 0; j < width; j++) { zfloat[j] = (float) (scale * ztemp[j]); @@ -284,12 +294,12 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, dst += dstStride; } } - else if (strb->surface->format == PIPE_FORMAT_Z32_UNORM) { + else if (surf->format == PIPE_FORMAT_Z32_UNORM) { for (i = 0; i < height; i++) { GLuint ztemp[MAX_WIDTH], j; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / 0xffffffff; - pipe_get_tile_raw(pipe, strb->surface, x, y, width, 1, ztemp, 0); + pipe_get_tile_raw(pipe, surf, x, y, width, 1, ztemp, 0); y += yStep; for (j = 0; j < width; j++) { zfloat[j] = (float) (scale * ztemp[j]); @@ -303,7 +313,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, /* RGBA format */ /* Do a row at a time to flip image data vertically */ for (i = 0; i < height; i++) { - pipe_get_tile_rgba(pipe, strb->surface, x, y, width, 1, df); + pipe_get_tile_rgba(pipe, surf, x, y, width, 1, df); y += yStep; df += dfStride; if (!dfStride) { @@ -315,6 +325,8 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } } + pipe_surface_reference(&surf, NULL); + _mesa_unmap_readpix_pbo(ctx, &clippedPacking); } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index f6f833a0db..828b2340f2 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -306,6 +306,11 @@ guess_and_alloc_texture(struct st_context *st, depth <<= 1; } + if (width == 0 || height == 0 || depth == 0) { + /* no texture needed */ + return; + } + /* Guess a reasonable value for lastLevel. This is probably going * to be wrong fairly often and might mean that we have to look at * resizable buffers, or require that buffers implement lazy @@ -1059,6 +1064,8 @@ fallback_copy_texsubimage(GLcontext *ctx, } src_surf = strb->surface; + src_surf = screen->get_tex_surface(screen, strb->texture, face, level, destZ, + PIPE_BUFFER_USAGE_CPU_READ); dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -1097,6 +1104,7 @@ fallback_copy_texsubimage(GLcontext *ctx, } screen->tex_surface_release(screen, &dest_surf); + screen->tex_surface_release(screen, &src_surf); } @@ -1164,7 +1172,7 @@ do_copy_texsubimage(GLcontext *ctx, stImage->level, destZ, PIPE_BUFFER_USAGE_CPU_WRITE); - if (ctx->_ImageTransferState == 0x0 && + if (0&& ctx->_ImageTransferState == 0x0 && strb->surface->buffer && dest_surface->buffer) { /* do blit-style copy */ -- cgit v1.2.3 From 6b0fd0f1572cd6af1e931b70f75852de25d32649 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 6 May 2008 15:39:39 -0600 Subject: gallium: re-enable temporarily disabled code in do_copy_texsubimage() --- src/mesa/state_tracker/st_cb_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 828b2340f2..3723b91210 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1172,7 +1172,7 @@ do_copy_texsubimage(GLcontext *ctx, stImage->level, destZ, PIPE_BUFFER_USAGE_CPU_WRITE); - if (0&& ctx->_ImageTransferState == 0x0 && + if (ctx->_ImageTransferState == 0x0 && strb->surface->buffer && dest_surface->buffer) { /* do blit-style copy */ -- cgit v1.2.3 From 10f6ae0355937615d137c79c060b9e5a923f0d65 Mon Sep 17 00:00:00 2001 From: Brian Date: Thu, 1 May 2008 19:29:25 -0600 Subject: mesa: comments, whitespace --- src/mesa/main/ffvertex_prog.c | 43 +++++++++++++++++++------------------------ 1 file changed, 19 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index f648d081c0..810af9e33e 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -946,35 +946,33 @@ static void build_lighting( struct tnl_program *p ) _bfc1 = _bfc0; } - /* If no lights, still need to emit the scenecolor. */ - { - struct ureg res0 = register_output( p, VERT_RESULT_COL0 ); - emit_op1(p, OPCODE_MOV, res0, 0, _col0); - } + { + struct ureg res0 = register_output( p, VERT_RESULT_COL0 ); + emit_op1(p, OPCODE_MOV, res0, 0, _col0); + } - if (separate) { - struct ureg res1 = register_output( p, VERT_RESULT_COL1 ); - emit_op1(p, OPCODE_MOV, res1, 0, _col1); - } + if (separate) { + struct ureg res1 = register_output( p, VERT_RESULT_COL1 ); + emit_op1(p, OPCODE_MOV, res1, 0, _col1); + } - if (twoside) { - struct ureg res0 = register_output( p, VERT_RESULT_BFC0 ); - emit_op1(p, OPCODE_MOV, res0, 0, _bfc0); - } + if (twoside) { + struct ureg res0 = register_output( p, VERT_RESULT_BFC0 ); + emit_op1(p, OPCODE_MOV, res0, 0, _bfc0); + } - if (twoside && separate) { - struct ureg res1 = register_output( p, VERT_RESULT_BFC1 ); - emit_op1(p, OPCODE_MOV, res1, 0, _bfc1); - } + if (twoside && separate) { + struct ureg res1 = register_output( p, VERT_RESULT_BFC1 ); + emit_op1(p, OPCODE_MOV, res1, 0, _bfc1); + } if (nr_lights == 0) { release_temps(p); return; } - for (i = 0; i < MAX_LIGHTS; i++) { if (p->state->unit[i].light_enabled) { struct ureg half = undef; @@ -1006,7 +1004,7 @@ static void build_lighting( struct tnl_program *p ) VPpli = get_temp(p); half = get_temp(p); - /* Calulate VPpli vector + /* Calculate VPpli vector */ emit_op2(p, OPCODE_SUB, VPpli, 0, Ppli, V); @@ -1017,15 +1015,13 @@ static void build_lighting( struct tnl_program *p ) emit_op1(p, OPCODE_RSQ, dist, 0, dist); emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist); - - /* Calculate attenuation: + /* Calculate attenuation: */ if (!p->state->unit[i].light_spotcutoff_is_180 || p->state->unit[i].light_attenuated) { att = calculate_light_attenuation(p, i, VPpli, dist); } - - + /* Calculate viewer direction, or use infinite viewer: */ if (p->state->light_local_viewer) { @@ -1047,7 +1043,6 @@ static void build_lighting( struct tnl_program *p ) emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half); - /* Front face lighting: */ { -- cgit v1.2.3 From 103ae5d16fd9fef566096570f731bb634a8025d4 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 May 2008 22:13:06 -0600 Subject: gallium: implement full reference counting for vertex/fragment programs Use _mesa_reference_vert/fragprog() wherever we assign program pointers. Fixes a memory corruption bug found with glean/api2 test. Another memory bug involving shaders yet to be fixed... --- src/mesa/main/context.c | 26 ++++---- src/mesa/main/mtypes.h | 4 +- src/mesa/main/state.c | 32 ++++++---- src/mesa/shader/prog_cache.c | 5 +- src/mesa/shader/program.c | 103 +++++++++++++++++++++----------- src/mesa/shader/program.h | 22 +++++++ src/mesa/shader/shader_api.c | 6 +- src/mesa/shader/slang/slang_link.c | 12 ++-- src/mesa/state_tracker/st_atom_shader.c | 7 ++- src/mesa/state_tracker/st_context.c | 3 + src/mesa/state_tracker/st_program.h | 24 +++++++- 11 files changed, 170 insertions(+), 74 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 7b8d934170..b9053344b2 100755 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -150,8 +150,6 @@ int MESA_DEBUG_FLAGS = 0; /* ubyte -> float conversion */ GLfloat _mesa_ubyte_to_float_color_tab[256]; -static void -free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ); /** @@ -423,12 +421,14 @@ alloc_shared_state( GLcontext *ctx ) #endif #if FEATURE_ARB_vertex_program - ss->DefaultVertexProgram = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); + ss->DefaultVertexProgram = (struct gl_vertex_program *) + ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); if (!ss->DefaultVertexProgram) goto cleanup; #endif #if FEATURE_ARB_fragment_program - ss->DefaultFragmentProgram = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + ss->DefaultFragmentProgram = (struct gl_fragment_program *) + ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!ss->DefaultFragmentProgram) goto cleanup; #endif @@ -513,12 +513,10 @@ alloc_shared_state( GLcontext *ctx ) _mesa_DeleteHashTable(ss->Programs); #endif #if FEATURE_ARB_vertex_program - if (ss->DefaultVertexProgram) - ctx->Driver.DeleteProgram(ctx, ss->DefaultVertexProgram); + _mesa_reference_vertprog(ctx, &ss->DefaultVertexProgram, NULL); #endif #if FEATURE_ARB_fragment_program - if (ss->DefaultFragmentProgram) - ctx->Driver.DeleteProgram(ctx, ss->DefaultFragmentProgram); + _mesa_reference_fragprog(ctx, &ss->DefaultFragmentProgram, NULL); #endif #if FEATURE_ATI_fragment_shader if (ss->DefaultFragmentShader) @@ -695,10 +693,10 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) _mesa_DeleteHashTable(ss->Programs); #endif #if FEATURE_ARB_vertex_program - ctx->Driver.DeleteProgram(ctx, ss->DefaultVertexProgram); + _mesa_reference_vertprog(ctx, &ss->DefaultVertexProgram, NULL); #endif #if FEATURE_ARB_fragment_program - ctx->Driver.DeleteProgram(ctx, ss->DefaultFragmentProgram); + _mesa_reference_fragprog(ctx, &ss->DefaultFragmentProgram, NULL); #endif #if FEATURE_ATI_fragment_shader @@ -1190,6 +1188,14 @@ _mesa_free_context_data( GLcontext *ctx ) _mesa_unreference_framebuffer(&ctx->ReadBuffer); } + _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, NULL); + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, NULL); + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._TnlProgram, NULL); + + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, NULL); + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL); + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, NULL); + _mesa_free_lighting_data( ctx ); _mesa_free_eval_data( ctx ); _mesa_free_texture_data( ctx ); diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 50b22d25bf..463142fe39 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2203,10 +2203,10 @@ struct gl_shared_state /*@{*/ struct _mesa_HashTable *Programs; /**< All vertex/fragment programs */ #if FEATURE_ARB_vertex_program - struct gl_program *DefaultVertexProgram; + struct gl_vertex_program *DefaultVertexProgram; #endif #if FEATURE_ARB_fragment_program - struct gl_program *DefaultFragmentProgram; + struct gl_fragment_program *DefaultFragmentProgram; #endif /*@}*/ diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 0b1c56fdd5..90379a1772 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -982,16 +982,20 @@ update_program(GLcontext *ctx) #endif if (shProg && shProg->LinkStatus && shProg->FragmentProgram) { /* user-defined fragment shader */ - ctx->FragmentProgram._Current = shProg->FragmentProgram; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, + shProg->FragmentProgram); } else if (ctx->FragmentProgram._Enabled) { /* use user-defined fragment program */ - ctx->FragmentProgram._Current = ctx->FragmentProgram.Current; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, + ctx->FragmentProgram.Current); } else if (ctx->FragmentProgram._MaintainTexEnvProgram) { /* fragment program generated from fixed-function state */ - ctx->FragmentProgram._Current = _mesa_get_fixed_func_fragment_program(ctx); - ctx->FragmentProgram._TexEnvProgram = ctx->FragmentProgram._Current; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, + _mesa_get_fixed_func_fragment_program(ctx)); + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, + ctx->FragmentProgram._Current); /* XXX get rid of this confusing stuff someday? */ ctx->FragmentProgram._Active = ctx->FragmentProgram._Enabled; @@ -1000,7 +1004,7 @@ update_program(GLcontext *ctx) } else { /* no fragment program */ - ctx->FragmentProgram._Current = NULL; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL); } if (ctx->FragmentProgram._Current != prevFP && ctx->Driver.BindProgram) { @@ -1013,29 +1017,33 @@ update_program(GLcontext *ctx) **/ #if 1 /* XXX get rid of this someday? */ - ctx->VertexProgram._TnlProgram = NULL; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._TnlProgram, NULL); #endif if (shProg && shProg->LinkStatus && shProg->VertexProgram) { /* user-defined vertex shader */ - ctx->VertexProgram._Current = shProg->VertexProgram; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, + shProg->VertexProgram); } else if (ctx->VertexProgram._Enabled) { /* use user-defined vertex program */ - ctx->VertexProgram._Current = ctx->VertexProgram.Current; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, + ctx->VertexProgram.Current); } else if (ctx->VertexProgram._MaintainTnlProgram) { /* vertex program generated from fixed-function state */ - ctx->VertexProgram._Current = _mesa_get_fixed_func_vertex_program(ctx); - ctx->VertexProgram._TnlProgram = ctx->VertexProgram._Current; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, + _mesa_get_fixed_func_vertex_program(ctx)); + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._TnlProgram, + ctx->VertexProgram._Current); } else { /* no vertex program / used fixed-function code */ - ctx->VertexProgram._Current = NULL; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, NULL); } if (ctx->VertexProgram._Current != prevVP && ctx->Driver.BindProgram) { ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB, - (struct gl_program *) ctx->VertexProgram._Current); + (struct gl_program *) ctx->VertexProgram._Current); } } diff --git a/src/mesa/shader/prog_cache.c b/src/mesa/shader/prog_cache.c index dd0241ef24..36a25377c5 100644 --- a/src/mesa/shader/prog_cache.c +++ b/src/mesa/shader/prog_cache.c @@ -30,6 +30,7 @@ #include "main/mtypes.h" #include "main/imports.h" #include "shader/prog_cache.h" +#include "shader/program.h" struct cache_item @@ -109,7 +110,7 @@ clear_cache(GLcontext *ctx, struct gl_program_cache *cache) for (c = cache->items[i]; c; c = next) { next = c->next; _mesa_free(c->key); - ctx->Driver.DeleteProgram(ctx, c->program); + _mesa_reference_program(ctx, &c->program, NULL); _mesa_free(c); } cache->items[i] = NULL; @@ -177,7 +178,7 @@ _mesa_program_cache_insert(GLcontext *ctx, c->key = _mesa_malloc(keysize); memcpy(c->key, key, keysize); - c->program = program; + c->program = program; /* no refcount change */ if (cache->n_items > cache->size * 1.5) { if (cache->size < 1000) diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 0ed7f833d2..9a23c5d7d3 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -60,9 +60,9 @@ _mesa_init_program(GLcontext *ctx) ctx->VertexProgram.Enabled = GL_FALSE; ctx->VertexProgram.PointSizeEnabled = GL_FALSE; ctx->VertexProgram.TwoSideEnabled = GL_FALSE; - ctx->VertexProgram.Current = (struct gl_vertex_program *) ctx->Shared->DefaultVertexProgram; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, + ctx->Shared->DefaultVertexProgram); assert(ctx->VertexProgram.Current); - ctx->VertexProgram.Current->Base.RefCount++; for (i = 0; i < MAX_NV_VERTEX_PROGRAM_PARAMS / 4; i++) { ctx->VertexProgram.TrackMatrix[i] = GL_NONE; ctx->VertexProgram.TrackMatrixTransform[i] = GL_IDENTITY_NV; @@ -72,9 +72,9 @@ _mesa_init_program(GLcontext *ctx) #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program ctx->FragmentProgram.Enabled = GL_FALSE; - ctx->FragmentProgram.Current = (struct gl_fragment_program *) ctx->Shared->DefaultFragmentProgram; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, + ctx->Shared->DefaultFragmentProgram); assert(ctx->FragmentProgram.Current); - ctx->FragmentProgram.Current->Base.RefCount++; ctx->FragmentProgram.Cache = _mesa_new_program_cache(); #endif @@ -96,19 +96,11 @@ void _mesa_free_program_data(GLcontext *ctx) { #if FEATURE_NV_vertex_program || FEATURE_ARB_vertex_program - if (ctx->VertexProgram.Current) { - ctx->VertexProgram.Current->Base.RefCount--; - if (ctx->VertexProgram.Current->Base.RefCount <= 0) - ctx->Driver.DeleteProgram(ctx, &(ctx->VertexProgram.Current->Base)); - } + _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, NULL); _mesa_delete_program_cache(ctx, ctx->VertexProgram.Cache); #endif #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program - if (ctx->FragmentProgram.Current) { - ctx->FragmentProgram.Current->Base.RefCount--; - if (ctx->FragmentProgram.Current->Base.RefCount <= 0) - ctx->Driver.DeleteProgram(ctx, &(ctx->FragmentProgram.Current->Base)); - } + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, NULL); _mesa_delete_program_cache(ctx, ctx->FragmentProgram.Cache); #endif /* XXX probably move this stuff */ @@ -325,6 +317,59 @@ _mesa_lookup_program(GLcontext *ctx, GLuint id) } +/** + * Reference counting for vertex/fragment programs + */ +void +_mesa_reference_program(GLcontext *ctx, + struct gl_program **ptr, + struct gl_program *prog) +{ + assert(ptr); + if (*ptr && prog) { + /* sanity check */ + ASSERT((*ptr)->Target == prog->Target); + } + if (*ptr == prog) { + return; /* no change */ + } + if (*ptr) { + GLboolean deleteFlag; + + /*_glthread_LOCK_MUTEX((*ptr)->Mutex);*/ +#if 0 + printf("Program %p %u 0x%x Refcount-- to %d\n", + *ptr, (*ptr)->Id, (*ptr)->Target, (*ptr)->RefCount - 1); +#endif + ASSERT((*ptr)->RefCount > 0); + (*ptr)->RefCount--; + + deleteFlag = ((*ptr)->RefCount == 0); + /*_glthread_UNLOCK_MUTEX((*ptr)->Mutex);*/ + + if (deleteFlag) { + ASSERT(ctx); + ctx->Driver.DeleteProgram(ctx, *ptr); + } + + *ptr = NULL; + } + + assert(!*ptr); + if (prog) { + /*_glthread_LOCK_MUTEX(prog->Mutex);*/ + prog->RefCount++; +#if 0 + printf("Program %p %u 0x%x Refcount++ to %d\n", + prog, prog->Id, prog->Target, prog->RefCount); +#endif + /*_glthread_UNLOCK_MUTEX(prog->Mutex);*/ + } + + *ptr = prog; +} + + /** * Return a copy of a program. * XXX Problem here if the program object is actually OO-derivation @@ -340,8 +385,9 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog) return NULL; assert(clone->Target == prog->Target); + assert(clone->RefCount == 1); + clone->String = (GLubyte *) _mesa_strdup((char *) prog->String); - clone->RefCount = 1; clone->Format = prog->Format; clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions); if (!clone->Instructions) { @@ -704,9 +750,9 @@ _mesa_BindProgram(GLenum target, GLuint id) /* Bind a default program */ newProg = NULL; if (target == GL_VERTEX_PROGRAM_ARB) /* == GL_VERTEX_PROGRAM_NV */ - newProg = ctx->Shared->DefaultVertexProgram; + newProg = &ctx->Shared->DefaultVertexProgram->Base; else - newProg = ctx->Shared->DefaultFragmentProgram; + newProg = &ctx->Shared->DefaultFragmentProgram->Base; } else { /* Bind a user program */ @@ -734,26 +780,16 @@ _mesa_BindProgram(GLenum target, GLuint id) return; } - /* unbind/delete oldProg */ - if (curProg->Id != 0) { - /* decrement refcount on previously bound fragment program */ - curProg->RefCount--; - /* and delete if refcount goes below one */ - if (curProg->RefCount <= 0) { - /* the program ID was already removed from the hash table */ - ctx->Driver.DeleteProgram(ctx, curProg); - } - } - /* bind newProg */ if (target == GL_VERTEX_PROGRAM_ARB) { /* == GL_VERTEX_PROGRAM_NV */ - ctx->VertexProgram.Current = (struct gl_vertex_program *) newProg; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, + (struct gl_vertex_program *) newProg); } else if (target == GL_FRAGMENT_PROGRAM_NV || target == GL_FRAGMENT_PROGRAM_ARB) { - ctx->FragmentProgram.Current = (struct gl_fragment_program *) newProg; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, + (struct gl_fragment_program *) newProg); } - newProg->RefCount++; /* Never null pointers */ ASSERT(ctx->VertexProgram.Current); @@ -811,10 +847,7 @@ _mesa_DeletePrograms(GLsizei n, const GLuint *ids) } /* The ID is immediately available for re-use now */ _mesa_HashRemove(ctx->Shared->Programs, ids[i]); - prog->RefCount--; - if (prog->RefCount <= 0) { - ctx->Driver.DeleteProgram(ctx, prog); - } + _mesa_reference_program(ctx, &prog, NULL); } } } diff --git a/src/mesa/shader/program.h b/src/mesa/shader/program.h index 414a57d39c..08fe576afc 100644 --- a/src/mesa/shader/program.h +++ b/src/mesa/shader/program.h @@ -83,6 +83,28 @@ _mesa_delete_program(GLcontext *ctx, struct gl_program *prog); extern struct gl_program * _mesa_lookup_program(GLcontext *ctx, GLuint id); +extern void +_mesa_reference_program(GLcontext *ctx, + struct gl_program **ptr, + struct gl_program *prog); + +static INLINE void +_mesa_reference_vertprog(GLcontext *ctx, + struct gl_vertex_program **ptr, + struct gl_vertex_program *prog) +{ + _mesa_reference_program(ctx, (struct gl_program **) ptr, + (struct gl_program *) prog); +} + +static INLINE void +_mesa_reference_fragprog(GLcontext *ctx, + struct gl_fragment_program **ptr, + struct gl_fragment_program *prog) +{ + _mesa_reference_program(ctx, (struct gl_program **) ptr, + (struct gl_program *) prog); +} extern struct gl_program * _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog); diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 9c419c9903..f12fa28d97 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -80,8 +80,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx, * original/unlinked program. */ shProg->VertexProgram->Base.Parameters = NULL; - ctx->Driver.DeleteProgram(ctx, &shProg->VertexProgram->Base); - shProg->VertexProgram = NULL; + _mesa_reference_vertprog(ctx, &shProg->VertexProgram, NULL); } if (shProg->FragmentProgram) { @@ -89,8 +88,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx, * original/unlinked program. */ shProg->FragmentProgram->Base.Parameters = NULL; - ctx->Driver.DeleteProgram(ctx, &shProg->FragmentProgram->Base); - shProg->FragmentProgram = NULL; + _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, NULL); } if (shProg->Uniforms) { diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index addff20421..ae581553dc 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -410,19 +410,19 @@ _slang_link(GLcontext *ctx, * changing src/dst registers after merging the uniforms and varying vars. */ if (vertProg) { - shProg->VertexProgram - = vertex_program(_mesa_clone_program(ctx, &vertProg->Base)); + _mesa_reference_vertprog(ctx, &shProg->VertexProgram, + vertex_program(_mesa_clone_program(ctx, &vertProg->Base))); } else { - shProg->VertexProgram = NULL; + _mesa_reference_vertprog(ctx, &shProg->VertexProgram, NULL); } if (fragProg) { - shProg->FragmentProgram - = fragment_program(_mesa_clone_program(ctx, &fragProg->Base)); + _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, + fragment_program(_mesa_clone_program(ctx, &fragProg->Base))); } else { - shProg->FragmentProgram = NULL; + _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, NULL); } /* link varying vars */ diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 652500f52a..7745591afb 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -39,6 +39,7 @@ #include "main/imports.h" #include "main/mtypes.h" +#include "shader/program.h" #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" @@ -264,14 +265,16 @@ update_linkage( struct st_context *st ) */ assert(st->ctx->VertexProgram._Current); stvp = st_vertex_program(st->ctx->VertexProgram._Current); + assert(stvp->Base.Base.Target == GL_VERTEX_PROGRAM_ARB); assert(st->ctx->FragmentProgram._Current); stfp = st_fragment_program(st->ctx->FragmentProgram._Current); + assert(stfp->Base.Base.Target == GL_FRAGMENT_PROGRAM_ARB); xvp = find_translated_vp(st, stvp, stfp); - st->vp = stvp; - st->fp = stfp; + st_reference_vertprog(st, &st->vp, stvp); + st_reference_fragprog(st, &st->fp, stfp); cso_set_vertex_shader_handle(st->cso_context, stvp->driver_shader); cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader); diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index c900064f2b..8db55a179f 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -158,6 +158,9 @@ static void st_destroy_context_priv( struct st_context *st ) { uint i; + st_reference_fragprog(st, &st->fp, NULL); + st_reference_vertprog(st, &st->vp, NULL); + draw_destroy(st->draw); st_destroy_atoms( st ); st_destroy_draw( st ); diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index d8f26da2ee..bf07a50789 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -34,7 +34,8 @@ #ifndef ST_PROGRAM_H #define ST_PROGRAM_H -#include "mtypes.h" +#include "main/mtypes.h" +#include "shader/program.h" #include "pipe/p_shader_tokens.h" @@ -115,6 +116,27 @@ st_vertex_program( struct gl_vertex_program *vp ) } +static INLINE void +st_reference_vertprog(struct st_context *st, + struct st_vertex_program **ptr, + struct st_vertex_program *prog) +{ + _mesa_reference_program(st->ctx, + (struct gl_program **) ptr, + (struct gl_program *) prog); +} + +static INLINE void +st_reference_fragprog(struct st_context *st, + struct st_fragment_program **ptr, + struct st_fragment_program *prog) +{ + _mesa_reference_program(st->ctx, + (struct gl_program **) ptr, + (struct gl_program *) prog); +} + + extern void st_translate_fragment_program(struct st_context *st, struct st_fragment_program *fp, -- cgit v1.2.3 From 05370685fedab9608d6b5b5de7042dac4289e522 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 6 May 2008 23:08:02 -0600 Subject: mesa: free shader program data before deleting shader objects. Picked from master. Fixes mem corruption seen when glean/api2 test exits. --- src/mesa/main/context.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index b9053344b2..893c79f28c 100755 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -630,6 +630,21 @@ delete_arrayobj_cb(GLuint id, void *data, void *userData) _mesa_delete_array_object(ctx, arrayObj); } +/** + * Callback for freeing shader program data. Call it before delete_shader_cb + * to avoid memory access error. + */ +static void +free_shader_program_data_cb(GLuint id, void *data, void *userData) +{ + GLcontext *ctx = (GLcontext *) userData; + struct gl_shader_program *shProg = (struct gl_shader_program *) data; + + if (shProg->Type == GL_SHADER_PROGRAM_MESA) { + _mesa_free_shader_program_data(ctx, shProg); + } +} + /** * Callback for deleting shader and shader programs objects. * Called by _mesa_HashDeleteAll(). @@ -714,6 +729,7 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) _mesa_DeleteHashTable(ss->ArrayObjects); #if FEATURE_ARB_shader_objects + _mesa_HashWalk(ss->ShaderObjects, free_shader_program_data_cb, ctx); _mesa_HashDeleteAll(ss->ShaderObjects, delete_shader_cb, ctx); _mesa_DeleteHashTable(ss->ShaderObjects); #endif -- cgit v1.2.3 From fe06a47361b2db3dfde537aeb0c193495a16905e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 May 2008 14:14:48 +0900 Subject: gallium: New PIPE_TEXTURE_USAGE_PRIMARY flag for primary surfaces. --- src/gallium/include/pipe/p_state.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 571ea8c6e1..e7ee8c97ed 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -287,8 +287,9 @@ struct pipe_surface #define PIPE_TEXTURE_USAGE_RENDER_TARGET 0x1 #define PIPE_TEXTURE_USAGE_DISPLAY_TARGET 0x2 /* ie a backbuffer */ -#define PIPE_TEXTURE_USAGE_DEPTH_STENCIL 0x4 -#define PIPE_TEXTURE_USAGE_SAMPLER 0x8 +#define PIPE_TEXTURE_USAGE_PRIMARY 0x4 /* ie a frontbuffer */ +#define PIPE_TEXTURE_USAGE_DEPTH_STENCIL 0x8 +#define PIPE_TEXTURE_USAGE_SAMPLER 0x10 /** * Texture object. -- cgit v1.2.3 From c7ad942c54c3892a98d248a15af817f256260e75 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 May 2008 14:17:59 +0900 Subject: gallium: Propagate tex_usage flags down to winsys. --- src/gallium/drivers/i915simple/i915_texture.c | 3 ++- src/gallium/drivers/softpipe/sp_texture.c | 3 ++- src/gallium/include/pipe/p_winsys.h | 3 ++- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 3 ++- src/gallium/winsys/xlib/xm_winsys.c | 3 ++- src/gallium/winsys/xlib/xm_winsys_aub.c | 3 ++- 6 files changed, 12 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index f668e2e7d7..3e23e540f9 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -127,7 +127,8 @@ i915_displaytarget_layout(struct pipe_screen *screen, tex->base.width[0], tex->base.height[0], tex->base.format, - flags); + flags, + tex->base.tex_usage); /* Now extract the goodies: */ diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 599ff2ac45..1d7a1fffe4 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -115,7 +115,8 @@ softpipe_displaytarget_layout(struct pipe_screen *screen, spt->base.width[0], spt->base.height[0], spt->base.format, - flags); + flags, + spt->base.tex_usage); /* Now extract the goodies: */ diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h index 87a66b66d7..7ebc285192 100644 --- a/src/gallium/include/pipe/p_winsys.h +++ b/src/gallium/include/pipe/p_winsys.h @@ -86,7 +86,8 @@ struct pipe_winsys struct pipe_surface *surf, unsigned width, unsigned height, enum pipe_format format, - unsigned flags); + unsigned flags, + unsigned tex_usage); void (*surface_release)(struct pipe_winsys *ws, struct pipe_surface **s); diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 77dec9488d..d15143acfd 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -206,7 +206,8 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, struct pipe_surface *surf, unsigned width, unsigned height, enum pipe_format format, - unsigned flags) + unsigned flags, + unsigned tex_usage) { const unsigned alignment = 64; int ret; diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index a70752428a..b14758f333 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -489,7 +489,8 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys, struct pipe_surface *surf, unsigned width, unsigned height, enum pipe_format format, - unsigned flags) + unsigned flags, + unsigned tex_usage) { const unsigned alignment = 64; diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c index f42f7fcc5f..77376099f0 100644 --- a/src/gallium/winsys/xlib/xm_winsys_aub.c +++ b/src/gallium/winsys/xlib/xm_winsys_aub.c @@ -276,7 +276,8 @@ aub_i915_surface_alloc_storage(struct pipe_winsys *winsys, struct pipe_surface *surf, unsigned width, unsigned height, enum pipe_format format, - unsigned flags) + unsigned flags, + unsigned tex_usage) { const unsigned alignment = 64; -- cgit v1.2.3 From 4d1bf8a85eae730ca875194864277602f57582ea Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 May 2008 16:29:36 +0900 Subject: gallium: Output the total of leaked memory. --- src/gallium/auxiliary/util/p_debug_mem.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c index 3b5e4fbaee..78497c5f6a 100644 --- a/src/gallium/auxiliary/util/p_debug_mem.c +++ b/src/gallium/auxiliary/util/p_debug_mem.c @@ -211,6 +211,7 @@ debug_memory_begin(void) void debug_memory_end(unsigned long start_no) { + size_t total_size = 0; struct list_head *entry; entry = list.prev; @@ -220,9 +221,15 @@ debug_memory_end(unsigned long start_no) hdr = LIST_ENTRY(struct debug_memory_header, entry, head); ptr = data_from_header(hdr); if(start_no <= hdr->no && hdr->no < last_no || - last_no < start_no && (hdr->no < last_no || start_no <= hdr->no)) + last_no < start_no && (hdr->no < last_no || start_no <= hdr->no)) { debug_printf("%s:%u:%s: %u bytes at %p not freed\n", hdr->file, hdr->line, hdr->function, hdr->size, ptr); + total_size += hdr->size; + } + } + if(total_size) { + debug_printf("Total of %u KB of system memory apparently leaked\n", + (total_size + 1023)/1024); } } -- cgit v1.2.3 From 79931e38abc62286151121a3f59127e296144551 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 7 May 2008 08:46:28 +0100 Subject: Revert "gallium: move setup of dest_surface in do_copy_texsubimage()" This reverts commit f7dbd18371f9cb6686b6a97642b3ca5577e83472. Looks like an accidental revert of commit 650c57f19398800dfdcf357b6e9ec7b68bfa34f1. --- src/mesa/state_tracker/st_cb_texture.c | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 3723b91210..3206215b2e 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1137,7 +1137,6 @@ do_copy_texsubimage(GLcontext *ctx, struct st_renderbuffer *strb; struct pipe_context *pipe = ctx->st->pipe; struct pipe_screen *screen = pipe->screen; - struct pipe_surface *dest_surface; uint dest_format, src_format; uint do_flip = FALSE; GLboolean use_fallback = GL_TRUE; @@ -1168,13 +1167,7 @@ do_copy_texsubimage(GLcontext *ctx, src_format = strb->surface->format; dest_format = stImage->pt->format; - dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, - stImage->level, destZ, - PIPE_BUFFER_USAGE_CPU_WRITE); - - if (ctx->_ImageTransferState == 0x0 && - strb->surface->buffer && - dest_surface->buffer) { + if (ctx->_ImageTransferState == 0x0) { /* do blit-style copy */ /* XXX may need to invert image depending on window @@ -1197,6 +1190,14 @@ do_copy_texsubimage(GLcontext *ctx, x, y + height, dstx, dsty, width, height, GL_COPY); /* ? */ #else + struct pipe_surface *dest_surface; + + dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, + stImage->level, destZ, + PIPE_BUFFER_USAGE_GPU_WRITE); + + assert(strb->surface->buffer); + assert(dest_surface->buffer); if (src_format == dest_format) { pipe->surface_copy(pipe, @@ -1224,6 +1225,8 @@ do_copy_texsubimage(GLcontext *ctx, 0.0, PIPE_TEX_MIPFILTER_NEAREST); use_fallback = GL_FALSE; } + + pipe_surface_reference(&dest_surface, NULL); #endif } @@ -1234,8 +1237,6 @@ do_copy_texsubimage(GLcontext *ctx, srcX, srcY, width, height); } - pipe_surface_reference(&dest_surface, NULL); - if (level == texObj->BaseLevel && texObj->GenerateMipmap) { ctx->Driver.GenerateMipmap(ctx, target, texObj); } -- cgit v1.2.3 From 942b02956e7889aab977cf465fddb0055b758af2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 May 2008 19:39:34 +0900 Subject: gallium: Use the u_string.h functions. --- src/gallium/auxiliary/util/p_debug.c | 28 ++++++++++++---------------- 1 file changed, 12 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index f971ee03ba..0e233580d6 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -67,7 +67,7 @@ void _debug_vprintf(const char *format, va_list ap) static char buf[512 + 1] = {'\0'}; size_t len = strlen(buf); int ret = util_vsnprintf(buf + len, sizeof(buf) - len, format, ap); - if(ret > (int)(sizeof(buf) - len - 1) || strchr(buf + len, '\n')) { + if(ret > (int)(sizeof(buf) - len - 1) || util_strchr(buf + len, '\n')) { _EngDebugPrint("%s", buf); buf[0] = '\0'; } @@ -203,15 +203,15 @@ debug_get_bool_option(const char *name, boolean dfault) if(str == NULL) result = dfault; - else if(!strcmp(str, "n")) + else if(!util_strcmp(str, "n")) result = FALSE; - else if(!strcmp(str, "no")) + else if(!util_strcmp(str, "no")) result = FALSE; - else if(!strcmp(str, "0")) + else if(!util_strcmp(str, "0")) result = FALSE; - else if(!strcmp(str, "f")) + else if(!util_strcmp(str, "f")) result = FALSE; - else if(!strcmp(str, "false")) + else if(!util_strcmp(str, "false")) result = FALSE; else result = TRUE; @@ -244,7 +244,7 @@ debug_get_flags_option(const char *name, else { result = 0; while( flags->name ) { - if (!strcmp(str, "all") || strstr(str, flags->name )) + if (!util_strcmp(str, "all") || util_strstr(str, flags->name )) result |= flags->value; ++flags; } @@ -299,10 +299,10 @@ debug_dump_flags(const struct debug_named_value *names, while(names->name) { if((names->value & value) == names->value) { if (!first) - strncat(output, "|", sizeof(output)); + util_strncat(output, "|", sizeof(output)); else first = 0; - strncat(output, names->name, sizeof(output)); + util_strncat(output, names->name, sizeof(output)); value &= ~names->value; } ++names; @@ -310,12 +310,12 @@ debug_dump_flags(const struct debug_named_value *names, if (value) { if (!first) - strncat(output, "|", sizeof(output)); + util_strncat(output, "|", sizeof(output)); else first = 0; util_snprintf(rest, sizeof(rest), "0x%08lx", value); - strncat(output, rest, sizeof(output)); + util_strncat(output, rest, sizeof(output)); } if(first) @@ -325,10 +325,6 @@ debug_dump_flags(const struct debug_named_value *names, } - - - - char *pf_sprint_name( char *str, enum pipe_format format ) { strcpy( str, "PIPE_FORMAT_" ); @@ -469,4 +465,4 @@ void debug_dump_image(const char *prefix, EngUnmapFile(iFile); #endif } -#endif \ No newline at end of file +#endif -- cgit v1.2.3 From 55c13f5af7903c2a0f5a1839259a0a0cc15d6e5e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 May 2008 20:27:42 +0900 Subject: gallium: Implement pf_sprint_name as a simple static table. --- src/gallium/auxiliary/util/p_debug.c | 180 +++++++++++++++++++---------------- 1 file changed, 97 insertions(+), 83 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 0e233580d6..3f3cf620fe 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -327,96 +327,110 @@ debug_dump_flags(const struct debug_named_value *names, char *pf_sprint_name( char *str, enum pipe_format format ) { - strcpy( str, "PIPE_FORMAT_" ); - switch (pf_layout( format )) { - case PIPE_FORMAT_LAYOUT_RGBAZS: - { - pipe_format_rgbazs_t rgbazs = (pipe_format_rgbazs_t) format; - uint i; - uint scale = 1 << (pf_exp8( rgbazs ) * 3); - - for (i = 0; i < 4; i++) { - uint size = pf_size_xyzw( rgbazs, i ); - - if (size == 0) { - break; - } - switch (pf_swizzle_xyzw( rgbazs, i )) { - case PIPE_FORMAT_COMP_R: - strcat( str, "R" ); - break; - case PIPE_FORMAT_COMP_G: - strcat( str, "G" ); - break; - case PIPE_FORMAT_COMP_B: - strcat( str, "B" ); - break; - case PIPE_FORMAT_COMP_A: - strcat( str, "A" ); - break; - case PIPE_FORMAT_COMP_0: - strcat( str, "0" ); - break; - case PIPE_FORMAT_COMP_1: - strcat( str, "1" ); - break; - case PIPE_FORMAT_COMP_Z: - strcat( str, "Z" ); - break; - case PIPE_FORMAT_COMP_S: - strcat( str, "S" ); - break; - } - util_snprintf( &str[strlen( str )], 32, "%u", size * scale ); - } - if (i != 0) { - strcat( str, "_" ); - } - switch (pf_type( rgbazs )) { - case PIPE_FORMAT_TYPE_UNKNOWN: - strcat( str, "NONE" ); - break; - case PIPE_FORMAT_TYPE_FLOAT: - strcat( str, "FLOAT" ); - break; - case PIPE_FORMAT_TYPE_UNORM: - strcat( str, "UNORM" ); - break; - case PIPE_FORMAT_TYPE_SNORM: - strcat( str, "SNORM" ); - break; - case PIPE_FORMAT_TYPE_USCALED: - strcat( str, "USCALED" ); - break; - case PIPE_FORMAT_TYPE_SSCALED: - strcat( str, "SSCALED" ); - break; - } - } - break; - case PIPE_FORMAT_LAYOUT_YCBCR: - { - pipe_format_ycbcr_t ycbcr = (pipe_format_ycbcr_t) format; - - strcat( str, "YCBCR" ); - if (pf_rev( ycbcr )) { - strcat( str, "_REV" ); - } - } - break; - } + strcpy( str, debug_dump_enum(pipe_format_names, fmt) ); return str; } #ifdef DEBUG +static const struct debug_named_value pipe_format_names[] = { + DEBUG_NAMED_VALUE(PIPE_FORMAT_NONE), + DEBUG_NAMED_VALUE(PIPE_FORMAT_A8R8G8B8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_X8R8G8B8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_B8G8R8A8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_B8G8R8X8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_A1R5G5B5_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_A4R4G4B4_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R5G6B5_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_L8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_A8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_I8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_A8L8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_YCBCR), + DEBUG_NAMED_VALUE(PIPE_FORMAT_YCBCR_REV), + DEBUG_NAMED_VALUE(PIPE_FORMAT_Z16_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_Z32_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_Z32_FLOAT), + DEBUG_NAMED_VALUE(PIPE_FORMAT_S8Z24_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_Z24S8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_X8Z24_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_Z24X8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_S8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R64_FLOAT), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R64G64_FLOAT), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R64G64B64_FLOAT), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R64G64B64A64_FLOAT), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_FLOAT), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_FLOAT), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_FLOAT), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_FLOAT), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R32G32B32A32_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R16G16B16A16_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_USCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SSCALED), + DEBUG_NAMED_VALUE(PIPE_FORMAT_L8_SRGB), + DEBUG_NAMED_VALUE(PIPE_FORMAT_A8_L8_SRGB), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SRGB), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SRGB), + DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SRGB), + DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_RGB), + DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_RGBA), + DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT3_RGBA), + DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT5_RGBA), + DEBUG_NAMED_VALUE_END +}; + void debug_print_format(const char *msg, unsigned fmt ) { - char fmtstr[80]; - - pf_sprint_name(fmtstr, (enum pipe_format)fmt); - - debug_printf("%s: %s\n", msg, fmtstr); + debug_printf("%s: %s\n", msg, debug_dump_enum(pipe_format_names, fmt)); } #endif -- cgit v1.2.3 From 5efd2d59eb19cc44624d3f842d3d1d291876ae7d Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 7 May 2008 14:24:14 +0200 Subject: gallium: Fix compilation errors. --- src/gallium/auxiliary/util/p_debug.c | 13 ++++++------- 1 file changed, 6 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 3f3cf620fe..0f3a99cf83 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -325,13 +325,6 @@ debug_dump_flags(const struct debug_named_value *names, } -char *pf_sprint_name( char *str, enum pipe_format format ) -{ - strcpy( str, debug_dump_enum(pipe_format_names, fmt) ); - return str; -} - - #ifdef DEBUG static const struct debug_named_value pipe_format_names[] = { DEBUG_NAMED_VALUE(PIPE_FORMAT_NONE), @@ -432,6 +425,12 @@ void debug_print_format(const char *msg, unsigned fmt ) { debug_printf("%s: %s\n", msg, debug_dump_enum(pipe_format_names, fmt)); } + +char *pf_sprint_name( char *str, enum pipe_format format ) +{ + strcpy( str, debug_dump_enum(pipe_format_names, format) ); + return str; +} #endif -- cgit v1.2.3 From 33cda1e5e57838845ec62714677832f7cdabc5dc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 May 2008 21:37:32 +0900 Subject: gallium: Fix release build. pf_sprint_name might be used there too. --- src/gallium/auxiliary/util/p_debug.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 0f3a99cf83..951dd5a2d4 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -325,8 +325,8 @@ debug_dump_flags(const struct debug_named_value *names, } -#ifdef DEBUG static const struct debug_named_value pipe_format_names[] = { +#ifdef DEBUG DEBUG_NAMED_VALUE(PIPE_FORMAT_NONE), DEBUG_NAMED_VALUE(PIPE_FORMAT_A8R8G8B8_UNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_X8R8G8B8_UNORM), @@ -418,20 +418,22 @@ static const struct debug_named_value pipe_format_names[] = { DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_RGBA), DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT3_RGBA), DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT5_RGBA), +#endif DEBUG_NAMED_VALUE_END }; +#ifdef DEBUG void debug_print_format(const char *msg, unsigned fmt ) { debug_printf("%s: %s\n", msg, debug_dump_enum(pipe_format_names, fmt)); } +#endif char *pf_sprint_name( char *str, enum pipe_format format ) { strcpy( str, debug_dump_enum(pipe_format_names, format) ); return str; } -#endif #ifdef DEBUG -- cgit v1.2.3 From 10b7192747087ec25f97cdfcfc062654a2d8fe6d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 7 May 2008 22:01:27 +0900 Subject: gallium: Implement util_pack_color for A8, L8, and I8 formats. --- src/gallium/auxiliary/util/u_pack_color.h | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 0b917c005f..655e2c8259 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -101,6 +101,19 @@ util_pack_color_ub(ubyte r, ubyte g, ubyte b, ubyte a, *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4); } return; + case PIPE_FORMAT_A8_UNORM: + { + ubyte *d = (ubyte *) dest; + *d = a; + } + return; + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_I8_UNORM: + { + ubyte *d = (ubyte *) dest; + *d = r; + } + return; case PIPE_FORMAT_R32G32B32A32_FLOAT: { float *d = (float *) dest; @@ -198,6 +211,19 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) *d = ((a & 0xf0) << 8) | ((r & 0xf0) << 4) | ((g & 0xf0) << 0) | (b >> 4); } return; + case PIPE_FORMAT_A8_UNORM: + { + ubyte *d = (ubyte *) dest; + *d = a; + } + return; + case PIPE_FORMAT_L8_UNORM: + case PIPE_FORMAT_I8_UNORM: + { + ubyte *d = (ubyte *) dest; + *d = r; + } + return; case PIPE_FORMAT_R32G32B32A32_FLOAT: { float *d = (float *) dest; -- cgit v1.2.3 From a56a59ce74b7f18f25a13992d2a2c1ae7cf973ce Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 7 May 2008 08:55:33 -0600 Subject: gallium: implement full reference counting for vertex/fragment programs Use _mesa_reference_vert/fragprog() wherever we assign program pointers. Fixes a memory corruption bug found with glean/api2 test. Another memory bug involving shaders yet to be fixed... Picked from gallium-0.1 --- src/mesa/main/context.c | 26 ++++---- src/mesa/main/mtypes.h | 4 +- src/mesa/main/state.c | 32 ++++++---- src/mesa/shader/prog_cache.c | 5 +- src/mesa/shader/program.c | 103 +++++++++++++++++++++----------- src/mesa/shader/program.h | 22 +++++++ src/mesa/shader/shader_api.c | 6 +- src/mesa/shader/slang/slang_link.c | 12 ++-- src/mesa/state_tracker/st_atom_shader.c | 7 ++- src/mesa/state_tracker/st_context.c | 3 + src/mesa/state_tracker/st_program.h | 24 +++++++- 11 files changed, 170 insertions(+), 74 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 7b8d934170..b9053344b2 100755 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -150,8 +150,6 @@ int MESA_DEBUG_FLAGS = 0; /* ubyte -> float conversion */ GLfloat _mesa_ubyte_to_float_color_tab[256]; -static void -free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ); /** @@ -423,12 +421,14 @@ alloc_shared_state( GLcontext *ctx ) #endif #if FEATURE_ARB_vertex_program - ss->DefaultVertexProgram = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); + ss->DefaultVertexProgram = (struct gl_vertex_program *) + ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); if (!ss->DefaultVertexProgram) goto cleanup; #endif #if FEATURE_ARB_fragment_program - ss->DefaultFragmentProgram = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); + ss->DefaultFragmentProgram = (struct gl_fragment_program *) + ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!ss->DefaultFragmentProgram) goto cleanup; #endif @@ -513,12 +513,10 @@ alloc_shared_state( GLcontext *ctx ) _mesa_DeleteHashTable(ss->Programs); #endif #if FEATURE_ARB_vertex_program - if (ss->DefaultVertexProgram) - ctx->Driver.DeleteProgram(ctx, ss->DefaultVertexProgram); + _mesa_reference_vertprog(ctx, &ss->DefaultVertexProgram, NULL); #endif #if FEATURE_ARB_fragment_program - if (ss->DefaultFragmentProgram) - ctx->Driver.DeleteProgram(ctx, ss->DefaultFragmentProgram); + _mesa_reference_fragprog(ctx, &ss->DefaultFragmentProgram, NULL); #endif #if FEATURE_ATI_fragment_shader if (ss->DefaultFragmentShader) @@ -695,10 +693,10 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) _mesa_DeleteHashTable(ss->Programs); #endif #if FEATURE_ARB_vertex_program - ctx->Driver.DeleteProgram(ctx, ss->DefaultVertexProgram); + _mesa_reference_vertprog(ctx, &ss->DefaultVertexProgram, NULL); #endif #if FEATURE_ARB_fragment_program - ctx->Driver.DeleteProgram(ctx, ss->DefaultFragmentProgram); + _mesa_reference_fragprog(ctx, &ss->DefaultFragmentProgram, NULL); #endif #if FEATURE_ATI_fragment_shader @@ -1190,6 +1188,14 @@ _mesa_free_context_data( GLcontext *ctx ) _mesa_unreference_framebuffer(&ctx->ReadBuffer); } + _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, NULL); + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, NULL); + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._TnlProgram, NULL); + + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, NULL); + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL); + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, NULL); + _mesa_free_lighting_data( ctx ); _mesa_free_eval_data( ctx ); _mesa_free_texture_data( ctx ); diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 50b22d25bf..463142fe39 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2203,10 +2203,10 @@ struct gl_shared_state /*@{*/ struct _mesa_HashTable *Programs; /**< All vertex/fragment programs */ #if FEATURE_ARB_vertex_program - struct gl_program *DefaultVertexProgram; + struct gl_vertex_program *DefaultVertexProgram; #endif #if FEATURE_ARB_fragment_program - struct gl_program *DefaultFragmentProgram; + struct gl_fragment_program *DefaultFragmentProgram; #endif /*@}*/ diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 0b1c56fdd5..90379a1772 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -982,16 +982,20 @@ update_program(GLcontext *ctx) #endif if (shProg && shProg->LinkStatus && shProg->FragmentProgram) { /* user-defined fragment shader */ - ctx->FragmentProgram._Current = shProg->FragmentProgram; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, + shProg->FragmentProgram); } else if (ctx->FragmentProgram._Enabled) { /* use user-defined fragment program */ - ctx->FragmentProgram._Current = ctx->FragmentProgram.Current; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, + ctx->FragmentProgram.Current); } else if (ctx->FragmentProgram._MaintainTexEnvProgram) { /* fragment program generated from fixed-function state */ - ctx->FragmentProgram._Current = _mesa_get_fixed_func_fragment_program(ctx); - ctx->FragmentProgram._TexEnvProgram = ctx->FragmentProgram._Current; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, + _mesa_get_fixed_func_fragment_program(ctx)); + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, + ctx->FragmentProgram._Current); /* XXX get rid of this confusing stuff someday? */ ctx->FragmentProgram._Active = ctx->FragmentProgram._Enabled; @@ -1000,7 +1004,7 @@ update_program(GLcontext *ctx) } else { /* no fragment program */ - ctx->FragmentProgram._Current = NULL; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._Current, NULL); } if (ctx->FragmentProgram._Current != prevFP && ctx->Driver.BindProgram) { @@ -1013,29 +1017,33 @@ update_program(GLcontext *ctx) **/ #if 1 /* XXX get rid of this someday? */ - ctx->VertexProgram._TnlProgram = NULL; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._TnlProgram, NULL); #endif if (shProg && shProg->LinkStatus && shProg->VertexProgram) { /* user-defined vertex shader */ - ctx->VertexProgram._Current = shProg->VertexProgram; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, + shProg->VertexProgram); } else if (ctx->VertexProgram._Enabled) { /* use user-defined vertex program */ - ctx->VertexProgram._Current = ctx->VertexProgram.Current; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, + ctx->VertexProgram.Current); } else if (ctx->VertexProgram._MaintainTnlProgram) { /* vertex program generated from fixed-function state */ - ctx->VertexProgram._Current = _mesa_get_fixed_func_vertex_program(ctx); - ctx->VertexProgram._TnlProgram = ctx->VertexProgram._Current; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, + _mesa_get_fixed_func_vertex_program(ctx)); + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._TnlProgram, + ctx->VertexProgram._Current); } else { /* no vertex program / used fixed-function code */ - ctx->VertexProgram._Current = NULL; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram._Current, NULL); } if (ctx->VertexProgram._Current != prevVP && ctx->Driver.BindProgram) { ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB, - (struct gl_program *) ctx->VertexProgram._Current); + (struct gl_program *) ctx->VertexProgram._Current); } } diff --git a/src/mesa/shader/prog_cache.c b/src/mesa/shader/prog_cache.c index dd0241ef24..36a25377c5 100644 --- a/src/mesa/shader/prog_cache.c +++ b/src/mesa/shader/prog_cache.c @@ -30,6 +30,7 @@ #include "main/mtypes.h" #include "main/imports.h" #include "shader/prog_cache.h" +#include "shader/program.h" struct cache_item @@ -109,7 +110,7 @@ clear_cache(GLcontext *ctx, struct gl_program_cache *cache) for (c = cache->items[i]; c; c = next) { next = c->next; _mesa_free(c->key); - ctx->Driver.DeleteProgram(ctx, c->program); + _mesa_reference_program(ctx, &c->program, NULL); _mesa_free(c); } cache->items[i] = NULL; @@ -177,7 +178,7 @@ _mesa_program_cache_insert(GLcontext *ctx, c->key = _mesa_malloc(keysize); memcpy(c->key, key, keysize); - c->program = program; + c->program = program; /* no refcount change */ if (cache->n_items > cache->size * 1.5) { if (cache->size < 1000) diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 0ed7f833d2..9a23c5d7d3 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -60,9 +60,9 @@ _mesa_init_program(GLcontext *ctx) ctx->VertexProgram.Enabled = GL_FALSE; ctx->VertexProgram.PointSizeEnabled = GL_FALSE; ctx->VertexProgram.TwoSideEnabled = GL_FALSE; - ctx->VertexProgram.Current = (struct gl_vertex_program *) ctx->Shared->DefaultVertexProgram; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, + ctx->Shared->DefaultVertexProgram); assert(ctx->VertexProgram.Current); - ctx->VertexProgram.Current->Base.RefCount++; for (i = 0; i < MAX_NV_VERTEX_PROGRAM_PARAMS / 4; i++) { ctx->VertexProgram.TrackMatrix[i] = GL_NONE; ctx->VertexProgram.TrackMatrixTransform[i] = GL_IDENTITY_NV; @@ -72,9 +72,9 @@ _mesa_init_program(GLcontext *ctx) #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program ctx->FragmentProgram.Enabled = GL_FALSE; - ctx->FragmentProgram.Current = (struct gl_fragment_program *) ctx->Shared->DefaultFragmentProgram; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, + ctx->Shared->DefaultFragmentProgram); assert(ctx->FragmentProgram.Current); - ctx->FragmentProgram.Current->Base.RefCount++; ctx->FragmentProgram.Cache = _mesa_new_program_cache(); #endif @@ -96,19 +96,11 @@ void _mesa_free_program_data(GLcontext *ctx) { #if FEATURE_NV_vertex_program || FEATURE_ARB_vertex_program - if (ctx->VertexProgram.Current) { - ctx->VertexProgram.Current->Base.RefCount--; - if (ctx->VertexProgram.Current->Base.RefCount <= 0) - ctx->Driver.DeleteProgram(ctx, &(ctx->VertexProgram.Current->Base)); - } + _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, NULL); _mesa_delete_program_cache(ctx, ctx->VertexProgram.Cache); #endif #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program - if (ctx->FragmentProgram.Current) { - ctx->FragmentProgram.Current->Base.RefCount--; - if (ctx->FragmentProgram.Current->Base.RefCount <= 0) - ctx->Driver.DeleteProgram(ctx, &(ctx->FragmentProgram.Current->Base)); - } + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, NULL); _mesa_delete_program_cache(ctx, ctx->FragmentProgram.Cache); #endif /* XXX probably move this stuff */ @@ -325,6 +317,59 @@ _mesa_lookup_program(GLcontext *ctx, GLuint id) } +/** + * Reference counting for vertex/fragment programs + */ +void +_mesa_reference_program(GLcontext *ctx, + struct gl_program **ptr, + struct gl_program *prog) +{ + assert(ptr); + if (*ptr && prog) { + /* sanity check */ + ASSERT((*ptr)->Target == prog->Target); + } + if (*ptr == prog) { + return; /* no change */ + } + if (*ptr) { + GLboolean deleteFlag; + + /*_glthread_LOCK_MUTEX((*ptr)->Mutex);*/ +#if 0 + printf("Program %p %u 0x%x Refcount-- to %d\n", + *ptr, (*ptr)->Id, (*ptr)->Target, (*ptr)->RefCount - 1); +#endif + ASSERT((*ptr)->RefCount > 0); + (*ptr)->RefCount--; + + deleteFlag = ((*ptr)->RefCount == 0); + /*_glthread_UNLOCK_MUTEX((*ptr)->Mutex);*/ + + if (deleteFlag) { + ASSERT(ctx); + ctx->Driver.DeleteProgram(ctx, *ptr); + } + + *ptr = NULL; + } + + assert(!*ptr); + if (prog) { + /*_glthread_LOCK_MUTEX(prog->Mutex);*/ + prog->RefCount++; +#if 0 + printf("Program %p %u 0x%x Refcount++ to %d\n", + prog, prog->Id, prog->Target, prog->RefCount); +#endif + /*_glthread_UNLOCK_MUTEX(prog->Mutex);*/ + } + + *ptr = prog; +} + + /** * Return a copy of a program. * XXX Problem here if the program object is actually OO-derivation @@ -340,8 +385,9 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog) return NULL; assert(clone->Target == prog->Target); + assert(clone->RefCount == 1); + clone->String = (GLubyte *) _mesa_strdup((char *) prog->String); - clone->RefCount = 1; clone->Format = prog->Format; clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions); if (!clone->Instructions) { @@ -704,9 +750,9 @@ _mesa_BindProgram(GLenum target, GLuint id) /* Bind a default program */ newProg = NULL; if (target == GL_VERTEX_PROGRAM_ARB) /* == GL_VERTEX_PROGRAM_NV */ - newProg = ctx->Shared->DefaultVertexProgram; + newProg = &ctx->Shared->DefaultVertexProgram->Base; else - newProg = ctx->Shared->DefaultFragmentProgram; + newProg = &ctx->Shared->DefaultFragmentProgram->Base; } else { /* Bind a user program */ @@ -734,26 +780,16 @@ _mesa_BindProgram(GLenum target, GLuint id) return; } - /* unbind/delete oldProg */ - if (curProg->Id != 0) { - /* decrement refcount on previously bound fragment program */ - curProg->RefCount--; - /* and delete if refcount goes below one */ - if (curProg->RefCount <= 0) { - /* the program ID was already removed from the hash table */ - ctx->Driver.DeleteProgram(ctx, curProg); - } - } - /* bind newProg */ if (target == GL_VERTEX_PROGRAM_ARB) { /* == GL_VERTEX_PROGRAM_NV */ - ctx->VertexProgram.Current = (struct gl_vertex_program *) newProg; + _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, + (struct gl_vertex_program *) newProg); } else if (target == GL_FRAGMENT_PROGRAM_NV || target == GL_FRAGMENT_PROGRAM_ARB) { - ctx->FragmentProgram.Current = (struct gl_fragment_program *) newProg; + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, + (struct gl_fragment_program *) newProg); } - newProg->RefCount++; /* Never null pointers */ ASSERT(ctx->VertexProgram.Current); @@ -811,10 +847,7 @@ _mesa_DeletePrograms(GLsizei n, const GLuint *ids) } /* The ID is immediately available for re-use now */ _mesa_HashRemove(ctx->Shared->Programs, ids[i]); - prog->RefCount--; - if (prog->RefCount <= 0) { - ctx->Driver.DeleteProgram(ctx, prog); - } + _mesa_reference_program(ctx, &prog, NULL); } } } diff --git a/src/mesa/shader/program.h b/src/mesa/shader/program.h index 414a57d39c..08fe576afc 100644 --- a/src/mesa/shader/program.h +++ b/src/mesa/shader/program.h @@ -83,6 +83,28 @@ _mesa_delete_program(GLcontext *ctx, struct gl_program *prog); extern struct gl_program * _mesa_lookup_program(GLcontext *ctx, GLuint id); +extern void +_mesa_reference_program(GLcontext *ctx, + struct gl_program **ptr, + struct gl_program *prog); + +static INLINE void +_mesa_reference_vertprog(GLcontext *ctx, + struct gl_vertex_program **ptr, + struct gl_vertex_program *prog) +{ + _mesa_reference_program(ctx, (struct gl_program **) ptr, + (struct gl_program *) prog); +} + +static INLINE void +_mesa_reference_fragprog(GLcontext *ctx, + struct gl_fragment_program **ptr, + struct gl_fragment_program *prog) +{ + _mesa_reference_program(ctx, (struct gl_program **) ptr, + (struct gl_program *) prog); +} extern struct gl_program * _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog); diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 9c419c9903..f12fa28d97 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -80,8 +80,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx, * original/unlinked program. */ shProg->VertexProgram->Base.Parameters = NULL; - ctx->Driver.DeleteProgram(ctx, &shProg->VertexProgram->Base); - shProg->VertexProgram = NULL; + _mesa_reference_vertprog(ctx, &shProg->VertexProgram, NULL); } if (shProg->FragmentProgram) { @@ -89,8 +88,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx, * original/unlinked program. */ shProg->FragmentProgram->Base.Parameters = NULL; - ctx->Driver.DeleteProgram(ctx, &shProg->FragmentProgram->Base); - shProg->FragmentProgram = NULL; + _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, NULL); } if (shProg->Uniforms) { diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index addff20421..ae581553dc 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -410,19 +410,19 @@ _slang_link(GLcontext *ctx, * changing src/dst registers after merging the uniforms and varying vars. */ if (vertProg) { - shProg->VertexProgram - = vertex_program(_mesa_clone_program(ctx, &vertProg->Base)); + _mesa_reference_vertprog(ctx, &shProg->VertexProgram, + vertex_program(_mesa_clone_program(ctx, &vertProg->Base))); } else { - shProg->VertexProgram = NULL; + _mesa_reference_vertprog(ctx, &shProg->VertexProgram, NULL); } if (fragProg) { - shProg->FragmentProgram - = fragment_program(_mesa_clone_program(ctx, &fragProg->Base)); + _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, + fragment_program(_mesa_clone_program(ctx, &fragProg->Base))); } else { - shProg->FragmentProgram = NULL; + _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, NULL); } /* link varying vars */ diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 652500f52a..7745591afb 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -39,6 +39,7 @@ #include "main/imports.h" #include "main/mtypes.h" +#include "shader/program.h" #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" @@ -264,14 +265,16 @@ update_linkage( struct st_context *st ) */ assert(st->ctx->VertexProgram._Current); stvp = st_vertex_program(st->ctx->VertexProgram._Current); + assert(stvp->Base.Base.Target == GL_VERTEX_PROGRAM_ARB); assert(st->ctx->FragmentProgram._Current); stfp = st_fragment_program(st->ctx->FragmentProgram._Current); + assert(stfp->Base.Base.Target == GL_FRAGMENT_PROGRAM_ARB); xvp = find_translated_vp(st, stvp, stfp); - st->vp = stvp; - st->fp = stfp; + st_reference_vertprog(st, &st->vp, stvp); + st_reference_fragprog(st, &st->fp, stfp); cso_set_vertex_shader_handle(st->cso_context, stvp->driver_shader); cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader); diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index c900064f2b..8db55a179f 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -158,6 +158,9 @@ static void st_destroy_context_priv( struct st_context *st ) { uint i; + st_reference_fragprog(st, &st->fp, NULL); + st_reference_vertprog(st, &st->vp, NULL); + draw_destroy(st->draw); st_destroy_atoms( st ); st_destroy_draw( st ); diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index d8f26da2ee..bf07a50789 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -34,7 +34,8 @@ #ifndef ST_PROGRAM_H #define ST_PROGRAM_H -#include "mtypes.h" +#include "main/mtypes.h" +#include "shader/program.h" #include "pipe/p_shader_tokens.h" @@ -115,6 +116,27 @@ st_vertex_program( struct gl_vertex_program *vp ) } +static INLINE void +st_reference_vertprog(struct st_context *st, + struct st_vertex_program **ptr, + struct st_vertex_program *prog) +{ + _mesa_reference_program(st->ctx, + (struct gl_program **) ptr, + (struct gl_program *) prog); +} + +static INLINE void +st_reference_fragprog(struct st_context *st, + struct st_fragment_program **ptr, + struct st_fragment_program *prog) +{ + _mesa_reference_program(st->ctx, + (struct gl_program **) ptr, + (struct gl_program *) prog); +} + + extern void st_translate_fragment_program(struct st_context *st, struct st_fragment_program *fp, -- cgit v1.2.3 From 3e1a4c286936abdb4ce1b62a9ebdd93db1777aff Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 7 May 2008 08:56:08 -0600 Subject: mesa: free shader program data before deleting shader objects. Picked from master. Fixes mem corruption seen when glean/api2 test exits. --- src/mesa/main/context.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index b9053344b2..893c79f28c 100755 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -630,6 +630,21 @@ delete_arrayobj_cb(GLuint id, void *data, void *userData) _mesa_delete_array_object(ctx, arrayObj); } +/** + * Callback for freeing shader program data. Call it before delete_shader_cb + * to avoid memory access error. + */ +static void +free_shader_program_data_cb(GLuint id, void *data, void *userData) +{ + GLcontext *ctx = (GLcontext *) userData; + struct gl_shader_program *shProg = (struct gl_shader_program *) data; + + if (shProg->Type == GL_SHADER_PROGRAM_MESA) { + _mesa_free_shader_program_data(ctx, shProg); + } +} + /** * Callback for deleting shader and shader programs objects. * Called by _mesa_HashDeleteAll(). @@ -714,6 +729,7 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) _mesa_DeleteHashTable(ss->ArrayObjects); #if FEATURE_ARB_shader_objects + _mesa_HashWalk(ss->ShaderObjects, free_shader_program_data_cb, ctx); _mesa_HashDeleteAll(ss->ShaderObjects, delete_shader_cb, ctx); _mesa_DeleteHashTable(ss->ShaderObjects); #endif -- cgit v1.2.3 From 1a82d9648b3db780e58e4966924157542d148c58 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 7 May 2008 16:44:33 -0600 Subject: gallium: fix some render to texture bugs Before, we were sometimes rendering into a stale texture because st_finalize_texture() would discard the old texture and create a new one. Moved st_update_framebuffer atom after texture validation so that we can create a new renderbuffer surface if the texture changes. Also, split texture validation into two parts: finalize_textures and update_textures. Do finalize_textures first to avoid getting into the situtation where we're doing a pipe->surface_copy() mid-way through state validation. Some debug code still in place, but disabled... --- src/mesa/state_tracker/st_atom.c | 3 +- src/mesa/state_tracker/st_atom.h | 1 + src/mesa/state_tracker/st_atom_framebuffer.c | 60 ++++++++++++++++++++- src/mesa/state_tracker/st_atom_shader.c | 26 ++++++++- src/mesa/state_tracker/st_atom_texture.c | 80 +++++++++++++++++----------- src/mesa/state_tracker/st_cb_fbo.c | 34 +++++++++--- src/mesa/state_tracker/st_cb_fbo.h | 3 ++ src/mesa/state_tracker/st_cb_readpixels.c | 2 + src/mesa/state_tracker/st_cb_texture.c | 1 + src/mesa/state_tracker/st_context.h | 3 ++ src/mesa/state_tracker/st_texture.c | 16 ++++++ 11 files changed, 189 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index 18063adc79..ecfd117918 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -45,10 +45,10 @@ */ static const struct st_tracked_state *atoms[] = { - &st_update_framebuffer, &st_update_depth_stencil_alpha, &st_update_clip, + &st_finalize_textures, &st_update_shader, &st_update_rasterizer, @@ -58,6 +58,7 @@ static const struct st_tracked_state *atoms[] = &st_update_blend, &st_update_sampler, &st_update_texture, + &st_update_framebuffer, &st_update_vs_constants, &st_update_fs_constants, &st_update_pixel_transfer diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h index c6c6eba812..c7cffd85c8 100644 --- a/src/mesa/state_tracker/st_atom.h +++ b/src/mesa/state_tracker/st_atom.h @@ -55,6 +55,7 @@ extern const struct st_tracked_state st_update_scissor; extern const struct st_tracked_state st_update_blend; extern const struct st_tracked_state st_update_sampler; extern const struct st_tracked_state st_update_texture; +extern const struct st_tracked_state st_finalize_textures; extern const struct st_tracked_state st_update_fs_constants; extern const struct st_tracked_state st_update_vs_constants; extern const struct st_tracked_state st_update_pixel_transfer; diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index 0a6974d8a7..c9a30e44b2 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -34,13 +34,60 @@ #include "st_context.h" #include "st_atom.h" #include "st_cb_fbo.h" +#include "st_texture.h" #include "pipe/p_context.h" +#include "pipe/p_inlines.h" #include "cso_cache/cso_context.h" + +/** + * When doing GL render to texture, we have to be sure that finalize_texture() + * didn't yank out the pipe_texture that we earlier created a surface for. + * Check for that here and create a new surface if needed. + */ +static void +update_renderbuffer_surface(struct st_context *st, + struct st_renderbuffer *strb) +{ + struct pipe_screen *screen = st->pipe->screen; + struct pipe_texture *texture = strb->rtt->pt; + int rtt_width = strb->Base.Width; + int rtt_height = strb->Base.Height; + + if (!strb->surface || + strb->surface->texture != texture || + strb->surface->width != rtt_width || + strb->surface->height != rtt_height) { + int level; + /* find matching mipmap level size */ + for (level = 0; level <= texture->last_level; level++) { + if (texture->width[level] == rtt_width && + texture->height[level] == rtt_height) { + + pipe_surface_reference(&strb->surface, NULL); + + strb->surface = screen->get_tex_surface(screen, + texture, + strb->rtt_face, + level, + strb->rtt_slice, + PIPE_BUFFER_USAGE_GPU_READ | + PIPE_BUFFER_USAGE_GPU_WRITE); +#if 0 + printf("-- alloc new surface %d x %d into tex %p\n", + strb->surface->width, strb->surface->height, + texture); +#endif + break; + } + } + } +} + + /** * Update framebuffer state (color, depth, stencil, etc. buffers) - * XXX someday: separate draw/read buffers. */ static void update_framebuffer_state( struct st_context *st ) @@ -55,6 +102,8 @@ update_framebuffer_state( struct st_context *st ) framebuffer->width = fb->Width; framebuffer->height = fb->Height; + /*printf("------ fb size %d x %d\n", fb->Width, fb->Height);*/ + /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state * to determine which surfaces to draw to */ @@ -62,6 +111,13 @@ update_framebuffer_state( struct st_context *st ) for (j = 0; j < MAX_DRAW_BUFFERS; j++) { for (i = 0; i < fb->_NumColorDrawBuffers[j]; i++) { strb = st_renderbuffer(fb->_ColorDrawBuffers[j][i]); + + /*printf("--------- framebuffer surface rtt %p\n", strb->rtt);*/ + if (strb->rtt) { + /* rendering to a GL texture, may have to update surface */ + update_renderbuffer_surface(st, strb); + } + assert(strb->surface); framebuffer->cbufs[framebuffer->num_cbufs] = strb->surface; framebuffer->num_cbufs++; @@ -99,7 +155,7 @@ const struct st_tracked_state st_update_framebuffer = { "st_update_framebuffer", /* name */ { /* dirty */ _NEW_BUFFERS, /* mesa */ - 0, /* st */ + ST_NEW_FRAMEBUFFER, /* st */ }, update_framebuffer_state /* update */ }; diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 7745591afb..8839ab380f 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -44,6 +44,8 @@ #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" +#include "util/u_simple_shaders.h" + #include "cso_cache/cso_context.h" #include "st_context.h" @@ -252,6 +254,20 @@ st_free_translated_vertex_programs(struct st_context *st, } +static void * +get_passthrough_fs(struct st_context *st) +{ + struct pipe_shader_state shader; + + if (!st->passthrough_fs) { + st->passthrough_fs = + util_make_fragment_passthrough_shader(st->pipe, &shader); + free((void *) shader.tokens); + } + + return st->passthrough_fs; +} + static void update_linkage( struct st_context *st ) @@ -277,7 +293,15 @@ update_linkage( struct st_context *st ) st_reference_fragprog(st, &st->fp, stfp); cso_set_vertex_shader_handle(st->cso_context, stvp->driver_shader); - cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader); + + if (st->missing_textures) { + /* use a pass-through frag shader that uses no textures */ + void *fs = get_passthrough_fs(st); + cso_set_fragment_shader_handle(st->cso_context, fs); + } + else { + cso_set_fragment_shader_handle(st->cso_context, stfp->driver_shader); + } st->vertex_result_to_slot = xvp->output_to_slot; } diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 767654f3d0..1ec671ed48 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -39,34 +39,13 @@ #include "pipe/p_context.h" #include "pipe/p_inlines.h" #include "cso_cache/cso_context.h" -#include "util/u_simple_shaders.h" -static void * -get_passthrough_fs(struct st_context *st) -{ - struct pipe_shader_state shader; - - if (!st->passthrough_fs) { - st->passthrough_fs = - util_make_fragment_passthrough_shader(st->pipe, &shader); - free((void *) shader.tokens); - } - - return st->passthrough_fs; -} - - -/** - * XXX This needs some work yet.... - * Need to "upload" texture images at appropriate times. - */ static void update_textures(struct st_context *st) { struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current; GLuint su; - GLboolean missing_textures = GL_FALSE; st->state.num_textures = 0; @@ -85,13 +64,11 @@ update_textures(struct st_context *st) retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); if (!retval) { /* out of mem */ - missing_textures = GL_TRUE; + /* missing texture */ continue; } st->state.num_textures = su + 1; - - stObj->teximage_realloc = TRUE; } pt = st_get_stobj_texture(stObj); @@ -103,12 +80,6 @@ update_textures(struct st_context *st) cso_set_sampler_textures(st->cso_context, st->state.num_textures, st->state.sampler_texture); - - if (missing_textures) { - /* use a pass-through frag shader that uses no textures */ - void *fs = get_passthrough_fs(st); - cso_set_fragment_shader_handle(st->cso_context, fs); - } } @@ -120,3 +91,52 @@ const struct st_tracked_state st_update_texture = { }, update_textures /* update */ }; + + + + +static void +finalize_textures(struct st_context *st) +{ + struct gl_fragment_program *fprog = st->ctx->FragmentProgram._Current; + const GLboolean prev_missing_textures = st->missing_textures; + GLuint su; + + st->missing_textures = GL_FALSE; + + for (su = 0; su < st->ctx->Const.MaxTextureCoordUnits; su++) { + if (fprog->Base.SamplersUsed & (1 << su)) { + const GLuint texUnit = fprog->Base.SamplerUnits[su]; + struct gl_texture_object *texObj + = st->ctx->Texture.Unit[texUnit]._Current; + struct st_texture_object *stObj = st_texture_object(texObj); + + if (texObj) { + GLboolean flush, retval; + + retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); + if (!retval) { + /* out of mem */ + st->missing_textures = GL_TRUE; + continue; + } + + stObj->teximage_realloc = TRUE; + } + } + } + + if (prev_missing_textures != st->missing_textures) + st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM; +} + + + +const struct st_tracked_state st_finalize_textures = { + "st_finalize_textures", /* name */ + { /* dirty */ + _NEW_TEXTURE, /* mesa */ + 0, /* st */ + }, + finalize_textures /* update */ +}; diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 747d4905e6..2368c31f4b 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -358,6 +358,10 @@ st_render_texture(GLcontext *ctx, struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; struct pipe_texture *pt; + struct st_texture_object *stObj; + const struct gl_texture_image *texImage = + att->Texture->Image[att->CubeMapFace][att->TextureLevel]; + assert(!att->Renderbuffer); @@ -374,27 +378,42 @@ st_render_texture(GLcontext *ctx, strb = st_renderbuffer(rb); /* get the texture for the texture object */ + stObj = st_texture_object(att->Texture); + + /* point renderbuffer at texobject */ + strb->rtt = stObj; + strb->rtt_level = att->TextureLevel; + strb->rtt_face = att->CubeMapFace; + strb->rtt_slice = att->Zoffset; + + rb->Width = texImage->Width2; + rb->Height = texImage->Height2; + /*printf("***** render to texture level %d: %d x %d\n", att->TextureLevel, rb->Width, rb->Height);*/ + pt = st_get_texobj_texture(att->Texture); assert(pt); - assert(pt->width[att->TextureLevel]); - - rb->Width = pt->width[att->TextureLevel]; - rb->Height = pt->height[att->TextureLevel]; + /*printf("***** pipe texture %d x %d\n", pt->width[0], pt->height[0]);*/ pipe_texture_reference( &strb->texture, pt ); + pipe_surface_reference(&strb->surface, NULL); + +#if 0 /* the renderbuffer's surface is inside the texture */ strb->surface = screen->get_tex_surface(screen, pt, att->CubeMapFace, - att->TextureLevel, + att->TextureLevel /*- att->Texture->BaseLevel*/, att->Zoffset, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE); + printf("***** surface size: %d x %d\n", strb->surface->width, strb->surface->height); + assert(strb->surface); assert(screen->is_format_supported(screen, strb->surface->format, PIPE_TEXTURE)); assert(screen->is_format_supported(screen, strb->surface->format, PIPE_SURFACE)); init_renderbuffer_bits(strb, pt->format); +#endif /* printf("RENDER TO TEXTURE obj=%p pt=%p surf=%p %d x %d\n", @@ -424,7 +443,10 @@ st_finish_render_texture(GLcontext *ctx, ctx->st->pipe->flush(ctx->st->pipe, PIPE_FLUSH_RENDER_CACHE, NULL); - screen->tex_surface_release( screen, &strb->surface ); + if (strb->surface) + screen->tex_surface_release( screen, &strb->surface ); + + strb->rtt = NULL; /* printf("FINISH RENDER TO TEXTURE surf=%p\n", strb->surface); diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index f9cec91314..87b0734a0c 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -44,6 +44,9 @@ struct st_renderbuffer struct pipe_texture *texture; struct pipe_surface *surface; /* temporary view into texture */ enum pipe_format format; /** preferred format, or PIPE_FORMAT_NONE */ + + struct st_texture_object *rtt; /**< GL render to texture's texture */ + int rtt_level, rtt_face, rtt_slice; }; diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 0b2b9d544d..3615fafc0a 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -183,6 +183,8 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct gl_pixelstore_attrib clippedPacking = *pack; struct pipe_surface *surf; + assert(ctx->ReadBuffer->Width > 0); + /* XXX convolution not done yet */ assert((transferOps & IMAGE_CONVOLUTION_BIT) == 0); diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 3206215b2e..3468b5f2a1 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1496,6 +1496,7 @@ st_finalize_texture(GLcontext *ctx, stObj->pt->cpp != cpp || stObj->pt->compressed != firstImage->base.IsCompressed) { pipe_texture_release(&stObj->pt); + ctx->st->dirty.st |= ST_NEW_FRAMEBUFFER; } } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 1ca779d0a9..69be4ebdd0 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -53,6 +53,7 @@ struct bitmap_cache; #define ST_NEW_MESA 0x1 /* Mesa state has changed */ #define ST_NEW_FRAGMENT_PROGRAM 0x2 #define ST_NEW_VERTEX_PROGRAM 0x4 +#define ST_NEW_FRAMEBUFFER 0x8 struct st_state_flags { @@ -121,6 +122,8 @@ struct st_context struct st_state_flags dirty; + GLboolean missing_textures; + GLfloat polygon_offset_scale; /* ?? */ /** Mapping from VERT_RESULT_x to post-transformed vertex slot */ diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 2b3742d4e5..d0f56c9717 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -315,6 +315,22 @@ st_texture_image_copy(struct pipe_context *pipe, assert(src->width[srcLevel] == width); assert(src->height[srcLevel] == height); +#if 0 + { + src_surface = screen->get_tex_surface(screen, src, face, srcLevel, i, + PIPE_BUFFER_USAGE_CPU_READ); + ubyte *map = screen->surface_map(screen, src_surface, PIPE_BUFFER_USAGE_CPU_READ); + map += src_surface->width * src_surface->height * 4 / 2; + printf("%s center pixel: %d %d %d %d (pt %p[%d] -> %p[%d])\n", + __FUNCTION__, + map[0], map[1], map[2], map[3], + src, srcLevel, dst, dstLevel); + + screen->surface_unmap(screen, src_surface); + pipe_surface_reference(&src_surface, NULL); + } +#endif + dst_surface = screen->get_tex_surface(screen, dst, face, dstLevel, i, PIPE_BUFFER_USAGE_GPU_WRITE); -- cgit v1.2.3 From a266af9200444a4f651e0ff0cf54e533be38d44d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 7 May 2008 20:18:49 -0600 Subject: mesa: minor fixes in _mesa_free_shader_program_data() and _mesa_link_program() From master: Set shProg->NumShaders = 0 after freeing program's list of shaders. Set _NEW_PROGRAM in _mesa_link_program() to fix a state validation bug. --- src/mesa/shader/shader_api.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index f12fa28d97..5ec03563ba 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -126,6 +126,8 @@ _mesa_free_shader_program_data(GLcontext *ctx, for (i = 0; i < shProg->NumShaders; i++) { _mesa_reference_shader(ctx, &shProg->Shaders[i], NULL); } + shProg->NumShaders = 0; + if (shProg->Shaders) { _mesa_free(shProg->Shaders); shProg->Shaders = NULL; @@ -1008,6 +1010,8 @@ _mesa_link_program(GLcontext *ctx, GLuint program) return; } + FLUSH_VERTICES(ctx, _NEW_PROGRAM); + _slang_link(ctx, program, shProg); } -- cgit v1.2.3 From 53996e562f8623ccb1defeaaa77efdff24477808 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 8 May 2008 14:56:03 +0900 Subject: pipebuffer: Don't include standard headers directly. --- src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index 8de286e3f9..f6cc7a525b 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -50,7 +50,8 @@ #define PB_BUFMGR_H_ -#include +#include "pipe/p_compiler.h" +#include "pipe/p_error.h" #ifdef __cplusplus @@ -68,7 +69,6 @@ struct pipe_winsys; */ struct pb_manager { - /* XXX: we will likely need more allocation flags */ struct pb_buffer * (*create_buffer)( struct pb_manager *mgr, size_t size, -- cgit v1.2.3 From 665b327a47ce80d136e91cfafedbc165227ea168 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 8 May 2008 15:19:07 +0900 Subject: gallium: Really free hash entries. Hook up to the new cso_hash_erase function. --- src/gallium/auxiliary/util/u_hash_table.c | 77 +++++++++++++++++++++++++++---- src/gallium/auxiliary/util/u_hash_table.h | 4 ++ 2 files changed, 72 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c index f3f16a8d94..dd5eca7fca 100644 --- a/src/gallium/auxiliary/util/u_hash_table.c +++ b/src/gallium/auxiliary/util/u_hash_table.c @@ -67,6 +67,13 @@ struct hash_table_item }; +static INLINE struct hash_table_item * +hash_table_item(struct cso_hash_iter iter) +{ + return (struct hash_table_item *)cso_hash_iter_data(iter); +} + + struct hash_table * hash_table_create(unsigned (*hash)(void *key), int (*compare)(void *key1, void *key2)) @@ -90,7 +97,27 @@ hash_table_create(unsigned (*hash)(void *key), } -static struct hash_table_item * +static INLINE struct cso_hash_iter +hash_table_find_iter(struct hash_table *ht, + void *key, + unsigned key_hash) +{ + struct cso_hash_iter iter; + struct hash_table_item *item; + + iter = cso_hash_find(ht->cso, key_hash); + while (!cso_hash_iter_is_null(iter)) { + item = (struct hash_table_item *)cso_hash_iter_data(iter); + if (!ht->compare(item->key, key)) + break; + iter = cso_hash_iter_next(iter); + } + + return iter; +} + + +static INLINE struct hash_table_item * hash_table_find_item(struct hash_table *ht, void *key, unsigned key_hash) @@ -117,6 +144,7 @@ hash_table_set(struct hash_table *ht, { unsigned key_hash; struct hash_table_item *item; + struct cso_hash_iter iter; assert(ht); @@ -136,9 +164,8 @@ hash_table_set(struct hash_table *ht, item->key = key; item->value = value; - cso_hash_insert(ht->cso, key_hash, item); - /* FIXME: there is no OOM propagation in cso_hash */ - if(0) { + iter = cso_hash_insert(ht->cso, key_hash, item); + if(cso_hash_iter_is_null(iter)) { FREE(item); return PIPE_ERROR_OUT_OF_MEMORY; } @@ -171,19 +198,39 @@ hash_table_remove(struct hash_table *ht, void *key) { unsigned key_hash; + struct cso_hash_iter iter; struct hash_table_item *item; assert(ht); key_hash = ht->hash(key); - item = hash_table_find_item(ht, key, key_hash); - if(!item) + iter = hash_table_find_iter(ht, key, key_hash); + if(cso_hash_iter_is_null(iter)) return; - /* FIXME: cso_hash_take takes the first element of the collision list - * indiscriminately, so we can not take the item down. */ - item->value = NULL; + item = hash_table_item(iter); + assert(item); + FREE(item); + + cso_hash_erase(ht->cso, iter); +} + + +void +hash_table_clear(struct hash_table *ht) +{ + struct cso_hash_iter iter; + struct hash_table_item *item; + + assert(ht); + + iter = cso_hash_first_node(ht->cso); + while (!cso_hash_iter_is_null(iter)) { + item = (struct hash_table_item *)cso_hash_take(ht->cso, cso_hash_iter_key(iter)); + FREE(item); + iter = cso_hash_first_node(ht->cso); + } } @@ -196,6 +243,8 @@ hash_table_foreach(struct hash_table *ht, struct hash_table_item *item; enum pipe_error result; + assert(ht); + iter = cso_hash_first_node(ht->cso); while (!cso_hash_iter_is_null(iter)) { item = (struct hash_table_item *)cso_hash_iter_data(iter); @@ -212,7 +261,17 @@ hash_table_foreach(struct hash_table *ht, void hash_table_destroy(struct hash_table *ht) { + struct cso_hash_iter iter; + struct hash_table_item *item; + assert(ht); + + iter = cso_hash_first_node(ht->cso); + while (!cso_hash_iter_is_null(iter)) { + item = (struct hash_table_item *)cso_hash_iter_data(iter); + FREE(item); + iter = cso_hash_iter_next(iter); + } cso_hash_delete(ht->cso); diff --git a/src/gallium/auxiliary/util/u_hash_table.h b/src/gallium/auxiliary/util/u_hash_table.h index 1583bd7548..feee881582 100644 --- a/src/gallium/auxiliary/util/u_hash_table.h +++ b/src/gallium/auxiliary/util/u_hash_table.h @@ -75,6 +75,10 @@ hash_table_remove(struct hash_table *ht, void *key); +void +hash_table_clear(struct hash_table *ht); + + enum pipe_error hash_table_foreach(struct hash_table *ht, enum pipe_error (*callback)(void *key, void *value, void *data), -- cgit v1.2.3 From 82dd0225e7e21a35ca66d439dce8cfa39d782470 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 8 May 2008 15:28:42 +0900 Subject: pipebuffer: Preliminary buffer validation. Use table to store a list of buffers to validate. Unfortunately cso_hash shrinks/regrows the hash every time, so still has to be addressed. Multi-thread validation is still WIP. --- src/gallium/auxiliary/pipebuffer/Makefile | 1 + src/gallium/auxiliary/pipebuffer/SConscript | 1 + src/gallium/auxiliary/pipebuffer/pb_validate.c | 170 +++++++++++++++++++++++++ src/gallium/auxiliary/pipebuffer/pb_validate.h | 91 +++++++++++++ 4 files changed, 263 insertions(+) create mode 100644 src/gallium/auxiliary/pipebuffer/pb_validate.c create mode 100644 src/gallium/auxiliary/pipebuffer/pb_validate.h (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile index d654dbcc91..ff09011b66 100644 --- a/src/gallium/auxiliary/pipebuffer/Makefile +++ b/src/gallium/auxiliary/pipebuffer/Makefile @@ -11,6 +11,7 @@ C_SOURCES = \ pb_bufmgr_mm.c \ pb_bufmgr_pool.c \ pb_bufmgr_slab.c \ + pb_validate.c \ pb_winsys.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript index 604a217982..9db0c0eae3 100644 --- a/src/gallium/auxiliary/pipebuffer/SConscript +++ b/src/gallium/auxiliary/pipebuffer/SConscript @@ -10,6 +10,7 @@ pipebuffer = env.ConvenienceLibrary( 'pb_bufmgr_mm.c', 'pb_bufmgr_pool.c', 'pb_bufmgr_slab.c', + 'pb_validate.c', 'pb_winsys.c', ]) diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.c b/src/gallium/auxiliary/pipebuffer/pb_validate.c new file mode 100644 index 0000000000..a0a0965a46 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_validate.c @@ -0,0 +1,170 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Buffer validation. + * + * @author Jose Fonseca + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_error.h" +#include "pipe/p_util.h" +#include "pipe/p_debug.h" + +#include "util/u_hash_table.h" + +#include "pb_buffer.h" +#include "pb_buffer_fenced.h" + + +struct pb_validate +{ + struct hash_table *buffer_table; +}; + + +static unsigned buffer_table_hash(void *pb_buf) +{ + return (unsigned)(uintptr_t)pb_buf; +} + + +static int buffer_table_compare(void *pb_buf1, void *pb_buf2) +{ + return (char *)pb_buf2 - (char *)pb_buf1; +} + + +enum pipe_error +pb_validate_add_buffer(struct pb_validate *vl, + struct pb_buffer *buf) +{ + enum pipe_error ret; + + assert(buf); + if(!buf) + return PIPE_ERROR; + + if(!hash_table_get(vl->buffer_table, buf)) { + struct pb_buffer *tmp = NULL; + + ret = hash_table_set(vl->buffer_table, buf, buf); + if(ret != PIPE_OK) + return ret; + + /* Increment reference count */ + pb_reference(&tmp, buf); + } + + return PIPE_OK; +} + + +enum pipe_error +pb_validate_validate(struct pb_validate *vl) +{ + /* FIXME: go through each buffer, ensure its not mapped, its address is + * available -- requires a new pb_buffer interface */ + return PIPE_OK; +} + + +struct pb_validate_fence_data { + struct pb_validate *vl; + struct pipe_fence_handle *fence; +}; + + +static enum pipe_error +pb_validate_fence_cb(void *key, void *value, void *_data) +{ + struct pb_buffer *buf = (struct pb_buffer *)key; + struct pb_validate_fence_data *data = (struct pb_validate_fence_data *)_data; + struct pb_validate *vl = data->vl; + struct pipe_fence_handle *fence = data->fence; + + assert(value == key); + + buffer_fence(buf, fence); + + /* Decrement the reference count -- table entry destroyed later */ + pb_reference(&buf, NULL); + + return PIPE_OK; +} + + +void +pb_validate_fence(struct pb_validate *vl, + struct pipe_fence_handle *fence) +{ + struct pb_validate_fence_data data; + + data.vl = vl; + data.fence = fence; + + hash_table_foreach(vl->buffer_table, + pb_validate_fence_cb, + &data); + + /* FIXME: cso_hash shrinks here, which is not desirable in this use case, + * as it will be refilled right soon */ + hash_table_clear(vl->buffer_table); +} + + +void +pb_validate_destroy(struct pb_validate *vl) +{ + pb_validate_fence(vl, NULL); + hash_table_destroy(vl->buffer_table); + FREE(vl); +} + + +struct pb_validate * +pb_validate_create() +{ + struct pb_validate *vl; + + vl = CALLOC_STRUCT(pb_validate); + if(!vl) + return NULL; + + vl->buffer_table = hash_table_create(buffer_table_hash, + buffer_table_compare); + if(!vl->buffer_table) { + FREE(vl); + return NULL; + } + + return vl; +} + diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.h b/src/gallium/auxiliary/pipebuffer/pb_validate.h new file mode 100644 index 0000000000..b0f05d3119 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_validate.h @@ -0,0 +1,91 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Buffer validation. + * + * @author Jose Fonseca + */ + +#ifndef PB_VALIDATE_H_ +#define PB_VALIDATE_H_ + + +#include "pipe/p_compiler.h" +#include "pipe/p_error.h" + +#ifdef __cplusplus +extern "C" { +#endif + + +struct pb_buffer; +struct pipe_fence_handle; + + +/** + * Buffer validation list. + * + * It holds a list of buffers to be validated and fenced when flushing. + */ +struct pb_validate; + + +enum pipe_error +pb_validate_add_buffer(struct pb_validate *vl, + struct pb_buffer *buf); + +/** + * Validate all buffers for hardware access. + * + * Should be called right before issuing commands to the hardware. + */ +enum pipe_error +pb_validate_validate(struct pb_validate *vl); + +/** + * Fence all buffers and clear the list. + * + * Should be called right before issuing commands to the hardware. + */ +void +pb_validate_fence(struct pb_validate *vl, + struct pipe_fence_handle *fence); + +struct pb_validate * +pb_validate_create(); + +void +pb_validate_destroy(struct pb_validate *vl); + + +#ifdef __cplusplus +} +#endif + +#endif /*PB_VALIDATE_H_*/ -- cgit v1.2.3 From 7ffbfaccfb1484a4ffd5aea0e0e1fbb407977a56 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 8 May 2008 13:26:23 +0100 Subject: gallium: Make sure functions have proper prototypes and remove unused variable. --- src/gallium/auxiliary/pipebuffer/pb_validate.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_validate.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.c b/src/gallium/auxiliary/pipebuffer/pb_validate.c index a0a0965a46..b32fe2252e 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_validate.c +++ b/src/gallium/auxiliary/pipebuffer/pb_validate.c @@ -42,6 +42,7 @@ #include "pb_buffer.h" #include "pb_buffer_fenced.h" +#include "pb_validate.h" struct pb_validate @@ -107,7 +108,6 @@ pb_validate_fence_cb(void *key, void *value, void *_data) { struct pb_buffer *buf = (struct pb_buffer *)key; struct pb_validate_fence_data *data = (struct pb_validate_fence_data *)_data; - struct pb_validate *vl = data->vl; struct pipe_fence_handle *fence = data->fence; assert(value == key); diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.h b/src/gallium/auxiliary/pipebuffer/pb_validate.h index b0f05d3119..3db1d5330b 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_validate.h +++ b/src/gallium/auxiliary/pipebuffer/pb_validate.h @@ -78,7 +78,7 @@ pb_validate_fence(struct pb_validate *vl, struct pipe_fence_handle *fence); struct pb_validate * -pb_validate_create(); +pb_validate_create(void); void pb_validate_destroy(struct pb_validate *vl); -- cgit v1.2.3 From 8eab7de888bb4056c34f80edfbc90a543736ea3b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 9 May 2008 00:08:57 +0900 Subject: pipebuffer: Temporarily reimplement validation as growable array. Jose --- src/gallium/auxiliary/pipebuffer/pb_validate.c | 113 +++++++++++-------------- 1 file changed, 48 insertions(+), 65 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.c b/src/gallium/auxiliary/pipebuffer/pb_validate.c index b32fe2252e..362fd896f3 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_validate.c +++ b/src/gallium/auxiliary/pipebuffer/pb_validate.c @@ -38,52 +38,63 @@ #include "pipe/p_util.h" #include "pipe/p_debug.h" -#include "util/u_hash_table.h" - #include "pb_buffer.h" #include "pb_buffer_fenced.h" #include "pb_validate.h" -struct pb_validate -{ - struct hash_table *buffer_table; -}; +#define PB_VALIDATE_INITIAL_SIZE 1 /* 512 */ -static unsigned buffer_table_hash(void *pb_buf) -{ - return (unsigned)(uintptr_t)pb_buf; -} - - -static int buffer_table_compare(void *pb_buf1, void *pb_buf2) +struct pb_validate { - return (char *)pb_buf2 - (char *)pb_buf1; -} + struct pb_buffer **buffers; + unsigned used; + unsigned size; +}; enum pipe_error pb_validate_add_buffer(struct pb_validate *vl, struct pb_buffer *buf) { - enum pipe_error ret; - assert(buf); if(!buf) return PIPE_ERROR; + + /* We only need to store one reference for each buffer, so avoid storing + * consecutive references for the same buffer. It might not be the more + * common pasttern, but it is easy to implement. + */ + if(vl->used && vl->buffers[vl->used - 1] == buf) { + return PIPE_OK; + } - if(!hash_table_get(vl->buffer_table, buf)) { - struct pb_buffer *tmp = NULL; + /* Grow the table */ + if(vl->used == vl->size) { + unsigned new_size; + struct pb_buffer **new_buffers; + + new_size = vl->size * 2; + if(!new_size) + return PIPE_ERROR_OUT_OF_MEMORY; + + new_buffers = (struct pb_buffer **)REALLOC(vl->buffers, + vl->size*sizeof(struct pb_buffer *), + new_size*sizeof(struct pb_buffer *)); + if(!new_buffers) + return PIPE_ERROR_OUT_OF_MEMORY; - ret = hash_table_set(vl->buffer_table, buf, buf); - if(ret != PIPE_OK) - return ret; + memset(new_buffers + vl->size, 0, (new_size - vl->size)*sizeof(struct pb_buffer *)); - /* Increment reference count */ - pb_reference(&tmp, buf); + vl->size = new_size; + vl->buffers = new_buffers; } + assert(!vl->buffers[vl->used]); + pb_reference(&vl->buffers[vl->used], buf); + ++vl->used; + return PIPE_OK; } @@ -97,54 +108,26 @@ pb_validate_validate(struct pb_validate *vl) } -struct pb_validate_fence_data { - struct pb_validate *vl; - struct pipe_fence_handle *fence; -}; - - -static enum pipe_error -pb_validate_fence_cb(void *key, void *value, void *_data) -{ - struct pb_buffer *buf = (struct pb_buffer *)key; - struct pb_validate_fence_data *data = (struct pb_validate_fence_data *)_data; - struct pipe_fence_handle *fence = data->fence; - - assert(value == key); - - buffer_fence(buf, fence); - - /* Decrement the reference count -- table entry destroyed later */ - pb_reference(&buf, NULL); - - return PIPE_OK; -} - - void pb_validate_fence(struct pb_validate *vl, struct pipe_fence_handle *fence) { - struct pb_validate_fence_data data; - - data.vl = vl; - data.fence = fence; - - hash_table_foreach(vl->buffer_table, - pb_validate_fence_cb, - &data); - - /* FIXME: cso_hash shrinks here, which is not desirable in this use case, - * as it will be refilled right soon */ - hash_table_clear(vl->buffer_table); + unsigned i; + for(i = 0; i < vl->used; ++i) { + buffer_fence(vl->buffers[i], fence); + pb_reference(&vl->buffers[i], NULL); + } + vl->used = 0; } void pb_validate_destroy(struct pb_validate *vl) { - pb_validate_fence(vl, NULL); - hash_table_destroy(vl->buffer_table); + unsigned i; + for(i = 0; i < vl->used; ++i) + pb_reference(&vl->buffers[i], NULL); + FREE(vl->buffers); FREE(vl); } @@ -158,9 +141,9 @@ pb_validate_create() if(!vl) return NULL; - vl->buffer_table = hash_table_create(buffer_table_hash, - buffer_table_compare); - if(!vl->buffer_table) { + vl->size = PB_VALIDATE_INITIAL_SIZE; + vl->buffers = (struct pb_buffer **)CALLOC(vl->size, sizeof(struct pb_buffer *)); + if(!vl->buffers) { FREE(vl); return NULL; } -- cgit v1.2.3 From 1dfb3d4729ce4cd71a593c14dbb2907cd987f8ab Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Thu, 8 May 2008 19:00:11 +0100 Subject: cso_cache: Fix test for currently bound blend state. --- src/gallium/auxiliary/cso_cache/cso_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index eef898f486..a1a3a9efaf 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -84,7 +84,7 @@ static boolean delete_blend_state(struct cso_context *ctx, void *state) { struct cso_blend *cso = (struct cso_blend *)state; - if (ctx->blend == state) + if (ctx->blend == cso->data) return FALSE; if (cso->delete_state) -- cgit v1.2.3 From 2abc1b3641e435e0b68490fa6b0a7ffa7c030c76 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 1 May 2008 12:38:51 -0400 Subject: abstract fetching elts --- src/gallium/auxiliary/draw/draw_pt_varray.c | 17 +++++++++++------ src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 6 ++---- 2 files changed, 13 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 355093f945..b0bd2b983e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -43,6 +43,8 @@ struct varray_frontend { unsigned draw_count; unsigned fetch_count; + unsigned fetch_start; + struct draw_pt_middle_end *middle; unsigned input_prim; @@ -68,15 +70,18 @@ static void varray_flush(struct varray_frontend *varray) varray->draw_count = 0; } -#if 0 -static void varray_check_flush(struct varray_frontend *varray) +static INLINE void fetch_init(struct varray_frontend *varray, + unsigned current_count, + unsigned count) { - if (varray->draw_count + 6 >= DRAW_MAX/* || - varray->fetch_count + 4 >= FETCH_MAX*/) { - varray_flush(varray); + unsigned idx; + const unsigned end = MIN2(FETCH_MAX, count - current_count); + for (idx = 0; idx < end; ++idx) { + varray->fetch_elts[idx] = varray->fetch_start + idx; } + varray->fetch_start += idx; + varray->fetch_count = idx; } -#endif static INLINE void add_draw_el(struct varray_frontend *varray, int idx, ushort flags) diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index b9a319b253..a3509613f5 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -20,10 +20,8 @@ static void FUNC(struct draw_pt_front_end *frontend, start, count); #endif - for (i = 0; i < count; ++i) { - varray->fetch_elts[i] = start + i; - } - varray->fetch_count = count; + varray->fetch_start = start; + fetch_init(varray, 0, count); switch (varray->input_prim) { case PIPE_PRIM_POINTS: -- cgit v1.2.3 From 90a46ed277cc887d49c8d8c627174c3bd693ecf7 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 1 May 2008 23:54:39 -0400 Subject: split larger primitives in the simple varray pt --- src/gallium/auxiliary/draw/draw_pt.c | 3 +- src/gallium/auxiliary/draw/draw_pt_varray.c | 62 ++++++++- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 169 +++++++++++++++++------- 3 files changed, 180 insertions(+), 54 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index c9c5d18313..bccde6c5fd 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -80,8 +80,7 @@ draw_pt_arrays(struct draw_context *draw, /* Pick the right frontend */ - if (draw->pt.user.elts || - count >= 256) { + if (draw->pt.user.elts) { frontend = draw->pt.front.vcache; } else { frontend = draw->pt.front.varray; diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index b0bd2b983e..c9843bded0 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -58,6 +58,11 @@ static void varray_flush(struct varray_frontend *varray) debug_printf("FLUSH fc = %d, dc = %d\n", varray->fetch_count, varray->draw_count); + debug_printf("\telt0 = %d, eltx = %d, draw0 = %d, drawx = %d\n", + varray->fetch_elts[0], + varray->fetch_elts[varray->fetch_count-1], + varray->draw_elts[0], + varray->draw_elts[varray->draw_count-1]); #endif varray->middle->run(varray->middle, varray->fetch_elts, @@ -71,18 +76,69 @@ static void varray_flush(struct varray_frontend *varray) } static INLINE void fetch_init(struct varray_frontend *varray, - unsigned current_count, unsigned count) { unsigned idx; - const unsigned end = MIN2(FETCH_MAX, count - current_count); - for (idx = 0; idx < end; ++idx) { +#if 0 + debug_printf("FETCH INIT c = %d, fs = %d\n", + count, + varray->fetch_start); +#endif + for (idx = 0; idx < count; ++idx) { varray->fetch_elts[idx] = varray->fetch_start + idx; } varray->fetch_start += idx; varray->fetch_count = idx; } + +static boolean split_prim_inplace(unsigned prim, unsigned *first, unsigned *incr) +{ + switch (prim) { + case PIPE_PRIM_POINTS: + *first = 1; + *incr = 1; + return TRUE; + case PIPE_PRIM_LINES: + *first = 2; + *incr = 2; + return TRUE; + case PIPE_PRIM_LINE_STRIP: + *first = 2; + *incr = 1; + return TRUE; + case PIPE_PRIM_TRIANGLES: + *first = 3; + *incr = 3; + return TRUE; + case PIPE_PRIM_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + return TRUE; + case PIPE_PRIM_TRIANGLE_FAN: + *first = 3; + *incr = 1; + return TRUE; + case PIPE_PRIM_QUADS: + *first = 4; + *incr = 4; + return TRUE; + case PIPE_PRIM_QUAD_STRIP: + *first = 4; + *incr = 2; + return TRUE; + case PIPE_PRIM_POLYGON: + *first = 3; + *incr = 1; + return TRUE; + default: + *first = 0; + *incr = 1; /* set to one so that count % incr works */ + return FALSE; + } +} + + static INLINE void add_draw_el(struct varray_frontend *varray, int idx, ushort flags) { diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index a3509613f5..073c1aadbf 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -10,30 +10,42 @@ static void FUNC(struct draw_pt_front_end *frontend, boolean flatfirst = (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first); - unsigned i, flags; + unsigned i, j, flags; + unsigned first, incr; -#if 0 - debug_printf("%s (%d) %d/%d\n", __FUNCTION__, draw->prim, start, count); -#endif -#if 0 - debug_printf("INPUT PRIM = %d (start = %d, count = %d)\n", varray->input_prim, + varray->fetch_start = start; + + split_prim_inplace(varray->input_prim, &first, &incr); + +#if 1 + debug_printf("%s (%d) %d/%d\n", __FUNCTION__, + varray->input_prim, start, count); #endif - varray->fetch_start = start; - fetch_init(varray, 0, count); - switch (varray->input_prim) { case PIPE_PRIM_POINTS: - for (i = 0; i < count; i ++) { - POINT(varray, i + 0); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i < count; i ++) { + POINT(varray, i + 0); + } + fetch_init(varray, end); + varray_flush(varray); } break; case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) { - LINE(varray, DRAW_PIPE_RESET_STIPPLE, - i + 0, i + 1); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+1 < end; i += 2) { + LINE(varray, DRAW_PIPE_RESET_STIPPLE, + i + 0, i + 1); + } + fetch_init(varray, end); + varray_flush(varray); } break; @@ -41,38 +53,68 @@ static void FUNC(struct draw_pt_front_end *frontend, if (count >= 2) { flags = DRAW_PIPE_RESET_STIPPLE; - for (i = 1; i < count; i++, flags = 0) { - LINE(varray, flags, i - 1, i); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 1; i < end; i++, flags = 0) { + LINE(varray, flags, i - 1, i); + } + LINE(varray, flags, i - 1, 0); + fetch_init(varray, end); + varray_flush(varray); } - LINE(varray, flags, i - 1, 0); } break; case PIPE_PRIM_LINE_STRIP: flags = DRAW_PIPE_RESET_STIPPLE; - for (i = 1; i < count; i++, flags = 0) { - LINE(varray, flags, i - 1, i); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 1; i < end; i++, flags = 0) { + LINE(varray, flags, i - 1, i); + } + fetch_init(varray, end); + varray_flush(varray); } break; case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) { - TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - i + 0, i + 1, i + 2); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i += 3) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0, i + 1, i + 2); + } + fetch_init(varray, end); + varray_flush(varray); } break; case PIPE_PRIM_TRIANGLE_STRIP: if (flatfirst) { - for (i = 0; i+2 < count; i++) { - TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - i + 0, i + 1 + (i&1), i + 2 - (i&1)); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0, i + 1 + (i&1), i + 2 - (i&1)); + } + fetch_init(varray, end); + varray_flush(varray); } } else { - for (i = 0; i+2 < count; i++) { - TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - i + 0 + (i&1), i + 1 - (i&1), i + 2); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0 + (i&1), i + 1 - (i&1), i + 2); + } + fetch_init(varray, end); + varray_flush(varray); } } break; @@ -81,51 +123,80 @@ static void FUNC(struct draw_pt_front_end *frontend, if (count >= 3) { if (flatfirst) { flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; - for (i = 0; i+2 < count; i++) { - TRIANGLE(varray, flags, i + 1, i + 2, 0); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, flags, i + 1, i + 2, 0); + } + fetch_init(varray, end); + varray_flush(varray); } } else { flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; - for (i = 0; i+2 < count; i++) { - TRIANGLE(varray, flags, 0, i + 1, i + 2); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, flags, 0, i + 1, i + 2); + } + fetch_init(varray, end); + varray_flush(varray); } } } break; case PIPE_PRIM_QUADS: - for (i = 0; i+3 < count; i += 4) { - QUAD(varray, i + 0, i + 1, i + 2, i + 3); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+3 < end; i += 4) { + QUAD(varray, i + 0, i + 1, i + 2, i + 3); + } + fetch_init(varray, end); + varray_flush(varray); } break; case PIPE_PRIM_QUAD_STRIP: - for (i = 0; i+3 < count; i += 2) { - QUAD(varray, i + 2, i + 0, i + 1, i + 3); + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+3 < end; i += 2) { + QUAD(varray, i + 2, i + 0, i + 1, i + 3); + } + fetch_init(varray, end); + varray_flush(varray); } break; case PIPE_PRIM_POLYGON: { - /* These bitflags look a little odd because we submit the - * vertices as (1,2,0) to satisfy flatshade requirements. - */ - const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; - const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; - const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; - - flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; - - for (i = 0; i+2 < count; i++, flags = edge_middle) { + /* These bitflags look a little odd because we submit the + * vertices as (1,2,0) to satisfy flatshade requirements. + */ + const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; + const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++, flags = edge_middle) { if (i + 3 == count) flags |= edge_last; - TRIANGLE(varray, flags, i + 1, i + 2, 0); - } + TRIANGLE(varray, flags, i + 1, i + 2, 0); + } + fetch_init(varray, end); + varray_flush(varray); } - break; + } + break; default: assert(0); -- cgit v1.2.3 From abb08e9335b5d7cb004dc9e6cec390ab6968abe5 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sat, 3 May 2008 22:32:17 -0400 Subject: implement linear emition and fetching and plug it in the varray paths --- src/gallium/auxiliary/draw/draw_pt.h | 18 +++++ src/gallium/auxiliary/draw/draw_pt_emit.c | 45 +++++++++++ src/gallium/auxiliary/draw/draw_pt_fetch.c | 36 +++++++++ .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 88 ++++++++++++++++++++-- src/gallium/auxiliary/draw/draw_pt_varray.c | 26 ++++++- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 4 +- 6 files changed, 208 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 2dec376cee..2f96ceaf00 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -92,6 +92,12 @@ struct draw_pt_middle_end { const ushort *draw_elts, unsigned draw_count ); + void (*run_linear)(struct draw_pt_middle_end *, + unsigned fetch_start, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count); + void (*finish)( struct draw_pt_middle_end * ); void (*destroy)( struct draw_pt_middle_end * ); }; @@ -152,6 +158,13 @@ void draw_pt_emit( struct pt_emit *emit, const ushort *elts, unsigned count ); +void draw_pt_emit_linear( struct pt_emit *emit, + const float (*vertex_data)[4], + unsigned vertex_count, + unsigned stride, + unsigned start, + unsigned count ); + void draw_pt_emit_destroy( struct pt_emit *emit ); struct pt_emit *draw_pt_emit_create( struct draw_context *draw ); @@ -170,6 +183,11 @@ void draw_pt_fetch_run( struct pt_fetch *fetch, unsigned count, char *verts ); +void draw_pt_fetch_run_linear( struct pt_fetch *fetch, + unsigned start, + unsigned count, + char *verts ); + void draw_pt_fetch_destroy( struct pt_fetch *fetch ); struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ); diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index f9ac16786e..2a961b7088 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -179,6 +179,51 @@ void draw_pt_emit( struct pt_emit *emit, } +void draw_pt_emit_linear(struct pt_emit *emit, + const float (*vertex_data)[4], + unsigned vertex_count, + unsigned stride, + unsigned start, + unsigned count) +{ + struct draw_context *draw = emit->draw; + struct translate *translate = emit->translate; + struct vbuf_render *render = draw->render; + void *hw_verts; + + debug_printf("XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX\n"); + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = render->allocate_vertices(render, + (ushort)translate->key.output_stride, + (ushort)count); + if (!hw_verts) { + assert(0); + return; + } + + translate->set_buffer(translate, 0, + vertex_data, stride); + + translate->set_buffer(translate, 1, + &draw->rasterizer->point_size, + 0); + + translate->run(translate, + 0, + vertex_count, + hw_verts); + + render->draw_arrays(render, start, count); + + render->release_vertices(render, + hw_verts, + translate->key.output_stride, + vertex_count); +} + struct pt_emit *draw_pt_emit_create( struct draw_context *draw ) { struct pt_emit *emit = CALLOC_STRUCT(pt_emit); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 1f765b73ad..d7cc1807d7 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -167,6 +167,42 @@ void draw_pt_fetch_run( struct pt_fetch *fetch, } +void draw_pt_fetch_run_linear( struct pt_fetch *fetch, + unsigned start, + unsigned count, + char *verts ) +{ + struct draw_context *draw = fetch->draw; + struct translate *translate = fetch->translate; + unsigned i; + + for (i = 0; i < draw->pt.nr_vertex_buffers; i++) { + translate->set_buffer(translate, + i, + ((char *)draw->pt.user.vbuffer[i] + + draw->pt.vertex_buffer[i].buffer_offset), + draw->pt.vertex_buffer[i].pitch ); + } + + translate->run( translate, + start, + count, + verts ); + + /* Edgeflags are hard to fit into a translate program, populate + * them separately if required. In the setup above they are + * defaulted to one, so only need this if there is reason to change + * that default: + */ + if (fetch->need_edgeflags) { + for (i = 0; i < count; i++) { + struct vertex_header *vh = (struct vertex_header *)(verts + i * fetch->vertex_size); + vh->edgeflag = draw_pt_get_edgeflag( draw, start + i ); + } + } +} + + struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) { struct pt_fetch *fetch = CALLOC_STRUCT(pt_fetch); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 4ec20493c4..b1e08a8f40 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -162,7 +162,7 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, fpme->vertex_size, draw_elts, draw_count ); - } + } else { draw_pt_emit( fpme->emit, (const float (*)[4])pipeline_verts->data, @@ -177,6 +177,83 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, } +static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, + unsigned fetch_start, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; + struct draw_context *draw = fpme->draw; + struct draw_vertex_shader *shader = draw->vertex_shader; + unsigned opt = fpme->opt; + unsigned alloc_count = align_int( fetch_count, 4 ); + + struct vertex_header *pipeline_verts = + (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count); + + if (!pipeline_verts) { + /* Not much we can do here - just skip the rendering. + */ + assert(0); + return; + } + + /* Fetch into our vertex buffer + */ + draw_pt_fetch_run_linear( fpme->fetch, + fetch_start, + fetch_count, + (char *)pipeline_verts ); + + /* Run the shader, note that this overwrites the data[] parts of + * the pipeline verts. If there is no shader, ie a bypass shader, + * then the inputs == outputs, and are already in the correct + * place. + */ + if (opt & PT_SHADE) + { + shader->run_linear(shader, + (const float (*)[4])pipeline_verts->data, + ( float (*)[4])pipeline_verts->data, + (const float (*)[4])draw->pt.user.constants, + fetch_count, + fpme->vertex_size, + fpme->vertex_size); + } + + if (draw_pt_post_vs_run( fpme->post_vs, + pipeline_verts, + fetch_count, + fpme->vertex_size )) + { + opt |= PT_PIPELINE; + } + + /* Do we need to run the pipeline? + */ + if (opt & PT_PIPELINE) { + draw_pipeline_run( fpme->draw, + fpme->prim, + pipeline_verts, + fetch_count, + fpme->vertex_size, + draw_elts, + draw_count ); + } + else { + draw_pt_emit_linear( fpme->emit, + (const float (*)[4])pipeline_verts->data, + fetch_count, + fpme->vertex_size, + 0, /*start*/ + draw_count ); + } + + FREE(pipeline_verts); +} + + static void fetch_pipeline_finish( struct draw_pt_middle_end *middle ) { @@ -206,10 +283,11 @@ struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit( struct draw_context * if (!fpme) goto fail; - fpme->base.prepare = fetch_pipeline_prepare; - fpme->base.run = fetch_pipeline_run; - fpme->base.finish = fetch_pipeline_finish; - fpme->base.destroy = fetch_pipeline_destroy; + fpme->base.prepare = fetch_pipeline_prepare; + fpme->base.run = fetch_pipeline_run; + fpme->base.run_linear = fetch_pipeline_linear_run; + fpme->base.finish = fetch_pipeline_finish; + fpme->base.destroy = fetch_pipeline_destroy; fpme->draw = draw; diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index c9843bded0..916373acc8 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -75,6 +75,28 @@ static void varray_flush(struct varray_frontend *varray) varray->draw_count = 0; } +static void varray_flush_linear(struct varray_frontend *varray) +{ + if (varray->draw_count) { + debug_printf("FLUSH LINEAR fc = %d, dc = %d\n", + varray->fetch_count, + varray->draw_count); + debug_printf("\telt0 = %d, eltx = %d, draw0 = %d, drawx = %d\n", + varray->fetch_elts[0], + varray->fetch_elts[varray->fetch_count-1], + varray->draw_elts[0], + varray->draw_elts[varray->draw_count-1]); + varray->middle->run_linear(varray->middle, + varray->fetch_elts[0], + varray->fetch_count, + varray->draw_elts, + varray->draw_count); + } + + varray->fetch_count = 0; + varray->draw_count = 0; +} + static INLINE void fetch_init(struct varray_frontend *varray, unsigned count) { @@ -265,8 +287,8 @@ static void varray_prepare(struct draw_pt_front_end *frontend, if (opt & PT_PIPELINE) { varray->base.run = varray_run_extras; - } - else + } + else { varray->base.run = varray_run; } diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index 073c1aadbf..67baafa3be 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -17,7 +17,7 @@ static void FUNC(struct draw_pt_front_end *frontend, split_prim_inplace(varray->input_prim, &first, &incr); -#if 1 +#if 0 debug_printf("%s (%d) %d/%d\n", __FUNCTION__, varray->input_prim, start, count); @@ -88,7 +88,7 @@ static void FUNC(struct draw_pt_front_end *frontend, i + 0, i + 1, i + 2); } fetch_init(varray, end); - varray_flush(varray); + varray_flush_linear(varray); } break; -- cgit v1.2.3 From ff1fee2cae9fabb47d6a2eb1f9f8094fec3c377f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sun, 4 May 2008 00:44:27 -0400 Subject: don't fill in linear fetch_elts --- src/gallium/auxiliary/draw/draw_pt_varray.c | 4 +++- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 3 ++- 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 916373acc8..fb1b59d53e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -78,6 +78,7 @@ static void varray_flush(struct varray_frontend *varray) static void varray_flush_linear(struct varray_frontend *varray) { if (varray->draw_count) { +#if 0 debug_printf("FLUSH LINEAR fc = %d, dc = %d\n", varray->fetch_count, varray->draw_count); @@ -86,8 +87,9 @@ static void varray_flush_linear(struct varray_frontend *varray) varray->fetch_elts[varray->fetch_count-1], varray->draw_elts[0], varray->draw_elts[varray->draw_count-1]); +#endif varray->middle->run_linear(varray->middle, - varray->fetch_elts[0], + varray->fetch_start, varray->fetch_count, varray->draw_elts, varray->draw_count); diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index 67baafa3be..a9f844a357 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -87,8 +87,9 @@ static void FUNC(struct draw_pt_front_end *frontend, TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, i + 0, i + 1, i + 2); } - fetch_init(varray, end); + varray->fetch_count = end; varray_flush_linear(varray); + varray->fetch_start += end; } break; -- cgit v1.2.3 From a24cb269e1ba5434acf8c94abd03517c149b9c51 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sun, 4 May 2008 01:23:01 -0400 Subject: implement linear path for fetch_emit pipeline --- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 64 +++++++++++++++++++++++-- src/gallium/auxiliary/draw/draw_pt_varray.c | 1 + 2 files changed, 61 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index a4de341df8..6d5a54cf0e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -257,6 +257,61 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, } +static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, + unsigned fetch_start, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; + struct draw_context *draw = feme->draw; + void *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)feme->translate->key.output_stride, + (ushort)fetch_count ); + if (!hw_verts) { + assert(0); + return; + } + + /* Single routine to fetch vertices and emit HW verts. + */ + feme->translate->run( feme->translate, + fetch_start, + fetch_count, + hw_verts ); + + if (0) { + unsigned i; + for (i = 0; i < fetch_count; i++) { + debug_printf("\n\nvertex %d:\n", i); + draw_dump_emitted_vertex( feme->vinfo, + (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i ); + } + } + + /* XXX: Draw arrays path to avoid re-emitting index list again and + * again. + */ + draw->render->draw_arrays( draw->render, + 0, /*start*/ + draw_count ); + + /* Done -- that was easy, wasn't it: + */ + draw->render->release_vertices( draw->render, + hw_verts, + feme->translate->key.output_stride, + fetch_count ); + +} + + static void fetch_emit_finish( struct draw_pt_middle_end *middle ) { @@ -285,10 +340,11 @@ struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ) return NULL; } - fetch_emit->base.prepare = fetch_emit_prepare; - fetch_emit->base.run = fetch_emit_run; - fetch_emit->base.finish = fetch_emit_finish; - fetch_emit->base.destroy = fetch_emit_destroy; + fetch_emit->base.prepare = fetch_emit_prepare; + fetch_emit->base.run = fetch_emit_run; + fetch_emit->base.run_linear = fetch_emit_run_linear; + fetch_emit->base.finish = fetch_emit_finish; + fetch_emit->base.destroy = fetch_emit_destroy; fetch_emit->draw = draw; diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index fb1b59d53e..e7e21e4bf6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -88,6 +88,7 @@ static void varray_flush_linear(struct varray_frontend *varray) varray->draw_elts[0], varray->draw_elts[varray->draw_count-1]); #endif + assert(varray->middle->run_linear); varray->middle->run_linear(varray->middle, varray->fetch_start, varray->fetch_count, -- cgit v1.2.3 From 66d72f176de2568f053c6dc54e93d423723ae8aa Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sun, 4 May 2008 01:37:32 -0400 Subject: silence debugging output --- src/gallium/auxiliary/draw/draw_pt_emit.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 2a961b7088..776ca32cfa 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -191,7 +191,9 @@ void draw_pt_emit_linear(struct pt_emit *emit, struct vbuf_render *render = draw->render; void *hw_verts; - debug_printf("XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX\n"); +#if 0 + debug_printf("Linear emit\n"); +#endif /* XXX: need to flush to get prim_vbuf.c to release its allocation?? */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); -- cgit v1.2.3 From e897fd6cd35c6b9e398e1903d2e79678fe85708a Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 5 May 2008 12:49:40 -0400 Subject: fix the regressions --- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index a9f844a357..10ac08ea30 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -28,9 +28,10 @@ static void FUNC(struct draw_pt_front_end *frontend, for (j = 0; j + first <= count; j += i) { unsigned end = MIN2(FETCH_MAX, count - j); end -= (end % incr); - for (i = 0; i < count; i ++) { + for (i = 0; i < count; i++) { POINT(varray, i + 0); } + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -44,6 +45,7 @@ static void FUNC(struct draw_pt_front_end *frontend, LINE(varray, DRAW_PIPE_RESET_STIPPLE, i + 0, i + 1); } + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -60,6 +62,7 @@ static void FUNC(struct draw_pt_front_end *frontend, LINE(varray, flags, i - 1, i); } LINE(varray, flags, i - 1, 0); + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -74,6 +77,7 @@ static void FUNC(struct draw_pt_front_end *frontend, for (i = 1; i < end; i++, flags = 0) { LINE(varray, flags, i - 1, i); } + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -87,6 +91,7 @@ static void FUNC(struct draw_pt_front_end *frontend, TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, i + 0, i + 1, i + 2); } + i = end; varray->fetch_count = end; varray_flush_linear(varray); varray->fetch_start += end; @@ -102,6 +107,7 @@ static void FUNC(struct draw_pt_front_end *frontend, TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, i + 0, i + 1 + (i&1), i + 2 - (i&1)); } + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -114,6 +120,7 @@ static void FUNC(struct draw_pt_front_end *frontend, TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, i + 0 + (i&1), i + 1 - (i&1), i + 2); } + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -130,6 +137,7 @@ static void FUNC(struct draw_pt_front_end *frontend, for (i = 0; i+2 < end; i++) { TRIANGLE(varray, flags, i + 1, i + 2, 0); } + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -142,6 +150,7 @@ static void FUNC(struct draw_pt_front_end *frontend, for (i = 0; i+2 < end; i++) { TRIANGLE(varray, flags, 0, i + 1, i + 2); } + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -156,6 +165,7 @@ static void FUNC(struct draw_pt_front_end *frontend, for (i = 0; i+3 < end; i += 4) { QUAD(varray, i + 0, i + 1, i + 2, i + 3); } + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -168,6 +178,7 @@ static void FUNC(struct draw_pt_front_end *frontend, for (i = 0; i+3 < end; i += 2) { QUAD(varray, i + 2, i + 0, i + 1, i + 3); } + i = end; fetch_init(varray, end); varray_flush(varray); } @@ -193,6 +204,7 @@ static void FUNC(struct draw_pt_front_end *frontend, TRIANGLE(varray, flags, i + 1, i + 2, 0); } + i = end; fetch_init(varray, end); varray_flush(varray); } -- cgit v1.2.3 From fe586f8612dd517b9a1f0d87fbaf3a75e3caf588 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 6 May 2008 18:59:45 -0400 Subject: redo the linear paths --- src/gallium/auxiliary/draw/draw_pipe.c | 39 ++++ src/gallium/auxiliary/draw/draw_private.h | 6 + src/gallium/auxiliary/draw/draw_pt.h | 6 +- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 18 +- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 36 ++-- src/gallium/auxiliary/draw/draw_pt_varray.c | 27 +-- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 5 +- .../auxiliary/draw/draw_pt_varray_tmp_linear.h | 198 +++++++++++++++++++++ 8 files changed, 279 insertions(+), 56 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 46afb0f41f..cb97f955b2 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -212,6 +212,45 @@ void draw_pipeline_run( struct draw_context *draw, draw->pipeline.vertex_count = 0; } +void draw_pipeline_run_linear( struct draw_context *draw, + unsigned prim, + struct vertex_header *vertices, + unsigned count, + unsigned stride ) +{ + char *verts = (char *)vertices; + unsigned i; + + draw->pipeline.verts = verts; + draw->pipeline.vertex_stride = stride; + draw->pipeline.vertex_count = count; + + switch (prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i++) + do_point( draw, + verts + stride * i ); + break; + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) + do_line( draw, + i+0, /* flags */ + verts + stride * ((i+0) & ~DRAW_PIPE_FLAG_MASK), + verts + stride * (i+1)); + break; + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) + do_triangle( draw, + (i+0), /* flags */ + verts + stride * ((i+0) & ~DRAW_PIPE_FLAG_MASK), + verts + stride * (i+1), + verts + stride * (i+2)); + break; + } + + draw->pipeline.verts = NULL; + draw->pipeline.vertex_count = 0; +} void draw_pipeline_flush( struct draw_context *draw, diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index cee58bbf73..e036d498b8 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -247,6 +247,12 @@ void draw_pipeline_run( struct draw_context *draw, const ushort *elts, unsigned count ); +void draw_pipeline_run_linear( struct draw_context *draw, + unsigned prim, + struct vertex_header *vertices, + unsigned count, + unsigned stride ); + void draw_pipeline_flush( struct draw_context *draw, diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 2f96ceaf00..312fdbe4f4 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -93,10 +93,8 @@ struct draw_pt_middle_end { unsigned draw_count ); void (*run_linear)(struct draw_pt_middle_end *, - unsigned fetch_start, - unsigned fetch_count, - const ushort *draw_elts, - unsigned draw_count); + unsigned start, + unsigned count); void (*finish)( struct draw_pt_middle_end * ); void (*destroy)( struct draw_pt_middle_end * ); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 6d5a54cf0e..8df4241b82 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -258,10 +258,8 @@ static void fetch_emit_run( struct draw_pt_middle_end *middle, static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, - unsigned fetch_start, - unsigned fetch_count, - const ushort *draw_elts, - unsigned draw_count ) + unsigned start, + unsigned count ) { struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; struct draw_context *draw = feme->draw; @@ -273,7 +271,7 @@ static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, hw_verts = draw->render->allocate_vertices( draw->render, (ushort)feme->translate->key.output_stride, - (ushort)fetch_count ); + (ushort)count ); if (!hw_verts) { assert(0); return; @@ -282,13 +280,13 @@ static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, /* Single routine to fetch vertices and emit HW verts. */ feme->translate->run( feme->translate, - fetch_start, - fetch_count, + start, + count, hw_verts ); if (0) { unsigned i; - for (i = 0; i < fetch_count; i++) { + for (i = 0; i < count; i++) { debug_printf("\n\nvertex %d:\n", i); draw_dump_emitted_vertex( feme->vinfo, (const uint8_t *)hw_verts + feme->vinfo->size * 4 * i ); @@ -300,14 +298,14 @@ static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, */ draw->render->draw_arrays( draw->render, 0, /*start*/ - draw_count ); + count ); /* Done -- that was easy, wasn't it: */ draw->render->release_vertices( draw->render, hw_verts, feme->translate->key.output_stride, - fetch_count ); + count ); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index b1e08a8f40..dad54690a5 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -178,18 +178,16 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, - unsigned fetch_start, - unsigned fetch_count, - const ushort *draw_elts, - unsigned draw_count ) + unsigned start, + unsigned count) { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; struct draw_vertex_shader *shader = draw->vertex_shader; unsigned opt = fpme->opt; - unsigned alloc_count = align_int( fetch_count, 4 ); + unsigned alloc_count = align_int( count, 4 ); - struct vertex_header *pipeline_verts = + struct vertex_header *pipeline_verts = (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count); if (!pipeline_verts) { @@ -202,8 +200,8 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, /* Fetch into our vertex buffer */ draw_pt_fetch_run_linear( fpme->fetch, - fetch_start, - fetch_count, + start, + count, (char *)pipeline_verts ); /* Run the shader, note that this overwrites the data[] parts of @@ -213,18 +211,18 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, */ if (opt & PT_SHADE) { - shader->run_linear(shader, + shader->run_linear(shader, (const float (*)[4])pipeline_verts->data, ( float (*)[4])pipeline_verts->data, (const float (*)[4])draw->pt.user.constants, - fetch_count, + count, fpme->vertex_size, fpme->vertex_size); } if (draw_pt_post_vs_run( fpme->post_vs, pipeline_verts, - fetch_count, + count, fpme->vertex_size )) { opt |= PT_PIPELINE; @@ -233,21 +231,19 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, /* Do we need to run the pipeline? */ if (opt & PT_PIPELINE) { - draw_pipeline_run( fpme->draw, - fpme->prim, - pipeline_verts, - fetch_count, - fpme->vertex_size, - draw_elts, - draw_count ); + draw_pipeline_run_linear( fpme->draw, + fpme->prim, + pipeline_verts, + count, + fpme->vertex_size); } else { draw_pt_emit_linear( fpme->emit, (const float (*)[4])pipeline_verts->data, - fetch_count, + count, fpme->vertex_size, 0, /*start*/ - draw_count ); + count ); } FREE(pipeline_verts); diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index e7e21e4bf6..59a9569270 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -75,29 +75,18 @@ static void varray_flush(struct varray_frontend *varray) varray->draw_count = 0; } -static void varray_flush_linear(struct varray_frontend *varray) +static void varray_flush_linear(struct varray_frontend *varray, + unsigned start, unsigned count) { - if (varray->draw_count) { + if (count) { #if 0 - debug_printf("FLUSH LINEAR fc = %d, dc = %d\n", - varray->fetch_count, - varray->draw_count); - debug_printf("\telt0 = %d, eltx = %d, draw0 = %d, drawx = %d\n", - varray->fetch_elts[0], - varray->fetch_elts[varray->fetch_count-1], - varray->draw_elts[0], - varray->draw_elts[varray->draw_count-1]); + debug_printf("FLUSH LINEAR start = %d, count = %d\n", + start, + count); #endif assert(varray->middle->run_linear); - varray->middle->run_linear(varray->middle, - varray->fetch_start, - varray->fetch_count, - varray->draw_elts, - varray->draw_count); + varray->middle->run_linear(varray->middle, start, count); } - - varray->fetch_count = 0; - varray->draw_count = 0; } static INLINE void fetch_init(struct varray_frontend *varray, @@ -261,7 +250,7 @@ static INLINE void varray_ef_quad( struct varray_frontend *varray, #define LINE(vc,flags,i0,i1) varray_line(vc,i0,i1) #define POINT(vc,i0) varray_point(vc,i0) #define FUNC varray_run -#include "draw_pt_varray_tmp.h" +#include "draw_pt_varray_tmp_linear.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index 10ac08ea30..335c4c89ca 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -92,9 +92,8 @@ static void FUNC(struct draw_pt_front_end *frontend, i + 0, i + 1, i + 2); } i = end; - varray->fetch_count = end; - varray_flush_linear(varray); - varray->fetch_start += end; + fetch_init(varray, end); + varray_flush(varray); } break; diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h new file mode 100644 index 0000000000..dfa4338407 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -0,0 +1,198 @@ + +static void FUNC(struct draw_pt_front_end *frontend, + pt_elt_func get_elt, + const void *elts, + unsigned count) +{ + struct varray_frontend *varray = (struct varray_frontend *)frontend; + struct draw_context *draw = varray->draw; + unsigned start = (unsigned)elts; + + boolean flatfirst = (draw->rasterizer->flatshade && + draw->rasterizer->flatshade_first); + unsigned i, j, flags; + unsigned first, incr; + + varray->fetch_start = start; + + split_prim_inplace(varray->input_prim, &first, &incr); + +#if 0 + debug_printf("%s (%d) %d/%d\n", __FUNCTION__, + varray->input_prim, + start, count); +#endif + + switch (varray->input_prim) { + case PIPE_PRIM_POINTS: + case PIPE_PRIM_LINES: + case PIPE_PRIM_TRIANGLES: + j = 0; + while (j + first <= count) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + varray_flush_linear(varray, start + j, end); + j += end; + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + flags = DRAW_PIPE_RESET_STIPPLE; + + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 1; i < end; i++, flags = 0) { + LINE(varray, flags, i - 1, i); + } + LINE(varray, flags, i - 1, 0); + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + } + break; + + case PIPE_PRIM_LINE_STRIP: + flags = DRAW_PIPE_RESET_STIPPLE; + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 1; i < end; i++, flags = 0) { + LINE(varray, flags, i - 1, i); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + if (flatfirst) { + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0, i + 1 + (i&1), i + 2 - (i&1)); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + } + else { + for (j = 0; j + first <= count;) { + unsigned end = MIN2(FETCH_MAX, count - j); + //end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + i + 0 + (i&1), i + 1 - (i&1), i + 2); + } + fetch_init(varray, end); + varray_flush(varray); + j += end; + if (j <= count) + j -= incr; + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + if (count >= 3) { + if (flatfirst) { + flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, flags, i + 1, i + 2, 0); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + } + else { + flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++) { + TRIANGLE(varray, flags, 0, i + 1, i + 2); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + } + } + break; + + case PIPE_PRIM_QUADS: + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+3 < end; i += 4) { + QUAD(varray, i + 0, i + 1, i + 2, i + 3); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + break; + + case PIPE_PRIM_QUAD_STRIP: + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+3 < end; i += 2) { + QUAD(varray, i + 2, i + 0, i + 1, i + 3); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + break; + + case PIPE_PRIM_POLYGON: + { + /* These bitflags look a little odd because we submit the + * vertices as (1,2,0) to satisfy flatshade requirements. + */ + const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; + const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + for (j = 0; j + first <= count; j += i) { + unsigned end = MIN2(FETCH_MAX, count - j); + end -= (end % incr); + for (i = 0; i+2 < end; i++, flags = edge_middle) { + + if (i + 3 == count) + flags |= edge_last; + + TRIANGLE(varray, flags, i + 1, i + 2, 0); + } + i = end; + fetch_init(varray, end); + varray_flush(varray); + } + } + break; + + default: + assert(0); + break; + } + + varray_flush(varray); +} + +#undef TRIANGLE +#undef QUAD +#undef POINT +#undef LINE +#undef FUNC -- cgit v1.2.3 From 22323af525d00022a1fa06fab7ee84df5ef2d1f0 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 7 May 2008 19:34:12 -0400 Subject: fix silly mistakes --- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 2 +- src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h | 4 +--- 2 files changed, 2 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index 335c4c89ca..fb49452d8b 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -28,7 +28,7 @@ static void FUNC(struct draw_pt_front_end *frontend, for (j = 0; j + first <= count; j += i) { unsigned end = MIN2(FETCH_MAX, count - j); end -= (end % incr); - for (i = 0; i < count; i++) { + for (i = 0; i < end; i++) { POINT(varray, i + 0); } i = end; diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h index dfa4338407..ab28859c35 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -85,7 +85,7 @@ static void FUNC(struct draw_pt_front_end *frontend, else { for (j = 0; j + first <= count;) { unsigned end = MIN2(FETCH_MAX, count - j); - //end -= (end % incr); + end -= (end % incr); for (i = 0; i+2 < end; i++) { TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, i + 0 + (i&1), i + 1 - (i&1), i + 2); @@ -93,8 +93,6 @@ static void FUNC(struct draw_pt_front_end *frontend, fetch_init(varray, end); varray_flush(varray); j += end; - if (j <= count) - j -= incr; } } break; -- cgit v1.2.3 From 8d709ae1595047b45a81f2fbd22850887fdbfea0 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 8 May 2008 12:10:24 -0400 Subject: fix triangle strips --- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 12 ++++++++++-- src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h | 12 ++++++++++-- 2 files changed, 20 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index fb49452d8b..d137a758e2 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -109,19 +109,27 @@ static void FUNC(struct draw_pt_front_end *frontend, i = end; fetch_init(varray, end); varray_flush(varray); + if (j + first + i <= count) { + varray->fetch_start -= 2; + i -= 2; + } } } else { for (j = 0; j + first <= count; j += i) { unsigned end = MIN2(FETCH_MAX, count - j); - end -= (end % incr); - for (i = 0; i+2 < end; i++) { + end -= (end % incr); + for (i = 0; i + 2 < end; i++) { TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, i + 0 + (i&1), i + 1 - (i&1), i + 2); } i = end; fetch_init(varray, end); varray_flush(varray); + if (j + first + i <= count) { + varray->fetch_start -= 2; + i -= 2; + } } } break; diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h index ab28859c35..4bf04fa62b 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -80,19 +80,27 @@ static void FUNC(struct draw_pt_front_end *frontend, i = end; fetch_init(varray, end); varray_flush(varray); + if (j + first + i <= count) { + varray->fetch_start -= 2; + i -= 2; + } } } else { - for (j = 0; j + first <= count;) { + for (j = 0; j + first <= count; j += i) { unsigned end = MIN2(FETCH_MAX, count - j); end -= (end % incr); for (i = 0; i+2 < end; i++) { TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, i + 0 + (i&1), i + 1 - (i&1), i + 2); } + i = end; fetch_init(varray, end); varray_flush(varray); - j += end; + if (j + first + i <= count) { + varray->fetch_start -= 2; + i -= 2; + } } } break; -- cgit v1.2.3 From 8ea6106f01f38853e9c0f1029da55eb449109aea Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 8 May 2008 15:11:16 -0400 Subject: fix quad strips --- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 4 ++++ src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h | 4 ++++ 2 files changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index d137a758e2..1395275897 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -188,6 +188,10 @@ static void FUNC(struct draw_pt_front_end *frontend, i = end; fetch_init(varray, end); varray_flush(varray); + if (j + first + i <= count) { + varray->fetch_start -= 2; + i -= 2; + } } break; diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h index 4bf04fa62b..6e2b16d9be 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -159,6 +159,10 @@ static void FUNC(struct draw_pt_front_end *frontend, i = end; fetch_init(varray, end); varray_flush(varray); + if (j + first + i <= count) { + varray->fetch_start -= 2; + i -= 2; + } } break; -- cgit v1.2.3 From 6548e9b0183d2ddfc8b57919d5be0e75ef79182e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 14:20:32 -0600 Subject: gallium: plug in dummy stage->destroy func --- src/mesa/state_tracker/st_cb_feedback.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c index 1b50792bd1..19021411cf 100644 --- a/src/mesa/state_tracker/st_cb_feedback.c +++ b/src/mesa/state_tracker/st_cb_feedback.c @@ -173,6 +173,12 @@ feedback_reset_stipple_counter( struct draw_stage *stage ) } +static void +feedback_destroy( struct draw_stage *stage ) +{ + /* no-op */ +} + /** * Create GL feedback drawing stage. */ @@ -188,6 +194,7 @@ draw_glfeedback_stage(GLcontext *ctx, struct draw_context *draw) fs->stage.tri = feedback_tri; fs->stage.flush = feedback_flush; fs->stage.reset_stipple_counter = feedback_reset_stipple_counter; + fs->stage.destroy = feedback_destroy; fs->ctx = ctx; return &fs->stage; -- cgit v1.2.3 From 9002cdb48e65c063ea00e1cb4917d432b22ae0ad Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 8 May 2008 22:07:52 +0100 Subject: softpipe: don't calc det if NO_RAST set --- src/gallium/drivers/softpipe/sp_setup.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index df7be01fcd..5370d85275 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -717,7 +717,7 @@ void setup_tri( struct setup_context *setup, const float (*v1)[4], const float (*v2)[4] ) { - float det = calc_det(v0, v1, v2); + float det; #if DEBUG_VERTS debug_printf("Setup triangle:\n"); @@ -728,7 +728,8 @@ void setup_tri( struct setup_context *setup, if (setup->softpipe->no_rast) return; - + + det = calc_det(v0, v1, v2); /* debug_printf("%s\n", __FUNCTION__ ); */ -- cgit v1.2.3 From fec1d215f623221cb52f22c8f10e5de99ebc9cc2 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 8 May 2008 22:10:15 +0100 Subject: xlib: more SP_NO_RAST support For some resaon normal (non-display-buffer) buffers are being allocated through Xshm... Bypass at least for SP_NO_RAST --- src/gallium/winsys/xlib/fakeglx.c | 9 +++++++++ src/gallium/winsys/xlib/xm_api.c | 3 +++ 2 files changed, 12 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/xlib/fakeglx.c b/src/gallium/winsys/xlib/fakeglx.c index ec77e81fed..2c0075e934 100644 --- a/src/gallium/winsys/xlib/fakeglx.c +++ b/src/gallium/winsys/xlib/fakeglx.c @@ -1689,6 +1689,15 @@ static void Fake_glXSwapBuffers( Display *dpy, GLXDrawable drawable ) { XMesaBuffer buffer = XMesaFindBuffer( dpy, drawable ); + static boolean firsttime = 1, no_rast = 0; + + if (firsttime) { + no_rast = getenv("SP_NO_RAST") != NULL; + firsttime = 0; + } + + if (no_rast) + return; if (buffer) { XMesaSwapBuffers(buffer); diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c index 0c248344b1..26b722f343 100644 --- a/src/gallium/winsys/xlib/xm_api.c +++ b/src/gallium/winsys/xlib/xm_api.c @@ -110,6 +110,9 @@ int xmesa_check_for_xshm( XMesaDisplay *display ) int major, minor, ignore; Bool pixmaps; + if (getenv("SP_NO_RAST")) + return 0; + if (getenv("MESA_NOSHM")) { return 0; } -- cgit v1.2.3 From b514f5f3ba4c9cf6c39cbcdf5bf0d2d8efb8d19b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 8 May 2008 22:13:21 +0100 Subject: draw: only fill in / compare the part of the translate key we're using. It's quite a big struct & we examine it a lot (too much). Reduce the impact of this by just looking at the active part where possible. --- src/gallium/auxiliary/draw/draw_pt_emit.c | 8 ++++---- src/gallium/auxiliary/draw/draw_pt_fetch.c | 8 +++++--- src/gallium/auxiliary/translate/translate.h | 5 ++--- 3 files changed, 11 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index f9ac16786e..671abc25ce 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -49,6 +49,7 @@ void draw_pt_emit_prepare( struct pt_emit *emit, const struct vertex_info *vinfo; unsigned dst_offset; struct translate_key hw_key; + unsigned keysize; unsigned i; boolean ok; @@ -58,12 +59,10 @@ void draw_pt_emit_prepare( struct pt_emit *emit, return; } - memset(&hw_key, 0, sizeof(hw_key)); - /* Must do this after set_primitive() above: */ vinfo = draw->render->get_vertex_info(draw->render); - + keysize = 2*4 + vinfo->num_attribs * sizeof(hw_key.element[0]); /* Translate from pipeline vertices to hw vertices. */ @@ -122,8 +121,9 @@ void draw_pt_emit_prepare( struct pt_emit *emit, hw_key.output_stride = vinfo->size * 4; if (!emit->translate || - memcmp(&emit->translate->key, &hw_key, sizeof(hw_key)) != 0) + memcmp(&emit->translate->key, &hw_key, keysize) != 0) { + memset((char *)&hw_key + keysize, 0, sizeof(hw_key) - keysize); emit->translate = translate_cache_find(emit->cache, &hw_key); } } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 1f765b73ad..a5bebb4ca1 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -62,10 +62,11 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, unsigned i, nr = 0; unsigned dst_offset = 0; struct translate_key key; + unsigned keysize; fetch->vertex_size = vertex_size; - - memset(&key, 0, sizeof(key)); + keysize = (2*4 + + (draw->pt.nr_vertex_elements + 1) * sizeof(key.element[0])); /* Always emit/leave space for a vertex header. * @@ -110,8 +111,9 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, if (!fetch->translate || - memcmp(&fetch->translate->key, &key, sizeof(key)) != 0) + memcmp(&fetch->translate->key, &key, keysize) != 0) { + memset((char *)&key + keysize, 0, sizeof(key) - keysize); fetch->translate = translate_cache_find(fetch->cache, &key); { diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h index 6c15d7e4dc..de6f09d18a 100644 --- a/src/gallium/auxiliary/translate/translate.h +++ b/src/gallium/auxiliary/translate/translate.h @@ -47,10 +47,9 @@ struct translate_element { enum pipe_format input_format; - unsigned input_buffer; - unsigned input_offset; - enum pipe_format output_format; + unsigned input_buffer; + unsigned input_offset; /* can't really reduce the size of these */ unsigned output_offset; }; -- cgit v1.2.3 From e93243cb80ee3ae834a50efe7bacd232d8846305 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 16:02:19 -0600 Subject: gallium: clean-up, minor fix --- src/mesa/state_tracker/st_cb_drawpixels.c | 10 +++------- 1 file changed, 3 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c967c989de..61cbd9e587 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -986,12 +986,7 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE); } -#if 0 - psRead = rbRead->surface; - srcFormat = psRead->format; -#else srcFormat = rbRead->texture->format; -#endif if (screen->is_format_supported(screen, srcFormat, PIPE_TEXTURE)) { texFormat = srcFormat; @@ -1003,8 +998,9 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, assert(texFormat != PIPE_FORMAT_NONE); /* XXX no depth texture formats??? */ } else { - /* todo */ - assert(0); + /* default color format */ + texFormat = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE); + assert(texFormat != PIPE_FORMAT_NONE); } } -- cgit v1.2.3 From 29d9f6b0d263b6ea2f89ea955d53d2671d9a6e43 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 17:11:05 -0600 Subject: gallium: no-op st_Bitmap() if width or height is zero --- src/mesa/state_tracker/st_cb_bitmap.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index f816e59104..56ca171a62 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -673,6 +673,9 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct st_context *st = ctx->st; struct pipe_texture *pt; + if (width == 0 || height == 0) + return; + st_validate_state(st); if (!st->bitmap.vs) { -- cgit v1.2.3 From f3ecd488ddb9ef0949466accbd7db686c8e4662e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 17:11:55 -0600 Subject: gallium: remove dead code --- src/mesa/state_tracker/st_cb_bitmap.c | 9 --------- 1 file changed, 9 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 56ca171a62..b17681a05b 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -579,18 +579,9 @@ st_flush_bitmap_cache(struct st_context *st) /* The texture surface has been mapped until now. * So unmap and release the texture surface before drawing. */ -#if 0 - pipe_surface_unmap(cache->surf); - pipe_surface_reference(&cache->surf, NULL); -#else screen->surface_unmap(screen, cache->surf); screen->tex_surface_release(screen, &cache->surf); -#endif -#if 0 - /* XXX is this needed? */ - pipe->texture_update(pipe, cache->texture, 0, 0x1); -#endif draw_bitmap_quad(st->ctx, cache->xpos, cache->ypos, -- cgit v1.2.3 From a7e8e31eeee320bbc563536389587520875c3b57 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 17:13:01 -0600 Subject: gallium: replace pipe_buffer_destroy() call with pipe_buffer_reference(NULL) --- src/mesa/state_tracker/st_cb_drawpixels.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 61cbd9e587..43e00e6acc 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -484,7 +484,7 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, ubyte *map; /* allocate/load buffer object with vertex data */ - buf = pipe_buffer_create(pipe,32, PIPE_BUFFER_USAGE_VERTEX, + buf = pipe_buffer_create(pipe, 32, PIPE_BUFFER_USAGE_VERTEX, sizeof(verts)); map = pipe_buffer_map(pipe, buf, PIPE_BUFFER_USAGE_CPU_WRITE); memcpy(map, verts, sizeof(verts)); @@ -494,8 +494,7 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, PIPE_PRIM_QUADS, 4, /* verts */ 3); /* attribs/vert */ - - pipe_buffer_destroy(pipe, buf); + pipe_buffer_reference(pipe->winsys, &buf, NULL); } } @@ -602,8 +601,10 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, x1 = x + width * ctx->Pixel.ZoomX; y0 = y; y1 = y + height * ctx->Pixel.ZoomY; + //if(!color) draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex); - + //else + //printf("skip draw quad\n"); /* restore state */ cso_restore_rasterizer(cso); cso_restore_viewport(cso); -- cgit v1.2.3 From 2f07e1caa02c9dcdddc673e61eb91b83b82d283b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 17:42:18 -0600 Subject: gallium: fix render to depth texture --- src/mesa/state_tracker/st_atom_framebuffer.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index c9a30e44b2..d46c3ee16c 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -127,6 +127,11 @@ update_framebuffer_state( struct st_context *st ) strb = st_renderbuffer(fb->Attachment[BUFFER_DEPTH].Renderbuffer); if (strb) { strb = st_renderbuffer(strb->Base.Wrapped); + if (strb->rtt) { + /* rendering to a GL texture, may have to update surface */ + update_renderbuffer_surface(st, strb); + } + assert(strb->surface); framebuffer->zsbuf = strb->surface; } -- cgit v1.2.3 From f9e2f26df3c16eaa0c56db11cd94b5af7a361ee8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 17:45:59 -0600 Subject: gallium: re-enable call to init_renderbuffer_bits(), remove dead code --- src/mesa/state_tracker/st_cb_fbo.c | 15 +-------------- 1 file changed, 1 insertion(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 2368c31f4b..3e592730f8 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -398,22 +398,9 @@ st_render_texture(GLcontext *ctx, pipe_surface_reference(&strb->surface, NULL); -#if 0 - /* the renderbuffer's surface is inside the texture */ - strb->surface = screen->get_tex_surface(screen, pt, - att->CubeMapFace, - att->TextureLevel /*- att->Texture->BaseLevel*/, - att->Zoffset, - PIPE_BUFFER_USAGE_GPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE); - printf("***** surface size: %d x %d\n", strb->surface->width, strb->surface->height); - - assert(strb->surface); - assert(screen->is_format_supported(screen, strb->surface->format, PIPE_TEXTURE)); - assert(screen->is_format_supported(screen, strb->surface->format, PIPE_SURFACE)); + /* the new surface will be created during framebuffer validation */ init_renderbuffer_bits(strb, pt->format); -#endif /* printf("RENDER TO TEXTURE obj=%p pt=%p surf=%p %d x %d\n", -- cgit v1.2.3 From 044d583ba12689cbe99098eb999854303de57f59 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 18:38:02 -0600 Subject: gallium: handle null ptrs --- src/gallium/auxiliary/util/u_draw_quad.c | 81 ++++++++++++++++---------------- 1 file changed, 40 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c index e659edb088..bf143815d8 100644 --- a/src/gallium/auxiliary/util/u_draw_quad.c +++ b/src/gallium/auxiliary/util/u_draw_quad.c @@ -82,52 +82,51 @@ util_draw_texquad(struct pipe_context *pipe, { struct pipe_buffer *vbuf; uint numAttribs = 2, vertexBytes, i, j; - float *v; vertexBytes = 4 * (4 * numAttribs * sizeof(float)); /* XXX create one-time */ vbuf = pipe->winsys->buffer_create(pipe->winsys, 32, PIPE_BUFFER_USAGE_VERTEX, vertexBytes); - assert(vbuf); - - v = (float *) pipe->winsys->buffer_map(pipe->winsys, vbuf, - PIPE_BUFFER_USAGE_CPU_WRITE); - - /* - * Load vertex buffer - */ - for (i = j = 0; i < 4; i++) { - v[j + 2] = z; /* z */ - v[j + 3] = 1.0; /* w */ - v[j + 6] = 0.0; /* r */ - v[j + 7] = 1.0; /* q */ - j += 8; + if (vbuf) { + float *v = (float *) pipe->winsys->buffer_map(pipe->winsys, vbuf, + PIPE_BUFFER_USAGE_CPU_WRITE); + if (v) { + /* + * Load vertex buffer + */ + for (i = j = 0; i < 4; i++) { + v[j + 2] = z; /* z */ + v[j + 3] = 1.0; /* w */ + v[j + 6] = 0.0; /* r */ + v[j + 7] = 1.0; /* q */ + j += 8; + } + + v[0] = x0; + v[1] = y0; + v[4] = 0.0; /*s*/ + v[5] = 0.0; /*t*/ + + v[8] = x1; + v[9] = y0; + v[12] = 1.0; + v[13] = 0.0; + + v[16] = x1; + v[17] = y1; + v[20] = 1.0; + v[21] = 1.0; + + v[24] = x0; + v[25] = y1; + v[28] = 0.0; + v[29] = 1.0; + + pipe->winsys->buffer_unmap(pipe->winsys, vbuf); + util_draw_vertex_buffer(pipe, vbuf, PIPE_PRIM_TRIANGLE_FAN, 4, 2); + } + + pipe_buffer_reference(pipe->winsys, &vbuf, NULL); } - - v[0] = x0; - v[1] = y0; - v[4] = 0.0; /*s*/ - v[5] = 0.0; /*t*/ - - v[8] = x1; - v[9] = y0; - v[12] = 1.0; - v[13] = 0.0; - - v[16] = x1; - v[17] = y1; - v[20] = 1.0; - v[21] = 1.0; - - v[24] = x0; - v[25] = y1; - v[28] = 0.0; - v[29] = 1.0; - - pipe->winsys->buffer_unmap(pipe->winsys, vbuf); - - util_draw_vertex_buffer(pipe, vbuf, PIPE_PRIM_TRIANGLE_FAN, 4, 2); - - pipe_buffer_reference(pipe->winsys, &vbuf, NULL); } -- cgit v1.2.3 From 2a39dbe7364af5444b1eb43650dfc31ed09257dc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 19:19:52 -0600 Subject: gallium: fix out of tex memory crashes --- src/mesa/state_tracker/st_cb_texture.c | 11 +++++++++-- src/mesa/state_tracker/st_texture.c | 5 ++++- 2 files changed, 13 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 3468b5f2a1..c0dba4cf2d 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -645,7 +645,8 @@ st_TexImage(GLcontext * ctx, if (stImage->pt) { texImage->Data = st_texture_image_map(ctx->st, stImage, 0, PIPE_BUFFER_USAGE_CPU_WRITE); - dstRowStride = stImage->surface->pitch * stImage->surface->cpp; + if (stImage->surface) + dstRowStride = stImage->surface->pitch * stImage->surface->cpp; } else { /* Allocate regular memory and store the image there temporarily. */ @@ -663,6 +664,11 @@ st_TexImage(GLcontext * ctx, texImage->Data = malloc(sizeInBytes); } + if (!texImage->Data) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage"); + return; + } + DBG("Upload image %dx%dx%d row_len %x pitch %x\n", width, height, depth, width * texelBytes, dstRowStride); @@ -906,7 +912,8 @@ st_TexSubimage(GLcontext * ctx, if (stImage->pt) { texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset, PIPE_BUFFER_USAGE_CPU_WRITE); - dstRowStride = stImage->surface->pitch * stImage->surface->cpp; + if (stImage->surface) + dstRowStride = stImage->surface->pitch * stImage->surface->cpp; } if (!texImage->Data) { diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index d0f56c9717..9553b34e31 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -197,7 +197,10 @@ st_texture_image_map(struct st_context *st, struct st_texture_image *stImage, stImage->level, zoffset, flags); - return screen->surface_map(screen, stImage->surface, flags); + if (stImage->surface) + return screen->surface_map(screen, stImage->surface, flags); + else + return NULL; } -- cgit v1.2.3 From f763cc3cb54ee419902bdf24065e7c1948e92b67 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 8 May 2008 19:32:49 -0600 Subject: gallium: don't free shader.tokens in get_passthrough_fs() --- src/mesa/state_tracker/st_atom_shader.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 8839ab380f..a62ea8161c 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -262,7 +262,9 @@ get_passthrough_fs(struct st_context *st) if (!st->passthrough_fs) { st->passthrough_fs = util_make_fragment_passthrough_shader(st->pipe, &shader); +#if 0 /* We actually need to keep the tokens around at this time */ free((void *) shader.tokens); +#endif } return st->passthrough_fs; -- cgit v1.2.3 From 2268306f58769dff4b2b1da8bb668bdff2856d70 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 9 May 2008 11:02:10 +0900 Subject: gallium: Don't serialize GPU writes. Only make sure the GPU is finished with a buffer before mapping. The opposite -- waiting for the CPU to be finished before handing to the CPU -- must be done before fencing. --- .../auxiliary/pipebuffer/pb_buffer_fenced.c | 100 ++++++++++----------- 1 file changed, 50 insertions(+), 50 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 2fa0842971..7f236887a9 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -168,6 +168,28 @@ _fenced_buffer_remove(struct fenced_buffer *fenced_buf) } +static INLINE enum pipe_error +_fenced_buffer_finish(struct fenced_buffer *fenced_buf) +{ + struct fenced_buffer_list *fenced_list = fenced_buf->list; + struct pipe_winsys *winsys = fenced_list->winsys; + + debug_warning("waiting for GPU"); + + assert(fenced_buf->fence); + if(fenced_buf->fence) { + if(winsys->fence_finish(winsys, fenced_buf->fence, 0) != 0) { + return PIPE_ERROR; + } + /* Remove from the fenced list */ + _fenced_buffer_remove(fenced_buf); /* TODO: remove consequents */ + } + + fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE; + return PIPE_OK; +} + + /** * Free as many fenced buffers from the list head as possible. */ @@ -207,40 +229,6 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, } -/** - * Serialize writes, but allow concurrent reads. - */ -static INLINE enum pipe_error -fenced_buffer_serialize(struct fenced_buffer *fenced_buf, unsigned flags) -{ - struct fenced_buffer_list *fenced_list = fenced_buf->list; - struct pipe_winsys *winsys = fenced_list->winsys; - - /* Allow concurrent reads */ - if(((fenced_buf->flags | flags) & PIPE_BUFFER_USAGE_WRITE) == 0) - return PIPE_OK; - - /* Wait for the CPU to finish */ - if(fenced_buf->mapcount) { - /* FIXME: Use thread conditions variables to signal when mapcount - * reaches zero */ - debug_warning("attemp to write concurrently to buffer"); - /* XXX: we must not fail here in order to support texture mipmap generation - return PIPE_ERROR_RETRY; - */ - } - - /* Wait for the GPU to finish */ - if(fenced_buf->fence) { - if(winsys->fence_finish(winsys, fenced_buf->fence, 0) != 0) - return PIPE_ERROR_RETRY; - _fenced_buffer_remove(fenced_buf); - } - - return PIPE_OK; -} - - static void fenced_buffer_destroy(struct pb_buffer *buf) { @@ -280,15 +268,28 @@ fenced_buffer_map(struct pb_buffer *buf, { struct fenced_buffer *fenced_buf = fenced_buffer(buf); void *map; - assert((flags & ~PIPE_BUFFER_USAGE_CPU_READ_WRITE) == 0); + + assert(!(flags & ~PIPE_BUFFER_USAGE_CPU_READ_WRITE)); + flags &= PIPE_BUFFER_USAGE_CPU_READ_WRITE; - if(fenced_buffer_serialize(fenced_buf, flags) != PIPE_OK) - return NULL; + /* Check for GPU read/write access */ + if(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) { + /* Wait for the GPU to finish writing */ + _fenced_buffer_finish(fenced_buf); + } + + /* Check for CPU write access (read is OK) */ + if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) { + /* this is legal -- just for debugging */ + debug_warning("concurrent CPU writes"); + } map = pb_map(fenced_buf->buffer, flags); - if(map) + if(map) { ++fenced_buf->mapcount; - fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE; + fenced_buf->flags |= flags; + } + return map; } @@ -298,10 +299,12 @@ fenced_buffer_unmap(struct pb_buffer *buf) { struct fenced_buffer *fenced_buf = fenced_buffer(buf); assert(fenced_buf->mapcount); - pb_unmap(fenced_buf->buffer); - --fenced_buf->mapcount; - if(!fenced_buf->mapcount) - fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE; + if(fenced_buf->mapcount) { + pb_unmap(fenced_buf->buffer); + --fenced_buf->mapcount; + if(!fenced_buf->mapcount) + fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE; + } } @@ -334,8 +337,10 @@ fenced_buffer_create(struct fenced_buffer_list *fenced_list, return NULL; buf = CALLOC_STRUCT(fenced_buffer); - if(!buf) + if(!buf) { + pb_reference(&buffer, NULL); return NULL; + } buf->base.base.refcount = 1; buf->base.base.alignment = buffer->base.alignment; @@ -374,7 +379,7 @@ buffer_fence(struct pb_buffer *buf, fenced_list = fenced_buf->list; winsys = fenced_list->winsys; - if(fence == fenced_buf->fence) { + if(!fence || fence == fenced_buf->fence) { /* Handle the same fence case specially, not only because it is a fast * path, but mostly to avoid serializing two writes with the same fence, * as that would bring the hardware down to synchronous operation without @@ -384,11 +389,6 @@ buffer_fence(struct pb_buffer *buf, return; } - if(fenced_buffer_serialize(fenced_buf, flags) != PIPE_OK) { - /* FIXME: propagate error */ - (void)0; - } - _glthread_LOCK_MUTEX(fenced_list->mutex); if (fenced_buf->fence) _fenced_buffer_remove(fenced_buf); -- cgit v1.2.3 From 47f639a62989cea4b3b14cd73bb39de85acec8ea Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 9 May 2008 14:34:51 +0900 Subject: gallium: Disable debug_get_option for release builds on Windows. It always creates the C:\gallium.cfg , even if it does not exists, which might be confusing. --- src/gallium/auxiliary/util/p_debug.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 951dd5a2d4..ce7fb58956 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -154,6 +154,7 @@ debug_get_option(const char *name, const char *dfault) { const char *result; #ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY +#ifdef DEBUG ULONG_PTR iFile = 0; const void *pMap = NULL; const char *sol, *eol, *sep; @@ -183,6 +184,9 @@ debug_get_option(const char *name, const char *dfault) } EngUnmapFile(iFile); } +#else + result = dfault; +#endif #else result = getenv(name); -- cgit v1.2.3 From 54777e124c38812e5e80319048b6c71009bcf9dd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 9 May 2008 18:47:44 +0900 Subject: gallium: Define util_vsprintf. --- src/gallium/auxiliary/util/u_string.h | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_string.h b/src/gallium/auxiliary/util/u_string.h index d7d925ee9b..73c88d87b4 100644 --- a/src/gallium/auxiliary/util/u_string.h +++ b/src/gallium/auxiliary/util/u_string.h @@ -54,6 +54,12 @@ extern "C" { int util_vsnprintf(char *, size_t, const char *, va_list); int util_snprintf(char *str, size_t size, const char *format, ...); +static INLINE void +util_vsprintf(char *str, const char *format, va_list ap) +{ + util_vsnprintf(str, (size_t)-1, format, ap); +} + static INLINE void util_sprintf(char *str, const char *format, ...) { @@ -158,6 +164,7 @@ util_memmove(void *dest, const void *src, size_t n) #define util_vsnprintf vsnprintf #define util_snprintf snprintf +#define util_vsprintf vsprintf #define util_sprintf sprintf #define util_strchr strchr #define util_strcmp strcmp -- cgit v1.2.3 From 80474d576c2e92441f6bcc18faae71a38b91bd70 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 9 May 2008 13:09:58 +0100 Subject: translate: helper functions for mimizing cost of key compares --- src/gallium/auxiliary/translate/translate.h | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h index de6f09d18a..b8210af50c 100644 --- a/src/gallium/auxiliary/translate/translate.h +++ b/src/gallium/auxiliary/translate/translate.h @@ -95,6 +95,27 @@ struct translate *translate_lookup_or_create( struct translate_context *tctx, struct translate *translate_create( const struct translate_key *key ); +static INLINE int translate_keysize( const struct translate_key *key ) +{ + return 2 * sizeof(int) + key->nr_elements * sizeof(struct translate_element); +} + +static INLINE int translate_key_compare( const struct translate_key *a, + const struct translate_key *b ) +{ + int keysize = translate_keysize(a); + return memcmp(a, b, keysize); +} + + +static INLINE void translate_key_sanitize( struct translate_key *a ) +{ + int keysize = translate_keysize(a); + char *ptr = (char *)a; + memset(ptr + keysize, 0, sizeof(*a) - keysize); +} + + /******************************************************************************* * Private: */ -- cgit v1.2.3 From 1a03812fb57e956b438cd42ac68978facb49a99d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 9 May 2008 13:10:15 +0100 Subject: draw: mimize cost of translate key compares, use cache universally --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 20 ++++++++++++++------ src/gallium/auxiliary/draw/draw_pt_emit.c | 10 +++++----- src/gallium/auxiliary/draw/draw_pt_fetch.c | 7 ++----- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 3 ++- 4 files changed, 23 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 2a19e6916a..86a7d1c730 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -42,6 +42,7 @@ #include "draw_vertex.h" #include "draw_pipe.h" #include "translate/translate.h" +#include "translate/translate_cache.h" /** @@ -75,6 +76,8 @@ struct vbuf_stage { /* Cache point size somewhere it's address won't change: */ float point_size; + + struct translate_cache *cache; }; @@ -220,7 +223,6 @@ vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) /* Translate from pipeline vertices to hw vertices. */ dst_offset = 0; - memset(&hw_key, 0, sizeof(hw_key)); for (i = 0; i < vbuf->vinfo->num_attribs; i++) { unsigned emit_sz = 0; @@ -277,12 +279,10 @@ vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) /* Don't bother with caching at this stage: */ if (!vbuf->translate || - memcmp(&vbuf->translate->key, &hw_key, sizeof(hw_key)) != 0) + translate_key_compare(&vbuf->translate->key, &hw_key) != 0) { - if (vbuf->translate) - vbuf->translate->release(vbuf->translate); - - vbuf->translate = translate_create( &hw_key ); + translate_key_sanitize(&hw_key); + vbuf->translate = translate_cache_find(vbuf->cache, &hw_key); vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0); } @@ -433,6 +433,9 @@ static void vbuf_destroy( struct draw_stage *stage ) if (vbuf->render) vbuf->render->destroy( vbuf->render ); + if (vbuf->cache) + translate_cache_destroy(vbuf->cache); + FREE( stage ); } @@ -463,6 +466,11 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, 16 ); if (!vbuf->indices) goto fail; + + vbuf->cache = translate_cache_create(); + if (!vbuf->cache) + goto fail; + vbuf->vertices = NULL; vbuf->vertex_ptr = vbuf->vertices; diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 671abc25ce..e458cbe533 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -49,7 +49,6 @@ void draw_pt_emit_prepare( struct pt_emit *emit, const struct vertex_info *vinfo; unsigned dst_offset; struct translate_key hw_key; - unsigned keysize; unsigned i; boolean ok; @@ -62,7 +61,7 @@ void draw_pt_emit_prepare( struct pt_emit *emit, /* Must do this after set_primitive() above: */ vinfo = draw->render->get_vertex_info(draw->render); - keysize = 2*4 + vinfo->num_attribs * sizeof(hw_key.element[0]); + /* Translate from pipeline vertices to hw vertices. */ @@ -121,9 +120,9 @@ void draw_pt_emit_prepare( struct pt_emit *emit, hw_key.output_stride = vinfo->size * 4; if (!emit->translate || - memcmp(&emit->translate->key, &hw_key, keysize) != 0) + translate_key_compare(&emit->translate->key, &hw_key) != 0) { - memset((char *)&hw_key + keysize, 0, sizeof(hw_key) - keysize); + translate_key_sanitize(&hw_key); emit->translate = translate_cache_find(emit->cache, &hw_key); } } @@ -197,7 +196,8 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw ) void draw_pt_emit_destroy( struct pt_emit *emit ) { - translate_cache_destroy(emit->cache); + if (emit->cache) + translate_cache_destroy(emit->cache); FREE(emit); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index a5bebb4ca1..100117a9ae 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -62,11 +62,8 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, unsigned i, nr = 0; unsigned dst_offset = 0; struct translate_key key; - unsigned keysize; fetch->vertex_size = vertex_size; - keysize = (2*4 + - (draw->pt.nr_vertex_elements + 1) * sizeof(key.element[0])); /* Always emit/leave space for a vertex header. * @@ -111,9 +108,9 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, if (!fetch->translate || - memcmp(&fetch->translate->key, &key, keysize) != 0) + translate_key_compare(&fetch->translate->key, &key) != 0) { - memset((char *)&key + keysize, 0, sizeof(key) - keysize); + translate_key_sanitize(&key); fetch->translate = translate_cache_find(fetch->cache, &key); { diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index a4de341df8..b7b970a297 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -174,8 +174,9 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, /* Don't bother with caching at this stage: */ if (!feme->translate || - memcmp(&feme->translate->key, &key, sizeof(key)) != 0) + translate_key_compare(&feme->translate->key, &key) != 0) { + translate_key_sanitize(&key); feme->translate = translate_cache_find(feme->cache, &key); -- cgit v1.2.3 From 5ee2b5bdcc62e844079829f4f4301aad5374c62e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 9 May 2008 15:02:59 +0100 Subject: draw: fix translate double-free, minor cleanups --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 3 --- src/gallium/auxiliary/draw/draw_pt_fetch.c | 3 ++- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 3 ++- 3 files changed, 4 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 86a7d1c730..67b9a9503d 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -427,9 +427,6 @@ static void vbuf_destroy( struct draw_stage *stage ) if(vbuf->indices) align_free( vbuf->indices ); - if(vbuf->translate) - vbuf->translate->release( vbuf->translate ); - if (vbuf->render) vbuf->render->destroy( vbuf->render ); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 100117a9ae..b96335b789 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -184,7 +184,8 @@ struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) void draw_pt_fetch_destroy( struct pt_fetch *fetch ) { - translate_cache_destroy(fetch->cache); + if (fetch->cache) + translate_cache_destroy(fetch->cache); FREE(fetch); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index b7b970a297..4ea7d4359f 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -268,7 +268,8 @@ static void fetch_emit_destroy( struct draw_pt_middle_end *middle ) { struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; - translate_cache_destroy(feme->cache); + if (feme->cache) + translate_cache_destroy(feme->cache); FREE(middle); } -- cgit v1.2.3 From 501be9c7dd0cc5f985c708fa0e5f35d7fd20deb4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 9 May 2008 15:02:59 +0100 Subject: draw: fix translate double-free, minor cleanups --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 3 --- src/gallium/auxiliary/draw/draw_pt_fetch.c | 3 ++- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 3 ++- 3 files changed, 4 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 2a19e6916a..64a9f9084f 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -427,9 +427,6 @@ static void vbuf_destroy( struct draw_stage *stage ) if(vbuf->indices) align_free( vbuf->indices ); - if(vbuf->translate) - vbuf->translate->release( vbuf->translate ); - if (vbuf->render) vbuf->render->destroy( vbuf->render ); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index d62cd9358b..034e0eccb2 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -223,7 +223,8 @@ struct pt_fetch *draw_pt_fetch_create( struct draw_context *draw ) void draw_pt_fetch_destroy( struct pt_fetch *fetch ) { - translate_cache_destroy(fetch->cache); + if (fetch->cache) + translate_cache_destroy(fetch->cache); FREE(fetch); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 8df4241b82..bdbb039f9e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -320,7 +320,8 @@ static void fetch_emit_destroy( struct draw_pt_middle_end *middle ) { struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; - translate_cache_destroy(feme->cache); + if (feme->cache) + translate_cache_destroy(feme->cache); FREE(middle); } -- cgit v1.2.3 From a1f95a8bf64f863289b6759caeec76d7e054400e Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 9 May 2008 15:04:33 +0200 Subject: gallium: depth textures have usage depth_stencil instead of render_target --- src/mesa/state_tracker/st_cb_fbo.c | 6 ------ src/mesa/state_tracker/st_cb_texture.c | 17 ++++++++++++----- src/mesa/state_tracker/st_texture.h | 5 +++++ 3 files changed, 17 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 3e592730f8..3560a040c8 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -77,12 +77,6 @@ init_renderbuffer_bits(struct st_renderbuffer *strb, return info.size; } -static INLINE GLboolean pf_is_depth_stencil( enum pipe_format format ) -{ - return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_S )) != 0; -} - /** * gl_renderbuffer::AllocStorage() * This is called to allocate the original drawing surface, and diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index c0dba4cf2d..fb78c87989 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -272,6 +272,7 @@ guess_and_alloc_texture(struct st_context *st, GLuint height = stImage->base.Height2; GLuint depth = stImage->base.Depth2; GLuint i, comp_byte = 0; + enum pipe_format fmt; DBG("%s\n", __FUNCTION__); @@ -331,15 +332,18 @@ guess_and_alloc_texture(struct st_context *st, if (stImage->base.IsCompressed) comp_byte = compressed_num_bytes(stImage->base.TexFormat->MesaFormat); + fmt = st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat); stObj->pt = st_texture_create(st, gl_target_to_pipe(stObj->base.Target), - st_mesa_format_to_pipe_format(stImage->base.TexFormat->MesaFormat), + fmt, lastLevel, width, height, depth, comp_byte, - ( PIPE_TEXTURE_USAGE_RENDER_TARGET | + ( (pf_is_depth_stencil(fmt) ? + PIPE_TEXTURE_USAGE_DEPTH_STENCIL : + PIPE_TEXTURE_USAGE_RENDER_TARGET) | PIPE_TEXTURE_USAGE_SAMPLER )); DBG("%s - success\n", __FUNCTION__); @@ -1510,16 +1514,19 @@ st_finalize_texture(GLcontext *ctx, /* May need to create a new gallium texture: */ if (!stObj->pt) { + const enum pipe_format fmt = + st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat); stObj->pt = st_texture_create(ctx->st, gl_target_to_pipe(stObj->base.Target), - st_mesa_format_to_pipe_format(firstImage->base.TexFormat->MesaFormat), + fmt, stObj->lastLevel, firstImage->base.Width2, firstImage->base.Height2, firstImage->base.Depth2, comp_byte, - - ( PIPE_TEXTURE_USAGE_RENDER_TARGET | + ( (pf_is_depth_stencil(fmt) ? + PIPE_TEXTURE_USAGE_DEPTH_STENCIL : + PIPE_TEXTURE_USAGE_RENDER_TARGET) | PIPE_TEXTURE_USAGE_SAMPLER )); if (!stObj->pt) { diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h index 55d1a367b5..3febe6a7cb 100644 --- a/src/mesa/state_tracker/st_texture.h +++ b/src/mesa/state_tracker/st_texture.h @@ -93,6 +93,11 @@ st_get_stobj_texture(struct st_texture_object *stObj) return stObj ? stObj->pt : NULL; } +static INLINE GLboolean pf_is_depth_stencil( enum pipe_format format ) +{ + return (pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + + pf_get_component_bits( format, PIPE_FORMAT_COMP_S )) != 0; +} extern struct pipe_texture * -- cgit v1.2.3 From 7462f0557f9cce73ff2d32e62ef110b5d8622f87 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Fri, 9 May 2008 18:04:16 +0100 Subject: draw: Fix number of vertices allocated in draw_pt_emit(). --- src/gallium/auxiliary/draw/draw_pt_emit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index e458cbe533..ce3a153f64 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -146,7 +146,7 @@ void draw_pt_emit( struct pt_emit *emit, hw_verts = render->allocate_vertices(render, (ushort)translate->key.output_stride, - (ushort)count); + (ushort)vertex_count); if (!hw_verts) { assert(0); return; -- cgit v1.2.3 From 140b3f7f9cc682809170d7c311f89e0477dba5aa Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 10 May 2008 12:16:19 -0600 Subject: gallium: remove unused code --- src/gallium/drivers/softpipe/sp_quad_fs.c | 15 --------------- 1 file changed, 15 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 625d0f9b48..8c88c192f8 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -53,7 +53,6 @@ struct quad_shade_stage struct tgsi_sampler samplers[PIPE_MAX_SAMPLERS]; struct tgsi_exec_machine machine; struct tgsi_exec_vector *inputs, *outputs; - int colorOutSlot, depthOutSlot; }; @@ -156,20 +155,6 @@ static void shade_begin(struct quad_stage *qs) qss->samplers[i].texture = softpipe->texture[i]; } - /* find output slots for depth, color */ - qss->colorOutSlot = -1; - qss->depthOutSlot = -1; - for (i = 0; i < qss->stage.softpipe->fs->info.num_outputs; i++) { - switch (qss->stage.softpipe->fs->info.output_semantic_name[i]) { - case TGSI_SEMANTIC_POSITION: - qss->depthOutSlot = i; - break; - case TGSI_SEMANTIC_COLOR: - qss->colorOutSlot = i; - break; - } - } - softpipe->fs->prepare( softpipe->fs, &qss->machine, qss->samplers ); -- cgit v1.2.3 From 6807b4f6b1fa6ef0412714622ff16fe9d1487a8e Mon Sep 17 00:00:00 2001 From: Brian Date: Sat, 10 May 2008 12:46:00 -0600 Subject: gallium: optimize the flush_spans() function --- src/gallium/drivers/softpipe/sp_setup.c | 81 ++++++++++++++++----------------- 1 file changed, 40 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 5370d85275..543d86a5cb 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -208,78 +208,77 @@ static INLINE int block( int x ) } -/** - * Compute mask which indicates which pixels in the 2x2 quad are actually inside - * the triangle's bounds. - * - * this is pretty nasty... may need to rework flush_spans again to - * fix it, if possible. - */ -static unsigned calculate_mask( struct setup_context *setup, int x ) -{ - unsigned mask = 0x0; - - if (x >= setup->span.left[0] && x < setup->span.right[0]) - mask |= MASK_TOP_LEFT; - - if (x >= setup->span.left[1] && x < setup->span.right[1]) - mask |= MASK_BOTTOM_LEFT; - - if (x+1 >= setup->span.left[0] && x+1 < setup->span.right[0]) - mask |= MASK_TOP_RIGHT; - - if (x+1 >= setup->span.left[1] && x+1 < setup->span.right[1]) - mask |= MASK_BOTTOM_RIGHT; - - return mask; -} - - /** * Render a horizontal span of quads */ static void flush_spans( struct setup_context *setup ) { + const int xleft0 = setup->span.left[0]; + const int xleft1 = setup->span.left[1]; + const int xright0 = setup->span.right[0]; + const int xright1 = setup->span.right[1]; int minleft, maxright; int x; switch (setup->span.y_flags) { case 0x3: /* both odd and even lines written (both quad rows) */ - minleft = MIN2(setup->span.left[0], setup->span.left[1]); - maxright = MAX2(setup->span.right[0], setup->span.right[1]); + minleft = block(MIN2(xleft0, xleft1)); + maxright = block(MAX2(xright0, xright1)); + for (x = minleft; x <= maxright; x += 2) { + /* determine which of the four pixels is inside the span bounds */ + uint mask = 0x0; + if (x >= xleft0 && x < xright0) + mask |= MASK_TOP_LEFT; + if (x >= xleft1 && x < xright1) + mask |= MASK_BOTTOM_LEFT; + if (x+1 >= xleft0 && x+1 < xright0) + mask |= MASK_TOP_RIGHT; + if (x+1 >= xleft1 && x+1 < xright1) + mask |= MASK_BOTTOM_RIGHT; + emit_quad( setup, x, setup->span.y, mask ); + } break; case 0x1: /* only even line written (quad top row) */ - minleft = setup->span.left[0]; - maxright = setup->span.right[0]; + minleft = block(xleft0); + maxright = block(xright0); + for (x = minleft; x <= maxright; x += 2) { + uint mask = 0x0; + if (x >= xleft0 && x < xright0) + mask |= MASK_TOP_LEFT; + if (x+1 >= xleft0 && x+1 < xright0) + mask |= MASK_TOP_RIGHT; + emit_quad( setup, x, setup->span.y, mask ); + } break; case 0x2: /* only odd line written (quad bottom row) */ - minleft = setup->span.left[1]; - maxright = setup->span.right[1]; + minleft = block(xleft1); + maxright = block(xright1); + for (x = minleft; x <= maxright; x += 2) { + uint mask = 0x0; + if (x >= xleft1 && x < xright1) + mask |= MASK_BOTTOM_LEFT; + if (x+1 >= xleft1 && x+1 < xright1) + mask |= MASK_BOTTOM_RIGHT; + emit_quad( setup, x, setup->span.y, mask ); + } break; default: return; } - /* XXX this loop could be moved into the above switch cases and - * calculate_mask() could be simplified a bit... - */ - for (x = block(minleft); x <= block(maxright); x += 2) { - emit_quad( setup, x, setup->span.y, - calculate_mask( setup, x ) ); - } - setup->span.y = 0; setup->span.y_flags = 0; setup->span.right[0] = 0; setup->span.right[1] = 0; } + #if DEBUG_VERTS static void print_vertex(const struct setup_context *setup, const float (*v)[4]) -- cgit v1.2.3 From c0a6040f568e0c9be07797b2dc2fdd8a3624ec34 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 9 May 2008 13:09:58 +0100 Subject: translate: helper functions for mimizing cost of key compares --- src/gallium/auxiliary/translate/translate.h | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h index de6f09d18a..b8210af50c 100644 --- a/src/gallium/auxiliary/translate/translate.h +++ b/src/gallium/auxiliary/translate/translate.h @@ -95,6 +95,27 @@ struct translate *translate_lookup_or_create( struct translate_context *tctx, struct translate *translate_create( const struct translate_key *key ); +static INLINE int translate_keysize( const struct translate_key *key ) +{ + return 2 * sizeof(int) + key->nr_elements * sizeof(struct translate_element); +} + +static INLINE int translate_key_compare( const struct translate_key *a, + const struct translate_key *b ) +{ + int keysize = translate_keysize(a); + return memcmp(a, b, keysize); +} + + +static INLINE void translate_key_sanitize( struct translate_key *a ) +{ + int keysize = translate_keysize(a); + char *ptr = (char *)a; + memset(ptr + keysize, 0, sizeof(*a) - keysize); +} + + /******************************************************************************* * Private: */ -- cgit v1.2.3 From 7ddb925b8bc6c18eba953e34d2b630a3a6593f05 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 9 May 2008 13:10:15 +0100 Subject: draw: mimize cost of translate key compares, use cache universally --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 20 ++++++++++++++------ src/gallium/auxiliary/draw/draw_pt_emit.c | 10 +++++----- src/gallium/auxiliary/draw/draw_pt_fetch.c | 7 ++----- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 3 ++- 4 files changed, 23 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 64a9f9084f..67b9a9503d 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -42,6 +42,7 @@ #include "draw_vertex.h" #include "draw_pipe.h" #include "translate/translate.h" +#include "translate/translate_cache.h" /** @@ -75,6 +76,8 @@ struct vbuf_stage { /* Cache point size somewhere it's address won't change: */ float point_size; + + struct translate_cache *cache; }; @@ -220,7 +223,6 @@ vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) /* Translate from pipeline vertices to hw vertices. */ dst_offset = 0; - memset(&hw_key, 0, sizeof(hw_key)); for (i = 0; i < vbuf->vinfo->num_attribs; i++) { unsigned emit_sz = 0; @@ -277,12 +279,10 @@ vbuf_set_prim( struct vbuf_stage *vbuf, uint prim ) /* Don't bother with caching at this stage: */ if (!vbuf->translate || - memcmp(&vbuf->translate->key, &hw_key, sizeof(hw_key)) != 0) + translate_key_compare(&vbuf->translate->key, &hw_key) != 0) { - if (vbuf->translate) - vbuf->translate->release(vbuf->translate); - - vbuf->translate = translate_create( &hw_key ); + translate_key_sanitize(&hw_key); + vbuf->translate = translate_cache_find(vbuf->cache, &hw_key); vbuf->translate->set_buffer(vbuf->translate, 1, &vbuf->point_size, 0); } @@ -430,6 +430,9 @@ static void vbuf_destroy( struct draw_stage *stage ) if (vbuf->render) vbuf->render->destroy( vbuf->render ); + if (vbuf->cache) + translate_cache_destroy(vbuf->cache); + FREE( stage ); } @@ -460,6 +463,11 @@ struct draw_stage *draw_vbuf_stage( struct draw_context *draw, 16 ); if (!vbuf->indices) goto fail; + + vbuf->cache = translate_cache_create(); + if (!vbuf->cache) + goto fail; + vbuf->vertices = NULL; vbuf->vertex_ptr = vbuf->vertices; diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 999b2007a2..4a854f4362 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -49,7 +49,6 @@ void draw_pt_emit_prepare( struct pt_emit *emit, const struct vertex_info *vinfo; unsigned dst_offset; struct translate_key hw_key; - unsigned keysize; unsigned i; boolean ok; @@ -62,7 +61,7 @@ void draw_pt_emit_prepare( struct pt_emit *emit, /* Must do this after set_primitive() above: */ vinfo = draw->render->get_vertex_info(draw->render); - keysize = 2*4 + vinfo->num_attribs * sizeof(hw_key.element[0]); + /* Translate from pipeline vertices to hw vertices. */ @@ -121,9 +120,9 @@ void draw_pt_emit_prepare( struct pt_emit *emit, hw_key.output_stride = vinfo->size * 4; if (!emit->translate || - memcmp(&emit->translate->key, &hw_key, keysize) != 0) + translate_key_compare(&emit->translate->key, &hw_key) != 0) { - memset((char *)&hw_key + keysize, 0, sizeof(hw_key) - keysize); + translate_key_sanitize(&hw_key); emit->translate = translate_cache_find(emit->cache, &hw_key); } } @@ -244,7 +243,8 @@ struct pt_emit *draw_pt_emit_create( struct draw_context *draw ) void draw_pt_emit_destroy( struct pt_emit *emit ) { - translate_cache_destroy(emit->cache); + if (emit->cache) + translate_cache_destroy(emit->cache); FREE(emit); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 034e0eccb2..07f4c99164 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -62,11 +62,8 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, unsigned i, nr = 0; unsigned dst_offset = 0; struct translate_key key; - unsigned keysize; fetch->vertex_size = vertex_size; - keysize = (2*4 + - (draw->pt.nr_vertex_elements + 1) * sizeof(key.element[0])); /* Always emit/leave space for a vertex header. * @@ -111,9 +108,9 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, if (!fetch->translate || - memcmp(&fetch->translate->key, &key, keysize) != 0) + translate_key_compare(&fetch->translate->key, &key) != 0) { - memset((char *)&key + keysize, 0, sizeof(key) - keysize); + translate_key_sanitize(&key); fetch->translate = translate_cache_find(fetch->cache, &key); { diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index bdbb039f9e..a1d041a74f 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -174,8 +174,9 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, /* Don't bother with caching at this stage: */ if (!feme->translate || - memcmp(&feme->translate->key, &key, sizeof(key)) != 0) + translate_key_compare(&feme->translate->key, &key) != 0) { + translate_key_sanitize(&key); feme->translate = translate_cache_find(feme->cache, &key); -- cgit v1.2.3 From 2258f6b437705860912be300e728efbde60e2140 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 12 May 2008 14:09:50 +0100 Subject: xlib: add failure paths for context creation --- src/gallium/winsys/xlib/xm_api.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c index 26b722f343..8a32c54349 100644 --- a/src/gallium/winsys/xlib/xm_api.c +++ b/src/gallium/winsys/xlib/xm_api.c @@ -797,8 +797,14 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) pipe = xmesa_create_i965simple(xmesa_get_pipe_winsys_aub(v)); } + if (pipe == NULL) + goto fail; + c->st = st_create_context(pipe, &v->mesa_visual, share_list ? share_list->st : NULL); + if (c->st == NULL) + goto fail; + mesaCtx = c->st->ctx; c->st->ctx->DriverCtx = c; @@ -818,6 +824,14 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) #endif return c; + + fail: + if (c->st) + st_destroy_context(c->st); + if (pipe) + pipe->destroy(pipe); + FREE(c); + return NULL; } -- cgit v1.2.3 From de818835de70961602bb9ceca86b98e9bbc63fc1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 12 May 2008 14:10:03 +0100 Subject: softpipe: add failure paths for context creation --- src/gallium/drivers/softpipe/sp_context.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index fe9cd8375e..a48e546139 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -88,7 +88,8 @@ static void softpipe_destroy( struct pipe_context *pipe ) struct pipe_winsys *ws = pipe->winsys; uint i; - draw_destroy( softpipe->draw ); + if (softpipe->draw) + draw_destroy( softpipe->draw ); softpipe->quad.polygon_stipple->destroy( softpipe->quad.polygon_stipple ); softpipe->quad.earlyz->destroy( softpipe->quad.earlyz ); @@ -216,8 +217,12 @@ softpipe_create( struct pipe_screen *screen, * Create drawing context and plug our rendering stage into it. */ softpipe->draw = draw_create(); - assert(softpipe->draw); + if (!softpipe->draw) + goto fail; + softpipe->setup = sp_draw_render_stage(softpipe); + if (!softpipe->setup) + goto fail; if (GETENV( "SP_NO_RAST" ) != NULL) softpipe->no_rast = TRUE; @@ -241,4 +246,8 @@ softpipe_create( struct pipe_screen *screen, sp_init_surface_functions(softpipe); return &softpipe->pipe; + + fail: + softpipe_destroy(&softpipe->pipe); + return NULL; } -- cgit v1.2.3 From b5e5369da5fc50d63a6ece931fac44b555eb0314 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 12 May 2008 15:20:38 +0100 Subject: draw: add fetch-shade-emit path Enable with TEST_FSE=t. Performs fetch from API-provided vertex buffers, transformation with one of three (two working) hard-coded shaders, and final emit to hardware vertices all in a single pass. Currently only really useful for profiling in conjunction with SP_NO_RAST=t. --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/draw_private.h | 3 + src/gallium/auxiliary/draw/draw_pt.c | 25 +- src/gallium/auxiliary/draw/draw_pt.h | 3 + .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 714 +++++++++++++++++++++ 5 files changed, 742 insertions(+), 4 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index da7eded21f..68e7744cc5 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -26,6 +26,7 @@ C_SOURCES = \ draw_pt_emit.c \ draw_pt_fetch.c \ draw_pt_fetch_emit.c \ + draw_pt_fetch_shade_emit.c \ draw_pt_fetch_shade_pipeline.c \ draw_pt_post_vs.c \ draw_pt_varray.c \ diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index e036d498b8..cbe64cd290 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -124,12 +124,14 @@ struct draw_context struct { struct { struct draw_pt_middle_end *fetch_emit; + /*struct draw_pt_middle_end *fetch_shade_emit;*/ struct draw_pt_middle_end *general; } middle; struct { struct draw_pt_front_end *vcache; struct draw_pt_front_end *varray; + struct draw_pt_front_end *fetch_shade_emit; /* temp hack */ } front; struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; @@ -154,6 +156,7 @@ struct draw_context const void *constants; } user; + boolean test_fse; } pt; struct { diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index bccde6c5fd..448deef98c 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -64,7 +64,7 @@ draw_pt_arrays(struct draw_context *draw, opt |= PT_PIPELINE; } - if (!draw->bypass_clipping) { + if (!draw->bypass_clipping && !draw->pt.test_fse) { opt |= PT_CLIPTEST; } @@ -72,16 +72,20 @@ draw_pt_arrays(struct draw_context *draw, opt |= PT_SHADE; } - if (opt) - middle = draw->pt.middle.general; - else + + if (opt == 0) middle = draw->pt.middle.fetch_emit; + else + middle = draw->pt.middle.general; /* Pick the right frontend */ if (draw->pt.user.elts) { frontend = draw->pt.front.vcache; + } else if (opt == PT_SHADE && draw->pt.test_fse) { + /* should be a middle end.. */ + frontend = draw->pt.front.fetch_shade_emit; } else { frontend = draw->pt.front.varray; } @@ -113,6 +117,14 @@ boolean draw_pt_init( struct draw_context *draw ) if (!draw->pt.middle.fetch_emit) return FALSE; + draw->pt.test_fse = GETENV("DRAW_FSE") != NULL; + if (draw->pt.test_fse) { + draw->pt.front.fetch_shade_emit = draw_pt_fetch_shade_emit( draw ); + if (!draw->pt.front.fetch_shade_emit) + return FALSE; + } + + draw->pt.middle.general = draw_pt_fetch_pipeline_or_emit( draw ); if (!draw->pt.middle.general) return FALSE; @@ -133,6 +145,11 @@ void draw_pt_destroy( struct draw_context *draw ) draw->pt.middle.fetch_emit = NULL; } + if (draw->pt.front.fetch_shade_emit) { + draw->pt.front.fetch_shade_emit->destroy( draw->pt.front.fetch_shade_emit ); + draw->pt.front.fetch_shade_emit = NULL; + } + if (draw->pt.front.vcache) { draw->pt.front.vcache->destroy( draw->pt.front.vcache ); draw->pt.front.vcache = NULL; diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 312fdbe4f4..bcd89f6bd6 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -121,6 +121,8 @@ const void *draw_pt_elt_ptr( struct draw_context *draw, struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ); struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw); +struct draw_pt_front_end *draw_pt_fetch_shade_emit( struct draw_context *draw ); + /* Middle-ends: * * Currently one general-purpose case which can do all possibilities, @@ -132,6 +134,7 @@ struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw); * vertex_elements. */ struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); +//struct draw_pt_middle_end *draw_pt_fetch_shade_emit( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c new file mode 100644 index 0000000000..9e1d1add36 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -0,0 +1,714 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" +#include "draw/draw_vs.h" + +#include "translate/translate.h" + +struct fetch_shade_emit; + +struct fse_shader { + struct translate_key key; + + void (*run_linear)( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ); +}; + +/* Prototype fetch, shade, emit-hw-verts all in one go. + */ +struct fetch_shade_emit { + struct draw_pt_front_end base; + + struct draw_context *draw; + + struct translate_key key; + + /* Temporaries: + */ + const float *constants; + unsigned pitch[PIPE_MAX_ATTRIBS]; + const ubyte *src[PIPE_MAX_ATTRIBS]; + unsigned prim; + + /* Points to one of the three hardwired example shaders, below: + */ + struct fse_shader *active; + + /* Temporary: A list of hard-wired shaders. Of course the plan + * would be to generate these for a given (vertex-shader, + * translate-key) pair... + */ + struct fse_shader shader[10]; + int nr_shaders; +}; + + + +/* Not quite passthrough yet -- we're still running the 'shader' here, + * inlined into the vertex fetch function. + */ +static void fetch_xyz_rgb_st( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + + const float *m = fse->constants; + const float m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12]; + const float m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13]; + const float m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14]; + const float m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15]; + + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *st = fse->src[2] + start * fse->pitch[2]; + + float *out = (float *)buffer; + + + assert(fse->pitch[1] == 0); + + /* loop over vertex attributes (vertex shader inputs) + */ + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m0 * ix + m4 * iy + m8 * iz + m12; + out[1] = m1 * ix + m5 * iy + m9 * iz + m13; + out[2] = m2 * ix + m6 * iy + m10 * iz + m14; + out[3] = m3 * ix + m7 * iy + m11 * iz + m15; + xyz += fse->pitch[0]; + } + + { + out[4] = 1.0f; + out[5] = 1.0f; + out[6] = 1.0f; + out[7] = 1.0f; + } + + { + const float *in = (const float *)st; st += fse->pitch[2]; + out[8] = in[0]; + out[9] = in[1]; + out[10] = 0.0f; + out[11] = 1.0f; + } + + out += 12; + } +} + + + +static void fetch_xyz_rgb( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + + const float *m = (const float *)fse->constants; + const float m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12]; + const float m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13]; + const float m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14]; + const float m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15]; + + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; + + float *out = (float *)buffer; + +// debug_printf("rgb %f %f %f\n", rgb[0], rgb[1], rgb[2]); + + + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m0 * ix + m4 * iy + m8 * iz + m12; + out[1] = m1 * ix + m5 * iy + m9 * iz + m13; + out[2] = m2 * ix + m6 * iy + m10 * iz + m14; + out[3] = m3 * ix + m7 * iy + m11 * iz + m15; + xyz += fse->pitch[0]; + } + + { + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + out += 8; + } +} + + + + +static void fetch_xyz_rgb_psiz( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + + const float *m = (const float *)fse->constants; + const float m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12]; + const float m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13]; + const float m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14]; + const float m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15]; + + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const float *rgb = (const float *)(fse->src[1] + start * fse->pitch[1]); + const float psiz = 1.0; + + float *out = (float *)buffer; + + + assert(fse->pitch[1] == 0); + + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m0 * ix + m4 * iy + m8 * iz + m12; + out[1] = m1 * ix + m5 * iy + m9 * iz + m13; + out[2] = m2 * ix + m6 * iy + m10 * iz + m14; + out[3] = m3 * ix + m7 * iy + m11 * iz + m15; + xyz += fse->pitch[0]; + } + + { + out[4] = rgb[0]; + out[5] = rgb[1]; + out[6] = rgb[2]; + out[7] = 1.0f; + } + + { + out[8] = psiz; + } + + out += 9; + } +} + + + + +static boolean set_prim( struct fetch_shade_emit *fse, + unsigned prim, + unsigned count ) +{ + struct draw_context *draw = fse->draw; + + fse->prim = prim; + + switch (prim) { + case PIPE_PRIM_LINE_LOOP: + if (count > 1024) + return FALSE; + draw->render->set_primitive( draw->render, PIPE_PRIM_LINE_STRIP ); + break; + + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + if (count > 1024) + return FALSE; + draw->render->set_primitive( draw->render, prim ); + break; + + case PIPE_PRIM_QUADS: + case PIPE_PRIM_QUAD_STRIP: + draw->render->set_primitive( draw->render, PIPE_PRIM_TRIANGLES ); + break; + + default: + draw->render->set_primitive( draw->render, prim ); + break; + } + + return TRUE; +} + + + + + + +static void fse_prepare( struct draw_pt_front_end *fe, + unsigned prim, + struct draw_pt_middle_end *unused, + unsigned opt ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)fe; + struct draw_context *draw = fse->draw; + unsigned num_vs_inputs = draw->vertex_shader->info.num_inputs; + unsigned num_vs_outputs = draw->vertex_shader->info.num_outputs; + const struct vertex_info *vinfo; + unsigned i; + boolean need_psize = 0; + + + if (draw->pt.user.elts) { + assert(0); + return ; + } + + if (!set_prim(fse, prim, /*count*/1022 )) { + assert(0); + return ; + } + + /* Must do this after set_primitive() above: + */ + vinfo = draw->render->get_vertex_info(draw->render); + + + + fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ + num_vs_inputs); /* inputs - fetch from api format */ + + fse->key.output_stride = vinfo->size * 4; + memset(fse->key.element, 0, + fse->key.nr_elements * sizeof(fse->key.element[0])); + + for (i = 0; i < num_vs_inputs; i++) { + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].input_format = src->src_format; + + /* Consider ignoring these at this point, ie make generated + * programs independent of this state: + */ + fse->key.element[i].input_buffer = 0; //src->vertex_buffer_index; + fse->key.element[i].input_offset = 0; //src->src_offset; + } + + + { + unsigned dst_offset = 0; + + for (i = 0; i < vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned output_format = PIPE_FORMAT_NONE; + unsigned vs_output = vinfo->src_index[i]; + + switch (vinfo->emit[i]) { + case EMIT_4F: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + need_psize = 1; + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + vs_output = num_vs_outputs + 1; + + break; + default: + assert(0); + break; + } + + /* The elements in the key correspond to vertex shader output + * numbers, not to positions in the hw vertex description -- + * that's handled by the output_offset field. + */ + fse->key.element[vs_output].output_format = output_format; + fse->key.element[vs_output].output_offset = dst_offset; + + dst_offset += emit_sz; + assert(fse->key.output_stride >= dst_offset); + } + } + + /* To make psize work, really need to tell the vertex shader to + * copy that value from input->output. For 'translate' this was + * implicit for all elements. + */ +#if 0 + if (need_psize) { + unsigned input = num_vs_inputs + 1; + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].input_format = PIPE_FORMAT_R32_FLOAT; + fse->key.element[i].input_buffer = 0; //nr_buffers + 1; + fse->key.element[i].input_offset = 0; + + fse->key.nr_elements += 1; + + } +#endif + + fse->constants = draw->pt.user.constants; + + /* Would normally look up a vertex shader and peruse its list of + * varients somehow. We omitted that step and put all the + * hardcoded "shaders" into an array. We're just making the + * assumption that this happens to be a matching shader... ie + * you're running isosurf, aren't you? + */ + fse->active = NULL; + for (i = 0; i < fse->nr_shaders; i++) { + if (translate_key_compare( &fse->key, &fse->shader[i].key) == 0) + fse->active = &fse->shader[i]; + } + + if (!fse->active) { + assert(0); + return ; + } + + /* Now set buffer pointers: + */ + for (i = 0; i < num_vs_inputs; i++) { + unsigned buf = draw->pt.vertex_element[i].vertex_buffer_index; + + fse->src[i] = ((const ubyte *) draw->pt.user.vbuffer[buf] + + draw->pt.vertex_buffer[buf].buffer_offset + + draw->pt.vertex_element[i].src_offset); + + fse->pitch[i] = draw->pt.vertex_buffer[buf].pitch; + + } + + + //return TRUE; +} + + +static boolean split_prim_inplace(unsigned prim, unsigned *first, unsigned *incr) +{ + switch (prim) { + case PIPE_PRIM_POINTS: + *first = 1; + *incr = 1; + return TRUE; + case PIPE_PRIM_LINES: + *first = 2; + *incr = 2; + return TRUE; + case PIPE_PRIM_LINE_STRIP: + *first = 2; + *incr = 1; + return TRUE; + case PIPE_PRIM_TRIANGLES: + *first = 3; + *incr = 3; + return TRUE; + case PIPE_PRIM_TRIANGLE_STRIP: + *first = 3; + *incr = 1; + return TRUE; + case PIPE_PRIM_QUADS: + *first = 4; + *incr = 4; + return TRUE; + case PIPE_PRIM_QUAD_STRIP: + *first = 4; + *incr = 2; + return TRUE; + default: + *first = 0; + *incr = 1; /* set to one so that count % incr works */ + return FALSE; + } +} + + + + +#define INDEX(i) (start + (i)) +static void fse_render_linear( struct vbuf_render *render, + unsigned prim, + unsigned start, + unsigned length ) +{ + ushort *tmp = NULL; + unsigned i, j; + + switch (prim) { + case PIPE_PRIM_LINE_LOOP: + tmp = MALLOC( sizeof(ushort) * (length + 1) ); + + for (i = 0; i < length; i++) + tmp[i] = INDEX(i); + tmp[length] = 0; + + render->draw( render, + tmp, + length+1 ); + break; + + + case PIPE_PRIM_QUAD_STRIP: + tmp = MALLOC( sizeof(ushort) * (length / 2 * 6) ); + + for (j = i = 0; i + 3 < length; i += 2, j += 6) { + tmp[j+0] = INDEX(i+0); + tmp[j+1] = INDEX(i+1); + tmp[j+2] = INDEX(i+3); + + tmp[j+3] = INDEX(i+2); + tmp[j+4] = INDEX(i+0); + tmp[j+5] = INDEX(i+3); + } + + if (j) + render->draw( render, tmp, j ); + break; + + case PIPE_PRIM_QUADS: + tmp = MALLOC( sizeof(int) * (length / 4 * 6) ); + + for (j = i = 0; i + 3 < length; i += 4, j += 6) { + tmp[j+0] = INDEX(i+0); + tmp[j+1] = INDEX(i+1); + tmp[j+2] = INDEX(i+3); + + tmp[j+3] = INDEX(i+1); + tmp[j+4] = INDEX(i+2); + tmp[j+5] = INDEX(i+3); + } + + if (j) + render->draw( render, tmp, j ); + break; + + default: + render->draw_arrays( render, + start, + length ); + break; + } + + if (tmp) + FREE(tmp); +} + + + +static boolean do_draw( struct fetch_shade_emit *fse, + unsigned start, unsigned count ) +{ + struct draw_context *draw = fse->draw; + + char *hw_verts = + draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)count ); + + if (!hw_verts) + return FALSE; + + /* Single routine to fetch vertices, run shader and emit HW verts. + * Clipping and viewport transformation are done on hardware. + */ + fse->active->run_linear( fse, + start, count, + hw_verts ); + + /* Draw arrays path to avoid re-emitting index list again and + * again. + */ + fse_render_linear( draw->render, + fse->prim, + 0, + count ); + + + draw->render->release_vertices( draw->render, + hw_verts, + fse->key.output_stride, + count ); + + return TRUE; +} + + +static void +fse_run(struct draw_pt_front_end *fe, + pt_elt_func elt_func, + const void *elt_ptr, + unsigned count) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)fe; + unsigned i = 0; + unsigned first, incr; + unsigned start = elt_func(elt_ptr, 0); + + //debug_printf("%s prim %d start %d count %d\n", __FUNCTION__, prim, start, count); + + split_prim_inplace(fse->prim, &first, &incr); + + count -= (count - first) % incr; + + while (i + first <= count) { + int nr = MIN2( count - i, 1024 ); + + /* snap to prim boundary + */ + nr -= (nr - first) % incr; + + if (!do_draw( fse, start + i, nr )) { + assert(0); + return ; + } + + /* increment allowing for repeated vertices + */ + i += nr - (first - incr); + } + + //return TRUE; +} + + +static void fse_finish( struct draw_pt_front_end *frontend ) +{ +} + + +static void +fse_destroy( struct draw_pt_front_end *frontend ) +{ + FREE(frontend); +} + +struct draw_pt_front_end *draw_pt_fetch_shade_emit( struct draw_context *draw ) +{ + struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit); + if (!fse) + return NULL; + + fse->base.prepare = fse_prepare; + fse->base.run = fse_run; + fse->base.finish = fse_finish; + fse->base.destroy = fse_destroy; + fse->draw = draw; + + fse->shader[0].run_linear = fetch_xyz_rgb_st; + fse->shader[0].key.nr_elements = 3; + fse->shader[0].key.output_stride = 12 * sizeof(float); + + fse->shader[0].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[0].key.element[0].input_buffer = 0; + fse->shader[0].key.element[0].input_offset = 0; + fse->shader[0].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[0].output_offset = 0; + + fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[0].key.element[1].input_buffer = 0; + fse->shader[0].key.element[1].input_offset = 0; + fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[1].output_offset = 16; + + fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32_FLOAT; + fse->shader[0].key.element[1].input_buffer = 0; + fse->shader[0].key.element[1].input_offset = 0; + fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[1].output_offset = 32; + + fse->shader[1].run_linear = fetch_xyz_rgb; + fse->shader[1].key.nr_elements = 2; + fse->shader[1].key.output_stride = 8 * sizeof(float); + + fse->shader[1].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[1].key.element[0].input_buffer = 0; + fse->shader[1].key.element[0].input_offset = 0; + fse->shader[1].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[1].key.element[0].output_offset = 0; + + fse->shader[1].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[1].key.element[1].input_buffer = 0; + fse->shader[1].key.element[1].input_offset = 0; + fse->shader[1].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[1].key.element[1].output_offset = 16; + + fse->shader[2].run_linear = fetch_xyz_rgb_psiz; + fse->shader[2].key.nr_elements = 3; + fse->shader[2].key.output_stride = 9 * sizeof(float); + + fse->shader[2].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[2].key.element[0].input_buffer = 0; + fse->shader[2].key.element[0].input_offset = 0; + fse->shader[2].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[2].key.element[0].output_offset = 0; + + fse->shader[2].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[2].key.element[1].input_buffer = 0; + fse->shader[2].key.element[1].input_offset = 0; + fse->shader[2].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[2].key.element[1].output_offset = 16; + + /* psize is special + * -- effectively add it here as another input!?! + * -- who knows how to add it as a buffer? + */ + fse->shader[2].key.element[2].input_format = PIPE_FORMAT_R32_FLOAT; + fse->shader[2].key.element[2].input_buffer = 0; + fse->shader[2].key.element[2].input_offset = 0; + fse->shader[2].key.element[2].output_format = PIPE_FORMAT_R32_FLOAT; + fse->shader[2].key.element[2].output_offset = 32; + + fse->nr_shaders = 3; + + return &fse->base; +} -- cgit v1.2.3 From 90e86363de7dbcfda3490b5c31d701350a0fa2ef Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 12 May 2008 16:16:04 +0100 Subject: softpipe: make vbuf handle all primitive types --- src/gallium/drivers/softpipe/sp_prim_setup.c | 15 +- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 255 +++++++++++++++------------ 2 files changed, 155 insertions(+), 115 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 1cf9ffa632..941ab62e00 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -64,16 +64,17 @@ static INLINE struct setup_stage *setup_stage( struct draw_stage *stage ) } +typedef const float (*cptrf4)[4]; static void do_tri(struct draw_stage *stage, struct prim_header *prim) { struct setup_stage *setup = setup_stage( stage ); - + setup_tri( setup->setup, - prim->v[0]->data, - prim->v[1]->data, - prim->v[2]->data ); + (cptrf4)prim->v[0]->data, + (cptrf4)prim->v[1]->data, + (cptrf4)prim->v[2]->data ); } static void @@ -82,8 +83,8 @@ do_line(struct draw_stage *stage, struct prim_header *prim) struct setup_stage *setup = setup_stage( stage ); setup_line( setup->setup, - prim->v[0]->data, - prim->v[1]->data ); + (cptrf4)prim->v[0]->data, + (cptrf4)prim->v[1]->data ); } static void @@ -92,7 +93,7 @@ do_point(struct draw_stage *stage, struct prim_header *prim) struct setup_stage *setup = setup_stage( stage ); setup_point( setup->setup, - prim->v[0]->data ); + (cptrf4)prim->v[0]->data ); } diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index e063fe82ef..1399776ff0 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -129,17 +129,22 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) } +static INLINE cptrf4 get_vert( const void *vertex_buffer, + int index, + int stride ) +{ + return (cptrf4)((char *)vertex_buffer + index * stride); +} static void -sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) +sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) { struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; - unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); - unsigned i, j; - void *vertex_buffer = cvbr->vertex_buffer; - cptrf4 v[3]; + unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float); + unsigned i; + const void *vertex_buffer = cvbr->vertex_buffer; /* XXX: break this dependency - make setup_context live under * softpipe, rename the old "setup" draw stage to something else. @@ -149,40 +154,98 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr_indices) switch (cvbr->prim) { - case PIPE_PRIM_TRIANGLES: - for (i = 0; i < nr_indices; i += 3) { - for (j = 0; j < 3; j++) - v[j] = (cptrf4)((char *)vertex_buffer + - indices[i+j] * vertex_size); - - setup_tri( setup_ctx, - v[0], - v[1], - v[2]); + case PIPE_PRIM_POINTS: + for (i = 0; i < nr; i++) { + setup_point( setup_ctx, + get_vert(vertex_buffer, indices[i-0], stride) ); } break; case PIPE_PRIM_LINES: - for (i = 0; i < nr_indices; i += 2) { - for (j = 0; j < 2; j++) - v[j] = (cptrf4)((char *)vertex_buffer + - indices[i+j] * vertex_size); + for (i = 1; i < nr; i += 2) { + setup_line( setup_ctx, + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + } + break; + case PIPE_PRIM_LINE_STRIP: + for (i = 1; i < nr; i ++) { setup_line( setup_ctx, - v[0], - v[1] ); + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); } break; - case PIPE_PRIM_POINTS: - for (i = 0; i < nr_indices; i++) { - v[0] = (cptrf4)((char *)vertex_buffer + - indices[i] * vertex_size); + case PIPE_PRIM_LINE_LOOP: + for (i = 1; i < nr; i ++) { + setup_line( setup_ctx, + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride) ); + } + if (nr) { + setup_line( setup_ctx, + get_vert(vertex_buffer, indices[nr-1], stride), + get_vert(vertex_buffer, indices[0], stride) ); + } + break; - setup_point( setup_ctx, - v[0] ); + + case PIPE_PRIM_TRIANGLES: + for (i = 2; i < nr; i += 3) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + for (i = 2; i < nr; i += 3) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i+(i&1)-2], stride), + get_vert(vertex_buffer, indices[i-(i&1)-1], stride), + get_vert(vertex_buffer, indices[i-0], stride)); } break; + + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + for (i = 2; i < nr; i += 3) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[0], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + } + break; + case PIPE_PRIM_QUADS: + for (i = 3; i < nr; i += 4) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-1], stride)); + + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + } + break; + case PIPE_PRIM_QUAD_STRIP: + for (i = 3; i < nr; i += 2) { + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-2], stride), + get_vert(vertex_buffer, indices[i-1], stride), + get_vert(vertex_buffer, indices[i-0], stride)); + + setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-0], stride), + get_vert(vertex_buffer, indices[i-2], stride)); + } + break; + default: + assert(0); } /* XXX: why are we calling this??? If we had to call something, it @@ -203,130 +266,106 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) struct softpipe_context *softpipe = cvbr->softpipe; struct draw_stage *setup = softpipe->setup; const void *vertex_buffer = cvbr->vertex_buffer; - const unsigned vertex_size = softpipe->vertex_info_vbuf.size * sizeof(float); + const unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float); unsigned i; struct setup_context *setup_ctx = sp_draw_setup_context(setup); - cptrf4 v[3]; - -#define VERTEX(I) \ - (cptrf4) ((char *) vertex_buffer + (I) * vertex_size) switch (cvbr->prim) { case PIPE_PRIM_POINTS: for (i = 0; i < nr; i++) { - v[0] = VERTEX(i); - setup_point( setup_ctx, v[0] ); + setup_point( setup_ctx, + get_vert(vertex_buffer, i-0, stride) ); } break; + case PIPE_PRIM_LINES: - assert(nr % 2 == 0); - for (i = 0; i < nr; i += 2) { - v[0] = VERTEX(i); - v[1] = VERTEX(i + 1); - setup_line( setup_ctx, v[0], v[1] ); + for (i = 1; i < nr; i += 2) { + setup_line( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride) ); } break; + case PIPE_PRIM_LINE_STRIP: - for (i = 1; i < nr; i++) { - v[0] = VERTEX(i - 1); - v[1] = VERTEX(i); - setup_line( setup_ctx, v[0], v[1] ); + for (i = 1; i < nr; i ++) { + setup_line( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride) ); } break; + + case PIPE_PRIM_LINE_LOOP: + for (i = 1; i < nr; i ++) { + setup_line( setup_ctx, + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride) ); + } + if (nr) { + setup_line( setup_ctx, + get_vert(vertex_buffer, nr-1, stride), + get_vert(vertex_buffer, 0, stride) ); + } + break; + + case PIPE_PRIM_TRIANGLES: - assert(nr % 3 == 0); - for (i = 0; i < nr; i += 3) { - v[0] = VERTEX(i + 0); - v[1] = VERTEX(i + 1); - v[2] = VERTEX(i + 2); + for (i = 2; i < nr; i += 3) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride)); } break; + case PIPE_PRIM_TRIANGLE_STRIP: - assert(nr >= 3); - for (i = 2; i < nr; i++) { - v[0] = VERTEX(i - 2); - v[1] = VERTEX(i - 1); - v[2] = VERTEX(i); + for (i = 2; i < nr; i += 3) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i+(i&1)-2, stride), + get_vert(vertex_buffer, i-(i&1)-1, stride), + get_vert(vertex_buffer, i-0, stride)); } break; + case PIPE_PRIM_TRIANGLE_FAN: - assert(nr >= 3); - for (i = 2; i < nr; i++) { - v[0] = VERTEX(0); - v[1] = VERTEX(i - 1); - v[2] = VERTEX(i); + case PIPE_PRIM_POLYGON: + for (i = 2; i < nr; i += 3) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, 0, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride)); } break; case PIPE_PRIM_QUADS: assert(nr % 4 == 0); - for (i = 0; i < nr; i += 4) { - v[0] = VERTEX(i + 0); - v[1] = VERTEX(i + 1); - v[2] = VERTEX(i + 2); + for (i = 3; i < nr; i += 4) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-1, stride)); - v[0] = VERTEX(i + 0); - v[1] = VERTEX(i + 2); - v[2] = VERTEX(i + 3); setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride)); } break; case PIPE_PRIM_QUAD_STRIP: assert(nr >= 4); - for (i = 2; i < nr; i += 2) { - v[0] = VERTEX(i - 2); - v[1] = VERTEX(i); - v[2] = VERTEX(i + 1); + for (i = 3; i < nr; i += 2) { setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-2, stride), + get_vert(vertex_buffer, i-1, stride), + get_vert(vertex_buffer, i-0, stride)); - v[0] = VERTEX(i - 2); - v[1] = VERTEX(i + 1); - v[2] = VERTEX(i - 1); - setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); - } - break; - case PIPE_PRIM_POLYGON: - /* draw as tri fan */ - for (i = 2; i < nr; i++) { - v[0] = VERTEX(0); - v[1] = VERTEX(i - 1); - v[2] = VERTEX(i); setup_tri( setup_ctx, - v[0], - v[1], - v[2] ); + get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-0, stride), + get_vert(vertex_buffer, i-2, stride)); } break; default: - /* XXX finish remaining prim types */ assert(0); } - -#undef VERTEX } -- cgit v1.2.3 From f116a149160d50d43a23b02a3416725d6f895d51 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 12 May 2008 17:30:05 +0100 Subject: softpipe: more work to get non-reduced primitives working in vbuf --- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 45 ++++++++++++----------------- 1 file changed, 19 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index 1399776ff0..e9fae951e0 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -116,15 +116,8 @@ sp_vbuf_set_primitive(struct vbuf_render *vbr, unsigned prim) - if (prim == PIPE_PRIM_TRIANGLES || - prim == PIPE_PRIM_LINES || - prim == PIPE_PRIM_POINTS) { - cvbr->prim = prim; - return TRUE; - } - else { - return FALSE; - } + cvbr->prim = prim; + return TRUE; } @@ -201,7 +194,7 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) break; case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 2; i < nr; i += 3) { + for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, get_vert(vertex_buffer, indices[i+(i&1)-2], stride), get_vert(vertex_buffer, indices[i-(i&1)-1], stride), @@ -211,7 +204,7 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) case PIPE_PRIM_TRIANGLE_FAN: case PIPE_PRIM_POLYGON: - for (i = 2; i < nr; i += 3) { + for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, get_vert(vertex_buffer, indices[0], stride), get_vert(vertex_buffer, indices[i-1], stride), @@ -223,10 +216,10 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) setup_tri( setup_ctx, get_vert(vertex_buffer, indices[i-3], stride), get_vert(vertex_buffer, indices[i-2], stride), - get_vert(vertex_buffer, indices[i-1], stride)); + get_vert(vertex_buffer, indices[i-0], stride)); setup_tri( setup_ctx, - get_vert(vertex_buffer, indices[i-3], stride), + get_vert(vertex_buffer, indices[i-2], stride), get_vert(vertex_buffer, indices[i-1], stride), get_vert(vertex_buffer, indices[i-0], stride)); } @@ -234,14 +227,14 @@ sp_vbuf_draw(struct vbuf_render *vbr, const ushort *indices, uint nr) case PIPE_PRIM_QUAD_STRIP: for (i = 3; i < nr; i += 2) { setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-3], stride), get_vert(vertex_buffer, indices[i-2], stride), - get_vert(vertex_buffer, indices[i-1], stride), get_vert(vertex_buffer, indices[i-0], stride)); setup_tri( setup_ctx, + get_vert(vertex_buffer, indices[i-1], stride), get_vert(vertex_buffer, indices[i-3], stride), - get_vert(vertex_buffer, indices[i-0], stride), - get_vert(vertex_buffer, indices[i-2], stride)); + get_vert(vertex_buffer, indices[i-0], stride)); } break; default: @@ -265,11 +258,13 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr); struct softpipe_context *softpipe = cvbr->softpipe; struct draw_stage *setup = softpipe->setup; - const void *vertex_buffer = cvbr->vertex_buffer; + const void *vertex_buffer = NULL; const unsigned stride = softpipe->vertex_info_vbuf.size * sizeof(float); unsigned i; struct setup_context *setup_ctx = sp_draw_setup_context(setup); + vertex_buffer = (void *)get_vert(cvbr->vertex_buffer, start, stride); + switch (cvbr->prim) { case PIPE_PRIM_POINTS: for (i = 0; i < nr; i++) { @@ -318,7 +313,7 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) break; case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 2; i < nr; i += 3) { + for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, get_vert(vertex_buffer, i+(i&1)-2, stride), get_vert(vertex_buffer, i-(i&1)-1, stride), @@ -328,7 +323,7 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) case PIPE_PRIM_TRIANGLE_FAN: case PIPE_PRIM_POLYGON: - for (i = 2; i < nr; i += 3) { + for (i = 2; i < nr; i += 1) { setup_tri( setup_ctx, get_vert(vertex_buffer, 0, stride), get_vert(vertex_buffer, i-1, stride), @@ -336,31 +331,29 @@ sp_vbuf_draw_arrays(struct vbuf_render *vbr, uint start, uint nr) } break; case PIPE_PRIM_QUADS: - assert(nr % 4 == 0); for (i = 3; i < nr; i += 4) { setup_tri( setup_ctx, get_vert(vertex_buffer, i-3, stride), get_vert(vertex_buffer, i-2, stride), - get_vert(vertex_buffer, i-1, stride)); + get_vert(vertex_buffer, i-0, stride)); setup_tri( setup_ctx, - get_vert(vertex_buffer, i-3, stride), + get_vert(vertex_buffer, i-2, stride), get_vert(vertex_buffer, i-1, stride), get_vert(vertex_buffer, i-0, stride)); } break; case PIPE_PRIM_QUAD_STRIP: - assert(nr >= 4); for (i = 3; i < nr; i += 2) { setup_tri( setup_ctx, + get_vert(vertex_buffer, i-3, stride), get_vert(vertex_buffer, i-2, stride), - get_vert(vertex_buffer, i-1, stride), get_vert(vertex_buffer, i-0, stride)); setup_tri( setup_ctx, + get_vert(vertex_buffer, i-1, stride), get_vert(vertex_buffer, i-3, stride), - get_vert(vertex_buffer, i-0, stride), - get_vert(vertex_buffer, i-2, stride)); + get_vert(vertex_buffer, i-0, stride)); } break; default: -- cgit v1.2.3 From 44463b2997826cd14def00abf724a7a65a4fc7cb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 12 May 2008 17:36:35 +0100 Subject: draw: streamline the varray path - drop support for running the pipeline (ie. don't populate the flags values) - pass through all split-able primitives intact to the middle end - only primitives that can't be split are shunted on the draw-element path --- src/gallium/auxiliary/draw/draw_pt.c | 2 +- src/gallium/auxiliary/draw/draw_pt_varray.c | 106 +++---------- .../auxiliary/draw/draw_pt_varray_tmp_linear.h | 165 +++------------------ 3 files changed, 44 insertions(+), 229 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 448deef98c..d9e73a2396 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -81,7 +81,7 @@ draw_pt_arrays(struct draw_context *draw, /* Pick the right frontend */ - if (draw->pt.user.elts) { + if (draw->pt.user.elts || (opt & PT_PIPELINE)) { frontend = draw->pt.front.vcache; } else if (opt == PT_SHADE && draw->pt.test_fse) { /* should be a middle end.. */ diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 59a9569270..d92ad4fda1 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -154,9 +154,9 @@ static boolean split_prim_inplace(unsigned prim, unsigned *first, unsigned *incr static INLINE void add_draw_el(struct varray_frontend *varray, - int idx, ushort flags) + int idx) { - varray->draw_elts[varray->draw_count++] = idx | flags; + varray->draw_elts[varray->draw_count++] = idx; } @@ -165,106 +165,47 @@ static INLINE void varray_triangle( struct varray_frontend *varray, unsigned i1, unsigned i2 ) { - add_draw_el(varray, i0, 0); - add_draw_el(varray, i1, 0); - add_draw_el(varray, i2, 0); -} - -static INLINE void varray_triangle_flags( struct varray_frontend *varray, - ushort flags, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - add_draw_el(varray, i0, flags); - add_draw_el(varray, i1, 0); - add_draw_el(varray, i2, 0); + add_draw_el(varray, i0); + add_draw_el(varray, i1); + add_draw_el(varray, i2); } static INLINE void varray_line( struct varray_frontend *varray, unsigned i0, unsigned i1 ) { - add_draw_el(varray, i0, 0); - add_draw_el(varray, i1, 0); -} - - -static INLINE void varray_line_flags( struct varray_frontend *varray, - ushort flags, - unsigned i0, - unsigned i1 ) -{ - add_draw_el(varray, i0, flags); - add_draw_el(varray, i1, 0); + add_draw_el(varray, i0); + add_draw_el(varray, i1); } static INLINE void varray_point( struct varray_frontend *varray, unsigned i0 ) { - add_draw_el(varray, i0, 0); -} - -static INLINE void varray_quad( struct varray_frontend *varray, - unsigned i0, - unsigned i1, - unsigned i2, - unsigned i3 ) -{ - varray_triangle( varray, i0, i1, i3 ); - varray_triangle( varray, i1, i2, i3 ); + add_draw_el(varray, i0); } -static INLINE void varray_ef_quad( struct varray_frontend *varray, - unsigned i0, - unsigned i1, - unsigned i2, - unsigned i3 ) -{ - const unsigned omitEdge1 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_2; - const unsigned omitEdge2 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_1; - - varray_triangle_flags( varray, - DRAW_PIPE_RESET_STIPPLE | omitEdge1, - i0, i1, i3 ); - - varray_triangle_flags( varray, - omitEdge2, - i1, i2, i3 ); -} -/* At least for now, we're back to using a template include file for - * this. The two paths aren't too different though - it may be - * possible to reunify them. - */ -#define TRIANGLE(vc,flags,i0,i1,i2) varray_triangle_flags(vc,flags,i0,i1,i2) -#define QUAD(vc,i0,i1,i2,i3) varray_ef_quad(vc,i0,i1,i2,i3) -#define LINE(vc,flags,i0,i1) varray_line_flags(vc,flags,i0,i1) -#define POINT(vc,i0) varray_point(vc,i0) -#define FUNC varray_run_extras -#include "draw_pt_varray_tmp.h" -#define TRIANGLE(vc,flags,i0,i1,i2) varray_triangle(vc,i0,i1,i2) -#define QUAD(vc,i0,i1,i2,i3) varray_quad(vc,i0,i1,i2,i3) -#define LINE(vc,flags,i0,i1) varray_line(vc,i0,i1) +#define TRIANGLE(vc,i0,i1,i2) varray_triangle(vc,i0,i1,i2) +#define LINE(vc,i0,i1) varray_line(vc,i0,i1) #define POINT(vc,i0) varray_point(vc,i0) #define FUNC varray_run #include "draw_pt_varray_tmp_linear.h" -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { +static unsigned decompose_prim[PIPE_PRIM_POLYGON + 1] = { PIPE_PRIM_POINTS, PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, + PIPE_PRIM_LINE_STRIP, + PIPE_PRIM_LINES, /* decomposed */ PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES + PIPE_PRIM_TRIANGLE_STRIP, + PIPE_PRIM_TRIANGLES, /* decomposed */ + PIPE_PRIM_QUADS, + PIPE_PRIM_QUAD_STRIP, + PIPE_PRIM_TRIANGLES /* decomposed */ }; @@ -276,17 +217,10 @@ static void varray_prepare(struct draw_pt_front_end *frontend, { struct varray_frontend *varray = (struct varray_frontend *)frontend; - if (opt & PT_PIPELINE) - { - varray->base.run = varray_run_extras; - } - else - { - varray->base.run = varray_run; - } + varray->base.run = varray_run; varray->input_prim = prim; - varray->output_prim = reduced_prim[prim]; + varray->output_prim = decompose_prim[prim]; varray->middle = middle; middle->prepare(middle, varray->output_prim, opt); diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h index 6e2b16d9be..b6f1f0cadc 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -1,3 +1,7 @@ +static unsigned trim( unsigned count, unsigned first, unsigned incr ) +{ + return count - (count - first) % incr; +} static void FUNC(struct draw_pt_front_end *frontend, pt_elt_func get_elt, @@ -5,12 +9,9 @@ static void FUNC(struct draw_pt_front_end *frontend, unsigned count) { struct varray_frontend *varray = (struct varray_frontend *)frontend; - struct draw_context *draw = varray->draw; unsigned start = (unsigned)elts; - boolean flatfirst = (draw->rasterizer->flatshade && - draw->rasterizer->flatshade_first); - unsigned i, j, flags; + unsigned i, j; unsigned first, incr; varray->fetch_start = start; @@ -27,26 +28,30 @@ static void FUNC(struct draw_pt_front_end *frontend, case PIPE_PRIM_POINTS: case PIPE_PRIM_LINES: case PIPE_PRIM_TRIANGLES: - j = 0; - while (j + first <= count) { - unsigned end = MIN2(FETCH_MAX, count - j); - end -= (end % incr); - varray_flush_linear(varray, start + j, end); - j += end; + case PIPE_PRIM_LINE_STRIP: + case PIPE_PRIM_TRIANGLE_STRIP: + case PIPE_PRIM_QUADS: + case PIPE_PRIM_QUAD_STRIP: + + for (j = 0; j < count;) { + unsigned remaining = count - j; + unsigned nr = trim( MIN2(FETCH_MAX, remaining), first, incr ); + varray_flush_linear(varray, start + j, nr); + j += nr; + if (nr != remaining) + j -= (first - incr); } break; case PIPE_PRIM_LINE_LOOP: if (count >= 2) { - flags = DRAW_PIPE_RESET_STIPPLE; - for (j = 0; j + first <= count; j += i) { unsigned end = MIN2(FETCH_MAX, count - j); end -= (end % incr); - for (i = 1; i < end; i++, flags = 0) { - LINE(varray, flags, i - 1, i); + for (i = 1; i < end; i++) { + LINE(varray, i - 1, i); } - LINE(varray, flags, i - 1, 0); + LINE(varray, i - 1, 0); i = end; fetch_init(varray, end); varray_flush(varray); @@ -54,145 +59,21 @@ static void FUNC(struct draw_pt_front_end *frontend, } break; - case PIPE_PRIM_LINE_STRIP: - flags = DRAW_PIPE_RESET_STIPPLE; - for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(FETCH_MAX, count - j); - end -= (end % incr); - for (i = 1; i < end; i++, flags = 0) { - LINE(varray, flags, i - 1, i); - } - i = end; - fetch_init(varray, end); - varray_flush(varray); - } - break; - - case PIPE_PRIM_TRIANGLE_STRIP: - if (flatfirst) { - for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(FETCH_MAX, count - j); - end -= (end % incr); - for (i = 0; i+2 < end; i++) { - TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - i + 0, i + 1 + (i&1), i + 2 - (i&1)); - } - i = end; - fetch_init(varray, end); - varray_flush(varray); - if (j + first + i <= count) { - varray->fetch_start -= 2; - i -= 2; - } - } - } - else { - for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(FETCH_MAX, count - j); - end -= (end % incr); - for (i = 0; i+2 < end; i++) { - TRIANGLE(varray, DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - i + 0 + (i&1), i + 1 - (i&1), i + 2); - } - i = end; - fetch_init(varray, end); - varray_flush(varray); - if (j + first + i <= count) { - varray->fetch_start -= 2; - i -= 2; - } - } - } - break; + case PIPE_PRIM_POLYGON: case PIPE_PRIM_TRIANGLE_FAN: - if (count >= 3) { - if (flatfirst) { - flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; - for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(FETCH_MAX, count - j); - end -= (end % incr); - for (i = 0; i+2 < end; i++) { - TRIANGLE(varray, flags, i + 1, i + 2, 0); - } - i = end; - fetch_init(varray, end); - varray_flush(varray); - } - } - else { - flags = DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL; - for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(FETCH_MAX, count - j); - end -= (end % incr); - for (i = 0; i+2 < end; i++) { - TRIANGLE(varray, flags, 0, i + 1, i + 2); - } - i = end; - fetch_init(varray, end); - varray_flush(varray); - } - } - } - break; - - case PIPE_PRIM_QUADS: - for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(FETCH_MAX, count - j); - end -= (end % incr); - for (i = 0; i+3 < end; i += 4) { - QUAD(varray, i + 0, i + 1, i + 2, i + 3); - } - i = end; - fetch_init(varray, end); - varray_flush(varray); - } - break; - - case PIPE_PRIM_QUAD_STRIP: for (j = 0; j + first <= count; j += i) { unsigned end = MIN2(FETCH_MAX, count - j); end -= (end % incr); - for (i = 0; i+3 < end; i += 2) { - QUAD(varray, i + 2, i + 0, i + 1, i + 3); + for (i = 2; i < end; i++) { + TRIANGLE(varray, 0, i - 1, i); } i = end; fetch_init(varray, end); varray_flush(varray); - if (j + first + i <= count) { - varray->fetch_start -= 2; - i -= 2; - } } break; - case PIPE_PRIM_POLYGON: - { - /* These bitflags look a little odd because we submit the - * vertices as (1,2,0) to satisfy flatshade requirements. - */ - const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; - const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; - const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; - - flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; - for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(FETCH_MAX, count - j); - end -= (end % incr); - for (i = 0; i+2 < end; i++, flags = edge_middle) { - - if (i + 3 == count) - flags |= edge_last; - - TRIANGLE(varray, flags, i + 1, i + 2, 0); - } - i = end; - fetch_init(varray, end); - varray_flush(varray); - } - } - break; - default: assert(0); break; -- cgit v1.2.3 From bbda45ec769120324f44febf00c6bb170f594f23 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 12 May 2008 19:40:20 +0100 Subject: draw: turn fse path into a middle end Also add some util functions in pt_util.c --- src/gallium/auxiliary/draw/Makefile | 2 + src/gallium/auxiliary/draw/draw_private.h | 2 +- src/gallium/auxiliary/draw/draw_pt.c | 34 +- src/gallium/auxiliary/draw/draw_pt.h | 9 +- .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 39 +- src/gallium/auxiliary/draw/draw_pt_middle_fse.c | 705 +++++++++++++++++++++ src/gallium/auxiliary/draw/draw_pt_util.c | 103 +++ src/gallium/auxiliary/draw/draw_pt_varray.c | 45 -- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 2 +- .../auxiliary/draw/draw_pt_varray_tmp_linear.h | 9 +- src/gallium/auxiliary/draw/draw_pt_vcache.c | 15 +- 11 files changed, 847 insertions(+), 118 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_middle_fse.c create mode 100644 src/gallium/auxiliary/draw/draw_pt_util.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 68e7744cc5..67d78bdbbd 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -27,8 +27,10 @@ C_SOURCES = \ draw_pt_fetch.c \ draw_pt_fetch_emit.c \ draw_pt_fetch_shade_emit.c \ + draw_pt_middle_fse.c \ draw_pt_fetch_shade_pipeline.c \ draw_pt_post_vs.c \ + draw_pt_util.c \ draw_pt_varray.c \ draw_pt_vcache.c \ draw_vertex.c \ diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index cbe64cd290..86b901a3c8 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -124,7 +124,7 @@ struct draw_context struct { struct { struct draw_pt_middle_end *fetch_emit; - /*struct draw_pt_middle_end *fetch_shade_emit;*/ + struct draw_pt_middle_end *fetch_shade_emit; struct draw_pt_middle_end *general; } middle; diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index d9e73a2396..91e35db819 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -75,6 +75,8 @@ draw_pt_arrays(struct draw_context *draw, if (opt == 0) middle = draw->pt.middle.fetch_emit; + else if (opt == PT_SHADE && draw->pt.test_fse) + middle = draw->pt.middle.fetch_shade_emit; else middle = draw->pt.middle.general; @@ -83,9 +85,11 @@ draw_pt_arrays(struct draw_context *draw, */ if (draw->pt.user.elts || (opt & PT_PIPELINE)) { frontend = draw->pt.front.vcache; +#if 0 } else if (opt == PT_SHADE && draw->pt.test_fse) { /* should be a middle end.. */ frontend = draw->pt.front.fetch_shade_emit; +#endif } else { frontend = draw->pt.front.varray; } @@ -105,6 +109,8 @@ draw_pt_arrays(struct draw_context *draw, boolean draw_pt_init( struct draw_context *draw ) { + draw->pt.test_fse = GETENV("DRAW_FSE") != NULL; + draw->pt.front.vcache = draw_pt_vcache( draw ); if (!draw->pt.front.vcache) return FALSE; @@ -117,8 +123,11 @@ boolean draw_pt_init( struct draw_context *draw ) if (!draw->pt.middle.fetch_emit) return FALSE; - draw->pt.test_fse = GETENV("DRAW_FSE") != NULL; if (draw->pt.test_fse) { + draw->pt.middle.fetch_shade_emit = draw_pt_middle_fse( draw ); + if (!draw->pt.middle.fetch_shade_emit) + return FALSE; + draw->pt.front.fetch_shade_emit = draw_pt_fetch_shade_emit( draw ); if (!draw->pt.front.fetch_shade_emit) return FALSE; @@ -145,6 +154,11 @@ void draw_pt_destroy( struct draw_context *draw ) draw->pt.middle.fetch_emit = NULL; } + if (draw->pt.middle.fetch_shade_emit) { + draw->pt.middle.fetch_shade_emit->destroy( draw->pt.middle.fetch_shade_emit ); + draw->pt.middle.fetch_shade_emit = NULL; + } + if (draw->pt.front.fetch_shade_emit) { draw->pt.front.fetch_shade_emit->destroy( draw->pt.front.fetch_shade_emit ); draw->pt.front.fetch_shade_emit = NULL; @@ -163,19 +177,6 @@ void draw_pt_destroy( struct draw_context *draw ) -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { - PIPE_PRIM_POINTS, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES -}; - /** * Draw vertex arrays @@ -188,9 +189,10 @@ void draw_arrays(struct draw_context *draw, unsigned prim, unsigned start, unsigned count) { - if (reduced_prim[prim] != draw->reduced_prim) { + unsigned reduced_prim = draw_pt_reduced_prim(prim); + if (reduced_prim != draw->reduced_prim) { draw_do_flush( draw, DRAW_FLUSH_STATE_CHANGE ); - draw->reduced_prim = reduced_prim[prim]; + draw->reduced_prim = reduced_prim; } /* drawing done here: */ diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index bcd89f6bd6..cdae46b8d2 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -134,7 +134,7 @@ struct draw_pt_front_end *draw_pt_fetch_shade_emit( struct draw_context *draw ); * vertex_elements. */ struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ); -//struct draw_pt_middle_end *draw_pt_fetch_shade_emit( struct draw_context *draw ); +struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw ); struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit(struct draw_context *draw); @@ -213,4 +213,11 @@ struct pt_post_vs *draw_pt_post_vs_create( struct draw_context *draw ); void draw_pt_post_vs_destroy( struct pt_post_vs *pvs ); +/******************************************************************************* + * Utils: + */ +void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr); +unsigned draw_pt_reduced_prim(unsigned prim); + + #endif diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 9e1d1add36..f756d3e0bb 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -434,43 +434,6 @@ static void fse_prepare( struct draw_pt_front_end *fe, } -static boolean split_prim_inplace(unsigned prim, unsigned *first, unsigned *incr) -{ - switch (prim) { - case PIPE_PRIM_POINTS: - *first = 1; - *incr = 1; - return TRUE; - case PIPE_PRIM_LINES: - *first = 2; - *incr = 2; - return TRUE; - case PIPE_PRIM_LINE_STRIP: - *first = 2; - *incr = 1; - return TRUE; - case PIPE_PRIM_TRIANGLES: - *first = 3; - *incr = 3; - return TRUE; - case PIPE_PRIM_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - return TRUE; - case PIPE_PRIM_QUADS: - *first = 4; - *incr = 4; - return TRUE; - case PIPE_PRIM_QUAD_STRIP: - *first = 4; - *incr = 2; - return TRUE; - default: - *first = 0; - *incr = 1; /* set to one so that count % incr works */ - return FALSE; - } -} @@ -596,7 +559,7 @@ fse_run(struct draw_pt_front_end *fe, //debug_printf("%s prim %d start %d count %d\n", __FUNCTION__, prim, start, count); - split_prim_inplace(fse->prim, &first, &incr); + draw_pt_split_prim(fse->prim, &first, &incr); count -= (count - first) % incr; diff --git a/src/gallium/auxiliary/draw/draw_pt_middle_fse.c b/src/gallium/auxiliary/draw/draw_pt_middle_fse.c new file mode 100644 index 0000000000..cdb7d260da --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_middle_fse.c @@ -0,0 +1,705 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" +#include "draw/draw_vs.h" + +#include "translate/translate.h" + +struct fetch_shade_emit; + +struct fse_shader { + struct translate_key key; + + void (*run_linear)( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ); + + void (*run_elts)( const struct fetch_shade_emit *fse, + const unsigned *fetch_elts, + unsigned fetch_count, + char *buffer ); + +}; + +/* Prototype fetch, shade, emit-hw-verts all in one go. + */ +struct fetch_shade_emit { + struct draw_pt_middle_end base; + struct draw_context *draw; + + struct translate_key key; + + /* Temporaries: + */ + const float *constants; + unsigned pitch[PIPE_MAX_ATTRIBS]; + const ubyte *src[PIPE_MAX_ATTRIBS]; + unsigned prim; + + /* Points to one of the three hardwired example shaders, below: + */ + struct fse_shader *active; + + /* Temporary: A list of hard-wired shaders. Of course the plan + * would be to generate these for a given (vertex-shader, + * translate-key) pair... + */ + struct fse_shader shader[10]; + int nr_shaders; +}; + + + +/* Not quite passthrough yet -- we're still running the 'shader' here, + * inlined into the vertex fetch function. + */ +static void shader0_run_linear( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + + const float *m = fse->constants; + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; + const ubyte *st = fse->src[2] + start * fse->pitch[2]; + + float *out = (float *)buffer; + + /* loop over vertex attributes (vertex shader inputs) + */ + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + xyz += fse->pitch[0]; + } + + { + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + { + const float *in = (const float *)st; + out[8] = in[0]; + out[9] = in[1]; + out[10] = 0.0f; + out[11] = 1.0f; + st += fse->pitch[2]; + } + + out += 12; + } +} + + + +static void shader1_run_linear( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = (const float *)fse->constants; + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; + float *out = (float *)buffer; + +// debug_printf("rgb %f %f %f\n", rgb[0], rgb[1], rgb[2]); + + + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + xyz += fse->pitch[0]; + } + + { + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + out += 8; + } +} + + + + +static void shader2_run_linear( const struct fetch_shade_emit *fse, + unsigned start, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = (const float *)fse->constants; + const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; + const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; + const float psiz = 1.0; + float *out = (float *)buffer; + + + assert(fse->pitch[1] == 0); + + for (i = 0; i < count; i++) { + { + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + xyz += fse->pitch[0]; + } + + { + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + { + out[8] = psiz; + } + + out += 9; + } +} + + + + +static void shader0_run_elts( const struct fetch_shade_emit *fse, + const unsigned *elts, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = fse->constants; + float *out = (float *)buffer; + + + /* loop over vertex attributes (vertex shader inputs) + */ + for (i = 0; i < count; i++) { + unsigned elt = elts[i]; + { + const ubyte *xyz = fse->src[0] + elt * fse->pitch[0]; + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + } + + { + const ubyte *rgb = fse->src[1] + elt * fse->pitch[1]; + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + } + + { + const ubyte *st = fse->src[2] + elt * fse->pitch[2]; + const float *in = (const float *)st; + out[8] = in[0]; + out[9] = in[1]; + out[10] = 0.0f; + out[11] = 1.0f; + } + + out += 12; + } +} + + + +static void shader1_run_elts( const struct fetch_shade_emit *fse, + const unsigned *elts, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = (const float *)fse->constants; + float *out = (float *)buffer; + + for (i = 0; i < count; i++) { + unsigned elt = elts[i]; + + { + const ubyte *xyz = fse->src[0] + elt * fse->pitch[0]; + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + xyz += fse->pitch[0]; + } + + { + const ubyte *rgb = fse->src[1] + elt * fse->pitch[1]; + const float *in = (const float *)rgb; + out[4] = in[0]; + out[5] = in[1]; + out[6] = in[2]; + out[7] = 1.0f; + rgb += fse->pitch[1]; + } + + out += 8; + } +} + + + + +static void shader2_run_elts( const struct fetch_shade_emit *fse, + const unsigned *elts, + unsigned count, + char *buffer ) +{ + unsigned i; + const float *m = (const float *)fse->constants; + const float psiz = 1.0; + float *out = (float *)buffer; + + for (i = 0; i < count; i++) { + unsigned elt = elts[i]; + { + const ubyte *xyz = fse->src[0] + elt * fse->pitch[0]; + const float *in = (const float *)xyz; + const float ix = in[0], iy = in[1], iz = in[2]; + + out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; + out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; + out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; + out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; + } + + { + const ubyte *rgb = fse->src[1] + elt * fse->pitch[1]; + out[4] = rgb[0]; + out[5] = rgb[1]; + out[6] = rgb[2]; + out[7] = 1.0f; + } + + { + out[8] = psiz; + } + + out += 9; + } +} + + + +static void fse_prepare( struct draw_pt_middle_end *middle, + unsigned prim, + unsigned opt ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + unsigned num_vs_inputs = draw->vertex_shader->info.num_inputs; + unsigned num_vs_outputs = draw->vertex_shader->info.num_outputs; + const struct vertex_info *vinfo; + unsigned i; + boolean need_psize = 0; + + + if (draw->pt.user.elts) { + assert(0); + return ; + } + + if (!draw->render->set_primitive( draw->render, + prim )) { + assert(0); + return; + } + + /* Must do this after set_primitive() above: + */ + vinfo = draw->render->get_vertex_info(draw->render); + + + + fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ + num_vs_inputs); /* inputs - fetch from api format */ + + fse->key.output_stride = vinfo->size * 4; + memset(fse->key.element, 0, + fse->key.nr_elements * sizeof(fse->key.element[0])); + + for (i = 0; i < num_vs_inputs; i++) { + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].input_format = src->src_format; + + /* Consider ignoring these at this point, ie make generated + * programs independent of this state: + */ + fse->key.element[i].input_buffer = 0; //src->vertex_buffer_index; + fse->key.element[i].input_offset = 0; //src->src_offset; + } + + + { + unsigned dst_offset = 0; + + for (i = 0; i < vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned output_format = PIPE_FORMAT_NONE; + unsigned vs_output = vinfo->src_index[i]; + + switch (vinfo->emit[i]) { + case EMIT_4F: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + need_psize = 1; + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + vs_output = num_vs_outputs + 1; + + break; + default: + assert(0); + break; + } + + /* The elements in the key correspond to vertex shader output + * numbers, not to positions in the hw vertex description -- + * that's handled by the output_offset field. + */ + fse->key.element[vs_output].output_format = output_format; + fse->key.element[vs_output].output_offset = dst_offset; + + dst_offset += emit_sz; + assert(fse->key.output_stride >= dst_offset); + } + } + + /* To make psize work, really need to tell the vertex shader to + * copy that value from input->output. For 'translate' this was + * implicit for all elements. + */ +#if 0 + if (need_psize) { + unsigned input = num_vs_inputs + 1; + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].input_format = PIPE_FORMAT_R32_FLOAT; + fse->key.element[i].input_buffer = 0; //nr_buffers + 1; + fse->key.element[i].input_offset = 0; + + fse->key.nr_elements += 1; + + } +#endif + + fse->constants = draw->pt.user.constants; + + /* Would normally look up a vertex shader and peruse its list of + * varients somehow. We omitted that step and put all the + * hardcoded "shaders" into an array. We're just making the + * assumption that this happens to be a matching shader... ie + * you're running isosurf, aren't you? + */ + fse->active = NULL; + for (i = 0; i < fse->nr_shaders; i++) { + if (translate_key_compare( &fse->key, &fse->shader[i].key) == 0) + fse->active = &fse->shader[i]; + } + + if (!fse->active) { + assert(0); + return ; + } + + /* Now set buffer pointers: + */ + for (i = 0; i < num_vs_inputs; i++) { + unsigned buf = draw->pt.vertex_element[i].vertex_buffer_index; + + fse->src[i] = ((const ubyte *) draw->pt.user.vbuffer[buf] + + draw->pt.vertex_buffer[buf].buffer_offset + + draw->pt.vertex_element[i].src_offset); + + fse->pitch[i] = draw->pt.vertex_buffer[buf].pitch; + + } + + + //return TRUE; +} + + + + + + + +static void fse_run_linear( struct draw_pt_middle_end *middle, + unsigned start, + unsigned count ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + + char *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)count ); + + if (!hw_verts) { + assert(0); + return; + } + + /* Single routine to fetch vertices, run shader and emit HW verts. + * Clipping and viewport transformation are done elsewhere -- + * either by the API or on hardware, or for some other reason not + * required... + */ + fse->active->run_linear( fse, + start, count, + hw_verts ); + + /* Draw arrays path to avoid re-emitting index list again and + * again. + */ + draw->render->draw_arrays( draw->render, + 0, + count ); + + + draw->render->release_vertices( draw->render, + hw_verts, + fse->key.output_stride, + count ); +} + + +static void +fse_run(struct draw_pt_middle_end *middle, + const unsigned *fetch_elts, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + void *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)fetch_count ); + if (!hw_verts) { + assert(0); + return; + } + + + /* Single routine to fetch vertices, run shader and emit HW verts. + */ + fse->active->run_elts( fse, + fetch_elts, + fetch_count, + hw_verts ); + + draw->render->draw( draw->render, + draw_elts, + draw_count ); + + draw->render->release_vertices( draw->render, + hw_verts, + fse->key.output_stride, + fetch_count ); + +} + + +static void fse_finish( struct draw_pt_middle_end *middle ) +{ +} + + +static void +fse_destroy( struct draw_pt_middle_end *middle ) +{ + FREE(middle); +} + +struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw ) +{ + struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit); + if (!fse) + return NULL; + + fse->base.prepare = fse_prepare; + fse->base.run = fse_run; + fse->base.run_linear = fse_run_linear; + fse->base.finish = fse_finish; + fse->base.destroy = fse_destroy; + fse->draw = draw; + + fse->shader[0].run_linear = shader0_run_linear; + fse->shader[0].run_elts = shader0_run_elts; + fse->shader[0].key.nr_elements = 3; + fse->shader[0].key.output_stride = 12 * sizeof(float); + + fse->shader[0].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[0].key.element[0].input_buffer = 0; + fse->shader[0].key.element[0].input_offset = 0; + fse->shader[0].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[0].output_offset = 0; + + fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[0].key.element[1].input_buffer = 0; + fse->shader[0].key.element[1].input_offset = 0; + fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[1].output_offset = 16; + + fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32_FLOAT; + fse->shader[0].key.element[1].input_buffer = 0; + fse->shader[0].key.element[1].input_offset = 0; + fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[0].key.element[1].output_offset = 32; + + fse->shader[1].run_linear = shader1_run_linear; + fse->shader[1].run_elts = shader1_run_elts; + fse->shader[1].key.nr_elements = 2; + fse->shader[1].key.output_stride = 8 * sizeof(float); + + fse->shader[1].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[1].key.element[0].input_buffer = 0; + fse->shader[1].key.element[0].input_offset = 0; + fse->shader[1].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[1].key.element[0].output_offset = 0; + + fse->shader[1].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[1].key.element[1].input_buffer = 0; + fse->shader[1].key.element[1].input_offset = 0; + fse->shader[1].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[1].key.element[1].output_offset = 16; + + fse->shader[2].run_linear = shader2_run_linear; + fse->shader[2].run_elts = shader2_run_elts; + fse->shader[2].key.nr_elements = 3; + fse->shader[2].key.output_stride = 9 * sizeof(float); + + fse->shader[2].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[2].key.element[0].input_buffer = 0; + fse->shader[2].key.element[0].input_offset = 0; + fse->shader[2].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[2].key.element[0].output_offset = 0; + + fse->shader[2].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; + fse->shader[2].key.element[1].input_buffer = 0; + fse->shader[2].key.element[1].input_offset = 0; + fse->shader[2].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fse->shader[2].key.element[1].output_offset = 16; + + /* psize is special + * -- effectively add it here as another input!?! + * -- who knows how to add it as a buffer? + */ + fse->shader[2].key.element[2].input_format = PIPE_FORMAT_R32_FLOAT; + fse->shader[2].key.element[2].input_buffer = 0; + fse->shader[2].key.element[2].input_offset = 0; + fse->shader[2].key.element[2].output_format = PIPE_FORMAT_R32_FLOAT; + fse->shader[2].key.element[2].output_offset = 32; + + fse->nr_shaders = 3; + + return &fse->base; +} diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c new file mode 100644 index 0000000000..32c8a9632c --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_util.c @@ -0,0 +1,103 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_pt.h" + +void draw_pt_split_prim(unsigned prim, unsigned *first, unsigned *incr) +{ + switch (prim) { + case PIPE_PRIM_POINTS: + *first = 1; + *incr = 1; + break; + case PIPE_PRIM_LINES: + *first = 2; + *incr = 2; + break; + case PIPE_PRIM_LINE_STRIP: + case PIPE_PRIM_LINE_LOOP: + *first = 2; + *incr = 1; + break; + case PIPE_PRIM_TRIANGLES: + *first = 3; + *incr = 3; + break; + case PIPE_PRIM_TRIANGLE_STRIP: + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + *first = 3; + *incr = 1; + break; + case PIPE_PRIM_QUADS: + *first = 4; + *incr = 4; + break; + case PIPE_PRIM_QUAD_STRIP: + *first = 4; + *incr = 2; + break; + default: + assert(0); + *first = 0; + *incr = 1; /* set to one so that count % incr works */ + break; + } +} + + +unsigned draw_pt_reduced_prim(unsigned prim) +{ + switch (prim) { + case PIPE_PRIM_POINTS: + return PIPE_PRIM_POINTS; + case PIPE_PRIM_LINES: + case PIPE_PRIM_LINE_STRIP: + case PIPE_PRIM_LINE_LOOP: + return PIPE_PRIM_LINES; + case PIPE_PRIM_TRIANGLES: + case PIPE_PRIM_TRIANGLE_STRIP: + case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_POLYGON: + case PIPE_PRIM_QUADS: + case PIPE_PRIM_QUAD_STRIP: + return PIPE_PRIM_TRIANGLES; + default: + assert(0); + return PIPE_PRIM_POINTS; + } +} + + diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index d92ad4fda1..af6e2d5157 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -106,51 +106,6 @@ static INLINE void fetch_init(struct varray_frontend *varray, } -static boolean split_prim_inplace(unsigned prim, unsigned *first, unsigned *incr) -{ - switch (prim) { - case PIPE_PRIM_POINTS: - *first = 1; - *incr = 1; - return TRUE; - case PIPE_PRIM_LINES: - *first = 2; - *incr = 2; - return TRUE; - case PIPE_PRIM_LINE_STRIP: - *first = 2; - *incr = 1; - return TRUE; - case PIPE_PRIM_TRIANGLES: - *first = 3; - *incr = 3; - return TRUE; - case PIPE_PRIM_TRIANGLE_STRIP: - *first = 3; - *incr = 1; - return TRUE; - case PIPE_PRIM_TRIANGLE_FAN: - *first = 3; - *incr = 1; - return TRUE; - case PIPE_PRIM_QUADS: - *first = 4; - *incr = 4; - return TRUE; - case PIPE_PRIM_QUAD_STRIP: - *first = 4; - *incr = 2; - return TRUE; - case PIPE_PRIM_POLYGON: - *first = 3; - *incr = 1; - return TRUE; - default: - *first = 0; - *incr = 1; /* set to one so that count % incr works */ - return FALSE; - } -} static INLINE void add_draw_el(struct varray_frontend *varray, diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index 1395275897..6979f6b544 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -15,7 +15,7 @@ static void FUNC(struct draw_pt_front_end *frontend, varray->fetch_start = start; - split_prim_inplace(varray->input_prim, &first, &incr); + draw_pt_split_prim(varray->input_prim, &first, &incr); #if 0 debug_printf("%s (%d) %d/%d\n", __FUNCTION__, diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h index b6f1f0cadc..114ed371a0 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -16,7 +16,13 @@ static void FUNC(struct draw_pt_front_end *frontend, varray->fetch_start = start; - split_prim_inplace(varray->input_prim, &first, &incr); + draw_pt_split_prim(varray->input_prim, &first, &incr); + + /* Sanitize primitive length: + */ + count = trim(count, first, incr); + if (count < first) + return; #if 0 debug_printf("%s (%d) %d/%d\n", __FUNCTION__, @@ -32,7 +38,6 @@ static void FUNC(struct draw_pt_front_end *frontend, case PIPE_PRIM_TRIANGLE_STRIP: case PIPE_PRIM_QUADS: case PIPE_PRIM_QUAD_STRIP: - for (j = 0; j < count;) { unsigned remaining = count - j; unsigned nr = trim( MIN2(FETCH_MAX, remaining), first, incr ); diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 6b3fb1406b..a3495f2a30 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -204,19 +204,6 @@ static void vcache_ef_quad( struct vcache_frontend *vcache, -static unsigned reduced_prim[PIPE_PRIM_POLYGON + 1] = { - PIPE_PRIM_POINTS, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_LINES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES, - PIPE_PRIM_TRIANGLES -}; - static void vcache_prepare( struct draw_pt_front_end *frontend, @@ -236,7 +223,7 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, } vcache->input_prim = prim; - vcache->output_prim = reduced_prim[prim]; + vcache->output_prim = draw_pt_reduced_prim(prim); vcache->middle = middle; middle->prepare( middle, vcache->output_prim, opt ); -- cgit v1.2.3 From 19f15277d1871b62902031f9fa9aabf2f1bc7c40 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Mon, 12 May 2008 17:17:18 -0400 Subject: adjust llvm code to the changes in 2.3 --- src/gallium/auxiliary/gallivm/instructions.cpp | 30 ++++++++++++-------------- src/gallium/auxiliary/gallivm/storage.cpp | 16 +++++++------- src/gallium/auxiliary/gallivm/storagesoa.cpp | 8 +++---- 3 files changed, 26 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/instructions.cpp b/src/gallium/auxiliary/gallivm/instructions.cpp index 95a670edaf..1a98491b82 100644 --- a/src/gallium/auxiliary/gallivm/instructions.cpp +++ b/src/gallium/auxiliary/gallivm/instructions.cpp @@ -166,10 +166,9 @@ llvm::Value * Instructions::rsq(llvm::Value *in1) Value *abs = callFAbs(x); Value *sqrt = callFSqrt(abs); - Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - sqrt, - name("rsqrt")); + Value *rsqrt = m_builder.CreateFDiv(ConstantFP::get(APFloat(1.f)), + sqrt, + name("rsqrt")); return vectorFromVals(rsqrt, rsqrt, rsqrt, rsqrt); } @@ -278,9 +277,8 @@ llvm::Value * Instructions::rcp(llvm::Value *in1) Value *x1 = m_builder.CreateExtractElement(in1, m_storage->constantInt(0), name("x1")); - Value *res = m_builder.CreateFDiv(ConstantFP::get(Type::FloatTy, - APFloat(1.f)), - x1, name("rcp")); + Value *res = m_builder.CreateFDiv(ConstantFP::get(APFloat(1.f)), + x1, name("rcp")); return vectorFromVals(res, res, res, res); } @@ -319,13 +317,13 @@ llvm::Value * Instructions::dst(llvm::Value *in1, llvm::Value *in2) m_storage->constantInt(3), name("w")); Value *ry = m_builder.CreateMul(y1, y2, name("tyuy")); - return vectorFromVals(ConstantFP::get(Type::FloatTy, APFloat(1.f)), + return vectorFromVals(ConstantFP::get(APFloat(1.f)), ry, z, w); } llvm::Value * Instructions::ex2(llvm::Value *in) { - llvm::Value *val = callPow(ConstantFP::get(Type::FloatTy, APFloat(2.f)), + llvm::Value *val = callPow(ConstantFP::get(APFloat(2.f)), m_builder.CreateExtractElement( in, m_storage->constantInt(0), name("x1"))); @@ -526,7 +524,7 @@ llvm::Function * Instructions::declarePrintf() llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) { - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f)); Constant *const0f = Constant::getNullValue(Type::FloatTy); std::vector vec1 = extractVector(in1); @@ -547,7 +545,7 @@ llvm::Value * Instructions::sgt(llvm::Value *in1, llvm::Value *in2) } llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) { - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f)); Constant *const0f = Constant::getNullValue(Type::FloatTy); std::vector vec1 = extractVector(in1); @@ -571,7 +569,7 @@ llvm::Value * Instructions::sge(llvm::Value *in1, llvm::Value *in2) llvm::Value * Instructions::slt(llvm::Value *in1, llvm::Value *in2) { - Constant *const1f = ConstantFP::get(Type::FloatTy, APFloat(1.000000e+00f)); + Constant *const1f = ConstantFP::get(APFloat(1.000000e+00f)); Constant *const0f = Constant::getNullValue(Type::FloatTy); std::vector vec1 = extractVector(in1); @@ -814,10 +812,10 @@ llvm::Function * Instructions::findFunction(int label) llvm::Value * Instructions::constVector(float x, float y, float z, float w) { std::vector vec(4); - vec[0] = ConstantFP::get(Type::FloatTy, APFloat(x)); - vec[1] = ConstantFP::get(Type::FloatTy, APFloat(y)); - vec[2] = ConstantFP::get(Type::FloatTy, APFloat(z)); - vec[3] = ConstantFP::get(Type::FloatTy, APFloat(w)); + vec[0] = ConstantFP::get(APFloat(x)); + vec[1] = ConstantFP::get(APFloat(y)); + vec[2] = ConstantFP::get(APFloat(z)); + vec[3] = ConstantFP::get(APFloat(w)); return ConstantVector::get(m_floatVecType, vec); } diff --git a/src/gallium/auxiliary/gallivm/storage.cpp b/src/gallium/auxiliary/gallivm/storage.cpp index 9d9fd12360..6f373f6dd5 100644 --- a/src/gallium/auxiliary/gallivm/storage.cpp +++ b/src/gallium/auxiliary/gallivm/storage.cpp @@ -69,10 +69,10 @@ llvm::Constant *Storage::shuffleMask(int vec) { if (!m_extSwizzleVec) { std::vector elems; - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(0.f))); - elems.push_back(ConstantFP::get(Type::FloatTy, APFloat(1.f))); + elems.push_back(ConstantFP::get(APFloat(0.f))); + elems.push_back(ConstantFP::get(APFloat(1.f))); + elems.push_back(ConstantFP::get(APFloat(0.f))); + elems.push_back(ConstantFP::get(APFloat(1.f))); m_extSwizzleVec = ConstantVector::get(m_floatVecType, elems); } @@ -295,10 +295,10 @@ llvm::Value * Storage::immediateElement(int idx) void Storage::addImmediate(float *val) { std::vector vec(4); - vec[0] = ConstantFP::get(Type::FloatTy, APFloat(val[0])); - vec[1] = ConstantFP::get(Type::FloatTy, APFloat(val[1])); - vec[2] = ConstantFP::get(Type::FloatTy, APFloat(val[2])); - vec[3] = ConstantFP::get(Type::FloatTy, APFloat(val[3])); + vec[0] = ConstantFP::get(APFloat(val[0])); + vec[1] = ConstantFP::get(APFloat(val[1])); + vec[2] = ConstantFP::get(APFloat(val[2])); + vec[3] = ConstantFP::get(APFloat(val[3])); m_immediates.push_back(ConstantVector::get(m_floatVecType, vec)); } diff --git a/src/gallium/auxiliary/gallivm/storagesoa.cpp b/src/gallium/auxiliary/gallivm/storagesoa.cpp index 0e6e68c9d7..78d754371f 100644 --- a/src/gallium/auxiliary/gallivm/storagesoa.cpp +++ b/src/gallium/auxiliary/gallivm/storagesoa.cpp @@ -264,10 +264,10 @@ llvm::Constant * StorageSoa::createConstGlobalVector(const std::vector &v { VectorType *vectorType = VectorType::get(Type::FloatTy, 4); std::vector immValues; - ConstantFP *constx = ConstantFP::get(Type::FloatTy, APFloat(vec[0])); - ConstantFP *consty = ConstantFP::get(Type::FloatTy, APFloat(vec[1])); - ConstantFP *constz = ConstantFP::get(Type::FloatTy, APFloat(vec[2])); - ConstantFP *constw = ConstantFP::get(Type::FloatTy, APFloat(vec[3])); + ConstantFP *constx = ConstantFP::get(APFloat(vec[0])); + ConstantFP *consty = ConstantFP::get(APFloat(vec[1])); + ConstantFP *constz = ConstantFP::get(APFloat(vec[2])); + ConstantFP *constw = ConstantFP::get(APFloat(vec[3])); immValues.push_back(constx); immValues.push_back(consty); immValues.push_back(constz); -- cgit v1.2.3 From 1c624846a81b0218b4a07328f485e295432c6312 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 13 May 2008 16:06:09 -0400 Subject: decomposition from keith, adds decomposition of more prim to the pipeline --- src/gallium/auxiliary/draw/draw_pipe.c | 74 ++++++++---- src/gallium/auxiliary/draw/draw_pt_decompose.h | 153 +++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_pt_varray.c | 11 +- src/gallium/auxiliary/draw/draw_pt_vcache.c | 117 +++++++++---------- 4 files changed, 263 insertions(+), 92 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_decompose.h (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index cb97f955b2..1d26706dee 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -212,6 +212,55 @@ void draw_pipeline_run( struct draw_context *draw, draw->pipeline.vertex_count = 0; } +#define QUAD(i0,i1,i2,i3) \ + do_triangle( draw, \ + ( DRAW_PIPE_RESET_STIPPLE | \ + DRAW_PIPE_EDGE_FLAG_0 | \ + DRAW_PIPE_EDGE_FLAG_2 ), \ + verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i1), \ + verts + stride * (i3)); \ + do_triangle( draw, \ + ( DRAW_PIPE_EDGE_FLAG_0 | \ + DRAW_PIPE_EDGE_FLAG_1 ), \ + verts + stride * ((i1) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i2), \ + verts + stride * (i3)) + +#define TRIANGLE(flags,i0,i1,i2) \ + do_triangle( draw, \ + flags, /* flags */ \ + verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i1), \ + verts + stride * (i2)) + +#define LINE(flags,i0,i1) \ + do_line( draw, \ + flags, \ + verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ + verts + stride * (i+1)) + +#define POINT(i0) \ + do_point( draw, \ + verts + stride * i0 ) + +#define FUNC pipe_run_linear +#define ARGS \ + struct draw_context *draw, \ + unsigned prim, \ + struct vertex_header *vertices, \ + unsigned stride + +#define LOCAL_VARS \ + char *verts = (char *)vertices; \ + boolean flatfirst = (draw->rasterizer->flatshade && \ + draw->rasterizer->flatshade_first); \ + unsigned i, flags + +#define FLUSH + +#include "draw_pt_decompose.h" + void draw_pipeline_run_linear( struct draw_context *draw, unsigned prim, struct vertex_header *vertices, @@ -219,34 +268,11 @@ void draw_pipeline_run_linear( struct draw_context *draw, unsigned stride ) { char *verts = (char *)vertices; - unsigned i; - draw->pipeline.verts = verts; draw->pipeline.vertex_stride = stride; draw->pipeline.vertex_count = count; - switch (prim) { - case PIPE_PRIM_POINTS: - for (i = 0; i < count; i++) - do_point( draw, - verts + stride * i ); - break; - case PIPE_PRIM_LINES: - for (i = 0; i+1 < count; i += 2) - do_line( draw, - i+0, /* flags */ - verts + stride * ((i+0) & ~DRAW_PIPE_FLAG_MASK), - verts + stride * (i+1)); - break; - case PIPE_PRIM_TRIANGLES: - for (i = 0; i+2 < count; i += 3) - do_triangle( draw, - (i+0), /* flags */ - verts + stride * ((i+0) & ~DRAW_PIPE_FLAG_MASK), - verts + stride * (i+1), - verts + stride * (i+2)); - break; - } + pipe_run_linear(draw, prim, vertices, stride, count); draw->pipeline.verts = NULL; draw->pipeline.vertex_count = 0; diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h new file mode 100644 index 0000000000..dccfde99dd --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h @@ -0,0 +1,153 @@ + + +static void FUNC( ARGS, + unsigned count ) +{ + LOCAL_VARS; + + switch (prim) { + case PIPE_PRIM_POINTS: + for (i = 0; i < count; i ++) { + POINT( (i + 0) ); + } + break; + + case PIPE_PRIM_LINES: + for (i = 0; i+1 < count; i += 2) { + LINE( DRAW_PIPE_RESET_STIPPLE, + (i + 0), + (i + 1)); + } + break; + + case PIPE_PRIM_LINE_LOOP: + if (count >= 2) { + flags = DRAW_PIPE_RESET_STIPPLE; + + for (i = 1; i < count; i++, flags = 0) { + LINE( flags, + (i - 1), + (i )); + } + + LINE( flags, + (i - 1), + (0 )); + } + break; + + case PIPE_PRIM_LINE_STRIP: + flags = DRAW_PIPE_RESET_STIPPLE; + for (i = 1; i < count; i++, flags = 0) { + LINE( flags, + (i - 1), + (i )); + } + break; + + case PIPE_PRIM_TRIANGLES: + for (i = 0; i+2 < count; i += 3) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 0), + (i + 1), + (i + 2 )); + } + break; + + case PIPE_PRIM_TRIANGLE_STRIP: + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 0), + (i + 1 + (i&1)), + (i + 2 - (i&1))); + } + } + else { + for (i = 0; i+2 < count; i++) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 0 + (i&1)), + (i + 1 - (i&1)), + (i + 2 )); + } + } + break; + + case PIPE_PRIM_TRIANGLE_FAN: + if (count >= 3) { + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (i + 1), + (i + 2), + (0 )); + } + } + else { + for (i = 0; i+2 < count; i++) { + TRIANGLE( DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + (0), + (i + 1), + (i + 2 )); + } + } + } + break; + + + case PIPE_PRIM_QUADS: + for (i = 0; i+3 < count; i += 4) { + QUAD( (i + 0), + (i + 1), + (i + 2), + (i + 3)); + } + break; + + case PIPE_PRIM_QUAD_STRIP: + for (i = 0; i+3 < count; i += 2) { + QUAD( (i + 2), + (i + 0), + (i + 1), + (i + 3)); + } + break; + + case PIPE_PRIM_POLYGON: + { + /* These bitflags look a little odd because we submit the + * vertices as (1,2,0) to satisfy flatshade requirements. + */ + const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; + const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + + for (i = 0; i+2 < count; i++, flags = edge_middle) { + + if (i + 3 == count) + flags |= edge_last; + + TRIANGLE( flags, + (i + 1), + (i + 2), + (0)); + } + } + break; + + default: + assert(0); + break; + } + + FLUSH; +} + + +#undef TRIANGLE +#undef QUAD +#undef POINT +#undef LINE +#undef FUNC diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index af6e2d5157..06fd866ccd 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -141,14 +141,19 @@ static INLINE void varray_point( struct varray_frontend *varray, } - +#if 0 +#define TRIANGLE(flags,i0,i1,i2) varray_triangle(varray,i0,i1,i2) +#define LINE(flags,i0,i1) varray_line(varray,i0,i1) +#define POINT(i0) varray_point(varray,i0) +#define FUNC varray_decompose +#include "draw_pt_decompose.h" +#else #define TRIANGLE(vc,i0,i1,i2) varray_triangle(vc,i0,i1,i2) #define LINE(vc,i0,i1) varray_line(vc,i0,i1) #define POINT(vc,i0) varray_point(vc,i0) #define FUNC varray_run #include "draw_pt_varray_tmp_linear.h" - - +#endif static unsigned decompose_prim[PIPE_PRIM_POLYGON + 1] = { PIPE_PRIM_POINTS, diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index a3495f2a30..6c17edba34 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -104,22 +104,10 @@ static INLINE void vcache_elt( struct vcache_frontend *vcache, static void vcache_triangle( struct vcache_frontend *vcache, + ushort flags, unsigned i0, unsigned i1, unsigned i2 ) -{ - vcache_elt(vcache, i0, 0); - vcache_elt(vcache, i1, 0); - vcache_elt(vcache, i2, 0); - vcache_check_flush(vcache); -} - - -static void vcache_triangle_flags( struct vcache_frontend *vcache, - ushort flags, - unsigned i0, - unsigned i1, - unsigned i2 ) { vcache_elt(vcache, i0, flags); vcache_elt(vcache, i1, 0); @@ -128,19 +116,9 @@ static void vcache_triangle_flags( struct vcache_frontend *vcache, } static void vcache_line( struct vcache_frontend *vcache, + ushort flags, unsigned i0, unsigned i1 ) -{ - vcache_elt(vcache, i0, 0); - vcache_elt(vcache, i1, 0); - vcache_check_flush(vcache); -} - - -static void vcache_line_flags( struct vcache_frontend *vcache, - ushort flags, - unsigned i0, - unsigned i1 ) { vcache_elt(vcache, i0, flags); vcache_elt(vcache, i1, 0); @@ -161,46 +139,63 @@ static void vcache_quad( struct vcache_frontend *vcache, unsigned i2, unsigned i3 ) { - vcache_triangle( vcache, i0, i1, i3 ); - vcache_triangle( vcache, i1, i2, i3 ); -} - -static void vcache_ef_quad( struct vcache_frontend *vcache, - unsigned i0, - unsigned i1, - unsigned i2, - unsigned i3 ) -{ - const unsigned omitEdge1 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_2; - const unsigned omitEdge2 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_1; - - vcache_triangle_flags( vcache, - DRAW_PIPE_RESET_STIPPLE | omitEdge1, - i0, i1, i3 ); - - vcache_triangle_flags( vcache, - omitEdge2, - i1, i2, i3 ); + vcache_triangle( vcache, + ( DRAW_PIPE_RESET_STIPPLE | + DRAW_PIPE_EDGE_FLAG_0 | + DRAW_PIPE_EDGE_FLAG_2 ), + i0, i1, i3 ); + + vcache_triangle( vcache, + ( DRAW_PIPE_EDGE_FLAG_0 | + DRAW_PIPE_EDGE_FLAG_1 ), + i1, i2, i3 ); } /* At least for now, we're back to using a template include file for * this. The two paths aren't too different though - it may be * possible to reunify them. */ -#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle_flags(vc,flags,i0,i1,i2) -#define QUAD(vc,i0,i1,i2,i3) vcache_ef_quad(vc,i0,i1,i2,i3) -#define LINE(vc,flags,i0,i1) vcache_line_flags(vc,flags,i0,i1) -#define POINT(vc,i0) vcache_point(vc,i0) -#define FUNC vcache_run_extras -#include "draw_pt_vcache_tmp.h" - -#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle(vc,i0,i1,i2) -#define QUAD(vc,i0,i1,i2,i3) vcache_quad(vc,i0,i1,i2,i3) -#define LINE(vc,flags,i0,i1) vcache_line(vc,i0,i1) -#define POINT(vc,i0) vcache_point(vc,i0) +#define TRIANGLE(flags,i0,i1,i2) \ + vcache_triangle(vcache, \ + flags, \ + get_elt(elts,i0), \ + get_elt(elts,i1), \ + get_elt(elts,i2)) + +#define QUAD(i0,i1,i2,i3) \ + vcache_quad(vcache, \ + get_elt(elts,i0), \ + get_elt(elts,i1), \ + get_elt(elts,i2), \ + get_elt(elts,i3)) + +#define LINE(flags,i0,i1) \ + vcache_line(vcache, \ + flags, \ + get_elt(elts,i0), \ + get_elt(elts,i1)) + +#define POINT(i0) \ + vcache_point(vcache, \ + get_elt(elts,i0)) + #define FUNC vcache_run -#include "draw_pt_vcache_tmp.h" +#define ARGS \ + struct draw_pt_front_end *frontend, \ + pt_elt_func get_elt, \ + const void *elts + +#define LOCAL_VARS \ + struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; \ + struct draw_context *draw = vcache->draw; \ + boolean flatfirst = (draw->rasterizer->flatshade && \ + draw->rasterizer->flatshade_first); \ + unsigned prim = vcache->input_prim; \ + unsigned i, flags; +#define FLUSH vcache_flush( vcache ) + +#include "draw_pt_decompose.h" @@ -213,15 +208,7 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; - if (opt & PT_PIPELINE) - { - vcache->base.run = vcache_run_extras; - } - else - { - vcache->base.run = vcache_run; - } - + vcache->base.run = vcache_run; vcache->input_prim = prim; vcache->output_prim = draw_pt_reduced_prim(prim); -- cgit v1.2.3 From d07fbdd93941d5027a1aa7f0726140c6648a1e27 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 14 May 2008 17:19:44 +0200 Subject: i915: Added EGL winsys --- src/gallium/winsys/egl_drm/Makefile | 38 + src/gallium/winsys/egl_drm/Makefile.template | 116 +++ src/gallium/winsys/egl_drm/intel/Makefile | 40 + src/gallium/winsys/egl_drm/intel/SConscript | 39 + .../winsys/egl_drm/intel/intel_batchbuffer.c | 465 ++++++++++ .../winsys/egl_drm/intel/intel_batchbuffer.h | 133 +++ src/gallium/winsys/egl_drm/intel/intel_context.c | 366 ++++++++ src/gallium/winsys/egl_drm/intel/intel_context.h | 162 ++++ src/gallium/winsys/egl_drm/intel/intel_egl.c | 605 +++++++++++++ src/gallium/winsys/egl_drm/intel/intel_egl.h | 42 + src/gallium/winsys/egl_drm/intel/intel_lock.c | 102 +++ src/gallium/winsys/egl_drm/intel/intel_reg.h | 53 ++ src/gallium/winsys/egl_drm/intel/intel_screen.c | 680 +++++++++++++++ src/gallium/winsys/egl_drm/intel/intel_screen.h | 133 +++ .../winsys/egl_drm/intel/intel_swapbuffers.c | 327 +++++++ .../winsys/egl_drm/intel/intel_swapbuffers.h | 47 + src/gallium/winsys/egl_drm/intel/intel_winsys.h | 73 ++ .../winsys/egl_drm/intel/intel_winsys_i915.c | 184 ++++ .../winsys/egl_drm/intel/intel_winsys_pipe.c | 338 +++++++ .../winsys/egl_drm/intel/intel_winsys_softpipe.c | 82 ++ src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c | 953 ++++++++++++++++++++ src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h | 138 +++ src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h | 102 +++ src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c | 268 ++++++ src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c | 377 ++++++++ src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h | 115 +++ .../winsys/egl_drm/intel/ws_dri_mallocpool.c | 162 ++++ src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c | 970 +++++++++++++++++++++ 28 files changed, 7110 insertions(+) create mode 100644 src/gallium/winsys/egl_drm/Makefile create mode 100644 src/gallium/winsys/egl_drm/Makefile.template create mode 100644 src/gallium/winsys/egl_drm/intel/Makefile create mode 100644 src/gallium/winsys/egl_drm/intel/SConscript create mode 100644 src/gallium/winsys/egl_drm/intel/intel_batchbuffer.c create mode 100644 src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h create mode 100644 src/gallium/winsys/egl_drm/intel/intel_context.c create mode 100644 src/gallium/winsys/egl_drm/intel/intel_context.h create mode 100644 src/gallium/winsys/egl_drm/intel/intel_egl.c create mode 100644 src/gallium/winsys/egl_drm/intel/intel_egl.h create mode 100644 src/gallium/winsys/egl_drm/intel/intel_lock.c create mode 100644 src/gallium/winsys/egl_drm/intel/intel_reg.h create mode 100644 src/gallium/winsys/egl_drm/intel/intel_screen.c create mode 100644 src/gallium/winsys/egl_drm/intel/intel_screen.h create mode 100644 src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c create mode 100644 src/gallium/winsys/egl_drm/intel/intel_swapbuffers.h create mode 100644 src/gallium/winsys/egl_drm/intel/intel_winsys.h create mode 100644 src/gallium/winsys/egl_drm/intel/intel_winsys_i915.c create mode 100644 src/gallium/winsys/egl_drm/intel/intel_winsys_pipe.c create mode 100644 src/gallium/winsys/egl_drm/intel/intel_winsys_softpipe.c create mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c create mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h create mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h create mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c create mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c create mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h create mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_mallocpool.c create mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/Makefile b/src/gallium/winsys/egl_drm/Makefile new file mode 100644 index 0000000000..4139d9e71f --- /dev/null +++ b/src/gallium/winsys/egl_drm/Makefile @@ -0,0 +1,38 @@ +# src/mesa/drivers/egl_drm/Makefile + +TOP = ../../../.. + +include $(TOP)/configs/current + + + +default: $(TOP)/$(LIB_DIR) subdirs + + +$(TOP)/$(LIB_DIR): + -mkdir $(TOP)/$(LIB_DIR) + + +subdirs: + @for dir in $(DRI_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +install: + @for dir in $(DRI_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) install) || exit 1 ; \ + fi \ + done + + +clean: + @for dir in $(DRI_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) clean) ; \ + fi \ + done + -rm -f common/*.o diff --git a/src/gallium/winsys/egl_drm/Makefile.template b/src/gallium/winsys/egl_drm/Makefile.template new file mode 100644 index 0000000000..3bc1fdd4d4 --- /dev/null +++ b/src/gallium/winsys/egl_drm/Makefile.template @@ -0,0 +1,116 @@ +# -*-makefile-*- + +MESA_MODULES = \ + $(TOP)/src/mesa/libmesa.a \ + $(GALLIUM_AUXILIARIES) + +COMMON_GALLIUM_SOURCES = \ + $(TOP)/src/mesa/drivers/dri/common/utils.c \ + $(TOP)/src/mesa/drivers/dri/common/vblank.c \ + $(TOP)/src/mesa/drivers/dri/common/dri_util.c \ + $(TOP)/src/mesa/drivers/dri/common/xmlconfig.c + +COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \ + $(TOP)/src/mesa/drivers/common/driverfuncs.c \ + $(TOP)/src/mesa/drivers/dri/common/texmem.c \ + $(TOP)/src/mesa/drivers/dri/common/drirenderbuffer.c + +COMMON_BM_SOURCES = \ + $(TOP)/src/mesa/drivers/dri/common/dri_bufmgr.c \ + $(TOP)/src/mesa/drivers/dri/common/dri_drmpool.c + + +ifeq ($(WINDOW_SYSTEM),dri) +WINOBJ= +WINLIB= +INCLUDES = $(SHARED_INCLUDES) $(EXPAT_INCLUDES) + +OBJECTS = \ + $(C_SOURCES:.c=.o) \ + $(ASM_SOURCES:.S=.o) + +else +# miniglx +WINOBJ= +WINLIB=-L$(MESA)/src/glx/mini +MINIGLX_INCLUDES = -I$(TOP)/src/glx/mini +INCLUDES = $(MINIGLX_INCLUDES) \ + $(SHARED_INCLUDES) \ + $(PCIACCESS_CFLAGS) + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(MINIGLX_SOURCES:.c=.o) \ + $(ASM_SOURCES:.S=.o) +endif + + +### Include directories +SHARED_INCLUDES = \ + -I. \ + -I$(TOP)/src/mesa/drivers/dri/common \ + -Iserver \ + -I$(TOP)/include \ + -I$(TOP)/include/GL/internal \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/auxiliary \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main \ + -I$(TOP)/src/mesa/glapi \ + -I$(TOP)/src/mesa/math \ + -I$(TOP)/src/mesa/transform \ + -I$(TOP)/src/mesa/shader \ + -I$(TOP)/src/mesa/swrast \ + -I$(TOP)/src/mesa/swrast_setup \ + -I$(TOP)/src/egl/main \ + -I$(TOP)/src/egl/drivers/dri \ + $(LIBDRM_CFLAGS) + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + + +##### TARGETS ##### + +default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) + + +$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template + $(TOP)/bin/mklib -noprefix -o $@ \ + $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) + + +$(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) + $(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR) + + +depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ + $(ASM_SOURCES) 2> /dev/null + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean: + -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) + -rm -f depend depend.bak + + +install: $(LIBNAME) + $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) + $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) + + +include depend diff --git a/src/gallium/winsys/egl_drm/intel/Makefile b/src/gallium/winsys/egl_drm/intel/Makefile new file mode 100644 index 0000000000..5778ba77c3 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/Makefile @@ -0,0 +1,40 @@ + +TOP = ../../../../.. +include $(TOP)/configs/current + +LIBNAME = EGL_i915.so + +PIPE_DRIVERS = \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a + +DRIVER_SOURCES = \ + intel_winsys_pipe.c \ + intel_winsys_softpipe.c \ + intel_winsys_i915.c \ + intel_batchbuffer.c \ + intel_swapbuffers.c \ + intel_context.c \ + intel_lock.c \ + intel_screen.c \ + ws_dri_bufmgr.c \ + ws_dri_drmpool.c \ + ws_dri_fencemgr.c \ + ws_dri_mallocpool.c \ + ws_dri_slabpool.c \ + intel_egl.c + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +DRIVER_DEFINES = -I$(TOP)/src/mesa/drivers/dri/intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") + +include ../Makefile.template + +intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c + +symlinks: diff --git a/src/gallium/winsys/egl_drm/intel/SConscript b/src/gallium/winsys/egl_drm/intel/SConscript new file mode 100644 index 0000000000..0ad19d42a8 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/SConscript @@ -0,0 +1,39 @@ +Import('*') + +env = drienv.Clone() + +env.Append(CPPPATH = [ + '../intel', + 'server' +]) + +#MINIGLX_SOURCES = server/intel_dri.c + +DRIVER_SOURCES = [ + 'intel_winsys_pipe.c', + 'intel_winsys_softpipe.c', + 'intel_winsys_i915.c', + 'intel_batchbuffer.c', + 'intel_swapbuffers.c', + 'intel_context.c', + 'intel_lock.c', + 'intel_screen.c', + 'intel_batchpool.c', +] + +sources = \ + COMMON_GALLIUM_SOURCES + \ + COMMON_BM_SOURCES + \ + DRIVER_SOURCES + +drivers = [ + softpipe, + i915simple +] + +# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions +env.SharedLibrary( + target ='i915tex_dri.so', + source = sources, + LIBS = drivers + mesa + auxiliaries + env['LIBS'], +) \ No newline at end of file diff --git a/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.c b/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.c new file mode 100644 index 0000000000..7ffa05a6e6 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.c @@ -0,0 +1,465 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_batchbuffer.h" +#include "intel_context.h" +#include "intel_egl.h" + +#include + +#if 0 +static void +intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) +{ + int i; + fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); + for (i = 0; i < count / 4; i += 4) + fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", + offset + i * 4, ptr[i], ptr[i + 1], ptr[i + 2], ptr[i + 3]); + fprintf(stderr, "END BATCH\n\n\n"); +} +#endif + +static void +intel_realloc_relocs(struct intel_batchbuffer *batch, int num_relocs) +{ + unsigned long size = num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER; + + size *= sizeof(uint32_t); + batch->reloc = realloc(batch->reloc, size); + batch->reloc_size = num_relocs; +} + + +void +intel_batchbuffer_reset(struct intel_batchbuffer *batch) +{ + /* + * Get a new, free batchbuffer. + */ + drmBO *bo; + struct drm_bo_info_req *req; + int ret; + + driBOUnrefUserList(batch->list); + driBOResetList(batch->list); + + batch->size = 4096; // ZZZ JB batch->intel->intelScreen->maxBatchSize; + ret = driBOData(batch->buffer, batch->size, NULL, NULL, 0); + assert(!ret); + + /* + * Add the batchbuffer to the validate list. + */ + + driBOAddListItem(batch->list, batch->buffer, + DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_EXE | DRM_BO_MASK_MEM, + &batch->dest_location, &batch->node); + + req = &batch->node->bo_arg.d.req.bo_req; + + /* + * Set up information needed for us to make relocations + * relative to the underlying drm buffer objects. + */ + + driReadLockKernelBO(); + bo = driBOKernel(batch->buffer); + req->presumed_offset = (uint64_t) bo->offset; + req->hint = DRM_BO_HINT_PRESUMED_OFFSET; + batch->drmBOVirtual = (uint8_t *) bo->virtual; + driReadUnlockKernelBO(); + + /* + * Adjust the relocation buffer size. + */ + + if (batch->reloc_size > INTEL_MAX_RELOCS || + batch->reloc == NULL) + intel_realloc_relocs(batch, INTEL_DEFAULT_RELOCS); + + assert(batch->reloc != NULL); + batch->reloc[0] = 0; /* No relocs yet. */ + batch->reloc[1] = 1; /* Reloc type 1 */ + batch->reloc[2] = 0; /* Only a single relocation list. */ + batch->reloc[3] = 0; /* Only a single relocation list. */ + + batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); + batch->poolOffset = driBOPoolOffset(batch->buffer); + batch->ptr = batch->map; + batch->dirty_state = ~0; + batch->nr_relocs = 0; + batch->flags = 0; + batch->id = 0;//batch->intel->intelScreen->batch_id++; +} + +/*====================================================================== + * Public functions + */ +struct intel_batchbuffer * +intel_batchbuffer_alloc(struct intel_screen *intel_screen) +{ + struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); + + batch->intel_screen = intel_screen; + + driGenBuffers(intel_screen->batchPool, "batchbuffer", 1, + &batch->buffer, 4096, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); + batch->last_fence = NULL; + batch->list = driBOCreateList(20); + batch->reloc = NULL; + intel_batchbuffer_reset(batch); + return batch; +} + +void +intel_batchbuffer_free(struct intel_batchbuffer *batch) +{ + if (batch->last_fence) { + driFenceFinish(batch->last_fence, + DRM_FENCE_TYPE_EXE, GL_FALSE); + driFenceUnReference(&batch->last_fence); + } + if (batch->map) { + driBOUnmap(batch->buffer); + batch->map = NULL; + } + driBOUnReference(batch->buffer); + driBOFreeList(batch->list); + if (batch->reloc) + free(batch->reloc); + batch->buffer = NULL; + free(batch); +} + +void +intel_offset_relocation(struct intel_batchbuffer *batch, + unsigned pre_add, + struct _DriBufferObject *driBO, + uint64_t val_flags, + uint64_t val_mask) +{ + int itemLoc; + struct _drmBONode *node; + uint32_t *reloc; + struct drm_bo_info_req *req; + + driBOAddListItem(batch->list, driBO, val_flags, val_mask, + &itemLoc, &node); + req = &node->bo_arg.d.req.bo_req; + + if (!(req->hint & DRM_BO_HINT_PRESUMED_OFFSET)) { + + /* + * Stop other threads from tampering with the underlying + * drmBO while we're reading its offset. + */ + + driReadLockKernelBO(); + req->presumed_offset = (uint64_t) driBOKernel(driBO)->offset; + driReadUnlockKernelBO(); + req->hint = DRM_BO_HINT_PRESUMED_OFFSET; + } + + pre_add += driBOPoolOffset(driBO); + + if (batch->nr_relocs == batch->reloc_size) + intel_realloc_relocs(batch, batch->reloc_size * 2); + + reloc = batch->reloc + + (I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE); + + reloc[0] = ((uint8_t *)batch->ptr - batch->drmBOVirtual); + intel_batchbuffer_emit_dword(batch, req->presumed_offset + pre_add); + reloc[1] = pre_add; + reloc[2] = itemLoc; + reloc[3] = batch->dest_location; + batch->nr_relocs++; +} + +static void +i915_drm_copy_reply(const struct drm_bo_info_rep * rep, drmBO * buf) +{ + buf->handle = rep->handle; + buf->flags = rep->flags; + buf->size = rep->size; + buf->offset = rep->offset; + buf->mapHandle = rep->arg_handle; + buf->proposedFlags = rep->proposed_flags; + buf->start = rep->buffer_start; + buf->fenceFlags = rep->fence_flags; + buf->replyFlags = rep->rep_flags; + buf->pageAlignment = rep->page_alignment; +} + +static int +i915_execbuf(struct intel_batchbuffer *batch, + GLuint used, + GLboolean ignore_cliprects, + drmBOList *list, + struct drm_i915_execbuffer *ea) +{ + struct intel_screen *intel_screen = batch->intel_screen; + drmBONode *node; + drmMMListHead *l; + struct drm_i915_op_arg *arg, *first; + struct drm_bo_op_req *req; + struct drm_bo_info_rep *rep; + uint64_t *prevNext = NULL; + drmBO *buf; + int ret = 0; + uint32_t count = 0; + + first = NULL; + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + + arg = &node->bo_arg; + req = &arg->d.req; + + if (!first) + first = arg; + + if (prevNext) + *prevNext = (unsigned long)arg; + + prevNext = &arg->next; + req->bo_req.handle = node->buf->handle; + req->op = drm_bo_validate; + req->bo_req.flags = node->arg0; + req->bo_req.mask = node->arg1; + req->bo_req.hint |= 0; + count++; + } + + memset(ea, 0, sizeof(*ea)); + ea->num_buffers = count; + ea->batch.start = batch->poolOffset; + ea->batch.used = used; +#if 0 /* ZZZ JB: no cliprects used */ + ea->batch.cliprects = intel->pClipRects; + ea->batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + ea->batch.DR1 = 0; + ea->batch.DR4 = 0;((((GLuint) intel->drawX) & 0xffff) | + (((GLuint) intel->drawY) << 16)); +#else + ea->batch.cliprects = NULL; + ea->batch.num_cliprects = 0; + ea->batch.DR1 = 0; + ea->batch.DR4 = 0; +#endif + ea->fence_arg.flags = DRM_I915_FENCE_FLAG_FLUSHED; + ea->ops_list = (unsigned long) first; + first->reloc_ptr = (unsigned long) batch->reloc; + batch->reloc[0] = batch->nr_relocs; + + //return -EFAULT; + do { + ret = drmCommandWriteRead(intel_screen->device->drmFD, DRM_I915_EXECBUFFER, ea, + sizeof(*ea)); + } while (ret == -EAGAIN); + + if (ret != 0) { + printf("%s somebody set us up the bomb\n", __FUNCTION__); + return ret; + } + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + arg = &node->bo_arg; + rep = &arg->d.rep.bo_info; + + if (!arg->handled) { + return -EFAULT; + } + if (arg->d.rep.ret) + return arg->d.rep.ret; + + buf = node->buf; + i915_drm_copy_reply(rep, buf); + } + return 0; +} + +/* TODO: Push this whole function into bufmgr. + */ +static struct _DriFenceObject * +do_flush_locked(struct intel_batchbuffer *batch, + GLuint used, + GLboolean ignore_cliprects, GLboolean allow_unlock) +{ + struct intel_screen *intel_screen = batch->intel_screen; + struct _DriFenceObject *fo; + drmFence fence; + drmBOList *boList; + struct drm_i915_execbuffer ea; + int ret = 0; + + driBOValidateUserList(batch->list); + boList = driGetdrmBOList(batch->list); + +#if 0 /* ZZZ JB Allways run */ + if (!(intel->numClipRects == 0 && !ignore_cliprects)) { +#else + if (1) { +#endif + ret = i915_execbuf(batch, used, ignore_cliprects, boList, &ea); + } else { + driPutdrmBOList(batch->list); + fo = NULL; + goto out; + } + driPutdrmBOList(batch->list); + if (ret) + abort(); + + if (ea.fence_arg.error != 0) { + + /* + * The hardware has been idled by the kernel. + * Don't fence the driBOs. + */ + + if (batch->last_fence) + driFenceUnReference(&batch->last_fence); +#if 0 /* ZZZ JB: no _mesa_* funcs in gallium */ + _mesa_printf("fence error\n"); +#endif + batch->last_fence = NULL; + fo = NULL; + goto out; + } + + fence.handle = ea.fence_arg.handle; + fence.fence_class = ea.fence_arg.fence_class; + fence.type = ea.fence_arg.type; + fence.flags = ea.fence_arg.flags; + fence.signaled = ea.fence_arg.signaled; + + fo = driBOFenceUserList(intel_screen->mgr, batch->list, + "SuperFence", &fence); + + if (driFenceType(fo) & DRM_I915_FENCE_TYPE_RW) { + if (batch->last_fence) + driFenceUnReference(&batch->last_fence); + /* + * FIXME: Context last fence?? + */ + batch->last_fence = fo; + driFenceReference(fo); + } + out: +#if 0 /* ZZZ JB: fix this */ + intel->vtbl.lost_hardware(intel); +#else +#endif + return fo; +} + + +struct _DriFenceObject * +intel_batchbuffer_flush(struct intel_batchbuffer *batch) +{ + //struct intel_context *intel = batch->intel; + GLuint used = batch->ptr - batch->map; + //GLboolean was_locked = 1; + struct _DriFenceObject *fence; + + if (used == 0) { + driFenceReference(batch->last_fence); + return batch->last_fence; + } + + /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a + * performance drain that we would like to avoid. + */ +#if 0 /* ZZZ JB: what should we do here? */ + if (used & 4) { + ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->ptr)[1] = 0; + ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; + used += 8; + } +#else + if (used & 4) { + ((int *) batch->ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->ptr)[1] = 0; + ((int *) batch->ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; + used += 8; + } +#endif + driBOUnmap(batch->buffer); + batch->ptr = NULL; + batch->map = NULL; + + /* TODO: Just pass the relocation list and dma buffer up to the + * kernel. + */ +/* if (!was_locked) + LOCK_HARDWARE(intel);*/ + + fence = do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), + GL_FALSE); + +/* if (!was_locked) + UNLOCK_HARDWARE(intel);*/ + + /* Reset the buffer: + */ + intel_batchbuffer_reset(batch); + return fence; +} + +void +intel_batchbuffer_finish(struct intel_batchbuffer *batch) +{ + struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); + driFenceFinish(fence, driFenceType(fence), GL_FALSE); + driFenceUnReference(&fence); +} + +void +intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, GLuint bytes, GLuint flags) +{ + assert((bytes & 3) == 0); + intel_batchbuffer_require_space(batch, bytes, flags); + memcpy(batch->ptr, data, bytes); + batch->ptr += bytes; +} diff --git a/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h b/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h new file mode 100644 index 0000000000..6d35cf8b96 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h @@ -0,0 +1,133 @@ +#ifndef INTEL_BATCHBUFFER_H +#define INTEL_BATCHBUFFER_H + +#include "mtypes.h" +#include "ws_dri_bufmgr.h" + +struct intel_screen; + +#define BATCH_SZ 16384 +#define BATCH_RESERVED 16 + +#define INTEL_DEFAULT_RELOCS 100 +#define INTEL_MAX_RELOCS 400 + +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 + +struct intel_batchbuffer +{ + struct bufmgr *bm; + struct intel_screen *intel_screen; + + struct _DriBufferObject *buffer; + struct _DriFenceObject *last_fence; + GLuint flags; + + struct _DriBufferList *list; + GLuint list_count; + GLubyte *map; + GLubyte *ptr; + + uint32_t *reloc; + GLuint reloc_size; + GLuint nr_relocs; + + GLuint size; + + GLuint dirty_state; + GLuint id; + + uint32_t poolOffset; + uint8_t *drmBOVirtual; + struct _drmBONode *node; /* Validation list node for this buffer */ + int dest_location; /* Validation list sequence for this buffer */ +}; + +struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_screen + *intel); + +void intel_batchbuffer_free(struct intel_batchbuffer *batch); + + +void intel_batchbuffer_finish(struct intel_batchbuffer *batch); + +struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer + *batch); + +void intel_batchbuffer_reset(struct intel_batchbuffer *batch); + + +/* Unlike bmBufferData, this currently requires the buffer be mapped. + * Consider it a convenience function wrapping multple + * intel_buffer_dword() calls. + */ +void intel_batchbuffer_data(struct intel_batchbuffer *batch, + const void *data, GLuint bytes, GLuint flags); + +void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, + GLuint bytes); + +void +intel_offset_relocation(struct intel_batchbuffer *batch, + unsigned pre_add, + struct _DriBufferObject *driBO, + uint64_t val_flags, + uint64_t val_mask); + +/* Inline functions - might actually be better off with these + * non-inlined. Certainly better off switching all command packets to + * be passed as structs rather than dwords, but that's a little bit of + * work... + */ +static INLINE GLuint +intel_batchbuffer_space(struct intel_batchbuffer *batch) +{ + return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); +} + + +static INLINE void +intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) +{ + assert(batch->map); + assert(intel_batchbuffer_space(batch) >= 4); + *(GLuint *) (batch->ptr) = dword; + batch->ptr += 4; +} + +static INLINE void +intel_batchbuffer_require_space(struct intel_batchbuffer *batch, + GLuint sz, GLuint flags) +{ + struct _DriFenceObject *fence; + + assert(sz < batch->size - 8); + if (intel_batchbuffer_space(batch) < sz || + (batch->flags != 0 && flags != 0 && batch->flags != flags)) { + fence = intel_batchbuffer_flush(batch); + driFenceUnReference(&fence); + } + + batch->flags |= flags; +} + +/* Here are the crusty old macros, to be removed: + */ +#define BATCH_LOCALS + +#define BEGIN_BATCH(n, flags) do { \ + assert(!intel->prim.flush); \ + intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ +} while (0) + +#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) + +#define OUT_RELOC(buf,flags,mask,delta) do { \ + assert((delta) >= 0); \ + intel_offset_relocation(intel->batch, delta, buf, flags, mask); \ +} while (0) + +#define ADVANCE_BATCH() do { } while(0) + +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.c b/src/gallium/winsys/egl_drm/intel/intel_context.c new file mode 100644 index 0000000000..394a40d714 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_context.c @@ -0,0 +1,366 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_swapbuffers.h" +#include "intel_winsys.h" +#include "intel_batchbuffer.h" + +#include "state_tracker/st_public.h" +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "intel_egl.h" +#include "utils.h" + +#ifdef DEBUG +int __intel_debug = 0; +#endif + + +#define need_GL_ARB_multisample +#define need_GL_ARB_point_parameters +#define need_GL_ARB_texture_compression +#define need_GL_ARB_vertex_buffer_object +#define need_GL_ARB_vertex_program +#define need_GL_ARB_window_pos +#define need_GL_EXT_blend_color +#define need_GL_EXT_blend_equation_separate +#define need_GL_EXT_blend_func_separate +#define need_GL_EXT_blend_minmax +#define need_GL_EXT_cull_vertex +#define need_GL_EXT_fog_coord +#define need_GL_EXT_framebuffer_object +#define need_GL_EXT_multi_draw_arrays +#define need_GL_EXT_secondary_color +#define need_GL_NV_vertex_program +#include "extension_helper.h" + + +/** + * Extension strings exported by the intel driver. + * + * \note + * It appears that ARB_texture_env_crossbar has "disappeared" compared to the + * old i830-specific driver. + */ +const struct dri_extension card_extensions[] = { + {"GL_ARB_multisample", GL_ARB_multisample_functions}, + {"GL_ARB_multitexture", NULL}, + {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_texture_border_clamp", NULL}, + {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, + {"GL_ARB_texture_cube_map", NULL}, + {"GL_ARB_texture_env_add", NULL}, + {"GL_ARB_texture_env_combine", NULL}, + {"GL_ARB_texture_env_dot3", NULL}, + {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_rectangle", NULL}, + {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, + {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, + {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, + {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, + {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_subtract", NULL}, + {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, + {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, + {"GL_EXT_packed_depth_stencil", NULL}, + {"GL_EXT_pixel_buffer_object", NULL}, + {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, + {"GL_EXT_stencil_wrap", NULL}, + {"GL_EXT_texture_edge_clamp", NULL}, + {"GL_EXT_texture_env_combine", NULL}, + {"GL_EXT_texture_env_dot3", NULL}, + {"GL_EXT_texture_filter_anisotropic", NULL}, + {"GL_EXT_texture_lod_bias", NULL}, + {"GL_3DFX_texture_compression_FXT1", NULL}, + {"GL_APPLE_client_storage", NULL}, + {"GL_MESA_pack_invert", NULL}, + {"GL_MESA_ycbcr_texture", NULL}, + {"GL_NV_blend_square", NULL}, + {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, + {"GL_NV_vertex_program1_1", NULL}, + {"GL_SGIS_generate_mipmap", NULL }, + {NULL, NULL} +}; + +int +intel_create_context(struct egl_drm_context *eglCon, const __GLcontextModes *visual, void *sharedContextPrivate) +{ + struct intel_context *iCon = CALLOC_STRUCT(intel_context); + struct intel_screen *iScrn = (struct intel_screen *)eglCon->device->priv; + struct pipe_context *pipe; + struct st_context *st_share = NULL; + + eglCon->priv = iCon; + + iCon->intel_screen = iScrn; + iCon->egl_context = eglCon; + iCon->egl_device = eglCon->device; + + iCon->batch = intel_batchbuffer_alloc(iScrn); + iCon->last_swap_fence = NULL; + iCon->first_swap_fence = NULL; + + pipe = intel_create_i915simple(iCon, iScrn->winsys); +// pipe = intel_create_softpipe(iCon, iScrn->winsys); + + pipe->priv = iCon; + + iCon->st = st_create_context(pipe, visual, st_share); + return TRUE; +} + +void +intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *draw, struct egl_drm_drawable *read) +{ + if (context) { + struct intel_context *intel = (struct intel_context *)context->priv; + struct intel_framebuffer *draw_fb = (struct intel_framebuffer *)draw->priv; + struct intel_framebuffer *read_fb = (struct intel_framebuffer *)read->priv; + + assert(draw_fb->stfb); + assert(read_fb->stfb); + + st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); + + intel->egl_drawable = draw; + + st_resize_framebuffer(draw_fb->stfb, draw->w, draw->h); + + if (draw != read) + st_resize_framebuffer(read_fb->stfb, read->w, read->h); + + //intelUpdateWindowSize(driDrawPriv); + } else { + st_make_current(NULL, NULL, NULL); + } +} + +void +intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer *front) +{ + struct intel_screen *intelScreen = (struct intel_screen *)draw->device->priv; + struct intel_framebuffer *draw_fb = (struct intel_framebuffer *)draw->priv; + + driBOUnReference(draw_fb->front_buffer); + draw_fb->front_buffer = NULL; + draw_fb->front = NULL; + + /* to unbind just call this function with front == NULL */ + if (!front) + return; + + draw_fb->front = front; + + driGenBuffers(intelScreen->staticPool, "front", 1, &draw_fb->front_buffer, 0, 0, 0); + driBOSetReferenced(draw_fb->front_buffer, front->handle); + + st_resize_framebuffer(draw_fb->stfb, draw->w, draw->h); +} + +#if 0 +GLboolean +intelCreateContext(const __GLcontextModes * visual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) +{ + struct intel_context *intel = CALLOC_STRUCT(intel_context); + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + struct intel_screen *intelScreen = intel_screen(sPriv); + drmI830Sarea *saPriv = intelScreen->sarea; + int fthrottle_mode; + GLboolean havePools; + struct pipe_context *pipe; + struct st_context *st_share = NULL; + + if (sharedContextPrivate) { + st_share = ((struct intel_context *) sharedContextPrivate)->st; + } + + driContextPriv->driverPrivate = intel; + intel->intelScreen = intelScreen; + intel->driScreen = sPriv; + intel->sarea = saPriv; + + driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, + intel->driScreen->myNum, "i915"); + + + /* + * memory pools + */ + DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + // ZZZ JB should be per screen and not be done per context + havePools = intelCreatePools(sPriv); + DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + if (!havePools) + return GL_FALSE; + + + /* Dri stuff */ + intel->hHWContext = driContextPriv->hHWContext; + intel->driFd = sPriv->fd; + intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; + + fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); + intel->iw.irq_seq = -1; + intel->irqsEmitted = 0; + + intel->batch = intel_batchbuffer_alloc(intel); + intel->last_swap_fence = NULL; + intel->first_swap_fence = NULL; + +#ifdef DEBUG + __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); +#endif + + /* + * Pipe-related setup + */ + if (getenv("INTEL_SP")) { + /* use softpipe driver instead of hw */ + pipe = intel_create_softpipe( intel, intelScreen->winsys ); + } + else { + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + pipe = intel_create_i915simple( intel, intelScreen->winsys ); + break; + default: + fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", + intel->intelScreen->deviceID, __FUNCTION__); + + pipe = intel_create_softpipe( intel, intelScreen->winsys ); + break; + } + } + + pipe->priv = intel; + + intel->st = st_create_context(pipe, visual, st_share); + + return GL_TRUE; +} + + +void +intelDestroyContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = intel_context(driContextPriv); + + assert(intel); /* should never be null */ + if (intel) { + st_finish(intel->st); + + intel_batchbuffer_free(intel->batch); + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(&intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + if (intel->first_swap_fence) { + driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(&intel->first_swap_fence); + intel->first_swap_fence = NULL; + } + + if (intel->intelScreen->dummyContext == intel) + intel->intelScreen->dummyContext = NULL; + + st_destroy_context(intel->st); + free(intel); + } +} + + +GLboolean +intelUnbindContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = intel_context(driContextPriv); + st_flush(intel->st, PIPE_FLUSH_RENDER_CACHE, NULL); + /* XXX make_current(NULL)? */ + return GL_TRUE; +} + + +GLboolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) +{ + if (driContextPriv) { + struct intel_context *intel = intel_context(driContextPriv); + struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv); + struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv); + + assert(draw_fb->stfb); + assert(read_fb->stfb); + + /* This is for situations in which we need a rendering context but + * there may not be any currently bound. + */ + intel->intelScreen->dummyContext = intel; + + st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); + + if ((intel->driDrawable != driDrawPriv) || + (intel->lastStamp != driDrawPriv->lastStamp)) { + intel->driDrawable = driDrawPriv; + intelUpdateWindowSize(driDrawPriv); + intel->lastStamp = driDrawPriv->lastStamp; + } + + /* The size of the draw buffer will have been updated above. + * If the readbuffer is a different window, check/update its size now. + */ + if (driReadPriv != driDrawPriv) { + intelUpdateWindowSize(driReadPriv); + } + + } + else { + st_make_current(NULL, NULL, NULL); + } + + return GL_TRUE; +} +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.h b/src/gallium/winsys/egl_drm/intel/intel_context.h new file mode 100644 index 0000000000..aa9903f274 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_context.h @@ -0,0 +1,162 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_CONTEXT_H +#define INTEL_CONTEXT_H + +#include +#include "drm.h" + +#include "pipe/p_debug.h" + +#include "intel_screen.h" +#include "i915_drm.h" + + +struct pipe_context; +struct intel_context; +struct _DriBufferObject; +struct st_context; +struct egl_drm_device; +struct egl_drm_context; +struct egl_drm_frontbuffer; + + +#define INTEL_MAX_FIXUP 64 + +/** + * Intel rendering context, contains a state tracker and intel-specific info. + */ +struct intel_context +{ + struct st_context *st; + + struct _DriFenceObject *last_swap_fence; + struct _DriFenceObject *first_swap_fence; + + struct intel_batchbuffer *batch; + +#if 0 + boolean locked; + char *prevLockFile; + int prevLockLine; +#endif + + /* pick this up from the screen instead + int drmFd; + */ + + struct intel_screen *intel_screen; + + uint lastStamp; + /* new egl stuff */ + struct egl_drm_device *egl_device; + struct egl_drm_context *egl_context; + struct egl_drm_drawable *egl_drawable; +}; + + + +/** + * Intel framebuffer. + */ +struct intel_framebuffer +{ + struct st_framebuffer *stfb; + + /* other fields TBD */ + int other; + struct _DriBufferObject *front_buffer; + struct egl_drm_frontbuffer *front; +}; + + + + +/* These are functions now: + */ +void LOCK_HARDWARE( struct intel_context *intel ); +void UNLOCK_HARDWARE( struct intel_context *intel ); + +extern char *__progname; + + + +/* ================================================================ + * Debugging: + */ +#ifdef DEBUG +extern int __intel_debug; + +#define DEBUG_SWAP 0x1 +#define DEBUG_LOCK 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_BATCH 0x8 + +#define DBG(flag, ...) do { \ + if (__intel_debug & (DEBUG_##flag)) \ + printf(__VA_ARGS__); \ +} while(0) + +#else +#define DBG(flag, ...) +#endif + + + +#define PCI_CHIP_845_G 0x2562 +#define PCI_CHIP_I830_M 0x3577 +#define PCI_CHIP_I855_GM 0x3582 +#define PCI_CHIP_I865_G 0x2572 +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GME 0x27AE +#define PCI_CHIP_G33_G 0x29C2 +#define PCI_CHIP_Q35_G 0x29B2 +#define PCI_CHIP_Q33_G 0x29D2 + + +#if 0 +/** Cast wrapper */ +static INLINE struct intel_context * +intel_context(__DRIcontextPrivate *driContextPriv) +{ + return (struct intel_context *) driContextPriv->driverPrivate; +} + + +/** Cast wrapper */ +static INLINE struct intel_framebuffer * +intel_framebuffer(__DRIdrawablePrivate * driDrawPriv) +{ + return (struct intel_framebuffer *) driDrawPriv->driverPrivate; +} +#endif + +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c new file mode 100644 index 0000000000..a9c0218455 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -0,0 +1,605 @@ + +#include +#include +#include +#include + +#include "eglconfig.h" +#include "eglcontext.h" +#include "egldisplay.h" +#include "egldriver.h" +#include "eglglobals.h" +#include "eglmode.h" +#include "eglscreen.h" +#include "eglsurface.h" + +#include "glapi.h" +#include "intel_egl.h" + +#include "xf86drm.h" +#include "xf86drmMode.h" + +#include "intel_context.h" + +#include "ws_dri_bufmgr.h" + +#include "intel_winsys.h" +#include "state_tracker/st_public.h" + +struct egl_drm_device* egl_drm_create_device(int drmFD); + +struct egl_drm_device* +egl_drm_create_device(int drmFD) +{ + struct egl_drm_device *device = malloc(sizeof(*device)); + memset(device, 0, sizeof(*device)); + device->drmFD = drmFD; + + if (!intel_init_driver(device)) { + printf("EGL: failed to initalize device\n"); + free(device); + } + + return device; +} + +__GLcontextModes* _gl_context_modes_create( unsigned count, size_t minimum_size ); + +struct drm_driver +{ + _EGLDriver base; /* base class/object */ + + drmModeResPtr res; + struct egl_drm_device *device; +}; + +struct drm_surface +{ + _EGLSurface base; /* base class/object */ + + struct egl_drm_drawable *drawable; +}; + +struct drm_context +{ + _EGLContext base; /* base class/object */ + + struct egl_drm_context *context; +}; + +struct drm_screen +{ + _EGLScreen base; + + /* backing buffer and crtc */ + drmBO buffer; + drmModeFBPtr fb; + uint32_t fbID; + drmModeCrtcPtr crtc; + + /* currently only support one output */ + drmModeOutputPtr output; + uint32_t outputID; + + struct drm_mode_modeinfo *mode; + + /* geometry of the screen */ + struct egl_drm_frontbuffer front; +}; + +static void +drm_update_res(struct drm_driver *drm_drv) +{ + drmModeFreeResources(drm_drv->res); + drm_drv->res = drmModeGetResources(drm_drv->device->drmFD); +} + +static EGLBoolean +drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) +{ + _EGLDisplay *disp = _eglLookupDisplay(dpy); + struct drm_driver *drm_drv = (struct drm_driver *)drv; + struct drm_screen *screen = NULL; + drmModeOutputPtr output = NULL; + drmModeResPtr res = NULL; + + EGLint i; + int fd; + + fd = drmOpen("i915", NULL); + if (fd < 0) { + return EGL_FALSE; + } + + drm_drv->device = egl_drm_create_device(fd); + if (!drm_drv->device) { + drmClose(fd); + return EGL_FALSE; + } + + drm_update_res(drm_drv); + res = drm_drv->res; + + for(i = 0; i < res->count_outputs; i++) { + output = drmModeGetOutput(fd, res->outputs[i]); + + if (!output) + continue; + + if (output->connection == DRM_MODE_DISCONNECTED) { + drmModeFreeOutput(output); + continue; + } + + screen = malloc(sizeof(struct drm_screen)); + memset(screen, 0, sizeof(*screen)); + screen->outputID = res->outputs[i]; + screen->output = output; + _eglInitScreen(&screen->base); + _eglAddScreen(disp, &screen->base); + _eglAddNewMode(&screen->base, 1024, 768, 60 * 1000, "1024x768-60"); + } + + /* for now we only have one config */ + _EGLConfig config; + _eglInitConfig(&config, i + 1); + _eglSetConfigAttrib(&config, EGL_RED_SIZE, 8); + _eglSetConfigAttrib(&config, EGL_GREEN_SIZE, 8); + _eglSetConfigAttrib(&config, EGL_BLUE_SIZE, 8); + _eglSetConfigAttrib(&config, EGL_ALPHA_SIZE, 8); + _eglSetConfigAttrib(&config, EGL_BUFFER_SIZE, 32); + _eglSetConfigAttrib(&config, EGL_DEPTH_SIZE, 24); + _eglSetConfigAttrib(&config, EGL_STENCIL_SIZE, 8); + _eglSetConfigAttrib(&config, EGL_SURFACE_TYPE, EGL_PBUFFER_BIT); + _eglAddConfig(disp, &config); + + drv->Initialized = EGL_TRUE; + + *major = 1; + *minor = 0; + + return EGL_TRUE; +} + + +static EGLBoolean +drm_terminate(_EGLDriver *drv, EGLDisplay dpy) +{ + /* TODO: clean up */ + free(drv); + return EGL_TRUE; +} + + +static struct drm_context * +lookup_drm_context(EGLContext context) +{ + _EGLContext *c = _eglLookupContext(context); + return (struct drm_context *) c; +} + + +static struct drm_surface * +lookup_drm_surface(EGLSurface surface) +{ + _EGLSurface *s = _eglLookupSurface(surface); + return (struct drm_surface *) s; +} + +static struct drm_screen * +lookup_drm_screen(EGLDisplay dpy, EGLScreenMESA screen) +{ + _EGLScreen *s = _eglLookupScreen(dpy, screen); + return (struct drm_screen *) s; +} + +static __GLcontextModes* +visual_from_config(_EGLConfig *conf) +{ + __GLcontextModes *visual; + (void)conf; + + visual = _gl_context_modes_create(1, sizeof(*visual)); + visual->redBits = 8; + visual->greenBits = 8; + visual->blueBits = 8; + visual->alphaBits = 8; + + visual->rgbBits = 32; + visual->doubleBufferMode = 1; + + visual->depthBits = 24; + visual->stencilBits = 8; + + return visual; +} + + + +static EGLContext +drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list) +{ + struct drm_driver *drm_drv = (struct drm_driver *)drv; + struct drm_context *c; + struct drm_egl_context *share = NULL; + _EGLConfig *conf; + int i; + int ret; + __GLcontextModes *visual; + struct egl_drm_context *context; + + conf = _eglLookupConfig(drv, dpy, config); + if (!conf) { + _eglError(EGL_BAD_CONFIG, "eglCreateContext"); + return EGL_NO_CONTEXT; + } + + for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { + switch (attrib_list[i]) { + /* no attribs defined for now */ + default: + _eglError(EGL_BAD_ATTRIBUTE, "eglCreateContext"); + return EGL_NO_CONTEXT; + } + } + + c = (struct drm_context *) calloc(1, sizeof(struct drm_context)); + if (!c) + return EGL_NO_CONTEXT; + + _eglInitContext(drv, dpy, &c->base, config, attrib_list); + + context = malloc(sizeof(*context)); + memset(context, 0, sizeof(*context)); + + if (!context) + goto err_c; + + context->device = drm_drv->device; + visual = visual_from_config(conf); + + ret = intel_create_context(context, visual, share); + free(visual); + + if (!ret) + goto err_gl; + + c->context = context; + + /* generate handle and insert into hash table */ + _eglSaveContext(&c->base); + assert(c->base.Handle); + + return c->base.Handle; +err_gl: + free(context); +err_c: + free(c); + return EGL_NO_CONTEXT; +} + +static EGLBoolean +drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context) +{ + struct drm_context *fc = lookup_drm_context(context); + _eglRemoveContext(&fc->base); + if (fc->base.IsBound) { + fc->base.DeletePending = EGL_TRUE; + } else { + free(fc); + } + return EGL_TRUE; +} + + +static EGLSurface +drm_create_window_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list) +{ + return EGL_NO_SURFACE; +} + + +static EGLSurface +drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list) +{ + return EGL_NO_SURFACE; +} + +static EGLSurface +drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, + const EGLint *attrib_list) +{ + struct drm_driver *drm_drv = (struct drm_driver *)drv; + int i; + int ret; + int width = -1; + int height = -1; + struct drm_surface *surf = NULL; + struct egl_drm_drawable *drawable = NULL; + __GLcontextModes *visual; + _EGLConfig *conf; + + conf = _eglLookupConfig(drv, dpy, config); + if (!conf) { + _eglError(EGL_BAD_CONFIG, "eglCreatePbufferSurface"); + return EGL_NO_CONTEXT; + } + + for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { + switch (attrib_list[i]) { + case EGL_WIDTH: + width = attrib_list[++i]; + break; + case EGL_HEIGHT: + height = attrib_list[++i]; + break; + default: + _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface"); + return EGL_NO_SURFACE; + } + } + + if (width < 1 || height < 1) { + _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface"); + return EGL_NO_SURFACE; + } + + surf = (struct drm_surface *) calloc(1, sizeof(struct drm_surface)); + if (!surf) + goto err; + + if (!_eglInitSurface(drv, dpy, &surf->base, EGL_PBUFFER_BIT, config, attrib_list)) + goto err_surf; + + drawable = malloc(sizeof(*drawable)); + memset(drawable, 0, sizeof(*drawable)); + + drawable->w = width; + drawable->h = height; + + visual = visual_from_config(conf); + + drawable->device = drm_drv->device; + ret = intel_create_drawable(drawable, visual); + free(visual); + + if (!ret) + goto err_draw; + + surf->drawable = drawable; + + _eglSaveSurface(&surf->base); + return surf->base.Handle; + +err_draw: + free(drawable); +err_surf: + free(surf); +err: + return EGL_NO_SURFACE; +} + +static EGLSurface +drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, + const EGLint *attrib_list) +{ + EGLSurface surf = drm_create_pbuffer_surface(drv, dpy, cfg, attrib_list); + + return surf; +} + +static struct drm_mode_modeinfo * +drm_find_mode(drmModeOutputPtr output, _EGLMode *mode) +{ + int i; + struct drm_mode_modeinfo *m; + + for (i = 0; i < output->count_modes; i++) { + m = &output->modes[i]; + if (m->hdisplay == mode->Width && m->vdisplay == mode->Height) + break; + m = NULL; + } + + return m; +} +static void +draw(size_t x, size_t y, size_t w, size_t h, size_t pitch, size_t v, unsigned int *ptr) +{ + int i, j; + + for (i = x; i < x + w; i++) + for(j = y; j < y + h; j++) + ptr[(i * pitch / 4) + j] = v; + +} + +static void +prettyColors(int fd, unsigned int handle, size_t pitch) +{ + drmBO bo; + unsigned int *ptr; + int i; + + drmBOReference(fd, handle, &bo); + drmBOMap(fd, &bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, (void**)&ptr); + + for (i = 0; i < (bo.size / 4); i++) + ptr[i] = 0xFFFFFFFF; + + for (i = 0; i < 4; i++) + draw(i * 40, i * 40, 40, 40, pitch, 0, ptr); + + + draw(200, 100, 40, 40, pitch, 0xff00ff, ptr); + draw(100, 200, 40, 40, pitch, 0xff00ff, ptr); + + drmBOUnmap(fd, &bo); +} + +static EGLBoolean +drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, + EGLScreenMESA screen, + EGLSurface surface, EGLModeMESA m) +{ + struct drm_driver *drm_drv = (struct drm_driver *)drv; + struct drm_surface *surf = lookup_drm_surface(surface); + struct drm_screen *scrn = lookup_drm_screen(dpy, screen); + //struct intel_framebuffer *intel_fb = NULL; + //struct pipe_surface *front_surf = NULL; + _EGLMode *mode = _eglLookupMode(dpy, m); + size_t pitch = 2048 * 4; + size_t size = mode->Height * pitch; + int ret; + + /* TODO if allready shown take down */ + + printf("setting mode to %i x %i\n", mode->Width, mode->Height); + + ret = drmBOCreate(drm_drv->device->drmFD, size, 0, 0, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT | + DRM_BO_FLAG_MEM_VRAM | + DRM_BO_FLAG_NO_EVICT, + DRM_BO_HINT_DONT_FENCE, &scrn->buffer); + + prettyColors(drm_drv->device->drmFD, scrn->buffer.handle, pitch); + if (ret) { + printf("failed to create framebuffer (ret %d)\n", ret); + return EGL_FALSE; + } + + ret = drmModeAddFB(drm_drv->device->drmFD, mode->Width, mode->Height, + 32, 32, pitch, + scrn->buffer.handle, + &scrn->fbID); + + if (ret) + goto err_bo; + + scrn->fb = drmModeGetFB(drm_drv->device->drmFD, scrn->fbID); + if (!scrn->fb) + goto err_bo; + + scrn->mode = drm_find_mode(scrn->output, mode); + if (!scrn->mode) { + printf("oh noes, no matching mode found\n"); + goto err_fb; + } + + ret = drmModeSetCrtc( + drm_drv->device->drmFD, + drm_drv->res->crtcs[1], + scrn->fbID, + 0, 0, + &scrn->outputID, 1, + scrn->mode); + + + scrn->front.handle = scrn->buffer.handle; + scrn->front.pitch = pitch; + scrn->front.width = mode->Width; + scrn->front.height = mode->Height; + + intel_bind_frontbuffer(surf->drawable, &scrn->front); + + return EGL_TRUE; + +err_fb: + /* TODO remove fb */ + +err_bo: + drmBOUnreference(drm_drv->device->drmFD, &scrn->buffer); + return EGL_FALSE; +} + +static EGLBoolean +drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) +{ + struct drm_surface *fs = lookup_drm_surface(surface); + _eglRemoveSurface(&fs->base); + if (fs->base.IsBound) { + fs->base.DeletePending = EGL_TRUE; + } else { + free(fs); + } + return EGL_TRUE; +} + + + + +static EGLBoolean +drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context) +{ + struct drm_surface *readSurf = lookup_drm_surface(read); + struct drm_surface *drawSurf = lookup_drm_surface(draw); + struct drm_context *ctx = lookup_drm_context(context); + EGLBoolean b; + + b = _eglMakeCurrent(drv, dpy, draw, read, context); + if (!b) + return EGL_FALSE; + + /* XXX this is where we'd do the hardware context switch */ + (void) drawSurf; + (void) readSurf; + (void) ctx; + + intel_make_current(ctx->context, drawSurf->drawable, readSurf->drawable); + return EGL_TRUE; +} + +static EGLBoolean +drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) +{ + struct drm_surface *surf = lookup_drm_surface(draw); + if (!surf) + return EGL_FALSE; + + /* error checking */ + if (!_eglSwapBuffers(drv, dpy, draw)) + return EGL_FALSE; + + intel_swap_buffers(surf->drawable); + return EGL_TRUE; +} + + +/** + * The bootstrap function. Return a new drm_driver object and + * plug in API functions. + */ +_EGLDriver * +_eglMain(_EGLDisplay *dpy) +{ + struct drm_driver *drm; + + drm = (struct drm_driver *) calloc(1, sizeof(struct drm_driver)); + if (!drm) { + return NULL; + } + + /* First fill in the dispatch table with defaults */ + _eglInitDriverFallbacks(&drm->base); + /* then plug in our Drm-specific functions */ + drm->base.API.Initialize = drm_initialize; + drm->base.API.Terminate = drm_terminate; + drm->base.API.CreateContext = drm_create_context; + drm->base.API.MakeCurrent = drm_make_current; + drm->base.API.CreateWindowSurface = drm_create_window_surface; + drm->base.API.CreatePixmapSurface = drm_create_pixmap_surface; + drm->base.API.CreatePbufferSurface = drm_create_pbuffer_surface; + drm->base.API.DestroySurface = drm_destroy_surface; + drm->base.API.DestroyContext = drm_destroy_context; + drm->base.API.CreateScreenSurfaceMESA = drm_create_screen_surface_mesa; + drm->base.API.ShowScreenSurfaceMESA = drm_show_screen_surface_mesa; + drm->base.API.SwapBuffers = drm_swap_buffers; + + /* enable supported extensions */ + drm->base.Extensions.MESA_screen_surface = EGL_TRUE; + drm->base.Extensions.MESA_copy_context = EGL_TRUE; + + return &drm->base; +} diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.h b/src/gallium/winsys/egl_drm/intel/intel_egl.h new file mode 100644 index 0000000000..18e84e8eea --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.h @@ -0,0 +1,42 @@ + +#ifndef _INTEL_EGL_H_ +#define _INTEL_EGL_H_ + +struct egl_drm_device +{ + void *priv; + int drmFD; +}; + +struct egl_drm_context +{ + void *priv; + struct egl_drm_device *device; +}; + +struct egl_drm_drawable +{ + void *priv; + struct egl_drm_device *device; + size_t h; + size_t w; +}; + +struct egl_drm_frontbuffer +{ + uint32_t handle; + uint32_t pitch; + uint32_t width; + uint32_t height; +}; + +#include "GL/internal/glcore.h" + +int intel_init_driver(struct egl_drm_device *device); +int intel_create_context(struct egl_drm_context *context, const __GLcontextModes *visual, void *sharedContextPrivate); +int intel_create_drawable(struct egl_drm_drawable *drawable, const __GLcontextModes * visual); +void intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *draw, struct egl_drm_drawable *read); +void intel_swap_buffers(struct egl_drm_drawable *draw); +void intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer *front); + +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_lock.c b/src/gallium/winsys/egl_drm/intel/intel_lock.c new file mode 100644 index 0000000000..cec83c7585 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_lock.c @@ -0,0 +1,102 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "main/glheader.h" +#include "glapi/glthread.h" +#include +#include "state_tracker/st_public.h" +#include "intel_context.h" + +#if 0 + +_glthread_DECLARE_STATIC_MUTEX( lockMutex ); + + +static void +intelContendedLock(struct intel_context *intel, uint flags) +{ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + __DRIscreenPrivate *sPriv = intel->driScreen; + struct intel_screen *intelScreen = intel_screen(sPriv); + drmI830Sarea *sarea = intel->sarea; + + drmGetLock(intel->driFd, intel->hHWContext, flags); + + DBG(LOCK, "%s - got contended lock\n", __progname); + + /* If the window moved, may need to set a new cliprect now. + * + * NOTE: This releases and regains the hw lock, so all state + * checking must be done *after* this call: + */ + if (dPriv) + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + + if (sarea->width != intelScreen->front.width || + sarea->height != intelScreen->front.height) { + + intelUpdateScreenRotation(sPriv, sarea); + } +} + + +/* Lock the hardware and validate our state. + */ +void LOCK_HARDWARE( struct intel_context *intel ) +{ + char __ret = 0; + + _glthread_LOCK_MUTEX(lockMutex); + assert(!intel->locked); + + DRM_CAS(intel->driHwLock, intel->hHWContext, + (DRM_LOCK_HELD|intel->hHWContext), __ret); + + if (__ret) + intelContendedLock( intel, 0 ); + + DBG(LOCK, "%s - locked\n", __progname); + + intel->locked = 1; +} + + +/* Unlock the hardware using the global current context + */ +void UNLOCK_HARDWARE( struct intel_context *intel ) +{ + assert(intel->locked); + intel->locked = 0; + + DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); + + _glthread_UNLOCK_MUTEX(lockMutex); + + DBG(LOCK, "%s - unlocked\n", __progname); +} +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_reg.h b/src/gallium/winsys/egl_drm/intel/intel_reg.h new file mode 100644 index 0000000000..f37c24fda9 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_reg.h @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef _INTEL_REG_H_ +#define _INTEL_REG_H_ + + +#define BR00_BITBLT_CLIENT 0x40000000 +#define BR00_OP_COLOR_BLT 0x10000000 +#define BR00_OP_SRC_COPY_BLT 0x10C00000 +#define BR13_SOLID_PATTERN 0x80000000 + +#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) +#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) +#define XY_COLOR_BLT_WRITE_RGB (1<<20) + +#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) +#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) +#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) + +#define MI_WAIT_FOR_EVENT ((0x3<<23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +#define MI_BATCH_BUFFER_END (0xA<<23) + + +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.c b/src/gallium/winsys/egl_drm/intel/intel_screen.c new file mode 100644 index 0000000000..38c4098087 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_screen.c @@ -0,0 +1,680 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "utils.h" +#include "vblank.h" +#include "xmlpool.h" + +#include "intel_context.h" +#include "intel_screen.h" +#include "intel_batchbuffer.h" +//#include "intel_batchpool.h" +#include "intel_swapbuffers.h" +#include "intel_winsys.h" + +#include "ws_dri_bufpool.h" + +#include "pipe/p_context.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_cb_fbo.h" +#include "intel_egl.h" + +static boolean +intel_create_pools(struct intel_screen *intel_screen) +{ + if (intel_screen->havePools) + return GL_TRUE; + + intel_screen->mgr = driFenceMgrTTMInit(intel_screen->device->drmFD); + if (!intel_screen->mgr) { + fprintf(stderr, "Failed to create fence manager.\n"); + return FALSE; + } + + intel_screen->fMan = driInitFreeSlabManager(10, 10); + if (!intel_screen->fMan) { + fprintf(stderr, "Failed to create free slab manager.\n"); + return FALSE; + } + + intel_screen->staticPool = driDRMPoolInit(intel_screen->device->drmFD); + intel_screen->batchPool = driSlabPoolInit(intel_screen->device->drmFD, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + 4096, //intelScreen->maxBatchSize, + 1, 40, 16*16384, 0, + intel_screen->fMan); + + intel_screen->havePools = GL_TRUE; + + return GL_TRUE; +} + +extern const struct dri_extension card_extensions[]; + +int +intel_init_driver(struct egl_drm_device *device) +{ + struct intel_screen *intel_screen; + + /* Allocate the private area */ + intel_screen = CALLOC_STRUCT(intel_screen); + if (!intel_screen) + return FALSE; + + device->priv = (void *)intel_screen; + intel_screen->device = device; + + if (!intel_create_pools(intel_screen)) + return FALSE; + + intel_screen->batch = intel_batchbuffer_alloc(intel_screen); + + intel_screen->winsys = intel_create_pipe_winsys(device->drmFD, intel_screen->fMan); + + /* hack */ + driInitExtensions(NULL, card_extensions, GL_FALSE); + + return TRUE; +} + +int +intel_create_drawable(struct egl_drm_drawable *drawable, + const __GLcontextModes * visual) +{ + enum pipe_format colorFormat, depthFormat, stencilFormat; + struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); + + if (!intelfb) + return GL_FALSE; + + if (visual->redBits == 5) + colorFormat = PIPE_FORMAT_R5G6B5_UNORM; + else + colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; + + if (visual->depthBits == 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (visual->depthBits == 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_NONE; + + if (visual->stencilBits == 8) + stencilFormat = PIPE_FORMAT_S8Z24_UNORM; + else + stencilFormat = PIPE_FORMAT_NONE; + + intelfb->stfb = st_create_framebuffer(visual, + colorFormat, + depthFormat, + stencilFormat, + drawable->w, + drawable->h, + (void*) intelfb); + + if (!intelfb->stfb) { + free(intelfb); + return GL_FALSE; + } + + drawable->priv = (void *) intelfb; + return GL_TRUE; +} + +#if 0 +PUBLIC const char __driConfigOptions[] = + DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE + DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) + DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) + DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY + DRI_CONF_FORCE_S3TC_ENABLE(false) + DRI_CONF_ALLOW_LARGE_TEXTURES(1) + DRI_CONF_SECTION_END DRI_CONF_END; + +const uint __driNConfigOptions = 4; + +#ifdef USE_NEW_INTERFACE +static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; +#endif /*USE_NEW_INTERFACE */ + +extern const struct dri_extension card_extensions[]; + + + + +static void +intelPrintDRIInfo(struct intel_screen * intelScreen, + __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) +{ + fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", + intelScreen->front.size, intelScreen->front.offset, + intelScreen->front.pitch); + fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); +} + +static void +intelPrintSAREA(const drmI830Sarea * sarea) +{ + fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, + sarea->height); + fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); + fprintf(stderr, + "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->front_offset, sarea->front_size, + (unsigned) sarea->front_handle); + fprintf(stderr, + "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->back_offset, sarea->back_size, + (unsigned) sarea->back_handle); + fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->depth_offset, sarea->depth_size, + (unsigned) sarea->depth_handle); + fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); + fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); + fprintf(stderr, + "SAREA: rotated offset: 0x%08x size: 0x%x\n", + sarea->rotated_offset, sarea->rotated_size); + fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); +} + +/** + * Use the information in the sarea to update the screen parameters + * related to screen rotation. Needs to be called locked. + */ +void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) +{ + struct intel_screen *intelScreen = intel_screen(sPriv); + + if (intelScreen->front.map) { + drmUnmap(intelScreen->front.map, intelScreen->front.size); + intelScreen->front.map = NULL; + } + + if (intelScreen->front.buffer) + driDeleteBuffers(1, &intelScreen->front.buffer); + + intelScreen->front.width = sarea->width; + intelScreen->front.height = sarea->height; + intelScreen->front.offset = sarea->front_offset; + intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; + intelScreen->front.size = sarea->front_size; + intelScreen->front.handle = sarea->front_handle; + + assert( sarea->front_size >= + intelScreen->front.pitch * intelScreen->front.height ); + +#if 0 /* JB not important */ + if (!sarea->front_handle) + return; + + if (drmMap(sPriv->fd, + sarea->front_handle, + intelScreen->front.size, + (drmAddress *) & intelScreen->front.map) != 0) { + fprintf(stderr, "drmMap(frontbuffer) failed!\n"); + return; + } +#endif + +#if 0 /* JB */ + if (intelScreen->staticPool) { + driGenBuffers(intelScreen->staticPool, "static region", 1, + &intelScreen->front.buffer, 64, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); + + driBOSetStatic(intelScreen->front.buffer, + intelScreen->front.offset, + intelScreen->front.pitch * intelScreen->front.height, + intelScreen->front.map, 0); + } +#else + if (intelScreen->staticPool) { + if (intelScreen->front.buffer) + driBOUnReference(intelScreen->front.buffer); + driGenBuffers(intelScreen->staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); + driBOSetReferenced(intelScreen->front.buffer, sarea->front_bo_handle); + } +#endif +} + + +boolean +intelCreatePools(__DRIscreenPrivate * sPriv) +{ + //unsigned batchPoolSize = 1024*1024; + struct intel_screen *intelScreen = intel_screen(sPriv); + + if (intelScreen->havePools) + return GL_TRUE; + +#if 0 /* ZZZ JB fix this */ + intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); + if (!intelScreen->staticPool) + return GL_FALSE; + + batchPoolSize /= BATCH_SZ; + intelScreen->batchPool = driBatchPoolInit(sPriv->fd, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT | + DRM_BO_FLAG_MEM_LOCAL, + BATCH_SZ, + batchPoolSize, 5); + if (!intelScreen->batchPool) { + fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); + return GL_FALSE; + } +#else + intelScreen->staticPool = driDRMPoolInit(sPriv->fd); + intelScreen->batchPool = driSlabPoolInit(sPriv->fd, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + 4096, //intelScreen->maxBatchSize, + 1, 40, 16*16384, 0, + intelScreen->fMan); +#endif + intelScreen->havePools = GL_TRUE; + + //intelUpdateScreenRotation(sPriv, intelScreen->sarea); + + return GL_TRUE; +} + + +static boolean +intelInitDriver(__DRIscreenPrivate * sPriv) +{ + struct intel_screen *intelScreen; + I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; + + PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = + (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> + getProcAddress("glxEnableExtension")); + void *const psc = sPriv->psc->screenConfigs; + + if (sPriv->devPrivSize != sizeof(I830DRIRec)) { + fprintf(stderr, + "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); + return GL_FALSE; + } + + /* Allocate the private area */ + intelScreen = CALLOC_STRUCT(intel_screen); + if (!intelScreen) + return GL_FALSE; + + /* parse information in __driConfigOptions */ + driParseOptionInfo(&intelScreen->optionCache, + __driConfigOptions, __driNConfigOptions); + + sPriv->private = (void *) intelScreen; + + intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + + gDRIPriv->sarea_priv_offset); + intelScreen->deviceID = gDRIPriv->deviceID; + intelScreen->front.cpp = gDRIPriv->cpp; + intelScreen->drmMinor = sPriv->drmMinor; + + + assert(gDRIPriv->bitsPerPixel == 16 || + gDRIPriv->bitsPerPixel == 32); + + + intelUpdateScreenRotation(sPriv, intelScreen->sarea); + + if (0) + intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); + + if (glx_enable_extension != NULL) { + (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); + (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); + (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); + } + + + +#if 1 // ZZZ JB + intelScreen->mgr = driFenceMgrTTMInit(sPriv->fd); + if (!intelScreen->mgr) { + fprintf(stderr, "Failed to create fence manager.\n"); + return GL_FALSE; + } + + intelScreen->fMan = driInitFreeSlabManager(10, 10); + if (!intelScreen->fMan) { + fprintf(stderr, "Failed to create free slab manager.\n"); + return GL_FALSE; + } + + if (!intelCreatePools(sPriv)) + return GL_FALSE; +#endif + + intelScreen->winsys = intel_create_pipe_winsys(sPriv->fd, intelScreen->fMan); + + return GL_TRUE; +} + + +static void +intelDestroyScreen(__DRIscreenPrivate * sPriv) +{ + struct intel_screen *intelScreen = intel_screen(sPriv); + + /* intelUnmapScreenRegions(intelScreen); */ + + if (intelScreen->havePools) { + driPoolTakeDown(intelScreen->staticPool); + driPoolTakeDown(intelScreen->batchPool); + } + FREE(intelScreen); + sPriv->private = NULL; +} + + +/** + * This is called when we need to set up GL rendering to a new X window. + */ +static boolean +intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, + __DRIdrawablePrivate * driDrawPriv, + const __GLcontextModes * visual, boolean isPixmap) +{ + if (isPixmap) { + return GL_FALSE; /* not implemented */ + } + else { + enum pipe_format colorFormat, depthFormat, stencilFormat; + struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); + + if (!intelfb) + return GL_FALSE; + + if (visual->redBits == 5) + colorFormat = PIPE_FORMAT_R5G6B5_UNORM; + else + colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; + + if (visual->depthBits == 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (visual->depthBits == 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_NONE; + + if (visual->stencilBits == 8) + stencilFormat = PIPE_FORMAT_S8Z24_UNORM; + else + stencilFormat = PIPE_FORMAT_NONE; + + intelfb->stfb = st_create_framebuffer(visual, + colorFormat, + depthFormat, + stencilFormat, + driDrawPriv->w, + driDrawPriv->h, + (void*) intelfb); + if (!intelfb->stfb) { + free(intelfb); + return GL_FALSE; + } + + driDrawPriv->driverPrivate = (void *) intelfb; + return GL_TRUE; + } +} + +static void +intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) +{ + struct intel_framebuffer *intelfb = intel_framebuffer(driDrawPriv); + assert(intelfb->stfb); + st_unreference_framebuffer(&intelfb->stfb); + free(intelfb); +} + + +/** + * Get information about previous buffer swaps. + */ +static int +intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) +{ + if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) + || (sInfo == NULL)) { + return -1; + } + + return 0; +} + + +static void +intelSetTexOffset(__DRIcontext *pDRICtx, int texname, + unsigned long long offset, int depth, uint pitch) +{ + abort(); +#if 0 + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct st_texture_object *stObj = st_texture_object(tObj); + + if (!stObj) + return; + + if (stObj->pt) + st->pipe->texture_release(intel->st->pipe, &stObj->pt); + + stObj->imageOverride = GL_TRUE; + stObj->depthOverride = depth; + stObj->pitchOverride = pitch; + + if (offset) + stObj->textureOffset = offset; +#endif +} + + +static const struct __DriverAPIRec intelAPI = { + .InitDriver = intelInitDriver, + .DestroyScreen = intelDestroyScreen, + .CreateContext = intelCreateContext, + .DestroyContext = intelDestroyContext, + .CreateBuffer = intelCreateBuffer, + .DestroyBuffer = intelDestroyBuffer, + .SwapBuffers = intelSwapBuffers, + .MakeCurrent = intelMakeCurrent, + .UnbindContext = intelUnbindContext, + .GetSwapInfo = intelGetSwapInfo, + .GetMSC = driGetMSC32, + .WaitForMSC = driWaitForMSC32, + .WaitForSBC = NULL, + .SwapBuffersMSC = NULL, + .CopySubBuffer = intelCopySubBuffer, + .setTexOffset = intelSetTexOffset, +}; + + +static __GLcontextModes * +intelFillInModes(unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, boolean have_back_buffer) +{ + __GLcontextModes *modes; + __GLcontextModes *m; + unsigned num_modes; + unsigned depth_buffer_factor; + unsigned back_buffer_factor; + GLenum fb_format; + GLenum fb_type; + + /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't + * support pageflipping at all. + */ + static const GLenum back_buffer_modes[] = { + GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML + }; + + u_int8_t depth_bits_array[3]; + u_int8_t stencil_bits_array[3]; + + + depth_bits_array[0] = 0; + depth_bits_array[1] = depth_bits; + depth_bits_array[2] = depth_bits; + + /* Just like with the accumulation buffer, always provide some modes + * with a stencil buffer. It will be a sw fallback, but some apps won't + * care about that. + */ + stencil_bits_array[0] = 0; + stencil_bits_array[1] = 0; + if (depth_bits == 24) + stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; + + stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; + + depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; + back_buffer_factor = (have_back_buffer) ? 3 : 1; + + num_modes = depth_buffer_factor * back_buffer_factor * 4; + + if (pixel_bits == 16) { + fb_format = GL_RGB; + fb_type = GL_UNSIGNED_SHORT_5_6_5; + } + else { + fb_format = GL_BGRA; + fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; + } + + modes = + (*dri_interface->createContextModes) (num_modes, + sizeof(__GLcontextModes)); + m = modes; + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_TRUE_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_DIRECT_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + + /* Mark the visual as slow if there are "fake" stencil bits. + */ + for (m = modes; m != NULL; m = m->next) { + if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { + m->visualRating = GLX_SLOW_CONFIG; + } + } + + return modes; +} + + +/** + * This is the bootstrap function for the driver. libGL supplies all of the + * requisite information about the system, and the driver initializes itself. + * This routine also fills in the linked list pointed to by \c driver_modes + * with the \c __GLcontextModes that the driver can support for windows or + * pbuffers. + * + * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on + * failure. + */ +PUBLIC void * +__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, + __DRIscreen * psc, + const __GLcontextModes * modes, + const __DRIversion * ddx_version, + const __DRIversion * dri_version, + const __DRIversion * drm_version, + const __DRIframebuffer * frame_buffer, + drmAddress pSAREA, int fd, + int internal_api_version, + const __DRIinterfaceMethods * interface, + __GLcontextModes ** driver_modes) +{ + __DRIscreenPrivate *psp; + static const __DRIversion ddx_expected = { 1, 7, 0 }; + static const __DRIversion dri_expected = { 4, 0, 0 }; + static const __DRIversion drm_expected = { 1, 7, 0 }; + + dri_interface = interface; + + if (!driCheckDriDdxDrmVersions2("i915", + dri_version, &dri_expected, + ddx_version, &ddx_expected, + drm_version, &drm_expected)) { + return NULL; + } + + psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, + ddx_version, dri_version, drm_version, + frame_buffer, pSAREA, fd, + internal_api_version, &intelAPI); + + if (psp != NULL) { + I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; + *driver_modes = intelFillInModes(dri_priv->cpp * 8, + (dri_priv->cpp == 2) ? 16 : 24, + (dri_priv->cpp == 2) ? 0 : 8, 1); + + /* Calling driInitExtensions here, with a NULL context pointer, + * does not actually enable the extensions. It just makes sure + * that all the dispatch offsets for all the extensions that + * *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create + * is called, but we can't enable the extensions until we have a + * context pointer. + * + * Hello chicken. Hello egg. How are you two today? + */ + driInitExtensions(NULL, card_extensions, GL_FALSE); + } + + return (void *) psp; +} +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.h b/src/gallium/winsys/egl_drm/intel/intel_screen.h new file mode 100644 index 0000000000..e8c1cdfca4 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_screen.h @@ -0,0 +1,133 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _INTEL_SCREEN_H_ +#define _INTEL_SCREEN_H_ + +#include "ws_dri_bufpool.h" + +#include "pipe/p_compiler.h" + +struct egl_drm_device *device; + +struct intel_screen +{ +#if 0 + struct { + drm_handle_t handle; + + /* We create a static dri buffer for the frontbuffer. + */ + struct _DriBufferObject *buffer; + + char *map; /* memory map */ + int offset; /* from start of video mem, in bytes */ + int pitch; /* row stride, in bytes */ + int width; + int height; + int size; + int cpp; /* for front and back buffers */ + } front; +#endif + + int drmFB; + +#if 0 + int deviceID; + int drmMinor; + + + drmI830Sarea *sarea;*/ + + + /** + * Configuration cache with default values for all contexts + */ + driOptionCache optionCache; +#endif + + struct _DriBufferPool *batchPool; + struct _DriBufferPool *staticPool; /** for the X screen/framebuffer */ + boolean havePools; + +#if 0 + /** + * Temporary(?) context to use for SwapBuffers or other situations in + * which we need a rendering context, but none is currently bound. + */ + struct intel_context *dummyContext; +#endif + + /* + * New stuff form the i915tex integration + */ + struct _DriFenceMgr *mgr; + struct _DriFreeSlabManager *fMan; + unsigned batch_id; + + struct pipe_winsys *winsys; + struct egl_drm_device *device; + + /* batch buffer used for swap buffers */ + struct intel_batchbuffer *batch; +}; + + + +/** cast wrapper */ +#if 0 +static INLINE struct intel_screen * +intel_screen(__DRIscreenPrivate *sPriv) +{ + return (struct intel_screen *) sPriv->private; +} + + +extern void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); + + +extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); + +extern boolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); + +extern boolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv); + + +extern boolean +intelCreatePools(__DRIscreenPrivate *sPriv); + +extern boolean +intelCreateContext(const __GLcontextModes * visual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); + +#endif +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c new file mode 100644 index 0000000000..1ce4b2754a --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c @@ -0,0 +1,327 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_swapbuffers.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" +#include "intel_winsys.h" + +#include "pipe/p_context.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_cb_fbo.h" +#include "intel_egl.h" + + +static void +intel_display_surface(struct egl_drm_drawable *draw, + struct pipe_surface *surf); + +void intel_swap_buffers(struct egl_drm_drawable *draw) +{ + struct intel_framebuffer *intel_fb = (struct intel_framebuffer *)draw->priv; + struct pipe_surface *back_surf; + + assert(intel_fb); + assert(intel_fb->stfb); + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, ST_SURFACE_BACK_LEFT); + if (back_surf) { + st_notify_swapbuffers(intel_fb->stfb); + intel_display_surface(draw, back_surf); + st_notify_swapbuffers_complete(intel_fb->stfb); + } +} + +static void +intel_display_surface(struct egl_drm_drawable *draw, + struct pipe_surface *surf) +{ + struct intel_screen *intel = (struct intel_screen *)draw->device->priv; + struct intel_framebuffer *intel_fb = (struct intel_framebuffer *)draw->priv; + struct _DriFenceObject *fence; + + //const int srcWidth = surf->width; + //const int srcHeight = surf->height; + const int srcPitch = surf->pitch; + + const int dstWidth = intel_fb->front->width; + const int dstHeight = intel_fb->front->height; + const int dstPitch = intel_fb->front->pitch / 4;//draw->front.cpp; + + const int cpp = 4;//intel_fb->front->cpp; + + int BR13, CMD; + //int i; + + + BR13 = (dstPitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + + printf("srcPitch: %u, dstWidth: %u, dstHeight: %u, dstPitch: %u, cpp: %u\n", srcPitch, dstWidth, dstHeight, dstPitch, cpp); + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((0 << 16) | 0); + OUT_BATCH((dstHeight << 16) | dstWidth); + + OUT_RELOC(intel_fb->front_buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + + OUT_BATCH((0 << 16) | 0); + OUT_BATCH((srcPitch * cpp) & 0xffff); + OUT_RELOC(dri_bo(surf->buffer), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + ADVANCE_BATCH(); + + fence = intel_batchbuffer_flush(intel->batch); + driFenceUnReference(&fence); + intel_batchbuffer_finish(intel->batch); +} + +#if 0 +/** + * Display a colorbuffer surface in an X window. + * Used for SwapBuffers and flushing front buffer rendering. + * + * \param dPriv the window/drawable to display into + * \param surf the surface to display + * \param rect optional subrect of surface to display (may be NULL). + */ +void +intelDisplaySurface(__DRIdrawablePrivate *dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t *rect) +{ + struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); + //struct intel_context *intel = intelScreen->dummyContext; + + DBG(SWAP, "%s\n", __FUNCTION__); + +#if 0 + if (!intel) { + /* XXX this is where some kind of extra/meta context could be useful */ + return; + } +#endif + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); + driFenceUnReference(&intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + * current context (which is what intelScreenContext should return) might + * not get a contended lock and thus cliprects not updated (tests/manywin) + */ + if (intel_context(dPriv->driContextPriv) != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + + + if (dPriv && dPriv->numClipRects) { + const int srcWidth = surf->width; + const int srcHeight = surf->height; + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; + const int cpp = intelScreen->front.cpp; + const int srcpitch = surf->pitch; + int BR13, CMD; + int i; + + ASSERT(surf->buffer); + ASSERT(surf->cpp == cpp); + + DBG(SWAP, "screen pitch %d src surface pitch %d\n", + pitch, surf->pitch); + + if (cpp == 2) { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + } + else { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + } + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + drm_clip_rect_t sbox; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->front.width || + pbox->y2 > intelScreen->front.height) { + /* invalid cliprect, skip it */ + continue; + } + + box = *pbox; + + if (rect) { + /* intersect cliprect with user-provided src rect */ + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > srcWidth) + box.x2 = srcWidth + box.x1; + if (box.y2 - box.y1 > srcHeight) + box.y2 = srcHeight + box.y1; + + DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + + assert(box.x1 < box.x2); + assert(box.y1 < box.y2); + + /* XXX this could be done with pipe->surface_copy() */ + BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((box.y1 << 16) | box.x1); + OUT_BATCH((box.y2 << 16) | box.x2); + + OUT_RELOC(intelScreen->front.buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + OUT_BATCH((sbox.y1 << 16) | sbox.x1); + OUT_BATCH((srcpitch * cpp) & 0xffff); + OUT_RELOC(dri_bo(surf->buffer), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + ADVANCE_BATCH(); + } + + if (intel->first_swap_fence) + driFenceUnReference(&intel->first_swap_fence); + intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); + } + + UNLOCK_HARDWARE(intel); + + if (intel->lastStamp != dPriv->lastStamp) { + intelUpdateWindowSize(dPriv); + intel->lastStamp = dPriv->lastStamp; + } +} + + + +/** + * This will be called whenever the currently bound window is moved/resized. + */ +void +intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) +{ + struct intel_framebuffer *intelfb = intel_framebuffer(dPriv); + assert(intelfb->stfb); + st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); +} + + + +void +intelSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); + struct pipe_surface *back_surf; + + assert(intel_fb); + assert(intel_fb->stfb); + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { + st_notify_swapbuffers(intel_fb->stfb); + intelDisplaySurface(dPriv, back_surf, NULL); + st_notify_swapbuffers_complete(intel_fb->stfb); + } +} + + +/** + * Called via glXCopySubBufferMESA() to copy a subrect of the back + * buffer to the front buffer/screen. + */ +void +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) +{ + struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); + struct pipe_surface *back_surf; + + assert(intel_fb); + assert(intel_fb->stfb); + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { + drm_clip_rect_t rect; + rect.x1 = x; + rect.y1 = y; + rect.x2 = w; + rect.y2 = h; + + st_notify_swapbuffers(intel_fb->stfb); + intelDisplaySurface(dPriv, back_surf, &rect); + } +} +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.h b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.h new file mode 100644 index 0000000000..904f26732e --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_SWAPBUFFERS_H +#define INTEL_SWAPBUFFERS_H + + +struct pipe_surface; + +#if 0 +extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t * rect); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); + +extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, + int x, int y, int w, int h); + +extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); +#endif + +#endif /* INTEL_SWAPBUFFERS_H */ diff --git a/src/gallium/winsys/egl_drm/intel/intel_winsys.h b/src/gallium/winsys/egl_drm/intel/intel_winsys.h new file mode 100644 index 0000000000..d0a319f9a4 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_winsys.h @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_WINSYS_H +#define INTEL_WINSYS_H + +#include "pipe/p_state.h" + +struct intel_context; +struct pipe_context; +struct pipe_winsys; +struct pipe_buffer; +struct _DriBufferObject; + +struct pipe_winsys * +intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ); + +void +intel_destroy_pipe_winsys( struct pipe_winsys *winsys ); + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel, + struct pipe_winsys *winsys ); + +struct pipe_context * +intel_create_i915simple( struct intel_context *intel, + struct pipe_winsys *winsys ); + + +struct intel_buffer { + struct pipe_buffer base; + struct _DriBufferObject *driBO; +}; + +static INLINE struct intel_buffer * +intel_buffer( struct pipe_buffer *buf ) +{ + return (struct intel_buffer *)buf; +} + +static INLINE struct _DriBufferObject * +dri_bo( struct pipe_buffer *buf ) +{ + return intel_buffer(buf)->driBO; +} + + + +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_winsys_i915.c b/src/gallium/winsys/egl_drm/intel/intel_winsys_i915.c new file mode 100644 index 0000000000..8ec5c7e82a --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_winsys_i915.c @@ -0,0 +1,184 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" + +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_winsys.h" + +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#include "i915simple/i915_winsys.h" +#include "i915simple/i915_screen.h" + + +struct intel_i915_winsys { + struct i915_winsys winsys; /**< batch buffer funcs */ + struct pipe_winsys *pws; + struct intel_context *intel; +}; + + +/* Turn a i915simple winsys into an intel/i915simple winsys: + */ +static inline struct intel_i915_winsys * +intel_i915_winsys( struct i915_winsys *sws ) +{ + return (struct intel_i915_winsys *)sws; +} + + +/* Simple batchbuffer interface: + */ + +static unsigned *intel_i915_batch_start( struct i915_winsys *sws, + unsigned dwords, + unsigned relocs ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + + /* XXX: check relocs. + */ + if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { + /* XXX: Hmm, the driver can't really do much with this pointer: + */ + return (unsigned *)intel->batch->ptr; + } + else + return NULL; +} + +static void intel_i915_batch_dword( struct i915_winsys *sws, + unsigned dword ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + intel_batchbuffer_emit_dword( intel->batch, dword ); +} + +static void intel_i915_batch_reloc( struct i915_winsys *sws, + struct pipe_buffer *buf, + unsigned access_flags, + unsigned delta ) +{ + struct intel_context *intel = intel_i915_winsys(sws)->intel; + unsigned flags = DRM_BO_FLAG_MEM_TT; + unsigned mask = DRM_BO_MASK_MEM; + + if (access_flags & I915_BUFFER_ACCESS_WRITE) { + flags |= DRM_BO_FLAG_WRITE; + mask |= DRM_BO_FLAG_WRITE; + } + + if (access_flags & I915_BUFFER_ACCESS_READ) { + flags |= DRM_BO_FLAG_READ; + mask |= DRM_BO_FLAG_READ; + } + +#if 0 /* JB old */ + intel_batchbuffer_emit_reloc( intel->batch, + dri_bo( buf ), + flags, mask, + delta ); +#else /* new */ + intel_offset_relocation( intel->batch, + delta, + dri_bo( buf ), + flags, + mask ); +#endif +} + + + +static void intel_i915_batch_flush( struct i915_winsys *sws, + struct pipe_fence_handle **fence ) +{ + struct intel_i915_winsys *iws = intel_i915_winsys(sws); + struct intel_context *intel = iws->intel; + union { + struct _DriFenceObject *dri; + struct pipe_fence_handle *pipe; + } fu; + + if (fence) + assert(!*fence); + + fu.dri = intel_batchbuffer_flush( intel->batch ); + + if (!fu.dri) { + assert(0); + *fence = NULL; + return; + } + + if (fu.dri) { + if (fence) + *fence = fu.pipe; + else + driFenceUnReference(&fu.dri); + } + +} + + +/** + * Create i915 hardware rendering context. + */ +struct pipe_context * +intel_create_i915simple( struct intel_context *intel, + struct pipe_winsys *winsys ) +{ + struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); + struct pipe_screen *screen; + + /* Fill in this struct with callbacks that i915simple will need to + * communicate with the window system, buffer manager, etc. + */ + iws->winsys.batch_start = intel_i915_batch_start; + iws->winsys.batch_dword = intel_i915_batch_dword; + iws->winsys.batch_reloc = intel_i915_batch_reloc; + iws->winsys.batch_flush = intel_i915_batch_flush; + iws->pws = winsys; + iws->intel = intel; + + screen = i915_create_screen(winsys, PCI_CHIP_I945_GM); + assert(screen); + + /* Create the i915simple context: + */ + return i915_create_context( screen, + winsys, + &iws->winsys ); +} diff --git a/src/gallium/winsys/egl_drm/intel/intel_winsys_pipe.c b/src/gallium/winsys/egl_drm/intel/intel_winsys_pipe.c new file mode 100644 index 0000000000..8bf8c21439 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_winsys_pipe.c @@ -0,0 +1,338 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include +#include +//#include "dri_bufpool.h" +//#include "dri_bufmgr.h" + +#include "intel_context.h" +#include "intel_winsys.h" +#include "intel_swapbuffers.h" +#include "intel_batchbuffer.h" + +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" + + + +struct intel_pipe_winsys { + struct pipe_winsys winsys; + struct _DriBufferPool *regionPool; + struct _DriFreeSlabManager *fMan; +}; + + + +/* Turn a pipe winsys into an intel/pipe winsys: + */ +static inline struct intel_pipe_winsys * +intel_pipe_winsys( struct pipe_winsys *winsys ) +{ + return (struct intel_pipe_winsys *)winsys; +} + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void *intel_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags ) +{ + unsigned drm_flags = 0; + + if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) + drm_flags |= DRM_BO_FLAG_WRITE; + + if (flags & PIPE_BUFFER_USAGE_CPU_READ) + drm_flags |= DRM_BO_FLAG_READ; + + return driBOMap( dri_bo(buf), drm_flags, 0 ); +} + +static void intel_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + driBOUnmap( dri_bo(buf) ); +} + + +static void +intel_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + driBOUnReference( dri_bo(buf) ); + FREE(buf); +} + + +/* Pipe has no concept of pools. We choose the tex/region pool + * for all buffers. + * Grabs the hardware lock! + */ +static struct pipe_buffer * +intel_buffer_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned usage, + unsigned size ) +{ + struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); + struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + unsigned flags = 0; + + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + + if (usage & (PIPE_BUFFER_USAGE_VERTEX /*| IWS_BUFFER_USAGE_LOCAL*/)) { + flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; + } else { + flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; + } + + if (usage & PIPE_BUFFER_USAGE_GPU_READ) + flags |= DRM_BO_FLAG_READ; + + if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) + flags |= DRM_BO_FLAG_WRITE; + + /* drm complains if we don't set any read/write flags. + */ + if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) + flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + +#if 0 + if (flags & IWS_BUFFER_USAGE_EXE) + flags |= DRM_BO_FLAG_EXE; + + if (usage & IWS_BUFFER_USAGE_CACHED) + flags |= DRM_BO_FLAG_CACHED; +#endif + + driGenBuffers( iws->regionPool, + "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); + + driBOData( buffer->driBO, size, NULL, iws->regionPool, 0 ); + + return &buffer->base; +} + + +static struct pipe_buffer * +intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) +{ + struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); + struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + + driGenUserBuffer( iws->regionPool, + "pipe user buffer", &buffer->driBO, ptr, bytes ); + + buffer->base.refcount = 1; + + return &buffer->base; +} + + +/* The state tracker (should!) keep track of whether the fake + * frontbuffer has been touched by any rendering since the last time + * we copied its contents to the real frontbuffer. Our task is easy: + */ +static void +intel_flush_frontbuffer( struct pipe_winsys *winsys, + struct pipe_surface *surf, + void *context_private) +{ +#if 0 + struct intel_context *intel = (struct intel_context *) context_private; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelDisplaySurface(dPriv, surf, NULL); +#endif +} + + +static struct pipe_surface * +intel_i915_surface_alloc(struct pipe_winsys *winsys) +{ + struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); + if (surf) { + surf->refcount = 1; + surf->winsys = winsys; + } + return surf; +} + + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + +/** + * Copied from xm_winsys.c + */ +static int +intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags) +{ + const unsigned alignment = 64; + //int ret; + + surf->width = width; + surf->height = height; + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = round_up(width, alignment / surf->cpp); + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->cpp * height); + if(!surf->buffer) + return -1; + + return 0; +} + + +static void +intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + surf->refcount--; + if (surf->refcount == 0) { + if (surf->buffer) + pipe_buffer_reference(winsys, &surf->buffer, NULL); + free(surf); + } + *s = NULL; +} + + + +static const char * +intel_get_name( struct pipe_winsys *winsys ) +{ + return "Intel/EGL/ttm"; +} + +static void +intel_fence_reference( struct pipe_winsys *sws, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence ) +{ + if (*ptr) + driFenceUnReference((struct _DriFenceObject **)ptr); + + if (fence) + *ptr = (struct pipe_fence_handle *)driFenceReference((struct _DriFenceObject *)fence); +} + +static int +intel_fence_signalled( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ) +{ + return driFenceSignaled((struct _DriFenceObject *)fence, flag); +} + +static int +intel_fence_finish( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ) +{ + /* JB: Lets allways lazy wait */ + return driFenceFinish((struct _DriFenceObject *)fence, flag, 1); +} + +struct pipe_winsys * +intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) +{ + struct intel_pipe_winsys *iws = CALLOC_STRUCT( intel_pipe_winsys ); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + * + * Pipe would be happy with a malloc based memory manager, but + * the SwapBuffers implementation in this winsys driver requires + * that rendering be done to an appropriate _DriBufferObject. + */ + iws->winsys.buffer_create = intel_buffer_create; + iws->winsys.user_buffer_create = intel_user_buffer_create; + iws->winsys.buffer_map = intel_buffer_map; + iws->winsys.buffer_unmap = intel_buffer_unmap; + iws->winsys.buffer_destroy = intel_buffer_destroy; + iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; + iws->winsys.get_name = intel_get_name; + iws->winsys.surface_alloc = intel_i915_surface_alloc; + iws->winsys.surface_alloc_storage = intel_i915_surface_alloc_storage; + iws->winsys.surface_release = intel_i915_surface_release; + + iws->winsys.fence_reference = intel_fence_reference; + iws->winsys.fence_signalled = intel_fence_signalled; + iws->winsys.fence_finish = intel_fence_finish; + + if (fd) + iws->regionPool = driSlabPoolInit(fd, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + 64, 6, 16, 4096, 0, + fMan); + + return &iws->winsys; +} + + +void +intel_destroy_pipe_winsys( struct pipe_winsys *winsys ) +{ + struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); + if (iws->regionPool) { + driPoolTakeDown(iws->regionPool); + } + free(iws); +} + diff --git a/src/gallium/winsys/egl_drm/intel/intel_winsys_softpipe.c b/src/gallium/winsys/egl_drm/intel/intel_winsys_softpipe.c new file mode 100644 index 0000000000..0bc2dc4002 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_winsys_softpipe.c @@ -0,0 +1,82 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include "intel_context.h" +#include "intel_winsys.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_format.h" +#include "softpipe/sp_winsys.h" + + +struct intel_softpipe_winsys { + struct softpipe_winsys sws; + struct intel_context *intel; +}; + +/** + * Return list of surface formats supported by this driver. + */ +static boolean +intel_is_format_supported(struct softpipe_winsys *sws, + enum pipe_format format) +{ + switch(format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + return TRUE; + default: + return FALSE; + } +} + + +/** + * Create rendering context which uses software rendering. + */ +struct pipe_context * +intel_create_softpipe( struct intel_context *intel, + struct pipe_winsys *winsys ) +{ + struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); + struct pipe_screen *screen = softpipe_create_screen(winsys); + + /* Fill in this struct with callbacks that softpipe will need to + * communicate with the window system, buffer manager, etc. + */ + isws->sws.is_format_supported = intel_is_format_supported; + isws->intel = intel; + + /* Create the softpipe context: + */ + return softpipe_create( screen, winsys, &isws->sws ); +} diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c b/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c new file mode 100644 index 0000000000..1bc1089352 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c @@ -0,0 +1,953 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + * Keith Whitwell + */ + +#include +#include +#include "glthread.h" +#include "errno.h" +#include "ws_dri_bufmgr.h" +#include "string.h" +#include "imports.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" + +/* + * This lock is here to protect drmBO structs changing underneath us during a + * validate list call, since validatelist cannot take individiual locks for + * each drmBO. Validatelist takes this lock in write mode. Any access to an + * individual drmBO should take this lock in read mode, since in that case, the + * driBufferObject mutex will protect the access. Locking order is + * driBufferObject mutex - > this rw lock. + */ + +_glthread_DECLARE_STATIC_MUTEX(bmMutex); +_glthread_DECLARE_STATIC_COND(bmCond); + +static int kernelReaders = 0; +static int num_buffers = 0; +static int num_user_buffers = 0; + +static drmBO *drmBOListBuf(void *iterator) +{ + drmBONode *node; + drmMMListHead *l = (drmMMListHead *) iterator; + node = DRMLISTENTRY(drmBONode, l, head); + return node->buf; +} + +static void *drmBOListIterator(drmBOList *list) +{ + void *ret = list->list.next; + + if (ret == &list->list) + return NULL; + return ret; +} + +static void *drmBOListNext(drmBOList *list, void *iterator) +{ + void *ret; + + drmMMListHead *l = (drmMMListHead *) iterator; + ret = l->next; + if (ret == &list->list) + return NULL; + return ret; +} + +static drmBONode *drmAddListItem(drmBOList *list, drmBO *item, + uint64_t arg0, + uint64_t arg1) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->free.next; + if (l == &list->free) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + return NULL; + } + list->numCurrent++; + } + else { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + } + node->buf = item; + node->arg0 = arg0; + node->arg1 = arg1; + DRMLISTADD(&node->head, &list->list); + list->numOnList++; + return node; +} + +static int drmAddValidateItem(drmBOList *list, drmBO *buf, uint64_t flags, + uint64_t mask, int *newItem) +{ + drmBONode *node, *cur; + drmMMListHead *l; + + *newItem = 0; + cur = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + if (node->buf == buf) { + cur = node; + break; + } + } + if (!cur) { + cur = drmAddListItem(list, buf, flags, mask); + if (!cur) { + return -ENOMEM; + } + *newItem = 1; + cur->arg0 = flags; + cur->arg1 = mask; + } + else { + uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; + uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; + + if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { + return -EINVAL; + } + + cur->arg1 |= mask; + cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); + + if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && + (cur->arg0 & DRM_BO_MASK_MEM) == 0) { + return -EINVAL; + } + } + return 0; +} + +static void drmBOFreeList(drmBOList *list) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->list.next; + while(l != &list->list) { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + l = list->list.next; + list->numCurrent--; + list->numOnList--; + } + + l = list->free.next; + while(l != &list->free) { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + l = list->free.next; + list->numCurrent--; + } +} + +static int drmAdjustListNodes(drmBOList *list) +{ + drmBONode *node; + drmMMListHead *l; + int ret = 0; + + while(list->numCurrent < list->numTarget) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + ret = -ENOMEM; + break; + } + list->numCurrent++; + DRMLISTADD(&node->head, &list->free); + } + + while(list->numCurrent > list->numTarget) { + l = list->free.next; + if (l == &list->free) + break; + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + list->numCurrent--; + } + return ret; +} + +static int drmBOCreateList(int numTarget, drmBOList *list) +{ + DRMINITLISTHEAD(&list->list); + DRMINITLISTHEAD(&list->free); + list->numTarget = numTarget; + list->numCurrent = 0; + list->numOnList = 0; + return drmAdjustListNodes(list); +} + +static int drmBOResetList(drmBOList *list) +{ + drmMMListHead *l; + int ret; + + ret = drmAdjustListNodes(list); + if (ret) + return ret; + + l = list->list.next; + while (l != &list->list) { + DRMLISTDEL(l); + DRMLISTADD(l, &list->free); + list->numOnList--; + l = list->list.next; + } + return drmAdjustListNodes(list); +} + +void driWriteLockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + while(kernelReaders != 0) + _glthread_COND_WAIT(bmCond, bmMutex); +} + +void driWriteUnlockKernelBO(void) +{ + _glthread_UNLOCK_MUTEX(bmMutex); +} + +void driReadLockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + kernelReaders++; + _glthread_UNLOCK_MUTEX(bmMutex); +} + +void driReadUnlockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + if (--kernelReaders == 0) + _glthread_COND_BROADCAST(bmCond); + _glthread_UNLOCK_MUTEX(bmMutex); +} + + + + +/* + * TODO: Introduce fence pools in the same way as + * buffer object pools. + */ + +typedef struct _DriBufferObject +{ + DriBufferPool *pool; + _glthread_Mutex mutex; + int refCount; + const char *name; + uint64_t flags; + unsigned hint; + unsigned alignment; + unsigned createdByReference; + void *private; + /* user-space buffer: */ + unsigned userBuffer; + void *userData; + unsigned userSize; +} DriBufferObject; + +typedef struct _DriBufferList { + drmBOList drmBuffers; /* List of kernel buffers needing validation */ + drmBOList driBuffers; /* List of user-space buffers needing validation */ +} DriBufferList; + + +void +bmError(int val, const char *file, const char *function, int line) +{ + _mesa_printf("Fatal video memory manager error \"%s\".\n" + "Check kernel logs or set the LIBGL_DEBUG\n" + "environment variable to \"verbose\" for more info.\n" + "Detected in file %s, line %d, function %s.\n", + strerror(-val), file, line, function); +#ifndef NDEBUG + abort(); +#else + abort(); +#endif +} + +extern drmBO * +driBOKernel(struct _DriBufferObject *buf) +{ + drmBO *ret; + + driReadLockKernelBO(); + _glthread_LOCK_MUTEX(buf->mutex); + assert(buf->private != NULL); + ret = buf->pool->kernel(buf->pool, buf->private); + if (!ret) + BM_CKFATAL(-EINVAL); + _glthread_UNLOCK_MUTEX(buf->mutex); + driReadUnlockKernelBO(); + + return ret; +} + +void +driBOWaitIdle(struct _DriBufferObject *buf, int lazy) +{ + + /* + * This function may block. Is it sane to keep the mutex held during + * that time?? + */ + + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy)); + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void * +driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint) +{ + void *virtual; + int retval; + + if (buf->userBuffer) { + return buf->userData; + } + + _glthread_LOCK_MUTEX(buf->mutex); + assert(buf->private != NULL); + retval = buf->pool->map(buf->pool, buf->private, flags, hint, + &buf->mutex, &virtual); + _glthread_UNLOCK_MUTEX(buf->mutex); + + return retval == 0 ? virtual : NULL; +} + +void +driBOUnmap(struct _DriBufferObject *buf) +{ + if (buf->userBuffer) + return; + + assert(buf->private != NULL); + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +unsigned long +driBOOffset(struct _DriBufferObject *buf) +{ + unsigned long ret; + + assert(buf->private != NULL); + + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->offset(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + return ret; +} + +unsigned long +driBOPoolOffset(struct _DriBufferObject *buf) +{ + unsigned long ret; + + assert(buf->private != NULL); + + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->poolOffset(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + return ret; +} + +uint64_t +driBOFlags(struct _DriBufferObject *buf) +{ + uint64_t ret; + + assert(buf->private != NULL); + + driReadLockKernelBO(); + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->flags(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + driReadUnlockKernelBO(); + return ret; +} + +struct _DriBufferObject * +driBOReference(struct _DriBufferObject *buf) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (++buf->refCount == 1) { + _glthread_UNLOCK_MUTEX(buf->mutex); + BM_CKFATAL(-EINVAL); + } + _glthread_UNLOCK_MUTEX(buf->mutex); + return buf; +} + +void +driBOUnReference(struct _DriBufferObject *buf) +{ + int tmp; + + if (!buf) + return; + + _glthread_LOCK_MUTEX(buf->mutex); + tmp = --buf->refCount; + if (!tmp) { + _glthread_UNLOCK_MUTEX(buf->mutex); + if (buf->private) { + if (buf->createdByReference) + buf->pool->unreference(buf->pool, buf->private); + else + buf->pool->destroy(buf->pool, buf->private); + } + if (buf->userBuffer) + num_user_buffers--; + else + num_buffers--; + free(buf); + } else + _glthread_UNLOCK_MUTEX(buf->mutex); + +} + + +int +driBOData(struct _DriBufferObject *buf, + unsigned size, const void *data, + DriBufferPool *newPool, + uint64_t flags) +{ + void *virtual = NULL; + int newBuffer; + int retval = 0; + struct _DriBufferPool *pool; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + pool = buf->pool; + + if (pool == NULL && newPool != NULL) { + buf->pool = newPool; + pool = newPool; + } + if (newPool == NULL) + newPool = pool; + + if (!pool->create) { + _mesa_error(NULL, GL_INVALID_OPERATION, + "driBOData called on invalid buffer\n"); + BM_CKFATAL(-EINVAL); + } + + newBuffer = (!buf->private || pool != newPool || + pool->size(pool, buf->private) < size); + + if (!flags) + flags = buf->flags; + + if (newBuffer) { + + if (buf->createdByReference) { + _mesa_error(NULL, GL_INVALID_OPERATION, + "driBOData requiring resizing called on " + "shared buffer.\n"); + BM_CKFATAL(-EINVAL); + } + + if (buf->private) + buf->pool->destroy(buf->pool, buf->private); + + pool = newPool; + buf->pool = newPool; + buf->private = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, + buf->alignment); + if (!buf->private) + retval = -ENOMEM; + + if (retval == 0) + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, + DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual); + } else if (pool->map(pool, buf->private, DRM_BO_FLAG_WRITE, + DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual)) { + /* + * Buffer is busy. need to create a new one. + */ + + void *newBuf; + + newBuf = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, + buf->alignment); + if (newBuf) { + buf->pool->destroy(buf->pool, buf->private); + buf->private = newBuf; + } + + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); + } else { + uint64_t flag_diff = flags ^ buf->flags; + + /* + * We might need to change buffer flags. + */ + + if (flag_diff){ + assert(pool->setStatus != NULL); + BM_CKFATAL(pool->unmap(pool, buf->private)); + BM_CKFATAL(pool->setStatus(pool, buf->private, flag_diff, + buf->flags)); + if (!data) + goto out; + + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); + } + } + + if (retval == 0) { + if (data) + memcpy(virtual, data, size); + + BM_CKFATAL(pool->unmap(pool, buf->private)); + } + + out: + _glthread_UNLOCK_MUTEX(buf->mutex); + + return retval; +} + +void +driBOSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, const void *data) +{ + void *virtual; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + if (size && data) { + BM_CKFATAL(buf->pool->map(buf->pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, + &virtual)); + memcpy((unsigned char *) virtual + offset, data, size); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + } + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void +driBOGetSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, void *data) +{ + void *virtual; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + if (size && data) { + BM_CKFATAL(buf->pool->map(buf->pool, buf->private, + DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual)); + memcpy(data, (unsigned char *) virtual + offset, size); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + } + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void +driBOSetReferenced(struct _DriBufferObject *buf, + unsigned long handle) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->private != NULL) { + _mesa_error(NULL, GL_INVALID_OPERATION, + "Invalid buffer for setReferenced\n"); + BM_CKFATAL(-EINVAL); + + } + if (buf->pool->reference == NULL) { + _mesa_error(NULL, GL_INVALID_OPERATION, + "Invalid buffer pool for setReferenced\n"); + BM_CKFATAL(-EINVAL); + } + buf->private = buf->pool->reference(buf->pool, handle); + if (!buf->private) { + _mesa_error(NULL, GL_OUT_OF_MEMORY, + "Invalid buffer pool for setStatic\n"); + BM_CKFATAL(-ENOMEM); + } + buf->createdByReference = GL_TRUE; + buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags; + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +int +driGenBuffers(struct _DriBufferPool *pool, + const char *name, + unsigned n, + struct _DriBufferObject *buffers[], + unsigned alignment, uint64_t flags, unsigned hint) +{ + struct _DriBufferObject *buf; + int i; + + flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM | + DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + + ++num_buffers; + + assert(pool); + + for (i = 0; i < n; ++i) { + buf = (struct _DriBufferObject *) calloc(1, sizeof(*buf)); + if (!buf) + return -ENOMEM; + + _glthread_INIT_MUTEX(buf->mutex); + _glthread_LOCK_MUTEX(buf->mutex); + buf->refCount = 1; + buf->flags = flags; + buf->hint = hint; + buf->name = name; + buf->alignment = alignment; + buf->pool = pool; + buf->createdByReference = 0; + _glthread_UNLOCK_MUTEX(buf->mutex); + buffers[i] = buf; + } + return 0; +} + +void +driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject **buffers, + void *ptr, unsigned bytes) +{ + const unsigned alignment = 1, flags = 0, hint = 0; + + --num_buffers; /* JB: is inced in GenBuffes */ + driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); + ++num_user_buffers; + + (*buffers)->userBuffer = 1; + (*buffers)->userData = ptr; + (*buffers)->userSize = bytes; +} + +void +driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]) +{ + int i; + + for (i = 0; i < n; ++i) { + driBOUnReference(buffers[i]); + } +} + + +void +driInitBufMgr(int fd) +{ + ; +} + +/* + * Note that lists are per-context and don't need mutex protection. + */ + +struct _DriBufferList * +driBOCreateList(int target) +{ + struct _DriBufferList *list = calloc(sizeof(*list), 1); + + BM_CKFATAL(drmBOCreateList(target, &list->drmBuffers)); + BM_CKFATAL(drmBOCreateList(target, &list->driBuffers)); + return list; +} + +int +driBOResetList(struct _DriBufferList * list) +{ + int ret; + ret = drmBOResetList(&list->drmBuffers); + if (ret) + return ret; + ret = drmBOResetList(&list->driBuffers); + return ret; +} + +void +driBOFreeList(struct _DriBufferList * list) +{ + drmBOFreeList(&list->drmBuffers); + drmBOFreeList(&list->driBuffers); + free(list); +} + + +/* + * Copied from libdrm, because it is needed by driAddValidateItem. + */ + +static drmBONode * +driAddListItem(drmBOList * list, drmBO * item, + uint64_t arg0, uint64_t arg1) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->free.next; + if (l == &list->free) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + return NULL; + } + list->numCurrent++; + } else { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + } + memset(&node->bo_arg, 0, sizeof(node->bo_arg)); + node->buf = item; + node->arg0 = arg0; + node->arg1 = arg1; + DRMLISTADDTAIL(&node->head, &list->list); + list->numOnList++; + return node; +} + +/* + * Slightly modified version compared to the libdrm version. + * This one returns the list index of the buffer put on the list. + */ + +static int +driAddValidateItem(drmBOList * list, drmBO * buf, uint64_t flags, + uint64_t mask, int *itemLoc, + struct _drmBONode **pnode) +{ + drmBONode *node, *cur; + drmMMListHead *l; + int count = 0; + + cur = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + if (node->buf == buf) { + cur = node; + break; + } + count++; + } + if (!cur) { + cur = driAddListItem(list, buf, flags, mask); + if (!cur) + return -ENOMEM; + + cur->arg0 = flags; + cur->arg1 = mask; + } else { + uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; + uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; + + if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { + return -EINVAL; + } + + cur->arg1 |= mask; + cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); + + if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && + (cur->arg0 & DRM_BO_MASK_MEM) == 0) { + return -EINVAL; + } + } + *itemLoc = count; + *pnode = cur; + return 0; +} + + +void +driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf, + uint64_t flags, uint64_t mask, int *itemLoc, + struct _drmBONode **node) +{ + int newItem; + + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(driAddValidateItem(&list->drmBuffers, + buf->pool->kernel(buf->pool, buf->private), + flags, mask, itemLoc, node)); + BM_CKFATAL(drmAddValidateItem(&list->driBuffers, (drmBO *) buf, + flags, mask, &newItem)); + if (newItem) + buf->refCount++; + + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +drmBOList *driGetdrmBOList(struct _DriBufferList *list) +{ + driWriteLockKernelBO(); + return &list->drmBuffers; +} + +void driPutdrmBOList(struct _DriBufferList *list) +{ + driWriteUnlockKernelBO(); +} + + +void +driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->pool->fence) + BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence)); + _glthread_UNLOCK_MUTEX(buf->mutex); + +} + +void +driBOUnrefUserList(struct _DriBufferList *list) +{ + struct _DriBufferObject *buf; + void *curBuf; + + curBuf = drmBOListIterator(&list->driBuffers); + while (curBuf) { + buf = (struct _DriBufferObject *)drmBOListBuf(curBuf); + driBOUnReference(buf); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } +} + +struct _DriFenceObject * +driBOFenceUserList(struct _DriFenceMgr *mgr, + struct _DriBufferList *list, const char *name, + drmFence *kFence) +{ + struct _DriFenceObject *fence; + struct _DriBufferObject *buf; + void *curBuf; + + fence = driFenceCreate(mgr, kFence->fence_class, kFence->type, + kFence, sizeof(*kFence)); + curBuf = drmBOListIterator(&list->driBuffers); + + /* + * User-space fencing callbacks. + */ + + while (curBuf) { + buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); + driBOFence(buf, fence); + driBOUnReference(buf); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } + + driBOResetList(list); + return fence; +} + +void +driBOValidateUserList(struct _DriBufferList * list) +{ + void *curBuf; + struct _DriBufferObject *buf; + + curBuf = drmBOListIterator(&list->driBuffers); + + /* + * User-space validation callbacks. + */ + + while (curBuf) { + buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->pool->validate) + BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex)); + _glthread_UNLOCK_MUTEX(buf->mutex); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } +} + + +void +driPoolTakeDown(struct _DriBufferPool *pool) +{ + pool->takeDown(pool); + +} + +unsigned long +driBOSize(struct _DriBufferObject *buf) +{ + unsigned long size; + + _glthread_LOCK_MUTEX(buf->mutex); + size = buf->pool->size(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + + return size; + +} + +drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list) +{ + return &list->drmBuffers; +} + +drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list) +{ + return &list->driBuffers; +} + diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h b/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h new file mode 100644 index 0000000000..fdaf5ee93a --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h @@ -0,0 +1,138 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + * Keith Whitwell + */ + +#ifndef _PSB_BUFMGR_H_ +#define _PSB_BUFMGR_H_ +#include +#include "i915_drm.h" +#include "ws_dri_fencemgr.h" + +typedef struct _drmBONode +{ + drmMMListHead head; + drmBO *buf; + struct drm_i915_op_arg bo_arg; + uint64_t arg0; + uint64_t arg1; +} drmBONode; + +typedef struct _drmBOList { + unsigned numTarget; + unsigned numCurrent; + unsigned numOnList; + drmMMListHead list; + drmMMListHead free; +} drmBOList; + + +struct _DriFenceObject; +struct _DriBufferObject; +struct _DriBufferPool; +struct _DriBufferList; + +/* + * Return a pointer to the libdrm buffer object this DriBufferObject + * uses. + */ + +extern drmBO *driBOKernel(struct _DriBufferObject *buf); +extern void *driBOMap(struct _DriBufferObject *buf, unsigned flags, + unsigned hint); +extern void driBOUnmap(struct _DriBufferObject *buf); +extern unsigned long driBOOffset(struct _DriBufferObject *buf); +extern unsigned long driBOPoolOffset(struct _DriBufferObject *buf); + +extern uint64_t driBOFlags(struct _DriBufferObject *buf); +extern struct _DriBufferObject *driBOReference(struct _DriBufferObject *buf); +extern void driBOUnReference(struct _DriBufferObject *buf); + +extern int driBOData(struct _DriBufferObject *r_buf, + unsigned size, const void *data, + struct _DriBufferPool *pool, uint64_t flags); + +extern void driBOSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, + const void *data); +extern void driBOGetSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, + void *data); +extern int driGenBuffers(struct _DriBufferPool *pool, + const char *name, + unsigned n, + struct _DriBufferObject *buffers[], + unsigned alignment, uint64_t flags, unsigned hint); +extern void driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject *buffers[], + void *ptr, unsigned bytes); +extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]); +extern void driInitBufMgr(int fd); +extern struct _DriBufferList *driBOCreateList(int target); +extern int driBOResetList(struct _DriBufferList * list); +extern void driBOAddListItem(struct _DriBufferList * list, + struct _DriBufferObject *buf, + uint64_t flags, uint64_t mask, int *itemLoc, + struct _drmBONode **node); + +extern void driBOValidateList(int fd, struct _DriBufferList * list); +extern void driBOFreeList(struct _DriBufferList * list); +extern struct _DriFenceObject *driBOFenceUserList(struct _DriFenceMgr *mgr, + struct _DriBufferList *list, + const char *name, + drmFence *kFence); +extern void driBOUnrefUserList(struct _DriBufferList *list); +extern void driBOValidateUserList(struct _DriBufferList * list); +extern drmBOList *driGetdrmBOList(struct _DriBufferList *list); +extern void driPutdrmBOList(struct _DriBufferList *list); + +extern void driBOFence(struct _DriBufferObject *buf, + struct _DriFenceObject *fence); + +extern void driPoolTakeDown(struct _DriBufferPool *pool); +extern void driBOSetReferenced(struct _DriBufferObject *buf, + unsigned long handle); +unsigned long driBOSize(struct _DriBufferObject *buf); +extern void driBOWaitIdle(struct _DriBufferObject *buf, int lazy); +extern void driPoolTakeDown(struct _DriBufferPool *pool); + +extern void driReadLockKernelBO(void); +extern void driReadUnlockKernelBO(void); +extern void driWriteLockKernelBO(void); +extern void driWriteUnlockKernelBO(void); + +/* + * For debugging purposes. + */ + +extern drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list); +extern drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list); +#endif diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h b/src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h new file mode 100644 index 0000000000..3a302e13d3 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h @@ -0,0 +1,102 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#ifndef _PSB_BUFPOOL_H_ +#define _PSB_BUFPOOL_H_ + +#include +#include +struct _DriFenceObject; + +typedef struct _DriBufferPool +{ + int fd; + int (*map) (struct _DriBufferPool * pool, void *private, + unsigned flags, int hint, _glthread_Mutex *mutex, + void **virtual); + int (*unmap) (struct _DriBufferPool * pool, void *private); + int (*destroy) (struct _DriBufferPool * pool, void *private); + unsigned long (*offset) (struct _DriBufferPool * pool, void *private); + unsigned long (*poolOffset) (struct _DriBufferPool * pool, void *private); + uint64_t (*flags) (struct _DriBufferPool * pool, void *private); + unsigned long (*size) (struct _DriBufferPool * pool, void *private); + void *(*create) (struct _DriBufferPool * pool, unsigned long size, + uint64_t flags, unsigned hint, unsigned alignment); + void *(*reference) (struct _DriBufferPool * pool, unsigned handle); + int (*unreference) (struct _DriBufferPool * pool, void *private); + int (*fence) (struct _DriBufferPool * pool, void *private, + struct _DriFenceObject * fence); + drmBO *(*kernel) (struct _DriBufferPool * pool, void *private); + int (*validate) (struct _DriBufferPool * pool, void *private, _glthread_Mutex *mutex); + int (*waitIdle) (struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, + int lazy); + int (*setStatus) (struct _DriBufferPool *pool, void *private, + uint64_t flag_diff, uint64_t old_flags); + void (*takeDown) (struct _DriBufferPool * pool); + void *data; +} DriBufferPool; + +extern void bmError(int val, const char *file, const char *function, + int line); +#define BM_CKFATAL(val) \ + do{ \ + int tstVal = (val); \ + if (tstVal) \ + bmError(tstVal, __FILE__, __FUNCTION__, __LINE__); \ + } while(0); + + +/* + * Builtin pools. + */ + +/* + * Kernel buffer objects. Size in multiples of page size. Page size aligned. + */ + +extern struct _DriBufferPool *driDRMPoolInit(int fd); +extern struct _DriBufferPool *driMallocPoolInit(void); + +struct _DriFreeSlabManager; +extern struct _DriBufferPool * driSlabPoolInit(int fd, uint64_t flags, + uint64_t validMask, + uint32_t smallestSize, + uint32_t numSizes, + uint32_t desiredNumBuffers, + uint32_t maxSlabSize, + uint32_t pageAlignment, + struct _DriFreeSlabManager *fMan); +extern void driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan); +extern struct _DriFreeSlabManager * +driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec); + + +#endif diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c b/src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c new file mode 100644 index 0000000000..7c55dbc674 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c @@ -0,0 +1,268 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" +#include "assert.h" + +/* + * Buffer pool implementation using DRM buffer objects as DRI buffer objects. + */ + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, uint64_t flags, unsigned hint, + unsigned alignment) +{ + drmBO *buf = (drmBO *) malloc(sizeof(*buf)); + int ret; + unsigned pageSize = getpagesize(); + + if (!buf) + return NULL; + + if ((alignment > pageSize) && (alignment % pageSize)) { + free(buf); + return NULL; + } + + ret = drmBOCreate(pool->fd, size, alignment / pageSize, + NULL, + flags, hint, buf); + if (ret) { + free(buf); + return NULL; + } + + return (void *) buf; +} + +static void * +pool_reference(struct _DriBufferPool *pool, unsigned handle) +{ + drmBO *buf = (drmBO *) malloc(sizeof(*buf)); + int ret; + + if (!buf) + return NULL; + + ret = drmBOReference(pool->fd, handle, buf); + + if (ret) { + free(buf); + return NULL; + } + + return (void *) buf; +} + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + int ret; + drmBO *buf = (drmBO *) private; + driReadLockKernelBO(); + ret = drmBOUnreference(pool->fd, buf); + free(buf); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_unreference(struct _DriBufferPool *pool, void *private) +{ + int ret; + drmBO *buf = (drmBO *) private; + driReadLockKernelBO(); + ret = drmBOUnreference(pool->fd, buf); + free(buf); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOMap(pool->fd, buf, flags, hint, virtual); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOUnmap(pool->fd, buf); + driReadUnlockKernelBO(); + + return ret; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + unsigned long offset; + + driReadLockKernelBO(); + assert(buf->flags & DRM_BO_FLAG_NO_MOVE); + offset = buf->offset; + driReadUnlockKernelBO(); + + return buf->offset; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + return 0; +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + uint64_t flags; + + driReadLockKernelBO(); + flags = buf->flags; + driReadUnlockKernelBO(); + + return flags; +} + + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + unsigned long size; + + driReadLockKernelBO(); + size = buf->size; + driReadUnlockKernelBO(); + + return buf->size; +} + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + /* + * Noop. The kernel handles all fencing. + */ + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + return (drmBO *) private; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, + int lazy) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOWaitIdle(pool->fd, buf, (lazy) ? DRM_BO_HINT_WAIT_LAZY:0); + driReadUnlockKernelBO(); + + return ret; +} + + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + free(pool); +} + +/*static int +pool_setStatus(struct _DriBufferPool *pool, void *private, + uint64_t flag_diff, uint64_t old_flags) +{ + drmBO *buf = (drmBO *) private; + uint64_t new_flags = old_flags ^ flag_diff; + int ret; + + driReadLockKernelBO(); + ret = drmBOSetStatus(pool->fd, buf, new_flags, flag_diff, + 0, 0, 0); + driReadUnlockKernelBO(); + return ret; +}*/ + +struct _DriBufferPool * +driDRMPoolInit(int fd) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + + if (!pool) + return NULL; + + pool->fd = fd; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->poolOffset = &pool_poolOffset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = NULL; + pool->waitIdle = &pool_waitIdle; + pool->takeDown = &pool_takedown; + pool->reference = &pool_reference; + pool->unreference = &pool_unreference; + pool->data = NULL; + return pool; +} diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c b/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c new file mode 100644 index 0000000000..1f893b47ce --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c @@ -0,0 +1,377 @@ +#include "ws_dri_fencemgr.h" +#include "glthread.h" +#include +#include +#include + +/* + * Note: Locking order is + * _DriFenceObject::mutex + * _DriFenceMgr::mutex + */ + +struct _DriFenceMgr { + /* + * Constant members. Need no mutex protection. + */ + struct _DriFenceMgrCreateInfo info; + void *private; + + /* + * These members are protected by this->mutex + */ + _glthread_Mutex mutex; + int refCount; + drmMMListHead *heads; + int num_fences; +}; + +struct _DriFenceObject { + + /* + * These members are constant and need no mutex protection. + */ + struct _DriFenceMgr *mgr; + uint32_t fence_class; + uint32_t fence_type; + + /* + * These members are protected by mgr->mutex. + */ + drmMMListHead head; + int refCount; + + /* + * These members are protected by this->mutex. + */ + _glthread_Mutex mutex; + uint32_t signaled_type; + void *private; +}; + +uint32_t +driFenceType(struct _DriFenceObject *fence) +{ + return fence->fence_type; +} + +struct _DriFenceMgr * +driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info) +{ + struct _DriFenceMgr *tmp; + uint32_t i; + + tmp = calloc(1, sizeof(*tmp)); + if (!tmp) + return NULL; + + _glthread_INIT_MUTEX(tmp->mutex); + _glthread_LOCK_MUTEX(tmp->mutex); + tmp->refCount = 1; + tmp->info = *info; + tmp->num_fences = 0; + tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads)); + if (!tmp->heads) + goto out_err; + + for (i=0; iinfo.num_classes; ++i) { + DRMINITLISTHEAD(&tmp->heads[i]); + } + _glthread_UNLOCK_MUTEX(tmp->mutex); + return tmp; + + out_err: + if (tmp) + free(tmp); + return NULL; +} + +static void +driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr) +{ + struct _DriFenceMgr *mgr = *pMgr; + + *pMgr = NULL; + if (--mgr->refCount == 0) + free(mgr); + else + _glthread_UNLOCK_MUTEX(mgr->mutex); +} + +void +driFenceMgrUnReference(struct _DriFenceMgr **pMgr) +{ + _glthread_LOCK_MUTEX((*pMgr)->mutex); + driFenceMgrUnrefUnlock(pMgr); +} + +static void +driFenceUnReferenceLocked(struct _DriFenceObject **pFence) +{ + struct _DriFenceObject *fence = *pFence; + struct _DriFenceMgr *mgr = fence->mgr; + + *pFence = NULL; + if (--fence->refCount == 0) { + DRMLISTDELINIT(&fence->head); + if (fence->private) + mgr->info.unreference(mgr, &fence->private); + --mgr->num_fences; + fence->mgr = NULL; + --mgr->refCount; + free(fence); + + } +} + + +static void +driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr, + drmMMListHead *list, + uint32_t fence_class, + uint32_t fence_type) +{ + struct _DriFenceObject *entry; + drmMMListHead *prev; + + while(list != &mgr->heads[fence_class]) { + entry = DRMLISTENTRY(struct _DriFenceObject, list, head); + + /* + * Up refcount so that entry doesn't disappear from under us + * when we unlock-relock mgr to get the correct locking order. + */ + + ++entry->refCount; + _glthread_UNLOCK_MUTEX(mgr->mutex); + _glthread_LOCK_MUTEX(entry->mutex); + _glthread_LOCK_MUTEX(mgr->mutex); + + prev = list->prev; + + + + if (list->prev == list) { + + /* + * Somebody else removed the entry from the list. + */ + + _glthread_UNLOCK_MUTEX(entry->mutex); + driFenceUnReferenceLocked(&entry); + return; + } + + entry->signaled_type |= (fence_type & entry->fence_type); + if (entry->signaled_type == entry->fence_type) { + DRMLISTDELINIT(list); + mgr->info.unreference(mgr, &entry->private); + } + _glthread_UNLOCK_MUTEX(entry->mutex); + driFenceUnReferenceLocked(&entry); + list = prev; + } +} + + +int +driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, + int lazy_hint) +{ + struct _DriFenceMgr *mgr = fence->mgr; + int ret = 0; + + _glthread_LOCK_MUTEX(fence->mutex); + + if ((fence->signaled_type & fence_type) == fence_type) + goto out0; + + ret = mgr->info.finish(mgr, fence->private, fence_type, lazy_hint); + if (ret) + goto out0; + + _glthread_LOCK_MUTEX(mgr->mutex); + _glthread_UNLOCK_MUTEX(fence->mutex); + + driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, + fence_type); + _glthread_UNLOCK_MUTEX(mgr->mutex); + return 0; + + out0: + _glthread_UNLOCK_MUTEX(fence->mutex); + return ret; +} + +uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence) +{ + uint32_t ret; + + _glthread_LOCK_MUTEX(fence->mutex); + ret = fence->signaled_type; + _glthread_UNLOCK_MUTEX(fence->mutex); + + return ret; +} + +int +driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type, + uint32_t *signaled) +{ + int ret = 0; + struct _DriFenceMgr *mgr; + + _glthread_LOCK_MUTEX(fence->mutex); + mgr = fence->mgr; + *signaled = fence->signaled_type; + if ((fence->signaled_type & flush_type) == flush_type) + goto out0; + + ret = mgr->info.signaled(mgr, fence->private, flush_type, signaled); + if (ret) { + *signaled = fence->signaled_type; + goto out0; + } + + if ((fence->signaled_type | *signaled) == fence->signaled_type) + goto out0; + + _glthread_LOCK_MUTEX(mgr->mutex); + _glthread_UNLOCK_MUTEX(fence->mutex); + + driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, + *signaled); + + _glthread_UNLOCK_MUTEX(mgr->mutex); + return 0; + out0: + _glthread_UNLOCK_MUTEX(fence->mutex); + return ret; +} + +struct _DriFenceObject * +driFenceReference(struct _DriFenceObject *fence) +{ + _glthread_LOCK_MUTEX(fence->mgr->mutex); + ++fence->refCount; + _glthread_UNLOCK_MUTEX(fence->mgr->mutex); + return fence; +} + +void +driFenceUnReference(struct _DriFenceObject **pFence) +{ + struct _DriFenceMgr *mgr; + + if (*pFence == NULL) + return; + + mgr = (*pFence)->mgr; + _glthread_LOCK_MUTEX(mgr->mutex); + ++mgr->refCount; + driFenceUnReferenceLocked(pFence); + driFenceMgrUnrefUnlock(&mgr); +} + +struct _DriFenceObject +*driFenceCreate(struct _DriFenceMgr *mgr, uint32_t fence_class, + uint32_t fence_type, void *private, size_t private_size) +{ + struct _DriFenceObject *fence; + size_t fence_size = sizeof(*fence); + + if (private_size) + fence_size = ((fence_size + 15) & ~15); + + fence = calloc(1, fence_size + private_size); + + if (!fence) { + int ret = mgr->info.finish(mgr, private, fence_type, 0); + + if (ret) + usleep(10000000); + + return NULL; + } + + _glthread_INIT_MUTEX(fence->mutex); + _glthread_LOCK_MUTEX(fence->mutex); + _glthread_LOCK_MUTEX(mgr->mutex); + fence->refCount = 1; + DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]); + fence->mgr = mgr; + ++mgr->refCount; + ++mgr->num_fences; + _glthread_UNLOCK_MUTEX(mgr->mutex); + fence->fence_class = fence_class; + fence->fence_type = fence_type; + fence->signaled_type = 0; + fence->private = private; + if (private_size) { + fence->private = (void *)(((uint8_t *) fence) + fence_size); + memcpy(fence->private, private, private_size); + } + + _glthread_UNLOCK_MUTEX(fence->mutex); + return fence; +} + + +static int +tSignaled(struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, + uint32_t *signaled_type) +{ + long fd = (long) mgr->private; + int dummy; + drmFence *fence = (drmFence *) private; + int ret; + + *signaled_type = 0; + ret = drmFenceSignaled((int) fd, fence, flush_type, &dummy); + if (ret) + return ret; + + *signaled_type = fence->signaled; + + return 0; +} + +static int +tFinish(struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, + int lazy_hint) +{ + long fd = (long) mgr->private; + unsigned flags = lazy_hint ? DRM_FENCE_FLAG_WAIT_LAZY : 0; + + return drmFenceWait((int)fd, flags, (drmFence *) private, fence_type); +} + +static int +tUnref(struct _DriFenceMgr *mgr, void **private) +{ + long fd = (long) mgr->private; + drmFence *fence = (drmFence *) *private; + *private = NULL; + + return drmFenceUnreference(fd, fence); +} + +struct _DriFenceMgr *driFenceMgrTTMInit(int fd) +{ + struct _DriFenceMgrCreateInfo info; + struct _DriFenceMgr *mgr; + + info.flags = DRI_FENCE_CLASS_ORDERED; + info.num_classes = 4; + info.signaled = tSignaled; + info.finish = tFinish; + info.unreference = tUnref; + + mgr = driFenceMgrCreate(&info); + if (mgr == NULL) + return NULL; + + mgr->private = (void *) (long) fd; + return mgr; +} + diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h b/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h new file mode 100644 index 0000000000..4ea58dfe18 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h @@ -0,0 +1,115 @@ +#ifndef DRI_FENCEMGR_H +#define DRI_FENCEMGR_H + +#include +#include + +struct _DriFenceObject; +struct _DriFenceMgr; + +/* + * Do a quick check to see if the fence manager has registered the fence + * object as signaled. Note that this function may return a false negative + * answer. + */ +extern uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence); + +/* + * Check if the fence object is signaled. This function can be substantially + * more expensive to call than the above function, but will not return a false + * negative answer. The argument "flush_type" sets the types that the + * underlying mechanism must make sure will eventually signal. + */ +extern int driFenceSignaledType(struct _DriFenceObject *fence, + uint32_t flush_type, uint32_t *signaled); + +/* + * Convenience functions. + */ + +static inline int driFenceSignaled(struct _DriFenceObject *fence, + uint32_t flush_type) +{ + uint32_t signaled_types; + int ret = driFenceSignaledType(fence, flush_type, &signaled_types); + if (ret) + return 0; + return ((signaled_types & flush_type) == flush_type); +} + +static inline int driFenceSignaledCached(struct _DriFenceObject *fence, + uint32_t flush_type) +{ + uint32_t signaled_types = + driFenceSignaledTypeCached(fence); + + return ((signaled_types & flush_type) == flush_type); +} + +/* + * Reference a fence object. + */ +extern struct _DriFenceObject *driFenceReference(struct _DriFenceObject *fence); + +/* + * Unreference a fence object. The fence object pointer will be reset to NULL. + */ + +extern void driFenceUnReference(struct _DriFenceObject **pFence); + + +/* + * Wait for a fence to signal the indicated fence_type. + * If "lazy_hint" is true, it indicates that the wait may sleep to avoid + * busy-wait polling. + */ +extern int driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, + int lazy_hint); + +/* + * Create a DriFenceObject for manager "mgr". + * + * "private" is a pointer that should be used for the callbacks in + * struct _DriFenceMgrCreateInfo. + * + * if private_size is nonzero, then the info stored at *private, with size + * private size will be copied and the fence manager will instead use a + * pointer to the copied data for the callbacks in + * struct _DriFenceMgrCreateInfo. In that case, the object pointed to by + * "private" may be destroyed after the call to driFenceCreate. + */ +extern struct _DriFenceObject *driFenceCreate(struct _DriFenceMgr *mgr, + uint32_t fence_class, + uint32_t fence_type, + void *private, + size_t private_size); + +extern uint32_t driFenceType(struct _DriFenceObject *fence); + +/* + * Fence creations are ordered. If a fence signals a fence_type, + * it is safe to assume that all fences of the same class that was + * created before that fence has signaled the same type. + */ + +#define DRI_FENCE_CLASS_ORDERED (1 << 0) + +struct _DriFenceMgrCreateInfo { + uint32_t flags; + uint32_t num_classes; + int (*signaled) (struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, + uint32_t *signaled_type); + int (*finish) (struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, int lazy_hint); + int (*unreference) (struct _DriFenceMgr *mgr, void **private); +}; + +extern struct _DriFenceMgr * +driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info); + +void +driFenceMgrUnReference(struct _DriFenceMgr **pMgr); + +extern struct _DriFenceMgr * +driFenceMgrTTMInit(int fd); + +#endif diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_mallocpool.c b/src/gallium/winsys/egl_drm/intel/ws_dri_mallocpool.c new file mode 100644 index 0000000000..bf97d7e440 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/ws_dri_mallocpool.c @@ -0,0 +1,162 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "imports.h" +#include "glthread.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" +#include "intel_screen.h" + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, uint64_t flags, unsigned hint, + unsigned alignment) +{ + unsigned long *private = malloc(size + 2*sizeof(unsigned long)); + if ((flags & DRM_BO_MASK_MEM) != DRM_BO_FLAG_MEM_LOCAL) + abort(); + + *private = size; + return (void *)private; +} + + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + free(private); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, + _glthread_Mutex *mutex, int lazy) +{ + return 0; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + *virtual = (void *)((unsigned long *)private + 2); + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + /* + * BUG + */ + abort(); + return 0UL; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + /* + * BUG + */ + abort(); +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + return DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + return *(unsigned long *) private; +} + + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + abort(); + return 0UL; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + abort(); + return NULL; +} + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + free(pool); +} + + +struct _DriBufferPool * +driMallocPoolInit(void) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + if (!pool) + return NULL; + + pool->data = NULL; + pool->fd = -1; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->poolOffset = &pool_poolOffset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = NULL; + pool->waitIdle = &pool_waitIdle; + pool->takeDown = &pool_takedown; + return pool; +} diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c b/src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c new file mode 100644 index 0000000000..e69519a01f --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c @@ -0,0 +1,970 @@ +/************************************************************************** + * + * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellstrom + */ + +#include +#include +#include +#include +#include +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" +#include "ws_dri_bufmgr.h" +#include "glthread.h" + +#define DRI_SLABPOOL_ALLOC_RETRIES 100 + +struct _DriSlab; + +struct _DriSlabBuffer { + int isSlabBuffer; + drmBO *bo; + struct _DriFenceObject *fence; + struct _DriSlab *parent; + drmMMListHead head; + uint32_t mapCount; + uint32_t start; + uint32_t fenceType; + int unFenced; + _glthread_Cond event; +}; + +struct _DriKernelBO { + int fd; + drmBO bo; + drmMMListHead timeoutHead; + drmMMListHead head; + struct timeval timeFreed; + uint32_t pageAlignment; + void *virtual; +}; + +struct _DriSlab{ + drmMMListHead head; + drmMMListHead freeBuffers; + uint32_t numBuffers; + uint32_t numFree; + struct _DriSlabBuffer *buffers; + struct _DriSlabSizeHeader *header; + struct _DriKernelBO *kbo; +}; + + +struct _DriSlabSizeHeader { + drmMMListHead slabs; + drmMMListHead freeSlabs; + drmMMListHead delayedBuffers; + uint32_t numDelayed; + struct _DriSlabPool *slabPool; + uint32_t bufSize; + _glthread_Mutex mutex; +}; + +struct _DriFreeSlabManager { + struct timeval slabTimeout; + struct timeval checkInterval; + struct timeval nextCheck; + drmMMListHead timeoutList; + drmMMListHead unCached; + drmMMListHead cached; + _glthread_Mutex mutex; +}; + + +struct _DriSlabPool { + + /* + * The data of this structure remains constant after + * initialization and thus needs no mutex protection. + */ + + struct _DriFreeSlabManager *fMan; + uint64_t proposedFlags; + uint64_t validMask; + uint32_t *bucketSizes; + uint32_t numBuckets; + uint32_t pageSize; + int fd; + int pageAlignment; + int maxSlabSize; + int desiredNumBuffers; + struct _DriSlabSizeHeader *headers; +}; + +/* + * FIXME: Perhaps arrange timeout slabs in size buckets for fast + * retreival?? + */ + + +static inline int +driTimeAfterEq(struct timeval *arg1, struct timeval *arg2) +{ + return ((arg1->tv_sec > arg2->tv_sec) || + ((arg1->tv_sec == arg2->tv_sec) && + (arg1->tv_usec > arg2->tv_usec))); +} + +static inline void +driTimeAdd(struct timeval *arg, struct timeval *add) +{ + unsigned int sec; + + arg->tv_sec += add->tv_sec; + arg->tv_usec += add->tv_usec; + sec = arg->tv_usec / 1000000; + arg->tv_sec += sec; + arg->tv_usec -= sec*1000000; +} + +static void +driFreeKernelBO(struct _DriKernelBO *kbo) +{ + if (!kbo) + return; + + (void) drmBOUnreference(kbo->fd, &kbo->bo); + free(kbo); +} + + +static void +driFreeTimeoutKBOsLocked(struct _DriFreeSlabManager *fMan, + struct timeval *time) +{ + drmMMListHead *list, *next; + struct _DriKernelBO *kbo; + + if (!driTimeAfterEq(time, &fMan->nextCheck)) + return; + + for (list = fMan->timeoutList.next, next = list->next; + list != &fMan->timeoutList; + list = next, next = list->next) { + + kbo = DRMLISTENTRY(struct _DriKernelBO, list, timeoutHead); + + if (!driTimeAfterEq(time, &kbo->timeFreed)) + break; + + DRMLISTDELINIT(&kbo->timeoutHead); + DRMLISTDELINIT(&kbo->head); + driFreeKernelBO(kbo); + } + + fMan->nextCheck = *time; + driTimeAdd(&fMan->nextCheck, &fMan->checkInterval); +} + + +/* + * Add a _DriKernelBO to the free slab manager. + * This means that it is available for reuse, but if it's not + * reused in a while, it will be freed. + */ + +static void +driSetKernelBOFree(struct _DriFreeSlabManager *fMan, + struct _DriKernelBO *kbo) +{ + struct timeval time; + + _glthread_LOCK_MUTEX(fMan->mutex); + gettimeofday(&time, NULL); + driTimeAdd(&time, &fMan->slabTimeout); + + kbo->timeFreed = time; + + if (kbo->bo.flags & DRM_BO_FLAG_CACHED) + DRMLISTADD(&kbo->head, &fMan->cached); + else + DRMLISTADD(&kbo->head, &fMan->unCached); + + DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList); + driFreeTimeoutKBOsLocked(fMan, &time); + + _glthread_UNLOCK_MUTEX(fMan->mutex); +} + +/* + * Get a _DriKernelBO for us to use as storage for a slab. + * + */ + +static struct _DriKernelBO * +driAllocKernelBO(struct _DriSlabSizeHeader *header) + +{ + struct _DriSlabPool *slabPool = header->slabPool; + struct _DriFreeSlabManager *fMan = slabPool->fMan; + drmMMListHead *list, *next, *head; + uint32_t size = header->bufSize * slabPool->desiredNumBuffers; + struct _DriKernelBO *kbo; + struct _DriKernelBO *kboTmp; + int ret; + + /* + * FIXME: We should perhaps allow some variation in slabsize in order + * to efficiently reuse slabs. + */ + + size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize; + size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1); + _glthread_LOCK_MUTEX(fMan->mutex); + + kbo = NULL; + + retry: + head = (slabPool->proposedFlags & DRM_BO_FLAG_CACHED) ? + &fMan->cached : &fMan->unCached; + + for (list = head->next, next = list->next; + list != head; + list = next, next = list->next) { + + kboTmp = DRMLISTENTRY(struct _DriKernelBO, list, head); + + if ((kboTmp->bo.size == size) && + (slabPool->pageAlignment == 0 || + (kboTmp->pageAlignment % slabPool->pageAlignment) == 0)) { + + if (!kbo) + kbo = kboTmp; + + if ((kbo->bo.proposedFlags ^ slabPool->proposedFlags) == 0) + break; + + } + } + + if (kbo) { + DRMLISTDELINIT(&kbo->head); + DRMLISTDELINIT(&kbo->timeoutHead); + } + + _glthread_UNLOCK_MUTEX(fMan->mutex); + + if (kbo) { + uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags; + + ret = 0; + if (new_mask) { + ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, + new_mask, DRM_BO_HINT_DONT_FENCE, 0, 0); + } + if (ret == 0) + return kbo; + + driFreeKernelBO(kbo); + kbo = NULL; + goto retry; + } + + kbo = calloc(1, sizeof(struct _DriKernelBO)); + if (!kbo) + return NULL; + + kbo->fd = slabPool->fd; + DRMINITLISTHEAD(&kbo->head); + DRMINITLISTHEAD(&kbo->timeoutHead); + ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL, + slabPool->proposedFlags, + DRM_BO_HINT_DONT_FENCE, &kbo->bo); + if (ret) + goto out_err0; + + ret = drmBOMap(kbo->fd, &kbo->bo, + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, &kbo->virtual); + + if (ret) + goto out_err1; + + ret = drmBOUnmap(kbo->fd, &kbo->bo); + if (ret) + goto out_err1; + + return kbo; + + out_err1: + drmBOUnreference(kbo->fd, &kbo->bo); + out_err0: + free(kbo); + return NULL; +} + + +static int +driAllocSlab(struct _DriSlabSizeHeader *header) +{ + struct _DriSlab *slab; + struct _DriSlabBuffer *buf; + uint32_t numBuffers; + int ret; + int i; + + slab = calloc(1, sizeof(*slab)); + if (!slab) + return -ENOMEM; + + slab->kbo = driAllocKernelBO(header); + if (!slab->kbo) { + ret = -ENOMEM; + goto out_err0; + } + + numBuffers = slab->kbo->bo.size / header->bufSize; + + slab->buffers = calloc(numBuffers, sizeof(*slab->buffers)); + if (!slab->buffers) { + ret = -ENOMEM; + goto out_err1; + } + + DRMINITLISTHEAD(&slab->head); + DRMINITLISTHEAD(&slab->freeBuffers); + slab->numBuffers = numBuffers; + slab->numFree = 0; + slab->header = header; + + buf = slab->buffers; + for (i=0; i < numBuffers; ++i) { + buf->parent = slab; + buf->start = i* header->bufSize; + buf->mapCount = 0; + buf->isSlabBuffer = 1; + _glthread_INIT_COND(buf->event); + DRMLISTADDTAIL(&buf->head, &slab->freeBuffers); + slab->numFree++; + buf++; + } + + DRMLISTADDTAIL(&slab->head, &header->slabs); + + return 0; + + out_err1: + driSetKernelBOFree(header->slabPool->fMan, slab->kbo); + free(slab->buffers); + out_err0: + free(slab); + return ret; +} + +/* + * Delete a buffer from the slab header delayed list and put + * it on the slab free list. + */ + +static void +driSlabFreeBufferLocked(struct _DriSlabBuffer *buf) +{ + struct _DriSlab *slab = buf->parent; + struct _DriSlabSizeHeader *header = slab->header; + drmMMListHead *list = &buf->head; + + DRMLISTDEL(list); + DRMLISTADDTAIL(list, &slab->freeBuffers); + slab->numFree++; + + if (slab->head.next == &slab->head) + DRMLISTADDTAIL(&slab->head, &header->slabs); + + if (slab->numFree == slab->numBuffers) { + list = &slab->head; + DRMLISTDEL(list); + DRMLISTADDTAIL(list, &header->freeSlabs); + } + + if (header->slabs.next == &header->slabs || + slab->numFree != slab->numBuffers) { + + drmMMListHead *next; + struct _DriFreeSlabManager *fMan = header->slabPool->fMan; + + for (list = header->freeSlabs.next, next = list->next; + list != &header->freeSlabs; + list = next, next = list->next) { + + slab = DRMLISTENTRY(struct _DriSlab, list, head); + + DRMLISTDELINIT(list); + driSetKernelBOFree(fMan, slab->kbo); + free(slab->buffers); + free(slab); + } + } +} + +static void +driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait) +{ + drmMMListHead *list, *prev, *first; + struct _DriSlabBuffer *buf; + struct _DriSlab *slab; + int firstWasSignaled = 1; + int signaled; + int i; + int ret; + + /* + * Rerun the freeing test if the youngest tested buffer + * was signaled, since there might be more idle buffers + * in the delay list. + */ + + while (firstWasSignaled) { + firstWasSignaled = 0; + signaled = 0; + first = header->delayedBuffers.next; + + /* Only examine the oldest 1/3 of delayed buffers: + */ + if (header->numDelayed > 3) { + for (i = 0; i < header->numDelayed; i += 3) { + first = first->next; + } + } + + for (list = first, prev = list->prev; + list != &header->delayedBuffers; + list = prev, prev = list->prev) { + buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); + slab = buf->parent; + + if (!signaled) { + if (wait) { + ret = driFenceFinish(buf->fence, buf->fenceType, 0); + if (ret) + break; + signaled = 1; + wait = 0; + } else { + signaled = driFenceSignaled(buf->fence, buf->fenceType); + } + if (signaled) { + if (list == first) + firstWasSignaled = 1; + driFenceUnReference(&buf->fence); + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { + driFenceUnReference(&buf->fence); + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + } + } +} + + +static struct _DriSlabBuffer * +driSlabAllocBuffer(struct _DriSlabSizeHeader *header) +{ + static struct _DriSlabBuffer *buf; + struct _DriSlab *slab; + drmMMListHead *list; + int count = DRI_SLABPOOL_ALLOC_RETRIES; + + _glthread_LOCK_MUTEX(header->mutex); + while(header->slabs.next == &header->slabs && count > 0) { + driSlabCheckFreeLocked(header, 0); + if (header->slabs.next != &header->slabs) + break; + + _glthread_UNLOCK_MUTEX(header->mutex); + if (count != DRI_SLABPOOL_ALLOC_RETRIES) + usleep(1); + _glthread_LOCK_MUTEX(header->mutex); + (void) driAllocSlab(header); + count--; + } + + list = header->slabs.next; + if (list == &header->slabs) { + _glthread_UNLOCK_MUTEX(header->mutex); + return NULL; + } + slab = DRMLISTENTRY(struct _DriSlab, list, head); + if (--slab->numFree == 0) + DRMLISTDELINIT(list); + + list = slab->freeBuffers.next; + DRMLISTDELINIT(list); + + _glthread_UNLOCK_MUTEX(header->mutex); + buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); + return buf; +} + +static void * +pool_create(struct _DriBufferPool *driPool, unsigned long size, + uint64_t flags, unsigned hint, unsigned alignment) +{ + struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; + struct _DriSlabSizeHeader *header; + struct _DriSlabBuffer *buf; + void *dummy; + int i; + int ret; + + /* + * FIXME: Check for compatibility. + */ + header = pool->headers; + for (i=0; inumBuckets; ++i) { + if (header->bufSize >= size) + break; + header++; + } + + if (i < pool->numBuckets) + return driSlabAllocBuffer(header); + + + /* + * Fall back to allocate a buffer object directly from DRM. + * and wrap it in a driBO structure. + */ + + + buf = calloc(1, sizeof(*buf)); + + if (!buf) + return NULL; + + buf->bo = calloc(1, sizeof(*buf->bo)); + if (!buf->bo) + goto out_err0; + + if (alignment) { + if ((alignment < pool->pageSize) && (pool->pageSize % alignment)) + goto out_err1; + if ((alignment > pool->pageSize) && (alignment % pool->pageSize)) + goto out_err1; + } + + ret = drmBOCreate(pool->fd, size, alignment / pool->pageSize, NULL, + flags, hint, buf->bo); + if (ret) + goto out_err1; + + ret = drmBOMap(pool->fd, buf->bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, &dummy); + if (ret) + goto out_err2; + + ret = drmBOUnmap(pool->fd, buf->bo); + if (ret) + goto out_err2; + + return buf; + out_err2: + drmBOUnreference(pool->fd, buf->bo); + out_err1: + free(buf->bo); + out_err0: + free(buf); + return NULL; +} + +static int +pool_destroy(struct _DriBufferPool *driPool, void *private) +{ + struct _DriSlabBuffer *buf = + (struct _DriSlabBuffer *) private; + struct _DriSlab *slab; + struct _DriSlabSizeHeader *header; + + if (!buf->isSlabBuffer) { + struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; + int ret; + + ret = drmBOUnreference(pool->fd, buf->bo); + free(buf->bo); + free(buf); + return ret; + } + + slab = buf->parent; + header = slab->header; + + _glthread_LOCK_MUTEX(header->mutex); + buf->unFenced = 0; + buf->mapCount = 0; + + if (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)) { + DRMLISTADDTAIL(&buf->head, &header->delayedBuffers); + header->numDelayed++; + } else { + if (buf->fence) + driFenceUnReference(&buf->fence); + driSlabFreeBufferLocked(buf); + } + + _glthread_UNLOCK_MUTEX(header->mutex); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *driPool, void *private, + _glthread_Mutex *mutex, int lazy) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + while(buf->unFenced) + _glthread_COND_WAIT(buf->event, *mutex); + + if (!buf->fence) + return 0; + + driFenceFinish(buf->fence, buf->fenceType, lazy); + driFenceUnReference(&buf->fence); + + return 0; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + int busy; + + if (buf->isSlabBuffer) + busy = buf->unFenced || (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)); + else + busy = buf->fence && !driFenceSignaled(buf->fence, buf->fenceType); + + + if (busy) { + if (hint & DRM_BO_HINT_DONT_BLOCK) + return -EBUSY; + else { + (void) pool_waitIdle(pool, private, mutex, 0); + } + } + + ++buf->mapCount; + *virtual = (buf->isSlabBuffer) ? + (void *) ((uint8_t *) buf->parent->kbo->virtual + buf->start) : + (void *) buf->bo->virtual; + + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + --buf->mapCount; + if (buf->mapCount == 0 && buf->isSlabBuffer) + _glthread_COND_BROADCAST(buf->event); + + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + struct _DriSlab *slab; + struct _DriSlabSizeHeader *header; + + if (!buf->isSlabBuffer) { + assert(buf->bo->proposedFlags & DRM_BO_FLAG_NO_MOVE); + return buf->bo->offset; + } + + slab = buf->parent; + header = slab->header; + + (void) header; + assert(header->slabPool->proposedFlags & DRM_BO_FLAG_NO_MOVE); + return slab->kbo->bo.offset + buf->start; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + return buf->start; +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + if (!buf->isSlabBuffer) + return buf->bo->flags; + + return buf->parent->kbo->bo.flags; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + if (!buf->isSlabBuffer) + return buf->bo->size; + + return buf->parent->header->bufSize; +} + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + drmBO *bo; + + if (buf->fence) + driFenceUnReference(&buf->fence); + + buf->fence = driFenceReference(fence); + bo = (buf->isSlabBuffer) ? + &buf->parent->kbo->bo: + buf->bo; + buf->fenceType = bo->fenceFlags; + + buf->unFenced = 0; + _glthread_COND_BROADCAST(buf->event); + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + if (!buf) + return NULL; + + return (buf->isSlabBuffer) ? &buf->parent->kbo->bo : buf->bo; +} + +static int +pool_validate(struct _DriBufferPool *pool, void *private, + _glthread_Mutex *mutex) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + if (!buf->isSlabBuffer) + return 0; + + while(buf->mapCount != 0) + _glthread_COND_WAIT(buf->event, *mutex); + + buf->unFenced = 1; + return 0; +} + + +struct _DriFreeSlabManager * +driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec) +{ + struct _DriFreeSlabManager *tmp; + + tmp = calloc(1, sizeof(*tmp)); + if (!tmp) + return NULL; + + _glthread_INIT_MUTEX(tmp->mutex); + _glthread_LOCK_MUTEX(tmp->mutex); + tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000; + tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000; + tmp->slabTimeout.tv_usec -= tmp->slabTimeout.tv_sec*1000000; + + tmp->checkInterval.tv_usec = checkIntervalMsec*1000; + tmp->checkInterval.tv_sec = tmp->checkInterval.tv_usec / 1000000; + tmp->checkInterval.tv_usec -= tmp->checkInterval.tv_sec*1000000; + + gettimeofday(&tmp->nextCheck, NULL); + driTimeAdd(&tmp->nextCheck, &tmp->checkInterval); + DRMINITLISTHEAD(&tmp->timeoutList); + DRMINITLISTHEAD(&tmp->unCached); + DRMINITLISTHEAD(&tmp->cached); + _glthread_UNLOCK_MUTEX(tmp->mutex); + + return tmp; +} + +void +driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan) +{ + struct timeval time; + + time = fMan->nextCheck; + driTimeAdd(&time, &fMan->checkInterval); + + _glthread_LOCK_MUTEX(fMan->mutex); + driFreeTimeoutKBOsLocked(fMan, &time); + _glthread_UNLOCK_MUTEX(fMan->mutex); + + assert(fMan->timeoutList.next == &fMan->timeoutList); + assert(fMan->unCached.next == &fMan->unCached); + assert(fMan->cached.next == &fMan->cached); + + free(fMan); +} + +static void +driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size, + struct _DriSlabSizeHeader *header) +{ + _glthread_INIT_MUTEX(header->mutex); + _glthread_LOCK_MUTEX(header->mutex); + + DRMINITLISTHEAD(&header->slabs); + DRMINITLISTHEAD(&header->freeSlabs); + DRMINITLISTHEAD(&header->delayedBuffers); + + header->numDelayed = 0; + header->slabPool = pool; + header->bufSize = size; + + _glthread_UNLOCK_MUTEX(header->mutex); +} + +static void +driFinishSizeHeader(struct _DriSlabSizeHeader *header) +{ + drmMMListHead *list, *next; + struct _DriSlabBuffer *buf; + + _glthread_LOCK_MUTEX(header->mutex); + for (list = header->delayedBuffers.next, next = list->next; + list != &header->delayedBuffers; + list = next, next = list->next) { + + buf = DRMLISTENTRY(struct _DriSlabBuffer, list , head); + if (buf->fence) { + (void) driFenceFinish(buf->fence, buf->fenceType, 0); + driFenceUnReference(&buf->fence); + } + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + _glthread_UNLOCK_MUTEX(header->mutex); +} + +static void +pool_takedown(struct _DriBufferPool *driPool) +{ + struct _DriSlabPool *pool = driPool->data; + int i; + + for (i=0; inumBuckets; ++i) { + driFinishSizeHeader(&pool->headers[i]); + } + + free(pool->headers); + free(pool->bucketSizes); + free(pool); + free(driPool); +} + +struct _DriBufferPool * +driSlabPoolInit(int fd, uint64_t flags, + uint64_t validMask, + uint32_t smallestSize, + uint32_t numSizes, + uint32_t desiredNumBuffers, + uint32_t maxSlabSize, + uint32_t pageAlignment, + struct _DriFreeSlabManager *fMan) +{ + struct _DriBufferPool *driPool; + struct _DriSlabPool *pool; + uint32_t i; + + driPool = calloc(1, sizeof(*driPool)); + if (!driPool) + return NULL; + + pool = calloc(1, sizeof(*pool)); + if (!pool) + goto out_err0; + + pool->bucketSizes = calloc(numSizes, sizeof(*pool->bucketSizes)); + if (!pool->bucketSizes) + goto out_err1; + + pool->headers = calloc(numSizes, sizeof(*pool->headers)); + if (!pool->headers) + goto out_err2; + + pool->fMan = fMan; + pool->proposedFlags = flags; + pool->validMask = validMask; + pool->numBuckets = numSizes; + pool->pageSize = getpagesize(); + pool->fd = fd; + pool->pageAlignment = pageAlignment; + pool->maxSlabSize = maxSlabSize; + pool->desiredNumBuffers = desiredNumBuffers; + + for (i=0; inumBuckets; ++i) { + pool->bucketSizes[i] = (smallestSize << i); + driInitSizeHeader(pool, pool->bucketSizes[i], + &pool->headers[i]); + } + + driPool->data = (void *) pool; + driPool->map = &pool_map; + driPool->unmap = &pool_unmap; + driPool->destroy = &pool_destroy; + driPool->offset = &pool_offset; + driPool->poolOffset = &pool_poolOffset; + driPool->flags = &pool_flags; + driPool->size = &pool_size; + driPool->create = &pool_create; + driPool->fence = &pool_fence; + driPool->kernel = &pool_kernel; + driPool->validate = &pool_validate; + driPool->waitIdle = &pool_waitIdle; + driPool->takeDown = &pool_takedown; + + return driPool; + + out_err2: + free(pool->bucketSizes); + out_err1: + free(pool); + out_err0: + free(driPool); + + return NULL; +} -- cgit v1.2.3 From d791386b09682dbccc68cc3803b6d71a0362124e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 12:19:22 -0600 Subject: fix some additional program refcounting bugs cherry-picked from master --- src/mesa/shader/program.c | 5 ++--- src/mesa/shader/shader_api.c | 6 ++---- src/mesa/shader/slang/slang_emit.c | 4 ++-- 3 files changed, 6 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 9a23c5d7d3..e06ab5aa23 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -116,8 +116,6 @@ _mesa_free_program_data(GLcontext *ctx) } - - /** * Set the vertex/fragment program error state (position and error string). * This is generally called from within the parsers. @@ -272,6 +270,7 @@ _mesa_delete_program(GLcontext *ctx, struct gl_program *prog) { (void) ctx; ASSERT(prog); + ASSERT(prog->RefCount==0); if (prog == &_mesa_DummyProgram) return; @@ -391,7 +390,7 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog) clone->Format = prog->Format; clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions); if (!clone->Instructions) { - ctx->Driver.DeleteProgram(ctx, clone); + _mesa_reference_program(ctx, &clone, NULL); return NULL; } _mesa_copy_instructions(clone->Instructions, prog->Instructions, diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index f12fa28d97..2b3dbff7a0 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -243,10 +243,8 @@ _mesa_free_shader(GLcontext *ctx, struct gl_shader *sh) _mesa_free((void *) sh->Source); if (sh->InfoLog) _mesa_free(sh->InfoLog); - for (i = 0; i < sh->NumPrograms; i++) { - assert(sh->Programs[i]); - ctx->Driver.DeleteProgram(ctx, sh->Programs[i]); - } + for (i = 0; i < sh->NumPrograms; i++) + _mesa_reference_program(ctx, &sh->Programs[i], NULL); if (sh->Programs) _mesa_free(sh->Programs); _mesa_free(sh); diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 382b13ff05..964274eba8 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -79,7 +79,7 @@ new_subroutine(slang_emit_info *emitInfo, GLuint *id) _mesa_realloc(emitInfo->Subroutines, n * sizeof(struct gl_program), (n + 1) * sizeof(struct gl_program)); - emitInfo->Subroutines[n] = _mesa_new_program(ctx, emitInfo->prog->Target, 0); + emitInfo->Subroutines[n] = ctx->Driver.NewProgram(ctx, emitInfo->prog->Target, 0); emitInfo->Subroutines[n]->Parameters = emitInfo->prog->Parameters; emitInfo->NumSubroutines++; *id = n; @@ -1787,7 +1787,7 @@ _slang_resolve_subroutines(slang_emit_info *emitInfo) sub->NumInstructions); /* delete subroutine code */ sub->Parameters = NULL; /* prevent double-free */ - _mesa_delete_program(ctx, sub); + _mesa_reference_program(ctx, &emitInfo->Subroutines[i], NULL); } /* free subroutine list */ -- cgit v1.2.3 From b2562736894b6a97e4f1be6c6861e250e9d0a765 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 12:19:22 -0600 Subject: fix some additional program refcounting bugs cherry-picked from master --- src/mesa/shader/program.c | 5 ++--- src/mesa/shader/shader_api.c | 6 ++---- src/mesa/shader/slang/slang_emit.c | 4 ++-- 3 files changed, 6 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 9a23c5d7d3..e06ab5aa23 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -116,8 +116,6 @@ _mesa_free_program_data(GLcontext *ctx) } - - /** * Set the vertex/fragment program error state (position and error string). * This is generally called from within the parsers. @@ -272,6 +270,7 @@ _mesa_delete_program(GLcontext *ctx, struct gl_program *prog) { (void) ctx; ASSERT(prog); + ASSERT(prog->RefCount==0); if (prog == &_mesa_DummyProgram) return; @@ -391,7 +390,7 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog) clone->Format = prog->Format; clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions); if (!clone->Instructions) { - ctx->Driver.DeleteProgram(ctx, clone); + _mesa_reference_program(ctx, &clone, NULL); return NULL; } _mesa_copy_instructions(clone->Instructions, prog->Instructions, diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 5ec03563ba..fb3c59b22d 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -245,10 +245,8 @@ _mesa_free_shader(GLcontext *ctx, struct gl_shader *sh) _mesa_free((void *) sh->Source); if (sh->InfoLog) _mesa_free(sh->InfoLog); - for (i = 0; i < sh->NumPrograms; i++) { - assert(sh->Programs[i]); - ctx->Driver.DeleteProgram(ctx, sh->Programs[i]); - } + for (i = 0; i < sh->NumPrograms; i++) + _mesa_reference_program(ctx, &sh->Programs[i], NULL); if (sh->Programs) _mesa_free(sh->Programs); _mesa_free(sh); diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 382b13ff05..964274eba8 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -79,7 +79,7 @@ new_subroutine(slang_emit_info *emitInfo, GLuint *id) _mesa_realloc(emitInfo->Subroutines, n * sizeof(struct gl_program), (n + 1) * sizeof(struct gl_program)); - emitInfo->Subroutines[n] = _mesa_new_program(ctx, emitInfo->prog->Target, 0); + emitInfo->Subroutines[n] = ctx->Driver.NewProgram(ctx, emitInfo->prog->Target, 0); emitInfo->Subroutines[n]->Parameters = emitInfo->prog->Parameters; emitInfo->NumSubroutines++; *id = n; @@ -1787,7 +1787,7 @@ _slang_resolve_subroutines(slang_emit_info *emitInfo) sub->NumInstructions); /* delete subroutine code */ sub->Parameters = NULL; /* prevent double-free */ - _mesa_delete_program(ctx, sub); + _mesa_reference_program(ctx, &emitInfo->Subroutines[i], NULL); } /* free subroutine list */ -- cgit v1.2.3 From f84005c0a114837a0c7209a3c598e02fb5f47da1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:01:31 -0600 Subject: mesa: glUniform(location==-1) is not an error --- src/mesa/shader/shader_api.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 2b3dbff7a0..d0512b8ed1 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -1151,6 +1151,10 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, return; } + if (location == -1) + return; /* The standard specifies this as a no-op */ + + if (location < 0 || location >= (GLint) shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniform(location)"); return; @@ -1246,11 +1250,16 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, GLboolean transpose, const GLfloat *values) { struct gl_shader_program *shProg = ctx->Shader.CurrentProgram; + if (!shProg || !shProg->LinkStatus) { _mesa_error(ctx, GL_INVALID_OPERATION, "glUniformMatrix(program not linked)"); return; } + + if (location == -1) + return; /* The standard specifies this as a no-op */ + if (location < 0 || location >= shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniformMatrix(location)"); return; -- cgit v1.2.3 From 6c536b9924a46e3e6f4eb93369b2e7006182ba39 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:02:20 -0600 Subject: mesa: fix errors in LightModelProduct state, other misc error cases --- src/mesa/shader/slang/slang_builtin.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_builtin.c b/src/mesa/shader/slang/slang_builtin.c index 1081d8ff8d..f0f59aa87a 100644 --- a/src/mesa/shader/slang/slang_builtin.c +++ b/src/mesa/shader/slang/slang_builtin.c @@ -250,7 +250,7 @@ lookup_statevar(const char *var, GLint index1, GLint index2, const char *field, } } else if (strcmp(var, "gl_FrontLightModelProduct") == 0) { - if (strcmp(field, "ambient") == 0) { + if (strcmp(field, "sceneColor") == 0) { tokens[0] = STATE_LIGHTMODEL_SCENECOLOR; tokens[1] = 0; } @@ -259,7 +259,7 @@ lookup_statevar(const char *var, GLint index1, GLint index2, const char *field, } } else if (strcmp(var, "gl_BackLightModelProduct") == 0) { - if (strcmp(field, "ambient") == 0) { + if (strcmp(field, "sceneColor") == 0) { tokens[0] = STATE_LIGHTMODEL_SCENECOLOR; tokens[1] = 1; } @@ -397,6 +397,8 @@ lookup_statevar(const char *var, GLint index1, GLint index2, const char *field, * var.field * var[i].field * var[i][j] + * + * \return -1 upon error, else position in paramList of the state var/data */ GLint _slang_alloc_statevar(slang_ir_node *n, @@ -414,9 +416,13 @@ _slang_alloc_statevar(slang_ir_node *n, if (n->Opcode == IR_ELEMENT) { /* XXX can only handle constant indexes for now */ - assert(n->Children[1]->Opcode == IR_FLOAT); - index1 = (GLint) n->Children[1]->Value[0]; - n = n->Children[0]; + if (n->Children[1]->Opcode == IR_FLOAT) { + index1 = (GLint) n->Children[1]->Value[0]; + n = n->Children[0]; + } + else { + return -1; + } } if (n->Opcode == IR_ELEMENT) { -- cgit v1.2.3 From c4ceb4e4f3bea1d6464c847669d5f9e8353fd5d3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:03:38 -0600 Subject: mesa: minor bug fixes, null ptr checks, dead code removal --- src/mesa/shader/slang/slang_codegen.c | 19 ++++++++----------- 1 file changed, 8 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 0de2b0a57e..15a0e3abcd 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -495,6 +495,9 @@ new_node0(slang_ir_opcode op) } +/** + * Create sequence of two nodes. + */ static slang_ir_node * new_seq(slang_ir_node *left, slang_ir_node *right) { @@ -1844,6 +1847,7 @@ _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) n->Store->File = PROGRAM_TEMPORARY; n->Store->Size = _slang_sizeof_type_specifier(&n->Var->type.specifier); + A->program->NumTemporaries++; assert(n->Store->Size > 0); } return n; @@ -2060,7 +2064,8 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) } /* XXX make copy of this initializer? */ rhs = _slang_gen_operation(A, &oper->children[0]); - assert(rhs); + if (!rhs) + return NULL; /* must have found an error */ init = new_node2(IR_MOVE, var, rhs); /*assert(rhs->Opcode != IR_SEQ);*/ n = new_seq(varDecl, init); @@ -2347,7 +2352,8 @@ _slang_gen_field(slang_assemble_ctx * A, slang_operation *oper) return n; } else if ( ti.spec.type == SLANG_SPEC_FLOAT - || ti.spec.type == SLANG_SPEC_INT) { + || ti.spec.type == SLANG_SPEC_INT + || ti.spec.type == SLANG_SPEC_BOOL) { const GLuint rows = 1; slang_swizzle swz; slang_ir_node *n; @@ -2879,17 +2885,8 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, * store->Index = sampler number (0..7, typically) * store->Size = texture type index (1D, 2D, 3D, cube, etc) */ -#if 0 - GLint samplerUniform - = _mesa_add_sampler(prog->Parameters, varName, datatype); -#elif 0 - GLint samplerUniform - = _mesa_add_sampler(prog->Samplers, varName, datatype); - (void) _mesa_add_sampler(prog->Parameters, varName, datatype); /* dummy entry */ -#else const GLint sampNum = A->numSamplers++; _mesa_add_sampler(prog->Parameters, varName, datatype, sampNum); -#endif store = _slang_new_ir_storage(PROGRAM_SAMPLER, sampNum, texIndex); if (dbg) printf("SAMPLER "); } -- cgit v1.2.3 From fcb23d7db7339a2b1ca51d929fc1b5065f7e0f77 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:04:00 -0600 Subject: mesa: dead code removal --- src/mesa/shader/slang/slang_compile.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 0df673085b..1449888f9f 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -2157,9 +2157,6 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) shader->Programs[0]->Parameters = _mesa_new_parameter_list(); shader->Programs[0]->Varying = _mesa_new_parameter_list(); shader->Programs[0]->Attributes = _mesa_new_parameter_list(); -#if 0 - shader->Programs[0]->Samplers = _mesa_new_parameter_list(); -#endif } slang_info_log_construct(&info_log); -- cgit v1.2.3 From 05cacc06ccbbf088ab3fed0bf648f4f5fc61c154 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:04:29 -0600 Subject: mesa: minor bug fixes from master --- src/mesa/shader/slang/slang_emit.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 964274eba8..ff63e05dd2 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1,6 +1,6 @@ /* * Mesa 3-D graphics library - * Version: 6.5.3 + * Version: 7.0.3 * * Copyright (C) 2005-2007 Brian Paul All Rights Reserved. * @@ -869,12 +869,18 @@ emit_return(slang_emit_info *emitInfo, slang_ir_node *n) static struct prog_instruction * emit_kill(slang_emit_info *emitInfo) { + struct gl_fragment_program *fp; struct prog_instruction *inst; /* NV-KILL - discard fragment depending on condition code. * Note that ARB-KILL depends on sign of vector operand. */ inst = new_instruction(emitInfo, OPCODE_KIL_NV); inst->DstReg.CondMask = COND_TR; /* always branch */ + + assert(emitInfo->prog->Target == GL_FRAGMENT_PROGRAM_ARB); + fp = (struct gl_fragment_program *) emitInfo->prog; + fp->UsesKill = GL_TRUE; + return inst; } @@ -1500,6 +1506,10 @@ emit_struct_field(slang_emit_info *emitInfo, slang_ir_node *n) { if (n->Store->File == PROGRAM_STATE_VAR) { n->Store->Index = _slang_alloc_statevar(n, emitInfo->prog->Parameters); + if (n->Store->Index < 0) { + slang_info_log_error(emitInfo->log, "Error parsing state variable"); + return NULL; + } } else { GLint offset = n->FieldOffset / 4; -- cgit v1.2.3 From 4f915b0f83861640e5e3f5ea62a70c0d6201a5a2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:04:51 -0600 Subject: mesa: updated comment --- src/mesa/shader/slang/slang_link.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index ae581553dc..8213b7772c 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -369,12 +369,9 @@ link_error(struct gl_shader_program *shProg, const char *msg) * 2. Varying vars in the two shaders are combined so their locations * agree between the vertex and fragment stages. They're treated as * vertex program output attribs and as fragment program input attribs. - * 3. Uniform vars (including state references, constants, etc) from the - * vertex and fragment shaders are merged into one group. Recall that - * GLSL uniforms are shared by all linked shaders. - * 4. The vertex and fragment programs are cloned and modified to update - * src/dst register references so they use the new, linked uniform/ - * varying storage locations. + * 3. The vertex and fragment programs are cloned and modified to update + * src/dst register references so they use the new, linked varying + * storage locations. */ void _slang_link(GLcontext *ctx, -- cgit v1.2.3 From afa4c792b00070a5c50730ba455df18629d4e550 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:05:25 -0600 Subject: mesa: sync up swrast/s_fragprog.c with master --- src/mesa/swrast/s_fragprog.c | 38 ++++++++++++++++++++++---------------- 1 file changed, 22 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/mesa/swrast/s_fragprog.c b/src/mesa/swrast/s_fragprog.c index 6ee8bfd0a5..40ed48696c 100644 --- a/src/mesa/swrast/s_fragprog.c +++ b/src/mesa/swrast/s_fragprog.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.2 + * Version: 7.0.3 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -33,18 +33,21 @@ /** - * Fetch a texel. + * Fetch a texel with given lod. + * Called via machine->FetchTexelLod() */ static void -fetch_texel( GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda, - GLuint unit, GLfloat color[4] ) +fetch_texel_lod( GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda, + GLuint unit, GLfloat color[4] ) { GLchan rgba[4]; SWcontext *swrast = SWRAST_CONTEXT(ctx); + const struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current; + + lambda = CLAMP(lambda, texObj->MinLod, texObj->MaxLod); /* XXX use a float-valued TextureSample routine here!!! */ - swrast->TextureSample[unit](ctx, ctx->Texture.Unit[unit]._Current, - 1, (const GLfloat (*)[4]) texcoord, + swrast->TextureSample[unit](ctx, texObj, 1, (const GLfloat (*)[4]) texcoord, &lambda, &rgba); color[0] = CHAN_TO_FLOAT(rgba[0]); color[1] = CHAN_TO_FLOAT(rgba[1]); @@ -56,6 +59,7 @@ fetch_texel( GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda, /** * Fetch a texel with the given partial derivatives to compute a level * of detail in the mipmap. + * Called via machine->FetchTexelDeriv() */ static void fetch_texel_deriv( GLcontext *ctx, const GLfloat texcoord[4], @@ -69,15 +73,17 @@ fetch_texel_deriv( GLcontext *ctx, const GLfloat texcoord[4], const GLfloat texH = (GLfloat) texImg->HeightScale; GLchan rgba[4]; - GLfloat lambda = _swrast_compute_lambda(texdx[0], texdy[0], /* ds/dx, ds/dy */ - texdx[1], texdy[1], /* dt/dx, dt/dy */ - texdx[3], texdy[2], /* dq/dx, dq/dy */ - texW, texH, - texcoord[0], texcoord[1], texcoord[3], - 1.0F / texcoord[3]); + GLfloat lambda + = _swrast_compute_lambda(texdx[0], texdy[0], /* ds/dx, ds/dy */ + texdx[1], texdy[1], /* dt/dx, dt/dy */ + texdx[3], texdy[2], /* dq/dx, dq/dy */ + texW, texH, + texcoord[0], texcoord[1], texcoord[3], + 1.0F / texcoord[3]) + lodBias; + + lambda = CLAMP(lambda, texObj->MinLod, texObj->MaxLod); - swrast->TextureSample[unit](ctx, ctx->Texture.Unit[unit]._Current, - 1, (const GLfloat (*)[4]) texcoord, + swrast->TextureSample[unit](ctx, texObj, 1, (const GLfloat (*)[4]) texcoord, &lambda, &rgba); color[0] = CHAN_TO_FLOAT(rgba[0]); color[1] = CHAN_TO_FLOAT(rgba[1]); @@ -132,7 +138,7 @@ init_machine(GLcontext *ctx, struct gl_program_machine *machine, /* init call stack */ machine->StackDepth = 0; - machine->FetchTexelLod = fetch_texel; + machine->FetchTexelLod = fetch_texel_lod; machine->FetchTexelDeriv = fetch_texel_deriv; } -- cgit v1.2.3 From 26988c170c61c21d0e76397689a270f485bc4086 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:05:48 -0600 Subject: mesa: use CALLOC_STRUCT() --- src/mesa/shader/prog_parameter.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index 4eb7313bec..0b61bac696 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -40,8 +40,7 @@ struct gl_program_parameter_list * _mesa_new_parameter_list(void) { - return (struct gl_program_parameter_list *) - _mesa_calloc(sizeof(struct gl_program_parameter_list)); + return CALLOC_STRUCT(gl_program_parameter_list); } -- cgit v1.2.3 From a9997657b9d66b54d184a899e49e35885928ef80 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:06:10 -0600 Subject: mesa: fix missing w assignment in asin(vec4) --- src/mesa/shader/slang/library/slang_common_builtin.gc | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index 45cf1c6fd0..26080cf26c 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -343,6 +343,7 @@ vec4 asin(const vec4 v) __retVal.x = asin(v.x); __retVal.y = asin(v.y); __retVal.z = asin(v.z); + __retVal.w = asin(v.w); } float acos(const float x) -- cgit v1.2.3 From 8cf38c7e6a42c66166549a2526f28d94b95b0c0a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:07:18 -0600 Subject: mesa: regenerate slang built-in data (asin(vec4) fix) --- .../shader/slang/library/slang_common_builtin_gc.h | 83 +++++++++++----------- 1 file changed, 42 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index a896c5a287..c8d1ffc174 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -181,47 +181,48 @@ 0,1,0,12,0,97,115,105,110,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97, 115,105,110,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105, 110,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,115,105,110,0, -18,118,0,59,122,0,0,0,20,0,0,1,0,9,0,97,99,111,115,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,104,97,108,102, -80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,48,0,53,0,0,48,0,0,9,18,95,95,114,101,116,86,97,108, -0,18,104,97,108,102,80,105,0,58,97,115,105,110,0,18,120,0,0,0,47,20,0,0,1,0,10,0,97,99,111,115,0,1, -1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120, -0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20, -0,0,1,0,11,0,97,99,111,115,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58, -97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99, -111,115,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115, -0,18,118,0,59,122,0,0,0,20,0,0,1,0,12,0,97,99,111,115,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,122,0,58,97,99,111,115,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0, -58,97,99,111,115,0,18,118,0,59,119,0,0,0,20,0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,120,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,58,97,115,105,110,0,18,120,0,58,105,110,118,101,114,115,101,115,113, -114,116,0,18,120,0,18,120,0,48,17,49,0,48,0,0,46,0,0,48,0,0,20,0,0,1,0,10,0,97,116,97,110,0,1,1,0, -10,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97, -110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97, -110,0,1,1,0,11,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0, -58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,122,0, -0,0,20,0,0,1,0,12,0,97,116,97,110,0,1,1,0,12,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120, -0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111, -118,101,114,95,120,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97, -110,0,18,121,95,111,118,101,114,95,120,0,59,119,0,0,0,20,0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,121,0, -0,1,1,0,9,120,0,0,0,1,10,18,120,0,17,48,0,48,0,0,38,0,8,17,48,0,48,0,0,0,9,14,0,3,2,0,9,1,122,0,2, -58,97,116,97,110,0,18,121,0,18,120,0,49,0,0,0,0,10,18,120,0,17,48,0,48,0,0,40,0,2,10,18,121,0,17, -48,0,48,0,0,40,0,8,18,122,0,17,51,0,49,52,49,53,57,51,0,0,47,0,9,14,0,8,18,122,0,17,51,0,49,52,49, -53,57,51,0,0,46,0,0,9,14,0,8,18,122,0,0,0,1,0,10,0,97,116,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0, -59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0, -0,18,118,0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0, -59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0, -0,18,118,0,59,122,0,0,0,20,0,0,1,0,12,0,97,116,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0, +18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,115,105,110,0,18,118, +0,59,119,0,0,0,20,0,0,1,0,9,0,97,99,111,115,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,104,97,108,102,80,105, +0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,48,0,53,0,0,48,0,0,9,18,95,95,114,101,116,86,97,108,0,18, +104,97,108,102,80,105,0,58,97,115,105,110,0,18,120,0,0,0,47,20,0,0,1,0,10,0,97,99,111,115,0,1,1,0, +10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0, +0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,0, +1,0,11,0,97,99,111,115,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99, +111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115, +0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115,0,18,118, +0,59,122,0,0,0,20,0,0,1,0,12,0,97,99,111,115,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, +121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58, +97,99,111,115,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,99, +111,115,0,18,118,0,59,119,0,0,0,20,0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,58,97,115,105,110,0,18,120,0,58,105,110,118,101,114,115,101,115,113,114,116,0, +18,120,0,18,120,0,48,17,49,0,48,0,0,46,0,0,48,0,0,20,0,0,1,0,10,0,97,116,97,110,0,1,1,0,10,121,95, +111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18, +121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, +97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97,110,0,1, +1,0,11,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116, +97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,122,0,0,0,20, +0,0,1,0,12,0,97,116,97,110,0,1,1,0,12,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59, +121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111,118, +101,114,95,120,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0, +18,121,95,111,118,101,114,95,120,0,59,119,0,0,0,20,0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,121,0,0,1,1, +0,9,120,0,0,0,1,10,18,120,0,17,48,0,48,0,0,38,0,8,17,48,0,48,0,0,0,9,14,0,3,2,0,9,1,122,0,2,58,97, +116,97,110,0,18,121,0,18,120,0,49,0,0,0,0,10,18,120,0,17,48,0,48,0,0,40,0,2,10,18,121,0,17,48,0,48, +0,0,40,0,8,18,122,0,17,51,0,49,52,49,53,57,51,0,0,47,0,9,14,0,8,18,122,0,17,51,0,49,52,49,53,57,51, +0,0,46,0,0,9,14,0,8,18,122,0,0,0,1,0,10,0,97,116,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0, +0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118, +0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121, +0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18, +118,0,59,122,0,0,0,20,0,0,1,0,12,0,97,116,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0, 59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0, 0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,18,117, 0,59,119,0,0,18,118,0,59,119,0,0,0,20,0,0,1,0,9,0,112,111,119,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1, -- cgit v1.2.3 From 753f979a32cdbbe76dfacf95cbc3b43545a7d1ad Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:20:28 -0600 Subject: rewrite some of the mat*mat, mat*vec intrinsics Also, remove obsolete matrix codegen code. cherry-picked from master --- src/mesa/shader/slang/library/slang_core.gc | 162 +----- src/mesa/shader/slang/library/slang_core_gc.h | 697 ++++++++++++-------------- src/mesa/shader/slang/slang_codegen.c | 44 -- 3 files changed, 336 insertions(+), 567 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 2e7ebc347a..ffa57f74c2 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -1545,15 +1545,8 @@ mat2 __operator - (const mat2 m, const mat2 n) mat2 __operator * (const mat2 m, const mat2 n) { - vec2 mRow0, mRow1; - mRow0.x = m[0].x; - mRow0.y = m[1].x; - mRow1.x = m[0].y; - mRow1.y = m[1].y; - __retVal[0].x = dot(mRow0, n[0]); - __retVal[1].x = dot(mRow0, n[1]); - __retVal[0].y = dot(mRow1, n[0]); - __retVal[1].y = dot(mRow1, n[1]); + __retVal[0] = m[0] * n[0].xx + m[1] * n[0].yy; + __retVal[1] = m[0] * n[1].xx + m[1] * n[1].yy; } mat2 __operator / (const mat2 m, const mat2 n) @@ -1581,34 +1574,9 @@ mat3 __operator - (const mat3 m, const mat3 n) mat3 __operator * (const mat3 m, const mat3 n) { - // sub-blocks to reduce register usage - { - vec3 mRow0; - mRow0.x = m[0].x; - mRow0.y = m[1].x; - mRow0.z = m[2].x; - __retVal[0].x = dot(mRow0, n[0]); - __retVal[1].x = dot(mRow0, n[1]); - __retVal[2].x = dot(mRow0, n[2]); - } - { - vec3 mRow1; - mRow1.x = m[0].y; - mRow1.y = m[1].y; - mRow1.z = m[2].y; - __retVal[0].y = dot(mRow1, n[0]); - __retVal[1].y = dot(mRow1, n[1]); - __retVal[2].y = dot(mRow1, n[2]); - } - { - vec3 mRow2; - mRow2.x = m[0].z; - mRow2.y = m[1].z; - mRow2.z = m[2].z; - __retVal[0].z = dot(mRow2, n[0]); - __retVal[1].z = dot(mRow2, n[1]); - __retVal[2].z = dot(mRow2, n[2]); - } + __retVal[0] = m[0] * n[0].xxx + m[1] * n[0].yyy + m[2] * n[0].zzz; + __retVal[1] = m[0] * n[1].xxx + m[1] * n[1].yyy + m[2] * n[1].zzz; + __retVal[2] = m[0] * n[2].xxx + m[1] * n[2].yyy + m[2] * n[2].zzz; } mat3 __operator / (const mat3 m, const mat3 n) @@ -1639,51 +1607,10 @@ mat4 __operator - (const mat4 m, const mat4 n) mat4 __operator * (const mat4 m, const mat4 n) { - // sub-blocks to reduce temporary usage - { - vec4 mRow0; - mRow0.x = m[0].x; - mRow0.y = m[1].x; - mRow0.z = m[2].x; - mRow0.w = m[3].x; - __retVal[0].x = dot(mRow0, n[0]); - __retVal[1].x = dot(mRow0, n[1]); - __retVal[2].x = dot(mRow0, n[2]); - __retVal[3].x = dot(mRow0, n[3]); - } - { - vec4 mRow1; - mRow1.x = m[0].y; - mRow1.y = m[1].y; - mRow1.z = m[2].y; - mRow1.w = m[3].y; - __retVal[0].y = dot(mRow1, n[0]); - __retVal[1].y = dot(mRow1, n[1]); - __retVal[2].y = dot(mRow1, n[2]); - __retVal[3].y = dot(mRow1, n[3]); - } - { - vec4 mRow2; - mRow2.x = m[0].z; - mRow2.y = m[1].z; - mRow2.z = m[2].z; - mRow2.w = m[3].z; - __retVal[0].z = dot(mRow2, n[0]); - __retVal[1].z = dot(mRow2, n[1]); - __retVal[2].z = dot(mRow2, n[2]); - __retVal[3].z = dot(mRow2, n[3]); - } - { - vec4 mRow3; - mRow3.x = m[0].w; - mRow3.y = m[1].w; - mRow3.z = m[2].w; - mRow3.w = m[3].w; - __retVal[0].w = dot(mRow3, n[0]); - __retVal[1].w = dot(mRow3, n[1]); - __retVal[2].w = dot(mRow3, n[2]); - __retVal[3].w = dot(mRow3, n[3]); - } + __retVal[0] = m[0] * n[0].xxxx + m[1] * n[0].yyyy + m[2] * n[0].zzzz + m[3] * n[0].wwww; + __retVal[1] = m[0] * n[1].xxxx + m[1] * n[1].yyyy + m[2] * n[1].zzzz + m[3] * n[1].wwww; + __retVal[2] = m[0] * n[2].xxxx + m[1] * n[2].yyyy + m[2] * n[2].zzzz + m[3] * n[2].wwww; + __retVal[3] = m[0] * n[3].xxxx + m[1] * n[3].yyyy + m[2] * n[3].zzzz + m[3] * n[3].wwww; } mat4 __operator / (const mat4 m, const mat4 n) @@ -1877,13 +1804,8 @@ mat4 __operator / (const mat4 m, const float b) vec2 __operator * (const mat2 m, const vec2 v) { - vec2 r0, r1; - r0.x = m[0].x; - r0.y = m[1].x; - r1.x = m[0].y; - r1.y = m[1].y; - __retVal.x = dot(r0, v); - __retVal.y = dot(r1, v); + __retVal = m[0] * v.xx + + m[1] * v.yy; } vec2 __operator * (const vec2 v, const mat2 m) @@ -1894,27 +1816,9 @@ vec2 __operator * (const vec2 v, const mat2 m) vec3 __operator * (const mat3 m, const vec3 v) { - { - vec3 r0; - r0.x = m[0].x; - r0.y = m[1].x; - r0.z = m[2].x; - __asm vec3_dot __retVal.x, r0, v; - } - { - vec3 r1; - r1.x = m[0].y; - r1.y = m[1].y; - r1.z = m[2].y; - __asm vec3_dot __retVal.y, r1, v; - } - { - vec3 r2; - r2.x = m[0].z; - r2.y = m[1].z; - r2.z = m[2].z; - __asm vec3_dot __retVal.z, r2, v; - } + __retVal = m[0] * v.xxx + + m[1] * v.yyy + + m[2] * v.zzz; } vec3 __operator * (const vec3 v, const mat3 m) @@ -1926,44 +1830,14 @@ vec3 __operator * (const vec3 v, const mat3 m) vec4 __operator * (const mat4 m, const vec4 v) { - // extract rows, then do dot product - { - vec4 r0; - r0.x = m[0].x; - r0.y = m[1].x; - r0.z = m[2].x; - r0.w = m[3].x; - __asm vec4_dot __retVal.x, r0, v; - } - { - vec4 r1; - r1.x = m[0].y; - r1.y = m[1].y; - r1.z = m[2].y; - r1.w = m[3].y; - __asm vec4_dot __retVal.y, r1, v; - } - { - vec4 r2; - r2.x = m[0].z; - r2.y = m[1].z; - r2.z = m[2].z; - r2.w = m[3].z; - __asm vec4_dot __retVal.z, r2, v; - } - { - vec4 r3; - r3.x = m[0].w; - r3.y = m[1].w; - r3.z = m[2].w; - r3.w = m[3].w; - __asm vec4_dot __retVal.w, r3, v; - } + __retVal = m[0] * v.xxxx + + m[1] * v.yyyy + + m[2] * v.zzzz + + m[3] * v.wwww; } vec4 __operator * (const vec4 v, const mat4 m) { - //mm __retVal.x = dot(v, m[0]); __retVal.y = dot(v, m[1]); __retVal.z = dot(v, m[2]); diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index f2b4fd6464..1604d5a034 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -453,151 +453,113 @@ 97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0, 13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, 57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,3,2,0, -10,1,109,82,111,119,48,0,0,1,1,109,82,111,119,49,0,0,0,9,18,109,82,111,119,48,0,59,120,0,18,109,0, -16,8,48,0,57,59,120,0,20,0,9,18,109,82,111,119,48,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,20,0, -9,18,109,82,111,119,49,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,20,0,9,18,109,82,111,119,49,0,59, -121,0,18,109,0,16,10,49,0,57,59,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120, -0,58,100,111,116,0,18,109,82,111,119,48,0,0,18,110,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,59,120,0,58,100,111,116,0,18,109,82,111,119,48,0,0,18,110,0,16,10,49,0, -57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,58,100,111,116,0,18,109,82, -111,119,49,0,0,18,110,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59, -121,0,58,100,111,116,0,18,109,82,111,119,49,0,0,18,110,0,16,10,49,0,57,0,0,20,0,0,1,0,13,2,22,1,1, -0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0, +48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0, +57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10, +49,0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27, +1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, +16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, +109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59, +120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16, +10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122, +122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110, +0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0, +48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1, +0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, 0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46, -20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,21,1,1,0,14,109,0, -0,1,1,0,14,110,0,0,0,1,2,3,2,0,11,1,109,82,111,119,48,0,0,0,9,18,109,82,111,119,48,0,59,120,0,18, -109,0,16,8,48,0,57,59,120,0,20,0,9,18,109,82,111,119,48,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0, -20,0,9,18,109,82,111,119,48,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,8,48,0,57,59,120,0,58,100,111,116,0,18,109,82,111,119,48,0,0,18,110,0,16,8,48,0,57, -0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,58,100,111,116,0,18,109,82,111, -119,48,0,0,18,110,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120, -0,58,100,111,116,0,18,109,82,111,119,48,0,0,18,110,0,16,10,50,0,57,0,0,20,0,0,2,3,2,0,11,1,109,82, -111,119,49,0,0,0,9,18,109,82,111,119,49,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,20,0,9,18,109,82, -111,119,49,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,20,0,9,18,109,82,111,119,49,0,59,122,0,18, -109,0,16,10,50,0,57,59,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,58,100, -111,116,0,18,109,82,111,119,49,0,0,18,110,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,49,0,57,59,121,0,58,100,111,116,0,18,109,82,111,119,49,0,0,18,110,0,16,10,49,0,57,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,58,100,111,116,0,18,109,82,111,119,49,0, -0,18,110,0,16,10,50,0,57,0,0,20,0,0,2,3,2,0,11,1,109,82,111,119,50,0,0,0,9,18,109,82,111,119,50,0, -59,120,0,18,109,0,16,8,48,0,57,59,122,0,20,0,9,18,109,82,111,119,50,0,59,121,0,18,109,0,16,10,49,0, -57,59,122,0,20,0,9,18,109,82,111,119,50,0,59,122,0,18,109,0,16,10,50,0,57,59,122,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,58,100,111,116,0,18,109,82,111,119,50,0,0,18,110,0, -16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,58,100,111,116,0, -18,109,82,111,119,50,0,0,18,110,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -50,0,57,59,122,0,58,100,111,116,0,18,109,82,111,119,50,0,0,18,110,0,16,10,50,0,57,0,0,20,0,0,0,1,0, -14,2,22,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, -0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15, -110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48, -0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16, -10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110, -0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18, -110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0, -0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,2,3,2,0,12,1,109,82,111,119,48,0,0,0,9,18,109, -82,111,119,48,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,20,0,9,18,109,82,111,119,48,0,59,121,0,18, -109,0,16,10,49,0,57,59,120,0,20,0,9,18,109,82,111,119,48,0,59,122,0,18,109,0,16,10,50,0,57,59,120, -0,20,0,9,18,109,82,111,119,48,0,59,119,0,18,109,0,16,10,51,0,57,59,120,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,59,120,0,58,100,111,116,0,18,109,82,111,119,48,0,0,18,110,0,16,8,48,0, -57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,58,100,111,116,0,18,109,82, -111,119,48,0,0,18,110,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59, -120,0,58,100,111,116,0,18,109,82,111,119,48,0,0,18,110,0,16,10,50,0,57,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,59,120,0,58,100,111,116,0,18,109,82,111,119,48,0,0,18,110,0,16,10,51, -0,57,0,0,20,0,0,2,3,2,0,12,1,109,82,111,119,49,0,0,0,9,18,109,82,111,119,49,0,59,120,0,18,109,0,16, -8,48,0,57,59,121,0,20,0,9,18,109,82,111,119,49,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,20,0,9, -18,109,82,111,119,49,0,59,122,0,18,109,0,16,10,50,0,57,59,121,0,20,0,9,18,109,82,111,119,49,0,59, -119,0,18,109,0,16,10,51,0,57,59,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121, -0,58,100,111,116,0,18,109,82,111,119,49,0,0,18,110,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,59,121,0,58,100,111,116,0,18,109,82,111,119,49,0,0,18,110,0,16,10,49,0, -57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,58,100,111,116,0,18,109,82, -111,119,49,0,0,18,110,0,16,10,50,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59, -121,0,58,100,111,116,0,18,109,82,111,119,49,0,0,18,110,0,16,10,51,0,57,0,0,20,0,0,2,3,2,0,12,1,109, -82,111,119,50,0,0,0,9,18,109,82,111,119,50,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,20,0,9,18,109, -82,111,119,50,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,20,0,9,18,109,82,111,119,50,0,59,122,0,18, -109,0,16,10,50,0,57,59,122,0,20,0,9,18,109,82,111,119,50,0,59,119,0,18,109,0,16,10,51,0,57,59,122, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,58,100,111,116,0,18,109,82,111,119, -50,0,0,18,110,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,58, -100,111,116,0,18,109,82,111,119,50,0,0,18,110,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,50,0,57,59,122,0,58,100,111,116,0,18,109,82,111,119,50,0,0,18,110,0,16,10,50,0,57,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,58,100,111,116,0,18,109,82,111, -119,50,0,0,18,110,0,16,10,51,0,57,0,0,20,0,0,2,3,2,0,12,1,109,82,111,119,51,0,0,0,9,18,109,82,111, -119,51,0,59,120,0,18,109,0,16,8,48,0,57,59,119,0,20,0,9,18,109,82,111,119,51,0,59,121,0,18,109,0, -16,10,49,0,57,59,119,0,20,0,9,18,109,82,111,119,51,0,59,122,0,18,109,0,16,10,50,0,57,59,119,0,20,0, -9,18,109,82,111,119,51,0,59,119,0,18,109,0,16,10,51,0,57,59,119,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,59,119,0,58,100,111,116,0,18,109,82,111,119,51,0,0,18,110,0,16,8,48,0,57,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0,58,100,111,116,0,18,109,82,111,119, -51,0,0,18,110,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,119,0, -58,100,111,116,0,18,109,82,111,119,51,0,0,18,110,0,16,10,50,0,57,0,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,51,0,57,59,119,0,58,100,111,116,0,18,109,82,111,119,51,0,0,18,110,0,16,10,51,0, -57,0,0,20,0,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,20,0,0,1,0,13,2,26, -1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110, -0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49, -0,57,46,20,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,27,1,1,0,13, -109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, -18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47, -20,0,0,1,0,13,2,21,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, +16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0, +57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10, +51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2, +21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48, +0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0, +48,46,18,109,0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120, +0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, +57,18,110,0,16,10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57, +59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48, +0,57,18,110,0,16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57, +59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48, +46,18,109,0,16,10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120, +0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, +57,18,110,0,16,10,51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, +59,119,119,119,119,0,48,46,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49, +20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, +0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0, +13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13, +2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, +57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97, +0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48, +0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20, +0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, +16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2,26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0, +1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, +47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2, +27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47, +20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, 0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, -0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97,0,0,1,1,0, -13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,0,1,0,13, -2,22,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,98,0,49,20,0,0,1,0,14,2,26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -97,0,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0,1,1,0,14,110,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,27,1,1,0,14, -109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, -18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,0,1,0, -14,2,21,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97, -0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0, -16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0, -57,48,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16, -8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, -16,10,50,0,57,18,98,0,48,20,0,0,1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9,97,0,0, -1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, -46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1,1,0, -15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, -46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2, +0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0, +16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9, +98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9, +97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8, +48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1, +1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98, +0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2, 27,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, 110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, 49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47, @@ -622,244 +584,221 @@ 97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16, 10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, 18,109,0,16,10,50,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, -16,10,51,0,57,18,98,0,49,20,0,0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,114, -48,0,0,1,1,114,49,0,0,0,9,18,114,48,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,20,0,9,18,114,48,0, -59,121,0,18,109,0,16,10,49,0,57,59,120,0,20,0,9,18,114,49,0,59,120,0,18,109,0,16,8,48,0,57,59,121, -0,20,0,9,18,114,49,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,120,0,58,100,111,116,0,18,114,48,0,0,18,118,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,121,0,58,100,111,116,0,18,114,49,0,0,18,118,0,0,0,20,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,13, -109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8, -48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0, -16,10,49,0,57,0,0,20,0,0,1,0,11,2,21,1,1,0,14,109,0,0,1,1,0,11,118,0,0,0,1,2,3,2,0,11,1,114,48,0,0, -0,9,18,114,48,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,20,0,9,18,114,48,0,59,121,0,18,109,0,16,10, -49,0,57,59,120,0,20,0,9,18,114,48,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,20,0,4,118,101,99,51, -95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,48,0,0,18,118,0,0,0,0,2,3,2,0, -11,1,114,49,0,0,0,9,18,114,49,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,20,0,9,18,114,49,0,59,121, -0,18,109,0,16,10,49,0,57,59,121,0,20,0,9,18,114,49,0,59,122,0,18,109,0,16,10,50,0,57,59,121,0,20,0, -4,118,101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,49,0,0,18,118, -0,0,0,0,2,3,2,0,11,1,114,50,0,0,0,9,18,114,50,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,20,0,9,18, -114,50,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,20,0,9,18,114,50,0,59,122,0,18,109,0,16,10,50,0, -57,59,122,0,20,0,4,118,101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18, -114,50,0,0,18,118,0,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0, -1,0,12,2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,2,3,2,0,12,1,114,48,0,0,0,9,18,114,48,0,59,120,0, -18,109,0,16,8,48,0,57,59,120,0,20,0,9,18,114,48,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,20,0,9, -18,114,48,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,20,0,9,18,114,48,0,59,119,0,18,109,0,16,10,51, -0,57,59,120,0,20,0,4,118,101,99,52,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,59,115,0,0,18, -114,48,0,0,18,118,0,0,0,0,2,3,2,0,12,1,114,49,0,0,0,9,18,114,49,0,59,120,0,18,109,0,16,8,48,0,57, -59,121,0,20,0,9,18,114,49,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,20,0,9,18,114,49,0,59,122,0, -18,109,0,16,10,50,0,57,59,121,0,20,0,9,18,114,49,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,20,0,4, -118,101,99,52,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,49,0,0,18,118,0, -0,0,0,2,3,2,0,12,1,114,50,0,0,0,9,18,114,50,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,20,0,9,18, -114,50,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,20,0,9,18,114,50,0,59,122,0,18,109,0,16,10,50,0, -57,59,122,0,20,0,9,18,114,50,0,59,119,0,18,109,0,16,10,51,0,57,59,122,0,20,0,4,118,101,99,52,95, -100,111,116,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,50,0,0,18,118,0,0,0,0,2,3,2,0,12, -1,114,51,0,0,0,9,18,114,51,0,59,120,0,18,109,0,16,8,48,0,57,59,119,0,20,0,9,18,114,51,0,59,121,0, -18,109,0,16,10,49,0,57,59,119,0,20,0,9,18,114,51,0,59,122,0,18,109,0,16,10,50,0,57,59,119,0,20,0,9, -18,114,51,0,59,119,0,18,109,0,16,10,51,0,57,59,119,0,20,0,4,118,101,99,52,95,100,111,116,0,18,95, -95,114,101,116,86,97,108,0,59,119,0,0,18,114,51,0,0,18,118,0,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0, -1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18, -109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0, -0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18, -118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116, -0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9, -18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0, -57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13, -110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1, -9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0, -57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16, -10,50,0,57,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0, -16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18, -110,0,16,10,50,0,57,22,0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18, -110,0,48,20,0,0,1,0,0,2,4,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0, -16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18, -110,0,16,10,50,0,57,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10, -50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2, -2,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18, +16,10,51,0,57,18,98,0,49,20,0,0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0, +59,121,121,0,48,46,20,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21, +1,1,0,14,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18, +118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50, +0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, +57,0,0,20,0,0,1,0,12,2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121, +121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57, +18,118,0,59,119,119,119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, +57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10, +51,0,57,0,0,20,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110, +0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109, +0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0, +57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109, +0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14, +109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2, +2,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18, 109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22, -0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0, -0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18, -109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0, -9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0, -57,24,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18, -109,0,16,10,49,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97, -0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0, -2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18, -97,0,24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9, -18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109, -0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0, -9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0, -23,0,0,1,0,0,2,4,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109, -0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1, -1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18, -109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0, -0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9, -18,109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109, -0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0, -9,18,109,0,16,10,50,0,57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15, -109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0, -24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2, -10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0, -1,1,0,14,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15, -109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0, -16,10,49,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0, -1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,20,0,0,1,0,7,2,25,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16, -10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0, -0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,20,0,0,1,0,9,2,25,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0, -58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1, -0,11,2,25,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18, -118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20, -0,0,1,0,13,2,25,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99, -50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0, -17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14, -109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47, -20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0, -9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48, -0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57, -18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18, -109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109, -0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0, -18,109,0,20,0,0,1,0,5,2,24,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,24,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118, -101,99,50,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1, -0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,24,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118, -101,99,52,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1, -0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97, -0,20,0,0,1,0,10,2,24,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18, -118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,0,1,0,12,2,24,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1, -9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18, -109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95, -95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,24,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109, -0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0, -18,109,0,20,0,0,1,0,15,2,24,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58, -118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118, -101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101, -99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99, -52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,0,95,95, -112,111,115,116,68,101,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0, -9,18,97,0,18,97,0,16,10,49,0,47,20,0,0,1,0,6,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,6,118,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0, -16,10,49,0,0,0,47,20,0,0,1,0,7,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0, -47,20,0,0,1,0,8,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0, -9,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -97,0,20,0,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,0,1,0,10,0,95,95,112,111,115,116,68,101,99,114, -0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118, -101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,11,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,11, -118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51, -0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,12,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48, -0,0,0,0,47,20,0,0,1,0,13,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99, -50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0, -17,49,0,48,0,0,0,0,47,20,0,0,1,0,14,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118, -101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101, -99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,15,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, +0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2, +4,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18, +109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24, +0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, +50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1, +0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51, +0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18, +109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57, +18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10, +50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2, +1,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, +57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22, +0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, +16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0, +9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2, +1,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, +57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0, +0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50, +0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2, +14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97, +0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18, +109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18, +97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9, +18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57, +18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1, +9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0, +57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0, +0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50, +0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109, +0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9, +18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18, +118,0,18,109,0,48,20,0,0,1,0,5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18, +95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58, +105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +7,2,25,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18, +95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0, +58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1, +0,9,2,25,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,18,97,0,20,0,0,1,0,10,2,25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49, +0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0, +0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,118,0,20,0,0,1,0,12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0, +17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13, +109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47, +20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0, +9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16, +8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49, +0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57, +18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,18,109,0,20,0,0,1,0,15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, 57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, 58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, 118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118, -101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,9,97, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0, -10,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95, -112,111,115,116,73,110,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0, -20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111, -115,116,73,110,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18, -118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,5,0,95,95,112,111,115,116,73, -110,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0, -16,10,49,0,46,20,0,0,1,0,6,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46, -20,0,0,1,0,7,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0, -95,95,112,111,115,116,73,110,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112, -111,115,116,73,110,99,114,0,1,0,2,13,109,0,0,0,1,3,2,0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8, +101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2, +24,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, +97,0,20,0,0,1,0,6,2,24,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0, +0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18, +118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,0,1,0,8,2,24,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0, +0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18, +97,0,18,97,0,17,49,0,48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24, +1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95, +114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +12,2,24,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, 48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95, -95,112,111,115,116,73,110,99,114,0,1,0,2,14,109,0,0,0,1,3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10, -49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95, -95,112,111,115,116,73,110,99,114,0,1,0,2,15,109,0,0,0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10, -49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57, -18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1, -0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,98,0,0,18,97,0,0,0,0,1,0,1,2,15,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97, -116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98, -0,0,0,1,3,2,0,1,1,99,0,0,0,4,102,108,111,97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0, -0,8,18,99,0,0,0,1,0,1,2,16,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0, -58,102,108,111,97,116,0,18,98,0,0,0,41,0,0,1,0,1,2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1, -103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4, -102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32, -0,0,1,0,1,2,18,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108, -111,97,116,0,18,98,0,0,0,43,0,0,1,0,1,2,17,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1, -101,0,0,0,4,102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111, -97,116,95,101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2, -17,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0, -18,98,0,0,0,42,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97, -116,95,112,114,105,110,116,0,18,102,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105, -0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83, -65,0,1,1,0,1,98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114, -105,110,116,77,69,83,65,0,1,1,0,10,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59, -120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105, -110,116,77,69,83,65,0,1,1,0,11,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120, -0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77, -69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0, -1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83, -65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58, -112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83, -65,0,1,1,0,6,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112, -114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0, -1,1,0,7,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, -59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105, -110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59, -121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,2,118, -0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77, -69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1, -9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83, -65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0, -0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,4,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0, -18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112, -114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114, -105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -109,0,16,10,49,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,14,109,0,0,0,1,9,58, -112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65, -0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0, -0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,15,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83, -65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0, -0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77, -69,83,65,0,18,109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,16,101,0, -0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83, -65,0,1,1,0,17,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114, -105,110,116,77,69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0, -0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0,1,4,105,110,116,95,112,114,105,110, -116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,20,101,0,0,0,1,4,105,110,116, -95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,21,101,0,0, -0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 +57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,109,0,20,0,0,1,0,14,2,24,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, +15,2,24,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, +0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101, +99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16, +10,49,0,47,20,0,0,1,0,6,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20, +0,0,1,0,7,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95, +95,112,111,115,116,68,101,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0, +20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115, +116,68,101,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0, +18,97,0,17,49,0,48,0,0,47,20,0,0,1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0, +48,0,0,0,0,47,20,0,0,1,0,11,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, +47,20,0,0,1,0,12,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0, +1,0,13,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0, +0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0, +0,47,20,0,0,1,0,14,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17, +49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0, +48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0, +0,0,0,47,20,0,0,1,0,15,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0, +17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115, +116,73,110,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118, +0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110, +99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0, +2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101, +99,52,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0, +95,95,112,111,115,116,73,110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112, +111,115,116,73,110,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, +18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73, +110,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, +118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114, +0,1,0,2,13,109,0,0,0,1,3,2,0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110, +99,114,0,1,0,2,14,109,0,0,0,1,3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110, +99,114,0,1,0,2,15,109,0,0,0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, +58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58, +118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0, +0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0, +0,0,0,1,0,1,2,15,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102, +108,111,97,116,0,18,98,0,0,0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0, +0,4,102,108,111,97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2, +16,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0, +18,98,0,0,0,41,0,0,1,0,1,2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4, +102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95, +101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0, +5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0, +0,0,43,0,0,1,0,1,2,17,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108, +111,97,116,95,108,101,115,115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113, +117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0, +1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0, +0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114, +105,110,116,0,18,102,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105, +110,116,95,112,114,105,110,116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1, +98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77, +69,83,65,0,1,1,0,10,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, +58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, +83,65,0,1,1,0,11,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58, +112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, +18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112, +114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114, +105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1, +0,6,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105, +110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7, +118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110, +116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0, +0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77, +69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0, +9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83, +65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58, +112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, +18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112, +114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112, +114,105,110,116,77,69,83,65,0,1,1,0,4,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, +59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110, +116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0, +0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77, +69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0, +57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116, +77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49, +0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114, +105,110,116,77,69,83,65,0,1,1,0,15,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16, +8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114, +105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, +109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105, +110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17, +101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77, +69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112, +114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0, +0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105, +110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110, +116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 15a0e3abcd..8ea6a8dcd9 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2514,49 +2514,6 @@ _slang_gen_subscript(slang_assemble_ctx * A, slang_operation *oper) } -/** - * Look for expressions such as: gl_ModelviewMatrix * gl_Vertex - * and replace with this: gl_Vertex * gl_ModelviewMatrixTranpose - * Since matrices are stored in column-major order, the second form of - * multiplication is much more efficient (just 4 dot products). - */ -static void -_slang_check_matmul_optimization(slang_assemble_ctx *A, slang_operation *oper) -{ - static const struct { - const char *orig; - const char *tranpose; - } matrices[] = { - {"gl_ModelViewMatrix", "gl_ModelViewMatrixTranspose"}, - {"gl_ProjectionMatrix", "gl_ProjectionMatrixTranspose"}, - {"gl_ModelViewProjectionMatrix", "gl_ModelViewProjectionMatrixTranspose"}, - {"gl_TextureMatrix", "gl_TextureMatrixTranspose"}, - {"gl_NormalMatrix", "__NormalMatrixTranspose"}, - { NULL, NULL } - }; - - assert(oper->type == SLANG_OPER_MULTIPLY); - if (oper->children[0].type == SLANG_OPER_IDENTIFIER) { - GLuint i; - for (i = 0; matrices[i].orig; i++) { - if (oper->children[0].a_id - == slang_atom_pool_atom(A->atoms, matrices[i].orig)) { - /* - _mesa_printf("Replace %s with %s\n", - matrices[i].orig, matrices[i].tranpose); - */ - assert(oper->children[0].type == SLANG_OPER_IDENTIFIER); - oper->children[0].a_id - = slang_atom_pool_atom(A->atoms, matrices[i].tranpose); - /* finally, swap the operands */ - _slang_operation_swap(&oper->children[0], &oper->children[1]); - return; - } - } - } -} - - /** * Generate IR tree for a slang_operation (AST node) */ @@ -2690,7 +2647,6 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) { slang_ir_node *n; assert(oper->num_children == 2); - _slang_check_matmul_optimization(A, oper); n = _slang_gen_function_call_name(A, "*", oper, NULL); return n; } -- cgit v1.2.3 From 215a847e3fe98def292d929baecb3876abe5027a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:27:34 -0600 Subject: mesa: prefix rm command with - --- src/mesa/shader/slang/library/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/Makefile b/src/mesa/shader/slang/library/Makefile index dc67b59088..0e03fac2ee 100644 --- a/src/mesa/shader/slang/library/Makefile +++ b/src/mesa/shader/slang/library/Makefile @@ -17,7 +17,7 @@ LIB_DEP = $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) default: syntax builtin clean: - rm -f syn_to_c gc_to_bin *_syn.h *_gc.h + -rm -f syn_to_c gc_to_bin *_syn.h *_gc.h syntax: slang_pp_directives_syn.h slang_pp_expression_syn.h slang_shader_syn.h slang_pp_version_syn.h -- cgit v1.2.3 From 7aed3751c7da47db0d2e5f7ebf58a915ee772bd3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:30:15 -0600 Subject: mesa: rm unneeded file --- src/mesa/shader/slang/library/slang_version_syn.h | 64 ----------------------- 1 file changed, 64 deletions(-) delete mode 100755 src/mesa/shader/slang/library/slang_version_syn.h (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_version_syn.h b/src/mesa/shader/slang/library/slang_version_syn.h deleted file mode 100755 index 3b94d85927..0000000000 --- a/src/mesa/shader/slang/library/slang_version_syn.h +++ /dev/null @@ -1,64 +0,0 @@ -".syntax version_directive;\n" -"version_directive\n" -" version_directive_1 .and .loop version_directive_2;\n" -"version_directive_1\n" -" prior_optional_spaces .and optional_version_directive .and .true .emit $;\n" -"version_directive_2\n" -" prior_optional_spaces .and version_directive_body .and .true .emit $;\n" -"optional_version_directive\n" -" version_directive_body .or .true .emit 10 .emit 1;\n" -"version_directive_body\n" -" '#' .and optional_space .and \"version\" .and space .and version_number .and optional_space .and\n" -" new_line;\n" -"version_number\n" -" version_number_110;\n" -"version_number_110\n" -" leading_zeroes .and \"110\" .emit 10 .emit 1;\n" -"leading_zeroes\n" -" .loop zero;\n" -"zero\n" -" '0';\n" -"space\n" -" single_space .and .loop single_space;\n" -"optional_space\n" -" .loop single_space;\n" -"single_space\n" -" ' ' .or '\\t';\n" -"prior_optional_spaces\n" -" .loop prior_space;\n" -"prior_space\n" -" c_style_comment_block .or cpp_style_comment_block .or space .or new_line;\n" -"c_style_comment_block\n" -" '/' .and '*' .and c_style_comment_rest;\n" -"c_style_comment_rest\n" -" .loop c_style_comment_char_no_star .and c_style_comment_rest_1;\n" -"c_style_comment_rest_1\n" -" c_style_comment_end .or c_style_comment_rest_2;\n" -"c_style_comment_rest_2\n" -" '*' .and c_style_comment_rest;\n" -"c_style_comment_char_no_star\n" -" '\\x2B'-'\\xFF' .or '\\x01'-'\\x29';\n" -"c_style_comment_end\n" -" '*' .and '/';\n" -"cpp_style_comment_block\n" -" '/' .and '/' .and cpp_style_comment_block_1;\n" -"cpp_style_comment_block_1\n" -" cpp_style_comment_block_2 .or cpp_style_comment_block_3;\n" -"cpp_style_comment_block_2\n" -" .loop cpp_style_comment_char .and new_line;\n" -"cpp_style_comment_block_3\n" -" .loop cpp_style_comment_char;\n" -"cpp_style_comment_char\n" -" '\\x0E'-'\\xFF' .or '\\x01'-'\\x09' .or '\\x0B'-'\\x0C';\n" -"new_line\n" -" cr_lf .or lf_cr .or '\\n' .or '\\r';\n" -"cr_lf\n" -" '\\r' .and '\\n';\n" -"lf_cr\n" -" '\\n' .and '\\r';\n" -".string __string_filter;\n" -"__string_filter\n" -" .loop __identifier_char;\n" -"__identifier_char\n" -" 'a'-'z' .or 'A'-'Z' .or '_' .or '0'-'9';\n" -"" -- cgit v1.2.3 From 087fa7e39075228fcb238537ee251ab47f77a0cb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:32:18 -0600 Subject: fix-up inlined/non-inlined function inconsistencies cherry-picked from master --- src/mesa/shader/slang/slang_codegen.c | 16 ++++++++-------- src/mesa/shader/slang/slang_compile_operation.h | 2 +- 2 files changed, 9 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 8ea6a8dcd9..8e4c10ff43 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -538,10 +538,10 @@ new_not(slang_ir_node *n) /** - * Inlined subroutine. + * Non-inlined function call. */ static slang_ir_node * -new_inlined_function_call(slang_ir_node *code, slang_label *name) +new_function_call(slang_ir_node *code, slang_label *name) { slang_ir_node *n = new_node1(IR_CALL, code); assert(name); @@ -1229,7 +1229,7 @@ _slang_gen_function_call(slang_assemble_ctx *A, slang_function *fun, else { callOper = inlined; } - callOper->type = SLANG_OPER_INLINED_CALL; + callOper->type = SLANG_OPER_NON_INLINED_CALL; callOper->fun = fun; callOper->label = _slang_label_new_unique((char*) fun->header.a_name); } @@ -2552,7 +2552,7 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) _slang_free_ir_tree(tree); return NULL; /* error must have occured */ } - tree = tree ? new_seq(tree, n) : n; + tree = new_seq(tree, n); } #if 00 @@ -2779,17 +2779,17 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) return n; } - case SLANG_OPER_INLINED_CALL: + case SLANG_OPER_NON_INLINED_CALL: case SLANG_OPER_SEQUENCE: { slang_ir_node *tree = NULL; GLuint i; for (i = 0; i < oper->num_children; i++) { slang_ir_node *n = _slang_gen_operation(A, &oper->children[i]); - tree = tree ? new_seq(tree, n) : n; + tree = new_seq(tree, n); } - if (oper->type == SLANG_OPER_INLINED_CALL) { - tree = new_inlined_function_call(tree, oper->label); + if (oper->type == SLANG_OPER_NON_INLINED_CALL) { + tree = new_function_call(tree, oper->label); } return tree; } diff --git a/src/mesa/shader/slang/slang_compile_operation.h b/src/mesa/shader/slang/slang_compile_operation.h index d497b6f66f..d5cbe779a6 100644 --- a/src/mesa/shader/slang/slang_compile_operation.h +++ b/src/mesa/shader/slang/slang_compile_operation.h @@ -93,7 +93,7 @@ typedef enum slang_operation_type_ SLANG_OPER_NOT, /* "!" [expr] */ SLANG_OPER_SUBSCRIPT, /* [expr] "[" [expr] "]" */ SLANG_OPER_CALL, /* [func name] [param] [param] [...] */ - SLANG_OPER_INLINED_CALL, /* inlined function call */ + SLANG_OPER_NON_INLINED_CALL, /* a real function call */ SLANG_OPER_FIELD, /* i.e.: ".next" or ".xzy" or ".xxx" etc */ SLANG_OPER_POSTINCREMENT, /* [var] "++" */ SLANG_OPER_POSTDECREMENT /* [var] "--" */ -- cgit v1.2.3 From 217e5b555189c1bd8213b439130944a18118d23b Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 7 Aug 2007 21:42:34 +0100 Subject: fix swizzle error test (bug 11881) --- src/mesa/shader/atifragshader.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/atifragshader.c b/src/mesa/shader/atifragshader.c index 4727c1a436..854c911874 100644 --- a/src/mesa/shader/atifragshader.c +++ b/src/mesa/shader/atifragshader.c @@ -440,7 +440,7 @@ _mesa_PassTexCoordATI(GLuint dst, GLuint coord, GLenum swizzle) _mesa_error(ctx, GL_INVALID_OPERATION, "glPassTexCoordATI(coord)"); return; } - if ((swizzle < GL_SWIZZLE_STR_ATI) && (swizzle > GL_SWIZZLE_STQ_DQ_ATI)) { + if (!(swizzle >= GL_SWIZZLE_STR_ATI) && (swizzle <= GL_SWIZZLE_STQ_DQ_ATI)) { _mesa_error(ctx, GL_INVALID_ENUM, "glPassTexCoordATI(swizzle)"); return; } @@ -513,7 +513,7 @@ _mesa_SampleMapATI(GLuint dst, GLuint interp, GLenum swizzle) _mesa_error(ctx, GL_INVALID_OPERATION, "glSampleMapATI(interp)"); return; } - if ((swizzle < GL_SWIZZLE_STR_ATI) && (swizzle > GL_SWIZZLE_STQ_DQ_ATI)) { + if (!(swizzle >= GL_SWIZZLE_STR_ATI) && (swizzle <= GL_SWIZZLE_STQ_DQ_ATI)) { _mesa_error(ctx, GL_INVALID_ENUM, "glSampleMapATI(swizzle)"); return; } -- cgit v1.2.3 From 23db6287d41494799a4d171bae628b0d2e9baf36 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Tue, 15 Jan 2008 15:08:34 -0800 Subject: Bug #13492: Only call ProgramStringNotify if program parsing succeeded. Wine intentionally tries some out-of-spec programs to test strictness, and calling ProgramStringNotify on the results of a failed program parse resulted in crashes in the 965 driver. --- src/mesa/shader/arbprogram.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c index 4b5f63ef92..ee75be315e 100644 --- a/src/mesa/shader/arbprogram.c +++ b/src/mesa/shader/arbprogram.c @@ -229,7 +229,7 @@ _mesa_ProgramStringARB(GLenum target, GLenum format, GLsizei len, struct gl_vertex_program *prog = ctx->VertexProgram.Current; _mesa_parse_arb_vertex_program(ctx, target, string, len, prog); - if (ctx->Driver.ProgramStringNotify) + if (ctx->Program.ErrorPos == -1 && ctx->Driver.ProgramStringNotify) ctx->Driver.ProgramStringNotify( ctx, target, &prog->Base ); } else if (target == GL_FRAGMENT_PROGRAM_ARB @@ -237,7 +237,7 @@ _mesa_ProgramStringARB(GLenum target, GLenum format, GLsizei len, struct gl_fragment_program *prog = ctx->FragmentProgram.Current; _mesa_parse_arb_fragment_program(ctx, target, string, len, prog); - if (ctx->Driver.ProgramStringNotify) + if (ctx->Program.ErrorPos == -1 && ctx->Driver.ProgramStringNotify) ctx->Driver.ProgramStringNotify( ctx, target, &prog->Base ); } else { -- cgit v1.2.3 From 887bfee6e3beeacb441b81ac225d99060d4b5dfc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 16:44:08 -0600 Subject: mesa: fix InfoLog memleak, misc sync with master --- src/mesa/shader/shader_api.c | 12 ++++++++---- 1 file changed, 8 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index d0512b8ed1..badc457c1c 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -126,10 +126,17 @@ _mesa_free_shader_program_data(GLcontext *ctx, for (i = 0; i < shProg->NumShaders; i++) { _mesa_reference_shader(ctx, &shProg->Shaders[i], NULL); } + shProg->NumShaders = 0; + if (shProg->Shaders) { _mesa_free(shProg->Shaders); shProg->Shaders = NULL; } + + if (shProg->InfoLog) { + _mesa_free(shProg->InfoLog); + shProg->InfoLog = NULL; + } } @@ -140,10 +147,7 @@ void _mesa_free_shader_program(GLcontext *ctx, struct gl_shader_program *shProg) { _mesa_free_shader_program_data(ctx, shProg); - if (shProg->Shaders) { - _mesa_free(shProg->Shaders); - shProg->Shaders = NULL; - } + _mesa_free(shProg); } -- cgit v1.2.3 From 826218d386a0193f7c6d576b768c4427f09bb109 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 19:14:41 -0600 Subject: Consolidate texture fetch code and use partial derivatives when possible. picked from master --- src/mesa/shader/prog_execute.c | 134 +++++++++++++++++++++-------------------- src/mesa/shader/prog_execute.h | 3 +- src/mesa/swrast/s_fragprog.c | 2 +- 3 files changed, 71 insertions(+), 68 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c index bd64b57eb9..8ce2ca3964 100644 --- a/src/mesa/shader/prog_execute.c +++ b/src/mesa/shader/prog_execute.c @@ -1,6 +1,6 @@ /* * Mesa 3-D graphics library - * Version: 6.5.3 + * Version: 7.0.3 * * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * @@ -46,9 +46,6 @@ #include "shader/slang/slang_library_noise.h" -/* See comments below for info about this */ -#define LAMBDA_ZERO 1 - /* debug predicate */ #define DEBUG_PROG 0 @@ -92,6 +89,7 @@ get_register_pointer(const struct prog_src_register *source, else { const struct gl_program_parameter_list *params; ASSERT(source->File == PROGRAM_LOCAL_PARAM || + source->File == PROGRAM_CONSTANT || source->File == PROGRAM_STATE_VAR); params = machine->CurProgram->Parameters; if (reg < 0 || reg >= params->NumParameters) @@ -302,6 +300,36 @@ fetch_vector1(const struct prog_src_register *source, } +/** + * Fetch texel from texture. Use partial derivatives when possible. + */ +static INLINE void +fetch_texel(GLcontext *ctx, + const struct gl_program_machine *machine, + const struct prog_instruction *inst, + const GLfloat texcoord[4], GLfloat lodBias, + GLfloat color[4]) +{ + const GLuint unit = machine->Samplers[inst->TexSrcUnit]; + + /* Note: we only have the right derivatives for fragment input attribs. + */ + if (machine->NumDeriv > 0 && + inst->SrcReg[0].File == PROGRAM_INPUT && + inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) { + /* simple texture fetch for which we should have derivatives */ + GLuint attr = inst->SrcReg[0].Index; + machine->FetchTexelDeriv(ctx, texcoord, + machine->DerivX[attr], + machine->DerivY[attr], + lodBias, unit, color); + } + else { + machine->FetchTexelLod(ctx, texcoord, lodBias, unit, color); + } +} + + /** * Test value against zero and return GT, LT, EQ or UN if NaN. */ @@ -1306,33 +1334,18 @@ _mesa_execute_program(GLcontext * ctx, } break; case OPCODE_TEX: /* Both ARB and NV frag prog */ - /* Texel lookup */ + /* Simple texel lookup */ { - /* Note: only use the precomputed lambda value when we're - * sampling texture unit [K] with texcoord[K]. - * Otherwise, the lambda value may have no relation to the - * instruction's texcoord or texture image. Using the wrong - * lambda is usually bad news. - * The rest of the time, just use zero (until we get a more - * sophisticated way of computing lambda). - */ - const GLuint unit = machine->Samplers[inst->TexSrcUnit]; - GLfloat coord[4], color[4], lambda; -#if 0 - if (inst->SrcReg[0].File == PROGRAM_INPUT && - inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + unit) - lambda = span->array->lambda[unit][column]; - else -#endif - lambda = 0.0; - fetch_vector4(&inst->SrcReg[0], machine, coord); - machine->FetchTexelLod(ctx, coord, lambda, unit, color); + GLfloat texcoord[4], color[4]; + fetch_vector4(&inst->SrcReg[0], machine, texcoord); + + fetch_texel(ctx, machine, inst, texcoord, 0.0, color); + if (DEBUG_PROG) { - printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g], " - "lod %f\n", + printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n", color[0], color[1], color[2], color[3], - unit, - coord[0], coord[1], coord[2], coord[3], lambda); + inst->TexSrcUnit, + texcoord[0], texcoord[1], texcoord[2], texcoord[3]); } store_vector4(inst, machine, color); } @@ -1342,21 +1355,18 @@ _mesa_execute_program(GLcontext * ctx, { const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[inst->TexSrcUnit]; - GLfloat coord[4], color[4], lambda, bias; -#if 0 - if (inst->SrcReg[0].File == PROGRAM_INPUT && - inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) - lambda = span->array->lambda[inst->TexSrcUnit][column]; - else -#endif - lambda = 0.0; - fetch_vector4(&inst->SrcReg[0], machine, coord); - /* coord[3] is the bias to add to lambda */ - bias = texUnit->LodBias + coord[3]; - if (texUnit->_Current) - bias += texUnit->_Current->LodBias; - machine->FetchTexelLod(ctx, coord, lambda + bias, - inst->TexSrcUnit, color); + GLfloat texcoord[4], color[4], lodBias; + + fetch_vector4(&inst->SrcReg[0], machine, texcoord); + + /* texcoord[3] is the bias to add to lambda */ + lodBias = texUnit->LodBias + texcoord[3]; + if (texUnit->_Current) { + lodBias += texUnit->_Current->LodBias; + } + + fetch_texel(ctx, machine, inst, texcoord, lodBias, color); + store_vector4(inst, machine, color); } break; @@ -1368,6 +1378,7 @@ _mesa_execute_program(GLcontext * ctx, fetch_vector4(&inst->SrcReg[1], machine, dtdx); fetch_vector4(&inst->SrcReg[2], machine, dtdy); machine->FetchTexelDeriv(ctx, texcoord, dtdx, dtdy, + 0.0, /* lodBias */ inst->TexSrcUnit, color); store_vector4(inst, machine, color); } @@ -1375,15 +1386,8 @@ _mesa_execute_program(GLcontext * ctx, case OPCODE_TXP: /* GL_ARB_fragment_program only */ /* Texture lookup w/ projective divide */ { - const GLuint unit = machine->Samplers[inst->TexSrcUnit]; - GLfloat texcoord[4], color[4], lambda; -#if 0 - if (inst->SrcReg[0].File == PROGRAM_INPUT && - inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) - lambda = span->array->lambda[unit][column]; - else -#endif - lambda = 0.0; + GLfloat texcoord[4], color[4]; + fetch_vector4(&inst->SrcReg[0], machine, texcoord); /* Not so sure about this test - if texcoord[3] is * zero, we'd probably be fine except for an ASSERT in @@ -1394,21 +1398,19 @@ _mesa_execute_program(GLcontext * ctx, texcoord[1] /= texcoord[3]; texcoord[2] /= texcoord[3]; } - machine->FetchTexelLod(ctx, texcoord, lambda, unit, color); + + fetch_texel(ctx, machine, inst, texcoord, 0.0, color); + store_vector4(inst, machine, color); } break; case OPCODE_TXP_NV: /* GL_NV_fragment_program only */ - /* Texture lookup w/ projective divide */ + /* Texture lookup w/ projective divide, as above, but do not + * do the divide by w if sampling from a cube map. + */ { - GLfloat texcoord[4], color[4], lambda; -#if 0 - if (inst->SrcReg[0].File == PROGRAM_INPUT && - inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) - lambda = span->array->lambda[inst->TexSrcUnit][column]; - else -#endif - lambda = 0.0; + GLfloat texcoord[4], color[4]; + fetch_vector4(&inst->SrcReg[0], machine, texcoord); if (inst->TexSrcTarget != TEXTURE_CUBE_INDEX && texcoord[3] != 0.0) { @@ -1416,8 +1418,9 @@ _mesa_execute_program(GLcontext * ctx, texcoord[1] /= texcoord[3]; texcoord[2] /= texcoord[3]; } - machine->FetchTexelLod(ctx, texcoord, lambda, - inst->TexSrcUnit, color); + + fetch_texel(ctx, machine, inst, texcoord, 0.0, color); + store_vector4(inst, machine, color); } break; @@ -1517,10 +1520,9 @@ _mesa_execute_program(GLcontext * ctx, case OPCODE_END: return GL_TRUE; default: - _mesa_problem(ctx, "Bad opcode %d in _mesa_exec_fragment_program", + _mesa_problem(ctx, "Bad opcode %d in _mesa_execute_program", inst->Opcode); return GL_TRUE; /* return value doesn't matter */ - } numExec++; diff --git a/src/mesa/shader/prog_execute.h b/src/mesa/shader/prog_execute.h index af77d4278b..18b13e11a4 100644 --- a/src/mesa/shader/prog_execute.h +++ b/src/mesa/shader/prog_execute.h @@ -1,6 +1,6 @@ /* * Mesa 3-D graphics library - * Version: 6.5.3 + * Version: 7.0.3 * * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * @@ -32,6 +32,7 @@ typedef void (*FetchTexelLodFunc)(GLcontext *ctx, const GLfloat texcoord[4], typedef void (*FetchTexelDerivFunc)(GLcontext *ctx, const GLfloat texcoord[4], const GLfloat texdx[4], const GLfloat texdy[4], + GLfloat lodBias, GLuint unit, GLfloat color[4]); diff --git a/src/mesa/swrast/s_fragprog.c b/src/mesa/swrast/s_fragprog.c index 40ed48696c..8eeb40f7c5 100644 --- a/src/mesa/swrast/s_fragprog.c +++ b/src/mesa/swrast/s_fragprog.c @@ -64,7 +64,7 @@ fetch_texel_lod( GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda, static void fetch_texel_deriv( GLcontext *ctx, const GLfloat texcoord[4], const GLfloat texdx[4], const GLfloat texdy[4], - GLuint unit, GLfloat color[4] ) + GLfloat lodBias, GLuint unit, GLfloat color[4] ) { SWcontext *swrast = SWRAST_CONTEXT(ctx); const struct gl_texture_object *texObj = ctx->Texture.Unit[unit]._Current; -- cgit v1.2.3 From 5f6a726d9f2274df1a968e5f3cd3862894b47a38 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 14 May 2008 19:19:53 -0600 Subject: fix parsing of state.texenv.color (bug 14931) cherry-picked from master --- src/mesa/shader/arbprogram.syn | 4 ++-- src/mesa/shader/arbprogram_syn.h | 2 ++ 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogram.syn b/src/mesa/shader/arbprogram.syn index ba17d638a3..b12c6a0eda 100644 --- a/src/mesa/shader/arbprogram.syn +++ b/src/mesa/shader/arbprogram.syn @@ -1942,10 +1942,10 @@ stateTexEnvProperty fragment program ::= "" | "[" "]" - -NOTE: is not optional. */ optLegacyTexUnitNum + optLegacyTexUnitNum_1 .or .true .emit 0x00; +optLegacyTexUnitNum_1 lbracket_ne .and legacyTexUnitNum .and rbracket; /* diff --git a/src/mesa/shader/arbprogram_syn.h b/src/mesa/shader/arbprogram_syn.h index e0c901ea8c..d95a5dede4 100644 --- a/src/mesa/shader/arbprogram_syn.h +++ b/src/mesa/shader/arbprogram_syn.h @@ -926,6 +926,8 @@ "stateTexEnvProperty\n" " \"color\" .emit TEX_ENV_COLOR;\n" "optLegacyTexUnitNum\n" +" optLegacyTexUnitNum_1 .or .true .emit 0x00;\n" +"optLegacyTexUnitNum_1\n" " lbracket_ne .and legacyTexUnitNum .and rbracket;\n" "legacyTexUnitNum\n" " integer;\n" -- cgit v1.2.3 From ffbd455a36f1d7600bd5b60e91768133d901d2f4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 16 May 2008 15:34:22 -0600 Subject: fix an attr/src mix-up when setting-up/binding vertex arrays This fixes problems with incorrect material coefficients when glMaterial is called per-vertex. --- src/mesa/vbo/vbo_exec_draw.c | 2 +- src/mesa/vbo/vbo_save_draw.c | 10 ++++++---- 2 files changed, 7 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c index 23a3658386..6bb6849a6a 100644 --- a/src/mesa/vbo/vbo_exec_draw.c +++ b/src/mesa/vbo/vbo_exec_draw.c @@ -198,7 +198,7 @@ static void vbo_exec_bind_arrays( GLcontext *ctx ) arrays[attr].BufferObj = exec->vtx.bufferobj; arrays[attr]._MaxElement = count; /* ??? */ - data += exec->vtx.attrsz[attr] * sizeof(GLfloat); + data += exec->vtx.attrsz[src] * sizeof(GLfloat); } } } diff --git a/src/mesa/vbo/vbo_save_draw.c b/src/mesa/vbo/vbo_save_draw.c index 3c6f0fccd9..bf5c6d4eef 100644 --- a/src/mesa/vbo/vbo_save_draw.c +++ b/src/mesa/vbo/vbo_save_draw.c @@ -131,10 +131,12 @@ static void vbo_bind_vertex_list( GLcontext *ctx, break; } - for (attr = 0; attr < VBO_ATTRIB_MAX; attr++) { - if (node->attrsz[attr]) { + for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) { + GLuint src = map[attr]; + + if (node->attrsz[src]) { arrays[attr].Ptr = (const GLubyte *)data; - arrays[attr].Size = node->attrsz[attr]; + arrays[attr].Size = node->attrsz[src]; arrays[attr].StrideB = node->vertex_size * sizeof(GLfloat); arrays[attr].Stride = node->vertex_size * sizeof(GLfloat); arrays[attr].Type = GL_FLOAT; @@ -144,7 +146,7 @@ static void vbo_bind_vertex_list( GLcontext *ctx, assert(arrays[attr].BufferObj->Name); - data += node->attrsz[attr] * sizeof(GLfloat); + data += node->attrsz[src] * sizeof(GLfloat); } } } -- cgit v1.2.3 From 470f69950577e05e3dfd09653c268cf22c034d7f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 16 May 2008 09:56:59 -0600 Subject: Fix a program refcounting error, don't share program parameter lists. The refcounting bug was causing a memleak (unfreed programs). The old parameter list sharing is not needed since the change in how uniforms are handled. --- src/mesa/shader/shader_api.c | 29 ++++++++++------------------- src/mesa/shader/slang/slang_link.c | 32 ++++++++++---------------------- 2 files changed, 20 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index badc457c1c..fbaaf8ddea 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -75,21 +75,8 @@ void _mesa_clear_shader_program_data(GLcontext *ctx, struct gl_shader_program *shProg) { - if (shProg->VertexProgram) { - /* Set ptr to NULL since the param list is shared with the - * original/unlinked program. - */ - shProg->VertexProgram->Base.Parameters = NULL; - _mesa_reference_vertprog(ctx, &shProg->VertexProgram, NULL); - } - - if (shProg->FragmentProgram) { - /* Set ptr to NULL since the param list is shared with the - * original/unlinked program. - */ - shProg->FragmentProgram->Base.Parameters = NULL; - _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, NULL); - } + _mesa_reference_vertprog(ctx, &shProg->VertexProgram, NULL); + _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, NULL); if (shProg->Uniforms) { _mesa_free_uniform_list(shProg->Uniforms); @@ -176,8 +163,10 @@ _mesa_reference_shader_program(GLcontext *ctx, ASSERT(old->RefCount > 0); old->RefCount--; - /*printf("SHPROG DECR %p (%d) to %d\n", - (void*) old, old->Name, old->RefCount);*/ +#if 0 + printf("ShaderProgram %p ID=%u RefCount-- to %d\n", + (void *) old, old->Name, old->RefCount); +#endif deleteFlag = (old->RefCount == 0); if (deleteFlag) { @@ -191,8 +180,10 @@ _mesa_reference_shader_program(GLcontext *ctx, if (shProg) { shProg->RefCount++; - /*printf("SHPROG INCR %p (%d) to %d\n", - (void*) shProg, shProg->Name, shProg->RefCount);*/ +#if 0 + printf("ShaderProgram %p ID=%u RefCount++ to %d\n", + (void *) shProg, shProg->Name, shProg->RefCount); +#endif *ptr = shProg; } } diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 8213b7772c..80cd4b6df6 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -406,20 +406,20 @@ _slang_link(GLcontext *ctx, * Make copies of the vertex/fragment programs now since we'll be * changing src/dst registers after merging the uniforms and varying vars. */ + _mesa_reference_vertprog(ctx, &shProg->VertexProgram, NULL); if (vertProg) { - _mesa_reference_vertprog(ctx, &shProg->VertexProgram, - vertex_program(_mesa_clone_program(ctx, &vertProg->Base))); - } - else { - _mesa_reference_vertprog(ctx, &shProg->VertexProgram, NULL); + struct gl_vertex_program *linked_vprog = + vertex_program(_mesa_clone_program(ctx, &vertProg->Base)); + shProg->VertexProgram = linked_vprog; /* refcount OK */ + ASSERT(shProg->VertexProgram->Base.RefCount == 1); } + _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, NULL); if (fragProg) { - _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, - fragment_program(_mesa_clone_program(ctx, &fragProg->Base))); - } - else { - _mesa_reference_fragprog(ctx, &shProg->FragmentProgram, NULL); + struct gl_fragment_program *linked_fprog = + fragment_program(_mesa_clone_program(ctx, &fragProg->Base)); + shProg->FragmentProgram = linked_fprog; /* refcount OK */ + ASSERT(shProg->FragmentProgram->Base.RefCount == 1); } /* link varying vars */ @@ -436,18 +436,6 @@ _slang_link(GLcontext *ctx, /*_mesa_print_uniforms(shProg->Uniforms);*/ - if (shProg->VertexProgram) { - /* Rather than cloning the parameter list here, just share it. - * We need to be careful _mesa_clear_shader_program_data() in - * to avoid double-freeing. - */ - shProg->VertexProgram->Base.Parameters = vertProg->Base.Parameters; - } - if (shProg->FragmentProgram) { - /* see comment just above */ - shProg->FragmentProgram->Base.Parameters = fragProg->Base.Parameters; - } - if (shProg->VertexProgram) { if (!_slang_resolve_attributes(shProg, &shProg->VertexProgram->Base)) { /*goto cleanup;*/ -- cgit v1.2.3 From fc0a48d587e3aaae4b0e1fc0b9652c7f3d8e1d4b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 16 May 2008 15:48:11 -0600 Subject: bring in fixes/changes from master --- src/mesa/shader/shader_api.c | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index fbaaf8ddea..24ab7568d6 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -369,7 +369,7 @@ _mesa_attach_shader(GLcontext *ctx, GLuint program, GLuint shader) struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, program); struct gl_shader *sh = _mesa_lookup_shader(ctx, shader); - const GLuint n = shProg->NumShaders; + GLuint n; GLuint i; if (!shProg || !sh) { @@ -378,6 +378,8 @@ _mesa_attach_shader(GLcontext *ctx, GLuint program, GLuint shader) return; } + n = shProg->NumShaders; + for (i = 0; i < n; i++) { if (shProg->Shaders[i] == sh) { /* already attached */ @@ -570,7 +572,7 @@ _mesa_detach_shader(GLcontext *ctx, GLuint program, GLuint shader) { struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, program); - const GLuint n = shProg->NumShaders; + GLuint n; GLuint i, j; if (!shProg) { @@ -579,6 +581,8 @@ _mesa_detach_shader(GLcontext *ctx, GLuint program, GLuint shader) return; } + n = shProg->NumShaders; + for (i = 0; i < n; i++) { if (shProg->Shaders[i]->Name == shader) { /* found it */ @@ -890,7 +894,7 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, if (shProg) { if (location < shProg->Uniforms->NumUniforms) { GLint progPos, i; - const struct gl_program *prog; + const struct gl_program *prog = NULL; progPos = shProg->Uniforms->Uniforms[location].VertPos; if (progPos >= 0) { @@ -903,8 +907,11 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, } } - for (i = 0; i < prog->Parameters->Parameters[progPos].Size; i++) { - params[i] = prog->Parameters->ParameterValues[progPos][i]; + ASSERT(prog); + if (prog) { + for (i = 0; i < prog->Parameters->Parameters[progPos].Size; i++) { + params[i] = prog->Parameters->ParameterValues[progPos][i]; + } } } else { @@ -1001,6 +1008,8 @@ _mesa_link_program(GLcontext *ctx, GLuint program) return; } + FLUSH_VERTICES(ctx, _NEW_PROGRAM); + _slang_link(ctx, program, shProg); } -- cgit v1.2.3 From f8dce51083042b4d3fabf48524835290b6725682 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 16 May 2008 16:04:04 -0600 Subject: gallium: ref counting clean-ups --- src/mesa/state_tracker/st_cb_texture.c | 13 +++++-------- 1 file changed, 5 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 866683c23e..cc17bebc6b 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -151,7 +151,7 @@ st_DeleteTextureObject(GLcontext *ctx, { struct st_texture_object *stObj = st_texture_object(texObj); if (stObj->pt) - pipe_texture_release(&stObj->pt); + pipe_texture_reference(&stObj->pt, NULL); _mesa_delete_texture_object(ctx, texObj); } @@ -165,7 +165,7 @@ st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) DBG("%s\n", __FUNCTION__); if (stImage->pt) { - pipe_texture_release(&stImage->pt); + pipe_texture_reference(&stImage->pt, NULL); } if (texImage->Data) { @@ -544,7 +544,7 @@ st_TexImage(GLcontext * ctx, * Release any old malloced memory. */ if (stImage->pt) { - pipe_texture_release(&stImage->pt); + pipe_texture_reference(&stImage->pt, NULL); assert(!texImage->Data); } else if (texImage->Data) { @@ -564,7 +564,7 @@ st_TexImage(GLcontext * ctx, stImage->face, stImage->level)))) { DBG("release it\n"); - pipe_texture_release(&stObj->pt); + pipe_texture_reference(&stObj->pt, NULL); assert(!stObj->pt); stObj->teximage_realloc = FALSE; } @@ -1396,7 +1396,7 @@ copy_image_data_to_texture(struct st_context *st, stImage->face ); - pipe_texture_release(&stImage->pt); + pipe_texture_reference(&stImage->pt, NULL); } else if (stImage->base.Data) { assert(stImage->base.Data != NULL); @@ -1457,9 +1457,6 @@ st_finalize_texture(GLcontext *ctx, firstImage->pt != stObj->pt && firstImage->pt->last_level >= stObj->lastLevel) { - if (stObj->pt) - pipe_texture_release(&stObj->pt); - pipe_texture_reference(&stObj->pt, firstImage->pt); } -- cgit v1.2.3 From 87afc9bcad9a440352c6c3bda234cf089ff7e265 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 16 May 2008 16:04:30 -0600 Subject: gallium: release textures during context tear-down (fix mem leak) --- src/mesa/state_tracker/st_context.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 8db55a179f..fb397ea510 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -171,6 +171,10 @@ static void st_destroy_context_priv( struct st_context *st ) _vbo_DestroyContext(st->ctx); + for (i = 0; i < Elements(st->state.sampler_texture); i++) { + pipe_texture_reference(&st->state.sampler_texture[i], NULL); + } + for (i = 0; i < Elements(st->state.constants); i++) { if (st->state.constants[i].buffer) { pipe_reference_buffer(st->pipe, &st->state.constants[i].buffer, NULL); -- cgit v1.2.3 From b4219e35e2de783ca03e53c311a8a56f8b633aa1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 16 May 2008 16:11:00 -0600 Subject: gallium: fix glBitmap surface memleak in context tear-down --- src/mesa/state_tracker/st_cb_bitmap.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index b17681a05b..d39fce1bb7 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -747,6 +747,11 @@ void st_destroy_bitmap(struct st_context *st) { struct pipe_context *pipe = st->pipe; + struct pipe_screen *screen = pipe->screen; + struct bitmap_cache *cache = st->bitmap.cache; + + screen->surface_unmap(screen, cache->surf); + screen->tex_surface_release(screen, &cache->surf); #if 0 if (st->bitmap.combined_prog) { -- cgit v1.2.3 From 1de0af303acc4fb973bbfea6641ba62a2f24b266 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 17 May 2008 08:46:43 -0600 Subject: s/PIPE_FORMAT_U_S8/PIPE_FORMAT_S8_UNORM/ --- src/mesa/state_tracker/st_cb_drawpixels.c | 4 ++-- src/mesa/state_tracker/st_cb_readpixels.c | 2 +- src/mesa/state_tracker/st_format.c | 4 ++-- 3 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 65bfd6cfcc..56b949cf96 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -770,7 +770,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, } switch (ps->format) { - case PIPE_FORMAT_U_S8: + case PIPE_FORMAT_S8_UNORM: { ubyte *dest = stmap + spanY * ps->pitch + spanX; memcpy(dest, values, spanWidth); @@ -914,7 +914,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, } } break; - case PIPE_FORMAT_U_S8: + case PIPE_FORMAT_S8_UNORM: memcpy(dst, src, width); break; default: diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index ddbe36106c..745755566e 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -87,7 +87,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, /* get stencil values */ switch (ps->format) { - case PIPE_FORMAT_U_S8: + case PIPE_FORMAT_S8_UNORM: { const ubyte *src = stmap + srcY * ps->pitch + x; memcpy(values, src, width); diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 9a385a0457..17a3cfd5a4 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -503,8 +503,8 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_STENCIL_INDEX4_EXT: case GL_STENCIL_INDEX8_EXT: case GL_STENCIL_INDEX16_EXT: - if (screen->is_format_supported( screen, PIPE_FORMAT_U_S8, surfType )) - return PIPE_FORMAT_U_S8; + if (screen->is_format_supported( screen, PIPE_FORMAT_S8_UNORM, surfType )) + return PIPE_FORMAT_S8_UNORM; if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, surfType )) return PIPE_FORMAT_S8Z24_UNORM; if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, surfType )) -- cgit v1.2.3 From 718a2d8c7a125609a8dca813703047e24de09653 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 17 May 2008 08:47:56 -0600 Subject: gallium: remove deprecated format names --- src/gallium/include/pipe/p_format.h | 8 -------- 1 file changed, 8 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index e32212d36c..a4bd23c302 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -327,14 +327,6 @@ enum pipe_format { }; -/** XXX remove these deprecated names */ -#define PIPE_FORMAT_U_L8 PIPE_FORMAT_L8_UNORM -#define PIPE_FORMAT_U_A8 PIPE_FORMAT_A8_UNORM -#define PIPE_FORMAT_U_I8 PIPE_FORMAT_I8_UNORM -#define PIPE_FORMAT_U_A8_L8 PIPE_FORMAT_A8L8_UNORM -#define PIPE_FORMAT_U_S8 PIPE_FORMAT_S8_UNORM - - /** * Builds pipe format name from format token. */ -- cgit v1.2.3 From 9671f7ae476cadb46f9f8f9d0363f24aabaf9f87 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 17 May 2008 10:30:21 -0600 Subject: gallium: in drivers, make copy of tokens passed to pipe->create_vs/fs_state() The caller can then free the token array immediately. --- src/gallium/auxiliary/draw/draw_vs_exec.c | 11 ++-- src/gallium/auxiliary/draw/draw_vs_llvm.c | 7 +- src/gallium/auxiliary/draw/draw_vs_sse.c | 5 +- src/gallium/auxiliary/tgsi/util/tgsi_parse.c | 15 +++++ src/gallium/auxiliary/tgsi/util/tgsi_parse.h | 31 ++++++++- src/gallium/drivers/cell/ppu/cell_state_shader.c | 84 +++++++++++++++--------- src/gallium/drivers/i915simple/i915_state.c | 5 +- src/gallium/drivers/i965simple/brw_state.c | 19 +++--- src/gallium/drivers/softpipe/sp_fs_exec.c | 5 +- src/gallium/drivers/softpipe/sp_state.h | 7 +- src/gallium/drivers/softpipe/sp_state_fs.c | 5 +- 11 files changed, 137 insertions(+), 57 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 54a2b2ab04..7a02f6334b 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -39,6 +39,7 @@ #include "draw_vs.h" #include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_scan.h" struct exec_vertex_shader { @@ -165,21 +166,23 @@ draw_create_vs_exec(struct draw_context *draw, const struct pipe_shader_state *state) { struct exec_vertex_shader *vs = CALLOC_STRUCT( exec_vertex_shader ); - uint nt = tgsi_num_tokens(state->tokens); if (vs == NULL) return NULL; /* we make a private copy of the tokens */ - vs->base.state.tokens = mem_dup(state->tokens, nt * sizeof(state->tokens[0])); - tgsi_scan_shader(state->tokens, &vs->base.info); + vs->base.state.tokens = tgsi_dup_tokens(state->tokens); + if (!vs->base.state.tokens) { + FREE(vs); + return NULL; + } + tgsi_scan_shader(state->tokens, &vs->base.info); vs->base.prepare = vs_exec_prepare; vs->base.run_linear = vs_exec_run_linear; vs->base.delete = vs_exec_delete; vs->machine = &draw->machine; - return &vs->base; } diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index dcada66514..be6907ed04 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -138,14 +138,17 @@ draw_create_vs_llvm(struct draw_context *draw, const struct pipe_shader_state *templ) { struct draw_llvm_vertex_shader *vs; - uint nt = tgsi_num_tokens(templ->tokens); vs = CALLOC_STRUCT( draw_llvm_vertex_shader ); if (vs == NULL) return NULL; /* we make a private copy of the tokens */ - vs->base.state.tokens = mem_dup(templ->tokens, nt * sizeof(templ->tokens[0])); + vs->base.state.tokens = tgsi_dup_tokens(templ->tokens); + if (!vs->base.state.tokens) { + FREE(vs); + return NULL; + } tgsi_scan_shader(vs->base.state.tokens, &vs->base.info); diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index a57c938fbf..5929ea76b2 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -188,7 +188,6 @@ draw_create_vs_sse(struct draw_context *draw, const struct pipe_shader_state *templ) { struct draw_sse_vertex_shader *vs; - uint nt = tgsi_num_tokens(templ->tokens); if (!rtasm_cpu_has_sse2()) return NULL; @@ -198,7 +197,9 @@ draw_create_vs_sse(struct draw_context *draw, return NULL; /* we make a private copy of the tokens */ - vs->base.state.tokens = mem_dup(templ->tokens, nt * sizeof(templ->tokens[0])); + vs->base.state.tokens = tgsi_dup_tokens(templ->tokens); + if (!vs->base.state.tokens) + goto fail; tgsi_scan_shader(templ->tokens, &vs->base.info); diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c index 5bea773840..5c0b0bfd61 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c @@ -330,3 +330,18 @@ tgsi_num_tokens(const struct tgsi_token *tokens) } return 0; } + + +/** + * Make a new copy of a token array. + */ +struct tgsi_token * +tgsi_dup_tokens(const struct tgsi_token *tokens) +{ + unsigned n = tgsi_num_tokens(tokens); + unsigned bytes = n * sizeof(struct tgsi_token); + struct tgsi_token *new_tokens = (struct tgsi_token *) MALLOC(bytes); + if (new_tokens) + memcpy(new_tokens, tokens, bytes); + return new_tokens; +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h index 15e76feb7c..4102101093 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h @@ -1,4 +1,31 @@ -#if !defined TGSI_PARSE_H +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_PARSE_H #define TGSI_PARSE_H #include "pipe/p_shader_tokens.h" @@ -118,6 +145,8 @@ tgsi_parse_token( unsigned tgsi_num_tokens(const struct tgsi_token *tokens); +struct tgsi_token * +tgsi_dup_tokens(const struct tgsi_token *tokens); #if defined __cplusplus } diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c index fb2e940348..c3a3fbd066 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -30,33 +30,50 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "draw/draw_context.h" -#if 0 -#include "pipe/p_shader_tokens.h" -#include "gallivm/gallivm.h" -#include "tgsi/util/tgsi_dump.h" -#include "tgsi/exec/tgsi_sse2.h" -#endif +#include "tgsi/util/tgsi_parse.h" #include "cell_context.h" #include "cell_state.h" + +/** cast wrapper */ +static INLINE struct cell_fragment_shader_state * +cell_fragment_shader_state(void *shader) +{ + return (struct pipe_shader_state *) shader; +} + + +/** cast wrapper */ +static INLINE struct cell_vertex_shader_state * +cell_vertex_shader_state(void *shader) +{ + return (struct pipe_shader_state *) shader; +} + + + static void * cell_create_fs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { /*struct cell_context *cell = cell_context(pipe);*/ - struct cell_fragment_shader_state *state; + struct cell_fragment_shader_state *cfs; - state = CALLOC_STRUCT(cell_fragment_shader_state); - if (!state) + cfs = CALLOC_STRUCT(cell_fragment_shader_state); + if (!cfs) return NULL; - state->shader = *templ; + cfs->shader.tokens = tgsi_dup_tokens(templ->tokens); + if (!cfs->shader.tokens) { + FREE(cfs); + return NULL; + } - tgsi_scan_shader(templ->tokens, &state->info); + tgsi_scan_shader(templ->tokens, &cfs->info); - return state; + return cfs; } @@ -65,7 +82,7 @@ cell_bind_fs_state(struct pipe_context *pipe, void *fs) { struct cell_context *cell = cell_context(pipe); - cell->fs = (struct cell_fragment_shader_state *) fs; + cell->fs = cell_fragment_shader_state(fs); cell->dirty |= CELL_NEW_FS; } @@ -74,10 +91,10 @@ cell_bind_fs_state(struct pipe_context *pipe, void *fs) static void cell_delete_fs_state(struct pipe_context *pipe, void *fs) { - struct cell_fragment_shader_state *state = - (struct cell_fragment_shader_state *) fs; + struct cell_fragment_shader_state *cfs = cell_fragment_shader_state(fs); - FREE( state ); + FREE((void *) cfs->shader.tokens); + FREE(cfs); } @@ -86,22 +103,28 @@ cell_create_vs_state(struct pipe_context *pipe, const struct pipe_shader_state *templ) { struct cell_context *cell = cell_context(pipe); - struct cell_vertex_shader_state *state; + struct cell_vertex_shader_state *cvs; - state = CALLOC_STRUCT(cell_vertex_shader_state); - if (!state) + cvs = CALLOC_STRUCT(cell_vertex_shader_state); + if (!cvs) return NULL; - state->shader = *templ; - tgsi_scan_shader(templ->tokens, &state->info); + cvs->shader.tokens = tgsi_dup_tokens(templ->tokens); + if (!cvs->shader.tokens) { + FREE(cvs); + return NULL; + } - state->draw_data = draw_create_vertex_shader(cell->draw, &state->shader); - if (state->draw_data == NULL) { - FREE( state ); + tgsi_scan_shader(templ->tokens, &cvs->info); + + cvs->draw_data = draw_create_vertex_shader(cell->draw, &cvs->shader); + if (cvs->draw_data == NULL) { + FREE( (void *) cvs->shader.tokens ); + FREE( cvs ); return NULL; } - return state; + return cvs; } @@ -110,7 +133,7 @@ cell_bind_vs_state(struct pipe_context *pipe, void *vs) { struct cell_context *cell = cell_context(pipe); - cell->vs = (const struct cell_vertex_shader_state *) vs; + cell->vs = cell_vertex_shader_state(vs); draw_bind_vertex_shader(cell->draw, (cell->vs ? cell->vs->draw_data : NULL)); @@ -123,12 +146,11 @@ static void cell_delete_vs_state(struct pipe_context *pipe, void *vs) { struct cell_context *cell = cell_context(pipe); + struct cell_vertex_shader_state *cvs = cell_vertex_shader_state(vs); - struct cell_vertex_shader_state *state = - (struct cell_vertex_shader_state *) vs; - - draw_delete_vertex_shader(cell->draw, state->draw_data); - FREE( state ); + draw_delete_vertex_shader(cell->draw, cvs->draw_data); + FREE( (void *) cvs->shader.tokens ); + FREE( cvs ); } diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 3d94b52366..4adeb37e86 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -33,6 +33,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "tgsi/util/tgsi_parse.h" #include "i915_context.h" #include "i915_reg.h" @@ -436,7 +437,7 @@ i915_create_fs_state(struct pipe_context *pipe, if (!ifs) return NULL; - ifs->state = *templ; + ifs->state.tokens = tgsi_dup_tokens(templ->tokens); tgsi_scan_shader(templ->tokens, &ifs->info); @@ -465,6 +466,8 @@ void i915_delete_fs_state(struct pipe_context *pipe, void *shader) FREE(ifs->program); ifs->program_len = 0; + FREE(ifs->state.tokens); + FREE(ifs); } diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 376f1487b2..ac243b7e4f 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -35,6 +35,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" #include "tgsi/util/tgsi_dump.h" +#include "tgsi/util/tgsi_parse.h" #include "brw_context.h" #include "brw_defines.h" @@ -182,9 +183,7 @@ static void * brw_create_fs_state(struct pipe_context *pipe, { struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); - /* XXX: Do I have to duplicate the tokens as well?? - */ - brw_fp->program = *shader; + brw_fp->program.tokens = tgsi_dup_tokens(shader->tokens); brw_fp->id = brw_context(pipe)->program_id++; tgsi_scan_shader(shader->tokens, &brw_fp->info); @@ -210,7 +209,10 @@ static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) { - FREE(shader); + struct brw_fragment_program *brw_fp = (struct brw_fragment_program *) shader; + + FREE((void *) brw_fp->program.tokens); + FREE(brw_fp); } @@ -223,9 +225,7 @@ static void *brw_create_vs_state(struct pipe_context *pipe, { struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); - /* XXX: Do I have to duplicate the tokens as well?? - */ - brw_vp->program = *shader; + brw_vp->program.tokens = tgsi_dup_tokens(shader->tokens); brw_vp->id = brw_context(pipe)->program_id++; tgsi_scan_shader(shader->tokens, &brw_vp->info); @@ -251,7 +251,10 @@ static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) { - FREE(shader); + struct brw_vertex_program *brw_vp = (struct brw_vertex_program *) shader; + + FREE((void *) brw_vp->program.tokens); + FREE(brw_vp); } diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index d5bd7a702f..0b199a2193 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -37,6 +37,7 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_parse.h" struct sp_exec_fragment_shader { struct sp_fragment_shader base; @@ -116,6 +117,7 @@ exec_run( const struct sp_fragment_shader *base, static void exec_delete( struct sp_fragment_shader *base ) { + FREE((void *) base->shader.tokens); FREE(base); } @@ -137,7 +139,8 @@ softpipe_create_fs_exec(struct softpipe_context *softpipe, if (!shader) return NULL; - shader->base.shader = *templ; + /* we need to keep a local copy of the tokens */ + shader->base.shader.tokens = tgsi_dup_tokens(templ->tokens); shader->base.prepare = exec_prepare; shader->base.run = exec_run; shader->base.delete = exec_delete; diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 45056502b8..452e51fa79 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -54,9 +54,11 @@ struct tgsi_sampler; struct tgsi_exec_machine; +struct vertex_info; -/** Subclass of pipe_shader_state (though it doesn't really need to be). +/** + * Subclass of pipe_shader_state (though it doesn't really need to be). * * This is starting to look an awful lot like a quad pipeline stage... */ @@ -80,11 +82,10 @@ struct sp_fragment_shader { void (*delete)( struct sp_fragment_shader * ); }; -struct vertex_info; /** Subclass of pipe_shader_state */ struct sp_vertex_shader { - struct pipe_shader_state shader; + struct pipe_shader_state shader; /* Note: this field not actually used */ struct draw_vertex_shader *draw_data; }; diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 9e77b7e91b..24b91fbc79 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -102,10 +102,7 @@ softpipe_create_vs_state(struct pipe_context *pipe, return NULL; } - state->shader = *templ; - - state->draw_data = draw_create_vertex_shader(softpipe->draw, - &state->shader); + state->draw_data = draw_create_vertex_shader(softpipe->draw, templ); if (state->draw_data == NULL) { FREE( state ); return NULL; -- cgit v1.2.3 From aeae57693b31bf42833a9d51844fe92e3ab61034 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 15 May 2008 12:17:46 -0400 Subject: move the swizzling code to gallivm in preperation for code-generating it also some minor cleanups --- src/gallium/auxiliary/draw/draw_vs_llvm.c | 58 +---- src/gallium/auxiliary/gallivm/gallivm.h | 13 +- src/gallium/auxiliary/gallivm/gallivm_builtins.cpp | 267 ++++++++++----------- src/gallium/auxiliary/gallivm/gallivm_cpu.cpp | 56 ++++- src/gallium/auxiliary/gallivm/llvm_builtins.c | 1 - 5 files changed, 201 insertions(+), 194 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index be6907ed04..171da51dd5 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ /* @@ -72,47 +72,9 @@ vs_llvm_run_linear( struct draw_vertex_shader *base, struct draw_llvm_vertex_shader *shader = (struct draw_llvm_vertex_shader *)base; - struct tgsi_exec_machine *machine = shader->machine; - unsigned int i, j; - unsigned slot; - - - for (i = 0; i < count; i += MAX_TGSI_VERTICES) { - unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); - - /* Swizzle inputs. - */ - for (j = 0; j < max_vertices; j++) { - for (slot = 0; slot < base->info.num_inputs; slot++) { - machine->Inputs[slot].xyzw[0].f[j] = input[slot][0]; - machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; - machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; - machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; - } - - input = (const float (*)[4])((const char *)input + input_stride); - } - - /* run shader */ - gallivm_cpu_vs_exec(shader->llvm_prog, - machine->Inputs, - machine->Outputs, - (float (*)[4]) constants, - machine->Temps); - - - /* Unswizzle all output results - */ - for (j = 0; j < max_vertices; j++) { - for (slot = 0; slot < base->info.num_outputs; slot++) { - output[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; - } - output = (float (*)[4])((char *)output + output_stride); - } - } + gallivm_cpu_vs_exec(shader->llvm_prog, shader->machine, + input, base->info.num_inputs, output, base->info.num_outputs, + constants, count, input_stride, output_stride); } @@ -120,7 +82,7 @@ vs_llvm_run_linear( struct draw_vertex_shader *base, static void vs_llvm_delete( struct draw_vertex_shader *base ) { - struct draw_llvm_vertex_shader *shader = + struct draw_llvm_vertex_shader *shader = (struct draw_llvm_vertex_shader *)base; /* Do something to free compiled shader: @@ -140,7 +102,7 @@ draw_create_vs_llvm(struct draw_context *draw, struct draw_llvm_vertex_shader *vs; vs = CALLOC_STRUCT( draw_llvm_vertex_shader ); - if (vs == NULL) + if (vs == NULL) return NULL; /* we make a private copy of the tokens */ diff --git a/src/gallium/auxiliary/gallivm/gallivm.h b/src/gallium/auxiliary/gallivm/gallivm.h index b4d6555d2f..36a64a7747 100644 --- a/src/gallium/auxiliary/gallivm/gallivm.h +++ b/src/gallium/auxiliary/gallivm/gallivm.h @@ -90,10 +90,15 @@ void gallivm_prog_dump(struct gallivm_prog *prog, const char *file_prefix); struct gallivm_cpu_engine *gallivm_cpu_engine_create(struct gallivm_prog *prog); struct gallivm_cpu_engine *gallivm_global_cpu_engine(); int gallivm_cpu_vs_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - struct tgsi_exec_vector *temps); + struct tgsi_exec_machine *machine, + const float (*input)[4], + unsigned num_inputs, + float (*output)[4], + unsigned num_outputs, + const float (*constants)[4], + unsigned count, + unsigned input_stride, + unsigned output_stride); int gallivm_cpu_fs_exec(struct gallivm_prog *prog, float x, float y, float (*dests)[PIPE_MAX_SHADER_INPUTS][4], diff --git a/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp b/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp index a6f8cd043b..0fc5c4ec5c 100644 --- a/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp +++ b/src/gallium/auxiliary/gallivm/gallivm_builtins.cpp @@ -1,141 +1,140 @@ static const unsigned char llvm_builtins_data[] = { -0x42,0x43,0xc0,0xde,0x21,0x0c,0x00,0x00,0x2b,0x02,0x00,0x00,0x01,0x10,0x00,0x00, +0x42,0x43,0xc0,0xde,0x21,0x0c,0x00,0x00,0x29,0x02,0x00,0x00,0x01,0x10,0x00,0x00, 0x10,0x00,0x00,0x00,0x07,0x81,0x23,0x91,0x41,0xc8,0x04,0x49,0x06,0x10,0x32,0x39, 0x92,0x01,0x84,0x0c,0x25,0x05,0x08,0x19,0x1e,0x04,0x8b,0x62,0x80,0x14,0x45,0x02, 0x42,0x92,0x0b,0x42,0xa4,0x10,0x32,0x14,0x38,0x08,0x18,0x49,0x0a,0x32,0x44,0x24, 0x48,0x0a,0x90,0x21,0x23,0x44,0x72,0x80,0x8c,0x14,0x21,0x86,0x0a,0x8a,0x0a,0x64, 0x0c,0x1f,0x00,0x00,0x49,0x18,0x00,0x00,0x02,0x00,0x00,0x00,0x0b,0x04,0x00,0x0c, -0x00,0x00,0x00,0x00,0x51,0x20,0x00,0x00,0x13,0x00,0x00,0x00,0x32,0x22,0x48,0x09, +0x00,0x00,0x00,0x00,0x51,0x20,0x00,0x00,0x12,0x00,0x00,0x00,0x32,0x22,0x48,0x09, 0x20,0x65,0x82,0x84,0x00,0x26,0x45,0x48,0x05,0x09,0x26,0x45,0xc6,0x05,0x42,0x52, -0x26,0x08,0xb0,0x19,0x80,0x61,0x04,0x02,0x98,0x23,0x00,0x83,0x21,0x80,0x39,0x82, -0x60,0x0a,0x80,0x2e,0xd5,0x61,0x04,0x42,0x20,0x49,0x90,0x22,0x4d,0xa2,0x73,0x04, -0x08,0xb9,0x32,0x00,0x00,0x8a,0x10,0xc2,0x65,0xb8,0x42,0x84,0x10,0x42,0x0d,0x44, -0x11,0x00,0x18,0x01,0x28,0x82,0x08,0x00,0x13,0xa2,0x74,0xb0,0x03,0x3c,0xb0,0x83, -0x36,0x80,0x87,0x71,0x68,0x03,0x76,0x48,0x07,0x77,0xa8,0x07,0x7c,0x68,0x83,0x73, -0x70,0x87,0x7a,0xd8,0x70,0x0f,0xe5,0xd0,0x06,0xf0,0xa0,0x07,0x73,0x20,0x07,0x7a, -0x30,0x07,0x72,0xa0,0x07,0x73,0x20,0x07,0x6d,0x90,0x0e,0x71,0xa0,0x07,0x78,0xa0, -0x07,0x78,0xd0,0x06,0xe9,0x80,0x07,0x7a,0x80,0x07,0x7a,0x80,0x07,0x6d,0x90,0x0e, -0x71,0x60,0x07,0x7a,0x10,0x07,0x76,0xa0,0x07,0x71,0x60,0x07,0x6d,0x90,0x0e,0x73, -0x20,0x07,0x7a,0x30,0x07,0x72,0xa0,0x07,0x73,0x20,0x07,0x6d,0x90,0x0e,0x76,0x40, -0x07,0x7a,0x30,0x07,0x72,0xa0,0x07,0x76,0x40,0x07,0x6d,0x60,0x0e,0x73,0x20,0x07, -0x7a,0x30,0x07,0x72,0xa0,0x07,0x73,0x20,0x07,0x6d,0x60,0x0e,0x76,0x40,0x07,0x7a, -0x30,0x07,0x72,0xa0,0x07,0x76,0x40,0x07,0x6d,0x60,0x0f,0x76,0x40,0x07,0x7a,0x60, -0x07,0x74,0xa0,0x07,0x76,0x40,0x07,0x6d,0x60,0x0f,0x71,0x20,0x07,0x78,0xa0,0x07, -0x71,0x20,0x07,0x78,0xa0,0x07,0x71,0x20,0x07,0x78,0xd0,0x06,0xe1,0x00,0x07,0x7a, -0x00,0x07,0x7a,0x60,0x07,0x74,0xd0,0x06,0xe6,0x80,0x07,0x70,0xa0,0x07,0x71,0x20, -0x07,0x78,0xa0,0x07,0x71,0x20,0x07,0x78,0xa0,0xf3,0x40,0x88,0x04,0x32,0x32,0x02, -0x04,0x60,0x76,0xc6,0xfc,0x6c,0x48,0xa2,0x00,0x40,0x00,0x00,0x00,0x00,0x0c,0x49, -0x14,0x20,0x00,0x00,0x00,0x00,0x80,0x21,0xc9,0x02,0x00,0x01,0x00,0x00,0x00,0x30, -0x24,0x61,0x00,0x20,0x08,0x00,0x00,0x00,0x86,0x24,0x0b,0x00,0x04,0x00,0x00,0x00, -0xc0,0x90,0xa4,0x01,0x02,0x00,0x00,0x00,0x00,0x18,0x92,0x1c,0x40,0x00,0x00,0x00, -0x00,0x00,0x43,0x92,0x05,0x00,0x02,0x00,0x00,0x00,0x60,0x48,0x72,0x00,0x01,0x00, -0x00,0x00,0x00,0x0c,0x49,0x16,0x00,0x08,0x00,0x00,0x00,0x80,0x21,0x89,0x01,0x00, -0x41,0x00,0x00,0x00,0x90,0x05,0x02,0x00,0x10,0x00,0x00,0x00,0x32,0x1e,0x98,0x10, -0x19,0x11,0x4c,0x90,0x8c,0x09,0x26,0x47,0xc6,0x04,0x43,0x92,0x8a,0x59,0x8b,0x43, -0x50,0xd2,0x09,0x02,0x81,0xd2,0x73,0x50,0xc9,0x0c,0x2a,0x99,0x41,0x25,0x33,0xa8, -0x64,0x56,0x28,0x66,0x2d,0x0e,0x41,0xcf,0x2a,0x15,0x04,0x4a,0xcf,0x41,0x25,0x33, -0xa8,0x64,0x06,0x95,0xcc,0xa0,0x92,0x59,0x01,0x00,0x00,0x00,0x53,0x82,0x26,0x0c, -0x04,0x00,0x00,0x00,0x22,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00, -0x05,0x00,0x00,0x00,0x04,0xc6,0x08,0x40,0x10,0x04,0xe1,0x70,0x18,0x23,0x00,0x41, -0x10,0x84,0xc3,0x60,0x04,0x00,0x00,0x00,0x93,0x0c,0xce,0x43,0x4c,0x31,0x3c,0x8e, -0x34,0xc9,0x30,0x41,0xc2,0x14,0x03,0x34,0x51,0x93,0x0c,0x4d,0x44,0x4c,0x31,0x44, -0x8d,0x35,0x56,0x01,0x04,0xc3,0x55,0x21,0x16,0x0e,0x04,0x00,0x0f,0x00,0x00,0x00, -0x46,0x41,0x08,0xcc,0x73,0x9b,0x05,0x21,0x30,0xcf,0x6e,0x18,0x84,0x00,0x2c,0x8b, -0x35,0x04,0x80,0x39,0x04,0x81,0x5d,0x20,0x80,0x0f,0x0c,0x43,0xe4,0xd3,0x36,0x81, -0x04,0x3e,0x30,0x0c,0x91,0x4f,0x5b,0x05,0x12,0xf8,0xc0,0x30,0x44,0x7e,0x7d,0x00, -0x05,0xd1,0x4c,0x11,0x66,0x12,0x83,0xc0,0x3c,0x00,0x00,0x00,0x00,0x00,0x00,0x00, -0x61,0x20,0x00,0x00,0x2a,0x00,0x00,0x00,0x13,0x04,0x43,0x2c,0x10,0x00,0x00,0x00, -0x08,0x00,0x00,0x00,0x24,0x8a,0xa0,0x0c,0x46,0x00,0x4a,0x80,0xc2,0x1c,0x84,0x55, -0x55,0xd6,0x1c,0x84,0x45,0x51,0x16,0x81,0x19,0x80,0x11,0x80,0x31,0x02,0x10,0x04, -0x41,0xfc,0x03,0x00,0x63,0x08,0x0d,0x34,0xc9,0x70,0x55,0xc2,0x2c,0x43,0x20,0x60, -0x73,0x0c,0xd3,0x15,0x8d,0x21,0x34,0xd1,0x18,0x42,0xf3,0x8c,0x55,0x00,0x81,0xa0, -0x6d,0x73,0x0c,0x19,0xe7,0x60,0x87,0x52,0x38,0x10,0x00,0x00,0x13,0x00,0x00,0x00, -0x17,0x60,0x20,0xc5,0x74,0x10,0x8d,0x65,0x14,0x13,0xf3,0xd4,0xb4,0x6d,0x14,0x13, -0xf3,0xd4,0xb8,0x69,0x14,0x13,0xf3,0xd4,0xb6,0x75,0x14,0x13,0xf3,0xd4,0xba,0x35, -0x0c,0x13,0xf3,0x9c,0x80,0xe4,0x36,0x48,0x81,0x10,0xc3,0x4a,0x4c,0x54,0xd4,0x6c, -0x8b,0x23,0x28,0x76,0x41,0x4c,0xcc,0xa3,0x1b,0x07,0x21,0x00,0xcb,0x72,0x00,0x05, -0xd1,0x4c,0x11,0x66,0x18,0x83,0xc0,0x3c,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00, -0x61,0x20,0x00,0x00,0x82,0x00,0x00,0x00,0x13,0x04,0x47,0x2c,0x10,0x00,0x00,0x00, -0x08,0x00,0x00,0x00,0x24,0x46,0x00,0x8a,0xa0,0x0c,0x4a,0xa0,0x14,0x8a,0xa1,0x1c, -0x68,0x8c,0x00,0x10,0x9a,0x83,0x80,0xa8,0x48,0x9a,0x83,0x80,0xa6,0x4a,0x9a,0x83, -0x80,0xa6,0xc8,0x02,0x63,0x08,0x0d,0x64,0xdb,0xc0,0x49,0x06,0xee,0x22,0xc6,0x10, -0x9a,0xc9,0xbc,0x81,0x93,0x0c,0xdf,0x45,0x4c,0x31,0x38,0x4f,0x37,0xcb,0x10,0x08, -0x60,0x30,0xc8,0x10,0x06,0x0e,0x36,0x86,0xd0,0x44,0x36,0x06,0x03,0x27,0x19,0xc8, -0xe0,0x22,0x66,0x19,0x06,0xa2,0x0c,0x06,0x19,0xc2,0xe0,0xc1,0xc6,0x10,0x9a,0xc8, -0xce,0x60,0xe0,0x24,0x03,0x1a,0x5c,0xc4,0x2c,0xc3,0x40,0xa4,0xc1,0x40,0x45,0x20, -0x06,0x81,0x19,0x08,0x83,0x0c,0x6a,0xe0,0x64,0x63,0x08,0x8d,0x64,0x6c,0x30,0x70, -0x92,0xa1,0x0d,0x2e,0x62,0x96,0xa1,0x30,0xdc,0x60,0xa0,0x22,0x10,0x83,0xc0,0x0c, -0x84,0x41,0x86,0x37,0x78,0xb2,0x31,0x84,0x46,0xb2,0x38,0x18,0x38,0xc9,0x20,0x07, -0x17,0x31,0xcb,0x50,0x18,0x73,0x30,0x50,0x11,0xac,0xc1,0x00,0x07,0xc4,0x20,0x03, -0x1d,0x38,0x1a,0xd6,0xc1,0x40,0x45,0xb0,0x06,0x03,0x1c,0x10,0x83,0x0c,0x76,0xf0, -0x68,0x78,0x07,0xe1,0x40,0x00,0x00,0x00,0x4c,0x00,0x00,0x00,0x56,0x62,0x08,0xcc, -0x63,0xef,0x3a,0xa9,0x00,0x19,0x7b,0x73,0x23,0x73,0xf9,0xa1,0x91,0x31,0x98,0x62, -0x62,0x9e,0x7b,0xb7,0x06,0x62,0x62,0x1e,0xda,0x1c,0x88,0x89,0x79,0x6a,0x7b,0x20, -0x26,0xe6,0xb1,0x6d,0x83,0x98,0x98,0xe7,0x36,0x92,0x43,0x70,0x9a,0xca,0xd6,0x73, -0xa3,0x79,0x26,0xe6,0xb9,0x77,0x3f,0x22,0x0c,0x9b,0x21,0x18,0x9f,0xb6,0x90,0x64, -0x62,0x9e,0xda,0x9f,0x98,0xc7,0x36,0x9b,0x67,0x62,0x9e,0x7b,0xf7,0x23,0xc2,0xb0, -0x19,0x82,0xf1,0x6b,0x53,0x79,0x26,0xe6,0xb1,0x6f,0x3f,0x22,0x0c,0x9b,0x21,0x18, -0x9f,0xb6,0x98,0x62,0x62,0x9e,0xbb,0xb7,0x97,0x67,0x62,0x1e,0xfb,0xf6,0x23,0xc2, -0xb0,0x19,0x82,0xf1,0x6b,0x43,0x31,0x04,0xa7,0xa9,0x6c,0xdd,0x66,0x0a,0x81,0x79, -0xf0,0xfa,0x08,0x16,0xc1,0x69,0x06,0x5f,0x70,0x9a,0xe9,0xc6,0x49,0x01,0xc8,0xd8, -0x9b,0x1b,0x99,0xcb,0x4f,0x0c,0x8d,0xad,0x18,0x13,0xf3,0xdc,0x3b,0x6f,0x35,0xc7, -0xc4,0x3c,0x79,0x5d,0xdf,0x06,0x52,0x08,0xcc,0x53,0xdf,0x26,0x62,0x4c,0xcc,0x63, -0xdf,0xb7,0xb9,0x1c,0x02,0xf3,0xe0,0x75,0x5d,0x5b,0xc7,0x20,0x30,0x8f,0x79,0x14, -0x13,0xf3,0xd4,0xf5,0x19,0x2c,0x82,0xd3,0x0c,0xbe,0xe0,0x34,0x13,0xce,0x5b,0x0b, -0x22,0x38,0x4d,0x85,0xd3,0x35,0x6d,0x37,0xc5,0xc4,0x3c,0x79,0x4d,0x1a,0x40,0xc6, -0xde,0xdc,0xc8,0x5c,0x7e,0x64,0x70,0x8c,0x83,0x10,0x9c,0xa6,0xb2,0x94,0x42,0x60, -0x1e,0x7b,0x37,0x19,0x43,0x70,0x9a,0x0a,0xa7,0xcd,0xa4,0x98,0x98,0xc7,0xbe,0x8d, -0xc5,0x98,0x98,0xe7,0xee,0x7b,0x3b,0x29,0x26,0xe6,0xb1,0xf3,0x13,0x58,0x04,0xa7, -0x19,0x7c,0xc1,0x69,0x26,0x9b,0xb6,0x0f,0x43,0x70,0x9a,0xaa,0xb6,0x6d,0xc4,0x98, -0x98,0xc7,0xce,0xf1,0x03,0x28,0x88,0x66,0x8a,0x30,0x00,0x00,0x00,0x00,0x00,0x00, -0x61,0x20,0x00,0x00,0x4a,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00, -0x07,0x00,0x00,0x00,0x24,0xca,0x60,0x04,0xa0,0x04,0x8a,0x80,0xc2,0x0c,0x00,0xbd, -0x61,0x8c,0x04,0x10,0x1e,0xe1,0x19,0xc6,0x48,0x02,0xe1,0x11,0x1e,0x00,0x00,0x00, -0x63,0x08,0xcd,0x63,0xd5,0xc0,0x31,0x84,0x06,0xb2,0x6b,0xe0,0x18,0x42,0x13,0x59, -0x36,0x70,0x0c,0xa1,0x71,0x6c,0x1b,0x38,0x16,0x02,0x04,0xc7,0x64,0x61,0x1a,0x37, -0x16,0x01,0x04,0x48,0x35,0xc7,0x20,0x79,0xcf,0x58,0x04,0x10,0x20,0xd5,0x1c,0xc3, -0x07,0x06,0xd0,0x58,0x04,0x10,0x20,0xd5,0x1c,0x43,0x18,0x88,0x41,0x34,0x16,0x01, -0x04,0x48,0x35,0xc7,0x30,0x06,0x64,0xe0,0x98,0x37,0xd0,0xc0,0x60,0xa0,0x89,0xc1, -0x40,0x23,0x83,0x81,0x63,0x21,0x40,0x70,0x50,0x66,0x70,0x06,0x68,0x90,0x06,0x58, -0x06,0xe1,0x40,0x00,0x25,0x00,0x00,0x00,0x56,0x52,0x4c,0xcc,0x73,0xd3,0x56,0x41, -0x4c,0xcc,0x53,0xdb,0x05,0x31,0x31,0xcf,0x6d,0x19,0xc4,0xc4,0x3c,0xba,0x6d,0x10, -0x13,0xf3,0xf4,0xd6,0x41,0x08,0xc0,0xb2,0x18,0x46,0x21,0x38,0x4d,0x85,0x9b,0x46, -0x21,0x38,0x4d,0xb5,0x9b,0x8a,0x21,0x00,0xcb,0x82,0xdf,0x66,0x62,0x08,0x4e,0x53, -0xdd,0xb7,0x9d,0x18,0x82,0xd3,0x54,0xb7,0x6e,0x28,0x86,0xe0,0x34,0xd5,0xdd,0xdb, -0x47,0x31,0x31,0x4f,0x9d,0x9b,0x87,0x21,0x00,0xcb,0x52,0xdf,0x06,0x62,0x08,0xc0, -0xb2,0xd4,0xbc,0x59,0x10,0x82,0xd3,0x54,0x96,0x62,0x08,0x4e,0x53,0xe1,0xb6,0x85, -0x14,0x13,0xf3,0xd8,0xb4,0x8d,0x14,0x13,0xf3,0xd8,0xb9,0x89,0x18,0x02,0xb0,0x2c, -0xf6,0x6d,0x24,0x86,0x00,0x2c,0x8b,0xcd,0x1b,0x87,0x21,0x38,0x4d,0x55,0xd3,0xd6, -0x30,0x54,0xc0,0x72,0x00,0x05,0xd1,0x4c,0x11,0x06,0x00,0x00,0x00,0x00,0x00,0x00, -0x61,0x20,0x00,0x00,0x19,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00, -0x03,0x00,0x00,0x00,0x24,0x4a,0x60,0x04,0x80,0xc2,0x0c,0x00,0x00,0x00,0x00,0x00, -0x63,0x08,0xcd,0x33,0x16,0x01,0x04,0x48,0x34,0xc7,0x00,0x49,0xcf,0x58,0x04,0x10, -0x28,0xd1,0x1c,0xc3,0x44,0x39,0x58,0x85,0x03,0x01,0x00,0x00,0x0a,0x00,0x00,0x00, -0x16,0x41,0x4c,0xcc,0x63,0xdb,0x04,0x31,0x31,0x4f,0x6e,0x0d,0x43,0x05,0x2c,0x07, -0x50,0x10,0xcd,0x14,0x61,0x56,0x41,0x4c,0xcc,0xd3,0x1b,0x45,0x21,0x00,0xcb,0xb2, -0x9b,0x04,0x21,0x00,0xcb,0x02,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00, -0x1b,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00,0x03,0x00,0x00,0x00, -0x24,0xca,0x60,0x04,0xa0,0x04,0x8a,0x80,0xc2,0x0c,0x00,0x00,0x63,0x08,0xcd,0x33, -0x16,0x01,0x04,0xca,0x34,0xc7,0x20,0x51,0xcf,0x1c,0x43,0x45,0x41,0x73,0x0c,0x16, -0x15,0xcd,0x31,0x5c,0x94,0x83,0x58,0x38,0x10,0x00,0x00,0x00,0x0b,0x00,0x00,0x00, -0x86,0x51,0x4c,0xcc,0x53,0xe7,0x76,0x51,0x4c,0xcc,0x53,0xdb,0x36,0x41,0x4c,0xcc, -0x63,0x5b,0x05,0x31,0x31,0x8f,0x6e,0x0d,0x43,0x05,0x2c,0x66,0x41,0x4c,0xcc,0xd3, -0x1f,0x40,0x41,0x34,0x53,0x84,0x19,0x05,0x21,0x00,0xcb,0x02,0x00,0x00,0x00,0x00, -0x61,0x20,0x00,0x00,0x2f,0x00,0x00,0x00,0x13,0x04,0x45,0x2c,0x10,0x00,0x00,0x00, -0x03,0x00,0x00,0x00,0x24,0xca,0xa0,0x04,0x46,0x00,0x8a,0x80,0xc0,0x08,0x00,0x00, -0x63,0x08,0x0d,0x34,0xc9,0x30,0x49,0xc4,0x2c,0x03,0x11,0x50,0x63,0x08,0xcd,0x33, -0xc9,0x50,0x49,0xc4,0x2c,0x03,0x21,0x58,0x63,0x08,0x4d,0x34,0xc9,0x70,0x49,0xc4, -0x2c,0x03,0x31,0x60,0x63,0x08,0x8d,0x33,0xc9,0x90,0x49,0x84,0x69,0x22,0x70,0xc3, -0x27,0x1c,0x08,0x00,0x1a,0x00,0x00,0x00,0x96,0x51,0x4c,0xcc,0x53,0xdf,0x66,0x41, -0x08,0xcc,0x83,0xdb,0x04,0x31,0x31,0x4f,0x6d,0x15,0xc4,0xc4,0x3c,0xb7,0x61,0x10, -0x02,0xf3,0xf0,0x47,0x20,0xb9,0x0d,0x52,0x20,0xc4,0xb0,0x12,0x13,0x15,0x35,0xdb, -0xe2,0x08,0x8a,0x5d,0x10,0x13,0xf3,0xec,0x37,0x90,0x2c,0x4e,0xf4,0x47,0x87,0x54, -0xd7,0x17,0x70,0x2c,0x4e,0xf4,0x47,0x87,0x74,0x02,0xc8,0xe2,0x44,0x7f,0x74,0x48, -0xb9,0x69,0x14,0x02,0xf3,0xd4,0xb8,0x6d,0x18,0x11,0x31,0x55,0xc0,0x62,0x0d,0x43, -0x05,0x2c,0x07,0x50,0x10,0xcd,0x14,0x61,0x46,0x31,0x08,0xcc,0x03,0x00,0x00,0x00, -0x00,0x00,0x00,0x00,0x71,0x20,0x00,0x00,0x12,0x00,0x00,0x00,0x66,0x40,0x54,0x82, -0x23,0x59,0xc2,0x20,0x09,0x92,0x1d,0x18,0x4f,0x84,0x34,0x53,0x61,0x03,0xc4,0xe3, -0x58,0x85,0x05,0x14,0xbe,0x34,0x45,0xb5,0x21,0x10,0x82,0x23,0x15,0x46,0x30,0x2c, -0xc8,0x64,0x02,0x06,0xf0,0x3c,0x91,0x73,0x19,0x00,0xe1,0x4b,0x53,0x64,0x0a,0x84, -0x84,0x34,0x85,0x31,0x10,0x0a,0xb2,0x3c,0x56,0x30,0x08,0xcc,0x63,0x0b,0x44,0x25, -0x21,0x0d,0x00,0x00,0x00,0x00,0x00,0x00}; +0x26,0x08,0xae,0x19,0x80,0x61,0x04,0x02,0x98,0x23,0x00,0x83,0x29,0x80,0x21,0x00, +0xb2,0x73,0x04,0x01,0x51,0x8a,0xf4,0x08,0x92,0xa4,0x39,0x47,0x80,0x50,0x2b,0x03, +0x00,0xa0,0x08,0x21,0x5c,0x46,0x2b,0x44,0x08,0x21,0xd4,0x40,0x14,0x01,0x80,0x11, +0x80,0x22,0x88,0x00,0x13,0xa2,0x74,0xb0,0x03,0x3c,0xb0,0x83,0x36,0x80,0x87,0x71, +0x68,0x03,0x76,0x48,0x07,0x77,0xa8,0x07,0x7c,0x68,0x83,0x73,0x70,0x87,0x7a,0xd8, +0x70,0x0f,0xe5,0xd0,0x06,0xf0,0xa0,0x07,0x73,0x20,0x07,0x7a,0x30,0x07,0x72,0xa0, +0x07,0x73,0x20,0x07,0x6d,0x90,0x0e,0x71,0xa0,0x07,0x78,0xa0,0x07,0x78,0xd0,0x06, +0xe9,0x80,0x07,0x7a,0x80,0x07,0x7a,0x80,0x07,0x6d,0x90,0x0e,0x71,0x60,0x07,0x7a, +0x10,0x07,0x76,0xa0,0x07,0x71,0x60,0x07,0x6d,0x90,0x0e,0x73,0x20,0x07,0x7a,0x30, +0x07,0x72,0xa0,0x07,0x73,0x20,0x07,0x6d,0x90,0x0e,0x76,0x40,0x07,0x7a,0x30,0x07, +0x72,0xa0,0x07,0x76,0x40,0x07,0x6d,0x60,0x0e,0x73,0x20,0x07,0x7a,0x30,0x07,0x72, +0xa0,0x07,0x73,0x20,0x07,0x6d,0x60,0x0e,0x76,0x40,0x07,0x7a,0x30,0x07,0x72,0xa0, +0x07,0x76,0x40,0x07,0x6d,0x60,0x0f,0x76,0x40,0x07,0x7a,0x60,0x07,0x74,0xa0,0x07, +0x76,0x40,0x07,0x6d,0x60,0x0f,0x71,0x20,0x07,0x78,0xa0,0x07,0x71,0x20,0x07,0x78, +0xa0,0x07,0x71,0x20,0x07,0x78,0xd0,0x06,0xe1,0x00,0x07,0x7a,0x00,0x07,0x7a,0x60, +0x07,0x74,0xd0,0x06,0xe6,0x80,0x07,0x70,0xa0,0x07,0x71,0x20,0x07,0x78,0xa0,0x07, +0x71,0x20,0x07,0x78,0xa0,0xf3,0x40,0x88,0x04,0x32,0x32,0x02,0x04,0x20,0x76,0x46, +0xfc,0x6c,0x48,0x92,0x00,0x40,0x00,0x00,0x00,0x00,0x0c,0x49,0x12,0x20,0x00,0x00, +0x00,0x00,0x80,0x21,0x89,0x02,0x00,0x01,0x00,0x00,0x00,0x30,0x24,0x59,0x00,0x20, +0x08,0x00,0x00,0x00,0x86,0x24,0x0a,0x00,0x04,0x00,0x00,0x00,0xc0,0x90,0x84,0x01, +0x02,0x00,0x00,0x00,0x00,0x18,0x92,0x1c,0x40,0x00,0x00,0x00,0x00,0x00,0x43,0x12, +0x05,0x00,0x02,0x00,0x00,0x00,0x60,0x48,0x72,0x00,0x01,0x00,0x00,0x00,0x00,0x0c, +0x49,0x14,0x00,0x08,0x00,0x00,0x00,0x80,0x21,0x49,0x01,0x00,0x41,0x00,0x00,0x00, +0x90,0x05,0x02,0x00,0x10,0x00,0x00,0x00,0x32,0x1e,0x98,0x10,0x19,0x11,0x4c,0x90, +0x8c,0x09,0x26,0x47,0xc6,0x04,0x43,0x8a,0x8a,0x59,0x8b,0x43,0x50,0xd2,0x09,0x02, +0x81,0xd2,0x73,0x50,0xc9,0x0c,0x2a,0x99,0x41,0x25,0x33,0xa8,0x64,0x56,0x28,0x66, +0x2d,0x0e,0x41,0xcf,0x2a,0x15,0x04,0x4a,0xcf,0x41,0x25,0x33,0xa8,0x64,0x06,0x95, +0xcc,0xa0,0x92,0x59,0x01,0x00,0x00,0x00,0x53,0x82,0x26,0x0c,0x04,0x00,0x00,0x00, +0x22,0x00,0x00,0x00,0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00,0x05,0x00,0x00,0x00, +0x04,0xc6,0x08,0x40,0x10,0x04,0xe1,0x70,0x18,0x23,0x00,0x41,0x10,0x84,0xc3,0x60, +0x04,0x00,0x00,0x00,0x93,0x0c,0xce,0x43,0x4c,0x31,0x3c,0x8e,0x34,0xc9,0x30,0x41, +0xc2,0x14,0x03,0x34,0x51,0x93,0x0c,0x4d,0x44,0x4c,0x31,0x44,0x8d,0x35,0x56,0x01, +0x04,0xc3,0x55,0x21,0x16,0x0e,0x04,0x00,0x0f,0x00,0x00,0x00,0x46,0x41,0x08,0xcc, +0x73,0x9b,0x05,0x21,0x30,0xcf,0x6e,0x18,0x84,0x00,0x2c,0x8b,0x35,0x04,0x80,0x39, +0x04,0x81,0x5d,0x20,0x80,0x0f,0x0c,0x43,0xe4,0xd3,0x36,0x81,0x04,0x3e,0x30,0x0c, +0x91,0x4f,0x5b,0x05,0x12,0xf8,0xc0,0x30,0x44,0x7e,0x7d,0x00,0x05,0xd1,0x4c,0x11, +0x66,0x12,0x83,0xc0,0x3c,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00, +0x2a,0x00,0x00,0x00,0x13,0x04,0x43,0x2c,0x10,0x00,0x00,0x00,0x08,0x00,0x00,0x00, +0x24,0x8a,0xa0,0x0c,0x46,0x00,0x4a,0x80,0xc2,0x1c,0x84,0x55,0x55,0xd6,0x1c,0x84, +0x45,0x51,0x16,0x81,0x19,0x80,0x11,0x80,0x31,0x02,0x10,0x04,0x41,0xfc,0x03,0x00, +0x63,0x08,0x0d,0x34,0xc9,0x70,0x55,0xc2,0x2c,0x43,0x20,0x60,0x73,0x0c,0xd3,0x15, +0x8d,0x21,0x34,0xd1,0x18,0x42,0xf3,0x8c,0x55,0x00,0x81,0xa0,0x6d,0x73,0x0c,0x19, +0xe7,0x60,0x87,0x52,0x38,0x10,0x00,0x00,0x13,0x00,0x00,0x00,0x17,0x60,0x20,0xc5, +0x74,0x10,0x8d,0x65,0x14,0x13,0xf3,0xd4,0xb4,0x6d,0x14,0x13,0xf3,0xd4,0xb8,0x69, +0x14,0x13,0xf3,0xd4,0xb6,0x75,0x14,0x13,0xf3,0xd4,0xba,0x35,0x0c,0x13,0xf3,0x9c, +0x80,0xe4,0x36,0x48,0x81,0x10,0xc3,0x4a,0x4c,0x54,0xd4,0x6c,0x8b,0x23,0x28,0x76, +0x41,0x4c,0xcc,0xa3,0x1b,0x07,0x21,0x00,0xcb,0x72,0x00,0x05,0xd1,0x4c,0x11,0x66, +0x18,0x83,0xc0,0x3c,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00, +0x81,0x00,0x00,0x00,0x13,0x04,0x4d,0x2c,0x10,0x00,0x00,0x00,0x04,0x00,0x00,0x00, +0x24,0xca,0x60,0x04,0xa0,0x04,0x8a,0x80,0xc2,0x0c,0x00,0x91,0x11,0x00,0x00,0x00, +0x63,0x08,0x4d,0x64,0x16,0xc1,0x49,0x86,0xab,0x22,0x66,0x19,0x02,0x01,0x1b,0x43, +0x70,0xa2,0x59,0x82,0x61,0x0c,0xe1,0x89,0x66,0x09,0x86,0x81,0x0a,0x20,0x0b,0x34, +0x61,0x8e,0x81,0xda,0xa2,0x31,0x84,0x46,0xb2,0x8e,0xe0,0x24,0x83,0x57,0x11,0xb3, +0x0c,0x44,0xf1,0x8d,0x21,0x38,0xd2,0x2c,0x81,0x31,0x86,0xf0,0x48,0xb3,0x04,0xc6, +0x40,0x05,0x00,0x06,0x44,0x18,0x14,0x73,0x0c,0x9c,0x18,0x48,0x63,0x08,0xcd,0x64, +0x64,0x40,0x70,0x92,0xa1,0x0c,0x2a,0x62,0x96,0xe1,0x40,0xcc,0x60,0x0c,0xc1,0x99, +0x66,0x09,0x92,0x31,0x84,0x67,0x9a,0x25,0x48,0x06,0x2a,0x80,0x33,0x38,0xd0,0x00, +0x99,0x63,0x18,0x83,0x34,0x98,0xc6,0x10,0x1a,0xc8,0xd6,0x80,0xe0,0x24,0x03,0x1b, +0x54,0xc4,0x2c,0x83,0xb2,0xb4,0xc1,0x18,0x82,0x03,0xcd,0x12,0x30,0x63,0x08,0x0f, +0x34,0x4b,0xc0,0x0c,0x54,0x00,0x6e,0xa0,0xbc,0xc1,0x32,0xc7,0xa0,0x06,0x70,0x00, +0x61,0x1c,0x84,0x03,0x01,0x00,0x00,0x00,0x4e,0x00,0x00,0x00,0x76,0x52,0x4c,0xcc, +0x73,0xd3,0x24,0x05,0x64,0xec,0xcd,0x8d,0xcc,0xe5,0x87,0x46,0xc6,0x50,0x8a,0x89, +0x79,0xee,0xdb,0x54,0x8a,0x89,0x79,0xee,0xdd,0x1a,0x88,0x89,0x79,0x68,0x73,0x20, +0x26,0xe6,0xa9,0xed,0x81,0x98,0x98,0xc7,0x36,0x0b,0x62,0x62,0x9e,0xdb,0x32,0x88, +0x89,0x79,0x72,0xd3,0x20,0x26,0xe6,0xd9,0x8d,0x83,0x98,0x98,0xa7,0xb7,0x95,0x62, +0x62,0x9e,0xbb,0x27,0x2d,0x20,0x63,0x6f,0x6e,0x64,0x2e,0x3a,0x34,0x35,0x56,0x62, +0x08,0x4e,0x53,0xd9,0xba,0xb5,0x14,0x02,0xf3,0xe0,0xf5,0x25,0x2c,0x82,0xd3,0x0c, +0xbe,0xe0,0x34,0xd3,0x8d,0x9b,0x88,0x21,0x38,0xcd,0x60,0xd7,0x24,0x01,0x63,0xec, +0xcd,0x8d,0xcc,0x45,0x87,0x44,0x80,0x8c,0xbd,0xb9,0x91,0xb9,0xfc,0xc4,0xd0,0x90, +0x02,0x8c,0xb1,0x37,0x37,0x32,0x97,0x1f,0x73,0x29,0x26,0xe6,0xc1,0x71,0x7b,0x29, +0x26,0xe6,0xc1,0x77,0xfb,0x28,0x04,0xe6,0xa9,0x6f,0x52,0x01,0x32,0xf6,0xe6,0x46, +0xe6,0xa2,0x13,0x73,0x63,0x18,0x83,0xc0,0x3c,0xb6,0x41,0x08,0x4e,0x33,0x58,0x47, +0x31,0x31,0x4f,0x5d,0x1f,0xc3,0x22,0x38,0xcd,0xe0,0x0b,0x4e,0x33,0xe1,0xbc,0xa5, +0x18,0x82,0xd3,0x0c,0x77,0x6e,0x20,0xc5,0xc4,0x3c,0xb5,0x4e,0x3a,0x40,0xc6,0xde, +0xdc,0xc8,0x5c,0x7e,0x64,0x70,0x2c,0xa4,0x98,0x98,0xa7,0xee,0xed,0x82,0x10,0x9c, +0xa6,0xba,0x81,0x44,0x70,0x9a,0xc1,0x17,0x9c,0x66,0x32,0x93,0x42,0x60,0x1e,0x7b, +0xb7,0x98,0x62,0x62,0x9e,0xbc,0x36,0x16,0x43,0x70,0x9a,0x0a,0xa7,0x6d,0xa4,0x98, +0x98,0xc7,0xbe,0x8d,0xa4,0x98,0x98,0xc7,0xce,0x0d,0xc6,0x10,0x9c,0x66,0xc0,0x7b, +0x12,0x02,0x32,0xf6,0xe6,0x46,0xe6,0xa2,0x33,0x13,0x73,0x06,0x8b,0xe0,0x34,0x83, +0x2f,0x38,0xcd,0x64,0xd3,0xe6,0x61,0x08,0x4e,0x53,0xd5,0xf6,0x01,0x14,0x44,0x33, +0x45,0x18,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00,0x4a,0x00,0x00,0x00, +0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00,0x07,0x00,0x00,0x00,0x24,0xca,0x60,0x04, +0xa0,0x04,0x8a,0x80,0xc2,0x0c,0x00,0xb9,0x61,0x0c,0x04,0x10,0x1e,0xe1,0x19,0xc6, +0x40,0x02,0xe1,0x11,0x1e,0x00,0x00,0x00,0x63,0x08,0xcd,0x63,0x15,0xc1,0x31,0x84, +0x06,0xb2,0x8b,0xe0,0x18,0x42,0x13,0x59,0x46,0x70,0x0c,0xa1,0x71,0x6c,0x23,0x38, +0x16,0x02,0x04,0xc7,0x64,0x61,0x1a,0x37,0x16,0x01,0x04,0x48,0x35,0xc7,0x20,0x79, +0xcf,0x58,0x04,0x10,0x20,0xd5,0x1c,0xc3,0x07,0x06,0xd0,0x58,0x04,0x10,0x20,0xd5, +0x1c,0x43,0x18,0x88,0x41,0x34,0x16,0x01,0x04,0x48,0x35,0xc7,0x30,0x06,0x64,0xe0, +0x98,0x47,0xd0,0xc0,0x80,0xa0,0x89,0x01,0x41,0x23,0x03,0x82,0x63,0x21,0x40,0x70, +0x50,0x66,0x70,0x06,0x68,0x90,0x06,0x58,0x06,0xe1,0x40,0x00,0x25,0x00,0x00,0x00, +0x56,0x52,0x4c,0xcc,0x73,0xd3,0x56,0x41,0x4c,0xcc,0x53,0xdb,0x05,0x31,0x31,0xcf, +0x6d,0x19,0xc4,0xc4,0x3c,0xba,0x6d,0x10,0x13,0xf3,0xf4,0xd6,0x41,0x08,0xc0,0xb2, +0x18,0x46,0x21,0x38,0x4d,0x85,0x9b,0x46,0x21,0x38,0x4d,0xb5,0x9b,0x8a,0x21,0x00, +0xcb,0x82,0xdf,0x66,0x62,0x08,0x4e,0x53,0xdd,0xb7,0x9d,0x18,0x82,0xd3,0x54,0xb7, +0x6e,0x28,0x86,0xe0,0x34,0xd5,0xdd,0xdb,0x47,0x31,0x31,0x4f,0x9d,0x9b,0x87,0x21, +0x00,0xcb,0x52,0xdf,0x06,0x62,0x08,0xc0,0xb2,0xd4,0xbc,0x59,0x10,0x82,0xd3,0x54, +0x96,0x62,0x08,0x4e,0x53,0xe1,0xb6,0x85,0x14,0x13,0xf3,0xd8,0xb4,0x8d,0x14,0x13, +0xf3,0xd8,0xb9,0x89,0x18,0x02,0xb0,0x2c,0xf6,0x6d,0x24,0x86,0x00,0x2c,0x8b,0xcd, +0x1b,0x87,0x21,0x38,0x4d,0x55,0xd3,0xd6,0x30,0x54,0xc0,0x72,0x00,0x05,0xd1,0x4c, +0x11,0x06,0x00,0x00,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00,0x19,0x00,0x00,0x00, +0x13,0x04,0x41,0x2c,0x10,0x00,0x00,0x00,0x03,0x00,0x00,0x00,0x24,0x4a,0x60,0x04, +0x80,0xc2,0x0c,0x00,0x00,0x00,0x00,0x00,0x63,0x08,0xcd,0x33,0x16,0x01,0x04,0x48, +0x34,0xc7,0x00,0x49,0xcf,0x58,0x04,0x10,0x28,0xd1,0x1c,0xc3,0x44,0x39,0x58,0x85, +0x03,0x01,0x00,0x00,0x0a,0x00,0x00,0x00,0x16,0x41,0x4c,0xcc,0x63,0xdb,0x04,0x31, +0x31,0x4f,0x6e,0x0d,0x43,0x05,0x2c,0x07,0x50,0x10,0xcd,0x14,0x61,0x56,0x41,0x4c, +0xcc,0xd3,0x1b,0x45,0x21,0x00,0xcb,0xb2,0x9b,0x04,0x21,0x00,0xcb,0x02,0x00,0x00, +0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00,0x1b,0x00,0x00,0x00,0x13,0x04,0x41,0x2c, +0x10,0x00,0x00,0x00,0x03,0x00,0x00,0x00,0x24,0xca,0x60,0x04,0xa0,0x04,0x8a,0x80, +0xc2,0x0c,0x00,0x00,0x63,0x08,0xcd,0x33,0x16,0x01,0x04,0xca,0x34,0xc7,0x20,0x51, +0xcf,0x1c,0x43,0x45,0x41,0x73,0x0c,0x16,0x15,0xcd,0x31,0x5c,0x94,0x83,0x58,0x38, +0x10,0x00,0x00,0x00,0x0b,0x00,0x00,0x00,0x86,0x51,0x4c,0xcc,0x53,0xe7,0x76,0x51, +0x4c,0xcc,0x53,0xdb,0x36,0x41,0x4c,0xcc,0x63,0x5b,0x05,0x31,0x31,0x8f,0x6e,0x0d, +0x43,0x05,0x2c,0x66,0x41,0x4c,0xcc,0xd3,0x1f,0x40,0x41,0x34,0x53,0x84,0x19,0x05, +0x21,0x00,0xcb,0x02,0x00,0x00,0x00,0x00,0x61,0x20,0x00,0x00,0x2f,0x00,0x00,0x00, +0x13,0x04,0x45,0x2c,0x10,0x00,0x00,0x00,0x03,0x00,0x00,0x00,0x24,0xca,0xa0,0x04, +0x46,0x00,0x8a,0x80,0xc0,0x08,0x00,0x00,0x63,0x08,0x0d,0x34,0xc9,0x30,0x49,0xc4, +0x2c,0x03,0x11,0x50,0x63,0x08,0xcd,0x33,0xc9,0x50,0x49,0xc4,0x2c,0x03,0x21,0x58, +0x63,0x08,0x4d,0x34,0xc9,0x70,0x49,0xc4,0x2c,0x03,0x31,0x60,0x63,0x08,0x8d,0x33, +0xc9,0x90,0x49,0x84,0x69,0x22,0x70,0xc3,0x27,0x1c,0x08,0x00,0x1a,0x00,0x00,0x00, +0x96,0x51,0x4c,0xcc,0x53,0xdf,0x66,0x41,0x08,0xcc,0x83,0xdb,0x04,0x31,0x31,0x4f, +0x6d,0x15,0xc4,0xc4,0x3c,0xb7,0x61,0x10,0x02,0xf3,0xf0,0x47,0x20,0xb9,0x0d,0x52, +0x20,0xc4,0xb0,0x12,0x13,0x15,0x35,0xdb,0xe2,0x08,0x8a,0x5d,0x10,0x13,0xf3,0xec, +0x37,0x90,0x2c,0x4e,0xf4,0x47,0x87,0x54,0xd7,0x17,0x70,0x2c,0x4e,0xf4,0x47,0x87, +0x74,0x02,0xc8,0xe2,0x44,0x7f,0x74,0x48,0xb9,0x69,0x14,0x02,0xf3,0xd4,0xb8,0x6d, +0x18,0x11,0x31,0x55,0xc0,0x62,0x0d,0x43,0x05,0x2c,0x07,0x50,0x10,0xcd,0x14,0x61, +0x46,0x31,0x08,0xcc,0x03,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x71,0x20,0x00,0x00, +0x12,0x00,0x00,0x00,0x66,0x40,0x54,0x82,0x23,0x59,0xc2,0x20,0x09,0x92,0x1d,0x18, +0x4f,0x84,0x34,0x53,0x61,0x03,0xc4,0xe3,0x58,0x85,0x05,0x14,0xbe,0x34,0x45,0xb5, +0x21,0x10,0x82,0x23,0x15,0x46,0x30,0x2c,0xc8,0x64,0x02,0x06,0xf0,0x3c,0x91,0x73, +0x19,0x00,0xe1,0x4b,0x53,0x64,0x0a,0x84,0x84,0x34,0x85,0x31,0x10,0x0a,0xb2,0x3c, +0x56,0x30,0x08,0xcc,0x63,0x0b,0x44,0x25,0x21,0x0d,0x00,0x00,0x00,0x00,0x00,0x00}; diff --git a/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp index 8f9830d0b1..857c190f7b 100644 --- a/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp +++ b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp @@ -41,6 +41,7 @@ #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" +#include "pipe/p_util.h" #include "tgsi/exec/tgsi_exec.h" #include "tgsi/util/tgsi_dump.h" @@ -179,22 +180,63 @@ typedef void (*vertex_shader_runner)(void *ainputs, float (*aconsts)[4], void *temps); - +#define MAX_TGSI_VERTICES 4 /*! This function is used to execute the gallivm_prog in software. Before calling this function the gallivm_prog has to be JIT compiled with the gallivm_cpu_jit_compile function. */ int gallivm_cpu_vs_exec(struct gallivm_prog *prog, - struct tgsi_exec_vector *inputs, - struct tgsi_exec_vector *dests, - float (*consts)[4], - struct tgsi_exec_vector *temps) + struct tgsi_exec_machine *machine, + const float (*input)[4], + unsigned num_inputs, + float (*output)[4], + unsigned num_outputs, + const float (*constants)[4], + unsigned count, + unsigned input_stride, + unsigned output_stride ) { + unsigned int i, j; + unsigned slot; vertex_shader_runner runner = reinterpret_cast(prog->function); + assert(runner); - /*FIXME*/ - runner(inputs, dests, consts, temps); + + for (i = 0; i < count; i += MAX_TGSI_VERTICES) { + unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); + + /* Swizzle inputs. + */ + for (j = 0; j < max_vertices; j++) { + for (slot = 0; slot < num_inputs; slot++) { + machine->Inputs[slot].xyzw[0].f[j] = input[slot][0]; + machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; + machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; + machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; + } + + input = (const float (*)[4])((const char *)input + input_stride); + } + + /* run shader */ + runner(machine->Inputs, + machine->Outputs, + (float (*)[4]) constants, + machine->Temps); + + /* Unswizzle all output results + */ + for (j = 0; j < max_vertices; j++) { + for (slot = 0; slot < num_outputs; slot++) { + output[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; + output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; + output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; + output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; + } + output = (float (*)[4])((char *)output + output_stride); + } + } return 0; } diff --git a/src/gallium/auxiliary/gallivm/llvm_builtins.c b/src/gallium/auxiliary/gallivm/llvm_builtins.c index 64b5d499a8..6b9d626ed4 100644 --- a/src/gallium/auxiliary/gallivm/llvm_builtins.c +++ b/src/gallium/auxiliary/gallivm/llvm_builtins.c @@ -1,4 +1,3 @@ -/*clang --emit-llvm llvm_builtins.c |llvm-as|opt -std-compile-opts|llvm2cpp -gen-contents -o=gallivm_builtins.cpp -f -for=shader -funcname=createGallivmBuiltins*/ /************************************************************************** * * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. -- cgit v1.2.3 From 735752e8dceeec9b202147d1d19ef3dc70e08673 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 15 May 2008 14:11:19 -0400 Subject: fix injections of functions from builtins into shaders --- src/gallium/auxiliary/gallivm/gallivm.cpp | 3 --- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 18 +++++++++++------- 2 files changed, 11 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/gallivm.cpp b/src/gallium/auxiliary/gallivm/gallivm.cpp index b6f641a3f8..48a3b18cdc 100644 --- a/src/gallium/auxiliary/gallivm/gallivm.cpp +++ b/src/gallium/auxiliary/gallivm/gallivm.cpp @@ -288,10 +288,7 @@ void gallivm_ir_fill_from_tgsi(struct gallivm_ir *ir, std::cout << "Creating llvm from: " <module = mod; gallivm_ir_dump(ir, 0); } diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index f0122802db..17e876e32d 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -38,7 +38,6 @@ #include #include #include -//#include #include #include @@ -193,9 +192,13 @@ llvm::Function * InstructionsSoa::function(int op) std::string name = m_functionsMap[op]; + std::cout <<"For op = "< deps = m_builtinDependencies[name]; for (unsigned int i = 0; i < deps.size(); ++i) { - injectFunction(m_builtins->getFunction(deps[i])); + llvm::Function *func = m_builtins->getFunction(deps[i]); + std::cout <<"\tinjecting dep = '"<getName()<<"'"<getFunction(name); @@ -216,8 +219,10 @@ void InstructionsSoa::createBuiltins() { MemoryBuffer *buffer = MemoryBuffer::getMemBuffer( (const char*)&soabuiltins_data[0], - (const char*)&soabuiltins_data[Elements(soabuiltins_data)-1]); + (const char*)&soabuiltins_data[Elements(soabuiltins_data)]); m_builtins = ParseBitcodeFile(buffer); + std::cout<<"Builtins created at "<isDeclaration()) { - std::cout << "function decleration" <getFunctionType(), GlobalValue::ExternalLinkage, originalFunc->getName(), currentModule()); func->setCallingConv(CallingConv::C); @@ -398,17 +402,17 @@ void InstructionsSoa::injectFunction(llvm::Function *originalFunc, int op) } else { DenseMap val; val[m_builtins->getFunction("powf")] = currentModule()->getFunction("powf"); + func = CloneFunction(originalFunc, val); +#if 0 std::cout <<" replacing "<getFunction("powf") <<", with " <getFunction("powf")<getFunctionList().push_back(func); - std::cout << "Func parent is "<getParent() - <<", cur is "< Date: Thu, 15 May 2008 17:46:20 -0400 Subject: llvm: implement sub and abs --- src/gallium/auxiliary/gallivm/gallivm.cpp | 6 ++-- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 35 +++++++++++++++++++-- src/gallium/auxiliary/gallivm/instructionssoa.h | 3 ++ src/gallium/auxiliary/gallivm/soabuiltins.c | 38 ++++++++++++++++------- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 2 ++ 5 files changed, 67 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/gallivm.cpp b/src/gallium/auxiliary/gallivm/gallivm.cpp index 48a3b18cdc..77900e342b 100644 --- a/src/gallium/auxiliary/gallivm/gallivm.cpp +++ b/src/gallium/auxiliary/gallivm/gallivm.cpp @@ -303,11 +303,11 @@ struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) { struct gallivm_prog *prog = (struct gallivm_prog *)calloc(1, sizeof(struct gallivm_prog)); - + std::cout << "Before optimizations:"<module->dump(); std::cout<<"-------------------------------"<module); @@ -315,7 +315,7 @@ struct gallivm_prog * gallivm_ir_compile(struct gallivm_ir *ir) prog->num_consts = ir->num_consts; memcpy(prog->interpolators, ir->interpolators, sizeof(prog->interpolators)); prog->num_interp = ir->num_interp; - + /* Run optimization passes over it */ PassManager passes; passes.add(new TargetData(mod)); diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 17e876e32d..4520559ba2 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -173,6 +173,7 @@ std::vector InstructionsSoa::extractVector(llvm::Value *vector) void InstructionsSoa::createFunctionMap() { + m_functionsMap[TGSI_OPCODE_ABS] = "abs"; m_functionsMap[TGSI_OPCODE_DP3] = "dp3"; m_functionsMap[TGSI_OPCODE_DP4] = "dp4"; m_functionsMap[TGSI_OPCODE_POWER] = "pow"; @@ -180,9 +181,16 @@ void InstructionsSoa::createFunctionMap() void InstructionsSoa::createDependencies() { - std::vector powDeps(1); - powDeps[0] = "powf"; - m_builtinDependencies["pow"] = powDeps; + { + std::vector powDeps(1); + powDeps[0] = "powf"; + m_builtinDependencies["pow"] = powDeps; + } + { + std::vector absDeps(1); + absDeps[0] = "fabsf"; + m_builtinDependencies["abs"] = absDeps; + } } llvm::Function * InstructionsSoa::function(int op) @@ -226,6 +234,13 @@ void InstructionsSoa::createBuiltins() createDependencies(); } + +std::vector InstructionsSoa::abs(const std::vector in1) +{ + llvm::Function *func = function(TGSI_OPCODE_ABS); + return callBuiltin(func, in1); +} + std::vector InstructionsSoa::dp3(const std::vector in1, const std::vector in2) { @@ -418,3 +433,17 @@ void InstructionsSoa::injectFunction(llvm::Function *originalFunc, int op) m_functions[op] = func; } } + +std::vector InstructionsSoa::sub(const std::vector in1, + const std::vector in2) +{ + std::vector res(4); + + res[0] = m_builder.CreateSub(in1[0], in2[0], name("subx")); + res[1] = m_builder.CreateSub(in1[1], in2[1], name("suby")); + res[2] = m_builder.CreateSub(in1[2], in2[2], name("subz")); + res[3] = m_builder.CreateSub(in1[3], in2[3], name("subw")); + + return res; +} + diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h index 060ee72f2e..02e5fab51f 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.h +++ b/src/gallium/auxiliary/gallivm/instructionssoa.h @@ -48,6 +48,7 @@ public: InstructionsSoa(llvm::Module *mod, llvm::Function *func, llvm::BasicBlock *block, StorageSoa *storage); + std::vector abs(const std::vector in1); std::vector arl(const std::vector in); std::vector add(const std::vector in1, const std::vector in2); @@ -62,6 +63,8 @@ public: const std::vector in2); std::vector pow(const std::vector in1, const std::vector in2); + std::vector sub(const std::vector in1, + const std::vector in2); void end(); std::vector extractVector(llvm::Value *vector); diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index 40addebd8c..f04e4c974d 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -33,6 +33,33 @@ */ typedef __attribute__(( ext_vector_type(4) )) float float4; + +extern float fabsf(float val); + +void abs(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w) +{ + res[0].x = fabsf(tmp0x.x); + res[0].y = fabsf(tmp0x.y); + res[0].z = fabsf(tmp0x.z); + res[0].w = fabsf(tmp0x.w); + + res[1].x = fabsf(tmp0y.x); + res[1].y = fabsf(tmp0y.y); + res[1].z = fabsf(tmp0y.z); + res[1].w = fabsf(tmp0y.w); + + res[2].x = fabsf(tmp0z.x); + res[2].y = fabsf(tmp0z.y); + res[2].z = fabsf(tmp0z.z); + res[2].w = fabsf(tmp0z.w); + + res[3].x = fabsf(tmp0w.x); + res[3].y = fabsf(tmp0w.y); + res[3].z = fabsf(tmp0w.z); + res[3].w = fabsf(tmp0w.w); +} + void dp3(float4 *res, float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) @@ -77,14 +104,3 @@ void pow(float4 *res, res[2] = p; res[3] = p; } - -#if 0 -void yo(float4 *out, float4 *in) -{ - float4 res[4]; - - dp3(res, in[0], in[1], in[2], in[3], - in[4], in[5], in[6], in[7]); - out[1] = res[1]; -} -#endif diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index ab8c851f14..5465d3a95e 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -740,6 +740,7 @@ translate_instructionir(llvm::Module *module, } break; case TGSI_OPCODE_SUB: { + out = instr->sub(inputs[0], inputs[1]); } break; case TGSI_OPCODE_LERP: { @@ -781,6 +782,7 @@ translate_instructionir(llvm::Module *module, case TGSI_OPCODE_MULTIPLYMATRIX: break; case TGSI_OPCODE_ABS: { + out = instr->abs(inputs[0]); } break; case TGSI_OPCODE_RCC: -- cgit v1.2.3 From ea1a607292ef31df70cda8c6476755e0224c9f7d Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 16 May 2008 14:54:40 -0400 Subject: implement min/max and abstract ops on vectors --- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 30 ++++++++- src/gallium/auxiliary/gallivm/instructionssoa.h | 4 ++ src/gallium/auxiliary/gallivm/soabuiltins.c | 75 ++++++++++++++++++++--- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 2 + 4 files changed, 100 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 4520559ba2..55fdda2791 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -176,14 +176,17 @@ void InstructionsSoa::createFunctionMap() m_functionsMap[TGSI_OPCODE_ABS] = "abs"; m_functionsMap[TGSI_OPCODE_DP3] = "dp3"; m_functionsMap[TGSI_OPCODE_DP4] = "dp4"; + m_functionsMap[TGSI_OPCODE_MIN] = "min"; + m_functionsMap[TGSI_OPCODE_MAX] = "max"; m_functionsMap[TGSI_OPCODE_POWER] = "pow"; } void InstructionsSoa::createDependencies() { { - std::vector powDeps(1); + std::vector powDeps(2); powDeps[0] = "powf"; + powDeps[1] = "powvec"; m_builtinDependencies["pow"] = powDeps; } { @@ -191,6 +194,16 @@ void InstructionsSoa::createDependencies() absDeps[0] = "fabsf"; m_builtinDependencies["abs"] = absDeps; } + { + std::vector maxDeps(1); + maxDeps[0] = "maxvec"; + m_builtinDependencies["max"] = maxDeps; + } + { + std::vector minDeps(1); + minDeps[0] = "minvec"; + m_builtinDependencies["min"] = minDeps; + } } llvm::Function * InstructionsSoa::function(int op) @@ -374,6 +387,21 @@ std::vector InstructionsSoa::pow(const std::vector i return callBuiltin(func, in1, in2); } +std::vector InstructionsSoa::min(const std::vector in1, + const std::vector in2) +{ + llvm::Function *func = function(TGSI_OPCODE_MIN); + return callBuiltin(func, in1, in2); +} + + +std::vector InstructionsSoa::max(const std::vector in1, + const std::vector in2) +{ + llvm::Function *func = function(TGSI_OPCODE_MAX); + return callBuiltin(func, in1, in2); +} + void checkFunction(Function *func) { for (Function::const_iterator BI = func->begin(), BE = func->end(); diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h index 02e5fab51f..5a7f8fdf72 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.h +++ b/src/gallium/auxiliary/gallivm/instructionssoa.h @@ -59,6 +59,10 @@ public: std::vector madd(const std::vector in1, const std::vector in2, const std::vector in3); + std::vector max(const std::vector in1, + const std::vector in2); + std::vector min(const std::vector in1, + const std::vector in2); std::vector mul(const std::vector in1, const std::vector in2); std::vector pow(const std::vector in1, diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index f04e4c974d..935283f962 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -89,18 +89,73 @@ void dp4(float4 *res, extern float powf(float num, float p); +float4 powvec(float4 vec, float4 q) +{ + float4 p; + p.x = powf(vec.x, q.x); + p.y = powf(vec.y, q.y); + p.z = powf(vec.z, q.z); + p.w = powf(vec.w, q.w); + return p; +} + void pow(float4 *res, float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) { - float4 p; - p.x = powf(tmp0x.x, tmp1x.x); - p.y = powf(tmp0x.y, tmp1x.y); - p.z = powf(tmp0x.z, tmp1x.z); - p.w = powf(tmp0x.w, tmp1x.w); - - res[0] = p; - res[1] = p; - res[2] = p; - res[3] = p; + res[0] = powvec(tmp0x, tmp1x); + res[1] = res[0]; + res[2] = res[0]; + res[3] = res[0]; +} + +void lit(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, + float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) +{ + const float4 zerovec = (float4) {0, 0, 0, 0}; + float4 tmpx = tmp0x; + float4 tmpy = tmp0y; + + res[0] = (float4){1.0, 1.0, 1.0, 1.0}; + res[1] = tmpx; + res[2] = tmpy; + res[3] = (float4){1.0, 1.0, 1.0, 1.0}; +} + +float4 minvec(float4 a, float4 b) +{ + return (float4){(a.x < b.x) ? a.x : b.x, + (a.y < b.y) ? a.y : b.y, + (a.z < b.z) ? a.z : b.z, + (a.w < b.w) ? a.w : b.w}; +} + +void min(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, + float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) +{ + res[0] = minvec(tmp0x, tmp1x); + res[1] = minvec(tmp0y, tmp1y); + res[2] = minvec(tmp0z, tmp1z); + res[3] = minvec(tmp0w, tmp1w); +} + + +float4 maxvec(float4 a, float4 b) +{ + return (float4){(a.x > b.x) ? a.x : b.x, + (a.y > b.y) ? a.y : b.y, + (a.z > b.z) ? a.z : b.z, + (a.w > b.w) ? a.w : b.w}; +} + +void max(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, + float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) +{ + res[0] = maxvec(tmp0x, tmp1x); + res[1] = maxvec(tmp0y, tmp1y); + res[2] = maxvec(tmp0z, tmp1z); + res[3] = maxvec(tmp0w, tmp1w); } diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index 5465d3a95e..007b5c169a 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -724,9 +724,11 @@ translate_instructionir(llvm::Module *module, } break; case TGSI_OPCODE_MIN: { + out = instr->min(inputs[0], inputs[1]); } break; case TGSI_OPCODE_MAX: { + out = instr->max(inputs[0], inputs[1]); } break; case TGSI_OPCODE_SLT: { -- cgit v1.2.3 From 1d1cf8edf6a0409caf9aa7d44e186eb51f51fa1f Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 16 May 2008 16:06:59 -0400 Subject: do the lit (some artifacts present) --- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 15 +++++++++ src/gallium/auxiliary/gallivm/instructionssoa.h | 1 + src/gallium/auxiliary/gallivm/soabuiltins.c | 37 ++++++++++++++--------- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 1 + 4 files changed, 40 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 55fdda2791..074dd0ecd6 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -179,6 +179,7 @@ void InstructionsSoa::createFunctionMap() m_functionsMap[TGSI_OPCODE_MIN] = "min"; m_functionsMap[TGSI_OPCODE_MAX] = "max"; m_functionsMap[TGSI_OPCODE_POWER] = "pow"; + m_functionsMap[TGSI_OPCODE_LIT] = "lit"; } void InstructionsSoa::createDependencies() @@ -204,6 +205,14 @@ void InstructionsSoa::createDependencies() minDeps[0] = "minvec"; m_builtinDependencies["min"] = minDeps; } + { + std::vector litDeps(4); + litDeps[0] = "minvec"; + litDeps[1] = "maxvec"; + litDeps[2] = "powf"; + litDeps[3] = "powvec"; + m_builtinDependencies["lit"] = litDeps; + } } llvm::Function * InstructionsSoa::function(int op) @@ -475,3 +484,9 @@ std::vector InstructionsSoa::sub(const std::vector i return res; } +std::vector InstructionsSoa::lit(const std::vector in) +{ + llvm::Function *func = function(TGSI_OPCODE_LIT); + return callBuiltin(func, in); +} + diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h index 5a7f8fdf72..477ef4a157 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.h +++ b/src/gallium/auxiliary/gallivm/instructionssoa.h @@ -56,6 +56,7 @@ public: const std::vector in2); std::vector dp4(const std::vector in1, const std::vector in2); + std::vector lit(const std::vector in); std::vector madd(const std::vector in1, const std::vector in2, const std::vector in3); diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index 935283f962..b3bfebfe50 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -109,20 +109,6 @@ void pow(float4 *res, res[3] = res[0]; } -void lit(float4 *res, - float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w, - float4 tmp1x, float4 tmp1y, float4 tmp1z, float4 tmp1w) -{ - const float4 zerovec = (float4) {0, 0, 0, 0}; - float4 tmpx = tmp0x; - float4 tmpy = tmp0y; - - res[0] = (float4){1.0, 1.0, 1.0, 1.0}; - res[1] = tmpx; - res[2] = tmpy; - res[3] = (float4){1.0, 1.0, 1.0, 1.0}; -} - float4 minvec(float4 a, float4 b) { return (float4){(a.x < b.x) ? a.x : b.x, @@ -159,3 +145,26 @@ void max(float4 *res, res[2] = maxvec(tmp0z, tmp1z); res[3] = maxvec(tmp0w, tmp1w); } + + +void lit(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w) +{ + const float4 zerovec = (float4) {0, 0, 0, 0}; + const float4 min128 = (float4) {-128.f, -128.f, -128.f, -128.f}; + const float4 plus128 = (float4) {128.f, 128.f, 128.f, 128.f}; + + res[0] = (float4){1.0, 1.0, 1.0, 1.0}; + if (tmp0x.x > 0) { + float4 tmpx = maxvec(tmpx, zerovec); + float4 tmpy = maxvec(tmp0y, zerovec); + float4 tmpw = minvec(tmp0w, plus128); + tmpw = maxvec(tmpw, min128); + res[1] = tmpx; + res[2] = powvec(tmpy, tmpw); + } else { + res[1] = zerovec; + res[2] = zerovec; + } + res[3] = (float4){1.0, 1.0, 1.0, 1.0}; +} diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index 007b5c169a..abcb240f46 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -692,6 +692,7 @@ translate_instructionir(llvm::Module *module, } break; case TGSI_OPCODE_LIT: { + out = instr->lit(inputs[0]); } break; case TGSI_OPCODE_RCP: { -- cgit v1.2.3 From 02e45b2dadd42c38247cb992a07eb520ac86519b Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 16 May 2008 17:10:52 -0400 Subject: fix abs and start on rsq --- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 20 +++++++- src/gallium/auxiliary/gallivm/instructionssoa.h | 1 + src/gallium/auxiliary/gallivm/soabuiltins.c | 56 +++++++++++++++-------- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 1 + 4 files changed, 58 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 074dd0ecd6..76049ade7c 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -180,6 +180,7 @@ void InstructionsSoa::createFunctionMap() m_functionsMap[TGSI_OPCODE_MAX] = "max"; m_functionsMap[TGSI_OPCODE_POWER] = "pow"; m_functionsMap[TGSI_OPCODE_LIT] = "lit"; + m_functionsMap[TGSI_OPCODE_RSQ] = "rsq"; } void InstructionsSoa::createDependencies() @@ -191,8 +192,9 @@ void InstructionsSoa::createDependencies() m_builtinDependencies["pow"] = powDeps; } { - std::vector absDeps(1); + std::vector absDeps(2); absDeps[0] = "fabsf"; + absDeps[1] = "absvec"; m_builtinDependencies["abs"] = absDeps; } { @@ -213,6 +215,14 @@ void InstructionsSoa::createDependencies() litDeps[3] = "powvec"; m_builtinDependencies["lit"] = litDeps; } + { + std::vector rsqDeps(4); + rsqDeps[0] = "sqrtf"; + rsqDeps[1] = "sqrtvec"; + rsqDeps[2] = "fabsf"; + rsqDeps[3] = "absvec"; + m_builtinDependencies["rsq"] = rsqDeps; + } } llvm::Function * InstructionsSoa::function(int op) @@ -453,7 +463,9 @@ void InstructionsSoa::injectFunction(llvm::Function *originalFunc, int op) currentModule()->dump(); } else { DenseMap val; + val[m_builtins->getFunction("fabsf")] = currentModule()->getFunction("fabsf"); val[m_builtins->getFunction("powf")] = currentModule()->getFunction("powf"); + val[m_builtins->getFunction("sqrtf")] = currentModule()->getFunction("sqrtf"); func = CloneFunction(originalFunc, val); #if 0 std::cout <<" replacing "<getFunction("powf") @@ -490,3 +502,9 @@ std::vector InstructionsSoa::lit(const std::vector i return callBuiltin(func, in); } +std::vector InstructionsSoa::rsq(const std::vector in) +{ + llvm::Function *func = function(TGSI_OPCODE_RSQ); + return callBuiltin(func, in); +} + diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.h b/src/gallium/auxiliary/gallivm/instructionssoa.h index 477ef4a157..3e20b652dd 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.h +++ b/src/gallium/auxiliary/gallivm/instructionssoa.h @@ -68,6 +68,7 @@ public: const std::vector in2); std::vector pow(const std::vector in1, const std::vector in2); + std::vector rsq(const std::vector in1); std::vector sub(const std::vector in1, const std::vector in2); void end(); diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index b3bfebfe50..64c02aa967 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -36,28 +36,24 @@ typedef __attribute__(( ext_vector_type(4) )) float float4; extern float fabsf(float val); +float4 absvec(float4 vec) +{ + float4 res; + res.x = fabsf(vec.x); + res.y = fabsf(vec.y); + res.z = fabsf(vec.z); + res.w = fabsf(vec.w); + + return res; +} + void abs(float4 *res, float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w) { - res[0].x = fabsf(tmp0x.x); - res[0].y = fabsf(tmp0x.y); - res[0].z = fabsf(tmp0x.z); - res[0].w = fabsf(tmp0x.w); - - res[1].x = fabsf(tmp0y.x); - res[1].y = fabsf(tmp0y.y); - res[1].z = fabsf(tmp0y.z); - res[1].w = fabsf(tmp0y.w); - - res[2].x = fabsf(tmp0z.x); - res[2].y = fabsf(tmp0z.y); - res[2].z = fabsf(tmp0z.z); - res[2].w = fabsf(tmp0z.w); - - res[3].x = fabsf(tmp0w.x); - res[3].y = fabsf(tmp0w.y); - res[3].z = fabsf(tmp0w.z); - res[3].w = fabsf(tmp0w.w); + res[0] = absvec(tmp0x); + res[1] = absvec(tmp0y); + res[2] = absvec(tmp0z); + res[3] = absvec(tmp0w); } void dp3(float4 *res, @@ -88,6 +84,7 @@ void dp4(float4 *res, } extern float powf(float num, float p); +extern float sqrtf(float x); float4 powvec(float4 vec, float4 q) { @@ -168,3 +165,24 @@ void lit(float4 *res, } res[3] = (float4){1.0, 1.0, 1.0, 1.0}; } + + +float4 sqrtvec(float4 vec) +{ + float4 p; + p.x = sqrtf(vec.x); + p.y = sqrtf(vec.y); + p.z = sqrtf(vec.z); + p.w = sqrtf(vec.w); + return p; +} + +void rsq(float4 *res, + float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w) +{ + const float4 onevec = (float4) {1., 1., 1., 1.}; + res[0] = onevec/sqrtvec(absvec(tmp0x)); + res[1] = res[0]; + res[2] = res[0]; + res[3] = res[0]; +} diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index abcb240f46..9695358ab8 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -699,6 +699,7 @@ translate_instructionir(llvm::Module *module, } break; case TGSI_OPCODE_RSQ: { + out = instr->rsq(inputs[0]); } break; case TGSI_OPCODE_EXP: -- cgit v1.2.3 From a7449d4d840148ccd9261b59e68d45e9d0d2be53 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 16 May 2008 17:35:47 -0400 Subject: fix rsq --- src/gallium/auxiliary/gallivm/soabuiltins.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index 64c02aa967..62c75f18f4 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -182,7 +182,7 @@ void rsq(float4 *res, { const float4 onevec = (float4) {1., 1., 1., 1.}; res[0] = onevec/sqrtvec(absvec(tmp0x)); - res[1] = res[0]; - res[2] = res[0]; - res[3] = res[0]; + res[1] = onevec/sqrtvec(absvec(tmp0y)); + res[2] = onevec/sqrtvec(absvec(tmp0z)); + res[3] = onevec/sqrtvec(absvec(tmp0w)); } -- cgit v1.2.3 From 0bf82c0111c9d5e33ffc76be2fd0d22eea316952 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 16 May 2008 17:56:38 -0400 Subject: cosmetic changes --- src/gallium/auxiliary/gallivm/soabuiltins.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/soabuiltins.c b/src/gallium/auxiliary/gallivm/soabuiltins.c index 62c75f18f4..78f84510e2 100644 --- a/src/gallium/auxiliary/gallivm/soabuiltins.c +++ b/src/gallium/auxiliary/gallivm/soabuiltins.c @@ -147,17 +147,16 @@ void max(float4 *res, void lit(float4 *res, float4 tmp0x, float4 tmp0y, float4 tmp0z, float4 tmp0w) { - const float4 zerovec = (float4) {0, 0, 0, 0}; + const float4 zerovec = (float4) {0.0, 0.0, 0.0, 0.0}; const float4 min128 = (float4) {-128.f, -128.f, -128.f, -128.f}; const float4 plus128 = (float4) {128.f, 128.f, 128.f, 128.f}; res[0] = (float4){1.0, 1.0, 1.0, 1.0}; if (tmp0x.x > 0) { - float4 tmpx = maxvec(tmpx, zerovec); float4 tmpy = maxvec(tmp0y, zerovec); float4 tmpw = minvec(tmp0w, plus128); tmpw = maxvec(tmpw, min128); - res[1] = tmpx; + res[1] = tmp0x; res[2] = powvec(tmpy, tmpw); } else { res[1] = zerovec; -- cgit v1.2.3 From f9e1ef2a5b7951d36db56913b1366cf65b9d0976 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sun, 18 May 2008 15:21:28 -0600 Subject: alias ProgramEnvParameter4xyARB and ProgramParameter4xyNV (bug #12935) these should be the same functions (as per spec). cherry-picked from master (86a4810b09097714942bf2b889e6c62357bba931) --- src/mesa/glapi/gl_API.xml | 14 ++- src/mesa/main/config.h | 39 ++++---- src/mesa/main/dlist.c | 205 ++++++++++++++++--------------------------- src/mesa/main/state.c | 8 +- src/mesa/shader/arbprogram.c | 29 +++++- src/mesa/shader/nvprogram.c | 69 --------------- src/mesa/shader/nvprogram.h | 12 --- 7 files changed, 132 insertions(+), 244 deletions(-) (limited to 'src') diff --git a/src/mesa/glapi/gl_API.xml b/src/mesa/glapi/gl_API.xml index 3d47e6f2ce..ef4a309cd6 100644 --- a/src/mesa/glapi/gl_API.xml +++ b/src/mesa/glapi/gl_API.xml @@ -11275,7 +11275,7 @@ - + @@ -11284,14 +11284,13 @@ - + - - + - + @@ -11300,11 +11299,10 @@ - + - - + diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h index 8c64248845..232e698f50 100644 --- a/src/mesa/main/config.h +++ b/src/mesa/main/config.h @@ -163,25 +163,6 @@ /** For GL_EXT_texture_lod_bias (typically MAX_TEXTURE_LEVELS - 1) */ #define MAX_TEXTURE_LOD_BIAS 11.0 -/** For GL_NV_vertex_program */ -/*@{*/ -#define MAX_NV_VERTEX_PROGRAM_INSTRUCTIONS 128 -#define MAX_NV_VERTEX_PROGRAM_TEMPS 12 -#define MAX_NV_VERTEX_PROGRAM_PARAMS 128 /* KW: power of two */ -#define MAX_NV_VERTEX_PROGRAM_INPUTS 16 -#define MAX_NV_VERTEX_PROGRAM_OUTPUTS 15 -/*@}*/ - -/** For GL_NV_fragment_program */ -/*@{*/ -#define MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS 1024 /* 72 for GL_ARB_f_p */ -#define MAX_NV_FRAGMENT_PROGRAM_TEMPS 96 -#define MAX_NV_FRAGMENT_PROGRAM_PARAMS 64 -#define MAX_NV_FRAGMENT_PROGRAM_INPUTS 12 -#define MAX_NV_FRAGMENT_PROGRAM_OUTPUTS 3 -#define MAX_NV_FRAGMENT_PROGRAM_WRITE_ONLYS 2 -/*@}*/ - /** For GL_ARB_vertex_program */ /*@{*/ #define MAX_VERTEX_PROGRAM_ADDRESS_REGS 1 @@ -210,6 +191,26 @@ #define MAX_SAMPLERS 8 /*@}*/ +/** For GL_NV_vertex_program */ +/*@{*/ +#define MAX_NV_VERTEX_PROGRAM_INSTRUCTIONS 128 +#define MAX_NV_VERTEX_PROGRAM_TEMPS 12 +#define MAX_NV_VERTEX_PROGRAM_PARAMS MAX_PROGRAM_ENV_PARAMS +#define MAX_NV_VERTEX_PROGRAM_INPUTS 16 +#define MAX_NV_VERTEX_PROGRAM_OUTPUTS 15 +/*@}*/ + +/** For GL_NV_fragment_program */ +/*@{*/ +#define MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS 1024 /* 72 for GL_ARB_f_p */ +#define MAX_NV_FRAGMENT_PROGRAM_TEMPS 96 +#define MAX_NV_FRAGMENT_PROGRAM_PARAMS 64 +#define MAX_NV_FRAGMENT_PROGRAM_INPUTS 12 +#define MAX_NV_FRAGMENT_PROGRAM_OUTPUTS 3 +#define MAX_NV_FRAGMENT_PROGRAM_WRITE_ONLYS 2 +/*@}*/ + + /** For GL_ARB_vertex_shader */ /*@{*/ #define MAX_VERTEX_ATTRIBS 16 diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c index 293ee5fa34..13ebd4dd0d 100644 --- a/src/mesa/main/dlist.c +++ b/src/mesa/main/dlist.c @@ -303,7 +303,6 @@ typedef enum OPCODE_EXECUTE_PROGRAM_NV, OPCODE_REQUEST_RESIDENT_PROGRAMS_NV, OPCODE_LOAD_PROGRAM_NV, - OPCODE_PROGRAM_PARAMETER4F_NV, OPCODE_TRACK_MATRIX_NV, /* GL_NV_fragment_program */ OPCODE_PROGRAM_LOCAL_PARAMETER_ARB, @@ -4248,77 +4247,111 @@ save_BindProgramNV(GLenum target, GLuint id) CALL_BindProgramNV(ctx->Exec, (target, id)); } } -#endif /* FEATURE_NV_vertex_program || FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program */ -#if FEATURE_NV_vertex_program static void GLAPIENTRY -save_ExecuteProgramNV(GLenum target, GLuint id, const GLfloat *params) +save_ProgramEnvParameter4fARB(GLenum target, GLuint index, + GLfloat x, GLfloat y, GLfloat z, GLfloat w) { GET_CURRENT_CONTEXT(ctx); Node *n; ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx); - n = ALLOC_INSTRUCTION(ctx, OPCODE_EXECUTE_PROGRAM_NV, 6); + n = ALLOC_INSTRUCTION(ctx, OPCODE_PROGRAM_ENV_PARAMETER_ARB, 6); if (n) { n[1].e = target; - n[2].ui = id; - n[3].f = params[0]; - n[4].f = params[1]; - n[5].f = params[2]; - n[6].f = params[3]; + n[2].ui = index; + n[3].f = x; + n[4].f = y; + n[5].f = z; + n[6].f = w; } if (ctx->ExecuteFlag) { - CALL_ExecuteProgramNV(ctx->Exec, (target, id, params)); + CALL_ProgramEnvParameter4fARB(ctx->Exec, (target, index, x, y, z, w)); } } static void GLAPIENTRY -save_ProgramParameter4fNV(GLenum target, GLuint index, - GLfloat x, GLfloat y, GLfloat z, GLfloat w) +save_ProgramEnvParameter4fvARB(GLenum target, GLuint index, + const GLfloat *params) +{ + save_ProgramEnvParameter4fARB(target, index, params[0], params[1], + params[2], params[3]); +} + + +static void GLAPIENTRY +save_ProgramEnvParameters4fvEXT(GLenum target, GLuint index, GLsizei count, + const GLfloat * params) { GET_CURRENT_CONTEXT(ctx); Node *n; ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx); - n = ALLOC_INSTRUCTION(ctx, OPCODE_PROGRAM_PARAMETER4F_NV, 6); - if (n) { - n[1].e = target; - n[2].ui = index; - n[3].f = x; - n[4].f = y; - n[5].f = z; - n[6].f = w; + + if (count > 0) { + GLint i; + const GLfloat * p = params; + + for (i = 0 ; i < count ; i++) { + n = ALLOC_INSTRUCTION(ctx, OPCODE_PROGRAM_ENV_PARAMETER_ARB, 6); + if (n) { + n[1].e = target; + n[2].ui = index; + n[3].f = p[0]; + n[4].f = p[1]; + n[5].f = p[2]; + n[6].f = p[3]; + p += 4; + } + } } + if (ctx->ExecuteFlag) { - CALL_ProgramParameter4fNV(ctx->Exec, (target, index, x, y, z, w)); + CALL_ProgramEnvParameters4fvEXT(ctx->Exec, (target, index, count, params)); } } static void GLAPIENTRY -save_ProgramParameter4fvNV(GLenum target, GLuint index, - const GLfloat *params) +save_ProgramEnvParameter4dARB(GLenum target, GLuint index, + GLdouble x, GLdouble y, GLdouble z, GLdouble w) { - save_ProgramParameter4fNV(target, index, params[0], params[1], - params[2], params[3]); + save_ProgramEnvParameter4fARB(target, index, + (GLfloat) x, + (GLfloat) y, (GLfloat) z, (GLfloat) w); } static void GLAPIENTRY -save_ProgramParameter4dNV(GLenum target, GLuint index, - GLdouble x, GLdouble y, GLdouble z, GLdouble w) +save_ProgramEnvParameter4dvARB(GLenum target, GLuint index, + const GLdouble *params) { - save_ProgramParameter4fNV(target, index, (GLfloat) x, (GLfloat) y, - (GLfloat) z, (GLfloat) w); + save_ProgramEnvParameter4fARB(target, index, + (GLfloat) params[0], + (GLfloat) params[1], + (GLfloat) params[2], (GLfloat) params[3]); } +#endif /* FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program || FEATURE_NV_vertex_program */ +#if FEATURE_NV_vertex_program static void GLAPIENTRY -save_ProgramParameter4dvNV(GLenum target, GLuint index, - const GLdouble *params) +save_ExecuteProgramNV(GLenum target, GLuint id, const GLfloat *params) { - save_ProgramParameter4fNV(target, index, (GLfloat) params[0], - (GLfloat) params[1], (GLfloat) params[2], - (GLfloat) params[3]); + GET_CURRENT_CONTEXT(ctx); + Node *n; + ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx); + n = ALLOC_INSTRUCTION(ctx, OPCODE_EXECUTE_PROGRAM_NV, 6); + if (n) { + n[1].e = target; + n[2].ui = id; + n[3].f = params[0]; + n[4].f = params[1]; + n[5].f = params[2]; + n[6].f = params[3]; + } + if (ctx->ExecuteFlag) { + CALL_ExecuteProgramNV(ctx->Exec, (target, id, params)); + } } @@ -4328,7 +4361,7 @@ save_ProgramParameters4dvNV(GLenum target, GLuint index, { GLuint i; for (i = 0; i < num; i++) { - save_ProgramParameter4dvNV(target, index + i, params + 4 * i); + save_ProgramEnvParameter4dvARB(target, index + i, params + 4 * i); } } @@ -4339,7 +4372,7 @@ save_ProgramParameters4fvNV(GLenum target, GLuint index, { GLuint i; for (i = 0; i < num; i++) { - save_ProgramParameter4fvNV(target, index + i, params + 4 * i); + save_ProgramEnvParameter4fvARB(target, index + i, params + 4 * i); } } @@ -4667,90 +4700,6 @@ save_ProgramStringARB(GLenum target, GLenum format, GLsizei len, } } - -static void GLAPIENTRY -save_ProgramEnvParameter4fARB(GLenum target, GLuint index, - GLfloat x, GLfloat y, GLfloat z, GLfloat w) -{ - GET_CURRENT_CONTEXT(ctx); - Node *n; - ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx); - n = ALLOC_INSTRUCTION(ctx, OPCODE_PROGRAM_ENV_PARAMETER_ARB, 6); - if (n) { - n[1].e = target; - n[2].ui = index; - n[3].f = x; - n[4].f = y; - n[5].f = z; - n[6].f = w; - } - if (ctx->ExecuteFlag) { - CALL_ProgramEnvParameter4fARB(ctx->Exec, (target, index, x, y, z, w)); - } -} - - -static void GLAPIENTRY -save_ProgramEnvParameter4fvARB(GLenum target, GLuint index, - const GLfloat *params) -{ - save_ProgramEnvParameter4fARB(target, index, params[0], params[1], - params[2], params[3]); -} - - -static void GLAPIENTRY -save_ProgramEnvParameters4fvEXT(GLenum target, GLuint index, GLsizei count, - const GLfloat * params) -{ - GET_CURRENT_CONTEXT(ctx); - Node *n; - ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx); - - if (count > 0) { - GLint i; - const GLfloat * p = params; - - for (i = 0 ; i < count ; i++) { - n = ALLOC_INSTRUCTION(ctx, OPCODE_PROGRAM_ENV_PARAMETER_ARB, 6); - if (n) { - n[1].e = target; - n[2].ui = index; - n[3].f = p[0]; - n[4].f = p[1]; - n[5].f = p[2]; - n[6].f = p[3]; - p += 4; - } - } - } - - if (ctx->ExecuteFlag) { - CALL_ProgramEnvParameters4fvEXT(ctx->Exec, (target, index, count, params)); - } -} - - -static void GLAPIENTRY -save_ProgramEnvParameter4dARB(GLenum target, GLuint index, - GLdouble x, GLdouble y, GLdouble z, GLdouble w) -{ - save_ProgramEnvParameter4fARB(target, index, - (GLfloat) x, - (GLfloat) y, (GLfloat) z, (GLfloat) w); -} - - -static void GLAPIENTRY -save_ProgramEnvParameter4dvARB(GLenum target, GLuint index, - const GLdouble *params) -{ - save_ProgramEnvParameter4fARB(target, index, - (GLfloat) params[0], - (GLfloat) params[1], - (GLfloat) params[2], (GLfloat) params[3]); -} - #endif /* FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program */ @@ -6425,10 +6374,6 @@ execute_list(GLcontext *ctx, GLuint list) CALL_LoadProgramNV(ctx->Exec, (n[1].e, n[2].ui, n[3].i, (const GLubyte *) n[4].data)); break; - case OPCODE_PROGRAM_PARAMETER4F_NV: - CALL_ProgramParameter4fNV(ctx->Exec, (n[1].e, n[2].ui, n[3].f, - n[4].f, n[5].f, n[6].f)); - break; case OPCODE_TRACK_MATRIX_NV: CALL_TrackMatrixNV(ctx->Exec, (n[1].e, n[2].ui, n[3].e, n[4].e)); break; @@ -6459,6 +6404,8 @@ execute_list(GLcontext *ctx, GLuint list) CALL_ProgramStringARB(ctx->Exec, (n[1].e, n[2].e, n[3].i, n[4].data)); break; +#endif +#if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program || FEATURE_NV_vertex_program case OPCODE_PROGRAM_ENV_PARAMETER_ARB: CALL_ProgramEnvParameter4fARB(ctx->Exec, (n[1].e, n[2].ui, n[3].f, n[4].f, n[5].f, @@ -8014,10 +7961,10 @@ _mesa_init_dlist_table(struct _glapi_table *table) SET_GetVertexAttribPointervNV(table, _mesa_GetVertexAttribPointervNV); SET_IsProgramNV(table, _mesa_IsProgramARB); SET_LoadProgramNV(table, save_LoadProgramNV); - SET_ProgramParameter4dNV(table, save_ProgramParameter4dNV); - SET_ProgramParameter4dvNV(table, save_ProgramParameter4dvNV); - SET_ProgramParameter4fNV(table, save_ProgramParameter4fNV); - SET_ProgramParameter4fvNV(table, save_ProgramParameter4fvNV); + SET_ProgramEnvParameter4dARB(table, save_ProgramEnvParameter4dARB); + SET_ProgramEnvParameter4dvARB(table, save_ProgramEnvParameter4dvARB); + SET_ProgramEnvParameter4fARB(table, save_ProgramEnvParameter4fARB); + SET_ProgramEnvParameter4fvARB(table, save_ProgramEnvParameter4fvARB); SET_ProgramParameters4dvNV(table, save_ProgramParameters4dvNV); SET_ProgramParameters4fvNV(table, save_ProgramParameters4fvNV); SET_TrackMatrixNV(table, save_TrackMatrixNV); diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 90379a1772..f8cb943e64 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -536,10 +536,10 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_GetVertexAttribPointervNV(exec, _mesa_GetVertexAttribPointervNV); SET_IsProgramNV(exec, _mesa_IsProgramARB); SET_LoadProgramNV(exec, _mesa_LoadProgramNV); - SET_ProgramParameter4dNV(exec, _mesa_ProgramParameter4dNV); - SET_ProgramParameter4dvNV(exec, _mesa_ProgramParameter4dvNV); - SET_ProgramParameter4fNV(exec, _mesa_ProgramParameter4fNV); - SET_ProgramParameter4fvNV(exec, _mesa_ProgramParameter4fvNV); + SET_ProgramEnvParameter4dARB(exec, _mesa_ProgramEnvParameter4dARB); /* alias to ProgramParameter4dNV */ + SET_ProgramEnvParameter4dvARB(exec, _mesa_ProgramEnvParameter4dvARB); /* alias to ProgramParameter4dvNV */ + SET_ProgramEnvParameter4fARB(exec, _mesa_ProgramEnvParameter4fARB); /* alias to ProgramParameter4fNV */ + SET_ProgramEnvParameter4fvARB(exec, _mesa_ProgramEnvParameter4fvARB); /* alias to ProgramParameter4fvNV */ SET_ProgramParameters4dvNV(exec, _mesa_ProgramParameters4dvNV); SET_ProgramParameters4fvNV(exec, _mesa_ProgramParameters4fvNV); SET_TrackMatrixNV(exec, _mesa_TrackMatrixNV); diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c index ee75be315e..1656dc9450 100644 --- a/src/mesa/shader/arbprogram.c +++ b/src/mesa/shader/arbprogram.c @@ -247,6 +247,12 @@ _mesa_ProgramStringARB(GLenum target, GLenum format, GLsizei len, } +/** + * Set a program env parameter register. + * \note Called from the GL API dispatcher. + * Note, this function is also used by the GL_NV_vertex_program extension + * (alias to ProgramParameterdNV) + */ void GLAPIENTRY _mesa_ProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w) @@ -256,6 +262,12 @@ _mesa_ProgramEnvParameter4dARB(GLenum target, GLuint index, } +/** + * Set a program env parameter register. + * \note Called from the GL API dispatcher. + * Note, this function is also used by the GL_NV_vertex_program extension + * (alias to ProgramParameterdvNV) + */ void GLAPIENTRY _mesa_ProgramEnvParameter4dvARB(GLenum target, GLuint index, const GLdouble *params) @@ -266,6 +278,12 @@ _mesa_ProgramEnvParameter4dvARB(GLenum target, GLuint index, } +/** + * Set a program env parameter register. + * \note Called from the GL API dispatcher. + * Note, this function is also used by the GL_NV_vertex_program extension + * (alias to ProgramParameterfNV) + */ void GLAPIENTRY _mesa_ProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w) @@ -283,8 +301,8 @@ _mesa_ProgramEnvParameter4fARB(GLenum target, GLuint index, } ASSIGN_4V(ctx->FragmentProgram.Parameters[index], x, y, z, w); } - else if (target == GL_VERTEX_PROGRAM_ARB - && ctx->Extensions.ARB_vertex_program) { + else if (target == GL_VERTEX_PROGRAM_ARB /* == GL_VERTEX_PROGRAM_NV */ + && (ctx->Extensions.ARB_vertex_program || ctx->Extensions.NV_vertex_program)) { if (index >= ctx->Const.VertexProgram.MaxEnvParams) { _mesa_error(ctx, GL_INVALID_VALUE, "glProgramEnvParameter(index)"); return; @@ -297,7 +315,12 @@ _mesa_ProgramEnvParameter4fARB(GLenum target, GLuint index, } } - +/** + * Set a program env parameter register. + * \note Called from the GL API dispatcher. + * Note, this function is also used by the GL_NV_vertex_program extension + * (alias to ProgramParameterfvNV) + */ void GLAPIENTRY _mesa_ProgramEnvParameter4fvARB(GLenum target, GLuint index, const GLfloat *params) diff --git a/src/mesa/shader/nvprogram.c b/src/mesa/shader/nvprogram.c index 337784bfe2..409c61cdc1 100644 --- a/src/mesa/shader/nvprogram.c +++ b/src/mesa/shader/nvprogram.c @@ -575,75 +575,6 @@ _mesa_LoadProgramNV(GLenum target, GLuint id, GLsizei len, -/** - * Set a program parameter register. - * \note Called from the GL API dispatcher. - */ -void GLAPIENTRY -_mesa_ProgramParameter4dNV(GLenum target, GLuint index, - GLdouble x, GLdouble y, GLdouble z, GLdouble w) -{ - _mesa_ProgramParameter4fNV(target, index, - (GLfloat)x, (GLfloat)y, (GLfloat)z, (GLfloat)w); -} - - -/** - * Set a program parameter register. - * \note Called from the GL API dispatcher. - */ -void GLAPIENTRY -_mesa_ProgramParameter4dvNV(GLenum target, GLuint index, - const GLdouble *params) -{ - _mesa_ProgramParameter4fNV(target, index, - (GLfloat)params[0], (GLfloat)params[1], - (GLfloat)params[2], (GLfloat)params[3]); -} - - -/** - * Set a program parameter register. - * \note Called from the GL API dispatcher. - */ -void GLAPIENTRY -_mesa_ProgramParameter4fNV(GLenum target, GLuint index, - GLfloat x, GLfloat y, GLfloat z, GLfloat w) -{ - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (target == GL_VERTEX_PROGRAM_NV && ctx->Extensions.NV_vertex_program) { - if (index < MAX_NV_VERTEX_PROGRAM_PARAMS) { - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - ASSIGN_4V(ctx->VertexProgram.Parameters[index], x, y, z, w); - } - else { - _mesa_error(ctx, GL_INVALID_VALUE, "glProgramParameterNV(index)"); - return; - } - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glProgramParameterNV"); - return; - } -} - - -/** - * Set a program parameter register. - * \note Called from the GL API dispatcher. - */ -void GLAPIENTRY -_mesa_ProgramParameter4fvNV(GLenum target, GLuint index, - const GLfloat *params) -{ - _mesa_ProgramParameter4fNV(target, index, - params[0], params[1], params[2], params[3]); -} - - - /** * Set a sequence of program parameter registers. * \note Called from the GL API dispatcher. diff --git a/src/mesa/shader/nvprogram.h b/src/mesa/shader/nvprogram.h index dcea7727e0..bfac165b5e 100644 --- a/src/mesa/shader/nvprogram.h +++ b/src/mesa/shader/nvprogram.h @@ -69,18 +69,6 @@ _mesa_GetVertexAttribPointervNV(GLuint index, GLenum pname, GLvoid **pointer); extern void GLAPIENTRY _mesa_LoadProgramNV(GLenum target, GLuint id, GLsizei len, const GLubyte *program); -extern void GLAPIENTRY -_mesa_ProgramParameter4dNV(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); - -extern void GLAPIENTRY -_mesa_ProgramParameter4dvNV(GLenum target, GLuint index, const GLdouble *params); - -extern void GLAPIENTRY -_mesa_ProgramParameter4fNV(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); - -extern void GLAPIENTRY -_mesa_ProgramParameter4fvNV(GLenum target, GLuint index, const GLfloat *params); - extern void GLAPIENTRY _mesa_ProgramParameters4dvNV(GLenum target, GLuint index, GLuint num, const GLdouble *params); -- cgit v1.2.3 From 1bb30b02dd7610fb1094491c232e8dd4497a931a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sun, 18 May 2008 15:23:03 -0600 Subject: Regenerated API dispatch files Follow on to cherry-pick from master --- src/glx/x11/indirect.c | 114 --- src/glx/x11/indirect.h | 6 - src/glx/x11/indirect_init.c | 6 - src/mesa/drivers/dri/common/extension_helper.h | 48 +- src/mesa/glapi/dispatch.h | 224 +++-- src/mesa/glapi/glapioffsets.h | 204 +++-- src/mesa/glapi/glapitable.h | 198 +++-- src/mesa/glapi/glapitemp.h | 122 +-- src/mesa/glapi/glprocs.h | 472 +++++----- src/mesa/sparc/glapi_sparc.S | 56 +- src/mesa/x86-64/glapi_x86-64.S | 1092 ++++++++++-------------- src/mesa/x86/glapi_x86.S | 60 +- 12 files changed, 1116 insertions(+), 1486 deletions(-) (limited to 'src') diff --git a/src/glx/x11/indirect.c b/src/glx/x11/indirect.c index 87c523383b..fbb2a91956 100644 --- a/src/glx/x11/indirect.c +++ b/src/glx/x11/indirect.c @@ -7468,26 +7468,6 @@ __indirect_glGetProgramivARB(GLenum target, GLenum pname, GLint * params) return; } -#define X_GLrop_ProgramEnvParameter4dvARB 4185 -void -__indirect_glProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x, - GLdouble y, GLdouble z, GLdouble w) -{ - __GLXcontext *const gc = __glXGetCurrentContext(); - const GLuint cmdlen = 44; - emit_header(gc->pc, X_GLrop_ProgramEnvParameter4dvARB, cmdlen); - (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); - (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); - (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 8); - (void) memcpy((void *) (gc->pc + 20), (void *) (&y), 8); - (void) memcpy((void *) (gc->pc + 28), (void *) (&z), 8); - (void) memcpy((void *) (gc->pc + 36), (void *) (&w), 8); - gc->pc += cmdlen; - if (__builtin_expect(gc->pc > gc->limit, 0)) { - (void) __glXFlushRenderBuffer(gc, gc->pc); - } -} - #define X_GLrop_ProgramEnvParameter4dvARB 4185 void __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, @@ -7505,26 +7485,6 @@ __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, } } -#define X_GLrop_ProgramEnvParameter4fvARB 4184 -void -__indirect_glProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x, - GLfloat y, GLfloat z, GLfloat w) -{ - __GLXcontext *const gc = __glXGetCurrentContext(); - const GLuint cmdlen = 28; - emit_header(gc->pc, X_GLrop_ProgramEnvParameter4fvARB, cmdlen); - (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); - (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); - (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 4); - (void) memcpy((void *) (gc->pc + 16), (void *) (&y), 4); - (void) memcpy((void *) (gc->pc + 20), (void *) (&z), 4); - (void) memcpy((void *) (gc->pc + 24), (void *) (&w), 4); - gc->pc += cmdlen; - if (__builtin_expect(gc->pc > gc->limit, 0)) { - (void) __glXFlushRenderBuffer(gc, gc->pc); - } -} - #define X_GLrop_ProgramEnvParameter4fvARB 4184 void __indirect_glProgramEnvParameter4fvARB(GLenum target, GLuint index, @@ -9104,80 +9064,6 @@ __indirect_glLoadProgramNV(GLenum target, GLuint id, GLsizei len, } } -#define X_GLrop_ProgramParameter4dvNV 4185 -void -__indirect_glProgramParameter4dNV(GLenum target, GLuint index, GLdouble x, - GLdouble y, GLdouble z, GLdouble w) -{ - __GLXcontext *const gc = __glXGetCurrentContext(); - const GLuint cmdlen = 44; - emit_header(gc->pc, X_GLrop_ProgramParameter4dvNV, cmdlen); - (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); - (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); - (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 8); - (void) memcpy((void *) (gc->pc + 20), (void *) (&y), 8); - (void) memcpy((void *) (gc->pc + 28), (void *) (&z), 8); - (void) memcpy((void *) (gc->pc + 36), (void *) (&w), 8); - gc->pc += cmdlen; - if (__builtin_expect(gc->pc > gc->limit, 0)) { - (void) __glXFlushRenderBuffer(gc, gc->pc); - } -} - -#define X_GLrop_ProgramParameter4dvNV 4185 -void -__indirect_glProgramParameter4dvNV(GLenum target, GLuint index, - const GLdouble * params) -{ - __GLXcontext *const gc = __glXGetCurrentContext(); - const GLuint cmdlen = 44; - emit_header(gc->pc, X_GLrop_ProgramParameter4dvNV, cmdlen); - (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); - (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); - (void) memcpy((void *) (gc->pc + 12), (void *) (params), 32); - gc->pc += cmdlen; - if (__builtin_expect(gc->pc > gc->limit, 0)) { - (void) __glXFlushRenderBuffer(gc, gc->pc); - } -} - -#define X_GLrop_ProgramParameter4fvNV 4184 -void -__indirect_glProgramParameter4fNV(GLenum target, GLuint index, GLfloat x, - GLfloat y, GLfloat z, GLfloat w) -{ - __GLXcontext *const gc = __glXGetCurrentContext(); - const GLuint cmdlen = 28; - emit_header(gc->pc, X_GLrop_ProgramParameter4fvNV, cmdlen); - (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); - (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); - (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 4); - (void) memcpy((void *) (gc->pc + 16), (void *) (&y), 4); - (void) memcpy((void *) (gc->pc + 20), (void *) (&z), 4); - (void) memcpy((void *) (gc->pc + 24), (void *) (&w), 4); - gc->pc += cmdlen; - if (__builtin_expect(gc->pc > gc->limit, 0)) { - (void) __glXFlushRenderBuffer(gc, gc->pc); - } -} - -#define X_GLrop_ProgramParameter4fvNV 4184 -void -__indirect_glProgramParameter4fvNV(GLenum target, GLuint index, - const GLfloat * params) -{ - __GLXcontext *const gc = __glXGetCurrentContext(); - const GLuint cmdlen = 28; - emit_header(gc->pc, X_GLrop_ProgramParameter4fvNV, cmdlen); - (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); - (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); - (void) memcpy((void *) (gc->pc + 12), (void *) (params), 16); - gc->pc += cmdlen; - if (__builtin_expect(gc->pc > gc->limit, 0)) { - (void) __glXFlushRenderBuffer(gc, gc->pc); - } -} - #define X_GLrop_ProgramParameters4dvNV 4187 void __indirect_glProgramParameters4dvNV(GLenum target, GLuint index, GLuint num, diff --git a/src/glx/x11/indirect.h b/src/glx/x11/indirect.h index 630062e233..f8c88b36bb 100644 --- a/src/glx/x11/indirect.h +++ b/src/glx/x11/indirect.h @@ -517,9 +517,7 @@ extern HIDDEN void __indirect_glGetProgramivARB(GLenum target, GLenum pname, GLi extern HIDDEN void __indirect_glGetVertexAttribdvARB(GLuint index, GLenum pname, GLdouble * params); extern HIDDEN void __indirect_glGetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat * params); extern HIDDEN void __indirect_glGetVertexAttribivARB(GLuint index, GLenum pname, GLint * params); -extern HIDDEN void __indirect_glProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); extern HIDDEN void __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, const GLdouble * params); -extern HIDDEN void __indirect_glProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); extern HIDDEN void __indirect_glProgramEnvParameter4fvARB(GLenum target, GLuint index, const GLfloat * params); extern HIDDEN void __indirect_glProgramLocalParameter4dARB(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); extern HIDDEN void __indirect_glProgramLocalParameter4dvARB(GLenum target, GLuint index, const GLdouble * params); @@ -639,10 +637,6 @@ extern HIDDEN void __indirect_glGetVertexAttribfvNV(GLuint index, GLenum pname, extern HIDDEN void __indirect_glGetVertexAttribivNV(GLuint index, GLenum pname, GLint * params); extern HIDDEN GLboolean __indirect_glIsProgramNV(GLuint program); extern HIDDEN void __indirect_glLoadProgramNV(GLenum target, GLuint id, GLsizei len, const GLubyte * program); -extern HIDDEN void __indirect_glProgramParameter4dNV(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); -extern HIDDEN void __indirect_glProgramParameter4dvNV(GLenum target, GLuint index, const GLdouble * params); -extern HIDDEN void __indirect_glProgramParameter4fNV(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); -extern HIDDEN void __indirect_glProgramParameter4fvNV(GLenum target, GLuint index, const GLfloat * params); extern HIDDEN void __indirect_glProgramParameters4dvNV(GLenum target, GLuint index, GLuint num, const GLdouble * params); extern HIDDEN void __indirect_glProgramParameters4fvNV(GLenum target, GLuint index, GLuint num, const GLfloat * params); extern HIDDEN void __indirect_glRequestResidentProgramsNV(GLsizei n, const GLuint * ids); diff --git a/src/glx/x11/indirect_init.c b/src/glx/x11/indirect_init.c index aaa70c8796..479184337c 100644 --- a/src/glx/x11/indirect_init.c +++ b/src/glx/x11/indirect_init.c @@ -526,9 +526,7 @@ __GLapi * __glXNewIndirectAPI( void ) glAPI->GetVertexAttribdvARB = __indirect_glGetVertexAttribdvARB; glAPI->GetVertexAttribfvARB = __indirect_glGetVertexAttribfvARB; glAPI->GetVertexAttribivARB = __indirect_glGetVertexAttribivARB; - glAPI->ProgramEnvParameter4dARB = __indirect_glProgramEnvParameter4dARB; glAPI->ProgramEnvParameter4dvARB = __indirect_glProgramEnvParameter4dvARB; - glAPI->ProgramEnvParameter4fARB = __indirect_glProgramEnvParameter4fARB; glAPI->ProgramEnvParameter4fvARB = __indirect_glProgramEnvParameter4fvARB; glAPI->ProgramLocalParameter4dARB = __indirect_glProgramLocalParameter4dARB; glAPI->ProgramLocalParameter4dvARB = __indirect_glProgramLocalParameter4dvARB; @@ -681,10 +679,6 @@ __GLapi * __glXNewIndirectAPI( void ) glAPI->GetVertexAttribivNV = __indirect_glGetVertexAttribivNV; glAPI->IsProgramNV = __indirect_glIsProgramNV; glAPI->LoadProgramNV = __indirect_glLoadProgramNV; - glAPI->ProgramParameter4dNV = __indirect_glProgramParameter4dNV; - glAPI->ProgramParameter4dvNV = __indirect_glProgramParameter4dvNV; - glAPI->ProgramParameter4fNV = __indirect_glProgramParameter4fNV; - glAPI->ProgramParameter4fvNV = __indirect_glProgramParameter4fvNV; glAPI->ProgramParameters4dvNV = __indirect_glProgramParameters4dvNV; glAPI->ProgramParameters4fvNV = __indirect_glProgramParameters4fvNV; glAPI->RequestResidentProgramsNV = __indirect_glRequestResidentProgramsNV; diff --git a/src/mesa/drivers/dri/common/extension_helper.h b/src/mesa/drivers/dri/common/extension_helper.h index 065c5d8dae..65e96657b8 100644 --- a/src/mesa/drivers/dri/common/extension_helper.h +++ b/src/mesa/drivers/dri/common/extension_helper.h @@ -40,13 +40,6 @@ static const char UniformMatrix3fvARB_names[] = ""; #endif -#if defined(need_GL_NV_vertex_program) -static const char ProgramParameter4fNV_names[] = - "iiffff\0" /* Parameter signature */ - "glProgramParameter4fNV\0" - ""; -#endif - #if defined(need_GL_VERSION_1_3) || defined(need_GL_ARB_multisample) static const char SampleCoverageARB_names[] = "fi\0" /* Parameter signature */ @@ -572,13 +565,6 @@ static const char MatrixIndexusvARB_names[] = ""; #endif -#if defined(need_GL_NV_vertex_program) -static const char ProgramParameter4dvNV_names[] = - "iip\0" /* Parameter signature */ - "glProgramParameter4dvNV\0" - ""; -#endif - #if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) static const char DisableVertexAttribArrayARB_names[] = "i\0" /* Parameter signature */ @@ -967,13 +953,6 @@ static const char GenerateMipmapEXT_names[] = ""; #endif -#if defined(need_GL_NV_vertex_program) -static const char ProgramParameter4dNV_names[] = - "iidddd\0" /* Parameter signature */ - "glProgramParameter4dNV\0" - ""; -#endif - #if defined(need_GL_ATI_fragment_shader) static const char SetFragmentShaderConstantATI_names[] = "ip\0" /* Parameter signature */ @@ -1311,10 +1290,11 @@ static const char Color3fVertex3fSUN_names[] = ""; #endif -#if defined(need_GL_ARB_vertex_program) +#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) static const char ProgramEnvParameter4fvARB_names[] = "iip\0" /* Parameter signature */ "glProgramEnvParameter4fvARB\0" + "glProgramParameter4fvNV\0" ""; #endif @@ -2035,13 +2015,6 @@ static const char WeightfvARB_names[] = ""; #endif -#if defined(need_GL_NV_vertex_program) -static const char ProgramParameter4fvNV_names[] = - "iip\0" /* Parameter signature */ - "glProgramParameter4fvNV\0" - ""; -#endif - #if defined(need_GL_MESA_window_pos) static const char WindowPos4fMESA_names[] = "ffff\0" /* Parameter signature */ @@ -2432,10 +2405,11 @@ static const char GetBufferPointervARB_names[] = ""; #endif -#if defined(need_GL_ARB_vertex_program) +#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) static const char ProgramEnvParameter4fARB_names[] = "iiffff\0" /* Parameter signature */ "glProgramEnvParameter4fARB\0" + "glProgramParameter4fNV\0" ""; #endif @@ -2803,10 +2777,11 @@ static const char ReplacementCodePointerSUN_names[] = ""; #endif -#if defined(need_GL_ARB_vertex_program) +#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) static const char ProgramEnvParameter4dARB_names[] = "iidddd\0" /* Parameter signature */ "glProgramEnvParameter4dARB\0" + "glProgramParameter4dNV\0" ""; #endif @@ -3660,10 +3635,11 @@ static const char GetColorTableParameteriv_names[] = ""; #endif -#if defined(need_GL_ARB_vertex_program) +#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program) static const char ProgramEnvParameter4dvARB_names[] = "iip\0" /* Parameter signature */ "glProgramEnvParameter4dvARB\0" + "glProgramParameter4dvNV\0" ""; #endif @@ -5748,12 +5724,10 @@ static const struct dri_extension_function GL_NV_vertex_array_range_functions[] #if defined(need_GL_NV_vertex_program) static const struct dri_extension_function GL_NV_vertex_program_functions[] = { - { ProgramParameter4fNV_names, ProgramParameter4fNV_remap_index, -1 }, { VertexAttrib4ubvNV_names, VertexAttrib4ubvNV_remap_index, -1 }, { VertexAttrib4svNV_names, VertexAttrib4svNV_remap_index, -1 }, { VertexAttribs1dvNV_names, VertexAttribs1dvNV_remap_index, -1 }, { VertexAttrib1fvNV_names, VertexAttrib1fvNV_remap_index, -1 }, - { ProgramParameter4dvNV_names, ProgramParameter4dvNV_remap_index, -1 }, { VertexAttrib4fNV_names, VertexAttrib4fNV_remap_index, -1 }, { VertexAttrib2dNV_names, VertexAttrib2dNV_remap_index, -1 }, { VertexAttrib4ubNV_names, VertexAttrib4ubNV_remap_index, -1 }, @@ -5761,7 +5735,7 @@ static const struct dri_extension_function GL_NV_vertex_program_functions[] = { { VertexAttribs4fvNV_names, VertexAttribs4fvNV_remap_index, -1 }, { VertexAttrib2sNV_names, VertexAttrib2sNV_remap_index, -1 }, { VertexAttribs3fvNV_names, VertexAttribs3fvNV_remap_index, -1 }, - { ProgramParameter4dNV_names, ProgramParameter4dNV_remap_index, -1 }, + { ProgramEnvParameter4fvARB_names, ProgramEnvParameter4fvARB_remap_index, -1 }, { LoadProgramNV_names, LoadProgramNV_remap_index, -1 }, { VertexAttrib4fvNV_names, VertexAttrib4fvNV_remap_index, -1 }, { VertexAttrib3fNV_names, VertexAttrib3fNV_remap_index, -1 }, @@ -5771,14 +5745,15 @@ static const struct dri_extension_function GL_NV_vertex_program_functions[] = { { VertexAttrib2fvNV_names, VertexAttrib2fvNV_remap_index, -1 }, { VertexAttrib2dvNV_names, VertexAttrib2dvNV_remap_index, -1 }, { VertexAttrib1dvNV_names, VertexAttrib1dvNV_remap_index, -1 }, - { ProgramParameter4fvNV_names, ProgramParameter4fvNV_remap_index, -1 }, { VertexAttrib1svNV_names, VertexAttrib1svNV_remap_index, -1 }, + { ProgramEnvParameter4fARB_names, ProgramEnvParameter4fARB_remap_index, -1 }, { VertexAttribs2svNV_names, VertexAttribs2svNV_remap_index, -1 }, { GetVertexAttribivNV_names, GetVertexAttribivNV_remap_index, -1 }, { GetVertexAttribfvNV_names, GetVertexAttribfvNV_remap_index, -1 }, { VertexAttrib2svNV_names, VertexAttrib2svNV_remap_index, -1 }, { VertexAttribs1fvNV_names, VertexAttribs1fvNV_remap_index, -1 }, { IsProgramNV_names, IsProgramNV_remap_index, -1 }, + { ProgramEnvParameter4dARB_names, ProgramEnvParameter4dARB_remap_index, -1 }, { VertexAttrib2fNV_names, VertexAttrib2fNV_remap_index, -1 }, { RequestResidentProgramsNV_names, RequestResidentProgramsNV_remap_index, -1 }, { ExecuteProgramNV_names, ExecuteProgramNV_remap_index, -1 }, @@ -5791,6 +5766,7 @@ static const struct dri_extension_function GL_NV_vertex_program_functions[] = { { GetProgramivNV_names, GetProgramivNV_remap_index, -1 }, { GetVertexAttribdvNV_names, GetVertexAttribdvNV_remap_index, -1 }, { VertexAttrib3fvNV_names, VertexAttrib3fvNV_remap_index, -1 }, + { ProgramEnvParameter4dvARB_names, ProgramEnvParameter4dvARB_remap_index, -1 }, { VertexAttribs2fvNV_names, VertexAttribs2fvNV_remap_index, -1 }, { DeleteProgramsNV_names, DeleteProgramsNV_remap_index, -1 }, { GetVertexAttribPointervNV_names, GetVertexAttribPointervNV_remap_index, -1 }, diff --git a/src/mesa/glapi/dispatch.h b/src/mesa/glapi/dispatch.h index db23e44396..7123156085 100644 --- a/src/mesa/glapi/dispatch.h +++ b/src/mesa/glapi/dispatch.h @@ -2077,18 +2077,6 @@ #define CALL_LoadProgramNV(disp, parameters) (*((disp)->LoadProgramNV)) parameters #define GET_LoadProgramNV(disp) ((disp)->LoadProgramNV) #define SET_LoadProgramNV(disp, fn) ((disp)->LoadProgramNV = fn) -#define CALL_ProgramParameter4dNV(disp, parameters) (*((disp)->ProgramParameter4dNV)) parameters -#define GET_ProgramParameter4dNV(disp) ((disp)->ProgramParameter4dNV) -#define SET_ProgramParameter4dNV(disp, fn) ((disp)->ProgramParameter4dNV = fn) -#define CALL_ProgramParameter4dvNV(disp, parameters) (*((disp)->ProgramParameter4dvNV)) parameters -#define GET_ProgramParameter4dvNV(disp) ((disp)->ProgramParameter4dvNV) -#define SET_ProgramParameter4dvNV(disp, fn) ((disp)->ProgramParameter4dvNV = fn) -#define CALL_ProgramParameter4fNV(disp, parameters) (*((disp)->ProgramParameter4fNV)) parameters -#define GET_ProgramParameter4fNV(disp) ((disp)->ProgramParameter4fNV) -#define SET_ProgramParameter4fNV(disp, fn) ((disp)->ProgramParameter4fNV = fn) -#define CALL_ProgramParameter4fvNV(disp, parameters) (*((disp)->ProgramParameter4fvNV)) parameters -#define GET_ProgramParameter4fvNV(disp) ((disp)->ProgramParameter4fvNV) -#define SET_ProgramParameter4fvNV(disp, fn) ((disp)->ProgramParameter4fvNV = fn) #define CALL_ProgramParameters4dvNV(disp, parameters) (*((disp)->ProgramParameters4dvNV)) parameters #define GET_ProgramParameters4dvNV(disp) ((disp)->ProgramParameters4dvNV) #define SET_ProgramParameters4dvNV(disp, fn) ((disp)->ProgramParameters4dvNV = fn) @@ -2383,7 +2371,7 @@ #else -#define driDispatchRemapTable_size 366 +#define driDispatchRemapTable_size 362 extern int driDispatchRemapTable[ driDispatchRemapTable_size ]; #define AttachShader_remap_index 0 @@ -2651,107 +2639,103 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ]; #define GetVertexAttribivNV_remap_index 262 #define IsProgramNV_remap_index 263 #define LoadProgramNV_remap_index 264 -#define ProgramParameter4dNV_remap_index 265 -#define ProgramParameter4dvNV_remap_index 266 -#define ProgramParameter4fNV_remap_index 267 -#define ProgramParameter4fvNV_remap_index 268 -#define ProgramParameters4dvNV_remap_index 269 -#define ProgramParameters4fvNV_remap_index 270 -#define RequestResidentProgramsNV_remap_index 271 -#define TrackMatrixNV_remap_index 272 -#define VertexAttrib1dNV_remap_index 273 -#define VertexAttrib1dvNV_remap_index 274 -#define VertexAttrib1fNV_remap_index 275 -#define VertexAttrib1fvNV_remap_index 276 -#define VertexAttrib1sNV_remap_index 277 -#define VertexAttrib1svNV_remap_index 278 -#define VertexAttrib2dNV_remap_index 279 -#define VertexAttrib2dvNV_remap_index 280 -#define VertexAttrib2fNV_remap_index 281 -#define VertexAttrib2fvNV_remap_index 282 -#define VertexAttrib2sNV_remap_index 283 -#define VertexAttrib2svNV_remap_index 284 -#define VertexAttrib3dNV_remap_index 285 -#define VertexAttrib3dvNV_remap_index 286 -#define VertexAttrib3fNV_remap_index 287 -#define VertexAttrib3fvNV_remap_index 288 -#define VertexAttrib3sNV_remap_index 289 -#define VertexAttrib3svNV_remap_index 290 -#define VertexAttrib4dNV_remap_index 291 -#define VertexAttrib4dvNV_remap_index 292 -#define VertexAttrib4fNV_remap_index 293 -#define VertexAttrib4fvNV_remap_index 294 -#define VertexAttrib4sNV_remap_index 295 -#define VertexAttrib4svNV_remap_index 296 -#define VertexAttrib4ubNV_remap_index 297 -#define VertexAttrib4ubvNV_remap_index 298 -#define VertexAttribPointerNV_remap_index 299 -#define VertexAttribs1dvNV_remap_index 300 -#define VertexAttribs1fvNV_remap_index 301 -#define VertexAttribs1svNV_remap_index 302 -#define VertexAttribs2dvNV_remap_index 303 -#define VertexAttribs2fvNV_remap_index 304 -#define VertexAttribs2svNV_remap_index 305 -#define VertexAttribs3dvNV_remap_index 306 -#define VertexAttribs3fvNV_remap_index 307 -#define VertexAttribs3svNV_remap_index 308 -#define VertexAttribs4dvNV_remap_index 309 -#define VertexAttribs4fvNV_remap_index 310 -#define VertexAttribs4svNV_remap_index 311 -#define VertexAttribs4ubvNV_remap_index 312 -#define AlphaFragmentOp1ATI_remap_index 313 -#define AlphaFragmentOp2ATI_remap_index 314 -#define AlphaFragmentOp3ATI_remap_index 315 -#define BeginFragmentShaderATI_remap_index 316 -#define BindFragmentShaderATI_remap_index 317 -#define ColorFragmentOp1ATI_remap_index 318 -#define ColorFragmentOp2ATI_remap_index 319 -#define ColorFragmentOp3ATI_remap_index 320 -#define DeleteFragmentShaderATI_remap_index 321 -#define EndFragmentShaderATI_remap_index 322 -#define GenFragmentShadersATI_remap_index 323 -#define PassTexCoordATI_remap_index 324 -#define SampleMapATI_remap_index 325 -#define SetFragmentShaderConstantATI_remap_index 326 -#define PointParameteriNV_remap_index 327 -#define PointParameterivNV_remap_index 328 -#define ActiveStencilFaceEXT_remap_index 329 -#define BindVertexArrayAPPLE_remap_index 330 -#define DeleteVertexArraysAPPLE_remap_index 331 -#define GenVertexArraysAPPLE_remap_index 332 -#define IsVertexArrayAPPLE_remap_index 333 -#define GetProgramNamedParameterdvNV_remap_index 334 -#define GetProgramNamedParameterfvNV_remap_index 335 -#define ProgramNamedParameter4dNV_remap_index 336 -#define ProgramNamedParameter4dvNV_remap_index 337 -#define ProgramNamedParameter4fNV_remap_index 338 -#define ProgramNamedParameter4fvNV_remap_index 339 -#define DepthBoundsEXT_remap_index 340 -#define BlendEquationSeparateEXT_remap_index 341 -#define BindFramebufferEXT_remap_index 342 -#define BindRenderbufferEXT_remap_index 343 -#define CheckFramebufferStatusEXT_remap_index 344 -#define DeleteFramebuffersEXT_remap_index 345 -#define DeleteRenderbuffersEXT_remap_index 346 -#define FramebufferRenderbufferEXT_remap_index 347 -#define FramebufferTexture1DEXT_remap_index 348 -#define FramebufferTexture2DEXT_remap_index 349 -#define FramebufferTexture3DEXT_remap_index 350 -#define GenFramebuffersEXT_remap_index 351 -#define GenRenderbuffersEXT_remap_index 352 -#define GenerateMipmapEXT_remap_index 353 -#define GetFramebufferAttachmentParameterivEXT_remap_index 354 -#define GetRenderbufferParameterivEXT_remap_index 355 -#define IsFramebufferEXT_remap_index 356 -#define IsRenderbufferEXT_remap_index 357 -#define RenderbufferStorageEXT_remap_index 358 -#define BlitFramebufferEXT_remap_index 359 -#define FramebufferTextureLayerEXT_remap_index 360 -#define StencilFuncSeparateATI_remap_index 361 -#define ProgramEnvParameters4fvEXT_remap_index 362 -#define ProgramLocalParameters4fvEXT_remap_index 363 -#define GetQueryObjecti64vEXT_remap_index 364 -#define GetQueryObjectui64vEXT_remap_index 365 +#define ProgramParameters4dvNV_remap_index 265 +#define ProgramParameters4fvNV_remap_index 266 +#define RequestResidentProgramsNV_remap_index 267 +#define TrackMatrixNV_remap_index 268 +#define VertexAttrib1dNV_remap_index 269 +#define VertexAttrib1dvNV_remap_index 270 +#define VertexAttrib1fNV_remap_index 271 +#define VertexAttrib1fvNV_remap_index 272 +#define VertexAttrib1sNV_remap_index 273 +#define VertexAttrib1svNV_remap_index 274 +#define VertexAttrib2dNV_remap_index 275 +#define VertexAttrib2dvNV_remap_index 276 +#define VertexAttrib2fNV_remap_index 277 +#define VertexAttrib2fvNV_remap_index 278 +#define VertexAttrib2sNV_remap_index 279 +#define VertexAttrib2svNV_remap_index 280 +#define VertexAttrib3dNV_remap_index 281 +#define VertexAttrib3dvNV_remap_index 282 +#define VertexAttrib3fNV_remap_index 283 +#define VertexAttrib3fvNV_remap_index 284 +#define VertexAttrib3sNV_remap_index 285 +#define VertexAttrib3svNV_remap_index 286 +#define VertexAttrib4dNV_remap_index 287 +#define VertexAttrib4dvNV_remap_index 288 +#define VertexAttrib4fNV_remap_index 289 +#define VertexAttrib4fvNV_remap_index 290 +#define VertexAttrib4sNV_remap_index 291 +#define VertexAttrib4svNV_remap_index 292 +#define VertexAttrib4ubNV_remap_index 293 +#define VertexAttrib4ubvNV_remap_index 294 +#define VertexAttribPointerNV_remap_index 295 +#define VertexAttribs1dvNV_remap_index 296 +#define VertexAttribs1fvNV_remap_index 297 +#define VertexAttribs1svNV_remap_index 298 +#define VertexAttribs2dvNV_remap_index 299 +#define VertexAttribs2fvNV_remap_index 300 +#define VertexAttribs2svNV_remap_index 301 +#define VertexAttribs3dvNV_remap_index 302 +#define VertexAttribs3fvNV_remap_index 303 +#define VertexAttribs3svNV_remap_index 304 +#define VertexAttribs4dvNV_remap_index 305 +#define VertexAttribs4fvNV_remap_index 306 +#define VertexAttribs4svNV_remap_index 307 +#define VertexAttribs4ubvNV_remap_index 308 +#define AlphaFragmentOp1ATI_remap_index 309 +#define AlphaFragmentOp2ATI_remap_index 310 +#define AlphaFragmentOp3ATI_remap_index 311 +#define BeginFragmentShaderATI_remap_index 312 +#define BindFragmentShaderATI_remap_index 313 +#define ColorFragmentOp1ATI_remap_index 314 +#define ColorFragmentOp2ATI_remap_index 315 +#define ColorFragmentOp3ATI_remap_index 316 +#define DeleteFragmentShaderATI_remap_index 317 +#define EndFragmentShaderATI_remap_index 318 +#define GenFragmentShadersATI_remap_index 319 +#define PassTexCoordATI_remap_index 320 +#define SampleMapATI_remap_index 321 +#define SetFragmentShaderConstantATI_remap_index 322 +#define PointParameteriNV_remap_index 323 +#define PointParameterivNV_remap_index 324 +#define ActiveStencilFaceEXT_remap_index 325 +#define BindVertexArrayAPPLE_remap_index 326 +#define DeleteVertexArraysAPPLE_remap_index 327 +#define GenVertexArraysAPPLE_remap_index 328 +#define IsVertexArrayAPPLE_remap_index 329 +#define GetProgramNamedParameterdvNV_remap_index 330 +#define GetProgramNamedParameterfvNV_remap_index 331 +#define ProgramNamedParameter4dNV_remap_index 332 +#define ProgramNamedParameter4dvNV_remap_index 333 +#define ProgramNamedParameter4fNV_remap_index 334 +#define ProgramNamedParameter4fvNV_remap_index 335 +#define DepthBoundsEXT_remap_index 336 +#define BlendEquationSeparateEXT_remap_index 337 +#define BindFramebufferEXT_remap_index 338 +#define BindRenderbufferEXT_remap_index 339 +#define CheckFramebufferStatusEXT_remap_index 340 +#define DeleteFramebuffersEXT_remap_index 341 +#define DeleteRenderbuffersEXT_remap_index 342 +#define FramebufferRenderbufferEXT_remap_index 343 +#define FramebufferTexture1DEXT_remap_index 344 +#define FramebufferTexture2DEXT_remap_index 345 +#define FramebufferTexture3DEXT_remap_index 346 +#define GenFramebuffersEXT_remap_index 347 +#define GenRenderbuffersEXT_remap_index 348 +#define GenerateMipmapEXT_remap_index 349 +#define GetFramebufferAttachmentParameterivEXT_remap_index 350 +#define GetRenderbufferParameterivEXT_remap_index 351 +#define IsFramebufferEXT_remap_index 352 +#define IsRenderbufferEXT_remap_index 353 +#define RenderbufferStorageEXT_remap_index 354 +#define BlitFramebufferEXT_remap_index 355 +#define FramebufferTextureLayerEXT_remap_index 356 +#define StencilFuncSeparateATI_remap_index 357 +#define ProgramEnvParameters4fvEXT_remap_index 358 +#define ProgramLocalParameters4fvEXT_remap_index 359 +#define GetQueryObjecti64vEXT_remap_index 360 +#define GetQueryObjectui64vEXT_remap_index 361 #define CALL_AttachShader(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLuint, GLuint)), driDispatchRemapTable[AttachShader_remap_index], parameters) #define GET_AttachShader(disp) GET_by_offset(disp, driDispatchRemapTable[AttachShader_remap_index]) @@ -3548,18 +3532,6 @@ extern int driDispatchRemapTable[ driDispatchRemapTable_size ]; #define CALL_LoadProgramNV(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLuint, GLsizei, const GLubyte *)), driDispatchRemapTable[LoadProgramNV_remap_index], parameters) #define GET_LoadProgramNV(disp) GET_by_offset(disp, driDispatchRemapTable[LoadProgramNV_remap_index]) #define SET_LoadProgramNV(disp, fn) SET_by_offset(disp, driDispatchRemapTable[LoadProgramNV_remap_index], fn) -#define CALL_ProgramParameter4dNV(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLuint, GLdouble, GLdouble, GLdouble, GLdouble)), driDispatchRemapTable[ProgramParameter4dNV_remap_index], parameters) -#define GET_ProgramParameter4dNV(disp) GET_by_offset(disp, driDispatchRemapTable[ProgramParameter4dNV_remap_index]) -#define SET_ProgramParameter4dNV(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ProgramParameter4dNV_remap_index], fn) -#define CALL_ProgramParameter4dvNV(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLuint, const GLdouble *)), driDispatchRemapTable[ProgramParameter4dvNV_remap_index], parameters) -#define GET_ProgramParameter4dvNV(disp) GET_by_offset(disp, driDispatchRemapTable[ProgramParameter4dvNV_remap_index]) -#define SET_ProgramParameter4dvNV(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ProgramParameter4dvNV_remap_index], fn) -#define CALL_ProgramParameter4fNV(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLuint, GLfloat, GLfloat, GLfloat, GLfloat)), driDispatchRemapTable[ProgramParameter4fNV_remap_index], parameters) -#define GET_ProgramParameter4fNV(disp) GET_by_offset(disp, driDispatchRemapTable[ProgramParameter4fNV_remap_index]) -#define SET_ProgramParameter4fNV(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ProgramParameter4fNV_remap_index], fn) -#define CALL_ProgramParameter4fvNV(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLuint, const GLfloat *)), driDispatchRemapTable[ProgramParameter4fvNV_remap_index], parameters) -#define GET_ProgramParameter4fvNV(disp) GET_by_offset(disp, driDispatchRemapTable[ProgramParameter4fvNV_remap_index]) -#define SET_ProgramParameter4fvNV(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ProgramParameter4fvNV_remap_index], fn) #define CALL_ProgramParameters4dvNV(disp, parameters) CALL_by_offset(disp, (void (GLAPIENTRYP)(GLenum, GLuint, GLuint, const GLdouble *)), driDispatchRemapTable[ProgramParameters4dvNV_remap_index], parameters) #define GET_ProgramParameters4dvNV(disp) GET_by_offset(disp, driDispatchRemapTable[ProgramParameters4dvNV_remap_index]) #define SET_ProgramParameters4dvNV(disp, fn) SET_by_offset(disp, driDispatchRemapTable[ProgramParameters4dvNV_remap_index], fn) diff --git a/src/mesa/glapi/glapioffsets.h b/src/mesa/glapi/glapioffsets.h index 4799fd3076..43c265924b 100644 --- a/src/mesa/glapi/glapioffsets.h +++ b/src/mesa/glapi/glapioffsets.h @@ -705,108 +705,104 @@ #define _gloffset_GetVertexAttribivNV 670 #define _gloffset_IsProgramNV 671 #define _gloffset_LoadProgramNV 672 -#define _gloffset_ProgramParameter4dNV 673 -#define _gloffset_ProgramParameter4dvNV 674 -#define _gloffset_ProgramParameter4fNV 675 -#define _gloffset_ProgramParameter4fvNV 676 -#define _gloffset_ProgramParameters4dvNV 677 -#define _gloffset_ProgramParameters4fvNV 678 -#define _gloffset_RequestResidentProgramsNV 679 -#define _gloffset_TrackMatrixNV 680 -#define _gloffset_VertexAttrib1dNV 681 -#define _gloffset_VertexAttrib1dvNV 682 -#define _gloffset_VertexAttrib1fNV 683 -#define _gloffset_VertexAttrib1fvNV 684 -#define _gloffset_VertexAttrib1sNV 685 -#define _gloffset_VertexAttrib1svNV 686 -#define _gloffset_VertexAttrib2dNV 687 -#define _gloffset_VertexAttrib2dvNV 688 -#define _gloffset_VertexAttrib2fNV 689 -#define _gloffset_VertexAttrib2fvNV 690 -#define _gloffset_VertexAttrib2sNV 691 -#define _gloffset_VertexAttrib2svNV 692 -#define _gloffset_VertexAttrib3dNV 693 -#define _gloffset_VertexAttrib3dvNV 694 -#define _gloffset_VertexAttrib3fNV 695 -#define _gloffset_VertexAttrib3fvNV 696 -#define _gloffset_VertexAttrib3sNV 697 -#define _gloffset_VertexAttrib3svNV 698 -#define _gloffset_VertexAttrib4dNV 699 -#define _gloffset_VertexAttrib4dvNV 700 -#define _gloffset_VertexAttrib4fNV 701 -#define _gloffset_VertexAttrib4fvNV 702 -#define _gloffset_VertexAttrib4sNV 703 -#define _gloffset_VertexAttrib4svNV 704 -#define _gloffset_VertexAttrib4ubNV 705 -#define _gloffset_VertexAttrib4ubvNV 706 -#define _gloffset_VertexAttribPointerNV 707 -#define _gloffset_VertexAttribs1dvNV 708 -#define _gloffset_VertexAttribs1fvNV 709 -#define _gloffset_VertexAttribs1svNV 710 -#define _gloffset_VertexAttribs2dvNV 711 -#define _gloffset_VertexAttribs2fvNV 712 -#define _gloffset_VertexAttribs2svNV 713 -#define _gloffset_VertexAttribs3dvNV 714 -#define _gloffset_VertexAttribs3fvNV 715 -#define _gloffset_VertexAttribs3svNV 716 -#define _gloffset_VertexAttribs4dvNV 717 -#define _gloffset_VertexAttribs4fvNV 718 -#define _gloffset_VertexAttribs4svNV 719 -#define _gloffset_VertexAttribs4ubvNV 720 -#define _gloffset_AlphaFragmentOp1ATI 721 -#define _gloffset_AlphaFragmentOp2ATI 722 -#define _gloffset_AlphaFragmentOp3ATI 723 -#define _gloffset_BeginFragmentShaderATI 724 -#define _gloffset_BindFragmentShaderATI 725 -#define _gloffset_ColorFragmentOp1ATI 726 -#define _gloffset_ColorFragmentOp2ATI 727 -#define _gloffset_ColorFragmentOp3ATI 728 -#define _gloffset_DeleteFragmentShaderATI 729 -#define _gloffset_EndFragmentShaderATI 730 -#define _gloffset_GenFragmentShadersATI 731 -#define _gloffset_PassTexCoordATI 732 -#define _gloffset_SampleMapATI 733 -#define _gloffset_SetFragmentShaderConstantATI 734 -#define _gloffset_PointParameteriNV 735 -#define _gloffset_PointParameterivNV 736 -#define _gloffset_ActiveStencilFaceEXT 737 -#define _gloffset_BindVertexArrayAPPLE 738 -#define _gloffset_DeleteVertexArraysAPPLE 739 -#define _gloffset_GenVertexArraysAPPLE 740 -#define _gloffset_IsVertexArrayAPPLE 741 -#define _gloffset_GetProgramNamedParameterdvNV 742 -#define _gloffset_GetProgramNamedParameterfvNV 743 -#define _gloffset_ProgramNamedParameter4dNV 744 -#define _gloffset_ProgramNamedParameter4dvNV 745 -#define _gloffset_ProgramNamedParameter4fNV 746 -#define _gloffset_ProgramNamedParameter4fvNV 747 -#define _gloffset_DepthBoundsEXT 748 -#define _gloffset_BlendEquationSeparateEXT 749 -#define _gloffset_BindFramebufferEXT 750 -#define _gloffset_BindRenderbufferEXT 751 -#define _gloffset_CheckFramebufferStatusEXT 752 -#define _gloffset_DeleteFramebuffersEXT 753 -#define _gloffset_DeleteRenderbuffersEXT 754 -#define _gloffset_FramebufferRenderbufferEXT 755 -#define _gloffset_FramebufferTexture1DEXT 756 -#define _gloffset_FramebufferTexture2DEXT 757 -#define _gloffset_FramebufferTexture3DEXT 758 -#define _gloffset_GenFramebuffersEXT 759 -#define _gloffset_GenRenderbuffersEXT 760 -#define _gloffset_GenerateMipmapEXT 761 -#define _gloffset_GetFramebufferAttachmentParameterivEXT 762 -#define _gloffset_GetRenderbufferParameterivEXT 763 -#define _gloffset_IsFramebufferEXT 764 -#define _gloffset_IsRenderbufferEXT 765 -#define _gloffset_RenderbufferStorageEXT 766 -#define _gloffset_BlitFramebufferEXT 767 -#define _gloffset_FramebufferTextureLayerEXT 768 -#define _gloffset_StencilFuncSeparateATI 769 -#define _gloffset_ProgramEnvParameters4fvEXT 770 -#define _gloffset_ProgramLocalParameters4fvEXT 771 -#define _gloffset_GetQueryObjecti64vEXT 772 -#define _gloffset_GetQueryObjectui64vEXT 773 -#define _gloffset_FIRST_DYNAMIC 774 +#define _gloffset_ProgramParameters4dvNV 673 +#define _gloffset_ProgramParameters4fvNV 674 +#define _gloffset_RequestResidentProgramsNV 675 +#define _gloffset_TrackMatrixNV 676 +#define _gloffset_VertexAttrib1dNV 677 +#define _gloffset_VertexAttrib1dvNV 678 +#define _gloffset_VertexAttrib1fNV 679 +#define _gloffset_VertexAttrib1fvNV 680 +#define _gloffset_VertexAttrib1sNV 681 +#define _gloffset_VertexAttrib1svNV 682 +#define _gloffset_VertexAttrib2dNV 683 +#define _gloffset_VertexAttrib2dvNV 684 +#define _gloffset_VertexAttrib2fNV 685 +#define _gloffset_VertexAttrib2fvNV 686 +#define _gloffset_VertexAttrib2sNV 687 +#define _gloffset_VertexAttrib2svNV 688 +#define _gloffset_VertexAttrib3dNV 689 +#define _gloffset_VertexAttrib3dvNV 690 +#define _gloffset_VertexAttrib3fNV 691 +#define _gloffset_VertexAttrib3fvNV 692 +#define _gloffset_VertexAttrib3sNV 693 +#define _gloffset_VertexAttrib3svNV 694 +#define _gloffset_VertexAttrib4dNV 695 +#define _gloffset_VertexAttrib4dvNV 696 +#define _gloffset_VertexAttrib4fNV 697 +#define _gloffset_VertexAttrib4fvNV 698 +#define _gloffset_VertexAttrib4sNV 699 +#define _gloffset_VertexAttrib4svNV 700 +#define _gloffset_VertexAttrib4ubNV 701 +#define _gloffset_VertexAttrib4ubvNV 702 +#define _gloffset_VertexAttribPointerNV 703 +#define _gloffset_VertexAttribs1dvNV 704 +#define _gloffset_VertexAttribs1fvNV 705 +#define _gloffset_VertexAttribs1svNV 706 +#define _gloffset_VertexAttribs2dvNV 707 +#define _gloffset_VertexAttribs2fvNV 708 +#define _gloffset_VertexAttribs2svNV 709 +#define _gloffset_VertexAttribs3dvNV 710 +#define _gloffset_VertexAttribs3fvNV 711 +#define _gloffset_VertexAttribs3svNV 712 +#define _gloffset_VertexAttribs4dvNV 713 +#define _gloffset_VertexAttribs4fvNV 714 +#define _gloffset_VertexAttribs4svNV 715 +#define _gloffset_VertexAttribs4ubvNV 716 +#define _gloffset_AlphaFragmentOp1ATI 717 +#define _gloffset_AlphaFragmentOp2ATI 718 +#define _gloffset_AlphaFragmentOp3ATI 719 +#define _gloffset_BeginFragmentShaderATI 720 +#define _gloffset_BindFragmentShaderATI 721 +#define _gloffset_ColorFragmentOp1ATI 722 +#define _gloffset_ColorFragmentOp2ATI 723 +#define _gloffset_ColorFragmentOp3ATI 724 +#define _gloffset_DeleteFragmentShaderATI 725 +#define _gloffset_EndFragmentShaderATI 726 +#define _gloffset_GenFragmentShadersATI 727 +#define _gloffset_PassTexCoordATI 728 +#define _gloffset_SampleMapATI 729 +#define _gloffset_SetFragmentShaderConstantATI 730 +#define _gloffset_PointParameteriNV 731 +#define _gloffset_PointParameterivNV 732 +#define _gloffset_ActiveStencilFaceEXT 733 +#define _gloffset_BindVertexArrayAPPLE 734 +#define _gloffset_DeleteVertexArraysAPPLE 735 +#define _gloffset_GenVertexArraysAPPLE 736 +#define _gloffset_IsVertexArrayAPPLE 737 +#define _gloffset_GetProgramNamedParameterdvNV 738 +#define _gloffset_GetProgramNamedParameterfvNV 739 +#define _gloffset_ProgramNamedParameter4dNV 740 +#define _gloffset_ProgramNamedParameter4dvNV 741 +#define _gloffset_ProgramNamedParameter4fNV 742 +#define _gloffset_ProgramNamedParameter4fvNV 743 +#define _gloffset_DepthBoundsEXT 744 +#define _gloffset_BlendEquationSeparateEXT 745 +#define _gloffset_BindFramebufferEXT 746 +#define _gloffset_BindRenderbufferEXT 747 +#define _gloffset_CheckFramebufferStatusEXT 748 +#define _gloffset_DeleteFramebuffersEXT 749 +#define _gloffset_DeleteRenderbuffersEXT 750 +#define _gloffset_FramebufferRenderbufferEXT 751 +#define _gloffset_FramebufferTexture1DEXT 752 +#define _gloffset_FramebufferTexture2DEXT 753 +#define _gloffset_FramebufferTexture3DEXT 754 +#define _gloffset_GenFramebuffersEXT 755 +#define _gloffset_GenRenderbuffersEXT 756 +#define _gloffset_GenerateMipmapEXT 757 +#define _gloffset_GetFramebufferAttachmentParameterivEXT 758 +#define _gloffset_GetRenderbufferParameterivEXT 759 +#define _gloffset_IsFramebufferEXT 760 +#define _gloffset_IsRenderbufferEXT 761 +#define _gloffset_RenderbufferStorageEXT 762 +#define _gloffset_BlitFramebufferEXT 763 +#define _gloffset_FramebufferTextureLayerEXT 764 +#define _gloffset_StencilFuncSeparateATI 765 +#define _gloffset_ProgramEnvParameters4fvEXT 766 +#define _gloffset_ProgramLocalParameters4fvEXT 767 +#define _gloffset_GetQueryObjecti64vEXT 768 +#define _gloffset_GetQueryObjectui64vEXT 769 +#define _gloffset_FIRST_DYNAMIC 770 #else @@ -1075,10 +1071,6 @@ #define _gloffset_GetVertexAttribivNV driDispatchRemapTable[GetVertexAttribivNV_remap_index] #define _gloffset_IsProgramNV driDispatchRemapTable[IsProgramNV_remap_index] #define _gloffset_LoadProgramNV driDispatchRemapTable[LoadProgramNV_remap_index] -#define _gloffset_ProgramParameter4dNV driDispatchRemapTable[ProgramParameter4dNV_remap_index] -#define _gloffset_ProgramParameter4dvNV driDispatchRemapTable[ProgramParameter4dvNV_remap_index] -#define _gloffset_ProgramParameter4fNV driDispatchRemapTable[ProgramParameter4fNV_remap_index] -#define _gloffset_ProgramParameter4fvNV driDispatchRemapTable[ProgramParameter4fvNV_remap_index] #define _gloffset_ProgramParameters4dvNV driDispatchRemapTable[ProgramParameters4dvNV_remap_index] #define _gloffset_ProgramParameters4fvNV driDispatchRemapTable[ProgramParameters4fvNV_remap_index] #define _gloffset_RequestResidentProgramsNV driDispatchRemapTable[RequestResidentProgramsNV_remap_index] diff --git a/src/mesa/glapi/glapitable.h b/src/mesa/glapi/glapitable.h index 22c2dc2f69..48941f5590 100644 --- a/src/mesa/glapi/glapitable.h +++ b/src/mesa/glapi/glapitable.h @@ -714,107 +714,103 @@ struct _glapi_table void (GLAPIENTRYP GetVertexAttribivNV)(GLuint index, GLenum pname, GLint * params); /* 670 */ GLboolean (GLAPIENTRYP IsProgramNV)(GLuint program); /* 671 */ void (GLAPIENTRYP LoadProgramNV)(GLenum target, GLuint id, GLsizei len, const GLubyte * program); /* 672 */ - void (GLAPIENTRYP ProgramParameter4dNV)(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 673 */ - void (GLAPIENTRYP ProgramParameter4dvNV)(GLenum target, GLuint index, const GLdouble * params); /* 674 */ - void (GLAPIENTRYP ProgramParameter4fNV)(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 675 */ - void (GLAPIENTRYP ProgramParameter4fvNV)(GLenum target, GLuint index, const GLfloat * params); /* 676 */ - void (GLAPIENTRYP ProgramParameters4dvNV)(GLenum target, GLuint index, GLuint num, const GLdouble * params); /* 677 */ - void (GLAPIENTRYP ProgramParameters4fvNV)(GLenum target, GLuint index, GLuint num, const GLfloat * params); /* 678 */ - void (GLAPIENTRYP RequestResidentProgramsNV)(GLsizei n, const GLuint * ids); /* 679 */ - void (GLAPIENTRYP TrackMatrixNV)(GLenum target, GLuint address, GLenum matrix, GLenum transform); /* 680 */ - void (GLAPIENTRYP VertexAttrib1dNV)(GLuint index, GLdouble x); /* 681 */ - void (GLAPIENTRYP VertexAttrib1dvNV)(GLuint index, const GLdouble * v); /* 682 */ - void (GLAPIENTRYP VertexAttrib1fNV)(GLuint index, GLfloat x); /* 683 */ - void (GLAPIENTRYP VertexAttrib1fvNV)(GLuint index, const GLfloat * v); /* 684 */ - void (GLAPIENTRYP VertexAttrib1sNV)(GLuint index, GLshort x); /* 685 */ - void (GLAPIENTRYP VertexAttrib1svNV)(GLuint index, const GLshort * v); /* 686 */ - void (GLAPIENTRYP VertexAttrib2dNV)(GLuint index, GLdouble x, GLdouble y); /* 687 */ - void (GLAPIENTRYP VertexAttrib2dvNV)(GLuint index, const GLdouble * v); /* 688 */ - void (GLAPIENTRYP VertexAttrib2fNV)(GLuint index, GLfloat x, GLfloat y); /* 689 */ - void (GLAPIENTRYP VertexAttrib2fvNV)(GLuint index, const GLfloat * v); /* 690 */ - void (GLAPIENTRYP VertexAttrib2sNV)(GLuint index, GLshort x, GLshort y); /* 691 */ - void (GLAPIENTRYP VertexAttrib2svNV)(GLuint index, const GLshort * v); /* 692 */ - void (GLAPIENTRYP VertexAttrib3dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z); /* 693 */ - void (GLAPIENTRYP VertexAttrib3dvNV)(GLuint index, const GLdouble * v); /* 694 */ - void (GLAPIENTRYP VertexAttrib3fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z); /* 695 */ - void (GLAPIENTRYP VertexAttrib3fvNV)(GLuint index, const GLfloat * v); /* 696 */ - void (GLAPIENTRYP VertexAttrib3sNV)(GLuint index, GLshort x, GLshort y, GLshort z); /* 697 */ - void (GLAPIENTRYP VertexAttrib3svNV)(GLuint index, const GLshort * v); /* 698 */ - void (GLAPIENTRYP VertexAttrib4dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 699 */ - void (GLAPIENTRYP VertexAttrib4dvNV)(GLuint index, const GLdouble * v); /* 700 */ - void (GLAPIENTRYP VertexAttrib4fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 701 */ - void (GLAPIENTRYP VertexAttrib4fvNV)(GLuint index, const GLfloat * v); /* 702 */ - void (GLAPIENTRYP VertexAttrib4sNV)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w); /* 703 */ - void (GLAPIENTRYP VertexAttrib4svNV)(GLuint index, const GLshort * v); /* 704 */ - void (GLAPIENTRYP VertexAttrib4ubNV)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w); /* 705 */ - void (GLAPIENTRYP VertexAttrib4ubvNV)(GLuint index, const GLubyte * v); /* 706 */ - void (GLAPIENTRYP VertexAttribPointerNV)(GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 707 */ - void (GLAPIENTRYP VertexAttribs1dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 708 */ - void (GLAPIENTRYP VertexAttribs1fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 709 */ - void (GLAPIENTRYP VertexAttribs1svNV)(GLuint index, GLsizei n, const GLshort * v); /* 710 */ - void (GLAPIENTRYP VertexAttribs2dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 711 */ - void (GLAPIENTRYP VertexAttribs2fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 712 */ - void (GLAPIENTRYP VertexAttribs2svNV)(GLuint index, GLsizei n, const GLshort * v); /* 713 */ - void (GLAPIENTRYP VertexAttribs3dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 714 */ - void (GLAPIENTRYP VertexAttribs3fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 715 */ - void (GLAPIENTRYP VertexAttribs3svNV)(GLuint index, GLsizei n, const GLshort * v); /* 716 */ - void (GLAPIENTRYP VertexAttribs4dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 717 */ - void (GLAPIENTRYP VertexAttribs4fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 718 */ - void (GLAPIENTRYP VertexAttribs4svNV)(GLuint index, GLsizei n, const GLshort * v); /* 719 */ - void (GLAPIENTRYP VertexAttribs4ubvNV)(GLuint index, GLsizei n, const GLubyte * v); /* 720 */ - void (GLAPIENTRYP AlphaFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 721 */ - void (GLAPIENTRYP AlphaFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 722 */ - void (GLAPIENTRYP AlphaFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 723 */ - void (GLAPIENTRYP BeginFragmentShaderATI)(void); /* 724 */ - void (GLAPIENTRYP BindFragmentShaderATI)(GLuint id); /* 725 */ - void (GLAPIENTRYP ColorFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 726 */ - void (GLAPIENTRYP ColorFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 727 */ - void (GLAPIENTRYP ColorFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 728 */ - void (GLAPIENTRYP DeleteFragmentShaderATI)(GLuint id); /* 729 */ - void (GLAPIENTRYP EndFragmentShaderATI)(void); /* 730 */ - GLuint (GLAPIENTRYP GenFragmentShadersATI)(GLuint range); /* 731 */ - void (GLAPIENTRYP PassTexCoordATI)(GLuint dst, GLuint coord, GLenum swizzle); /* 732 */ - void (GLAPIENTRYP SampleMapATI)(GLuint dst, GLuint interp, GLenum swizzle); /* 733 */ - void (GLAPIENTRYP SetFragmentShaderConstantATI)(GLuint dst, const GLfloat * value); /* 734 */ - void (GLAPIENTRYP PointParameteriNV)(GLenum pname, GLint param); /* 735 */ - void (GLAPIENTRYP PointParameterivNV)(GLenum pname, const GLint * params); /* 736 */ - void (GLAPIENTRYP ActiveStencilFaceEXT)(GLenum face); /* 737 */ - void (GLAPIENTRYP BindVertexArrayAPPLE)(GLuint array); /* 738 */ - void (GLAPIENTRYP DeleteVertexArraysAPPLE)(GLsizei n, const GLuint * arrays); /* 739 */ - void (GLAPIENTRYP GenVertexArraysAPPLE)(GLsizei n, GLuint * arrays); /* 740 */ - GLboolean (GLAPIENTRYP IsVertexArrayAPPLE)(GLuint array); /* 741 */ - void (GLAPIENTRYP GetProgramNamedParameterdvNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble * params); /* 742 */ - void (GLAPIENTRYP GetProgramNamedParameterfvNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat * params); /* 743 */ - void (GLAPIENTRYP ProgramNamedParameter4dNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 744 */ - void (GLAPIENTRYP ProgramNamedParameter4dvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLdouble * v); /* 745 */ - void (GLAPIENTRYP ProgramNamedParameter4fNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 746 */ - void (GLAPIENTRYP ProgramNamedParameter4fvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLfloat * v); /* 747 */ - void (GLAPIENTRYP DepthBoundsEXT)(GLclampd zmin, GLclampd zmax); /* 748 */ - void (GLAPIENTRYP BlendEquationSeparateEXT)(GLenum modeRGB, GLenum modeA); /* 749 */ - void (GLAPIENTRYP BindFramebufferEXT)(GLenum target, GLuint framebuffer); /* 750 */ - void (GLAPIENTRYP BindRenderbufferEXT)(GLenum target, GLuint renderbuffer); /* 751 */ - GLenum (GLAPIENTRYP CheckFramebufferStatusEXT)(GLenum target); /* 752 */ - void (GLAPIENTRYP DeleteFramebuffersEXT)(GLsizei n, const GLuint * framebuffers); /* 753 */ - void (GLAPIENTRYP DeleteRenderbuffersEXT)(GLsizei n, const GLuint * renderbuffers); /* 754 */ - void (GLAPIENTRYP FramebufferRenderbufferEXT)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer); /* 755 */ - void (GLAPIENTRYP FramebufferTexture1DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 756 */ - void (GLAPIENTRYP FramebufferTexture2DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 757 */ - void (GLAPIENTRYP FramebufferTexture3DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset); /* 758 */ - void (GLAPIENTRYP GenFramebuffersEXT)(GLsizei n, GLuint * framebuffers); /* 759 */ - void (GLAPIENTRYP GenRenderbuffersEXT)(GLsizei n, GLuint * renderbuffers); /* 760 */ - void (GLAPIENTRYP GenerateMipmapEXT)(GLenum target); /* 761 */ - void (GLAPIENTRYP GetFramebufferAttachmentParameterivEXT)(GLenum target, GLenum attachment, GLenum pname, GLint * params); /* 762 */ - void (GLAPIENTRYP GetRenderbufferParameterivEXT)(GLenum target, GLenum pname, GLint * params); /* 763 */ - GLboolean (GLAPIENTRYP IsFramebufferEXT)(GLuint framebuffer); /* 764 */ - GLboolean (GLAPIENTRYP IsRenderbufferEXT)(GLuint renderbuffer); /* 765 */ - void (GLAPIENTRYP RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height); /* 766 */ - void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 767 */ - void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 768 */ - void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 769 */ - void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 770 */ - void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 771 */ - void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 772 */ - void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 773 */ + void (GLAPIENTRYP ProgramParameters4dvNV)(GLenum target, GLuint index, GLuint num, const GLdouble * params); /* 673 */ + void (GLAPIENTRYP ProgramParameters4fvNV)(GLenum target, GLuint index, GLuint num, const GLfloat * params); /* 674 */ + void (GLAPIENTRYP RequestResidentProgramsNV)(GLsizei n, const GLuint * ids); /* 675 */ + void (GLAPIENTRYP TrackMatrixNV)(GLenum target, GLuint address, GLenum matrix, GLenum transform); /* 676 */ + void (GLAPIENTRYP VertexAttrib1dNV)(GLuint index, GLdouble x); /* 677 */ + void (GLAPIENTRYP VertexAttrib1dvNV)(GLuint index, const GLdouble * v); /* 678 */ + void (GLAPIENTRYP VertexAttrib1fNV)(GLuint index, GLfloat x); /* 679 */ + void (GLAPIENTRYP VertexAttrib1fvNV)(GLuint index, const GLfloat * v); /* 680 */ + void (GLAPIENTRYP VertexAttrib1sNV)(GLuint index, GLshort x); /* 681 */ + void (GLAPIENTRYP VertexAttrib1svNV)(GLuint index, const GLshort * v); /* 682 */ + void (GLAPIENTRYP VertexAttrib2dNV)(GLuint index, GLdouble x, GLdouble y); /* 683 */ + void (GLAPIENTRYP VertexAttrib2dvNV)(GLuint index, const GLdouble * v); /* 684 */ + void (GLAPIENTRYP VertexAttrib2fNV)(GLuint index, GLfloat x, GLfloat y); /* 685 */ + void (GLAPIENTRYP VertexAttrib2fvNV)(GLuint index, const GLfloat * v); /* 686 */ + void (GLAPIENTRYP VertexAttrib2sNV)(GLuint index, GLshort x, GLshort y); /* 687 */ + void (GLAPIENTRYP VertexAttrib2svNV)(GLuint index, const GLshort * v); /* 688 */ + void (GLAPIENTRYP VertexAttrib3dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z); /* 689 */ + void (GLAPIENTRYP VertexAttrib3dvNV)(GLuint index, const GLdouble * v); /* 690 */ + void (GLAPIENTRYP VertexAttrib3fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z); /* 691 */ + void (GLAPIENTRYP VertexAttrib3fvNV)(GLuint index, const GLfloat * v); /* 692 */ + void (GLAPIENTRYP VertexAttrib3sNV)(GLuint index, GLshort x, GLshort y, GLshort z); /* 693 */ + void (GLAPIENTRYP VertexAttrib3svNV)(GLuint index, const GLshort * v); /* 694 */ + void (GLAPIENTRYP VertexAttrib4dNV)(GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 695 */ + void (GLAPIENTRYP VertexAttrib4dvNV)(GLuint index, const GLdouble * v); /* 696 */ + void (GLAPIENTRYP VertexAttrib4fNV)(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 697 */ + void (GLAPIENTRYP VertexAttrib4fvNV)(GLuint index, const GLfloat * v); /* 698 */ + void (GLAPIENTRYP VertexAttrib4sNV)(GLuint index, GLshort x, GLshort y, GLshort z, GLshort w); /* 699 */ + void (GLAPIENTRYP VertexAttrib4svNV)(GLuint index, const GLshort * v); /* 700 */ + void (GLAPIENTRYP VertexAttrib4ubNV)(GLuint index, GLubyte x, GLubyte y, GLubyte z, GLubyte w); /* 701 */ + void (GLAPIENTRYP VertexAttrib4ubvNV)(GLuint index, const GLubyte * v); /* 702 */ + void (GLAPIENTRYP VertexAttribPointerNV)(GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid * pointer); /* 703 */ + void (GLAPIENTRYP VertexAttribs1dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 704 */ + void (GLAPIENTRYP VertexAttribs1fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 705 */ + void (GLAPIENTRYP VertexAttribs1svNV)(GLuint index, GLsizei n, const GLshort * v); /* 706 */ + void (GLAPIENTRYP VertexAttribs2dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 707 */ + void (GLAPIENTRYP VertexAttribs2fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 708 */ + void (GLAPIENTRYP VertexAttribs2svNV)(GLuint index, GLsizei n, const GLshort * v); /* 709 */ + void (GLAPIENTRYP VertexAttribs3dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 710 */ + void (GLAPIENTRYP VertexAttribs3fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 711 */ + void (GLAPIENTRYP VertexAttribs3svNV)(GLuint index, GLsizei n, const GLshort * v); /* 712 */ + void (GLAPIENTRYP VertexAttribs4dvNV)(GLuint index, GLsizei n, const GLdouble * v); /* 713 */ + void (GLAPIENTRYP VertexAttribs4fvNV)(GLuint index, GLsizei n, const GLfloat * v); /* 714 */ + void (GLAPIENTRYP VertexAttribs4svNV)(GLuint index, GLsizei n, const GLshort * v); /* 715 */ + void (GLAPIENTRYP VertexAttribs4ubvNV)(GLuint index, GLsizei n, const GLubyte * v); /* 716 */ + void (GLAPIENTRYP AlphaFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 717 */ + void (GLAPIENTRYP AlphaFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 718 */ + void (GLAPIENTRYP AlphaFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 719 */ + void (GLAPIENTRYP BeginFragmentShaderATI)(void); /* 720 */ + void (GLAPIENTRYP BindFragmentShaderATI)(GLuint id); /* 721 */ + void (GLAPIENTRYP ColorFragmentOp1ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod); /* 722 */ + void (GLAPIENTRYP ColorFragmentOp2ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod); /* 723 */ + void (GLAPIENTRYP ColorFragmentOp3ATI)(GLenum op, GLuint dst, GLuint dstMask, GLuint dstMod, GLuint arg1, GLuint arg1Rep, GLuint arg1Mod, GLuint arg2, GLuint arg2Rep, GLuint arg2Mod, GLuint arg3, GLuint arg3Rep, GLuint arg3Mod); /* 724 */ + void (GLAPIENTRYP DeleteFragmentShaderATI)(GLuint id); /* 725 */ + void (GLAPIENTRYP EndFragmentShaderATI)(void); /* 726 */ + GLuint (GLAPIENTRYP GenFragmentShadersATI)(GLuint range); /* 727 */ + void (GLAPIENTRYP PassTexCoordATI)(GLuint dst, GLuint coord, GLenum swizzle); /* 728 */ + void (GLAPIENTRYP SampleMapATI)(GLuint dst, GLuint interp, GLenum swizzle); /* 729 */ + void (GLAPIENTRYP SetFragmentShaderConstantATI)(GLuint dst, const GLfloat * value); /* 730 */ + void (GLAPIENTRYP PointParameteriNV)(GLenum pname, GLint param); /* 731 */ + void (GLAPIENTRYP PointParameterivNV)(GLenum pname, const GLint * params); /* 732 */ + void (GLAPIENTRYP ActiveStencilFaceEXT)(GLenum face); /* 733 */ + void (GLAPIENTRYP BindVertexArrayAPPLE)(GLuint array); /* 734 */ + void (GLAPIENTRYP DeleteVertexArraysAPPLE)(GLsizei n, const GLuint * arrays); /* 735 */ + void (GLAPIENTRYP GenVertexArraysAPPLE)(GLsizei n, GLuint * arrays); /* 736 */ + GLboolean (GLAPIENTRYP IsVertexArrayAPPLE)(GLuint array); /* 737 */ + void (GLAPIENTRYP GetProgramNamedParameterdvNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble * params); /* 738 */ + void (GLAPIENTRYP GetProgramNamedParameterfvNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat * params); /* 739 */ + void (GLAPIENTRYP ProgramNamedParameter4dNV)(GLuint id, GLsizei len, const GLubyte * name, GLdouble x, GLdouble y, GLdouble z, GLdouble w); /* 740 */ + void (GLAPIENTRYP ProgramNamedParameter4dvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLdouble * v); /* 741 */ + void (GLAPIENTRYP ProgramNamedParameter4fNV)(GLuint id, GLsizei len, const GLubyte * name, GLfloat x, GLfloat y, GLfloat z, GLfloat w); /* 742 */ + void (GLAPIENTRYP ProgramNamedParameter4fvNV)(GLuint id, GLsizei len, const GLubyte * name, const GLfloat * v); /* 743 */ + void (GLAPIENTRYP DepthBoundsEXT)(GLclampd zmin, GLclampd zmax); /* 744 */ + void (GLAPIENTRYP BlendEquationSeparateEXT)(GLenum modeRGB, GLenum modeA); /* 745 */ + void (GLAPIENTRYP BindFramebufferEXT)(GLenum target, GLuint framebuffer); /* 746 */ + void (GLAPIENTRYP BindRenderbufferEXT)(GLenum target, GLuint renderbuffer); /* 747 */ + GLenum (GLAPIENTRYP CheckFramebufferStatusEXT)(GLenum target); /* 748 */ + void (GLAPIENTRYP DeleteFramebuffersEXT)(GLsizei n, const GLuint * framebuffers); /* 749 */ + void (GLAPIENTRYP DeleteRenderbuffersEXT)(GLsizei n, const GLuint * renderbuffers); /* 750 */ + void (GLAPIENTRYP FramebufferRenderbufferEXT)(GLenum target, GLenum attachment, GLenum renderbuffertarget, GLuint renderbuffer); /* 751 */ + void (GLAPIENTRYP FramebufferTexture1DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 752 */ + void (GLAPIENTRYP FramebufferTexture2DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level); /* 753 */ + void (GLAPIENTRYP FramebufferTexture3DEXT)(GLenum target, GLenum attachment, GLenum textarget, GLuint texture, GLint level, GLint zoffset); /* 754 */ + void (GLAPIENTRYP GenFramebuffersEXT)(GLsizei n, GLuint * framebuffers); /* 755 */ + void (GLAPIENTRYP GenRenderbuffersEXT)(GLsizei n, GLuint * renderbuffers); /* 756 */ + void (GLAPIENTRYP GenerateMipmapEXT)(GLenum target); /* 757 */ + void (GLAPIENTRYP GetFramebufferAttachmentParameterivEXT)(GLenum target, GLenum attachment, GLenum pname, GLint * params); /* 758 */ + void (GLAPIENTRYP GetRenderbufferParameterivEXT)(GLenum target, GLenum pname, GLint * params); /* 759 */ + GLboolean (GLAPIENTRYP IsFramebufferEXT)(GLuint framebuffer); /* 760 */ + GLboolean (GLAPIENTRYP IsRenderbufferEXT)(GLuint renderbuffer); /* 761 */ + void (GLAPIENTRYP RenderbufferStorageEXT)(GLenum target, GLenum internalformat, GLsizei width, GLsizei height); /* 762 */ + void (GLAPIENTRYP BlitFramebufferEXT)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); /* 763 */ + void (GLAPIENTRYP FramebufferTextureLayerEXT)(GLenum target, GLenum attachment, GLuint texture, GLint level, GLint layer); /* 764 */ + void (GLAPIENTRYP StencilFuncSeparateATI)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); /* 765 */ + void (GLAPIENTRYP ProgramEnvParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 766 */ + void (GLAPIENTRYP ProgramLocalParameters4fvEXT)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); /* 767 */ + void (GLAPIENTRYP GetQueryObjecti64vEXT)(GLuint id, GLenum pname, GLint64EXT * params); /* 768 */ + void (GLAPIENTRYP GetQueryObjectui64vEXT)(GLuint id, GLenum pname, GLuint64EXT * params); /* 769 */ }; #endif /* !defined( _GLAPI_TABLE_H_ ) */ diff --git a/src/mesa/glapi/glapitemp.h b/src/mesa/glapi/glapitemp.h index 6ded362c0e..09259f4704 100644 --- a/src/mesa/glapi/glapitemp.h +++ b/src/mesa/glapi/glapitemp.h @@ -3001,21 +3001,41 @@ KEYWORD1 void KEYWORD2 NAME(ProgramEnvParameter4dARB)(GLenum target, GLuint inde DISPATCH(ProgramEnvParameter4dARB, (target, index, x, y, z, w), (F, "glProgramEnvParameter4dARB(0x%x, %d, %f, %f, %f, %f);\n", target, index, x, y, z, w)); } +KEYWORD1 void KEYWORD2 NAME(ProgramParameter4dNV)(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w) +{ + DISPATCH(ProgramEnvParameter4dARB, (target, index, x, y, z, w), (F, "glProgramParameter4dNV(0x%x, %d, %f, %f, %f, %f);\n", target, index, x, y, z, w)); +} + KEYWORD1 void KEYWORD2 NAME(ProgramEnvParameter4dvARB)(GLenum target, GLuint index, const GLdouble * params) { DISPATCH(ProgramEnvParameter4dvARB, (target, index, params), (F, "glProgramEnvParameter4dvARB(0x%x, %d, %p);\n", target, index, (const void *) params)); } +KEYWORD1 void KEYWORD2 NAME(ProgramParameter4dvNV)(GLenum target, GLuint index, const GLdouble * params) +{ + DISPATCH(ProgramEnvParameter4dvARB, (target, index, params), (F, "glProgramParameter4dvNV(0x%x, %d, %p);\n", target, index, (const void *) params)); +} + KEYWORD1 void KEYWORD2 NAME(ProgramEnvParameter4fARB)(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w) { DISPATCH(ProgramEnvParameter4fARB, (target, index, x, y, z, w), (F, "glProgramEnvParameter4fARB(0x%x, %d, %f, %f, %f, %f);\n", target, index, x, y, z, w)); } +KEYWORD1 void KEYWORD2 NAME(ProgramParameter4fNV)(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w) +{ + DISPATCH(ProgramEnvParameter4fARB, (target, index, x, y, z, w), (F, "glProgramParameter4fNV(0x%x, %d, %f, %f, %f, %f);\n", target, index, x, y, z, w)); +} + KEYWORD1 void KEYWORD2 NAME(ProgramEnvParameter4fvARB)(GLenum target, GLuint index, const GLfloat * params) { DISPATCH(ProgramEnvParameter4fvARB, (target, index, params), (F, "glProgramEnvParameter4fvARB(0x%x, %d, %p);\n", target, index, (const void *) params)); } +KEYWORD1 void KEYWORD2 NAME(ProgramParameter4fvNV)(GLenum target, GLuint index, const GLfloat * params) +{ + DISPATCH(ProgramEnvParameter4fvARB, (target, index, params), (F, "glProgramParameter4fvNV(0x%x, %d, %p);\n", target, index, (const void *) params)); +} + KEYWORD1 void KEYWORD2 NAME(ProgramLocalParameter4dARB)(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w) { DISPATCH(ProgramLocalParameter4dARB, (target, index, x, y, z, w), (F, "glProgramLocalParameter4dARB(0x%x, %d, %f, %f, %f, %f);\n", target, index, x, y, z, w)); @@ -4942,26 +4962,6 @@ KEYWORD1 void KEYWORD2 NAME(LoadProgramNV)(GLenum target, GLuint id, GLsizei len DISPATCH(LoadProgramNV, (target, id, len, program), (F, "glLoadProgramNV(0x%x, %d, %d, %p);\n", target, id, len, (const void *) program)); } -KEYWORD1 void KEYWORD2 NAME(ProgramParameter4dNV)(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w) -{ - DISPATCH(ProgramParameter4dNV, (target, index, x, y, z, w), (F, "glProgramParameter4dNV(0x%x, %d, %f, %f, %f, %f);\n", target, index, x, y, z, w)); -} - -KEYWORD1 void KEYWORD2 NAME(ProgramParameter4dvNV)(GLenum target, GLuint index, const GLdouble * params) -{ - DISPATCH(ProgramParameter4dvNV, (target, index, params), (F, "glProgramParameter4dvNV(0x%x, %d, %p);\n", target, index, (const void *) params)); -} - -KEYWORD1 void KEYWORD2 NAME(ProgramParameter4fNV)(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w) -{ - DISPATCH(ProgramParameter4fNV, (target, index, x, y, z, w), (F, "glProgramParameter4fNV(0x%x, %d, %f, %f, %f, %f);\n", target, index, x, y, z, w)); -} - -KEYWORD1 void KEYWORD2 NAME(ProgramParameter4fvNV)(GLenum target, GLuint index, const GLfloat * params) -{ - DISPATCH(ProgramParameter4fvNV, (target, index, params), (F, "glProgramParameter4fvNV(0x%x, %d, %p);\n", target, index, (const void *) params)); -} - KEYWORD1 void KEYWORD2 NAME(ProgramParameters4dvNV)(GLenum target, GLuint index, GLuint num, const GLdouble * params) { DISPATCH(ProgramParameters4dvNV, (target, index, num, params), (F, "glProgramParameters4dvNV(0x%x, %d, %d, %p);\n", target, index, num, (const void *) params)); @@ -5272,37 +5272,37 @@ KEYWORD1 void KEYWORD2 NAME(PointParameterivNV)(GLenum pname, const GLint * para DISPATCH(PointParameterivNV, (pname, params), (F, "glPointParameterivNV(0x%x, %p);\n", pname, (const void *) params)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_737)(GLenum face); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_733)(GLenum face); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_737)(GLenum face) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_733)(GLenum face) { DISPATCH(ActiveStencilFaceEXT, (face), (F, "glActiveStencilFaceEXT(0x%x);\n", face)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_738)(GLuint array); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_734)(GLuint array); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_738)(GLuint array) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_734)(GLuint array) { DISPATCH(BindVertexArrayAPPLE, (array), (F, "glBindVertexArrayAPPLE(%d);\n", array)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_739)(GLsizei n, const GLuint * arrays); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_735)(GLsizei n, const GLuint * arrays); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_739)(GLsizei n, const GLuint * arrays) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_735)(GLsizei n, const GLuint * arrays) { DISPATCH(DeleteVertexArraysAPPLE, (n, arrays), (F, "glDeleteVertexArraysAPPLE(%d, %p);\n", n, (const void *) arrays)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_740)(GLsizei n, GLuint * arrays); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_736)(GLsizei n, GLuint * arrays); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_740)(GLsizei n, GLuint * arrays) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_736)(GLsizei n, GLuint * arrays) { DISPATCH(GenVertexArraysAPPLE, (n, arrays), (F, "glGenVertexArraysAPPLE(%d, %p);\n", n, (const void *) arrays)); } -KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_741)(GLuint array); +KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_737)(GLuint array); -KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_741)(GLuint array) +KEYWORD1_ALT GLboolean KEYWORD2 NAME(_dispatch_stub_737)(GLuint array) { RETURN_DISPATCH(IsVertexArrayAPPLE, (array), (F, "glIsVertexArrayAPPLE(%d);\n", array)); } @@ -5337,9 +5337,9 @@ KEYWORD1 void KEYWORD2 NAME(ProgramNamedParameter4fvNV)(GLuint id, GLsizei len, DISPATCH(ProgramNamedParameter4fvNV, (id, len, name, v), (F, "glProgramNamedParameter4fvNV(%d, %d, %p, %p);\n", id, len, (const void *) name, (const void *) v)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_748)(GLclampd zmin, GLclampd zmax); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_744)(GLclampd zmin, GLclampd zmax); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_748)(GLclampd zmin, GLclampd zmax) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_744)(GLclampd zmin, GLclampd zmax) { DISPATCH(DepthBoundsEXT, (zmin, zmax), (F, "glDepthBoundsEXT(%f, %f);\n", zmin, zmax)); } @@ -5349,9 +5349,9 @@ KEYWORD1 void KEYWORD2 NAME(BlendEquationSeparate)(GLenum modeRGB, GLenum modeA) DISPATCH(BlendEquationSeparateEXT, (modeRGB, modeA), (F, "glBlendEquationSeparate(0x%x, 0x%x);\n", modeRGB, modeA)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_749)(GLenum modeRGB, GLenum modeA); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_745)(GLenum modeRGB, GLenum modeA); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_749)(GLenum modeRGB, GLenum modeA) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_745)(GLenum modeRGB, GLenum modeA) { DISPATCH(BlendEquationSeparateEXT, (modeRGB, modeA), (F, "glBlendEquationSeparateEXT(0x%x, 0x%x);\n", modeRGB, modeA)); } @@ -5441,9 +5441,9 @@ KEYWORD1 void KEYWORD2 NAME(RenderbufferStorageEXT)(GLenum target, GLenum intern DISPATCH(RenderbufferStorageEXT, (target, internalformat, width, height), (F, "glRenderbufferStorageEXT(0x%x, 0x%x, %d, %d);\n", target, internalformat, width, height)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_767)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_763)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_767)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_763)(GLint srcX0, GLint srcY0, GLint srcX1, GLint srcY1, GLint dstX0, GLint dstY0, GLint dstX1, GLint dstY1, GLbitfield mask, GLenum filter) { DISPATCH(BlitFramebufferEXT, (srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter), (F, "glBlitFramebufferEXT(%d, %d, %d, %d, %d, %d, %d, %d, %d, 0x%x);\n", srcX0, srcY0, srcX1, srcY1, dstX0, dstY0, dstX1, dstY1, mask, filter)); } @@ -5453,37 +5453,37 @@ KEYWORD1 void KEYWORD2 NAME(FramebufferTextureLayerEXT)(GLenum target, GLenum at DISPATCH(FramebufferTextureLayerEXT, (target, attachment, texture, level, layer), (F, "glFramebufferTextureLayerEXT(0x%x, 0x%x, %d, %d, %d);\n", target, attachment, texture, level, layer)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_765)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_765)(GLenum frontfunc, GLenum backfunc, GLint ref, GLuint mask) { DISPATCH(StencilFuncSeparateATI, (frontfunc, backfunc, ref, mask), (F, "glStencilFuncSeparateATI(0x%x, 0x%x, %d, %d);\n", frontfunc, backfunc, ref, mask)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_766)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_770)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_766)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) { DISPATCH(ProgramEnvParameters4fvEXT, (target, index, count, params), (F, "glProgramEnvParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_767)(GLenum target, GLuint index, GLsizei count, const GLfloat * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_771)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_767)(GLenum target, GLuint index, GLsizei count, const GLfloat * params) { DISPATCH(ProgramLocalParameters4fvEXT, (target, index, count, params), (F, "glProgramLocalParameters4fvEXT(0x%x, %d, %d, %p);\n", target, index, count, (const void *) params)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLuint id, GLenum pname, GLint64EXT * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLuint id, GLenum pname, GLint64EXT * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_772)(GLuint id, GLenum pname, GLint64EXT * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_768)(GLuint id, GLenum pname, GLint64EXT * params) { DISPATCH(GetQueryObjecti64vEXT, (id, pname, params), (F, "glGetQueryObjecti64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params)); } -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLuint64EXT * params); +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLuint id, GLenum pname, GLuint64EXT * params); -KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_773)(GLuint id, GLenum pname, GLuint64EXT * params) +KEYWORD1_ALT void KEYWORD2 NAME(_dispatch_stub_769)(GLuint id, GLenum pname, GLuint64EXT * params) { DISPATCH(GetQueryObjectui64vEXT, (id, pname, params), (F, "glGetQueryObjectui64vEXT(%d, 0x%x, %p);\n", id, pname, (const void *) params)); } @@ -6175,10 +6175,6 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = { TABLE_ENTRY(GetVertexAttribivNV), TABLE_ENTRY(IsProgramNV), TABLE_ENTRY(LoadProgramNV), - TABLE_ENTRY(ProgramParameter4dNV), - TABLE_ENTRY(ProgramParameter4dvNV), - TABLE_ENTRY(ProgramParameter4fNV), - TABLE_ENTRY(ProgramParameter4fvNV), TABLE_ENTRY(ProgramParameters4dvNV), TABLE_ENTRY(ProgramParameters4fvNV), TABLE_ENTRY(RequestResidentProgramsNV), @@ -6239,19 +6235,19 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = { TABLE_ENTRY(SetFragmentShaderConstantATI), TABLE_ENTRY(PointParameteriNV), TABLE_ENTRY(PointParameterivNV), + TABLE_ENTRY(_dispatch_stub_733), + TABLE_ENTRY(_dispatch_stub_734), + TABLE_ENTRY(_dispatch_stub_735), + TABLE_ENTRY(_dispatch_stub_736), TABLE_ENTRY(_dispatch_stub_737), - TABLE_ENTRY(_dispatch_stub_738), - TABLE_ENTRY(_dispatch_stub_739), - TABLE_ENTRY(_dispatch_stub_740), - TABLE_ENTRY(_dispatch_stub_741), TABLE_ENTRY(GetProgramNamedParameterdvNV), TABLE_ENTRY(GetProgramNamedParameterfvNV), TABLE_ENTRY(ProgramNamedParameter4dNV), TABLE_ENTRY(ProgramNamedParameter4dvNV), TABLE_ENTRY(ProgramNamedParameter4fNV), TABLE_ENTRY(ProgramNamedParameter4fvNV), - TABLE_ENTRY(_dispatch_stub_748), - TABLE_ENTRY(_dispatch_stub_749), + TABLE_ENTRY(_dispatch_stub_744), + TABLE_ENTRY(_dispatch_stub_745), TABLE_ENTRY(BindFramebufferEXT), TABLE_ENTRY(BindRenderbufferEXT), TABLE_ENTRY(CheckFramebufferStatusEXT), @@ -6269,13 +6265,13 @@ static _glapi_proc DISPATCH_TABLE_NAME[] = { TABLE_ENTRY(IsFramebufferEXT), TABLE_ENTRY(IsRenderbufferEXT), TABLE_ENTRY(RenderbufferStorageEXT), - TABLE_ENTRY(_dispatch_stub_767), + TABLE_ENTRY(_dispatch_stub_763), TABLE_ENTRY(FramebufferTextureLayerEXT), + TABLE_ENTRY(_dispatch_stub_765), + TABLE_ENTRY(_dispatch_stub_766), + TABLE_ENTRY(_dispatch_stub_767), + TABLE_ENTRY(_dispatch_stub_768), TABLE_ENTRY(_dispatch_stub_769), - TABLE_ENTRY(_dispatch_stub_770), - TABLE_ENTRY(_dispatch_stub_771), - TABLE_ENTRY(_dispatch_stub_772), - TABLE_ENTRY(_dispatch_stub_773), /* A whole bunch of no-op functions. These might be called * when someone tries to call a dynamically-registered * extension function without a current rendering context. @@ -6479,6 +6475,10 @@ static _glapi_proc UNUSED_TABLE_NAME[] = { TABLE_ENTRY(GetVertexAttribdv), TABLE_ENTRY(GetVertexAttribfv), TABLE_ENTRY(GetVertexAttribiv), + TABLE_ENTRY(ProgramParameter4dNV), + TABLE_ENTRY(ProgramParameter4dvNV), + TABLE_ENTRY(ProgramParameter4fNV), + TABLE_ENTRY(ProgramParameter4fvNV), TABLE_ENTRY(VertexAttrib1d), TABLE_ENTRY(VertexAttrib1dv), TABLE_ENTRY(VertexAttrib1f), diff --git a/src/mesa/glapi/glprocs.h b/src/mesa/glapi/glprocs.h index abc22a9a04..5a99d505ec 100644 --- a/src/mesa/glapi/glprocs.h +++ b/src/mesa/glapi/glprocs.h @@ -725,10 +725,6 @@ static const char gl_string_table[] = "glGetVertexAttribivNV\0" "glIsProgramNV\0" "glLoadProgramNV\0" - "glProgramParameter4dNV\0" - "glProgramParameter4dvNV\0" - "glProgramParameter4fNV\0" - "glProgramParameter4fvNV\0" "glProgramParameters4dvNV\0" "glProgramParameters4fvNV\0" "glRequestResidentProgramsNV\0" @@ -935,6 +931,10 @@ static const char gl_string_table[] = "glGetVertexAttribdv\0" "glGetVertexAttribfv\0" "glGetVertexAttribiv\0" + "glProgramParameter4dNV\0" + "glProgramParameter4dvNV\0" + "glProgramParameter4fNV\0" + "glProgramParameter4fvNV\0" "glVertexAttrib1d\0" "glVertexAttrib1dv\0" "glVertexAttrib1f\0" @@ -1138,19 +1138,19 @@ static const char gl_string_table[] = #define gl_dispatch_stub_654 mgl_dispatch_stub_654 #define gl_dispatch_stub_655 mgl_dispatch_stub_655 #define gl_dispatch_stub_656 mgl_dispatch_stub_656 +#define gl_dispatch_stub_733 mgl_dispatch_stub_733 +#define gl_dispatch_stub_734 mgl_dispatch_stub_734 +#define gl_dispatch_stub_735 mgl_dispatch_stub_735 +#define gl_dispatch_stub_736 mgl_dispatch_stub_736 #define gl_dispatch_stub_737 mgl_dispatch_stub_737 -#define gl_dispatch_stub_738 mgl_dispatch_stub_738 -#define gl_dispatch_stub_739 mgl_dispatch_stub_739 -#define gl_dispatch_stub_740 mgl_dispatch_stub_740 -#define gl_dispatch_stub_741 mgl_dispatch_stub_741 -#define gl_dispatch_stub_748 mgl_dispatch_stub_748 -#define gl_dispatch_stub_749 mgl_dispatch_stub_749 +#define gl_dispatch_stub_744 mgl_dispatch_stub_744 +#define gl_dispatch_stub_745 mgl_dispatch_stub_745 +#define gl_dispatch_stub_763 mgl_dispatch_stub_763 +#define gl_dispatch_stub_765 mgl_dispatch_stub_765 +#define gl_dispatch_stub_766 mgl_dispatch_stub_766 #define gl_dispatch_stub_767 mgl_dispatch_stub_767 +#define gl_dispatch_stub_768 mgl_dispatch_stub_768 #define gl_dispatch_stub_769 mgl_dispatch_stub_769 -#define gl_dispatch_stub_770 mgl_dispatch_stub_770 -#define gl_dispatch_stub_771 mgl_dispatch_stub_771 -#define gl_dispatch_stub_772 mgl_dispatch_stub_772 -#define gl_dispatch_stub_773 mgl_dispatch_stub_773 #endif /* USE_MGL_NAMESPACE */ @@ -1189,19 +1189,19 @@ extern void gl_dispatch_stub_653(void); extern void gl_dispatch_stub_654(void); extern void gl_dispatch_stub_655(void); extern void gl_dispatch_stub_656(void); +extern void gl_dispatch_stub_733(void); +extern void gl_dispatch_stub_734(void); +extern void gl_dispatch_stub_735(void); +extern void gl_dispatch_stub_736(void); extern void gl_dispatch_stub_737(void); -extern void gl_dispatch_stub_738(void); -extern void gl_dispatch_stub_739(void); -extern void gl_dispatch_stub_740(void); -extern void gl_dispatch_stub_741(void); -extern void gl_dispatch_stub_748(void); -extern void gl_dispatch_stub_749(void); +extern void gl_dispatch_stub_744(void); +extern void gl_dispatch_stub_745(void); +extern void gl_dispatch_stub_763(void); +extern void gl_dispatch_stub_765(void); +extern void gl_dispatch_stub_766(void); extern void gl_dispatch_stub_767(void); +extern void gl_dispatch_stub_768(void); extern void gl_dispatch_stub_769(void); -extern void gl_dispatch_stub_770(void); -extern void gl_dispatch_stub_771(void); -extern void gl_dispatch_stub_772(void); -extern void gl_dispatch_stub_773(void); #endif /* defined(NEED_FUNCTION_POINTER) || defined(GLX_INDIRECT_RENDERING) */ static const glprocs_table_t static_functions[] = { @@ -1878,216 +1878,216 @@ static const glprocs_table_t static_functions[] = { NAME_FUNC_OFFSET(11280, glGetVertexAttribivNV, glGetVertexAttribivNV, NULL, _gloffset_GetVertexAttribivNV), NAME_FUNC_OFFSET(11302, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV), NAME_FUNC_OFFSET(11316, glLoadProgramNV, glLoadProgramNV, NULL, _gloffset_LoadProgramNV), - NAME_FUNC_OFFSET(11332, glProgramParameter4dNV, glProgramParameter4dNV, NULL, _gloffset_ProgramParameter4dNV), - NAME_FUNC_OFFSET(11355, glProgramParameter4dvNV, glProgramParameter4dvNV, NULL, _gloffset_ProgramParameter4dvNV), - NAME_FUNC_OFFSET(11379, glProgramParameter4fNV, glProgramParameter4fNV, NULL, _gloffset_ProgramParameter4fNV), - NAME_FUNC_OFFSET(11402, glProgramParameter4fvNV, glProgramParameter4fvNV, NULL, _gloffset_ProgramParameter4fvNV), - NAME_FUNC_OFFSET(11426, glProgramParameters4dvNV, glProgramParameters4dvNV, NULL, _gloffset_ProgramParameters4dvNV), - NAME_FUNC_OFFSET(11451, glProgramParameters4fvNV, glProgramParameters4fvNV, NULL, _gloffset_ProgramParameters4fvNV), - NAME_FUNC_OFFSET(11476, glRequestResidentProgramsNV, glRequestResidentProgramsNV, NULL, _gloffset_RequestResidentProgramsNV), - NAME_FUNC_OFFSET(11504, glTrackMatrixNV, glTrackMatrixNV, NULL, _gloffset_TrackMatrixNV), - NAME_FUNC_OFFSET(11520, glVertexAttrib1dNV, glVertexAttrib1dNV, NULL, _gloffset_VertexAttrib1dNV), - NAME_FUNC_OFFSET(11539, glVertexAttrib1dvNV, glVertexAttrib1dvNV, NULL, _gloffset_VertexAttrib1dvNV), - NAME_FUNC_OFFSET(11559, glVertexAttrib1fNV, glVertexAttrib1fNV, NULL, _gloffset_VertexAttrib1fNV), - NAME_FUNC_OFFSET(11578, glVertexAttrib1fvNV, glVertexAttrib1fvNV, NULL, _gloffset_VertexAttrib1fvNV), - NAME_FUNC_OFFSET(11598, glVertexAttrib1sNV, glVertexAttrib1sNV, NULL, _gloffset_VertexAttrib1sNV), - NAME_FUNC_OFFSET(11617, glVertexAttrib1svNV, glVertexAttrib1svNV, NULL, _gloffset_VertexAttrib1svNV), - NAME_FUNC_OFFSET(11637, glVertexAttrib2dNV, glVertexAttrib2dNV, NULL, _gloffset_VertexAttrib2dNV), - NAME_FUNC_OFFSET(11656, glVertexAttrib2dvNV, glVertexAttrib2dvNV, NULL, _gloffset_VertexAttrib2dvNV), - NAME_FUNC_OFFSET(11676, glVertexAttrib2fNV, glVertexAttrib2fNV, NULL, _gloffset_VertexAttrib2fNV), - NAME_FUNC_OFFSET(11695, glVertexAttrib2fvNV, glVertexAttrib2fvNV, NULL, _gloffset_VertexAttrib2fvNV), - NAME_FUNC_OFFSET(11715, glVertexAttrib2sNV, glVertexAttrib2sNV, NULL, _gloffset_VertexAttrib2sNV), - NAME_FUNC_OFFSET(11734, glVertexAttrib2svNV, glVertexAttrib2svNV, NULL, _gloffset_VertexAttrib2svNV), - NAME_FUNC_OFFSET(11754, glVertexAttrib3dNV, glVertexAttrib3dNV, NULL, _gloffset_VertexAttrib3dNV), - NAME_FUNC_OFFSET(11773, glVertexAttrib3dvNV, glVertexAttrib3dvNV, NULL, _gloffset_VertexAttrib3dvNV), - NAME_FUNC_OFFSET(11793, glVertexAttrib3fNV, glVertexAttrib3fNV, NULL, _gloffset_VertexAttrib3fNV), - NAME_FUNC_OFFSET(11812, glVertexAttrib3fvNV, glVertexAttrib3fvNV, NULL, _gloffset_VertexAttrib3fvNV), - NAME_FUNC_OFFSET(11832, glVertexAttrib3sNV, glVertexAttrib3sNV, NULL, _gloffset_VertexAttrib3sNV), - NAME_FUNC_OFFSET(11851, glVertexAttrib3svNV, glVertexAttrib3svNV, NULL, _gloffset_VertexAttrib3svNV), - NAME_FUNC_OFFSET(11871, glVertexAttrib4dNV, glVertexAttrib4dNV, NULL, _gloffset_VertexAttrib4dNV), - NAME_FUNC_OFFSET(11890, glVertexAttrib4dvNV, glVertexAttrib4dvNV, NULL, _gloffset_VertexAttrib4dvNV), - NAME_FUNC_OFFSET(11910, glVertexAttrib4fNV, glVertexAttrib4fNV, NULL, _gloffset_VertexAttrib4fNV), - NAME_FUNC_OFFSET(11929, glVertexAttrib4fvNV, glVertexAttrib4fvNV, NULL, _gloffset_VertexAttrib4fvNV), - NAME_FUNC_OFFSET(11949, glVertexAttrib4sNV, glVertexAttrib4sNV, NULL, _gloffset_VertexAttrib4sNV), - NAME_FUNC_OFFSET(11968, glVertexAttrib4svNV, glVertexAttrib4svNV, NULL, _gloffset_VertexAttrib4svNV), - NAME_FUNC_OFFSET(11988, glVertexAttrib4ubNV, glVertexAttrib4ubNV, NULL, _gloffset_VertexAttrib4ubNV), - NAME_FUNC_OFFSET(12008, glVertexAttrib4ubvNV, glVertexAttrib4ubvNV, NULL, _gloffset_VertexAttrib4ubvNV), - NAME_FUNC_OFFSET(12029, glVertexAttribPointerNV, glVertexAttribPointerNV, NULL, _gloffset_VertexAttribPointerNV), - NAME_FUNC_OFFSET(12053, glVertexAttribs1dvNV, glVertexAttribs1dvNV, NULL, _gloffset_VertexAttribs1dvNV), - NAME_FUNC_OFFSET(12074, glVertexAttribs1fvNV, glVertexAttribs1fvNV, NULL, _gloffset_VertexAttribs1fvNV), - NAME_FUNC_OFFSET(12095, glVertexAttribs1svNV, glVertexAttribs1svNV, NULL, _gloffset_VertexAttribs1svNV), - NAME_FUNC_OFFSET(12116, glVertexAttribs2dvNV, glVertexAttribs2dvNV, NULL, _gloffset_VertexAttribs2dvNV), - NAME_FUNC_OFFSET(12137, glVertexAttribs2fvNV, glVertexAttribs2fvNV, NULL, _gloffset_VertexAttribs2fvNV), - NAME_FUNC_OFFSET(12158, glVertexAttribs2svNV, glVertexAttribs2svNV, NULL, _gloffset_VertexAttribs2svNV), - NAME_FUNC_OFFSET(12179, glVertexAttribs3dvNV, glVertexAttribs3dvNV, NULL, _gloffset_VertexAttribs3dvNV), - NAME_FUNC_OFFSET(12200, glVertexAttribs3fvNV, glVertexAttribs3fvNV, NULL, _gloffset_VertexAttribs3fvNV), - NAME_FUNC_OFFSET(12221, glVertexAttribs3svNV, glVertexAttribs3svNV, NULL, _gloffset_VertexAttribs3svNV), - NAME_FUNC_OFFSET(12242, glVertexAttribs4dvNV, glVertexAttribs4dvNV, NULL, _gloffset_VertexAttribs4dvNV), - NAME_FUNC_OFFSET(12263, glVertexAttribs4fvNV, glVertexAttribs4fvNV, NULL, _gloffset_VertexAttribs4fvNV), - NAME_FUNC_OFFSET(12284, glVertexAttribs4svNV, glVertexAttribs4svNV, NULL, _gloffset_VertexAttribs4svNV), - NAME_FUNC_OFFSET(12305, glVertexAttribs4ubvNV, glVertexAttribs4ubvNV, NULL, _gloffset_VertexAttribs4ubvNV), - NAME_FUNC_OFFSET(12327, glAlphaFragmentOp1ATI, glAlphaFragmentOp1ATI, NULL, _gloffset_AlphaFragmentOp1ATI), - NAME_FUNC_OFFSET(12349, glAlphaFragmentOp2ATI, glAlphaFragmentOp2ATI, NULL, _gloffset_AlphaFragmentOp2ATI), - NAME_FUNC_OFFSET(12371, glAlphaFragmentOp3ATI, glAlphaFragmentOp3ATI, NULL, _gloffset_AlphaFragmentOp3ATI), - NAME_FUNC_OFFSET(12393, glBeginFragmentShaderATI, glBeginFragmentShaderATI, NULL, _gloffset_BeginFragmentShaderATI), - NAME_FUNC_OFFSET(12418, glBindFragmentShaderATI, glBindFragmentShaderATI, NULL, _gloffset_BindFragmentShaderATI), - NAME_FUNC_OFFSET(12442, glColorFragmentOp1ATI, glColorFragmentOp1ATI, NULL, _gloffset_ColorFragmentOp1ATI), - NAME_FUNC_OFFSET(12464, glColorFragmentOp2ATI, glColorFragmentOp2ATI, NULL, _gloffset_ColorFragmentOp2ATI), - NAME_FUNC_OFFSET(12486, glColorFragmentOp3ATI, glColorFragmentOp3ATI, NULL, _gloffset_ColorFragmentOp3ATI), - NAME_FUNC_OFFSET(12508, glDeleteFragmentShaderATI, glDeleteFragmentShaderATI, NULL, _gloffset_DeleteFragmentShaderATI), - NAME_FUNC_OFFSET(12534, glEndFragmentShaderATI, glEndFragmentShaderATI, NULL, _gloffset_EndFragmentShaderATI), - NAME_FUNC_OFFSET(12557, glGenFragmentShadersATI, glGenFragmentShadersATI, NULL, _gloffset_GenFragmentShadersATI), - NAME_FUNC_OFFSET(12581, glPassTexCoordATI, glPassTexCoordATI, NULL, _gloffset_PassTexCoordATI), - NAME_FUNC_OFFSET(12599, glSampleMapATI, glSampleMapATI, NULL, _gloffset_SampleMapATI), - NAME_FUNC_OFFSET(12614, glSetFragmentShaderConstantATI, glSetFragmentShaderConstantATI, NULL, _gloffset_SetFragmentShaderConstantATI), - NAME_FUNC_OFFSET(12645, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV), - NAME_FUNC_OFFSET(12665, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV), - NAME_FUNC_OFFSET(12686, gl_dispatch_stub_737, gl_dispatch_stub_737, NULL, _gloffset_ActiveStencilFaceEXT), - NAME_FUNC_OFFSET(12709, gl_dispatch_stub_738, gl_dispatch_stub_738, NULL, _gloffset_BindVertexArrayAPPLE), - NAME_FUNC_OFFSET(12732, gl_dispatch_stub_739, gl_dispatch_stub_739, NULL, _gloffset_DeleteVertexArraysAPPLE), - NAME_FUNC_OFFSET(12758, gl_dispatch_stub_740, gl_dispatch_stub_740, NULL, _gloffset_GenVertexArraysAPPLE), - NAME_FUNC_OFFSET(12781, gl_dispatch_stub_741, gl_dispatch_stub_741, NULL, _gloffset_IsVertexArrayAPPLE), - NAME_FUNC_OFFSET(12802, glGetProgramNamedParameterdvNV, glGetProgramNamedParameterdvNV, NULL, _gloffset_GetProgramNamedParameterdvNV), - NAME_FUNC_OFFSET(12833, glGetProgramNamedParameterfvNV, glGetProgramNamedParameterfvNV, NULL, _gloffset_GetProgramNamedParameterfvNV), - NAME_FUNC_OFFSET(12864, glProgramNamedParameter4dNV, glProgramNamedParameter4dNV, NULL, _gloffset_ProgramNamedParameter4dNV), - NAME_FUNC_OFFSET(12892, glProgramNamedParameter4dvNV, glProgramNamedParameter4dvNV, NULL, _gloffset_ProgramNamedParameter4dvNV), - NAME_FUNC_OFFSET(12921, glProgramNamedParameter4fNV, glProgramNamedParameter4fNV, NULL, _gloffset_ProgramNamedParameter4fNV), - NAME_FUNC_OFFSET(12949, glProgramNamedParameter4fvNV, glProgramNamedParameter4fvNV, NULL, _gloffset_ProgramNamedParameter4fvNV), - NAME_FUNC_OFFSET(12978, gl_dispatch_stub_748, gl_dispatch_stub_748, NULL, _gloffset_DepthBoundsEXT), - NAME_FUNC_OFFSET(12995, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT), - NAME_FUNC_OFFSET(13022, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT), - NAME_FUNC_OFFSET(13043, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT), - NAME_FUNC_OFFSET(13065, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT), - NAME_FUNC_OFFSET(13093, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT), - NAME_FUNC_OFFSET(13117, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT), - NAME_FUNC_OFFSET(13142, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT), - NAME_FUNC_OFFSET(13171, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT), - NAME_FUNC_OFFSET(13197, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT), - NAME_FUNC_OFFSET(13223, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT), - NAME_FUNC_OFFSET(13249, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT), - NAME_FUNC_OFFSET(13270, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT), - NAME_FUNC_OFFSET(13292, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT), - NAME_FUNC_OFFSET(13312, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT), - NAME_FUNC_OFFSET(13353, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT), - NAME_FUNC_OFFSET(13385, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT), - NAME_FUNC_OFFSET(13404, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT), - NAME_FUNC_OFFSET(13424, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT), - NAME_FUNC_OFFSET(13449, gl_dispatch_stub_767, gl_dispatch_stub_767, NULL, _gloffset_BlitFramebufferEXT), - NAME_FUNC_OFFSET(13470, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT), - NAME_FUNC_OFFSET(13499, gl_dispatch_stub_769, gl_dispatch_stub_769, NULL, _gloffset_StencilFuncSeparateATI), - NAME_FUNC_OFFSET(13524, gl_dispatch_stub_770, gl_dispatch_stub_770, NULL, _gloffset_ProgramEnvParameters4fvEXT), - NAME_FUNC_OFFSET(13553, gl_dispatch_stub_771, gl_dispatch_stub_771, NULL, _gloffset_ProgramLocalParameters4fvEXT), - NAME_FUNC_OFFSET(13584, gl_dispatch_stub_772, gl_dispatch_stub_772, NULL, _gloffset_GetQueryObjecti64vEXT), - NAME_FUNC_OFFSET(13608, gl_dispatch_stub_773, gl_dispatch_stub_773, NULL, _gloffset_GetQueryObjectui64vEXT), - NAME_FUNC_OFFSET(13633, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement), - NAME_FUNC_OFFSET(13651, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture), - NAME_FUNC_OFFSET(13668, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays), - NAME_FUNC_OFFSET(13684, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident), - NAME_FUNC_OFFSET(13709, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D), - NAME_FUNC_OFFSET(13729, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D), - NAME_FUNC_OFFSET(13749, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D), - NAME_FUNC_OFFSET(13772, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D), - NAME_FUNC_OFFSET(13795, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures), - NAME_FUNC_OFFSET(13815, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures), - NAME_FUNC_OFFSET(13832, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv), - NAME_FUNC_OFFSET(13849, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture), - NAME_FUNC_OFFSET(13864, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures), - NAME_FUNC_OFFSET(13888, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D), - NAME_FUNC_OFFSET(13907, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D), - NAME_FUNC_OFFSET(13926, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor), - NAME_FUNC_OFFSET(13942, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation), - NAME_FUNC_OFFSET(13961, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements), - NAME_FUNC_OFFSET(13984, glColorTable, glColorTable, NULL, _gloffset_ColorTable), - NAME_FUNC_OFFSET(14000, glColorTable, glColorTable, NULL, _gloffset_ColorTable), - NAME_FUNC_OFFSET(14016, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv), - NAME_FUNC_OFFSET(14043, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv), - NAME_FUNC_OFFSET(14070, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable), - NAME_FUNC_OFFSET(14090, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), - NAME_FUNC_OFFSET(14109, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), - NAME_FUNC_OFFSET(14128, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), - NAME_FUNC_OFFSET(14158, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), - NAME_FUNC_OFFSET(14188, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), - NAME_FUNC_OFFSET(14218, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), - NAME_FUNC_OFFSET(14248, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable), - NAME_FUNC_OFFSET(14267, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable), - NAME_FUNC_OFFSET(14290, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D), - NAME_FUNC_OFFSET(14315, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D), - NAME_FUNC_OFFSET(14340, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf), - NAME_FUNC_OFFSET(14367, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv), - NAME_FUNC_OFFSET(14395, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri), - NAME_FUNC_OFFSET(14422, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv), - NAME_FUNC_OFFSET(14450, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D), - NAME_FUNC_OFFSET(14479, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D), - NAME_FUNC_OFFSET(14508, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter), - NAME_FUNC_OFFSET(14534, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv), - NAME_FUNC_OFFSET(14565, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv), - NAME_FUNC_OFFSET(14596, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter), - NAME_FUNC_OFFSET(14620, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D), - NAME_FUNC_OFFSET(14643, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram), - NAME_FUNC_OFFSET(14661, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv), - NAME_FUNC_OFFSET(14690, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv), - NAME_FUNC_OFFSET(14719, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax), - NAME_FUNC_OFFSET(14734, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv), - NAME_FUNC_OFFSET(14760, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv), - NAME_FUNC_OFFSET(14786, glHistogram, glHistogram, NULL, _gloffset_Histogram), - NAME_FUNC_OFFSET(14801, glMinmax, glMinmax, NULL, _gloffset_Minmax), - NAME_FUNC_OFFSET(14813, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram), - NAME_FUNC_OFFSET(14833, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax), - NAME_FUNC_OFFSET(14850, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D), - NAME_FUNC_OFFSET(14866, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D), - NAME_FUNC_OFFSET(14885, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D), - NAME_FUNC_OFFSET(14908, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB), - NAME_FUNC_OFFSET(14924, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB), - NAME_FUNC_OFFSET(14946, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB), - NAME_FUNC_OFFSET(14964, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB), - NAME_FUNC_OFFSET(14983, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB), - NAME_FUNC_OFFSET(15001, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB), - NAME_FUNC_OFFSET(15020, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB), - NAME_FUNC_OFFSET(15038, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB), - NAME_FUNC_OFFSET(15057, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB), - NAME_FUNC_OFFSET(15075, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB), - NAME_FUNC_OFFSET(15094, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB), - NAME_FUNC_OFFSET(15112, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB), - NAME_FUNC_OFFSET(15131, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB), - NAME_FUNC_OFFSET(15149, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB), - NAME_FUNC_OFFSET(15168, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB), - NAME_FUNC_OFFSET(15186, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB), - NAME_FUNC_OFFSET(15205, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB), - NAME_FUNC_OFFSET(15223, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB), - NAME_FUNC_OFFSET(15242, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB), - NAME_FUNC_OFFSET(15260, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB), - NAME_FUNC_OFFSET(15279, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB), - NAME_FUNC_OFFSET(15297, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB), - NAME_FUNC_OFFSET(15316, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB), - NAME_FUNC_OFFSET(15334, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB), - NAME_FUNC_OFFSET(15353, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB), - NAME_FUNC_OFFSET(15371, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB), - NAME_FUNC_OFFSET(15390, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB), - NAME_FUNC_OFFSET(15408, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB), - NAME_FUNC_OFFSET(15427, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB), - NAME_FUNC_OFFSET(15445, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB), - NAME_FUNC_OFFSET(15464, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB), - NAME_FUNC_OFFSET(15482, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB), - NAME_FUNC_OFFSET(15501, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB), - NAME_FUNC_OFFSET(15519, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB), - NAME_FUNC_OFFSET(15538, glStencilOpSeparate, glStencilOpSeparate, NULL, _gloffset_StencilOpSeparate), - NAME_FUNC_OFFSET(15561, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB), - NAME_FUNC_OFFSET(15584, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB), - NAME_FUNC_OFFSET(15607, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB), - NAME_FUNC_OFFSET(15630, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB), - NAME_FUNC_OFFSET(15653, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB), - NAME_FUNC_OFFSET(15670, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB), - NAME_FUNC_OFFSET(15693, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB), - NAME_FUNC_OFFSET(15716, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB), - NAME_FUNC_OFFSET(15739, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB), - NAME_FUNC_OFFSET(15765, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB), - NAME_FUNC_OFFSET(15791, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB), - NAME_FUNC_OFFSET(15817, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB), - NAME_FUNC_OFFSET(15841, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB), - NAME_FUNC_OFFSET(15868, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB), - NAME_FUNC_OFFSET(15894, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB), - NAME_FUNC_OFFSET(15914, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB), - NAME_FUNC_OFFSET(15934, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB), + NAME_FUNC_OFFSET(11332, glProgramParameters4dvNV, glProgramParameters4dvNV, NULL, _gloffset_ProgramParameters4dvNV), + NAME_FUNC_OFFSET(11357, glProgramParameters4fvNV, glProgramParameters4fvNV, NULL, _gloffset_ProgramParameters4fvNV), + NAME_FUNC_OFFSET(11382, glRequestResidentProgramsNV, glRequestResidentProgramsNV, NULL, _gloffset_RequestResidentProgramsNV), + NAME_FUNC_OFFSET(11410, glTrackMatrixNV, glTrackMatrixNV, NULL, _gloffset_TrackMatrixNV), + NAME_FUNC_OFFSET(11426, glVertexAttrib1dNV, glVertexAttrib1dNV, NULL, _gloffset_VertexAttrib1dNV), + NAME_FUNC_OFFSET(11445, glVertexAttrib1dvNV, glVertexAttrib1dvNV, NULL, _gloffset_VertexAttrib1dvNV), + NAME_FUNC_OFFSET(11465, glVertexAttrib1fNV, glVertexAttrib1fNV, NULL, _gloffset_VertexAttrib1fNV), + NAME_FUNC_OFFSET(11484, glVertexAttrib1fvNV, glVertexAttrib1fvNV, NULL, _gloffset_VertexAttrib1fvNV), + NAME_FUNC_OFFSET(11504, glVertexAttrib1sNV, glVertexAttrib1sNV, NULL, _gloffset_VertexAttrib1sNV), + NAME_FUNC_OFFSET(11523, glVertexAttrib1svNV, glVertexAttrib1svNV, NULL, _gloffset_VertexAttrib1svNV), + NAME_FUNC_OFFSET(11543, glVertexAttrib2dNV, glVertexAttrib2dNV, NULL, _gloffset_VertexAttrib2dNV), + NAME_FUNC_OFFSET(11562, glVertexAttrib2dvNV, glVertexAttrib2dvNV, NULL, _gloffset_VertexAttrib2dvNV), + NAME_FUNC_OFFSET(11582, glVertexAttrib2fNV, glVertexAttrib2fNV, NULL, _gloffset_VertexAttrib2fNV), + NAME_FUNC_OFFSET(11601, glVertexAttrib2fvNV, glVertexAttrib2fvNV, NULL, _gloffset_VertexAttrib2fvNV), + NAME_FUNC_OFFSET(11621, glVertexAttrib2sNV, glVertexAttrib2sNV, NULL, _gloffset_VertexAttrib2sNV), + NAME_FUNC_OFFSET(11640, glVertexAttrib2svNV, glVertexAttrib2svNV, NULL, _gloffset_VertexAttrib2svNV), + NAME_FUNC_OFFSET(11660, glVertexAttrib3dNV, glVertexAttrib3dNV, NULL, _gloffset_VertexAttrib3dNV), + NAME_FUNC_OFFSET(11679, glVertexAttrib3dvNV, glVertexAttrib3dvNV, NULL, _gloffset_VertexAttrib3dvNV), + NAME_FUNC_OFFSET(11699, glVertexAttrib3fNV, glVertexAttrib3fNV, NULL, _gloffset_VertexAttrib3fNV), + NAME_FUNC_OFFSET(11718, glVertexAttrib3fvNV, glVertexAttrib3fvNV, NULL, _gloffset_VertexAttrib3fvNV), + NAME_FUNC_OFFSET(11738, glVertexAttrib3sNV, glVertexAttrib3sNV, NULL, _gloffset_VertexAttrib3sNV), + NAME_FUNC_OFFSET(11757, glVertexAttrib3svNV, glVertexAttrib3svNV, NULL, _gloffset_VertexAttrib3svNV), + NAME_FUNC_OFFSET(11777, glVertexAttrib4dNV, glVertexAttrib4dNV, NULL, _gloffset_VertexAttrib4dNV), + NAME_FUNC_OFFSET(11796, glVertexAttrib4dvNV, glVertexAttrib4dvNV, NULL, _gloffset_VertexAttrib4dvNV), + NAME_FUNC_OFFSET(11816, glVertexAttrib4fNV, glVertexAttrib4fNV, NULL, _gloffset_VertexAttrib4fNV), + NAME_FUNC_OFFSET(11835, glVertexAttrib4fvNV, glVertexAttrib4fvNV, NULL, _gloffset_VertexAttrib4fvNV), + NAME_FUNC_OFFSET(11855, glVertexAttrib4sNV, glVertexAttrib4sNV, NULL, _gloffset_VertexAttrib4sNV), + NAME_FUNC_OFFSET(11874, glVertexAttrib4svNV, glVertexAttrib4svNV, NULL, _gloffset_VertexAttrib4svNV), + NAME_FUNC_OFFSET(11894, glVertexAttrib4ubNV, glVertexAttrib4ubNV, NULL, _gloffset_VertexAttrib4ubNV), + NAME_FUNC_OFFSET(11914, glVertexAttrib4ubvNV, glVertexAttrib4ubvNV, NULL, _gloffset_VertexAttrib4ubvNV), + NAME_FUNC_OFFSET(11935, glVertexAttribPointerNV, glVertexAttribPointerNV, NULL, _gloffset_VertexAttribPointerNV), + NAME_FUNC_OFFSET(11959, glVertexAttribs1dvNV, glVertexAttribs1dvNV, NULL, _gloffset_VertexAttribs1dvNV), + NAME_FUNC_OFFSET(11980, glVertexAttribs1fvNV, glVertexAttribs1fvNV, NULL, _gloffset_VertexAttribs1fvNV), + NAME_FUNC_OFFSET(12001, glVertexAttribs1svNV, glVertexAttribs1svNV, NULL, _gloffset_VertexAttribs1svNV), + NAME_FUNC_OFFSET(12022, glVertexAttribs2dvNV, glVertexAttribs2dvNV, NULL, _gloffset_VertexAttribs2dvNV), + NAME_FUNC_OFFSET(12043, glVertexAttribs2fvNV, glVertexAttribs2fvNV, NULL, _gloffset_VertexAttribs2fvNV), + NAME_FUNC_OFFSET(12064, glVertexAttribs2svNV, glVertexAttribs2svNV, NULL, _gloffset_VertexAttribs2svNV), + NAME_FUNC_OFFSET(12085, glVertexAttribs3dvNV, glVertexAttribs3dvNV, NULL, _gloffset_VertexAttribs3dvNV), + NAME_FUNC_OFFSET(12106, glVertexAttribs3fvNV, glVertexAttribs3fvNV, NULL, _gloffset_VertexAttribs3fvNV), + NAME_FUNC_OFFSET(12127, glVertexAttribs3svNV, glVertexAttribs3svNV, NULL, _gloffset_VertexAttribs3svNV), + NAME_FUNC_OFFSET(12148, glVertexAttribs4dvNV, glVertexAttribs4dvNV, NULL, _gloffset_VertexAttribs4dvNV), + NAME_FUNC_OFFSET(12169, glVertexAttribs4fvNV, glVertexAttribs4fvNV, NULL, _gloffset_VertexAttribs4fvNV), + NAME_FUNC_OFFSET(12190, glVertexAttribs4svNV, glVertexAttribs4svNV, NULL, _gloffset_VertexAttribs4svNV), + NAME_FUNC_OFFSET(12211, glVertexAttribs4ubvNV, glVertexAttribs4ubvNV, NULL, _gloffset_VertexAttribs4ubvNV), + NAME_FUNC_OFFSET(12233, glAlphaFragmentOp1ATI, glAlphaFragmentOp1ATI, NULL, _gloffset_AlphaFragmentOp1ATI), + NAME_FUNC_OFFSET(12255, glAlphaFragmentOp2ATI, glAlphaFragmentOp2ATI, NULL, _gloffset_AlphaFragmentOp2ATI), + NAME_FUNC_OFFSET(12277, glAlphaFragmentOp3ATI, glAlphaFragmentOp3ATI, NULL, _gloffset_AlphaFragmentOp3ATI), + NAME_FUNC_OFFSET(12299, glBeginFragmentShaderATI, glBeginFragmentShaderATI, NULL, _gloffset_BeginFragmentShaderATI), + NAME_FUNC_OFFSET(12324, glBindFragmentShaderATI, glBindFragmentShaderATI, NULL, _gloffset_BindFragmentShaderATI), + NAME_FUNC_OFFSET(12348, glColorFragmentOp1ATI, glColorFragmentOp1ATI, NULL, _gloffset_ColorFragmentOp1ATI), + NAME_FUNC_OFFSET(12370, glColorFragmentOp2ATI, glColorFragmentOp2ATI, NULL, _gloffset_ColorFragmentOp2ATI), + NAME_FUNC_OFFSET(12392, glColorFragmentOp3ATI, glColorFragmentOp3ATI, NULL, _gloffset_ColorFragmentOp3ATI), + NAME_FUNC_OFFSET(12414, glDeleteFragmentShaderATI, glDeleteFragmentShaderATI, NULL, _gloffset_DeleteFragmentShaderATI), + NAME_FUNC_OFFSET(12440, glEndFragmentShaderATI, glEndFragmentShaderATI, NULL, _gloffset_EndFragmentShaderATI), + NAME_FUNC_OFFSET(12463, glGenFragmentShadersATI, glGenFragmentShadersATI, NULL, _gloffset_GenFragmentShadersATI), + NAME_FUNC_OFFSET(12487, glPassTexCoordATI, glPassTexCoordATI, NULL, _gloffset_PassTexCoordATI), + NAME_FUNC_OFFSET(12505, glSampleMapATI, glSampleMapATI, NULL, _gloffset_SampleMapATI), + NAME_FUNC_OFFSET(12520, glSetFragmentShaderConstantATI, glSetFragmentShaderConstantATI, NULL, _gloffset_SetFragmentShaderConstantATI), + NAME_FUNC_OFFSET(12551, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV), + NAME_FUNC_OFFSET(12571, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV), + NAME_FUNC_OFFSET(12592, gl_dispatch_stub_733, gl_dispatch_stub_733, NULL, _gloffset_ActiveStencilFaceEXT), + NAME_FUNC_OFFSET(12615, gl_dispatch_stub_734, gl_dispatch_stub_734, NULL, _gloffset_BindVertexArrayAPPLE), + NAME_FUNC_OFFSET(12638, gl_dispatch_stub_735, gl_dispatch_stub_735, NULL, _gloffset_DeleteVertexArraysAPPLE), + NAME_FUNC_OFFSET(12664, gl_dispatch_stub_736, gl_dispatch_stub_736, NULL, _gloffset_GenVertexArraysAPPLE), + NAME_FUNC_OFFSET(12687, gl_dispatch_stub_737, gl_dispatch_stub_737, NULL, _gloffset_IsVertexArrayAPPLE), + NAME_FUNC_OFFSET(12708, glGetProgramNamedParameterdvNV, glGetProgramNamedParameterdvNV, NULL, _gloffset_GetProgramNamedParameterdvNV), + NAME_FUNC_OFFSET(12739, glGetProgramNamedParameterfvNV, glGetProgramNamedParameterfvNV, NULL, _gloffset_GetProgramNamedParameterfvNV), + NAME_FUNC_OFFSET(12770, glProgramNamedParameter4dNV, glProgramNamedParameter4dNV, NULL, _gloffset_ProgramNamedParameter4dNV), + NAME_FUNC_OFFSET(12798, glProgramNamedParameter4dvNV, glProgramNamedParameter4dvNV, NULL, _gloffset_ProgramNamedParameter4dvNV), + NAME_FUNC_OFFSET(12827, glProgramNamedParameter4fNV, glProgramNamedParameter4fNV, NULL, _gloffset_ProgramNamedParameter4fNV), + NAME_FUNC_OFFSET(12855, glProgramNamedParameter4fvNV, glProgramNamedParameter4fvNV, NULL, _gloffset_ProgramNamedParameter4fvNV), + NAME_FUNC_OFFSET(12884, gl_dispatch_stub_744, gl_dispatch_stub_744, NULL, _gloffset_DepthBoundsEXT), + NAME_FUNC_OFFSET(12901, gl_dispatch_stub_745, gl_dispatch_stub_745, NULL, _gloffset_BlendEquationSeparateEXT), + NAME_FUNC_OFFSET(12928, glBindFramebufferEXT, glBindFramebufferEXT, NULL, _gloffset_BindFramebufferEXT), + NAME_FUNC_OFFSET(12949, glBindRenderbufferEXT, glBindRenderbufferEXT, NULL, _gloffset_BindRenderbufferEXT), + NAME_FUNC_OFFSET(12971, glCheckFramebufferStatusEXT, glCheckFramebufferStatusEXT, NULL, _gloffset_CheckFramebufferStatusEXT), + NAME_FUNC_OFFSET(12999, glDeleteFramebuffersEXT, glDeleteFramebuffersEXT, NULL, _gloffset_DeleteFramebuffersEXT), + NAME_FUNC_OFFSET(13023, glDeleteRenderbuffersEXT, glDeleteRenderbuffersEXT, NULL, _gloffset_DeleteRenderbuffersEXT), + NAME_FUNC_OFFSET(13048, glFramebufferRenderbufferEXT, glFramebufferRenderbufferEXT, NULL, _gloffset_FramebufferRenderbufferEXT), + NAME_FUNC_OFFSET(13077, glFramebufferTexture1DEXT, glFramebufferTexture1DEXT, NULL, _gloffset_FramebufferTexture1DEXT), + NAME_FUNC_OFFSET(13103, glFramebufferTexture2DEXT, glFramebufferTexture2DEXT, NULL, _gloffset_FramebufferTexture2DEXT), + NAME_FUNC_OFFSET(13129, glFramebufferTexture3DEXT, glFramebufferTexture3DEXT, NULL, _gloffset_FramebufferTexture3DEXT), + NAME_FUNC_OFFSET(13155, glGenFramebuffersEXT, glGenFramebuffersEXT, NULL, _gloffset_GenFramebuffersEXT), + NAME_FUNC_OFFSET(13176, glGenRenderbuffersEXT, glGenRenderbuffersEXT, NULL, _gloffset_GenRenderbuffersEXT), + NAME_FUNC_OFFSET(13198, glGenerateMipmapEXT, glGenerateMipmapEXT, NULL, _gloffset_GenerateMipmapEXT), + NAME_FUNC_OFFSET(13218, glGetFramebufferAttachmentParameterivEXT, glGetFramebufferAttachmentParameterivEXT, NULL, _gloffset_GetFramebufferAttachmentParameterivEXT), + NAME_FUNC_OFFSET(13259, glGetRenderbufferParameterivEXT, glGetRenderbufferParameterivEXT, NULL, _gloffset_GetRenderbufferParameterivEXT), + NAME_FUNC_OFFSET(13291, glIsFramebufferEXT, glIsFramebufferEXT, NULL, _gloffset_IsFramebufferEXT), + NAME_FUNC_OFFSET(13310, glIsRenderbufferEXT, glIsRenderbufferEXT, NULL, _gloffset_IsRenderbufferEXT), + NAME_FUNC_OFFSET(13330, glRenderbufferStorageEXT, glRenderbufferStorageEXT, NULL, _gloffset_RenderbufferStorageEXT), + NAME_FUNC_OFFSET(13355, gl_dispatch_stub_763, gl_dispatch_stub_763, NULL, _gloffset_BlitFramebufferEXT), + NAME_FUNC_OFFSET(13376, glFramebufferTextureLayerEXT, glFramebufferTextureLayerEXT, NULL, _gloffset_FramebufferTextureLayerEXT), + NAME_FUNC_OFFSET(13405, gl_dispatch_stub_765, gl_dispatch_stub_765, NULL, _gloffset_StencilFuncSeparateATI), + NAME_FUNC_OFFSET(13430, gl_dispatch_stub_766, gl_dispatch_stub_766, NULL, _gloffset_ProgramEnvParameters4fvEXT), + NAME_FUNC_OFFSET(13459, gl_dispatch_stub_767, gl_dispatch_stub_767, NULL, _gloffset_ProgramLocalParameters4fvEXT), + NAME_FUNC_OFFSET(13490, gl_dispatch_stub_768, gl_dispatch_stub_768, NULL, _gloffset_GetQueryObjecti64vEXT), + NAME_FUNC_OFFSET(13514, gl_dispatch_stub_769, gl_dispatch_stub_769, NULL, _gloffset_GetQueryObjectui64vEXT), + NAME_FUNC_OFFSET(13539, glArrayElement, glArrayElement, NULL, _gloffset_ArrayElement), + NAME_FUNC_OFFSET(13557, glBindTexture, glBindTexture, NULL, _gloffset_BindTexture), + NAME_FUNC_OFFSET(13574, glDrawArrays, glDrawArrays, NULL, _gloffset_DrawArrays), + NAME_FUNC_OFFSET(13590, glAreTexturesResident, glAreTexturesResidentEXT, glAreTexturesResidentEXT, _gloffset_AreTexturesResident), + NAME_FUNC_OFFSET(13615, glCopyTexImage1D, glCopyTexImage1D, NULL, _gloffset_CopyTexImage1D), + NAME_FUNC_OFFSET(13635, glCopyTexImage2D, glCopyTexImage2D, NULL, _gloffset_CopyTexImage2D), + NAME_FUNC_OFFSET(13655, glCopyTexSubImage1D, glCopyTexSubImage1D, NULL, _gloffset_CopyTexSubImage1D), + NAME_FUNC_OFFSET(13678, glCopyTexSubImage2D, glCopyTexSubImage2D, NULL, _gloffset_CopyTexSubImage2D), + NAME_FUNC_OFFSET(13701, glDeleteTextures, glDeleteTexturesEXT, glDeleteTexturesEXT, _gloffset_DeleteTextures), + NAME_FUNC_OFFSET(13721, glGenTextures, glGenTexturesEXT, glGenTexturesEXT, _gloffset_GenTextures), + NAME_FUNC_OFFSET(13738, glGetPointerv, glGetPointerv, NULL, _gloffset_GetPointerv), + NAME_FUNC_OFFSET(13755, glIsTexture, glIsTextureEXT, glIsTextureEXT, _gloffset_IsTexture), + NAME_FUNC_OFFSET(13770, glPrioritizeTextures, glPrioritizeTextures, NULL, _gloffset_PrioritizeTextures), + NAME_FUNC_OFFSET(13794, glTexSubImage1D, glTexSubImage1D, NULL, _gloffset_TexSubImage1D), + NAME_FUNC_OFFSET(13813, glTexSubImage2D, glTexSubImage2D, NULL, _gloffset_TexSubImage2D), + NAME_FUNC_OFFSET(13832, glBlendColor, glBlendColor, NULL, _gloffset_BlendColor), + NAME_FUNC_OFFSET(13848, glBlendEquation, glBlendEquation, NULL, _gloffset_BlendEquation), + NAME_FUNC_OFFSET(13867, glDrawRangeElements, glDrawRangeElements, NULL, _gloffset_DrawRangeElements), + NAME_FUNC_OFFSET(13890, glColorTable, glColorTable, NULL, _gloffset_ColorTable), + NAME_FUNC_OFFSET(13906, glColorTable, glColorTable, NULL, _gloffset_ColorTable), + NAME_FUNC_OFFSET(13922, glColorTableParameterfv, glColorTableParameterfv, NULL, _gloffset_ColorTableParameterfv), + NAME_FUNC_OFFSET(13949, glColorTableParameteriv, glColorTableParameteriv, NULL, _gloffset_ColorTableParameteriv), + NAME_FUNC_OFFSET(13976, glCopyColorTable, glCopyColorTable, NULL, _gloffset_CopyColorTable), + NAME_FUNC_OFFSET(13996, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), + NAME_FUNC_OFFSET(14015, glGetColorTable, glGetColorTableEXT, glGetColorTableEXT, _gloffset_GetColorTable), + NAME_FUNC_OFFSET(14034, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), + NAME_FUNC_OFFSET(14064, glGetColorTableParameterfv, glGetColorTableParameterfvEXT, glGetColorTableParameterfvEXT, _gloffset_GetColorTableParameterfv), + NAME_FUNC_OFFSET(14094, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), + NAME_FUNC_OFFSET(14124, glGetColorTableParameteriv, glGetColorTableParameterivEXT, glGetColorTableParameterivEXT, _gloffset_GetColorTableParameteriv), + NAME_FUNC_OFFSET(14154, glColorSubTable, glColorSubTable, NULL, _gloffset_ColorSubTable), + NAME_FUNC_OFFSET(14173, glCopyColorSubTable, glCopyColorSubTable, NULL, _gloffset_CopyColorSubTable), + NAME_FUNC_OFFSET(14196, glConvolutionFilter1D, glConvolutionFilter1D, NULL, _gloffset_ConvolutionFilter1D), + NAME_FUNC_OFFSET(14221, glConvolutionFilter2D, glConvolutionFilter2D, NULL, _gloffset_ConvolutionFilter2D), + NAME_FUNC_OFFSET(14246, glConvolutionParameterf, glConvolutionParameterf, NULL, _gloffset_ConvolutionParameterf), + NAME_FUNC_OFFSET(14273, glConvolutionParameterfv, glConvolutionParameterfv, NULL, _gloffset_ConvolutionParameterfv), + NAME_FUNC_OFFSET(14301, glConvolutionParameteri, glConvolutionParameteri, NULL, _gloffset_ConvolutionParameteri), + NAME_FUNC_OFFSET(14328, glConvolutionParameteriv, glConvolutionParameteriv, NULL, _gloffset_ConvolutionParameteriv), + NAME_FUNC_OFFSET(14356, glCopyConvolutionFilter1D, glCopyConvolutionFilter1D, NULL, _gloffset_CopyConvolutionFilter1D), + NAME_FUNC_OFFSET(14385, glCopyConvolutionFilter2D, glCopyConvolutionFilter2D, NULL, _gloffset_CopyConvolutionFilter2D), + NAME_FUNC_OFFSET(14414, glGetConvolutionFilter, gl_dispatch_stub_356, gl_dispatch_stub_356, _gloffset_GetConvolutionFilter), + NAME_FUNC_OFFSET(14440, glGetConvolutionParameterfv, gl_dispatch_stub_357, gl_dispatch_stub_357, _gloffset_GetConvolutionParameterfv), + NAME_FUNC_OFFSET(14471, glGetConvolutionParameteriv, gl_dispatch_stub_358, gl_dispatch_stub_358, _gloffset_GetConvolutionParameteriv), + NAME_FUNC_OFFSET(14502, glGetSeparableFilter, gl_dispatch_stub_359, gl_dispatch_stub_359, _gloffset_GetSeparableFilter), + NAME_FUNC_OFFSET(14526, glSeparableFilter2D, glSeparableFilter2D, NULL, _gloffset_SeparableFilter2D), + NAME_FUNC_OFFSET(14549, glGetHistogram, gl_dispatch_stub_361, gl_dispatch_stub_361, _gloffset_GetHistogram), + NAME_FUNC_OFFSET(14567, glGetHistogramParameterfv, gl_dispatch_stub_362, gl_dispatch_stub_362, _gloffset_GetHistogramParameterfv), + NAME_FUNC_OFFSET(14596, glGetHistogramParameteriv, gl_dispatch_stub_363, gl_dispatch_stub_363, _gloffset_GetHistogramParameteriv), + NAME_FUNC_OFFSET(14625, glGetMinmax, gl_dispatch_stub_364, gl_dispatch_stub_364, _gloffset_GetMinmax), + NAME_FUNC_OFFSET(14640, glGetMinmaxParameterfv, gl_dispatch_stub_365, gl_dispatch_stub_365, _gloffset_GetMinmaxParameterfv), + NAME_FUNC_OFFSET(14666, glGetMinmaxParameteriv, gl_dispatch_stub_366, gl_dispatch_stub_366, _gloffset_GetMinmaxParameteriv), + NAME_FUNC_OFFSET(14692, glHistogram, glHistogram, NULL, _gloffset_Histogram), + NAME_FUNC_OFFSET(14707, glMinmax, glMinmax, NULL, _gloffset_Minmax), + NAME_FUNC_OFFSET(14719, glResetHistogram, glResetHistogram, NULL, _gloffset_ResetHistogram), + NAME_FUNC_OFFSET(14739, glResetMinmax, glResetMinmax, NULL, _gloffset_ResetMinmax), + NAME_FUNC_OFFSET(14756, glTexImage3D, glTexImage3D, NULL, _gloffset_TexImage3D), + NAME_FUNC_OFFSET(14772, glTexSubImage3D, glTexSubImage3D, NULL, _gloffset_TexSubImage3D), + NAME_FUNC_OFFSET(14791, glCopyTexSubImage3D, glCopyTexSubImage3D, NULL, _gloffset_CopyTexSubImage3D), + NAME_FUNC_OFFSET(14814, glActiveTextureARB, glActiveTextureARB, NULL, _gloffset_ActiveTextureARB), + NAME_FUNC_OFFSET(14830, glClientActiveTextureARB, glClientActiveTextureARB, NULL, _gloffset_ClientActiveTextureARB), + NAME_FUNC_OFFSET(14852, glMultiTexCoord1dARB, glMultiTexCoord1dARB, NULL, _gloffset_MultiTexCoord1dARB), + NAME_FUNC_OFFSET(14870, glMultiTexCoord1dvARB, glMultiTexCoord1dvARB, NULL, _gloffset_MultiTexCoord1dvARB), + NAME_FUNC_OFFSET(14889, glMultiTexCoord1fARB, glMultiTexCoord1fARB, NULL, _gloffset_MultiTexCoord1fARB), + NAME_FUNC_OFFSET(14907, glMultiTexCoord1fvARB, glMultiTexCoord1fvARB, NULL, _gloffset_MultiTexCoord1fvARB), + NAME_FUNC_OFFSET(14926, glMultiTexCoord1iARB, glMultiTexCoord1iARB, NULL, _gloffset_MultiTexCoord1iARB), + NAME_FUNC_OFFSET(14944, glMultiTexCoord1ivARB, glMultiTexCoord1ivARB, NULL, _gloffset_MultiTexCoord1ivARB), + NAME_FUNC_OFFSET(14963, glMultiTexCoord1sARB, glMultiTexCoord1sARB, NULL, _gloffset_MultiTexCoord1sARB), + NAME_FUNC_OFFSET(14981, glMultiTexCoord1svARB, glMultiTexCoord1svARB, NULL, _gloffset_MultiTexCoord1svARB), + NAME_FUNC_OFFSET(15000, glMultiTexCoord2dARB, glMultiTexCoord2dARB, NULL, _gloffset_MultiTexCoord2dARB), + NAME_FUNC_OFFSET(15018, glMultiTexCoord2dvARB, glMultiTexCoord2dvARB, NULL, _gloffset_MultiTexCoord2dvARB), + NAME_FUNC_OFFSET(15037, glMultiTexCoord2fARB, glMultiTexCoord2fARB, NULL, _gloffset_MultiTexCoord2fARB), + NAME_FUNC_OFFSET(15055, glMultiTexCoord2fvARB, glMultiTexCoord2fvARB, NULL, _gloffset_MultiTexCoord2fvARB), + NAME_FUNC_OFFSET(15074, glMultiTexCoord2iARB, glMultiTexCoord2iARB, NULL, _gloffset_MultiTexCoord2iARB), + NAME_FUNC_OFFSET(15092, glMultiTexCoord2ivARB, glMultiTexCoord2ivARB, NULL, _gloffset_MultiTexCoord2ivARB), + NAME_FUNC_OFFSET(15111, glMultiTexCoord2sARB, glMultiTexCoord2sARB, NULL, _gloffset_MultiTexCoord2sARB), + NAME_FUNC_OFFSET(15129, glMultiTexCoord2svARB, glMultiTexCoord2svARB, NULL, _gloffset_MultiTexCoord2svARB), + NAME_FUNC_OFFSET(15148, glMultiTexCoord3dARB, glMultiTexCoord3dARB, NULL, _gloffset_MultiTexCoord3dARB), + NAME_FUNC_OFFSET(15166, glMultiTexCoord3dvARB, glMultiTexCoord3dvARB, NULL, _gloffset_MultiTexCoord3dvARB), + NAME_FUNC_OFFSET(15185, glMultiTexCoord3fARB, glMultiTexCoord3fARB, NULL, _gloffset_MultiTexCoord3fARB), + NAME_FUNC_OFFSET(15203, glMultiTexCoord3fvARB, glMultiTexCoord3fvARB, NULL, _gloffset_MultiTexCoord3fvARB), + NAME_FUNC_OFFSET(15222, glMultiTexCoord3iARB, glMultiTexCoord3iARB, NULL, _gloffset_MultiTexCoord3iARB), + NAME_FUNC_OFFSET(15240, glMultiTexCoord3ivARB, glMultiTexCoord3ivARB, NULL, _gloffset_MultiTexCoord3ivARB), + NAME_FUNC_OFFSET(15259, glMultiTexCoord3sARB, glMultiTexCoord3sARB, NULL, _gloffset_MultiTexCoord3sARB), + NAME_FUNC_OFFSET(15277, glMultiTexCoord3svARB, glMultiTexCoord3svARB, NULL, _gloffset_MultiTexCoord3svARB), + NAME_FUNC_OFFSET(15296, glMultiTexCoord4dARB, glMultiTexCoord4dARB, NULL, _gloffset_MultiTexCoord4dARB), + NAME_FUNC_OFFSET(15314, glMultiTexCoord4dvARB, glMultiTexCoord4dvARB, NULL, _gloffset_MultiTexCoord4dvARB), + NAME_FUNC_OFFSET(15333, glMultiTexCoord4fARB, glMultiTexCoord4fARB, NULL, _gloffset_MultiTexCoord4fARB), + NAME_FUNC_OFFSET(15351, glMultiTexCoord4fvARB, glMultiTexCoord4fvARB, NULL, _gloffset_MultiTexCoord4fvARB), + NAME_FUNC_OFFSET(15370, glMultiTexCoord4iARB, glMultiTexCoord4iARB, NULL, _gloffset_MultiTexCoord4iARB), + NAME_FUNC_OFFSET(15388, glMultiTexCoord4ivARB, glMultiTexCoord4ivARB, NULL, _gloffset_MultiTexCoord4ivARB), + NAME_FUNC_OFFSET(15407, glMultiTexCoord4sARB, glMultiTexCoord4sARB, NULL, _gloffset_MultiTexCoord4sARB), + NAME_FUNC_OFFSET(15425, glMultiTexCoord4svARB, glMultiTexCoord4svARB, NULL, _gloffset_MultiTexCoord4svARB), + NAME_FUNC_OFFSET(15444, glStencilOpSeparate, glStencilOpSeparate, NULL, _gloffset_StencilOpSeparate), + NAME_FUNC_OFFSET(15467, glLoadTransposeMatrixdARB, glLoadTransposeMatrixdARB, NULL, _gloffset_LoadTransposeMatrixdARB), + NAME_FUNC_OFFSET(15490, glLoadTransposeMatrixfARB, glLoadTransposeMatrixfARB, NULL, _gloffset_LoadTransposeMatrixfARB), + NAME_FUNC_OFFSET(15513, glMultTransposeMatrixdARB, glMultTransposeMatrixdARB, NULL, _gloffset_MultTransposeMatrixdARB), + NAME_FUNC_OFFSET(15536, glMultTransposeMatrixfARB, glMultTransposeMatrixfARB, NULL, _gloffset_MultTransposeMatrixfARB), + NAME_FUNC_OFFSET(15559, glSampleCoverageARB, glSampleCoverageARB, NULL, _gloffset_SampleCoverageARB), + NAME_FUNC_OFFSET(15576, glCompressedTexImage1DARB, glCompressedTexImage1DARB, NULL, _gloffset_CompressedTexImage1DARB), + NAME_FUNC_OFFSET(15599, glCompressedTexImage2DARB, glCompressedTexImage2DARB, NULL, _gloffset_CompressedTexImage2DARB), + NAME_FUNC_OFFSET(15622, glCompressedTexImage3DARB, glCompressedTexImage3DARB, NULL, _gloffset_CompressedTexImage3DARB), + NAME_FUNC_OFFSET(15645, glCompressedTexSubImage1DARB, glCompressedTexSubImage1DARB, NULL, _gloffset_CompressedTexSubImage1DARB), + NAME_FUNC_OFFSET(15671, glCompressedTexSubImage2DARB, glCompressedTexSubImage2DARB, NULL, _gloffset_CompressedTexSubImage2DARB), + NAME_FUNC_OFFSET(15697, glCompressedTexSubImage3DARB, glCompressedTexSubImage3DARB, NULL, _gloffset_CompressedTexSubImage3DARB), + NAME_FUNC_OFFSET(15723, glGetCompressedTexImageARB, glGetCompressedTexImageARB, NULL, _gloffset_GetCompressedTexImageARB), + NAME_FUNC_OFFSET(15747, glDisableVertexAttribArrayARB, glDisableVertexAttribArrayARB, NULL, _gloffset_DisableVertexAttribArrayARB), + NAME_FUNC_OFFSET(15774, glEnableVertexAttribArrayARB, glEnableVertexAttribArrayARB, NULL, _gloffset_EnableVertexAttribArrayARB), + NAME_FUNC_OFFSET(15800, glGetVertexAttribdvARB, glGetVertexAttribdvARB, NULL, _gloffset_GetVertexAttribdvARB), + NAME_FUNC_OFFSET(15820, glGetVertexAttribfvARB, glGetVertexAttribfvARB, NULL, _gloffset_GetVertexAttribfvARB), + NAME_FUNC_OFFSET(15840, glGetVertexAttribivARB, glGetVertexAttribivARB, NULL, _gloffset_GetVertexAttribivARB), + NAME_FUNC_OFFSET(15860, glProgramEnvParameter4dARB, glProgramEnvParameter4dARB, NULL, _gloffset_ProgramEnvParameter4dARB), + NAME_FUNC_OFFSET(15883, glProgramEnvParameter4dvARB, glProgramEnvParameter4dvARB, NULL, _gloffset_ProgramEnvParameter4dvARB), + NAME_FUNC_OFFSET(15907, glProgramEnvParameter4fARB, glProgramEnvParameter4fARB, NULL, _gloffset_ProgramEnvParameter4fARB), + NAME_FUNC_OFFSET(15930, glProgramEnvParameter4fvARB, glProgramEnvParameter4fvARB, NULL, _gloffset_ProgramEnvParameter4fvARB), NAME_FUNC_OFFSET(15954, glVertexAttrib1dARB, glVertexAttrib1dARB, NULL, _gloffset_VertexAttrib1dARB), NAME_FUNC_OFFSET(15971, glVertexAttrib1dvARB, glVertexAttrib1dvARB, NULL, _gloffset_VertexAttrib1dvARB), NAME_FUNC_OFFSET(15989, glVertexAttrib1fARB, glVertexAttrib1fARB, NULL, _gloffset_VertexAttrib1fARB), @@ -2252,8 +2252,8 @@ static const glprocs_table_t static_functions[] = { NAME_FUNC_OFFSET(18694, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV), NAME_FUNC_OFFSET(18709, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV), NAME_FUNC_OFFSET(18727, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV), - NAME_FUNC_OFFSET(18746, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT), - NAME_FUNC_OFFSET(18770, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT), + NAME_FUNC_OFFSET(18746, gl_dispatch_stub_745, gl_dispatch_stub_745, NULL, _gloffset_BlendEquationSeparateEXT), + NAME_FUNC_OFFSET(18770, gl_dispatch_stub_745, gl_dispatch_stub_745, NULL, _gloffset_BlendEquationSeparateEXT), NAME_FUNC_OFFSET(-1, NULL, NULL, NULL, 0) }; diff --git a/src/mesa/sparc/glapi_sparc.S b/src/mesa/sparc/glapi_sparc.S index b214bef54d..3d1d55fd76 100644 --- a/src/mesa/sparc/glapi_sparc.S +++ b/src/mesa/sparc/glapi_sparc.S @@ -738,10 +738,6 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */ .globl glGetVertexAttribivNV ; .type glGetVertexAttribivNV,#function .globl glIsProgramNV ; .type glIsProgramNV,#function .globl glLoadProgramNV ; .type glLoadProgramNV,#function - .globl glProgramParameter4dNV ; .type glProgramParameter4dNV,#function - .globl glProgramParameter4dvNV ; .type glProgramParameter4dvNV,#function - .globl glProgramParameter4fNV ; .type glProgramParameter4fNV,#function - .globl glProgramParameter4fvNV ; .type glProgramParameter4fvNV,#function .globl glProgramParameters4dvNV ; .type glProgramParameters4dvNV,#function .globl glProgramParameters4fvNV ; .type glProgramParameters4fvNV,#function .globl glRequestResidentProgramsNV ; .type glRequestResidentProgramsNV,#function @@ -802,19 +798,19 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */ .globl glSetFragmentShaderConstantATI ; .type glSetFragmentShaderConstantATI,#function .globl glPointParameteriNV ; .type glPointParameteriNV,#function .globl glPointParameterivNV ; .type glPointParameterivNV,#function + .globl gl_dispatch_stub_733 ; .type gl_dispatch_stub_733,#function + .globl gl_dispatch_stub_734 ; .type gl_dispatch_stub_734,#function + .globl gl_dispatch_stub_735 ; .type gl_dispatch_stub_735,#function + .globl gl_dispatch_stub_736 ; .type gl_dispatch_stub_736,#function .globl gl_dispatch_stub_737 ; .type gl_dispatch_stub_737,#function - .globl gl_dispatch_stub_738 ; .type gl_dispatch_stub_738,#function - .globl gl_dispatch_stub_739 ; .type gl_dispatch_stub_739,#function - .globl gl_dispatch_stub_740 ; .type gl_dispatch_stub_740,#function - .globl gl_dispatch_stub_741 ; .type gl_dispatch_stub_741,#function .globl glGetProgramNamedParameterdvNV ; .type glGetProgramNamedParameterdvNV,#function .globl glGetProgramNamedParameterfvNV ; .type glGetProgramNamedParameterfvNV,#function .globl glProgramNamedParameter4dNV ; .type glProgramNamedParameter4dNV,#function .globl glProgramNamedParameter4dvNV ; .type glProgramNamedParameter4dvNV,#function .globl glProgramNamedParameter4fNV ; .type glProgramNamedParameter4fNV,#function .globl glProgramNamedParameter4fvNV ; .type glProgramNamedParameter4fvNV,#function - .globl gl_dispatch_stub_748 ; .type gl_dispatch_stub_748,#function - .globl gl_dispatch_stub_749 ; .type gl_dispatch_stub_749,#function + .globl gl_dispatch_stub_744 ; .type gl_dispatch_stub_744,#function + .globl gl_dispatch_stub_745 ; .type gl_dispatch_stub_745,#function .globl glBindFramebufferEXT ; .type glBindFramebufferEXT,#function .globl glBindRenderbufferEXT ; .type glBindRenderbufferEXT,#function .globl glCheckFramebufferStatusEXT ; .type glCheckFramebufferStatusEXT,#function @@ -832,13 +828,13 @@ __glapi_sparc_icache_flush: /* %o0 = insn_addr */ .globl glIsFramebufferEXT ; .type glIsFramebufferEXT,#function .globl glIsRenderbufferEXT ; .type glIsRenderbufferEXT,#function .globl glRenderbufferStorageEXT ; .type glRenderbufferStorageEXT,#function - .globl gl_dispatch_stub_767 ; .type gl_dispatch_stub_767,#function + .globl gl_dispatch_stub_763 ; .type gl_dispatch_stub_763,#function .globl glFramebufferTextureLayerEXT ; .type glFramebufferTextureLayerEXT,#function + .globl gl_dispatch_stub_765 ; .type gl_dispatch_stub_765,#function + .globl gl_dispatch_stub_766 ; .type gl_dispatch_stub_766,#function + .globl gl_dispatch_stub_767 ; .type gl_dispatch_stub_767,#function + .globl gl_dispatch_stub_768 ; .type gl_dispatch_stub_768,#function .globl gl_dispatch_stub_769 ; .type gl_dispatch_stub_769,#function - .globl gl_dispatch_stub_770 ; .type gl_dispatch_stub_770,#function - .globl gl_dispatch_stub_771 ; .type gl_dispatch_stub_771,#function - .globl gl_dispatch_stub_772 ; .type gl_dispatch_stub_772,#function - .globl gl_dispatch_stub_773 ; .type gl_dispatch_stub_773,#function .globl _mesa_sparc_glapi_begin ; .type _mesa_sparc_glapi_begin,#function _mesa_sparc_glapi_begin: @@ -1515,10 +1511,6 @@ _mesa_sparc_glapi_begin: GL_STUB(glGetVertexAttribivNV, _gloffset_GetVertexAttribivNV) GL_STUB(glIsProgramNV, _gloffset_IsProgramNV) GL_STUB(glLoadProgramNV, _gloffset_LoadProgramNV) - GL_STUB(glProgramParameter4dNV, _gloffset_ProgramParameter4dNV) - GL_STUB(glProgramParameter4dvNV, _gloffset_ProgramParameter4dvNV) - GL_STUB(glProgramParameter4fNV, _gloffset_ProgramParameter4fNV) - GL_STUB(glProgramParameter4fvNV, _gloffset_ProgramParameter4fvNV) GL_STUB(glProgramParameters4dvNV, _gloffset_ProgramParameters4dvNV) GL_STUB(glProgramParameters4fvNV, _gloffset_ProgramParameters4fvNV) GL_STUB(glRequestResidentProgramsNV, _gloffset_RequestResidentProgramsNV) @@ -1579,19 +1571,19 @@ _mesa_sparc_glapi_begin: GL_STUB(glSetFragmentShaderConstantATI, _gloffset_SetFragmentShaderConstantATI) GL_STUB(glPointParameteriNV, _gloffset_PointParameteriNV) GL_STUB(glPointParameterivNV, _gloffset_PointParameterivNV) + GL_STUB(gl_dispatch_stub_733, _gloffset__dispatch_stub_733) + GL_STUB(gl_dispatch_stub_734, _gloffset__dispatch_stub_734) + GL_STUB(gl_dispatch_stub_735, _gloffset__dispatch_stub_735) + GL_STUB(gl_dispatch_stub_736, _gloffset__dispatch_stub_736) GL_STUB(gl_dispatch_stub_737, _gloffset__dispatch_stub_737) - GL_STUB(gl_dispatch_stub_738, _gloffset__dispatch_stub_738) - GL_STUB(gl_dispatch_stub_739, _gloffset__dispatch_stub_739) - GL_STUB(gl_dispatch_stub_740, _gloffset__dispatch_stub_740) - GL_STUB(gl_dispatch_stub_741, _gloffset__dispatch_stub_741) GL_STUB(glGetProgramNamedParameterdvNV, _gloffset_GetProgramNamedParameterdvNV) GL_STUB(glGetProgramNamedParameterfvNV, _gloffset_GetProgramNamedParameterfvNV) GL_STUB(glProgramNamedParameter4dNV, _gloffset_ProgramNamedParameter4dNV) GL_STUB(glProgramNamedParameter4dvNV, _gloffset_ProgramNamedParameter4dvNV) GL_STUB(glProgramNamedParameter4fNV, _gloffset_ProgramNamedParameter4fNV) GL_STUB(glProgramNamedParameter4fvNV, _gloffset_ProgramNamedParameter4fvNV) - GL_STUB(gl_dispatch_stub_748, _gloffset__dispatch_stub_748) - GL_STUB(gl_dispatch_stub_749, _gloffset__dispatch_stub_749) + GL_STUB(gl_dispatch_stub_744, _gloffset__dispatch_stub_744) + GL_STUB(gl_dispatch_stub_745, _gloffset__dispatch_stub_745) GL_STUB(glBindFramebufferEXT, _gloffset_BindFramebufferEXT) GL_STUB(glBindRenderbufferEXT, _gloffset_BindRenderbufferEXT) GL_STUB(glCheckFramebufferStatusEXT, _gloffset_CheckFramebufferStatusEXT) @@ -1609,13 +1601,13 @@ _mesa_sparc_glapi_begin: GL_STUB(glIsFramebufferEXT, _gloffset_IsFramebufferEXT) GL_STUB(glIsRenderbufferEXT, _gloffset_IsRenderbufferEXT) GL_STUB(glRenderbufferStorageEXT, _gloffset_RenderbufferStorageEXT) - GL_STUB(gl_dispatch_stub_767, _gloffset__dispatch_stub_767) + GL_STUB(gl_dispatch_stub_763, _gloffset__dispatch_stub_763) GL_STUB(glFramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT) + GL_STUB(gl_dispatch_stub_765, _gloffset__dispatch_stub_765) + GL_STUB(gl_dispatch_stub_766, _gloffset__dispatch_stub_766) + GL_STUB(gl_dispatch_stub_767, _gloffset__dispatch_stub_767) + GL_STUB(gl_dispatch_stub_768, _gloffset__dispatch_stub_768) GL_STUB(gl_dispatch_stub_769, _gloffset__dispatch_stub_769) - GL_STUB(gl_dispatch_stub_770, _gloffset__dispatch_stub_770) - GL_STUB(gl_dispatch_stub_771, _gloffset__dispatch_stub_771) - GL_STUB(gl_dispatch_stub_772, _gloffset__dispatch_stub_772) - GL_STUB(gl_dispatch_stub_773, _gloffset__dispatch_stub_773) .globl _mesa_sparc_glapi_end ; .type _mesa_sparc_glapi_end,#function _mesa_sparc_glapi_end: @@ -1710,6 +1702,10 @@ _mesa_sparc_glapi_end: .globl glGetVertexAttribdv ; .type glGetVertexAttribdv,#function ; glGetVertexAttribdv = glGetVertexAttribdvARB .globl glGetVertexAttribfv ; .type glGetVertexAttribfv,#function ; glGetVertexAttribfv = glGetVertexAttribfvARB .globl glGetVertexAttribiv ; .type glGetVertexAttribiv,#function ; glGetVertexAttribiv = glGetVertexAttribivARB + .globl glProgramParameter4dNV ; .type glProgramParameter4dNV,#function ; glProgramParameter4dNV = glProgramEnvParameter4dARB + .globl glProgramParameter4dvNV ; .type glProgramParameter4dvNV,#function ; glProgramParameter4dvNV = glProgramEnvParameter4dvARB + .globl glProgramParameter4fNV ; .type glProgramParameter4fNV,#function ; glProgramParameter4fNV = glProgramEnvParameter4fARB + .globl glProgramParameter4fvNV ; .type glProgramParameter4fvNV,#function ; glProgramParameter4fvNV = glProgramEnvParameter4fvARB .globl glVertexAttrib1d ; .type glVertexAttrib1d,#function ; glVertexAttrib1d = glVertexAttrib1dARB .globl glVertexAttrib1dv ; .type glVertexAttrib1dv,#function ; glVertexAttrib1dv = glVertexAttrib1dvARB .globl glVertexAttrib1f ; .type glVertexAttrib1f,#function ; glVertexAttrib1f = glVertexAttrib1fARB diff --git a/src/mesa/x86-64/glapi_x86-64.S b/src/mesa/x86-64/glapi_x86-64.S index a239a6749d..cb34061b36 100644 --- a/src/mesa/x86-64/glapi_x86-64.S +++ b/src/mesa/x86-64/glapi_x86-64.S @@ -25267,193 +25267,13 @@ GL_PREFIX(LoadProgramNV): #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(LoadProgramNV), .-GL_PREFIX(LoadProgramNV) - .p2align 4,,15 - .globl GL_PREFIX(ProgramParameter4dNV) - .type GL_PREFIX(ProgramParameter4dNV), @function -GL_PREFIX(ProgramParameter4dNV): -#if defined(GLX_USE_TLS) - call _x86_64_get_dispatch@PLT - movq 5384(%rax), %r11 - jmp *%r11 -#elif defined(PTHREADS) - subq $56, %rsp - movq %rdi, (%rsp) - movq %rsi, 8(%rsp) - movq %xmm0, 16(%rsp) - movq %xmm1, 24(%rsp) - movq %xmm2, 32(%rsp) - movq %xmm3, 40(%rsp) - call _x86_64_get_dispatch@PLT - movq 40(%rsp), %xmm3 - movq 32(%rsp), %xmm2 - movq 24(%rsp), %xmm1 - movq 16(%rsp), %xmm0 - movq 8(%rsp), %rsi - movq (%rsp), %rdi - addq $56, %rsp - movq 5384(%rax), %r11 - jmp *%r11 -#else - movq _glapi_Dispatch(%rip), %rax - testq %rax, %rax - je 1f - movq 5384(%rax), %r11 - jmp *%r11 -1: - subq $56, %rsp - movq %rdi, (%rsp) - movq %rsi, 8(%rsp) - movq %xmm0, 16(%rsp) - movq %xmm1, 24(%rsp) - movq %xmm2, 32(%rsp) - movq %xmm3, 40(%rsp) - call _glapi_get_dispatch - movq 40(%rsp), %xmm3 - movq 32(%rsp), %xmm2 - movq 24(%rsp), %xmm1 - movq 16(%rsp), %xmm0 - movq 8(%rsp), %rsi - movq (%rsp), %rdi - addq $56, %rsp - movq 5384(%rax), %r11 - jmp *%r11 -#endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(ProgramParameter4dNV), .-GL_PREFIX(ProgramParameter4dNV) - - .p2align 4,,15 - .globl GL_PREFIX(ProgramParameter4dvNV) - .type GL_PREFIX(ProgramParameter4dvNV), @function -GL_PREFIX(ProgramParameter4dvNV): -#if defined(GLX_USE_TLS) - call _x86_64_get_dispatch@PLT - movq 5392(%rax), %r11 - jmp *%r11 -#elif defined(PTHREADS) - pushq %rdi - pushq %rsi - pushq %rdx - call _x86_64_get_dispatch@PLT - popq %rdx - popq %rsi - popq %rdi - movq 5392(%rax), %r11 - jmp *%r11 -#else - movq _glapi_Dispatch(%rip), %rax - testq %rax, %rax - je 1f - movq 5392(%rax), %r11 - jmp *%r11 -1: - pushq %rdi - pushq %rsi - pushq %rdx - call _glapi_get_dispatch - popq %rdx - popq %rsi - popq %rdi - movq 5392(%rax), %r11 - jmp *%r11 -#endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(ProgramParameter4dvNV), .-GL_PREFIX(ProgramParameter4dvNV) - - .p2align 4,,15 - .globl GL_PREFIX(ProgramParameter4fNV) - .type GL_PREFIX(ProgramParameter4fNV), @function -GL_PREFIX(ProgramParameter4fNV): -#if defined(GLX_USE_TLS) - call _x86_64_get_dispatch@PLT - movq 5400(%rax), %r11 - jmp *%r11 -#elif defined(PTHREADS) - subq $56, %rsp - movq %rdi, (%rsp) - movq %rsi, 8(%rsp) - movq %xmm0, 16(%rsp) - movq %xmm1, 24(%rsp) - movq %xmm2, 32(%rsp) - movq %xmm3, 40(%rsp) - call _x86_64_get_dispatch@PLT - movq 40(%rsp), %xmm3 - movq 32(%rsp), %xmm2 - movq 24(%rsp), %xmm1 - movq 16(%rsp), %xmm0 - movq 8(%rsp), %rsi - movq (%rsp), %rdi - addq $56, %rsp - movq 5400(%rax), %r11 - jmp *%r11 -#else - movq _glapi_Dispatch(%rip), %rax - testq %rax, %rax - je 1f - movq 5400(%rax), %r11 - jmp *%r11 -1: - subq $56, %rsp - movq %rdi, (%rsp) - movq %rsi, 8(%rsp) - movq %xmm0, 16(%rsp) - movq %xmm1, 24(%rsp) - movq %xmm2, 32(%rsp) - movq %xmm3, 40(%rsp) - call _glapi_get_dispatch - movq 40(%rsp), %xmm3 - movq 32(%rsp), %xmm2 - movq 24(%rsp), %xmm1 - movq 16(%rsp), %xmm0 - movq 8(%rsp), %rsi - movq (%rsp), %rdi - addq $56, %rsp - movq 5400(%rax), %r11 - jmp *%r11 -#endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(ProgramParameter4fNV), .-GL_PREFIX(ProgramParameter4fNV) - - .p2align 4,,15 - .globl GL_PREFIX(ProgramParameter4fvNV) - .type GL_PREFIX(ProgramParameter4fvNV), @function -GL_PREFIX(ProgramParameter4fvNV): -#if defined(GLX_USE_TLS) - call _x86_64_get_dispatch@PLT - movq 5408(%rax), %r11 - jmp *%r11 -#elif defined(PTHREADS) - pushq %rdi - pushq %rsi - pushq %rdx - call _x86_64_get_dispatch@PLT - popq %rdx - popq %rsi - popq %rdi - movq 5408(%rax), %r11 - jmp *%r11 -#else - movq _glapi_Dispatch(%rip), %rax - testq %rax, %rax - je 1f - movq 5408(%rax), %r11 - jmp *%r11 -1: - pushq %rdi - pushq %rsi - pushq %rdx - call _glapi_get_dispatch - popq %rdx - popq %rsi - popq %rdi - movq 5408(%rax), %r11 - jmp *%r11 -#endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(ProgramParameter4fvNV), .-GL_PREFIX(ProgramParameter4fvNV) - .p2align 4,,15 .globl GL_PREFIX(ProgramParameters4dvNV) .type GL_PREFIX(ProgramParameters4dvNV), @function GL_PREFIX(ProgramParameters4dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5416(%rax), %r11 + movq 5384(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25467,13 +25287,13 @@ GL_PREFIX(ProgramParameters4dvNV): popq %rdx popq %rsi popq %rdi - movq 5416(%rax), %r11 + movq 5384(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5416(%rax), %r11 + movq 5384(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25487,7 +25307,7 @@ GL_PREFIX(ProgramParameters4dvNV): popq %rdx popq %rsi popq %rdi - movq 5416(%rax), %r11 + movq 5384(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(ProgramParameters4dvNV), .-GL_PREFIX(ProgramParameters4dvNV) @@ -25498,7 +25318,7 @@ GL_PREFIX(ProgramParameters4dvNV): GL_PREFIX(ProgramParameters4fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5424(%rax), %r11 + movq 5392(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25512,13 +25332,13 @@ GL_PREFIX(ProgramParameters4fvNV): popq %rdx popq %rsi popq %rdi - movq 5424(%rax), %r11 + movq 5392(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5424(%rax), %r11 + movq 5392(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25532,7 +25352,7 @@ GL_PREFIX(ProgramParameters4fvNV): popq %rdx popq %rsi popq %rdi - movq 5424(%rax), %r11 + movq 5392(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(ProgramParameters4fvNV), .-GL_PREFIX(ProgramParameters4fvNV) @@ -25543,7 +25363,7 @@ GL_PREFIX(ProgramParameters4fvNV): GL_PREFIX(RequestResidentProgramsNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5432(%rax), %r11 + movq 5400(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25553,13 +25373,13 @@ GL_PREFIX(RequestResidentProgramsNV): popq %rbp popq %rsi popq %rdi - movq 5432(%rax), %r11 + movq 5400(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5432(%rax), %r11 + movq 5400(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25569,7 +25389,7 @@ GL_PREFIX(RequestResidentProgramsNV): popq %rbp popq %rsi popq %rdi - movq 5432(%rax), %r11 + movq 5400(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(RequestResidentProgramsNV), .-GL_PREFIX(RequestResidentProgramsNV) @@ -25580,7 +25400,7 @@ GL_PREFIX(RequestResidentProgramsNV): GL_PREFIX(TrackMatrixNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5440(%rax), %r11 + movq 5408(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25594,13 +25414,13 @@ GL_PREFIX(TrackMatrixNV): popq %rdx popq %rsi popq %rdi - movq 5440(%rax), %r11 + movq 5408(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5440(%rax), %r11 + movq 5408(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25614,7 +25434,7 @@ GL_PREFIX(TrackMatrixNV): popq %rdx popq %rsi popq %rdi - movq 5440(%rax), %r11 + movq 5408(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(TrackMatrixNV), .-GL_PREFIX(TrackMatrixNV) @@ -25625,7 +25445,7 @@ GL_PREFIX(TrackMatrixNV): GL_PREFIX(VertexAttrib1dNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5448(%rax), %r11 + movq 5416(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $24, %rsp @@ -25635,13 +25455,13 @@ GL_PREFIX(VertexAttrib1dNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $24, %rsp - movq 5448(%rax), %r11 + movq 5416(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5448(%rax), %r11 + movq 5416(%rax), %r11 jmp *%r11 1: subq $24, %rsp @@ -25651,7 +25471,7 @@ GL_PREFIX(VertexAttrib1dNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $24, %rsp - movq 5448(%rax), %r11 + movq 5416(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib1dNV), .-GL_PREFIX(VertexAttrib1dNV) @@ -25662,7 +25482,7 @@ GL_PREFIX(VertexAttrib1dNV): GL_PREFIX(VertexAttrib1dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5456(%rax), %r11 + movq 5424(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25672,13 +25492,13 @@ GL_PREFIX(VertexAttrib1dvNV): popq %rbp popq %rsi popq %rdi - movq 5456(%rax), %r11 + movq 5424(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5456(%rax), %r11 + movq 5424(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25688,7 +25508,7 @@ GL_PREFIX(VertexAttrib1dvNV): popq %rbp popq %rsi popq %rdi - movq 5456(%rax), %r11 + movq 5424(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib1dvNV), .-GL_PREFIX(VertexAttrib1dvNV) @@ -25699,7 +25519,7 @@ GL_PREFIX(VertexAttrib1dvNV): GL_PREFIX(VertexAttrib1fNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5464(%rax), %r11 + movq 5432(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $24, %rsp @@ -25709,13 +25529,13 @@ GL_PREFIX(VertexAttrib1fNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $24, %rsp - movq 5464(%rax), %r11 + movq 5432(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5464(%rax), %r11 + movq 5432(%rax), %r11 jmp *%r11 1: subq $24, %rsp @@ -25725,7 +25545,7 @@ GL_PREFIX(VertexAttrib1fNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $24, %rsp - movq 5464(%rax), %r11 + movq 5432(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib1fNV), .-GL_PREFIX(VertexAttrib1fNV) @@ -25736,7 +25556,7 @@ GL_PREFIX(VertexAttrib1fNV): GL_PREFIX(VertexAttrib1fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5472(%rax), %r11 + movq 5440(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25746,13 +25566,13 @@ GL_PREFIX(VertexAttrib1fvNV): popq %rbp popq %rsi popq %rdi - movq 5472(%rax), %r11 + movq 5440(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5472(%rax), %r11 + movq 5440(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25762,7 +25582,7 @@ GL_PREFIX(VertexAttrib1fvNV): popq %rbp popq %rsi popq %rdi - movq 5472(%rax), %r11 + movq 5440(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib1fvNV), .-GL_PREFIX(VertexAttrib1fvNV) @@ -25773,7 +25593,7 @@ GL_PREFIX(VertexAttrib1fvNV): GL_PREFIX(VertexAttrib1sNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5480(%rax), %r11 + movq 5448(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25783,13 +25603,13 @@ GL_PREFIX(VertexAttrib1sNV): popq %rbp popq %rsi popq %rdi - movq 5480(%rax), %r11 + movq 5448(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5480(%rax), %r11 + movq 5448(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25799,7 +25619,7 @@ GL_PREFIX(VertexAttrib1sNV): popq %rbp popq %rsi popq %rdi - movq 5480(%rax), %r11 + movq 5448(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib1sNV), .-GL_PREFIX(VertexAttrib1sNV) @@ -25810,7 +25630,7 @@ GL_PREFIX(VertexAttrib1sNV): GL_PREFIX(VertexAttrib1svNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5488(%rax), %r11 + movq 5456(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25820,13 +25640,13 @@ GL_PREFIX(VertexAttrib1svNV): popq %rbp popq %rsi popq %rdi - movq 5488(%rax), %r11 + movq 5456(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5488(%rax), %r11 + movq 5456(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25836,7 +25656,7 @@ GL_PREFIX(VertexAttrib1svNV): popq %rbp popq %rsi popq %rdi - movq 5488(%rax), %r11 + movq 5456(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib1svNV), .-GL_PREFIX(VertexAttrib1svNV) @@ -25847,7 +25667,7 @@ GL_PREFIX(VertexAttrib1svNV): GL_PREFIX(VertexAttrib2dNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5496(%rax), %r11 + movq 5464(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $24, %rsp @@ -25859,13 +25679,13 @@ GL_PREFIX(VertexAttrib2dNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $24, %rsp - movq 5496(%rax), %r11 + movq 5464(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5496(%rax), %r11 + movq 5464(%rax), %r11 jmp *%r11 1: subq $24, %rsp @@ -25877,7 +25697,7 @@ GL_PREFIX(VertexAttrib2dNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $24, %rsp - movq 5496(%rax), %r11 + movq 5464(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib2dNV), .-GL_PREFIX(VertexAttrib2dNV) @@ -25888,7 +25708,7 @@ GL_PREFIX(VertexAttrib2dNV): GL_PREFIX(VertexAttrib2dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5504(%rax), %r11 + movq 5472(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25898,13 +25718,13 @@ GL_PREFIX(VertexAttrib2dvNV): popq %rbp popq %rsi popq %rdi - movq 5504(%rax), %r11 + movq 5472(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5504(%rax), %r11 + movq 5472(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25914,7 +25734,7 @@ GL_PREFIX(VertexAttrib2dvNV): popq %rbp popq %rsi popq %rdi - movq 5504(%rax), %r11 + movq 5472(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib2dvNV), .-GL_PREFIX(VertexAttrib2dvNV) @@ -25925,7 +25745,7 @@ GL_PREFIX(VertexAttrib2dvNV): GL_PREFIX(VertexAttrib2fNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5512(%rax), %r11 + movq 5480(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $24, %rsp @@ -25937,13 +25757,13 @@ GL_PREFIX(VertexAttrib2fNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $24, %rsp - movq 5512(%rax), %r11 + movq 5480(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5512(%rax), %r11 + movq 5480(%rax), %r11 jmp *%r11 1: subq $24, %rsp @@ -25955,7 +25775,7 @@ GL_PREFIX(VertexAttrib2fNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $24, %rsp - movq 5512(%rax), %r11 + movq 5480(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib2fNV), .-GL_PREFIX(VertexAttrib2fNV) @@ -25966,7 +25786,7 @@ GL_PREFIX(VertexAttrib2fNV): GL_PREFIX(VertexAttrib2fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5520(%rax), %r11 + movq 5488(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -25976,13 +25796,13 @@ GL_PREFIX(VertexAttrib2fvNV): popq %rbp popq %rsi popq %rdi - movq 5520(%rax), %r11 + movq 5488(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5520(%rax), %r11 + movq 5488(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -25992,7 +25812,7 @@ GL_PREFIX(VertexAttrib2fvNV): popq %rbp popq %rsi popq %rdi - movq 5520(%rax), %r11 + movq 5488(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib2fvNV), .-GL_PREFIX(VertexAttrib2fvNV) @@ -26003,7 +25823,7 @@ GL_PREFIX(VertexAttrib2fvNV): GL_PREFIX(VertexAttrib2sNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5528(%rax), %r11 + movq 5496(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26013,13 +25833,13 @@ GL_PREFIX(VertexAttrib2sNV): popq %rdx popq %rsi popq %rdi - movq 5528(%rax), %r11 + movq 5496(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5528(%rax), %r11 + movq 5496(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26029,7 +25849,7 @@ GL_PREFIX(VertexAttrib2sNV): popq %rdx popq %rsi popq %rdi - movq 5528(%rax), %r11 + movq 5496(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib2sNV), .-GL_PREFIX(VertexAttrib2sNV) @@ -26040,7 +25860,7 @@ GL_PREFIX(VertexAttrib2sNV): GL_PREFIX(VertexAttrib2svNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5536(%rax), %r11 + movq 5504(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26050,13 +25870,13 @@ GL_PREFIX(VertexAttrib2svNV): popq %rbp popq %rsi popq %rdi - movq 5536(%rax), %r11 + movq 5504(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5536(%rax), %r11 + movq 5504(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26066,7 +25886,7 @@ GL_PREFIX(VertexAttrib2svNV): popq %rbp popq %rsi popq %rdi - movq 5536(%rax), %r11 + movq 5504(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib2svNV), .-GL_PREFIX(VertexAttrib2svNV) @@ -26077,7 +25897,7 @@ GL_PREFIX(VertexAttrib2svNV): GL_PREFIX(VertexAttrib3dNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5544(%rax), %r11 + movq 5512(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $40, %rsp @@ -26091,13 +25911,13 @@ GL_PREFIX(VertexAttrib3dNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $40, %rsp - movq 5544(%rax), %r11 + movq 5512(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5544(%rax), %r11 + movq 5512(%rax), %r11 jmp *%r11 1: subq $40, %rsp @@ -26111,7 +25931,7 @@ GL_PREFIX(VertexAttrib3dNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $40, %rsp - movq 5544(%rax), %r11 + movq 5512(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib3dNV), .-GL_PREFIX(VertexAttrib3dNV) @@ -26122,7 +25942,7 @@ GL_PREFIX(VertexAttrib3dNV): GL_PREFIX(VertexAttrib3dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5552(%rax), %r11 + movq 5520(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26132,13 +25952,13 @@ GL_PREFIX(VertexAttrib3dvNV): popq %rbp popq %rsi popq %rdi - movq 5552(%rax), %r11 + movq 5520(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5552(%rax), %r11 + movq 5520(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26148,7 +25968,7 @@ GL_PREFIX(VertexAttrib3dvNV): popq %rbp popq %rsi popq %rdi - movq 5552(%rax), %r11 + movq 5520(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib3dvNV), .-GL_PREFIX(VertexAttrib3dvNV) @@ -26159,7 +25979,7 @@ GL_PREFIX(VertexAttrib3dvNV): GL_PREFIX(VertexAttrib3fNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5560(%rax), %r11 + movq 5528(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $40, %rsp @@ -26173,13 +25993,13 @@ GL_PREFIX(VertexAttrib3fNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $40, %rsp - movq 5560(%rax), %r11 + movq 5528(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5560(%rax), %r11 + movq 5528(%rax), %r11 jmp *%r11 1: subq $40, %rsp @@ -26193,7 +26013,7 @@ GL_PREFIX(VertexAttrib3fNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $40, %rsp - movq 5560(%rax), %r11 + movq 5528(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib3fNV), .-GL_PREFIX(VertexAttrib3fNV) @@ -26204,7 +26024,7 @@ GL_PREFIX(VertexAttrib3fNV): GL_PREFIX(VertexAttrib3fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5568(%rax), %r11 + movq 5536(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26214,13 +26034,13 @@ GL_PREFIX(VertexAttrib3fvNV): popq %rbp popq %rsi popq %rdi - movq 5568(%rax), %r11 + movq 5536(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5568(%rax), %r11 + movq 5536(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26230,7 +26050,7 @@ GL_PREFIX(VertexAttrib3fvNV): popq %rbp popq %rsi popq %rdi - movq 5568(%rax), %r11 + movq 5536(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib3fvNV), .-GL_PREFIX(VertexAttrib3fvNV) @@ -26241,7 +26061,7 @@ GL_PREFIX(VertexAttrib3fvNV): GL_PREFIX(VertexAttrib3sNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5576(%rax), %r11 + movq 5544(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26255,13 +26075,13 @@ GL_PREFIX(VertexAttrib3sNV): popq %rdx popq %rsi popq %rdi - movq 5576(%rax), %r11 + movq 5544(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5576(%rax), %r11 + movq 5544(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26275,7 +26095,7 @@ GL_PREFIX(VertexAttrib3sNV): popq %rdx popq %rsi popq %rdi - movq 5576(%rax), %r11 + movq 5544(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib3sNV), .-GL_PREFIX(VertexAttrib3sNV) @@ -26286,7 +26106,7 @@ GL_PREFIX(VertexAttrib3sNV): GL_PREFIX(VertexAttrib3svNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5584(%rax), %r11 + movq 5552(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26296,13 +26116,13 @@ GL_PREFIX(VertexAttrib3svNV): popq %rbp popq %rsi popq %rdi - movq 5584(%rax), %r11 + movq 5552(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5584(%rax), %r11 + movq 5552(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26312,7 +26132,7 @@ GL_PREFIX(VertexAttrib3svNV): popq %rbp popq %rsi popq %rdi - movq 5584(%rax), %r11 + movq 5552(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib3svNV), .-GL_PREFIX(VertexAttrib3svNV) @@ -26323,7 +26143,7 @@ GL_PREFIX(VertexAttrib3svNV): GL_PREFIX(VertexAttrib4dNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5592(%rax), %r11 + movq 5560(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $40, %rsp @@ -26339,13 +26159,13 @@ GL_PREFIX(VertexAttrib4dNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $40, %rsp - movq 5592(%rax), %r11 + movq 5560(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5592(%rax), %r11 + movq 5560(%rax), %r11 jmp *%r11 1: subq $40, %rsp @@ -26361,7 +26181,7 @@ GL_PREFIX(VertexAttrib4dNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $40, %rsp - movq 5592(%rax), %r11 + movq 5560(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib4dNV), .-GL_PREFIX(VertexAttrib4dNV) @@ -26372,7 +26192,7 @@ GL_PREFIX(VertexAttrib4dNV): GL_PREFIX(VertexAttrib4dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5600(%rax), %r11 + movq 5568(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26382,13 +26202,13 @@ GL_PREFIX(VertexAttrib4dvNV): popq %rbp popq %rsi popq %rdi - movq 5600(%rax), %r11 + movq 5568(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5600(%rax), %r11 + movq 5568(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26398,7 +26218,7 @@ GL_PREFIX(VertexAttrib4dvNV): popq %rbp popq %rsi popq %rdi - movq 5600(%rax), %r11 + movq 5568(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib4dvNV), .-GL_PREFIX(VertexAttrib4dvNV) @@ -26409,7 +26229,7 @@ GL_PREFIX(VertexAttrib4dvNV): GL_PREFIX(VertexAttrib4fNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5608(%rax), %r11 + movq 5576(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $40, %rsp @@ -26425,13 +26245,13 @@ GL_PREFIX(VertexAttrib4fNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $40, %rsp - movq 5608(%rax), %r11 + movq 5576(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5608(%rax), %r11 + movq 5576(%rax), %r11 jmp *%r11 1: subq $40, %rsp @@ -26447,7 +26267,7 @@ GL_PREFIX(VertexAttrib4fNV): movq 8(%rsp), %xmm0 movq (%rsp), %rdi addq $40, %rsp - movq 5608(%rax), %r11 + movq 5576(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib4fNV), .-GL_PREFIX(VertexAttrib4fNV) @@ -26458,7 +26278,7 @@ GL_PREFIX(VertexAttrib4fNV): GL_PREFIX(VertexAttrib4fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5616(%rax), %r11 + movq 5584(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26468,13 +26288,13 @@ GL_PREFIX(VertexAttrib4fvNV): popq %rbp popq %rsi popq %rdi - movq 5616(%rax), %r11 + movq 5584(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5616(%rax), %r11 + movq 5584(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26484,7 +26304,7 @@ GL_PREFIX(VertexAttrib4fvNV): popq %rbp popq %rsi popq %rdi - movq 5616(%rax), %r11 + movq 5584(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib4fvNV), .-GL_PREFIX(VertexAttrib4fvNV) @@ -26495,7 +26315,7 @@ GL_PREFIX(VertexAttrib4fvNV): GL_PREFIX(VertexAttrib4sNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5624(%rax), %r11 + movq 5592(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26509,13 +26329,13 @@ GL_PREFIX(VertexAttrib4sNV): popq %rdx popq %rsi popq %rdi - movq 5624(%rax), %r11 + movq 5592(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5624(%rax), %r11 + movq 5592(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26529,7 +26349,7 @@ GL_PREFIX(VertexAttrib4sNV): popq %rdx popq %rsi popq %rdi - movq 5624(%rax), %r11 + movq 5592(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib4sNV), .-GL_PREFIX(VertexAttrib4sNV) @@ -26540,7 +26360,7 @@ GL_PREFIX(VertexAttrib4sNV): GL_PREFIX(VertexAttrib4svNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5632(%rax), %r11 + movq 5600(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26550,13 +26370,13 @@ GL_PREFIX(VertexAttrib4svNV): popq %rbp popq %rsi popq %rdi - movq 5632(%rax), %r11 + movq 5600(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5632(%rax), %r11 + movq 5600(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26566,7 +26386,7 @@ GL_PREFIX(VertexAttrib4svNV): popq %rbp popq %rsi popq %rdi - movq 5632(%rax), %r11 + movq 5600(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib4svNV), .-GL_PREFIX(VertexAttrib4svNV) @@ -26577,7 +26397,7 @@ GL_PREFIX(VertexAttrib4svNV): GL_PREFIX(VertexAttrib4ubNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5640(%rax), %r11 + movq 5608(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26591,13 +26411,13 @@ GL_PREFIX(VertexAttrib4ubNV): popq %rdx popq %rsi popq %rdi - movq 5640(%rax), %r11 + movq 5608(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5640(%rax), %r11 + movq 5608(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26611,7 +26431,7 @@ GL_PREFIX(VertexAttrib4ubNV): popq %rdx popq %rsi popq %rdi - movq 5640(%rax), %r11 + movq 5608(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib4ubNV), .-GL_PREFIX(VertexAttrib4ubNV) @@ -26622,7 +26442,7 @@ GL_PREFIX(VertexAttrib4ubNV): GL_PREFIX(VertexAttrib4ubvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5648(%rax), %r11 + movq 5616(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26632,13 +26452,13 @@ GL_PREFIX(VertexAttrib4ubvNV): popq %rbp popq %rsi popq %rdi - movq 5648(%rax), %r11 + movq 5616(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5648(%rax), %r11 + movq 5616(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26648,7 +26468,7 @@ GL_PREFIX(VertexAttrib4ubvNV): popq %rbp popq %rsi popq %rdi - movq 5648(%rax), %r11 + movq 5616(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttrib4ubvNV), .-GL_PREFIX(VertexAttrib4ubvNV) @@ -26659,7 +26479,7 @@ GL_PREFIX(VertexAttrib4ubvNV): GL_PREFIX(VertexAttribPointerNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5656(%rax), %r11 + movq 5624(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26673,13 +26493,13 @@ GL_PREFIX(VertexAttribPointerNV): popq %rdx popq %rsi popq %rdi - movq 5656(%rax), %r11 + movq 5624(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5656(%rax), %r11 + movq 5624(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26693,7 +26513,7 @@ GL_PREFIX(VertexAttribPointerNV): popq %rdx popq %rsi popq %rdi - movq 5656(%rax), %r11 + movq 5624(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribPointerNV), .-GL_PREFIX(VertexAttribPointerNV) @@ -26704,7 +26524,7 @@ GL_PREFIX(VertexAttribPointerNV): GL_PREFIX(VertexAttribs1dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5664(%rax), %r11 + movq 5632(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26714,13 +26534,13 @@ GL_PREFIX(VertexAttribs1dvNV): popq %rdx popq %rsi popq %rdi - movq 5664(%rax), %r11 + movq 5632(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5664(%rax), %r11 + movq 5632(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26730,7 +26550,7 @@ GL_PREFIX(VertexAttribs1dvNV): popq %rdx popq %rsi popq %rdi - movq 5664(%rax), %r11 + movq 5632(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs1dvNV), .-GL_PREFIX(VertexAttribs1dvNV) @@ -26741,7 +26561,7 @@ GL_PREFIX(VertexAttribs1dvNV): GL_PREFIX(VertexAttribs1fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5672(%rax), %r11 + movq 5640(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26751,13 +26571,13 @@ GL_PREFIX(VertexAttribs1fvNV): popq %rdx popq %rsi popq %rdi - movq 5672(%rax), %r11 + movq 5640(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5672(%rax), %r11 + movq 5640(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26767,7 +26587,7 @@ GL_PREFIX(VertexAttribs1fvNV): popq %rdx popq %rsi popq %rdi - movq 5672(%rax), %r11 + movq 5640(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs1fvNV), .-GL_PREFIX(VertexAttribs1fvNV) @@ -26778,7 +26598,7 @@ GL_PREFIX(VertexAttribs1fvNV): GL_PREFIX(VertexAttribs1svNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5680(%rax), %r11 + movq 5648(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26788,13 +26608,13 @@ GL_PREFIX(VertexAttribs1svNV): popq %rdx popq %rsi popq %rdi - movq 5680(%rax), %r11 + movq 5648(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5680(%rax), %r11 + movq 5648(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26804,7 +26624,7 @@ GL_PREFIX(VertexAttribs1svNV): popq %rdx popq %rsi popq %rdi - movq 5680(%rax), %r11 + movq 5648(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs1svNV), .-GL_PREFIX(VertexAttribs1svNV) @@ -26815,7 +26635,7 @@ GL_PREFIX(VertexAttribs1svNV): GL_PREFIX(VertexAttribs2dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5688(%rax), %r11 + movq 5656(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26825,13 +26645,13 @@ GL_PREFIX(VertexAttribs2dvNV): popq %rdx popq %rsi popq %rdi - movq 5688(%rax), %r11 + movq 5656(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5688(%rax), %r11 + movq 5656(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26841,7 +26661,7 @@ GL_PREFIX(VertexAttribs2dvNV): popq %rdx popq %rsi popq %rdi - movq 5688(%rax), %r11 + movq 5656(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs2dvNV), .-GL_PREFIX(VertexAttribs2dvNV) @@ -26852,7 +26672,7 @@ GL_PREFIX(VertexAttribs2dvNV): GL_PREFIX(VertexAttribs2fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5696(%rax), %r11 + movq 5664(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26862,13 +26682,13 @@ GL_PREFIX(VertexAttribs2fvNV): popq %rdx popq %rsi popq %rdi - movq 5696(%rax), %r11 + movq 5664(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5696(%rax), %r11 + movq 5664(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26878,7 +26698,7 @@ GL_PREFIX(VertexAttribs2fvNV): popq %rdx popq %rsi popq %rdi - movq 5696(%rax), %r11 + movq 5664(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs2fvNV), .-GL_PREFIX(VertexAttribs2fvNV) @@ -26889,7 +26709,7 @@ GL_PREFIX(VertexAttribs2fvNV): GL_PREFIX(VertexAttribs2svNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5704(%rax), %r11 + movq 5672(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26899,13 +26719,13 @@ GL_PREFIX(VertexAttribs2svNV): popq %rdx popq %rsi popq %rdi - movq 5704(%rax), %r11 + movq 5672(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5704(%rax), %r11 + movq 5672(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26915,7 +26735,7 @@ GL_PREFIX(VertexAttribs2svNV): popq %rdx popq %rsi popq %rdi - movq 5704(%rax), %r11 + movq 5672(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs2svNV), .-GL_PREFIX(VertexAttribs2svNV) @@ -26926,7 +26746,7 @@ GL_PREFIX(VertexAttribs2svNV): GL_PREFIX(VertexAttribs3dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5712(%rax), %r11 + movq 5680(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26936,13 +26756,13 @@ GL_PREFIX(VertexAttribs3dvNV): popq %rdx popq %rsi popq %rdi - movq 5712(%rax), %r11 + movq 5680(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5712(%rax), %r11 + movq 5680(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26952,7 +26772,7 @@ GL_PREFIX(VertexAttribs3dvNV): popq %rdx popq %rsi popq %rdi - movq 5712(%rax), %r11 + movq 5680(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs3dvNV), .-GL_PREFIX(VertexAttribs3dvNV) @@ -26963,7 +26783,7 @@ GL_PREFIX(VertexAttribs3dvNV): GL_PREFIX(VertexAttribs3fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5720(%rax), %r11 + movq 5688(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -26973,13 +26793,13 @@ GL_PREFIX(VertexAttribs3fvNV): popq %rdx popq %rsi popq %rdi - movq 5720(%rax), %r11 + movq 5688(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5720(%rax), %r11 + movq 5688(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -26989,7 +26809,7 @@ GL_PREFIX(VertexAttribs3fvNV): popq %rdx popq %rsi popq %rdi - movq 5720(%rax), %r11 + movq 5688(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs3fvNV), .-GL_PREFIX(VertexAttribs3fvNV) @@ -27000,7 +26820,7 @@ GL_PREFIX(VertexAttribs3fvNV): GL_PREFIX(VertexAttribs3svNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5728(%rax), %r11 + movq 5696(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27010,13 +26830,13 @@ GL_PREFIX(VertexAttribs3svNV): popq %rdx popq %rsi popq %rdi - movq 5728(%rax), %r11 + movq 5696(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5728(%rax), %r11 + movq 5696(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27026,7 +26846,7 @@ GL_PREFIX(VertexAttribs3svNV): popq %rdx popq %rsi popq %rdi - movq 5728(%rax), %r11 + movq 5696(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs3svNV), .-GL_PREFIX(VertexAttribs3svNV) @@ -27037,7 +26857,7 @@ GL_PREFIX(VertexAttribs3svNV): GL_PREFIX(VertexAttribs4dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5736(%rax), %r11 + movq 5704(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27047,13 +26867,13 @@ GL_PREFIX(VertexAttribs4dvNV): popq %rdx popq %rsi popq %rdi - movq 5736(%rax), %r11 + movq 5704(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5736(%rax), %r11 + movq 5704(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27063,7 +26883,7 @@ GL_PREFIX(VertexAttribs4dvNV): popq %rdx popq %rsi popq %rdi - movq 5736(%rax), %r11 + movq 5704(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs4dvNV), .-GL_PREFIX(VertexAttribs4dvNV) @@ -27074,7 +26894,7 @@ GL_PREFIX(VertexAttribs4dvNV): GL_PREFIX(VertexAttribs4fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5744(%rax), %r11 + movq 5712(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27084,13 +26904,13 @@ GL_PREFIX(VertexAttribs4fvNV): popq %rdx popq %rsi popq %rdi - movq 5744(%rax), %r11 + movq 5712(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5744(%rax), %r11 + movq 5712(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27100,7 +26920,7 @@ GL_PREFIX(VertexAttribs4fvNV): popq %rdx popq %rsi popq %rdi - movq 5744(%rax), %r11 + movq 5712(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs4fvNV), .-GL_PREFIX(VertexAttribs4fvNV) @@ -27111,7 +26931,7 @@ GL_PREFIX(VertexAttribs4fvNV): GL_PREFIX(VertexAttribs4svNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5752(%rax), %r11 + movq 5720(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27121,13 +26941,13 @@ GL_PREFIX(VertexAttribs4svNV): popq %rdx popq %rsi popq %rdi - movq 5752(%rax), %r11 + movq 5720(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5752(%rax), %r11 + movq 5720(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27137,7 +26957,7 @@ GL_PREFIX(VertexAttribs4svNV): popq %rdx popq %rsi popq %rdi - movq 5752(%rax), %r11 + movq 5720(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs4svNV), .-GL_PREFIX(VertexAttribs4svNV) @@ -27148,7 +26968,7 @@ GL_PREFIX(VertexAttribs4svNV): GL_PREFIX(VertexAttribs4ubvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5760(%rax), %r11 + movq 5728(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27158,13 +26978,13 @@ GL_PREFIX(VertexAttribs4ubvNV): popq %rdx popq %rsi popq %rdi - movq 5760(%rax), %r11 + movq 5728(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5760(%rax), %r11 + movq 5728(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27174,7 +26994,7 @@ GL_PREFIX(VertexAttribs4ubvNV): popq %rdx popq %rsi popq %rdi - movq 5760(%rax), %r11 + movq 5728(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(VertexAttribs4ubvNV), .-GL_PREFIX(VertexAttribs4ubvNV) @@ -27185,7 +27005,7 @@ GL_PREFIX(VertexAttribs4ubvNV): GL_PREFIX(AlphaFragmentOp1ATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5768(%rax), %r11 + movq 5736(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27203,13 +27023,13 @@ GL_PREFIX(AlphaFragmentOp1ATI): popq %rdx popq %rsi popq %rdi - movq 5768(%rax), %r11 + movq 5736(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5768(%rax), %r11 + movq 5736(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27227,7 +27047,7 @@ GL_PREFIX(AlphaFragmentOp1ATI): popq %rdx popq %rsi popq %rdi - movq 5768(%rax), %r11 + movq 5736(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(AlphaFragmentOp1ATI), .-GL_PREFIX(AlphaFragmentOp1ATI) @@ -27238,7 +27058,7 @@ GL_PREFIX(AlphaFragmentOp1ATI): GL_PREFIX(AlphaFragmentOp2ATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5776(%rax), %r11 + movq 5744(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27256,13 +27076,13 @@ GL_PREFIX(AlphaFragmentOp2ATI): popq %rdx popq %rsi popq %rdi - movq 5776(%rax), %r11 + movq 5744(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5776(%rax), %r11 + movq 5744(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27280,7 +27100,7 @@ GL_PREFIX(AlphaFragmentOp2ATI): popq %rdx popq %rsi popq %rdi - movq 5776(%rax), %r11 + movq 5744(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(AlphaFragmentOp2ATI), .-GL_PREFIX(AlphaFragmentOp2ATI) @@ -27291,7 +27111,7 @@ GL_PREFIX(AlphaFragmentOp2ATI): GL_PREFIX(AlphaFragmentOp3ATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5784(%rax), %r11 + movq 5752(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27309,13 +27129,13 @@ GL_PREFIX(AlphaFragmentOp3ATI): popq %rdx popq %rsi popq %rdi - movq 5784(%rax), %r11 + movq 5752(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5784(%rax), %r11 + movq 5752(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27333,7 +27153,7 @@ GL_PREFIX(AlphaFragmentOp3ATI): popq %rdx popq %rsi popq %rdi - movq 5784(%rax), %r11 + movq 5752(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(AlphaFragmentOp3ATI), .-GL_PREFIX(AlphaFragmentOp3ATI) @@ -27344,25 +27164,25 @@ GL_PREFIX(AlphaFragmentOp3ATI): GL_PREFIX(BeginFragmentShaderATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5792(%rax), %r11 + movq 5760(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rbp call _x86_64_get_dispatch@PLT popq %rbp - movq 5792(%rax), %r11 + movq 5760(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5792(%rax), %r11 + movq 5760(%rax), %r11 jmp *%r11 1: pushq %rbp call _glapi_get_dispatch popq %rbp - movq 5792(%rax), %r11 + movq 5760(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(BeginFragmentShaderATI), .-GL_PREFIX(BeginFragmentShaderATI) @@ -27373,25 +27193,25 @@ GL_PREFIX(BeginFragmentShaderATI): GL_PREFIX(BindFragmentShaderATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5800(%rax), %r11 + movq 5768(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 5800(%rax), %r11 + movq 5768(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5800(%rax), %r11 + movq 5768(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 5800(%rax), %r11 + movq 5768(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(BindFragmentShaderATI), .-GL_PREFIX(BindFragmentShaderATI) @@ -27402,7 +27222,7 @@ GL_PREFIX(BindFragmentShaderATI): GL_PREFIX(ColorFragmentOp1ATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5808(%rax), %r11 + movq 5776(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27420,13 +27240,13 @@ GL_PREFIX(ColorFragmentOp1ATI): popq %rdx popq %rsi popq %rdi - movq 5808(%rax), %r11 + movq 5776(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5808(%rax), %r11 + movq 5776(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27444,7 +27264,7 @@ GL_PREFIX(ColorFragmentOp1ATI): popq %rdx popq %rsi popq %rdi - movq 5808(%rax), %r11 + movq 5776(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(ColorFragmentOp1ATI), .-GL_PREFIX(ColorFragmentOp1ATI) @@ -27455,7 +27275,7 @@ GL_PREFIX(ColorFragmentOp1ATI): GL_PREFIX(ColorFragmentOp2ATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5816(%rax), %r11 + movq 5784(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27473,13 +27293,13 @@ GL_PREFIX(ColorFragmentOp2ATI): popq %rdx popq %rsi popq %rdi - movq 5816(%rax), %r11 + movq 5784(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5816(%rax), %r11 + movq 5784(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27497,7 +27317,7 @@ GL_PREFIX(ColorFragmentOp2ATI): popq %rdx popq %rsi popq %rdi - movq 5816(%rax), %r11 + movq 5784(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(ColorFragmentOp2ATI), .-GL_PREFIX(ColorFragmentOp2ATI) @@ -27508,7 +27328,7 @@ GL_PREFIX(ColorFragmentOp2ATI): GL_PREFIX(ColorFragmentOp3ATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5824(%rax), %r11 + movq 5792(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27526,13 +27346,13 @@ GL_PREFIX(ColorFragmentOp3ATI): popq %rdx popq %rsi popq %rdi - movq 5824(%rax), %r11 + movq 5792(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5824(%rax), %r11 + movq 5792(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27550,7 +27370,7 @@ GL_PREFIX(ColorFragmentOp3ATI): popq %rdx popq %rsi popq %rdi - movq 5824(%rax), %r11 + movq 5792(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(ColorFragmentOp3ATI), .-GL_PREFIX(ColorFragmentOp3ATI) @@ -27561,25 +27381,25 @@ GL_PREFIX(ColorFragmentOp3ATI): GL_PREFIX(DeleteFragmentShaderATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5832(%rax), %r11 + movq 5800(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 5832(%rax), %r11 + movq 5800(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5832(%rax), %r11 + movq 5800(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 5832(%rax), %r11 + movq 5800(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(DeleteFragmentShaderATI), .-GL_PREFIX(DeleteFragmentShaderATI) @@ -27590,25 +27410,25 @@ GL_PREFIX(DeleteFragmentShaderATI): GL_PREFIX(EndFragmentShaderATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5840(%rax), %r11 + movq 5808(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rbp call _x86_64_get_dispatch@PLT popq %rbp - movq 5840(%rax), %r11 + movq 5808(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5840(%rax), %r11 + movq 5808(%rax), %r11 jmp *%r11 1: pushq %rbp call _glapi_get_dispatch popq %rbp - movq 5840(%rax), %r11 + movq 5808(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(EndFragmentShaderATI), .-GL_PREFIX(EndFragmentShaderATI) @@ -27619,25 +27439,25 @@ GL_PREFIX(EndFragmentShaderATI): GL_PREFIX(GenFragmentShadersATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5848(%rax), %r11 + movq 5816(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 5848(%rax), %r11 + movq 5816(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5848(%rax), %r11 + movq 5816(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 5848(%rax), %r11 + movq 5816(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(GenFragmentShadersATI), .-GL_PREFIX(GenFragmentShadersATI) @@ -27648,7 +27468,7 @@ GL_PREFIX(GenFragmentShadersATI): GL_PREFIX(PassTexCoordATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5856(%rax), %r11 + movq 5824(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27658,13 +27478,13 @@ GL_PREFIX(PassTexCoordATI): popq %rdx popq %rsi popq %rdi - movq 5856(%rax), %r11 + movq 5824(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5856(%rax), %r11 + movq 5824(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27674,7 +27494,7 @@ GL_PREFIX(PassTexCoordATI): popq %rdx popq %rsi popq %rdi - movq 5856(%rax), %r11 + movq 5824(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(PassTexCoordATI), .-GL_PREFIX(PassTexCoordATI) @@ -27685,7 +27505,7 @@ GL_PREFIX(PassTexCoordATI): GL_PREFIX(SampleMapATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5864(%rax), %r11 + movq 5832(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27695,13 +27515,13 @@ GL_PREFIX(SampleMapATI): popq %rdx popq %rsi popq %rdi - movq 5864(%rax), %r11 + movq 5832(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5864(%rax), %r11 + movq 5832(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27711,7 +27531,7 @@ GL_PREFIX(SampleMapATI): popq %rdx popq %rsi popq %rdi - movq 5864(%rax), %r11 + movq 5832(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(SampleMapATI), .-GL_PREFIX(SampleMapATI) @@ -27722,7 +27542,7 @@ GL_PREFIX(SampleMapATI): GL_PREFIX(SetFragmentShaderConstantATI): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5872(%rax), %r11 + movq 5840(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27732,13 +27552,13 @@ GL_PREFIX(SetFragmentShaderConstantATI): popq %rbp popq %rsi popq %rdi - movq 5872(%rax), %r11 + movq 5840(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5872(%rax), %r11 + movq 5840(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27748,7 +27568,7 @@ GL_PREFIX(SetFragmentShaderConstantATI): popq %rbp popq %rsi popq %rdi - movq 5872(%rax), %r11 + movq 5840(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(SetFragmentShaderConstantATI), .-GL_PREFIX(SetFragmentShaderConstantATI) @@ -27759,7 +27579,7 @@ GL_PREFIX(SetFragmentShaderConstantATI): GL_PREFIX(PointParameteriNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5880(%rax), %r11 + movq 5848(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27769,13 +27589,13 @@ GL_PREFIX(PointParameteriNV): popq %rbp popq %rsi popq %rdi - movq 5880(%rax), %r11 + movq 5848(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5880(%rax), %r11 + movq 5848(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27785,7 +27605,7 @@ GL_PREFIX(PointParameteriNV): popq %rbp popq %rsi popq %rdi - movq 5880(%rax), %r11 + movq 5848(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(PointParameteriNV), .-GL_PREFIX(PointParameteriNV) @@ -27796,7 +27616,7 @@ GL_PREFIX(PointParameteriNV): GL_PREFIX(PointParameterivNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5888(%rax), %r11 + movq 5856(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27806,13 +27626,13 @@ GL_PREFIX(PointParameterivNV): popq %rbp popq %rsi popq %rdi - movq 5888(%rax), %r11 + movq 5856(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5888(%rax), %r11 + movq 5856(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27822,79 +27642,79 @@ GL_PREFIX(PointParameterivNV): popq %rbp popq %rsi popq %rdi - movq 5888(%rax), %r11 + movq 5856(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(PointParameterivNV), .-GL_PREFIX(PointParameterivNV) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_737) - .type GL_PREFIX(_dispatch_stub_737), @function - HIDDEN(GL_PREFIX(_dispatch_stub_737)) -GL_PREFIX(_dispatch_stub_737): + .globl GL_PREFIX(_dispatch_stub_733) + .type GL_PREFIX(_dispatch_stub_733), @function + HIDDEN(GL_PREFIX(_dispatch_stub_733)) +GL_PREFIX(_dispatch_stub_733): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5896(%rax), %r11 + movq 5864(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 5896(%rax), %r11 + movq 5864(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5896(%rax), %r11 + movq 5864(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 5896(%rax), %r11 + movq 5864(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_737), .-GL_PREFIX(_dispatch_stub_737) + .size GL_PREFIX(_dispatch_stub_733), .-GL_PREFIX(_dispatch_stub_733) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_738) - .type GL_PREFIX(_dispatch_stub_738), @function - HIDDEN(GL_PREFIX(_dispatch_stub_738)) -GL_PREFIX(_dispatch_stub_738): + .globl GL_PREFIX(_dispatch_stub_734) + .type GL_PREFIX(_dispatch_stub_734), @function + HIDDEN(GL_PREFIX(_dispatch_stub_734)) +GL_PREFIX(_dispatch_stub_734): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5904(%rax), %r11 + movq 5872(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 5904(%rax), %r11 + movq 5872(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5904(%rax), %r11 + movq 5872(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 5904(%rax), %r11 + movq 5872(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_738), .-GL_PREFIX(_dispatch_stub_738) + .size GL_PREFIX(_dispatch_stub_734), .-GL_PREFIX(_dispatch_stub_734) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_739) - .type GL_PREFIX(_dispatch_stub_739), @function - HIDDEN(GL_PREFIX(_dispatch_stub_739)) -GL_PREFIX(_dispatch_stub_739): + .globl GL_PREFIX(_dispatch_stub_735) + .type GL_PREFIX(_dispatch_stub_735), @function + HIDDEN(GL_PREFIX(_dispatch_stub_735)) +GL_PREFIX(_dispatch_stub_735): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5912(%rax), %r11 + movq 5880(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27904,13 +27724,13 @@ GL_PREFIX(_dispatch_stub_739): popq %rbp popq %rsi popq %rdi - movq 5912(%rax), %r11 + movq 5880(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5912(%rax), %r11 + movq 5880(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27920,19 +27740,19 @@ GL_PREFIX(_dispatch_stub_739): popq %rbp popq %rsi popq %rdi - movq 5912(%rax), %r11 + movq 5880(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_739), .-GL_PREFIX(_dispatch_stub_739) + .size GL_PREFIX(_dispatch_stub_735), .-GL_PREFIX(_dispatch_stub_735) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_740) - .type GL_PREFIX(_dispatch_stub_740), @function - HIDDEN(GL_PREFIX(_dispatch_stub_740)) -GL_PREFIX(_dispatch_stub_740): + .globl GL_PREFIX(_dispatch_stub_736) + .type GL_PREFIX(_dispatch_stub_736), @function + HIDDEN(GL_PREFIX(_dispatch_stub_736)) +GL_PREFIX(_dispatch_stub_736): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5920(%rax), %r11 + movq 5888(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -27942,13 +27762,13 @@ GL_PREFIX(_dispatch_stub_740): popq %rbp popq %rsi popq %rdi - movq 5920(%rax), %r11 + movq 5888(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5920(%rax), %r11 + movq 5888(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -27958,40 +27778,40 @@ GL_PREFIX(_dispatch_stub_740): popq %rbp popq %rsi popq %rdi - movq 5920(%rax), %r11 + movq 5888(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_740), .-GL_PREFIX(_dispatch_stub_740) + .size GL_PREFIX(_dispatch_stub_736), .-GL_PREFIX(_dispatch_stub_736) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_741) - .type GL_PREFIX(_dispatch_stub_741), @function - HIDDEN(GL_PREFIX(_dispatch_stub_741)) -GL_PREFIX(_dispatch_stub_741): + .globl GL_PREFIX(_dispatch_stub_737) + .type GL_PREFIX(_dispatch_stub_737), @function + HIDDEN(GL_PREFIX(_dispatch_stub_737)) +GL_PREFIX(_dispatch_stub_737): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5928(%rax), %r11 + movq 5896(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 5928(%rax), %r11 + movq 5896(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5928(%rax), %r11 + movq 5896(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 5928(%rax), %r11 + movq 5896(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_741), .-GL_PREFIX(_dispatch_stub_741) + .size GL_PREFIX(_dispatch_stub_737), .-GL_PREFIX(_dispatch_stub_737) .p2align 4,,15 .globl GL_PREFIX(GetProgramNamedParameterdvNV) @@ -27999,7 +27819,7 @@ GL_PREFIX(_dispatch_stub_741): GL_PREFIX(GetProgramNamedParameterdvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5936(%rax), %r11 + movq 5904(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28013,13 +27833,13 @@ GL_PREFIX(GetProgramNamedParameterdvNV): popq %rdx popq %rsi popq %rdi - movq 5936(%rax), %r11 + movq 5904(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5936(%rax), %r11 + movq 5904(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28033,7 +27853,7 @@ GL_PREFIX(GetProgramNamedParameterdvNV): popq %rdx popq %rsi popq %rdi - movq 5936(%rax), %r11 + movq 5904(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(GetProgramNamedParameterdvNV), .-GL_PREFIX(GetProgramNamedParameterdvNV) @@ -28044,7 +27864,7 @@ GL_PREFIX(GetProgramNamedParameterdvNV): GL_PREFIX(GetProgramNamedParameterfvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5944(%rax), %r11 + movq 5912(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28058,13 +27878,13 @@ GL_PREFIX(GetProgramNamedParameterfvNV): popq %rdx popq %rsi popq %rdi - movq 5944(%rax), %r11 + movq 5912(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5944(%rax), %r11 + movq 5912(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28078,7 +27898,7 @@ GL_PREFIX(GetProgramNamedParameterfvNV): popq %rdx popq %rsi popq %rdi - movq 5944(%rax), %r11 + movq 5912(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(GetProgramNamedParameterfvNV), .-GL_PREFIX(GetProgramNamedParameterfvNV) @@ -28089,7 +27909,7 @@ GL_PREFIX(GetProgramNamedParameterfvNV): GL_PREFIX(ProgramNamedParameter4dNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5952(%rax), %r11 + movq 5920(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $56, %rsp @@ -28109,13 +27929,13 @@ GL_PREFIX(ProgramNamedParameter4dNV): movq 8(%rsp), %rsi movq (%rsp), %rdi addq $56, %rsp - movq 5952(%rax), %r11 + movq 5920(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5952(%rax), %r11 + movq 5920(%rax), %r11 jmp *%r11 1: subq $56, %rsp @@ -28135,7 +27955,7 @@ GL_PREFIX(ProgramNamedParameter4dNV): movq 8(%rsp), %rsi movq (%rsp), %rdi addq $56, %rsp - movq 5952(%rax), %r11 + movq 5920(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(ProgramNamedParameter4dNV), .-GL_PREFIX(ProgramNamedParameter4dNV) @@ -28146,7 +27966,7 @@ GL_PREFIX(ProgramNamedParameter4dNV): GL_PREFIX(ProgramNamedParameter4dvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5960(%rax), %r11 + movq 5928(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28160,13 +27980,13 @@ GL_PREFIX(ProgramNamedParameter4dvNV): popq %rdx popq %rsi popq %rdi - movq 5960(%rax), %r11 + movq 5928(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5960(%rax), %r11 + movq 5928(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28180,7 +28000,7 @@ GL_PREFIX(ProgramNamedParameter4dvNV): popq %rdx popq %rsi popq %rdi - movq 5960(%rax), %r11 + movq 5928(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(ProgramNamedParameter4dvNV), .-GL_PREFIX(ProgramNamedParameter4dvNV) @@ -28191,7 +28011,7 @@ GL_PREFIX(ProgramNamedParameter4dvNV): GL_PREFIX(ProgramNamedParameter4fNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5968(%rax), %r11 + movq 5936(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) subq $56, %rsp @@ -28211,13 +28031,13 @@ GL_PREFIX(ProgramNamedParameter4fNV): movq 8(%rsp), %rsi movq (%rsp), %rdi addq $56, %rsp - movq 5968(%rax), %r11 + movq 5936(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5968(%rax), %r11 + movq 5936(%rax), %r11 jmp *%r11 1: subq $56, %rsp @@ -28237,7 +28057,7 @@ GL_PREFIX(ProgramNamedParameter4fNV): movq 8(%rsp), %rsi movq (%rsp), %rdi addq $56, %rsp - movq 5968(%rax), %r11 + movq 5936(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(ProgramNamedParameter4fNV), .-GL_PREFIX(ProgramNamedParameter4fNV) @@ -28248,7 +28068,7 @@ GL_PREFIX(ProgramNamedParameter4fNV): GL_PREFIX(ProgramNamedParameter4fvNV): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5976(%rax), %r11 + movq 5944(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28262,13 +28082,13 @@ GL_PREFIX(ProgramNamedParameter4fvNV): popq %rdx popq %rsi popq %rdi - movq 5976(%rax), %r11 + movq 5944(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5976(%rax), %r11 + movq 5944(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28282,19 +28102,19 @@ GL_PREFIX(ProgramNamedParameter4fvNV): popq %rdx popq %rsi popq %rdi - movq 5976(%rax), %r11 + movq 5944(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(ProgramNamedParameter4fvNV), .-GL_PREFIX(ProgramNamedParameter4fvNV) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_748) - .type GL_PREFIX(_dispatch_stub_748), @function - HIDDEN(GL_PREFIX(_dispatch_stub_748)) -GL_PREFIX(_dispatch_stub_748): + .globl GL_PREFIX(_dispatch_stub_744) + .type GL_PREFIX(_dispatch_stub_744), @function + HIDDEN(GL_PREFIX(_dispatch_stub_744)) +GL_PREFIX(_dispatch_stub_744): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5984(%rax), %r11 + movq 5952(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28304,13 +28124,13 @@ GL_PREFIX(_dispatch_stub_748): popq %rbp popq %rsi popq %rdi - movq 5984(%rax), %r11 + movq 5952(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5984(%rax), %r11 + movq 5952(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28320,19 +28140,19 @@ GL_PREFIX(_dispatch_stub_748): popq %rbp popq %rsi popq %rdi - movq 5984(%rax), %r11 + movq 5952(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_748), .-GL_PREFIX(_dispatch_stub_748) + .size GL_PREFIX(_dispatch_stub_744), .-GL_PREFIX(_dispatch_stub_744) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_749) - .type GL_PREFIX(_dispatch_stub_749), @function - HIDDEN(GL_PREFIX(_dispatch_stub_749)) -GL_PREFIX(_dispatch_stub_749): + .globl GL_PREFIX(_dispatch_stub_745) + .type GL_PREFIX(_dispatch_stub_745), @function + HIDDEN(GL_PREFIX(_dispatch_stub_745)) +GL_PREFIX(_dispatch_stub_745): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 5992(%rax), %r11 + movq 5960(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28342,13 +28162,13 @@ GL_PREFIX(_dispatch_stub_749): popq %rbp popq %rsi popq %rdi - movq 5992(%rax), %r11 + movq 5960(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 5992(%rax), %r11 + movq 5960(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28358,10 +28178,10 @@ GL_PREFIX(_dispatch_stub_749): popq %rbp popq %rsi popq %rdi - movq 5992(%rax), %r11 + movq 5960(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_749), .-GL_PREFIX(_dispatch_stub_749) + .size GL_PREFIX(_dispatch_stub_745), .-GL_PREFIX(_dispatch_stub_745) .p2align 4,,15 .globl GL_PREFIX(BindFramebufferEXT) @@ -28369,7 +28189,7 @@ GL_PREFIX(_dispatch_stub_749): GL_PREFIX(BindFramebufferEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6000(%rax), %r11 + movq 5968(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28379,13 +28199,13 @@ GL_PREFIX(BindFramebufferEXT): popq %rbp popq %rsi popq %rdi - movq 6000(%rax), %r11 + movq 5968(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6000(%rax), %r11 + movq 5968(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28395,7 +28215,7 @@ GL_PREFIX(BindFramebufferEXT): popq %rbp popq %rsi popq %rdi - movq 6000(%rax), %r11 + movq 5968(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(BindFramebufferEXT), .-GL_PREFIX(BindFramebufferEXT) @@ -28406,7 +28226,7 @@ GL_PREFIX(BindFramebufferEXT): GL_PREFIX(BindRenderbufferEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6008(%rax), %r11 + movq 5976(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28416,13 +28236,13 @@ GL_PREFIX(BindRenderbufferEXT): popq %rbp popq %rsi popq %rdi - movq 6008(%rax), %r11 + movq 5976(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6008(%rax), %r11 + movq 5976(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28432,7 +28252,7 @@ GL_PREFIX(BindRenderbufferEXT): popq %rbp popq %rsi popq %rdi - movq 6008(%rax), %r11 + movq 5976(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(BindRenderbufferEXT), .-GL_PREFIX(BindRenderbufferEXT) @@ -28443,25 +28263,25 @@ GL_PREFIX(BindRenderbufferEXT): GL_PREFIX(CheckFramebufferStatusEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6016(%rax), %r11 + movq 5984(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 6016(%rax), %r11 + movq 5984(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6016(%rax), %r11 + movq 5984(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 6016(%rax), %r11 + movq 5984(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(CheckFramebufferStatusEXT), .-GL_PREFIX(CheckFramebufferStatusEXT) @@ -28472,7 +28292,7 @@ GL_PREFIX(CheckFramebufferStatusEXT): GL_PREFIX(DeleteFramebuffersEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6024(%rax), %r11 + movq 5992(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28482,13 +28302,13 @@ GL_PREFIX(DeleteFramebuffersEXT): popq %rbp popq %rsi popq %rdi - movq 6024(%rax), %r11 + movq 5992(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6024(%rax), %r11 + movq 5992(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28498,7 +28318,7 @@ GL_PREFIX(DeleteFramebuffersEXT): popq %rbp popq %rsi popq %rdi - movq 6024(%rax), %r11 + movq 5992(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(DeleteFramebuffersEXT), .-GL_PREFIX(DeleteFramebuffersEXT) @@ -28509,7 +28329,7 @@ GL_PREFIX(DeleteFramebuffersEXT): GL_PREFIX(DeleteRenderbuffersEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6032(%rax), %r11 + movq 6000(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28519,13 +28339,13 @@ GL_PREFIX(DeleteRenderbuffersEXT): popq %rbp popq %rsi popq %rdi - movq 6032(%rax), %r11 + movq 6000(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6032(%rax), %r11 + movq 6000(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28535,7 +28355,7 @@ GL_PREFIX(DeleteRenderbuffersEXT): popq %rbp popq %rsi popq %rdi - movq 6032(%rax), %r11 + movq 6000(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(DeleteRenderbuffersEXT), .-GL_PREFIX(DeleteRenderbuffersEXT) @@ -28546,7 +28366,7 @@ GL_PREFIX(DeleteRenderbuffersEXT): GL_PREFIX(FramebufferRenderbufferEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6040(%rax), %r11 + movq 6008(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28560,13 +28380,13 @@ GL_PREFIX(FramebufferRenderbufferEXT): popq %rdx popq %rsi popq %rdi - movq 6040(%rax), %r11 + movq 6008(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6040(%rax), %r11 + movq 6008(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28580,7 +28400,7 @@ GL_PREFIX(FramebufferRenderbufferEXT): popq %rdx popq %rsi popq %rdi - movq 6040(%rax), %r11 + movq 6008(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(FramebufferRenderbufferEXT), .-GL_PREFIX(FramebufferRenderbufferEXT) @@ -28591,7 +28411,7 @@ GL_PREFIX(FramebufferRenderbufferEXT): GL_PREFIX(FramebufferTexture1DEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6048(%rax), %r11 + movq 6016(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28605,13 +28425,13 @@ GL_PREFIX(FramebufferTexture1DEXT): popq %rdx popq %rsi popq %rdi - movq 6048(%rax), %r11 + movq 6016(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6048(%rax), %r11 + movq 6016(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28625,7 +28445,7 @@ GL_PREFIX(FramebufferTexture1DEXT): popq %rdx popq %rsi popq %rdi - movq 6048(%rax), %r11 + movq 6016(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(FramebufferTexture1DEXT), .-GL_PREFIX(FramebufferTexture1DEXT) @@ -28636,7 +28456,7 @@ GL_PREFIX(FramebufferTexture1DEXT): GL_PREFIX(FramebufferTexture2DEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6056(%rax), %r11 + movq 6024(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28650,13 +28470,13 @@ GL_PREFIX(FramebufferTexture2DEXT): popq %rdx popq %rsi popq %rdi - movq 6056(%rax), %r11 + movq 6024(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6056(%rax), %r11 + movq 6024(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28670,7 +28490,7 @@ GL_PREFIX(FramebufferTexture2DEXT): popq %rdx popq %rsi popq %rdi - movq 6056(%rax), %r11 + movq 6024(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(FramebufferTexture2DEXT), .-GL_PREFIX(FramebufferTexture2DEXT) @@ -28681,7 +28501,7 @@ GL_PREFIX(FramebufferTexture2DEXT): GL_PREFIX(FramebufferTexture3DEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6064(%rax), %r11 + movq 6032(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28699,13 +28519,13 @@ GL_PREFIX(FramebufferTexture3DEXT): popq %rdx popq %rsi popq %rdi - movq 6064(%rax), %r11 + movq 6032(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6064(%rax), %r11 + movq 6032(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28723,7 +28543,7 @@ GL_PREFIX(FramebufferTexture3DEXT): popq %rdx popq %rsi popq %rdi - movq 6064(%rax), %r11 + movq 6032(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(FramebufferTexture3DEXT), .-GL_PREFIX(FramebufferTexture3DEXT) @@ -28734,7 +28554,7 @@ GL_PREFIX(FramebufferTexture3DEXT): GL_PREFIX(GenFramebuffersEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6072(%rax), %r11 + movq 6040(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28744,13 +28564,13 @@ GL_PREFIX(GenFramebuffersEXT): popq %rbp popq %rsi popq %rdi - movq 6072(%rax), %r11 + movq 6040(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6072(%rax), %r11 + movq 6040(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28760,7 +28580,7 @@ GL_PREFIX(GenFramebuffersEXT): popq %rbp popq %rsi popq %rdi - movq 6072(%rax), %r11 + movq 6040(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(GenFramebuffersEXT), .-GL_PREFIX(GenFramebuffersEXT) @@ -28771,7 +28591,7 @@ GL_PREFIX(GenFramebuffersEXT): GL_PREFIX(GenRenderbuffersEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6080(%rax), %r11 + movq 6048(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28781,13 +28601,13 @@ GL_PREFIX(GenRenderbuffersEXT): popq %rbp popq %rsi popq %rdi - movq 6080(%rax), %r11 + movq 6048(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6080(%rax), %r11 + movq 6048(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28797,7 +28617,7 @@ GL_PREFIX(GenRenderbuffersEXT): popq %rbp popq %rsi popq %rdi - movq 6080(%rax), %r11 + movq 6048(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(GenRenderbuffersEXT), .-GL_PREFIX(GenRenderbuffersEXT) @@ -28808,25 +28628,25 @@ GL_PREFIX(GenRenderbuffersEXT): GL_PREFIX(GenerateMipmapEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6088(%rax), %r11 + movq 6056(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 6088(%rax), %r11 + movq 6056(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6088(%rax), %r11 + movq 6056(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 6088(%rax), %r11 + movq 6056(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(GenerateMipmapEXT), .-GL_PREFIX(GenerateMipmapEXT) @@ -28837,7 +28657,7 @@ GL_PREFIX(GenerateMipmapEXT): GL_PREFIX(GetFramebufferAttachmentParameterivEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6096(%rax), %r11 + movq 6064(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28851,13 +28671,13 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT): popq %rdx popq %rsi popq %rdi - movq 6096(%rax), %r11 + movq 6064(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6096(%rax), %r11 + movq 6064(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28871,7 +28691,7 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT): popq %rdx popq %rsi popq %rdi - movq 6096(%rax), %r11 + movq 6064(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(GetFramebufferAttachmentParameterivEXT), .-GL_PREFIX(GetFramebufferAttachmentParameterivEXT) @@ -28882,7 +28702,7 @@ GL_PREFIX(GetFramebufferAttachmentParameterivEXT): GL_PREFIX(GetRenderbufferParameterivEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6104(%rax), %r11 + movq 6072(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28892,13 +28712,13 @@ GL_PREFIX(GetRenderbufferParameterivEXT): popq %rdx popq %rsi popq %rdi - movq 6104(%rax), %r11 + movq 6072(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6104(%rax), %r11 + movq 6072(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -28908,7 +28728,7 @@ GL_PREFIX(GetRenderbufferParameterivEXT): popq %rdx popq %rsi popq %rdi - movq 6104(%rax), %r11 + movq 6072(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(GetRenderbufferParameterivEXT), .-GL_PREFIX(GetRenderbufferParameterivEXT) @@ -28919,25 +28739,25 @@ GL_PREFIX(GetRenderbufferParameterivEXT): GL_PREFIX(IsFramebufferEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6112(%rax), %r11 + movq 6080(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 6112(%rax), %r11 + movq 6080(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6112(%rax), %r11 + movq 6080(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 6112(%rax), %r11 + movq 6080(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(IsFramebufferEXT), .-GL_PREFIX(IsFramebufferEXT) @@ -28948,25 +28768,25 @@ GL_PREFIX(IsFramebufferEXT): GL_PREFIX(IsRenderbufferEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6120(%rax), %r11 + movq 6088(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi call _x86_64_get_dispatch@PLT popq %rdi - movq 6120(%rax), %r11 + movq 6088(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6120(%rax), %r11 + movq 6088(%rax), %r11 jmp *%r11 1: pushq %rdi call _glapi_get_dispatch popq %rdi - movq 6120(%rax), %r11 + movq 6088(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(IsRenderbufferEXT), .-GL_PREFIX(IsRenderbufferEXT) @@ -28977,7 +28797,7 @@ GL_PREFIX(IsRenderbufferEXT): GL_PREFIX(RenderbufferStorageEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6128(%rax), %r11 + movq 6096(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -28991,13 +28811,13 @@ GL_PREFIX(RenderbufferStorageEXT): popq %rdx popq %rsi popq %rdi - movq 6128(%rax), %r11 + movq 6096(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6128(%rax), %r11 + movq 6096(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -29011,19 +28831,19 @@ GL_PREFIX(RenderbufferStorageEXT): popq %rdx popq %rsi popq %rdi - movq 6128(%rax), %r11 + movq 6096(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(RenderbufferStorageEXT), .-GL_PREFIX(RenderbufferStorageEXT) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_767) - .type GL_PREFIX(_dispatch_stub_767), @function - HIDDEN(GL_PREFIX(_dispatch_stub_767)) -GL_PREFIX(_dispatch_stub_767): + .globl GL_PREFIX(_dispatch_stub_763) + .type GL_PREFIX(_dispatch_stub_763), @function + HIDDEN(GL_PREFIX(_dispatch_stub_763)) +GL_PREFIX(_dispatch_stub_763): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6136(%rax), %r11 + movq 6104(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -29041,13 +28861,13 @@ GL_PREFIX(_dispatch_stub_767): popq %rdx popq %rsi popq %rdi - movq 6136(%rax), %r11 + movq 6104(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6136(%rax), %r11 + movq 6104(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -29065,10 +28885,10 @@ GL_PREFIX(_dispatch_stub_767): popq %rdx popq %rsi popq %rdi - movq 6136(%rax), %r11 + movq 6104(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_767), .-GL_PREFIX(_dispatch_stub_767) + .size GL_PREFIX(_dispatch_stub_763), .-GL_PREFIX(_dispatch_stub_763) .p2align 4,,15 .globl GL_PREFIX(FramebufferTextureLayerEXT) @@ -29076,7 +28896,7 @@ GL_PREFIX(_dispatch_stub_767): GL_PREFIX(FramebufferTextureLayerEXT): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6144(%rax), %r11 + movq 6112(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -29090,13 +28910,13 @@ GL_PREFIX(FramebufferTextureLayerEXT): popq %rdx popq %rsi popq %rdi - movq 6144(%rax), %r11 + movq 6112(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6144(%rax), %r11 + movq 6112(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -29110,19 +28930,19 @@ GL_PREFIX(FramebufferTextureLayerEXT): popq %rdx popq %rsi popq %rdi - movq 6144(%rax), %r11 + movq 6112(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ .size GL_PREFIX(FramebufferTextureLayerEXT), .-GL_PREFIX(FramebufferTextureLayerEXT) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_769) - .type GL_PREFIX(_dispatch_stub_769), @function - HIDDEN(GL_PREFIX(_dispatch_stub_769)) -GL_PREFIX(_dispatch_stub_769): + .globl GL_PREFIX(_dispatch_stub_765) + .type GL_PREFIX(_dispatch_stub_765), @function + HIDDEN(GL_PREFIX(_dispatch_stub_765)) +GL_PREFIX(_dispatch_stub_765): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6152(%rax), %r11 + movq 6120(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -29136,13 +28956,13 @@ GL_PREFIX(_dispatch_stub_769): popq %rdx popq %rsi popq %rdi - movq 6152(%rax), %r11 + movq 6120(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6152(%rax), %r11 + movq 6120(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -29156,19 +28976,19 @@ GL_PREFIX(_dispatch_stub_769): popq %rdx popq %rsi popq %rdi - movq 6152(%rax), %r11 + movq 6120(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_769), .-GL_PREFIX(_dispatch_stub_769) + .size GL_PREFIX(_dispatch_stub_765), .-GL_PREFIX(_dispatch_stub_765) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_770) - .type GL_PREFIX(_dispatch_stub_770), @function - HIDDEN(GL_PREFIX(_dispatch_stub_770)) -GL_PREFIX(_dispatch_stub_770): + .globl GL_PREFIX(_dispatch_stub_766) + .type GL_PREFIX(_dispatch_stub_766), @function + HIDDEN(GL_PREFIX(_dispatch_stub_766)) +GL_PREFIX(_dispatch_stub_766): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6160(%rax), %r11 + movq 6128(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -29182,13 +29002,13 @@ GL_PREFIX(_dispatch_stub_770): popq %rdx popq %rsi popq %rdi - movq 6160(%rax), %r11 + movq 6128(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6160(%rax), %r11 + movq 6128(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -29202,19 +29022,19 @@ GL_PREFIX(_dispatch_stub_770): popq %rdx popq %rsi popq %rdi - movq 6160(%rax), %r11 + movq 6128(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_770), .-GL_PREFIX(_dispatch_stub_770) + .size GL_PREFIX(_dispatch_stub_766), .-GL_PREFIX(_dispatch_stub_766) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_771) - .type GL_PREFIX(_dispatch_stub_771), @function - HIDDEN(GL_PREFIX(_dispatch_stub_771)) -GL_PREFIX(_dispatch_stub_771): + .globl GL_PREFIX(_dispatch_stub_767) + .type GL_PREFIX(_dispatch_stub_767), @function + HIDDEN(GL_PREFIX(_dispatch_stub_767)) +GL_PREFIX(_dispatch_stub_767): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6168(%rax), %r11 + movq 6136(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -29228,13 +29048,13 @@ GL_PREFIX(_dispatch_stub_771): popq %rdx popq %rsi popq %rdi - movq 6168(%rax), %r11 + movq 6136(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6168(%rax), %r11 + movq 6136(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -29248,19 +29068,19 @@ GL_PREFIX(_dispatch_stub_771): popq %rdx popq %rsi popq %rdi - movq 6168(%rax), %r11 + movq 6136(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_771), .-GL_PREFIX(_dispatch_stub_771) + .size GL_PREFIX(_dispatch_stub_767), .-GL_PREFIX(_dispatch_stub_767) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_772) - .type GL_PREFIX(_dispatch_stub_772), @function - HIDDEN(GL_PREFIX(_dispatch_stub_772)) -GL_PREFIX(_dispatch_stub_772): + .globl GL_PREFIX(_dispatch_stub_768) + .type GL_PREFIX(_dispatch_stub_768), @function + HIDDEN(GL_PREFIX(_dispatch_stub_768)) +GL_PREFIX(_dispatch_stub_768): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6176(%rax), %r11 + movq 6144(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -29270,13 +29090,13 @@ GL_PREFIX(_dispatch_stub_772): popq %rdx popq %rsi popq %rdi - movq 6176(%rax), %r11 + movq 6144(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6176(%rax), %r11 + movq 6144(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -29286,19 +29106,19 @@ GL_PREFIX(_dispatch_stub_772): popq %rdx popq %rsi popq %rdi - movq 6176(%rax), %r11 + movq 6144(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_772), .-GL_PREFIX(_dispatch_stub_772) + .size GL_PREFIX(_dispatch_stub_768), .-GL_PREFIX(_dispatch_stub_768) .p2align 4,,15 - .globl GL_PREFIX(_dispatch_stub_773) - .type GL_PREFIX(_dispatch_stub_773), @function - HIDDEN(GL_PREFIX(_dispatch_stub_773)) -GL_PREFIX(_dispatch_stub_773): + .globl GL_PREFIX(_dispatch_stub_769) + .type GL_PREFIX(_dispatch_stub_769), @function + HIDDEN(GL_PREFIX(_dispatch_stub_769)) +GL_PREFIX(_dispatch_stub_769): #if defined(GLX_USE_TLS) call _x86_64_get_dispatch@PLT - movq 6184(%rax), %r11 + movq 6152(%rax), %r11 jmp *%r11 #elif defined(PTHREADS) pushq %rdi @@ -29308,13 +29128,13 @@ GL_PREFIX(_dispatch_stub_773): popq %rdx popq %rsi popq %rdi - movq 6184(%rax), %r11 + movq 6152(%rax), %r11 jmp *%r11 #else movq _glapi_Dispatch(%rip), %rax testq %rax, %rax je 1f - movq 6184(%rax), %r11 + movq 6152(%rax), %r11 jmp *%r11 1: pushq %rdi @@ -29324,10 +29144,10 @@ GL_PREFIX(_dispatch_stub_773): popq %rdx popq %rsi popq %rdi - movq 6184(%rax), %r11 + movq 6152(%rax), %r11 jmp *%r11 #endif /* defined(GLX_USE_TLS) */ - .size GL_PREFIX(_dispatch_stub_773), .-GL_PREFIX(_dispatch_stub_773) + .size GL_PREFIX(_dispatch_stub_769), .-GL_PREFIX(_dispatch_stub_769) .globl GL_PREFIX(ArrayElementEXT) ; .set GL_PREFIX(ArrayElementEXT), GL_PREFIX(ArrayElement) .globl GL_PREFIX(BindTextureEXT) ; .set GL_PREFIX(BindTextureEXT), GL_PREFIX(BindTexture) @@ -29419,6 +29239,10 @@ GL_PREFIX(_dispatch_stub_773): .globl GL_PREFIX(GetVertexAttribdv) ; .set GL_PREFIX(GetVertexAttribdv), GL_PREFIX(GetVertexAttribdvARB) .globl GL_PREFIX(GetVertexAttribfv) ; .set GL_PREFIX(GetVertexAttribfv), GL_PREFIX(GetVertexAttribfvARB) .globl GL_PREFIX(GetVertexAttribiv) ; .set GL_PREFIX(GetVertexAttribiv), GL_PREFIX(GetVertexAttribivARB) + .globl GL_PREFIX(ProgramParameter4dNV) ; .set GL_PREFIX(ProgramParameter4dNV), GL_PREFIX(ProgramEnvParameter4dARB) + .globl GL_PREFIX(ProgramParameter4dvNV) ; .set GL_PREFIX(ProgramParameter4dvNV), GL_PREFIX(ProgramEnvParameter4dvARB) + .globl GL_PREFIX(ProgramParameter4fNV) ; .set GL_PREFIX(ProgramParameter4fNV), GL_PREFIX(ProgramEnvParameter4fARB) + .globl GL_PREFIX(ProgramParameter4fvNV) ; .set GL_PREFIX(ProgramParameter4fvNV), GL_PREFIX(ProgramEnvParameter4fvARB) .globl GL_PREFIX(VertexAttrib1d) ; .set GL_PREFIX(VertexAttrib1d), GL_PREFIX(VertexAttrib1dARB) .globl GL_PREFIX(VertexAttrib1dv) ; .set GL_PREFIX(VertexAttrib1dv), GL_PREFIX(VertexAttrib1dvARB) .globl GL_PREFIX(VertexAttrib1f) ; .set GL_PREFIX(VertexAttrib1f), GL_PREFIX(VertexAttrib1fARB) @@ -29578,7 +29402,7 @@ GL_PREFIX(_dispatch_stub_773): .globl GL_PREFIX(IsProgramARB) ; .set GL_PREFIX(IsProgramARB), GL_PREFIX(IsProgramNV) .globl GL_PREFIX(PointParameteri) ; .set GL_PREFIX(PointParameteri), GL_PREFIX(PointParameteriNV) .globl GL_PREFIX(PointParameteriv) ; .set GL_PREFIX(PointParameteriv), GL_PREFIX(PointParameterivNV) - .globl GL_PREFIX(BlendEquationSeparate) ; .set GL_PREFIX(BlendEquationSeparate), GL_PREFIX(_dispatch_stub_749) + .globl GL_PREFIX(BlendEquationSeparate) ; .set GL_PREFIX(BlendEquationSeparate), GL_PREFIX(_dispatch_stub_745) #if defined(GLX_USE_TLS) && defined(__linux__) .section ".note.ABI-tag", "a" diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S index 40ecc20753..9a5d875e21 100644 --- a/src/mesa/x86/glapi_x86.S +++ b/src/mesa/x86/glapi_x86.S @@ -27,7 +27,7 @@ */ #include "assyntax.h" -#include "glapi/glapioffsets.h" +#include "glapioffsets.h" #if defined(STDCALL_API) # if defined(USE_MGL_NAMESPACE) @@ -844,10 +844,6 @@ GLNAME(gl_dispatch_functions_start): GL_STUB(GetVertexAttribivNV, _gloffset_GetVertexAttribivNV, GetVertexAttribivNV@12) GL_STUB(IsProgramNV, _gloffset_IsProgramNV, IsProgramNV@4) GL_STUB(LoadProgramNV, _gloffset_LoadProgramNV, LoadProgramNV@16) - GL_STUB(ProgramParameter4dNV, _gloffset_ProgramParameter4dNV, ProgramParameter4dNV@40) - GL_STUB(ProgramParameter4dvNV, _gloffset_ProgramParameter4dvNV, ProgramParameter4dvNV@12) - GL_STUB(ProgramParameter4fNV, _gloffset_ProgramParameter4fNV, ProgramParameter4fNV@24) - GL_STUB(ProgramParameter4fvNV, _gloffset_ProgramParameter4fvNV, ProgramParameter4fvNV@12) GL_STUB(ProgramParameters4dvNV, _gloffset_ProgramParameters4dvNV, ProgramParameters4dvNV@16) GL_STUB(ProgramParameters4fvNV, _gloffset_ProgramParameters4fvNV, ProgramParameters4fvNV@16) GL_STUB(RequestResidentProgramsNV, _gloffset_RequestResidentProgramsNV, RequestResidentProgramsNV@8) @@ -908,26 +904,26 @@ GLNAME(gl_dispatch_functions_start): GL_STUB(SetFragmentShaderConstantATI, _gloffset_SetFragmentShaderConstantATI, SetFragmentShaderConstantATI@8) GL_STUB(PointParameteriNV, _gloffset_PointParameteriNV, PointParameteriNV@8) GL_STUB(PointParameterivNV, _gloffset_PointParameterivNV, PointParameterivNV@8) - GL_STUB(_dispatch_stub_737, _gloffset_ActiveStencilFaceEXT, _dispatch_stub_737@4) + GL_STUB(_dispatch_stub_733, _gloffset_ActiveStencilFaceEXT, _dispatch_stub_733@4) + HIDDEN(GL_PREFIX(_dispatch_stub_733, _dispatch_stub_733@4)) + GL_STUB(_dispatch_stub_734, _gloffset_BindVertexArrayAPPLE, _dispatch_stub_734@4) + HIDDEN(GL_PREFIX(_dispatch_stub_734, _dispatch_stub_734@4)) + GL_STUB(_dispatch_stub_735, _gloffset_DeleteVertexArraysAPPLE, _dispatch_stub_735@8) + HIDDEN(GL_PREFIX(_dispatch_stub_735, _dispatch_stub_735@8)) + GL_STUB(_dispatch_stub_736, _gloffset_GenVertexArraysAPPLE, _dispatch_stub_736@8) + HIDDEN(GL_PREFIX(_dispatch_stub_736, _dispatch_stub_736@8)) + GL_STUB(_dispatch_stub_737, _gloffset_IsVertexArrayAPPLE, _dispatch_stub_737@4) HIDDEN(GL_PREFIX(_dispatch_stub_737, _dispatch_stub_737@4)) - GL_STUB(_dispatch_stub_738, _gloffset_BindVertexArrayAPPLE, _dispatch_stub_738@4) - HIDDEN(GL_PREFIX(_dispatch_stub_738, _dispatch_stub_738@4)) - GL_STUB(_dispatch_stub_739, _gloffset_DeleteVertexArraysAPPLE, _dispatch_stub_739@8) - HIDDEN(GL_PREFIX(_dispatch_stub_739, _dispatch_stub_739@8)) - GL_STUB(_dispatch_stub_740, _gloffset_GenVertexArraysAPPLE, _dispatch_stub_740@8) - HIDDEN(GL_PREFIX(_dispatch_stub_740, _dispatch_stub_740@8)) - GL_STUB(_dispatch_stub_741, _gloffset_IsVertexArrayAPPLE, _dispatch_stub_741@4) - HIDDEN(GL_PREFIX(_dispatch_stub_741, _dispatch_stub_741@4)) GL_STUB(GetProgramNamedParameterdvNV, _gloffset_GetProgramNamedParameterdvNV, GetProgramNamedParameterdvNV@16) GL_STUB(GetProgramNamedParameterfvNV, _gloffset_GetProgramNamedParameterfvNV, GetProgramNamedParameterfvNV@16) GL_STUB(ProgramNamedParameter4dNV, _gloffset_ProgramNamedParameter4dNV, ProgramNamedParameter4dNV@44) GL_STUB(ProgramNamedParameter4dvNV, _gloffset_ProgramNamedParameter4dvNV, ProgramNamedParameter4dvNV@16) GL_STUB(ProgramNamedParameter4fNV, _gloffset_ProgramNamedParameter4fNV, ProgramNamedParameter4fNV@28) GL_STUB(ProgramNamedParameter4fvNV, _gloffset_ProgramNamedParameter4fvNV, ProgramNamedParameter4fvNV@16) - GL_STUB(_dispatch_stub_748, _gloffset_DepthBoundsEXT, _dispatch_stub_748@16) - HIDDEN(GL_PREFIX(_dispatch_stub_748, _dispatch_stub_748@16)) - GL_STUB(_dispatch_stub_749, _gloffset_BlendEquationSeparateEXT, _dispatch_stub_749@8) - HIDDEN(GL_PREFIX(_dispatch_stub_749, _dispatch_stub_749@8)) + GL_STUB(_dispatch_stub_744, _gloffset_DepthBoundsEXT, _dispatch_stub_744@16) + HIDDEN(GL_PREFIX(_dispatch_stub_744, _dispatch_stub_744@16)) + GL_STUB(_dispatch_stub_745, _gloffset_BlendEquationSeparateEXT, _dispatch_stub_745@8) + HIDDEN(GL_PREFIX(_dispatch_stub_745, _dispatch_stub_745@8)) GL_STUB(BindFramebufferEXT, _gloffset_BindFramebufferEXT, BindFramebufferEXT@8) GL_STUB(BindRenderbufferEXT, _gloffset_BindRenderbufferEXT, BindRenderbufferEXT@8) GL_STUB(CheckFramebufferStatusEXT, _gloffset_CheckFramebufferStatusEXT, CheckFramebufferStatusEXT@4) @@ -945,19 +941,19 @@ GLNAME(gl_dispatch_functions_start): GL_STUB(IsFramebufferEXT, _gloffset_IsFramebufferEXT, IsFramebufferEXT@4) GL_STUB(IsRenderbufferEXT, _gloffset_IsRenderbufferEXT, IsRenderbufferEXT@4) GL_STUB(RenderbufferStorageEXT, _gloffset_RenderbufferStorageEXT, RenderbufferStorageEXT@16) - GL_STUB(_dispatch_stub_767, _gloffset_BlitFramebufferEXT, _dispatch_stub_767@40) - HIDDEN(GL_PREFIX(_dispatch_stub_767, _dispatch_stub_767@40)) + GL_STUB(_dispatch_stub_763, _gloffset_BlitFramebufferEXT, _dispatch_stub_763@40) + HIDDEN(GL_PREFIX(_dispatch_stub_763, _dispatch_stub_763@40)) GL_STUB(FramebufferTextureLayerEXT, _gloffset_FramebufferTextureLayerEXT, FramebufferTextureLayerEXT@20) - GL_STUB(_dispatch_stub_769, _gloffset_StencilFuncSeparateATI, _dispatch_stub_769@16) - HIDDEN(GL_PREFIX(_dispatch_stub_769, _dispatch_stub_769@16)) - GL_STUB(_dispatch_stub_770, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_770@16) - HIDDEN(GL_PREFIX(_dispatch_stub_770, _dispatch_stub_770@16)) - GL_STUB(_dispatch_stub_771, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_771@16) - HIDDEN(GL_PREFIX(_dispatch_stub_771, _dispatch_stub_771@16)) - GL_STUB(_dispatch_stub_772, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_772@12) - HIDDEN(GL_PREFIX(_dispatch_stub_772, _dispatch_stub_772@12)) - GL_STUB(_dispatch_stub_773, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_773@12) - HIDDEN(GL_PREFIX(_dispatch_stub_773, _dispatch_stub_773@12)) + GL_STUB(_dispatch_stub_765, _gloffset_StencilFuncSeparateATI, _dispatch_stub_765@16) + HIDDEN(GL_PREFIX(_dispatch_stub_765, _dispatch_stub_765@16)) + GL_STUB(_dispatch_stub_766, _gloffset_ProgramEnvParameters4fvEXT, _dispatch_stub_766@16) + HIDDEN(GL_PREFIX(_dispatch_stub_766, _dispatch_stub_766@16)) + GL_STUB(_dispatch_stub_767, _gloffset_ProgramLocalParameters4fvEXT, _dispatch_stub_767@16) + HIDDEN(GL_PREFIX(_dispatch_stub_767, _dispatch_stub_767@16)) + GL_STUB(_dispatch_stub_768, _gloffset_GetQueryObjecti64vEXT, _dispatch_stub_768@12) + HIDDEN(GL_PREFIX(_dispatch_stub_768, _dispatch_stub_768@12)) + GL_STUB(_dispatch_stub_769, _gloffset_GetQueryObjectui64vEXT, _dispatch_stub_769@12) + HIDDEN(GL_PREFIX(_dispatch_stub_769, _dispatch_stub_769@12)) GL_STUB_ALIAS(ArrayElementEXT, _gloffset_ArrayElement, ArrayElementEXT@4, ArrayElement, ArrayElement@4) GL_STUB_ALIAS(BindTextureEXT, _gloffset_BindTexture, BindTextureEXT@8, BindTexture, BindTexture@8) GL_STUB_ALIAS(DrawArraysEXT, _gloffset_DrawArrays, DrawArraysEXT@12, DrawArrays, DrawArrays@12) @@ -1107,6 +1103,10 @@ GLNAME(gl_dispatch_functions_start): GL_STUB_ALIAS(GetVertexAttribdv, _gloffset_GetVertexAttribdvARB, GetVertexAttribdv@12, GetVertexAttribdvARB, GetVertexAttribdvARB@12) GL_STUB_ALIAS(GetVertexAttribfv, _gloffset_GetVertexAttribfvARB, GetVertexAttribfv@12, GetVertexAttribfvARB, GetVertexAttribfvARB@12) GL_STUB_ALIAS(GetVertexAttribiv, _gloffset_GetVertexAttribivARB, GetVertexAttribiv@12, GetVertexAttribivARB, GetVertexAttribivARB@12) + GL_STUB_ALIAS(ProgramParameter4dNV, _gloffset_ProgramEnvParameter4dARB, ProgramParameter4dNV@40, ProgramEnvParameter4dARB, ProgramEnvParameter4dARB@40) + GL_STUB_ALIAS(ProgramParameter4dvNV, _gloffset_ProgramEnvParameter4dvARB, ProgramParameter4dvNV@12, ProgramEnvParameter4dvARB, ProgramEnvParameter4dvARB@12) + GL_STUB_ALIAS(ProgramParameter4fNV, _gloffset_ProgramEnvParameter4fARB, ProgramParameter4fNV@24, ProgramEnvParameter4fARB, ProgramEnvParameter4fARB@24) + GL_STUB_ALIAS(ProgramParameter4fvNV, _gloffset_ProgramEnvParameter4fvARB, ProgramParameter4fvNV@12, ProgramEnvParameter4fvARB, ProgramEnvParameter4fvARB@12) GL_STUB_ALIAS(VertexAttrib1d, _gloffset_VertexAttrib1dARB, VertexAttrib1d@12, VertexAttrib1dARB, VertexAttrib1dARB@12) GL_STUB_ALIAS(VertexAttrib1dv, _gloffset_VertexAttrib1dvARB, VertexAttrib1dv@8, VertexAttrib1dvARB, VertexAttrib1dvARB@8) GL_STUB_ALIAS(VertexAttrib1f, _gloffset_VertexAttrib1fARB, VertexAttrib1f@8, VertexAttrib1fARB, VertexAttrib1fARB@8) -- cgit v1.2.3 From 1036ec94e5a30f9068a1b41f2e8668daa2ee4a2c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sun, 18 May 2008 15:32:58 -0600 Subject: Fix some bugs in material, depth range state picked from master --- src/mesa/shader/prog_statevars.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index eee4d2d925..ba3c988445 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -1,6 +1,6 @@ /* * Mesa 3-D graphics library - * Version: 7.0 + * Version: 7.1 * * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * @@ -181,7 +181,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_AMBIENT+face][i]; } /* [3] = material alpha */ - value[3] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE+face][3]; + value[3] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_AMBIENT+face][3]; return; case STATE_DIFFUSE: for (i = 0; i < 3; i++) { @@ -197,7 +197,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_SPECULAR+face][i]; } /* [3] = material alpha */ - value[3] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE+face][3]; + value[3] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_SPECULAR+face][3]; return; default: _mesa_problem(ctx, "Invalid lightprod state in fetch_state"); @@ -354,7 +354,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], value[0] = ctx->Viewport.Near; /* near */ value[1] = ctx->Viewport.Far; /* far */ value[2] = ctx->Viewport.Far - ctx->Viewport.Near; /* far - near */ - value[3] = 0; + value[3] = 1.0; return; case STATE_FRAGMENT_PROGRAM: { -- cgit v1.2.3 From a0bfeb0c3ca58a1f4d978f2aaa343ed4009de079 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sun, 18 May 2008 15:46:58 -0600 Subject: print vertex.attrib[n] (from master) --- src/mesa/shader/prog_print.c | 22 +++++++++++++++++++--- 1 file changed, 19 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index 1145f0c7a4..1c35ce3fec 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -99,7 +99,23 @@ arb_input_attrib_string(GLint index, GLenum progType) "vertex.texcoord[4]", "vertex.texcoord[5]", "vertex.texcoord[6]", - "vertex.texcoord[7]" + "vertex.texcoord[7]", + "vertex.attrib[0]", + "vertex.attrib[1]", + "vertex.attrib[2]", + "vertex.attrib[3]", + "vertex.attrib[4]", + "vertex.attrib[5]", + "vertex.attrib[6]", + "vertex.attrib[7]", + "vertex.attrib[8]", + "vertex.attrib[9]", + "vertex.attrib[10]", + "vertex.attrib[11]", + "vertex.attrib[12]", + "vertex.attrib[13]", + "vertex.attrib[14]", + "vertex.attrib[15]" }; const char *fragAttribs[] = { "fragment.position", @@ -751,9 +767,9 @@ _mesa_print_program_parameters(GLcontext *ctx, const struct gl_program *prog) _mesa_printf("%d ", prog->SamplerUnits[i]); } _mesa_printf("]\n"); - + _mesa_load_state_parameters(ctx, prog->Parameters); - + #if 0 _mesa_printf("Local Params:\n"); for (i = 0; i < MAX_PROGRAM_LOCAL_PARAMS; i++){ -- cgit v1.2.3 From 88cdc5e23fadc81dcf8ab068c0620cf67712d01f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sun, 18 May 2008 16:37:35 -0600 Subject: fix program delete/refcount bug --- src/mesa/state_tracker/st_cb_program.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index a293ec3f0b..3154fc85aa 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -159,7 +159,9 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) } if (stfp->bitmap_program) { - st_delete_program(ctx, &stfp->bitmap_program->Base.Base); + struct gl_program *prg = &stfp->bitmap_program->Base.Base; + _mesa_reference_program(ctx, &prg, NULL); + stfp->bitmap_program = NULL; } st_free_translated_vertex_programs(st, stfp->vertex_programs); @@ -169,6 +171,7 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) assert(0); /* problem */ } + /* delete base class */ _mesa_delete_program( ctx, prog ); } -- cgit v1.2.3 From 820fe368635c82ed1c3459b0f23cfd9c0cc762de Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 17 May 2008 00:12:43 +0900 Subject: gallium: Fix typo. --- src/gallium/README.portability | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/README.portability b/src/gallium/README.portability index 18a97f449b..d5d5987a7f 100644 --- a/src/gallium/README.portability +++ b/src/gallium/README.portability @@ -6,7 +6,7 @@ The state tracker and winsys driver support a rather limited number of platforms. However, the pipe drivers are meant to run in a wide number of platforms. Hence the pipe drivers, the auxiliary modules, and all public -headers in general, should stricly follow these guidelines to ensure +headers in general, should strictly follow these guidelines to ensure = Compiler Support = -- cgit v1.2.3 From 5ca4f9e97ce153ace0fb8f40c772f44895cb20fe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 19 May 2008 14:47:37 +0900 Subject: gallium: Additional formats for bump mapping. --- src/gallium/include/pipe/p_format.h | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index a4bd23c302..9ba00f8d7b 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -307,6 +307,9 @@ enum pipe_format { PIPE_FORMAT_R8G8B8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SNORM ), PIPE_FORMAT_R8G8B8A8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), PIPE_FORMAT_R8G8B8X8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_B6G5R5_SNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_BGR1, 6, 5, 5, 0, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_A8B8G8R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_BGRA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), + PIPE_FORMAT_X8B8G8R8_SNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SNORM ), PIPE_FORMAT_R8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R8G8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R8G8B8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ), -- cgit v1.2.3 From aafe3ddee276230243ed6767803ce4b6ce2aeec6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 19 May 2008 15:18:40 +0900 Subject: i915simple: Fix win9x build. --- src/gallium/drivers/i915simple/i915_state.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 4adeb37e86..e09c225900 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -466,7 +466,7 @@ void i915_delete_fs_state(struct pipe_context *pipe, void *shader) FREE(ifs->program); ifs->program_len = 0; - FREE(ifs->state.tokens); + FREE((struct tgsi_token *)ifs->state.tokens); FREE(ifs); } -- cgit v1.2.3 From 62f96ddbbc2549bd4d50016f571cd4d1f6f6a7d6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 19 May 2008 08:59:41 -0600 Subject: Fix program refcounting assertion failure during context tear-down When purging the program hash table, the refcount _should_ be one since the program is referenced by the hash table. Need to explicitly set to zero before calling delete(). Also, purge high-level shader hash tables before low-level program hash tabl cherry-picked from master --- src/mesa/main/context.c | 46 ++++++++++++++++++++++++---------------------- 1 file changed, 24 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 893c79f28c..2158eb6873 100755 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -593,6 +593,8 @@ delete_program_cb(GLuint id, void *data, void *userData) { struct gl_program *prog = (struct gl_program *) data; GLcontext *ctx = (GLcontext *) userData; + ASSERT(prog->RefCount == 1); /* should only be referenced by hash table */ + prog->RefCount = 0; /* now going away */ ctx->Driver.DeleteProgram(ctx, prog); } @@ -686,22 +688,11 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) _mesa_HashDeleteAll(ss->DisplayList, delete_displaylist_cb, ctx); _mesa_DeleteHashTable(ss->DisplayList); - /* - * Free texture objects - */ - ASSERT(ctx->Driver.DeleteTexture); - /* the default textures */ - ctx->Driver.DeleteTexture(ctx, ss->Default1D); - ctx->Driver.DeleteTexture(ctx, ss->Default2D); - ctx->Driver.DeleteTexture(ctx, ss->Default3D); - ctx->Driver.DeleteTexture(ctx, ss->DefaultCubeMap); - ctx->Driver.DeleteTexture(ctx, ss->DefaultRect); - ctx->Driver.DeleteTexture(ctx, ss->Default1DArray); - ctx->Driver.DeleteTexture(ctx, ss->Default2DArray); - - /* all other textures */ - _mesa_HashDeleteAll(ss->TexObjects, delete_texture_cb, ctx); - _mesa_DeleteHashTable(ss->TexObjects); +#if FEATURE_ARB_shader_objects + _mesa_HashWalk(ss->ShaderObjects, free_shader_program_data_cb, ctx); + _mesa_HashDeleteAll(ss->ShaderObjects, delete_shader_cb, ctx); + _mesa_DeleteHashTable(ss->ShaderObjects); +#endif #if defined(FEATURE_NV_vertex_program) || defined(FEATURE_NV_fragment_program) _mesa_HashDeleteAll(ss->Programs, delete_program_cb, ctx); @@ -728,17 +719,28 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) _mesa_HashDeleteAll(ss->ArrayObjects, delete_arrayobj_cb, ctx); _mesa_DeleteHashTable(ss->ArrayObjects); -#if FEATURE_ARB_shader_objects - _mesa_HashWalk(ss->ShaderObjects, free_shader_program_data_cb, ctx); - _mesa_HashDeleteAll(ss->ShaderObjects, delete_shader_cb, ctx); - _mesa_DeleteHashTable(ss->ShaderObjects); -#endif - #if FEATURE_EXT_framebuffer_object _mesa_DeleteHashTable(ss->FrameBuffers); _mesa_DeleteHashTable(ss->RenderBuffers); #endif + /* + * Free texture objects (after FBOs since some textures might have + * been bound to FBOs). + */ + ASSERT(ctx->Driver.DeleteTexture); + /* the default textures */ + ctx->Driver.DeleteTexture(ctx, ss->Default1D); + ctx->Driver.DeleteTexture(ctx, ss->Default2D); + ctx->Driver.DeleteTexture(ctx, ss->Default3D); + ctx->Driver.DeleteTexture(ctx, ss->DefaultCubeMap); + ctx->Driver.DeleteTexture(ctx, ss->DefaultRect); + ctx->Driver.DeleteTexture(ctx, ss->Default1DArray); + ctx->Driver.DeleteTexture(ctx, ss->Default2DArray); + /* all other textures */ + _mesa_HashDeleteAll(ss->TexObjects, delete_texture_cb, ctx); + _mesa_DeleteHashTable(ss->TexObjects); + _glthread_DESTROY_MUTEX(ss->Mutex); _mesa_free(ss); -- cgit v1.2.3 From 38fb20b2240e826394dbf56dc54f226c5f887631 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 19 May 2008 09:04:02 -0600 Subject: fix program delete/refcount bug cherry-picked from gallium-tex-surfaces --- src/mesa/state_tracker/st_cb_program.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index a293ec3f0b..b23e000a0b 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -159,7 +159,9 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) } if (stfp->bitmap_program) { - st_delete_program(ctx, &stfp->bitmap_program->Base.Base); + struct gl_program *prg = &stfp->bitmap_program->Base.Base; + _mesa_reference_program(ctx, &prg, NULL); + stfp->bitmap_program = NULL; } st_free_translated_vertex_programs(st, stfp->vertex_programs); -- cgit v1.2.3 From 59007a811de2d76ea00164e8f1cacb4a375d1458 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 19 May 2008 09:34:28 -0600 Subject: if x86_get_func() returns NULL, handle it properly instead of aborting --- src/gallium/auxiliary/draw/draw_vs_sse.c | 3 +++ src/gallium/drivers/softpipe/sp_fs_sse.c | 6 +++++- 2 files changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 5929ea76b2..e3f4e67472 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -215,6 +215,9 @@ draw_create_vs_sse(struct draw_context *draw, goto fail; vs->func = (codegen_function) x86_get_func( &vs->sse2_program ); + if (!vs->func) { + goto fail; + } return &vs->base; diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 25fdfea491..55741cc1df 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -139,7 +139,11 @@ softpipe_create_fs_sse(struct softpipe_context *softpipe, } shader->func = (codegen_function) x86_get_func( &shader->sse2_program ); - assert(shader->func); + if (!shader->func) { + x86_release_func( &shader->sse2_program ); + FREE(shader); + return NULL; + } shader->base.shader = *templ; shader->base.prepare = fs_sse_prepare; -- cgit v1.2.3 From 21e614eabc5e6a502504f307f3710b4dd0417923 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 19 May 2008 12:40:11 -0600 Subject: gallium: fix some texture object leaks --- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 8 ++++++++ src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 5 +++++ 2 files changed, 13 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index f501b2aed4..346a96e37f 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -713,6 +713,11 @@ static void aaline_destroy(struct draw_stage *stage) { struct aaline_stage *aaline = aaline_stage(stage); + uint i; + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + pipe_texture_reference(&aaline->state.texture[i], NULL); + } if (aaline->sampler_cso) aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso); @@ -836,6 +841,9 @@ aaline_set_sampler_textures(struct pipe_context *pipe, for (i = 0; i < num; i++) { pipe_texture_reference(&aaline->state.texture[i], texture[i]); } + for ( ; i < PIPE_MAX_SAMPLERS; i++) { + pipe_texture_reference(&aaline->state.texture[i], NULL); + } aaline->num_textures = num; /* pass-through */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 73ee419858..2dfd1800d5 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -559,6 +559,11 @@ static void pstip_destroy(struct draw_stage *stage) { struct pstip_stage *pstip = pstip_stage(stage); + uint i; + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) { + pipe_texture_reference(&pstip->state.textures[i], NULL); + } pstip->pipe->delete_sampler_state(pstip->pipe, pstip->sampler_cso); -- cgit v1.2.3 From 8493f0849f6fa6ef87388c3a7e179a329447a763 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 19 May 2008 16:04:02 -0600 Subject: fix tempReg test in _mesa_combine_programs() --- src/mesa/shader/program.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index e06ab5aa23..a0817a91ec 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -602,7 +602,7 @@ _mesa_combine_programs(GLcontext *ctx, if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) && (progB->InputsRead & (1 << FRAG_ATTRIB_COL0))) { GLint tempReg = _mesa_find_free_register(newProg, PROGRAM_TEMPORARY); - if (!tempReg) { + if (tempReg < 0) { _mesa_problem(ctx, "No free temp regs found in " "_mesa_combine_programs(), using 31"); tempReg = 31; -- cgit v1.2.3 From 655374bda7dc7ae6fc1a0ef64ab868e3c8be9aa0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 19 May 2008 16:05:41 -0600 Subject: gallium: remove dead code --- src/mesa/state_tracker/st_cb_bitmap.c | 9 --------- 1 file changed, 9 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index ce8fefe703..edf4be172b 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -743,15 +743,6 @@ st_destroy_bitmap(struct st_context *st) { struct pipe_context *pipe = st->pipe; -#if 0 - if (st->bitmap.combined_prog) { - st_delete_program(st->ctx, &st->bitmap.combined_prog->Base.Base); - } - - if (st->bitmap.program) { - st_delete_program(st->ctx, &st->bitmap.program->Base.Base); - } -#endif if (st->bitmap.vs) { cso_delete_vertex_shader(st->cso_context, st->bitmap.vs); st->bitmap.vs = NULL; -- cgit v1.2.3 From 65075da8c39f686ec4ab6593d7b0574982d21a23 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 19 May 2008 16:04:02 -0600 Subject: fix tempReg test in _mesa_combine_programs() --- src/mesa/shader/program.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index e06ab5aa23..a0817a91ec 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -602,7 +602,7 @@ _mesa_combine_programs(GLcontext *ctx, if ((progA->OutputsWritten & (1 << FRAG_RESULT_COLR)) && (progB->InputsRead & (1 << FRAG_ATTRIB_COL0))) { GLint tempReg = _mesa_find_free_register(newProg, PROGRAM_TEMPORARY); - if (!tempReg) { + if (tempReg < 0) { _mesa_problem(ctx, "No free temp regs found in " "_mesa_combine_programs(), using 31"); tempReg = 31; -- cgit v1.2.3 From de03e8e9604afd7b61002e8d7e0f94a859734abf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 19 May 2008 16:05:41 -0600 Subject: gallium: remove dead code --- src/mesa/state_tracker/st_cb_bitmap.c | 9 --------- 1 file changed, 9 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index d39fce1bb7..c8e2d7d118 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -753,15 +753,6 @@ st_destroy_bitmap(struct st_context *st) screen->surface_unmap(screen, cache->surf); screen->tex_surface_release(screen, &cache->surf); -#if 0 - if (st->bitmap.combined_prog) { - st_delete_program(st->ctx, &st->bitmap.combined_prog->Base.Base); - } - - if (st->bitmap.program) { - st_delete_program(st->ctx, &st->bitmap.program->Base.Base); - } -#endif if (st->bitmap.vs) { cso_delete_vertex_shader(st->cso_context, st->bitmap.vs); st->bitmap.vs = NULL; -- cgit v1.2.3 From c95c4efa54175f0adcffacb71e3d66839617ceba Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 11:29:58 -0600 Subject: fix incorrect sampler numbering/indexing. All samplers indexes were zero. cherry-picked from master (b6fb0940c226373ac235a5d327d3fcfd742bc6b9) --- src/mesa/shader/prog_parameter.c | 20 +++++++++++++------- src/mesa/shader/prog_parameter.h | 2 +- src/mesa/shader/slang/slang_codegen.c | 3 +-- src/mesa/shader/slang/slang_compile.c | 3 --- src/mesa/shader/slang/slang_typeinfo.h | 1 - 5 files changed, 15 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index 0b61bac696..75cf5023df 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -283,25 +283,31 @@ _mesa_add_uniform(struct gl_program_parameter_list *paramList, * \param name uniform's name * \param datatype GL_SAMPLER_2D, GL_SAMPLER_2D_RECT_ARB, etc. * \param index the sampler number (as seen in TEX instructions) + * \return sampler index (starting at zero) or -1 if error */ GLint _mesa_add_sampler(struct gl_program_parameter_list *paramList, - const char *name, GLenum datatype, GLuint index) + const char *name, GLenum datatype) { GLint i = _mesa_lookup_parameter_index(paramList, -1, name); if (i >= 0 && paramList->Parameters[i].Type == PROGRAM_SAMPLER) { ASSERT(paramList->Parameters[i].Size == 1); ASSERT(paramList->Parameters[i].DataType == datatype); - ASSERT(paramList->ParameterValues[i][0] == index); /* already in list */ - return i; + return (GLint) paramList->ParameterValues[i][0]; } else { - GLfloat indexf = index; const GLint size = 1; /* a sampler is basically a texture unit number */ - i = _mesa_add_parameter(paramList, PROGRAM_SAMPLER, name, - size, datatype, &indexf, NULL); - return i; + GLfloat value; + GLint numSamplers = 0; + for (i = 0; i < paramList->NumParameters; i++) { + if (paramList->Parameters[i].Type == PROGRAM_SAMPLER) + numSamplers++; + } + value = (GLfloat) numSamplers; + (void) _mesa_add_parameter(paramList, PROGRAM_SAMPLER, name, + size, datatype, &value, NULL); + return numSamplers; } } diff --git a/src/mesa/shader/prog_parameter.h b/src/mesa/shader/prog_parameter.h index d6cc03448c..105f6f24de 100644 --- a/src/mesa/shader/prog_parameter.h +++ b/src/mesa/shader/prog_parameter.h @@ -114,7 +114,7 @@ _mesa_add_uniform(struct gl_program_parameter_list *paramList, extern GLint _mesa_add_sampler(struct gl_program_parameter_list *paramList, - const char *name, GLenum datatype, GLuint index); + const char *name, GLenum datatype); extern GLint _mesa_add_varying(struct gl_program_parameter_list *paramList, diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 8e4c10ff43..d19d5a0abb 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2841,8 +2841,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, * store->Index = sampler number (0..7, typically) * store->Size = texture type index (1D, 2D, 3D, cube, etc) */ - const GLint sampNum = A->numSamplers++; - _mesa_add_sampler(prog->Parameters, varName, datatype, sampNum); + GLint sampNum = _mesa_add_sampler(prog->Parameters, varName, datatype); store = _slang_new_ir_storage(PROGRAM_SAMPLER, sampNum, texIndex); if (dbg) printf("SAMPLER "); } diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 1449888f9f..46b5c54bbe 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -1620,7 +1620,6 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O, A.program = O->program; A.vartable = O->vartable; A.curFuncEndLabel = NULL; - A.numSamplers = 0; if (!_slang_codegen_global_variable(&A, var, C->type)) return 0; } @@ -1643,7 +1642,6 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O, A.space.funcs = O->funs; A.space.structs = O->structs; A.space.vars = O->vars; - A.numSamplers = 0; if (!initialize_global(&A, var)) return 0; } @@ -1777,7 +1775,6 @@ parse_function(slang_parse_ctx * C, slang_output_ctx * O, int definition, A.program = O->program; A.vartable = O->vartable; A.log = C->L; - A.numSamplers = 0; _slang_codegen_function(&A, *parsed_func_ret); } diff --git a/src/mesa/shader/slang/slang_typeinfo.h b/src/mesa/shader/slang/slang_typeinfo.h index ad5aa3e195..587331e8b1 100644 --- a/src/mesa/shader/slang/slang_typeinfo.h +++ b/src/mesa/shader/slang/slang_typeinfo.h @@ -65,7 +65,6 @@ typedef struct slang_assemble_ctx_ struct slang_label_ *curFuncEndLabel; struct slang_ir_node_ *CurLoop; struct slang_function_ *CurFunction; - GLuint numSamplers; } slang_assemble_ctx; -- cgit v1.2.3 From 34149ec89d71aa550ac4477050fa79d66bb94f25 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 11:31:20 -0600 Subject: copy StateFlags in _mesa_clone_parameter_list() cherry-picked from master (b539b6132141b0f583076355ad158852667b79cd) --- src/mesa/shader/prog_parameter.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index 75cf5023df..152bd79f69 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -608,6 +608,8 @@ _mesa_clone_parameter_list(const struct gl_program_parameter_list *list) } } + clone->StateFlags = list->StateFlags; + return clone; } -- cgit v1.2.3 From 5c431c22227fdc552b34a5feabf1d339dcfe9848 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 13:38:45 -0600 Subject: gallium: clean-up glDraw/CopyPixels shaders when destroying context --- src/mesa/state_tracker/st_cb_drawpixels.c | 39 +++++++++++++++++++++---------- src/mesa/state_tracker/st_cb_drawpixels.h | 3 +++ src/mesa/state_tracker/st_context.c | 1 + src/mesa/state_tracker/st_context.h | 6 +++++ 4 files changed, 37 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 56b949cf96..fbbe8d2d64 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -164,14 +164,16 @@ static struct st_fragment_program * make_fragment_shader_z(struct st_context *st) { GLcontext *ctx = st->ctx; - /* only make programs once and re-use */ - static struct st_fragment_program *stfp = NULL; struct gl_program *p; GLuint ic = 0; - if (stfp) - return stfp; + if (st->drawpix.z_shader) { + return st->drawpix.z_shader; + } + /* + * Create shader now + */ p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) return NULL; @@ -213,10 +215,10 @@ make_fragment_shader_z(struct st_context *st) p->OutputsWritten = (1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR); p->SamplersUsed = 0x1; /* sampler 0 (bit 0) is used */ - stfp = (struct st_fragment_program *) p; - st_translate_fragment_program(st, stfp, NULL); + st->drawpix.z_shader = (struct st_fragment_program *) p; + st_translate_fragment_program(st, st->drawpix.z_shader, NULL); - return stfp; + return st->drawpix.z_shader; } @@ -228,16 +230,17 @@ make_fragment_shader_z(struct st_context *st) static struct st_vertex_program * st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor) { - /* only make programs once and re-use */ - static struct st_vertex_program *progs[2] = { NULL, NULL }; GLcontext *ctx = st->ctx; struct st_vertex_program *stvp; struct gl_program *p; GLuint ic = 0; - if (progs[passColor]) - return progs[passColor]; + if (st->drawpix.vert_shaders[passColor]) + return st->drawpix.vert_shaders[passColor]; + /* + * Create shader now + */ p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); if (!p) return NULL; @@ -293,7 +296,7 @@ st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor) stvp = (struct st_vertex_program *) p; st_translate_vertex_program(st, stvp, NULL); - progs[passColor] = stvp; + st->drawpix.vert_shaders[passColor] = stvp; return stvp; } @@ -1042,3 +1045,15 @@ void st_init_drawpixels_functions(struct dd_function_table *functions) functions->DrawPixels = st_DrawPixels; functions->CopyPixels = st_CopyPixels; } + + +void +st_destroy_drawpix(struct st_context *st) +{ + st_reference_fragprog(st, &st->drawpix.z_shader, NULL); + st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL); + st_reference_vertprog(st, &st->drawpix.vert_shaders[0], NULL); + st_reference_vertprog(st, &st->drawpix.vert_shaders[1], NULL); +} + + diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h index 71ba487020..26fe864d18 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.h +++ b/src/mesa/state_tracker/st_cb_drawpixels.h @@ -32,5 +32,8 @@ extern void st_init_drawpixels_functions(struct dd_function_table *functions); +extern void +st_destroy_drawpix(struct st_context *st); + #endif /* ST_CB_DRAWPIXELS_H */ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index fb397ea510..b407fd884b 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -168,6 +168,7 @@ static void st_destroy_context_priv( struct st_context *st ) st_destroy_bitmap(st); st_destroy_blit(st); st_destroy_clear(st); + st_destroy_drawpix(st); _vbo_DestroyContext(st->ctx); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 1ca779d0a9..46c16e45e6 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -152,6 +152,12 @@ struct st_context struct bitmap_cache *cache; } bitmap; + /** for glDraw/CopyPixels */ + struct { + struct st_fragment_program *z_shader; + struct st_vertex_program *vert_shaders[2]; + } drawpix; + /** for glClear */ struct { struct pipe_shader_state vert_shader; -- cgit v1.2.3 From 90ce8467a8cb2b819a44d00e4a9d315f8aa43b7c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 13:38:45 -0600 Subject: gallium: clean-up glDraw/CopyPixels shaders when destroying context --- src/mesa/state_tracker/st_cb_drawpixels.c | 39 +++++++++++++++++++++---------- src/mesa/state_tracker/st_cb_drawpixels.h | 3 +++ src/mesa/state_tracker/st_context.c | 1 + src/mesa/state_tracker/st_context.h | 6 +++++ 4 files changed, 37 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 56b949cf96..fbbe8d2d64 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -164,14 +164,16 @@ static struct st_fragment_program * make_fragment_shader_z(struct st_context *st) { GLcontext *ctx = st->ctx; - /* only make programs once and re-use */ - static struct st_fragment_program *stfp = NULL; struct gl_program *p; GLuint ic = 0; - if (stfp) - return stfp; + if (st->drawpix.z_shader) { + return st->drawpix.z_shader; + } + /* + * Create shader now + */ p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0); if (!p) return NULL; @@ -213,10 +215,10 @@ make_fragment_shader_z(struct st_context *st) p->OutputsWritten = (1 << FRAG_RESULT_COLR) | (1 << FRAG_RESULT_DEPR); p->SamplersUsed = 0x1; /* sampler 0 (bit 0) is used */ - stfp = (struct st_fragment_program *) p; - st_translate_fragment_program(st, stfp, NULL); + st->drawpix.z_shader = (struct st_fragment_program *) p; + st_translate_fragment_program(st, st->drawpix.z_shader, NULL); - return stfp; + return st->drawpix.z_shader; } @@ -228,16 +230,17 @@ make_fragment_shader_z(struct st_context *st) static struct st_vertex_program * st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor) { - /* only make programs once and re-use */ - static struct st_vertex_program *progs[2] = { NULL, NULL }; GLcontext *ctx = st->ctx; struct st_vertex_program *stvp; struct gl_program *p; GLuint ic = 0; - if (progs[passColor]) - return progs[passColor]; + if (st->drawpix.vert_shaders[passColor]) + return st->drawpix.vert_shaders[passColor]; + /* + * Create shader now + */ p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); if (!p) return NULL; @@ -293,7 +296,7 @@ st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor) stvp = (struct st_vertex_program *) p; st_translate_vertex_program(st, stvp, NULL); - progs[passColor] = stvp; + st->drawpix.vert_shaders[passColor] = stvp; return stvp; } @@ -1042,3 +1045,15 @@ void st_init_drawpixels_functions(struct dd_function_table *functions) functions->DrawPixels = st_DrawPixels; functions->CopyPixels = st_CopyPixels; } + + +void +st_destroy_drawpix(struct st_context *st) +{ + st_reference_fragprog(st, &st->drawpix.z_shader, NULL); + st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL); + st_reference_vertprog(st, &st->drawpix.vert_shaders[0], NULL); + st_reference_vertprog(st, &st->drawpix.vert_shaders[1], NULL); +} + + diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h index 71ba487020..26fe864d18 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.h +++ b/src/mesa/state_tracker/st_cb_drawpixels.h @@ -32,5 +32,8 @@ extern void st_init_drawpixels_functions(struct dd_function_table *functions); +extern void +st_destroy_drawpix(struct st_context *st); + #endif /* ST_CB_DRAWPIXELS_H */ diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index fb397ea510..b407fd884b 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -168,6 +168,7 @@ static void st_destroy_context_priv( struct st_context *st ) st_destroy_bitmap(st); st_destroy_blit(st); st_destroy_clear(st); + st_destroy_drawpix(st); _vbo_DestroyContext(st->ctx); diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 1ca779d0a9..46c16e45e6 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -152,6 +152,12 @@ struct st_context struct bitmap_cache *cache; } bitmap; + /** for glDraw/CopyPixels */ + struct { + struct st_fragment_program *z_shader; + struct st_vertex_program *vert_shaders[2]; + } drawpix; + /** for glClear */ struct { struct pipe_shader_state vert_shader; -- cgit v1.2.3 From 4b4ac9ed3497d0cbf58311b83ed4a08a98bb854c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 13:48:34 -0600 Subject: gallium: fix mem leaks --- src/mesa/state_tracker/st_cb_clear.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index b7d7204633..cdfcdcee72 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -97,6 +97,16 @@ st_destroy_clear(struct st_context *st) { struct pipe_context *pipe = st->pipe; + if (st->clear.vert_shader.tokens) { + FREE((void *) st->clear.vert_shader.tokens); + st->clear.vert_shader.tokens = NULL; + } + + if (st->clear.frag_shader.tokens) { + FREE((void *) st->clear.frag_shader.tokens); + st->clear.frag_shader.tokens = NULL; + } + if (st->clear.fs) { cso_delete_fragment_shader(st->cso_context, st->clear.fs); st->clear.fs = NULL; -- cgit v1.2.3 From 13c74914067b7e49be7761534c1f6d3fb90f6f1f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 14:38:00 -0600 Subject: gallium: fix default case in st_new_program() --- src/mesa/state_tracker/st_cb_program.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index b23e000a0b..a1694d93c5 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -89,8 +89,12 @@ static void st_use_program( GLcontext *ctx, +/** + * Called via ctx->Driver.NewProgram() to allocate a new vertex or + * fragment program. + */ static struct gl_program *st_new_program( GLcontext *ctx, - GLenum target, + GLenum target, GLuint id ) { switch (target) { @@ -118,7 +122,8 @@ static struct gl_program *st_new_program( GLcontext *ctx, } default: - return _mesa_new_program(ctx, target, id); + assert(0); + return NULL; } } -- cgit v1.2.3 From 69fd676240f91b1a42a355b768f86d397b899002 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 14:38:22 -0600 Subject: gallium: move vertex/fragment program unbinding --- src/mesa/state_tracker/st_context.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index b407fd884b..4081722bc7 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -158,9 +158,6 @@ static void st_destroy_context_priv( struct st_context *st ) { uint i; - st_reference_fragprog(st, &st->fp, NULL); - st_reference_vertprog(st, &st->vp, NULL); - draw_destroy(st->draw); st_destroy_atoms( st ); st_destroy_draw( st ); @@ -195,6 +192,9 @@ void st_destroy_context( struct st_context *st ) /* need to unbind and destroy CSO objects before anything else */ cso_release_all(st->cso_context); + st_reference_fragprog(st, &st->fp, NULL); + st_reference_vertprog(st, &st->vp, NULL); + _mesa_delete_program_cache(st->ctx, st->pixel_xfer.cache); _mesa_free_context_data(ctx); -- cgit v1.2.3 From ac09b0e0eb30f6691e35638f210c04ac0f867ea8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 15:11:53 -0600 Subject: gallium: fix shader mem leak --- src/mesa/state_tracker/st_cb_bitmap.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index edf4be172b..e95ff5e2e0 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -198,6 +198,10 @@ combined_bitmap_fragment_program(GLcontext *ctx) _mesa_combine_programs(ctx, &bitmap_prog->Base.Base, &stfp->Base.Base); stfp->bitmap_program->bitmap_sampler = sampler; + + /* done with this after combining */ + st_reference_fragprog(st, &bitmap_prog, NULL); + #if 0 { struct gl_program *p = &stfp->bitmap_program->Base.Base; -- cgit v1.2.3 From 7899ecdd6502a323b052f9ad4acd23cbb9ba88db Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 15:12:26 -0600 Subject: gallium: replace assignment with pipe_texture_reference() --- src/mesa/state_tracker/st_gen_mipmap.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 1a0e19c2f9..047743689d 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -197,6 +197,6 @@ st_generate_mipmap(GLcontext *ctx, GLenum target, dstImage->TexFormat = srcImage->TexFormat; stImage = (struct st_texture_image *) dstImage; - stImage->pt = pt; + pipe_texture_reference(&stImage->pt, pt); } } -- cgit v1.2.3 From 26f874e967e5bcbd0e0c73674df3d3900b98d35b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 20 May 2008 15:20:29 -0600 Subject: gallium: fix vertex shader mem leak (glRasterPos) --- src/mesa/state_tracker/st_cb_program.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index a1694d93c5..5f753afdd3 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -143,6 +143,12 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) stvp->driver_shader = NULL; } + if (stvp->draw_shader) { + /* this would only have been allocated for the RasterPos path */ + draw_delete_vertex_shader(st->draw, stvp->draw_shader); + stvp->draw_shader = NULL; + } + if (stvp->state.tokens) { FREE((void *) stvp->state.tokens); stvp->state.tokens = NULL; @@ -225,6 +231,7 @@ static void st_program_string_notify( GLcontext *ctx, } if (stvp->draw_shader) { + /* this would only have been allocated for the RasterPos path */ draw_delete_vertex_shader(st->draw, stvp->draw_shader); stvp->draw_shader = NULL; } -- cgit v1.2.3 From 13a2be4446bd6612b5b56dab1cb9d8975a4f6f1f Mon Sep 17 00:00:00 2001 From: Brian Date: Wed, 21 May 2008 06:59:49 -0600 Subject: mesa: fix #include path for glapioffsets.h (fixes x86 build) --- src/mesa/glapi/gl_x86_asm.py | 2 +- src/mesa/x86/glapi_x86.S | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/glapi/gl_x86_asm.py b/src/mesa/glapi/gl_x86_asm.py index 403e87261b..651cb03f14 100644 --- a/src/mesa/glapi/gl_x86_asm.py +++ b/src/mesa/glapi/gl_x86_asm.py @@ -51,7 +51,7 @@ class PrintGenericStubs(gl_XML.gl_print_base): def printRealHeader(self): print '#include "assyntax.h"' - print '#include "glapioffsets.h"' + print '#include "glapi/glapioffsets.h"' print '' print '#if defined(STDCALL_API)' print '# if defined(USE_MGL_NAMESPACE)' diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S index 9a5d875e21..b99c2b6520 100644 --- a/src/mesa/x86/glapi_x86.S +++ b/src/mesa/x86/glapi_x86.S @@ -27,7 +27,7 @@ */ #include "assyntax.h" -#include "glapioffsets.h" +#include "glapi/glapioffsets.h" #if defined(STDCALL_API) # if defined(USE_MGL_NAMESPACE) -- cgit v1.2.3 From bd4eec0561fb021849ac4047fdbf40a616fb68b3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 22 May 2008 20:33:17 +0900 Subject: pipebuffer: Don't retry allocating in slab suballocator. In pipebuffer, fencing is done at on a level above sub-allocation, so no matter how many times slab allocator retries no buffer will be freed. The pipebuffer fencing implemention already retries allocating. --- src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c | 23 +++++------------------ 1 file changed, 5 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c index b931455056..45ba158a4d 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c @@ -47,9 +47,6 @@ #include "pb_bufmgr.h" -#define DRI_SLABPOOL_ALLOC_RETRIES 100 - - struct pb_slab; struct pb_slab_buffer @@ -313,7 +310,6 @@ pb_slab_manager_create_buffer(struct pb_manager *_mgr, static struct pb_slab_buffer *buf; struct pb_slab *slab; struct list_head *list; - int count = DRI_SLABPOOL_ALLOC_RETRIES; /* check size */ assert(size == mgr->bufSize); @@ -331,23 +327,14 @@ pb_slab_manager_create_buffer(struct pb_manager *_mgr, /* XXX: check for compatible buffer usage too? */ _glthread_LOCK_MUTEX(mgr->mutex); - while (mgr->slabs.next == &mgr->slabs && count > 0) { - if (mgr->slabs.next != &mgr->slabs) - break; - - _glthread_UNLOCK_MUTEX(mgr->mutex); - if (count != DRI_SLABPOOL_ALLOC_RETRIES) - util_time_sleep(1); - _glthread_LOCK_MUTEX(mgr->mutex); + if (mgr->slabs.next == &mgr->slabs) { (void) pb_slab_create(mgr); - count--; + if (mgr->slabs.next == &mgr->slabs) { + _glthread_UNLOCK_MUTEX(mgr->mutex); + return NULL; + } } - list = mgr->slabs.next; - if (list == &mgr->slabs) { - _glthread_UNLOCK_MUTEX(mgr->mutex); - return NULL; - } slab = LIST_ENTRY(struct pb_slab, list, head); if (--slab->numFree == 0) LIST_DELINIT(list); -- cgit v1.2.3 From 781676c7cc5ae7586ee8edd07de880892c5a2d86 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 22 May 2008 21:54:41 +0900 Subject: pipebuffer: More robust face null pointers. It is really the caller responsibility not to call pipebuffer with null buffers, etc. But don't let the crash happen here, and still asserting early. --- src/gallium/auxiliary/pipebuffer/pb_buffer.h | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h index 49705cb862..857ea53c78 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h @@ -146,6 +146,8 @@ pb_map(struct pb_buffer *buf, unsigned flags) { assert(buf); + if(!buf) + return NULL; return buf->vtbl->map(buf, flags); } @@ -154,6 +156,8 @@ static INLINE void pb_unmap(struct pb_buffer *buf) { assert(buf); + if(!buf) + return; buf->vtbl->unmap(buf); } @@ -163,6 +167,12 @@ pb_get_base_buffer( struct pb_buffer *buf, struct pb_buffer **base_buf, unsigned *offset ) { + assert(buf); + if(!buf) { + base_buf = NULL; + offset = 0; + return; + } buf->vtbl->get_base_buffer(buf, base_buf, offset); } @@ -171,7 +181,8 @@ static INLINE void pb_destroy(struct pb_buffer *buf) { assert(buf); - assert(buf->vtbl); + if(!buf) + return; buf->vtbl->destroy(buf); } -- cgit v1.2.3 From 8b25b5256fad23e8ea11c6931ecac658ca60c0b0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 13 May 2008 09:46:53 +0100 Subject: draw: remove disabled non-sse swizzle code --- src/gallium/auxiliary/draw/draw_vs_sse.c | 50 +------------------------------- 1 file changed, 1 insertion(+), 49 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index e3f4e67472..edf235cddc 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -47,9 +47,7 @@ #include "tgsi/util/tgsi_parse.h" #define SSE_MAX_VERTICES 4 -#define SSE_SWIZZLES 1 -#if SSE_SWIZZLES typedef void (XSTDCALL *codegen_function) ( const struct tgsi_exec_vector *input, /* 1 */ struct tgsi_exec_vector *output, /* 2 */ @@ -62,14 +60,6 @@ typedef void (XSTDCALL *codegen_function) ( float (*aos_output)[4], /* 9 */ uint num_outputs, /* 10 */ uint output_stride ); /* 11 */ -#else -typedef void (XSTDCALL *codegen_function) ( - const struct tgsi_exec_vector *input, - struct tgsi_exec_vector *output, - float (*constant)[4], - struct tgsi_exec_vector *temporary, - float (*immediates)[4] ); -#endif struct draw_sse_vertex_shader { struct draw_vertex_shader base; @@ -111,7 +101,6 @@ vs_sse_run_linear( struct draw_vertex_shader *base, for (i = 0; i < count; i += MAX_TGSI_VERTICES) { unsigned int max_vertices = MIN2(MAX_TGSI_VERTICES, count - i); -#if SSE_SWIZZLES /* run compiled shader */ shader->func(machine->Inputs, @@ -128,43 +117,6 @@ vs_sse_run_linear( struct draw_vertex_shader *base, input = (const float (*)[4])((const char *)input + input_stride * max_vertices); output = (float (*)[4])((char *)output + output_stride * max_vertices); -#else - unsigned int j, slot; - - /* Swizzle inputs. - */ - for (j = 0; j < max_vertices; j++) { - for (slot = 0; slot < base->info.num_inputs; slot++) { - machine->Inputs[slot].xyzw[0].f[j] = input[slot][0]; - machine->Inputs[slot].xyzw[1].f[j] = input[slot][1]; - machine->Inputs[slot].xyzw[2].f[j] = input[slot][2]; - machine->Inputs[slot].xyzw[3].f[j] = input[slot][3]; - } - - input = (const float (*)[4])((const char *)input + input_stride); - } - - /* run compiled shader - */ - shader->func(machine->Inputs, - machine->Outputs, - (float (*)[4])constants, - machine->Temps, - shader->immediates); - - /* Unswizzle all output results. - */ - for (j = 0; j < max_vertices; j++) { - for (slot = 0; slot < base->info.num_outputs; slot++) { - output[slot][0] = machine->Outputs[slot].xyzw[0].f[j]; - output[slot][1] = machine->Outputs[slot].xyzw[1].f[j]; - output[slot][2] = machine->Outputs[slot].xyzw[2].f[j]; - output[slot][3] = machine->Outputs[slot].xyzw[3].f[j]; - } - - output = (float (*)[4])((char *)output + output_stride); - } -#endif } } @@ -211,7 +163,7 @@ draw_create_vs_sse(struct draw_context *draw, x86_init_func( &vs->sse2_program ); if (!tgsi_emit_sse2( (struct tgsi_token *) vs->base.state.tokens, - &vs->sse2_program, vs->immediates, SSE_SWIZZLES )) + &vs->sse2_program, vs->immediates, TRUE )) goto fail; vs->func = (codegen_function) x86_get_func( &vs->sse2_program ); -- cgit v1.2.3 From b23706454bb165a62888d264e95a98a2e4cf139c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 13 May 2008 13:35:14 +0100 Subject: draw: get rid of fetch-shade-emit frontend hack The code is now living in it's intended place as a pt middle end. --- src/gallium/auxiliary/draw/Makefile | 1 - src/gallium/auxiliary/draw/draw_private.h | 1 - src/gallium/auxiliary/draw/draw_pt.c | 14 - src/gallium/auxiliary/draw/draw_pt.h | 1 - .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 677 --------------------- 5 files changed, 694 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 67d78bdbbd..3053682da8 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -26,7 +26,6 @@ C_SOURCES = \ draw_pt_emit.c \ draw_pt_fetch.c \ draw_pt_fetch_emit.c \ - draw_pt_fetch_shade_emit.c \ draw_pt_middle_fse.c \ draw_pt_fetch_shade_pipeline.c \ draw_pt_post_vs.c \ diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 86b901a3c8..fd51a57781 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -131,7 +131,6 @@ struct draw_context struct { struct draw_pt_front_end *vcache; struct draw_pt_front_end *varray; - struct draw_pt_front_end *fetch_shade_emit; /* temp hack */ } front; struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 91e35db819..75f44d503e 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -85,11 +85,6 @@ draw_pt_arrays(struct draw_context *draw, */ if (draw->pt.user.elts || (opt & PT_PIPELINE)) { frontend = draw->pt.front.vcache; -#if 0 - } else if (opt == PT_SHADE && draw->pt.test_fse) { - /* should be a middle end.. */ - frontend = draw->pt.front.fetch_shade_emit; -#endif } else { frontend = draw->pt.front.varray; } @@ -127,10 +122,6 @@ boolean draw_pt_init( struct draw_context *draw ) draw->pt.middle.fetch_shade_emit = draw_pt_middle_fse( draw ); if (!draw->pt.middle.fetch_shade_emit) return FALSE; - - draw->pt.front.fetch_shade_emit = draw_pt_fetch_shade_emit( draw ); - if (!draw->pt.front.fetch_shade_emit) - return FALSE; } @@ -159,11 +150,6 @@ void draw_pt_destroy( struct draw_context *draw ) draw->pt.middle.fetch_shade_emit = NULL; } - if (draw->pt.front.fetch_shade_emit) { - draw->pt.front.fetch_shade_emit->destroy( draw->pt.front.fetch_shade_emit ); - draw->pt.front.fetch_shade_emit = NULL; - } - if (draw->pt.front.vcache) { draw->pt.front.vcache->destroy( draw->pt.front.vcache ); draw->pt.front.vcache = NULL; diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index cdae46b8d2..e03816ebbc 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -121,7 +121,6 @@ const void *draw_pt_elt_ptr( struct draw_context *draw, struct draw_pt_front_end *draw_pt_vcache( struct draw_context *draw ); struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw); -struct draw_pt_front_end *draw_pt_fetch_shade_emit( struct draw_context *draw ); /* Middle-ends: * diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c deleted file mode 100644 index f756d3e0bb..0000000000 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ /dev/null @@ -1,677 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "draw/draw_context.h" -#include "draw/draw_private.h" -#include "draw/draw_vbuf.h" -#include "draw/draw_vertex.h" -#include "draw/draw_pt.h" -#include "draw/draw_vs.h" - -#include "translate/translate.h" - -struct fetch_shade_emit; - -struct fse_shader { - struct translate_key key; - - void (*run_linear)( const struct fetch_shade_emit *fse, - unsigned start, - unsigned count, - char *buffer ); -}; - -/* Prototype fetch, shade, emit-hw-verts all in one go. - */ -struct fetch_shade_emit { - struct draw_pt_front_end base; - - struct draw_context *draw; - - struct translate_key key; - - /* Temporaries: - */ - const float *constants; - unsigned pitch[PIPE_MAX_ATTRIBS]; - const ubyte *src[PIPE_MAX_ATTRIBS]; - unsigned prim; - - /* Points to one of the three hardwired example shaders, below: - */ - struct fse_shader *active; - - /* Temporary: A list of hard-wired shaders. Of course the plan - * would be to generate these for a given (vertex-shader, - * translate-key) pair... - */ - struct fse_shader shader[10]; - int nr_shaders; -}; - - - -/* Not quite passthrough yet -- we're still running the 'shader' here, - * inlined into the vertex fetch function. - */ -static void fetch_xyz_rgb_st( const struct fetch_shade_emit *fse, - unsigned start, - unsigned count, - char *buffer ) -{ - unsigned i; - - const float *m = fse->constants; - const float m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12]; - const float m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13]; - const float m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14]; - const float m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15]; - - const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; - const ubyte *st = fse->src[2] + start * fse->pitch[2]; - - float *out = (float *)buffer; - - - assert(fse->pitch[1] == 0); - - /* loop over vertex attributes (vertex shader inputs) - */ - for (i = 0; i < count; i++) { - { - const float *in = (const float *)xyz; - const float ix = in[0], iy = in[1], iz = in[2]; - - out[0] = m0 * ix + m4 * iy + m8 * iz + m12; - out[1] = m1 * ix + m5 * iy + m9 * iz + m13; - out[2] = m2 * ix + m6 * iy + m10 * iz + m14; - out[3] = m3 * ix + m7 * iy + m11 * iz + m15; - xyz += fse->pitch[0]; - } - - { - out[4] = 1.0f; - out[5] = 1.0f; - out[6] = 1.0f; - out[7] = 1.0f; - } - - { - const float *in = (const float *)st; st += fse->pitch[2]; - out[8] = in[0]; - out[9] = in[1]; - out[10] = 0.0f; - out[11] = 1.0f; - } - - out += 12; - } -} - - - -static void fetch_xyz_rgb( const struct fetch_shade_emit *fse, - unsigned start, - unsigned count, - char *buffer ) -{ - unsigned i; - - const float *m = (const float *)fse->constants; - const float m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12]; - const float m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13]; - const float m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14]; - const float m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15]; - - const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; - const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; - - float *out = (float *)buffer; - -// debug_printf("rgb %f %f %f\n", rgb[0], rgb[1], rgb[2]); - - - for (i = 0; i < count; i++) { - { - const float *in = (const float *)xyz; - const float ix = in[0], iy = in[1], iz = in[2]; - - out[0] = m0 * ix + m4 * iy + m8 * iz + m12; - out[1] = m1 * ix + m5 * iy + m9 * iz + m13; - out[2] = m2 * ix + m6 * iy + m10 * iz + m14; - out[3] = m3 * ix + m7 * iy + m11 * iz + m15; - xyz += fse->pitch[0]; - } - - { - const float *in = (const float *)rgb; - out[4] = in[0]; - out[5] = in[1]; - out[6] = in[2]; - out[7] = 1.0f; - rgb += fse->pitch[1]; - } - - out += 8; - } -} - - - - -static void fetch_xyz_rgb_psiz( const struct fetch_shade_emit *fse, - unsigned start, - unsigned count, - char *buffer ) -{ - unsigned i; - - const float *m = (const float *)fse->constants; - const float m0 = m[0], m4 = m[4], m8 = m[8], m12 = m[12]; - const float m1 = m[1], m5 = m[5], m9 = m[9], m13 = m[13]; - const float m2 = m[2], m6 = m[6], m10 = m[10], m14 = m[14]; - const float m3 = m[3], m7 = m[7], m11 = m[11], m15 = m[15]; - - const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; - const float *rgb = (const float *)(fse->src[1] + start * fse->pitch[1]); - const float psiz = 1.0; - - float *out = (float *)buffer; - - - assert(fse->pitch[1] == 0); - - for (i = 0; i < count; i++) { - { - const float *in = (const float *)xyz; - const float ix = in[0], iy = in[1], iz = in[2]; - - out[0] = m0 * ix + m4 * iy + m8 * iz + m12; - out[1] = m1 * ix + m5 * iy + m9 * iz + m13; - out[2] = m2 * ix + m6 * iy + m10 * iz + m14; - out[3] = m3 * ix + m7 * iy + m11 * iz + m15; - xyz += fse->pitch[0]; - } - - { - out[4] = rgb[0]; - out[5] = rgb[1]; - out[6] = rgb[2]; - out[7] = 1.0f; - } - - { - out[8] = psiz; - } - - out += 9; - } -} - - - - -static boolean set_prim( struct fetch_shade_emit *fse, - unsigned prim, - unsigned count ) -{ - struct draw_context *draw = fse->draw; - - fse->prim = prim; - - switch (prim) { - case PIPE_PRIM_LINE_LOOP: - if (count > 1024) - return FALSE; - draw->render->set_primitive( draw->render, PIPE_PRIM_LINE_STRIP ); - break; - - case PIPE_PRIM_TRIANGLE_FAN: - case PIPE_PRIM_POLYGON: - if (count > 1024) - return FALSE; - draw->render->set_primitive( draw->render, prim ); - break; - - case PIPE_PRIM_QUADS: - case PIPE_PRIM_QUAD_STRIP: - draw->render->set_primitive( draw->render, PIPE_PRIM_TRIANGLES ); - break; - - default: - draw->render->set_primitive( draw->render, prim ); - break; - } - - return TRUE; -} - - - - - - -static void fse_prepare( struct draw_pt_front_end *fe, - unsigned prim, - struct draw_pt_middle_end *unused, - unsigned opt ) -{ - struct fetch_shade_emit *fse = (struct fetch_shade_emit *)fe; - struct draw_context *draw = fse->draw; - unsigned num_vs_inputs = draw->vertex_shader->info.num_inputs; - unsigned num_vs_outputs = draw->vertex_shader->info.num_outputs; - const struct vertex_info *vinfo; - unsigned i; - boolean need_psize = 0; - - - if (draw->pt.user.elts) { - assert(0); - return ; - } - - if (!set_prim(fse, prim, /*count*/1022 )) { - assert(0); - return ; - } - - /* Must do this after set_primitive() above: - */ - vinfo = draw->render->get_vertex_info(draw->render); - - - - fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ - num_vs_inputs); /* inputs - fetch from api format */ - - fse->key.output_stride = vinfo->size * 4; - memset(fse->key.element, 0, - fse->key.nr_elements * sizeof(fse->key.element[0])); - - for (i = 0; i < num_vs_inputs; i++) { - const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; - fse->key.element[i].input_format = src->src_format; - - /* Consider ignoring these at this point, ie make generated - * programs independent of this state: - */ - fse->key.element[i].input_buffer = 0; //src->vertex_buffer_index; - fse->key.element[i].input_offset = 0; //src->src_offset; - } - - - { - unsigned dst_offset = 0; - - for (i = 0; i < vinfo->num_attribs; i++) { - unsigned emit_sz = 0; - unsigned output_format = PIPE_FORMAT_NONE; - unsigned vs_output = vinfo->src_index[i]; - - switch (vinfo->emit[i]) { - case EMIT_4F: - output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - emit_sz = 4 * sizeof(float); - break; - case EMIT_3F: - output_format = PIPE_FORMAT_R32G32B32_FLOAT; - emit_sz = 3 * sizeof(float); - break; - case EMIT_2F: - output_format = PIPE_FORMAT_R32G32_FLOAT; - emit_sz = 2 * sizeof(float); - break; - case EMIT_1F: - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - break; - case EMIT_1F_PSIZE: - need_psize = 1; - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - vs_output = num_vs_outputs + 1; - - break; - default: - assert(0); - break; - } - - /* The elements in the key correspond to vertex shader output - * numbers, not to positions in the hw vertex description -- - * that's handled by the output_offset field. - */ - fse->key.element[vs_output].output_format = output_format; - fse->key.element[vs_output].output_offset = dst_offset; - - dst_offset += emit_sz; - assert(fse->key.output_stride >= dst_offset); - } - } - - /* To make psize work, really need to tell the vertex shader to - * copy that value from input->output. For 'translate' this was - * implicit for all elements. - */ -#if 0 - if (need_psize) { - unsigned input = num_vs_inputs + 1; - const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; - fse->key.element[i].input_format = PIPE_FORMAT_R32_FLOAT; - fse->key.element[i].input_buffer = 0; //nr_buffers + 1; - fse->key.element[i].input_offset = 0; - - fse->key.nr_elements += 1; - - } -#endif - - fse->constants = draw->pt.user.constants; - - /* Would normally look up a vertex shader and peruse its list of - * varients somehow. We omitted that step and put all the - * hardcoded "shaders" into an array. We're just making the - * assumption that this happens to be a matching shader... ie - * you're running isosurf, aren't you? - */ - fse->active = NULL; - for (i = 0; i < fse->nr_shaders; i++) { - if (translate_key_compare( &fse->key, &fse->shader[i].key) == 0) - fse->active = &fse->shader[i]; - } - - if (!fse->active) { - assert(0); - return ; - } - - /* Now set buffer pointers: - */ - for (i = 0; i < num_vs_inputs; i++) { - unsigned buf = draw->pt.vertex_element[i].vertex_buffer_index; - - fse->src[i] = ((const ubyte *) draw->pt.user.vbuffer[buf] + - draw->pt.vertex_buffer[buf].buffer_offset + - draw->pt.vertex_element[i].src_offset); - - fse->pitch[i] = draw->pt.vertex_buffer[buf].pitch; - - } - - - //return TRUE; -} - - - - - - -#define INDEX(i) (start + (i)) -static void fse_render_linear( struct vbuf_render *render, - unsigned prim, - unsigned start, - unsigned length ) -{ - ushort *tmp = NULL; - unsigned i, j; - - switch (prim) { - case PIPE_PRIM_LINE_LOOP: - tmp = MALLOC( sizeof(ushort) * (length + 1) ); - - for (i = 0; i < length; i++) - tmp[i] = INDEX(i); - tmp[length] = 0; - - render->draw( render, - tmp, - length+1 ); - break; - - - case PIPE_PRIM_QUAD_STRIP: - tmp = MALLOC( sizeof(ushort) * (length / 2 * 6) ); - - for (j = i = 0; i + 3 < length; i += 2, j += 6) { - tmp[j+0] = INDEX(i+0); - tmp[j+1] = INDEX(i+1); - tmp[j+2] = INDEX(i+3); - - tmp[j+3] = INDEX(i+2); - tmp[j+4] = INDEX(i+0); - tmp[j+5] = INDEX(i+3); - } - - if (j) - render->draw( render, tmp, j ); - break; - - case PIPE_PRIM_QUADS: - tmp = MALLOC( sizeof(int) * (length / 4 * 6) ); - - for (j = i = 0; i + 3 < length; i += 4, j += 6) { - tmp[j+0] = INDEX(i+0); - tmp[j+1] = INDEX(i+1); - tmp[j+2] = INDEX(i+3); - - tmp[j+3] = INDEX(i+1); - tmp[j+4] = INDEX(i+2); - tmp[j+5] = INDEX(i+3); - } - - if (j) - render->draw( render, tmp, j ); - break; - - default: - render->draw_arrays( render, - start, - length ); - break; - } - - if (tmp) - FREE(tmp); -} - - - -static boolean do_draw( struct fetch_shade_emit *fse, - unsigned start, unsigned count ) -{ - struct draw_context *draw = fse->draw; - - char *hw_verts = - draw->render->allocate_vertices( draw->render, - (ushort)fse->key.output_stride, - (ushort)count ); - - if (!hw_verts) - return FALSE; - - /* Single routine to fetch vertices, run shader and emit HW verts. - * Clipping and viewport transformation are done on hardware. - */ - fse->active->run_linear( fse, - start, count, - hw_verts ); - - /* Draw arrays path to avoid re-emitting index list again and - * again. - */ - fse_render_linear( draw->render, - fse->prim, - 0, - count ); - - - draw->render->release_vertices( draw->render, - hw_verts, - fse->key.output_stride, - count ); - - return TRUE; -} - - -static void -fse_run(struct draw_pt_front_end *fe, - pt_elt_func elt_func, - const void *elt_ptr, - unsigned count) -{ - struct fetch_shade_emit *fse = (struct fetch_shade_emit *)fe; - unsigned i = 0; - unsigned first, incr; - unsigned start = elt_func(elt_ptr, 0); - - //debug_printf("%s prim %d start %d count %d\n", __FUNCTION__, prim, start, count); - - draw_pt_split_prim(fse->prim, &first, &incr); - - count -= (count - first) % incr; - - while (i + first <= count) { - int nr = MIN2( count - i, 1024 ); - - /* snap to prim boundary - */ - nr -= (nr - first) % incr; - - if (!do_draw( fse, start + i, nr )) { - assert(0); - return ; - } - - /* increment allowing for repeated vertices - */ - i += nr - (first - incr); - } - - //return TRUE; -} - - -static void fse_finish( struct draw_pt_front_end *frontend ) -{ -} - - -static void -fse_destroy( struct draw_pt_front_end *frontend ) -{ - FREE(frontend); -} - -struct draw_pt_front_end *draw_pt_fetch_shade_emit( struct draw_context *draw ) -{ - struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit); - if (!fse) - return NULL; - - fse->base.prepare = fse_prepare; - fse->base.run = fse_run; - fse->base.finish = fse_finish; - fse->base.destroy = fse_destroy; - fse->draw = draw; - - fse->shader[0].run_linear = fetch_xyz_rgb_st; - fse->shader[0].key.nr_elements = 3; - fse->shader[0].key.output_stride = 12 * sizeof(float); - - fse->shader[0].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[0].key.element[0].input_buffer = 0; - fse->shader[0].key.element[0].input_offset = 0; - fse->shader[0].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[0].key.element[0].output_offset = 0; - - fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[0].key.element[1].input_buffer = 0; - fse->shader[0].key.element[1].input_offset = 0; - fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[0].key.element[1].output_offset = 16; - - fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32_FLOAT; - fse->shader[0].key.element[1].input_buffer = 0; - fse->shader[0].key.element[1].input_offset = 0; - fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[0].key.element[1].output_offset = 32; - - fse->shader[1].run_linear = fetch_xyz_rgb; - fse->shader[1].key.nr_elements = 2; - fse->shader[1].key.output_stride = 8 * sizeof(float); - - fse->shader[1].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[1].key.element[0].input_buffer = 0; - fse->shader[1].key.element[0].input_offset = 0; - fse->shader[1].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[1].key.element[0].output_offset = 0; - - fse->shader[1].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[1].key.element[1].input_buffer = 0; - fse->shader[1].key.element[1].input_offset = 0; - fse->shader[1].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[1].key.element[1].output_offset = 16; - - fse->shader[2].run_linear = fetch_xyz_rgb_psiz; - fse->shader[2].key.nr_elements = 3; - fse->shader[2].key.output_stride = 9 * sizeof(float); - - fse->shader[2].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[2].key.element[0].input_buffer = 0; - fse->shader[2].key.element[0].input_offset = 0; - fse->shader[2].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[2].key.element[0].output_offset = 0; - - fse->shader[2].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[2].key.element[1].input_buffer = 0; - fse->shader[2].key.element[1].input_offset = 0; - fse->shader[2].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[2].key.element[1].output_offset = 16; - - /* psize is special - * -- effectively add it here as another input!?! - * -- who knows how to add it as a buffer? - */ - fse->shader[2].key.element[2].input_format = PIPE_FORMAT_R32_FLOAT; - fse->shader[2].key.element[2].input_buffer = 0; - fse->shader[2].key.element[2].input_offset = 0; - fse->shader[2].key.element[2].output_format = PIPE_FORMAT_R32_FLOAT; - fse->shader[2].key.element[2].output_offset = 32; - - fse->nr_shaders = 3; - - return &fse->base; -} -- cgit v1.2.3 From 2f0d1396e4c1626b3b1ac799bd29e86a9530369e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 13 May 2008 13:40:22 +0100 Subject: draw: move some state into a new 'vs' area --- src/gallium/auxiliary/draw/draw_context.c | 21 ++++--------- src/gallium/auxiliary/draw/draw_pipe.h | 2 +- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 4 +-- src/gallium/auxiliary/draw/draw_pipe_clip.c | 6 ++-- src/gallium/auxiliary/draw/draw_pipe_flatshade.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_stipple.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_twoside.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_wide_point.c | 4 +-- src/gallium/auxiliary/draw/draw_private.h | 28 +++++++++++------ .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 6 ++-- src/gallium/auxiliary/draw/draw_pt_middle_fse.c | 4 +-- src/gallium/auxiliary/draw/draw_vs.c | 35 +++++++++++++++++++--- src/gallium/auxiliary/draw/draw_vs_exec.c | 2 +- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- 15 files changed, 75 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 98e23fa830..2242074965 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -56,12 +56,6 @@ struct draw_context *draw_create( void ) draw->reduced_prim = ~0; /* != any of PIPE_PRIM_x */ - tgsi_exec_machine_init(&draw->machine); - - /* FIXME: give this machine thing a proper constructor: - */ - draw->machine.Inputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); - draw->machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); if (!draw_pipeline_init( draw )) goto fail; @@ -69,6 +63,9 @@ struct draw_context *draw_create( void ) if (!draw_pt_init( draw )) goto fail; + if (!draw_vs_init( draw )) + goto fail; + return draw; fail: @@ -83,13 +80,6 @@ void draw_destroy( struct draw_context *draw ) return; - if (draw->machine.Inputs) - align_free(draw->machine.Inputs); - - if (draw->machine.Outputs) - align_free(draw->machine.Outputs); - - tgsi_exec_machine_free_data(&draw->machine); /* Not so fast -- we're just borrowing this at the moment. * @@ -99,6 +89,7 @@ void draw_destroy( struct draw_context *draw ) draw_pipeline_destroy( draw ); draw_pt_destroy( draw ); + draw_vs_destroy( draw ); FREE( draw ); } @@ -295,7 +286,7 @@ int draw_find_vs_output(struct draw_context *draw, uint semantic_name, uint semantic_index) { - const struct draw_vertex_shader *vs = draw->vertex_shader; + const struct draw_vertex_shader *vs = draw->vs.vertex_shader; uint i; for (i = 0; i < vs->info.num_outputs; i++) { if (vs->info.output_semantic_name[i] == semantic_name && @@ -320,7 +311,7 @@ draw_find_vs_output(struct draw_context *draw, uint draw_num_vs_outputs(struct draw_context *draw) { - uint count = draw->vertex_shader->info.num_outputs; + uint count = draw->vs.vertex_shader->info.num_outputs; if (draw->extra_vp_outputs.slot > 0) count++; return count; diff --git a/src/gallium/auxiliary/draw/draw_pipe.h b/src/gallium/auxiliary/draw/draw_pipe.h index f1cb0891ca..dbad8f98ac 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.h +++ b/src/gallium/auxiliary/draw/draw_pipe.h @@ -116,7 +116,7 @@ dup_vert( struct draw_stage *stage, { struct vertex_header *tmp = stage->tmp[idx]; const uint vsize = sizeof(struct vertex_header) - + stage->draw->num_vs_outputs * 4 * sizeof(float); + + stage->draw->vs.num_vs_outputs * 4 * sizeof(float); memcpy(tmp, vert, vsize); tmp->vertex_id = UNDEFINED_VERTEX_ID; return tmp; diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index f501b2aed4..d93708ad3c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -651,7 +651,7 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) } /* update vertex attrib info */ - aaline->tex_slot = draw->num_vs_outputs; + aaline->tex_slot = draw->vs.num_vs_outputs; assert(aaline->tex_slot > 0); /* output[0] is vertex pos */ /* advertise the extra post-transformed vertex attribute */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 122a48660a..97d74ad693 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -681,7 +681,7 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header) bind_aapoint_fragment_shader(aapoint); /* update vertex attrib info */ - aapoint->tex_slot = draw->num_vs_outputs; + aapoint->tex_slot = draw->vs.num_vs_outputs; assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */ draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; @@ -692,7 +692,7 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header) aapoint->psize_slot = -1; if (draw->rasterizer->point_size_per_vertex) { /* find PSIZ vertex output */ - const struct draw_vertex_shader *vs = draw->vertex_shader; + const struct draw_vertex_shader *vs = draw->vs.vertex_shader; uint i; for (i = 0; i < vs->info.num_outputs; i++) { if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c index ce80c94163..c11ed934a4 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_clip.c +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -112,7 +112,7 @@ static void interp( const struct clipper *clip, const struct vertex_header *out, const struct vertex_header *in ) { - const unsigned nr_attrs = clip->stage.draw->num_vs_outputs; + const unsigned nr_attrs = clip->stage.draw->vs.num_vs_outputs; unsigned j; /* Vertex header. @@ -180,7 +180,7 @@ static void emit_poly( struct draw_stage *stage, header.flags |= edge_last; if (0) { - const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + const struct draw_vertex_shader *vs = stage->draw->vs.vertex_shader; uint j, k; debug_printf("Clipped tri:\n"); for (j = 0; j < 3; j++) { @@ -425,7 +425,7 @@ clip_init_state( struct draw_stage *stage ) clipper->flat = stage->draw->rasterizer->flatshade ? TRUE : FALSE; if (clipper->flat) { - const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + const struct draw_vertex_shader *vs = stage->draw->vs.vertex_shader; uint i; clipper->num_color_attribs = 0; diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c index 09b68c4559..21a9c3b77f 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c @@ -159,7 +159,7 @@ static void flatshade_line_1( struct draw_stage *stage, static void flatshade_init_state( struct draw_stage *stage ) { struct flat_stage *flat = flat_stage(stage); - const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + const struct draw_vertex_shader *vs = stage->draw->vs.vertex_shader; uint i; /* Find which vertex shader outputs are colors, make a list */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c index 3cbced362e..4673d5dcba 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c @@ -71,7 +71,7 @@ screen_interp( struct draw_context *draw, const struct vertex_header *v1 ) { uint attr; - for (attr = 0; attr < draw->num_vs_outputs; attr++) { + for (attr = 0; attr < draw->vs.num_vs_outputs; attr++) { const float *val0 = v0->data[attr]; const float *val1 = v1->data[attr]; float *newv = dst->data[attr]; diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c index 50872fdbe9..3ac825f565 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_twoside.c +++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c @@ -105,7 +105,7 @@ static void twoside_first_tri( struct draw_stage *stage, struct prim_header *header ) { struct twoside_stage *twoside = twoside_stage(stage); - const struct draw_vertex_shader *vs = stage->draw->vertex_shader; + const struct draw_vertex_shader *vs = stage->draw->vs.vertex_shader; uint i; twoside->attrib_front0 = 0; diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index ed08573382..df92e3f2d0 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -197,7 +197,7 @@ static void widepoint_first_point( struct draw_stage *stage, if (draw->rasterizer->point_sprite) { /* find vertex shader texcoord outputs */ - const struct draw_vertex_shader *vs = draw->vertex_shader; + const struct draw_vertex_shader *vs = draw->vs.vertex_shader; uint i, j = 0; for (i = 0; i < vs->info.num_outputs; i++) { if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_GENERIC) { @@ -212,7 +212,7 @@ static void widepoint_first_point( struct draw_stage *stage, wide->psize_slot = -1; if (draw->rasterizer->point_size_per_vertex) { /* find PSIZ vertex output */ - const struct draw_vertex_shader *vs = draw->vertex_shader; + const struct draw_vertex_shader *vs = draw->vs.vertex_shader; uint i; for (i = 0; i < vs->info.num_outputs; i++) { if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_PSIZE) { diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index fd51a57781..3418ee2b88 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -169,13 +169,24 @@ struct draw_context /* pipe state that we need: */ const struct pipe_rasterizer_state *rasterizer; struct pipe_viewport_state viewport; + boolean identity_viewport; - struct draw_vertex_shader *vertex_shader; + struct { + struct draw_vertex_shader *vertex_shader; + uint num_vs_outputs; /**< convenience, from vertex_shader */ - boolean identity_viewport; - uint num_vs_outputs; /**< convenience, from vertex_shader */ + /** TGSI program interpreter runtime state */ + struct tgsi_exec_machine machine; + + /* This (and the tgsi_exec_machine struct) probably need to be moved somewhere private. + */ + struct gallivm_cpu_engine *engine; + + struct translate_cache *fetch_cache; + struct translate_cache *emit_cache; + } vs; /* Clip derived state: */ @@ -192,16 +203,15 @@ struct draw_context unsigned reduced_prim; - /** TGSI program interpreter runtime state */ - struct tgsi_exec_machine machine; - - /* This (and the tgsi_exec_machine struct) probably need to be moved somewhere private. - */ - struct gallivm_cpu_engine *engine; void *driver_private; }; +/******************************************************************************* + * Vertex shader code: + */ +boolean draw_vs_init( struct draw_context *draw ); +void draw_vs_destroy( struct draw_context *draw ); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index dad54690a5..06718779a5 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -55,7 +55,7 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; - struct draw_vertex_shader *vs = draw->vertex_shader; + struct draw_vertex_shader *vs = draw->vs.vertex_shader; /* Add one to num_outputs because the pipeline occasionally tags on * an additional texcoord, eg for AA lines. @@ -107,7 +107,7 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; - struct draw_vertex_shader *shader = draw->vertex_shader; + struct draw_vertex_shader *shader = draw->vs.vertex_shader; unsigned opt = fpme->opt; unsigned alloc_count = align_int( fetch_count, 4 ); @@ -183,7 +183,7 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; - struct draw_vertex_shader *shader = draw->vertex_shader; + struct draw_vertex_shader *shader = draw->vs.vertex_shader; unsigned opt = fpme->opt; unsigned alloc_count = align_int( count, 4 ); diff --git a/src/gallium/auxiliary/draw/draw_pt_middle_fse.c b/src/gallium/auxiliary/draw/draw_pt_middle_fse.c index cdb7d260da..643ea151c1 100644 --- a/src/gallium/auxiliary/draw/draw_pt_middle_fse.c +++ b/src/gallium/auxiliary/draw/draw_pt_middle_fse.c @@ -368,8 +368,8 @@ static void fse_prepare( struct draw_pt_middle_end *middle, { struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; struct draw_context *draw = fse->draw; - unsigned num_vs_inputs = draw->vertex_shader->info.num_inputs; - unsigned num_vs_outputs = draw->vertex_shader->info.num_outputs; + unsigned num_vs_inputs = draw->vs.vertex_shader->info.num_inputs; + unsigned num_vs_outputs = draw->vs.vertex_shader->info.num_outputs; const struct vertex_info *vinfo; unsigned i; boolean need_psize = 0; diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index 03fe00a951..4142dd9589 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -66,13 +66,13 @@ draw_bind_vertex_shader(struct draw_context *draw, if (dvs) { - draw->vertex_shader = dvs; - draw->num_vs_outputs = dvs->info.num_outputs; + draw->vs.vertex_shader = dvs; + draw->vs.num_vs_outputs = dvs->info.num_outputs; dvs->prepare( dvs, draw ); } else { - draw->vertex_shader = NULL; - draw->num_vs_outputs = 0; + draw->vs.vertex_shader = NULL; + draw->vs.num_vs_outputs = 0; } } @@ -83,3 +83,30 @@ draw_delete_vertex_shader(struct draw_context *draw, { dvs->delete( dvs ); } + + + +boolean +draw_vs_init( struct draw_context *draw ) +{ + tgsi_exec_machine_init(&draw->vs.machine); + /* FIXME: give this machine thing a proper constructor: + */ + draw->vs.machine.Inputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); + draw->vs.machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); + + return TRUE; +} + +void +draw_vs_destroy( struct draw_context *draw ) +{ + if (draw->vs.machine.Inputs) + align_free(draw->vs.machine.Inputs); + + if (draw->vs.machine.Outputs) + align_free(draw->vs.machine.Outputs); + + tgsi_exec_machine_free_data(&draw->vs.machine); + +} diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 7a02f6334b..cb80d008cd 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -182,7 +182,7 @@ draw_create_vs_exec(struct draw_context *draw, vs->base.prepare = vs_exec_prepare; vs->base.run_linear = vs_exec_run_linear; vs->base.delete = vs_exec_delete; - vs->machine = &draw->machine; + vs->machine = &draw->vs.machine; return &vs->base; } diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index edf235cddc..13ad032bd3 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -158,7 +158,7 @@ draw_create_vs_sse(struct draw_context *draw, vs->base.prepare = vs_sse_prepare; vs->base.run_linear = vs_sse_run_linear; vs->base.delete = vs_sse_delete; - vs->machine = &draw->machine; + vs->machine = &draw->vs.machine; x86_init_func( &vs->sse2_program ); -- cgit v1.2.3 From 7c99d7fe60e7bb0b7cf103a851aeef4614278ca6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 15 May 2008 12:39:08 +0100 Subject: draw: create specialized vs varients incorporating fetch & emit --- src/gallium/auxiliary/draw/Makefile | 3 +- src/gallium/auxiliary/draw/draw_private.h | 2 + .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 338 +++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs.c | 83 ++++- src/gallium/auxiliary/draw/draw_vs.h | 105 +++++++ src/gallium/auxiliary/draw/draw_vs_exec.c | 2 + src/gallium/auxiliary/draw/draw_vs_llvm.c | 2 + src/gallium/auxiliary/draw/draw_vs_sse.c | 2 + src/gallium/auxiliary/draw/draw_vs_varient.c | 229 ++++++++++++++ 9 files changed, 764 insertions(+), 2 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c create mode 100644 src/gallium/auxiliary/draw/draw_vs_varient.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 3053682da8..84877994fb 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -26,7 +26,7 @@ C_SOURCES = \ draw_pt_emit.c \ draw_pt_fetch.c \ draw_pt_fetch_emit.c \ - draw_pt_middle_fse.c \ + draw_pt_fetch_shade_emit.c \ draw_pt_fetch_shade_pipeline.c \ draw_pt_post_vs.c \ draw_pt_util.c \ @@ -34,6 +34,7 @@ C_SOURCES = \ draw_pt_vcache.c \ draw_vertex.c \ draw_vs.c \ + draw_vs_varient.c \ draw_vs_exec.c \ draw_vs_llvm.c \ draw_vs_sse.c diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 3418ee2b88..c095bf3d7b 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -184,7 +184,9 @@ struct draw_context struct gallivm_cpu_engine *engine; + struct translate *fetch; struct translate_cache *fetch_cache; + struct translate *emit; struct translate_cache *emit_cache; } vs; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c new file mode 100644 index 0000000000..74945dcfe9 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -0,0 +1,338 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_pt.h" +#include "draw/draw_vs.h" + +#include "translate/translate.h" + +struct fetch_shade_emit; + + +/* Prototype fetch, shade, emit-hw-verts all in one go. + */ +struct fetch_shade_emit { + struct draw_pt_middle_end base; + struct draw_context *draw; + + + /* Temporaries: + */ + const float *constants; + unsigned pitch[PIPE_MAX_ATTRIBS]; + const ubyte *src[PIPE_MAX_ATTRIBS]; + unsigned prim; + + struct draw_vs_varient_key key; + struct draw_vs_varient *active; +}; + + + + +static void fse_prepare( struct draw_pt_middle_end *middle, + unsigned prim, + unsigned opt ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + unsigned num_vs_inputs = draw->vs.vertex_shader->info.num_inputs; + unsigned num_vs_outputs = draw->vs.vertex_shader->info.num_outputs; + const struct vertex_info *vinfo; + unsigned i; + boolean need_psize = 0; + + + if (draw->pt.user.elts) { + assert(0); + return ; + } + + if (!draw->render->set_primitive( draw->render, + prim )) { + assert(0); + return; + } + + /* Must do this after set_primitive() above: + */ + vinfo = draw->render->get_vertex_info(draw->render); + + + + fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ + num_vs_inputs); /* inputs - fetch from api format */ + + fse->key.output_stride = vinfo->size * 4; + memset(fse->key.element, 0, + fse->key.nr_elements * sizeof(fse->key.element[0])); + + for (i = 0; i < num_vs_inputs; i++) { + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].in.format = src->src_format; + + /* Consider ignoring these, ie make generated programs + * independent of this state: + */ + fse->key.element[i].in.buffer = src->vertex_buffer_index; + fse->key.element[i].in.offset = src->src_offset; + } + + + { + unsigned dst_offset = 0; + + for (i = 0; i < vinfo->num_attribs; i++) { + unsigned emit_sz = 0; + unsigned output_format = PIPE_FORMAT_NONE; + unsigned vs_output = vinfo->src_index[i]; + + switch (vinfo->emit[i]) { + case EMIT_4F: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit_sz = 4 * sizeof(float); + break; + case EMIT_3F: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + emit_sz = 3 * sizeof(float); + break; + case EMIT_2F: + output_format = PIPE_FORMAT_R32G32_FLOAT; + emit_sz = 2 * sizeof(float); + break; + case EMIT_1F: + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + break; + case EMIT_1F_PSIZE: + need_psize = 1; + output_format = PIPE_FORMAT_R32_FLOAT; + emit_sz = 1 * sizeof(float); + vs_output = num_vs_outputs + 1; + + break; + default: + assert(0); + break; + } + + /* The elements in the key correspond to vertex shader output + * numbers, not to positions in the hw vertex description -- + * that's handled by the output_offset field. + */ + fse->key.element[vs_output].out.format = output_format; + fse->key.element[vs_output].out.offset = dst_offset; + + dst_offset += emit_sz; + assert(fse->key.output_stride >= dst_offset); + } + } + + /* To make psize work, really need to tell the vertex shader to + * copy that value from input->output. For 'translate' this was + * implicit for all elements. + */ +#if 0 + if (need_psize) { + unsigned input = num_vs_inputs + 1; + const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; + fse->key.element[i].input_format = PIPE_FORMAT_R32_FLOAT; + fse->key.element[i].input_buffer = 0; //nr_buffers + 1; + fse->key.element[i].input_offset = 0; + + fse->key.nr_elements += 1; + + } +#endif + + /* Would normally look up a vertex shader and peruse its list of + * varients somehow. We omitted that step and put all the + * hardcoded "shaders" into an array. We're just making the + * assumption that this happens to be a matching shader... ie + * you're running isosurf, aren't you? + */ + fse->active = draw_vs_lookup_varient( draw->vs.vertex_shader, + &fse->key ); + + if (!fse->active) { + assert(0); + return ; + } + + /* Now set buffer pointers: + */ + for (i = 0; i < num_vs_inputs; i++) { + unsigned buf = draw->pt.vertex_element[i].vertex_buffer_index; + + fse->active->set_input( fse->active, + i, + + ((const ubyte *) draw->pt.user.vbuffer[buf] + + draw->pt.vertex_buffer[buf].buffer_offset), + + draw->pt.vertex_buffer[buf].pitch ); + } + + fse->active->set_constants( fse->active, + (const float (*)[4])draw->pt.user.constants ); + + //return TRUE; +} + + + + + + + +static void fse_run_linear( struct draw_pt_middle_end *middle, + unsigned start, + unsigned count ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + unsigned alloc_count = align(count, 4); + char *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)alloc_count ); + + if (!hw_verts) { + assert(0); + return; + } + + /* Single routine to fetch vertices, run shader and emit HW verts. + * Clipping and viewport transformation are done elsewhere -- + * either by the API or on hardware, or for some other reason not + * required... + */ + fse->active->run_linear( fse->active, + start, count, + hw_verts ); + + /* Draw arrays path to avoid re-emitting index list again and + * again. + */ + draw->render->draw_arrays( draw->render, + 0, + count ); + + + draw->render->release_vertices( draw->render, + hw_verts, + fse->key.output_stride, + count ); +} + + +static void +fse_run(struct draw_pt_middle_end *middle, + const unsigned *fetch_elts, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + unsigned alloc_count = align(fetch_count, 4); + void *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)alloc_count ); + if (!hw_verts) { + assert(0); + return; + } + + + /* Single routine to fetch vertices, run shader and emit HW verts. + */ + fse->active->run_elts( fse->active, + fetch_elts, + fetch_count, + hw_verts ); + + draw->render->draw( draw->render, + draw_elts, + draw_count ); + + draw->render->release_vertices( draw->render, + hw_verts, + fse->key.output_stride, + fetch_count ); + +} + + +static void fse_finish( struct draw_pt_middle_end *middle ) +{ +} + + +static void +fse_destroy( struct draw_pt_middle_end *middle ) +{ + FREE(middle); +} + +struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw ) +{ + struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit); + if (!fse) + return NULL; + + fse->base.prepare = fse_prepare; + fse->base.run = fse_run; + fse->base.run_linear = fse_run_linear; + fse->base.finish = fse_finish; + fse->base.destroy = fse_destroy; + fse->draw = draw; + + return &fse->base; +} diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index 4142dd9589..9b899d404e 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -36,6 +36,8 @@ #include "draw_private.h" #include "draw_context.h" #include "draw_vs.h" +#include "translate/translate.h" +#include "translate/translate_cache.h" @@ -90,11 +92,25 @@ boolean draw_vs_init( struct draw_context *draw ) { tgsi_exec_machine_init(&draw->vs.machine); + /* FIXME: give this machine thing a proper constructor: */ draw->vs.machine.Inputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); - draw->vs.machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); + if (!draw->vs.machine.Inputs) + return FALSE; + draw->vs.machine.Outputs = align_malloc(PIPE_MAX_ATTRIBS * sizeof(struct tgsi_exec_vector), 16); + if (!draw->vs.machine.Outputs) + return FALSE; + + draw->vs.emit_cache = translate_cache_create(); + if (!draw->vs.emit_cache) + return FALSE; + + draw->vs.fetch_cache = translate_cache_create(); + if (!draw->vs.fetch_cache) + return FALSE; + return TRUE; } @@ -107,6 +123,71 @@ draw_vs_destroy( struct draw_context *draw ) if (draw->vs.machine.Outputs) align_free(draw->vs.machine.Outputs); + if (draw->vs.fetch_cache) + translate_cache_destroy(draw->vs.fetch_cache); + + if (draw->vs.emit_cache) + translate_cache_destroy(draw->vs.emit_cache); + tgsi_exec_machine_free_data(&draw->vs.machine); } + + +struct draw_vs_varient * +draw_vs_lookup_varient( struct draw_vertex_shader *vs, + const struct draw_vs_varient_key *key ) +{ + struct draw_vs_varient *varient; + unsigned i; + + /* Lookup existing varient: + */ + for (i = 0; i < vs->nr_varients; i++) + if (draw_vs_varient_key_compare(key, &vs->varient[i]->key) == 0) + return vs->varient[i]; + + /* Else have to create a new one: + */ + varient = vs->create_varient( vs, key ); + if (varient == NULL) + return NULL; + + /* Add it to our list: + */ + assert(vs->nr_varients < Elements(vs->varient)); + vs->varient[vs->nr_varients++] = varient; + + /* Done + */ + return varient; +} + + +struct translate * +draw_vs_get_fetch( struct draw_context *draw, + struct translate_key *key ) +{ + if (!draw->vs.fetch || + translate_key_compare(&draw->vs.fetch->key, key) != 0) + { + translate_key_sanitize(key); + draw->vs.fetch = translate_cache_find(draw->vs.fetch_cache, key); + } + + return draw->vs.fetch; +} + +struct translate * +draw_vs_get_emit( struct draw_context *draw, + struct translate_key *key ) +{ + if (!draw->vs.emit || + translate_key_compare(&draw->vs.emit->key, key) != 0) + { + translate_key_sanitize(key); + draw->vs.emit = translate_cache_find(draw->vs.emit_cache, key); + } + + return draw->vs.emit; +} diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index f9772b83b8..677be0d28d 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -38,10 +38,63 @@ struct draw_context; struct pipe_shader_state; +struct draw_vs_input +{ + enum pipe_format format; + unsigned buffer; + unsigned offset; +}; + +struct draw_vs_output +{ + enum pipe_format format; + unsigned offset; +}; + +struct draw_vs_element { + struct draw_vs_input in; + struct draw_vs_output out; +}; + +struct draw_vs_varient_key { + unsigned output_stride; + unsigned nr_elements; + struct draw_vs_element element[PIPE_MAX_ATTRIBS]; +}; + +struct draw_vs_varient { + struct draw_vs_varient_key key; + + struct draw_vertex_shader *vs; + + void (*set_input)( struct draw_vs_varient *, + unsigned i, + const void *ptr, + unsigned stride ); + + void (*set_constants)( struct draw_vs_varient *, + const float (*constants)[4] ); + + + void (*run_linear)( struct draw_vs_varient *shader, + unsigned start, + unsigned count, + void *output_buffer ); + + void (*run_elts)( struct draw_vs_varient *shader, + const unsigned *elts, + unsigned count, + void *output_buffer ); + + void (*destroy)( struct draw_vs_varient * ); +}; + + /** * Private version of the compiled vertex_shader */ struct draw_vertex_shader { + struct draw_context *draw; /* This member will disappear shortly: */ @@ -49,6 +102,14 @@ struct draw_vertex_shader { struct tgsi_shader_info info; + /* + */ + struct draw_vs_varient *varient[16]; + unsigned nr_varients; + struct draw_vs_varient *(*create_varient)( struct draw_vertex_shader *shader, + const struct draw_vs_varient_key *key ); + + void (*prepare)( struct draw_vertex_shader *shader, struct draw_context *draw ); @@ -68,6 +129,15 @@ struct draw_vertex_shader { }; +struct draw_vs_varient * +draw_vs_lookup_varient( struct draw_vertex_shader *base, + const struct draw_vs_varient_key *key ); + + +/******************************************************************************** + * Internal functions: + */ + struct draw_vertex_shader * draw_create_vs_exec(struct draw_context *draw, const struct pipe_shader_state *templ); @@ -80,8 +150,43 @@ struct draw_vertex_shader * draw_create_vs_llvm(struct draw_context *draw, const struct pipe_shader_state *templ); +/******************************************************************************** + * Helpers for vs implementations that don't do their own fetch/emit varients. + * Means these can be shared between shaders. + */ +struct translate; +struct translate_key; + +struct translate *draw_vs_get_fetch( struct draw_context *draw, + struct translate_key *key ); + + +struct translate *draw_vs_get_emit( struct draw_context *draw, + struct translate_key *key ); + +struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, + const struct draw_vs_varient_key *key ); + + + +static INLINE int draw_vs_varient_keysize( const struct draw_vs_varient_key *key ) +{ + return 2 * sizeof(int) + key->nr_elements * sizeof(struct draw_vs_element); +} + +static INLINE int draw_vs_varient_key_compare( const struct draw_vs_varient_key *a, + const struct draw_vs_varient_key *b ) +{ + int keysize = draw_vs_varient_keysize(a); + return memcmp(a, b, keysize); +} + + + + #define MAX_TGSI_VERTICES 4 + #endif diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index cb80d008cd..4501877efc 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -179,9 +179,11 @@ draw_create_vs_exec(struct draw_context *draw, tgsi_scan_shader(state->tokens, &vs->base.info); + vs->base.draw = draw; vs->base.prepare = vs_exec_prepare; vs->base.run_linear = vs_exec_run_linear; vs->base.delete = vs_exec_delete; + vs->base.create_varient = draw_vs_varient_generic; vs->machine = &draw->vs.machine; return &vs->base; diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 171da51dd5..621472ec7c 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -114,7 +114,9 @@ draw_create_vs_llvm(struct draw_context *draw, tgsi_scan_shader(vs->base.state.tokens, &vs->base.info); + vs->base.draw = draw; vs->base.prepare = vs_llvm_prepare; + vs->base.create_varient = draw_vs_varient_generic; vs->base.run_linear = vs_llvm_run_linear; vs->base.delete = vs_llvm_delete; vs->machine = &draw->machine; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 13ad032bd3..df94a7e0c7 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -155,6 +155,8 @@ draw_create_vs_sse(struct draw_context *draw, tgsi_scan_shader(templ->tokens, &vs->base.info); + vs->base.draw = draw; + vs->base.create_varient = draw_vs_varient_generic; vs->base.prepare = vs_sse_prepare; vs->base.run_linear = vs_sse_run_linear; vs->base.delete = vs_sse_delete; diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c new file mode 100644 index 0000000000..d27b0f6187 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -0,0 +1,229 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Keith Whitwell + */ + +#include "pipe/p_util.h" +#include "draw/draw_context.h" +#include "draw/draw_private.h" +#include "draw/draw_vbuf.h" +#include "draw/draw_vertex.h" +#include "draw/draw_vs.h" +#include "translate/translate.h" +#include "translate/translate_cache.h" + +/* A first pass at incorporating vertex fetch/emit functionality into + */ +struct draw_vs_varient_generic { + struct draw_vs_varient base; + + + + struct draw_vertex_shader *shader; + struct draw_context *draw; + + /* Basic plan is to run these two translate functions before/after + * the vertex shader's existing run_linear() routine to simulate + * the inclusion of this functionality into the shader... + * + * Next will look at actually including it. + */ + struct translate *fetch; + struct translate *emit; + + const float (*constants)[4]; +}; + + + + +static void vsvg_set_constants( struct draw_vs_varient *varient, + const float (*constants)[4] ) +{ + struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; + + vsvg->constants = constants; +} + + +static void vsvg_set_input( struct draw_vs_varient *varient, + unsigned buffer, + const void *ptr, + unsigned stride ) +{ + struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; + + vsvg->fetch->set_buffer(vsvg->fetch, + buffer, + ptr, + stride); +} + + +static void vsvg_run_elts( struct draw_vs_varient *varient, + const unsigned *elts, + unsigned count, + void *output_buffer) +{ + struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; + + /* Want to do this in small batches for cache locality? + */ + + vsvg->fetch->run_elts( vsvg->fetch, + elts, + count, + output_buffer ); + + //if (!vsvg->base.vs->is_passthrough) + { + vsvg->base.vs->run_linear( vsvg->base.vs, + output_buffer, + output_buffer, + vsvg->constants, + count, + vsvg->base.key.output_stride, + vsvg->base.key.output_stride); + + //if (!vsvg->already_in_emit_format) + + vsvg->emit->set_buffer( vsvg->emit, + 0, + output_buffer, + vsvg->base.key.output_stride ); + + + vsvg->emit->run( vsvg->emit, + 0, count, + output_buffer ); + } +} + + +static void vsvg_run_linear( struct draw_vs_varient *varient, + unsigned start, + unsigned count, + void *output_buffer ) +{ + struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; + + //debug_printf("%s %d %d\n", __FUNCTION__, start, count); + + + vsvg->fetch->run( vsvg->fetch, + start, + count, + output_buffer ); + + //if (!vsvg->base.vs->is_passthrough) + { + vsvg->base.vs->run_linear( vsvg->base.vs, + output_buffer, + output_buffer, + vsvg->constants, + count, + vsvg->base.key.output_stride, + vsvg->base.key.output_stride); + + //if (!vsvg->already_in_emit_format) + vsvg->emit->set_buffer( vsvg->emit, + 0, + output_buffer, + vsvg->base.key.output_stride ); + + + vsvg->emit->run( vsvg->emit, + 0, count, + output_buffer ); + } +} + + + +static void vsvg_destroy( struct draw_vs_varient *varient ) +{ + FREE(varient); +} + + +struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, + const struct draw_vs_varient_key *key ) +{ + unsigned i; + struct translate_key fetch, emit; + + struct draw_vs_varient_generic *vsvg = CALLOC_STRUCT( draw_vs_varient_generic ); + if (vsvg == NULL) + return NULL; + + vsvg->base.key = *key; + vsvg->base.vs = vs; + vsvg->base.set_input = vsvg_set_input; + vsvg->base.set_constants = vsvg_set_constants; + vsvg->base.run_elts = vsvg_run_elts; + vsvg->base.run_linear = vsvg_run_linear; + vsvg->base.destroy = vsvg_destroy; + + + + /* OK, have to build a new one: + */ + fetch.nr_elements = vs->info.num_inputs; + fetch.output_stride = 0; + for (i = 0; i < vs->info.num_inputs; i++) { + fetch.element[i].input_format = key->element[i].in.format; + fetch.element[i].input_buffer = key->element[i].in.buffer; + fetch.element[i].input_offset = key->element[i].in.offset; + fetch.element[i].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + fetch.element[i].output_offset = fetch.output_stride; + fetch.output_stride += 4 * sizeof(float); + } + + + emit.nr_elements = vs->info.num_outputs; + emit.output_stride = key->output_stride; + for (i = 0; i < vs->info.num_outputs; i++) { + emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit.element[i].input_buffer = 0; + emit.element[i].input_offset = i * 4 * sizeof(float); + emit.element[i].output_format = key->element[i].out.format; + emit.element[i].output_offset = key->element[i].out.offset; + } + + vsvg->fetch = draw_vs_get_fetch( vs->draw, &fetch ); + vsvg->emit = draw_vs_get_emit( vs->draw, &emit ); + + return &vsvg->base; +} + + + + + -- cgit v1.2.3 From 9232f0c023af060b12f77dee5e8b6a533c48e146 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 19 May 2008 16:28:53 +0100 Subject: rtasm: remove unused struct member --- src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index eacaeeaf6f..baa10b7d4a 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -44,7 +44,6 @@ struct x86_function { unsigned stack_offset; int need_emms; unsigned char error_overflow[4]; - const char *fn; }; enum x86_reg_file { -- cgit v1.2.3 From 8618e6aa16bdba2c8b08124261bbaedaf7e22447 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 20 May 2008 14:34:06 +0100 Subject: translate: remove spurious comment --- src/gallium/auxiliary/translate/translate_sse.c | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index a54ac5a82f..582d6f6466 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -472,13 +472,7 @@ static boolean build_vertex_emit( struct translate_sse *p, x86_lea(p->func, vertexECX, x86_make_disp(vertexECX, p->translate.key.output_stride)); /* Incr index - */ /* Emit code for each of the attributes. Currently routes - * everything through SSE registers, even when it might be more - * efficient to stick with regular old x86. No optimization or - * other tricks - enough new ground to cover here just getting - * things working. - */ - + */ if (linear) { x86_inc(p->func, idxEBX); } -- cgit v1.2.3 From d3e64caef6f8654af1a84825803e517ab8221c68 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 08:28:16 +0100 Subject: rtasm: export debug reg print function --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 20 +++++++++----------- src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 3 +++ 2 files changed, 12 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 4e036d9032..68ac91ed13 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -36,11 +36,8 @@ #define DUMP_SSE 0 -#if DUMP_SSE -static void -_print_reg( - struct x86_reg reg ) +void x86_print_reg( struct x86_reg reg ) { if (reg.mod != mod_REG) debug_printf( "[" ); @@ -77,6 +74,7 @@ _print_reg( debug_printf( "]" ); } +#if DUMP_SSE #define DUMP_START() debug_printf( "\n" ) #define DUMP_END() debug_printf( "\n" ) @@ -87,7 +85,7 @@ _print_reg( foo++; \ if (*foo) \ foo++; \ - debug_printf( "\n% 15s ", foo ); \ + debug_printf( "\n% 4x% 15s ", p->csr - p->store, foo ); \ } while (0) #define DUMP_I( I ) do { \ @@ -97,27 +95,27 @@ _print_reg( #define DUMP_R( R0 ) do { \ DUMP(); \ - _print_reg( R0 ); \ + x86_print_reg( R0 ); \ } while( 0 ) #define DUMP_RR( R0, R1 ) do { \ DUMP(); \ - _print_reg( R0 ); \ + x86_print_reg( R0 ); \ debug_printf( ", " ); \ - _print_reg( R1 ); \ + x86_print_reg( R1 ); \ } while( 0 ) #define DUMP_RI( R0, I ) do { \ DUMP(); \ - _print_reg( R0 ); \ + x86_print_reg( R0 ); \ debug_printf( ", %u", I ); \ } while( 0 ) #define DUMP_RRI( R0, R1, I ) do { \ DUMP(); \ - _print_reg( R0 ); \ + x86_print_reg( R0 ); \ debug_printf( ", " ); \ - _print_reg( R1 ); \ + x86_print_reg( R1 ); \ debug_printf( ", %u", I ); \ } while( 0 ) diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index baa10b7d4a..1e02c6e73b 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -106,6 +106,9 @@ void x86_init_func_size( struct x86_function *p, unsigned code_size ); void x86_release_func( struct x86_function *p ); void (*x86_get_func( struct x86_function *p ))( void ); +/* Debugging: + */ +void x86_print_reg( struct x86_reg reg ); /* Create and manipulate registers and regmem values: -- cgit v1.2.3 From 9343779a8c800cf72e38b09b6f5087a0df258c08 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 08:28:53 +0100 Subject: gallium: define PIPE_CDECL calling convention, which really is cdecl everywhere --- src/gallium/include/pipe/p_compiler.h | 21 ++++++++++++++++++--- 1 file changed, 18 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index a4b772bc4f..01d1807b1c 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -103,6 +103,17 @@ typedef unsigned int uintptr_t; #endif +/* This should match linux gcc cdecl semantics everywhere, so that we + * just codegen one calling convention on all platforms. + */ +#ifdef WIN32 +#define PIPE_CDECL __cdecl +#else +#define PIPE_CDECL +#endif + + + #if defined __GNUC__ #define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___aligned[SIZE] __attribute__(( aligned( 16 ) )) #define ALIGN16_ASSIGN(NAME) NAME##___aligned @@ -115,12 +126,16 @@ typedef unsigned int uintptr_t; -/** For calling code-gen'd functions */ +/** + * For calling code-gen'd functions, phase out in favor of + * PIPE_CDECL, above, which really means cdecl on all platforms, not + * like the below... + */ #if !defined(XSTDCALL) #if defined(WIN32) -#define XSTDCALL __stdcall +#define XSTDCALL __stdcall /* phase this out */ #else -#define XSTDCALL +#define XSTDCALL /* XXX: NOTE! not STDCALL! */ #endif #endif -- cgit v1.2.3 From b5c8b3fba6ac90a0d83e02bfe432142f1adee9e5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 08:29:19 +0100 Subject: translate: mark functions as PIPE_CDECL --- src/gallium/auxiliary/translate/translate.h | 18 +++++++++--------- src/gallium/auxiliary/translate/translate_generic.c | 16 ++++++++-------- src/gallium/auxiliary/translate/translate_sse.c | 20 +++++++------------- 3 files changed, 24 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h index b8210af50c..c3b754a902 100644 --- a/src/gallium/auxiliary/translate/translate.h +++ b/src/gallium/auxiliary/translate/translate.h @@ -71,15 +71,15 @@ struct translate { const void *ptr, unsigned stride ); - void (*run_elts)( struct translate *, - const unsigned *elts, - unsigned count, - void *output_buffer); - - void (*run)( struct translate *, - unsigned start, - unsigned count, - void *output_buffer); + void (PIPE_CDECL *run_elts)( struct translate *, + const unsigned *elts, + unsigned count, + void *output_buffer); + + void (PIPE_CDECL *run)( struct translate *, + unsigned start, + unsigned count, + void *output_buffer); }; diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index 402780ee53..a25d94f2ca 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -541,10 +541,10 @@ static emit_func get_emit_func( enum pipe_format format ) /** * Fetch vertex attributes for 'count' vertices. */ -static void generic_run_elts( struct translate *translate, - const unsigned *elts, - unsigned count, - void *output_buffer ) +static void PIPE_CDECL generic_run_elts( struct translate *translate, + const unsigned *elts, + unsigned count, + void *output_buffer ) { struct translate_generic *tg = translate_generic(translate); char *vert = output_buffer; @@ -580,10 +580,10 @@ static void generic_run_elts( struct translate *translate, -static void generic_run( struct translate *translate, - unsigned start, - unsigned count, - void *output_buffer ) +static void PIPE_CDECL generic_run( struct translate *translate, + unsigned start, + unsigned count, + void *output_buffer ) { struct translate_generic *tg = translate_generic(translate); char *vert = output_buffer; diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index 582d6f6466..2fc8b9d3d0 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -45,22 +45,16 @@ #define W 3 -#ifdef WIN32 -#define RTASM __cdecl -#else -#define RTASM -#endif - -typedef void (RTASM *run_func)( struct translate *translate, - unsigned start, - unsigned count, - void *output_buffer ); - -typedef void (RTASM *run_elts_func)( struct translate *translate, - const unsigned *elts, +typedef void (PIPE_CDECL *run_func)( struct translate *translate, + unsigned start, unsigned count, void *output_buffer ); +typedef void (PIPE_CDECL *run_elts_func)( struct translate *translate, + const unsigned *elts, + unsigned count, + void *output_buffer ); + struct translate_sse { -- cgit v1.2.3 From ba738a3135415de8b381cd8845cd6c435d5747a8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 09:43:30 +0100 Subject: draw: mark varient functions as PIPE_CDECL --- src/gallium/auxiliary/draw/draw_vs.h | 30 +++++++++++++++++++++--------- 1 file changed, 21 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 677be0d28d..6bfc2c8d75 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -62,6 +62,19 @@ struct draw_vs_varient_key { struct draw_vs_element element[PIPE_MAX_ATTRIBS]; }; +struct draw_vs_varient; + +typedef void (PIPE_CDECL *vsv_run_elts_func)( struct draw_vs_varient *, + const unsigned *elts, + unsigned count, + void *output_buffer); + +typedef void (PIPE_CDECL *vsv_run_linear_func)( struct draw_vs_varient *, + unsigned start, + unsigned count, + void *output_buffer); + + struct draw_vs_varient { struct draw_vs_varient_key key; @@ -75,16 +88,15 @@ struct draw_vs_varient { void (*set_constants)( struct draw_vs_varient *, const float (*constants)[4] ); + void (PIPE_CDECL *run_linear)( struct draw_vs_varient *shader, + unsigned start, + unsigned count, + void *output_buffer ); - void (*run_linear)( struct draw_vs_varient *shader, - unsigned start, - unsigned count, - void *output_buffer ); - - void (*run_elts)( struct draw_vs_varient *shader, - const unsigned *elts, - unsigned count, - void *output_buffer ); + void (PIPE_CDECL *run_elts)( struct draw_vs_varient *shader, + const unsigned *elts, + unsigned count, + void *output_buffer ); void (*destroy)( struct draw_vs_varient * ); }; -- cgit v1.2.3 From 1ba10e5ccf5cd0c990922e982e1e9bc6be48a5e4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 09:44:16 +0100 Subject: draw: add aos vertex shader varient --- src/gallium/auxiliary/draw/Makefile | 2 + src/gallium/auxiliary/draw/draw_vs.h | 10 + src/gallium/auxiliary/draw/draw_vs_aos.c | 1739 +++++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs_aos.h | 181 +++ src/gallium/auxiliary/draw/draw_vs_aos_io.c | 314 +++++ src/gallium/auxiliary/draw/draw_vs_sse.c | 1 + 6 files changed, 2247 insertions(+) create mode 100644 src/gallium/auxiliary/draw/draw_vs_aos.c create mode 100644 src/gallium/auxiliary/draw/draw_vs_aos.h create mode 100644 src/gallium/auxiliary/draw/draw_vs_aos_io.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 84877994fb..9a88ecc070 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -35,6 +35,8 @@ C_SOURCES = \ draw_vertex.c \ draw_vs.c \ draw_vs_varient.c \ + draw_vs_aos.c \ + draw_vs_aos_io.c \ draw_vs_exec.c \ draw_vs_llvm.c \ draw_vs_sse.c diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 6bfc2c8d75..5a8d0da06d 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -162,6 +162,16 @@ struct draw_vertex_shader * draw_create_vs_llvm(struct draw_context *draw, const struct pipe_shader_state *templ); + + +struct draw_vs_varient_key; +struct draw_vertex_shader; + +struct draw_vs_varient *draw_vs_varient_aos_sse( struct draw_vertex_shader *vs, + const struct draw_vs_varient_key *key ); + + + /******************************************************************************** * Helpers for vs implementations that don't do their own fetch/emit varients. * Means these can be shared between shaders. diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c new file mode 100644 index 0000000000..620f5e3592 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -0,0 +1,1739 @@ +/* + * Mesa 3-D graphics library + * Version: 6.3 + * + * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * Translate tgsi vertex programs to x86/x87/SSE/SSE2 machine code + * using the rtasm runtime assembler. Based on the old + * t_vb_arb_program_sse.c + */ + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_util.h" +#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/util/tgsi_dump.h" + +#include "draw_vs.h" +#include "draw_vs_aos.h" + +#include "rtasm/rtasm_x86sse.h" + +#ifdef PIPE_ARCH_X86 + + +#define DISASSEM 0 + + + + + +static INLINE boolean eq( struct x86_reg a, + struct x86_reg b ) +{ + return (a.file == b.file && + a.idx == b.idx && + a.mod == b.mod && + a.disp == b.disp); +} + + +static struct x86_reg get_reg_ptr(struct aos_compilation *cp, + unsigned file, + unsigned idx ) +{ + struct x86_reg ptr = cp->machine_EDX; + + switch (file) { + case TGSI_FILE_INPUT: + return x86_make_disp(ptr, Offset(struct aos_machine, input[idx])); + + case TGSI_FILE_OUTPUT: + return x86_make_disp(ptr, Offset(struct aos_machine, output[idx])); + + case TGSI_FILE_TEMPORARY: + return x86_make_disp(ptr, Offset(struct aos_machine, temp[idx])); + + case TGSI_FILE_IMMEDIATE: + return x86_make_disp(ptr, Offset(struct aos_machine, immediate[idx])); + + case TGSI_FILE_CONSTANT: + return x86_make_disp(ptr, Offset(struct aos_machine, constant[idx])); + + case AOS_FILE_INTERNAL: + return x86_make_disp(ptr, Offset(struct aos_machine, immediate[idx])); + + default: + ERROR(cp, "unknown reg file"); + return x86_make_reg(0,0); + } +} + + +struct x86_reg aos_get_internal( struct aos_compilation *cp, + unsigned imm ) +{ + return get_reg_ptr( cp, + AOS_FILE_INTERNAL, + imm + 1 ); +} + +static void spill( struct aos_compilation *cp, unsigned idx ) +{ + if (!cp->xmm[idx].dirty || + (cp->xmm[idx].file != TGSI_FILE_INPUT && /* inputs are fetched into xmm & set dirty */ + cp->xmm[idx].file != TGSI_FILE_OUTPUT && + cp->xmm[idx].file != TGSI_FILE_TEMPORARY)) { + ERROR(cp, "invalid spill"); + return; + } + else { + struct x86_reg oldval = get_reg_ptr(cp, + cp->xmm[idx].file, + cp->xmm[idx].idx); + + assert(cp->xmm[idx].dirty); + sse_movups(cp->func, oldval, x86_make_reg(file_XMM, idx)); + cp->xmm[idx].dirty = 0; + } +} + +struct x86_reg aos_get_xmm_reg( struct aos_compilation *cp ) +{ + unsigned i; + unsigned oldest = 0; + + for (i = 0; i < 8; i++) + if (cp->xmm[i].last_used < cp->xmm[oldest].last_used) + oldest = i; + + /* Need to write out the old value? + */ + if (cp->xmm[oldest].dirty) + spill(cp, oldest); + + assert(cp->xmm[oldest].last_used != cp->insn_counter); + + cp->xmm[oldest].file = TGSI_FILE_NULL; + cp->xmm[oldest].idx = 0; + cp->xmm[oldest].last_used = cp->insn_counter; + return x86_make_reg(file_XMM, oldest); +} + +void aos_release_xmm_reg( struct aos_compilation *cp, + unsigned idx ) +{ + cp->xmm[idx].file = TGSI_FILE_NULL; + cp->xmm[idx].idx = 0; + cp->xmm[idx].dirty = 0; + cp->xmm[idx].last_used = 0; +} + +static void invalidate_xmm( struct aos_compilation *cp, + unsigned file, unsigned idx ) +{ + unsigned i; + + /* Invalidate any old copy of this register in XMM0-7. + */ + for (i = 0; i < 8; i++) { + if (cp->xmm[i].file == file && cp->xmm[i].idx == idx) { + + if (cp->xmm[i].dirty) + spill(cp, i); + + aos_release_xmm_reg(cp, i); + break; + } + } + + for (; i < 8; i++) { + if (cp->xmm[i].file == file && cp->xmm[i].idx == idx) { + assert(0); + } + } +} + + +void aos_adopt_xmm_reg( struct aos_compilation *cp, + struct x86_reg reg, + unsigned file, + unsigned idx, + unsigned dirty ) +{ + if (reg.file != file_XMM) { + assert(0); + return; + } + + invalidate_xmm(cp, file, idx); + cp->xmm[reg.idx].file = file; + cp->xmm[reg.idx].idx = idx; + cp->xmm[reg.idx].dirty = dirty; +} + + + +static struct x86_reg aos_get_shader_reg_ptr( struct aos_compilation *cp, + unsigned file, + unsigned idx ) +{ + invalidate_xmm( cp, file, idx ); + return get_reg_ptr( cp, file, idx ); +} + + +/* As above, but return a pointer. Note - this pointer may alias + * those returned by get_arg_ptr(). + */ +static struct x86_reg get_dst_ptr( struct aos_compilation *cp, + const struct tgsi_full_dst_register *dst ) +{ + return aos_get_shader_reg_ptr( cp, dst->DstRegister.File, dst->DstRegister.Index ); +} + + + + + +/* Return an XMM reg if the argument is resident, otherwise return a + * base+offset pointer to the saved value. + */ +struct x86_reg aos_get_shader_reg( struct aos_compilation *cp, + unsigned file, + unsigned idx ) +{ + unsigned i; + + for (i = 0; i < 8; i++) { + if (cp->xmm[i].file == file && + cp->xmm[i].idx == idx) + { + cp->xmm[i].last_used = cp->insn_counter; + return x86_make_reg(file_XMM, i); + } + } + + /* If not found in the XMM register file, return an indirect + * reference to the in-memory copy: + */ + return get_reg_ptr( cp, file, idx ); +} + + + + + +/* Emulate pshufd insn in regular SSE, if necessary: + */ +static void emit_pshufd( struct aos_compilation *cp, + struct x86_reg dst, + struct x86_reg arg0, + ubyte shuf ) +{ + if (cp->have_sse2) { + sse2_pshufd(cp->func, dst, arg0, shuf); + } + else { + if (!eq(dst, arg0)) + sse_movups(cp->func, dst, arg0); + + sse_shufps(cp->func, dst, dst, shuf); + } +} + + + + +/* Helper for writemask: + */ +static boolean emit_shuf_copy1( struct aos_compilation *cp, + struct x86_reg dst, + struct x86_reg arg0, + struct x86_reg arg1, + ubyte shuf ) +{ + struct x86_reg tmp = aos_get_xmm_reg(cp); + sse_movups(cp->func, dst, arg1); + emit_pshufd(cp, dst, dst, shuf); + emit_pshufd(cp, tmp, arg0, shuf); + + sse_movss(cp->func, dst, tmp); + + emit_pshufd(cp, dst, dst, shuf); + + aos_release_xmm_reg(cp, tmp.idx); + return TRUE; +} + + +/* Helper for writemask: + */ +static boolean emit_shuf_copy2( struct aos_compilation *cp, + struct x86_reg dst, + struct x86_reg arg0, + struct x86_reg arg1, + ubyte shuf ) +{ + struct x86_reg tmp = aos_get_xmm_reg(cp); + emit_pshufd(cp, dst, arg1, shuf); + emit_pshufd(cp, tmp, arg0, shuf); + + sse_shufps(cp->func, dst, tmp, SHUF(X, Y, Z, W)); + + emit_pshufd(cp, dst, dst, shuf); + + aos_release_xmm_reg(cp, tmp.idx); + return TRUE; +} + +#define SSE_SWIZZLE_NOOP ((0<<0) | (1<<2) | (2<<4) | (3<<6)) + + +/* Locate a source register and perform any required (simple) swizzle. + * + * Just fail on complex swizzles at this point. + */ +static struct x86_reg fetch_src( struct aos_compilation *cp, + const struct tgsi_full_src_register *src ) +{ + struct x86_reg arg0 = aos_get_shader_reg(cp, + src->SrcRegister.File, + src->SrcRegister.Index); + unsigned i; + unsigned swz = 0; + unsigned negs = 0; + unsigned abs = 0; + + for (i = 0; i < 4; i++) { + unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( src, i ); + unsigned neg = tgsi_util_get_full_src_register_sign_mode( src, i ); + + switch (swizzle) { + case TGSI_EXTSWIZZLE_ZERO: + case TGSI_EXTSWIZZLE_ONE: + ERROR(cp, "not supporting full swizzles yet in tgsi_aos_sse2"); + break; + + default: + swz |= (swizzle & 0x3) << (i * 2); + break; + } + + switch (neg) { + case TGSI_UTIL_SIGN_TOGGLE: + negs |= (1<func, dst, arg0); + + aos_release_xmm_reg(cp, tmp.idx); + arg0 = dst; + } + + if (abs && abs != 0xf) { + ERROR(cp, "unsupported partial abs"); + } + + if (abs) { + struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); + struct x86_reg tmp = aos_get_xmm_reg(cp); + + sse_movups(cp->func, tmp, arg0); + sse_mulps(cp->func, tmp, neg); + sse_maxps(cp->func, dst, arg0); + + aos_release_xmm_reg(cp, tmp.idx); + arg0 = dst; + } + } + + return arg0; +} + +static void x87_fld_src( struct aos_compilation *cp, + const struct tgsi_full_src_register *src, + unsigned channel ) +{ + struct x86_reg arg0 = aos_get_shader_reg_ptr(cp, + src->SrcRegister.File, + src->SrcRegister.Index); + + unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( src, channel ); + unsigned neg = tgsi_util_get_full_src_register_sign_mode( src, channel ); + + switch (swizzle) { + case TGSI_EXTSWIZZLE_ZERO: + x87_fldz( cp->func ); + break; + + case TGSI_EXTSWIZZLE_ONE: + x87_fld1( cp->func ); + break; + + default: + x87_fld( cp->func, x86_make_disp(arg0, (swizzle & 3) * sizeof(float)) ); + break; + } + + + switch (neg) { + case TGSI_UTIL_SIGN_TOGGLE: + /* Flip the sign: + */ + x87_fchs( cp->func ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + + case TGSI_UTIL_SIGN_CLEAR: + x87_fabs( cp->func ); + break; + + case TGSI_UTIL_SIGN_SET: + x87_fabs( cp->func ); + x87_fchs( cp->func ); + break; + + default: + ERROR(cp, "unsupported sign-mode"); + break; + } +} + + + + + + +/* Used to implement write masking. This and most of the other instructions + * here would be easier to implement if there had been a translation + * to a 2 argument format (dst/arg0, arg1) at the shader level before + * attempting to translate to x86/sse code. + */ +static void store_dest( struct aos_compilation *cp, + const struct tgsi_full_dst_register *reg, + struct x86_reg result ) +{ + if (reg->DstRegister.WriteMask == 0) + { + return; + } + else if (reg->DstRegister.WriteMask == TGSI_WRITEMASK_XYZW) + { + if (result.file == file_XMM) { + aos_adopt_xmm_reg(cp, + result, + reg->DstRegister.File, + reg->DstRegister.Index, + TRUE); + } + else { + struct x86_reg dst = aos_get_xmm_reg(cp); + aos_adopt_xmm_reg(cp, + dst, + reg->DstRegister.File, + reg->DstRegister.Index, + TRUE); + sse_movups(cp->func, dst, result); + } + } + else + { + /* Previous value of the dest register: + */ + struct x86_reg old_dst = aos_get_shader_reg(cp, + reg->DstRegister.File, + reg->DstRegister.Index); + + + /* Alloc an xmm reg to hold the new value of the dest register: + */ + struct x86_reg dst = aos_get_xmm_reg(cp); + + aos_adopt_xmm_reg(cp, + dst, + reg->DstRegister.File, + reg->DstRegister.Index, + TRUE ); + + switch (reg->DstRegister.WriteMask) { + case TGSI_WRITEMASK_X: + if (result.file == file_XMM) { + sse_movups(cp->func, dst, old_dst); + sse_movss(cp->func, dst, result); + } + else { + struct x86_reg tmp = aos_get_xmm_reg(cp); + sse_movups(cp->func, dst, old_dst); + sse_movss(cp->func, tmp, result); + sse_movss(cp->func, dst, tmp); + aos_release_xmm_reg(cp, tmp.idx); + } + break; + + case TGSI_WRITEMASK_XY: + sse_movups(cp->func, dst, old_dst); + sse_shufps(cp->func, dst, result, SHUF(X, Y, Z, W)); + break; + + case TGSI_WRITEMASK_ZW: + sse_movups(cp->func, dst, result); + sse_shufps(cp->func, dst, old_dst, SHUF(X, Y, Z, W)); + break; + + case TGSI_WRITEMASK_YZW: + if (old_dst.file == file_XMM) { + sse_movups(cp->func, dst, result); + sse_movss(cp->func, dst, old_dst); + } + else { + struct x86_reg tmp = aos_get_xmm_reg(cp); + sse_movups(cp->func, dst, result); + sse_movss(cp->func, tmp, old_dst); + sse_movss(cp->func, dst, tmp); + aos_release_xmm_reg(cp, tmp.idx); + } + break; + + case TGSI_WRITEMASK_Y: + emit_shuf_copy1(cp, dst, result, old_dst, SHUF(Y,X,Z,W)); + break; + + case TGSI_WRITEMASK_Z: + emit_shuf_copy1(cp, dst, result, old_dst, SHUF(Z,Y,X,W)); + break; + + case TGSI_WRITEMASK_W: + emit_shuf_copy1(cp, dst, result, old_dst, SHUF(W,Y,Z,X)); + break; + + case TGSI_WRITEMASK_XZ: + emit_shuf_copy2(cp, dst, result, old_dst, SHUF(X,Z,Y,W)); + break; + + case TGSI_WRITEMASK_XW: + emit_shuf_copy2(cp, dst, result, old_dst, SHUF(X,W,Z,Y)); + + case TGSI_WRITEMASK_YZ: + emit_shuf_copy2(cp, dst, result, old_dst, SHUF(Z,Y,X,W)); + break; + + case TGSI_WRITEMASK_YW: + emit_shuf_copy2(cp, dst, result, old_dst, SHUF(W,Y,Z,X)); + break; + + case TGSI_WRITEMASK_XZW: + emit_shuf_copy1(cp, dst, old_dst, result, SHUF(Y,X,Z,W)); + break; + + case TGSI_WRITEMASK_XYW: + emit_shuf_copy1(cp, dst, old_dst, result, SHUF(Z,Y,X,W)); + break; + + case TGSI_WRITEMASK_XYZ: + emit_shuf_copy1(cp, dst, old_dst, result, SHUF(W,Y,Z,X)); + break; + + default: + assert(0); /* not possible */ + break; + } + } +} + + +static void x87_fst_or_nop( struct x86_function *func, + unsigned writemask, + unsigned channel, + struct x86_reg ptr ) +{ + if (writemask & (1<DstRegister.WriteMask; + + x87_fst_or_nop(cp->func, writemask, 0, ptr); + x87_fst_or_nop(cp->func, writemask, 1, ptr); + x87_fst_or_nop(cp->func, writemask, 2, ptr); + x87_fstp_or_pop(cp->func, writemask, 3, ptr); +} + +/* Save current x87 state and put it into single precision mode. + */ +static void save_fpu_state( struct aos_compilation *cp ) +{ +#if 0 + x87_fnstcw( cp->func, x86_make_disp(regEDX, get_offset(m, &m->fpucntl_restore))); + x87_fldcw( cp->func, ); +#endif +} + +static void restore_fpu_state( struct aos_compilation *cp ) +{ +#if 0 + x87_fnclex(cp->func); + x87_fldcw(cp->func, x86_make_disp(regEDX, get_offset(m, &m->fpucntl_restore))); +#endif +} + +static void set_fpu_round_neg_inf( struct aos_compilation *cp ) +{ +#if 0 + if (cp->fpucntl != RND_NEG_FPU) { + struct x86_reg regEDX = x86_make_reg(file_REG32, reg_DX); + struct arb_vp_machine *m = NULL; + + cp->fpucntl = RND_NEG_FPU; + x87_fnclex(cp->func); + x87_fldcw(cp->func, x86_make_disp(regEDX, get_offset(m, &m->fpucntl_rnd_neg))); + } +#endif +} + +static void set_fpu_round_nearest( struct aos_compilation *cp ) +{ +#if 0 +#endif +} + + +static void emit_x87_ex2( struct aos_compilation *cp ) +{ + struct x86_reg st0 = x86_make_reg(file_x87, 0); + struct x86_reg st1 = x86_make_reg(file_x87, 1); + struct x86_reg st3 = x86_make_reg(file_x87, 3); + + set_fpu_round_neg_inf( cp ); + + x87_fld(cp->func, st0); /* a a */ + x87_fprndint( cp->func ); /* int(a) a */ + x87_fld(cp->func, st0); /* int(a) int(a) a */ + x87_fstp(cp->func, st3); /* int(a) a int(a)*/ + x87_fsubp(cp->func, st1); /* frac(a) int(a) */ + x87_f2xm1(cp->func); /* (2^frac(a))-1 int(a)*/ + x87_fld1(cp->func); /* 1 (2^frac(a))-1 int(a)*/ + x87_faddp(cp->func, st1); /* 2^frac(a) int(a) */ + x87_fscale(cp->func); /* 2^a */ +} + + + +/** + * The traditional instructions. All operate on internal registers + * and ignore write masks and swizzling issues. + */ + +static boolean emit_ABS( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); + + sse_movups(cp->func, dst, arg0); + sse_mulps(cp->func, dst, neg); + sse_maxps(cp->func, dst, arg0); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_ADD( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + sse_addps(cp->func, dst, arg1); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_COS( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + x87_fld_src(cp, &op->FullSrcRegisters[0], 0); + x87_fcos(cp->func); + x87_fstp_dest4(cp, &op->FullDstRegisters[0]); + return TRUE; +} + + +/* The dotproduct instructions don't really do that well in sse: + */ +static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg tmp = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + sse_mulps(cp->func, dst, arg1); + + /* Now the hard bit: sum the first 3 values: + */ + sse_movhlps(cp->func, tmp, dst); + sse_addss(cp->func, dst, tmp); /* a*x+c*z, b*y, ?, ? */ + emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z)); + sse_addss(cp->func, dst, tmp); + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + + aos_release_xmm_reg(cp, tmp.idx); + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + + + +static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg tmp = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + sse_mulps(cp->func, dst, arg1); + + /* Now the hard bit: sum the values: + */ + sse_movhlps(cp->func, tmp, dst); + sse_addps(cp->func, dst, tmp); /* a*x+c*z, b*y+d*w, a*x+c*z, b*y+d*w */ + emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z)); + sse_addss(cp->func, dst, tmp); + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + + aos_release_xmm_reg(cp, tmp.idx); + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_DPH( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg tmp = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + sse_mulps(cp->func, dst, arg1); + + /* Now the hard bit: sum the values (from DP3): + */ + sse_movhlps(cp->func, tmp, dst); + sse_addss(cp->func, dst, tmp); /* a*x+c*z, b*y, ?, ? */ + emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z)); + sse_addss(cp->func, dst, tmp); + emit_pshufd(cp, tmp, arg1, SHUF(W,W,W,W)); + sse_addss(cp->func, dst, tmp); + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + + aos_release_xmm_reg(cp, tmp.idx); + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_DST( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg tmp = aos_get_xmm_reg(cp); + struct x86_reg ones = aos_get_internal(cp, IMM_ONES); + +/* dst[0] = 1.0 * 1.0F; */ +/* dst[1] = arg0[1] * arg1[1]; */ +/* dst[2] = arg0[2] * 1.0; */ +/* dst[3] = 1.0 * arg1[3]; */ + + emit_shuf_copy2(cp, dst, arg0, ones, SHUF(X,W,Z,Y)); + emit_shuf_copy2(cp, tmp, arg1, ones, SHUF(X,Z,Y,W)); + sse_mulps(cp->func, dst, tmp); + + aos_release_xmm_reg(cp, tmp.idx); + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_LG2( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + x87_fld1(cp->func); /* 1 */ + x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* a0 1 */ + x87_fyl2x(cp->func); /* log2(a0) */ + x87_fstp_dest4(cp, &op->FullDstRegisters[0]); + return TRUE; +} + + +static boolean emit_EX2( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + x87_fld_src(cp, &op->FullSrcRegisters[0], 0); + + emit_x87_ex2(cp); + + x87_fstp_dest4(cp, &op->FullDstRegisters[0]); + return TRUE; +} + +static boolean emit_EXP( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); + struct x86_reg st0 = x86_make_reg(file_x87, 0); + struct x86_reg st1 = x86_make_reg(file_x87, 1); + struct x86_reg st3 = x86_make_reg(file_x87, 3); + unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; + + /* CAUTION: dst may alias arg0! + */ + x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* arg0.x */ + x87_fld(cp->func, st0); /* arg arg */ + + /* by default, fpu is setup to round-to-nearest. We want to + * change this now, and track the state through to the end of the + * generated function so that it isn't repeated unnecessarily. + * Alternately, could subtract .5 to get round to -inf behaviour. + */ + set_fpu_round_neg_inf( cp ); + x87_fprndint( cp->func ); /* flr(a) a */ + x87_fld(cp->func, st0); /* flr(a) flr(a) a */ + x87_fld1(cp->func); /* 1 floor(a) floor(a) a */ + x87_fst_or_nop(cp->func, writemask, 3, dst); /* stack unchanged */ + + x87_fscale(cp->func); /* 2^floor(a) floor(a) a */ + x87_fst(cp->func, st3); /* 2^floor(a) floor(a) a 2^floor(a)*/ + + x87_fstp_or_pop(cp->func, writemask, 0, dst); /* flr(a) a 2^flr(a) */ + + x87_fsubrp(cp->func, st1); /* frac(a) 2^flr(a) */ + + x87_fst_or_nop(cp->func, writemask, 1, dst); /* frac(a) 2^flr(a) */ + + x87_f2xm1(cp->func); /* (2^frac(a))-1 2^flr(a)*/ + x87_fld1(cp->func); /* 1 (2^frac(a))-1 2^flr(a)*/ + x87_faddp(cp->func, st1); /* 2^frac(a) 2^flr(a) */ + x87_fmulp(cp->func, st1); /* 2^a */ + + x87_fstp_or_pop(cp->func, writemask, 2, dst); + +/* dst[0] = 2^floor(tmp); */ +/* dst[1] = frac(tmp); */ +/* dst[2] = 2^floor(tmp) * 2^frac(tmp); */ +/* dst[3] = 1.0F; */ + return TRUE; +} + +static boolean emit_LOG( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); + struct x86_reg st0 = x86_make_reg(file_x87, 0); + struct x86_reg st1 = x86_make_reg(file_x87, 1); + struct x86_reg st2 = x86_make_reg(file_x87, 2); + unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; + + /* CAUTION: dst may alias arg0! + */ + x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* arg0.x */ + x87_fabs(cp->func); /* |arg0.x| */ + x87_fxtract(cp->func); /* mantissa(arg0.x), exponent(arg0.x) */ + x87_fst(cp->func, st2); /* mantissa, exponent, mantissa */ + x87_fld1(cp->func); /* 1, mantissa, exponent, mantissa */ + x87_fyl2x(cp->func); /* log2(mantissa), exponent, mantissa */ + x87_fadd(cp->func, st0, st1); /* e+l2(m), e, m */ + + x87_fstp_or_pop(cp->func, writemask, 2, dst); /* e, m */ + + x87_fld1(cp->func); /* 1, e, m */ + x87_fsub(cp->func, st1, st0); /* 1, e-1, m */ + + x87_fstp_or_pop(cp->func, writemask, 3, dst); /* e-1,m */ + x87_fstp_or_pop(cp->func, writemask, 0, dst); /* m */ + + x87_fadd(cp->func, st0, st0); /* 2m */ + + x87_fstp_or_pop( cp->func, writemask, 1, dst ); + + return TRUE; +} + +static boolean emit_FLR( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); + unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; + int i; + + set_fpu_round_neg_inf( cp ); + + /* Load all sources first to avoid aliasing + */ + for (i = 0; i < 4; i++) { + if (writemask & (1<FullSrcRegisters[0], i); + } + } + + for (i = 0; i < 4; i++) { + if (writemask & (1<func ); + x87_fstp(cp->func, x86_make_disp(dst, i*4)); + } + } + + return TRUE; +} + + +static boolean emit_RND( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); + unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; + int i; + + set_fpu_round_nearest( cp ); + + /* Load all sources first to avoid aliasing + */ + for (i = 0; i < 4; i++) { + if (writemask & (1<FullSrcRegisters[0], i); + } + } + + for (i = 0; i < 4; i++) { + if (writemask & (1<func ); + x87_fstp(cp->func, x86_make_disp(dst, i*4)); + } + } + + return TRUE; +} + + +static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); + struct x86_reg st0 = x86_make_reg(file_x87, 0); + struct x86_reg st1 = x86_make_reg(file_x87, 1); + unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; + int i; + + set_fpu_round_neg_inf( cp ); + + /* suck all the source values onto the stack before writing out any + * dst, which may alias... + */ + for (i = 0; i < 4; i++) { + if (writemask & (1<FullSrcRegisters[0], i); + } + } + + for (i = 0; i < 4; i++) { + if (writemask & (1<func, st0); /* a a */ + x87_fprndint( cp->func ); /* flr(a) a */ + x87_fsubrp(cp->func, st1); /* frc(a) */ + x87_fstp(cp->func, x86_make_disp(dst, i*4)); + } + } + + return TRUE; +} + + + +static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); + struct x86_reg st1 = x86_make_reg(file_x87, 1); + unsigned fixup1, fixup2; + unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; + + + /* Load the interesting parts of arg0: + */ + x87_fld_src(cp, &op->FullSrcRegisters[0], 3); + x87_fld_src(cp, &op->FullSrcRegisters[0], 1); + x87_fld_src(cp, &op->FullSrcRegisters[0], 0); + + + if (writemask & TGSI_WRITEMASK_XW) { + x87_fld1(cp->func); + x87_fst_or_nop(cp->func, writemask, 0, dst); + x87_fstp_or_pop(cp->func, writemask, 3, dst); + } + + if (writemask & TGSI_WRITEMASK_YZ) { + + /* Pre-zero destinations, may be overwritten later... fixme. + */ + x87_fldz(cp->func); + x87_fst_or_nop(cp->func, writemask, 1, dst); + x87_fstp_or_pop(cp->func, writemask, 2, dst); + + + /* Check arg0[0]: + */ + x87_fldz(cp->func); /* 0 a0 a1 a3 */ + x87_fucomp(cp->func, st1); /* a0 a1 a3 */ + x87_fnstsw(cp->func, cp->tmp_EAX); + x86_sahf(cp->func); + fixup1 = x86_jcc_forward(cp->func, cc_AE); + + x87_fstp_or_pop(cp->func, writemask, 1, dst); /* a1 a3 */ + + /* Check arg0[1]: + */ + x87_fldz(cp->func); /* 0 a1 a3 */ + x87_fucomp(cp->func, st1); /* a1 a3 */ + x87_fnstsw(cp->func, cp->tmp_EAX); + x86_sahf(cp->func); + fixup2 = x86_jcc_forward(cp->func, cc_AE); + + /* Compute pow(a1, a3) + */ + x87_fyl2x(cp->func); /* a3*log2(a1) */ + + emit_x87_ex2( cp ); /* 2^(a3*log2(a1)) */ + + x87_fstp_or_pop(cp->func, writemask, 2, dst); + + /* Land jumps: + */ + x86_fixup_fwd_jump(cp->func, fixup1); + x86_fixup_fwd_jump(cp->func, fixup2); + } + + return TRUE; +} + + + +static boolean emit_MAX( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + sse_maxps(cp->func, dst, arg1); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + + +static boolean emit_MIN( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + sse_minps(cp->func, dst, arg1); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_MOV( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg dst = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_MUL( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + sse_mulps(cp->func, dst, arg1); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + + +static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg arg2 = fetch_src(cp, &op->FullSrcRegisters[2]); + struct x86_reg dst = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + sse_mulps(cp->func, dst, arg1); + sse_addps(cp->func, dst, arg2); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + + +static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + x87_fld_src(cp, &op->FullSrcRegisters[1], 0); /* a1.x */ + x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* a0.x a1.x */ + x87_fyl2x(cp->func); /* a1*log2(a0) */ + + emit_x87_ex2( cp ); /* 2^(a1*log2(a0)) */ + + x87_fstp_dest4(cp, &op->FullDstRegisters[0]); + return TRUE; +} + + +static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg dst = aos_get_xmm_reg(cp); + + if (cp->have_sse2) { + sse2_rcpss(cp->func, dst, arg0); + /* extend precision here... + */ + } + else { + struct x86_reg ones = aos_get_internal(cp, IMM_ONES); + sse_movss(cp->func, dst, ones); + sse_divss(cp->func, dst, arg0); + } + + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg dst = aos_get_xmm_reg(cp); + + sse_rsqrtss(cp->func, dst, arg0); + + /* Extend precision here... + */ + + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + + +static boolean emit_SGE( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg ones = aos_get_internal(cp, IMM_ONES); + + sse_movups(cp->func, dst, arg0); + sse_cmpps(cp->func, dst, arg1, cc_NotLessThan); + sse_andps(cp->func, dst, ones); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_SIN( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + x87_fld_src(cp, &op->FullSrcRegisters[0], 0); + x87_fsin(cp->func); + x87_fstp_dest4(cp, &op->FullDstRegisters[0]); + return TRUE; +} + + + +static boolean emit_SLT( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg ones = aos_get_internal(cp, IMM_ONES); + + sse_movups(cp->func, dst, arg0); + sse_cmpps(cp->func, dst, arg1, cc_LessThan); + sse_andps(cp->func, dst, ones); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + +static boolean emit_SUB( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + + sse_movups(cp->func, dst, arg0); + sse_subps(cp->func, dst, arg1); + + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + + +static boolean emit_XPD( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); + struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg tmp0 = aos_get_xmm_reg(cp); + struct x86_reg tmp1 = aos_get_xmm_reg(cp); + + /* Could avoid tmp0, tmp1 if we overwrote arg0, arg1. Need a way + * to invalidate registers. This will come with better analysis + * (liveness analysis) of the incoming program. + */ + emit_pshufd(cp, dst, arg0, SHUF(Y, Z, X, W)); + emit_pshufd(cp, tmp1, arg1, SHUF(Z, X, Y, W)); + sse_mulps(cp->func, dst, tmp1); + emit_pshufd(cp, tmp0, arg0, SHUF(Z, X, Y, W)); + emit_pshufd(cp, tmp1, arg1, SHUF(Y, Z, X, W)); + sse_mulps(cp->func, tmp0, tmp1); + sse_subps(cp->func, dst, tmp0); + +/* dst[0] = arg0[1] * arg1[2] - arg0[2] * arg1[1]; */ +/* dst[1] = arg0[2] * arg1[0] - arg0[0] * arg1[2]; */ +/* dst[2] = arg0[0] * arg1[1] - arg0[1] * arg1[0]; */ +/* dst[3] is undef */ + + + aos_release_xmm_reg(cp, tmp0.idx); + aos_release_xmm_reg(cp, tmp1.idx); + store_dest(cp, &op->FullDstRegisters[0], dst); + return TRUE; +} + + + +static boolean +emit_instruction( struct aos_compilation *cp, + struct tgsi_full_instruction *inst ) +{ + switch( inst->Instruction.Opcode ) { + case TGSI_OPCODE_MOV: + return emit_MOV( cp, inst ); + + case TGSI_OPCODE_LIT: + return emit_LIT(cp, inst); + + case TGSI_OPCODE_RCP: + return emit_RCP(cp, inst); + + case TGSI_OPCODE_RSQ: + return emit_RSQ(cp, inst); + + case TGSI_OPCODE_EXP: + return emit_EXP(cp, inst); + + case TGSI_OPCODE_LOG: + return emit_LOG(cp, inst); + + case TGSI_OPCODE_MUL: + return emit_MUL(cp, inst); + + case TGSI_OPCODE_ADD: + return emit_ADD(cp, inst); + + case TGSI_OPCODE_DP3: + return emit_DP3(cp, inst); + + case TGSI_OPCODE_DP4: + return emit_DP4(cp, inst); + + case TGSI_OPCODE_DST: + return emit_DST(cp, inst); + + case TGSI_OPCODE_MIN: + return emit_MIN(cp, inst); + + case TGSI_OPCODE_MAX: + return emit_MAX(cp, inst); + + case TGSI_OPCODE_SLT: + return emit_SLT(cp, inst); + + case TGSI_OPCODE_SGE: + return emit_SGE(cp, inst); + + case TGSI_OPCODE_MAD: + return emit_MAD(cp, inst); + + case TGSI_OPCODE_SUB: + return emit_SUB(cp, inst); + + case TGSI_OPCODE_LERP: +// return emit_LERP(cp, inst); + return FALSE; + + case TGSI_OPCODE_FRAC: + return emit_FRC(cp, inst); + + case TGSI_OPCODE_CLAMP: +// return emit_CLAMP(cp, inst); + return FALSE; + + case TGSI_OPCODE_FLOOR: + return emit_FLR(cp, inst); + + case TGSI_OPCODE_ROUND: + return emit_RND(cp, inst); + + case TGSI_OPCODE_EXPBASE2: + return emit_EX2(cp, inst); + + case TGSI_OPCODE_LOGBASE2: + return emit_LG2(cp, inst); + + case TGSI_OPCODE_POWER: + return emit_POW(cp, inst); + + case TGSI_OPCODE_CROSSPRODUCT: + return emit_XPD(cp, inst); + + case TGSI_OPCODE_ABS: + return emit_ABS(cp, inst); + + case TGSI_OPCODE_DPH: + return emit_DPH(cp, inst); + + case TGSI_OPCODE_COS: + return emit_COS(cp, inst); + + case TGSI_OPCODE_SIN: + return emit_SIN(cp, inst); + + case TGSI_OPCODE_END: + return TRUE; + + default: + return FALSE; + } +} + +static boolean note_immediate( struct aos_compilation *cp, + struct tgsi_full_immediate *imm ) +{ + unsigned pos = cp->num_immediates++; + unsigned j; + + for (j = 0; j < imm->Immediate.Size; j++) { + cp->vaos->machine->immediate[pos][j] = imm->u.ImmediateFloat32[j].Float; + } + + return TRUE; +} + + + + +static void find_last_write_outputs( struct aos_compilation *cp ) +{ + struct tgsi_parse_context parse; + unsigned this_instruction = 0; + unsigned i; + + tgsi_parse_init( &parse, cp->vaos->base.vs->state.tokens ); + + while (!tgsi_parse_end_of_tokens( &parse )) { + + tgsi_parse_token( &parse ); + + if (parse.FullToken.Token.Type != TGSI_TOKEN_TYPE_INSTRUCTION) + continue; + + for (i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++) { + if (parse.FullToken.FullInstruction.FullDstRegisters[i].DstRegister.File == + TGSI_FILE_OUTPUT) + { + unsigned idx = parse.FullToken.FullInstruction.FullDstRegisters[i].DstRegister.Index; + cp->output_last_write[idx] = this_instruction; + } + } + + this_instruction++; + } + + tgsi_parse_free( &parse ); +} + + +#define ARG_VARIENT 1 +#define ARG_START_ELTS 2 +#define ARG_COUNT 3 +#define ARG_OUTBUF 4 + + +static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, + boolean linear ) +{ + struct tgsi_parse_context parse; + struct aos_compilation cp; + unsigned fixup, label; + + tgsi_parse_init( &parse, varient->base.vs->state.tokens ); + + memset(&cp, 0, sizeof(cp)); + + cp.insn_counter = 1; + cp.vaos = varient; + cp.have_sse2 = 1; + cp.func = &varient->func[ linear ? 0 : 1 ]; + + cp.tmp_EAX = x86_make_reg(file_REG32, reg_AX); + cp.idx_EBX = x86_make_reg(file_REG32, reg_BX); + cp.outbuf_ECX = x86_make_reg(file_REG32, reg_CX); + cp.machine_EDX = x86_make_reg(file_REG32, reg_DX); + cp.count_ESI = x86_make_reg(file_REG32, reg_SI); + + x86_init_func(cp.func); + + find_last_write_outputs(&cp); + + x86_push(cp.func, cp.idx_EBX); + x86_push(cp.func, cp.count_ESI); + + + /* Load arguments into regs: + */ + x86_mov(cp.func, cp.machine_EDX, x86_fn_arg(cp.func, ARG_VARIENT)); + x86_mov(cp.func, cp.idx_EBX, x86_fn_arg(cp.func, ARG_START_ELTS)); + x86_mov(cp.func, cp.count_ESI, x86_fn_arg(cp.func, ARG_COUNT)); + x86_mov(cp.func, cp.outbuf_ECX, x86_fn_arg(cp.func, ARG_OUTBUF)); + + + /* Compare count to zero and possibly bail. + */ + x86_xor(cp.func, cp.tmp_EAX, cp.tmp_EAX); + x86_cmp(cp.func, cp.count_ESI, cp.tmp_EAX); + fixup = x86_jcc_forward(cp.func, cc_E); + + /* Dig out the machine pointer from inside the varient arg + */ + x86_mov(cp.func, cp.machine_EDX, + x86_make_disp(cp.machine_EDX, + Offset( struct draw_vs_varient_aos_sse, machine ))); + + save_fpu_state( &cp ); + + /* Note address for loop jump + */ + label = x86_get_label(cp.func); + { + /* Fetch inputs... TODO: fetch lazily... + */ + if (!aos_fetch_inputs( &cp, linear )) + goto fail; + + /* Emit the shader: + */ + while( !tgsi_parse_end_of_tokens( &parse ) && !cp.error ) + { + tgsi_parse_token( &parse ); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_IMMEDIATE: + if (!note_immediate( &cp, &parse.FullToken.FullImmediate )) + goto fail; + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + if (!emit_instruction( &cp, &parse.FullToken.FullInstruction )) + goto fail; + break; + } + + cp.insn_counter++; + debug_printf("\n"); + } + + if (cp.error) + goto fail; + + /* Emit output... TODO: do this eagerly after the last write to a + * given output. + */ + if (!aos_emit_outputs( &cp )) + goto fail; + + + /* Next vertex: + */ + x86_lea(cp.func, + cp.outbuf_ECX, + x86_make_disp(cp.outbuf_ECX, + cp.vaos->base.key.output_stride)); + + /* Incr index + */ + if (linear) { + x86_inc(cp.func, cp.idx_EBX); + } + else { + x86_lea(cp.func, cp.idx_EBX, x86_make_disp(cp.idx_EBX, 4)); + } + + } + /* decr count, loop if not zero + */ + x86_dec(cp.func, cp.count_ESI); +/* x86_test(cp.func, cp.count_ESI, cp.count_ESI); */ + x86_jcc(cp.func, cc_NZ, label); + + restore_fpu_state(&cp); + + /* Land forward jump here: + */ + x86_fixup_fwd_jump(cp.func, fixup); + + /* Exit mmx state? + */ + if (cp.func->need_emms) + mmx_emms(cp.func); + + x86_pop(cp.func, cp.count_ESI); + x86_pop(cp.func, cp.idx_EBX); + + x86_ret(cp.func); + + tgsi_parse_free( &parse ); + return !cp.error; + + fail: + tgsi_parse_free( &parse ); + return FALSE; +} + + + +static void vaos_set_buffer( struct draw_vs_varient *varient, + unsigned buf, + const void *ptr, + unsigned stride ) +{ + struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + unsigned i; + + for (i = 0; i < vaos->base.vs->info.num_inputs; i++) { + if (vaos->base.key.element[i].in.buffer == buf) { + vaos->machine->attrib[i].input_ptr = ((char *)ptr + + vaos->base.key.element[i].in.offset); + vaos->machine->attrib[i].input_stride = stride; + } + } +} + + +static void vaos_destroy( struct draw_vs_varient *varient ) +{ + struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + + if (vaos->machine) + align_free( vaos->machine ); + + x86_release_func( &vaos->func[0] ); + x86_release_func( &vaos->func[1] ); + + FREE(vaos); +} + +static void vaos_run_elts( struct draw_vs_varient *varient, + const unsigned *elts, + unsigned count, + void *output_buffer ) +{ + struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + + vaos->gen_run_elts( varient, + elts, + count, + output_buffer ); +} + +static void vaos_run_linear( struct draw_vs_varient *varient, + unsigned start, + unsigned count, + void *output_buffer ) +{ + struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + + vaos->gen_run_linear( varient, + start, + count, + output_buffer ); +} + + +static void vaos_set_constants( struct draw_vs_varient *varient, + const float (*constants)[4] ) +{ + struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + + memcpy(vaos->machine->constant, + constants, + (vaos->base.vs->info.file_max[TGSI_FILE_CONSTANT] + 1) * 4 * sizeof(float)); +} + + +static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, + const struct draw_vs_varient_key *key ) +{ + struct draw_vs_varient_aos_sse *vaos = CALLOC_STRUCT(draw_vs_varient_aos_sse); + + if (!vaos) + goto fail; + + vaos->base.key = *key; + vaos->base.vs = vs; + vaos->base.set_input = vaos_set_buffer; + vaos->base.set_constants = vaos_set_constants; + vaos->base.destroy = vaos_destroy; + vaos->base.run_linear = vaos_run_linear; + vaos->base.run_elts = vaos_run_elts; + + vaos->machine = align_malloc( sizeof(struct aos_machine), 16 ); + if (!vaos->machine) + goto fail; + + memset(vaos->machine, 0, sizeof(struct aos_machine)); + + tgsi_dump(vs->state.tokens, 0); + + if (!build_vertex_program( vaos, TRUE )) + goto fail; + + if (!build_vertex_program( vaos, FALSE )) + goto fail; + + vaos->gen_run_linear = (vsv_run_linear_func)x86_get_func(&vaos->func[0]); + if (!vaos->gen_run_linear) + goto fail; + + vaos->gen_run_elts = (vsv_run_elts_func)x86_get_func(&vaos->func[1]); + if (!vaos->gen_run_elts) + goto fail; + + return &vaos->base; + + fail: + if (vaos->machine) + align_free( vaos->machine ); + + if (vaos) + x86_release_func( &vaos->func[0] ); + + if (vaos) + x86_release_func( &vaos->func[1] ); + + FREE(vaos); + + return NULL; +} + + +struct draw_vs_varient *draw_vs_varient_aos_sse( struct draw_vertex_shader *vs, + const struct draw_vs_varient_key *key ) +{ + struct draw_vs_varient *varient = varient_aos_sse( vs, key ); + + if (varient == NULL) { + assert(0); + varient = draw_vs_varient_generic( vs, key ); + } + + return varient; +} + + + +#endif diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h new file mode 100644 index 0000000000..1d8a055a90 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -0,0 +1,181 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Keith Whitwell + */ + +#ifndef DRAW_VS_AOS_H +#define DRAW_VS_AOS_H + + +struct tgsi_token; +struct x86_function; + +#include "pipe/p_state.h" +#include "rtasm/rtasm_x86sse.h" + + + + + +#define X 0 +#define Y 1 +#define Z 2 +#define W 3 + +#define MAX_INPUTS PIPE_MAX_ATTRIBS +#define MAX_OUTPUTS PIPE_MAX_ATTRIBS +#define MAX_TEMPS PIPE_MAX_ATTRIBS /* say */ +#define MAX_CONSTANTS PIPE_MAX_ATTRIBS /* say */ +#define MAX_IMMEDIATES PIPE_MAX_ATTRIBS /* say */ +#define MAX_INTERNALS 4 + +#define AOS_FILE_INTERNAL TGSI_FILE_COUNT + +/* This is the temporary storage used by all the aos_sse vs varients. + * Create one per context and reuse by passing a pointer in at + * vs_varient creation?? + */ +struct aos_machine { + float input [MAX_INPUTS ][4]; + float output [MAX_OUTPUTS ][4]; + float temp [MAX_TEMPS ][4]; + float constant [MAX_CONSTANTS ][4]; /* fixme -- should just be a pointer */ + float immediate[MAX_IMMEDIATES][4]; /* fixme -- should just be a pointer */ + float internal [MAX_INTERNALS ][4]; + + unsigned fpu_round_nearest; + unsigned fpu_round_neg_inf; + + struct { + const void *input_ptr; + unsigned input_stride; + + unsigned output_offset; + } attrib[PIPE_MAX_ATTRIBS]; +}; + + + + +struct aos_compilation { + struct x86_function *func; + struct draw_vs_varient_aos_sse *vaos; + + unsigned insn_counter; + unsigned num_immediates; + + struct { + unsigned idx:16; + unsigned file:8; + unsigned dirty:8; + unsigned last_used; + } xmm[8]; + + + boolean input_fetched[PIPE_MAX_ATTRIBS]; + unsigned output_last_write[PIPE_MAX_ATTRIBS]; + + boolean have_sse2; + boolean error; + short fpucntl; + + /* these are actually known values, but putting them in a struct + * like this is helpful to keep them in sync across the file. + */ + struct x86_reg tmp_EAX; + struct x86_reg idx_EBX; /* either start+i or &elt[i] */ + struct x86_reg outbuf_ECX; + struct x86_reg machine_EDX; + struct x86_reg count_ESI; /* decrements to zero */ +}; + +struct x86_reg aos_get_xmm_reg( struct aos_compilation *cp ); +void aos_release_xmm_reg( struct aos_compilation *cp, unsigned idx ); + +void aos_adopt_xmm_reg( struct aos_compilation *cp, + struct x86_reg reg, + unsigned file, + unsigned idx, + unsigned dirty ); + +struct x86_reg aos_get_shader_reg( struct aos_compilation *cp, + unsigned file, + unsigned idx ); + +boolean aos_fetch_inputs( struct aos_compilation *cp, + boolean linear ); + +boolean aos_emit_outputs( struct aos_compilation *cp ); + + +#define IMM_ONES 0 /* 1, 1,1,1 */ +#define IMM_NEGS 1 /* 1,-1,0,0 */ +#define IMM_IDENTITY 2 /* 0, 0,0,1 */ +#define IMM_INV_255 3 /* 1/255, 1/255, 1/255, 1/255 */ +#define IMM_255 4 /* 255, 255, 255, 255 */ + +struct x86_reg aos_get_internal( struct aos_compilation *cp, + unsigned imm ); + + +#define ERROR(cp, msg) \ +do { \ + debug_printf("%s: x86 translation failed: %s\n", __FUNCTION__, msg); \ + cp->error = 1; \ + assert(0); \ +} while (0) + + + + + + +struct draw_vs_varient_aos_sse { + struct draw_vs_varient base; + struct draw_context *draw; + +#if 0 + struct { + const void *ptr; + unsigned stride; + } attrib[PIPE_MAX_ATTRIBS]; +#endif + + struct aos_machine *machine; /* XXX: temporarily unshared */ + + vsv_run_linear_func gen_run_linear; + vsv_run_elts_func gen_run_elts; + + + struct x86_function func[2]; +}; + + + +#endif + diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c new file mode 100644 index 0000000000..72b2b3d11d --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -0,0 +1,314 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_util.h" +#include "tgsi/exec/tgsi_exec.h" +#include "draw_vs.h" +#include "draw_vs_aos.h" + +#include "rtasm/rtasm_x86sse.h" + +#ifdef PIPE_ARCH_X86 + +/* Note - don't yet have to worry about interacting with the code in + * draw_vs_aos.c as there is no intermingling of generated code... + * That may have to change, we'll see. + */ +static void emit_load_R32G32B32A32( struct aos_compilation *cp, + struct x86_reg data, + struct x86_reg src_ptr ) +{ + sse_movups(cp->func, data, src_ptr); +} + +static void emit_load_R32G32B32( struct aos_compilation *cp, + struct x86_reg data, + struct x86_reg src_ptr ) +{ + sse_movss(cp->func, data, x86_make_disp(src_ptr, 8)); + sse_shufps(cp->func, data, aos_get_internal( cp, IMM_IDENTITY ), SHUF(X,Y,Z,W) ); + sse_shufps(cp->func, data, data, SHUF(Y,Z,X,W) ); + sse_movlps(cp->func, data, src_ptr); +} + +static void emit_load_R32G32( struct aos_compilation *cp, + struct x86_reg data, + struct x86_reg src_ptr ) +{ + sse_movups(cp->func, data, aos_get_internal( cp, IMM_IDENTITY ) ); + sse_movlps(cp->func, data, src_ptr); +} + + +static void emit_load_R32( struct aos_compilation *cp, + struct x86_reg data, + struct x86_reg src_ptr ) +{ + sse_movss(cp->func, data, src_ptr); + sse_orps(cp->func, data, aos_get_internal( cp, IMM_IDENTITY ) ); +} + + +static void emit_load_R8G8B8A8_UNORM( struct aos_compilation *cp, + struct x86_reg data, + struct x86_reg src_ptr ) +{ + sse_movss(cp->func, data, src_ptr); + sse2_punpcklbw(cp->func, data, aos_get_internal( cp, IMM_IDENTITY )); + sse2_punpcklbw(cp->func, data, aos_get_internal( cp, IMM_IDENTITY )); + sse2_cvtdq2ps(cp->func, data, data); + sse_mulps(cp->func, data, aos_get_internal(cp, IMM_INV_255)); +} + + + +static void get_src_ptr( struct x86_function *func, + struct x86_reg src, + struct x86_reg machine, + struct x86_reg elt, + unsigned a ) +{ + struct x86_reg input_ptr = + x86_make_disp(machine, + Offset(struct aos_machine, attrib[a].input_ptr)); + + struct x86_reg input_stride = + x86_make_disp(machine, + Offset(struct aos_machine, attrib[a].input_stride)); + + /* Calculate pointer to current attrib: + */ + x86_mov(func, src, input_stride); + x86_imul(func, src, elt); + x86_add(func, src, input_ptr); +} + + +/* Extended swizzles? Maybe later. + */ +static void emit_swizzle( struct aos_compilation *cp, + struct x86_reg dest, + struct x86_reg src, + unsigned shuffle ) +{ + sse_shufps(cp->func, dest, src, shuffle); +} + + +static boolean load_input( struct aos_compilation *cp, + unsigned idx, + boolean linear ) +{ + unsigned format = cp->vaos->base.key.element[idx].in.format; + struct x86_reg src = cp->tmp_EAX; + struct x86_reg dataXMM = aos_get_xmm_reg(cp); + + /* Figure out source pointer address: + */ + get_src_ptr(cp->func, + src, + cp->machine_EDX, + linear ? cp->idx_EBX : x86_deref(cp->idx_EBX), + idx); + + src = x86_deref(src); + + aos_adopt_xmm_reg( cp, + dataXMM, + TGSI_FILE_INPUT, + idx, + TRUE ); + + switch (format) { + case PIPE_FORMAT_R32_FLOAT: + emit_load_R32(cp, dataXMM, src); + break; + case PIPE_FORMAT_R32G32_FLOAT: + emit_load_R32G32(cp, dataXMM, src); + break; + case PIPE_FORMAT_R32G32B32_FLOAT: + emit_load_R32G32B32(cp, dataXMM, src); + break; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + emit_load_R32G32B32A32(cp, dataXMM, src); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + emit_load_R8G8B8A8_UNORM(cp, dataXMM, src); + emit_swizzle(cp, dataXMM, dataXMM, SHUF(Z,Y,X,W)); + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + emit_load_R8G8B8A8_UNORM(cp, dataXMM, src); + break; + default: + ERROR(cp, "unhandled input format"); + return FALSE; + } + + return TRUE; +} + + +boolean aos_fetch_inputs( struct aos_compilation *cp, boolean linear ) +{ + unsigned i; + + for (i = 0; i < cp->vaos->base.vs->info.num_inputs; i++) { + if (!load_input( cp, i, linear )) + return FALSE; + cp->insn_counter++; + debug_printf("\n"); + } + + return TRUE; +} + + + + + + + +static void emit_store_R32G32B32A32( struct aos_compilation *cp, + struct x86_reg dst_ptr, + struct x86_reg dataXMM ) +{ + sse_movups(cp->func, dst_ptr, dataXMM); +} + +static void emit_store_R32G32B32( struct aos_compilation *cp, + struct x86_reg dst_ptr, + struct x86_reg dataXMM ) +{ + sse_movlps(cp->func, dst_ptr, dataXMM); + sse_shufps(cp->func, dataXMM, dataXMM, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */ + sse_movss(cp->func, x86_make_disp(dst_ptr,8), dataXMM); +} + +static void emit_store_R32G32( struct aos_compilation *cp, + struct x86_reg dst_ptr, + struct x86_reg dataXMM ) +{ + sse_movlps(cp->func, dst_ptr, dataXMM); +} + +static void emit_store_R32( struct aos_compilation *cp, + struct x86_reg dst_ptr, + struct x86_reg dataXMM ) +{ + sse_movss(cp->func, dst_ptr, dataXMM); +} + + + +static void emit_store_R8G8B8A8_UNORM( struct aos_compilation *cp, + struct x86_reg dst_ptr, + struct x86_reg dataXMM ) +{ + sse_mulps(cp->func, dataXMM, aos_get_internal(cp, IMM_255)); + sse2_cvtps2dq(cp->func, dataXMM, dataXMM); + sse2_packssdw(cp->func, dataXMM, dataXMM); + sse2_packuswb(cp->func, dataXMM, dataXMM); + sse_movss(cp->func, dst_ptr, dataXMM); +} + + + + + +static boolean emit_output( struct aos_compilation *cp, + struct x86_reg ptr, + struct x86_reg dataXMM, + unsigned format ) +{ + switch (format) { + case PIPE_FORMAT_R32_FLOAT: + emit_store_R32(cp, ptr, dataXMM); + break; + case PIPE_FORMAT_R32G32_FLOAT: + emit_store_R32G32(cp, ptr, dataXMM); + break; + case PIPE_FORMAT_R32G32B32_FLOAT: + emit_store_R32G32B32(cp, ptr, dataXMM); + break; + case PIPE_FORMAT_R32G32B32A32_FLOAT: + emit_store_R32G32B32A32(cp, ptr, dataXMM); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + emit_swizzle(cp, dataXMM, dataXMM, SHUF(Z,Y,X,W)); + emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM); + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM); + break; + default: + ERROR(cp, "unhandled output format"); + return FALSE; + } + + return TRUE; +} + + + +boolean aos_emit_outputs( struct aos_compilation *cp ) +{ + unsigned i; + + for (i = 0; i < cp->vaos->base.vs->info.num_inputs; i++) { + unsigned format = cp->vaos->base.key.element[i].out.format; + unsigned offset = cp->vaos->base.key.element[i].out.offset; + + struct x86_reg data = aos_get_shader_reg( cp, + TGSI_FILE_OUTPUT, + i ); + + if (data.file != file_XMM) { + struct x86_reg tmp = aos_get_xmm_reg( cp ); + sse_movups(cp->func, tmp, data); + data = tmp; + } + + if (!emit_output( cp, + x86_make_disp( cp->outbuf_ECX, offset ), + data, + format )) + return FALSE; + + aos_release_xmm_reg( cp, data.idx ); + + cp->insn_counter++; + debug_printf("\n"); + } + + return TRUE; +} + +#endif diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index df94a7e0c7..0581c3042f 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -157,6 +157,7 @@ draw_create_vs_sse(struct draw_context *draw, vs->base.draw = draw; vs->base.create_varient = draw_vs_varient_generic; +// vs->base.create_varient = draw_vs_varient_aos_sse; vs->base.prepare = vs_sse_prepare; vs->base.run_linear = vs_sse_run_linear; vs->base.delete = vs_sse_delete; -- cgit v1.2.3 From 030af06691bc5bc82ca141a576da7a2edffe9d1c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 20:14:55 +0100 Subject: rtasm: add x87 instructions and debug-check for x87 stack usage --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 120 +++++++++++++++++++++++++++++ src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 20 ++++- 2 files changed, 138 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 68ac91ed13..a2e8af343b 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -986,6 +986,26 @@ void sse2_movd( struct x86_function *p, /*********************************************************************** * x87 instructions */ +static void note_x87_pop( struct x86_function *p ) +{ + p->x87_stack--; + assert(p->x87_stack >= 0); + debug_printf("\nstack: %d\n", p->x87_stack); +} + +static void note_x87_push( struct x86_function *p ) +{ + p->x87_stack++; + assert(p->x87_stack <= 7); + debug_printf("\nstack: %d\n", p->x87_stack); +} + +void x87_assert_stack_empty( struct x86_function *p ) +{ + assert (p->x87_stack == 0); +} + + void x87_fist( struct x86_function *p, struct x86_reg dst ) { DUMP_R( dst ); @@ -998,6 +1018,7 @@ void x87_fistp( struct x86_function *p, struct x86_reg dst ) DUMP_R( dst ); emit_1ub(p, 0xdb); emit_modrm_noreg(p, 3, dst); + note_x87_pop(p); } void x87_fild( struct x86_function *p, struct x86_reg arg ) @@ -1005,12 +1026,14 @@ void x87_fild( struct x86_function *p, struct x86_reg arg ) DUMP_R( arg ); emit_1ub(p, 0xdf); emit_modrm_noreg(p, 0, arg); + note_x87_push(p); } void x87_fldz( struct x86_function *p ) { DUMP(); emit_2ub(p, 0xd9, 0xee); + note_x87_push(p); } @@ -1027,18 +1050,21 @@ void x87_fld1( struct x86_function *p ) { DUMP(); emit_2ub(p, 0xd9, 0xe8); + note_x87_push(p); } void x87_fldl2e( struct x86_function *p ) { DUMP(); emit_2ub(p, 0xd9, 0xea); + note_x87_push(p); } void x87_fldln2( struct x86_function *p ) { DUMP(); emit_2ub(p, 0xd9, 0xed); + note_x87_push(p); } void x87_fwait( struct x86_function *p ) @@ -1059,6 +1085,49 @@ void x87_fclex( struct x86_function *p ) x87_fnclex(p); } +void x87_fcmovb( struct x86_function *p, struct x86_reg arg ) +{ + DUMP_R( arg ); + assert(arg.file == file_x87); + emit_2ub(p, 0xda, 0xc0+arg.idx); +} + +void x87_fcmove( struct x86_function *p, struct x86_reg arg ) +{ + DUMP_R( arg ); + assert(arg.file == file_x87); + emit_2ub(p, 0xda, 0xc8+arg.idx); +} + +void x87_fcmovbe( struct x86_function *p, struct x86_reg arg ) +{ + DUMP_R( arg ); + assert(arg.file == file_x87); + emit_2ub(p, 0xda, 0xd0+arg.idx); +} + +void x87_fcmovnb( struct x86_function *p, struct x86_reg arg ) +{ + DUMP_R( arg ); + assert(arg.file == file_x87); + emit_2ub(p, 0xdb, 0xc0+arg.idx); +} + +void x87_fcmovne( struct x86_function *p, struct x86_reg arg ) +{ + DUMP_R( arg ); + assert(arg.file == file_x87); + emit_2ub(p, 0xdb, 0xc8+arg.idx); +} + +void x87_fcmovnbe( struct x86_function *p, struct x86_reg arg ) +{ + DUMP_R( arg ); + assert(arg.file == file_x87); + emit_2ub(p, 0xdb, 0xd0+arg.idx); +} + + static void x87_arith_op( struct x86_function *p, struct x86_reg dst, struct x86_reg arg, unsigned char dst0ub0, @@ -1146,6 +1215,7 @@ void x87_fmulp( struct x86_function *p, struct x86_reg dst ) assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xc8+dst.idx); + note_x87_pop(p); } void x87_fsubp( struct x86_function *p, struct x86_reg dst ) @@ -1154,6 +1224,7 @@ void x87_fsubp( struct x86_function *p, struct x86_reg dst ) assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xe8+dst.idx); + note_x87_pop(p); } void x87_fsubrp( struct x86_function *p, struct x86_reg dst ) @@ -1162,6 +1233,7 @@ void x87_fsubrp( struct x86_function *p, struct x86_reg dst ) assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xe0+dst.idx); + note_x87_pop(p); } void x87_faddp( struct x86_function *p, struct x86_reg dst ) @@ -1170,6 +1242,7 @@ void x87_faddp( struct x86_function *p, struct x86_reg dst ) assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xc0+dst.idx); + note_x87_pop(p); } void x87_fdivp( struct x86_function *p, struct x86_reg dst ) @@ -1178,6 +1251,7 @@ void x87_fdivp( struct x86_function *p, struct x86_reg dst ) assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xf8+dst.idx); + note_x87_pop(p); } void x87_fdivrp( struct x86_function *p, struct x86_reg dst ) @@ -1186,6 +1260,13 @@ void x87_fdivrp( struct x86_function *p, struct x86_reg dst ) assert(dst.file == file_x87); assert(dst.idx >= 1); emit_2ub(p, 0xde, 0xf0+dst.idx); + note_x87_pop(p); +} + +void x87_ftst( struct x86_function *p ) +{ + DUMP(); + emit_2ub(p, 0xd9, 0xe4); } void x87_fucom( struct x86_function *p, struct x86_reg arg ) @@ -1200,12 +1281,15 @@ void x87_fucomp( struct x86_function *p, struct x86_reg arg ) DUMP_R( arg ); assert(arg.file == file_x87); emit_2ub(p, 0xdd, 0xe8+arg.idx); + note_x87_pop(p); } void x87_fucompp( struct x86_function *p ) { DUMP(); emit_2ub(p, 0xda, 0xe9); + note_x87_pop(p); /* pop twice */ + note_x87_pop(p); /* pop twice */ } void x87_fxch( struct x86_function *p, struct x86_reg arg ) @@ -1287,6 +1371,7 @@ void x87_fyl2x( struct x86_function *p ) { DUMP(); emit_2ub(p, 0xd9, 0xf1); + note_x87_pop(p); } /* st1 = st1 * log2(st0 + 1.0); @@ -1298,6 +1383,7 @@ void x87_fyl2xp1( struct x86_function *p ) { DUMP(); emit_2ub(p, 0xd9, 0xf9); + note_x87_pop(p); } @@ -1310,6 +1396,7 @@ void x87_fld( struct x86_function *p, struct x86_reg arg ) emit_1ub(p, 0xd9); emit_modrm_noreg(p, 0, arg); } + note_x87_push(p); } void x87_fst( struct x86_function *p, struct x86_reg dst ) @@ -1332,8 +1419,15 @@ void x87_fstp( struct x86_function *p, struct x86_reg dst ) emit_1ub(p, 0xd9); emit_modrm_noreg(p, 3, dst); } + note_x87_pop(p); +} + +void x87_fpop( struct x86_function *p ) +{ + x87_fstp( p, x86_make_reg( file_x87, 0 )); } + void x87_fcom( struct x86_function *p, struct x86_reg dst ) { DUMP_R( dst ); @@ -1345,6 +1439,7 @@ void x87_fcom( struct x86_function *p, struct x86_reg dst ) } } + void x87_fcomp( struct x86_function *p, struct x86_reg dst ) { DUMP_R( dst ); @@ -1354,6 +1449,20 @@ void x87_fcomp( struct x86_function *p, struct x86_reg dst ) emit_1ub(p, 0xd8); emit_modrm_noreg(p, 3, dst); } + note_x87_pop(p); +} + +void x87_fcomi( struct x86_function *p, struct x86_reg arg ) +{ + DUMP_R( arg ); + emit_2ub(p, 0xdb, 0xf0+arg.idx); +} + +void x87_fcomip( struct x86_function *p, struct x86_reg arg ) +{ + DUMP_R( arg ); + emit_2ub(p, 0xdb, 0xf0+arg.idx); + note_x87_pop(p); } @@ -1372,6 +1481,17 @@ void x87_fnstsw( struct x86_function *p, struct x86_reg dst ) } +void x87_fnstcw( struct x86_function *p, struct x86_reg dst ) +{ + DUMP_R( dst ); + assert(dst.file == file_REG32); + + emit_1ub(p, 0x9b); /* WAIT -- needed? */ + emit_1ub(p, 0xd9); + emit_modrm_noreg(p, 7, dst); +} + + /*********************************************************************** diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index 1e02c6e73b..9f7e31e055 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -41,8 +41,11 @@ struct x86_function { unsigned size; unsigned char *store; unsigned char *csr; - unsigned stack_offset; - int need_emms; + + unsigned stack_offset:16; + unsigned need_emms:8; + int x87_stack:8; + unsigned char error_overflow[4]; }; @@ -229,13 +232,23 @@ void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_sahf( struct x86_function *p ); +void x87_assert_stack_empty( struct x86_function *p ); + void x87_f2xm1( struct x86_function *p ); void x87_fabs( struct x86_function *p ); void x87_fadd( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); void x87_faddp( struct x86_function *p, struct x86_reg dst ); void x87_fchs( struct x86_function *p ); void x87_fclex( struct x86_function *p ); +void x87_fcmovb( struct x86_function *p, struct x86_reg src ); +void x87_fcmovbe( struct x86_function *p, struct x86_reg src ); +void x87_fcmove( struct x86_function *p, struct x86_reg src ); +void x87_fcmovnb( struct x86_function *p, struct x86_reg src ); +void x87_fcmovnbe( struct x86_function *p, struct x86_reg src ); +void x87_fcmovne( struct x86_function *p, struct x86_reg src ); void x87_fcom( struct x86_function *p, struct x86_reg dst ); +void x87_fcomi( struct x86_function *p, struct x86_reg dst ); +void x87_fcomip( struct x86_function *p, struct x86_reg dst ); void x87_fcomp( struct x86_function *p, struct x86_reg dst ); void x87_fcos( struct x86_function *p ); void x87_fdiv( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); @@ -255,6 +268,7 @@ void x87_fmul( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); void x87_fmulp( struct x86_function *p, struct x86_reg dst ); void x87_fnclex( struct x86_function *p ); void x87_fprndint( struct x86_function *p ); +void x87_fpop( struct x86_function *p ); void x87_fscale( struct x86_function *p ); void x87_fsin( struct x86_function *p ); void x87_fsincos( struct x86_function *p ); @@ -265,11 +279,13 @@ void x87_fsub( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); void x87_fsubp( struct x86_function *p, struct x86_reg dst ); void x87_fsubr( struct x86_function *p, struct x86_reg dst, struct x86_reg arg ); void x87_fsubrp( struct x86_function *p, struct x86_reg dst ); +void x87_ftst( struct x86_function *p ); void x87_fxch( struct x86_function *p, struct x86_reg dst ); void x87_fxtract( struct x86_function *p ); void x87_fyl2x( struct x86_function *p ); void x87_fyl2xp1( struct x86_function *p ); void x87_fwait( struct x86_function *p ); +void x87_fnstcw( struct x86_function *p, struct x86_reg dst ); void x87_fnstsw( struct x86_function *p, struct x86_reg dst ); void x87_fucompp( struct x86_function *p ); void x87_fucomp( struct x86_function *p, struct x86_reg arg ); -- cgit v1.2.3 From 889473b3f5a216bd753c357974d6bae29fe3c41d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 20:28:56 +0100 Subject: draw: add viewport to varient state --- .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 9 +++- src/gallium/auxiliary/draw/draw_vs.h | 8 +++- src/gallium/auxiliary/draw/draw_vs_aos.c | 50 ++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs_aos.h | 9 +++- src/gallium/auxiliary/draw/draw_vs_sse.c | 4 +- src/gallium/auxiliary/draw/draw_vs_varient.c | 10 +++++ 6 files changed, 84 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 74945dcfe9..984fbb6767 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -95,10 +95,14 @@ static void fse_prepare( struct draw_pt_middle_end *middle, + fse->key.output_stride = vinfo->size * 4; fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ num_vs_inputs); /* inputs - fetch from api format */ - fse->key.output_stride = vinfo->size * 4; + fse->key.viewport = 1; + fse->key.clip = 0; + fse->key.pad = 0; + memset(fse->key.element, 0, fse->key.nr_elements * sizeof(fse->key.element[0])); @@ -211,6 +215,9 @@ static void fse_prepare( struct draw_pt_middle_end *middle, fse->active->set_constants( fse->active, (const float (*)[4])draw->pt.user.constants ); + fse->active->set_viewport( fse->active, + &draw->viewport ); + //return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 5a8d0da06d..ff3e19b2a8 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -58,7 +58,10 @@ struct draw_vs_element { struct draw_vs_varient_key { unsigned output_stride; - unsigned nr_elements; + unsigned nr_elements:16; + unsigned viewport:1; + unsigned clip:1; + unsigned pad:14; struct draw_vs_element element[PIPE_MAX_ATTRIBS]; }; @@ -88,6 +91,9 @@ struct draw_vs_varient { void (*set_constants)( struct draw_vs_varient *, const float (*constants)[4] ); + void (*set_viewport)( struct draw_vs_varient *, + const struct pipe_viewport_state * ); + void (PIPE_CDECL *run_linear)( struct draw_vs_varient *shader, unsigned start, unsigned count, diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 620f5e3592..b8e66e8b78 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1401,6 +1401,37 @@ emit_instruction( struct aos_compilation *cp, } } + +static boolean emit_viewport( struct aos_compilation *cp ) +{ + struct x86_reg pos = aos_get_shader_reg(cp, + TGSI_FILE_OUTPUT, + 0); + + struct x86_reg scale = x86_make_disp(cp->machine_EDX, + Offset(struct aos_machine, scale)); + + struct x86_reg translate = x86_make_disp(cp->machine_EDX, + Offset(struct aos_machine, translate)); + + if (pos.file != file_XMM) { + struct x86_reg dst = aos_get_xmm_reg(cp); + sse_movups(cp->func, dst, pos); + pos = dst; + } + + sse_mulps(cp->func, pos, scale); + sse_addps(cp->func, pos, translate); + + aos_adopt_xmm_reg( cp, + pos, + TGSI_FILE_OUTPUT, + 0, + TRUE ); + return TRUE; +} + + static boolean note_immediate( struct aos_compilation *cp, struct tgsi_full_immediate *imm ) { @@ -1540,6 +1571,10 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, if (cp.error) goto fail; + if (cp.vaos->base.key.viewport) { + emit_viewport(&cp); + } + /* Emit output... TODO: do this eagerly after the last write to a * given output. */ @@ -1665,11 +1700,25 @@ static void vaos_set_constants( struct draw_vs_varient *varient, } +static void vaos_set_viewport( struct draw_vs_varient *varient, + const struct pipe_viewport_state *viewport ) +{ + struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + + memcpy(vaos->machine->scale, viewport->scale, 4 * sizeof(float)); + memcpy(vaos->machine->translate, viewport->translate, 4 * sizeof(float)); +} + + + static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, const struct draw_vs_varient_key *key ) { struct draw_vs_varient_aos_sse *vaos = CALLOC_STRUCT(draw_vs_varient_aos_sse); + if (key->clip) + return NULL; + if (!vaos) goto fail; @@ -1677,6 +1726,7 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, vaos->base.vs = vs; vaos->base.set_input = vaos_set_buffer; vaos->base.set_constants = vaos_set_constants; + vaos->base.set_viewport = vaos_set_viewport; vaos->base.destroy = vaos_destroy; vaos->base.run_linear = vaos_run_linear; vaos->base.run_elts = vaos_run_elts; diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index 1d8a055a90..16fef6451c 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -68,8 +68,13 @@ struct aos_machine { float immediate[MAX_IMMEDIATES][4]; /* fixme -- should just be a pointer */ float internal [MAX_INTERNALS ][4]; - unsigned fpu_round_nearest; - unsigned fpu_round_neg_inf; + float scale[4]; /* viewport */ + float translate[4]; /* viewport */ + + ushort fpu_round_nearest; + ushort fpu_round_neg_inf; + ushort fpu_restore; + ushort fpucntl; /* one of FPU_* above */ struct { const void *input_ptr; diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 0581c3042f..7781782ae8 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -156,8 +156,8 @@ draw_create_vs_sse(struct draw_context *draw, tgsi_scan_shader(templ->tokens, &vs->base.info); vs->base.draw = draw; - vs->base.create_varient = draw_vs_varient_generic; -// vs->base.create_varient = draw_vs_varient_aos_sse; + vs->base.create_varient = draw_vs_varient_aos_sse; +// vs->base.create_varient = draw_vs_varient_generic; vs->base.prepare = vs_sse_prepare; vs->base.run_linear = vs_sse_run_linear; vs->base.delete = vs_sse_delete; diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index d27b0f6187..f6f621a748 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -167,6 +167,12 @@ static void vsvg_run_linear( struct draw_vs_varient *varient, + +static void vsvg_set_viewport( struct draw_vs_varient *varient, + const struct pipe_viewport_state *viewport ) +{ +} + static void vsvg_destroy( struct draw_vs_varient *varient ) { FREE(varient); @@ -179,6 +185,9 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, unsigned i; struct translate_key fetch, emit; + if (key->viewport || key->clip) + return NULL; + struct draw_vs_varient_generic *vsvg = CALLOC_STRUCT( draw_vs_varient_generic ); if (vsvg == NULL) return NULL; @@ -187,6 +196,7 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, vsvg->base.vs = vs; vsvg->base.set_input = vsvg_set_input; vsvg->base.set_constants = vsvg_set_constants; + vsvg->base.set_viewport = vsvg_set_viewport; vsvg->base.run_elts = vsvg_run_elts; vsvg->base.run_linear = vsvg_run_linear; vsvg->base.destroy = vsvg_destroy; -- cgit v1.2.3 From 194a7be28f6eed502f2475d9a637cb3610ca75f6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 20:31:08 +0100 Subject: draw: fix vs aos internal/machine state --- src/gallium/auxiliary/draw/draw_vs_aos.c | 59 ++++++++++++++++++++++++++++++-- src/gallium/auxiliary/draw/draw_vs_aos.h | 9 +++-- 2 files changed, 63 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index b8e66e8b78..67761f881d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -83,7 +83,7 @@ static struct x86_reg get_reg_ptr(struct aos_compilation *cp, return x86_make_disp(ptr, Offset(struct aos_machine, constant[idx])); case AOS_FILE_INTERNAL: - return x86_make_disp(ptr, Offset(struct aos_machine, immediate[idx])); + return x86_make_disp(ptr, Offset(struct aos_machine, internal[idx])); default: ERROR(cp, "unknown reg file"); @@ -97,9 +97,63 @@ struct x86_reg aos_get_internal( struct aos_compilation *cp, { return get_reg_ptr( cp, AOS_FILE_INTERNAL, - imm + 1 ); + imm ); +} + +#define X87_CW_EXCEPTION_INV_OP (1<<0) +#define X87_CW_EXCEPTION_DENORM_OP (1<<1) +#define X87_CW_EXCEPTION_ZERO_DIVIDE (1<<2) +#define X87_CW_EXCEPTION_OVERFLOW (1<<3) +#define X87_CW_EXCEPTION_UNDERFLOW (1<<4) +#define X87_CW_EXCEPTION_PRECISION (1<<5) +#define X87_CW_PRECISION_SINGLE (0<<8) +#define X87_CW_PRECISION_RESERVED (1<<8) +#define X87_CW_PRECISION_DOUBLE (2<<8) +#define X87_CW_PRECISION_DOUBLE_EXT (3<<8) +#define X87_CW_PRECISION_MASK (3<<8) +#define X87_CW_ROUND_NEAREST (0<<10) +#define X87_CW_ROUND_DOWN (1<<10) +#define X87_CW_ROUND_UP (2<<10) +#define X87_CW_ROUND_ZERO (3<<10) +#define X87_CW_ROUND_MASK (3<<10) +#define X87_CW_INFINITY (1<<12) + +static void init_internals( struct aos_machine *machine ) +{ + float inv = 1.0f/255.0f; + float f255 = 255.0f; + + ASSIGN_4V(machine->internal[IMM_ONES], 1.0f, 1.0f, 1.0f, 1.0f); + ASSIGN_4V(machine->internal[IMM_NEGS], -1.0f, -1.0f, -1.0f, -1.0f); + ASSIGN_4V(machine->internal[IMM_IDENTITY], 0.0f, 0.0f, 0.0f, 1.0f); + ASSIGN_4V(machine->internal[IMM_INV_255], inv, inv, inv, inv); + ASSIGN_4V(machine->internal[IMM_255], f255, f255, f255, f255); + + + machine->fpu_rnd_nearest = (X87_CW_EXCEPTION_INV_OP | + X87_CW_EXCEPTION_DENORM_OP | + X87_CW_EXCEPTION_ZERO_DIVIDE | + X87_CW_EXCEPTION_OVERFLOW | + X87_CW_EXCEPTION_UNDERFLOW | + X87_CW_EXCEPTION_PRECISION | + (1<<6) | + X87_CW_ROUND_NEAREST | + X87_CW_PRECISION_DOUBLE_EXT); + + assert(machine->fpu_rnd_nearest == 0x37f); + + machine->fpu_rnd_neg_inf = (X87_CW_EXCEPTION_INV_OP | + X87_CW_EXCEPTION_DENORM_OP | + X87_CW_EXCEPTION_ZERO_DIVIDE | + X87_CW_EXCEPTION_OVERFLOW | + X87_CW_EXCEPTION_UNDERFLOW | + X87_CW_EXCEPTION_PRECISION | + (1<<6) | + X87_CW_ROUND_DOWN | + X87_CW_PRECISION_DOUBLE_EXT); } + static void spill( struct aos_compilation *cp, unsigned idx ) { if (!cp->xmm[idx].dirty || @@ -1736,6 +1790,7 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, goto fail; memset(vaos->machine, 0, sizeof(struct aos_machine)); + init_internals(vaos->machine); tgsi_dump(vs->state.tokens, 0); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index 16fef6451c..c2afd4e9a0 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -52,10 +52,13 @@ struct x86_function; #define MAX_TEMPS PIPE_MAX_ATTRIBS /* say */ #define MAX_CONSTANTS PIPE_MAX_ATTRIBS /* say */ #define MAX_IMMEDIATES PIPE_MAX_ATTRIBS /* say */ -#define MAX_INTERNALS 4 +#define MAX_INTERNALS 8 #define AOS_FILE_INTERNAL TGSI_FILE_COUNT +#define FPU_RND_NEG 1 +#define FPU_RND_NEAREST 2 + /* This is the temporary storage used by all the aos_sse vs varients. * Create one per context and reuse by passing a pointer in at * vs_varient creation?? @@ -71,8 +74,8 @@ struct aos_machine { float scale[4]; /* viewport */ float translate[4]; /* viewport */ - ushort fpu_round_nearest; - ushort fpu_round_neg_inf; + ushort fpu_rnd_nearest; + ushort fpu_rnd_neg_inf; ushort fpu_restore; ushort fpucntl; /* one of FPU_* above */ -- cgit v1.2.3 From 2302a5d3c1ea2c682dfc034012a054b8327a81de Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 20:32:43 +0100 Subject: draw: fix fpu control word manipulations --- src/gallium/auxiliary/draw/draw_vs_aos.c | 33 +++++++++++++++++++------------- 1 file changed, 20 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 67761f881d..e736990acc 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -690,41 +690,47 @@ static void x87_fstp_dest4( struct aos_compilation *cp, x87_fstp_or_pop(cp->func, writemask, 3, ptr); } +#define FPU_MANIP 1 /* Save current x87 state and put it into single precision mode. */ static void save_fpu_state( struct aos_compilation *cp ) { -#if 0 - x87_fnstcw( cp->func, x86_make_disp(regEDX, get_offset(m, &m->fpucntl_restore))); - x87_fldcw( cp->func, ); +#if FPU_MANIP + x87_fnstcw( cp->func, x86_make_disp(cp->machine_EDX, + Offset(struct aos_machine, fpu_restore))); #endif } static void restore_fpu_state( struct aos_compilation *cp ) { -#if 0 +#if FPU_MANIP x87_fnclex(cp->func); - x87_fldcw(cp->func, x86_make_disp(regEDX, get_offset(m, &m->fpucntl_restore))); + x87_fldcw( cp->func, x86_make_disp(cp->machine_EDX, + Offset(struct aos_machine, fpu_restore))); #endif } static void set_fpu_round_neg_inf( struct aos_compilation *cp ) { -#if 0 - if (cp->fpucntl != RND_NEG_FPU) { - struct x86_reg regEDX = x86_make_reg(file_REG32, reg_DX); - struct arb_vp_machine *m = NULL; - - cp->fpucntl = RND_NEG_FPU; +#if FPU_MANIP + if (cp->fpucntl != FPU_RND_NEG) { + cp->fpucntl = FPU_RND_NEG; x87_fnclex(cp->func); - x87_fldcw(cp->func, x86_make_disp(regEDX, get_offset(m, &m->fpucntl_rnd_neg))); + x87_fldcw( cp->func, x86_make_disp(cp->machine_EDX, + Offset(struct aos_machine, fpu_rnd_neg_inf))); } #endif } static void set_fpu_round_nearest( struct aos_compilation *cp ) { -#if 0 +#if FPU_MANIP + if (cp->fpucntl != FPU_RND_NEAREST) { + cp->fpucntl = FPU_RND_NEAREST; + x87_fnclex(cp->func); + x87_fldcw( cp->func, x86_make_disp(cp->machine_EDX, + Offset(struct aos_machine, fpu_rnd_nearest))); + } #endif } @@ -1590,6 +1596,7 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, Offset( struct draw_vs_varient_aos_sse, machine ))); save_fpu_state( &cp ); + set_fpu_round_nearest( &cp ); /* Note address for loop jump */ -- cgit v1.2.3 From 0a7a0d79f64de9794878c42bc5b79a04772d7ed8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 20:34:52 +0100 Subject: draw: fix x87_ex2 and partially fix lit insn --- src/gallium/auxiliary/draw/draw_vs_aos.c | 112 +++++++++++++++---------------- 1 file changed, 56 insertions(+), 56 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index e736990acc..a365d456d1 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -735,23 +735,26 @@ static void set_fpu_round_nearest( struct aos_compilation *cp ) } -static void emit_x87_ex2( struct aos_compilation *cp ) +static void x87_emit_ex2( struct aos_compilation *cp ) { struct x86_reg st0 = x86_make_reg(file_x87, 0); struct x86_reg st1 = x86_make_reg(file_x87, 1); - struct x86_reg st3 = x86_make_reg(file_x87, 3); + int stack = cp->func->x87_stack; set_fpu_round_neg_inf( cp ); - x87_fld(cp->func, st0); /* a a */ - x87_fprndint( cp->func ); /* int(a) a */ - x87_fld(cp->func, st0); /* int(a) int(a) a */ - x87_fstp(cp->func, st3); /* int(a) a int(a)*/ - x87_fsubp(cp->func, st1); /* frac(a) int(a) */ - x87_f2xm1(cp->func); /* (2^frac(a))-1 int(a)*/ - x87_fld1(cp->func); /* 1 (2^frac(a))-1 int(a)*/ - x87_faddp(cp->func, st1); /* 2^frac(a) int(a) */ - x87_fscale(cp->func); /* 2^a */ + x87_fld(cp->func, st0); /* a a */ + x87_fld(cp->func, st0); /* a a a */ + x87_fprndint( cp->func ); /* flr(a) a a*/ + x87_fsubp(cp->func, st1); /* frac(a) a */ + x87_f2xm1(cp->func); /* (2^frac(a))-1 a */ + x87_fld1(cp->func); /* 1 (2^frac(a))-1 a */ + x87_faddp(cp->func, st1); /* 2^frac(a) a */ + x87_fscale(cp->func); /* 2^a a */ + x87_fstp(cp->func, st1); + + assert( stack == cp->func->x87_stack); + } @@ -907,9 +910,7 @@ static boolean emit_LG2( struct aos_compilation *cp, const struct tgsi_full_inst static boolean emit_EX2( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { x87_fld_src(cp, &op->FullSrcRegisters[0], 0); - - emit_x87_ex2(cp); - + x87_emit_ex2(cp); x87_fstp_dest4(cp, &op->FullDstRegisters[0]); return TRUE; } @@ -1084,63 +1085,62 @@ static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_inst static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); - struct x86_reg st1 = x86_make_reg(file_x87, 1); - unsigned fixup1, fixup2; unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; - /* Load the interesting parts of arg0: - */ - x87_fld_src(cp, &op->FullSrcRegisters[0], 3); - x87_fld_src(cp, &op->FullSrcRegisters[0], 1); - x87_fld_src(cp, &op->FullSrcRegisters[0], 0); - - if (writemask & TGSI_WRITEMASK_XW) { - x87_fld1(cp->func); - x87_fst_or_nop(cp->func, writemask, 0, dst); - x87_fstp_or_pop(cp->func, writemask, 3, dst); - } if (writemask & TGSI_WRITEMASK_YZ) { + struct x86_reg st1 = x86_make_reg(file_x87, 1); + struct x86_reg st2 = x86_make_reg(file_x87, 2); + + - /* Pre-zero destinations, may be overwritten later... fixme. - */ - x87_fldz(cp->func); - x87_fst_or_nop(cp->func, writemask, 1, dst); - x87_fstp_or_pop(cp->func, writemask, 2, dst); + /* a1' = a1 <= 0 ? 1 : a1; + */ + x87_fldz(cp->func); /* 0 */ + x87_fld1(cp->func); /* 1 0 */ + x87_fld_src(cp, &op->FullSrcRegisters[0], 1); /* a1 1 0 */ + x87_fcomi(cp->func, st2); /* a1 1 0 */ + x87_fcmovb(cp->func, st1); /* a1' 1 0 */ + x87_fstp(cp->func, st1); /* a1' 0 */ + x87_fstp(cp->func, st1); /* a1' */ + + x87_fld_src(cp, &op->FullSrcRegisters[0], 3); /* a3 a1' */ + x87_fxch(cp->func, st1); /* a1' a3 */ + - /* Check arg0[0]: + /* Compute pow(a1, a3) */ - x87_fldz(cp->func); /* 0 a0 a1 a3 */ - x87_fucomp(cp->func, st1); /* a0 a1 a3 */ - x87_fnstsw(cp->func, cp->tmp_EAX); - x86_sahf(cp->func); - fixup1 = x86_jcc_forward(cp->func, cc_AE); - - x87_fstp_or_pop(cp->func, writemask, 1, dst); /* a1 a3 */ + x87_fyl2x(cp->func); /* a3*log2(a1) */ + x87_emit_ex2( cp ); /* 2^(a3*log2(a1)) */ - /* Check arg0[1]: - */ - x87_fldz(cp->func); /* 0 a1 a3 */ - x87_fucomp(cp->func, st1); /* a1 a3 */ - x87_fnstsw(cp->func, cp->tmp_EAX); - x86_sahf(cp->func); - fixup2 = x86_jcc_forward(cp->func, cc_AE); - /* Compute pow(a1, a3) + /* a0' = max2(a0, 0): */ - x87_fyl2x(cp->func); /* a3*log2(a1) */ + x87_fldz(cp->func); /* 0 r2 */ + x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* a0 0 r2 */ + x87_fcomi(cp->func, st1); + x87_fcmovb(cp->func, st1); /* a0' 0 r2 */ + x87_fstp(cp->func, st1); /* a0' r2 */ - emit_x87_ex2( cp ); /* 2^(a3*log2(a1)) */ + x87_fxch(cp->func, st1); /* a0' r2 */ + x87_fst_or_nop(cp->func, writemask, 1, dst); /* result[1] = a0' */ + + x87_fldz(cp->func); /* 0 a0' r2 */ + x87_fcomi(cp->func, st1); /* 0 a0' r2 */ + x87_fcmovnbe(cp->func, st2); /* r2' a0' r2 */ x87_fstp_or_pop(cp->func, writemask, 2, dst); - - /* Land jumps: - */ - x86_fixup_fwd_jump(cp->func, fixup1); - x86_fixup_fwd_jump(cp->func, fixup2); + x87_fpop(cp->func); + x87_fpop(cp->func); + } + + if (writemask & TGSI_WRITEMASK_XW) { + x87_fld1(cp->func); + x87_fst_or_nop(cp->func, writemask, 0, dst); + x87_fstp_or_pop(cp->func, writemask, 3, dst); } return TRUE; @@ -1222,7 +1222,7 @@ static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_inst x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* a0.x a1.x */ x87_fyl2x(cp->func); /* a1*log2(a0) */ - emit_x87_ex2( cp ); /* 2^(a1*log2(a0)) */ + x87_emit_ex2( cp ); /* 2^(a1*log2(a0)) */ x87_fstp_dest4(cp, &op->FullDstRegisters[0]); return TRUE; -- cgit v1.2.3 From 083f3f5c32a28d2993a8a5a8b4f5ef81224a5ec3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 20:38:23 +0100 Subject: draw: avoid a pointless mov in many sse opcodes --- src/gallium/auxiliary/draw/draw_vs_aos.c | 94 +++++++++++++++++++++----------- 1 file changed, 63 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index a365d456d1..97de43c232 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -174,14 +174,44 @@ static void spill( struct aos_compilation *cp, unsigned idx ) } } +static boolean is_xmm_tmp( struct aos_compilation *cp, + struct x86_reg reg ) +{ + return (reg.file == file_XMM && + cp->xmm[reg.idx].file == TGSI_FILE_NULL); +} + +static struct x86_reg get_xmm_tmp( struct aos_compilation *cp, + struct x86_reg reg ) +{ + if (!is_xmm_tmp(cp, reg)) { + struct x86_reg tmp = aos_get_xmm_reg(cp); + sse_movups(cp->func, tmp, reg); + reg = tmp; + } + + return reg; +} + + struct x86_reg aos_get_xmm_reg( struct aos_compilation *cp ) { unsigned i; unsigned oldest = 0; + boolean found = FALSE; for (i = 0; i < 8; i++) - if (cp->xmm[i].last_used < cp->xmm[oldest].last_used) + if (cp->xmm[i].last_used != cp->insn_counter && + cp->xmm[i].file == TGSI_FILE_NULL) { oldest = i; + found = TRUE; + } + + if (!found) { + for (i = 0; i < 8; i++) + if (cp->xmm[i].last_used < cp->xmm[oldest].last_used) + oldest = i; + } /* Need to write out the old value? */ @@ -237,15 +267,24 @@ void aos_adopt_xmm_reg( struct aos_compilation *cp, unsigned idx, unsigned dirty ) { + unsigned i; + if (reg.file != file_XMM) { assert(0); return; } - invalidate_xmm(cp, file, idx); + for (i = 0; i < 8; i++) { + if (cp->xmm[i].file == file && + cp->xmm[i].idx == idx) { + aos_release_xmm_reg(cp, i); + } + } + cp->xmm[reg.idx].file = file; cp->xmm[reg.idx].idx = idx; cp->xmm[reg.idx].dirty = dirty; + cp->xmm[reg.idx].last_used = cp->insn_counter; } @@ -659,6 +698,7 @@ static void x87_fst_or_nop( struct x86_function *func, unsigned channel, struct x86_reg ptr ) { + assert(ptr.file == file_REG32); if (writemask & (1<FullSrcRegisters[0]); - struct x86_reg dst = aos_get_xmm_reg(cp); struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_mulps(cp->func, dst, neg); sse_maxps(cp->func, dst, arg0); @@ -782,9 +822,8 @@ static boolean emit_ADD( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_addps(cp->func, dst, arg1); store_dest(cp, &op->FullDstRegisters[0], dst); @@ -806,10 +845,9 @@ static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); struct x86_reg tmp = aos_get_xmm_reg(cp); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_mulps(cp->func, dst, arg1); /* Now the hard bit: sum the first 3 values: @@ -831,10 +869,9 @@ static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); struct x86_reg tmp = aos_get_xmm_reg(cp); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_mulps(cp->func, dst, arg1); /* Now the hard bit: sum the values: @@ -854,10 +891,9 @@ static boolean emit_DPH( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); struct x86_reg tmp = aos_get_xmm_reg(cp); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_mulps(cp->func, dst, arg1); /* Now the hard bit: sum the values (from DP3): @@ -1152,9 +1188,8 @@ static boolean emit_MAX( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_maxps(cp->func, dst, arg1); store_dest(cp, &op->FullDstRegisters[0], dst); @@ -1166,9 +1201,8 @@ static boolean emit_MIN( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_minps(cp->func, dst, arg1); store_dest(cp, &op->FullDstRegisters[0], dst); @@ -1178,9 +1212,9 @@ static boolean emit_MIN( struct aos_compilation *cp, const struct tgsi_full_inst static boolean emit_MOV( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); - struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); + /* potentially nothing to do */ store_dest(cp, &op->FullDstRegisters[0], dst); return TRUE; @@ -1190,9 +1224,8 @@ static boolean emit_MUL( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_mulps(cp->func, dst, arg1); store_dest(cp, &op->FullDstRegisters[0], dst); @@ -1205,13 +1238,15 @@ static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg arg2 = fetch_src(cp, &op->FullSrcRegisters[2]); - struct x86_reg dst = aos_get_xmm_reg(cp); - sse_movups(cp->func, dst, arg0); - sse_mulps(cp->func, dst, arg1); - sse_addps(cp->func, dst, arg2); + /* If we can't clobber old contents of arg0, get a temporary & copy + * it there, then clobber it... + */ + arg0 = get_xmm_tmp(cp, arg0); - store_dest(cp, &op->FullDstRegisters[0], dst); + sse_mulps(cp->func, arg0, arg1); + sse_addps(cp->func, arg0, arg2); + store_dest(cp, &op->FullDstRegisters[0], arg0); return TRUE; } @@ -1272,10 +1307,9 @@ static boolean emit_SGE( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); struct x86_reg ones = aos_get_internal(cp, IMM_ONES); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_cmpps(cp->func, dst, arg1, cc_NotLessThan); sse_andps(cp->func, dst, ones); @@ -1297,10 +1331,9 @@ static boolean emit_SLT( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); struct x86_reg ones = aos_get_internal(cp, IMM_ONES); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_cmpps(cp->func, dst, arg1, cc_LessThan); sse_andps(cp->func, dst, ones); @@ -1312,9 +1345,8 @@ static boolean emit_SUB( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg dst = get_xmm_tmp(cp, arg0); - sse_movups(cp->func, dst, arg0); sse_subps(cp->func, dst, arg1); store_dest(cp, &op->FullDstRegisters[0], dst); -- cgit v1.2.3 From 5b1bd30f22ffa3955150ec008631d0f4754d340f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 20:41:01 +0100 Subject: draw: when preloading args to x87 stack, need to use reverse order --- src/gallium/auxiliary/draw/draw_vs_aos.c | 18 +++++++++++------- 1 file changed, 11 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 97de43c232..fde92c7226 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -980,7 +980,7 @@ static boolean emit_EXP( struct aos_compilation *cp, const struct tgsi_full_inst x87_fstp_or_pop(cp->func, writemask, 0, dst); /* flr(a) a 2^flr(a) */ - x87_fsubrp(cp->func, st1); /* frac(a) 2^flr(a) */ + x87_fsubp(cp->func, st1); /* frac(a) 2^flr(a) */ x87_fst_or_nop(cp->func, writemask, 1, dst); /* frac(a) 2^flr(a) */ @@ -1041,9 +1041,9 @@ static boolean emit_FLR( struct aos_compilation *cp, const struct tgsi_full_inst /* Load all sources first to avoid aliasing */ - for (i = 0; i < 4; i++) { + for (i = 3; i >= 0; i--) { if (writemask & (1<FullSrcRegisters[0], i); + x87_fld_src(cp, &op->FullSrcRegisters[0], i); } } @@ -1068,9 +1068,9 @@ static boolean emit_RND( struct aos_compilation *cp, const struct tgsi_full_inst /* Load all sources first to avoid aliasing */ - for (i = 0; i < 4; i++) { + for (i = 3; i >= 0; i--) { if (writemask & (1<FullSrcRegisters[0], i); + x87_fld_src(cp, &op->FullSrcRegisters[0], i); } } @@ -1098,7 +1098,7 @@ static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_inst /* suck all the source values onto the stack before writing out any * dst, which may alias... */ - for (i = 0; i < 4; i++) { + for (i = 3; i >= 0; i--) { if (writemask & (1<FullSrcRegisters[0], i); } @@ -1108,7 +1108,7 @@ static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_inst if (writemask & (1<func, st0); /* a a */ x87_fprndint( cp->func ); /* flr(a) a */ - x87_fsubrp(cp->func, st1); /* frc(a) */ + x87_fsubp(cp->func, st1); /* frc(a) */ x87_fstp(cp->func, x86_make_disp(dst, i*4)); } } @@ -1392,6 +1392,8 @@ static boolean emit_instruction( struct aos_compilation *cp, struct tgsi_full_instruction *inst ) { + x87_assert_stack_empty(cp->func); + switch( inst->Instruction.Opcode ) { case TGSI_OPCODE_MOV: return emit_MOV( cp, inst ); @@ -1657,6 +1659,7 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, break; } + x87_assert_stack_empty(cp.func); cp.insn_counter++; debug_printf("\n"); } @@ -1712,6 +1715,7 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, x86_pop(cp.func, cp.count_ESI); x86_pop(cp.func, cp.idx_EBX); + x87_assert_stack_empty(cp.func); x86_ret(cp.func); tgsi_parse_free( &parse ); -- cgit v1.2.3 From 6f407b072453eb2bb7077a952257a099db4da025 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 21 May 2008 20:50:36 +0100 Subject: rtasm: remove debug --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index a2e8af343b..d78676b8f3 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -990,14 +990,12 @@ static void note_x87_pop( struct x86_function *p ) { p->x87_stack--; assert(p->x87_stack >= 0); - debug_printf("\nstack: %d\n", p->x87_stack); } static void note_x87_push( struct x86_function *p ) { p->x87_stack++; assert(p->x87_stack <= 7); - debug_printf("\nstack: %d\n", p->x87_stack); } void x87_assert_stack_empty( struct x86_function *p ) -- cgit v1.2.3 From a5c3b499fa40f46298389900e74f1db04f99166a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 22 May 2008 13:37:48 +0100 Subject: draw: fse works with elts, remove assert --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c | 10 ++-------- 1 file changed, 2 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 984fbb6767..7fefd391a6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -78,11 +78,6 @@ static void fse_prepare( struct draw_pt_middle_end *middle, boolean need_psize = 0; - if (draw->pt.user.elts) { - assert(0); - return ; - } - if (!draw->render->set_primitive( draw->render, prim )) { assert(0); @@ -250,9 +245,8 @@ static void fse_run_linear( struct draw_pt_middle_end *middle, } /* Single routine to fetch vertices, run shader and emit HW verts. - * Clipping and viewport transformation are done elsewhere -- - * either by the API or on hardware, or for some other reason not - * required... + * Clipping is done elsewhere -- either by the API or on hardware, + * or for some other reason not required... */ fse->active->run_linear( fse->active, start, count, -- cgit v1.2.3 From c684ffa02d8d43ee04b99ee63ccd1adb66e81c1a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 22 May 2008 13:41:49 +0100 Subject: draw: clean up internal immediates in aos sse --- src/gallium/auxiliary/draw/draw_vs_aos.c | 64 +++++++++++++++++++++-------- src/gallium/auxiliary/draw/draw_vs_aos.h | 5 ++- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 10 ++--- 3 files changed, 55 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index fde92c7226..0b8600696a 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -44,12 +44,6 @@ #ifdef PIPE_ARCH_X86 -#define DISASSEM 0 - - - - - static INLINE boolean eq( struct x86_reg a, struct x86_reg b ) { @@ -92,13 +86,6 @@ static struct x86_reg get_reg_ptr(struct aos_compilation *cp, } -struct x86_reg aos_get_internal( struct aos_compilation *cp, - unsigned imm ) -{ - return get_reg_ptr( cp, - AOS_FILE_INTERNAL, - imm ); -} #define X87_CW_EXCEPTION_INV_OP (1<<0) #define X87_CW_EXCEPTION_DENORM_OP (1<<1) @@ -123,6 +110,9 @@ static void init_internals( struct aos_machine *machine ) float inv = 1.0f/255.0f; float f255 = 255.0f; + ASSIGN_4V(machine->internal[IMM_SWZ], 1.0f, -1.0f, 0.0f, 1.0f); + *(unsigned *)&machine->internal[IMM_SWZ][3] = 0xffffffff; + ASSIGN_4V(machine->internal[IMM_ONES], 1.0f, 1.0f, 1.0f, 1.0f); ASSIGN_4V(machine->internal[IMM_NEGS], -1.0f, -1.0f, -1.0f, -1.0f); ASSIGN_4V(machine->internal[IMM_IDENTITY], 0.0f, 0.0f, 0.0f, 1.0f); @@ -337,6 +327,39 @@ struct x86_reg aos_get_shader_reg( struct aos_compilation *cp, +static struct x86_reg aos_get_shader_reg_xmm( struct aos_compilation *cp, + unsigned file, + unsigned idx ) +{ + struct x86_reg reg = aos_get_shader_reg( cp, file, idx ); + + if (reg.file != file_XMM) { + struct x86_reg tmp = aos_get_xmm_reg(cp); + sse_movups(cp->func, tmp, reg); + aos_adopt_xmm_reg( cp, tmp, file, idx, FALSE ); + reg = tmp; + } + + return reg; +} + + + +struct x86_reg aos_get_internal_xmm( struct aos_compilation *cp, + unsigned imm ) +{ + return aos_get_shader_reg_xmm( cp, AOS_FILE_INTERNAL, imm ); +} + + +struct x86_reg aos_get_internal( struct aos_compilation *cp, + unsigned imm ) +{ + return aos_get_shader_reg( cp, AOS_FILE_INTERNAL, imm ); +} + + + /* Emulate pshufd insn in regular SSE, if necessary: @@ -461,15 +484,15 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, arg0 = dst; } - if (negs) { - struct x86_reg imm_negs = aos_get_internal(cp, IMM_NEGS); + if (negs && negs != 0xf) { + struct x86_reg imm_swz = aos_get_internal_xmm(cp, IMM_SWZ); struct x86_reg tmp = aos_get_xmm_reg(cp); /* Load 1,-1,0,0 * Use neg as arg to pshufd * Multiply */ - emit_pshufd(cp, tmp, imm_negs, + emit_pshufd(cp, tmp, imm_swz, SHUF((negs & 1) ? 1 : 0, (negs & 2) ? 1 : 0, (negs & 4) ? 1 : 0, @@ -479,12 +502,17 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, aos_release_xmm_reg(cp, tmp.idx); arg0 = dst; } + else if (negs) { + struct x86_reg imm_negs = aos_get_internal_xmm(cp, IMM_NEGS); + sse_mulps(cp->func, dst, imm_negs); + arg0 = dst; + } + if (abs && abs != 0xf) { ERROR(cp, "unsupported partial abs"); } - - if (abs) { + else if (abs) { struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); struct x86_reg tmp = aos_get_xmm_reg(cp); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index c2afd4e9a0..efdc9a38f4 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -142,13 +142,16 @@ boolean aos_emit_outputs( struct aos_compilation *cp ); #define IMM_ONES 0 /* 1, 1,1,1 */ -#define IMM_NEGS 1 /* 1,-1,0,0 */ +#define IMM_SWZ 1 /* 1,-1,0, 0xffffffff */ #define IMM_IDENTITY 2 /* 0, 0,0,1 */ #define IMM_INV_255 3 /* 1/255, 1/255, 1/255, 1/255 */ #define IMM_255 4 /* 255, 255, 255, 255 */ +#define IMM_NEGS 5 /* -1,-1,-1,-1 */ struct x86_reg aos_get_internal( struct aos_compilation *cp, unsigned imm ); +struct x86_reg aos_get_internal_xmm( struct aos_compilation *cp, + unsigned imm ); #define ERROR(cp, msg) \ diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index 72b2b3d11d..0dda9df97d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -54,7 +54,7 @@ static void emit_load_R32G32B32( struct aos_compilation *cp, struct x86_reg src_ptr ) { sse_movss(cp->func, data, x86_make_disp(src_ptr, 8)); - sse_shufps(cp->func, data, aos_get_internal( cp, IMM_IDENTITY ), SHUF(X,Y,Z,W) ); + sse_shufps(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY ), SHUF(X,Y,Z,W) ); sse_shufps(cp->func, data, data, SHUF(Y,Z,X,W) ); sse_movlps(cp->func, data, src_ptr); } @@ -63,7 +63,7 @@ static void emit_load_R32G32( struct aos_compilation *cp, struct x86_reg data, struct x86_reg src_ptr ) { - sse_movups(cp->func, data, aos_get_internal( cp, IMM_IDENTITY ) ); + sse_movups(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY ) ); sse_movlps(cp->func, data, src_ptr); } @@ -73,7 +73,7 @@ static void emit_load_R32( struct aos_compilation *cp, struct x86_reg src_ptr ) { sse_movss(cp->func, data, src_ptr); - sse_orps(cp->func, data, aos_get_internal( cp, IMM_IDENTITY ) ); + sse_orps(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY ) ); } @@ -82,8 +82,8 @@ static void emit_load_R8G8B8A8_UNORM( struct aos_compilation *cp, struct x86_reg src_ptr ) { sse_movss(cp->func, data, src_ptr); - sse2_punpcklbw(cp->func, data, aos_get_internal( cp, IMM_IDENTITY )); - sse2_punpcklbw(cp->func, data, aos_get_internal( cp, IMM_IDENTITY )); + sse2_punpcklbw(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY )); + sse2_punpcklbw(cp->func, data, aos_get_internal_xmm( cp, IMM_IDENTITY )); sse2_cvtdq2ps(cp->func, data, data); sse_mulps(cp->func, data, aos_get_internal(cp, IMM_INV_255)); } -- cgit v1.2.3 From 05029c919d46299ca259ee8af880d0a65f95ce7c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 22 May 2008 13:46:06 +0100 Subject: draw: clean up masked writes in aos sse, make some xmm function names clearer --- src/gallium/auxiliary/draw/draw_vs_aos.c | 236 +++++++++++-------------------- 1 file changed, 82 insertions(+), 154 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 0b8600696a..708ecadbac 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -171,7 +171,7 @@ static boolean is_xmm_tmp( struct aos_compilation *cp, cp->xmm[reg.idx].file == TGSI_FILE_NULL); } -static struct x86_reg get_xmm_tmp( struct aos_compilation *cp, +static struct x86_reg get_xmm_clone( struct aos_compilation *cp, struct x86_reg reg ) { if (!is_xmm_tmp(cp, reg)) { @@ -380,31 +380,37 @@ static void emit_pshufd( struct aos_compilation *cp, } } - - - -/* Helper for writemask: +/* load masks (pack into negs??) + * pshufd - shuffle according to writemask + * and - result, mask + * nand - dest, mask + * or - dest, result */ -static boolean emit_shuf_copy1( struct aos_compilation *cp, - struct x86_reg dst, - struct x86_reg arg0, - struct x86_reg arg1, - ubyte shuf ) +static boolean mask_write( struct aos_compilation *cp, + struct x86_reg dst, + struct x86_reg result, + unsigned mask ) { + struct x86_reg imm_swz = aos_get_internal_xmm(cp, IMM_SWZ); struct x86_reg tmp = aos_get_xmm_reg(cp); - sse_movups(cp->func, dst, arg1); - emit_pshufd(cp, dst, dst, shuf); - emit_pshufd(cp, tmp, arg0, shuf); - - sse_movss(cp->func, dst, tmp); + + emit_pshufd(cp, tmp, imm_swz, + SHUF((mask & 1) ? 2 : 3, + (mask & 2) ? 2 : 3, + (mask & 4) ? 2 : 3, + (mask & 8) ? 2 : 3)); - emit_pshufd(cp, dst, dst, shuf); + sse_andps(cp->func, dst, tmp); + sse_andnps(cp->func, tmp, result); + sse_orps(cp->func, dst, tmp); aos_release_xmm_reg(cp, tmp.idx); return TRUE; } + + /* Helper for writemask: */ static boolean emit_shuf_copy2( struct aos_compilation *cp, @@ -414,17 +420,18 @@ static boolean emit_shuf_copy2( struct aos_compilation *cp, ubyte shuf ) { struct x86_reg tmp = aos_get_xmm_reg(cp); + emit_pshufd(cp, dst, arg1, shuf); emit_pshufd(cp, tmp, arg0, shuf); - sse_shufps(cp->func, dst, tmp, SHUF(X, Y, Z, W)); - emit_pshufd(cp, dst, dst, shuf); aos_release_xmm_reg(cp, tmp.idx); return TRUE; } + + #define SSE_SWIZZLE_NOOP ((0<<0) | (1<<2) | (2<<4) | (3<<6)) @@ -593,131 +600,58 @@ static void store_dest( struct aos_compilation *cp, const struct tgsi_full_dst_register *reg, struct x86_reg result ) { - if (reg->DstRegister.WriteMask == 0) - { - return; - } - else if (reg->DstRegister.WriteMask == TGSI_WRITEMASK_XYZW) - { - if (result.file == file_XMM) { - aos_adopt_xmm_reg(cp, - result, - reg->DstRegister.File, - reg->DstRegister.Index, - TRUE); - } - else { - struct x86_reg dst = aos_get_xmm_reg(cp); - aos_adopt_xmm_reg(cp, - dst, - reg->DstRegister.File, - reg->DstRegister.Index, - TRUE); - sse_movups(cp->func, dst, result); - } - } - else - { - /* Previous value of the dest register: - */ - struct x86_reg old_dst = aos_get_shader_reg(cp, - reg->DstRegister.File, - reg->DstRegister.Index); - - - /* Alloc an xmm reg to hold the new value of the dest register: - */ - struct x86_reg dst = aos_get_xmm_reg(cp); + struct x86_reg dst; + switch (reg->DstRegister.WriteMask) { + case 0: + return; + + case TGSI_WRITEMASK_XYZW: aos_adopt_xmm_reg(cp, - dst, + get_xmm_clone(cp, result), reg->DstRegister.File, reg->DstRegister.Index, - TRUE ); - - switch (reg->DstRegister.WriteMask) { - case TGSI_WRITEMASK_X: - if (result.file == file_XMM) { - sse_movups(cp->func, dst, old_dst); - sse_movss(cp->func, dst, result); - } - else { - struct x86_reg tmp = aos_get_xmm_reg(cp); - sse_movups(cp->func, dst, old_dst); - sse_movss(cp->func, tmp, result); - sse_movss(cp->func, dst, tmp); - aos_release_xmm_reg(cp, tmp.idx); - } - break; - - case TGSI_WRITEMASK_XY: - sse_movups(cp->func, dst, old_dst); - sse_shufps(cp->func, dst, result, SHUF(X, Y, Z, W)); - break; - - case TGSI_WRITEMASK_ZW: - sse_movups(cp->func, dst, result); - sse_shufps(cp->func, dst, old_dst, SHUF(X, Y, Z, W)); - break; - - case TGSI_WRITEMASK_YZW: - if (old_dst.file == file_XMM) { - sse_movups(cp->func, dst, result); - sse_movss(cp->func, dst, old_dst); - } - else { - struct x86_reg tmp = aos_get_xmm_reg(cp); - sse_movups(cp->func, dst, result); - sse_movss(cp->func, tmp, old_dst); - sse_movss(cp->func, dst, tmp); - aos_release_xmm_reg(cp, tmp.idx); - } - break; - - case TGSI_WRITEMASK_Y: - emit_shuf_copy1(cp, dst, result, old_dst, SHUF(Y,X,Z,W)); - break; - - case TGSI_WRITEMASK_Z: - emit_shuf_copy1(cp, dst, result, old_dst, SHUF(Z,Y,X,W)); - break; - - case TGSI_WRITEMASK_W: - emit_shuf_copy1(cp, dst, result, old_dst, SHUF(W,Y,Z,X)); - break; - - case TGSI_WRITEMASK_XZ: - emit_shuf_copy2(cp, dst, result, old_dst, SHUF(X,Z,Y,W)); - break; + TRUE); + return; + default: + break; + } - case TGSI_WRITEMASK_XW: - emit_shuf_copy2(cp, dst, result, old_dst, SHUF(X,W,Z,Y)); + dst = aos_get_shader_reg_xmm(cp, + reg->DstRegister.File, + reg->DstRegister.Index); - case TGSI_WRITEMASK_YZ: - emit_shuf_copy2(cp, dst, result, old_dst, SHUF(Z,Y,X,W)); - break; + switch (reg->DstRegister.WriteMask) { + case TGSI_WRITEMASK_X: + sse_movss(cp->func, dst, get_xmm_clone(cp, result)); + break; + + case TGSI_WRITEMASK_XY: + sse_shufps(cp->func, dst, get_xmm_clone(cp, result), SHUF(X, Y, Z, W)); + break; - case TGSI_WRITEMASK_YW: - emit_shuf_copy2(cp, dst, result, old_dst, SHUF(W,Y,Z,X)); - break; + case TGSI_WRITEMASK_ZW: + result = get_xmm_clone(cp, result); + sse_shufps(cp->func, result, dst, SHUF(X, Y, Z, W)); + dst = result; + break; - case TGSI_WRITEMASK_XZW: - emit_shuf_copy1(cp, dst, old_dst, result, SHUF(Y,X,Z,W)); - break; + case TGSI_WRITEMASK_YZW: + sse_movss(cp->func, result, dst); + dst = result; + break; - case TGSI_WRITEMASK_XYW: - emit_shuf_copy1(cp, dst, old_dst, result, SHUF(Z,Y,X,W)); - break; + default: + mask_write(cp, dst, result, reg->DstRegister.WriteMask); + break; + } - case TGSI_WRITEMASK_XYZ: - emit_shuf_copy1(cp, dst, old_dst, result, SHUF(W,Y,Z,X)); - break; + aos_adopt_xmm_reg(cp, + dst, + reg->DstRegister.File, + reg->DstRegister.Index, + TRUE); - default: - assert(0); /* not possible */ - break; - } - } } @@ -837,7 +771,7 @@ static boolean emit_ABS( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_mulps(cp->func, dst, neg); sse_maxps(cp->func, dst, arg0); @@ -850,7 +784,7 @@ static boolean emit_ADD( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_addps(cp->func, dst, arg1); @@ -874,7 +808,7 @@ static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg tmp = aos_get_xmm_reg(cp); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_mulps(cp->func, dst, arg1); @@ -898,7 +832,7 @@ static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg tmp = aos_get_xmm_reg(cp); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_mulps(cp->func, dst, arg1); @@ -920,7 +854,7 @@ static boolean emit_DPH( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg tmp = aos_get_xmm_reg(cp); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_mulps(cp->func, dst, arg1); @@ -1216,7 +1150,7 @@ static boolean emit_MAX( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_maxps(cp->func, dst, arg1); @@ -1229,7 +1163,7 @@ static boolean emit_MIN( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_minps(cp->func, dst, arg1); @@ -1240,7 +1174,7 @@ static boolean emit_MIN( struct aos_compilation *cp, const struct tgsi_full_inst static boolean emit_MOV( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); /* potentially nothing to do */ @@ -1252,7 +1186,7 @@ static boolean emit_MUL( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_mulps(cp->func, dst, arg1); @@ -1270,7 +1204,7 @@ static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_inst /* If we can't clobber old contents of arg0, get a temporary & copy * it there, then clobber it... */ - arg0 = get_xmm_tmp(cp, arg0); + arg0 = get_xmm_clone(cp, arg0); sse_mulps(cp->func, arg0, arg1); sse_addps(cp->func, arg0, arg2); @@ -1336,7 +1270,7 @@ static boolean emit_SGE( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg ones = aos_get_internal(cp, IMM_ONES); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_cmpps(cp->func, dst, arg1, cc_NotLessThan); sse_andps(cp->func, dst, ones); @@ -1360,7 +1294,7 @@ static boolean emit_SLT( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg ones = aos_get_internal(cp, IMM_ONES); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_cmpps(cp->func, dst, arg1, cc_LessThan); sse_andps(cp->func, dst, ones); @@ -1373,7 +1307,7 @@ static boolean emit_SUB( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_tmp(cp, arg0); + struct x86_reg dst = get_xmm_clone(cp, arg0); sse_subps(cp->func, dst, arg1); @@ -1526,9 +1460,9 @@ emit_instruction( struct aos_compilation *cp, static boolean emit_viewport( struct aos_compilation *cp ) { - struct x86_reg pos = aos_get_shader_reg(cp, - TGSI_FILE_OUTPUT, - 0); + struct x86_reg pos = aos_get_shader_reg_xmm(cp, + TGSI_FILE_OUTPUT, + 0); struct x86_reg scale = x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, scale)); @@ -1536,12 +1470,6 @@ static boolean emit_viewport( struct aos_compilation *cp ) struct x86_reg translate = x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, translate)); - if (pos.file != file_XMM) { - struct x86_reg dst = aos_get_xmm_reg(cp); - sse_movups(cp->func, dst, pos); - pos = dst; - } - sse_mulps(cp->func, pos, scale); sse_addps(cp->func, pos, translate); -- cgit v1.2.3 From 7b25c1a4032960752d8a8e950bdf75740b2de2e8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 22 May 2008 13:47:08 +0100 Subject: draw: remove FPU_MANIP ifdef --- src/gallium/auxiliary/draw/draw_vs_aos.c | 11 +---------- 1 file changed, 1 insertion(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 708ecadbac..d60940bb7a 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -693,48 +693,39 @@ static void x87_fstp_dest4( struct aos_compilation *cp, x87_fstp_or_pop(cp->func, writemask, 3, ptr); } -#define FPU_MANIP 1 /* Save current x87 state and put it into single precision mode. */ static void save_fpu_state( struct aos_compilation *cp ) { -#if FPU_MANIP x87_fnstcw( cp->func, x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, fpu_restore))); -#endif } static void restore_fpu_state( struct aos_compilation *cp ) { -#if FPU_MANIP x87_fnclex(cp->func); x87_fldcw( cp->func, x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, fpu_restore))); -#endif } static void set_fpu_round_neg_inf( struct aos_compilation *cp ) { -#if FPU_MANIP if (cp->fpucntl != FPU_RND_NEG) { cp->fpucntl = FPU_RND_NEG; x87_fnclex(cp->func); x87_fldcw( cp->func, x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, fpu_rnd_neg_inf))); } -#endif } static void set_fpu_round_nearest( struct aos_compilation *cp ) { -#if FPU_MANIP if (cp->fpucntl != FPU_RND_NEAREST) { cp->fpucntl = FPU_RND_NEAREST; x87_fnclex(cp->func); x87_fldcw( cp->func, x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, fpu_rnd_nearest))); } -#endif } @@ -754,7 +745,7 @@ static void x87_emit_ex2( struct aos_compilation *cp ) x87_fld1(cp->func); /* 1 (2^frac(a))-1 a */ x87_faddp(cp->func, st1); /* 2^frac(a) a */ x87_fscale(cp->func); /* 2^a a */ - x87_fstp(cp->func, st1); + x87_fstp(cp->func, st1); /* 2^a */ assert( stack == cp->func->x87_stack); -- cgit v1.2.3 From 6780a6dede31e7f2eb465e1d7b507b3e64fe6ec9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 22 May 2008 13:48:07 +0100 Subject: draw: shortcircuit shuffle in aos_sse when possible --- src/gallium/auxiliary/draw/draw_vs_aos.c | 18 +++++++++++++----- 1 file changed, 13 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index d60940bb7a..b8fad231ca 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -809,7 +809,9 @@ static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_inst sse_addss(cp->func, dst, tmp); /* a*x+c*z, b*y, ?, ? */ emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z)); sse_addss(cp->func, dst, tmp); - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + + if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); aos_release_xmm_reg(cp, tmp.idx); store_dest(cp, &op->FullDstRegisters[0], dst); @@ -833,7 +835,9 @@ static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_inst sse_addps(cp->func, dst, tmp); /* a*x+c*z, b*y+d*w, a*x+c*z, b*y+d*w */ emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z)); sse_addss(cp->func, dst, tmp); - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + + if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); aos_release_xmm_reg(cp, tmp.idx); store_dest(cp, &op->FullDstRegisters[0], dst); @@ -857,7 +861,9 @@ static boolean emit_DPH( struct aos_compilation *cp, const struct tgsi_full_inst sse_addss(cp->func, dst, tmp); emit_pshufd(cp, tmp, arg1, SHUF(W,W,W,W)); sse_addss(cp->func, dst, tmp); - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + + if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); aos_release_xmm_reg(cp, tmp.idx); store_dest(cp, &op->FullDstRegisters[0], dst); @@ -1233,7 +1239,8 @@ static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_inst sse_divss(cp->func, dst, arg0); } - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); store_dest(cp, &op->FullDstRegisters[0], dst); return TRUE; @@ -1249,7 +1256,8 @@ static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_inst /* Extend precision here... */ - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) + sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); store_dest(cp, &op->FullDstRegisters[0], dst); return TRUE; -- cgit v1.2.3 From 65cb09249e750b45ec3fc9a57670fc77250efc5e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 22 May 2008 13:49:38 +0100 Subject: draw: for debug, do rhw divide in aos_sse viewport calcs --- src/gallium/auxiliary/draw/draw_vs_aos.c | 45 +++++++++++++++++++++++++++++++- 1 file changed, 44 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index b8fad231ca..40de13a98c 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1481,6 +1481,46 @@ static boolean emit_viewport( struct aos_compilation *cp ) } +/* This is useful to be able to see the results on softpipe. Doesn't + * do proper clipping, just assumes the backend can do it during + * rasterization -- for debug only... + */ +static boolean emit_rhw_viewport( struct aos_compilation *cp ) +{ + struct x86_reg tmp = aos_get_xmm_reg(cp); + struct x86_reg pos = aos_get_shader_reg_xmm(cp, + TGSI_FILE_OUTPUT, + 0); + + struct x86_reg scale = x86_make_disp(cp->machine_EDX, + Offset(struct aos_machine, scale)); + + struct x86_reg translate = x86_make_disp(cp->machine_EDX, + Offset(struct aos_machine, translate)); + + + + emit_pshufd(cp, tmp, pos, SHUF(W, W, W, W)); + sse2_rcpss(cp->func, tmp, tmp); + sse_shufps(cp->func, tmp, tmp, SHUF(X, X, X, X)); + + sse_mulps(cp->func, pos, scale); + sse_mulps(cp->func, pos, tmp); + sse_addps(cp->func, pos, translate); + + /* Set pos[3] = w + */ + mask_write(cp, pos, tmp, TGSI_WRITEMASK_W); + + aos_adopt_xmm_reg( cp, + pos, + TGSI_FILE_OUTPUT, + 0, + TRUE ); + return TRUE; +} + + static boolean note_immediate( struct aos_compilation *cp, struct tgsi_full_immediate *imm ) { @@ -1623,7 +1663,10 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, goto fail; if (cp.vaos->base.key.viewport) { - emit_viewport(&cp); + if (0) + emit_viewport(&cp); + else + emit_rhw_viewport(&cp); } /* Emit output... TODO: do this eagerly after the last write to a -- cgit v1.2.3 From 260001430bbd28ea17201f1980ab1ebed93b246f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 22 May 2008 15:24:02 +0100 Subject: draw: use aligned movs within draw_vs_aos.c --- src/gallium/auxiliary/draw/draw_vs_aos.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 40de13a98c..039e233fe8 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -159,7 +159,7 @@ static void spill( struct aos_compilation *cp, unsigned idx ) cp->xmm[idx].idx); assert(cp->xmm[idx].dirty); - sse_movups(cp->func, oldval, x86_make_reg(file_XMM, idx)); + sse_movaps(cp->func, oldval, x86_make_reg(file_XMM, idx)); cp->xmm[idx].dirty = 0; } } @@ -176,7 +176,7 @@ static struct x86_reg get_xmm_clone( struct aos_compilation *cp, { if (!is_xmm_tmp(cp, reg)) { struct x86_reg tmp = aos_get_xmm_reg(cp); - sse_movups(cp->func, tmp, reg); + sse_movaps(cp->func, tmp, reg); reg = tmp; } @@ -335,7 +335,7 @@ static struct x86_reg aos_get_shader_reg_xmm( struct aos_compilation *cp, if (reg.file != file_XMM) { struct x86_reg tmp = aos_get_xmm_reg(cp); - sse_movups(cp->func, tmp, reg); + sse_movaps(cp->func, tmp, reg); aos_adopt_xmm_reg( cp, tmp, file, idx, FALSE ); reg = tmp; } @@ -374,7 +374,7 @@ static void emit_pshufd( struct aos_compilation *cp, } else { if (!eq(dst, arg0)) - sse_movups(cp->func, dst, arg0); + sse_movaps(cp->func, dst, arg0); sse_shufps(cp->func, dst, dst, shuf); } @@ -523,7 +523,7 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); struct x86_reg tmp = aos_get_xmm_reg(cp); - sse_movups(cp->func, tmp, arg0); + sse_movaps(cp->func, tmp, arg0); sse_mulps(cp->func, tmp, neg); sse_maxps(cp->func, dst, arg0); -- cgit v1.2.3 From 43df4642f1d2f3d2673a1d5e4f5126f5175fb899 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 22 May 2008 20:21:49 +0100 Subject: draw: tweak x87_emit_ex2 to avoid changing x87 fpu settings --- src/gallium/auxiliary/draw/draw_vs_aos.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 039e233fe8..93bb4f9bc0 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -735,16 +735,17 @@ static void x87_emit_ex2( struct aos_compilation *cp ) struct x86_reg st1 = x86_make_reg(file_x87, 1); int stack = cp->func->x87_stack; - set_fpu_round_neg_inf( cp ); +// set_fpu_round_neg_inf( cp ); x87_fld(cp->func, st0); /* a a */ - x87_fld(cp->func, st0); /* a a a */ - x87_fprndint( cp->func ); /* flr(a) a a*/ - x87_fsubp(cp->func, st1); /* frac(a) a */ - x87_f2xm1(cp->func); /* (2^frac(a))-1 a */ - x87_fld1(cp->func); /* 1 (2^frac(a))-1 a */ - x87_faddp(cp->func, st1); /* 2^frac(a) a */ - x87_fscale(cp->func); /* 2^a a */ + x87_fprndint( cp->func ); /* int(a) a*/ + x87_fsubr(cp->func, st1, st0); /* int(a) frc(a) */ + x87_fxch(cp->func, st1); /* frc(a) int(a) */ + x87_f2xm1(cp->func); /* (2^frc(a))-1 int(a) */ + x87_fld1(cp->func); /* 1 (2^frc(a))-1 int(a) */ + x87_faddp(cp->func, st1); /* 2^frac(a) int(a) */ + x87_fscale(cp->func); /* (2^frac(a)*2^int(int(a))) int(a) */ + /* 2^a int(a) */ x87_fstp(cp->func, st1); /* 2^a */ assert( stack == cp->func->x87_stack); -- cgit v1.2.3 From 7106da136069f865747e03c30ca245bc030b241b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 22 May 2008 20:22:15 +0100 Subject: draw: correct but slow LIT() in aos varient --- src/gallium/auxiliary/draw/draw_vs_aos.c | 19 +++++++++---------- 1 file changed, 9 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 93bb4f9bc0..930914f609 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1093,10 +1093,12 @@ static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_inst - /* a1' = a1 <= 0 ? 1 : a1; + /* a1' = a1 <= 0 ? 1 : a1; + * + * Note: use 1.0 to avoid passing zero to */ - x87_fldz(cp->func); /* 0 */ - x87_fld1(cp->func); /* 1 0 */ + x87_fldz(cp->func); /* 1 0 */ + x87_fldz(cp->func); /* 1 0 */ x87_fld_src(cp, &op->FullSrcRegisters[0], 1); /* a1 1 0 */ x87_fcomi(cp->func, st2); /* a1 1 0 */ x87_fcmovb(cp->func, st1); /* a1' 1 0 */ @@ -1119,17 +1121,14 @@ static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_inst x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* a0 0 r2 */ x87_fcomi(cp->func, st1); x87_fcmovb(cp->func, st1); /* a0' 0 r2 */ - x87_fstp(cp->func, st1); /* a0' r2 */ - x87_fxch(cp->func, st1); /* a0' r2 */ x87_fst_or_nop(cp->func, writemask, 1, dst); /* result[1] = a0' */ - x87_fldz(cp->func); /* 0 a0' r2 */ - x87_fcomi(cp->func, st1); /* 0 a0' r2 */ - x87_fcmovnbe(cp->func, st2); /* r2' a0' r2 */ + x87_fcomi(cp->func, st1); /* a0' 0 r2 */ + x87_fcmovnbe(cp->func, st2); /* r2' 0' r2 */ - x87_fstp_or_pop(cp->func, writemask, 2, dst); - x87_fpop(cp->func); + x87_fstp_or_pop(cp->func, writemask, 2, dst); /* 0 r2 */ + x87_fpop(cp->func); /* r2 */ x87_fpop(cp->func); } -- cgit v1.2.3 From adc1f88fc9278bdbb3b24a6d48f91a0bd98e9f1c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 May 2008 09:10:59 +0100 Subject: mesa: do object-space lighting in ffvertex_prog.c Start pulling over some of the optimizations from the fixed function paths. --- src/mesa/main/ffvertex_prog.c | 79 +++++++++++++++++++++++++--------------- src/mesa/shader/prog_statevars.c | 42 +++++++++++++++++---- src/mesa/shader/prog_statevars.h | 6 ++- 3 files changed, 88 insertions(+), 39 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 810af9e33e..adf15b03c2 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -54,6 +54,7 @@ struct state_key { unsigned light_color_material_mask:12; unsigned light_material_mask:12; + unsigned need_eye_coords:1; unsigned normalize:1; unsigned rescale_normals:1; unsigned fog_source_is_depth:1; @@ -167,6 +168,8 @@ static struct state_key *make_state_key( GLcontext *ctx ) */ assert(fp); + key->need_eye_coords = ctx->_NeedEyeCoords; + key->fragprog_inputs_read = fp->Base.InputsRead; if (ctx->RenderMode == GL_FEEDBACK) { @@ -310,7 +313,7 @@ struct tnl_program { struct ureg eye_position; struct ureg eye_position_normalized; - struct ureg eye_normal; + struct ureg transformed_normal; struct ureg identity; GLuint materials; @@ -653,9 +656,9 @@ static void emit_normalize_vec3( struct tnl_program *p, struct ureg src ) { struct ureg tmp = get_temp(p); - emit_op2(p, OPCODE_DP3, tmp, 0, src, src); - emit_op1(p, OPCODE_RSQ, tmp, 0, tmp); - emit_op2(p, OPCODE_MUL, dest, 0, src, tmp); + emit_op2(p, OPCODE_DP3, tmp, WRITEMASK_X, src, src); + emit_op1(p, OPCODE_RSQ, tmp, WRITEMASK_X, tmp); + emit_op2(p, OPCODE_MUL, dest, 0, src, swizzle1(tmp, X)); release_temp(p, tmp); } @@ -705,36 +708,53 @@ static struct ureg get_eye_position_normalized( struct tnl_program *p ) } -static struct ureg get_eye_normal( struct tnl_program *p ) +static struct ureg get_transformed_normal( struct tnl_program *p ) { - if (is_undef(p->eye_normal)) { + if (is_undef(p->transformed_normal) && + !p->state->need_eye_coords && + !p->state->normalize && + !(p->state->need_eye_coords == p->state->rescale_normals)) + { + p->transformed_normal = register_input(p, VERT_ATTRIB_NORMAL ); + } + else if (is_undef(p->transformed_normal)) + { struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL ); struct ureg mvinv[3]; + struct ureg transformed_normal = reserve_temp(p); - register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 2, - STATE_MATRIX_INVTRANS, mvinv ); + if (p->state->need_eye_coords) { + register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 2, + STATE_MATRIX_INVTRANS, mvinv ); - p->eye_normal = reserve_temp(p); - - /* Transform to eye space: - */ - emit_matrix_transform_vec3( p, p->eye_normal, mvinv, normal ); + /* Transform to eye space: + */ + emit_matrix_transform_vec3( p, transformed_normal, mvinv, normal ); + normal = transformed_normal; + } /* Normalize/Rescale: */ if (p->state->normalize) { - emit_normalize_vec3( p, p->eye_normal, p->eye_normal ); + emit_normalize_vec3( p, transformed_normal, normal ); + normal = transformed_normal; } - else if (p->state->rescale_normals) { + else if (p->state->need_eye_coords == p->state->rescale_normals) { + /* This is already adjusted for eye/non-eye rendering: + */ struct ureg rescale = register_param2(p, STATE_INTERNAL, - STATE_NORMAL_SCALE); + STATE_NORMAL_SCALE); - emit_op2( p, OPCODE_MUL, p->eye_normal, 0, p->eye_normal, + emit_op2( p, OPCODE_MUL, transformed_normal, 0, normal, swizzle1(rescale, X)); + normal = transformed_normal; } + + assert(normal.file == PROGRAM_TEMPORARY); + p->transformed_normal = normal; } - return p->eye_normal; + return p->transformed_normal; } @@ -856,7 +876,7 @@ static struct ureg calculate_light_attenuation( struct tnl_program *p, */ if (!p->state->unit[i].light_spotcutoff_is_180) { struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL, - STATE_SPOT_DIR_NORMALIZED, i); + STATE_LIGHT_SPOT_DIR_NORMALIZED, i); struct ureg spot = get_temp(p); struct ureg slt = get_temp(p); @@ -907,7 +927,7 @@ static void build_lighting( struct tnl_program *p ) const GLboolean twoside = p->state->light_twoside; const GLboolean separate = p->state->separate_specular; GLuint nr_lights = 0, count = 0; - struct ureg normal = get_eye_normal(p); + struct ureg normal = get_transformed_normal(p); struct ureg lit = get_temp(p); struct ureg dots = get_temp(p); struct ureg _col0 = undef, _col1 = undef; @@ -984,20 +1004,21 @@ static void build_lighting( struct tnl_program *p ) /* Can used precomputed constants in this case. * Attenuation never applies to infinite lights. */ - VPpli = register_param3(p, STATE_LIGHT, i, - STATE_POSITION_NORMALIZED); + VPpli = register_param3(p, STATE_INTERNAL, + STATE_LIGHT_POSITION_NORMALIZED, i); if (p->state->light_local_viewer) { struct ureg eye_hat = get_eye_position_normalized(p); half = get_temp(p); emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); emit_normalize_vec3(p, half, half); } else { - half = register_param3(p, STATE_LIGHT, i, STATE_HALF_VECTOR); + half = register_param3(p, STATE_INTERNAL, + STATE_LIGHT_HALF_VECTOR, i); } } else { - struct ureg Ppli = register_param3(p, STATE_LIGHT, i, - STATE_POSITION); + struct ureg Ppli = register_param3(p, STATE_INTERNAL, + STATE_LIGHT_POSITION, i); struct ureg V = get_eye_position(p); struct ureg dist = get_temp(p); @@ -1201,7 +1222,7 @@ static void build_reflect_texgen( struct tnl_program *p, struct ureg dest, GLuint writemask ) { - struct ureg normal = get_eye_normal(p); + struct ureg normal = get_transformed_normal(p); struct ureg eye_hat = get_eye_position_normalized(p); struct ureg tmp = get_temp(p); @@ -1219,7 +1240,7 @@ static void build_sphere_texgen( struct tnl_program *p, struct ureg dest, GLuint writemask ) { - struct ureg normal = get_eye_normal(p); + struct ureg normal = get_transformed_normal(p); struct ureg eye_hat = get_eye_position_normalized(p); struct ureg tmp = get_temp(p); struct ureg half = register_scalar_const(p, .5); @@ -1338,7 +1359,7 @@ static void build_texture_transform( struct tnl_program *p ) } if (normal_mask) { - struct ureg normal = get_eye_normal(p); + struct ureg normal = get_transformed_normal(p); emit_op1(p, OPCODE_MOV, out_texgen, normal_mask, normal ); } @@ -1475,7 +1496,7 @@ create_new_program( const struct state_key *key, p.program = program; p.eye_position = undef; p.eye_position_normalized = undef; - p.eye_normal = undef; + p.transformed_normal = undef; p.identity = undef; p.temp_in_use = 0; diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index ba3c988445..37bd17ba4a 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -134,10 +134,6 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], value[3] = 1.0; } return; - case STATE_POSITION_NORMALIZED: - COPY_4V(value, ctx->Light.Light[ln].EyePosition); - NORMALIZE_3FV( value ); - return; default: _mesa_problem(ctx, "Invalid light state in fetch_state"); return; @@ -431,15 +427,46 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], value[2] = ctx->Fog.Density * ONE_DIV_LN2; value[3] = ctx->Fog.Density * ONE_DIV_SQRT_LN2; return; - case STATE_SPOT_DIR_NORMALIZED: { + + case STATE_LIGHT_SPOT_DIR_NORMALIZED: { /* here, state[2] is the light number */ /* pre-normalize spot dir */ const GLuint ln = (GLuint) state[2]; - COPY_3V(value, ctx->Light.Light[ln].EyeDirection); - NORMALIZE_3FV(value); + COPY_3V(value, ctx->Light.Light[ln]._NormDirection); value[3] = ctx->Light.Light[ln]._CosCutoff; return; } + + case STATE_LIGHT_POSITION: { + const GLuint ln = (GLuint) state[2]; + COPY_4V(value, ctx->Light.Light[ln]._Position); + return; + } + + case STATE_LIGHT_POSITION_NORMALIZED: { + const GLuint ln = (GLuint) state[2]; + COPY_4V(value, ctx->Light.Light[ln]._Position); + NORMALIZE_3FV( value ); + return; + } + + case STATE_LIGHT_HALF_VECTOR: { + const GLuint ln = (GLuint) state[2]; + GLfloat p[3]; + /* Compute infinite half angle vector: + * halfVector = normalize(normalize(lightPos) + (0, 0, 1)) + * light.EyePosition.w should be 0 for infinite lights. + */ + COPY_3V(p, ctx->Light.Light[ln]._Position); + NORMALIZE_3FV(p); + ADD_3V(value, p, ctx->_EyeZDir); + NORMALIZE_3FV(value); + value[3] = 1.0; + return; + } + + + case STATE_PT_SCALE: value[0] = ctx->Pixel.RedScale; value[1] = ctx->Pixel.GreenScale; @@ -696,7 +723,6 @@ append_token(char *dst, gl_state_index k) append(dst, "normalScale"); break; case STATE_INTERNAL: - case STATE_POSITION_NORMALIZED: append(dst, "(internal)"); break; case STATE_PT_SCALE: diff --git a/src/mesa/shader/prog_statevars.h b/src/mesa/shader/prog_statevars.h index d12142055f..a515fda3aa 100644 --- a/src/mesa/shader/prog_statevars.h +++ b/src/mesa/shader/prog_statevars.h @@ -106,9 +106,11 @@ typedef enum gl_state_index_ { STATE_INTERNAL, /* Mesa additions */ STATE_NORMAL_SCALE, STATE_TEXRECT_SCALE, - STATE_POSITION_NORMALIZED, /* normalized light position */ STATE_FOG_PARAMS_OPTIMIZED, /* for faster fog calc */ - STATE_SPOT_DIR_NORMALIZED, /* pre-normalized spot dir */ + STATE_LIGHT_SPOT_DIR_NORMALIZED, /* pre-normalized spot dir */ + STATE_LIGHT_POSITION, /* object vs eye space */ + STATE_LIGHT_POSITION_NORMALIZED, /* object vs eye space */ + STATE_LIGHT_HALF_VECTOR, /* object vs eye space */ STATE_PT_SCALE, /**< Pixel transfer RGBA scale */ STATE_PT_BIAS, /**< Pixel transfer RGBA bias */ STATE_PCM_SCALE, /**< Post color matrix RGBA scale */ -- cgit v1.2.3 From 3b41d619a1b7cc8c356c32af777486461ddd7926 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 May 2008 09:14:17 +0100 Subject: draw: faster LIT(), incorrect though --- src/gallium/auxiliary/draw/draw_vs_aos.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 930914f609..b0c3ac49d2 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1098,7 +1098,13 @@ static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_inst * Note: use 1.0 to avoid passing zero to */ x87_fldz(cp->func); /* 1 0 */ +#if 1 + x87_fld1(cp->func); /* 1 0 */ +#else + /* Correct but slow due to fp exceptions generated in fyl2x - fix me. + */ x87_fldz(cp->func); /* 1 0 */ +#endif x87_fld_src(cp, &op->FullSrcRegisters[0], 1); /* a1 1 0 */ x87_fcomi(cp->func, st2); /* a1 1 0 */ x87_fcmovb(cp->func, st1); /* a1' 1 0 */ -- cgit v1.2.3 From 45fd9ec462bf338b286d3b757c3e67d54f663dad Mon Sep 17 00:00:00 2001 From: Thomas Hellstrom Date: Thu, 22 May 2008 16:42:53 +0200 Subject: i915: Use the malloc pool for constant buffers since they don't go to the GPU directly. --- src/gallium/winsys/dri/intel/intel_winsys.h | 2 +- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 29 ++++++++++++------------ 2 files changed, 16 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_winsys.h b/src/gallium/winsys/dri/intel/intel_winsys.h index d0a319f9a4..3d32db10a4 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys.h +++ b/src/gallium/winsys/dri/intel/intel_winsys.h @@ -53,6 +53,7 @@ intel_create_i915simple( struct intel_context *intel, struct intel_buffer { struct pipe_buffer base; + struct _DriBufferPool *pool; struct _DriBufferObject *driBO; }; @@ -69,5 +70,4 @@ dri_bo( struct pipe_buffer *buf ) } - #endif diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 44baa6d9fa..7b1c0acfc9 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -50,6 +50,7 @@ struct intel_pipe_winsys { struct pipe_winsys winsys; struct _DriBufferPool *regionPool; + struct _DriBufferPool *mallocPool; struct _DriFreeSlabManager *fMan; }; @@ -110,16 +111,19 @@ intel_buffer_create(struct pipe_winsys *winsys, struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); unsigned flags = 0; + struct _DriBufferPool *pool; buffer->base.refcount = 1; buffer->base.alignment = alignment; buffer->base.usage = usage; buffer->base.size = size; - if (usage & (PIPE_BUFFER_USAGE_VERTEX /*| IWS_BUFFER_USAGE_LOCAL*/)) { + if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; + pool = iws->mallocPool; } else { flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; + pool = iws->regionPool; } if (usage & PIPE_BUFFER_USAGE_GPU_READ) @@ -141,10 +145,11 @@ intel_buffer_create(struct pipe_winsys *winsys, flags |= DRM_BO_FLAG_CACHED; #endif - driGenBuffers( iws->regionPool, + buffer->pool = pool; + driGenBuffers( buffer->pool, "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); - driBOData( buffer->driBO, size, NULL, iws->regionPool, 0 ); + driBOData( buffer->driBO, size, NULL, buffer->pool, 0 ); return &buffer->base; } @@ -278,8 +283,7 @@ intel_fence_finish( struct pipe_winsys *sws, struct pipe_fence_handle *fence, unsigned flag ) { - /* JB: Lets allways lazy wait */ - return driFenceFinish((struct _DriFenceObject *)fence, flag, 1); + return driFenceFinish((struct _DriFenceObject *)fence, flag, 0); } struct pipe_winsys * @@ -310,15 +314,9 @@ intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) iws->winsys.fence_finish = intel_fence_finish; if (fd) - iws->regionPool = driSlabPoolInit(fd, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - 64, 6, 16, 4096, 0, - fMan); + iws->regionPool = driDRMPoolInit(fd); + + iws->mallocPool = driMallocPoolInit(); return &iws->winsys; } @@ -331,6 +329,9 @@ intel_destroy_pipe_winsys( struct pipe_winsys *winsys ) if (iws->regionPool) { driPoolTakeDown(iws->regionPool); } + if (iws->mallocPool) { + driPoolTakeDown(iws->mallocPool); + } free(iws); } -- cgit v1.2.3 From 01cc1eebe92441ec33e951b95f0a5b52721741f0 Mon Sep 17 00:00:00 2001 From: Thomas Hellstrom Date: Fri, 23 May 2008 11:14:12 +0200 Subject: i915: Make batchbuffers larger. --- src/gallium/winsys/dri/intel/intel_batchbuffer.c | 2 +- src/gallium/winsys/dri/intel/intel_screen.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c index 3a8591b580..aa2eed53b8 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.c +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.c @@ -65,7 +65,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) driBOUnrefUserList(batch->list); driBOResetList(batch->list); - batch->size = 4096; // ZZZ JB batch->intel->intelScreen->maxBatchSize; + batch->size = 4 * 4096; // ZZZ JB batch->intel->intelScreen->maxBatchSize; driBOData(batch->buffer, batch->size, NULL, NULL, 0); /* diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index a3fc6d7344..58bebbe972 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -200,7 +200,7 @@ intelCreatePools(__DRIscreenPrivate * sPriv) DRM_BO_FLAG_MEM_TT, DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, - 4096, //intelScreen->maxBatchSize, + 4 * 4096, //intelScreen->maxBatchSize, 1, 40, 16*16384, 0, intelScreen->fMan); #endif -- cgit v1.2.3 From d607a02da929cd192a3b0896c38dbb468ffa5b0d Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 23 May 2008 11:24:33 +0200 Subject: i915: Made EGL report the actual modes on the screen/output pair --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index a9c0218455..ac1825210e 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -94,6 +94,18 @@ drm_update_res(struct drm_driver *drm_drv) drm_drv->res = drmModeGetResources(drm_drv->device->drmFD); } +static void +drm_add_modes_from_output(_EGLScreen *screen, drmModeOutputPtr output) +{ + struct drm_mode_modeinfo *m; + int i; + + for (i = 0; i < output->count_modes; i++) { + m = &output->modes[i]; + _eglAddNewMode(screen, m->hdisplay, m->vdisplay, m->vrefresh, m->name); + } +} + static EGLBoolean drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) { @@ -137,7 +149,7 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) screen->output = output; _eglInitScreen(&screen->base); _eglAddScreen(disp, &screen->base); - _eglAddNewMode(&screen->base, 1024, 768, 60 * 1000, "1024x768-60"); + drm_add_modes_from_output(&screen->base, output); } /* for now we only have one config */ @@ -396,7 +408,7 @@ drm_find_mode(drmModeOutputPtr output, _EGLMode *mode) for (i = 0; i < output->count_modes; i++) { m = &output->modes[i]; - if (m->hdisplay == mode->Width && m->vdisplay == mode->Height) + if (m->hdisplay == mode->Width && m->vdisplay == mode->Height && m->vrefresh == mode->RefreshRate) break; m = NULL; } -- cgit v1.2.3 From 845db16dbe66c0f741063333a70af79bca5b6544 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 22 May 2008 15:02:54 +0200 Subject: i915: Fix for edgeflags --- src/gallium/drivers/i915simple/i915_state.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 4adeb37e86..964bd7c871 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -717,10 +717,16 @@ static void i915_set_vertex_elements(struct pipe_context *pipe, } +static void i915_set_edgeflags(struct pipe_context *pipe, + const unsigned *bitfield) +{ + /* TODO do something here */ +} void i915_init_state_functions( struct i915_context *i915 ) { + i915->pipe.set_edgeflags = i915_set_edgeflags; i915->pipe.create_blend_state = i915_create_blend_state; i915->pipe.bind_blend_state = i915_bind_blend_state; i915->pipe.delete_blend_state = i915_delete_blend_state; -- cgit v1.2.3 From 5bf6ffb0b2be34ffb000e9700861b177afa9df97 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 23 May 2008 16:11:38 +0200 Subject: i915: Improved vertex buffer performance --- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 22 +++++++++++++++++++--- 1 file changed, 19 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 7b1c0acfc9..62b43cc395 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -51,7 +51,8 @@ struct intel_pipe_winsys { struct pipe_winsys winsys; struct _DriBufferPool *regionPool; struct _DriBufferPool *mallocPool; - struct _DriFreeSlabManager *fMan; + struct _DriBufferPool *vertexPool; + struct _DriFreeSlabManager *fMan; /** shared between all pipes */ }; @@ -119,8 +120,12 @@ intel_buffer_create(struct pipe_winsys *winsys, buffer->base.size = size; if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { - flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; + flags |= DRM_BO_FLAG_MEM_LOCAL; pool = iws->mallocPool; + } else if (usage & PIPE_BUFFER_USAGE_CUSTOM) { + /* For vertex buffers */ + flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; + pool = iws->vertexPool; } else { flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; pool = iws->regionPool; @@ -313,8 +318,19 @@ intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) iws->winsys.fence_signalled = intel_fence_signalled; iws->winsys.fence_finish = intel_fence_finish; - if (fd) + if (fd) { iws->regionPool = driDRMPoolInit(fd); + iws->vertexPool = driSlabPoolInit(fd, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + 32 * 4096, + 1, 40, 32 * 4096 * 2, 0, + fMan); + } iws->mallocPool = driMallocPoolInit(); -- cgit v1.2.3 From 7fbb61eedd4b07f07007a172cea227d5c363b908 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 23 May 2008 16:55:30 +0200 Subject: i915: Revert accidental change --- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 62b43cc395..6a42c79033 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -120,7 +120,7 @@ intel_buffer_create(struct pipe_winsys *winsys, buffer->base.size = size; if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { - flags |= DRM_BO_FLAG_MEM_LOCAL; + flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; pool = iws->mallocPool; } else if (usage & PIPE_BUFFER_USAGE_CUSTOM) { /* For vertex buffers */ -- cgit v1.2.3 From 0ac2f7955c01749e122f67ff03e79a0d8bd0f8e5 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 May 2008 19:17:02 +0100 Subject: mesa: don't emit LIT instruction when mat shininess known to be zero Use a faster path in that case & make gears go faster. --- src/mesa/main/ffvertex_prog.c | 133 ++++++++++++++++++++++++++++++++---------- 1 file changed, 102 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index adf15b03c2..623c2a64b5 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -53,6 +53,7 @@ struct state_key { unsigned light_color_material:1; unsigned light_color_material_mask:12; unsigned light_material_mask:12; + unsigned material_shininess_is_zero:1; unsigned need_eye_coords:1; unsigned normalize:1; @@ -155,6 +156,26 @@ tnl_get_per_vertex_fog(GLcontext *ctx) #endif } +static GLboolean check_active_shininess( GLcontext *ctx, + const struct state_key *key, + GLuint side ) +{ + GLuint bit = 1 << (MAT_ATTRIB_FRONT_SHININESS + side); + + if (key->light_color_material_mask & bit) + return GL_TRUE; + + if (key->light_material_mask & bit) + return GL_TRUE; + + if (ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_SHININESS + side][0] != 0.0F) + return GL_TRUE; + + return GL_FALSE; +} + + + static struct state_key *make_state_key( GLcontext *ctx ) { @@ -214,6 +235,17 @@ static struct state_key *make_state_key( GLcontext *ctx ) key->unit[i].light_attenuated = 1; } } + + if (check_active_shininess(ctx, key, 0)) { + key->material_shininess_is_zero = 0; + } + else if (key->light_twoside && + check_active_shininess(ctx, key, 1)) { + key->material_shininess_is_zero = 0; + } + else { + key->material_shininess_is_zero = 1; + } } if (ctx->Transform.Normalize) @@ -915,7 +947,26 @@ static struct ureg calculate_light_attenuation( struct tnl_program *p, } +static void emit_degenerate_lit( struct tnl_program *p, + struct ureg lit, + struct ureg dots ) +{ + struct ureg id = get_identity_param(p); + + /* 1, 0, 0, 1 + */ + emit_op1(p, OPCODE_MOV, lit, 0, swizzle(id, Z, X, X, Z)); + /* 1, MAX2(in[0], 0), 0, 1 + */ + emit_op2(p, OPCODE_MAX, lit, WRITEMASK_Y, lit, swizzle1(dots, X)); + + /* 1, MAX2(in[0], 0), (in[0] > 0 ? 1 : 0), 1 + */ + emit_op2(p, OPCODE_SLT, lit, WRITEMASK_Z, + lit, /* 0 */ + swizzle1(dots, X)); /* in[0] */ +} /* Need to add some addtional parameters to allow lighting in object @@ -941,9 +992,11 @@ static void build_lighting( struct tnl_program *p ) set_material_flags(p); { - struct ureg shininess = get_material(p, 0, STATE_SHININESS); - emit_op1(p, OPCODE_MOV, dots, WRITEMASK_W, swizzle1(shininess,X)); - release_temp(p, shininess); + if (!p->state->material_shininess_is_zero) { + struct ureg shininess = get_material(p, 0, STATE_SHININESS); + emit_op1(p, OPCODE_MOV, dots, WRITEMASK_W, swizzle1(shininess,X)); + release_temp(p, shininess); + } _col0 = make_temp(p, get_scenecolor(p, 0)); if (separate) @@ -954,10 +1007,12 @@ static void build_lighting( struct tnl_program *p ) } if (twoside) { - struct ureg shininess = get_material(p, 1, STATE_SHININESS); - emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z, - negate(swizzle1(shininess,X))); - release_temp(p, shininess); + if (!p->state->material_shininess_is_zero) { + struct ureg shininess = get_material(p, 1, STATE_SHININESS); + emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z, + negate(swizzle1(shininess,X))); + release_temp(p, shininess); + } _bfc0 = make_temp(p, get_scenecolor(p, 1)); if (separate) @@ -1006,14 +1061,17 @@ static void build_lighting( struct tnl_program *p ) */ VPpli = register_param3(p, STATE_INTERNAL, STATE_LIGHT_POSITION_NORMALIZED, i); - if (p->state->light_local_viewer) { - struct ureg eye_hat = get_eye_position_normalized(p); - half = get_temp(p); - emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); - emit_normalize_vec3(p, half, half); - } else { - half = register_param3(p, STATE_INTERNAL, - STATE_LIGHT_HALF_VECTOR, i); + + if (!p->state->material_shininess_is_zero) { + if (p->state->light_local_viewer) { + struct ureg eye_hat = get_eye_position_normalized(p); + half = get_temp(p); + emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); + emit_normalize_vec3(p, half, half); + } else { + half = register_param3(p, STATE_INTERNAL, + STATE_LIGHT_HALF_VECTOR, i); + } } } else { @@ -1023,7 +1081,6 @@ static void build_lighting( struct tnl_program *p ) struct ureg dist = get_temp(p); VPpli = get_temp(p); - half = get_temp(p); /* Calculate VPpli vector */ @@ -1045,16 +1102,20 @@ static void build_lighting( struct tnl_program *p ) /* Calculate viewer direction, or use infinite viewer: */ - if (p->state->light_local_viewer) { - struct ureg eye_hat = get_eye_position_normalized(p); - emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); - } - else { - struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z); - emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir); - } - - emit_normalize_vec3(p, half, half); + if (!p->state->material_shininess_is_zero) { + half = get_temp(p); + + if (p->state->light_local_viewer) { + struct ureg eye_hat = get_eye_position_normalized(p); + emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat); + } + else { + struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z); + emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir); + } + + emit_normalize_vec3(p, half, half); + } release_temp(p, dist); } @@ -1062,7 +1123,9 @@ static void build_lighting( struct tnl_program *p ) /* Calculate dot products: */ emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); - emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half); + + if (!p->state->material_shininess_is_zero) + emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half); /* Front face lighting: */ @@ -1073,7 +1136,11 @@ static void build_lighting( struct tnl_program *p ) struct ureg res0, res1; GLuint mask0, mask1; - emit_op1(p, OPCODE_LIT, lit, 0, dots); + if (p->state->material_shininess_is_zero) { + emit_degenerate_lit(p, lit, dots); + } else { + emit_op1(p, OPCODE_LIT, lit, 0, dots); + } if (!is_undef(att)) emit_op2(p, OPCODE_MUL, lit, 0, lit, att); @@ -1099,7 +1166,7 @@ static void build_lighting( struct tnl_program *p ) res1 = _col1; } - emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0); + emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0); emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0); emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1); @@ -1117,7 +1184,11 @@ static void build_lighting( struct tnl_program *p ) struct ureg res0, res1; GLuint mask0, mask1; - emit_op1(p, OPCODE_LIT, lit, 0, negate(swizzle(dots,X,Y,W,Z))); + if (p->state->material_shininess_is_zero) { + emit_degenerate_lit(p, lit, negate(swizzle(dots,X,Y,W,Z))); + } else { + emit_op1(p, OPCODE_LIT, lit, 0, negate(swizzle(dots,X,Y,W,Z))); + } if (!is_undef(att)) emit_op2(p, OPCODE_MUL, lit, 0, lit, att); @@ -1142,7 +1213,7 @@ static void build_lighting( struct tnl_program *p ) mask1 = 0; } - emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0); + emit_op2(p, OPCODE_ADD, _bfc0, 0, ambient, _bfc0); emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0); emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1); -- cgit v1.2.3 From 333d377bbda4f598292108f91cd8ec4f0f647c20 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 May 2008 19:17:57 +0100 Subject: glapi: fix include path & make build work --- src/mesa/x86/glapi_x86.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S index 9a5d875e21..b99c2b6520 100644 --- a/src/mesa/x86/glapi_x86.S +++ b/src/mesa/x86/glapi_x86.S @@ -27,7 +27,7 @@ */ #include "assyntax.h" -#include "glapioffsets.h" +#include "glapi/glapioffsets.h" #if defined(STDCALL_API) # if defined(USE_MGL_NAMESPACE) -- cgit v1.2.3 From e841b92d9c8bf48085b4996df828ae745977f931 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 May 2008 20:05:36 +0100 Subject: mesa: further degenerate the special case lit substitute --- src/mesa/main/ffvertex_prog.c | 21 ++++++++++++--------- 1 file changed, 12 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 623c2a64b5..90b156f812 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -953,19 +953,19 @@ static void emit_degenerate_lit( struct tnl_program *p, { struct ureg id = get_identity_param(p); - /* 1, 0, 0, 1 + /* Note that result.x & result.w will not be examined. Note also that + * dots.xyzw == dots.xxxx. */ - emit_op1(p, OPCODE_MOV, lit, 0, swizzle(id, Z, X, X, Z)); - /* 1, MAX2(in[0], 0), 0, 1 + /* result[1] = MAX2(in, 0) */ - emit_op2(p, OPCODE_MAX, lit, WRITEMASK_Y, lit, swizzle1(dots, X)); + emit_op2(p, OPCODE_MAX, lit, 0, id, dots); - /* 1, MAX2(in[0], 0), (in[0] > 0 ? 1 : 0), 1 + /* result[2] = (in > 0 ? 1 : 0) */ emit_op2(p, OPCODE_SLT, lit, WRITEMASK_Z, lit, /* 0 */ - swizzle1(dots, X)); /* in[0] */ + dots); /* in[0] */ } @@ -1122,10 +1122,13 @@ static void build_lighting( struct tnl_program *p ) /* Calculate dot products: */ - emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); - - if (!p->state->material_shininess_is_zero) + if (p->state->material_shininess_is_zero) { + emit_op2(p, OPCODE_DP3, dots, 0, normal, VPpli); + } + else { + emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half); + } /* Front face lighting: */ -- cgit v1.2.3 From feceb43948f76cc4d4c8ecbb86b1b1f438c6daee Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 23 May 2008 20:37:50 +0100 Subject: mesa: save a temp on normalizes --- src/mesa/main/ffvertex_prog.c | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 90b156f812..e36f1f69a4 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -305,7 +305,7 @@ static struct state_key *make_state_key( GLcontext *ctx ) * generated program with line/function references for each * instruction back into this file: */ -#define DISASSEM (MESA_VERBOSE&VERBOSE_DISASSEM) +#define DISASSEM 1 /* Should be tunable by the driver - do we want to do matrix * multiplications with DP4's or with MUL/MAD's? SSE works better @@ -687,11 +687,9 @@ static void emit_normalize_vec3( struct tnl_program *p, struct ureg dest, struct ureg src ) { - struct ureg tmp = get_temp(p); - emit_op2(p, OPCODE_DP3, tmp, WRITEMASK_X, src, src); - emit_op1(p, OPCODE_RSQ, tmp, WRITEMASK_X, tmp); - emit_op2(p, OPCODE_MUL, dest, 0, src, swizzle1(tmp, X)); - release_temp(p, tmp); + emit_op2(p, OPCODE_DP3, dest, WRITEMASK_X, src, src); + emit_op1(p, OPCODE_RSQ, dest, WRITEMASK_X, dest); + emit_op2(p, OPCODE_MUL, dest, 0, src, swizzle1(dest, X)); } static void emit_passthrough( struct tnl_program *p, -- cgit v1.2.3 From 059a652d64da470ccc7f2f3266fd64721848a7be Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 24 May 2008 19:25:02 +0900 Subject: scons: New profile build. --- src/gallium/auxiliary/util/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index d55d2c7081..0309de1ac2 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -5,6 +5,7 @@ util = env.ConvenienceLibrary( source = [ 'p_debug.c', 'p_debug_mem.c', + 'p_debug_prof.c', 'p_tile.c', 'p_util.c', 'u_blit.c', -- cgit v1.2.3 From 345eb7fb70840829571cbacdb3980181df8e018a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 24 May 2008 19:25:33 +0900 Subject: gallium: Poor-man profiler for win32 kernel. --- src/gallium/auxiliary/util/p_debug_prof.c | 175 ++++++++++++++++++++++++++++++ src/gallium/include/pipe/p_debug.h | 11 ++ 2 files changed, 186 insertions(+) create mode 100644 src/gallium/auxiliary/util/p_debug_prof.c (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug_prof.c b/src/gallium/auxiliary/util/p_debug_prof.c new file mode 100644 index 0000000000..958f99c327 --- /dev/null +++ b/src/gallium/auxiliary/util/p_debug_prof.c @@ -0,0 +1,175 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Poor-man profiling. + * + * @author José Fonseca + * + * @sa http://blogs.msdn.com/joshpoley/archive/2008/03/12/poor-man-s-profiler.aspx + * @sa http://www.johnpanzer.com/aci_cuj/index.html + */ + +#include "pipe/p_config.h" + +#if defined(PROFILE) && defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) + +#include +#include + +#include "pipe/p_debug.h" +#include "util/u_string.h" + + +#define PROFILE_FILE_SIZE 4*1024*1024 +#define FILE_NAME_SIZE 256 + +static WCHAR wFileName[FILE_NAME_SIZE]; +static ULONG_PTR iFile = 0; +static void *pMap = NULL; +static void *pMapEnd = NULL; + + +/** + * Called at the start of every method or function. + * + * @sa http://msdn.microsoft.com/en-us/library/c63a9b7h.aspx + */ +void __declspec(naked) __cdecl +_penter(void) { + _asm { + push ebx + mov ebx, [pMap] + test ebx, ebx + jz done + cmp ebx, [pMapEnd] + je done + push eax + push edx + mov eax, [esp+12] + and eax, 0xfffffffe + mov [ebx], eax + add ebx, 4 + rdtsc + mov [ebx], eax + add ebx, 4 + mov [pMap], ebx + pop edx + pop eax +done: + pop ebx + ret + } +} + + +/** + * Called at the end of Calls the end of every method or function. + * + * @sa http://msdn.microsoft.com/en-us/library/xc11y76y.aspx + */ +void __declspec(naked) __cdecl +_pexit(void) { + _asm { + push ebx + mov ebx, [pMap] + test ebx, ebx + jz done + cmp ebx, [pMapEnd] + je done + push eax + push edx + mov eax, [esp+12] + or eax, 0x00000001 + mov [ebx], eax + add ebx, 4 + rdtsc + mov [ebx], eax + add ebx, 4 + mov [pMap], ebx + pop edx + pop eax +done: + pop ebx + ret + } +} + + +void __declspec(naked) +__debug_profile_reference1(void) { + _asm { + call _penter + call _pexit + ret + } +} + + +void __declspec(naked) +__debug_profile_reference2(void) { + _asm { + call _penter + call __debug_profile_reference1 + call _pexit + ret + } +} + + +void +debug_profile_start(void) +{ + static unsigned no = 0; + char filename[FILE_NAME_SIZE]; + unsigned i; + + util_snprintf(filename, sizeof(filename), "\\??\\c:\\%03u.prof", ++no); + for(i = 0; i < FILE_NAME_SIZE; ++i) + wFileName[i] = (WCHAR)filename[i]; + + pMap = EngMapFile(wFileName, PROFILE_FILE_SIZE, &iFile); + if(pMap) { + pMapEnd = (unsigned char*)pMap + PROFILE_FILE_SIZE; + /* reference functions for calibration purposes */ + __debug_profile_reference2(); + } +} + +void +debug_profile_stop(void) +{ + if(iFile) { + EngUnmapFile(iFile); + /* TODO: truncate file */ + } + iFile = 0; + pMapEnd = pMap = NULL; +} + +#endif /* PROFILE */ diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 1263d0da61..0af635be57 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -313,6 +313,17 @@ void debug_memory_end(unsigned long beginning); +#if defined(PROFILE) && defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) + +void +debug_profile_start(void); + +void +debug_profile_stop(void); + +#endif + + #ifdef DEBUG void debug_dump_image(const char *prefix, unsigned format, unsigned cpp, -- cgit v1.2.3 From 6b3723ee8d084a1abbc971b21c58f7c1e66949a7 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 24 May 2008 13:22:15 +0100 Subject: rtasm: add some helpers for calling out from generated code --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 44 ++++++++++++++++++++++++++++-- src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 6 ++++ 2 files changed, 47 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index d78676b8f3..2415b0156b 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -218,6 +218,8 @@ static void emit_3ub( struct x86_function *p, unsigned char b0, unsigned char b1 /* Build a modRM byte + possible displacement. No treatment of SIB * indexing. BZZT - no way to encode an absolute address. + * + * This is the "/r" field in the x86 manuals... */ static void emit_modrm( struct x86_function *p, struct x86_reg reg, @@ -256,7 +258,8 @@ static void emit_modrm( struct x86_function *p, } } - +/* Emits the "/0".."/7" specialized versions of the modrm ("/r") bytes. + */ static void emit_modrm_noreg( struct x86_function *p, unsigned op, struct x86_reg regmem ) @@ -365,8 +368,7 @@ void x86_jcc( struct x86_function *p, DUMP_I(cc); if (offset < 0) { - int amt = p->csr - p->store; - assert(amt > -offset); + assert(p->csr - p->store > -offset); } if (offset <= 127 && offset >= -128) { @@ -443,6 +445,16 @@ void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ) emit_1i(p, imm); } +void x86_add_reg_imm8( struct x86_function *p, struct x86_reg dst, ubyte imm ) +{ + DUMP_RI( dst, imm ); + assert(dst.mod == mod_REG); + emit_1ub(p, 0x80); + emit_modrm_noreg(p, 0, dst); + emit_1ub(p, imm); +} + + void x86_push( struct x86_function *p, struct x86_reg reg ) { @@ -459,6 +471,17 @@ void x86_push( struct x86_function *p, p->stack_offset += 4; } +void x86_push_imm32( struct x86_function *p, + int imm32 ) +{ + DUMP_I( imm32 ); + emit_1ub(p, 0x68); + emit_1i(p, imm32); + + p->stack_offset += 4; +} + + void x86_pop( struct x86_function *p, struct x86_reg reg ) { @@ -1558,6 +1581,21 @@ void mmx_movq( struct x86_function *p, */ +void x86_cdecl_caller_push_regs( struct x86_function *p ) +{ + x86_push(p, x86_make_reg(file_REG32, reg_AX)); + x86_push(p, x86_make_reg(file_REG32, reg_CX)); + x86_push(p, x86_make_reg(file_REG32, reg_DX)); +} + +void x86_cdecl_caller_pop_regs( struct x86_function *p ) +{ + x86_pop(p, x86_make_reg(file_REG32, reg_DX)); + x86_pop(p, x86_make_reg(file_REG32, reg_CX)); + x86_pop(p, x86_make_reg(file_REG32, reg_AX)); +} + + /* Retreive a reference to one of the function arguments, taking into * account any push/pop activity: */ diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index 9f7e31e055..63e812fac9 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -155,6 +155,7 @@ void x86_call( struct x86_function *p, struct x86_reg reg); * I load the immediate into general purpose register and use it. */ void x86_mov_reg_imm( struct x86_function *p, struct x86_reg dst, int imm ); +void x86_add_reg_imm8( struct x86_function *p, struct x86_reg dst, ubyte imm ); /* Macro for sse_shufps() and sse2_pshufd(): @@ -225,6 +226,7 @@ void x86_imul( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_or( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_pop( struct x86_function *p, struct x86_reg reg ); void x86_push( struct x86_function *p, struct x86_reg reg ); +void x86_push_imm32( struct x86_function *p, int imm ); void x86_ret( struct x86_function *p ); void x86_retw( struct x86_function *p, unsigned short imm ); void x86_sub( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); @@ -232,6 +234,10 @@ void x86_test( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_xor( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void x86_sahf( struct x86_function *p ); + +void x86_cdecl_caller_push_regs( struct x86_function *p ); +void x86_cdecl_caller_pop_regs( struct x86_function *p ); + void x87_assert_stack_empty( struct x86_function *p ); void x87_f2xm1( struct x86_function *p ); -- cgit v1.2.3 From 6172f1295cf812108d8ceba15a83ba87880360d3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 24 May 2008 13:22:29 +0100 Subject: draw: add a debug-print which can be called from inside generated shaders --- src/gallium/auxiliary/draw/draw_vs_aos.c | 67 ++++++++++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs_aos.h | 1 + 2 files changed, 68 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index b0c3ac49d2..aa119f242e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -752,7 +752,63 @@ static void x87_emit_ex2( struct aos_compilation *cp ) } +static void PIPE_CDECL print_reg( const char *msg, + const float *reg ) +{ + debug_printf("%s: %f %f %f %f\n", msg, reg[0], reg[1], reg[2], reg[3]); +} + +static void emit_print( struct aos_compilation *cp, + const char *message, /* must point to a static string! */ + unsigned file, + unsigned idx ) +{ + struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); + struct x86_reg arg = get_reg_ptr( cp, file, idx ); + unsigned i; + + /* There shouldn't be anything on the x87 stack. Can add this + * capacity later if need be. + */ + assert(cp->func->x87_stack == 0); + + /* For absolute correctness, need to spill/invalidate all XMM regs + * too. We're obviously not concerned about performance on this + * debug path, so here goes: + */ + for (i = 0; i < 8; i++) { + if (cp->xmm[i].dirty) + spill(cp, i); + aos_release_xmm_reg(cp, i); + } + + /* Push caller-save (ie scratch) regs. + */ + x86_cdecl_caller_push_regs( cp->func ); + + + /* Push the arguments: + */ + x86_lea( cp->func, ecx, arg ); + x86_push( cp->func, ecx ); + x86_push_imm32( cp->func, (int)message ); + + /* Call the helper. Could call debug_printf directly, but + * print_reg is a nice place to put a breakpoint if need be. + */ + x86_mov_reg_imm( cp->func, ecx, (int)print_reg ); + x86_call( cp->func, ecx ); + x86_pop( cp->func, ecx ); + x86_pop( cp->func, ecx ); + + /* Pop caller-save regs + */ + x86_cdecl_caller_pop_regs( cp->func ); + + /* Done... + */ +} /** * The traditional instructions. All operate on internal registers @@ -1798,6 +1854,17 @@ static void vaos_set_constants( struct draw_vs_varient *varient, memcpy(vaos->machine->constant, constants, (vaos->base.vs->info.file_max[TGSI_FILE_CONSTANT] + 1) * 4 * sizeof(float)); + +#if 0 + unsigned i; + for (i =0; i < vaos->base.vs->info.file_max[TGSI_FILE_CONSTANT] + 1; i++) + debug_printf("state %d: %f %f %f %f\n", + i, + constants[i][0], + constants[i][1], + constants[i][2], + constants[i][3]); +#endif } diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index efdc9a38f4..a0680ec63d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -96,6 +96,7 @@ struct aos_compilation { unsigned insn_counter; unsigned num_immediates; + unsigned count; struct { unsigned idx:16; -- cgit v1.2.3 From a2b1c46535d02bb4cc154f26481eda264a65abe8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 24 May 2008 13:22:39 +0100 Subject: mesa: evaluate _NeedEyeCoords prior to generating internal vertex shader --- src/mesa/main/state.c | 26 ++++++++++++++------------ 1 file changed, 14 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index f8cb943e64..cdf1249cd0 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -1209,18 +1209,6 @@ _mesa_update_state_locked( GLcontext *ctx ) | _NEW_STENCIL | _DD_NEW_SEPARATE_SPECULAR)) update_tricaps( ctx, new_state ); - if (ctx->FragmentProgram._MaintainTexEnvProgram) { - prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR); - } - if (ctx->VertexProgram._MaintainTnlProgram) { - prog_flags |= (_NEW_TEXTURE | _NEW_TEXTURE_MATRIX | - _NEW_TRANSFORM | _NEW_POINT | - _NEW_FOG | _NEW_LIGHT); - } - if (new_state & prog_flags) - update_program( ctx ); - - /* ctx->_NeedEyeCoords is now up to date. * * If the truth value of this variable has changed, update for the @@ -1233,6 +1221,20 @@ _mesa_update_state_locked( GLcontext *ctx ) if (new_state & _MESA_NEW_NEED_EYE_COORDS) _mesa_update_tnl_spaces( ctx, new_state ); + if (ctx->FragmentProgram._MaintainTexEnvProgram) { + prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR); + } + if (ctx->VertexProgram._MaintainTnlProgram) { + prog_flags |= (_NEW_TEXTURE | _NEW_TEXTURE_MATRIX | + _NEW_TRANSFORM | _NEW_POINT | + _NEW_FOG | _NEW_LIGHT | + _MESA_NEW_NEED_EYE_COORDS); + } + if (new_state & prog_flags) + update_program( ctx ); + + + /* * Give the driver a chance to act upon the new_state flags. * The driver might plug in different span functions, for example. -- cgit v1.2.3 From e1590abb17f1effd92c136207f363de6cf52df18 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 24 May 2008 13:23:06 +0100 Subject: mesa: pre-swizzle normal scale state value --- src/mesa/main/ffvertex_prog.c | 3 +-- src/mesa/shader/prog_statevars.c | 6 +++++- 2 files changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index e36f1f69a4..7a099b2376 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -775,8 +775,7 @@ static struct ureg get_transformed_normal( struct tnl_program *p ) struct ureg rescale = register_param2(p, STATE_INTERNAL, STATE_NORMAL_SCALE); - emit_op2( p, OPCODE_MUL, transformed_normal, 0, normal, - swizzle1(rescale, X)); + emit_op2( p, OPCODE_MUL, transformed_normal, 0, normal, rescale ); normal = transformed_normal; } diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 37bd17ba4a..44fbfdcd04 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -397,7 +397,11 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], case STATE_INTERNAL: switch (state[1]) { case STATE_NORMAL_SCALE: - ASSIGN_4V(value, ctx->_ModelViewInvScale, 0, 0, 1); + ASSIGN_4V(value, + ctx->_ModelViewInvScale, + ctx->_ModelViewInvScale, + ctx->_ModelViewInvScale, + 1); return; case STATE_TEXRECT_SCALE: { -- cgit v1.2.3 From 86e529ad90411d21bca3d70984b2db202e7a0cd6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 24 May 2008 16:31:11 +0100 Subject: draw: use lookup tables to avoid calling pow() in LIT opcode --- src/gallium/auxiliary/draw/draw_vs_aos.c | 251 ++++++++++++++++++++++++++++++- src/gallium/auxiliary/draw/draw_vs_aos.h | 27 ++++ 2 files changed, 272 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index aa119f242e..1fbb7088ca 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -105,8 +105,31 @@ static struct x86_reg get_reg_ptr(struct aos_compilation *cp, #define X87_CW_ROUND_MASK (3<<10) #define X87_CW_INFINITY (1<<12) +static void do_populate_lut( struct shine_tab *tab, + float unclamped_exponent ) +{ + const float epsilon = 1.0F / 256.0F; + float exponent = CLAMP(unclamped_exponent, -(128.0F - epsilon), (128.0F - epsilon)); + unsigned i; + + tab->exponent = unclamped_exponent; /* for later comparison */ + + tab->values[0] = 0; + if (exponent == 0) { + for (i = 1; i < 258; i++) { + tab->values[i] = 1.0; + } + } + else { + for (i = 1; i < 258; i++) { + tab->values[i] = powf((float)i * epsilon, exponent); + } + } +} + static void init_internals( struct aos_machine *machine ) { + unsigned i; float inv = 1.0f/255.0f; float f255 = 255.0f; @@ -141,6 +164,9 @@ static void init_internals( struct aos_machine *machine ) (1<<6) | X87_CW_ROUND_DOWN | X87_CW_PRECISION_DOUBLE_EXT); + + for (i = 0; i < MAX_SHINE_TAB; i++) + do_populate_lut( &machine->shine_tab[i], 1.0f ); } @@ -1132,26 +1158,231 @@ static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } +static PIPE_CDECL void do_lit( struct aos_machine *machine, + float *result, + const float *in, + unsigned count ) +{ + if (in[0] > 0) + { + if (in[1] <= 0.0) + { + result[0] = 1.0F; + result[1] = in[0]; + result[2] = 1.0; + result[3] = 1.0F; + } + else + { + const float epsilon = 1.0F / 256.0F; + float exponent = CLAMP(in[3], -(128.0F - epsilon), (128.0F - epsilon)); + result[0] = 1.0F; + result[1] = in[0]; + result[2] = powf(in[1], exponent); + result[3] = 1.0; + } + } + else + { + result[0] = 1.0F; + result[1] = 0.0; + result[2] = 0.0; + result[3] = 1.0F; + } +} + + +static PIPE_CDECL void do_lit_lut( struct aos_machine *machine, + float *result, + const float *in, + unsigned count ) +{ + if (in[0] > 0) + { + if (in[1] <= 0.0) + { + result[0] = 1.0F; + result[1] = in[0]; + result[2] = 1.0; + result[3] = 1.0F; + return; + } + + if (machine->lit_info[count].shine_tab->exponent != in[3]) { + machine->lit_info[count].func = do_lit; + goto no_luck; + } + + if (in[1] <= 1.0) + { + const float *tab = machine->lit_info[count].shine_tab->values; + float f = in[1] * 256; + int k = (int)f; + float frac = f - (float)k; + + result[0] = 1.0F; + result[1] = in[0]; + result[2] = tab[k] + frac*(tab[k+1]-tab[k]); + result[3] = 1.0; + return; + } + + no_luck: + { + const float epsilon = 1.0F / 256.0F; + float exponent = CLAMP(in[3], -(128.0F - epsilon), (128.0F - epsilon)); + result[0] = 1.0F; + result[1] = in[0]; + result[2] = powf(in[1], exponent); + result[3] = 1.0; + } + } + else + { + result[0] = 1.0F; + result[1] = 0.0; + result[2] = 0.0; + result[3] = 1.0F; + } +} + + + +static void PIPE_CDECL populate_lut( struct aos_machine *machine, + float *result, + const float *in, + unsigned count ) +{ + unsigned i, tab; + + /* Search for an existing table for this value. Note that without + * static analysis we don't really know if in[3] will be constant, + * but it usually is... + */ + for (tab = 0; tab < 4; tab++) { + if (machine->shine_tab[tab].exponent == in[3]) { + goto found; + } + } + + for (tab = 0, i = 1; i < 4; i++) { + if (machine->shine_tab[i].last_used < machine->shine_tab[tab].last_used) + tab = i; + } + + if (machine->shine_tab[tab].last_used == machine->now) { + /* No unused tables (this is not a ffvertex program...). Just + * call pow each time: + */ + machine->lit_info[count].func = do_lit; + machine->lit_info[count].func( machine, result, in, count ); + return; + } + else { + do_populate_lut( &machine->shine_tab[tab], in[3] ); + } + + found: + machine->shine_tab[tab].last_used = machine->now; + machine->lit_info[count].shine_tab = &machine->shine_tab[tab]; + machine->lit_info[count].func = do_lit_lut; + machine->lit_info[count].func( machine, result, in, count ); +} + + + static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { - struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); + struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; + unsigned lit_count = cp->lit_count++; + struct x86_reg result, arg0; + unsigned i; + +#if 1 + /* For absolute correctness, need to spill/invalidate all XMM regs + * too. + */ + for (i = 0; i < 8; i++) { + if (cp->xmm[i].dirty) + spill(cp, i); + aos_release_xmm_reg(cp, i); + } +#endif + + if (writemask != TGSI_WRITEMASK_XYZW) + result = x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, tmp[0])); + else + result = get_dst_ptr(cp, &op->FullDstRegisters[0]); + + + arg0 = fetch_src( cp, &op->FullSrcRegisters[0] ); + if (arg0.file == file_XMM) { + struct x86_reg tmp = x86_make_disp(cp->machine_EDX, + Offset(struct aos_machine, tmp[1])); + sse_movaps( cp->func, tmp, arg0 ); + arg0 = tmp; + } + + + + /* Push caller-save (ie scratch) regs. + */ + x86_cdecl_caller_push_regs( cp->func ); + + /* Push the arguments: + */ + x86_push_imm32( cp->func, lit_count ); + + x86_lea( cp->func, ecx, arg0 ); + x86_push( cp->func, ecx ); + x86_lea( cp->func, ecx, result ); + x86_push( cp->func, ecx ); + x86_push( cp->func, cp->machine_EDX ); + if (lit_count < MAX_LIT_INFO) { + x86_mov( cp->func, ecx, x86_make_disp( cp->machine_EDX, + Offset(struct aos_machine, lit_info) + + lit_count * sizeof(struct lit_info) + + Offset(struct lit_info, func))); + } + else { + x86_mov_reg_imm( cp->func, ecx, (int)do_lit ); + } + + x86_call( cp->func, ecx ); + + x86_pop( cp->func, ecx ); /* fixme... */ + x86_pop( cp->func, ecx ); + x86_pop( cp->func, ecx ); + x86_pop( cp->func, ecx ); + + x86_cdecl_caller_pop_regs( cp->func ); + + if (writemask != TGSI_WRITEMASK_XYZW) { + store_dest( cp, + &op->FullDstRegisters[0], + get_xmm_clone( cp, result ) ); + } + + return TRUE; +} + + +static boolean emit_inline_LIT( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); + unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; if (writemask & TGSI_WRITEMASK_YZ) { struct x86_reg st1 = x86_make_reg(file_x87, 1); struct x86_reg st2 = x86_make_reg(file_x87, 2); - - - /* a1' = a1 <= 0 ? 1 : a1; - * - * Note: use 1.0 to avoid passing zero to */ x87_fldz(cp->func); /* 1 0 */ #if 1 @@ -1865,6 +2096,14 @@ static void vaos_set_constants( struct draw_vs_varient *varient, constants[i][2], constants[i][3]); #endif + + { + unsigned i; + for (i = 0; i < MAX_LIT_INFO; i++) { + vaos->machine->lit_info[i].func = populate_lut; + vaos->machine->now++; + } + } } diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index a0680ec63d..c08c73d4bc 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -59,6 +59,25 @@ struct x86_function; #define FPU_RND_NEG 1 #define FPU_RND_NEAREST 2 +struct aos_machine; +typedef void PIPE_CDECL (*lit_func)( struct aos_machine *, + float *result, + const float *in, + unsigned count ); +struct shine_tab { + float exponent; + float values[258]; + unsigned last_used; +}; + +struct lit_info { + lit_func func; + struct shine_tab *shine_tab; +}; + +#define MAX_SHINE_TAB 4 +#define MAX_LIT_INFO 16 + /* This is the temporary storage used by all the aos_sse vs varients. * Create one per context and reuse by passing a pointer in at * vs_varient creation?? @@ -74,6 +93,13 @@ struct aos_machine { float scale[4]; /* viewport */ float translate[4]; /* viewport */ + float tmp[2][4]; /* scratch space for LIT */ + + struct shine_tab shine_tab[MAX_SHINE_TAB]; + struct lit_info lit_info[MAX_LIT_INFO]; + unsigned now; + + ushort fpu_rnd_nearest; ushort fpu_rnd_neg_inf; ushort fpu_restore; @@ -97,6 +123,7 @@ struct aos_compilation { unsigned insn_counter; unsigned num_immediates; unsigned count; + unsigned lit_count; struct { unsigned idx:16; -- cgit v1.2.3 From 48a24f0ff7e3aad000b8acc55c16bbeaca58abe6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 24 May 2008 16:32:08 +0100 Subject: Revert "mesa: save a temp on normalizes" This reverts commit feceb43948f76cc4d4c8ecbb86b1b1f438c6daee. --- src/mesa/main/ffvertex_prog.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 7a099b2376..a627a21f65 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -305,7 +305,7 @@ static struct state_key *make_state_key( GLcontext *ctx ) * generated program with line/function references for each * instruction back into this file: */ -#define DISASSEM 1 +#define DISASSEM (MESA_VERBOSE&VERBOSE_DISASSEM) /* Should be tunable by the driver - do we want to do matrix * multiplications with DP4's or with MUL/MAD's? SSE works better @@ -687,9 +687,11 @@ static void emit_normalize_vec3( struct tnl_program *p, struct ureg dest, struct ureg src ) { - emit_op2(p, OPCODE_DP3, dest, WRITEMASK_X, src, src); - emit_op1(p, OPCODE_RSQ, dest, WRITEMASK_X, dest); - emit_op2(p, OPCODE_MUL, dest, 0, src, swizzle1(dest, X)); + struct ureg tmp = get_temp(p); + emit_op2(p, OPCODE_DP3, tmp, WRITEMASK_X, src, src); + emit_op1(p, OPCODE_RSQ, tmp, WRITEMASK_X, tmp); + emit_op2(p, OPCODE_MUL, dest, 0, src, swizzle1(tmp, X)); + release_temp(p, tmp); } static void emit_passthrough( struct tnl_program *p, -- cgit v1.2.3 From a6fca8acb5e8ce0e5e6ce91a524e2bb4c180d3ac Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 25 May 2008 11:20:38 +0100 Subject: draw: fix input vs output typo in emit --- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index 0dda9df97d..f39ebb7a17 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -282,7 +282,7 @@ boolean aos_emit_outputs( struct aos_compilation *cp ) { unsigned i; - for (i = 0; i < cp->vaos->base.vs->info.num_inputs; i++) { + for (i = 0; i < cp->vaos->base.vs->info.num_outputs; i++) { unsigned format = cp->vaos->base.key.element[i].out.format; unsigned offset = cp->vaos->base.key.element[i].out.offset; -- cgit v1.2.3 From 584a3dcf8e4042cc1a5d48d83ea63d0a3c9706c1 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 25 May 2008 11:25:09 +0100 Subject: draw: add viewport support to generic vs varient code --- src/gallium/auxiliary/draw/draw_vs_varient.c | 43 ++++++++++++++++++++++++++-- 1 file changed, 41 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index f6f621a748..c15c648527 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -44,7 +44,7 @@ struct draw_vs_varient_generic { struct draw_vs_varient base; - + struct pipe_viewport_state viewport; struct draw_vertex_shader *shader; struct draw_context *draw; @@ -87,6 +87,30 @@ static void vsvg_set_input( struct draw_vs_varient *varient, } +/* Mainly for debug at this stage: + */ +static void do_viewport( struct draw_vs_varient_generic *vsvg, + unsigned count, + void *output_buffer ) +{ + char *ptr = (char *)output_buffer; + const float *scale = vsvg->viewport.scale; + const float *trans = vsvg->viewport.translate; + unsigned stride = vsvg->base.key.output_stride; + unsigned j; + + for (j = 0; j < count; j++, ptr += stride) { + float *data = (float *)ptr; + float w = 1.0f / data[3]; + + data[0] = data[0] * w * scale[0] + trans[0]; + data[1] = data[1] * w * scale[1] + trans[1]; + data[2] = data[2] * w * scale[2] + trans[2]; + data[3] = w; + } +} + + static void vsvg_run_elts( struct draw_vs_varient *varient, const unsigned *elts, unsigned count, @@ -112,6 +136,12 @@ static void vsvg_run_elts( struct draw_vs_varient *varient, vsvg->base.key.output_stride, vsvg->base.key.output_stride); + if (vsvg->base.key.viewport) + do_viewport( vsvg, + count, + output_buffer ); + + //if (!vsvg->already_in_emit_format) vsvg->emit->set_buffer( vsvg->emit, @@ -152,6 +182,12 @@ static void vsvg_run_linear( struct draw_vs_varient *varient, vsvg->base.key.output_stride, vsvg->base.key.output_stride); + if (vsvg->base.key.viewport) + do_viewport( vsvg, + count, + output_buffer ); + + //if (!vsvg->already_in_emit_format) vsvg->emit->set_buffer( vsvg->emit, 0, @@ -171,6 +207,9 @@ static void vsvg_run_linear( struct draw_vs_varient *varient, static void vsvg_set_viewport( struct draw_vs_varient *varient, const struct pipe_viewport_state *viewport ) { + struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; + + vsvg->viewport = *viewport; } static void vsvg_destroy( struct draw_vs_varient *varient ) @@ -185,7 +224,7 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, unsigned i; struct translate_key fetch, emit; - if (key->viewport || key->clip) + if (key->clip) return NULL; struct draw_vs_varient_generic *vsvg = CALLOC_STRUCT( draw_vs_varient_generic ); -- cgit v1.2.3 From dc52622fcf5660a9675ed61c359cf7068aa4861b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 25 May 2008 15:28:30 +0100 Subject: cso: use memcpy rather than structure asignment for copying Apparently gcc will omit to copy hidden padding bytes under some circumstances, which means later on memcmp() will indicate a difference between structs even though all the visible members are identical. --- src/gallium/auxiliary/cso_cache/cso_context.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index a1a3a9efaf..7236bff592 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -292,7 +292,7 @@ enum pipe_error cso_set_blend(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(&cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_blend_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_blend_state; cso->context = ctx->pipe; @@ -350,7 +350,7 @@ enum pipe_error cso_single_sampler(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(&cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state; cso->context = ctx->pipe; @@ -508,7 +508,7 @@ enum pipe_error cso_set_depth_stencil_alpha(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(&cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_depth_stencil_alpha_state; cso->context = ctx->pipe; @@ -564,7 +564,7 @@ enum pipe_error cso_set_rasterizer(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(&cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_rasterizer_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_rasterizer_state; cso->context = ctx->pipe; @@ -726,7 +726,7 @@ enum pipe_error cso_set_vertex_shader(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_vs_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_vs_state; cso->context = ctx->pipe; -- cgit v1.2.3 From caadc8d944c558e1fa9f23c3616d726337a19862 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 25 May 2008 15:37:47 +0100 Subject: draw: clean up some of the xmm register manipulation function names --- src/gallium/auxiliary/draw/draw_vs_aos.c | 141 +++++++++++++++++-------------- 1 file changed, 78 insertions(+), 63 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 1fbb7088ca..17b9442d6b 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -190,17 +190,26 @@ static void spill( struct aos_compilation *cp, unsigned idx ) } } -static boolean is_xmm_tmp( struct aos_compilation *cp, - struct x86_reg reg ) + +static struct x86_reg get_xmm_writable( struct aos_compilation *cp, + struct x86_reg reg ) { - return (reg.file == file_XMM && - cp->xmm[reg.idx].file == TGSI_FILE_NULL); + if (reg.file != file_XMM || + cp->xmm[reg.idx].file != TGSI_FILE_NULL) + { + struct x86_reg tmp = aos_get_xmm_reg(cp); + sse_movaps(cp->func, tmp, reg); + reg = tmp; + } + + return reg; } -static struct x86_reg get_xmm_clone( struct aos_compilation *cp, - struct x86_reg reg ) +static struct x86_reg get_xmm( struct aos_compilation *cp, + struct x86_reg reg ) { - if (!is_xmm_tmp(cp, reg)) { + if (reg.file != file_XMM) + { struct x86_reg tmp = aos_get_xmm_reg(cp); sse_movaps(cp->func, tmp, reg); reg = tmp; @@ -210,6 +219,9 @@ static struct x86_reg get_xmm_clone( struct aos_compilation *cp, } +/* Allocate an empty xmm register, either as a temporary or later to + * "adopt" as a shader reg. + */ struct x86_reg aos_get_xmm_reg( struct aos_compilation *cp ) { unsigned i; @@ -251,32 +263,11 @@ void aos_release_xmm_reg( struct aos_compilation *cp, cp->xmm[idx].last_used = 0; } -static void invalidate_xmm( struct aos_compilation *cp, - unsigned file, unsigned idx ) -{ - unsigned i; - - /* Invalidate any old copy of this register in XMM0-7. - */ - for (i = 0; i < 8; i++) { - if (cp->xmm[i].file == file && cp->xmm[i].idx == idx) { - - if (cp->xmm[i].dirty) - spill(cp, i); - - aos_release_xmm_reg(cp, i); - break; - } - } - for (; i < 8; i++) { - if (cp->xmm[i].file == file && cp->xmm[i].idx == idx) { - assert(0); - } - } -} - + +/* Mark an xmm reg as holding the current copy of a shader reg. + */ void aos_adopt_xmm_reg( struct aos_compilation *cp, struct x86_reg reg, unsigned file, @@ -290,6 +281,9 @@ void aos_adopt_xmm_reg( struct aos_compilation *cp, return; } + /* If any xmm reg thinks it holds this shader reg, break the + * illusion. + */ for (i = 0; i < 8; i++) { if (cp->xmm[i].file == file && cp->xmm[i].idx == idx) { @@ -304,12 +298,24 @@ void aos_adopt_xmm_reg( struct aos_compilation *cp, } - +/* Return a pointer to the in-memory copy of the reg, making sure it is uptodate. + */ static struct x86_reg aos_get_shader_reg_ptr( struct aos_compilation *cp, unsigned file, unsigned idx ) { - invalidate_xmm( cp, file, idx ); + unsigned i; + + /* Ensure the in-memory copy of this reg is up-to-date + */ + for (i = 0; i < 8; i++) { + if (cp->xmm[i].file == file && + cp->xmm[i].idx == idx && + cp->xmm[i].dirty) { + spill(cp, i); + } + } + return get_reg_ptr( cp, file, idx ); } @@ -320,7 +326,26 @@ static struct x86_reg aos_get_shader_reg_ptr( struct aos_compilation *cp, static struct x86_reg get_dst_ptr( struct aos_compilation *cp, const struct tgsi_full_dst_register *dst ) { - return aos_get_shader_reg_ptr( cp, dst->DstRegister.File, dst->DstRegister.Index ); + unsigned file = dst->DstRegister.File; + unsigned idx = dst->DstRegister.Index; + unsigned i; + + + /* Ensure in-memory copy of this reg is up-to-date and invalidate + * any xmm copies. + */ + for (i = 0; i < 8; i++) { + if (cp->xmm[i].file == file && + cp->xmm[i].idx == idx) + { + if (cp->xmm[i].dirty) + spill(cp, i); + + aos_release_xmm_reg(cp, i); + } + } + + return get_reg_ptr( cp, file, idx ); } @@ -358,15 +383,7 @@ static struct x86_reg aos_get_shader_reg_xmm( struct aos_compilation *cp, unsigned idx ) { struct x86_reg reg = aos_get_shader_reg( cp, file, idx ); - - if (reg.file != file_XMM) { - struct x86_reg tmp = aos_get_xmm_reg(cp); - sse_movaps(cp->func, tmp, reg); - aos_adopt_xmm_reg( cp, tmp, file, idx, FALSE ); - reg = tmp; - } - - return reg; + return get_xmm( cp, reg ); } @@ -634,7 +651,7 @@ static void store_dest( struct aos_compilation *cp, case TGSI_WRITEMASK_XYZW: aos_adopt_xmm_reg(cp, - get_xmm_clone(cp, result), + get_xmm_writable(cp, result), reg->DstRegister.File, reg->DstRegister.Index, TRUE); @@ -649,15 +666,15 @@ static void store_dest( struct aos_compilation *cp, switch (reg->DstRegister.WriteMask) { case TGSI_WRITEMASK_X: - sse_movss(cp->func, dst, get_xmm_clone(cp, result)); + sse_movss(cp->func, dst, get_xmm(cp, result)); break; case TGSI_WRITEMASK_XY: - sse_shufps(cp->func, dst, get_xmm_clone(cp, result), SHUF(X, Y, Z, W)); + sse_shufps(cp->func, dst, get_xmm(cp, result), SHUF(X, Y, Z, W)); break; case TGSI_WRITEMASK_ZW: - result = get_xmm_clone(cp, result); + result = get_xmm_writable(cp, result); sse_shufps(cp->func, result, dst, SHUF(X, Y, Z, W)); dst = result; break; @@ -845,7 +862,7 @@ static boolean emit_ABS( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_mulps(cp->func, dst, neg); sse_maxps(cp->func, dst, arg0); @@ -858,7 +875,7 @@ static boolean emit_ADD( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_addps(cp->func, dst, arg1); @@ -882,10 +899,9 @@ static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg tmp = aos_get_xmm_reg(cp); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_mulps(cp->func, dst, arg1); - /* Now the hard bit: sum the first 3 values: */ sse_movhlps(cp->func, tmp, dst); @@ -908,7 +924,7 @@ static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg tmp = aos_get_xmm_reg(cp); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_mulps(cp->func, dst, arg1); @@ -932,7 +948,7 @@ static boolean emit_DPH( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg tmp = aos_get_xmm_reg(cp); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_mulps(cp->func, dst, arg1); @@ -1366,7 +1382,7 @@ static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_inst if (writemask != TGSI_WRITEMASK_XYZW) { store_dest( cp, &op->FullDstRegisters[0], - get_xmm_clone( cp, result ) ); + get_xmm_writable( cp, result ) ); } return TRUE; @@ -1440,7 +1456,7 @@ static boolean emit_MAX( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_maxps(cp->func, dst, arg1); @@ -1453,7 +1469,7 @@ static boolean emit_MIN( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_minps(cp->func, dst, arg1); @@ -1464,7 +1480,7 @@ static boolean emit_MIN( struct aos_compilation *cp, const struct tgsi_full_inst static boolean emit_MOV( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); /* potentially nothing to do */ @@ -1476,7 +1492,7 @@ static boolean emit_MUL( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_mulps(cp->func, dst, arg1); @@ -1494,7 +1510,7 @@ static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_inst /* If we can't clobber old contents of arg0, get a temporary & copy * it there, then clobber it... */ - arg0 = get_xmm_clone(cp, arg0); + arg0 = get_xmm_writable(cp, arg0); sse_mulps(cp->func, arg0, arg1); sse_addps(cp->func, arg0, arg2); @@ -1562,7 +1578,7 @@ static boolean emit_SGE( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg ones = aos_get_internal(cp, IMM_ONES); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_cmpps(cp->func, dst, arg1, cc_NotLessThan); sse_andps(cp->func, dst, ones); @@ -1586,7 +1602,7 @@ static boolean emit_SLT( struct aos_compilation *cp, const struct tgsi_full_inst struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); struct x86_reg ones = aos_get_internal(cp, IMM_ONES); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_cmpps(cp->func, dst, arg1, cc_LessThan); sse_andps(cp->func, dst, ones); @@ -1599,7 +1615,7 @@ static boolean emit_SUB( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = get_xmm_clone(cp, arg0); + struct x86_reg dst = get_xmm_writable(cp, arg0); sse_subps(cp->func, dst, arg1); @@ -1989,7 +2005,6 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, /* decr count, loop if not zero */ x86_dec(cp.func, cp.count_ESI); -/* x86_test(cp.func, cp.count_ESI, cp.count_ESI); */ x86_jcc(cp.func, cc_NZ, label); restore_fpu_state(&cp); -- cgit v1.2.3 From ce331e3a5e2a0505e01637861bdd7f5e6cfbd041 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 25 May 2008 15:44:17 +0100 Subject: draw: special case for writing out scalar results --- src/gallium/auxiliary/draw/draw_vs_aos.c | 127 +++++++++++++++++++++++++------ 1 file changed, 102 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 17b9442d6b..aebc230858 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -697,6 +697,72 @@ static void store_dest( struct aos_compilation *cp, } +static void inject_scalar( struct aos_compilation *cp, + struct x86_reg dst, + struct x86_reg result, + unsigned swizzle ) +{ + sse_shufps(cp->func, dst, dst, swizzle); + sse_movss(cp->func, dst, result); + sse_shufps(cp->func, dst, dst, swizzle); +} + + +static void store_scalar_dest( struct aos_compilation *cp, + const struct tgsi_full_dst_register *reg, + struct x86_reg result ) +{ + unsigned writemask = reg->DstRegister.WriteMask; + struct x86_reg dst; + + if (writemask != TGSI_WRITEMASK_X && + writemask != TGSI_WRITEMASK_Y && + writemask != TGSI_WRITEMASK_Z && + writemask != TGSI_WRITEMASK_W && + writemask != 0) + { + result = get_xmm_writable(cp, result); /* already true, right? */ + sse_shufps(cp->func, result, result, SHUF(X,X,X,X)); + store_dest(cp, reg, result); + return; + } + + result = get_xmm(cp, result); + dst = aos_get_shader_reg_xmm(cp, + reg->DstRegister.File, + reg->DstRegister.Index); + + + + switch (reg->DstRegister.WriteMask) { + case TGSI_WRITEMASK_X: + sse_movss(cp->func, dst, result); + break; + + case TGSI_WRITEMASK_Y: + inject_scalar(cp, dst, result, SHUF(Y, X, Z, W)); + break; + + case TGSI_WRITEMASK_Z: + inject_scalar(cp, dst, result, SHUF(Z, Y, X, W)); + break; + + case TGSI_WRITEMASK_W: + inject_scalar(cp, dst, result, SHUF(W, Y, Z, X)); + break; + + default: + break; + } + + aos_adopt_xmm_reg(cp, + dst, + reg->DstRegister.File, + reg->DstRegister.Index, + TRUE); +} + + static void x87_fst_or_nop( struct x86_function *func, unsigned writemask, @@ -909,11 +975,8 @@ static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_inst emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z)); sse_addss(cp->func, dst, tmp); - if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); - aos_release_xmm_reg(cp, tmp.idx); - store_dest(cp, &op->FullDstRegisters[0], dst); + store_scalar_dest(cp, &op->FullDstRegisters[0], dst); return TRUE; } @@ -935,11 +998,8 @@ static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_inst emit_pshufd(cp, tmp, dst, SHUF(Y,X,W,Z)); sse_addss(cp->func, dst, tmp); - if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); - aos_release_xmm_reg(cp, tmp.idx); - store_dest(cp, &op->FullDstRegisters[0], dst); + store_scalar_dest(cp, &op->FullDstRegisters[0], dst); return TRUE; } @@ -961,11 +1021,8 @@ static boolean emit_DPH( struct aos_compilation *cp, const struct tgsi_full_inst emit_pshufd(cp, tmp, arg1, SHUF(W,W,W,W)); sse_addss(cp->func, dst, tmp); - if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); - aos_release_xmm_reg(cp, tmp.idx); - store_dest(cp, &op->FullDstRegisters[0], dst); + store_scalar_dest(cp, &op->FullDstRegisters[0], dst); return TRUE; } @@ -1518,7 +1575,9 @@ static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } - +/* Really not sufficient -- need to check for conditions that could + * generate inf/nan values, which will slow things down hugely. + */ static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { x87_fld_src(cp, &op->FullSrcRegisters[1], 0); /* a1.x */ @@ -1548,27 +1607,45 @@ static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_inst sse_divss(cp->func, dst, arg0); } - if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); - - store_dest(cp, &op->FullDstRegisters[0], dst); + store_scalar_dest(cp, &op->FullDstRegisters[0], dst); return TRUE; } + +/* Although rsqrtps() and rcpps() are low precision on some/all SSE + * implementations, it is possible to improve its precision at + * fairly low cost, using a newton/raphson step, as below: + * + * x1 = 2 * rcpps(a) - a * rcpps(a) * rcpps(a) + * x1 = 0.5 * rsqrtps(a) * [3.0 - (a * rsqrtps(a))* rsqrtps(a)] + * + * See: http://softwarecommunity.intel.com/articles/eng/1818.htm + */ static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg dst = aos_get_xmm_reg(cp); - sse_rsqrtss(cp->func, dst, arg0); - - /* Extend precision here... - */ - - if (op->FullDstRegisters[0].DstRegister.WriteMask != 0x1) - sse_shufps(cp->func, dst, dst, SHUF(X, X, X, X)); + if (1) { + sse_rsqrtss(cp->func, dst, arg0); + } + else { +#if 0 + /* Extend precision here... + */ + sse_movaps( func, dst, get_temp( TGSI_EXEC_TEMP_HALF_I, TGSI_EXEC_TEMP_HALF_C ) ); + sse_movaps( func, tmp0, get_temp( TGSI_EXEC_TEMP_THREE_I, TGSI_EXEC_TEMP_THREE_C ) ); + + sse_rsqrtss( func, tmp1, src ); /* rsqrtss(a) */ + sse_mulss( func, src, tmp1 ); /* a * rsqrtss(a) */ + sse_mulss( func, dst, tmp1 ); /* .5 * rsqrtss(a) */ + sse_mulss( func, src, tmp1 ); /* a * rsqrtss(a) * rsqrtss(a) */ + sse_subss( func, tmp0, src ); /* 3.0 - (a * rsqrtss(a) * rsqrtss(a)) */ + sse_mulss( func, dst, tmp0 ); /* .5 * r * (3.0 - (a * r * r)) */ +#endif + } - store_dest(cp, &op->FullDstRegisters[0], dst); + store_scalar_dest(cp, &op->FullDstRegisters[0], dst); return TRUE; } -- cgit v1.2.3 From 3afb7198e01516dba38bb3248d4c0161e54650fe Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 25 May 2008 15:45:27 +0100 Subject: draw: remove EXP & LOG from vs_aos.c These don't get hit & look like bug magnets to me... --- src/gallium/auxiliary/draw/draw_vs_aos.c | 85 ++------------------------------ 1 file changed, 4 insertions(+), 81 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index aebc230858..34dc09ead7 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1066,85 +1066,6 @@ static boolean emit_EX2( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } -static boolean emit_EXP( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) -{ - struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); - struct x86_reg st0 = x86_make_reg(file_x87, 0); - struct x86_reg st1 = x86_make_reg(file_x87, 1); - struct x86_reg st3 = x86_make_reg(file_x87, 3); - unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; - - /* CAUTION: dst may alias arg0! - */ - x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* arg0.x */ - x87_fld(cp->func, st0); /* arg arg */ - - /* by default, fpu is setup to round-to-nearest. We want to - * change this now, and track the state through to the end of the - * generated function so that it isn't repeated unnecessarily. - * Alternately, could subtract .5 to get round to -inf behaviour. - */ - set_fpu_round_neg_inf( cp ); - x87_fprndint( cp->func ); /* flr(a) a */ - x87_fld(cp->func, st0); /* flr(a) flr(a) a */ - x87_fld1(cp->func); /* 1 floor(a) floor(a) a */ - x87_fst_or_nop(cp->func, writemask, 3, dst); /* stack unchanged */ - - x87_fscale(cp->func); /* 2^floor(a) floor(a) a */ - x87_fst(cp->func, st3); /* 2^floor(a) floor(a) a 2^floor(a)*/ - - x87_fstp_or_pop(cp->func, writemask, 0, dst); /* flr(a) a 2^flr(a) */ - - x87_fsubp(cp->func, st1); /* frac(a) 2^flr(a) */ - - x87_fst_or_nop(cp->func, writemask, 1, dst); /* frac(a) 2^flr(a) */ - - x87_f2xm1(cp->func); /* (2^frac(a))-1 2^flr(a)*/ - x87_fld1(cp->func); /* 1 (2^frac(a))-1 2^flr(a)*/ - x87_faddp(cp->func, st1); /* 2^frac(a) 2^flr(a) */ - x87_fmulp(cp->func, st1); /* 2^a */ - - x87_fstp_or_pop(cp->func, writemask, 2, dst); - -/* dst[0] = 2^floor(tmp); */ -/* dst[1] = frac(tmp); */ -/* dst[2] = 2^floor(tmp) * 2^frac(tmp); */ -/* dst[3] = 1.0F; */ - return TRUE; -} - -static boolean emit_LOG( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) -{ - struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); - struct x86_reg st0 = x86_make_reg(file_x87, 0); - struct x86_reg st1 = x86_make_reg(file_x87, 1); - struct x86_reg st2 = x86_make_reg(file_x87, 2); - unsigned writemask = op->FullDstRegisters[0].DstRegister.WriteMask; - - /* CAUTION: dst may alias arg0! - */ - x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* arg0.x */ - x87_fabs(cp->func); /* |arg0.x| */ - x87_fxtract(cp->func); /* mantissa(arg0.x), exponent(arg0.x) */ - x87_fst(cp->func, st2); /* mantissa, exponent, mantissa */ - x87_fld1(cp->func); /* 1, mantissa, exponent, mantissa */ - x87_fyl2x(cp->func); /* log2(mantissa), exponent, mantissa */ - x87_fadd(cp->func, st0, st1); /* e+l2(m), e, m */ - - x87_fstp_or_pop(cp->func, writemask, 2, dst); /* e, m */ - - x87_fld1(cp->func); /* 1, e, m */ - x87_fsub(cp->func, st1, st0); /* 1, e-1, m */ - - x87_fstp_or_pop(cp->func, writemask, 3, dst); /* e-1,m */ - x87_fstp_or_pop(cp->func, writemask, 0, dst); /* m */ - - x87_fadd(cp->func, st0, st0); /* 2m */ - - x87_fstp_or_pop( cp->func, writemask, 1, dst ); - - return TRUE; -} static boolean emit_FLR( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { @@ -1755,10 +1676,12 @@ emit_instruction( struct aos_compilation *cp, return emit_RSQ(cp, inst); case TGSI_OPCODE_EXP: - return emit_EXP(cp, inst); + /*return emit_EXP(cp, inst);*/ + return FALSE; case TGSI_OPCODE_LOG: - return emit_LOG(cp, inst); + /*return emit_LOG(cp, inst);*/ + return FALSE; case TGSI_OPCODE_MUL: return emit_MUL(cp, inst); -- cgit v1.2.3 From 9c7568965c00dcc2e9403a2f94f1cd09dcd783ae Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 25 May 2008 15:47:04 +0100 Subject: draw: slight tweak for XPD opcode --- src/gallium/auxiliary/draw/draw_vs_aos.c | 21 +++++++-------------- 1 file changed, 7 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 34dc09ead7..37d04e45a6 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1626,31 +1626,24 @@ static boolean emit_XPD( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg arg1 = fetch_src(cp, &op->FullSrcRegisters[1]); - struct x86_reg dst = aos_get_xmm_reg(cp); struct x86_reg tmp0 = aos_get_xmm_reg(cp); struct x86_reg tmp1 = aos_get_xmm_reg(cp); - /* Could avoid tmp0, tmp1 if we overwrote arg0, arg1. Need a way - * to invalidate registers. This will come with better analysis - * (liveness analysis) of the incoming program. - */ - emit_pshufd(cp, dst, arg0, SHUF(Y, Z, X, W)); - emit_pshufd(cp, tmp1, arg1, SHUF(Z, X, Y, W)); - sse_mulps(cp->func, dst, tmp1); - emit_pshufd(cp, tmp0, arg0, SHUF(Z, X, Y, W)); emit_pshufd(cp, tmp1, arg1, SHUF(Y, Z, X, W)); - sse_mulps(cp->func, tmp0, tmp1); - sse_subps(cp->func, dst, tmp0); + sse_mulps(cp->func, tmp1, arg0); + emit_pshufd(cp, tmp0, arg0, SHUF(Y, Z, X, W)); + sse_mulps(cp->func, tmp0, arg1); + sse_subps(cp->func, tmp1, tmp0); + sse_shufps(cp->func, tmp1, tmp1, SHUF(Y, Z, X, W)); +/* dst[2] = arg0[0] * arg1[1] - arg0[1] * arg1[0]; */ /* dst[0] = arg0[1] * arg1[2] - arg0[2] * arg1[1]; */ /* dst[1] = arg0[2] * arg1[0] - arg0[0] * arg1[2]; */ -/* dst[2] = arg0[0] * arg1[1] - arg0[1] * arg1[0]; */ /* dst[3] is undef */ aos_release_xmm_reg(cp, tmp0.idx); - aos_release_xmm_reg(cp, tmp1.idx); - store_dest(cp, &op->FullDstRegisters[0], dst); + store_dest(cp, &op->FullDstRegisters[0], tmp1); return TRUE; } -- cgit v1.2.3 From 359058e7b77ddbac5eec7e8d1c77232bcbb1adbf Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 25 May 2008 15:28:30 +0100 Subject: cso: use memcpy rather than structure asignment for copying Apparently gcc will omit to copy hidden padding bytes under some circumstances, which means later on memcmp() will indicate a difference between structs even though all the visible members are identical. --- src/gallium/auxiliary/cso_cache/cso_context.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index dc5987df44..af4af8ac1d 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -292,7 +292,7 @@ enum pipe_error cso_set_blend(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(&cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_blend_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_blend_state; cso->context = ctx->pipe; @@ -350,7 +350,7 @@ enum pipe_error cso_single_sampler(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(&cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_sampler_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_sampler_state; cso->context = ctx->pipe; @@ -508,7 +508,7 @@ enum pipe_error cso_set_depth_stencil_alpha(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(&cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_depth_stencil_alpha_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_depth_stencil_alpha_state; cso->context = ctx->pipe; @@ -564,7 +564,7 @@ enum pipe_error cso_set_rasterizer(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(&cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_rasterizer_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_rasterizer_state; cso->context = ctx->pipe; @@ -726,7 +726,7 @@ enum pipe_error cso_set_vertex_shader(struct cso_context *ctx, if (!cso) return PIPE_ERROR_OUT_OF_MEMORY; - cso->state = *templ; + memcpy(cso->state, templ, sizeof(*templ)); cso->data = ctx->pipe->create_vs_state(ctx->pipe, &cso->state); cso->delete_state = (cso_state_callback)ctx->pipe->delete_vs_state; cso->context = ctx->pipe; -- cgit v1.2.3 From 721fb5597e687fc1446119002ab03cc428104b29 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 26 May 2008 00:09:02 +0100 Subject: draw: more aos tweaks --- src/gallium/auxiliary/draw/draw_vs_aos.c | 26 ++++++++++++++++++++++++-- 1 file changed, 24 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 37d04e45a6..916203c66b 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -202,6 +202,7 @@ static struct x86_reg get_xmm_writable( struct aos_compilation *cp, reg = tmp; } + cp->xmm[reg.idx].last_used = cp->insn_counter; return reg; } @@ -215,6 +216,7 @@ static struct x86_reg get_xmm( struct aos_compilation *cp, reg = tmp; } + cp->xmm[reg.idx].last_used = cp->insn_counter; return reg; } @@ -281,6 +283,18 @@ void aos_adopt_xmm_reg( struct aos_compilation *cp, return; } + /* If this xmm reg is already holding this shader reg, just update + * last_used, and don't clobber the dirty flag... + */ + if (cp->xmm[reg.idx].file == file && + cp->xmm[reg.idx].idx == idx) + { + cp->xmm[reg.idx].dirty |= dirty; + cp->xmm[reg.idx].last_used = cp->insn_counter; + return; + } + + /* If any xmm reg thinks it holds this shader reg, break the * illusion. */ @@ -382,8 +396,16 @@ static struct x86_reg aos_get_shader_reg_xmm( struct aos_compilation *cp, unsigned file, unsigned idx ) { - struct x86_reg reg = aos_get_shader_reg( cp, file, idx ); - return get_xmm( cp, reg ); + struct x86_reg reg = get_xmm( cp, + aos_get_shader_reg( cp, file, idx ) ); + + aos_adopt_xmm_reg( cp, + reg, + file, + idx, + FALSE ); + + return reg; } -- cgit v1.2.3 From 351eca365c0ba488000c3826d5093de6170381e4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 26 May 2008 11:03:00 +0100 Subject: draw: extend precision in RSQ opcode --- src/gallium/auxiliary/draw/draw_vs_aos.c | 48 ++++++++++++++++++-------------- src/gallium/auxiliary/draw/draw_vs_aos.h | 1 + 2 files changed, 28 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 916203c66b..1622358ae1 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -140,7 +140,8 @@ static void init_internals( struct aos_machine *machine ) ASSIGN_4V(machine->internal[IMM_NEGS], -1.0f, -1.0f, -1.0f, -1.0f); ASSIGN_4V(machine->internal[IMM_IDENTITY], 0.0f, 0.0f, 0.0f, 1.0f); ASSIGN_4V(machine->internal[IMM_INV_255], inv, inv, inv, inv); - ASSIGN_4V(machine->internal[IMM_255], f255, f255, f255, f255); + ASSIGN_4V(machine->internal[IMM_255], f255, f255, f255, f255); + ASSIGN_4V(machine->internal[IMM_RSQ], -.5f, 1.5f, 0.0f, 0.0f); machine->fpu_rnd_nearest = (X87_CW_EXCEPTION_INV_OP | @@ -1561,35 +1562,40 @@ static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_inst * * x1 = 2 * rcpps(a) - a * rcpps(a) * rcpps(a) * x1 = 0.5 * rsqrtps(a) * [3.0 - (a * rsqrtps(a))* rsqrtps(a)] + * or: + * x1 = rsqrtps(a) * [1.5 - .5 * a * rsqrtps(a) * rsqrtps(a)] + * * * See: http://softwarecommunity.intel.com/articles/eng/1818.htm */ static boolean emit_RSQ( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { - struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); - struct x86_reg dst = aos_get_xmm_reg(cp); - if (1) { - sse_rsqrtss(cp->func, dst, arg0); + if (0) { + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg r = aos_get_xmm_reg(cp); + sse_rsqrtss(cp->func, r, arg0); + store_scalar_dest(cp, &op->FullDstRegisters[0], r); + return TRUE; } else { -#if 0 - /* Extend precision here... - */ - sse_movaps( func, dst, get_temp( TGSI_EXEC_TEMP_HALF_I, TGSI_EXEC_TEMP_HALF_C ) ); - sse_movaps( func, tmp0, get_temp( TGSI_EXEC_TEMP_THREE_I, TGSI_EXEC_TEMP_THREE_C ) ); - - sse_rsqrtss( func, tmp1, src ); /* rsqrtss(a) */ - sse_mulss( func, src, tmp1 ); /* a * rsqrtss(a) */ - sse_mulss( func, dst, tmp1 ); /* .5 * rsqrtss(a) */ - sse_mulss( func, src, tmp1 ); /* a * rsqrtss(a) * rsqrtss(a) */ - sse_subss( func, tmp0, src ); /* 3.0 - (a * rsqrtss(a) * rsqrtss(a)) */ - sse_mulss( func, dst, tmp0 ); /* .5 * r * (3.0 - (a * r * r)) */ -#endif - } + struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); + struct x86_reg r = aos_get_xmm_reg(cp); - store_scalar_dest(cp, &op->FullDstRegisters[0], dst); - return TRUE; + struct x86_reg neg_half = get_reg_ptr( cp, AOS_FILE_INTERNAL, IMM_RSQ ); + struct x86_reg one_point_five = x86_make_disp( neg_half, 4 ); + struct x86_reg src = get_xmm_writable( cp, arg0 ); + + sse_rsqrtss( cp->func, r, src ); /* rsqrtss(a) */ + sse_mulss( cp->func, src, neg_half ); /* -.5 * a */ + sse_mulss( cp->func, src, r ); /* -.5 * a * r */ + sse_mulss( cp->func, src, r ); /* -.5 * a * r * r */ + sse_addss( cp->func, src, one_point_five ); /* 1.5 - .5 * a * r * r */ + sse_mulss( cp->func, r, src ); /* r * (1.5 - .5 * a * r * r) */ + + store_scalar_dest(cp, &op->FullDstRegisters[0], r); + return TRUE; + } } diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index c08c73d4bc..fffe2e4658 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -175,6 +175,7 @@ boolean aos_emit_outputs( struct aos_compilation *cp ); #define IMM_INV_255 3 /* 1/255, 1/255, 1/255, 1/255 */ #define IMM_255 4 /* 255, 255, 255, 255 */ #define IMM_NEGS 5 /* -1,-1,-1,-1 */ +#define IMM_RSQ 6 /* -.5,1.5,_,_ */ struct x86_reg aos_get_internal( struct aos_compilation *cp, unsigned imm ); -- cgit v1.2.3 From e0fd3449f824ed8eaec49e83a0f90b7fe47e09a6 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Sun, 25 May 2008 17:26:07 +0200 Subject: i915: Fix for tex-surface merge --- src/gallium/drivers/i915simple/i915_texture.c | 29 +++++++++++++++++++++++++-- 1 file changed, 27 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 3e23e540f9..b08385e7db 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -132,11 +132,11 @@ i915_displaytarget_layout(struct pipe_screen *screen, /* Now extract the goodies: */ - i915_miptree_set_image_offset( tex, 0, 0, 0, 0 ); - i915_miptree_set_level_info( tex, 0, 0, 0, 0, + i915_miptree_set_level_info( tex, 0, 1, 0, 0, tex->base.width[0], tex->base.height[0], 1 ); + i915_miptree_set_image_offset( tex, 0, 0, 0, 0 ); tex->buffer = surf.buffer; tex->pitch = surf.pitch; @@ -633,6 +633,7 @@ i915_get_tex_surface_screen(struct pipe_screen *screen, if (ps) { assert(ps->refcount); assert(ps->winsys); + pipe_texture_reference(&ps->texture, pt); pipe_buffer_reference(ws, &ps->buffer, tex->buffer); ps->format = pt->format; ps->cpp = pt->cpp; @@ -652,6 +653,29 @@ i915_init_texture_functions(struct i915_context *i915) // i915->pipe.texture_update = i915_texture_update; } +static void +i915_tex_surface_release_screen(struct pipe_screen *screen, + struct pipe_surface **surface) +{ + struct pipe_surface *surf = *surface; + + if (--surf->refcount == 0) { + + /* This really should not be possible, but it's actually + * happening quite a bit... Will fix. + */ + if (surf->status == PIPE_SURFACE_STATUS_CLEAR) { + debug_printf("XXX destroying a surface with pending clears...\n"); + assert(0); + } + + pipe_texture_reference(&surf->texture, NULL); + pipe_buffer_reference(screen->winsys, &surf->buffer, NULL); + FREE(surf); + } + + *surface = NULL; +} void i915_init_screen_texture_functions(struct pipe_screen *screen) @@ -659,4 +683,5 @@ i915_init_screen_texture_functions(struct pipe_screen *screen) screen->texture_create = i915_texture_create_screen; screen->texture_release = i915_texture_release_screen; screen->get_tex_surface = i915_get_tex_surface_screen; + screen->tex_surface_release = i915_tex_surface_release_screen; } -- cgit v1.2.3 From 791eee64e03c7323c8a8907f54b09a015c046e2f Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 26 May 2008 11:20:51 +0200 Subject: i915: Removed screen sufixes on texture functions --- src/gallium/drivers/i915simple/i915_texture.c | 28 +++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index b08385e7db..df11ba0544 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -525,8 +525,8 @@ i945_miptree_layout(struct i915_texture * tex) static struct pipe_texture * -i915_texture_create_screen(struct pipe_screen *screen, - const struct pipe_texture *templat) +i915_texture_create(struct pipe_screen *screen, + const struct pipe_texture *templat) { struct i915_screen *i915screen = i915_screen(screen); struct pipe_winsys *ws = screen->winsys; @@ -571,8 +571,8 @@ i915_texture_create_screen(struct pipe_screen *screen, static void -i915_texture_release_screen(struct pipe_screen *screen, - struct pipe_texture **pt) +i915_texture_release(struct pipe_screen *screen, + struct pipe_texture **pt) { if (!*pt) return; @@ -606,10 +606,10 @@ i915_texture_release_screen(struct pipe_screen *screen, * XXX note: same as code in sp_surface.c */ static struct pipe_surface * -i915_get_tex_surface_screen(struct pipe_screen *screen, - struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice, - unsigned flags) +i915_get_tex_surface(struct pipe_screen *screen, + struct pipe_texture *pt, + unsigned face, unsigned level, unsigned zslice, + unsigned flags) { struct i915_texture *tex = (struct i915_texture *)pt; struct pipe_winsys *ws = screen->winsys; @@ -654,8 +654,8 @@ i915_init_texture_functions(struct i915_context *i915) } static void -i915_tex_surface_release_screen(struct pipe_screen *screen, - struct pipe_surface **surface) +i915_tex_surface_release(struct pipe_screen *screen, + struct pipe_surface **surface) { struct pipe_surface *surf = *surface; @@ -680,8 +680,8 @@ i915_tex_surface_release_screen(struct pipe_screen *screen, void i915_init_screen_texture_functions(struct pipe_screen *screen) { - screen->texture_create = i915_texture_create_screen; - screen->texture_release = i915_texture_release_screen; - screen->get_tex_surface = i915_get_tex_surface_screen; - screen->tex_surface_release = i915_tex_surface_release_screen; + screen->texture_create = i915_texture_create; + screen->texture_release = i915_texture_release; + screen->get_tex_surface = i915_get_tex_surface; + screen->tex_surface_release = i915_tex_surface_release; } -- cgit v1.2.3 From 77ce568ff704e6cdcfaa557965c894752d19e462 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 26 May 2008 20:14:40 +0900 Subject: Remove CVS keywords. --- src/gallium/winsys/dri/intel/server/i830_common.h | 1 - src/gallium/winsys/dri/intel/server/i830_dri.h | 1 - src/glu/mini/all.h | 1 - src/glu/mini/glu.c | 1 - src/glu/mini/gluP.h | 1 - src/glu/mini/mipmap.c | 1 - src/glu/mini/nurbs.c | 1 - src/glu/mini/nurbs.h | 1 - src/glu/mini/nurbscrv.c | 1 - src/glu/mini/polytest.c | 1 - src/glu/mini/project.c | 1 - src/glu/mini/quadric.c | 1 - src/glu/mini/tess.c | 1 - src/glu/mini/tess.h | 1 - src/glu/mini/tesselat.c | 1 - src/glu/sgi/dummy.cc | 1 - src/glu/sgi/libnurbs/interface/bezierEval.h | 2 -- src/glu/sgi/libnurbs/interface/bezierPatch.cc | 2 -- src/glu/sgi/libnurbs/interface/bezierPatch.h | 2 -- src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc | 2 -- src/glu/sgi/libnurbs/interface/bezierPatchMesh.h | 2 -- src/glu/sgi/libnurbs/interface/glcurveval.cc | 2 -- src/glu/sgi/libnurbs/interface/glimports.h | 2 -- src/glu/sgi/libnurbs/interface/glinterface.cc | 2 -- src/glu/sgi/libnurbs/interface/glrenderer.h | 2 -- src/glu/sgi/libnurbs/interface/incurveeval.cc | 2 -- src/glu/sgi/libnurbs/interface/insurfeval.cc | 2 -- src/glu/sgi/libnurbs/interface/mystdio.h | 2 -- src/glu/sgi/libnurbs/interface/mystdlib.h | 2 -- src/glu/sgi/libnurbs/internals/arc.h | 2 -- src/glu/sgi/libnurbs/internals/arcsorter.cc | 2 -- src/glu/sgi/libnurbs/internals/arcsorter.h | 2 -- src/glu/sgi/libnurbs/internals/arctess.h | 2 -- src/glu/sgi/libnurbs/internals/backend.cc | 2 -- src/glu/sgi/libnurbs/internals/backend.h | 2 -- src/glu/sgi/libnurbs/internals/basiccrveval.h | 2 -- src/glu/sgi/libnurbs/internals/basicsurfeval.h | 2 -- src/glu/sgi/libnurbs/internals/bezierarc.h | 2 -- src/glu/sgi/libnurbs/internals/bin.cc | 2 -- src/glu/sgi/libnurbs/internals/bin.h | 2 -- src/glu/sgi/libnurbs/internals/bufpool.cc | 2 -- src/glu/sgi/libnurbs/internals/bufpool.h | 2 -- src/glu/sgi/libnurbs/internals/cachingeval.cc | 2 -- src/glu/sgi/libnurbs/internals/cachingeval.h | 2 -- src/glu/sgi/libnurbs/internals/ccw.cc | 2 -- src/glu/sgi/libnurbs/internals/coveandtiler.h | 2 -- src/glu/sgi/libnurbs/internals/curve.cc | 2 -- src/glu/sgi/libnurbs/internals/curve.h | 2 -- src/glu/sgi/libnurbs/internals/curvelist.cc | 2 -- src/glu/sgi/libnurbs/internals/curvelist.h | 2 -- src/glu/sgi/libnurbs/internals/curvesub.cc | 2 -- src/glu/sgi/libnurbs/internals/dataTransform.cc | 2 -- src/glu/sgi/libnurbs/internals/dataTransform.h | 2 -- src/glu/sgi/libnurbs/internals/defines.h | 2 -- src/glu/sgi/libnurbs/internals/displaylist.cc | 2 -- src/glu/sgi/libnurbs/internals/displaylist.h | 2 -- src/glu/sgi/libnurbs/internals/displaymode.h | 2 -- src/glu/sgi/libnurbs/internals/flist.cc | 2 -- src/glu/sgi/libnurbs/internals/flist.h | 2 -- src/glu/sgi/libnurbs/internals/flistsorter.cc | 2 -- src/glu/sgi/libnurbs/internals/flistsorter.h | 2 -- src/glu/sgi/libnurbs/internals/gridline.h | 2 -- src/glu/sgi/libnurbs/internals/gridtrimvertex.h | 2 -- src/glu/sgi/libnurbs/internals/gridvertex.h | 2 -- src/glu/sgi/libnurbs/internals/hull.cc | 2 -- src/glu/sgi/libnurbs/internals/hull.h | 2 -- src/glu/sgi/libnurbs/internals/intersect.cc | 2 -- src/glu/sgi/libnurbs/internals/jarcloc.h | 2 -- src/glu/sgi/libnurbs/internals/knotvector.h | 2 -- src/glu/sgi/libnurbs/internals/mapdesc.cc | 2 -- src/glu/sgi/libnurbs/internals/mapdesc.h | 2 -- src/glu/sgi/libnurbs/internals/mapdescv.cc | 2 -- src/glu/sgi/libnurbs/internals/maplist.cc | 2 -- src/glu/sgi/libnurbs/internals/maplist.h | 2 -- src/glu/sgi/libnurbs/internals/mesher.cc | 2 -- src/glu/sgi/libnurbs/internals/mesher.h | 2 -- src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc | 2 -- src/glu/sgi/libnurbs/internals/monotonizer.cc | 2 -- src/glu/sgi/libnurbs/internals/monotonizer.h | 1 - src/glu/sgi/libnurbs/internals/myassert.h | 2 -- src/glu/sgi/libnurbs/internals/mycode.cc | 2 -- src/glu/sgi/libnurbs/internals/mystring.h | 2 -- src/glu/sgi/libnurbs/internals/nurbsconsts.h | 2 -- src/glu/sgi/libnurbs/internals/nurbstess.cc | 2 -- src/glu/sgi/libnurbs/internals/patch.cc | 2 -- src/glu/sgi/libnurbs/internals/patch.h | 2 -- src/glu/sgi/libnurbs/internals/patchlist.cc | 2 -- src/glu/sgi/libnurbs/internals/patchlist.h | 2 -- src/glu/sgi/libnurbs/internals/pwlarc.h | 2 -- src/glu/sgi/libnurbs/internals/quilt.cc | 2 -- src/glu/sgi/libnurbs/internals/quilt.h | 2 -- src/glu/sgi/libnurbs/internals/reader.cc | 2 -- src/glu/sgi/libnurbs/internals/reader.h | 2 -- src/glu/sgi/libnurbs/internals/renderhints.cc | 2 -- src/glu/sgi/libnurbs/internals/renderhints.h | 2 -- src/glu/sgi/libnurbs/internals/simplemath.h | 2 -- src/glu/sgi/libnurbs/internals/slicer.cc | 2 -- src/glu/sgi/libnurbs/internals/slicer.h | 2 -- src/glu/sgi/libnurbs/internals/sorter.cc | 2 -- src/glu/sgi/libnurbs/internals/sorter.h | 2 -- src/glu/sgi/libnurbs/internals/splitarcs.cc | 2 -- src/glu/sgi/libnurbs/internals/subdivider.h | 2 -- src/glu/sgi/libnurbs/internals/tobezier.cc | 2 -- src/glu/sgi/libnurbs/internals/trimline.cc | 2 -- src/glu/sgi/libnurbs/internals/trimline.h | 2 -- src/glu/sgi/libnurbs/internals/trimregion.cc | 2 -- src/glu/sgi/libnurbs/internals/trimregion.h | 2 -- src/glu/sgi/libnurbs/internals/trimvertex.h | 2 -- src/glu/sgi/libnurbs/internals/trimvertpool.cc | 2 -- src/glu/sgi/libnurbs/internals/trimvertpool.h | 2 -- src/glu/sgi/libnurbs/internals/types.h | 2 -- src/glu/sgi/libnurbs/internals/uarray.cc | 2 -- src/glu/sgi/libnurbs/internals/uarray.h | 2 -- src/glu/sgi/libnurbs/internals/varray.cc | 2 -- src/glu/sgi/libnurbs/internals/varray.h | 2 -- src/glu/sgi/libnurbs/nurbtess/definitions.h | 2 -- src/glu/sgi/libnurbs/nurbtess/directedLine.h | 2 -- src/glu/sgi/libnurbs/nurbtess/glimports.h | 2 -- src/glu/sgi/libnurbs/nurbtess/gridWrap.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/gridWrap.h | 2 -- src/glu/sgi/libnurbs/nurbtess/monoChain.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/monoChain.h | 2 -- src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc | 1 - src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h | 1 - src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h | 2 -- src/glu/sgi/libnurbs/nurbtess/mystdio.h | 2 -- src/glu/sgi/libnurbs/nurbtess/mystdlib.h | 2 -- src/glu/sgi/libnurbs/nurbtess/partitionX.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/partitionX.h | 2 -- src/glu/sgi/libnurbs/nurbtess/partitionY.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/partitionY.h | 2 -- src/glu/sgi/libnurbs/nurbtess/polyDBG.h | 2 -- src/glu/sgi/libnurbs/nurbtess/polyUtil.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/polyUtil.h | 2 -- src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/primitiveStream.h | 2 -- src/glu/sgi/libnurbs/nurbtess/quicksort.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/quicksort.h | 2 -- src/glu/sgi/libnurbs/nurbtess/rectBlock.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/rectBlock.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleComp.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleComp.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampledLine.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampledLine.h | 2 -- src/glu/sgi/libnurbs/nurbtess/searchTree.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/searchTree.h | 2 -- src/glu/sgi/libnurbs/nurbtess/zlassert.h | 2 -- src/glu/sgi/libtess/README | 1 - src/glu/sgi/libtess/alg-outline | 1 - src/glu/sgi/libtess/dict-list.h | 2 -- src/glu/sgi/libtess/dict.c | 2 -- src/glu/sgi/libtess/dict.h | 2 -- src/glu/sgi/libtess/geom.c | 2 -- src/glu/sgi/libtess/memalloc.c | 2 -- src/glu/sgi/libtess/memalloc.h | 2 -- src/glu/sgi/libtess/mesh.c | 2 -- src/glu/sgi/libtess/mesh.h | 2 -- src/glu/sgi/libtess/normal.h | 2 -- src/glu/sgi/libtess/priorityq-heap.c | 2 -- src/glu/sgi/libtess/priorityq-heap.h | 2 -- src/glu/sgi/libtess/priorityq-sort.h | 2 -- src/glu/sgi/libtess/priorityq.c | 2 -- src/glu/sgi/libtess/priorityq.h | 2 -- src/glu/sgi/libtess/render.c | 2 -- src/glu/sgi/libtess/render.h | 2 -- src/glu/sgi/libtess/sweep.h | 2 -- src/glu/sgi/libtess/tess.h | 2 -- src/glu/sgi/libtess/tessmono.c | 2 -- src/glu/sgi/libtess/tessmono.h | 2 -- src/glu/sgi/libutil/error.c | 2 -- src/glu/sgi/libutil/glue.c | 2 -- src/glu/sgi/libutil/gluint.h | 2 -- src/glu/sgi/libutil/project.c | 2 -- src/glu/sgi/libutil/registry.c | 2 -- src/glut/beos/beos_x11.cpp | 1 - src/glut/ggi/debug.h | 2 +- src/glut/glx/stroke.h | 1 - src/glut/glx/win32_x11.c | 1 - src/glx/mini/miniglx_events.c | 1 - src/glx/x11/XF86dri.c | 1 - src/glx/x11/clientattrib.c | 1 - src/glx/x11/compsize.c | 1 - src/glx/x11/dri_glx.c | 1 - src/glx/x11/eval.c | 1 - src/glx/x11/glxclient.h | 1 - src/glx/x11/glxcmds.c | 1 - src/glx/x11/glxext.c | 1 - src/glx/x11/indirect_init.h | 1 - src/glx/x11/packrender.h | 1 - src/glx/x11/packsingle.h | 1 - src/glx/x11/pixel.c | 1 - src/glx/x11/pixelstore.c | 1 - src/glx/x11/render2.c | 1 - src/glx/x11/renderpix.c | 1 - src/glx/x11/single2.c | 1 - src/glx/x11/singlepix.c | 1 - src/glx/x11/vertarr.c | 1 - src/glx/x11/xf86dri.h | 1 - src/glx/x11/xf86dristr.h | 1 - src/glx/x11/xfont.c | 1 - src/mesa/drivers/dri/common/stenciltmp.h | 1 - src/mesa/drivers/dri/common/texmem.c | 1 - src/mesa/drivers/dri/common/texmem.h | 1 - src/mesa/drivers/dri/common/utils.h | 1 - src/mesa/drivers/dri/common/vblank.c | 1 - src/mesa/drivers/dri/common/vblank.h | 1 - src/mesa/drivers/dri/ffb/ffb_bitmap.c | 2 +- src/mesa/drivers/dri/ffb/ffb_bitmap.h | 1 - src/mesa/drivers/dri/ffb/ffb_clear.c | 2 +- src/mesa/drivers/dri/ffb/ffb_context.h | 1 - src/mesa/drivers/dri/ffb/ffb_dd.c | 2 +- src/mesa/drivers/dri/ffb/ffb_dd.h | 2 +- src/mesa/drivers/dri/ffb/ffb_depth.c | 2 +- src/mesa/drivers/dri/ffb/ffb_depth.h | 1 - src/mesa/drivers/dri/ffb/ffb_fifo.h | 1 - src/mesa/drivers/dri/ffb/ffb_lines.c | 2 +- src/mesa/drivers/dri/ffb/ffb_lines.h | 1 - src/mesa/drivers/dri/ffb/ffb_linetmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_lock.h | 1 - src/mesa/drivers/dri/ffb/ffb_points.c | 2 +- src/mesa/drivers/dri/ffb/ffb_points.h | 1 - src/mesa/drivers/dri/ffb/ffb_pointtmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_rendertmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_span.c | 2 +- src/mesa/drivers/dri/ffb/ffb_span.h | 1 - src/mesa/drivers/dri/ffb/ffb_state.c | 2 +- src/mesa/drivers/dri/ffb/ffb_state.h | 1 - src/mesa/drivers/dri/ffb/ffb_stencil.c | 2 +- src/mesa/drivers/dri/ffb/ffb_stencil.h | 1 - src/mesa/drivers/dri/ffb/ffb_tex.c | 2 +- src/mesa/drivers/dri/ffb/ffb_tex.h | 2 +- src/mesa/drivers/dri/ffb/ffb_tris.c | 2 +- src/mesa/drivers/dri/ffb/ffb_tris.h | 1 - src/mesa/drivers/dri/ffb/ffb_tritmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_vb.c | 2 +- src/mesa/drivers/dri/ffb/ffb_vb.h | 1 - src/mesa/drivers/dri/ffb/ffb_vbtmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_vtxfmt.c | 2 +- src/mesa/drivers/dri/ffb/ffb_vtxfmt.h | 1 - src/mesa/drivers/dri/ffb/ffb_xmesa.c | 2 +- src/mesa/drivers/dri/ffb/ffb_xmesa.h | 1 - src/mesa/drivers/dri/ffb/server/ffb_dac.h | 1 - src/mesa/drivers/dri/ffb/server/ffb_drishare.h | 1 - src/mesa/drivers/dri/ffb/server/ffb_regs.h | 1 - src/mesa/drivers/dri/gamma/gamma_client.h | 1 - src/mesa/drivers/dri/gamma/gamma_context.h | 1 - src/mesa/drivers/dri/gamma/gamma_inithw.c | 1 - src/mesa/drivers/dri/gamma/gamma_lock.c | 1 - src/mesa/drivers/dri/gamma/gamma_macros.h | 1 - src/mesa/drivers/dri/gamma/gamma_regs.h | 1 - src/mesa/drivers/dri/gamma/gamma_span.c | 1 - src/mesa/drivers/dri/gamma/gamma_state.c | 1 - src/mesa/drivers/dri/gamma/gamma_tex.c | 1 - src/mesa/drivers/dri/gamma/gamma_texmem.c | 1 - src/mesa/drivers/dri/gamma/gamma_texstate.c | 1 - src/mesa/drivers/dri/gamma/gamma_tritmp.h | 1 - src/mesa/drivers/dri/gamma/gamma_vb.c | 1 - src/mesa/drivers/dri/gamma/gamma_xmesa.c | 1 - src/mesa/drivers/dri/gamma/server/glint_common.h | 1 - src/mesa/drivers/dri/gamma/server/glint_dri.h | 1 - src/mesa/drivers/dri/i810/i810_3d_reg.h | 1 - src/mesa/drivers/dri/i810/i810context.c | 1 - src/mesa/drivers/dri/i810/i810context.h | 1 - src/mesa/drivers/dri/i810/i810ioctl.c | 1 - src/mesa/drivers/dri/i810/i810ioctl.h | 1 - src/mesa/drivers/dri/i810/i810screen.c | 1 - src/mesa/drivers/dri/i810/i810state.c | 1 - src/mesa/drivers/dri/i810/i810tex.c | 1 - src/mesa/drivers/dri/i810/i810tris.c | 1 - src/mesa/drivers/dri/i810/i810tris.h | 1 - src/mesa/drivers/dri/i810/i810vb.c | 1 - src/mesa/drivers/dri/i810/i810vb.h | 1 - src/mesa/drivers/dri/i810/server/i810_common.h | 1 - src/mesa/drivers/dri/i810/server/i810_dri.h | 1 - src/mesa/drivers/dri/i810/server/i810_reg.h | 1 - src/mesa/drivers/dri/i915/server/i830_common.h | 1 - src/mesa/drivers/dri/i915/server/i830_dri.h | 1 - src/mesa/drivers/dri/i965/server/i830_common.h | 1 - src/mesa/drivers/dri/i965/server/i830_dri.h | 1 - src/mesa/drivers/dri/mach64/mach64_context.c | 2 +- src/mesa/drivers/dri/mach64/mach64_context.h | 2 +- src/mesa/drivers/dri/mach64/mach64_dd.c | 2 +- src/mesa/drivers/dri/mach64/mach64_dd.h | 2 +- src/mesa/drivers/dri/mach64/mach64_ioctl.c | 2 +- src/mesa/drivers/dri/mach64/mach64_ioctl.h | 2 +- src/mesa/drivers/dri/mach64/mach64_lock.c | 2 +- src/mesa/drivers/dri/mach64/mach64_lock.h | 2 +- src/mesa/drivers/dri/mach64/mach64_native_vb.c | 2 +- src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h | 2 +- src/mesa/drivers/dri/mach64/mach64_reg.h | 2 +- src/mesa/drivers/dri/mach64/mach64_screen.c | 2 +- src/mesa/drivers/dri/mach64/mach64_screen.h | 2 +- src/mesa/drivers/dri/mach64/mach64_span.c | 2 +- src/mesa/drivers/dri/mach64/mach64_span.h | 2 +- src/mesa/drivers/dri/mach64/mach64_state.c | 2 +- src/mesa/drivers/dri/mach64/mach64_state.h | 2 +- src/mesa/drivers/dri/mach64/mach64_tex.c | 2 +- src/mesa/drivers/dri/mach64/mach64_tex.h | 2 +- src/mesa/drivers/dri/mach64/mach64_texmem.c | 2 +- src/mesa/drivers/dri/mach64/mach64_texstate.c | 2 +- src/mesa/drivers/dri/mach64/mach64_tris.c | 2 +- src/mesa/drivers/dri/mach64/mach64_tris.h | 2 +- src/mesa/drivers/dri/mach64/mach64_vb.c | 2 +- src/mesa/drivers/dri/mach64/mach64_vb.h | 2 +- src/mesa/drivers/dri/mach64/mach64_vbtmp.h | 2 +- src/mesa/drivers/dri/mach64/server/mach64_dri.h | 2 +- src/mesa/drivers/dri/mga/mga_texstate.c | 1 - src/mesa/drivers/dri/mga/mga_xmesa.c | 1 - src/mesa/drivers/dri/mga/mga_xmesa.h | 1 - src/mesa/drivers/dri/mga/mgacontext.h | 1 - src/mesa/drivers/dri/mga/mgadd.c | 1 - src/mesa/drivers/dri/mga/mgadd.h | 1 - src/mesa/drivers/dri/mga/mgaioctl.h | 1 - src/mesa/drivers/dri/mga/mgapixel.c | 1 - src/mesa/drivers/dri/mga/mgapixel.h | 1 - src/mesa/drivers/dri/mga/mgaregs.h | 1 - src/mesa/drivers/dri/mga/mgarender.c | 1 - src/mesa/drivers/dri/mga/mgaspan.h | 1 - src/mesa/drivers/dri/mga/mgastate.h | 1 - src/mesa/drivers/dri/mga/mgatex.c | 1 - src/mesa/drivers/dri/mga/mgatex.h | 1 - src/mesa/drivers/dri/mga/mgatexmem.c | 1 - src/mesa/drivers/dri/mga/mgatris.c | 1 - src/mesa/drivers/dri/mga/mgatris.h | 1 - src/mesa/drivers/dri/mga/mgavb.c | 1 - src/mesa/drivers/dri/mga/mgavb.h | 1 - src/mesa/drivers/dri/mga/server/mga.h | 1 - src/mesa/drivers/dri/mga/server/mga_bios.h | 2 -- src/mesa/drivers/dri/mga/server/mga_dri.c | 1 - src/mesa/drivers/dri/mga/server/mga_dri.h | 1 - src/mesa/drivers/dri/mga/server/mga_macros.h | 1 - src/mesa/drivers/dri/mga/server/mga_reg.h | 2 -- src/mesa/drivers/dri/r128/r128_context.c | 1 - src/mesa/drivers/dri/r128/r128_context.h | 1 - src/mesa/drivers/dri/r128/r128_dd.c | 1 - src/mesa/drivers/dri/r128/r128_dd.h | 1 - src/mesa/drivers/dri/r128/r128_ioctl.c | 1 - src/mesa/drivers/dri/r128/r128_ioctl.h | 1 - src/mesa/drivers/dri/r128/r128_lock.c | 1 - src/mesa/drivers/dri/r128/r128_lock.h | 1 - src/mesa/drivers/dri/r128/r128_screen.c | 1 - src/mesa/drivers/dri/r128/r128_screen.h | 1 - src/mesa/drivers/dri/r128/r128_span.c | 1 - src/mesa/drivers/dri/r128/r128_span.h | 1 - src/mesa/drivers/dri/r128/r128_state.c | 1 - src/mesa/drivers/dri/r128/r128_state.h | 1 - src/mesa/drivers/dri/r128/r128_tex.c | 1 - src/mesa/drivers/dri/r128/r128_tex.h | 1 - src/mesa/drivers/dri/r128/r128_texmem.c | 1 - src/mesa/drivers/dri/r128/r128_texobj.h | 1 - src/mesa/drivers/dri/r128/r128_texstate.c | 1 - src/mesa/drivers/dri/r128/r128_tris.c | 2 +- src/mesa/drivers/dri/r128/r128_tris.h | 1 - src/mesa/drivers/dri/r128/server/r128.h | 1 - src/mesa/drivers/dri/r128/server/r128_dri.c | 1 - src/mesa/drivers/dri/r128/server/r128_dri.h | 1 - src/mesa/drivers/dri/r128/server/r128_macros.h | 1 - src/mesa/drivers/dri/r128/server/r128_reg.h | 1 - src/mesa/drivers/dri/r128/server/r128_version.h | 1 - src/mesa/drivers/dri/radeon/radeon_compat.c | 1 - src/mesa/drivers/dri/radeon/radeon_context.c | 1 - src/mesa/drivers/dri/radeon/radeon_ioctl.c | 1 - src/mesa/drivers/dri/radeon/radeon_ioctl.h | 1 - src/mesa/drivers/dri/radeon/radeon_lighting.c | 1 - src/mesa/drivers/dri/radeon/radeon_maos.h | 1 - src/mesa/drivers/dri/radeon/radeon_maos_arrays.c | 1 - src/mesa/drivers/dri/radeon/radeon_maos_verts.c | 1 - src/mesa/drivers/dri/radeon/radeon_sanity.c | 1 - src/mesa/drivers/dri/radeon/radeon_screen.c | 1 - src/mesa/drivers/dri/radeon/radeon_screen.h | 1 - src/mesa/drivers/dri/radeon/radeon_state.c | 1 - src/mesa/drivers/dri/radeon/radeon_state.h | 1 - src/mesa/drivers/dri/radeon/radeon_state_init.c | 1 - src/mesa/drivers/dri/radeon/radeon_swtcl.c | 1 - src/mesa/drivers/dri/radeon/radeon_swtcl.h | 1 - src/mesa/drivers/dri/radeon/radeon_tcl.c | 1 - src/mesa/drivers/dri/radeon/radeon_tcl.h | 1 - src/mesa/drivers/dri/radeon/radeon_tex.c | 1 - src/mesa/drivers/dri/radeon/radeon_tex.h | 1 - src/mesa/drivers/dri/radeon/radeon_texmem.c | 1 - src/mesa/drivers/dri/radeon/radeon_texstate.c | 1 - src/mesa/drivers/dri/radeon/server/radeon.h | 1 - src/mesa/drivers/dri/radeon/server/radeon_dri.h | 1 - src/mesa/drivers/dri/radeon/server/radeon_macros.h | 1 - src/mesa/drivers/dri/radeon/server/radeon_reg.h | 1 - src/mesa/drivers/dri/savage/savagetris.c | 2 +- src/mesa/drivers/dri/savage/savagetris.h | 1 - src/mesa/drivers/dri/sis/server/sis_common.h | 1 - src/mesa/drivers/dri/sis/server/sis_dri.h | 1 - src/mesa/drivers/dri/sis/sis_alloc.c | 1 - src/mesa/drivers/dri/sis/sis_alloc.h | 1 - src/mesa/drivers/dri/sis/sis_clear.c | 1 - src/mesa/drivers/dri/sis/sis_context.c | 1 - src/mesa/drivers/dri/sis/sis_context.h | 1 - src/mesa/drivers/dri/sis/sis_dd.c | 1 - src/mesa/drivers/dri/sis/sis_dd.h | 1 - src/mesa/drivers/dri/sis/sis_fog.c | 1 - src/mesa/drivers/dri/sis/sis_lock.c | 1 - src/mesa/drivers/dri/sis/sis_lock.h | 1 - src/mesa/drivers/dri/sis/sis_reg.h | 1 - src/mesa/drivers/dri/sis/sis_screen.c | 1 - src/mesa/drivers/dri/sis/sis_screen.h | 1 - src/mesa/drivers/dri/sis/sis_span.c | 1 - src/mesa/drivers/dri/sis/sis_span.h | 1 - src/mesa/drivers/dri/sis/sis_state.c | 1 - src/mesa/drivers/dri/sis/sis_state.h | 1 - src/mesa/drivers/dri/sis/sis_stencil.c | 1 - src/mesa/drivers/dri/sis/sis_stencil.h | 1 - src/mesa/drivers/dri/sis/sis_tex.c | 1 - src/mesa/drivers/dri/sis/sis_tex.h | 1 - src/mesa/drivers/dri/sis/sis_texstate.c | 1 - src/mesa/drivers/dri/sis/sis_tris.h | 1 - src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S | 1 - src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h | 1 - src/mesa/drivers/dri/tdfx/dri_glide.h | 1 - src/mesa/drivers/dri/tdfx/server/tdfx_dri.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_context.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_dd.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_glide.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_lock.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_lock.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_pixels.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_pixels.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_render.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_render.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_screen.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_screen.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_span.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_span.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_state.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_state.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_tex.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_tex.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_texman.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_texman.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_texstate.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_texstate.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_tris.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_tris.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_vb.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_vb.h | 1 - src/mesa/drivers/dri/unichrome/server/via_dri.c | 1 - src/mesa/drivers/dri/unichrome/server/via_driver.h | 1 - src/mesa/drivers/dri/unichrome/server/via_priv.h | 1 - src/mesa/drivers/ggi/default/genkgi.h | 2 +- src/mesa/drivers/ggi/default/genkgi_mode.c | 2 +- src/mesa/drivers/ggi/default/genkgi_visual.c | 2 +- src/mesa/drivers/ggi/include/ggi/mesa/debug.h | 2 +- src/mesa/drivers/svga/svgamesa.c | 1 - src/mesa/drivers/svga/svgamesa15.c | 1 - src/mesa/drivers/svga/svgamesa15.h | 1 - src/mesa/drivers/svga/svgamesa16.c | 1 - src/mesa/drivers/svga/svgamesa16.h | 1 - src/mesa/drivers/svga/svgamesa24.c | 1 - src/mesa/drivers/svga/svgamesa24.h | 1 - src/mesa/drivers/svga/svgamesa32.c | 1 - src/mesa/drivers/svga/svgamesa32.h | 1 - src/mesa/drivers/svga/svgamesa8.c | 1 - src/mesa/drivers/svga/svgamesa8.h | 1 - src/mesa/drivers/svga/svgapix.h | 1 - src/mesa/drivers/windows/gdi/wgl.c | 1 - src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c | 1 - src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c | 1 - src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c | 1 - src/mesa/drivers/windows/gldirect/gld_debug_clip.c | 1 - src/mesa/drivers/windows/gldirect/gld_debug_norm.c | 1 - src/mesa/drivers/windows/gldirect/gld_debug_xform.c | 1 - src/mesa/drivers/windows/gldirect/mesasw/colors.h | 7 ++----- src/mesa/glapi/mesadef.py | 1 - src/mesa/sparc/norm.S | 1 - src/mesa/sparc/sparc.h | 1 - src/mesa/sparc/xform.S | 1 - src/mesa/x86-64/x86-64.c | 1 - src/mesa/x86-64/x86-64.h | 1 - src/mesa/x86-64/xform4.S | 1 - src/mesa/x86/3dnow.c | 1 - src/mesa/x86/3dnow.h | 1 - src/mesa/x86/3dnow_normal.S | 1 - src/mesa/x86/3dnow_xform1.S | 1 - src/mesa/x86/3dnow_xform2.S | 1 - src/mesa/x86/3dnow_xform3.S | 1 - src/mesa/x86/3dnow_xform4.S | 1 - src/mesa/x86/clip_args.h | 1 - src/mesa/x86/common_x86_asm.h | 1 - src/mesa/x86/common_x86_features.h | 1 - src/mesa/x86/common_x86_macros.h | 1 - src/mesa/x86/norm_args.h | 1 - src/mesa/x86/sse.h | 1 - src/mesa/x86/sse_normal.S | 1 - src/mesa/x86/sse_xform1.S | 1 - src/mesa/x86/sse_xform2.S | 1 - src/mesa/x86/sse_xform3.S | 1 - src/mesa/x86/sse_xform4.S | 1 - src/mesa/x86/x86.c | 1 - src/mesa/x86/x86.h | 1 - src/mesa/x86/x86_cliptest.S | 1 - src/mesa/x86/x86_xform2.S | 1 - src/mesa/x86/x86_xform3.S | 1 - src/mesa/x86/x86_xform4.S | 1 - src/mesa/x86/xform_args.h | 1 - 509 files changed, 52 insertions(+), 677 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/server/i830_common.h b/src/gallium/winsys/dri/intel/server/i830_common.h index f1fd3939ab..f84f453309 100644 --- a/src/gallium/winsys/dri/intel/server/i830_common.h +++ b/src/gallium/winsys/dri/intel/server/i830_common.h @@ -26,7 +26,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ #ifndef _I830_COMMON_H_ #define _I830_COMMON_H_ diff --git a/src/gallium/winsys/dri/intel/server/i830_dri.h b/src/gallium/winsys/dri/intel/server/i830_dri.h index c2a3af8cbf..685de4a551 100644 --- a/src/gallium/winsys/dri/intel/server/i830_dri.h +++ b/src/gallium/winsys/dri/intel/server/i830_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ #ifndef _I830_DRI_H #define _I830_DRI_H diff --git a/src/glu/mini/all.h b/src/glu/mini/all.h index d626bee937..874c935925 100644 --- a/src/glu/mini/all.h +++ b/src/glu/mini/all.h @@ -1,4 +1,3 @@ -/* $Id: all.h,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/glu.c b/src/glu/mini/glu.c index 5c7722c5f0..31429e3343 100644 --- a/src/glu/mini/glu.c +++ b/src/glu/mini/glu.c @@ -1,4 +1,3 @@ -/* $Id: glu.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/gluP.h b/src/glu/mini/gluP.h index 85fbc33c62..a39edce41f 100644 --- a/src/glu/mini/gluP.h +++ b/src/glu/mini/gluP.h @@ -1,4 +1,3 @@ -/* $Id: gluP.h,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/mipmap.c b/src/glu/mini/mipmap.c index 97297729e7..a655d214e3 100644 --- a/src/glu/mini/mipmap.c +++ b/src/glu/mini/mipmap.c @@ -1,4 +1,3 @@ -/* $Id: mipmap.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/nurbs.c b/src/glu/mini/nurbs.c index 93c0dd3ce2..9f39cacb41 100644 --- a/src/glu/mini/nurbs.c +++ b/src/glu/mini/nurbs.c @@ -1,4 +1,3 @@ -/* $Id: nurbs.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/nurbs.h b/src/glu/mini/nurbs.h index c9c9c094f1..3642e213a8 100644 --- a/src/glu/mini/nurbs.h +++ b/src/glu/mini/nurbs.h @@ -1,4 +1,3 @@ -/* $Id: nurbs.h,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/nurbscrv.c b/src/glu/mini/nurbscrv.c index 62d91b46d3..e80468fdb0 100644 --- a/src/glu/mini/nurbscrv.c +++ b/src/glu/mini/nurbscrv.c @@ -1,4 +1,3 @@ -/* $Id: nurbscrv.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/polytest.c b/src/glu/mini/polytest.c index 52f272a3cb..1ff966f61c 100644 --- a/src/glu/mini/polytest.c +++ b/src/glu/mini/polytest.c @@ -1,4 +1,3 @@ -/* $Id: polytest.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/project.c b/src/glu/mini/project.c index a2747de55f..6fa03267e5 100644 --- a/src/glu/mini/project.c +++ b/src/glu/mini/project.c @@ -1,4 +1,3 @@ -/* $Id: project.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/quadric.c b/src/glu/mini/quadric.c index 015552e123..0484890ef6 100644 --- a/src/glu/mini/quadric.c +++ b/src/glu/mini/quadric.c @@ -1,4 +1,3 @@ -/* $Id: quadric.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/tess.c b/src/glu/mini/tess.c index 1a384239be..341d29bae3 100644 --- a/src/glu/mini/tess.c +++ b/src/glu/mini/tess.c @@ -1,4 +1,3 @@ -/* $Id: tess.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/tess.h b/src/glu/mini/tess.h index 908e20972c..4e51dddd37 100644 --- a/src/glu/mini/tess.h +++ b/src/glu/mini/tess.h @@ -1,4 +1,3 @@ -/* $Id: tess.h,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/tesselat.c b/src/glu/mini/tesselat.c index a1102e6e5a..47d230073f 100644 --- a/src/glu/mini/tesselat.c +++ b/src/glu/mini/tesselat.c @@ -1,4 +1,3 @@ -/* $Id: tesselat.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/sgi/dummy.cc b/src/glu/sgi/dummy.cc index fac5a63b76..bd905a2608 100644 --- a/src/glu/sgi/dummy.cc +++ b/src/glu/sgi/dummy.cc @@ -1,4 +1,3 @@ -/* $Id: dummy.cc,v 1.1 2001/03/18 13:06:19 pesco Exp $ */ /* * This file contains nothing. It's just there so there's at least a single * source file for libGLU.la in this directory. diff --git a/src/glu/sgi/libnurbs/interface/bezierEval.h b/src/glu/sgi/libnurbs/interface/bezierEval.h index 1a9f3c78e7..adecfe9b2f 100644 --- a/src/glu/sgi/libnurbs/interface/bezierEval.h +++ b/src/glu/sgi/libnurbs/interface/bezierEval.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierEval.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef _BEZIEREVAL_H diff --git a/src/glu/sgi/libnurbs/interface/bezierPatch.cc b/src/glu/sgi/libnurbs/interface/bezierPatch.cc index 836ae94e0a..fa1daed52e 100644 --- a/src/glu/sgi/libnurbs/interface/bezierPatch.cc +++ b/src/glu/sgi/libnurbs/interface/bezierPatch.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierPatch.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/interface/bezierPatch.h b/src/glu/sgi/libnurbs/interface/bezierPatch.h index 31c97ba08f..ad0f8b0d2a 100644 --- a/src/glu/sgi/libnurbs/interface/bezierPatch.h +++ b/src/glu/sgi/libnurbs/interface/bezierPatch.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierPatch.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef _BEZIERPATCH_H diff --git a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc index 9ff416ad6e..3dc16313ff 100644 --- a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc +++ b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h index 74cf098858..2ab24dff5b 100644 --- a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h +++ b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef _BEZIERPATCHMESH_H diff --git a/src/glu/sgi/libnurbs/interface/glcurveval.cc b/src/glu/sgi/libnurbs/interface/glcurveval.cc index 32e4704137..b6591dba0d 100644 --- a/src/glu/sgi/libnurbs/interface/glcurveval.cc +++ b/src/glu/sgi/libnurbs/interface/glcurveval.cc @@ -35,8 +35,6 @@ /* * glcurveval.c++ * - * $Date: 2006/03/29 18:46:46 $ $Revision: 1.7 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/glcurveval.cc,v 1.7 2006/03/29 18:46:46 brianp Exp $ */ /* Polynomial Evaluator Interface */ diff --git a/src/glu/sgi/libnurbs/interface/glimports.h b/src/glu/sgi/libnurbs/interface/glimports.h index 9a9d3e32c9..2c307f63e8 100644 --- a/src/glu/sgi/libnurbs/interface/glimports.h +++ b/src/glu/sgi/libnurbs/interface/glimports.h @@ -35,8 +35,6 @@ /* * glimports.h * - * $Date: 2001/03/19 17:52:02 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/glimports.h,v 1.3 2001/03/19 17:52:02 pesco Exp $ */ #ifndef __gluimports_h_ diff --git a/src/glu/sgi/libnurbs/interface/glinterface.cc b/src/glu/sgi/libnurbs/interface/glinterface.cc index dfd16d1722..ba64bcd2dc 100644 --- a/src/glu/sgi/libnurbs/interface/glinterface.cc +++ b/src/glu/sgi/libnurbs/interface/glinterface.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/07/16 15:46:42 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/glinterface.cc,v 1.2 2001/07/16 15:46:42 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/interface/glrenderer.h b/src/glu/sgi/libnurbs/interface/glrenderer.h index 30f07632a4..8fc23125e0 100644 --- a/src/glu/sgi/libnurbs/interface/glrenderer.h +++ b/src/glu/sgi/libnurbs/interface/glrenderer.h @@ -35,8 +35,6 @@ /* * glrenderer.h * - * $Date: 2004/02/26 14:58:11 $ $Revision: 1.4 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/glrenderer.h,v 1.4 2004/02/26 14:58:11 brianp Exp $ */ #ifndef __gluglrenderer_h_ diff --git a/src/glu/sgi/libnurbs/interface/incurveeval.cc b/src/glu/sgi/libnurbs/interface/incurveeval.cc index 336cca0508..96ea8896ae 100644 --- a/src/glu/sgi/libnurbs/interface/incurveeval.cc +++ b/src/glu/sgi/libnurbs/interface/incurveeval.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2004/05/12 15:29:36 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/incurveeval.cc,v 1.2 2004/05/12 15:29:36 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/interface/insurfeval.cc b/src/glu/sgi/libnurbs/interface/insurfeval.cc index b314699c7a..78d8bece13 100644 --- a/src/glu/sgi/libnurbs/interface/insurfeval.cc +++ b/src/glu/sgi/libnurbs/interface/insurfeval.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2004/05/12 15:29:36 $ $Revision: 1.3 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/insurfeval.cc,v 1.3 2004/05/12 15:29:36 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/interface/mystdio.h b/src/glu/sgi/libnurbs/interface/mystdio.h index 6d737257f7..e9947ea393 100644 --- a/src/glu/sgi/libnurbs/interface/mystdio.h +++ b/src/glu/sgi/libnurbs/interface/mystdio.h @@ -35,8 +35,6 @@ /* * mystdio.h * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.4 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/mystdio.h,v 1.4 2006/03/14 15:08:52 brianp Exp $ */ #ifndef __glumystdio_h_ diff --git a/src/glu/sgi/libnurbs/interface/mystdlib.h b/src/glu/sgi/libnurbs/interface/mystdlib.h index 0ebbc1299f..2520b41e0a 100644 --- a/src/glu/sgi/libnurbs/interface/mystdlib.h +++ b/src/glu/sgi/libnurbs/interface/mystdlib.h @@ -35,8 +35,6 @@ /* * mystdlib.h * - * $Date: 2001/03/19 17:52:02 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/mystdlib.h,v 1.3 2001/03/19 17:52:02 pesco Exp $ */ #ifndef __glumystdlib_h_ diff --git a/src/glu/sgi/libnurbs/internals/arc.h b/src/glu/sgi/libnurbs/internals/arc.h index b700a1e826..bbed33c649 100644 --- a/src/glu/sgi/libnurbs/internals/arc.h +++ b/src/glu/sgi/libnurbs/internals/arc.h @@ -35,8 +35,6 @@ /* * arc.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/arc.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __gluarc_h_ diff --git a/src/glu/sgi/libnurbs/internals/arcsorter.cc b/src/glu/sgi/libnurbs/internals/arcsorter.cc index 1a7f4c6911..1f85cb7108 100644 --- a/src/glu/sgi/libnurbs/internals/arcsorter.cc +++ b/src/glu/sgi/libnurbs/internals/arcsorter.cc @@ -35,8 +35,6 @@ /* * arcsorter.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/arcsorter.cc,v 1.2 2006/03/14 15:08:52 brianp Exp $ */ #ifndef __gluarcsorter_c_ diff --git a/src/glu/sgi/libnurbs/internals/arcsorter.h b/src/glu/sgi/libnurbs/internals/arcsorter.h index 989f80a43c..1025d30b5d 100644 --- a/src/glu/sgi/libnurbs/internals/arcsorter.h +++ b/src/glu/sgi/libnurbs/internals/arcsorter.h @@ -35,8 +35,6 @@ /* * arcsorter.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/arcsorter.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gluarcsorter_h_ diff --git a/src/glu/sgi/libnurbs/internals/arctess.h b/src/glu/sgi/libnurbs/internals/arctess.h index fc42ea5eb7..d3ea2071ea 100644 --- a/src/glu/sgi/libnurbs/internals/arctess.h +++ b/src/glu/sgi/libnurbs/internals/arctess.h @@ -35,8 +35,6 @@ /* * arctess.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/arctess.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __gluarctess_h_ diff --git a/src/glu/sgi/libnurbs/internals/backend.cc b/src/glu/sgi/libnurbs/internals/backend.cc index 97775a9768..69c46b2d52 100644 --- a/src/glu/sgi/libnurbs/internals/backend.cc +++ b/src/glu/sgi/libnurbs/internals/backend.cc @@ -35,8 +35,6 @@ /* * backend.c++ * - * $Date: 2004/05/12 15:29:36 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/backend.cc,v 1.2 2004/05/12 15:29:36 brianp Exp $ */ /* Bezier surface backend diff --git a/src/glu/sgi/libnurbs/internals/backend.h b/src/glu/sgi/libnurbs/internals/backend.h index c1f00b1a01..fb03859f27 100644 --- a/src/glu/sgi/libnurbs/internals/backend.h +++ b/src/glu/sgi/libnurbs/internals/backend.h @@ -35,8 +35,6 @@ /* * backend.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/backend.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glubackend_h_ diff --git a/src/glu/sgi/libnurbs/internals/basiccrveval.h b/src/glu/sgi/libnurbs/internals/basiccrveval.h index 0a5f66c201..41abedbb20 100644 --- a/src/glu/sgi/libnurbs/internals/basiccrveval.h +++ b/src/glu/sgi/libnurbs/internals/basiccrveval.h @@ -35,8 +35,6 @@ /* * basiccurveeval.h * - * $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/basiccrveval.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __glubasiccrveval_h_ diff --git a/src/glu/sgi/libnurbs/internals/basicsurfeval.h b/src/glu/sgi/libnurbs/internals/basicsurfeval.h index a67ded97b5..2fe76ad67d 100644 --- a/src/glu/sgi/libnurbs/internals/basicsurfeval.h +++ b/src/glu/sgi/libnurbs/internals/basicsurfeval.h @@ -35,8 +35,6 @@ /* * basicsurfeval.h * - * $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/basicsurfeval.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __glubasicsurfeval_h_ diff --git a/src/glu/sgi/libnurbs/internals/bezierarc.h b/src/glu/sgi/libnurbs/internals/bezierarc.h index 64dd31d87d..a6d5a13ee6 100644 --- a/src/glu/sgi/libnurbs/internals/bezierarc.h +++ b/src/glu/sgi/libnurbs/internals/bezierarc.h @@ -35,8 +35,6 @@ /* * bezierarc.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bezierarc.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glubezierarc_h diff --git a/src/glu/sgi/libnurbs/internals/bin.cc b/src/glu/sgi/libnurbs/internals/bin.cc index ed427567f9..54b406147b 100644 --- a/src/glu/sgi/libnurbs/internals/bin.cc +++ b/src/glu/sgi/libnurbs/internals/bin.cc @@ -35,8 +35,6 @@ /* * bin.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bin.cc,v 1.3 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/bin.h b/src/glu/sgi/libnurbs/internals/bin.h index 17d146fdf1..ecdf9b83b8 100644 --- a/src/glu/sgi/libnurbs/internals/bin.h +++ b/src/glu/sgi/libnurbs/internals/bin.h @@ -35,8 +35,6 @@ /* * bin.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bin.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glubin_h_ diff --git a/src/glu/sgi/libnurbs/internals/bufpool.cc b/src/glu/sgi/libnurbs/internals/bufpool.cc index d8d9c23db3..f60f7dc7b1 100644 --- a/src/glu/sgi/libnurbs/internals/bufpool.cc +++ b/src/glu/sgi/libnurbs/internals/bufpool.cc @@ -35,8 +35,6 @@ /* * bufpool.c++ * - * $Date: 2004/05/12 15:29:36 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bufpool.cc,v 1.2 2004/05/12 15:29:36 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/bufpool.h b/src/glu/sgi/libnurbs/internals/bufpool.h index 02e4ff247b..8eaafc4fd0 100644 --- a/src/glu/sgi/libnurbs/internals/bufpool.h +++ b/src/glu/sgi/libnurbs/internals/bufpool.h @@ -35,8 +35,6 @@ /* * bufpool.h * - * $Date: 2006/03/29 18:46:46 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bufpool.h,v 1.3 2006/03/29 18:46:46 brianp Exp $ */ #ifndef __glubufpool_h_ diff --git a/src/glu/sgi/libnurbs/internals/cachingeval.cc b/src/glu/sgi/libnurbs/internals/cachingeval.cc index 7245ee3a18..3fab38c106 100644 --- a/src/glu/sgi/libnurbs/internals/cachingeval.cc +++ b/src/glu/sgi/libnurbs/internals/cachingeval.cc @@ -35,8 +35,6 @@ /* * cachingeval.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/cachingeval.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "cachingeval.h" diff --git a/src/glu/sgi/libnurbs/internals/cachingeval.h b/src/glu/sgi/libnurbs/internals/cachingeval.h index 578391707a..cb4c83501a 100644 --- a/src/glu/sgi/libnurbs/internals/cachingeval.h +++ b/src/glu/sgi/libnurbs/internals/cachingeval.h @@ -35,8 +35,6 @@ /* * cachingeval.h * - * $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/cachingeval.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __glucachingval_h_ diff --git a/src/glu/sgi/libnurbs/internals/ccw.cc b/src/glu/sgi/libnurbs/internals/ccw.cc index b1bb6276f7..eb01b7781a 100644 --- a/src/glu/sgi/libnurbs/internals/ccw.cc +++ b/src/glu/sgi/libnurbs/internals/ccw.cc @@ -35,8 +35,6 @@ /* * ccw.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/ccw.cc,v 1.3 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/coveandtiler.h b/src/glu/sgi/libnurbs/internals/coveandtiler.h index 4f4077e208..bb682b75c7 100644 --- a/src/glu/sgi/libnurbs/internals/coveandtiler.h +++ b/src/glu/sgi/libnurbs/internals/coveandtiler.h @@ -35,8 +35,6 @@ /* * coveandtiler.h * - * $Date: 2001/07/16 15:46:42 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/coveandtiler.h,v 1.2 2001/07/16 15:46:42 brianp Exp $ */ #ifndef __glucoveandtiler_h diff --git a/src/glu/sgi/libnurbs/internals/curve.cc b/src/glu/sgi/libnurbs/internals/curve.cc index 5517afa2db..33e2752643 100644 --- a/src/glu/sgi/libnurbs/internals/curve.cc +++ b/src/glu/sgi/libnurbs/internals/curve.cc @@ -35,8 +35,6 @@ /* * curve.c++ * - * $Date: 2004/05/12 15:29:36 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curve.cc,v 1.3 2004/05/12 15:29:36 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/curve.h b/src/glu/sgi/libnurbs/internals/curve.h index 7b7bd3dc89..6f7b1de9c0 100644 --- a/src/glu/sgi/libnurbs/internals/curve.h +++ b/src/glu/sgi/libnurbs/internals/curve.h @@ -35,8 +35,6 @@ /* * curve.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curve.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glucurve_h_ diff --git a/src/glu/sgi/libnurbs/internals/curvelist.cc b/src/glu/sgi/libnurbs/internals/curvelist.cc index e763c62945..872eb5816d 100644 --- a/src/glu/sgi/libnurbs/internals/curvelist.cc +++ b/src/glu/sgi/libnurbs/internals/curvelist.cc @@ -35,8 +35,6 @@ /* * curvelist.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curvelist.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/curvelist.h b/src/glu/sgi/libnurbs/internals/curvelist.h index d285fb5b98..afbaa353ec 100644 --- a/src/glu/sgi/libnurbs/internals/curvelist.h +++ b/src/glu/sgi/libnurbs/internals/curvelist.h @@ -35,8 +35,6 @@ /* * curvelist.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curvelist.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glucurvelist_h_ diff --git a/src/glu/sgi/libnurbs/internals/curvesub.cc b/src/glu/sgi/libnurbs/internals/curvesub.cc index 11b15e4174..f85acc269a 100644 --- a/src/glu/sgi/libnurbs/internals/curvesub.cc +++ b/src/glu/sgi/libnurbs/internals/curvesub.cc @@ -35,8 +35,6 @@ /* * curvesub.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curvesub.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/dataTransform.cc b/src/glu/sgi/libnurbs/internals/dataTransform.cc index 822da02228..55c0fbb159 100644 --- a/src/glu/sgi/libnurbs/internals/dataTransform.cc +++ b/src/glu/sgi/libnurbs/internals/dataTransform.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/dataTransform.cc,v 1.2 2005/10/28 13:09:23 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/dataTransform.h b/src/glu/sgi/libnurbs/internals/dataTransform.h index 1032896f13..08730e174e 100644 --- a/src/glu/sgi/libnurbs/internals/dataTransform.h +++ b/src/glu/sgi/libnurbs/internals/dataTransform.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/dataTransform.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef _DATA_TRANSFORM_H diff --git a/src/glu/sgi/libnurbs/internals/defines.h b/src/glu/sgi/libnurbs/internals/defines.h index 77b6088acc..aae1682e39 100644 --- a/src/glu/sgi/libnurbs/internals/defines.h +++ b/src/glu/sgi/libnurbs/internals/defines.h @@ -35,8 +35,6 @@ /* * defines.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/defines.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gludefines_h_ diff --git a/src/glu/sgi/libnurbs/internals/displaylist.cc b/src/glu/sgi/libnurbs/internals/displaylist.cc index 4b39a8991d..48593c6371 100644 --- a/src/glu/sgi/libnurbs/internals/displaylist.cc +++ b/src/glu/sgi/libnurbs/internals/displaylist.cc @@ -35,8 +35,6 @@ /* * displaylist.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/displaylist.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/displaylist.h b/src/glu/sgi/libnurbs/internals/displaylist.h index 13cadaeae8..4bd6d76384 100644 --- a/src/glu/sgi/libnurbs/internals/displaylist.h +++ b/src/glu/sgi/libnurbs/internals/displaylist.h @@ -35,8 +35,6 @@ /* * displaylist.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/displaylist.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gludisplaylist_h_ diff --git a/src/glu/sgi/libnurbs/internals/displaymode.h b/src/glu/sgi/libnurbs/internals/displaymode.h index 791434e6d2..9289b99b89 100644 --- a/src/glu/sgi/libnurbs/internals/displaymode.h +++ b/src/glu/sgi/libnurbs/internals/displaymode.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/displaymode.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gludisplaymode_h_ diff --git a/src/glu/sgi/libnurbs/internals/flist.cc b/src/glu/sgi/libnurbs/internals/flist.cc index 21414fd736..d3162b9f5f 100644 --- a/src/glu/sgi/libnurbs/internals/flist.cc +++ b/src/glu/sgi/libnurbs/internals/flist.cc @@ -35,8 +35,6 @@ /* * flist.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/flist.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/flist.h b/src/glu/sgi/libnurbs/internals/flist.h index 8450caad45..a643db52b8 100644 --- a/src/glu/sgi/libnurbs/internals/flist.h +++ b/src/glu/sgi/libnurbs/internals/flist.h @@ -35,8 +35,6 @@ /* * flist.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/flist.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gluflist_h_ diff --git a/src/glu/sgi/libnurbs/internals/flistsorter.cc b/src/glu/sgi/libnurbs/internals/flistsorter.cc index 730613224c..d49bdea3e0 100644 --- a/src/glu/sgi/libnurbs/internals/flistsorter.cc +++ b/src/glu/sgi/libnurbs/internals/flistsorter.cc @@ -35,8 +35,6 @@ /* * flistsorter.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/flistsorter.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/flistsorter.h b/src/glu/sgi/libnurbs/internals/flistsorter.h index 753ed255b5..d9fe81a85f 100644 --- a/src/glu/sgi/libnurbs/internals/flistsorter.h +++ b/src/glu/sgi/libnurbs/internals/flistsorter.h @@ -35,8 +35,6 @@ /* * flistsorter.h * - * $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/flistsorter.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __gluflistsorter_h_ diff --git a/src/glu/sgi/libnurbs/internals/gridline.h b/src/glu/sgi/libnurbs/internals/gridline.h index 32b70bbb29..eaa8797217 100644 --- a/src/glu/sgi/libnurbs/internals/gridline.h +++ b/src/glu/sgi/libnurbs/internals/gridline.h @@ -35,8 +35,6 @@ /* * gridline.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/gridline.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glugridline_h_ diff --git a/src/glu/sgi/libnurbs/internals/gridtrimvertex.h b/src/glu/sgi/libnurbs/internals/gridtrimvertex.h index 70a7029c54..72f737a9dc 100644 --- a/src/glu/sgi/libnurbs/internals/gridtrimvertex.h +++ b/src/glu/sgi/libnurbs/internals/gridtrimvertex.h @@ -35,8 +35,6 @@ /* * gridtrimvertex.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/gridtrimvertex.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glugridtrimvertex_h_ diff --git a/src/glu/sgi/libnurbs/internals/gridvertex.h b/src/glu/sgi/libnurbs/internals/gridvertex.h index 2eac57386a..23035a00c5 100644 --- a/src/glu/sgi/libnurbs/internals/gridvertex.h +++ b/src/glu/sgi/libnurbs/internals/gridvertex.h @@ -35,8 +35,6 @@ /* * gridvertex.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/gridvertex.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glugridvertex_h_ diff --git a/src/glu/sgi/libnurbs/internals/hull.cc b/src/glu/sgi/libnurbs/internals/hull.cc index 75f7c160d6..389ba66fb8 100644 --- a/src/glu/sgi/libnurbs/internals/hull.cc +++ b/src/glu/sgi/libnurbs/internals/hull.cc @@ -35,8 +35,6 @@ /* * hull.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/hull.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/hull.h b/src/glu/sgi/libnurbs/internals/hull.h index 34f1593a3e..30ffd6bac3 100644 --- a/src/glu/sgi/libnurbs/internals/hull.h +++ b/src/glu/sgi/libnurbs/internals/hull.h @@ -35,8 +35,6 @@ /* * hull.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/hull.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __gluhull_h_ diff --git a/src/glu/sgi/libnurbs/internals/intersect.cc b/src/glu/sgi/libnurbs/internals/intersect.cc index 6fb7e3239b..b39ea2121e 100644 --- a/src/glu/sgi/libnurbs/internals/intersect.cc +++ b/src/glu/sgi/libnurbs/internals/intersect.cc @@ -35,8 +35,6 @@ /* * intersect.c++ * - * $Date: 2005/10/28 13:09:23 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/intersect.cc,v 1.3 2005/10/28 13:09:23 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/jarcloc.h b/src/glu/sgi/libnurbs/internals/jarcloc.h index 785234f6c0..3582a607a7 100644 --- a/src/glu/sgi/libnurbs/internals/jarcloc.h +++ b/src/glu/sgi/libnurbs/internals/jarcloc.h @@ -35,8 +35,6 @@ /* * jarcloc.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/jarcloc.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glujarcloc_h_ diff --git a/src/glu/sgi/libnurbs/internals/knotvector.h b/src/glu/sgi/libnurbs/internals/knotvector.h index bb1e593326..508fc4f345 100644 --- a/src/glu/sgi/libnurbs/internals/knotvector.h +++ b/src/glu/sgi/libnurbs/internals/knotvector.h @@ -35,8 +35,6 @@ /* * knotvector.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/knotvector.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __gluknotvector_h_ diff --git a/src/glu/sgi/libnurbs/internals/mapdesc.cc b/src/glu/sgi/libnurbs/internals/mapdesc.cc index 14d01582b0..d59f8fd395 100644 --- a/src/glu/sgi/libnurbs/internals/mapdesc.cc +++ b/src/glu/sgi/libnurbs/internals/mapdesc.cc @@ -35,8 +35,6 @@ /* * mapdesc.c++ * - * $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mapdesc.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/mapdesc.h b/src/glu/sgi/libnurbs/internals/mapdesc.h index 3c4ef6ff6c..fe5d650a2a 100644 --- a/src/glu/sgi/libnurbs/internals/mapdesc.h +++ b/src/glu/sgi/libnurbs/internals/mapdesc.h @@ -35,8 +35,6 @@ /* * mapdesc.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mapdesc.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glumapdesc_h_ diff --git a/src/glu/sgi/libnurbs/internals/mapdescv.cc b/src/glu/sgi/libnurbs/internals/mapdescv.cc index 6e4bb40c90..35b38b141b 100644 --- a/src/glu/sgi/libnurbs/internals/mapdescv.cc +++ b/src/glu/sgi/libnurbs/internals/mapdescv.cc @@ -35,8 +35,6 @@ /* * mapdescv.c++ * - * $Date: 2004/05/12 15:29:36 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mapdescv.cc,v 1.3 2004/05/12 15:29:36 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/maplist.cc b/src/glu/sgi/libnurbs/internals/maplist.cc index 44f8666b7a..f944d1529e 100644 --- a/src/glu/sgi/libnurbs/internals/maplist.cc +++ b/src/glu/sgi/libnurbs/internals/maplist.cc @@ -35,8 +35,6 @@ /* * maplist.c++ * - * $Date: 2005/10/28 13:09:23 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/maplist.cc,v 1.2 2005/10/28 13:09:23 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/maplist.h b/src/glu/sgi/libnurbs/internals/maplist.h index ca92def8ca..e86253966d 100644 --- a/src/glu/sgi/libnurbs/internals/maplist.h +++ b/src/glu/sgi/libnurbs/internals/maplist.h @@ -35,8 +35,6 @@ /* * maplist.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/maplist.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glumaplist_h_ diff --git a/src/glu/sgi/libnurbs/internals/mesher.cc b/src/glu/sgi/libnurbs/internals/mesher.cc index 1178eeb516..9cc436adbf 100644 --- a/src/glu/sgi/libnurbs/internals/mesher.cc +++ b/src/glu/sgi/libnurbs/internals/mesher.cc @@ -35,8 +35,6 @@ /* * mesher.c++ * - * $Date: 2001/11/29 16:16:55 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mesher.cc,v 1.3 2001/11/29 16:16:55 kschultz Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/mesher.h b/src/glu/sgi/libnurbs/internals/mesher.h index e4cb4466bc..b9f74f3819 100644 --- a/src/glu/sgi/libnurbs/internals/mesher.h +++ b/src/glu/sgi/libnurbs/internals/mesher.h @@ -35,8 +35,6 @@ /* * mesher.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mesher.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __glumesher_h_ diff --git a/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc b/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc index b08cd91570..2830cc743c 100644 --- a/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc +++ b/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.3 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc,v 1.3 2005/10/28 13:09:23 brianp Exp $ */ #include "monoTriangulation.h" diff --git a/src/glu/sgi/libnurbs/internals/monotonizer.cc b/src/glu/sgi/libnurbs/internals/monotonizer.cc index 7b6685dd06..5845d310ba 100644 --- a/src/glu/sgi/libnurbs/internals/monotonizer.cc +++ b/src/glu/sgi/libnurbs/internals/monotonizer.cc @@ -35,8 +35,6 @@ /* * monotonizer.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/monotonizer.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/monotonizer.h b/src/glu/sgi/libnurbs/internals/monotonizer.h index f516207cb5..7282a8c491 100644 --- a/src/glu/sgi/libnurbs/internals/monotonizer.h +++ b/src/glu/sgi/libnurbs/internals/monotonizer.h @@ -35,7 +35,6 @@ /* * monotonizer.h * - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/monotonizer.h,v 1.2 2006/04/03 22:23:52 ajax Exp $ */ #ifndef __glumonotonizer_h_ diff --git a/src/glu/sgi/libnurbs/internals/myassert.h b/src/glu/sgi/libnurbs/internals/myassert.h index e222c7e43e..9b5ee0f353 100644 --- a/src/glu/sgi/libnurbs/internals/myassert.h +++ b/src/glu/sgi/libnurbs/internals/myassert.h @@ -35,8 +35,6 @@ /* * myassert.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/myassert.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glumyassert_h_ diff --git a/src/glu/sgi/libnurbs/internals/mycode.cc b/src/glu/sgi/libnurbs/internals/mycode.cc index c49f64fa21..3625cacd74 100644 --- a/src/glu/sgi/libnurbs/internals/mycode.cc +++ b/src/glu/sgi/libnurbs/internals/mycode.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mycode.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "mymath.h" diff --git a/src/glu/sgi/libnurbs/internals/mystring.h b/src/glu/sgi/libnurbs/internals/mystring.h index 8b9032bf23..fedf32f114 100644 --- a/src/glu/sgi/libnurbs/internals/mystring.h +++ b/src/glu/sgi/libnurbs/internals/mystring.h @@ -35,8 +35,6 @@ /* * mystring.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mystring.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glumystring_h_ diff --git a/src/glu/sgi/libnurbs/internals/nurbsconsts.h b/src/glu/sgi/libnurbs/internals/nurbsconsts.h index 7b9dcc39cd..30277d6892 100644 --- a/src/glu/sgi/libnurbs/internals/nurbsconsts.h +++ b/src/glu/sgi/libnurbs/internals/nurbsconsts.h @@ -35,8 +35,6 @@ /* * nurbsconsts.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/nurbsconsts.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glunurbsconsts_h_ diff --git a/src/glu/sgi/libnurbs/internals/nurbstess.cc b/src/glu/sgi/libnurbs/internals/nurbstess.cc index adf7c74626..a5bd060fb0 100644 --- a/src/glu/sgi/libnurbs/internals/nurbstess.cc +++ b/src/glu/sgi/libnurbs/internals/nurbstess.cc @@ -35,8 +35,6 @@ /* * nurbstess.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/nurbstess.cc,v 1.2 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/patch.cc b/src/glu/sgi/libnurbs/internals/patch.cc index 4a524f1de2..808baa69e4 100644 --- a/src/glu/sgi/libnurbs/internals/patch.cc +++ b/src/glu/sgi/libnurbs/internals/patch.cc @@ -35,8 +35,6 @@ /* * patch.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.4 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/patch.cc,v 1.4 2006/03/14 15:08:52 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/patch.h b/src/glu/sgi/libnurbs/internals/patch.h index a214b571f9..d42613b67e 100644 --- a/src/glu/sgi/libnurbs/internals/patch.h +++ b/src/glu/sgi/libnurbs/internals/patch.h @@ -35,8 +35,6 @@ /* * patch.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/patch.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glupatch_h_ diff --git a/src/glu/sgi/libnurbs/internals/patchlist.cc b/src/glu/sgi/libnurbs/internals/patchlist.cc index a640893f1b..989d2dd00a 100644 --- a/src/glu/sgi/libnurbs/internals/patchlist.cc +++ b/src/glu/sgi/libnurbs/internals/patchlist.cc @@ -35,8 +35,6 @@ /* * patchlist.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/patchlist.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/patchlist.h b/src/glu/sgi/libnurbs/internals/patchlist.h index 9fb3795c09..04701c292b 100644 --- a/src/glu/sgi/libnurbs/internals/patchlist.h +++ b/src/glu/sgi/libnurbs/internals/patchlist.h @@ -35,8 +35,6 @@ /* * patchlist.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/patchlist.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glupatchlist_h_ diff --git a/src/glu/sgi/libnurbs/internals/pwlarc.h b/src/glu/sgi/libnurbs/internals/pwlarc.h index 83b7c3f813..b0422b4ded 100644 --- a/src/glu/sgi/libnurbs/internals/pwlarc.h +++ b/src/glu/sgi/libnurbs/internals/pwlarc.h @@ -35,8 +35,6 @@ /* * pwlarc.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/pwlarc.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glupwlarc_h_ diff --git a/src/glu/sgi/libnurbs/internals/quilt.cc b/src/glu/sgi/libnurbs/internals/quilt.cc index f693b370ba..4fc58b7473 100644 --- a/src/glu/sgi/libnurbs/internals/quilt.cc +++ b/src/glu/sgi/libnurbs/internals/quilt.cc @@ -35,8 +35,6 @@ /* * quilt.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/quilt.cc,v 1.3 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/quilt.h b/src/glu/sgi/libnurbs/internals/quilt.h index 336c2574d2..a23c3c11b1 100644 --- a/src/glu/sgi/libnurbs/internals/quilt.h +++ b/src/glu/sgi/libnurbs/internals/quilt.h @@ -35,8 +35,6 @@ /* * quilt.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/quilt.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __gluquilt_h_ diff --git a/src/glu/sgi/libnurbs/internals/reader.cc b/src/glu/sgi/libnurbs/internals/reader.cc index 271a32fbc1..6135eef60e 100644 --- a/src/glu/sgi/libnurbs/internals/reader.cc +++ b/src/glu/sgi/libnurbs/internals/reader.cc @@ -35,8 +35,6 @@ /* * reader.c++ * - * $Date: 2002/11/01 23:35:07 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/reader.cc,v 1.2 2002/11/01 23:35:07 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/reader.h b/src/glu/sgi/libnurbs/internals/reader.h index ac86f8a29f..c826d3812f 100644 --- a/src/glu/sgi/libnurbs/internals/reader.h +++ b/src/glu/sgi/libnurbs/internals/reader.h @@ -35,8 +35,6 @@ /* * reader.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/reader.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __glureader_h_ diff --git a/src/glu/sgi/libnurbs/internals/renderhints.cc b/src/glu/sgi/libnurbs/internals/renderhints.cc index 6a9d37e013..a3aa62d42c 100644 --- a/src/glu/sgi/libnurbs/internals/renderhints.cc +++ b/src/glu/sgi/libnurbs/internals/renderhints.cc @@ -35,8 +35,6 @@ /* * renderhints.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/renderhints.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/renderhints.h b/src/glu/sgi/libnurbs/internals/renderhints.h index efa959e3ab..e248422925 100644 --- a/src/glu/sgi/libnurbs/internals/renderhints.h +++ b/src/glu/sgi/libnurbs/internals/renderhints.h @@ -35,8 +35,6 @@ /* * renderhints.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/renderhints.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glurenderhints_h_ diff --git a/src/glu/sgi/libnurbs/internals/simplemath.h b/src/glu/sgi/libnurbs/internals/simplemath.h index f2efee35f1..195471e23e 100644 --- a/src/glu/sgi/libnurbs/internals/simplemath.h +++ b/src/glu/sgi/libnurbs/internals/simplemath.h @@ -35,8 +35,6 @@ /* * simplemath.h * - * $Date: 2002/11/01 23:35:07 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/simplemath.h,v 1.2 2002/11/01 23:35:07 brianp Exp $ */ #ifndef __glusimplemath_h_ diff --git a/src/glu/sgi/libnurbs/internals/slicer.cc b/src/glu/sgi/libnurbs/internals/slicer.cc index 3fc7e2723a..27d2a650d1 100644 --- a/src/glu/sgi/libnurbs/internals/slicer.cc +++ b/src/glu/sgi/libnurbs/internals/slicer.cc @@ -35,8 +35,6 @@ /* * slicer.c++ * - * $Date: 2005/10/28 13:09:23 $ $Revision: 1.5 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/slicer.cc,v 1.5 2005/10/28 13:09:23 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/slicer.h b/src/glu/sgi/libnurbs/internals/slicer.h index 6027eaa1c0..6700024ba2 100644 --- a/src/glu/sgi/libnurbs/internals/slicer.h +++ b/src/glu/sgi/libnurbs/internals/slicer.h @@ -35,8 +35,6 @@ /* * slicer.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/slicer.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __gluslicer_h_ diff --git a/src/glu/sgi/libnurbs/internals/sorter.cc b/src/glu/sgi/libnurbs/internals/sorter.cc index bf13a68d72..7a79941492 100644 --- a/src/glu/sgi/libnurbs/internals/sorter.cc +++ b/src/glu/sgi/libnurbs/internals/sorter.cc @@ -35,8 +35,6 @@ /* * sorter.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/sorter.cc,v 1.3 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/sorter.h b/src/glu/sgi/libnurbs/internals/sorter.h index e9c98affa7..0f6b43be37 100644 --- a/src/glu/sgi/libnurbs/internals/sorter.h +++ b/src/glu/sgi/libnurbs/internals/sorter.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/sorter.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __glusorter_h_ diff --git a/src/glu/sgi/libnurbs/internals/splitarcs.cc b/src/glu/sgi/libnurbs/internals/splitarcs.cc index 716f6b9aae..1f79d543fb 100644 --- a/src/glu/sgi/libnurbs/internals/splitarcs.cc +++ b/src/glu/sgi/libnurbs/internals/splitarcs.cc @@ -35,8 +35,6 @@ /* * splitarcs.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/splitarcs.cc,v 1.2 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/subdivider.h b/src/glu/sgi/libnurbs/internals/subdivider.h index 48aff36b44..37970d6942 100644 --- a/src/glu/sgi/libnurbs/internals/subdivider.h +++ b/src/glu/sgi/libnurbs/internals/subdivider.h @@ -35,8 +35,6 @@ /* * subdivider.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/subdivider.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __glusubdivider_h_ diff --git a/src/glu/sgi/libnurbs/internals/tobezier.cc b/src/glu/sgi/libnurbs/internals/tobezier.cc index 95ef3b68b4..531f26bc78 100644 --- a/src/glu/sgi/libnurbs/internals/tobezier.cc +++ b/src/glu/sgi/libnurbs/internals/tobezier.cc @@ -35,8 +35,6 @@ /* * tobezier.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/tobezier.cc,v 1.2 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/trimline.cc b/src/glu/sgi/libnurbs/internals/trimline.cc index 231369b6ef..61f34cd38a 100644 --- a/src/glu/sgi/libnurbs/internals/trimline.cc +++ b/src/glu/sgi/libnurbs/internals/trimline.cc @@ -35,8 +35,6 @@ /* * trimline.c++ * - * $Date: 2005/10/28 13:09:23 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimline.cc,v 1.2 2005/10/28 13:09:23 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/trimline.h b/src/glu/sgi/libnurbs/internals/trimline.h index d28574d3e5..5d52e30aba 100644 --- a/src/glu/sgi/libnurbs/internals/trimline.h +++ b/src/glu/sgi/libnurbs/internals/trimline.h @@ -35,8 +35,6 @@ /* * trimline.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimline.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutrimline_h_ diff --git a/src/glu/sgi/libnurbs/internals/trimregion.cc b/src/glu/sgi/libnurbs/internals/trimregion.cc index 64b4ffc10e..efe7893569 100644 --- a/src/glu/sgi/libnurbs/internals/trimregion.cc +++ b/src/glu/sgi/libnurbs/internals/trimregion.cc @@ -35,8 +35,6 @@ /* * trimregion.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimregion.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/trimregion.h b/src/glu/sgi/libnurbs/internals/trimregion.h index 263b8d4719..6534a8c1da 100644 --- a/src/glu/sgi/libnurbs/internals/trimregion.h +++ b/src/glu/sgi/libnurbs/internals/trimregion.h @@ -35,8 +35,6 @@ /* * trimregion.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimregion.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutrimregion_h_ diff --git a/src/glu/sgi/libnurbs/internals/trimvertex.h b/src/glu/sgi/libnurbs/internals/trimvertex.h index 8ded26a648..85f1162167 100644 --- a/src/glu/sgi/libnurbs/internals/trimvertex.h +++ b/src/glu/sgi/libnurbs/internals/trimvertex.h @@ -35,8 +35,6 @@ /* * trimvertex.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimvertex.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutrimvertex_h_ diff --git a/src/glu/sgi/libnurbs/internals/trimvertpool.cc b/src/glu/sgi/libnurbs/internals/trimvertpool.cc index 7c12ab3999..3e5bd70380 100644 --- a/src/glu/sgi/libnurbs/internals/trimvertpool.cc +++ b/src/glu/sgi/libnurbs/internals/trimvertpool.cc @@ -35,8 +35,6 @@ /* * trimvertexpool.c++ * - * $Date: 2003/05/08 15:47:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimvertpool.cc,v 1.2 2003/05/08 15:47:00 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/trimvertpool.h b/src/glu/sgi/libnurbs/internals/trimvertpool.h index deb8d4c534..2420e8cca4 100644 --- a/src/glu/sgi/libnurbs/internals/trimvertpool.h +++ b/src/glu/sgi/libnurbs/internals/trimvertpool.h @@ -35,8 +35,6 @@ /* * trimvertexpool.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimvertpool.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutrimvertpool_h_ diff --git a/src/glu/sgi/libnurbs/internals/types.h b/src/glu/sgi/libnurbs/internals/types.h index d8e7751d0b..3f89e52593 100644 --- a/src/glu/sgi/libnurbs/internals/types.h +++ b/src/glu/sgi/libnurbs/internals/types.h @@ -35,8 +35,6 @@ /* * types.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/types.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutypes_h_ diff --git a/src/glu/sgi/libnurbs/internals/uarray.cc b/src/glu/sgi/libnurbs/internals/uarray.cc index 0cc3c8d273..f0e2364373 100644 --- a/src/glu/sgi/libnurbs/internals/uarray.cc +++ b/src/glu/sgi/libnurbs/internals/uarray.cc @@ -35,8 +35,6 @@ /* * uarray.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/uarray.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/uarray.h b/src/glu/sgi/libnurbs/internals/uarray.h index e7a7e00d10..908b8ccfc8 100644 --- a/src/glu/sgi/libnurbs/internals/uarray.h +++ b/src/glu/sgi/libnurbs/internals/uarray.h @@ -35,8 +35,6 @@ /* * uarray.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/uarray.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __gluuarray_h_ diff --git a/src/glu/sgi/libnurbs/internals/varray.cc b/src/glu/sgi/libnurbs/internals/varray.cc index 969bba080e..31cc73a9d0 100644 --- a/src/glu/sgi/libnurbs/internals/varray.cc +++ b/src/glu/sgi/libnurbs/internals/varray.cc @@ -35,8 +35,6 @@ /* * varray.c++ * - * $Date: 2002/11/01 23:35:07 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/varray.cc,v 1.2 2002/11/01 23:35:07 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/varray.h b/src/glu/sgi/libnurbs/internals/varray.h index 5fb2541425..8408f27bae 100644 --- a/src/glu/sgi/libnurbs/internals/varray.h +++ b/src/glu/sgi/libnurbs/internals/varray.h @@ -35,8 +35,6 @@ /* * varray.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/varray.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __gluvarray_h_ diff --git a/src/glu/sgi/libnurbs/nurbtess/definitions.h b/src/glu/sgi/libnurbs/nurbtess/definitions.h index 216d479b1a..8dcbf20050 100644 --- a/src/glu/sgi/libnurbs/nurbtess/definitions.h +++ b/src/glu/sgi/libnurbs/nurbtess/definitions.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/definitions.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _DEFINITIONS_H diff --git a/src/glu/sgi/libnurbs/nurbtess/directedLine.h b/src/glu/sgi/libnurbs/nurbtess/directedLine.h index 009295e61e..9d68183ad3 100644 --- a/src/glu/sgi/libnurbs/nurbtess/directedLine.h +++ b/src/glu/sgi/libnurbs/nurbtess/directedLine.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/directedLine.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _DIRECTEDLINE_H diff --git a/src/glu/sgi/libnurbs/nurbtess/glimports.h b/src/glu/sgi/libnurbs/nurbtess/glimports.h index cb370218c0..2c307f63e8 100644 --- a/src/glu/sgi/libnurbs/nurbtess/glimports.h +++ b/src/glu/sgi/libnurbs/nurbtess/glimports.h @@ -35,8 +35,6 @@ /* * glimports.h * - * $Date: 2001/03/19 17:52:03 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/glimports.h,v 1.3 2001/03/19 17:52:03 pesco Exp $ */ #ifndef __gluimports_h_ diff --git a/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc b/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc index 6df10c4385..3c92039bae 100644 --- a/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc +++ b/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/gridWrap.h b/src/glu/sgi/libnurbs/nurbtess/gridWrap.h index 1c8237fe27..723988d2d0 100644 --- a/src/glu/sgi/libnurbs/nurbtess/gridWrap.h +++ b/src/glu/sgi/libnurbs/nurbtess/gridWrap.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/gridWrap.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _GRIDWRAP_H diff --git a/src/glu/sgi/libnurbs/nurbtess/monoChain.cc b/src/glu/sgi/libnurbs/nurbtess/monoChain.cc index dccbb2bbc0..814bf32fae 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoChain.cc +++ b/src/glu/sgi/libnurbs/nurbtess/monoChain.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.3 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/monoChain.cc,v 1.3 2005/10/28 13:09:23 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/monoChain.h b/src/glu/sgi/libnurbs/nurbtess/monoChain.h index e25b18028c..0302ff9ce2 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoChain.h +++ b/src/glu/sgi/libnurbs/nurbtess/monoChain.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/monoChain.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _MONO_CHAIN_H diff --git a/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc b/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc index 6405d277fe..8391205bf7 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc +++ b/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc @@ -31,7 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* *monoPolyPart.C diff --git a/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h b/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h index b760862dcb..51a664de34 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h +++ b/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h @@ -31,7 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* *monoPolyPart.h diff --git a/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc b/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc index d168374c98..8e8d49dda7 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc +++ b/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/03/29 18:46:46 $ $Revision: 1.5 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc,v 1.5 2006/03/29 18:46:46 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h b/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h index 002549ecbd..86b8b7164b 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h +++ b/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _MONO_TRIANGULATION_H diff --git a/src/glu/sgi/libnurbs/nurbtess/mystdio.h b/src/glu/sgi/libnurbs/nurbtess/mystdio.h index a7594eecd6..e9947ea393 100644 --- a/src/glu/sgi/libnurbs/nurbtess/mystdio.h +++ b/src/glu/sgi/libnurbs/nurbtess/mystdio.h @@ -35,8 +35,6 @@ /* * mystdio.h * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.4 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/mystdio.h,v 1.4 2006/03/14 15:08:52 brianp Exp $ */ #ifndef __glumystdio_h_ diff --git a/src/glu/sgi/libnurbs/nurbtess/mystdlib.h b/src/glu/sgi/libnurbs/nurbtess/mystdlib.h index d28e70bd51..2520b41e0a 100644 --- a/src/glu/sgi/libnurbs/nurbtess/mystdlib.h +++ b/src/glu/sgi/libnurbs/nurbtess/mystdlib.h @@ -35,8 +35,6 @@ /* * mystdlib.h * - * $Date: 2001/03/19 17:52:03 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/mystdlib.h,v 1.3 2001/03/19 17:52:03 pesco Exp $ */ #ifndef __glumystdlib_h_ diff --git a/src/glu/sgi/libnurbs/nurbtess/partitionX.cc b/src/glu/sgi/libnurbs/nurbtess/partitionX.cc index bfe77123c4..e25e30b0f3 100644 --- a/src/glu/sgi/libnurbs/nurbtess/partitionX.cc +++ b/src/glu/sgi/libnurbs/nurbtess/partitionX.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/partitionX.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/partitionX.h b/src/glu/sgi/libnurbs/nurbtess/partitionX.h index cd18f0eb02..bef724fe1f 100644 --- a/src/glu/sgi/libnurbs/nurbtess/partitionX.h +++ b/src/glu/sgi/libnurbs/nurbtess/partitionX.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/partitionX.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _PARTITIONX_H diff --git a/src/glu/sgi/libnurbs/nurbtess/partitionY.cc b/src/glu/sgi/libnurbs/nurbtess/partitionY.cc index 216ac07e06..297c629976 100644 --- a/src/glu/sgi/libnurbs/nurbtess/partitionY.cc +++ b/src/glu/sgi/libnurbs/nurbtess/partitionY.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/partitionY.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/partitionY.h b/src/glu/sgi/libnurbs/nurbtess/partitionY.h index b810693a5c..7e62aeaa9d 100644 --- a/src/glu/sgi/libnurbs/nurbtess/partitionY.h +++ b/src/glu/sgi/libnurbs/nurbtess/partitionY.h @@ -31,7 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* *partitionY.h: @@ -53,7 +52,6 @@ *A vertex is an interior cusp if it is a cusp and a reflex. *A vertex is an exterior cusp if it is a cusp but not a reflex. * - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/partitionY.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _PARTITIONY_H diff --git a/src/glu/sgi/libnurbs/nurbtess/polyDBG.h b/src/glu/sgi/libnurbs/nurbtess/polyDBG.h index a5125a50d1..832fe05093 100644 --- a/src/glu/sgi/libnurbs/nurbtess/polyDBG.h +++ b/src/glu/sgi/libnurbs/nurbtess/polyDBG.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/polyDBG.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _POLYDBG_H diff --git a/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc b/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc index 1a17bcc78a..f9a27f402c 100644 --- a/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc +++ b/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/polyUtil.h b/src/glu/sgi/libnurbs/nurbtess/polyUtil.h index 19c76d37d3..010838a9cc 100644 --- a/src/glu/sgi/libnurbs/nurbtess/polyUtil.h +++ b/src/glu/sgi/libnurbs/nurbtess/polyUtil.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/polyUtil.h,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #ifndef _POLYUTIL_H diff --git a/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc b/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc index 2d54b155ee..26d05342f9 100644 --- a/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc +++ b/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h b/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h index 438d4ad6b0..8063dcd622 100644 --- a/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h +++ b/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ /*we do not use the constans GL_... so that this file is independent of diff --git a/src/glu/sgi/libnurbs/nurbtess/quicksort.cc b/src/glu/sgi/libnurbs/nurbtess/quicksort.cc index f411aaa82a..9d0b290b39 100644 --- a/src/glu/sgi/libnurbs/nurbtess/quicksort.cc +++ b/src/glu/sgi/libnurbs/nurbtess/quicksort.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2002/04/17 19:30:41 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/quicksort.cc,v 1.2 2002/04/17 19:30:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/quicksort.h b/src/glu/sgi/libnurbs/nurbtess/quicksort.h index af245615b3..1a32188ee1 100644 --- a/src/glu/sgi/libnurbs/nurbtess/quicksort.h +++ b/src/glu/sgi/libnurbs/nurbtess/quicksort.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/quicksort.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _QUICKSORT_H diff --git a/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc b/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc index 932683ccac..f457b15733 100644 --- a/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc +++ b/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/rectBlock.h b/src/glu/sgi/libnurbs/nurbtess/rectBlock.h index d98b5a03e1..ce546442df 100644 --- a/src/glu/sgi/libnurbs/nurbtess/rectBlock.h +++ b/src/glu/sgi/libnurbs/nurbtess/rectBlock.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/rectBlock.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _RECTBLOCK_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc b/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc index b58de10af7..861c71bb38 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc,v 1.2 2005/10/28 13:09:23 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleComp.h b/src/glu/sgi/libnurbs/nurbtess/sampleComp.h index 8bdc4c41eb..e35e5e291d 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleComp.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleComp.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleComp.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLECOMP_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc b/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc index b66647aa99..e12f88bab1 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h b/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h index f48dceaea6..6debef9119 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLECOMPBOT_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc b/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc index e25b53c1a9..d01e50018b 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/08/30 19:02:45 $ $Revision: 1.4 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc,v 1.4 2006/08/30 19:02:45 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h b/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h index 747e35e6ad..b4b0e0732e 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLECOMPRIGHT_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc b/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc index 0d012d47ce..b7b929623a 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h b/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h index 6875ad57e2..695092c586 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLECOMPTOP_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc b/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc index c1b045437c..051f241083 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.5 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc,v 1.5 2005/10/28 13:09:23 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h b/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h index 3bfa0d4393..777a28fa2b 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLEMONOPOLY_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc b/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc index 15332eb41c..6253a7c09d 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampledLine.h b/src/glu/sgi/libnurbs/nurbtess/sampledLine.h index 8925197ab3..147b8a5e12 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampledLine.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampledLine.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampledLine.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLEDLINE_H diff --git a/src/glu/sgi/libnurbs/nurbtess/searchTree.cc b/src/glu/sgi/libnurbs/nurbtess/searchTree.cc index 45c2412b48..1865755a48 100644 --- a/src/glu/sgi/libnurbs/nurbtess/searchTree.cc +++ b/src/glu/sgi/libnurbs/nurbtess/searchTree.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/searchTree.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/searchTree.h b/src/glu/sgi/libnurbs/nurbtess/searchTree.h index 4272248528..246099fbac 100644 --- a/src/glu/sgi/libnurbs/nurbtess/searchTree.h +++ b/src/glu/sgi/libnurbs/nurbtess/searchTree.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/searchTree.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SEARCHTREE_H diff --git a/src/glu/sgi/libnurbs/nurbtess/zlassert.h b/src/glu/sgi/libnurbs/nurbtess/zlassert.h index 6a3720853b..6891385196 100644 --- a/src/glu/sgi/libnurbs/nurbtess/zlassert.h +++ b/src/glu/sgi/libnurbs/nurbtess/zlassert.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/zlassert.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ /*XXXblythe this file should be deleted*/ diff --git a/src/glu/sgi/libtess/README b/src/glu/sgi/libtess/README index 7c314b74a0..66a6011e25 100644 --- a/src/glu/sgi/libtess/README +++ b/src/glu/sgi/libtess/README @@ -1,5 +1,4 @@ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/README,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ General Polygon Tesselation diff --git a/src/glu/sgi/libtess/alg-outline b/src/glu/sgi/libtess/alg-outline index f51d68ce3b..33fd69728a 100644 --- a/src/glu/sgi/libtess/alg-outline +++ b/src/glu/sgi/libtess/alg-outline @@ -1,5 +1,4 @@ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/alg-outline,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ This is only a very brief overview. There is quite a bit of diff --git a/src/glu/sgi/libtess/dict-list.h b/src/glu/sgi/libtess/dict-list.h index f5b82116d8..8cc1069c52 100644 --- a/src/glu/sgi/libtess/dict-list.h +++ b/src/glu/sgi/libtess/dict-list.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/dict-list.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __dict_list_h_ diff --git a/src/glu/sgi/libtess/dict.c b/src/glu/sgi/libtess/dict.c index e3750eea22..f42565f2ff 100644 --- a/src/glu/sgi/libtess/dict.c +++ b/src/glu/sgi/libtess/dict.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2006/04/19 14:42:01 $ $Revision: 1.3 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/dict.c,v 1.3 2006/04/19 14:42:01 brianp Exp $ */ #include diff --git a/src/glu/sgi/libtess/dict.h b/src/glu/sgi/libtess/dict.h index ea3b4064ff..8cc1069c52 100644 --- a/src/glu/sgi/libtess/dict.h +++ b/src/glu/sgi/libtess/dict.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/dict.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __dict_list_h_ diff --git a/src/glu/sgi/libtess/geom.c b/src/glu/sgi/libtess/geom.c index d009e143ad..4551531751 100644 --- a/src/glu/sgi/libtess/geom.c +++ b/src/glu/sgi/libtess/geom.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/geom.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/memalloc.c b/src/glu/sgi/libtess/memalloc.c index 61fd59aaed..11fd33b3d7 100644 --- a/src/glu/sgi/libtess/memalloc.c +++ b/src/glu/sgi/libtess/memalloc.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/memalloc.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "memalloc.h" diff --git a/src/glu/sgi/libtess/memalloc.h b/src/glu/sgi/libtess/memalloc.h index 97f223759d..5cbdb5342e 100644 --- a/src/glu/sgi/libtess/memalloc.h +++ b/src/glu/sgi/libtess/memalloc.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2003/07/24 22:41:17 $ $Revision: 1.4 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/memalloc.h,v 1.4 2003/07/24 22:41:17 brianp Exp $ */ #ifndef __memalloc_simple_h_ diff --git a/src/glu/sgi/libtess/mesh.c b/src/glu/sgi/libtess/mesh.c index 045954db91..4ffe1a6142 100644 --- a/src/glu/sgi/libtess/mesh.c +++ b/src/glu/sgi/libtess/mesh.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/mesh.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/mesh.h b/src/glu/sgi/libtess/mesh.h index 6224df415b..299b1bacb0 100644 --- a/src/glu/sgi/libtess/mesh.h +++ b/src/glu/sgi/libtess/mesh.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/mesh.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __mesh_h_ diff --git a/src/glu/sgi/libtess/normal.h b/src/glu/sgi/libtess/normal.h index c8e334f45f..9a805d56f9 100644 --- a/src/glu/sgi/libtess/normal.h +++ b/src/glu/sgi/libtess/normal.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/normal.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __normal_h_ diff --git a/src/glu/sgi/libtess/priorityq-heap.c b/src/glu/sgi/libtess/priorityq-heap.c index 6b77155e1e..3f8a6f5f0d 100644 --- a/src/glu/sgi/libtess/priorityq-heap.c +++ b/src/glu/sgi/libtess/priorityq-heap.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq-heap.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libtess/priorityq-heap.h b/src/glu/sgi/libtess/priorityq-heap.h index 39c33c3921..095cc3456c 100644 --- a/src/glu/sgi/libtess/priorityq-heap.h +++ b/src/glu/sgi/libtess/priorityq-heap.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq-heap.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __priorityq_heap_h_ diff --git a/src/glu/sgi/libtess/priorityq-sort.h b/src/glu/sgi/libtess/priorityq-sort.h index 2439238793..9e62e983e9 100644 --- a/src/glu/sgi/libtess/priorityq-sort.h +++ b/src/glu/sgi/libtess/priorityq-sort.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq-sort.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __priorityq_sort_h_ diff --git a/src/glu/sgi/libtess/priorityq.c b/src/glu/sgi/libtess/priorityq.c index fffa1d5255..7eac424e96 100644 --- a/src/glu/sgi/libtess/priorityq.c +++ b/src/glu/sgi/libtess/priorityq.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/priorityq.h b/src/glu/sgi/libtess/priorityq.h index 97ed707578..9e62e983e9 100644 --- a/src/glu/sgi/libtess/priorityq.h +++ b/src/glu/sgi/libtess/priorityq.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __priorityq_sort_h_ diff --git a/src/glu/sgi/libtess/render.c b/src/glu/sgi/libtess/render.c index 97751dc810..c2b12b35c3 100644 --- a/src/glu/sgi/libtess/render.c +++ b/src/glu/sgi/libtess/render.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/render.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/render.h b/src/glu/sgi/libtess/render.h index 956569bb77..271e616c19 100644 --- a/src/glu/sgi/libtess/render.h +++ b/src/glu/sgi/libtess/render.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/render.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __render_h_ diff --git a/src/glu/sgi/libtess/sweep.h b/src/glu/sgi/libtess/sweep.h index 2223f52f59..74c375c9b3 100644 --- a/src/glu/sgi/libtess/sweep.h +++ b/src/glu/sgi/libtess/sweep.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/sweep.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __sweep_h_ diff --git a/src/glu/sgi/libtess/tess.h b/src/glu/sgi/libtess/tess.h index 2ba00b6ddb..d705d04c6e 100644 --- a/src/glu/sgi/libtess/tess.h +++ b/src/glu/sgi/libtess/tess.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/tess.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __tess_h_ diff --git a/src/glu/sgi/libtess/tessmono.c b/src/glu/sgi/libtess/tessmono.c index 77fe0ac619..a2b6eccbac 100644 --- a/src/glu/sgi/libtess/tessmono.c +++ b/src/glu/sgi/libtess/tessmono.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/tessmono.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/tessmono.h b/src/glu/sgi/libtess/tessmono.h index 01f244f6ec..cbe8950d20 100644 --- a/src/glu/sgi/libtess/tessmono.h +++ b/src/glu/sgi/libtess/tessmono.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/tessmono.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __tessmono_h_ diff --git a/src/glu/sgi/libutil/error.c b/src/glu/sgi/libutil/error.c index 3d1ce9b210..24d8b70f88 100644 --- a/src/glu/sgi/libutil/error.c +++ b/src/glu/sgi/libutil/error.c @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/06/20 15:30:26 $ $Revision: 1.3 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/error.c,v 1.3 2006/06/20 15:30:26 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libutil/glue.c b/src/glu/sgi/libutil/glue.c index a0471bbe2e..6a4e6c7c6f 100644 --- a/src/glu/sgi/libutil/glue.c +++ b/src/glu/sgi/libutil/glue.c @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/09/24 09:40:40 $ $Revision: 1.3 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/glue.c,v 1.3 2001/09/24 09:40:40 joukj Exp $ */ #include diff --git a/src/glu/sgi/libutil/gluint.h b/src/glu/sgi/libutil/gluint.h index f08401df7a..cd2a56fed9 100644 --- a/src/glu/sgi/libutil/gluint.h +++ b/src/glu/sgi/libutil/gluint.h @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/09/20 21:50:53 $ $Revision: 1.2 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/gluint.h,v 1.2 2001/09/20 21:50:53 kschultz Exp $ */ #ifndef __gluint_h__ diff --git a/src/glu/sgi/libutil/project.c b/src/glu/sgi/libutil/project.c index 2b20ad4fb3..5ba396ca1c 100644 --- a/src/glu/sgi/libutil/project.c +++ b/src/glu/sgi/libutil/project.c @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/05/01 16:01:17 $ $Revision: 1.6 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/project.c,v 1.6 2006/05/01 16:01:17 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libutil/registry.c b/src/glu/sgi/libutil/registry.c index e486ffa8ca..d83d2fef11 100644 --- a/src/glu/sgi/libutil/registry.c +++ b/src/glu/sgi/libutil/registry.c @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/registry.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glut/beos/beos_x11.cpp b/src/glut/beos/beos_x11.cpp index 2d1bc655cb..4f7ec48ac8 100644 --- a/src/glut/beos/beos_x11.cpp +++ b/src/glut/beos/beos_x11.cpp @@ -23,7 +23,6 @@ int DisplayHeight() { /* the following function was stolen from the X sources as indicated. */ /* Copyright Massachusetts Institute of Technology 1985, 1986, 1987 */ -/* $XConsortium: XParseGeom.c,v 11.18 91/02/21 17:23:05 rws Exp $ */ /* Permission to use, copy, modify, distribute, and sell this software and its diff --git a/src/glut/ggi/debug.h b/src/glut/ggi/debug.h index da329a1d9b..09fa960670 100644 --- a/src/glut/ggi/debug.h +++ b/src/glut/ggi/debug.h @@ -1,4 +1,4 @@ -/* $Id: debug.h,v 1.1 2000/11/19 07:41:26 jtaylor Exp $ +/* ****************************************************************************** GGIMesa debugging macros diff --git a/src/glut/glx/stroke.h b/src/glut/glx/stroke.h index fc29680bea..602b2fae9f 100644 --- a/src/glut/glx/stroke.h +++ b/src/glut/glx/stroke.h @@ -1,4 +1,3 @@ -/* $XConsortium: wfont.h,v 5.1 91/02/16 09:46:37 rws Exp $ */ /***************************************************************** Copyright (c) 1989,1990, 1991 by Sun Microsystems, Inc. and the X Consortium. diff --git a/src/glut/glx/win32_x11.c b/src/glut/glx/win32_x11.c index 1d138cfa2a..d00ccdb121 100644 --- a/src/glut/glx/win32_x11.c +++ b/src/glut/glx/win32_x11.c @@ -263,7 +263,6 @@ XPending(Display* display) /* the following function was stolen from the X sources as indicated. */ /* Copyright Massachusetts Institute of Technology 1985, 1986, 1987 */ -/* $XConsortium: XParseGeom.c,v 11.18 91/02/21 17:23:05 rws Exp $ */ /* Permission to use, copy, modify, distribute, and sell this software and its diff --git a/src/glx/mini/miniglx_events.c b/src/glx/mini/miniglx_events.c index 969398bc16..a20d5847b3 100644 --- a/src/glx/mini/miniglx_events.c +++ b/src/glx/mini/miniglx_events.c @@ -38,7 +38,6 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/* $Id: miniglx_events.c,v 1.6 2006/04/03 07:31:27 airlied Exp $ */ #include diff --git a/src/glx/x11/XF86dri.c b/src/glx/x11/XF86dri.c index 8909a04772..9919a40977 100644 --- a/src/glx/x11/XF86dri.c +++ b/src/glx/x11/XF86dri.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/dri/XF86dri.c,v 1.13 2002/10/30 12:51:25 alanh Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/glx/x11/clientattrib.c b/src/glx/x11/clientattrib.c index bfb263ced1..888f8e3187 100644 --- a/src/glx/x11/clientattrib.c +++ b/src/glx/x11/clientattrib.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/clientattrib.c,v 1.5 2001/03/21 16:04:39 dawes Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/compsize.c b/src/glx/x11/compsize.c index b8c162e8ac..2d124573ef 100644 --- a/src/glx/x11/compsize.c +++ b/src/glx/x11/compsize.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/compsize.c,v 1.6 2004/01/28 18:11:38 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/dri_glx.c b/src/glx/x11/dri_glx.c index 5cf9923979..21e07c1935 100644 --- a/src/glx/x11/dri_glx.c +++ b/src/glx/x11/dri_glx.c @@ -24,7 +24,6 @@ TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/dri/dri_glx.c,v 1.14 2003/07/16 00:54:00 dawes Exp $ */ /* * Authors: diff --git a/src/glx/x11/eval.c b/src/glx/x11/eval.c index 0f94e6da6f..2544c50fce 100644 --- a/src/glx/x11/eval.c +++ b/src/glx/x11/eval.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/glxclient.h b/src/glx/x11/glxclient.h index 477566cc46..03e44e5d04 100644 --- a/src/glx/x11/glxclient.h +++ b/src/glx/x11/glxclient.h @@ -31,7 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. */ -/* $XFree86: xc/lib/GL/glx/glxclient.h,v 1.21 2004/02/09 23:46:31 alanh Exp $ */ /** * \file glxclient.h diff --git a/src/glx/x11/glxcmds.c b/src/glx/x11/glxcmds.c index f52b71ffcd..80281896f6 100644 --- a/src/glx/x11/glxcmds.c +++ b/src/glx/x11/glxcmds.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/glxcmds.c,v 1.30 2004/01/30 20:33:06 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/glxext.c b/src/glx/x11/glxext.c index af3a5166dc..2852217ba3 100644 --- a/src/glx/x11/glxext.c +++ b/src/glx/x11/glxext.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/glxext.c,v 1.22 2003/12/08 17:35:28 dawes Exp $ */ /* ** License Applicability. Except to the extent portions of this file are diff --git a/src/glx/x11/indirect_init.h b/src/glx/x11/indirect_init.h index 62d04ba6dc..72255f1301 100644 --- a/src/glx/x11/indirect_init.h +++ b/src/glx/x11/indirect_init.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/indirect_init.h,v 1.2 2000/02/08 17:18:33 dawes Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/glx/x11/packrender.h b/src/glx/x11/packrender.h index ce2a1616de..8e3119d1b2 100644 --- a/src/glx/x11/packrender.h +++ b/src/glx/x11/packrender.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/packrender.h,v 1.7tsi Exp $ */ #ifndef __GLX_packrender_h__ #define __GLX_packrender_h__ diff --git a/src/glx/x11/packsingle.h b/src/glx/x11/packsingle.h index 16b054f1e0..c69c543921 100644 --- a/src/glx/x11/packsingle.h +++ b/src/glx/x11/packsingle.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/packsingle.h,v 1.5tsi Exp $ */ #ifndef __GLX_packsingle_h__ #define __GLX_packsingle_h__ diff --git a/src/glx/x11/pixel.c b/src/glx/x11/pixel.c index 3b3a1811ab..279555bdfd 100644 --- a/src/glx/x11/pixel.c +++ b/src/glx/x11/pixel.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/pixel.c,v 1.8 2003/09/28 20:15:04 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/pixelstore.c b/src/glx/x11/pixelstore.c index 3bf1b35ba3..6f25ed786e 100644 --- a/src/glx/x11/pixelstore.c +++ b/src/glx/x11/pixelstore.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/pixelstore.c,v 1.4 2004/01/28 18:11:43 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/render2.c b/src/glx/x11/render2.c index 21ba270998..b17ad974c8 100644 --- a/src/glx/x11/render2.c +++ b/src/glx/x11/render2.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/render2.c,v 1.6 2004/01/31 09:29:33 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/renderpix.c b/src/glx/x11/renderpix.c index b7d01dc679..41a7a2d762 100644 --- a/src/glx/x11/renderpix.c +++ b/src/glx/x11/renderpix.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/renderpix.c,v 1.5 2003/09/28 20:15:04 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/single2.c b/src/glx/x11/single2.c index d535757a9e..35fe417b62 100644 --- a/src/glx/x11/single2.c +++ b/src/glx/x11/single2.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/single2.c,v 1.10 2004/02/11 19:48:16 dawes Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/singlepix.c b/src/glx/x11/singlepix.c index a7b5b79870..cd88684f70 100644 --- a/src/glx/x11/singlepix.c +++ b/src/glx/x11/singlepix.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/singlepix.c,v 1.3 2001/03/21 16:04:39 dawes Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/vertarr.c b/src/glx/x11/vertarr.c index 483a166ea2..d50560ba1a 100644 --- a/src/glx/x11/vertarr.c +++ b/src/glx/x11/vertarr.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/vertarr.c,v 1.4 2001/03/25 05:32:00 tsi Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/xf86dri.h b/src/glx/x11/xf86dri.h index 0a2bb24971..c8c878f127 100644 --- a/src/glx/x11/xf86dri.h +++ b/src/glx/x11/xf86dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/dri/xf86dri.h,v 1.8 2002/10/30 12:51:25 alanh Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/glx/x11/xf86dristr.h b/src/glx/x11/xf86dristr.h index ac05b183b3..b834bd1a1a 100644 --- a/src/glx/x11/xf86dristr.h +++ b/src/glx/x11/xf86dristr.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/dri/xf86dristr.h,v 1.10 2002/10/30 12:51:25 alanh Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/glx/x11/xfont.c b/src/glx/x11/xfont.c index 5f23a79622..f3e3da3e79 100644 --- a/src/glx/x11/xfont.c +++ b/src/glx/x11/xfont.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/xfont.c,v 1.6 2001/05/02 15:06:02 dawes Exp $ */ /* * Mesa 3-D graphics library * Version: 3.1 diff --git a/src/mesa/drivers/dri/common/stenciltmp.h b/src/mesa/drivers/dri/common/stenciltmp.h index 324fc873d3..2b10b9ecfe 100644 --- a/src/mesa/drivers/dri/common/stenciltmp.h +++ b/src/mesa/drivers/dri/common/stenciltmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/common/stenciltmp.h,v 1.3 2001/03/21 16:14:20 dawes Exp $ */ #include "spantmp_common.h" diff --git a/src/mesa/drivers/dri/common/texmem.c b/src/mesa/drivers/dri/common/texmem.c index b0e8c4c1c2..a81cc2413d 100644 --- a/src/mesa/drivers/dri/common/texmem.c +++ b/src/mesa/drivers/dri/common/texmem.c @@ -28,7 +28,6 @@ * Kevin E. Martin * Gareth Hughes */ -/* $XFree86:$ */ /** \file texmem.c * Implements all of the device-independent texture memory management. diff --git a/src/mesa/drivers/dri/common/texmem.h b/src/mesa/drivers/dri/common/texmem.h index 6692efcc30..ffed7dd66e 100644 --- a/src/mesa/drivers/dri/common/texmem.h +++ b/src/mesa/drivers/dri/common/texmem.h @@ -28,7 +28,6 @@ * Kevin E. Martin * Gareth Hughes */ -/* $XFree86:$ */ /** \file texmem.h * Public interface to the DRI texture memory management routines. diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h index b2bab86e66..b28b895627 100644 --- a/src/mesa/drivers/dri/common/utils.h +++ b/src/mesa/drivers/dri/common/utils.h @@ -24,7 +24,6 @@ * Authors: * Ian Romanick */ -/* $XFree86:$ */ #ifndef DRI_DEBUG_H #define DRI_DEBUG_H diff --git a/src/mesa/drivers/dri/common/vblank.c b/src/mesa/drivers/dri/common/vblank.c index e7ed545f13..094950d362 100644 --- a/src/mesa/drivers/dri/common/vblank.c +++ b/src/mesa/drivers/dri/common/vblank.c @@ -25,7 +25,6 @@ * Authors: * Ian Romanick */ -/* $XFree86:$ */ #include "glheader.h" #include "xf86drm.h" diff --git a/src/mesa/drivers/dri/common/vblank.h b/src/mesa/drivers/dri/common/vblank.h index ec83adc78d..52c1933ca5 100644 --- a/src/mesa/drivers/dri/common/vblank.h +++ b/src/mesa/drivers/dri/common/vblank.h @@ -25,7 +25,6 @@ * Authors: * Ian Romanick */ -/* $XFree86:$ */ #ifndef DRI_VBLANK_H #define DRI_VBLANK_H diff --git a/src/mesa/drivers/dri/ffb/ffb_bitmap.c b/src/mesa/drivers/dri/ffb/ffb_bitmap.c index 7263e83813..1aa66859a6 100644 --- a/src/mesa/drivers/dri/ffb/ffb_bitmap.c +++ b/src/mesa/drivers/dri/ffb/ffb_bitmap.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_bitmap.c,v 1.1 2002/02/22 21:32:58 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_bitmap.h b/src/mesa/drivers/dri/ffb/ffb_bitmap.h index 4f8d2ea2a6..0ccbc57bd0 100644 --- a/src/mesa/drivers/dri/ffb/ffb_bitmap.h +++ b/src/mesa/drivers/dri/ffb/ffb_bitmap.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_bitmap.h,v 1.1 2002/02/22 21:32:58 dawes Exp $ */ #ifndef _FFB_BITMAP_H #define _FFB_BITMAP_H diff --git a/src/mesa/drivers/dri/ffb/ffb_clear.c b/src/mesa/drivers/dri/ffb/ffb_clear.c index e8dfcbe254..7de05b5cf0 100644 --- a/src/mesa/drivers/dri/ffb/ffb_clear.c +++ b/src/mesa/drivers/dri/ffb/ffb_clear.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_clear.c,v 1.2 2002/02/22 21:32:58 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_context.h b/src/mesa/drivers/dri/ffb/ffb_context.h index df1b65d748..0ab75fce47 100644 --- a/src/mesa/drivers/dri/ffb/ffb_context.h +++ b/src/mesa/drivers/dri/ffb/ffb_context.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_context.h,v 1.2 2002/02/22 21:32:58 dawes Exp $ */ #ifndef _FFB_CONTEXT_H #define _FFB_CONTEXT_H diff --git a/src/mesa/drivers/dri/ffb/ffb_dd.c b/src/mesa/drivers/dri/ffb/ffb_dd.c index 53423bbae4..f64a577d1f 100644 --- a/src/mesa/drivers/dri/ffb/ffb_dd.c +++ b/src/mesa/drivers/dri/ffb/ffb_dd.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_dd.c,v 1.4 2002/09/11 19:49:07 tsi Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_dd.h b/src/mesa/drivers/dri/ffb/ffb_dd.h index 4ffcbe6666..e065ebbecd 100644 --- a/src/mesa/drivers/dri/ffb/ffb_dd.h +++ b/src/mesa/drivers/dri/ffb/ffb_dd.h @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_dd.h,v 1.1 2000/06/20 05:08:38 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D. * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_depth.c b/src/mesa/drivers/dri/ffb/ffb_depth.c index 68a2450eb7..cca6212f50 100644 --- a/src/mesa/drivers/dri/ffb/ffb_depth.c +++ b/src/mesa/drivers/dri/ffb/ffb_depth.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_depth.c,v 1.2 2002/02/22 21:32:58 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_depth.h b/src/mesa/drivers/dri/ffb/ffb_depth.h index db908e7a63..8a1829ed49 100644 --- a/src/mesa/drivers/dri/ffb/ffb_depth.h +++ b/src/mesa/drivers/dri/ffb/ffb_depth.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_depth.h,v 1.1 2000/06/20 05:08:38 dawes Exp $ */ #ifndef _FFB_DEPTH_H #define _FFB_DEPTH_H diff --git a/src/mesa/drivers/dri/ffb/ffb_fifo.h b/src/mesa/drivers/dri/ffb/ffb_fifo.h index 886d71b76e..a175f38643 100644 --- a/src/mesa/drivers/dri/ffb/ffb_fifo.h +++ b/src/mesa/drivers/dri/ffb/ffb_fifo.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_fifo.h,v 1.2 2002/02/22 21:32:58 dawes Exp $ */ #ifndef _FFB_FIFO_H #define _FFB_FIFO_H diff --git a/src/mesa/drivers/dri/ffb/ffb_lines.c b/src/mesa/drivers/dri/ffb/ffb_lines.c index da1de18f36..8294701464 100644 --- a/src/mesa/drivers/dri/ffb/ffb_lines.c +++ b/src/mesa/drivers/dri/ffb/ffb_lines.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_lines.c,v 1.2 2002/02/22 21:32:58 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_lines.h b/src/mesa/drivers/dri/ffb/ffb_lines.h index d508c243ea..ddb9365653 100644 --- a/src/mesa/drivers/dri/ffb/ffb_lines.h +++ b/src/mesa/drivers/dri/ffb/ffb_lines.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_lines.h,v 1.2 2002/02/22 21:32:58 dawes Exp $ */ #ifndef _FFB_LINES_H #define _FFB_LINES_H diff --git a/src/mesa/drivers/dri/ffb/ffb_linetmp.h b/src/mesa/drivers/dri/ffb/ffb_linetmp.h index 0951513ca1..e9d8260e1a 100644 --- a/src/mesa/drivers/dri/ffb/ffb_linetmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_linetmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_linetmp.h,v 1.2 2002/02/22 21:32:58 dawes Exp $ */ static __inline void TAG(ffb_line)(GLcontext *ctx, ffb_vertex *v0, ffb_vertex *v1 ) diff --git a/src/mesa/drivers/dri/ffb/ffb_lock.h b/src/mesa/drivers/dri/ffb/ffb_lock.h index 7c49f740f8..1fd3eb5512 100644 --- a/src/mesa/drivers/dri/ffb/ffb_lock.h +++ b/src/mesa/drivers/dri/ffb/ffb_lock.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_lock.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_LOCK_H #define _FFB_LOCK_H diff --git a/src/mesa/drivers/dri/ffb/ffb_points.c b/src/mesa/drivers/dri/ffb/ffb_points.c index a7496dd1d6..d00255ccee 100644 --- a/src/mesa/drivers/dri/ffb/ffb_points.c +++ b/src/mesa/drivers/dri/ffb/ffb_points.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_points.c,v 1.2 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_points.h b/src/mesa/drivers/dri/ffb/ffb_points.h index 7d5c1f8a03..a7229de7f1 100644 --- a/src/mesa/drivers/dri/ffb/ffb_points.h +++ b/src/mesa/drivers/dri/ffb/ffb_points.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_points.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_POINTS_H #define _FFB_POINTS_H diff --git a/src/mesa/drivers/dri/ffb/ffb_pointtmp.h b/src/mesa/drivers/dri/ffb/ffb_pointtmp.h index 310c95d89b..2c91426b3a 100644 --- a/src/mesa/drivers/dri/ffb/ffb_pointtmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_pointtmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_pointtmp.h,v 1.3 2002/02/22 21:32:59 dawes Exp $ */ static __inline void TAG(ffb_draw_point)(GLcontext *ctx, ffb_vertex *tmp ) { diff --git a/src/mesa/drivers/dri/ffb/ffb_rendertmp.h b/src/mesa/drivers/dri/ffb/ffb_rendertmp.h index 26d991b081..64141c2c5f 100644 --- a/src/mesa/drivers/dri/ffb/ffb_rendertmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_rendertmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_rendertmp.h,v 1.2 2003/01/29 23:00:40 dawes Exp $ */ #define IMPL_LOCAL_VARS \ ffbContextPtr fmesa = FFB_CONTEXT(ctx); \ diff --git a/src/mesa/drivers/dri/ffb/ffb_span.c b/src/mesa/drivers/dri/ffb/ffb_span.c index fff7fa1d3f..59ac414678 100644 --- a/src/mesa/drivers/dri/ffb/ffb_span.c +++ b/src/mesa/drivers/dri/ffb/ffb_span.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_span.c,v 1.2 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_span.h b/src/mesa/drivers/dri/ffb/ffb_span.h index 5ae227910d..37506cf30e 100644 --- a/src/mesa/drivers/dri/ffb/ffb_span.h +++ b/src/mesa/drivers/dri/ffb/ffb_span.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_span.h,v 1.1 2000/06/20 05:08:39 dawes Exp $ */ #ifndef _FFB_SPAN_H #define _FFB_SPAN_H diff --git a/src/mesa/drivers/dri/ffb/ffb_state.c b/src/mesa/drivers/dri/ffb/ffb_state.c index eb13478166..880ad8be0a 100644 --- a/src/mesa/drivers/dri/ffb/ffb_state.c +++ b/src/mesa/drivers/dri/ffb/ffb_state.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_state.c,v 1.5 2002/10/30 12:51:27 alanh Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_state.h b/src/mesa/drivers/dri/ffb/ffb_state.h index 17b6fa20ab..19e72085fd 100644 --- a/src/mesa/drivers/dri/ffb/ffb_state.h +++ b/src/mesa/drivers/dri/ffb/ffb_state.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_state.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_STATE_H #define _FFB_STATE_H diff --git a/src/mesa/drivers/dri/ffb/ffb_stencil.c b/src/mesa/drivers/dri/ffb/ffb_stencil.c index 2f13ee9210..d535b1b778 100644 --- a/src/mesa/drivers/dri/ffb/ffb_stencil.c +++ b/src/mesa/drivers/dri/ffb/ffb_stencil.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_stencil.c,v 1.2 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_stencil.h b/src/mesa/drivers/dri/ffb/ffb_stencil.h index c7da1ca681..2d529980d1 100644 --- a/src/mesa/drivers/dri/ffb/ffb_stencil.h +++ b/src/mesa/drivers/dri/ffb/ffb_stencil.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_stencil.h,v 1.1 2000/06/20 05:08:39 dawes Exp $ */ #ifndef _FFB_STENCIL_H #define _FFB_STENCIL_H diff --git a/src/mesa/drivers/dri/ffb/ffb_tex.c b/src/mesa/drivers/dri/ffb/ffb_tex.c index d6763b7cd3..6503b0f4e7 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tex.c +++ b/src/mesa/drivers/dri/ffb/ffb_tex.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tex.c,v 1.1 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_tex.h b/src/mesa/drivers/dri/ffb/ffb_tex.h index dba0e08af6..4032e73209 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tex.h +++ b/src/mesa/drivers/dri/ffb/ffb_tex.h @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tex.h,v 1.1 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D. * Copyright (C) 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_tris.c b/src/mesa/drivers/dri/ffb/ffb_tris.c index 9fae8c8283..c2857f61bd 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tris.c +++ b/src/mesa/drivers/dri/ffb/ffb_tris.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tris.c,v 1.3 2002/10/30 12:51:28 alanh Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_tris.h b/src/mesa/drivers/dri/ffb/ffb_tris.h index a803174b3e..116b8e07f1 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tris.h +++ b/src/mesa/drivers/dri/ffb/ffb_tris.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tris.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_TRIS_H #define _FFB_TRIS_H diff --git a/src/mesa/drivers/dri/ffb/ffb_tritmp.h b/src/mesa/drivers/dri/ffb/ffb_tritmp.h index 612ef2433f..324a871ec4 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tritmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_tritmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tritmp.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ static void TAG(ffb_triangle)( GLcontext *ctx, ffb_vertex *v0, diff --git a/src/mesa/drivers/dri/ffb/ffb_vb.c b/src/mesa/drivers/dri/ffb/ffb_vb.c index 6ba1eabbf2..edc9d79124 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vb.c +++ b/src/mesa/drivers/dri/ffb/ffb_vb.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vb.c,v 1.4 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_vb.h b/src/mesa/drivers/dri/ffb/ffb_vb.h index 9eb6759f61..af669bce30 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vb.h +++ b/src/mesa/drivers/dri/ffb/ffb_vb.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vb.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_VB_H #define _FFB_VB_H diff --git a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h index a1d1254d97..0495d0e276 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vbtmp.h,v 1.1 2002/02/22 21:32:59 dawes Exp $ */ static void TAG(emit)(GLcontext *ctx, GLuint start, GLuint end) { diff --git a/src/mesa/drivers/dri/ffb/ffb_vtxfmt.c b/src/mesa/drivers/dri/ffb/ffb_vtxfmt.c index 9c1b770fbd..8b60f095c9 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vtxfmt.c +++ b/src/mesa/drivers/dri/ffb/ffb_vtxfmt.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vtxfmt.c,v 1.1 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_vtxfmt.h b/src/mesa/drivers/dri/ffb/ffb_vtxfmt.h index 063bb4923e..4d9125cd15 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vtxfmt.h +++ b/src/mesa/drivers/dri/ffb/ffb_vtxfmt.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vtxfmt.h,v 1.1 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_VTXFMT_H #define _FFB_VTXFMT_H diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.c b/src/mesa/drivers/dri/ffb/ffb_xmesa.c index 4c5323d230..f521de63c0 100644 --- a/src/mesa/drivers/dri/ffb/ffb_xmesa.c +++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_xmesa.c,v 1.4 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.h b/src/mesa/drivers/dri/ffb/ffb_xmesa.h index b7580780a6..bc8cfe9f21 100644 --- a/src/mesa/drivers/dri/ffb/ffb_xmesa.h +++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_xmesa.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_XMESA_H_ #define _FFB_XMESA_H_ diff --git a/src/mesa/drivers/dri/ffb/server/ffb_dac.h b/src/mesa/drivers/dri/ffb/server/ffb_dac.h index 08114282e5..ac4a75b459 100644 --- a/src/mesa/drivers/dri/ffb/server/ffb_dac.h +++ b/src/mesa/drivers/dri/ffb/server/ffb_dac.h @@ -21,7 +21,6 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sunffb/ffb_dac.h,v 1.2 2001/04/05 17:42:33 dawes Exp $ */ #ifndef _FFB_DAC_H #define _FFB_DAC_H diff --git a/src/mesa/drivers/dri/ffb/server/ffb_drishare.h b/src/mesa/drivers/dri/ffb/server/ffb_drishare.h index baf2f0d0a6..69fefa3f0a 100644 --- a/src/mesa/drivers/dri/ffb/server/ffb_drishare.h +++ b/src/mesa/drivers/dri/ffb/server/ffb_drishare.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sunffb/ffb_drishare.h,v 1.2 2000/06/21 00:47:37 dawes Exp $ */ #ifndef _FFB_DRISHARE_H #define _FFB_DRISHARE_H diff --git a/src/mesa/drivers/dri/ffb/server/ffb_regs.h b/src/mesa/drivers/dri/ffb/server/ffb_regs.h index 7f383d38d6..bda5840d60 100644 --- a/src/mesa/drivers/dri/ffb/server/ffb_regs.h +++ b/src/mesa/drivers/dri/ffb/server/ffb_regs.h @@ -24,7 +24,6 @@ * USE OR OTHER DEALINGS IN THE SOFTWARE. * */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sunffb/ffb_regs.h,v 1.1 2000/05/18 23:21:37 dawes Exp $ */ #ifndef FFBREGS_H #define FFBREGS_H diff --git a/src/mesa/drivers/dri/gamma/gamma_client.h b/src/mesa/drivers/dri/gamma/gamma_client.h index 1c1a22ebc4..6dcf2e9438 100644 --- a/src/mesa/drivers/dri/gamma/gamma_client.h +++ b/src/mesa/drivers/dri/gamma/gamma_client.h @@ -31,7 +31,6 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_client.h,v 1.3 2002/02/22 21:33:00 dawes Exp $ * */ diff --git a/src/mesa/drivers/dri/gamma/gamma_context.h b/src/mesa/drivers/dri/gamma/gamma_context.h index f0ab1c4f05..fb70df6c37 100644 --- a/src/mesa/drivers/dri/gamma/gamma_context.h +++ b/src/mesa/drivers/dri/gamma/gamma_context.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_context.h,v 1.6 2002/12/16 16:18:50 dawes Exp $ */ /* * Copyright 2001 by Alan Hourihane. * diff --git a/src/mesa/drivers/dri/gamma/gamma_inithw.c b/src/mesa/drivers/dri/gamma/gamma_inithw.c index 47eb802b4e..79b54aacb5 100644 --- a/src/mesa/drivers/dri/gamma/gamma_inithw.c +++ b/src/mesa/drivers/dri/gamma/gamma_inithw.c @@ -23,7 +23,6 @@ * Kevin E. Martin * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_inithw.c,v 1.9 2002/10/30 12:51:29 alanh Exp $ */ #include "gamma_context.h" #include "glint_dri.h" diff --git a/src/mesa/drivers/dri/gamma/gamma_lock.c b/src/mesa/drivers/dri/gamma/gamma_lock.c index 2ab387fa27..97eea75541 100644 --- a/src/mesa/drivers/dri/gamma/gamma_lock.c +++ b/src/mesa/drivers/dri/gamma/gamma_lock.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_lock.c,v 1.4 2002/11/05 17:46:07 tsi Exp $ */ #include "gamma_context.h" #include "gamma_lock.h" diff --git a/src/mesa/drivers/dri/gamma/gamma_macros.h b/src/mesa/drivers/dri/gamma/gamma_macros.h index 974fe569df..c15483b770 100644 --- a/src/mesa/drivers/dri/gamma/gamma_macros.h +++ b/src/mesa/drivers/dri/gamma/gamma_macros.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_macros.h,v 1.5 2002/02/22 21:33:02 dawes Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/gamma/gamma_regs.h b/src/mesa/drivers/dri/gamma/gamma_regs.h index 2edda07227..9e1c735019 100644 --- a/src/mesa/drivers/dri/gamma/gamma_regs.h +++ b/src/mesa/drivers/dri/gamma/gamma_regs.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_regs.h,v 1.5 2002/02/22 21:33:02 dawes Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/gamma/gamma_span.c b/src/mesa/drivers/dri/gamma/gamma_span.c index f62bea9b66..012d77782b 100644 --- a/src/mesa/drivers/dri/gamma/gamma_span.c +++ b/src/mesa/drivers/dri/gamma/gamma_span.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_span.c,v 1.4 2002/11/05 17:46:07 tsi Exp $ */ #include "gamma_context.h" #include "gamma_lock.h" diff --git a/src/mesa/drivers/dri/gamma/gamma_state.c b/src/mesa/drivers/dri/gamma/gamma_state.c index 8dbe0a97ca..a0690f64d0 100644 --- a/src/mesa/drivers/dri/gamma/gamma_state.c +++ b/src/mesa/drivers/dri/gamma/gamma_state.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_state.c,v 1.5 2002/11/05 17:46:07 tsi Exp $ */ /* * Copyright 2001 by Alan Hourihane. * diff --git a/src/mesa/drivers/dri/gamma/gamma_tex.c b/src/mesa/drivers/dri/gamma/gamma_tex.c index d4fc93f86b..0770cbf694 100644 --- a/src/mesa/drivers/dri/gamma/gamma_tex.c +++ b/src/mesa/drivers/dri/gamma/gamma_tex.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_tex.c,v 1.4 2002/11/05 17:46:07 tsi Exp $ */ #include #include diff --git a/src/mesa/drivers/dri/gamma/gamma_texmem.c b/src/mesa/drivers/dri/gamma/gamma_texmem.c index 506b5c4c8f..94ecb5c2f6 100644 --- a/src/mesa/drivers/dri/gamma/gamma_texmem.c +++ b/src/mesa/drivers/dri/gamma/gamma_texmem.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_texmem.c,v 1.5 2002/11/05 17:46:07 tsi Exp $ */ #include #include diff --git a/src/mesa/drivers/dri/gamma/gamma_texstate.c b/src/mesa/drivers/dri/gamma/gamma_texstate.c index a8d1b253c7..b9bd6d4cee 100644 --- a/src/mesa/drivers/dri/gamma/gamma_texstate.c +++ b/src/mesa/drivers/dri/gamma/gamma_texstate.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_texstate.c,v 1.5 2002/11/05 17:46:07 tsi Exp $ */ #include #include diff --git a/src/mesa/drivers/dri/gamma/gamma_tritmp.h b/src/mesa/drivers/dri/gamma/gamma_tritmp.h index 23459ff156..56e0a850c8 100644 --- a/src/mesa/drivers/dri/gamma/gamma_tritmp.h +++ b/src/mesa/drivers/dri/gamma/gamma_tritmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/extras/Mesa/src/mesa/drivers/dri/gamma/gamma_tritmp.h,v 1.2 2004/12/13 22:40:49 tsi Exp $ */ static void TAG(gamma_point)( gammaContextPtr gmesa, const gammaVertex *v0 ) diff --git a/src/mesa/drivers/dri/gamma/gamma_vb.c b/src/mesa/drivers/dri/gamma/gamma_vb.c index 80d35cba9e..f23f585fc0 100644 --- a/src/mesa/drivers/dri/gamma/gamma_vb.c +++ b/src/mesa/drivers/dri/gamma/gamma_vb.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_vb.c,v 1.4 2003/03/26 20:43:48 tsi Exp $ */ /* * Copyright 2001 by Alan Hourihane. * diff --git a/src/mesa/drivers/dri/gamma/gamma_xmesa.c b/src/mesa/drivers/dri/gamma/gamma_xmesa.c index f41682cea7..4c0ebe1899 100644 --- a/src/mesa/drivers/dri/gamma/gamma_xmesa.c +++ b/src/mesa/drivers/dri/gamma/gamma_xmesa.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_xmesa.c,v 1.14 2002/10/30 12:51:30 alanh Exp $ */ /* * Copyright 2001 by Alan Hourihane. * diff --git a/src/mesa/drivers/dri/gamma/server/glint_common.h b/src/mesa/drivers/dri/gamma/server/glint_common.h index ec601f942d..36554e4ac2 100644 --- a/src/mesa/drivers/dri/gamma/server/glint_common.h +++ b/src/mesa/drivers/dri/gamma/server/glint_common.h @@ -25,7 +25,6 @@ * Converted to common header format: * Jens Owen * - * $XFree86: xc/programs/Xserver/hw/xfree86/drivers/glint/glint_common.h,v 1.2 2003/04/03 16:52:18 dawes Exp $ * */ diff --git a/src/mesa/drivers/dri/gamma/server/glint_dri.h b/src/mesa/drivers/dri/gamma/server/glint_dri.h index 3952759f83..df1992a5d1 100644 --- a/src/mesa/drivers/dri/gamma/server/glint_dri.h +++ b/src/mesa/drivers/dri/gamma/server/glint_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/glint/glint_dri.h,v 1.7 2002/10/30 12:52:16 alanh Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/i810/i810_3d_reg.h b/src/mesa/drivers/dri/i810/i810_3d_reg.h index 7cc59d5c86..2fbeb64978 100644 --- a/src/mesa/drivers/dri/i810/i810_3d_reg.h +++ b/src/mesa/drivers/dri/i810/i810_3d_reg.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810_3d_reg.h,v 1.7 2002/02/22 21:33:03 dawes Exp $ */ #ifndef I810_3D_REG_H #define I810_3D_REG_H diff --git a/src/mesa/drivers/dri/i810/i810context.c b/src/mesa/drivers/dri/i810/i810context.c index 3f7f2cc8a4..f90b3682f8 100644 --- a/src/mesa/drivers/dri/i810/i810context.c +++ b/src/mesa/drivers/dri/i810/i810context.c @@ -24,7 +24,6 @@ TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810context.c,v 1.3 2002/10/30 12:51:33 alanh Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/i810/i810context.h b/src/mesa/drivers/dri/i810/i810context.h index b83500bbd0..4708042059 100644 --- a/src/mesa/drivers/dri/i810/i810context.h +++ b/src/mesa/drivers/dri/i810/i810context.h @@ -21,7 +21,6 @@ * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810context.h,v 1.9 2002/12/16 16:18:51 dawes Exp $ */ #ifndef I810CONTEXT_INC #define I810CONTEXT_INC diff --git a/src/mesa/drivers/dri/i810/i810ioctl.c b/src/mesa/drivers/dri/i810/i810ioctl.c index 57c84193fa..95726fb252 100644 --- a/src/mesa/drivers/dri/i810/i810ioctl.c +++ b/src/mesa/drivers/dri/i810/i810ioctl.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810ioctl.c,v 1.7 2002/10/30 12:51:33 alanh Exp $ */ #include /* for usleep() */ diff --git a/src/mesa/drivers/dri/i810/i810ioctl.h b/src/mesa/drivers/dri/i810/i810ioctl.h index 61399ee7b7..748d29ae36 100644 --- a/src/mesa/drivers/dri/i810/i810ioctl.h +++ b/src/mesa/drivers/dri/i810/i810ioctl.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810ioctl.h,v 1.7 2002/10/30 12:51:33 alanh Exp $ */ #ifndef I810_IOCTL_H #define I810_IOCTL_H diff --git a/src/mesa/drivers/dri/i810/i810screen.c b/src/mesa/drivers/dri/i810/i810screen.c index f64c10a9ae..695b996319 100644 --- a/src/mesa/drivers/dri/i810/i810screen.c +++ b/src/mesa/drivers/dri/i810/i810screen.c @@ -24,7 +24,6 @@ TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810screen.c,v 1.2 2002/10/30 12:51:33 alanh Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/i810/i810state.c b/src/mesa/drivers/dri/i810/i810state.c index e0d5b2b487..e203c74f52 100644 --- a/src/mesa/drivers/dri/i810/i810state.c +++ b/src/mesa/drivers/dri/i810/i810state.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810state.c,v 1.9 2002/10/30 12:51:33 alanh Exp $ */ #include diff --git a/src/mesa/drivers/dri/i810/i810tex.c b/src/mesa/drivers/dri/i810/i810tex.c index f657abe671..730bc90eaf 100644 --- a/src/mesa/drivers/dri/i810/i810tex.c +++ b/src/mesa/drivers/dri/i810/i810tex.c @@ -21,7 +21,6 @@ * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810tex.c,v 1.9 2002/10/30 12:51:33 alanh Exp $ */ #include "glheader.h" #include "mtypes.h" diff --git a/src/mesa/drivers/dri/i810/i810tris.c b/src/mesa/drivers/dri/i810/i810tris.c index 2c4ee06633..40ab436b95 100644 --- a/src/mesa/drivers/dri/i810/i810tris.c +++ b/src/mesa/drivers/dri/i810/i810tris.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810tris.c,v 1.7 2002/10/30 12:51:33 alanh Exp $ */ /************************************************************************** Copyright 2001 VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/i810/i810tris.h b/src/mesa/drivers/dri/i810/i810tris.h index 06c8b3fcd5..3d0dd916ca 100644 --- a/src/mesa/drivers/dri/i810/i810tris.h +++ b/src/mesa/drivers/dri/i810/i810tris.h @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810tris.h,v 1.10 2002/02/22 21:33:04 dawes Exp $ */ #ifndef I810TRIS_INC #define I810TRIS_INC diff --git a/src/mesa/drivers/dri/i810/i810vb.c b/src/mesa/drivers/dri/i810/i810vb.c index 5ce98a991d..3439192b0d 100644 --- a/src/mesa/drivers/dri/i810/i810vb.c +++ b/src/mesa/drivers/dri/i810/i810vb.c @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810vb.c,v 1.13 2003/03/26 20:43:48 tsi Exp $ */ #include "glheader.h" diff --git a/src/mesa/drivers/dri/i810/i810vb.h b/src/mesa/drivers/dri/i810/i810vb.h index 1cced86ab2..55d0d2409e 100644 --- a/src/mesa/drivers/dri/i810/i810vb.h +++ b/src/mesa/drivers/dri/i810/i810vb.h @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810vb.h,v 1.4 2002/02/22 21:33:04 dawes Exp $ */ #ifndef I810VB_INC #define I810VB_INC diff --git a/src/mesa/drivers/dri/i810/server/i810_common.h b/src/mesa/drivers/dri/i810/server/i810_common.h index 02e548be0e..29be444b45 100644 --- a/src/mesa/drivers/dri/i810/server/i810_common.h +++ b/src/mesa/drivers/dri/i810/server/i810_common.h @@ -25,7 +25,6 @@ * Converted to common header format: * Jens Owen * - * $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i810_common.h,v 1.1 2002/09/11 00:29:31 dawes Exp $ * */ diff --git a/src/mesa/drivers/dri/i810/server/i810_dri.h b/src/mesa/drivers/dri/i810/server/i810_dri.h index 408a4ebb4d..4a714f0306 100644 --- a/src/mesa/drivers/dri/i810/server/i810_dri.h +++ b/src/mesa/drivers/dri/i810/server/i810_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i810_dri.h,v 1.10 2002/12/10 01:27:04 dawes Exp $ */ #ifndef _I810_DRI_ #define _I810_DRI_ diff --git a/src/mesa/drivers/dri/i810/server/i810_reg.h b/src/mesa/drivers/dri/i810/server/i810_reg.h index c935982a78..e7e5081038 100644 --- a/src/mesa/drivers/dri/i810/server/i810_reg.h +++ b/src/mesa/drivers/dri/i810/server/i810_reg.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i810_reg.h,v 1.13 2003/02/06 04:18:04 dawes Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/i915/server/i830_common.h b/src/mesa/drivers/dri/i915/server/i830_common.h index fb6ceaa52d..2b0fee82a8 100644 --- a/src/mesa/drivers/dri/i915/server/i830_common.h +++ b/src/mesa/drivers/dri/i915/server/i830_common.h @@ -26,7 +26,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ #ifndef _I830_COMMON_H_ #define _I830_COMMON_H_ diff --git a/src/mesa/drivers/dri/i915/server/i830_dri.h b/src/mesa/drivers/dri/i915/server/i830_dri.h index 6c9a709021..313eb759b0 100644 --- a/src/mesa/drivers/dri/i915/server/i830_dri.h +++ b/src/mesa/drivers/dri/i915/server/i830_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ #ifndef _I830_DRI_H #define _I830_DRI_H diff --git a/src/mesa/drivers/dri/i965/server/i830_common.h b/src/mesa/drivers/dri/i965/server/i830_common.h index f320378c2a..49eb145f8b 100644 --- a/src/mesa/drivers/dri/i965/server/i830_common.h +++ b/src/mesa/drivers/dri/i965/server/i830_common.h @@ -26,7 +26,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ #ifndef _I830_COMMON_H_ #define _I830_COMMON_H_ diff --git a/src/mesa/drivers/dri/i965/server/i830_dri.h b/src/mesa/drivers/dri/i965/server/i830_dri.h index 22951812ad..68213f69f5 100644 --- a/src/mesa/drivers/dri/i965/server/i830_dri.h +++ b/src/mesa/drivers/dri/i965/server/i830_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ #ifndef _I830_DRI_H #define _I830_DRI_H diff --git a/src/mesa/drivers/dri/mach64/mach64_context.c b/src/mesa/drivers/dri/mach64/mach64_context.c index ad661e198c..7f558e92bc 100644 --- a/src/mesa/drivers/dri/mach64/mach64_context.c +++ b/src/mesa/drivers/dri/mach64/mach64_context.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_context.h b/src/mesa/drivers/dri/mach64/mach64_context.h index 8d89452412..e925f18c11 100644 --- a/src/mesa/drivers/dri/mach64/mach64_context.h +++ b/src/mesa/drivers/dri/mach64/mach64_context.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_dd.c b/src/mesa/drivers/dri/mach64/mach64_dd.c index 17e8d74d9f..7d225ebc88 100644 --- a/src/mesa/drivers/dri/mach64/mach64_dd.c +++ b/src/mesa/drivers/dri/mach64/mach64_dd.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_dd.h b/src/mesa/drivers/dri/mach64/mach64_dd.h index 74cf1d304f..0a2ce06412 100644 --- a/src/mesa/drivers/dri/mach64/mach64_dd.h +++ b/src/mesa/drivers/dri/mach64/mach64_dd.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_ioctl.c b/src/mesa/drivers/dri/mach64/mach64_ioctl.c index 36e7d3c5d3..6bc2b58ce9 100644 --- a/src/mesa/drivers/dri/mach64/mach64_ioctl.c +++ b/src/mesa/drivers/dri/mach64/mach64_ioctl.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_ioctl.h b/src/mesa/drivers/dri/mach64/mach64_ioctl.h index 52fe863484..2153ab80d7 100644 --- a/src/mesa/drivers/dri/mach64/mach64_ioctl.h +++ b/src/mesa/drivers/dri/mach64/mach64_ioctl.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_lock.c b/src/mesa/drivers/dri/mach64/mach64_lock.c index b73e350111..ea605fb061 100644 --- a/src/mesa/drivers/dri/mach64/mach64_lock.c +++ b/src/mesa/drivers/dri/mach64/mach64_lock.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_lock.h b/src/mesa/drivers/dri/mach64/mach64_lock.h index 973880ee25..3130b183e3 100644 --- a/src/mesa/drivers/dri/mach64/mach64_lock.h +++ b/src/mesa/drivers/dri/mach64/mach64_lock.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vb.c b/src/mesa/drivers/dri/mach64/mach64_native_vb.c index 248fa2a9a2..99f1a14e17 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vb.c +++ b/src/mesa/drivers/dri/mach64/mach64_native_vb.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Mesa 3-D graphics library * Version: 3.5 diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h index f64b808ee7..684f2acc89 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Mesa 3-D graphics library * Version: 3.5 diff --git a/src/mesa/drivers/dri/mach64/mach64_reg.h b/src/mesa/drivers/dri/mach64/mach64_reg.h index abbba295a5..cb944e1023 100644 --- a/src/mesa/drivers/dri/mach64/mach64_reg.h +++ b/src/mesa/drivers/dri/mach64/mach64_reg.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_screen.c b/src/mesa/drivers/dri/mach64/mach64_screen.c index 4e9e216e7d..b780ba65ea 100644 --- a/src/mesa/drivers/dri/mach64/mach64_screen.c +++ b/src/mesa/drivers/dri/mach64/mach64_screen.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_screen.h b/src/mesa/drivers/dri/mach64/mach64_screen.h index 5305058e2f..7bf7dc474d 100644 --- a/src/mesa/drivers/dri/mach64/mach64_screen.h +++ b/src/mesa/drivers/dri/mach64/mach64_screen.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_span.c b/src/mesa/drivers/dri/mach64/mach64_span.c index 3830a28165..5c2403f587 100644 --- a/src/mesa/drivers/dri/mach64/mach64_span.c +++ b/src/mesa/drivers/dri/mach64/mach64_span.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_span.h b/src/mesa/drivers/dri/mach64/mach64_span.h index 0f4c766477..65141d05c3 100644 --- a/src/mesa/drivers/dri/mach64/mach64_span.h +++ b/src/mesa/drivers/dri/mach64/mach64_span.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_state.c b/src/mesa/drivers/dri/mach64/mach64_state.c index 667a394520..9ac51ee5b1 100644 --- a/src/mesa/drivers/dri/mach64/mach64_state.c +++ b/src/mesa/drivers/dri/mach64/mach64_state.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_state.h b/src/mesa/drivers/dri/mach64/mach64_state.h index 95bcab3653..23081cb2fe 100644 --- a/src/mesa/drivers/dri/mach64/mach64_state.h +++ b/src/mesa/drivers/dri/mach64/mach64_state.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_tex.c b/src/mesa/drivers/dri/mach64/mach64_tex.c index 5288d321ce..c42588e064 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tex.c +++ b/src/mesa/drivers/dri/mach64/mach64_tex.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_tex.h b/src/mesa/drivers/dri/mach64/mach64_tex.h index f6cf1cf802..e67661b970 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tex.h +++ b/src/mesa/drivers/dri/mach64/mach64_tex.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_texmem.c b/src/mesa/drivers/dri/mach64/mach64_texmem.c index 3b7b93b984..d65b2cda6a 100644 --- a/src/mesa/drivers/dri/mach64/mach64_texmem.c +++ b/src/mesa/drivers/dri/mach64/mach64_texmem.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., * Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/mach64/mach64_texstate.c b/src/mesa/drivers/dri/mach64/mach64_texstate.c index 3ace370d70..80c84d6774 100644 --- a/src/mesa/drivers/dri/mach64/mach64_texstate.c +++ b/src/mesa/drivers/dri/mach64/mach64_texstate.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_tris.c b/src/mesa/drivers/dri/mach64/mach64_tris.c index 369f610442..e4df01106d 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tris.c +++ b/src/mesa/drivers/dri/mach64/mach64_tris.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_tris.h b/src/mesa/drivers/dri/mach64/mach64_tris.h index 208703289d..4780765a18 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tris.h +++ b/src/mesa/drivers/dri/mach64/mach64_tris.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_vb.c b/src/mesa/drivers/dri/mach64/mach64_vb.c index 83a5f73e6b..8aab72a3f3 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vb.c +++ b/src/mesa/drivers/dri/mach64/mach64_vb.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_vb.h b/src/mesa/drivers/dri/mach64/mach64_vb.h index bcc4759af3..0d923abce0 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vb.h +++ b/src/mesa/drivers/dri/mach64/mach64_vb.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h index c1207cacd1..938804af9e 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Mesa 3-D graphics library * Version: 3.5 diff --git a/src/mesa/drivers/dri/mach64/server/mach64_dri.h b/src/mesa/drivers/dri/mach64/server/mach64_dri.h index 139668e3f3..1477443f79 100644 --- a/src/mesa/drivers/dri/mach64/server/mach64_dri.h +++ b/src/mesa/drivers/dri/mach64/server/mach64_dri.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mga_texstate.c b/src/mesa/drivers/dri/mga/mga_texstate.c index 71d264b0f1..c14ddc95c9 100644 --- a/src/mesa/drivers/dri/mga/mga_texstate.c +++ b/src/mesa/drivers/dri/mga/mga_texstate.c @@ -26,7 +26,6 @@ * Ian Romanick * Keith Whitwell */ -/* $XFree86:$ */ #include #include "mm.h" diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.c b/src/mesa/drivers/dri/mga/mga_xmesa.c index f4e651afa0..6148f6b488 100644 --- a/src/mesa/drivers/dri/mga/mga_xmesa.c +++ b/src/mesa/drivers/dri/mga/mga_xmesa.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mga_xmesa.c,v 1.19 2003/03/26 20:43:49 tsi Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.h b/src/mesa/drivers/dri/mga/mga_xmesa.h index 0ab0c63f78..0f81c9cbec 100644 --- a/src/mesa/drivers/dri/mga/mga_xmesa.h +++ b/src/mesa/drivers/dri/mga/mga_xmesa.h @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mga_xmesa.h,v 1.12 2002/12/16 16:18:52 dawes Exp $ */ #ifndef _MGA_INIT_H_ #define _MGA_INIT_H_ diff --git a/src/mesa/drivers/dri/mga/mgacontext.h b/src/mesa/drivers/dri/mga/mgacontext.h index 2124006ade..6aa92355b8 100644 --- a/src/mesa/drivers/dri/mga/mgacontext.h +++ b/src/mesa/drivers/dri/mga/mgacontext.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgacontext.h,v 1.7 2002/12/16 16:18:52 dawes Exp $*/ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgadd.c b/src/mesa/drivers/dri/mga/mgadd.c index b1d5e0c48f..04336b5ac7 100644 --- a/src/mesa/drivers/dri/mga/mgadd.c +++ b/src/mesa/drivers/dri/mga/mgadd.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgadd.c,v 1.14 2002/10/30 12:51:35 alanh Exp $ */ #include "mtypes.h" diff --git a/src/mesa/drivers/dri/mga/mgadd.h b/src/mesa/drivers/dri/mga/mgadd.h index f98bfdc878..6830ca67ad 100644 --- a/src/mesa/drivers/dri/mga/mgadd.h +++ b/src/mesa/drivers/dri/mga/mgadd.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgadd.h,v 1.3 2002/10/30 12:51:35 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgaioctl.h b/src/mesa/drivers/dri/mga/mgaioctl.h index f3ae749ca9..9aa08c5158 100644 --- a/src/mesa/drivers/dri/mga/mgaioctl.h +++ b/src/mesa/drivers/dri/mga/mgaioctl.h @@ -25,7 +25,6 @@ * Keith Whitwell * Gareth Hughes */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgaioctl.h,v 1.11 2002/10/30 12:51:36 alanh Exp $ */ #ifndef MGA_IOCTL_H #define MGA_IOCTL_H diff --git a/src/mesa/drivers/dri/mga/mgapixel.c b/src/mesa/drivers/dri/mga/mgapixel.c index 2b9da8c181..f309aabbc8 100644 --- a/src/mesa/drivers/dri/mga/mgapixel.c +++ b/src/mesa/drivers/dri/mga/mgapixel.c @@ -34,7 +34,6 @@ * \author Keith Whitwell * \author Gareth Hughes */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgapixel.c,v 1.9 2002/11/05 17:46:08 tsi Exp $ */ #include "mtypes.h" #include "macros.h" diff --git a/src/mesa/drivers/dri/mga/mgapixel.h b/src/mesa/drivers/dri/mga/mgapixel.h index c44fd769a8..b52c8670f3 100644 --- a/src/mesa/drivers/dri/mga/mgapixel.h +++ b/src/mesa/drivers/dri/mga/mgapixel.h @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgapixel.h,v 1.5 2002/10/30 12:51:36 alanh Exp $ */ #ifndef MGA_PIXELS_H #define MGA_PIXELS_H diff --git a/src/mesa/drivers/dri/mga/mgaregs.h b/src/mesa/drivers/dri/mga/mgaregs.h index e1291ca01b..1ef1e6d24c 100644 --- a/src/mesa/drivers/dri/mga/mgaregs.h +++ b/src/mesa/drivers/dri/mga/mgaregs.h @@ -19,7 +19,6 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgaregs.h,v 1.6 2003/01/12 03:55:46 tsi Exp $ */ #ifndef _MGAREGS_H_ #define _MGAREGS_H_ diff --git a/src/mesa/drivers/dri/mga/mgarender.c b/src/mesa/drivers/dri/mga/mgarender.c index 3080cea79f..c9e42a8040 100644 --- a/src/mesa/drivers/dri/mga/mgarender.c +++ b/src/mesa/drivers/dri/mga/mgarender.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgarender.c,v 1.4 2002/10/30 12:51:36 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/mga/mgaspan.h b/src/mesa/drivers/dri/mga/mgaspan.h index f133a51c08..f5e2e49b8a 100644 --- a/src/mesa/drivers/dri/mga/mgaspan.h +++ b/src/mesa/drivers/dri/mga/mgaspan.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgaspan.h,v 1.3 2002/10/30 12:51:36 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgastate.h b/src/mesa/drivers/dri/mga/mgastate.h index afbe0aaf90..ec65d4e6cd 100644 --- a/src/mesa/drivers/dri/mga/mgastate.h +++ b/src/mesa/drivers/dri/mga/mgastate.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgastate.h,v 1.5 2002/10/30 12:51:36 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgatex.c b/src/mesa/drivers/dri/mga/mgatex.c index a7d74317a5..31ea5046df 100644 --- a/src/mesa/drivers/dri/mga/mgatex.c +++ b/src/mesa/drivers/dri/mga/mgatex.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatex.c,v 1.14 2002/10/30 12:51:36 alanh Exp $ */ #include "glheader.h" #include "mm.h" diff --git a/src/mesa/drivers/dri/mga/mgatex.h b/src/mesa/drivers/dri/mga/mgatex.h index fb7ffcff16..789034964a 100644 --- a/src/mesa/drivers/dri/mga/mgatex.h +++ b/src/mesa/drivers/dri/mga/mgatex.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatex.h,v 1.7 2002/10/30 12:51:36 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgatexmem.c b/src/mesa/drivers/dri/mga/mgatexmem.c index 18743331c6..559813f5de 100644 --- a/src/mesa/drivers/dri/mga/mgatexmem.c +++ b/src/mesa/drivers/dri/mga/mgatexmem.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatexmem.c,v 1.7 2002/10/30 12:51:36 alanh Exp $ */ #include "glheader.h" diff --git a/src/mesa/drivers/dri/mga/mgatris.c b/src/mesa/drivers/dri/mga/mgatris.c index 91b413ae76..0c8081cfb9 100644 --- a/src/mesa/drivers/dri/mga/mgatris.c +++ b/src/mesa/drivers/dri/mga/mgatris.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatris.c,v 1.10 2002/10/30 12:51:36 alanh Exp $ */ #include "mtypes.h" #include "macros.h" diff --git a/src/mesa/drivers/dri/mga/mgatris.h b/src/mesa/drivers/dri/mga/mgatris.h index f3ece3a053..a40fef8307 100644 --- a/src/mesa/drivers/dri/mga/mgatris.h +++ b/src/mesa/drivers/dri/mga/mgatris.h @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatris.h,v 1.10 2002/10/30 12:51:36 alanh Exp $ */ #ifndef MGATRIS_INC #define MGATRIS_INC diff --git a/src/mesa/drivers/dri/mga/mgavb.c b/src/mesa/drivers/dri/mga/mgavb.c index 902d8bd1c1..954fd53ae3 100644 --- a/src/mesa/drivers/dri/mga/mgavb.c +++ b/src/mesa/drivers/dri/mga/mgavb.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgavb.c,v 1.15 2003/03/26 20:43:49 tsi Exp $ */ #include #include "mgacontext.h" diff --git a/src/mesa/drivers/dri/mga/mgavb.h b/src/mesa/drivers/dri/mga/mgavb.h index 5f6454aca9..f6580e0db9 100644 --- a/src/mesa/drivers/dri/mga/mgavb.h +++ b/src/mesa/drivers/dri/mga/mgavb.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgavb.h,v 1.8 2002/10/30 12:51:36 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/server/mga.h b/src/mesa/drivers/dri/mga/server/mga.h index 830d48d859..d7790e4779 100644 --- a/src/mesa/drivers/dri/mga/server/mga.h +++ b/src/mesa/drivers/dri/mga/server/mga.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga.h,v 1.85 2002/12/16 16:19:17 dawes Exp $ */ /* * MGA Millennium (MGA2064W) functions * diff --git a/src/mesa/drivers/dri/mga/server/mga_bios.h b/src/mesa/drivers/dri/mga/server/mga_bios.h index 8fbf619e34..5dcfc1614d 100644 --- a/src/mesa/drivers/dri/mga/server/mga_bios.h +++ b/src/mesa/drivers/dri/mga/server/mga_bios.h @@ -1,8 +1,6 @@ -/* $XConsortium: mga_bios.h /main/2 1996/10/28 04:48:23 kaleb $ */ #ifndef MGA_BIOS_H #define MGA_BIOS_H -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_bios.h,v 1.3 1998/07/25 16:55:51 dawes Exp $ */ /* * MGABiosInfo - This struct describes the video BIOS info block. diff --git a/src/mesa/drivers/dri/mga/server/mga_dri.c b/src/mesa/drivers/dri/mga/server/mga_dri.c index 258ace83a0..bc575e62ee 100644 --- a/src/mesa/drivers/dri/mga/server/mga_dri.c +++ b/src/mesa/drivers/dri/mga/server/mga_dri.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_dri.c,v 1.28 2003/02/08 21:26:58 dawes Exp $ */ /* * Copyright 2000 VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/mga/server/mga_dri.h b/src/mesa/drivers/dri/mga/server/mga_dri.h index 03b8414603..1ce07028f1 100644 --- a/src/mesa/drivers/dri/mga/server/mga_dri.h +++ b/src/mesa/drivers/dri/mga/server/mga_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_dri.h,v 1.8 2002/11/29 11:06:42 eich Exp $ */ /* * Copyright 2000 VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/mga/server/mga_macros.h b/src/mesa/drivers/dri/mga/server/mga_macros.h index d985081ab6..189e1415d0 100644 --- a/src/mesa/drivers/dri/mga/server/mga_macros.h +++ b/src/mesa/drivers/dri/mga/server/mga_macros.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_macros.h,v 1.22 2002/02/20 17:17:50 dawes Exp $ */ #ifndef _MGA_MACROS_H_ #define _MGA_MACROS_H_ diff --git a/src/mesa/drivers/dri/mga/server/mga_reg.h b/src/mesa/drivers/dri/mga/server/mga_reg.h index b8e3499235..d51366d44e 100644 --- a/src/mesa/drivers/dri/mga/server/mga_reg.h +++ b/src/mesa/drivers/dri/mga/server/mga_reg.h @@ -1,8 +1,6 @@ -/* $XConsortium: mgareg.h /main/2 1996/10/25 10:33:21 kaleb $ */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_reg.h,v 1.18 2001/09/26 12:59:18 alanh Exp $ */ diff --git a/src/mesa/drivers/dri/r128/r128_context.c b/src/mesa/drivers/dri/r128/r128_context.c index 95e54a6af5..dfc89a2da7 100644 --- a/src/mesa/drivers/dri/r128/r128_context.c +++ b/src/mesa/drivers/dri/r128/r128_context.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_context.c,v 1.8 2002/10/30 12:51:38 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_context.h b/src/mesa/drivers/dri/r128/r128_context.h index c51dd7fa58..3f96836df1 100644 --- a/src/mesa/drivers/dri/r128/r128_context.h +++ b/src/mesa/drivers/dri/r128/r128_context.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_context.h,v 1.12 2002/12/16 16:18:52 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_dd.c b/src/mesa/drivers/dri/r128/r128_dd.c index 54f2b21b5d..d8e1c70ab7 100644 --- a/src/mesa/drivers/dri/r128/r128_dd.c +++ b/src/mesa/drivers/dri/r128/r128_dd.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_dd.c,v 1.15 2002/10/30 12:51:38 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_dd.h b/src/mesa/drivers/dri/r128/r128_dd.h index 7a0abb73f8..ce038853c4 100644 --- a/src/mesa/drivers/dri/r128/r128_dd.h +++ b/src/mesa/drivers/dri/r128/r128_dd.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_dd.h,v 1.3 2001/01/08 01:07:20 martin Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_ioctl.c b/src/mesa/drivers/dri/r128/r128_ioctl.c index b0dba7d04e..25188061a0 100644 --- a/src/mesa/drivers/dri/r128/r128_ioctl.c +++ b/src/mesa/drivers/dri/r128/r128_ioctl.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_ioctl.c,v 1.10 2002/12/16 16:18:53 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_ioctl.h b/src/mesa/drivers/dri/r128/r128_ioctl.h index 95779f09be..57063c41f5 100644 --- a/src/mesa/drivers/dri/r128/r128_ioctl.h +++ b/src/mesa/drivers/dri/r128/r128_ioctl.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_ioctl.h,v 1.6 2002/12/16 16:18:53 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_lock.c b/src/mesa/drivers/dri/r128/r128_lock.c index ea23b007f3..3478e12ad0 100644 --- a/src/mesa/drivers/dri/r128/r128_lock.c +++ b/src/mesa/drivers/dri/r128/r128_lock.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_lock.c,v 1.5 2002/10/30 12:51:38 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_lock.h b/src/mesa/drivers/dri/r128/r128_lock.h index 39bdde9820..1fc8cbe29f 100644 --- a/src/mesa/drivers/dri/r128/r128_lock.h +++ b/src/mesa/drivers/dri/r128/r128_lock.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_lock.h,v 1.4 2001/01/08 01:07:21 martin Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_screen.c b/src/mesa/drivers/dri/r128/r128_screen.c index 880dee85c2..0722b80ee5 100644 --- a/src/mesa/drivers/dri/r128/r128_screen.c +++ b/src/mesa/drivers/dri/r128/r128_screen.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_screen.c,v 1.9 2003/03/26 20:43:49 tsi Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_screen.h b/src/mesa/drivers/dri/r128/r128_screen.h index 8db8eea358..b31e87661b 100644 --- a/src/mesa/drivers/dri/r128/r128_screen.h +++ b/src/mesa/drivers/dri/r128/r128_screen.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_screen.h,v 1.7 2002/12/16 16:18:53 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_span.c b/src/mesa/drivers/dri/r128/r128_span.c index 85798c1601..c5b6480db9 100644 --- a/src/mesa/drivers/dri/r128/r128_span.c +++ b/src/mesa/drivers/dri/r128/r128_span.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_span.c,v 1.8 2002/10/30 12:51:39 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_span.h b/src/mesa/drivers/dri/r128/r128_span.h index fd7c2d1394..9af4058129 100644 --- a/src/mesa/drivers/dri/r128/r128_span.h +++ b/src/mesa/drivers/dri/r128/r128_span.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_span.h,v 1.3 2001/01/08 01:07:21 martin Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_state.c b/src/mesa/drivers/dri/r128/r128_state.c index e476afa5d8..58c3a27ee8 100644 --- a/src/mesa/drivers/dri/r128/r128_state.c +++ b/src/mesa/drivers/dri/r128/r128_state.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_state.c,v 1.11 2002/10/30 12:51:39 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_state.h b/src/mesa/drivers/dri/r128/r128_state.h index 6f0a6a6557..a44327dfb3 100644 --- a/src/mesa/drivers/dri/r128/r128_state.h +++ b/src/mesa/drivers/dri/r128/r128_state.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_state.h,v 1.3 2001/01/08 01:07:21 martin Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_tex.c b/src/mesa/drivers/dri/r128/r128_tex.c index 3b2d017c1f..554a92287f 100644 --- a/src/mesa/drivers/dri/r128/r128_tex.c +++ b/src/mesa/drivers/dri/r128/r128_tex.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tex.c,v 1.14 2002/11/05 17:46:08 tsi Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_tex.h b/src/mesa/drivers/dri/r128/r128_tex.h index 54053b8b31..994dffb5a9 100644 --- a/src/mesa/drivers/dri/r128/r128_tex.h +++ b/src/mesa/drivers/dri/r128/r128_tex.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tex.h,v 1.7 2002/02/22 21:44:58 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_texmem.c b/src/mesa/drivers/dri/r128/r128_texmem.c index d011a75671..a7d0280636 100644 --- a/src/mesa/drivers/dri/r128/r128_texmem.c +++ b/src/mesa/drivers/dri/r128/r128_texmem.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_texmem.c,v 1.1 2002/02/22 21:44:58 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_texobj.h b/src/mesa/drivers/dri/r128/r128_texobj.h index 282e887149..08eac87758 100644 --- a/src/mesa/drivers/dri/r128/r128_texobj.h +++ b/src/mesa/drivers/dri/r128/r128_texobj.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_texobj.h,v 1.5 2002/02/22 21:44:58 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_texstate.c b/src/mesa/drivers/dri/r128/r128_texstate.c index 6b43f21cd4..211b9ea2a9 100644 --- a/src/mesa/drivers/dri/r128/r128_texstate.c +++ b/src/mesa/drivers/dri/r128/r128_texstate.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_texstate.c,v 1.1 2002/02/22 21:44:58 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_tris.c b/src/mesa/drivers/dri/r128/r128_tris.c index f406e928c5..f2f124360c 100644 --- a/src/mesa/drivers/dri/r128/r128_tris.c +++ b/src/mesa/drivers/dri/r128/r128_tris.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tris.c,v 1.8 2002/10/30 12:51:43 alanh Exp $ */ /* -*- c-basic-offset: 3 -*- */ +/* -*- c-basic-offset: 3 -*- */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/r128/r128_tris.h b/src/mesa/drivers/dri/r128/r128_tris.h index 755d3320b0..c8f0a4809b 100644 --- a/src/mesa/drivers/dri/r128/r128_tris.h +++ b/src/mesa/drivers/dri/r128/r128_tris.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tris.h,v 1.8 2002/10/30 12:51:43 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/r128/server/r128.h b/src/mesa/drivers/dri/r128/server/r128.h index ce98b1b915..ca08d7c86a 100644 --- a/src/mesa/drivers/dri/r128/server/r128.h +++ b/src/mesa/drivers/dri/r128/server/r128.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128.h,v 1.24 2002/12/16 16:19:10 dawes Exp $ */ /* * Copyright 1999, 2000 ATI Technologies Inc., Markham, Ontario, * Precision Insight, Inc., Cedar Park, Texas, and diff --git a/src/mesa/drivers/dri/r128/server/r128_dri.c b/src/mesa/drivers/dri/r128/server/r128_dri.c index 5edf1e1003..efe9232dc2 100644 --- a/src/mesa/drivers/dri/r128/server/r128_dri.c +++ b/src/mesa/drivers/dri/r128/server/r128_dri.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128_dri.c,v 1.28 2003/02/07 20:41:14 martin Exp $ */ /* * Copyright 1999, 2000 ATI Technologies Inc., Markham, Ontario, * Precision Insight, Inc., Cedar Park, Texas, and diff --git a/src/mesa/drivers/dri/r128/server/r128_dri.h b/src/mesa/drivers/dri/r128/server/r128_dri.h index 67ade70de4..430e5f580b 100644 --- a/src/mesa/drivers/dri/r128/server/r128_dri.h +++ b/src/mesa/drivers/dri/r128/server/r128_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128_dri.h,v 1.7 2002/10/30 12:52:12 alanh Exp $ */ /* * Copyright 1999, 2000 ATI Technologies Inc., Markham, Ontario, * Precision Insight, Inc., Cedar Park, Texas, and diff --git a/src/mesa/drivers/dri/r128/server/r128_macros.h b/src/mesa/drivers/dri/r128/server/r128_macros.h index 93b7feb02c..f7b945da93 100644 --- a/src/mesa/drivers/dri/r128/server/r128_macros.h +++ b/src/mesa/drivers/dri/r128/server/r128_macros.h @@ -35,7 +35,6 @@ * DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/R128_reg.h,v 1.20 2002/10/12 01:38:07 martin Exp $ */ #ifndef _R128_MACROS_H_ #define _R128_MACROS_H_ diff --git a/src/mesa/drivers/dri/r128/server/r128_reg.h b/src/mesa/drivers/dri/r128/server/r128_reg.h index 5669452d74..50033540b9 100644 --- a/src/mesa/drivers/dri/r128/server/r128_reg.h +++ b/src/mesa/drivers/dri/r128/server/r128_reg.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128_reg.h,v 1.15 2002/12/16 16:19:11 dawes Exp $ */ /* * Copyright 1999, 2000 ATI Technologies Inc., Markham, Ontario, * Precision Insight, Inc., Cedar Park, Texas, and diff --git a/src/mesa/drivers/dri/r128/server/r128_version.h b/src/mesa/drivers/dri/r128/server/r128_version.h index 589d8d40bc..783711ef97 100644 --- a/src/mesa/drivers/dri/r128/server/r128_version.h +++ b/src/mesa/drivers/dri/r128/server/r128_version.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128_version.h,v 1.6 2003/01/01 19:16:35 tsi Exp $ */ /* * Copyright 2000 through 2003 by Marc Aurele La France (TSI @ UQV), tsi@xfree86.org * diff --git a/src/mesa/drivers/dri/radeon/radeon_compat.c b/src/mesa/drivers/dri/radeon/radeon_compat.c index 1cbe3407ba..bd467fb15b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_compat.c +++ b/src/mesa/drivers/dri/radeon/radeon_compat.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2002 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c index 9451ec4aa5..ba93a054ae 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_context.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_context.c,v 1.9 2003/09/24 02:43:12 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index 4c64bc201a..f7e461239e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_ioctl.c,v 1.11 2003/01/29 22:04:59 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.h b/src/mesa/drivers/dri/radeon/radeon_ioctl.h index 11a7d02b1b..020a5c21e2 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.h +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_ioctl.h,v 1.6 2002/12/16 16:18:58 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_lighting.c b/src/mesa/drivers/dri/radeon/radeon_lighting.c index 44e00af0ef..5e9b9c3051 100644 --- a/src/mesa/drivers/dri/radeon/radeon_lighting.c +++ b/src/mesa/drivers/dri/radeon/radeon_lighting.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_state.c,v 1.5 2002/09/16 18:05:20 eich Exp $ */ /* * Copyright 2000, 2001 VA Linux Systems Inc., Fremont, California. * diff --git a/src/mesa/drivers/dri/radeon/radeon_maos.h b/src/mesa/drivers/dri/radeon/radeon_maos.h index 09039d6840..b8935e84a0 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos.h +++ b/src/mesa/drivers/dri/radeon/radeon_maos.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_maos.h,v 1.1 2002/10/30 12:51:55 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index 49118b5e37..b61f5e0f3e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_maos_arrays.c,v 1.1 2002/10/30 12:51:55 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 65dbecf7a6..d5ceedfa24 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_maos_verts.c,v 1.1 2002/10/30 12:51:55 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_sanity.c b/src/mesa/drivers/dri/radeon/radeon_sanity.c index 557057784c..bdfb7240d7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_sanity.c +++ b/src/mesa/drivers/dri/radeon/radeon_sanity.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_sanity.c,v 1.1 2002/10/30 12:51:55 alanh Exp $ */ /************************************************************************** Copyright 2002 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index aa7fb633dd..4a45948608 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_screen.c,v 1.7 2003/03/26 20:43:51 tsi Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.h b/src/mesa/drivers/dri/radeon/radeon_screen.h index 25e6fcf399..f8c0cc96df 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.h +++ b/src/mesa/drivers/dri/radeon/radeon_screen.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_screen.h,v 1.5 2002/12/16 16:18:58 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c index 4de05c7697..856d27df75 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.c +++ b/src/mesa/drivers/dri/radeon/radeon_state.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_state.c,v 1.8 2002/12/16 16:18:58 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/radeon/radeon_state.h b/src/mesa/drivers/dri/radeon/radeon_state.h index ad7db3b677..2171879f75 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.h +++ b/src/mesa/drivers/dri/radeon/radeon_state.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_state.h,v 1.5 2002/11/05 17:46:09 tsi Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index 5fc34f0933..c876a596e6 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_state_init.c,v 1.3 2003/02/22 06:21:11 dawes Exp $ */ /* * Copyright 2000, 2001 VA Linux Systems Inc., Fremont, California. * diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 7ce1fa67cf..2b3ae14ff7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_swtcl.c,v 1.6 2003/05/06 23:52:08 daenzer Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.h b/src/mesa/drivers/dri/radeon/radeon_swtcl.h index 64f9019513..1feedf185d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.h +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.h @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index 5ad044c262..d35be1ca88 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.h b/src/mesa/drivers/dri/radeon/radeon_tcl.h index 168ab958a2..dccbea5fdb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.h +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_tcl.h,v 1.2 2003/02/08 21:26:45 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c index edaea6c209..f3eb9d8eef 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.c +++ b/src/mesa/drivers/dri/radeon/radeon_tex.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_tex.c,v 1.6 2002/09/16 18:05:20 eich Exp $ */ /* Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.h b/src/mesa/drivers/dri/radeon/radeon_tex.h index a806981ae6..bdf086dfee 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.h +++ b/src/mesa/drivers/dri/radeon/radeon_tex.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_tex.h,v 1.3 2002/02/22 21:45:01 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_texmem.c b/src/mesa/drivers/dri/radeon/radeon_texmem.c index 20f25dd34b..f7520f1dea 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texmem.c +++ b/src/mesa/drivers/dri/radeon/radeon_texmem.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_texmem.c,v 1.7 2002/12/16 16:18:59 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c index 37bb749223..ae8d527cf4 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texstate.c +++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_texstate.c,v 1.6 2002/12/16 16:18:59 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/server/radeon.h b/src/mesa/drivers/dri/radeon/server/radeon.h index 6f6c2e6d25..3fb1e37c53 100644 --- a/src/mesa/drivers/dri/radeon/server/radeon.h +++ b/src/mesa/drivers/dri/radeon/server/radeon.h @@ -31,7 +31,6 @@ * DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/radeon.h,v 1.29 2002/10/12 01:38:07 martin Exp $ */ #ifndef _RADEON_H_ #define _RADEON_H_ diff --git a/src/mesa/drivers/dri/radeon/server/radeon_dri.h b/src/mesa/drivers/dri/radeon/server/radeon_dri.h index ecd5323339..dc51372107 100644 --- a/src/mesa/drivers/dri/radeon/server/radeon_dri.h +++ b/src/mesa/drivers/dri/radeon/server/radeon_dri.h @@ -34,7 +34,6 @@ * DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/radeon_dri.h,v 1.3 2002/04/24 16:20:40 martin Exp $ */ #ifndef _RADEON_DRI_ #define _RADEON_DRI_ diff --git a/src/mesa/drivers/dri/radeon/server/radeon_macros.h b/src/mesa/drivers/dri/radeon/server/radeon_macros.h index 60f0fa2d35..355262c9ba 100644 --- a/src/mesa/drivers/dri/radeon/server/radeon_macros.h +++ b/src/mesa/drivers/dri/radeon/server/radeon_macros.h @@ -35,7 +35,6 @@ * DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/radeon_reg.h,v 1.20 2002/10/12 01:38:07 martin Exp $ */ #ifndef _RADEON_MACROS_H_ #define _RADEON_MACROS_H_ diff --git a/src/mesa/drivers/dri/radeon/server/radeon_reg.h b/src/mesa/drivers/dri/radeon/server/radeon_reg.h index 4dcce63846..596a8aa715 100644 --- a/src/mesa/drivers/dri/radeon/server/radeon_reg.h +++ b/src/mesa/drivers/dri/radeon/server/radeon_reg.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/radeon_reg.h,v 1.30 2003/10/07 22:47:12 martin Exp $ */ /* * Copyright 2000 ATI Technologies Inc., Markham, Ontario, and * VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/savage/savagetris.c b/src/mesa/drivers/dri/savage/savagetris.c index 4ce2f60b4f..52c7f5fa76 100644 --- a/src/mesa/drivers/dri/savage/savagetris.c +++ b/src/mesa/drivers/dri/savage/savagetris.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- c-basic-offset: 3 -*- */ +/* -*- c-basic-offset: 3 -*- */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/savage/savagetris.h b/src/mesa/drivers/dri/savage/savagetris.h index 00803e7ff3..b2b3d951c6 100644 --- a/src/mesa/drivers/dri/savage/savagetris.h +++ b/src/mesa/drivers/dri/savage/savagetris.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tris.h,v 1.4 2001/01/08 01:07:24 martin Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/sis/server/sis_common.h b/src/mesa/drivers/dri/sis/server/sis_common.h index cbddf0c737..bd9bab846f 100644 --- a/src/mesa/drivers/dri/sis/server/sis_common.h +++ b/src/mesa/drivers/dri/sis/server/sis_common.h @@ -1,4 +1,3 @@ -/* * $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sis/sis_common.h,v 1.1 2003/08/29 08:52:12 twini Exp $ */ /* * Common header definitions for SiS 2D/3D/DRM suite * diff --git a/src/mesa/drivers/dri/sis/server/sis_dri.h b/src/mesa/drivers/dri/sis/server/sis_dri.h index a05662430e..f0171f3c0f 100644 --- a/src/mesa/drivers/dri/sis/server/sis_dri.h +++ b/src/mesa/drivers/dri/sis/server/sis_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sis/sis_dri.h,v 1.9 2003/08/29 08:50:54 twini Exp $ */ /* modified from tdfx_dri.h */ diff --git a/src/mesa/drivers/dri/sis/sis_alloc.c b/src/mesa/drivers/dri/sis/sis_alloc.c index b696eeb51a..4ca4052803 100644 --- a/src/mesa/drivers/dri/sis/sis_alloc.c +++ b/src/mesa/drivers/dri/sis/sis_alloc.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_alloc.c,v 1.7 2001/01/08 01:07:29 martin Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_alloc.h b/src/mesa/drivers/dri/sis/sis_alloc.h index e76fc53fe2..eb784afad9 100644 --- a/src/mesa/drivers/dri/sis/sis_alloc.h +++ b/src/mesa/drivers/dri/sis/sis_alloc.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_clear.c b/src/mesa/drivers/dri/sis/sis_clear.c index fb92d06c73..174f3c0768 100644 --- a/src/mesa/drivers/dri/sis/sis_clear.c +++ b/src/mesa/drivers/dri/sis/sis_clear.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_clear.c,v 1.5 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_context.c b/src/mesa/drivers/dri/sis/sis_context.c index b21df0a61e..04c7464c5e 100644 --- a/src/mesa/drivers/dri/sis/sis_context.c +++ b/src/mesa/drivers/dri/sis/sis_context.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_ctx.c,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_context.h b/src/mesa/drivers/dri/sis/sis_context.h index c349bf96ed..b81812d6ce 100644 --- a/src/mesa/drivers/dri/sis/sis_context.h +++ b/src/mesa/drivers/dri/sis/sis_context.h @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_dd.c b/src/mesa/drivers/dri/sis/sis_dd.c index 8fc7896b87..989c159a80 100644 --- a/src/mesa/drivers/dri/sis/sis_dd.c +++ b/src/mesa/drivers/dri/sis/sis_dd.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_ctx.c,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_dd.h b/src/mesa/drivers/dri/sis/sis_dd.h index da76596e92..b141243a59 100644 --- a/src/mesa/drivers/dri/sis/sis_dd.h +++ b/src/mesa/drivers/dri/sis/sis_dd.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_fog.c b/src/mesa/drivers/dri/sis/sis_fog.c index fe9a3c95d6..ba5ac90851 100644 --- a/src/mesa/drivers/dri/sis/sis_fog.c +++ b/src/mesa/drivers/dri/sis/sis_fog.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_fog.c,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_lock.c b/src/mesa/drivers/dri/sis/sis_lock.c index 70ca8e6cbc..0ea64e3498 100644 --- a/src/mesa/drivers/dri/sis/sis_lock.c +++ b/src/mesa/drivers/dri/sis/sis_lock.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2003 Eric Anholt diff --git a/src/mesa/drivers/dri/sis/sis_lock.h b/src/mesa/drivers/dri/sis/sis_lock.h index fef9931963..54844e9b09 100644 --- a/src/mesa/drivers/dri/sis/sis_lock.h +++ b/src/mesa/drivers/dri/sis/sis_lock.h @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_reg.h b/src/mesa/drivers/dri/sis/sis_reg.h index 78c6660181..e40c4371bf 100644 --- a/src/mesa/drivers/dri/sis/sis_reg.h +++ b/src/mesa/drivers/dri/sis/sis_reg.h @@ -25,7 +25,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_reg.h,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_screen.c b/src/mesa/drivers/dri/sis/sis_screen.c index 89d734ba78..d90482f3d7 100644 --- a/src/mesa/drivers/dri/sis/sis_screen.c +++ b/src/mesa/drivers/dri/sis/sis_screen.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2003 Eric Anholt diff --git a/src/mesa/drivers/dri/sis/sis_screen.h b/src/mesa/drivers/dri/sis/sis_screen.h index d5b2101e98..07c29cfa09 100644 --- a/src/mesa/drivers/dri/sis/sis_screen.h +++ b/src/mesa/drivers/dri/sis/sis_screen.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_span.c b/src/mesa/drivers/dri/sis/sis_span.c index ea6db6781d..dc50bda877 100644 --- a/src/mesa/drivers/dri/sis/sis_span.c +++ b/src/mesa/drivers/dri/sis/sis_span.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_span.c,v 1.5 2001/03/21 16:14:26 dawes Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_span.h b/src/mesa/drivers/dri/sis/sis_span.h index 4b0add2ac2..a1f817c44c 100644 --- a/src/mesa/drivers/dri/sis/sis_span.h +++ b/src/mesa/drivers/dri/sis/sis_span.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_state.c b/src/mesa/drivers/dri/sis/sis_state.c index 33a2f089b8..305c63f73f 100644 --- a/src/mesa/drivers/dri/sis/sis_state.c +++ b/src/mesa/drivers/dri/sis/sis_state.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_ctx.c,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_state.h b/src/mesa/drivers/dri/sis/sis_state.h index 8f7e2acb92..2d0ea9c5fb 100644 --- a/src/mesa/drivers/dri/sis/sis_state.h +++ b/src/mesa/drivers/dri/sis/sis_state.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_stencil.c b/src/mesa/drivers/dri/sis/sis_stencil.c index a1ce2966e8..55c0440eba 100644 --- a/src/mesa/drivers/dri/sis/sis_stencil.c +++ b/src/mesa/drivers/dri/sis/sis_stencil.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_stencil.c,v 1.3 2000/09/26 15:56:49 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_stencil.h b/src/mesa/drivers/dri/sis/sis_stencil.h index 4a36c98f3d..6b556c4378 100644 --- a/src/mesa/drivers/dri/sis/sis_stencil.h +++ b/src/mesa/drivers/dri/sis/sis_stencil.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_tex.c b/src/mesa/drivers/dri/sis/sis_tex.c index be87f16e29..5e10c610f8 100644 --- a/src/mesa/drivers/dri/sis/sis_tex.c +++ b/src/mesa/drivers/dri/sis/sis_tex.c @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_tex.h b/src/mesa/drivers/dri/sis/sis_tex.h index 8ddc7c469e..c499e80e86 100644 --- a/src/mesa/drivers/dri/sis/sis_tex.h +++ b/src/mesa/drivers/dri/sis/sis_tex.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_texstate.c b/src/mesa/drivers/dri/sis/sis_texstate.c index 7ef20f880c..4f813bb81c 100644 --- a/src/mesa/drivers/dri/sis/sis_texstate.c +++ b/src/mesa/drivers/dri/sis/sis_texstate.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_tris.h b/src/mesa/drivers/dri/sis/sis_tris.h index 5e07acc211..499eb4d24d 100644 --- a/src/mesa/drivers/dri/sis/sis_tris.h +++ b/src/mesa/drivers/dri/sis/sis_tris.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tris.h,v 1.8 2002/10/30 12:51:43 alanh Exp $ */ /************************************************************************** Copyright 2003 Eric Anholt diff --git a/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S b/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S index 0f4cc45089..500c97c536 100644 --- a/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S +++ b/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/X86/fx_3dnow_fastpath.S,v 1.2 2000/09/26 15:56:51 tsi Exp $ */ #include "../../X86/assyntax.h" diff --git a/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h b/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h index 9ec4935d78..78c5fef746 100644 --- a/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h +++ b/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/X86/fx_3dnow_fasttmp.h,v 1.2 2000/09/26 15:56:51 tsi Exp $ */ #if !defined(NASM_ASSEMBLER) && !defined(MASM_ASSEMBLER) #define TAGLLBL(a) TAG(.L##a) diff --git a/src/mesa/drivers/dri/tdfx/dri_glide.h b/src/mesa/drivers/dri/tdfx/dri_glide.h index 52a53f7dd3..3ad2bf68c6 100644 --- a/src/mesa/drivers/dri/tdfx/dri_glide.h +++ b/src/mesa/drivers/dri/tdfx/dri_glide.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/dri_glide.h,v 1.1 2001/03/21 16:14:26 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/server/tdfx_dri.h b/src/mesa/drivers/dri/tdfx/server/tdfx_dri.h index acd0b9ae5b..dc29984a27 100644 --- a/src/mesa/drivers/dri/tdfx/server/tdfx_dri.h +++ b/src/mesa/drivers/dri/tdfx/server/tdfx_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/tdfx/tdfx_dri.h,v 1.5 2001/03/21 17:02:26 dawes Exp $ */ #ifndef _TDFX_DRI_ #define _TDFX_DRI_ diff --git a/src/mesa/drivers/dri/tdfx/tdfx_context.h b/src/mesa/drivers/dri/tdfx/tdfx_context.h index 89a7a9d6c4..05673cd186 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_context.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_context.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_context.h,v 1.5 2002/02/24 21:51:10 dawes Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_dd.h b/src/mesa/drivers/dri/tdfx/tdfx_dd.h index 5ceba9d5f0..bd61e10605 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_dd.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_dd.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_dd.h,v 1.1 2001/03/21 16:14:27 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_glide.h b/src/mesa/drivers/dri/tdfx/tdfx_glide.h index f077aa678b..69e5399e72 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_glide.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_glide.h @@ -2,7 +2,6 @@ * This file defines macros and types necessary for accessing glide3. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_glide.h,v 1.1 2002/02/22 21:45:03 dawes Exp $ */ #ifndef NEWGLIDE_H #define NEWGLIDE_H diff --git a/src/mesa/drivers/dri/tdfx/tdfx_lock.c b/src/mesa/drivers/dri/tdfx/tdfx_lock.c index a20c91d030..17cdc51ee1 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_lock.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_lock.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_lock.c,v 1.5 2002/12/16 16:19:00 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_lock.h b/src/mesa/drivers/dri/tdfx/tdfx_lock.h index 616e65b2a1..74e3f5c9cc 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_lock.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_lock.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_lock.h,v 1.3 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_pixels.c b/src/mesa/drivers/dri/tdfx/tdfx_pixels.c index 732270b2bd..b5c01f6ef2 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_pixels.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_pixels.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_pixels.c,v 1.4 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_pixels.h b/src/mesa/drivers/dri/tdfx/tdfx_pixels.h index c38ce070ca..55f7eedef8 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_pixels.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_pixels.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_pixels.h,v 1.2 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_render.c b/src/mesa/drivers/dri/tdfx/tdfx_render.c index f36c97bfeb..e374f09df3 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_render.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_render.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_render.c,v 1.4 2002/02/22 21:45:03 dawes Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_render.h b/src/mesa/drivers/dri/tdfx/tdfx_render.h index 09d0d90197..18c6168333 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_render.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_render.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_render.h,v 1.1 2001/03/21 16:14:28 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_screen.c b/src/mesa/drivers/dri/tdfx/tdfx_screen.c index 1f9ff4e30c..7761664394 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_screen.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_screen.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_screen.c,v 1.3 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_screen.h b/src/mesa/drivers/dri/tdfx/tdfx_screen.h index 90be89a352..5a68898b36 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_screen.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_screen.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_screen.h,v 1.2 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_span.c b/src/mesa/drivers/dri/tdfx/tdfx_span.c index d9d52d2b6f..6b38fa5a01 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_span.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_span.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_span.c,v 1.7 2002/10/30 12:52:00 alanh Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_span.h b/src/mesa/drivers/dri/tdfx/tdfx_span.h index 62044144f0..5af9f9b301 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_span.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_span.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_span.h,v 1.1 2001/03/21 16:14:28 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_state.c b/src/mesa/drivers/dri/tdfx/tdfx_state.c index 42cb5dfaa3..3688c76a5c 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_state.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_state.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_state.c,v 1.7 2002/10/30 12:52:00 alanh Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_state.h b/src/mesa/drivers/dri/tdfx/tdfx_state.h index b10c38f591..591ea5b083 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_state.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_state.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_state.h,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tex.c b/src/mesa/drivers/dri/tdfx/tdfx_tex.c index 89865d9637..65e665ee39 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tex.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_tex.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_tex.c,v 1.7 2002/11/05 17:46:10 tsi Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tex.h b/src/mesa/drivers/dri/tdfx/tdfx_tex.h index f536c25a2f..a445935a01 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tex.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_tex.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_tex.h,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_texman.c b/src/mesa/drivers/dri/tdfx/tdfx_texman.c index 6f782f687f..f9b2726da2 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_texman.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_texman.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_texman.c,v 1.5 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_texman.h b/src/mesa/drivers/dri/tdfx/tdfx_texman.h index 739d4e142f..a9af4cb7c5 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_texman.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_texman.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_texman.h,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_texstate.c b/src/mesa/drivers/dri/tdfx/tdfx_texstate.c index fda9ce5684..bbd2c8cfee 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_texstate.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_texstate.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_texstate.c,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_texstate.h b/src/mesa/drivers/dri/tdfx/tdfx_texstate.h index 234ed4439a..0c5c4101ca 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_texstate.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_texstate.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_texstate.h,v 1.1 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tris.c b/src/mesa/drivers/dri/tdfx/tdfx_tris.c index 7252a7e7dc..59ff35a7fa 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tris.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_tris.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_tris.c,v 1.4 2002/10/30 12:52:01 alanh Exp $ */ /* New fixes: * Daniel Borca , 19 Jul 2004 diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tris.h b/src/mesa/drivers/dri/tdfx/tdfx_tris.h index 57e5d9b0ae..a591decf1d 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tris.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_tris.h @@ -29,7 +29,6 @@ * Keith Whitwell * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_tris.h,v 1.5 2002/10/30 12:52:01 alanh Exp $ */ #ifndef TDFX_TRIS_INC #define TDFX_TRIS_INC diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vb.c b/src/mesa/drivers/dri/tdfx/tdfx_vb.c index 0580135d1b..62885daaa5 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vb.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_vb.c @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_vb.c,v 1.3 2002/10/30 12:52:01 alanh Exp $ */ #include "glheader.h" #include "mtypes.h" diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vb.h b/src/mesa/drivers/dri/tdfx/tdfx_vb.h index 7b7cd9065a..6389ec95b1 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vb.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_vb.h @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_vb.h,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ #ifndef TDFXVB_INC #define TDFXVB_INC diff --git a/src/mesa/drivers/dri/unichrome/server/via_dri.c b/src/mesa/drivers/dri/unichrome/server/via_dri.c index 6944bd66f9..9833145940 100644 --- a/src/mesa/drivers/dri/unichrome/server/via_dri.c +++ b/src/mesa/drivers/dri/unichrome/server/via_dri.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/via/via_dri.c,v 1.4 2003/09/24 02:43:30 dawes Exp $ */ /* * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved. * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved. diff --git a/src/mesa/drivers/dri/unichrome/server/via_driver.h b/src/mesa/drivers/dri/unichrome/server/via_driver.h index 997b2e41a7..a643fd9fbb 100644 --- a/src/mesa/drivers/dri/unichrome/server/via_driver.h +++ b/src/mesa/drivers/dri/unichrome/server/via_driver.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/via/via_driver.h,v 1.7 2003/11/06 18:38:11 tsi Exp $ */ /* * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved. * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved. diff --git a/src/mesa/drivers/dri/unichrome/server/via_priv.h b/src/mesa/drivers/dri/unichrome/server/via_priv.h index 587531b37c..352eac0597 100644 --- a/src/mesa/drivers/dri/unichrome/server/via_priv.h +++ b/src/mesa/drivers/dri/unichrome/server/via_priv.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/via/via_priv.h,v 1.3 2003/08/27 15:16:12 tsi Exp $ */ #ifndef _VIA_PRIV_H_ #define _VIA_PRIV_H_ 1 diff --git a/src/mesa/drivers/ggi/default/genkgi.h b/src/mesa/drivers/ggi/default/genkgi.h index 022189138f..6d0963416f 100644 --- a/src/mesa/drivers/ggi/default/genkgi.h +++ b/src/mesa/drivers/ggi/default/genkgi.h @@ -1,4 +1,4 @@ -/* $Id: genkgi.h,v 1.3 1999/08/22 08:56:50 jtaylor Exp $ +/* ****************************************************************************** GGIMesa - KGIcon specific overrides for fbcon-mesa diff --git a/src/mesa/drivers/ggi/default/genkgi_mode.c b/src/mesa/drivers/ggi/default/genkgi_mode.c index 938024789f..f81d6a45bd 100644 --- a/src/mesa/drivers/ggi/default/genkgi_mode.c +++ b/src/mesa/drivers/ggi/default/genkgi_mode.c @@ -1,4 +1,4 @@ -/* $Id: genkgi_mode.c,v 1.4 2000/01/07 08:34:44 jtaylor Exp $ +/* ****************************************************************************** display-fbdev-kgicon-generic-mesa diff --git a/src/mesa/drivers/ggi/default/genkgi_visual.c b/src/mesa/drivers/ggi/default/genkgi_visual.c index 17ef9679bb..d7838cae6e 100644 --- a/src/mesa/drivers/ggi/default/genkgi_visual.c +++ b/src/mesa/drivers/ggi/default/genkgi_visual.c @@ -1,4 +1,4 @@ -/* $Id: genkgi_visual.c,v 1.7 2000/06/11 20:11:55 jtaylor Exp $ +/* ****************************************************************************** genkgi_visual.c: visual handling for the generic KGI helper diff --git a/src/mesa/drivers/ggi/include/ggi/mesa/debug.h b/src/mesa/drivers/ggi/include/ggi/mesa/debug.h index 35d11624c6..f461fee72c 100644 --- a/src/mesa/drivers/ggi/include/ggi/mesa/debug.h +++ b/src/mesa/drivers/ggi/include/ggi/mesa/debug.h @@ -1,4 +1,4 @@ -/* $Id: debug.h,v 1.5 2003/09/22 15:18:51 brianp Exp $ +/* ****************************************************************************** GGIMesa debugging macros diff --git a/src/mesa/drivers/svga/svgamesa.c b/src/mesa/drivers/svga/svgamesa.c index d138587569..1e4e185d65 100644 --- a/src/mesa/drivers/svga/svgamesa.c +++ b/src/mesa/drivers/svga/svgamesa.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa.c,v 1.27 2006/10/15 18:51:22 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa15.c b/src/mesa/drivers/svga/svgamesa15.c index ae5104d0c0..934aaa33fb 100644 --- a/src/mesa/drivers/svga/svgamesa15.c +++ b/src/mesa/drivers/svga/svgamesa15.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa15.c,v 1.11.36.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa15.h b/src/mesa/drivers/svga/svgamesa15.h index 3ed7db82ee..d453fb8d35 100644 --- a/src/mesa/drivers/svga/svgamesa15.h +++ b/src/mesa/drivers/svga/svgamesa15.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa15.h,v 1.7 2002/11/11 18:42:39 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa16.c b/src/mesa/drivers/svga/svgamesa16.c index a59937bfb4..9fc8c786e8 100644 --- a/src/mesa/drivers/svga/svgamesa16.c +++ b/src/mesa/drivers/svga/svgamesa16.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa16.c,v 1.11.36.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa16.h b/src/mesa/drivers/svga/svgamesa16.h index 247c1f4045..b80cd3dd7e 100644 --- a/src/mesa/drivers/svga/svgamesa16.h +++ b/src/mesa/drivers/svga/svgamesa16.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa16.h,v 1.6 2002/11/11 18:42:41 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa24.c b/src/mesa/drivers/svga/svgamesa24.c index dd15bf38db..c7c095333f 100644 --- a/src/mesa/drivers/svga/svgamesa24.c +++ b/src/mesa/drivers/svga/svgamesa24.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa24.c,v 1.12.36.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa24.h b/src/mesa/drivers/svga/svgamesa24.h index 54d1a8298b..df5fa68c44 100644 --- a/src/mesa/drivers/svga/svgamesa24.h +++ b/src/mesa/drivers/svga/svgamesa24.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa24.h,v 1.6 2002/11/11 18:42:41 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa32.c b/src/mesa/drivers/svga/svgamesa32.c index 4da18795d8..d089c20c05 100644 --- a/src/mesa/drivers/svga/svgamesa32.c +++ b/src/mesa/drivers/svga/svgamesa32.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa32.c,v 1.12.36.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa32.h b/src/mesa/drivers/svga/svgamesa32.h index f518e11ad5..6cf8315300 100644 --- a/src/mesa/drivers/svga/svgamesa32.h +++ b/src/mesa/drivers/svga/svgamesa32.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa32.h,v 1.6 2002/11/11 18:42:42 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa8.c b/src/mesa/drivers/svga/svgamesa8.c index 4264fcd959..2f7048a930 100644 --- a/src/mesa/drivers/svga/svgamesa8.c +++ b/src/mesa/drivers/svga/svgamesa8.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa8.c,v 1.9.10.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa8.h b/src/mesa/drivers/svga/svgamesa8.h index 1aa25f93fc..d2b0509480 100644 --- a/src/mesa/drivers/svga/svgamesa8.h +++ b/src/mesa/drivers/svga/svgamesa8.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa8.h,v 1.4 2001/02/06 00:03:48 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgapix.h b/src/mesa/drivers/svga/svgapix.h index 0b19551bf6..19cb74487d 100644 --- a/src/mesa/drivers/svga/svgapix.h +++ b/src/mesa/drivers/svga/svgapix.h @@ -1,4 +1,3 @@ -/* $Id: svgapix.h,v 1.5 2002/11/11 18:42:44 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gdi/wgl.c b/src/mesa/drivers/windows/gdi/wgl.c index dad3dc1160..6e00d08aba 100644 --- a/src/mesa/drivers/windows/gdi/wgl.c +++ b/src/mesa/drivers/windows/gdi/wgl.c @@ -1,4 +1,3 @@ -/* $Id: wgl.c,v 1.12 2006/03/30 07:58:24 kschultz Exp $ */ /* * This library is free software; you can redistribute it and/or diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c index ecc40e8f8b..72e5e1308c 100644 --- a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c +++ b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c @@ -1,4 +1,3 @@ -/* $Id: gld_vb_mesa_render_dx7.c,v 1.6 2005/08/27 13:56:08 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c index 414a2f64bf..9ab562010c 100644 --- a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c +++ b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c @@ -1,4 +1,3 @@ -/* $Id: gld_vb_mesa_render_dx8.c,v 1.6 2005/08/27 13:56:08 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c index c71fdefbae..64acab2d2a 100644 --- a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c +++ b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c @@ -1,4 +1,3 @@ -/* $Id: gld_vb_mesa_render_dx9.c,v 1.6 2005/08/27 13:56:08 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/gld_debug_clip.c b/src/mesa/drivers/windows/gldirect/gld_debug_clip.c index 1eb19ca84b..044d2e66f4 100644 --- a/src/mesa/drivers/windows/gldirect/gld_debug_clip.c +++ b/src/mesa/drivers/windows/gldirect/gld_debug_clip.c @@ -1,4 +1,3 @@ -/* $Id: gld_debug_clip.c,v 1.1 2004/04/20 11:13:11 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/gld_debug_norm.c b/src/mesa/drivers/windows/gldirect/gld_debug_norm.c index 00c428bd26..c20362bb24 100644 --- a/src/mesa/drivers/windows/gldirect/gld_debug_norm.c +++ b/src/mesa/drivers/windows/gldirect/gld_debug_norm.c @@ -1,4 +1,3 @@ -/* $Id: gld_debug_norm.c,v 1.1 2004/04/20 11:13:11 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/gld_debug_xform.c b/src/mesa/drivers/windows/gldirect/gld_debug_xform.c index d6e64b8ffd..73439dc3b6 100644 --- a/src/mesa/drivers/windows/gldirect/gld_debug_xform.c +++ b/src/mesa/drivers/windows/gldirect/gld_debug_xform.c @@ -1,4 +1,3 @@ -/* $Id: gld_debug_xform.c,v 1.1 2004/04/20 11:13:11 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/mesasw/colors.h b/src/mesa/drivers/windows/gldirect/mesasw/colors.h index 17371a96cc..9c1f2a0540 100644 --- a/src/mesa/drivers/windows/gldirect/mesasw/colors.h +++ b/src/mesa/drivers/windows/gldirect/mesasw/colors.h @@ -18,12 +18,11 @@ * (mark@rsinc.com). */ -/* $Log: ddcolors.h 1997/6/14 by Li Wei(liwei@aiar.xjtu.edu.cn) +/* * Macros for pixel format defined */ /* - * $Log: colors.h,v $ * Revision 1.1 2004/04/20 11:13:11 alanh * add SciTech's GLDirect driver for Windows. * @@ -46,7 +45,6 @@ */ /* - * $Log: colors.h,v $ * Revision 1.1 2004/04/20 11:13:11 alanh * add SciTech's GLDirect driver for Windows. * @@ -69,7 +67,6 @@ */ /* - * $Log: colors.h,v $ * Revision 1.1 2004/04/20 11:13:11 alanh * add SciTech's GLDirect driver for Windows. * @@ -520,4 +517,4 @@ char unsigned const aWinGHalftoneTranslation[216] = 225, 226, 255, -}; \ No newline at end of file +}; diff --git a/src/mesa/glapi/mesadef.py b/src/mesa/glapi/mesadef.py index 097348dae0..0f410fc482 100644 --- a/src/mesa/glapi/mesadef.py +++ b/src/mesa/glapi/mesadef.py @@ -1,6 +1,5 @@ #!/usr/bin/env python -# $Id: mesadef.py,v 1.4 2006/01/25 15:05:36 brianp Exp $ # Mesa 3-D graphics library # Version: 4.1 diff --git a/src/mesa/sparc/norm.S b/src/mesa/sparc/norm.S index 713cd5b375..44950a10a5 100644 --- a/src/mesa/sparc/norm.S +++ b/src/mesa/sparc/norm.S @@ -1,4 +1,3 @@ -/* $Id: norm.S,v 1.5 2005/07/28 00:11:11 idr Exp $ */ #include "sparc_matrix.h" diff --git a/src/mesa/sparc/sparc.h b/src/mesa/sparc/sparc.h index 55ab12122d..a98e4d0e40 100644 --- a/src/mesa/sparc/sparc.h +++ b/src/mesa/sparc/sparc.h @@ -1,4 +1,3 @@ -/* $Id: sparc.h,v 1.3 2001/06/06 22:55:28 davem69 Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/sparc/xform.S b/src/mesa/sparc/xform.S index f44ec794e9..f2b9674bf2 100644 --- a/src/mesa/sparc/xform.S +++ b/src/mesa/sparc/xform.S @@ -1,4 +1,3 @@ -/* $Id: xform.S,v 1.4 2005/07/28 00:11:11 idr Exp $ */ /* TODO * diff --git a/src/mesa/x86-64/x86-64.c b/src/mesa/x86-64/x86-64.c index 09508b66d5..dee09fd648 100644 --- a/src/mesa/x86-64/x86-64.c +++ b/src/mesa/x86-64/x86-64.c @@ -1,4 +1,3 @@ -/* $Id: x86-64.c,v 1.4 2006/10/17 17:03:21 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86-64/x86-64.h b/src/mesa/x86-64/x86-64.h index fdbd154d5d..1d931fa345 100644 --- a/src/mesa/x86-64/x86-64.h +++ b/src/mesa/x86-64/x86-64.h @@ -1,4 +1,3 @@ -/* $Id: x86-64.h,v 1.1 2005/05/07 16:59:59 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86-64/xform4.S b/src/mesa/x86-64/xform4.S index 65328f6666..667ecf6e58 100644 --- a/src/mesa/x86-64/xform4.S +++ b/src/mesa/x86-64/xform4.S @@ -1,4 +1,3 @@ -/* $Id: xform4.S,v 1.2 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow.c b/src/mesa/x86/3dnow.c index 032aa661f4..4122ee4b00 100644 --- a/src/mesa/x86/3dnow.c +++ b/src/mesa/x86/3dnow.c @@ -1,4 +1,3 @@ -/* $Id: 3dnow.c,v 1.24 2005/10/07 17:18:52 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow.h b/src/mesa/x86/3dnow.h index 1f2fd8e8b4..df9f2638d7 100644 --- a/src/mesa/x86/3dnow.h +++ b/src/mesa/x86/3dnow.h @@ -1,4 +1,3 @@ -/* $Id: 3dnow.h,v 1.6 2002/04/09 14:58:03 keithw Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_normal.S b/src/mesa/x86/3dnow_normal.S index f3bbcb27b7..693a7864db 100644 --- a/src/mesa/x86/3dnow_normal.S +++ b/src/mesa/x86/3dnow_normal.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_normal.S,v 1.10 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_xform1.S b/src/mesa/x86/3dnow_xform1.S index 22b12cca06..7665c0ff8b 100644 --- a/src/mesa/x86/3dnow_xform1.S +++ b/src/mesa/x86/3dnow_xform1.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_xform1.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_xform2.S b/src/mesa/x86/3dnow_xform2.S index d9e96d04e2..b201d1e901 100644 --- a/src/mesa/x86/3dnow_xform2.S +++ b/src/mesa/x86/3dnow_xform2.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_xform2.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_xform3.S b/src/mesa/x86/3dnow_xform3.S index babee1caa0..46f155697d 100644 --- a/src/mesa/x86/3dnow_xform3.S +++ b/src/mesa/x86/3dnow_xform3.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_xform3.S,v 1.5 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_xform4.S b/src/mesa/x86/3dnow_xform4.S index b16d2b12dd..a0c6b193cd 100644 --- a/src/mesa/x86/3dnow_xform4.S +++ b/src/mesa/x86/3dnow_xform4.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_xform4.S,v 1.5 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/clip_args.h b/src/mesa/x86/clip_args.h index cccf801981..796611fbfd 100644 --- a/src/mesa/x86/clip_args.h +++ b/src/mesa/x86/clip_args.h @@ -1,4 +1,3 @@ -/* $Id: clip_args.h,v 1.5 2002/10/29 20:28:57 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/common_x86_asm.h b/src/mesa/x86/common_x86_asm.h index 9977298328..89312b2437 100644 --- a/src/mesa/x86/common_x86_asm.h +++ b/src/mesa/x86/common_x86_asm.h @@ -1,4 +1,3 @@ -/* $Id: common_x86_asm.h,v 1.12 2005/07/16 00:56:20 ajax Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/common_x86_features.h b/src/mesa/x86/common_x86_features.h index 90509775cf..676af8c1f8 100644 --- a/src/mesa/x86/common_x86_features.h +++ b/src/mesa/x86/common_x86_features.h @@ -1,4 +1,3 @@ -/* $Id: common_x86_features.h,v 1.6 2003/01/21 16:14:00 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/common_x86_macros.h b/src/mesa/x86/common_x86_macros.h index ba155caae1..462f32b3f2 100644 --- a/src/mesa/x86/common_x86_macros.h +++ b/src/mesa/x86/common_x86_macros.h @@ -1,4 +1,3 @@ -/* $Id: common_x86_macros.h,v 1.3 2002/10/29 20:28:58 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/norm_args.h b/src/mesa/x86/norm_args.h index 1b43d57a20..5d352838be 100644 --- a/src/mesa/x86/norm_args.h +++ b/src/mesa/x86/norm_args.h @@ -1,4 +1,3 @@ -/* $Id: norm_args.h,v 1.4 2003/11/26 08:32:36 dborca Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse.h b/src/mesa/x86/sse.h index 98146a9047..521f91e411 100644 --- a/src/mesa/x86/sse.h +++ b/src/mesa/x86/sse.h @@ -1,4 +1,3 @@ -/* $Id: sse.h,v 1.2 2002/04/09 14:58:03 keithw Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_normal.S b/src/mesa/x86/sse_normal.S index 066d46e5ef..1c32e3b2fe 100644 --- a/src/mesa/x86/sse_normal.S +++ b/src/mesa/x86/sse_normal.S @@ -1,4 +1,3 @@ -/* $Id: sse_normal.S,v 1.6 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_xform1.S b/src/mesa/x86/sse_xform1.S index 4051f606a7..22fd8dd27b 100644 --- a/src/mesa/x86/sse_xform1.S +++ b/src/mesa/x86/sse_xform1.S @@ -1,4 +1,3 @@ -/* $Id: sse_xform1.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_xform2.S b/src/mesa/x86/sse_xform2.S index 06fe086bd4..52eeb27ef5 100644 --- a/src/mesa/x86/sse_xform2.S +++ b/src/mesa/x86/sse_xform2.S @@ -1,4 +1,3 @@ -/* $Id: sse_xform2.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_xform3.S b/src/mesa/x86/sse_xform3.S index eafbe34288..5e0cd8b666 100644 --- a/src/mesa/x86/sse_xform3.S +++ b/src/mesa/x86/sse_xform3.S @@ -1,4 +1,3 @@ -/* $Id: sse_xform3.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_xform4.S b/src/mesa/x86/sse_xform4.S index 24c323194f..13680528db 100644 --- a/src/mesa/x86/sse_xform4.S +++ b/src/mesa/x86/sse_xform4.S @@ -1,4 +1,3 @@ -/* $Id: sse_xform4.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86.c b/src/mesa/x86/x86.c index 6b74e9e375..82caa42dbd 100644 --- a/src/mesa/x86/x86.c +++ b/src/mesa/x86/x86.c @@ -1,4 +1,3 @@ -/* $Id: x86.c,v 1.26 2005/10/07 17:18:52 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86.h b/src/mesa/x86/x86.h index a646aff46b..97651ec6ee 100644 --- a/src/mesa/x86/x86.h +++ b/src/mesa/x86/x86.h @@ -1,4 +1,3 @@ -/* $Id: x86.h,v 1.5 2002/04/09 14:58:03 keithw Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86_cliptest.S b/src/mesa/x86/x86_cliptest.S index 5a45ee6ae6..c7a3a9b57e 100644 --- a/src/mesa/x86/x86_cliptest.S +++ b/src/mesa/x86/x86_cliptest.S @@ -1,4 +1,3 @@ -/* $Id: x86_cliptest.S,v 1.12 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86_xform2.S b/src/mesa/x86/x86_xform2.S index 94f6989d42..e41661d546 100644 --- a/src/mesa/x86/x86_xform2.S +++ b/src/mesa/x86/x86_xform2.S @@ -1,4 +1,3 @@ -/* $Id: x86_xform2.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86_xform3.S b/src/mesa/x86/x86_xform3.S index 747e2f4d28..067ddd7d7c 100644 --- a/src/mesa/x86/x86_xform3.S +++ b/src/mesa/x86/x86_xform3.S @@ -1,4 +1,3 @@ -/* $Id: x86_xform3.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86_xform4.S b/src/mesa/x86/x86_xform4.S index def3c1ceb9..77621ac4bd 100644 --- a/src/mesa/x86/x86_xform4.S +++ b/src/mesa/x86/x86_xform4.S @@ -1,4 +1,3 @@ -/* $Id: x86_xform4.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/xform_args.h b/src/mesa/x86/xform_args.h index 89a04205c1..b773f5198d 100644 --- a/src/mesa/x86/xform_args.h +++ b/src/mesa/x86/xform_args.h @@ -1,4 +1,3 @@ -/* $Id: xform_args.h,v 1.5 2002/10/29 20:28:58 brianp Exp $ */ /* * Mesa 3-D graphics library -- cgit v1.2.3 From fc72d7e032fc0a4130fae53106f03aa3fbe4e99e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 26 May 2008 20:39:26 +0900 Subject: Remove CVS keywords. --- src/gallium/winsys/dri/intel/server/i830_common.h | 1 - src/gallium/winsys/dri/intel/server/i830_dri.h | 1 - src/glu/mini/all.h | 1 - src/glu/mini/glu.c | 1 - src/glu/mini/gluP.h | 1 - src/glu/mini/mipmap.c | 1 - src/glu/mini/nurbs.c | 1 - src/glu/mini/nurbs.h | 1 - src/glu/mini/nurbscrv.c | 1 - src/glu/mini/polytest.c | 1 - src/glu/mini/project.c | 1 - src/glu/mini/quadric.c | 1 - src/glu/mini/tess.c | 1 - src/glu/mini/tess.h | 1 - src/glu/mini/tesselat.c | 1 - src/glu/sgi/dummy.cc | 1 - src/glu/sgi/libnurbs/interface/bezierEval.h | 2 -- src/glu/sgi/libnurbs/interface/bezierPatch.cc | 2 -- src/glu/sgi/libnurbs/interface/bezierPatch.h | 2 -- src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc | 2 -- src/glu/sgi/libnurbs/interface/bezierPatchMesh.h | 2 -- src/glu/sgi/libnurbs/interface/glcurveval.cc | 2 -- src/glu/sgi/libnurbs/interface/glimports.h | 2 -- src/glu/sgi/libnurbs/interface/glinterface.cc | 2 -- src/glu/sgi/libnurbs/interface/glrenderer.h | 2 -- src/glu/sgi/libnurbs/interface/incurveeval.cc | 2 -- src/glu/sgi/libnurbs/interface/insurfeval.cc | 2 -- src/glu/sgi/libnurbs/interface/mystdio.h | 2 -- src/glu/sgi/libnurbs/interface/mystdlib.h | 2 -- src/glu/sgi/libnurbs/internals/arc.h | 2 -- src/glu/sgi/libnurbs/internals/arcsorter.cc | 2 -- src/glu/sgi/libnurbs/internals/arcsorter.h | 2 -- src/glu/sgi/libnurbs/internals/arctess.h | 2 -- src/glu/sgi/libnurbs/internals/backend.cc | 2 -- src/glu/sgi/libnurbs/internals/backend.h | 2 -- src/glu/sgi/libnurbs/internals/basiccrveval.h | 2 -- src/glu/sgi/libnurbs/internals/basicsurfeval.h | 2 -- src/glu/sgi/libnurbs/internals/bezierarc.h | 2 -- src/glu/sgi/libnurbs/internals/bin.cc | 2 -- src/glu/sgi/libnurbs/internals/bin.h | 2 -- src/glu/sgi/libnurbs/internals/bufpool.cc | 2 -- src/glu/sgi/libnurbs/internals/bufpool.h | 2 -- src/glu/sgi/libnurbs/internals/cachingeval.cc | 2 -- src/glu/sgi/libnurbs/internals/cachingeval.h | 2 -- src/glu/sgi/libnurbs/internals/ccw.cc | 2 -- src/glu/sgi/libnurbs/internals/coveandtiler.h | 2 -- src/glu/sgi/libnurbs/internals/curve.cc | 2 -- src/glu/sgi/libnurbs/internals/curve.h | 2 -- src/glu/sgi/libnurbs/internals/curvelist.cc | 2 -- src/glu/sgi/libnurbs/internals/curvelist.h | 2 -- src/glu/sgi/libnurbs/internals/curvesub.cc | 2 -- src/glu/sgi/libnurbs/internals/dataTransform.cc | 2 -- src/glu/sgi/libnurbs/internals/dataTransform.h | 2 -- src/glu/sgi/libnurbs/internals/defines.h | 2 -- src/glu/sgi/libnurbs/internals/displaylist.cc | 2 -- src/glu/sgi/libnurbs/internals/displaylist.h | 2 -- src/glu/sgi/libnurbs/internals/displaymode.h | 2 -- src/glu/sgi/libnurbs/internals/flist.cc | 2 -- src/glu/sgi/libnurbs/internals/flist.h | 2 -- src/glu/sgi/libnurbs/internals/flistsorter.cc | 2 -- src/glu/sgi/libnurbs/internals/flistsorter.h | 2 -- src/glu/sgi/libnurbs/internals/gridline.h | 2 -- src/glu/sgi/libnurbs/internals/gridtrimvertex.h | 2 -- src/glu/sgi/libnurbs/internals/gridvertex.h | 2 -- src/glu/sgi/libnurbs/internals/hull.cc | 2 -- src/glu/sgi/libnurbs/internals/hull.h | 2 -- src/glu/sgi/libnurbs/internals/intersect.cc | 2 -- src/glu/sgi/libnurbs/internals/jarcloc.h | 2 -- src/glu/sgi/libnurbs/internals/knotvector.h | 2 -- src/glu/sgi/libnurbs/internals/mapdesc.cc | 2 -- src/glu/sgi/libnurbs/internals/mapdesc.h | 2 -- src/glu/sgi/libnurbs/internals/mapdescv.cc | 2 -- src/glu/sgi/libnurbs/internals/maplist.cc | 2 -- src/glu/sgi/libnurbs/internals/maplist.h | 2 -- src/glu/sgi/libnurbs/internals/mesher.cc | 2 -- src/glu/sgi/libnurbs/internals/mesher.h | 2 -- src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc | 2 -- src/glu/sgi/libnurbs/internals/monotonizer.cc | 2 -- src/glu/sgi/libnurbs/internals/monotonizer.h | 1 - src/glu/sgi/libnurbs/internals/myassert.h | 2 -- src/glu/sgi/libnurbs/internals/mycode.cc | 2 -- src/glu/sgi/libnurbs/internals/mystring.h | 2 -- src/glu/sgi/libnurbs/internals/nurbsconsts.h | 2 -- src/glu/sgi/libnurbs/internals/nurbstess.cc | 2 -- src/glu/sgi/libnurbs/internals/patch.cc | 2 -- src/glu/sgi/libnurbs/internals/patch.h | 2 -- src/glu/sgi/libnurbs/internals/patchlist.cc | 2 -- src/glu/sgi/libnurbs/internals/patchlist.h | 2 -- src/glu/sgi/libnurbs/internals/pwlarc.h | 2 -- src/glu/sgi/libnurbs/internals/quilt.cc | 2 -- src/glu/sgi/libnurbs/internals/quilt.h | 2 -- src/glu/sgi/libnurbs/internals/reader.cc | 2 -- src/glu/sgi/libnurbs/internals/reader.h | 2 -- src/glu/sgi/libnurbs/internals/renderhints.cc | 2 -- src/glu/sgi/libnurbs/internals/renderhints.h | 2 -- src/glu/sgi/libnurbs/internals/simplemath.h | 2 -- src/glu/sgi/libnurbs/internals/slicer.cc | 2 -- src/glu/sgi/libnurbs/internals/slicer.h | 2 -- src/glu/sgi/libnurbs/internals/sorter.cc | 2 -- src/glu/sgi/libnurbs/internals/sorter.h | 2 -- src/glu/sgi/libnurbs/internals/splitarcs.cc | 2 -- src/glu/sgi/libnurbs/internals/subdivider.h | 2 -- src/glu/sgi/libnurbs/internals/tobezier.cc | 2 -- src/glu/sgi/libnurbs/internals/trimline.cc | 2 -- src/glu/sgi/libnurbs/internals/trimline.h | 2 -- src/glu/sgi/libnurbs/internals/trimregion.cc | 2 -- src/glu/sgi/libnurbs/internals/trimregion.h | 2 -- src/glu/sgi/libnurbs/internals/trimvertex.h | 2 -- src/glu/sgi/libnurbs/internals/trimvertpool.cc | 2 -- src/glu/sgi/libnurbs/internals/trimvertpool.h | 2 -- src/glu/sgi/libnurbs/internals/types.h | 2 -- src/glu/sgi/libnurbs/internals/uarray.cc | 2 -- src/glu/sgi/libnurbs/internals/uarray.h | 2 -- src/glu/sgi/libnurbs/internals/varray.cc | 2 -- src/glu/sgi/libnurbs/internals/varray.h | 2 -- src/glu/sgi/libnurbs/nurbtess/definitions.h | 2 -- src/glu/sgi/libnurbs/nurbtess/directedLine.h | 2 -- src/glu/sgi/libnurbs/nurbtess/glimports.h | 2 -- src/glu/sgi/libnurbs/nurbtess/gridWrap.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/gridWrap.h | 2 -- src/glu/sgi/libnurbs/nurbtess/monoChain.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/monoChain.h | 2 -- src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc | 1 - src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h | 1 - src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h | 2 -- src/glu/sgi/libnurbs/nurbtess/mystdio.h | 2 -- src/glu/sgi/libnurbs/nurbtess/mystdlib.h | 2 -- src/glu/sgi/libnurbs/nurbtess/partitionX.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/partitionX.h | 2 -- src/glu/sgi/libnurbs/nurbtess/partitionY.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/partitionY.h | 2 -- src/glu/sgi/libnurbs/nurbtess/polyDBG.h | 2 -- src/glu/sgi/libnurbs/nurbtess/polyUtil.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/polyUtil.h | 2 -- src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/primitiveStream.h | 2 -- src/glu/sgi/libnurbs/nurbtess/quicksort.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/quicksort.h | 2 -- src/glu/sgi/libnurbs/nurbtess/rectBlock.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/rectBlock.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleComp.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleComp.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h | 2 -- src/glu/sgi/libnurbs/nurbtess/sampledLine.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/sampledLine.h | 2 -- src/glu/sgi/libnurbs/nurbtess/searchTree.cc | 2 -- src/glu/sgi/libnurbs/nurbtess/searchTree.h | 2 -- src/glu/sgi/libnurbs/nurbtess/zlassert.h | 2 -- src/glu/sgi/libtess/README | 1 - src/glu/sgi/libtess/alg-outline | 1 - src/glu/sgi/libtess/dict-list.h | 2 -- src/glu/sgi/libtess/dict.c | 2 -- src/glu/sgi/libtess/dict.h | 2 -- src/glu/sgi/libtess/geom.c | 2 -- src/glu/sgi/libtess/memalloc.c | 2 -- src/glu/sgi/libtess/memalloc.h | 2 -- src/glu/sgi/libtess/mesh.c | 2 -- src/glu/sgi/libtess/mesh.h | 2 -- src/glu/sgi/libtess/normal.h | 2 -- src/glu/sgi/libtess/priorityq-heap.c | 2 -- src/glu/sgi/libtess/priorityq-heap.h | 2 -- src/glu/sgi/libtess/priorityq-sort.h | 2 -- src/glu/sgi/libtess/priorityq.c | 2 -- src/glu/sgi/libtess/priorityq.h | 2 -- src/glu/sgi/libtess/render.c | 2 -- src/glu/sgi/libtess/render.h | 2 -- src/glu/sgi/libtess/sweep.h | 2 -- src/glu/sgi/libtess/tess.h | 2 -- src/glu/sgi/libtess/tessmono.c | 2 -- src/glu/sgi/libtess/tessmono.h | 2 -- src/glu/sgi/libutil/error.c | 2 -- src/glu/sgi/libutil/glue.c | 2 -- src/glu/sgi/libutil/gluint.h | 2 -- src/glu/sgi/libutil/project.c | 2 -- src/glu/sgi/libutil/registry.c | 2 -- src/glut/beos/beos_x11.cpp | 1 - src/glut/ggi/debug.h | 2 +- src/glut/glx/stroke.h | 1 - src/glut/glx/win32_x11.c | 1 - src/glx/mini/miniglx_events.c | 1 - src/glx/x11/XF86dri.c | 1 - src/glx/x11/clientattrib.c | 1 - src/glx/x11/compsize.c | 1 - src/glx/x11/dri_glx.c | 1 - src/glx/x11/eval.c | 1 - src/glx/x11/glxclient.h | 1 - src/glx/x11/glxcmds.c | 1 - src/glx/x11/glxext.c | 1 - src/glx/x11/indirect_init.h | 1 - src/glx/x11/packrender.h | 1 - src/glx/x11/packsingle.h | 1 - src/glx/x11/pixel.c | 1 - src/glx/x11/pixelstore.c | 1 - src/glx/x11/render2.c | 1 - src/glx/x11/renderpix.c | 1 - src/glx/x11/single2.c | 1 - src/glx/x11/singlepix.c | 1 - src/glx/x11/vertarr.c | 1 - src/glx/x11/xf86dri.h | 1 - src/glx/x11/xf86dristr.h | 1 - src/glx/x11/xfont.c | 1 - src/mesa/drivers/dri/common/stenciltmp.h | 1 - src/mesa/drivers/dri/common/texmem.c | 1 - src/mesa/drivers/dri/common/texmem.h | 1 - src/mesa/drivers/dri/common/utils.h | 1 - src/mesa/drivers/dri/common/vblank.c | 1 - src/mesa/drivers/dri/common/vblank.h | 1 - src/mesa/drivers/dri/ffb/ffb_bitmap.c | 2 +- src/mesa/drivers/dri/ffb/ffb_bitmap.h | 1 - src/mesa/drivers/dri/ffb/ffb_clear.c | 2 +- src/mesa/drivers/dri/ffb/ffb_context.h | 1 - src/mesa/drivers/dri/ffb/ffb_dd.c | 2 +- src/mesa/drivers/dri/ffb/ffb_dd.h | 2 +- src/mesa/drivers/dri/ffb/ffb_depth.c | 2 +- src/mesa/drivers/dri/ffb/ffb_depth.h | 1 - src/mesa/drivers/dri/ffb/ffb_fifo.h | 1 - src/mesa/drivers/dri/ffb/ffb_lines.c | 2 +- src/mesa/drivers/dri/ffb/ffb_lines.h | 1 - src/mesa/drivers/dri/ffb/ffb_linetmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_lock.h | 1 - src/mesa/drivers/dri/ffb/ffb_points.c | 2 +- src/mesa/drivers/dri/ffb/ffb_points.h | 1 - src/mesa/drivers/dri/ffb/ffb_pointtmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_rendertmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_span.c | 2 +- src/mesa/drivers/dri/ffb/ffb_span.h | 1 - src/mesa/drivers/dri/ffb/ffb_state.c | 2 +- src/mesa/drivers/dri/ffb/ffb_state.h | 1 - src/mesa/drivers/dri/ffb/ffb_stencil.c | 2 +- src/mesa/drivers/dri/ffb/ffb_stencil.h | 1 - src/mesa/drivers/dri/ffb/ffb_tex.c | 2 +- src/mesa/drivers/dri/ffb/ffb_tex.h | 2 +- src/mesa/drivers/dri/ffb/ffb_tris.c | 2 +- src/mesa/drivers/dri/ffb/ffb_tris.h | 1 - src/mesa/drivers/dri/ffb/ffb_tritmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_vb.c | 2 +- src/mesa/drivers/dri/ffb/ffb_vb.h | 1 - src/mesa/drivers/dri/ffb/ffb_vbtmp.h | 1 - src/mesa/drivers/dri/ffb/ffb_vtxfmt.c | 2 +- src/mesa/drivers/dri/ffb/ffb_vtxfmt.h | 1 - src/mesa/drivers/dri/ffb/ffb_xmesa.c | 2 +- src/mesa/drivers/dri/ffb/ffb_xmesa.h | 1 - src/mesa/drivers/dri/ffb/server/ffb_dac.h | 1 - src/mesa/drivers/dri/ffb/server/ffb_drishare.h | 1 - src/mesa/drivers/dri/ffb/server/ffb_regs.h | 1 - src/mesa/drivers/dri/gamma/gamma_client.h | 1 - src/mesa/drivers/dri/gamma/gamma_context.h | 1 - src/mesa/drivers/dri/gamma/gamma_inithw.c | 1 - src/mesa/drivers/dri/gamma/gamma_lock.c | 1 - src/mesa/drivers/dri/gamma/gamma_macros.h | 1 - src/mesa/drivers/dri/gamma/gamma_regs.h | 1 - src/mesa/drivers/dri/gamma/gamma_span.c | 1 - src/mesa/drivers/dri/gamma/gamma_state.c | 1 - src/mesa/drivers/dri/gamma/gamma_tex.c | 1 - src/mesa/drivers/dri/gamma/gamma_texmem.c | 1 - src/mesa/drivers/dri/gamma/gamma_texstate.c | 1 - src/mesa/drivers/dri/gamma/gamma_tritmp.h | 1 - src/mesa/drivers/dri/gamma/gamma_vb.c | 1 - src/mesa/drivers/dri/gamma/gamma_xmesa.c | 1 - src/mesa/drivers/dri/gamma/server/glint_common.h | 1 - src/mesa/drivers/dri/gamma/server/glint_dri.h | 1 - src/mesa/drivers/dri/i810/i810_3d_reg.h | 1 - src/mesa/drivers/dri/i810/i810context.c | 1 - src/mesa/drivers/dri/i810/i810context.h | 1 - src/mesa/drivers/dri/i810/i810ioctl.c | 1 - src/mesa/drivers/dri/i810/i810ioctl.h | 1 - src/mesa/drivers/dri/i810/i810screen.c | 1 - src/mesa/drivers/dri/i810/i810state.c | 1 - src/mesa/drivers/dri/i810/i810tex.c | 1 - src/mesa/drivers/dri/i810/i810tris.c | 1 - src/mesa/drivers/dri/i810/i810tris.h | 1 - src/mesa/drivers/dri/i810/i810vb.c | 1 - src/mesa/drivers/dri/i810/i810vb.h | 1 - src/mesa/drivers/dri/i810/server/i810_common.h | 1 - src/mesa/drivers/dri/i810/server/i810_dri.h | 1 - src/mesa/drivers/dri/i810/server/i810_reg.h | 1 - src/mesa/drivers/dri/i915/server/i830_common.h | 1 - src/mesa/drivers/dri/i915/server/i830_dri.h | 1 - src/mesa/drivers/dri/i965/server/i830_common.h | 1 - src/mesa/drivers/dri/i965/server/i830_dri.h | 1 - src/mesa/drivers/dri/mach64/mach64_context.c | 2 +- src/mesa/drivers/dri/mach64/mach64_context.h | 2 +- src/mesa/drivers/dri/mach64/mach64_dd.c | 2 +- src/mesa/drivers/dri/mach64/mach64_dd.h | 2 +- src/mesa/drivers/dri/mach64/mach64_ioctl.c | 2 +- src/mesa/drivers/dri/mach64/mach64_ioctl.h | 2 +- src/mesa/drivers/dri/mach64/mach64_lock.c | 2 +- src/mesa/drivers/dri/mach64/mach64_lock.h | 2 +- src/mesa/drivers/dri/mach64/mach64_native_vb.c | 2 +- src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h | 2 +- src/mesa/drivers/dri/mach64/mach64_reg.h | 2 +- src/mesa/drivers/dri/mach64/mach64_screen.c | 2 +- src/mesa/drivers/dri/mach64/mach64_screen.h | 2 +- src/mesa/drivers/dri/mach64/mach64_span.c | 2 +- src/mesa/drivers/dri/mach64/mach64_span.h | 2 +- src/mesa/drivers/dri/mach64/mach64_state.c | 2 +- src/mesa/drivers/dri/mach64/mach64_state.h | 2 +- src/mesa/drivers/dri/mach64/mach64_tex.c | 2 +- src/mesa/drivers/dri/mach64/mach64_tex.h | 2 +- src/mesa/drivers/dri/mach64/mach64_texmem.c | 2 +- src/mesa/drivers/dri/mach64/mach64_texstate.c | 2 +- src/mesa/drivers/dri/mach64/mach64_tris.c | 2 +- src/mesa/drivers/dri/mach64/mach64_tris.h | 2 +- src/mesa/drivers/dri/mach64/mach64_vb.c | 2 +- src/mesa/drivers/dri/mach64/mach64_vb.h | 2 +- src/mesa/drivers/dri/mach64/mach64_vbtmp.h | 2 +- src/mesa/drivers/dri/mach64/server/mach64_dri.h | 2 +- src/mesa/drivers/dri/mga/mga_texstate.c | 1 - src/mesa/drivers/dri/mga/mga_xmesa.c | 1 - src/mesa/drivers/dri/mga/mga_xmesa.h | 1 - src/mesa/drivers/dri/mga/mgacontext.h | 1 - src/mesa/drivers/dri/mga/mgadd.c | 1 - src/mesa/drivers/dri/mga/mgadd.h | 1 - src/mesa/drivers/dri/mga/mgaioctl.h | 1 - src/mesa/drivers/dri/mga/mgapixel.c | 1 - src/mesa/drivers/dri/mga/mgapixel.h | 1 - src/mesa/drivers/dri/mga/mgaregs.h | 1 - src/mesa/drivers/dri/mga/mgarender.c | 1 - src/mesa/drivers/dri/mga/mgaspan.h | 1 - src/mesa/drivers/dri/mga/mgastate.h | 1 - src/mesa/drivers/dri/mga/mgatex.c | 1 - src/mesa/drivers/dri/mga/mgatex.h | 1 - src/mesa/drivers/dri/mga/mgatexmem.c | 1 - src/mesa/drivers/dri/mga/mgatris.c | 1 - src/mesa/drivers/dri/mga/mgatris.h | 1 - src/mesa/drivers/dri/mga/mgavb.c | 1 - src/mesa/drivers/dri/mga/mgavb.h | 1 - src/mesa/drivers/dri/mga/server/mga.h | 1 - src/mesa/drivers/dri/mga/server/mga_bios.h | 2 -- src/mesa/drivers/dri/mga/server/mga_dri.c | 1 - src/mesa/drivers/dri/mga/server/mga_dri.h | 1 - src/mesa/drivers/dri/mga/server/mga_macros.h | 1 - src/mesa/drivers/dri/mga/server/mga_reg.h | 2 -- src/mesa/drivers/dri/r128/r128_context.c | 1 - src/mesa/drivers/dri/r128/r128_context.h | 1 - src/mesa/drivers/dri/r128/r128_dd.c | 1 - src/mesa/drivers/dri/r128/r128_dd.h | 1 - src/mesa/drivers/dri/r128/r128_ioctl.c | 1 - src/mesa/drivers/dri/r128/r128_ioctl.h | 1 - src/mesa/drivers/dri/r128/r128_lock.c | 1 - src/mesa/drivers/dri/r128/r128_lock.h | 1 - src/mesa/drivers/dri/r128/r128_screen.c | 1 - src/mesa/drivers/dri/r128/r128_screen.h | 1 - src/mesa/drivers/dri/r128/r128_span.c | 1 - src/mesa/drivers/dri/r128/r128_span.h | 1 - src/mesa/drivers/dri/r128/r128_state.c | 1 - src/mesa/drivers/dri/r128/r128_state.h | 1 - src/mesa/drivers/dri/r128/r128_tex.c | 1 - src/mesa/drivers/dri/r128/r128_tex.h | 1 - src/mesa/drivers/dri/r128/r128_texmem.c | 1 - src/mesa/drivers/dri/r128/r128_texobj.h | 1 - src/mesa/drivers/dri/r128/r128_texstate.c | 1 - src/mesa/drivers/dri/r128/r128_tris.c | 2 +- src/mesa/drivers/dri/r128/r128_tris.h | 1 - src/mesa/drivers/dri/r128/server/r128.h | 1 - src/mesa/drivers/dri/r128/server/r128_dri.c | 1 - src/mesa/drivers/dri/r128/server/r128_dri.h | 1 - src/mesa/drivers/dri/r128/server/r128_macros.h | 1 - src/mesa/drivers/dri/r128/server/r128_reg.h | 1 - src/mesa/drivers/dri/r128/server/r128_version.h | 1 - src/mesa/drivers/dri/radeon/radeon_compat.c | 1 - src/mesa/drivers/dri/radeon/radeon_context.c | 1 - src/mesa/drivers/dri/radeon/radeon_ioctl.c | 1 - src/mesa/drivers/dri/radeon/radeon_ioctl.h | 1 - src/mesa/drivers/dri/radeon/radeon_lighting.c | 1 - src/mesa/drivers/dri/radeon/radeon_maos.h | 1 - src/mesa/drivers/dri/radeon/radeon_maos_arrays.c | 1 - src/mesa/drivers/dri/radeon/radeon_maos_verts.c | 1 - src/mesa/drivers/dri/radeon/radeon_sanity.c | 1 - src/mesa/drivers/dri/radeon/radeon_screen.c | 1 - src/mesa/drivers/dri/radeon/radeon_screen.h | 1 - src/mesa/drivers/dri/radeon/radeon_state.c | 1 - src/mesa/drivers/dri/radeon/radeon_state.h | 1 - src/mesa/drivers/dri/radeon/radeon_state_init.c | 1 - src/mesa/drivers/dri/radeon/radeon_swtcl.c | 1 - src/mesa/drivers/dri/radeon/radeon_swtcl.h | 1 - src/mesa/drivers/dri/radeon/radeon_tcl.c | 1 - src/mesa/drivers/dri/radeon/radeon_tcl.h | 1 - src/mesa/drivers/dri/radeon/radeon_tex.c | 1 - src/mesa/drivers/dri/radeon/radeon_tex.h | 1 - src/mesa/drivers/dri/radeon/radeon_texmem.c | 1 - src/mesa/drivers/dri/radeon/radeon_texstate.c | 1 - src/mesa/drivers/dri/radeon/server/radeon.h | 1 - src/mesa/drivers/dri/radeon/server/radeon_dri.h | 1 - src/mesa/drivers/dri/radeon/server/radeon_macros.h | 1 - src/mesa/drivers/dri/radeon/server/radeon_reg.h | 1 - src/mesa/drivers/dri/savage/savagetris.c | 2 +- src/mesa/drivers/dri/savage/savagetris.h | 1 - src/mesa/drivers/dri/sis/server/sis_common.h | 2 +- src/mesa/drivers/dri/sis/server/sis_dri.h | 1 - src/mesa/drivers/dri/sis/sis_alloc.c | 1 - src/mesa/drivers/dri/sis/sis_alloc.h | 1 - src/mesa/drivers/dri/sis/sis_clear.c | 1 - src/mesa/drivers/dri/sis/sis_context.c | 1 - src/mesa/drivers/dri/sis/sis_context.h | 1 - src/mesa/drivers/dri/sis/sis_dd.c | 1 - src/mesa/drivers/dri/sis/sis_dd.h | 1 - src/mesa/drivers/dri/sis/sis_fog.c | 1 - src/mesa/drivers/dri/sis/sis_lock.c | 1 - src/mesa/drivers/dri/sis/sis_lock.h | 1 - src/mesa/drivers/dri/sis/sis_reg.h | 1 - src/mesa/drivers/dri/sis/sis_screen.c | 1 - src/mesa/drivers/dri/sis/sis_screen.h | 1 - src/mesa/drivers/dri/sis/sis_span.c | 1 - src/mesa/drivers/dri/sis/sis_span.h | 1 - src/mesa/drivers/dri/sis/sis_state.c | 1 - src/mesa/drivers/dri/sis/sis_state.h | 1 - src/mesa/drivers/dri/sis/sis_stencil.c | 1 - src/mesa/drivers/dri/sis/sis_stencil.h | 1 - src/mesa/drivers/dri/sis/sis_tex.c | 1 - src/mesa/drivers/dri/sis/sis_tex.h | 1 - src/mesa/drivers/dri/sis/sis_texstate.c | 1 - src/mesa/drivers/dri/sis/sis_tris.h | 1 - src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S | 1 - src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h | 1 - src/mesa/drivers/dri/tdfx/dri_glide.h | 1 - src/mesa/drivers/dri/tdfx/server/tdfx_dri.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_context.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_dd.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_glide.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_lock.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_lock.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_pixels.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_pixels.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_render.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_render.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_screen.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_screen.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_span.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_span.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_state.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_state.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_tex.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_tex.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_texman.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_texman.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_texstate.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_texstate.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_tris.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_tris.h | 1 - src/mesa/drivers/dri/tdfx/tdfx_vb.c | 1 - src/mesa/drivers/dri/tdfx/tdfx_vb.h | 1 - src/mesa/drivers/dri/unichrome/server/via_dri.c | 1 - src/mesa/drivers/dri/unichrome/server/via_driver.h | 1 - src/mesa/drivers/dri/unichrome/server/via_priv.h | 1 - src/mesa/drivers/ggi/default/genkgi.h | 2 +- src/mesa/drivers/ggi/default/genkgi_mode.c | 2 +- src/mesa/drivers/ggi/default/genkgi_visual.c | 2 +- src/mesa/drivers/ggi/include/ggi/mesa/debug.h | 2 +- src/mesa/drivers/svga/svgamesa.c | 1 - src/mesa/drivers/svga/svgamesa15.c | 1 - src/mesa/drivers/svga/svgamesa15.h | 1 - src/mesa/drivers/svga/svgamesa16.c | 1 - src/mesa/drivers/svga/svgamesa16.h | 1 - src/mesa/drivers/svga/svgamesa24.c | 1 - src/mesa/drivers/svga/svgamesa24.h | 1 - src/mesa/drivers/svga/svgamesa32.c | 1 - src/mesa/drivers/svga/svgamesa32.h | 1 - src/mesa/drivers/svga/svgamesa8.c | 1 - src/mesa/drivers/svga/svgamesa8.h | 1 - src/mesa/drivers/svga/svgapix.h | 1 - src/mesa/drivers/windows/gdi/wgl.c | 1 - src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c | 1 - src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c | 1 - src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c | 1 - src/mesa/drivers/windows/gldirect/gld_debug_clip.c | 1 - src/mesa/drivers/windows/gldirect/gld_debug_norm.c | 1 - src/mesa/drivers/windows/gldirect/gld_debug_xform.c | 1 - src/mesa/drivers/windows/gldirect/mesasw/colors.h | 7 ++----- src/mesa/glapi/mesadef.py | 1 - src/mesa/sparc/norm.S | 1 - src/mesa/sparc/sparc.h | 1 - src/mesa/sparc/xform.S | 1 - src/mesa/x86-64/x86-64.c | 1 - src/mesa/x86-64/x86-64.h | 1 - src/mesa/x86-64/xform4.S | 1 - src/mesa/x86/3dnow.c | 1 - src/mesa/x86/3dnow.h | 1 - src/mesa/x86/3dnow_normal.S | 1 - src/mesa/x86/3dnow_xform1.S | 1 - src/mesa/x86/3dnow_xform2.S | 1 - src/mesa/x86/3dnow_xform3.S | 1 - src/mesa/x86/3dnow_xform4.S | 1 - src/mesa/x86/clip_args.h | 1 - src/mesa/x86/common_x86_asm.h | 1 - src/mesa/x86/common_x86_features.h | 1 - src/mesa/x86/common_x86_macros.h | 1 - src/mesa/x86/norm_args.h | 1 - src/mesa/x86/sse.h | 1 - src/mesa/x86/sse_normal.S | 1 - src/mesa/x86/sse_xform1.S | 1 - src/mesa/x86/sse_xform2.S | 1 - src/mesa/x86/sse_xform3.S | 1 - src/mesa/x86/sse_xform4.S | 1 - src/mesa/x86/x86.c | 1 - src/mesa/x86/x86.h | 1 - src/mesa/x86/x86_cliptest.S | 1 - src/mesa/x86/x86_xform2.S | 1 - src/mesa/x86/x86_xform3.S | 1 - src/mesa/x86/x86_xform4.S | 1 - src/mesa/x86/xform_args.h | 1 - 509 files changed, 53 insertions(+), 677 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/server/i830_common.h b/src/gallium/winsys/dri/intel/server/i830_common.h index d4d58886ce..50bb751709 100644 --- a/src/gallium/winsys/dri/intel/server/i830_common.h +++ b/src/gallium/winsys/dri/intel/server/i830_common.h @@ -26,7 +26,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ #ifndef _I830_COMMON_H_ #define _I830_COMMON_H_ diff --git a/src/gallium/winsys/dri/intel/server/i830_dri.h b/src/gallium/winsys/dri/intel/server/i830_dri.h index c2a3af8cbf..685de4a551 100644 --- a/src/gallium/winsys/dri/intel/server/i830_dri.h +++ b/src/gallium/winsys/dri/intel/server/i830_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ #ifndef _I830_DRI_H #define _I830_DRI_H diff --git a/src/glu/mini/all.h b/src/glu/mini/all.h index d626bee937..874c935925 100644 --- a/src/glu/mini/all.h +++ b/src/glu/mini/all.h @@ -1,4 +1,3 @@ -/* $Id: all.h,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/glu.c b/src/glu/mini/glu.c index 5c7722c5f0..31429e3343 100644 --- a/src/glu/mini/glu.c +++ b/src/glu/mini/glu.c @@ -1,4 +1,3 @@ -/* $Id: glu.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/gluP.h b/src/glu/mini/gluP.h index 85fbc33c62..a39edce41f 100644 --- a/src/glu/mini/gluP.h +++ b/src/glu/mini/gluP.h @@ -1,4 +1,3 @@ -/* $Id: gluP.h,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/mipmap.c b/src/glu/mini/mipmap.c index 97297729e7..a655d214e3 100644 --- a/src/glu/mini/mipmap.c +++ b/src/glu/mini/mipmap.c @@ -1,4 +1,3 @@ -/* $Id: mipmap.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/nurbs.c b/src/glu/mini/nurbs.c index 93c0dd3ce2..9f39cacb41 100644 --- a/src/glu/mini/nurbs.c +++ b/src/glu/mini/nurbs.c @@ -1,4 +1,3 @@ -/* $Id: nurbs.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/nurbs.h b/src/glu/mini/nurbs.h index c9c9c094f1..3642e213a8 100644 --- a/src/glu/mini/nurbs.h +++ b/src/glu/mini/nurbs.h @@ -1,4 +1,3 @@ -/* $Id: nurbs.h,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/nurbscrv.c b/src/glu/mini/nurbscrv.c index 62d91b46d3..e80468fdb0 100644 --- a/src/glu/mini/nurbscrv.c +++ b/src/glu/mini/nurbscrv.c @@ -1,4 +1,3 @@ -/* $Id: nurbscrv.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/polytest.c b/src/glu/mini/polytest.c index 52f272a3cb..1ff966f61c 100644 --- a/src/glu/mini/polytest.c +++ b/src/glu/mini/polytest.c @@ -1,4 +1,3 @@ -/* $Id: polytest.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/project.c b/src/glu/mini/project.c index a2747de55f..6fa03267e5 100644 --- a/src/glu/mini/project.c +++ b/src/glu/mini/project.c @@ -1,4 +1,3 @@ -/* $Id: project.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/quadric.c b/src/glu/mini/quadric.c index 015552e123..0484890ef6 100644 --- a/src/glu/mini/quadric.c +++ b/src/glu/mini/quadric.c @@ -1,4 +1,3 @@ -/* $Id: quadric.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/tess.c b/src/glu/mini/tess.c index 1a384239be..341d29bae3 100644 --- a/src/glu/mini/tess.c +++ b/src/glu/mini/tess.c @@ -1,4 +1,3 @@ -/* $Id: tess.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/tess.h b/src/glu/mini/tess.h index 908e20972c..4e51dddd37 100644 --- a/src/glu/mini/tess.h +++ b/src/glu/mini/tess.h @@ -1,4 +1,3 @@ -/* $Id: tess.h,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/mini/tesselat.c b/src/glu/mini/tesselat.c index a1102e6e5a..47d230073f 100644 --- a/src/glu/mini/tesselat.c +++ b/src/glu/mini/tesselat.c @@ -1,4 +1,3 @@ -/* $Id: tesselat.c,v 1.2 2003/08/22 20:11:43 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/glu/sgi/dummy.cc b/src/glu/sgi/dummy.cc index fac5a63b76..bd905a2608 100644 --- a/src/glu/sgi/dummy.cc +++ b/src/glu/sgi/dummy.cc @@ -1,4 +1,3 @@ -/* $Id: dummy.cc,v 1.1 2001/03/18 13:06:19 pesco Exp $ */ /* * This file contains nothing. It's just there so there's at least a single * source file for libGLU.la in this directory. diff --git a/src/glu/sgi/libnurbs/interface/bezierEval.h b/src/glu/sgi/libnurbs/interface/bezierEval.h index 1a9f3c78e7..adecfe9b2f 100644 --- a/src/glu/sgi/libnurbs/interface/bezierEval.h +++ b/src/glu/sgi/libnurbs/interface/bezierEval.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierEval.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef _BEZIEREVAL_H diff --git a/src/glu/sgi/libnurbs/interface/bezierPatch.cc b/src/glu/sgi/libnurbs/interface/bezierPatch.cc index 836ae94e0a..fa1daed52e 100644 --- a/src/glu/sgi/libnurbs/interface/bezierPatch.cc +++ b/src/glu/sgi/libnurbs/interface/bezierPatch.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierPatch.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/interface/bezierPatch.h b/src/glu/sgi/libnurbs/interface/bezierPatch.h index 31c97ba08f..ad0f8b0d2a 100644 --- a/src/glu/sgi/libnurbs/interface/bezierPatch.h +++ b/src/glu/sgi/libnurbs/interface/bezierPatch.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierPatch.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef _BEZIERPATCH_H diff --git a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc index 9ff416ad6e..3dc16313ff 100644 --- a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc +++ b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierPatchMesh.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h index 74cf098858..2ab24dff5b 100644 --- a/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h +++ b/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/bezierPatchMesh.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef _BEZIERPATCHMESH_H diff --git a/src/glu/sgi/libnurbs/interface/glcurveval.cc b/src/glu/sgi/libnurbs/interface/glcurveval.cc index 32e4704137..b6591dba0d 100644 --- a/src/glu/sgi/libnurbs/interface/glcurveval.cc +++ b/src/glu/sgi/libnurbs/interface/glcurveval.cc @@ -35,8 +35,6 @@ /* * glcurveval.c++ * - * $Date: 2006/03/29 18:46:46 $ $Revision: 1.7 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/glcurveval.cc,v 1.7 2006/03/29 18:46:46 brianp Exp $ */ /* Polynomial Evaluator Interface */ diff --git a/src/glu/sgi/libnurbs/interface/glimports.h b/src/glu/sgi/libnurbs/interface/glimports.h index 9a9d3e32c9..2c307f63e8 100644 --- a/src/glu/sgi/libnurbs/interface/glimports.h +++ b/src/glu/sgi/libnurbs/interface/glimports.h @@ -35,8 +35,6 @@ /* * glimports.h * - * $Date: 2001/03/19 17:52:02 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/glimports.h,v 1.3 2001/03/19 17:52:02 pesco Exp $ */ #ifndef __gluimports_h_ diff --git a/src/glu/sgi/libnurbs/interface/glinterface.cc b/src/glu/sgi/libnurbs/interface/glinterface.cc index dfd16d1722..ba64bcd2dc 100644 --- a/src/glu/sgi/libnurbs/interface/glinterface.cc +++ b/src/glu/sgi/libnurbs/interface/glinterface.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/07/16 15:46:42 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/glinterface.cc,v 1.2 2001/07/16 15:46:42 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/interface/glrenderer.h b/src/glu/sgi/libnurbs/interface/glrenderer.h index 30f07632a4..8fc23125e0 100644 --- a/src/glu/sgi/libnurbs/interface/glrenderer.h +++ b/src/glu/sgi/libnurbs/interface/glrenderer.h @@ -35,8 +35,6 @@ /* * glrenderer.h * - * $Date: 2004/02/26 14:58:11 $ $Revision: 1.4 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/glrenderer.h,v 1.4 2004/02/26 14:58:11 brianp Exp $ */ #ifndef __gluglrenderer_h_ diff --git a/src/glu/sgi/libnurbs/interface/incurveeval.cc b/src/glu/sgi/libnurbs/interface/incurveeval.cc index 336cca0508..96ea8896ae 100644 --- a/src/glu/sgi/libnurbs/interface/incurveeval.cc +++ b/src/glu/sgi/libnurbs/interface/incurveeval.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2004/05/12 15:29:36 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/incurveeval.cc,v 1.2 2004/05/12 15:29:36 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/interface/insurfeval.cc b/src/glu/sgi/libnurbs/interface/insurfeval.cc index b314699c7a..78d8bece13 100644 --- a/src/glu/sgi/libnurbs/interface/insurfeval.cc +++ b/src/glu/sgi/libnurbs/interface/insurfeval.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2004/05/12 15:29:36 $ $Revision: 1.3 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/insurfeval.cc,v 1.3 2004/05/12 15:29:36 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/interface/mystdio.h b/src/glu/sgi/libnurbs/interface/mystdio.h index 6d737257f7..e9947ea393 100644 --- a/src/glu/sgi/libnurbs/interface/mystdio.h +++ b/src/glu/sgi/libnurbs/interface/mystdio.h @@ -35,8 +35,6 @@ /* * mystdio.h * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.4 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/mystdio.h,v 1.4 2006/03/14 15:08:52 brianp Exp $ */ #ifndef __glumystdio_h_ diff --git a/src/glu/sgi/libnurbs/interface/mystdlib.h b/src/glu/sgi/libnurbs/interface/mystdlib.h index 0ebbc1299f..2520b41e0a 100644 --- a/src/glu/sgi/libnurbs/interface/mystdlib.h +++ b/src/glu/sgi/libnurbs/interface/mystdlib.h @@ -35,8 +35,6 @@ /* * mystdlib.h * - * $Date: 2001/03/19 17:52:02 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/interface/mystdlib.h,v 1.3 2001/03/19 17:52:02 pesco Exp $ */ #ifndef __glumystdlib_h_ diff --git a/src/glu/sgi/libnurbs/internals/arc.h b/src/glu/sgi/libnurbs/internals/arc.h index b700a1e826..bbed33c649 100644 --- a/src/glu/sgi/libnurbs/internals/arc.h +++ b/src/glu/sgi/libnurbs/internals/arc.h @@ -35,8 +35,6 @@ /* * arc.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/arc.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __gluarc_h_ diff --git a/src/glu/sgi/libnurbs/internals/arcsorter.cc b/src/glu/sgi/libnurbs/internals/arcsorter.cc index 1a7f4c6911..1f85cb7108 100644 --- a/src/glu/sgi/libnurbs/internals/arcsorter.cc +++ b/src/glu/sgi/libnurbs/internals/arcsorter.cc @@ -35,8 +35,6 @@ /* * arcsorter.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/arcsorter.cc,v 1.2 2006/03/14 15:08:52 brianp Exp $ */ #ifndef __gluarcsorter_c_ diff --git a/src/glu/sgi/libnurbs/internals/arcsorter.h b/src/glu/sgi/libnurbs/internals/arcsorter.h index 989f80a43c..1025d30b5d 100644 --- a/src/glu/sgi/libnurbs/internals/arcsorter.h +++ b/src/glu/sgi/libnurbs/internals/arcsorter.h @@ -35,8 +35,6 @@ /* * arcsorter.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/arcsorter.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gluarcsorter_h_ diff --git a/src/glu/sgi/libnurbs/internals/arctess.h b/src/glu/sgi/libnurbs/internals/arctess.h index fc42ea5eb7..d3ea2071ea 100644 --- a/src/glu/sgi/libnurbs/internals/arctess.h +++ b/src/glu/sgi/libnurbs/internals/arctess.h @@ -35,8 +35,6 @@ /* * arctess.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/arctess.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __gluarctess_h_ diff --git a/src/glu/sgi/libnurbs/internals/backend.cc b/src/glu/sgi/libnurbs/internals/backend.cc index 97775a9768..69c46b2d52 100644 --- a/src/glu/sgi/libnurbs/internals/backend.cc +++ b/src/glu/sgi/libnurbs/internals/backend.cc @@ -35,8 +35,6 @@ /* * backend.c++ * - * $Date: 2004/05/12 15:29:36 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/backend.cc,v 1.2 2004/05/12 15:29:36 brianp Exp $ */ /* Bezier surface backend diff --git a/src/glu/sgi/libnurbs/internals/backend.h b/src/glu/sgi/libnurbs/internals/backend.h index c1f00b1a01..fb03859f27 100644 --- a/src/glu/sgi/libnurbs/internals/backend.h +++ b/src/glu/sgi/libnurbs/internals/backend.h @@ -35,8 +35,6 @@ /* * backend.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/backend.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glubackend_h_ diff --git a/src/glu/sgi/libnurbs/internals/basiccrveval.h b/src/glu/sgi/libnurbs/internals/basiccrveval.h index 0a5f66c201..41abedbb20 100644 --- a/src/glu/sgi/libnurbs/internals/basiccrveval.h +++ b/src/glu/sgi/libnurbs/internals/basiccrveval.h @@ -35,8 +35,6 @@ /* * basiccurveeval.h * - * $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/basiccrveval.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __glubasiccrveval_h_ diff --git a/src/glu/sgi/libnurbs/internals/basicsurfeval.h b/src/glu/sgi/libnurbs/internals/basicsurfeval.h index a67ded97b5..2fe76ad67d 100644 --- a/src/glu/sgi/libnurbs/internals/basicsurfeval.h +++ b/src/glu/sgi/libnurbs/internals/basicsurfeval.h @@ -35,8 +35,6 @@ /* * basicsurfeval.h * - * $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/basicsurfeval.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __glubasicsurfeval_h_ diff --git a/src/glu/sgi/libnurbs/internals/bezierarc.h b/src/glu/sgi/libnurbs/internals/bezierarc.h index 64dd31d87d..a6d5a13ee6 100644 --- a/src/glu/sgi/libnurbs/internals/bezierarc.h +++ b/src/glu/sgi/libnurbs/internals/bezierarc.h @@ -35,8 +35,6 @@ /* * bezierarc.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bezierarc.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glubezierarc_h diff --git a/src/glu/sgi/libnurbs/internals/bin.cc b/src/glu/sgi/libnurbs/internals/bin.cc index ed427567f9..54b406147b 100644 --- a/src/glu/sgi/libnurbs/internals/bin.cc +++ b/src/glu/sgi/libnurbs/internals/bin.cc @@ -35,8 +35,6 @@ /* * bin.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bin.cc,v 1.3 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/bin.h b/src/glu/sgi/libnurbs/internals/bin.h index 17d146fdf1..ecdf9b83b8 100644 --- a/src/glu/sgi/libnurbs/internals/bin.h +++ b/src/glu/sgi/libnurbs/internals/bin.h @@ -35,8 +35,6 @@ /* * bin.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bin.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glubin_h_ diff --git a/src/glu/sgi/libnurbs/internals/bufpool.cc b/src/glu/sgi/libnurbs/internals/bufpool.cc index d8d9c23db3..f60f7dc7b1 100644 --- a/src/glu/sgi/libnurbs/internals/bufpool.cc +++ b/src/glu/sgi/libnurbs/internals/bufpool.cc @@ -35,8 +35,6 @@ /* * bufpool.c++ * - * $Date: 2004/05/12 15:29:36 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bufpool.cc,v 1.2 2004/05/12 15:29:36 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/bufpool.h b/src/glu/sgi/libnurbs/internals/bufpool.h index 02e4ff247b..8eaafc4fd0 100644 --- a/src/glu/sgi/libnurbs/internals/bufpool.h +++ b/src/glu/sgi/libnurbs/internals/bufpool.h @@ -35,8 +35,6 @@ /* * bufpool.h * - * $Date: 2006/03/29 18:46:46 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/bufpool.h,v 1.3 2006/03/29 18:46:46 brianp Exp $ */ #ifndef __glubufpool_h_ diff --git a/src/glu/sgi/libnurbs/internals/cachingeval.cc b/src/glu/sgi/libnurbs/internals/cachingeval.cc index 7245ee3a18..3fab38c106 100644 --- a/src/glu/sgi/libnurbs/internals/cachingeval.cc +++ b/src/glu/sgi/libnurbs/internals/cachingeval.cc @@ -35,8 +35,6 @@ /* * cachingeval.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/cachingeval.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "cachingeval.h" diff --git a/src/glu/sgi/libnurbs/internals/cachingeval.h b/src/glu/sgi/libnurbs/internals/cachingeval.h index 578391707a..cb4c83501a 100644 --- a/src/glu/sgi/libnurbs/internals/cachingeval.h +++ b/src/glu/sgi/libnurbs/internals/cachingeval.h @@ -35,8 +35,6 @@ /* * cachingeval.h * - * $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/cachingeval.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __glucachingval_h_ diff --git a/src/glu/sgi/libnurbs/internals/ccw.cc b/src/glu/sgi/libnurbs/internals/ccw.cc index b1bb6276f7..eb01b7781a 100644 --- a/src/glu/sgi/libnurbs/internals/ccw.cc +++ b/src/glu/sgi/libnurbs/internals/ccw.cc @@ -35,8 +35,6 @@ /* * ccw.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/ccw.cc,v 1.3 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/coveandtiler.h b/src/glu/sgi/libnurbs/internals/coveandtiler.h index 4f4077e208..bb682b75c7 100644 --- a/src/glu/sgi/libnurbs/internals/coveandtiler.h +++ b/src/glu/sgi/libnurbs/internals/coveandtiler.h @@ -35,8 +35,6 @@ /* * coveandtiler.h * - * $Date: 2001/07/16 15:46:42 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/coveandtiler.h,v 1.2 2001/07/16 15:46:42 brianp Exp $ */ #ifndef __glucoveandtiler_h diff --git a/src/glu/sgi/libnurbs/internals/curve.cc b/src/glu/sgi/libnurbs/internals/curve.cc index 5517afa2db..33e2752643 100644 --- a/src/glu/sgi/libnurbs/internals/curve.cc +++ b/src/glu/sgi/libnurbs/internals/curve.cc @@ -35,8 +35,6 @@ /* * curve.c++ * - * $Date: 2004/05/12 15:29:36 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curve.cc,v 1.3 2004/05/12 15:29:36 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/curve.h b/src/glu/sgi/libnurbs/internals/curve.h index 7b7bd3dc89..6f7b1de9c0 100644 --- a/src/glu/sgi/libnurbs/internals/curve.h +++ b/src/glu/sgi/libnurbs/internals/curve.h @@ -35,8 +35,6 @@ /* * curve.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curve.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glucurve_h_ diff --git a/src/glu/sgi/libnurbs/internals/curvelist.cc b/src/glu/sgi/libnurbs/internals/curvelist.cc index e763c62945..872eb5816d 100644 --- a/src/glu/sgi/libnurbs/internals/curvelist.cc +++ b/src/glu/sgi/libnurbs/internals/curvelist.cc @@ -35,8 +35,6 @@ /* * curvelist.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curvelist.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/curvelist.h b/src/glu/sgi/libnurbs/internals/curvelist.h index d285fb5b98..afbaa353ec 100644 --- a/src/glu/sgi/libnurbs/internals/curvelist.h +++ b/src/glu/sgi/libnurbs/internals/curvelist.h @@ -35,8 +35,6 @@ /* * curvelist.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curvelist.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __glucurvelist_h_ diff --git a/src/glu/sgi/libnurbs/internals/curvesub.cc b/src/glu/sgi/libnurbs/internals/curvesub.cc index 11b15e4174..f85acc269a 100644 --- a/src/glu/sgi/libnurbs/internals/curvesub.cc +++ b/src/glu/sgi/libnurbs/internals/curvesub.cc @@ -35,8 +35,6 @@ /* * curvesub.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/curvesub.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/dataTransform.cc b/src/glu/sgi/libnurbs/internals/dataTransform.cc index 822da02228..55c0fbb159 100644 --- a/src/glu/sgi/libnurbs/internals/dataTransform.cc +++ b/src/glu/sgi/libnurbs/internals/dataTransform.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/dataTransform.cc,v 1.2 2005/10/28 13:09:23 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/dataTransform.h b/src/glu/sgi/libnurbs/internals/dataTransform.h index 1032896f13..08730e174e 100644 --- a/src/glu/sgi/libnurbs/internals/dataTransform.h +++ b/src/glu/sgi/libnurbs/internals/dataTransform.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/dataTransform.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef _DATA_TRANSFORM_H diff --git a/src/glu/sgi/libnurbs/internals/defines.h b/src/glu/sgi/libnurbs/internals/defines.h index 77b6088acc..aae1682e39 100644 --- a/src/glu/sgi/libnurbs/internals/defines.h +++ b/src/glu/sgi/libnurbs/internals/defines.h @@ -35,8 +35,6 @@ /* * defines.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/defines.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gludefines_h_ diff --git a/src/glu/sgi/libnurbs/internals/displaylist.cc b/src/glu/sgi/libnurbs/internals/displaylist.cc index 4b39a8991d..48593c6371 100644 --- a/src/glu/sgi/libnurbs/internals/displaylist.cc +++ b/src/glu/sgi/libnurbs/internals/displaylist.cc @@ -35,8 +35,6 @@ /* * displaylist.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/displaylist.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/displaylist.h b/src/glu/sgi/libnurbs/internals/displaylist.h index 13cadaeae8..4bd6d76384 100644 --- a/src/glu/sgi/libnurbs/internals/displaylist.h +++ b/src/glu/sgi/libnurbs/internals/displaylist.h @@ -35,8 +35,6 @@ /* * displaylist.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/displaylist.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gludisplaylist_h_ diff --git a/src/glu/sgi/libnurbs/internals/displaymode.h b/src/glu/sgi/libnurbs/internals/displaymode.h index 791434e6d2..9289b99b89 100644 --- a/src/glu/sgi/libnurbs/internals/displaymode.h +++ b/src/glu/sgi/libnurbs/internals/displaymode.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/displaymode.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gludisplaymode_h_ diff --git a/src/glu/sgi/libnurbs/internals/flist.cc b/src/glu/sgi/libnurbs/internals/flist.cc index 21414fd736..d3162b9f5f 100644 --- a/src/glu/sgi/libnurbs/internals/flist.cc +++ b/src/glu/sgi/libnurbs/internals/flist.cc @@ -35,8 +35,6 @@ /* * flist.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/flist.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/flist.h b/src/glu/sgi/libnurbs/internals/flist.h index 8450caad45..a643db52b8 100644 --- a/src/glu/sgi/libnurbs/internals/flist.h +++ b/src/glu/sgi/libnurbs/internals/flist.h @@ -35,8 +35,6 @@ /* * flist.h * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/flist.h,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #ifndef __gluflist_h_ diff --git a/src/glu/sgi/libnurbs/internals/flistsorter.cc b/src/glu/sgi/libnurbs/internals/flistsorter.cc index 730613224c..d49bdea3e0 100644 --- a/src/glu/sgi/libnurbs/internals/flistsorter.cc +++ b/src/glu/sgi/libnurbs/internals/flistsorter.cc @@ -35,8 +35,6 @@ /* * flistsorter.c++ * - * $Date: 2001/03/17 00:25:40 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/flistsorter.cc,v 1.1 2001/03/17 00:25:40 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/flistsorter.h b/src/glu/sgi/libnurbs/internals/flistsorter.h index 753ed255b5..d9fe81a85f 100644 --- a/src/glu/sgi/libnurbs/internals/flistsorter.h +++ b/src/glu/sgi/libnurbs/internals/flistsorter.h @@ -35,8 +35,6 @@ /* * flistsorter.h * - * $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/flistsorter.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __gluflistsorter_h_ diff --git a/src/glu/sgi/libnurbs/internals/gridline.h b/src/glu/sgi/libnurbs/internals/gridline.h index 32b70bbb29..eaa8797217 100644 --- a/src/glu/sgi/libnurbs/internals/gridline.h +++ b/src/glu/sgi/libnurbs/internals/gridline.h @@ -35,8 +35,6 @@ /* * gridline.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/gridline.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glugridline_h_ diff --git a/src/glu/sgi/libnurbs/internals/gridtrimvertex.h b/src/glu/sgi/libnurbs/internals/gridtrimvertex.h index 70a7029c54..72f737a9dc 100644 --- a/src/glu/sgi/libnurbs/internals/gridtrimvertex.h +++ b/src/glu/sgi/libnurbs/internals/gridtrimvertex.h @@ -35,8 +35,6 @@ /* * gridtrimvertex.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/gridtrimvertex.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glugridtrimvertex_h_ diff --git a/src/glu/sgi/libnurbs/internals/gridvertex.h b/src/glu/sgi/libnurbs/internals/gridvertex.h index 2eac57386a..23035a00c5 100644 --- a/src/glu/sgi/libnurbs/internals/gridvertex.h +++ b/src/glu/sgi/libnurbs/internals/gridvertex.h @@ -35,8 +35,6 @@ /* * gridvertex.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/gridvertex.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glugridvertex_h_ diff --git a/src/glu/sgi/libnurbs/internals/hull.cc b/src/glu/sgi/libnurbs/internals/hull.cc index 75f7c160d6..389ba66fb8 100644 --- a/src/glu/sgi/libnurbs/internals/hull.cc +++ b/src/glu/sgi/libnurbs/internals/hull.cc @@ -35,8 +35,6 @@ /* * hull.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/hull.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/hull.h b/src/glu/sgi/libnurbs/internals/hull.h index 34f1593a3e..30ffd6bac3 100644 --- a/src/glu/sgi/libnurbs/internals/hull.h +++ b/src/glu/sgi/libnurbs/internals/hull.h @@ -35,8 +35,6 @@ /* * hull.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/hull.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __gluhull_h_ diff --git a/src/glu/sgi/libnurbs/internals/intersect.cc b/src/glu/sgi/libnurbs/internals/intersect.cc index 6fb7e3239b..b39ea2121e 100644 --- a/src/glu/sgi/libnurbs/internals/intersect.cc +++ b/src/glu/sgi/libnurbs/internals/intersect.cc @@ -35,8 +35,6 @@ /* * intersect.c++ * - * $Date: 2005/10/28 13:09:23 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/intersect.cc,v 1.3 2005/10/28 13:09:23 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/jarcloc.h b/src/glu/sgi/libnurbs/internals/jarcloc.h index 785234f6c0..3582a607a7 100644 --- a/src/glu/sgi/libnurbs/internals/jarcloc.h +++ b/src/glu/sgi/libnurbs/internals/jarcloc.h @@ -35,8 +35,6 @@ /* * jarcloc.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/jarcloc.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glujarcloc_h_ diff --git a/src/glu/sgi/libnurbs/internals/knotvector.h b/src/glu/sgi/libnurbs/internals/knotvector.h index bb1e593326..508fc4f345 100644 --- a/src/glu/sgi/libnurbs/internals/knotvector.h +++ b/src/glu/sgi/libnurbs/internals/knotvector.h @@ -35,8 +35,6 @@ /* * knotvector.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/knotvector.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __gluknotvector_h_ diff --git a/src/glu/sgi/libnurbs/internals/mapdesc.cc b/src/glu/sgi/libnurbs/internals/mapdesc.cc index 14d01582b0..d59f8fd395 100644 --- a/src/glu/sgi/libnurbs/internals/mapdesc.cc +++ b/src/glu/sgi/libnurbs/internals/mapdesc.cc @@ -35,8 +35,6 @@ /* * mapdesc.c++ * - * $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mapdesc.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/mapdesc.h b/src/glu/sgi/libnurbs/internals/mapdesc.h index 3c4ef6ff6c..fe5d650a2a 100644 --- a/src/glu/sgi/libnurbs/internals/mapdesc.h +++ b/src/glu/sgi/libnurbs/internals/mapdesc.h @@ -35,8 +35,6 @@ /* * mapdesc.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mapdesc.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glumapdesc_h_ diff --git a/src/glu/sgi/libnurbs/internals/mapdescv.cc b/src/glu/sgi/libnurbs/internals/mapdescv.cc index 6e4bb40c90..35b38b141b 100644 --- a/src/glu/sgi/libnurbs/internals/mapdescv.cc +++ b/src/glu/sgi/libnurbs/internals/mapdescv.cc @@ -35,8 +35,6 @@ /* * mapdescv.c++ * - * $Date: 2004/05/12 15:29:36 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mapdescv.cc,v 1.3 2004/05/12 15:29:36 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/maplist.cc b/src/glu/sgi/libnurbs/internals/maplist.cc index 44f8666b7a..f944d1529e 100644 --- a/src/glu/sgi/libnurbs/internals/maplist.cc +++ b/src/glu/sgi/libnurbs/internals/maplist.cc @@ -35,8 +35,6 @@ /* * maplist.c++ * - * $Date: 2005/10/28 13:09:23 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/maplist.cc,v 1.2 2005/10/28 13:09:23 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/maplist.h b/src/glu/sgi/libnurbs/internals/maplist.h index ca92def8ca..e86253966d 100644 --- a/src/glu/sgi/libnurbs/internals/maplist.h +++ b/src/glu/sgi/libnurbs/internals/maplist.h @@ -35,8 +35,6 @@ /* * maplist.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/maplist.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glumaplist_h_ diff --git a/src/glu/sgi/libnurbs/internals/mesher.cc b/src/glu/sgi/libnurbs/internals/mesher.cc index 1178eeb516..9cc436adbf 100644 --- a/src/glu/sgi/libnurbs/internals/mesher.cc +++ b/src/glu/sgi/libnurbs/internals/mesher.cc @@ -35,8 +35,6 @@ /* * mesher.c++ * - * $Date: 2001/11/29 16:16:55 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mesher.cc,v 1.3 2001/11/29 16:16:55 kschultz Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/mesher.h b/src/glu/sgi/libnurbs/internals/mesher.h index e4cb4466bc..b9f74f3819 100644 --- a/src/glu/sgi/libnurbs/internals/mesher.h +++ b/src/glu/sgi/libnurbs/internals/mesher.h @@ -35,8 +35,6 @@ /* * mesher.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mesher.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __glumesher_h_ diff --git a/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc b/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc index b08cd91570..2830cc743c 100644 --- a/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc +++ b/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.3 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/monoTriangulationBackend.cc,v 1.3 2005/10/28 13:09:23 brianp Exp $ */ #include "monoTriangulation.h" diff --git a/src/glu/sgi/libnurbs/internals/monotonizer.cc b/src/glu/sgi/libnurbs/internals/monotonizer.cc index 7b6685dd06..5845d310ba 100644 --- a/src/glu/sgi/libnurbs/internals/monotonizer.cc +++ b/src/glu/sgi/libnurbs/internals/monotonizer.cc @@ -35,8 +35,6 @@ /* * monotonizer.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/monotonizer.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/monotonizer.h b/src/glu/sgi/libnurbs/internals/monotonizer.h index f516207cb5..7282a8c491 100644 --- a/src/glu/sgi/libnurbs/internals/monotonizer.h +++ b/src/glu/sgi/libnurbs/internals/monotonizer.h @@ -35,7 +35,6 @@ /* * monotonizer.h * - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/monotonizer.h,v 1.2 2006/04/03 22:23:52 ajax Exp $ */ #ifndef __glumonotonizer_h_ diff --git a/src/glu/sgi/libnurbs/internals/myassert.h b/src/glu/sgi/libnurbs/internals/myassert.h index e222c7e43e..9b5ee0f353 100644 --- a/src/glu/sgi/libnurbs/internals/myassert.h +++ b/src/glu/sgi/libnurbs/internals/myassert.h @@ -35,8 +35,6 @@ /* * myassert.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/myassert.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glumyassert_h_ diff --git a/src/glu/sgi/libnurbs/internals/mycode.cc b/src/glu/sgi/libnurbs/internals/mycode.cc index c49f64fa21..3625cacd74 100644 --- a/src/glu/sgi/libnurbs/internals/mycode.cc +++ b/src/glu/sgi/libnurbs/internals/mycode.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mycode.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "mymath.h" diff --git a/src/glu/sgi/libnurbs/internals/mystring.h b/src/glu/sgi/libnurbs/internals/mystring.h index 8b9032bf23..fedf32f114 100644 --- a/src/glu/sgi/libnurbs/internals/mystring.h +++ b/src/glu/sgi/libnurbs/internals/mystring.h @@ -35,8 +35,6 @@ /* * mystring.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/mystring.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glumystring_h_ diff --git a/src/glu/sgi/libnurbs/internals/nurbsconsts.h b/src/glu/sgi/libnurbs/internals/nurbsconsts.h index 7b9dcc39cd..30277d6892 100644 --- a/src/glu/sgi/libnurbs/internals/nurbsconsts.h +++ b/src/glu/sgi/libnurbs/internals/nurbsconsts.h @@ -35,8 +35,6 @@ /* * nurbsconsts.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/nurbsconsts.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glunurbsconsts_h_ diff --git a/src/glu/sgi/libnurbs/internals/nurbstess.cc b/src/glu/sgi/libnurbs/internals/nurbstess.cc index adf7c74626..a5bd060fb0 100644 --- a/src/glu/sgi/libnurbs/internals/nurbstess.cc +++ b/src/glu/sgi/libnurbs/internals/nurbstess.cc @@ -35,8 +35,6 @@ /* * nurbstess.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/nurbstess.cc,v 1.2 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/patch.cc b/src/glu/sgi/libnurbs/internals/patch.cc index 4a524f1de2..808baa69e4 100644 --- a/src/glu/sgi/libnurbs/internals/patch.cc +++ b/src/glu/sgi/libnurbs/internals/patch.cc @@ -35,8 +35,6 @@ /* * patch.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.4 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/patch.cc,v 1.4 2006/03/14 15:08:52 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/patch.h b/src/glu/sgi/libnurbs/internals/patch.h index a214b571f9..d42613b67e 100644 --- a/src/glu/sgi/libnurbs/internals/patch.h +++ b/src/glu/sgi/libnurbs/internals/patch.h @@ -35,8 +35,6 @@ /* * patch.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/patch.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glupatch_h_ diff --git a/src/glu/sgi/libnurbs/internals/patchlist.cc b/src/glu/sgi/libnurbs/internals/patchlist.cc index a640893f1b..989d2dd00a 100644 --- a/src/glu/sgi/libnurbs/internals/patchlist.cc +++ b/src/glu/sgi/libnurbs/internals/patchlist.cc @@ -35,8 +35,6 @@ /* * patchlist.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/patchlist.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/patchlist.h b/src/glu/sgi/libnurbs/internals/patchlist.h index 9fb3795c09..04701c292b 100644 --- a/src/glu/sgi/libnurbs/internals/patchlist.h +++ b/src/glu/sgi/libnurbs/internals/patchlist.h @@ -35,8 +35,6 @@ /* * patchlist.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/patchlist.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glupatchlist_h_ diff --git a/src/glu/sgi/libnurbs/internals/pwlarc.h b/src/glu/sgi/libnurbs/internals/pwlarc.h index 83b7c3f813..b0422b4ded 100644 --- a/src/glu/sgi/libnurbs/internals/pwlarc.h +++ b/src/glu/sgi/libnurbs/internals/pwlarc.h @@ -35,8 +35,6 @@ /* * pwlarc.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/pwlarc.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glupwlarc_h_ diff --git a/src/glu/sgi/libnurbs/internals/quilt.cc b/src/glu/sgi/libnurbs/internals/quilt.cc index f693b370ba..4fc58b7473 100644 --- a/src/glu/sgi/libnurbs/internals/quilt.cc +++ b/src/glu/sgi/libnurbs/internals/quilt.cc @@ -35,8 +35,6 @@ /* * quilt.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/quilt.cc,v 1.3 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/quilt.h b/src/glu/sgi/libnurbs/internals/quilt.h index 336c2574d2..a23c3c11b1 100644 --- a/src/glu/sgi/libnurbs/internals/quilt.h +++ b/src/glu/sgi/libnurbs/internals/quilt.h @@ -35,8 +35,6 @@ /* * quilt.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/quilt.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __gluquilt_h_ diff --git a/src/glu/sgi/libnurbs/internals/reader.cc b/src/glu/sgi/libnurbs/internals/reader.cc index 271a32fbc1..6135eef60e 100644 --- a/src/glu/sgi/libnurbs/internals/reader.cc +++ b/src/glu/sgi/libnurbs/internals/reader.cc @@ -35,8 +35,6 @@ /* * reader.c++ * - * $Date: 2002/11/01 23:35:07 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/reader.cc,v 1.2 2002/11/01 23:35:07 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/reader.h b/src/glu/sgi/libnurbs/internals/reader.h index ac86f8a29f..c826d3812f 100644 --- a/src/glu/sgi/libnurbs/internals/reader.h +++ b/src/glu/sgi/libnurbs/internals/reader.h @@ -35,8 +35,6 @@ /* * reader.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/reader.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __glureader_h_ diff --git a/src/glu/sgi/libnurbs/internals/renderhints.cc b/src/glu/sgi/libnurbs/internals/renderhints.cc index 6a9d37e013..a3aa62d42c 100644 --- a/src/glu/sgi/libnurbs/internals/renderhints.cc +++ b/src/glu/sgi/libnurbs/internals/renderhints.cc @@ -35,8 +35,6 @@ /* * renderhints.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/renderhints.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/renderhints.h b/src/glu/sgi/libnurbs/internals/renderhints.h index efa959e3ab..e248422925 100644 --- a/src/glu/sgi/libnurbs/internals/renderhints.h +++ b/src/glu/sgi/libnurbs/internals/renderhints.h @@ -35,8 +35,6 @@ /* * renderhints.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/renderhints.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glurenderhints_h_ diff --git a/src/glu/sgi/libnurbs/internals/simplemath.h b/src/glu/sgi/libnurbs/internals/simplemath.h index f2efee35f1..195471e23e 100644 --- a/src/glu/sgi/libnurbs/internals/simplemath.h +++ b/src/glu/sgi/libnurbs/internals/simplemath.h @@ -35,8 +35,6 @@ /* * simplemath.h * - * $Date: 2002/11/01 23:35:07 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/simplemath.h,v 1.2 2002/11/01 23:35:07 brianp Exp $ */ #ifndef __glusimplemath_h_ diff --git a/src/glu/sgi/libnurbs/internals/slicer.cc b/src/glu/sgi/libnurbs/internals/slicer.cc index 3fc7e2723a..27d2a650d1 100644 --- a/src/glu/sgi/libnurbs/internals/slicer.cc +++ b/src/glu/sgi/libnurbs/internals/slicer.cc @@ -35,8 +35,6 @@ /* * slicer.c++ * - * $Date: 2005/10/28 13:09:23 $ $Revision: 1.5 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/slicer.cc,v 1.5 2005/10/28 13:09:23 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/internals/slicer.h b/src/glu/sgi/libnurbs/internals/slicer.h index 6027eaa1c0..6700024ba2 100644 --- a/src/glu/sgi/libnurbs/internals/slicer.h +++ b/src/glu/sgi/libnurbs/internals/slicer.h @@ -35,8 +35,6 @@ /* * slicer.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/slicer.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __gluslicer_h_ diff --git a/src/glu/sgi/libnurbs/internals/sorter.cc b/src/glu/sgi/libnurbs/internals/sorter.cc index bf13a68d72..7a79941492 100644 --- a/src/glu/sgi/libnurbs/internals/sorter.cc +++ b/src/glu/sgi/libnurbs/internals/sorter.cc @@ -35,8 +35,6 @@ /* * sorter.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/sorter.cc,v 1.3 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/sorter.h b/src/glu/sgi/libnurbs/internals/sorter.h index e9c98affa7..0f6b43be37 100644 --- a/src/glu/sgi/libnurbs/internals/sorter.h +++ b/src/glu/sgi/libnurbs/internals/sorter.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/03/29 18:54:00 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/sorter.h,v 1.2 2006/03/29 18:54:00 brianp Exp $ */ #ifndef __glusorter_h_ diff --git a/src/glu/sgi/libnurbs/internals/splitarcs.cc b/src/glu/sgi/libnurbs/internals/splitarcs.cc index 716f6b9aae..1f79d543fb 100644 --- a/src/glu/sgi/libnurbs/internals/splitarcs.cc +++ b/src/glu/sgi/libnurbs/internals/splitarcs.cc @@ -35,8 +35,6 @@ /* * splitarcs.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/splitarcs.cc,v 1.2 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/subdivider.h b/src/glu/sgi/libnurbs/internals/subdivider.h index 48aff36b44..37970d6942 100644 --- a/src/glu/sgi/libnurbs/internals/subdivider.h +++ b/src/glu/sgi/libnurbs/internals/subdivider.h @@ -35,8 +35,6 @@ /* * subdivider.h * - * $Date: 2001/08/07 17:34:11 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/subdivider.h,v 1.2 2001/08/07 17:34:11 brianp Exp $ */ #ifndef __glusubdivider_h_ diff --git a/src/glu/sgi/libnurbs/internals/tobezier.cc b/src/glu/sgi/libnurbs/internals/tobezier.cc index 95ef3b68b4..531f26bc78 100644 --- a/src/glu/sgi/libnurbs/internals/tobezier.cc +++ b/src/glu/sgi/libnurbs/internals/tobezier.cc @@ -35,8 +35,6 @@ /* * tobezier.c++ * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/tobezier.cc,v 1.2 2006/03/14 15:08:52 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/trimline.cc b/src/glu/sgi/libnurbs/internals/trimline.cc index 231369b6ef..61f34cd38a 100644 --- a/src/glu/sgi/libnurbs/internals/trimline.cc +++ b/src/glu/sgi/libnurbs/internals/trimline.cc @@ -35,8 +35,6 @@ /* * trimline.c++ * - * $Date: 2005/10/28 13:09:23 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimline.cc,v 1.2 2005/10/28 13:09:23 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/trimline.h b/src/glu/sgi/libnurbs/internals/trimline.h index d28574d3e5..5d52e30aba 100644 --- a/src/glu/sgi/libnurbs/internals/trimline.h +++ b/src/glu/sgi/libnurbs/internals/trimline.h @@ -35,8 +35,6 @@ /* * trimline.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimline.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutrimline_h_ diff --git a/src/glu/sgi/libnurbs/internals/trimregion.cc b/src/glu/sgi/libnurbs/internals/trimregion.cc index 64b4ffc10e..efe7893569 100644 --- a/src/glu/sgi/libnurbs/internals/trimregion.cc +++ b/src/glu/sgi/libnurbs/internals/trimregion.cc @@ -35,8 +35,6 @@ /* * trimregion.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimregion.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/trimregion.h b/src/glu/sgi/libnurbs/internals/trimregion.h index 263b8d4719..6534a8c1da 100644 --- a/src/glu/sgi/libnurbs/internals/trimregion.h +++ b/src/glu/sgi/libnurbs/internals/trimregion.h @@ -35,8 +35,6 @@ /* * trimregion.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimregion.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutrimregion_h_ diff --git a/src/glu/sgi/libnurbs/internals/trimvertex.h b/src/glu/sgi/libnurbs/internals/trimvertex.h index 8ded26a648..85f1162167 100644 --- a/src/glu/sgi/libnurbs/internals/trimvertex.h +++ b/src/glu/sgi/libnurbs/internals/trimvertex.h @@ -35,8 +35,6 @@ /* * trimvertex.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimvertex.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutrimvertex_h_ diff --git a/src/glu/sgi/libnurbs/internals/trimvertpool.cc b/src/glu/sgi/libnurbs/internals/trimvertpool.cc index 7c12ab3999..3e5bd70380 100644 --- a/src/glu/sgi/libnurbs/internals/trimvertpool.cc +++ b/src/glu/sgi/libnurbs/internals/trimvertpool.cc @@ -35,8 +35,6 @@ /* * trimvertexpool.c++ * - * $Date: 2003/05/08 15:47:00 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimvertpool.cc,v 1.2 2003/05/08 15:47:00 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/trimvertpool.h b/src/glu/sgi/libnurbs/internals/trimvertpool.h index deb8d4c534..2420e8cca4 100644 --- a/src/glu/sgi/libnurbs/internals/trimvertpool.h +++ b/src/glu/sgi/libnurbs/internals/trimvertpool.h @@ -35,8 +35,6 @@ /* * trimvertexpool.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/trimvertpool.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutrimvertpool_h_ diff --git a/src/glu/sgi/libnurbs/internals/types.h b/src/glu/sgi/libnurbs/internals/types.h index d8e7751d0b..3f89e52593 100644 --- a/src/glu/sgi/libnurbs/internals/types.h +++ b/src/glu/sgi/libnurbs/internals/types.h @@ -35,8 +35,6 @@ /* * types.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/types.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __glutypes_h_ diff --git a/src/glu/sgi/libnurbs/internals/uarray.cc b/src/glu/sgi/libnurbs/internals/uarray.cc index 0cc3c8d273..f0e2364373 100644 --- a/src/glu/sgi/libnurbs/internals/uarray.cc +++ b/src/glu/sgi/libnurbs/internals/uarray.cc @@ -35,8 +35,6 @@ /* * uarray.c++ * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/uarray.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/uarray.h b/src/glu/sgi/libnurbs/internals/uarray.h index e7a7e00d10..908b8ccfc8 100644 --- a/src/glu/sgi/libnurbs/internals/uarray.h +++ b/src/glu/sgi/libnurbs/internals/uarray.h @@ -35,8 +35,6 @@ /* * uarray.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/uarray.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __gluuarray_h_ diff --git a/src/glu/sgi/libnurbs/internals/varray.cc b/src/glu/sgi/libnurbs/internals/varray.cc index 969bba080e..31cc73a9d0 100644 --- a/src/glu/sgi/libnurbs/internals/varray.cc +++ b/src/glu/sgi/libnurbs/internals/varray.cc @@ -35,8 +35,6 @@ /* * varray.c++ * - * $Date: 2002/11/01 23:35:07 $ $Revision: 1.2 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/varray.cc,v 1.2 2002/11/01 23:35:07 brianp Exp $ */ #include "glimports.h" diff --git a/src/glu/sgi/libnurbs/internals/varray.h b/src/glu/sgi/libnurbs/internals/varray.h index 5fb2541425..8408f27bae 100644 --- a/src/glu/sgi/libnurbs/internals/varray.h +++ b/src/glu/sgi/libnurbs/internals/varray.h @@ -35,8 +35,6 @@ /* * varray.h * - * $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/internals/varray.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __gluvarray_h_ diff --git a/src/glu/sgi/libnurbs/nurbtess/definitions.h b/src/glu/sgi/libnurbs/nurbtess/definitions.h index 216d479b1a..8dcbf20050 100644 --- a/src/glu/sgi/libnurbs/nurbtess/definitions.h +++ b/src/glu/sgi/libnurbs/nurbtess/definitions.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/definitions.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _DEFINITIONS_H diff --git a/src/glu/sgi/libnurbs/nurbtess/directedLine.h b/src/glu/sgi/libnurbs/nurbtess/directedLine.h index 009295e61e..9d68183ad3 100644 --- a/src/glu/sgi/libnurbs/nurbtess/directedLine.h +++ b/src/glu/sgi/libnurbs/nurbtess/directedLine.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/directedLine.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _DIRECTEDLINE_H diff --git a/src/glu/sgi/libnurbs/nurbtess/glimports.h b/src/glu/sgi/libnurbs/nurbtess/glimports.h index cb370218c0..2c307f63e8 100644 --- a/src/glu/sgi/libnurbs/nurbtess/glimports.h +++ b/src/glu/sgi/libnurbs/nurbtess/glimports.h @@ -35,8 +35,6 @@ /* * glimports.h * - * $Date: 2001/03/19 17:52:03 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/glimports.h,v 1.3 2001/03/19 17:52:03 pesco Exp $ */ #ifndef __gluimports_h_ diff --git a/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc b/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc index 6df10c4385..3c92039bae 100644 --- a/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc +++ b/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/gridWrap.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/gridWrap.h b/src/glu/sgi/libnurbs/nurbtess/gridWrap.h index 1c8237fe27..723988d2d0 100644 --- a/src/glu/sgi/libnurbs/nurbtess/gridWrap.h +++ b/src/glu/sgi/libnurbs/nurbtess/gridWrap.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/gridWrap.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _GRIDWRAP_H diff --git a/src/glu/sgi/libnurbs/nurbtess/monoChain.cc b/src/glu/sgi/libnurbs/nurbtess/monoChain.cc index dccbb2bbc0..814bf32fae 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoChain.cc +++ b/src/glu/sgi/libnurbs/nurbtess/monoChain.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.3 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/monoChain.cc,v 1.3 2005/10/28 13:09:23 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/monoChain.h b/src/glu/sgi/libnurbs/nurbtess/monoChain.h index e25b18028c..0302ff9ce2 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoChain.h +++ b/src/glu/sgi/libnurbs/nurbtess/monoChain.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/monoChain.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _MONO_CHAIN_H diff --git a/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc b/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc index 6405d277fe..8391205bf7 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc +++ b/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.cc @@ -31,7 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* *monoPolyPart.C diff --git a/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h b/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h index b760862dcb..51a664de34 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h +++ b/src/glu/sgi/libnurbs/nurbtess/monoPolyPart.h @@ -31,7 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* *monoPolyPart.h diff --git a/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc b/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc index d168374c98..8e8d49dda7 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc +++ b/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/03/29 18:46:46 $ $Revision: 1.5 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.cc,v 1.5 2006/03/29 18:46:46 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h b/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h index 002549ecbd..86b8b7164b 100644 --- a/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h +++ b/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/monoTriangulation.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _MONO_TRIANGULATION_H diff --git a/src/glu/sgi/libnurbs/nurbtess/mystdio.h b/src/glu/sgi/libnurbs/nurbtess/mystdio.h index a7594eecd6..e9947ea393 100644 --- a/src/glu/sgi/libnurbs/nurbtess/mystdio.h +++ b/src/glu/sgi/libnurbs/nurbtess/mystdio.h @@ -35,8 +35,6 @@ /* * mystdio.h * - * $Date: 2006/03/14 15:08:52 $ $Revision: 1.4 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/mystdio.h,v 1.4 2006/03/14 15:08:52 brianp Exp $ */ #ifndef __glumystdio_h_ diff --git a/src/glu/sgi/libnurbs/nurbtess/mystdlib.h b/src/glu/sgi/libnurbs/nurbtess/mystdlib.h index d28e70bd51..2520b41e0a 100644 --- a/src/glu/sgi/libnurbs/nurbtess/mystdlib.h +++ b/src/glu/sgi/libnurbs/nurbtess/mystdlib.h @@ -35,8 +35,6 @@ /* * mystdlib.h * - * $Date: 2001/03/19 17:52:03 $ $Revision: 1.3 $ - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/mystdlib.h,v 1.3 2001/03/19 17:52:03 pesco Exp $ */ #ifndef __glumystdlib_h_ diff --git a/src/glu/sgi/libnurbs/nurbtess/partitionX.cc b/src/glu/sgi/libnurbs/nurbtess/partitionX.cc index bfe77123c4..e25e30b0f3 100644 --- a/src/glu/sgi/libnurbs/nurbtess/partitionX.cc +++ b/src/glu/sgi/libnurbs/nurbtess/partitionX.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/partitionX.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/partitionX.h b/src/glu/sgi/libnurbs/nurbtess/partitionX.h index cd18f0eb02..bef724fe1f 100644 --- a/src/glu/sgi/libnurbs/nurbtess/partitionX.h +++ b/src/glu/sgi/libnurbs/nurbtess/partitionX.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/partitionX.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _PARTITIONX_H diff --git a/src/glu/sgi/libnurbs/nurbtess/partitionY.cc b/src/glu/sgi/libnurbs/nurbtess/partitionY.cc index 216ac07e06..297c629976 100644 --- a/src/glu/sgi/libnurbs/nurbtess/partitionY.cc +++ b/src/glu/sgi/libnurbs/nurbtess/partitionY.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/partitionY.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/partitionY.h b/src/glu/sgi/libnurbs/nurbtess/partitionY.h index b810693a5c..7e62aeaa9d 100644 --- a/src/glu/sgi/libnurbs/nurbtess/partitionY.h +++ b/src/glu/sgi/libnurbs/nurbtess/partitionY.h @@ -31,7 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* *partitionY.h: @@ -53,7 +52,6 @@ *A vertex is an interior cusp if it is a cusp and a reflex. *A vertex is an exterior cusp if it is a cusp but not a reflex. * - * $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/partitionY.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _PARTITIONY_H diff --git a/src/glu/sgi/libnurbs/nurbtess/polyDBG.h b/src/glu/sgi/libnurbs/nurbtess/polyDBG.h index a5125a50d1..832fe05093 100644 --- a/src/glu/sgi/libnurbs/nurbtess/polyDBG.h +++ b/src/glu/sgi/libnurbs/nurbtess/polyDBG.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/polyDBG.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _POLYDBG_H diff --git a/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc b/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc index 1a17bcc78a..f9a27f402c 100644 --- a/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc +++ b/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/polyUtil.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/polyUtil.h b/src/glu/sgi/libnurbs/nurbtess/polyUtil.h index 19c76d37d3..010838a9cc 100644 --- a/src/glu/sgi/libnurbs/nurbtess/polyUtil.h +++ b/src/glu/sgi/libnurbs/nurbtess/polyUtil.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/polyUtil.h,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #ifndef _POLYUTIL_H diff --git a/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc b/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc index 2d54b155ee..26d05342f9 100644 --- a/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc +++ b/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/primitiveStream.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h b/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h index 438d4ad6b0..8063dcd622 100644 --- a/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h +++ b/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/primitiveStream.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ /*we do not use the constans GL_... so that this file is independent of diff --git a/src/glu/sgi/libnurbs/nurbtess/quicksort.cc b/src/glu/sgi/libnurbs/nurbtess/quicksort.cc index f411aaa82a..9d0b290b39 100644 --- a/src/glu/sgi/libnurbs/nurbtess/quicksort.cc +++ b/src/glu/sgi/libnurbs/nurbtess/quicksort.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2002/04/17 19:30:41 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/quicksort.cc,v 1.2 2002/04/17 19:30:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/quicksort.h b/src/glu/sgi/libnurbs/nurbtess/quicksort.h index af245615b3..1a32188ee1 100644 --- a/src/glu/sgi/libnurbs/nurbtess/quicksort.h +++ b/src/glu/sgi/libnurbs/nurbtess/quicksort.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/quicksort.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _QUICKSORT_H diff --git a/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc b/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc index 932683ccac..f457b15733 100644 --- a/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc +++ b/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/rectBlock.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/rectBlock.h b/src/glu/sgi/libnurbs/nurbtess/rectBlock.h index d98b5a03e1..ce546442df 100644 --- a/src/glu/sgi/libnurbs/nurbtess/rectBlock.h +++ b/src/glu/sgi/libnurbs/nurbtess/rectBlock.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/rectBlock.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _RECTBLOCK_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc b/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc index b58de10af7..861c71bb38 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleComp.cc,v 1.2 2005/10/28 13:09:23 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleComp.h b/src/glu/sgi/libnurbs/nurbtess/sampleComp.h index 8bdc4c41eb..e35e5e291d 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleComp.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleComp.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleComp.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLECOMP_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc b/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc index b66647aa99..e12f88bab1 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h b/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h index f48dceaea6..6debef9119 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompBot.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLECOMPBOT_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc b/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc index e25b53c1a9..d01e50018b 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/08/30 19:02:45 $ $Revision: 1.4 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.cc,v 1.4 2006/08/30 19:02:45 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h b/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h index 747e35e6ad..b4b0e0732e 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompRight.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLECOMPRIGHT_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc b/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc index 0d012d47ce..b7b929623a 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h b/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h index 6875ad57e2..695092c586 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleCompTop.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLECOMPTOP_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc b/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc index c1b045437c..051f241083 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2005/10/28 13:09:23 $ $Revision: 1.5 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.cc,v 1.5 2005/10/28 13:09:23 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h b/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h index 3bfa0d4393..777a28fa2b 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampleMonoPoly.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLEMONOPOLY_H diff --git a/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc b/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc index 15332eb41c..6253a7c09d 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc +++ b/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/11/29 16:16:55 $ $Revision: 1.2 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampledLine.cc,v 1.2 2001/11/29 16:16:55 kschultz Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/sampledLine.h b/src/glu/sgi/libnurbs/nurbtess/sampledLine.h index 8925197ab3..147b8a5e12 100644 --- a/src/glu/sgi/libnurbs/nurbtess/sampledLine.h +++ b/src/glu/sgi/libnurbs/nurbtess/sampledLine.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/sampledLine.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SAMPLEDLINE_H diff --git a/src/glu/sgi/libnurbs/nurbtess/searchTree.cc b/src/glu/sgi/libnurbs/nurbtess/searchTree.cc index 45c2412b48..1865755a48 100644 --- a/src/glu/sgi/libnurbs/nurbtess/searchTree.cc +++ b/src/glu/sgi/libnurbs/nurbtess/searchTree.cc @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/searchTree.cc,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libnurbs/nurbtess/searchTree.h b/src/glu/sgi/libnurbs/nurbtess/searchTree.h index 4272248528..246099fbac 100644 --- a/src/glu/sgi/libnurbs/nurbtess/searchTree.h +++ b/src/glu/sgi/libnurbs/nurbtess/searchTree.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/searchTree.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef _SEARCHTREE_H diff --git a/src/glu/sgi/libnurbs/nurbtess/zlassert.h b/src/glu/sgi/libnurbs/nurbtess/zlassert.h index 6a3720853b..6891385196 100644 --- a/src/glu/sgi/libnurbs/nurbtess/zlassert.h +++ b/src/glu/sgi/libnurbs/nurbtess/zlassert.h @@ -31,10 +31,8 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ */ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libnurbs/nurbtess/zlassert.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ /*XXXblythe this file should be deleted*/ diff --git a/src/glu/sgi/libtess/README b/src/glu/sgi/libtess/README index 7c314b74a0..66a6011e25 100644 --- a/src/glu/sgi/libtess/README +++ b/src/glu/sgi/libtess/README @@ -1,5 +1,4 @@ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/README,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ General Polygon Tesselation diff --git a/src/glu/sgi/libtess/alg-outline b/src/glu/sgi/libtess/alg-outline index f51d68ce3b..33fd69728a 100644 --- a/src/glu/sgi/libtess/alg-outline +++ b/src/glu/sgi/libtess/alg-outline @@ -1,5 +1,4 @@ /* -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/alg-outline,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ This is only a very brief overview. There is quite a bit of diff --git a/src/glu/sgi/libtess/dict-list.h b/src/glu/sgi/libtess/dict-list.h index f5b82116d8..8cc1069c52 100644 --- a/src/glu/sgi/libtess/dict-list.h +++ b/src/glu/sgi/libtess/dict-list.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/dict-list.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __dict_list_h_ diff --git a/src/glu/sgi/libtess/dict.c b/src/glu/sgi/libtess/dict.c index e3750eea22..f42565f2ff 100644 --- a/src/glu/sgi/libtess/dict.c +++ b/src/glu/sgi/libtess/dict.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2006/04/19 14:42:01 $ $Revision: 1.3 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/dict.c,v 1.3 2006/04/19 14:42:01 brianp Exp $ */ #include diff --git a/src/glu/sgi/libtess/dict.h b/src/glu/sgi/libtess/dict.h index ea3b4064ff..8cc1069c52 100644 --- a/src/glu/sgi/libtess/dict.h +++ b/src/glu/sgi/libtess/dict.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/dict.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __dict_list_h_ diff --git a/src/glu/sgi/libtess/geom.c b/src/glu/sgi/libtess/geom.c index d009e143ad..4551531751 100644 --- a/src/glu/sgi/libtess/geom.c +++ b/src/glu/sgi/libtess/geom.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/geom.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/memalloc.c b/src/glu/sgi/libtess/memalloc.c index 61fd59aaed..11fd33b3d7 100644 --- a/src/glu/sgi/libtess/memalloc.c +++ b/src/glu/sgi/libtess/memalloc.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/memalloc.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "memalloc.h" diff --git a/src/glu/sgi/libtess/memalloc.h b/src/glu/sgi/libtess/memalloc.h index 97f223759d..5cbdb5342e 100644 --- a/src/glu/sgi/libtess/memalloc.h +++ b/src/glu/sgi/libtess/memalloc.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2003/07/24 22:41:17 $ $Revision: 1.4 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/memalloc.h,v 1.4 2003/07/24 22:41:17 brianp Exp $ */ #ifndef __memalloc_simple_h_ diff --git a/src/glu/sgi/libtess/mesh.c b/src/glu/sgi/libtess/mesh.c index 045954db91..4ffe1a6142 100644 --- a/src/glu/sgi/libtess/mesh.c +++ b/src/glu/sgi/libtess/mesh.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/mesh.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/mesh.h b/src/glu/sgi/libtess/mesh.h index 6224df415b..299b1bacb0 100644 --- a/src/glu/sgi/libtess/mesh.h +++ b/src/glu/sgi/libtess/mesh.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/mesh.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __mesh_h_ diff --git a/src/glu/sgi/libtess/normal.h b/src/glu/sgi/libtess/normal.h index c8e334f45f..9a805d56f9 100644 --- a/src/glu/sgi/libtess/normal.h +++ b/src/glu/sgi/libtess/normal.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/normal.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __normal_h_ diff --git a/src/glu/sgi/libtess/priorityq-heap.c b/src/glu/sgi/libtess/priorityq-heap.c index 6b77155e1e..3f8a6f5f0d 100644 --- a/src/glu/sgi/libtess/priorityq-heap.c +++ b/src/glu/sgi/libtess/priorityq-heap.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq-heap.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include diff --git a/src/glu/sgi/libtess/priorityq-heap.h b/src/glu/sgi/libtess/priorityq-heap.h index 39c33c3921..095cc3456c 100644 --- a/src/glu/sgi/libtess/priorityq-heap.h +++ b/src/glu/sgi/libtess/priorityq-heap.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq-heap.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __priorityq_heap_h_ diff --git a/src/glu/sgi/libtess/priorityq-sort.h b/src/glu/sgi/libtess/priorityq-sort.h index 2439238793..9e62e983e9 100644 --- a/src/glu/sgi/libtess/priorityq-sort.h +++ b/src/glu/sgi/libtess/priorityq-sort.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq-sort.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __priorityq_sort_h_ diff --git a/src/glu/sgi/libtess/priorityq.c b/src/glu/sgi/libtess/priorityq.c index fffa1d5255..7eac424e96 100644 --- a/src/glu/sgi/libtess/priorityq.c +++ b/src/glu/sgi/libtess/priorityq.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/priorityq.h b/src/glu/sgi/libtess/priorityq.h index 97ed707578..9e62e983e9 100644 --- a/src/glu/sgi/libtess/priorityq.h +++ b/src/glu/sgi/libtess/priorityq.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/priorityq.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __priorityq_sort_h_ diff --git a/src/glu/sgi/libtess/render.c b/src/glu/sgi/libtess/render.c index 97751dc810..c2b12b35c3 100644 --- a/src/glu/sgi/libtess/render.c +++ b/src/glu/sgi/libtess/render.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/render.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/render.h b/src/glu/sgi/libtess/render.h index 956569bb77..271e616c19 100644 --- a/src/glu/sgi/libtess/render.h +++ b/src/glu/sgi/libtess/render.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/render.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __render_h_ diff --git a/src/glu/sgi/libtess/sweep.h b/src/glu/sgi/libtess/sweep.h index 2223f52f59..74c375c9b3 100644 --- a/src/glu/sgi/libtess/sweep.h +++ b/src/glu/sgi/libtess/sweep.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/sweep.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __sweep_h_ diff --git a/src/glu/sgi/libtess/tess.h b/src/glu/sgi/libtess/tess.h index 2ba00b6ddb..d705d04c6e 100644 --- a/src/glu/sgi/libtess/tess.h +++ b/src/glu/sgi/libtess/tess.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/tess.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __tess_h_ diff --git a/src/glu/sgi/libtess/tessmono.c b/src/glu/sgi/libtess/tessmono.c index 77fe0ac619..a2b6eccbac 100644 --- a/src/glu/sgi/libtess/tessmono.c +++ b/src/glu/sgi/libtess/tessmono.c @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/tessmono.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libtess/tessmono.h b/src/glu/sgi/libtess/tessmono.h index 01f244f6ec..cbe8950d20 100644 --- a/src/glu/sgi/libtess/tessmono.h +++ b/src/glu/sgi/libtess/tessmono.h @@ -35,8 +35,6 @@ /* ** Author: Eric Veach, July 1994. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libtess/tessmono.h,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #ifndef __tessmono_h_ diff --git a/src/glu/sgi/libutil/error.c b/src/glu/sgi/libutil/error.c index 3d1ce9b210..24d8b70f88 100644 --- a/src/glu/sgi/libutil/error.c +++ b/src/glu/sgi/libutil/error.c @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/06/20 15:30:26 $ $Revision: 1.3 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/error.c,v 1.3 2006/06/20 15:30:26 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libutil/glue.c b/src/glu/sgi/libutil/glue.c index a0471bbe2e..6a4e6c7c6f 100644 --- a/src/glu/sgi/libutil/glue.c +++ b/src/glu/sgi/libutil/glue.c @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/09/24 09:40:40 $ $Revision: 1.3 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/glue.c,v 1.3 2001/09/24 09:40:40 joukj Exp $ */ #include diff --git a/src/glu/sgi/libutil/gluint.h b/src/glu/sgi/libutil/gluint.h index f08401df7a..cd2a56fed9 100644 --- a/src/glu/sgi/libutil/gluint.h +++ b/src/glu/sgi/libutil/gluint.h @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/09/20 21:50:53 $ $Revision: 1.2 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/gluint.h,v 1.2 2001/09/20 21:50:53 kschultz Exp $ */ #ifndef __gluint_h__ diff --git a/src/glu/sgi/libutil/project.c b/src/glu/sgi/libutil/project.c index 2b20ad4fb3..5ba396ca1c 100644 --- a/src/glu/sgi/libutil/project.c +++ b/src/glu/sgi/libutil/project.c @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2006/05/01 16:01:17 $ $Revision: 1.6 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/project.c,v 1.6 2006/05/01 16:01:17 brianp Exp $ */ #include "gluos.h" diff --git a/src/glu/sgi/libutil/registry.c b/src/glu/sgi/libutil/registry.c index e486ffa8ca..d83d2fef11 100644 --- a/src/glu/sgi/libutil/registry.c +++ b/src/glu/sgi/libutil/registry.c @@ -31,8 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. ** -** $Date: 2001/03/17 00:25:41 $ $Revision: 1.1 $ -** $Header: /home/krh/git/sync/mesa-cvs-repo/Mesa/src/glu/sgi/libutil/registry.c,v 1.1 2001/03/17 00:25:41 brianp Exp $ */ #include "gluos.h" diff --git a/src/glut/beos/beos_x11.cpp b/src/glut/beos/beos_x11.cpp index 2d1bc655cb..4f7ec48ac8 100644 --- a/src/glut/beos/beos_x11.cpp +++ b/src/glut/beos/beos_x11.cpp @@ -23,7 +23,6 @@ int DisplayHeight() { /* the following function was stolen from the X sources as indicated. */ /* Copyright Massachusetts Institute of Technology 1985, 1986, 1987 */ -/* $XConsortium: XParseGeom.c,v 11.18 91/02/21 17:23:05 rws Exp $ */ /* Permission to use, copy, modify, distribute, and sell this software and its diff --git a/src/glut/ggi/debug.h b/src/glut/ggi/debug.h index da329a1d9b..09fa960670 100644 --- a/src/glut/ggi/debug.h +++ b/src/glut/ggi/debug.h @@ -1,4 +1,4 @@ -/* $Id: debug.h,v 1.1 2000/11/19 07:41:26 jtaylor Exp $ +/* ****************************************************************************** GGIMesa debugging macros diff --git a/src/glut/glx/stroke.h b/src/glut/glx/stroke.h index fc29680bea..602b2fae9f 100644 --- a/src/glut/glx/stroke.h +++ b/src/glut/glx/stroke.h @@ -1,4 +1,3 @@ -/* $XConsortium: wfont.h,v 5.1 91/02/16 09:46:37 rws Exp $ */ /***************************************************************** Copyright (c) 1989,1990, 1991 by Sun Microsystems, Inc. and the X Consortium. diff --git a/src/glut/glx/win32_x11.c b/src/glut/glx/win32_x11.c index 1d138cfa2a..d00ccdb121 100644 --- a/src/glut/glx/win32_x11.c +++ b/src/glut/glx/win32_x11.c @@ -263,7 +263,6 @@ XPending(Display* display) /* the following function was stolen from the X sources as indicated. */ /* Copyright Massachusetts Institute of Technology 1985, 1986, 1987 */ -/* $XConsortium: XParseGeom.c,v 11.18 91/02/21 17:23:05 rws Exp $ */ /* Permission to use, copy, modify, distribute, and sell this software and its diff --git a/src/glx/mini/miniglx_events.c b/src/glx/mini/miniglx_events.c index 969398bc16..a20d5847b3 100644 --- a/src/glx/mini/miniglx_events.c +++ b/src/glx/mini/miniglx_events.c @@ -38,7 +38,6 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/* $Id: miniglx_events.c,v 1.6 2006/04/03 07:31:27 airlied Exp $ */ #include diff --git a/src/glx/x11/XF86dri.c b/src/glx/x11/XF86dri.c index 8909a04772..9919a40977 100644 --- a/src/glx/x11/XF86dri.c +++ b/src/glx/x11/XF86dri.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/dri/XF86dri.c,v 1.13 2002/10/30 12:51:25 alanh Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/glx/x11/clientattrib.c b/src/glx/x11/clientattrib.c index bfb263ced1..888f8e3187 100644 --- a/src/glx/x11/clientattrib.c +++ b/src/glx/x11/clientattrib.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/clientattrib.c,v 1.5 2001/03/21 16:04:39 dawes Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/compsize.c b/src/glx/x11/compsize.c index b8c162e8ac..2d124573ef 100644 --- a/src/glx/x11/compsize.c +++ b/src/glx/x11/compsize.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/compsize.c,v 1.6 2004/01/28 18:11:38 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/dri_glx.c b/src/glx/x11/dri_glx.c index 5cf9923979..21e07c1935 100644 --- a/src/glx/x11/dri_glx.c +++ b/src/glx/x11/dri_glx.c @@ -24,7 +24,6 @@ TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/dri/dri_glx.c,v 1.14 2003/07/16 00:54:00 dawes Exp $ */ /* * Authors: diff --git a/src/glx/x11/eval.c b/src/glx/x11/eval.c index 0f94e6da6f..2544c50fce 100644 --- a/src/glx/x11/eval.c +++ b/src/glx/x11/eval.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/glxclient.h b/src/glx/x11/glxclient.h index 477566cc46..03e44e5d04 100644 --- a/src/glx/x11/glxclient.h +++ b/src/glx/x11/glxclient.h @@ -31,7 +31,6 @@ ** published by SGI, but has not been independently verified as being ** compliant with the OpenGL(R) version 1.2.1 Specification. */ -/* $XFree86: xc/lib/GL/glx/glxclient.h,v 1.21 2004/02/09 23:46:31 alanh Exp $ */ /** * \file glxclient.h diff --git a/src/glx/x11/glxcmds.c b/src/glx/x11/glxcmds.c index f52b71ffcd..80281896f6 100644 --- a/src/glx/x11/glxcmds.c +++ b/src/glx/x11/glxcmds.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/glxcmds.c,v 1.30 2004/01/30 20:33:06 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/glxext.c b/src/glx/x11/glxext.c index af3a5166dc..2852217ba3 100644 --- a/src/glx/x11/glxext.c +++ b/src/glx/x11/glxext.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/glxext.c,v 1.22 2003/12/08 17:35:28 dawes Exp $ */ /* ** License Applicability. Except to the extent portions of this file are diff --git a/src/glx/x11/indirect_init.h b/src/glx/x11/indirect_init.h index 62d04ba6dc..72255f1301 100644 --- a/src/glx/x11/indirect_init.h +++ b/src/glx/x11/indirect_init.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/indirect_init.h,v 1.2 2000/02/08 17:18:33 dawes Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/glx/x11/packrender.h b/src/glx/x11/packrender.h index ce2a1616de..8e3119d1b2 100644 --- a/src/glx/x11/packrender.h +++ b/src/glx/x11/packrender.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/packrender.h,v 1.7tsi Exp $ */ #ifndef __GLX_packrender_h__ #define __GLX_packrender_h__ diff --git a/src/glx/x11/packsingle.h b/src/glx/x11/packsingle.h index 16b054f1e0..c69c543921 100644 --- a/src/glx/x11/packsingle.h +++ b/src/glx/x11/packsingle.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/packsingle.h,v 1.5tsi Exp $ */ #ifndef __GLX_packsingle_h__ #define __GLX_packsingle_h__ diff --git a/src/glx/x11/pixel.c b/src/glx/x11/pixel.c index 3b3a1811ab..279555bdfd 100644 --- a/src/glx/x11/pixel.c +++ b/src/glx/x11/pixel.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/pixel.c,v 1.8 2003/09/28 20:15:04 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/pixelstore.c b/src/glx/x11/pixelstore.c index 3bf1b35ba3..6f25ed786e 100644 --- a/src/glx/x11/pixelstore.c +++ b/src/glx/x11/pixelstore.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/pixelstore.c,v 1.4 2004/01/28 18:11:43 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/render2.c b/src/glx/x11/render2.c index 21ba270998..b17ad974c8 100644 --- a/src/glx/x11/render2.c +++ b/src/glx/x11/render2.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/render2.c,v 1.6 2004/01/31 09:29:33 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/renderpix.c b/src/glx/x11/renderpix.c index b7d01dc679..41a7a2d762 100644 --- a/src/glx/x11/renderpix.c +++ b/src/glx/x11/renderpix.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/renderpix.c,v 1.5 2003/09/28 20:15:04 alanh Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/single2.c b/src/glx/x11/single2.c index d535757a9e..35fe417b62 100644 --- a/src/glx/x11/single2.c +++ b/src/glx/x11/single2.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/single2.c,v 1.10 2004/02/11 19:48:16 dawes Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/singlepix.c b/src/glx/x11/singlepix.c index a7b5b79870..cd88684f70 100644 --- a/src/glx/x11/singlepix.c +++ b/src/glx/x11/singlepix.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/singlepix.c,v 1.3 2001/03/21 16:04:39 dawes Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/vertarr.c b/src/glx/x11/vertarr.c index 483a166ea2..d50560ba1a 100644 --- a/src/glx/x11/vertarr.c +++ b/src/glx/x11/vertarr.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/vertarr.c,v 1.4 2001/03/25 05:32:00 tsi Exp $ */ /* ** License Applicability. Except to the extent portions of this file are ** made subject to an alternative license as permitted in the SGI Free diff --git a/src/glx/x11/xf86dri.h b/src/glx/x11/xf86dri.h index 0a2bb24971..c8c878f127 100644 --- a/src/glx/x11/xf86dri.h +++ b/src/glx/x11/xf86dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/dri/xf86dri.h,v 1.8 2002/10/30 12:51:25 alanh Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/glx/x11/xf86dristr.h b/src/glx/x11/xf86dristr.h index ac05b183b3..b834bd1a1a 100644 --- a/src/glx/x11/xf86dristr.h +++ b/src/glx/x11/xf86dristr.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/dri/xf86dristr.h,v 1.10 2002/10/30 12:51:25 alanh Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/glx/x11/xfont.c b/src/glx/x11/xfont.c index 5f23a79622..f3e3da3e79 100644 --- a/src/glx/x11/xfont.c +++ b/src/glx/x11/xfont.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/glx/xfont.c,v 1.6 2001/05/02 15:06:02 dawes Exp $ */ /* * Mesa 3-D graphics library * Version: 3.1 diff --git a/src/mesa/drivers/dri/common/stenciltmp.h b/src/mesa/drivers/dri/common/stenciltmp.h index 324fc873d3..2b10b9ecfe 100644 --- a/src/mesa/drivers/dri/common/stenciltmp.h +++ b/src/mesa/drivers/dri/common/stenciltmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/common/stenciltmp.h,v 1.3 2001/03/21 16:14:20 dawes Exp $ */ #include "spantmp_common.h" diff --git a/src/mesa/drivers/dri/common/texmem.c b/src/mesa/drivers/dri/common/texmem.c index b0e8c4c1c2..a81cc2413d 100644 --- a/src/mesa/drivers/dri/common/texmem.c +++ b/src/mesa/drivers/dri/common/texmem.c @@ -28,7 +28,6 @@ * Kevin E. Martin * Gareth Hughes */ -/* $XFree86:$ */ /** \file texmem.c * Implements all of the device-independent texture memory management. diff --git a/src/mesa/drivers/dri/common/texmem.h b/src/mesa/drivers/dri/common/texmem.h index 6692efcc30..ffed7dd66e 100644 --- a/src/mesa/drivers/dri/common/texmem.h +++ b/src/mesa/drivers/dri/common/texmem.h @@ -28,7 +28,6 @@ * Kevin E. Martin * Gareth Hughes */ -/* $XFree86:$ */ /** \file texmem.h * Public interface to the DRI texture memory management routines. diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h index b2bab86e66..b28b895627 100644 --- a/src/mesa/drivers/dri/common/utils.h +++ b/src/mesa/drivers/dri/common/utils.h @@ -24,7 +24,6 @@ * Authors: * Ian Romanick */ -/* $XFree86:$ */ #ifndef DRI_DEBUG_H #define DRI_DEBUG_H diff --git a/src/mesa/drivers/dri/common/vblank.c b/src/mesa/drivers/dri/common/vblank.c index e7ed545f13..094950d362 100644 --- a/src/mesa/drivers/dri/common/vblank.c +++ b/src/mesa/drivers/dri/common/vblank.c @@ -25,7 +25,6 @@ * Authors: * Ian Romanick */ -/* $XFree86:$ */ #include "glheader.h" #include "xf86drm.h" diff --git a/src/mesa/drivers/dri/common/vblank.h b/src/mesa/drivers/dri/common/vblank.h index ec83adc78d..52c1933ca5 100644 --- a/src/mesa/drivers/dri/common/vblank.h +++ b/src/mesa/drivers/dri/common/vblank.h @@ -25,7 +25,6 @@ * Authors: * Ian Romanick */ -/* $XFree86:$ */ #ifndef DRI_VBLANK_H #define DRI_VBLANK_H diff --git a/src/mesa/drivers/dri/ffb/ffb_bitmap.c b/src/mesa/drivers/dri/ffb/ffb_bitmap.c index 7263e83813..1aa66859a6 100644 --- a/src/mesa/drivers/dri/ffb/ffb_bitmap.c +++ b/src/mesa/drivers/dri/ffb/ffb_bitmap.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_bitmap.c,v 1.1 2002/02/22 21:32:58 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_bitmap.h b/src/mesa/drivers/dri/ffb/ffb_bitmap.h index 4f8d2ea2a6..0ccbc57bd0 100644 --- a/src/mesa/drivers/dri/ffb/ffb_bitmap.h +++ b/src/mesa/drivers/dri/ffb/ffb_bitmap.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_bitmap.h,v 1.1 2002/02/22 21:32:58 dawes Exp $ */ #ifndef _FFB_BITMAP_H #define _FFB_BITMAP_H diff --git a/src/mesa/drivers/dri/ffb/ffb_clear.c b/src/mesa/drivers/dri/ffb/ffb_clear.c index e8dfcbe254..7de05b5cf0 100644 --- a/src/mesa/drivers/dri/ffb/ffb_clear.c +++ b/src/mesa/drivers/dri/ffb/ffb_clear.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_clear.c,v 1.2 2002/02/22 21:32:58 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_context.h b/src/mesa/drivers/dri/ffb/ffb_context.h index df1b65d748..0ab75fce47 100644 --- a/src/mesa/drivers/dri/ffb/ffb_context.h +++ b/src/mesa/drivers/dri/ffb/ffb_context.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_context.h,v 1.2 2002/02/22 21:32:58 dawes Exp $ */ #ifndef _FFB_CONTEXT_H #define _FFB_CONTEXT_H diff --git a/src/mesa/drivers/dri/ffb/ffb_dd.c b/src/mesa/drivers/dri/ffb/ffb_dd.c index 53423bbae4..f64a577d1f 100644 --- a/src/mesa/drivers/dri/ffb/ffb_dd.c +++ b/src/mesa/drivers/dri/ffb/ffb_dd.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_dd.c,v 1.4 2002/09/11 19:49:07 tsi Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_dd.h b/src/mesa/drivers/dri/ffb/ffb_dd.h index 4ffcbe6666..e065ebbecd 100644 --- a/src/mesa/drivers/dri/ffb/ffb_dd.h +++ b/src/mesa/drivers/dri/ffb/ffb_dd.h @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_dd.h,v 1.1 2000/06/20 05:08:38 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D. * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_depth.c b/src/mesa/drivers/dri/ffb/ffb_depth.c index 68a2450eb7..cca6212f50 100644 --- a/src/mesa/drivers/dri/ffb/ffb_depth.c +++ b/src/mesa/drivers/dri/ffb/ffb_depth.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_depth.c,v 1.2 2002/02/22 21:32:58 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_depth.h b/src/mesa/drivers/dri/ffb/ffb_depth.h index db908e7a63..8a1829ed49 100644 --- a/src/mesa/drivers/dri/ffb/ffb_depth.h +++ b/src/mesa/drivers/dri/ffb/ffb_depth.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_depth.h,v 1.1 2000/06/20 05:08:38 dawes Exp $ */ #ifndef _FFB_DEPTH_H #define _FFB_DEPTH_H diff --git a/src/mesa/drivers/dri/ffb/ffb_fifo.h b/src/mesa/drivers/dri/ffb/ffb_fifo.h index 886d71b76e..a175f38643 100644 --- a/src/mesa/drivers/dri/ffb/ffb_fifo.h +++ b/src/mesa/drivers/dri/ffb/ffb_fifo.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_fifo.h,v 1.2 2002/02/22 21:32:58 dawes Exp $ */ #ifndef _FFB_FIFO_H #define _FFB_FIFO_H diff --git a/src/mesa/drivers/dri/ffb/ffb_lines.c b/src/mesa/drivers/dri/ffb/ffb_lines.c index da1de18f36..8294701464 100644 --- a/src/mesa/drivers/dri/ffb/ffb_lines.c +++ b/src/mesa/drivers/dri/ffb/ffb_lines.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_lines.c,v 1.2 2002/02/22 21:32:58 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_lines.h b/src/mesa/drivers/dri/ffb/ffb_lines.h index d508c243ea..ddb9365653 100644 --- a/src/mesa/drivers/dri/ffb/ffb_lines.h +++ b/src/mesa/drivers/dri/ffb/ffb_lines.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_lines.h,v 1.2 2002/02/22 21:32:58 dawes Exp $ */ #ifndef _FFB_LINES_H #define _FFB_LINES_H diff --git a/src/mesa/drivers/dri/ffb/ffb_linetmp.h b/src/mesa/drivers/dri/ffb/ffb_linetmp.h index 0951513ca1..e9d8260e1a 100644 --- a/src/mesa/drivers/dri/ffb/ffb_linetmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_linetmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_linetmp.h,v 1.2 2002/02/22 21:32:58 dawes Exp $ */ static __inline void TAG(ffb_line)(GLcontext *ctx, ffb_vertex *v0, ffb_vertex *v1 ) diff --git a/src/mesa/drivers/dri/ffb/ffb_lock.h b/src/mesa/drivers/dri/ffb/ffb_lock.h index 7c49f740f8..1fd3eb5512 100644 --- a/src/mesa/drivers/dri/ffb/ffb_lock.h +++ b/src/mesa/drivers/dri/ffb/ffb_lock.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_lock.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_LOCK_H #define _FFB_LOCK_H diff --git a/src/mesa/drivers/dri/ffb/ffb_points.c b/src/mesa/drivers/dri/ffb/ffb_points.c index a7496dd1d6..d00255ccee 100644 --- a/src/mesa/drivers/dri/ffb/ffb_points.c +++ b/src/mesa/drivers/dri/ffb/ffb_points.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_points.c,v 1.2 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_points.h b/src/mesa/drivers/dri/ffb/ffb_points.h index 7d5c1f8a03..a7229de7f1 100644 --- a/src/mesa/drivers/dri/ffb/ffb_points.h +++ b/src/mesa/drivers/dri/ffb/ffb_points.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_points.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_POINTS_H #define _FFB_POINTS_H diff --git a/src/mesa/drivers/dri/ffb/ffb_pointtmp.h b/src/mesa/drivers/dri/ffb/ffb_pointtmp.h index 310c95d89b..2c91426b3a 100644 --- a/src/mesa/drivers/dri/ffb/ffb_pointtmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_pointtmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_pointtmp.h,v 1.3 2002/02/22 21:32:59 dawes Exp $ */ static __inline void TAG(ffb_draw_point)(GLcontext *ctx, ffb_vertex *tmp ) { diff --git a/src/mesa/drivers/dri/ffb/ffb_rendertmp.h b/src/mesa/drivers/dri/ffb/ffb_rendertmp.h index 26d991b081..64141c2c5f 100644 --- a/src/mesa/drivers/dri/ffb/ffb_rendertmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_rendertmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_rendertmp.h,v 1.2 2003/01/29 23:00:40 dawes Exp $ */ #define IMPL_LOCAL_VARS \ ffbContextPtr fmesa = FFB_CONTEXT(ctx); \ diff --git a/src/mesa/drivers/dri/ffb/ffb_span.c b/src/mesa/drivers/dri/ffb/ffb_span.c index fff7fa1d3f..59ac414678 100644 --- a/src/mesa/drivers/dri/ffb/ffb_span.c +++ b/src/mesa/drivers/dri/ffb/ffb_span.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_span.c,v 1.2 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_span.h b/src/mesa/drivers/dri/ffb/ffb_span.h index 5ae227910d..37506cf30e 100644 --- a/src/mesa/drivers/dri/ffb/ffb_span.h +++ b/src/mesa/drivers/dri/ffb/ffb_span.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_span.h,v 1.1 2000/06/20 05:08:39 dawes Exp $ */ #ifndef _FFB_SPAN_H #define _FFB_SPAN_H diff --git a/src/mesa/drivers/dri/ffb/ffb_state.c b/src/mesa/drivers/dri/ffb/ffb_state.c index eb13478166..880ad8be0a 100644 --- a/src/mesa/drivers/dri/ffb/ffb_state.c +++ b/src/mesa/drivers/dri/ffb/ffb_state.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_state.c,v 1.5 2002/10/30 12:51:27 alanh Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_state.h b/src/mesa/drivers/dri/ffb/ffb_state.h index 17b6fa20ab..19e72085fd 100644 --- a/src/mesa/drivers/dri/ffb/ffb_state.h +++ b/src/mesa/drivers/dri/ffb/ffb_state.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_state.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_STATE_H #define _FFB_STATE_H diff --git a/src/mesa/drivers/dri/ffb/ffb_stencil.c b/src/mesa/drivers/dri/ffb/ffb_stencil.c index 2f13ee9210..d535b1b778 100644 --- a/src/mesa/drivers/dri/ffb/ffb_stencil.c +++ b/src/mesa/drivers/dri/ffb/ffb_stencil.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_stencil.c,v 1.2 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_stencil.h b/src/mesa/drivers/dri/ffb/ffb_stencil.h index c7da1ca681..2d529980d1 100644 --- a/src/mesa/drivers/dri/ffb/ffb_stencil.h +++ b/src/mesa/drivers/dri/ffb/ffb_stencil.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_stencil.h,v 1.1 2000/06/20 05:08:39 dawes Exp $ */ #ifndef _FFB_STENCIL_H #define _FFB_STENCIL_H diff --git a/src/mesa/drivers/dri/ffb/ffb_tex.c b/src/mesa/drivers/dri/ffb/ffb_tex.c index d6763b7cd3..6503b0f4e7 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tex.c +++ b/src/mesa/drivers/dri/ffb/ffb_tex.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tex.c,v 1.1 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_tex.h b/src/mesa/drivers/dri/ffb/ffb_tex.h index dba0e08af6..4032e73209 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tex.h +++ b/src/mesa/drivers/dri/ffb/ffb_tex.h @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tex.h,v 1.1 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D. * Copyright (C) 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_tris.c b/src/mesa/drivers/dri/ffb/ffb_tris.c index 9fae8c8283..c2857f61bd 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tris.c +++ b/src/mesa/drivers/dri/ffb/ffb_tris.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tris.c,v 1.3 2002/10/30 12:51:28 alanh Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_tris.h b/src/mesa/drivers/dri/ffb/ffb_tris.h index a803174b3e..116b8e07f1 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tris.h +++ b/src/mesa/drivers/dri/ffb/ffb_tris.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tris.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_TRIS_H #define _FFB_TRIS_H diff --git a/src/mesa/drivers/dri/ffb/ffb_tritmp.h b/src/mesa/drivers/dri/ffb/ffb_tritmp.h index 612ef2433f..324a871ec4 100644 --- a/src/mesa/drivers/dri/ffb/ffb_tritmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_tritmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_tritmp.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ static void TAG(ffb_triangle)( GLcontext *ctx, ffb_vertex *v0, diff --git a/src/mesa/drivers/dri/ffb/ffb_vb.c b/src/mesa/drivers/dri/ffb/ffb_vb.c index 6ba1eabbf2..edc9d79124 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vb.c +++ b/src/mesa/drivers/dri/ffb/ffb_vb.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vb.c,v 1.4 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_vb.h b/src/mesa/drivers/dri/ffb/ffb_vb.h index 9eb6759f61..af669bce30 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vb.h +++ b/src/mesa/drivers/dri/ffb/ffb_vb.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vb.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_VB_H #define _FFB_VB_H diff --git a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h index a1d1254d97..0495d0e276 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vbtmp.h +++ b/src/mesa/drivers/dri/ffb/ffb_vbtmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vbtmp.h,v 1.1 2002/02/22 21:32:59 dawes Exp $ */ static void TAG(emit)(GLcontext *ctx, GLuint start, GLuint end) { diff --git a/src/mesa/drivers/dri/ffb/ffb_vtxfmt.c b/src/mesa/drivers/dri/ffb/ffb_vtxfmt.c index 9c1b770fbd..8b60f095c9 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vtxfmt.c +++ b/src/mesa/drivers/dri/ffb/ffb_vtxfmt.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vtxfmt.c,v 1.1 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_vtxfmt.h b/src/mesa/drivers/dri/ffb/ffb_vtxfmt.h index 063bb4923e..4d9125cd15 100644 --- a/src/mesa/drivers/dri/ffb/ffb_vtxfmt.h +++ b/src/mesa/drivers/dri/ffb/ffb_vtxfmt.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_vtxfmt.h,v 1.1 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_VTXFMT_H #define _FFB_VTXFMT_H diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.c b/src/mesa/drivers/dri/ffb/ffb_xmesa.c index 4c5323d230..f521de63c0 100644 --- a/src/mesa/drivers/dri/ffb/ffb_xmesa.c +++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_xmesa.c,v 1.4 2002/02/22 21:32:59 dawes Exp $ +/* * * GLX Hardware Device Driver for Sun Creator/Creator3D * Copyright (C) 2000, 2001 David S. Miller diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.h b/src/mesa/drivers/dri/ffb/ffb_xmesa.h index b7580780a6..bc8cfe9f21 100644 --- a/src/mesa/drivers/dri/ffb/ffb_xmesa.h +++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/ffb/ffb_xmesa.h,v 1.2 2002/02/22 21:32:59 dawes Exp $ */ #ifndef _FFB_XMESA_H_ #define _FFB_XMESA_H_ diff --git a/src/mesa/drivers/dri/ffb/server/ffb_dac.h b/src/mesa/drivers/dri/ffb/server/ffb_dac.h index 08114282e5..ac4a75b459 100644 --- a/src/mesa/drivers/dri/ffb/server/ffb_dac.h +++ b/src/mesa/drivers/dri/ffb/server/ffb_dac.h @@ -21,7 +21,6 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sunffb/ffb_dac.h,v 1.2 2001/04/05 17:42:33 dawes Exp $ */ #ifndef _FFB_DAC_H #define _FFB_DAC_H diff --git a/src/mesa/drivers/dri/ffb/server/ffb_drishare.h b/src/mesa/drivers/dri/ffb/server/ffb_drishare.h index baf2f0d0a6..69fefa3f0a 100644 --- a/src/mesa/drivers/dri/ffb/server/ffb_drishare.h +++ b/src/mesa/drivers/dri/ffb/server/ffb_drishare.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sunffb/ffb_drishare.h,v 1.2 2000/06/21 00:47:37 dawes Exp $ */ #ifndef _FFB_DRISHARE_H #define _FFB_DRISHARE_H diff --git a/src/mesa/drivers/dri/ffb/server/ffb_regs.h b/src/mesa/drivers/dri/ffb/server/ffb_regs.h index 7f383d38d6..bda5840d60 100644 --- a/src/mesa/drivers/dri/ffb/server/ffb_regs.h +++ b/src/mesa/drivers/dri/ffb/server/ffb_regs.h @@ -24,7 +24,6 @@ * USE OR OTHER DEALINGS IN THE SOFTWARE. * */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sunffb/ffb_regs.h,v 1.1 2000/05/18 23:21:37 dawes Exp $ */ #ifndef FFBREGS_H #define FFBREGS_H diff --git a/src/mesa/drivers/dri/gamma/gamma_client.h b/src/mesa/drivers/dri/gamma/gamma_client.h index 1c1a22ebc4..6dcf2e9438 100644 --- a/src/mesa/drivers/dri/gamma/gamma_client.h +++ b/src/mesa/drivers/dri/gamma/gamma_client.h @@ -31,7 +31,6 @@ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER * DEALINGS IN THE SOFTWARE. * - * $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_client.h,v 1.3 2002/02/22 21:33:00 dawes Exp $ * */ diff --git a/src/mesa/drivers/dri/gamma/gamma_context.h b/src/mesa/drivers/dri/gamma/gamma_context.h index f0ab1c4f05..fb70df6c37 100644 --- a/src/mesa/drivers/dri/gamma/gamma_context.h +++ b/src/mesa/drivers/dri/gamma/gamma_context.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_context.h,v 1.6 2002/12/16 16:18:50 dawes Exp $ */ /* * Copyright 2001 by Alan Hourihane. * diff --git a/src/mesa/drivers/dri/gamma/gamma_inithw.c b/src/mesa/drivers/dri/gamma/gamma_inithw.c index 47eb802b4e..79b54aacb5 100644 --- a/src/mesa/drivers/dri/gamma/gamma_inithw.c +++ b/src/mesa/drivers/dri/gamma/gamma_inithw.c @@ -23,7 +23,6 @@ * Kevin E. Martin * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_inithw.c,v 1.9 2002/10/30 12:51:29 alanh Exp $ */ #include "gamma_context.h" #include "glint_dri.h" diff --git a/src/mesa/drivers/dri/gamma/gamma_lock.c b/src/mesa/drivers/dri/gamma/gamma_lock.c index 2ab387fa27..97eea75541 100644 --- a/src/mesa/drivers/dri/gamma/gamma_lock.c +++ b/src/mesa/drivers/dri/gamma/gamma_lock.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_lock.c,v 1.4 2002/11/05 17:46:07 tsi Exp $ */ #include "gamma_context.h" #include "gamma_lock.h" diff --git a/src/mesa/drivers/dri/gamma/gamma_macros.h b/src/mesa/drivers/dri/gamma/gamma_macros.h index 974fe569df..c15483b770 100644 --- a/src/mesa/drivers/dri/gamma/gamma_macros.h +++ b/src/mesa/drivers/dri/gamma/gamma_macros.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_macros.h,v 1.5 2002/02/22 21:33:02 dawes Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/gamma/gamma_regs.h b/src/mesa/drivers/dri/gamma/gamma_regs.h index 2edda07227..9e1c735019 100644 --- a/src/mesa/drivers/dri/gamma/gamma_regs.h +++ b/src/mesa/drivers/dri/gamma/gamma_regs.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_regs.h,v 1.5 2002/02/22 21:33:02 dawes Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/gamma/gamma_span.c b/src/mesa/drivers/dri/gamma/gamma_span.c index f62bea9b66..012d77782b 100644 --- a/src/mesa/drivers/dri/gamma/gamma_span.c +++ b/src/mesa/drivers/dri/gamma/gamma_span.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_span.c,v 1.4 2002/11/05 17:46:07 tsi Exp $ */ #include "gamma_context.h" #include "gamma_lock.h" diff --git a/src/mesa/drivers/dri/gamma/gamma_state.c b/src/mesa/drivers/dri/gamma/gamma_state.c index 8dbe0a97ca..a0690f64d0 100644 --- a/src/mesa/drivers/dri/gamma/gamma_state.c +++ b/src/mesa/drivers/dri/gamma/gamma_state.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_state.c,v 1.5 2002/11/05 17:46:07 tsi Exp $ */ /* * Copyright 2001 by Alan Hourihane. * diff --git a/src/mesa/drivers/dri/gamma/gamma_tex.c b/src/mesa/drivers/dri/gamma/gamma_tex.c index d4fc93f86b..0770cbf694 100644 --- a/src/mesa/drivers/dri/gamma/gamma_tex.c +++ b/src/mesa/drivers/dri/gamma/gamma_tex.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_tex.c,v 1.4 2002/11/05 17:46:07 tsi Exp $ */ #include #include diff --git a/src/mesa/drivers/dri/gamma/gamma_texmem.c b/src/mesa/drivers/dri/gamma/gamma_texmem.c index 506b5c4c8f..94ecb5c2f6 100644 --- a/src/mesa/drivers/dri/gamma/gamma_texmem.c +++ b/src/mesa/drivers/dri/gamma/gamma_texmem.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_texmem.c,v 1.5 2002/11/05 17:46:07 tsi Exp $ */ #include #include diff --git a/src/mesa/drivers/dri/gamma/gamma_texstate.c b/src/mesa/drivers/dri/gamma/gamma_texstate.c index a8d1b253c7..b9bd6d4cee 100644 --- a/src/mesa/drivers/dri/gamma/gamma_texstate.c +++ b/src/mesa/drivers/dri/gamma/gamma_texstate.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_texstate.c,v 1.5 2002/11/05 17:46:07 tsi Exp $ */ #include #include diff --git a/src/mesa/drivers/dri/gamma/gamma_tritmp.h b/src/mesa/drivers/dri/gamma/gamma_tritmp.h index 23459ff156..56e0a850c8 100644 --- a/src/mesa/drivers/dri/gamma/gamma_tritmp.h +++ b/src/mesa/drivers/dri/gamma/gamma_tritmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/extras/Mesa/src/mesa/drivers/dri/gamma/gamma_tritmp.h,v 1.2 2004/12/13 22:40:49 tsi Exp $ */ static void TAG(gamma_point)( gammaContextPtr gmesa, const gammaVertex *v0 ) diff --git a/src/mesa/drivers/dri/gamma/gamma_vb.c b/src/mesa/drivers/dri/gamma/gamma_vb.c index 80d35cba9e..f23f585fc0 100644 --- a/src/mesa/drivers/dri/gamma/gamma_vb.c +++ b/src/mesa/drivers/dri/gamma/gamma_vb.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_vb.c,v 1.4 2003/03/26 20:43:48 tsi Exp $ */ /* * Copyright 2001 by Alan Hourihane. * diff --git a/src/mesa/drivers/dri/gamma/gamma_xmesa.c b/src/mesa/drivers/dri/gamma/gamma_xmesa.c index f41682cea7..4c0ebe1899 100644 --- a/src/mesa/drivers/dri/gamma/gamma_xmesa.c +++ b/src/mesa/drivers/dri/gamma/gamma_xmesa.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/gamma/gamma_xmesa.c,v 1.14 2002/10/30 12:51:30 alanh Exp $ */ /* * Copyright 2001 by Alan Hourihane. * diff --git a/src/mesa/drivers/dri/gamma/server/glint_common.h b/src/mesa/drivers/dri/gamma/server/glint_common.h index ec601f942d..36554e4ac2 100644 --- a/src/mesa/drivers/dri/gamma/server/glint_common.h +++ b/src/mesa/drivers/dri/gamma/server/glint_common.h @@ -25,7 +25,6 @@ * Converted to common header format: * Jens Owen * - * $XFree86: xc/programs/Xserver/hw/xfree86/drivers/glint/glint_common.h,v 1.2 2003/04/03 16:52:18 dawes Exp $ * */ diff --git a/src/mesa/drivers/dri/gamma/server/glint_dri.h b/src/mesa/drivers/dri/gamma/server/glint_dri.h index 3952759f83..df1992a5d1 100644 --- a/src/mesa/drivers/dri/gamma/server/glint_dri.h +++ b/src/mesa/drivers/dri/gamma/server/glint_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/glint/glint_dri.h,v 1.7 2002/10/30 12:52:16 alanh Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/i810/i810_3d_reg.h b/src/mesa/drivers/dri/i810/i810_3d_reg.h index 7cc59d5c86..2fbeb64978 100644 --- a/src/mesa/drivers/dri/i810/i810_3d_reg.h +++ b/src/mesa/drivers/dri/i810/i810_3d_reg.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810_3d_reg.h,v 1.7 2002/02/22 21:33:03 dawes Exp $ */ #ifndef I810_3D_REG_H #define I810_3D_REG_H diff --git a/src/mesa/drivers/dri/i810/i810context.c b/src/mesa/drivers/dri/i810/i810context.c index 3f7f2cc8a4..f90b3682f8 100644 --- a/src/mesa/drivers/dri/i810/i810context.c +++ b/src/mesa/drivers/dri/i810/i810context.c @@ -24,7 +24,6 @@ TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810context.c,v 1.3 2002/10/30 12:51:33 alanh Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/i810/i810context.h b/src/mesa/drivers/dri/i810/i810context.h index b83500bbd0..4708042059 100644 --- a/src/mesa/drivers/dri/i810/i810context.h +++ b/src/mesa/drivers/dri/i810/i810context.h @@ -21,7 +21,6 @@ * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810context.h,v 1.9 2002/12/16 16:18:51 dawes Exp $ */ #ifndef I810CONTEXT_INC #define I810CONTEXT_INC diff --git a/src/mesa/drivers/dri/i810/i810ioctl.c b/src/mesa/drivers/dri/i810/i810ioctl.c index 57c84193fa..95726fb252 100644 --- a/src/mesa/drivers/dri/i810/i810ioctl.c +++ b/src/mesa/drivers/dri/i810/i810ioctl.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810ioctl.c,v 1.7 2002/10/30 12:51:33 alanh Exp $ */ #include /* for usleep() */ diff --git a/src/mesa/drivers/dri/i810/i810ioctl.h b/src/mesa/drivers/dri/i810/i810ioctl.h index 61399ee7b7..748d29ae36 100644 --- a/src/mesa/drivers/dri/i810/i810ioctl.h +++ b/src/mesa/drivers/dri/i810/i810ioctl.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810ioctl.h,v 1.7 2002/10/30 12:51:33 alanh Exp $ */ #ifndef I810_IOCTL_H #define I810_IOCTL_H diff --git a/src/mesa/drivers/dri/i810/i810screen.c b/src/mesa/drivers/dri/i810/i810screen.c index f64c10a9ae..695b996319 100644 --- a/src/mesa/drivers/dri/i810/i810screen.c +++ b/src/mesa/drivers/dri/i810/i810screen.c @@ -24,7 +24,6 @@ TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810screen.c,v 1.2 2002/10/30 12:51:33 alanh Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/i810/i810state.c b/src/mesa/drivers/dri/i810/i810state.c index e0d5b2b487..e203c74f52 100644 --- a/src/mesa/drivers/dri/i810/i810state.c +++ b/src/mesa/drivers/dri/i810/i810state.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810state.c,v 1.9 2002/10/30 12:51:33 alanh Exp $ */ #include diff --git a/src/mesa/drivers/dri/i810/i810tex.c b/src/mesa/drivers/dri/i810/i810tex.c index f657abe671..730bc90eaf 100644 --- a/src/mesa/drivers/dri/i810/i810tex.c +++ b/src/mesa/drivers/dri/i810/i810tex.c @@ -21,7 +21,6 @@ * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810tex.c,v 1.9 2002/10/30 12:51:33 alanh Exp $ */ #include "glheader.h" #include "mtypes.h" diff --git a/src/mesa/drivers/dri/i810/i810tris.c b/src/mesa/drivers/dri/i810/i810tris.c index 2c4ee06633..40ab436b95 100644 --- a/src/mesa/drivers/dri/i810/i810tris.c +++ b/src/mesa/drivers/dri/i810/i810tris.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810tris.c,v 1.7 2002/10/30 12:51:33 alanh Exp $ */ /************************************************************************** Copyright 2001 VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/i810/i810tris.h b/src/mesa/drivers/dri/i810/i810tris.h index 06c8b3fcd5..3d0dd916ca 100644 --- a/src/mesa/drivers/dri/i810/i810tris.h +++ b/src/mesa/drivers/dri/i810/i810tris.h @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810tris.h,v 1.10 2002/02/22 21:33:04 dawes Exp $ */ #ifndef I810TRIS_INC #define I810TRIS_INC diff --git a/src/mesa/drivers/dri/i810/i810vb.c b/src/mesa/drivers/dri/i810/i810vb.c index 5ce98a991d..3439192b0d 100644 --- a/src/mesa/drivers/dri/i810/i810vb.c +++ b/src/mesa/drivers/dri/i810/i810vb.c @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810vb.c,v 1.13 2003/03/26 20:43:48 tsi Exp $ */ #include "glheader.h" diff --git a/src/mesa/drivers/dri/i810/i810vb.h b/src/mesa/drivers/dri/i810/i810vb.h index 1cced86ab2..55d0d2409e 100644 --- a/src/mesa/drivers/dri/i810/i810vb.h +++ b/src/mesa/drivers/dri/i810/i810vb.h @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/i810/i810vb.h,v 1.4 2002/02/22 21:33:04 dawes Exp $ */ #ifndef I810VB_INC #define I810VB_INC diff --git a/src/mesa/drivers/dri/i810/server/i810_common.h b/src/mesa/drivers/dri/i810/server/i810_common.h index 02e548be0e..29be444b45 100644 --- a/src/mesa/drivers/dri/i810/server/i810_common.h +++ b/src/mesa/drivers/dri/i810/server/i810_common.h @@ -25,7 +25,6 @@ * Converted to common header format: * Jens Owen * - * $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i810_common.h,v 1.1 2002/09/11 00:29:31 dawes Exp $ * */ diff --git a/src/mesa/drivers/dri/i810/server/i810_dri.h b/src/mesa/drivers/dri/i810/server/i810_dri.h index 408a4ebb4d..4a714f0306 100644 --- a/src/mesa/drivers/dri/i810/server/i810_dri.h +++ b/src/mesa/drivers/dri/i810/server/i810_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i810_dri.h,v 1.10 2002/12/10 01:27:04 dawes Exp $ */ #ifndef _I810_DRI_ #define _I810_DRI_ diff --git a/src/mesa/drivers/dri/i810/server/i810_reg.h b/src/mesa/drivers/dri/i810/server/i810_reg.h index c935982a78..e7e5081038 100644 --- a/src/mesa/drivers/dri/i810/server/i810_reg.h +++ b/src/mesa/drivers/dri/i810/server/i810_reg.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i810_reg.h,v 1.13 2003/02/06 04:18:04 dawes Exp $ */ /************************************************************************** Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/i915/server/i830_common.h b/src/mesa/drivers/dri/i915/server/i830_common.h index fb6ceaa52d..2b0fee82a8 100644 --- a/src/mesa/drivers/dri/i915/server/i830_common.h +++ b/src/mesa/drivers/dri/i915/server/i830_common.h @@ -26,7 +26,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ #ifndef _I830_COMMON_H_ #define _I830_COMMON_H_ diff --git a/src/mesa/drivers/dri/i915/server/i830_dri.h b/src/mesa/drivers/dri/i915/server/i830_dri.h index 6c9a709021..313eb759b0 100644 --- a/src/mesa/drivers/dri/i915/server/i830_dri.h +++ b/src/mesa/drivers/dri/i915/server/i830_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ #ifndef _I830_DRI_H #define _I830_DRI_H diff --git a/src/mesa/drivers/dri/i965/server/i830_common.h b/src/mesa/drivers/dri/i965/server/i830_common.h index f320378c2a..49eb145f8b 100644 --- a/src/mesa/drivers/dri/i965/server/i830_common.h +++ b/src/mesa/drivers/dri/i965/server/i830_common.h @@ -26,7 +26,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_common.h,v 1.1 2002/09/11 00:29:32 dawes Exp $ */ #ifndef _I830_COMMON_H_ #define _I830_COMMON_H_ diff --git a/src/mesa/drivers/dri/i965/server/i830_dri.h b/src/mesa/drivers/dri/i965/server/i830_dri.h index 22951812ad..68213f69f5 100644 --- a/src/mesa/drivers/dri/i965/server/i830_dri.h +++ b/src/mesa/drivers/dri/i965/server/i830_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/i810/i830_dri.h,v 1.4 2002/10/30 12:52:18 alanh Exp $ */ #ifndef _I830_DRI_H #define _I830_DRI_H diff --git a/src/mesa/drivers/dri/mach64/mach64_context.c b/src/mesa/drivers/dri/mach64/mach64_context.c index ad661e198c..7f558e92bc 100644 --- a/src/mesa/drivers/dri/mach64/mach64_context.c +++ b/src/mesa/drivers/dri/mach64/mach64_context.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_context.h b/src/mesa/drivers/dri/mach64/mach64_context.h index 8d89452412..e925f18c11 100644 --- a/src/mesa/drivers/dri/mach64/mach64_context.h +++ b/src/mesa/drivers/dri/mach64/mach64_context.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_dd.c b/src/mesa/drivers/dri/mach64/mach64_dd.c index 17e8d74d9f..7d225ebc88 100644 --- a/src/mesa/drivers/dri/mach64/mach64_dd.c +++ b/src/mesa/drivers/dri/mach64/mach64_dd.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_dd.h b/src/mesa/drivers/dri/mach64/mach64_dd.h index 74cf1d304f..0a2ce06412 100644 --- a/src/mesa/drivers/dri/mach64/mach64_dd.h +++ b/src/mesa/drivers/dri/mach64/mach64_dd.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_ioctl.c b/src/mesa/drivers/dri/mach64/mach64_ioctl.c index 36e7d3c5d3..6bc2b58ce9 100644 --- a/src/mesa/drivers/dri/mach64/mach64_ioctl.c +++ b/src/mesa/drivers/dri/mach64/mach64_ioctl.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_ioctl.h b/src/mesa/drivers/dri/mach64/mach64_ioctl.h index 52fe863484..2153ab80d7 100644 --- a/src/mesa/drivers/dri/mach64/mach64_ioctl.h +++ b/src/mesa/drivers/dri/mach64/mach64_ioctl.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_lock.c b/src/mesa/drivers/dri/mach64/mach64_lock.c index b73e350111..ea605fb061 100644 --- a/src/mesa/drivers/dri/mach64/mach64_lock.c +++ b/src/mesa/drivers/dri/mach64/mach64_lock.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_lock.h b/src/mesa/drivers/dri/mach64/mach64_lock.h index 973880ee25..3130b183e3 100644 --- a/src/mesa/drivers/dri/mach64/mach64_lock.h +++ b/src/mesa/drivers/dri/mach64/mach64_lock.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vb.c b/src/mesa/drivers/dri/mach64/mach64_native_vb.c index 248fa2a9a2..99f1a14e17 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vb.c +++ b/src/mesa/drivers/dri/mach64/mach64_native_vb.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Mesa 3-D graphics library * Version: 3.5 diff --git a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h index f64b808ee7..684f2acc89 100644 --- a/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_native_vbtmp.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Mesa 3-D graphics library * Version: 3.5 diff --git a/src/mesa/drivers/dri/mach64/mach64_reg.h b/src/mesa/drivers/dri/mach64/mach64_reg.h index abbba295a5..cb944e1023 100644 --- a/src/mesa/drivers/dri/mach64/mach64_reg.h +++ b/src/mesa/drivers/dri/mach64/mach64_reg.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_screen.c b/src/mesa/drivers/dri/mach64/mach64_screen.c index 4e9e216e7d..b780ba65ea 100644 --- a/src/mesa/drivers/dri/mach64/mach64_screen.c +++ b/src/mesa/drivers/dri/mach64/mach64_screen.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_screen.h b/src/mesa/drivers/dri/mach64/mach64_screen.h index 5305058e2f..7bf7dc474d 100644 --- a/src/mesa/drivers/dri/mach64/mach64_screen.h +++ b/src/mesa/drivers/dri/mach64/mach64_screen.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_span.c b/src/mesa/drivers/dri/mach64/mach64_span.c index 3830a28165..5c2403f587 100644 --- a/src/mesa/drivers/dri/mach64/mach64_span.c +++ b/src/mesa/drivers/dri/mach64/mach64_span.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_span.h b/src/mesa/drivers/dri/mach64/mach64_span.h index 0f4c766477..65141d05c3 100644 --- a/src/mesa/drivers/dri/mach64/mach64_span.h +++ b/src/mesa/drivers/dri/mach64/mach64_span.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_state.c b/src/mesa/drivers/dri/mach64/mach64_state.c index 667a394520..9ac51ee5b1 100644 --- a/src/mesa/drivers/dri/mach64/mach64_state.c +++ b/src/mesa/drivers/dri/mach64/mach64_state.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_state.h b/src/mesa/drivers/dri/mach64/mach64_state.h index 95bcab3653..23081cb2fe 100644 --- a/src/mesa/drivers/dri/mach64/mach64_state.h +++ b/src/mesa/drivers/dri/mach64/mach64_state.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_tex.c b/src/mesa/drivers/dri/mach64/mach64_tex.c index 5288d321ce..c42588e064 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tex.c +++ b/src/mesa/drivers/dri/mach64/mach64_tex.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_tex.h b/src/mesa/drivers/dri/mach64/mach64_tex.h index f6cf1cf802..e67661b970 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tex.h +++ b/src/mesa/drivers/dri/mach64/mach64_tex.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_texmem.c b/src/mesa/drivers/dri/mach64/mach64_texmem.c index 3b7b93b984..d65b2cda6a 100644 --- a/src/mesa/drivers/dri/mach64/mach64_texmem.c +++ b/src/mesa/drivers/dri/mach64/mach64_texmem.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., * Cedar Park, Texas. diff --git a/src/mesa/drivers/dri/mach64/mach64_texstate.c b/src/mesa/drivers/dri/mach64/mach64_texstate.c index 3ace370d70..80c84d6774 100644 --- a/src/mesa/drivers/dri/mach64/mach64_texstate.c +++ b/src/mesa/drivers/dri/mach64/mach64_texstate.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_tris.c b/src/mesa/drivers/dri/mach64/mach64_tris.c index 369f610442..e4df01106d 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tris.c +++ b/src/mesa/drivers/dri/mach64/mach64_tris.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_tris.h b/src/mesa/drivers/dri/mach64/mach64_tris.h index 208703289d..4780765a18 100644 --- a/src/mesa/drivers/dri/mach64/mach64_tris.h +++ b/src/mesa/drivers/dri/mach64/mach64_tris.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_vb.c b/src/mesa/drivers/dri/mach64/mach64_vb.c index 83a5f73e6b..8aab72a3f3 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vb.c +++ b/src/mesa/drivers/dri/mach64/mach64_vb.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_vb.h b/src/mesa/drivers/dri/mach64/mach64_vb.h index bcc4759af3..0d923abce0 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vb.h +++ b/src/mesa/drivers/dri/mach64/mach64_vb.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h index c1207cacd1..938804af9e 100644 --- a/src/mesa/drivers/dri/mach64/mach64_vbtmp.h +++ b/src/mesa/drivers/dri/mach64/mach64_vbtmp.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Mesa 3-D graphics library * Version: 3.5 diff --git a/src/mesa/drivers/dri/mach64/server/mach64_dri.h b/src/mesa/drivers/dri/mach64/server/mach64_dri.h index 139668e3f3..1477443f79 100644 --- a/src/mesa/drivers/dri/mach64/server/mach64_dri.h +++ b/src/mesa/drivers/dri/mach64/server/mach64_dri.h @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- mode: c; c-basic-offset: 3 -*- */ +/* -*- mode: c; c-basic-offset: 3 -*- */ /* * Copyright 2000 Gareth Hughes * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mga_texstate.c b/src/mesa/drivers/dri/mga/mga_texstate.c index 71d264b0f1..c14ddc95c9 100644 --- a/src/mesa/drivers/dri/mga/mga_texstate.c +++ b/src/mesa/drivers/dri/mga/mga_texstate.c @@ -26,7 +26,6 @@ * Ian Romanick * Keith Whitwell */ -/* $XFree86:$ */ #include #include "mm.h" diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.c b/src/mesa/drivers/dri/mga/mga_xmesa.c index f4e651afa0..6148f6b488 100644 --- a/src/mesa/drivers/dri/mga/mga_xmesa.c +++ b/src/mesa/drivers/dri/mga/mga_xmesa.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mga_xmesa.c,v 1.19 2003/03/26 20:43:49 tsi Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.h b/src/mesa/drivers/dri/mga/mga_xmesa.h index 0ab0c63f78..0f81c9cbec 100644 --- a/src/mesa/drivers/dri/mga/mga_xmesa.h +++ b/src/mesa/drivers/dri/mga/mga_xmesa.h @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mga_xmesa.h,v 1.12 2002/12/16 16:18:52 dawes Exp $ */ #ifndef _MGA_INIT_H_ #define _MGA_INIT_H_ diff --git a/src/mesa/drivers/dri/mga/mgacontext.h b/src/mesa/drivers/dri/mga/mgacontext.h index 2124006ade..6aa92355b8 100644 --- a/src/mesa/drivers/dri/mga/mgacontext.h +++ b/src/mesa/drivers/dri/mga/mgacontext.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgacontext.h,v 1.7 2002/12/16 16:18:52 dawes Exp $*/ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgadd.c b/src/mesa/drivers/dri/mga/mgadd.c index b1d5e0c48f..04336b5ac7 100644 --- a/src/mesa/drivers/dri/mga/mgadd.c +++ b/src/mesa/drivers/dri/mga/mgadd.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgadd.c,v 1.14 2002/10/30 12:51:35 alanh Exp $ */ #include "mtypes.h" diff --git a/src/mesa/drivers/dri/mga/mgadd.h b/src/mesa/drivers/dri/mga/mgadd.h index f98bfdc878..6830ca67ad 100644 --- a/src/mesa/drivers/dri/mga/mgadd.h +++ b/src/mesa/drivers/dri/mga/mgadd.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgadd.h,v 1.3 2002/10/30 12:51:35 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgaioctl.h b/src/mesa/drivers/dri/mga/mgaioctl.h index f3ae749ca9..9aa08c5158 100644 --- a/src/mesa/drivers/dri/mga/mgaioctl.h +++ b/src/mesa/drivers/dri/mga/mgaioctl.h @@ -25,7 +25,6 @@ * Keith Whitwell * Gareth Hughes */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgaioctl.h,v 1.11 2002/10/30 12:51:36 alanh Exp $ */ #ifndef MGA_IOCTL_H #define MGA_IOCTL_H diff --git a/src/mesa/drivers/dri/mga/mgapixel.c b/src/mesa/drivers/dri/mga/mgapixel.c index 2b9da8c181..f309aabbc8 100644 --- a/src/mesa/drivers/dri/mga/mgapixel.c +++ b/src/mesa/drivers/dri/mga/mgapixel.c @@ -34,7 +34,6 @@ * \author Keith Whitwell * \author Gareth Hughes */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgapixel.c,v 1.9 2002/11/05 17:46:08 tsi Exp $ */ #include "mtypes.h" #include "macros.h" diff --git a/src/mesa/drivers/dri/mga/mgapixel.h b/src/mesa/drivers/dri/mga/mgapixel.h index c44fd769a8..b52c8670f3 100644 --- a/src/mesa/drivers/dri/mga/mgapixel.h +++ b/src/mesa/drivers/dri/mga/mgapixel.h @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgapixel.h,v 1.5 2002/10/30 12:51:36 alanh Exp $ */ #ifndef MGA_PIXELS_H #define MGA_PIXELS_H diff --git a/src/mesa/drivers/dri/mga/mgaregs.h b/src/mesa/drivers/dri/mga/mgaregs.h index e1291ca01b..1ef1e6d24c 100644 --- a/src/mesa/drivers/dri/mga/mgaregs.h +++ b/src/mesa/drivers/dri/mga/mgaregs.h @@ -19,7 +19,6 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE * OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgaregs.h,v 1.6 2003/01/12 03:55:46 tsi Exp $ */ #ifndef _MGAREGS_H_ #define _MGAREGS_H_ diff --git a/src/mesa/drivers/dri/mga/mgarender.c b/src/mesa/drivers/dri/mga/mgarender.c index 3080cea79f..c9e42a8040 100644 --- a/src/mesa/drivers/dri/mga/mgarender.c +++ b/src/mesa/drivers/dri/mga/mgarender.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgarender.c,v 1.4 2002/10/30 12:51:36 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/mga/mgaspan.h b/src/mesa/drivers/dri/mga/mgaspan.h index f133a51c08..f5e2e49b8a 100644 --- a/src/mesa/drivers/dri/mga/mgaspan.h +++ b/src/mesa/drivers/dri/mga/mgaspan.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgaspan.h,v 1.3 2002/10/30 12:51:36 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgastate.h b/src/mesa/drivers/dri/mga/mgastate.h index afbe0aaf90..ec65d4e6cd 100644 --- a/src/mesa/drivers/dri/mga/mgastate.h +++ b/src/mesa/drivers/dri/mga/mgastate.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgastate.h,v 1.5 2002/10/30 12:51:36 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgatex.c b/src/mesa/drivers/dri/mga/mgatex.c index a7d74317a5..31ea5046df 100644 --- a/src/mesa/drivers/dri/mga/mgatex.c +++ b/src/mesa/drivers/dri/mga/mgatex.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatex.c,v 1.14 2002/10/30 12:51:36 alanh Exp $ */ #include "glheader.h" #include "mm.h" diff --git a/src/mesa/drivers/dri/mga/mgatex.h b/src/mesa/drivers/dri/mga/mgatex.h index fb7ffcff16..789034964a 100644 --- a/src/mesa/drivers/dri/mga/mgatex.h +++ b/src/mesa/drivers/dri/mga/mgatex.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatex.h,v 1.7 2002/10/30 12:51:36 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/mgatexmem.c b/src/mesa/drivers/dri/mga/mgatexmem.c index 18743331c6..559813f5de 100644 --- a/src/mesa/drivers/dri/mga/mgatexmem.c +++ b/src/mesa/drivers/dri/mga/mgatexmem.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatexmem.c,v 1.7 2002/10/30 12:51:36 alanh Exp $ */ #include "glheader.h" diff --git a/src/mesa/drivers/dri/mga/mgatris.c b/src/mesa/drivers/dri/mga/mgatris.c index 91b413ae76..0c8081cfb9 100644 --- a/src/mesa/drivers/dri/mga/mgatris.c +++ b/src/mesa/drivers/dri/mga/mgatris.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatris.c,v 1.10 2002/10/30 12:51:36 alanh Exp $ */ #include "mtypes.h" #include "macros.h" diff --git a/src/mesa/drivers/dri/mga/mgatris.h b/src/mesa/drivers/dri/mga/mgatris.h index f3ece3a053..a40fef8307 100644 --- a/src/mesa/drivers/dri/mga/mgatris.h +++ b/src/mesa/drivers/dri/mga/mgatris.h @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgatris.h,v 1.10 2002/10/30 12:51:36 alanh Exp $ */ #ifndef MGATRIS_INC #define MGATRIS_INC diff --git a/src/mesa/drivers/dri/mga/mgavb.c b/src/mesa/drivers/dri/mga/mgavb.c index 902d8bd1c1..954fd53ae3 100644 --- a/src/mesa/drivers/dri/mga/mgavb.c +++ b/src/mesa/drivers/dri/mga/mgavb.c @@ -24,7 +24,6 @@ * Authors: * Keith Whitwell */ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgavb.c,v 1.15 2003/03/26 20:43:49 tsi Exp $ */ #include #include "mgacontext.h" diff --git a/src/mesa/drivers/dri/mga/mgavb.h b/src/mesa/drivers/dri/mga/mgavb.h index 5f6454aca9..f6580e0db9 100644 --- a/src/mesa/drivers/dri/mga/mgavb.h +++ b/src/mesa/drivers/dri/mga/mgavb.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/mga/mgavb.h,v 1.8 2002/10/30 12:51:36 alanh Exp $ */ /* * Copyright 2000-2001 VA Linux Systems, Inc. * All Rights Reserved. diff --git a/src/mesa/drivers/dri/mga/server/mga.h b/src/mesa/drivers/dri/mga/server/mga.h index 830d48d859..d7790e4779 100644 --- a/src/mesa/drivers/dri/mga/server/mga.h +++ b/src/mesa/drivers/dri/mga/server/mga.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga.h,v 1.85 2002/12/16 16:19:17 dawes Exp $ */ /* * MGA Millennium (MGA2064W) functions * diff --git a/src/mesa/drivers/dri/mga/server/mga_bios.h b/src/mesa/drivers/dri/mga/server/mga_bios.h index 8fbf619e34..5dcfc1614d 100644 --- a/src/mesa/drivers/dri/mga/server/mga_bios.h +++ b/src/mesa/drivers/dri/mga/server/mga_bios.h @@ -1,8 +1,6 @@ -/* $XConsortium: mga_bios.h /main/2 1996/10/28 04:48:23 kaleb $ */ #ifndef MGA_BIOS_H #define MGA_BIOS_H -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_bios.h,v 1.3 1998/07/25 16:55:51 dawes Exp $ */ /* * MGABiosInfo - This struct describes the video BIOS info block. diff --git a/src/mesa/drivers/dri/mga/server/mga_dri.c b/src/mesa/drivers/dri/mga/server/mga_dri.c index 258ace83a0..bc575e62ee 100644 --- a/src/mesa/drivers/dri/mga/server/mga_dri.c +++ b/src/mesa/drivers/dri/mga/server/mga_dri.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_dri.c,v 1.28 2003/02/08 21:26:58 dawes Exp $ */ /* * Copyright 2000 VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/mga/server/mga_dri.h b/src/mesa/drivers/dri/mga/server/mga_dri.h index 03b8414603..1ce07028f1 100644 --- a/src/mesa/drivers/dri/mga/server/mga_dri.h +++ b/src/mesa/drivers/dri/mga/server/mga_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_dri.h,v 1.8 2002/11/29 11:06:42 eich Exp $ */ /* * Copyright 2000 VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/mga/server/mga_macros.h b/src/mesa/drivers/dri/mga/server/mga_macros.h index d985081ab6..189e1415d0 100644 --- a/src/mesa/drivers/dri/mga/server/mga_macros.h +++ b/src/mesa/drivers/dri/mga/server/mga_macros.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_macros.h,v 1.22 2002/02/20 17:17:50 dawes Exp $ */ #ifndef _MGA_MACROS_H_ #define _MGA_MACROS_H_ diff --git a/src/mesa/drivers/dri/mga/server/mga_reg.h b/src/mesa/drivers/dri/mga/server/mga_reg.h index b8e3499235..d51366d44e 100644 --- a/src/mesa/drivers/dri/mga/server/mga_reg.h +++ b/src/mesa/drivers/dri/mga/server/mga_reg.h @@ -1,8 +1,6 @@ -/* $XConsortium: mgareg.h /main/2 1996/10/25 10:33:21 kaleb $ */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/mga/mga_reg.h,v 1.18 2001/09/26 12:59:18 alanh Exp $ */ diff --git a/src/mesa/drivers/dri/r128/r128_context.c b/src/mesa/drivers/dri/r128/r128_context.c index 95e54a6af5..dfc89a2da7 100644 --- a/src/mesa/drivers/dri/r128/r128_context.c +++ b/src/mesa/drivers/dri/r128/r128_context.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_context.c,v 1.8 2002/10/30 12:51:38 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_context.h b/src/mesa/drivers/dri/r128/r128_context.h index c51dd7fa58..3f96836df1 100644 --- a/src/mesa/drivers/dri/r128/r128_context.h +++ b/src/mesa/drivers/dri/r128/r128_context.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_context.h,v 1.12 2002/12/16 16:18:52 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_dd.c b/src/mesa/drivers/dri/r128/r128_dd.c index 54f2b21b5d..d8e1c70ab7 100644 --- a/src/mesa/drivers/dri/r128/r128_dd.c +++ b/src/mesa/drivers/dri/r128/r128_dd.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_dd.c,v 1.15 2002/10/30 12:51:38 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_dd.h b/src/mesa/drivers/dri/r128/r128_dd.h index 7a0abb73f8..ce038853c4 100644 --- a/src/mesa/drivers/dri/r128/r128_dd.h +++ b/src/mesa/drivers/dri/r128/r128_dd.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_dd.h,v 1.3 2001/01/08 01:07:20 martin Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_ioctl.c b/src/mesa/drivers/dri/r128/r128_ioctl.c index b0dba7d04e..25188061a0 100644 --- a/src/mesa/drivers/dri/r128/r128_ioctl.c +++ b/src/mesa/drivers/dri/r128/r128_ioctl.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_ioctl.c,v 1.10 2002/12/16 16:18:53 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_ioctl.h b/src/mesa/drivers/dri/r128/r128_ioctl.h index 95779f09be..57063c41f5 100644 --- a/src/mesa/drivers/dri/r128/r128_ioctl.h +++ b/src/mesa/drivers/dri/r128/r128_ioctl.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_ioctl.h,v 1.6 2002/12/16 16:18:53 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_lock.c b/src/mesa/drivers/dri/r128/r128_lock.c index ea23b007f3..3478e12ad0 100644 --- a/src/mesa/drivers/dri/r128/r128_lock.c +++ b/src/mesa/drivers/dri/r128/r128_lock.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_lock.c,v 1.5 2002/10/30 12:51:38 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_lock.h b/src/mesa/drivers/dri/r128/r128_lock.h index 39bdde9820..1fc8cbe29f 100644 --- a/src/mesa/drivers/dri/r128/r128_lock.h +++ b/src/mesa/drivers/dri/r128/r128_lock.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_lock.h,v 1.4 2001/01/08 01:07:21 martin Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_screen.c b/src/mesa/drivers/dri/r128/r128_screen.c index 880dee85c2..0722b80ee5 100644 --- a/src/mesa/drivers/dri/r128/r128_screen.c +++ b/src/mesa/drivers/dri/r128/r128_screen.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_screen.c,v 1.9 2003/03/26 20:43:49 tsi Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_screen.h b/src/mesa/drivers/dri/r128/r128_screen.h index 8db8eea358..b31e87661b 100644 --- a/src/mesa/drivers/dri/r128/r128_screen.h +++ b/src/mesa/drivers/dri/r128/r128_screen.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_screen.h,v 1.7 2002/12/16 16:18:53 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_span.c b/src/mesa/drivers/dri/r128/r128_span.c index 85798c1601..c5b6480db9 100644 --- a/src/mesa/drivers/dri/r128/r128_span.c +++ b/src/mesa/drivers/dri/r128/r128_span.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_span.c,v 1.8 2002/10/30 12:51:39 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_span.h b/src/mesa/drivers/dri/r128/r128_span.h index fd7c2d1394..9af4058129 100644 --- a/src/mesa/drivers/dri/r128/r128_span.h +++ b/src/mesa/drivers/dri/r128/r128_span.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_span.h,v 1.3 2001/01/08 01:07:21 martin Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_state.c b/src/mesa/drivers/dri/r128/r128_state.c index e476afa5d8..58c3a27ee8 100644 --- a/src/mesa/drivers/dri/r128/r128_state.c +++ b/src/mesa/drivers/dri/r128/r128_state.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_state.c,v 1.11 2002/10/30 12:51:39 alanh Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_state.h b/src/mesa/drivers/dri/r128/r128_state.h index 6f0a6a6557..a44327dfb3 100644 --- a/src/mesa/drivers/dri/r128/r128_state.h +++ b/src/mesa/drivers/dri/r128/r128_state.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_state.h,v 1.3 2001/01/08 01:07:21 martin Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_tex.c b/src/mesa/drivers/dri/r128/r128_tex.c index 3b2d017c1f..554a92287f 100644 --- a/src/mesa/drivers/dri/r128/r128_tex.c +++ b/src/mesa/drivers/dri/r128/r128_tex.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tex.c,v 1.14 2002/11/05 17:46:08 tsi Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_tex.h b/src/mesa/drivers/dri/r128/r128_tex.h index 54053b8b31..994dffb5a9 100644 --- a/src/mesa/drivers/dri/r128/r128_tex.h +++ b/src/mesa/drivers/dri/r128/r128_tex.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tex.h,v 1.7 2002/02/22 21:44:58 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_texmem.c b/src/mesa/drivers/dri/r128/r128_texmem.c index d011a75671..a7d0280636 100644 --- a/src/mesa/drivers/dri/r128/r128_texmem.c +++ b/src/mesa/drivers/dri/r128/r128_texmem.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_texmem.c,v 1.1 2002/02/22 21:44:58 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_texobj.h b/src/mesa/drivers/dri/r128/r128_texobj.h index 282e887149..08eac87758 100644 --- a/src/mesa/drivers/dri/r128/r128_texobj.h +++ b/src/mesa/drivers/dri/r128/r128_texobj.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_texobj.h,v 1.5 2002/02/22 21:44:58 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_texstate.c b/src/mesa/drivers/dri/r128/r128_texstate.c index 6b43f21cd4..211b9ea2a9 100644 --- a/src/mesa/drivers/dri/r128/r128_texstate.c +++ b/src/mesa/drivers/dri/r128/r128_texstate.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_texstate.c,v 1.1 2002/02/22 21:44:58 dawes Exp $ */ /************************************************************************** Copyright 1999, 2000 ATI Technologies Inc. and Precision Insight, Inc., diff --git a/src/mesa/drivers/dri/r128/r128_tris.c b/src/mesa/drivers/dri/r128/r128_tris.c index f406e928c5..f2f124360c 100644 --- a/src/mesa/drivers/dri/r128/r128_tris.c +++ b/src/mesa/drivers/dri/r128/r128_tris.c @@ -1,4 +1,4 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tris.c,v 1.8 2002/10/30 12:51:43 alanh Exp $ */ /* -*- c-basic-offset: 3 -*- */ +/* -*- c-basic-offset: 3 -*- */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/r128/r128_tris.h b/src/mesa/drivers/dri/r128/r128_tris.h index 755d3320b0..c8f0a4809b 100644 --- a/src/mesa/drivers/dri/r128/r128_tris.h +++ b/src/mesa/drivers/dri/r128/r128_tris.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tris.h,v 1.8 2002/10/30 12:51:43 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/r128/server/r128.h b/src/mesa/drivers/dri/r128/server/r128.h index ce98b1b915..ca08d7c86a 100644 --- a/src/mesa/drivers/dri/r128/server/r128.h +++ b/src/mesa/drivers/dri/r128/server/r128.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128.h,v 1.24 2002/12/16 16:19:10 dawes Exp $ */ /* * Copyright 1999, 2000 ATI Technologies Inc., Markham, Ontario, * Precision Insight, Inc., Cedar Park, Texas, and diff --git a/src/mesa/drivers/dri/r128/server/r128_dri.c b/src/mesa/drivers/dri/r128/server/r128_dri.c index 5edf1e1003..efe9232dc2 100644 --- a/src/mesa/drivers/dri/r128/server/r128_dri.c +++ b/src/mesa/drivers/dri/r128/server/r128_dri.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128_dri.c,v 1.28 2003/02/07 20:41:14 martin Exp $ */ /* * Copyright 1999, 2000 ATI Technologies Inc., Markham, Ontario, * Precision Insight, Inc., Cedar Park, Texas, and diff --git a/src/mesa/drivers/dri/r128/server/r128_dri.h b/src/mesa/drivers/dri/r128/server/r128_dri.h index 67ade70de4..430e5f580b 100644 --- a/src/mesa/drivers/dri/r128/server/r128_dri.h +++ b/src/mesa/drivers/dri/r128/server/r128_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128_dri.h,v 1.7 2002/10/30 12:52:12 alanh Exp $ */ /* * Copyright 1999, 2000 ATI Technologies Inc., Markham, Ontario, * Precision Insight, Inc., Cedar Park, Texas, and diff --git a/src/mesa/drivers/dri/r128/server/r128_macros.h b/src/mesa/drivers/dri/r128/server/r128_macros.h index 93b7feb02c..f7b945da93 100644 --- a/src/mesa/drivers/dri/r128/server/r128_macros.h +++ b/src/mesa/drivers/dri/r128/server/r128_macros.h @@ -35,7 +35,6 @@ * DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/R128_reg.h,v 1.20 2002/10/12 01:38:07 martin Exp $ */ #ifndef _R128_MACROS_H_ #define _R128_MACROS_H_ diff --git a/src/mesa/drivers/dri/r128/server/r128_reg.h b/src/mesa/drivers/dri/r128/server/r128_reg.h index 5669452d74..50033540b9 100644 --- a/src/mesa/drivers/dri/r128/server/r128_reg.h +++ b/src/mesa/drivers/dri/r128/server/r128_reg.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128_reg.h,v 1.15 2002/12/16 16:19:11 dawes Exp $ */ /* * Copyright 1999, 2000 ATI Technologies Inc., Markham, Ontario, * Precision Insight, Inc., Cedar Park, Texas, and diff --git a/src/mesa/drivers/dri/r128/server/r128_version.h b/src/mesa/drivers/dri/r128/server/r128_version.h index 589d8d40bc..783711ef97 100644 --- a/src/mesa/drivers/dri/r128/server/r128_version.h +++ b/src/mesa/drivers/dri/r128/server/r128_version.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/r128_version.h,v 1.6 2003/01/01 19:16:35 tsi Exp $ */ /* * Copyright 2000 through 2003 by Marc Aurele La France (TSI @ UQV), tsi@xfree86.org * diff --git a/src/mesa/drivers/dri/radeon/radeon_compat.c b/src/mesa/drivers/dri/radeon/radeon_compat.c index 1cbe3407ba..bd467fb15b 100644 --- a/src/mesa/drivers/dri/radeon/radeon_compat.c +++ b/src/mesa/drivers/dri/radeon/radeon_compat.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2002 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c index 9451ec4aa5..ba93a054ae 100644 --- a/src/mesa/drivers/dri/radeon/radeon_context.c +++ b/src/mesa/drivers/dri/radeon/radeon_context.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_context.c,v 1.9 2003/09/24 02:43:12 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index 4c64bc201a..f7e461239e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_ioctl.c,v 1.11 2003/01/29 22:04:59 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.h b/src/mesa/drivers/dri/radeon/radeon_ioctl.h index 11a7d02b1b..020a5c21e2 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.h +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_ioctl.h,v 1.6 2002/12/16 16:18:58 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_lighting.c b/src/mesa/drivers/dri/radeon/radeon_lighting.c index 44e00af0ef..5e9b9c3051 100644 --- a/src/mesa/drivers/dri/radeon/radeon_lighting.c +++ b/src/mesa/drivers/dri/radeon/radeon_lighting.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_state.c,v 1.5 2002/09/16 18:05:20 eich Exp $ */ /* * Copyright 2000, 2001 VA Linux Systems Inc., Fremont, California. * diff --git a/src/mesa/drivers/dri/radeon/radeon_maos.h b/src/mesa/drivers/dri/radeon/radeon_maos.h index 09039d6840..b8935e84a0 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos.h +++ b/src/mesa/drivers/dri/radeon/radeon_maos.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_maos.h,v 1.1 2002/10/30 12:51:55 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c index 49118b5e37..b61f5e0f3e 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_arrays.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_maos_arrays.c,v 1.1 2002/10/30 12:51:55 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c index 65dbecf7a6..d5ceedfa24 100644 --- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c +++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_maos_verts.c,v 1.1 2002/10/30 12:51:55 alanh Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_sanity.c b/src/mesa/drivers/dri/radeon/radeon_sanity.c index 557057784c..bdfb7240d7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_sanity.c +++ b/src/mesa/drivers/dri/radeon/radeon_sanity.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_sanity.c,v 1.1 2002/10/30 12:51:55 alanh Exp $ */ /************************************************************************** Copyright 2002 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index aa7fb633dd..4a45948608 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_screen.c,v 1.7 2003/03/26 20:43:51 tsi Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.h b/src/mesa/drivers/dri/radeon/radeon_screen.h index 25e6fcf399..f8c0cc96df 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.h +++ b/src/mesa/drivers/dri/radeon/radeon_screen.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_screen.h,v 1.5 2002/12/16 16:18:58 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c index 4de05c7697..856d27df75 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.c +++ b/src/mesa/drivers/dri/radeon/radeon_state.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_state.c,v 1.8 2002/12/16 16:18:58 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/radeon/radeon_state.h b/src/mesa/drivers/dri/radeon/radeon_state.h index ad7db3b677..2171879f75 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.h +++ b/src/mesa/drivers/dri/radeon/radeon_state.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_state.h,v 1.5 2002/11/05 17:46:09 tsi Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c index 5fc34f0933..c876a596e6 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state_init.c +++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_state_init.c,v 1.3 2003/02/22 06:21:11 dawes Exp $ */ /* * Copyright 2000, 2001 VA Linux Systems Inc., Fremont, California. * diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c index 7ce1fa67cf..2b3ae14ff7 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_swtcl.c,v 1.6 2003/05/06 23:52:08 daenzer Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.h b/src/mesa/drivers/dri/radeon/radeon_swtcl.h index 64f9019513..1feedf185d 100644 --- a/src/mesa/drivers/dri/radeon/radeon_swtcl.h +++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.h @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c index 5ad044c262..d35be1ca88 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.c +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.h b/src/mesa/drivers/dri/radeon/radeon_tcl.h index 168ab958a2..dccbea5fdb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tcl.h +++ b/src/mesa/drivers/dri/radeon/radeon_tcl.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_tcl.h,v 1.2 2003/02/08 21:26:45 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.c b/src/mesa/drivers/dri/radeon/radeon_tex.c index edaea6c209..f3eb9d8eef 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.c +++ b/src/mesa/drivers/dri/radeon/radeon_tex.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_tex.c,v 1.6 2002/09/16 18:05:20 eich Exp $ */ /* Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/radeon/radeon_tex.h b/src/mesa/drivers/dri/radeon/radeon_tex.h index a806981ae6..bdf086dfee 100644 --- a/src/mesa/drivers/dri/radeon/radeon_tex.h +++ b/src/mesa/drivers/dri/radeon/radeon_tex.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_tex.h,v 1.3 2002/02/22 21:45:01 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_texmem.c b/src/mesa/drivers/dri/radeon/radeon_texmem.c index 20f25dd34b..f7520f1dea 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texmem.c +++ b/src/mesa/drivers/dri/radeon/radeon_texmem.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_texmem.c,v 1.7 2002/12/16 16:18:59 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/radeon_texstate.c b/src/mesa/drivers/dri/radeon/radeon_texstate.c index 37bb749223..ae8d527cf4 100644 --- a/src/mesa/drivers/dri/radeon/radeon_texstate.c +++ b/src/mesa/drivers/dri/radeon/radeon_texstate.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_texstate.c,v 1.6 2002/12/16 16:18:59 dawes Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/radeon/server/radeon.h b/src/mesa/drivers/dri/radeon/server/radeon.h index 6f6c2e6d25..3fb1e37c53 100644 --- a/src/mesa/drivers/dri/radeon/server/radeon.h +++ b/src/mesa/drivers/dri/radeon/server/radeon.h @@ -31,7 +31,6 @@ * DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/radeon.h,v 1.29 2002/10/12 01:38:07 martin Exp $ */ #ifndef _RADEON_H_ #define _RADEON_H_ diff --git a/src/mesa/drivers/dri/radeon/server/radeon_dri.h b/src/mesa/drivers/dri/radeon/server/radeon_dri.h index ecd5323339..dc51372107 100644 --- a/src/mesa/drivers/dri/radeon/server/radeon_dri.h +++ b/src/mesa/drivers/dri/radeon/server/radeon_dri.h @@ -34,7 +34,6 @@ * DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/radeon_dri.h,v 1.3 2002/04/24 16:20:40 martin Exp $ */ #ifndef _RADEON_DRI_ #define _RADEON_DRI_ diff --git a/src/mesa/drivers/dri/radeon/server/radeon_macros.h b/src/mesa/drivers/dri/radeon/server/radeon_macros.h index 60f0fa2d35..355262c9ba 100644 --- a/src/mesa/drivers/dri/radeon/server/radeon_macros.h +++ b/src/mesa/drivers/dri/radeon/server/radeon_macros.h @@ -35,7 +35,6 @@ * DEALINGS IN THE SOFTWARE. */ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/radeon_reg.h,v 1.20 2002/10/12 01:38:07 martin Exp $ */ #ifndef _RADEON_MACROS_H_ #define _RADEON_MACROS_H_ diff --git a/src/mesa/drivers/dri/radeon/server/radeon_reg.h b/src/mesa/drivers/dri/radeon/server/radeon_reg.h index 4dcce63846..596a8aa715 100644 --- a/src/mesa/drivers/dri/radeon/server/radeon_reg.h +++ b/src/mesa/drivers/dri/radeon/server/radeon_reg.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/ati/radeon_reg.h,v 1.30 2003/10/07 22:47:12 martin Exp $ */ /* * Copyright 2000 ATI Technologies Inc., Markham, Ontario, and * VA Linux Systems Inc., Fremont, California. diff --git a/src/mesa/drivers/dri/savage/savagetris.c b/src/mesa/drivers/dri/savage/savagetris.c index 4ce2f60b4f..52c7f5fa76 100644 --- a/src/mesa/drivers/dri/savage/savagetris.c +++ b/src/mesa/drivers/dri/savage/savagetris.c @@ -1,4 +1,4 @@ -/* $XFree86$ */ /* -*- c-basic-offset: 3 -*- */ +/* -*- c-basic-offset: 3 -*- */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/savage/savagetris.h b/src/mesa/drivers/dri/savage/savagetris.h index 00803e7ff3..b2b3d951c6 100644 --- a/src/mesa/drivers/dri/savage/savagetris.h +++ b/src/mesa/drivers/dri/savage/savagetris.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tris.h,v 1.4 2001/01/08 01:07:24 martin Exp $ */ /************************************************************************** Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and diff --git a/src/mesa/drivers/dri/sis/server/sis_common.h b/src/mesa/drivers/dri/sis/server/sis_common.h index cbddf0c737..04f9ccdbf8 100644 --- a/src/mesa/drivers/dri/sis/server/sis_common.h +++ b/src/mesa/drivers/dri/sis/server/sis_common.h @@ -1,4 +1,4 @@ -/* * $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sis/sis_common.h,v 1.1 2003/08/29 08:52:12 twini Exp $ */ +/* * */ /* * Common header definitions for SiS 2D/3D/DRM suite * diff --git a/src/mesa/drivers/dri/sis/server/sis_dri.h b/src/mesa/drivers/dri/sis/server/sis_dri.h index a05662430e..f0171f3c0f 100644 --- a/src/mesa/drivers/dri/sis/server/sis_dri.h +++ b/src/mesa/drivers/dri/sis/server/sis_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/sis/sis_dri.h,v 1.9 2003/08/29 08:50:54 twini Exp $ */ /* modified from tdfx_dri.h */ diff --git a/src/mesa/drivers/dri/sis/sis_alloc.c b/src/mesa/drivers/dri/sis/sis_alloc.c index b696eeb51a..4ca4052803 100644 --- a/src/mesa/drivers/dri/sis/sis_alloc.c +++ b/src/mesa/drivers/dri/sis/sis_alloc.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_alloc.c,v 1.7 2001/01/08 01:07:29 martin Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_alloc.h b/src/mesa/drivers/dri/sis/sis_alloc.h index e76fc53fe2..eb784afad9 100644 --- a/src/mesa/drivers/dri/sis/sis_alloc.h +++ b/src/mesa/drivers/dri/sis/sis_alloc.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_clear.c b/src/mesa/drivers/dri/sis/sis_clear.c index fb92d06c73..174f3c0768 100644 --- a/src/mesa/drivers/dri/sis/sis_clear.c +++ b/src/mesa/drivers/dri/sis/sis_clear.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_clear.c,v 1.5 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_context.c b/src/mesa/drivers/dri/sis/sis_context.c index b21df0a61e..04c7464c5e 100644 --- a/src/mesa/drivers/dri/sis/sis_context.c +++ b/src/mesa/drivers/dri/sis/sis_context.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_ctx.c,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_context.h b/src/mesa/drivers/dri/sis/sis_context.h index c349bf96ed..b81812d6ce 100644 --- a/src/mesa/drivers/dri/sis/sis_context.h +++ b/src/mesa/drivers/dri/sis/sis_context.h @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_dd.c b/src/mesa/drivers/dri/sis/sis_dd.c index 8fc7896b87..989c159a80 100644 --- a/src/mesa/drivers/dri/sis/sis_dd.c +++ b/src/mesa/drivers/dri/sis/sis_dd.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_ctx.c,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_dd.h b/src/mesa/drivers/dri/sis/sis_dd.h index da76596e92..b141243a59 100644 --- a/src/mesa/drivers/dri/sis/sis_dd.h +++ b/src/mesa/drivers/dri/sis/sis_dd.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_fog.c b/src/mesa/drivers/dri/sis/sis_fog.c index fe9a3c95d6..ba5ac90851 100644 --- a/src/mesa/drivers/dri/sis/sis_fog.c +++ b/src/mesa/drivers/dri/sis/sis_fog.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_fog.c,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_lock.c b/src/mesa/drivers/dri/sis/sis_lock.c index 70ca8e6cbc..0ea64e3498 100644 --- a/src/mesa/drivers/dri/sis/sis_lock.c +++ b/src/mesa/drivers/dri/sis/sis_lock.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2003 Eric Anholt diff --git a/src/mesa/drivers/dri/sis/sis_lock.h b/src/mesa/drivers/dri/sis/sis_lock.h index fef9931963..54844e9b09 100644 --- a/src/mesa/drivers/dri/sis/sis_lock.h +++ b/src/mesa/drivers/dri/sis/sis_lock.h @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_reg.h b/src/mesa/drivers/dri/sis/sis_reg.h index 78c6660181..e40c4371bf 100644 --- a/src/mesa/drivers/dri/sis/sis_reg.h +++ b/src/mesa/drivers/dri/sis/sis_reg.h @@ -25,7 +25,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_reg.h,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_screen.c b/src/mesa/drivers/dri/sis/sis_screen.c index 89d734ba78..d90482f3d7 100644 --- a/src/mesa/drivers/dri/sis/sis_screen.c +++ b/src/mesa/drivers/dri/sis/sis_screen.c @@ -1,4 +1,3 @@ -/* $XFree86$ */ /************************************************************************** Copyright 2003 Eric Anholt diff --git a/src/mesa/drivers/dri/sis/sis_screen.h b/src/mesa/drivers/dri/sis/sis_screen.h index d5b2101e98..07c29cfa09 100644 --- a/src/mesa/drivers/dri/sis/sis_screen.h +++ b/src/mesa/drivers/dri/sis/sis_screen.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_span.c b/src/mesa/drivers/dri/sis/sis_span.c index ea6db6781d..dc50bda877 100644 --- a/src/mesa/drivers/dri/sis/sis_span.c +++ b/src/mesa/drivers/dri/sis/sis_span.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_span.c,v 1.5 2001/03/21 16:14:26 dawes Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_span.h b/src/mesa/drivers/dri/sis/sis_span.h index 4b0add2ac2..a1f817c44c 100644 --- a/src/mesa/drivers/dri/sis/sis_span.h +++ b/src/mesa/drivers/dri/sis/sis_span.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_state.c b/src/mesa/drivers/dri/sis/sis_state.c index 33a2f089b8..305c63f73f 100644 --- a/src/mesa/drivers/dri/sis/sis_state.c +++ b/src/mesa/drivers/dri/sis/sis_state.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_ctx.c,v 1.3 2000/09/26 15:56:48 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_state.h b/src/mesa/drivers/dri/sis/sis_state.h index 8f7e2acb92..2d0ea9c5fb 100644 --- a/src/mesa/drivers/dri/sis/sis_state.h +++ b/src/mesa/drivers/dri/sis/sis_state.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_stencil.c b/src/mesa/drivers/dri/sis/sis_stencil.c index a1ce2966e8..55c0440eba 100644 --- a/src/mesa/drivers/dri/sis/sis_stencil.c +++ b/src/mesa/drivers/dri/sis/sis_stencil.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86: xc/lib/GL/mesa/src/drv/sis/sis_stencil.c,v 1.3 2000/09/26 15:56:49 tsi Exp $ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_stencil.h b/src/mesa/drivers/dri/sis/sis_stencil.h index 4a36c98f3d..6b556c4378 100644 --- a/src/mesa/drivers/dri/sis/sis_stencil.h +++ b/src/mesa/drivers/dri/sis/sis_stencil.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_tex.c b/src/mesa/drivers/dri/sis/sis_tex.c index be87f16e29..5e10c610f8 100644 --- a/src/mesa/drivers/dri/sis/sis_tex.c +++ b/src/mesa/drivers/dri/sis/sis_tex.c @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_tex.h b/src/mesa/drivers/dri/sis/sis_tex.h index 8ddc7c469e..c499e80e86 100644 --- a/src/mesa/drivers/dri/sis/sis_tex.h +++ b/src/mesa/drivers/dri/sis/sis_tex.h @@ -22,7 +22,6 @@ AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_texstate.c b/src/mesa/drivers/dri/sis/sis_texstate.c index 7ef20f880c..4f813bb81c 100644 --- a/src/mesa/drivers/dri/sis/sis_texstate.c +++ b/src/mesa/drivers/dri/sis/sis_texstate.c @@ -24,7 +24,6 @@ OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. **************************************************************************/ -/* $XFree86$ */ /* * Authors: diff --git a/src/mesa/drivers/dri/sis/sis_tris.h b/src/mesa/drivers/dri/sis/sis_tris.h index 5e07acc211..499eb4d24d 100644 --- a/src/mesa/drivers/dri/sis/sis_tris.h +++ b/src/mesa/drivers/dri/sis/sis_tris.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/r128/r128_tris.h,v 1.8 2002/10/30 12:51:43 alanh Exp $ */ /************************************************************************** Copyright 2003 Eric Anholt diff --git a/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S b/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S index 0f4cc45089..500c97c536 100644 --- a/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S +++ b/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fastpath.S @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/X86/fx_3dnow_fastpath.S,v 1.2 2000/09/26 15:56:51 tsi Exp $ */ #include "../../X86/assyntax.h" diff --git a/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h b/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h index 9ec4935d78..78c5fef746 100644 --- a/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h +++ b/src/mesa/drivers/dri/tdfx/X86/fx_3dnow_fasttmp.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/X86/fx_3dnow_fasttmp.h,v 1.2 2000/09/26 15:56:51 tsi Exp $ */ #if !defined(NASM_ASSEMBLER) && !defined(MASM_ASSEMBLER) #define TAGLLBL(a) TAG(.L##a) diff --git a/src/mesa/drivers/dri/tdfx/dri_glide.h b/src/mesa/drivers/dri/tdfx/dri_glide.h index 52a53f7dd3..3ad2bf68c6 100644 --- a/src/mesa/drivers/dri/tdfx/dri_glide.h +++ b/src/mesa/drivers/dri/tdfx/dri_glide.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/dri_glide.h,v 1.1 2001/03/21 16:14:26 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/server/tdfx_dri.h b/src/mesa/drivers/dri/tdfx/server/tdfx_dri.h index acd0b9ae5b..dc29984a27 100644 --- a/src/mesa/drivers/dri/tdfx/server/tdfx_dri.h +++ b/src/mesa/drivers/dri/tdfx/server/tdfx_dri.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/tdfx/tdfx_dri.h,v 1.5 2001/03/21 17:02:26 dawes Exp $ */ #ifndef _TDFX_DRI_ #define _TDFX_DRI_ diff --git a/src/mesa/drivers/dri/tdfx/tdfx_context.h b/src/mesa/drivers/dri/tdfx/tdfx_context.h index 89a7a9d6c4..05673cd186 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_context.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_context.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_context.h,v 1.5 2002/02/24 21:51:10 dawes Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_dd.h b/src/mesa/drivers/dri/tdfx/tdfx_dd.h index 5ceba9d5f0..bd61e10605 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_dd.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_dd.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_dd.h,v 1.1 2001/03/21 16:14:27 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_glide.h b/src/mesa/drivers/dri/tdfx/tdfx_glide.h index f077aa678b..69e5399e72 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_glide.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_glide.h @@ -2,7 +2,6 @@ * This file defines macros and types necessary for accessing glide3. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_glide.h,v 1.1 2002/02/22 21:45:03 dawes Exp $ */ #ifndef NEWGLIDE_H #define NEWGLIDE_H diff --git a/src/mesa/drivers/dri/tdfx/tdfx_lock.c b/src/mesa/drivers/dri/tdfx/tdfx_lock.c index a20c91d030..17cdc51ee1 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_lock.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_lock.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_lock.c,v 1.5 2002/12/16 16:19:00 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_lock.h b/src/mesa/drivers/dri/tdfx/tdfx_lock.h index 616e65b2a1..74e3f5c9cc 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_lock.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_lock.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_lock.h,v 1.3 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_pixels.c b/src/mesa/drivers/dri/tdfx/tdfx_pixels.c index 732270b2bd..b5c01f6ef2 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_pixels.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_pixels.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_pixels.c,v 1.4 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_pixels.h b/src/mesa/drivers/dri/tdfx/tdfx_pixels.h index c38ce070ca..55f7eedef8 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_pixels.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_pixels.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_pixels.h,v 1.2 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_render.c b/src/mesa/drivers/dri/tdfx/tdfx_render.c index f36c97bfeb..e374f09df3 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_render.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_render.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_render.c,v 1.4 2002/02/22 21:45:03 dawes Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_render.h b/src/mesa/drivers/dri/tdfx/tdfx_render.h index 09d0d90197..18c6168333 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_render.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_render.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_render.h,v 1.1 2001/03/21 16:14:28 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_screen.c b/src/mesa/drivers/dri/tdfx/tdfx_screen.c index 1f9ff4e30c..7761664394 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_screen.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_screen.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_screen.c,v 1.3 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_screen.h b/src/mesa/drivers/dri/tdfx/tdfx_screen.h index 90be89a352..5a68898b36 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_screen.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_screen.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_screen.h,v 1.2 2002/02/22 21:45:03 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_span.c b/src/mesa/drivers/dri/tdfx/tdfx_span.c index d9d52d2b6f..6b38fa5a01 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_span.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_span.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_span.c,v 1.7 2002/10/30 12:52:00 alanh Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_span.h b/src/mesa/drivers/dri/tdfx/tdfx_span.h index 62044144f0..5af9f9b301 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_span.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_span.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_span.h,v 1.1 2001/03/21 16:14:28 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_state.c b/src/mesa/drivers/dri/tdfx/tdfx_state.c index 42cb5dfaa3..3688c76a5c 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_state.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_state.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_state.c,v 1.7 2002/10/30 12:52:00 alanh Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_state.h b/src/mesa/drivers/dri/tdfx/tdfx_state.h index b10c38f591..591ea5b083 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_state.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_state.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_state.h,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tex.c b/src/mesa/drivers/dri/tdfx/tdfx_tex.c index 89865d9637..65e665ee39 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tex.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_tex.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_tex.c,v 1.7 2002/11/05 17:46:10 tsi Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tex.h b/src/mesa/drivers/dri/tdfx/tdfx_tex.h index f536c25a2f..a445935a01 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tex.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_tex.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_tex.h,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_texman.c b/src/mesa/drivers/dri/tdfx/tdfx_texman.c index 6f782f687f..f9b2726da2 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_texman.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_texman.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_texman.c,v 1.5 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_texman.h b/src/mesa/drivers/dri/tdfx/tdfx_texman.h index 739d4e142f..a9af4cb7c5 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_texman.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_texman.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_texman.h,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_texstate.c b/src/mesa/drivers/dri/tdfx/tdfx_texstate.c index fda9ce5684..bbd2c8cfee 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_texstate.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_texstate.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_texstate.c,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ /* * New fixes: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_texstate.h b/src/mesa/drivers/dri/tdfx/tdfx_texstate.h index 234ed4439a..0c5c4101ca 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_texstate.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_texstate.h @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_texstate.h,v 1.1 2002/02/22 21:45:04 dawes Exp $ */ /* * Original rewrite: diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tris.c b/src/mesa/drivers/dri/tdfx/tdfx_tris.c index 7252a7e7dc..59ff35a7fa 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tris.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_tris.c @@ -23,7 +23,6 @@ * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE * SOFTWARE. */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_tris.c,v 1.4 2002/10/30 12:52:01 alanh Exp $ */ /* New fixes: * Daniel Borca , 19 Jul 2004 diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tris.h b/src/mesa/drivers/dri/tdfx/tdfx_tris.h index 57e5d9b0ae..a591decf1d 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tris.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_tris.h @@ -29,7 +29,6 @@ * Keith Whitwell * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_tris.h,v 1.5 2002/10/30 12:52:01 alanh Exp $ */ #ifndef TDFX_TRIS_INC #define TDFX_TRIS_INC diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vb.c b/src/mesa/drivers/dri/tdfx/tdfx_vb.c index 0580135d1b..62885daaa5 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vb.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_vb.c @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_vb.c,v 1.3 2002/10/30 12:52:01 alanh Exp $ */ #include "glheader.h" #include "mtypes.h" diff --git a/src/mesa/drivers/dri/tdfx/tdfx_vb.h b/src/mesa/drivers/dri/tdfx/tdfx_vb.h index 7b7cd9065a..6389ec95b1 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_vb.h +++ b/src/mesa/drivers/dri/tdfx/tdfx_vb.h @@ -22,7 +22,6 @@ * * */ -/* $XFree86: xc/lib/GL/mesa/src/drv/tdfx/tdfx_vb.h,v 1.2 2002/02/22 21:45:04 dawes Exp $ */ #ifndef TDFXVB_INC #define TDFXVB_INC diff --git a/src/mesa/drivers/dri/unichrome/server/via_dri.c b/src/mesa/drivers/dri/unichrome/server/via_dri.c index 6944bd66f9..9833145940 100644 --- a/src/mesa/drivers/dri/unichrome/server/via_dri.c +++ b/src/mesa/drivers/dri/unichrome/server/via_dri.c @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/via/via_dri.c,v 1.4 2003/09/24 02:43:30 dawes Exp $ */ /* * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved. * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved. diff --git a/src/mesa/drivers/dri/unichrome/server/via_driver.h b/src/mesa/drivers/dri/unichrome/server/via_driver.h index 997b2e41a7..a643fd9fbb 100644 --- a/src/mesa/drivers/dri/unichrome/server/via_driver.h +++ b/src/mesa/drivers/dri/unichrome/server/via_driver.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/via/via_driver.h,v 1.7 2003/11/06 18:38:11 tsi Exp $ */ /* * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved. * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved. diff --git a/src/mesa/drivers/dri/unichrome/server/via_priv.h b/src/mesa/drivers/dri/unichrome/server/via_priv.h index 587531b37c..352eac0597 100644 --- a/src/mesa/drivers/dri/unichrome/server/via_priv.h +++ b/src/mesa/drivers/dri/unichrome/server/via_priv.h @@ -1,4 +1,3 @@ -/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/via/via_priv.h,v 1.3 2003/08/27 15:16:12 tsi Exp $ */ #ifndef _VIA_PRIV_H_ #define _VIA_PRIV_H_ 1 diff --git a/src/mesa/drivers/ggi/default/genkgi.h b/src/mesa/drivers/ggi/default/genkgi.h index 022189138f..6d0963416f 100644 --- a/src/mesa/drivers/ggi/default/genkgi.h +++ b/src/mesa/drivers/ggi/default/genkgi.h @@ -1,4 +1,4 @@ -/* $Id: genkgi.h,v 1.3 1999/08/22 08:56:50 jtaylor Exp $ +/* ****************************************************************************** GGIMesa - KGIcon specific overrides for fbcon-mesa diff --git a/src/mesa/drivers/ggi/default/genkgi_mode.c b/src/mesa/drivers/ggi/default/genkgi_mode.c index 938024789f..f81d6a45bd 100644 --- a/src/mesa/drivers/ggi/default/genkgi_mode.c +++ b/src/mesa/drivers/ggi/default/genkgi_mode.c @@ -1,4 +1,4 @@ -/* $Id: genkgi_mode.c,v 1.4 2000/01/07 08:34:44 jtaylor Exp $ +/* ****************************************************************************** display-fbdev-kgicon-generic-mesa diff --git a/src/mesa/drivers/ggi/default/genkgi_visual.c b/src/mesa/drivers/ggi/default/genkgi_visual.c index 17ef9679bb..d7838cae6e 100644 --- a/src/mesa/drivers/ggi/default/genkgi_visual.c +++ b/src/mesa/drivers/ggi/default/genkgi_visual.c @@ -1,4 +1,4 @@ -/* $Id: genkgi_visual.c,v 1.7 2000/06/11 20:11:55 jtaylor Exp $ +/* ****************************************************************************** genkgi_visual.c: visual handling for the generic KGI helper diff --git a/src/mesa/drivers/ggi/include/ggi/mesa/debug.h b/src/mesa/drivers/ggi/include/ggi/mesa/debug.h index 35d11624c6..f461fee72c 100644 --- a/src/mesa/drivers/ggi/include/ggi/mesa/debug.h +++ b/src/mesa/drivers/ggi/include/ggi/mesa/debug.h @@ -1,4 +1,4 @@ -/* $Id: debug.h,v 1.5 2003/09/22 15:18:51 brianp Exp $ +/* ****************************************************************************** GGIMesa debugging macros diff --git a/src/mesa/drivers/svga/svgamesa.c b/src/mesa/drivers/svga/svgamesa.c index d138587569..1e4e185d65 100644 --- a/src/mesa/drivers/svga/svgamesa.c +++ b/src/mesa/drivers/svga/svgamesa.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa.c,v 1.27 2006/10/15 18:51:22 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa15.c b/src/mesa/drivers/svga/svgamesa15.c index ae5104d0c0..934aaa33fb 100644 --- a/src/mesa/drivers/svga/svgamesa15.c +++ b/src/mesa/drivers/svga/svgamesa15.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa15.c,v 1.11.36.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa15.h b/src/mesa/drivers/svga/svgamesa15.h index 3ed7db82ee..d453fb8d35 100644 --- a/src/mesa/drivers/svga/svgamesa15.h +++ b/src/mesa/drivers/svga/svgamesa15.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa15.h,v 1.7 2002/11/11 18:42:39 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa16.c b/src/mesa/drivers/svga/svgamesa16.c index a59937bfb4..9fc8c786e8 100644 --- a/src/mesa/drivers/svga/svgamesa16.c +++ b/src/mesa/drivers/svga/svgamesa16.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa16.c,v 1.11.36.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa16.h b/src/mesa/drivers/svga/svgamesa16.h index 247c1f4045..b80cd3dd7e 100644 --- a/src/mesa/drivers/svga/svgamesa16.h +++ b/src/mesa/drivers/svga/svgamesa16.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa16.h,v 1.6 2002/11/11 18:42:41 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa24.c b/src/mesa/drivers/svga/svgamesa24.c index dd15bf38db..c7c095333f 100644 --- a/src/mesa/drivers/svga/svgamesa24.c +++ b/src/mesa/drivers/svga/svgamesa24.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa24.c,v 1.12.36.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa24.h b/src/mesa/drivers/svga/svgamesa24.h index 54d1a8298b..df5fa68c44 100644 --- a/src/mesa/drivers/svga/svgamesa24.h +++ b/src/mesa/drivers/svga/svgamesa24.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa24.h,v 1.6 2002/11/11 18:42:41 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa32.c b/src/mesa/drivers/svga/svgamesa32.c index 4da18795d8..d089c20c05 100644 --- a/src/mesa/drivers/svga/svgamesa32.c +++ b/src/mesa/drivers/svga/svgamesa32.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa32.c,v 1.12.36.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa32.h b/src/mesa/drivers/svga/svgamesa32.h index f518e11ad5..6cf8315300 100644 --- a/src/mesa/drivers/svga/svgamesa32.h +++ b/src/mesa/drivers/svga/svgamesa32.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa32.h,v 1.6 2002/11/11 18:42:42 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa8.c b/src/mesa/drivers/svga/svgamesa8.c index 4264fcd959..2f7048a930 100644 --- a/src/mesa/drivers/svga/svgamesa8.c +++ b/src/mesa/drivers/svga/svgamesa8.c @@ -1,4 +1,3 @@ -/* $Id: svgamesa8.c,v 1.9.10.1 2006/11/02 12:02:17 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgamesa8.h b/src/mesa/drivers/svga/svgamesa8.h index 1aa25f93fc..d2b0509480 100644 --- a/src/mesa/drivers/svga/svgamesa8.h +++ b/src/mesa/drivers/svga/svgamesa8.h @@ -1,4 +1,3 @@ -/* $Id: svgamesa8.h,v 1.4 2001/02/06 00:03:48 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/svga/svgapix.h b/src/mesa/drivers/svga/svgapix.h index 0b19551bf6..19cb74487d 100644 --- a/src/mesa/drivers/svga/svgapix.h +++ b/src/mesa/drivers/svga/svgapix.h @@ -1,4 +1,3 @@ -/* $Id: svgapix.h,v 1.5 2002/11/11 18:42:44 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gdi/wgl.c b/src/mesa/drivers/windows/gdi/wgl.c index dad3dc1160..6e00d08aba 100644 --- a/src/mesa/drivers/windows/gdi/wgl.c +++ b/src/mesa/drivers/windows/gdi/wgl.c @@ -1,4 +1,3 @@ -/* $Id: wgl.c,v 1.12 2006/03/30 07:58:24 kschultz Exp $ */ /* * This library is free software; you can redistribute it and/or diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c index ecc40e8f8b..72e5e1308c 100644 --- a/src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c +++ b/src/mesa/drivers/windows/gldirect/dx7/gld_vb_mesa_render_dx7.c @@ -1,4 +1,3 @@ -/* $Id: gld_vb_mesa_render_dx7.c,v 1.6 2005/08/27 13:56:08 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c index 414a2f64bf..9ab562010c 100644 --- a/src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c +++ b/src/mesa/drivers/windows/gldirect/dx8/gld_vb_mesa_render_dx8.c @@ -1,4 +1,3 @@ -/* $Id: gld_vb_mesa_render_dx8.c,v 1.6 2005/08/27 13:56:08 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c index c71fdefbae..64acab2d2a 100644 --- a/src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c +++ b/src/mesa/drivers/windows/gldirect/dx9/gld_vb_mesa_render_dx9.c @@ -1,4 +1,3 @@ -/* $Id: gld_vb_mesa_render_dx9.c,v 1.6 2005/08/27 13:56:08 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/gld_debug_clip.c b/src/mesa/drivers/windows/gldirect/gld_debug_clip.c index 1eb19ca84b..044d2e66f4 100644 --- a/src/mesa/drivers/windows/gldirect/gld_debug_clip.c +++ b/src/mesa/drivers/windows/gldirect/gld_debug_clip.c @@ -1,4 +1,3 @@ -/* $Id: gld_debug_clip.c,v 1.1 2004/04/20 11:13:11 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/gld_debug_norm.c b/src/mesa/drivers/windows/gldirect/gld_debug_norm.c index 00c428bd26..c20362bb24 100644 --- a/src/mesa/drivers/windows/gldirect/gld_debug_norm.c +++ b/src/mesa/drivers/windows/gldirect/gld_debug_norm.c @@ -1,4 +1,3 @@ -/* $Id: gld_debug_norm.c,v 1.1 2004/04/20 11:13:11 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/gld_debug_xform.c b/src/mesa/drivers/windows/gldirect/gld_debug_xform.c index d6e64b8ffd..73439dc3b6 100644 --- a/src/mesa/drivers/windows/gldirect/gld_debug_xform.c +++ b/src/mesa/drivers/windows/gldirect/gld_debug_xform.c @@ -1,4 +1,3 @@ -/* $Id: gld_debug_xform.c,v 1.1 2004/04/20 11:13:11 alanh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/drivers/windows/gldirect/mesasw/colors.h b/src/mesa/drivers/windows/gldirect/mesasw/colors.h index 17371a96cc..9c1f2a0540 100644 --- a/src/mesa/drivers/windows/gldirect/mesasw/colors.h +++ b/src/mesa/drivers/windows/gldirect/mesasw/colors.h @@ -18,12 +18,11 @@ * (mark@rsinc.com). */ -/* $Log: ddcolors.h 1997/6/14 by Li Wei(liwei@aiar.xjtu.edu.cn) +/* * Macros for pixel format defined */ /* - * $Log: colors.h,v $ * Revision 1.1 2004/04/20 11:13:11 alanh * add SciTech's GLDirect driver for Windows. * @@ -46,7 +45,6 @@ */ /* - * $Log: colors.h,v $ * Revision 1.1 2004/04/20 11:13:11 alanh * add SciTech's GLDirect driver for Windows. * @@ -69,7 +67,6 @@ */ /* - * $Log: colors.h,v $ * Revision 1.1 2004/04/20 11:13:11 alanh * add SciTech's GLDirect driver for Windows. * @@ -520,4 +517,4 @@ char unsigned const aWinGHalftoneTranslation[216] = 225, 226, 255, -}; \ No newline at end of file +}; diff --git a/src/mesa/glapi/mesadef.py b/src/mesa/glapi/mesadef.py index 097348dae0..0f410fc482 100644 --- a/src/mesa/glapi/mesadef.py +++ b/src/mesa/glapi/mesadef.py @@ -1,6 +1,5 @@ #!/usr/bin/env python -# $Id: mesadef.py,v 1.4 2006/01/25 15:05:36 brianp Exp $ # Mesa 3-D graphics library # Version: 4.1 diff --git a/src/mesa/sparc/norm.S b/src/mesa/sparc/norm.S index 713cd5b375..44950a10a5 100644 --- a/src/mesa/sparc/norm.S +++ b/src/mesa/sparc/norm.S @@ -1,4 +1,3 @@ -/* $Id: norm.S,v 1.5 2005/07/28 00:11:11 idr Exp $ */ #include "sparc_matrix.h" diff --git a/src/mesa/sparc/sparc.h b/src/mesa/sparc/sparc.h index 55ab12122d..a98e4d0e40 100644 --- a/src/mesa/sparc/sparc.h +++ b/src/mesa/sparc/sparc.h @@ -1,4 +1,3 @@ -/* $Id: sparc.h,v 1.3 2001/06/06 22:55:28 davem69 Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/sparc/xform.S b/src/mesa/sparc/xform.S index f44ec794e9..f2b9674bf2 100644 --- a/src/mesa/sparc/xform.S +++ b/src/mesa/sparc/xform.S @@ -1,4 +1,3 @@ -/* $Id: xform.S,v 1.4 2005/07/28 00:11:11 idr Exp $ */ /* TODO * diff --git a/src/mesa/x86-64/x86-64.c b/src/mesa/x86-64/x86-64.c index 09508b66d5..dee09fd648 100644 --- a/src/mesa/x86-64/x86-64.c +++ b/src/mesa/x86-64/x86-64.c @@ -1,4 +1,3 @@ -/* $Id: x86-64.c,v 1.4 2006/10/17 17:03:21 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86-64/x86-64.h b/src/mesa/x86-64/x86-64.h index fdbd154d5d..1d931fa345 100644 --- a/src/mesa/x86-64/x86-64.h +++ b/src/mesa/x86-64/x86-64.h @@ -1,4 +1,3 @@ -/* $Id: x86-64.h,v 1.1 2005/05/07 16:59:59 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86-64/xform4.S b/src/mesa/x86-64/xform4.S index 65328f6666..667ecf6e58 100644 --- a/src/mesa/x86-64/xform4.S +++ b/src/mesa/x86-64/xform4.S @@ -1,4 +1,3 @@ -/* $Id: xform4.S,v 1.2 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow.c b/src/mesa/x86/3dnow.c index 032aa661f4..4122ee4b00 100644 --- a/src/mesa/x86/3dnow.c +++ b/src/mesa/x86/3dnow.c @@ -1,4 +1,3 @@ -/* $Id: 3dnow.c,v 1.24 2005/10/07 17:18:52 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow.h b/src/mesa/x86/3dnow.h index 1f2fd8e8b4..df9f2638d7 100644 --- a/src/mesa/x86/3dnow.h +++ b/src/mesa/x86/3dnow.h @@ -1,4 +1,3 @@ -/* $Id: 3dnow.h,v 1.6 2002/04/09 14:58:03 keithw Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_normal.S b/src/mesa/x86/3dnow_normal.S index f3bbcb27b7..693a7864db 100644 --- a/src/mesa/x86/3dnow_normal.S +++ b/src/mesa/x86/3dnow_normal.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_normal.S,v 1.10 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_xform1.S b/src/mesa/x86/3dnow_xform1.S index 22b12cca06..7665c0ff8b 100644 --- a/src/mesa/x86/3dnow_xform1.S +++ b/src/mesa/x86/3dnow_xform1.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_xform1.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_xform2.S b/src/mesa/x86/3dnow_xform2.S index d9e96d04e2..b201d1e901 100644 --- a/src/mesa/x86/3dnow_xform2.S +++ b/src/mesa/x86/3dnow_xform2.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_xform2.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_xform3.S b/src/mesa/x86/3dnow_xform3.S index babee1caa0..46f155697d 100644 --- a/src/mesa/x86/3dnow_xform3.S +++ b/src/mesa/x86/3dnow_xform3.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_xform3.S,v 1.5 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/3dnow_xform4.S b/src/mesa/x86/3dnow_xform4.S index b16d2b12dd..a0c6b193cd 100644 --- a/src/mesa/x86/3dnow_xform4.S +++ b/src/mesa/x86/3dnow_xform4.S @@ -1,4 +1,3 @@ -/* $Id: 3dnow_xform4.S,v 1.5 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/clip_args.h b/src/mesa/x86/clip_args.h index cccf801981..796611fbfd 100644 --- a/src/mesa/x86/clip_args.h +++ b/src/mesa/x86/clip_args.h @@ -1,4 +1,3 @@ -/* $Id: clip_args.h,v 1.5 2002/10/29 20:28:57 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/common_x86_asm.h b/src/mesa/x86/common_x86_asm.h index 9977298328..89312b2437 100644 --- a/src/mesa/x86/common_x86_asm.h +++ b/src/mesa/x86/common_x86_asm.h @@ -1,4 +1,3 @@ -/* $Id: common_x86_asm.h,v 1.12 2005/07/16 00:56:20 ajax Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/common_x86_features.h b/src/mesa/x86/common_x86_features.h index 90509775cf..676af8c1f8 100644 --- a/src/mesa/x86/common_x86_features.h +++ b/src/mesa/x86/common_x86_features.h @@ -1,4 +1,3 @@ -/* $Id: common_x86_features.h,v 1.6 2003/01/21 16:14:00 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/common_x86_macros.h b/src/mesa/x86/common_x86_macros.h index ba155caae1..462f32b3f2 100644 --- a/src/mesa/x86/common_x86_macros.h +++ b/src/mesa/x86/common_x86_macros.h @@ -1,4 +1,3 @@ -/* $Id: common_x86_macros.h,v 1.3 2002/10/29 20:28:58 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/norm_args.h b/src/mesa/x86/norm_args.h index 1b43d57a20..5d352838be 100644 --- a/src/mesa/x86/norm_args.h +++ b/src/mesa/x86/norm_args.h @@ -1,4 +1,3 @@ -/* $Id: norm_args.h,v 1.4 2003/11/26 08:32:36 dborca Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse.h b/src/mesa/x86/sse.h index 98146a9047..521f91e411 100644 --- a/src/mesa/x86/sse.h +++ b/src/mesa/x86/sse.h @@ -1,4 +1,3 @@ -/* $Id: sse.h,v 1.2 2002/04/09 14:58:03 keithw Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_normal.S b/src/mesa/x86/sse_normal.S index 066d46e5ef..1c32e3b2fe 100644 --- a/src/mesa/x86/sse_normal.S +++ b/src/mesa/x86/sse_normal.S @@ -1,4 +1,3 @@ -/* $Id: sse_normal.S,v 1.6 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_xform1.S b/src/mesa/x86/sse_xform1.S index 4051f606a7..22fd8dd27b 100644 --- a/src/mesa/x86/sse_xform1.S +++ b/src/mesa/x86/sse_xform1.S @@ -1,4 +1,3 @@ -/* $Id: sse_xform1.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_xform2.S b/src/mesa/x86/sse_xform2.S index 06fe086bd4..52eeb27ef5 100644 --- a/src/mesa/x86/sse_xform2.S +++ b/src/mesa/x86/sse_xform2.S @@ -1,4 +1,3 @@ -/* $Id: sse_xform2.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_xform3.S b/src/mesa/x86/sse_xform3.S index eafbe34288..5e0cd8b666 100644 --- a/src/mesa/x86/sse_xform3.S +++ b/src/mesa/x86/sse_xform3.S @@ -1,4 +1,3 @@ -/* $Id: sse_xform3.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/sse_xform4.S b/src/mesa/x86/sse_xform4.S index 24c323194f..13680528db 100644 --- a/src/mesa/x86/sse_xform4.S +++ b/src/mesa/x86/sse_xform4.S @@ -1,4 +1,3 @@ -/* $Id: sse_xform4.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86.c b/src/mesa/x86/x86.c index 6b74e9e375..82caa42dbd 100644 --- a/src/mesa/x86/x86.c +++ b/src/mesa/x86/x86.c @@ -1,4 +1,3 @@ -/* $Id: x86.c,v 1.26 2005/10/07 17:18:52 brianp Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86.h b/src/mesa/x86/x86.h index a646aff46b..97651ec6ee 100644 --- a/src/mesa/x86/x86.h +++ b/src/mesa/x86/x86.h @@ -1,4 +1,3 @@ -/* $Id: x86.h,v 1.5 2002/04/09 14:58:03 keithw Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86_cliptest.S b/src/mesa/x86/x86_cliptest.S index 5a45ee6ae6..c7a3a9b57e 100644 --- a/src/mesa/x86/x86_cliptest.S +++ b/src/mesa/x86/x86_cliptest.S @@ -1,4 +1,3 @@ -/* $Id: x86_cliptest.S,v 1.12 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86_xform2.S b/src/mesa/x86/x86_xform2.S index 94f6989d42..e41661d546 100644 --- a/src/mesa/x86/x86_xform2.S +++ b/src/mesa/x86/x86_xform2.S @@ -1,4 +1,3 @@ -/* $Id: x86_xform2.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86_xform3.S b/src/mesa/x86/x86_xform3.S index 747e2f4d28..067ddd7d7c 100644 --- a/src/mesa/x86/x86_xform3.S +++ b/src/mesa/x86/x86_xform3.S @@ -1,4 +1,3 @@ -/* $Id: x86_xform3.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/x86_xform4.S b/src/mesa/x86/x86_xform4.S index def3c1ceb9..77621ac4bd 100644 --- a/src/mesa/x86/x86_xform4.S +++ b/src/mesa/x86/x86_xform4.S @@ -1,4 +1,3 @@ -/* $Id: x86_xform4.S,v 1.4 2006/04/17 18:58:24 krh Exp $ */ /* * Mesa 3-D graphics library diff --git a/src/mesa/x86/xform_args.h b/src/mesa/x86/xform_args.h index 89a04205c1..b773f5198d 100644 --- a/src/mesa/x86/xform_args.h +++ b/src/mesa/x86/xform_args.h @@ -1,4 +1,3 @@ -/* $Id: xform_args.h,v 1.5 2002/10/29 20:28:58 brianp Exp $ */ /* * Mesa 3-D graphics library -- cgit v1.2.3 From 253066d716e3039522eeb7b072811cccd89b4a82 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 26 May 2008 23:29:38 +0900 Subject: Revert DOS line endings. --- src/gallium/drivers/i915simple/i915_context.h | 682 ++++++------ src/gallium/drivers/i965simple/brw_context.h | 1368 ++++++++++++------------- src/gallium/drivers/i965simple/brw_state.c | 938 ++++++++--------- 3 files changed, 1494 insertions(+), 1494 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 38e6a34884..53fc5ed079 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -1,341 +1,341 @@ - /************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef I915_CONTEXT_H -#define I915_CONTEXT_H - - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" - -#include "draw/draw_vertex.h" - -#include "tgsi/util/tgsi_scan.h" - - -#define I915_TEX_UNITS 8 - -#define I915_DYNAMIC_MODES4 0 -#define I915_DYNAMIC_DEPTHSCALE_0 1 /* just the header */ -#define I915_DYNAMIC_DEPTHSCALE_1 2 -#define I915_DYNAMIC_IAB 3 -#define I915_DYNAMIC_BC_0 4 /* just the header */ -#define I915_DYNAMIC_BC_1 5 -#define I915_DYNAMIC_BFO_0 6 -#define I915_DYNAMIC_BFO_1 7 -#define I915_DYNAMIC_STP_0 8 -#define I915_DYNAMIC_STP_1 9 -#define I915_DYNAMIC_SC_ENA_0 10 -#define I915_DYNAMIC_SC_RECT_0 11 -#define I915_DYNAMIC_SC_RECT_1 12 -#define I915_DYNAMIC_SC_RECT_2 13 -#define I915_MAX_DYNAMIC 14 - - -#define I915_IMMEDIATE_S0 0 -#define I915_IMMEDIATE_S1 1 -#define I915_IMMEDIATE_S2 2 -#define I915_IMMEDIATE_S3 3 -#define I915_IMMEDIATE_S4 4 -#define I915_IMMEDIATE_S5 5 -#define I915_IMMEDIATE_S6 6 -#define I915_IMMEDIATE_S7 7 -#define I915_MAX_IMMEDIATE 8 - -/* These must mach the order of LI0_STATE_* bits, as they will be used - * to generate hardware packets: - */ -#define I915_CACHE_STATIC 0 -#define I915_CACHE_DYNAMIC 1 /* handled specially */ -#define I915_CACHE_SAMPLER 2 -#define I915_CACHE_MAP 3 -#define I915_CACHE_PROGRAM 4 -#define I915_CACHE_CONSTANTS 5 -#define I915_MAX_CACHE 6 - -#define I915_MAX_CONSTANT 32 - - -/** See constant_flags[] below */ -#define I915_CONSTFLAG_USER 0x1f - - -/** - * Subclass of pipe_shader_state - */ -struct i915_fragment_shader -{ - struct pipe_shader_state state; - - struct tgsi_shader_info info; - - uint *program; - uint program_len; - - /** - * constants introduced during translation. - * These are placed at the end of the constant buffer and grow toward - * the beginning (eg: slot 31, 30 29, ...) - * User-provided constants start at 0. - * This allows both types of constants to co-exist (until there's too many) - * and doesn't require regenerating/changing the fragment program to - * shuffle constants around. - */ - uint num_constants; - float constants[I915_MAX_CONSTANT][4]; - - /** - * Status of each constant - * if I915_CONSTFLAG_PARAM, the value must be taken from the corresponding - * slot of the user's constant buffer. (set by pipe->set_constant_buffer()) - * Else, the bitmask indicates which components are occupied by immediates. - */ - ubyte constant_flags[I915_MAX_CONSTANT]; -}; - - -struct i915_cache_context; - -/* Use to calculate differences between state emitted to hardware and - * current driver-calculated state. - */ -struct i915_state -{ - unsigned immediate[I915_MAX_IMMEDIATE]; - unsigned dynamic[I915_MAX_DYNAMIC]; - - float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; - /** number of constants passed in through a constant buffer */ - uint num_user_constants[PIPE_SHADER_TYPES]; - - /* texture sampler state */ - unsigned sampler[I915_TEX_UNITS][3]; - unsigned sampler_enable_flags; - unsigned sampler_enable_nr; - - /* texture image buffers */ - unsigned texbuffer[I915_TEX_UNITS][2]; - - /** Describes the current hardware vertex layout */ - struct vertex_info vertex_info; - - unsigned id; /* track lost context events */ -}; - -struct i915_blend_state { - unsigned iab; - unsigned modes4; - unsigned LIS5; - unsigned LIS6; -}; - -struct i915_depth_stencil_state { - unsigned stencil_modes4; - unsigned bfo[2]; - unsigned stencil_LIS5; - unsigned depth_LIS6; -}; - -struct i915_rasterizer_state { - int light_twoside : 1; - unsigned st; - enum interp_mode color_interp; - - unsigned LIS4; - unsigned LIS7; - unsigned sc[1]; - - const struct pipe_rasterizer_state *templ; - - union { float f; unsigned u; } ds[2]; -}; - -struct i915_sampler_state { - unsigned state[3]; - const struct pipe_sampler_state *templ; -}; - - -struct i915_texture { - struct pipe_texture base; - - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ - unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; -}; - -struct i915_context -{ - struct pipe_context pipe; - struct i915_winsys *winsys; - struct draw_context *draw; - - /* The most recent drawing state as set by the driver: - */ - const struct i915_blend_state *blend; - const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct i915_depth_stencil_state *depth_stencil; - const struct i915_rasterizer_state *rasterizer; - - struct i915_fragment_shader *fs; - - struct pipe_blend_color blend_color; - struct pipe_clip_state clip; - struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; - struct pipe_framebuffer_state framebuffer; - struct pipe_poly_stipple poly_stipple; - struct pipe_scissor_state scissor; - struct i915_texture *texture[PIPE_MAX_SAMPLERS]; - struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; - - unsigned dirty; - - unsigned num_samplers; - unsigned num_textures; - unsigned num_vertex_elements; - unsigned num_vertex_buffers; - - unsigned *batch_start; - - /** Vertex buffer */ - struct pipe_buffer *vbo; - - struct i915_state current; - unsigned hardware_dirty; - - unsigned debug; -}; - -/* A flag for each state_tracker state object: - */ -#define I915_NEW_VIEWPORT 0x1 -#define I915_NEW_RASTERIZER 0x2 -#define I915_NEW_FS 0x4 -#define I915_NEW_BLEND 0x8 -#define I915_NEW_CLIP 0x10 -#define I915_NEW_SCISSOR 0x20 -#define I915_NEW_STIPPLE 0x40 -#define I915_NEW_FRAMEBUFFER 0x80 -#define I915_NEW_ALPHA_TEST 0x100 -#define I915_NEW_DEPTH_STENCIL 0x200 -#define I915_NEW_SAMPLER 0x400 -#define I915_NEW_TEXTURE 0x800 -#define I915_NEW_CONSTANTS 0x1000 -#define I915_NEW_VBO 0x2000 -#define I915_NEW_VS 0x4000 - - -/* Driver's internally generated state flags: - */ -#define I915_NEW_VERTEX_FORMAT 0x10000 - - -/* Dirty flags for hardware emit - */ -#define I915_HW_STATIC (1<set_constant_buffer()) + * Else, the bitmask indicates which components are occupied by immediates. + */ + ubyte constant_flags[I915_MAX_CONSTANT]; +}; + + +struct i915_cache_context; + +/* Use to calculate differences between state emitted to hardware and + * current driver-calculated state. + */ +struct i915_state +{ + unsigned immediate[I915_MAX_IMMEDIATE]; + unsigned dynamic[I915_MAX_DYNAMIC]; + + float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; + /** number of constants passed in through a constant buffer */ + uint num_user_constants[PIPE_SHADER_TYPES]; + + /* texture sampler state */ + unsigned sampler[I915_TEX_UNITS][3]; + unsigned sampler_enable_flags; + unsigned sampler_enable_nr; + + /* texture image buffers */ + unsigned texbuffer[I915_TEX_UNITS][2]; + + /** Describes the current hardware vertex layout */ + struct vertex_info vertex_info; + + unsigned id; /* track lost context events */ +}; + +struct i915_blend_state { + unsigned iab; + unsigned modes4; + unsigned LIS5; + unsigned LIS6; +}; + +struct i915_depth_stencil_state { + unsigned stencil_modes4; + unsigned bfo[2]; + unsigned stencil_LIS5; + unsigned depth_LIS6; +}; + +struct i915_rasterizer_state { + int light_twoside : 1; + unsigned st; + enum interp_mode color_interp; + + unsigned LIS4; + unsigned LIS7; + unsigned sc[1]; + + const struct pipe_rasterizer_state *templ; + + union { float f; unsigned u; } ds[2]; +}; + +struct i915_sampler_state { + unsigned state[3]; + const struct pipe_sampler_state *templ; +}; + + +struct i915_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; +}; + +struct i915_context +{ + struct pipe_context pipe; + struct i915_winsys *winsys; + struct draw_context *draw; + + /* The most recent drawing state as set by the driver: + */ + const struct i915_blend_state *blend; + const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct i915_depth_stencil_state *depth_stencil; + const struct i915_rasterizer_state *rasterizer; + + struct i915_fragment_shader *fs; + + struct pipe_blend_color blend_color; + struct pipe_clip_state clip; + struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; + struct pipe_framebuffer_state framebuffer; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct i915_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + + unsigned dirty; + + unsigned num_samplers; + unsigned num_textures; + unsigned num_vertex_elements; + unsigned num_vertex_buffers; + + unsigned *batch_start; + + /** Vertex buffer */ + struct pipe_buffer *vbo; + + struct i915_state current; + unsigned hardware_dirty; + + unsigned debug; +}; + +/* A flag for each state_tracker state object: + */ +#define I915_NEW_VIEWPORT 0x1 +#define I915_NEW_RASTERIZER 0x2 +#define I915_NEW_FS 0x4 +#define I915_NEW_BLEND 0x8 +#define I915_NEW_CLIP 0x10 +#define I915_NEW_SCISSOR 0x20 +#define I915_NEW_STIPPLE 0x40 +#define I915_NEW_FRAMEBUFFER 0x80 +#define I915_NEW_ALPHA_TEST 0x100 +#define I915_NEW_DEPTH_STENCIL 0x200 +#define I915_NEW_SAMPLER 0x400 +#define I915_NEW_TEXTURE 0x800 +#define I915_NEW_CONSTANTS 0x1000 +#define I915_NEW_VBO 0x2000 +#define I915_NEW_VS 0x4000 + + +/* Driver's internally generated state flags: + */ +#define I915_NEW_VERTEX_FORMAT 0x10000 + + +/* Dirty flags for hardware emit + */ +#define I915_HW_STATIC (1< - */ - - -#ifndef BRWCONTEXT_INC -#define BRWCONTEXT_INC - - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" - -#include "tgsi/util/tgsi_scan.h" - -#include "brw_structs.h" -#include "brw_winsys.h" - - -/* Glossary: - * - * URB - uniform resource buffer. A mid-sized buffer which is - * partitioned between the fixed function units and used for passing - * values (vertices, primitives, constants) between them. - * - * CURBE - constant URB entry. An urb region (entry) used to hold - * constant values which the fixed function units can be instructed to - * preload into the GRF when spawining a thread. - * - * VUE - vertex URB entry. An urb entry holding a vertex and usually - * a vertex header. The header contains control information and - * things like primitive type, Begin/end flags and clip codes. - * - * PUE - primitive URB entry. An urb entry produced by the setup (SF) - * unit holding rasterization and interpolation parameters. - * - * GRF - general register file. One of several register files - * addressable by programmed threads. The inputs (r0, payload, curbe, - * urb) of the thread are preloaded to this area before the thread is - * spawned. The registers are individually 8 dwords wide and suitable - * for general usage. Registers holding thread input values are not - * special and may be overwritten. - * - * MRF - message register file. Threads communicate (and terminate) - * by sending messages. Message parameters are placed in contigous - * MRF registers. All program output is via these messages. URB - * entries are populated by sending a message to the shared URB - * function containing the new data, together with a control word, - * often an unmodified copy of R0. - * - * R0 - GRF register 0. Typically holds control information used when - * sending messages to other threads. - * - * EU or GEN4 EU: The name of the programmable subsystem of the - * i965 hardware. Threads are executed by the EU, the registers - * described above are part of the EU architecture. - * - * Fixed function units: - * - * CS - Command streamer. Notional first unit, little software - * interaction. Holds the URB entries used for constant data, ie the - * CURBEs. - * - * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of - * this unit is responsible for pulling vertices out of vertex buffers - * in vram and injecting them into the processing pipe as VUEs. If - * enabled, it first passes them to a VS thread which is a good place - * for the driver to implement any active vertex shader. - * - * GS - Geometry Shader. This corresponds to a new DX10 concept. If - * enabled, incoming strips etc are passed to GS threads in individual - * line/triangle/point units. The GS thread may perform arbitary - * computation and emit whatever primtives with whatever vertices it - * chooses. This makes GS an excellent place to implement GL's - * unfilled polygon modes, though of course it is capable of much - * more. Additionally, GS is used to translate away primitives not - * handled by latter units, including Quads and Lineloops. - * - * CS - Clipper. Mesa's clipping algorithms are imported to run on - * this unit. The fixed function part performs cliptesting against - * the 6 fixed clipplanes and makes descisions on whether or not the - * incoming primitive needs to be passed to a thread for clipping. - * User clip planes are handled via cooperation with the VS thread. - * - * SF - Strips Fans or Setup: Triangles are prepared for - * rasterization. Interpolation coefficients are calculated. - * Flatshading and two-side lighting usually performed here. - * - * WM - Windower. Interpolation of vertex attributes performed here. - * Fragment shader implemented here. SIMD aspects of EU taken full - * advantage of, as pixels are processed in blocks of 16. - * - * CC - Color Calculator. No EU threads associated with this unit. - * Handles blending and (presumably) depth and stencil testing. - */ - -#define BRW_MAX_CURBE (32*16) - -struct brw_context; -struct brw_winsys; - - -/* Raised when we receive new state across the pipe interface: - */ -#define BRW_NEW_VIEWPORT 0x1 -#define BRW_NEW_RASTERIZER 0x2 -#define BRW_NEW_FS 0x4 -#define BRW_NEW_BLEND 0x8 -#define BRW_NEW_CLIP 0x10 -#define BRW_NEW_SCISSOR 0x20 -#define BRW_NEW_STIPPLE 0x40 -#define BRW_NEW_FRAMEBUFFER 0x80 -#define BRW_NEW_ALPHA_TEST 0x100 -#define BRW_NEW_DEPTH_STENCIL 0x200 -#define BRW_NEW_SAMPLER 0x400 -#define BRW_NEW_TEXTURE 0x800 -#define BRW_NEW_CONSTANTS 0x1000 -#define BRW_NEW_VBO 0x2000 -#define BRW_NEW_VS 0x4000 - -/* Raised for other internal events: - */ -#define BRW_NEW_URB_FENCE 0x10000 -#define BRW_NEW_PSP 0x20000 -#define BRW_NEW_CURBE_OFFSETS 0x40000 -#define BRW_NEW_REDUCED_PRIMITIVE 0x80000 -#define BRW_NEW_PRIMITIVE 0x100000 -#define BRW_NEW_SCENE 0x200000 -#define BRW_NEW_SF_LINKAGE 0x400000 - -extern int BRW_DEBUG; - -#define DEBUG_TEXTURE 0x1 -#define DEBUG_STATE 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_PRIMS 0x8 -#define DEBUG_VERTS 0x10 -#define DEBUG_FALLBACKS 0x20 -#define DEBUG_VERBOSE 0x40 -#define DEBUG_DRI 0x80 -#define DEBUG_DMA 0x100 -#define DEBUG_SANITY 0x200 -#define DEBUG_SYNC 0x400 -#define DEBUG_SLEEP 0x800 -#define DEBUG_PIXEL 0x1000 -#define DEBUG_STATS 0x2000 -#define DEBUG_TILE 0x4000 -#define DEBUG_SINGLE_THREAD 0x8000 -#define DEBUG_WM 0x10000 -#define DEBUG_URB 0x20000 -#define DEBUG_VS 0x40000 -#define DEBUG_BATCH 0x80000 -#define DEBUG_BUFMGR 0x100000 -#define DEBUG_BLIT 0x200000 -#define DEBUG_REGION 0x400000 -#define DEBUG_MIPTREE 0x800000 - -#define DBG(...) do { \ - if (BRW_DEBUG & FILE_DEBUG_FLAG) \ - debug_printf(__VA_ARGS__); \ -} while(0) - -#define PRINT(...) do { \ - debug_printf(brw->pipe.winsys, __VA_ARGS__); \ -} while(0) - -struct brw_state_flags { - unsigned cache; - unsigned brw; -}; - - -struct brw_vertex_program { - struct pipe_shader_state program; - struct tgsi_shader_info info; - int id; -}; - - -struct brw_fragment_program { - struct pipe_shader_state program; - struct tgsi_shader_info info; - - boolean UsesDepth; /* XXX add this to tgsi_shader_info? */ - int id; -}; - - -struct pipe_setup_linkage { - struct { - unsigned vp_output:5; - unsigned interp_mode:4; - unsigned bf_vp_output:5; - } fp_input[PIPE_MAX_SHADER_INPUTS]; - - unsigned fp_input_count:5; - unsigned max_vp_output:5; -}; - - - -struct brw_texture { - struct pipe_texture base; - - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ - unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; -}; - -/* Data about a particular attempt to compile a program. Note that - * there can be many of these, each in a different GL state - * corresponding to a different brw_wm_prog_key struct, with different - * compiled programs: - */ -/* Data about a particular attempt to compile a program. Note that - * there can be many of these, each in a different GL state - * corresponding to a different brw_wm_prog_key struct, with different - * compiled programs: - */ - -struct brw_wm_prog_data { - unsigned curb_read_length; - unsigned urb_read_length; - - unsigned first_curbe_grf; - unsigned total_grf; - unsigned total_scratch; - - /* Internally generated constants for the CURBE. These are loaded - * ahead of the data from the constant buffer. - */ - const float internal_const[8]; - unsigned nr_internal_consts; - unsigned max_const; - - boolean error; -}; - -struct brw_sf_prog_data { - unsigned urb_read_length; - unsigned total_grf; - - /* Each vertex may have upto 12 attributes, 4 components each, - * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 - * rows. - * - * Actually we use 4 for each, so call it 12 rows. - */ - unsigned urb_entry_size; -}; - -struct brw_clip_prog_data { - unsigned curb_read_length; /* user planes? */ - unsigned clip_mode; - unsigned urb_read_length; - unsigned total_grf; -}; - -struct brw_gs_prog_data { - unsigned urb_read_length; - unsigned total_grf; -}; - -struct brw_vs_prog_data { - unsigned curb_read_length; - unsigned urb_read_length; - unsigned total_grf; - unsigned outputs_written; - - unsigned inputs_read; - - unsigned max_const; - - float imm_buf[PIPE_MAX_CONSTANT][4]; - unsigned num_imm; - unsigned num_consts; - - /* Used for calculating urb partitions: - */ - unsigned urb_entry_size; -}; - - -#define BRW_MAX_TEX_UNIT 8 -#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 - -/* Create a fixed sized struct for caching binding tables: - */ -struct brw_surface_binding_table { - unsigned surf_ss_offset[BRW_WM_MAX_SURF]; -}; - - -struct brw_cache; - -struct brw_mem_pool { - struct pipe_buffer *buffer; - - unsigned size; - unsigned offset; /* offset of first free byte */ - - struct brw_context *brw; -}; - -struct brw_cache_item { - unsigned hash; - unsigned key_size; /* for variable-sized keys */ - const void *key; - - unsigned offset; /* offset within pool's buffer */ - unsigned data_size; - - struct brw_cache_item *next; -}; - - - -struct brw_cache { - unsigned id; - - const char *name; - - struct brw_context *brw; - struct brw_mem_pool *pool; - - struct brw_cache_item **items; - unsigned size, n_items; - - unsigned key_size; /* for fixed-size keys */ - unsigned aux_size; - - unsigned last_addr; /* offset of active item */ -}; - - - - -/* Considered adding a member to this struct to document which flags - * an update might raise so that ordering of the state atoms can be - * checked or derived at runtime. Dropped the idea in favor of having - * a debug mode where the state is monitored for flags which are - * raised that have already been tested against. - */ -struct brw_tracked_state { - struct brw_state_flags dirty; - void (*update)( struct brw_context *brw ); -}; - - -/* Flags for brw->state.cache. - */ -#define CACHE_NEW_CC_VP (1< 32. Wouldn't life - * be easier if C allowed arrays of packed elements? - */ -#define ATTRIB_BIT_DWORDS ((PIPE_MAX_ATTRIBS+31)/32) - - - - -struct brw_vertex_info { - unsigned varying; /* varying:1[PIPE_MAX_ATTRIBS] */ - unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_MAX_ATTRIBS] */ -}; - - - - - -struct brw_context -{ - struct pipe_context pipe; - struct brw_winsys *winsys; - - unsigned primitive; - unsigned reduced_primitive; - - boolean emit_state_always; - - struct { - struct brw_state_flags dirty; - } state; - - - struct { - const struct pipe_blend_state *Blend; - const struct pipe_depth_stencil_alpha_state *DepthStencil; - const struct pipe_poly_stipple *PolygonStipple; - const struct pipe_rasterizer_state *Raster; - const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; - const struct brw_vertex_program *VertexProgram; - const struct brw_fragment_program *FragmentProgram; - - struct pipe_clip_state Clip; - struct pipe_blend_color BlendColor; - struct pipe_scissor_state Scissor; - struct pipe_viewport_state Viewport; - struct pipe_framebuffer_state FrameBuffer; - - const struct pipe_constant_buffer *Constants[2]; - const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; - } attribs; - - unsigned num_samplers; - unsigned num_textures; - - struct brw_mem_pool pool[BRW_MAX_POOL]; - struct brw_cache cache[BRW_MAX_CACHE]; - struct brw_cached_batch_item *cached_batch_items; - - struct { - - /* Arrays with buffer objects to copy non-bufferobj arrays into - * for upload: - */ - const struct pipe_vertex_buffer *vbo_array[PIPE_MAX_ATTRIBS]; - - struct brw_vertex_element_state inputs[PIPE_MAX_ATTRIBS]; - -#define BRW_NR_UPLOAD_BUFS 17 -#define BRW_UPLOAD_INIT_SIZE (128*1024) - - /* Summary of size and varying of active arrays, so we can check - * for changes to this state: - */ - struct brw_vertex_info info; - } vb; - - - unsigned hardware_dirty; - unsigned dirty; - unsigned pci_id; - /* BRW_NEW_URB_ALLOCATIONS: - */ - struct { - unsigned vsize; /* vertex size plus header in urb registers */ - unsigned csize; /* constant buffer size in urb registers */ - unsigned sfsize; /* setup data size in urb registers */ - - boolean constrained; - - unsigned nr_vs_entries; - unsigned nr_gs_entries; - unsigned nr_clip_entries; - unsigned nr_sf_entries; - unsigned nr_cs_entries; - -/* unsigned vs_size; */ -/* unsigned gs_size; */ -/* unsigned clip_size; */ -/* unsigned sf_size; */ -/* unsigned cs_size; */ - - unsigned vs_start; - unsigned gs_start; - unsigned clip_start; - unsigned sf_start; - unsigned cs_start; - } urb; - - - /* BRW_NEW_CURBE_OFFSETS: - */ - struct { - unsigned wm_start; - unsigned wm_size; - unsigned clip_start; - unsigned clip_size; - unsigned vs_start; - unsigned vs_size; - unsigned total_size; - - unsigned gs_offset; - - float *last_buf; - unsigned last_bufsz; - } curbe; - - struct { - struct brw_vs_prog_data *prog_data; - - unsigned prog_gs_offset; - unsigned state_gs_offset; - } vs; - - struct { - struct brw_gs_prog_data *prog_data; - - boolean prog_active; - unsigned prog_gs_offset; - unsigned state_gs_offset; - } gs; - - struct { - struct brw_clip_prog_data *prog_data; - - unsigned prog_gs_offset; - unsigned vp_gs_offset; - unsigned state_gs_offset; - } clip; - - - struct { - struct brw_sf_prog_data *prog_data; - - struct pipe_setup_linkage linkage; - - unsigned prog_gs_offset; - unsigned vp_gs_offset; - unsigned state_gs_offset; - } sf; - - struct { - struct brw_wm_prog_data *prog_data; - -// struct brw_wm_compiler *compile_data; - - - /** - * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER - * cache - */ - struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; - - unsigned render_surf; - unsigned nr_surfaces; - - unsigned max_threads; - struct pipe_buffer *scratch_buffer; - unsigned scratch_buffer_size; - - unsigned sampler_count; - unsigned sampler_gs_offset; - - struct brw_surface_binding_table bind; - unsigned bind_ss_offset; - - unsigned prog_gs_offset; - unsigned state_gs_offset; - } wm; - - - struct { - unsigned vp_gs_offset; - unsigned state_gs_offset; - } cc; - - - /* Used to give every program string a unique id - */ - unsigned program_id; -}; - - -#define BRW_PACKCOLOR8888(r,g,b,a) ((r<<24) | (g<<16) | (b<<8) | a) - - -/*====================================================================== - * brw_vtbl.c - */ -void brw_do_flush( struct brw_context *brw, - unsigned flags ); - - -/*====================================================================== - * brw_state.c - */ -void brw_validate_state(struct brw_context *brw); -void brw_init_state(struct brw_context *brw); -void brw_destroy_state(struct brw_context *brw); - - -/*====================================================================== - * brw_tex.c - */ -void brwUpdateTextureState( struct brw_context *brw ); - - -/* brw_urb.c - */ -void brw_upload_urb_fence(struct brw_context *brw); - -void brw_upload_constant_buffer_state(struct brw_context *brw); - -void brw_init_surface_functions(struct brw_context *brw); -void brw_init_state_functions(struct brw_context *brw); -void brw_init_flush_functions(struct brw_context *brw); -void brw_init_string_functions(struct brw_context *brw); - -/*====================================================================== - * Inline conversion functions. These are better-typed than the - * macros used previously: - */ -static inline struct brw_context * -brw_context( struct pipe_context *ctx ) -{ - return (struct brw_context *)ctx; -} - -#endif - +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRWCONTEXT_INC +#define BRWCONTEXT_INC + + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "tgsi/util/tgsi_scan.h" + +#include "brw_structs.h" +#include "brw_winsys.h" + + +/* Glossary: + * + * URB - uniform resource buffer. A mid-sized buffer which is + * partitioned between the fixed function units and used for passing + * values (vertices, primitives, constants) between them. + * + * CURBE - constant URB entry. An urb region (entry) used to hold + * constant values which the fixed function units can be instructed to + * preload into the GRF when spawining a thread. + * + * VUE - vertex URB entry. An urb entry holding a vertex and usually + * a vertex header. The header contains control information and + * things like primitive type, Begin/end flags and clip codes. + * + * PUE - primitive URB entry. An urb entry produced by the setup (SF) + * unit holding rasterization and interpolation parameters. + * + * GRF - general register file. One of several register files + * addressable by programmed threads. The inputs (r0, payload, curbe, + * urb) of the thread are preloaded to this area before the thread is + * spawned. The registers are individually 8 dwords wide and suitable + * for general usage. Registers holding thread input values are not + * special and may be overwritten. + * + * MRF - message register file. Threads communicate (and terminate) + * by sending messages. Message parameters are placed in contigous + * MRF registers. All program output is via these messages. URB + * entries are populated by sending a message to the shared URB + * function containing the new data, together with a control word, + * often an unmodified copy of R0. + * + * R0 - GRF register 0. Typically holds control information used when + * sending messages to other threads. + * + * EU or GEN4 EU: The name of the programmable subsystem of the + * i965 hardware. Threads are executed by the EU, the registers + * described above are part of the EU architecture. + * + * Fixed function units: + * + * CS - Command streamer. Notional first unit, little software + * interaction. Holds the URB entries used for constant data, ie the + * CURBEs. + * + * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of + * this unit is responsible for pulling vertices out of vertex buffers + * in vram and injecting them into the processing pipe as VUEs. If + * enabled, it first passes them to a VS thread which is a good place + * for the driver to implement any active vertex shader. + * + * GS - Geometry Shader. This corresponds to a new DX10 concept. If + * enabled, incoming strips etc are passed to GS threads in individual + * line/triangle/point units. The GS thread may perform arbitary + * computation and emit whatever primtives with whatever vertices it + * chooses. This makes GS an excellent place to implement GL's + * unfilled polygon modes, though of course it is capable of much + * more. Additionally, GS is used to translate away primitives not + * handled by latter units, including Quads and Lineloops. + * + * CS - Clipper. Mesa's clipping algorithms are imported to run on + * this unit. The fixed function part performs cliptesting against + * the 6 fixed clipplanes and makes descisions on whether or not the + * incoming primitive needs to be passed to a thread for clipping. + * User clip planes are handled via cooperation with the VS thread. + * + * SF - Strips Fans or Setup: Triangles are prepared for + * rasterization. Interpolation coefficients are calculated. + * Flatshading and two-side lighting usually performed here. + * + * WM - Windower. Interpolation of vertex attributes performed here. + * Fragment shader implemented here. SIMD aspects of EU taken full + * advantage of, as pixels are processed in blocks of 16. + * + * CC - Color Calculator. No EU threads associated with this unit. + * Handles blending and (presumably) depth and stencil testing. + */ + +#define BRW_MAX_CURBE (32*16) + +struct brw_context; +struct brw_winsys; + + +/* Raised when we receive new state across the pipe interface: + */ +#define BRW_NEW_VIEWPORT 0x1 +#define BRW_NEW_RASTERIZER 0x2 +#define BRW_NEW_FS 0x4 +#define BRW_NEW_BLEND 0x8 +#define BRW_NEW_CLIP 0x10 +#define BRW_NEW_SCISSOR 0x20 +#define BRW_NEW_STIPPLE 0x40 +#define BRW_NEW_FRAMEBUFFER 0x80 +#define BRW_NEW_ALPHA_TEST 0x100 +#define BRW_NEW_DEPTH_STENCIL 0x200 +#define BRW_NEW_SAMPLER 0x400 +#define BRW_NEW_TEXTURE 0x800 +#define BRW_NEW_CONSTANTS 0x1000 +#define BRW_NEW_VBO 0x2000 +#define BRW_NEW_VS 0x4000 + +/* Raised for other internal events: + */ +#define BRW_NEW_URB_FENCE 0x10000 +#define BRW_NEW_PSP 0x20000 +#define BRW_NEW_CURBE_OFFSETS 0x40000 +#define BRW_NEW_REDUCED_PRIMITIVE 0x80000 +#define BRW_NEW_PRIMITIVE 0x100000 +#define BRW_NEW_SCENE 0x200000 +#define BRW_NEW_SF_LINKAGE 0x400000 + +extern int BRW_DEBUG; + +#define DEBUG_TEXTURE 0x1 +#define DEBUG_STATE 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_PRIMS 0x8 +#define DEBUG_VERTS 0x10 +#define DEBUG_FALLBACKS 0x20 +#define DEBUG_VERBOSE 0x40 +#define DEBUG_DRI 0x80 +#define DEBUG_DMA 0x100 +#define DEBUG_SANITY 0x200 +#define DEBUG_SYNC 0x400 +#define DEBUG_SLEEP 0x800 +#define DEBUG_PIXEL 0x1000 +#define DEBUG_STATS 0x2000 +#define DEBUG_TILE 0x4000 +#define DEBUG_SINGLE_THREAD 0x8000 +#define DEBUG_WM 0x10000 +#define DEBUG_URB 0x20000 +#define DEBUG_VS 0x40000 +#define DEBUG_BATCH 0x80000 +#define DEBUG_BUFMGR 0x100000 +#define DEBUG_BLIT 0x200000 +#define DEBUG_REGION 0x400000 +#define DEBUG_MIPTREE 0x800000 + +#define DBG(...) do { \ + if (BRW_DEBUG & FILE_DEBUG_FLAG) \ + debug_printf(__VA_ARGS__); \ +} while(0) + +#define PRINT(...) do { \ + debug_printf(brw->pipe.winsys, __VA_ARGS__); \ +} while(0) + +struct brw_state_flags { + unsigned cache; + unsigned brw; +}; + + +struct brw_vertex_program { + struct pipe_shader_state program; + struct tgsi_shader_info info; + int id; +}; + + +struct brw_fragment_program { + struct pipe_shader_state program; + struct tgsi_shader_info info; + + boolean UsesDepth; /* XXX add this to tgsi_shader_info? */ + int id; +}; + + +struct pipe_setup_linkage { + struct { + unsigned vp_output:5; + unsigned interp_mode:4; + unsigned bf_vp_output:5; + } fp_input[PIPE_MAX_SHADER_INPUTS]; + + unsigned fp_input_count:5; + unsigned max_vp_output:5; +}; + + + +struct brw_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; +}; + +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ + +struct brw_wm_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + + unsigned first_curbe_grf; + unsigned total_grf; + unsigned total_scratch; + + /* Internally generated constants for the CURBE. These are loaded + * ahead of the data from the constant buffer. + */ + const float internal_const[8]; + unsigned nr_internal_consts; + unsigned max_const; + + boolean error; +}; + +struct brw_sf_prog_data { + unsigned urb_read_length; + unsigned total_grf; + + /* Each vertex may have upto 12 attributes, 4 components each, + * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 + * rows. + * + * Actually we use 4 for each, so call it 12 rows. + */ + unsigned urb_entry_size; +}; + +struct brw_clip_prog_data { + unsigned curb_read_length; /* user planes? */ + unsigned clip_mode; + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_gs_prog_data { + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_vs_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + unsigned total_grf; + unsigned outputs_written; + + unsigned inputs_read; + + unsigned max_const; + + float imm_buf[PIPE_MAX_CONSTANT][4]; + unsigned num_imm; + unsigned num_consts; + + /* Used for calculating urb partitions: + */ + unsigned urb_entry_size; +}; + + +#define BRW_MAX_TEX_UNIT 8 +#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 + +/* Create a fixed sized struct for caching binding tables: + */ +struct brw_surface_binding_table { + unsigned surf_ss_offset[BRW_WM_MAX_SURF]; +}; + + +struct brw_cache; + +struct brw_mem_pool { + struct pipe_buffer *buffer; + + unsigned size; + unsigned offset; /* offset of first free byte */ + + struct brw_context *brw; +}; + +struct brw_cache_item { + unsigned hash; + unsigned key_size; /* for variable-sized keys */ + const void *key; + + unsigned offset; /* offset within pool's buffer */ + unsigned data_size; + + struct brw_cache_item *next; +}; + + + +struct brw_cache { + unsigned id; + + const char *name; + + struct brw_context *brw; + struct brw_mem_pool *pool; + + struct brw_cache_item **items; + unsigned size, n_items; + + unsigned key_size; /* for fixed-size keys */ + unsigned aux_size; + + unsigned last_addr; /* offset of active item */ +}; + + + + +/* Considered adding a member to this struct to document which flags + * an update might raise so that ordering of the state atoms can be + * checked or derived at runtime. Dropped the idea in favor of having + * a debug mode where the state is monitored for flags which are + * raised that have already been tested against. + */ +struct brw_tracked_state { + struct brw_state_flags dirty; + void (*update)( struct brw_context *brw ); +}; + + +/* Flags for brw->state.cache. + */ +#define CACHE_NEW_CC_VP (1< 32. Wouldn't life + * be easier if C allowed arrays of packed elements? + */ +#define ATTRIB_BIT_DWORDS ((PIPE_MAX_ATTRIBS+31)/32) + + + + +struct brw_vertex_info { + unsigned varying; /* varying:1[PIPE_MAX_ATTRIBS] */ + unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_MAX_ATTRIBS] */ +}; + + + + + +struct brw_context +{ + struct pipe_context pipe; + struct brw_winsys *winsys; + + unsigned primitive; + unsigned reduced_primitive; + + boolean emit_state_always; + + struct { + struct brw_state_flags dirty; + } state; + + + struct { + const struct pipe_blend_state *Blend; + const struct pipe_depth_stencil_alpha_state *DepthStencil; + const struct pipe_poly_stipple *PolygonStipple; + const struct pipe_rasterizer_state *Raster; + const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; + const struct brw_vertex_program *VertexProgram; + const struct brw_fragment_program *FragmentProgram; + + struct pipe_clip_state Clip; + struct pipe_blend_color BlendColor; + struct pipe_scissor_state Scissor; + struct pipe_viewport_state Viewport; + struct pipe_framebuffer_state FrameBuffer; + + const struct pipe_constant_buffer *Constants[2]; + const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; + } attribs; + + unsigned num_samplers; + unsigned num_textures; + + struct brw_mem_pool pool[BRW_MAX_POOL]; + struct brw_cache cache[BRW_MAX_CACHE]; + struct brw_cached_batch_item *cached_batch_items; + + struct { + + /* Arrays with buffer objects to copy non-bufferobj arrays into + * for upload: + */ + const struct pipe_vertex_buffer *vbo_array[PIPE_MAX_ATTRIBS]; + + struct brw_vertex_element_state inputs[PIPE_MAX_ATTRIBS]; + +#define BRW_NR_UPLOAD_BUFS 17 +#define BRW_UPLOAD_INIT_SIZE (128*1024) + + /* Summary of size and varying of active arrays, so we can check + * for changes to this state: + */ + struct brw_vertex_info info; + } vb; + + + unsigned hardware_dirty; + unsigned dirty; + unsigned pci_id; + /* BRW_NEW_URB_ALLOCATIONS: + */ + struct { + unsigned vsize; /* vertex size plus header in urb registers */ + unsigned csize; /* constant buffer size in urb registers */ + unsigned sfsize; /* setup data size in urb registers */ + + boolean constrained; + + unsigned nr_vs_entries; + unsigned nr_gs_entries; + unsigned nr_clip_entries; + unsigned nr_sf_entries; + unsigned nr_cs_entries; + +/* unsigned vs_size; */ +/* unsigned gs_size; */ +/* unsigned clip_size; */ +/* unsigned sf_size; */ +/* unsigned cs_size; */ + + unsigned vs_start; + unsigned gs_start; + unsigned clip_start; + unsigned sf_start; + unsigned cs_start; + } urb; + + + /* BRW_NEW_CURBE_OFFSETS: + */ + struct { + unsigned wm_start; + unsigned wm_size; + unsigned clip_start; + unsigned clip_size; + unsigned vs_start; + unsigned vs_size; + unsigned total_size; + + unsigned gs_offset; + + float *last_buf; + unsigned last_bufsz; + } curbe; + + struct { + struct brw_vs_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } vs; + + struct { + struct brw_gs_prog_data *prog_data; + + boolean prog_active; + unsigned prog_gs_offset; + unsigned state_gs_offset; + } gs; + + struct { + struct brw_clip_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } clip; + + + struct { + struct brw_sf_prog_data *prog_data; + + struct pipe_setup_linkage linkage; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } sf; + + struct { + struct brw_wm_prog_data *prog_data; + +// struct brw_wm_compiler *compile_data; + + + /** + * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER + * cache + */ + struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; + + unsigned render_surf; + unsigned nr_surfaces; + + unsigned max_threads; + struct pipe_buffer *scratch_buffer; + unsigned scratch_buffer_size; + + unsigned sampler_count; + unsigned sampler_gs_offset; + + struct brw_surface_binding_table bind; + unsigned bind_ss_offset; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } wm; + + + struct { + unsigned vp_gs_offset; + unsigned state_gs_offset; + } cc; + + + /* Used to give every program string a unique id + */ + unsigned program_id; +}; + + +#define BRW_PACKCOLOR8888(r,g,b,a) ((r<<24) | (g<<16) | (b<<8) | a) + + +/*====================================================================== + * brw_vtbl.c + */ +void brw_do_flush( struct brw_context *brw, + unsigned flags ); + + +/*====================================================================== + * brw_state.c + */ +void brw_validate_state(struct brw_context *brw); +void brw_init_state(struct brw_context *brw); +void brw_destroy_state(struct brw_context *brw); + + +/*====================================================================== + * brw_tex.c + */ +void brwUpdateTextureState( struct brw_context *brw ); + + +/* brw_urb.c + */ +void brw_upload_urb_fence(struct brw_context *brw); + +void brw_upload_constant_buffer_state(struct brw_context *brw); + +void brw_init_surface_functions(struct brw_context *brw); +void brw_init_state_functions(struct brw_context *brw); +void brw_init_flush_functions(struct brw_context *brw); +void brw_init_string_functions(struct brw_context *brw); + +/*====================================================================== + * Inline conversion functions. These are better-typed than the + * macros used previously: + */ +static inline struct brw_context * +brw_context( struct pipe_context *ctx ) +{ + return (struct brw_context *)ctx; +} + +#endif + diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index ac243b7e4f..caeeba4630 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -1,469 +1,469 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Zack Rusin - * Keith Whitwell - */ - - -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_dump.h" -#include "tgsi/util/tgsi_parse.h" - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_state.h" -#include "brw_draw.h" - - -#define DUP( TYPE, VAL ) \ -do { \ - struct TYPE *x = malloc(sizeof(*x)); \ - memcpy(x, VAL, sizeof(*x) ); \ - return x; \ -} while (0) - -/************************************************************************ - * Blend - */ -static void * -brw_create_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend) -{ - DUP( pipe_blend_state, blend ); -} - -static void brw_bind_blend_state(struct pipe_context *pipe, - void *blend) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Blend = (struct pipe_blend_state*)blend; - brw->state.dirty.brw |= BRW_NEW_BLEND; -} - - -static void brw_delete_blend_state(struct pipe_context *pipe, void *blend) -{ - free(blend); -} - -static void brw_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.BlendColor = *blend_color; - - brw->state.dirty.brw |= BRW_NEW_BLEND; -} - -/************************************************************************ - * Sampler - */ - -static void * -brw_create_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) -{ - DUP( pipe_sampler_state, sampler ); -} - -static void brw_bind_sampler_states(struct pipe_context *pipe, - unsigned num, void **sampler) -{ - struct brw_context *brw = brw_context(pipe); - - assert(num <= PIPE_MAX_SAMPLERS); - - /* Check for no-op */ - if (num == brw->num_samplers && - !memcmp(brw->attribs.Samplers, sampler, num * sizeof(void *))) - return; - - memcpy(brw->attribs.Samplers, sampler, num * sizeof(void *)); - memset(&brw->attribs.Samplers[num], 0, (PIPE_MAX_SAMPLERS - num) * - sizeof(void *)); - - brw->num_samplers = num; - - brw->state.dirty.brw |= BRW_NEW_SAMPLER; -} - -static void brw_delete_sampler_state(struct pipe_context *pipe, - void *sampler) -{ - free(sampler); -} - - -/************************************************************************ - * Depth stencil - */ - -static void * -brw_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) -{ - DUP( pipe_depth_stencil_alpha_state, depth_stencil ); -} - -static void brw_bind_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; - - brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; -} - -static void brw_delete_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - free(depth_stencil); -} - -/************************************************************************ - * Scissor - */ -static void brw_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct brw_context *brw = brw_context(pipe); - - memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) ); - brw->state.dirty.brw |= BRW_NEW_SCISSOR; -} - - -/************************************************************************ - * Stipple - */ - -static void brw_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ -} - - -/************************************************************************ - * Fragment shader - */ - -static void * brw_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *shader) -{ - struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); - - brw_fp->program.tokens = tgsi_dup_tokens(shader->tokens); - brw_fp->id = brw_context(pipe)->program_id++; - - tgsi_scan_shader(shader->tokens, &brw_fp->info); - -#if 0 - brw_shader_info(shader->tokens, - &brw_fp->info2); -#endif - - tgsi_dump(shader->tokens, 0); - - - return (void *)brw_fp; -} - -static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader; - brw->state.dirty.brw |= BRW_NEW_FS; -} - -static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) -{ - struct brw_fragment_program *brw_fp = (struct brw_fragment_program *) shader; - - FREE((void *) brw_fp->program.tokens); - FREE(brw_fp); -} - - -/************************************************************************ - * Vertex shader and other TNL state - */ - -static void *brw_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *shader) -{ - struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); - - brw_vp->program.tokens = tgsi_dup_tokens(shader->tokens); - brw_vp->id = brw_context(pipe)->program_id++; - - tgsi_scan_shader(shader->tokens, &brw_vp->info); - -#if 0 - brw_shader_info(shader->tokens, - &brw_vp->info2); -#endif - tgsi_dump(shader->tokens, 0); - - return (void *)brw_vp; -} - -static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; - brw->state.dirty.brw |= BRW_NEW_VS; - - debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); -} - -static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) -{ - struct brw_vertex_program *brw_vp = (struct brw_vertex_program *) shader; - - FREE((void *) brw_vp->program.tokens); - FREE(brw_vp); -} - - -static void brw_set_clip_state( struct pipe_context *pipe, - const struct pipe_clip_state *clip ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Clip = *clip; -} - - -static void brw_set_viewport_state( struct pipe_context *pipe, - const struct pipe_viewport_state *viewport ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Viewport = *viewport; /* struct copy */ - brw->state.dirty.brw |= BRW_NEW_VIEWPORT; - - /* pass the viewport info to the draw module */ - //draw_set_viewport_state(brw->draw, viewport); -} - - -static void brw_set_vertex_buffers(struct pipe_context *pipe, - unsigned count, - const struct pipe_vertex_buffer *buffers) -{ - struct brw_context *brw = brw_context(pipe); - memcpy(brw->vb.vbo_array, buffers, count * sizeof(buffers[0])); -} - -static void brw_set_vertex_elements(struct pipe_context *pipe, - unsigned count, - const struct pipe_vertex_element *elements) -{ - /* flush ? */ - struct brw_context *brw = brw_context(pipe); - uint i; - - assert(count <= PIPE_MAX_ATTRIBS); - - for (i = 0; i < count; i++) { - struct brw_vertex_element_state el; - memset(&el, 0, sizeof(el)); - - el.ve0.src_offset = elements[i].src_offset; - el.ve0.src_format = brw_translate_surface_format(elements[i].src_format); - el.ve0.valid = 1; - el.ve0.vertex_buffer_index = elements[i].vertex_buffer_index; - - el.ve1.dst_offset = i * 4; - - el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; - - switch (elements[i].nr_components) { - case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; - case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; - case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; - break; - } - - brw->vb.inputs[i] = el; - } -} - - - -/************************************************************************ - * Constant buffers - */ - -static void brw_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) -{ - struct brw_context *brw = brw_context(pipe); - - assert(buf == 0 || index == 0); - - brw->attribs.Constants[shader] = buf; - brw->state.dirty.brw |= BRW_NEW_CONSTANTS; -} - - -/************************************************************************ - * Texture surfaces - */ - - -static void brw_set_sampler_textures(struct pipe_context *pipe, - unsigned num, - struct pipe_texture **texture) -{ - struct brw_context *brw = brw_context(pipe); - uint i; - - assert(num <= PIPE_MAX_SAMPLERS); - - /* Check for no-op */ - if (num == brw->num_textures && - !memcmp(brw->attribs.Texture, texture, num * - sizeof(struct pipe_texture *))) - return; - - for (i = 0; i < num; i++) - pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], - texture[i]); - - for (i = num; i < brw->num_textures; i++) - pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], - NULL); - - brw->num_textures = num; - - brw->state.dirty.brw |= BRW_NEW_TEXTURE; -} - - -/************************************************************************ - * Render targets, etc - */ - -static void brw_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.FrameBuffer = *fb; /* struct copy */ - - brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER; -} - - - -/************************************************************************ - * Rasterizer state - */ - -static void * -brw_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *rasterizer) -{ - DUP(pipe_rasterizer_state, rasterizer); -} - -static void brw_bind_rasterizer_state( struct pipe_context *pipe, - void *setup ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Raster = (struct pipe_rasterizer_state *)setup; - - /* Also pass-through to draw module: - */ - //draw_set_rasterizer_state(brw->draw, setup); - - brw->state.dirty.brw |= BRW_NEW_RASTERIZER; -} - -static void brw_delete_rasterizer_state(struct pipe_context *pipe, - void *setup) -{ - free(setup); -} - - - -void -brw_init_state_functions( struct brw_context *brw ) -{ - brw->pipe.create_blend_state = brw_create_blend_state; - brw->pipe.bind_blend_state = brw_bind_blend_state; - brw->pipe.delete_blend_state = brw_delete_blend_state; - - brw->pipe.create_sampler_state = brw_create_sampler_state; - brw->pipe.bind_sampler_states = brw_bind_sampler_states; - brw->pipe.delete_sampler_state = brw_delete_sampler_state; - - brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; - brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; - brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; - - brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; - brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; - brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state; - brw->pipe.create_fs_state = brw_create_fs_state; - brw->pipe.bind_fs_state = brw_bind_fs_state; - brw->pipe.delete_fs_state = brw_delete_fs_state; - brw->pipe.create_vs_state = brw_create_vs_state; - brw->pipe.bind_vs_state = brw_bind_vs_state; - brw->pipe.delete_vs_state = brw_delete_vs_state; - - brw->pipe.set_blend_color = brw_set_blend_color; - brw->pipe.set_clip_state = brw_set_clip_state; - brw->pipe.set_constant_buffer = brw_set_constant_buffer; - brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; - -// brw->pipe.set_feedback_state = brw_set_feedback_state; -// brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; - - brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; - brw->pipe.set_scissor_state = brw_set_scissor_state; - brw->pipe.set_sampler_textures = brw_set_sampler_textures; - brw->pipe.set_viewport_state = brw_set_viewport_state; - brw->pipe.set_vertex_buffers = brw_set_vertex_buffers; - brw->pipe.set_vertex_elements = brw_set_vertex_elements; -} +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Zack Rusin + * Keith Whitwell + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_dump.h" +#include "tgsi/util/tgsi_parse.h" + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_state.h" +#include "brw_draw.h" + + +#define DUP( TYPE, VAL ) \ +do { \ + struct TYPE *x = malloc(sizeof(*x)); \ + memcpy(x, VAL, sizeof(*x) ); \ + return x; \ +} while (0) + +/************************************************************************ + * Blend + */ +static void * +brw_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + DUP( pipe_blend_state, blend ); +} + +static void brw_bind_blend_state(struct pipe_context *pipe, + void *blend) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Blend = (struct pipe_blend_state*)blend; + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + + +static void brw_delete_blend_state(struct pipe_context *pipe, void *blend) +{ + free(blend); +} + +static void brw_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.BlendColor = *blend_color; + + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + +/************************************************************************ + * Sampler + */ + +static void * +brw_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + DUP( pipe_sampler_state, sampler ); +} + +static void brw_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **sampler) +{ + struct brw_context *brw = brw_context(pipe); + + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == brw->num_samplers && + !memcmp(brw->attribs.Samplers, sampler, num * sizeof(void *))) + return; + + memcpy(brw->attribs.Samplers, sampler, num * sizeof(void *)); + memset(&brw->attribs.Samplers[num], 0, (PIPE_MAX_SAMPLERS - num) * + sizeof(void *)); + + brw->num_samplers = num; + + brw->state.dirty.brw |= BRW_NEW_SAMPLER; +} + +static void brw_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + free(sampler); +} + + +/************************************************************************ + * Depth stencil + */ + +static void * +brw_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil) +{ + DUP( pipe_depth_stencil_alpha_state, depth_stencil ); +} + +static void brw_bind_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; + + brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; +} + +static void brw_delete_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + free(depth_stencil); +} + +/************************************************************************ + * Scissor + */ +static void brw_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct brw_context *brw = brw_context(pipe); + + memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) ); + brw->state.dirty.brw |= BRW_NEW_SCISSOR; +} + + +/************************************************************************ + * Stipple + */ + +static void brw_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ +} + + +/************************************************************************ + * Fragment shader + */ + +static void * brw_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); + + brw_fp->program.tokens = tgsi_dup_tokens(shader->tokens); + brw_fp->id = brw_context(pipe)->program_id++; + + tgsi_scan_shader(shader->tokens, &brw_fp->info); + +#if 0 + brw_shader_info(shader->tokens, + &brw_fp->info2); +#endif + + tgsi_dump(shader->tokens, 0); + + + return (void *)brw_fp; +} + +static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader; + brw->state.dirty.brw |= BRW_NEW_FS; +} + +static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) +{ + struct brw_fragment_program *brw_fp = (struct brw_fragment_program *) shader; + + FREE((void *) brw_fp->program.tokens); + FREE(brw_fp); +} + + +/************************************************************************ + * Vertex shader and other TNL state + */ + +static void *brw_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); + + brw_vp->program.tokens = tgsi_dup_tokens(shader->tokens); + brw_vp->id = brw_context(pipe)->program_id++; + + tgsi_scan_shader(shader->tokens, &brw_vp->info); + +#if 0 + brw_shader_info(shader->tokens, + &brw_vp->info2); +#endif + tgsi_dump(shader->tokens, 0); + + return (void *)brw_vp; +} + +static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; + brw->state.dirty.brw |= BRW_NEW_VS; + + debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); +} + +static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) +{ + struct brw_vertex_program *brw_vp = (struct brw_vertex_program *) shader; + + FREE((void *) brw_vp->program.tokens); + FREE(brw_vp); +} + + +static void brw_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Clip = *clip; +} + + +static void brw_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Viewport = *viewport; /* struct copy */ + brw->state.dirty.brw |= BRW_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + //draw_set_viewport_state(brw->draw, viewport); +} + + +static void brw_set_vertex_buffers(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_buffer *buffers) +{ + struct brw_context *brw = brw_context(pipe); + memcpy(brw->vb.vbo_array, buffers, count * sizeof(buffers[0])); +} + +static void brw_set_vertex_elements(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_element *elements) +{ + /* flush ? */ + struct brw_context *brw = brw_context(pipe); + uint i; + + assert(count <= PIPE_MAX_ATTRIBS); + + for (i = 0; i < count; i++) { + struct brw_vertex_element_state el; + memset(&el, 0, sizeof(el)); + + el.ve0.src_offset = elements[i].src_offset; + el.ve0.src_format = brw_translate_surface_format(elements[i].src_format); + el.ve0.valid = 1; + el.ve0.vertex_buffer_index = elements[i].vertex_buffer_index; + + el.ve1.dst_offset = i * 4; + + el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; + + switch (elements[i].nr_components) { + case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; + case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; + case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; + break; + } + + brw->vb.inputs[i] = el; + } +} + + + +/************************************************************************ + * Constant buffers + */ + +static void brw_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct brw_context *brw = brw_context(pipe); + + assert(buf == 0 || index == 0); + + brw->attribs.Constants[shader] = buf; + brw->state.dirty.brw |= BRW_NEW_CONSTANTS; +} + + +/************************************************************************ + * Texture surfaces + */ + + +static void brw_set_sampler_textures(struct pipe_context *pipe, + unsigned num, + struct pipe_texture **texture) +{ + struct brw_context *brw = brw_context(pipe); + uint i; + + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == brw->num_textures && + !memcmp(brw->attribs.Texture, texture, num * + sizeof(struct pipe_texture *))) + return; + + for (i = 0; i < num; i++) + pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], + texture[i]); + + for (i = num; i < brw->num_textures; i++) + pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], + NULL); + + brw->num_textures = num; + + brw->state.dirty.brw |= BRW_NEW_TEXTURE; +} + + +/************************************************************************ + * Render targets, etc + */ + +static void brw_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FrameBuffer = *fb; /* struct copy */ + + brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER; +} + + + +/************************************************************************ + * Rasterizer state + */ + +static void * +brw_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *rasterizer) +{ + DUP(pipe_rasterizer_state, rasterizer); +} + +static void brw_bind_rasterizer_state( struct pipe_context *pipe, + void *setup ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Raster = (struct pipe_rasterizer_state *)setup; + + /* Also pass-through to draw module: + */ + //draw_set_rasterizer_state(brw->draw, setup); + + brw->state.dirty.brw |= BRW_NEW_RASTERIZER; +} + +static void brw_delete_rasterizer_state(struct pipe_context *pipe, + void *setup) +{ + free(setup); +} + + + +void +brw_init_state_functions( struct brw_context *brw ) +{ + brw->pipe.create_blend_state = brw_create_blend_state; + brw->pipe.bind_blend_state = brw_bind_blend_state; + brw->pipe.delete_blend_state = brw_delete_blend_state; + + brw->pipe.create_sampler_state = brw_create_sampler_state; + brw->pipe.bind_sampler_states = brw_bind_sampler_states; + brw->pipe.delete_sampler_state = brw_delete_sampler_state; + + brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; + brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; + brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; + + brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; + brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; + brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state; + brw->pipe.create_fs_state = brw_create_fs_state; + brw->pipe.bind_fs_state = brw_bind_fs_state; + brw->pipe.delete_fs_state = brw_delete_fs_state; + brw->pipe.create_vs_state = brw_create_vs_state; + brw->pipe.bind_vs_state = brw_bind_vs_state; + brw->pipe.delete_vs_state = brw_delete_vs_state; + + brw->pipe.set_blend_color = brw_set_blend_color; + brw->pipe.set_clip_state = brw_set_clip_state; + brw->pipe.set_constant_buffer = brw_set_constant_buffer; + brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; + +// brw->pipe.set_feedback_state = brw_set_feedback_state; +// brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; + + brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; + brw->pipe.set_scissor_state = brw_set_scissor_state; + brw->pipe.set_sampler_textures = brw_set_sampler_textures; + brw->pipe.set_viewport_state = brw_set_viewport_state; + brw->pipe.set_vertex_buffers = brw_set_vertex_buffers; + brw->pipe.set_vertex_elements = brw_set_vertex_elements; +} -- cgit v1.2.3 From c428997a521475c46ccd1df4e8ed8ccc6c4f8d61 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 26 May 2008 23:29:38 +0900 Subject: Revert DOS line endings. --- src/gallium/drivers/i915simple/i915_context.h | 682 ++++++------ src/gallium/drivers/i965simple/brw_context.h | 1368 ++++++++++++------------- src/gallium/drivers/i965simple/brw_state.c | 938 ++++++++--------- 3 files changed, 1494 insertions(+), 1494 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 38e6a34884..53fc5ed079 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -1,341 +1,341 @@ - /************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef I915_CONTEXT_H -#define I915_CONTEXT_H - - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" - -#include "draw/draw_vertex.h" - -#include "tgsi/util/tgsi_scan.h" - - -#define I915_TEX_UNITS 8 - -#define I915_DYNAMIC_MODES4 0 -#define I915_DYNAMIC_DEPTHSCALE_0 1 /* just the header */ -#define I915_DYNAMIC_DEPTHSCALE_1 2 -#define I915_DYNAMIC_IAB 3 -#define I915_DYNAMIC_BC_0 4 /* just the header */ -#define I915_DYNAMIC_BC_1 5 -#define I915_DYNAMIC_BFO_0 6 -#define I915_DYNAMIC_BFO_1 7 -#define I915_DYNAMIC_STP_0 8 -#define I915_DYNAMIC_STP_1 9 -#define I915_DYNAMIC_SC_ENA_0 10 -#define I915_DYNAMIC_SC_RECT_0 11 -#define I915_DYNAMIC_SC_RECT_1 12 -#define I915_DYNAMIC_SC_RECT_2 13 -#define I915_MAX_DYNAMIC 14 - - -#define I915_IMMEDIATE_S0 0 -#define I915_IMMEDIATE_S1 1 -#define I915_IMMEDIATE_S2 2 -#define I915_IMMEDIATE_S3 3 -#define I915_IMMEDIATE_S4 4 -#define I915_IMMEDIATE_S5 5 -#define I915_IMMEDIATE_S6 6 -#define I915_IMMEDIATE_S7 7 -#define I915_MAX_IMMEDIATE 8 - -/* These must mach the order of LI0_STATE_* bits, as they will be used - * to generate hardware packets: - */ -#define I915_CACHE_STATIC 0 -#define I915_CACHE_DYNAMIC 1 /* handled specially */ -#define I915_CACHE_SAMPLER 2 -#define I915_CACHE_MAP 3 -#define I915_CACHE_PROGRAM 4 -#define I915_CACHE_CONSTANTS 5 -#define I915_MAX_CACHE 6 - -#define I915_MAX_CONSTANT 32 - - -/** See constant_flags[] below */ -#define I915_CONSTFLAG_USER 0x1f - - -/** - * Subclass of pipe_shader_state - */ -struct i915_fragment_shader -{ - struct pipe_shader_state state; - - struct tgsi_shader_info info; - - uint *program; - uint program_len; - - /** - * constants introduced during translation. - * These are placed at the end of the constant buffer and grow toward - * the beginning (eg: slot 31, 30 29, ...) - * User-provided constants start at 0. - * This allows both types of constants to co-exist (until there's too many) - * and doesn't require regenerating/changing the fragment program to - * shuffle constants around. - */ - uint num_constants; - float constants[I915_MAX_CONSTANT][4]; - - /** - * Status of each constant - * if I915_CONSTFLAG_PARAM, the value must be taken from the corresponding - * slot of the user's constant buffer. (set by pipe->set_constant_buffer()) - * Else, the bitmask indicates which components are occupied by immediates. - */ - ubyte constant_flags[I915_MAX_CONSTANT]; -}; - - -struct i915_cache_context; - -/* Use to calculate differences between state emitted to hardware and - * current driver-calculated state. - */ -struct i915_state -{ - unsigned immediate[I915_MAX_IMMEDIATE]; - unsigned dynamic[I915_MAX_DYNAMIC]; - - float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; - /** number of constants passed in through a constant buffer */ - uint num_user_constants[PIPE_SHADER_TYPES]; - - /* texture sampler state */ - unsigned sampler[I915_TEX_UNITS][3]; - unsigned sampler_enable_flags; - unsigned sampler_enable_nr; - - /* texture image buffers */ - unsigned texbuffer[I915_TEX_UNITS][2]; - - /** Describes the current hardware vertex layout */ - struct vertex_info vertex_info; - - unsigned id; /* track lost context events */ -}; - -struct i915_blend_state { - unsigned iab; - unsigned modes4; - unsigned LIS5; - unsigned LIS6; -}; - -struct i915_depth_stencil_state { - unsigned stencil_modes4; - unsigned bfo[2]; - unsigned stencil_LIS5; - unsigned depth_LIS6; -}; - -struct i915_rasterizer_state { - int light_twoside : 1; - unsigned st; - enum interp_mode color_interp; - - unsigned LIS4; - unsigned LIS7; - unsigned sc[1]; - - const struct pipe_rasterizer_state *templ; - - union { float f; unsigned u; } ds[2]; -}; - -struct i915_sampler_state { - unsigned state[3]; - const struct pipe_sampler_state *templ; -}; - - -struct i915_texture { - struct pipe_texture base; - - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ - unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; -}; - -struct i915_context -{ - struct pipe_context pipe; - struct i915_winsys *winsys; - struct draw_context *draw; - - /* The most recent drawing state as set by the driver: - */ - const struct i915_blend_state *blend; - const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; - const struct i915_depth_stencil_state *depth_stencil; - const struct i915_rasterizer_state *rasterizer; - - struct i915_fragment_shader *fs; - - struct pipe_blend_color blend_color; - struct pipe_clip_state clip; - struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; - struct pipe_framebuffer_state framebuffer; - struct pipe_poly_stipple poly_stipple; - struct pipe_scissor_state scissor; - struct i915_texture *texture[PIPE_MAX_SAMPLERS]; - struct pipe_viewport_state viewport; - struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; - - unsigned dirty; - - unsigned num_samplers; - unsigned num_textures; - unsigned num_vertex_elements; - unsigned num_vertex_buffers; - - unsigned *batch_start; - - /** Vertex buffer */ - struct pipe_buffer *vbo; - - struct i915_state current; - unsigned hardware_dirty; - - unsigned debug; -}; - -/* A flag for each state_tracker state object: - */ -#define I915_NEW_VIEWPORT 0x1 -#define I915_NEW_RASTERIZER 0x2 -#define I915_NEW_FS 0x4 -#define I915_NEW_BLEND 0x8 -#define I915_NEW_CLIP 0x10 -#define I915_NEW_SCISSOR 0x20 -#define I915_NEW_STIPPLE 0x40 -#define I915_NEW_FRAMEBUFFER 0x80 -#define I915_NEW_ALPHA_TEST 0x100 -#define I915_NEW_DEPTH_STENCIL 0x200 -#define I915_NEW_SAMPLER 0x400 -#define I915_NEW_TEXTURE 0x800 -#define I915_NEW_CONSTANTS 0x1000 -#define I915_NEW_VBO 0x2000 -#define I915_NEW_VS 0x4000 - - -/* Driver's internally generated state flags: - */ -#define I915_NEW_VERTEX_FORMAT 0x10000 - - -/* Dirty flags for hardware emit - */ -#define I915_HW_STATIC (1<set_constant_buffer()) + * Else, the bitmask indicates which components are occupied by immediates. + */ + ubyte constant_flags[I915_MAX_CONSTANT]; +}; + + +struct i915_cache_context; + +/* Use to calculate differences between state emitted to hardware and + * current driver-calculated state. + */ +struct i915_state +{ + unsigned immediate[I915_MAX_IMMEDIATE]; + unsigned dynamic[I915_MAX_DYNAMIC]; + + float constants[PIPE_SHADER_TYPES][I915_MAX_CONSTANT][4]; + /** number of constants passed in through a constant buffer */ + uint num_user_constants[PIPE_SHADER_TYPES]; + + /* texture sampler state */ + unsigned sampler[I915_TEX_UNITS][3]; + unsigned sampler_enable_flags; + unsigned sampler_enable_nr; + + /* texture image buffers */ + unsigned texbuffer[I915_TEX_UNITS][2]; + + /** Describes the current hardware vertex layout */ + struct vertex_info vertex_info; + + unsigned id; /* track lost context events */ +}; + +struct i915_blend_state { + unsigned iab; + unsigned modes4; + unsigned LIS5; + unsigned LIS6; +}; + +struct i915_depth_stencil_state { + unsigned stencil_modes4; + unsigned bfo[2]; + unsigned stencil_LIS5; + unsigned depth_LIS6; +}; + +struct i915_rasterizer_state { + int light_twoside : 1; + unsigned st; + enum interp_mode color_interp; + + unsigned LIS4; + unsigned LIS7; + unsigned sc[1]; + + const struct pipe_rasterizer_state *templ; + + union { float f; unsigned u; } ds[2]; +}; + +struct i915_sampler_state { + unsigned state[3]; + const struct pipe_sampler_state *templ; +}; + + +struct i915_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; +}; + +struct i915_context +{ + struct pipe_context pipe; + struct i915_winsys *winsys; + struct draw_context *draw; + + /* The most recent drawing state as set by the driver: + */ + const struct i915_blend_state *blend; + const struct i915_sampler_state *sampler[PIPE_MAX_SAMPLERS]; + const struct i915_depth_stencil_state *depth_stencil; + const struct i915_rasterizer_state *rasterizer; + + struct i915_fragment_shader *fs; + + struct pipe_blend_color blend_color; + struct pipe_clip_state clip; + struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; + struct pipe_framebuffer_state framebuffer; + struct pipe_poly_stipple poly_stipple; + struct pipe_scissor_state scissor; + struct i915_texture *texture[PIPE_MAX_SAMPLERS]; + struct pipe_viewport_state viewport; + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + + unsigned dirty; + + unsigned num_samplers; + unsigned num_textures; + unsigned num_vertex_elements; + unsigned num_vertex_buffers; + + unsigned *batch_start; + + /** Vertex buffer */ + struct pipe_buffer *vbo; + + struct i915_state current; + unsigned hardware_dirty; + + unsigned debug; +}; + +/* A flag for each state_tracker state object: + */ +#define I915_NEW_VIEWPORT 0x1 +#define I915_NEW_RASTERIZER 0x2 +#define I915_NEW_FS 0x4 +#define I915_NEW_BLEND 0x8 +#define I915_NEW_CLIP 0x10 +#define I915_NEW_SCISSOR 0x20 +#define I915_NEW_STIPPLE 0x40 +#define I915_NEW_FRAMEBUFFER 0x80 +#define I915_NEW_ALPHA_TEST 0x100 +#define I915_NEW_DEPTH_STENCIL 0x200 +#define I915_NEW_SAMPLER 0x400 +#define I915_NEW_TEXTURE 0x800 +#define I915_NEW_CONSTANTS 0x1000 +#define I915_NEW_VBO 0x2000 +#define I915_NEW_VS 0x4000 + + +/* Driver's internally generated state flags: + */ +#define I915_NEW_VERTEX_FORMAT 0x10000 + + +/* Dirty flags for hardware emit + */ +#define I915_HW_STATIC (1< - */ - - -#ifndef BRWCONTEXT_INC -#define BRWCONTEXT_INC - - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" - -#include "tgsi/util/tgsi_scan.h" - -#include "brw_structs.h" -#include "brw_winsys.h" - - -/* Glossary: - * - * URB - uniform resource buffer. A mid-sized buffer which is - * partitioned between the fixed function units and used for passing - * values (vertices, primitives, constants) between them. - * - * CURBE - constant URB entry. An urb region (entry) used to hold - * constant values which the fixed function units can be instructed to - * preload into the GRF when spawining a thread. - * - * VUE - vertex URB entry. An urb entry holding a vertex and usually - * a vertex header. The header contains control information and - * things like primitive type, Begin/end flags and clip codes. - * - * PUE - primitive URB entry. An urb entry produced by the setup (SF) - * unit holding rasterization and interpolation parameters. - * - * GRF - general register file. One of several register files - * addressable by programmed threads. The inputs (r0, payload, curbe, - * urb) of the thread are preloaded to this area before the thread is - * spawned. The registers are individually 8 dwords wide and suitable - * for general usage. Registers holding thread input values are not - * special and may be overwritten. - * - * MRF - message register file. Threads communicate (and terminate) - * by sending messages. Message parameters are placed in contigous - * MRF registers. All program output is via these messages. URB - * entries are populated by sending a message to the shared URB - * function containing the new data, together with a control word, - * often an unmodified copy of R0. - * - * R0 - GRF register 0. Typically holds control information used when - * sending messages to other threads. - * - * EU or GEN4 EU: The name of the programmable subsystem of the - * i965 hardware. Threads are executed by the EU, the registers - * described above are part of the EU architecture. - * - * Fixed function units: - * - * CS - Command streamer. Notional first unit, little software - * interaction. Holds the URB entries used for constant data, ie the - * CURBEs. - * - * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of - * this unit is responsible for pulling vertices out of vertex buffers - * in vram and injecting them into the processing pipe as VUEs. If - * enabled, it first passes them to a VS thread which is a good place - * for the driver to implement any active vertex shader. - * - * GS - Geometry Shader. This corresponds to a new DX10 concept. If - * enabled, incoming strips etc are passed to GS threads in individual - * line/triangle/point units. The GS thread may perform arbitary - * computation and emit whatever primtives with whatever vertices it - * chooses. This makes GS an excellent place to implement GL's - * unfilled polygon modes, though of course it is capable of much - * more. Additionally, GS is used to translate away primitives not - * handled by latter units, including Quads and Lineloops. - * - * CS - Clipper. Mesa's clipping algorithms are imported to run on - * this unit. The fixed function part performs cliptesting against - * the 6 fixed clipplanes and makes descisions on whether or not the - * incoming primitive needs to be passed to a thread for clipping. - * User clip planes are handled via cooperation with the VS thread. - * - * SF - Strips Fans or Setup: Triangles are prepared for - * rasterization. Interpolation coefficients are calculated. - * Flatshading and two-side lighting usually performed here. - * - * WM - Windower. Interpolation of vertex attributes performed here. - * Fragment shader implemented here. SIMD aspects of EU taken full - * advantage of, as pixels are processed in blocks of 16. - * - * CC - Color Calculator. No EU threads associated with this unit. - * Handles blending and (presumably) depth and stencil testing. - */ - -#define BRW_MAX_CURBE (32*16) - -struct brw_context; -struct brw_winsys; - - -/* Raised when we receive new state across the pipe interface: - */ -#define BRW_NEW_VIEWPORT 0x1 -#define BRW_NEW_RASTERIZER 0x2 -#define BRW_NEW_FS 0x4 -#define BRW_NEW_BLEND 0x8 -#define BRW_NEW_CLIP 0x10 -#define BRW_NEW_SCISSOR 0x20 -#define BRW_NEW_STIPPLE 0x40 -#define BRW_NEW_FRAMEBUFFER 0x80 -#define BRW_NEW_ALPHA_TEST 0x100 -#define BRW_NEW_DEPTH_STENCIL 0x200 -#define BRW_NEW_SAMPLER 0x400 -#define BRW_NEW_TEXTURE 0x800 -#define BRW_NEW_CONSTANTS 0x1000 -#define BRW_NEW_VBO 0x2000 -#define BRW_NEW_VS 0x4000 - -/* Raised for other internal events: - */ -#define BRW_NEW_URB_FENCE 0x10000 -#define BRW_NEW_PSP 0x20000 -#define BRW_NEW_CURBE_OFFSETS 0x40000 -#define BRW_NEW_REDUCED_PRIMITIVE 0x80000 -#define BRW_NEW_PRIMITIVE 0x100000 -#define BRW_NEW_SCENE 0x200000 -#define BRW_NEW_SF_LINKAGE 0x400000 - -extern int BRW_DEBUG; - -#define DEBUG_TEXTURE 0x1 -#define DEBUG_STATE 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_PRIMS 0x8 -#define DEBUG_VERTS 0x10 -#define DEBUG_FALLBACKS 0x20 -#define DEBUG_VERBOSE 0x40 -#define DEBUG_DRI 0x80 -#define DEBUG_DMA 0x100 -#define DEBUG_SANITY 0x200 -#define DEBUG_SYNC 0x400 -#define DEBUG_SLEEP 0x800 -#define DEBUG_PIXEL 0x1000 -#define DEBUG_STATS 0x2000 -#define DEBUG_TILE 0x4000 -#define DEBUG_SINGLE_THREAD 0x8000 -#define DEBUG_WM 0x10000 -#define DEBUG_URB 0x20000 -#define DEBUG_VS 0x40000 -#define DEBUG_BATCH 0x80000 -#define DEBUG_BUFMGR 0x100000 -#define DEBUG_BLIT 0x200000 -#define DEBUG_REGION 0x400000 -#define DEBUG_MIPTREE 0x800000 - -#define DBG(...) do { \ - if (BRW_DEBUG & FILE_DEBUG_FLAG) \ - debug_printf(__VA_ARGS__); \ -} while(0) - -#define PRINT(...) do { \ - debug_printf(brw->pipe.winsys, __VA_ARGS__); \ -} while(0) - -struct brw_state_flags { - unsigned cache; - unsigned brw; -}; - - -struct brw_vertex_program { - struct pipe_shader_state program; - struct tgsi_shader_info info; - int id; -}; - - -struct brw_fragment_program { - struct pipe_shader_state program; - struct tgsi_shader_info info; - - boolean UsesDepth; /* XXX add this to tgsi_shader_info? */ - int id; -}; - - -struct pipe_setup_linkage { - struct { - unsigned vp_output:5; - unsigned interp_mode:4; - unsigned bf_vp_output:5; - } fp_input[PIPE_MAX_SHADER_INPUTS]; - - unsigned fp_input_count:5; - unsigned max_vp_output:5; -}; - - - -struct brw_texture { - struct pipe_texture base; - - /* Derived from the above: - */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; - - unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; - - /* Explicitly store the offset of each image for each cube face or - * depth value. Pretty much have to accept that hardware formats - * are going to be so diverse that there is no unified way to - * compute the offsets of depth/cube images within a mipmap level, - * so have to store them as a lookup table: - */ - unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - - /* Includes image offset tables: - */ - unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; - - /* The data is held here: - */ - struct pipe_buffer *buffer; -}; - -/* Data about a particular attempt to compile a program. Note that - * there can be many of these, each in a different GL state - * corresponding to a different brw_wm_prog_key struct, with different - * compiled programs: - */ -/* Data about a particular attempt to compile a program. Note that - * there can be many of these, each in a different GL state - * corresponding to a different brw_wm_prog_key struct, with different - * compiled programs: - */ - -struct brw_wm_prog_data { - unsigned curb_read_length; - unsigned urb_read_length; - - unsigned first_curbe_grf; - unsigned total_grf; - unsigned total_scratch; - - /* Internally generated constants for the CURBE. These are loaded - * ahead of the data from the constant buffer. - */ - const float internal_const[8]; - unsigned nr_internal_consts; - unsigned max_const; - - boolean error; -}; - -struct brw_sf_prog_data { - unsigned urb_read_length; - unsigned total_grf; - - /* Each vertex may have upto 12 attributes, 4 components each, - * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 - * rows. - * - * Actually we use 4 for each, so call it 12 rows. - */ - unsigned urb_entry_size; -}; - -struct brw_clip_prog_data { - unsigned curb_read_length; /* user planes? */ - unsigned clip_mode; - unsigned urb_read_length; - unsigned total_grf; -}; - -struct brw_gs_prog_data { - unsigned urb_read_length; - unsigned total_grf; -}; - -struct brw_vs_prog_data { - unsigned curb_read_length; - unsigned urb_read_length; - unsigned total_grf; - unsigned outputs_written; - - unsigned inputs_read; - - unsigned max_const; - - float imm_buf[PIPE_MAX_CONSTANT][4]; - unsigned num_imm; - unsigned num_consts; - - /* Used for calculating urb partitions: - */ - unsigned urb_entry_size; -}; - - -#define BRW_MAX_TEX_UNIT 8 -#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 - -/* Create a fixed sized struct for caching binding tables: - */ -struct brw_surface_binding_table { - unsigned surf_ss_offset[BRW_WM_MAX_SURF]; -}; - - -struct brw_cache; - -struct brw_mem_pool { - struct pipe_buffer *buffer; - - unsigned size; - unsigned offset; /* offset of first free byte */ - - struct brw_context *brw; -}; - -struct brw_cache_item { - unsigned hash; - unsigned key_size; /* for variable-sized keys */ - const void *key; - - unsigned offset; /* offset within pool's buffer */ - unsigned data_size; - - struct brw_cache_item *next; -}; - - - -struct brw_cache { - unsigned id; - - const char *name; - - struct brw_context *brw; - struct brw_mem_pool *pool; - - struct brw_cache_item **items; - unsigned size, n_items; - - unsigned key_size; /* for fixed-size keys */ - unsigned aux_size; - - unsigned last_addr; /* offset of active item */ -}; - - - - -/* Considered adding a member to this struct to document which flags - * an update might raise so that ordering of the state atoms can be - * checked or derived at runtime. Dropped the idea in favor of having - * a debug mode where the state is monitored for flags which are - * raised that have already been tested against. - */ -struct brw_tracked_state { - struct brw_state_flags dirty; - void (*update)( struct brw_context *brw ); -}; - - -/* Flags for brw->state.cache. - */ -#define CACHE_NEW_CC_VP (1< 32. Wouldn't life - * be easier if C allowed arrays of packed elements? - */ -#define ATTRIB_BIT_DWORDS ((PIPE_MAX_ATTRIBS+31)/32) - - - - -struct brw_vertex_info { - unsigned varying; /* varying:1[PIPE_MAX_ATTRIBS] */ - unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_MAX_ATTRIBS] */ -}; - - - - - -struct brw_context -{ - struct pipe_context pipe; - struct brw_winsys *winsys; - - unsigned primitive; - unsigned reduced_primitive; - - boolean emit_state_always; - - struct { - struct brw_state_flags dirty; - } state; - - - struct { - const struct pipe_blend_state *Blend; - const struct pipe_depth_stencil_alpha_state *DepthStencil; - const struct pipe_poly_stipple *PolygonStipple; - const struct pipe_rasterizer_state *Raster; - const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; - const struct brw_vertex_program *VertexProgram; - const struct brw_fragment_program *FragmentProgram; - - struct pipe_clip_state Clip; - struct pipe_blend_color BlendColor; - struct pipe_scissor_state Scissor; - struct pipe_viewport_state Viewport; - struct pipe_framebuffer_state FrameBuffer; - - const struct pipe_constant_buffer *Constants[2]; - const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; - } attribs; - - unsigned num_samplers; - unsigned num_textures; - - struct brw_mem_pool pool[BRW_MAX_POOL]; - struct brw_cache cache[BRW_MAX_CACHE]; - struct brw_cached_batch_item *cached_batch_items; - - struct { - - /* Arrays with buffer objects to copy non-bufferobj arrays into - * for upload: - */ - const struct pipe_vertex_buffer *vbo_array[PIPE_MAX_ATTRIBS]; - - struct brw_vertex_element_state inputs[PIPE_MAX_ATTRIBS]; - -#define BRW_NR_UPLOAD_BUFS 17 -#define BRW_UPLOAD_INIT_SIZE (128*1024) - - /* Summary of size and varying of active arrays, so we can check - * for changes to this state: - */ - struct brw_vertex_info info; - } vb; - - - unsigned hardware_dirty; - unsigned dirty; - unsigned pci_id; - /* BRW_NEW_URB_ALLOCATIONS: - */ - struct { - unsigned vsize; /* vertex size plus header in urb registers */ - unsigned csize; /* constant buffer size in urb registers */ - unsigned sfsize; /* setup data size in urb registers */ - - boolean constrained; - - unsigned nr_vs_entries; - unsigned nr_gs_entries; - unsigned nr_clip_entries; - unsigned nr_sf_entries; - unsigned nr_cs_entries; - -/* unsigned vs_size; */ -/* unsigned gs_size; */ -/* unsigned clip_size; */ -/* unsigned sf_size; */ -/* unsigned cs_size; */ - - unsigned vs_start; - unsigned gs_start; - unsigned clip_start; - unsigned sf_start; - unsigned cs_start; - } urb; - - - /* BRW_NEW_CURBE_OFFSETS: - */ - struct { - unsigned wm_start; - unsigned wm_size; - unsigned clip_start; - unsigned clip_size; - unsigned vs_start; - unsigned vs_size; - unsigned total_size; - - unsigned gs_offset; - - float *last_buf; - unsigned last_bufsz; - } curbe; - - struct { - struct brw_vs_prog_data *prog_data; - - unsigned prog_gs_offset; - unsigned state_gs_offset; - } vs; - - struct { - struct brw_gs_prog_data *prog_data; - - boolean prog_active; - unsigned prog_gs_offset; - unsigned state_gs_offset; - } gs; - - struct { - struct brw_clip_prog_data *prog_data; - - unsigned prog_gs_offset; - unsigned vp_gs_offset; - unsigned state_gs_offset; - } clip; - - - struct { - struct brw_sf_prog_data *prog_data; - - struct pipe_setup_linkage linkage; - - unsigned prog_gs_offset; - unsigned vp_gs_offset; - unsigned state_gs_offset; - } sf; - - struct { - struct brw_wm_prog_data *prog_data; - -// struct brw_wm_compiler *compile_data; - - - /** - * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER - * cache - */ - struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; - - unsigned render_surf; - unsigned nr_surfaces; - - unsigned max_threads; - struct pipe_buffer *scratch_buffer; - unsigned scratch_buffer_size; - - unsigned sampler_count; - unsigned sampler_gs_offset; - - struct brw_surface_binding_table bind; - unsigned bind_ss_offset; - - unsigned prog_gs_offset; - unsigned state_gs_offset; - } wm; - - - struct { - unsigned vp_gs_offset; - unsigned state_gs_offset; - } cc; - - - /* Used to give every program string a unique id - */ - unsigned program_id; -}; - - -#define BRW_PACKCOLOR8888(r,g,b,a) ((r<<24) | (g<<16) | (b<<8) | a) - - -/*====================================================================== - * brw_vtbl.c - */ -void brw_do_flush( struct brw_context *brw, - unsigned flags ); - - -/*====================================================================== - * brw_state.c - */ -void brw_validate_state(struct brw_context *brw); -void brw_init_state(struct brw_context *brw); -void brw_destroy_state(struct brw_context *brw); - - -/*====================================================================== - * brw_tex.c - */ -void brwUpdateTextureState( struct brw_context *brw ); - - -/* brw_urb.c - */ -void brw_upload_urb_fence(struct brw_context *brw); - -void brw_upload_constant_buffer_state(struct brw_context *brw); - -void brw_init_surface_functions(struct brw_context *brw); -void brw_init_state_functions(struct brw_context *brw); -void brw_init_flush_functions(struct brw_context *brw); -void brw_init_string_functions(struct brw_context *brw); - -/*====================================================================== - * Inline conversion functions. These are better-typed than the - * macros used previously: - */ -static inline struct brw_context * -brw_context( struct pipe_context *ctx ) -{ - return (struct brw_context *)ctx; -} - -#endif - +/* + Copyright (C) Intel Corp. 2006. All Rights Reserved. + Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to + develop this 3D driver. + + Permission is hereby granted, free of charge, to any person obtaining + a copy of this software and associated documentation files (the + "Software"), to deal in the Software without restriction, including + without limitation the rights to use, copy, modify, merge, publish, + distribute, sublicense, and/or sell copies of the Software, and to + permit persons to whom the Software is furnished to do so, subject to + the following conditions: + + The above copyright notice and this permission notice (including the + next paragraph) shall be included in all copies or substantial + portions of the Software. + + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. + IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE + LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION + OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION + WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + + **********************************************************************/ + /* + * Authors: + * Keith Whitwell + */ + + +#ifndef BRWCONTEXT_INC +#define BRWCONTEXT_INC + + +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" + +#include "tgsi/util/tgsi_scan.h" + +#include "brw_structs.h" +#include "brw_winsys.h" + + +/* Glossary: + * + * URB - uniform resource buffer. A mid-sized buffer which is + * partitioned between the fixed function units and used for passing + * values (vertices, primitives, constants) between them. + * + * CURBE - constant URB entry. An urb region (entry) used to hold + * constant values which the fixed function units can be instructed to + * preload into the GRF when spawining a thread. + * + * VUE - vertex URB entry. An urb entry holding a vertex and usually + * a vertex header. The header contains control information and + * things like primitive type, Begin/end flags and clip codes. + * + * PUE - primitive URB entry. An urb entry produced by the setup (SF) + * unit holding rasterization and interpolation parameters. + * + * GRF - general register file. One of several register files + * addressable by programmed threads. The inputs (r0, payload, curbe, + * urb) of the thread are preloaded to this area before the thread is + * spawned. The registers are individually 8 dwords wide and suitable + * for general usage. Registers holding thread input values are not + * special and may be overwritten. + * + * MRF - message register file. Threads communicate (and terminate) + * by sending messages. Message parameters are placed in contigous + * MRF registers. All program output is via these messages. URB + * entries are populated by sending a message to the shared URB + * function containing the new data, together with a control word, + * often an unmodified copy of R0. + * + * R0 - GRF register 0. Typically holds control information used when + * sending messages to other threads. + * + * EU or GEN4 EU: The name of the programmable subsystem of the + * i965 hardware. Threads are executed by the EU, the registers + * described above are part of the EU architecture. + * + * Fixed function units: + * + * CS - Command streamer. Notional first unit, little software + * interaction. Holds the URB entries used for constant data, ie the + * CURBEs. + * + * VF/VS - Vertex Fetch / Vertex Shader. The fixed function part of + * this unit is responsible for pulling vertices out of vertex buffers + * in vram and injecting them into the processing pipe as VUEs. If + * enabled, it first passes them to a VS thread which is a good place + * for the driver to implement any active vertex shader. + * + * GS - Geometry Shader. This corresponds to a new DX10 concept. If + * enabled, incoming strips etc are passed to GS threads in individual + * line/triangle/point units. The GS thread may perform arbitary + * computation and emit whatever primtives with whatever vertices it + * chooses. This makes GS an excellent place to implement GL's + * unfilled polygon modes, though of course it is capable of much + * more. Additionally, GS is used to translate away primitives not + * handled by latter units, including Quads and Lineloops. + * + * CS - Clipper. Mesa's clipping algorithms are imported to run on + * this unit. The fixed function part performs cliptesting against + * the 6 fixed clipplanes and makes descisions on whether or not the + * incoming primitive needs to be passed to a thread for clipping. + * User clip planes are handled via cooperation with the VS thread. + * + * SF - Strips Fans or Setup: Triangles are prepared for + * rasterization. Interpolation coefficients are calculated. + * Flatshading and two-side lighting usually performed here. + * + * WM - Windower. Interpolation of vertex attributes performed here. + * Fragment shader implemented here. SIMD aspects of EU taken full + * advantage of, as pixels are processed in blocks of 16. + * + * CC - Color Calculator. No EU threads associated with this unit. + * Handles blending and (presumably) depth and stencil testing. + */ + +#define BRW_MAX_CURBE (32*16) + +struct brw_context; +struct brw_winsys; + + +/* Raised when we receive new state across the pipe interface: + */ +#define BRW_NEW_VIEWPORT 0x1 +#define BRW_NEW_RASTERIZER 0x2 +#define BRW_NEW_FS 0x4 +#define BRW_NEW_BLEND 0x8 +#define BRW_NEW_CLIP 0x10 +#define BRW_NEW_SCISSOR 0x20 +#define BRW_NEW_STIPPLE 0x40 +#define BRW_NEW_FRAMEBUFFER 0x80 +#define BRW_NEW_ALPHA_TEST 0x100 +#define BRW_NEW_DEPTH_STENCIL 0x200 +#define BRW_NEW_SAMPLER 0x400 +#define BRW_NEW_TEXTURE 0x800 +#define BRW_NEW_CONSTANTS 0x1000 +#define BRW_NEW_VBO 0x2000 +#define BRW_NEW_VS 0x4000 + +/* Raised for other internal events: + */ +#define BRW_NEW_URB_FENCE 0x10000 +#define BRW_NEW_PSP 0x20000 +#define BRW_NEW_CURBE_OFFSETS 0x40000 +#define BRW_NEW_REDUCED_PRIMITIVE 0x80000 +#define BRW_NEW_PRIMITIVE 0x100000 +#define BRW_NEW_SCENE 0x200000 +#define BRW_NEW_SF_LINKAGE 0x400000 + +extern int BRW_DEBUG; + +#define DEBUG_TEXTURE 0x1 +#define DEBUG_STATE 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_PRIMS 0x8 +#define DEBUG_VERTS 0x10 +#define DEBUG_FALLBACKS 0x20 +#define DEBUG_VERBOSE 0x40 +#define DEBUG_DRI 0x80 +#define DEBUG_DMA 0x100 +#define DEBUG_SANITY 0x200 +#define DEBUG_SYNC 0x400 +#define DEBUG_SLEEP 0x800 +#define DEBUG_PIXEL 0x1000 +#define DEBUG_STATS 0x2000 +#define DEBUG_TILE 0x4000 +#define DEBUG_SINGLE_THREAD 0x8000 +#define DEBUG_WM 0x10000 +#define DEBUG_URB 0x20000 +#define DEBUG_VS 0x40000 +#define DEBUG_BATCH 0x80000 +#define DEBUG_BUFMGR 0x100000 +#define DEBUG_BLIT 0x200000 +#define DEBUG_REGION 0x400000 +#define DEBUG_MIPTREE 0x800000 + +#define DBG(...) do { \ + if (BRW_DEBUG & FILE_DEBUG_FLAG) \ + debug_printf(__VA_ARGS__); \ +} while(0) + +#define PRINT(...) do { \ + debug_printf(brw->pipe.winsys, __VA_ARGS__); \ +} while(0) + +struct brw_state_flags { + unsigned cache; + unsigned brw; +}; + + +struct brw_vertex_program { + struct pipe_shader_state program; + struct tgsi_shader_info info; + int id; +}; + + +struct brw_fragment_program { + struct pipe_shader_state program; + struct tgsi_shader_info info; + + boolean UsesDepth; /* XXX add this to tgsi_shader_info? */ + int id; +}; + + +struct pipe_setup_linkage { + struct { + unsigned vp_output:5; + unsigned interp_mode:4; + unsigned bf_vp_output:5; + } fp_input[PIPE_MAX_SHADER_INPUTS]; + + unsigned fp_input_count:5; + unsigned max_vp_output:5; +}; + + + +struct brw_texture { + struct pipe_texture base; + + /* Derived from the above: + */ + unsigned pitch; + unsigned depth_pitch; /* per-image on i945? */ + unsigned total_height; + + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; + + /* Explicitly store the offset of each image for each cube face or + * depth value. Pretty much have to accept that hardware formats + * are going to be so diverse that there is no unified way to + * compute the offsets of depth/cube images within a mipmap level, + * so have to store them as a lookup table: + */ + unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ + + /* Includes image offset tables: + */ + unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; + + /* The data is held here: + */ + struct pipe_buffer *buffer; +}; + +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ +/* Data about a particular attempt to compile a program. Note that + * there can be many of these, each in a different GL state + * corresponding to a different brw_wm_prog_key struct, with different + * compiled programs: + */ + +struct brw_wm_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + + unsigned first_curbe_grf; + unsigned total_grf; + unsigned total_scratch; + + /* Internally generated constants for the CURBE. These are loaded + * ahead of the data from the constant buffer. + */ + const float internal_const[8]; + unsigned nr_internal_consts; + unsigned max_const; + + boolean error; +}; + +struct brw_sf_prog_data { + unsigned urb_read_length; + unsigned total_grf; + + /* Each vertex may have upto 12 attributes, 4 components each, + * except WPOS which requires only 2. (11*4 + 2) == 44 ==> 11 + * rows. + * + * Actually we use 4 for each, so call it 12 rows. + */ + unsigned urb_entry_size; +}; + +struct brw_clip_prog_data { + unsigned curb_read_length; /* user planes? */ + unsigned clip_mode; + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_gs_prog_data { + unsigned urb_read_length; + unsigned total_grf; +}; + +struct brw_vs_prog_data { + unsigned curb_read_length; + unsigned urb_read_length; + unsigned total_grf; + unsigned outputs_written; + + unsigned inputs_read; + + unsigned max_const; + + float imm_buf[PIPE_MAX_CONSTANT][4]; + unsigned num_imm; + unsigned num_consts; + + /* Used for calculating urb partitions: + */ + unsigned urb_entry_size; +}; + + +#define BRW_MAX_TEX_UNIT 8 +#define BRW_WM_MAX_SURF BRW_MAX_TEX_UNIT + 1 + +/* Create a fixed sized struct for caching binding tables: + */ +struct brw_surface_binding_table { + unsigned surf_ss_offset[BRW_WM_MAX_SURF]; +}; + + +struct brw_cache; + +struct brw_mem_pool { + struct pipe_buffer *buffer; + + unsigned size; + unsigned offset; /* offset of first free byte */ + + struct brw_context *brw; +}; + +struct brw_cache_item { + unsigned hash; + unsigned key_size; /* for variable-sized keys */ + const void *key; + + unsigned offset; /* offset within pool's buffer */ + unsigned data_size; + + struct brw_cache_item *next; +}; + + + +struct brw_cache { + unsigned id; + + const char *name; + + struct brw_context *brw; + struct brw_mem_pool *pool; + + struct brw_cache_item **items; + unsigned size, n_items; + + unsigned key_size; /* for fixed-size keys */ + unsigned aux_size; + + unsigned last_addr; /* offset of active item */ +}; + + + + +/* Considered adding a member to this struct to document which flags + * an update might raise so that ordering of the state atoms can be + * checked or derived at runtime. Dropped the idea in favor of having + * a debug mode where the state is monitored for flags which are + * raised that have already been tested against. + */ +struct brw_tracked_state { + struct brw_state_flags dirty; + void (*update)( struct brw_context *brw ); +}; + + +/* Flags for brw->state.cache. + */ +#define CACHE_NEW_CC_VP (1< 32. Wouldn't life + * be easier if C allowed arrays of packed elements? + */ +#define ATTRIB_BIT_DWORDS ((PIPE_MAX_ATTRIBS+31)/32) + + + + +struct brw_vertex_info { + unsigned varying; /* varying:1[PIPE_MAX_ATTRIBS] */ + unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_MAX_ATTRIBS] */ +}; + + + + + +struct brw_context +{ + struct pipe_context pipe; + struct brw_winsys *winsys; + + unsigned primitive; + unsigned reduced_primitive; + + boolean emit_state_always; + + struct { + struct brw_state_flags dirty; + } state; + + + struct { + const struct pipe_blend_state *Blend; + const struct pipe_depth_stencil_alpha_state *DepthStencil; + const struct pipe_poly_stipple *PolygonStipple; + const struct pipe_rasterizer_state *Raster; + const struct pipe_sampler_state *Samplers[PIPE_MAX_SAMPLERS]; + const struct brw_vertex_program *VertexProgram; + const struct brw_fragment_program *FragmentProgram; + + struct pipe_clip_state Clip; + struct pipe_blend_color BlendColor; + struct pipe_scissor_state Scissor; + struct pipe_viewport_state Viewport; + struct pipe_framebuffer_state FrameBuffer; + + const struct pipe_constant_buffer *Constants[2]; + const struct brw_texture *Texture[PIPE_MAX_SAMPLERS]; + } attribs; + + unsigned num_samplers; + unsigned num_textures; + + struct brw_mem_pool pool[BRW_MAX_POOL]; + struct brw_cache cache[BRW_MAX_CACHE]; + struct brw_cached_batch_item *cached_batch_items; + + struct { + + /* Arrays with buffer objects to copy non-bufferobj arrays into + * for upload: + */ + const struct pipe_vertex_buffer *vbo_array[PIPE_MAX_ATTRIBS]; + + struct brw_vertex_element_state inputs[PIPE_MAX_ATTRIBS]; + +#define BRW_NR_UPLOAD_BUFS 17 +#define BRW_UPLOAD_INIT_SIZE (128*1024) + + /* Summary of size and varying of active arrays, so we can check + * for changes to this state: + */ + struct brw_vertex_info info; + } vb; + + + unsigned hardware_dirty; + unsigned dirty; + unsigned pci_id; + /* BRW_NEW_URB_ALLOCATIONS: + */ + struct { + unsigned vsize; /* vertex size plus header in urb registers */ + unsigned csize; /* constant buffer size in urb registers */ + unsigned sfsize; /* setup data size in urb registers */ + + boolean constrained; + + unsigned nr_vs_entries; + unsigned nr_gs_entries; + unsigned nr_clip_entries; + unsigned nr_sf_entries; + unsigned nr_cs_entries; + +/* unsigned vs_size; */ +/* unsigned gs_size; */ +/* unsigned clip_size; */ +/* unsigned sf_size; */ +/* unsigned cs_size; */ + + unsigned vs_start; + unsigned gs_start; + unsigned clip_start; + unsigned sf_start; + unsigned cs_start; + } urb; + + + /* BRW_NEW_CURBE_OFFSETS: + */ + struct { + unsigned wm_start; + unsigned wm_size; + unsigned clip_start; + unsigned clip_size; + unsigned vs_start; + unsigned vs_size; + unsigned total_size; + + unsigned gs_offset; + + float *last_buf; + unsigned last_bufsz; + } curbe; + + struct { + struct brw_vs_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } vs; + + struct { + struct brw_gs_prog_data *prog_data; + + boolean prog_active; + unsigned prog_gs_offset; + unsigned state_gs_offset; + } gs; + + struct { + struct brw_clip_prog_data *prog_data; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } clip; + + + struct { + struct brw_sf_prog_data *prog_data; + + struct pipe_setup_linkage linkage; + + unsigned prog_gs_offset; + unsigned vp_gs_offset; + unsigned state_gs_offset; + } sf; + + struct { + struct brw_wm_prog_data *prog_data; + +// struct brw_wm_compiler *compile_data; + + + /** + * Array of sampler state uploaded at sampler_gs_offset of BRW_SAMPLER + * cache + */ + struct brw_sampler_state sampler[BRW_MAX_TEX_UNIT]; + + unsigned render_surf; + unsigned nr_surfaces; + + unsigned max_threads; + struct pipe_buffer *scratch_buffer; + unsigned scratch_buffer_size; + + unsigned sampler_count; + unsigned sampler_gs_offset; + + struct brw_surface_binding_table bind; + unsigned bind_ss_offset; + + unsigned prog_gs_offset; + unsigned state_gs_offset; + } wm; + + + struct { + unsigned vp_gs_offset; + unsigned state_gs_offset; + } cc; + + + /* Used to give every program string a unique id + */ + unsigned program_id; +}; + + +#define BRW_PACKCOLOR8888(r,g,b,a) ((r<<24) | (g<<16) | (b<<8) | a) + + +/*====================================================================== + * brw_vtbl.c + */ +void brw_do_flush( struct brw_context *brw, + unsigned flags ); + + +/*====================================================================== + * brw_state.c + */ +void brw_validate_state(struct brw_context *brw); +void brw_init_state(struct brw_context *brw); +void brw_destroy_state(struct brw_context *brw); + + +/*====================================================================== + * brw_tex.c + */ +void brwUpdateTextureState( struct brw_context *brw ); + + +/* brw_urb.c + */ +void brw_upload_urb_fence(struct brw_context *brw); + +void brw_upload_constant_buffer_state(struct brw_context *brw); + +void brw_init_surface_functions(struct brw_context *brw); +void brw_init_state_functions(struct brw_context *brw); +void brw_init_flush_functions(struct brw_context *brw); +void brw_init_string_functions(struct brw_context *brw); + +/*====================================================================== + * Inline conversion functions. These are better-typed than the + * macros used previously: + */ +static inline struct brw_context * +brw_context( struct pipe_context *ctx ) +{ + return (struct brw_context *)ctx; +} + +#endif + diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index ac243b7e4f..caeeba4630 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -1,469 +1,469 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/* Authors: Zack Rusin - * Keith Whitwell - */ - - -#include "pipe/p_winsys.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_dump.h" -#include "tgsi/util/tgsi_parse.h" - -#include "brw_context.h" -#include "brw_defines.h" -#include "brw_state.h" -#include "brw_draw.h" - - -#define DUP( TYPE, VAL ) \ -do { \ - struct TYPE *x = malloc(sizeof(*x)); \ - memcpy(x, VAL, sizeof(*x) ); \ - return x; \ -} while (0) - -/************************************************************************ - * Blend - */ -static void * -brw_create_blend_state(struct pipe_context *pipe, - const struct pipe_blend_state *blend) -{ - DUP( pipe_blend_state, blend ); -} - -static void brw_bind_blend_state(struct pipe_context *pipe, - void *blend) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Blend = (struct pipe_blend_state*)blend; - brw->state.dirty.brw |= BRW_NEW_BLEND; -} - - -static void brw_delete_blend_state(struct pipe_context *pipe, void *blend) -{ - free(blend); -} - -static void brw_set_blend_color( struct pipe_context *pipe, - const struct pipe_blend_color *blend_color ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.BlendColor = *blend_color; - - brw->state.dirty.brw |= BRW_NEW_BLEND; -} - -/************************************************************************ - * Sampler - */ - -static void * -brw_create_sampler_state(struct pipe_context *pipe, - const struct pipe_sampler_state *sampler) -{ - DUP( pipe_sampler_state, sampler ); -} - -static void brw_bind_sampler_states(struct pipe_context *pipe, - unsigned num, void **sampler) -{ - struct brw_context *brw = brw_context(pipe); - - assert(num <= PIPE_MAX_SAMPLERS); - - /* Check for no-op */ - if (num == brw->num_samplers && - !memcmp(brw->attribs.Samplers, sampler, num * sizeof(void *))) - return; - - memcpy(brw->attribs.Samplers, sampler, num * sizeof(void *)); - memset(&brw->attribs.Samplers[num], 0, (PIPE_MAX_SAMPLERS - num) * - sizeof(void *)); - - brw->num_samplers = num; - - brw->state.dirty.brw |= BRW_NEW_SAMPLER; -} - -static void brw_delete_sampler_state(struct pipe_context *pipe, - void *sampler) -{ - free(sampler); -} - - -/************************************************************************ - * Depth stencil - */ - -static void * -brw_create_depth_stencil_state(struct pipe_context *pipe, - const struct pipe_depth_stencil_alpha_state *depth_stencil) -{ - DUP( pipe_depth_stencil_alpha_state, depth_stencil ); -} - -static void brw_bind_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; - - brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; -} - -static void brw_delete_depth_stencil_state(struct pipe_context *pipe, - void *depth_stencil) -{ - free(depth_stencil); -} - -/************************************************************************ - * Scissor - */ -static void brw_set_scissor_state( struct pipe_context *pipe, - const struct pipe_scissor_state *scissor ) -{ - struct brw_context *brw = brw_context(pipe); - - memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) ); - brw->state.dirty.brw |= BRW_NEW_SCISSOR; -} - - -/************************************************************************ - * Stipple - */ - -static void brw_set_polygon_stipple( struct pipe_context *pipe, - const struct pipe_poly_stipple *stipple ) -{ -} - - -/************************************************************************ - * Fragment shader - */ - -static void * brw_create_fs_state(struct pipe_context *pipe, - const struct pipe_shader_state *shader) -{ - struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); - - brw_fp->program.tokens = tgsi_dup_tokens(shader->tokens); - brw_fp->id = brw_context(pipe)->program_id++; - - tgsi_scan_shader(shader->tokens, &brw_fp->info); - -#if 0 - brw_shader_info(shader->tokens, - &brw_fp->info2); -#endif - - tgsi_dump(shader->tokens, 0); - - - return (void *)brw_fp; -} - -static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader; - brw->state.dirty.brw |= BRW_NEW_FS; -} - -static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) -{ - struct brw_fragment_program *brw_fp = (struct brw_fragment_program *) shader; - - FREE((void *) brw_fp->program.tokens); - FREE(brw_fp); -} - - -/************************************************************************ - * Vertex shader and other TNL state - */ - -static void *brw_create_vs_state(struct pipe_context *pipe, - const struct pipe_shader_state *shader) -{ - struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); - - brw_vp->program.tokens = tgsi_dup_tokens(shader->tokens); - brw_vp->id = brw_context(pipe)->program_id++; - - tgsi_scan_shader(shader->tokens, &brw_vp->info); - -#if 0 - brw_shader_info(shader->tokens, - &brw_vp->info2); -#endif - tgsi_dump(shader->tokens, 0); - - return (void *)brw_vp; -} - -static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; - brw->state.dirty.brw |= BRW_NEW_VS; - - debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); -} - -static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) -{ - struct brw_vertex_program *brw_vp = (struct brw_vertex_program *) shader; - - FREE((void *) brw_vp->program.tokens); - FREE(brw_vp); -} - - -static void brw_set_clip_state( struct pipe_context *pipe, - const struct pipe_clip_state *clip ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Clip = *clip; -} - - -static void brw_set_viewport_state( struct pipe_context *pipe, - const struct pipe_viewport_state *viewport ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Viewport = *viewport; /* struct copy */ - brw->state.dirty.brw |= BRW_NEW_VIEWPORT; - - /* pass the viewport info to the draw module */ - //draw_set_viewport_state(brw->draw, viewport); -} - - -static void brw_set_vertex_buffers(struct pipe_context *pipe, - unsigned count, - const struct pipe_vertex_buffer *buffers) -{ - struct brw_context *brw = brw_context(pipe); - memcpy(brw->vb.vbo_array, buffers, count * sizeof(buffers[0])); -} - -static void brw_set_vertex_elements(struct pipe_context *pipe, - unsigned count, - const struct pipe_vertex_element *elements) -{ - /* flush ? */ - struct brw_context *brw = brw_context(pipe); - uint i; - - assert(count <= PIPE_MAX_ATTRIBS); - - for (i = 0; i < count; i++) { - struct brw_vertex_element_state el; - memset(&el, 0, sizeof(el)); - - el.ve0.src_offset = elements[i].src_offset; - el.ve0.src_format = brw_translate_surface_format(elements[i].src_format); - el.ve0.valid = 1; - el.ve0.vertex_buffer_index = elements[i].vertex_buffer_index; - - el.ve1.dst_offset = i * 4; - - el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; - el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; - - switch (elements[i].nr_components) { - case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; - case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; - case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; - break; - } - - brw->vb.inputs[i] = el; - } -} - - - -/************************************************************************ - * Constant buffers - */ - -static void brw_set_constant_buffer(struct pipe_context *pipe, - uint shader, uint index, - const struct pipe_constant_buffer *buf) -{ - struct brw_context *brw = brw_context(pipe); - - assert(buf == 0 || index == 0); - - brw->attribs.Constants[shader] = buf; - brw->state.dirty.brw |= BRW_NEW_CONSTANTS; -} - - -/************************************************************************ - * Texture surfaces - */ - - -static void brw_set_sampler_textures(struct pipe_context *pipe, - unsigned num, - struct pipe_texture **texture) -{ - struct brw_context *brw = brw_context(pipe); - uint i; - - assert(num <= PIPE_MAX_SAMPLERS); - - /* Check for no-op */ - if (num == brw->num_textures && - !memcmp(brw->attribs.Texture, texture, num * - sizeof(struct pipe_texture *))) - return; - - for (i = 0; i < num; i++) - pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], - texture[i]); - - for (i = num; i < brw->num_textures; i++) - pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], - NULL); - - brw->num_textures = num; - - brw->state.dirty.brw |= BRW_NEW_TEXTURE; -} - - -/************************************************************************ - * Render targets, etc - */ - -static void brw_set_framebuffer_state(struct pipe_context *pipe, - const struct pipe_framebuffer_state *fb) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.FrameBuffer = *fb; /* struct copy */ - - brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER; -} - - - -/************************************************************************ - * Rasterizer state - */ - -static void * -brw_create_rasterizer_state(struct pipe_context *pipe, - const struct pipe_rasterizer_state *rasterizer) -{ - DUP(pipe_rasterizer_state, rasterizer); -} - -static void brw_bind_rasterizer_state( struct pipe_context *pipe, - void *setup ) -{ - struct brw_context *brw = brw_context(pipe); - - brw->attribs.Raster = (struct pipe_rasterizer_state *)setup; - - /* Also pass-through to draw module: - */ - //draw_set_rasterizer_state(brw->draw, setup); - - brw->state.dirty.brw |= BRW_NEW_RASTERIZER; -} - -static void brw_delete_rasterizer_state(struct pipe_context *pipe, - void *setup) -{ - free(setup); -} - - - -void -brw_init_state_functions( struct brw_context *brw ) -{ - brw->pipe.create_blend_state = brw_create_blend_state; - brw->pipe.bind_blend_state = brw_bind_blend_state; - brw->pipe.delete_blend_state = brw_delete_blend_state; - - brw->pipe.create_sampler_state = brw_create_sampler_state; - brw->pipe.bind_sampler_states = brw_bind_sampler_states; - brw->pipe.delete_sampler_state = brw_delete_sampler_state; - - brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; - brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; - brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; - - brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; - brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; - brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state; - brw->pipe.create_fs_state = brw_create_fs_state; - brw->pipe.bind_fs_state = brw_bind_fs_state; - brw->pipe.delete_fs_state = brw_delete_fs_state; - brw->pipe.create_vs_state = brw_create_vs_state; - brw->pipe.bind_vs_state = brw_bind_vs_state; - brw->pipe.delete_vs_state = brw_delete_vs_state; - - brw->pipe.set_blend_color = brw_set_blend_color; - brw->pipe.set_clip_state = brw_set_clip_state; - brw->pipe.set_constant_buffer = brw_set_constant_buffer; - brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; - -// brw->pipe.set_feedback_state = brw_set_feedback_state; -// brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; - - brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; - brw->pipe.set_scissor_state = brw_set_scissor_state; - brw->pipe.set_sampler_textures = brw_set_sampler_textures; - brw->pipe.set_viewport_state = brw_set_viewport_state; - brw->pipe.set_vertex_buffers = brw_set_vertex_buffers; - brw->pipe.set_vertex_elements = brw_set_vertex_elements; -} +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* Authors: Zack Rusin + * Keith Whitwell + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_dump.h" +#include "tgsi/util/tgsi_parse.h" + +#include "brw_context.h" +#include "brw_defines.h" +#include "brw_state.h" +#include "brw_draw.h" + + +#define DUP( TYPE, VAL ) \ +do { \ + struct TYPE *x = malloc(sizeof(*x)); \ + memcpy(x, VAL, sizeof(*x) ); \ + return x; \ +} while (0) + +/************************************************************************ + * Blend + */ +static void * +brw_create_blend_state(struct pipe_context *pipe, + const struct pipe_blend_state *blend) +{ + DUP( pipe_blend_state, blend ); +} + +static void brw_bind_blend_state(struct pipe_context *pipe, + void *blend) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Blend = (struct pipe_blend_state*)blend; + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + + +static void brw_delete_blend_state(struct pipe_context *pipe, void *blend) +{ + free(blend); +} + +static void brw_set_blend_color( struct pipe_context *pipe, + const struct pipe_blend_color *blend_color ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.BlendColor = *blend_color; + + brw->state.dirty.brw |= BRW_NEW_BLEND; +} + +/************************************************************************ + * Sampler + */ + +static void * +brw_create_sampler_state(struct pipe_context *pipe, + const struct pipe_sampler_state *sampler) +{ + DUP( pipe_sampler_state, sampler ); +} + +static void brw_bind_sampler_states(struct pipe_context *pipe, + unsigned num, void **sampler) +{ + struct brw_context *brw = brw_context(pipe); + + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == brw->num_samplers && + !memcmp(brw->attribs.Samplers, sampler, num * sizeof(void *))) + return; + + memcpy(brw->attribs.Samplers, sampler, num * sizeof(void *)); + memset(&brw->attribs.Samplers[num], 0, (PIPE_MAX_SAMPLERS - num) * + sizeof(void *)); + + brw->num_samplers = num; + + brw->state.dirty.brw |= BRW_NEW_SAMPLER; +} + +static void brw_delete_sampler_state(struct pipe_context *pipe, + void *sampler) +{ + free(sampler); +} + + +/************************************************************************ + * Depth stencil + */ + +static void * +brw_create_depth_stencil_state(struct pipe_context *pipe, + const struct pipe_depth_stencil_alpha_state *depth_stencil) +{ + DUP( pipe_depth_stencil_alpha_state, depth_stencil ); +} + +static void brw_bind_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.DepthStencil = (const struct pipe_depth_stencil_alpha_state *)depth_stencil; + + brw->state.dirty.brw |= BRW_NEW_DEPTH_STENCIL; +} + +static void brw_delete_depth_stencil_state(struct pipe_context *pipe, + void *depth_stencil) +{ + free(depth_stencil); +} + +/************************************************************************ + * Scissor + */ +static void brw_set_scissor_state( struct pipe_context *pipe, + const struct pipe_scissor_state *scissor ) +{ + struct brw_context *brw = brw_context(pipe); + + memcpy( &brw->attribs.Scissor, scissor, sizeof(*scissor) ); + brw->state.dirty.brw |= BRW_NEW_SCISSOR; +} + + +/************************************************************************ + * Stipple + */ + +static void brw_set_polygon_stipple( struct pipe_context *pipe, + const struct pipe_poly_stipple *stipple ) +{ +} + + +/************************************************************************ + * Fragment shader + */ + +static void * brw_create_fs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_fragment_program *brw_fp = CALLOC_STRUCT(brw_fragment_program); + + brw_fp->program.tokens = tgsi_dup_tokens(shader->tokens); + brw_fp->id = brw_context(pipe)->program_id++; + + tgsi_scan_shader(shader->tokens, &brw_fp->info); + +#if 0 + brw_shader_info(shader->tokens, + &brw_fp->info2); +#endif + + tgsi_dump(shader->tokens, 0); + + + return (void *)brw_fp; +} + +static void brw_bind_fs_state(struct pipe_context *pipe, void *shader) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FragmentProgram = (struct brw_fragment_program *)shader; + brw->state.dirty.brw |= BRW_NEW_FS; +} + +static void brw_delete_fs_state(struct pipe_context *pipe, void *shader) +{ + struct brw_fragment_program *brw_fp = (struct brw_fragment_program *) shader; + + FREE((void *) brw_fp->program.tokens); + FREE(brw_fp); +} + + +/************************************************************************ + * Vertex shader and other TNL state + */ + +static void *brw_create_vs_state(struct pipe_context *pipe, + const struct pipe_shader_state *shader) +{ + struct brw_vertex_program *brw_vp = CALLOC_STRUCT(brw_vertex_program); + + brw_vp->program.tokens = tgsi_dup_tokens(shader->tokens); + brw_vp->id = brw_context(pipe)->program_id++; + + tgsi_scan_shader(shader->tokens, &brw_vp->info); + +#if 0 + brw_shader_info(shader->tokens, + &brw_vp->info2); +#endif + tgsi_dump(shader->tokens, 0); + + return (void *)brw_vp; +} + +static void brw_bind_vs_state(struct pipe_context *pipe, void *vs) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.VertexProgram = (struct brw_vertex_program *)vs; + brw->state.dirty.brw |= BRW_NEW_VS; + + debug_printf("YYYYYYYYYYYYY BINDING VERTEX SHADER\n"); +} + +static void brw_delete_vs_state(struct pipe_context *pipe, void *shader) +{ + struct brw_vertex_program *brw_vp = (struct brw_vertex_program *) shader; + + FREE((void *) brw_vp->program.tokens); + FREE(brw_vp); +} + + +static void brw_set_clip_state( struct pipe_context *pipe, + const struct pipe_clip_state *clip ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Clip = *clip; +} + + +static void brw_set_viewport_state( struct pipe_context *pipe, + const struct pipe_viewport_state *viewport ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Viewport = *viewport; /* struct copy */ + brw->state.dirty.brw |= BRW_NEW_VIEWPORT; + + /* pass the viewport info to the draw module */ + //draw_set_viewport_state(brw->draw, viewport); +} + + +static void brw_set_vertex_buffers(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_buffer *buffers) +{ + struct brw_context *brw = brw_context(pipe); + memcpy(brw->vb.vbo_array, buffers, count * sizeof(buffers[0])); +} + +static void brw_set_vertex_elements(struct pipe_context *pipe, + unsigned count, + const struct pipe_vertex_element *elements) +{ + /* flush ? */ + struct brw_context *brw = brw_context(pipe); + uint i; + + assert(count <= PIPE_MAX_ATTRIBS); + + for (i = 0; i < count; i++) { + struct brw_vertex_element_state el; + memset(&el, 0, sizeof(el)); + + el.ve0.src_offset = elements[i].src_offset; + el.ve0.src_format = brw_translate_surface_format(elements[i].src_format); + el.ve0.valid = 1; + el.ve0.vertex_buffer_index = elements[i].vertex_buffer_index; + + el.ve1.dst_offset = i * 4; + + el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC; + el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC; + + switch (elements[i].nr_components) { + case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0; + case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0; + case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT; + break; + } + + brw->vb.inputs[i] = el; + } +} + + + +/************************************************************************ + * Constant buffers + */ + +static void brw_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct brw_context *brw = brw_context(pipe); + + assert(buf == 0 || index == 0); + + brw->attribs.Constants[shader] = buf; + brw->state.dirty.brw |= BRW_NEW_CONSTANTS; +} + + +/************************************************************************ + * Texture surfaces + */ + + +static void brw_set_sampler_textures(struct pipe_context *pipe, + unsigned num, + struct pipe_texture **texture) +{ + struct brw_context *brw = brw_context(pipe); + uint i; + + assert(num <= PIPE_MAX_SAMPLERS); + + /* Check for no-op */ + if (num == brw->num_textures && + !memcmp(brw->attribs.Texture, texture, num * + sizeof(struct pipe_texture *))) + return; + + for (i = 0; i < num; i++) + pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], + texture[i]); + + for (i = num; i < brw->num_textures; i++) + pipe_texture_reference((struct pipe_texture **) &brw->attribs.Texture[i], + NULL); + + brw->num_textures = num; + + brw->state.dirty.brw |= BRW_NEW_TEXTURE; +} + + +/************************************************************************ + * Render targets, etc + */ + +static void brw_set_framebuffer_state(struct pipe_context *pipe, + const struct pipe_framebuffer_state *fb) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.FrameBuffer = *fb; /* struct copy */ + + brw->state.dirty.brw |= BRW_NEW_FRAMEBUFFER; +} + + + +/************************************************************************ + * Rasterizer state + */ + +static void * +brw_create_rasterizer_state(struct pipe_context *pipe, + const struct pipe_rasterizer_state *rasterizer) +{ + DUP(pipe_rasterizer_state, rasterizer); +} + +static void brw_bind_rasterizer_state( struct pipe_context *pipe, + void *setup ) +{ + struct brw_context *brw = brw_context(pipe); + + brw->attribs.Raster = (struct pipe_rasterizer_state *)setup; + + /* Also pass-through to draw module: + */ + //draw_set_rasterizer_state(brw->draw, setup); + + brw->state.dirty.brw |= BRW_NEW_RASTERIZER; +} + +static void brw_delete_rasterizer_state(struct pipe_context *pipe, + void *setup) +{ + free(setup); +} + + + +void +brw_init_state_functions( struct brw_context *brw ) +{ + brw->pipe.create_blend_state = brw_create_blend_state; + brw->pipe.bind_blend_state = brw_bind_blend_state; + brw->pipe.delete_blend_state = brw_delete_blend_state; + + brw->pipe.create_sampler_state = brw_create_sampler_state; + brw->pipe.bind_sampler_states = brw_bind_sampler_states; + brw->pipe.delete_sampler_state = brw_delete_sampler_state; + + brw->pipe.create_depth_stencil_alpha_state = brw_create_depth_stencil_state; + brw->pipe.bind_depth_stencil_alpha_state = brw_bind_depth_stencil_state; + brw->pipe.delete_depth_stencil_alpha_state = brw_delete_depth_stencil_state; + + brw->pipe.create_rasterizer_state = brw_create_rasterizer_state; + brw->pipe.bind_rasterizer_state = brw_bind_rasterizer_state; + brw->pipe.delete_rasterizer_state = brw_delete_rasterizer_state; + brw->pipe.create_fs_state = brw_create_fs_state; + brw->pipe.bind_fs_state = brw_bind_fs_state; + brw->pipe.delete_fs_state = brw_delete_fs_state; + brw->pipe.create_vs_state = brw_create_vs_state; + brw->pipe.bind_vs_state = brw_bind_vs_state; + brw->pipe.delete_vs_state = brw_delete_vs_state; + + brw->pipe.set_blend_color = brw_set_blend_color; + brw->pipe.set_clip_state = brw_set_clip_state; + brw->pipe.set_constant_buffer = brw_set_constant_buffer; + brw->pipe.set_framebuffer_state = brw_set_framebuffer_state; + +// brw->pipe.set_feedback_state = brw_set_feedback_state; +// brw->pipe.set_feedback_buffer = brw_set_feedback_buffer; + + brw->pipe.set_polygon_stipple = brw_set_polygon_stipple; + brw->pipe.set_scissor_state = brw_set_scissor_state; + brw->pipe.set_sampler_textures = brw_set_sampler_textures; + brw->pipe.set_viewport_state = brw_set_viewport_state; + brw->pipe.set_vertex_buffers = brw_set_vertex_buffers; + brw->pipe.set_vertex_elements = brw_set_vertex_elements; +} -- cgit v1.2.3 From a0e2955a16a8a04afe7f84b1c8551211542a6fbd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 27 May 2008 00:13:57 +0900 Subject: pipebuffer: Allow slab allocations of buffers of inequal size. --- src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c index 45ba158a4d..b9dff09804 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c @@ -312,8 +312,8 @@ pb_slab_manager_create_buffer(struct pb_manager *_mgr, struct list_head *list; /* check size */ - assert(size == mgr->bufSize); - if(size != mgr->bufSize) + assert(size <= mgr->bufSize); + if(size > mgr->bufSize) return NULL; /* check if we can provide the requested alignment */ -- cgit v1.2.3 From 35cfd0a4900750f67cba4f64929ff3347f0bd46f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 27 May 2008 00:19:41 +0900 Subject: pipebuffer: Malloc buffer provider. Simple wrapper around pb_malloc_buffer_create for convenience. --- .../auxiliary/pipebuffer/pb_buffer_malloc.c | 33 +++++++++++++++++++++- src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 9 ++++++ 2 files changed, 41 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c index 9e8244f909..6c3502eea7 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c @@ -30,13 +30,14 @@ * Implementation of malloc-based buffers to store data that can't be processed * by the hardware. * - * \author José Fonseca + * \author Jose Fonseca */ #include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pb_buffer.h" +#include "pb_bufmgr.h" struct malloc_buffer @@ -125,3 +126,33 @@ pb_malloc_buffer_create(size_t size, return &buf->base; } + + +static struct pb_buffer * +pb_malloc_buffer_create_buffer(struct pb_manager *mgr, + size_t size, + const struct pb_desc *desc) +{ + return pb_malloc_buffer_create(size, desc); +} + + +static void +pb_malloc_bufmgr_destroy(struct pb_manager *mgr) +{ + /* No-op */ +} + + +static struct pb_manager +pb_malloc_bufmgr = { + pb_malloc_buffer_create_buffer, + pb_malloc_bufmgr_destroy +}; + + +struct pb_manager * +pb_malloc_bufmgr_create(void) +{ + return &pb_malloc_bufmgr; +} diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index f6cc7a525b..4a922d16c1 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -79,6 +79,15 @@ struct pb_manager }; +/** + * Malloc buffer provider. + * + * Simple wrapper around pb_malloc_buffer_create for convenience. + */ +struct pb_manager * +pb_malloc_bufmgr_create(void); + + /** * Static buffer pool sub-allocator. * -- cgit v1.2.3 From 529762d5df6d9427f9fa0927e38b9886e412a6bc Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 26 May 2008 18:29:47 +0100 Subject: draw: add missing break --- src/gallium/auxiliary/draw/draw_pt_emit.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 22a83ec78f..e21af3d287 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -100,6 +100,7 @@ void draw_pt_emit_prepare( struct pt_emit *emit, case EMIT_4UB: output_format = PIPE_FORMAT_B8G8R8A8_UNORM; emit_sz = 4 * sizeof(ubyte); + break; default: assert(0); output_format = PIPE_FORMAT_NONE; -- cgit v1.2.3 From d80c24a81a9a46c132fe877dde6919a57cacf8c0 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 26 May 2008 18:37:34 +0100 Subject: draw: defensively reset render primitive, which can get clobbered by clipping --- src/gallium/auxiliary/draw/draw_pt_emit.c | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index e21af3d287..01396a749e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -40,6 +40,7 @@ struct pt_emit { struct translate *translate; struct translate_cache *cache; + unsigned prim; }; void draw_pt_emit_prepare( struct pt_emit *emit, @@ -51,8 +52,14 @@ void draw_pt_emit_prepare( struct pt_emit *emit, struct translate_key hw_key; unsigned i; boolean ok; + - ok = draw->render->set_primitive(draw->render, prim); + /* XXX: may need to defensively reset this later on as clipping can + * clobber this state in the render backend. + */ + emit->prim = prim; + + ok = draw->render->set_primitive(draw->render, emit->prim); if (!ok) { assert(0); return; @@ -145,6 +152,14 @@ void draw_pt_emit( struct pt_emit *emit, */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + /* XXX: and work out some way to coordinate the render primitive + * between vbuf.c and here... + */ + if (!draw->render->set_primitive(draw->render, emit->prim)) { + assert(0); + return; + } + hw_verts = render->allocate_vertices(render, (ushort)translate->key.output_stride, (ushort)vertex_count); -- cgit v1.2.3 From 93bfc94c351a2eafd43ac7a20b362d969f98d86a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 26 May 2008 18:54:35 +0100 Subject: draw: defensively reset render primitive some more --- src/gallium/auxiliary/draw/draw_pt_emit.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 01396a749e..d0c9577ee6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -213,6 +213,14 @@ void draw_pt_emit_linear(struct pt_emit *emit, */ draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + /* XXX: and work out some way to coordinate the render primitive + * between vbuf.c and here... + */ + if (!draw->render->set_primitive(draw->render, emit->prim)) { + assert(0); + return; + } + hw_verts = render->allocate_vertices(render, (ushort)translate->key.output_stride, (ushort)count); -- cgit v1.2.3 From aa16e3a2750993afdba16c24237bb6d8d8e4d91a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 26 May 2008 19:10:44 +0100 Subject: draw: defensively flush pipeline backend when setting primitive --- src/gallium/auxiliary/draw/draw_pt_emit.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index d0c9577ee6..e37256b1ff 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -53,6 +53,10 @@ void draw_pt_emit_prepare( struct pt_emit *emit, unsigned i; boolean ok; + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + /* XXX: may need to defensively reset this later on as clipping can * clobber this state in the render backend. -- cgit v1.2.3 From 635abed109b26ded34954c379b80e306c602384b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 26 May 2008 13:30:10 -0600 Subject: gallium: free renderbuffer's old texture in st_renderbuffer_alloc_storage() --- src/mesa/state_tracker/st_cb_fbo.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 3560a040c8..76145a4d5c 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -92,10 +92,11 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, struct pipe_texture template; unsigned surface_usage; - /* Free the old surface (and texture if we hold the last - * reference): + /* Free the old surface and texture */ pipe_surface_reference( &strb->surface, NULL ); + pipe_texture_reference( &strb->texture, NULL ); + memset(&template, 0, sizeof(template)); -- cgit v1.2.3 From f172148c6e3246ffd7c3a33f16d2ac6faca9d6cf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 26 May 2008 13:35:29 -0600 Subject: gallium: free renderbuffer's old texture in st_renderbuffer_alloc_storage() --- src/mesa/state_tracker/st_cb_fbo.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 3560a040c8..76145a4d5c 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -92,10 +92,11 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, struct pipe_texture template; unsigned surface_usage; - /* Free the old surface (and texture if we hold the last - * reference): + /* Free the old surface and texture */ pipe_surface_reference( &strb->surface, NULL ); + pipe_texture_reference( &strb->texture, NULL ); + memset(&template, 0, sizeof(template)); -- cgit v1.2.3 From dc1537bc25c7cbff0a41034ece0830146616f036 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 09:48:32 +0100 Subject: ffvertex: don't compute whole eye vector if only eye.z is required --- src/mesa/main/ffvertex_prog.c | 30 +++++++++++++++++++++++++++--- 1 file changed, 27 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index a627a21f65..a790d4b142 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -305,7 +305,7 @@ static struct state_key *make_state_key( GLcontext *ctx ) * generated program with line/function references for each * instruction back into this file: */ -#define DISASSEM (MESA_VERBOSE&VERBOSE_DISASSEM) +#define DISASSEM 1 /* Should be tunable by the driver - do we want to do matrix * multiplications with DP4's or with MUL/MAD's? SSE works better @@ -344,6 +344,7 @@ struct tnl_program { GLuint temp_reserved; struct ureg eye_position; + struct ureg eye_position_z; struct ureg eye_position_normalized; struct ureg transformed_normal; struct ureg identity; @@ -728,6 +729,28 @@ static struct ureg get_eye_position( struct tnl_program *p ) } +static struct ureg get_eye_position_z( struct tnl_program *p ) +{ + if (!is_undef(p->eye_position)) + return swizzle1(p->eye_position, Z); + + if (is_undef(p->eye_position_z)) { + struct ureg pos = register_input( p, VERT_ATTRIB_POS ); + struct ureg modelview[4]; + + p->eye_position_z = reserve_temp(p); + + register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3, + 0, modelview ); + + emit_op2(p, OPCODE_DP4, p->eye_position_z, 0, pos, modelview[2]); + } + + return p->eye_position_z; +} + + + static struct ureg get_eye_position_normalized( struct tnl_program *p ) { if (is_undef(p->eye_position_normalized)) { @@ -1240,7 +1263,7 @@ static void build_fog( struct tnl_program *p ) struct ureg input; if (p->state->fog_source_is_depth) { - input = swizzle1(get_eye_position(p), Z); + input = get_eye_position_z(p); } else { input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X); @@ -1470,7 +1493,7 @@ static void build_texture_transform( struct tnl_program *p ) static void build_pointsize( struct tnl_program *p ) { - struct ureg eye = get_eye_position(p); + struct ureg eye = get_eye_position_z(p); struct ureg state_size = register_param1(p, STATE_POINT_SIZE); struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION); struct ureg out = register_output(p, VERT_RESULT_PSIZ); @@ -1568,6 +1591,7 @@ create_new_program( const struct state_key *key, p.state = key; p.program = program; p.eye_position = undef; + p.eye_position_z = undef; p.eye_position_normalized = undef; p.transformed_normal = undef; p.identity = undef; -- cgit v1.2.3 From 2109ddafefde26dd20a1c6a25f594984143944a3 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 10:35:33 +0100 Subject: ffvertex: emit full LIT when attenuating (needs the 1 in X position) --- src/mesa/main/ffvertex_prog.c | 50 +++++++++++++++++++++++++++---------------- src/mesa/main/light.c | 1 + 2 files changed, 32 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index a790d4b142..2ef3286e57 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -1161,15 +1161,6 @@ static void build_lighting( struct tnl_program *p ) struct ureg res0, res1; GLuint mask0, mask1; - if (p->state->material_shininess_is_zero) { - emit_degenerate_lit(p, lit, dots); - } else { - emit_op1(p, OPCODE_LIT, lit, 0, dots); - } - - if (!is_undef(att)) - emit_op2(p, OPCODE_MUL, lit, 0, lit, att); - if (count == nr_lights) { if (separate) { @@ -1191,7 +1182,21 @@ static void build_lighting( struct tnl_program *p ) res1 = _col1; } - emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0); + + if (!is_undef(att)) { + emit_op1(p, OPCODE_LIT, lit, 0, dots); + emit_op2(p, OPCODE_MUL, lit, 0, lit, att); + emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0); + } + else if (!p->state->material_shininess_is_zero) { + emit_op1(p, OPCODE_LIT, lit, 0, dots); + emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0); + } + else { + emit_degenerate_lit(p, lit, dots); + emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0); + } + emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0); emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1); @@ -1209,15 +1214,6 @@ static void build_lighting( struct tnl_program *p ) struct ureg res0, res1; GLuint mask0, mask1; - if (p->state->material_shininess_is_zero) { - emit_degenerate_lit(p, lit, negate(swizzle(dots,X,Y,W,Z))); - } else { - emit_op1(p, OPCODE_LIT, lit, 0, negate(swizzle(dots,X,Y,W,Z))); - } - - if (!is_undef(att)) - emit_op2(p, OPCODE_MUL, lit, 0, lit, att); - if (count == nr_lights) { if (separate) { mask0 = WRITEMASK_XYZ; @@ -1238,6 +1234,22 @@ static void build_lighting( struct tnl_program *p ) mask1 = 0; } + dots = negate(swizzle(dots,X,Y,W,Z)); + + if (!is_undef(att)) { + emit_op1(p, OPCODE_LIT, lit, 0, dots); + emit_op2(p, OPCODE_MUL, lit, 0, lit, att); + emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0); + } + else if (!p->state->material_shininess_is_zero) { + emit_op1(p, OPCODE_LIT, lit, 0, dots); + emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0); + } + else { + emit_degenerate_lit(p, lit, dots); + emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0); + } + emit_op2(p, OPCODE_ADD, _bfc0, 0, ambient, _bfc0); emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0); emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1); diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c index 6e057614ba..0af647e13c 100644 --- a/src/mesa/main/light.c +++ b/src/mesa/main/light.c @@ -1357,6 +1357,7 @@ _mesa_init_lighting( GLcontext *ctx ) /* Miscellaneous */ ctx->Light._NeedEyeCoords = GL_FALSE; ctx->_NeedEyeCoords = GL_FALSE; + ctx->_ForceEyeCoords = GL_TRUE; ctx->_ModelViewInvScale = 1.0; } -- cgit v1.2.3 From 4e2567f0ab6afd701bea4c35e388663e90f5cb6c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 10:42:58 +0100 Subject: draw: some possible fixes for spilling --- src/gallium/auxiliary/draw/draw_vs_aos.c | 30 ++++++++++++++++++------------ 1 file changed, 18 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 1622358ae1..99630e4f75 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -253,6 +253,7 @@ struct x86_reg aos_get_xmm_reg( struct aos_compilation *cp ) cp->xmm[oldest].file = TGSI_FILE_NULL; cp->xmm[oldest].idx = 0; + cp->xmm[oldest].dirty = 0; cp->xmm[oldest].last_used = cp->insn_counter; return x86_make_reg(file_XMM, oldest); } @@ -284,24 +285,18 @@ void aos_adopt_xmm_reg( struct aos_compilation *cp, return; } - /* If this xmm reg is already holding this shader reg, just update - * last_used, and don't clobber the dirty flag... - */ - if (cp->xmm[reg.idx].file == file && - cp->xmm[reg.idx].idx == idx) - { - cp->xmm[reg.idx].dirty |= dirty; - cp->xmm[reg.idx].last_used = cp->insn_counter; - return; - } - /* If any xmm reg thinks it holds this shader reg, break the * illusion. */ for (i = 0; i < 8; i++) { if (cp->xmm[i].file == file && - cp->xmm[i].idx == idx) { + cp->xmm[i].idx == idx) + { + /* If an xmm reg is already holding this shader reg, take into account its + * dirty flag... + */ + dirty |= cp->xmm[i].dirty; aos_release_xmm_reg(cp, i); } } @@ -1989,6 +1984,17 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, debug_printf("\n"); } + + { + unsigned i; + for (i = 0; i < 8; i++) { + if (cp.xmm[i].file != TGSI_FILE_OUTPUT) { + cp.xmm[i].file = TGSI_FILE_NULL; + cp.xmm[i].dirty = 0; + } + } + } + if (cp.error) goto fail; -- cgit v1.2.3 From 5dc44184fa9f07465b7ff2be94394c55392ce5e9 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 11:10:50 +0100 Subject: draw: fix writemask/shufps confusion --- src/gallium/auxiliary/draw/draw_vs_aos.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 99630e4f75..434bd2a9ab 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -687,17 +687,18 @@ static void store_dest( struct aos_compilation *cp, sse_movss(cp->func, dst, get_xmm(cp, result)); break; - case TGSI_WRITEMASK_XY: + case TGSI_WRITEMASK_ZW: sse_shufps(cp->func, dst, get_xmm(cp, result), SHUF(X, Y, Z, W)); break; - case TGSI_WRITEMASK_ZW: + case TGSI_WRITEMASK_XY: result = get_xmm_writable(cp, result); sse_shufps(cp->func, result, dst, SHUF(X, Y, Z, W)); dst = result; break; case TGSI_WRITEMASK_YZW: + result = get_xmm_writable(cp, result); sse_movss(cp->func, result, dst); dst = result; break; @@ -891,7 +892,7 @@ static void emit_print( struct aos_compilation *cp, unsigned idx ) { struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); - struct x86_reg arg = get_reg_ptr( cp, file, idx ); + struct x86_reg arg = aos_get_shader_reg_ptr( cp, file, idx ); unsigned i; /* There shouldn't be anything on the x87 stack. Can add this -- cgit v1.2.3 From adaaa29218f1babad874f50681ca971fdd3b8a40 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 11:12:42 +0100 Subject: tgsi: export utils for dumping individual instructions --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 24 ++++++++++++------------ src/gallium/auxiliary/tgsi/util/tgsi_dump.h | 18 ++++++++++++++++++ 2 files changed, 30 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 4c65ffd780..648afa2a51 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -539,9 +539,9 @@ static const char *TGSI_MODULATES[] = "MODULATE_EIGHTH" }; -static void -dump_declaration_short( - struct tgsi_full_declaration *decl ) +void +tgsi_dump_declaration( + const struct tgsi_full_declaration *decl ) { TXT( "\nDCL " ); ENM( decl->Declaration.File, TGSI_FILES_SHORT ); @@ -672,9 +672,9 @@ dump_declaration_verbose( } } -static void -dump_immediate_short( - struct tgsi_full_immediate *imm ) +void +tgsi_dump_immediate( + const struct tgsi_full_immediate *imm ) { unsigned i; @@ -727,9 +727,9 @@ dump_immediate_verbose( } } -static void -dump_instruction_short( - struct tgsi_full_instruction *inst, +void +tgsi_dump_instruction( + const struct tgsi_full_instruction *inst, unsigned instno ) { unsigned i; @@ -1281,17 +1281,17 @@ tgsi_dump( switch( parse.FullToken.Token.Type ) { case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_short( + tgsi_dump_declaration( &parse.FullToken.FullDeclaration ); break; case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_short( + tgsi_dump_immediate( &parse.FullToken.FullImmediate ); break; case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_short( + tgsi_dump_instruction( &parse.FullToken.FullInstruction, instno ); instno++; diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h index beb0155d56..ca83bdef20 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h @@ -14,6 +14,24 @@ tgsi_dump( const struct tgsi_token *tokens, unsigned flags ); +struct tgsi_full_immediate; +struct tgsi_full_instruction; +struct tgsi_full_declaration; + +void +tgsi_dump_immediate( + const struct tgsi_full_immediate *imm ); + +void +tgsi_dump_instruction( + const struct tgsi_full_instruction *inst, + unsigned instno ); + +void +tgsi_dump_declaration( + const struct tgsi_full_declaration *decl ); + + #if defined __cplusplus } #endif -- cgit v1.2.3 From f7946bc7c0435ab2926cd729dfd8312222a3aa2a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 11:15:31 +0100 Subject: draw: dump individual instructions as they are processed --- src/gallium/auxiliary/draw/draw_vs_aos.c | 26 ++++++++++++++++++++++++-- 1 file changed, 24 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 434bd2a9ab..d3989fe107 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -42,7 +42,20 @@ #include "rtasm/rtasm_x86sse.h" #ifdef PIPE_ARCH_X86 +#define DISASSEM 0 +static const char *files[] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM", + "INTERNAL", +}; static INLINE boolean eq( struct x86_reg a, struct x86_reg b ) @@ -184,7 +197,11 @@ static void spill( struct aos_compilation *cp, unsigned idx ) struct x86_reg oldval = get_reg_ptr(cp, cp->xmm[idx].file, cp->xmm[idx].idx); - + + if (0) debug_printf("\nspill %s[%d]", + files[cp->xmm[idx].file], + cp->xmm[idx].idx); + assert(cp->xmm[idx].dirty); sse_movaps(cp->func, oldval, x86_make_reg(file_XMM, idx)); cp->xmm[idx].dirty = 0; @@ -1975,6 +1992,9 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, break; case TGSI_TOKEN_TYPE_INSTRUCTION: + if (DISASSEM) + tgsi_dump_instruction( &parse.FullToken.FullInstruction, cp.insn_counter ); + if (!emit_instruction( &cp, &parse.FullToken.FullInstruction )) goto fail; break; @@ -1982,7 +2002,9 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, x87_assert_stack_empty(cp.func); cp.insn_counter++; - debug_printf("\n"); + + if (DISASSEM) + debug_printf("\n"); } -- cgit v1.2.3 From 50c1d329b95ad78e03ca4d537daee4d11f308c7a Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 11:58:55 +0100 Subject: draw: restore extras path in draw_pt_vcache.c, keep pipeline flags out of non-pipeline elts --- src/gallium/auxiliary/draw/draw_pt_vcache.c | 117 +++++++++++++++------------- 1 file changed, 65 insertions(+), 52 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 6c17edba34..96e02fbf3a 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -104,10 +104,22 @@ static INLINE void vcache_elt( struct vcache_frontend *vcache, static void vcache_triangle( struct vcache_frontend *vcache, - ushort flags, unsigned i0, unsigned i1, unsigned i2 ) +{ + vcache_elt(vcache, i0, 0); + vcache_elt(vcache, i1, 0); + vcache_elt(vcache, i2, 0); + vcache_check_flush(vcache); +} + + +static void vcache_triangle_flags( struct vcache_frontend *vcache, + ushort flags, + unsigned i0, + unsigned i1, + unsigned i2 ) { vcache_elt(vcache, i0, flags); vcache_elt(vcache, i1, 0); @@ -116,9 +128,19 @@ static void vcache_triangle( struct vcache_frontend *vcache, } static void vcache_line( struct vcache_frontend *vcache, - ushort flags, unsigned i0, unsigned i1 ) +{ + vcache_elt(vcache, i0, 0); + vcache_elt(vcache, i1, 0); + vcache_check_flush(vcache); +} + + +static void vcache_line_flags( struct vcache_frontend *vcache, + ushort flags, + unsigned i0, + unsigned i1 ) { vcache_elt(vcache, i0, flags); vcache_elt(vcache, i1, 0); @@ -139,63 +161,46 @@ static void vcache_quad( struct vcache_frontend *vcache, unsigned i2, unsigned i3 ) { - vcache_triangle( vcache, - ( DRAW_PIPE_RESET_STIPPLE | - DRAW_PIPE_EDGE_FLAG_0 | - DRAW_PIPE_EDGE_FLAG_2 ), - i0, i1, i3 ); - - vcache_triangle( vcache, - ( DRAW_PIPE_EDGE_FLAG_0 | - DRAW_PIPE_EDGE_FLAG_1 ), - i1, i2, i3 ); + vcache_triangle( vcache, i0, i1, i3 ); + vcache_triangle( vcache, i1, i2, i3 ); +} + +static void vcache_ef_quad( struct vcache_frontend *vcache, + unsigned i0, + unsigned i1, + unsigned i2, + unsigned i3 ) +{ + vcache_triangle_flags( vcache, + ( DRAW_PIPE_RESET_STIPPLE | + DRAW_PIPE_EDGE_FLAG_0 | + DRAW_PIPE_EDGE_FLAG_2 ), + i0, i1, i3 ); + + vcache_triangle_flags( vcache, + ( DRAW_PIPE_EDGE_FLAG_0 | + DRAW_PIPE_EDGE_FLAG_1 ), + i1, i2, i3 ); } /* At least for now, we're back to using a template include file for * this. The two paths aren't too different though - it may be * possible to reunify them. */ -#define TRIANGLE(flags,i0,i1,i2) \ - vcache_triangle(vcache, \ - flags, \ - get_elt(elts,i0), \ - get_elt(elts,i1), \ - get_elt(elts,i2)) - -#define QUAD(i0,i1,i2,i3) \ - vcache_quad(vcache, \ - get_elt(elts,i0), \ - get_elt(elts,i1), \ - get_elt(elts,i2), \ - get_elt(elts,i3)) - -#define LINE(flags,i0,i1) \ - vcache_line(vcache, \ - flags, \ - get_elt(elts,i0), \ - get_elt(elts,i1)) - -#define POINT(i0) \ - vcache_point(vcache, \ - get_elt(elts,i0)) - +#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle_flags(vc,flags,i0,i1,i2) +#define QUAD(vc,i0,i1,i2,i3) vcache_ef_quad(vc,i0,i1,i2,i3) +#define LINE(vc,flags,i0,i1) vcache_line_flags(vc,flags,i0,i1) +#define POINT(vc,i0) vcache_point(vc,i0) +#define FUNC vcache_run_extras +#include "draw_pt_vcache_tmp.h" + +#define TRIANGLE(vc,flags,i0,i1,i2) vcache_triangle(vc,i0,i1,i2) +#define QUAD(vc,i0,i1,i2,i3) vcache_quad(vc,i0,i1,i2,i3) +#define LINE(vc,flags,i0,i1) vcache_line(vc,i0,i1) +#define POINT(vc,i0) vcache_point(vc,i0) #define FUNC vcache_run -#define ARGS \ - struct draw_pt_front_end *frontend, \ - pt_elt_func get_elt, \ - const void *elts - -#define LOCAL_VARS \ - struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; \ - struct draw_context *draw = vcache->draw; \ - boolean flatfirst = (draw->rasterizer->flatshade && \ - draw->rasterizer->flatshade_first); \ - unsigned prim = vcache->input_prim; \ - unsigned i, flags; +#include "draw_pt_vcache_tmp.h" -#define FLUSH vcache_flush( vcache ) - -#include "draw_pt_decompose.h" @@ -208,7 +213,15 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; - vcache->base.run = vcache_run; + if (opt & PT_PIPELINE) + { + vcache->base.run = vcache_run_extras; + } + else + { + vcache->base.run = vcache_run; + } + vcache->input_prim = prim; vcache->output_prim = draw_pt_reduced_prim(prim); -- cgit v1.2.3 From a08c574bfcf72c7f7ffbeb35c10347b491ef87fb Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 12:26:23 +0100 Subject: draw: hook up viewport / rhw emit to varient key state --- .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 4 +- src/gallium/auxiliary/draw/draw_vs_aos.c | 16 +++---- src/gallium/auxiliary/draw/draw_vs_varient.c | 52 ++++++++++++++++++---- 3 files changed, 53 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 7fefd391a6..2f2e7195b3 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -94,8 +94,8 @@ static void fse_prepare( struct draw_pt_middle_end *middle, fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ num_vs_inputs); /* inputs - fetch from api format */ - fse->key.viewport = 1; - fse->key.clip = 0; + fse->key.viewport = !draw->identity_viewport; + fse->key.clip = !draw->bypass_clipping; fse->key.pad = 0; memset(fse->key.element, 0, diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index d3989fe107..c63553216c 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -2021,11 +2021,14 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, if (cp.error) goto fail; - if (cp.vaos->base.key.viewport) { - if (0) - emit_viewport(&cp); - else - emit_rhw_viewport(&cp); + if (cp.vaos->base.key.clip) { + /* not really handling clipping, just do the rhw so we can + * see the results... + */ + emit_rhw_viewport(&cp); + } + else if (cp.vaos->base.key.viewport) { + emit_viewport(&cp); } /* Emit output... TODO: do this eagerly after the last write to a @@ -2188,9 +2191,6 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, { struct draw_vs_varient_aos_sse *vaos = CALLOC_STRUCT(draw_vs_varient_aos_sse); - if (key->clip) - return NULL; - if (!vaos) goto fail; diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index c15c648527..d4deabfff3 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -89,9 +89,9 @@ static void vsvg_set_input( struct draw_vs_varient *varient, /* Mainly for debug at this stage: */ -static void do_viewport( struct draw_vs_varient_generic *vsvg, - unsigned count, - void *output_buffer ) +static void do_rhw_viewport( struct draw_vs_varient_generic *vsvg, + unsigned count, + void *output_buffer ) { char *ptr = (char *)output_buffer; const float *scale = vsvg->viewport.scale; @@ -109,6 +109,25 @@ static void do_viewport( struct draw_vs_varient_generic *vsvg, data[3] = w; } } + +static void do_viewport( struct draw_vs_varient_generic *vsvg, + unsigned count, + void *output_buffer ) +{ + char *ptr = (char *)output_buffer; + const float *scale = vsvg->viewport.scale; + const float *trans = vsvg->viewport.translate; + unsigned stride = vsvg->base.key.output_stride; + unsigned j; + + for (j = 0; j < count; j++, ptr += stride) { + float *data = (float *)ptr; + + data[0] = data[0] * scale[0] + trans[0]; + data[1] = data[1] * scale[1] + trans[1]; + data[2] = data[2] * scale[2] + trans[2]; + } +} static void vsvg_run_elts( struct draw_vs_varient *varient, @@ -136,10 +155,20 @@ static void vsvg_run_elts( struct draw_vs_varient *varient, vsvg->base.key.output_stride, vsvg->base.key.output_stride); - if (vsvg->base.key.viewport) + + if (vsvg->base.key.clip) { + /* not really handling clipping, just do the rhw so we can + * see the results... + */ + do_rhw_viewport( vsvg, + count, + output_buffer ); + } + else if (vsvg->base.key.viewport) { do_viewport( vsvg, count, output_buffer ); + } //if (!vsvg->already_in_emit_format) @@ -182,11 +211,19 @@ static void vsvg_run_linear( struct draw_vs_varient *varient, vsvg->base.key.output_stride, vsvg->base.key.output_stride); - if (vsvg->base.key.viewport) + if (vsvg->base.key.clip) { + /* not really handling clipping, just do the rhw so we can + * see the results... + */ + do_rhw_viewport( vsvg, + count, + output_buffer ); + } + else if (vsvg->base.key.viewport) { do_viewport( vsvg, count, output_buffer ); - + } //if (!vsvg->already_in_emit_format) vsvg->emit->set_buffer( vsvg->emit, @@ -224,9 +261,6 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, unsigned i; struct translate_key fetch, emit; - if (key->clip) - return NULL; - struct draw_vs_varient_generic *vsvg = CALLOC_STRUCT( draw_vs_varient_generic ); if (vsvg == NULL) return NULL; -- cgit v1.2.3 From 9752ebd99e16646fed3c14712fc0af2c34c9e48f Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 27 May 2008 12:59:46 +0200 Subject: draw: Fix for EMIT_4UB case --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 2f2e7195b3..c6249b4b41 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -143,7 +143,10 @@ static void fse_prepare( struct draw_pt_middle_end *middle, output_format = PIPE_FORMAT_R32_FLOAT; emit_sz = 1 * sizeof(float); vs_output = num_vs_outputs + 1; - + break; + case EMIT_4UB: + output_format = PIPE_FORMAT_B8G8R8A8_UNORM; + emit_sz = 4 * sizeof(ubyte); break; default: assert(0); -- cgit v1.2.3 From 9f9f6c21be105de41a58128605b911e679efe8f0 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 27 May 2008 13:15:52 +0200 Subject: draw: Fixed typo in draw_pt_varray and added comments --- src/gallium/auxiliary/draw/draw_pt_varray.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 06fd866ccd..260f28f284 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -158,14 +158,14 @@ static INLINE void varray_point( struct varray_frontend *varray, static unsigned decompose_prim[PIPE_PRIM_POLYGON + 1] = { PIPE_PRIM_POINTS, PIPE_PRIM_LINES, + PIPE_PRIM_LINES, /* decomposed LINELOOP */ PIPE_PRIM_LINE_STRIP, - PIPE_PRIM_LINES, /* decomposed */ PIPE_PRIM_TRIANGLES, PIPE_PRIM_TRIANGLE_STRIP, - PIPE_PRIM_TRIANGLES, /* decomposed */ + PIPE_PRIM_TRIANGLES, /* decomposed TRI_FAN */ PIPE_PRIM_QUADS, PIPE_PRIM_QUAD_STRIP, - PIPE_PRIM_TRIANGLES /* decomposed */ + PIPE_PRIM_TRIANGLES /* decomposed POLYGON */ }; -- cgit v1.2.3 From f64c44ad3e55467ce964871502445cf5a0fb46d6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 14:17:57 +0100 Subject: draw: remove dead file --- src/gallium/auxiliary/draw/draw_pt_middle_fse.c | 705 ------------------------ 1 file changed, 705 deletions(-) delete mode 100644 src/gallium/auxiliary/draw/draw_pt_middle_fse.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_middle_fse.c b/src/gallium/auxiliary/draw/draw_pt_middle_fse.c deleted file mode 100644 index 643ea151c1..0000000000 --- a/src/gallium/auxiliary/draw/draw_pt_middle_fse.c +++ /dev/null @@ -1,705 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - /* - * Authors: - * Keith Whitwell - */ - - -#include "pipe/p_util.h" -#include "draw/draw_context.h" -#include "draw/draw_private.h" -#include "draw/draw_vbuf.h" -#include "draw/draw_vertex.h" -#include "draw/draw_pt.h" -#include "draw/draw_vs.h" - -#include "translate/translate.h" - -struct fetch_shade_emit; - -struct fse_shader { - struct translate_key key; - - void (*run_linear)( const struct fetch_shade_emit *fse, - unsigned start, - unsigned count, - char *buffer ); - - void (*run_elts)( const struct fetch_shade_emit *fse, - const unsigned *fetch_elts, - unsigned fetch_count, - char *buffer ); - -}; - -/* Prototype fetch, shade, emit-hw-verts all in one go. - */ -struct fetch_shade_emit { - struct draw_pt_middle_end base; - struct draw_context *draw; - - struct translate_key key; - - /* Temporaries: - */ - const float *constants; - unsigned pitch[PIPE_MAX_ATTRIBS]; - const ubyte *src[PIPE_MAX_ATTRIBS]; - unsigned prim; - - /* Points to one of the three hardwired example shaders, below: - */ - struct fse_shader *active; - - /* Temporary: A list of hard-wired shaders. Of course the plan - * would be to generate these for a given (vertex-shader, - * translate-key) pair... - */ - struct fse_shader shader[10]; - int nr_shaders; -}; - - - -/* Not quite passthrough yet -- we're still running the 'shader' here, - * inlined into the vertex fetch function. - */ -static void shader0_run_linear( const struct fetch_shade_emit *fse, - unsigned start, - unsigned count, - char *buffer ) -{ - unsigned i; - - const float *m = fse->constants; - const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; - const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; - const ubyte *st = fse->src[2] + start * fse->pitch[2]; - - float *out = (float *)buffer; - - /* loop over vertex attributes (vertex shader inputs) - */ - for (i = 0; i < count; i++) { - { - const float *in = (const float *)xyz; - const float ix = in[0], iy = in[1], iz = in[2]; - - out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; - out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; - out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; - out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; - xyz += fse->pitch[0]; - } - - { - const float *in = (const float *)rgb; - out[4] = in[0]; - out[5] = in[1]; - out[6] = in[2]; - out[7] = 1.0f; - rgb += fse->pitch[1]; - } - - { - const float *in = (const float *)st; - out[8] = in[0]; - out[9] = in[1]; - out[10] = 0.0f; - out[11] = 1.0f; - st += fse->pitch[2]; - } - - out += 12; - } -} - - - -static void shader1_run_linear( const struct fetch_shade_emit *fse, - unsigned start, - unsigned count, - char *buffer ) -{ - unsigned i; - const float *m = (const float *)fse->constants; - const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; - const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; - float *out = (float *)buffer; - -// debug_printf("rgb %f %f %f\n", rgb[0], rgb[1], rgb[2]); - - - for (i = 0; i < count; i++) { - { - const float *in = (const float *)xyz; - const float ix = in[0], iy = in[1], iz = in[2]; - - out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; - out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; - out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; - out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; - xyz += fse->pitch[0]; - } - - { - const float *in = (const float *)rgb; - out[4] = in[0]; - out[5] = in[1]; - out[6] = in[2]; - out[7] = 1.0f; - rgb += fse->pitch[1]; - } - - out += 8; - } -} - - - - -static void shader2_run_linear( const struct fetch_shade_emit *fse, - unsigned start, - unsigned count, - char *buffer ) -{ - unsigned i; - const float *m = (const float *)fse->constants; - const ubyte *xyz = fse->src[0] + start * fse->pitch[0]; - const ubyte *rgb = fse->src[1] + start * fse->pitch[1]; - const float psiz = 1.0; - float *out = (float *)buffer; - - - assert(fse->pitch[1] == 0); - - for (i = 0; i < count; i++) { - { - const float *in = (const float *)xyz; - const float ix = in[0], iy = in[1], iz = in[2]; - - out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; - out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; - out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; - out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; - xyz += fse->pitch[0]; - } - - { - const float *in = (const float *)rgb; - out[4] = in[0]; - out[5] = in[1]; - out[6] = in[2]; - out[7] = 1.0f; - rgb += fse->pitch[1]; - } - - { - out[8] = psiz; - } - - out += 9; - } -} - - - - -static void shader0_run_elts( const struct fetch_shade_emit *fse, - const unsigned *elts, - unsigned count, - char *buffer ) -{ - unsigned i; - const float *m = fse->constants; - float *out = (float *)buffer; - - - /* loop over vertex attributes (vertex shader inputs) - */ - for (i = 0; i < count; i++) { - unsigned elt = elts[i]; - { - const ubyte *xyz = fse->src[0] + elt * fse->pitch[0]; - const float *in = (const float *)xyz; - const float ix = in[0], iy = in[1], iz = in[2]; - - out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; - out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; - out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; - out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; - } - - { - const ubyte *rgb = fse->src[1] + elt * fse->pitch[1]; - const float *in = (const float *)rgb; - out[4] = in[0]; - out[5] = in[1]; - out[6] = in[2]; - out[7] = 1.0f; - } - - { - const ubyte *st = fse->src[2] + elt * fse->pitch[2]; - const float *in = (const float *)st; - out[8] = in[0]; - out[9] = in[1]; - out[10] = 0.0f; - out[11] = 1.0f; - } - - out += 12; - } -} - - - -static void shader1_run_elts( const struct fetch_shade_emit *fse, - const unsigned *elts, - unsigned count, - char *buffer ) -{ - unsigned i; - const float *m = (const float *)fse->constants; - float *out = (float *)buffer; - - for (i = 0; i < count; i++) { - unsigned elt = elts[i]; - - { - const ubyte *xyz = fse->src[0] + elt * fse->pitch[0]; - const float *in = (const float *)xyz; - const float ix = in[0], iy = in[1], iz = in[2]; - - out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; - out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; - out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; - out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; - xyz += fse->pitch[0]; - } - - { - const ubyte *rgb = fse->src[1] + elt * fse->pitch[1]; - const float *in = (const float *)rgb; - out[4] = in[0]; - out[5] = in[1]; - out[6] = in[2]; - out[7] = 1.0f; - rgb += fse->pitch[1]; - } - - out += 8; - } -} - - - - -static void shader2_run_elts( const struct fetch_shade_emit *fse, - const unsigned *elts, - unsigned count, - char *buffer ) -{ - unsigned i; - const float *m = (const float *)fse->constants; - const float psiz = 1.0; - float *out = (float *)buffer; - - for (i = 0; i < count; i++) { - unsigned elt = elts[i]; - { - const ubyte *xyz = fse->src[0] + elt * fse->pitch[0]; - const float *in = (const float *)xyz; - const float ix = in[0], iy = in[1], iz = in[2]; - - out[0] = m[0] * ix + m[4] * iy + m[8] * iz + m[12]; - out[1] = m[1] * ix + m[5] * iy + m[9] * iz + m[13]; - out[2] = m[2] * ix + m[6] * iy + m[10] * iz + m[14]; - out[3] = m[3] * ix + m[7] * iy + m[11] * iz + m[15]; - } - - { - const ubyte *rgb = fse->src[1] + elt * fse->pitch[1]; - out[4] = rgb[0]; - out[5] = rgb[1]; - out[6] = rgb[2]; - out[7] = 1.0f; - } - - { - out[8] = psiz; - } - - out += 9; - } -} - - - -static void fse_prepare( struct draw_pt_middle_end *middle, - unsigned prim, - unsigned opt ) -{ - struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; - struct draw_context *draw = fse->draw; - unsigned num_vs_inputs = draw->vs.vertex_shader->info.num_inputs; - unsigned num_vs_outputs = draw->vs.vertex_shader->info.num_outputs; - const struct vertex_info *vinfo; - unsigned i; - boolean need_psize = 0; - - - if (draw->pt.user.elts) { - assert(0); - return ; - } - - if (!draw->render->set_primitive( draw->render, - prim )) { - assert(0); - return; - } - - /* Must do this after set_primitive() above: - */ - vinfo = draw->render->get_vertex_info(draw->render); - - - - fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ - num_vs_inputs); /* inputs - fetch from api format */ - - fse->key.output_stride = vinfo->size * 4; - memset(fse->key.element, 0, - fse->key.nr_elements * sizeof(fse->key.element[0])); - - for (i = 0; i < num_vs_inputs; i++) { - const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; - fse->key.element[i].input_format = src->src_format; - - /* Consider ignoring these at this point, ie make generated - * programs independent of this state: - */ - fse->key.element[i].input_buffer = 0; //src->vertex_buffer_index; - fse->key.element[i].input_offset = 0; //src->src_offset; - } - - - { - unsigned dst_offset = 0; - - for (i = 0; i < vinfo->num_attribs; i++) { - unsigned emit_sz = 0; - unsigned output_format = PIPE_FORMAT_NONE; - unsigned vs_output = vinfo->src_index[i]; - - switch (vinfo->emit[i]) { - case EMIT_4F: - output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - emit_sz = 4 * sizeof(float); - break; - case EMIT_3F: - output_format = PIPE_FORMAT_R32G32B32_FLOAT; - emit_sz = 3 * sizeof(float); - break; - case EMIT_2F: - output_format = PIPE_FORMAT_R32G32_FLOAT; - emit_sz = 2 * sizeof(float); - break; - case EMIT_1F: - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - break; - case EMIT_1F_PSIZE: - need_psize = 1; - output_format = PIPE_FORMAT_R32_FLOAT; - emit_sz = 1 * sizeof(float); - vs_output = num_vs_outputs + 1; - - break; - default: - assert(0); - break; - } - - /* The elements in the key correspond to vertex shader output - * numbers, not to positions in the hw vertex description -- - * that's handled by the output_offset field. - */ - fse->key.element[vs_output].output_format = output_format; - fse->key.element[vs_output].output_offset = dst_offset; - - dst_offset += emit_sz; - assert(fse->key.output_stride >= dst_offset); - } - } - - /* To make psize work, really need to tell the vertex shader to - * copy that value from input->output. For 'translate' this was - * implicit for all elements. - */ -#if 0 - if (need_psize) { - unsigned input = num_vs_inputs + 1; - const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; - fse->key.element[i].input_format = PIPE_FORMAT_R32_FLOAT; - fse->key.element[i].input_buffer = 0; //nr_buffers + 1; - fse->key.element[i].input_offset = 0; - - fse->key.nr_elements += 1; - - } -#endif - - fse->constants = draw->pt.user.constants; - - /* Would normally look up a vertex shader and peruse its list of - * varients somehow. We omitted that step and put all the - * hardcoded "shaders" into an array. We're just making the - * assumption that this happens to be a matching shader... ie - * you're running isosurf, aren't you? - */ - fse->active = NULL; - for (i = 0; i < fse->nr_shaders; i++) { - if (translate_key_compare( &fse->key, &fse->shader[i].key) == 0) - fse->active = &fse->shader[i]; - } - - if (!fse->active) { - assert(0); - return ; - } - - /* Now set buffer pointers: - */ - for (i = 0; i < num_vs_inputs; i++) { - unsigned buf = draw->pt.vertex_element[i].vertex_buffer_index; - - fse->src[i] = ((const ubyte *) draw->pt.user.vbuffer[buf] + - draw->pt.vertex_buffer[buf].buffer_offset + - draw->pt.vertex_element[i].src_offset); - - fse->pitch[i] = draw->pt.vertex_buffer[buf].pitch; - - } - - - //return TRUE; -} - - - - - - - -static void fse_run_linear( struct draw_pt_middle_end *middle, - unsigned start, - unsigned count ) -{ - struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; - struct draw_context *draw = fse->draw; - - char *hw_verts; - - /* XXX: need to flush to get prim_vbuf.c to release its allocation?? - */ - draw_do_flush( draw, DRAW_FLUSH_BACKEND ); - - hw_verts = draw->render->allocate_vertices( draw->render, - (ushort)fse->key.output_stride, - (ushort)count ); - - if (!hw_verts) { - assert(0); - return; - } - - /* Single routine to fetch vertices, run shader and emit HW verts. - * Clipping and viewport transformation are done elsewhere -- - * either by the API or on hardware, or for some other reason not - * required... - */ - fse->active->run_linear( fse, - start, count, - hw_verts ); - - /* Draw arrays path to avoid re-emitting index list again and - * again. - */ - draw->render->draw_arrays( draw->render, - 0, - count ); - - - draw->render->release_vertices( draw->render, - hw_verts, - fse->key.output_stride, - count ); -} - - -static void -fse_run(struct draw_pt_middle_end *middle, - const unsigned *fetch_elts, - unsigned fetch_count, - const ushort *draw_elts, - unsigned draw_count ) -{ - struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; - struct draw_context *draw = fse->draw; - void *hw_verts; - - /* XXX: need to flush to get prim_vbuf.c to release its allocation?? - */ - draw_do_flush( draw, DRAW_FLUSH_BACKEND ); - - hw_verts = draw->render->allocate_vertices( draw->render, - (ushort)fse->key.output_stride, - (ushort)fetch_count ); - if (!hw_verts) { - assert(0); - return; - } - - - /* Single routine to fetch vertices, run shader and emit HW verts. - */ - fse->active->run_elts( fse, - fetch_elts, - fetch_count, - hw_verts ); - - draw->render->draw( draw->render, - draw_elts, - draw_count ); - - draw->render->release_vertices( draw->render, - hw_verts, - fse->key.output_stride, - fetch_count ); - -} - - -static void fse_finish( struct draw_pt_middle_end *middle ) -{ -} - - -static void -fse_destroy( struct draw_pt_middle_end *middle ) -{ - FREE(middle); -} - -struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw ) -{ - struct fetch_shade_emit *fse = CALLOC_STRUCT(fetch_shade_emit); - if (!fse) - return NULL; - - fse->base.prepare = fse_prepare; - fse->base.run = fse_run; - fse->base.run_linear = fse_run_linear; - fse->base.finish = fse_finish; - fse->base.destroy = fse_destroy; - fse->draw = draw; - - fse->shader[0].run_linear = shader0_run_linear; - fse->shader[0].run_elts = shader0_run_elts; - fse->shader[0].key.nr_elements = 3; - fse->shader[0].key.output_stride = 12 * sizeof(float); - - fse->shader[0].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[0].key.element[0].input_buffer = 0; - fse->shader[0].key.element[0].input_offset = 0; - fse->shader[0].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[0].key.element[0].output_offset = 0; - - fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[0].key.element[1].input_buffer = 0; - fse->shader[0].key.element[1].input_offset = 0; - fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[0].key.element[1].output_offset = 16; - - fse->shader[0].key.element[1].input_format = PIPE_FORMAT_R32G32_FLOAT; - fse->shader[0].key.element[1].input_buffer = 0; - fse->shader[0].key.element[1].input_offset = 0; - fse->shader[0].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[0].key.element[1].output_offset = 32; - - fse->shader[1].run_linear = shader1_run_linear; - fse->shader[1].run_elts = shader1_run_elts; - fse->shader[1].key.nr_elements = 2; - fse->shader[1].key.output_stride = 8 * sizeof(float); - - fse->shader[1].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[1].key.element[0].input_buffer = 0; - fse->shader[1].key.element[0].input_offset = 0; - fse->shader[1].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[1].key.element[0].output_offset = 0; - - fse->shader[1].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[1].key.element[1].input_buffer = 0; - fse->shader[1].key.element[1].input_offset = 0; - fse->shader[1].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[1].key.element[1].output_offset = 16; - - fse->shader[2].run_linear = shader2_run_linear; - fse->shader[2].run_elts = shader2_run_elts; - fse->shader[2].key.nr_elements = 3; - fse->shader[2].key.output_stride = 9 * sizeof(float); - - fse->shader[2].key.element[0].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[2].key.element[0].input_buffer = 0; - fse->shader[2].key.element[0].input_offset = 0; - fse->shader[2].key.element[0].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[2].key.element[0].output_offset = 0; - - fse->shader[2].key.element[1].input_format = PIPE_FORMAT_R32G32B32_FLOAT; - fse->shader[2].key.element[1].input_buffer = 0; - fse->shader[2].key.element[1].input_offset = 0; - fse->shader[2].key.element[1].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fse->shader[2].key.element[1].output_offset = 16; - - /* psize is special - * -- effectively add it here as another input!?! - * -- who knows how to add it as a buffer? - */ - fse->shader[2].key.element[2].input_format = PIPE_FORMAT_R32_FLOAT; - fse->shader[2].key.element[2].input_buffer = 0; - fse->shader[2].key.element[2].input_offset = 0; - fse->shader[2].key.element[2].output_format = PIPE_FORMAT_R32_FLOAT; - fse->shader[2].key.element[2].output_offset = 32; - - fse->nr_shaders = 3; - - return &fse->base; -} -- cgit v1.2.3 From f8762ba5234fd1b44e11e76bb5f58d2305c90572 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 14:42:15 +0100 Subject: draw: explicitly list nr_inputs, outputs in varient key --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c | 14 +++++++++----- src/gallium/auxiliary/draw/draw_vs.h | 6 ++++-- src/gallium/auxiliary/draw/draw_vs_aos.c | 2 +- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 4 ++-- src/gallium/auxiliary/draw/draw_vs_varient.c | 10 +++++----- 5 files changed, 21 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index c6249b4b41..581026dcb0 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -72,7 +72,6 @@ static void fse_prepare( struct draw_pt_middle_end *middle, struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; struct draw_context *draw = fse->draw; unsigned num_vs_inputs = draw->vs.vertex_shader->info.num_inputs; - unsigned num_vs_outputs = draw->vs.vertex_shader->info.num_outputs; const struct vertex_info *vinfo; unsigned i; boolean need_psize = 0; @@ -91,8 +90,11 @@ static void fse_prepare( struct draw_pt_middle_end *middle, fse->key.output_stride = vinfo->size * 4; - fse->key.nr_elements = MAX2(num_vs_outputs, /* outputs - translate to hw format */ - num_vs_inputs); /* inputs - fetch from api format */ + fse->key.nr_outputs = vinfo->num_attribs; + fse->key.nr_inputs = num_vs_inputs; + + fse->key.nr_elements = MAX2(fse->key.nr_outputs, /* outputs - translate to hw format */ + fse->key.nr_inputs); /* inputs - fetch from api format */ fse->key.viewport = !draw->identity_viewport; fse->key.clip = !draw->bypass_clipping; @@ -142,7 +144,7 @@ static void fse_prepare( struct draw_pt_middle_end *middle, need_psize = 1; output_format = PIPE_FORMAT_R32_FLOAT; emit_sz = 1 * sizeof(float); - vs_output = num_vs_outputs + 1; + vs_output = vinfo->num_attribs + 1; break; case EMIT_4UB: output_format = PIPE_FORMAT_B8G8R8A8_UNORM; @@ -177,7 +179,9 @@ static void fse_prepare( struct draw_pt_middle_end *middle, fse->key.element[i].input_buffer = 0; //nr_buffers + 1; fse->key.element[i].input_offset = 0; - fse->key.nr_elements += 1; + fse->key.nr_inputs += 1; + fse->key.nr_elements = MAX2(fse->key.nr_inputs, + fse->key.nr_outputs); } #endif diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index ff3e19b2a8..17902ab86a 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -58,10 +58,12 @@ struct draw_vs_element { struct draw_vs_varient_key { unsigned output_stride; - unsigned nr_elements:16; + unsigned nr_elements:8; /* max2(nr_inputs, nr_outputs) */ + unsigned nr_inputs:8; + unsigned nr_outputs:8; unsigned viewport:1; unsigned clip:1; - unsigned pad:14; + unsigned pad:5; struct draw_vs_element element[PIPE_MAX_ATTRIBS]; }; diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index c63553216c..e2e96470f7 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -2095,7 +2095,7 @@ static void vaos_set_buffer( struct draw_vs_varient *varient, struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; unsigned i; - for (i = 0; i < vaos->base.vs->info.num_inputs; i++) { + for (i = 0; i < vaos->base.key.nr_inputs; i++) { if (vaos->base.key.element[i].in.buffer == buf) { vaos->machine->attrib[i].input_ptr = ((char *)ptr + vaos->base.key.element[i].in.offset); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index f39ebb7a17..ef265d61cf 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -180,7 +180,7 @@ boolean aos_fetch_inputs( struct aos_compilation *cp, boolean linear ) { unsigned i; - for (i = 0; i < cp->vaos->base.vs->info.num_inputs; i++) { + for (i = 0; i < cp->vaos->base.key.nr_inputs; i++) { if (!load_input( cp, i, linear )) return FALSE; cp->insn_counter++; @@ -282,7 +282,7 @@ boolean aos_emit_outputs( struct aos_compilation *cp ) { unsigned i; - for (i = 0; i < cp->vaos->base.vs->info.num_outputs; i++) { + for (i = 0; i < cp->vaos->base.key.nr_outputs; i++) { unsigned format = cp->vaos->base.key.element[i].out.format; unsigned offset = cp->vaos->base.key.element[i].out.offset; diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index d4deabfff3..dab46e8eed 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -276,11 +276,11 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, - /* OK, have to build a new one: + /* Build free-standing fetch and emit functions: */ - fetch.nr_elements = vs->info.num_inputs; + fetch.nr_elements = key->nr_inputs; fetch.output_stride = 0; - for (i = 0; i < vs->info.num_inputs; i++) { + for (i = 0; i < key->nr_inputs; i++) { fetch.element[i].input_format = key->element[i].in.format; fetch.element[i].input_buffer = key->element[i].in.buffer; fetch.element[i].input_offset = key->element[i].in.offset; @@ -290,9 +290,9 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, } - emit.nr_elements = vs->info.num_outputs; + emit.nr_elements = key->nr_outputs; emit.output_stride = key->output_stride; - for (i = 0; i < vs->info.num_outputs; i++) { + for (i = 0; i < key->nr_outputs; i++) { emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; emit.element[i].input_buffer = 0; emit.element[i].input_offset = i * 4 * sizeof(float); -- cgit v1.2.3 From 44a7bd0019b9af9ff01336df0aa6eb206f5dc2e9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Tue, 27 May 2008 17:00:59 +0100 Subject: gallium: Always unreference bitmap cache texture before resetting the cache. Fixes assertion failure with tunnel2 demo. --- src/mesa/state_tracker/st_cb_bitmap.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 3a102df67f..593938f8cf 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -567,8 +567,9 @@ void st_flush_bitmap_cache(struct st_context *st) { if (!st->bitmap.cache->empty) { + struct bitmap_cache *cache = st->bitmap.cache; + if (st->ctx->DrawBuffer) { - struct bitmap_cache *cache = st->bitmap.cache; struct pipe_context *pipe = st->pipe; struct pipe_screen *screen = pipe->screen; @@ -592,10 +593,11 @@ st_flush_bitmap_cache(struct st_context *st) st->ctx->Current.RasterPos[2], BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, cache->texture); - - /* release/free the texture */ - pipe_texture_reference(&cache->texture, NULL); } + + /* release/free the texture */ + pipe_texture_reference(&cache->texture, NULL); + reset_cache(st); } } -- cgit v1.2.3 From 26f34dcff2a6ad0d44605c8ff34ae2850f655fd9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 28 May 2008 01:24:26 +0900 Subject: gallium: Autodetect WINCE. --- src/gallium/include/pipe/p_config.h | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h index 5c030bdfff..6ba211a1fc 100644 --- a/src/gallium/include/pipe/p_config.h +++ b/src/gallium/include/pipe/p_config.h @@ -100,6 +100,7 @@ #endif #if defined(PIPE_OS_WINDOWS) +#ifndef _WIN32_WCE #if !defined(PIPE_SUBSYSTEM_USER) && !defined(PIPE_SUBSYSTEM_KERNEL) #error Neither PIPE_SUBSYSTEM_USER or PIPE_SUBSYSTEM_KERNEL defined. #endif @@ -109,12 +110,12 @@ #if 0 /* FIXME */ #define PIPE_SUBSYSTEM_WINDOWS_MINIPORT #endif -#if 0 /* FIXME */ -#define PIPE_SUBSYSTEM_WINDOWS_CE -#endif #if defined(PIPE_SUBSYSTEM_USER) #define PIPE_SUBSYSTEM_WINDOWS_USER #endif +#else /* _WIN32_WCE */ +#define PIPE_SUBSYSTEM_WINDOWS_CE +#endif /* _WIN32_WCE */ #endif /* PIPE_OS_WINDOWS */ -- cgit v1.2.3 From 2ec419d40dba43305c28fca9658ea00541f67821 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 17:45:54 +0100 Subject: draw: fix ABS aliasing bug --- src/gallium/auxiliary/draw/draw_vs_aos.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index e2e96470f7..1c63677e6e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -964,12 +964,13 @@ static boolean emit_ABS( struct aos_compilation *cp, const struct tgsi_full_inst { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); - struct x86_reg dst = get_xmm_writable(cp, arg0); + struct x86_reg tmp = aos_get_xmm_reg(cp); - sse_mulps(cp->func, dst, neg); - sse_maxps(cp->func, dst, arg0); + sse_movaps(cp->func, tmp, arg0); + sse_mulps(cp->func, tmp, neg); + sse_maxps(cp->func, tmp, arg0); - store_dest(cp, &op->FullDstRegisters[0], dst); + store_dest(cp, &op->FullDstRegisters[0], tmp); return TRUE; } -- cgit v1.2.3 From 660fee8351542dadc0d5550164e753f7c2d67261 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 17:49:13 +0100 Subject: draw: ensure vs outputs mapped correctly to vinfo attribs --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c | 1 + src/gallium/auxiliary/draw/draw_vs.h | 19 ++++++++++--------- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 3 ++- 3 files changed, 13 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 581026dcb0..43d7095f76 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -160,6 +160,7 @@ static void fse_prepare( struct draw_pt_middle_end *middle, * that's handled by the output_offset field. */ fse->key.element[vs_output].out.format = output_format; + fse->key.element[vs_output].out.vs_output = vs_output; fse->key.element[vs_output].out.offset = dst_offset; dst_offset += emit_sz; diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 17902ab86a..01171bc23d 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -38,22 +38,23 @@ struct draw_context; struct pipe_shader_state; -struct draw_vs_input +struct draw_varient_input { enum pipe_format format; unsigned buffer; unsigned offset; }; -struct draw_vs_output +struct draw_varient_output { - enum pipe_format format; - unsigned offset; + enum pipe_format format; /* output format */ + unsigned vs_output:8; /* which vertex shader output is this? */ + unsigned offset:24; /* offset into output vertex */ }; -struct draw_vs_element { - struct draw_vs_input in; - struct draw_vs_output out; +struct draw_varient_element { + struct draw_varient_input in; + struct draw_varient_output out; }; struct draw_vs_varient_key { @@ -64,7 +65,7 @@ struct draw_vs_varient_key { unsigned viewport:1; unsigned clip:1; unsigned pad:5; - struct draw_vs_element element[PIPE_MAX_ATTRIBS]; + struct draw_varient_element element[PIPE_MAX_ATTRIBS]; }; struct draw_vs_varient; @@ -201,7 +202,7 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, static INLINE int draw_vs_varient_keysize( const struct draw_vs_varient_key *key ) { - return 2 * sizeof(int) + key->nr_elements * sizeof(struct draw_vs_element); + return 2 * sizeof(int) + key->nr_elements * sizeof(struct draw_varient_element); } static INLINE int draw_vs_varient_key_compare( const struct draw_vs_varient_key *a, diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index ef265d61cf..cebfaf6474 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -285,10 +285,11 @@ boolean aos_emit_outputs( struct aos_compilation *cp ) for (i = 0; i < cp->vaos->base.key.nr_outputs; i++) { unsigned format = cp->vaos->base.key.element[i].out.format; unsigned offset = cp->vaos->base.key.element[i].out.offset; + unsigned vs_output = cp->vaos->base.key.element[i].out.vs_output; struct x86_reg data = aos_get_shader_reg( cp, TGSI_FILE_OUTPUT, - i ); + vs_output ); if (data.file != file_XMM) { struct x86_reg tmp = aos_get_xmm_reg( cp ); -- cgit v1.2.3 From fd20d1c7e8bbe2f40d73679b1514023772cfd8f6 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 17:51:48 +0100 Subject: draw: add disabled debug code --- src/gallium/auxiliary/draw/draw_pt_emit.c | 15 +++++++++++- .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 28 +++++++++++++++++++++- 2 files changed, 41 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index e37256b1ff..cf87cde996 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -41,6 +41,8 @@ struct pt_emit { struct translate_cache *cache; unsigned prim; + + const struct vertex_info *vinfo; }; void draw_pt_emit_prepare( struct pt_emit *emit, @@ -71,7 +73,7 @@ void draw_pt_emit_prepare( struct pt_emit *emit, /* Must do this after set_primitive() above: */ - vinfo = draw->render->get_vertex_info(draw->render); + emit->vinfo = vinfo = draw->render->get_vertex_info(draw->render); /* Translate from pipeline vertices to hw vertices. @@ -245,6 +247,17 @@ void draw_pt_emit_linear(struct pt_emit *emit, vertex_count, hw_verts); + if (0) { + unsigned i; + for (i = 0; i < vertex_count; i++) { + debug_printf("\n\n%s vertex %d:\n", __FUNCTION__, i); + draw_dump_emitted_vertex( emit->vinfo, + (const uint8_t *)hw_verts + + translate->key.output_stride * i ); + } + } + + render->draw_arrays(render, start, count); render->release_vertices(render, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 43d7095f76..85d0bdfcab 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -60,6 +60,9 @@ struct fetch_shade_emit { struct draw_vs_varient_key key; struct draw_vs_varient *active; + + + const struct vertex_info *vinfo; }; @@ -85,7 +88,7 @@ static void fse_prepare( struct draw_pt_middle_end *middle, /* Must do this after set_primitive() above: */ - vinfo = draw->render->get_vertex_info(draw->render); + fse->vinfo = vinfo = draw->render->get_vertex_info(draw->render); @@ -267,6 +270,18 @@ static void fse_run_linear( struct draw_pt_middle_end *middle, 0, count ); + if (0) { + unsigned i; + for (i = 0; i < count; i++) { + debug_printf("\n\n%s vertex %d: (stride %d, offset %d)\n", __FUNCTION__, i, + fse->key.output_stride, + fse->key.output_stride * i); + + draw_dump_emitted_vertex( fse->vinfo, + (const uint8_t *)hw_verts + fse->key.output_stride * i ); + } + } + draw->render->release_vertices( draw->render, hw_verts, @@ -311,6 +326,17 @@ fse_run(struct draw_pt_middle_end *middle, draw_elts, draw_count ); + if (0) { + unsigned i; + for (i = 0; i < fetch_count; i++) { + debug_printf("\n\n%s vertex %d:\n", __FUNCTION__, i); + draw_dump_emitted_vertex( fse->vinfo, + (const uint8_t *)hw_verts + + fse->key.output_stride * i ); + } + } + + draw->render->release_vertices( draw->render, hw_verts, fse->key.output_stride, -- cgit v1.2.3 From 7b85ea19de09d4e7e077ca147528e90e52683690 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 19:01:57 +0100 Subject: draw: support psize in vs_varient paths Preserve the vinfo "EMIT_*" format descriptors in the varient key, and deal with PSIZE directly in each implementation. --- .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 35 ++------------------ src/gallium/auxiliary/draw/draw_vertex.h | 21 ++++++++++++ src/gallium/auxiliary/draw/draw_vs_aos.c | 3 ++ src/gallium/auxiliary/draw/draw_vs_aos.h | 1 + src/gallium/auxiliary/draw/draw_vs_aos_io.c | 37 ++++++++++++++-------- src/gallium/auxiliary/draw/draw_vs_varient.c | 24 +++++++++++--- 6 files changed, 71 insertions(+), 50 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 85d0bdfcab..729c7db999 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -77,7 +77,6 @@ static void fse_prepare( struct draw_pt_middle_end *middle, unsigned num_vs_inputs = draw->vs.vertex_shader->info.num_inputs; const struct vertex_info *vinfo; unsigned i; - boolean need_psize = 0; if (!draw->render->set_primitive( draw->render, @@ -123,34 +122,24 @@ static void fse_prepare( struct draw_pt_middle_end *middle, for (i = 0; i < vinfo->num_attribs; i++) { unsigned emit_sz = 0; - unsigned output_format = PIPE_FORMAT_NONE; - unsigned vs_output = vinfo->src_index[i]; switch (vinfo->emit[i]) { case EMIT_4F: - output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; emit_sz = 4 * sizeof(float); break; case EMIT_3F: - output_format = PIPE_FORMAT_R32G32B32_FLOAT; emit_sz = 3 * sizeof(float); break; case EMIT_2F: - output_format = PIPE_FORMAT_R32G32_FLOAT; emit_sz = 2 * sizeof(float); break; case EMIT_1F: - output_format = PIPE_FORMAT_R32_FLOAT; emit_sz = 1 * sizeof(float); break; case EMIT_1F_PSIZE: - need_psize = 1; - output_format = PIPE_FORMAT_R32_FLOAT; emit_sz = 1 * sizeof(float); - vs_output = vinfo->num_attribs + 1; break; case EMIT_4UB: - output_format = PIPE_FORMAT_B8G8R8A8_UNORM; emit_sz = 4 * sizeof(ubyte); break; default: @@ -162,33 +151,15 @@ static void fse_prepare( struct draw_pt_middle_end *middle, * numbers, not to positions in the hw vertex description -- * that's handled by the output_offset field. */ - fse->key.element[vs_output].out.format = output_format; - fse->key.element[vs_output].out.vs_output = vs_output; - fse->key.element[vs_output].out.offset = dst_offset; + fse->key.element[i].out.format = vinfo->emit[i]; + fse->key.element[i].out.vs_output = vinfo->src_index[i]; + fse->key.element[i].out.offset = dst_offset; dst_offset += emit_sz; assert(fse->key.output_stride >= dst_offset); } } - /* To make psize work, really need to tell the vertex shader to - * copy that value from input->output. For 'translate' this was - * implicit for all elements. - */ -#if 0 - if (need_psize) { - unsigned input = num_vs_inputs + 1; - const struct pipe_vertex_element *src = &draw->pt.vertex_element[i]; - fse->key.element[i].input_format = PIPE_FORMAT_R32_FLOAT; - fse->key.element[i].input_buffer = 0; //nr_buffers + 1; - fse->key.element[i].input_offset = 0; - - fse->key.nr_inputs += 1; - fse->key.nr_elements = MAX2(fse->key.nr_inputs, - fse->key.nr_outputs); - - } -#endif /* Would normally look up a vertex shader and peruse its list of * varients somehow. We omitted that step and put all the diff --git a/src/gallium/auxiliary/draw/draw_vertex.h b/src/gallium/auxiliary/draw/draw_vertex.h index 6d8bac5138..16c65c4317 100644 --- a/src/gallium/auxiliary/draw/draw_vertex.h +++ b/src/gallium/auxiliary/draw/draw_vertex.h @@ -109,4 +109,25 @@ extern void draw_compute_vertex_size(struct vertex_info *vinfo); void draw_dump_emitted_vertex(const struct vertex_info *vinfo, const uint8_t *data); + +static INLINE unsigned draw_translate_vinfo_format(unsigned format ) +{ + switch (format) { + case EMIT_1F: + case EMIT_1F_PSIZE: + return PIPE_FORMAT_R32_FLOAT; + case EMIT_2F: + return PIPE_FORMAT_R32G32_FLOAT; + case EMIT_3F: + return PIPE_FORMAT_R32G32B32_FLOAT; + case EMIT_4F: + return PIPE_FORMAT_R32G32B32A32_FLOAT; + case EMIT_4UB: + return PIPE_FORMAT_R8G8B8A8_UNORM; + default: + return PIPE_FORMAT_NONE; + } +} + + #endif /* DRAW_VERTEX_H */ diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 1c63677e6e..d3770b2c53 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -2126,6 +2126,7 @@ static void vaos_run_elts( struct draw_vs_varient *varient, { struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + vaos->machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; vaos->gen_run_elts( varient, elts, count, @@ -2139,6 +2140,7 @@ static void vaos_run_linear( struct draw_vs_varient *varient, { struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + vaos->machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; vaos->gen_run_linear( varient, start, count, @@ -2204,6 +2206,7 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, vaos->base.run_linear = vaos_run_linear; vaos->base.run_elts = vaos_run_elts; + vaos->draw = vs->draw; vaos->machine = align_malloc( sizeof(struct aos_machine), 16 ); if (!vaos->machine) goto fail; diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index fffe2e4658..b47413ff43 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -176,6 +176,7 @@ boolean aos_emit_outputs( struct aos_compilation *cp ); #define IMM_255 4 /* 255, 255, 255, 255 */ #define IMM_NEGS 5 /* -1,-1,-1,-1 */ #define IMM_RSQ 6 /* -.5,1.5,_,_ */ +#define IMM_PSIZE 7 /* not really an immediate - updated each run */ struct x86_reg aos_get_internal( struct aos_compilation *cp, unsigned imm ); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index cebfaf6474..836110f382 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -33,6 +33,7 @@ #include "tgsi/exec/tgsi_exec.h" #include "draw_vs.h" #include "draw_vs_aos.h" +#include "draw_vertex.h" #include "rtasm/rtasm_x86sse.h" @@ -249,24 +250,27 @@ static boolean emit_output( struct aos_compilation *cp, unsigned format ) { switch (format) { - case PIPE_FORMAT_R32_FLOAT: + case EMIT_1F: + case EMIT_1F_PSIZE: emit_store_R32(cp, ptr, dataXMM); break; - case PIPE_FORMAT_R32G32_FLOAT: + case EMIT_2F: emit_store_R32G32(cp, ptr, dataXMM); break; - case PIPE_FORMAT_R32G32B32_FLOAT: + case EMIT_3F: emit_store_R32G32B32(cp, ptr, dataXMM); break; - case PIPE_FORMAT_R32G32B32A32_FLOAT: + case EMIT_4F: emit_store_R32G32B32A32(cp, ptr, dataXMM); break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - emit_swizzle(cp, dataXMM, dataXMM, SHUF(Z,Y,X,W)); - emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM); - break; - case PIPE_FORMAT_R8G8B8A8_UNORM: - emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM); + case EMIT_4UB: + if (1) { + emit_swizzle(cp, dataXMM, dataXMM, SHUF(Z,Y,X,W)); + emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM); + } + else { + emit_store_R8G8B8A8_UNORM(cp, ptr, dataXMM); + } break; default: ERROR(cp, "unhandled output format"); @@ -287,9 +291,16 @@ boolean aos_emit_outputs( struct aos_compilation *cp ) unsigned offset = cp->vaos->base.key.element[i].out.offset; unsigned vs_output = cp->vaos->base.key.element[i].out.vs_output; - struct x86_reg data = aos_get_shader_reg( cp, - TGSI_FILE_OUTPUT, - vs_output ); + struct x86_reg data; + + if (format == EMIT_1F_PSIZE) { + data = aos_get_internal_xmm( cp, IMM_PSIZE ); + } + else { + data = aos_get_shader_reg( cp, + TGSI_FILE_OUTPUT, + vs_output ); + } if (data.file != file_XMM) { struct x86_reg tmp = aos_get_xmm_reg( cp ); diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index dab46e8eed..119a3a04b5 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -231,6 +231,10 @@ static void vsvg_run_linear( struct draw_vs_varient *varient, output_buffer, vsvg->base.key.output_stride ); + vsvg->emit->set_buffer( vsvg->emit, + 1, + &vsvg->draw->rasterizer->point_size, + 0); vsvg->emit->run( vsvg->emit, 0, count, @@ -293,11 +297,21 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, emit.nr_elements = key->nr_outputs; emit.output_stride = key->output_stride; for (i = 0; i < key->nr_outputs; i++) { - emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - emit.element[i].input_buffer = 0; - emit.element[i].input_offset = i * 4 * sizeof(float); - emit.element[i].output_format = key->element[i].out.format; - emit.element[i].output_offset = key->element[i].out.offset; + if (key->element[i].out.format != EMIT_1F_PSIZE) + { + emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + emit.element[i].input_buffer = 0; + emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float); + emit.element[i].output_format = draw_translate_vinfo_format(key->element[i].out.format); + emit.element[i].output_offset = key->element[i].out.offset; + } + else { + emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT; + emit.element[i].input_buffer = 1; + emit.element[i].input_offset = 0; + emit.element[i].output_format = PIPE_FORMAT_R32_FLOAT; + emit.element[i].output_offset = key->element[i].out.offset; + } } vsvg->fetch = draw_vs_get_fetch( vs->draw, &fetch ); -- cgit v1.2.3 From 3e4128963d4e1a27ef29658a3aa260f492091628 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 13:45:16 -0600 Subject: pipe stderr to /dev/null --- src/egl/main/Makefile | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index e6844d4852..e1058a23f7 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -61,7 +61,8 @@ depend: $(SOURCES) $(HEADERS) @ rm -f depend @ touch depend $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) \ - $(SOURCES) $(HEADERS) > /dev/null + $(SOURCES) $(HEADERS) > /dev/null 2>/dev/null + include depend # DO NOT DELETE -- cgit v1.2.3 From 1ed1027e886980b9b0f48fa6bfcf3d6e209c7787 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 13:45:41 -0600 Subject: assorted changes to compile with new EGL 1.4 headers (untested) --- src/egl/main/eglapi.c | 66 +++++++++++++++++++++++----------------------- src/egl/main/eglconfig.c | 8 ++++-- src/egl/main/eglconfig.h | 20 +++++++++----- src/egl/main/eglcontext.c | 16 ++++++----- src/egl/main/egldisplay.c | 10 ++++--- src/egl/main/eglmode.h | 3 +++ src/egl/main/eglsurface.c | 12 +++++---- src/egl/main/egltypedefs.h | 2 ++ 8 files changed, 80 insertions(+), 57 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index bfa580e6c3..f9b3f5921a 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -47,7 +47,7 @@ * This will probably change! * See _eglChooseDriver() for details! */ -EGLDisplay APIENTRY +EGLDisplay EGLAPIENTRY eglGetDisplay(NativeDisplayType displayName) { _EGLDisplay *dpy; @@ -60,7 +60,7 @@ eglGetDisplay(NativeDisplayType displayName) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor) { if (dpy) { @@ -72,7 +72,7 @@ eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglTerminate(EGLDisplay dpy) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -83,7 +83,7 @@ eglTerminate(EGLDisplay dpy) } -const char * APIENTRY +const char * EGLAPIENTRY eglQueryString(EGLDisplay dpy, EGLint name) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -94,7 +94,7 @@ eglQueryString(EGLDisplay dpy, EGLint name) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglGetConfigs(EGLDisplay dpy, EGLConfig *configs, EGLint config_size, EGLint *num_config) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -103,7 +103,7 @@ eglGetConfigs(EGLDisplay dpy, EGLConfig *configs, EGLint config_size, EGLint *nu } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglChooseConfig(EGLDisplay dpy, const EGLint *attrib_list, EGLConfig *configs, EGLint config_size, EGLint *num_config) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -111,7 +111,7 @@ eglChooseConfig(EGLDisplay dpy, const EGLint *attrib_list, EGLConfig *configs, E } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglGetConfigAttrib(EGLDisplay dpy, EGLConfig config, EGLint attribute, EGLint *value) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -119,7 +119,7 @@ eglGetConfigAttrib(EGLDisplay dpy, EGLConfig config, EGLint attribute, EGLint *v } -EGLContext APIENTRY +EGLContext EGLAPIENTRY eglCreateContext(EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -127,7 +127,7 @@ eglCreateContext(EGLDisplay dpy, EGLConfig config, EGLContext share_list, const } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglDestroyContext(EGLDisplay dpy, EGLContext ctx) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -135,7 +135,7 @@ eglDestroyContext(EGLDisplay dpy, EGLContext ctx) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglMakeCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -143,7 +143,7 @@ eglMakeCurrent(EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext ctx) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglQueryContext(EGLDisplay dpy, EGLContext ctx, EGLint attribute, EGLint *value) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -151,7 +151,7 @@ eglQueryContext(EGLDisplay dpy, EGLContext ctx, EGLint attribute, EGLint *value) } -EGLSurface APIENTRY +EGLSurface EGLAPIENTRY eglCreateWindowSurface(EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -159,7 +159,7 @@ eglCreateWindowSurface(EGLDisplay dpy, EGLConfig config, NativeWindowType window } -EGLSurface APIENTRY +EGLSurface EGLAPIENTRY eglCreatePixmapSurface(EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -167,7 +167,7 @@ eglCreatePixmapSurface(EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap } -EGLSurface APIENTRY +EGLSurface EGLAPIENTRY eglCreatePbufferSurface(EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -175,7 +175,7 @@ eglCreatePbufferSurface(EGLDisplay dpy, EGLConfig config, const EGLint *attrib_l } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglDestroySurface(EGLDisplay dpy, EGLSurface surface) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -183,7 +183,7 @@ eglDestroySurface(EGLDisplay dpy, EGLSurface surface) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglQuerySurface(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint *value) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -191,7 +191,7 @@ eglQuerySurface(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint *va } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglSurfaceAttrib(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint value) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -199,7 +199,7 @@ eglSurfaceAttrib(EGLDisplay dpy, EGLSurface surface, EGLint attribute, EGLint va } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglBindTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -207,7 +207,7 @@ eglBindTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglReleaseTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -215,7 +215,7 @@ eglReleaseTexImage(EGLDisplay dpy, EGLSurface surface, EGLint buffer) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglSwapInterval(EGLDisplay dpy, EGLint interval) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -223,7 +223,7 @@ eglSwapInterval(EGLDisplay dpy, EGLint interval) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglSwapBuffers(EGLDisplay dpy, EGLSurface draw) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -231,7 +231,7 @@ eglSwapBuffers(EGLDisplay dpy, EGLSurface draw) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglCopyBuffers(EGLDisplay dpy, EGLSurface surface, NativePixmapType target) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -239,7 +239,7 @@ eglCopyBuffers(EGLDisplay dpy, EGLSurface surface, NativePixmapType target) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglWaitGL(void) { EGLDisplay dpy = eglGetCurrentDisplay(); @@ -252,7 +252,7 @@ eglWaitGL(void) } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglWaitNative(EGLint engine) { EGLDisplay dpy = eglGetCurrentDisplay(); @@ -265,7 +265,7 @@ eglWaitNative(EGLint engine) } -EGLDisplay APIENTRY +EGLDisplay EGLAPIENTRY eglGetCurrentDisplay(void) { _EGLDisplay *dpy = _eglGetCurrentDisplay(); @@ -276,7 +276,7 @@ eglGetCurrentDisplay(void) } -EGLContext APIENTRY +EGLContext EGLAPIENTRY eglGetCurrentContext(void) { _EGLContext *ctx = _eglGetCurrentContext(); @@ -287,7 +287,7 @@ eglGetCurrentContext(void) } -EGLSurface APIENTRY +EGLSurface EGLAPIENTRY eglGetCurrentSurface(EGLint readdraw) { _EGLSurface *s = _eglGetCurrentSurface(readdraw); @@ -298,7 +298,7 @@ eglGetCurrentSurface(EGLint readdraw) } -EGLint APIENTRY +EGLint EGLAPIENTRY eglGetError(void) { _EGLThreadInfo *t = _eglGetCurrentThread(); @@ -308,7 +308,7 @@ eglGetError(void) } -void (* APIENTRY eglGetProcAddress(const char *procname))() +void (* EGLAPIENTRY eglGetProcAddress(const char *procname))() { typedef void (*genericFunc)(); struct name_function { @@ -389,7 +389,7 @@ void (* APIENTRY eglGetProcAddress(const char *procname))() * EGL_MESA_screen extension */ -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglChooseModeMESA(EGLDisplay dpy, EGLScreenMESA screen, const EGLint *attrib_list, EGLModeMESA *modes, EGLint modes_size, EGLint *num_modes) @@ -402,7 +402,7 @@ eglChooseModeMESA(EGLDisplay dpy, EGLScreenMESA screen, } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglGetModesMESA(EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *modes, EGLint mode_size, EGLint *num_mode) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -413,7 +413,7 @@ eglGetModesMESA(EGLDisplay dpy, EGLScreenMESA screen, EGLModeMESA *modes, EGLint } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglGetModeAttribMESA(EGLDisplay dpy, EGLModeMESA mode, EGLint attribute, EGLint *value) { _EGLDriver *drv = _eglLookupDriver(dpy); @@ -424,7 +424,7 @@ eglGetModeAttribMESA(EGLDisplay dpy, EGLModeMESA mode, EGLint attribute, EGLint } -EGLBoolean APIENTRY +EGLBoolean EGLAPIENTRY eglCopyContextMESA(EGLDisplay dpy, EGLContext source, EGLContext dest, EGLint mask) { _EGLDriver *drv = _eglLookupDriver(dpy); diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c index c180e30d7f..a74cd23113 100644 --- a/src/egl/main/eglconfig.c +++ b/src/egl/main/eglconfig.c @@ -17,6 +17,7 @@ #define MIN2(A, B) (((A) < (B)) ? (A) : (B)) +#if 0 /** * Convert an _EGLConfig to a __GLcontextModes object. * NOTE: This routine may be incomplete - we're only making sure that @@ -57,6 +58,7 @@ _eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode) mode->visualType = GLX_TRUE_COLOR; mode->renderType = GLX_RGBA_BIT; } +#endif void @@ -76,7 +78,7 @@ void _eglInitConfig(_EGLConfig *config, EGLint id) { memset(config, 0, sizeof(*config)); - config->Handle = id; + config->Handle = (EGLConfig) id; _eglSetConfigAttrib(config, EGL_CONFIG_ID, id); _eglSetConfigAttrib(config, EGL_BIND_TO_TEXTURE_RGB, EGL_DONT_CARE); _eglSetConfigAttrib(config, EGL_BIND_TO_TEXTURE_RGBA, EGL_DONT_CARE); @@ -132,7 +134,7 @@ _eglAddConfig(_EGLDisplay *display, const _EGLConfig *config) if (newConfigs) { display->Configs = newConfigs; display->Configs[n] = *config; /* copy struct */ - display->Configs[n].Handle = n; + display->Configs[n].Handle = (EGLConfig) n; display->NumConfigs++; return display->Configs + n; } @@ -498,6 +500,7 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, * \c GL_UNSIGNED_3BYTE_8_8_8, \c GL_4FLOAT_32_32_32_32, * \c GL_4HALF_16_16_16_16, etc. We can cross that bridge when we come to it. */ +#if 0 GLboolean _eglFillInConfigs(_EGLConfig * configs, GLenum fb_format, GLenum fb_type, @@ -636,3 +639,4 @@ _eglFillInConfigs(_EGLConfig * configs, } return GL_TRUE; } +#endif diff --git a/src/egl/main/eglconfig.h b/src/egl/main/eglconfig.h index 1fb976e5b4..88e44dfc7d 100644 --- a/src/egl/main/eglconfig.h +++ b/src/egl/main/eglconfig.h @@ -3,7 +3,9 @@ #include "egltypedefs.h" +#if 0 #include "GL/internal/glcore.h" +#endif #define MAX_ATTRIBS 100 @@ -52,16 +54,20 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, EGLint confi extern void _eglSetConfigAttrib(_EGLConfig *config, EGLint attr, EGLint val); -extern GLboolean +#if 0 +extern EGLBoolean _eglFillInConfigs( _EGLConfig *configs, - GLenum fb_format, GLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, - unsigned num_depth_stencil_bits, - const GLenum * db_modes, unsigned num_db_modes, - int visType ); - + EGLenum fb_format, EGLenum fb_type, + const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + unsigned num_depth_stencil_bits, + const EGLenum * db_modes, unsigned num_db_modes, + int visType ); +#endif + +#if 0 extern void _eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode); +#endif #endif /* EGLCONFIG_INCLUDED */ diff --git a/src/egl/main/eglcontext.c b/src/egl/main/eglcontext.c index 374c006dae..bf26c1faa4 100644 --- a/src/egl/main/eglcontext.c +++ b/src/egl/main/eglcontext.c @@ -32,7 +32,7 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, /* no attribs defined for now */ default: _eglError(EGL_BAD_ATTRIBUTE, "eglCreateContext"); - return EGL_NO_CONTEXT; + return EGL_FALSE; } } @@ -53,9 +53,10 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, void _eglSaveContext(_EGLContext *ctx) { + EGLuint key = _eglHashGenKey(_eglGlobal.Contexts); assert(ctx); - ctx->Handle = _eglHashGenKey(_eglGlobal.Contexts); - _eglHashInsert(_eglGlobal.Contexts, ctx->Handle, ctx); + ctx->Handle = (EGLContext) key; + _eglHashInsert(_eglGlobal.Contexts, key, ctx); } @@ -65,7 +66,8 @@ _eglSaveContext(_EGLContext *ctx) void _eglRemoveContext(_EGLContext *ctx) { - _eglHashRemove(_eglGlobal.Contexts, ctx->Handle); + EGLuint key = (EGLuint) ctx->Handle; + _eglHashRemove(_eglGlobal.Contexts, key); } @@ -76,7 +78,8 @@ _eglRemoveContext(_EGLContext *ctx) _EGLContext * _eglLookupContext(EGLContext ctx) { - _EGLContext *c = (_EGLContext *) _eglHashLookup(_eglGlobal.Contexts, ctx); + EGLuint key = (EGLuint) ctx; + _EGLContext *c = (_EGLContext *) _eglHashLookup(_eglGlobal.Contexts, key); return c; } @@ -126,7 +129,8 @@ _eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx) { _EGLContext *context = _eglLookupContext(ctx); if (context) { - _eglHashRemove(_eglGlobal.Contexts, ctx); + EGLuint key = (EGLuint) ctx; + _eglHashRemove(_eglGlobal.Contexts, key); if (context->IsBound) { context->DeletePending = EGL_TRUE; } diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c index 074a85bf26..59d0bd3dc3 100644 --- a/src/egl/main/egldisplay.c +++ b/src/egl/main/egldisplay.c @@ -26,10 +26,11 @@ _eglNewDisplay(NativeDisplayType displayName) { _EGLDisplay *dpy = (_EGLDisplay *) calloc(1, sizeof(_EGLDisplay)); if (dpy) { - dpy->Handle = _eglHashGenKey(_eglGlobal.Displays); - _eglHashInsert(_eglGlobal.Displays, dpy->Handle, dpy); + EGLuint key = _eglHashGenKey(_eglGlobal.Displays); + dpy->Handle = (EGLDisplay) key; + _eglHashInsert(_eglGlobal.Displays, key, dpy); if (displayName) - dpy->Name = my_strdup(displayName); + dpy->Name = my_strdup((char *) displayName); else dpy->Name = NULL; dpy->Driver = NULL; /* this gets set later */ @@ -45,7 +46,8 @@ _eglNewDisplay(NativeDisplayType displayName) _EGLDisplay * _eglLookupDisplay(EGLDisplay dpy) { - _EGLDisplay *d = (_EGLDisplay *) _eglHashLookup(_eglGlobal.Displays, dpy); + EGLuint key = (EGLuint) dpy; + _EGLDisplay *d = (_EGLDisplay *) _eglHashLookup(_eglGlobal.Displays, key); return d; } diff --git a/src/egl/main/eglmode.h b/src/egl/main/eglmode.h index e70da85759..52d4875676 100644 --- a/src/egl/main/eglmode.h +++ b/src/egl/main/eglmode.h @@ -4,6 +4,9 @@ #include "egltypedefs.h" +#define EGL_NO_MODE_MESA 0 + + /** * Data structure which corresponds to an EGLModeMESA. */ diff --git a/src/egl/main/eglsurface.c b/src/egl/main/eglsurface.c index 874f318e96..cd5d9757f9 100644 --- a/src/egl/main/eglsurface.c +++ b/src/egl/main/eglsurface.c @@ -207,25 +207,27 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy, void _eglSaveSurface(_EGLSurface *surf) { + EGLuint key = _eglHashGenKey(_eglGlobal.Contexts); assert(surf); assert(!surf->Handle); - surf->Handle = _eglHashGenKey(_eglGlobal.Contexts); + surf->Handle = (EGLSurface) key; assert(surf->Handle); - _eglHashInsert(_eglGlobal.Surfaces, surf->Handle, surf); + _eglHashInsert(_eglGlobal.Surfaces, key, surf); } void _eglRemoveSurface(_EGLSurface *surf) { - _eglHashRemove(_eglGlobal.Surfaces, surf->Handle); + _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surf->Handle); } _EGLSurface * _eglLookupSurface(EGLSurface surf) { - _EGLSurface *c = (_EGLSurface *) _eglHashLookup(_eglGlobal.Surfaces, surf); + _EGLSurface *c = (_EGLSurface *) _eglHashLookup(_eglGlobal.Surfaces, + (EGLuint) surf); return c; } @@ -439,7 +441,7 @@ _eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) { _EGLSurface *surf = _eglLookupSurface(surface); if (surf) { - _eglHashRemove(_eglGlobal.Surfaces, surface); + _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface); if (surf->IsBound) { surf->DeletePending = EGL_TRUE; } diff --git a/src/egl/main/egltypedefs.h b/src/egl/main/egltypedefs.h index fa8cb496f8..53810a5a44 100644 --- a/src/egl/main/egltypedefs.h +++ b/src/egl/main/egltypedefs.h @@ -1,8 +1,10 @@ #ifndef EGLTYPEDEFS_INCLUDED #define EGLTYPEDEFS_INCLUDED +#define EGL_EGLEXT_PROTOTYPES #include +#include typedef struct _egl_api _EGLAPI; -- cgit v1.2.3 From fcce6c068cf53fd394a2c82080338860dae8777b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 13:47:44 -0600 Subject: s/GLuint/unsigned/ --- src/egl/drivers/demo/demo.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/demo/demo.c b/src/egl/drivers/demo/demo.c index 45545755c0..0a13de6d87 100644 --- a/src/egl/drivers/demo/demo.c +++ b/src/egl/drivers/demo/demo.c @@ -21,7 +21,7 @@ typedef struct demo_driver { _EGLDriver Base; /* base class/object */ - GLuint DemoStuff; + unsigned DemoStuff; } DemoDriver; #define DEMO_DRIVER(D) ((DemoDriver *) (D)) @@ -33,7 +33,7 @@ typedef struct demo_driver typedef struct demo_surface { _EGLSurface Base; /* base class/object */ - GLuint DemoStuff; + unsigned DemoStuff; } DemoSurface; @@ -43,7 +43,7 @@ typedef struct demo_surface typedef struct demo_context { _EGLContext Base; /* base class/object */ - GLuint DemoStuff; + unsigned DemoStuff; } DemoContext; -- cgit v1.2.3 From 209a557574b9833da3d6ac299c83f4cddfff6910 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 14:17:40 -0600 Subject: fix typo: s/Contexts/Surfaces/ --- src/egl/main/eglsurface.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/eglsurface.c b/src/egl/main/eglsurface.c index cd5d9757f9..134afa7cbd 100644 --- a/src/egl/main/eglsurface.c +++ b/src/egl/main/eglsurface.c @@ -207,7 +207,7 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy, void _eglSaveSurface(_EGLSurface *surf) { - EGLuint key = _eglHashGenKey(_eglGlobal.Contexts); + EGLuint key = _eglHashGenKey(_eglGlobal.Surfaces); assert(surf); assert(!surf->Handle); surf->Handle = (EGLSurface) key; -- cgit v1.2.3 From 5e7dba541298a29f175f9d077bf6f63030465d94 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 14:21:25 -0600 Subject: eliminate the context hash table In EGL 1.4 the opaque EGLContext type is a pointer so we can just cast between public EGLContext handles and private _EGLContext pointers. --- src/egl/main/eglapi.c | 5 +---- src/egl/main/eglcontext.c | 53 ++++++++++++++++++++++++++++------------------- src/egl/main/eglcontext.h | 6 ++++-- src/egl/main/eglglobals.c | 2 -- src/egl/main/eglglobals.h | 1 - 5 files changed, 37 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index f9b3f5921a..c57d29dda6 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -280,10 +280,7 @@ EGLContext EGLAPIENTRY eglGetCurrentContext(void) { _EGLContext *ctx = _eglGetCurrentContext(); - if (ctx) - return ctx->Handle; - else - return EGL_NO_CONTEXT; + return _eglGetContextHandle(ctx); } diff --git a/src/egl/main/eglcontext.c b/src/egl/main/eglcontext.c index bf26c1faa4..86d1d84334 100644 --- a/src/egl/main/eglcontext.c +++ b/src/egl/main/eglcontext.c @@ -6,12 +6,12 @@ #include "egldisplay.h" #include "egldriver.h" #include "eglglobals.h" -#include "eglhash.h" #include "eglsurface.h" /** - * Initialize the given _EGLContext object to defaults. + * Initialize the given _EGLContext object to defaults and/or the values + * in the attrib_list. */ EGLBoolean _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, @@ -23,15 +23,17 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, conf = _eglLookupConfig(drv, dpy, config); if (!conf) { - _eglError(EGL_BAD_CONFIG, "eglCreateContext"); + _eglError(EGL_BAD_CONFIG, "_eglInitContext"); return EGL_FALSE; } for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { switch (attrib_list[i]) { - /* no attribs defined for now */ + case EGL_CONTEXT_CLIENT_VERSION: + /* xxx todo */ + break; default: - _eglError(EGL_BAD_ATTRIBUTE, "eglCreateContext"); + _eglError(EGL_BAD_ATTRIBUTE, "_eglInitContext"); return EGL_FALSE; } } @@ -46,17 +48,15 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, } -/* - * Assign an EGLContext handle to the _EGLContext object then put it into - * the hash table. +/** + * Save a new _EGLContext into the hash table. */ void _eglSaveContext(_EGLContext *ctx) { - EGLuint key = _eglHashGenKey(_eglGlobal.Contexts); - assert(ctx); - ctx->Handle = (EGLContext) key; - _eglHashInsert(_eglGlobal.Contexts, key, ctx); + /* no-op. + * Public EGLContext handle and private _EGLContext are the same. + */ } @@ -66,21 +66,34 @@ _eglSaveContext(_EGLContext *ctx) void _eglRemoveContext(_EGLContext *ctx) { - EGLuint key = (EGLuint) ctx->Handle; - _eglHashRemove(_eglGlobal.Contexts, key); + /* no-op. + * Public EGLContext handle and private _EGLContext are the same. + */ +} + + +/** + * Return the public handle for the given private context ptr. + * This is the inverse of _eglLookupContext(). + */ +EGLContext +_eglGetContextHandle(_EGLContext *ctx) +{ + /* just a cast! */ + return (EGLContext) ctx; } /** * Return the _EGLContext object that corresponds to the given * EGLContext handle. + * This is the inverse of _eglGetContextHandle(). */ _EGLContext * _eglLookupContext(EGLContext ctx) { - EGLuint key = (EGLuint) ctx; - _EGLContext *c = (_EGLContext *) _eglHashLookup(_eglGlobal.Contexts, key); - return c; + /* just a cast since EGLContext is just a void ptr */ + return (_EGLContext *) ctx; } @@ -115,7 +128,7 @@ _eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, } _eglSaveContext(context); - return context->Handle; + return (EGLContext) context; #endif return EGL_NO_CONTEXT; } @@ -129,8 +142,6 @@ _eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx) { _EGLContext *context = _eglLookupContext(ctx); if (context) { - EGLuint key = (EGLuint) ctx; - _eglHashRemove(_eglGlobal.Contexts, key); if (context->IsBound) { context->DeletePending = EGL_TRUE; } @@ -243,7 +254,7 @@ _eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d, ctx = NULL; } /* really delete context now */ - drv->API.DestroyContext(drv, dpy, oldContext->Handle); + drv->API.DestroyContext(drv, dpy, _eglGetContextHandle(oldContext)); } } diff --git a/src/egl/main/eglcontext.h b/src/egl/main/eglcontext.h index 82bfde151f..093f65fbfb 100644 --- a/src/egl/main/eglcontext.h +++ b/src/egl/main/eglcontext.h @@ -11,8 +11,6 @@ */ struct _egl_context { - EGLContext Handle; /* The public/opaque handle which names this object */ - _EGLDisplay *Display; /* who do I belong to? */ _EGLConfig *Config; @@ -41,6 +39,10 @@ extern void _eglRemoveContext(_EGLContext *ctx); +extern EGLContext +_eglGetContextHandle(_EGLContext *ctx); + + extern _EGLContext * _eglLookupContext(EGLContext ctx); diff --git a/src/egl/main/eglglobals.c b/src/egl/main/eglglobals.c index 608311d749..4bdc0329bf 100644 --- a/src/egl/main/eglglobals.c +++ b/src/egl/main/eglglobals.c @@ -15,7 +15,6 @@ _eglInitGlobals(void) { if (!_eglGlobal.Initialized) { _eglGlobal.Displays = _eglNewHashTable(); - _eglGlobal.Contexts = _eglNewHashTable(); _eglGlobal.Surfaces = _eglNewHashTable(); _eglGlobal.FreeScreenHandle = 1; _eglGlobal.Initialized = EGL_TRUE; @@ -37,7 +36,6 @@ _eglDestroyGlobals(void) { /* XXX TODO walk over table entries, deleting each */ _eglDeleteHashTable(_eglGlobal.Displays); - _eglDeleteHashTable(_eglGlobal.Contexts); _eglDeleteHashTable(_eglGlobal.Surfaces); } diff --git a/src/egl/main/eglglobals.h b/src/egl/main/eglglobals.h index c16baa2d6b..4fccd226b7 100644 --- a/src/egl/main/eglglobals.h +++ b/src/egl/main/eglglobals.h @@ -24,7 +24,6 @@ struct _egl_global EGLBoolean Initialized; _EGLHashtable *Displays; - _EGLHashtable *Contexts; _EGLHashtable *Surfaces; EGLScreenMESA FreeScreenHandle; -- cgit v1.2.3 From 5f8a4f3e5e8fe78f1abe9ca6dd1131ad53d3d943 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 14:23:56 -0600 Subject: added EGL_OPENGL_API case --- src/egl/main/eglapi.c | 11 ++++++++++- src/egl/main/eglglobals.h | 1 + 2 files changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index c57d29dda6..4df31cc03f 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -511,6 +511,15 @@ eglBindAPI(EGLenum api) _EGLThreadInfo *t = _eglGetCurrentThread(); switch (api) { +#ifdef EGL_VERSION_1_4 + case EGL_OPENGL_API: + if (_eglGlobal.OpenGLAPISupported) { + t->CurrentAPI = api; + return EGL_TRUE; + } + _eglError(EGL_BAD_PARAMETER, "eglBindAPI"); + return EGL_FALSE; +#endif case EGL_OPENGL_ES_API: if (_eglGlobal.OpenGLESAPISupported) { t->CurrentAPI = api; @@ -546,7 +555,7 @@ eglCreatePbufferFromClientBuffer(EGLDisplay dpy, EGLenum buftype, EGLenum eglQueryAPI(void) { - /* returns one of EGL_OPENGL_ES_API or EGL_OPENVG_API */ + /* returns one of EGL_OPENGL_API, EGL_OPENGL_ES_API or EGL_OPENVG_API */ _EGLThreadInfo *t = _eglGetCurrentThread(); return t->CurrentAPI; } diff --git a/src/egl/main/eglglobals.h b/src/egl/main/eglglobals.h index 4fccd226b7..6b858b6b08 100644 --- a/src/egl/main/eglglobals.h +++ b/src/egl/main/eglglobals.h @@ -29,6 +29,7 @@ struct _egl_global EGLScreenMESA FreeScreenHandle; /* XXX these may be temporary */ + EGLBoolean OpenGLAPISupported; EGLBoolean OpenGLESAPISupported; EGLBoolean OpenVGAPISupported; -- cgit v1.2.3 From 721ba15bf4596b2e9589e7656005b387724875c3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 14:33:54 -0600 Subject: added _eglGet*Handle() functions These are the inverse of the _eglLookup*() functions. Returns the public handle for a private surface/config/display/etc. Removes glapi.c's direct access of private fields. --- src/egl/main/eglapi.c | 15 +++------------ src/egl/main/eglconfig.c | 12 ++++++++++++ src/egl/main/eglconfig.h | 4 ++++ src/egl/main/egldisplay.c | 15 +++++++++++++++ src/egl/main/egldisplay.h | 4 ++++ src/egl/main/eglsurface.c | 20 ++++++++++++++++++++ src/egl/main/eglsurface.h | 4 ++++ 7 files changed, 62 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index 4df31cc03f..e4eec26de0 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -53,10 +53,7 @@ eglGetDisplay(NativeDisplayType displayName) _EGLDisplay *dpy; _eglInitGlobals(); dpy = _eglNewDisplay(displayName); - if (dpy) - return dpy->Handle; - else - return EGL_NO_DISPLAY; + return _eglGetDisplayHandle(dpy); } @@ -269,10 +266,7 @@ EGLDisplay EGLAPIENTRY eglGetCurrentDisplay(void) { _EGLDisplay *dpy = _eglGetCurrentDisplay(); - if (dpy) - return dpy->Handle; - else - return EGL_NO_DISPLAY; + return _eglGetDisplayHandle(dpy); } @@ -288,10 +282,7 @@ EGLSurface EGLAPIENTRY eglGetCurrentSurface(EGLint readdraw) { _EGLSurface *s = _eglGetCurrentSurface(readdraw); - if (s) - return s->Handle; - else - return EGL_NO_SURFACE; + return _eglGetSurfaceHandle(s); } diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c index a74cd23113..1b49f7afd9 100644 --- a/src/egl/main/eglconfig.c +++ b/src/egl/main/eglconfig.c @@ -101,8 +101,20 @@ _eglInitConfig(_EGLConfig *config, EGLint id) } +/** + * Return the public handle for an internal _EGLConfig. + * This is the inverse of _eglLookupConfig(). + */ +EGLConfig +_eglGetConfigHandle(_EGLConfig *config) +{ + return config ? config->Handle : 0; +} + + /** * Given an EGLConfig handle, return the corresponding _EGLConfig object. + * This is the inverse of _eglGetConfigHandle(). */ _EGLConfig * _eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config) diff --git a/src/egl/main/eglconfig.h b/src/egl/main/eglconfig.h index 88e44dfc7d..4a80612980 100644 --- a/src/egl/main/eglconfig.h +++ b/src/egl/main/eglconfig.h @@ -27,6 +27,10 @@ extern void _eglInitConfig(_EGLConfig *config, EGLint id); +extern EGLConfig +_eglGetConfigHandle(_EGLConfig *config); + + extern _EGLConfig * _eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config); diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c index 59d0bd3dc3..fd24f22273 100644 --- a/src/egl/main/egldisplay.c +++ b/src/egl/main/egldisplay.c @@ -39,9 +39,24 @@ _eglNewDisplay(NativeDisplayType displayName) } +/** + * Return the public handle for an internal _EGLDisplay. + * This is the inverse of _eglLookupDisplay(). + */ +EGLDisplay +_eglGetDisplayHandle(_EGLDisplay *display) +{ + if (display) + return display->Handle; + else + return EGL_NO_DISPLAY; +} + + /** * Return the _EGLDisplay object that corresponds to the given public/ * opaque display handle. + * This is the inverse of _eglGetDisplayHandle(). */ _EGLDisplay * _eglLookupDisplay(EGLDisplay dpy) diff --git a/src/egl/main/egldisplay.h b/src/egl/main/egldisplay.h index 1a03fdd4ad..fe7b788455 100644 --- a/src/egl/main/egldisplay.h +++ b/src/egl/main/egldisplay.h @@ -24,6 +24,10 @@ extern _EGLDisplay * _eglNewDisplay(NativeDisplayType displayName); +EGLDisplay +_eglGetDisplayHandle(_EGLDisplay *display); + + extern _EGLDisplay * _eglLookupDisplay(EGLDisplay dpy); diff --git a/src/egl/main/eglsurface.c b/src/egl/main/eglsurface.c index 134afa7cbd..1dbb12ecfe 100644 --- a/src/egl/main/eglsurface.c +++ b/src/egl/main/eglsurface.c @@ -223,6 +223,26 @@ _eglRemoveSurface(_EGLSurface *surf) } + +/** + * Return the public handle for an internal _EGLSurface. + * This is the inverse of _eglLookupSurface(). + */ +EGLSurface +_eglGetSurfaceHandle(_EGLSurface *surface) +{ + if (surface) + return surface->Handle; + else + return EGL_NO_SURFACE; +} + + +/** + * Return the private _EGLSurface which corresponds to a public EGLSurface + * handle. + * This is the inverse of _eglGetSurfaceHandle(). + */ _EGLSurface * _eglLookupSurface(EGLSurface surf) { diff --git a/src/egl/main/eglsurface.h b/src/egl/main/eglsurface.h index 79abeca0b2..df1e70122e 100644 --- a/src/egl/main/eglsurface.h +++ b/src/egl/main/eglsurface.h @@ -51,6 +51,10 @@ extern void _eglRemoveSurface(_EGLSurface *surf); +extern EGLSurface +_eglGetSurfaceHandle(_EGLSurface *surface); + + extern _EGLSurface * _eglLookupSurface(EGLSurface surf); -- cgit v1.2.3 From e38bb10824fc3dc194d7cc6987f3f4957784310e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 23:21:50 +0100 Subject: draw: reset stipple counter whenever non-line prim encountered --- src/gallium/auxiliary/draw/draw_pipe_stipple.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c index 4673d5dcba..9522b79582 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c @@ -175,6 +175,22 @@ reset_stipple_counter(struct draw_stage *stage) stage->next->reset_stipple_counter( stage->next ); } +static void +stipple_reset_point(struct draw_stage *stage, struct prim_header *header) +{ + struct stipple_stage *stipple = stipple_stage(stage); + stipple->counter = 0; + stage->next->point(stage->next, header); +} + +static void +stipple_reset_tri(struct draw_stage *stage, struct prim_header *header) +{ + struct stipple_stage *stipple = stipple_stage(stage); + stipple->counter = 0; + stage->next->tri(stage->next, header); +} + static void stipple_first_line(struct draw_stage *stage, @@ -220,9 +236,9 @@ struct draw_stage *draw_stipple_stage( struct draw_context *draw ) stipple->stage.draw = draw; stipple->stage.next = NULL; - stipple->stage.point = draw_pipe_passthrough_point; + stipple->stage.point = stipple_reset_point; stipple->stage.line = stipple_first_line; - stipple->stage.tri = draw_pipe_passthrough_tri; + stipple->stage.tri = stipple_reset_tri; stipple->stage.reset_stipple_counter = reset_stipple_counter; stipple->stage.flush = stipple_flush; stipple->stage.destroy = stipple_destroy; -- cgit v1.2.3 From b7b9ce0f8677993c3cd5376add72a684a5653341 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 27 May 2008 23:23:37 +0100 Subject: softpipe: enable vbuf by default The non-vbuf option is going away... --- src/gallium/drivers/softpipe/sp_context.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index a48e546139..ed9322a109 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -227,11 +227,13 @@ softpipe_create( struct pipe_screen *screen, if (GETENV( "SP_NO_RAST" ) != NULL) softpipe->no_rast = TRUE; - if (GETENV( "SP_VBUF" ) != NULL) { - sp_init_vbuf(softpipe); + if (GETENV( "SP_NO_VBUF" ) != NULL) { + /* Deprecated path -- vbuf is the intended interface to the draw module: + */ + draw_set_rasterize_stage(softpipe->draw, softpipe->setup); } else { - draw_set_rasterize_stage(softpipe->draw, softpipe->setup); + sp_init_vbuf(softpipe); } /* plug in AA line/point stages */ -- cgit v1.2.3 From 6052af172f0241e6678cd16efac0a0f14f40146c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 16:48:23 -0600 Subject: minor overhaul/re-org of driver selection/loading code --- src/egl/main/Makefile | 11 ++++++--- src/egl/main/eglapi.c | 27 ++++++++++++++------- src/egl/main/eglconfig.c | 4 +--- src/egl/main/eglconfig.h | 6 ++--- src/egl/main/egldisplay.c | 57 ++++++++++++++++++++++++++++++++----------- src/egl/main/egldisplay.h | 14 ++++++++++- src/egl/main/egldriver.c | 61 +++++++++++++++++++++++++++-------------------- src/egl/main/egldriver.h | 6 ++--- 8 files changed, 124 insertions(+), 62 deletions(-) (limited to 'src') diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index e1058a23f7..0efcd4e605 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -16,7 +16,8 @@ HEADERS = \ eglhash.h \ eglmode.h \ eglscreen.h \ - eglsurface.h + eglsurface.h \ + eglx.h SOURCES = \ eglapi.c \ @@ -29,13 +30,17 @@ SOURCES = \ eglhash.c \ eglmode.c \ eglscreen.c \ - eglsurface.c + eglsurface.c \ + eglx.c OBJECTS = $(SOURCES:.c=.o) +LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 + + .c.o: - $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(LOCAL_CFLAGS) $< -o $@ diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index e4eec26de0..fe63d36b80 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -43,27 +43,38 @@ /** - * NOTE: displayName is treated as a string in _eglChooseDriver()!!! - * This will probably change! - * See _eglChooseDriver() for details! + * This is typically the first EGL function that an application calls. + * We initialize our global vars and create a private _EGLDisplay object. */ EGLDisplay EGLAPIENTRY -eglGetDisplay(NativeDisplayType displayName) +eglGetDisplay(NativeDisplayType nativeDisplay) { _EGLDisplay *dpy; _eglInitGlobals(); - dpy = _eglNewDisplay(displayName); + dpy = _eglNewDisplay(nativeDisplay); return _eglGetDisplayHandle(dpy); } +/** + * This is typically the second EGL function that an application calls. + * Here we load/initialize the actual hardware driver. + */ EGLBoolean EGLAPIENTRY eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor) { if (dpy) { - _EGLDriver *drv = _eglChooseDriver(dpy); - if (drv) - return drv->API.Initialize(drv, dpy, major, minor); + _EGLDisplay *dpyPriv = _eglLookupDisplay(dpy); + if (!dpyPriv) { + return EGL_FALSE; + } + dpyPriv->Driver = _eglOpenDriver(dpyPriv, dpyPriv->DriverName); + if (!dpyPriv->Driver) { + return EGL_FALSE; + } + /* Initialize the particular driver now */ + return dpyPriv->Driver->API.Initialize(dpyPriv->Driver, dpy, + major, minor); } return EGL_FALSE; } diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c index 1b49f7afd9..eb2c34a802 100644 --- a/src/egl/main/eglconfig.c +++ b/src/egl/main/eglconfig.c @@ -17,7 +17,6 @@ #define MIN2(A, B) (((A) < (B)) ? (A) : (B)) -#if 0 /** * Convert an _EGLConfig to a __GLcontextModes object. * NOTE: This routine may be incomplete - we're only making sure that @@ -58,7 +57,6 @@ _eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode) mode->visualType = GLX_TRUE_COLOR; mode->renderType = GLX_RGBA_BIT; } -#endif void @@ -445,6 +443,7 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, } +#if 0 /** * Creates a set of \c __GLcontextModes that a driver will expose. * @@ -512,7 +511,6 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, * \c GL_UNSIGNED_3BYTE_8_8_8, \c GL_4FLOAT_32_32_32_32, * \c GL_4HALF_16_16_16_16, etc. We can cross that bridge when we come to it. */ -#if 0 GLboolean _eglFillInConfigs(_EGLConfig * configs, GLenum fb_format, GLenum fb_type, diff --git a/src/egl/main/eglconfig.h b/src/egl/main/eglconfig.h index 4a80612980..e025f7f845 100644 --- a/src/egl/main/eglconfig.h +++ b/src/egl/main/eglconfig.h @@ -3,9 +3,8 @@ #include "egltypedefs.h" -#if 0 +#include #include "GL/internal/glcore.h" -#endif #define MAX_ATTRIBS 100 @@ -68,10 +67,9 @@ _eglFillInConfigs( _EGLConfig *configs, int visType ); #endif -#if 0 + extern void _eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode); -#endif #endif /* EGLCONFIG_INCLUDED */ diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c index fd24f22273..9c42194c61 100644 --- a/src/egl/main/egldisplay.c +++ b/src/egl/main/egldisplay.c @@ -1,7 +1,14 @@ + +/** + * Functions related to EGLDisplay. + */ + +#include #include #include #include "eglcontext.h" #include "egldisplay.h" +#include "egldriver.h" #include "eglglobals.h" #include "eglhash.h" @@ -9,31 +16,41 @@ static char * my_strdup(const char *s) { - int l = strlen(s); - char *s2 = malloc(l + 1); - strcpy(s2, s); - return s2; + if (s) { + int l = strlen(s); + char *s2 = malloc(l + 1); + if (s2) + strcpy(s2, s); + return s2; + } + return NULL; } /** - * We're assuming that the NativeDisplayType parameter is actually - * a string. - * Return a new _EGLDisplay object for the given displayName + * Allocate a new _EGLDisplay object for the given nativeDisplay handle. + * We'll also try to determine the device driver name at this time. */ _EGLDisplay * -_eglNewDisplay(NativeDisplayType displayName) +_eglNewDisplay(NativeDisplayType nativeDisplay) { _EGLDisplay *dpy = (_EGLDisplay *) calloc(1, sizeof(_EGLDisplay)); if (dpy) { EGLuint key = _eglHashGenKey(_eglGlobal.Displays); + dpy->Handle = (EGLDisplay) key; _eglHashInsert(_eglGlobal.Displays, key, dpy); - if (displayName) - dpy->Name = my_strdup((char *) displayName); - else - dpy->Name = NULL; - dpy->Driver = NULL; /* this gets set later */ + + dpy->NativeDisplay = nativeDisplay; +#if defined(_EGL_PLATFORM_X) + dpy->Xdpy = (Display *) nativeDisplay; +#endif + + dpy->DriverName = my_strdup(_eglChooseDriver(dpy)); + if (!dpy->DriverName) { + free(dpy); + return NULL; + } } return dpy; } @@ -67,6 +84,18 @@ _eglLookupDisplay(EGLDisplay dpy) } +void +_eglSaveDisplay(_EGLDisplay *dpy) +{ + EGLuint key = _eglHashGenKey(_eglGlobal.Displays); + assert(dpy); + assert(!dpy->Handle); + dpy->Handle = (EGLDisplay) key; + assert(dpy->Handle); + _eglHashInsert(_eglGlobal.Displays, key, dpy); +} + + _EGLDisplay * _eglGetCurrentDisplay(void) { @@ -83,6 +112,6 @@ _eglCleanupDisplay(_EGLDisplay *disp) { /* XXX incomplete */ free(disp->Configs); - free(disp->Name); + free((void *) disp->DriverName); /* driver deletes _EGLDisplay */ } diff --git a/src/egl/main/egldisplay.h b/src/egl/main/egldisplay.h index fe7b788455..be134374ca 100644 --- a/src/egl/main/egldisplay.h +++ b/src/egl/main/egldisplay.h @@ -1,15 +1,19 @@ #ifndef EGLDISPLAY_INCLUDED #define EGLDISPLAY_INCLUDED +#ifdef _EGL_PLATFORM_X +#include +#endif #include "egltypedefs.h" struct _egl_display { + EGLNativeDisplayType NativeDisplay; EGLDisplay Handle; - char *Name; + const char *DriverName; _EGLDriver *Driver; EGLint NumScreens; @@ -17,6 +21,10 @@ struct _egl_display EGLint NumConfigs; _EGLConfig *Configs; /* array [NumConfigs] */ + +#ifdef _EGL_PLATFORM_X + Display *Xdpy; +#endif }; @@ -32,6 +40,10 @@ extern _EGLDisplay * _eglLookupDisplay(EGLDisplay dpy); +extern void +_eglSaveDisplay(_EGLDisplay *dpy); + + extern _EGLDisplay * _eglGetCurrentDisplay(void); diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index bda06dd827..50c466c258 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -1,3 +1,8 @@ +/** + * Functions for choosing and opening/loading device drivers. + */ + + #include #include #include @@ -11,62 +16,64 @@ #include "eglmode.h" #include "eglscreen.h" #include "eglsurface.h" +#include "eglx.h" const char *DefaultDriverName = "demodriver"; /** - * Choose and open/init the hardware driver for the given EGLDisplay. - * Previously, the EGLDisplay was created with _eglNewDisplay() where - * we recorded the user's NativeDisplayType parameter. + * Determine/return the name of the driver to use for the given _EGLDisplay. * - * Now we'll use the NativeDisplayType value. + * Try to be clever and determine if nativeDisplay is an Xlib Display + * ptr or a string (naming a driver or screen number, etc). * - * Currently, the native display value is treated as a string. * If the first character is ':' we interpret it as a screen or card index * number (i.e. ":0" or ":1", etc) * Else if the first character is '!' we interpret it as specific driver name * (i.e. "!r200" or "!i830". + * + * The caller should make a copy of the returned string. */ -_EGLDriver * -_eglChooseDriver(EGLDisplay display) +const char * +_eglChooseDriver(_EGLDisplay *dpy) { - _EGLDisplay *dpy = _eglLookupDisplay(display); - _EGLDriver *drv; - const char *driverName = DefaultDriverName; - const char *name; + const char *name = (const char *) dpy->NativeDisplay; + const char *driverName = NULL; - assert(dpy); - - name = dpy->Name; - if (!name) { - /* use default */ + if (!dpy->NativeDisplay) { + /* choose a default */ + driverName = DefaultDriverName; } - else if (name[0] == ':' && (name[1] >= '0' && name[1] <= '9') && !name[2]) { + else if (name && name[0] == ':' && + (name[1] >= '0' && name[1] <= '9') && !name[2]) { /* XXX probe hardware here to determine which driver to open */ driverName = "libEGLdri"; } - else if (name[0] == '!') { + else if (name && name[0] == '!') { /* use specified driver name */ driverName = name + 1; } else { - /* Maybe display was returned by XOpenDisplay? */ - _eglLog(_EGL_FATAL, "eglChooseDriver() bad name"); +#if defined(_EGL_PLATFORM_X) + driverName = _xeglChooseDriver(dpy); +#elif defined(_EGL_PLATFORM_WINDOWS) + /* XXX to do */ + driverName = _weglChooseDriver(dpy); +#elif defined(_EGL_PLATFORM_WINCE) + /* XXX to do */ +#endif } - _eglLog(_EGL_INFO, "eglChooseDriver() choosing %s", driverName); - - drv = _eglOpenDriver(dpy, driverName); - dpy->Driver = drv; - - return drv; + return driverName; } /** * Open/load the named driver and call its bootstrap function: _eglMain(). + * By the time this function is called, the dpy->DriverName should have + * been determined. + * * \return new _EGLDriver object. */ _EGLDriver * @@ -77,6 +84,8 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName) void *lib; char driverFilename[1000]; + assert(driverName); + /* XXX also prepend a directory path??? */ sprintf(driverFilename, "%s.so", driverName); diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index 88526e973d..bde726e25e 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -45,12 +45,12 @@ struct _egl_driver extern _EGLDriver *_eglMain(_EGLDisplay *dpy); -extern _EGLDriver * -_eglChooseDriver(EGLDisplay dpy); +extern const char * +_eglChooseDriver(_EGLDisplay *dpy); extern _EGLDriver * -_eglOpenDriver(_EGLDisplay *dpy, const char *driverName); +_eglOpenDriver(_EGLDisplay *dpy, const char *DriverName); extern EGLBoolean -- cgit v1.2.3 From 07e92b174214e6bdaa22eecc0ed87cf1dbc8fc10 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 16:49:11 -0600 Subject: misc updates, clean-ups, remove dependency on mtypes.h --- src/egl/drivers/dri/Makefile | 2 +- src/egl/drivers/dri/egldri.c | 28 +++++++++++++++++----------- src/egl/drivers/dri/egldri.h | 8 +++++--- 3 files changed, 23 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/dri/Makefile b/src/egl/drivers/dri/Makefile index fa3720a16b..13aacd4389 100644 --- a/src/egl/drivers/dri/Makefile +++ b/src/egl/drivers/dri/Makefile @@ -55,7 +55,7 @@ depend: $(SOURCES) $(HEADERS) @ rm -f depend @ touch depend $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) \ - $(SOURCES) $(HEADERS) > /dev/null + $(SOURCES) $(HEADERS) >/dev/null 2>/dev/null include depend # DO NOT DELETE diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c index cab0be2bd1..81c12b7ca6 100644 --- a/src/egl/drivers/dri/egldri.c +++ b/src/egl/drivers/dri/egldri.c @@ -32,7 +32,7 @@ #include "egldri.h" const char *sysfs = "/sys/class"; -#define None 0 + static const int empty_attribute_list[1] = { None }; @@ -62,10 +62,11 @@ _eglMain(_EGLDisplay *dpy) if (strncmp(&dirent->d_name[0], "card", 4) != 0) continue; - if (strcmp(&dirent->d_name[4], &dpy->Name[1]) != 0) + if (strcmp(&dirent->d_name[4], &dpy->DriverName[1]) != 0) continue; - snprintf(path, sizeof(path), "%s/drm/card%s/dri_library_name", sysfs, &dpy->Name[1]); + snprintf(path, sizeof(path), "%s/drm/card%s/dri_library_name", + sysfs, &dpy->DriverName[1]); _eglLog(_EGL_INFO, "Opening %s", path); #if 1 file = fopen(path, "r"); @@ -141,7 +142,7 @@ _eglDRICreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, /* generate handle and insert into hash table */ _eglSaveContext(&c->Base); - return c->Base.Handle; + return _eglGetContextHandle(&c->Base); } @@ -152,13 +153,15 @@ _eglDRIMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, driDisplay *disp = Lookup_driDisplay(dpy); driContext *ctx = Lookup_driContext(context); EGLBoolean b; + __DRIid drawBuf = (__DRIid) draw; + __DRIid readBuf = (__DRIid) read; b = _eglMakeCurrent(drv, dpy, draw, read, context); if (!b) return EGL_FALSE; if (ctx) { - ctx->driContext.bindContext(disp, 0, read, draw, &ctx->driContext); + ctx->driContext.bindContext(disp, 0, drawBuf, readBuf, &ctx->driContext); } else { /* what's this??? */ @@ -190,7 +193,7 @@ _eglDRICreatePbufferSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, #if 0 GLcontext *ctx = NULL; /* this _should_ be OK */ #endif - GLvisual visMode; + __GLcontextModes visMode; _EGLConfig *conf = _eglLookupConfig(drv, dpy, config); assert(conf); /* bad config should be caught earlier */ _eglConfigToContextModesRec(conf, &visMode); @@ -267,7 +270,8 @@ _eglDRICreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, _EGLConfig *config = _eglLookupConfig(drv, dpy, cfg); driDisplay *disp = Lookup_driDisplay(dpy); driSurface *surface; - GLvisual visMode; + __GLcontextModes visMode; + __DRIid drawBuf; surface = (driSurface *) calloc(1, sizeof(*surface)); if (!surface) { @@ -292,8 +296,10 @@ _eglDRICreateScreenSurfaceMESA(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, /* convert EGLConfig to GLvisual */ _eglConfigToContextModesRec(config, &visMode); + drawBuf = (__DRIid) _eglGetSurfaceHandle(&surface->Base); + /* Create a new DRI drawable */ - if (!disp->driScreen.createNewDrawable(disp, &visMode, surface->Base.Handle, + if (!disp->driScreen.createNewDrawable(disp, &visMode, drawBuf, &surface->drawable, GLX_WINDOW_BIT, empty_attribute_list)) { _eglRemoveSurface(&surface->Base); @@ -715,7 +721,7 @@ __eglGetDrawableInfo(__DRInativeDisplay * ndpy, int screen, __DRIid drawable, { __DRIscreen *pDRIScreen; __DRIscreenPrivate *psp; - driSurface *surf = Lookup_driSurface(drawable); + driSurface *surf = Lookup_driSurface((EGLSurface) drawable); pDRIScreen = __eglFindDRIScreen(ndpy, screen); @@ -1088,13 +1094,13 @@ _eglDRIInitialize(_EGLDriver *drv, EGLDisplay dpy, */ display = calloc(1, sizeof(*display)); display->Base = *disp; - _eglHashInsert(_eglGlobal.Displays, disp->Handle, display); + _eglSaveDisplay(&display->Base); free(disp); *major = 1; *minor = 0; - sscanf(&disp->Name[1], "%d", &display->minor); + sscanf(&disp->DriverName[1], "%d", &display->minor); drv->Initialized = EGL_TRUE; return EGL_TRUE; diff --git a/src/egl/drivers/dri/egldri.h b/src/egl/drivers/dri/egldri.h index 34b12d64fc..49e1cf4d99 100644 --- a/src/egl/drivers/dri/egldri.h +++ b/src/egl/drivers/dri/egldri.h @@ -1,11 +1,13 @@ #ifndef EGLDRI_INCLUDED #define EGLDRI_INCLUDED +#include +#include #include "egldisplay.h" #include "eglscreen.h" #include "eglsurface.h" #include "eglcontext.h" -#include "mtypes.h" + #include "dri_util.h" #include "drm_sarea.h" @@ -14,7 +16,7 @@ */ typedef struct dri_display { - _EGLDisplay Base; /* base class/object */ + _EGLDisplay Base; /**< base class */ void *pFB; int drmFD; /**< \brief DRM device file descriptor */ int minor; @@ -32,7 +34,7 @@ typedef struct dri_display unsigned long FBStart; /**< \brief physical address of the framebuffer */ void *driverClientMsg; int driverClientMsgSize; - int chipset; + unsigned chipset; void *driverPrivate; drm_magic_t magic; -- cgit v1.2.3 From a772bbb16ec91a8714a498e8089f96f45730153c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 16:57:49 -0600 Subject: Move some config-related utility functions into new eglconfigutil.c file. --- src/egl/main/Makefile | 2 + src/egl/main/eglconfig.c | 253 +---------------------------------------- src/egl/main/eglconfig.h | 15 --- src/egl/main/eglconfigutil.c | 260 +++++++++++++++++++++++++++++++++++++++++++ src/egl/main/eglconfigutil.h | 23 ++++ 5 files changed, 286 insertions(+), 267 deletions(-) create mode 100644 src/egl/main/eglconfigutil.c create mode 100644 src/egl/main/eglconfigutil.h (limited to 'src') diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index 0efcd4e605..4fb0344e63 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -8,6 +8,7 @@ INCLUDE_DIRS = -I$(TOP)/include -I$(TOP)/src/mesa/glapi HEADERS = \ eglconfig.h \ + eglconfigutil.h \ eglcontext.h \ egldisplay.h \ egldriver.h \ @@ -22,6 +23,7 @@ HEADERS = \ SOURCES = \ eglapi.c \ eglconfig.c \ + eglconfigutil.c \ eglcontext.c \ egldisplay.c \ egldriver.c \ diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c index eb2c34a802..c6369e7adf 100644 --- a/src/egl/main/eglconfig.c +++ b/src/egl/main/eglconfig.c @@ -17,48 +17,6 @@ #define MIN2(A, B) (((A) < (B)) ? (A) : (B)) -/** - * Convert an _EGLConfig to a __GLcontextModes object. - * NOTE: This routine may be incomplete - we're only making sure that - * the fields needed by Mesa (for _mesa_create_context/framebuffer) are - * set correctly. - */ -void -_eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode) -{ - memset(mode, 0, sizeof(*mode)); - - mode->rgbMode = GL_TRUE; /* no color index */ - mode->colorIndexMode = GL_FALSE; - mode->doubleBufferMode = GL_TRUE; /* always DB for now */ - mode->stereoMode = GL_FALSE; - - mode->redBits = GET_CONFIG_ATTRIB(config, EGL_RED_SIZE); - mode->greenBits = GET_CONFIG_ATTRIB(config, EGL_GREEN_SIZE); - mode->blueBits = GET_CONFIG_ATTRIB(config, EGL_BLUE_SIZE); - mode->alphaBits = GET_CONFIG_ATTRIB(config, EGL_ALPHA_SIZE); - mode->rgbBits = GET_CONFIG_ATTRIB(config, EGL_BUFFER_SIZE); - - /* no rgba masks - fix? */ - - mode->depthBits = GET_CONFIG_ATTRIB(config, EGL_DEPTH_SIZE); - mode->haveDepthBuffer = mode->depthBits > 0; - - mode->stencilBits = GET_CONFIG_ATTRIB(config, EGL_STENCIL_SIZE); - mode->haveStencilBuffer = mode->stencilBits > 0; - - /* no accum */ - - mode->level = GET_CONFIG_ATTRIB(config, EGL_LEVEL); - mode->samples = GET_CONFIG_ATTRIB(config, EGL_SAMPLES); - mode->sampleBuffers = GET_CONFIG_ATTRIB(config, EGL_SAMPLE_BUFFERS); - - /* surface type - not really needed */ - mode->visualType = GLX_TRUE_COLOR; - mode->renderType = GLX_RGBA_BIT; -} - - void _eglSetConfigAttrib(_EGLConfig *config, EGLint attr, EGLint val) { @@ -106,7 +64,7 @@ _eglInitConfig(_EGLConfig *config, EGLint id) EGLConfig _eglGetConfigHandle(_EGLConfig *config) { - return config ? config->Handle : 0; + return config ? config->Handle : 0; } @@ -441,212 +399,3 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, return EGL_TRUE; } - - -#if 0 -/** - * Creates a set of \c __GLcontextModes that a driver will expose. - * - * A set of \c __GLcontextModes will be created based on the supplied - * parameters. The number of modes processed will be 2 * - * \c num_depth_stencil_bits * \c num_db_modes. - * - * For the most part, data is just copied from \c depth_bits, \c stencil_bits, - * \c db_modes, and \c visType into each \c __GLcontextModes element. - * However, the meanings of \c fb_format and \c fb_type require further - * explanation. The \c fb_format specifies which color components are in - * each pixel and what the default order is. For example, \c GL_RGB specifies - * that red, green, blue are available and red is in the "most significant" - * position and blue is in the "least significant". The \c fb_type specifies - * the bit sizes of each component and the actual ordering. For example, if - * \c GL_UNSIGNED_SHORT_5_6_5_REV is specified with \c GL_RGB, bits [15:11] - * are the blue value, bits [10:5] are the green value, and bits [4:0] are - * the red value. - * - * One sublte issue is the combination of \c GL_RGB or \c GL_BGR and either - * of the \c GL_UNSIGNED_INT_8_8_8_8 modes. The resulting mask values in the - * \c __GLcontextModes structure is \b identical to the \c GL_RGBA or - * \c GL_BGRA case, except the \c alphaMask is zero. This means that, as - * far as this routine is concerned, \c GL_RGB with \c GL_UNSIGNED_INT_8_8_8_8 - * still uses 32-bits. - * - * If in doubt, look at the tables used in the function. - * - * \param ptr_to_modes Pointer to a pointer to a linked list of - * \c __GLcontextModes. Upon completion, a pointer to - * the next element to be process will be stored here. - * If the function fails and returns \c GL_FALSE, this - * value will be unmodified, but some elements in the - * linked list may be modified. - * \param fb_format Format of the framebuffer. Currently only \c GL_RGB, - * \c GL_RGBA, \c GL_BGR, and \c GL_BGRA are supported. - * \param fb_type Type of the pixels in the framebuffer. Currently only - * \c GL_UNSIGNED_SHORT_5_6_5, - * \c GL_UNSIGNED_SHORT_5_6_5_REV, - * \c GL_UNSIGNED_INT_8_8_8_8, and - * \c GL_UNSIGNED_INT_8_8_8_8_REV are supported. - * \param depth_bits Array of depth buffer sizes to be exposed. - * \param stencil_bits Array of stencil buffer sizes to be exposed. - * \param num_depth_stencil_bits Number of entries in both \c depth_bits and - * \c stencil_bits. - * \param db_modes Array of buffer swap modes. If an element has a - * value of \c GLX_NONE, then it represents a - * single-buffered mode. Other valid values are - * \c GLX_SWAP_EXCHANGE_OML, \c GLX_SWAP_COPY_OML, and - * \c GLX_SWAP_UNDEFINED_OML. See the - * GLX_OML_swap_method extension spec for more details. - * \param num_db_modes Number of entries in \c db_modes. - * \param visType GLX visual type. Usually either \c GLX_TRUE_COLOR or - * \c GLX_DIRECT_COLOR. - * - * \returns - * \c GL_TRUE on success or \c GL_FALSE on failure. Currently the only - * cause of failure is a bad parameter (i.e., unsupported \c fb_format or - * \c fb_type). - * - * \todo - * There is currently no way to support packed RGB modes (i.e., modes with - * exactly 3 bytes per pixel) or floating-point modes. This could probably - * be done by creating some new, private enums with clever names likes - * \c GL_UNSIGNED_3BYTE_8_8_8, \c GL_4FLOAT_32_32_32_32, - * \c GL_4HALF_16_16_16_16, etc. We can cross that bridge when we come to it. - */ -GLboolean -_eglFillInConfigs(_EGLConfig * configs, - GLenum fb_format, GLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, - unsigned num_depth_stencil_bits, - const GLenum * db_modes, unsigned num_db_modes, - int visType) -{ - static const u_int8_t bits_table[3][4] = { - /* R G B A */ - { 5, 6, 5, 0 }, /* Any GL_UNSIGNED_SHORT_5_6_5 */ - { 8, 8, 8, 0 }, /* Any RGB with any GL_UNSIGNED_INT_8_8_8_8 */ - { 8, 8, 8, 8 } /* Any RGBA with any GL_UNSIGNED_INT_8_8_8_8 */ - }; - - /* The following arrays are all indexed by the fb_type masked with 0x07. - * Given the four supported fb_type values, this results in valid array - * indices of 3, 4, 5, and 7. - */ - static const u_int32_t masks_table_rgb[8][4] = { - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000}, /* 5_6_5 */ - {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000}, /* 5_6_5_REV */ - {0xFF000000, 0x00FF0000, 0x0000FF00, 0x00000000}, /* 8_8_8_8 */ - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x000000FF, 0x0000FF00, 0x00FF0000, 0x00000000} /* 8_8_8_8_REV */ - }; - - static const u_int32_t masks_table_rgba[8][4] = { - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000}, /* 5_6_5 */ - {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000}, /* 5_6_5_REV */ - {0xFF000000, 0x00FF0000, 0x0000FF00, 0x000000FF}, /* 8_8_8_8 */ - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x000000FF, 0x0000FF00, 0x00FF0000, 0xFF000000}, /* 8_8_8_8_REV */ - }; - - static const u_int32_t masks_table_bgr[8][4] = { - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000}, /* 5_6_5 */ - {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000}, /* 5_6_5_REV */ - {0x0000FF00, 0x00FF0000, 0xFF000000, 0x00000000}, /* 8_8_8_8 */ - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00FF0000, 0x0000FF00, 0x000000FF, 0x00000000}, /* 8_8_8_8_REV */ - }; - - static const u_int32_t masks_table_bgra[8][4] = { - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000}, /* 5_6_5 */ - {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000}, /* 5_6_5_REV */ - {0x0000FF00, 0x00FF0000, 0xFF000000, 0x000000FF}, /* 8_8_8_8 */ - {0x00000000, 0x00000000, 0x00000000, 0x00000000}, - {0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000}, /* 8_8_8_8_REV */ - }; - - static const u_int8_t bytes_per_pixel[8] = { - 0, 0, 0, 2, 2, 4, 0, 4 - }; - - const u_int8_t * bits; - const u_int32_t * masks; - const int index = fb_type & 0x07; - _EGLConfig *config; - unsigned i; - unsigned j; - unsigned k; - - if ( bytes_per_pixel[index] == 0 ) { - _eglLog(_EGL_INFO, - "[%s:%u] Framebuffer type 0x%04x has 0 bytes per pixel.", - __FUNCTION__, __LINE__, fb_type); - return GL_FALSE; - } - - /* Valid types are GL_UNSIGNED_SHORT_5_6_5 and GL_UNSIGNED_INT_8_8_8_8 and - * the _REV versions. - * - * Valid formats are GL_RGBA, GL_RGB, and GL_BGRA. - */ - switch ( fb_format ) { - case GL_RGB: - bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[1]; - masks = masks_table_rgb[index]; - break; - - case GL_RGBA: - bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[2]; - masks = masks_table_rgba[index]; - break; - - case GL_BGR: - bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[1]; - masks = masks_table_bgr[index]; - break; - - case GL_BGRA: - bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[2]; - masks = masks_table_bgra[index]; - break; - - default: - _eglLog(_EGL_WARNING, - "[%s:%u] Framebuffer format 0x%04x is not GL_RGB, GL_RGBA, GL_BGR, or GL_BGRA.", - __FUNCTION__, __LINE__, fb_format); - return GL_FALSE; - } - - config = configs; - for (k = 0; k < num_depth_stencil_bits; k++) { - for (i = 0; i < num_db_modes; i++) { - for (j = 0; j < 2; j++) { - _eglSetConfigAttrib(config, EGL_RED_SIZE, bits[0]); - _eglSetConfigAttrib(config, EGL_GREEN_SIZE, bits[1]); - _eglSetConfigAttrib(config, EGL_BLUE_SIZE, bits[2]); - _eglSetConfigAttrib(config, EGL_ALPHA_SIZE, bits[3]); - _eglSetConfigAttrib(config, EGL_BUFFER_SIZE, - bits[0] + bits[1] + bits[2] + bits[3]); - - _eglSetConfigAttrib(config, EGL_STENCIL_SIZE, stencil_bits[k]); - _eglSetConfigAttrib(config, EGL_DEPTH_SIZE, depth_bits[i]); - - _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_SCREEN_BIT_MESA | - EGL_PBUFFER_BIT | EGL_PIXMAP_BIT | EGL_WINDOW_BIT); - - config++; - } - } - } - return GL_TRUE; -} -#endif diff --git a/src/egl/main/eglconfig.h b/src/egl/main/eglconfig.h index e025f7f845..b10a61985d 100644 --- a/src/egl/main/eglconfig.h +++ b/src/egl/main/eglconfig.h @@ -4,7 +4,6 @@ #include "egltypedefs.h" #include -#include "GL/internal/glcore.h" #define MAX_ATTRIBS 100 @@ -57,19 +56,5 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, EGLint confi extern void _eglSetConfigAttrib(_EGLConfig *config, EGLint attr, EGLint val); -#if 0 -extern EGLBoolean -_eglFillInConfigs( _EGLConfig *configs, - EGLenum fb_format, EGLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, - unsigned num_depth_stencil_bits, - const EGLenum * db_modes, unsigned num_db_modes, - int visType ); -#endif - - -extern void -_eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode); - #endif /* EGLCONFIG_INCLUDED */ diff --git a/src/egl/main/eglconfigutil.c b/src/egl/main/eglconfigutil.c new file mode 100644 index 0000000000..b6585619a5 --- /dev/null +++ b/src/egl/main/eglconfigutil.c @@ -0,0 +1,260 @@ +/** + * Extra utility functions related to EGL configs. + */ + + +#include +#include +#include +#include "eglconfigutil.h" +#include "egllog.h" + + +/** + * Convert an _EGLConfig to a __GLcontextModes object. + * NOTE: This routine may be incomplete - we're only making sure that + * the fields needed by Mesa (for _mesa_create_context/framebuffer) are + * set correctly. + */ +void +_eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode) +{ + memset(mode, 0, sizeof(*mode)); + + mode->rgbMode = GL_TRUE; /* no color index */ + mode->colorIndexMode = GL_FALSE; + mode->doubleBufferMode = GL_TRUE; /* always DB for now */ + mode->stereoMode = GL_FALSE; + + mode->redBits = GET_CONFIG_ATTRIB(config, EGL_RED_SIZE); + mode->greenBits = GET_CONFIG_ATTRIB(config, EGL_GREEN_SIZE); + mode->blueBits = GET_CONFIG_ATTRIB(config, EGL_BLUE_SIZE); + mode->alphaBits = GET_CONFIG_ATTRIB(config, EGL_ALPHA_SIZE); + mode->rgbBits = GET_CONFIG_ATTRIB(config, EGL_BUFFER_SIZE); + + /* no rgba masks - fix? */ + + mode->depthBits = GET_CONFIG_ATTRIB(config, EGL_DEPTH_SIZE); + mode->haveDepthBuffer = mode->depthBits > 0; + + mode->stencilBits = GET_CONFIG_ATTRIB(config, EGL_STENCIL_SIZE); + mode->haveStencilBuffer = mode->stencilBits > 0; + + /* no accum */ + + mode->level = GET_CONFIG_ATTRIB(config, EGL_LEVEL); + mode->samples = GET_CONFIG_ATTRIB(config, EGL_SAMPLES); + mode->sampleBuffers = GET_CONFIG_ATTRIB(config, EGL_SAMPLE_BUFFERS); + + /* surface type - not really needed */ + mode->visualType = GLX_TRUE_COLOR; + mode->renderType = GLX_RGBA_BIT; +} + + + +/** + * Creates a set of \c _EGLConfigs that a driver will expose. + * + * A set of \c __GLcontextModes will be created based on the supplied + * parameters. The number of modes processed will be 2 * + * \c num_depth_stencil_bits * \c num_db_modes. + * + * For the most part, data is just copied from \c depth_bits, \c stencil_bits, + * \c db_modes, and \c visType into each \c __GLcontextModes element. + * However, the meanings of \c fb_format and \c fb_type require further + * explanation. The \c fb_format specifies which color components are in + * each pixel and what the default order is. For example, \c GL_RGB specifies + * that red, green, blue are available and red is in the "most significant" + * position and blue is in the "least significant". The \c fb_type specifies + * the bit sizes of each component and the actual ordering. For example, if + * \c GL_UNSIGNED_SHORT_5_6_5_REV is specified with \c GL_RGB, bits [15:11] + * are the blue value, bits [10:5] are the green value, and bits [4:0] are + * the red value. + * + * One sublte issue is the combination of \c GL_RGB or \c GL_BGR and either + * of the \c GL_UNSIGNED_INT_8_8_8_8 modes. The resulting mask values in the + * \c __GLcontextModes structure is \b identical to the \c GL_RGBA or + * \c GL_BGRA case, except the \c alphaMask is zero. This means that, as + * far as this routine is concerned, \c GL_RGB with \c GL_UNSIGNED_INT_8_8_8_8 + * still uses 32-bits. + * + * If in doubt, look at the tables used in the function. + * + * \param configs the array of configs generated + * \param fb_format Format of the framebuffer. Currently only \c GL_RGB, + * \c GL_RGBA, \c GL_BGR, and \c GL_BGRA are supported. + * \param fb_type Type of the pixels in the framebuffer. Currently only + * \c GL_UNSIGNED_SHORT_5_6_5, + * \c GL_UNSIGNED_SHORT_5_6_5_REV, + * \c GL_UNSIGNED_INT_8_8_8_8, and + * \c GL_UNSIGNED_INT_8_8_8_8_REV are supported. + * \param depth_bits Array of depth buffer sizes to be exposed. + * \param stencil_bits Array of stencil buffer sizes to be exposed. + * \param num_depth_stencil_bits Number of entries in both \c depth_bits and + * \c stencil_bits. + * \param db_modes Array of buffer swap modes. If an element has a + * value of \c GLX_NONE, then it represents a + * single-buffered mode. Other valid values are + * \c GLX_SWAP_EXCHANGE_OML, \c GLX_SWAP_COPY_OML, and + * \c GLX_SWAP_UNDEFINED_OML. See the + * GLX_OML_swap_method extension spec for more details. + * \param num_db_modes Number of entries in \c db_modes. + * \param visType GLX visual type. Usually either \c GLX_TRUE_COLOR or + * \c GLX_DIRECT_COLOR. + * + * \returns + * \c GL_TRUE on success or \c GL_FALSE on failure. Currently the only + * cause of failure is a bad parameter (i.e., unsupported \c fb_format or + * \c fb_type). + * + * \todo + * There is currently no way to support packed RGB modes (i.e., modes with + * exactly 3 bytes per pixel) or floating-point modes. This could probably + * be done by creating some new, private enums with clever names likes + * \c GL_UNSIGNED_3BYTE_8_8_8, \c GL_4FLOAT_32_32_32_32, + * \c GL_4HALF_16_16_16_16, etc. We can cross that bridge when we come to it. + */ +EGLBoolean +_eglFillInConfigs(_EGLConfig * configs, + GLenum fb_format, GLenum fb_type, + const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + unsigned num_depth_stencil_bits, + const GLenum * db_modes, unsigned num_db_modes, + int visType) +{ +#if 0 + static const u_int8_t bits_table[3][4] = { + /* R G B A */ + { 5, 6, 5, 0 }, /* Any GL_UNSIGNED_SHORT_5_6_5 */ + { 8, 8, 8, 0 }, /* Any RGB with any GL_UNSIGNED_INT_8_8_8_8 */ + { 8, 8, 8, 8 } /* Any RGBA with any GL_UNSIGNED_INT_8_8_8_8 */ + }; + + /* The following arrays are all indexed by the fb_type masked with 0x07. + * Given the four supported fb_type values, this results in valid array + * indices of 3, 4, 5, and 7. + */ + static const u_int32_t masks_table_rgb[8][4] = { + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000}, /* 5_6_5 */ + {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000}, /* 5_6_5_REV */ + {0xFF000000, 0x00FF0000, 0x0000FF00, 0x00000000}, /* 8_8_8_8 */ + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x000000FF, 0x0000FF00, 0x00FF0000, 0x00000000} /* 8_8_8_8_REV */ + }; + + static const u_int32_t masks_table_rgba[8][4] = { + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000}, /* 5_6_5 */ + {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000}, /* 5_6_5_REV */ + {0xFF000000, 0x00FF0000, 0x0000FF00, 0x000000FF}, /* 8_8_8_8 */ + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x000000FF, 0x0000FF00, 0x00FF0000, 0xFF000000}, /* 8_8_8_8_REV */ + }; + + static const u_int32_t masks_table_bgr[8][4] = { + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000}, /* 5_6_5 */ + {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000}, /* 5_6_5_REV */ + {0x0000FF00, 0x00FF0000, 0xFF000000, 0x00000000}, /* 8_8_8_8 */ + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00FF0000, 0x0000FF00, 0x000000FF, 0x00000000}, /* 8_8_8_8_REV */ + }; + + static const u_int32_t masks_table_bgra[8][4] = { + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x0000001F, 0x000007E0, 0x0000F800, 0x00000000}, /* 5_6_5 */ + {0x0000F800, 0x000007E0, 0x0000001F, 0x00000000}, /* 5_6_5_REV */ + {0x0000FF00, 0x00FF0000, 0xFF000000, 0x000000FF}, /* 8_8_8_8 */ + {0x00000000, 0x00000000, 0x00000000, 0x00000000}, + {0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000}, /* 8_8_8_8_REV */ + }; + + static const u_int8_t bytes_per_pixel[8] = { + 0, 0, 0, 2, 2, 4, 0, 4 + }; + + const u_int8_t * bits; + const u_int32_t * masks; + const int index = fb_type & 0x07; + _EGLConfig *config; + unsigned i; + unsigned j; + unsigned k; + + if ( bytes_per_pixel[index] == 0 ) { + _eglLog(_EGL_INFO, + "[%s:%u] Framebuffer type 0x%04x has 0 bytes per pixel.", + __FUNCTION__, __LINE__, fb_type); + return GL_FALSE; + } + + /* Valid types are GL_UNSIGNED_SHORT_5_6_5 and GL_UNSIGNED_INT_8_8_8_8 and + * the _REV versions. + * + * Valid formats are GL_RGBA, GL_RGB, and GL_BGRA. + */ + switch ( fb_format ) { + case GL_RGB: + bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[1]; + masks = masks_table_rgb[index]; + break; + + case GL_RGBA: + bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[2]; + masks = masks_table_rgba[index]; + break; + + case GL_BGR: + bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[1]; + masks = masks_table_bgr[index]; + break; + + case GL_BGRA: + bits = (bytes_per_pixel[index] == 2) ? bits_table[0] : bits_table[2]; + masks = masks_table_bgra[index]; + break; + + default: + _eglLog(_EGL_WARNING, + "[%s:%u] Framebuffer format 0x%04x is not GL_RGB, GL_RGBA, GL_BGR, or GL_BGRA.", + __FUNCTION__, __LINE__, fb_format); + return GL_FALSE; + } + + config = configs; + for (k = 0; k < num_depth_stencil_bits; k++) { + for (i = 0; i < num_db_modes; i++) { + for (j = 0; j < 2; j++) { + _eglSetConfigAttrib(config, EGL_RED_SIZE, bits[0]); + _eglSetConfigAttrib(config, EGL_GREEN_SIZE, bits[1]); + _eglSetConfigAttrib(config, EGL_BLUE_SIZE, bits[2]); + _eglSetConfigAttrib(config, EGL_ALPHA_SIZE, bits[3]); + _eglSetConfigAttrib(config, EGL_BUFFER_SIZE, + bits[0] + bits[1] + bits[2] + bits[3]); + + _eglSetConfigAttrib(config, EGL_STENCIL_SIZE, stencil_bits[k]); + _eglSetConfigAttrib(config, EGL_DEPTH_SIZE, depth_bits[i]); + + _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_SCREEN_BIT_MESA | + EGL_PBUFFER_BIT | EGL_PIXMAP_BIT | EGL_WINDOW_BIT); + + config++; + } + } + } + return GL_TRUE; +#else + return GL_FALSE; +#endif +} + diff --git a/src/egl/main/eglconfigutil.h b/src/egl/main/eglconfigutil.h new file mode 100644 index 0000000000..5db906db65 --- /dev/null +++ b/src/egl/main/eglconfigutil.h @@ -0,0 +1,23 @@ + +#ifndef EGLCONFIGUTIL_INCLUDED +#define EGLCONFIGUTIL_INCLUDED + +#include "eglconfig.h" +#include "GL/internal/glcore.h" + + +extern void +_eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode); + + +extern EGLBoolean +_eglFillInConfigs( _EGLConfig *configs, + EGLenum fb_format, EGLenum fb_type, + const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + unsigned num_depth_stencil_bits, + const EGLenum * db_modes, unsigned num_db_modes, + int visType ); + + + +#endif /* EGLCONFIGUTIL_INCLUDED */ -- cgit v1.2.3 From 96a40345bb3b7c87d9742d0b5683355f42596823 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 27 May 2008 16:58:23 -0600 Subject: include eglconfigutil.h --- src/egl/drivers/dri/egldri.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c index 81c12b7ca6..79a2eda3ae 100644 --- a/src/egl/drivers/dri/egldri.c +++ b/src/egl/drivers/dri/egldri.c @@ -23,6 +23,7 @@ #include "egldisplay.h" #include "eglcontext.h" #include "eglconfig.h" +#include "eglconfigutil.h" #include "eglsurface.h" #include "eglscreen.h" #include "eglglobals.h" -- cgit v1.2.3 From 55d29a8d48663982a1aeea414f69a5896b97d1ea Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 28 May 2008 16:12:14 +0900 Subject: gallium: Windows CE portability fixes. --- src/gallium/auxiliary/draw/draw_pt_elts.c | 8 +- src/gallium/auxiliary/draw/draw_pt_varray.c | 4 +- src/gallium/auxiliary/draw/draw_vs_sse.c | 4 +- src/gallium/auxiliary/rtasm/rtasm_cpu.c | 4 +- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 4 +- src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 4 +- src/gallium/auxiliary/tgsi/util/tgsi_util.c | 2 +- src/gallium/auxiliary/translate/translate.c | 3 +- .../auxiliary/translate/translate_generic.c | 236 ++++++++++----------- src/gallium/auxiliary/translate/translate_sse.c | 5 +- src/gallium/auxiliary/util/u_time.h | 2 +- src/gallium/include/pipe/p_compiler.h | 54 +++-- src/gallium/include/pipe/p_config.h | 8 +- src/gallium/include/pipe/p_debug.h | 11 +- 14 files changed, 192 insertions(+), 157 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_elts.c b/src/gallium/auxiliary/draw/draw_pt_elts.c index 2094c081ed..b7780fb507 100644 --- a/src/gallium/auxiliary/draw/draw_pt_elts.c +++ b/src/gallium/auxiliary/draw/draw_pt_elts.c @@ -60,10 +60,10 @@ static unsigned elt_vert( const void *elts, unsigned idx ) pt_elt_func draw_pt_elt_func( struct draw_context *draw ) { switch (draw->pt.user.eltSize) { - case 0: return elt_vert; - case 1: return elt_ubyte; - case 2: return elt_ushort; - case 4: return elt_uint; + case 0: return &elt_vert; + case 1: return &elt_ubyte; + case 2: return &elt_ushort; + case 4: return &elt_uint; default: return NULL; } } diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 355093f945..c7c66b34d4 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -147,8 +147,8 @@ static INLINE void varray_ef_quad( struct varray_frontend *varray, unsigned i2, unsigned i3 ) { - const unsigned omitEdge1 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_2; - const unsigned omitEdge2 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_1; + const ushort omitEdge1 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_2; + const ushort omitEdge2 = DRAW_PIPE_EDGE_FLAG_0 | DRAW_PIPE_EDGE_FLAG_1; varray_triangle_flags( varray, DRAW_PIPE_RESET_STIPPLE | omitEdge1, diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index e3f4e67472..c88bc137ee 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -31,9 +31,11 @@ * Brian Paul */ +#include "pipe/p_config.h" + #include "draw_vs.h" -#if defined(__i386__) || defined(__386__) +#if defined(PIPE_ARCH_X86) #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" diff --git a/src/gallium/auxiliary/rtasm/rtasm_cpu.c b/src/gallium/auxiliary/rtasm/rtasm_cpu.c index f01e12faa0..5499018b21 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_cpu.c +++ b/src/gallium/auxiliary/rtasm/rtasm_cpu.c @@ -47,7 +47,7 @@ static boolean rtasm_sse_enabled(void) int rtasm_cpu_has_sse(void) { /* FIXME: actually detect this at run-time */ -#if defined(__i386__) || defined(__386__) || defined(i386) +#if defined(PIPE_ARCH_X86) return rtasm_sse_enabled(); #else return 0; @@ -57,7 +57,7 @@ int rtasm_cpu_has_sse(void) int rtasm_cpu_has_sse2(void) { /* FIXME: actually detect this at run-time */ -#if defined(__i386__) || defined(__386__) || defined(i386) +#if defined(PIPE_ARCH_X86) return rtasm_sse_enabled(); #else return 0; diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 4e036d9032..6cd88ebca3 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -21,7 +21,9 @@ * **************************************************************************/ -#if defined(__i386__) || defined(__386__) || defined(i386) +#include "pipe/p_config.h" + +#if defined(PIPE_ARCH_X86) #include "pipe/p_compiler.h" #include "pipe/p_debug.h" diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index eacaeeaf6f..a5afa16395 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -24,7 +24,9 @@ #ifndef _RTASM_X86SSE_H_ #define _RTASM_X86SSE_H_ -#if defined(__i386__) || defined(__386__) || defined(i386) +#include "pipe/p_config.h" + +#if defined(PIPE_ARCH_X86) /* It is up to the caller to ensure that instructions issued are * suitable for the host cpu. There are no checks made in this module diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_util.c b/src/gallium/auxiliary/tgsi/util/tgsi_util.c index 4cdd89182a..56a50d3b21 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_util.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_util.c @@ -8,7 +8,7 @@ union pointer_hack { void *pointer; - unsigned long long uint64; + uint64_t uint64; }; void * diff --git a/src/gallium/auxiliary/translate/translate.c b/src/gallium/auxiliary/translate/translate.c index b04bc6eefd..b93fbf9033 100644 --- a/src/gallium/auxiliary/translate/translate.c +++ b/src/gallium/auxiliary/translate/translate.c @@ -30,6 +30,7 @@ * Keith Whitwell */ +#include "pipe/p_config.h" #include "pipe/p_util.h" #include "pipe/p_state.h" #include "translate.h" @@ -38,7 +39,7 @@ struct translate *translate_create( const struct translate_key *key ) { struct translate *translate = NULL; -#if defined(__i386__) || defined(__386__) || defined(i386) +#if defined(PIPE_ARCH_X86) translate = translate_sse2_create( key ); if (translate) return translate; diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index 402780ee53..8f3b470333 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -255,140 +255,140 @@ static fetch_func get_fetch_func( enum pipe_format format ) { switch (format) { case PIPE_FORMAT_R64_FLOAT: - return fetch_R64_FLOAT; + return &fetch_R64_FLOAT; case PIPE_FORMAT_R64G64_FLOAT: - return fetch_R64G64_FLOAT; + return &fetch_R64G64_FLOAT; case PIPE_FORMAT_R64G64B64_FLOAT: - return fetch_R64G64B64_FLOAT; + return &fetch_R64G64B64_FLOAT; case PIPE_FORMAT_R64G64B64A64_FLOAT: - return fetch_R64G64B64A64_FLOAT; + return &fetch_R64G64B64A64_FLOAT; case PIPE_FORMAT_R32_FLOAT: - return fetch_R32_FLOAT; + return &fetch_R32_FLOAT; case PIPE_FORMAT_R32G32_FLOAT: - return fetch_R32G32_FLOAT; + return &fetch_R32G32_FLOAT; case PIPE_FORMAT_R32G32B32_FLOAT: - return fetch_R32G32B32_FLOAT; + return &fetch_R32G32B32_FLOAT; case PIPE_FORMAT_R32G32B32A32_FLOAT: - return fetch_R32G32B32A32_FLOAT; + return &fetch_R32G32B32A32_FLOAT; case PIPE_FORMAT_R32_UNORM: - return fetch_R32_UNORM; + return &fetch_R32_UNORM; case PIPE_FORMAT_R32G32_UNORM: - return fetch_R32G32_UNORM; + return &fetch_R32G32_UNORM; case PIPE_FORMAT_R32G32B32_UNORM: - return fetch_R32G32B32_UNORM; + return &fetch_R32G32B32_UNORM; case PIPE_FORMAT_R32G32B32A32_UNORM: - return fetch_R32G32B32A32_UNORM; + return &fetch_R32G32B32A32_UNORM; case PIPE_FORMAT_R32_USCALED: - return fetch_R32_USCALED; + return &fetch_R32_USCALED; case PIPE_FORMAT_R32G32_USCALED: - return fetch_R32G32_USCALED; + return &fetch_R32G32_USCALED; case PIPE_FORMAT_R32G32B32_USCALED: - return fetch_R32G32B32_USCALED; + return &fetch_R32G32B32_USCALED; case PIPE_FORMAT_R32G32B32A32_USCALED: - return fetch_R32G32B32A32_USCALED; + return &fetch_R32G32B32A32_USCALED; case PIPE_FORMAT_R32_SNORM: - return fetch_R32_SNORM; + return &fetch_R32_SNORM; case PIPE_FORMAT_R32G32_SNORM: - return fetch_R32G32_SNORM; + return &fetch_R32G32_SNORM; case PIPE_FORMAT_R32G32B32_SNORM: - return fetch_R32G32B32_SNORM; + return &fetch_R32G32B32_SNORM; case PIPE_FORMAT_R32G32B32A32_SNORM: - return fetch_R32G32B32A32_SNORM; + return &fetch_R32G32B32A32_SNORM; case PIPE_FORMAT_R32_SSCALED: - return fetch_R32_SSCALED; + return &fetch_R32_SSCALED; case PIPE_FORMAT_R32G32_SSCALED: - return fetch_R32G32_SSCALED; + return &fetch_R32G32_SSCALED; case PIPE_FORMAT_R32G32B32_SSCALED: - return fetch_R32G32B32_SSCALED; + return &fetch_R32G32B32_SSCALED; case PIPE_FORMAT_R32G32B32A32_SSCALED: - return fetch_R32G32B32A32_SSCALED; + return &fetch_R32G32B32A32_SSCALED; case PIPE_FORMAT_R16_UNORM: - return fetch_R16_UNORM; + return &fetch_R16_UNORM; case PIPE_FORMAT_R16G16_UNORM: - return fetch_R16G16_UNORM; + return &fetch_R16G16_UNORM; case PIPE_FORMAT_R16G16B16_UNORM: - return fetch_R16G16B16_UNORM; + return &fetch_R16G16B16_UNORM; case PIPE_FORMAT_R16G16B16A16_UNORM: - return fetch_R16G16B16A16_UNORM; + return &fetch_R16G16B16A16_UNORM; case PIPE_FORMAT_R16_USCALED: - return fetch_R16_USCALED; + return &fetch_R16_USCALED; case PIPE_FORMAT_R16G16_USCALED: - return fetch_R16G16_USCALED; + return &fetch_R16G16_USCALED; case PIPE_FORMAT_R16G16B16_USCALED: - return fetch_R16G16B16_USCALED; + return &fetch_R16G16B16_USCALED; case PIPE_FORMAT_R16G16B16A16_USCALED: - return fetch_R16G16B16A16_USCALED; + return &fetch_R16G16B16A16_USCALED; case PIPE_FORMAT_R16_SNORM: - return fetch_R16_SNORM; + return &fetch_R16_SNORM; case PIPE_FORMAT_R16G16_SNORM: - return fetch_R16G16_SNORM; + return &fetch_R16G16_SNORM; case PIPE_FORMAT_R16G16B16_SNORM: - return fetch_R16G16B16_SNORM; + return &fetch_R16G16B16_SNORM; case PIPE_FORMAT_R16G16B16A16_SNORM: - return fetch_R16G16B16A16_SNORM; + return &fetch_R16G16B16A16_SNORM; case PIPE_FORMAT_R16_SSCALED: - return fetch_R16_SSCALED; + return &fetch_R16_SSCALED; case PIPE_FORMAT_R16G16_SSCALED: - return fetch_R16G16_SSCALED; + return &fetch_R16G16_SSCALED; case PIPE_FORMAT_R16G16B16_SSCALED: - return fetch_R16G16B16_SSCALED; + return &fetch_R16G16B16_SSCALED; case PIPE_FORMAT_R16G16B16A16_SSCALED: - return fetch_R16G16B16A16_SSCALED; + return &fetch_R16G16B16A16_SSCALED; case PIPE_FORMAT_R8_UNORM: - return fetch_R8_UNORM; + return &fetch_R8_UNORM; case PIPE_FORMAT_R8G8_UNORM: - return fetch_R8G8_UNORM; + return &fetch_R8G8_UNORM; case PIPE_FORMAT_R8G8B8_UNORM: - return fetch_R8G8B8_UNORM; + return &fetch_R8G8B8_UNORM; case PIPE_FORMAT_R8G8B8A8_UNORM: - return fetch_R8G8B8A8_UNORM; + return &fetch_R8G8B8A8_UNORM; case PIPE_FORMAT_R8_USCALED: - return fetch_R8_USCALED; + return &fetch_R8_USCALED; case PIPE_FORMAT_R8G8_USCALED: - return fetch_R8G8_USCALED; + return &fetch_R8G8_USCALED; case PIPE_FORMAT_R8G8B8_USCALED: - return fetch_R8G8B8_USCALED; + return &fetch_R8G8B8_USCALED; case PIPE_FORMAT_R8G8B8A8_USCALED: - return fetch_R8G8B8A8_USCALED; + return &fetch_R8G8B8A8_USCALED; case PIPE_FORMAT_R8_SNORM: - return fetch_R8_SNORM; + return &fetch_R8_SNORM; case PIPE_FORMAT_R8G8_SNORM: - return fetch_R8G8_SNORM; + return &fetch_R8G8_SNORM; case PIPE_FORMAT_R8G8B8_SNORM: - return fetch_R8G8B8_SNORM; + return &fetch_R8G8B8_SNORM; case PIPE_FORMAT_R8G8B8A8_SNORM: - return fetch_R8G8B8A8_SNORM; + return &fetch_R8G8B8A8_SNORM; case PIPE_FORMAT_R8_SSCALED: - return fetch_R8_SSCALED; + return &fetch_R8_SSCALED; case PIPE_FORMAT_R8G8_SSCALED: - return fetch_R8G8_SSCALED; + return &fetch_R8G8_SSCALED; case PIPE_FORMAT_R8G8B8_SSCALED: - return fetch_R8G8B8_SSCALED; + return &fetch_R8G8B8_SSCALED; case PIPE_FORMAT_R8G8B8A8_SSCALED: - return fetch_R8G8B8A8_SSCALED; + return &fetch_R8G8B8A8_SSCALED; case PIPE_FORMAT_A8R8G8B8_UNORM: - return fetch_A8R8G8B8_UNORM; + return &fetch_A8R8G8B8_UNORM; case PIPE_FORMAT_B8G8R8A8_UNORM: - return fetch_B8G8R8A8_UNORM; + return &fetch_B8G8R8A8_UNORM; default: assert(0); - return fetch_NULL; + return &fetch_NULL; } } @@ -399,140 +399,140 @@ static emit_func get_emit_func( enum pipe_format format ) { switch (format) { case PIPE_FORMAT_R64_FLOAT: - return emit_R64_FLOAT; + return &emit_R64_FLOAT; case PIPE_FORMAT_R64G64_FLOAT: - return emit_R64G64_FLOAT; + return &emit_R64G64_FLOAT; case PIPE_FORMAT_R64G64B64_FLOAT: - return emit_R64G64B64_FLOAT; + return &emit_R64G64B64_FLOAT; case PIPE_FORMAT_R64G64B64A64_FLOAT: - return emit_R64G64B64A64_FLOAT; + return &emit_R64G64B64A64_FLOAT; case PIPE_FORMAT_R32_FLOAT: - return emit_R32_FLOAT; + return &emit_R32_FLOAT; case PIPE_FORMAT_R32G32_FLOAT: - return emit_R32G32_FLOAT; + return &emit_R32G32_FLOAT; case PIPE_FORMAT_R32G32B32_FLOAT: - return emit_R32G32B32_FLOAT; + return &emit_R32G32B32_FLOAT; case PIPE_FORMAT_R32G32B32A32_FLOAT: - return emit_R32G32B32A32_FLOAT; + return &emit_R32G32B32A32_FLOAT; case PIPE_FORMAT_R32_UNORM: - return emit_R32_UNORM; + return &emit_R32_UNORM; case PIPE_FORMAT_R32G32_UNORM: - return emit_R32G32_UNORM; + return &emit_R32G32_UNORM; case PIPE_FORMAT_R32G32B32_UNORM: - return emit_R32G32B32_UNORM; + return &emit_R32G32B32_UNORM; case PIPE_FORMAT_R32G32B32A32_UNORM: - return emit_R32G32B32A32_UNORM; + return &emit_R32G32B32A32_UNORM; case PIPE_FORMAT_R32_USCALED: - return emit_R32_USCALED; + return &emit_R32_USCALED; case PIPE_FORMAT_R32G32_USCALED: - return emit_R32G32_USCALED; + return &emit_R32G32_USCALED; case PIPE_FORMAT_R32G32B32_USCALED: - return emit_R32G32B32_USCALED; + return &emit_R32G32B32_USCALED; case PIPE_FORMAT_R32G32B32A32_USCALED: - return emit_R32G32B32A32_USCALED; + return &emit_R32G32B32A32_USCALED; case PIPE_FORMAT_R32_SNORM: - return emit_R32_SNORM; + return &emit_R32_SNORM; case PIPE_FORMAT_R32G32_SNORM: - return emit_R32G32_SNORM; + return &emit_R32G32_SNORM; case PIPE_FORMAT_R32G32B32_SNORM: - return emit_R32G32B32_SNORM; + return &emit_R32G32B32_SNORM; case PIPE_FORMAT_R32G32B32A32_SNORM: - return emit_R32G32B32A32_SNORM; + return &emit_R32G32B32A32_SNORM; case PIPE_FORMAT_R32_SSCALED: - return emit_R32_SSCALED; + return &emit_R32_SSCALED; case PIPE_FORMAT_R32G32_SSCALED: - return emit_R32G32_SSCALED; + return &emit_R32G32_SSCALED; case PIPE_FORMAT_R32G32B32_SSCALED: - return emit_R32G32B32_SSCALED; + return &emit_R32G32B32_SSCALED; case PIPE_FORMAT_R32G32B32A32_SSCALED: - return emit_R32G32B32A32_SSCALED; + return &emit_R32G32B32A32_SSCALED; case PIPE_FORMAT_R16_UNORM: - return emit_R16_UNORM; + return &emit_R16_UNORM; case PIPE_FORMAT_R16G16_UNORM: - return emit_R16G16_UNORM; + return &emit_R16G16_UNORM; case PIPE_FORMAT_R16G16B16_UNORM: - return emit_R16G16B16_UNORM; + return &emit_R16G16B16_UNORM; case PIPE_FORMAT_R16G16B16A16_UNORM: - return emit_R16G16B16A16_UNORM; + return &emit_R16G16B16A16_UNORM; case PIPE_FORMAT_R16_USCALED: - return emit_R16_USCALED; + return &emit_R16_USCALED; case PIPE_FORMAT_R16G16_USCALED: - return emit_R16G16_USCALED; + return &emit_R16G16_USCALED; case PIPE_FORMAT_R16G16B16_USCALED: - return emit_R16G16B16_USCALED; + return &emit_R16G16B16_USCALED; case PIPE_FORMAT_R16G16B16A16_USCALED: - return emit_R16G16B16A16_USCALED; + return &emit_R16G16B16A16_USCALED; case PIPE_FORMAT_R16_SNORM: - return emit_R16_SNORM; + return &emit_R16_SNORM; case PIPE_FORMAT_R16G16_SNORM: - return emit_R16G16_SNORM; + return &emit_R16G16_SNORM; case PIPE_FORMAT_R16G16B16_SNORM: - return emit_R16G16B16_SNORM; + return &emit_R16G16B16_SNORM; case PIPE_FORMAT_R16G16B16A16_SNORM: - return emit_R16G16B16A16_SNORM; + return &emit_R16G16B16A16_SNORM; case PIPE_FORMAT_R16_SSCALED: - return emit_R16_SSCALED; + return &emit_R16_SSCALED; case PIPE_FORMAT_R16G16_SSCALED: - return emit_R16G16_SSCALED; + return &emit_R16G16_SSCALED; case PIPE_FORMAT_R16G16B16_SSCALED: - return emit_R16G16B16_SSCALED; + return &emit_R16G16B16_SSCALED; case PIPE_FORMAT_R16G16B16A16_SSCALED: - return emit_R16G16B16A16_SSCALED; + return &emit_R16G16B16A16_SSCALED; case PIPE_FORMAT_R8_UNORM: - return emit_R8_UNORM; + return &emit_R8_UNORM; case PIPE_FORMAT_R8G8_UNORM: - return emit_R8G8_UNORM; + return &emit_R8G8_UNORM; case PIPE_FORMAT_R8G8B8_UNORM: - return emit_R8G8B8_UNORM; + return &emit_R8G8B8_UNORM; case PIPE_FORMAT_R8G8B8A8_UNORM: - return emit_R8G8B8A8_UNORM; + return &emit_R8G8B8A8_UNORM; case PIPE_FORMAT_R8_USCALED: - return emit_R8_USCALED; + return &emit_R8_USCALED; case PIPE_FORMAT_R8G8_USCALED: - return emit_R8G8_USCALED; + return &emit_R8G8_USCALED; case PIPE_FORMAT_R8G8B8_USCALED: - return emit_R8G8B8_USCALED; + return &emit_R8G8B8_USCALED; case PIPE_FORMAT_R8G8B8A8_USCALED: - return emit_R8G8B8A8_USCALED; + return &emit_R8G8B8A8_USCALED; case PIPE_FORMAT_R8_SNORM: - return emit_R8_SNORM; + return &emit_R8_SNORM; case PIPE_FORMAT_R8G8_SNORM: - return emit_R8G8_SNORM; + return &emit_R8G8_SNORM; case PIPE_FORMAT_R8G8B8_SNORM: - return emit_R8G8B8_SNORM; + return &emit_R8G8B8_SNORM; case PIPE_FORMAT_R8G8B8A8_SNORM: - return emit_R8G8B8A8_SNORM; + return &emit_R8G8B8A8_SNORM; case PIPE_FORMAT_R8_SSCALED: - return emit_R8_SSCALED; + return &emit_R8_SSCALED; case PIPE_FORMAT_R8G8_SSCALED: - return emit_R8G8_SSCALED; + return &emit_R8G8_SSCALED; case PIPE_FORMAT_R8G8B8_SSCALED: - return emit_R8G8B8_SSCALED; + return &emit_R8G8B8_SSCALED; case PIPE_FORMAT_R8G8B8A8_SSCALED: - return emit_R8G8B8A8_SSCALED; + return &emit_R8G8B8A8_SSCALED; case PIPE_FORMAT_A8R8G8B8_UNORM: - return emit_A8R8G8B8_UNORM; + return &emit_A8R8G8B8_UNORM; case PIPE_FORMAT_B8G8R8A8_UNORM: - return emit_B8G8R8A8_UNORM; + return &emit_B8G8R8A8_UNORM; default: assert(0); - return emit_NULL; + return &emit_NULL; } } diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index a54ac5a82f..634b05b8a9 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -26,6 +26,7 @@ */ +#include "pipe/p_config.h" #include "pipe/p_compiler.h" #include "pipe/p_util.h" #include "util/u_simple_list.h" @@ -33,7 +34,7 @@ #include "translate.h" -#if defined(__i386__) || defined(__386__) || defined(i386) +#if defined(PIPE_ARCH_X86) #include "rtasm/rtasm_cpu.h" #include "rtasm/rtasm_x86sse.h" @@ -617,7 +618,7 @@ struct translate *translate_sse2_create( const struct translate_key *key ) #else -void translate_create_sse( const struct translate_key *key ) +struct translate *translate_sse2_create( const struct translate_key *key ) { return NULL; } diff --git a/src/gallium/auxiliary/util/u_time.h b/src/gallium/auxiliary/util/u_time.h index 48ec7a4a96..f9963ce0e2 100644 --- a/src/gallium/auxiliary/util/u_time.h +++ b/src/gallium/auxiliary/util/u_time.h @@ -61,7 +61,7 @@ struct util_time #if defined(PIPE_OS_LINUX) struct timeval tv; #else - long long counter; + int64_t counter; #endif }; diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index a4b772bc4f..96b21d998d 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -52,39 +52,55 @@ #endif /* __MSC__ */ -typedef unsigned int uint; -typedef unsigned char ubyte; -typedef unsigned char boolean; -typedef unsigned short ushort; -typedef unsigned long long uint64; - - #if defined(__MSC__) -typedef char int8_t; -typedef unsigned char uint8_t; -typedef short int16_t; -typedef unsigned short uint16_t; -typedef long int32_t; -typedef unsigned long uint32_t; -typedef long long int64_t; -typedef unsigned long long uint64_t; +typedef __int8 int8_t; +typedef unsigned __int8 uint8_t; +typedef __int16 int16_t; +typedef unsigned __int16 uint16_t; +typedef __int32 int32_t; +typedef unsigned __int32 uint32_t; +typedef __int64 int64_t; +typedef unsigned __int64 uint64_t; #if defined(_WIN64) typedef __int64 intptr_t; typedef unsigned __int64 uintptr_t; #else -typedef int intptr_t; -typedef unsigned int uintptr_t; +typedef __int32 intptr_t; +typedef unsigned __int32 uintptr_t; #endif +#ifndef __cplusplus +#define false 0 +#define true 1 +#define bool _Bool +typedef int _Bool; +#define __bool_true_false_are_defined 1 +#endif /* !__cplusplus */ + #else #include +#include #endif -#define TRUE 1 -#define FALSE 0 +typedef unsigned int uint; +typedef unsigned char ubyte; +typedef unsigned short ushort; +typedef uint64_t uint64; + +#if 0 +#define boolean bool +#else +typedef unsigned char boolean; +#endif +#ifndef TRUE +#define TRUE true +#endif +#ifndef FALSE +#define FALSE false +#endif /* Function inlining */ diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h index 6ba211a1fc..d2d2ae1617 100644 --- a/src/gallium/include/pipe/p_config.h +++ b/src/gallium/include/pipe/p_config.h @@ -35,6 +35,10 @@ * this file is auto-generated by an autoconf-like tool at some point, as some * things cannot be determined by existing defines alone. * + * See also: + * - http://gcc.gnu.org/onlinedocs/cpp/Common-Predefined-Macros.html + * - echo | gcc -dM -E - | sort + * - http://msdn.microsoft.com/en-us/library/b0084kay.aspx * @author José Fonseca */ @@ -63,11 +67,11 @@ * Processor architecture */ -#if defined(_X86_) || defined(__i386__) || defined(__386__) || defined(i386) +#if defined(__i386__) /* gcc */ || defined(_M_IX86) /* msvc */ || defined(_X86_) || defined(__386__) || defined(i386) #define PIPE_ARCH_X86 #endif -#if 0 /* FIXME */ +#if defined(__x86_64__) /* gcc */ || defined(_M_X64) /* msvc */ || defined(_M_AMD64) /* msvc */ #define PIPE_ARCH_X86_64 #endif diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 0af635be57..05eca75201 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -59,6 +59,13 @@ extern "C" { #endif #endif + +/* MSVC bebore VC7 does not have the __FUNCTION__ macro */ +#if defined(_MSC_VER) && _MSC_VER < 1300 +#define __FUNCTION__ "???" +#endif + + void _debug_vprintf(const char *format, va_list ap); @@ -127,8 +134,8 @@ void _debug_break(void); #ifdef DEBUG #if (defined(__i386__) || defined(__386__)) && defined(__GNUC__) #define debug_break() __asm("int3") -#elif (defined(__i386__) || defined(__386__)) && defined(__MSC__) -#define debug_break() _asm {int 3} +#elif defined(_M_IX86) && defined(_MSC_VER) +#define debug_break() do { _asm {int 3} } while(0) #else #define debug_break() _debug_break() #endif -- cgit v1.2.3 From 003afbe5bdb3d3c45e0f4f06d0b4239237df934c Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 28 May 2008 11:32:18 +0200 Subject: egl: Make demo driver compile --- src/egl/drivers/demo/demo.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/demo/demo.c b/src/egl/drivers/demo/demo.c index 0a13de6d87..1033f1b4ed 100644 --- a/src/egl/drivers/demo/demo.c +++ b/src/egl/drivers/demo/demo.c @@ -152,9 +152,9 @@ demoCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext /* generate handle and insert into hash table */ _eglSaveContext(&c->Base); - assert(c->Base.Handle); + assert(_eglGetContextHandle(&c->Base)); - return c->Base.Handle; + return _eglGetContextHandle(&c->Base); } -- cgit v1.2.3 From 68ff873c20a6e32eef0fa4c6d259a5a415308be5 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 28 May 2008 11:33:52 +0200 Subject: egl: Make dri driver compile --- src/egl/drivers/dri/egldri.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/egl/drivers/dri/egldri.h b/src/egl/drivers/dri/egldri.h index 49e1cf4d99..54a9a4ea26 100644 --- a/src/egl/drivers/dri/egldri.h +++ b/src/egl/drivers/dri/egldri.h @@ -3,6 +3,7 @@ #include #include +#include #include "egldisplay.h" #include "eglscreen.h" #include "eglsurface.h" -- cgit v1.2.3 From b870bf79b5387f26668285f44ccbf5812ad62e10 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 28 May 2008 11:40:41 +0200 Subject: egl: Temporarily disable eglx.[c|h] building --- src/egl/main/Makefile | 7 ++++--- src/egl/main/egldriver.c | 8 +++++++- 2 files changed, 11 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index 4fb0344e63..5f3d65f72a 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -18,7 +18,7 @@ HEADERS = \ eglmode.h \ eglscreen.h \ eglsurface.h \ - eglx.h +# eglx.h SOURCES = \ eglapi.c \ @@ -33,12 +33,13 @@ SOURCES = \ eglmode.c \ eglscreen.c \ eglsurface.c \ - eglx.c +# eglx.c OBJECTS = $(SOURCES:.c=.o) -LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 +# Undefined for now +#LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 .c.o: diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 50c466c258..cbdd47948d 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -16,8 +16,14 @@ #include "eglmode.h" #include "eglscreen.h" #include "eglsurface.h" -#include "eglx.h" +#if defined(_EGL_PLATFORM_X) +#include "eglx.h" +#elif defined(_EGL_PLATFORM_WINDOWS) +/* XXX to do */ +#elif defined(_EGL_PLATFORM_WINCE) +/* XXX to do */ +#endif const char *DefaultDriverName = "demodriver"; -- cgit v1.2.3 From c2bd95abf60a8b6ea83dce3fc67603b36f937484 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 28 May 2008 01:18:45 +0200 Subject: i915: Make EGL_i915 segfault if modesetting is not working --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index ac1825210e..12eeff2b23 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -114,6 +114,7 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) struct drm_screen *screen = NULL; drmModeOutputPtr output = NULL; drmModeResPtr res = NULL; + unsigned count_outputs = 0; EGLint i; int fd; @@ -131,8 +132,10 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) drm_update_res(drm_drv); res = drm_drv->res; + if (res) + count_outputs = res->count_outputs; - for(i = 0; i < res->count_outputs; i++) { + for(i = 0; i < count_outputs; i++) { output = drmModeGetOutput(fd, res->outputs[i]); if (!output) -- cgit v1.2.3 From ae3795a968f07fc150d4c34aa1a9cd067f33b914 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 28 May 2008 12:27:27 +0200 Subject: i915: Make EGL_i915 compile --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 12eeff2b23..98dbe26376 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -3,6 +3,7 @@ #include #include #include +#include #include "eglconfig.h" #include "eglcontext.h" @@ -13,7 +14,6 @@ #include "eglscreen.h" #include "eglsurface.h" -#include "glapi.h" #include "intel_egl.h" #include "xf86drm.h" @@ -283,9 +283,9 @@ drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext /* generate handle and insert into hash table */ _eglSaveContext(&c->base); - assert(c->base.Handle); + assert(_eglGetContextHandle(&c->base)); - return c->base.Handle; + return _eglGetContextHandle(&c->base); err_gl: free(context); err_c: -- cgit v1.2.3 From 5d90f97f48f8ba231d52bb1a4758dd37f81ec8d6 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 28 May 2008 12:27:52 +0200 Subject: i915: Don't segfault on buffer allocation error --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 98dbe26376..809371fdeb 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -479,12 +479,13 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, DRM_BO_FLAG_NO_EVICT, DRM_BO_HINT_DONT_FENCE, &scrn->buffer); - prettyColors(drm_drv->device->drmFD, scrn->buffer.handle, pitch); if (ret) { printf("failed to create framebuffer (ret %d)\n", ret); return EGL_FALSE; } + prettyColors(drm_drv->device->drmFD, scrn->buffer.handle, pitch); + ret = drmModeAddFB(drm_drv->device->drmFD, mode->Width, mode->Height, 32, 32, pitch, scrn->buffer.handle, -- cgit v1.2.3 From 08130512b9961da76a6385403d56387125df5e8c Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 27 May 2008 13:18:25 +0200 Subject: i915: Made vbuf work --- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 248 +++++++++++++++++++++-- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 4 +- 2 files changed, 228 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index 7fb2adbb53..81293d0d1f 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -62,8 +62,14 @@ struct i915_vbuf_render { /** Vertex size in bytes */ unsigned vertex_size; + /** Software primitive */ + unsigned prim; + /** Hardware primitive */ unsigned hwprim; + + /** Genereate a vertex list */ + unsigned fallback; }; @@ -95,8 +101,8 @@ i915_vbuf_render_get_vertex_info( struct vbuf_render *render ) static void * i915_vbuf_render_allocate_vertices( struct vbuf_render *render, - ushort vertex_size, - ushort nr_vertices ) + ushort vertex_size, + ushort nr_vertices ) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; @@ -107,9 +113,9 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, assert(!i915->vbo); i915->vbo = winsys->buffer_create(winsys, 64, I915_BUFFER_USAGE_LIT_VERTEX, size); - + i915->dirty |= I915_NEW_VBO; - + return winsys->buffer_map(winsys, i915->vbo, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -121,16 +127,36 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render, unsigned prim ) { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); - + i915_render->prim = prim; + switch(prim) { case PIPE_PRIM_POINTS: i915_render->hwprim = PRIM3D_POINTLIST; + i915_render->fallback = 0; return TRUE; case PIPE_PRIM_LINES: i915_render->hwprim = PRIM3D_LINELIST; + i915_render->fallback = 0; + return TRUE; + case PIPE_PRIM_LINE_STRIP: + i915_render->hwprim = PRIM3D_LINESTRIP; + i915_render->fallback = 0; return TRUE; case PIPE_PRIM_TRIANGLES: i915_render->hwprim = PRIM3D_TRILIST; + i915_render->fallback = 0; + return TRUE; + case PIPE_PRIM_TRIANGLE_STRIP: + i915_render->hwprim = PRIM3D_TRISTRIP; + i915_render->fallback = 0; + return TRUE; + case PIPE_PRIM_QUADS: + i915_render->hwprim = PRIM3D_TRILIST; + i915_render->fallback = PIPE_PRIM_QUADS; + return TRUE; + case PIPE_PRIM_QUAD_STRIP: + i915_render->hwprim = PRIM3D_TRILIST; + i915_render->fallback = PIPE_PRIM_QUAD_STRIP; return TRUE; default: /* Actually, can handle a lot more just fine... Fixme. @@ -140,6 +166,179 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render, } + +/** + * Used for fallbacks in draw_arrays + */ +static void +draw_arrays_generate_indices( struct vbuf_render *render, + unsigned start, uint nr, + unsigned type ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + unsigned i; + unsigned end = start + nr; + switch(type) { + case 0: + for (i = start; i+1 < end; i += 2) + OUT_BATCH( (i+0) | (i+1) << 16 ); + if (i < end) + OUT_BATCH( i ); + break; + case PIPE_PRIM_QUADS: + for (i = start; i + 3 < end; i += 4) { + OUT_BATCH( (i+0) | (i+1) << 16 ); + OUT_BATCH( (i+3) | (i+1) << 16 ); + OUT_BATCH( (i+2) | (i+3) << 16 ); + } + break; + case PIPE_PRIM_QUAD_STRIP: + for (i = start; i + 3 < end; i += 2) { + OUT_BATCH( (i+0) | (i+1) << 16 ); + OUT_BATCH( (i+3) | (i+2) << 16 ); + OUT_BATCH( (i+0) | (i+3) << 16 ); + } + break; + default: + assert(0); + } +} + +static unsigned +draw_arrays_calc_nr_indices( uint nr, unsigned type ) +{ + switch (type) { + case 0: + return nr; + case PIPE_PRIM_QUADS: + return (nr / 4) * 6; + case PIPE_PRIM_QUAD_STRIP: + return ((nr - 2) / 2) * 6; + default: + assert(0); + return 0; + } +} + +static void +draw_arrays_fallback( struct vbuf_render *render, + unsigned start, + uint nr ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + struct pipe_winsys *winsys = i915->pipe.winsys; + unsigned nr_indices; + + winsys->buffer_unmap( winsys, i915->vbo ); + if (i915->dirty) + i915_update_derived( i915 ); + + if (i915->hardware_dirty) + i915_emit_hardware_state( i915 ); + + nr_indices = draw_arrays_calc_nr_indices( nr, i915_render->fallback ); + + if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { + FLUSH_BATCH(NULL); + + /* Make sure state is re-emitted after a flush: + */ + i915_update_derived( i915 ); + i915_emit_hardware_state( i915 ); + + if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { + assert(0); + goto out; + } + } + OUT_BATCH( _3DPRIMITIVE | + PRIM_INDIRECT | + i915_render->hwprim | + PRIM_INDIRECT_ELTS | + nr_indices ); + + draw_arrays_generate_indices( render, start, nr, i915_render->fallback ); +out: + winsys->buffer_map( winsys, i915->vbo, PIPE_BUFFER_USAGE_CPU_WRITE ); +} + +static void +i915_vbuf_render_draw_arrays( struct vbuf_render *render, + unsigned start, + uint nr ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + + if (i915_render->fallback) { + draw_arrays_fallback( render, start, nr ); + return; + } + + /* JB: TODO submit direct cmds */ + draw_arrays_fallback( render, start, nr ); +} + +/** + * Used for normal and fallback emitting of indices + * If type is zero normal operation assumed. + */ +static void +draw_generate_indices( struct vbuf_render *render, + const ushort *indices, + uint nr_indices, + unsigned type ) +{ + struct i915_vbuf_render *i915_render = i915_vbuf_render(render); + struct i915_context *i915 = i915_render->i915; + unsigned i; + + switch(type) { + case 0: + for (i = 0; i + 1 < nr_indices; i += 2) { + OUT_BATCH( (indices[i] & 0x0FFF) | ((indices[i+1] & 0x0FFF) << 16) ); + } + if (i < nr_indices) { + OUT_BATCH( indices[i] & 0x0FFF ); + } + break; + case PIPE_PRIM_QUADS: + for (i = 0; i + 3 < nr_indices; i += 4) { + OUT_BATCH( (indices[i+0] & 0x0FFF) | (indices[i+1] & 0x0FFF) << 16 ); + OUT_BATCH( (indices[i+3] & 0x0FFF) | (indices[i+1] & 0x0FFF) << 16 ); + OUT_BATCH( (indices[i+2] & 0x0FFF) | (indices[i+3] & 0x0FFF) << 16 ); + } + break; + case PIPE_PRIM_QUAD_STRIP: + for (i = 0; i + 3 < nr_indices; i += 2) { + OUT_BATCH( (indices[i+0] & 0x0FFF) | (indices[i+1] & 0x0FFF) << 16 ); + OUT_BATCH( (indices[i+3] & 0x0FFF) | (indices[i+2] & 0x0FFF) << 16 ); + OUT_BATCH( (indices[i+0] & 0x0FFF) | (indices[i+3] & 0x0FFF) << 16 ); + } + break; + default: + assert(0); + break; + } +} + +static unsigned +draw_calc_nr_indices( uint nr_indices, unsigned type ) +{ + switch (type) { + case 0: + return nr_indices; + case PIPE_PRIM_QUADS: + return (nr_indices / 4) * 6; + case PIPE_PRIM_QUAD_STRIP: + return ((nr_indices - 2) / 2) * 6; + default: + assert(0); + return 0; + } +} + static void i915_vbuf_render_draw( struct vbuf_render *render, const ushort *indices, @@ -147,13 +346,15 @@ i915_vbuf_render_draw( struct vbuf_render *render, { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; - unsigned i; + struct pipe_winsys *winsys = i915->pipe.winsys; + unsigned save_nr_indices; + + save_nr_indices = nr_indices; + nr_indices = draw_calc_nr_indices( nr_indices, i915_render->fallback ); assert(nr_indices); + winsys->buffer_unmap( winsys, i915->vbo ); - /* this seems to be bogus, since we validate state right after this */ - /*assert((i915->dirty & ~I915_NEW_VBO) == 0);*/ - if (i915->dirty) i915_update_derived( i915 ); @@ -170,22 +371,23 @@ i915_vbuf_render_draw( struct vbuf_render *render, if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { assert(0); - return; + goto out; } } OUT_BATCH( _3DPRIMITIVE | - PRIM_INDIRECT | - i915_render->hwprim | - PRIM_INDIRECT_ELTS | - nr_indices ); - for (i = 0; i + 1 < nr_indices; i += 2) { - OUT_BATCH( indices[i] | - (indices[i + 1] << 16) ); - } - if (i < nr_indices) { - OUT_BATCH( indices[i] ); - } + PRIM_INDIRECT | + i915_render->hwprim | + PRIM_INDIRECT_ELTS | + nr_indices ); + draw_generate_indices( render, + indices, + save_nr_indices, + i915_render->fallback ); + +out: + winsys->buffer_map( winsys, i915->vbo, PIPE_BUFFER_USAGE_CPU_WRITE ); + return; } @@ -234,6 +436,7 @@ i915_vbuf_render_create( struct i915_context *i915 ) i915_render->base.allocate_vertices = i915_vbuf_render_allocate_vertices; i915_render->base.set_primitive = i915_vbuf_render_set_primitive; i915_render->base.draw = i915_vbuf_render_draw; + i915_render->base.draw_arrays = i915_vbuf_render_draw_arrays; i915_render->base.release_vertices = i915_vbuf_render_release_vertices; i915_render->base.destroy = i915_vbuf_render_destroy; @@ -258,6 +461,7 @@ struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) render->destroy(render); return NULL; } - + draw_set_render(i915->draw, render); + return stage; } diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 10e1a84cc1..059b16be3b 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -328,8 +328,8 @@ intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | DRM_BO_FLAG_MEM_TT, - 32 * 4096, - 1, 40, 32 * 4096 * 2, 0, + 128, + 6, 120, 32 * 4096, 0, fMan); } -- cgit v1.2.3 From cb5e05d99c40d4f7ab1ecbb42a6390caf3966ba4 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 27 May 2008 19:00:16 +0200 Subject: i915: Made vertex submission eaven faster --- src/gallium/drivers/i915simple/i915_context.h | 2 + src/gallium/drivers/i915simple/i915_flush.c | 1 + src/gallium/drivers/i915simple/i915_prim_vbuf.c | 66 +++++++++++++++++----- .../drivers/i915simple/i915_state_immediate.c | 2 +- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 8 +-- 5 files changed, 61 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 53fc5ed079..2da90ae49d 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -245,6 +245,8 @@ struct i915_context /** Vertex buffer */ struct pipe_buffer *vbo; + size_t vbo_offset; + unsigned vbo_flushed; struct i915_state current; unsigned hardware_dirty; diff --git a/src/gallium/drivers/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c index 7d23e6b6b9..4c4718d68e 100644 --- a/src/gallium/drivers/i915simple/i915_flush.c +++ b/src/gallium/drivers/i915simple/i915_flush.c @@ -68,6 +68,7 @@ static void i915_flush( struct pipe_context *pipe, /* If there are no flags, just flush pending commands to hardware: */ FLUSH_BATCH(fence); + i915->vbo_flushed = 1; } diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index 81293d0d1f..4f36c2a22a 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -70,6 +70,13 @@ struct i915_vbuf_render { /** Genereate a vertex list */ unsigned fallback; + + /* Stuff for the vbo */ + struct pipe_buffer *vbo; + size_t vbo_size; + size_t vbo_offset; + void *vbo_ptr; + size_t vbo_alloc_size; }; @@ -111,14 +118,31 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, /* FIXME: handle failure */ assert(!i915->vbo); - i915->vbo = winsys->buffer_create(winsys, 64, I915_BUFFER_USAGE_LIT_VERTEX, - size); + if (i915_render->vbo_size > size + i915_render->vbo_offset && !i915->vbo_flushed) { + } else { + i915->vbo_flushed = 0; + pipe_buffer_reference(winsys, &i915_render->vbo, NULL); + } + + if (!i915_render->vbo) { + i915_render->vbo_size = MAX2(size, i915_render->vbo_alloc_size); + i915_render->vbo_offset = 0; + i915_render->vbo = winsys->buffer_create(winsys, + 64, + I915_BUFFER_USAGE_LIT_VERTEX, + i915_render->vbo_size); + i915_render->vbo_ptr = winsys->buffer_map(winsys, + i915_render->vbo, + PIPE_BUFFER_USAGE_CPU_WRITE); + winsys->buffer_unmap(winsys, i915_render->vbo); + } + + i915->vbo = i915_render->vbo; + i915->vbo_offset = i915_render->vbo_offset; i915->dirty |= I915_NEW_VBO; - return winsys->buffer_map(winsys, - i915->vbo, - PIPE_BUFFER_USAGE_CPU_WRITE); + return i915_render->vbo_ptr + i915->vbo_offset; } @@ -231,7 +255,6 @@ draw_arrays_fallback( struct vbuf_render *render, struct pipe_winsys *winsys = i915->pipe.winsys; unsigned nr_indices; - winsys->buffer_unmap( winsys, i915->vbo ); if (i915->dirty) i915_update_derived( i915 ); @@ -247,6 +270,7 @@ draw_arrays_fallback( struct vbuf_render *render, */ i915_update_derived( i915 ); i915_emit_hardware_state( i915 ); + i915->vbo_flushed = 1; if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { assert(0); @@ -260,8 +284,9 @@ draw_arrays_fallback( struct vbuf_render *render, nr_indices ); draw_arrays_generate_indices( render, start, nr, i915_render->fallback ); + out: - winsys->buffer_map( winsys, i915->vbo, PIPE_BUFFER_USAGE_CPU_WRITE ); + return; } static void @@ -353,7 +378,6 @@ i915_vbuf_render_draw( struct vbuf_render *render, nr_indices = draw_calc_nr_indices( nr_indices, i915_render->fallback ); assert(nr_indices); - winsys->buffer_unmap( winsys, i915->vbo ); if (i915->dirty) i915_update_derived( i915 ); @@ -368,6 +392,7 @@ i915_vbuf_render_draw( struct vbuf_render *render, */ i915_update_derived( i915 ); i915_emit_hardware_state( i915 ); + i915->vbo_flushed = 1; if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { assert(0); @@ -386,7 +411,6 @@ i915_vbuf_render_draw( struct vbuf_render *render, i915_render->fallback ); out: - winsys->buffer_map( winsys, i915->vbo, PIPE_BUFFER_USAGE_CPU_WRITE ); return; } @@ -400,10 +424,13 @@ i915_vbuf_render_release_vertices( struct vbuf_render *render, struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; struct pipe_winsys *winsys = i915->pipe.winsys; + size_t size = (size_t)vertex_size * (size_t)vertices_used; assert(i915->vbo); - winsys->buffer_unmap(winsys, i915->vbo); - pipe_buffer_reference(winsys, &i915->vbo, NULL); + + i915_render->vbo_offset += size; + i915->vbo = NULL; + i915->dirty |= I915_NEW_VBO; } @@ -422,6 +449,7 @@ static struct vbuf_render * i915_vbuf_render_create( struct i915_context *i915 ) { struct i915_vbuf_render *i915_render = CALLOC_STRUCT(i915_vbuf_render); + struct pipe_winsys *winsys = i915->pipe.winsys; i915_render->i915 = i915; @@ -431,7 +459,7 @@ i915_vbuf_render_create( struct i915_context *i915 ) * batch buffer. */ i915_render->base.max_indices = 16*1024; - + i915_render->base.get_vertex_info = i915_vbuf_render_get_vertex_info; i915_render->base.allocate_vertices = i915_vbuf_render_allocate_vertices; i915_render->base.set_primitive = i915_vbuf_render_set_primitive; @@ -439,7 +467,19 @@ i915_vbuf_render_create( struct i915_context *i915 ) i915_render->base.draw_arrays = i915_vbuf_render_draw_arrays; i915_render->base.release_vertices = i915_vbuf_render_release_vertices; i915_render->base.destroy = i915_vbuf_render_destroy; - + + i915_render->vbo_alloc_size = 128 * 4096; + i915_render->vbo_size = i915_render->vbo_alloc_size; + i915_render->vbo_offset = 0; + i915_render->vbo = winsys->buffer_create(winsys, + 64, + I915_BUFFER_USAGE_LIT_VERTEX, + i915_render->vbo_size); + i915_render->vbo_ptr = winsys->buffer_map(winsys, + i915_render->vbo, + PIPE_BUFFER_USAGE_CPU_WRITE); + winsys->buffer_unmap(winsys, i915_render->vbo); + return &i915_render->base; } diff --git a/src/gallium/drivers/i915simple/i915_state_immediate.c b/src/gallium/drivers/i915simple/i915_state_immediate.c index dfbbcab624..704ea4d838 100644 --- a/src/gallium/drivers/i915simple/i915_state_immediate.c +++ b/src/gallium/drivers/i915simple/i915_state_immediate.c @@ -54,7 +54,7 @@ static void upload_S0S1(struct i915_context *i915) /* INTEL_NEW_VBO */ /* TODO: re-use vertex buffers here? */ - LIS0 = 0; + LIS0 = i915->vbo_offset; /* INTEL_NEW_VERTEX_SIZE -- do this where the vertex size is calculated! */ diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 059b16be3b..fb8f44c845 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -224,7 +224,6 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, unsigned tex_usage) { const unsigned alignment = 64; - //int ret; surf->width = width; surf->height = height; @@ -235,7 +234,8 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, assert(!surf->buffer); surf->buffer = winsys->buffer_create(winsys, alignment, PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * height); + surf->pitch * surf->cpp * surf->height); + if(!surf->buffer) return -1; @@ -328,8 +328,8 @@ intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | DRM_BO_FLAG_MEM_TT, - 128, - 6, 120, 32 * 4096, 0, + 128 * 4096, + 1, 120, 128 * 4096 * 4, 0, fMan); } -- cgit v1.2.3 From ce56bcb640072496809d2aefede5a32fe9256ace Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 27 May 2008 21:15:36 +0200 Subject: i915: Prepare for tiled private front and back buffers --- src/gallium/drivers/i915simple/i915_state_emit.c | 1 + src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 26 ++++++++++++++++++++++-- 2 files changed, 25 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index 6f947d4346..8bcc26ad35 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -217,6 +217,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_BATCH(BUF_3D_ID_COLOR_BACK | BUF_3D_PITCH(pitch) | /* pitch in bytes */ +// BUF_3D_TILED_SURFACE); /* JB: Used to force tileing */ BUF_3D_USE_FENCE); OUT_RELOC(cbuf_surface->buffer, diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index fb8f44c845..2e6c452cd1 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -212,6 +212,15 @@ round_up(unsigned n, unsigned multiple) return (n + multiple - 1) & ~(multiple - 1); } +static unsigned +power_of_two(unsigned x) +{ + int value = 1; + while (value <= x) + value = value << 1; + return value; +} + /*pipe_buffer_reference(winsys, &i915_render->vbo, NULL);*/ /** * Copied from xm_winsys.c */ @@ -224,18 +233,31 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, unsigned tex_usage) { const unsigned alignment = 64; + assert(!surf->buffer); +#if 0 + surf->width = width; + surf->height = round_up(height, 8); + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = power_of_two(MAX2(width * surf->cpp, 512)); + surf->buffer = winsys->buffer_create(winsys, 2*4096, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->height); + + surf->pitch = surf->pitch / surf->cpp; + surf->height = height; +#else surf->width = width; surf->height = height; surf->format = format; surf->cpp = pf_get_size(format); surf->pitch = round_up(width, alignment / surf->cpp); - assert(!surf->buffer); surf->buffer = winsys->buffer_create(winsys, alignment, PIPE_BUFFER_USAGE_PIXEL, surf->pitch * surf->cpp * surf->height); - +#endif if(!surf->buffer) return -1; -- cgit v1.2.3 From 4b1f382e13ba2d8ad08bdf8dac1738ad1d22acf3 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 28 May 2008 14:22:07 +0200 Subject: i915: Remove workaround for buggy draw module --- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index 4f36c2a22a..faf5971abb 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -322,24 +322,24 @@ draw_generate_indices( struct vbuf_render *render, switch(type) { case 0: for (i = 0; i + 1 < nr_indices; i += 2) { - OUT_BATCH( (indices[i] & 0x0FFF) | ((indices[i+1] & 0x0FFF) << 16) ); + OUT_BATCH( indices[i] | indices[i+1] << 16 ); } if (i < nr_indices) { - OUT_BATCH( indices[i] & 0x0FFF ); + OUT_BATCH( indices[i] ); } break; case PIPE_PRIM_QUADS: for (i = 0; i + 3 < nr_indices; i += 4) { - OUT_BATCH( (indices[i+0] & 0x0FFF) | (indices[i+1] & 0x0FFF) << 16 ); - OUT_BATCH( (indices[i+3] & 0x0FFF) | (indices[i+1] & 0x0FFF) << 16 ); - OUT_BATCH( (indices[i+2] & 0x0FFF) | (indices[i+3] & 0x0FFF) << 16 ); + OUT_BATCH( indices[i+0] | indices[i+1] << 16 ); + OUT_BATCH( indices[i+3] | indices[i+1] << 16 ); + OUT_BATCH( indices[i+2] | indices[i+3] << 16 ); } break; case PIPE_PRIM_QUAD_STRIP: for (i = 0; i + 3 < nr_indices; i += 2) { - OUT_BATCH( (indices[i+0] & 0x0FFF) | (indices[i+1] & 0x0FFF) << 16 ); - OUT_BATCH( (indices[i+3] & 0x0FFF) | (indices[i+2] & 0x0FFF) << 16 ); - OUT_BATCH( (indices[i+0] & 0x0FFF) | (indices[i+3] & 0x0FFF) << 16 ); + OUT_BATCH( indices[i+0] | indices[i+1] << 16 ); + OUT_BATCH( indices[i+3] | indices[i+2] << 16 ); + OUT_BATCH( indices[i+0] | indices[i+3] << 16 ); } break; default: -- cgit v1.2.3 From 2c004a4bece706c590f5ab03432d2aa65d6d0263 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 28 May 2008 14:22:50 +0200 Subject: i915: DriConfigOptions started complaining on this line --- src/gallium/winsys/dri/intel/intel_screen.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 58bebbe972..429a2cf8f8 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -50,11 +50,11 @@ PUBLIC const char __driConfigOptions[] = DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY - DRI_CONF_FORCE_S3TC_ENABLE(false) +// DRI_CONF_FORCE_S3TC_ENABLE(false) DRI_CONF_ALLOW_LARGE_TEXTURES(1) DRI_CONF_SECTION_END DRI_CONF_END; -const uint __driNConfigOptions = 4; +const uint __driNConfigOptions = 3; #ifdef USE_NEW_INTERFACE static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; -- cgit v1.2.3 From 4767c10cbb87a77e6cb24c53815c38bd5887c771 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 28 May 2008 21:47:25 +0900 Subject: translate: Mark functions as PIPE_CDECL. --- src/gallium/auxiliary/translate/translate_sse.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index 2fc8b9d3d0..69dd4d3dff 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -534,7 +534,7 @@ static void translate_sse_release( struct translate *translate ) FREE(p); } -static void translate_sse_run_elts( struct translate *translate, +static void PIPE_CDECL translate_sse_run_elts( struct translate *translate, const unsigned *elts, unsigned count, void *output_buffer ) @@ -547,7 +547,7 @@ static void translate_sse_run_elts( struct translate *translate, output_buffer ); } -static void translate_sse_run( struct translate *translate, +static void PIPE_CDECL translate_sse_run( struct translate *translate, unsigned start, unsigned count, void *output_buffer ) -- cgit v1.2.3 From 7a986792dabe6556c63b2f2a997c7c6217604e2d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 28 May 2008 21:48:30 +0900 Subject: tgsi: Observe constness. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 648afa2a51..b018ea9fa1 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -754,7 +754,7 @@ tgsi_dump_instruction( } for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + const struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; if( !first_reg ) { CHR( ',' ); @@ -812,7 +812,7 @@ tgsi_dump_instruction( } for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + const struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; if( !first_reg ) { CHR( ',' ); -- cgit v1.2.3 From 276552c0dd8d9d68b8324c42b05c768c45a9db76 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 28 May 2008 14:52:10 +0200 Subject: i915: Add draw_flush to state changes --- src/gallium/drivers/i915simple/i915_state.c | 20 +++++++++++++++++++- 1 file changed, 19 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index e6c4671700..39d5d5e9d6 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -41,7 +41,6 @@ #include "i915_state_inlines.h" #include "i915_fpc.h" - /* The i915 (and related graphics cores) do not support GL_CLAMP. The * Intel drivers for "other operating systems" implement GL_CLAMP as * GL_CLAMP_TO_EDGE, so the same is done here. @@ -178,6 +177,7 @@ static void i915_bind_blend_state(struct pipe_context *pipe, void *blend) { struct i915_context *i915 = i915_context(pipe); + draw_flush(i915->draw); i915->blend = (struct i915_blend_state*)blend; @@ -194,6 +194,7 @@ static void i915_set_blend_color( struct pipe_context *pipe, const struct pipe_blend_color *blend_color ) { struct i915_context *i915 = i915_context(pipe); + draw_flush(i915->draw); i915->blend_color = *blend_color; @@ -280,6 +281,8 @@ static void i915_bind_sampler_states(struct pipe_context *pipe, !memcmp(i915->sampler, sampler, num * sizeof(void *))) return; + draw_flush(i915->draw); + for (i = 0; i < num; ++i) i915->sampler[i] = sampler[i]; for (i = num; i < PIPE_MAX_SAMPLERS; ++i) @@ -398,6 +401,7 @@ static void i915_bind_depth_stencil_state(struct pipe_context *pipe, void *depth_stencil) { struct i915_context *i915 = i915_context(pipe); + draw_flush(i915->draw); i915->depth_stencil = (const struct i915_depth_stencil_state *)depth_stencil; @@ -415,6 +419,7 @@ static void i915_set_scissor_state( struct pipe_context *pipe, const struct pipe_scissor_state *scissor ) { struct i915_context *i915 = i915_context(pipe); + draw_flush(i915->draw); memcpy( &i915->scissor, scissor, sizeof(*scissor) ); i915->dirty |= I915_NEW_SCISSOR; @@ -451,6 +456,7 @@ static void i915_bind_fs_state(struct pipe_context *pipe, void *shader) { struct i915_context *i915 = i915_context(pipe); + draw_flush(i915->draw); i915->fs = (struct i915_fragment_shader*) shader; @@ -506,6 +512,7 @@ static void i915_set_constant_buffer(struct pipe_context *pipe, { struct i915_context *i915 = i915_context(pipe); struct pipe_winsys *ws = pipe->winsys; + draw_flush(i915->draw); assert(shader < PIPE_SHADER_TYPES); assert(index == 0); @@ -574,6 +581,7 @@ static void i915_set_framebuffer_state(struct pipe_context *pipe, const struct pipe_framebuffer_state *fb) { struct i915_context *i915 = i915_context(pipe); + draw_flush(i915->draw); i915->framebuffer = *fb; /* struct copy */ @@ -586,6 +594,7 @@ static void i915_set_clip_state( struct pipe_context *pipe, const struct pipe_clip_state *clip ) { struct i915_context *i915 = i915_context(pipe); + draw_flush(i915->draw); draw_set_clip_state(i915->draw, clip); @@ -698,6 +707,10 @@ static void i915_set_vertex_buffers(struct pipe_context *pipe, const struct pipe_vertex_buffer *buffers) { struct i915_context *i915 = i915_context(pipe); + /* Because we change state before the draw_set_vertex_buffers call + * we need a flush here, just to be sure. + */ + draw_flush(i915->draw); memcpy(i915->vertex_buffer, buffers, count * sizeof(buffers[0])); i915->num_vertex_buffers = count; @@ -711,6 +724,11 @@ static void i915_set_vertex_elements(struct pipe_context *pipe, const struct pipe_vertex_element *elements) { struct i915_context *i915 = i915_context(pipe); + /* Because we change state before the draw_set_vertex_buffers call + * we need a flush here, just to be sure. + */ + draw_flush(i915->draw); + i915->num_vertex_elements = count; /* pass-through to draw module */ draw_set_vertex_elements(i915->draw, count, elements); -- cgit v1.2.3 From ff44dd5cde47c81f90c5293b904e1c7edd006be5 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 28 May 2008 15:10:19 +0200 Subject: draw: Decorate callbacks with PIPE_CDECL. --- src/gallium/auxiliary/draw/draw_vs_varient.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index 119a3a04b5..784ae41205 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -130,10 +130,10 @@ static void do_viewport( struct draw_vs_varient_generic *vsvg, } -static void vsvg_run_elts( struct draw_vs_varient *varient, - const unsigned *elts, - unsigned count, - void *output_buffer) +static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, + const unsigned *elts, + unsigned count, + void *output_buffer ) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; @@ -186,10 +186,10 @@ static void vsvg_run_elts( struct draw_vs_varient *varient, } -static void vsvg_run_linear( struct draw_vs_varient *varient, - unsigned start, - unsigned count, - void *output_buffer ) +static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, + unsigned start, + unsigned count, + void *output_buffer ) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; -- cgit v1.2.3 From e407e83966f2d7d6d9751fc0069ebacd4808d89d Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 28 May 2008 15:10:39 +0200 Subject: scons: List missing files for draw module. --- src/gallium/auxiliary/draw/SConscript | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 3b5d5ed492..96f1ae248f 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -15,7 +15,7 @@ draw = env.ConvenienceLibrary( 'draw_pipe_stipple.c', 'draw_pipe_twoside.c', 'draw_pipe_unfilled.c', - 'draw_pipe_util.c', + 'draw_pipe_util.c', 'draw_pipe_validate.c', 'draw_pipe_vbuf.c', 'draw_pipe_wide_line.c', @@ -25,8 +25,10 @@ draw = env.ConvenienceLibrary( 'draw_pt_emit.c', 'draw_pt_fetch.c', 'draw_pt_fetch_emit.c', + 'draw_pt_fetch_shade_emit.c', 'draw_pt_fetch_shade_pipeline.c', 'draw_pt_post_vs.c', + 'draw_pt_util.c', 'draw_pt_varray.c', 'draw_pt_vcache.c', 'draw_vertex.c', @@ -34,6 +36,7 @@ draw = env.ConvenienceLibrary( 'draw_vs_exec.c', 'draw_vs_llvm.c', 'draw_vs_sse.c', + 'draw_vs_varient.c' ]) auxiliaries.insert(0, draw) -- cgit v1.2.3 From 25cec212d05b613a83eb2bc2167e3dab88da0967 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 28 May 2008 22:19:26 +0900 Subject: scons: Add new files. --- src/gallium/auxiliary/draw/SConscript | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 96f1ae248f..26919a2298 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -31,8 +31,11 @@ draw = env.ConvenienceLibrary( 'draw_pt_util.c', 'draw_pt_varray.c', 'draw_pt_vcache.c', + 'draw_pt_util.c', 'draw_vertex.c', 'draw_vs.c', + 'draw_vs_aos.c', + 'draw_vs_aos_io.c', 'draw_vs_exec.c', 'draw_vs_llvm.c', 'draw_vs_sse.c', -- cgit v1.2.3 From 364f75d9dbc2b8e19c884b9cc74676ab6cbee60d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 28 May 2008 22:19:58 +0900 Subject: draw: Use PIPE_CDECL. --- src/gallium/auxiliary/draw/draw_vs_aos.c | 20 ++++++++++---------- src/gallium/auxiliary/draw/draw_vs_aos.h | 2 +- 2 files changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index d3770b2c53..0cd82ff599 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1189,7 +1189,7 @@ static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } -static PIPE_CDECL void do_lit( struct aos_machine *machine, +static void PIPE_CDECL do_lit( struct aos_machine *machine, float *result, const float *in, unsigned count ) @@ -1223,7 +1223,7 @@ static PIPE_CDECL void do_lit( struct aos_machine *machine, } -static PIPE_CDECL void do_lit_lut( struct aos_machine *machine, +static void PIPE_CDECL do_lit_lut( struct aos_machine *machine, float *result, const float *in, unsigned count ) @@ -2119,10 +2119,10 @@ static void vaos_destroy( struct draw_vs_varient *varient ) FREE(vaos); } -static void vaos_run_elts( struct draw_vs_varient *varient, - const unsigned *elts, - unsigned count, - void *output_buffer ) +static void PIPE_CDECL vaos_run_elts( struct draw_vs_varient *varient, + const unsigned *elts, + unsigned count, + void *output_buffer ) { struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; @@ -2133,10 +2133,10 @@ static void vaos_run_elts( struct draw_vs_varient *varient, output_buffer ); } -static void vaos_run_linear( struct draw_vs_varient *varient, - unsigned start, - unsigned count, - void *output_buffer ) +static void PIPE_CDECL vaos_run_linear( struct draw_vs_varient *varient, + unsigned start, + unsigned count, + void *output_buffer ) { struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index b47413ff43..837b32794f 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -60,7 +60,7 @@ struct x86_function; #define FPU_RND_NEAREST 2 struct aos_machine; -typedef void PIPE_CDECL (*lit_func)( struct aos_machine *, +typedef void (PIPE_CDECL *lit_func)( struct aos_machine *, float *result, const float *in, unsigned count ); -- cgit v1.2.3 From 95c917f7dd6e837cb1b280be617a795dd57c9f13 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 08:42:13 -0600 Subject: x-related EGL code (stub) --- src/egl/main/eglx.c | 37 +++++++++++++++++++++++++++++++++++++ src/egl/main/eglx.h | 12 ++++++++++++ 2 files changed, 49 insertions(+) create mode 100644 src/egl/main/eglx.c create mode 100644 src/egl/main/eglx.h (limited to 'src') diff --git a/src/egl/main/eglx.c b/src/egl/main/eglx.c new file mode 100644 index 0000000000..4b33080f1a --- /dev/null +++ b/src/egl/main/eglx.c @@ -0,0 +1,37 @@ + +/** + * X-specific EGL code. + * + * Any glue code needed to make EGL work with X is placed in this file. + */ + + +#include +#include +#include +#include "eglx.h" + + + +/** + * Given an X Display ptr (at dpy->Xdpy) try to determine the appropriate + * device driver. Return its name. + */ +const char * +_xeglChooseDriver(_EGLDisplay *dpy) +{ +#ifdef _EGL_PLATFORM_X + _XPrivDisplay xdpy = (_XPrivDisplay) dpy->Xdpy; + + assert(dpy); + assert(dpy->Xdpy); + + printf("%s\n", xdpy->display_name); + + return "foo"; /* XXX todo */ +#else + return NULL; +#endif +} + + diff --git a/src/egl/main/eglx.h b/src/egl/main/eglx.h new file mode 100644 index 0000000000..4323d55838 --- /dev/null +++ b/src/egl/main/eglx.h @@ -0,0 +1,12 @@ +#ifndef EGLX_INCLUDED +#define EGLX_INCLUDED + + +#include "egldisplay.h" + + +extern const char * +_xeglChooseDriver(_EGLDisplay *dpy); + + +#endif /* EGLX_INCLUDED */ -- cgit v1.2.3 From 648da5158e5f418bf859aee6aa4532b6899b0d94 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 28 May 2008 16:36:45 +0100 Subject: rtasm: special case for [ebp] --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 2415b0156b..672d2ff554 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -328,7 +328,7 @@ struct x86_reg x86_make_disp( struct x86_reg reg, else reg.disp += disp; - if (reg.disp == 0) + if (reg.disp == 0 && reg.idx != reg_BP) reg.mod = mod_INDIRECT; else if (reg.disp <= 127 && reg.disp >= -128) reg.mod = mod_DISP8; -- cgit v1.2.3 From aa4ca9119d1dbaf542cf5971f845b434ef008aa0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 11:44:47 -0600 Subject: link libEGLdri.so with libdrm --- src/egl/drivers/dri/Makefile | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/drivers/dri/Makefile b/src/egl/drivers/dri/Makefile index 13aacd4389..bdc683a0c1 100644 --- a/src/egl/drivers/dri/Makefile +++ b/src/egl/drivers/dri/Makefile @@ -29,6 +29,8 @@ SOURCES = egldri.c OBJECTS = $(SOURCES:.c=.o) +DRM_LIB = `pkg-config --libs libdrm` + .c.o: $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ @@ -43,7 +45,7 @@ library: $(TOP)/$(LIB_DIR)/libEGLdri.so $(TOP)/$(LIB_DIR)/libEGLdri.so: $(OBJECTS) $(TOP)/bin/mklib -o EGLdri -major 1 -minor 0 \ - -install $(TOP)/$(LIB_DIR) -ldl $(OBJECTS) + -install $(TOP)/$(LIB_DIR) -ldl $(OBJECTS) $(DRM_LIB) clean: -- cgit v1.2.3 From 17ec3b3cc5100deb6ef96977d87b2a7f05814f72 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 11:45:50 -0600 Subject: prototype new code to map card number to driver name --- src/egl/drivers/dri/egldri.c | 66 +++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 62 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c index 79a2eda3ae..f00625a059 100644 --- a/src/egl/drivers/dri/egldri.c +++ b/src/egl/drivers/dri/egldri.c @@ -1,5 +1,6 @@ /** - * Generic EGL driver for DRI. + * Generic EGL driver for DRI. This is basically an "adaptor" driver + * that allows libEGL to load/use regular DRI drivers. * * This file contains all the code needed to interface DRI-based drivers * with libEGL. @@ -37,6 +38,41 @@ const char *sysfs = "/sys/class"; static const int empty_attribute_list[1] = { None }; + +/** + * Given a card number, return the name of the DRI driver to use. + * This generally means reading the contents of + * /sys/class/drm/cardX/dri_library_name, where X is the card number + */ +static EGLBoolean +driver_name_from_card_number(int card, char *driverName, int maxDriverName) +{ + char path[2000]; + FILE *f; + int length; + + snprintf(path, sizeof(path), "%s/drm/card%d/dri_library_name", sysfs, card); + + f = fopen(path, "r"); + if (!f) + return EGL_FALSE; + + fgets(driverName, maxDriverName, f); + fclose(f); + + if ((length = strlen(driverName)) > 1) { + /* remove the trailing newline from sysfs */ + driverName[length - 1] = '\0'; + strncat(driverName, "_dri", maxDriverName); + return EGL_TRUE; + } + else { + return EGL_FALSE; + } +} + + + /** * The bootstrap function. * Return a new driDriver object and plug in API functions. @@ -45,6 +81,26 @@ static const int empty_attribute_list[1] = { None }; _EGLDriver * _eglMain(_EGLDisplay *dpy) { +#if 1 + const char *displayString = (const char *) dpy->NativeDisplay; + const int card = atoi(displayString + 1); + _EGLDriver *driver = NULL; + char driverName[1000]; + + if (!driver_name_from_card_number(card, driverName, sizeof(driverName))) { + _eglLog(_EGL_WARNING, + "Unable to determine driver name for card %d\n", card); + return NULL; + } + + _eglLog(_EGL_DEBUG, "Driver name: %s\n", driverName); + + driver = _eglOpenDriver(dpy, driverName); + + return driver; + +#else + int length; char path[NAME_MAX]; struct dirent *dirent; @@ -63,11 +119,11 @@ _eglMain(_EGLDisplay *dpy) if (strncmp(&dirent->d_name[0], "card", 4) != 0) continue; - if (strcmp(&dirent->d_name[4], &dpy->DriverName[1]) != 0) + if (strcmp(&dirent->d_name[4], &driverName[1]) != 0) continue; snprintf(path, sizeof(path), "%s/drm/card%s/dri_library_name", - sysfs, &dpy->DriverName[1]); + sysfs, &driverName[1]); _eglLog(_EGL_INFO, "Opening %s", path); #if 1 file = fopen(path, "r"); @@ -91,6 +147,7 @@ _eglMain(_EGLDisplay *dpy) closedir(dir); return driver; +#endif } @@ -1087,6 +1144,7 @@ _eglDRIInitialize(_EGLDriver *drv, EGLDisplay dpy, { _EGLDisplay *disp = _eglLookupDisplay(dpy); driDisplay *display; + const char *driverName = (const char *) disp->NativeDisplay; assert(disp); @@ -1101,7 +1159,7 @@ _eglDRIInitialize(_EGLDriver *drv, EGLDisplay dpy, *major = 1; *minor = 0; - sscanf(&disp->DriverName[1], "%d", &display->minor); + sscanf(driverName + 1, "%d", &display->minor); drv->Initialized = EGL_TRUE; return EGL_TRUE; -- cgit v1.2.3 From e94d383b9ba7964da9fefac2a55e10c00ee72392 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 12:54:33 -0600 Subject: egl: added eglstring.[ch] --- src/egl/main/Makefile | 6 ++++-- src/egl/main/eglstring.c | 24 ++++++++++++++++++++++++ src/egl/main/eglstring.h | 9 +++++++++ 3 files changed, 37 insertions(+), 2 deletions(-) create mode 100644 src/egl/main/eglstring.c create mode 100644 src/egl/main/eglstring.h (limited to 'src') diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index 5f3d65f72a..eabfb41450 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -17,8 +17,9 @@ HEADERS = \ eglhash.h \ eglmode.h \ eglscreen.h \ + eglstring.h \ eglsurface.h \ -# eglx.h + eglx.h SOURCES = \ eglapi.c \ @@ -32,8 +33,9 @@ SOURCES = \ eglhash.c \ eglmode.c \ eglscreen.c \ + eglstring.c \ eglsurface.c \ -# eglx.c + eglx.c OBJECTS = $(SOURCES:.c=.o) diff --git a/src/egl/main/eglstring.c b/src/egl/main/eglstring.c new file mode 100644 index 0000000000..ba7406158c --- /dev/null +++ b/src/egl/main/eglstring.c @@ -0,0 +1,24 @@ +/** + * String utils. + */ + +#include +#include +#include "eglstring.h" + + +char * +_eglstrdup(const char *s) +{ + if (s) { + int l = strlen(s); + char *s2 = malloc(l + 1); + if (s2) + strcpy(s2, s); + return s2; + } + return NULL; +} + + + diff --git a/src/egl/main/eglstring.h b/src/egl/main/eglstring.h new file mode 100644 index 0000000000..10468636e8 --- /dev/null +++ b/src/egl/main/eglstring.h @@ -0,0 +1,9 @@ +#ifndef EGLSTRING_INCLUDED +#define EGLSTRING_INCLUDED + + +extern char * +_eglstrdup(const char *s); + + +#endif /* EGLSTRING_INCLUDED */ -- cgit v1.2.3 From 0c8908c411c434eda318b41b4f2a370a1e794831 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 12:56:36 -0600 Subject: egl: added args string to _eglMain() --- src/egl/drivers/demo/demo.c | 2 +- src/egl/drivers/dri/egldri.c | 4 ++-- src/egl/main/eglapi.c | 4 +++- src/egl/main/egldisplay.c | 19 ++++--------------- src/egl/main/egldisplay.h | 1 + src/egl/main/egldriver.c | 35 +++++++++++++++++++++++++---------- src/egl/main/egldriver.h | 4 ++-- src/egl/main/egltypedefs.h | 2 +- 8 files changed, 39 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/demo/demo.c b/src/egl/drivers/demo/demo.c index 1033f1b4ed..6b8b71d16b 100644 --- a/src/egl/drivers/demo/demo.c +++ b/src/egl/drivers/demo/demo.c @@ -286,7 +286,7 @@ demoMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface rea * plug in API functions. */ _EGLDriver * -_eglMain(_EGLDisplay *dpy) +_eglMain(_EGLDisplay *dpy, const char *args) { DemoDriver *demo; diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c index f00625a059..677073fb3a 100644 --- a/src/egl/drivers/dri/egldri.c +++ b/src/egl/drivers/dri/egldri.c @@ -79,11 +79,11 @@ driver_name_from_card_number(int card, char *driverName, int maxDriverName) * This function, in turn, loads a specific DRI driver (ex: r200_dri.so). */ _EGLDriver * -_eglMain(_EGLDisplay *dpy) +_eglMain(_EGLDisplay *dpy, const char *args) { #if 1 const char *displayString = (const char *) dpy->NativeDisplay; - const int card = atoi(displayString + 1); + const int card = atoi(args); _EGLDriver *driver = NULL; char driverName[1000]; diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index fe63d36b80..984af4ea22 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -68,7 +68,9 @@ eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor) if (!dpyPriv) { return EGL_FALSE; } - dpyPriv->Driver = _eglOpenDriver(dpyPriv, dpyPriv->DriverName); + dpyPriv->Driver = _eglOpenDriver(dpyPriv, + dpyPriv->DriverName, + dpyPriv->DriverArgs); if (!dpyPriv->Driver) { return EGL_FALSE; } diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c index 9c42194c61..b2d30d4274 100644 --- a/src/egl/main/egldisplay.c +++ b/src/egl/main/egldisplay.c @@ -11,25 +11,14 @@ #include "egldriver.h" #include "eglglobals.h" #include "eglhash.h" - - -static char * -my_strdup(const char *s) -{ - if (s) { - int l = strlen(s); - char *s2 = malloc(l + 1); - if (s2) - strcpy(s2, s); - return s2; - } - return NULL; -} +#include "eglstring.h" /** * Allocate a new _EGLDisplay object for the given nativeDisplay handle. * We'll also try to determine the device driver name at this time. + * + * Note that nativeDisplay may be an X Display ptr, or a string. */ _EGLDisplay * _eglNewDisplay(NativeDisplayType nativeDisplay) @@ -46,7 +35,7 @@ _eglNewDisplay(NativeDisplayType nativeDisplay) dpy->Xdpy = (Display *) nativeDisplay; #endif - dpy->DriverName = my_strdup(_eglChooseDriver(dpy)); + dpy->DriverName = _eglstrdup(_eglChooseDriver(dpy)); if (!dpy->DriverName) { free(dpy); return NULL; diff --git a/src/egl/main/egldisplay.h b/src/egl/main/egldisplay.h index be134374ca..e2ebab0b21 100644 --- a/src/egl/main/egldisplay.h +++ b/src/egl/main/egldisplay.h @@ -14,6 +14,7 @@ struct _egl_display EGLDisplay Handle; const char *DriverName; + const char *DriverArgs; _EGLDriver *Driver; EGLint NumScreens; diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index cbdd47948d..599077190a 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -6,6 +6,7 @@ #include #include #include +#include #include #include "eglconfig.h" #include "eglcontext.h" @@ -15,6 +16,7 @@ #include "egllog.h" #include "eglmode.h" #include "eglscreen.h" +#include "eglstring.h" #include "eglsurface.h" #if defined(_EGL_PLATFORM_X) @@ -25,7 +27,7 @@ /* XXX to do */ #endif -const char *DefaultDriverName = "demodriver"; +const char *DefaultDriverName = ":0"; /** @@ -44,23 +46,36 @@ const char *DefaultDriverName = "demodriver"; const char * _eglChooseDriver(_EGLDisplay *dpy) { - const char *name = (const char *) dpy->NativeDisplay; + const char *displayString = (const char *) dpy->NativeDisplay; const char *driverName = NULL; - if (!dpy->NativeDisplay) { + if (!displayString) { /* choose a default */ - driverName = DefaultDriverName; + displayString = DefaultDriverName; } - else if (name && name[0] == ':' && - (name[1] >= '0' && name[1] <= '9') && !name[2]) { + + /* extract default DriverArgs = whatever follows ':' */ + if (displayString[0] == '!' || + displayString[0] == ':') { + const char *args = strchr(displayString, ':'); + if (args) + dpy->DriverArgs = _eglstrdup(args + 1); + } + + + if (displayString && displayString[0] == ':' && + (displayString[1] >= '0' && displayString[1] <= '9') && + !displayString[2]) { /* XXX probe hardware here to determine which driver to open */ driverName = "libEGLdri"; } - else if (name && name[0] == '!') { + else if (displayString && displayString[0] == '!') { /* use specified driver name */ - driverName = name + 1; + driverName = displayString + 1; } else { + /* NativeDisplay is not a string! */ + #if defined(_EGL_PLATFORM_X) driverName = _xeglChooseDriver(dpy); #elif defined(_EGL_PLATFORM_WINDOWS) @@ -83,7 +98,7 @@ _eglChooseDriver(_EGLDisplay *dpy) * \return new _EGLDriver object. */ _EGLDriver * -_eglOpenDriver(_EGLDisplay *dpy, const char *driverName) +_eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) { _EGLDriver *drv; _EGLMain_t mainFunc; @@ -110,7 +125,7 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName) return NULL; } - drv = mainFunc(dpy); + drv = mainFunc(dpy, args); if (!drv) { dlclose(lib); return NULL; diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index bde726e25e..9c505880b7 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -42,7 +42,7 @@ struct _egl_driver }; -extern _EGLDriver *_eglMain(_EGLDisplay *dpy); +extern _EGLDriver *_eglMain(_EGLDisplay *dpy, const char *args); extern const char * @@ -50,7 +50,7 @@ _eglChooseDriver(_EGLDisplay *dpy); extern _EGLDriver * -_eglOpenDriver(_EGLDisplay *dpy, const char *DriverName); +_eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args); extern EGLBoolean diff --git a/src/egl/main/egltypedefs.h b/src/egl/main/egltypedefs.h index 53810a5a44..b1c8ec1f02 100644 --- a/src/egl/main/egltypedefs.h +++ b/src/egl/main/egltypedefs.h @@ -30,7 +30,7 @@ typedef struct _egl_thread_info _EGLThreadInfo; typedef void (*_EGLProc)(); -typedef _EGLDriver *(*_EGLMain_t)(_EGLDisplay *dpy); +typedef _EGLDriver *(*_EGLMain_t)(_EGLDisplay *dpy, const char *args); #endif /* EGLTYPEDEFS_INCLUDED */ -- cgit v1.2.3 From aa6e350bed8e6930bef761fb85ad524e5a4aa3a3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 13:03:59 -0600 Subject: egl: pass args to _eglOpenDriver() --- src/egl/drivers/dri/egldri.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c index 677073fb3a..35b6d0d3d1 100644 --- a/src/egl/drivers/dri/egldri.c +++ b/src/egl/drivers/dri/egldri.c @@ -82,7 +82,6 @@ _EGLDriver * _eglMain(_EGLDisplay *dpy, const char *args) { #if 1 - const char *displayString = (const char *) dpy->NativeDisplay; const int card = atoi(args); _EGLDriver *driver = NULL; char driverName[1000]; @@ -95,7 +94,7 @@ _eglMain(_EGLDisplay *dpy, const char *args) _eglLog(_EGL_DEBUG, "Driver name: %s\n", driverName); - driver = _eglOpenDriver(dpy, driverName); + driver = _eglOpenDriver(dpy, driverName, args); return driver; -- cgit v1.2.3 From 9843c6420d88db0c43b831cf79a3d1872c636225 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 14:51:40 -0600 Subject: egl: bring card->driver lookup code into egldriver.c --- src/egl/main/egldisplay.c | 2 +- src/egl/main/egldriver.c | 60 +++++++++++++++++++++++++++++++++++++++++------ 2 files changed, 54 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c index b2d30d4274..8fd29b8421 100644 --- a/src/egl/main/egldisplay.c +++ b/src/egl/main/egldisplay.c @@ -35,7 +35,7 @@ _eglNewDisplay(NativeDisplayType nativeDisplay) dpy->Xdpy = (Display *) nativeDisplay; #endif - dpy->DriverName = _eglstrdup(_eglChooseDriver(dpy)); + dpy->DriverName = _eglChooseDriver(dpy); if (!dpy->DriverName) { free(dpy); return NULL; diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 599077190a..b58222e48e 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -28,6 +28,44 @@ #endif const char *DefaultDriverName = ":0"; +const char *SysFS = "/sys/class"; + + + + +/** + * Given a card number, use sysfs to determine the DRI driver name. + */ +static const char * +_eglChooseDRMDriver(int card) +{ +#if 0 + return _eglstrdup("libEGLdri"); +#else + char path[2000], driverName[2000]; + FILE *f; + int length; + + snprintf(path, sizeof(path), "%s/drm/card%d/dri_library_name", SysFS, card); + + f = fopen(path, "r"); + if (!f) + return NULL; + + fgets(driverName, sizeof(driverName), f); + fclose(f); + + if ((length = strlen(driverName)) > 1) { + /* remove the trailing newline from sysfs */ + driverName[length - 1] = '\0'; + strncat(driverName, "_dri", sizeof(driverName)); + return _eglstrdup(driverName); + } + else { + return NULL; + } +#endif +} /** @@ -41,7 +79,9 @@ const char *DefaultDriverName = ":0"; * Else if the first character is '!' we interpret it as specific driver name * (i.e. "!r200" or "!i830". * - * The caller should make a copy of the returned string. + * Whatever follows ':' is copied and put into dpy->DriverArgs. + * + * The caller may free() the returned string. */ const char * _eglChooseDriver(_EGLDisplay *dpy) @@ -62,20 +102,26 @@ _eglChooseDriver(_EGLDisplay *dpy) dpy->DriverArgs = _eglstrdup(args + 1); } - + /* determine driver name now */ if (displayString && displayString[0] == ':' && (displayString[1] >= '0' && displayString[1] <= '9') && !displayString[2]) { - /* XXX probe hardware here to determine which driver to open */ - driverName = "libEGLdri"; + int card = atoi(displayString + 1); + driverName = _eglChooseDRMDriver(card); } else if (displayString && displayString[0] == '!') { - /* use specified driver name */ - driverName = displayString + 1; + /* use user-specified driver name */ + driverName = _eglstrdup(displayString + 1); + /* truncate driverName at ':' if present */ + { + char *args = strchr(driverName, ':'); + if (args) { + *args = 0; + } + } } else { /* NativeDisplay is not a string! */ - #if defined(_EGL_PLATFORM_X) driverName = _xeglChooseDriver(dpy); #elif defined(_EGL_PLATFORM_WINDOWS) -- cgit v1.2.3 From cf3c7d74ccee8e21ebfa04159ab64fec3b989d83 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 15:21:52 -0600 Subject: egl: added egldefines.h --- src/egl/main/Makefile | 1 + src/egl/main/egldefines.h | 47 +++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 48 insertions(+) create mode 100644 src/egl/main/egldefines.h (limited to 'src') diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index eabfb41450..1c395ec928 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -10,6 +10,7 @@ HEADERS = \ eglconfig.h \ eglconfigutil.h \ eglcontext.h \ + egldefines.h \ egldisplay.h \ egldriver.h \ eglglobals.h \ diff --git a/src/egl/main/egldefines.h b/src/egl/main/egldefines.h new file mode 100644 index 0000000000..f5d37a2fc0 --- /dev/null +++ b/src/egl/main/egldefines.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + + +/** + * Internal EGL defines + */ + + +#ifndef EGLDEFINES_INCLUDED +#define EGLDEFINES_INCLUDED + + +#define _EGL_MAX_EXTENSIONS_LEN 1000 + +#define _EGL_VERSION_STRING "1.4" + +#define _EGL_VENDOR_STRING "Mesa Project" + + + +#endif /* EGLDEFINES_INCLUDED */ -- cgit v1.2.3 From e084fe54f93c9d51df99812b76d3299b0cff57a3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 15:22:17 -0600 Subject: egl: include egldefines.h --- src/egl/main/egldriver.c | 8 +++++--- src/egl/main/egldriver.h | 6 ++---- 2 files changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index b58222e48e..6d533625b7 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -10,6 +10,7 @@ #include #include "eglconfig.h" #include "eglcontext.h" +#include "egldefines.h" #include "egldisplay.h" #include "egldriver.h" #include "eglglobals.h" @@ -27,6 +28,7 @@ /* XXX to do */ #endif + const char *DefaultDriverName = ":0"; const char *SysFS = "/sys/class"; @@ -290,7 +292,7 @@ _eglUpdateExtensionsString(_EGLDriver *drv) strcat(drv->Extensions.String, "EGL_MESA_screen_surface "); if (drv->Extensions.MESA_copy_context) strcat(drv->Extensions.String, "EGL_MESA_copy_context "); - assert(strlen(drv->Extensions.String) < MAX_EXTENSIONS_LEN); + assert(strlen(drv->Extensions.String) < _EGL_MAX_EXTENSIONS_LEN); } @@ -302,9 +304,9 @@ _eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name) (void) dpy; switch (name) { case EGL_VENDOR: - return "Mesa Project"; + return _EGL_VENDOR_STRING; case EGL_VERSION: - return "1.0"; + return _EGL_VERSION_STRING; case EGL_EXTENSIONS: _eglUpdateExtensionsString(drv); return drv->Extensions.String; diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index 9c505880b7..67cfb02f46 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -4,9 +4,7 @@ #include "egltypedefs.h" #include "eglapi.h" - -/* should probably use a dynamic-length string, but this will do */ -#define MAX_EXTENSIONS_LEN 1000 +#include "egldefines.h" /** @@ -17,7 +15,7 @@ struct _egl_extensions EGLBoolean MESA_screen_surface; EGLBoolean MESA_copy_context; - char String[MAX_EXTENSIONS_LEN]; + char String[_EGL_MAX_EXTENSIONS_LEN]; }; -- cgit v1.2.3 From c56e15b093c367e7e17ebd2e153baab8cafd213a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 15:43:41 -0600 Subject: egl: make sure EGL_VERSION_STRING query returns same version as eglInitialize() --- src/egl/main/eglapi.c | 11 +++++++++-- src/egl/main/egldefines.h | 2 -- src/egl/main/egldriver.c | 2 +- src/egl/main/egldriver.h | 11 ++++++----- 4 files changed, 16 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index 984af4ea22..3151b35156 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -64,6 +64,7 @@ EGLBoolean EGLAPIENTRY eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor) { if (dpy) { + EGLBoolean retVal; _EGLDisplay *dpyPriv = _eglLookupDisplay(dpy); if (!dpyPriv) { return EGL_FALSE; @@ -75,8 +76,14 @@ eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor) return EGL_FALSE; } /* Initialize the particular driver now */ - return dpyPriv->Driver->API.Initialize(dpyPriv->Driver, dpy, - major, minor); + retVal = dpyPriv->Driver->API.Initialize(dpyPriv->Driver, dpy, + major, minor); + + dpyPriv->Driver->APImajor = *major; + dpyPriv->Driver->APIminor = *minor; + sprintf(dpyPriv->Driver->Version, "%d.%d", *major, *minor); + + return retVal; } return EGL_FALSE; } diff --git a/src/egl/main/egldefines.h b/src/egl/main/egldefines.h index f5d37a2fc0..8fc2301b79 100644 --- a/src/egl/main/egldefines.h +++ b/src/egl/main/egldefines.h @@ -38,8 +38,6 @@ #define _EGL_MAX_EXTENSIONS_LEN 1000 -#define _EGL_VERSION_STRING "1.4" - #define _EGL_VENDOR_STRING "Mesa Project" diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 6d533625b7..2257a68630 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -306,7 +306,7 @@ _eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name) case EGL_VENDOR: return _EGL_VENDOR_STRING; case EGL_VERSION: - return _EGL_VERSION_STRING; + return drv->Version; case EGL_EXTENSIONS: _eglUpdateExtensionsString(drv); return drv->Extensions.String; diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index 67cfb02f46..9395a667ab 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -24,17 +24,18 @@ struct _egl_extensions */ struct _egl_driver { - EGLBoolean Initialized; /* set by driver after initialized */ + EGLBoolean Initialized; /**< set by driver after initialized */ - void *LibHandle; /* dlopen handle */ + void *LibHandle; /**< dlopen handle */ _EGLDisplay *Display; - int ABIversion; - int APImajor, APIminor; /* returned through eglInitialize */ + int APImajor, APIminor; /**< as returned by eglInitialize() */ + char Version[10]; /**< initialized from APImajor/minor */ + const char *ClientAPIs; - _EGLAPI API; + _EGLAPI API; /**< EGL API dispatch table */ _EGLExtensions Extensions; }; -- cgit v1.2.3 From 11a261ef4f1d4100c46f73ad51e7e4ed57cc1b5e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 15:50:58 -0600 Subject: egl: move a few small functions into new eglmisc.[ch] files --- src/egl/main/Makefile | 2 + src/egl/main/egldriver.c | 74 +------------------------------- src/egl/main/egldriver.h | 13 ------ src/egl/main/eglmisc.c | 108 +++++++++++++++++++++++++++++++++++++++++++++++ src/egl/main/eglmisc.h | 47 +++++++++++++++++++++ 5 files changed, 159 insertions(+), 85 deletions(-) create mode 100644 src/egl/main/eglmisc.c create mode 100644 src/egl/main/eglmisc.h (limited to 'src') diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index 1c395ec928..6676e7e031 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -16,6 +16,7 @@ HEADERS = \ eglglobals.h \ egllog.h \ eglhash.h \ + eglmisc.h \ eglmode.h \ eglscreen.h \ eglstring.h \ @@ -32,6 +33,7 @@ SOURCES = \ eglglobals.c \ egllog.c \ eglhash.c \ + eglmisc.c \ eglmode.c \ eglscreen.c \ eglstring.c \ diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 2257a68630..9f3c52fd2f 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -4,10 +4,10 @@ #include +#include #include #include #include -#include #include "eglconfig.h" #include "eglcontext.h" #include "egldefines.h" @@ -15,6 +15,7 @@ #include "egldriver.h" #include "eglglobals.h" #include "egllog.h" +#include "eglmisc.h" #include "eglmode.h" #include "eglscreen.h" #include "eglstring.h" @@ -277,74 +278,3 @@ _eglInitDriverFallbacks(_EGLDriver *drv) drv->API.CreatePbufferFromClientBuffer = _eglCreatePbufferFromClientBuffer; #endif /* EGL_VERSION_1_2 */ } - - -/** - * Examine the individual extension enable/disable flags and recompute - * the driver's Extensions string. - */ -static void -_eglUpdateExtensionsString(_EGLDriver *drv) -{ - drv->Extensions.String[0] = 0; - - if (drv->Extensions.MESA_screen_surface) - strcat(drv->Extensions.String, "EGL_MESA_screen_surface "); - if (drv->Extensions.MESA_copy_context) - strcat(drv->Extensions.String, "EGL_MESA_copy_context "); - assert(strlen(drv->Extensions.String) < _EGL_MAX_EXTENSIONS_LEN); -} - - - -const char * -_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name) -{ - (void) drv; - (void) dpy; - switch (name) { - case EGL_VENDOR: - return _EGL_VENDOR_STRING; - case EGL_VERSION: - return drv->Version; - case EGL_EXTENSIONS: - _eglUpdateExtensionsString(drv); - return drv->Extensions.String; -#ifdef EGL_VERSION_1_2 - case EGL_CLIENT_APIS: - /* XXX need to initialize somewhere */ - return drv->ClientAPIs; -#endif - default: - _eglError(EGL_BAD_PARAMETER, "eglQueryString"); - return NULL; - } -} - - -EGLBoolean -_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy) -{ - /* just a placeholder */ - (void) drv; - (void) dpy; - return EGL_TRUE; -} - - -EGLBoolean -_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine) -{ - /* just a placeholder */ - (void) drv; - (void) dpy; - switch (engine) { - case EGL_CORE_NATIVE_ENGINE: - break; - default: - _eglError(EGL_BAD_PARAMETER, "eglWaitNative(engine)"); - return EGL_FALSE; - } - - return EGL_TRUE; -} diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index 9395a667ab..df0eb7fc4c 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -64,17 +64,4 @@ extern void _eglInitDriverFallbacks(_EGLDriver *drv); -extern const char * -_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name); - - -extern EGLBoolean -_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy); - - -extern EGLBoolean -_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine); - - - #endif /* EGLDRIVER_INCLUDED */ diff --git a/src/egl/main/eglmisc.c b/src/egl/main/eglmisc.c new file mode 100644 index 0000000000..fb32544b15 --- /dev/null +++ b/src/egl/main/eglmisc.c @@ -0,0 +1,108 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Small/misc EGL functions + */ + + +#include +#include +#include "eglglobals.h" +#include "eglmisc.h" + + +/** + * Examine the individual extension enable/disable flags and recompute + * the driver's Extensions string. + */ +static void +_eglUpdateExtensionsString(_EGLDriver *drv) +{ + drv->Extensions.String[0] = 0; + + if (drv->Extensions.MESA_screen_surface) + strcat(drv->Extensions.String, "EGL_MESA_screen_surface "); + if (drv->Extensions.MESA_copy_context) + strcat(drv->Extensions.String, "EGL_MESA_copy_context "); + assert(strlen(drv->Extensions.String) < _EGL_MAX_EXTENSIONS_LEN); +} + + + +const char * +_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name) +{ + (void) drv; + (void) dpy; + switch (name) { + case EGL_VENDOR: + return _EGL_VENDOR_STRING; + case EGL_VERSION: + return drv->Version; + case EGL_EXTENSIONS: + _eglUpdateExtensionsString(drv); + return drv->Extensions.String; +#ifdef EGL_VERSION_1_2 + case EGL_CLIENT_APIS: + /* XXX need to initialize somewhere */ + return drv->ClientAPIs; +#endif + default: + _eglError(EGL_BAD_PARAMETER, "eglQueryString"); + return NULL; + } +} + + +EGLBoolean +_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy) +{ + /* just a placeholder */ + (void) drv; + (void) dpy; + return EGL_TRUE; +} + + +EGLBoolean +_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine) +{ + /* just a placeholder */ + (void) drv; + (void) dpy; + switch (engine) { + case EGL_CORE_NATIVE_ENGINE: + break; + default: + _eglError(EGL_BAD_PARAMETER, "eglWaitNative(engine)"); + return EGL_FALSE; + } + + return EGL_TRUE; +} diff --git a/src/egl/main/eglmisc.h b/src/egl/main/eglmisc.h new file mode 100644 index 0000000000..4e2a40ea99 --- /dev/null +++ b/src/egl/main/eglmisc.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef EGLMISC_INCLUDED +#define EGLMISC_INCLUDED + +#include "egldriver.h" + + +extern const char * +_eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name); + + +extern EGLBoolean +_eglWaitGL(_EGLDriver *drv, EGLDisplay dpy); + + +extern EGLBoolean +_eglWaitNative(_EGLDriver *drv, EGLDisplay dpy, EGLint engine); + + +#endif /* EGLMISC_INCLUDED */ -- cgit v1.2.3 From 260ac3c983644e1e51f705c00ceec1866a0c9611 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 16:08:50 -0600 Subject: egl: remove unused Display field --- src/egl/main/egldriver.c | 1 - src/egl/main/egldriver.h | 2 -- 2 files changed, 3 deletions(-) (limited to 'src') diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 9f3c52fd2f..32842933ce 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -185,7 +185,6 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) else dlclose(lib); - drv->Display = dpy; return drv; } diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index df0eb7fc4c..b0a5fadf64 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -28,8 +28,6 @@ struct _egl_driver void *LibHandle; /**< dlopen handle */ - _EGLDisplay *Display; - int APImajor, APIminor; /**< as returned by eglInitialize() */ char Version[10]; /**< initialized from APImajor/minor */ -- cgit v1.2.3 From f751371028cc425bef83418224c23ece67f5b8b7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 16:48:35 -0600 Subject: egl: re-enable LOCAL_CFLAGS --- src/egl/main/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index 6676e7e031..5b341a19be 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -44,7 +44,7 @@ OBJECTS = $(SOURCES:.c=.o) # Undefined for now -#LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 +LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 .c.o: -- cgit v1.2.3 From 728d1f7f43b6db9f4f42c2d16ba223c492d1147d Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 28 May 2008 23:54:18 +0100 Subject: draw: enable FSE by default --- src/gallium/auxiliary/draw/draw_pt.c | 11 ++- src/gallium/auxiliary/draw/draw_vs.h | 4 ++ src/gallium/auxiliary/draw/draw_vs_aos.c | 105 +++++++++++++++++++++------- src/gallium/auxiliary/draw/draw_vs_aos.h | 34 +++++---- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 25 ++++--- src/gallium/auxiliary/draw/draw_vs_sse.c | 14 ++-- 6 files changed, 129 insertions(+), 64 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 75f44d503e..d48c6c220d 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -75,7 +75,7 @@ draw_pt_arrays(struct draw_context *draw, if (opt == 0) middle = draw->pt.middle.fetch_emit; - else if (opt == PT_SHADE && draw->pt.test_fse) + else if (opt == PT_SHADE) middle = draw->pt.middle.fetch_shade_emit; else middle = draw->pt.middle.general; @@ -118,12 +118,9 @@ boolean draw_pt_init( struct draw_context *draw ) if (!draw->pt.middle.fetch_emit) return FALSE; - if (draw->pt.test_fse) { - draw->pt.middle.fetch_shade_emit = draw_pt_middle_fse( draw ); - if (!draw->pt.middle.fetch_shade_emit) - return FALSE; - } - + draw->pt.middle.fetch_shade_emit = draw_pt_middle_fse( draw ); + if (!draw->pt.middle.fetch_shade_emit) + return FALSE; draw->pt.middle.general = draw_pt_fetch_pipeline_or_emit( draw ); if (!draw->pt.middle.general) diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 01171bc23d..7aa0415baf 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -123,6 +123,10 @@ struct draw_vertex_shader { struct tgsi_shader_info info; + /* Extracted from shader: + */ + const float (*immediates)[4]; + /* */ struct draw_vs_varient *varient[16]; diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 0cd82ff599..9056785e7a 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -66,6 +66,37 @@ static INLINE boolean eq( struct x86_reg a, a.disp == b.disp); } +struct x86_reg aos_get_x86( struct aos_compilation *cp, + unsigned value ) +{ + if (cp->ebp != value) { + unsigned offset; + + switch (value) { + case X86_IMMEDIATES: + offset = Offset(struct aos_machine, immediates); + break; + case X86_CONSTANTS: + offset = Offset(struct aos_machine, constants); + break; + case X86_ATTRIBS: + offset = Offset(struct aos_machine, attrib); + break; + default: + assert(0); + offset = 0; + } + + x86_mov(cp->func, cp->temp_EBP, + x86_make_disp(cp->machine_EDX, offset)); + /* x86_deref(x86_make_disp(cp->machine_EDX, offset))); */ + + cp->ebp = value; + } + + return cp->temp_EBP; +} + static struct x86_reg get_reg_ptr(struct aos_compilation *cp, unsigned file, @@ -83,15 +114,15 @@ static struct x86_reg get_reg_ptr(struct aos_compilation *cp, case TGSI_FILE_TEMPORARY: return x86_make_disp(ptr, Offset(struct aos_machine, temp[idx])); - case TGSI_FILE_IMMEDIATE: - return x86_make_disp(ptr, Offset(struct aos_machine, immediate[idx])); - - case TGSI_FILE_CONSTANT: - return x86_make_disp(ptr, Offset(struct aos_machine, constant[idx])); - case AOS_FILE_INTERNAL: return x86_make_disp(ptr, Offset(struct aos_machine, internal[idx])); + case TGSI_FILE_IMMEDIATE: + return x86_make_disp(aos_get_x86(cp, X86_IMMEDIATES), idx * 4 * sizeof(float)); + + case TGSI_FILE_CONSTANT: + return x86_make_disp(aos_get_x86(cp, X86_CONSTANTS), idx * 4 * sizeof(float)); + default: ERROR(cp, "unknown reg file"); return x86_make_reg(0,0); @@ -1865,6 +1896,7 @@ static boolean emit_rhw_viewport( struct aos_compilation *cp ) } +#if 0 static boolean note_immediate( struct aos_compilation *cp, struct tgsi_full_immediate *imm ) { @@ -1877,6 +1909,7 @@ static boolean note_immediate( struct aos_compilation *cp, return TRUE; } +#endif @@ -1939,6 +1972,7 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, cp.outbuf_ECX = x86_make_reg(file_REG32, reg_CX); cp.machine_EDX = x86_make_reg(file_REG32, reg_DX); cp.count_ESI = x86_make_reg(file_REG32, reg_SI); + cp.temp_EBP = x86_make_reg(file_REG32, reg_BP); x86_init_func(cp.func); @@ -1946,6 +1980,7 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, x86_push(cp.func, cp.idx_EBX); x86_push(cp.func, cp.count_ESI); + x86_push(cp.func, cp.temp_EBP); /* Load arguments into regs: @@ -1988,8 +2023,10 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, switch (parse.FullToken.Token.Type) { case TGSI_TOKEN_TYPE_IMMEDIATE: +#if 0 if (!note_immediate( &cp, &parse.FullToken.FullImmediate )) goto fail; +#endif break; case TGSI_TOKEN_TYPE_INSTRUCTION: @@ -2072,6 +2109,7 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, if (cp.func->need_emms) mmx_emms(cp.func); + x86_pop(cp.func, cp.temp_EBP); x86_pop(cp.func, cp.count_ESI); x86_pop(cp.func, cp.idx_EBX); @@ -2098,26 +2136,14 @@ static void vaos_set_buffer( struct draw_vs_varient *varient, for (i = 0; i < vaos->base.key.nr_inputs; i++) { if (vaos->base.key.element[i].in.buffer == buf) { - vaos->machine->attrib[i].input_ptr = ((char *)ptr + - vaos->base.key.element[i].in.offset); - vaos->machine->attrib[i].input_stride = stride; + vaos->attrib[i].input_ptr = ((char *)ptr + + vaos->base.key.element[i].in.offset); + vaos->attrib[i].input_stride = stride; } } } -static void vaos_destroy( struct draw_vs_varient *varient ) -{ - struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; - - if (vaos->machine) - align_free( vaos->machine ); - - x86_release_func( &vaos->func[0] ); - x86_release_func( &vaos->func[1] ); - - FREE(vaos); -} static void PIPE_CDECL vaos_run_elts( struct draw_vs_varient *varient, const unsigned *elts, @@ -2127,6 +2153,10 @@ static void PIPE_CDECL vaos_run_elts( struct draw_vs_varient *varient, struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; vaos->machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; + vaos->machine->constants = vaos->draw->pt.user.constants; + vaos->machine->immediates = vaos->base.vs->immediates; + vaos->machine->attrib = vaos->attrib; + vaos->gen_run_elts( varient, elts, count, @@ -2141,6 +2171,10 @@ static void PIPE_CDECL vaos_run_linear( struct draw_vs_varient *varient, struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; vaos->machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; + vaos->machine->constants = vaos->draw->pt.user.constants; + vaos->machine->immediates = vaos->base.vs->immediates; + vaos->machine->attrib = vaos->attrib; + vaos->gen_run_linear( varient, start, count, @@ -2153,10 +2187,6 @@ static void vaos_set_constants( struct draw_vs_varient *varient, { struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; - memcpy(vaos->machine->constant, - constants, - (vaos->base.vs->info.file_max[TGSI_FILE_CONSTANT] + 1) * 4 * sizeof(float)); - #if 0 unsigned i; for (i =0; i < vaos->base.vs->info.file_max[TGSI_FILE_CONSTANT] + 1; i++) @@ -2187,6 +2217,21 @@ static void vaos_set_viewport( struct draw_vs_varient *varient, memcpy(vaos->machine->translate, viewport->translate, 4 * sizeof(float)); } +static void vaos_destroy( struct draw_vs_varient *varient ) +{ + struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + + if (vaos->machine) + align_free( vaos->machine ); + + FREE( vaos->attrib ); + + x86_release_func( &vaos->func[0] ); + x86_release_func( &vaos->func[1] ); + + FREE(vaos); +} + static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, @@ -2207,6 +2252,11 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, vaos->base.run_elts = vaos_run_elts; vaos->draw = vs->draw; + + vaos->attrib = MALLOC( key->nr_inputs * sizeof(vaos->attrib[0]) ); + if (!vaos->attrib) + goto fail; + vaos->machine = align_malloc( sizeof(struct aos_machine), 16 ); if (!vaos->machine) goto fail; @@ -2233,7 +2283,10 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, return &vaos->base; fail: - if (vaos->machine) + if (vaos && vaos->attrib) + FREE(vaos->attrib); + + if (vaos && vaos->machine) align_free( vaos->machine ); if (vaos) diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index 837b32794f..295d2cb3fe 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -78,6 +78,14 @@ struct lit_info { #define MAX_SHINE_TAB 4 #define MAX_LIT_INFO 16 +struct aos_attrib { + const void *input_ptr; + unsigned input_stride; +}; + + + + /* This is the temporary storage used by all the aos_sse vs varients. * Create one per context and reuse by passing a pointer in at * vs_varient creation?? @@ -86,8 +94,6 @@ struct aos_machine { float input [MAX_INPUTS ][4]; float output [MAX_OUTPUTS ][4]; float temp [MAX_TEMPS ][4]; - float constant [MAX_CONSTANTS ][4]; /* fixme -- should just be a pointer */ - float immediate[MAX_IMMEDIATES][4]; /* fixme -- should just be a pointer */ float internal [MAX_INTERNALS ][4]; float scale[4]; /* viewport */ @@ -105,12 +111,10 @@ struct aos_machine { ushort fpu_restore; ushort fpucntl; /* one of FPU_* above */ - struct { - const void *input_ptr; - unsigned input_stride; + const float (*immediates)[4]; /* points to shader data */ + const float (*constants)[4]; /* points to draw data */ - unsigned output_offset; - } attrib[PIPE_MAX_ATTRIBS]; + const struct aos_attrib *attrib; /* points to ? */ }; @@ -132,6 +136,7 @@ struct aos_compilation { unsigned last_used; } xmm[8]; + unsigned ebp; /* one of X86_* */ boolean input_fetched[PIPE_MAX_ATTRIBS]; unsigned output_last_write[PIPE_MAX_ATTRIBS]; @@ -148,6 +153,7 @@ struct aos_compilation { struct x86_reg outbuf_ECX; struct x86_reg machine_EDX; struct x86_reg count_ESI; /* decrements to zero */ + struct x86_reg temp_EBP; }; struct x86_reg aos_get_xmm_reg( struct aos_compilation *cp ); @@ -192,20 +198,20 @@ do { \ } while (0) +#define X86_NULL 0 +#define X86_IMMEDIATES 1 +#define X86_CONSTANTS 2 +#define X86_ATTRIBS 3 - +struct x86_reg aos_get_x86( struct aos_compilation *cp, + unsigned value ); struct draw_vs_varient_aos_sse { struct draw_vs_varient base; struct draw_context *draw; -#if 0 - struct { - const void *ptr; - unsigned stride; - } attrib[PIPE_MAX_ATTRIBS]; -#endif + struct aos_attrib *attrib; struct aos_machine *machine; /* XXX: temporarily unshared */ diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index 836110f382..45e2092209 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -91,25 +91,25 @@ static void emit_load_R8G8B8A8_UNORM( struct aos_compilation *cp, -static void get_src_ptr( struct x86_function *func, +static void get_src_ptr( struct aos_compilation *cp, struct x86_reg src, - struct x86_reg machine, struct x86_reg elt, unsigned a ) { - struct x86_reg input_ptr = - x86_make_disp(machine, - Offset(struct aos_machine, attrib[a].input_ptr)); + struct x86_reg attrib = x86_make_disp(aos_get_x86( cp, X86_ATTRIBS ), + a * sizeof(struct aos_attrib)); - struct x86_reg input_stride = - x86_make_disp(machine, - Offset(struct aos_machine, attrib[a].input_stride)); + struct x86_reg input_ptr = x86_make_disp(attrib, + Offset(struct aos_attrib, input_ptr)); + + struct x86_reg input_stride = x86_make_disp(attrib, + Offset(struct aos_attrib, input_stride)); /* Calculate pointer to current attrib: */ - x86_mov(func, src, input_stride); - x86_imul(func, src, elt); - x86_add(func, src, input_ptr); + x86_mov(cp->func, src, input_stride); + x86_imul(cp->func, src, elt); + x86_add(cp->func, src, input_ptr); } @@ -134,9 +134,8 @@ static boolean load_input( struct aos_compilation *cp, /* Figure out source pointer address: */ - get_src_ptr(cp->func, + get_src_ptr(cp, src, - cp->machine_EDX, linear ? cp->idx_EBX : x86_deref(cp->idx_EBX), idx); diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 7781782ae8..24f619a278 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -68,8 +68,6 @@ struct draw_sse_vertex_shader { codegen_function func; struct tgsi_exec_machine *machine; - - float immediates[TGSI_EXEC_NUM_IMMEDIATES][4]; }; @@ -107,7 +105,7 @@ vs_sse_run_linear( struct draw_vertex_shader *base, machine->Outputs, (float (*)[4])constants, machine->Temps, - shader->immediates, + (float (*)[4])shader->base.immediates, input, base->info.num_inputs, input_stride, @@ -130,6 +128,8 @@ vs_sse_delete( struct draw_vertex_shader *base ) x86_release_func( &shader->sse2_program ); + align_free(shader->base.immediates); + FREE( (void*) shader->base.state.tokens ); FREE( shader ); } @@ -161,12 +161,18 @@ draw_create_vs_sse(struct draw_context *draw, vs->base.prepare = vs_sse_prepare; vs->base.run_linear = vs_sse_run_linear; vs->base.delete = vs_sse_delete; + + vs->base.immediates = align_malloc(TGSI_EXEC_NUM_IMMEDIATES * 4 * + sizeof(float), 16); + vs->machine = &draw->vs.machine; x86_init_func( &vs->sse2_program ); if (!tgsi_emit_sse2( (struct tgsi_token *) vs->base.state.tokens, - &vs->sse2_program, vs->immediates, TRUE )) + &vs->sse2_program, + (float (*)[4])vs->base.immediates, + TRUE )) goto fail; vs->func = (codegen_function) x86_get_func( &vs->sse2_program ); -- cgit v1.2.3 From 62628c4d3d497cbca73fde869c9069fa90e6453e Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 29 May 2008 00:17:53 +0100 Subject: draw: share machine --- src/gallium/auxiliary/draw/Makefile | 1 + src/gallium/auxiliary/draw/SConscript | 1 + src/gallium/auxiliary/draw/draw_context.c | 3 + src/gallium/auxiliary/draw/draw_private.h | 9 + .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 6 - src/gallium/auxiliary/draw/draw_vs.c | 43 ++- src/gallium/auxiliary/draw/draw_vs.h | 24 +- src/gallium/auxiliary/draw/draw_vs_aos.c | 282 ++----------------- src/gallium/auxiliary/draw/draw_vs_aos.h | 25 +- src/gallium/auxiliary/draw/draw_vs_aos_machine.c | 297 +++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs_varient.c | 49 +--- 11 files changed, 412 insertions(+), 328 deletions(-) create mode 100644 src/gallium/auxiliary/draw/draw_vs_aos_machine.c (limited to 'src') diff --git a/src/gallium/auxiliary/draw/Makefile b/src/gallium/auxiliary/draw/Makefile index 9a88ecc070..f2e36a89e9 100644 --- a/src/gallium/auxiliary/draw/Makefile +++ b/src/gallium/auxiliary/draw/Makefile @@ -37,6 +37,7 @@ C_SOURCES = \ draw_vs_varient.c \ draw_vs_aos.c \ draw_vs_aos_io.c \ + draw_vs_aos_machine.c \ draw_vs_exec.c \ draw_vs_llvm.c \ draw_vs_sse.c diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 26919a2298..925e668f22 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -36,6 +36,7 @@ draw = env.ConvenienceLibrary( 'draw_vs.c', 'draw_vs_aos.c', 'draw_vs_aos_io.c', + 'draw_vs_aos_machine.c', 'draw_vs_exec.c', 'draw_vs_llvm.c', 'draw_vs_sse.c', diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 2242074965..8509baf865 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -174,6 +174,8 @@ void draw_set_viewport_state( struct draw_context *draw, viewport->translate[1] == 0.0f && viewport->translate[2] == 0.0f && viewport->translate[3] == 0.0f); + + draw_vs_set_viewport( draw, viewport ); } @@ -218,6 +220,7 @@ draw_set_mapped_constant_buffer(struct draw_context *draw, const void *buffer) { draw->pt.user.constants = buffer; + draw_vs_set_constants( draw, (const float (*)[4])buffer ); } diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index c095bf3d7b..4cbccc8b5b 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -183,6 +183,10 @@ struct draw_context */ struct gallivm_cpu_engine *engine; + /* Here's another one: + */ + struct aos_machine *aos_machine; + struct translate *fetch; struct translate_cache *fetch_cache; @@ -215,6 +219,11 @@ struct draw_context boolean draw_vs_init( struct draw_context *draw ); void draw_vs_destroy( struct draw_context *draw ); +void draw_vs_set_viewport( struct draw_context *, + const struct pipe_viewport_state * ); + +void draw_vs_set_constants( struct draw_context *, + const float (*constants)[4] ); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 729c7db999..5265a13160 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -189,12 +189,6 @@ static void fse_prepare( struct draw_pt_middle_end *middle, draw->pt.vertex_buffer[buf].pitch ); } - fse->active->set_constants( fse->active, - (const float (*)[4])draw->pt.user.constants ); - - fse->active->set_viewport( fse->active, - &draw->viewport ); - //return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index 9b899d404e..a8b6d0c90d 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -41,6 +41,22 @@ + +void draw_vs_set_constants( struct draw_context *draw, + const float (*constants)[4] ) +{ + draw_vs_aos_machine_constants( draw->vs.aos_machine, constants ); +} + + +void draw_vs_set_viewport( struct draw_context *draw, + const struct pipe_viewport_state *viewport ) +{ + draw_vs_aos_machine_viewport( draw->vs.aos_machine, viewport ); +} + + + struct draw_vertex_shader * draw_create_vertex_shader(struct draw_context *draw, const struct pipe_shader_state *shader) @@ -83,6 +99,13 @@ void draw_delete_vertex_shader(struct draw_context *draw, struct draw_vertex_shader *dvs) { + unsigned i; + + for (i = 0; i < dvs->nr_varients; i++) + dvs->varient[i]->destroy( dvs->varient[i] ); + + dvs->nr_varients = 0; + dvs->delete( dvs ); } @@ -110,6 +133,10 @@ draw_vs_init( struct draw_context *draw ) draw->vs.fetch_cache = translate_cache_create(); if (!draw->vs.fetch_cache) return FALSE; + + draw->vs.aos_machine = draw_vs_aos_machine(); + if (!draw->vs.aos_machine) + return FALSE; return TRUE; } @@ -129,6 +156,9 @@ draw_vs_destroy( struct draw_context *draw ) if (draw->vs.emit_cache) translate_cache_destroy(draw->vs.emit_cache); + if (draw->vs.aos_machine) + draw_vs_aos_machine_destroy(draw->vs.aos_machine); + tgsi_exec_machine_free_data(&draw->vs.machine); } @@ -153,10 +183,17 @@ draw_vs_lookup_varient( struct draw_vertex_shader *vs, if (varient == NULL) return NULL; - /* Add it to our list: + /* Add it to our list, could be smarter: */ - assert(vs->nr_varients < Elements(vs->varient)); - vs->varient[vs->nr_varients++] = varient; + if (vs->nr_varients < Elements(vs->varient)) { + vs->varient[vs->nr_varients++] = varient; + } + else { + vs->last_varient++; + vs->last_varient %= Elements(vs->varient); + vs->varient[vs->last_varient]->destroy(vs->varient[vs->last_varient]); + vs->varient[vs->last_varient] = varient; + } /* Done */ diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 7aa0415baf..08c6de8ba8 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -70,16 +70,6 @@ struct draw_vs_varient_key { struct draw_vs_varient; -typedef void (PIPE_CDECL *vsv_run_elts_func)( struct draw_vs_varient *, - const unsigned *elts, - unsigned count, - void *output_buffer); - -typedef void (PIPE_CDECL *vsv_run_linear_func)( struct draw_vs_varient *, - unsigned start, - unsigned count, - void *output_buffer); - struct draw_vs_varient { struct draw_vs_varient_key key; @@ -91,12 +81,6 @@ struct draw_vs_varient { const void *ptr, unsigned stride ); - void (*set_constants)( struct draw_vs_varient *, - const float (*constants)[4] ); - - void (*set_viewport)( struct draw_vs_varient *, - const struct pipe_viewport_state * ); - void (PIPE_CDECL *run_linear)( struct draw_vs_varient *shader, unsigned start, unsigned count, @@ -131,6 +115,7 @@ struct draw_vertex_shader { */ struct draw_vs_varient *varient[16]; unsigned nr_varients; + unsigned last_varient; struct draw_vs_varient *(*create_varient)( struct draw_vertex_shader *shader, const struct draw_vs_varient_key *key ); @@ -217,7 +202,14 @@ static INLINE int draw_vs_varient_key_compare( const struct draw_vs_varient_key } +struct aos_machine *draw_vs_aos_machine( void ); +void draw_vs_aos_machine_destroy( struct aos_machine *machine ); + +void draw_vs_aos_machine_constants( struct aos_machine *machine, + const float (*constants)[4] ); +void draw_vs_aos_machine_viewport( struct aos_machine *machine, + const struct pipe_viewport_state *viewport ); #define MAX_TGSI_VERTICES 4 diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 9056785e7a..b5e4e1e7b1 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -149,70 +149,7 @@ static struct x86_reg get_reg_ptr(struct aos_compilation *cp, #define X87_CW_ROUND_MASK (3<<10) #define X87_CW_INFINITY (1<<12) -static void do_populate_lut( struct shine_tab *tab, - float unclamped_exponent ) -{ - const float epsilon = 1.0F / 256.0F; - float exponent = CLAMP(unclamped_exponent, -(128.0F - epsilon), (128.0F - epsilon)); - unsigned i; - tab->exponent = unclamped_exponent; /* for later comparison */ - - tab->values[0] = 0; - if (exponent == 0) { - for (i = 1; i < 258; i++) { - tab->values[i] = 1.0; - } - } - else { - for (i = 1; i < 258; i++) { - tab->values[i] = powf((float)i * epsilon, exponent); - } - } -} - -static void init_internals( struct aos_machine *machine ) -{ - unsigned i; - float inv = 1.0f/255.0f; - float f255 = 255.0f; - - ASSIGN_4V(machine->internal[IMM_SWZ], 1.0f, -1.0f, 0.0f, 1.0f); - *(unsigned *)&machine->internal[IMM_SWZ][3] = 0xffffffff; - - ASSIGN_4V(machine->internal[IMM_ONES], 1.0f, 1.0f, 1.0f, 1.0f); - ASSIGN_4V(machine->internal[IMM_NEGS], -1.0f, -1.0f, -1.0f, -1.0f); - ASSIGN_4V(machine->internal[IMM_IDENTITY], 0.0f, 0.0f, 0.0f, 1.0f); - ASSIGN_4V(machine->internal[IMM_INV_255], inv, inv, inv, inv); - ASSIGN_4V(machine->internal[IMM_255], f255, f255, f255, f255); - ASSIGN_4V(machine->internal[IMM_RSQ], -.5f, 1.5f, 0.0f, 0.0f); - - - machine->fpu_rnd_nearest = (X87_CW_EXCEPTION_INV_OP | - X87_CW_EXCEPTION_DENORM_OP | - X87_CW_EXCEPTION_ZERO_DIVIDE | - X87_CW_EXCEPTION_OVERFLOW | - X87_CW_EXCEPTION_UNDERFLOW | - X87_CW_EXCEPTION_PRECISION | - (1<<6) | - X87_CW_ROUND_NEAREST | - X87_CW_PRECISION_DOUBLE_EXT); - - assert(machine->fpu_rnd_nearest == 0x37f); - - machine->fpu_rnd_neg_inf = (X87_CW_EXCEPTION_INV_OP | - X87_CW_EXCEPTION_DENORM_OP | - X87_CW_EXCEPTION_ZERO_DIVIDE | - X87_CW_EXCEPTION_OVERFLOW | - X87_CW_EXCEPTION_UNDERFLOW | - X87_CW_EXCEPTION_PRECISION | - (1<<6) | - X87_CW_ROUND_DOWN | - X87_CW_PRECISION_DOUBLE_EXT); - - for (i = 0; i < MAX_SHINE_TAB; i++) - do_populate_lut( &machine->shine_tab[i], 1.0f ); -} static void spill( struct aos_compilation *cp, unsigned idx ) @@ -1220,136 +1157,6 @@ static boolean emit_FRC( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } -static void PIPE_CDECL do_lit( struct aos_machine *machine, - float *result, - const float *in, - unsigned count ) -{ - if (in[0] > 0) - { - if (in[1] <= 0.0) - { - result[0] = 1.0F; - result[1] = in[0]; - result[2] = 1.0; - result[3] = 1.0F; - } - else - { - const float epsilon = 1.0F / 256.0F; - float exponent = CLAMP(in[3], -(128.0F - epsilon), (128.0F - epsilon)); - result[0] = 1.0F; - result[1] = in[0]; - result[2] = powf(in[1], exponent); - result[3] = 1.0; - } - } - else - { - result[0] = 1.0F; - result[1] = 0.0; - result[2] = 0.0; - result[3] = 1.0F; - } -} - - -static void PIPE_CDECL do_lit_lut( struct aos_machine *machine, - float *result, - const float *in, - unsigned count ) -{ - if (in[0] > 0) - { - if (in[1] <= 0.0) - { - result[0] = 1.0F; - result[1] = in[0]; - result[2] = 1.0; - result[3] = 1.0F; - return; - } - - if (machine->lit_info[count].shine_tab->exponent != in[3]) { - machine->lit_info[count].func = do_lit; - goto no_luck; - } - - if (in[1] <= 1.0) - { - const float *tab = machine->lit_info[count].shine_tab->values; - float f = in[1] * 256; - int k = (int)f; - float frac = f - (float)k; - - result[0] = 1.0F; - result[1] = in[0]; - result[2] = tab[k] + frac*(tab[k+1]-tab[k]); - result[3] = 1.0; - return; - } - - no_luck: - { - const float epsilon = 1.0F / 256.0F; - float exponent = CLAMP(in[3], -(128.0F - epsilon), (128.0F - epsilon)); - result[0] = 1.0F; - result[1] = in[0]; - result[2] = powf(in[1], exponent); - result[3] = 1.0; - } - } - else - { - result[0] = 1.0F; - result[1] = 0.0; - result[2] = 0.0; - result[3] = 1.0F; - } -} - - - -static void PIPE_CDECL populate_lut( struct aos_machine *machine, - float *result, - const float *in, - unsigned count ) -{ - unsigned i, tab; - - /* Search for an existing table for this value. Note that without - * static analysis we don't really know if in[3] will be constant, - * but it usually is... - */ - for (tab = 0; tab < 4; tab++) { - if (machine->shine_tab[tab].exponent == in[3]) { - goto found; - } - } - - for (tab = 0, i = 1; i < 4; i++) { - if (machine->shine_tab[i].last_used < machine->shine_tab[tab].last_used) - tab = i; - } - - if (machine->shine_tab[tab].last_used == machine->now) { - /* No unused tables (this is not a ffvertex program...). Just - * call pow each time: - */ - machine->lit_info[count].func = do_lit; - machine->lit_info[count].func( machine, result, in, count ); - return; - } - else { - do_populate_lut( &machine->shine_tab[tab], in[3] ); - } - - found: - machine->shine_tab[tab].last_used = machine->now; - machine->lit_info[count].shine_tab = &machine->shine_tab[tab]; - machine->lit_info[count].func = do_lit_lut; - machine->lit_info[count].func( machine, result, in, count ); -} @@ -1413,7 +1220,7 @@ static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_inst Offset(struct lit_info, func))); } else { - x86_mov_reg_imm( cp->func, ecx, (int)do_lit ); + x86_mov_reg_imm( cp->func, ecx, (int)aos_do_lit ); } x86_call( cp->func, ecx ); @@ -1434,7 +1241,7 @@ static boolean emit_LIT( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } - +#if 0 static boolean emit_inline_LIT( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { struct x86_reg dst = get_dst_ptr(cp, &op->FullDstRegisters[0]); @@ -1495,6 +1302,7 @@ static boolean emit_inline_LIT( struct aos_compilation *cp, const struct tgsi_fu return TRUE; } +#endif @@ -1945,7 +1753,7 @@ static void find_last_write_outputs( struct aos_compilation *cp ) } -#define ARG_VARIENT 1 +#define ARG_MACHINE 1 #define ARG_START_ELTS 2 #define ARG_COUNT 3 #define ARG_OUTBUF 4 @@ -1985,7 +1793,7 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, /* Load arguments into regs: */ - x86_mov(cp.func, cp.machine_EDX, x86_fn_arg(cp.func, ARG_VARIENT)); + x86_mov(cp.func, cp.machine_EDX, x86_fn_arg(cp.func, ARG_MACHINE)); x86_mov(cp.func, cp.idx_EBX, x86_fn_arg(cp.func, ARG_START_ELTS)); x86_mov(cp.func, cp.count_ESI, x86_fn_arg(cp.func, ARG_COUNT)); x86_mov(cp.func, cp.outbuf_ECX, x86_fn_arg(cp.func, ARG_OUTBUF)); @@ -1997,11 +1805,6 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, x86_cmp(cp.func, cp.count_ESI, cp.tmp_EAX); fixup = x86_jcc_forward(cp.func, cc_E); - /* Dig out the machine pointer from inside the varient arg - */ - x86_mov(cp.func, cp.machine_EDX, - x86_make_disp(cp.machine_EDX, - Offset( struct draw_vs_varient_aos_sse, machine ))); save_fpu_state( &cp ); set_fpu_round_nearest( &cp ); @@ -2151,13 +1954,14 @@ static void PIPE_CDECL vaos_run_elts( struct draw_vs_varient *varient, void *output_buffer ) { struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + struct aos_machine *machine = vaos->draw->vs.aos_machine; - vaos->machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; - vaos->machine->constants = vaos->draw->pt.user.constants; - vaos->machine->immediates = vaos->base.vs->immediates; - vaos->machine->attrib = vaos->attrib; + machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; + machine->constants = (const float (*)[4])vaos->draw->pt.user.constants; + machine->immediates = vaos->base.vs->immediates; + machine->attrib = vaos->attrib; - vaos->gen_run_elts( varient, + vaos->gen_run_elts( machine, elts, count, output_buffer ); @@ -2169,61 +1973,25 @@ static void PIPE_CDECL vaos_run_linear( struct draw_vs_varient *varient, void *output_buffer ) { struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; + struct aos_machine *machine = vaos->draw->vs.aos_machine; - vaos->machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; - vaos->machine->constants = vaos->draw->pt.user.constants; - vaos->machine->immediates = vaos->base.vs->immediates; - vaos->machine->attrib = vaos->attrib; + machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; + machine->constants = (const float (*)[4])vaos->draw->pt.user.constants; + machine->immediates = vaos->base.vs->immediates; + machine->attrib = vaos->attrib; - vaos->gen_run_linear( varient, + vaos->gen_run_linear( machine, start, count, output_buffer ); } -static void vaos_set_constants( struct draw_vs_varient *varient, - const float (*constants)[4] ) -{ - struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; - -#if 0 - unsigned i; - for (i =0; i < vaos->base.vs->info.file_max[TGSI_FILE_CONSTANT] + 1; i++) - debug_printf("state %d: %f %f %f %f\n", - i, - constants[i][0], - constants[i][1], - constants[i][2], - constants[i][3]); -#endif - - { - unsigned i; - for (i = 0; i < MAX_LIT_INFO; i++) { - vaos->machine->lit_info[i].func = populate_lut; - vaos->machine->now++; - } - } -} - - -static void vaos_set_viewport( struct draw_vs_varient *varient, - const struct pipe_viewport_state *viewport ) -{ - struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; - - memcpy(vaos->machine->scale, viewport->scale, 4 * sizeof(float)); - memcpy(vaos->machine->translate, viewport->translate, 4 * sizeof(float)); -} static void vaos_destroy( struct draw_vs_varient *varient ) { struct draw_vs_varient_aos_sse *vaos = (struct draw_vs_varient_aos_sse *)varient; - if (vaos->machine) - align_free( vaos->machine ); - FREE( vaos->attrib ); x86_release_func( &vaos->func[0] ); @@ -2245,8 +2013,6 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, vaos->base.key = *key; vaos->base.vs = vs; vaos->base.set_input = vaos_set_buffer; - vaos->base.set_constants = vaos_set_constants; - vaos->base.set_viewport = vaos_set_viewport; vaos->base.destroy = vaos_destroy; vaos->base.run_linear = vaos_run_linear; vaos->base.run_elts = vaos_run_elts; @@ -2257,13 +2023,6 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, if (!vaos->attrib) goto fail; - vaos->machine = align_malloc( sizeof(struct aos_machine), 16 ); - if (!vaos->machine) - goto fail; - - memset(vaos->machine, 0, sizeof(struct aos_machine)); - init_internals(vaos->machine); - tgsi_dump(vs->state.tokens, 0); if (!build_vertex_program( vaos, TRUE )) @@ -2272,11 +2031,11 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, if (!build_vertex_program( vaos, FALSE )) goto fail; - vaos->gen_run_linear = (vsv_run_linear_func)x86_get_func(&vaos->func[0]); + vaos->gen_run_linear = (vaos_run_linear_func)x86_get_func(&vaos->func[0]); if (!vaos->gen_run_linear) goto fail; - vaos->gen_run_elts = (vsv_run_elts_func)x86_get_func(&vaos->func[1]); + vaos->gen_run_elts = (vaos_run_elts_func)x86_get_func(&vaos->func[1]); if (!vaos->gen_run_elts) goto fail; @@ -2286,9 +2045,6 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, if (vaos && vaos->attrib) FREE(vaos->attrib); - if (vaos && vaos->machine) - align_free( vaos->machine ); - if (vaos) x86_release_func( &vaos->func[0] ); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index 295d2cb3fe..89a9174151 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -60,10 +60,16 @@ struct x86_function; #define FPU_RND_NEAREST 2 struct aos_machine; -typedef void (PIPE_CDECL *lit_func)( struct aos_machine *, +typedef void PIPE_CDECL (*lit_func)( struct aos_machine *, float *result, const float *in, unsigned count ); + +PIPE_CDECL void aos_do_lit( struct aos_machine *machine, + float *result, + const float *in, + unsigned count ); + struct shine_tab { float exponent; float values[258]; @@ -207,16 +213,25 @@ struct x86_reg aos_get_x86( struct aos_compilation *cp, unsigned value ); +typedef void (PIPE_CDECL *vaos_run_elts_func)( struct aos_machine *, + const unsigned *elts, + unsigned count, + void *output_buffer); + +typedef void (PIPE_CDECL *vaos_run_linear_func)( struct aos_machine *, + unsigned start, + unsigned count, + void *output_buffer); + + struct draw_vs_varient_aos_sse { struct draw_vs_varient base; struct draw_context *draw; struct aos_attrib *attrib; - struct aos_machine *machine; /* XXX: temporarily unshared */ - - vsv_run_linear_func gen_run_linear; - vsv_run_elts_func gen_run_elts; + vaos_run_linear_func gen_run_linear; + vaos_run_elts_func gen_run_elts; struct x86_function func[2]; diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c new file mode 100644 index 0000000000..53e999b191 --- /dev/null +++ b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c @@ -0,0 +1,297 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_parse.h" +#include "tgsi/util/tgsi_util.h" +#include "tgsi/exec/tgsi_exec.h" +#include "draw_vs.h" +#include "draw_vs_aos.h" +#include "draw_vertex.h" + +#include "rtasm/rtasm_x86sse.h" + + +#define X87_CW_EXCEPTION_INV_OP (1<<0) +#define X87_CW_EXCEPTION_DENORM_OP (1<<1) +#define X87_CW_EXCEPTION_ZERO_DIVIDE (1<<2) +#define X87_CW_EXCEPTION_OVERFLOW (1<<3) +#define X87_CW_EXCEPTION_UNDERFLOW (1<<4) +#define X87_CW_EXCEPTION_PRECISION (1<<5) +#define X87_CW_PRECISION_SINGLE (0<<8) +#define X87_CW_PRECISION_RESERVED (1<<8) +#define X87_CW_PRECISION_DOUBLE (2<<8) +#define X87_CW_PRECISION_DOUBLE_EXT (3<<8) +#define X87_CW_PRECISION_MASK (3<<8) +#define X87_CW_ROUND_NEAREST (0<<10) +#define X87_CW_ROUND_DOWN (1<<10) +#define X87_CW_ROUND_UP (2<<10) +#define X87_CW_ROUND_ZERO (3<<10) +#define X87_CW_ROUND_MASK (3<<10) +#define X87_CW_INFINITY (1<<12) + + +PIPE_CDECL void aos_do_lit( struct aos_machine *machine, + float *result, + const float *in, + unsigned count ) +{ + if (in[0] > 0) + { + if (in[1] <= 0.0) + { + result[0] = 1.0F; + result[1] = in[0]; + result[2] = 1.0; + result[3] = 1.0F; + } + else + { + const float epsilon = 1.0F / 256.0F; + float exponent = CLAMP(in[3], -(128.0F - epsilon), (128.0F - epsilon)); + result[0] = 1.0F; + result[1] = in[0]; + result[2] = powf(in[1], exponent); + result[3] = 1.0; + } + } + else + { + result[0] = 1.0F; + result[1] = 0.0; + result[2] = 0.0; + result[3] = 1.0F; + } +} + + +static PIPE_CDECL void do_lit_lut( struct aos_machine *machine, + float *result, + const float *in, + unsigned count ) +{ + if (in[0] > 0) + { + if (in[1] <= 0.0) + { + result[0] = 1.0F; + result[1] = in[0]; + result[2] = 1.0; + result[3] = 1.0F; + return; + } + + if (machine->lit_info[count].shine_tab->exponent != in[3]) { + machine->lit_info[count].func = aos_do_lit; + goto no_luck; + } + + if (in[1] <= 1.0) + { + const float *tab = machine->lit_info[count].shine_tab->values; + float f = in[1] * 256; + int k = (int)f; + float frac = f - (float)k; + + result[0] = 1.0F; + result[1] = in[0]; + result[2] = tab[k] + frac*(tab[k+1]-tab[k]); + result[3] = 1.0; + return; + } + + no_luck: + { + const float epsilon = 1.0F / 256.0F; + float exponent = CLAMP(in[3], -(128.0F - epsilon), (128.0F - epsilon)); + result[0] = 1.0F; + result[1] = in[0]; + result[2] = powf(in[1], exponent); + result[3] = 1.0; + } + } + else + { + result[0] = 1.0F; + result[1] = 0.0; + result[2] = 0.0; + result[3] = 1.0F; + } +} + + +static void do_populate_lut( struct shine_tab *tab, + float unclamped_exponent ) +{ + const float epsilon = 1.0F / 256.0F; + float exponent = CLAMP(unclamped_exponent, -(128.0F - epsilon), (128.0F - epsilon)); + unsigned i; + + tab->exponent = unclamped_exponent; /* for later comparison */ + + tab->values[0] = 0; + if (exponent == 0) { + for (i = 1; i < 258; i++) { + tab->values[i] = 1.0; + } + } + else { + for (i = 1; i < 258; i++) { + tab->values[i] = powf((float)i * epsilon, exponent); + } + } +} + + + + +static void PIPE_CDECL populate_lut( struct aos_machine *machine, + float *result, + const float *in, + unsigned count ) +{ + unsigned i, tab; + + /* Search for an existing table for this value. Note that without + * static analysis we don't really know if in[3] will be constant, + * but it usually is... + */ + for (tab = 0; tab < 4; tab++) { + if (machine->shine_tab[tab].exponent == in[3]) { + goto found; + } + } + + for (tab = 0, i = 1; i < 4; i++) { + if (machine->shine_tab[i].last_used < machine->shine_tab[tab].last_used) + tab = i; + } + + if (machine->shine_tab[tab].last_used == machine->now) { + /* No unused tables (this is not a ffvertex program...). Just + * call pow each time: + */ + machine->lit_info[count].func = aos_do_lit; + machine->lit_info[count].func( machine, result, in, count ); + return; + } + else { + do_populate_lut( &machine->shine_tab[tab], in[3] ); + } + + found: + machine->shine_tab[tab].last_used = machine->now; + machine->lit_info[count].shine_tab = &machine->shine_tab[tab]; + machine->lit_info[count].func = do_lit_lut; + machine->lit_info[count].func( machine, result, in, count ); +} + + +void draw_vs_aos_machine_constants( struct aos_machine *machine, + const float (*constants)[4] ) +{ + machine->constants = constants; + + { + unsigned i; + for (i = 0; i < MAX_LIT_INFO; i++) { + machine->lit_info[i].func = populate_lut; + machine->now++; + } + } +} + + +void draw_vs_aos_machine_viewport( struct aos_machine *machine, + const struct pipe_viewport_state *viewport ) +{ + memcpy(machine->scale, viewport->scale, 4 * sizeof(float)); + memcpy(machine->translate, viewport->translate, 4 * sizeof(float)); +} + + + +void draw_vs_aos_machine_destroy( struct aos_machine *machine ) +{ + align_free(machine); +} + +struct aos_machine *draw_vs_aos_machine( void ) +{ + struct aos_machine *machine; + unsigned i; + float inv = 1.0f/255.0f; + float f255 = 255.0f; + + machine = align_malloc(sizeof(struct aos_machine), 16); + if (!machine) + return NULL; + + memset(machine, 0, sizeof(*machine)); + + ASSIGN_4V(machine->internal[IMM_SWZ], 1.0f, -1.0f, 0.0f, 1.0f); + *(unsigned *)&machine->internal[IMM_SWZ][3] = 0xffffffff; + + ASSIGN_4V(machine->internal[IMM_ONES], 1.0f, 1.0f, 1.0f, 1.0f); + ASSIGN_4V(machine->internal[IMM_NEGS], -1.0f, -1.0f, -1.0f, -1.0f); + ASSIGN_4V(machine->internal[IMM_IDENTITY], 0.0f, 0.0f, 0.0f, 1.0f); + ASSIGN_4V(machine->internal[IMM_INV_255], inv, inv, inv, inv); + ASSIGN_4V(machine->internal[IMM_255], f255, f255, f255, f255); + ASSIGN_4V(machine->internal[IMM_RSQ], -.5f, 1.5f, 0.0f, 0.0f); + + + machine->fpu_rnd_nearest = (X87_CW_EXCEPTION_INV_OP | + X87_CW_EXCEPTION_DENORM_OP | + X87_CW_EXCEPTION_ZERO_DIVIDE | + X87_CW_EXCEPTION_OVERFLOW | + X87_CW_EXCEPTION_UNDERFLOW | + X87_CW_EXCEPTION_PRECISION | + (1<<6) | + X87_CW_ROUND_NEAREST | + X87_CW_PRECISION_DOUBLE_EXT); + + assert(machine->fpu_rnd_nearest == 0x37f); + + machine->fpu_rnd_neg_inf = (X87_CW_EXCEPTION_INV_OP | + X87_CW_EXCEPTION_DENORM_OP | + X87_CW_EXCEPTION_ZERO_DIVIDE | + X87_CW_EXCEPTION_OVERFLOW | + X87_CW_EXCEPTION_UNDERFLOW | + X87_CW_EXCEPTION_PRECISION | + (1<<6) | + X87_CW_ROUND_DOWN | + X87_CW_PRECISION_DOUBLE_EXT); + + for (i = 0; i < MAX_SHINE_TAB; i++) + do_populate_lut( &machine->shine_tab[i], 1.0f ); + + return machine; +} + + diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index 784ae41205..18cb06e374 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -44,8 +44,6 @@ struct draw_vs_varient_generic { struct draw_vs_varient base; - struct pipe_viewport_state viewport; - struct draw_vertex_shader *shader; struct draw_context *draw; @@ -57,21 +55,11 @@ struct draw_vs_varient_generic { */ struct translate *fetch; struct translate *emit; - - const float (*constants)[4]; }; -static void vsvg_set_constants( struct draw_vs_varient *varient, - const float (*constants)[4] ) -{ - struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; - - vsvg->constants = constants; -} - static void vsvg_set_input( struct draw_vs_varient *varient, unsigned buffer, @@ -94,8 +82,8 @@ static void do_rhw_viewport( struct draw_vs_varient_generic *vsvg, void *output_buffer ) { char *ptr = (char *)output_buffer; - const float *scale = vsvg->viewport.scale; - const float *trans = vsvg->viewport.translate; + const float *scale = vsvg->base.vs->draw->viewport.scale; + const float *trans = vsvg->base.vs->draw->viewport.translate; unsigned stride = vsvg->base.key.output_stride; unsigned j; @@ -115,8 +103,8 @@ static void do_viewport( struct draw_vs_varient_generic *vsvg, void *output_buffer ) { char *ptr = (char *)output_buffer; - const float *scale = vsvg->viewport.scale; - const float *trans = vsvg->viewport.translate; + const float *scale = vsvg->base.vs->draw->viewport.scale; + const float *trans = vsvg->base.vs->draw->viewport.translate; unsigned stride = vsvg->base.key.output_stride; unsigned j; @@ -130,10 +118,10 @@ static void do_viewport( struct draw_vs_varient_generic *vsvg, } -static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, - const unsigned *elts, - unsigned count, - void *output_buffer ) +static void vsvg_run_elts( struct draw_vs_varient *varient, + const unsigned *elts, + unsigned count, + void *output_buffer) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; @@ -150,7 +138,7 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, vsvg->base.vs->run_linear( vsvg->base.vs, output_buffer, output_buffer, - vsvg->constants, + (const float (*)[4])vsvg->base.vs->draw->pt.user.constants, count, vsvg->base.key.output_stride, vsvg->base.key.output_stride); @@ -186,10 +174,10 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, } -static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, - unsigned start, - unsigned count, - void *output_buffer ) +static void vsvg_run_linear( struct draw_vs_varient *varient, + unsigned start, + unsigned count, + void *output_buffer ) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; @@ -206,7 +194,7 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, vsvg->base.vs->run_linear( vsvg->base.vs, output_buffer, output_buffer, - vsvg->constants, + (const float (*)[4])vsvg->base.vs->draw->pt.user.constants, count, vsvg->base.key.output_stride, vsvg->base.key.output_stride); @@ -245,13 +233,6 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, -static void vsvg_set_viewport( struct draw_vs_varient *varient, - const struct pipe_viewport_state *viewport ) -{ - struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; - - vsvg->viewport = *viewport; -} static void vsvg_destroy( struct draw_vs_varient *varient ) { @@ -272,8 +253,6 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, vsvg->base.key = *key; vsvg->base.vs = vs; vsvg->base.set_input = vsvg_set_input; - vsvg->base.set_constants = vsvg_set_constants; - vsvg->base.set_viewport = vsvg_set_viewport; vsvg->base.run_elts = vsvg_run_elts; vsvg->base.run_linear = vsvg_run_linear; vsvg->base.destroy = vsvg_destroy; -- cgit v1.2.3 From 6945bcb89370501e0a218bc656e68e30e4dadcda Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 29 May 2008 18:28:02 +0900 Subject: draw: Put PIPE_CDECL in the right places. MSVC seems picky about this. --- src/gallium/auxiliary/draw/draw_vs_aos.h | 4 ++-- src/gallium/auxiliary/draw/draw_vs_aos_machine.c | 4 ++-- src/gallium/auxiliary/draw/draw_vs_varient.c | 16 ++++++++-------- 3 files changed, 12 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index 89a9174151..665b425e68 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -60,12 +60,12 @@ struct x86_function; #define FPU_RND_NEAREST 2 struct aos_machine; -typedef void PIPE_CDECL (*lit_func)( struct aos_machine *, +typedef void (PIPE_CDECL *lit_func)( struct aos_machine *, float *result, const float *in, unsigned count ); -PIPE_CDECL void aos_do_lit( struct aos_machine *machine, +void PIPE_CDECL aos_do_lit( struct aos_machine *machine, float *result, const float *in, unsigned count ); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c index 53e999b191..b7864e9f2f 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c @@ -57,7 +57,7 @@ #define X87_CW_INFINITY (1<<12) -PIPE_CDECL void aos_do_lit( struct aos_machine *machine, +void PIPE_CDECL aos_do_lit( struct aos_machine *machine, float *result, const float *in, unsigned count ) @@ -91,7 +91,7 @@ PIPE_CDECL void aos_do_lit( struct aos_machine *machine, } -static PIPE_CDECL void do_lit_lut( struct aos_machine *machine, +static void PIPE_CDECL do_lit_lut( struct aos_machine *machine, float *result, const float *in, unsigned count ) diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index 18cb06e374..4a155251b5 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -118,10 +118,10 @@ static void do_viewport( struct draw_vs_varient_generic *vsvg, } -static void vsvg_run_elts( struct draw_vs_varient *varient, - const unsigned *elts, - unsigned count, - void *output_buffer) +static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, + const unsigned *elts, + unsigned count, + void *output_buffer) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; @@ -174,10 +174,10 @@ static void vsvg_run_elts( struct draw_vs_varient *varient, } -static void vsvg_run_linear( struct draw_vs_varient *varient, - unsigned start, - unsigned count, - void *output_buffer ) +static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, + unsigned start, + unsigned count, + void *output_buffer ) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; -- cgit v1.2.3 From cb87d7e44a6d6b1b4239b4e38c76c6bb848d2ef6 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 29 May 2008 12:11:37 +0200 Subject: scons: Remove duplicate entry. --- src/gallium/auxiliary/draw/SConscript | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/SConscript b/src/gallium/auxiliary/draw/SConscript index 925e668f22..544a04918b 100644 --- a/src/gallium/auxiliary/draw/SConscript +++ b/src/gallium/auxiliary/draw/SConscript @@ -31,7 +31,6 @@ draw = env.ConvenienceLibrary( 'draw_pt_util.c', 'draw_pt_varray.c', 'draw_pt_vcache.c', - 'draw_pt_util.c', 'draw_vertex.c', 'draw_vs.c', 'draw_vs_aos.c', -- cgit v1.2.3 From 82605d7bcd533d7c96cc619c45970efd7229dc3b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 29 May 2008 11:46:43 +0100 Subject: draw: draw_range_elements trial --- src/gallium/auxiliary/draw/draw_context.c | 20 +++- src/gallium/auxiliary/draw/draw_context.h | 10 +- src/gallium/auxiliary/draw/draw_private.h | 2 + src/gallium/auxiliary/draw/draw_pt.h | 9 ++ src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 48 ++++++++ .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 49 ++++++++ .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 80 +++++++++++++- src/gallium/auxiliary/draw/draw_pt_vcache.c | 123 ++++++++++++++++++++- src/gallium/drivers/softpipe/sp_context.c | 1 + src/gallium/drivers/softpipe/sp_draw_arrays.c | 31 +++++- src/gallium/drivers/softpipe/sp_state.h | 7 ++ src/gallium/include/pipe/p_context.h | 14 +++ src/mesa/state_tracker/st_draw.c | 27 ++++- 13 files changed, 404 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 8509baf865..bcec85c2ef 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -348,14 +348,30 @@ void draw_set_edgeflags( struct draw_context *draw, * \param elements the element buffer ptr */ void -draw_set_mapped_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ) +draw_set_mapped_element_buffer_range( struct draw_context *draw, + unsigned eltSize, + unsigned min_index, + unsigned max_index, + void *elements ) { draw->pt.user.elts = elements; draw->pt.user.eltSize = eltSize; + draw->pt.user.min_index = min_index; + draw->pt.user.max_index = max_index; } +void +draw_set_mapped_element_buffer( struct draw_context *draw, + unsigned eltSize, + void *elements ) +{ + draw->pt.user.elts = elements; + draw->pt.user.eltSize = eltSize; + draw->pt.user.min_index = 0; + draw->pt.user.max_index = 0xffffffff; +} + /* Revamp me please: */ diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index c5c3d3b09e..8dd03cb79e 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -118,8 +118,16 @@ void draw_set_vertex_elements(struct draw_context *draw, unsigned count, const struct pipe_vertex_element *elements); +void +draw_set_mapped_element_buffer_range( struct draw_context *draw, + unsigned eltSize, + unsigned min_index, + unsigned max_index, + void *elements ); + void draw_set_mapped_element_buffer( struct draw_context *draw, - unsigned eltSize, void *elements ); + unsigned eltSize, + void *elements ); void draw_set_mapped_vertex_buffer(struct draw_context *draw, unsigned attr, const void *buffer); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 4cbccc8b5b..40f1d978f2 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -147,6 +147,8 @@ struct draw_context const void *elts; /** bytes per index (0, 1, 2 or 4) */ unsigned eltSize; + unsigned min_index; + unsigned max_index; /** vertex arrays */ const void *vbuffer[PIPE_MAX_ATTRIBS]; diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index e03816ebbc..6b8ba1d171 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -96,6 +96,15 @@ struct draw_pt_middle_end { unsigned start, unsigned count); + /* Transform all vertices in a linear range and then draw them with + * the supplied element list. + */ + void (*run_linear_elts)( struct draw_pt_middle_end *, + unsigned fetch_start, + unsigned fetch_count, + const ushort *draw_elts, + unsigned draw_count ); + void (*finish)( struct draw_pt_middle_end * ); void (*destroy)( struct draw_pt_middle_end * ); }; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index a1d041a74f..09bdc5fb5e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -311,6 +311,53 @@ static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, } +static void fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle, + unsigned start, + unsigned count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; + struct draw_context *draw = feme->draw; + void *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)feme->translate->key.output_stride, + (ushort)count ); + if (!hw_verts) { + assert(0); + return; + } + + /* Single routine to fetch vertices and emit HW verts. + */ + feme->translate->run( feme->translate, + start, + count, + hw_verts ); + + /* XXX: Draw arrays path to avoid re-emitting index list again and + * again. + */ + draw->render->draw( draw->render, + draw_elts, + draw_count ); + + /* Done -- that was easy, wasn't it: + */ + draw->render->release_vertices( draw->render, + hw_verts, + feme->translate->key.output_stride, + count ); + +} + + + static void fetch_emit_finish( struct draw_pt_middle_end *middle ) { @@ -343,6 +390,7 @@ struct draw_pt_middle_end *draw_pt_fetch_emit( struct draw_context *draw ) fetch_emit->base.prepare = fetch_emit_prepare; fetch_emit->base.run = fetch_emit_run; fetch_emit->base.run_linear = fetch_emit_run_linear; + fetch_emit->base.run_linear_elts = fetch_emit_run_linear_elts; fetch_emit->base.finish = fetch_emit_finish; fetch_emit->base.destroy = fetch_emit_destroy; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 5265a13160..efa6dddbda 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -310,6 +310,54 @@ fse_run(struct draw_pt_middle_end *middle, } + +static void fse_run_linear_elts( struct draw_pt_middle_end *middle, + unsigned start, + unsigned count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; + struct draw_context *draw = fse->draw; + unsigned alloc_count = align(count, 4); + char *hw_verts; + + /* XXX: need to flush to get prim_vbuf.c to release its allocation?? + */ + draw_do_flush( draw, DRAW_FLUSH_BACKEND ); + + hw_verts = draw->render->allocate_vertices( draw->render, + (ushort)fse->key.output_stride, + (ushort)alloc_count ); + + if (!hw_verts) { + assert(0); + return; + } + + /* Single routine to fetch vertices, run shader and emit HW verts. + * Clipping is done elsewhere -- either by the API or on hardware, + * or for some other reason not required... + */ + fse->active->run_linear( fse->active, + start, count, + hw_verts ); + + + draw->render->draw( draw->render, + draw_elts, + draw_count ); + + + + draw->render->release_vertices( draw->render, + hw_verts, + fse->key.output_stride, + count ); +} + + + static void fse_finish( struct draw_pt_middle_end *middle ) { } @@ -330,6 +378,7 @@ struct draw_pt_middle_end *draw_pt_middle_fse( struct draw_context *draw ) fse->base.prepare = fse_prepare; fse->base.run = fse_run; fse->base.run_linear = fse_run_linear; + fse->base.run_linear_elts = fse_run_linear_elts; fse->base.finish = fse_finish; fse->base.destroy = fse_destroy; fse->draw = draw; diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 06718779a5..c58a900867 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -98,7 +98,6 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, - static void fetch_pipeline_run( struct draw_pt_middle_end *middle, const unsigned *fetch_elts, unsigned fetch_count, @@ -251,6 +250,84 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, +static void fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle, + unsigned start, + unsigned count, + const ushort *draw_elts, + unsigned draw_count ) +{ + struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; + struct draw_context *draw = fpme->draw; + struct draw_vertex_shader *shader = draw->vs.vertex_shader; + unsigned opt = fpme->opt; + unsigned alloc_count = align_int( count, 4 ); + + struct vertex_header *pipeline_verts = + (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count); + + if (!pipeline_verts) { + /* Not much we can do here - just skip the rendering. + */ + assert(0); + return; + } + + /* Fetch into our vertex buffer + */ + draw_pt_fetch_run_linear( fpme->fetch, + start, + count, + (char *)pipeline_verts ); + + /* Run the shader, note that this overwrites the data[] parts of + * the pipeline verts. If there is no shader, ie a bypass shader, + * then the inputs == outputs, and are already in the correct + * place. + */ + if (opt & PT_SHADE) + { + shader->run_linear(shader, + (const float (*)[4])pipeline_verts->data, + ( float (*)[4])pipeline_verts->data, + (const float (*)[4])draw->pt.user.constants, + count, + fpme->vertex_size, + fpme->vertex_size); + } + + if (draw_pt_post_vs_run( fpme->post_vs, + pipeline_verts, + count, + fpme->vertex_size )) + { + opt |= PT_PIPELINE; + } + + /* Do we need to run the pipeline? + */ + if (opt & PT_PIPELINE) { + draw_pipeline_run( fpme->draw, + fpme->prim, + pipeline_verts, + count, + fpme->vertex_size, + draw_elts, + draw_count ); + } + else { + draw_pt_emit( fpme->emit, + (const float (*)[4])pipeline_verts->data, + count, + fpme->vertex_size, + draw_elts, + draw_count ); + } + + FREE(pipeline_verts); +} + + + static void fetch_pipeline_finish( struct draw_pt_middle_end *middle ) { /* nothing to do */ @@ -282,6 +359,7 @@ struct draw_pt_middle_end *draw_pt_fetch_pipeline_or_emit( struct draw_context * fpme->base.prepare = fetch_pipeline_prepare; fpme->base.run = fetch_pipeline_run; fpme->base.run_linear = fetch_pipeline_linear_run; + fpme->base.run_linear_elts = fetch_pipeline_linear_run_elts; fpme->base.finish = fetch_pipeline_finish; fpme->base.destroy = fetch_pipeline_destroy; diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 96e02fbf3a..720b91b8e6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -36,8 +36,8 @@ #include "draw/draw_pt.h" -#define CACHE_MAX 32 -#define FETCH_MAX 128 +#define CACHE_MAX 1024 +#define FETCH_MAX 4096 #define DRAW_MAX (16*1024) struct vcache_frontend { @@ -201,7 +201,124 @@ static void vcache_ef_quad( struct vcache_frontend *vcache, #define FUNC vcache_run #include "draw_pt_vcache_tmp.h" +static void translate_uint_elts( const unsigned *src, + unsigned count, + int delta, + ushort *dest ) +{ + unsigned i; + + for (i = 0; i < count; i++) + dest[i] = (ushort)(src[i] + delta); +} + +static void translate_ushort_elts( const ushort *src, + unsigned count, + int delta, + ushort *dest ) +{ + unsigned i; + + for (i = 0; i < count; i++) + dest[i] = (ushort)(src[i] + delta); +} +static void translate_ubyte_elts( const ubyte *src, + unsigned count, + int delta, + ushort *dest ) +{ + unsigned i; + + for (i = 0; i < count; i++) + dest[i] = (ushort)(src[i] + delta); +} + +#if 0 +static enum pipe_format format_from_get_elt( pt_elt_func get_elt ) +{ + switch (draw->pt.user.eltSize) { + case 1: return PIPE_FORMAT_R8_UNORM; + case 2: return PIPE_FORMAT_R16_UNORM; + case 4: return PIPE_FORMAT_R32_UNORM; + default: return PIPE_FORMAT_NONE; + } +} +#endif + +static void vcache_check_run( struct draw_pt_front_end *frontend, + pt_elt_func get_elt, + const void *elts, + unsigned draw_count ) +{ + struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; + struct draw_context *draw = vcache->draw; + unsigned min_index = draw->pt.user.min_index; + unsigned max_index = draw->pt.user.max_index; + unsigned index_size = draw->pt.user.eltSize; + unsigned fetch_count = MAX2(max_index, max_index + 1 - min_index); + const ushort *transformed_elts; + ushort *storage = NULL; + + printf("fetch_count %x\n", fetch_count); + + if (fetch_count >= FETCH_MAX || + fetch_count > draw_count) + goto fail; + + + if (min_index == 0 && + index_size == 2) + { + transformed_elts = (const ushort *)elts; + } + else + { + storage = MALLOC( draw_count * sizeof(ushort) ); + if (!storage) + goto fail; + + switch(index_size) { + case 1: + translate_ubyte_elts( (const ubyte *)elts, + draw_count, + 0 - (int)min_index, + storage ); + break; + + case 2: + translate_ushort_elts( (const ushort *)elts, + draw_count, + 0 - (int)min_index, + storage ); + break; + + case 4: + translate_uint_elts( (const uint *)elts, + draw_count, + 0 - (int)min_index, + storage ); + break; + + default: + assert(0); + return; + } + transformed_elts = storage; + } + + vcache->middle->run_linear_elts( vcache->middle, + min_index, /* start */ + fetch_count, + transformed_elts, + draw_count ); + + FREE(storage); + return; + + fail: + vcache_run( frontend, get_elt, elts, draw_count ); +} @@ -219,7 +336,7 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, } else { - vcache->base.run = vcache_run; + vcache->base.run = vcache_check_run; } vcache->input_prim = prim; diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 045a1f74a9..1e0106b86c 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -179,6 +179,7 @@ softpipe_create( struct pipe_screen *screen, softpipe->pipe.draw_arrays = softpipe_draw_arrays; softpipe->pipe.draw_elements = softpipe_draw_elements; + softpipe->pipe.draw_range_elements = softpipe_draw_range_elements; softpipe->pipe.set_edgeflags = softpipe_set_edgeflags; diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index 6c58f9909d..dbecf6865f 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -108,11 +108,14 @@ softpipe_draw_arrays(struct pipe_context *pipe, unsigned mode, * * XXX should the element buffer be specified/bound with a separate function? */ + boolean -softpipe_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count) +softpipe_draw_range_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned min_index, + unsigned max_index, + unsigned mode, unsigned start, unsigned count) { struct softpipe_context *sp = softpipe_context(pipe); struct draw_context *draw = sp->draw; @@ -141,11 +144,14 @@ softpipe_draw_elements(struct pipe_context *pipe, void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + draw_set_mapped_element_buffer_range(draw, indexSize, + min_index, + max_index, + mapped_indexes); } else { /* no index/element buffer */ - draw_set_mapped_element_buffer(draw, 0, NULL); + draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL); } @@ -171,6 +177,19 @@ softpipe_draw_elements(struct pipe_context *pipe, return TRUE; } +boolean +softpipe_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + return softpipe_draw_range_elements( pipe, indexBuffer, + indexSize, + 0, 0xffffffff, + mode, start, count ); +} + + void softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags) diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 452e51fa79..701e02b295 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -171,6 +171,13 @@ boolean softpipe_draw_elements(struct pipe_context *pipe, struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count); +boolean +softpipe_draw_range_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned min_index, + unsigned max_index, + unsigned mode, unsigned start, unsigned count); void softpipe_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags); diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index 0f68f592f7..faf112c6d6 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -76,6 +76,20 @@ struct pipe_context { struct pipe_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count); + + /* XXX: this is (probably) a temporary entrypoint, as the range + * information should be available from the vertex_buffer state. + * Using this to quickly evaluate a specialized path in the draw + * module. + */ + boolean (*draw_range_elements)( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned mode, + unsigned start, + unsigned count); /*@}*/ diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index a3bffbfc95..551860452a 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -365,14 +365,33 @@ st_draw_vbo(GLcontext *ctx, } /* draw */ - for (i = 0; i < nr_prims; i++) { + if (nr_prims == 1 && pipe->draw_range_elements != NULL) { + i = 0; + + /* XXX: exercise temporary path to pass min/max directly + * through to driver & draw module. These interfaces still + * need a bit of work... + */ setup_edgeflags(ctx, prims[i].mode, prims[i].start + indexOffset, prims[i].count, arrays[VERT_ATTRIB_EDGEFLAG]); - pipe->draw_elements(pipe, indexBuf, indexSize, - prims[i].mode, - prims[i].start + indexOffset, prims[i].count); + pipe->draw_range_elements(pipe, indexBuf, indexSize, + min_index, + max_index, + prims[i].mode, + prims[i].start + indexOffset, prims[i].count); + } + else { + for (i = 0; i < nr_prims; i++) { + setup_edgeflags(ctx, prims[i].mode, + prims[i].start + indexOffset, prims[i].count, + arrays[VERT_ATTRIB_EDGEFLAG]); + + pipe->draw_elements(pipe, indexBuf, indexSize, + prims[i].mode, + prims[i].start + indexOffset, prims[i].count); + } } pipe_reference_buffer(pipe, &indexBuf, NULL); -- cgit v1.2.3 From 0cd75a4c99ec63b514b6fbb53152858fa20f53ec Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 29 May 2008 12:54:01 +0200 Subject: i915: Get up to date with the latest EGL changes --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 809371fdeb..41b7eac4ed 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -588,7 +588,7 @@ drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) * plug in API functions. */ _EGLDriver * -_eglMain(_EGLDisplay *dpy) +_eglMain(_EGLDisplay *dpy, const char *args) { struct drm_driver *drm; -- cgit v1.2.3 From 837601af522b9d1e687040015672769392884fa1 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 29 May 2008 12:43:09 +0200 Subject: i915: Ops how did that get there --- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index 2e6c452cd1..b3e5f29c3e 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -220,7 +220,7 @@ power_of_two(unsigned x) value = value << 1; return value; } - /*pipe_buffer_reference(winsys, &i915_render->vbo, NULL);*/ + /** * Copied from xm_winsys.c */ -- cgit v1.2.3 From 79b67d8408d1ba5b2232791ae35e731a5953b52d Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 29 May 2008 12:44:53 +0200 Subject: i915: Remove last usage of surface_alloc_storage --- src/gallium/drivers/i915simple/i915_texture.c | 42 +++++++++--------------- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 31 +---------------- 2 files changed, 17 insertions(+), 56 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index df11ba0544..122f88fe3a 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -104,7 +104,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex, */ } - +#if 0 /* Hack it up to use the old winsys->surface_alloc_storage() * method for now: */ @@ -145,10 +145,7 @@ i915_displaytarget_layout(struct pipe_screen *screen, return tex->buffer != NULL; } - - - - +#endif static void i945_miptree_layout_2d( struct i915_texture *tex ) @@ -539,32 +536,25 @@ i915_texture_create(struct pipe_screen *screen, tex->base.refcount = 1; tex->base.screen = screen; - if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { - if (!i915_displaytarget_layout(screen, tex)) - goto fail; + if (i915screen->is_i945) { + if (!i945_miptree_layout(tex)) + goto fail; + } else { + if (!i915_miptree_layout(tex)) + goto fail; } - else { - if (i915screen->is_i945) { - if (!i945_miptree_layout(tex)) - goto fail; - } - else { - if (!i915_miptree_layout(tex)) - goto fail; - } - - tex->buffer = ws->buffer_create(ws, 64, - PIPE_BUFFER_USAGE_PIXEL, - tex->pitch * tex->base.cpp * - tex->total_height); - if (!tex->buffer) - goto fail; - } + tex->buffer = ws->buffer_create(ws, 64, + PIPE_BUFFER_USAGE_PIXEL, + tex->pitch * tex->base.cpp * + tex->total_height); + + if (!tex->buffer) + goto fail; return &tex->base; - fail: +fail: FREE(tex); return NULL; } diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index b3e5f29c3e..a7f3047352 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -232,36 +232,7 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, unsigned flags, unsigned tex_usage) { - const unsigned alignment = 64; - assert(!surf->buffer); -#if 0 - surf->width = width; - surf->height = round_up(height, 8); - surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = power_of_two(MAX2(width * surf->cpp, 512)); - - surf->buffer = winsys->buffer_create(winsys, 2*4096, - PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->height); - - surf->pitch = surf->pitch / surf->cpp; - surf->height = height; -#else - surf->width = width; - surf->height = height; - surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = round_up(width, alignment / surf->cpp); - - surf->buffer = winsys->buffer_create(winsys, alignment, - PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * surf->height); -#endif - if(!surf->buffer) - return -1; - - return 0; + return -1; } -- cgit v1.2.3 From bb2e13b9e82b68ec3b9fc56a4c35e7ead8fd138f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 29 May 2008 12:38:49 +0100 Subject: draw: make sure constant buffer data is aligned before passing to aos.c --- src/gallium/auxiliary/draw/draw_context.c | 5 +++-- src/gallium/auxiliary/draw/draw_context.h | 3 ++- src/gallium/auxiliary/draw/draw_private.h | 9 ++++++++- src/gallium/auxiliary/draw/draw_vs.c | 19 ++++++++++++++++++- src/gallium/auxiliary/draw/draw_vs_aos.c | 4 ++-- src/gallium/drivers/i915simple/i915_context.c | 4 +++- src/gallium/drivers/softpipe/sp_draw_arrays.c | 5 +++-- src/mesa/state_tracker/st_draw.c | 7 ++++--- 8 files changed, 43 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index bcec85c2ef..2f263cf06a 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -217,10 +217,11 @@ draw_set_mapped_vertex_buffer(struct draw_context *draw, void draw_set_mapped_constant_buffer(struct draw_context *draw, - const void *buffer) + const void *buffer, + unsigned size ) { draw->pt.user.constants = buffer; - draw_vs_set_constants( draw, (const float (*)[4])buffer ); + draw_vs_set_constants( draw, (const float (*)[4])buffer, size ); } diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index 8dd03cb79e..b8f2bfa332 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -133,7 +133,8 @@ void draw_set_mapped_vertex_buffer(struct draw_context *draw, unsigned attr, const void *buffer); void draw_set_mapped_constant_buffer(struct draw_context *draw, - const void *buffer); + const void *buffer, + unsigned size ); void draw_set_edgeflags( struct draw_context *draw, const unsigned *edgeflag ); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 40f1d978f2..88a7224b62 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -190,6 +190,12 @@ struct draw_context struct aos_machine *aos_machine; + const float (*aligned_constants)[4]; + + const float (*aligned_constant_storage)[4]; + unsigned const_storage_size; + + struct translate *fetch; struct translate_cache *fetch_cache; struct translate *emit; @@ -225,7 +231,8 @@ void draw_vs_set_viewport( struct draw_context *, const struct pipe_viewport_state * ); void draw_vs_set_constants( struct draw_context *, - const float (*constants)[4] ); + const float (*constants)[4], + unsigned size ); diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index a8b6d0c90d..ce35112fc1 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -43,8 +43,22 @@ void draw_vs_set_constants( struct draw_context *draw, - const float (*constants)[4] ) + const float (*constants)[4], + unsigned size ) { + if (((unsigned)constants) & 0xf) { + if (size > draw->vs.const_storage_size) { + if (draw->vs.aligned_constant_storage) + align_free(draw->vs.aligned_constant_storage); + draw->vs.aligned_constant_storage = align_malloc( size, 16 ); + } + memcpy( draw->vs.aligned_constant_storage, + constants, + size ); + constants = draw->vs.aligned_constant_storage; + } + + draw->vs.aligned_constants = constants; draw_vs_aos_machine_constants( draw->vs.aos_machine, constants ); } @@ -159,6 +173,9 @@ draw_vs_destroy( struct draw_context *draw ) if (draw->vs.aos_machine) draw_vs_aos_machine_destroy(draw->vs.aos_machine); + if (draw->vs.aligned_constant_storage) + align_free(draw->vs.aligned_constant_storage); + tgsi_exec_machine_free_data(&draw->vs.machine); } diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index b5e4e1e7b1..55cabb6df9 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1957,7 +1957,7 @@ static void PIPE_CDECL vaos_run_elts( struct draw_vs_varient *varient, struct aos_machine *machine = vaos->draw->vs.aos_machine; machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; - machine->constants = (const float (*)[4])vaos->draw->pt.user.constants; + machine->constants = vaos->draw->vs.aligned_constants; machine->immediates = vaos->base.vs->immediates; machine->attrib = vaos->attrib; @@ -1976,7 +1976,7 @@ static void PIPE_CDECL vaos_run_linear( struct draw_vs_varient *varient, struct aos_machine *machine = vaos->draw->vs.aos_machine; machine->internal[IMM_PSIZE][0] = vaos->draw->rasterizer->point_size; - machine->constants = (const float (*)[4])vaos->draw->pt.user.constants; + machine->constants = vaos->draw->vs.aligned_constants; machine->immediates = vaos->base.vs->immediates; machine->attrib = vaos->attrib; diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 4bef21619c..c609d16a5a 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -86,7 +86,9 @@ i915_draw_elements( struct pipe_context *pipe, draw_set_mapped_constant_buffer(draw, - i915->current.constants[PIPE_SHADER_VERTEX]); + i915->current.constants[PIPE_SHADER_VERTEX], + ( i915->current.num_user_constants[PIPE_SHADER_VERTEX] * + 4 * sizeof(float) )); /* draw! */ draw_arrays(i915->draw, prim, start, count); diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index dbecf6865f..d4d5fa744f 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -54,7 +54,8 @@ softpipe_map_constant_buffers(struct softpipe_context *sp) } draw_set_mapped_constant_buffer(sp->draw, - sp->mapped_constants[PIPE_SHADER_VERTEX]); + sp->mapped_constants[PIPE_SHADER_VERTEX], + sp->constants[i].size); } static void @@ -68,7 +69,7 @@ softpipe_unmap_constant_buffers(struct softpipe_context *sp) */ draw_flush(sp->draw); - draw_set_mapped_constant_buffer(sp->draw, NULL); + draw_set_mapped_constant_buffer(sp->draw, NULL, 0); for (i = 0; i < 2; i++) { if (sp->constants[i].size) diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 551860452a..5300848ef6 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -596,9 +596,10 @@ st_feedback_draw_vbo(GLcontext *ctx, /* map constant buffers */ mapped_constants = pipe_buffer_map(pipe, - st->state.constants[PIPE_SHADER_VERTEX].buffer, - PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_constant_buffer(st->draw, mapped_constants); + st->state.constants[PIPE_SHADER_VERTEX].buffer, + PIPE_BUFFER_USAGE_CPU_READ); + draw_set_mapped_constant_buffer(st->draw, mapped_constants, + st->state.constants[PIPE_SHADER_VERTEX].buffer->size); /* draw here */ -- cgit v1.2.3 From 45eecb93920c5a33e71b3c152749273908cb62fd Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 29 May 2008 12:42:39 +0100 Subject: draw: remove printf --- src/gallium/auxiliary/draw/draw_pt_vcache.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 720b91b8e6..64fab61271 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -260,8 +260,6 @@ static void vcache_check_run( struct draw_pt_front_end *frontend, const ushort *transformed_elts; ushort *storage = NULL; - printf("fetch_count %x\n", fetch_count); - if (fetch_count >= FETCH_MAX || fetch_count > draw_count) goto fail; -- cgit v1.2.3 From a233f65f9b79734498c120e8052aa8d6255586e4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 29 May 2008 13:08:15 +0100 Subject: draw: better calculation of fetch_count --- src/gallium/auxiliary/draw/draw_pt_vcache.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 64fab61271..957b8edec2 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -256,13 +256,19 @@ static void vcache_check_run( struct draw_pt_front_end *frontend, unsigned min_index = draw->pt.user.min_index; unsigned max_index = draw->pt.user.max_index; unsigned index_size = draw->pt.user.eltSize; - unsigned fetch_count = MAX2(max_index, max_index + 1 - min_index); + unsigned fetch_count = max_index + 1 - min_index; const ushort *transformed_elts; ushort *storage = NULL; - if (fetch_count >= FETCH_MAX || - fetch_count > draw_count) + + if (0) debug_printf("fetch_count %d draw_count %d\n", fetch_count, draw_count); + + if (max_index == 0xffffffff || + fetch_count >= FETCH_MAX || + fetch_count > draw_count) { + debug_printf("fail\n"); goto fail; + } if (min_index == 0 && -- cgit v1.2.3 From 8f887b4252208e60e7e86217ec3b72fb639a4e82 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 29 May 2008 13:26:01 +0100 Subject: draw: michal's patch for calling powf... teapot still not quite right --- src/gallium/auxiliary/draw/draw_vs_aos.c | 46 ++++++++++++++++++++++++++++++++ src/gallium/auxiliary/draw/draw_vs_aos.h | 1 + 2 files changed, 47 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 55cabb6df9..725f36b502 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1372,11 +1372,20 @@ static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } +/* A wrapper for powf(). + * Makes sure it is cdecl and operates on floats. + */ +static float PIPE_CDECL _powerf( float x, float y ) +{ + return powf( x, y ); +} + /* Really not sufficient -- need to check for conditions that could * generate inf/nan values, which will slow things down hugely. */ static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { +#if 0 x87_fld_src(cp, &op->FullSrcRegisters[1], 0); /* a1.x */ x87_fld_src(cp, &op->FullSrcRegisters[0], 0); /* a0.x a1.x */ x87_fyl2x(cp->func); /* a1*log2(a0) */ @@ -1384,6 +1393,42 @@ static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_inst x87_emit_ex2( cp ); /* 2^(a1*log2(a0)) */ x87_fstp_dest4(cp, &op->FullDstRegisters[0]); +#else + uint i; + + /* For absolute correctness, need to spill/invalidate all XMM regs + * too. + */ + for (i = 0; i < 8; i++) { + if (cp->xmm[i].dirty) + spill(cp, i); + aos_release_xmm_reg(cp, i); + } + + /* Push caller-save (ie scratch) regs. + */ + x86_cdecl_caller_push_regs( cp->func ); + + x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, -8) ); + + x87_fld_src( cp, &op->FullSrcRegisters[1], 0 ); + x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 4 ) ); + x87_fld_src( cp, &op->FullSrcRegisters[0], 0 ); + x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 0 ) ); + + x86_mov_reg_imm( cp->func, cp->tmp_EAX, (unsigned long) _powerf ); + x86_call( cp->func, cp->tmp_EAX ); + + x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, 8) ); + + x86_cdecl_caller_pop_regs( cp->func ); + + /* Note retval on x87 stack: + */ + cp->func->x87_stack++; + + x87_fstp_dest4( cp, &op->FullDstRegisters[0] ); +#endif return TRUE; } @@ -1781,6 +1826,7 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, cp.machine_EDX = x86_make_reg(file_REG32, reg_DX); cp.count_ESI = x86_make_reg(file_REG32, reg_SI); cp.temp_EBP = x86_make_reg(file_REG32, reg_BP); + cp.stack_ESP = x86_make_reg( file_REG32, reg_SP ); x86_init_func(cp.func); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index 665b425e68..fb6d43d32e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -160,6 +160,7 @@ struct aos_compilation { struct x86_reg machine_EDX; struct x86_reg count_ESI; /* decrements to zero */ struct x86_reg temp_EBP; + struct x86_reg stack_ESP; }; struct x86_reg aos_get_xmm_reg( struct aos_compilation *cp ); -- cgit v1.2.3 From 4a7198fdcfe3256bdefff5d1d766327ae7f18b35 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 29 May 2008 22:20:51 +0900 Subject: psb: Make msvc happy. Conflicts: src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index c58a900867..eb6988ff03 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -81,9 +81,9 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, * but gl vs dx9 clip spaces. */ draw_pt_post_vs_prepare( fpme->post_vs, - draw->bypass_clipping, - draw->identity_viewport, - draw->rasterizer->gl_rasterization_rules ); + (boolean)draw->bypass_clipping, + (boolean)draw->identity_viewport, + (boolean)draw->rasterizer->gl_rasterization_rules ); if (!(opt & PT_PIPELINE)) -- cgit v1.2.3 From 8808d62f608d1397ee75d0087301d0b0a0278244 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 29 May 2008 22:26:56 +0900 Subject: gallium: MSVC warning fixes. Conflicts: src/gallium/auxiliary/draw/draw_pt_varray.c src/gallium/auxiliary/draw/draw_pt_varray_tmp.h src/gallium/auxiliary/draw/draw_pt_vcache.c --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 2 +- src/gallium/auxiliary/draw/draw_pt_varray_tmp.h | 9 +++++---- src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h | 9 +++++---- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2 +- src/gallium/auxiliary/translate/translate_sse.c | 2 +- 5 files changed, 13 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 67b9a9503d..d514e28b77 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -145,7 +145,7 @@ emit_vertex( struct vbuf_stage *vbuf, vertex->vertex_id = vbuf->nr_vertices++; } - return vertex->vertex_id; + return (ushort)vertex->vertex_id; } diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h index 6979f6b544..7c722457c3 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp.h @@ -10,7 +10,8 @@ static void FUNC(struct draw_pt_front_end *frontend, boolean flatfirst = (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first); - unsigned i, j, flags; + unsigned i, j; + ushort flags; unsigned first, incr; varray->fetch_start = start; @@ -200,9 +201,9 @@ static void FUNC(struct draw_pt_front_end *frontend, /* These bitflags look a little odd because we submit the * vertices as (1,2,0) to satisfy flatshade requirements. */ - const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; - const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; - const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + const ushort edge_first = DRAW_PIPE_EDGE_FLAG_2; + const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const ushort edge_last = DRAW_PIPE_EDGE_FLAG_1; flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; for (j = 0; j + first <= count; j += i) { diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h index cf9f394aa3..5c99a47e49 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h @@ -10,7 +10,8 @@ static void FUNC( struct draw_pt_front_end *frontend, boolean flatfirst = (draw->rasterizer->flatshade && draw->rasterizer->flatshade_first); - unsigned i, flags; + unsigned i; + ushort flags; switch (vcache->input_prim) { @@ -138,9 +139,9 @@ static void FUNC( struct draw_pt_front_end *frontend, /* These bitflags look a little odd because we submit the * vertices as (1,2,0) to satisfy flatshade requirements. */ - const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; - const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; - const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + const ushort edge_first = DRAW_PIPE_EDGE_FLAG_2; + const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const ushort edge_last = DRAW_PIPE_EDGE_FLAG_1; flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 8018bd7fa4..189dc6024d 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -198,7 +198,7 @@ get_coef( static void emit_retw( struct x86_function *func, - unsigned size ) + unsigned short size ) { x86_retw( func, size ); } diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index 69dd4d3dff..be48a14737 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -309,7 +309,7 @@ static void get_src_ptr( struct translate_sse *p, static void emit_swizzle( struct translate_sse *p, struct x86_reg dest, struct x86_reg src, - unsigned shuffle ) + unsigned char shuffle ) { sse_shufps(p->func, dest, src, shuffle); } -- cgit v1.2.3 From 807e7c4ccfdaebf8e568357fb1fd8090ccae638c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 29 May 2008 14:35:30 +0100 Subject: draw: add more switches to turn FSE on/off --- src/gallium/auxiliary/draw/draw_private.h | 3 ++- src/gallium/auxiliary/draw/draw_pt.c | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 88a7224b62..4f8cceee1e 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -157,7 +157,8 @@ struct draw_context const void *constants; } user; - boolean test_fse; + boolean test_fse; /* enable FSE even though its not correct (eg for softpipe) */ + boolean no_fse; /* disable FSE even when it is correct */ } pt; struct { diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index d48c6c220d..723077159b 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -75,7 +75,7 @@ draw_pt_arrays(struct draw_context *draw, if (opt == 0) middle = draw->pt.middle.fetch_emit; - else if (opt == PT_SHADE) + else if (opt == PT_SHADE && !draw->pt.no_fse) middle = draw->pt.middle.fetch_shade_emit; else middle = draw->pt.middle.general; @@ -105,6 +105,7 @@ draw_pt_arrays(struct draw_context *draw, boolean draw_pt_init( struct draw_context *draw ) { draw->pt.test_fse = GETENV("DRAW_FSE") != NULL; + draw->pt.no_fse = GETENV("DRAW_NO_FSE") != NULL; draw->pt.front.vcache = draw_pt_vcache( draw ); if (!draw->pt.front.vcache) -- cgit v1.2.3 From 056bea86b013858832a6ab390fe6efe310a7bc8f Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 29 May 2008 15:49:23 +0200 Subject: i915: Deprecated pipe_winsys buffer functions --- src/gallium/drivers/i915simple/i915_texture.c | 2 +- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 45 ++++-------------------- 2 files changed, 8 insertions(+), 39 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 122f88fe3a..935bc742b1 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -619,7 +619,7 @@ i915_get_tex_surface(struct pipe_screen *screen, assert(zslice == 0); } - ps = ws->surface_alloc(ws); + ps = CALLOC_STRUCT(pipe_surface);//ws->surface_alloc(ws); if (ps) { assert(ps->refcount); assert(ps->winsys); diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index a7f3047352..af439a796f 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -190,40 +190,16 @@ intel_flush_frontbuffer( struct pipe_winsys *winsys, intelDisplaySurface(dPriv, surf, NULL); } - +/* + * Deprecated surface functions + */ static struct pipe_surface * intel_i915_surface_alloc(struct pipe_winsys *winsys) { - struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); - if (surf) { - surf->refcount = 1; - surf->winsys = winsys; - } - return surf; + assert("intel_i915_surface_alloc is deprecated" & 0); + return NULL; } - -/** - * Round n up to next multiple. - */ -static INLINE unsigned -round_up(unsigned n, unsigned multiple) -{ - return (n + multiple - 1) & ~(multiple - 1); -} - -static unsigned -power_of_two(unsigned x) -{ - int value = 1; - while (value <= x) - value = value << 1; - return value; -} - -/** - * Copied from xm_winsys.c - */ static int intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, struct pipe_surface *surf, @@ -232,21 +208,14 @@ intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, unsigned flags, unsigned tex_usage) { + assert("intel_i915_surface_alloc_storage is deprecated" & 0); return -1; } - static void intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) { - struct pipe_surface *surf = *s; - surf->refcount--; - if (surf->refcount == 0) { - if (surf->buffer) - pipe_buffer_reference(winsys, &surf->buffer, NULL); - free(surf); - } - *s = NULL; + assert("intel_i915_surface_release is deprecated" & 0); } -- cgit v1.2.3 From 643cc9387d03b5002a4a1e1a95f0dbcfff83b2d2 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 29 May 2008 16:03:21 +0200 Subject: i915: Cleaned up intel_winsys_pipe.c --- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 76 ++++++++++++------------ 1 file changed, 39 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c index af439a796f..51a79ca535 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c @@ -31,8 +31,6 @@ #include #include -//#include "dri_bufpool.h" -//#include "dri_bufmgr.h" #include "intel_context.h" #include "intel_winsys.h" @@ -45,8 +43,6 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" - - struct intel_pipe_winsys { struct pipe_winsys winsys; struct _DriBufferPool *regionPool; @@ -56,7 +52,6 @@ struct intel_pipe_winsys { }; - /* Turn a pipe winsys into an intel/pipe winsys: */ static inline struct intel_pipe_winsys * @@ -66,8 +61,12 @@ intel_pipe_winsys( struct pipe_winsys *winsys ) } -/* Most callbacks map direcly onto dri_bufmgr operations: +/* + * Buffer functions. + * + * Most callbacks map direcly onto dri_bufmgr operations: */ + static void *intel_buffer_map(struct pipe_winsys *winsys, struct pipe_buffer *buf, unsigned flags ) @@ -89,7 +88,6 @@ static void intel_buffer_unmap(struct pipe_winsys *winsys, driBOUnmap( dri_bo(buf) ); } - static void intel_buffer_destroy(struct pipe_winsys *winsys, struct pipe_buffer *buf) @@ -98,11 +96,6 @@ intel_buffer_destroy(struct pipe_winsys *winsys, FREE(buf); } - -/* Pipe has no concept of pools. We choose the tex/region pool - * for all buffers. - * Grabs the hardware lock! - */ static struct pipe_buffer * intel_buffer_create(struct pipe_winsys *winsys, unsigned alignment, @@ -175,24 +168,12 @@ intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) } -/* The state tracker (should!) keep track of whether the fake - * frontbuffer has been touched by any rendering since the last time - * we copied its contents to the real frontbuffer. Our task is easy: - */ -static void -intel_flush_frontbuffer( struct pipe_winsys *winsys, - struct pipe_surface *surf, - void *context_private) -{ - struct intel_context *intel = (struct intel_context *) context_private; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelDisplaySurface(dPriv, surf, NULL); -} - /* - * Deprecated surface functions + * Surface functions. + * + * Deprecated! */ + static struct pipe_surface * intel_i915_surface_alloc(struct pipe_winsys *winsys) { @@ -218,13 +199,9 @@ intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) assert("intel_i915_surface_release is deprecated" & 0); } - - -static const char * -intel_get_name( struct pipe_winsys *winsys ) -{ - return "Intel/DRI/ttm"; -} +/* + * Fence functions + */ static void intel_fence_reference( struct pipe_winsys *sws, @@ -254,6 +231,33 @@ intel_fence_finish( struct pipe_winsys *sws, return driFenceFinish((struct _DriFenceObject *)fence, flag, 0); } + +/* + * Mixed functions + */ + +static const char * +intel_get_name( struct pipe_winsys *winsys ) +{ + return "Intel/DRI/ttm"; +} + +/* + * The state tracker (should!) keep track of whether the fake + * frontbuffer has been touched by any rendering since the last time + * we copied its contents to the real frontbuffer. Our task is easy: + */ +static void +intel_flush_frontbuffer( struct pipe_winsys *winsys, + struct pipe_surface *surf, + void *context_private) +{ + struct intel_context *intel = (struct intel_context *) context_private; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelDisplaySurface(dPriv, surf, NULL); +} + struct pipe_winsys * intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) { @@ -300,7 +304,6 @@ intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) return &iws->winsys; } - void intel_destroy_pipe_winsys( struct pipe_winsys *winsys ) { @@ -313,4 +316,3 @@ intel_destroy_pipe_winsys( struct pipe_winsys *winsys ) } free(iws); } - -- cgit v1.2.3 From 529b3f4cc0ce3a3219daf5e1e8d4248cd1afe286 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 29 May 2008 19:07:40 +0200 Subject: i915: Fix GPU lockup on resize --- src/gallium/drivers/i915simple/i915_texture.c | 66 ++++++++++++++------------- 1 file changed, 35 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 935bc742b1..b216e3bfb9 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -104,48 +104,49 @@ i915_miptree_set_image_offset(struct i915_texture *tex, */ } -#if 0 -/* Hack it up to use the old winsys->surface_alloc_storage() - * method for now: +static unsigned +power_of_two(unsigned x) +{ + int value = 1; + while (value <= x) + value = value << 1; + return value; +} + +static unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + +/** + * Special case to deal with display targets. */ static boolean i915_displaytarget_layout(struct pipe_screen *screen, struct i915_texture *tex) { - struct pipe_winsys *ws = screen->winsys; - struct pipe_surface surf; - unsigned flags = (PIPE_BUFFER_USAGE_CPU_READ | - PIPE_BUFFER_USAGE_CPU_WRITE | - PIPE_BUFFER_USAGE_GPU_READ | - PIPE_BUFFER_USAGE_GPU_WRITE); - - - memset(&surf, 0, sizeof(surf)); - - ws->surface_alloc_storage( ws, - &surf, - tex->base.width[0], - tex->base.height[0], - tex->base.format, - flags, - tex->base.tex_usage); - - /* Now extract the goodies: - */ + struct pipe_texture *pt = &tex->base; + + if (pt->last_level > 1 || pt->cpp != 4) + return 0; + i915_miptree_set_level_info( tex, 0, 1, 0, 0, tex->base.width[0], tex->base.height[0], 1 ); i915_miptree_set_image_offset( tex, 0, 0, 0, 0 ); - tex->buffer = surf.buffer; - tex->pitch = surf.pitch; - tex->total_height = 0; - +#if 0 + tex->pitch = MAX2(512, power_of_two(tex->base.width[0] * pt->cpp)) / pt->cpp; + tex->total_height = round_up(tex->base.height[0], 8); +#else + tex->pitch = round_up(tex->base.width[0], 64 / pt->cpp); + tex->total_height = tex->base.height[0]; +#endif - return tex->buffer != NULL; + return 1; } -#endif static void i945_miptree_layout_2d( struct i915_texture *tex ) @@ -529,14 +530,17 @@ i915_texture_create(struct pipe_screen *screen, struct pipe_winsys *ws = screen->winsys; struct i915_texture *tex = CALLOC_STRUCT(i915_texture); - if (!tex) + if (!tex) return NULL; tex->base = *templat; tex->base.refcount = 1; tex->base.screen = screen; - if (i915screen->is_i945) { + if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { + if (!i915_displaytarget_layout(screen, tex)) + goto fail; + } else if (i915screen->is_i945) { if (!i945_miptree_layout(tex)) goto fail; } else { -- cgit v1.2.3 From 22be9ea4a63ff051604123de5e82cf4050792ef8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 29 May 2008 18:18:56 +0100 Subject: draw: quieten some debug --- src/gallium/auxiliary/draw/draw_pt_vcache.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 957b8edec2..9ffc99abf9 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -266,7 +266,7 @@ static void vcache_check_run( struct draw_pt_front_end *frontend, if (max_index == 0xffffffff || fetch_count >= FETCH_MAX || fetch_count > draw_count) { - debug_printf("fail\n"); + if (0) debug_printf("fail\n"); goto fail; } -- cgit v1.2.3 From 2691c228701e0d3a048eaaa698ba2236c2650d08 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 29 May 2008 19:24:53 +0900 Subject: gallium: Replace getenv by placeholder code on WinCE. WinCE processes supposedly have environment, but it is not clear which API to use. --- src/gallium/auxiliary/util/p_debug.c | 13 +++++++------ 1 file changed, 7 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index ce7fb58956..d1dfc377f8 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -153,7 +153,9 @@ const char * debug_get_option(const char *name, const char *dfault) { const char *result; -#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) + /* EngMapFile creates the file if it does not exists, so it must either be + * disabled on release versions (or put in a less conspicuous place). */ #ifdef DEBUG ULONG_PTR iFile = 0; const void *pMap = NULL; @@ -161,9 +163,6 @@ debug_get_option(const char *name, const char *dfault) static char output[1024]; result = dfault; - /* XXX: this creates the file if it does not exists, so it must either be - * disabled on release versions, or put in a less conspicuous place. - */ pMap = EngMapFile(L"\\??\\c:\\gallium.cfg", 0, &iFile); if(pMap) { sol = (const char *)pMap; @@ -187,13 +186,15 @@ debug_get_option(const char *name, const char *dfault) #else result = dfault; #endif +#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) + /* TODO: implement */ + result = dfault; #else - result = getenv(name); if(!result) result = dfault; #endif - + debug_printf("%s: %s = %s\n", __FUNCTION__, name, result ? result : "(null)"); return result; -- cgit v1.2.3 From ae10775b731c8c58aa42f8046b6b557b4659cb7c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 29 May 2008 21:09:06 +0900 Subject: gallium: Provide the INT64_C/UINT64_C macros. "long long" types and 1234LL constants are not supported by eVC. --- src/gallium/include/pipe/p_compiler.h | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index 29b7f9c262..f6707385cd 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -71,6 +71,9 @@ typedef __int32 intptr_t; typedef unsigned __int32 uintptr_t; #endif +#define INT64_C(__val) __val##i64 +#define UINT64_C(__val) __val##ui64 + #ifndef __cplusplus #define false 0 #define true 1 @@ -80,6 +83,9 @@ typedef int _Bool; #endif /* !__cplusplus */ #else +#ifndef __STDC_LIMIT_MACROS +#define __STDC_LIMIT_MACROS 1 +#endif #include #include #endif -- cgit v1.2.3 From 31e70fa275a48b8020c14017919e08d21f5e982f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 28 May 2008 10:40:25 -0600 Subject: comments in _eglMain() --- src/egl/drivers/dri/egldri.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c index 35b6d0d3d1..1a922a7471 100644 --- a/src/egl/drivers/dri/egldri.c +++ b/src/egl/drivers/dri/egldri.c @@ -114,6 +114,10 @@ _eglMain(_EGLDisplay *dpy, const char *args) _eglLog(_EGL_WARNING, "%s DRM devices not found.", path); return EGL_FALSE; } + + /* loop over dir entries looking for cardX where "X" is in the + * dpy->DriverName ":X" string. + */ while ((dirent = readdir(dir))) { if (strncmp(&dirent->d_name[0], "card", 4) != 0) -- cgit v1.2.3 From d1a0faffc3fde42b030f3eb18d6f31c10dfdc8f6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 29 May 2008 17:01:19 -0600 Subject: egl: use softpipe_egl driver by default --- src/egl/main/eglx.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglx.c b/src/egl/main/eglx.c index 4b33080f1a..085da33e7d 100644 --- a/src/egl/main/eglx.c +++ b/src/egl/main/eglx.c @@ -12,6 +12,8 @@ #include "eglx.h" +static const char *DefaultXDriver = "softpipe_egl"; + /** * Given an X Display ptr (at dpy->Xdpy) try to determine the appropriate @@ -21,14 +23,24 @@ const char * _xeglChooseDriver(_EGLDisplay *dpy) { #ifdef _EGL_PLATFORM_X - _XPrivDisplay xdpy = (_XPrivDisplay) dpy->Xdpy; + _XPrivDisplay xdpy; assert(dpy); + + if (!dpy->Xdpy) { + dpy->Xdpy = XOpenDisplay(NULL); + if (!dpy->Xdpy) { + /* can't open X display -> can't use X-based driver */ + return NULL; + } + } + xdpy = (_XPrivDisplay) dpy->Xdpy; + assert(dpy->Xdpy); printf("%s\n", xdpy->display_name); - return "foo"; /* XXX todo */ + return DefaultXDriver; /* XXX temporary */ #else return NULL; #endif -- cgit v1.2.3 From 1c73b4ba86b424ba66a16c4006a57db505bd3ca3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 29 May 2008 17:02:31 -0600 Subject: egl: added Name field to _EGLDriver so a driver name string can propogate up through EGL_VERSION string --- src/egl/main/eglapi.c | 3 ++- src/egl/main/egldriver.h | 4 +++- 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index 3151b35156..367787b243 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -81,7 +81,8 @@ eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor) dpyPriv->Driver->APImajor = *major; dpyPriv->Driver->APIminor = *minor; - sprintf(dpyPriv->Driver->Version, "%d.%d", *major, *minor); + snprintf(dpyPriv->Driver->Version, sizeof(dpyPriv->Driver->Version), + "%d.%d (%s)", *major, *minor, dpyPriv->Driver->Name); return retVal; } diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index b0a5fadf64..e0cedb22c3 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -28,8 +28,10 @@ struct _egl_driver void *LibHandle; /**< dlopen handle */ + const char *Name; /**< name of this driver */ + int APImajor, APIminor; /**< as returned by eglInitialize() */ - char Version[10]; /**< initialized from APImajor/minor */ + char Version[1000]; /**< initialized from APImajor/minor, Name */ const char *ClientAPIs; -- cgit v1.2.3 From bee79eb9b9ef649585dc507dc756b1e79eb00d2c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 29 May 2008 17:02:55 -0600 Subject: egl: new EGL/gallium/softpipe/xlib winsys Checkpoint commit. Most required code is in place, and compiles, but totally untested. --- src/gallium/winsys/egl_xlib/Makefile | 87 ++++++ src/gallium/winsys/egl_xlib/egl_xlib.c | 461 ++++++++++++++++++++++++++++++++ src/gallium/winsys/egl_xlib/sw_winsys.c | 244 +++++++++++++++++ src/gallium/winsys/egl_xlib/sw_winsys.h | 40 +++ 4 files changed, 832 insertions(+) create mode 100644 src/gallium/winsys/egl_xlib/Makefile create mode 100644 src/gallium/winsys/egl_xlib/egl_xlib.c create mode 100644 src/gallium/winsys/egl_xlib/sw_winsys.c create mode 100644 src/gallium/winsys/egl_xlib/sw_winsys.h (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile new file mode 100644 index 0000000000..fe4b931e30 --- /dev/null +++ b/src/gallium/winsys/egl_xlib/Makefile @@ -0,0 +1,87 @@ +# src/gallium/winsys/egl_xlib/Makefile + +# Build softpipe/xlib/EGL driver library/object: "softpipe_egl.so" + + +TOP = ../../../.. +include $(TOP)/configs/current + + +LIBNAME = softpipe_egl.so + + +INCLUDE_DIRS = \ + -I$(TOP)/include \ + -I$(TOP)/src/egl/main \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/auxiliary + +WINSYS_SOURCES = \ + egl_xlib.c \ + sw_winsys.c + +WINSYS_OBJECTS = $(WINSYS_SOURCES:.c=.o) + + +LIBS = \ + $(GALLIUM_DRIVERS) \ + $(TOP)/src/mesa/libglapi.a \ + $(TOP)/src/mesa/libmesa.a \ + $(GALLIUM_AUXILIARIES) + + +LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 + + +.SUFFIXES : .cpp + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(LOCAL_CFLAGS) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDE_DIRS) $(CXXFLAGS) $(LOCAL_CFLAGS) $< -o $@ + + + +default: $(TOP)/$(LIB_DIR)/$(LIBNAME) + + +# Make the softpipe_egl.so library +$(TOP)/$(LIB_DIR)/$(LIBNAME): $(WINSYS_OBJECTS) $(LIBS) + $(TOP)/bin/mklib -o $(LIBNAME) \ + -linker "$(CC)" \ + -noprefix \ + -install $(TOP)/$(LIB_DIR) \ + $(MKLIB_OPTIONS) $(WINSYS_OBJECTS) \ + --start-group $(LIBS) --end-group + +# $(GL_LIB_DEPS) + + +depend: $(ALL_SOURCES) + @ echo "running $(MKDEP)" + @ rm -f depend # workaround oops on gutsy?!? + @ touch depend + @ $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \ + > /dev/null 2>/dev/null + + +install: default + $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) + @if [ -e $(TOP)/$(LIB_DIR) ]; then \ + $(INSTALL) $(TOP)/$(LIB_DIR)/$(LIBNAME) $(INSTALL_DIR)/$(LIB_DIR); \ + fi + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` $(TOP)/include/GL/*.h + +clean: + -rm -f *.o *~ *.bak + + +include depend diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c new file mode 100644 index 0000000000..4741da87cd --- /dev/null +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -0,0 +1,461 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * EGL / softpipe / xlib winsys module + * + * Authors: Brian Paul + */ + + +#include + +#include "pipe/p_compiler.h" +#include "pipe/p_format.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#include "softpipe/sp_winsys.h" + +#include "eglconfig.h" +#include "eglconfigutil.h" +#include "eglcontext.h" +#include "egldisplay.h" +#include "egldriver.h" +#include "eglglobals.h" +#include "egllog.h" +#include "eglsurface.h" + +#include "state_tracker/st_public.h" + +#include "sw_winsys.h" + + +/** subclass of _EGLDriver */ +struct xlib_egl_driver +{ + _EGLDriver Base; /**< base class */ + + struct pipe_winsys *winsys; + struct pipe_screen *screen; +}; + + +/** subclass of _EGLContext */ +struct xlib_egl_context +{ + _EGLContext Base; /**< base class */ + + struct pipe_context *pipe; /**< Gallium driver context */ + struct st_context *Context; /**< Mesa/gallium state tracker context */ +}; + + +/** subclass of _EGLSurface */ +struct xlib_egl_surface +{ + _EGLSurface Base; /**< base class */ + + Display *Dpy; /**< The X Display of the window */ + Window Win; /**< The user-created window ID */ + GC Gc; + XVisualInfo VisInfo; + + struct pipe_winsys *winsys; + + struct st_framebuffer *Framebuffer; +}; + + +/** cast wrapper */ +static INLINE struct xlib_egl_driver * +xlib_egl_driver(_EGLDriver *drv) +{ + return (struct xlib_egl_driver *) drv; +} + + +static struct xlib_egl_surface * +lookup_surface(EGLSurface surf) +{ + _EGLSurface *surface = _eglLookupSurface(surf); + return (struct xlib_egl_surface *) surface; +} + + +static struct xlib_egl_context * +lookup_context(EGLContext surf) +{ + _EGLContext *context = _eglLookupContext(surf); + return (struct xlib_egl_context *) context; +} + + +/** + * XXX temporary + * Need to query X server's GLX visuals. + */ +static void +init_configs(_EGLDriver *drv, EGLDisplay dpy) +{ + _EGLDisplay *disp = _eglLookupDisplay(dpy); + int i; + + for (i = 0; i < 2; i++) { + _EGLConfig config; + int id = i + 1; + _eglInitConfig(&config, id); + SET_CONFIG_ATTRIB(&config, EGL_RED_SIZE, 8); + SET_CONFIG_ATTRIB(&config, EGL_GREEN_SIZE, 8); + SET_CONFIG_ATTRIB(&config, EGL_BLUE_SIZE, 8); + SET_CONFIG_ATTRIB(&config, EGL_ALPHA_SIZE, 8); + if (i > 0) { + SET_CONFIG_ATTRIB(&config, EGL_DEPTH_SIZE, 24); + SET_CONFIG_ATTRIB(&config, EGL_STENCIL_SIZE, 8); + } + _eglAddConfig(disp, &config); + } +} + + + +/** + * Called via eglInitialize(), drv->API.Initialize(). + */ +static EGLBoolean +xlib_eglInitialize(_EGLDriver *drv, EGLDisplay dpy, + EGLint *minor, EGLint *major) +{ + /* visual configs */ + + init_configs(drv, dpy); + + + drv->Initialized = EGL_TRUE; + + /* we're supporting EGL 1.4 */ + *minor = 1; + *major = 4; + + return EGL_TRUE; +} + + +/** + * Called via eglTerminate(), drv->API.Terminate(). + */ +static EGLBoolean +xlib_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) +{ + + return EGL_TRUE; +} + + +/** Get size of given window */ +static Status +get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height) +{ + Window root; + Status stat; + int xpos, ypos; + unsigned int w, h, bw, depth; + stat = XGetGeometry(dpy, d, &root, &xpos, &ypos, &w, &h, &bw, &depth); + *width = w; + *height = h; + return stat; +} + + +static void +display_surface(struct pipe_winsys *pws, + struct pipe_surface *psurf, + struct xlib_egl_surface *xsurf) +{ + XImage *ximage; + void *data; + + ximage = XCreateImage(xsurf->Dpy, + xsurf->VisInfo.visual, + xsurf->VisInfo.depth, + ZPixmap, 0, /* format, offset */ + NULL, /* data */ + 0, 0, /* size */ + 32, /* bitmap_pad */ + 0); /* bytes_per_line */ + + + assert(ximage->format); + assert(ximage->bitmap_unit); + + data = pws->buffer_map(pws, psurf->buffer, 0); + + /* update XImage's fields */ + ximage->data = data; + ximage->width = psurf->width; + ximage->height = psurf->height; + ximage->bytes_per_line = psurf->pitch * psurf->cpp; + + XPutImage(xsurf->Dpy, xsurf->Win, xsurf->Gc, + ximage, 0, 0, 0, 0, psurf->width, psurf->height); + + ximage->data = NULL; + XDestroyImage(ximage); + + pws->buffer_unmap(pws, psurf->buffer); +} + + + +/** Display gallium surface in X window */ +static void +flush_frontbuffer(struct pipe_winsys *pws, + struct pipe_surface *psurf, + void *context_private) +{ + struct xlib_egl_surface *xsurf = (struct xlib_egl_surface *) context_private; + display_surface(pws, psurf, xsurf); +} + + + +/** + * Called via eglCreateContext(), drv->API.CreateContext(). + */ +static EGLContext +xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, + EGLContext share_list, const EGLint *attrib_list) +{ + struct xlib_egl_driver *xdrv = xlib_egl_driver(drv); + _EGLConfig *conf = _eglLookupConfig(drv, dpy, config); + struct xlib_egl_context *ctx; + struct st_context *share_ctx = NULL; /* XXX fix */ + __GLcontextModes visual; + + ctx = CALLOC_STRUCT(xlib_egl_context); + if (!ctx) + return EGL_NO_CONTEXT; + + /* let EGL lib init the common stuff */ + if (!_eglInitContext(drv, dpy, &ctx->Base, config, attrib_list)) { + free(ctx); + return EGL_NO_CONTEXT; + } + + /* create a softpipe context */ + ctx->pipe = softpipe_create(xdrv->screen, xdrv->winsys, NULL); + + /* Now do xlib / state tracker inits here */ + _eglConfigToContextModesRec(conf, &visual); + ctx->Context = st_create_context(ctx->pipe, &visual, share_ctx); + + + return _eglGetContextHandle(&ctx->Base); +} + + +static EGLBoolean +xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx) +{ + struct xlib_egl_context *context = lookup_context(ctx); + if (context) { + if (context->Base.IsBound) { + context->Base.DeletePending = EGL_TRUE; + } + else { + st_destroy_context(context->Context); + free(context); + } + return EGL_TRUE; + } + else { + _eglError(EGL_BAD_CONTEXT, "eglDestroyContext"); + return EGL_TRUE; + } +} + + +/** + * Called via eglMakeCurrent(), drv->API.MakeCurrent(). + */ +static EGLBoolean +xlib_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d, + EGLSurface r, EGLContext context) +{ + if (!_eglMakeCurrent(drv, dpy, d, r, context)) + return EGL_FALSE; + + /* XXX anything todo? */ + + return EGL_TRUE; +} + + +static enum pipe_format +choose_color_format(const __GLcontextModes *visual) +{ + if (visual->redBits == 8 && + visual->greenBits == 8 && + visual->blueBits == 8 && + visual->alphaBits == 8) { + /* XXX this really also depends on the ordering of R,G,B,A */ + return PIPE_FORMAT_A8R8G8B8_UNORM; + } + else { + assert(0); + return PIPE_FORMAT_NONE; + } +} + + +static enum pipe_format +choose_depth_format(const __GLcontextModes *visual) +{ + if (visual->depthBits > 0) + return PIPE_FORMAT_Z24S8_UNORM; + else + return PIPE_FORMAT_NONE; +} + + +static enum pipe_format +choose_stencil_format(const __GLcontextModes *visual) +{ + if (visual->stencilBits > 0) + return PIPE_FORMAT_Z24S8_UNORM; + else + return PIPE_FORMAT_NONE; +} + + +/** + * Called via eglCreateWindowSurface(), drv->API.CreateWindowSurface(). + */ +static EGLSurface +xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, + NativeWindowType window, const EGLint *attrib_list) +{ + struct xlib_egl_driver *xdrv = xlib_egl_driver(drv); + _EGLDisplay *disp = _eglLookupDisplay(dpy); + _EGLConfig *conf = _eglLookupConfig(drv, dpy, config); + + struct xlib_egl_surface *surf; + __GLcontextModes visual; + uint width, height; + + surf = CALLOC_STRUCT(xlib_egl_surface); + if (!surf) + return EGL_NO_SURFACE; + + /* Let EGL lib init the common stuff */ + if (!_eglInitSurface(drv, dpy, &surf->Base, EGL_WINDOW_BIT, + config, attrib_list)) { + free(surf); + return EGL_NO_SURFACE; + } + + /* + * Now init the Xlib and gallium stuff + */ + surf->Win = (Window) window; /* The X window ID */ + surf->Dpy = disp->Xdpy; /* The X display */ + surf->Gc = XCreateGC(surf->Dpy, surf->Win, 0, NULL); + + surf->winsys = xdrv->winsys; + + _eglConfigToContextModesRec(conf, &visual); + get_drawable_size(surf->Dpy, surf->Win, &width, &height); + + /* Create GL statetracker framebuffer */ + surf->Framebuffer = st_create_framebuffer(&visual, + choose_color_format(&visual), + choose_depth_format(&visual), + choose_stencil_format(&visual), + width, height, + (void *) surf); + return surf; +} + + +static EGLBoolean +xlib_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) +{ + /* error checking step: */ + if (!_eglSwapBuffers(drv, dpy, draw)) + return EGL_FALSE; + + { + struct xlib_egl_surface *xsurf = lookup_surface(draw); + struct pipe_winsys *pws = xsurf->winsys; + struct pipe_surface *psurf = + st_get_framebuffer_surface(xsurf->Framebuffer, ST_SURFACE_BACK_LEFT); + + display_surface(pws, psurf, xsurf); + } + + return EGL_TRUE; +} + + +/** + * This is the main entrypoint into the driver. + * Called by libEGL to instantiate an _EGLDriver object. + */ +_EGLDriver * +_eglMain(_EGLDisplay *dpy, const char *args) +{ + struct xlib_egl_driver *xdrv; + + _eglLog(_EGL_INFO, "Entering EGL/Xlib _eglMain(%s)", args); + + xdrv = CALLOC_STRUCT(xlib_egl_driver); + if (!xdrv) + return NULL; + + _eglInitDriverFallbacks(&xdrv->Base); + xdrv->Base.API.Initialize = xlib_eglInitialize; + xdrv->Base.API.Terminate = xlib_eglTerminate; + xdrv->Base.API.CreateContext = xlib_eglCreateContext; + xdrv->Base.API.DestroyContext = xlib_eglDestroyContext; + xdrv->Base.API.CreateWindowSurface = xlib_eglCreateWindowSurface; + xdrv->Base.API.MakeCurrent = xlib_eglMakeCurrent; + xdrv->Base.API.SwapBuffers = xlib_eglSwapBuffers; + + + xdrv->Base.ClientAPIs = "OpenGL"; /* "OpenGL_ES" */ + xdrv->Base.Name = "Xlib/softpipe"; + + /* create one winsys and use it for all contexts/surfaces */ + xdrv->winsys = create_sw_winsys(); + xdrv->winsys->flush_frontbuffer = flush_frontbuffer; + + xdrv->screen = softpipe_create_screen(xdrv->winsys); + + + return &xdrv->Base; +} + diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c new file mode 100644 index 0000000000..0e71bdee9d --- /dev/null +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -0,0 +1,244 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Totally software-based winsys layer. + * Note that the one winsys function that we can't implement here + * is flush_frontbuffer(). + * Whoever uses this code will have to provide that. + * + * Authors: Brian Paul + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" + +#include "sw_winsys.h" + + + +/** Subclass of pipe_winsys */ +struct sw_pipe_winsys +{ + struct pipe_winsys Base; + /* no extra fields for now */ +}; + + +/** subclass of pipe_buffer */ +struct sw_pipe_buffer +{ + struct pipe_buffer Base; + boolean UserBuffer; /** Is this a user-space buffer? */ + void *Data; + void *Mapped; +}; + + +/** cast wrapper */ +static INLINE struct sw_pipe_buffer * +sw_pipe_buffer(struct pipe_buffer *b) +{ + return (struct sw_pipe_buffer *) b; +} + + + +static const char * +get_name(struct pipe_winsys *pws) +{ + return "software"; +} + + +/** Create new pipe_buffer and allocate storage of given size */ +static struct pipe_buffer * +buffer_create(struct pipe_winsys *pws, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct sw_pipe_buffer *buffer = CALLOC_STRUCT(sw_pipe_buffer); + if (!buffer) + return NULL; + + buffer->Base.refcount = 1; + buffer->Base.alignment = alignment; + buffer->Base.usage = usage; + buffer->Base.size = size; + + /* align to 16-byte multiple for Cell */ + buffer->Data = align_malloc(size, MAX2(alignment, 16)); + + return &buffer->Base; +} + + +/** + * Create buffer which wraps user-space data. + */ +static struct pipe_buffer * +user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) +{ + struct sw_pipe_buffer *buffer = CALLOC_STRUCT(sw_pipe_buffer); + if (!buffer) + return NULL; + + buffer->Base.refcount = 1; + buffer->Base.size = bytes; + buffer->UserBuffer = TRUE; + buffer->Data = ptr; + + return &buffer->Base; +} + + +static void * +buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buf, unsigned flags) +{ + struct sw_pipe_buffer *buffer = sw_pipe_buffer(buf); + buffer->Mapped = buffer->Data; + return buffer->Mapped; +} + + +static void +buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) +{ + struct sw_pipe_buffer *buffer = sw_pipe_buffer(buf); + buffer->Mapped = NULL; +} + + +static void +buffer_destroy(struct pipe_winsys *pws, struct pipe_buffer *buf) +{ + struct sw_pipe_buffer *buffer = sw_pipe_buffer(buf); + + if (buffer->Data && !buffer->UserBuffer) { + align_free(buffer->Data); + buffer->Data = NULL; + } + + free(buffer); +} + + +/** + * Called via winsys->surface_alloc() to create new surfaces. + */ +static struct pipe_surface * +surface_alloc(struct pipe_winsys *ws) +{ + struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); + if (!surf) + return NULL; + + surf->refcount = 1; + surf->winsys = ws; + + return surf; +} + + +static void +surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + assert(!surf->texture); + surf->refcount--; + if (surf->refcount == 0) { + if (surf->buffer) + pipe_buffer_reference(winsys, &surf->buffer, NULL); + free(surf); + } + *s = NULL; +} + + +static void +fence_reference(struct pipe_winsys *sws, struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) +{ + /* no-op */ +} + + +static int +fence_signalled(struct pipe_winsys *sws, struct pipe_fence_handle *fence, + unsigned flag) +{ + /* no-op */ + return 0; +} + + +static int +fence_finish(struct pipe_winsys *sws, struct pipe_fence_handle *fence, + unsigned flag) +{ + /* no-op */ + return 0; +} + + +/** + * Create/return a new pipe_winsys object. + */ +struct pipe_winsys * +create_sw_winsys(void) +{ + struct sw_pipe_winsys *ws = CALLOC_STRUCT(sw_pipe_winsys); + if (!ws) + return NULL; + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + */ + ws->Base.buffer_create = buffer_create; + ws->Base.user_buffer_create = user_buffer_create; + ws->Base.buffer_map = buffer_map; + ws->Base.buffer_unmap = buffer_unmap; + ws->Base.buffer_destroy = buffer_destroy; + + ws->Base.surface_alloc = surface_alloc; + ws->Base.surface_alloc_storage = NULL; /* unused */ + ws->Base.surface_release = surface_release; + + ws->Base.fence_reference = fence_reference; + ws->Base.fence_signalled = fence_signalled; + ws->Base.fence_finish = fence_finish; + + ws->Base.flush_frontbuffer = NULL; /* not implemented here! */ + + ws->Base.get_name = get_name; + + return &ws->Base; +} diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.h b/src/gallium/winsys/egl_xlib/sw_winsys.h new file mode 100644 index 0000000000..f96c5a14b0 --- /dev/null +++ b/src/gallium/winsys/egl_xlib/sw_winsys.h @@ -0,0 +1,40 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef SW_WINSYS_H +#define SW_WINSYS_H + + +struct pipe_winsys; + + +extern struct pipe_winsys * +create_sw_winsys(void); + + +#endif /* SW_WINSYS_H */ -- cgit v1.2.3 From 012c0dd6321e858af22389f98201c8dcb8f8780c Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 30 May 2008 13:36:52 +0200 Subject: sis: Removed useless comment --- src/mesa/drivers/dri/sis/server/sis_common.h | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/sis/server/sis_common.h b/src/mesa/drivers/dri/sis/server/sis_common.h index 04f9ccdbf8..bd9bab846f 100644 --- a/src/mesa/drivers/dri/sis/server/sis_common.h +++ b/src/mesa/drivers/dri/sis/server/sis_common.h @@ -1,4 +1,3 @@ -/* * */ /* * Common header definitions for SiS 2D/3D/DRM suite * -- cgit v1.2.3 From 9760ab234951d18c9606e962c6e3ac1f56fc6385 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 30 May 2008 13:30:51 +0200 Subject: i915: Fix extentions not being set --- src/gallium/winsys/dri/intel/intel_context.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index cc2fc11c5e..6a0c381849 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -35,6 +35,7 @@ #include "intel_batchbuffer.h" #include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" #include "pipe/p_defines.h" #include "pipe/p_context.h" @@ -218,6 +219,8 @@ intelCreateContext(const __GLcontextModes * visual, intel->st = st_create_context(pipe, visual, st_share); + driInitExtensions( intel->st->ctx, card_extensions, GL_TRUE ); + return GL_TRUE; } -- cgit v1.2.3 From 2ade5268dca67a73d3f5f8cc41ea86d1e48de9f0 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 30 May 2008 13:31:42 +0200 Subject: i915: Remade texture allocation code again --- src/gallium/drivers/i915simple/i915_texture.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index b216e3bfb9..4c0c7ce3f3 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -128,7 +128,7 @@ i915_displaytarget_layout(struct pipe_screen *screen, { struct pipe_texture *pt = &tex->base; - if (pt->last_level > 1 || pt->cpp != 4) + if (pt->last_level > 0 || pt->cpp != 4) return 0; i915_miptree_set_level_info( tex, 0, 1, 0, 0, @@ -177,7 +177,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) /* Pitch must be a whole number of dwords, even though we * express it in texels. */ - tex->pitch = align_int(tex->pitch * pt->cpp, 4) / pt->cpp; + tex->pitch = align_int(tex->pitch * pt->cpp, 64) / pt->cpp; tex->total_height = 0; for (level = 0; level <= pt->last_level; level++) { @@ -537,10 +537,7 @@ i915_texture_create(struct pipe_screen *screen, tex->base.refcount = 1; tex->base.screen = screen; - if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) { - if (!i915_displaytarget_layout(screen, tex)) - goto fail; - } else if (i915screen->is_i945) { + if (i915screen->is_i945) { if (!i945_miptree_layout(tex)) goto fail; } else { -- cgit v1.2.3 From 7b0a551c4cebc44cc06face712eb9659ccdebcc9 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 29 May 2008 16:56:42 +0200 Subject: i915: Made batch buffer size much larger --- src/gallium/winsys/dri/intel/intel_batchbuffer.c | 2 +- src/gallium/winsys/dri/intel/intel_screen.c | 6 +++--- src/gallium/winsys/dri/intel/intel_screen.h | 1 + 3 files changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c index aa2eed53b8..09d4eef484 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.c +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.c @@ -65,7 +65,7 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) driBOUnrefUserList(batch->list); driBOResetList(batch->list); - batch->size = 4 * 4096; // ZZZ JB batch->intel->intelScreen->maxBatchSize; + batch->size = batch->intel->intelScreen->max_batch_size; driBOData(batch->buffer, batch->size, NULL, NULL, 0); /* diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 429a2cf8f8..0be88974f4 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -200,8 +200,8 @@ intelCreatePools(__DRIscreenPrivate * sPriv) DRM_BO_FLAG_MEM_TT, DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, - 4 * 4096, //intelScreen->maxBatchSize, - 1, 40, 16*16384, 0, + intelScreen->max_batch_size, + 1, 40, intelScreen->max_batch_size * 16, 0, intelScreen->fMan); #endif intelScreen->havePools = GL_TRUE; @@ -262,7 +262,7 @@ intelInitDriver(__DRIscreenPrivate * sPriv) (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); } - + intelScreen->max_batch_size = 16 * 4096; #if 1 // ZZZ JB intelScreen->mgr = driFenceMgrTTMInit(sPriv->fd); diff --git a/src/gallium/winsys/dri/intel/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h index c17f0202e4..1db0502b13 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.h +++ b/src/gallium/winsys/dri/intel/intel_screen.h @@ -80,6 +80,7 @@ struct intel_screen struct _DriFenceMgr *mgr; struct _DriFreeSlabManager *fMan; unsigned batch_id; + unsigned max_batch_size; struct pipe_winsys *winsys; }; -- cgit v1.2.3 From b12a31f59d80a1d5703181b00b42df3fa712497b Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 29 May 2008 16:57:34 +0200 Subject: i915: Now using draw_elements_range --- src/gallium/drivers/i915simple/i915_context.c | 29 +++++++++++++++++++++------ 1 file changed, 23 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index c609d16a5a..81cab75738 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -50,10 +50,12 @@ static void i915_destroy( struct pipe_context *pipe ) static boolean -i915_draw_elements( struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned prim, unsigned start, unsigned count) +i915_draw_range_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned min_index, + unsigned max_index, + unsigned prim, unsigned start, unsigned count) { struct i915_context *i915 = i915_context( pipe ); struct draw_context *draw = i915->draw; @@ -77,7 +79,10 @@ i915_draw_elements( struct pipe_context *pipe, void *mapped_indexes = pipe->winsys->buffer_map(pipe->winsys, indexBuffer, PIPE_BUFFER_USAGE_CPU_READ); - draw_set_mapped_element_buffer(draw, indexSize, mapped_indexes); + draw_set_mapped_element_buffer_range(draw, indexSize, + min_index, + max_index, + mapped_indexes); } else { /* no index/element buffer */ @@ -102,12 +107,23 @@ i915_draw_elements( struct pipe_context *pipe, } if (indexBuffer) { pipe->winsys->buffer_unmap(pipe->winsys, indexBuffer); - draw_set_mapped_element_buffer(draw, 0, NULL); + draw_set_mapped_element_buffer_range(draw, 0, start, start + count - 1, NULL); } return TRUE; } +static boolean +i915_draw_elements( struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned prim, unsigned start, unsigned count) +{ + return i915_draw_range_elements( pipe, indexBuffer, + indexSize, + 0, 0xffffffff, + prim, start, count ); +} static boolean i915_draw_arrays( struct pipe_context *pipe, unsigned prim, unsigned start, unsigned count) @@ -138,6 +154,7 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, i915->pipe.draw_arrays = i915_draw_arrays; i915->pipe.draw_elements = i915_draw_elements; + i915->pipe.draw_range_elements = i915_draw_range_elements; /* * Create drawing context and plug our rendering stage into it. -- cgit v1.2.3 From df7d1ee2a5270be2d329fd0614a608357917bb74 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 08:53:42 -0600 Subject: Fix segfault in _save_OBE_DrawElements() when using VBO and display list (bug 16156) This was previously fixed in master by commit 982dcb74fd19b88208d127b8019e2a2af979cac2 by Haihao Xiang. --- src/mesa/vbo/vbo_save_api.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/vbo/vbo_save_api.c b/src/mesa/vbo/vbo_save_api.c index aded738143..f62be5c14c 100644 --- a/src/mesa/vbo/vbo_save_api.c +++ b/src/mesa/vbo/vbo_save_api.c @@ -864,6 +864,9 @@ static void GLAPIENTRY _save_OBE_DrawElements(GLenum mode, GLsizei count, GLenum _ae_map_vbos( ctx ); + if (ctx->Array.ElementArrayBufferObj->Name) + indices = ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Pointer, indices); + vbo_save_NotifyBegin( ctx, mode | VBO_SAVE_PRIM_WEAK ); switch (type) { -- cgit v1.2.3 From 63faab0150c3394bd9532e621947d2a31b9712ea Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 30 May 2008 14:47:38 +0100 Subject: mesa: undo accidental setting of _ForceEyeCoords --- src/mesa/main/light.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c index 0af647e13c..f9715b4865 100644 --- a/src/mesa/main/light.c +++ b/src/mesa/main/light.c @@ -1357,7 +1357,7 @@ _mesa_init_lighting( GLcontext *ctx ) /* Miscellaneous */ ctx->Light._NeedEyeCoords = GL_FALSE; ctx->_NeedEyeCoords = GL_FALSE; - ctx->_ForceEyeCoords = GL_TRUE; + ctx->_ForceEyeCoords = GL_FALSE; ctx->_ModelViewInvScale = 1.0; } -- cgit v1.2.3 From 1529a2c983f70ed7ff661ae6e8995cdfbca6d9cf Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 30 May 2008 14:51:09 +0100 Subject: draw: trim incoming primitives --- src/gallium/auxiliary/draw/draw_pt.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 723077159b..f0d7b51ad7 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -35,6 +35,10 @@ #include "draw/draw_private.h" #include "draw/draw_pt.h" +static unsigned trim( unsigned count, unsigned first, unsigned incr ) +{ + return count - (count - first) % incr; +} @@ -54,6 +58,17 @@ draw_pt_arrays(struct draw_context *draw, struct draw_pt_middle_end *middle = NULL; unsigned opt = 0; + /* Sanitize primitive length: + */ + { + unsigned first, incr; + draw_pt_split_prim(prim, &first, &incr); + count = trim(count, first, incr); + if (count < first) + return TRUE; + } + + if (!draw->render) { opt |= PT_PIPELINE; } -- cgit v1.2.3 From d0de5a2dafb5143f01a3c6dc367c85aa0800466b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 11:40:23 -0600 Subject: link libEGL with Xlib --- src/egl/main/Makefile | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/Makefile b/src/egl/main/Makefile index 5b341a19be..96742bb4bc 100644 --- a/src/egl/main/Makefile +++ b/src/egl/main/Makefile @@ -46,6 +46,8 @@ OBJECTS = $(SOURCES:.c=.o) # Undefined for now LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 +LIBS = -lX11 + .c.o: $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(LOCAL_CFLAGS) $< -o $@ @@ -60,7 +62,7 @@ library: $(TOP)/$(LIB_DIR)/libEGL.so $(TOP)/$(LIB_DIR)/libEGL.so: $(OBJECTS) $(TOP)/bin/mklib -o EGL -major 1 -minor 0 \ - -install $(TOP)/$(LIB_DIR) -ldl $(OBJECTS) + -install $(TOP)/$(LIB_DIR) -ldl $(OBJECTS) $(LIBS) -- cgit v1.2.3 From 274dd381a30072ecb8341cfc41e63bb6e39419ac Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 11:40:48 -0600 Subject: egl: fix width/height tests --- src/egl/main/eglsurface.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/eglsurface.c b/src/egl/main/eglsurface.c index 1dbb12ecfe..796d62f2b9 100644 --- a/src/egl/main/eglsurface.c +++ b/src/egl/main/eglsurface.c @@ -175,7 +175,7 @@ _eglInitSurface(_EGLDriver *drv, EGLDisplay dpy, } } - if (width <= 0 || height <= 0) { + if (width < 0 || height < 0) { _eglError(EGL_BAD_ATTRIBUTE, func); return EGL_FALSE; } -- cgit v1.2.3 From 057ee55720faee0e1269937d70bbafe6df5c57e6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 11:41:18 -0600 Subject: egl: added surface_alloc_storage() --- src/gallium/winsys/egl_xlib/sw_winsys.c | 39 ++++++++++++++++++++++++++++++++- 1 file changed, 38 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c index 0e71bdee9d..fd0cb9fb12 100644 --- a/src/gallium/winsys/egl_xlib/sw_winsys.c +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -70,6 +70,15 @@ sw_pipe_buffer(struct pipe_buffer *b) } +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + static const char * get_name(struct pipe_winsys *pws) @@ -168,6 +177,34 @@ surface_alloc(struct pipe_winsys *ws) } +static int +surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags, + unsigned tex_usage) +{ + const unsigned alignment = 64; + + surf->width = width; + surf->height = height; + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = round_up(width, alignment / surf->cpp); + surf->usage = flags; + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->cpp * height); + if(!surf->buffer) + return -1; + + return 0; +} + + static void surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) { @@ -229,7 +266,7 @@ create_sw_winsys(void) ws->Base.buffer_destroy = buffer_destroy; ws->Base.surface_alloc = surface_alloc; - ws->Base.surface_alloc_storage = NULL; /* unused */ + ws->Base.surface_alloc_storage = surface_alloc_storage; ws->Base.surface_release = surface_release; ws->Base.fence_reference = fence_reference; -- cgit v1.2.3 From 750782a327a1d295f4ed00a590c362f32912d597 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 11:42:08 -0600 Subject: egl: assorted fixes. The code works now. --- src/gallium/winsys/egl_xlib/egl_xlib.c | 67 ++++++++++++++++++++++++++++++---- 1 file changed, 60 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 4741da87cd..555bb8b7b4 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -176,6 +176,28 @@ xlib_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) } +static void +get_drawable_visual_info(Display *dpy, Drawable d, XVisualInfo *visInfo) +{ + XWindowAttributes attr; + XVisualInfo visTemp, *vis; + int num_visuals; + + XGetWindowAttributes(dpy, d, &attr); + + visTemp.screen = DefaultScreen(dpy); + visTemp.visualid = attr.visual->visualid; + vis = XGetVisualInfo(dpy, + (VisualScreenMask | VisualIDMask), + &visTemp, &num_visuals); + if (vis) + *visInfo = *vis; + + XFree(vis); +} + + + /** Get size of given window */ static Status get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height) @@ -191,6 +213,16 @@ get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height) } +static void +check_and_update_buffer_size(struct xlib_egl_surface *surface) +{ + uint width, height; + get_drawable_size(surface->Dpy, surface->Win, &width, &height); + st_resize_framebuffer(surface->Framebuffer, width, height); +} + + + static void display_surface(struct pipe_winsys *pws, struct pipe_surface *psurf, @@ -223,6 +255,8 @@ display_surface(struct pipe_winsys *pws, XPutImage(xsurf->Dpy, xsurf->Win, xsurf->Gc, ximage, 0, 0, 0, 0, psurf->width, psurf->height); + XSync(xsurf->Dpy, 0); + ximage->data = NULL; XDestroyImage(ximage); @@ -303,13 +337,21 @@ xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx) * Called via eglMakeCurrent(), drv->API.MakeCurrent(). */ static EGLBoolean -xlib_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d, - EGLSurface r, EGLContext context) +xlib_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, + EGLSurface draw, EGLSurface read, EGLContext ctx) { - if (!_eglMakeCurrent(drv, dpy, d, r, context)) + struct xlib_egl_context *context = lookup_context(ctx); + struct xlib_egl_surface *draw_surf = lookup_surface(draw); + struct xlib_egl_surface *read_surf = lookup_surface(read); + + if (!_eglMakeCurrent(drv, dpy, draw, read, context)) return EGL_FALSE; - /* XXX anything todo? */ + st_make_current((context ? context->Context : NULL), + (draw_surf ? draw_surf->Framebuffer : NULL), + (read_surf ? read_surf->Framebuffer : NULL)); + + check_and_update_buffer_size(draw_surf); return EGL_TRUE; } @@ -336,7 +378,7 @@ static enum pipe_format choose_depth_format(const __GLcontextModes *visual) { if (visual->depthBits > 0) - return PIPE_FORMAT_Z24S8_UNORM; + return PIPE_FORMAT_S8Z24_UNORM; else return PIPE_FORMAT_NONE; } @@ -346,7 +388,7 @@ static enum pipe_format choose_stencil_format(const __GLcontextModes *visual) { if (visual->stencilBits > 0) - return PIPE_FORMAT_Z24S8_UNORM; + return PIPE_FORMAT_S8Z24_UNORM; else return PIPE_FORMAT_NONE; } @@ -378,6 +420,8 @@ xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, return EGL_NO_SURFACE; } + _eglSaveSurface(&surf->Base); + /* * Now init the Xlib and gallium stuff */ @@ -389,6 +433,10 @@ xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, _eglConfigToContextModesRec(conf, &visual); get_drawable_size(surf->Dpy, surf->Win, &width, &height); + get_drawable_visual_info(surf->Dpy, surf->Win, &surf->VisInfo); + + surf->Base.Width = width; + surf->Base.Height = height; /* Create GL statetracker framebuffer */ surf->Framebuffer = st_create_framebuffer(&visual, @@ -397,7 +445,10 @@ xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, choose_stencil_format(&visual), width, height, (void *) surf); - return surf; + + st_resize_framebuffer(surf->Framebuffer, width, height); + + return _eglGetSurfaceHandle(&surf->Base); } @@ -414,6 +465,8 @@ xlib_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) struct pipe_surface *psurf = st_get_framebuffer_surface(xsurf->Framebuffer, ST_SURFACE_BACK_LEFT); + st_notify_swapbuffers(xsurf->Framebuffer); + display_surface(pws, psurf, xsurf); } -- cgit v1.2.3 From 834aa8cfbc25622256889624ee1768c10224feb2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 11:46:43 -0600 Subject: egl: check for EGL_DRIVER env var to override normal driver selection process --- src/egl/main/egldriver.c | 11 +++++++++-- 1 file changed, 9 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 32842933ce..e155f5fb98 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -92,14 +92,21 @@ _eglChooseDriver(_EGLDisplay *dpy) const char *displayString = (const char *) dpy->NativeDisplay; const char *driverName = NULL; + /* First, if the EGL_DRIVER env var is set, use that */ + driverName = getenv("EGL_DRIVER"); + if (driverName) + return _eglstrdup(driverName); + + if (!displayString) { /* choose a default */ displayString = DefaultDriverName; } /* extract default DriverArgs = whatever follows ':' */ - if (displayString[0] == '!' || - displayString[0] == ':') { + if (displayString && + (displayString[0] == '!' || + displayString[0] == ':')) { const char *args = strchr(displayString, ':'); if (args) dpy->DriverArgs = _eglstrdup(args + 1); -- cgit v1.2.3 From 64d0eb111d5a19bc251dfc59884c945cf93c6bb0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 11:51:50 -0600 Subject: egl: fix-up window resizes --- src/gallium/winsys/egl_xlib/egl_xlib.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 555bb8b7b4..b683898dcb 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -219,6 +219,8 @@ check_and_update_buffer_size(struct xlib_egl_surface *surface) uint width, height; get_drawable_size(surface->Dpy, surface->Win, &width, &height); st_resize_framebuffer(surface->Framebuffer, width, height); + surface->Base.Width = width; + surface->Base.Height = height; } @@ -468,6 +470,8 @@ xlib_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) st_notify_swapbuffers(xsurf->Framebuffer); display_surface(pws, psurf, xsurf); + + check_and_update_buffer_size(xsurf); } return EGL_TRUE; -- cgit v1.2.3 From 52676207e338b4c2b4953747521921ab79d70f06 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 12:06:06 -0600 Subject: egl: added eglDestroySurface function --- src/gallium/winsys/egl_xlib/egl_xlib.c | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index b683898dcb..bd2acd0411 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -454,6 +454,28 @@ xlib_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, } +static EGLBoolean +xlib_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) +{ + struct xlib_egl_surface *surf = lookup_surface(surface); + if (surf) { + _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface); + if (surf->Base.IsBound) { + surf->Base.DeletePending = EGL_TRUE; + } + else { + st_unreference_framebuffer(&surf->Framebuffer); + free(surf); + } + return EGL_TRUE; + } + else { + _eglError(EGL_BAD_SURFACE, "eglDestroySurface"); + return EGL_FALSE; + } +} + + static EGLBoolean xlib_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) { @@ -499,6 +521,7 @@ _eglMain(_EGLDisplay *dpy, const char *args) xdrv->Base.API.CreateContext = xlib_eglCreateContext; xdrv->Base.API.DestroyContext = xlib_eglDestroyContext; xdrv->Base.API.CreateWindowSurface = xlib_eglCreateWindowSurface; + xdrv->Base.API.DestroySurface = xlib_eglDestroySurface; xdrv->Base.API.MakeCurrent = xlib_eglMakeCurrent; xdrv->Base.API.SwapBuffers = xlib_eglSwapBuffers; -- cgit v1.2.3 From d5078b94323241a6482f54797756116b1c864327 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 13:45:40 -0600 Subject: egl: clean-up re-org of the client API state --- src/egl/main/eglapi.c | 39 +++++++++++++++++++++++++++------------ src/egl/main/eglcontext.c | 10 +++++++++- src/egl/main/eglcontext.h | 6 +++--- src/egl/main/egldriver.c | 7 +++++-- src/egl/main/egldriver.h | 3 ++- src/egl/main/eglglobals.c | 3 +-- src/egl/main/eglglobals.h | 8 ++++---- src/egl/main/eglmisc.c | 25 +++++++++++++++++++++++-- 8 files changed, 74 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index 367787b243..aacbb6b08e 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -517,6 +517,18 @@ eglQueryModeStringMESA(EGLDisplay dpy, EGLModeMESA mode) #ifdef EGL_VERSION_1_2 + +/** + * Specify the client API to use for subsequent calls including: + * eglCreateContext() + * eglGetCurrentContext() + * eglGetCurrentDisplay() + * eglGetCurrentSurface() + * eglMakeCurrent(when the ctx parameter is EGL NO CONTEXT) + * eglWaitClient() + * eglWaitNative() + * See section 3.7 "Rendering Context" in the EGL specification for details. + */ EGLBoolean eglBindAPI(EGLenum api) { @@ -525,7 +537,7 @@ eglBindAPI(EGLenum api) switch (api) { #ifdef EGL_VERSION_1_4 case EGL_OPENGL_API: - if (_eglGlobal.OpenGLAPISupported) { + if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_BIT) { t->CurrentAPI = api; return EGL_TRUE; } @@ -533,14 +545,14 @@ eglBindAPI(EGLenum api) return EGL_FALSE; #endif case EGL_OPENGL_ES_API: - if (_eglGlobal.OpenGLESAPISupported) { + if (_eglGlobal.ClientAPIsMask & (EGL_OPENGL_ES_BIT | EGL_OPENGL_ES2_BIT)) { t->CurrentAPI = api; return EGL_TRUE; } _eglError(EGL_BAD_PARAMETER, "eglBindAPI"); return EGL_FALSE; case EGL_OPENVG_API: - if (_eglGlobal.OpenVGAPISupported) { + if (_eglGlobal.ClientAPIsMask & EGL_OPENVG_BIT) { t->CurrentAPI = api; return EGL_TRUE; } @@ -553,6 +565,18 @@ eglBindAPI(EGLenum api) } +/** + * Return the last value set with eglBindAPI(). + */ +EGLenum +eglQueryAPI(void) +{ + /* returns one of EGL_OPENGL_API, EGL_OPENGL_ES_API or EGL_OPENVG_API */ + _EGLThreadInfo *t = _eglGetCurrentThread(); + return t->CurrentAPI; +} + + EGLSurface eglCreatePbufferFromClientBuffer(EGLDisplay dpy, EGLenum buftype, EGLClientBuffer buffer, EGLConfig config, @@ -564,15 +588,6 @@ eglCreatePbufferFromClientBuffer(EGLDisplay dpy, EGLenum buftype, } -EGLenum -eglQueryAPI(void) -{ - /* returns one of EGL_OPENGL_API, EGL_OPENGL_ES_API or EGL_OPENVG_API */ - _EGLThreadInfo *t = _eglGetCurrentThread(); - return t->CurrentAPI; -} - - EGLBoolean eglReleaseThread(void) { diff --git a/src/egl/main/eglcontext.c b/src/egl/main/eglcontext.c index 86d1d84334..bf1addf241 100644 --- a/src/egl/main/eglcontext.c +++ b/src/egl/main/eglcontext.c @@ -20,6 +20,12 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, _EGLConfig *conf; _EGLDisplay *display = _eglLookupDisplay(dpy); EGLint i; + const EGLenum api = eglQueryAPI(); + + if (api == EGL_NONE) { + _eglError(EGL_BAD_MATCH, "eglCreateContext(no client API)"); + return EGL_FALSE; + } conf = _eglLookupConfig(drv, dpy, config); if (!conf) { @@ -30,7 +36,8 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { switch (attrib_list[i]) { case EGL_CONTEXT_CLIENT_VERSION: - /* xxx todo */ + i++; + ctx->ClientVersion = attrib_list[i]; break; default: _eglError(EGL_BAD_ATTRIBUTE, "_eglInitContext"); @@ -43,6 +50,7 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, ctx->Config = conf; ctx->DrawSurface = EGL_NO_SURFACE; ctx->ReadSurface = EGL_NO_SURFACE; + ctx->ClientAPI = api; return EGL_TRUE; } diff --git a/src/egl/main/eglcontext.h b/src/egl/main/eglcontext.h index 093f65fbfb..34fee9c637 100644 --- a/src/egl/main/eglcontext.h +++ b/src/egl/main/eglcontext.h @@ -20,9 +20,9 @@ struct _egl_context EGLBoolean IsBound; EGLBoolean DeletePending; -#ifdef EGL_VERSION_1_2 - EGLint ClientAPI; /* Either EGL_OPENGL_ES_API or EGL_OPENVG_API */ -#endif /* EGL_VERSION_1_2 */ + + EGLint ClientAPI; /**< EGL_OPENGL_ES_API, EGL_OPENGL_API, EGL_OPENVG_API */ + EGLint ClientVersion; /**< 1 = OpenGLES 1.x, 2 = OpenGLES 2.x */ }; diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index e155f5fb98..681be47202 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -97,12 +97,12 @@ _eglChooseDriver(_EGLDisplay *dpy) if (driverName) return _eglstrdup(driverName); - +#if 0 if (!displayString) { /* choose a default */ displayString = DefaultDriverName; } - +#endif /* extract default DriverArgs = whatever follows ':' */ if (displayString && (displayString[0] == '!' || @@ -192,6 +192,9 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) else dlclose(lib); + /* update the global notion of supported APIs */ + _eglGlobal.ClientAPIsMask |= drv->ClientAPIsMask; + return drv; } diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index e0cedb22c3..09d7755f5d 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -33,7 +33,8 @@ struct _egl_driver int APImajor, APIminor; /**< as returned by eglInitialize() */ char Version[1000]; /**< initialized from APImajor/minor, Name */ - const char *ClientAPIs; + /** Bitmask of supported APIs (EGL_xx_BIT) set by the driver during init */ + EGLint ClientAPIsMask; _EGLAPI API; /**< EGL API dispatch table */ diff --git a/src/egl/main/eglglobals.c b/src/egl/main/eglglobals.c index 4bdc0329bf..9a077edbd7 100644 --- a/src/egl/main/eglglobals.c +++ b/src/egl/main/eglglobals.c @@ -19,8 +19,7 @@ _eglInitGlobals(void) _eglGlobal.FreeScreenHandle = 1; _eglGlobal.Initialized = EGL_TRUE; - _eglGlobal.OpenGLESAPISupported = EGL_TRUE; - _eglGlobal.OpenVGAPISupported = EGL_FALSE; + _eglGlobal.ClientAPIsMask = 0x0; /* XXX temporary */ _eglGlobal.ThreadInfo = _eglNewThreadInfo(); diff --git a/src/egl/main/eglglobals.h b/src/egl/main/eglglobals.h index 6b858b6b08..ac85f8b8a4 100644 --- a/src/egl/main/eglglobals.h +++ b/src/egl/main/eglglobals.h @@ -28,10 +28,10 @@ struct _egl_global EGLScreenMESA FreeScreenHandle; - /* XXX these may be temporary */ - EGLBoolean OpenGLAPISupported; - EGLBoolean OpenGLESAPISupported; - EGLBoolean OpenVGAPISupported; + /* bitmaks of supported APIs (supported by _some_ driver) */ + EGLint ClientAPIsMask; + + char ClientAPIs[1000]; /**< updated by eglQueryString */ /* XXX temporary - should be thread-specific data (TSD) */ _EGLThreadInfo *ThreadInfo; diff --git a/src/egl/main/eglmisc.c b/src/egl/main/eglmisc.c index fb32544b15..b5bdc3ea4b 100644 --- a/src/egl/main/eglmisc.c +++ b/src/egl/main/eglmisc.c @@ -54,6 +54,27 @@ _eglUpdateExtensionsString(_EGLDriver *drv) } +static void +_eglUpdateAPIsString(_EGLDriver *drv) +{ + _eglGlobal.ClientAPIs[0] = 0; + + if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_BIT) + strcat(_eglGlobal.ClientAPIs, "OpenGL "); + + if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_ES_BIT) + strcat(_eglGlobal.ClientAPIs, "OpenGL_ES "); + + if (_eglGlobal.ClientAPIsMask & EGL_OPENGL_ES2_BIT) + strcat(_eglGlobal.ClientAPIs, "OpenGL_ES2 "); + + if (_eglGlobal.ClientAPIsMask & EGL_OPENVG_BIT) + strcat(_eglGlobal.ClientAPIs, "OpenVG "); + + assert(strlen(_eglGlobal.ClientAPIs) < sizeof(_eglGlobal.ClientAPIs)); +} + + const char * _eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name) @@ -70,8 +91,8 @@ _eglQueryString(_EGLDriver *drv, EGLDisplay dpy, EGLint name) return drv->Extensions.String; #ifdef EGL_VERSION_1_2 case EGL_CLIENT_APIS: - /* XXX need to initialize somewhere */ - return drv->ClientAPIs; + _eglUpdateAPIsString(drv); + return _eglGlobal.ClientAPIs; #endif default: _eglError(EGL_BAD_PARAMETER, "eglQueryString"); -- cgit v1.2.3 From ba7aeb8b34c2e310b6ba0d3dbfdc68b0d6ab82c8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 13:46:31 -0600 Subject: egl: specify client API by bit flag, not string --- src/gallium/winsys/egl_xlib/egl_xlib.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index bd2acd0411..1458f83c02 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -302,6 +302,12 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, return EGL_NO_CONTEXT; } + if (ctx->Base.ClientAPI != EGL_OPENGL_API) { + _eglError(EGL_BAD_MATCH, "eglCreateContext(only OpenGL API supported)"); + free(ctx); + return EGL_NO_CONTEXT; + } + /* create a softpipe context */ ctx->pipe = softpipe_create(xdrv->screen, xdrv->winsys, NULL); @@ -309,6 +315,7 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, _eglConfigToContextModesRec(conf, &visual); ctx->Context = st_create_context(ctx->pipe, &visual, share_ctx); + _eglSaveContext(&ctx->Base); return _eglGetContextHandle(&ctx->Base); } @@ -525,8 +532,8 @@ _eglMain(_EGLDisplay *dpy, const char *args) xdrv->Base.API.MakeCurrent = xlib_eglMakeCurrent; xdrv->Base.API.SwapBuffers = xlib_eglSwapBuffers; + xdrv->Base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; - xdrv->Base.ClientAPIs = "OpenGL"; /* "OpenGL_ES" */ xdrv->Base.Name = "Xlib/softpipe"; /* create one winsys and use it for all contexts/surfaces */ @@ -535,7 +542,6 @@ _eglMain(_EGLDisplay *dpy, const char *args) xdrv->screen = softpipe_create_screen(xdrv->winsys); - return &xdrv->Base; } -- cgit v1.2.3 From e3805cad0d15ed25ce8f6c5a1f1ea913e5d0986a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 14:50:33 -0600 Subject: egl: new eglGetProcAddress() code The idea is to pass the call down to the device driver where an API-specific query can be made. Untested. --- src/egl/main/eglapi.c | 13 ++++++++----- src/egl/main/eglapi.h | 11 ++++++++++- src/egl/main/egldriver.c | 23 +++++++++++++++++++++-- src/egl/main/egldriver.h | 4 ++++ src/egl/main/eglglobals.h | 3 +++ src/egl/main/egltypedefs.h | 2 -- 6 files changed, 46 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index aacbb6b08e..49d1f3d0eb 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -385,12 +385,15 @@ void (* EGLAPIENTRY eglGetProcAddress(const char *procname))() return (genericFunc) egl_functions[i].function; } } -#if 0 - /* XXX enable this code someday */ - return (genericFunc) _glapi_get_proc_address(procname); -#else + + /* now loop over drivers to query their procs */ + for (i = 0; i < _eglGlobal.NumDrivers; i++) { + _EGLProc p = _eglGlobal.Drivers[i]->API.GetProcAddress(procname); + if (p) + return p; + } + return NULL; -#endif } diff --git a/src/egl/main/eglapi.h b/src/egl/main/eglapi.h index 555aa5dd9e..f6163a0c7a 100644 --- a/src/egl/main/eglapi.h +++ b/src/egl/main/eglapi.h @@ -2,10 +2,15 @@ #define EGLAPI_INCLUDED /** - * Typedefs for all EGL API entrypoint functions. + * A generic function ptr type */ +typedef void (*_EGLProc)(); +/** + * Typedefs for all EGL API entrypoint functions. + */ + /* driver funcs */ typedef EGLBoolean (*Initialize_t)(_EGLDriver *, EGLDisplay dpy, EGLint *major, EGLint *minor); typedef EGLBoolean (*Terminate_t)(_EGLDriver *, EGLDisplay dpy); @@ -39,6 +44,9 @@ typedef const char *(*QueryString_t)(_EGLDriver *drv, EGLDisplay dpy, EGLint nam typedef EGLBoolean (*WaitGL_t)(_EGLDriver *drv, EGLDisplay dpy); typedef EGLBoolean (*WaitNative_t)(_EGLDriver *drv, EGLDisplay dpy, EGLint engine); +typedef _EGLProc (*GetProcAddress_t)(const char *procname); + + #ifdef EGL_MESA_screen_surface typedef EGLBoolean (*ChooseModeMESA_t)(_EGLDriver *drv, EGLDisplay dpy, EGLScreenMESA screen, const EGLint *attrib_list, EGLModeMESA *modes, EGLint modes_size, EGLint *num_modes); @@ -95,6 +103,7 @@ struct _egl_api QueryString_t QueryString; WaitGL_t WaitGL; WaitNative_t WaitNative; + GetProcAddress_t GetProcAddress; /* EGL_MESA_screen extension */ ChooseModeMESA_t ChooseModeMESA; diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 681be47202..80fa49b7ab 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -30,8 +30,8 @@ #endif -const char *DefaultDriverName = ":0"; -const char *SysFS = "/sys/class"; +static const char *DefaultDriverName = ":0"; +static const char *SysFS = "/sys/class"; @@ -72,6 +72,9 @@ _eglChooseDRMDriver(int card) /** + * XXX this function is totally subject change!!! + * + * * Determine/return the name of the driver to use for the given _EGLDisplay. * * Try to be clever and determine if nativeDisplay is an Xlib Display @@ -92,6 +95,8 @@ _eglChooseDriver(_EGLDisplay *dpy) const char *displayString = (const char *) dpy->NativeDisplay; const char *driverName = NULL; + (void) DefaultDriverName; + /* First, if the EGL_DRIVER env var is set, use that */ driverName = getenv("EGL_DRIVER"); if (driverName) @@ -139,6 +144,8 @@ _eglChooseDriver(_EGLDisplay *dpy) driverName = _weglChooseDriver(dpy); #elif defined(_EGL_PLATFORM_WINCE) /* XXX to do */ +#else + driverName = DefaultDriverName; #endif } @@ -195,6 +202,8 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) /* update the global notion of supported APIs */ _eglGlobal.ClientAPIsMask |= drv->ClientAPIsMask; + _eglSaveDriver(drv); + return drv; } @@ -217,6 +226,16 @@ _eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy) } +/** + * Save the given driver pointer in the list of all known drivers. + */ +void +_eglSaveDriver(_EGLDriver *drv) +{ + _eglGlobal.Drivers[ _eglGlobal.NumDrivers++ ] = drv; +} + + /** * Given a display handle, return the _EGLDriver for that display. */ diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index 09d7755f5d..af028eaf22 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -57,6 +57,10 @@ extern EGLBoolean _eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy); +extern void +_eglSaveDriver(_EGLDriver *drv); + + extern _EGLDriver * _eglLookupDriver(EGLDisplay d); diff --git a/src/egl/main/eglglobals.h b/src/egl/main/eglglobals.h index ac85f8b8a4..14d8ea487a 100644 --- a/src/egl/main/eglglobals.h +++ b/src/egl/main/eglglobals.h @@ -35,6 +35,9 @@ struct _egl_global /* XXX temporary - should be thread-specific data (TSD) */ _EGLThreadInfo *ThreadInfo; + + EGLint NumDrivers; + _EGLDriver *Drivers[10]; }; diff --git a/src/egl/main/egltypedefs.h b/src/egl/main/egltypedefs.h index b1c8ec1f02..efbb17a4dc 100644 --- a/src/egl/main/egltypedefs.h +++ b/src/egl/main/egltypedefs.h @@ -28,8 +28,6 @@ typedef struct _egl_surface _EGLSurface; typedef struct _egl_thread_info _EGLThreadInfo; -typedef void (*_EGLProc)(); - typedef _EGLDriver *(*_EGLMain_t)(_EGLDisplay *dpy, const char *args); -- cgit v1.2.3 From 3c4162ed2afe4da252d2a32c14e486515935ba39 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 14:51:40 -0600 Subject: egl: eglGetProcAddress() stub --- src/gallium/winsys/egl_xlib/egl_xlib.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 1458f83c02..99389970f5 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -176,6 +176,17 @@ xlib_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) } +static _EGLProc +xlib_eglGetProcAddress(const char *procname) +{ + /* XXX for each supported API, evaluate GetProcAddress(name) */ + /* + return _glapi_get_proc_address(procname); + */ + return NULL; +} + + static void get_drawable_visual_info(Display *dpy, Drawable d, XVisualInfo *visInfo) { @@ -525,6 +536,7 @@ _eglMain(_EGLDisplay *dpy, const char *args) _eglInitDriverFallbacks(&xdrv->Base); xdrv->Base.API.Initialize = xlib_eglInitialize; xdrv->Base.API.Terminate = xlib_eglTerminate; + xdrv->Base.API.GetProcAddress = xlib_eglGetProcAddress; xdrv->Base.API.CreateContext = xlib_eglCreateContext; xdrv->Base.API.DestroyContext = xlib_eglDestroyContext; xdrv->Base.API.CreateWindowSurface = xlib_eglCreateWindowSurface; -- cgit v1.2.3 From 12bb9075221da6bf8121f3d705888f29ca11cc0f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 15:07:58 -0600 Subject: gallium: fix some PIPE_FORMAT_Z24S8_UNORM bugs in softpipe driver --- src/gallium/drivers/softpipe/sp_quad_depth_test.c | 2 +- src/gallium/drivers/softpipe/sp_tile_cache.c | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c index a1859f9883..33888abcc5 100644 --- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c @@ -230,7 +230,7 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) int x = quad->x0 % TILE_SIZE + (j & 1); int y = quad->y0 % TILE_SIZE + (j >> 1); uint z24s8 = tile->data.depth32[y][x]; - z24s8 = (z24s8 & 0xff) | (bzzzz[j] << 24); + z24s8 = (z24s8 & 0xff) | (bzzzz[j] << 8); tile->data.depth32[y][x] = z24s8; } break; diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 28c29da87c..0e4c8c41ee 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -169,6 +169,7 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, PIPE_BUFFER_USAGE_CPU_WRITE); tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || + ps->format == PIPE_FORMAT_Z24S8_UNORM || ps->format == PIPE_FORMAT_Z16_UNORM || ps->format == PIPE_FORMAT_Z32_UNORM || ps->format == PIPE_FORMAT_S8_UNORM); -- cgit v1.2.3 From a9d6b1afa5b548a98c2b95db50236c4139d9f569 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 31 May 2008 15:18:55 +0900 Subject: gallium: Port timing functions to WinCE. --- src/gallium/auxiliary/util/u_time.c | 74 ++++++++++++++++++++++++++++--------- 1 file changed, 56 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index 9b97050d51..c6ce0afab0 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -33,26 +33,44 @@ */ -#include "util/u_time.h" +#include "pipe/p_config.h" #if defined(PIPE_OS_LINUX) #include #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) #include #include -#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) +#include +extern VOID KeQuerySystemTime(PLARGE_INTEGER); +#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) #include #else #error Unsupported OS #endif +#include "util/u_time.h" + -#if defined(PIPE_OS_WINDOWS) -static LONGLONG frequency = 0; -#if !defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) -#define EngQueryPerformanceFrequency(p) QueryPerformanceFrequency((LARGE_INTEGER*)(p)) -#define EngQueryPerformanceCounter(p) QueryPerformanceCounter((LARGE_INTEGER*)(p)) +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) + +static int64_t frequency = 0; + +static INLINE void +util_time_get_frequency(void) +{ + if(!frequency) { +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) + LONGLONG temp; + EngQueryPerformanceFrequency(&temp); + frequency = temp; +#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) + LARGE_INTEGER temp; + QueryPerformanceFrequency(&temp); + frequency = temp.QuadPart; #endif + } +} #endif @@ -61,8 +79,20 @@ util_time_get(struct util_time *t) { #if defined(PIPE_OS_LINUX) gettimeofday(&t->tv, NULL); -#elif defined(PIPE_OS_WINDOWS) - EngQueryPerformanceCounter(&t->counter); +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) + LONGLONG temp; + EngQueryPerformanceCounter(&temp); + t->counter = temp; +#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) + /* Updated every 10 miliseconds, measured in units of 100 nanoseconds. + * http://msdn.microsoft.com/en-us/library/ms801642.aspx */ + LARGE_INTEGER temp; + KeQuerySystemTime(&temp); + t->counter = temp.QuadPart; +#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) + LARGE_INTEGER temp; + QueryPerformanceCounter(&temp); + t->counter = temp.QuadPart; #endif } @@ -75,10 +105,17 @@ util_time_add(const struct util_time *t1, #if defined(PIPE_OS_LINUX) t2->tv.tv_sec = t1->tv.tv_sec + usecs / 1000000; t2->tv.tv_usec = t1->tv.tv_usec + usecs % 1000000; -#elif defined(PIPE_OS_WINDOWS) - if(!frequency) - EngQueryPerformanceFrequency(&frequency); - t2->counter = t1->counter + (usecs * frequency + 999999LL)/1000000LL; +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) + util_time_get_frequency(); + t2->counter = t1->counter + (usecs * frequency + INT64_C(999999))/INT64_C(1000000); +#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) + /* 1 tick = 100 nano seconds. */ + t2->counter = t1->counter + usecs * 10; +#elif + LARGE_INTEGER temp; + LONGLONG freq; + freq = temp.QuadPart; + t2->counter = t1->counter + (usecs * freq)/1000000L; #endif } @@ -90,10 +127,11 @@ util_time_diff(const struct util_time *t1, #if defined(PIPE_OS_LINUX) return (t2->tv.tv_usec - t1->tv.tv_usec) + (t2->tv.tv_sec - t1->tv.tv_sec)*1000000; -#elif defined(PIPE_OS_WINDOWS) - if(!frequency) - EngQueryPerformanceFrequency(&frequency); - return (t2->counter - t1->counter)*1000000LL/frequency; +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) + util_time_get_frequency(); + return (t2->counter - t1->counter)*INT64_C(1000000)/frequency; +#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) + return (t2->counter - t1->counter)/10; #endif } @@ -142,7 +180,7 @@ util_time_timeout(const struct util_time *start, } -#if defined(PIPE_OS_WINDOWS) +#if defined(PIPE_SUBSYSYEM_WINDOWS_DISPLAY) void util_time_sleep(unsigned usecs) { LONGLONG start, curr, end; -- cgit v1.2.3 From 054189e87abcd399aadb449d13b2e331c060e65f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 31 May 2008 18:07:39 +0900 Subject: gallium: MSVC 8.0 already defines the cosf, sinf, etc. --- src/gallium/include/pipe/p_util.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 597354ca96..fe4ba3a689 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -410,8 +410,7 @@ extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, -#ifdef WIN32 - +#if !defined(_MSC_VER) && _MSC_VER < 0x800 #if !defined(_INC_MATH) || !defined(__cplusplus) static INLINE float cosf( float f ) @@ -453,6 +452,7 @@ static INLINE float logf( float f ) { return (float) cos( (double) f ); } + #endif /* _INC_MATH */ #endif -- cgit v1.2.3 From 45b2c23d7a2f1bd723d0719b13470125de09c243 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 31 May 2008 18:08:24 +0900 Subject: i915: Eliminate void pointer arithmetic. --- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index faf5971abb..cdbb5feb23 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -142,7 +142,7 @@ i915_vbuf_render_allocate_vertices( struct vbuf_render *render, i915->vbo_offset = i915_render->vbo_offset; i915->dirty |= I915_NEW_VBO; - return i915_render->vbo_ptr + i915->vbo_offset; + return (unsigned char *)i915_render->vbo_ptr + i915->vbo_offset; } -- cgit v1.2.3 From 53174afeeb68a79e471185cb463c13ff90af698f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 31 May 2008 18:14:09 +0900 Subject: mesa: Apply MSVC portability fixes from Alan Hourihane. --- src/mesa/main/api_arrayelt.c | 96 +++++++++++++------------- src/mesa/main/api_validate.c | 2 +- src/mesa/main/context.c | 2 +- src/mesa/main/dlist.c | 2 +- src/mesa/main/drawpix.c | 2 +- src/mesa/main/framebuffer.c | 2 +- src/mesa/main/get.c | 4 +- src/mesa/main/image.c | 6 +- src/mesa/main/light.c | 2 +- src/mesa/main/mipmap.c | 2 +- src/mesa/main/pixel.c | 4 +- src/mesa/main/queryobj.c | 4 +- src/mesa/main/rastpos.c | 2 +- src/mesa/main/texenvprogram.c | 10 +-- src/mesa/shader/prog_execute.c | 6 +- src/mesa/shader/prog_statevars.c | 10 +-- src/mesa/shader/prog_uniform.c | 2 +- src/mesa/shader/shader_api.c | 6 +- src/mesa/state_tracker/st_atom_pixeltransfer.c | 2 +- src/mesa/state_tracker/st_atom_scissor.c | 8 +-- src/mesa/state_tracker/st_atom_viewport.c | 12 ++-- src/mesa/state_tracker/st_cb_bitmap.c | 38 +++++----- src/mesa/vbo/vbo_exec_array.c | 2 +- src/mesa/vbo/vbo_split_inplace.c | 2 +- 24 files changed, 114 insertions(+), 114 deletions(-) (limited to 'src') diff --git a/src/mesa/main/api_arrayelt.c b/src/mesa/main/api_arrayelt.c index 72091b0789..d124c724c9 100644 --- a/src/mesa/main/api_arrayelt.c +++ b/src/mesa/main/api_arrayelt.c @@ -166,7 +166,7 @@ static void GLAPIENTRY VertexAttrib1NbvNV(GLuint index, const GLbyte *v) static void GLAPIENTRY VertexAttrib1bvNV(GLuint index, const GLbyte *v) { - CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NbvNV(GLuint index, const GLbyte *v) @@ -176,7 +176,7 @@ static void GLAPIENTRY VertexAttrib2NbvNV(GLuint index, const GLbyte *v) static void GLAPIENTRY VertexAttrib2bvNV(GLuint index, const GLbyte *v) { - CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NbvNV(GLuint index, const GLbyte *v) @@ -188,7 +188,7 @@ static void GLAPIENTRY VertexAttrib3NbvNV(GLuint index, const GLbyte *v) static void GLAPIENTRY VertexAttrib3bvNV(GLuint index, const GLbyte *v) { - CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NbvNV(GLuint index, const GLbyte *v) @@ -201,7 +201,7 @@ static void GLAPIENTRY VertexAttrib4NbvNV(GLuint index, const GLbyte *v) static void GLAPIENTRY VertexAttrib4bvNV(GLuint index, const GLbyte *v) { - CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_UNSIGNED_BYTE attributes */ @@ -213,7 +213,7 @@ static void GLAPIENTRY VertexAttrib1NubvNV(GLuint index, const GLubyte *v) static void GLAPIENTRY VertexAttrib1ubvNV(GLuint index, const GLubyte *v) { - CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NubvNV(GLuint index, const GLubyte *v) @@ -224,7 +224,7 @@ static void GLAPIENTRY VertexAttrib2NubvNV(GLuint index, const GLubyte *v) static void GLAPIENTRY VertexAttrib2ubvNV(GLuint index, const GLubyte *v) { - CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NubvNV(GLuint index, const GLubyte *v) @@ -235,7 +235,7 @@ static void GLAPIENTRY VertexAttrib3NubvNV(GLuint index, const GLubyte *v) } static void GLAPIENTRY VertexAttrib3ubvNV(GLuint index, const GLubyte *v) { - CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NubvNV(GLuint index, const GLubyte *v) @@ -248,7 +248,7 @@ static void GLAPIENTRY VertexAttrib4NubvNV(GLuint index, const GLubyte *v) static void GLAPIENTRY VertexAttrib4ubvNV(GLuint index, const GLubyte *v) { - CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_SHORT attributes */ @@ -260,7 +260,7 @@ static void GLAPIENTRY VertexAttrib1NsvNV(GLuint index, const GLshort *v) static void GLAPIENTRY VertexAttrib1svNV(GLuint index, const GLshort *v) { - CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NsvNV(GLuint index, const GLshort *v) @@ -271,7 +271,7 @@ static void GLAPIENTRY VertexAttrib2NsvNV(GLuint index, const GLshort *v) static void GLAPIENTRY VertexAttrib2svNV(GLuint index, const GLshort *v) { - CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NsvNV(GLuint index, const GLshort *v) @@ -283,7 +283,7 @@ static void GLAPIENTRY VertexAttrib3NsvNV(GLuint index, const GLshort *v) static void GLAPIENTRY VertexAttrib3svNV(GLuint index, const GLshort *v) { - CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NsvNV(GLuint index, const GLshort *v) @@ -296,7 +296,7 @@ static void GLAPIENTRY VertexAttrib4NsvNV(GLuint index, const GLshort *v) static void GLAPIENTRY VertexAttrib4svNV(GLuint index, const GLshort *v) { - CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_UNSIGNED_SHORT attributes */ @@ -308,7 +308,7 @@ static void GLAPIENTRY VertexAttrib1NusvNV(GLuint index, const GLushort *v) static void GLAPIENTRY VertexAttrib1usvNV(GLuint index, const GLushort *v) { - CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NusvNV(GLuint index, const GLushort *v) @@ -319,7 +319,7 @@ static void GLAPIENTRY VertexAttrib2NusvNV(GLuint index, const GLushort *v) static void GLAPIENTRY VertexAttrib2usvNV(GLuint index, const GLushort *v) { - CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NusvNV(GLuint index, const GLushort *v) @@ -331,7 +331,7 @@ static void GLAPIENTRY VertexAttrib3NusvNV(GLuint index, const GLushort *v) static void GLAPIENTRY VertexAttrib3usvNV(GLuint index, const GLushort *v) { - CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NusvNV(GLuint index, const GLushort *v) @@ -344,7 +344,7 @@ static void GLAPIENTRY VertexAttrib4NusvNV(GLuint index, const GLushort *v) static void GLAPIENTRY VertexAttrib4usvNV(GLuint index, const GLushort *v) { - CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_INT attributes */ @@ -356,7 +356,7 @@ static void GLAPIENTRY VertexAttrib1NivNV(GLuint index, const GLint *v) static void GLAPIENTRY VertexAttrib1ivNV(GLuint index, const GLint *v) { - CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NivNV(GLuint index, const GLint *v) @@ -367,7 +367,7 @@ static void GLAPIENTRY VertexAttrib2NivNV(GLuint index, const GLint *v) static void GLAPIENTRY VertexAttrib2ivNV(GLuint index, const GLint *v) { - CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NivNV(GLuint index, const GLint *v) @@ -379,7 +379,7 @@ static void GLAPIENTRY VertexAttrib3NivNV(GLuint index, const GLint *v) static void GLAPIENTRY VertexAttrib3ivNV(GLuint index, const GLint *v) { - CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NivNV(GLuint index, const GLint *v) @@ -392,7 +392,7 @@ static void GLAPIENTRY VertexAttrib4NivNV(GLuint index, const GLint *v) static void GLAPIENTRY VertexAttrib4ivNV(GLuint index, const GLint *v) { - CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_UNSIGNED_INT attributes */ @@ -404,7 +404,7 @@ static void GLAPIENTRY VertexAttrib1NuivNV(GLuint index, const GLuint *v) static void GLAPIENTRY VertexAttrib1uivNV(GLuint index, const GLuint *v) { - CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fNV(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NuivNV(GLuint index, const GLuint *v) @@ -415,7 +415,7 @@ static void GLAPIENTRY VertexAttrib2NuivNV(GLuint index, const GLuint *v) static void GLAPIENTRY VertexAttrib2uivNV(GLuint index, const GLuint *v) { - CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NuivNV(GLuint index, const GLuint *v) @@ -427,7 +427,7 @@ static void GLAPIENTRY VertexAttrib3NuivNV(GLuint index, const GLuint *v) static void GLAPIENTRY VertexAttrib3uivNV(GLuint index, const GLuint *v) { - CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NuivNV(GLuint index, const GLuint *v) @@ -440,7 +440,7 @@ static void GLAPIENTRY VertexAttrib4NuivNV(GLuint index, const GLuint *v) static void GLAPIENTRY VertexAttrib4uivNV(GLuint index, const GLuint *v) { - CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fNV(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_FLOAT attributes */ @@ -602,7 +602,7 @@ static void GLAPIENTRY VertexAttrib1NbvARB(GLuint index, const GLbyte *v) static void GLAPIENTRY VertexAttrib1bvARB(GLuint index, const GLbyte *v) { - CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NbvARB(GLuint index, const GLbyte *v) @@ -612,7 +612,7 @@ static void GLAPIENTRY VertexAttrib2NbvARB(GLuint index, const GLbyte *v) static void GLAPIENTRY VertexAttrib2bvARB(GLuint index, const GLbyte *v) { - CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NbvARB(GLuint index, const GLbyte *v) @@ -624,7 +624,7 @@ static void GLAPIENTRY VertexAttrib3NbvARB(GLuint index, const GLbyte *v) static void GLAPIENTRY VertexAttrib3bvARB(GLuint index, const GLbyte *v) { - CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NbvARB(GLuint index, const GLbyte *v) @@ -637,7 +637,7 @@ static void GLAPIENTRY VertexAttrib4NbvARB(GLuint index, const GLbyte *v) static void GLAPIENTRY VertexAttrib4bvARB(GLuint index, const GLbyte *v) { - CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_UNSIGNED_BYTE attributes */ @@ -649,7 +649,7 @@ static void GLAPIENTRY VertexAttrib1NubvARB(GLuint index, const GLubyte *v) static void GLAPIENTRY VertexAttrib1ubvARB(GLuint index, const GLubyte *v) { - CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NubvARB(GLuint index, const GLubyte *v) @@ -660,7 +660,7 @@ static void GLAPIENTRY VertexAttrib2NubvARB(GLuint index, const GLubyte *v) static void GLAPIENTRY VertexAttrib2ubvARB(GLuint index, const GLubyte *v) { - CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NubvARB(GLuint index, const GLubyte *v) @@ -671,7 +671,7 @@ static void GLAPIENTRY VertexAttrib3NubvARB(GLuint index, const GLubyte *v) } static void GLAPIENTRY VertexAttrib3ubvARB(GLuint index, const GLubyte *v) { - CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NubvARB(GLuint index, const GLubyte *v) @@ -684,7 +684,7 @@ static void GLAPIENTRY VertexAttrib4NubvARB(GLuint index, const GLubyte *v) static void GLAPIENTRY VertexAttrib4ubvARB(GLuint index, const GLubyte *v) { - CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_SHORT attributes */ @@ -696,7 +696,7 @@ static void GLAPIENTRY VertexAttrib1NsvARB(GLuint index, const GLshort *v) static void GLAPIENTRY VertexAttrib1svARB(GLuint index, const GLshort *v) { - CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NsvARB(GLuint index, const GLshort *v) @@ -707,7 +707,7 @@ static void GLAPIENTRY VertexAttrib2NsvARB(GLuint index, const GLshort *v) static void GLAPIENTRY VertexAttrib2svARB(GLuint index, const GLshort *v) { - CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NsvARB(GLuint index, const GLshort *v) @@ -719,7 +719,7 @@ static void GLAPIENTRY VertexAttrib3NsvARB(GLuint index, const GLshort *v) static void GLAPIENTRY VertexAttrib3svARB(GLuint index, const GLshort *v) { - CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NsvARB(GLuint index, const GLshort *v) @@ -732,7 +732,7 @@ static void GLAPIENTRY VertexAttrib4NsvARB(GLuint index, const GLshort *v) static void GLAPIENTRY VertexAttrib4svARB(GLuint index, const GLshort *v) { - CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_UNSIGNED_SHORT attributes */ @@ -744,7 +744,7 @@ static void GLAPIENTRY VertexAttrib1NusvARB(GLuint index, const GLushort *v) static void GLAPIENTRY VertexAttrib1usvARB(GLuint index, const GLushort *v) { - CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NusvARB(GLuint index, const GLushort *v) @@ -755,7 +755,7 @@ static void GLAPIENTRY VertexAttrib2NusvARB(GLuint index, const GLushort *v) static void GLAPIENTRY VertexAttrib2usvARB(GLuint index, const GLushort *v) { - CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NusvARB(GLuint index, const GLushort *v) @@ -767,7 +767,7 @@ static void GLAPIENTRY VertexAttrib3NusvARB(GLuint index, const GLushort *v) static void GLAPIENTRY VertexAttrib3usvARB(GLuint index, const GLushort *v) { - CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NusvARB(GLuint index, const GLushort *v) @@ -780,7 +780,7 @@ static void GLAPIENTRY VertexAttrib4NusvARB(GLuint index, const GLushort *v) static void GLAPIENTRY VertexAttrib4usvARB(GLuint index, const GLushort *v) { - CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_INT attributes */ @@ -792,7 +792,7 @@ static void GLAPIENTRY VertexAttrib1NivARB(GLuint index, const GLint *v) static void GLAPIENTRY VertexAttrib1ivARB(GLuint index, const GLint *v) { - CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NivARB(GLuint index, const GLint *v) @@ -803,7 +803,7 @@ static void GLAPIENTRY VertexAttrib2NivARB(GLuint index, const GLint *v) static void GLAPIENTRY VertexAttrib2ivARB(GLuint index, const GLint *v) { - CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NivARB(GLuint index, const GLint *v) @@ -815,7 +815,7 @@ static void GLAPIENTRY VertexAttrib3NivARB(GLuint index, const GLint *v) static void GLAPIENTRY VertexAttrib3ivARB(GLuint index, const GLint *v) { - CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NivARB(GLuint index, const GLint *v) @@ -828,7 +828,7 @@ static void GLAPIENTRY VertexAttrib4NivARB(GLuint index, const GLint *v) static void GLAPIENTRY VertexAttrib4ivARB(GLuint index, const GLint *v) { - CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_UNSIGNED_INT attributes */ @@ -840,7 +840,7 @@ static void GLAPIENTRY VertexAttrib1NuivARB(GLuint index, const GLuint *v) static void GLAPIENTRY VertexAttrib1uivARB(GLuint index, const GLuint *v) { - CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, v[0])); + CALL_VertexAttrib1fARB(GET_DISPATCH(), (index, (GLfloat)v[0])); } static void GLAPIENTRY VertexAttrib2NuivARB(GLuint index, const GLuint *v) @@ -851,7 +851,7 @@ static void GLAPIENTRY VertexAttrib2NuivARB(GLuint index, const GLuint *v) static void GLAPIENTRY VertexAttrib2uivARB(GLuint index, const GLuint *v) { - CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, v[0], v[1])); + CALL_VertexAttrib2fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1])); } static void GLAPIENTRY VertexAttrib3NuivARB(GLuint index, const GLuint *v) @@ -863,7 +863,7 @@ static void GLAPIENTRY VertexAttrib3NuivARB(GLuint index, const GLuint *v) static void GLAPIENTRY VertexAttrib3uivARB(GLuint index, const GLuint *v) { - CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, v[0], v[1], v[2])); + CALL_VertexAttrib3fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2])); } static void GLAPIENTRY VertexAttrib4NuivARB(GLuint index, const GLuint *v) @@ -876,7 +876,7 @@ static void GLAPIENTRY VertexAttrib4NuivARB(GLuint index, const GLuint *v) static void GLAPIENTRY VertexAttrib4uivARB(GLuint index, const GLuint *v) { - CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, v[0], v[1], v[2], v[3])); + CALL_VertexAttrib4fARB(GET_DISPATCH(), (index, (GLfloat)v[0], (GLfloat)v[1], (GLfloat)v[2], (GLfloat)v[3])); } /* GL_FLOAT attributes */ diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c index 64ab324af2..9144b4bc66 100644 --- a/src/mesa/main/api_validate.c +++ b/src/mesa/main/api_validate.c @@ -87,7 +87,7 @@ _mesa_validate_DrawElements(GLcontext *ctx, indexBytes = count * sizeof(GLushort); } - if (indexBytes > ctx->Array.ElementArrayBufferObj->Size) { + if (indexBytes > (GLuint) ctx->Array.ElementArrayBufferObj->Size) { _mesa_warning(ctx, "glDrawElements index out of buffer bounds"); return GL_FALSE; } diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 2158eb6873..0053180000 100755 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1532,7 +1532,7 @@ _mesa_make_current( GLcontext *newCtx, GLframebuffer *drawBuffer, * if the DRIdrawable changes, and everything relies on them. * This is a bit messy (same as needed in _mesa_BindFramebufferEXT) */ - int i; + unsigned int i; GLenum buffers[MAX_DRAW_BUFFERS]; _mesa_reference_framebuffer(&newCtx->DrawBuffer, drawBuffer); diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c index 13ebd4dd0d..f933580b2d 100644 --- a/src/mesa/main/dlist.c +++ b/src/mesa/main/dlist.c @@ -5069,7 +5069,7 @@ save_Indexfv(const GLfloat * v) static void GLAPIENTRY save_EdgeFlag(GLboolean x) { - save_Attr1fNV(VERT_ATTRIB_EDGEFLAG, x ? 1.0 : 0.0); + save_Attr1fNV(VERT_ATTRIB_EDGEFLAG, x ? (GLfloat)1.0 : (GLfloat)0.0); } static void GLAPIENTRY diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c index fa422bb3c7..016ddd0a81 100644 --- a/src/mesa/main/drawpix.c +++ b/src/mesa/main/drawpix.c @@ -377,7 +377,7 @@ _mesa_Bitmap( GLsizei width, GLsizei height, if (ctx->RenderMode == GL_RENDER) { /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */ - const GLfloat epsilon = 0.0001; + const GLfloat epsilon = (const GLfloat)0.0001; GLint x = IFLOOR(ctx->Current.RasterPos[0] + epsilon - xorig); GLint y = IFLOOR(ctx->Current.RasterPos[1] + epsilon - yorig); diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c index c9b30d3252..894d99afd2 100644 --- a/src/mesa/main/framebuffer.c +++ b/src/mesa/main/framebuffer.c @@ -67,7 +67,7 @@ compute_depth_max(struct gl_framebuffer *fb) fb->_DepthMaxF = (GLfloat) fb->_DepthMax; /* Minimum resolvable depth value, for polygon offset */ - fb->_MRD = 1.0 / fb->_DepthMaxF; + fb->_MRD = (GLfloat)1.0 / fb->_DepthMaxF; } diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c index eb81ee4a52..ee48b78318 100644 --- a/src/mesa/main/get.c +++ b/src/mesa/main/get.c @@ -2124,7 +2124,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) params[0] = (GLfloat)(ctx->DrawBuffer->Visual.depthBits); break; case GL_DEPTH_CLEAR_VALUE: - params[0] = ctx->Depth.Clear; + params[0] = (GLfloat)ctx->Depth.Clear; break; case GL_DEPTH_FUNC: params[0] = ENUM_TO_FLOAT(ctx->Depth.Func); @@ -2914,7 +2914,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) GLuint i, n = _mesa_get_compressed_formats(ctx, formats, GL_FALSE); ASSERT(n <= 100); for (i = 0; i < n; i++) - params[i] = ENUM_TO_INT(formats[i]); + params[i] = (GLfloat)(ENUM_TO_INT(formats[i])); } break; case GL_ARRAY_ELEMENT_LOCK_FIRST_EXT: diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c index 76e105e65e..285c8346a5 100644 --- a/src/mesa/main/image.c +++ b/src/mesa/main/image.c @@ -1169,7 +1169,7 @@ _mesa_apply_stencil_transfer_ops(const GLcontext *ctx, GLuint n, GLuint mask = ctx->PixelMaps.StoS.Size - 1; GLuint i; for (i = 0; i < n; i++) { - stencil[i] = ctx->PixelMaps.StoS.Map[ stencil[i] & mask ]; + stencil[i] = (GLstencil)ctx->PixelMaps.StoS.Map[ stencil[i] & mask ]; } } } @@ -3680,7 +3680,7 @@ _mesa_unpack_stencil_span( const GLcontext *ctx, GLuint n, const GLuint mask = ctx->PixelMaps.StoS.Size - 1; GLuint i; for (i = 0; i < n; i++) { - indexes[i] = ctx->PixelMaps.StoS.Map[ indexes[i] & mask ]; + indexes[i] = (GLuint)ctx->PixelMaps.StoS.Map[ indexes[i] & mask ]; } } @@ -4035,7 +4035,7 @@ _mesa_unpack_depth_span( const GLcontext *ctx, GLuint n, if (needClamp) { GLuint i; for (i = 0; i < n; i++) { - depthValues[i] = CLAMP(depthValues[i], 0.0, 1.0); + depthValues[i] = (GLfloat)CLAMP(depthValues[i], 0.0, 1.0); } } diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c index f9715b4865..f37d1f216f 100644 --- a/src/mesa/main/light.c +++ b/src/mesa/main/light.c @@ -1119,7 +1119,7 @@ compute_light_positions( GLcontext *ctx ) } else { /* positional light w/ homogeneous coordinate, divide by W */ - GLfloat wInv = 1.0 / light->_Position[3]; + GLfloat wInv = (GLfloat)1.0 / light->_Position[3]; light->_Position[0] *= wInv; light->_Position[1] *= wInv; light->_Position[2] *= wInv; diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c index d3d1958951..061378f3b7 100644 --- a/src/mesa/main/mipmap.c +++ b/src/mesa/main/mipmap.c @@ -292,7 +292,7 @@ do_row(GLenum datatype, GLuint comps, GLint srcWidth, GLfloat *dst = (GLfloat *) dstRow; for (i = j = 0, k = k0; i < (GLuint) dstWidth; i++, j += colStride, k += colStride) { - dst[i] = rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4; + dst[i] = (GLfloat)(rowA[j] / 4 + rowA[k] / 4 + rowB[j] / 4 + rowB[k] / 4); } } diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c index 0e9915dd38..7eeae05dbd 100644 --- a/src/mesa/main/pixel.c +++ b/src/mesa/main/pixel.c @@ -304,7 +304,7 @@ store_pixelmap(GLcontext *ctx, GLenum map, GLsizei mapsize, /* special case */ ctx->PixelMaps.StoS.Size = mapsize; for (i = 0; i < mapsize; i++) { - ctx->PixelMaps.StoS.Map[i] = IROUND(values[i]); + ctx->PixelMaps.StoS.Map[i] = (GLfloat)IROUND(values[i]); } break; case GL_PIXEL_MAP_I_TO_I: @@ -1142,7 +1142,7 @@ _mesa_lookup_rgba_ubyte(const struct gl_color_table *table, GLuint n, GLubyte rgba[][4]) { const GLubyte *lut = table->TableUB; - const GLfloat scale = (GLfloat) (table->Size - 1) / 255.0; + const GLfloat scale = (GLfloat) (table->Size - 1) / (GLfloat)255.0; GLuint i; if (!table->TableUB || table->Size == 0) diff --git a/src/mesa/main/queryobj.c b/src/mesa/main/queryobj.c index e30f5480da..a1e32e70ba 100644 --- a/src/mesa/main/queryobj.c +++ b/src/mesa/main/queryobj.c @@ -382,7 +382,7 @@ _mesa_GetQueryObjectivARB(GLuint id, GLenum pname, GLint *params) *params = 0x7fffffff; } else { - *params = q->Result; + *params = (GLint)q->Result; } break; case GL_QUERY_RESULT_AVAILABLE_ARB: @@ -422,7 +422,7 @@ _mesa_GetQueryObjectuivARB(GLuint id, GLenum pname, GLuint *params) *params = 0xffffffff; } else { - *params = q->Result; + *params = (GLuint)q->Result; } break; case GL_QUERY_RESULT_AVAILABLE_ARB: diff --git a/src/mesa/main/rastpos.c b/src/mesa/main/rastpos.c index ee163e0c71..f0500083ec 100644 --- a/src/mesa/main/rastpos.c +++ b/src/mesa/main/rastpos.c @@ -63,7 +63,7 @@ rasterpos(GLfloat x, GLfloat y, GLfloat z, GLfloat w) void GLAPIENTRY _mesa_RasterPos2d(GLdouble x, GLdouble y) { - rasterpos(x, y, 0.0F, 1.0F); + rasterpos((GLfloat)x, (GLfloat)y, (GLfloat)0.0, (GLfloat)1.0); } void GLAPIENTRY diff --git a/src/mesa/main/texenvprogram.c b/src/mesa/main/texenvprogram.c index 644b1f39c7..6877ef96f2 100644 --- a/src/mesa/main/texenvprogram.c +++ b/src/mesa/main/texenvprogram.c @@ -904,14 +904,14 @@ emit_texenv(struct texenv_fragment_program *p, GLuint unit) */ if (alpha_shift || rgb_shift) { if (rgb_shift == alpha_shift) { - shift = register_scalar_const(p, 1<File == PROGRAM_CONSTANT || source->File == PROGRAM_STATE_VAR); params = machine->CurProgram->Parameters; - if (reg < 0 || reg >= params->NumParameters) + if (reg < 0 || reg >= (GLint)params->NumParameters) return ZeroVec; else return params->ParameterValues[reg]; @@ -227,7 +227,7 @@ fetch_vector4_deriv(GLcontext * ctx, const struct gl_program_machine *machine, char xOrY, GLfloat result[4]) { - if (source->File == PROGRAM_INPUT && source->Index < machine->NumDeriv) { + if (source->File == PROGRAM_INPUT && source->Index < (GLint)machine->NumDeriv) { const GLint col = machine->CurElement; const GLfloat w = machine->Attribs[FRAG_ATTRIB_WPOS][col][3]; const GLfloat invQ = 1.0f / w; @@ -506,7 +506,7 @@ _mesa_execute_program(GLcontext * ctx, { const GLuint numInst = program->NumInstructions; const GLuint maxExec = 10000; - GLint pc, numExec = 0; + GLuint pc, numExec = 0; machine->CurProgram = program; diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 44fbfdcd04..8f48155825 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -250,7 +250,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], value[1] = ctx->Fog.Start; value[2] = ctx->Fog.End; value[3] = (ctx->Fog.End == ctx->Fog.Start) - ? 1.0 : 1.0F / (ctx->Fog.End - ctx->Fog.Start); + ? 1.0 : (GLfloat)(1.0 / (ctx->Fog.End - ctx->Fog.Start)); return; case STATE_CLIPPLANE: { @@ -411,7 +411,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], if (texObj) { struct gl_texture_image *texImage = texObj->Image[0][0]; ASSIGN_4V(value, 1.0 / texImage->Width, - 1.0 / texImage->Height, + (GLfloat)(1.0 / texImage->Height), 0.0, 1.0); } } @@ -426,10 +426,10 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], * exp2: 2^-((density/(ln(2)^2) * fogcoord)^2) */ value[0] = (ctx->Fog.End == ctx->Fog.Start) - ? 1.0 : -1.0F / (ctx->Fog.End - ctx->Fog.Start); + ? 1.0 : (GLfloat)(-1.0F / (ctx->Fog.End - ctx->Fog.Start)); value[1] = ctx->Fog.End * -value[0]; - value[2] = ctx->Fog.Density * ONE_DIV_LN2; - value[3] = ctx->Fog.Density * ONE_DIV_SQRT_LN2; + value[2] = (GLfloat)(ctx->Fog.Density * ONE_DIV_LN2); + value[3] = (GLfloat)(ctx->Fog.Density * ONE_DIV_SQRT_LN2); return; case STATE_LIGHT_SPOT_DIR_NORMALIZED: { diff --git a/src/mesa/shader/prog_uniform.c b/src/mesa/shader/prog_uniform.c index 20e004b350..d96a916533 100644 --- a/src/mesa/shader/prog_uniform.c +++ b/src/mesa/shader/prog_uniform.c @@ -135,7 +135,7 @@ _mesa_longest_uniform_name(const struct gl_uniform_list *list) GLuint i; for (i = 0; i < list->NumUniforms; i++) { GLuint len = _mesa_strlen(list->Uniforms[i].Name); - if (len > max) + if (len > (GLuint)max) max = len; } return max; diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 24ab7568d6..856179e1d5 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -719,7 +719,7 @@ _mesa_get_attached_shaders(GLcontext *ctx, GLuint program, GLsizei maxCount, struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, program); if (shProg) { - GLint i; + GLuint i; for (i = 0; i < maxCount && i < shProg->NumShaders; i++) { obj[i] = shProg->Shaders[i]->Name; } @@ -893,7 +893,7 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, = _mesa_lookup_shader_program(ctx, program); if (shProg) { if (location < shProg->Uniforms->NumUniforms) { - GLint progPos, i; + GLuint progPos, i; const struct gl_program *prog = NULL; progPos = shProg->Uniforms->Uniforms[location].VertPos; @@ -1111,7 +1111,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, } else { /* ordinary uniform variable */ - GLint k, i; + GLuint k, i; if (count * elems > program->Parameters->Parameters[location].Size) { _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(count too large)"); diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index e500ac8684..e4de875e8c 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -77,7 +77,7 @@ is_identity(const GLfloat m[16]) GLuint i; for (i = 0; i < 16; i++) { const int row = i % 4, col = i / 4; - const float val = (row == col); + const float val = (GLfloat)(row == col); if (m[i] != val) return GL_FALSE; } diff --git a/src/mesa/state_tracker/st_atom_scissor.c b/src/mesa/state_tracker/st_atom_scissor.c index f5db492403..3fd59e1945 100644 --- a/src/mesa/state_tracker/st_atom_scissor.c +++ b/src/mesa/state_tracker/st_atom_scissor.c @@ -52,14 +52,14 @@ update_scissor( struct st_context *st ) scissor.maxy = fb->Height; if (st->ctx->Scissor.Enabled) { - if (st->ctx->Scissor.X > scissor.minx) + if ((GLuint)st->ctx->Scissor.X > scissor.minx) scissor.minx = st->ctx->Scissor.X; - if (st->ctx->Scissor.Y > scissor.miny) + if ((GLuint)st->ctx->Scissor.Y > scissor.miny) scissor.miny = st->ctx->Scissor.Y; - if (st->ctx->Scissor.X + st->ctx->Scissor.Width < scissor.maxx) + if ((GLuint)st->ctx->Scissor.X + st->ctx->Scissor.Width < scissor.maxx) scissor.maxx = st->ctx->Scissor.X + st->ctx->Scissor.Width; - if (st->ctx->Scissor.Y + st->ctx->Scissor.Height < scissor.maxy) + if ((GLuint)st->ctx->Scissor.Y + st->ctx->Scissor.Height < scissor.maxy) scissor.maxy = st->ctx->Scissor.Y + st->ctx->Scissor.Height; /* check for null space */ diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index 4b51521470..b105909e96 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -49,7 +49,7 @@ update_viewport( struct st_context *st ) */ if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { yScale = -1; - yBias = ctx->DrawBuffer->Height ; + yBias = (GLfloat)ctx->DrawBuffer->Height; } else { yScale = 1.0; @@ -59,12 +59,12 @@ update_viewport( struct st_context *st ) /* _NEW_VIEWPORT */ { - GLfloat x = ctx->Viewport.X; - GLfloat y = ctx->Viewport.Y; + GLfloat x = (GLfloat)ctx->Viewport.X; + GLfloat y = (GLfloat)ctx->Viewport.Y; GLfloat z = ctx->Viewport.Near; - GLfloat half_width = ctx->Viewport.Width / 2.0; - GLfloat half_height = ctx->Viewport.Height / 2.0; - GLfloat half_depth = (ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; + GLfloat half_width = (GLfloat)ctx->Viewport.Width / 2.0; + GLfloat half_height = (GLfloat)ctx->Viewport.Height / 2.0; + GLfloat half_depth = (GLfloat)(ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; st->state.viewport.scale[0] = half_width; st->state.viewport.scale[1] = half_height * yScale; diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 593938f8cf..9763eebe54 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -360,18 +360,18 @@ setup_bitmap_vertex_data(struct st_context *st, { struct pipe_context *pipe = st->pipe; const struct gl_framebuffer *fb = st->ctx->DrawBuffer; - const GLfloat fb_width = fb->Width; - const GLfloat fb_height = fb->Height; - const GLfloat x0 = x; - const GLfloat x1 = x + width; - const GLfloat y0 = y; - const GLfloat y1 = y + height; - const GLfloat sLeft = 0.0F, sRight = 1.0F; - const GLfloat tTop = 0.0, tBot = 1.0 - tTop; - const GLfloat clip_x0 = x0 / fb_width * 2.0 - 1.0; - const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; - const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; - const GLfloat clip_y1 = y1 / fb_height * 2.0 - 1.0; + const GLfloat fb_width = (GLfloat)fb->Width; + const GLfloat fb_height = (GLfloat)fb->Height; + const GLfloat x0 = (GLfloat)x; + const GLfloat x1 = (GLfloat)(x + width); + const GLfloat y0 = (GLfloat)y; + const GLfloat y1 = (GLfloat)(y + height); + const GLfloat sLeft = (GLfloat)0.0, sRight = (GLfloat)1.0; + const GLfloat tTop = (GLfloat)0.0, tBot = (GLfloat)1.0 - tTop; + const GLfloat clip_x0 = (GLfloat)(x0 / fb_width * 2.0 - 1.0); + const GLfloat clip_y0 = (GLfloat)(y0 / fb_height * 2.0 - 1.0); + const GLfloat clip_x1 = (GLfloat)(x1 / fb_width * 2.0 - 1.0); + const GLfloat clip_y1 = (GLfloat)(y1 / fb_height * 2.0 - 1.0); GLuint i; void *buf; @@ -444,8 +444,8 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, * it up into chunks. */ maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1); - assert(width <= maxSize); - assert(height <= maxSize); + assert(width <= (GLsizei)maxSize); + assert(height <= (GLsizei)maxSize); cso_save_rasterizer(cso); cso_save_samplers(cso); @@ -488,15 +488,15 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, { const struct gl_framebuffer *fb = st->ctx->DrawBuffer; const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP); - const float width = fb->Width; - const float height = fb->Height; + const GLfloat width = (GLfloat)fb->Width; + const GLfloat height = (GLfloat)fb->Height; struct pipe_viewport_state vp; vp.scale[0] = 0.5 * width; - vp.scale[1] = height * (invert ? -0.5 : 0.5); + vp.scale[1] = (GLfloat)(height * (invert ? -0.5 : 0.5)); vp.scale[2] = 1.0; vp.scale[3] = 1.0; - vp.translate[0] = 0.5 * width; - vp.translate[1] = 0.5 * height; + vp.translate[0] = (GLfloat)(0.5 * width); + vp.translate[1] = (GLfloat)(0.5 * height); vp.translate[2] = 0.0; vp.translate[3] = 0.0; cso_set_viewport(cso, &vp); diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c index a52521db64..dbee2188ee 100644 --- a/src/mesa/vbo/vbo_exec_array.c +++ b/src/mesa/vbo/vbo_exec_array.c @@ -41,7 +41,7 @@ static void get_minmax_index( GLuint count, GLuint type, GLuint *min_index, GLuint *max_index) { - GLint i; + GLuint i; switch(type) { case GL_UNSIGNED_INT: { diff --git a/src/mesa/vbo/vbo_split_inplace.c b/src/mesa/vbo/vbo_split_inplace.c index 958afccd0c..fbc856e93b 100644 --- a/src/mesa/vbo/vbo_split_inplace.c +++ b/src/mesa/vbo/vbo_split_inplace.c @@ -58,7 +58,7 @@ struct split_context { static void flush_vertex( struct split_context *split ) { - GLint min_index, max_index; + GLuint min_index, max_index; if (!split->dstprim_nr) return; -- cgit v1.2.3 From 3b77f391db7827d1fb19a5dc3d8e8d3d705185ae Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 31 May 2008 18:51:44 +0900 Subject: draw: Eliminate stdio usage. --- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 37cbe9bf9e..243d3eee23 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -185,7 +185,7 @@ draw_create_vs_sse(struct draw_context *draw, return &vs->base; fail: - fprintf(stderr, "tgsi_emit_sse2() failed, falling back to interpreter\n"); + debug_error("tgsi_emit_sse2() failed, falling back to interpreter\n"); x86_release_func( &vs->sse2_program ); -- cgit v1.2.3 From ffc6afcdbf14b5d26bd3665a269fdd28d944afed Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 31 May 2008 18:55:25 +0900 Subject: scons: Enable mesa state tracker in all platforms by default. But only actually build it on some. This results in a better default scons experience. --- src/mesa/SConscript | 604 ++++++++++++++++++++++++++-------------------------- 1 file changed, 303 insertions(+), 301 deletions(-) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index 6b99dacf2a..04b771a7e1 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -4,307 +4,309 @@ Import('*') -env = env.Clone() +if env['platform'] != 'winddk': -# Includes -env.Append(CPPPATH = [ - '#/src/mesa', - '#/src/mesa/main', -]) - -if gcc: - env.Append(CFLAGS = [ - '-std=c99', - ]) - -# x86 assembly -if x86 and gcc: - env.Append(CPPDEFINES = [ - 'USE_X86_ASM', - 'USE_MMX_ASM', - 'USE_3DNOW_ASM', - 'USE_SSE_ASM', + env = env.Clone() + + # Includes + env.Append(CPPPATH = [ + '#/src/mesa', + '#/src/mesa/main', ]) - - -####################################################################### -# Core sources - -MAIN_SOURCES = [ - 'main/api_arrayelt.c', - 'main/api_loopback.c', - 'main/api_noop.c', - 'main/api_validate.c', - 'main/accum.c', - 'main/attrib.c', - 'main/arrayobj.c', - 'main/blend.c', - 'main/bufferobj.c', - 'main/buffers.c', - 'main/clip.c', - 'main/colortab.c', - 'main/context.c', - 'main/convolve.c', - 'main/debug.c', - 'main/depth.c', - 'main/depthstencil.c', - 'main/dlist.c', - 'main/drawpix.c', - 'main/enable.c', - 'main/enums.c', - 'main/eval.c', - 'main/execmem.c', - 'main/extensions.c', - 'main/fbobject.c', - 'main/feedback.c', - 'main/ffvertex_prog.c', - 'main/fog.c', - 'main/framebuffer.c', - 'main/get.c', - 'main/getstring.c', - 'main/hash.c', - 'main/hint.c', - 'main/histogram.c', - 'main/image.c', - 'main/imports.c', - 'main/light.c', - 'main/lines.c', - 'main/matrix.c', - 'main/mipmap.c', - 'main/mm.c', - 'main/pixel.c', - 'main/points.c', - 'main/polygon.c', - 'main/queryobj.c', - 'main/rastpos.c', - 'main/rbadaptors.c', - 'main/renderbuffer.c', - 'main/shaders.c', - 'main/state.c', - 'main/stencil.c', - 'main/texcompress.c', - 'main/texcompress_s3tc.c', - 'main/texcompress_fxt1.c', - 'main/texenvprogram.c', - 'main/texformat.c', - 'main/teximage.c', - 'main/texobj.c', - 'main/texrender.c', - 'main/texstate.c', - 'main/texstore.c', - 'main/varray.c', - 'main/vtxfmt.c', -] - -GLAPI_SOURCES = [ - 'main/dispatch.c', - 'glapi/glapi.c', - 'glapi/glthread.c', -] - -MATH_SOURCES = [ - 'math/m_debug_clip.c', - 'math/m_debug_norm.c', - 'math/m_debug_xform.c', - 'math/m_eval.c', - 'math/m_matrix.c', - 'math/m_translate.c', - 'math/m_vector.c', - 'math/m_xform.c', -] - -VBO_SOURCES = [ - 'vbo/vbo_context.c', - 'vbo/vbo_exec.c', - 'vbo/vbo_exec_api.c', - 'vbo/vbo_exec_array.c', - 'vbo/vbo_exec_draw.c', - 'vbo/vbo_exec_eval.c', - 'vbo/vbo_rebase.c', - 'vbo/vbo_split.c', - 'vbo/vbo_split_copy.c', - 'vbo/vbo_split_inplace.c', - 'vbo/vbo_save.c', - 'vbo/vbo_save_api.c', - 'vbo/vbo_save_draw.c', - 'vbo/vbo_save_loopback.c', -] - -VF_SOURCES = [ - 'vf/vf.c', - 'vf/vf_generic.c', - 'vf/vf_sse.c', -] - -STATETRACKER_SOURCES = [ - 'state_tracker/st_atom.c', - 'state_tracker/st_atom_blend.c', - 'state_tracker/st_atom_clip.c', - 'state_tracker/st_atom_constbuf.c', - 'state_tracker/st_atom_depth.c', - 'state_tracker/st_atom_framebuffer.c', - 'state_tracker/st_atom_pixeltransfer.c', - 'state_tracker/st_atom_sampler.c', - 'state_tracker/st_atom_scissor.c', - 'state_tracker/st_atom_shader.c', - 'state_tracker/st_atom_rasterizer.c', - 'state_tracker/st_atom_stipple.c', - 'state_tracker/st_atom_texture.c', - 'state_tracker/st_atom_viewport.c', - 'state_tracker/st_cb_accum.c', - 'state_tracker/st_cb_bitmap.c', - 'state_tracker/st_cb_blit.c', - 'state_tracker/st_cb_bufferobjects.c', - 'state_tracker/st_cb_clear.c', - 'state_tracker/st_cb_flush.c', - 'state_tracker/st_cb_drawpixels.c', - 'state_tracker/st_cb_fbo.c', - 'state_tracker/st_cb_feedback.c', - 'state_tracker/st_cb_program.c', - 'state_tracker/st_cb_queryobj.c', - 'state_tracker/st_cb_rasterpos.c', - 'state_tracker/st_cb_readpixels.c', - 'state_tracker/st_cb_strings.c', - 'state_tracker/st_cb_texture.c', - 'state_tracker/st_context.c', - 'state_tracker/st_debug.c', - 'state_tracker/st_draw.c', - 'state_tracker/st_extensions.c', - 'state_tracker/st_format.c', - 'state_tracker/st_framebuffer.c', - 'state_tracker/st_gen_mipmap.c', - 'state_tracker/st_mesa_to_tgsi.c', - 'state_tracker/st_program.c', - 'state_tracker/st_texture.c', -] - -SHADER_SOURCES = [ - 'shader/arbprogparse.c', - 'shader/arbprogram.c', - 'shader/atifragshader.c', - 'shader/grammar/grammar_mesa.c', - 'shader/nvfragparse.c', - 'shader/nvprogram.c', - 'shader/nvvertparse.c', - 'shader/program.c', - 'shader/prog_cache.c', - 'shader/prog_debug.c', - 'shader/prog_execute.c', - 'shader/prog_instruction.c', - 'shader/prog_parameter.c', - 'shader/prog_print.c', - 'shader/prog_statevars.c', - 'shader/prog_uniform.c', - 'shader/programopt.c', - 'shader/shader_api.c', -] - -SLANG_SOURCES = [ - 'shader/slang/slang_builtin.c', - 'shader/slang/slang_codegen.c', - 'shader/slang/slang_compile.c', - 'shader/slang/slang_compile_function.c', - 'shader/slang/slang_compile_operation.c', - 'shader/slang/slang_compile_struct.c', - 'shader/slang/slang_compile_variable.c', - 'shader/slang/slang_emit.c', - 'shader/slang/slang_ir.c', - 'shader/slang/slang_label.c', - 'shader/slang/slang_library_noise.c', - 'shader/slang/slang_link.c', - 'shader/slang/slang_log.c', - 'shader/slang/slang_mem.c', - 'shader/slang/slang_preprocess.c', - 'shader/slang/slang_print.c', - 'shader/slang/slang_simplify.c', - 'shader/slang/slang_storage.c', - 'shader/slang/slang_typeinfo.c', - 'shader/slang/slang_vartable.c', - 'shader/slang/slang_utility.c', -] - - -####################################################################### -# Assembly sources - -ASM_C_SOURCES = [ - 'x86/common_x86.c', - 'x86/x86.c', - 'x86/3dnow.c', - 'x86/sse.c', - 'sparc/sparc.c', - 'ppc/common_ppc.c', - 'x86-64/x86-64.c', -] - -X86_SOURCES = [ - 'x86/common_x86_asm.S', - 'x86/x86_xform2.S', - 'x86/x86_xform3.S', - 'x86/x86_xform4.S', - 'x86/x86_cliptest.S', - 'x86/mmx_blend.S', - 'x86/3dnow_xform1.S', - 'x86/3dnow_xform2.S', - 'x86/3dnow_xform3.S', - 'x86/3dnow_xform4.S', - 'x86/3dnow_normal.S', - 'x86/sse_xform1.S', - 'x86/sse_xform2.S', - 'x86/sse_xform3.S', - 'x86/sse_xform4.S', - 'x86/sse_normal.S', - 'x86/read_rgba_span_x86.S', -] - -X86_API = [ - 'x86/glapi_x86.S', -] - -X86_64_SOURCES = [ - 'x86-64/xform4.S', -] - -X86_64_API = [ - 'x86-64/glapi_x86-64.S', -] - -SPARC_SOURCES = [ - 'sparc/clip.S', - 'sparc/norm.S', - 'sparc/xform.S', -] - -SPARC_API = [ - 'sparc/glapi_sparc.S', -] - -if x86 and gcc: - ASM_SOURCES = ASM_C_SOURCES + X86_SOURCES - API_SOURCES = X86_API -else: - ASM_SOURCES = [] - API_SOURCES = [] - -SOLO_SOURCES = \ - MAIN_SOURCES + \ - MATH_SOURCES + \ - VBO_SOURCES + \ - VF_SOURCES + \ - STATETRACKER_SOURCES + \ - SHADER_SOURCES + \ - ASM_SOURCES + \ - SLANG_SOURCES - -mesa = env.ConvenienceLibrary( - target = 'mesa', - source = SOLO_SOURCES, -) -Export('mesa') - -if not dri: - glapi = env.ConvenienceLibrary( - target = 'glapi', - source = GLAPI_SOURCES + API_SOURCES, + + if gcc: + env.Append(CFLAGS = [ + '-std=c99', + ]) + + # x86 assembly + if x86 and gcc: + env.Append(CPPDEFINES = [ + 'USE_X86_ASM', + 'USE_MMX_ASM', + 'USE_3DNOW_ASM', + 'USE_SSE_ASM', + ]) + + + ####################################################################### + # Core sources + + MAIN_SOURCES = [ + 'main/api_arrayelt.c', + 'main/api_loopback.c', + 'main/api_noop.c', + 'main/api_validate.c', + 'main/accum.c', + 'main/attrib.c', + 'main/arrayobj.c', + 'main/blend.c', + 'main/bufferobj.c', + 'main/buffers.c', + 'main/clip.c', + 'main/colortab.c', + 'main/context.c', + 'main/convolve.c', + 'main/debug.c', + 'main/depth.c', + 'main/depthstencil.c', + 'main/dlist.c', + 'main/drawpix.c', + 'main/enable.c', + 'main/enums.c', + 'main/eval.c', + 'main/execmem.c', + 'main/extensions.c', + 'main/fbobject.c', + 'main/feedback.c', + 'main/ffvertex_prog.c', + 'main/fog.c', + 'main/framebuffer.c', + 'main/get.c', + 'main/getstring.c', + 'main/hash.c', + 'main/hint.c', + 'main/histogram.c', + 'main/image.c', + 'main/imports.c', + 'main/light.c', + 'main/lines.c', + 'main/matrix.c', + 'main/mipmap.c', + 'main/mm.c', + 'main/pixel.c', + 'main/points.c', + 'main/polygon.c', + 'main/queryobj.c', + 'main/rastpos.c', + 'main/rbadaptors.c', + 'main/renderbuffer.c', + 'main/shaders.c', + 'main/state.c', + 'main/stencil.c', + 'main/texcompress.c', + 'main/texcompress_s3tc.c', + 'main/texcompress_fxt1.c', + 'main/texenvprogram.c', + 'main/texformat.c', + 'main/teximage.c', + 'main/texobj.c', + 'main/texrender.c', + 'main/texstate.c', + 'main/texstore.c', + 'main/varray.c', + 'main/vtxfmt.c', + ] + + GLAPI_SOURCES = [ + 'main/dispatch.c', + 'glapi/glapi.c', + 'glapi/glthread.c', + ] + + MATH_SOURCES = [ + 'math/m_debug_clip.c', + 'math/m_debug_norm.c', + 'math/m_debug_xform.c', + 'math/m_eval.c', + 'math/m_matrix.c', + 'math/m_translate.c', + 'math/m_vector.c', + 'math/m_xform.c', + ] + + VBO_SOURCES = [ + 'vbo/vbo_context.c', + 'vbo/vbo_exec.c', + 'vbo/vbo_exec_api.c', + 'vbo/vbo_exec_array.c', + 'vbo/vbo_exec_draw.c', + 'vbo/vbo_exec_eval.c', + 'vbo/vbo_rebase.c', + 'vbo/vbo_split.c', + 'vbo/vbo_split_copy.c', + 'vbo/vbo_split_inplace.c', + 'vbo/vbo_save.c', + 'vbo/vbo_save_api.c', + 'vbo/vbo_save_draw.c', + 'vbo/vbo_save_loopback.c', + ] + + VF_SOURCES = [ + 'vf/vf.c', + 'vf/vf_generic.c', + 'vf/vf_sse.c', + ] + + STATETRACKER_SOURCES = [ + 'state_tracker/st_atom.c', + 'state_tracker/st_atom_blend.c', + 'state_tracker/st_atom_clip.c', + 'state_tracker/st_atom_constbuf.c', + 'state_tracker/st_atom_depth.c', + 'state_tracker/st_atom_framebuffer.c', + 'state_tracker/st_atom_pixeltransfer.c', + 'state_tracker/st_atom_sampler.c', + 'state_tracker/st_atom_scissor.c', + 'state_tracker/st_atom_shader.c', + 'state_tracker/st_atom_rasterizer.c', + 'state_tracker/st_atom_stipple.c', + 'state_tracker/st_atom_texture.c', + 'state_tracker/st_atom_viewport.c', + 'state_tracker/st_cb_accum.c', + 'state_tracker/st_cb_bitmap.c', + 'state_tracker/st_cb_blit.c', + 'state_tracker/st_cb_bufferobjects.c', + 'state_tracker/st_cb_clear.c', + 'state_tracker/st_cb_flush.c', + 'state_tracker/st_cb_drawpixels.c', + 'state_tracker/st_cb_fbo.c', + 'state_tracker/st_cb_feedback.c', + 'state_tracker/st_cb_program.c', + 'state_tracker/st_cb_queryobj.c', + 'state_tracker/st_cb_rasterpos.c', + 'state_tracker/st_cb_readpixels.c', + 'state_tracker/st_cb_strings.c', + 'state_tracker/st_cb_texture.c', + 'state_tracker/st_context.c', + 'state_tracker/st_debug.c', + 'state_tracker/st_draw.c', + 'state_tracker/st_extensions.c', + 'state_tracker/st_format.c', + 'state_tracker/st_framebuffer.c', + 'state_tracker/st_gen_mipmap.c', + 'state_tracker/st_mesa_to_tgsi.c', + 'state_tracker/st_program.c', + 'state_tracker/st_texture.c', + ] + + SHADER_SOURCES = [ + 'shader/arbprogparse.c', + 'shader/arbprogram.c', + 'shader/atifragshader.c', + 'shader/grammar/grammar_mesa.c', + 'shader/nvfragparse.c', + 'shader/nvprogram.c', + 'shader/nvvertparse.c', + 'shader/program.c', + 'shader/prog_cache.c', + 'shader/prog_debug.c', + 'shader/prog_execute.c', + 'shader/prog_instruction.c', + 'shader/prog_parameter.c', + 'shader/prog_print.c', + 'shader/prog_statevars.c', + 'shader/prog_uniform.c', + 'shader/programopt.c', + 'shader/shader_api.c', + ] + + SLANG_SOURCES = [ + 'shader/slang/slang_builtin.c', + 'shader/slang/slang_codegen.c', + 'shader/slang/slang_compile.c', + 'shader/slang/slang_compile_function.c', + 'shader/slang/slang_compile_operation.c', + 'shader/slang/slang_compile_struct.c', + 'shader/slang/slang_compile_variable.c', + 'shader/slang/slang_emit.c', + 'shader/slang/slang_ir.c', + 'shader/slang/slang_label.c', + 'shader/slang/slang_library_noise.c', + 'shader/slang/slang_link.c', + 'shader/slang/slang_log.c', + 'shader/slang/slang_mem.c', + 'shader/slang/slang_preprocess.c', + 'shader/slang/slang_print.c', + 'shader/slang/slang_simplify.c', + 'shader/slang/slang_storage.c', + 'shader/slang/slang_typeinfo.c', + 'shader/slang/slang_vartable.c', + 'shader/slang/slang_utility.c', + ] + + + ####################################################################### + # Assembly sources + + ASM_C_SOURCES = [ + 'x86/common_x86.c', + 'x86/x86.c', + 'x86/3dnow.c', + 'x86/sse.c', + 'sparc/sparc.c', + 'ppc/common_ppc.c', + 'x86-64/x86-64.c', + ] + + X86_SOURCES = [ + 'x86/common_x86_asm.S', + 'x86/x86_xform2.S', + 'x86/x86_xform3.S', + 'x86/x86_xform4.S', + 'x86/x86_cliptest.S', + 'x86/mmx_blend.S', + 'x86/3dnow_xform1.S', + 'x86/3dnow_xform2.S', + 'x86/3dnow_xform3.S', + 'x86/3dnow_xform4.S', + 'x86/3dnow_normal.S', + 'x86/sse_xform1.S', + 'x86/sse_xform2.S', + 'x86/sse_xform3.S', + 'x86/sse_xform4.S', + 'x86/sse_normal.S', + 'x86/read_rgba_span_x86.S', + ] + + X86_API = [ + 'x86/glapi_x86.S', + ] + + X86_64_SOURCES = [ + 'x86-64/xform4.S', + ] + + X86_64_API = [ + 'x86-64/glapi_x86-64.S', + ] + + SPARC_SOURCES = [ + 'sparc/clip.S', + 'sparc/norm.S', + 'sparc/xform.S', + ] + + SPARC_API = [ + 'sparc/glapi_sparc.S', + ] + + if x86 and gcc: + ASM_SOURCES = ASM_C_SOURCES + X86_SOURCES + API_SOURCES = X86_API + else: + ASM_SOURCES = [] + API_SOURCES = [] + + SOLO_SOURCES = \ + MAIN_SOURCES + \ + MATH_SOURCES + \ + VBO_SOURCES + \ + VF_SOURCES + \ + STATETRACKER_SOURCES + \ + SHADER_SOURCES + \ + ASM_SOURCES + \ + SLANG_SOURCES + + mesa = env.ConvenienceLibrary( + target = 'mesa', + source = SOLO_SOURCES, ) - Export('glapi') + Export('mesa') + + if not dri: + glapi = env.ConvenienceLibrary( + target = 'glapi', + source = GLAPI_SOURCES + API_SOURCES, + ) + Export('glapi') -- cgit v1.2.3 From 13581958bd99396ab8ec314f10cf61f717b18a9b Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 18:19:21 +0200 Subject: draw: Remove const qualifier. --- src/gallium/auxiliary/draw/draw_private.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 4f8cceee1e..423f64262b 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -193,7 +193,7 @@ struct draw_context const float (*aligned_constants)[4]; - const float (*aligned_constant_storage)[4]; + float (*aligned_constant_storage)[4]; unsigned const_storage_size; -- cgit v1.2.3 From 9046d1acfa91621ee83b3933fe6e4b52deb00cbf Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 18:40:00 +0200 Subject: gallium: Fix preprocessor logic. --- src/gallium/include/pipe/p_util.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index fe4ba3a689..9497eeeec9 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -410,7 +410,7 @@ extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, -#if !defined(_MSC_VER) && _MSC_VER < 0x800 +#if !defined(_MSC_VER) || _MSC_VER < 0x800 #if !defined(_INC_MATH) || !defined(__cplusplus) static INLINE float cosf( float f ) -- cgit v1.2.3 From a4abedc4f56e0442083aee9b39980900abf6ef40 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 18:40:39 +0200 Subject: draw: Remove const qualifier. --- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 243d3eee23..f638208bf5 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -130,7 +130,7 @@ vs_sse_delete( struct draw_vertex_shader *base ) x86_release_func( &shader->sse2_program ); - align_free(shader->base.immediates); + align_free( (void *) shader->base.immediates ); FREE( (void*) shader->base.state.tokens ); FREE( shader ); -- cgit v1.2.3 From 9b3c1582befd0c64d9b48f7ed39566f9ae2e24d9 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 18:42:17 +0200 Subject: i915: Add mising include. --- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index cdbb5feb23..f9297e29fc 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -38,6 +38,7 @@ */ +#include "draw/draw_context.h" #include "draw/draw_vbuf.h" #include "pipe/p_debug.h" #include "pipe/p_util.h" -- cgit v1.2.3 From 140e0e071890f24653e333e309778f6242c8d8f9 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 18:44:47 +0200 Subject: i915: Comment out dead & problematic code. --- src/gallium/drivers/i915simple/i915_texture.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 4c0c7ce3f3..f0d00280c2 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -104,6 +104,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex, */ } +#if 0 static unsigned power_of_two(unsigned x) { @@ -112,6 +113,7 @@ power_of_two(unsigned x) value = value << 1; return value; } +#endif static unsigned round_up(unsigned n, unsigned multiple) -- cgit v1.2.3 From 4e33edfd06999e7c65b761be2d1cdc16e5b659c5 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 18:56:20 +0200 Subject: gallium: Refactor TGSI decalaration tokens. * Incorporate declaration_interpolation into declaration itself. * Remove declaration_mask -- always use declaration_range. --- src/gallium/include/pipe/p_shader_tokens.h | 27 ++++++--------------------- 1 file changed, 6 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index f05e1a34b3..282d32910e 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -55,9 +55,6 @@ enum tgsi_file_type { }; -#define TGSI_DECLARE_RANGE 0 -#define TGSI_DECLARE_MASK 1 - #define TGSI_WRITEMASK_NONE 0x00 #define TGSI_WRITEMASK_X 0x01 #define TGSI_WRITEMASK_Y 0x02 @@ -75,16 +72,19 @@ enum tgsi_file_type { #define TGSI_WRITEMASK_YZW 0x0E #define TGSI_WRITEMASK_XYZW 0x0F +#define TGSI_INTERPOLATE_CONSTANT 0 +#define TGSI_INTERPOLATE_LINEAR 1 +#define TGSI_INTERPOLATE_PERSPECTIVE 2 + struct tgsi_declaration { unsigned Type : 4; /* TGSI_TOKEN_TYPE_DECLARATION */ unsigned Size : 8; /* UINT */ unsigned File : 4; /* one of TGSI_FILE_x */ - unsigned Declare : 4; /* one of TGSI_DECLARE_x */ unsigned UsageMask : 4; /* bitmask of TGSI_WRITEMASK_x flags */ - unsigned Interpolate : 1; /* BOOL, any interpolation info? */ + unsigned Interpolate : 4; /* TGSI_INTERPOLATE_ */ unsigned Semantic : 1; /* BOOL, any semantic info? */ - unsigned Padding : 5; + unsigned Padding : 6; unsigned Extended : 1; /* BOOL */ }; @@ -94,21 +94,6 @@ struct tgsi_declaration_range unsigned Last : 16; /* UINT */ }; -struct tgsi_declaration_mask -{ - unsigned Mask : 32; /* UINT */ -}; - -#define TGSI_INTERPOLATE_CONSTANT 0 -#define TGSI_INTERPOLATE_LINEAR 1 -#define TGSI_INTERPOLATE_PERSPECTIVE 2 - -struct tgsi_declaration_interpolation -{ - unsigned Interpolate : 4; /* TGSI_INTERPOLATE_ */ - unsigned Padding : 28; -}; - #define TGSI_SEMANTIC_POSITION 0 #define TGSI_SEMANTIC_COLOR 1 #define TGSI_SEMANTIC_BCOLOR 2 /**< back-face color */ -- cgit v1.2.3 From c2ff3a66a1d9fe0b5303ded0503323a73a6a7391 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 19:40:36 +0200 Subject: draw: Fix build after TGSI declaration interface changes. --- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 35 ++++++++++++------------- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 27 +++++++++---------- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 27 +++++++++---------- 3 files changed, 43 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index fd48b224b4..634bf067f1 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -141,18 +141,18 @@ aa_transform_decl(struct tgsi_transform_context *ctx, if (decl->Declaration.File == TGSI_FILE_OUTPUT && decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR && decl->Semantic.SemanticIndex == 0) { - aactx->colorOutput = decl->u.DeclarationRange.First; + aactx->colorOutput = decl->DeclarationRange.First; } else if (decl->Declaration.File == TGSI_FILE_SAMPLER) { uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { + for (i = decl->DeclarationRange.First; + i <= decl->DeclarationRange.Last; i++) { aactx->samplersUsed |= 1 << i; } } else if (decl->Declaration.File == TGSI_FILE_INPUT) { - if ((int) decl->u.DeclarationRange.Last > aactx->maxInput) - aactx->maxInput = decl->u.DeclarationRange.Last; + if ((int) decl->DeclarationRange.Last > aactx->maxInput) + aactx->maxInput = decl->DeclarationRange.Last; if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) { aactx->maxGeneric = decl->Semantic.SemanticIndex; @@ -160,8 +160,8 @@ aa_transform_decl(struct tgsi_transform_context *ctx, } else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { + for (i = decl->DeclarationRange.First; + i <= decl->DeclarationRange.Last; i++) { aactx->tempsUsed |= (1 << i); } } @@ -225,34 +225,33 @@ aa_transform_inst(struct tgsi_transform_context *ctx, /* declare new generic input/texcoord */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_INPUT; + /* XXX this could be linear... */ + decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; decl.Declaration.Semantic = 1; decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; decl.Semantic.SemanticIndex = aactx->maxGeneric + 1; - decl.Declaration.Interpolate = 1; - /* XXX this could be linear... */ - decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->maxInput + 1; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = aactx->maxInput + 1; ctx->emit_declaration(ctx, &decl); /* declare new sampler */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_SAMPLER; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->freeSampler; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = aactx->freeSampler; ctx->emit_declaration(ctx, &decl); /* declare new temp regs */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->texTemp; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = aactx->texTemp; ctx->emit_declaration(ctx, &decl); decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->colorTemp; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = aactx->colorTemp; ctx->emit_declaration(ctx, &decl); aactx->firstInstruction = FALSE; diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 97d74ad693..96dcdb43d5 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -131,11 +131,11 @@ aa_transform_decl(struct tgsi_transform_context *ctx, if (decl->Declaration.File == TGSI_FILE_OUTPUT && decl->Semantic.SemanticName == TGSI_SEMANTIC_COLOR && decl->Semantic.SemanticIndex == 0) { - aactx->colorOutput = decl->u.DeclarationRange.First; + aactx->colorOutput = decl->DeclarationRange.First; } else if (decl->Declaration.File == TGSI_FILE_INPUT) { - if ((int) decl->u.DeclarationRange.Last > aactx->maxInput) - aactx->maxInput = decl->u.DeclarationRange.Last; + if ((int) decl->DeclarationRange.Last > aactx->maxInput) + aactx->maxInput = decl->DeclarationRange.Last; if (decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC && (int) decl->Semantic.SemanticIndex > aactx->maxGeneric) { aactx->maxGeneric = decl->Semantic.SemanticIndex; @@ -143,8 +143,8 @@ aa_transform_decl(struct tgsi_transform_context *ctx, } else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { + for (i = decl->DeclarationRange.First; + i <= decl->DeclarationRange.Last; i++) { aactx->tempsUsed |= (1 << i); } } @@ -193,27 +193,26 @@ aa_transform_inst(struct tgsi_transform_context *ctx, /* declare new generic input/texcoord */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_INPUT; + /* XXX this could be linear... */ + decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; decl.Declaration.Semantic = 1; decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; decl.Semantic.SemanticIndex = aactx->maxGeneric + 1; - decl.Declaration.Interpolate = 1; - /* XXX this could be linear... */ - decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = texInput; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = texInput; ctx->emit_declaration(ctx, &decl); /* declare new temp regs */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = tmp0; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = tmp0; ctx->emit_declaration(ctx, &decl); decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = aactx->colorTemp; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = aactx->colorTemp; ctx->emit_declaration(ctx, &decl); aactx->firstInstruction = FALSE; diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 4c92416eb1..4087cf7a49 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -132,20 +132,20 @@ pstip_transform_decl(struct tgsi_transform_context *ctx, if (decl->Declaration.File == TGSI_FILE_SAMPLER) { uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { + for (i = decl->DeclarationRange.First; + i <= decl->DeclarationRange.Last; i++) { pctx->samplersUsed |= 1 << i; } } else if (decl->Declaration.File == TGSI_FILE_INPUT) { - pctx->maxInput = MAX2(pctx->maxInput, (int) decl->u.DeclarationRange.Last); + pctx->maxInput = MAX2(pctx->maxInput, (int) decl->DeclarationRange.Last); if (decl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) - pctx->wincoordInput = (int) decl->u.DeclarationRange.First; + pctx->wincoordInput = (int) decl->DeclarationRange.First; } else if (decl->Declaration.File == TGSI_FILE_TEMPORARY) { uint i; - for (i = decl->u.DeclarationRange.First; - i <= decl->u.DeclarationRange.Last; i++) { + for (i = decl->DeclarationRange.First; + i <= decl->DeclarationRange.Last; i++) { pctx->tempsUsed |= (1 << i); } } @@ -223,28 +223,27 @@ pstip_transform_inst(struct tgsi_transform_context *ctx, /* declare new position input reg */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_INPUT; + decl.Declaration.Interpolate = TGSI_INTERPOLATE_LINEAR; /* XXX? */ decl.Declaration.Semantic = 1; decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION; decl.Semantic.SemanticIndex = 0; - decl.Declaration.Interpolate = 1; - decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR; /* XXX? */ - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = wincoordInput; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = wincoordInput; ctx->emit_declaration(ctx, &decl); } /* declare new sampler */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_SAMPLER; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = pctx->freeSampler; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = pctx->freeSampler; ctx->emit_declaration(ctx, &decl); /* declare new temp regs */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = pctx->texTemp; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = pctx->texTemp; ctx->emit_declaration(ctx, &decl); /* emit immediate = {1/32, 1/32, 1, 1} -- cgit v1.2.3 From 3de18c2ac3cf679875d22d7ae9e62a11f5ea03c9 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 19:41:29 +0200 Subject: gallivm: Fix build after TGSI declaration interface changes. --- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 10 ++++------ 1 file changed, 4 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index 9695358ab8..98014bdaa1 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -96,10 +96,8 @@ translate_declaration(struct gallivm_ir *prog, unsigned first, last, mask; uint interp_method; - assert(decl->Declaration.Declare == TGSI_DECLARE_RANGE); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; + first = decl->DeclarationRange.First; + last = decl->DeclarationRange.Last; mask = decl->Declaration.UsageMask; /* Do not touch WPOS.xy */ @@ -113,7 +111,7 @@ translate_declaration(struct gallivm_ir *prog, } } - interp_method = decl->Interpolation.Interpolate; + interp_method = decl->Declaration.Interpolate; if (mask == TGSI_WRITEMASK_XYZW) { unsigned i, j; @@ -153,7 +151,7 @@ translate_declarationir(struct gallivm_ir *, struct tgsi_full_declaration *) { if (decl->Declaration.File == TGSI_FILE_ADDRESS) { - int idx = decl->u.DeclarationRange.First; + int idx = decl->DeclarationRange.First; storage->addAddress(idx); } } -- cgit v1.2.3 From a49381587f73c67469ec7546419cfc41387f938c Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 19:48:13 +0200 Subject: tgsi: Fix build after TGSI declaration interface changes. --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 8 +- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 8 +- src/gallium/auxiliary/tgsi/util/tgsi_build.c | 126 ++++++--------------------- src/gallium/auxiliary/tgsi/util/tgsi_build.h | 19 +--- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 58 +++--------- src/gallium/auxiliary/tgsi/util/tgsi_parse.c | 17 +--- src/gallium/auxiliary/tgsi/util/tgsi_parse.h | 9 +- src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 4 +- 8 files changed, 52 insertions(+), 197 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 826b432f09..6ba0949183 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -1411,13 +1411,11 @@ exec_declaration( unsigned first, last, mask; eval_coef_func eval; - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; + first = decl->DeclarationRange.First; + last = decl->DeclarationRange.Last; mask = decl->Declaration.UsageMask; - switch( decl->Interpolation.Interpolate ) { + switch( decl->Declaration.Interpolate ) { case TGSI_INTERPOLATE_CONSTANT: eval = eval_constant_coef; break; diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index 189dc6024d..a59e22b6ea 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -1923,16 +1923,14 @@ emit_declaration( unsigned first, last, mask; unsigned i, j; - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; + first = decl->DeclarationRange.First; + last = decl->DeclarationRange.Last; mask = decl->Declaration.UsageMask; for( i = first; i <= last; i++ ) { for( j = 0; j < NUM_CHANNELS; j++ ) { if( mask & (1 << j) ) { - switch( decl->Interpolation.Interpolate ) { + switch( decl->Declaration.Interpolate ) { case TGSI_INTERPOLATE_CONSTANT: emit_coef_a0( func, 0, i, j ); emit_inputs( func, 0, i, j ); diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.c b/src/gallium/auxiliary/tgsi/util/tgsi_build.c index 9c883ab704..63cc27becc 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.c @@ -90,9 +90,8 @@ tgsi_default_declaration( void ) declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; declaration.Size = 1; declaration.File = TGSI_FILE_NULL; - declaration.Declare = TGSI_DECLARE_RANGE; declaration.UsageMask = TGSI_WRITEMASK_XYZW; - declaration.Interpolate = 0; + declaration.Interpolate = TGSI_INTERPOLATE_CONSTANT; declaration.Semantic = 0; declaration.Padding = 0; declaration.Extended = 0; @@ -103,7 +102,6 @@ tgsi_default_declaration( void ) struct tgsi_declaration tgsi_build_declaration( unsigned file, - unsigned declare, unsigned usage_mask, unsigned interpolate, unsigned semantic, @@ -112,11 +110,10 @@ tgsi_build_declaration( struct tgsi_declaration declaration; assert( file <= TGSI_FILE_IMMEDIATE ); - assert( declare <= TGSI_DECLARE_MASK ); + assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); declaration = tgsi_default_declaration(); declaration.File = file; - declaration.Declare = declare; declaration.UsageMask = usage_mask; declaration.Interpolate = interpolate; declaration.Semantic = semantic; @@ -144,7 +141,7 @@ tgsi_default_full_declaration( void ) struct tgsi_full_declaration full_declaration; full_declaration.Declaration = tgsi_default_declaration(); - full_declaration.Interpolation = tgsi_default_declaration_interpolation(); + full_declaration.DeclarationRange = tgsi_default_declaration_range(); full_declaration.Semantic = tgsi_default_declaration_semantic(); return full_declaration; @@ -159,6 +156,7 @@ tgsi_build_full_declaration( { unsigned size = 0; struct tgsi_declaration *declaration; + struct tgsi_declaration_range *dr; if( maxsize <= size ) return 0; @@ -167,63 +165,21 @@ tgsi_build_full_declaration( *declaration = tgsi_build_declaration( full_decl->Declaration.File, - full_decl->Declaration.Declare, full_decl->Declaration.UsageMask, full_decl->Declaration.Interpolate, full_decl->Declaration.Semantic, header ); - switch( full_decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - { - struct tgsi_declaration_range *dr; - - if( maxsize <= size ) - return 0; - dr = (struct tgsi_declaration_range *) &tokens[size]; - size++; - - *dr = tgsi_build_declaration_range( - full_decl->u.DeclarationRange.First, - full_decl->u.DeclarationRange.Last, - declaration, - header ); - break; - } - - case TGSI_DECLARE_MASK: - { - struct tgsi_declaration_mask *dm; - - if( maxsize <= size ) - return 0; - dm = (struct tgsi_declaration_mask *) &tokens[size]; - size++; - - *dm = tgsi_build_declaration_mask( - full_decl->u.DeclarationMask.Mask, - declaration, - header ); - break; - } - - default: - assert( 0 ); - } - - if( full_decl->Declaration.Interpolate ) { - struct tgsi_declaration_interpolation *di; - - if( maxsize <= size ) - return 0; - di = (struct tgsi_declaration_interpolation *) &tokens[size]; - size++; + if (maxsize <= size) + return 0; + dr = (struct tgsi_declaration_range *) &tokens[size]; + size++; - *di = tgsi_build_declaration_interpolation( - full_decl->Interpolation.Interpolate, - declaration, - header ); - } + *dr = tgsi_build_declaration_range( + full_decl->DeclarationRange.First, + full_decl->DeclarationRange.Last, + declaration, + header ); if( full_decl->Declaration.Semantic ) { struct tgsi_declaration_semantic *ds; @@ -243,6 +199,17 @@ tgsi_build_full_declaration( return size; } +struct tgsi_declaration_range +tgsi_default_declaration_range( void ) +{ + struct tgsi_declaration_range dr; + + dr.First = 0; + dr.Last = 0; + + return dr; +} + struct tgsi_declaration_range tgsi_build_declaration_range( unsigned first, @@ -255,6 +222,7 @@ tgsi_build_declaration_range( assert( last >= first ); assert( last <= 0xFFFF ); + declaration_range = tgsi_default_declaration_range(); declaration_range.First = first; declaration_range.Last = last; @@ -263,50 +231,6 @@ tgsi_build_declaration_range( return declaration_range; } -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_mask declaration_mask; - - declaration_mask.Mask = mask; - - declaration_grow( declaration, header ); - - return declaration_mask; -} - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ) -{ - struct tgsi_declaration_interpolation di; - - di.Interpolate = TGSI_INTERPOLATE_CONSTANT; - di.Padding = 0; - - return di; -} - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_interpolation di; - - assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); - - di = tgsi_default_declaration_interpolation(); - di.Interpolate = interpolate; - - declaration_grow( declaration, header ); - - return di; -} - struct tgsi_declaration_semantic tgsi_default_declaration_semantic( void ) { diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.h b/src/gallium/auxiliary/tgsi/util/tgsi_build.h index 80bffc4ae7..423cf141f5 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.h @@ -37,7 +37,6 @@ tgsi_default_declaration( void ); struct tgsi_declaration tgsi_build_declaration( unsigned file, - unsigned declare, unsigned usage_mask, unsigned interpolate, unsigned semantic, @@ -53,6 +52,9 @@ tgsi_build_full_declaration( struct tgsi_header *header, unsigned maxsize ); +struct tgsi_declaration_range +tgsi_default_declaration_range( void ); + struct tgsi_declaration_range tgsi_build_declaration_range( unsigned first, @@ -60,21 +62,6 @@ tgsi_build_declaration_range( struct tgsi_declaration *declaration, struct tgsi_header *header ); -struct tgsi_declaration_mask -tgsi_build_declaration_mask( - unsigned mask, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_interpolation -tgsi_default_declaration_interpolation( void ); - -struct tgsi_declaration_interpolation -tgsi_build_declaration_interpolation( - unsigned interpolate, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - struct tgsi_declaration_semantic tgsi_default_declaration_semantic( void ); diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index b018ea9fa1..d1a3dfd9c7 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -546,19 +546,13 @@ tgsi_dump_declaration( TXT( "\nDCL " ); ENM( decl->Declaration.File, TGSI_FILES_SHORT ); - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - CHR( '[' ); - UID( decl->u.DeclarationRange.First ); - if( decl->u.DeclarationRange.First != decl->u.DeclarationRange.Last ) { - TXT( ".." ); - UID( decl->u.DeclarationRange.Last ); - } - CHR( ']' ); - break; - default: - assert( 0 ); + CHR( '[' ); + UID( decl->DeclarationRange.First ); + if (decl->DeclarationRange.First != decl->DeclarationRange.Last) { + TXT( ".." ); + UID( decl->DeclarationRange.Last ); } + CHR( ']' ); if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { CHR( '.' ); @@ -586,10 +580,8 @@ tgsi_dump_declaration( } } - if (decl->Declaration.Interpolate) { - TXT( ", " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES_SHORT ); - } + TXT( ", " ); + ENM( decl->Declaration.Interpolate, TGSI_INTERPOLATES_SHORT ); } static void @@ -601,8 +593,6 @@ dump_declaration_verbose( { TXT( "\nFile : " ); ENM( decl->Declaration.File, TGSI_FILES ); - TXT( "\nDeclare : " ); - ENM( decl->Declaration.Declare, TGSI_DECLARES ); if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { TXT( "\nUsageMask : " ); if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { @@ -620,7 +610,7 @@ dump_declaration_verbose( } if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { TXT( "\nInterpolate: " ); - UID( decl->Declaration.Interpolate ); + ENM( decl->Declaration.Interpolate, TGSI_INTERPOLATES ); } if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { TXT( "\nSemantic : " ); @@ -632,32 +622,10 @@ dump_declaration_verbose( } EOL(); - switch( decl->Declaration.Declare ) { - case TGSI_DECLARE_RANGE: - TXT( "\nFirst: " ); - UID( decl->u.DeclarationRange.First ); - TXT( "\nLast : " ); - UID( decl->u.DeclarationRange.Last ); - break; - - case TGSI_DECLARE_MASK: - TXT( "\nMask: " ); - UIX( decl->u.DeclarationMask.Mask ); - break; - - default: - assert( 0 ); - } - - if( decl->Declaration.Interpolate ) { - EOL(); - TXT( "\nInterpolate: " ); - ENM( decl->Interpolation.Interpolate, TGSI_INTERPOLATES ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Interpolation.Padding ); - } - } + TXT( "\nFirst: " ); + UID( decl->DeclarationRange.First ); + TXT( "\nLast : " ); + UID( decl->DeclarationRange.Last ); if( decl->Declaration.Semantic ) { EOL(); diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c index 5c0b0bfd61..d16f0cdcad 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c @@ -118,22 +118,7 @@ tgsi_parse_token( *decl = tgsi_default_full_declaration(); decl->Declaration = *(struct tgsi_declaration *) &token; - switch( decl->Declaration.Type ) { - case TGSI_DECLARE_RANGE: - next_token( ctx, &decl->u.DeclarationRange ); - break; - - case TGSI_DECLARE_MASK: - next_token( ctx, &decl->u.DeclarationMask ); - break; - - default: - assert (0); - } - - if( decl->Declaration.Interpolate ) { - next_token( ctx, &decl->Interpolation ); - } + next_token( ctx, &decl->DeclarationRange ); if( decl->Declaration.Semantic ) { next_token( ctx, &decl->Semantic ); diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h index 4102101093..054350712d 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h @@ -65,13 +65,8 @@ struct tgsi_full_src_register struct tgsi_full_declaration { struct tgsi_declaration Declaration; - union - { - struct tgsi_declaration_range DeclarationRange; - struct tgsi_declaration_mask DeclarationMask; - } u; - struct tgsi_declaration_interpolation Interpolation; - struct tgsi_declaration_semantic Semantic; + struct tgsi_declaration_range DeclarationRange; + struct tgsi_declaration_semantic Semantic; }; struct tgsi_full_immediate diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c index 65650ed22a..bda7bc2e2e 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -93,8 +93,8 @@ tgsi_scan_shader(const struct tgsi_token *tokens, = &parse.FullToken.FullDeclaration; uint file = fulldecl->Declaration.File; uint i; - for (i = fulldecl->u.DeclarationRange.First; - i <= fulldecl->u.DeclarationRange.Last; + for (i = fulldecl->DeclarationRange.First; + i <= fulldecl->DeclarationRange.Last; i++) { /* only first 32 regs will appear in this bitfield */ -- cgit v1.2.3 From 56fc7690d791819d81ff1c6e6e22d22017c68919 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 19:50:58 +0200 Subject: util: Fix build after TGSI declaration interface changes. --- src/gallium/auxiliary/util/u_simple_shaders.c | 33 +++++++++++++-------------- 1 file changed, 16 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index 5f8d12191d..505d93d727 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -92,8 +92,8 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, decl.Semantic.SemanticName = semantic_names[i]; decl.Semantic.SemanticIndex = semantic_indexes[i]; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = i; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = i; ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, @@ -107,8 +107,8 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe, decl.Declaration.Semantic = 1; decl.Semantic.SemanticName = semantic_names[i]; decl.Semantic.SemanticIndex = semantic_indexes[i]; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = i; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = i; ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, @@ -190,14 +190,13 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, /* declare TEX[0] input */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_INPUT; + /* XXX this could be linear... */ + decl.Declaration.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; decl.Declaration.Semantic = 1; decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC; decl.Semantic.SemanticIndex = 0; - /* XXX this could be linear... */ - decl.Declaration.Interpolate = 1; - decl.Interpolation.Interpolate = TGSI_INTERPOLATE_PERSPECTIVE; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = 0; ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, @@ -209,8 +208,8 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, decl.Declaration.Semantic = 1; decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; decl.Semantic.SemanticIndex = 0; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = 0; ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, @@ -219,8 +218,8 @@ util_make_fragment_tex_shader(struct pipe_context *pipe, /* declare sampler */ decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_SAMPLER; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = 0; ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, @@ -303,8 +302,8 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe, decl.Declaration.Semantic = 1; decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; decl.Semantic.SemanticIndex = 0; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = 0; ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, @@ -316,8 +315,8 @@ util_make_fragment_passthrough_shader(struct pipe_context *pipe, decl.Declaration.Semantic = 1; decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR; decl.Semantic.SemanticIndex = 0; - decl.u.DeclarationRange.First = - decl.u.DeclarationRange.Last = 0; + decl.DeclarationRange.First = + decl.DeclarationRange.Last = 0; ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, -- cgit v1.2.3 From 347d28fd20645674c3509b9fb8ebf8c31a24c239 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 19:51:50 +0200 Subject: cell: Fix build after TGSI declaration interface changes. --- src/gallium/drivers/cell/spu/spu_exec.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 48edc62f49..69b0526120 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -830,13 +830,11 @@ exec_declaration(struct spu_exec_machine *mach, unsigned first, last, mask; interpolation_func interp; - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); - - first = decl->u.DeclarationRange.First; - last = decl->u.DeclarationRange.Last; + first = decl->DeclarationRange.First; + last = decl->DeclarationRange.Last; mask = decl->Declaration.UsageMask; - switch( decl->Interpolation.Interpolate ) { + switch( decl->Declaration.Interpolate ) { case TGSI_INTERPOLATE_CONSTANT: interp = constant_interpolation; break; -- cgit v1.2.3 From 99b46555499005bd9454fb4a91d28d4e7d93dba4 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 19:52:41 +0200 Subject: i915: Fix build after TGSI declaration interface changes. --- src/gallium/drivers/i915simple/i915_fpc_translate.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 3ccf74c72c..48f3ceb2e8 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -943,8 +943,8 @@ i915_translate_instructions(struct i915_fp_compile *p, if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_CONSTANT) { uint i; - for (i = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - i <= parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + for (i = parse.FullToken.FullDeclaration.DeclarationRange.First; + i <= parse.FullToken.FullDeclaration.DeclarationRange.Last; i++) { assert(ifs->constant_flags[i] == 0x0); ifs->constant_flags[i] = I915_CONSTFLAG_USER; @@ -954,8 +954,8 @@ i915_translate_instructions(struct i915_fp_compile *p, else if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_TEMPORARY) { uint i; - for (i = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - i <= parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + for (i = parse.FullToken.FullDeclaration.DeclarationRange.First; + i <= parse.FullToken.FullDeclaration.DeclarationRange.Last; i++) { assert(i < I915_MAX_TEMPORARY); /* XXX just use shader->info->file_mask[TGSI_FILE_TEMPORARY] */ -- cgit v1.2.3 From 01122116144619a93c7ebb852eaffb3a6c96fe67 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 19:54:20 +0200 Subject: brw: Fix build after TGSI declaration interface changes. --- src/gallium/drivers/i965simple/brw_sf.c | 10 +++++----- src/gallium/drivers/i965simple/brw_shader_info.c | 4 +--- src/gallium/drivers/i965simple/brw_vs_emit.c | 6 ++---- src/gallium/drivers/i965simple/brw_wm_decl.c | 8 +++----- 4 files changed, 11 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_sf.c b/src/gallium/drivers/i965simple/brw_sf.c index c3b815a82b..96f8fb87a3 100644 --- a/src/gallium/drivers/i965simple/brw_sf.c +++ b/src/gallium/drivers/i965simple/brw_sf.c @@ -169,9 +169,9 @@ static void upload_sf_prog( struct brw_context *brw ) case TGSI_TOKEN_TYPE_DECLARATION: if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) { - int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; - int interp_mode = parse.FullToken.FullDeclaration.Interpolation.Interpolate; + int first = parse.FullToken.FullDeclaration.DeclarationRange.First; + int last = parse.FullToken.FullDeclaration.DeclarationRange.Last; + int interp_mode = parse.FullToken.FullDeclaration.Declaration.Interpolate; //int semantic = parse.FullToken.FullDeclaration.Semantic.SemanticName; //int semantic_index = parse.FullToken.FullDeclaration.Semantic.SemanticIndex; @@ -291,8 +291,8 @@ static void update_sf_linkage( struct brw_context *brw ) case TGSI_TOKEN_TYPE_DECLARATION: if (parse.FullToken.FullDeclaration.Declaration.File == TGSI_FILE_INPUT) { - int first = parse.FullToken.FullDeclaration.u.DeclarationRange.First; - int last = parse.FullToken.FullDeclaration.u.DeclarationRange.Last; + int first = parse.FullToken.FullDeclaration.DeclarationRange.First; + int last = parse.FullToken.FullDeclaration.DeclarationRange.Last; for (i = first; i < last; i++) { vp_semantic[i].semantic = diff --git a/src/gallium/drivers/i965simple/brw_shader_info.c b/src/gallium/drivers/i965simple/brw_shader_info.c index f4694a4433..fb3da92421 100644 --- a/src/gallium/drivers/i965simple/brw_shader_info.c +++ b/src/gallium/drivers/i965simple/brw_shader_info.c @@ -26,9 +26,7 @@ void brw_shader_info(const struct tgsi_token *tokens, case TGSI_TOKEN_TYPE_DECLARATION: { const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; - unsigned last = decl->u.DeclarationRange.Last; - - assert( decl->Declaration.Declare == TGSI_DECLARE_RANGE ); + unsigned last = decl->DeclarationRange.Last; // Broken by crazy wpos init: //assert( info->nr_regs[decl->Declaration.File] <= last); diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c index 9020fcc001..a1432fece1 100644 --- a/src/gallium/drivers/i965simple/brw_vs_emit.c +++ b/src/gallium/drivers/i965simple/brw_vs_emit.c @@ -988,10 +988,8 @@ post_vs_emit( struct brw_vs_compile *c, struct brw_instruction *end_inst ) static void process_declaration(const struct tgsi_full_declaration *decl, struct brw_prog_info *info) { - int first = decl->u.DeclarationRange.First; - int last = decl->u.DeclarationRange.Last; - - assert (decl->Declaration.Declare != TGSI_DECLARE_MASK); + int first = decl->DeclarationRange.First; + int last = decl->DeclarationRange.Last; switch(decl->Declaration.File) { case TGSI_FILE_CONSTANT: diff --git a/src/gallium/drivers/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c index 74ccfd494a..bf1b4d961a 100644 --- a/src/gallium/drivers/i965simple/brw_wm_decl.c +++ b/src/gallium/drivers/i965simple/brw_wm_decl.c @@ -351,18 +351,16 @@ void brw_wm_emit_decls(struct brw_wm_compile *c) case TGSI_TOKEN_TYPE_DECLARATION: { const struct tgsi_full_declaration *decl = &parse.FullToken.FullDeclaration; - unsigned first = decl->u.DeclarationRange.First; - unsigned last = decl->u.DeclarationRange.Last; + unsigned first = decl->DeclarationRange.First; + unsigned last = decl->DeclarationRange.Last; unsigned mask = decl->Declaration.UsageMask; /* ? */ unsigned i; if (decl->Declaration.File != TGSI_FILE_INPUT) break; - assert(decl->Declaration.Interpolate); - for( i = first; i <= last; i++ ) { - switch (decl->Interpolation.Interpolate) { + switch (decl->Declaration.Interpolate) { case TGSI_INTERPOLATE_CONSTANT: emit_cinterp(c, i, mask); break; -- cgit v1.2.3 From a3212ee313c52ff4fc15804a3b4ddfb186556575 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 31 May 2008 19:55:27 +0200 Subject: mesa: Fix build after TGSI declaration interface changes. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 28 +++++++++++----------------- 1 file changed, 11 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index d2be450bdb..12979de523 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -550,18 +550,16 @@ make_input_decl( decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_INPUT; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; decl.Declaration.UsageMask = usage_mask; decl.Declaration.Semantic = semantic_info; - decl.u.DeclarationRange.First = index; - decl.u.DeclarationRange.Last = index; + decl.DeclarationRange.First = index; + decl.DeclarationRange.Last = index; if (semantic_info) { decl.Semantic.SemanticName = semantic_name; decl.Semantic.SemanticIndex = semantic_index; } if (interpolate_info) { - decl.Declaration.Interpolate = 1; - decl.Interpolation.Interpolate = interpolate; + decl.Declaration.Interpolate = interpolate; } return decl; @@ -583,11 +581,10 @@ make_output_decl( decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_OUTPUT; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; decl.Declaration.UsageMask = usage_mask; decl.Declaration.Semantic = 1; - decl.u.DeclarationRange.First = index; - decl.u.DeclarationRange.Last = index; + decl.DeclarationRange.First = index; + decl.DeclarationRange.Last = index; decl.Semantic.SemanticName = semantic_name; decl.Semantic.SemanticIndex = semantic_index; @@ -603,9 +600,8 @@ make_temp_decl( struct tgsi_full_declaration decl; decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_TEMPORARY; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.u.DeclarationRange.First = start_index; - decl.u.DeclarationRange.Last = end_index; + decl.DeclarationRange.First = start_index; + decl.DeclarationRange.Last = end_index; return decl; } @@ -616,9 +612,8 @@ make_sampler_decl(GLuint index) struct tgsi_full_declaration decl; decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_SAMPLER; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.u.DeclarationRange.First = index; - decl.u.DeclarationRange.Last = index; + decl.DeclarationRange.First = index; + decl.DeclarationRange.Last = index; return decl; } @@ -629,9 +624,8 @@ make_constant_decl(GLuint first, GLuint last) struct tgsi_full_declaration decl; decl = tgsi_default_full_declaration(); decl.Declaration.File = TGSI_FILE_CONSTANT; - decl.Declaration.Declare = TGSI_DECLARE_RANGE; - decl.u.DeclarationRange.First = first; - decl.u.DeclarationRange.Last = last; + decl.DeclarationRange.First = first; + decl.DeclarationRange.Last = last; return decl; } -- cgit v1.2.3 From 837d49a84e24420dbc06924f4862a5f7dcf4cace Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 2 Jun 2008 11:31:52 +0200 Subject: gallium: The SWZ opcode no longer aliases MOV. Also, when the extended swizzle token is used, the simple swizzle and negate are set to X,Y,Z,W and FALSE, respectively. --- src/gallium/include/pipe/p_shader_tokens.h | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 282d32910e..84e5096418 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -232,7 +232,7 @@ struct tgsi_immediate_float32 /* * GL_ARB_vertex_program */ -#define TGSI_OPCODE_SWZ TGSI_OPCODE_MOV +#define TGSI_OPCODE_SWZ 118 #define TGSI_OPCODE_XPD TGSI_OPCODE_CROSSPRODUCT /* @@ -388,7 +388,7 @@ struct tgsi_immediate_float32 #define TGSI_OPCODE_KIL 116 /* unpredicated kill */ #define TGSI_OPCODE_END 117 /* aka HALT */ -#define TGSI_OPCODE_LAST 118 +#define TGSI_OPCODE_LAST 119 #define TGSI_SAT_NONE 0 /* do not saturate */ #define TGSI_SAT_ZERO_ONE 1 /* clamp to [0,1] */ @@ -621,6 +621,10 @@ struct tgsi_src_register_ext * * NegateX, NegateY, NegateZ and NegateW negate individual components of the * source register. + * + * NOTE: To simplify matter, if this token is present, the corresponding Swizzle + * and Negate fields in tgsi_src_register should be set to X,Y,Z,W + * and FALSE, respectively. */ struct tgsi_src_register_ext_swz -- cgit v1.2.3 From c6ae627fdca417318d27a8c26e6d9bc23577aabe Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 2 Jun 2008 11:39:59 +0200 Subject: tgsi: SWZ no longer aliases to MOV. --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 2 +- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 6ba0949183..6689c3f1fb 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -1477,7 +1477,7 @@ exec_instruction( break; case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ + case TGSI_OPCODE_SWZ: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); STORE( &r[0], 0, chan_index ); diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index a59e22b6ea..cdce5ea9c5 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -1190,7 +1190,7 @@ emit_instruction( break; case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ + case TGSI_OPCODE_SWZ: FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); STORE( func, *inst, 0, 0, chan_index ); -- cgit v1.2.3 From dc6068a8bcd66e2cbcf76962c70ba202e0078a49 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 2 Jun 2008 11:40:44 +0200 Subject: cell: SWZ no longer aliases MOV. --- src/gallium/drivers/cell/spu/spu_exec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 69b0526120..3a80df427d 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -896,7 +896,7 @@ exec_instruction( break; case TGSI_OPCODE_MOV: - /* TGSI_OPCODE_SWZ */ + case TGSI_OPCODE_SWZ: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( &r[0], 0, chan_index ); STORE( &r[0], 0, chan_index ); -- cgit v1.2.3 From dfd30b878680dd6dca96928a06a301b837b7a650 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 2 Jun 2008 11:41:17 +0200 Subject: i915: SWZ no longer aliases MOV. --- src/gallium/drivers/i915simple/i915_fpc_translate.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 48f3ceb2e8..23cd909337 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -676,7 +676,7 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_MOV: - /* aka TGSI_OPCODE_SWZ */ + case TGSI_OPCODE_SWZ: emit_simple_arith(p, inst, A0_MOV, 1); break; -- cgit v1.2.3 From 49ed85d6b1cdb74a7985e2d743635c73151bbfdb Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 2 Jun 2008 11:42:16 +0200 Subject: brw: SWZ no longer aliases MOV. --- src/gallium/drivers/i965simple/brw_vs_emit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c index a1432fece1..81423e2d7d 100644 --- a/src/gallium/drivers/i965simple/brw_vs_emit.c +++ b/src/gallium/drivers/i965simple/brw_vs_emit.c @@ -1135,8 +1135,8 @@ static void process_instruction(struct brw_vs_compile *c, emit_min(p, dst, args[0], args[1]); break; case TGSI_OPCODE_MOV: -#if 0 case TGSI_OPCODE_SWZ: +#if 0 /* The args[0] value can't be used here as it won't have * correctly encoded the full swizzle: */ -- cgit v1.2.3 From 2c7ae3371b6058988f7f10bf031d630b649f3831 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 2 Jun 2008 11:59:04 +0200 Subject: tgsi: Add assertions to the new rule that when an extended swizzle is used, the simple swizzle must be set to identity. --- src/gallium/auxiliary/tgsi/util/tgsi_build.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.c b/src/gallium/auxiliary/tgsi/util/tgsi_build.c index 63cc27becc..18e44b38c2 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.c @@ -628,6 +628,14 @@ tgsi_build_full_instruction( tgsi_default_src_register_ext_swz() ) ) { struct tgsi_src_register_ext_swz *src_register_ext_swz; + /* Use of the extended swizzle requires the simple swizzle to be identity. + */ + assert( reg->SrcRegister.SwizzleX == TGSI_SWIZZLE_X ); + assert( reg->SrcRegister.SwizzleY == TGSI_SWIZZLE_Y ); + assert( reg->SrcRegister.SwizzleZ == TGSI_SWIZZLE_Z ); + assert( reg->SrcRegister.SwizzleW == TGSI_SWIZZLE_W ); + assert( reg->SrcRegister.Negate == FALSE ); + if( maxsize <= size ) return 0; src_register_ext_swz = -- cgit v1.2.3 From 5b86ae60fe339ae0b813d16ec328a68ccb2b9514 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 30 May 2008 01:46:59 +0900 Subject: draw: Fix MSVC warnings. --- src/gallium/auxiliary/draw/draw_pipe.c | 3 ++- src/gallium/auxiliary/draw/draw_pt_decompose.h | 6 +++--- src/gallium/auxiliary/draw/draw_pt_varray.c | 4 ++-- src/gallium/auxiliary/draw/draw_vs.c | 6 +++--- src/gallium/auxiliary/draw/draw_vs_aos.c | 4 ++-- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 2 +- 6 files changed, 13 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 1d26706dee..3355c871ee 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -255,7 +255,8 @@ void draw_pipeline_run( struct draw_context *draw, char *verts = (char *)vertices; \ boolean flatfirst = (draw->rasterizer->flatshade && \ draw->rasterizer->flatshade_first); \ - unsigned i, flags + unsigned i; \ + ushort flags #define FLUSH diff --git a/src/gallium/auxiliary/draw/draw_pt_decompose.h b/src/gallium/auxiliary/draw/draw_pt_decompose.h index dccfde99dd..3fb0695687 100644 --- a/src/gallium/auxiliary/draw/draw_pt_decompose.h +++ b/src/gallium/auxiliary/draw/draw_pt_decompose.h @@ -118,9 +118,9 @@ static void FUNC( ARGS, /* These bitflags look a little odd because we submit the * vertices as (1,2,0) to satisfy flatshade requirements. */ - const unsigned edge_first = DRAW_PIPE_EDGE_FLAG_2; - const unsigned edge_middle = DRAW_PIPE_EDGE_FLAG_0; - const unsigned edge_last = DRAW_PIPE_EDGE_FLAG_1; + const ushort edge_first = DRAW_PIPE_EDGE_FLAG_2; + const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const ushort edge_last = DRAW_PIPE_EDGE_FLAG_1; flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 260f28f284..f19e8850b3 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -109,9 +109,9 @@ static INLINE void fetch_init(struct varray_frontend *varray, static INLINE void add_draw_el(struct varray_frontend *varray, - int idx) + unsigned idx) { - varray->draw_elts[varray->draw_count++] = idx; + varray->draw_elts[varray->draw_count++] = (ushort)idx; } diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index ce35112fc1..979f9864fd 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -49,10 +49,10 @@ void draw_vs_set_constants( struct draw_context *draw, if (((unsigned)constants) & 0xf) { if (size > draw->vs.const_storage_size) { if (draw->vs.aligned_constant_storage) - align_free(draw->vs.aligned_constant_storage); + align_free((void *)draw->vs.aligned_constant_storage); draw->vs.aligned_constant_storage = align_malloc( size, 16 ); } - memcpy( draw->vs.aligned_constant_storage, + memcpy( (void*)draw->vs.aligned_constant_storage, constants, size ); constants = draw->vs.aligned_constant_storage; @@ -174,7 +174,7 @@ draw_vs_destroy( struct draw_context *draw ) draw_vs_aos_machine_destroy(draw->vs.aos_machine); if (draw->vs.aligned_constant_storage) - align_free(draw->vs.aligned_constant_storage); + align_free((void*)draw->vs.aligned_constant_storage); tgsi_exec_machine_free_data(&draw->vs.machine); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 725f36b502..9e9f8bac1e 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -492,7 +492,7 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, src->SrcRegister.File, src->SrcRegister.Index); unsigned i; - unsigned swz = 0; + ubyte swz = 0; unsigned negs = 0; unsigned abs = 0; @@ -704,7 +704,7 @@ static void store_dest( struct aos_compilation *cp, static void inject_scalar( struct aos_compilation *cp, struct x86_reg dst, struct x86_reg result, - unsigned swizzle ) + ubyte swizzle ) { sse_shufps(cp->func, dst, dst, swizzle); sse_movss(cp->func, dst, result); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index 45e2092209..b720185709 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -118,7 +118,7 @@ static void get_src_ptr( struct aos_compilation *cp, static void emit_swizzle( struct aos_compilation *cp, struct x86_reg dest, struct x86_reg src, - unsigned shuffle ) + ubyte shuffle ) { sse_shufps(cp->func, dest, src, shuffle); } -- cgit v1.2.3 From 275fc32d588fb6d2b78038f5a97cc2bcd2cd61dc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 2 Jun 2008 19:36:53 +0900 Subject: gallium: Identify each Windows platform individually from scons. --- src/gallium/include/pipe/p_config.h | 51 ++++++++++++++++++++++--------------- 1 file changed, 31 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_config.h b/src/gallium/include/pipe/p_config.h index d2d2ae1617..af3746c026 100644 --- a/src/gallium/include/pipe/p_config.h +++ b/src/gallium/include/pipe/p_config.h @@ -33,12 +33,13 @@ * architecture, and operating system being used. These defines should be used * throughout the code to facilitate porting to new platforms. It is likely that * this file is auto-generated by an autoconf-like tool at some point, as some - * things cannot be determined by existing defines alone. + * things cannot be determined by pre-defined environment alone. * * See also: * - http://gcc.gnu.org/onlinedocs/cpp/Common-Predefined-Macros.html * - echo | gcc -dM -E - | sort * - http://msdn.microsoft.com/en-us/library/b0084kay.aspx + * * @author José Fonseca */ @@ -54,6 +55,15 @@ #define PIPE_CC_GCC #endif +/* + * Meaning of _MSC_VER value: + * - 1400: Visual C++ 2005 + * - 1310: Visual C++ .NET 2003 + * - 1300: Visual C++ .NET 2002 + * + * __MSC__ seems to be an old macro -- it is not pre-defined on recent MSVC + * versions. + */ #if defined(_MSC_VER) || defined(__MSC__) #define PIPE_CC_MSVC #endif @@ -81,7 +91,9 @@ /* - * Operating system + * Operating system family. + * + * See subsystem below for a more fine-grained distinction. */ #if defined(__linux__) @@ -94,32 +106,31 @@ /* - * Subsystem + * Subsystem. * - * XXX: There is no way to autodetect this. + * NOTE: There is no way to auto-detect most of these. */ #if defined(PIPE_OS_LINUX) #define PIPE_SUBSYSTEM_DRI -#endif +#endif /* PIPE_OS_LINUX */ #if defined(PIPE_OS_WINDOWS) -#ifndef _WIN32_WCE -#if !defined(PIPE_SUBSYSTEM_USER) && !defined(PIPE_SUBSYSTEM_KERNEL) -#error Neither PIPE_SUBSYSTEM_USER or PIPE_SUBSYSTEM_KERNEL defined. -#endif -#if defined(PIPE_SUBSYSTEM_KERNEL) -#define PIPE_SUBSYSTEM_WINDOWS_DISPLAY -#endif -#if 0 /* FIXME */ -#define PIPE_SUBSYSTEM_WINDOWS_MINIPORT -#endif -#if defined(PIPE_SUBSYSTEM_USER) -#define PIPE_SUBSYSTEM_WINDOWS_USER -#endif -#else /* _WIN32_WCE */ +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) +/* Windows 2000/XP Display Driver */ +#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) +/* Windows 2000/XP Miniport Driver */ +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) +/* Windows User-space Library */ +#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) +/* Windows CE 5.0/6.0 */ +#else +#ifdef _WIN32_WCE #define PIPE_SUBSYSTEM_WINDOWS_CE -#endif /* _WIN32_WCE */ +#else /* !_WIN32_WCE */ +#error No PIPE_SUBSYSTEM_WINDOWS_xxx subsystem defined. +#endif /* !_WIN32_WCE */ +#endif #endif /* PIPE_OS_WINDOWS */ -- cgit v1.2.3 From 9b50043ea9e20c15a1be6735e533f5cc25a253ca Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 2 Jun 2008 19:46:05 +0900 Subject: gallium: Hopefully fix the cosf/sinf/etc. conditional compolation logic for good. --- src/gallium/include/pipe/p_util.h | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 9497eeeec9..cb3cd264e9 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -410,8 +410,7 @@ extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, -#if !defined(_MSC_VER) || _MSC_VER < 0x800 -#if !defined(_INC_MATH) || !defined(__cplusplus) +#if defined(_MSC_VER) && !defined(__cplusplus) static INLINE float cosf( float f ) { @@ -453,7 +452,6 @@ static INLINE float logf( float f ) return (float) cos( (double) f ); } -#endif /* _INC_MATH */ #endif -- cgit v1.2.3 From f4364cd1a6f9e825ea183c1fa6d1050b7e113695 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 30 May 2008 18:46:40 +0200 Subject: i915: Fixed initialization of surface --- src/gallium/drivers/i915simple/i915_texture.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index f0d00280c2..16354dce50 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -624,8 +624,8 @@ i915_get_tex_surface(struct pipe_screen *screen, ps = CALLOC_STRUCT(pipe_surface);//ws->surface_alloc(ws); if (ps) { - assert(ps->refcount); - assert(ps->winsys); + ps->refcount = 1; + ps->winsys = ws; pipe_texture_reference(&ps->texture, pt); pipe_buffer_reference(ws, &ps->buffer, tex->buffer); ps->format = pt->format; -- cgit v1.2.3 From c1949e2bd3acc45c23cc434eef2b0d6aae9092ca Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 2 Jun 2008 12:55:35 +0200 Subject: i915: Fixed some warnings --- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 4 +--- src/gallium/drivers/i915simple/i915_state_emit.c | 1 + 2 files changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index f9297e29fc..adfb16fbc5 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -253,7 +253,6 @@ draw_arrays_fallback( struct vbuf_render *render, { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; - struct pipe_winsys *winsys = i915->pipe.winsys; unsigned nr_indices; if (i915->dirty) @@ -372,7 +371,6 @@ i915_vbuf_render_draw( struct vbuf_render *render, { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; - struct pipe_winsys *winsys = i915->pipe.winsys; unsigned save_nr_indices; save_nr_indices = nr_indices; @@ -424,7 +422,6 @@ i915_vbuf_render_release_vertices( struct vbuf_render *render, { struct i915_vbuf_render *i915_render = i915_vbuf_render(render); struct i915_context *i915 = i915_render->i915; - struct pipe_winsys *winsys = i915->pipe.winsys; size_t size = (size_t)vertex_size * (size_t)vertices_used; assert(i915->vbo); @@ -502,6 +499,7 @@ struct draw_stage *i915_draw_vbuf_stage( struct i915_context *i915 ) render->destroy(render); return NULL; } + /** TODO JB: this shouldn't be here */ draw_set_render(i915->draw, render); return stage; diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index 8bcc26ad35..de0c1a787a 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -377,6 +377,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) { uint w, h; boolean k = framebuffer_size(&i915->framebuffer, &w, &h); + (void)k; assert(k); OUT_BATCH(_3DSTATE_DRAW_RECT_CMD); -- cgit v1.2.3 From 969a207fe356d152b65085a9113502c7fbb5712e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 2 Jun 2008 20:16:49 +0900 Subject: gallium: Fix log<->cos typo in logf. --- src/gallium/include/pipe/p_util.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 0d8ed167b2..3d8ad48d4f 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -452,7 +452,7 @@ static INLINE float fabsf( float f ) static INLINE float logf( float f ) { - return (float) cos( (double) f ); + return (float) log( (double) f ); } #endif /* _INC_MATH */ #endif -- cgit v1.2.3 From acdf24e53047892b83dc5b92567694600ffb8cf5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 2 Jun 2008 20:16:49 +0900 Subject: gallium: Fix log<->cos typo in logf. --- src/gallium/include/pipe/p_util.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index cb3cd264e9..0a9e2ef1b6 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -449,7 +449,7 @@ static INLINE float fabsf( float f ) static INLINE float logf( float f ) { - return (float) cos( (double) f ); + return (float) log( (double) f ); } #endif -- cgit v1.2.3 From aa1a39d1a742c1bb346ba14814d6bf7b44e646cb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 2 Jun 2008 20:46:05 +0900 Subject: rtasm: Use enum sse_cc in sse_cmpps. --- src/gallium/auxiliary/rtasm/rtasm_x86sse.c | 2 +- src/gallium/auxiliary/rtasm/rtasm_x86sse.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c index 664a69a537..f4ca282dd9 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.c +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.c @@ -888,7 +888,7 @@ void sse_unpcklps( struct x86_function *p, struct x86_reg dst, struct x86_reg sr void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, - unsigned char cc) + enum sse_cc cc) { DUMP_RRI( dst, src, cc ); emit_2ub(p, X86_TWOB, 0xC2); diff --git a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h index bd76e1729c..af94577aab 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_x86sse.h +++ b/src/gallium/auxiliary/rtasm/rtasm_x86sse.h @@ -191,7 +191,7 @@ void sse_divss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ) void sse_andnps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_andps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_cmpps( struct x86_function *p, struct x86_reg dst, struct x86_reg src, - unsigned char cc ); + enum sse_cc cc ); void sse_maxps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_maxss( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); void sse_minps( struct x86_function *p, struct x86_reg dst, struct x86_reg src ); -- cgit v1.2.3 From 50274111341e82e1f26b1f3316042e5fe610ec8a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 2 Jun 2008 21:43:31 +0900 Subject: gallium: More tweaks for the cosf/sinf logic. --- src/gallium/include/pipe/p_util.h | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 0a9e2ef1b6..5547e57833 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -410,8 +410,9 @@ extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, -#if defined(_MSC_VER) && !defined(__cplusplus) - +#if defined(_MSC_VER) +#if _MSC_VER < 1400 && !defined(__cplusplus) + static INLINE float cosf( float f ) { return (float) cos( (double) f ); @@ -452,7 +453,14 @@ static INLINE float logf( float f ) return (float) log( (double) f ); } +#else +/* Work-around an extra semi-colon in VS 2005 logf definition */ +#ifdef logf +#undef logf +#define logf(x) ((float)log((double)(x))) +#endif /* logf */ #endif +#endif /* _MSC_VER */ #ifdef __cplusplus -- cgit v1.2.3 From 8d9a96386a5be7f15968bed63ca8b3e5555bbeeb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 2 Jun 2008 21:45:25 +0900 Subject: gallium: Port util_time functions to windows userspace. --- src/gallium/auxiliary/util/u_time.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_time.c b/src/gallium/auxiliary/util/u_time.c index c6ce0afab0..49dce75289 100644 --- a/src/gallium/auxiliary/util/u_time.c +++ b/src/gallium/auxiliary/util/u_time.c @@ -43,7 +43,7 @@ #elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) #include extern VOID KeQuerySystemTime(PLARGE_INTEGER); -#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) #include #else #error Unsupported OS @@ -52,7 +52,7 @@ extern VOID KeQuerySystemTime(PLARGE_INTEGER); #include "util/u_time.h" -#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) static int64_t frequency = 0; @@ -64,7 +64,7 @@ util_time_get_frequency(void) LONGLONG temp; EngQueryPerformanceFrequency(&temp); frequency = temp; -#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) LARGE_INTEGER temp; QueryPerformanceFrequency(&temp); frequency = temp.QuadPart; @@ -89,7 +89,7 @@ util_time_get(struct util_time *t) LARGE_INTEGER temp; KeQuerySystemTime(&temp); t->counter = temp.QuadPart; -#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) LARGE_INTEGER temp; QueryPerformanceCounter(&temp); t->counter = temp.QuadPart; @@ -105,7 +105,7 @@ util_time_add(const struct util_time *t1, #if defined(PIPE_OS_LINUX) t2->tv.tv_sec = t1->tv.tv_sec + usecs / 1000000; t2->tv.tv_usec = t1->tv.tv_usec + usecs % 1000000; -#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) util_time_get_frequency(); t2->counter = t1->counter + (usecs * frequency + INT64_C(999999))/INT64_C(1000000); #elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) @@ -127,7 +127,7 @@ util_time_diff(const struct util_time *t1, #if defined(PIPE_OS_LINUX) return (t2->tv.tv_usec - t1->tv.tv_usec) + (t2->tv.tv_sec - t1->tv.tv_sec)*1000000; -#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) || defined(PIPE_SUBSYSTEM_WINDOWS_USER) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) util_time_get_frequency(); return (t2->counter - t1->counter)*INT64_C(1000000)/frequency; #elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) -- cgit v1.2.3 From e0860518dfb5a5c6ba6584e3c1b5d7b203277dac Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 2 Jun 2008 22:31:02 +0900 Subject: gallium: Replace XSTDCALL by PIPE_CDECL. --- src/gallium/auxiliary/draw/draw_vs_sse.c | 2 +- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 38 +++++++---------------------- src/gallium/drivers/softpipe/sp_fs_sse.c | 2 +- src/gallium/include/pipe/p_compiler.h | 16 +----------- 4 files changed, 12 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index f638208bf5..0aa0c9a76b 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -50,7 +50,7 @@ #define SSE_MAX_VERTICES 4 -typedef void (XSTDCALL *codegen_function) ( +typedef void (PIPE_CDECL *codegen_function) ( const struct tgsi_exec_vector *input, /* 1 */ struct tgsi_exec_vector *output, /* 2 */ float (*constant)[4], /* 3 */ diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c index cdce5ea9c5..cdbdf5c882 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c @@ -194,22 +194,12 @@ get_coef( } -#ifdef WIN32 -static void -emit_retw( - struct x86_function *func, - unsigned short size ) -{ - x86_retw( func, size ); -} -#else static void emit_ret( struct x86_function *func ) { x86_ret( func ); } -#endif /** @@ -475,7 +465,7 @@ static void emit_func_call_dst( struct x86_function *func, unsigned xmm_dst, - void (*code)() ) + void (PIPE_CDECL *code)() ) { sse_movaps( func, @@ -496,9 +486,7 @@ emit_func_call_dst( x86_push( func, ecx ); x86_mov_reg_imm( func, ecx, (unsigned long) code ); x86_call( func, ecx ); -#ifndef WIN32 x86_pop(func, ecx ); -#endif } @@ -516,7 +504,7 @@ emit_func_call_dst_src( struct x86_function *func, unsigned xmm_dst, unsigned xmm_src, - void (*code)() ) + void (PIPE_CDECL *code)() ) { sse_movaps( func, @@ -558,7 +546,7 @@ emit_add( make_xmm( xmm_src ) ); } -static void XSTDCALL +static void PIPE_CDECL cos4f( float *store ) { @@ -581,7 +569,7 @@ emit_cos( cos4f ); } -static void XSTDCALL +static void PIPE_CDECL ex24f( float *store ) { @@ -615,7 +603,7 @@ emit_f2it( make_xmm( xmm ) ); } -static void XSTDCALL +static void PIPE_CDECL flr4f( float *store ) { @@ -638,7 +626,7 @@ emit_flr( flr4f ); } -static void XSTDCALL +static void PIPE_CDECL frc4f( float *store ) { @@ -661,7 +649,7 @@ emit_frc( frc4f ); } -static void XSTDCALL +static void PIPE_CDECL lg24f( float *store ) { @@ -720,7 +708,7 @@ emit_neg( TGSI_EXEC_TEMP_80000000_C ) ); } -static void XSTDCALL +static void PIPE_CDECL pow4f( float *store ) { @@ -820,7 +808,7 @@ emit_setsign( TGSI_EXEC_TEMP_80000000_C ) ); } -static void XSTDCALL +static void PIPE_CDECL sin4f( float *store ) { @@ -1736,11 +1724,7 @@ emit_instruction( break; case TGSI_OPCODE_RET: -#ifdef WIN32 - emit_retw( func, 16 ); -#else emit_ret( func ); -#endif break; case TGSI_OPCODE_END: @@ -2281,11 +2265,7 @@ tgsi_emit_sse2( func, get_immediate_base() ); -#ifdef WIN32 - emit_retw( func, 16 ); -#else emit_ret( func ); -#endif tgsi_parse_free( &parse ); diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 55741cc1df..69f7f960aa 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -46,7 +46,7 @@ /* Surely this should be defined somewhere in a tgsi header: */ -typedef void (XSTDCALL *codegen_function)( +typedef void (PIPE_CDECL *codegen_function)( const struct tgsi_exec_vector *input, struct tgsi_exec_vector *output, const float (*constant)[4], diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index f6707385cd..b14260dd90 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -128,7 +128,7 @@ typedef unsigned char boolean; /* This should match linux gcc cdecl semantics everywhere, so that we * just codegen one calling convention on all platforms. */ -#ifdef WIN32 +#ifdef _MSC_VER #define PIPE_CDECL __cdecl #else #define PIPE_CDECL @@ -148,18 +148,4 @@ typedef unsigned char boolean; -/** - * For calling code-gen'd functions, phase out in favor of - * PIPE_CDECL, above, which really means cdecl on all platforms, not - * like the below... - */ -#if !defined(XSTDCALL) -#if defined(WIN32) -#define XSTDCALL __stdcall /* phase this out */ -#else -#define XSTDCALL /* XXX: NOTE! not STDCALL! */ -#endif -#endif - - #endif /* P_COMPILER_H */ -- cgit v1.2.3 From 6a39bcf3752df7c22cdd38b4645a885eb318add8 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 2 Jun 2008 14:36:27 +0100 Subject: draw: fast element translate path without delta --- src/gallium/auxiliary/draw/draw_pt_vcache.c | 124 +++++++++++++++++++++------- 1 file changed, 93 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 9ffc99abf9..ad86ab4292 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -201,39 +201,74 @@ static void vcache_ef_quad( struct vcache_frontend *vcache, #define FUNC vcache_run #include "draw_pt_vcache_tmp.h" +static void rebase_uint_elts( const unsigned *src, + unsigned count, + int delta, + ushort *dest ) +{ + unsigned i; + + for (i = 0; i < count; i++) + dest[i] = (ushort)(src[i] + delta); +} + +static void rebase_ushort_elts( const ushort *src, + unsigned count, + int delta, + ushort *dest ) +{ + unsigned i; + + for (i = 0; i < count; i++) + dest[i] = (ushort)(src[i] + delta); +} + +static void rebase_ubyte_elts( const ubyte *src, + unsigned count, + int delta, + ushort *dest ) +{ + unsigned i; + + for (i = 0; i < count; i++) + dest[i] = (ushort)(src[i] + delta); +} + + + static void translate_uint_elts( const unsigned *src, unsigned count, - int delta, ushort *dest ) { unsigned i; for (i = 0; i < count; i++) - dest[i] = (ushort)(src[i] + delta); + dest[i] = (ushort)(src[i]); } static void translate_ushort_elts( const ushort *src, unsigned count, - int delta, ushort *dest ) { unsigned i; for (i = 0; i < count; i++) - dest[i] = (ushort)(src[i] + delta); + dest[i] = (ushort)(src[i]); } static void translate_ubyte_elts( const ubyte *src, unsigned count, - int delta, ushort *dest ) { unsigned i; for (i = 0; i < count; i++) - dest[i] = (ushort)(src[i] + delta); + dest[i] = (ushort)(src[i]); } + + + #if 0 static enum pipe_format format_from_get_elt( pt_elt_func get_elt ) { @@ -282,31 +317,58 @@ static void vcache_check_run( struct draw_pt_front_end *frontend, if (!storage) goto fail; - switch(index_size) { - case 1: - translate_ubyte_elts( (const ubyte *)elts, - draw_count, - 0 - (int)min_index, - storage ); - break; - - case 2: - translate_ushort_elts( (const ushort *)elts, - draw_count, - 0 - (int)min_index, - storage ); - break; - - case 4: - translate_uint_elts( (const uint *)elts, - draw_count, - 0 - (int)min_index, - storage ); - break; - - default: - assert(0); - return; + if (min_index == 0) { + switch(index_size) { + case 1: + translate_ubyte_elts( (const ubyte *)elts, + draw_count, + storage ); + break; + + case 2: + translate_ushort_elts( (const ushort *)elts, + draw_count, + storage ); + break; + + case 4: + translate_uint_elts( (const uint *)elts, + draw_count, + storage ); + break; + + default: + assert(0); + return; + } + } + else { + switch(index_size) { + case 1: + rebase_ubyte_elts( (const ubyte *)elts, + draw_count, + 0 - (int)min_index, + storage ); + break; + + case 2: + rebase_ushort_elts( (const ushort *)elts, + draw_count, + 0 - (int)min_index, + storage ); + break; + + case 4: + rebase_uint_elts( (const uint *)elts, + draw_count, + 0 - (int)min_index, + storage ); + break; + + default: + assert(0); + return; + } } transformed_elts = storage; } -- cgit v1.2.3 From 7c22bb383a8fcccf71d7916ce91ae6cb17ac1e3c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 3 Jun 2008 00:03:17 +0900 Subject: gdi: Port of the gdi mesa driver to gallium (Alan Hourihane). --- src/gallium/winsys/gdi/SConscript | 33 ++ src/gallium/winsys/gdi/colors.h | 29 ++ src/gallium/winsys/gdi/opengl32.def | 859 ++++++++++++++++++++++++++++++++++ src/gallium/winsys/gdi/wgl.c | 703 ++++++++++++++++++++++++++++ src/gallium/winsys/gdi/wmesa.c | 890 ++++++++++++++++++++++++++++++++++++ src/gallium/winsys/gdi/wmesadef.h | 40 ++ 6 files changed, 2554 insertions(+) create mode 100644 src/gallium/winsys/gdi/SConscript create mode 100644 src/gallium/winsys/gdi/colors.h create mode 100644 src/gallium/winsys/gdi/opengl32.def create mode 100644 src/gallium/winsys/gdi/wgl.c create mode 100644 src/gallium/winsys/gdi/wmesa.c create mode 100644 src/gallium/winsys/gdi/wmesadef.h (limited to 'src') diff --git a/src/gallium/winsys/gdi/SConscript b/src/gallium/winsys/gdi/SConscript new file mode 100644 index 0000000000..170fdf5127 --- /dev/null +++ b/src/gallium/winsys/gdi/SConscript @@ -0,0 +1,33 @@ +####################################################################### +# SConscript for gdi winsys + +Import('*') + +if env['platform'] == 'windows': + + env = env.Clone() + + env.Append(CPPPATH = [ + '#src/mesa/glapi', + '#src/mesa', + '#src/mesa/main', + ]) + + sources = [ + 'opengl32.def', + 'wgl.c', + 'wmesa.c', + ] + + drivers = [ + softpipe, + ] + + env.Append(LIBS = ['gdi32', 'user32']) + + # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions + env.SharedLibrary( + target ='opengl32', + source = sources, + LIBS = glapi + mesa + drivers + auxiliaries + env['LIBS'], + ) diff --git a/src/gallium/winsys/gdi/colors.h b/src/gallium/winsys/gdi/colors.h new file mode 100644 index 0000000000..03e512c1fa --- /dev/null +++ b/src/gallium/winsys/gdi/colors.h @@ -0,0 +1,29 @@ +/* Values for wmesa->pixelformat: */ + +#define PF_8A8B8G8R 3 /* 32-bit TrueColor: 8-A, 8-B, 8-G, 8-R */ +#define PF_8R8G8B 4 /* 32-bit TrueColor: 8-R, 8-G, 8-B */ +#define PF_5R6G5B 5 /* 16-bit TrueColor: 5-R, 6-G, 5-B bits */ +#define PF_DITHER8 6 /* Dithered RGB using a lookup table */ +#define PF_LOOKUP 7 /* Undithered RGB using a lookup table */ +#define PF_GRAYSCALE 10 /* Grayscale or StaticGray */ +#define PF_BADFORMAT 11 +#define PF_INDEX8 12 + + +#define BGR8(r,g,b) (unsigned)(((BYTE)((b & 0xc0) | ((g & 0xe0)>>2) | \ + ((r & 0xe0)>>5)))) + +/* Windows uses 5,5,5 for 16-bit */ +#define BGR16(r,g,b) ( (((unsigned short)b ) >> 3) | \ + (((unsigned short)g & 0xf8) << 2) | \ + (((unsigned short)r & 0xf8) << 7) ) + +#define BGR24(r,g,b) (unsigned long)((DWORD)(((BYTE)(b)| \ + ((WORD)((BYTE)(g))<<8))| \ + (((DWORD)(BYTE)(r))<<16))) + +#define BGR32(r,g,b) (unsigned long)((DWORD)(((BYTE)(b)| \ + ((WORD)((BYTE)(g))<<8))| \ + (((DWORD)(BYTE)(r))<<16))) + + diff --git a/src/gallium/winsys/gdi/opengl32.def b/src/gallium/winsys/gdi/opengl32.def new file mode 100644 index 0000000000..54e72f57b1 --- /dev/null +++ b/src/gallium/winsys/gdi/opengl32.def @@ -0,0 +1,859 @@ +; DO NOT EDIT - This file generated automatically by mesadef.py script +;DESCRIPTION 'Mesa (OpenGL work-alike) for Win32' +VERSION 6.5 +; +; Module definition file for Mesa (OPENGL32.DLL) +; +; Note: The OpenGL functions use the STDCALL +; function calling convention. Microsoft's +; OPENGL32 uses this convention and so must the +; Mesa OPENGL32 so that the Mesa DLL can be used +; as a drop-in replacement. +; +; The linker exports STDCALL entry points with +; 'decorated' names; e.g., _glBegin@0, where the +; trailing number is the number of bytes of +; parameter data pushed onto the stack. The +; callee is responsible for popping this data +; off the stack, usually via a RETF n instruction. +; +; However, the Microsoft OPENGL32.DLL does not export +; the decorated names, even though the calling convention +; is STDCALL. So, this module definition file is +; needed to force the Mesa OPENGL32.DLL to export the +; symbols in the same manner as the Microsoft DLL. +; Were it not for this problem, this file would not +; be needed (for the gl* functions) since the entry +; points are compiled with dllexport declspec. +; +; However, this file is still needed to export "internal" +; Mesa symbols for the benefit of the OSMESA32.DLL. +; +EXPORTS + glNewList + glEndList + glCallList + glCallLists + glDeleteLists + glGenLists + glListBase + glBegin + glBitmap + glColor3b + glColor3bv + glColor3d + glColor3dv + glColor3f + glColor3fv + glColor3i + glColor3iv + glColor3s + glColor3sv + glColor3ub + glColor3ubv + glColor3ui + glColor3uiv + glColor3us + glColor3usv + glColor4b + glColor4bv + glColor4d + glColor4dv + glColor4f + glColor4fv + glColor4i + glColor4iv + glColor4s + glColor4sv + glColor4ub + glColor4ubv + glColor4ui + glColor4uiv + glColor4us + glColor4usv + glEdgeFlag + glEdgeFlagv + glEnd + glIndexd + glIndexdv + glIndexf + glIndexfv + glIndexi + glIndexiv + glIndexs + glIndexsv + glNormal3b + glNormal3bv + glNormal3d + glNormal3dv + glNormal3f + glNormal3fv + glNormal3i + glNormal3iv + glNormal3s + glNormal3sv + glRasterPos2d + glRasterPos2dv + glRasterPos2f + glRasterPos2fv + glRasterPos2i + glRasterPos2iv + glRasterPos2s + glRasterPos2sv + glRasterPos3d + glRasterPos3dv + glRasterPos3f + glRasterPos3fv + glRasterPos3i + glRasterPos3iv + glRasterPos3s + glRasterPos3sv + glRasterPos4d + glRasterPos4dv + glRasterPos4f + glRasterPos4fv + glRasterPos4i + glRasterPos4iv + glRasterPos4s + glRasterPos4sv + glRectd + glRectdv + glRectf + glRectfv + glRecti + glRectiv + glRects + glRectsv + glTexCoord1d + glTexCoord1dv + glTexCoord1f + glTexCoord1fv + glTexCoord1i + glTexCoord1iv + glTexCoord1s + glTexCoord1sv + glTexCoord2d + glTexCoord2dv + glTexCoord2f + glTexCoord2fv + glTexCoord2i + glTexCoord2iv + glTexCoord2s + glTexCoord2sv + glTexCoord3d + glTexCoord3dv + glTexCoord3f + glTexCoord3fv + glTexCoord3i + glTexCoord3iv + glTexCoord3s + glTexCoord3sv + glTexCoord4d + glTexCoord4dv + glTexCoord4f + glTexCoord4fv + glTexCoord4i + glTexCoord4iv + glTexCoord4s + glTexCoord4sv + glVertex2d + glVertex2dv + glVertex2f + glVertex2fv + glVertex2i + glVertex2iv + glVertex2s + glVertex2sv + glVertex3d + glVertex3dv + glVertex3f + glVertex3fv + glVertex3i + glVertex3iv + glVertex3s + glVertex3sv + glVertex4d + glVertex4dv + glVertex4f + glVertex4fv + glVertex4i + glVertex4iv + glVertex4s + glVertex4sv + glClipPlane + glColorMaterial + glCullFace + glFogf + glFogfv + glFogi + glFogiv + glFrontFace + glHint + glLightf + glLightfv + glLighti + glLightiv + glLightModelf + glLightModelfv + glLightModeli + glLightModeliv + glLineStipple + glLineWidth + glMaterialf + glMaterialfv + glMateriali + glMaterialiv + glPointSize + glPolygonMode + glPolygonStipple + glScissor + glShadeModel + glTexParameterf + glTexParameterfv + glTexParameteri + glTexParameteriv + glTexImage1D + glTexImage2D + glTexEnvf + glTexEnvfv + glTexEnvi + glTexEnviv + glTexGend + glTexGendv + glTexGenf + glTexGenfv + glTexGeni + glTexGeniv + glFeedbackBuffer + glSelectBuffer + glRenderMode + glInitNames + glLoadName + glPassThrough + glPopName + glPushName + glDrawBuffer + glClear + glClearAccum + glClearIndex + glClearColor + glClearStencil + glClearDepth + glStencilMask + glColorMask + glDepthMask + glIndexMask + glAccum + glDisable + glEnable + glFinish + glFlush + glPopAttrib + glPushAttrib + glMap1d + glMap1f + glMap2d + glMap2f + glMapGrid1d + glMapGrid1f + glMapGrid2d + glMapGrid2f + glEvalCoord1d + glEvalCoord1dv + glEvalCoord1f + glEvalCoord1fv + glEvalCoord2d + glEvalCoord2dv + glEvalCoord2f + glEvalCoord2fv + glEvalMesh1 + glEvalPoint1 + glEvalMesh2 + glEvalPoint2 + glAlphaFunc + glBlendFunc + glLogicOp + glStencilFunc + glStencilOp + glDepthFunc + glPixelZoom + glPixelTransferf + glPixelTransferi + glPixelStoref + glPixelStorei + glPixelMapfv + glPixelMapuiv + glPixelMapusv + glReadBuffer + glCopyPixels + glReadPixels + glDrawPixels + glGetBooleanv + glGetClipPlane + glGetDoublev + glGetError + glGetFloatv + glGetIntegerv + glGetLightfv + glGetLightiv + glGetMapdv + glGetMapfv + glGetMapiv + glGetMaterialfv + glGetMaterialiv + glGetPixelMapfv + glGetPixelMapuiv + glGetPixelMapusv + glGetPolygonStipple + glGetString + glGetTexEnvfv + glGetTexEnviv + glGetTexGendv + glGetTexGenfv + glGetTexGeniv + glGetTexImage + glGetTexParameterfv + glGetTexParameteriv + glGetTexLevelParameterfv + glGetTexLevelParameteriv + glIsEnabled + glIsList + glDepthRange + glFrustum + glLoadIdentity + glLoadMatrixf + glLoadMatrixd + glMatrixMode + glMultMatrixf + glMultMatrixd + glOrtho + glPopMatrix + glPushMatrix + glRotated + glRotatef + glScaled + glScalef + glTranslated + glTranslatef + glViewport + glArrayElement + glColorPointer + glDisableClientState + glDrawArrays + glDrawElements + glEdgeFlagPointer + glEnableClientState + glGetPointerv + glIndexPointer + glInterleavedArrays + glNormalPointer + glTexCoordPointer + glVertexPointer + glPolygonOffset + glCopyTexImage1D + glCopyTexImage2D + glCopyTexSubImage1D + glCopyTexSubImage2D + glTexSubImage1D + glTexSubImage2D + glAreTexturesResident + glBindTexture + glDeleteTextures + glGenTextures + glIsTexture + glPrioritizeTextures + glIndexub + glIndexubv + glPopClientAttrib + glPushClientAttrib + glBlendColor + glBlendEquation + glDrawRangeElements + glColorTable + glColorTableParameterfv + glColorTableParameteriv + glCopyColorTable + glGetColorTable + glGetColorTableParameterfv + glGetColorTableParameteriv + glColorSubTable + glCopyColorSubTable + glConvolutionFilter1D + glConvolutionFilter2D + glConvolutionParameterf + glConvolutionParameterfv + glConvolutionParameteri + glConvolutionParameteriv + glCopyConvolutionFilter1D + glCopyConvolutionFilter2D + glGetConvolutionFilter + glGetConvolutionParameterfv + glGetConvolutionParameteriv + glGetSeparableFilter + glSeparableFilter2D + glGetHistogram + glGetHistogramParameterfv + glGetHistogramParameteriv + glGetMinmax + glGetMinmaxParameterfv + glGetMinmaxParameteriv + glHistogram + glMinmax + glResetHistogram + glResetMinmax + glTexImage3D + glTexSubImage3D + glCopyTexSubImage3D + glActiveTextureARB + glClientActiveTextureARB + glMultiTexCoord1dARB + glMultiTexCoord1dvARB + glMultiTexCoord1fARB + glMultiTexCoord1fvARB + glMultiTexCoord1iARB + glMultiTexCoord1ivARB + glMultiTexCoord1sARB + glMultiTexCoord1svARB + glMultiTexCoord2dARB + glMultiTexCoord2dvARB + glMultiTexCoord2fARB + glMultiTexCoord2fvARB + glMultiTexCoord2iARB + glMultiTexCoord2ivARB + glMultiTexCoord2sARB + glMultiTexCoord2svARB + glMultiTexCoord3dARB + glMultiTexCoord3dvARB + glMultiTexCoord3fARB + glMultiTexCoord3fvARB + glMultiTexCoord3iARB + glMultiTexCoord3ivARB + glMultiTexCoord3sARB + glMultiTexCoord3svARB + glMultiTexCoord4dARB + glMultiTexCoord4dvARB + glMultiTexCoord4fARB + glMultiTexCoord4fvARB + glMultiTexCoord4iARB + glMultiTexCoord4ivARB + glMultiTexCoord4sARB + glMultiTexCoord4svARB + glLoadTransposeMatrixfARB + glLoadTransposeMatrixdARB + glMultTransposeMatrixfARB + glMultTransposeMatrixdARB + glSampleCoverageARB + glCompressedTexImage3DARB + glCompressedTexImage2DARB + glCompressedTexImage1DARB + glCompressedTexSubImage3DARB + glCompressedTexSubImage2DARB + glCompressedTexSubImage1DARB + glGetCompressedTexImageARB + glActiveTexture + glClientActiveTexture + glMultiTexCoord1d + glMultiTexCoord1dv + glMultiTexCoord1f + glMultiTexCoord1fv + glMultiTexCoord1i + glMultiTexCoord1iv + glMultiTexCoord1s + glMultiTexCoord1sv + glMultiTexCoord2d + glMultiTexCoord2dv + glMultiTexCoord2f + glMultiTexCoord2fv + glMultiTexCoord2i + glMultiTexCoord2iv + glMultiTexCoord2s + glMultiTexCoord2sv + glMultiTexCoord3d + glMultiTexCoord3dv + glMultiTexCoord3f + glMultiTexCoord3fv + glMultiTexCoord3i + glMultiTexCoord3iv + glMultiTexCoord3s + glMultiTexCoord3sv + glMultiTexCoord4d + glMultiTexCoord4dv + glMultiTexCoord4f + glMultiTexCoord4fv + glMultiTexCoord4i + glMultiTexCoord4iv + glMultiTexCoord4s + glMultiTexCoord4sv + glLoadTransposeMatrixf + glLoadTransposeMatrixd + glMultTransposeMatrixf + glMultTransposeMatrixd + glSampleCoverage + glCompressedTexImage3D + glCompressedTexImage2D + glCompressedTexImage1D + glCompressedTexSubImage3D + glCompressedTexSubImage2D + glCompressedTexSubImage1D + glGetCompressedTexImage + glBlendColorEXT + glPolygonOffsetEXT + glTexImage3DEXT + glTexSubImage3DEXT + glTexSubImage1DEXT + glTexSubImage2DEXT + glCopyTexImage1DEXT + glCopyTexImage2DEXT + glCopyTexSubImage1DEXT + glCopyTexSubImage2DEXT + glCopyTexSubImage3DEXT + glAreTexturesResidentEXT + glBindTextureEXT + glDeleteTexturesEXT + glGenTexturesEXT + glIsTextureEXT + glPrioritizeTexturesEXT + glArrayElementEXT + glColorPointerEXT + glDrawArraysEXT + glEdgeFlagPointerEXT + glGetPointervEXT + glIndexPointerEXT + glNormalPointerEXT + glTexCoordPointerEXT + glVertexPointerEXT + glBlendEquationEXT + glPointParameterfEXT + glPointParameterfvEXT + glPointParameterfARB + glPointParameterfvARB + glColorTableEXT + glGetColorTableEXT + glGetColorTableParameterivEXT + glGetColorTableParameterfvEXT + glLockArraysEXT + glUnlockArraysEXT + glDrawRangeElementsEXT + glSecondaryColor3bEXT + glSecondaryColor3bvEXT + glSecondaryColor3dEXT + glSecondaryColor3dvEXT + glSecondaryColor3fEXT + glSecondaryColor3fvEXT + glSecondaryColor3iEXT + glSecondaryColor3ivEXT + glSecondaryColor3sEXT + glSecondaryColor3svEXT + glSecondaryColor3ubEXT + glSecondaryColor3ubvEXT + glSecondaryColor3uiEXT + glSecondaryColor3uivEXT + glSecondaryColor3usEXT + glSecondaryColor3usvEXT + glSecondaryColorPointerEXT + glMultiDrawArraysEXT + glMultiDrawElementsEXT + glFogCoordfEXT + glFogCoordfvEXT + glFogCoorddEXT + glFogCoorddvEXT + glFogCoordPointerEXT + glBlendFuncSeparateEXT + glFlushVertexArrayRangeNV + glVertexArrayRangeNV + glCombinerParameterfvNV + glCombinerParameterfNV + glCombinerParameterivNV + glCombinerParameteriNV + glCombinerInputNV + glCombinerOutputNV + glFinalCombinerInputNV + glGetCombinerInputParameterfvNV + glGetCombinerInputParameterivNV + glGetCombinerOutputParameterfvNV + glGetCombinerOutputParameterivNV + glGetFinalCombinerInputParameterfvNV + glGetFinalCombinerInputParameterivNV + glResizeBuffersMESA + glWindowPos2dMESA + glWindowPos2dvMESA + glWindowPos2fMESA + glWindowPos2fvMESA + glWindowPos2iMESA + glWindowPos2ivMESA + glWindowPos2sMESA + glWindowPos2svMESA + glWindowPos3dMESA + glWindowPos3dvMESA + glWindowPos3fMESA + glWindowPos3fvMESA + glWindowPos3iMESA + glWindowPos3ivMESA + glWindowPos3sMESA + glWindowPos3svMESA + glWindowPos4dMESA + glWindowPos4dvMESA + glWindowPos4fMESA + glWindowPos4fvMESA + glWindowPos4iMESA + glWindowPos4ivMESA + glWindowPos4sMESA + glWindowPos4svMESA + glWindowPos2dARB + glWindowPos2fARB + glWindowPos2iARB + glWindowPos2sARB + glWindowPos2dvARB + glWindowPos2fvARB + glWindowPos2ivARB + glWindowPos2svARB + glWindowPos3dARB + glWindowPos3fARB + glWindowPos3iARB + glWindowPos3sARB + glWindowPos3dvARB + glWindowPos3fvARB + glWindowPos3ivARB + glWindowPos3svARB + glAreProgramsResidentNV + glBindProgramNV + glDeleteProgramsNV + glExecuteProgramNV + glGenProgramsNV + glGetProgramParameterdvNV + glGetProgramParameterfvNV + glGetProgramivNV + glGetProgramStringNV + glGetTrackMatrixivNV + glGetVertexAttribdvNV + glGetVertexAttribfvNV + glGetVertexAttribivNV + glGetVertexAttribPointervNV + glIsProgramNV + glLoadProgramNV + glProgramParameter4dNV + glProgramParameter4dvNV + glProgramParameter4fNV + glProgramParameter4fvNV + glProgramParameters4dvNV + glProgramParameters4fvNV + glRequestResidentProgramsNV + glTrackMatrixNV + glVertexAttribPointerNV + glVertexAttrib1dNV + glVertexAttrib1dvNV + glVertexAttrib1fNV + glVertexAttrib1fvNV + glVertexAttrib1sNV + glVertexAttrib1svNV + glVertexAttrib2dNV + glVertexAttrib2dvNV + glVertexAttrib2fNV + glVertexAttrib2fvNV + glVertexAttrib2sNV + glVertexAttrib2svNV + glVertexAttrib3dNV + glVertexAttrib3dvNV + glVertexAttrib3fNV + glVertexAttrib3fvNV + glVertexAttrib3sNV + glVertexAttrib3svNV + glVertexAttrib4dNV + glVertexAttrib4dvNV + glVertexAttrib4fNV + glVertexAttrib4fvNV + glVertexAttrib4sNV + glVertexAttrib4svNV + glVertexAttrib4ubNV + glVertexAttrib4ubvNV + glVertexAttribs1dvNV + glVertexAttribs1fvNV + glVertexAttribs1svNV + glVertexAttribs2dvNV + glVertexAttribs2fvNV + glVertexAttribs2svNV + glVertexAttribs3dvNV + glVertexAttribs3fvNV + glVertexAttribs3svNV + glVertexAttribs4dvNV + glVertexAttribs4fvNV + glVertexAttribs4svNV + glVertexAttribs4ubvNV + glPointParameteriNV + glPointParameterivNV + glFogCoordf + glFogCoordfv + glFogCoordd + glFogCoorddv + glFogCoordPointer + glMultiDrawArrays + glMultiDrawElements + glPointParameterf + glPointParameterfv + glPointParameteri + glPointParameteriv + glSecondaryColor3b + glSecondaryColor3bv + glSecondaryColor3d + glSecondaryColor3dv + glSecondaryColor3f + glSecondaryColor3fv + glSecondaryColor3i + glSecondaryColor3iv + glSecondaryColor3s + glSecondaryColor3sv + glSecondaryColor3ub + glSecondaryColor3ubv + glSecondaryColor3ui + glSecondaryColor3uiv + glSecondaryColor3us + glSecondaryColor3usv + glSecondaryColorPointer + glWindowPos2d + glWindowPos2dv + glWindowPos2f + glWindowPos2fv + glWindowPos2i + glWindowPos2iv + glWindowPos2s + glWindowPos2sv + glWindowPos3d + glWindowPos3dv + glWindowPos3f + glWindowPos3fv + glWindowPos3i + glWindowPos3iv + glWindowPos3s + glWindowPos3sv + glVertexAttrib1sARB + glVertexAttrib1fARB + glVertexAttrib1dARB + glVertexAttrib2sARB + glVertexAttrib2fARB + glVertexAttrib2dARB + glVertexAttrib3sARB + glVertexAttrib3fARB + glVertexAttrib3dARB + glVertexAttrib4sARB + glVertexAttrib4fARB + glVertexAttrib4dARB + glVertexAttrib4NubARB + glVertexAttrib1svARB + glVertexAttrib1fvARB + glVertexAttrib1dvARB + glVertexAttrib2svARB + glVertexAttrib2fvARB + glVertexAttrib2dvARB + glVertexAttrib3svARB + glVertexAttrib3fvARB + glVertexAttrib3dvARB + glVertexAttrib4bvARB + glVertexAttrib4svARB + glVertexAttrib4ivARB + glVertexAttrib4ubvARB + glVertexAttrib4usvARB + glVertexAttrib4uivARB + glVertexAttrib4fvARB + glVertexAttrib4dvARB + glVertexAttrib4NbvARB + glVertexAttrib4NsvARB + glVertexAttrib4NivARB + glVertexAttrib4NubvARB + glVertexAttrib4NusvARB + glVertexAttrib4NuivARB + glVertexAttribPointerARB + glEnableVertexAttribArrayARB + glDisableVertexAttribArrayARB + glProgramStringARB + glBindProgramARB + glDeleteProgramsARB + glGenProgramsARB + glIsProgramARB + glProgramEnvParameter4dARB + glProgramEnvParameter4dvARB + glProgramEnvParameter4fARB + glProgramEnvParameter4fvARB + glProgramLocalParameter4dARB + glProgramLocalParameter4dvARB + glProgramLocalParameter4fARB + glProgramLocalParameter4fvARB + glGetProgramEnvParameterdvARB + glGetProgramEnvParameterfvARB + glGetProgramLocalParameterdvARB + glGetProgramLocalParameterfvARB + glGetProgramivARB + glGetProgramStringARB + glGetVertexAttribdvARB + glGetVertexAttribfvARB + glGetVertexAttribivARB + glGetVertexAttribPointervARB + glProgramNamedParameter4fNV + glProgramNamedParameter4dNV + glProgramNamedParameter4fvNV + glProgramNamedParameter4dvNV + glGetProgramNamedParameterfvNV + glGetProgramNamedParameterdvNV + glBindBufferARB + glBufferDataARB + glBufferSubDataARB + glDeleteBuffersARB + glGenBuffersARB + glGetBufferParameterivARB + glGetBufferPointervARB + glGetBufferSubDataARB + glIsBufferARB + glMapBufferARB + glUnmapBufferARB + glGenQueriesARB + glDeleteQueriesARB + glIsQueryARB + glBeginQueryARB + glEndQueryARB + glGetQueryivARB + glGetQueryObjectivARB + glGetQueryObjectuivARB + glBindBuffer + glBufferData + glBufferSubData + glDeleteBuffers + glGenBuffers + glGetBufferParameteriv + glGetBufferPointerv + glGetBufferSubData + glIsBuffer + glMapBuffer + glUnmapBuffer + glGenQueries + glDeleteQueries + glIsQuery + glBeginQuery + glEndQuery + glGetQueryiv + glGetQueryObjectiv + glGetQueryObjectuiv +; +; WGL API + wglChoosePixelFormat + wglCopyContext + wglCreateContext + wglCreateLayerContext + wglDeleteContext + wglDescribeLayerPlane + wglDescribePixelFormat + wglGetCurrentContext + wglGetCurrentDC + wglGetLayerPaletteEntries + wglGetPixelFormat + wglGetProcAddress + wglMakeCurrent + wglRealizeLayerPalette + wglSetLayerPaletteEntries + wglSetPixelFormat + wglShareLists + wglSwapBuffers + wglSwapLayerBuffers + wglUseFontBitmapsA + wglUseFontBitmapsW + wglUseFontOutlinesA + wglUseFontOutlinesW + wglGetExtensionsStringARB diff --git a/src/gallium/winsys/gdi/wgl.c b/src/gallium/winsys/gdi/wgl.c new file mode 100644 index 0000000000..50126014a8 --- /dev/null +++ b/src/gallium/winsys/gdi/wgl.c @@ -0,0 +1,703 @@ +/* $Id: wgl.c,v 1.12 2006/03/30 07:58:24 kschultz Exp $ */ + +/* + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Library General Public + * License as published by the Free Software Foundation; either + * version 2 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Library General Public License for more details. + * + * You should have received a copy of the GNU Library General Public + * License along with this library; if not, write to the Free + * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * + */ + +/* + * File name : wgl.c + * WGL stuff. Added by Oleg Letsinsky, ajl@ultersys.ru + * Some things originated from the 3Dfx WGL functions + */ + +/* + * This file contains the implementation of the wgl* functions for + * Mesa on Windows. Since these functions are provided by Windows in + * GDI/OpenGL, we must supply our versions that work with Mesa here. + */ + + +/* We're essentially building part of GDI here, so define this so that + * we get the right export linkage. */ +#ifdef __MINGW32__ + +#include +#include +#include +#include + +# if defined(BUILD_GL32) +# define WINGDIAPI __declspec(dllexport) +# else +# define __W32API_USE_DLLIMPORT__ +# endif + +#include +#include "GL/mesa_wgl.h" +#include + +#else + +#define _GDI32_ +#include + +#endif + +#include "glapi.h" +#include "GL/wmesa.h" /* protos for wmesa* functions */ + +/* + * Pixel Format Descriptors + */ + +/* Extend the PFD to include DB flag */ +struct __pixelformat__ +{ + PIXELFORMATDESCRIPTOR pfd; + GLboolean doubleBuffered; +}; + +/* These are the PFD's supported by this driver. */ +struct __pixelformat__ pfd[] = +{ + /* Double Buffer, alpha */ + { + { + sizeof(PIXELFORMATDESCRIPTOR), 1, + PFD_DRAW_TO_WINDOW|PFD_SUPPORT_OPENGL| + PFD_GENERIC_FORMAT|PFD_DOUBLEBUFFER|PFD_SWAP_COPY, + PFD_TYPE_RGBA, + 24, + 8, 0, + 8, 8, + 8, 16, + 8, 24, + 0, 0, 0, 0, 0, + 16, 8, + 0, 0, 0, + 0, 0, 0 + }, + GL_TRUE + }, + /* Single Buffer, alpha */ + { + { + sizeof(PIXELFORMATDESCRIPTOR), 1, + PFD_DRAW_TO_WINDOW|PFD_SUPPORT_OPENGL| + PFD_GENERIC_FORMAT, + PFD_TYPE_RGBA, + 24, + 8, 0, + 8, 8, + 8, 16, + 8, 24, + 0, 0, 0, 0, 0, + 16, 8, + 0, 0, 0, + 0, 0, 0 + }, + GL_FALSE + }, + /* Double Buffer, no alpha */ + { + { + sizeof(PIXELFORMATDESCRIPTOR), 1, + PFD_DRAW_TO_WINDOW|PFD_SUPPORT_OPENGL| + PFD_GENERIC_FORMAT|PFD_DOUBLEBUFFER|PFD_SWAP_COPY, + PFD_TYPE_RGBA, + 24, + 8, 0, + 8, 8, + 8, 16, + 0, 0, + 0, 0, 0, 0, 0, + 16, 8, + 0, 0, 0, + 0, 0, 0 + }, + GL_TRUE + }, + /* Single Buffer, no alpha */ + { + { + sizeof(PIXELFORMATDESCRIPTOR), 1, + PFD_DRAW_TO_WINDOW|PFD_SUPPORT_OPENGL| + PFD_GENERIC_FORMAT, + PFD_TYPE_RGBA, + 24, + 8, 0, + 8, 8, + 8, 16, + 0, 0, + 0, 0, 0, 0, 0, + 16, 8, + 0, 0, 0, + 0, 0, 0 + }, + GL_FALSE + }, +}; + +int npfd = sizeof(pfd) / sizeof(pfd[0]); + + +/* + * Contexts + */ + +typedef struct { + WMesaContext ctx; +} MesaWglCtx; + +#define MESAWGL_CTX_MAX_COUNT 20 + +static MesaWglCtx wgl_ctx[MESAWGL_CTX_MAX_COUNT]; + +static unsigned ctx_count = 0; +static int ctx_current = -1; +static unsigned curPFD = 0; + +static HDC CurrentHDC = 0; + + +WINGDIAPI HGLRC GLAPIENTRY wglCreateContext(HDC hdc) +{ + int i = 0; + if (!ctx_count) { + for(i=0;inSize != sizeof(PIXELFORMATDESCRIPTOR) || ppfd->nVersion != 1) + { + SetLastError(0); + return(0); + } + for(i = 0; i < npfd;i++) + { + delta = 0; + if( + (ppfd->dwFlags & PFD_DRAW_TO_WINDOW) && + !(pfd[i].pfd.dwFlags & PFD_DRAW_TO_WINDOW)) + continue; + if( + (ppfd->dwFlags & PFD_DRAW_TO_BITMAP) && + !(pfd[i].pfd.dwFlags & PFD_DRAW_TO_BITMAP)) + continue; + if( + (ppfd->dwFlags & PFD_SUPPORT_GDI) && + !(pfd[i].pfd.dwFlags & PFD_SUPPORT_GDI)) + continue; + if( + (ppfd->dwFlags & PFD_SUPPORT_OPENGL) && + !(pfd[i].pfd.dwFlags & PFD_SUPPORT_OPENGL)) + continue; + if( + !(ppfd->dwFlags & PFD_DOUBLEBUFFER_DONTCARE) && + ((ppfd->dwFlags & PFD_DOUBLEBUFFER) != + (pfd[i].pfd.dwFlags & PFD_DOUBLEBUFFER))) + continue; + if( + !(ppfd->dwFlags & PFD_STEREO_DONTCARE) && + ((ppfd->dwFlags & PFD_STEREO) != + (pfd[i].pfd.dwFlags & PFD_STEREO))) + continue; + if(ppfd->iPixelType != pfd[i].pfd.iPixelType) + delta++; + if(ppfd->cAlphaBits != pfd[i].pfd.cAlphaBits) + delta++; + if(delta < bestdelta) + { + best = i + 1; + bestdelta = delta; + if(bestdelta == 0) + break; + } + } + if(best == -1) + { + SetLastError(0); + return(0); + } + return(best); +} + +WINGDIAPI int GLAPIENTRY wglDescribePixelFormat(HDC hdc, + int iPixelFormat, + UINT nBytes, + LPPIXELFORMATDESCRIPTOR ppfd) +{ + (void) hdc; + + if(ppfd == NULL) + return(npfd); + if(iPixelFormat < 1 || iPixelFormat > npfd || + nBytes != sizeof(PIXELFORMATDESCRIPTOR)) + { + SetLastError(0); + return(0); + } + *ppfd = pfd[iPixelFormat - 1].pfd; + return(npfd); +} + +WINGDIAPI PROC GLAPIENTRY wglGetProcAddress(LPCSTR lpszProc) +{ + PROC p = (PROC) _glapi_get_proc_address((const char *) lpszProc); + if (p) + return p; + + SetLastError(0); + return(NULL); +} + +WINGDIAPI int GLAPIENTRY wglGetPixelFormat(HDC hdc) +{ + (void) hdc; + if(curPFD == 0) { + SetLastError(0); + return(0); + } + return(curPFD); +} + +WINGDIAPI BOOL GLAPIENTRY wglSetPixelFormat(HDC hdc,int iPixelFormat, + const PIXELFORMATDESCRIPTOR *ppfd) +{ + (void) hdc; + + if(iPixelFormat < 1 || iPixelFormat > npfd || + ppfd->nSize != sizeof(PIXELFORMATDESCRIPTOR)) { + SetLastError(0); + return(FALSE); + } + curPFD = iPixelFormat; + return(TRUE); +} + +WINGDIAPI BOOL GLAPIENTRY wglSwapBuffers(HDC hdc) +{ + WMesaSwapBuffers(hdc); + return TRUE; +} + +static FIXED FixedFromDouble(double d) +{ + long l = (long) (d * 65536L); + return *(FIXED *) (void *) &l; +} + + +/* +** This is cribbed from FX/fxwgl.c, and seems to implement support +** for bitmap fonts where the wglUseFontBitmapsA() code implements +** support for outline fonts. In combination they hopefully give +** fairly generic support for fonts. +*/ +static BOOL wglUseFontBitmaps_FX(HDC fontDevice, DWORD firstChar, + DWORD numChars, DWORD listBase) +{ +#define VERIFY(a) a + + TEXTMETRIC metric; + BITMAPINFO *dibInfo; + HDC bitDevice; + COLORREF tempColor; + int i; + + VERIFY(GetTextMetrics(fontDevice, &metric)); + + dibInfo = (BITMAPINFO *) calloc(sizeof(BITMAPINFO) + sizeof(RGBQUAD), 1); + dibInfo->bmiHeader.biSize = sizeof(BITMAPINFOHEADER); + dibInfo->bmiHeader.biPlanes = 1; + dibInfo->bmiHeader.biBitCount = 1; + dibInfo->bmiHeader.biCompression = BI_RGB; + + bitDevice = CreateCompatibleDC(fontDevice); + + /* Swap fore and back colors so the bitmap has the right polarity */ + tempColor = GetBkColor(bitDevice); + SetBkColor(bitDevice, GetTextColor(bitDevice)); + SetTextColor(bitDevice, tempColor); + + /* Place chars based on base line */ + VERIFY(SetTextAlign(bitDevice, TA_BASELINE) != GDI_ERROR ? 1 : 0); + + for(i = 0; i < (int)numChars; i++) { + SIZE size; + char curChar; + int charWidth,charHeight,bmapWidth,bmapHeight,numBytes,res; + HBITMAP bitObject; + HGDIOBJ origBmap; + unsigned char *bmap; + + curChar = (char)(i + firstChar); + + /* Find how high/wide this character is */ + VERIFY(GetTextExtentPoint32(bitDevice, (LPCWSTR)&curChar, 1, &size)); + + /* Create the output bitmap */ + charWidth = size.cx; + charHeight = size.cy; + /* Round up to the next multiple of 32 bits */ + bmapWidth = ((charWidth + 31) / 32) * 32; + bmapHeight = charHeight; + bitObject = CreateCompatibleBitmap(bitDevice, + bmapWidth, + bmapHeight); + /* VERIFY(bitObject); */ + + /* Assign the output bitmap to the device */ + origBmap = SelectObject(bitDevice, bitObject); + (void) VERIFY(origBmap); + + VERIFY( PatBlt( bitDevice, 0, 0, bmapWidth, bmapHeight,BLACKNESS ) ); + + /* Use our source font on the device */ + VERIFY(SelectObject(bitDevice, GetCurrentObject(fontDevice,OBJ_FONT))); + + /* Draw the character */ + VERIFY(TextOut(bitDevice, 0, metric.tmAscent, (LPCWSTR)&curChar, 1)); + + /* Unselect our bmap object */ + VERIFY(SelectObject(bitDevice, origBmap)); + + /* Convert the display dependant representation to a 1 bit deep DIB */ + numBytes = (bmapWidth * bmapHeight) / 8; + bmap = (unsigned char *)malloc(numBytes); + dibInfo->bmiHeader.biWidth = bmapWidth; + dibInfo->bmiHeader.biHeight = bmapHeight; + res = GetDIBits(bitDevice, bitObject, 0, bmapHeight, bmap, + dibInfo, + DIB_RGB_COLORS); + /* VERIFY(res); */ + + /* Create the GL object */ + glNewList(i + listBase, GL_COMPILE); + glBitmap(bmapWidth, bmapHeight, 0.0, (GLfloat)metric.tmDescent, + (GLfloat)charWidth, 0.0, + bmap); + glEndList(); + /* CheckGL(); */ + + /* Destroy the bmap object */ + DeleteObject(bitObject); + + /* Deallocate the bitmap data */ + free(bmap); + } + + /* Destroy the DC */ + VERIFY(DeleteDC(bitDevice)); + + free(dibInfo); + + return TRUE; +#undef VERIFY +} + +WINGDIAPI BOOL GLAPIENTRY wglUseFontBitmapsA(HDC hdc, DWORD first, + DWORD count, DWORD listBase) +{ + int i; + GLuint font_list; + DWORD size; + GLYPHMETRICS gm; + HANDLE hBits; + LPSTR lpBits; + MAT2 mat; + int success = TRUE; + + if (count == 0) + return FALSE; + + font_list = listBase; + + mat.eM11 = FixedFromDouble(1); + mat.eM12 = FixedFromDouble(0); + mat.eM21 = FixedFromDouble(0); + mat.eM22 = FixedFromDouble(-1); + + memset(&gm,0,sizeof(gm)); + + /* + ** If we can't get the glyph outline, it may be because this is a fixed + ** font. Try processing it that way. + */ + if( GetGlyphOutline(hdc, first, GGO_BITMAP, &gm, 0, NULL, &mat) + == GDI_ERROR ) { + return wglUseFontBitmaps_FX( hdc, first, count, listBase ); + } + + /* + ** Otherwise process all desired characters. + */ + for (i = 0; i < (int)count; i++) { + DWORD err; + + glNewList( font_list+i, GL_COMPILE ); + + /* allocate space for the bitmap/outline */ + size = GetGlyphOutline(hdc, first + i, GGO_BITMAP, + &gm, 0, NULL, &mat); + if (size == GDI_ERROR) { + glEndList( ); + err = GetLastError(); + success = FALSE; + continue; + } + + hBits = GlobalAlloc(GHND, size+1); + lpBits = GlobalLock(hBits); + + err = + GetGlyphOutline(hdc, /* handle to device context */ + first + i, /* character to query */ + GGO_BITMAP, /* format of data to return */ + &gm, /* ptr to structure for metrics*/ + size, /* size of buffer for data */ + lpBits, /* pointer to buffer for data */ + &mat /* pointer to transformation */ + /* matrix structure */ + ); + + if (err == GDI_ERROR) { + GlobalUnlock(hBits); + GlobalFree(hBits); + + glEndList( ); + err = GetLastError(); + success = FALSE; + continue; + } + + glBitmap(gm.gmBlackBoxX,gm.gmBlackBoxY, + (GLfloat)-gm.gmptGlyphOrigin.x, + (GLfloat)gm.gmptGlyphOrigin.y, + (GLfloat)gm.gmCellIncX, + (GLfloat)gm.gmCellIncY, + (const GLubyte * )lpBits); + + GlobalUnlock(hBits); + GlobalFree(hBits); + + glEndList( ); + } + + return success; +} + + + +/* NOT IMPLEMENTED YET */ +WINGDIAPI BOOL GLAPIENTRY wglCopyContext(HGLRC hglrcSrc, + HGLRC hglrcDst, + UINT mask) +{ + (void) hglrcSrc; (void) hglrcDst; (void) mask; + return(FALSE); +} + +WINGDIAPI HGLRC GLAPIENTRY wglCreateLayerContext(HDC hdc, + int iLayerPlane) +{ + (void) hdc; (void) iLayerPlane; + SetLastError(0); + return(NULL); +} + +WINGDIAPI BOOL GLAPIENTRY wglShareLists(HGLRC hglrc1, + HGLRC hglrc2) +{ + (void) hglrc1; (void) hglrc2; + return(TRUE); +} + + +WINGDIAPI BOOL GLAPIENTRY wglUseFontBitmapsW(HDC hdc, + DWORD first, + DWORD count, + DWORD listBase) +{ + (void) hdc; (void) first; (void) count; (void) listBase; + return FALSE; +} + +WINGDIAPI BOOL GLAPIENTRY wglUseFontOutlinesA(HDC hdc, + DWORD first, + DWORD count, + DWORD listBase, + FLOAT deviation, + FLOAT extrusion, + int format, + LPGLYPHMETRICSFLOAT lpgmf) +{ + (void) hdc; (void) first; (void) count; + (void) listBase; (void) deviation; (void) extrusion; (void) format; + (void) lpgmf; + SetLastError(0); + return(FALSE); +} + +WINGDIAPI BOOL GLAPIENTRY wglUseFontOutlinesW(HDC hdc, + DWORD first, + DWORD count, + DWORD listBase, + FLOAT deviation, + FLOAT extrusion, + int format, + LPGLYPHMETRICSFLOAT lpgmf) +{ + (void) hdc; (void) first; (void) count; + (void) listBase; (void) deviation; (void) extrusion; (void) format; + (void) lpgmf; + SetLastError(0); + return(FALSE); +} + +WINGDIAPI BOOL GLAPIENTRY wglDescribeLayerPlane(HDC hdc, + int iPixelFormat, + int iLayerPlane, + UINT nBytes, + LPLAYERPLANEDESCRIPTOR plpd) +{ + (void) hdc; (void) iPixelFormat; (void) iLayerPlane; + (void) nBytes; (void) plpd; + SetLastError(0); + return(FALSE); +} + +WINGDIAPI int GLAPIENTRY wglSetLayerPaletteEntries(HDC hdc, + int iLayerPlane, + int iStart, + int cEntries, + CONST COLORREF *pcr) +{ + (void) hdc; (void) iLayerPlane; (void) iStart; + (void) cEntries; (void) pcr; + SetLastError(0); + return(0); +} + +WINGDIAPI int GLAPIENTRY wglGetLayerPaletteEntries(HDC hdc, + int iLayerPlane, + int iStart, + int cEntries, + COLORREF *pcr) +{ + (void) hdc; (void) iLayerPlane; (void) iStart; (void) cEntries; (void) pcr; + SetLastError(0); + return(0); +} + +WINGDIAPI BOOL GLAPIENTRY wglRealizeLayerPalette(HDC hdc, + int iLayerPlane, + BOOL bRealize) +{ + (void) hdc; (void) iLayerPlane; (void) bRealize; + SetLastError(0); + return(FALSE); +} + +WINGDIAPI BOOL GLAPIENTRY wglSwapLayerBuffers(HDC hdc, + UINT fuPlanes) +{ + (void) hdc; (void) fuPlanes; + SetLastError(0); + return(FALSE); +} + +WINGDIAPI const char * GLAPIENTRY wglGetExtensionsStringARB(HDC hdc) +{ + return "WGL_ARB_extensions_string"; +} diff --git a/src/gallium/winsys/gdi/wmesa.c b/src/gallium/winsys/gdi/wmesa.c new file mode 100644 index 0000000000..74a8292352 --- /dev/null +++ b/src/gallium/winsys/gdi/wmesa.c @@ -0,0 +1,890 @@ +/* + * Windows (Win32/Win64) device driver for Mesa + * + */ + +#include "mtypes.h" +#include +#include "wmesadef.h" + +#undef Elements + +#include "pipe/p_winsys.h" +#include "pipe/p_format.h" +#include "pipe/p_context.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "softpipe/sp_winsys.h" +#include "glapi/glapi.h" +#include "colors.h" + +extern GLvisual * +_mesa_create_visual( GLboolean rgbFlag, + GLboolean dbFlag, + GLboolean stereoFlag, + GLint redBits, + GLint greenBits, + GLint blueBits, + GLint alphaBits, + GLint indexBits, + GLint depthBits, + GLint stencilBits, + GLint accumRedBits, + GLint accumGreenBits, + GLint accumBlueBits, + GLint accumAlphaBits, + GLint numSamples ); + +/* linked list of our Framebuffers (windows) */ +WMesaFramebuffer FirstFramebuffer = NULL; + +struct wmesa_pipe_winsys +{ + struct pipe_winsys base; +}; + +/** + * Choose the pixel format for the given visual. + * This will tell the gallium driver how to pack pixel data into + * drawing surfaces. + */ +static GLuint +choose_pixel_format(GLvisual *v) +{ +#if 1 + return PIPE_FORMAT_A8R8G8B8_UNORM; +#else + if ( GET_REDMASK(v) == 0x0000ff + && GET_GREENMASK(v) == 0x00ff00 + && GET_BLUEMASK(v) == 0xff0000 + && v->BitsPerPixel == 32) { + if (CHECK_BYTE_ORDER(v)) { + /* no byteswapping needed */ + return 0 /* PIXEL_FORMAT_U_A8_B8_G8_R8 */; + } + else { + return PIPE_FORMAT_R8G8B8A8_UNORM; + } + } + else if ( GET_REDMASK(v) == 0xff0000 + && GET_GREENMASK(v) == 0x00ff00 + && GET_BLUEMASK(v) == 0x0000ff + && v->BitsPerPixel == 32) { + if (CHECK_BYTE_ORDER(v)) { + /* no byteswapping needed */ + return PIPE_FORMAT_A8R8G8B8_UNORM; + } + else { + return PIPE_FORMAT_B8G8R8A8_UNORM; + } + } + else if ( GET_REDMASK(v) == 0xf800 + && GET_GREENMASK(v) == 0x07e0 + && GET_BLUEMASK(v) == 0x001f + && CHECK_BYTE_ORDER(v) + && v->BitsPerPixel == 16) { + /* 5-6-5 RGB */ + return PIPE_FORMAT_R5G6B5_UNORM; + } + +printf("BITS %d\n",v->BitsPerPixel); + assert(0); + return 0; +#endif +} + +/* + * Determine the pixel format based on the pixel size. + */ +static void wmSetPixelFormat(WMesaFramebuffer pwfb, HDC hDC) +{ + /* Only 16 and 32 bit targets are supported now */ + assert(pwfb->cColorBits == 0 || + pwfb->cColorBits == 16 || + pwfb->cColorBits == 32); + + switch(pwfb->cColorBits){ + case 8: + pwfb->pixelformat = PF_INDEX8; + break; + case 16: + pwfb->pixelformat = PF_5R6G5B; + break; + case 32: + pwfb->pixelformat = PF_8R8G8B; + break; + default: + pwfb->pixelformat = PF_BADFORMAT; + } +} + + +/** + * Create DIB for back buffer. + * We write into this memory with the span routines and then blit it + * to the window on a buffer swap. + */ +BOOL wmCreateBackingStore(WMesaFramebuffer pwfb, long lxSize, long lySize) +{ + HDC hdc = pwfb->hDC; + BITMAPINFO bmi; + LPBITMAPINFO pbmi = &bmi; + HDC hic; + + pbmi->bmiHeader.biSize = sizeof(BITMAPINFOHEADER); + pbmi->bmiHeader.biWidth = lxSize; + pbmi->bmiHeader.biHeight= -lySize; + pbmi->bmiHeader.biPlanes = 1; + pbmi->bmiHeader.biBitCount = pwfb->cColorBits; + pbmi->bmiHeader.biCompression = BI_RGB; + pbmi->bmiHeader.biSizeImage = 0; + pbmi->bmiHeader.biXPelsPerMeter = 0; + pbmi->bmiHeader.biYPelsPerMeter = 0; + pbmi->bmiHeader.biClrUsed = 0; + pbmi->bmiHeader.biClrImportant = 0; + + hic = CreateIC("display", NULL, NULL, NULL); + pwfb->dib_hDC = CreateCompatibleDC(hic); + + pwfb->hbmDIB = CreateDIBSection(hic, + pbmi, + DIB_RGB_COLORS, + (void **)&(pwfb->pbPixels), + 0, + 0); + pwfb->hOldBitmap = SelectObject(pwfb->dib_hDC, pwfb->hbmDIB); + + DeleteDC(hic); + + wmSetPixelFormat(pwfb, pwfb->hDC); + return TRUE; +} + +/** + * Create a new WMesaFramebuffer object which will correspond to the + * given HDC (Window handle). + */ +WMesaFramebuffer +wmesa_new_framebuffer(HDC hdc, GLvisual *visual, GLuint width, GLuint height) +{ + WMesaFramebuffer pwfb + = (WMesaFramebuffer) malloc(sizeof(struct wmesa_framebuffer)); + if (pwfb) { + enum pipe_format colorFormat, depthFormat, stencilFormat; + + /* determine PIPE_FORMATs for buffers */ + colorFormat = choose_pixel_format(visual); + + if (visual->depthBits == 0) + depthFormat = PIPE_FORMAT_NONE; + else if (visual->depthBits <= 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (visual->depthBits <= 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_Z32_UNORM; + + if (visual->stencilBits == 8) { + if (depthFormat == PIPE_FORMAT_S8Z24_UNORM) + stencilFormat = depthFormat; + else + stencilFormat = PIPE_FORMAT_S8_UNORM; + } + else { + stencilFormat = PIPE_FORMAT_NONE; + } + + pwfb->stfb = st_create_framebuffer(visual, + colorFormat, depthFormat, stencilFormat, + width, height, + (void *) pwfb); + + pwfb->cColorBits = GetDeviceCaps(hdc, BITSPIXEL); + +#if 0 + wmCreateBackingStore(pwfb, width, height); +#endif + + pwfb->hDC = hdc; + /* insert at head of list */ + pwfb->next = FirstFramebuffer; + FirstFramebuffer = pwfb; + } + return pwfb; +} + +/** + * Given an hdc, free the corresponding WMesaFramebuffer + */ +void +wmesa_free_framebuffer(HDC hdc) +{ + WMesaFramebuffer pwfb, prev; + for (pwfb = FirstFramebuffer; pwfb; pwfb = pwfb->next) { + if (pwfb->hDC == hdc) + break; + prev = pwfb; + } + if (pwfb) { + if (pwfb == FirstFramebuffer) + FirstFramebuffer = pwfb->next; + else + prev->next = pwfb->next; + free(pwfb); + } +} + +/** + * Given an hdc, return the corresponding WMesaFramebuffer + */ +WMesaFramebuffer +wmesa_lookup_framebuffer(HDC hdc) +{ + WMesaFramebuffer pwfb; + for (pwfb = FirstFramebuffer; pwfb; pwfb = pwfb->next) { + if (pwfb->hDC == hdc) + return pwfb; + } + return NULL; +} + + +/** + * Given a GLframebuffer, return the corresponding WMesaFramebuffer. + */ +static WMesaFramebuffer wmesa_framebuffer(GLframebuffer *fb) +{ + return (WMesaFramebuffer) fb; +} + + +/** + * Given a GLcontext, return the corresponding WMesaContext. + */ +static WMesaContext wmesa_context(const GLcontext *ctx) +{ + return (WMesaContext) ctx; +} + +static wmDeleteBackingStore(WMesaFramebuffer pwfb) +{ + if (pwfb->hbmDIB) { + SelectObject(pwfb->dib_hDC, pwfb->hOldBitmap); + DeleteDC(pwfb->dib_hDC); + DeleteObject(pwfb->hbmDIB); + } +} + + +/** + * Find the width and height of the window named by hdc. + */ +static void +get_window_size(HDC hdc, GLuint *width, GLuint *height) +{ + if (WindowFromDC(hdc)) { + RECT rect; + GetClientRect(WindowFromDC(hdc), &rect); + *width = rect.right - rect.left; + *height = rect.bottom - rect.top; + } + else { /* Memory context */ + /* From contributed code - use the size of the desktop + * for the size of a memory context (?) */ + *width = GetDeviceCaps(hdc, HORZRES); + *height = GetDeviceCaps(hdc, VERTRES); + } +} + +/** + * Low-level OS/window system memory buffer + */ +struct wm_buffer +{ + struct pipe_buffer base; + boolean userBuffer; /** Is this a user-space buffer? */ + void *data; + void *mapped; +}; + +struct wmesa_surface +{ + struct pipe_surface surface; + + int no_swap; +}; + + +/** Cast wrapper */ +static INLINE struct wmesa_surface * +wmesa_surface(struct pipe_surface *ps) +{ +// assert(0); + return (struct wmesa_surface *) ps; +} + +/** + * Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque + * buffer pointer... + */ +static INLINE struct wm_buffer * +wm_buffer( struct pipe_buffer *buf ) +{ + return (struct wm_buffer *)buf; +} + + + +/* Most callbacks map direcly onto dri_bufmgr operations: + */ +static void * +wm_buffer_map(struct pipe_winsys *pws, struct pipe_buffer *buf, + unsigned flags) +{ + struct wm_buffer *wm_buf = wm_buffer(buf); + wm_buf->mapped = wm_buf->data; + return wm_buf->mapped; +} + +static void +wm_buffer_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf) +{ + struct wm_buffer *wm_buf = wm_buffer(buf); + wm_buf->mapped = NULL; +} + +static void +wm_buffer_destroy(struct pipe_winsys *pws, + struct pipe_buffer *buf) +{ + struct wm_buffer *oldBuf = wm_buffer(buf); + + if (oldBuf->data) { + { + if (!oldBuf->userBuffer) { + align_free(oldBuf->data); + } + } + + oldBuf->data = NULL; + } + + free(oldBuf); +} + + +static void +wm_flush_frontbuffer(struct pipe_winsys *pws, + struct pipe_surface *surf, + void *context_private) +{ + WMesaContext pwc = context_private; + WMesaFramebuffer pwfb = wmesa_lookup_framebuffer(pwc->hDC); + struct wm_buffer *wm_buf; + BITMAPINFO bmi, *pbmi; + +#if 0 + if (pwfb) + BitBlt(pwfb->hDC, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, + pwfb->dib_hDC, 0, 0, SRCCOPY); +#else + wm_buf = wm_buffer(surf->buffer); + + pbmi = &bmi; + memset(pbmi, 0, sizeof(BITMAPINFO)); + pbmi->bmiHeader.biSize = sizeof(BITMAPINFOHEADER); + pbmi->bmiHeader.biWidth = pwfb->stfb->Base.Width; + pbmi->bmiHeader.biHeight= -((long)pwfb->stfb->Base.Height); + pbmi->bmiHeader.biPlanes = 1; + pbmi->bmiHeader.biBitCount = pwfb->cColorBits; + pbmi->bmiHeader.biCompression = BI_RGB; + pbmi->bmiHeader.biSizeImage = 0; + pbmi->bmiHeader.biXPelsPerMeter = 0; + pbmi->bmiHeader.biYPelsPerMeter = 0; + pbmi->bmiHeader.biClrUsed = 0; + pbmi->bmiHeader.biClrImportant = 0; + + StretchDIBits(pwfb->hDC, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, wm_buf->data, pbmi, 0, SRCCOPY); +#endif +} + + + +static const char * +wm_get_name(struct pipe_winsys *pws) +{ + return "gdi"; +} + +static struct pipe_buffer * +wm_buffer_create(struct pipe_winsys *pws, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct wm_buffer *buffer = CALLOC_STRUCT(wm_buffer); + + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + + if (buffer->data == NULL) { + /* align to 16-byte multiple for Cell */ + buffer->data = align_malloc(size, max(alignment, 16)); + } + + return &buffer->base; +} + + +/** + * Create buffer which wraps user-space data. + */ +static struct pipe_buffer * +wm_user_buffer_create(struct pipe_winsys *pws, void *ptr, unsigned bytes) +{ + struct wm_buffer *buffer = CALLOC_STRUCT(wm_buffer); + buffer->base.refcount = 1; + buffer->base.size = bytes; + buffer->userBuffer = TRUE; + buffer->data = ptr; + + return &buffer->base; +} + + + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + +static int +wm_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags) +{ + const unsigned alignment = 64; + + surf->width = width; + surf->height = height; + surf->format = format; + surf->cpp = pf_get_size(format); + surf->pitch = round_up(width, alignment / surf->cpp); + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->pitch * surf->cpp * height); + if(!surf->buffer) + return -1; + + return 0; +} + + +/** + * Called via winsys->surface_alloc() to create new surfaces. + */ +static struct pipe_surface * +wm_surface_alloc(struct pipe_winsys *ws) +{ + struct wmesa_surface *wms = CALLOC_STRUCT(wmesa_surface); + static boolean no_swap = 0; + static boolean firsttime = 1; + + if (firsttime) { + no_swap = getenv("SP_NO_RAST") != NULL; + firsttime = 0; + } + + assert(ws); + + wms->surface.refcount = 1; + wms->surface.winsys = ws; + + wms->no_swap = no_swap; + + return &wms->surface; +} + +static void +wm_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + surf->refcount--; + if (surf->refcount == 0) { + if (surf->buffer) + pipe_buffer_reference(winsys, &surf->buffer, NULL); + free(surf); + } + *s = NULL; +} + + +/* + * Fence functions - basically nothing to do, as we don't create any actual + * fence objects. + */ + +static void +wm_fence_reference(struct pipe_winsys *sws, struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) +{ +} + + +static int +wm_fence_signalled(struct pipe_winsys *sws, struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + + +static int +wm_fence_finish(struct pipe_winsys *sws, struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + + + +struct pipe_winsys * +wmesa_get_pipe_winsys(GLvisual *visual) +{ + static struct wmesa_pipe_winsys *ws = NULL; + + if (!ws) { + ws = CALLOC_STRUCT(wmesa_pipe_winsys); + + /* Fill in this struct with callbacks that pipe will need to + * communicate with the window system, buffer manager, etc. + */ + ws->base.buffer_create = wm_buffer_create; + ws->base.user_buffer_create = wm_user_buffer_create; + ws->base.buffer_map = wm_buffer_map; + ws->base.buffer_unmap = wm_buffer_unmap; + ws->base.buffer_destroy = wm_buffer_destroy; + + ws->base.surface_alloc = wm_surface_alloc; + ws->base.surface_alloc_storage = wm_surface_alloc_storage; + ws->base.surface_release = wm_surface_release; + + ws->base.fence_reference = wm_fence_reference; + ws->base.fence_signalled = wm_fence_signalled; + ws->base.fence_finish = wm_fence_finish; + + ws->base.flush_frontbuffer = wm_flush_frontbuffer; + ws->base.get_name = wm_get_name; + } + + return &ws->base; +} + + + +/**********************************************************************/ +/***** WMESA Functions *****/ +/**********************************************************************/ + +WMesaContext WMesaCreateContext(HDC hDC, + HPALETTE* Pal, + GLboolean rgb_flag, + GLboolean db_flag, + GLboolean alpha_flag) +{ + WMesaContext c; + struct pipe_winsys *pws; + struct pipe_context *pipe; + struct pipe_screen *screen; + GLint red_bits, green_bits, blue_bits, alpha_bits; + GLvisual *visual; + + (void) Pal; + + /* Indexed mode not supported */ + if (!rgb_flag) + return NULL; + + /* Allocate wmesa context */ + c = CALLOC_STRUCT(wmesa_context); + if (!c) + return NULL; + + c->hDC = hDC; + + /* Get data for visual */ + /* Dealing with this is actually a bit of overkill because Mesa will end + * up treating all color component size requests less than 8 by using + * a single byte per channel. In addition, the interface to the span + * routines passes colors as an entire byte per channel anyway, so there + * is nothing to be saved by telling the visual to be 16 bits if the device + * is 16 bits. That is, Mesa is going to compute colors down to 8 bits per + * channel anyway. + * But we go through the motions here anyway. + */ + c->cColorBits = GetDeviceCaps(c->hDC, BITSPIXEL); + + switch (c->cColorBits) { + case 16: + red_bits = green_bits = blue_bits = 5; + alpha_bits = 0; + break; + default: + red_bits = green_bits = blue_bits = 8; + alpha_bits = 8; + break; + } + /* Create visual based on flags */ + visual = _mesa_create_visual(rgb_flag, + db_flag, /* db_flag */ + GL_FALSE, /* stereo */ + red_bits, green_bits, blue_bits, /* color RGB */ + alpha_flag ? alpha_bits : 0, /* color A */ + 0, /* index bits */ + DEFAULT_SOFTWARE_DEPTH_BITS, /* depth_bits */ + 8, /* stencil_bits */ + 16,16,16, /* accum RGB */ + alpha_flag ? 16 : 0, /* accum A */ + 1); /* num samples */ + + if (!visual) { + _mesa_free(c); + return NULL; + } + + pws = wmesa_get_pipe_winsys(visual); + + screen = softpipe_create_screen(pws); + + if (!screen) { + _mesa_free(c); + return NULL; + } + + pipe = softpipe_create(screen, pws, NULL); + + if (!pipe) { + /* FIXME - free screen */ + _mesa_free(c); + return NULL; + } + + pipe->priv = c; + + c->st = st_create_context(pipe, visual, NULL); + + c->st->ctx->DriverCtx = c; + + return c; +} + + +void WMesaDestroyContext( WMesaContext pwc ) +{ + GLcontext *ctx = pwc->st->ctx; + WMesaFramebuffer pwfb; + GET_CURRENT_CONTEXT(cur_ctx); + + if (cur_ctx == ctx) { + /* unbind current if deleting current context */ + WMesaMakeCurrent(NULL, NULL); + } + + /* clean up frame buffer resources */ + pwfb = wmesa_lookup_framebuffer(pwc->hDC); + if (pwfb) { +#if 0 + wmDeleteBackingStore(pwfb); +#endif + wmesa_free_framebuffer(pwc->hDC); + } + + /* Release for device, not memory contexts */ + if (WindowFromDC(pwc->hDC) != NULL) + { + ReleaseDC(WindowFromDC(pwc->hDC), pwc->hDC); + } + + st_destroy_context(pwc->st); + _mesa_free(pwc); +} + + +void WMesaMakeCurrent(WMesaContext c, HDC hdc) +{ + GLuint width = 0, height = 0; + WMesaFramebuffer pwfb; + + { + /* return if already current */ + GET_CURRENT_CONTEXT(ctx); + WMesaContext pwc = wmesa_context(ctx); + if (pwc && c == pwc && pwc->hDC == hdc) + return; + } + + pwfb = wmesa_lookup_framebuffer(hdc); + + if (hdc) { + get_window_size(hdc, &width, &height); + } + + /* Lazy creation of framebuffers */ + if (c && !pwfb && (hdc != 0)) { + GLvisual *visual = &c->st->ctx->Visual; + + pwfb = wmesa_new_framebuffer(hdc, visual, width, height); + } + + if (c && pwfb) { + st_make_current(c->st, pwfb->stfb, pwfb->stfb); + + st_resize_framebuffer(pwfb->stfb, width, height); + } + else { + /* Detach */ + st_make_current( NULL, NULL, NULL ); + } +} + + +void WMesaSwapBuffers( HDC hdc ) +{ + struct pipe_surface *surf; + struct wm_buffer *wm_buf; + WMesaFramebuffer pwfb = wmesa_lookup_framebuffer(hdc); + BITMAPINFO bmi, *pbmi; + + if (!pwfb) { + _mesa_problem(NULL, "wmesa: swapbuffers on unknown hdc"); + return; + } + + + /* If we're swapping the buffer associated with the current context + * we have to flush any pending rendering commands first. + */ + st_notify_swapbuffers(pwfb->stfb); + +#if 0 + BitBlt(pwfb->hDC, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, + pwfb->dib_hDC, 0, 0, SRCCOPY); +#else + surf = st_get_framebuffer_surface(pwfb->stfb, ST_SURFACE_BACK_LEFT); + wm_buf = wm_buffer(surf->buffer); + + pbmi = &bmi; + memset(pbmi, 0, sizeof(BITMAPINFO)); + pbmi->bmiHeader.biSize = sizeof(BITMAPINFOHEADER); + pbmi->bmiHeader.biWidth = pwfb->stfb->Base.Width; + pbmi->bmiHeader.biHeight= -((long)pwfb->stfb->Base.Height); + pbmi->bmiHeader.biPlanes = 1; + pbmi->bmiHeader.biBitCount = pwfb->cColorBits; + pbmi->bmiHeader.biCompression = BI_RGB; + pbmi->bmiHeader.biSizeImage = 0; + pbmi->bmiHeader.biXPelsPerMeter = 0; + pbmi->bmiHeader.biYPelsPerMeter = 0; + pbmi->bmiHeader.biClrUsed = 0; + pbmi->bmiHeader.biClrImportant = 0; + + StretchDIBits(pwfb->hDC, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, wm_buf->data, pbmi, 0, SRCCOPY); + + { + GLuint width = 0, height = 0; + + get_window_size(pwfb->hDC, &width, &height); + + st_resize_framebuffer(pwfb->stfb, width, height); + } +#endif +} + +/* This is hopefully a temporary hack to define some needed dispatch + * table entries. Hopefully, I'll find a better solution. The + * dispatch table generation scripts ought to be making these dummy + * stubs as well. */ +#if !defined(__MINGW32__) || !defined(GL_NO_STDCALL) +void gl_dispatch_stub_543(void){} +void gl_dispatch_stub_544(void){} +void gl_dispatch_stub_545(void){} +void gl_dispatch_stub_546(void){} +void gl_dispatch_stub_547(void){} +void gl_dispatch_stub_548(void){} +void gl_dispatch_stub_549(void){} +void gl_dispatch_stub_550(void){} +void gl_dispatch_stub_551(void){} +void gl_dispatch_stub_552(void){} +void gl_dispatch_stub_553(void){} +void gl_dispatch_stub_554(void){} +void gl_dispatch_stub_555(void){} +void gl_dispatch_stub_556(void){} +void gl_dispatch_stub_557(void){} +void gl_dispatch_stub_558(void){} +void gl_dispatch_stub_559(void){} +void gl_dispatch_stub_560(void){} +void gl_dispatch_stub_561(void){} +void gl_dispatch_stub_565(void){} +void gl_dispatch_stub_566(void){} +void gl_dispatch_stub_577(void){} +void gl_dispatch_stub_578(void){} +void gl_dispatch_stub_603(void){} +void gl_dispatch_stub_645(void){} +void gl_dispatch_stub_646(void){} +void gl_dispatch_stub_647(void){} +void gl_dispatch_stub_648(void){} +void gl_dispatch_stub_649(void){} +void gl_dispatch_stub_650(void){} +void gl_dispatch_stub_651(void){} +void gl_dispatch_stub_652(void){} +void gl_dispatch_stub_653(void){} +void gl_dispatch_stub_733(void){} +void gl_dispatch_stub_734(void){} +void gl_dispatch_stub_735(void){} +void gl_dispatch_stub_736(void){} +void gl_dispatch_stub_737(void){} +void gl_dispatch_stub_738(void){} +void gl_dispatch_stub_744(void){} +void gl_dispatch_stub_745(void){} +void gl_dispatch_stub_746(void){} +void gl_dispatch_stub_760(void){} +void gl_dispatch_stub_761(void){} +void gl_dispatch_stub_763(void){} +void gl_dispatch_stub_765(void){} +void gl_dispatch_stub_766(void){} +void gl_dispatch_stub_767(void){} +void gl_dispatch_stub_768(void){} + +void gl_dispatch_stub_562(void){} +void gl_dispatch_stub_563(void){} +void gl_dispatch_stub_564(void){} +void gl_dispatch_stub_567(void){} +void gl_dispatch_stub_568(void){} +void gl_dispatch_stub_569(void){} +void gl_dispatch_stub_580(void){} +void gl_dispatch_stub_581(void){} +void gl_dispatch_stub_606(void){} +void gl_dispatch_stub_654(void){} +void gl_dispatch_stub_655(void){} +void gl_dispatch_stub_656(void){} +void gl_dispatch_stub_739(void){} +void gl_dispatch_stub_740(void){} +void gl_dispatch_stub_741(void){} +void gl_dispatch_stub_748(void){} +void gl_dispatch_stub_749(void){} +void gl_dispatch_stub_769(void){} +void gl_dispatch_stub_770(void){} +void gl_dispatch_stub_771(void){} +void gl_dispatch_stub_772(void){} +void gl_dispatch_stub_773(void){} + +#endif diff --git a/src/gallium/winsys/gdi/wmesadef.h b/src/gallium/winsys/gdi/wmesadef.h new file mode 100644 index 0000000000..fb8ce30a08 --- /dev/null +++ b/src/gallium/winsys/gdi/wmesadef.h @@ -0,0 +1,40 @@ +#ifndef WMESADEF_H +#define WMESADEF_H +#ifdef __MINGW32__ +#include +#endif +#if 0 +#include "context.h" +#endif +#include "state_tracker/st_context.h" +#include "state_tracker/st_public.h" + + +/** + * The Windows Mesa rendering context, derived from GLcontext. + */ +struct wmesa_context { + struct st_context *st; + HDC hDC; + BYTE cColorBits; +}; + +/** + * Windows framebuffer, derived from gl_framebuffer + */ +struct wmesa_framebuffer +{ + struct st_framebuffer *stfb; + HDC hDC; + int pixelformat; + BYTE cColorBits; + HDC dib_hDC; + HBITMAP hbmDIB; + HBITMAP hOldBitmap; + PBYTE pbPixels; + struct wmesa_framebuffer *next; +}; + +typedef struct wmesa_framebuffer *WMesaFramebuffer; + +#endif /* WMESADEF_H */ -- cgit v1.2.3 From 5a67df6d7cc8c74bfb71a8f19b8f6fdfb525091b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 3 Jun 2008 00:04:19 +0900 Subject: scons: Integrate gdi winsys. Conditional build of the winsys based on the platform. --- src/gallium/winsys/SConscript | 9 ++- src/gallium/winsys/dri/SConscript | 101 ++++++++++++++++---------------- src/gallium/winsys/dri/intel/SConscript | 76 ++++++++++++------------ src/gallium/winsys/xlib/SConscript | 50 ++++++++-------- 4 files changed, 124 insertions(+), 112 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript index e8a581adb2..bf1718e7a5 100644 --- a/src/gallium/winsys/SConscript +++ b/src/gallium/winsys/SConscript @@ -1,11 +1,16 @@ Import('*') -if 'intel' in env['winsys'] and dri: +if env['dri']: SConscript([ 'dri/SConscript', ]) -if 'xlib' in env['winsys'] and not dri: +if 'xlib' in env['winsys']: SConscript([ 'xlib/SConscript', ]) + +if 'gdi' in env['winsys']: + SConscript([ + 'gdi/SConscript', + ]) diff --git a/src/gallium/winsys/dri/SConscript b/src/gallium/winsys/dri/SConscript index 8c56ce917c..aef5210a32 100644 --- a/src/gallium/winsys/dri/SConscript +++ b/src/gallium/winsys/dri/SConscript @@ -1,51 +1,54 @@ Import('*') -drienv = env.Clone() - -drienv.Replace(CPPPATH = [ - '#src/mesa/drivers/dri/common', - '#include', - '#include/GL/internal', - '#src/gallium/include', - '#src/gallium/auxiliary', - '#src/gallium/drivers', - '#src/mesa', - '#src/mesa/main', - '#src/mesa/glapi', - '#src/mesa/math', - '#src/mesa/transform', - '#src/mesa/shader', - '#src/mesa/swrast', - '#src/mesa/swrast_setup', - '#src/egl/main', - '#src/egl/drivers/dri', -]) - -drienv.ParseConfig('pkg-config --cflags --libs libdrm') - -COMMON_GALLIUM_SOURCES = [ - '#src/mesa/drivers/dri/common/utils.c', - '#src/mesa/drivers/dri/common/vblank.c', - '#src/mesa/drivers/dri/common/dri_util.c', - '#src/mesa/drivers/dri/common/xmlconfig.c', -] - -COMMON_BM_SOURCES = [ - '#src/mesa/drivers/dri/common/dri_bufmgr.c', - '#src/mesa/drivers/dri/common/dri_drmpool.c', -] - -Export([ - 'drienv', - 'COMMON_GALLIUM_SOURCES', - 'COMMON_BM_SOURCES', -]) - -# TODO: Installation -#install: $(LIBNAME) -# $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) -# $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) - -SConscript([ - 'intel/SConscript', -]) +if env['dri']: + + drienv = env.Clone() + + drienv.Replace(CPPPATH = [ + '#src/mesa/drivers/dri/common', + '#include', + '#include/GL/internal', + '#src/gallium/include', + '#src/gallium/auxiliary', + '#src/gallium/drivers', + '#src/mesa', + '#src/mesa/main', + '#src/mesa/glapi', + '#src/mesa/math', + '#src/mesa/transform', + '#src/mesa/shader', + '#src/mesa/swrast', + '#src/mesa/swrast_setup', + '#src/egl/main', + '#src/egl/drivers/dri', + ]) + + drienv.ParseConfig('pkg-config --cflags --libs libdrm') + + COMMON_GALLIUM_SOURCES = [ + '#src/mesa/drivers/dri/common/utils.c', + '#src/mesa/drivers/dri/common/vblank.c', + '#src/mesa/drivers/dri/common/dri_util.c', + '#src/mesa/drivers/dri/common/xmlconfig.c', + ] + + COMMON_BM_SOURCES = [ + '#src/mesa/drivers/dri/common/dri_bufmgr.c', + '#src/mesa/drivers/dri/common/dri_drmpool.c', + ] + + Export([ + 'drienv', + 'COMMON_GALLIUM_SOURCES', + 'COMMON_BM_SOURCES', + ]) + + # TODO: Installation + #install: $(LIBNAME) + # $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) + # $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) + + if 'intel' in env['winsys']: + SConscript([ + 'intel/SConscript', + ]) diff --git a/src/gallium/winsys/dri/intel/SConscript b/src/gallium/winsys/dri/intel/SConscript index 0ad19d42a8..6a4f50afcc 100644 --- a/src/gallium/winsys/dri/intel/SConscript +++ b/src/gallium/winsys/dri/intel/SConscript @@ -1,39 +1,41 @@ Import('*') -env = drienv.Clone() - -env.Append(CPPPATH = [ - '../intel', - 'server' -]) - -#MINIGLX_SOURCES = server/intel_dri.c - -DRIVER_SOURCES = [ - 'intel_winsys_pipe.c', - 'intel_winsys_softpipe.c', - 'intel_winsys_i915.c', - 'intel_batchbuffer.c', - 'intel_swapbuffers.c', - 'intel_context.c', - 'intel_lock.c', - 'intel_screen.c', - 'intel_batchpool.c', -] - -sources = \ - COMMON_GALLIUM_SOURCES + \ - COMMON_BM_SOURCES + \ - DRIVER_SOURCES - -drivers = [ - softpipe, - i915simple -] - -# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions -env.SharedLibrary( - target ='i915tex_dri.so', - source = sources, - LIBS = drivers + mesa + auxiliaries + env['LIBS'], -) \ No newline at end of file +if 'mesa' in env['statetrackers']: + + env = drienv.Clone() + + env.Append(CPPPATH = [ + '../intel', + 'server' + ]) + + #MINIGLX_SOURCES = server/intel_dri.c + + DRIVER_SOURCES = [ + 'intel_winsys_pipe.c', + 'intel_winsys_softpipe.c', + 'intel_winsys_i915.c', + 'intel_batchbuffer.c', + 'intel_swapbuffers.c', + 'intel_context.c', + 'intel_lock.c', + 'intel_screen.c', + 'intel_batchpool.c', + ] + + sources = \ + COMMON_GALLIUM_SOURCES + \ + COMMON_BM_SOURCES + \ + DRIVER_SOURCES + + drivers = [ + softpipe, + i915simple + ] + + # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions + env.SharedLibrary( + target ='i915tex_dri.so', + source = sources, + LIBS = drivers + mesa + auxiliaries + env['LIBS'], + ) diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript index 218b89285f..5e98a36abc 100644 --- a/src/gallium/winsys/xlib/SConscript +++ b/src/gallium/winsys/xlib/SConscript @@ -3,30 +3,32 @@ Import('*') -env = env.Clone() +if env['platform'] == 'linux' and 'mesa' in env['statetrackers'] and not env['dri']: -env.Append(CPPPATH = [ - '#/src/mesa', - '#/src/mesa/main', -]) + env = env.Clone() -sources = [ - 'glxapi.c', - 'fakeglx.c', - 'xfonts.c', - 'xm_api.c', - 'xm_winsys.c', - 'xm_winsys_aub.c', - 'brw_aub.c', -] - -drivers = [ - softpipe, -] + env.Append(CPPPATH = [ + '#/src/mesa', + '#/src/mesa/main', + ]) -# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions -env.SharedLibrary( - target ='GL', - source = sources, - LIBS = glapi + mesa + drivers + auxiliaries + env['LIBS'], -) + sources = [ + 'glxapi.c', + 'fakeglx.c', + 'xfonts.c', + 'xm_api.c', + 'xm_winsys.c', + 'xm_winsys_aub.c', + 'brw_aub.c', + ] + + drivers = [ + softpipe, + ] + + # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions + env.SharedLibrary( + target ='GL', + source = sources, + LIBS = glapi + mesa + drivers + auxiliaries + env['LIBS'], + ) -- cgit v1.2.3 From 4ee14279f3a466093869f1f40819e6c6d5af378d Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 2 Jun 2008 14:55:06 +0200 Subject: i915: Rework of batchbuffer code --- src/gallium/drivers/i915simple/i915_batch.h | 92 +++++++++++++++++++++++++-- src/gallium/drivers/i915simple/i915_context.c | 3 +- src/gallium/drivers/i915simple/i915_context.h | 4 +- src/gallium/drivers/i915simple/i915_debug.c | 2 +- 4 files changed, 94 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h index 4ea06ce02b..6c62e84bc3 100644 --- a/src/gallium/drivers/i915simple/i915_batch.h +++ b/src/gallium/drivers/i915simple/i915_batch.h @@ -31,23 +31,107 @@ #include "i915_winsys.h" #include "i915_debug.h" +struct i915_batchbuffer +{ + struct pipe_buffer *buffer; + struct i915_winsys *winsys; + + unsigned char *map; + unsigned char *ptr; + + size_t size; + + size_t relocs; + size_t max_relocs; +}; + +static INLINE boolean +i915_batchbuffer_check( struct i915_batchbuffer *batch, + size_t dwords, + size_t relocs ) +{ +#if 0 /* To be used */ + /** TODO JB: Check relocs */ + return dwords * 4 <= batch->size - (batch->ptr - batch->map); +#else + if (batch->winsys->batch_start( batch->winsys, dwords, relocs )) + return 1; + return 0; +#endif +} + +static INLINE size_t +i915_batchbuffer_space( struct i915_batchbuffer *batch ) +{ +#if 0 /* To be used */ + return batch->size - (batch->ptr - batch->map); +#else + return 0; +#endif +} + +static INLINE void +i915_batchbuffer_dword( struct i915_batchbuffer *batch, + unsigned dword ) +{ +#if 0 /* To be used */ + if (i915_batchbuffer_space(batch) < 4) + return; + + *(unsigned *)batch->ptr = dword; + batch->ptr += 4; +#else + batch->winsys->batch_dword( batch->winsys, dword ); +#endif +} + +static INLINE void +i915_batchbuffer_write( struct i915_batchbuffer *batch, + void *data, + size_t size ) +{ +#if 0 /* To be used */ + if (i915_batchbuffer_space(batch) < size) + return; + + memcpy(data, batch->ptr, size); + batch->ptr += size; +#else +#endif +} + +static INLINE void +i915_batchbuffer_reloc( struct i915_batchbuffer *batch, + struct pipe_buffer *buffer, + size_t flags, + size_t offset ) +{ + batch->winsys->batch_reloc( batch->winsys, buffer, flags, offset ); +} + +static INLINE void +i915_batchbuffer_flush( struct i915_batchbuffer *batch, + struct pipe_fence_handle **fence ) +{ + batch->winsys->batch_flush( batch->winsys, fence ); +} + #define BATCH_LOCALS #define BEGIN_BATCH( dwords, relocs ) \ - (i915->batch_start = i915->winsys->batch_start( i915->winsys, dwords, relocs )) + (i915_batchbuffer_check( i915->batch, dwords, relocs )) #define OUT_BATCH( dword ) \ - i915->winsys->batch_dword( i915->winsys, dword ) + i915_batchbuffer_dword( i915->batch, dword ) #define OUT_RELOC( buf, flags, delta ) \ - i915->winsys->batch_reloc( i915->winsys, buf, flags, delta ) + i915_batchbuffer_reloc( i915->batch, buf, flags, delta ) #define ADVANCE_BATCH() #define FLUSH_BATCH(fence) do { \ if (0) i915_dump_batchbuffer( i915 ); \ i915->winsys->batch_flush( i915->winsys, fence ); \ - i915->batch_start = NULL; \ i915->hardware_dirty = ~0; \ } while (0) diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 81cab75738..378e4a5d4d 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -181,7 +181,8 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, /* Batch stream debugging is a bit hacked up at the moment: */ - i915->batch_start = NULL; + i915->batch = CALLOC_STRUCT(i915_batchbuffer); + i915->batch->winsys = i915_winsys; return &i915->pipe; } diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 2da90ae49d..2ee0381648 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -209,6 +209,8 @@ struct i915_texture { struct pipe_buffer *buffer; }; +struct i915_batchbuffer; + struct i915_context { struct pipe_context pipe; @@ -241,7 +243,7 @@ struct i915_context unsigned num_vertex_elements; unsigned num_vertex_buffers; - unsigned *batch_start; + struct i915_batchbuffer *batch; /** Vertex buffer */ struct pipe_buffer *vbo; diff --git a/src/gallium/drivers/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c index 9b9111167f..a121dc0af4 100644 --- a/src/gallium/drivers/i915simple/i915_debug.c +++ b/src/gallium/drivers/i915simple/i915_debug.c @@ -861,7 +861,7 @@ void i915_dump_batchbuffer( struct i915_context *i915 ) { struct debug_stream stream; - unsigned *start = i915->batch_start; + unsigned *start = 0;/*i915->batch_start;*/ unsigned *end = i915->winsys->batch_start( i915->winsys, 0, 0 ); unsigned long bytes = (unsigned long) (end - start) * 4; boolean done = FALSE; -- cgit v1.2.3 From 7cc23a9eaebc788ae34f6e06c6227524d08a7693 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 2 Jun 2008 17:22:45 +0200 Subject: i915: Implement and use the reworked batchbuffer code --- src/gallium/drivers/i915simple/i915_batch.h | 20 ++--------- src/gallium/drivers/i915simple/i915_context.c | 2 +- src/gallium/drivers/i915simple/i915_debug.c | 3 +- src/gallium/drivers/i915simple/i915_winsys.h | 22 +++++------- src/gallium/winsys/dri/intel/intel_batchbuffer.c | 46 ++++++++++++------------ src/gallium/winsys/dri/intel/intel_batchbuffer.h | 19 +++++----- src/gallium/winsys/dri/intel/intel_winsys_i915.c | 35 +++--------------- 7 files changed, 52 insertions(+), 95 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h index 6c62e84bc3..9379d1fb5d 100644 --- a/src/gallium/drivers/i915simple/i915_batch.h +++ b/src/gallium/drivers/i915simple/i915_batch.h @@ -40,6 +40,7 @@ struct i915_batchbuffer unsigned char *ptr; size_t size; + size_t actual_size; size_t relocs; size_t max_relocs; @@ -50,39 +51,25 @@ i915_batchbuffer_check( struct i915_batchbuffer *batch, size_t dwords, size_t relocs ) { -#if 0 /* To be used */ /** TODO JB: Check relocs */ return dwords * 4 <= batch->size - (batch->ptr - batch->map); -#else - if (batch->winsys->batch_start( batch->winsys, dwords, relocs )) - return 1; - return 0; -#endif } static INLINE size_t i915_batchbuffer_space( struct i915_batchbuffer *batch ) { -#if 0 /* To be used */ return batch->size - (batch->ptr - batch->map); -#else - return 0; -#endif } static INLINE void i915_batchbuffer_dword( struct i915_batchbuffer *batch, unsigned dword ) { -#if 0 /* To be used */ if (i915_batchbuffer_space(batch) < 4) return; *(unsigned *)batch->ptr = dword; batch->ptr += 4; -#else - batch->winsys->batch_dword( batch->winsys, dword ); -#endif } static INLINE void @@ -90,14 +77,11 @@ i915_batchbuffer_write( struct i915_batchbuffer *batch, void *data, size_t size ) { -#if 0 /* To be used */ if (i915_batchbuffer_space(batch) < size) return; memcpy(data, batch->ptr, size); batch->ptr += size; -#else -#endif } static INLINE void @@ -135,4 +119,4 @@ i915_batchbuffer_flush( struct i915_batchbuffer *batch, i915->hardware_dirty = ~0; \ } while (0) -#endif +#endif diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 378e4a5d4d..243b4a6852 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -181,7 +181,7 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, /* Batch stream debugging is a bit hacked up at the moment: */ - i915->batch = CALLOC_STRUCT(i915_batchbuffer); + i915->batch = i915_winsys->batch_get(i915_winsys); i915->batch->winsys = i915_winsys; return &i915->pipe; diff --git a/src/gallium/drivers/i915simple/i915_debug.c b/src/gallium/drivers/i915simple/i915_debug.c index a121dc0af4..5e26d1b905 100644 --- a/src/gallium/drivers/i915simple/i915_debug.c +++ b/src/gallium/drivers/i915simple/i915_debug.c @@ -861,8 +861,9 @@ void i915_dump_batchbuffer( struct i915_context *i915 ) { struct debug_stream stream; + /* TODO fix me */ unsigned *start = 0;/*i915->batch_start;*/ - unsigned *end = i915->winsys->batch_start( i915->winsys, 0, 0 ); + unsigned *end = 0;/*i915->winsys->batch_start( i915->winsys, 0, 0 );*/ unsigned long bytes = (unsigned long) (end - start) * 4; boolean done = FALSE; diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h index 5e16543f4e..9afaa16a62 100644 --- a/src/gallium/drivers/i915simple/i915_winsys.h +++ b/src/gallium/drivers/i915simple/i915_winsys.h @@ -55,6 +55,7 @@ extern "C" { * etc. */ +struct i915_batchbuffer; struct pipe_buffer; struct pipe_fence_handle; struct pipe_winsys; @@ -75,20 +76,10 @@ struct pipe_screen; struct i915_winsys { /** - * Reserve space on batch buffer. - * - * Returns a null pointer if there is insufficient space in the batch buffer - * to hold the requested number of dwords and relocations. - * - * The number of dwords should also include the number of relocations. + * Get the current batch buffer from the winsys. */ - unsigned *(*batch_start)( struct i915_winsys *sws, - unsigned dwords, - unsigned relocs ); - - void (*batch_dword)( struct i915_winsys *sws, - unsigned dword ); - + struct i915_batchbuffer *(*batch_get)( struct i915_winsys *sws ); + /** * Emit a relocation to a buffer. * @@ -103,7 +94,10 @@ struct i915_winsys { struct pipe_buffer *buf, unsigned access_flags, unsigned delta ); - + + /** + * Flush the batch. + */ void (*batch_flush)( struct i915_winsys *sws, struct pipe_fence_handle **fence ); }; diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c index 09d4eef484..aa33c12045 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.c +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.c @@ -65,8 +65,10 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) driBOUnrefUserList(batch->list); driBOResetList(batch->list); - batch->size = batch->intel->intelScreen->max_batch_size; - driBOData(batch->buffer, batch->size, NULL, NULL, 0); + /* base.size is the size available to the i915simple driver */ + batch->base.size = batch->intel->intelScreen->max_batch_size - BATCH_RESERVED; + batch->base.actual_size = batch->intel->intelScreen->max_batch_size; + driBOData(batch->buffer, batch->base.actual_size, NULL, NULL, 0); /* * Add the batchbuffer to the validate list. @@ -105,9 +107,9 @@ intel_batchbuffer_reset(struct intel_batchbuffer *batch) batch->reloc[2] = 0; /* Only a single relocation list. */ batch->reloc[3] = 0; /* Only a single relocation list. */ - batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); + batch->base.map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); batch->poolOffset = driBOPoolOffset(batch->buffer); - batch->ptr = batch->map; + batch->base.ptr = batch->base.map; batch->dirty_state = ~0; batch->nr_relocs = 0; batch->flags = 0; @@ -142,9 +144,9 @@ intel_batchbuffer_free(struct intel_batchbuffer *batch) DRM_FENCE_TYPE_EXE, GL_FALSE); driFenceUnReference(&batch->last_fence); } - if (batch->map) { + if (batch->base.map) { driBOUnmap(batch->buffer); - batch->map = NULL; + batch->base.map = NULL; } driBOUnReference(batch->buffer); driBOFreeList(batch->list); @@ -191,7 +193,7 @@ intel_offset_relocation(struct intel_batchbuffer *batch, reloc = batch->reloc + (I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE); - reloc[0] = ((uint8_t *)batch->ptr - batch->drmBOVirtual); + reloc[0] = ((uint8_t *)batch->base.ptr - batch->drmBOVirtual); intel_batchbuffer_emit_dword(batch, req->presumed_offset + pre_add); reloc[1] = pre_add; reloc[2] = itemLoc; @@ -382,7 +384,7 @@ struct _DriFenceObject * intel_batchbuffer_flush(struct intel_batchbuffer *batch) { struct intel_context *intel = batch->intel; - GLuint used = batch->ptr - batch->map; + GLuint used = batch->base.ptr - batch->base.map; GLboolean was_locked = intel->locked; struct _DriFenceObject *fence; @@ -396,32 +398,32 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch) */ #if 0 /* ZZZ JB: what should we do here? */ if (used & 4) { - ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->ptr)[1] = 0; - ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; + ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->base.ptr)[1] = 0; + ((int *) batch->base.ptr)[2] = MI_BATCH_BUFFER_END; used += 12; } else { - ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; + ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->base.ptr)[1] = MI_BATCH_BUFFER_END; used += 8; } #else if (used & 4) { - ((int *) batch->ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->ptr)[1] = 0; - ((int *) batch->ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; + ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->base.ptr)[1] = 0; + ((int *) batch->base.ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; used += 12; } else { - ((int *) batch->ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; + ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->base.ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; used += 8; } #endif driBOUnmap(batch->buffer); - batch->ptr = NULL; - batch->map = NULL; + batch->base.ptr = NULL; + batch->base.map = NULL; /* TODO: Just pass the relocation list and dma buffer up to the * kernel. @@ -455,6 +457,6 @@ intel_batchbuffer_data(struct intel_batchbuffer *batch, { assert((bytes & 3) == 0); intel_batchbuffer_require_space(batch, bytes, flags); - memcpy(batch->ptr, data, bytes); - batch->ptr += bytes; + memcpy(batch->base.ptr, data, bytes); + batch->base.ptr += bytes; } diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.h b/src/gallium/winsys/dri/intel/intel_batchbuffer.h index 9e4b8043bf..dcb2121ddf 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.h +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.h @@ -3,6 +3,7 @@ #include "mtypes.h" #include "ws_dri_bufmgr.h" +#include "i915simple/i915_batch.h" struct intel_context; @@ -17,7 +18,8 @@ struct intel_context; struct intel_batchbuffer { - struct bufmgr *bm; + struct i915_batchbuffer base; + struct intel_context *intel; struct _DriBufferObject *buffer; @@ -26,15 +28,11 @@ struct intel_batchbuffer struct _DriBufferList *list; GLuint list_count; - GLubyte *map; - GLubyte *ptr; uint32_t *reloc; GLuint reloc_size; GLuint nr_relocs; - GLuint size; - GLuint dirty_state; GLuint id; @@ -83,7 +81,7 @@ intel_offset_relocation(struct intel_batchbuffer *batch, static INLINE GLuint intel_batchbuffer_space(struct intel_batchbuffer *batch) { - return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); + return (batch->base.size - BATCH_RESERVED) - (batch->base.ptr - batch->base.map); } @@ -92,8 +90,8 @@ intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) { assert(batch->map); assert(intel_batchbuffer_space(batch) >= 4); - *(GLuint *) (batch->ptr) = dword; - batch->ptr += 4; + *(GLuint *) (batch->base.ptr) = dword; + batch->base.ptr += 4; } static INLINE void @@ -114,20 +112,25 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch, /* Here are the crusty old macros, to be removed: */ +#undef BATCH_LOCALS #define BATCH_LOCALS +#undef BEGIN_BATCH #define BEGIN_BATCH(n, flags) do { \ assert(!intel->prim.flush); \ intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ } while (0) +#undef OUT_BATCH #define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) +#undef OUT_RELOC #define OUT_RELOC(buf,flags,mask,delta) do { \ assert((delta) >= 0); \ intel_offset_relocation(intel->batch, delta, buf, flags, mask); \ } while (0) +#undef ADVANCE_BATCH #define ADVANCE_BATCH() do { } while(0) diff --git a/src/gallium/winsys/dri/intel/intel_winsys_i915.c b/src/gallium/winsys/dri/intel/intel_winsys_i915.c index a35825d36a..013291364a 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_i915.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_i915.c @@ -63,28 +63,11 @@ intel_i915_winsys( struct i915_winsys *sws ) /* Simple batchbuffer interface: */ -static unsigned *intel_i915_batch_start( struct i915_winsys *sws, - unsigned dwords, - unsigned relocs ) +static struct i915_batchbuffer* +intel_i915_batch_get( struct i915_winsys *sws ) { struct intel_context *intel = intel_i915_winsys(sws)->intel; - - /* XXX: check relocs. - */ - if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { - /* XXX: Hmm, the driver can't really do much with this pointer: - */ - return (unsigned *)intel->batch->ptr; - } - else - return NULL; -} - -static void intel_i915_batch_dword( struct i915_winsys *sws, - unsigned dword ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - intel_batchbuffer_emit_dword( intel->batch, dword ); + return &intel->batch->base; } static void intel_i915_batch_reloc( struct i915_winsys *sws, @@ -106,22 +89,13 @@ static void intel_i915_batch_reloc( struct i915_winsys *sws, mask |= DRM_BO_FLAG_READ; } -#if 0 /* JB old */ - intel_batchbuffer_emit_reloc( intel->batch, - dri_bo( buf ), - flags, mask, - delta ); -#else /* new */ intel_offset_relocation( intel->batch, delta, dri_bo( buf ), flags, mask ); -#endif } - - static void intel_i915_batch_flush( struct i915_winsys *sws, struct pipe_fence_handle **fence ) { @@ -166,8 +140,7 @@ intel_create_i915simple( struct intel_context *intel, /* Fill in this struct with callbacks that i915simple will need to * communicate with the window system, buffer manager, etc. */ - iws->winsys.batch_start = intel_i915_batch_start; - iws->winsys.batch_dword = intel_i915_batch_dword; + iws->winsys.batch_get = intel_i915_batch_get; iws->winsys.batch_reloc = intel_i915_batch_reloc; iws->winsys.batch_flush = intel_i915_batch_flush; iws->pws = winsys; -- cgit v1.2.3 From 6a3aab1983d999a3a9eccbeaab88195820abf467 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 2 Jun 2008 17:23:30 +0200 Subject: i915: Removed useless defines --- src/gallium/drivers/i915simple/i915_batch.h | 4 ---- src/gallium/drivers/i915simple/i915_blit.c | 4 ---- src/gallium/drivers/i915simple/i915_flush.c | 1 - 3 files changed, 9 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h index 9379d1fb5d..987d0b4d28 100644 --- a/src/gallium/drivers/i915simple/i915_batch.h +++ b/src/gallium/drivers/i915simple/i915_batch.h @@ -100,8 +100,6 @@ i915_batchbuffer_flush( struct i915_batchbuffer *batch, batch->winsys->batch_flush( batch->winsys, fence ); } -#define BATCH_LOCALS - #define BEGIN_BATCH( dwords, relocs ) \ (i915_batchbuffer_check( i915->batch, dwords, relocs )) @@ -111,8 +109,6 @@ i915_batchbuffer_flush( struct i915_batchbuffer *batch, #define OUT_RELOC( buf, flags, delta ) \ i915_batchbuffer_reloc( i915->batch, buf, flags, delta ) -#define ADVANCE_BATCH() - #define FLUSH_BATCH(fence) do { \ if (0) i915_dump_batchbuffer( i915 ); \ i915->winsys->batch_flush( i915->winsys, fence ); \ diff --git a/src/gallium/drivers/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c index 24449e3fb3..b500ff36a8 100644 --- a/src/gallium/drivers/i915simple/i915_blit.c +++ b/src/gallium/drivers/i915simple/i915_blit.c @@ -45,7 +45,6 @@ i915_fill_blit(struct i915_context *i915, unsigned color) { unsigned BR13, CMD; - BATCH_LOCALS; dst_pitch *= (short) cpp; @@ -79,7 +78,6 @@ i915_fill_blit(struct i915_context *i915, OUT_BATCH(((y + h) << 16) | (x + w)); OUT_RELOC( dst_buffer, I915_BUFFER_ACCESS_WRITE, dst_offset); OUT_BATCH(color); - ADVANCE_BATCH(); } @@ -100,7 +98,6 @@ i915_copy_blit( struct i915_context *i915, unsigned CMD, BR13; int dst_y2 = dst_y + h; int dst_x2 = dst_x + w; - BATCH_LOCALS; I915_DBG(i915, @@ -156,7 +153,6 @@ i915_copy_blit( struct i915_context *i915, OUT_BATCH((src_y << 16) | src_x); OUT_BATCH(((int) src_pitch & 0xffff)); OUT_RELOC(src_buffer, I915_BUFFER_ACCESS_READ, src_offset); - ADVANCE_BATCH(); } diff --git a/src/gallium/drivers/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c index 4c4718d68e..472e0ab774 100644 --- a/src/gallium/drivers/i915simple/i915_flush.c +++ b/src/gallium/drivers/i915simple/i915_flush.c @@ -62,7 +62,6 @@ static void i915_flush( struct pipe_context *pipe, assert(BEGIN_BATCH(1, 0)); } OUT_BATCH( flush ); - ADVANCE_BATCH(); } /* If there are no flags, just flush pending commands to hardware: -- cgit v1.2.3 From 58cccc8d6b49c75eeabe9b61055e69de824ff757 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 2 Jun 2008 21:51:31 +0200 Subject: draw: Fix DP3 implementation by replacing SSE with x87 version. --- src/gallium/auxiliary/draw/draw_vs_aos.c | 29 ++++++++++++++++++++++++++++- 1 file changed, 28 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 725f36b502..5bfcd96ac3 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -962,8 +962,35 @@ static boolean emit_COS( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } +#if 1 + +/* The x87 version. + */ +static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + struct x86_reg st1 = x86_make_reg( file_x87, 1 ); + + x87_fld_src( cp, &op->FullSrcRegisters[0], 0 ); + x87_fld_src( cp, &op->FullSrcRegisters[1], 0 ); + x87_fmulp( cp->func, st1 ); + x87_fld_src( cp, &op->FullSrcRegisters[0], 1 ); + x87_fld_src( cp, &op->FullSrcRegisters[1], 1 ); + x87_fmulp( cp->func, st1 ); + x87_faddp( cp->func, st1 ); + x87_fld_src( cp, &op->FullSrcRegisters[0], 2 ); + x87_fld_src( cp, &op->FullSrcRegisters[1], 2 ); + x87_fmulp( cp->func, st1 ); + x87_faddp( cp->func, st1 ); + + x87_fstp_dest4( cp, &op->FullDstRegisters[0] ); + + return TRUE; +} + +#else /* The dotproduct instructions don't really do that well in sse: + * XXX: produces wrong results -- disabled. */ static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { @@ -985,7 +1012,7 @@ static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } - +#endif static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { -- cgit v1.2.3 From 183d490ab139483c88d0b0f541714919de86235c Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 3 Jun 2008 10:59:46 +0200 Subject: draw: Fix fetch_src(). Resurrect SSE version of DP3. --- src/gallium/auxiliary/draw/draw_vs_aos.c | 43 +++++--------------------------- 1 file changed, 6 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 5bfcd96ac3..891f8c211a 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -532,10 +532,11 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, if (swz != SSE_SWIZZLE_NOOP || negs != 0 || abs != 0) { struct x86_reg dst = aos_get_xmm_reg(cp); - if (swz != SSE_SWIZZLE_NOOP) { + if (swz != SSE_SWIZZLE_NOOP) emit_pshufd(cp, dst, arg0, swz); - arg0 = dst; - } + else + sse_movaps(cp->func, dst, arg0); + arg0 = dst; if (negs && negs != 0xf) { struct x86_reg imm_swz = aos_get_internal_xmm(cp, IMM_SWZ); @@ -550,15 +551,13 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, (negs & 2) ? 1 : 0, (negs & 4) ? 1 : 0, (negs & 8) ? 1 : 0)); - sse_mulps(cp->func, dst, arg0); + sse_mulps(cp->func, dst, tmp); aos_release_xmm_reg(cp, tmp.idx); - arg0 = dst; } else if (negs) { struct x86_reg imm_negs = aos_get_internal_xmm(cp, IMM_NEGS); sse_mulps(cp->func, dst, imm_negs); - arg0 = dst; } @@ -571,10 +570,9 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, sse_movaps(cp->func, tmp, arg0); sse_mulps(cp->func, tmp, neg); - sse_maxps(cp->func, dst, arg0); + sse_maxps(cp->func, dst, tmp); aos_release_xmm_reg(cp, tmp.idx); - arg0 = dst; } } @@ -962,33 +960,6 @@ static boolean emit_COS( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } -#if 1 - -/* The x87 version. - */ -static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) -{ - struct x86_reg st1 = x86_make_reg( file_x87, 1 ); - - x87_fld_src( cp, &op->FullSrcRegisters[0], 0 ); - x87_fld_src( cp, &op->FullSrcRegisters[1], 0 ); - x87_fmulp( cp->func, st1 ); - x87_fld_src( cp, &op->FullSrcRegisters[0], 1 ); - x87_fld_src( cp, &op->FullSrcRegisters[1], 1 ); - x87_fmulp( cp->func, st1 ); - x87_faddp( cp->func, st1 ); - x87_fld_src( cp, &op->FullSrcRegisters[0], 2 ); - x87_fld_src( cp, &op->FullSrcRegisters[1], 2 ); - x87_fmulp( cp->func, st1 ); - x87_faddp( cp->func, st1 ); - - x87_fstp_dest4( cp, &op->FullDstRegisters[0] ); - - return TRUE; -} - -#else - /* The dotproduct instructions don't really do that well in sse: * XXX: produces wrong results -- disabled. */ @@ -1012,8 +983,6 @@ static boolean emit_DP3( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } -#endif - static boolean emit_DP4( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); -- cgit v1.2.3 From 0a5df5bc7d711a766c9d0963fb2029d60cf70a8b Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 3 Jun 2008 11:22:00 +0200 Subject: draw: Use register names more consistently. --- src/gallium/auxiliary/draw/draw_vs_aos.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 891f8c211a..3cf4a4d125 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -536,7 +536,6 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, emit_pshufd(cp, dst, arg0, swz); else sse_movaps(cp->func, dst, arg0); - arg0 = dst; if (negs && negs != 0xf) { struct x86_reg imm_swz = aos_get_internal_xmm(cp, IMM_SWZ); @@ -568,12 +567,14 @@ static struct x86_reg fetch_src( struct aos_compilation *cp, struct x86_reg neg = aos_get_internal(cp, IMM_NEGS); struct x86_reg tmp = aos_get_xmm_reg(cp); - sse_movaps(cp->func, tmp, arg0); + sse_movaps(cp->func, tmp, dst); sse_mulps(cp->func, tmp, neg); sse_maxps(cp->func, dst, tmp); aos_release_xmm_reg(cp, tmp.idx); } + + return dst; } return arg0; -- cgit v1.2.3 From afd635a95056abc0909ebd1503131660d6e78cf6 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 3 Jun 2008 12:35:01 +0200 Subject: i915: Fix compile on linux-dri-debug --- src/gallium/drivers/i915simple/i915_batch.h | 2 -- src/gallium/drivers/i915simple/i915_blit.c | 1 + src/gallium/winsys/dri/intel/intel_batchbuffer.h | 4 ++-- 3 files changed, 3 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h index 987d0b4d28..45bf4f4028 100644 --- a/src/gallium/drivers/i915simple/i915_batch.h +++ b/src/gallium/drivers/i915simple/i915_batch.h @@ -29,7 +29,6 @@ #define I915_BATCH_H #include "i915_winsys.h" -#include "i915_debug.h" struct i915_batchbuffer { @@ -110,7 +109,6 @@ i915_batchbuffer_flush( struct i915_batchbuffer *batch, i915_batchbuffer_reloc( i915->batch, buf, flags, delta ) #define FLUSH_BATCH(fence) do { \ - if (0) i915_dump_batchbuffer( i915 ); \ i915->winsys->batch_flush( i915->winsys, fence ); \ i915->hardware_dirty = ~0; \ } while (0) diff --git a/src/gallium/drivers/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c index b500ff36a8..22f91fab92 100644 --- a/src/gallium/drivers/i915simple/i915_blit.c +++ b/src/gallium/drivers/i915simple/i915_blit.c @@ -31,6 +31,7 @@ #include "i915_blit.h" #include "i915_reg.h" #include "i915_batch.h" +#include "i915_debug.h" #define FILE_DEBUG_FLAG DEBUG_BLIT diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.h b/src/gallium/winsys/dri/intel/intel_batchbuffer.h index dcb2121ddf..abb7a624f5 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.h +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.h @@ -88,7 +88,7 @@ intel_batchbuffer_space(struct intel_batchbuffer *batch) static INLINE void intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) { - assert(batch->map); + assert(batch->base.map); assert(intel_batchbuffer_space(batch) >= 4); *(GLuint *) (batch->base.ptr) = dword; batch->base.ptr += 4; @@ -100,7 +100,7 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch, { struct _DriFenceObject *fence; - assert(sz < batch->size - 8); + assert(sz < batch->base.size - 8); if (intel_batchbuffer_space(batch) < sz || (batch->flags != 0 && flags != 0 && batch->flags != flags)) { fence = intel_batchbuffer_flush(batch); -- cgit v1.2.3 From 0a4aea0e86a897d9afb9f2a0ec27f03faf8f1b21 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 2 Jun 2008 12:59:16 +0100 Subject: draw: respect driver's max vertex buffer size --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 6 ++++++ src/gallium/auxiliary/draw/draw_private.h | 1 + src/gallium/auxiliary/draw/draw_pt.h | 8 ++++++-- src/gallium/auxiliary/draw/draw_pt_emit.c | 6 +++++- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 6 +++++- src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c | 17 ++++++++++------- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 16 ++++++++++++---- src/gallium/auxiliary/draw/draw_pt_varray.c | 5 ++++- src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h | 11 +++++++---- src/gallium/auxiliary/draw/draw_pt_vcache.c | 13 ++++++++----- 10 files changed, 64 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index d514e28b77..9e5597c32c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -387,6 +387,12 @@ vbuf_alloc_vertices( struct vbuf_stage *vbuf ) /* Allocate a new vertex buffer */ vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; + + /* Must always succeed -- driver gives us a + * 'max_vertex_buffer_bytes' which it guarantees it can allocate, + * and it will flush itself if necessary to do so. If this does + * fail, we are basically without usable hardware. + */ vbuf->vertices = (uint *) vbuf->render->allocate_vertices(vbuf->render, (ushort) vbuf->vertex_size, (ushort) vbuf->max_vertices); diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 423f64262b..1865601cc0 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -265,6 +265,7 @@ void draw_pipeline_destroy( struct draw_context *draw ); * These flags expected at first vertex of lines & triangles when * unfilled and/or line stipple modes are operational. */ +#define DRAW_PIPE_MAX_VERTICES (0x1<<12) #define DRAW_PIPE_EDGE_FLAG_0 (0x1<<12) #define DRAW_PIPE_EDGE_FLAG_1 (0x2<<12) #define DRAW_PIPE_EDGE_FLAG_2 (0x4<<12) diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 6b8ba1d171..3d2a9c78b7 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -84,7 +84,8 @@ struct draw_pt_front_end { struct draw_pt_middle_end { void (*prepare)( struct draw_pt_middle_end *, unsigned prim, - unsigned opt ); + unsigned opt, + unsigned *max_vertices ); void (*run)( struct draw_pt_middle_end *, const unsigned *fetch_elts, @@ -105,6 +106,8 @@ struct draw_pt_middle_end { const ushort *draw_elts, unsigned draw_count ); + int (*get_max_vertex_count)( struct draw_pt_middle_end * ); + void (*finish)( struct draw_pt_middle_end * ); void (*destroy)( struct draw_pt_middle_end * ); }; @@ -158,7 +161,8 @@ boolean draw_pt_get_edgeflag( struct draw_context *draw, struct pt_emit; void draw_pt_emit_prepare( struct pt_emit *emit, - unsigned prim ); + unsigned prim, + unsigned *max_vertices ); void draw_pt_emit( struct pt_emit *emit, const float (*vertex_data)[4], diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index cf87cde996..a02f1f46fe 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -46,7 +46,8 @@ struct pt_emit { }; void draw_pt_emit_prepare( struct pt_emit *emit, - unsigned prim ) + unsigned prim, + unsigned *max_vertices ) { struct draw_context *draw = emit->draw; const struct vertex_info *vinfo; @@ -139,6 +140,9 @@ void draw_pt_emit_prepare( struct pt_emit *emit, translate_key_sanitize(&hw_key); emit->translate = translate_cache_find(emit->cache, &hw_key); } + + *max_vertices = (draw->render->max_vertex_buffer_bytes / + (vinfo->size * 4)); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 09bdc5fb5e..083ce105bf 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -90,7 +90,8 @@ struct fetch_emit_middle_end { static void fetch_emit_prepare( struct draw_pt_middle_end *middle, unsigned prim, - unsigned opt ) + unsigned opt, + unsigned *max_vertices ) { struct fetch_emit_middle_end *feme = (struct fetch_emit_middle_end *)middle; struct draw_context *draw = feme->draw; @@ -196,6 +197,9 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, draw->pt.vertex_buffer[i].buffer_offset), draw->pt.vertex_buffer[i].pitch ); } + + *max_vertices = (draw->render->max_vertex_buffer_bytes / + (vinfo->size * 4)); } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index efa6dddbda..5a1d79d996 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -70,7 +70,8 @@ struct fetch_shade_emit { static void fse_prepare( struct draw_pt_middle_end *middle, unsigned prim, - unsigned opt ) + unsigned opt, + unsigned *max_vertices ) { struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; struct draw_context *draw = fse->draw; @@ -189,6 +190,11 @@ static void fse_prepare( struct draw_pt_middle_end *middle, draw->pt.vertex_buffer[buf].pitch ); } + *max_vertices = (draw->render->max_vertex_buffer_bytes / + (vinfo->size * 4)); + + + //return TRUE; } @@ -204,7 +210,6 @@ static void fse_run_linear( struct draw_pt_middle_end *middle, { struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; struct draw_context *draw = fse->draw; - unsigned alloc_count = align(count, 4); char *hw_verts; /* XXX: need to flush to get prim_vbuf.c to release its allocation?? @@ -213,7 +218,7 @@ static void fse_run_linear( struct draw_pt_middle_end *middle, hw_verts = draw->render->allocate_vertices( draw->render, (ushort)fse->key.output_stride, - (ushort)alloc_count ); + (ushort)count ); if (!hw_verts) { assert(0); @@ -264,7 +269,6 @@ fse_run(struct draw_pt_middle_end *middle, { struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; struct draw_context *draw = fse->draw; - unsigned alloc_count = align(fetch_count, 4); void *hw_verts; /* XXX: need to flush to get prim_vbuf.c to release its allocation?? @@ -273,7 +277,7 @@ fse_run(struct draw_pt_middle_end *middle, hw_verts = draw->render->allocate_vertices( draw->render, (ushort)fse->key.output_stride, - (ushort)alloc_count ); + (ushort)fetch_count ); if (!hw_verts) { assert(0); return; @@ -319,7 +323,6 @@ static void fse_run_linear_elts( struct draw_pt_middle_end *middle, { struct fetch_shade_emit *fse = (struct fetch_shade_emit *)middle; struct draw_context *draw = fse->draw; - unsigned alloc_count = align(count, 4); char *hw_verts; /* XXX: need to flush to get prim_vbuf.c to release its allocation?? @@ -328,7 +331,7 @@ static void fse_run_linear_elts( struct draw_pt_middle_end *middle, hw_verts = draw->render->allocate_vertices( draw->render, (ushort)fse->key.output_stride, - (ushort)alloc_count ); + (ushort)count ); if (!hw_verts) { assert(0); diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index eb6988ff03..25118712a6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -51,7 +51,8 @@ struct fetch_pipeline_middle_end { static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, unsigned prim, - unsigned opt ) + unsigned opt, + unsigned *max_vertices ) { struct fetch_pipeline_middle_end *fpme = (struct fetch_pipeline_middle_end *)middle; struct draw_context *draw = fpme->draw; @@ -86,14 +87,21 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, (boolean)draw->rasterizer->gl_rasterization_rules ); - if (!(opt & PT_PIPELINE)) + if (!(opt & PT_PIPELINE)) { draw_pt_emit_prepare( fpme->emit, - prim ); + prim, + max_vertices ); + + *max_vertices = MAX2( *max_vertices, + DRAW_PIPE_MAX_VERTICES ); + } + else { + *max_vertices = DRAW_PIPE_MAX_VERTICES; + } /* No need to prepare the shader. */ vs->prepare(vs, draw); - } diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index f19e8850b3..f5495a80c7 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -45,6 +45,9 @@ struct varray_frontend { unsigned fetch_start; + unsigned driver_fetch_max; + unsigned fetch_max; + struct draw_pt_middle_end *middle; unsigned input_prim; @@ -183,7 +186,7 @@ static void varray_prepare(struct draw_pt_front_end *frontend, varray->output_prim = decompose_prim[prim]; varray->middle = middle; - middle->prepare(middle, varray->output_prim, opt); + middle->prepare(middle, varray->output_prim, opt, &varray->fetch_max ); } diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h index 114ed371a0..6e5e30f38e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -40,7 +40,7 @@ static void FUNC(struct draw_pt_front_end *frontend, case PIPE_PRIM_QUAD_STRIP: for (j = 0; j < count;) { unsigned remaining = count - j; - unsigned nr = trim( MIN2(FETCH_MAX, remaining), first, incr ); + unsigned nr = trim( MIN2(varray->fetch_max, remaining), first, incr ); varray_flush_linear(varray, start + j, nr); j += nr; if (nr != remaining) @@ -50,8 +50,9 @@ static void FUNC(struct draw_pt_front_end *frontend, case PIPE_PRIM_LINE_LOOP: if (count >= 2) { + unsigned fetch_max = MIN2(FETCH_MAX, varray->fetch_max); for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(FETCH_MAX, count - j); + unsigned end = MIN2(fetch_max, count - j); end -= (end % incr); for (i = 1; i < end; i++) { LINE(varray, i - 1, i); @@ -66,9 +67,10 @@ static void FUNC(struct draw_pt_front_end *frontend, case PIPE_PRIM_POLYGON: - case PIPE_PRIM_TRIANGLE_FAN: + case PIPE_PRIM_TRIANGLE_FAN: { + unsigned fetch_max = MIN2(FETCH_MAX, varray->fetch_max); for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(FETCH_MAX, count - j); + unsigned end = MIN2(fetch_max, count - j); end -= (end % incr); for (i = 2; i < end; i++) { TRIANGLE(varray, 0, i - 1, i); @@ -78,6 +80,7 @@ static void FUNC(struct draw_pt_front_end *frontend, varray_flush(varray); } break; + } default: assert(0); diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index ad86ab4292..2eafe270bc 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -36,8 +36,8 @@ #include "draw/draw_pt.h" -#define CACHE_MAX 1024 -#define FETCH_MAX 4096 +#define CACHE_MAX 256 +#define FETCH_MAX 256 #define DRAW_MAX (16*1024) struct vcache_frontend { @@ -52,6 +52,7 @@ struct vcache_frontend { unsigned draw_count; unsigned fetch_count; + unsigned fetch_max; struct draw_pt_middle_end *middle; @@ -296,10 +297,12 @@ static void vcache_check_run( struct draw_pt_front_end *frontend, ushort *storage = NULL; - if (0) debug_printf("fetch_count %d draw_count %d\n", fetch_count, draw_count); + if (0) debug_printf("fetch_count %d fetch_max %d draw_count %d\n", fetch_count, + vcache->fetch_max, + draw_count); if (max_index == 0xffffffff || - fetch_count >= FETCH_MAX || + fetch_count >= vcache->fetch_max || fetch_count > draw_count) { if (0) debug_printf("fail\n"); goto fail; @@ -409,7 +412,7 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, vcache->output_prim = draw_pt_reduced_prim(prim); vcache->middle = middle; - middle->prepare( middle, vcache->output_prim, opt ); + middle->prepare( middle, vcache->output_prim, opt, &vcache->fetch_max ); } -- cgit v1.2.3 From 43b92a6424a4d4f4f29b47c35092264c60822f1b Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 30 May 2008 18:36:16 +0200 Subject: gallium: Define PIPE_CAP_GUARD_BAND_* capabilities. --- src/gallium/include/pipe/p_defines.h | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index 83330ef22f..f8fadf31b6 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -274,6 +274,10 @@ enum pipe_texture_target { #define PIPE_CAP_MAX_POINT_WIDTH_AA 17 #define PIPE_CAP_MAX_TEXTURE_ANISOTROPY 18 #define PIPE_CAP_MAX_TEXTURE_LOD_BIAS 19 +#define PIPE_CAP_GUARD_BAND_LEFT 20 /*< float */ +#define PIPE_CAP_GUARD_BAND_TOP 21 /*< float */ +#define PIPE_CAP_GUARD_BAND_RIGHT 22 /*< float */ +#define PIPE_CAP_GUARD_BAND_BOTTOM 23 /*< float */ #ifdef __cplusplus -- cgit v1.2.3 From b98ac1d47257bf7b2661ae7c1a8904b7bc5d623c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 4 Jun 2008 15:49:18 +0100 Subject: draw: init vsvg draw pointer --- src/gallium/auxiliary/draw/draw_vs_sse.c | 6 ++++-- src/gallium/auxiliary/draw/draw_vs_varient.c | 5 ++++- 2 files changed, 8 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 0aa0c9a76b..c3189c707d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -158,8 +158,10 @@ draw_create_vs_sse(struct draw_context *draw, tgsi_scan_shader(templ->tokens, &vs->base.info); vs->base.draw = draw; - vs->base.create_varient = draw_vs_varient_aos_sse; -// vs->base.create_varient = draw_vs_varient_generic; + if (1) + vs->base.create_varient = draw_vs_varient_aos_sse; + else + vs->base.create_varient = draw_vs_varient_generic; vs->base.prepare = vs_sse_prepare; vs->base.run_linear = vs_sse_run_linear; vs->base.delete = vs_sse_delete; diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index 4a155251b5..3ceca5e849 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -124,6 +124,8 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, void *output_buffer) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; + + if (0) debug_printf("%s %d \n", __FUNCTION__, count); /* Want to do this in small batches for cache locality? */ @@ -181,7 +183,7 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; - //debug_printf("%s %d %d\n", __FUNCTION__, start, count); + if (0) debug_printf("%s %d %d\n", __FUNCTION__, start, count); vsvg->fetch->run( vsvg->fetch, @@ -257,6 +259,7 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, vsvg->base.run_linear = vsvg_run_linear; vsvg->base.destroy = vsvg_destroy; + vsvg->draw = vs->draw; /* Build free-standing fetch and emit functions: -- cgit v1.2.3 From 4ebfc3c8ffccadaed98c4e032b7691eaf299b0bc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 30 May 2008 15:23:03 -0600 Subject: egl: prototype some multi-API code --- src/gallium/winsys/egl_xlib/egl_xlib.c | 29 +++++++++++++++++++---------- 1 file changed, 19 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 99389970f5..ce3a0ae0da 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -313,19 +313,21 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, return EGL_NO_CONTEXT; } - if (ctx->Base.ClientAPI != EGL_OPENGL_API) { - _eglError(EGL_BAD_MATCH, "eglCreateContext(only OpenGL API supported)"); + /* API-dependent context creation */ + switch (ctx->Base.ClientAPI) { + case EGL_OPENGL_API: + /* create a softpipe context */ + ctx->pipe = softpipe_create(xdrv->screen, xdrv->winsys, NULL); + /* Now do xlib / state tracker inits here */ + _eglConfigToContextModesRec(conf, &visual); + ctx->Context = st_create_context(ctx->pipe, &visual, share_ctx); + break; + default: + _eglError(EGL_BAD_MATCH, "eglCreateContext(unsupported API)"); free(ctx); return EGL_NO_CONTEXT; } - /* create a softpipe context */ - ctx->pipe = softpipe_create(xdrv->screen, xdrv->winsys, NULL); - - /* Now do xlib / state tracker inits here */ - _eglConfigToContextModesRec(conf, &visual); - ctx->Context = st_create_context(ctx->pipe, &visual, share_ctx); - _eglSaveContext(&ctx->Base); return _eglGetContextHandle(&ctx->Base); @@ -341,7 +343,14 @@ xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx) context->Base.DeletePending = EGL_TRUE; } else { - st_destroy_context(context->Context); + /* API-dependent clean-up */ + switch (context->Base.ClientAPI) { + case EGL_OPENGL_API: + st_destroy_context(context->Context); + break; + default: + assert(0); + } free(context); } return EGL_TRUE; -- cgit v1.2.3 From 8223add3304451d5e75737a6d1be1739e4517943 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 4 Jun 2008 08:56:06 -0600 Subject: gallium: added tgsi_is_passthrough_shader() function Checks if all instructions are of the form MOV OUT[n], IN[n] Untested at this time. --- src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 83 +++++++++++++++++++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_scan.h | 4 ++ 2 files changed, 87 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c index bda7bc2e2e..240aaaf362 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c @@ -141,3 +141,86 @@ tgsi_scan_shader(const struct tgsi_token *tokens, tgsi_parse_free (&parse); } + + + +/** + * Check if the given shader is a "passthrough" shader consisting of only + * MOV instructions of the form: MOV OUT[n], IN[n] + * + */ +boolean +tgsi_is_passthrough_shader(const struct tgsi_token *tokens) +{ + struct tgsi_parse_context parse; + + /** + ** Setup to begin parsing input shader + **/ + if (tgsi_parse_init(&parse, tokens) != TGSI_PARSE_OK) { + debug_printf("tgsi_parse_init() failed in tgsi_is_passthrough_shader()!\n"); + return FALSE; + } + + /** + ** Loop over incoming program tokens/instructions + */ + while (!tgsi_parse_end_of_tokens(&parse)) { + + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *fullinst = + &parse.FullToken.FullInstruction; + const struct tgsi_full_src_register *src = + &fullinst->FullSrcRegisters[0]; + const struct tgsi_full_dst_register *dst = + &fullinst->FullDstRegisters[0]; + + /* Do a whole bunch of checks for a simple move */ + if (fullinst->Instruction.Opcode != TGSI_OPCODE_MOV || + src->SrcRegister.File != TGSI_FILE_INPUT || + dst->DstRegister.File != TGSI_FILE_OUTPUT || + src->SrcRegister.Index != dst->DstRegister.Index || + + src->SrcRegister.Negate || + src->SrcRegisterExtMod.Negate || + src->SrcRegisterExtMod.Absolute || + src->SrcRegisterExtMod.Scale2X || + src->SrcRegisterExtMod.Bias || + src->SrcRegisterExtMod.Complement || + + src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W || + + src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W || + + dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW) + { + tgsi_parse_free(&parse); + return FALSE; + } + } + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + /* fall-through */ + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* fall-through */ + default: + ; /* no-op */ + } + } + + tgsi_parse_free(&parse); + + /* if we get here, it's a pass-through shader */ + return TRUE; +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h index 0530bc6b51..5cb6efb343 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h @@ -67,4 +67,8 @@ tgsi_scan_shader(const struct tgsi_token *tokens, struct tgsi_shader_info *info); +extern boolean +tgsi_is_passthrough_shader(const struct tgsi_token *tokens); + + #endif /* TGSI_SCAN_H */ -- cgit v1.2.3 From fc286ff3ba089af5f1aca2d35b93b53a9383b1cf Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 4 Jun 2008 23:55:55 +0900 Subject: mesa: Remove assyntax.h include from generated matypes.h. matypes.h doesn't really depend on assyntax.h; assyntax.h is only present on x86; and this way we remove the need of grep being available (which is relevant for scons and non-unix platforms). --- src/mesa/x86-64/Makefile | 2 +- src/mesa/x86/3dnow_normal.S | 1 + src/mesa/x86/3dnow_xform1.S | 1 + src/mesa/x86/3dnow_xform2.S | 1 + src/mesa/x86/3dnow_xform3.S | 1 + src/mesa/x86/3dnow_xform4.S | 1 + src/mesa/x86/gen_matypes.c | 1 - src/mesa/x86/mmx_blend.S | 3 ++- src/mesa/x86/sse_normal.S | 1 + src/mesa/x86/sse_xform1.S | 1 + src/mesa/x86/sse_xform2.S | 1 + src/mesa/x86/sse_xform3.S | 1 + src/mesa/x86/sse_xform4.S | 1 + src/mesa/x86/x86_cliptest.S | 1 + src/mesa/x86/x86_xform2.S | 1 + src/mesa/x86/x86_xform3.S | 1 + src/mesa/x86/x86_xform4.S | 1 + 17 files changed, 17 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/x86-64/Makefile b/src/mesa/x86-64/Makefile index 252218ca86..ab58aa6c23 100644 --- a/src/mesa/x86-64/Makefile +++ b/src/mesa/x86-64/Makefile @@ -24,6 +24,6 @@ clean: # need some special rules here, unfortunately matypes.h: ../main/mtypes.h ../tnl/t_context.h ../x86/gen_matypes - ../x86/gen_matypes | grep -v '#include "assyntax.h' > matypes.h + ../x86/gen_matypes > matypes.h xform4.o: matypes.h diff --git a/src/mesa/x86/3dnow_normal.S b/src/mesa/x86/3dnow_normal.S index 693a7864db..7f5f6b357f 100644 --- a/src/mesa/x86/3dnow_normal.S +++ b/src/mesa/x86/3dnow_normal.S @@ -29,6 +29,7 @@ #ifdef USE_3DNOW_ASM +#include "assyntax.h" #include "matypes.h" #include "norm_args.h" diff --git a/src/mesa/x86/3dnow_xform1.S b/src/mesa/x86/3dnow_xform1.S index 7665c0ff8b..a73301a8d6 100644 --- a/src/mesa/x86/3dnow_xform1.S +++ b/src/mesa/x86/3dnow_xform1.S @@ -24,6 +24,7 @@ */ #ifdef USE_3DNOW_ASM +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/3dnow_xform2.S b/src/mesa/x86/3dnow_xform2.S index b201d1e901..2988fb7bfd 100644 --- a/src/mesa/x86/3dnow_xform2.S +++ b/src/mesa/x86/3dnow_xform2.S @@ -24,6 +24,7 @@ */ #ifdef USE_3DNOW_ASM +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/3dnow_xform3.S b/src/mesa/x86/3dnow_xform3.S index 46f155697d..a356aaee76 100644 --- a/src/mesa/x86/3dnow_xform3.S +++ b/src/mesa/x86/3dnow_xform3.S @@ -24,6 +24,7 @@ */ #ifdef USE_3DNOW_ASM +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/3dnow_xform4.S b/src/mesa/x86/3dnow_xform4.S index a0c6b193cd..b2b7c64f23 100644 --- a/src/mesa/x86/3dnow_xform4.S +++ b/src/mesa/x86/3dnow_xform4.S @@ -24,6 +24,7 @@ */ #ifdef USE_3DNOW_ASM +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/gen_matypes.c b/src/mesa/x86/gen_matypes.c index e95f3d4f3d..afb4b11529 100644 --- a/src/mesa/x86/gen_matypes.c +++ b/src/mesa/x86/gen_matypes.c @@ -92,7 +92,6 @@ int main( int argc, char **argv ) printf( "#ifndef __ASM_TYPES_H__\n" ); printf( "#define __ASM_TYPES_H__\n" ); printf( "\n" ); - printf( "#include \"assyntax.h\"\n" ); /* GLcontext offsets: diff --git a/src/mesa/x86/mmx_blend.S b/src/mesa/x86/mmx_blend.S index 20ac5a20ad..eeaf43ea93 100644 --- a/src/mesa/x86/mmx_blend.S +++ b/src/mesa/x86/mmx_blend.S @@ -1,10 +1,11 @@ ; /* - * Written by José Fonseca + * Written by Jos� Fonseca */ #ifdef USE_MMX_ASM +#include "assyntax.h" #include "matypes.h" /* integer multiplication - alpha plus one diff --git a/src/mesa/x86/sse_normal.S b/src/mesa/x86/sse_normal.S index 1c32e3b2fe..a8c0d38c7e 100644 --- a/src/mesa/x86/sse_normal.S +++ b/src/mesa/x86/sse_normal.S @@ -31,6 +31,7 @@ */ #ifdef USE_SSE_ASM +#include "assyntax.h" #include "matypes.h" #include "norm_args.h" diff --git a/src/mesa/x86/sse_xform1.S b/src/mesa/x86/sse_xform1.S index 22fd8dd27b..4aa9de607c 100644 --- a/src/mesa/x86/sse_xform1.S +++ b/src/mesa/x86/sse_xform1.S @@ -31,6 +31,7 @@ */ #ifdef USE_SSE_ASM +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/sse_xform2.S b/src/mesa/x86/sse_xform2.S index 52eeb27ef5..91b82e7297 100644 --- a/src/mesa/x86/sse_xform2.S +++ b/src/mesa/x86/sse_xform2.S @@ -31,6 +31,7 @@ */ #ifdef USE_SSE_ASM +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/sse_xform3.S b/src/mesa/x86/sse_xform3.S index 5e0cd8b666..1fc79ef21b 100644 --- a/src/mesa/x86/sse_xform3.S +++ b/src/mesa/x86/sse_xform3.S @@ -31,6 +31,7 @@ */ #ifdef USE_SSE_ASM +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/sse_xform4.S b/src/mesa/x86/sse_xform4.S index 13680528db..fb1fa741c0 100644 --- a/src/mesa/x86/sse_xform4.S +++ b/src/mesa/x86/sse_xform4.S @@ -24,6 +24,7 @@ */ #ifdef USE_SSE_ASM +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/x86_cliptest.S b/src/mesa/x86/x86_cliptest.S index c7a3a9b57e..e413aee61e 100644 --- a/src/mesa/x86/x86_cliptest.S +++ b/src/mesa/x86/x86_cliptest.S @@ -29,6 +29,7 @@ * in there will break the build on some platforms. */ +#include "assyntax.h" #include "matypes.h" #include "clip_args.h" diff --git a/src/mesa/x86/x86_xform2.S b/src/mesa/x86/x86_xform2.S index e41661d546..980725ef51 100644 --- a/src/mesa/x86/x86_xform2.S +++ b/src/mesa/x86/x86_xform2.S @@ -29,6 +29,7 @@ * in there will break the build on some platforms. */ +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/x86_xform3.S b/src/mesa/x86/x86_xform3.S index 067ddd7d7c..1c782f1c5c 100644 --- a/src/mesa/x86/x86_xform3.S +++ b/src/mesa/x86/x86_xform3.S @@ -29,6 +29,7 @@ * in there will break the build on some platforms. */ +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" diff --git a/src/mesa/x86/x86_xform4.S b/src/mesa/x86/x86_xform4.S index 77621ac4bd..97a841138e 100644 --- a/src/mesa/x86/x86_xform4.S +++ b/src/mesa/x86/x86_xform4.S @@ -29,6 +29,7 @@ * in there will break the build on some platforms. */ +#include "assyntax.h" #include "matypes.h" #include "xform_args.h" -- cgit v1.2.3 From 5c9c6b0d2398715d035e939eb8672ed7e95cfec5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 4 Jun 2008 23:56:57 +0900 Subject: scons: Generate matypes.h at build time. Also cleanup mesa SConscript. --- src/mesa/SConscript | 207 ++++++++++++++++++++++++++++------------------------ 1 file changed, 110 insertions(+), 97 deletions(-) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index 04b771a7e1..1beb81c595 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -8,7 +8,6 @@ if env['platform'] != 'winddk': env = env.Clone() - # Includes env.Append(CPPPATH = [ '#/src/mesa', '#/src/mesa/main', @@ -18,21 +17,12 @@ if env['platform'] != 'winddk': env.Append(CFLAGS = [ '-std=c99', ]) + + # + # Source files + # - # x86 assembly - if x86 and gcc: - env.Append(CPPDEFINES = [ - 'USE_X86_ASM', - 'USE_MMX_ASM', - 'USE_3DNOW_ASM', - 'USE_SSE_ASM', - ]) - - - ####################################################################### - # Core sources - - MAIN_SOURCES = [ + main_sources = [ 'main/api_arrayelt.c', 'main/api_loopback.c', 'main/api_noop.c', @@ -98,13 +88,7 @@ if env['platform'] != 'winddk': 'main/vtxfmt.c', ] - GLAPI_SOURCES = [ - 'main/dispatch.c', - 'glapi/glapi.c', - 'glapi/glthread.c', - ] - - MATH_SOURCES = [ + math_sources = [ 'math/m_debug_clip.c', 'math/m_debug_norm.c', 'math/m_debug_xform.c', @@ -115,7 +99,7 @@ if env['platform'] != 'winddk': 'math/m_xform.c', ] - VBO_SOURCES = [ + vbo_sources = [ 'vbo/vbo_context.c', 'vbo/vbo_exec.c', 'vbo/vbo_exec_api.c', @@ -132,13 +116,13 @@ if env['platform'] != 'winddk': 'vbo/vbo_save_loopback.c', ] - VF_SOURCES = [ + vf_sources = [ 'vf/vf.c', 'vf/vf_generic.c', 'vf/vf_sse.c', ] - STATETRACKER_SOURCES = [ + statetracker_sources = [ 'state_tracker/st_atom.c', 'state_tracker/st_atom_blend.c', 'state_tracker/st_atom_clip.c', @@ -180,7 +164,7 @@ if env['platform'] != 'winddk': 'state_tracker/st_texture.c', ] - SHADER_SOURCES = [ + shader_sources = [ 'shader/arbprogparse.c', 'shader/arbprogram.c', 'shader/atifragshader.c', @@ -201,7 +185,7 @@ if env['platform'] != 'winddk': 'shader/shader_api.c', ] - SLANG_SOURCES = [ + slang_sources = [ 'shader/slang/slang_builtin.c', 'shader/slang/slang_codegen.c', 'shader/slang/slang_compile.c', @@ -225,88 +209,117 @@ if env['platform'] != 'winddk': 'shader/slang/slang_utility.c', ] - - ####################################################################### - # Assembly sources - - ASM_C_SOURCES = [ - 'x86/common_x86.c', - 'x86/x86.c', - 'x86/3dnow.c', - 'x86/sse.c', - 'sparc/sparc.c', - 'ppc/common_ppc.c', - 'x86-64/x86-64.c', - ] - - X86_SOURCES = [ - 'x86/common_x86_asm.S', - 'x86/x86_xform2.S', - 'x86/x86_xform3.S', - 'x86/x86_xform4.S', - 'x86/x86_cliptest.S', - 'x86/mmx_blend.S', - 'x86/3dnow_xform1.S', - 'x86/3dnow_xform2.S', - 'x86/3dnow_xform3.S', - 'x86/3dnow_xform4.S', - 'x86/3dnow_normal.S', - 'x86/sse_xform1.S', - 'x86/sse_xform2.S', - 'x86/sse_xform3.S', - 'x86/sse_xform4.S', - 'x86/sse_normal.S', - 'x86/read_rgba_span_x86.S', - ] - - X86_API = [ - 'x86/glapi_x86.S', - ] - - X86_64_SOURCES = [ - 'x86-64/xform4.S', - ] - - X86_64_API = [ - 'x86-64/glapi_x86-64.S', - ] - - SPARC_SOURCES = [ - 'sparc/clip.S', - 'sparc/norm.S', - 'sparc/xform.S', - ] - - SPARC_API = [ - 'sparc/glapi_sparc.S', + mesa_sources = ( + main_sources + + math_sources + + vbo_sources + + vf_sources + + statetracker_sources + + shader_sources + + slang_sources + ) + + glapi_sources = [ + 'main/dispatch.c', + 'glapi/glapi.c', + 'glapi/glthread.c', ] - if x86 and gcc: - ASM_SOURCES = ASM_C_SOURCES + X86_SOURCES - API_SOURCES = X86_API + # + # Assembly sources + # + if gcc and env['machine'] == 'x86': + env.Append(CPPDEFINES = [ + 'USE_X86_ASM', + 'USE_MMX_ASM', + 'USE_3DNOW_ASM', + 'USE_SSE_ASM', + ]) + mesa_sources += [ + 'x86/common_x86.c', + 'x86/x86.c', + 'x86/3dnow.c', + 'x86/sse.c', + 'x86/common_x86_asm.S', + 'x86/x86_xform2.S', + 'x86/x86_xform3.S', + 'x86/x86_xform4.S', + 'x86/x86_cliptest.S', + 'x86/mmx_blend.S', + 'x86/3dnow_xform1.S', + 'x86/3dnow_xform2.S', + 'x86/3dnow_xform3.S', + 'x86/3dnow_xform4.S', + 'x86/3dnow_normal.S', + 'x86/sse_xform1.S', + 'x86/sse_xform2.S', + 'x86/sse_xform3.S', + 'x86/sse_xform4.S', + 'x86/sse_normal.S', + 'x86/read_rgba_span_x86.S', + ] + glapi_sources += [ + 'x86/glapi_x86.S', + ] + elif gcc and env['machine'] == 'x86_64': + env.Append(CPPDEFINES = [ + 'USE_X86_64_ASM', + ]) + mesa_sources += [ + 'x86-64/x86-64.c', + 'x86-64/xform4.S', + ] + glapi_sources += [ + 'x86-64/glapi_x86-64.S' + ] + elif gcc and env['machine'] == 'ppc': + mesa_sources += [ + 'ppc/common_ppc.c', + ] + glapi_sources += [ + ] + elif gcc and env['machine'] == 'sparc': + mesa_sources += [ + 'sparc/sparc.c', + 'sparc/clip.S', + 'sparc/norm.S', + 'sparc/xform.S', + ] + glapi_sources += [ + 'sparc/glapi_sparc.S' + ] else: - ASM_SOURCES = [] - API_SOURCES = [] + pass - SOLO_SOURCES = \ - MAIN_SOURCES + \ - MATH_SOURCES + \ - VBO_SOURCES + \ - VF_SOURCES + \ - STATETRACKER_SOURCES + \ - SHADER_SOURCES + \ - ASM_SOURCES + \ - SLANG_SOURCES + # Generate matypes.h + if gcc and env['machine'] in ('x86', 'x86_64'): + # See http://www.scons.org/wiki/UsingCodeGenerators + gen_matypes = env.Program( + target = 'gen_matypes', + source = 'x86/gen_matypes.c', + ) + matypes = env.Command( + 'matypes.h', + gen_matypes, + gen_matypes[0].abspath + ' > $TARGET', + ) + # Add the dir containing the generated header (somewhere inside the + # build dir) to the include path + env.Append(CPPPATH = [matypes[0].dir]) + # + # Libraries + # + mesa = env.ConvenienceLibrary( target = 'mesa', - source = SOLO_SOURCES, + source = mesa_sources, ) Export('mesa') if not dri: glapi = env.ConvenienceLibrary( target = 'glapi', - source = GLAPI_SOURCES + API_SOURCES, + source = glapi_sources, ) Export('glapi') -- cgit v1.2.3 From 19ca6ce573b409346a397453e7d4fa43a0f98368 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 4 Jun 2008 10:09:49 -0600 Subject: egl: updated include paths (s/GLES/EGL/) --- src/egl/main/egltypedefs.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/main/egltypedefs.h b/src/egl/main/egltypedefs.h index efbb17a4dc..9fbc55352c 100644 --- a/src/egl/main/egltypedefs.h +++ b/src/egl/main/egltypedefs.h @@ -3,8 +3,8 @@ #define EGL_EGLEXT_PROTOTYPES -#include -#include +#include +#include typedef struct _egl_api _EGLAPI; -- cgit v1.2.3 From 97035cb19aaf508aad45446651a80da9af1d0e8c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 4 Jun 2008 11:34:10 -0600 Subject: egl: in _eglAddConfig() just save a pointer to the config; don't copy the config This allows subclassing by drivers. --- src/egl/main/eglconfig.c | 25 +++++++++++++------------ src/egl/main/eglconfig.h | 2 +- src/egl/main/egldisplay.c | 18 ++++++++++++++++-- src/egl/main/egldisplay.h | 2 +- 4 files changed, 31 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c index c6369e7adf..794a783b9e 100644 --- a/src/egl/main/eglconfig.c +++ b/src/egl/main/eglconfig.c @@ -78,8 +78,8 @@ _eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config) EGLint i; _EGLDisplay *disp = _eglLookupDisplay(dpy); for (i = 0; i < disp->NumConfigs; i++) { - if (disp->Configs[i].Handle == config) { - return disp->Configs + i; + if (disp->Configs[i]->Handle == config) { + return disp->Configs[i]; } } return NULL; @@ -88,23 +88,24 @@ _eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config) /** * Add the given _EGLConfig to the given display. + * Note that we just save the ptr to the config (we don't copy the config). */ _EGLConfig * -_eglAddConfig(_EGLDisplay *display, const _EGLConfig *config) +_eglAddConfig(_EGLDisplay *display, _EGLConfig *config) { - _EGLConfig *newConfigs; + _EGLConfig **newConfigs; EGLint n; n = display->NumConfigs; - newConfigs = (_EGLConfig *) realloc(display->Configs, - (n + 1) * sizeof(_EGLConfig)); + /* realloc array of ptrs */ + newConfigs = (_EGLConfig **) realloc(display->Configs, + (n + 1) * sizeof(_EGLConfig *)); if (newConfigs) { display->Configs = newConfigs; - display->Configs[n] = *config; /* copy struct */ - display->Configs[n].Handle = (EGLConfig) n; + display->Configs[n] = config; display->NumConfigs++; - return display->Configs + n; + return config; } else { return NULL; @@ -330,8 +331,8 @@ _eglChooseConfig(_EGLDriver *drv, EGLDisplay dpy, const EGLint *attrib_list, /* make array of pointers to qualifying configs */ for (i = count = 0; i < disp->NumConfigs && count < config_size; i++) { - if (_eglConfigQualifies(disp->Configs + i, &criteria)) { - configList[count++] = disp->Configs + i; + if (_eglConfigQualifies(disp->Configs[i], &criteria)) { + configList[count++] = disp->Configs[i]; } } @@ -389,7 +390,7 @@ _eglGetConfigs(_EGLDriver *drv, EGLDisplay dpy, EGLConfig *configs, EGLint i; *num_config = MIN2(disp->NumConfigs, config_size); for (i = 0; i < *num_config; i++) { - configs[i] = disp->Configs[i].Handle; + configs[i] = disp->Configs[i]->Handle; } } else { diff --git a/src/egl/main/eglconfig.h b/src/egl/main/eglconfig.h index b10a61985d..d12f66245c 100644 --- a/src/egl/main/eglconfig.h +++ b/src/egl/main/eglconfig.h @@ -34,7 +34,7 @@ _eglLookupConfig(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config); extern _EGLConfig * -_eglAddConfig(_EGLDisplay *display, const _EGLConfig *config); +_eglAddConfig(_EGLDisplay *display, _EGLConfig *config); extern EGLBoolean diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c index 8fd29b8421..540efd4fee 100644 --- a/src/egl/main/egldisplay.c +++ b/src/egl/main/egldisplay.c @@ -96,11 +96,25 @@ _eglGetCurrentDisplay(void) } +/** + * Free all the data hanging of an _EGLDisplay object, but not + * the object itself. + */ void _eglCleanupDisplay(_EGLDisplay *disp) { - /* XXX incomplete */ + EGLint i; + + for (i = 0; i < disp->NumConfigs; i++) { + free(disp->Configs[i]); + } free(disp->Configs); + disp->Configs = NULL; + + /* XXX incomplete */ + free((void *) disp->DriverName); - /* driver deletes _EGLDisplay */ + disp->DriverName = NULL; + + /* driver deletes the _EGLDisplay object */ } diff --git a/src/egl/main/egldisplay.h b/src/egl/main/egldisplay.h index e2ebab0b21..ff623ee1c6 100644 --- a/src/egl/main/egldisplay.h +++ b/src/egl/main/egldisplay.h @@ -21,7 +21,7 @@ struct _egl_display _EGLScreen **Screens; /* array [NumScreens] */ EGLint NumConfigs; - _EGLConfig *Configs; /* array [NumConfigs] */ + _EGLConfig **Configs; /* array [NumConfigs] of ptr to _EGLConfig */ #ifdef _EGL_PLATFORM_X Display *Xdpy; -- cgit v1.2.3 From af31e5d42982088821bdf8deaeb33d45f01fd004 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 4 Jun 2008 11:35:41 -0600 Subject: egl: do proper setup/init of EGL configs --- src/gallium/winsys/egl_xlib/egl_xlib.c | 73 ++++++++++++++++++++++++---------- 1 file changed, 52 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index ce3a0ae0da..07b3c0eb00 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -115,34 +115,68 @@ lookup_context(EGLContext surf) } +static unsigned int +bitcount(unsigned int n) +{ + unsigned int bits; + for (bits = 0; n > 0; n = n >> 1) { + bits += (n & 1); + } + return bits; +} + + /** - * XXX temporary - * Need to query X server's GLX visuals. + * Create the EGLConfigs. (one per X visual) */ static void -init_configs(_EGLDriver *drv, EGLDisplay dpy) +create_configs(_EGLDriver *drv, EGLDisplay dpy) { _EGLDisplay *disp = _eglLookupDisplay(dpy); - int i; + XVisualInfo *visInfo, visTemplate; + int num_visuals, i; + + /* get list of all X visuals, create an EGL config for each */ + visTemplate.screen = DefaultScreen(disp->Xdpy); + visInfo = XGetVisualInfo(disp->Xdpy, VisualScreenMask, + &visTemplate, &num_visuals); + if (!visInfo) { + printf("egl_xlib.c: couldn't get any X visuals\n"); + abort(); + } - for (i = 0; i < 2; i++) { - _EGLConfig config; + for (i = 0; i < num_visuals; i++) { + _EGLConfig *config = calloc(1, sizeof(_EGLConfig)); int id = i + 1; - _eglInitConfig(&config, id); - SET_CONFIG_ATTRIB(&config, EGL_RED_SIZE, 8); - SET_CONFIG_ATTRIB(&config, EGL_GREEN_SIZE, 8); - SET_CONFIG_ATTRIB(&config, EGL_BLUE_SIZE, 8); - SET_CONFIG_ATTRIB(&config, EGL_ALPHA_SIZE, 8); - if (i > 0) { - SET_CONFIG_ATTRIB(&config, EGL_DEPTH_SIZE, 24); - SET_CONFIG_ATTRIB(&config, EGL_STENCIL_SIZE, 8); - } - _eglAddConfig(disp, &config); + int rbits = bitcount(visInfo[i].red_mask); + int gbits = bitcount(visInfo[i].green_mask); + int bbits = bitcount(visInfo[i].blue_mask); + int abits = bbits == 8 ? 8 : 0; + int zbits = 24; + int sbits = 8; + int visid = visInfo[i].visualid; +#if defined(__cplusplus) || defined(c_plusplus) + int vistype = visInfo[i].c_class; +#else + int vistype = visInfo[i].class; +#endif + + _eglInitConfig(config, id); + SET_CONFIG_ATTRIB(config, EGL_BUFFER_SIZE, rbits + gbits + bbits + abits); + SET_CONFIG_ATTRIB(config, EGL_RED_SIZE, rbits); + SET_CONFIG_ATTRIB(config, EGL_GREEN_SIZE, gbits); + SET_CONFIG_ATTRIB(config, EGL_BLUE_SIZE, bbits); + SET_CONFIG_ATTRIB(config, EGL_ALPHA_SIZE, abits); + SET_CONFIG_ATTRIB(config, EGL_DEPTH_SIZE, zbits); + SET_CONFIG_ATTRIB(config, EGL_STENCIL_SIZE, sbits); + SET_CONFIG_ATTRIB(config, EGL_NATIVE_VISUAL_ID, visid); + SET_CONFIG_ATTRIB(config, EGL_NATIVE_VISUAL_TYPE, vistype); + + _eglAddConfig(disp, config); } } - /** * Called via eglInitialize(), drv->API.Initialize(). */ @@ -150,10 +184,7 @@ static EGLBoolean xlib_eglInitialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *minor, EGLint *major) { - /* visual configs */ - - init_configs(drv, dpy); - + create_configs(drv, dpy); drv->Initialized = EGL_TRUE; -- cgit v1.2.3 From 25641d599fcf770e214bf078ee35b6aae1d7c73e Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Wed, 4 Jun 2008 21:25:08 +0100 Subject: Set the attribute as used. --- src/mesa/shader/slang/slang_link.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 80cd4b6df6..dd4990ba02 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -262,6 +262,9 @@ _slang_resolve_attributes(struct gl_shader_program *shProg, return GL_FALSE; } _mesa_add_attribute(shProg->Attributes, name, size, attr); + + /* set the attribute as used */ + usedAttributes |= 1<SrcReg[j].Index = VERT_ATTRIB_GENERIC0 + attr; -- cgit v1.2.3 From f6bb2f87b96bd05661020cf8e73e80f5ef25353a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 4 Jun 2008 11:50:18 -0600 Subject: egl: minor tweaks --- src/egl/drivers/dri/egldri.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/dri/egldri.c b/src/egl/drivers/dri/egldri.c index 1a922a7471..57661cc3ab 100644 --- a/src/egl/drivers/dri/egldri.c +++ b/src/egl/drivers/dri/egldri.c @@ -82,7 +82,7 @@ _EGLDriver * _eglMain(_EGLDisplay *dpy, const char *args) { #if 1 - const int card = atoi(args); + const int card = args ? atoi(args) : 0; _EGLDriver *driver = NULL; char driverName[1000]; @@ -1086,8 +1086,10 @@ _eglDRICreateDisplay(driDisplay *dpy, __DRIframebuffer *framebuffer) api_ver, & interface_methods, NULL); - if (!dpy->driScreen.private) + if (!dpy->driScreen.private) { + _eglLog(_EGL_WARNING, "egldri.c: DRI create new screen failed"); return EGL_FALSE; + } DRM_UNLOCK( dpy->drmFD, dpy->pSAREA, dpy->serverContext ); -- cgit v1.2.3 From feb722fa98f04a4487b7ec4746bcc8c7296899c8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 5 Jun 2008 12:01:00 -0600 Subject: mesa: added _mesa_DrawArrays, DrawElements, DrawRangeElements() wrappers for VBO funcs --- src/mesa/main/varray.h | 14 ++++++++++++++ src/mesa/vbo/vbo_exec_array.c | 27 +++++++++++++++++++++++++++ 2 files changed, 41 insertions(+) (limited to 'src') diff --git a/src/mesa/main/varray.h b/src/mesa/main/varray.h index 3589735f68..ba91ecf5f6 100644 --- a/src/mesa/main/varray.h +++ b/src/mesa/main/varray.h @@ -147,6 +147,20 @@ _mesa_MultiModeDrawElementsIBM( const GLenum * mode, const GLsizei * count, GLsizei primcount, GLint modestride ); + +extern void GLAPIENTRY +_mesa_DrawArrays(GLenum mode, GLint first, GLsizei count); + +extern void GLAPIENTRY +_mesa_DrawElements(GLenum mode, GLsizei count, GLenum type, + const GLvoid *indices); + +extern void GLAPIENTRY +_mesa_DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count, + GLenum type, const GLvoid *indices); + + + extern void _mesa_init_varray( GLcontext * ctx ); diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c index dbee2188ee..e3d2fc51cb 100644 --- a/src/mesa/vbo/vbo_exec_array.c +++ b/src/mesa/vbo/vbo_exec_array.c @@ -30,6 +30,7 @@ #include "main/state.h" #include "main/api_validate.h" #include "main/api_noop.h" +#include "main/varray.h" #include "glapi/dispatch.h" #include "vbo_context.h" @@ -400,3 +401,29 @@ void vbo_exec_array_destroy( struct vbo_exec_context *exec ) { /* nothing to do */ } + + +/* This API entrypoint is not ordinarily used */ +void GLAPIENTRY +_mesa_DrawArrays(GLenum mode, GLint first, GLsizei count) +{ + vbo_exec_DrawArrays(mode, first, count); +} + + +/* This API entrypoint is not ordinarily used */ +void GLAPIENTRY +_mesa_DrawElements(GLenum mode, GLsizei count, GLenum type, + const GLvoid *indices) +{ + vbo_exec_DrawElements(mode, count, type, indices); +} + + +/* This API entrypoint is not ordinarily used */ +void GLAPIENTRY +_mesa_DrawRangeElements(GLenum mode, GLuint start, GLuint end, GLsizei count, + GLenum type, const GLvoid *indices) +{ + vbo_exec_DrawRangeElements(mode, start, end, count, type, indices); +} -- cgit v1.2.3 From 16e8ee33bd3db1b14a97b3ddd2a7a8833851f8fb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 5 Jun 2008 12:08:19 -0600 Subject: mesa: remove EXT/NV suffixes from _mesa_PointParameter functions --- src/mesa/main/attrib.c | 24 ++++++++++++------------ src/mesa/main/points.c | 29 ++++++++--------------------- src/mesa/main/points.h | 8 ++++---- src/mesa/main/state.c | 8 ++++---- 4 files changed, 28 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index b422198f92..39666d175a 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -1074,14 +1074,14 @@ _mesa_PopAttrib(void) _mesa_PointSize(point->Size); _mesa_set_enable(ctx, GL_POINT_SMOOTH, point->SmoothFlag); if (ctx->Extensions.EXT_point_parameters) { - _mesa_PointParameterfvEXT(GL_DISTANCE_ATTENUATION_EXT, - point->Params); - _mesa_PointParameterfEXT(GL_POINT_SIZE_MIN_EXT, - point->MinSize); - _mesa_PointParameterfEXT(GL_POINT_SIZE_MAX_EXT, - point->MaxSize); - _mesa_PointParameterfEXT(GL_POINT_FADE_THRESHOLD_SIZE_EXT, - point->Threshold); + _mesa_PointParameterfv(GL_DISTANCE_ATTENUATION_EXT, + point->Params); + _mesa_PointParameterf(GL_POINT_SIZE_MIN_EXT, + point->MinSize); + _mesa_PointParameterf(GL_POINT_SIZE_MAX_EXT, + point->MaxSize); + _mesa_PointParameterf(GL_POINT_FADE_THRESHOLD_SIZE_EXT, + point->Threshold); } if (ctx->Extensions.NV_point_sprite || ctx->Extensions.ARB_point_sprite) { @@ -1091,10 +1091,10 @@ _mesa_PopAttrib(void) (GLint) point->CoordReplace[u]); } _mesa_set_enable(ctx, GL_POINT_SPRITE_NV,point->PointSprite); - _mesa_PointParameteriNV(GL_POINT_SPRITE_R_MODE_NV, - ctx->Point.SpriteRMode); - _mesa_PointParameterfEXT(GL_POINT_SPRITE_COORD_ORIGIN, - (GLfloat)ctx->Point.SpriteOrigin); + _mesa_PointParameteri(GL_POINT_SPRITE_R_MODE_NV, + ctx->Point.SpriteRMode); + _mesa_PointParameterf(GL_POINT_SPRITE_COORD_ORIGIN, + (GLfloat)ctx->Point.SpriteOrigin); } } break; diff --git a/src/mesa/main/points.c b/src/mesa/main/points.c index e83db5de78..fbedbcb22c 100644 --- a/src/mesa/main/points.c +++ b/src/mesa/main/points.c @@ -65,45 +65,32 @@ _mesa_PointSize( GLfloat size ) #if _HAVE_FULL_GL -/* - * Added by GL_NV_point_sprite - */ + void GLAPIENTRY -_mesa_PointParameteriNV( GLenum pname, GLint param ) +_mesa_PointParameteri( GLenum pname, GLint param ) { const GLfloat value = (GLfloat) param; - _mesa_PointParameterfvEXT(pname, &value); + _mesa_PointParameterfv(pname, &value); } -/* - * Added by GL_NV_point_sprite - */ void GLAPIENTRY -_mesa_PointParameterivNV( GLenum pname, const GLint *params ) +_mesa_PointParameteriv( GLenum pname, const GLint *params ) { const GLfloat value = (GLfloat) params[0]; - _mesa_PointParameterfvEXT(pname, &value); + _mesa_PointParameterfv(pname, &value); } - -/* - * Same for both GL_EXT_point_parameters and GL_ARB_point_parameters. - */ void GLAPIENTRY -_mesa_PointParameterfEXT( GLenum pname, GLfloat param) +_mesa_PointParameterf( GLenum pname, GLfloat param) { - _mesa_PointParameterfvEXT(pname, ¶m); + _mesa_PointParameterfv(pname, ¶m); } - -/* - * Same for both GL_EXT_point_parameters and GL_ARB_point_parameters. - */ void GLAPIENTRY -_mesa_PointParameterfvEXT( GLenum pname, const GLfloat *params) +_mesa_PointParameterfv( GLenum pname, const GLfloat *params) { GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); diff --git a/src/mesa/main/points.h b/src/mesa/main/points.h index 951ff677db..156641eab9 100644 --- a/src/mesa/main/points.h +++ b/src/mesa/main/points.h @@ -39,16 +39,16 @@ extern void GLAPIENTRY _mesa_PointSize( GLfloat size ); extern void GLAPIENTRY -_mesa_PointParameteriNV( GLenum pname, GLint param ); +_mesa_PointParameteri( GLenum pname, GLint param ); extern void GLAPIENTRY -_mesa_PointParameterivNV( GLenum pname, const GLint *params ); +_mesa_PointParameteriv( GLenum pname, const GLint *params ); extern void GLAPIENTRY -_mesa_PointParameterfEXT( GLenum pname, GLfloat param ); +_mesa_PointParameterf( GLenum pname, GLfloat param ); extern void GLAPIENTRY -_mesa_PointParameterfvEXT( GLenum pname, const GLfloat *params ); +_mesa_PointParameterfv( GLenum pname, const GLfloat *params ); extern void _mesa_init_point( GLcontext * ctx ); diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index cdf1249cd0..7e8f964bb1 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -457,8 +457,8 @@ _mesa_init_exec_table(struct _glapi_table *exec) /* 54. GL_EXT_point_parameters */ #if _HAVE_FULL_GL - SET_PointParameterfEXT(exec, _mesa_PointParameterfEXT); - SET_PointParameterfvEXT(exec, _mesa_PointParameterfvEXT); + SET_PointParameterfEXT(exec, _mesa_PointParameterf); + SET_PointParameterfvEXT(exec, _mesa_PointParameterfv); #endif /* 97. GL_EXT_compiled_vertex_array */ @@ -571,8 +571,8 @@ _mesa_init_exec_table(struct _glapi_table *exec) /* 262. GL_NV_point_sprite */ #if _HAVE_FULL_GL - SET_PointParameteriNV(exec, _mesa_PointParameteriNV); - SET_PointParameterivNV(exec, _mesa_PointParameterivNV); + SET_PointParameteriNV(exec, _mesa_PointParameteri); + SET_PointParameterivNV(exec, _mesa_PointParameteriv); #endif /* 268. GL_EXT_stencil_two_side */ -- cgit v1.2.3 From cca1ae79aa8a06205c93a39be780b32adb194022 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 5 Jun 2008 14:10:42 -0600 Subject: mesa: Makefile comment --- src/mesa/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 2403223db2..d663cbabfe 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -38,6 +38,7 @@ libmesa.a: $(SOLO_OBJECTS) mimeset -f "$@" ; \ fi +# Make archive of gl* API dispatcher functions (when not building DRI/GLX) libglapi.a: $(GLAPI_OBJECTS) @ $(TOP)/bin/mklib -o glapi -static $(GLAPI_OBJECTS) -- cgit v1.2.3 From d5cf57de2de9692e9dc194bea1bbf8abfd446770 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 5 Jun 2008 14:11:39 -0600 Subject: egl: new src/egl/drivers/Makefile --- src/egl/drivers/Makefile | 42 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 42 insertions(+) create mode 100644 src/egl/drivers/Makefile (limited to 'src') diff --git a/src/egl/drivers/Makefile b/src/egl/drivers/Makefile new file mode 100644 index 0000000000..837626e3d6 --- /dev/null +++ b/src/egl/drivers/Makefile @@ -0,0 +1,42 @@ +# src/egl/drivers/Makefile + +TOP = ../../.. + +SUBDIRS = demo dri + + +default: conditional_subdirs + + +# depending on $DRIVER_DIRS... +conditional_subdirs: + @echo "DRIVER_DIRS = " $(DRIVER_DIRS) + @case "$$DRIVER_DIRS" in \ + dri) $(MAKE) dri_subdir || exit 1 ;; \ + *) ;; \ + esac; + + +dri_subdir: + (cd dri ; $(MAKE)) || exit 1 ; \ + + +demo_subdir: + (cd demo ; $(MAKE)) || exit 1 ; \ + + + +subdirs: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir ; $(MAKE)) || exit 1 ; \ + fi \ + done + + +clean: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir ; $(MAKE) clean) ; \ + fi \ + done -- cgit v1.2.3 From 3b28aab1614bb22ca12ce7ebd5aa5ff0d87309ec Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 5 Jun 2008 14:11:54 -0600 Subject: egl: updated SUBDIRS --- src/egl/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/Makefile b/src/egl/Makefile index 931e9d0cb8..082d762183 100644 --- a/src/egl/Makefile +++ b/src/egl/Makefile @@ -2,7 +2,7 @@ TOP = ../.. -SUBDIRS = main drivers/demo drivers/dri +SUBDIRS = main drivers default: subdirs -- cgit v1.2.3 From d607e30232add2358e915027d030ac094394b08e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 5 Jun 2008 14:40:27 -0600 Subject: mesa: always build libglapi.a, updated subdir, gl.pc rule code from master --- src/mesa/Makefile | 22 ++++++++++++++++------ 1 file changed, 16 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index d663cbabfe..e1ff0b2fa2 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -23,11 +23,8 @@ GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ -default: depend subdirs libmesa.a +default: depend subdirs libmesa.a libglapi.a -ifneq ($(DRIVER_DIRS),dri) -default: libglapi.a -endif ###################################################################### @@ -108,8 +105,21 @@ depend: $(ALL_SOURCES) subdirs: - @ (cd x86 ; $(MAKE)) - @ (cd x86-64 ; $(MAKE)) + @ if echo "$(ASM_FLAGS)" | grep -q USE_X86_ASM ; then \ + (cd x86 && $(MAKE)) || exit 1 ; \ + fi + @ if echo "$(ASM_FLAGS)" | grep -q USE_X86_64_ASM ; then \ + (cd x86 && $(MAKE)) || exit 1 ; \ + (cd x86-64 && $(MAKE)) || exit 1 ; \ + fi + +pcedit = sed \ + -e 's,@INSTALL_DIR@,$(INSTALL_DIR),' \ + -e 's,@LIB_DIR@,$(LIB_DIR),' \ + -e 's,@VERSION@,$(MESA_MAJOR).$(MESA_MINOR).$(MESA_TINY),' + +gl.pc: gl.pc.in + $(pcedit) $< > $@ install: default $(INSTALL) -d $(INSTALL_DIR)/include/GL -- cgit v1.2.3 From fe1a2d1fffe69018e30158ee21ed9842384fd233 Mon Sep 17 00:00:00 2001 From: Jonathan White Date: Thu, 5 Jun 2008 15:07:03 -0600 Subject: egl: assorted fixes for Windows Note that int32_t is typedef'd both in p_compiler.h and eglplatform.h --- src/gallium/include/pipe/p_compiler.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index b14260dd90..521ef2d189 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -58,7 +58,9 @@ typedef __int8 int8_t; typedef unsigned __int8 uint8_t; typedef __int16 int16_t; typedef unsigned __int16 uint16_t; +#ifndef __eglplatform_h_ typedef __int32 int32_t; +#endif typedef unsigned __int32 uint32_t; typedef __int64 int64_t; typedef unsigned __int64 uint64_t; -- cgit v1.2.3 From 20e2bb3e8aa8f41246be6319a67f89de42f50344 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 5 Jun 2008 15:38:29 -0600 Subject: insert mesa/ before include files (picked from master) --- src/mesa/drivers/osmesa/osmesa.c | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/osmesa/osmesa.c b/src/mesa/drivers/osmesa/osmesa.c index ad8b20283c..399997bdec 100644 --- a/src/mesa/drivers/osmesa/osmesa.c +++ b/src/mesa/drivers/osmesa/osmesa.c @@ -33,14 +33,14 @@ */ -#include "glheader.h" +#include "main/glheader.h" #include "GL/osmesa.h" -#include "context.h" -#include "extensions.h" -#include "framebuffer.h" -#include "imports.h" -#include "mtypes.h" -#include "renderbuffer.h" +#include "main/context.h" +#include "main/extensions.h" +#include "main/framebuffer.h" +#include "main/imports.h" +#include "main/mtypes.h" +#include "main/renderbuffer.h" #include "swrast/swrast.h" #include "swrast_setup/swrast_setup.h" #include "swrast/s_context.h" -- cgit v1.2.3 From 949600b69356ac470d7df14e96e72c2333568ee1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 5 Jun 2008 15:42:25 -0600 Subject: Move construction of libOSMesa.so into src/mesa/drivers/osmesa/Makefile This removes some cruft from src/mesa/Makefile. Something similar could be done for stand-alone / Xlib-Mesa libGL... --- src/mesa/Makefile | 32 +++-------------- src/mesa/drivers/osmesa/Makefile | 74 ++++++++++++++++++++++++++++++++++++++++ 2 files changed, 78 insertions(+), 28 deletions(-) create mode 100644 src/mesa/drivers/osmesa/Makefile (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index e1ff0b2fa2..d80459b248 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -41,7 +41,7 @@ libglapi.a: $(GLAPI_OBJECTS) ###################################################################### -# Stand-alone Mesa libGL and libOSMesa +# Stand-alone Mesa libGL STAND_ALONE_DRIVER_SOURCES = \ $(COMMON_DRIVER_SOURCES) \ $(X11_DRIVER_SOURCES) @@ -52,17 +52,12 @@ STAND_ALONE_OBJECTS = \ $(CORE_OBJECTS) \ $(STAND_ALONE_DRIVER_OBJECTS) -# For libOSMesa16 or libOSMesa32 we link _all_ the objects into the library, -# not just the osmesa.o object (i.e. we don't have a libGL). -OSMESA16_OBJECTS = \ - $(CORE_OBJECTS) \ - $(COMMON_DRIVER_OBJECTS) \ - $(OSMESA_DRIVER_OBJECTS) +stand-alone: depend subdirs $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) -stand-alone: depend subdirs $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) +osmesa-only: depend subdirs libmesa.a libglapi.a + cd drivers/osmesa && $(MAKE) -osmesa-only: depend subdirs $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) # Make the GL library $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) @@ -73,25 +68,6 @@ $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) -# Make the OSMesa library -$(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME): $(OSMESA_DRIVER_OBJECTS) \ - $(OSMESA16_OBJECTS) $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) - @ if [ "${DRIVER_DIRS}" = "osmesa" ] ; then \ - $(TOP)/bin/mklib -o $(OSMESA_LIB) \ - -linker "$(CC)" \ - -major $(MESA_MAJOR) \ - -minor $(MESA_MINOR) -patch $(MESA_TINY) \ - -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ - $(OSMESA_LIB_DEPS) $(OSMESA16_OBJECTS) ; \ - else \ - $(TOP)/bin/mklib -o $(OSMESA_LIB) \ - -linker "$(CC)" \ - -major $(MESA_MAJOR) \ - -minor $(MESA_MINOR) -patch $(GL_TINY) \ - -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ - $(OSMESA_LIB_DEPS) $(OSMESA_DRIVER_OBJECTS) ; \ - fi - ###################################################################### # Generic stuff diff --git a/src/mesa/drivers/osmesa/Makefile b/src/mesa/drivers/osmesa/Makefile new file mode 100644 index 0000000000..fa8dffcb3e --- /dev/null +++ b/src/mesa/drivers/osmesa/Makefile @@ -0,0 +1,74 @@ +# src/mesa/drivers/osmesa/Makefile for libOSMesa.so + +# Note that we may generate libOSMesa.so or libOSMesa16.so or libOSMesa32.so +# with this Makefile + + +TOP = ../../../.. + +include $(TOP)/configs/current + + + +SOURCES = osmesa.c + +OBJECTS = $(SOURCES:.c=.o) + +INCLUDE_DIRS = \ + -I$(TOP)/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main + +CORE_MESA = $(TOP)/src/mesa/libmesa.a $(TOP)/src/mesa/libglapi.a + + +.PHONY: osmesa8 +.PHONY: osmesa16 + + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + + +default: +# $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) + @ if [ "${DRIVER_DIRS}" = "osmesa" ] ; then \ + $(MAKE) osmesa16 ; \ + else \ + $(MAKE) osmesa8 ; \ + fi + + + + +# The normal libOSMesa is used in conjuction with libGL +osmesa8: $(TOP)/lib/$(OSMESA_LIB_NAME) + +$(TOP)/lib/$(OSMESA_LIB_NAME): $(OBJECTS) + $(TOP)/bin/mklib -o $(OSMESA_LIB) \ + -linker "$(CC)" \ + -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ + -install $(TOP)/$(LIB_DIR) \ + $(MKLIB_OPTIONS) \ + $(OSMESA_LIB_DEPS) $(OBJECTS) + + + + +# The libOSMesa16/libOSMesa32 libraries do not use libGL but rather are built +# with all the other Mesa sources (compiled with -DCHAN_BITS=16/32 +osmesa16: $(OBJECTS) $(CORE_MESA) + $(TOP)/bin/mklib -o $(OSMESA_LIB) \ + -linker "$(CC)" \ + -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ + -install $(TOP)/$(LIB_DIR) \ + $(MKLIB_OPTIONS) \ + $(OSMESA_LIB_DEPS) $(OBJECTS) $(CORE_MESA) + + + +clean: + -rm -f *.o *~ + + +# XXX todo install rule? -- cgit v1.2.3 From eb189c856d5fa4e4b083a2b265e30aa7548d0613 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 5 Jun 2008 16:46:29 -0600 Subject: sync up with Makefile changes on master --- src/mesa/Makefile | 62 ++++++++++++++++++++---------- src/mesa/drivers/fbdev/Makefile | 36 ++++++++++++++++++ src/mesa/drivers/x11/Makefile | 83 ++++++++++++++++++++++++++++++++++++++++- src/mesa/sources | 33 +--------------- 4 files changed, 161 insertions(+), 53 deletions(-) create mode 100644 src/mesa/drivers/fbdev/Makefile (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index d80459b248..42096f8088 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -28,45 +28,67 @@ default: depend subdirs libmesa.a libglapi.a ###################################################################### -# Make archive of core object files +# Helper libraries used by many drivers: + +# Make archive of core mesa object files libmesa.a: $(SOLO_OBJECTS) @ $(TOP)/bin/mklib -o mesa -static $(SOLO_OBJECTS); @if [ "${CONFIG_NAME}" = "beos" ] ; then \ mimeset -f "$@" ; \ fi -# Make archive of gl* API dispatcher functions (when not building DRI/GLX) +# Make archive of gl* API dispatcher functions only libglapi.a: $(GLAPI_OBJECTS) @ $(TOP)/bin/mklib -o glapi -static $(GLAPI_OBJECTS) + +###################################################################### +# BeOS driver target + +beos-driver: depend subdirs libmesa.a + cd drivers/beos && $(MAKE) + + ###################################################################### -# Stand-alone Mesa libGL -STAND_ALONE_DRIVER_SOURCES = \ - $(COMMON_DRIVER_SOURCES) \ - $(X11_DRIVER_SOURCES) +# DRI drivers -STAND_ALONE_DRIVER_OBJECTS = $(STAND_ALONE_DRIVER_SOURCES:.c=.o) +dri-drivers: depend subdirs libmesa.a + cd drivers/dri && $(MAKE) -STAND_ALONE_OBJECTS = \ - $(CORE_OBJECTS) \ - $(STAND_ALONE_DRIVER_OBJECTS) +###################################################################### +# Xlib driver (libGL.so) + +xlib-driver: depend subdirs libmesa.a libglapi.a + cd drivers/x11 && $(MAKE) -stand-alone: depend subdirs $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) -osmesa-only: depend subdirs libmesa.a libglapi.a +###################################################################### +# osmesa driver (libOSMesa.so) + +osmesa-driver: depend subdirs libmesa.a libglapi.a cd drivers/osmesa && $(MAKE) -# Make the GL library -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(STAND_ALONE_OBJECTS) $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) - @ $(TOP)/bin/mklib -o $(GL_LIB) \ - -linker "$(CC)" \ - -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ - -install $(TOP)/$(LIB_DIR) \ - $(MKLIB_OPTIONS) $(STAND_ALONE_OBJECTS) \ - $(PIPE_LIB) $(CELL_LIB) $(CELL_LIB_SPU) $(GL_LIB_DEPS) +##################################################################### +# fbdev driver (libGL.so) + +fbdev-driver: depend subdirs libmesa.a libglapi.a + cd drivers/fbdev && $(MAKE) + + +##################################################################### +# DirectFB driver (libGL.so) + +directfb-libgl: $(CORE_OBJECTS) + @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' -ldflags '$(LDFLAGS)' \ + -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ + -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) $(CORE_OBJECTS) \ + $(GL_LIB_DEPS) + +directfb-driver: depend subdirs directfb-libgl + cd drivers/directfb && $(MAKE) ###################################################################### diff --git a/src/mesa/drivers/fbdev/Makefile b/src/mesa/drivers/fbdev/Makefile new file mode 100644 index 0000000000..ac43dfdb67 --- /dev/null +++ b/src/mesa/drivers/fbdev/Makefile @@ -0,0 +1,36 @@ +# src/mesa/drivers/fbdev/Makefile for libGL.so + +TOP = ../../../.. + +include $(TOP)/configs/current + + +SOURCES = glfbdev.c + +OBJECTS = $(SOURCES:.c=.o) + +INCLUDE_DIRS = \ + -I$(TOP)/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main + +CORE_MESA = $(TOP)/src/mesa/libmesa.a $(TOP)/src/mesa/libglapi.a + + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + + +default: $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) + + +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(CORE_OBJECTS) $(OBJECTS) + @ $(TOP)/bin/mklib -o $(GL_LIB) \ + -linker '$(CC)' -ldflags '$(LDFLAGS)' \ + -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ + -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) \ + $(CORE_MESA) $(OBJECTS) $(GL_LIB_DEPS) + + +clean: + -rm -f $(OBJECTS) diff --git a/src/mesa/drivers/x11/Makefile b/src/mesa/drivers/x11/Makefile index 0ab1dc6e6b..8ee9b61808 100644 --- a/src/mesa/drivers/x11/Makefile +++ b/src/mesa/drivers/x11/Makefile @@ -1,2 +1,81 @@ -default: - cd ../.. ; make \ No newline at end of file +# src/mesa/drivers/x11/Makefile for libGL.so + +# This builds "stand-alone" Mesa, a version of libGL that does not need the +# GLX extension. All rendering is converted to Xlib calls. No hardware +# acceleration. + + +TOP = ../../../.. + +include $(TOP)/configs/current + + +GL_MAJOR = 1 +GL_MINOR = 5 +GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) + + +HEADERS = \ + glxapi.h \ + glxheader.h \ + realglx.h \ + xfonts.h \ + xmesaP.h \ + xm_glide.h \ + xm_image.h + +SOURCES = \ + fakeglx.c \ + glxapi.c \ + realglx.c \ + xfonts.c \ + xm_api.c \ + xm_buffer.c \ + xm_dd.c \ + xm_glide.c \ + xm_image.c \ + xm_line.c \ + xm_span.c \ + xm_tri.c + +OBJECTS = $(SOURCES:.c=.o) + +INCLUDE_DIRS = \ + -I$(TOP)/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main + +CORE_MESA = $(TOP)/src/mesa/libmesa.a $(TOP)/src/mesa/libglapi.a + + + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + + +default: $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) + + +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(OBJECTS) $(CORE_MESA) + @ $(TOP)/bin/mklib -o $(GL_LIB) \ + -linker '$(CC)' -ldflags '$(LDFLAGS)' \ + -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ + -install $(TOP)/$(LIB_DIR) \ + $(MKLIB_OPTIONS) $(GL_LIB_DEPS) $(OBJECTS) + + + +clean: + -rm -f *.o *~ + -rm -f depend depend.bak + + + +depend: $(SOURCES) $(HEADERS) + @ echo "running $(MKDEP)" + @ touch depend + @$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(SOURCES) $(HEADERS) \ + > /dev/null 2>/dev/null + + +-include depend diff --git a/src/mesa/sources b/src/mesa/sources index 5c7ff50cc6..c9ecd7dbf2 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -288,21 +288,6 @@ SPARC_API = \ __COMMON_DRIVER_SOURCES = \ drivers/common/driverfuncs.c -OSMESA_DRIVER_SOURCES = \ - drivers/osmesa/osmesa.c - -GLIDE_DRIVER_SOURCES = \ - drivers/glide/fxapi.c \ - drivers/glide/fxdd.c \ - drivers/glide/fxddspan.c \ - drivers/glide/fxddtex.c \ - drivers/glide/fxsetup.c \ - drivers/glide/fxtexman.c \ - drivers/glide/fxtris.c \ - drivers/glide/fxvb.c \ - drivers/glide/fxglidew.c \ - drivers/glide/fxg.c - SVGA_DRIVER_SOURCES = \ drivers/svga/svgamesa.c \ drivers/svga/svgamesa8.c \ @@ -320,11 +305,9 @@ FBDEV_DRIVER_SOURCES = \ ALL_SOURCES = \ $(GLAPI_SOURCES) \ $(SOLO_SOURCES) \ - $(MESA_ASM_SOURCES) \ + $(MESA_ASM_SOURCES) \ $(COMMON_DRIVER_SOURCES)\ - $(X11_DRIVER_SOURCES) \ - $(FBDEV_DRIVER_SOURCES) \ - $(OSMESA_DRIVER_SOURCES) + $(FBDEV_DRIVER_SOURCES) SOLO_SOURCES = \ $(MAIN_SOURCES) \ @@ -361,8 +344,6 @@ GLAPI_OBJECTS = \ CORE_OBJECTS = $(SOLO_OBJECTS) $(GLAPI_OBJECTS) -OSMESA_DRIVER_OBJECTS = $(OSMESA_DRIVER_SOURCES:.c=.o) - COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o) FBDEV_DRIVER_OBJECTS = $(FBDEV_DRIVER_SOURCES:.c=.o) @@ -378,13 +359,3 @@ INCLUDE_DIRS = \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/gallium/drivers \ -I$(TOP)/src/gallium/auxiliary - -OLD_INCLUDE_DIRS = \ - -I$(TOP)/src/mesa/tnl \ - -I$(TOP)/src/mesa/math \ - -I$(TOP)/src/mesa/swrast \ - -I$(TOP)/src/mesa/swrast_setup \ - -I$(TOP)/src/mesa/shader \ - -I$(TOP)/src/mesa/shader/grammar \ - -I$(TOP)/src/mesa/shader/slang \ - -I$(TOP)/s$(TOP)/src/gallium/auxiliary/tgsi -- cgit v1.2.3 From f27c7729a98937a761eacceabdfd03f9d694d257 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 6 Jun 2008 13:29:59 +0900 Subject: draw: Compile draw_vs_aos only on x86. --- src/gallium/auxiliary/draw/draw_vs_aos.h | 4 ++++ src/gallium/auxiliary/draw/draw_vs_aos_machine.c | 6 ++++++ 2 files changed, 10 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index fb6d43d32e..f6f6af2dd9 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -31,6 +31,9 @@ #ifndef DRAW_VS_AOS_H #define DRAW_VS_AOS_H +#include "pipe/p_config.h" + +#ifdef PIPE_ARCH_X86 struct tgsi_token; struct x86_function; @@ -239,6 +242,7 @@ struct draw_vs_varient_aos_sse { }; +#endif #endif diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c index b7864e9f2f..dcb6c2c8d4 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c @@ -26,6 +26,10 @@ **************************************************************************/ +#include "pipe/p_config.h" + +#ifdef PIPE_ARCH_X86 + #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi/util/tgsi_parse.h" @@ -295,3 +299,5 @@ struct aos_machine *draw_vs_aos_machine( void ) } +#endif + -- cgit v1.2.3 From adbdabb85ae322d5c80cadcee931e36d5c688d98 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 6 Jun 2008 14:49:02 +0200 Subject: i915: Moved pipe_buffer and i915_winsys functions to a common folder --- src/gallium/Makefile | 2 +- src/gallium/winsys/common/Makefile | 20 + src/gallium/winsys/common/Makefile.template | 64 + src/gallium/winsys/common/intel_drm/Makefile | 23 + src/gallium/winsys/common/intel_drm/glthread.h | 359 ++++++ .../winsys/common/intel_drm/intel_be_batchbuffer.c | 429 +++++++ .../winsys/common/intel_drm/intel_be_batchbuffer.h | 69 ++ .../winsys/common/intel_drm/intel_be_context.c | 107 ++ .../winsys/common/intel_drm/intel_be_context.h | 40 + .../winsys/common/intel_drm/intel_be_device.c | 257 ++++ .../winsys/common/intel_drm/intel_be_device.h | 58 + .../winsys/common/intel_drm/ws_dri_bufmgr.c | 949 ++++++++++++++ .../winsys/common/intel_drm/ws_dri_bufmgr.h | 138 +++ .../winsys/common/intel_drm/ws_dri_bufpool.h | 102 ++ .../winsys/common/intel_drm/ws_dri_drmpool.c | 268 ++++ .../winsys/common/intel_drm/ws_dri_fencemgr.c | 377 ++++++ .../winsys/common/intel_drm/ws_dri_fencemgr.h | 115 ++ .../winsys/common/intel_drm/ws_dri_mallocpool.c | 161 +++ .../winsys/common/intel_drm/ws_dri_slabpool.c | 968 +++++++++++++++ src/gallium/winsys/dri/Makefile.template | 1 + src/gallium/winsys/dri/intel/Makefile | 17 +- src/gallium/winsys/dri/intel/intel_batchbuffer.c | 462 ------- src/gallium/winsys/dri/intel/intel_batchbuffer.h | 131 +- src/gallium/winsys/dri/intel/intel_context.c | 71 +- src/gallium/winsys/dri/intel/intel_context.h | 17 +- src/gallium/winsys/dri/intel/intel_lock.c | 14 +- src/gallium/winsys/dri/intel/intel_reg.h | 10 +- src/gallium/winsys/dri/intel/intel_screen.c | 106 +- src/gallium/winsys/dri/intel/intel_screen.h | 25 +- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 33 +- src/gallium/winsys/dri/intel/intel_swapbuffers.h | 10 +- src/gallium/winsys/dri/intel/intel_winsys.h | 73 -- src/gallium/winsys/dri/intel/intel_winsys_i915.c | 156 --- src/gallium/winsys/dri/intel/intel_winsys_pipe.c | 318 ----- .../winsys/dri/intel/intel_winsys_softpipe.c | 20 +- .../winsys/dri/intel/intel_winsys_softpipe.h | 39 + src/gallium/winsys/dri/intel/server/i830_common.h | 8 +- src/gallium/winsys/dri/intel/server/i830_dri.h | 2 +- src/gallium/winsys/dri/intel/server/intel.h | 331 ----- src/gallium/winsys/dri/intel/server/intel_dri.c | 1306 -------------------- src/gallium/winsys/dri/intel/ws_dri_bufmgr.c | 953 -------------- src/gallium/winsys/dri/intel/ws_dri_bufmgr.h | 138 --- src/gallium/winsys/dri/intel/ws_dri_bufpool.h | 102 -- src/gallium/winsys/dri/intel/ws_dri_drmpool.c | 268 ---- src/gallium/winsys/dri/intel/ws_dri_fencemgr.c | 377 ------ src/gallium/winsys/dri/intel/ws_dri_fencemgr.h | 115 -- src/gallium/winsys/dri/intel/ws_dri_mallocpool.c | 162 --- src/gallium/winsys/dri/intel/ws_dri_slabpool.c | 968 --------------- 48 files changed, 4730 insertions(+), 6009 deletions(-) create mode 100644 src/gallium/winsys/common/Makefile create mode 100644 src/gallium/winsys/common/Makefile.template create mode 100644 src/gallium/winsys/common/intel_drm/Makefile create mode 100644 src/gallium/winsys/common/intel_drm/glthread.h create mode 100644 src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.c create mode 100644 src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.h create mode 100644 src/gallium/winsys/common/intel_drm/intel_be_context.c create mode 100644 src/gallium/winsys/common/intel_drm/intel_be_context.h create mode 100644 src/gallium/winsys/common/intel_drm/intel_be_device.c create mode 100644 src/gallium/winsys/common/intel_drm/intel_be_device.h create mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.c create mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.h create mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_bufpool.h create mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_drmpool.c create mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.c create mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.h create mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_mallocpool.c create mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_slabpool.c delete mode 100644 src/gallium/winsys/dri/intel/intel_batchbuffer.c delete mode 100644 src/gallium/winsys/dri/intel/intel_winsys.h delete mode 100644 src/gallium/winsys/dri/intel/intel_winsys_i915.c delete mode 100644 src/gallium/winsys/dri/intel/intel_winsys_pipe.c create mode 100644 src/gallium/winsys/dri/intel/intel_winsys_softpipe.h delete mode 100644 src/gallium/winsys/dri/intel/server/intel.h delete mode 100644 src/gallium/winsys/dri/intel/server/intel_dri.c delete mode 100644 src/gallium/winsys/dri/intel/ws_dri_bufmgr.c delete mode 100644 src/gallium/winsys/dri/intel/ws_dri_bufmgr.h delete mode 100644 src/gallium/winsys/dri/intel/ws_dri_bufpool.h delete mode 100644 src/gallium/winsys/dri/intel/ws_dri_drmpool.c delete mode 100644 src/gallium/winsys/dri/intel/ws_dri_fencemgr.c delete mode 100644 src/gallium/winsys/dri/intel/ws_dri_fencemgr.h delete mode 100644 src/gallium/winsys/dri/intel/ws_dri_mallocpool.c delete mode 100644 src/gallium/winsys/dri/intel/ws_dri_slabpool.c (limited to 'src') diff --git a/src/gallium/Makefile b/src/gallium/Makefile index aa77021daf..291973c904 100644 --- a/src/gallium/Makefile +++ b/src/gallium/Makefile @@ -2,7 +2,7 @@ TOP = ../.. include $(TOP)/configs/current -SUBDIRS = auxiliary drivers +SUBDIRS = auxiliary drivers winsys/common default: subdirs diff --git a/src/gallium/winsys/common/Makefile b/src/gallium/winsys/common/Makefile new file mode 100644 index 0000000000..4c0f3545a5 --- /dev/null +++ b/src/gallium/winsys/common/Makefile @@ -0,0 +1,20 @@ +TOP = ../../../.. +include $(TOP)/configs/current + + +SUBDIRS = $(GALLIUM_WINSYS_COMMON_DIRS) + + +default: subdirs + + +subdirs: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +clean: + rm -f `find . -name \*.[oa]` diff --git a/src/gallium/winsys/common/Makefile.template b/src/gallium/winsys/common/Makefile.template new file mode 100644 index 0000000000..67af778157 --- /dev/null +++ b/src/gallium/winsys/common/Makefile.template @@ -0,0 +1,64 @@ +# -*-makefile-*- + + +# We still have a dependency on the "dri" buffer manager. Most likely +# the interface can be reused in non-dri environments, and also as a +# frontend to simpler memory managers. +# +COMMON_SOURCES = + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(CPP_SOURCES:.cpp=.o) \ + $(ASM_SOURCES:.S=.o) + + +### Include directories +INCLUDES = \ + -I. \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/auxiliary \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/include \ + $(DRIVER_INCLUDES) + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + + +##### TARGETS ##### + +default: depend symlinks $(LIBNAME) + + +$(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/gallium/winsys/common/Makefile.template + $(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS) + + +depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ + $(ASM_SOURCES) 2> /dev/null + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean:: + -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) + -rm -f depend depend.bak + + +include depend diff --git a/src/gallium/winsys/common/intel_drm/Makefile b/src/gallium/winsys/common/intel_drm/Makefile new file mode 100644 index 0000000000..913dbeff20 --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/Makefile @@ -0,0 +1,23 @@ +TOP = ../../../../.. +include $(TOP)/configs/current + +LIBNAME = inteldrm + +C_SOURCES = \ + intel_be_batchbuffer.c \ + intel_be_context.c \ + intel_be_device.c \ + ws_dri_bufmgr.c \ + ws_dri_drmpool.c \ + ws_dri_fencemgr.c \ + ws_dri_mallocpool.c \ + ws_dri_slabpool.c + + +include ../Makefile.template + +DRIVER_DEFINES = $(shell pkg-config libdrm --cflags \ + && pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") +symlinks: + diff --git a/src/gallium/winsys/common/intel_drm/glthread.h b/src/gallium/winsys/common/intel_drm/glthread.h new file mode 100644 index 0000000000..b8e9d5f59b --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/glthread.h @@ -0,0 +1,359 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5.2 + * + * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* + * Thread support for gl dispatch. + * + * Initial version by John Stone (j.stone@acm.org) (johns@cs.umr.edu) + * and Christoph Poliwoda (poliwoda@volumegraphics.com) + * Revised by Keith Whitwell + * Adapted for new gl dispatcher by Brian Paul + * + * + * + * DOCUMENTATION + * + * This thread module exports the following types: + * _glthread_TSD Thread-specific data area + * _glthread_Thread Thread datatype + * _glthread_Mutex Mutual exclusion lock + * + * Macros: + * _glthread_DECLARE_STATIC_MUTEX(name) Declare a non-local mutex + * _glthread_INIT_MUTEX(name) Initialize a mutex + * _glthread_LOCK_MUTEX(name) Lock a mutex + * _glthread_UNLOCK_MUTEX(name) Unlock a mutex + * + * Functions: + * _glthread_GetID(v) Get integer thread ID + * _glthread_InitTSD() Initialize thread-specific data + * _glthread_GetTSD() Get thread-specific data + * _glthread_SetTSD() Set thread-specific data + * + */ + +/* + * If this file is accidentally included by a non-threaded build, + * it should not cause the build to fail, or otherwise cause problems. + * In general, it should only be included when needed however. + */ + +#ifndef GLTHREAD_H +#define GLTHREAD_H + + +#if defined(USE_MGL_NAMESPACE) +#define _glapi_Dispatch _mglapi_Dispatch +#endif + + + +#if (defined(PTHREADS) || defined(SOLARIS_THREADS) ||\ + defined(WIN32_THREADS) || defined(USE_XTHREADS) || defined(BEOS_THREADS)) \ + && !defined(THREADS) +# define THREADS +#endif + +#ifdef VMS +#include +#endif + +/* + * POSIX threads. This should be your choice in the Unix world + * whenever possible. When building with POSIX threads, be sure + * to enable any compiler flags which will cause the MT-safe + * libc (if one exists) to be used when linking, as well as any + * header macros for MT-safe errno, etc. For Solaris, this is the -mt + * compiler flag. On Solaris with gcc, use -D_REENTRANT to enable + * proper compiling for MT-safe libc etc. + */ +#if defined(PTHREADS) +#include /* POSIX threads headers */ + +typedef struct { + pthread_key_t key; + int initMagic; +} _glthread_TSD; + +typedef pthread_t _glthread_Thread; + +typedef pthread_mutex_t _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) \ + static _glthread_Mutex name = PTHREAD_MUTEX_INITIALIZER + +#define _glthread_INIT_MUTEX(name) \ + pthread_mutex_init(&(name), NULL) + +#define _glthread_DESTROY_MUTEX(name) \ + pthread_mutex_destroy(&(name)) + +#define _glthread_LOCK_MUTEX(name) \ + (void) pthread_mutex_lock(&(name)) + +#define _glthread_UNLOCK_MUTEX(name) \ + (void) pthread_mutex_unlock(&(name)) + +typedef pthread_cond_t _glthread_Cond; + +#define _glthread_DECLARE_STATIC_COND(name) \ + static _glthread_Cond name = PTHREAD_COND_INITIALIZER + +#define _glthread_INIT_COND(cond) \ + pthread_cond_init(&(cond), NULL) + +#define _glthread_DESTROY_COND(name) \ + pthread_cond_destroy(&(name)) + +#define _glthread_COND_WAIT(cond, mutex) \ + pthread_cond_wait(&(cond), &(mutex)) + +#define _glthread_COND_SIGNAL(cond) \ + pthread_cond_signal(&(cond)) + +#define _glthread_COND_BROADCAST(cond) \ + pthread_cond_broadcast(&(cond)) + + +#else /* PTHREADS */ + +typedef unsigned int _glthread_Cond; +#define _glthread_DECLARE_STATIC_COND(name) \ +// #warning Condition variables not implemented. + +#define _glthread_INIT_COND(cond) \ + abort(); + +#define _glthread_DESTROY_COND(name) \ + abort(); + +#define _glthread_COND_WAIT(cond, mutex) \ + abort(); + +#define _glthread_COND_SIGNAL(cond) \ + abort(); + +#define _glthread_COND_BROADCAST(cond) \ + abort(); + +#endif + + +/* + * Solaris threads. Use only up to Solaris 2.4. + * Solaris 2.5 and higher provide POSIX threads. + * Be sure to compile with -mt on the Solaris compilers, or + * use -D_REENTRANT if using gcc. + */ +#ifdef SOLARIS_THREADS +#include + +typedef struct { + thread_key_t key; + mutex_t keylock; + int initMagic; +} _glthread_TSD; + +typedef thread_t _glthread_Thread; + +typedef mutex_t _glthread_Mutex; + +/* XXX need to really implement mutex-related macros */ +#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 +#define _glthread_INIT_MUTEX(name) (void) name +#define _glthread_DESTROY_MUTEX(name) (void) name +#define _glthread_LOCK_MUTEX(name) (void) name +#define _glthread_UNLOCK_MUTEX(name) (void) name + +#endif /* SOLARIS_THREADS */ + + + + +/* + * Windows threads. Should work with Windows NT and 95. + * IMPORTANT: Link with multithreaded runtime library when THREADS are + * used! + */ +#ifdef WIN32_THREADS +#include + +typedef struct { + DWORD key; + int initMagic; +} _glthread_TSD; + +typedef HANDLE _glthread_Thread; + +typedef CRITICAL_SECTION _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) /*static*/ _glthread_Mutex name = {0,0,0,0,0,0} +#define _glthread_INIT_MUTEX(name) InitializeCriticalSection(&name) +#define _glthread_DESTROY_MUTEX(name) DeleteCriticalSection(&name) +#define _glthread_LOCK_MUTEX(name) EnterCriticalSection(&name) +#define _glthread_UNLOCK_MUTEX(name) LeaveCriticalSection(&name) + +#endif /* WIN32_THREADS */ + + + + +/* + * XFree86 has its own thread wrapper, Xthreads.h + * We wrap it again for GL. + */ +#ifdef USE_XTHREADS +#include + +typedef struct { + xthread_key_t key; + int initMagic; +} _glthread_TSD; + +typedef xthread_t _glthread_Thread; + +typedef xmutex_rec _glthread_Mutex; + +#ifdef XMUTEX_INITIALIZER +#define _glthread_DECLARE_STATIC_MUTEX(name) \ + static _glthread_Mutex name = XMUTEX_INITIALIZER +#else +#define _glthread_DECLARE_STATIC_MUTEX(name) \ + static _glthread_Mutex name +#endif + +#define _glthread_INIT_MUTEX(name) \ + xmutex_init(&(name)) + +#define _glthread_DESTROY_MUTEX(name) \ + xmutex_clear(&(name)) + +#define _glthread_LOCK_MUTEX(name) \ + (void) xmutex_lock(&(name)) + +#define _glthread_UNLOCK_MUTEX(name) \ + (void) xmutex_unlock(&(name)) + +#endif /* USE_XTHREADS */ + + + +/* + * BeOS threads. R5.x required. + */ +#ifdef BEOS_THREADS + +#include +#include + +typedef struct { + int32 key; + int initMagic; +} _glthread_TSD; + +typedef thread_id _glthread_Thread; + +/* Use Benaphore, aka speeder semaphore */ +typedef struct { + int32 lock; + sem_id sem; +} benaphore; +typedef benaphore _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = { 0, 0 } +#define _glthread_INIT_MUTEX(name) name.sem = create_sem(0, #name"_benaphore"), name.lock = 0 +#define _glthread_DESTROY_MUTEX(name) delete_sem(name.sem), name.lock = 0 +#define _glthread_LOCK_MUTEX(name) if (name.sem == 0) _glthread_INIT_MUTEX(name); \ + if (atomic_add(&(name.lock), 1) >= 1) acquire_sem(name.sem) +#define _glthread_UNLOCK_MUTEX(name) if (atomic_add(&(name.lock), -1) > 1) release_sem(name.sem) + +#endif /* BEOS_THREADS */ + + + +#ifndef THREADS + +/* + * THREADS not defined + */ + +typedef GLuint _glthread_TSD; + +typedef GLuint _glthread_Thread; + +typedef GLuint _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 + +#define _glthread_INIT_MUTEX(name) (void) name + +#define _glthread_DESTROY_MUTEX(name) (void) name + +#define _glthread_LOCK_MUTEX(name) (void) name + +#define _glthread_UNLOCK_MUTEX(name) (void) name + +#endif /* THREADS */ + + + +/* + * Platform independent thread specific data API. + */ + +extern unsigned long +_glthread_GetID(void); + + +extern void +_glthread_InitTSD(_glthread_TSD *); + + +extern void * +_glthread_GetTSD(_glthread_TSD *); + + +extern void +_glthread_SetTSD(_glthread_TSD *, void *); + +#if defined(GLX_USE_TLS) + +extern __thread struct _glapi_table * _glapi_tls_Dispatch + __attribute__((tls_model("initial-exec"))); + +#define GET_DISPATCH() _glapi_tls_Dispatch + +#elif !defined(GL_CALL) +# if defined(THREADS) +# define GET_DISPATCH() \ + ((__builtin_expect( _glapi_Dispatch != NULL, 1 )) \ + ? _glapi_Dispatch : _glapi_get_dispatch()) +# else +# define GET_DISPATCH() _glapi_Dispatch +# endif /* defined(THREADS) */ +#endif /* ndef GL_CALL */ + + +#endif /* THREADS_H */ diff --git a/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.c b/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.c new file mode 100644 index 0000000000..bc13a5761e --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.c @@ -0,0 +1,429 @@ + +#include "intel_be_batchbuffer.h" +#include "intel_be_context.h" +#include "intel_be_device.h" +#include + +#include "xf86drm.h" + +static void +intel_realloc_relocs(struct intel_be_batchbuffer *batch, int num_relocs) +{ + unsigned long size = num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER; + + size *= sizeof(uint32_t); + batch->reloc = realloc(batch->reloc, size); + batch->reloc_size = num_relocs; +} + + +void +intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch) +{ + /* + * Get a new, free batchbuffer. + */ + drmBO *bo; + struct drm_bo_info_req *req; + + driBOUnrefUserList(batch->list); + driBOResetList(batch->list); + + /* base.size is the size available to the i915simple driver */ + batch->base.size = batch->device->max_batch_size - BATCH_RESERVED; + batch->base.actual_size = batch->device->max_batch_size; + driBOData(batch->buffer, batch->base.actual_size, NULL, NULL, 0); + + /* + * Add the batchbuffer to the validate list. + */ + + driBOAddListItem(batch->list, batch->buffer, + DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_EXE | DRM_BO_MASK_MEM, + &batch->dest_location, &batch->node); + + req = &batch->node->bo_arg.d.req.bo_req; + + /* + * Set up information needed for us to make relocations + * relative to the underlying drm buffer objects. + */ + + driReadLockKernelBO(); + bo = driBOKernel(batch->buffer); + req->presumed_offset = (uint64_t) bo->offset; + req->hint = DRM_BO_HINT_PRESUMED_OFFSET; + batch->drmBOVirtual = (uint8_t *) bo->virtual; + driReadUnlockKernelBO(); + + /* + * Adjust the relocation buffer size. + */ + + if (batch->reloc_size > INTEL_MAX_RELOCS || + batch->reloc == NULL) + intel_realloc_relocs(batch, INTEL_DEFAULT_RELOCS); + + assert(batch->reloc != NULL); + batch->reloc[0] = 0; /* No relocs yet. */ + batch->reloc[1] = 1; /* Reloc type 1 */ + batch->reloc[2] = 0; /* Only a single relocation list. */ + batch->reloc[3] = 0; /* Only a single relocation list. */ + + batch->base.map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); + batch->poolOffset = driBOPoolOffset(batch->buffer); + batch->base.ptr = batch->base.map; + batch->dirty_state = ~0; + batch->nr_relocs = 0; + batch->flags = 0; + batch->id = 0;//batch->intel->intelScreen->batch_id++; +} + +/*====================================================================== + * Public functions + */ +struct intel_be_batchbuffer * +intel_be_batchbuffer_alloc(struct intel_be_context *intel) +{ + struct intel_be_batchbuffer *batch = calloc(sizeof(*batch), 1); + + batch->intel = intel; + batch->device = intel->device; + + driGenBuffers(intel->device->batchPool, "batchbuffer", 1, + &batch->buffer, 4096, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); + batch->last_fence = NULL; + batch->list = driBOCreateList(20); + batch->reloc = NULL; + intel_be_batchbuffer_reset(batch); + return batch; +} + +void +intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch) +{ + if (batch->last_fence) { + driFenceFinish(batch->last_fence, + DRM_FENCE_TYPE_EXE, FALSE); + driFenceUnReference(&batch->last_fence); + } + if (batch->base.map) { + driBOUnmap(batch->buffer); + batch->base.map = NULL; + } + driBOUnReference(batch->buffer); + driBOFreeList(batch->list); + if (batch->reloc) + free(batch->reloc); + batch->buffer = NULL; + free(batch); +} + +void +intel_be_offset_relocation(struct intel_be_batchbuffer *batch, + unsigned pre_add, + struct _DriBufferObject *driBO, + uint64_t val_flags, + uint64_t val_mask) +{ + int itemLoc; + struct _drmBONode *node; + uint32_t *reloc; + struct drm_bo_info_req *req; + + driBOAddListItem(batch->list, driBO, val_flags, val_mask, + &itemLoc, &node); + req = &node->bo_arg.d.req.bo_req; + + if (!(req->hint & DRM_BO_HINT_PRESUMED_OFFSET)) { + + /* + * Stop other threads from tampering with the underlying + * drmBO while we're reading its offset. + */ + + driReadLockKernelBO(); + req->presumed_offset = (uint64_t) driBOKernel(driBO)->offset; + driReadUnlockKernelBO(); + req->hint = DRM_BO_HINT_PRESUMED_OFFSET; + } + + pre_add += driBOPoolOffset(driBO); + + if (batch->nr_relocs == batch->reloc_size) + intel_realloc_relocs(batch, batch->reloc_size * 2); + + reloc = batch->reloc + + (I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE); + + reloc[0] = ((uint8_t *)batch->base.ptr - batch->drmBOVirtual); + i915_batchbuffer_dword(&batch->base, req->presumed_offset + pre_add); + reloc[1] = pre_add; + reloc[2] = itemLoc; + reloc[3] = batch->dest_location; + batch->nr_relocs++; +} + +static void +i915_drm_copy_reply(const struct drm_bo_info_rep * rep, drmBO * buf) +{ + buf->handle = rep->handle; + buf->flags = rep->flags; + buf->size = rep->size; + buf->offset = rep->offset; + buf->mapHandle = rep->arg_handle; + buf->proposedFlags = rep->proposed_flags; + buf->start = rep->buffer_start; + buf->fenceFlags = rep->fence_flags; + buf->replyFlags = rep->rep_flags; + buf->pageAlignment = rep->page_alignment; +} + +static int +i915_execbuf(struct intel_be_batchbuffer *batch, + unsigned int used, + boolean ignore_cliprects, + drmBOList *list, + struct drm_i915_execbuffer *ea) +{ +// struct intel_be_context *intel = batch->intel; + drmBONode *node; + drmMMListHead *l; + struct drm_i915_op_arg *arg, *first; + struct drm_bo_op_req *req; + struct drm_bo_info_rep *rep; + uint64_t *prevNext = NULL; + drmBO *buf; + int ret = 0; + uint32_t count = 0; + + first = NULL; + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + + arg = &node->bo_arg; + req = &arg->d.req; + + if (!first) + first = arg; + + if (prevNext) + *prevNext = (unsigned long)arg; + + prevNext = &arg->next; + req->bo_req.handle = node->buf->handle; + req->op = drm_bo_validate; + req->bo_req.flags = node->arg0; + req->bo_req.mask = node->arg1; + req->bo_req.hint |= 0; + count++; + } + + memset(ea, 0, sizeof(*ea)); + ea->num_buffers = count; + ea->batch.start = batch->poolOffset; + ea->batch.used = used; +#if 0 /* ZZZ JB: no cliprects used */ + ea->batch.cliprects = intel->pClipRects; + ea->batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + ea->batch.DR1 = 0; + ea->batch.DR4 = 0;((((GLuint) intel->drawX) & 0xffff) | + (((GLuint) intel->drawY) << 16)); +#else + ea->batch.cliprects = NULL; + ea->batch.num_cliprects = 0; + ea->batch.DR1 = 0; + ea->batch.DR4 = 0; +#endif + ea->fence_arg.flags = DRM_I915_FENCE_FLAG_FLUSHED; + ea->ops_list = (unsigned long) first; + first->reloc_ptr = (unsigned long) batch->reloc; + batch->reloc[0] = batch->nr_relocs; + + //return -EFAULT; + do { + ret = drmCommandWriteRead(batch->device->fd, DRM_I915_EXECBUFFER, ea, + sizeof(*ea)); + } while (ret == -EAGAIN); + + if (ret != 0) + return ret; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + arg = &node->bo_arg; + rep = &arg->d.rep.bo_info; + + if (!arg->handled) { + return -EFAULT; + } + if (arg->d.rep.ret) + return arg->d.rep.ret; + + buf = node->buf; + i915_drm_copy_reply(rep, buf); + } + return 0; +} + +/* TODO: Push this whole function into bufmgr. + */ +static struct _DriFenceObject * +do_flush_locked(struct intel_be_batchbuffer *batch, + unsigned int used, + boolean ignore_cliprects, boolean allow_unlock) +{ + struct intel_be_context *intel = batch->intel; + struct _DriFenceObject *fo; + drmFence fence; + drmBOList *boList; + struct drm_i915_execbuffer ea; + int ret = 0; + + driBOValidateUserList(batch->list); + boList = driGetdrmBOList(batch->list); + +#if 0 /* ZZZ JB Allways run */ + if (!(intel->numClipRects == 0 && !ignore_cliprects)) { +#else + if (1) { +#endif + ret = i915_execbuf(batch, used, ignore_cliprects, boList, &ea); + } else { + driPutdrmBOList(batch->list); + fo = NULL; + goto out; + } + driPutdrmBOList(batch->list); + if (ret) + abort(); + + if (ea.fence_arg.error != 0) { + + /* + * The hardware has been idled by the kernel. + * Don't fence the driBOs. + */ + + if (batch->last_fence) + driFenceUnReference(&batch->last_fence); +#if 0 /* ZZZ JB: no _mesa_* funcs in gallium */ + _mesa_printf("fence error\n"); +#endif + batch->last_fence = NULL; + fo = NULL; + goto out; + } + + fence.handle = ea.fence_arg.handle; + fence.fence_class = ea.fence_arg.fence_class; + fence.type = ea.fence_arg.type; + fence.flags = ea.fence_arg.flags; + fence.signaled = ea.fence_arg.signaled; + + fo = driBOFenceUserList(batch->device->fenceMgr, batch->list, + "SuperFence", &fence); + + if (driFenceType(fo) & DRM_I915_FENCE_TYPE_RW) { + if (batch->last_fence) + driFenceUnReference(&batch->last_fence); + /* + * FIXME: Context last fence?? + */ + batch->last_fence = fo; + driFenceReference(fo); + } + out: +#if 0 /* ZZZ JB: fix this */ + intel->vtbl.lost_hardware(intel); +#else + (void)intel; +#endif + return fo; +} + + +struct _DriFenceObject * +intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch) +{ + struct intel_be_context *intel = batch->intel; + unsigned int used = batch->base.ptr - batch->base.map; + boolean was_locked = batch->intel->hardware_locked(intel); + struct _DriFenceObject *fence; + + if (used == 0) { + driFenceReference(batch->last_fence); + return batch->last_fence; + } + + /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a + * performance drain that we would like to avoid. + */ +#if 0 /* ZZZ JB: what should we do here? */ + if (used & 4) { + ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->base.ptr)[1] = 0; + ((int *) batch->base.ptr)[2] = MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->base.ptr)[1] = MI_BATCH_BUFFER_END; + used += 8; + } +#else + if (used & 4) { + ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->base.ptr)[1] = 0; + ((int *) batch->base.ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->base.ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; + used += 8; + } +#endif + driBOUnmap(batch->buffer); + batch->base.ptr = NULL; + batch->base.map = NULL; + + /* TODO: Just pass the relocation list and dma buffer up to the + * kernel. + */ + if (!was_locked) + intel->hardware_lock(intel); + + fence = do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), + FALSE); + + if (!was_locked) + intel->hardware_unlock(intel); + + /* Reset the buffer: + */ + intel_be_batchbuffer_reset(batch); + return fence; +} + +void +intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch) +{ + struct _DriFenceObject *fence = intel_be_batchbuffer_flush(batch); + driFenceFinish(fence, driFenceType(fence), FALSE); + driFenceUnReference(&fence); +} + +#if 0 +void +intel_be_batchbuffer_data(struct intel_be_batchbuffer *batch, + const void *data, unsigned int bytes, unsigned int flags) +{ + assert((bytes & 3) == 0); + intel_batchbuffer_require_space(batch, bytes, flags); + memcpy(batch->base.ptr, data, bytes); + batch->base.ptr += bytes; +} +#endif diff --git a/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.h b/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.h new file mode 100644 index 0000000000..f150e3a674 --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.h @@ -0,0 +1,69 @@ + +#ifndef INTEL_BE_BATCHBUFFER_H +#define INTEL_BE_BATCHBUFFER_H + +#include "i915simple/i915_batch.h" + +#include "ws_dri_bufmgr.h" + +#define BATCH_RESERVED 16 + +#define INTEL_DEFAULT_RELOCS 100 +#define INTEL_MAX_RELOCS 400 + +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 + +struct intel_be_context; +struct intel_be_device; + +struct intel_be_batchbuffer +{ + struct i915_batchbuffer base; + + struct intel_be_context *intel; + struct intel_be_device *device; + + struct _DriBufferObject *buffer; + struct _DriFenceObject *last_fence; + uint32_t flags; + + struct _DriBufferList *list; + size_t list_count; + + uint32_t *reloc; + size_t reloc_size; + size_t nr_relocs; + + uint32_t dirty_state; + uint32_t id; + + uint32_t poolOffset; + uint8_t *drmBOVirtual; + struct _drmBONode *node; /* Validation list node for this buffer */ + int dest_location; /* Validation list sequence for this buffer */ +}; + +struct intel_be_batchbuffer * +intel_be_batchbuffer_alloc(struct intel_be_context *intel); + +void +intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch); + +void +intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch); + +struct _DriFenceObject * +intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch); + +void +intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch); + +void +intel_be_offset_relocation(struct intel_be_batchbuffer *batch, + unsigned pre_add, + struct _DriBufferObject *driBO, + uint64_t val_flags, + uint64_t val_mask); + +#endif diff --git a/src/gallium/winsys/common/intel_drm/intel_be_context.c b/src/gallium/winsys/common/intel_drm/intel_be_context.c new file mode 100644 index 0000000000..1af39674f4 --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/intel_be_context.c @@ -0,0 +1,107 @@ + +/* + * Authors: Jakob Bornecrantz + */ + +#include "ws_dri_fencemgr.h" +#include "intel_be_device.h" +#include "intel_be_context.h" +#include "intel_be_batchbuffer.h" + +static INLINE struct intel_be_context * +intel_be_context(struct i915_winsys *sws) +{ + return (struct intel_be_context *)sws; +} + +/* Simple batchbuffer interface: + */ + +static struct i915_batchbuffer* +intel_i915_batch_get(struct i915_winsys *sws) +{ + struct intel_be_context *intel = intel_be_context(sws); + return &intel->batch->base; +} + +static void intel_i915_batch_reloc(struct i915_winsys *sws, + struct pipe_buffer *buf, + unsigned access_flags, + unsigned delta) +{ + struct intel_be_context *intel = intel_be_context(sws); + + unsigned flags = DRM_BO_FLAG_MEM_TT; + unsigned mask = DRM_BO_MASK_MEM; + + if (access_flags & I915_BUFFER_ACCESS_WRITE) { + flags |= DRM_BO_FLAG_WRITE; + mask |= DRM_BO_FLAG_WRITE; + } + + if (access_flags & I915_BUFFER_ACCESS_READ) { + flags |= DRM_BO_FLAG_READ; + mask |= DRM_BO_FLAG_READ; + } + + intel_be_offset_relocation(intel->batch, + delta, + dri_bo(buf), + flags, + mask); +} + +static void intel_i915_batch_flush(struct i915_winsys *sws, + struct pipe_fence_handle **fence) +{ + struct intel_be_context *intel = intel_be_context(sws); + + union { + struct _DriFenceObject *dri; + struct pipe_fence_handle *pipe; + } fu; + + if (fence) + assert(!*fence); + + fu.dri = intel_be_batchbuffer_flush(intel->batch); + + if (!fu.dri) { + assert(0); + *fence = NULL; + return; + } + + if (fu.dri) { + if (fence) + *fence = fu.pipe; + else + driFenceUnReference(&fu.dri); + } + +} + +boolean +intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device) +{ + assert(intel); + assert(device); + + intel->device = device; + + /* TODO move framebuffer createion to the driver */ + + intel->base.batch_get = intel_i915_batch_get; + intel->base.batch_reloc = intel_i915_batch_reloc; + intel->base.batch_flush = intel_i915_batch_flush; + + intel->batch = intel_be_batchbuffer_alloc(intel); + + return true; +} + +void +intel_be_destroy_context(struct intel_be_context *intel) +{ + intel_be_batchbuffer_free(intel->batch); +} diff --git a/src/gallium/winsys/common/intel_drm/intel_be_context.h b/src/gallium/winsys/common/intel_drm/intel_be_context.h new file mode 100644 index 0000000000..d5cbc93594 --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/intel_be_context.h @@ -0,0 +1,40 @@ +/* These need to be diffrent from the intel winsys */ +#ifndef INTEL_BE_CONTEXT_H +#define INTEL_BE_CONTEXT_H + +#include "i915simple/i915_winsys.h" + +struct intel_be_context +{ + /** Interface to i915simple driver */ + struct i915_winsys base; + + struct intel_be_device *device; + struct intel_be_batchbuffer *batch; + + /* + * Hardware lock functions. + * + * Needs to be filled in by the winsys. + */ + void (*hardware_lock)(struct intel_be_context *context); + void (*hardware_unlock)(struct intel_be_context *context); + boolean (*hardware_locked)(struct intel_be_context *context); +}; + +/** + * Intialize a allocated intel_be_context struct. + * + * Remember to set the hardware_* functions. + */ +boolean +intel_be_init_context(struct intel_be_context *intel, + struct intel_be_device *device); + +/** + * Destroy a intel_be_context. + * Does not free the struct that is up to the winsys. + */ +void +intel_be_destroy_context(struct intel_be_context *intel); +#endif diff --git a/src/gallium/winsys/common/intel_drm/intel_be_device.c b/src/gallium/winsys/common/intel_drm/intel_be_device.c new file mode 100644 index 0000000000..0fc1894eaa --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/intel_be_device.c @@ -0,0 +1,257 @@ + + +/* + * Authors: Keith Whitwell + * Jakob Bornecrantz + */ + +#include "intel_be_device.h" +#include "ws_dri_bufmgr.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" + +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" + +/* Turn a pipe winsys into an intel/pipe winsys: + */ +static INLINE struct intel_be_device * +intel_be_device( struct pipe_winsys *winsys ) +{ + return (struct intel_be_device *)winsys; +} + + +/* + * Buffer functions. + * + * Most callbacks map direcly onto dri_bufmgr operations: + */ + +static void *intel_be_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags ) +{ + unsigned drm_flags = 0; + + if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) + drm_flags |= DRM_BO_FLAG_WRITE; + + if (flags & PIPE_BUFFER_USAGE_CPU_READ) + drm_flags |= DRM_BO_FLAG_READ; + + return driBOMap( dri_bo(buf), drm_flags, 0 ); +} + +static void intel_be_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + driBOUnmap( dri_bo(buf) ); +} + +static void +intel_be_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + driBOUnReference( dri_bo(buf) ); + FREE(buf); +} + +static struct pipe_buffer * +intel_be_buffer_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned usage, + unsigned size ) +{ + struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer ); + struct intel_be_device *iws = intel_be_device(winsys); + unsigned flags = 0; + struct _DriBufferPool *pool; + + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + + if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { + flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; + pool = iws->mallocPool; + } else if (usage & PIPE_BUFFER_USAGE_CUSTOM) { + /* For vertex buffers */ + flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; + pool = iws->vertexPool; + } else { + flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; + pool = iws->regionPool; + } + + if (usage & PIPE_BUFFER_USAGE_GPU_READ) + flags |= DRM_BO_FLAG_READ; + + if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) + flags |= DRM_BO_FLAG_WRITE; + + /* drm complains if we don't set any read/write flags. + */ + if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) + flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + + buffer->pool = pool; + driGenBuffers( buffer->pool, + "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); + + driBOData( buffer->driBO, size, NULL, buffer->pool, 0 ); + + return &buffer->base; +} + + +static struct pipe_buffer * +intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) +{ + struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer ); + struct intel_be_device *iws = intel_be_device(winsys); + + driGenUserBuffer( iws->regionPool, + "pipe user buffer", &buffer->driBO, ptr, bytes ); + + buffer->base.refcount = 1; + + return &buffer->base; +} + + +/* + * Surface functions. + * + * Deprecated! + */ + +static struct pipe_surface * +intel_i915_surface_alloc(struct pipe_winsys *winsys) +{ + assert((size_t)"intel_i915_surface_alloc is deprecated" & 0); + return NULL; +} + +static int +intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags, + unsigned tex_usage) +{ + assert((size_t)"intel_i915_surface_alloc_storage is deprecated" & 0); + return -1; +} + +static void +intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + assert((size_t)"intel_i915_surface_release is deprecated" & 0); +} + +/* + * Fence functions + */ + +static void +intel_be_fence_reference( struct pipe_winsys *sws, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence ) +{ + if (*ptr) + driFenceUnReference((struct _DriFenceObject **)ptr); + + if (fence) + *ptr = (struct pipe_fence_handle *)driFenceReference((struct _DriFenceObject *)fence); +} + +static int +intel_be_fence_signalled( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ) +{ + return driFenceSignaled((struct _DriFenceObject *)fence, flag); +} + +static int +intel_be_fence_finish( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ) +{ + return driFenceFinish((struct _DriFenceObject *)fence, flag, 0); +} + +/* + * Misc functions + */ + +boolean +intel_be_init_device(struct intel_be_device *dev, int fd) +{ + dev->fd = fd; + dev->max_batch_size = 16 * 4096; + dev->max_vertex_size = 128 * 4096; + + dev->base.buffer_create = intel_be_buffer_create; + dev->base.user_buffer_create = intel_be_user_buffer_create; + dev->base.buffer_map = intel_be_buffer_map; + dev->base.buffer_unmap = intel_be_buffer_unmap; + dev->base.buffer_destroy = intel_be_buffer_destroy; + dev->base.surface_alloc = intel_i915_surface_alloc; + dev->base.surface_alloc_storage = intel_i915_surface_alloc_storage; + dev->base.surface_release = intel_i915_surface_release; + dev->base.fence_reference = intel_be_fence_reference; + dev->base.fence_signalled = intel_be_fence_signalled; + dev->base.fence_finish = intel_be_fence_finish; + +#if 0 /* Set by the winsys */ + dev->base.flush_frontbuffer = intel_flush_frontbuffer; + dev->base.get_name = intel_get_name; +#endif + + dev->fMan = driInitFreeSlabManager(10, 10); + dev->fenceMgr = driFenceMgrTTMInit(dev->fd); + + dev->mallocPool = driMallocPoolInit(); + dev->staticPool = driDRMPoolInit(dev->fd); + dev->regionPool = driDRMPoolInit(dev->fd); + dev->vertexPool = driSlabPoolInit(dev->fd, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + dev->max_vertex_size, + 1, 120, dev->max_vertex_size * 4, 0, + dev->fMan); + + dev->batchPool = driSlabPoolInit(dev->fd, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + dev->max_batch_size, + 1, 40, dev->max_batch_size * 16, 0, + dev->fMan); + + return true; +} + +void +intel_be_destroy_device(struct intel_be_device *dev) +{ + driPoolTakeDown(dev->mallocPool); + driPoolTakeDown(dev->staticPool); + driPoolTakeDown(dev->regionPool); + driPoolTakeDown(dev->vertexPool); + driPoolTakeDown(dev->batchPool); + + /** TODO takedown fenceMgr and fMan */ +} diff --git a/src/gallium/winsys/common/intel_drm/intel_be_device.h b/src/gallium/winsys/common/intel_drm/intel_be_device.h new file mode 100644 index 0000000000..ec5cace71c --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/intel_be_device.h @@ -0,0 +1,58 @@ +#ifndef INTEL_DRM_DEVICE_H +#define INTEL_DRM_DEVICE_H + +#include "pipe/p_winsys.h" +#include "pipe/p_context.h" + +/* + * Device + */ + +struct intel_be_device +{ + struct pipe_winsys base; + + int fd; /**< Drm file discriptor */ + + size_t max_batch_size; + size_t max_vertex_size; + + struct _DriFenceMgr *fenceMgr; + + struct _DriBufferPool *batchPool; + struct _DriBufferPool *regionPool; + struct _DriBufferPool *mallocPool; + struct _DriBufferPool *vertexPool; + struct _DriBufferPool *staticPool; + struct _DriFreeSlabManager *fMan; +}; + +boolean +intel_be_init_device(struct intel_be_device *device, int fd); + +void +intel_be_destroy_device(struct intel_be_device *dev); + +/* + * Buffer + */ + +struct intel_be_buffer { + struct pipe_buffer base; + struct _DriBufferPool *pool; + struct _DriBufferObject *driBO; +}; + +static INLINE struct intel_be_buffer * +intel_be_buffer( struct pipe_buffer *buf ) +{ + return (struct intel_be_buffer *)buf; +} + +static INLINE struct _DriBufferObject * +dri_bo( struct pipe_buffer *buf ) +{ + return intel_be_buffer(buf)->driBO; +} + +#endif diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.c b/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.c new file mode 100644 index 0000000000..b6d901f85e --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.c @@ -0,0 +1,949 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + * Keith Whitwell + */ + +#include +#include +#include +#include "glthread.h" +#include "errno.h" +#include "ws_dri_bufmgr.h" +#include "string.h" +#include "pipe/p_debug.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" + + +/* + * This lock is here to protect drmBO structs changing underneath us during a + * validate list call, since validatelist cannot take individiual locks for + * each drmBO. Validatelist takes this lock in write mode. Any access to an + * individual drmBO should take this lock in read mode, since in that case, the + * driBufferObject mutex will protect the access. Locking order is + * driBufferObject mutex - > this rw lock. + */ + +_glthread_DECLARE_STATIC_MUTEX(bmMutex); +_glthread_DECLARE_STATIC_COND(bmCond); + +static int kernelReaders = 0; +static int num_buffers = 0; +static int num_user_buffers = 0; + +static drmBO *drmBOListBuf(void *iterator) +{ + drmBONode *node; + drmMMListHead *l = (drmMMListHead *) iterator; + node = DRMLISTENTRY(drmBONode, l, head); + return node->buf; +} + +static void *drmBOListIterator(drmBOList *list) +{ + void *ret = list->list.next; + + if (ret == &list->list) + return NULL; + return ret; +} + +static void *drmBOListNext(drmBOList *list, void *iterator) +{ + void *ret; + + drmMMListHead *l = (drmMMListHead *) iterator; + ret = l->next; + if (ret == &list->list) + return NULL; + return ret; +} + +static drmBONode *drmAddListItem(drmBOList *list, drmBO *item, + uint64_t arg0, + uint64_t arg1) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->free.next; + if (l == &list->free) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + return NULL; + } + list->numCurrent++; + } + else { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + } + node->buf = item; + node->arg0 = arg0; + node->arg1 = arg1; + DRMLISTADD(&node->head, &list->list); + list->numOnList++; + return node; +} + +static int drmAddValidateItem(drmBOList *list, drmBO *buf, uint64_t flags, + uint64_t mask, int *newItem) +{ + drmBONode *node, *cur; + drmMMListHead *l; + + *newItem = 0; + cur = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + if (node->buf == buf) { + cur = node; + break; + } + } + if (!cur) { + cur = drmAddListItem(list, buf, flags, mask); + if (!cur) { + return -ENOMEM; + } + *newItem = 1; + cur->arg0 = flags; + cur->arg1 = mask; + } + else { + uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; + uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; + + if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { + return -EINVAL; + } + + cur->arg1 |= mask; + cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); + + if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && + (cur->arg0 & DRM_BO_MASK_MEM) == 0) { + return -EINVAL; + } + } + return 0; +} + +static void drmBOFreeList(drmBOList *list) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->list.next; + while(l != &list->list) { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + l = list->list.next; + list->numCurrent--; + list->numOnList--; + } + + l = list->free.next; + while(l != &list->free) { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + l = list->free.next; + list->numCurrent--; + } +} + +static int drmAdjustListNodes(drmBOList *list) +{ + drmBONode *node; + drmMMListHead *l; + int ret = 0; + + while(list->numCurrent < list->numTarget) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + ret = -ENOMEM; + break; + } + list->numCurrent++; + DRMLISTADD(&node->head, &list->free); + } + + while(list->numCurrent > list->numTarget) { + l = list->free.next; + if (l == &list->free) + break; + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + list->numCurrent--; + } + return ret; +} + +static int drmBOCreateList(int numTarget, drmBOList *list) +{ + DRMINITLISTHEAD(&list->list); + DRMINITLISTHEAD(&list->free); + list->numTarget = numTarget; + list->numCurrent = 0; + list->numOnList = 0; + return drmAdjustListNodes(list); +} + +static int drmBOResetList(drmBOList *list) +{ + drmMMListHead *l; + int ret; + + ret = drmAdjustListNodes(list); + if (ret) + return ret; + + l = list->list.next; + while (l != &list->list) { + DRMLISTDEL(l); + DRMLISTADD(l, &list->free); + list->numOnList--; + l = list->list.next; + } + return drmAdjustListNodes(list); +} + +void driWriteLockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + while(kernelReaders != 0) + _glthread_COND_WAIT(bmCond, bmMutex); +} + +void driWriteUnlockKernelBO(void) +{ + _glthread_UNLOCK_MUTEX(bmMutex); +} + +void driReadLockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + kernelReaders++; + _glthread_UNLOCK_MUTEX(bmMutex); +} + +void driReadUnlockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + if (--kernelReaders == 0) + _glthread_COND_BROADCAST(bmCond); + _glthread_UNLOCK_MUTEX(bmMutex); +} + + + + +/* + * TODO: Introduce fence pools in the same way as + * buffer object pools. + */ + +typedef struct _DriBufferObject +{ + DriBufferPool *pool; + _glthread_Mutex mutex; + int refCount; + const char *name; + uint64_t flags; + unsigned hint; + unsigned alignment; + unsigned createdByReference; + void *private; + /* user-space buffer: */ + unsigned userBuffer; + void *userData; + unsigned userSize; +} DriBufferObject; + +typedef struct _DriBufferList { + drmBOList drmBuffers; /* List of kernel buffers needing validation */ + drmBOList driBuffers; /* List of user-space buffers needing validation */ +} DriBufferList; + + +void +bmError(int val, const char *file, const char *function, int line) +{ + printf("Fatal video memory manager error \"%s\".\n" + "Check kernel logs or set the LIBGL_DEBUG\n" + "environment variable to \"verbose\" for more info.\n" + "Detected in file %s, line %d, function %s.\n", + strerror(-val), file, line, function); +#ifndef NDEBUG + abort(); +#else + abort(); +#endif +} + +extern drmBO * +driBOKernel(struct _DriBufferObject *buf) +{ + drmBO *ret; + + driReadLockKernelBO(); + _glthread_LOCK_MUTEX(buf->mutex); + assert(buf->private != NULL); + ret = buf->pool->kernel(buf->pool, buf->private); + if (!ret) + BM_CKFATAL(-EINVAL); + _glthread_UNLOCK_MUTEX(buf->mutex); + driReadUnlockKernelBO(); + + return ret; +} + +void +driBOWaitIdle(struct _DriBufferObject *buf, int lazy) +{ + + /* + * This function may block. Is it sane to keep the mutex held during + * that time?? + */ + + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy)); + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void * +driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint) +{ + void *virtual; + int retval; + + if (buf->userBuffer) { + return buf->userData; + } + + _glthread_LOCK_MUTEX(buf->mutex); + assert(buf->private != NULL); + retval = buf->pool->map(buf->pool, buf->private, flags, hint, + &buf->mutex, &virtual); + _glthread_UNLOCK_MUTEX(buf->mutex); + + return retval == 0 ? virtual : NULL; +} + +void +driBOUnmap(struct _DriBufferObject *buf) +{ + if (buf->userBuffer) + return; + + assert(buf->private != NULL); + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +unsigned long +driBOOffset(struct _DriBufferObject *buf) +{ + unsigned long ret; + + assert(buf->private != NULL); + + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->offset(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + return ret; +} + +unsigned long +driBOPoolOffset(struct _DriBufferObject *buf) +{ + unsigned long ret; + + assert(buf->private != NULL); + + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->poolOffset(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + return ret; +} + +uint64_t +driBOFlags(struct _DriBufferObject *buf) +{ + uint64_t ret; + + assert(buf->private != NULL); + + driReadLockKernelBO(); + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->flags(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + driReadUnlockKernelBO(); + return ret; +} + +struct _DriBufferObject * +driBOReference(struct _DriBufferObject *buf) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (++buf->refCount == 1) { + _glthread_UNLOCK_MUTEX(buf->mutex); + BM_CKFATAL(-EINVAL); + } + _glthread_UNLOCK_MUTEX(buf->mutex); + return buf; +} + +void +driBOUnReference(struct _DriBufferObject *buf) +{ + int tmp; + + if (!buf) + return; + + _glthread_LOCK_MUTEX(buf->mutex); + tmp = --buf->refCount; + if (!tmp) { + _glthread_UNLOCK_MUTEX(buf->mutex); + if (buf->private) { + if (buf->createdByReference) + buf->pool->unreference(buf->pool, buf->private); + else + buf->pool->destroy(buf->pool, buf->private); + } + if (buf->userBuffer) + num_user_buffers--; + else + num_buffers--; + free(buf); + } else + _glthread_UNLOCK_MUTEX(buf->mutex); + +} + + +int +driBOData(struct _DriBufferObject *buf, + unsigned size, const void *data, + DriBufferPool *newPool, + uint64_t flags) +{ + void *virtual = NULL; + int newBuffer; + int retval = 0; + struct _DriBufferPool *pool; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + pool = buf->pool; + + if (pool == NULL && newPool != NULL) { + buf->pool = newPool; + pool = newPool; + } + if (newPool == NULL) + newPool = pool; + + if (!pool->create) { + assert((size_t)"driBOData called on invalid buffer\n" & 0); + BM_CKFATAL(-EINVAL); + } + + newBuffer = (!buf->private || pool != newPool || + pool->size(pool, buf->private) < size); + + if (!flags) + flags = buf->flags; + + if (newBuffer) { + + if (buf->createdByReference) { + assert((size_t)"driBOData requiring resizing called on shared buffer.\n" & 0); + BM_CKFATAL(-EINVAL); + } + + if (buf->private) + buf->pool->destroy(buf->pool, buf->private); + + pool = newPool; + buf->pool = newPool; + buf->private = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, + buf->alignment); + if (!buf->private) + retval = -ENOMEM; + + if (retval == 0) + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, + DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual); + } else if (pool->map(pool, buf->private, DRM_BO_FLAG_WRITE, + DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual)) { + /* + * Buffer is busy. need to create a new one. + */ + + void *newBuf; + + newBuf = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, + buf->alignment); + if (newBuf) { + buf->pool->destroy(buf->pool, buf->private); + buf->private = newBuf; + } + + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); + } else { + uint64_t flag_diff = flags ^ buf->flags; + + /* + * We might need to change buffer flags. + */ + + if (flag_diff){ + assert(pool->setStatus != NULL); + BM_CKFATAL(pool->unmap(pool, buf->private)); + BM_CKFATAL(pool->setStatus(pool, buf->private, flag_diff, + buf->flags)); + if (!data) + goto out; + + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); + } + } + + if (retval == 0) { + if (data) + memcpy(virtual, data, size); + + BM_CKFATAL(pool->unmap(pool, buf->private)); + } + + out: + _glthread_UNLOCK_MUTEX(buf->mutex); + + return retval; +} + +void +driBOSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, const void *data) +{ + void *virtual; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + if (size && data) { + BM_CKFATAL(buf->pool->map(buf->pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, + &virtual)); + memcpy((unsigned char *) virtual + offset, data, size); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + } + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void +driBOGetSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, void *data) +{ + void *virtual; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + if (size && data) { + BM_CKFATAL(buf->pool->map(buf->pool, buf->private, + DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual)); + memcpy(data, (unsigned char *) virtual + offset, size); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + } + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void +driBOSetReferenced(struct _DriBufferObject *buf, + unsigned long handle) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->private != NULL) { + assert((size_t)"Invalid buffer for setReferenced\n" & 0); + BM_CKFATAL(-EINVAL); + + } + if (buf->pool->reference == NULL) { + assert((size_t)"Invalid buffer pool for setReferenced\n" & 0); + BM_CKFATAL(-EINVAL); + } + buf->private = buf->pool->reference(buf->pool, handle); + if (!buf->private) { + assert((size_t)"Invalid buffer pool for setStatic\n" & 0); + BM_CKFATAL(-ENOMEM); + } + buf->createdByReference = TRUE; + buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags; + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +int +driGenBuffers(struct _DriBufferPool *pool, + const char *name, + unsigned n, + struct _DriBufferObject *buffers[], + unsigned alignment, uint64_t flags, unsigned hint) +{ + struct _DriBufferObject *buf; + int i; + + flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM | + DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + + ++num_buffers; + + assert(pool); + + for (i = 0; i < n; ++i) { + buf = (struct _DriBufferObject *) calloc(1, sizeof(*buf)); + if (!buf) + return -ENOMEM; + + _glthread_INIT_MUTEX(buf->mutex); + _glthread_LOCK_MUTEX(buf->mutex); + buf->refCount = 1; + buf->flags = flags; + buf->hint = hint; + buf->name = name; + buf->alignment = alignment; + buf->pool = pool; + buf->createdByReference = 0; + _glthread_UNLOCK_MUTEX(buf->mutex); + buffers[i] = buf; + } + return 0; +} + +void +driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject **buffers, + void *ptr, unsigned bytes) +{ + const unsigned alignment = 1, flags = 0, hint = 0; + + --num_buffers; /* JB: is inced in GenBuffes */ + driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); + ++num_user_buffers; + + (*buffers)->userBuffer = 1; + (*buffers)->userData = ptr; + (*buffers)->userSize = bytes; +} + +void +driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]) +{ + int i; + + for (i = 0; i < n; ++i) { + driBOUnReference(buffers[i]); + } +} + + +void +driInitBufMgr(int fd) +{ + ; +} + +/* + * Note that lists are per-context and don't need mutex protection. + */ + +struct _DriBufferList * +driBOCreateList(int target) +{ + struct _DriBufferList *list = calloc(sizeof(*list), 1); + + BM_CKFATAL(drmBOCreateList(target, &list->drmBuffers)); + BM_CKFATAL(drmBOCreateList(target, &list->driBuffers)); + return list; +} + +int +driBOResetList(struct _DriBufferList * list) +{ + int ret; + ret = drmBOResetList(&list->drmBuffers); + if (ret) + return ret; + ret = drmBOResetList(&list->driBuffers); + return ret; +} + +void +driBOFreeList(struct _DriBufferList * list) +{ + drmBOFreeList(&list->drmBuffers); + drmBOFreeList(&list->driBuffers); + free(list); +} + + +/* + * Copied from libdrm, because it is needed by driAddValidateItem. + */ + +static drmBONode * +driAddListItem(drmBOList * list, drmBO * item, + uint64_t arg0, uint64_t arg1) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->free.next; + if (l == &list->free) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + return NULL; + } + list->numCurrent++; + } else { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + } + memset(&node->bo_arg, 0, sizeof(node->bo_arg)); + node->buf = item; + node->arg0 = arg0; + node->arg1 = arg1; + DRMLISTADDTAIL(&node->head, &list->list); + list->numOnList++; + return node; +} + +/* + * Slightly modified version compared to the libdrm version. + * This one returns the list index of the buffer put on the list. + */ + +static int +driAddValidateItem(drmBOList * list, drmBO * buf, uint64_t flags, + uint64_t mask, int *itemLoc, + struct _drmBONode **pnode) +{ + drmBONode *node, *cur; + drmMMListHead *l; + int count = 0; + + cur = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + if (node->buf == buf) { + cur = node; + break; + } + count++; + } + if (!cur) { + cur = driAddListItem(list, buf, flags, mask); + if (!cur) + return -ENOMEM; + + cur->arg0 = flags; + cur->arg1 = mask; + } else { + uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; + uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; + + if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { + return -EINVAL; + } + + cur->arg1 |= mask; + cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); + + if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && + (cur->arg0 & DRM_BO_MASK_MEM) == 0) { + return -EINVAL; + } + } + *itemLoc = count; + *pnode = cur; + return 0; +} + + +void +driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf, + uint64_t flags, uint64_t mask, int *itemLoc, + struct _drmBONode **node) +{ + int newItem; + + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(driAddValidateItem(&list->drmBuffers, + buf->pool->kernel(buf->pool, buf->private), + flags, mask, itemLoc, node)); + BM_CKFATAL(drmAddValidateItem(&list->driBuffers, (drmBO *) buf, + flags, mask, &newItem)); + if (newItem) + buf->refCount++; + + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +drmBOList *driGetdrmBOList(struct _DriBufferList *list) +{ + driWriteLockKernelBO(); + return &list->drmBuffers; +} + +void driPutdrmBOList(struct _DriBufferList *list) +{ + driWriteUnlockKernelBO(); +} + + +void +driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->pool->fence) + BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence)); + _glthread_UNLOCK_MUTEX(buf->mutex); + +} + +void +driBOUnrefUserList(struct _DriBufferList *list) +{ + struct _DriBufferObject *buf; + void *curBuf; + + curBuf = drmBOListIterator(&list->driBuffers); + while (curBuf) { + buf = (struct _DriBufferObject *)drmBOListBuf(curBuf); + driBOUnReference(buf); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } +} + +struct _DriFenceObject * +driBOFenceUserList(struct _DriFenceMgr *mgr, + struct _DriBufferList *list, const char *name, + drmFence *kFence) +{ + struct _DriFenceObject *fence; + struct _DriBufferObject *buf; + void *curBuf; + + fence = driFenceCreate(mgr, kFence->fence_class, kFence->type, + kFence, sizeof(*kFence)); + curBuf = drmBOListIterator(&list->driBuffers); + + /* + * User-space fencing callbacks. + */ + + while (curBuf) { + buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); + driBOFence(buf, fence); + driBOUnReference(buf); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } + + driBOResetList(list); + return fence; +} + +void +driBOValidateUserList(struct _DriBufferList * list) +{ + void *curBuf; + struct _DriBufferObject *buf; + + curBuf = drmBOListIterator(&list->driBuffers); + + /* + * User-space validation callbacks. + */ + + while (curBuf) { + buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->pool->validate) + BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex)); + _glthread_UNLOCK_MUTEX(buf->mutex); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } +} + + +void +driPoolTakeDown(struct _DriBufferPool *pool) +{ + pool->takeDown(pool); + +} + +unsigned long +driBOSize(struct _DriBufferObject *buf) +{ + unsigned long size; + + _glthread_LOCK_MUTEX(buf->mutex); + size = buf->pool->size(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + + return size; + +} + +drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list) +{ + return &list->drmBuffers; +} + +drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list) +{ + return &list->driBuffers; +} + diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.h b/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.h new file mode 100644 index 0000000000..e6c0cff0a0 --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.h @@ -0,0 +1,138 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + * Keith Whitwell + */ + +#ifndef _PSB_BUFMGR_H_ +#define _PSB_BUFMGR_H_ +#include +#include "i915_drm.h" +#include "ws_dri_fencemgr.h" + +typedef struct _drmBONode +{ + drmMMListHead head; + drmBO *buf; + struct drm_i915_op_arg bo_arg; + uint64_t arg0; + uint64_t arg1; +} drmBONode; + +typedef struct _drmBOList { + unsigned numTarget; + unsigned numCurrent; + unsigned numOnList; + drmMMListHead list; + drmMMListHead free; +} drmBOList; + + +struct _DriFenceObject; +struct _DriBufferObject; +struct _DriBufferPool; +struct _DriBufferList; + +/* + * Return a pointer to the libdrm buffer object this DriBufferObject + * uses. + */ + +extern drmBO *driBOKernel(struct _DriBufferObject *buf); +extern void *driBOMap(struct _DriBufferObject *buf, unsigned flags, + unsigned hint); +extern void driBOUnmap(struct _DriBufferObject *buf); +extern unsigned long driBOOffset(struct _DriBufferObject *buf); +extern unsigned long driBOPoolOffset(struct _DriBufferObject *buf); + +extern uint64_t driBOFlags(struct _DriBufferObject *buf); +extern struct _DriBufferObject *driBOReference(struct _DriBufferObject *buf); +extern void driBOUnReference(struct _DriBufferObject *buf); + +extern int driBOData(struct _DriBufferObject *r_buf, + unsigned size, const void *data, + struct _DriBufferPool *pool, uint64_t flags); + +extern void driBOSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, + const void *data); +extern void driBOGetSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, + void *data); +extern int driGenBuffers(struct _DriBufferPool *pool, + const char *name, + unsigned n, + struct _DriBufferObject *buffers[], + unsigned alignment, uint64_t flags, unsigned hint); +extern void driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject *buffers[], + void *ptr, unsigned bytes); +extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]); +extern void driInitBufMgr(int fd); +extern struct _DriBufferList *driBOCreateList(int target); +extern int driBOResetList(struct _DriBufferList * list); +extern void driBOAddListItem(struct _DriBufferList * list, + struct _DriBufferObject *buf, + uint64_t flags, uint64_t mask, int *itemLoc, + struct _drmBONode **node); + +extern void driBOValidateList(int fd, struct _DriBufferList * list); +extern void driBOFreeList(struct _DriBufferList * list); +extern struct _DriFenceObject *driBOFenceUserList(struct _DriFenceMgr *mgr, + struct _DriBufferList *list, + const char *name, + drmFence *kFence); +extern void driBOUnrefUserList(struct _DriBufferList *list); +extern void driBOValidateUserList(struct _DriBufferList * list); +extern drmBOList *driGetdrmBOList(struct _DriBufferList *list); +extern void driPutdrmBOList(struct _DriBufferList *list); + +extern void driBOFence(struct _DriBufferObject *buf, + struct _DriFenceObject *fence); + +extern void driPoolTakeDown(struct _DriBufferPool *pool); +extern void driBOSetReferenced(struct _DriBufferObject *buf, + unsigned long handle); +unsigned long driBOSize(struct _DriBufferObject *buf); +extern void driBOWaitIdle(struct _DriBufferObject *buf, int lazy); +extern void driPoolTakeDown(struct _DriBufferPool *pool); + +extern void driReadLockKernelBO(void); +extern void driReadUnlockKernelBO(void); +extern void driWriteLockKernelBO(void); +extern void driWriteUnlockKernelBO(void); + +/* + * For debugging purposes. + */ + +extern drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list); +extern drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list); +#endif diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_bufpool.h b/src/gallium/winsys/common/intel_drm/ws_dri_bufpool.h new file mode 100644 index 0000000000..bf60798924 --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/ws_dri_bufpool.h @@ -0,0 +1,102 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#ifndef _PSB_BUFPOOL_H_ +#define _PSB_BUFPOOL_H_ + +#include +#include +struct _DriFenceObject; + +typedef struct _DriBufferPool +{ + int fd; + int (*map) (struct _DriBufferPool * pool, void *private, + unsigned flags, int hint, _glthread_Mutex *mutex, + void **virtual); + int (*unmap) (struct _DriBufferPool * pool, void *private); + int (*destroy) (struct _DriBufferPool * pool, void *private); + unsigned long (*offset) (struct _DriBufferPool * pool, void *private); + unsigned long (*poolOffset) (struct _DriBufferPool * pool, void *private); + uint64_t (*flags) (struct _DriBufferPool * pool, void *private); + unsigned long (*size) (struct _DriBufferPool * pool, void *private); + void *(*create) (struct _DriBufferPool * pool, unsigned long size, + uint64_t flags, unsigned hint, unsigned alignment); + void *(*reference) (struct _DriBufferPool * pool, unsigned handle); + int (*unreference) (struct _DriBufferPool * pool, void *private); + int (*fence) (struct _DriBufferPool * pool, void *private, + struct _DriFenceObject * fence); + drmBO *(*kernel) (struct _DriBufferPool * pool, void *private); + int (*validate) (struct _DriBufferPool * pool, void *private, _glthread_Mutex *mutex); + int (*waitIdle) (struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, + int lazy); + int (*setStatus) (struct _DriBufferPool *pool, void *private, + uint64_t flag_diff, uint64_t old_flags); + void (*takeDown) (struct _DriBufferPool * pool); + void *data; +} DriBufferPool; + +extern void bmError(int val, const char *file, const char *function, + int line); +#define BM_CKFATAL(val) \ + do{ \ + int tstVal = (val); \ + if (tstVal) \ + bmError(tstVal, __FILE__, __FUNCTION__, __LINE__); \ + } while(0); + + +/* + * Builtin pools. + */ + +/* + * Kernel buffer objects. Size in multiples of page size. Page size aligned. + */ + +extern struct _DriBufferPool *driDRMPoolInit(int fd); +extern struct _DriBufferPool *driMallocPoolInit(void); + +struct _DriFreeSlabManager; +extern struct _DriBufferPool * driSlabPoolInit(int fd, uint64_t flags, + uint64_t validMask, + uint32_t smallestSize, + uint32_t numSizes, + uint32_t desiredNumBuffers, + uint32_t maxSlabSize, + uint32_t pageAlignment, + struct _DriFreeSlabManager *fMan); +extern void driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan); +extern struct _DriFreeSlabManager * +driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec); + + +#endif diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_drmpool.c b/src/gallium/winsys/common/intel_drm/ws_dri_drmpool.c new file mode 100644 index 0000000000..40929efa2f --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/ws_dri_drmpool.c @@ -0,0 +1,268 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" +#include "assert.h" + +/* + * Buffer pool implementation using DRM buffer objects as DRI buffer objects. + */ + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, uint64_t flags, unsigned hint, + unsigned alignment) +{ + drmBO *buf = (drmBO *) malloc(sizeof(*buf)); + int ret; + unsigned pageSize = getpagesize(); + + if (!buf) + return NULL; + + if ((alignment > pageSize) && (alignment % pageSize)) { + free(buf); + return NULL; + } + + ret = drmBOCreate(pool->fd, size, alignment / pageSize, + NULL, + flags, hint, buf); + if (ret) { + free(buf); + return NULL; + } + + return (void *) buf; +} + +static void * +pool_reference(struct _DriBufferPool *pool, unsigned handle) +{ + drmBO *buf = (drmBO *) malloc(sizeof(*buf)); + int ret; + + if (!buf) + return NULL; + + ret = drmBOReference(pool->fd, handle, buf); + + if (ret) { + free(buf); + return NULL; + } + + return (void *) buf; +} + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + int ret; + drmBO *buf = (drmBO *) private; + driReadLockKernelBO(); + ret = drmBOUnreference(pool->fd, buf); + free(buf); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_unreference(struct _DriBufferPool *pool, void *private) +{ + int ret; + drmBO *buf = (drmBO *) private; + driReadLockKernelBO(); + ret = drmBOUnreference(pool->fd, buf); + free(buf); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOMap(pool->fd, buf, flags, hint, virtual); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOUnmap(pool->fd, buf); + driReadUnlockKernelBO(); + + return ret; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + unsigned long offset; + + driReadLockKernelBO(); + assert(buf->flags & DRM_BO_FLAG_NO_MOVE); + offset = buf->offset; + driReadUnlockKernelBO(); + + return buf->offset; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + return 0; +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + uint64_t flags; + + driReadLockKernelBO(); + flags = buf->flags; + driReadUnlockKernelBO(); + + return flags; +} + + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + unsigned long size; + + driReadLockKernelBO(); + size = buf->size; + driReadUnlockKernelBO(); + + return buf->size; +} + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + /* + * Noop. The kernel handles all fencing. + */ + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + return (drmBO *) private; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, + int lazy) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOWaitIdle(pool->fd, buf, (lazy) ? DRM_BO_HINT_WAIT_LAZY:0); + driReadUnlockKernelBO(); + + return ret; +} + + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + free(pool); +} + +/*static int +pool_setStatus(struct _DriBufferPool *pool, void *private, + uint64_t flag_diff, uint64_t old_flags) +{ + drmBO *buf = (drmBO *) private; + uint64_t new_flags = old_flags ^ flag_diff; + int ret; + + driReadLockKernelBO(); + ret = drmBOSetStatus(pool->fd, buf, new_flags, flag_diff, + 0, 0, 0); + driReadUnlockKernelBO(); + return ret; +}*/ + +struct _DriBufferPool * +driDRMPoolInit(int fd) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + + if (!pool) + return NULL; + + pool->fd = fd; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->poolOffset = &pool_poolOffset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = NULL; + pool->waitIdle = &pool_waitIdle; + pool->takeDown = &pool_takedown; + pool->reference = &pool_reference; + pool->unreference = &pool_unreference; + pool->data = NULL; + return pool; +} diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.c b/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.c new file mode 100644 index 0000000000..b56bc269da --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.c @@ -0,0 +1,377 @@ +#include "ws_dri_fencemgr.h" +#include "glthread.h" +#include +#include +#include + +/* + * Note: Locking order is + * _DriFenceObject::mutex + * _DriFenceMgr::mutex + */ + +struct _DriFenceMgr { + /* + * Constant members. Need no mutex protection. + */ + struct _DriFenceMgrCreateInfo info; + void *private; + + /* + * These members are protected by this->mutex + */ + _glthread_Mutex mutex; + int refCount; + drmMMListHead *heads; + int num_fences; +}; + +struct _DriFenceObject { + + /* + * These members are constant and need no mutex protection. + */ + struct _DriFenceMgr *mgr; + uint32_t fence_class; + uint32_t fence_type; + + /* + * These members are protected by mgr->mutex. + */ + drmMMListHead head; + int refCount; + + /* + * These members are protected by this->mutex. + */ + _glthread_Mutex mutex; + uint32_t signaled_type; + void *private; +}; + +uint32_t +driFenceType(struct _DriFenceObject *fence) +{ + return fence->fence_type; +} + +struct _DriFenceMgr * +driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info) +{ + struct _DriFenceMgr *tmp; + uint32_t i; + + tmp = calloc(1, sizeof(*tmp)); + if (!tmp) + return NULL; + + _glthread_INIT_MUTEX(tmp->mutex); + _glthread_LOCK_MUTEX(tmp->mutex); + tmp->refCount = 1; + tmp->info = *info; + tmp->num_fences = 0; + tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads)); + if (!tmp->heads) + goto out_err; + + for (i=0; iinfo.num_classes; ++i) { + DRMINITLISTHEAD(&tmp->heads[i]); + } + _glthread_UNLOCK_MUTEX(tmp->mutex); + return tmp; + + out_err: + if (tmp) + free(tmp); + return NULL; +} + +static void +driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr) +{ + struct _DriFenceMgr *mgr = *pMgr; + + *pMgr = NULL; + if (--mgr->refCount == 0) + free(mgr); + else + _glthread_UNLOCK_MUTEX(mgr->mutex); +} + +void +driFenceMgrUnReference(struct _DriFenceMgr **pMgr) +{ + _glthread_LOCK_MUTEX((*pMgr)->mutex); + driFenceMgrUnrefUnlock(pMgr); +} + +static void +driFenceUnReferenceLocked(struct _DriFenceObject **pFence) +{ + struct _DriFenceObject *fence = *pFence; + struct _DriFenceMgr *mgr = fence->mgr; + + *pFence = NULL; + if (--fence->refCount == 0) { + DRMLISTDELINIT(&fence->head); + if (fence->private) + mgr->info.unreference(mgr, &fence->private); + --mgr->num_fences; + fence->mgr = NULL; + --mgr->refCount; + free(fence); + + } +} + + +static void +driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr, + drmMMListHead *list, + uint32_t fence_class, + uint32_t fence_type) +{ + struct _DriFenceObject *entry; + drmMMListHead *prev; + + while(list != &mgr->heads[fence_class]) { + entry = DRMLISTENTRY(struct _DriFenceObject, list, head); + + /* + * Up refcount so that entry doesn't disappear from under us + * when we unlock-relock mgr to get the correct locking order. + */ + + ++entry->refCount; + _glthread_UNLOCK_MUTEX(mgr->mutex); + _glthread_LOCK_MUTEX(entry->mutex); + _glthread_LOCK_MUTEX(mgr->mutex); + + prev = list->prev; + + + + if (list->prev == list) { + + /* + * Somebody else removed the entry from the list. + */ + + _glthread_UNLOCK_MUTEX(entry->mutex); + driFenceUnReferenceLocked(&entry); + return; + } + + entry->signaled_type |= (fence_type & entry->fence_type); + if (entry->signaled_type == entry->fence_type) { + DRMLISTDELINIT(list); + mgr->info.unreference(mgr, &entry->private); + } + _glthread_UNLOCK_MUTEX(entry->mutex); + driFenceUnReferenceLocked(&entry); + list = prev; + } +} + + +int +driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, + int lazy_hint) +{ + struct _DriFenceMgr *mgr = fence->mgr; + int ret = 0; + + _glthread_LOCK_MUTEX(fence->mutex); + + if ((fence->signaled_type & fence_type) == fence_type) + goto out0; + + ret = mgr->info.finish(mgr, fence->private, fence_type, lazy_hint); + if (ret) + goto out0; + + _glthread_LOCK_MUTEX(mgr->mutex); + _glthread_UNLOCK_MUTEX(fence->mutex); + + driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, + fence_type); + _glthread_UNLOCK_MUTEX(mgr->mutex); + return 0; + + out0: + _glthread_UNLOCK_MUTEX(fence->mutex); + return ret; +} + +uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence) +{ + uint32_t ret; + + _glthread_LOCK_MUTEX(fence->mutex); + ret = fence->signaled_type; + _glthread_UNLOCK_MUTEX(fence->mutex); + + return ret; +} + +int +driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type, + uint32_t *signaled) +{ + int ret = 0; + struct _DriFenceMgr *mgr; + + _glthread_LOCK_MUTEX(fence->mutex); + mgr = fence->mgr; + *signaled = fence->signaled_type; + if ((fence->signaled_type & flush_type) == flush_type) + goto out0; + + ret = mgr->info.signaled(mgr, fence->private, flush_type, signaled); + if (ret) { + *signaled = fence->signaled_type; + goto out0; + } + + if ((fence->signaled_type | *signaled) == fence->signaled_type) + goto out0; + + _glthread_LOCK_MUTEX(mgr->mutex); + _glthread_UNLOCK_MUTEX(fence->mutex); + + driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, + *signaled); + + _glthread_UNLOCK_MUTEX(mgr->mutex); + return 0; + out0: + _glthread_UNLOCK_MUTEX(fence->mutex); + return ret; +} + +struct _DriFenceObject * +driFenceReference(struct _DriFenceObject *fence) +{ + _glthread_LOCK_MUTEX(fence->mgr->mutex); + ++fence->refCount; + _glthread_UNLOCK_MUTEX(fence->mgr->mutex); + return fence; +} + +void +driFenceUnReference(struct _DriFenceObject **pFence) +{ + struct _DriFenceMgr *mgr; + + if (*pFence == NULL) + return; + + mgr = (*pFence)->mgr; + _glthread_LOCK_MUTEX(mgr->mutex); + ++mgr->refCount; + driFenceUnReferenceLocked(pFence); + driFenceMgrUnrefUnlock(&mgr); +} + +struct _DriFenceObject +*driFenceCreate(struct _DriFenceMgr *mgr, uint32_t fence_class, + uint32_t fence_type, void *private, size_t private_size) +{ + struct _DriFenceObject *fence; + size_t fence_size = sizeof(*fence); + + if (private_size) + fence_size = ((fence_size + 15) & ~15); + + fence = calloc(1, fence_size + private_size); + + if (!fence) { + int ret = mgr->info.finish(mgr, private, fence_type, 0); + + if (ret) + usleep(10000000); + + return NULL; + } + + _glthread_INIT_MUTEX(fence->mutex); + _glthread_LOCK_MUTEX(fence->mutex); + _glthread_LOCK_MUTEX(mgr->mutex); + fence->refCount = 1; + DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]); + fence->mgr = mgr; + ++mgr->refCount; + ++mgr->num_fences; + _glthread_UNLOCK_MUTEX(mgr->mutex); + fence->fence_class = fence_class; + fence->fence_type = fence_type; + fence->signaled_type = 0; + fence->private = private; + if (private_size) { + fence->private = (void *)(((uint8_t *) fence) + fence_size); + memcpy(fence->private, private, private_size); + } + + _glthread_UNLOCK_MUTEX(fence->mutex); + return fence; +} + + +static int +tSignaled(struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, + uint32_t *signaled_type) +{ + long fd = (long) mgr->private; + int dummy; + drmFence *fence = (drmFence *) private; + int ret; + + *signaled_type = 0; + ret = drmFenceSignaled((int) fd, fence, flush_type, &dummy); + if (ret) + return ret; + + *signaled_type = fence->signaled; + + return 0; +} + +static int +tFinish(struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, + int lazy_hint) +{ + long fd = (long) mgr->private; + unsigned flags = lazy_hint ? DRM_FENCE_FLAG_WAIT_LAZY : 0; + + return drmFenceWait((int)fd, flags, (drmFence *) private, fence_type); +} + +static int +tUnref(struct _DriFenceMgr *mgr, void **private) +{ + long fd = (long) mgr->private; + drmFence *fence = (drmFence *) *private; + *private = NULL; + + return drmFenceUnreference(fd, fence); +} + +struct _DriFenceMgr *driFenceMgrTTMInit(int fd) +{ + struct _DriFenceMgrCreateInfo info; + struct _DriFenceMgr *mgr; + + info.flags = DRI_FENCE_CLASS_ORDERED; + info.num_classes = 4; + info.signaled = tSignaled; + info.finish = tFinish; + info.unreference = tUnref; + + mgr = driFenceMgrCreate(&info); + if (mgr == NULL) + return NULL; + + mgr->private = (void *) (long) fd; + return mgr; +} + diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.h b/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.h new file mode 100644 index 0000000000..4ea58dfe18 --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.h @@ -0,0 +1,115 @@ +#ifndef DRI_FENCEMGR_H +#define DRI_FENCEMGR_H + +#include +#include + +struct _DriFenceObject; +struct _DriFenceMgr; + +/* + * Do a quick check to see if the fence manager has registered the fence + * object as signaled. Note that this function may return a false negative + * answer. + */ +extern uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence); + +/* + * Check if the fence object is signaled. This function can be substantially + * more expensive to call than the above function, but will not return a false + * negative answer. The argument "flush_type" sets the types that the + * underlying mechanism must make sure will eventually signal. + */ +extern int driFenceSignaledType(struct _DriFenceObject *fence, + uint32_t flush_type, uint32_t *signaled); + +/* + * Convenience functions. + */ + +static inline int driFenceSignaled(struct _DriFenceObject *fence, + uint32_t flush_type) +{ + uint32_t signaled_types; + int ret = driFenceSignaledType(fence, flush_type, &signaled_types); + if (ret) + return 0; + return ((signaled_types & flush_type) == flush_type); +} + +static inline int driFenceSignaledCached(struct _DriFenceObject *fence, + uint32_t flush_type) +{ + uint32_t signaled_types = + driFenceSignaledTypeCached(fence); + + return ((signaled_types & flush_type) == flush_type); +} + +/* + * Reference a fence object. + */ +extern struct _DriFenceObject *driFenceReference(struct _DriFenceObject *fence); + +/* + * Unreference a fence object. The fence object pointer will be reset to NULL. + */ + +extern void driFenceUnReference(struct _DriFenceObject **pFence); + + +/* + * Wait for a fence to signal the indicated fence_type. + * If "lazy_hint" is true, it indicates that the wait may sleep to avoid + * busy-wait polling. + */ +extern int driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, + int lazy_hint); + +/* + * Create a DriFenceObject for manager "mgr". + * + * "private" is a pointer that should be used for the callbacks in + * struct _DriFenceMgrCreateInfo. + * + * if private_size is nonzero, then the info stored at *private, with size + * private size will be copied and the fence manager will instead use a + * pointer to the copied data for the callbacks in + * struct _DriFenceMgrCreateInfo. In that case, the object pointed to by + * "private" may be destroyed after the call to driFenceCreate. + */ +extern struct _DriFenceObject *driFenceCreate(struct _DriFenceMgr *mgr, + uint32_t fence_class, + uint32_t fence_type, + void *private, + size_t private_size); + +extern uint32_t driFenceType(struct _DriFenceObject *fence); + +/* + * Fence creations are ordered. If a fence signals a fence_type, + * it is safe to assume that all fences of the same class that was + * created before that fence has signaled the same type. + */ + +#define DRI_FENCE_CLASS_ORDERED (1 << 0) + +struct _DriFenceMgrCreateInfo { + uint32_t flags; + uint32_t num_classes; + int (*signaled) (struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, + uint32_t *signaled_type); + int (*finish) (struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, int lazy_hint); + int (*unreference) (struct _DriFenceMgr *mgr, void **private); +}; + +extern struct _DriFenceMgr * +driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info); + +void +driFenceMgrUnReference(struct _DriFenceMgr **pMgr); + +extern struct _DriFenceMgr * +driFenceMgrTTMInit(int fd); + +#endif diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_mallocpool.c b/src/gallium/winsys/common/intel_drm/ws_dri_mallocpool.c new file mode 100644 index 0000000000..a80555c9c7 --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/ws_dri_mallocpool.c @@ -0,0 +1,161 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "pipe/p_debug.h" +#include "glthread.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, uint64_t flags, unsigned hint, + unsigned alignment) +{ + unsigned long *private = malloc(size + 2*sizeof(unsigned long)); + if ((flags & DRM_BO_MASK_MEM) != DRM_BO_FLAG_MEM_LOCAL) + abort(); + + *private = size; + return (void *)private; +} + + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + free(private); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, + _glthread_Mutex *mutex, int lazy) +{ + return 0; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + *virtual = (void *)((unsigned long *)private + 2); + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + /* + * BUG + */ + abort(); + return 0UL; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + /* + * BUG + */ + abort(); +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + return DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + return *(unsigned long *) private; +} + + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + abort(); + return 0UL; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + abort(); + return NULL; +} + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + free(pool); +} + + +struct _DriBufferPool * +driMallocPoolInit(void) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + if (!pool) + return NULL; + + pool->data = NULL; + pool->fd = -1; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->poolOffset = &pool_poolOffset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = NULL; + pool->waitIdle = &pool_waitIdle; + pool->takeDown = &pool_takedown; + return pool; +} diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_slabpool.c b/src/gallium/winsys/common/intel_drm/ws_dri_slabpool.c new file mode 100644 index 0000000000..dfcf6d6b19 --- /dev/null +++ b/src/gallium/winsys/common/intel_drm/ws_dri_slabpool.c @@ -0,0 +1,968 @@ +/************************************************************************** + * + * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellstrom + */ + +#include +#include +#include +#include +#include +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" +#include "ws_dri_bufmgr.h" +#include "glthread.h" + +#define DRI_SLABPOOL_ALLOC_RETRIES 100 + +struct _DriSlab; + +struct _DriSlabBuffer { + int isSlabBuffer; + drmBO *bo; + struct _DriFenceObject *fence; + struct _DriSlab *parent; + drmMMListHead head; + uint32_t mapCount; + uint32_t start; + uint32_t fenceType; + int unFenced; + _glthread_Cond event; +}; + +struct _DriKernelBO { + int fd; + drmBO bo; + drmMMListHead timeoutHead; + drmMMListHead head; + struct timeval timeFreed; + uint32_t pageAlignment; + void *virtual; +}; + +struct _DriSlab{ + drmMMListHead head; + drmMMListHead freeBuffers; + uint32_t numBuffers; + uint32_t numFree; + struct _DriSlabBuffer *buffers; + struct _DriSlabSizeHeader *header; + struct _DriKernelBO *kbo; +}; + + +struct _DriSlabSizeHeader { + drmMMListHead slabs; + drmMMListHead freeSlabs; + drmMMListHead delayedBuffers; + uint32_t numDelayed; + struct _DriSlabPool *slabPool; + uint32_t bufSize; + _glthread_Mutex mutex; +}; + +struct _DriFreeSlabManager { + struct timeval slabTimeout; + struct timeval checkInterval; + struct timeval nextCheck; + drmMMListHead timeoutList; + drmMMListHead unCached; + drmMMListHead cached; + _glthread_Mutex mutex; +}; + + +struct _DriSlabPool { + + /* + * The data of this structure remains constant after + * initialization and thus needs no mutex protection. + */ + + struct _DriFreeSlabManager *fMan; + uint64_t proposedFlags; + uint64_t validMask; + uint32_t *bucketSizes; + uint32_t numBuckets; + uint32_t pageSize; + int fd; + int pageAlignment; + int maxSlabSize; + int desiredNumBuffers; + struct _DriSlabSizeHeader *headers; +}; + +/* + * FIXME: Perhaps arrange timeout slabs in size buckets for fast + * retreival?? + */ + + +static inline int +driTimeAfterEq(struct timeval *arg1, struct timeval *arg2) +{ + return ((arg1->tv_sec > arg2->tv_sec) || + ((arg1->tv_sec == arg2->tv_sec) && + (arg1->tv_usec > arg2->tv_usec))); +} + +static inline void +driTimeAdd(struct timeval *arg, struct timeval *add) +{ + unsigned int sec; + + arg->tv_sec += add->tv_sec; + arg->tv_usec += add->tv_usec; + sec = arg->tv_usec / 1000000; + arg->tv_sec += sec; + arg->tv_usec -= sec*1000000; +} + +static void +driFreeKernelBO(struct _DriKernelBO *kbo) +{ + if (!kbo) + return; + + (void) drmBOUnreference(kbo->fd, &kbo->bo); + free(kbo); +} + + +static void +driFreeTimeoutKBOsLocked(struct _DriFreeSlabManager *fMan, + struct timeval *time) +{ + drmMMListHead *list, *next; + struct _DriKernelBO *kbo; + + if (!driTimeAfterEq(time, &fMan->nextCheck)) + return; + + for (list = fMan->timeoutList.next, next = list->next; + list != &fMan->timeoutList; + list = next, next = list->next) { + + kbo = DRMLISTENTRY(struct _DriKernelBO, list, timeoutHead); + + if (!driTimeAfterEq(time, &kbo->timeFreed)) + break; + + DRMLISTDELINIT(&kbo->timeoutHead); + DRMLISTDELINIT(&kbo->head); + driFreeKernelBO(kbo); + } + + fMan->nextCheck = *time; + driTimeAdd(&fMan->nextCheck, &fMan->checkInterval); +} + + +/* + * Add a _DriKernelBO to the free slab manager. + * This means that it is available for reuse, but if it's not + * reused in a while, it will be freed. + */ + +static void +driSetKernelBOFree(struct _DriFreeSlabManager *fMan, + struct _DriKernelBO *kbo) +{ + struct timeval time; + + _glthread_LOCK_MUTEX(fMan->mutex); + gettimeofday(&time, NULL); + driTimeAdd(&time, &fMan->slabTimeout); + + kbo->timeFreed = time; + + if (kbo->bo.flags & DRM_BO_FLAG_CACHED) + DRMLISTADD(&kbo->head, &fMan->cached); + else + DRMLISTADD(&kbo->head, &fMan->unCached); + + DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList); + driFreeTimeoutKBOsLocked(fMan, &time); + + _glthread_UNLOCK_MUTEX(fMan->mutex); +} + +/* + * Get a _DriKernelBO for us to use as storage for a slab. + * + */ + +static struct _DriKernelBO * +driAllocKernelBO(struct _DriSlabSizeHeader *header) + +{ + struct _DriSlabPool *slabPool = header->slabPool; + struct _DriFreeSlabManager *fMan = slabPool->fMan; + drmMMListHead *list, *next, *head; + uint32_t size = header->bufSize * slabPool->desiredNumBuffers; + struct _DriKernelBO *kbo; + struct _DriKernelBO *kboTmp; + int ret; + + /* + * FIXME: We should perhaps allow some variation in slabsize in order + * to efficiently reuse slabs. + */ + + size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize; + size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1); + _glthread_LOCK_MUTEX(fMan->mutex); + + kbo = NULL; + + retry: + head = (slabPool->proposedFlags & DRM_BO_FLAG_CACHED) ? + &fMan->cached : &fMan->unCached; + + for (list = head->next, next = list->next; + list != head; + list = next, next = list->next) { + + kboTmp = DRMLISTENTRY(struct _DriKernelBO, list, head); + + if ((kboTmp->bo.size == size) && + (slabPool->pageAlignment == 0 || + (kboTmp->pageAlignment % slabPool->pageAlignment) == 0)) { + + if (!kbo) + kbo = kboTmp; + + if ((kbo->bo.proposedFlags ^ slabPool->proposedFlags) == 0) + break; + + } + } + + if (kbo) { + DRMLISTDELINIT(&kbo->head); + DRMLISTDELINIT(&kbo->timeoutHead); + } + + _glthread_UNLOCK_MUTEX(fMan->mutex); + + if (kbo) { + uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags; + + ret = 0; + if (new_mask) { + ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, + new_mask, DRM_BO_HINT_DONT_FENCE, 0, 0); + } + if (ret == 0) + return kbo; + + driFreeKernelBO(kbo); + kbo = NULL; + goto retry; + } + + kbo = calloc(1, sizeof(struct _DriKernelBO)); + if (!kbo) + return NULL; + + kbo->fd = slabPool->fd; + DRMINITLISTHEAD(&kbo->head); + DRMINITLISTHEAD(&kbo->timeoutHead); + ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL, + slabPool->proposedFlags, + DRM_BO_HINT_DONT_FENCE, &kbo->bo); + if (ret) + goto out_err0; + + ret = drmBOMap(kbo->fd, &kbo->bo, + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, &kbo->virtual); + + if (ret) + goto out_err1; + + ret = drmBOUnmap(kbo->fd, &kbo->bo); + if (ret) + goto out_err1; + + return kbo; + + out_err1: + drmBOUnreference(kbo->fd, &kbo->bo); + out_err0: + free(kbo); + return NULL; +} + + +static int +driAllocSlab(struct _DriSlabSizeHeader *header) +{ + struct _DriSlab *slab; + struct _DriSlabBuffer *buf; + uint32_t numBuffers; + int ret; + int i; + + slab = calloc(1, sizeof(*slab)); + if (!slab) + return -ENOMEM; + + slab->kbo = driAllocKernelBO(header); + if (!slab->kbo) { + ret = -ENOMEM; + goto out_err0; + } + + numBuffers = slab->kbo->bo.size / header->bufSize; + + slab->buffers = calloc(numBuffers, sizeof(*slab->buffers)); + if (!slab->buffers) { + ret = -ENOMEM; + goto out_err1; + } + + DRMINITLISTHEAD(&slab->head); + DRMINITLISTHEAD(&slab->freeBuffers); + slab->numBuffers = numBuffers; + slab->numFree = 0; + slab->header = header; + + buf = slab->buffers; + for (i=0; i < numBuffers; ++i) { + buf->parent = slab; + buf->start = i* header->bufSize; + buf->mapCount = 0; + buf->isSlabBuffer = 1; + _glthread_INIT_COND(buf->event); + DRMLISTADDTAIL(&buf->head, &slab->freeBuffers); + slab->numFree++; + buf++; + } + + DRMLISTADDTAIL(&slab->head, &header->slabs); + + return 0; + + out_err1: + driSetKernelBOFree(header->slabPool->fMan, slab->kbo); + free(slab->buffers); + out_err0: + free(slab); + return ret; +} + +/* + * Delete a buffer from the slab header delayed list and put + * it on the slab free list. + */ + +static void +driSlabFreeBufferLocked(struct _DriSlabBuffer *buf) +{ + struct _DriSlab *slab = buf->parent; + struct _DriSlabSizeHeader *header = slab->header; + drmMMListHead *list = &buf->head; + + DRMLISTDEL(list); + DRMLISTADDTAIL(list, &slab->freeBuffers); + slab->numFree++; + + if (slab->head.next == &slab->head) + DRMLISTADDTAIL(&slab->head, &header->slabs); + + if (slab->numFree == slab->numBuffers) { + list = &slab->head; + DRMLISTDEL(list); + DRMLISTADDTAIL(list, &header->freeSlabs); + } + + if (header->slabs.next == &header->slabs || + slab->numFree != slab->numBuffers) { + + drmMMListHead *next; + struct _DriFreeSlabManager *fMan = header->slabPool->fMan; + + for (list = header->freeSlabs.next, next = list->next; + list != &header->freeSlabs; + list = next, next = list->next) { + + slab = DRMLISTENTRY(struct _DriSlab, list, head); + + DRMLISTDELINIT(list); + driSetKernelBOFree(fMan, slab->kbo); + free(slab->buffers); + free(slab); + } + } +} + +static void +driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait) +{ + drmMMListHead *list, *prev, *first; + struct _DriSlabBuffer *buf; + struct _DriSlab *slab; + int firstWasSignaled = 1; + int signaled; + int i; + int ret; + + /* + * Rerun the freeing test if the youngest tested buffer + * was signaled, since there might be more idle buffers + * in the delay list. + */ + + while (firstWasSignaled) { + firstWasSignaled = 0; + signaled = 0; + first = header->delayedBuffers.next; + + /* Only examine the oldest 1/3 of delayed buffers: + */ + if (header->numDelayed > 3) { + for (i = 0; i < header->numDelayed; i += 3) { + first = first->next; + } + } + + for (list = first, prev = list->prev; + list != &header->delayedBuffers; + list = prev, prev = list->prev) { + buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); + slab = buf->parent; + + if (!signaled) { + if (wait) { + ret = driFenceFinish(buf->fence, buf->fenceType, 0); + if (ret) + break; + signaled = 1; + wait = 0; + } else { + signaled = driFenceSignaled(buf->fence, buf->fenceType); + } + if (signaled) { + if (list == first) + firstWasSignaled = 1; + driFenceUnReference(&buf->fence); + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { + driFenceUnReference(&buf->fence); + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + } + } +} + + +static struct _DriSlabBuffer * +driSlabAllocBuffer(struct _DriSlabSizeHeader *header) +{ + static struct _DriSlabBuffer *buf; + struct _DriSlab *slab; + drmMMListHead *list; + int count = DRI_SLABPOOL_ALLOC_RETRIES; + + _glthread_LOCK_MUTEX(header->mutex); + while(header->slabs.next == &header->slabs && count > 0) { + driSlabCheckFreeLocked(header, 0); + if (header->slabs.next != &header->slabs) + break; + + _glthread_UNLOCK_MUTEX(header->mutex); + if (count != DRI_SLABPOOL_ALLOC_RETRIES) + usleep(1); + _glthread_LOCK_MUTEX(header->mutex); + (void) driAllocSlab(header); + count--; + } + + list = header->slabs.next; + if (list == &header->slabs) { + _glthread_UNLOCK_MUTEX(header->mutex); + return NULL; + } + slab = DRMLISTENTRY(struct _DriSlab, list, head); + if (--slab->numFree == 0) + DRMLISTDELINIT(list); + + list = slab->freeBuffers.next; + DRMLISTDELINIT(list); + + _glthread_UNLOCK_MUTEX(header->mutex); + buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); + return buf; +} + +static void * +pool_create(struct _DriBufferPool *driPool, unsigned long size, + uint64_t flags, unsigned hint, unsigned alignment) +{ + struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; + struct _DriSlabSizeHeader *header; + struct _DriSlabBuffer *buf; + void *dummy; + int i; + int ret; + + /* + * FIXME: Check for compatibility. + */ + + header = pool->headers; + for (i=0; inumBuckets; ++i) { + if (header->bufSize >= size) + break; + header++; + } + + if (i < pool->numBuckets) + return driSlabAllocBuffer(header); + + + /* + * Fall back to allocate a buffer object directly from DRM. + * and wrap it in a driBO structure. + */ + + + buf = calloc(1, sizeof(*buf)); + + if (!buf) + return NULL; + + buf->bo = calloc(1, sizeof(*buf->bo)); + if (!buf->bo) + goto out_err0; + + if (alignment) { + if ((alignment < pool->pageSize) && (pool->pageSize % alignment)) + goto out_err1; + if ((alignment > pool->pageSize) && (alignment % pool->pageSize)) + goto out_err1; + } + + ret = drmBOCreate(pool->fd, size, alignment / pool->pageSize, NULL, + flags, hint, buf->bo); + if (ret) + goto out_err1; + + ret = drmBOMap(pool->fd, buf->bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, &dummy); + if (ret) + goto out_err2; + + ret = drmBOUnmap(pool->fd, buf->bo); + if (ret) + goto out_err2; + + return buf; + out_err2: + drmBOUnreference(pool->fd, buf->bo); + out_err1: + free(buf->bo); + out_err0: + free(buf); + return NULL; +} + +static int +pool_destroy(struct _DriBufferPool *driPool, void *private) +{ + struct _DriSlabBuffer *buf = + (struct _DriSlabBuffer *) private; + struct _DriSlab *slab; + struct _DriSlabSizeHeader *header; + + if (!buf->isSlabBuffer) { + struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; + int ret; + + ret = drmBOUnreference(pool->fd, buf->bo); + free(buf->bo); + free(buf); + return ret; + } + + slab = buf->parent; + header = slab->header; + + _glthread_LOCK_MUTEX(header->mutex); + buf->unFenced = 0; + buf->mapCount = 0; + + if (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)) { + DRMLISTADDTAIL(&buf->head, &header->delayedBuffers); + header->numDelayed++; + } else { + if (buf->fence) + driFenceUnReference(&buf->fence); + driSlabFreeBufferLocked(buf); + } + + _glthread_UNLOCK_MUTEX(header->mutex); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *driPool, void *private, + _glthread_Mutex *mutex, int lazy) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + while(buf->unFenced) + _glthread_COND_WAIT(buf->event, *mutex); + + if (!buf->fence) + return 0; + + driFenceFinish(buf->fence, buf->fenceType, lazy); + driFenceUnReference(&buf->fence); + + return 0; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + int busy; + + if (buf->isSlabBuffer) + busy = buf->unFenced || (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)); + else + busy = buf->fence && !driFenceSignaled(buf->fence, buf->fenceType); + + + if (busy) { + if (hint & DRM_BO_HINT_DONT_BLOCK) + return -EBUSY; + else { + (void) pool_waitIdle(pool, private, mutex, 0); + } + } + + ++buf->mapCount; + *virtual = (buf->isSlabBuffer) ? + (void *) ((uint8_t *) buf->parent->kbo->virtual + buf->start) : + (void *) buf->bo->virtual; + + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + --buf->mapCount; + if (buf->mapCount == 0 && buf->isSlabBuffer) + _glthread_COND_BROADCAST(buf->event); + + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + struct _DriSlab *slab; + struct _DriSlabSizeHeader *header; + + if (!buf->isSlabBuffer) { + assert(buf->bo->proposedFlags & DRM_BO_FLAG_NO_MOVE); + return buf->bo->offset; + } + + slab = buf->parent; + header = slab->header; + + (void) header; + assert(header->slabPool->proposedFlags & DRM_BO_FLAG_NO_MOVE); + return slab->kbo->bo.offset + buf->start; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + return buf->start; +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + if (!buf->isSlabBuffer) + return buf->bo->flags; + + return buf->parent->kbo->bo.flags; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + if (!buf->isSlabBuffer) + return buf->bo->size; + + return buf->parent->header->bufSize; +} + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + drmBO *bo; + + if (buf->fence) + driFenceUnReference(&buf->fence); + + buf->fence = driFenceReference(fence); + bo = (buf->isSlabBuffer) ? + &buf->parent->kbo->bo: + buf->bo; + buf->fenceType = bo->fenceFlags; + + buf->unFenced = 0; + _glthread_COND_BROADCAST(buf->event); + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + return (buf->isSlabBuffer) ? &buf->parent->kbo->bo : buf->bo; +} + +static int +pool_validate(struct _DriBufferPool *pool, void *private, + _glthread_Mutex *mutex) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + if (!buf->isSlabBuffer) + return 0; + + while(buf->mapCount != 0) + _glthread_COND_WAIT(buf->event, *mutex); + + buf->unFenced = 1; + return 0; +} + + +struct _DriFreeSlabManager * +driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec) +{ + struct _DriFreeSlabManager *tmp; + + tmp = calloc(1, sizeof(*tmp)); + if (!tmp) + return NULL; + + _glthread_INIT_MUTEX(tmp->mutex); + _glthread_LOCK_MUTEX(tmp->mutex); + tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000; + tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000; + tmp->slabTimeout.tv_usec -= tmp->slabTimeout.tv_sec*1000000; + + tmp->checkInterval.tv_usec = checkIntervalMsec*1000; + tmp->checkInterval.tv_sec = tmp->checkInterval.tv_usec / 1000000; + tmp->checkInterval.tv_usec -= tmp->checkInterval.tv_sec*1000000; + + gettimeofday(&tmp->nextCheck, NULL); + driTimeAdd(&tmp->nextCheck, &tmp->checkInterval); + DRMINITLISTHEAD(&tmp->timeoutList); + DRMINITLISTHEAD(&tmp->unCached); + DRMINITLISTHEAD(&tmp->cached); + _glthread_UNLOCK_MUTEX(tmp->mutex); + + return tmp; +} + +void +driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan) +{ + struct timeval time; + + time = fMan->nextCheck; + driTimeAdd(&time, &fMan->checkInterval); + + _glthread_LOCK_MUTEX(fMan->mutex); + driFreeTimeoutKBOsLocked(fMan, &time); + _glthread_UNLOCK_MUTEX(fMan->mutex); + + assert(fMan->timeoutList.next == &fMan->timeoutList); + assert(fMan->unCached.next == &fMan->unCached); + assert(fMan->cached.next == &fMan->cached); + + free(fMan); +} + +static void +driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size, + struct _DriSlabSizeHeader *header) +{ + _glthread_INIT_MUTEX(header->mutex); + _glthread_LOCK_MUTEX(header->mutex); + + DRMINITLISTHEAD(&header->slabs); + DRMINITLISTHEAD(&header->freeSlabs); + DRMINITLISTHEAD(&header->delayedBuffers); + + header->numDelayed = 0; + header->slabPool = pool; + header->bufSize = size; + + _glthread_UNLOCK_MUTEX(header->mutex); +} + +static void +driFinishSizeHeader(struct _DriSlabSizeHeader *header) +{ + drmMMListHead *list, *next; + struct _DriSlabBuffer *buf; + + _glthread_LOCK_MUTEX(header->mutex); + for (list = header->delayedBuffers.next, next = list->next; + list != &header->delayedBuffers; + list = next, next = list->next) { + + buf = DRMLISTENTRY(struct _DriSlabBuffer, list , head); + if (buf->fence) { + (void) driFenceFinish(buf->fence, buf->fenceType, 0); + driFenceUnReference(&buf->fence); + } + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + _glthread_UNLOCK_MUTEX(header->mutex); +} + +static void +pool_takedown(struct _DriBufferPool *driPool) +{ + struct _DriSlabPool *pool = driPool->data; + int i; + + for (i=0; inumBuckets; ++i) { + driFinishSizeHeader(&pool->headers[i]); + } + + free(pool->headers); + free(pool->bucketSizes); + free(pool); + free(driPool); +} + +struct _DriBufferPool * +driSlabPoolInit(int fd, uint64_t flags, + uint64_t validMask, + uint32_t smallestSize, + uint32_t numSizes, + uint32_t desiredNumBuffers, + uint32_t maxSlabSize, + uint32_t pageAlignment, + struct _DriFreeSlabManager *fMan) +{ + struct _DriBufferPool *driPool; + struct _DriSlabPool *pool; + uint32_t i; + + driPool = calloc(1, sizeof(*driPool)); + if (!driPool) + return NULL; + + pool = calloc(1, sizeof(*pool)); + if (!pool) + goto out_err0; + + pool->bucketSizes = calloc(numSizes, sizeof(*pool->bucketSizes)); + if (!pool->bucketSizes) + goto out_err1; + + pool->headers = calloc(numSizes, sizeof(*pool->headers)); + if (!pool->headers) + goto out_err2; + + pool->fMan = fMan; + pool->proposedFlags = flags; + pool->validMask = validMask; + pool->numBuckets = numSizes; + pool->pageSize = getpagesize(); + pool->fd = fd; + pool->pageAlignment = pageAlignment; + pool->maxSlabSize = maxSlabSize; + pool->desiredNumBuffers = desiredNumBuffers; + + for (i=0; inumBuckets; ++i) { + pool->bucketSizes[i] = (smallestSize << i); + driInitSizeHeader(pool, pool->bucketSizes[i], + &pool->headers[i]); + } + + driPool->data = (void *) pool; + driPool->map = &pool_map; + driPool->unmap = &pool_unmap; + driPool->destroy = &pool_destroy; + driPool->offset = &pool_offset; + driPool->poolOffset = &pool_poolOffset; + driPool->flags = &pool_flags; + driPool->size = &pool_size; + driPool->create = &pool_create; + driPool->fence = &pool_fence; + driPool->kernel = &pool_kernel; + driPool->validate = &pool_validate; + driPool->waitIdle = &pool_waitIdle; + driPool->takeDown = &pool_takedown; + + return driPool; + + out_err2: + free(pool->bucketSizes); + out_err1: + free(pool); + out_err0: + free(driPool); + + return NULL; +} diff --git a/src/gallium/winsys/dri/Makefile.template b/src/gallium/winsys/dri/Makefile.template index 3bc1fdd4d4..07abfa53f3 100644 --- a/src/gallium/winsys/dri/Makefile.template +++ b/src/gallium/winsys/dri/Makefile.template @@ -54,6 +54,7 @@ SHARED_INCLUDES = \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/winsys/common \ -I$(TOP)/src/mesa \ -I$(TOP)/src/mesa/main \ -I$(TOP)/src/mesa/glapi \ diff --git a/src/gallium/winsys/dri/intel/Makefile b/src/gallium/winsys/dri/intel/Makefile index c0ce2f927b..00ce21f690 100644 --- a/src/gallium/winsys/dri/intel/Makefile +++ b/src/gallium/winsys/dri/intel/Makefile @@ -4,26 +4,17 @@ include $(TOP)/configs/current LIBNAME = i915_dri.so -MINIGLX_SOURCES = server/intel_dri.c - PIPE_DRIVERS = \ $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ - $(TOP)/src/gallium/drivers/i915simple/libi915simple.a + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a \ + $(TOP)/src/gallium/winsys/common/intel_drm/libinteldrm.a DRIVER_SOURCES = \ - intel_winsys_pipe.c \ intel_winsys_softpipe.c \ - intel_winsys_i915.c \ - intel_batchbuffer.c \ intel_swapbuffers.c \ intel_context.c \ intel_lock.c \ - intel_screen.c \ - ws_dri_bufmgr.c \ - ws_dri_drmpool.c \ - ws_dri_fencemgr.c \ - ws_dri_mallocpool.c \ - ws_dri_slabpool.c + intel_screen.c C_SOURCES = \ $(COMMON_GALLIUM_SOURCES) \ @@ -36,6 +27,6 @@ DRIVER_DEFINES = -I$(TOP)/src/mesa/drivers/dri/intel $(shell pkg-config libdrm - include ../Makefile.template -intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c +#intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c symlinks: diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.c b/src/gallium/winsys/dri/intel/intel_batchbuffer.c deleted file mode 100644 index aa33c12045..0000000000 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.c +++ /dev/null @@ -1,462 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_batchbuffer.h" -#include "intel_context.h" -#include - -#if 0 -static void -intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) -{ - int i; - fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); - for (i = 0; i < count / 4; i += 4) - fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", - offset + i * 4, ptr[i], ptr[i + 1], ptr[i + 2], ptr[i + 3]); - fprintf(stderr, "END BATCH\n\n\n"); -} -#endif - -static void -intel_realloc_relocs(struct intel_batchbuffer *batch, int num_relocs) -{ - unsigned long size = num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER; - - size *= sizeof(uint32_t); - batch->reloc = realloc(batch->reloc, size); - batch->reloc_size = num_relocs; -} - - -void -intel_batchbuffer_reset(struct intel_batchbuffer *batch) -{ - /* - * Get a new, free batchbuffer. - */ - drmBO *bo; - struct drm_bo_info_req *req; - - driBOUnrefUserList(batch->list); - driBOResetList(batch->list); - - /* base.size is the size available to the i915simple driver */ - batch->base.size = batch->intel->intelScreen->max_batch_size - BATCH_RESERVED; - batch->base.actual_size = batch->intel->intelScreen->max_batch_size; - driBOData(batch->buffer, batch->base.actual_size, NULL, NULL, 0); - - /* - * Add the batchbuffer to the validate list. - */ - - driBOAddListItem(batch->list, batch->buffer, - DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_EXE | DRM_BO_MASK_MEM, - &batch->dest_location, &batch->node); - - req = &batch->node->bo_arg.d.req.bo_req; - - /* - * Set up information needed for us to make relocations - * relative to the underlying drm buffer objects. - */ - - driReadLockKernelBO(); - bo = driBOKernel(batch->buffer); - req->presumed_offset = (uint64_t) bo->offset; - req->hint = DRM_BO_HINT_PRESUMED_OFFSET; - batch->drmBOVirtual = (uint8_t *) bo->virtual; - driReadUnlockKernelBO(); - - /* - * Adjust the relocation buffer size. - */ - - if (batch->reloc_size > INTEL_MAX_RELOCS || - batch->reloc == NULL) - intel_realloc_relocs(batch, INTEL_DEFAULT_RELOCS); - - assert(batch->reloc != NULL); - batch->reloc[0] = 0; /* No relocs yet. */ - batch->reloc[1] = 1; /* Reloc type 1 */ - batch->reloc[2] = 0; /* Only a single relocation list. */ - batch->reloc[3] = 0; /* Only a single relocation list. */ - - batch->base.map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); - batch->poolOffset = driBOPoolOffset(batch->buffer); - batch->base.ptr = batch->base.map; - batch->dirty_state = ~0; - batch->nr_relocs = 0; - batch->flags = 0; - batch->id = 0;//batch->intel->intelScreen->batch_id++; -} - -/*====================================================================== - * Public functions - */ -struct intel_batchbuffer * -intel_batchbuffer_alloc(struct intel_context *intel) -{ - struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); - - batch->intel = intel; - - driGenBuffers(intel->intelScreen->batchPool, "batchbuffer", 1, - &batch->buffer, 4096, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); - batch->last_fence = NULL; - batch->list = driBOCreateList(20); - batch->reloc = NULL; - intel_batchbuffer_reset(batch); - return batch; -} - -void -intel_batchbuffer_free(struct intel_batchbuffer *batch) -{ - if (batch->last_fence) { - driFenceFinish(batch->last_fence, - DRM_FENCE_TYPE_EXE, GL_FALSE); - driFenceUnReference(&batch->last_fence); - } - if (batch->base.map) { - driBOUnmap(batch->buffer); - batch->base.map = NULL; - } - driBOUnReference(batch->buffer); - driBOFreeList(batch->list); - if (batch->reloc) - free(batch->reloc); - batch->buffer = NULL; - free(batch); -} - -void -intel_offset_relocation(struct intel_batchbuffer *batch, - unsigned pre_add, - struct _DriBufferObject *driBO, - uint64_t val_flags, - uint64_t val_mask) -{ - int itemLoc; - struct _drmBONode *node; - uint32_t *reloc; - struct drm_bo_info_req *req; - - driBOAddListItem(batch->list, driBO, val_flags, val_mask, - &itemLoc, &node); - req = &node->bo_arg.d.req.bo_req; - - if (!(req->hint & DRM_BO_HINT_PRESUMED_OFFSET)) { - - /* - * Stop other threads from tampering with the underlying - * drmBO while we're reading its offset. - */ - - driReadLockKernelBO(); - req->presumed_offset = (uint64_t) driBOKernel(driBO)->offset; - driReadUnlockKernelBO(); - req->hint = DRM_BO_HINT_PRESUMED_OFFSET; - } - - pre_add += driBOPoolOffset(driBO); - - if (batch->nr_relocs == batch->reloc_size) - intel_realloc_relocs(batch, batch->reloc_size * 2); - - reloc = batch->reloc + - (I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE); - - reloc[0] = ((uint8_t *)batch->base.ptr - batch->drmBOVirtual); - intel_batchbuffer_emit_dword(batch, req->presumed_offset + pre_add); - reloc[1] = pre_add; - reloc[2] = itemLoc; - reloc[3] = batch->dest_location; - batch->nr_relocs++; -} - -static void -i915_drm_copy_reply(const struct drm_bo_info_rep * rep, drmBO * buf) -{ - buf->handle = rep->handle; - buf->flags = rep->flags; - buf->size = rep->size; - buf->offset = rep->offset; - buf->mapHandle = rep->arg_handle; - buf->proposedFlags = rep->proposed_flags; - buf->start = rep->buffer_start; - buf->fenceFlags = rep->fence_flags; - buf->replyFlags = rep->rep_flags; - buf->pageAlignment = rep->page_alignment; -} - -static int -i915_execbuf(struct intel_batchbuffer *batch, - GLuint used, - GLboolean ignore_cliprects, - drmBOList *list, - struct drm_i915_execbuffer *ea) -{ - struct intel_context *intel = batch->intel; - drmBONode *node; - drmMMListHead *l; - struct drm_i915_op_arg *arg, *first; - struct drm_bo_op_req *req; - struct drm_bo_info_rep *rep; - uint64_t *prevNext = NULL; - drmBO *buf; - int ret = 0; - uint32_t count = 0; - - first = NULL; - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - - arg = &node->bo_arg; - req = &arg->d.req; - - if (!first) - first = arg; - - if (prevNext) - *prevNext = (unsigned long)arg; - - prevNext = &arg->next; - req->bo_req.handle = node->buf->handle; - req->op = drm_bo_validate; - req->bo_req.flags = node->arg0; - req->bo_req.mask = node->arg1; - req->bo_req.hint |= 0; - count++; - } - - memset(ea, 0, sizeof(*ea)); - ea->num_buffers = count; - ea->batch.start = batch->poolOffset; - ea->batch.used = used; -#if 0 /* ZZZ JB: no cliprects used */ - ea->batch.cliprects = intel->pClipRects; - ea->batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; - ea->batch.DR1 = 0; - ea->batch.DR4 = 0;((((GLuint) intel->drawX) & 0xffff) | - (((GLuint) intel->drawY) << 16)); -#else - ea->batch.cliprects = NULL; - ea->batch.num_cliprects = 0; - ea->batch.DR1 = 0; - ea->batch.DR4 = 0; -#endif - ea->fence_arg.flags = DRM_I915_FENCE_FLAG_FLUSHED; - ea->ops_list = (unsigned long) first; - first->reloc_ptr = (unsigned long) batch->reloc; - batch->reloc[0] = batch->nr_relocs; - - //return -EFAULT; - do { - ret = drmCommandWriteRead(intel->driFd, DRM_I915_EXECBUFFER, ea, - sizeof(*ea)); - } while (ret == -EAGAIN); - - if (ret != 0) - return ret; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - arg = &node->bo_arg; - rep = &arg->d.rep.bo_info; - - if (!arg->handled) { - return -EFAULT; - } - if (arg->d.rep.ret) - return arg->d.rep.ret; - - buf = node->buf; - i915_drm_copy_reply(rep, buf); - } - return 0; -} - -/* TODO: Push this whole function into bufmgr. - */ -static struct _DriFenceObject * -do_flush_locked(struct intel_batchbuffer *batch, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) -{ - struct intel_context *intel = batch->intel; - struct _DriFenceObject *fo; - drmFence fence; - drmBOList *boList; - struct drm_i915_execbuffer ea; - int ret = 0; - - driBOValidateUserList(batch->list); - boList = driGetdrmBOList(batch->list); - -#if 0 /* ZZZ JB Allways run */ - if (!(intel->numClipRects == 0 && !ignore_cliprects)) { -#else - if (1) { -#endif - ret = i915_execbuf(batch, used, ignore_cliprects, boList, &ea); - } else { - driPutdrmBOList(batch->list); - fo = NULL; - goto out; - } - driPutdrmBOList(batch->list); - if (ret) - abort(); - - if (ea.fence_arg.error != 0) { - - /* - * The hardware has been idled by the kernel. - * Don't fence the driBOs. - */ - - if (batch->last_fence) - driFenceUnReference(&batch->last_fence); -#if 0 /* ZZZ JB: no _mesa_* funcs in gallium */ - _mesa_printf("fence error\n"); -#endif - batch->last_fence = NULL; - fo = NULL; - goto out; - } - - fence.handle = ea.fence_arg.handle; - fence.fence_class = ea.fence_arg.fence_class; - fence.type = ea.fence_arg.type; - fence.flags = ea.fence_arg.flags; - fence.signaled = ea.fence_arg.signaled; - - fo = driBOFenceUserList(batch->intel->intelScreen->mgr, batch->list, - "SuperFence", &fence); - - if (driFenceType(fo) & DRM_I915_FENCE_TYPE_RW) { - if (batch->last_fence) - driFenceUnReference(&batch->last_fence); - /* - * FIXME: Context last fence?? - */ - batch->last_fence = fo; - driFenceReference(fo); - } - out: -#if 0 /* ZZZ JB: fix this */ - intel->vtbl.lost_hardware(intel); -#else - (void)intel; -#endif - return fo; -} - - -struct _DriFenceObject * -intel_batchbuffer_flush(struct intel_batchbuffer *batch) -{ - struct intel_context *intel = batch->intel; - GLuint used = batch->base.ptr - batch->base.map; - GLboolean was_locked = intel->locked; - struct _DriFenceObject *fence; - - if (used == 0) { - driFenceReference(batch->last_fence); - return batch->last_fence; - } - - /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a - * performance drain that we would like to avoid. - */ -#if 0 /* ZZZ JB: what should we do here? */ - if (used & 4) { - ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->base.ptr)[1] = 0; - ((int *) batch->base.ptr)[2] = MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->base.ptr)[1] = MI_BATCH_BUFFER_END; - used += 8; - } -#else - if (used & 4) { - ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->base.ptr)[1] = 0; - ((int *) batch->base.ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->base.ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; - used += 8; - } -#endif - driBOUnmap(batch->buffer); - batch->base.ptr = NULL; - batch->base.map = NULL; - - /* TODO: Just pass the relocation list and dma buffer up to the - * kernel. - */ - if (!was_locked) - LOCK_HARDWARE(intel); - - fence = do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), - GL_FALSE); - - if (!was_locked) - UNLOCK_HARDWARE(intel); - - /* Reset the buffer: - */ - intel_batchbuffer_reset(batch); - return fence; -} - -void -intel_batchbuffer_finish(struct intel_batchbuffer *batch) -{ - struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); - driFenceFinish(fence, driFenceType(fence), GL_FALSE); - driFenceUnReference(&fence); -} - -void -intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags) -{ - assert((bytes & 3) == 0); - intel_batchbuffer_require_space(batch, bytes, flags); - memcpy(batch->base.ptr, data, bytes); - batch->base.ptr += bytes; -} diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.h b/src/gallium/winsys/dri/intel/intel_batchbuffer.h index abb7a624f5..1fa2719845 100644 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.h +++ b/src/gallium/winsys/dri/intel/intel_batchbuffer.h @@ -1,137 +1,24 @@ #ifndef INTEL_BATCHBUFFER_H #define INTEL_BATCHBUFFER_H -#include "mtypes.h" -#include "ws_dri_bufmgr.h" -#include "i915simple/i915_batch.h" +#include "intel_drm/intel_be_batchbuffer.h" -struct intel_context; - -#define BATCH_SZ 16384 -#define BATCH_RESERVED 16 - -#define INTEL_DEFAULT_RELOCS 100 -#define INTEL_MAX_RELOCS 400 - -#define INTEL_BATCH_NO_CLIPRECTS 0x1 -#define INTEL_BATCH_CLIPRECTS 0x2 - -struct intel_batchbuffer -{ - struct i915_batchbuffer base; - - struct intel_context *intel; - - struct _DriBufferObject *buffer; - struct _DriFenceObject *last_fence; - GLuint flags; - - struct _DriBufferList *list; - GLuint list_count; - - uint32_t *reloc; - GLuint reloc_size; - GLuint nr_relocs; - - GLuint dirty_state; - GLuint id; - - uint32_t poolOffset; - uint8_t *drmBOVirtual; - struct _drmBONode *node; /* Validation list node for this buffer */ - int dest_location; /* Validation list sequence for this buffer */ -}; - -struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_context - *intel); - -void intel_batchbuffer_free(struct intel_batchbuffer *batch); - - -void intel_batchbuffer_finish(struct intel_batchbuffer *batch); - -struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer - *batch); - -void intel_batchbuffer_reset(struct intel_batchbuffer *batch); - - -/* Unlike bmBufferData, this currently requires the buffer be mapped. - * Consider it a convenience function wrapping multple - * intel_buffer_dword() calls. - */ -void intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags); - -void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, - GLuint bytes); - -void -intel_offset_relocation(struct intel_batchbuffer *batch, - unsigned pre_add, - struct _DriBufferObject *driBO, - uint64_t val_flags, - uint64_t val_mask); - -/* Inline functions - might actually be better off with these - * non-inlined. Certainly better off switching all command packets to - * be passed as structs rather than dwords, but that's a little bit of - * work... - */ -static INLINE GLuint -intel_batchbuffer_space(struct intel_batchbuffer *batch) -{ - return (batch->base.size - BATCH_RESERVED) - (batch->base.ptr - batch->base.map); -} - - -static INLINE void -intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) -{ - assert(batch->base.map); - assert(intel_batchbuffer_space(batch) >= 4); - *(GLuint *) (batch->base.ptr) = dword; - batch->base.ptr += 4; -} - -static INLINE void -intel_batchbuffer_require_space(struct intel_batchbuffer *batch, - GLuint sz, GLuint flags) -{ - struct _DriFenceObject *fence; - - assert(sz < batch->base.size - 8); - if (intel_batchbuffer_space(batch) < sz || - (batch->flags != 0 && flags != 0 && batch->flags != flags)) { - fence = intel_batchbuffer_flush(batch); - driFenceUnReference(&fence); - } - - batch->flags |= flags; -} - -/* Here are the crusty old macros, to be removed: +/* + * Need to redefine the BATCH defines */ -#undef BATCH_LOCALS -#define BATCH_LOCALS #undef BEGIN_BATCH -#define BEGIN_BATCH(n, flags) do { \ - assert(!intel->prim.flush); \ - intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ -} while (0) +#define BEGIN_BATCH(dwords, relocs) \ + (i915_batchbuffer_check(&intel->base.batch->base, dwords, relocs)) #undef OUT_BATCH -#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) +#define OUT_BATCH(d) \ + i915_batchbuffer_dword(&intel->base.batch->base, d) #undef OUT_RELOC -#define OUT_RELOC(buf,flags,mask,delta) do { \ +#define OUT_RELOC(buf,flags,mask,delta) do { \ assert((delta) >= 0); \ - intel_offset_relocation(intel->batch, delta, buf, flags, mask); \ + intel_be_offset_relocation(intel->base.batch, delta, buf, flags, mask); \ } while (0) -#undef ADVANCE_BATCH -#define ADVANCE_BATCH() do { } while(0) - - #endif diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index 6a0c381849..8284e0edbb 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ @@ -31,8 +31,10 @@ #include "intel_screen.h" #include "intel_context.h" #include "intel_swapbuffers.h" -#include "intel_winsys.h" #include "intel_batchbuffer.h" +#include "intel_winsys_softpipe.h" + +#include "i915simple/i915_screen.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" @@ -132,6 +134,46 @@ static const struct dri_debug_control debug_control[] = { +/** + * Create i915 hardware rendering context. + */ +static struct pipe_context * +intel_create_i915simple(struct intel_context *intel, + struct pipe_winsys *winsys) +{ + struct pipe_screen *screen; + + /* Fill in this struct with callbacks that i915simple will need to + * communicate with the window system, buffer manager, etc. + */ + screen = i915_create_screen(winsys, intel->intelScreen->deviceID); + + /* Create the i915simple context: + */ + return i915_create_context(screen, winsys, &intel->base.base ); +} + +static void +intel_lock_hardware(struct intel_be_context *context) +{ + struct intel_context *intel = (struct intel_context *)context; + LOCK_HARDWARE(intel); +} + +static void +intel_unlock_hardware(struct intel_be_context *context) +{ + struct intel_context *intel = (struct intel_context *)context; + UNLOCK_HARDWARE(intel); +} + +static boolean +intel_locked_hardware(struct intel_be_context *context) +{ + struct intel_context *intel = (struct intel_context *)context; + return intel->locked ? TRUE : FALSE; +} + GLboolean intelCreateContext(const __GLcontextModes * visual, __DRIcontextPrivate * driContextPriv, @@ -179,20 +221,24 @@ intelCreateContext(const __GLcontextModes * visual, intel->iw.irq_seq = -1; intel->irqsEmitted = 0; - intel->batch = intel_batchbuffer_alloc(intel); intel->last_swap_fence = NULL; intel->first_swap_fence = NULL; #ifdef DEBUG __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); #endif + intel->base.hardware_lock = intel_lock_hardware; + intel->base.hardware_unlock = intel_unlock_hardware; + intel->base.hardware_locked = intel_locked_hardware; + + intel_be_init_context(&intel->base, &intelScreen->base); /* * Pipe-related setup */ if (getenv("INTEL_SP")) { /* use softpipe driver instead of hw */ - pipe = intel_create_softpipe( intel, intelScreen->winsys ); + pipe = intel_create_softpipe( intel, &intelScreen->base.base ); } else { switch (intel->intelScreen->deviceID) { @@ -204,13 +250,13 @@ intelCreateContext(const __GLcontextModes * visual, case PCI_CHIP_Q35_G: case PCI_CHIP_I915_G: case PCI_CHIP_I915_GM: - pipe = intel_create_i915simple( intel, intelScreen->winsys ); + pipe = intel_create_i915simple( intel, &intelScreen->base.base ); break; default: - fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", + fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", intel->intelScreen->deviceID, __FUNCTION__); - pipe = intel_create_softpipe( intel, intelScreen->winsys ); + pipe = intel_create_softpipe( intel, &intelScreen->base.base ); break; } } @@ -234,8 +280,6 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) if (intel) { st_finish(intel->st); - intel_batchbuffer_free(intel->batch); - if (intel->last_swap_fence) { driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); driFenceUnReference(&intel->last_swap_fence); @@ -251,6 +295,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv) intel->intelScreen->dummyContext = NULL; st_destroy_context(intel->st); + intel_be_destroy_context(&intel->base); free(intel); } } diff --git a/src/gallium/winsys/dri/intel/intel_context.h b/src/gallium/winsys/dri/intel/intel_context.h index 597dc13830..ced18da143 100644 --- a/src/gallium/winsys/dri/intel/intel_context.h +++ b/src/gallium/winsys/dri/intel/intel_context.h @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ #ifndef INTEL_CONTEXT_H @@ -36,6 +36,8 @@ #include "intel_screen.h" #include "i915_drm.h" +#include "intel_drm/intel_be_context.h" + struct pipe_context; struct intel_context; @@ -50,12 +52,13 @@ struct st_context; */ struct intel_context { + struct intel_be_context base; struct st_context *st; struct _DriFenceObject *last_swap_fence; struct _DriFenceObject *first_swap_fence; - struct intel_batchbuffer *batch; +// struct intel_batchbuffer *batch; boolean locked; char *prevLockFile; @@ -123,7 +126,7 @@ extern int __intel_debug; } while(0) #else -#define DBG(flag, ...) +#define DBG(flag, ...) #endif diff --git a/src/gallium/winsys/dri/intel/intel_lock.c b/src/gallium/winsys/dri/intel/intel_lock.c index 70aa7ea5f4..406284c98f 100644 --- a/src/gallium/winsys/dri/intel/intel_lock.c +++ b/src/gallium/winsys/dri/intel/intel_lock.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ @@ -87,7 +87,7 @@ void LOCK_HARDWARE( struct intel_context *intel ) } -/* Unlock the hardware using the global current context +/* Unlock the hardware using the global current context */ void UNLOCK_HARDWARE( struct intel_context *intel ) { @@ -99,4 +99,4 @@ void UNLOCK_HARDWARE( struct intel_context *intel ) _glthread_UNLOCK_MUTEX(lockMutex); DBG(LOCK, "%s - unlocked\n", __progname); -} +} diff --git a/src/gallium/winsys/dri/intel/intel_reg.h b/src/gallium/winsys/dri/intel/intel_reg.h index f37c24fda9..4f33bee438 100644 --- a/src/gallium/winsys/dri/intel/intel_reg.h +++ b/src/gallium/winsys/dri/intel/intel_reg.h @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 0be88974f4..8817a80a5a 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ #include "utils.h" @@ -32,12 +32,10 @@ #include "intel_context.h" #include "intel_screen.h" #include "intel_batchbuffer.h" -//#include "intel_batchpool.h" #include "intel_swapbuffers.h" -#include "intel_winsys.h" #include "i830_dri.h" -#include "ws_dri_bufpool.h" +#include "intel_drm/ws_dri_bufpool.h" #include "pipe/p_context.h" #include "state_tracker/st_public.h" @@ -152,16 +150,16 @@ intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - driBOSetStatic(intelScreen->front.buffer, - intelScreen->front.offset, - intelScreen->front.pitch * intelScreen->front.height, + driBOSetStatic(intelScreen->front.buffer, + intelScreen->front.offset, + intelScreen->front.pitch * intelScreen->front.height, intelScreen->front.map, 0); } #else - if (intelScreen->staticPool) { + if (intelScreen->base.staticPool) { if (intelScreen->front.buffer) driBOUnReference(intelScreen->front.buffer); - driGenBuffers(intelScreen->staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); + driGenBuffers(intelScreen->base.staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); driBOSetReferenced(intelScreen->front.buffer, sarea->front_bo_handle); } #endif @@ -177,33 +175,6 @@ intelCreatePools(__DRIscreenPrivate * sPriv) if (intelScreen->havePools) return GL_TRUE; -#if 0 /* ZZZ JB fix this */ - intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); - if (!intelScreen->staticPool) - return GL_FALSE; - - batchPoolSize /= BATCH_SZ; - intelScreen->batchPool = driBatchPoolInit(sPriv->fd, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT | - DRM_BO_FLAG_MEM_LOCAL, - BATCH_SZ, - batchPoolSize, 5); - if (!intelScreen->batchPool) { - fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); - return GL_FALSE; - } -#else - intelScreen->staticPool = driDRMPoolInit(sPriv->fd); - intelScreen->batchPool = driSlabPoolInit(sPriv->fd, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - intelScreen->max_batch_size, - 1, 40, intelScreen->max_batch_size * 16, 0, - intelScreen->fMan); -#endif intelScreen->havePools = GL_TRUE; intelUpdateScreenRotation(sPriv, intelScreen->sarea); @@ -211,6 +182,27 @@ intelCreatePools(__DRIscreenPrivate * sPriv) return GL_TRUE; } +static const char * +intel_get_name( struct pipe_winsys *winsys ) +{ + return "Intel/DRI/ttm"; +} + +/* + * The state tracker (should!) keep track of whether the fake + * frontbuffer has been touched by any rendering since the last time + * we copied its contents to the real frontbuffer. Our task is easy: + */ +static void +intel_flush_frontbuffer( struct pipe_winsys *winsys, + struct pipe_surface *surf, + void *context_private) +{ + struct intel_context *intel = (struct intel_context *) context_private; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelDisplaySurface(dPriv, surf, NULL); +} static boolean intelInitDriver(__DRIscreenPrivate * sPriv) @@ -231,7 +223,7 @@ intelInitDriver(__DRIscreenPrivate * sPriv) /* Allocate the private area */ intelScreen = CALLOC_STRUCT(intel_screen); - if (!intelScreen) + if (!intelScreen) return GL_FALSE; /* parse information in __driConfigOptions */ @@ -262,26 +254,9 @@ intelInitDriver(__DRIscreenPrivate * sPriv) (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); } - intelScreen->max_batch_size = 16 * 4096; - -#if 1 // ZZZ JB - intelScreen->mgr = driFenceMgrTTMInit(sPriv->fd); - if (!intelScreen->mgr) { - fprintf(stderr, "Failed to create fence manager.\n"); - return GL_FALSE; - } - - intelScreen->fMan = driInitFreeSlabManager(10, 10); - if (!intelScreen->fMan) { - fprintf(stderr, "Failed to create free slab manager.\n"); - return GL_FALSE; - } - - if (!intelCreatePools(sPriv)) - return GL_FALSE; -#endif - - intelScreen->winsys = intel_create_pipe_winsys(sPriv->fd, intelScreen->fMan); + intel_be_init_device(&intelScreen->base, sPriv->fd); + intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; + intelScreen->base.base.get_name = intel_get_name; return GL_TRUE; } @@ -292,12 +267,9 @@ intelDestroyScreen(__DRIscreenPrivate * sPriv) { struct intel_screen *intelScreen = intel_screen(sPriv); + intel_be_destroy_device(&intelScreen->base); /* intelUnmapScreenRegions(intelScreen); */ - if (intelScreen->havePools) { - driPoolTakeDown(intelScreen->staticPool); - driPoolTakeDown(intelScreen->batchPool); - } FREE(intelScreen); sPriv->private = NULL; } @@ -518,8 +490,8 @@ intelFillInModes(unsigned pixel_bits, unsigned depth_bits, * This routine also fills in the linked list pointed to by \c driver_modes * with the \c __GLcontextModes that the driver can support for windows or * pbuffers. - * - * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on + * + * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on * failure. */ PUBLIC void * diff --git a/src/gallium/winsys/dri/intel/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h index 1db0502b13..8036917903 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.h +++ b/src/gallium/winsys/dri/intel/intel_screen.h @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ #ifndef _INTEL_SCREEN_H_ @@ -31,13 +31,16 @@ #include "dri_util.h" #include "i830_common.h" #include "xmlconfig.h" -#include "ws_dri_bufpool.h" +#include "intel_drm/ws_dri_bufpool.h" #include "pipe/p_compiler.h" +#include "intel_drm/intel_be_device.h" struct intel_screen { + struct intel_be_device base; + struct { drm_handle_t handle; @@ -51,7 +54,7 @@ struct intel_screen int width; int height; int size; - int cpp; /* for front and back buffers */ + int cpp; /* for front and back buffers */ } front; int deviceID; @@ -64,8 +67,6 @@ struct intel_screen */ driOptionCache optionCache; - struct _DriBufferPool *batchPool; - struct _DriBufferPool *staticPool; /** for the X screen/framebuffer */ boolean havePools; /** @@ -74,13 +75,11 @@ struct intel_screen */ struct intel_context *dummyContext; - /* + /* * New stuff form the i915tex integration */ - struct _DriFenceMgr *mgr; - struct _DriFreeSlabManager *fMan; unsigned batch_id; - unsigned max_batch_size; + struct pipe_winsys *winsys; }; diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index 923b542771..7f3babd98e 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,21 +22,22 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ #include "intel_screen.h" #include "intel_context.h" #include "intel_swapbuffers.h" -#include "intel_batchbuffer.h" + #include "intel_reg.h" -#include "intel_winsys.h" #include "pipe/p_context.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" #include "state_tracker/st_cb_fbo.h" +#include "intel_drm/ws_dri_bufmgr.h" +#include "intel_batchbuffer.h" /** * Display a colorbuffer surface in an X window. @@ -114,7 +115,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, if (pbox->x1 > pbox->x2 || pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || + pbox->x2 > intelScreen->front.width || pbox->y2 > intelScreen->front.height) { /* invalid cliprect, skip it */ continue; @@ -159,13 +160,22 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, assert(box.y1 < box.y2); /* XXX this could be done with pipe->surface_copy() */ - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + /* XXX should have its own batch buffer */ + if (!BEGIN_BATCH(8, 2)) { + /* + * Since we share this batch buffer with a context + * we can't flush it since that risks a GPU lockup + */ + assert(0); + continue; + } + OUT_BATCH(CMD); OUT_BATCH(BR13); OUT_BATCH((box.y1 << 16) | box.x1); OUT_BATCH((box.y2 << 16) | box.x2); - OUT_RELOC(intelScreen->front.buffer, + OUT_RELOC(intelScreen->front.buffer, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); OUT_BATCH((sbox.y1 << 16) | sbox.x1); @@ -174,12 +184,11 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - ADVANCE_BATCH(); } if (intel->first_swap_fence) driFenceUnReference(&intel->first_swap_fence); - intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); + intel->first_swap_fence = intel_be_batchbuffer_flush(intel->base.batch); } UNLOCK_HARDWARE(intel); diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.h b/src/gallium/winsys/dri/intel/intel_swapbuffers.h index 7ae5fd15a5..46c9bab3af 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.h +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.h @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ #ifndef INTEL_SWAPBUFFERS_H diff --git a/src/gallium/winsys/dri/intel/intel_winsys.h b/src/gallium/winsys/dri/intel/intel_winsys.h deleted file mode 100644 index 3d32db10a4..0000000000 --- a/src/gallium/winsys/dri/intel/intel_winsys.h +++ /dev/null @@ -1,73 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_WINSYS_H -#define INTEL_WINSYS_H - -#include "pipe/p_state.h" - -struct intel_context; -struct pipe_context; -struct pipe_winsys; -struct pipe_buffer; -struct _DriBufferObject; - -struct pipe_winsys * -intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ); - -void -intel_destroy_pipe_winsys( struct pipe_winsys *winsys ); - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel, - struct pipe_winsys *winsys ); - -struct pipe_context * -intel_create_i915simple( struct intel_context *intel, - struct pipe_winsys *winsys ); - - -struct intel_buffer { - struct pipe_buffer base; - struct _DriBufferPool *pool; - struct _DriBufferObject *driBO; -}; - -static INLINE struct intel_buffer * -intel_buffer( struct pipe_buffer *buf ) -{ - return (struct intel_buffer *)buf; -} - -static INLINE struct _DriBufferObject * -dri_bo( struct pipe_buffer *buf ) -{ - return intel_buffer(buf)->driBO; -} - - -#endif diff --git a/src/gallium/winsys/dri/intel/intel_winsys_i915.c b/src/gallium/winsys/dri/intel/intel_winsys_i915.c deleted file mode 100644 index 013291364a..0000000000 --- a/src/gallium/winsys/dri/intel/intel_winsys_i915.c +++ /dev/null @@ -1,156 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" - -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_winsys.h" - -#include "pipe/p_util.h" -#include "pipe/p_winsys.h" -#include "i915simple/i915_winsys.h" -#include "i915simple/i915_screen.h" - - -struct intel_i915_winsys { - struct i915_winsys winsys; /**< batch buffer funcs */ - struct pipe_winsys *pws; - struct intel_context *intel; -}; - - -/* Turn a i915simple winsys into an intel/i915simple winsys: - */ -static inline struct intel_i915_winsys * -intel_i915_winsys( struct i915_winsys *sws ) -{ - return (struct intel_i915_winsys *)sws; -} - - -/* Simple batchbuffer interface: - */ - -static struct i915_batchbuffer* -intel_i915_batch_get( struct i915_winsys *sws ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - return &intel->batch->base; -} - -static void intel_i915_batch_reloc( struct i915_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - unsigned flags = DRM_BO_FLAG_MEM_TT; - unsigned mask = DRM_BO_MASK_MEM; - - if (access_flags & I915_BUFFER_ACCESS_WRITE) { - flags |= DRM_BO_FLAG_WRITE; - mask |= DRM_BO_FLAG_WRITE; - } - - if (access_flags & I915_BUFFER_ACCESS_READ) { - flags |= DRM_BO_FLAG_READ; - mask |= DRM_BO_FLAG_READ; - } - - intel_offset_relocation( intel->batch, - delta, - dri_bo( buf ), - flags, - mask ); -} - -static void intel_i915_batch_flush( struct i915_winsys *sws, - struct pipe_fence_handle **fence ) -{ - struct intel_i915_winsys *iws = intel_i915_winsys(sws); - struct intel_context *intel = iws->intel; - union { - struct _DriFenceObject *dri; - struct pipe_fence_handle *pipe; - } fu; - - if (fence) - assert(!*fence); - - fu.dri = intel_batchbuffer_flush( intel->batch ); - - if (!fu.dri) { - assert(0); - *fence = NULL; - return; - } - - if (fu.dri) { - if (fence) - *fence = fu.pipe; - else - driFenceUnReference(&fu.dri); - } - -} - - -/** - * Create i915 hardware rendering context. - */ -struct pipe_context * -intel_create_i915simple( struct intel_context *intel, - struct pipe_winsys *winsys ) -{ - struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); - struct pipe_screen *screen; - - /* Fill in this struct with callbacks that i915simple will need to - * communicate with the window system, buffer manager, etc. - */ - iws->winsys.batch_get = intel_i915_batch_get; - iws->winsys.batch_reloc = intel_i915_batch_reloc; - iws->winsys.batch_flush = intel_i915_batch_flush; - iws->pws = winsys; - iws->intel = intel; - - screen = i915_create_screen(winsys, intel->intelScreen->deviceID); - - /* Create the i915simple context: - */ - return i915_create_context( screen, - winsys, - &iws->winsys ); -} diff --git a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c b/src/gallium/winsys/dri/intel/intel_winsys_pipe.c deleted file mode 100644 index 51a79ca535..0000000000 --- a/src/gallium/winsys/dri/intel/intel_winsys_pipe.c +++ /dev/null @@ -1,318 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include - -#include "intel_context.h" -#include "intel_winsys.h" -#include "intel_swapbuffers.h" -#include "intel_batchbuffer.h" - -#include "pipe/p_winsys.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" - -struct intel_pipe_winsys { - struct pipe_winsys winsys; - struct _DriBufferPool *regionPool; - struct _DriBufferPool *mallocPool; - struct _DriBufferPool *vertexPool; - struct _DriFreeSlabManager *fMan; /** shared between all pipes */ -}; - - -/* Turn a pipe winsys into an intel/pipe winsys: - */ -static inline struct intel_pipe_winsys * -intel_pipe_winsys( struct pipe_winsys *winsys ) -{ - return (struct intel_pipe_winsys *)winsys; -} - - -/* - * Buffer functions. - * - * Most callbacks map direcly onto dri_bufmgr operations: - */ - -static void *intel_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned flags ) -{ - unsigned drm_flags = 0; - - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - drm_flags |= DRM_BO_FLAG_WRITE; - - if (flags & PIPE_BUFFER_USAGE_CPU_READ) - drm_flags |= DRM_BO_FLAG_READ; - - return driBOMap( dri_bo(buf), drm_flags, 0 ); -} - -static void intel_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnmap( dri_bo(buf) ); -} - -static void -intel_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnReference( dri_bo(buf) ); - FREE(buf); -} - -static struct pipe_buffer * -intel_buffer_create(struct pipe_winsys *winsys, - unsigned alignment, - unsigned usage, - unsigned size ) -{ - struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); - struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); - unsigned flags = 0; - struct _DriBufferPool *pool; - - buffer->base.refcount = 1; - buffer->base.alignment = alignment; - buffer->base.usage = usage; - buffer->base.size = size; - - if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { - flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; - pool = iws->mallocPool; - } else if (usage & PIPE_BUFFER_USAGE_CUSTOM) { - /* For vertex buffers */ - flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; - pool = iws->vertexPool; - } else { - flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; - pool = iws->regionPool; - } - - if (usage & PIPE_BUFFER_USAGE_GPU_READ) - flags |= DRM_BO_FLAG_READ; - - if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) - flags |= DRM_BO_FLAG_WRITE; - - /* drm complains if we don't set any read/write flags. - */ - if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) - flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; - -#if 0 - if (flags & IWS_BUFFER_USAGE_EXE) - flags |= DRM_BO_FLAG_EXE; - - if (usage & IWS_BUFFER_USAGE_CACHED) - flags |= DRM_BO_FLAG_CACHED; -#endif - - buffer->pool = pool; - driGenBuffers( buffer->pool, - "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); - - driBOData( buffer->driBO, size, NULL, buffer->pool, 0 ); - - return &buffer->base; -} - - -static struct pipe_buffer * -intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) -{ - struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); - struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); - - driGenUserBuffer( iws->regionPool, - "pipe user buffer", &buffer->driBO, ptr, bytes ); - - buffer->base.refcount = 1; - - return &buffer->base; -} - - -/* - * Surface functions. - * - * Deprecated! - */ - -static struct pipe_surface * -intel_i915_surface_alloc(struct pipe_winsys *winsys) -{ - assert("intel_i915_surface_alloc is deprecated" & 0); - return NULL; -} - -static int -intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, - struct pipe_surface *surf, - unsigned width, unsigned height, - enum pipe_format format, - unsigned flags, - unsigned tex_usage) -{ - assert("intel_i915_surface_alloc_storage is deprecated" & 0); - return -1; -} - -static void -intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) -{ - assert("intel_i915_surface_release is deprecated" & 0); -} - -/* - * Fence functions - */ - -static void -intel_fence_reference( struct pipe_winsys *sws, - struct pipe_fence_handle **ptr, - struct pipe_fence_handle *fence ) -{ - if (*ptr) - driFenceUnReference((struct _DriFenceObject **)ptr); - - if (fence) - *ptr = (struct pipe_fence_handle *)driFenceReference((struct _DriFenceObject *)fence); -} - -static int -intel_fence_signalled( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ) -{ - return driFenceSignaled((struct _DriFenceObject *)fence, flag); -} - -static int -intel_fence_finish( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ) -{ - return driFenceFinish((struct _DriFenceObject *)fence, flag, 0); -} - - -/* - * Mixed functions - */ - -static const char * -intel_get_name( struct pipe_winsys *winsys ) -{ - return "Intel/DRI/ttm"; -} - -/* - * The state tracker (should!) keep track of whether the fake - * frontbuffer has been touched by any rendering since the last time - * we copied its contents to the real frontbuffer. Our task is easy: - */ -static void -intel_flush_frontbuffer( struct pipe_winsys *winsys, - struct pipe_surface *surf, - void *context_private) -{ - struct intel_context *intel = (struct intel_context *) context_private; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelDisplaySurface(dPriv, surf, NULL); -} - -struct pipe_winsys * -intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) -{ - struct intel_pipe_winsys *iws = CALLOC_STRUCT( intel_pipe_winsys ); - - /* Fill in this struct with callbacks that pipe will need to - * communicate with the window system, buffer manager, etc. - * - * Pipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate _DriBufferObject. - */ - iws->winsys.buffer_create = intel_buffer_create; - iws->winsys.user_buffer_create = intel_user_buffer_create; - iws->winsys.buffer_map = intel_buffer_map; - iws->winsys.buffer_unmap = intel_buffer_unmap; - iws->winsys.buffer_destroy = intel_buffer_destroy; - iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; - iws->winsys.get_name = intel_get_name; - iws->winsys.surface_alloc = intel_i915_surface_alloc; - iws->winsys.surface_alloc_storage = intel_i915_surface_alloc_storage; - iws->winsys.surface_release = intel_i915_surface_release; - - iws->winsys.fence_reference = intel_fence_reference; - iws->winsys.fence_signalled = intel_fence_signalled; - iws->winsys.fence_finish = intel_fence_finish; - - if (fd) { - iws->regionPool = driDRMPoolInit(fd); - iws->vertexPool = driSlabPoolInit(fd, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - 128 * 4096, - 1, 120, 128 * 4096 * 4, 0, - fMan); - } - - iws->mallocPool = driMallocPoolInit(); - - return &iws->winsys; -} - -void -intel_destroy_pipe_winsys( struct pipe_winsys *winsys ) -{ - struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); - if (iws->regionPool) { - driPoolTakeDown(iws->regionPool); - } - if (iws->mallocPool) { - driPoolTakeDown(iws->mallocPool); - } - free(iws); -} diff --git a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c index 0bc2dc4002..0d98d16cf1 100644 --- a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c +++ b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,27 +10,27 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * - * + * + * **************************************************************************/ /* * Authors: Keith Whitwell */ #include "intel_context.h" -#include "intel_winsys.h" +#include "intel_winsys_softpipe.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "pipe/p_format.h" @@ -69,9 +69,9 @@ intel_create_softpipe( struct intel_context *intel, { struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); struct pipe_screen *screen = softpipe_create_screen(winsys); - + /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. + * communicate with the window system, buffer manager, etc. */ isws->sws.is_format_supported = intel_is_format_supported; isws->intel = intel; diff --git a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.h b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.h new file mode 100644 index 0000000000..5fa14cb749 --- /dev/null +++ b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.h @@ -0,0 +1,39 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_SOFTPIPE_H +#define INTEL_SOFTPIPE_H + +struct pipe_winsys; +struct pipe_context; +struct intel_context; + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel, + struct pipe_winsys *winsys ); + +#endif diff --git a/src/gallium/winsys/dri/intel/server/i830_common.h b/src/gallium/winsys/dri/intel/server/i830_common.h index f84f453309..3452ddb3c9 100644 --- a/src/gallium/winsys/dri/intel/server/i830_common.h +++ b/src/gallium/winsys/dri/intel/server/i830_common.h @@ -88,9 +88,9 @@ typedef struct { /** Last context that used the buffer manager. */ int texAge; int pf_enabled; /* is pageflipping allowed? */ - int pf_active; + int pf_active; int pf_current_page; /* which buffer is being displayed? */ - int perf_boxes; /* performance boxes to be displayed */ + int perf_boxes; /* performance boxes to be displayed */ int width, height; /* screen size in pixels */ drm_handle_t front_handle; @@ -173,7 +173,7 @@ typedef struct { int num_cliprects; /* mulitpass with multiple cliprects? */ drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ } drmI830CmdBuffer; - + typedef struct { int *irq_seq; } drmI830IrqEmit; @@ -188,7 +188,7 @@ typedef struct { } drmI830GetParam; #define I830_PARAM_IRQ_ACTIVE 1 -#define I830_PARAM_ALLOW_BATCHBUFFER 2 +#define I830_PARAM_ALLOW_BATCHBUFFER 2 typedef struct { int param; diff --git a/src/gallium/winsys/dri/intel/server/i830_dri.h b/src/gallium/winsys/dri/intel/server/i830_dri.h index 685de4a551..0d514b6c38 100644 --- a/src/gallium/winsys/dri/intel/server/i830_dri.h +++ b/src/gallium/winsys/dri/intel/server/i830_dri.h @@ -40,7 +40,7 @@ typedef struct _I830DRIRec { int bitsPerPixel; int unused11[8]; /* was front/back/depth/rotated offset/pitch */ - + int unused12; /* logTextureGranularity */ int unused13; /* textureOffset */ diff --git a/src/gallium/winsys/dri/intel/server/intel.h b/src/gallium/winsys/dri/intel/server/intel.h deleted file mode 100644 index 6ea72499c1..0000000000 --- a/src/gallium/winsys/dri/intel/server/intel.h +++ /dev/null @@ -1,331 +0,0 @@ -#ifndef _INTEL_H_ -#define _INTEL_H_ - -#include "xf86drm.h" /* drm_handle_t, etc */ - -/* Intel */ -#ifndef PCI_CHIP_I810 -#define PCI_CHIP_I810 0x7121 -#define PCI_CHIP_I810_DC100 0x7123 -#define PCI_CHIP_I810_E 0x7125 -#define PCI_CHIP_I815 0x1132 -#define PCI_CHIP_I810_BRIDGE 0x7120 -#define PCI_CHIP_I810_DC100_BRIDGE 0x7122 -#define PCI_CHIP_I810_E_BRIDGE 0x7124 -#define PCI_CHIP_I815_BRIDGE 0x1130 -#endif - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 - -#ifndef PCI_CHIP_I855_GM -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I855_GM_BRIDGE 0x3580 -#endif - -#ifndef PCI_CHIP_I865_G -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I865_G_BRIDGE 0x2570 -#endif - -#ifndef PCI_CHIP_I915_G -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I915_GM -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I915_GM_BRIDGE 0x2590 -#endif - -#ifndef PCI_CHIP_E7221_G -#define PCI_CHIP_E7221_G 0x258A -/* Same as I915_G_BRIDGE */ -#define PCI_CHIP_E7221_G_BRIDGE 0x2580 -#endif - -#ifndef PCI_CHIP_I945_G -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_G_BRIDGE 0x2770 -#endif - -#ifndef PCI_CHIP_I945_GM -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GM_BRIDGE 0x27A0 -#endif - -#define IS_I810(pI810) (pI810->Chipset == PCI_CHIP_I810 || \ - pI810->Chipset == PCI_CHIP_I810_DC100 || \ - pI810->Chipset == PCI_CHIP_I810_E) -#define IS_I815(pI810) (pI810->Chipset == PCI_CHIP_I815) -#define IS_I830(pI810) (pI810->Chipset == PCI_CHIP_I830_M) -#define IS_845G(pI810) (pI810->Chipset == PCI_CHIP_845_G) -#define IS_I85X(pI810) (pI810->Chipset == PCI_CHIP_I855_GM) -#define IS_I852(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I852_GM || pI810->variant == I852_GME)) -#define IS_I855(pI810) (pI810->Chipset == PCI_CHIP_I855_GM && (pI810->variant == I855_GM || pI810->variant == I855_GME)) -#define IS_I865G(pI810) (pI810->Chipset == PCI_CHIP_I865_G) - -#define IS_I915G(pI810) (pI810->Chipset == PCI_CHIP_I915_G || pI810->Chipset == PCI_CHIP_E7221_G) -#define IS_I915GM(pI810) (pI810->Chipset == PCI_CHIP_I915_GM) -#define IS_I945G(pI810) (pI810->Chipset == PCI_CHIP_I945_G) -#define IS_I945GM(pI810) (pI810->Chipset == PCI_CHIP_I945_GM) -#define IS_I9XX(pI810) (IS_I915G(pI810) || IS_I915GM(pI810) || IS_I945G(pI810) || IS_I945GM(pI810)) - -#define IS_MOBILE(pI810) (IS_I830(pI810) || IS_I85X(pI810) || IS_I915GM(pI810) || IS_I945GM(pI810)) - -#define I830_GMCH_CTRL 0x52 - -#define I830_GMCH_MEM_MASK 0x1 -#define I830_GMCH_MEM_64M 0x1 -#define I830_GMCH_MEM_128M 0 - -#define I830_GMCH_GMS_MASK 0x70 -#define I830_GMCH_GMS_DISABLED 0x00 -#define I830_GMCH_GMS_LOCAL 0x10 -#define I830_GMCH_GMS_STOLEN_512 0x20 -#define I830_GMCH_GMS_STOLEN_1024 0x30 -#define I830_GMCH_GMS_STOLEN_8192 0x40 - -#define I855_GMCH_GMS_MASK (0x7 << 4) -#define I855_GMCH_GMS_DISABLED 0x00 -#define I855_GMCH_GMS_STOLEN_1M (0x1 << 4) -#define I855_GMCH_GMS_STOLEN_4M (0x2 << 4) -#define I855_GMCH_GMS_STOLEN_8M (0x3 << 4) -#define I855_GMCH_GMS_STOLEN_16M (0x4 << 4) -#define I855_GMCH_GMS_STOLEN_32M (0x5 << 4) -#define I915G_GMCH_GMS_STOLEN_48M (0x6 << 4) -#define I915G_GMCH_GMS_STOLEN_64M (0x7 << 4) - -typedef unsigned char Bool; -#define TRUE 1 -#define FALSE 0 - -#define PIPE_NONE 0<<0 -#define PIPE_CRT 1<<0 -#define PIPE_TV 1<<1 -#define PIPE_DFP 1<<2 -#define PIPE_LFP 1<<3 -#define PIPE_CRT2 1<<4 -#define PIPE_TV2 1<<5 -#define PIPE_DFP2 1<<6 -#define PIPE_LFP2 1<<7 - -typedef struct _I830MemPool *I830MemPoolPtr; -typedef struct _I830MemRange *I830MemRangePtr; -typedef struct _I830MemRange { - long Start; - long End; - long Size; - unsigned long Physical; - unsigned long Offset; /* Offset of AGP-allocated portion */ - unsigned long Alignment; - drm_handle_t Key; - unsigned long Pitch; // add pitch - I830MemPoolPtr Pool; -} I830MemRange; - -typedef struct _I830MemPool { - I830MemRange Total; - I830MemRange Free; - I830MemRange Fixed; - I830MemRange Allocated; -} I830MemPool; - -typedef struct { - int tail_mask; - I830MemRange mem; - unsigned char *virtual_start; - int head; - int tail; - int space; -} I830RingBuffer; - -typedef struct _I830Rec { - unsigned char *MMIOBase; - unsigned char *FbBase; - int cpp; - uint32_t aper_size; - unsigned int bios_version; - - /* These are set in PreInit and never changed. */ - long FbMapSize; - long TotalVideoRam; - I830MemRange StolenMemory; /* pre-allocated memory */ - long BIOSMemorySize; /* min stolen pool size */ - int BIOSMemSizeLoc; - - /* These change according to what has been allocated. */ - long FreeMemory; - I830MemRange MemoryAperture; - I830MemPool StolenPool; - long allocatedMemory; - - /* Regions allocated either from the above pools, or from agpgart. */ - /* for single and dual head configurations */ - I830MemRange FrontBuffer; - I830MemRange FrontBuffer2; - I830MemRange Scratch; - I830MemRange Scratch2; - - I830RingBuffer *LpRing; - - I830MemRange BackBuffer; - I830MemRange DepthBuffer; - I830MemRange TexMem; - int TexGranularity; - I830MemRange ContextMem; - int drmMinor; - Bool have3DWindows; - - Bool NeedRingBufferLow; - Bool allowPageFlip; - Bool disableTiling; - - int Chipset; - unsigned long LinearAddr; - unsigned long MMIOAddr; - - drmSize registerSize; /**< \brief MMIO register map size */ - drm_handle_t registerHandle; /**< \brief MMIO register map handle */ - // IOADDRESS ioBase; - int irq; /**< \brief IRQ number */ - int GttBound; - - drm_handle_t ring_map; - unsigned int Fence[8]; - -} I830Rec; - -/* - * 12288 is set as the maximum, chosen because it is enough for - * 1920x1440@32bpp with a 2048 pixel line pitch with some to spare. - */ -#define I830_MAXIMUM_VBIOS_MEM 12288 -#define I830_DEFAULT_VIDEOMEM_2D (MB(32) / 1024) -#define I830_DEFAULT_VIDEOMEM_3D (MB(64) / 1024) - -/* Flags for memory allocation function */ -#define FROM_ANYWHERE 0x00000000 -#define FROM_POOL_ONLY 0x00000001 -#define FROM_NEW_ONLY 0x00000002 -#define FROM_MASK 0x0000000f - -#define ALLOCATE_AT_TOP 0x00000010 -#define ALLOCATE_AT_BOTTOM 0x00000020 -#define FORCE_GAPS 0x00000040 - -#define NEED_PHYSICAL_ADDR 0x00000100 -#define ALIGN_BOTH_ENDS 0x00000200 -#define FORCE_LOW 0x00000400 - -#define ALLOC_NO_TILING 0x00001000 -#define ALLOC_INITIAL 0x00002000 - -#define ALLOCATE_DRY_RUN 0x80000000 - -/* Chipset registers for VIDEO BIOS memory RW access */ -#define _855_DRAM_RW_CONTROL 0x58 -#define _845_DRAM_RW_CONTROL 0x90 -#define DRAM_WRITE 0x33330000 - -#define KB(x) ((x) * 1024) -#define MB(x) ((x) * KB(1024)) - -#define GTT_PAGE_SIZE KB(4) -#define ROUND_TO(x, y) (((x) + (y) - 1) / (y) * (y)) -#define ROUND_DOWN_TO(x, y) ((x) / (y) * (y)) -#define ROUND_TO_PAGE(x) ROUND_TO((x), GTT_PAGE_SIZE) -#define ROUND_TO_MB(x) ROUND_TO((x), MB(1)) -#define PRIMARY_RINGBUFFER_SIZE KB(128) - - -/* Ring buffer registers, p277, overview p19 - */ -#define LP_RING 0x2030 -#define HP_RING 0x2040 - -#define RING_TAIL 0x00 -#define TAIL_ADDR 0x000FFFF8 -#define I830_TAIL_MASK 0x001FFFF8 - -#define RING_HEAD 0x04 -#define HEAD_WRAP_COUNT 0xFFE00000 -#define HEAD_WRAP_ONE 0x00200000 -#define HEAD_ADDR 0x001FFFFC -#define I830_HEAD_MASK 0x001FFFFC - -#define RING_START 0x08 -#define START_ADDR 0x03FFFFF8 -#define I830_RING_START_MASK 0xFFFFF000 - -#define RING_LEN 0x0C -#define RING_NR_PAGES 0x001FF000 -#define I830_RING_NR_PAGES 0x001FF000 -#define RING_REPORT_MASK 0x00000006 -#define RING_REPORT_64K 0x00000002 -#define RING_REPORT_128K 0x00000004 -#define RING_NO_REPORT 0x00000000 -#define RING_VALID_MASK 0x00000001 -#define RING_VALID 0x00000001 -#define RING_INVALID 0x00000000 - - -/* Fence/Tiling ranges [0..7] - */ -#define FENCE 0x2000 -#define FENCE_NR 8 - -#define I915G_FENCE_START_MASK 0x0ff00000 - -#define I830_FENCE_START_MASK 0x07f80000 - -#define FENCE_START_MASK 0x03F80000 -#define FENCE_X_MAJOR 0x00000000 -#define FENCE_Y_MAJOR 0x00001000 -#define FENCE_SIZE_MASK 0x00000700 -#define FENCE_SIZE_512K 0x00000000 -#define FENCE_SIZE_1M 0x00000100 -#define FENCE_SIZE_2M 0x00000200 -#define FENCE_SIZE_4M 0x00000300 -#define FENCE_SIZE_8M 0x00000400 -#define FENCE_SIZE_16M 0x00000500 -#define FENCE_SIZE_32M 0x00000600 -#define FENCE_SIZE_64M 0x00000700 -#define I915G_FENCE_SIZE_1M 0x00000000 -#define I915G_FENCE_SIZE_2M 0x00000100 -#define I915G_FENCE_SIZE_4M 0x00000200 -#define I915G_FENCE_SIZE_8M 0x00000300 -#define I915G_FENCE_SIZE_16M 0x00000400 -#define I915G_FENCE_SIZE_32M 0x00000500 -#define I915G_FENCE_SIZE_64M 0x00000600 -#define I915G_FENCE_SIZE_128M 0x00000700 -#define FENCE_PITCH_1 0x00000000 -#define FENCE_PITCH_2 0x00000010 -#define FENCE_PITCH_4 0x00000020 -#define FENCE_PITCH_8 0x00000030 -#define FENCE_PITCH_16 0x00000040 -#define FENCE_PITCH_32 0x00000050 -#define FENCE_PITCH_64 0x00000060 -#define FENCE_VALID 0x00000001 - -#include - -# define MMIO_IN8(base, offset) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) -# define MMIO_IN32(base, offset) \ - read_MMIO_LE32(base, offset) -# define MMIO_OUT8(base, offset, val) \ - *(volatile unsigned char *)(((unsigned char*)(base)) + (offset)) = (val) -# define MMIO_OUT32(base, offset, val) \ - *(volatile unsigned int *)(void *)(((unsigned char*)(base)) + (offset)) = CPU_TO_LE32(val) - - - /* Memory mapped register access macros */ -#define INREG8(addr) MMIO_IN8(MMIO, addr) -#define INREG(addr) MMIO_IN32(MMIO, addr) -#define OUTREG8(addr, val) MMIO_OUT8(MMIO, addr, val) -#define OUTREG(addr, val) MMIO_OUT32(MMIO, addr, val) - -#define DSPABASE 0x70184 - -#endif diff --git a/src/gallium/winsys/dri/intel/server/intel_dri.c b/src/gallium/winsys/dri/intel/server/intel_dri.c deleted file mode 100644 index e49c4214ad..0000000000 --- a/src/gallium/winsys/dri/intel/server/intel_dri.c +++ /dev/null @@ -1,1306 +0,0 @@ -/** - * \file server/intel_dri.c - * \brief File to perform the device-specific initialization tasks typically - * done in the X server. - * - * Here they are converted to run in the client (or perhaps a standalone - * process), and to work with the frame buffer device rather than the X - * server infrastructure. - * - * Copyright (C) 2006 Dave Airlie (airlied@linux.ie) - - Permission is hereby granted, free of charge, to any person obtaining a - copy of this software and associated documentation files (the - "Software"), to deal in the Software without restriction, including - without limitation the rights to use, copy, modify, merge, publish, - distribute, sub license, and/or sell copies of the Software, and to - permit persons to whom the Software is furnished to do so, subject to - the following conditions: - - The above copyright notice and this permission notice (including the - next paragraph) shall be included in all copies or substantial portions - of the Software. - - THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - IN NO EVENT SHALL THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR - ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. -*/ - -#include -#include -#include -#include -#include - -#include "driver.h" -#include "drm.h" - -#include "intel.h" -#include "i830_dri.h" - -#include "memops.h" -#include "pciaccess.h" - -static size_t drm_page_size; -static int nextTile = 0; -#define xf86DrvMsg(...) do {} while(0) - -static const int pitches[] = { - 128 * 8, - 128 * 16, - 128 * 32, - 128 * 64, - 0 -}; - -static Bool I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea); - -static unsigned long -GetBestTileAlignment(unsigned long size) -{ - unsigned long i; - - for (i = KB(512); i < size; i <<= 1) - ; - - if (i > MB(64)) - i = MB(64); - - return i; -} - -static void SetFenceRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - unsigned char *MMIO = ctx->MMIOAddress; - - for (i = 0; i < 8; i++) { - OUTREG(FENCE + i * 4, pI830->Fence[i]); - // if (I810_DEBUG & DEBUG_VERBOSE_VGA) - fprintf(stderr,"Fence Register : %x\n", pI830->Fence[i]); - } -} - -/* Tiled memory is good... really, really good... - * - * Need to make it less likely that we miss out on this - probably - * need to move the frontbuffer away from the 'guarenteed' alignment - * of the first memory segment, or perhaps allocate a discontigous - * framebuffer to get more alignment 'sweet spots'. - */ -static void -SetFence(const DRIDriverContext *ctx, I830Rec *pI830, - int nr, unsigned int start, unsigned int pitch, - unsigned int size) -{ - unsigned int val; - unsigned int fence_mask = 0; - unsigned int fence_pitch; - - if (nr < 0 || nr > 7) { - fprintf(stderr, - "SetFence: fence %d out of range\n",nr); - return; - } - - pI830->Fence[nr] = 0; - - if (IS_I9XX(pI830)) - fence_mask = ~I915G_FENCE_START_MASK; - else - fence_mask = ~I830_FENCE_START_MASK; - - if (start & fence_mask) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not %s aligned\n", - nr, start, (IS_I9XX(pI830)) ? "1MB" : "512k"); - return; - } - - if (start % size) { - fprintf(stderr, - "SetFence: %d: start (0x%08x) is not size (%dk) aligned\n", - nr, start, size / 1024); - return; - } - - if (pitch & 127) { - fprintf(stderr, - "SetFence: %d: pitch (%d) not a multiple of 128 bytes\n", - nr, pitch); - return; - } - - val = (start | FENCE_X_MAJOR | FENCE_VALID); - - if (IS_I9XX(pI830)) { - switch (size) { - case MB(1): - val |= I915G_FENCE_SIZE_1M; - break; - case MB(2): - val |= I915G_FENCE_SIZE_2M; - break; - case MB(4): - val |= I915G_FENCE_SIZE_4M; - break; - case MB(8): - val |= I915G_FENCE_SIZE_8M; - break; - case MB(16): - val |= I915G_FENCE_SIZE_16M; - break; - case MB(32): - val |= I915G_FENCE_SIZE_32M; - break; - case MB(64): - val |= I915G_FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } else { - switch (size) { - case KB(512): - val |= FENCE_SIZE_512K; - break; - case MB(1): - val |= FENCE_SIZE_1M; - break; - case MB(2): - val |= FENCE_SIZE_2M; - break; - case MB(4): - val |= FENCE_SIZE_4M; - break; - case MB(8): - val |= FENCE_SIZE_8M; - break; - case MB(16): - val |= FENCE_SIZE_16M; - break; - case MB(32): - val |= FENCE_SIZE_32M; - break; - case MB(64): - val |= FENCE_SIZE_64M; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal size (%d kByte)\n", nr, size / 1024); - return; - } - } - - if (IS_I9XX(pI830)) - fence_pitch = pitch / 512; - else - fence_pitch = pitch / 128; - - switch (fence_pitch) { - case 1: - val |= FENCE_PITCH_1; - break; - case 2: - val |= FENCE_PITCH_2; - break; - case 4: - val |= FENCE_PITCH_4; - break; - case 8: - val |= FENCE_PITCH_8; - break; - case 16: - val |= FENCE_PITCH_16; - break; - case 32: - val |= FENCE_PITCH_32; - break; - case 64: - val |= FENCE_PITCH_64; - break; - default: - fprintf(stderr, - "SetFence: %d: illegal pitch (%d)\n", nr, pitch); - return; - } - - pI830->Fence[nr] = val; -} - -static Bool -MakeTiles(const DRIDriverContext *ctx, I830Rec *pI830, I830MemRange *pMem) -{ - int pitch, ntiles, i; - - pitch = pMem->Pitch * ctx->cpp; - /* - * Simply try to break the region up into at most four pieces of size - * equal to the alignment. - */ - ntiles = ROUND_TO(pMem->Size, pMem->Alignment) / pMem->Alignment; - if (ntiles >= 4) { - return FALSE; - } - - for (i = 0; i < ntiles; i++, nextTile++) { - SetFence(ctx, pI830, nextTile, pMem->Start + i * pMem->Alignment, - pitch, pMem->Alignment); - } - return TRUE; -} - -static void I830SetupMemoryTiling(const DRIDriverContext *ctx, I830Rec *pI830) -{ - int i; - - /* Clear out */ - for (i = 0; i < 8; i++) - pI830->Fence[i] = 0; - - nextTile = 0; - - if (pI830->BackBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->BackBuffer))) { - fprintf(stderr, - "Activating tiled memory for the back buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the back buffer.\n"); - pI830->allowPageFlip = FALSE; - } - } - - if (pI830->DepthBuffer.Alignment >= KB(512)) { - if (MakeTiles(ctx, pI830, &(pI830->DepthBuffer))) { - fprintf(stderr, - "Activating tiled memory for the depth buffer.\n"); - } else { - fprintf(stderr, - "MakeTiles failed for the depth buffer.\n"); - } - } - - return; -} - -static int I830DetectMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - struct pci_device host_bridge, ig_dev; - uint32_t gmch_ctrl; - int memsize = 0; - int range; - uint32_t aper_size; - uint32_t membase2 = 0; - - memset(&host_bridge, 0, sizeof(host_bridge)); - memset(&ig_dev, 0, sizeof(ig_dev)); - - ig_dev.dev = 2; - - pci_device_cfg_read_u32(&host_bridge, &gmch_ctrl, I830_GMCH_CTRL); - - if (IS_I830(pI830) || IS_845G(pI830)) { - if ((gmch_ctrl & I830_GMCH_MEM_MASK) == I830_GMCH_MEM_128M) { - aper_size = 0x80000000; - } else { - aper_size = 0x40000000; - } - } else { - if (IS_I9XX(pI830)) { - int ret; - ret = pci_device_cfg_read_u32(&ig_dev, &membase2, 0x18); - if (membase2 & 0x08000000) - aper_size = 0x8000000; - else - aper_size = 0x10000000; - - fprintf(stderr,"aper size is %08X %08x %d\n", aper_size, membase2, ret); - } else - aper_size = 0x8000000; - } - - pI830->aper_size = aper_size; - - - /* We need to reduce the stolen size, by the GTT and the popup. - * The GTT varying according the the FbMapSize and the popup is 4KB */ - range = (ctx->shared.fbSize / (1024*1024)) + 4; - - if (IS_I85X(pI830) || IS_I865G(pI830) || IS_I9XX(pI830)) { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I855_GMCH_GMS_STOLEN_1M: - memsize = MB(1) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_4M: - memsize = MB(4) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_8M: - memsize = MB(8) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_16M: - memsize = MB(16) - KB(range); - break; - case I855_GMCH_GMS_STOLEN_32M: - memsize = MB(32) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_48M: - if (IS_I9XX(pI830)) - memsize = MB(48) - KB(range); - break; - case I915G_GMCH_GMS_STOLEN_64M: - if (IS_I9XX(pI830)) - memsize = MB(64) - KB(range); - break; - } - } else { - switch (gmch_ctrl & I830_GMCH_GMS_MASK) { - case I830_GMCH_GMS_STOLEN_512: - memsize = KB(512) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_1024: - memsize = MB(1) - KB(range); - break; - case I830_GMCH_GMS_STOLEN_8192: - memsize = MB(8) - KB(range); - break; - case I830_GMCH_GMS_LOCAL: - memsize = 0; - xf86DrvMsg(pScrn->scrnIndex, X_WARNING, - "Local memory found, but won't be used.\n"); - break; - } - } - if (memsize > 0) { - fprintf(stderr, - "detected %d kB stolen memory.\n", memsize / 1024); - } else { - fprintf(stderr, - "no video memory detected.\n"); - } - return memsize; -} - -static int AgpInit(const DRIDriverContext *ctx, I830Rec *info) -{ - unsigned long mode = 0x4; - - if (drmAgpAcquire(ctx->drmFD) < 0) { - fprintf(stderr, "[gart] AGP not available\n"); - return 0; - } - - if (drmAgpEnable(ctx->drmFD, mode) < 0) { - fprintf(stderr, "[gart] AGP not enabled\n"); - drmAgpRelease(ctx->drmFD); - return 0; - } - else - fprintf(stderr, "[gart] AGP enabled at %dx\n", ctx->agpmode); - - return 1; -} - -/* - * Allocate memory from the given pool. Grow the pool if needed and if - * possible. - */ -static unsigned long -AllocFromPool(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, - long size, unsigned long alignment, int flags) -{ - long needed, start, end; - - if (!result || !pool || !size) - return 0; - - /* Calculate how much space is needed. */ - if (alignment <= GTT_PAGE_SIZE) - needed = size; - else { - start = ROUND_TO(pool->Free.Start, alignment); - end = ROUND_TO(start + size, alignment); - needed = end - pool->Free.Start; - } - if (needed > pool->Free.Size) { - return 0; - } - - result->Start = ROUND_TO(pool->Free.Start, alignment); - pool->Free.Start += needed; - result->End = pool->Free.Start; - - pool->Free.Size = pool->Free.End - pool->Free.Start; - result->Size = result->End - result->Start; - result->Pool = pool; - result->Alignment = alignment; - return needed; -} - -static unsigned long AllocFromAGP(const DRIDriverContext *ctx, I830Rec *pI830, long size, unsigned long alignment, I830MemRange *result) -{ - unsigned long start, end; - unsigned long newApStart, newApEnd; - int ret; - if (!result || !size) - return 0; - - if (!alignment) - alignment = 4; - - start = ROUND_TO(pI830->MemoryAperture.Start, alignment); - end = ROUND_TO(start + size, alignment); - newApStart = end; - newApEnd = pI830->MemoryAperture.End; - - ret=drmAgpAlloc(ctx->drmFD, size, 0, &(result->Physical), (drm_handle_t *)&(result->Key)); - - if (ret) - { - fprintf(stderr,"drmAgpAlloc failed %d\n", ret); - return 0; - } - pI830->allocatedMemory += size; - pI830->MemoryAperture.Start = newApStart; - pI830->MemoryAperture.End = newApEnd; - pI830->MemoryAperture.Size = newApEnd - newApStart; - // pI830->FreeMemory -= size; - result->Start = start; - result->End = start + size; - result->Size = size; - result->Offset = start; - result->Alignment = alignment; - result->Pool = NULL; - - return size; -} - -unsigned long -I830AllocVidMem(const DRIDriverContext *ctx, I830Rec *pI830, - I830MemRange *result, I830MemPool *pool, long size, - unsigned long alignment, int flags) -{ - unsigned long ret; - - if (!result) - return 0; - - /* Make sure these are initialised. */ - result->Size = 0; - result->Key = -1; - - if (!size) { - return 0; - } - - if (pool->Free.Size < size) { - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - else { - ret = AllocFromPool(ctx, pI830, result, pool, size, alignment, flags); - if (ret == 0) - ret = AllocFromAGP(ctx, pI830, size, alignment, result); - } - return ret; -} - -static Bool BindAgpRange(const DRIDriverContext *ctx, I830MemRange *mem) -{ - if (!mem) - return FALSE; - - if (mem->Key == -1) - return TRUE; - - return !drmAgpBind(ctx->drmFD, mem->Key, mem->Offset); -} - -/* simple memory allocation routines needed */ -/* put ring buffer in low memory */ -/* need to allocate front, back, depth buffers aligned correctly, - allocate ring buffer, -*/ - -/* */ -static Bool -I830AllocateMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned long size, ret; - unsigned long lines, lineSize, align; - - /* allocate ring buffer */ - memset(pI830->LpRing, 0, sizeof(I830RingBuffer)); - pI830->LpRing->mem.Key = -1; - - size = PRIMARY_RINGBUFFER_SIZE; - - ret = I830AllocVidMem(ctx, pI830, &pI830->LpRing->mem, &pI830->StolenPool, size, 0x1000, 0); - - if (ret != size) - { - fprintf(stderr,"unable to allocate ring buffer %ld\n", ret); - return FALSE; - } - - pI830->LpRing->tail_mask = pI830->LpRing->mem.Size - 1; - - - /* allocate front buffer */ - memset(&(pI830->FrontBuffer), 0, sizeof(pI830->FrontBuffer)); - pI830->FrontBuffer.Key = -1; - pI830->FrontBuffer.Pitch = ctx->shared.virtualWidth; - - align = KB(512); - - lineSize = ctx->shared.virtualWidth * ctx->cpp; - lines = (ctx->shared.virtualHeight + 15) / 16 * 16; - size = lineSize * lines; - size = ROUND_TO_PAGE(size); - - align = GetBestTileAlignment(size); - - ret = I830AllocVidMem(ctx, pI830, &pI830->FrontBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate front buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->BackBuffer), 0, sizeof(pI830->BackBuffer)); - pI830->BackBuffer.Key = -1; - pI830->BackBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->BackBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate back buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->DepthBuffer), 0, sizeof(pI830->DepthBuffer)); - pI830->DepthBuffer.Key = -1; - pI830->DepthBuffer.Pitch = ctx->shared.virtualWidth; - - ret = I830AllocVidMem(ctx, pI830, &pI830->DepthBuffer, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate depth buffer %ld\n", ret); - return FALSE; - } - - memset(&(pI830->ContextMem), 0, sizeof(pI830->ContextMem)); - pI830->ContextMem.Key = -1; - size = KB(32); - - ret = I830AllocVidMem(ctx, pI830, &pI830->ContextMem, &pI830->StolenPool, size, align, 0); - if (ret < size) - { - fprintf(stderr,"unable to allocate context buffer %ld\n", ret); - return FALSE; - } - -#if 0 - memset(&(pI830->TexMem), 0, sizeof(pI830->TexMem)); - pI830->TexMem.Key = -1; - - size = 32768 * 1024; - ret = AllocFromAGP(ctx, pI830, size, align, &pI830->TexMem); - if (ret < size) - { - fprintf(stderr,"unable to allocate texture memory %ld\n", ret); - return FALSE; - } -#endif - - return TRUE; -} - -static Bool -I830BindMemory(const DRIDriverContext *ctx, I830Rec *pI830) -{ - if (!BindAgpRange(ctx, &pI830->LpRing->mem)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->FrontBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->BackBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->DepthBuffer)) - return FALSE; - if (!BindAgpRange(ctx, &pI830->ContextMem)) - return FALSE; -#if 0 - if (!BindAgpRange(ctx, &pI830->TexMem)) - return FALSE; -#endif - return TRUE; -} - -static void SetupDRIMM(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned long aperEnd = ROUND_DOWN_TO(pI830->aper_size, GTT_PAGE_SIZE) / GTT_PAGE_SIZE; - unsigned long aperStart = ROUND_TO(pI830->aper_size - KB(32768), GTT_PAGE_SIZE) / GTT_PAGE_SIZE; - - fprintf(stderr, "aper size is %08X\n", ctx->shared.fbSize); - if (drmMMInit(ctx->drmFD, aperStart, aperEnd - aperStart, DRM_BO_MEM_TT)) { - fprintf(stderr, - "DRM MM Initialization Failed\n"); - } else { - fprintf(stderr, - "DRM MM Initialized at offset 0x%lx length %d page\n", aperStart, aperEnd-aperStart); - } - -} - -static Bool -I830CleanupDma(const DRIDriverContext *ctx) -{ - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_CLEANUP_DMA; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, "I830 Dma Cleanup Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830InitDma(const DRIDriverContext *ctx, I830Rec *pI830) -{ - I830RingBuffer *ring = pI830->LpRing; - drmI830Init info; - - memset(&info, 0, sizeof(drmI830Init)); - info.func = I830_INIT_DMA; - - info.ring_start = ring->mem.Start + pI830->LinearAddr; - info.ring_end = ring->mem.End + pI830->LinearAddr; - info.ring_size = ring->mem.Size; - - info.mmio_offset = (unsigned int)ctx->MMIOStart; - - info.sarea_priv_offset = sizeof(drm_sarea_t); - - info.front_offset = pI830->FrontBuffer.Start; - info.back_offset = pI830->BackBuffer.Start; - info.depth_offset = pI830->DepthBuffer.Start; - info.w = ctx->shared.virtualWidth; - info.h = ctx->shared.virtualHeight; - info.pitch = ctx->shared.virtualWidth; - info.back_pitch = pI830->BackBuffer.Pitch; - info.depth_pitch = pI830->DepthBuffer.Pitch; - info.cpp = ctx->cpp; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_INIT, - &info, sizeof(drmI830Init))) { - fprintf(stderr, - "I830 Dma Initialization Failed\n"); - return FALSE; - } - - return TRUE; -} - -static int I830CheckDRMVersion( const DRIDriverContext *ctx, - I830Rec *pI830 ) -{ - drmVersionPtr version; - - version = drmGetVersion(ctx->drmFD); - - if (version) { - int req_minor, req_patch; - - req_minor = 4; - req_patch = 0; - - if (version->version_major != 1 || - version->version_minor < req_minor || - (version->version_minor == req_minor && - version->version_patchlevel < req_patch)) { - /* Incompatible drm version */ - fprintf(stderr, - "[dri] I830DRIScreenInit failed because of a version " - "mismatch.\n" - "[dri] i915.o kernel module version is %d.%d.%d " - "but version 1.%d.%d or newer is needed.\n" - "[dri] Disabling DRI.\n", - version->version_major, - version->version_minor, - version->version_patchlevel, - req_minor, - req_patch); - drmFreeVersion(version); - return 0; - } - - pI830->drmMinor = version->version_minor; - drmFreeVersion(version); - } - return 1; -} - -static void -I830SetRingRegs(const DRIDriverContext *ctx, I830Rec *pI830) -{ - unsigned int itemp; - unsigned char *MMIO = ctx->MMIOAddress; - - OUTREG(LP_RING + RING_LEN, 0); - OUTREG(LP_RING + RING_TAIL, 0); - OUTREG(LP_RING + RING_HEAD, 0); - - if ((long)(pI830->LpRing->mem.Start & I830_RING_START_MASK) != - pI830->LpRing->mem.Start) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer start (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Start, I830_RING_START_MASK); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = pI830->LpRing->mem.Start & I830_RING_START_MASK; - OUTREG(LP_RING + RING_START, itemp); - - if (((pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES) != - pI830->LpRing->mem.Size - 4096) { - fprintf(stderr, - "I830SetRingRegs: Ring buffer size - 4096 (%lx) violates its " - "mask (%x)\n", pI830->LpRing->mem.Size - 4096, - I830_RING_NR_PAGES); - } - /* Don't care about the old value. Reserved bits must be zero anyway. */ - itemp = (pI830->LpRing->mem.Size - 4096) & I830_RING_NR_PAGES; - itemp |= (RING_NO_REPORT | RING_VALID); - OUTREG(LP_RING + RING_LEN, itemp); - - pI830->LpRing->head = INREG(LP_RING + RING_HEAD) & I830_HEAD_MASK; - pI830->LpRing->tail = INREG(LP_RING + RING_TAIL); - pI830->LpRing->space = pI830->LpRing->head - (pI830->LpRing->tail + 8); - if (pI830->LpRing->space < 0) - pI830->LpRing->space += pI830->LpRing->mem.Size; - - SetFenceRegs(ctx, pI830); - - /* RESET THE DISPLAY PIPE TO POINT TO THE FRONTBUFFER - hacky - hacky hacky */ - OUTREG(DSPABASE, pI830->FrontBuffer.Start + pI830->LinearAddr); - -} - -static Bool -I830SetParam(const DRIDriverContext *ctx, int param, int value) -{ - drmI830SetParam sp; - - memset(&sp, 0, sizeof(sp)); - sp.param = param; - sp.value = value; - - if (drmCommandWrite(ctx->drmFD, DRM_I830_SETPARAM, &sp, sizeof(sp))) { - fprintf(stderr, "I830 SetParam Failed\n"); - return FALSE; - } - - return TRUE; -} - -static Bool -I830DRIMapScreenRegions(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - fprintf(stderr, - "[drm] Mapping front buffer\n"); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->front_offset + pI830->LinearAddr), - sarea->front_size, - DRM_FRAME_BUFFER, /*DRM_AGP,*/ - 0, - &sarea->front_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(front_handle) failed. Disabling DRI\n"); - return FALSE; - } - ctx->shared.hFrameBuffer = sarea->front_handle; - ctx->shared.fbSize = sarea->front_size; - fprintf(stderr, "[drm] Front Buffer = 0x%08x\n", - sarea->front_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)(sarea->back_offset), - sarea->back_size, DRM_AGP, 0, - &sarea->back_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(back_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Back Buffer = 0x%08x\n", - sarea->back_handle); - - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->depth_offset, - sarea->depth_size, DRM_AGP, 0, - &sarea->depth_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(depth_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] Depth Buffer = 0x%08x\n", - sarea->depth_handle); - -#if 0 - if (drmAddMap(ctx->drmFD, - (drm_handle_t)sarea->tex_offset, - sarea->tex_size, DRM_AGP, 0, - &sarea->tex_handle) < 0) { - fprintf(stderr, - "[drm] drmAddMap(tex_handle) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] textures = 0x%08x\n", - sarea->tex_handle); -#endif - return TRUE; -} - - -static void -I830DRIUnmapScreenRegions(const DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ -#if 1 - if (sarea->front_handle) { - drmRmMap(ctx->drmFD, sarea->front_handle); - sarea->front_handle = 0; - } -#endif - if (sarea->back_handle) { - drmRmMap(ctx->drmFD, sarea->back_handle); - sarea->back_handle = 0; - } - if (sarea->depth_handle) { - drmRmMap(ctx->drmFD, sarea->depth_handle); - sarea->depth_handle = 0; - } - if (sarea->tex_handle) { - drmRmMap(ctx->drmFD, sarea->tex_handle); - sarea->tex_handle = 0; - } -} - -static Bool -I830DRIDoMappings(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - if (drmAddMap(ctx->drmFD, - (drm_handle_t)pI830->LpRing->mem.Start, - pI830->LpRing->mem.Size, DRM_AGP, 0, - &pI830->ring_map) < 0) { - fprintf(stderr, - "[drm] drmAddMap(ring_map) failed. Disabling DRI\n"); - return FALSE; - } - fprintf(stderr, "[drm] ring buffer = 0x%08x\n", - pI830->ring_map); - - if (I830InitDma(ctx, pI830) == FALSE) { - return FALSE; - } - - /* init to zero to be safe */ - - I830DRIMapScreenRegions(ctx, pI830, sarea); - SetupDRIMM(ctx, pI830); - - if (ctx->pciDevice != PCI_CHIP_845_G && - ctx->pciDevice != PCI_CHIP_I830_M) { - I830SetParam(ctx, I830_SETPARAM_USE_MI_BATCHBUFFER_START, 1 ); - } - - /* Okay now initialize the dma engine */ - { - pI830->irq = drmGetInterruptFromBusID(ctx->drmFD, - ctx->pciBus, - ctx->pciDevice, - ctx->pciFunc); - - if (drmCtlInstHandler(ctx->drmFD, pI830->irq)) { - fprintf(stderr, - "[drm] failure adding irq handler\n"); - pI830->irq = 0; - return FALSE; - } - else - fprintf(stderr, - "[drm] dma control initialized, using IRQ %d\n", - pI830->irq); - } - - fprintf(stderr, "[dri] visual configs initialized\n"); - - return TRUE; -} - -static Bool -I830ClearScreen(DRIDriverContext *ctx, I830Rec *pI830, drmI830Sarea *sarea) -{ - /* need to drmMap front and back buffers and zero them */ - drmAddress map_addr; - int ret; - - ret = drmMap(ctx->drmFD, - sarea->front_handle, - sarea->front_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map front buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->front_size); - drmUnmap(map_addr, sarea->front_size); - - - ret = drmMap(ctx->drmFD, - sarea->back_handle, - sarea->back_size, - &map_addr); - - if (ret) - { - fprintf(stderr, "Unable to map back buffer\n"); - return FALSE; - } - - drimemsetio((char *)map_addr, - 0, - sarea->back_size); - drmUnmap(map_addr, sarea->back_size); - - return TRUE; -} - -static Bool -I830ScreenInit(DRIDriverContext *ctx, I830Rec *pI830) - -{ - I830DRIPtr pI830DRI; - drmI830Sarea *pSAREAPriv; - int err; - - drm_page_size = getpagesize(); - - pI830->registerSize = ctx->MMIOSize; - /* This is a hack for now. We have to have more than a 4k page here - * because of the size of the state. However, the state should be - * in a per-context mapping. This will be added in the Mesa 3.5 port - * of the I830 driver. - */ - ctx->shared.SAREASize = SAREA_MAX; - - /* Note that drmOpen will try to load the kernel module, if needed. */ - ctx->drmFD = drmOpen("i915", NULL ); - if (ctx->drmFD < 0) { - fprintf(stderr, "[drm] drmOpen failed\n"); - return 0; - } - - if ((err = drmSetBusid(ctx->drmFD, ctx->pciBusID)) < 0) { - fprintf(stderr, "[drm] drmSetBusid failed (%d, %s), %s\n", - ctx->drmFD, ctx->pciBusID, strerror(-err)); - return 0; - } - - if (drmAddMap( ctx->drmFD, - 0, - ctx->shared.SAREASize, - DRM_SHM, - DRM_CONTAINS_LOCK, - &ctx->shared.hSAREA) < 0) - { - fprintf(stderr, "[drm] drmAddMap failed\n"); - return 0; - } - - fprintf(stderr, "[drm] added %d byte SAREA at 0x%08x\n", - ctx->shared.SAREASize, ctx->shared.hSAREA); - - if (drmMap( ctx->drmFD, - ctx->shared.hSAREA, - ctx->shared.SAREASize, - (drmAddressPtr)(&ctx->pSAREA)) < 0) - { - fprintf(stderr, "[drm] drmMap failed\n"); - return 0; - - } - - memset(ctx->pSAREA, 0, ctx->shared.SAREASize); - fprintf(stderr, "[drm] mapped SAREA 0x%08x to %p, size %d\n", - ctx->shared.hSAREA, ctx->pSAREA, ctx->shared.SAREASize); - - - if (drmAddMap(ctx->drmFD, - ctx->MMIOStart, - ctx->MMIOSize, - DRM_REGISTERS, - DRM_READ_ONLY, - &pI830->registerHandle) < 0) { - fprintf(stderr, "[drm] drmAddMap mmio failed\n"); - return 0; - } - fprintf(stderr, - "[drm] register handle = 0x%08x\n", pI830->registerHandle); - - - if (!I830CheckDRMVersion(ctx, pI830)) { - return FALSE; - } - - /* Create a 'server' context so we can grab the lock for - * initialization ioctls. - */ - if ((err = drmCreateContext(ctx->drmFD, &ctx->serverContext)) != 0) { - fprintf(stderr, "%s: drmCreateContext failed %d\n", __FUNCTION__, err); - return 0; - } - - DRM_LOCK(ctx->drmFD, ctx->pSAREA, ctx->serverContext, 0); - - /* Initialize the SAREA private data structure */ - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - memset(pSAREAPriv, 0, sizeof(*pSAREAPriv)); - - pI830->StolenMemory.Size = I830DetectMemory(ctx, pI830); - pI830->StolenMemory.Start = 0; - pI830->StolenMemory.End = pI830->StolenMemory.Size; - - pI830->MemoryAperture.Start = pI830->StolenMemory.End; - pI830->MemoryAperture.End = KB(40000); - pI830->MemoryAperture.Size = pI830->MemoryAperture.End - pI830->MemoryAperture.Start; - - pI830->StolenPool.Fixed = pI830->StolenMemory; - pI830->StolenPool.Total = pI830->StolenMemory; - pI830->StolenPool.Free = pI830->StolenPool.Total; - pI830->FreeMemory = pI830->StolenPool.Total.Size; - - if (!AgpInit(ctx, pI830)) - return FALSE; - - if (I830AllocateMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - if (I830BindMemory(ctx, pI830) == FALSE) - { - return FALSE; - } - - pSAREAPriv->rotated_offset = -1; - pSAREAPriv->rotated_size = 0; - pSAREAPriv->rotated_pitch = ctx->shared.virtualWidth; - - pSAREAPriv->front_offset = pI830->FrontBuffer.Start; - pSAREAPriv->front_size = pI830->FrontBuffer.Size; - pSAREAPriv->width = ctx->shared.virtualWidth; - pSAREAPriv->height = ctx->shared.virtualHeight; - pSAREAPriv->pitch = ctx->shared.virtualWidth; - pSAREAPriv->virtualX = ctx->shared.virtualWidth; - pSAREAPriv->virtualY = ctx->shared.virtualHeight; - pSAREAPriv->back_offset = pI830->BackBuffer.Start; - pSAREAPriv->back_size = pI830->BackBuffer.Size; - pSAREAPriv->depth_offset = pI830->DepthBuffer.Start; - pSAREAPriv->depth_size = pI830->DepthBuffer.Size; -#if 0 - pSAREAPriv->tex_offset = pI830->TexMem.Start; - pSAREAPriv->tex_size = pI830->TexMem.Size; -#endif - pSAREAPriv->log_tex_granularity = pI830->TexGranularity; - - ctx->driverClientMsg = malloc(sizeof(I830DRIRec)); - ctx->driverClientMsgSize = sizeof(I830DRIRec); - pI830DRI = (I830DRIPtr)ctx->driverClientMsg; - pI830DRI->deviceID = pI830->Chipset; - pI830DRI->regsSize = I830_REG_SIZE; - pI830DRI->width = ctx->shared.virtualWidth; - pI830DRI->height = ctx->shared.virtualHeight; - pI830DRI->mem = ctx->shared.fbSize; - pI830DRI->cpp = ctx->cpp; - - pI830DRI->bitsPerPixel = ctx->bpp; - pI830DRI->sarea_priv_offset = sizeof(drm_sarea_t); - - err = I830DRIDoMappings(ctx, pI830, pSAREAPriv); - if (err == FALSE) - return FALSE; - - I830SetupMemoryTiling(ctx, pI830); - - /* Quick hack to clear the front & back buffers. Could also use - * the clear ioctl to do this, but would need to setup hw state - * first. - */ - I830ClearScreen(ctx, pI830, pSAREAPriv); - - I830SetRingRegs(ctx, pI830); - - return TRUE; -} - - -/** - * \brief Validate the fbdev mode. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Saves some registers and returns 1. - * - * \sa radeonValidateMode(). - */ -static int i830ValidateMode( const DRIDriverContext *ctx ) -{ - return 1; -} - -/** - * \brief Examine mode returned by fbdev. - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Restores registers that fbdev has clobbered and returns 1. - * - * \sa i810ValidateMode(). - */ -static int i830PostValidateMode( const DRIDriverContext *ctx ) -{ - I830Rec *pI830 = ctx->driverPrivate; - - I830SetRingRegs(ctx, pI830); - return 1; -} - - -/** - * \brief Initialize the framebuffer device mode - * - * \param ctx display handle. - * - * \return one on success, or zero on failure. - * - * Fills in \p info with some default values and some information from \p ctx - * and then calls I810ScreenInit() for the screen initialization. - * - * Before exiting clears the framebuffer memory accessing it directly. - */ -static int i830InitFBDev( DRIDriverContext *ctx ) -{ - I830Rec *pI830 = calloc(1, sizeof(I830Rec)); - int i; - - { - int dummy = ctx->shared.virtualWidth; - - switch (ctx->bpp / 8) { - case 1: dummy = (ctx->shared.virtualWidth + 127) & ~127; break; - case 2: dummy = (ctx->shared.virtualWidth + 31) & ~31; break; - case 3: - case 4: dummy = (ctx->shared.virtualWidth + 15) & ~15; break; - } - - ctx->shared.virtualWidth = dummy; - ctx->shared.Width = ctx->shared.virtualWidth; - } - - - for (i = 0; pitches[i] != 0; i++) { - if (pitches[i] >= ctx->shared.virtualWidth) { - ctx->shared.virtualWidth = pitches[i]; - break; - } - } - - ctx->driverPrivate = (void *)pI830; - - pI830->LpRing = calloc(1, sizeof(I830RingBuffer)); - pI830->Chipset = ctx->chipset; - pI830->LinearAddr = ctx->FBStart; - - if (!I830ScreenInit( ctx, pI830 )) - return 0; - - - return 1; -} - - -/** - * \brief The screen is being closed, so clean up any state and free any - * resources used by the DRI. - * - * \param ctx display handle. - * - * Unmaps the SAREA, closes the DRM device file descriptor and frees the driver - * private data. - */ -static void i830HaltFBDev( DRIDriverContext *ctx ) -{ - drmI830Sarea *pSAREAPriv; - I830Rec *pI830 = ctx->driverPrivate; - - if (pI830->irq) { - drmCtlUninstHandler(ctx->drmFD); - pI830->irq = 0; } - - I830CleanupDma(ctx); - - pSAREAPriv = (drmI830Sarea *)(((char*)ctx->pSAREA) + - sizeof(drm_sarea_t)); - - I830DRIUnmapScreenRegions(ctx, pI830, pSAREAPriv); - drmUnmap( ctx->pSAREA, ctx->shared.SAREASize ); - drmClose(ctx->drmFD); - - if (ctx->driverPrivate) { - free(ctx->driverPrivate); - ctx->driverPrivate = 0; - } -} - - -extern void i810NotifyFocus( int ); - -/** - * \brief Exported driver interface for Mini GLX. - * - * \sa DRIDriverRec. - */ -const struct DRIDriverRec __driDriver = { - i830ValidateMode, - i830PostValidateMode, - i830InitFBDev, - i830HaltFBDev, - NULL,//I830EngineShutdown, - NULL, //I830EngineRestore, -#ifndef _EMBEDDED - 0, -#else - i810NotifyFocus, -#endif -}; diff --git a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c deleted file mode 100644 index 1bc1089352..0000000000 --- a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.c +++ /dev/null @@ -1,953 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - * Keith Whitwell - */ - -#include -#include -#include "glthread.h" -#include "errno.h" -#include "ws_dri_bufmgr.h" -#include "string.h" -#include "imports.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" - -/* - * This lock is here to protect drmBO structs changing underneath us during a - * validate list call, since validatelist cannot take individiual locks for - * each drmBO. Validatelist takes this lock in write mode. Any access to an - * individual drmBO should take this lock in read mode, since in that case, the - * driBufferObject mutex will protect the access. Locking order is - * driBufferObject mutex - > this rw lock. - */ - -_glthread_DECLARE_STATIC_MUTEX(bmMutex); -_glthread_DECLARE_STATIC_COND(bmCond); - -static int kernelReaders = 0; -static int num_buffers = 0; -static int num_user_buffers = 0; - -static drmBO *drmBOListBuf(void *iterator) -{ - drmBONode *node; - drmMMListHead *l = (drmMMListHead *) iterator; - node = DRMLISTENTRY(drmBONode, l, head); - return node->buf; -} - -static void *drmBOListIterator(drmBOList *list) -{ - void *ret = list->list.next; - - if (ret == &list->list) - return NULL; - return ret; -} - -static void *drmBOListNext(drmBOList *list, void *iterator) -{ - void *ret; - - drmMMListHead *l = (drmMMListHead *) iterator; - ret = l->next; - if (ret == &list->list) - return NULL; - return ret; -} - -static drmBONode *drmAddListItem(drmBOList *list, drmBO *item, - uint64_t arg0, - uint64_t arg1) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->free.next; - if (l == &list->free) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - return NULL; - } - list->numCurrent++; - } - else { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - } - node->buf = item; - node->arg0 = arg0; - node->arg1 = arg1; - DRMLISTADD(&node->head, &list->list); - list->numOnList++; - return node; -} - -static int drmAddValidateItem(drmBOList *list, drmBO *buf, uint64_t flags, - uint64_t mask, int *newItem) -{ - drmBONode *node, *cur; - drmMMListHead *l; - - *newItem = 0; - cur = NULL; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - if (node->buf == buf) { - cur = node; - break; - } - } - if (!cur) { - cur = drmAddListItem(list, buf, flags, mask); - if (!cur) { - return -ENOMEM; - } - *newItem = 1; - cur->arg0 = flags; - cur->arg1 = mask; - } - else { - uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; - uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; - - if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { - return -EINVAL; - } - - cur->arg1 |= mask; - cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); - - if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && - (cur->arg0 & DRM_BO_MASK_MEM) == 0) { - return -EINVAL; - } - } - return 0; -} - -static void drmBOFreeList(drmBOList *list) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->list.next; - while(l != &list->list) { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - l = list->list.next; - list->numCurrent--; - list->numOnList--; - } - - l = list->free.next; - while(l != &list->free) { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - l = list->free.next; - list->numCurrent--; - } -} - -static int drmAdjustListNodes(drmBOList *list) -{ - drmBONode *node; - drmMMListHead *l; - int ret = 0; - - while(list->numCurrent < list->numTarget) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - ret = -ENOMEM; - break; - } - list->numCurrent++; - DRMLISTADD(&node->head, &list->free); - } - - while(list->numCurrent > list->numTarget) { - l = list->free.next; - if (l == &list->free) - break; - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - list->numCurrent--; - } - return ret; -} - -static int drmBOCreateList(int numTarget, drmBOList *list) -{ - DRMINITLISTHEAD(&list->list); - DRMINITLISTHEAD(&list->free); - list->numTarget = numTarget; - list->numCurrent = 0; - list->numOnList = 0; - return drmAdjustListNodes(list); -} - -static int drmBOResetList(drmBOList *list) -{ - drmMMListHead *l; - int ret; - - ret = drmAdjustListNodes(list); - if (ret) - return ret; - - l = list->list.next; - while (l != &list->list) { - DRMLISTDEL(l); - DRMLISTADD(l, &list->free); - list->numOnList--; - l = list->list.next; - } - return drmAdjustListNodes(list); -} - -void driWriteLockKernelBO(void) -{ - _glthread_LOCK_MUTEX(bmMutex); - while(kernelReaders != 0) - _glthread_COND_WAIT(bmCond, bmMutex); -} - -void driWriteUnlockKernelBO(void) -{ - _glthread_UNLOCK_MUTEX(bmMutex); -} - -void driReadLockKernelBO(void) -{ - _glthread_LOCK_MUTEX(bmMutex); - kernelReaders++; - _glthread_UNLOCK_MUTEX(bmMutex); -} - -void driReadUnlockKernelBO(void) -{ - _glthread_LOCK_MUTEX(bmMutex); - if (--kernelReaders == 0) - _glthread_COND_BROADCAST(bmCond); - _glthread_UNLOCK_MUTEX(bmMutex); -} - - - - -/* - * TODO: Introduce fence pools in the same way as - * buffer object pools. - */ - -typedef struct _DriBufferObject -{ - DriBufferPool *pool; - _glthread_Mutex mutex; - int refCount; - const char *name; - uint64_t flags; - unsigned hint; - unsigned alignment; - unsigned createdByReference; - void *private; - /* user-space buffer: */ - unsigned userBuffer; - void *userData; - unsigned userSize; -} DriBufferObject; - -typedef struct _DriBufferList { - drmBOList drmBuffers; /* List of kernel buffers needing validation */ - drmBOList driBuffers; /* List of user-space buffers needing validation */ -} DriBufferList; - - -void -bmError(int val, const char *file, const char *function, int line) -{ - _mesa_printf("Fatal video memory manager error \"%s\".\n" - "Check kernel logs or set the LIBGL_DEBUG\n" - "environment variable to \"verbose\" for more info.\n" - "Detected in file %s, line %d, function %s.\n", - strerror(-val), file, line, function); -#ifndef NDEBUG - abort(); -#else - abort(); -#endif -} - -extern drmBO * -driBOKernel(struct _DriBufferObject *buf) -{ - drmBO *ret; - - driReadLockKernelBO(); - _glthread_LOCK_MUTEX(buf->mutex); - assert(buf->private != NULL); - ret = buf->pool->kernel(buf->pool, buf->private); - if (!ret) - BM_CKFATAL(-EINVAL); - _glthread_UNLOCK_MUTEX(buf->mutex); - driReadUnlockKernelBO(); - - return ret; -} - -void -driBOWaitIdle(struct _DriBufferObject *buf, int lazy) -{ - - /* - * This function may block. Is it sane to keep the mutex held during - * that time?? - */ - - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy)); - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -void * -driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint) -{ - void *virtual; - int retval; - - if (buf->userBuffer) { - return buf->userData; - } - - _glthread_LOCK_MUTEX(buf->mutex); - assert(buf->private != NULL); - retval = buf->pool->map(buf->pool, buf->private, flags, hint, - &buf->mutex, &virtual); - _glthread_UNLOCK_MUTEX(buf->mutex); - - return retval == 0 ? virtual : NULL; -} - -void -driBOUnmap(struct _DriBufferObject *buf) -{ - if (buf->userBuffer) - return; - - assert(buf->private != NULL); - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -unsigned long -driBOOffset(struct _DriBufferObject *buf) -{ - unsigned long ret; - - assert(buf->private != NULL); - - _glthread_LOCK_MUTEX(buf->mutex); - ret = buf->pool->offset(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - return ret; -} - -unsigned long -driBOPoolOffset(struct _DriBufferObject *buf) -{ - unsigned long ret; - - assert(buf->private != NULL); - - _glthread_LOCK_MUTEX(buf->mutex); - ret = buf->pool->poolOffset(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - return ret; -} - -uint64_t -driBOFlags(struct _DriBufferObject *buf) -{ - uint64_t ret; - - assert(buf->private != NULL); - - driReadLockKernelBO(); - _glthread_LOCK_MUTEX(buf->mutex); - ret = buf->pool->flags(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - driReadUnlockKernelBO(); - return ret; -} - -struct _DriBufferObject * -driBOReference(struct _DriBufferObject *buf) -{ - _glthread_LOCK_MUTEX(buf->mutex); - if (++buf->refCount == 1) { - _glthread_UNLOCK_MUTEX(buf->mutex); - BM_CKFATAL(-EINVAL); - } - _glthread_UNLOCK_MUTEX(buf->mutex); - return buf; -} - -void -driBOUnReference(struct _DriBufferObject *buf) -{ - int tmp; - - if (!buf) - return; - - _glthread_LOCK_MUTEX(buf->mutex); - tmp = --buf->refCount; - if (!tmp) { - _glthread_UNLOCK_MUTEX(buf->mutex); - if (buf->private) { - if (buf->createdByReference) - buf->pool->unreference(buf->pool, buf->private); - else - buf->pool->destroy(buf->pool, buf->private); - } - if (buf->userBuffer) - num_user_buffers--; - else - num_buffers--; - free(buf); - } else - _glthread_UNLOCK_MUTEX(buf->mutex); - -} - - -int -driBOData(struct _DriBufferObject *buf, - unsigned size, const void *data, - DriBufferPool *newPool, - uint64_t flags) -{ - void *virtual = NULL; - int newBuffer; - int retval = 0; - struct _DriBufferPool *pool; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - _glthread_LOCK_MUTEX(buf->mutex); - pool = buf->pool; - - if (pool == NULL && newPool != NULL) { - buf->pool = newPool; - pool = newPool; - } - if (newPool == NULL) - newPool = pool; - - if (!pool->create) { - _mesa_error(NULL, GL_INVALID_OPERATION, - "driBOData called on invalid buffer\n"); - BM_CKFATAL(-EINVAL); - } - - newBuffer = (!buf->private || pool != newPool || - pool->size(pool, buf->private) < size); - - if (!flags) - flags = buf->flags; - - if (newBuffer) { - - if (buf->createdByReference) { - _mesa_error(NULL, GL_INVALID_OPERATION, - "driBOData requiring resizing called on " - "shared buffer.\n"); - BM_CKFATAL(-EINVAL); - } - - if (buf->private) - buf->pool->destroy(buf->pool, buf->private); - - pool = newPool; - buf->pool = newPool; - buf->private = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, - buf->alignment); - if (!buf->private) - retval = -ENOMEM; - - if (retval == 0) - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, - DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual); - } else if (pool->map(pool, buf->private, DRM_BO_FLAG_WRITE, - DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual)) { - /* - * Buffer is busy. need to create a new one. - */ - - void *newBuf; - - newBuf = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, - buf->alignment); - if (newBuf) { - buf->pool->destroy(buf->pool, buf->private); - buf->private = newBuf; - } - - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); - } else { - uint64_t flag_diff = flags ^ buf->flags; - - /* - * We might need to change buffer flags. - */ - - if (flag_diff){ - assert(pool->setStatus != NULL); - BM_CKFATAL(pool->unmap(pool, buf->private)); - BM_CKFATAL(pool->setStatus(pool, buf->private, flag_diff, - buf->flags)); - if (!data) - goto out; - - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); - } - } - - if (retval == 0) { - if (data) - memcpy(virtual, data, size); - - BM_CKFATAL(pool->unmap(pool, buf->private)); - } - - out: - _glthread_UNLOCK_MUTEX(buf->mutex); - - return retval; -} - -void -driBOSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, const void *data) -{ - void *virtual; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - _glthread_LOCK_MUTEX(buf->mutex); - if (size && data) { - BM_CKFATAL(buf->pool->map(buf->pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, - &virtual)); - memcpy((unsigned char *) virtual + offset, data, size); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - } - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -void -driBOGetSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, void *data) -{ - void *virtual; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - _glthread_LOCK_MUTEX(buf->mutex); - if (size && data) { - BM_CKFATAL(buf->pool->map(buf->pool, buf->private, - DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual)); - memcpy(data, (unsigned char *) virtual + offset, size); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - } - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -void -driBOSetReferenced(struct _DriBufferObject *buf, - unsigned long handle) -{ - _glthread_LOCK_MUTEX(buf->mutex); - if (buf->private != NULL) { - _mesa_error(NULL, GL_INVALID_OPERATION, - "Invalid buffer for setReferenced\n"); - BM_CKFATAL(-EINVAL); - - } - if (buf->pool->reference == NULL) { - _mesa_error(NULL, GL_INVALID_OPERATION, - "Invalid buffer pool for setReferenced\n"); - BM_CKFATAL(-EINVAL); - } - buf->private = buf->pool->reference(buf->pool, handle); - if (!buf->private) { - _mesa_error(NULL, GL_OUT_OF_MEMORY, - "Invalid buffer pool for setStatic\n"); - BM_CKFATAL(-ENOMEM); - } - buf->createdByReference = GL_TRUE; - buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags; - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -int -driGenBuffers(struct _DriBufferPool *pool, - const char *name, - unsigned n, - struct _DriBufferObject *buffers[], - unsigned alignment, uint64_t flags, unsigned hint) -{ - struct _DriBufferObject *buf; - int i; - - flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM | - DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; - - ++num_buffers; - - assert(pool); - - for (i = 0; i < n; ++i) { - buf = (struct _DriBufferObject *) calloc(1, sizeof(*buf)); - if (!buf) - return -ENOMEM; - - _glthread_INIT_MUTEX(buf->mutex); - _glthread_LOCK_MUTEX(buf->mutex); - buf->refCount = 1; - buf->flags = flags; - buf->hint = hint; - buf->name = name; - buf->alignment = alignment; - buf->pool = pool; - buf->createdByReference = 0; - _glthread_UNLOCK_MUTEX(buf->mutex); - buffers[i] = buf; - } - return 0; -} - -void -driGenUserBuffer(struct _DriBufferPool *pool, - const char *name, - struct _DriBufferObject **buffers, - void *ptr, unsigned bytes) -{ - const unsigned alignment = 1, flags = 0, hint = 0; - - --num_buffers; /* JB: is inced in GenBuffes */ - driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); - ++num_user_buffers; - - (*buffers)->userBuffer = 1; - (*buffers)->userData = ptr; - (*buffers)->userSize = bytes; -} - -void -driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]) -{ - int i; - - for (i = 0; i < n; ++i) { - driBOUnReference(buffers[i]); - } -} - - -void -driInitBufMgr(int fd) -{ - ; -} - -/* - * Note that lists are per-context and don't need mutex protection. - */ - -struct _DriBufferList * -driBOCreateList(int target) -{ - struct _DriBufferList *list = calloc(sizeof(*list), 1); - - BM_CKFATAL(drmBOCreateList(target, &list->drmBuffers)); - BM_CKFATAL(drmBOCreateList(target, &list->driBuffers)); - return list; -} - -int -driBOResetList(struct _DriBufferList * list) -{ - int ret; - ret = drmBOResetList(&list->drmBuffers); - if (ret) - return ret; - ret = drmBOResetList(&list->driBuffers); - return ret; -} - -void -driBOFreeList(struct _DriBufferList * list) -{ - drmBOFreeList(&list->drmBuffers); - drmBOFreeList(&list->driBuffers); - free(list); -} - - -/* - * Copied from libdrm, because it is needed by driAddValidateItem. - */ - -static drmBONode * -driAddListItem(drmBOList * list, drmBO * item, - uint64_t arg0, uint64_t arg1) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->free.next; - if (l == &list->free) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - return NULL; - } - list->numCurrent++; - } else { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - } - memset(&node->bo_arg, 0, sizeof(node->bo_arg)); - node->buf = item; - node->arg0 = arg0; - node->arg1 = arg1; - DRMLISTADDTAIL(&node->head, &list->list); - list->numOnList++; - return node; -} - -/* - * Slightly modified version compared to the libdrm version. - * This one returns the list index of the buffer put on the list. - */ - -static int -driAddValidateItem(drmBOList * list, drmBO * buf, uint64_t flags, - uint64_t mask, int *itemLoc, - struct _drmBONode **pnode) -{ - drmBONode *node, *cur; - drmMMListHead *l; - int count = 0; - - cur = NULL; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - if (node->buf == buf) { - cur = node; - break; - } - count++; - } - if (!cur) { - cur = driAddListItem(list, buf, flags, mask); - if (!cur) - return -ENOMEM; - - cur->arg0 = flags; - cur->arg1 = mask; - } else { - uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; - uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; - - if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { - return -EINVAL; - } - - cur->arg1 |= mask; - cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); - - if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && - (cur->arg0 & DRM_BO_MASK_MEM) == 0) { - return -EINVAL; - } - } - *itemLoc = count; - *pnode = cur; - return 0; -} - - -void -driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf, - uint64_t flags, uint64_t mask, int *itemLoc, - struct _drmBONode **node) -{ - int newItem; - - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(driAddValidateItem(&list->drmBuffers, - buf->pool->kernel(buf->pool, buf->private), - flags, mask, itemLoc, node)); - BM_CKFATAL(drmAddValidateItem(&list->driBuffers, (drmBO *) buf, - flags, mask, &newItem)); - if (newItem) - buf->refCount++; - - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -drmBOList *driGetdrmBOList(struct _DriBufferList *list) -{ - driWriteLockKernelBO(); - return &list->drmBuffers; -} - -void driPutdrmBOList(struct _DriBufferList *list) -{ - driWriteUnlockKernelBO(); -} - - -void -driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence) -{ - _glthread_LOCK_MUTEX(buf->mutex); - if (buf->pool->fence) - BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence)); - _glthread_UNLOCK_MUTEX(buf->mutex); - -} - -void -driBOUnrefUserList(struct _DriBufferList *list) -{ - struct _DriBufferObject *buf; - void *curBuf; - - curBuf = drmBOListIterator(&list->driBuffers); - while (curBuf) { - buf = (struct _DriBufferObject *)drmBOListBuf(curBuf); - driBOUnReference(buf); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } -} - -struct _DriFenceObject * -driBOFenceUserList(struct _DriFenceMgr *mgr, - struct _DriBufferList *list, const char *name, - drmFence *kFence) -{ - struct _DriFenceObject *fence; - struct _DriBufferObject *buf; - void *curBuf; - - fence = driFenceCreate(mgr, kFence->fence_class, kFence->type, - kFence, sizeof(*kFence)); - curBuf = drmBOListIterator(&list->driBuffers); - - /* - * User-space fencing callbacks. - */ - - while (curBuf) { - buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); - driBOFence(buf, fence); - driBOUnReference(buf); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } - - driBOResetList(list); - return fence; -} - -void -driBOValidateUserList(struct _DriBufferList * list) -{ - void *curBuf; - struct _DriBufferObject *buf; - - curBuf = drmBOListIterator(&list->driBuffers); - - /* - * User-space validation callbacks. - */ - - while (curBuf) { - buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); - _glthread_LOCK_MUTEX(buf->mutex); - if (buf->pool->validate) - BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex)); - _glthread_UNLOCK_MUTEX(buf->mutex); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } -} - - -void -driPoolTakeDown(struct _DriBufferPool *pool) -{ - pool->takeDown(pool); - -} - -unsigned long -driBOSize(struct _DriBufferObject *buf) -{ - unsigned long size; - - _glthread_LOCK_MUTEX(buf->mutex); - size = buf->pool->size(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - - return size; - -} - -drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list) -{ - return &list->drmBuffers; -} - -drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list) -{ - return &list->driBuffers; -} - diff --git a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.h b/src/gallium/winsys/dri/intel/ws_dri_bufmgr.h deleted file mode 100644 index fdaf5ee93a..0000000000 --- a/src/gallium/winsys/dri/intel/ws_dri_bufmgr.h +++ /dev/null @@ -1,138 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - * Keith Whitwell - */ - -#ifndef _PSB_BUFMGR_H_ -#define _PSB_BUFMGR_H_ -#include -#include "i915_drm.h" -#include "ws_dri_fencemgr.h" - -typedef struct _drmBONode -{ - drmMMListHead head; - drmBO *buf; - struct drm_i915_op_arg bo_arg; - uint64_t arg0; - uint64_t arg1; -} drmBONode; - -typedef struct _drmBOList { - unsigned numTarget; - unsigned numCurrent; - unsigned numOnList; - drmMMListHead list; - drmMMListHead free; -} drmBOList; - - -struct _DriFenceObject; -struct _DriBufferObject; -struct _DriBufferPool; -struct _DriBufferList; - -/* - * Return a pointer to the libdrm buffer object this DriBufferObject - * uses. - */ - -extern drmBO *driBOKernel(struct _DriBufferObject *buf); -extern void *driBOMap(struct _DriBufferObject *buf, unsigned flags, - unsigned hint); -extern void driBOUnmap(struct _DriBufferObject *buf); -extern unsigned long driBOOffset(struct _DriBufferObject *buf); -extern unsigned long driBOPoolOffset(struct _DriBufferObject *buf); - -extern uint64_t driBOFlags(struct _DriBufferObject *buf); -extern struct _DriBufferObject *driBOReference(struct _DriBufferObject *buf); -extern void driBOUnReference(struct _DriBufferObject *buf); - -extern int driBOData(struct _DriBufferObject *r_buf, - unsigned size, const void *data, - struct _DriBufferPool *pool, uint64_t flags); - -extern void driBOSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, - const void *data); -extern void driBOGetSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, - void *data); -extern int driGenBuffers(struct _DriBufferPool *pool, - const char *name, - unsigned n, - struct _DriBufferObject *buffers[], - unsigned alignment, uint64_t flags, unsigned hint); -extern void driGenUserBuffer(struct _DriBufferPool *pool, - const char *name, - struct _DriBufferObject *buffers[], - void *ptr, unsigned bytes); -extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]); -extern void driInitBufMgr(int fd); -extern struct _DriBufferList *driBOCreateList(int target); -extern int driBOResetList(struct _DriBufferList * list); -extern void driBOAddListItem(struct _DriBufferList * list, - struct _DriBufferObject *buf, - uint64_t flags, uint64_t mask, int *itemLoc, - struct _drmBONode **node); - -extern void driBOValidateList(int fd, struct _DriBufferList * list); -extern void driBOFreeList(struct _DriBufferList * list); -extern struct _DriFenceObject *driBOFenceUserList(struct _DriFenceMgr *mgr, - struct _DriBufferList *list, - const char *name, - drmFence *kFence); -extern void driBOUnrefUserList(struct _DriBufferList *list); -extern void driBOValidateUserList(struct _DriBufferList * list); -extern drmBOList *driGetdrmBOList(struct _DriBufferList *list); -extern void driPutdrmBOList(struct _DriBufferList *list); - -extern void driBOFence(struct _DriBufferObject *buf, - struct _DriFenceObject *fence); - -extern void driPoolTakeDown(struct _DriBufferPool *pool); -extern void driBOSetReferenced(struct _DriBufferObject *buf, - unsigned long handle); -unsigned long driBOSize(struct _DriBufferObject *buf); -extern void driBOWaitIdle(struct _DriBufferObject *buf, int lazy); -extern void driPoolTakeDown(struct _DriBufferPool *pool); - -extern void driReadLockKernelBO(void); -extern void driReadUnlockKernelBO(void); -extern void driWriteLockKernelBO(void); -extern void driWriteUnlockKernelBO(void); - -/* - * For debugging purposes. - */ - -extern drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list); -extern drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list); -#endif diff --git a/src/gallium/winsys/dri/intel/ws_dri_bufpool.h b/src/gallium/winsys/dri/intel/ws_dri_bufpool.h deleted file mode 100644 index 3a302e13d3..0000000000 --- a/src/gallium/winsys/dri/intel/ws_dri_bufpool.h +++ /dev/null @@ -1,102 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#ifndef _PSB_BUFPOOL_H_ -#define _PSB_BUFPOOL_H_ - -#include -#include -struct _DriFenceObject; - -typedef struct _DriBufferPool -{ - int fd; - int (*map) (struct _DriBufferPool * pool, void *private, - unsigned flags, int hint, _glthread_Mutex *mutex, - void **virtual); - int (*unmap) (struct _DriBufferPool * pool, void *private); - int (*destroy) (struct _DriBufferPool * pool, void *private); - unsigned long (*offset) (struct _DriBufferPool * pool, void *private); - unsigned long (*poolOffset) (struct _DriBufferPool * pool, void *private); - uint64_t (*flags) (struct _DriBufferPool * pool, void *private); - unsigned long (*size) (struct _DriBufferPool * pool, void *private); - void *(*create) (struct _DriBufferPool * pool, unsigned long size, - uint64_t flags, unsigned hint, unsigned alignment); - void *(*reference) (struct _DriBufferPool * pool, unsigned handle); - int (*unreference) (struct _DriBufferPool * pool, void *private); - int (*fence) (struct _DriBufferPool * pool, void *private, - struct _DriFenceObject * fence); - drmBO *(*kernel) (struct _DriBufferPool * pool, void *private); - int (*validate) (struct _DriBufferPool * pool, void *private, _glthread_Mutex *mutex); - int (*waitIdle) (struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, - int lazy); - int (*setStatus) (struct _DriBufferPool *pool, void *private, - uint64_t flag_diff, uint64_t old_flags); - void (*takeDown) (struct _DriBufferPool * pool); - void *data; -} DriBufferPool; - -extern void bmError(int val, const char *file, const char *function, - int line); -#define BM_CKFATAL(val) \ - do{ \ - int tstVal = (val); \ - if (tstVal) \ - bmError(tstVal, __FILE__, __FUNCTION__, __LINE__); \ - } while(0); - - -/* - * Builtin pools. - */ - -/* - * Kernel buffer objects. Size in multiples of page size. Page size aligned. - */ - -extern struct _DriBufferPool *driDRMPoolInit(int fd); -extern struct _DriBufferPool *driMallocPoolInit(void); - -struct _DriFreeSlabManager; -extern struct _DriBufferPool * driSlabPoolInit(int fd, uint64_t flags, - uint64_t validMask, - uint32_t smallestSize, - uint32_t numSizes, - uint32_t desiredNumBuffers, - uint32_t maxSlabSize, - uint32_t pageAlignment, - struct _DriFreeSlabManager *fMan); -extern void driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan); -extern struct _DriFreeSlabManager * -driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec); - - -#endif diff --git a/src/gallium/winsys/dri/intel/ws_dri_drmpool.c b/src/gallium/winsys/dri/intel/ws_dri_drmpool.c deleted file mode 100644 index 7c55dbc674..0000000000 --- a/src/gallium/winsys/dri/intel/ws_dri_drmpool.c +++ /dev/null @@ -1,268 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#include -#include -#include -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" -#include "assert.h" - -/* - * Buffer pool implementation using DRM buffer objects as DRI buffer objects. - */ - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, uint64_t flags, unsigned hint, - unsigned alignment) -{ - drmBO *buf = (drmBO *) malloc(sizeof(*buf)); - int ret; - unsigned pageSize = getpagesize(); - - if (!buf) - return NULL; - - if ((alignment > pageSize) && (alignment % pageSize)) { - free(buf); - return NULL; - } - - ret = drmBOCreate(pool->fd, size, alignment / pageSize, - NULL, - flags, hint, buf); - if (ret) { - free(buf); - return NULL; - } - - return (void *) buf; -} - -static void * -pool_reference(struct _DriBufferPool *pool, unsigned handle) -{ - drmBO *buf = (drmBO *) malloc(sizeof(*buf)); - int ret; - - if (!buf) - return NULL; - - ret = drmBOReference(pool->fd, handle, buf); - - if (ret) { - free(buf); - return NULL; - } - - return (void *) buf; -} - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - int ret; - drmBO *buf = (drmBO *) private; - driReadLockKernelBO(); - ret = drmBOUnreference(pool->fd, buf); - free(buf); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_unreference(struct _DriBufferPool *pool, void *private) -{ - int ret; - drmBO *buf = (drmBO *) private; - driReadLockKernelBO(); - ret = drmBOUnreference(pool->fd, buf); - free(buf); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, _glthread_Mutex *mutex, void **virtual) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOMap(pool->fd, buf, flags, hint, virtual); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOUnmap(pool->fd, buf); - driReadUnlockKernelBO(); - - return ret; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - unsigned long offset; - - driReadLockKernelBO(); - assert(buf->flags & DRM_BO_FLAG_NO_MOVE); - offset = buf->offset; - driReadUnlockKernelBO(); - - return buf->offset; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - return 0; -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - uint64_t flags; - - driReadLockKernelBO(); - flags = buf->flags; - driReadUnlockKernelBO(); - - return flags; -} - - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - unsigned long size; - - driReadLockKernelBO(); - size = buf->size; - driReadUnlockKernelBO(); - - return buf->size; -} - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - /* - * Noop. The kernel handles all fencing. - */ - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - return (drmBO *) private; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, - int lazy) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOWaitIdle(pool->fd, buf, (lazy) ? DRM_BO_HINT_WAIT_LAZY:0); - driReadUnlockKernelBO(); - - return ret; -} - - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - free(pool); -} - -/*static int -pool_setStatus(struct _DriBufferPool *pool, void *private, - uint64_t flag_diff, uint64_t old_flags) -{ - drmBO *buf = (drmBO *) private; - uint64_t new_flags = old_flags ^ flag_diff; - int ret; - - driReadLockKernelBO(); - ret = drmBOSetStatus(pool->fd, buf, new_flags, flag_diff, - 0, 0, 0); - driReadUnlockKernelBO(); - return ret; -}*/ - -struct _DriBufferPool * -driDRMPoolInit(int fd) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - - if (!pool) - return NULL; - - pool->fd = fd; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->poolOffset = &pool_poolOffset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = NULL; - pool->waitIdle = &pool_waitIdle; - pool->takeDown = &pool_takedown; - pool->reference = &pool_reference; - pool->unreference = &pool_unreference; - pool->data = NULL; - return pool; -} diff --git a/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c b/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c deleted file mode 100644 index 1f893b47ce..0000000000 --- a/src/gallium/winsys/dri/intel/ws_dri_fencemgr.c +++ /dev/null @@ -1,377 +0,0 @@ -#include "ws_dri_fencemgr.h" -#include "glthread.h" -#include -#include -#include - -/* - * Note: Locking order is - * _DriFenceObject::mutex - * _DriFenceMgr::mutex - */ - -struct _DriFenceMgr { - /* - * Constant members. Need no mutex protection. - */ - struct _DriFenceMgrCreateInfo info; - void *private; - - /* - * These members are protected by this->mutex - */ - _glthread_Mutex mutex; - int refCount; - drmMMListHead *heads; - int num_fences; -}; - -struct _DriFenceObject { - - /* - * These members are constant and need no mutex protection. - */ - struct _DriFenceMgr *mgr; - uint32_t fence_class; - uint32_t fence_type; - - /* - * These members are protected by mgr->mutex. - */ - drmMMListHead head; - int refCount; - - /* - * These members are protected by this->mutex. - */ - _glthread_Mutex mutex; - uint32_t signaled_type; - void *private; -}; - -uint32_t -driFenceType(struct _DriFenceObject *fence) -{ - return fence->fence_type; -} - -struct _DriFenceMgr * -driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info) -{ - struct _DriFenceMgr *tmp; - uint32_t i; - - tmp = calloc(1, sizeof(*tmp)); - if (!tmp) - return NULL; - - _glthread_INIT_MUTEX(tmp->mutex); - _glthread_LOCK_MUTEX(tmp->mutex); - tmp->refCount = 1; - tmp->info = *info; - tmp->num_fences = 0; - tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads)); - if (!tmp->heads) - goto out_err; - - for (i=0; iinfo.num_classes; ++i) { - DRMINITLISTHEAD(&tmp->heads[i]); - } - _glthread_UNLOCK_MUTEX(tmp->mutex); - return tmp; - - out_err: - if (tmp) - free(tmp); - return NULL; -} - -static void -driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr) -{ - struct _DriFenceMgr *mgr = *pMgr; - - *pMgr = NULL; - if (--mgr->refCount == 0) - free(mgr); - else - _glthread_UNLOCK_MUTEX(mgr->mutex); -} - -void -driFenceMgrUnReference(struct _DriFenceMgr **pMgr) -{ - _glthread_LOCK_MUTEX((*pMgr)->mutex); - driFenceMgrUnrefUnlock(pMgr); -} - -static void -driFenceUnReferenceLocked(struct _DriFenceObject **pFence) -{ - struct _DriFenceObject *fence = *pFence; - struct _DriFenceMgr *mgr = fence->mgr; - - *pFence = NULL; - if (--fence->refCount == 0) { - DRMLISTDELINIT(&fence->head); - if (fence->private) - mgr->info.unreference(mgr, &fence->private); - --mgr->num_fences; - fence->mgr = NULL; - --mgr->refCount; - free(fence); - - } -} - - -static void -driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr, - drmMMListHead *list, - uint32_t fence_class, - uint32_t fence_type) -{ - struct _DriFenceObject *entry; - drmMMListHead *prev; - - while(list != &mgr->heads[fence_class]) { - entry = DRMLISTENTRY(struct _DriFenceObject, list, head); - - /* - * Up refcount so that entry doesn't disappear from under us - * when we unlock-relock mgr to get the correct locking order. - */ - - ++entry->refCount; - _glthread_UNLOCK_MUTEX(mgr->mutex); - _glthread_LOCK_MUTEX(entry->mutex); - _glthread_LOCK_MUTEX(mgr->mutex); - - prev = list->prev; - - - - if (list->prev == list) { - - /* - * Somebody else removed the entry from the list. - */ - - _glthread_UNLOCK_MUTEX(entry->mutex); - driFenceUnReferenceLocked(&entry); - return; - } - - entry->signaled_type |= (fence_type & entry->fence_type); - if (entry->signaled_type == entry->fence_type) { - DRMLISTDELINIT(list); - mgr->info.unreference(mgr, &entry->private); - } - _glthread_UNLOCK_MUTEX(entry->mutex); - driFenceUnReferenceLocked(&entry); - list = prev; - } -} - - -int -driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, - int lazy_hint) -{ - struct _DriFenceMgr *mgr = fence->mgr; - int ret = 0; - - _glthread_LOCK_MUTEX(fence->mutex); - - if ((fence->signaled_type & fence_type) == fence_type) - goto out0; - - ret = mgr->info.finish(mgr, fence->private, fence_type, lazy_hint); - if (ret) - goto out0; - - _glthread_LOCK_MUTEX(mgr->mutex); - _glthread_UNLOCK_MUTEX(fence->mutex); - - driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, - fence_type); - _glthread_UNLOCK_MUTEX(mgr->mutex); - return 0; - - out0: - _glthread_UNLOCK_MUTEX(fence->mutex); - return ret; -} - -uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence) -{ - uint32_t ret; - - _glthread_LOCK_MUTEX(fence->mutex); - ret = fence->signaled_type; - _glthread_UNLOCK_MUTEX(fence->mutex); - - return ret; -} - -int -driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type, - uint32_t *signaled) -{ - int ret = 0; - struct _DriFenceMgr *mgr; - - _glthread_LOCK_MUTEX(fence->mutex); - mgr = fence->mgr; - *signaled = fence->signaled_type; - if ((fence->signaled_type & flush_type) == flush_type) - goto out0; - - ret = mgr->info.signaled(mgr, fence->private, flush_type, signaled); - if (ret) { - *signaled = fence->signaled_type; - goto out0; - } - - if ((fence->signaled_type | *signaled) == fence->signaled_type) - goto out0; - - _glthread_LOCK_MUTEX(mgr->mutex); - _glthread_UNLOCK_MUTEX(fence->mutex); - - driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, - *signaled); - - _glthread_UNLOCK_MUTEX(mgr->mutex); - return 0; - out0: - _glthread_UNLOCK_MUTEX(fence->mutex); - return ret; -} - -struct _DriFenceObject * -driFenceReference(struct _DriFenceObject *fence) -{ - _glthread_LOCK_MUTEX(fence->mgr->mutex); - ++fence->refCount; - _glthread_UNLOCK_MUTEX(fence->mgr->mutex); - return fence; -} - -void -driFenceUnReference(struct _DriFenceObject **pFence) -{ - struct _DriFenceMgr *mgr; - - if (*pFence == NULL) - return; - - mgr = (*pFence)->mgr; - _glthread_LOCK_MUTEX(mgr->mutex); - ++mgr->refCount; - driFenceUnReferenceLocked(pFence); - driFenceMgrUnrefUnlock(&mgr); -} - -struct _DriFenceObject -*driFenceCreate(struct _DriFenceMgr *mgr, uint32_t fence_class, - uint32_t fence_type, void *private, size_t private_size) -{ - struct _DriFenceObject *fence; - size_t fence_size = sizeof(*fence); - - if (private_size) - fence_size = ((fence_size + 15) & ~15); - - fence = calloc(1, fence_size + private_size); - - if (!fence) { - int ret = mgr->info.finish(mgr, private, fence_type, 0); - - if (ret) - usleep(10000000); - - return NULL; - } - - _glthread_INIT_MUTEX(fence->mutex); - _glthread_LOCK_MUTEX(fence->mutex); - _glthread_LOCK_MUTEX(mgr->mutex); - fence->refCount = 1; - DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]); - fence->mgr = mgr; - ++mgr->refCount; - ++mgr->num_fences; - _glthread_UNLOCK_MUTEX(mgr->mutex); - fence->fence_class = fence_class; - fence->fence_type = fence_type; - fence->signaled_type = 0; - fence->private = private; - if (private_size) { - fence->private = (void *)(((uint8_t *) fence) + fence_size); - memcpy(fence->private, private, private_size); - } - - _glthread_UNLOCK_MUTEX(fence->mutex); - return fence; -} - - -static int -tSignaled(struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, - uint32_t *signaled_type) -{ - long fd = (long) mgr->private; - int dummy; - drmFence *fence = (drmFence *) private; - int ret; - - *signaled_type = 0; - ret = drmFenceSignaled((int) fd, fence, flush_type, &dummy); - if (ret) - return ret; - - *signaled_type = fence->signaled; - - return 0; -} - -static int -tFinish(struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, - int lazy_hint) -{ - long fd = (long) mgr->private; - unsigned flags = lazy_hint ? DRM_FENCE_FLAG_WAIT_LAZY : 0; - - return drmFenceWait((int)fd, flags, (drmFence *) private, fence_type); -} - -static int -tUnref(struct _DriFenceMgr *mgr, void **private) -{ - long fd = (long) mgr->private; - drmFence *fence = (drmFence *) *private; - *private = NULL; - - return drmFenceUnreference(fd, fence); -} - -struct _DriFenceMgr *driFenceMgrTTMInit(int fd) -{ - struct _DriFenceMgrCreateInfo info; - struct _DriFenceMgr *mgr; - - info.flags = DRI_FENCE_CLASS_ORDERED; - info.num_classes = 4; - info.signaled = tSignaled; - info.finish = tFinish; - info.unreference = tUnref; - - mgr = driFenceMgrCreate(&info); - if (mgr == NULL) - return NULL; - - mgr->private = (void *) (long) fd; - return mgr; -} - diff --git a/src/gallium/winsys/dri/intel/ws_dri_fencemgr.h b/src/gallium/winsys/dri/intel/ws_dri_fencemgr.h deleted file mode 100644 index 4ea58dfe18..0000000000 --- a/src/gallium/winsys/dri/intel/ws_dri_fencemgr.h +++ /dev/null @@ -1,115 +0,0 @@ -#ifndef DRI_FENCEMGR_H -#define DRI_FENCEMGR_H - -#include -#include - -struct _DriFenceObject; -struct _DriFenceMgr; - -/* - * Do a quick check to see if the fence manager has registered the fence - * object as signaled. Note that this function may return a false negative - * answer. - */ -extern uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence); - -/* - * Check if the fence object is signaled. This function can be substantially - * more expensive to call than the above function, but will not return a false - * negative answer. The argument "flush_type" sets the types that the - * underlying mechanism must make sure will eventually signal. - */ -extern int driFenceSignaledType(struct _DriFenceObject *fence, - uint32_t flush_type, uint32_t *signaled); - -/* - * Convenience functions. - */ - -static inline int driFenceSignaled(struct _DriFenceObject *fence, - uint32_t flush_type) -{ - uint32_t signaled_types; - int ret = driFenceSignaledType(fence, flush_type, &signaled_types); - if (ret) - return 0; - return ((signaled_types & flush_type) == flush_type); -} - -static inline int driFenceSignaledCached(struct _DriFenceObject *fence, - uint32_t flush_type) -{ - uint32_t signaled_types = - driFenceSignaledTypeCached(fence); - - return ((signaled_types & flush_type) == flush_type); -} - -/* - * Reference a fence object. - */ -extern struct _DriFenceObject *driFenceReference(struct _DriFenceObject *fence); - -/* - * Unreference a fence object. The fence object pointer will be reset to NULL. - */ - -extern void driFenceUnReference(struct _DriFenceObject **pFence); - - -/* - * Wait for a fence to signal the indicated fence_type. - * If "lazy_hint" is true, it indicates that the wait may sleep to avoid - * busy-wait polling. - */ -extern int driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, - int lazy_hint); - -/* - * Create a DriFenceObject for manager "mgr". - * - * "private" is a pointer that should be used for the callbacks in - * struct _DriFenceMgrCreateInfo. - * - * if private_size is nonzero, then the info stored at *private, with size - * private size will be copied and the fence manager will instead use a - * pointer to the copied data for the callbacks in - * struct _DriFenceMgrCreateInfo. In that case, the object pointed to by - * "private" may be destroyed after the call to driFenceCreate. - */ -extern struct _DriFenceObject *driFenceCreate(struct _DriFenceMgr *mgr, - uint32_t fence_class, - uint32_t fence_type, - void *private, - size_t private_size); - -extern uint32_t driFenceType(struct _DriFenceObject *fence); - -/* - * Fence creations are ordered. If a fence signals a fence_type, - * it is safe to assume that all fences of the same class that was - * created before that fence has signaled the same type. - */ - -#define DRI_FENCE_CLASS_ORDERED (1 << 0) - -struct _DriFenceMgrCreateInfo { - uint32_t flags; - uint32_t num_classes; - int (*signaled) (struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, - uint32_t *signaled_type); - int (*finish) (struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, int lazy_hint); - int (*unreference) (struct _DriFenceMgr *mgr, void **private); -}; - -extern struct _DriFenceMgr * -driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info); - -void -driFenceMgrUnReference(struct _DriFenceMgr **pMgr); - -extern struct _DriFenceMgr * -driFenceMgrTTMInit(int fd); - -#endif diff --git a/src/gallium/winsys/dri/intel/ws_dri_mallocpool.c b/src/gallium/winsys/dri/intel/ws_dri_mallocpool.c deleted file mode 100644 index bf97d7e440..0000000000 --- a/src/gallium/winsys/dri/intel/ws_dri_mallocpool.c +++ /dev/null @@ -1,162 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, TX., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#include -#include -#include -#include "imports.h" -#include "glthread.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" -#include "intel_screen.h" - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, uint64_t flags, unsigned hint, - unsigned alignment) -{ - unsigned long *private = malloc(size + 2*sizeof(unsigned long)); - if ((flags & DRM_BO_MASK_MEM) != DRM_BO_FLAG_MEM_LOCAL) - abort(); - - *private = size; - return (void *)private; -} - - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - free(private); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, - _glthread_Mutex *mutex, int lazy) -{ - return 0; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, _glthread_Mutex *mutex, void **virtual) -{ - *virtual = (void *)((unsigned long *)private + 2); - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - /* - * BUG - */ - abort(); - return 0UL; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - /* - * BUG - */ - abort(); -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - return DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - return *(unsigned long *) private; -} - - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - abort(); - return 0UL; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - abort(); - return NULL; -} - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - free(pool); -} - - -struct _DriBufferPool * -driMallocPoolInit(void) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - if (!pool) - return NULL; - - pool->data = NULL; - pool->fd = -1; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->poolOffset = &pool_poolOffset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = NULL; - pool->waitIdle = &pool_waitIdle; - pool->takeDown = &pool_takedown; - return pool; -} diff --git a/src/gallium/winsys/dri/intel/ws_dri_slabpool.c b/src/gallium/winsys/dri/intel/ws_dri_slabpool.c deleted file mode 100644 index 62d82bbd94..0000000000 --- a/src/gallium/winsys/dri/intel/ws_dri_slabpool.c +++ /dev/null @@ -1,968 +0,0 @@ -/************************************************************************** - * - * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellstrom - */ - -#include -#include -#include -#include -#include -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" -#include "ws_dri_bufmgr.h" -#include "glthread.h" - -#define DRI_SLABPOOL_ALLOC_RETRIES 100 - -struct _DriSlab; - -struct _DriSlabBuffer { - int isSlabBuffer; - drmBO *bo; - struct _DriFenceObject *fence; - struct _DriSlab *parent; - drmMMListHead head; - uint32_t mapCount; - uint32_t start; - uint32_t fenceType; - int unFenced; - _glthread_Cond event; -}; - -struct _DriKernelBO { - int fd; - drmBO bo; - drmMMListHead timeoutHead; - drmMMListHead head; - struct timeval timeFreed; - uint32_t pageAlignment; - void *virtual; -}; - -struct _DriSlab{ - drmMMListHead head; - drmMMListHead freeBuffers; - uint32_t numBuffers; - uint32_t numFree; - struct _DriSlabBuffer *buffers; - struct _DriSlabSizeHeader *header; - struct _DriKernelBO *kbo; -}; - - -struct _DriSlabSizeHeader { - drmMMListHead slabs; - drmMMListHead freeSlabs; - drmMMListHead delayedBuffers; - uint32_t numDelayed; - struct _DriSlabPool *slabPool; - uint32_t bufSize; - _glthread_Mutex mutex; -}; - -struct _DriFreeSlabManager { - struct timeval slabTimeout; - struct timeval checkInterval; - struct timeval nextCheck; - drmMMListHead timeoutList; - drmMMListHead unCached; - drmMMListHead cached; - _glthread_Mutex mutex; -}; - - -struct _DriSlabPool { - - /* - * The data of this structure remains constant after - * initialization and thus needs no mutex protection. - */ - - struct _DriFreeSlabManager *fMan; - uint64_t proposedFlags; - uint64_t validMask; - uint32_t *bucketSizes; - uint32_t numBuckets; - uint32_t pageSize; - int fd; - int pageAlignment; - int maxSlabSize; - int desiredNumBuffers; - struct _DriSlabSizeHeader *headers; -}; - -/* - * FIXME: Perhaps arrange timeout slabs in size buckets for fast - * retreival?? - */ - - -static inline int -driTimeAfterEq(struct timeval *arg1, struct timeval *arg2) -{ - return ((arg1->tv_sec > arg2->tv_sec) || - ((arg1->tv_sec == arg2->tv_sec) && - (arg1->tv_usec > arg2->tv_usec))); -} - -static inline void -driTimeAdd(struct timeval *arg, struct timeval *add) -{ - unsigned int sec; - - arg->tv_sec += add->tv_sec; - arg->tv_usec += add->tv_usec; - sec = arg->tv_usec / 1000000; - arg->tv_sec += sec; - arg->tv_usec -= sec*1000000; -} - -static void -driFreeKernelBO(struct _DriKernelBO *kbo) -{ - if (!kbo) - return; - - (void) drmBOUnreference(kbo->fd, &kbo->bo); - free(kbo); -} - - -static void -driFreeTimeoutKBOsLocked(struct _DriFreeSlabManager *fMan, - struct timeval *time) -{ - drmMMListHead *list, *next; - struct _DriKernelBO *kbo; - - if (!driTimeAfterEq(time, &fMan->nextCheck)) - return; - - for (list = fMan->timeoutList.next, next = list->next; - list != &fMan->timeoutList; - list = next, next = list->next) { - - kbo = DRMLISTENTRY(struct _DriKernelBO, list, timeoutHead); - - if (!driTimeAfterEq(time, &kbo->timeFreed)) - break; - - DRMLISTDELINIT(&kbo->timeoutHead); - DRMLISTDELINIT(&kbo->head); - driFreeKernelBO(kbo); - } - - fMan->nextCheck = *time; - driTimeAdd(&fMan->nextCheck, &fMan->checkInterval); -} - - -/* - * Add a _DriKernelBO to the free slab manager. - * This means that it is available for reuse, but if it's not - * reused in a while, it will be freed. - */ - -static void -driSetKernelBOFree(struct _DriFreeSlabManager *fMan, - struct _DriKernelBO *kbo) -{ - struct timeval time; - - _glthread_LOCK_MUTEX(fMan->mutex); - gettimeofday(&time, NULL); - driTimeAdd(&time, &fMan->slabTimeout); - - kbo->timeFreed = time; - - if (kbo->bo.flags & DRM_BO_FLAG_CACHED) - DRMLISTADD(&kbo->head, &fMan->cached); - else - DRMLISTADD(&kbo->head, &fMan->unCached); - - DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList); - driFreeTimeoutKBOsLocked(fMan, &time); - - _glthread_UNLOCK_MUTEX(fMan->mutex); -} - -/* - * Get a _DriKernelBO for us to use as storage for a slab. - * - */ - -static struct _DriKernelBO * -driAllocKernelBO(struct _DriSlabSizeHeader *header) - -{ - struct _DriSlabPool *slabPool = header->slabPool; - struct _DriFreeSlabManager *fMan = slabPool->fMan; - drmMMListHead *list, *next, *head; - uint32_t size = header->bufSize * slabPool->desiredNumBuffers; - struct _DriKernelBO *kbo; - struct _DriKernelBO *kboTmp; - int ret; - - /* - * FIXME: We should perhaps allow some variation in slabsize in order - * to efficiently reuse slabs. - */ - - size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize; - size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1); - _glthread_LOCK_MUTEX(fMan->mutex); - - kbo = NULL; - - retry: - head = (slabPool->proposedFlags & DRM_BO_FLAG_CACHED) ? - &fMan->cached : &fMan->unCached; - - for (list = head->next, next = list->next; - list != head; - list = next, next = list->next) { - - kboTmp = DRMLISTENTRY(struct _DriKernelBO, list, head); - - if ((kboTmp->bo.size == size) && - (slabPool->pageAlignment == 0 || - (kboTmp->pageAlignment % slabPool->pageAlignment) == 0)) { - - if (!kbo) - kbo = kboTmp; - - if ((kbo->bo.proposedFlags ^ slabPool->proposedFlags) == 0) - break; - - } - } - - if (kbo) { - DRMLISTDELINIT(&kbo->head); - DRMLISTDELINIT(&kbo->timeoutHead); - } - - _glthread_UNLOCK_MUTEX(fMan->mutex); - - if (kbo) { - uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags; - - ret = 0; - if (new_mask) { - ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, - new_mask, DRM_BO_HINT_DONT_FENCE, 0, 0); - } - if (ret == 0) - return kbo; - - driFreeKernelBO(kbo); - kbo = NULL; - goto retry; - } - - kbo = calloc(1, sizeof(struct _DriKernelBO)); - if (!kbo) - return NULL; - - kbo->fd = slabPool->fd; - DRMINITLISTHEAD(&kbo->head); - DRMINITLISTHEAD(&kbo->timeoutHead); - ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL, - slabPool->proposedFlags, - DRM_BO_HINT_DONT_FENCE, &kbo->bo); - if (ret) - goto out_err0; - - ret = drmBOMap(kbo->fd, &kbo->bo, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0, &kbo->virtual); - - if (ret) - goto out_err1; - - ret = drmBOUnmap(kbo->fd, &kbo->bo); - if (ret) - goto out_err1; - - return kbo; - - out_err1: - drmBOUnreference(kbo->fd, &kbo->bo); - out_err0: - free(kbo); - return NULL; -} - - -static int -driAllocSlab(struct _DriSlabSizeHeader *header) -{ - struct _DriSlab *slab; - struct _DriSlabBuffer *buf; - uint32_t numBuffers; - int ret; - int i; - - slab = calloc(1, sizeof(*slab)); - if (!slab) - return -ENOMEM; - - slab->kbo = driAllocKernelBO(header); - if (!slab->kbo) { - ret = -ENOMEM; - goto out_err0; - } - - numBuffers = slab->kbo->bo.size / header->bufSize; - - slab->buffers = calloc(numBuffers, sizeof(*slab->buffers)); - if (!slab->buffers) { - ret = -ENOMEM; - goto out_err1; - } - - DRMINITLISTHEAD(&slab->head); - DRMINITLISTHEAD(&slab->freeBuffers); - slab->numBuffers = numBuffers; - slab->numFree = 0; - slab->header = header; - - buf = slab->buffers; - for (i=0; i < numBuffers; ++i) { - buf->parent = slab; - buf->start = i* header->bufSize; - buf->mapCount = 0; - buf->isSlabBuffer = 1; - _glthread_INIT_COND(buf->event); - DRMLISTADDTAIL(&buf->head, &slab->freeBuffers); - slab->numFree++; - buf++; - } - - DRMLISTADDTAIL(&slab->head, &header->slabs); - - return 0; - - out_err1: - driSetKernelBOFree(header->slabPool->fMan, slab->kbo); - free(slab->buffers); - out_err0: - free(slab); - return ret; -} - -/* - * Delete a buffer from the slab header delayed list and put - * it on the slab free list. - */ - -static void -driSlabFreeBufferLocked(struct _DriSlabBuffer *buf) -{ - struct _DriSlab *slab = buf->parent; - struct _DriSlabSizeHeader *header = slab->header; - drmMMListHead *list = &buf->head; - - DRMLISTDEL(list); - DRMLISTADDTAIL(list, &slab->freeBuffers); - slab->numFree++; - - if (slab->head.next == &slab->head) - DRMLISTADDTAIL(&slab->head, &header->slabs); - - if (slab->numFree == slab->numBuffers) { - list = &slab->head; - DRMLISTDEL(list); - DRMLISTADDTAIL(list, &header->freeSlabs); - } - - if (header->slabs.next == &header->slabs || - slab->numFree != slab->numBuffers) { - - drmMMListHead *next; - struct _DriFreeSlabManager *fMan = header->slabPool->fMan; - - for (list = header->freeSlabs.next, next = list->next; - list != &header->freeSlabs; - list = next, next = list->next) { - - slab = DRMLISTENTRY(struct _DriSlab, list, head); - - DRMLISTDELINIT(list); - driSetKernelBOFree(fMan, slab->kbo); - free(slab->buffers); - free(slab); - } - } -} - -static void -driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait) -{ - drmMMListHead *list, *prev, *first; - struct _DriSlabBuffer *buf; - struct _DriSlab *slab; - int firstWasSignaled = 1; - int signaled; - int i; - int ret; - - /* - * Rerun the freeing test if the youngest tested buffer - * was signaled, since there might be more idle buffers - * in the delay list. - */ - - while (firstWasSignaled) { - firstWasSignaled = 0; - signaled = 0; - first = header->delayedBuffers.next; - - /* Only examine the oldest 1/3 of delayed buffers: - */ - if (header->numDelayed > 3) { - for (i = 0; i < header->numDelayed; i += 3) { - first = first->next; - } - } - - for (list = first, prev = list->prev; - list != &header->delayedBuffers; - list = prev, prev = list->prev) { - buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); - slab = buf->parent; - - if (!signaled) { - if (wait) { - ret = driFenceFinish(buf->fence, buf->fenceType, 0); - if (ret) - break; - signaled = 1; - wait = 0; - } else { - signaled = driFenceSignaled(buf->fence, buf->fenceType); - } - if (signaled) { - if (list == first) - firstWasSignaled = 1; - driFenceUnReference(&buf->fence); - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { - driFenceUnReference(&buf->fence); - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - } - } -} - - -static struct _DriSlabBuffer * -driSlabAllocBuffer(struct _DriSlabSizeHeader *header) -{ - static struct _DriSlabBuffer *buf; - struct _DriSlab *slab; - drmMMListHead *list; - int count = DRI_SLABPOOL_ALLOC_RETRIES; - - _glthread_LOCK_MUTEX(header->mutex); - while(header->slabs.next == &header->slabs && count > 0) { - driSlabCheckFreeLocked(header, 0); - if (header->slabs.next != &header->slabs) - break; - - _glthread_UNLOCK_MUTEX(header->mutex); - if (count != DRI_SLABPOOL_ALLOC_RETRIES) - usleep(1); - _glthread_LOCK_MUTEX(header->mutex); - (void) driAllocSlab(header); - count--; - } - - list = header->slabs.next; - if (list == &header->slabs) { - _glthread_UNLOCK_MUTEX(header->mutex); - return NULL; - } - slab = DRMLISTENTRY(struct _DriSlab, list, head); - if (--slab->numFree == 0) - DRMLISTDELINIT(list); - - list = slab->freeBuffers.next; - DRMLISTDELINIT(list); - - _glthread_UNLOCK_MUTEX(header->mutex); - buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); - return buf; -} - -static void * -pool_create(struct _DriBufferPool *driPool, unsigned long size, - uint64_t flags, unsigned hint, unsigned alignment) -{ - struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; - struct _DriSlabSizeHeader *header; - struct _DriSlabBuffer *buf; - void *dummy; - int i; - int ret; - - /* - * FIXME: Check for compatibility. - */ - - header = pool->headers; - for (i=0; inumBuckets; ++i) { - if (header->bufSize >= size) - break; - header++; - } - - if (i < pool->numBuckets) - return driSlabAllocBuffer(header); - - - /* - * Fall back to allocate a buffer object directly from DRM. - * and wrap it in a driBO structure. - */ - - - buf = calloc(1, sizeof(*buf)); - - if (!buf) - return NULL; - - buf->bo = calloc(1, sizeof(*buf->bo)); - if (!buf->bo) - goto out_err0; - - if (alignment) { - if ((alignment < pool->pageSize) && (pool->pageSize % alignment)) - goto out_err1; - if ((alignment > pool->pageSize) && (alignment % pool->pageSize)) - goto out_err1; - } - - ret = drmBOCreate(pool->fd, size, alignment / pool->pageSize, NULL, - flags, hint, buf->bo); - if (ret) - goto out_err1; - - ret = drmBOMap(pool->fd, buf->bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0, &dummy); - if (ret) - goto out_err2; - - ret = drmBOUnmap(pool->fd, buf->bo); - if (ret) - goto out_err2; - - return buf; - out_err2: - drmBOUnreference(pool->fd, buf->bo); - out_err1: - free(buf->bo); - out_err0: - free(buf); - return NULL; -} - -static int -pool_destroy(struct _DriBufferPool *driPool, void *private) -{ - struct _DriSlabBuffer *buf = - (struct _DriSlabBuffer *) private; - struct _DriSlab *slab; - struct _DriSlabSizeHeader *header; - - if (!buf->isSlabBuffer) { - struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; - int ret; - - ret = drmBOUnreference(pool->fd, buf->bo); - free(buf->bo); - free(buf); - return ret; - } - - slab = buf->parent; - header = slab->header; - - _glthread_LOCK_MUTEX(header->mutex); - buf->unFenced = 0; - buf->mapCount = 0; - - if (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)) { - DRMLISTADDTAIL(&buf->head, &header->delayedBuffers); - header->numDelayed++; - } else { - if (buf->fence) - driFenceUnReference(&buf->fence); - driSlabFreeBufferLocked(buf); - } - - _glthread_UNLOCK_MUTEX(header->mutex); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *driPool, void *private, - _glthread_Mutex *mutex, int lazy) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - while(buf->unFenced) - _glthread_COND_WAIT(buf->event, *mutex); - - if (!buf->fence) - return 0; - - driFenceFinish(buf->fence, buf->fenceType, lazy); - driFenceUnReference(&buf->fence); - - return 0; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, _glthread_Mutex *mutex, void **virtual) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - int busy; - - if (buf->isSlabBuffer) - busy = buf->unFenced || (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)); - else - busy = buf->fence && !driFenceSignaled(buf->fence, buf->fenceType); - - - if (busy) { - if (hint & DRM_BO_HINT_DONT_BLOCK) - return -EBUSY; - else { - (void) pool_waitIdle(pool, private, mutex, 0); - } - } - - ++buf->mapCount; - *virtual = (buf->isSlabBuffer) ? - (void *) ((uint8_t *) buf->parent->kbo->virtual + buf->start) : - (void *) buf->bo->virtual; - - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - --buf->mapCount; - if (buf->mapCount == 0 && buf->isSlabBuffer) - _glthread_COND_BROADCAST(buf->event); - - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - struct _DriSlab *slab; - struct _DriSlabSizeHeader *header; - - if (!buf->isSlabBuffer) { - assert(buf->bo->proposedFlags & DRM_BO_FLAG_NO_MOVE); - return buf->bo->offset; - } - - slab = buf->parent; - header = slab->header; - - (void) header; - assert(header->slabPool->proposedFlags & DRM_BO_FLAG_NO_MOVE); - return slab->kbo->bo.offset + buf->start; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - return buf->start; -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - if (!buf->isSlabBuffer) - return buf->bo->flags; - - return buf->parent->kbo->bo.flags; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - if (!buf->isSlabBuffer) - return buf->bo->size; - - return buf->parent->header->bufSize; -} - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - drmBO *bo; - - if (buf->fence) - driFenceUnReference(&buf->fence); - - buf->fence = driFenceReference(fence); - bo = (buf->isSlabBuffer) ? - &buf->parent->kbo->bo: - buf->bo; - buf->fenceType = bo->fenceFlags; - - buf->unFenced = 0; - _glthread_COND_BROADCAST(buf->event); - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - return (buf->isSlabBuffer) ? &buf->parent->kbo->bo : buf->bo; -} - -static int -pool_validate(struct _DriBufferPool *pool, void *private, - _glthread_Mutex *mutex) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - if (!buf->isSlabBuffer) - return 0; - - while(buf->mapCount != 0) - _glthread_COND_WAIT(buf->event, *mutex); - - buf->unFenced = 1; - return 0; -} - - -struct _DriFreeSlabManager * -driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec) -{ - struct _DriFreeSlabManager *tmp; - - tmp = calloc(1, sizeof(*tmp)); - if (!tmp) - return NULL; - - _glthread_INIT_MUTEX(tmp->mutex); - _glthread_LOCK_MUTEX(tmp->mutex); - tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000; - tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000; - tmp->slabTimeout.tv_usec -= tmp->slabTimeout.tv_sec*1000000; - - tmp->checkInterval.tv_usec = checkIntervalMsec*1000; - tmp->checkInterval.tv_sec = tmp->checkInterval.tv_usec / 1000000; - tmp->checkInterval.tv_usec -= tmp->checkInterval.tv_sec*1000000; - - gettimeofday(&tmp->nextCheck, NULL); - driTimeAdd(&tmp->nextCheck, &tmp->checkInterval); - DRMINITLISTHEAD(&tmp->timeoutList); - DRMINITLISTHEAD(&tmp->unCached); - DRMINITLISTHEAD(&tmp->cached); - _glthread_UNLOCK_MUTEX(tmp->mutex); - - return tmp; -} - -void -driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan) -{ - struct timeval time; - - time = fMan->nextCheck; - driTimeAdd(&time, &fMan->checkInterval); - - _glthread_LOCK_MUTEX(fMan->mutex); - driFreeTimeoutKBOsLocked(fMan, &time); - _glthread_UNLOCK_MUTEX(fMan->mutex); - - assert(fMan->timeoutList.next == &fMan->timeoutList); - assert(fMan->unCached.next == &fMan->unCached); - assert(fMan->cached.next == &fMan->cached); - - free(fMan); -} - -static void -driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size, - struct _DriSlabSizeHeader *header) -{ - _glthread_INIT_MUTEX(header->mutex); - _glthread_LOCK_MUTEX(header->mutex); - - DRMINITLISTHEAD(&header->slabs); - DRMINITLISTHEAD(&header->freeSlabs); - DRMINITLISTHEAD(&header->delayedBuffers); - - header->numDelayed = 0; - header->slabPool = pool; - header->bufSize = size; - - _glthread_UNLOCK_MUTEX(header->mutex); -} - -static void -driFinishSizeHeader(struct _DriSlabSizeHeader *header) -{ - drmMMListHead *list, *next; - struct _DriSlabBuffer *buf; - - _glthread_LOCK_MUTEX(header->mutex); - for (list = header->delayedBuffers.next, next = list->next; - list != &header->delayedBuffers; - list = next, next = list->next) { - - buf = DRMLISTENTRY(struct _DriSlabBuffer, list , head); - if (buf->fence) { - (void) driFenceFinish(buf->fence, buf->fenceType, 0); - driFenceUnReference(&buf->fence); - } - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - _glthread_UNLOCK_MUTEX(header->mutex); -} - -static void -pool_takedown(struct _DriBufferPool *driPool) -{ - struct _DriSlabPool *pool = driPool->data; - int i; - - for (i=0; inumBuckets; ++i) { - driFinishSizeHeader(&pool->headers[i]); - } - - free(pool->headers); - free(pool->bucketSizes); - free(pool); - free(driPool); -} - -struct _DriBufferPool * -driSlabPoolInit(int fd, uint64_t flags, - uint64_t validMask, - uint32_t smallestSize, - uint32_t numSizes, - uint32_t desiredNumBuffers, - uint32_t maxSlabSize, - uint32_t pageAlignment, - struct _DriFreeSlabManager *fMan) -{ - struct _DriBufferPool *driPool; - struct _DriSlabPool *pool; - uint32_t i; - - driPool = calloc(1, sizeof(*driPool)); - if (!driPool) - return NULL; - - pool = calloc(1, sizeof(*pool)); - if (!pool) - goto out_err0; - - pool->bucketSizes = calloc(numSizes, sizeof(*pool->bucketSizes)); - if (!pool->bucketSizes) - goto out_err1; - - pool->headers = calloc(numSizes, sizeof(*pool->headers)); - if (!pool->headers) - goto out_err2; - - pool->fMan = fMan; - pool->proposedFlags = flags; - pool->validMask = validMask; - pool->numBuckets = numSizes; - pool->pageSize = getpagesize(); - pool->fd = fd; - pool->pageAlignment = pageAlignment; - pool->maxSlabSize = maxSlabSize; - pool->desiredNumBuffers = desiredNumBuffers; - - for (i=0; inumBuckets; ++i) { - pool->bucketSizes[i] = (smallestSize << i); - driInitSizeHeader(pool, pool->bucketSizes[i], - &pool->headers[i]); - } - - driPool->data = (void *) pool; - driPool->map = &pool_map; - driPool->unmap = &pool_unmap; - driPool->destroy = &pool_destroy; - driPool->offset = &pool_offset; - driPool->poolOffset = &pool_poolOffset; - driPool->flags = &pool_flags; - driPool->size = &pool_size; - driPool->create = &pool_create; - driPool->fence = &pool_fence; - driPool->kernel = &pool_kernel; - driPool->validate = &pool_validate; - driPool->waitIdle = &pool_waitIdle; - driPool->takeDown = &pool_takedown; - - return driPool; - - out_err2: - free(pool->bucketSizes); - out_err1: - free(pool); - out_err0: - free(driPool); - - return NULL; -} -- cgit v1.2.3 From 1a2c445baf9f493b1590d715314a1c240e4c9ef7 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 6 Jun 2008 14:51:11 +0200 Subject: i915: Moved EGL_i915 to the common code --- src/gallium/winsys/egl_drm/Makefile.template | 1 + src/gallium/winsys/egl_drm/intel/Makefile | 15 +- .../winsys/egl_drm/intel/intel_batchbuffer.c | 465 ---------- .../winsys/egl_drm/intel/intel_batchbuffer.h | 133 +-- src/gallium/winsys/egl_drm/intel/intel_context.c | 343 +++----- src/gallium/winsys/egl_drm/intel/intel_context.h | 81 +- src/gallium/winsys/egl_drm/intel/intel_egl.c | 15 +- src/gallium/winsys/egl_drm/intel/intel_lock.c | 102 --- src/gallium/winsys/egl_drm/intel/intel_reg.h | 10 +- src/gallium/winsys/egl_drm/intel/intel_screen.c | 599 +------------ src/gallium/winsys/egl_drm/intel/intel_screen.h | 110 +-- .../winsys/egl_drm/intel/intel_swapbuffers.c | 243 +----- .../winsys/egl_drm/intel/intel_swapbuffers.h | 47 - src/gallium/winsys/egl_drm/intel/intel_winsys.h | 73 -- .../winsys/egl_drm/intel/intel_winsys_i915.c | 184 ---- .../winsys/egl_drm/intel/intel_winsys_pipe.c | 338 ------- .../winsys/egl_drm/intel/intel_winsys_softpipe.c | 82 -- src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c | 953 -------------------- src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h | 138 --- src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h | 102 --- src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c | 268 ------ src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c | 377 -------- src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h | 115 --- .../winsys/egl_drm/intel/ws_dri_mallocpool.c | 162 ---- src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c | 970 --------------------- 25 files changed, 185 insertions(+), 5741 deletions(-) delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_batchbuffer.c delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_lock.c delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_swapbuffers.h delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_winsys.h delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_winsys_i915.c delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_winsys_pipe.c delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_winsys_softpipe.c delete mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c delete mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h delete mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h delete mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c delete mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c delete mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h delete mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_mallocpool.c delete mode 100644 src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/Makefile.template b/src/gallium/winsys/egl_drm/Makefile.template index 3bc1fdd4d4..07abfa53f3 100644 --- a/src/gallium/winsys/egl_drm/Makefile.template +++ b/src/gallium/winsys/egl_drm/Makefile.template @@ -54,6 +54,7 @@ SHARED_INCLUDES = \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/gallium/auxiliary \ -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/winsys/common \ -I$(TOP)/src/mesa \ -I$(TOP)/src/mesa/main \ -I$(TOP)/src/mesa/glapi \ diff --git a/src/gallium/winsys/egl_drm/intel/Makefile b/src/gallium/winsys/egl_drm/intel/Makefile index 5778ba77c3..9c7ff065ff 100644 --- a/src/gallium/winsys/egl_drm/intel/Makefile +++ b/src/gallium/winsys/egl_drm/intel/Makefile @@ -6,22 +6,13 @@ LIBNAME = EGL_i915.so PIPE_DRIVERS = \ $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ - $(TOP)/src/gallium/drivers/i915simple/libi915simple.a + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a \ + $(TOP)/src/gallium/winsys/common/intel_drm/libinteldrm.a DRIVER_SOURCES = \ - intel_winsys_pipe.c \ - intel_winsys_softpipe.c \ - intel_winsys_i915.c \ - intel_batchbuffer.c \ intel_swapbuffers.c \ intel_context.c \ - intel_lock.c \ intel_screen.c \ - ws_dri_bufmgr.c \ - ws_dri_drmpool.c \ - ws_dri_fencemgr.c \ - ws_dri_mallocpool.c \ - ws_dri_slabpool.c \ intel_egl.c C_SOURCES = \ @@ -35,6 +26,4 @@ DRIVER_DEFINES = -I$(TOP)/src/mesa/drivers/dri/intel $(shell pkg-config libdrm - include ../Makefile.template -intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c - symlinks: diff --git a/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.c b/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.c deleted file mode 100644 index 7ffa05a6e6..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.c +++ /dev/null @@ -1,465 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_batchbuffer.h" -#include "intel_context.h" -#include "intel_egl.h" - -#include - -#if 0 -static void -intel_dump_batchbuffer(GLuint offset, GLuint * ptr, GLuint count) -{ - int i; - fprintf(stderr, "\n\n\nSTART BATCH (%d dwords):\n", count / 4); - for (i = 0; i < count / 4; i += 4) - fprintf(stderr, "0x%x:\t0x%08x 0x%08x 0x%08x 0x%08x\n", - offset + i * 4, ptr[i], ptr[i + 1], ptr[i + 2], ptr[i + 3]); - fprintf(stderr, "END BATCH\n\n\n"); -} -#endif - -static void -intel_realloc_relocs(struct intel_batchbuffer *batch, int num_relocs) -{ - unsigned long size = num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER; - - size *= sizeof(uint32_t); - batch->reloc = realloc(batch->reloc, size); - batch->reloc_size = num_relocs; -} - - -void -intel_batchbuffer_reset(struct intel_batchbuffer *batch) -{ - /* - * Get a new, free batchbuffer. - */ - drmBO *bo; - struct drm_bo_info_req *req; - int ret; - - driBOUnrefUserList(batch->list); - driBOResetList(batch->list); - - batch->size = 4096; // ZZZ JB batch->intel->intelScreen->maxBatchSize; - ret = driBOData(batch->buffer, batch->size, NULL, NULL, 0); - assert(!ret); - - /* - * Add the batchbuffer to the validate list. - */ - - driBOAddListItem(batch->list, batch->buffer, - DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_EXE | DRM_BO_MASK_MEM, - &batch->dest_location, &batch->node); - - req = &batch->node->bo_arg.d.req.bo_req; - - /* - * Set up information needed for us to make relocations - * relative to the underlying drm buffer objects. - */ - - driReadLockKernelBO(); - bo = driBOKernel(batch->buffer); - req->presumed_offset = (uint64_t) bo->offset; - req->hint = DRM_BO_HINT_PRESUMED_OFFSET; - batch->drmBOVirtual = (uint8_t *) bo->virtual; - driReadUnlockKernelBO(); - - /* - * Adjust the relocation buffer size. - */ - - if (batch->reloc_size > INTEL_MAX_RELOCS || - batch->reloc == NULL) - intel_realloc_relocs(batch, INTEL_DEFAULT_RELOCS); - - assert(batch->reloc != NULL); - batch->reloc[0] = 0; /* No relocs yet. */ - batch->reloc[1] = 1; /* Reloc type 1 */ - batch->reloc[2] = 0; /* Only a single relocation list. */ - batch->reloc[3] = 0; /* Only a single relocation list. */ - - batch->map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); - batch->poolOffset = driBOPoolOffset(batch->buffer); - batch->ptr = batch->map; - batch->dirty_state = ~0; - batch->nr_relocs = 0; - batch->flags = 0; - batch->id = 0;//batch->intel->intelScreen->batch_id++; -} - -/*====================================================================== - * Public functions - */ -struct intel_batchbuffer * -intel_batchbuffer_alloc(struct intel_screen *intel_screen) -{ - struct intel_batchbuffer *batch = calloc(sizeof(*batch), 1); - - batch->intel_screen = intel_screen; - - driGenBuffers(intel_screen->batchPool, "batchbuffer", 1, - &batch->buffer, 4096, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); - batch->last_fence = NULL; - batch->list = driBOCreateList(20); - batch->reloc = NULL; - intel_batchbuffer_reset(batch); - return batch; -} - -void -intel_batchbuffer_free(struct intel_batchbuffer *batch) -{ - if (batch->last_fence) { - driFenceFinish(batch->last_fence, - DRM_FENCE_TYPE_EXE, GL_FALSE); - driFenceUnReference(&batch->last_fence); - } - if (batch->map) { - driBOUnmap(batch->buffer); - batch->map = NULL; - } - driBOUnReference(batch->buffer); - driBOFreeList(batch->list); - if (batch->reloc) - free(batch->reloc); - batch->buffer = NULL; - free(batch); -} - -void -intel_offset_relocation(struct intel_batchbuffer *batch, - unsigned pre_add, - struct _DriBufferObject *driBO, - uint64_t val_flags, - uint64_t val_mask) -{ - int itemLoc; - struct _drmBONode *node; - uint32_t *reloc; - struct drm_bo_info_req *req; - - driBOAddListItem(batch->list, driBO, val_flags, val_mask, - &itemLoc, &node); - req = &node->bo_arg.d.req.bo_req; - - if (!(req->hint & DRM_BO_HINT_PRESUMED_OFFSET)) { - - /* - * Stop other threads from tampering with the underlying - * drmBO while we're reading its offset. - */ - - driReadLockKernelBO(); - req->presumed_offset = (uint64_t) driBOKernel(driBO)->offset; - driReadUnlockKernelBO(); - req->hint = DRM_BO_HINT_PRESUMED_OFFSET; - } - - pre_add += driBOPoolOffset(driBO); - - if (batch->nr_relocs == batch->reloc_size) - intel_realloc_relocs(batch, batch->reloc_size * 2); - - reloc = batch->reloc + - (I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE); - - reloc[0] = ((uint8_t *)batch->ptr - batch->drmBOVirtual); - intel_batchbuffer_emit_dword(batch, req->presumed_offset + pre_add); - reloc[1] = pre_add; - reloc[2] = itemLoc; - reloc[3] = batch->dest_location; - batch->nr_relocs++; -} - -static void -i915_drm_copy_reply(const struct drm_bo_info_rep * rep, drmBO * buf) -{ - buf->handle = rep->handle; - buf->flags = rep->flags; - buf->size = rep->size; - buf->offset = rep->offset; - buf->mapHandle = rep->arg_handle; - buf->proposedFlags = rep->proposed_flags; - buf->start = rep->buffer_start; - buf->fenceFlags = rep->fence_flags; - buf->replyFlags = rep->rep_flags; - buf->pageAlignment = rep->page_alignment; -} - -static int -i915_execbuf(struct intel_batchbuffer *batch, - GLuint used, - GLboolean ignore_cliprects, - drmBOList *list, - struct drm_i915_execbuffer *ea) -{ - struct intel_screen *intel_screen = batch->intel_screen; - drmBONode *node; - drmMMListHead *l; - struct drm_i915_op_arg *arg, *first; - struct drm_bo_op_req *req; - struct drm_bo_info_rep *rep; - uint64_t *prevNext = NULL; - drmBO *buf; - int ret = 0; - uint32_t count = 0; - - first = NULL; - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - - arg = &node->bo_arg; - req = &arg->d.req; - - if (!first) - first = arg; - - if (prevNext) - *prevNext = (unsigned long)arg; - - prevNext = &arg->next; - req->bo_req.handle = node->buf->handle; - req->op = drm_bo_validate; - req->bo_req.flags = node->arg0; - req->bo_req.mask = node->arg1; - req->bo_req.hint |= 0; - count++; - } - - memset(ea, 0, sizeof(*ea)); - ea->num_buffers = count; - ea->batch.start = batch->poolOffset; - ea->batch.used = used; -#if 0 /* ZZZ JB: no cliprects used */ - ea->batch.cliprects = intel->pClipRects; - ea->batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; - ea->batch.DR1 = 0; - ea->batch.DR4 = 0;((((GLuint) intel->drawX) & 0xffff) | - (((GLuint) intel->drawY) << 16)); -#else - ea->batch.cliprects = NULL; - ea->batch.num_cliprects = 0; - ea->batch.DR1 = 0; - ea->batch.DR4 = 0; -#endif - ea->fence_arg.flags = DRM_I915_FENCE_FLAG_FLUSHED; - ea->ops_list = (unsigned long) first; - first->reloc_ptr = (unsigned long) batch->reloc; - batch->reloc[0] = batch->nr_relocs; - - //return -EFAULT; - do { - ret = drmCommandWriteRead(intel_screen->device->drmFD, DRM_I915_EXECBUFFER, ea, - sizeof(*ea)); - } while (ret == -EAGAIN); - - if (ret != 0) { - printf("%s somebody set us up the bomb\n", __FUNCTION__); - return ret; - } - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - arg = &node->bo_arg; - rep = &arg->d.rep.bo_info; - - if (!arg->handled) { - return -EFAULT; - } - if (arg->d.rep.ret) - return arg->d.rep.ret; - - buf = node->buf; - i915_drm_copy_reply(rep, buf); - } - return 0; -} - -/* TODO: Push this whole function into bufmgr. - */ -static struct _DriFenceObject * -do_flush_locked(struct intel_batchbuffer *batch, - GLuint used, - GLboolean ignore_cliprects, GLboolean allow_unlock) -{ - struct intel_screen *intel_screen = batch->intel_screen; - struct _DriFenceObject *fo; - drmFence fence; - drmBOList *boList; - struct drm_i915_execbuffer ea; - int ret = 0; - - driBOValidateUserList(batch->list); - boList = driGetdrmBOList(batch->list); - -#if 0 /* ZZZ JB Allways run */ - if (!(intel->numClipRects == 0 && !ignore_cliprects)) { -#else - if (1) { -#endif - ret = i915_execbuf(batch, used, ignore_cliprects, boList, &ea); - } else { - driPutdrmBOList(batch->list); - fo = NULL; - goto out; - } - driPutdrmBOList(batch->list); - if (ret) - abort(); - - if (ea.fence_arg.error != 0) { - - /* - * The hardware has been idled by the kernel. - * Don't fence the driBOs. - */ - - if (batch->last_fence) - driFenceUnReference(&batch->last_fence); -#if 0 /* ZZZ JB: no _mesa_* funcs in gallium */ - _mesa_printf("fence error\n"); -#endif - batch->last_fence = NULL; - fo = NULL; - goto out; - } - - fence.handle = ea.fence_arg.handle; - fence.fence_class = ea.fence_arg.fence_class; - fence.type = ea.fence_arg.type; - fence.flags = ea.fence_arg.flags; - fence.signaled = ea.fence_arg.signaled; - - fo = driBOFenceUserList(intel_screen->mgr, batch->list, - "SuperFence", &fence); - - if (driFenceType(fo) & DRM_I915_FENCE_TYPE_RW) { - if (batch->last_fence) - driFenceUnReference(&batch->last_fence); - /* - * FIXME: Context last fence?? - */ - batch->last_fence = fo; - driFenceReference(fo); - } - out: -#if 0 /* ZZZ JB: fix this */ - intel->vtbl.lost_hardware(intel); -#else -#endif - return fo; -} - - -struct _DriFenceObject * -intel_batchbuffer_flush(struct intel_batchbuffer *batch) -{ - //struct intel_context *intel = batch->intel; - GLuint used = batch->ptr - batch->map; - //GLboolean was_locked = 1; - struct _DriFenceObject *fence; - - if (used == 0) { - driFenceReference(batch->last_fence); - return batch->last_fence; - } - - /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a - * performance drain that we would like to avoid. - */ -#if 0 /* ZZZ JB: what should we do here? */ - if (used & 4) { - ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->ptr)[1] = 0; - ((int *) batch->ptr)[2] = MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->ptr)[1] = MI_BATCH_BUFFER_END; - used += 8; - } -#else - if (used & 4) { - ((int *) batch->ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->ptr)[1] = 0; - ((int *) batch->ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; - used += 8; - } -#endif - driBOUnmap(batch->buffer); - batch->ptr = NULL; - batch->map = NULL; - - /* TODO: Just pass the relocation list and dma buffer up to the - * kernel. - */ -/* if (!was_locked) - LOCK_HARDWARE(intel);*/ - - fence = do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), - GL_FALSE); - -/* if (!was_locked) - UNLOCK_HARDWARE(intel);*/ - - /* Reset the buffer: - */ - intel_batchbuffer_reset(batch); - return fence; -} - -void -intel_batchbuffer_finish(struct intel_batchbuffer *batch) -{ - struct _DriFenceObject *fence = intel_batchbuffer_flush(batch); - driFenceFinish(fence, driFenceType(fence), GL_FALSE); - driFenceUnReference(&fence); -} - -void -intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags) -{ - assert((bytes & 3) == 0); - intel_batchbuffer_require_space(batch, bytes, flags); - memcpy(batch->ptr, data, bytes); - batch->ptr += bytes; -} diff --git a/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h b/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h index 6d35cf8b96..1fa2719845 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h +++ b/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h @@ -1,133 +1,24 @@ #ifndef INTEL_BATCHBUFFER_H #define INTEL_BATCHBUFFER_H -#include "mtypes.h" -#include "ws_dri_bufmgr.h" +#include "intel_drm/intel_be_batchbuffer.h" -struct intel_screen; - -#define BATCH_SZ 16384 -#define BATCH_RESERVED 16 - -#define INTEL_DEFAULT_RELOCS 100 -#define INTEL_MAX_RELOCS 400 - -#define INTEL_BATCH_NO_CLIPRECTS 0x1 -#define INTEL_BATCH_CLIPRECTS 0x2 - -struct intel_batchbuffer -{ - struct bufmgr *bm; - struct intel_screen *intel_screen; - - struct _DriBufferObject *buffer; - struct _DriFenceObject *last_fence; - GLuint flags; - - struct _DriBufferList *list; - GLuint list_count; - GLubyte *map; - GLubyte *ptr; - - uint32_t *reloc; - GLuint reloc_size; - GLuint nr_relocs; - - GLuint size; - - GLuint dirty_state; - GLuint id; - - uint32_t poolOffset; - uint8_t *drmBOVirtual; - struct _drmBONode *node; /* Validation list node for this buffer */ - int dest_location; /* Validation list sequence for this buffer */ -}; - -struct intel_batchbuffer *intel_batchbuffer_alloc(struct intel_screen - *intel); - -void intel_batchbuffer_free(struct intel_batchbuffer *batch); - - -void intel_batchbuffer_finish(struct intel_batchbuffer *batch); - -struct _DriFenceObject *intel_batchbuffer_flush(struct intel_batchbuffer - *batch); - -void intel_batchbuffer_reset(struct intel_batchbuffer *batch); - - -/* Unlike bmBufferData, this currently requires the buffer be mapped. - * Consider it a convenience function wrapping multple - * intel_buffer_dword() calls. - */ -void intel_batchbuffer_data(struct intel_batchbuffer *batch, - const void *data, GLuint bytes, GLuint flags); - -void intel_batchbuffer_release_space(struct intel_batchbuffer *batch, - GLuint bytes); - -void -intel_offset_relocation(struct intel_batchbuffer *batch, - unsigned pre_add, - struct _DriBufferObject *driBO, - uint64_t val_flags, - uint64_t val_mask); - -/* Inline functions - might actually be better off with these - * non-inlined. Certainly better off switching all command packets to - * be passed as structs rather than dwords, but that's a little bit of - * work... +/* + * Need to redefine the BATCH defines */ -static INLINE GLuint -intel_batchbuffer_space(struct intel_batchbuffer *batch) -{ - return (batch->size - BATCH_RESERVED) - (batch->ptr - batch->map); -} - - -static INLINE void -intel_batchbuffer_emit_dword(struct intel_batchbuffer *batch, GLuint dword) -{ - assert(batch->map); - assert(intel_batchbuffer_space(batch) >= 4); - *(GLuint *) (batch->ptr) = dword; - batch->ptr += 4; -} -static INLINE void -intel_batchbuffer_require_space(struct intel_batchbuffer *batch, - GLuint sz, GLuint flags) -{ - struct _DriFenceObject *fence; +#undef BEGIN_BATCH +#define BEGIN_BATCH(dwords, relocs) \ + (i915_batchbuffer_check(&intel->base.batch->base, dwords, relocs)) - assert(sz < batch->size - 8); - if (intel_batchbuffer_space(batch) < sz || - (batch->flags != 0 && flags != 0 && batch->flags != flags)) { - fence = intel_batchbuffer_flush(batch); - driFenceUnReference(&fence); - } +#undef OUT_BATCH +#define OUT_BATCH(d) \ + i915_batchbuffer_dword(&intel->base.batch->base, d) - batch->flags |= flags; -} - -/* Here are the crusty old macros, to be removed: - */ -#define BATCH_LOCALS - -#define BEGIN_BATCH(n, flags) do { \ - assert(!intel->prim.flush); \ - intel_batchbuffer_require_space(intel->batch, (n)*4, flags); \ -} while (0) - -#define OUT_BATCH(d) intel_batchbuffer_emit_dword(intel->batch, d) - -#define OUT_RELOC(buf,flags,mask,delta) do { \ +#undef OUT_RELOC +#define OUT_RELOC(buf,flags,mask,delta) do { \ assert((delta) >= 0); \ - intel_offset_relocation(intel->batch, delta, buf, flags, mask); \ + intel_be_offset_relocation(intel->base.batch, delta, buf, flags, mask); \ } while (0) -#define ADVANCE_BATCH() do { } while(0) - #endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.c b/src/gallium/winsys/egl_drm/intel/intel_context.c index 394a40d714..fdbaa230e5 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_context.c +++ b/src/gallium/winsys/egl_drm/intel/intel_context.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,14 +22,13 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ +#include "i915simple/i915_screen.h" #include "intel_screen.h" #include "intel_context.h" -#include "intel_swapbuffers.h" -#include "intel_winsys.h" #include "intel_batchbuffer.h" #include "state_tracker/st_public.h" @@ -70,74 +69,109 @@ int __intel_debug = 0; * old i830-specific driver. */ const struct dri_extension card_extensions[] = { - {"GL_ARB_multisample", GL_ARB_multisample_functions}, - {"GL_ARB_multitexture", NULL}, - {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, - {"GL_ARB_texture_border_clamp", NULL}, - {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, - {"GL_ARB_texture_cube_map", NULL}, - {"GL_ARB_texture_env_add", NULL}, - {"GL_ARB_texture_env_combine", NULL}, - {"GL_ARB_texture_env_dot3", NULL}, - {"GL_ARB_texture_mirrored_repeat", NULL}, - {"GL_ARB_texture_rectangle", NULL}, - {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, - {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, - {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, - {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, - {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, - {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_subtract", NULL}, - {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, - {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, - {"GL_EXT_packed_depth_stencil", NULL}, - {"GL_EXT_pixel_buffer_object", NULL}, - {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, - {"GL_EXT_stencil_wrap", NULL}, - {"GL_EXT_texture_edge_clamp", NULL}, - {"GL_EXT_texture_env_combine", NULL}, - {"GL_EXT_texture_env_dot3", NULL}, - {"GL_EXT_texture_filter_anisotropic", NULL}, - {"GL_EXT_texture_lod_bias", NULL}, - {"GL_3DFX_texture_compression_FXT1", NULL}, - {"GL_APPLE_client_storage", NULL}, - {"GL_MESA_pack_invert", NULL}, - {"GL_MESA_ycbcr_texture", NULL}, - {"GL_NV_blend_square", NULL}, - {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, - {"GL_NV_vertex_program1_1", NULL}, - {"GL_SGIS_generate_mipmap", NULL }, - {NULL, NULL} + {"GL_ARB_multisample", GL_ARB_multisample_functions}, + {"GL_ARB_multitexture", NULL}, + {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_texture_border_clamp", NULL}, + {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, + {"GL_ARB_texture_cube_map", NULL}, + {"GL_ARB_texture_env_add", NULL}, + {"GL_ARB_texture_env_combine", NULL}, + {"GL_ARB_texture_env_dot3", NULL}, + {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_rectangle", NULL}, + {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, + {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, + {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, + {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, + {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_subtract", NULL}, + {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, + {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, + {"GL_EXT_packed_depth_stencil", NULL}, + {"GL_EXT_pixel_buffer_object", NULL}, + {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, + {"GL_EXT_stencil_wrap", NULL}, + {"GL_EXT_texture_edge_clamp", NULL}, + {"GL_EXT_texture_env_combine", NULL}, + {"GL_EXT_texture_env_dot3", NULL}, + {"GL_EXT_texture_filter_anisotropic", NULL}, + {"GL_EXT_texture_lod_bias", NULL}, + {"GL_3DFX_texture_compression_FXT1", NULL}, + {"GL_APPLE_client_storage", NULL}, + {"GL_MESA_pack_invert", NULL}, + {"GL_MESA_ycbcr_texture", NULL}, + {"GL_NV_blend_square", NULL}, + {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, + {"GL_NV_vertex_program1_1", NULL}, + {"GL_SGIS_generate_mipmap", NULL }, + {NULL, NULL} }; + +/* + * Hardware lock functions. + * Doesn't do anything in EGL + */ + +static void +intel_lock_hardware(struct intel_be_context *context) +{ + (void)context; +} + +static void +intel_unlock_hardware(struct intel_be_context *context) +{ + (void)context; +} + +static boolean +intel_locked_hardware(struct intel_be_context *context) +{ + (void)context; + return FALSE; +} + + +/* + * Misc functions. + */ + int -intel_create_context(struct egl_drm_context *eglCon, const __GLcontextModes *visual, void *sharedContextPrivate) +intel_create_context(struct egl_drm_context *egl_context, const __GLcontextModes *visual, void *sharedContextPrivate) { - struct intel_context *iCon = CALLOC_STRUCT(intel_context); - struct intel_screen *iScrn = (struct intel_screen *)eglCon->device->priv; + struct intel_context *intel = CALLOC_STRUCT(intel_context); + struct intel_screen *screen = (struct intel_screen *)egl_context->device->priv; struct pipe_context *pipe; struct st_context *st_share = NULL; - eglCon->priv = iCon; + egl_context->priv = intel; + + intel->intel_screen = screen; + intel->egl_context = egl_context; + intel->egl_device = egl_context->device; - iCon->intel_screen = iScrn; - iCon->egl_context = eglCon; - iCon->egl_device = eglCon->device; + intel->base.hardware_lock = intel_lock_hardware; + intel->base.hardware_unlock = intel_unlock_hardware; + intel->base.hardware_locked = intel_locked_hardware; - iCon->batch = intel_batchbuffer_alloc(iScrn); - iCon->last_swap_fence = NULL; - iCon->first_swap_fence = NULL; + intel_be_init_context(&intel->base, &screen->base); - pipe = intel_create_i915simple(iCon, iScrn->winsys); -// pipe = intel_create_softpipe(iCon, iScrn->winsys); +#if 0 + pipe = intel_create_softpipe(intel, screen->winsys); +#else + pipe = i915_create_context(screen->pipe, &screen->base.base, &intel->base.base); +#endif - pipe->priv = iCon; + pipe->priv = intel; - iCon->st = st_create_context(pipe, visual, st_share); + intel->st = st_create_context(pipe, visual, st_share); return TRUE; } @@ -161,7 +195,6 @@ intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *dra if (draw != read) st_resize_framebuffer(read_fb->stfb, read->w, read->h); - //intelUpdateWindowSize(driDrawPriv); } else { st_make_current(NULL, NULL, NULL); } @@ -173,7 +206,9 @@ intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer struct intel_screen *intelScreen = (struct intel_screen *)draw->device->priv; struct intel_framebuffer *draw_fb = (struct intel_framebuffer *)draw->priv; - driBOUnReference(draw_fb->front_buffer); + if (draw_fb->front_buffer) + driBOUnReference(draw_fb->front_buffer); + draw_fb->front_buffer = NULL; draw_fb->front = NULL; @@ -183,184 +218,8 @@ intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer draw_fb->front = front; - driGenBuffers(intelScreen->staticPool, "front", 1, &draw_fb->front_buffer, 0, 0, 0); + driGenBuffers(intelScreen->base.staticPool, "front", 1, &draw_fb->front_buffer, 0, 0, 0); driBOSetReferenced(draw_fb->front_buffer, front->handle); st_resize_framebuffer(draw_fb->stfb, draw->w, draw->h); } - -#if 0 -GLboolean -intelCreateContext(const __GLcontextModes * visual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate) -{ - struct intel_context *intel = CALLOC_STRUCT(intel_context); - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - struct intel_screen *intelScreen = intel_screen(sPriv); - drmI830Sarea *saPriv = intelScreen->sarea; - int fthrottle_mode; - GLboolean havePools; - struct pipe_context *pipe; - struct st_context *st_share = NULL; - - if (sharedContextPrivate) { - st_share = ((struct intel_context *) sharedContextPrivate)->st; - } - - driContextPriv->driverPrivate = intel; - intel->intelScreen = intelScreen; - intel->driScreen = sPriv; - intel->sarea = saPriv; - - driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, - intel->driScreen->myNum, "i915"); - - - /* - * memory pools - */ - DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - // ZZZ JB should be per screen and not be done per context - havePools = intelCreatePools(sPriv); - DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - if (!havePools) - return GL_FALSE; - - - /* Dri stuff */ - intel->hHWContext = driContextPriv->hHWContext; - intel->driFd = sPriv->fd; - intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; - - fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->iw.irq_seq = -1; - intel->irqsEmitted = 0; - - intel->batch = intel_batchbuffer_alloc(intel); - intel->last_swap_fence = NULL; - intel->first_swap_fence = NULL; - -#ifdef DEBUG - __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); -#endif - - /* - * Pipe-related setup - */ - if (getenv("INTEL_SP")) { - /* use softpipe driver instead of hw */ - pipe = intel_create_softpipe( intel, intelScreen->winsys ); - } - else { - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - pipe = intel_create_i915simple( intel, intelScreen->winsys ); - break; - default: - fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", - intel->intelScreen->deviceID, __FUNCTION__); - - pipe = intel_create_softpipe( intel, intelScreen->winsys ); - break; - } - } - - pipe->priv = intel; - - intel->st = st_create_context(pipe, visual, st_share); - - return GL_TRUE; -} - - -void -intelDestroyContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = intel_context(driContextPriv); - - assert(intel); /* should never be null */ - if (intel) { - st_finish(intel->st); - - intel_batchbuffer_free(intel->batch); - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(&intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - if (intel->first_swap_fence) { - driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(&intel->first_swap_fence); - intel->first_swap_fence = NULL; - } - - if (intel->intelScreen->dummyContext == intel) - intel->intelScreen->dummyContext = NULL; - - st_destroy_context(intel->st); - free(intel); - } -} - - -GLboolean -intelUnbindContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = intel_context(driContextPriv); - st_flush(intel->st, PIPE_FLUSH_RENDER_CACHE, NULL); - /* XXX make_current(NULL)? */ - return GL_TRUE; -} - - -GLboolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv) -{ - if (driContextPriv) { - struct intel_context *intel = intel_context(driContextPriv); - struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv); - struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv); - - assert(draw_fb->stfb); - assert(read_fb->stfb); - - /* This is for situations in which we need a rendering context but - * there may not be any currently bound. - */ - intel->intelScreen->dummyContext = intel; - - st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); - - if ((intel->driDrawable != driDrawPriv) || - (intel->lastStamp != driDrawPriv->lastStamp)) { - intel->driDrawable = driDrawPriv; - intelUpdateWindowSize(driDrawPriv); - intel->lastStamp = driDrawPriv->lastStamp; - } - - /* The size of the draw buffer will have been updated above. - * If the readbuffer is a different window, check/update its size now. - */ - if (driReadPriv != driDrawPriv) { - intelUpdateWindowSize(driReadPriv); - } - - } - else { - st_make_current(NULL, NULL, NULL); - } - - return GL_TRUE; -} -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.h b/src/gallium/winsys/egl_drm/intel/intel_context.h index aa9903f274..ccf8120761 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_context.h +++ b/src/gallium/winsys/egl_drm/intel/intel_context.h @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,61 +22,37 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ #ifndef INTEL_CONTEXT_H #define INTEL_CONTEXT_H -#include -#include "drm.h" - #include "pipe/p_debug.h" - -#include "intel_screen.h" -#include "i915_drm.h" +#include "intel_drm/intel_be_context.h" -struct pipe_context; -struct intel_context; -struct _DriBufferObject; struct st_context; struct egl_drm_device; struct egl_drm_context; struct egl_drm_frontbuffer; -#define INTEL_MAX_FIXUP 64 - /** * Intel rendering context, contains a state tracker and intel-specific info. */ struct intel_context { - struct st_context *st; - - struct _DriFenceObject *last_swap_fence; - struct _DriFenceObject *first_swap_fence; + struct intel_be_context base; - struct intel_batchbuffer *batch; + struct st_context *st; -#if 0 - boolean locked; - char *prevLockFile; - int prevLockLine; -#endif - - /* pick this up from the screen instead - int drmFd; - */ + struct intel_screen *intel_screen; - struct intel_screen *intel_screen; - - uint lastStamp; - /* new egl stuff */ - struct egl_drm_device *egl_device; - struct egl_drm_context *egl_context; - struct egl_drm_drawable *egl_drawable; + /* new egl stuff */ + struct egl_drm_device *egl_device; + struct egl_drm_context *egl_context; + struct egl_drm_drawable *egl_drawable; }; @@ -86,12 +62,12 @@ struct intel_context */ struct intel_framebuffer { - struct st_framebuffer *stfb; + struct st_framebuffer *stfb; - /* other fields TBD */ - int other; - struct _DriBufferObject *front_buffer; - struct egl_drm_frontbuffer *front; + /* other fields TBD */ + int other; + struct _DriBufferObject *front_buffer; + struct egl_drm_frontbuffer *front; }; @@ -123,11 +99,10 @@ extern int __intel_debug; } while(0) #else -#define DBG(flag, ...) +#define DBG(flag, ...) #endif - #define PCI_CHIP_845_G 0x2562 #define PCI_CHIP_I830_M 0x3577 #define PCI_CHIP_I855_GM 0x3582 @@ -141,22 +116,4 @@ extern int __intel_debug; #define PCI_CHIP_Q35_G 0x29B2 #define PCI_CHIP_Q33_G 0x29D2 - -#if 0 -/** Cast wrapper */ -static INLINE struct intel_context * -intel_context(__DRIcontextPrivate *driContextPriv) -{ - return (struct intel_context *) driContextPriv->driverPrivate; -} - - -/** Cast wrapper */ -static INLINE struct intel_framebuffer * -intel_framebuffer(__DRIdrawablePrivate * driDrawPriv) -{ - return (struct intel_framebuffer *) driDrawPriv->driverPrivate; -} -#endif - #endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 41b7eac4ed..3b4ab330f5 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -21,9 +21,6 @@ #include "intel_context.h" -#include "ws_dri_bufmgr.h" - -#include "intel_winsys.h" #include "state_tracker/st_public.h" struct egl_drm_device* egl_drm_create_device(int drmFD); @@ -434,10 +431,12 @@ prettyColors(int fd, unsigned int handle, size_t pitch) { drmBO bo; unsigned int *ptr; + void *p; int i; drmBOReference(fd, handle, &bo); - drmBOMap(fd, &bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, (void**)&ptr); + drmBOMap(fd, &bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, &p); + ptr = (unsigned int*)p; for (i = 0; i < (bo.size / 4); i++) ptr[i] = 0xFFFFFFFF; @@ -460,8 +459,6 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, struct drm_driver *drm_drv = (struct drm_driver *)drv; struct drm_surface *surf = lookup_drm_surface(surface); struct drm_screen *scrn = lookup_drm_screen(dpy, screen); - //struct intel_framebuffer *intel_fb = NULL; - //struct pipe_surface *front_surf = NULL; _EGLMode *mode = _eglLookupMode(dpy, m); size_t pitch = 2048 * 4; size_t size = mode->Height * pitch; @@ -469,8 +466,6 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, /* TODO if allready shown take down */ - printf("setting mode to %i x %i\n", mode->Width, mode->Height); - ret = drmBOCreate(drm_drv->device->drmFD, size, 0, 0, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | @@ -554,6 +549,7 @@ drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface re struct drm_context *ctx = lookup_drm_context(context); EGLBoolean b; + printf("drm_make_current\n"); b = _eglMakeCurrent(drv, dpy, draw, read, context); if (!b) return EGL_FALSE; @@ -563,6 +559,7 @@ drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface re (void) readSurf; (void) ctx; + printf("enter intel_make_current\n"); intel_make_current(ctx->context, drawSurf->drawable, readSurf->drawable); return EGL_TRUE; } @@ -613,6 +610,8 @@ _eglMain(_EGLDisplay *dpy, const char *args) drm->base.API.ShowScreenSurfaceMESA = drm_show_screen_surface_mesa; drm->base.API.SwapBuffers = drm_swap_buffers; + drm->base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; + /* enable supported extensions */ drm->base.Extensions.MESA_screen_surface = EGL_TRUE; drm->base.Extensions.MESA_copy_context = EGL_TRUE; diff --git a/src/gallium/winsys/egl_drm/intel/intel_lock.c b/src/gallium/winsys/egl_drm/intel/intel_lock.c deleted file mode 100644 index cec83c7585..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_lock.c +++ /dev/null @@ -1,102 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "main/glheader.h" -#include "glapi/glthread.h" -#include -#include "state_tracker/st_public.h" -#include "intel_context.h" - -#if 0 - -_glthread_DECLARE_STATIC_MUTEX( lockMutex ); - - -static void -intelContendedLock(struct intel_context *intel, uint flags) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - __DRIscreenPrivate *sPriv = intel->driScreen; - struct intel_screen *intelScreen = intel_screen(sPriv); - drmI830Sarea *sarea = intel->sarea; - - drmGetLock(intel->driFd, intel->hHWContext, flags); - - DBG(LOCK, "%s - got contended lock\n", __progname); - - /* If the window moved, may need to set a new cliprect now. - * - * NOTE: This releases and regains the hw lock, so all state - * checking must be done *after* this call: - */ - if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - - if (sarea->width != intelScreen->front.width || - sarea->height != intelScreen->front.height) { - - intelUpdateScreenRotation(sPriv, sarea); - } -} - - -/* Lock the hardware and validate our state. - */ -void LOCK_HARDWARE( struct intel_context *intel ) -{ - char __ret = 0; - - _glthread_LOCK_MUTEX(lockMutex); - assert(!intel->locked); - - DRM_CAS(intel->driHwLock, intel->hHWContext, - (DRM_LOCK_HELD|intel->hHWContext), __ret); - - if (__ret) - intelContendedLock( intel, 0 ); - - DBG(LOCK, "%s - locked\n", __progname); - - intel->locked = 1; -} - - -/* Unlock the hardware using the global current context - */ -void UNLOCK_HARDWARE( struct intel_context *intel ) -{ - assert(intel->locked); - intel->locked = 0; - - DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); - - _glthread_UNLOCK_MUTEX(lockMutex); - - DBG(LOCK, "%s - unlocked\n", __progname); -} -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_reg.h b/src/gallium/winsys/egl_drm/intel/intel_reg.h index f37c24fda9..4f33bee438 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_reg.h +++ b/src/gallium/winsys/egl_drm/intel/intel_reg.h @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,7 +22,7 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.c b/src/gallium/winsys/egl_drm/intel/intel_screen.c index 38c4098087..96b0bf1b85 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_screen.c +++ b/src/gallium/winsys/egl_drm/intel/intel_screen.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,62 +22,24 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ + #include "utils.h" -#include "vblank.h" -#include "xmlpool.h" + +#include "state_tracker/st_public.h" +#include "i915simple/i915_screen.h" #include "intel_context.h" #include "intel_screen.h" #include "intel_batchbuffer.h" -//#include "intel_batchpool.h" -#include "intel_swapbuffers.h" -#include "intel_winsys.h" - -#include "ws_dri_bufpool.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_cb_fbo.h" #include "intel_egl.h" -static boolean -intel_create_pools(struct intel_screen *intel_screen) -{ - if (intel_screen->havePools) - return GL_TRUE; - - intel_screen->mgr = driFenceMgrTTMInit(intel_screen->device->drmFD); - if (!intel_screen->mgr) { - fprintf(stderr, "Failed to create fence manager.\n"); - return FALSE; - } - - intel_screen->fMan = driInitFreeSlabManager(10, 10); - if (!intel_screen->fMan) { - fprintf(stderr, "Failed to create free slab manager.\n"); - return FALSE; - } - - intel_screen->staticPool = driDRMPoolInit(intel_screen->device->drmFD); - intel_screen->batchPool = driSlabPoolInit(intel_screen->device->drmFD, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - 4096, //intelScreen->maxBatchSize, - 1, 40, 16*16384, 0, - intel_screen->fMan); - - intel_screen->havePools = GL_TRUE; - - return GL_TRUE; -} extern const struct dri_extension card_extensions[]; + int intel_init_driver(struct egl_drm_device *device) { @@ -85,18 +47,18 @@ intel_init_driver(struct egl_drm_device *device) /* Allocate the private area */ intel_screen = CALLOC_STRUCT(intel_screen); - if (!intel_screen) + if (!intel_screen) return FALSE; device->priv = (void *)intel_screen; intel_screen->device = device; - if (!intel_create_pools(intel_screen)) - return FALSE; + /** TODO JB: ugly hack */ + intel_screen->deviceID = PCI_CHIP_I945_GM; - intel_screen->batch = intel_batchbuffer_alloc(intel_screen); + intel_be_init_device(&intel_screen->base, device->drmFD); - intel_screen->winsys = intel_create_pipe_winsys(device->drmFD, intel_screen->fMan); + intel_screen->pipe = i915_create_screen(&intel_screen->base.base, intel_screen->deviceID); /* hack */ driInitExtensions(NULL, card_extensions, GL_FALSE); @@ -147,534 +109,3 @@ intel_create_drawable(struct egl_drm_drawable *drawable, drawable->priv = (void *) intelfb; return GL_TRUE; } - -#if 0 -PUBLIC const char __driConfigOptions[] = - DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE - DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) - DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) - DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY - DRI_CONF_FORCE_S3TC_ENABLE(false) - DRI_CONF_ALLOW_LARGE_TEXTURES(1) - DRI_CONF_SECTION_END DRI_CONF_END; - -const uint __driNConfigOptions = 4; - -#ifdef USE_NEW_INTERFACE -static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; -#endif /*USE_NEW_INTERFACE */ - -extern const struct dri_extension card_extensions[]; - - - - -static void -intelPrintDRIInfo(struct intel_screen * intelScreen, - __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) -{ - fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", - intelScreen->front.size, intelScreen->front.offset, - intelScreen->front.pitch); - fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); -} - -static void -intelPrintSAREA(const drmI830Sarea * sarea) -{ - fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, - sarea->height); - fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); - fprintf(stderr, - "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->front_offset, sarea->front_size, - (unsigned) sarea->front_handle); - fprintf(stderr, - "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->back_offset, sarea->back_size, - (unsigned) sarea->back_handle); - fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->depth_offset, sarea->depth_size, - (unsigned) sarea->depth_handle); - fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); - fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); - fprintf(stderr, - "SAREA: rotated offset: 0x%08x size: 0x%x\n", - sarea->rotated_offset, sarea->rotated_size); - fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); -} - -/** - * Use the information in the sarea to update the screen parameters - * related to screen rotation. Needs to be called locked. - */ -void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) -{ - struct intel_screen *intelScreen = intel_screen(sPriv); - - if (intelScreen->front.map) { - drmUnmap(intelScreen->front.map, intelScreen->front.size); - intelScreen->front.map = NULL; - } - - if (intelScreen->front.buffer) - driDeleteBuffers(1, &intelScreen->front.buffer); - - intelScreen->front.width = sarea->width; - intelScreen->front.height = sarea->height; - intelScreen->front.offset = sarea->front_offset; - intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; - intelScreen->front.size = sarea->front_size; - intelScreen->front.handle = sarea->front_handle; - - assert( sarea->front_size >= - intelScreen->front.pitch * intelScreen->front.height ); - -#if 0 /* JB not important */ - if (!sarea->front_handle) - return; - - if (drmMap(sPriv->fd, - sarea->front_handle, - intelScreen->front.size, - (drmAddress *) & intelScreen->front.map) != 0) { - fprintf(stderr, "drmMap(frontbuffer) failed!\n"); - return; - } -#endif - -#if 0 /* JB */ - if (intelScreen->staticPool) { - driGenBuffers(intelScreen->staticPool, "static region", 1, - &intelScreen->front.buffer, 64, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - - driBOSetStatic(intelScreen->front.buffer, - intelScreen->front.offset, - intelScreen->front.pitch * intelScreen->front.height, - intelScreen->front.map, 0); - } -#else - if (intelScreen->staticPool) { - if (intelScreen->front.buffer) - driBOUnReference(intelScreen->front.buffer); - driGenBuffers(intelScreen->staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); - driBOSetReferenced(intelScreen->front.buffer, sarea->front_bo_handle); - } -#endif -} - - -boolean -intelCreatePools(__DRIscreenPrivate * sPriv) -{ - //unsigned batchPoolSize = 1024*1024; - struct intel_screen *intelScreen = intel_screen(sPriv); - - if (intelScreen->havePools) - return GL_TRUE; - -#if 0 /* ZZZ JB fix this */ - intelScreen->staticPool = driDRMStaticPoolInit(sPriv->fd); - if (!intelScreen->staticPool) - return GL_FALSE; - - batchPoolSize /= BATCH_SZ; - intelScreen->batchPool = driBatchPoolInit(sPriv->fd, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT | - DRM_BO_FLAG_MEM_LOCAL, - BATCH_SZ, - batchPoolSize, 5); - if (!intelScreen->batchPool) { - fprintf(stderr, "Failed to initialize batch pool - possible incorrect agpgart installed\n"); - return GL_FALSE; - } -#else - intelScreen->staticPool = driDRMPoolInit(sPriv->fd); - intelScreen->batchPool = driSlabPoolInit(sPriv->fd, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - 4096, //intelScreen->maxBatchSize, - 1, 40, 16*16384, 0, - intelScreen->fMan); -#endif - intelScreen->havePools = GL_TRUE; - - //intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - return GL_TRUE; -} - - -static boolean -intelInitDriver(__DRIscreenPrivate * sPriv) -{ - struct intel_screen *intelScreen; - I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; - - PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = - (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> - getProcAddress("glxEnableExtension")); - void *const psc = sPriv->psc->screenConfigs; - - if (sPriv->devPrivSize != sizeof(I830DRIRec)) { - fprintf(stderr, - "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); - return GL_FALSE; - } - - /* Allocate the private area */ - intelScreen = CALLOC_STRUCT(intel_screen); - if (!intelScreen) - return GL_FALSE; - - /* parse information in __driConfigOptions */ - driParseOptionInfo(&intelScreen->optionCache, - __driConfigOptions, __driNConfigOptions); - - sPriv->private = (void *) intelScreen; - - intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + - gDRIPriv->sarea_priv_offset); - intelScreen->deviceID = gDRIPriv->deviceID; - intelScreen->front.cpp = gDRIPriv->cpp; - intelScreen->drmMinor = sPriv->drmMinor; - - - assert(gDRIPriv->bitsPerPixel == 16 || - gDRIPriv->bitsPerPixel == 32); - - - intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - if (0) - intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); - - if (glx_enable_extension != NULL) { - (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); - (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); - (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); - } - - - -#if 1 // ZZZ JB - intelScreen->mgr = driFenceMgrTTMInit(sPriv->fd); - if (!intelScreen->mgr) { - fprintf(stderr, "Failed to create fence manager.\n"); - return GL_FALSE; - } - - intelScreen->fMan = driInitFreeSlabManager(10, 10); - if (!intelScreen->fMan) { - fprintf(stderr, "Failed to create free slab manager.\n"); - return GL_FALSE; - } - - if (!intelCreatePools(sPriv)) - return GL_FALSE; -#endif - - intelScreen->winsys = intel_create_pipe_winsys(sPriv->fd, intelScreen->fMan); - - return GL_TRUE; -} - - -static void -intelDestroyScreen(__DRIscreenPrivate * sPriv) -{ - struct intel_screen *intelScreen = intel_screen(sPriv); - - /* intelUnmapScreenRegions(intelScreen); */ - - if (intelScreen->havePools) { - driPoolTakeDown(intelScreen->staticPool); - driPoolTakeDown(intelScreen->batchPool); - } - FREE(intelScreen); - sPriv->private = NULL; -} - - -/** - * This is called when we need to set up GL rendering to a new X window. - */ -static boolean -intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, - __DRIdrawablePrivate * driDrawPriv, - const __GLcontextModes * visual, boolean isPixmap) -{ - if (isPixmap) { - return GL_FALSE; /* not implemented */ - } - else { - enum pipe_format colorFormat, depthFormat, stencilFormat; - struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); - - if (!intelfb) - return GL_FALSE; - - if (visual->redBits == 5) - colorFormat = PIPE_FORMAT_R5G6B5_UNORM; - else - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (visual->depthBits == 16) - depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (visual->depthBits == 24) - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - else - depthFormat = PIPE_FORMAT_NONE; - - if (visual->stencilBits == 8) - stencilFormat = PIPE_FORMAT_S8Z24_UNORM; - else - stencilFormat = PIPE_FORMAT_NONE; - - intelfb->stfb = st_create_framebuffer(visual, - colorFormat, - depthFormat, - stencilFormat, - driDrawPriv->w, - driDrawPriv->h, - (void*) intelfb); - if (!intelfb->stfb) { - free(intelfb); - return GL_FALSE; - } - - driDrawPriv->driverPrivate = (void *) intelfb; - return GL_TRUE; - } -} - -static void -intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) -{ - struct intel_framebuffer *intelfb = intel_framebuffer(driDrawPriv); - assert(intelfb->stfb); - st_unreference_framebuffer(&intelfb->stfb); - free(intelfb); -} - - -/** - * Get information about previous buffer swaps. - */ -static int -intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) -{ - if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) - || (sInfo == NULL)) { - return -1; - } - - return 0; -} - - -static void -intelSetTexOffset(__DRIcontext *pDRICtx, int texname, - unsigned long long offset, int depth, uint pitch) -{ - abort(); -#if 0 - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct st_texture_object *stObj = st_texture_object(tObj); - - if (!stObj) - return; - - if (stObj->pt) - st->pipe->texture_release(intel->st->pipe, &stObj->pt); - - stObj->imageOverride = GL_TRUE; - stObj->depthOverride = depth; - stObj->pitchOverride = pitch; - - if (offset) - stObj->textureOffset = offset; -#endif -} - - -static const struct __DriverAPIRec intelAPI = { - .InitDriver = intelInitDriver, - .DestroyScreen = intelDestroyScreen, - .CreateContext = intelCreateContext, - .DestroyContext = intelDestroyContext, - .CreateBuffer = intelCreateBuffer, - .DestroyBuffer = intelDestroyBuffer, - .SwapBuffers = intelSwapBuffers, - .MakeCurrent = intelMakeCurrent, - .UnbindContext = intelUnbindContext, - .GetSwapInfo = intelGetSwapInfo, - .GetMSC = driGetMSC32, - .WaitForMSC = driWaitForMSC32, - .WaitForSBC = NULL, - .SwapBuffersMSC = NULL, - .CopySubBuffer = intelCopySubBuffer, - .setTexOffset = intelSetTexOffset, -}; - - -static __GLcontextModes * -intelFillInModes(unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, boolean have_back_buffer) -{ - __GLcontextModes *modes; - __GLcontextModes *m; - unsigned num_modes; - unsigned depth_buffer_factor; - unsigned back_buffer_factor; - GLenum fb_format; - GLenum fb_type; - - /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't - * support pageflipping at all. - */ - static const GLenum back_buffer_modes[] = { - GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML - }; - - u_int8_t depth_bits_array[3]; - u_int8_t stencil_bits_array[3]; - - - depth_bits_array[0] = 0; - depth_bits_array[1] = depth_bits; - depth_bits_array[2] = depth_bits; - - /* Just like with the accumulation buffer, always provide some modes - * with a stencil buffer. It will be a sw fallback, but some apps won't - * care about that. - */ - stencil_bits_array[0] = 0; - stencil_bits_array[1] = 0; - if (depth_bits == 24) - stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; - - stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; - - depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; - back_buffer_factor = (have_back_buffer) ? 3 : 1; - - num_modes = depth_buffer_factor * back_buffer_factor * 4; - - if (pixel_bits == 16) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; - } - - modes = - (*dri_interface->createContextModes) (num_modes, - sizeof(__GLcontextModes)); - m = modes; - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_TRUE_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_DIRECT_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - - /* Mark the visual as slow if there are "fake" stencil bits. - */ - for (m = modes; m != NULL; m = m->next) { - if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { - m->visualRating = GLX_SLOW_CONFIG; - } - } - - return modes; -} - - -/** - * This is the bootstrap function for the driver. libGL supplies all of the - * requisite information about the system, and the driver initializes itself. - * This routine also fills in the linked list pointed to by \c driver_modes - * with the \c __GLcontextModes that the driver can support for windows or - * pbuffers. - * - * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on - * failure. - */ -PUBLIC void * -__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, - __DRIscreen * psc, - const __GLcontextModes * modes, - const __DRIversion * ddx_version, - const __DRIversion * dri_version, - const __DRIversion * drm_version, - const __DRIframebuffer * frame_buffer, - drmAddress pSAREA, int fd, - int internal_api_version, - const __DRIinterfaceMethods * interface, - __GLcontextModes ** driver_modes) -{ - __DRIscreenPrivate *psp; - static const __DRIversion ddx_expected = { 1, 7, 0 }; - static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 7, 0 }; - - dri_interface = interface; - - if (!driCheckDriDdxDrmVersions2("i915", - dri_version, &dri_expected, - ddx_version, &ddx_expected, - drm_version, &drm_expected)) { - return NULL; - } - - psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, - ddx_version, dri_version, drm_version, - frame_buffer, pSAREA, fd, - internal_api_version, &intelAPI); - - if (psp != NULL) { - I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; - *driver_modes = intelFillInModes(dri_priv->cpp * 8, - (dri_priv->cpp == 2) ? 16 : 24, - (dri_priv->cpp == 2) ? 0 : 8, 1); - - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create - * is called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions(NULL, card_extensions, GL_FALSE); - } - - return (void *) psp; -} -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.h b/src/gallium/winsys/egl_drm/intel/intel_screen.h index e8c1cdfca4..87c4406f54 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_screen.h +++ b/src/gallium/winsys/egl_drm/intel/intel_screen.h @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,112 +22,26 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ #ifndef _INTEL_SCREEN_H_ #define _INTEL_SCREEN_H_ -#include "ws_dri_bufpool.h" +#include "intel_drm/intel_be_device.h" #include "pipe/p_compiler.h" -struct egl_drm_device *device; +struct pipe_screen; +struct egl_drm_device; struct intel_screen { -#if 0 - struct { - drm_handle_t handle; + struct intel_be_device base; + struct pipe_screen *pipe; - /* We create a static dri buffer for the frontbuffer. - */ - struct _DriBufferObject *buffer; - - char *map; /* memory map */ - int offset; /* from start of video mem, in bytes */ - int pitch; /* row stride, in bytes */ - int width; - int height; - int size; - int cpp; /* for front and back buffers */ - } front; -#endif - - int drmFB; - -#if 0 - int deviceID; - int drmMinor; - - - drmI830Sarea *sarea;*/ - - - /** - * Configuration cache with default values for all contexts - */ - driOptionCache optionCache; -#endif - - struct _DriBufferPool *batchPool; - struct _DriBufferPool *staticPool; /** for the X screen/framebuffer */ - boolean havePools; - -#if 0 - /** - * Temporary(?) context to use for SwapBuffers or other situations in - * which we need a rendering context, but none is currently bound. - */ - struct intel_context *dummyContext; -#endif - - /* - * New stuff form the i915tex integration - */ - struct _DriFenceMgr *mgr; - struct _DriFreeSlabManager *fMan; - unsigned batch_id; - - struct pipe_winsys *winsys; - struct egl_drm_device *device; - - /* batch buffer used for swap buffers */ - struct intel_batchbuffer *batch; + int deviceID; + struct egl_drm_device *device; }; - - -/** cast wrapper */ -#if 0 -static INLINE struct intel_screen * -intel_screen(__DRIscreenPrivate *sPriv) -{ - return (struct intel_screen *) sPriv->private; -} - - -extern void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); - - -extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); - -extern boolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); - -extern boolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); - - -extern boolean -intelCreatePools(__DRIscreenPrivate *sPriv); - -extern boolean -intelCreateContext(const __GLcontextModes * visual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - -#endif #endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c index 1ce4b2754a..24e55f1568 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c @@ -1,8 +1,8 @@ /************************************************************************** - * + * * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. - * + * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including @@ -10,11 +10,11 @@ * distribute, sub license, and/or sell copies of the Software, and to * permit persons to whom the Software is furnished to do so, subject to * the following conditions: - * + * * The above copyright notice and this permission notice (including the * next paragraph) shall be included in all copies or substantial portions * of the Software. - * + * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. @@ -22,15 +22,13 @@ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * + * **************************************************************************/ #include "intel_screen.h" #include "intel_context.h" -#include "intel_swapbuffers.h" #include "intel_batchbuffer.h" #include "intel_reg.h" -#include "intel_winsys.h" #include "pipe/p_context.h" #include "state_tracker/st_public.h" @@ -54,7 +52,8 @@ void intel_swap_buffers(struct egl_drm_drawable *draw) back_surf = st_get_framebuffer_surface(intel_fb->stfb, ST_SURFACE_BACK_LEFT); if (back_surf) { st_notify_swapbuffers(intel_fb->stfb); - intel_display_surface(draw, back_surf); + if (intel_fb->front) + intel_display_surface(draw, back_surf); st_notify_swapbuffers_complete(intel_fb->stfb); } } @@ -63,7 +62,7 @@ static void intel_display_surface(struct egl_drm_drawable *draw, struct pipe_surface *surf) { - struct intel_screen *intel = (struct intel_screen *)draw->device->priv; + struct intel_context *intel = NULL; struct intel_framebuffer *intel_fb = (struct intel_framebuffer *)draw->priv; struct _DriFenceObject *fence; @@ -85,8 +84,7 @@ intel_display_surface(struct egl_drm_drawable *draw, CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | XY_SRC_COPY_BLT_WRITE_RGB); - printf("srcPitch: %u, dstWidth: %u, dstHeight: %u, dstPitch: %u, cpp: %u\n", srcPitch, dstWidth, dstHeight, dstPitch, cpp); - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); + BEGIN_BATCH(8, 2); OUT_BATCH(CMD); OUT_BATCH(BR13); OUT_BATCH((0 << 16) | 0); @@ -102,226 +100,7 @@ intel_display_surface(struct egl_drm_drawable *draw, DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - ADVANCE_BATCH(); - - fence = intel_batchbuffer_flush(intel->batch); + fence = intel_be_batchbuffer_flush(intel->base.batch); driFenceUnReference(&fence); - intel_batchbuffer_finish(intel->batch); -} - -#if 0 -/** - * Display a colorbuffer surface in an X window. - * Used for SwapBuffers and flushing front buffer rendering. - * - * \param dPriv the window/drawable to display into - * \param surf the surface to display - * \param rect optional subrect of surface to display (may be NULL). - */ -void -intelDisplaySurface(__DRIdrawablePrivate *dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t *rect) -{ - struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - //struct intel_context *intel = intelScreen->dummyContext; - - DBG(SWAP, "%s\n", __FUNCTION__); - -#if 0 - if (!intel) { - /* XXX this is where some kind of extra/meta context could be useful */ - return; - } -#endif - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); - driFenceUnReference(&intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - * current context (which is what intelScreenContext should return) might - * not get a contended lock and thus cliprects not updated (tests/manywin) - */ - if (intel_context(dPriv->driContextPriv) != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - - if (dPriv && dPriv->numClipRects) { - const int srcWidth = surf->width; - const int srcHeight = surf->height; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; - const int cpp = intelScreen->front.cpp; - const int srcpitch = surf->pitch; - int BR13, CMD; - int i; - - ASSERT(surf->buffer); - ASSERT(surf->cpp == cpp); - - DBG(SWAP, "screen pitch %d src surface pitch %d\n", - pitch, surf->pitch); - - if (cpp == 2) { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - } - else { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - } - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) { - /* invalid cliprect, skip it */ - continue; - } - - box = *pbox; - - if (rect) { - /* intersect cliprect with user-provided src rect */ - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > srcWidth) - box.x2 = srcWidth + box.x1; - if (box.y2 - box.y1 > srcHeight) - box.y2 = srcHeight + box.y1; - - DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - assert(box.x1 < box.x2); - assert(box.y1 < box.y2); - - /* XXX this could be done with pipe->surface_copy() */ - BEGIN_BATCH(8, INTEL_BATCH_NO_CLIPRECTS); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((box.y1 << 16) | box.x1); - OUT_BATCH((box.y2 << 16) | box.x2); - - OUT_RELOC(intelScreen->front.buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((sbox.y1 << 16) | sbox.x1); - OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(dri_bo(surf->buffer), - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - ADVANCE_BATCH(); - } - - if (intel->first_swap_fence) - driFenceUnReference(&intel->first_swap_fence); - intel->first_swap_fence = intel_batchbuffer_flush(intel->batch); - } - - UNLOCK_HARDWARE(intel); - - if (intel->lastStamp != dPriv->lastStamp) { - intelUpdateWindowSize(dPriv); - intel->lastStamp = dPriv->lastStamp; - } -} - - - -/** - * This will be called whenever the currently bound window is moved/resized. - */ -void -intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) -{ - struct intel_framebuffer *intelfb = intel_framebuffer(dPriv); - assert(intelfb->stfb); - st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); -} - - - -void -intelSwapBuffers(__DRIdrawablePrivate * dPriv) -{ - struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; - - assert(intel_fb); - assert(intel_fb->stfb); - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, NULL); - st_notify_swapbuffers_complete(intel_fb->stfb); - } -} - - -/** - * Called via glXCopySubBufferMESA() to copy a subrect of the back - * buffer to the front buffer/screen. - */ -void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) -{ - struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; - - assert(intel_fb); - assert(intel_fb->stfb); - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - drm_clip_rect_t rect; - rect.x1 = x; - rect.y1 = y; - rect.x2 = w; - rect.y2 = h; - - st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, &rect); - } + intel_be_batchbuffer_finish(intel->base.batch); } -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.h b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.h deleted file mode 100644 index 904f26732e..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.h +++ /dev/null @@ -1,47 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_SWAPBUFFERS_H -#define INTEL_SWAPBUFFERS_H - - -struct pipe_surface; - -#if 0 -extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t * rect); - -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, - int x, int y, int w, int h); - -extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); -#endif - -#endif /* INTEL_SWAPBUFFERS_H */ diff --git a/src/gallium/winsys/egl_drm/intel/intel_winsys.h b/src/gallium/winsys/egl_drm/intel/intel_winsys.h deleted file mode 100644 index d0a319f9a4..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_winsys.h +++ /dev/null @@ -1,73 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_WINSYS_H -#define INTEL_WINSYS_H - -#include "pipe/p_state.h" - -struct intel_context; -struct pipe_context; -struct pipe_winsys; -struct pipe_buffer; -struct _DriBufferObject; - -struct pipe_winsys * -intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ); - -void -intel_destroy_pipe_winsys( struct pipe_winsys *winsys ); - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel, - struct pipe_winsys *winsys ); - -struct pipe_context * -intel_create_i915simple( struct intel_context *intel, - struct pipe_winsys *winsys ); - - -struct intel_buffer { - struct pipe_buffer base; - struct _DriBufferObject *driBO; -}; - -static INLINE struct intel_buffer * -intel_buffer( struct pipe_buffer *buf ) -{ - return (struct intel_buffer *)buf; -} - -static INLINE struct _DriBufferObject * -dri_bo( struct pipe_buffer *buf ) -{ - return intel_buffer(buf)->driBO; -} - - - -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_winsys_i915.c b/src/gallium/winsys/egl_drm/intel/intel_winsys_i915.c deleted file mode 100644 index 8ec5c7e82a..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_winsys_i915.c +++ /dev/null @@ -1,184 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" - -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_winsys.h" - -#include "pipe/p_util.h" -#include "pipe/p_winsys.h" -#include "i915simple/i915_winsys.h" -#include "i915simple/i915_screen.h" - - -struct intel_i915_winsys { - struct i915_winsys winsys; /**< batch buffer funcs */ - struct pipe_winsys *pws; - struct intel_context *intel; -}; - - -/* Turn a i915simple winsys into an intel/i915simple winsys: - */ -static inline struct intel_i915_winsys * -intel_i915_winsys( struct i915_winsys *sws ) -{ - return (struct intel_i915_winsys *)sws; -} - - -/* Simple batchbuffer interface: - */ - -static unsigned *intel_i915_batch_start( struct i915_winsys *sws, - unsigned dwords, - unsigned relocs ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - - /* XXX: check relocs. - */ - if (intel_batchbuffer_space( intel->batch ) >= dwords * 4) { - /* XXX: Hmm, the driver can't really do much with this pointer: - */ - return (unsigned *)intel->batch->ptr; - } - else - return NULL; -} - -static void intel_i915_batch_dword( struct i915_winsys *sws, - unsigned dword ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - intel_batchbuffer_emit_dword( intel->batch, dword ); -} - -static void intel_i915_batch_reloc( struct i915_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta ) -{ - struct intel_context *intel = intel_i915_winsys(sws)->intel; - unsigned flags = DRM_BO_FLAG_MEM_TT; - unsigned mask = DRM_BO_MASK_MEM; - - if (access_flags & I915_BUFFER_ACCESS_WRITE) { - flags |= DRM_BO_FLAG_WRITE; - mask |= DRM_BO_FLAG_WRITE; - } - - if (access_flags & I915_BUFFER_ACCESS_READ) { - flags |= DRM_BO_FLAG_READ; - mask |= DRM_BO_FLAG_READ; - } - -#if 0 /* JB old */ - intel_batchbuffer_emit_reloc( intel->batch, - dri_bo( buf ), - flags, mask, - delta ); -#else /* new */ - intel_offset_relocation( intel->batch, - delta, - dri_bo( buf ), - flags, - mask ); -#endif -} - - - -static void intel_i915_batch_flush( struct i915_winsys *sws, - struct pipe_fence_handle **fence ) -{ - struct intel_i915_winsys *iws = intel_i915_winsys(sws); - struct intel_context *intel = iws->intel; - union { - struct _DriFenceObject *dri; - struct pipe_fence_handle *pipe; - } fu; - - if (fence) - assert(!*fence); - - fu.dri = intel_batchbuffer_flush( intel->batch ); - - if (!fu.dri) { - assert(0); - *fence = NULL; - return; - } - - if (fu.dri) { - if (fence) - *fence = fu.pipe; - else - driFenceUnReference(&fu.dri); - } - -} - - -/** - * Create i915 hardware rendering context. - */ -struct pipe_context * -intel_create_i915simple( struct intel_context *intel, - struct pipe_winsys *winsys ) -{ - struct intel_i915_winsys *iws = CALLOC_STRUCT( intel_i915_winsys ); - struct pipe_screen *screen; - - /* Fill in this struct with callbacks that i915simple will need to - * communicate with the window system, buffer manager, etc. - */ - iws->winsys.batch_start = intel_i915_batch_start; - iws->winsys.batch_dword = intel_i915_batch_dword; - iws->winsys.batch_reloc = intel_i915_batch_reloc; - iws->winsys.batch_flush = intel_i915_batch_flush; - iws->pws = winsys; - iws->intel = intel; - - screen = i915_create_screen(winsys, PCI_CHIP_I945_GM); - assert(screen); - - /* Create the i915simple context: - */ - return i915_create_context( screen, - winsys, - &iws->winsys ); -} diff --git a/src/gallium/winsys/egl_drm/intel/intel_winsys_pipe.c b/src/gallium/winsys/egl_drm/intel/intel_winsys_pipe.c deleted file mode 100644 index 8bf8c21439..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_winsys_pipe.c +++ /dev/null @@ -1,338 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include -#include -//#include "dri_bufpool.h" -//#include "dri_bufmgr.h" - -#include "intel_context.h" -#include "intel_winsys.h" -#include "intel_swapbuffers.h" -#include "intel_batchbuffer.h" - -#include "pipe/p_winsys.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" - - - -struct intel_pipe_winsys { - struct pipe_winsys winsys; - struct _DriBufferPool *regionPool; - struct _DriFreeSlabManager *fMan; -}; - - - -/* Turn a pipe winsys into an intel/pipe winsys: - */ -static inline struct intel_pipe_winsys * -intel_pipe_winsys( struct pipe_winsys *winsys ) -{ - return (struct intel_pipe_winsys *)winsys; -} - - -/* Most callbacks map direcly onto dri_bufmgr operations: - */ -static void *intel_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned flags ) -{ - unsigned drm_flags = 0; - - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - drm_flags |= DRM_BO_FLAG_WRITE; - - if (flags & PIPE_BUFFER_USAGE_CPU_READ) - drm_flags |= DRM_BO_FLAG_READ; - - return driBOMap( dri_bo(buf), drm_flags, 0 ); -} - -static void intel_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnmap( dri_bo(buf) ); -} - - -static void -intel_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnReference( dri_bo(buf) ); - FREE(buf); -} - - -/* Pipe has no concept of pools. We choose the tex/region pool - * for all buffers. - * Grabs the hardware lock! - */ -static struct pipe_buffer * -intel_buffer_create(struct pipe_winsys *winsys, - unsigned alignment, - unsigned usage, - unsigned size ) -{ - struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); - struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); - unsigned flags = 0; - - buffer->base.refcount = 1; - buffer->base.alignment = alignment; - buffer->base.usage = usage; - buffer->base.size = size; - - if (usage & (PIPE_BUFFER_USAGE_VERTEX /*| IWS_BUFFER_USAGE_LOCAL*/)) { - flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; - } else { - flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; - } - - if (usage & PIPE_BUFFER_USAGE_GPU_READ) - flags |= DRM_BO_FLAG_READ; - - if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) - flags |= DRM_BO_FLAG_WRITE; - - /* drm complains if we don't set any read/write flags. - */ - if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) - flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; - -#if 0 - if (flags & IWS_BUFFER_USAGE_EXE) - flags |= DRM_BO_FLAG_EXE; - - if (usage & IWS_BUFFER_USAGE_CACHED) - flags |= DRM_BO_FLAG_CACHED; -#endif - - driGenBuffers( iws->regionPool, - "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); - - driBOData( buffer->driBO, size, NULL, iws->regionPool, 0 ); - - return &buffer->base; -} - - -static struct pipe_buffer * -intel_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) -{ - struct intel_buffer *buffer = CALLOC_STRUCT( intel_buffer ); - struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); - - driGenUserBuffer( iws->regionPool, - "pipe user buffer", &buffer->driBO, ptr, bytes ); - - buffer->base.refcount = 1; - - return &buffer->base; -} - - -/* The state tracker (should!) keep track of whether the fake - * frontbuffer has been touched by any rendering since the last time - * we copied its contents to the real frontbuffer. Our task is easy: - */ -static void -intel_flush_frontbuffer( struct pipe_winsys *winsys, - struct pipe_surface *surf, - void *context_private) -{ -#if 0 - struct intel_context *intel = (struct intel_context *) context_private; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelDisplaySurface(dPriv, surf, NULL); -#endif -} - - -static struct pipe_surface * -intel_i915_surface_alloc(struct pipe_winsys *winsys) -{ - struct pipe_surface *surf = CALLOC_STRUCT(pipe_surface); - if (surf) { - surf->refcount = 1; - surf->winsys = winsys; - } - return surf; -} - - -/** - * Round n up to next multiple. - */ -static INLINE unsigned -round_up(unsigned n, unsigned multiple) -{ - return (n + multiple - 1) & ~(multiple - 1); -} - -/** - * Copied from xm_winsys.c - */ -static int -intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, - struct pipe_surface *surf, - unsigned width, unsigned height, - enum pipe_format format, - unsigned flags) -{ - const unsigned alignment = 64; - //int ret; - - surf->width = width; - surf->height = height; - surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = round_up(width, alignment / surf->cpp); - - assert(!surf->buffer); - surf->buffer = winsys->buffer_create(winsys, alignment, - PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * height); - if(!surf->buffer) - return -1; - - return 0; -} - - -static void -intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) -{ - struct pipe_surface *surf = *s; - surf->refcount--; - if (surf->refcount == 0) { - if (surf->buffer) - pipe_buffer_reference(winsys, &surf->buffer, NULL); - free(surf); - } - *s = NULL; -} - - - -static const char * -intel_get_name( struct pipe_winsys *winsys ) -{ - return "Intel/EGL/ttm"; -} - -static void -intel_fence_reference( struct pipe_winsys *sws, - struct pipe_fence_handle **ptr, - struct pipe_fence_handle *fence ) -{ - if (*ptr) - driFenceUnReference((struct _DriFenceObject **)ptr); - - if (fence) - *ptr = (struct pipe_fence_handle *)driFenceReference((struct _DriFenceObject *)fence); -} - -static int -intel_fence_signalled( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ) -{ - return driFenceSignaled((struct _DriFenceObject *)fence, flag); -} - -static int -intel_fence_finish( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ) -{ - /* JB: Lets allways lazy wait */ - return driFenceFinish((struct _DriFenceObject *)fence, flag, 1); -} - -struct pipe_winsys * -intel_create_pipe_winsys( int fd, struct _DriFreeSlabManager *fMan ) -{ - struct intel_pipe_winsys *iws = CALLOC_STRUCT( intel_pipe_winsys ); - - /* Fill in this struct with callbacks that pipe will need to - * communicate with the window system, buffer manager, etc. - * - * Pipe would be happy with a malloc based memory manager, but - * the SwapBuffers implementation in this winsys driver requires - * that rendering be done to an appropriate _DriBufferObject. - */ - iws->winsys.buffer_create = intel_buffer_create; - iws->winsys.user_buffer_create = intel_user_buffer_create; - iws->winsys.buffer_map = intel_buffer_map; - iws->winsys.buffer_unmap = intel_buffer_unmap; - iws->winsys.buffer_destroy = intel_buffer_destroy; - iws->winsys.flush_frontbuffer = intel_flush_frontbuffer; - iws->winsys.get_name = intel_get_name; - iws->winsys.surface_alloc = intel_i915_surface_alloc; - iws->winsys.surface_alloc_storage = intel_i915_surface_alloc_storage; - iws->winsys.surface_release = intel_i915_surface_release; - - iws->winsys.fence_reference = intel_fence_reference; - iws->winsys.fence_signalled = intel_fence_signalled; - iws->winsys.fence_finish = intel_fence_finish; - - if (fd) - iws->regionPool = driSlabPoolInit(fd, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - 64, 6, 16, 4096, 0, - fMan); - - return &iws->winsys; -} - - -void -intel_destroy_pipe_winsys( struct pipe_winsys *winsys ) -{ - struct intel_pipe_winsys *iws = intel_pipe_winsys(winsys); - if (iws->regionPool) { - driPoolTakeDown(iws->regionPool); - } - free(iws); -} - diff --git a/src/gallium/winsys/egl_drm/intel/intel_winsys_softpipe.c b/src/gallium/winsys/egl_drm/intel/intel_winsys_softpipe.c deleted file mode 100644 index 0bc2dc4002..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_winsys_softpipe.c +++ /dev/null @@ -1,82 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include "intel_context.h" -#include "intel_winsys.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_format.h" -#include "softpipe/sp_winsys.h" - - -struct intel_softpipe_winsys { - struct softpipe_winsys sws; - struct intel_context *intel; -}; - -/** - * Return list of surface formats supported by this driver. - */ -static boolean -intel_is_format_supported(struct softpipe_winsys *sws, - enum pipe_format format) -{ - switch(format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - return TRUE; - default: - return FALSE; - } -} - - -/** - * Create rendering context which uses software rendering. - */ -struct pipe_context * -intel_create_softpipe( struct intel_context *intel, - struct pipe_winsys *winsys ) -{ - struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); - struct pipe_screen *screen = softpipe_create_screen(winsys); - - /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. - */ - isws->sws.is_format_supported = intel_is_format_supported; - isws->intel = intel; - - /* Create the softpipe context: - */ - return softpipe_create( screen, winsys, &isws->sws ); -} diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c b/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c deleted file mode 100644 index 1bc1089352..0000000000 --- a/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.c +++ /dev/null @@ -1,953 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - * Keith Whitwell - */ - -#include -#include -#include "glthread.h" -#include "errno.h" -#include "ws_dri_bufmgr.h" -#include "string.h" -#include "imports.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" - -/* - * This lock is here to protect drmBO structs changing underneath us during a - * validate list call, since validatelist cannot take individiual locks for - * each drmBO. Validatelist takes this lock in write mode. Any access to an - * individual drmBO should take this lock in read mode, since in that case, the - * driBufferObject mutex will protect the access. Locking order is - * driBufferObject mutex - > this rw lock. - */ - -_glthread_DECLARE_STATIC_MUTEX(bmMutex); -_glthread_DECLARE_STATIC_COND(bmCond); - -static int kernelReaders = 0; -static int num_buffers = 0; -static int num_user_buffers = 0; - -static drmBO *drmBOListBuf(void *iterator) -{ - drmBONode *node; - drmMMListHead *l = (drmMMListHead *) iterator; - node = DRMLISTENTRY(drmBONode, l, head); - return node->buf; -} - -static void *drmBOListIterator(drmBOList *list) -{ - void *ret = list->list.next; - - if (ret == &list->list) - return NULL; - return ret; -} - -static void *drmBOListNext(drmBOList *list, void *iterator) -{ - void *ret; - - drmMMListHead *l = (drmMMListHead *) iterator; - ret = l->next; - if (ret == &list->list) - return NULL; - return ret; -} - -static drmBONode *drmAddListItem(drmBOList *list, drmBO *item, - uint64_t arg0, - uint64_t arg1) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->free.next; - if (l == &list->free) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - return NULL; - } - list->numCurrent++; - } - else { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - } - node->buf = item; - node->arg0 = arg0; - node->arg1 = arg1; - DRMLISTADD(&node->head, &list->list); - list->numOnList++; - return node; -} - -static int drmAddValidateItem(drmBOList *list, drmBO *buf, uint64_t flags, - uint64_t mask, int *newItem) -{ - drmBONode *node, *cur; - drmMMListHead *l; - - *newItem = 0; - cur = NULL; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - if (node->buf == buf) { - cur = node; - break; - } - } - if (!cur) { - cur = drmAddListItem(list, buf, flags, mask); - if (!cur) { - return -ENOMEM; - } - *newItem = 1; - cur->arg0 = flags; - cur->arg1 = mask; - } - else { - uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; - uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; - - if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { - return -EINVAL; - } - - cur->arg1 |= mask; - cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); - - if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && - (cur->arg0 & DRM_BO_MASK_MEM) == 0) { - return -EINVAL; - } - } - return 0; -} - -static void drmBOFreeList(drmBOList *list) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->list.next; - while(l != &list->list) { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - l = list->list.next; - list->numCurrent--; - list->numOnList--; - } - - l = list->free.next; - while(l != &list->free) { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - l = list->free.next; - list->numCurrent--; - } -} - -static int drmAdjustListNodes(drmBOList *list) -{ - drmBONode *node; - drmMMListHead *l; - int ret = 0; - - while(list->numCurrent < list->numTarget) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - ret = -ENOMEM; - break; - } - list->numCurrent++; - DRMLISTADD(&node->head, &list->free); - } - - while(list->numCurrent > list->numTarget) { - l = list->free.next; - if (l == &list->free) - break; - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - list->numCurrent--; - } - return ret; -} - -static int drmBOCreateList(int numTarget, drmBOList *list) -{ - DRMINITLISTHEAD(&list->list); - DRMINITLISTHEAD(&list->free); - list->numTarget = numTarget; - list->numCurrent = 0; - list->numOnList = 0; - return drmAdjustListNodes(list); -} - -static int drmBOResetList(drmBOList *list) -{ - drmMMListHead *l; - int ret; - - ret = drmAdjustListNodes(list); - if (ret) - return ret; - - l = list->list.next; - while (l != &list->list) { - DRMLISTDEL(l); - DRMLISTADD(l, &list->free); - list->numOnList--; - l = list->list.next; - } - return drmAdjustListNodes(list); -} - -void driWriteLockKernelBO(void) -{ - _glthread_LOCK_MUTEX(bmMutex); - while(kernelReaders != 0) - _glthread_COND_WAIT(bmCond, bmMutex); -} - -void driWriteUnlockKernelBO(void) -{ - _glthread_UNLOCK_MUTEX(bmMutex); -} - -void driReadLockKernelBO(void) -{ - _glthread_LOCK_MUTEX(bmMutex); - kernelReaders++; - _glthread_UNLOCK_MUTEX(bmMutex); -} - -void driReadUnlockKernelBO(void) -{ - _glthread_LOCK_MUTEX(bmMutex); - if (--kernelReaders == 0) - _glthread_COND_BROADCAST(bmCond); - _glthread_UNLOCK_MUTEX(bmMutex); -} - - - - -/* - * TODO: Introduce fence pools in the same way as - * buffer object pools. - */ - -typedef struct _DriBufferObject -{ - DriBufferPool *pool; - _glthread_Mutex mutex; - int refCount; - const char *name; - uint64_t flags; - unsigned hint; - unsigned alignment; - unsigned createdByReference; - void *private; - /* user-space buffer: */ - unsigned userBuffer; - void *userData; - unsigned userSize; -} DriBufferObject; - -typedef struct _DriBufferList { - drmBOList drmBuffers; /* List of kernel buffers needing validation */ - drmBOList driBuffers; /* List of user-space buffers needing validation */ -} DriBufferList; - - -void -bmError(int val, const char *file, const char *function, int line) -{ - _mesa_printf("Fatal video memory manager error \"%s\".\n" - "Check kernel logs or set the LIBGL_DEBUG\n" - "environment variable to \"verbose\" for more info.\n" - "Detected in file %s, line %d, function %s.\n", - strerror(-val), file, line, function); -#ifndef NDEBUG - abort(); -#else - abort(); -#endif -} - -extern drmBO * -driBOKernel(struct _DriBufferObject *buf) -{ - drmBO *ret; - - driReadLockKernelBO(); - _glthread_LOCK_MUTEX(buf->mutex); - assert(buf->private != NULL); - ret = buf->pool->kernel(buf->pool, buf->private); - if (!ret) - BM_CKFATAL(-EINVAL); - _glthread_UNLOCK_MUTEX(buf->mutex); - driReadUnlockKernelBO(); - - return ret; -} - -void -driBOWaitIdle(struct _DriBufferObject *buf, int lazy) -{ - - /* - * This function may block. Is it sane to keep the mutex held during - * that time?? - */ - - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy)); - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -void * -driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint) -{ - void *virtual; - int retval; - - if (buf->userBuffer) { - return buf->userData; - } - - _glthread_LOCK_MUTEX(buf->mutex); - assert(buf->private != NULL); - retval = buf->pool->map(buf->pool, buf->private, flags, hint, - &buf->mutex, &virtual); - _glthread_UNLOCK_MUTEX(buf->mutex); - - return retval == 0 ? virtual : NULL; -} - -void -driBOUnmap(struct _DriBufferObject *buf) -{ - if (buf->userBuffer) - return; - - assert(buf->private != NULL); - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -unsigned long -driBOOffset(struct _DriBufferObject *buf) -{ - unsigned long ret; - - assert(buf->private != NULL); - - _glthread_LOCK_MUTEX(buf->mutex); - ret = buf->pool->offset(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - return ret; -} - -unsigned long -driBOPoolOffset(struct _DriBufferObject *buf) -{ - unsigned long ret; - - assert(buf->private != NULL); - - _glthread_LOCK_MUTEX(buf->mutex); - ret = buf->pool->poolOffset(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - return ret; -} - -uint64_t -driBOFlags(struct _DriBufferObject *buf) -{ - uint64_t ret; - - assert(buf->private != NULL); - - driReadLockKernelBO(); - _glthread_LOCK_MUTEX(buf->mutex); - ret = buf->pool->flags(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - driReadUnlockKernelBO(); - return ret; -} - -struct _DriBufferObject * -driBOReference(struct _DriBufferObject *buf) -{ - _glthread_LOCK_MUTEX(buf->mutex); - if (++buf->refCount == 1) { - _glthread_UNLOCK_MUTEX(buf->mutex); - BM_CKFATAL(-EINVAL); - } - _glthread_UNLOCK_MUTEX(buf->mutex); - return buf; -} - -void -driBOUnReference(struct _DriBufferObject *buf) -{ - int tmp; - - if (!buf) - return; - - _glthread_LOCK_MUTEX(buf->mutex); - tmp = --buf->refCount; - if (!tmp) { - _glthread_UNLOCK_MUTEX(buf->mutex); - if (buf->private) { - if (buf->createdByReference) - buf->pool->unreference(buf->pool, buf->private); - else - buf->pool->destroy(buf->pool, buf->private); - } - if (buf->userBuffer) - num_user_buffers--; - else - num_buffers--; - free(buf); - } else - _glthread_UNLOCK_MUTEX(buf->mutex); - -} - - -int -driBOData(struct _DriBufferObject *buf, - unsigned size, const void *data, - DriBufferPool *newPool, - uint64_t flags) -{ - void *virtual = NULL; - int newBuffer; - int retval = 0; - struct _DriBufferPool *pool; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - _glthread_LOCK_MUTEX(buf->mutex); - pool = buf->pool; - - if (pool == NULL && newPool != NULL) { - buf->pool = newPool; - pool = newPool; - } - if (newPool == NULL) - newPool = pool; - - if (!pool->create) { - _mesa_error(NULL, GL_INVALID_OPERATION, - "driBOData called on invalid buffer\n"); - BM_CKFATAL(-EINVAL); - } - - newBuffer = (!buf->private || pool != newPool || - pool->size(pool, buf->private) < size); - - if (!flags) - flags = buf->flags; - - if (newBuffer) { - - if (buf->createdByReference) { - _mesa_error(NULL, GL_INVALID_OPERATION, - "driBOData requiring resizing called on " - "shared buffer.\n"); - BM_CKFATAL(-EINVAL); - } - - if (buf->private) - buf->pool->destroy(buf->pool, buf->private); - - pool = newPool; - buf->pool = newPool; - buf->private = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, - buf->alignment); - if (!buf->private) - retval = -ENOMEM; - - if (retval == 0) - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, - DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual); - } else if (pool->map(pool, buf->private, DRM_BO_FLAG_WRITE, - DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual)) { - /* - * Buffer is busy. need to create a new one. - */ - - void *newBuf; - - newBuf = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, - buf->alignment); - if (newBuf) { - buf->pool->destroy(buf->pool, buf->private); - buf->private = newBuf; - } - - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); - } else { - uint64_t flag_diff = flags ^ buf->flags; - - /* - * We might need to change buffer flags. - */ - - if (flag_diff){ - assert(pool->setStatus != NULL); - BM_CKFATAL(pool->unmap(pool, buf->private)); - BM_CKFATAL(pool->setStatus(pool, buf->private, flag_diff, - buf->flags)); - if (!data) - goto out; - - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); - } - } - - if (retval == 0) { - if (data) - memcpy(virtual, data, size); - - BM_CKFATAL(pool->unmap(pool, buf->private)); - } - - out: - _glthread_UNLOCK_MUTEX(buf->mutex); - - return retval; -} - -void -driBOSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, const void *data) -{ - void *virtual; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - _glthread_LOCK_MUTEX(buf->mutex); - if (size && data) { - BM_CKFATAL(buf->pool->map(buf->pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, - &virtual)); - memcpy((unsigned char *) virtual + offset, data, size); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - } - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -void -driBOGetSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, void *data) -{ - void *virtual; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - _glthread_LOCK_MUTEX(buf->mutex); - if (size && data) { - BM_CKFATAL(buf->pool->map(buf->pool, buf->private, - DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual)); - memcpy(data, (unsigned char *) virtual + offset, size); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - } - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -void -driBOSetReferenced(struct _DriBufferObject *buf, - unsigned long handle) -{ - _glthread_LOCK_MUTEX(buf->mutex); - if (buf->private != NULL) { - _mesa_error(NULL, GL_INVALID_OPERATION, - "Invalid buffer for setReferenced\n"); - BM_CKFATAL(-EINVAL); - - } - if (buf->pool->reference == NULL) { - _mesa_error(NULL, GL_INVALID_OPERATION, - "Invalid buffer pool for setReferenced\n"); - BM_CKFATAL(-EINVAL); - } - buf->private = buf->pool->reference(buf->pool, handle); - if (!buf->private) { - _mesa_error(NULL, GL_OUT_OF_MEMORY, - "Invalid buffer pool for setStatic\n"); - BM_CKFATAL(-ENOMEM); - } - buf->createdByReference = GL_TRUE; - buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags; - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -int -driGenBuffers(struct _DriBufferPool *pool, - const char *name, - unsigned n, - struct _DriBufferObject *buffers[], - unsigned alignment, uint64_t flags, unsigned hint) -{ - struct _DriBufferObject *buf; - int i; - - flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM | - DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; - - ++num_buffers; - - assert(pool); - - for (i = 0; i < n; ++i) { - buf = (struct _DriBufferObject *) calloc(1, sizeof(*buf)); - if (!buf) - return -ENOMEM; - - _glthread_INIT_MUTEX(buf->mutex); - _glthread_LOCK_MUTEX(buf->mutex); - buf->refCount = 1; - buf->flags = flags; - buf->hint = hint; - buf->name = name; - buf->alignment = alignment; - buf->pool = pool; - buf->createdByReference = 0; - _glthread_UNLOCK_MUTEX(buf->mutex); - buffers[i] = buf; - } - return 0; -} - -void -driGenUserBuffer(struct _DriBufferPool *pool, - const char *name, - struct _DriBufferObject **buffers, - void *ptr, unsigned bytes) -{ - const unsigned alignment = 1, flags = 0, hint = 0; - - --num_buffers; /* JB: is inced in GenBuffes */ - driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); - ++num_user_buffers; - - (*buffers)->userBuffer = 1; - (*buffers)->userData = ptr; - (*buffers)->userSize = bytes; -} - -void -driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]) -{ - int i; - - for (i = 0; i < n; ++i) { - driBOUnReference(buffers[i]); - } -} - - -void -driInitBufMgr(int fd) -{ - ; -} - -/* - * Note that lists are per-context and don't need mutex protection. - */ - -struct _DriBufferList * -driBOCreateList(int target) -{ - struct _DriBufferList *list = calloc(sizeof(*list), 1); - - BM_CKFATAL(drmBOCreateList(target, &list->drmBuffers)); - BM_CKFATAL(drmBOCreateList(target, &list->driBuffers)); - return list; -} - -int -driBOResetList(struct _DriBufferList * list) -{ - int ret; - ret = drmBOResetList(&list->drmBuffers); - if (ret) - return ret; - ret = drmBOResetList(&list->driBuffers); - return ret; -} - -void -driBOFreeList(struct _DriBufferList * list) -{ - drmBOFreeList(&list->drmBuffers); - drmBOFreeList(&list->driBuffers); - free(list); -} - - -/* - * Copied from libdrm, because it is needed by driAddValidateItem. - */ - -static drmBONode * -driAddListItem(drmBOList * list, drmBO * item, - uint64_t arg0, uint64_t arg1) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->free.next; - if (l == &list->free) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - return NULL; - } - list->numCurrent++; - } else { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - } - memset(&node->bo_arg, 0, sizeof(node->bo_arg)); - node->buf = item; - node->arg0 = arg0; - node->arg1 = arg1; - DRMLISTADDTAIL(&node->head, &list->list); - list->numOnList++; - return node; -} - -/* - * Slightly modified version compared to the libdrm version. - * This one returns the list index of the buffer put on the list. - */ - -static int -driAddValidateItem(drmBOList * list, drmBO * buf, uint64_t flags, - uint64_t mask, int *itemLoc, - struct _drmBONode **pnode) -{ - drmBONode *node, *cur; - drmMMListHead *l; - int count = 0; - - cur = NULL; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - if (node->buf == buf) { - cur = node; - break; - } - count++; - } - if (!cur) { - cur = driAddListItem(list, buf, flags, mask); - if (!cur) - return -ENOMEM; - - cur->arg0 = flags; - cur->arg1 = mask; - } else { - uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; - uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; - - if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { - return -EINVAL; - } - - cur->arg1 |= mask; - cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); - - if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && - (cur->arg0 & DRM_BO_MASK_MEM) == 0) { - return -EINVAL; - } - } - *itemLoc = count; - *pnode = cur; - return 0; -} - - -void -driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf, - uint64_t flags, uint64_t mask, int *itemLoc, - struct _drmBONode **node) -{ - int newItem; - - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(driAddValidateItem(&list->drmBuffers, - buf->pool->kernel(buf->pool, buf->private), - flags, mask, itemLoc, node)); - BM_CKFATAL(drmAddValidateItem(&list->driBuffers, (drmBO *) buf, - flags, mask, &newItem)); - if (newItem) - buf->refCount++; - - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -drmBOList *driGetdrmBOList(struct _DriBufferList *list) -{ - driWriteLockKernelBO(); - return &list->drmBuffers; -} - -void driPutdrmBOList(struct _DriBufferList *list) -{ - driWriteUnlockKernelBO(); -} - - -void -driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence) -{ - _glthread_LOCK_MUTEX(buf->mutex); - if (buf->pool->fence) - BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence)); - _glthread_UNLOCK_MUTEX(buf->mutex); - -} - -void -driBOUnrefUserList(struct _DriBufferList *list) -{ - struct _DriBufferObject *buf; - void *curBuf; - - curBuf = drmBOListIterator(&list->driBuffers); - while (curBuf) { - buf = (struct _DriBufferObject *)drmBOListBuf(curBuf); - driBOUnReference(buf); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } -} - -struct _DriFenceObject * -driBOFenceUserList(struct _DriFenceMgr *mgr, - struct _DriBufferList *list, const char *name, - drmFence *kFence) -{ - struct _DriFenceObject *fence; - struct _DriBufferObject *buf; - void *curBuf; - - fence = driFenceCreate(mgr, kFence->fence_class, kFence->type, - kFence, sizeof(*kFence)); - curBuf = drmBOListIterator(&list->driBuffers); - - /* - * User-space fencing callbacks. - */ - - while (curBuf) { - buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); - driBOFence(buf, fence); - driBOUnReference(buf); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } - - driBOResetList(list); - return fence; -} - -void -driBOValidateUserList(struct _DriBufferList * list) -{ - void *curBuf; - struct _DriBufferObject *buf; - - curBuf = drmBOListIterator(&list->driBuffers); - - /* - * User-space validation callbacks. - */ - - while (curBuf) { - buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); - _glthread_LOCK_MUTEX(buf->mutex); - if (buf->pool->validate) - BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex)); - _glthread_UNLOCK_MUTEX(buf->mutex); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } -} - - -void -driPoolTakeDown(struct _DriBufferPool *pool) -{ - pool->takeDown(pool); - -} - -unsigned long -driBOSize(struct _DriBufferObject *buf) -{ - unsigned long size; - - _glthread_LOCK_MUTEX(buf->mutex); - size = buf->pool->size(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - - return size; - -} - -drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list) -{ - return &list->drmBuffers; -} - -drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list) -{ - return &list->driBuffers; -} - diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h b/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h deleted file mode 100644 index fdaf5ee93a..0000000000 --- a/src/gallium/winsys/egl_drm/intel/ws_dri_bufmgr.h +++ /dev/null @@ -1,138 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - * Keith Whitwell - */ - -#ifndef _PSB_BUFMGR_H_ -#define _PSB_BUFMGR_H_ -#include -#include "i915_drm.h" -#include "ws_dri_fencemgr.h" - -typedef struct _drmBONode -{ - drmMMListHead head; - drmBO *buf; - struct drm_i915_op_arg bo_arg; - uint64_t arg0; - uint64_t arg1; -} drmBONode; - -typedef struct _drmBOList { - unsigned numTarget; - unsigned numCurrent; - unsigned numOnList; - drmMMListHead list; - drmMMListHead free; -} drmBOList; - - -struct _DriFenceObject; -struct _DriBufferObject; -struct _DriBufferPool; -struct _DriBufferList; - -/* - * Return a pointer to the libdrm buffer object this DriBufferObject - * uses. - */ - -extern drmBO *driBOKernel(struct _DriBufferObject *buf); -extern void *driBOMap(struct _DriBufferObject *buf, unsigned flags, - unsigned hint); -extern void driBOUnmap(struct _DriBufferObject *buf); -extern unsigned long driBOOffset(struct _DriBufferObject *buf); -extern unsigned long driBOPoolOffset(struct _DriBufferObject *buf); - -extern uint64_t driBOFlags(struct _DriBufferObject *buf); -extern struct _DriBufferObject *driBOReference(struct _DriBufferObject *buf); -extern void driBOUnReference(struct _DriBufferObject *buf); - -extern int driBOData(struct _DriBufferObject *r_buf, - unsigned size, const void *data, - struct _DriBufferPool *pool, uint64_t flags); - -extern void driBOSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, - const void *data); -extern void driBOGetSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, - void *data); -extern int driGenBuffers(struct _DriBufferPool *pool, - const char *name, - unsigned n, - struct _DriBufferObject *buffers[], - unsigned alignment, uint64_t flags, unsigned hint); -extern void driGenUserBuffer(struct _DriBufferPool *pool, - const char *name, - struct _DriBufferObject *buffers[], - void *ptr, unsigned bytes); -extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]); -extern void driInitBufMgr(int fd); -extern struct _DriBufferList *driBOCreateList(int target); -extern int driBOResetList(struct _DriBufferList * list); -extern void driBOAddListItem(struct _DriBufferList * list, - struct _DriBufferObject *buf, - uint64_t flags, uint64_t mask, int *itemLoc, - struct _drmBONode **node); - -extern void driBOValidateList(int fd, struct _DriBufferList * list); -extern void driBOFreeList(struct _DriBufferList * list); -extern struct _DriFenceObject *driBOFenceUserList(struct _DriFenceMgr *mgr, - struct _DriBufferList *list, - const char *name, - drmFence *kFence); -extern void driBOUnrefUserList(struct _DriBufferList *list); -extern void driBOValidateUserList(struct _DriBufferList * list); -extern drmBOList *driGetdrmBOList(struct _DriBufferList *list); -extern void driPutdrmBOList(struct _DriBufferList *list); - -extern void driBOFence(struct _DriBufferObject *buf, - struct _DriFenceObject *fence); - -extern void driPoolTakeDown(struct _DriBufferPool *pool); -extern void driBOSetReferenced(struct _DriBufferObject *buf, - unsigned long handle); -unsigned long driBOSize(struct _DriBufferObject *buf); -extern void driBOWaitIdle(struct _DriBufferObject *buf, int lazy); -extern void driPoolTakeDown(struct _DriBufferPool *pool); - -extern void driReadLockKernelBO(void); -extern void driReadUnlockKernelBO(void); -extern void driWriteLockKernelBO(void); -extern void driWriteUnlockKernelBO(void); - -/* - * For debugging purposes. - */ - -extern drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list); -extern drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list); -#endif diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h b/src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h deleted file mode 100644 index 3a302e13d3..0000000000 --- a/src/gallium/winsys/egl_drm/intel/ws_dri_bufpool.h +++ /dev/null @@ -1,102 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#ifndef _PSB_BUFPOOL_H_ -#define _PSB_BUFPOOL_H_ - -#include -#include -struct _DriFenceObject; - -typedef struct _DriBufferPool -{ - int fd; - int (*map) (struct _DriBufferPool * pool, void *private, - unsigned flags, int hint, _glthread_Mutex *mutex, - void **virtual); - int (*unmap) (struct _DriBufferPool * pool, void *private); - int (*destroy) (struct _DriBufferPool * pool, void *private); - unsigned long (*offset) (struct _DriBufferPool * pool, void *private); - unsigned long (*poolOffset) (struct _DriBufferPool * pool, void *private); - uint64_t (*flags) (struct _DriBufferPool * pool, void *private); - unsigned long (*size) (struct _DriBufferPool * pool, void *private); - void *(*create) (struct _DriBufferPool * pool, unsigned long size, - uint64_t flags, unsigned hint, unsigned alignment); - void *(*reference) (struct _DriBufferPool * pool, unsigned handle); - int (*unreference) (struct _DriBufferPool * pool, void *private); - int (*fence) (struct _DriBufferPool * pool, void *private, - struct _DriFenceObject * fence); - drmBO *(*kernel) (struct _DriBufferPool * pool, void *private); - int (*validate) (struct _DriBufferPool * pool, void *private, _glthread_Mutex *mutex); - int (*waitIdle) (struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, - int lazy); - int (*setStatus) (struct _DriBufferPool *pool, void *private, - uint64_t flag_diff, uint64_t old_flags); - void (*takeDown) (struct _DriBufferPool * pool); - void *data; -} DriBufferPool; - -extern void bmError(int val, const char *file, const char *function, - int line); -#define BM_CKFATAL(val) \ - do{ \ - int tstVal = (val); \ - if (tstVal) \ - bmError(tstVal, __FILE__, __FUNCTION__, __LINE__); \ - } while(0); - - -/* - * Builtin pools. - */ - -/* - * Kernel buffer objects. Size in multiples of page size. Page size aligned. - */ - -extern struct _DriBufferPool *driDRMPoolInit(int fd); -extern struct _DriBufferPool *driMallocPoolInit(void); - -struct _DriFreeSlabManager; -extern struct _DriBufferPool * driSlabPoolInit(int fd, uint64_t flags, - uint64_t validMask, - uint32_t smallestSize, - uint32_t numSizes, - uint32_t desiredNumBuffers, - uint32_t maxSlabSize, - uint32_t pageAlignment, - struct _DriFreeSlabManager *fMan); -extern void driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan); -extern struct _DriFreeSlabManager * -driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec); - - -#endif diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c b/src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c deleted file mode 100644 index 7c55dbc674..0000000000 --- a/src/gallium/winsys/egl_drm/intel/ws_dri_drmpool.c +++ /dev/null @@ -1,268 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#include -#include -#include -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" -#include "assert.h" - -/* - * Buffer pool implementation using DRM buffer objects as DRI buffer objects. - */ - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, uint64_t flags, unsigned hint, - unsigned alignment) -{ - drmBO *buf = (drmBO *) malloc(sizeof(*buf)); - int ret; - unsigned pageSize = getpagesize(); - - if (!buf) - return NULL; - - if ((alignment > pageSize) && (alignment % pageSize)) { - free(buf); - return NULL; - } - - ret = drmBOCreate(pool->fd, size, alignment / pageSize, - NULL, - flags, hint, buf); - if (ret) { - free(buf); - return NULL; - } - - return (void *) buf; -} - -static void * -pool_reference(struct _DriBufferPool *pool, unsigned handle) -{ - drmBO *buf = (drmBO *) malloc(sizeof(*buf)); - int ret; - - if (!buf) - return NULL; - - ret = drmBOReference(pool->fd, handle, buf); - - if (ret) { - free(buf); - return NULL; - } - - return (void *) buf; -} - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - int ret; - drmBO *buf = (drmBO *) private; - driReadLockKernelBO(); - ret = drmBOUnreference(pool->fd, buf); - free(buf); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_unreference(struct _DriBufferPool *pool, void *private) -{ - int ret; - drmBO *buf = (drmBO *) private; - driReadLockKernelBO(); - ret = drmBOUnreference(pool->fd, buf); - free(buf); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, _glthread_Mutex *mutex, void **virtual) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOMap(pool->fd, buf, flags, hint, virtual); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOUnmap(pool->fd, buf); - driReadUnlockKernelBO(); - - return ret; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - unsigned long offset; - - driReadLockKernelBO(); - assert(buf->flags & DRM_BO_FLAG_NO_MOVE); - offset = buf->offset; - driReadUnlockKernelBO(); - - return buf->offset; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - return 0; -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - uint64_t flags; - - driReadLockKernelBO(); - flags = buf->flags; - driReadUnlockKernelBO(); - - return flags; -} - - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - unsigned long size; - - driReadLockKernelBO(); - size = buf->size; - driReadUnlockKernelBO(); - - return buf->size; -} - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - /* - * Noop. The kernel handles all fencing. - */ - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - return (drmBO *) private; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, - int lazy) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOWaitIdle(pool->fd, buf, (lazy) ? DRM_BO_HINT_WAIT_LAZY:0); - driReadUnlockKernelBO(); - - return ret; -} - - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - free(pool); -} - -/*static int -pool_setStatus(struct _DriBufferPool *pool, void *private, - uint64_t flag_diff, uint64_t old_flags) -{ - drmBO *buf = (drmBO *) private; - uint64_t new_flags = old_flags ^ flag_diff; - int ret; - - driReadLockKernelBO(); - ret = drmBOSetStatus(pool->fd, buf, new_flags, flag_diff, - 0, 0, 0); - driReadUnlockKernelBO(); - return ret; -}*/ - -struct _DriBufferPool * -driDRMPoolInit(int fd) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - - if (!pool) - return NULL; - - pool->fd = fd; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->poolOffset = &pool_poolOffset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = NULL; - pool->waitIdle = &pool_waitIdle; - pool->takeDown = &pool_takedown; - pool->reference = &pool_reference; - pool->unreference = &pool_unreference; - pool->data = NULL; - return pool; -} diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c b/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c deleted file mode 100644 index 1f893b47ce..0000000000 --- a/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.c +++ /dev/null @@ -1,377 +0,0 @@ -#include "ws_dri_fencemgr.h" -#include "glthread.h" -#include -#include -#include - -/* - * Note: Locking order is - * _DriFenceObject::mutex - * _DriFenceMgr::mutex - */ - -struct _DriFenceMgr { - /* - * Constant members. Need no mutex protection. - */ - struct _DriFenceMgrCreateInfo info; - void *private; - - /* - * These members are protected by this->mutex - */ - _glthread_Mutex mutex; - int refCount; - drmMMListHead *heads; - int num_fences; -}; - -struct _DriFenceObject { - - /* - * These members are constant and need no mutex protection. - */ - struct _DriFenceMgr *mgr; - uint32_t fence_class; - uint32_t fence_type; - - /* - * These members are protected by mgr->mutex. - */ - drmMMListHead head; - int refCount; - - /* - * These members are protected by this->mutex. - */ - _glthread_Mutex mutex; - uint32_t signaled_type; - void *private; -}; - -uint32_t -driFenceType(struct _DriFenceObject *fence) -{ - return fence->fence_type; -} - -struct _DriFenceMgr * -driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info) -{ - struct _DriFenceMgr *tmp; - uint32_t i; - - tmp = calloc(1, sizeof(*tmp)); - if (!tmp) - return NULL; - - _glthread_INIT_MUTEX(tmp->mutex); - _glthread_LOCK_MUTEX(tmp->mutex); - tmp->refCount = 1; - tmp->info = *info; - tmp->num_fences = 0; - tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads)); - if (!tmp->heads) - goto out_err; - - for (i=0; iinfo.num_classes; ++i) { - DRMINITLISTHEAD(&tmp->heads[i]); - } - _glthread_UNLOCK_MUTEX(tmp->mutex); - return tmp; - - out_err: - if (tmp) - free(tmp); - return NULL; -} - -static void -driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr) -{ - struct _DriFenceMgr *mgr = *pMgr; - - *pMgr = NULL; - if (--mgr->refCount == 0) - free(mgr); - else - _glthread_UNLOCK_MUTEX(mgr->mutex); -} - -void -driFenceMgrUnReference(struct _DriFenceMgr **pMgr) -{ - _glthread_LOCK_MUTEX((*pMgr)->mutex); - driFenceMgrUnrefUnlock(pMgr); -} - -static void -driFenceUnReferenceLocked(struct _DriFenceObject **pFence) -{ - struct _DriFenceObject *fence = *pFence; - struct _DriFenceMgr *mgr = fence->mgr; - - *pFence = NULL; - if (--fence->refCount == 0) { - DRMLISTDELINIT(&fence->head); - if (fence->private) - mgr->info.unreference(mgr, &fence->private); - --mgr->num_fences; - fence->mgr = NULL; - --mgr->refCount; - free(fence); - - } -} - - -static void -driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr, - drmMMListHead *list, - uint32_t fence_class, - uint32_t fence_type) -{ - struct _DriFenceObject *entry; - drmMMListHead *prev; - - while(list != &mgr->heads[fence_class]) { - entry = DRMLISTENTRY(struct _DriFenceObject, list, head); - - /* - * Up refcount so that entry doesn't disappear from under us - * when we unlock-relock mgr to get the correct locking order. - */ - - ++entry->refCount; - _glthread_UNLOCK_MUTEX(mgr->mutex); - _glthread_LOCK_MUTEX(entry->mutex); - _glthread_LOCK_MUTEX(mgr->mutex); - - prev = list->prev; - - - - if (list->prev == list) { - - /* - * Somebody else removed the entry from the list. - */ - - _glthread_UNLOCK_MUTEX(entry->mutex); - driFenceUnReferenceLocked(&entry); - return; - } - - entry->signaled_type |= (fence_type & entry->fence_type); - if (entry->signaled_type == entry->fence_type) { - DRMLISTDELINIT(list); - mgr->info.unreference(mgr, &entry->private); - } - _glthread_UNLOCK_MUTEX(entry->mutex); - driFenceUnReferenceLocked(&entry); - list = prev; - } -} - - -int -driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, - int lazy_hint) -{ - struct _DriFenceMgr *mgr = fence->mgr; - int ret = 0; - - _glthread_LOCK_MUTEX(fence->mutex); - - if ((fence->signaled_type & fence_type) == fence_type) - goto out0; - - ret = mgr->info.finish(mgr, fence->private, fence_type, lazy_hint); - if (ret) - goto out0; - - _glthread_LOCK_MUTEX(mgr->mutex); - _glthread_UNLOCK_MUTEX(fence->mutex); - - driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, - fence_type); - _glthread_UNLOCK_MUTEX(mgr->mutex); - return 0; - - out0: - _glthread_UNLOCK_MUTEX(fence->mutex); - return ret; -} - -uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence) -{ - uint32_t ret; - - _glthread_LOCK_MUTEX(fence->mutex); - ret = fence->signaled_type; - _glthread_UNLOCK_MUTEX(fence->mutex); - - return ret; -} - -int -driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type, - uint32_t *signaled) -{ - int ret = 0; - struct _DriFenceMgr *mgr; - - _glthread_LOCK_MUTEX(fence->mutex); - mgr = fence->mgr; - *signaled = fence->signaled_type; - if ((fence->signaled_type & flush_type) == flush_type) - goto out0; - - ret = mgr->info.signaled(mgr, fence->private, flush_type, signaled); - if (ret) { - *signaled = fence->signaled_type; - goto out0; - } - - if ((fence->signaled_type | *signaled) == fence->signaled_type) - goto out0; - - _glthread_LOCK_MUTEX(mgr->mutex); - _glthread_UNLOCK_MUTEX(fence->mutex); - - driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, - *signaled); - - _glthread_UNLOCK_MUTEX(mgr->mutex); - return 0; - out0: - _glthread_UNLOCK_MUTEX(fence->mutex); - return ret; -} - -struct _DriFenceObject * -driFenceReference(struct _DriFenceObject *fence) -{ - _glthread_LOCK_MUTEX(fence->mgr->mutex); - ++fence->refCount; - _glthread_UNLOCK_MUTEX(fence->mgr->mutex); - return fence; -} - -void -driFenceUnReference(struct _DriFenceObject **pFence) -{ - struct _DriFenceMgr *mgr; - - if (*pFence == NULL) - return; - - mgr = (*pFence)->mgr; - _glthread_LOCK_MUTEX(mgr->mutex); - ++mgr->refCount; - driFenceUnReferenceLocked(pFence); - driFenceMgrUnrefUnlock(&mgr); -} - -struct _DriFenceObject -*driFenceCreate(struct _DriFenceMgr *mgr, uint32_t fence_class, - uint32_t fence_type, void *private, size_t private_size) -{ - struct _DriFenceObject *fence; - size_t fence_size = sizeof(*fence); - - if (private_size) - fence_size = ((fence_size + 15) & ~15); - - fence = calloc(1, fence_size + private_size); - - if (!fence) { - int ret = mgr->info.finish(mgr, private, fence_type, 0); - - if (ret) - usleep(10000000); - - return NULL; - } - - _glthread_INIT_MUTEX(fence->mutex); - _glthread_LOCK_MUTEX(fence->mutex); - _glthread_LOCK_MUTEX(mgr->mutex); - fence->refCount = 1; - DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]); - fence->mgr = mgr; - ++mgr->refCount; - ++mgr->num_fences; - _glthread_UNLOCK_MUTEX(mgr->mutex); - fence->fence_class = fence_class; - fence->fence_type = fence_type; - fence->signaled_type = 0; - fence->private = private; - if (private_size) { - fence->private = (void *)(((uint8_t *) fence) + fence_size); - memcpy(fence->private, private, private_size); - } - - _glthread_UNLOCK_MUTEX(fence->mutex); - return fence; -} - - -static int -tSignaled(struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, - uint32_t *signaled_type) -{ - long fd = (long) mgr->private; - int dummy; - drmFence *fence = (drmFence *) private; - int ret; - - *signaled_type = 0; - ret = drmFenceSignaled((int) fd, fence, flush_type, &dummy); - if (ret) - return ret; - - *signaled_type = fence->signaled; - - return 0; -} - -static int -tFinish(struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, - int lazy_hint) -{ - long fd = (long) mgr->private; - unsigned flags = lazy_hint ? DRM_FENCE_FLAG_WAIT_LAZY : 0; - - return drmFenceWait((int)fd, flags, (drmFence *) private, fence_type); -} - -static int -tUnref(struct _DriFenceMgr *mgr, void **private) -{ - long fd = (long) mgr->private; - drmFence *fence = (drmFence *) *private; - *private = NULL; - - return drmFenceUnreference(fd, fence); -} - -struct _DriFenceMgr *driFenceMgrTTMInit(int fd) -{ - struct _DriFenceMgrCreateInfo info; - struct _DriFenceMgr *mgr; - - info.flags = DRI_FENCE_CLASS_ORDERED; - info.num_classes = 4; - info.signaled = tSignaled; - info.finish = tFinish; - info.unreference = tUnref; - - mgr = driFenceMgrCreate(&info); - if (mgr == NULL) - return NULL; - - mgr->private = (void *) (long) fd; - return mgr; -} - diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h b/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h deleted file mode 100644 index 4ea58dfe18..0000000000 --- a/src/gallium/winsys/egl_drm/intel/ws_dri_fencemgr.h +++ /dev/null @@ -1,115 +0,0 @@ -#ifndef DRI_FENCEMGR_H -#define DRI_FENCEMGR_H - -#include -#include - -struct _DriFenceObject; -struct _DriFenceMgr; - -/* - * Do a quick check to see if the fence manager has registered the fence - * object as signaled. Note that this function may return a false negative - * answer. - */ -extern uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence); - -/* - * Check if the fence object is signaled. This function can be substantially - * more expensive to call than the above function, but will not return a false - * negative answer. The argument "flush_type" sets the types that the - * underlying mechanism must make sure will eventually signal. - */ -extern int driFenceSignaledType(struct _DriFenceObject *fence, - uint32_t flush_type, uint32_t *signaled); - -/* - * Convenience functions. - */ - -static inline int driFenceSignaled(struct _DriFenceObject *fence, - uint32_t flush_type) -{ - uint32_t signaled_types; - int ret = driFenceSignaledType(fence, flush_type, &signaled_types); - if (ret) - return 0; - return ((signaled_types & flush_type) == flush_type); -} - -static inline int driFenceSignaledCached(struct _DriFenceObject *fence, - uint32_t flush_type) -{ - uint32_t signaled_types = - driFenceSignaledTypeCached(fence); - - return ((signaled_types & flush_type) == flush_type); -} - -/* - * Reference a fence object. - */ -extern struct _DriFenceObject *driFenceReference(struct _DriFenceObject *fence); - -/* - * Unreference a fence object. The fence object pointer will be reset to NULL. - */ - -extern void driFenceUnReference(struct _DriFenceObject **pFence); - - -/* - * Wait for a fence to signal the indicated fence_type. - * If "lazy_hint" is true, it indicates that the wait may sleep to avoid - * busy-wait polling. - */ -extern int driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, - int lazy_hint); - -/* - * Create a DriFenceObject for manager "mgr". - * - * "private" is a pointer that should be used for the callbacks in - * struct _DriFenceMgrCreateInfo. - * - * if private_size is nonzero, then the info stored at *private, with size - * private size will be copied and the fence manager will instead use a - * pointer to the copied data for the callbacks in - * struct _DriFenceMgrCreateInfo. In that case, the object pointed to by - * "private" may be destroyed after the call to driFenceCreate. - */ -extern struct _DriFenceObject *driFenceCreate(struct _DriFenceMgr *mgr, - uint32_t fence_class, - uint32_t fence_type, - void *private, - size_t private_size); - -extern uint32_t driFenceType(struct _DriFenceObject *fence); - -/* - * Fence creations are ordered. If a fence signals a fence_type, - * it is safe to assume that all fences of the same class that was - * created before that fence has signaled the same type. - */ - -#define DRI_FENCE_CLASS_ORDERED (1 << 0) - -struct _DriFenceMgrCreateInfo { - uint32_t flags; - uint32_t num_classes; - int (*signaled) (struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, - uint32_t *signaled_type); - int (*finish) (struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, int lazy_hint); - int (*unreference) (struct _DriFenceMgr *mgr, void **private); -}; - -extern struct _DriFenceMgr * -driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info); - -void -driFenceMgrUnReference(struct _DriFenceMgr **pMgr); - -extern struct _DriFenceMgr * -driFenceMgrTTMInit(int fd); - -#endif diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_mallocpool.c b/src/gallium/winsys/egl_drm/intel/ws_dri_mallocpool.c deleted file mode 100644 index bf97d7e440..0000000000 --- a/src/gallium/winsys/egl_drm/intel/ws_dri_mallocpool.c +++ /dev/null @@ -1,162 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, TX., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#include -#include -#include -#include "imports.h" -#include "glthread.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" -#include "intel_screen.h" - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, uint64_t flags, unsigned hint, - unsigned alignment) -{ - unsigned long *private = malloc(size + 2*sizeof(unsigned long)); - if ((flags & DRM_BO_MASK_MEM) != DRM_BO_FLAG_MEM_LOCAL) - abort(); - - *private = size; - return (void *)private; -} - - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - free(private); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, - _glthread_Mutex *mutex, int lazy) -{ - return 0; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, _glthread_Mutex *mutex, void **virtual) -{ - *virtual = (void *)((unsigned long *)private + 2); - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - /* - * BUG - */ - abort(); - return 0UL; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - /* - * BUG - */ - abort(); -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - return DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - return *(unsigned long *) private; -} - - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - abort(); - return 0UL; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - abort(); - return NULL; -} - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - free(pool); -} - - -struct _DriBufferPool * -driMallocPoolInit(void) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - if (!pool) - return NULL; - - pool->data = NULL; - pool->fd = -1; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->poolOffset = &pool_poolOffset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = NULL; - pool->waitIdle = &pool_waitIdle; - pool->takeDown = &pool_takedown; - return pool; -} diff --git a/src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c b/src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c deleted file mode 100644 index e69519a01f..0000000000 --- a/src/gallium/winsys/egl_drm/intel/ws_dri_slabpool.c +++ /dev/null @@ -1,970 +0,0 @@ -/************************************************************************** - * - * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellstrom - */ - -#include -#include -#include -#include -#include -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" -#include "ws_dri_bufmgr.h" -#include "glthread.h" - -#define DRI_SLABPOOL_ALLOC_RETRIES 100 - -struct _DriSlab; - -struct _DriSlabBuffer { - int isSlabBuffer; - drmBO *bo; - struct _DriFenceObject *fence; - struct _DriSlab *parent; - drmMMListHead head; - uint32_t mapCount; - uint32_t start; - uint32_t fenceType; - int unFenced; - _glthread_Cond event; -}; - -struct _DriKernelBO { - int fd; - drmBO bo; - drmMMListHead timeoutHead; - drmMMListHead head; - struct timeval timeFreed; - uint32_t pageAlignment; - void *virtual; -}; - -struct _DriSlab{ - drmMMListHead head; - drmMMListHead freeBuffers; - uint32_t numBuffers; - uint32_t numFree; - struct _DriSlabBuffer *buffers; - struct _DriSlabSizeHeader *header; - struct _DriKernelBO *kbo; -}; - - -struct _DriSlabSizeHeader { - drmMMListHead slabs; - drmMMListHead freeSlabs; - drmMMListHead delayedBuffers; - uint32_t numDelayed; - struct _DriSlabPool *slabPool; - uint32_t bufSize; - _glthread_Mutex mutex; -}; - -struct _DriFreeSlabManager { - struct timeval slabTimeout; - struct timeval checkInterval; - struct timeval nextCheck; - drmMMListHead timeoutList; - drmMMListHead unCached; - drmMMListHead cached; - _glthread_Mutex mutex; -}; - - -struct _DriSlabPool { - - /* - * The data of this structure remains constant after - * initialization and thus needs no mutex protection. - */ - - struct _DriFreeSlabManager *fMan; - uint64_t proposedFlags; - uint64_t validMask; - uint32_t *bucketSizes; - uint32_t numBuckets; - uint32_t pageSize; - int fd; - int pageAlignment; - int maxSlabSize; - int desiredNumBuffers; - struct _DriSlabSizeHeader *headers; -}; - -/* - * FIXME: Perhaps arrange timeout slabs in size buckets for fast - * retreival?? - */ - - -static inline int -driTimeAfterEq(struct timeval *arg1, struct timeval *arg2) -{ - return ((arg1->tv_sec > arg2->tv_sec) || - ((arg1->tv_sec == arg2->tv_sec) && - (arg1->tv_usec > arg2->tv_usec))); -} - -static inline void -driTimeAdd(struct timeval *arg, struct timeval *add) -{ - unsigned int sec; - - arg->tv_sec += add->tv_sec; - arg->tv_usec += add->tv_usec; - sec = arg->tv_usec / 1000000; - arg->tv_sec += sec; - arg->tv_usec -= sec*1000000; -} - -static void -driFreeKernelBO(struct _DriKernelBO *kbo) -{ - if (!kbo) - return; - - (void) drmBOUnreference(kbo->fd, &kbo->bo); - free(kbo); -} - - -static void -driFreeTimeoutKBOsLocked(struct _DriFreeSlabManager *fMan, - struct timeval *time) -{ - drmMMListHead *list, *next; - struct _DriKernelBO *kbo; - - if (!driTimeAfterEq(time, &fMan->nextCheck)) - return; - - for (list = fMan->timeoutList.next, next = list->next; - list != &fMan->timeoutList; - list = next, next = list->next) { - - kbo = DRMLISTENTRY(struct _DriKernelBO, list, timeoutHead); - - if (!driTimeAfterEq(time, &kbo->timeFreed)) - break; - - DRMLISTDELINIT(&kbo->timeoutHead); - DRMLISTDELINIT(&kbo->head); - driFreeKernelBO(kbo); - } - - fMan->nextCheck = *time; - driTimeAdd(&fMan->nextCheck, &fMan->checkInterval); -} - - -/* - * Add a _DriKernelBO to the free slab manager. - * This means that it is available for reuse, but if it's not - * reused in a while, it will be freed. - */ - -static void -driSetKernelBOFree(struct _DriFreeSlabManager *fMan, - struct _DriKernelBO *kbo) -{ - struct timeval time; - - _glthread_LOCK_MUTEX(fMan->mutex); - gettimeofday(&time, NULL); - driTimeAdd(&time, &fMan->slabTimeout); - - kbo->timeFreed = time; - - if (kbo->bo.flags & DRM_BO_FLAG_CACHED) - DRMLISTADD(&kbo->head, &fMan->cached); - else - DRMLISTADD(&kbo->head, &fMan->unCached); - - DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList); - driFreeTimeoutKBOsLocked(fMan, &time); - - _glthread_UNLOCK_MUTEX(fMan->mutex); -} - -/* - * Get a _DriKernelBO for us to use as storage for a slab. - * - */ - -static struct _DriKernelBO * -driAllocKernelBO(struct _DriSlabSizeHeader *header) - -{ - struct _DriSlabPool *slabPool = header->slabPool; - struct _DriFreeSlabManager *fMan = slabPool->fMan; - drmMMListHead *list, *next, *head; - uint32_t size = header->bufSize * slabPool->desiredNumBuffers; - struct _DriKernelBO *kbo; - struct _DriKernelBO *kboTmp; - int ret; - - /* - * FIXME: We should perhaps allow some variation in slabsize in order - * to efficiently reuse slabs. - */ - - size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize; - size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1); - _glthread_LOCK_MUTEX(fMan->mutex); - - kbo = NULL; - - retry: - head = (slabPool->proposedFlags & DRM_BO_FLAG_CACHED) ? - &fMan->cached : &fMan->unCached; - - for (list = head->next, next = list->next; - list != head; - list = next, next = list->next) { - - kboTmp = DRMLISTENTRY(struct _DriKernelBO, list, head); - - if ((kboTmp->bo.size == size) && - (slabPool->pageAlignment == 0 || - (kboTmp->pageAlignment % slabPool->pageAlignment) == 0)) { - - if (!kbo) - kbo = kboTmp; - - if ((kbo->bo.proposedFlags ^ slabPool->proposedFlags) == 0) - break; - - } - } - - if (kbo) { - DRMLISTDELINIT(&kbo->head); - DRMLISTDELINIT(&kbo->timeoutHead); - } - - _glthread_UNLOCK_MUTEX(fMan->mutex); - - if (kbo) { - uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags; - - ret = 0; - if (new_mask) { - ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, - new_mask, DRM_BO_HINT_DONT_FENCE, 0, 0); - } - if (ret == 0) - return kbo; - - driFreeKernelBO(kbo); - kbo = NULL; - goto retry; - } - - kbo = calloc(1, sizeof(struct _DriKernelBO)); - if (!kbo) - return NULL; - - kbo->fd = slabPool->fd; - DRMINITLISTHEAD(&kbo->head); - DRMINITLISTHEAD(&kbo->timeoutHead); - ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL, - slabPool->proposedFlags, - DRM_BO_HINT_DONT_FENCE, &kbo->bo); - if (ret) - goto out_err0; - - ret = drmBOMap(kbo->fd, &kbo->bo, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0, &kbo->virtual); - - if (ret) - goto out_err1; - - ret = drmBOUnmap(kbo->fd, &kbo->bo); - if (ret) - goto out_err1; - - return kbo; - - out_err1: - drmBOUnreference(kbo->fd, &kbo->bo); - out_err0: - free(kbo); - return NULL; -} - - -static int -driAllocSlab(struct _DriSlabSizeHeader *header) -{ - struct _DriSlab *slab; - struct _DriSlabBuffer *buf; - uint32_t numBuffers; - int ret; - int i; - - slab = calloc(1, sizeof(*slab)); - if (!slab) - return -ENOMEM; - - slab->kbo = driAllocKernelBO(header); - if (!slab->kbo) { - ret = -ENOMEM; - goto out_err0; - } - - numBuffers = slab->kbo->bo.size / header->bufSize; - - slab->buffers = calloc(numBuffers, sizeof(*slab->buffers)); - if (!slab->buffers) { - ret = -ENOMEM; - goto out_err1; - } - - DRMINITLISTHEAD(&slab->head); - DRMINITLISTHEAD(&slab->freeBuffers); - slab->numBuffers = numBuffers; - slab->numFree = 0; - slab->header = header; - - buf = slab->buffers; - for (i=0; i < numBuffers; ++i) { - buf->parent = slab; - buf->start = i* header->bufSize; - buf->mapCount = 0; - buf->isSlabBuffer = 1; - _glthread_INIT_COND(buf->event); - DRMLISTADDTAIL(&buf->head, &slab->freeBuffers); - slab->numFree++; - buf++; - } - - DRMLISTADDTAIL(&slab->head, &header->slabs); - - return 0; - - out_err1: - driSetKernelBOFree(header->slabPool->fMan, slab->kbo); - free(slab->buffers); - out_err0: - free(slab); - return ret; -} - -/* - * Delete a buffer from the slab header delayed list and put - * it on the slab free list. - */ - -static void -driSlabFreeBufferLocked(struct _DriSlabBuffer *buf) -{ - struct _DriSlab *slab = buf->parent; - struct _DriSlabSizeHeader *header = slab->header; - drmMMListHead *list = &buf->head; - - DRMLISTDEL(list); - DRMLISTADDTAIL(list, &slab->freeBuffers); - slab->numFree++; - - if (slab->head.next == &slab->head) - DRMLISTADDTAIL(&slab->head, &header->slabs); - - if (slab->numFree == slab->numBuffers) { - list = &slab->head; - DRMLISTDEL(list); - DRMLISTADDTAIL(list, &header->freeSlabs); - } - - if (header->slabs.next == &header->slabs || - slab->numFree != slab->numBuffers) { - - drmMMListHead *next; - struct _DriFreeSlabManager *fMan = header->slabPool->fMan; - - for (list = header->freeSlabs.next, next = list->next; - list != &header->freeSlabs; - list = next, next = list->next) { - - slab = DRMLISTENTRY(struct _DriSlab, list, head); - - DRMLISTDELINIT(list); - driSetKernelBOFree(fMan, slab->kbo); - free(slab->buffers); - free(slab); - } - } -} - -static void -driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait) -{ - drmMMListHead *list, *prev, *first; - struct _DriSlabBuffer *buf; - struct _DriSlab *slab; - int firstWasSignaled = 1; - int signaled; - int i; - int ret; - - /* - * Rerun the freeing test if the youngest tested buffer - * was signaled, since there might be more idle buffers - * in the delay list. - */ - - while (firstWasSignaled) { - firstWasSignaled = 0; - signaled = 0; - first = header->delayedBuffers.next; - - /* Only examine the oldest 1/3 of delayed buffers: - */ - if (header->numDelayed > 3) { - for (i = 0; i < header->numDelayed; i += 3) { - first = first->next; - } - } - - for (list = first, prev = list->prev; - list != &header->delayedBuffers; - list = prev, prev = list->prev) { - buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); - slab = buf->parent; - - if (!signaled) { - if (wait) { - ret = driFenceFinish(buf->fence, buf->fenceType, 0); - if (ret) - break; - signaled = 1; - wait = 0; - } else { - signaled = driFenceSignaled(buf->fence, buf->fenceType); - } - if (signaled) { - if (list == first) - firstWasSignaled = 1; - driFenceUnReference(&buf->fence); - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { - driFenceUnReference(&buf->fence); - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - } - } -} - - -static struct _DriSlabBuffer * -driSlabAllocBuffer(struct _DriSlabSizeHeader *header) -{ - static struct _DriSlabBuffer *buf; - struct _DriSlab *slab; - drmMMListHead *list; - int count = DRI_SLABPOOL_ALLOC_RETRIES; - - _glthread_LOCK_MUTEX(header->mutex); - while(header->slabs.next == &header->slabs && count > 0) { - driSlabCheckFreeLocked(header, 0); - if (header->slabs.next != &header->slabs) - break; - - _glthread_UNLOCK_MUTEX(header->mutex); - if (count != DRI_SLABPOOL_ALLOC_RETRIES) - usleep(1); - _glthread_LOCK_MUTEX(header->mutex); - (void) driAllocSlab(header); - count--; - } - - list = header->slabs.next; - if (list == &header->slabs) { - _glthread_UNLOCK_MUTEX(header->mutex); - return NULL; - } - slab = DRMLISTENTRY(struct _DriSlab, list, head); - if (--slab->numFree == 0) - DRMLISTDELINIT(list); - - list = slab->freeBuffers.next; - DRMLISTDELINIT(list); - - _glthread_UNLOCK_MUTEX(header->mutex); - buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); - return buf; -} - -static void * -pool_create(struct _DriBufferPool *driPool, unsigned long size, - uint64_t flags, unsigned hint, unsigned alignment) -{ - struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; - struct _DriSlabSizeHeader *header; - struct _DriSlabBuffer *buf; - void *dummy; - int i; - int ret; - - /* - * FIXME: Check for compatibility. - */ - header = pool->headers; - for (i=0; inumBuckets; ++i) { - if (header->bufSize >= size) - break; - header++; - } - - if (i < pool->numBuckets) - return driSlabAllocBuffer(header); - - - /* - * Fall back to allocate a buffer object directly from DRM. - * and wrap it in a driBO structure. - */ - - - buf = calloc(1, sizeof(*buf)); - - if (!buf) - return NULL; - - buf->bo = calloc(1, sizeof(*buf->bo)); - if (!buf->bo) - goto out_err0; - - if (alignment) { - if ((alignment < pool->pageSize) && (pool->pageSize % alignment)) - goto out_err1; - if ((alignment > pool->pageSize) && (alignment % pool->pageSize)) - goto out_err1; - } - - ret = drmBOCreate(pool->fd, size, alignment / pool->pageSize, NULL, - flags, hint, buf->bo); - if (ret) - goto out_err1; - - ret = drmBOMap(pool->fd, buf->bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0, &dummy); - if (ret) - goto out_err2; - - ret = drmBOUnmap(pool->fd, buf->bo); - if (ret) - goto out_err2; - - return buf; - out_err2: - drmBOUnreference(pool->fd, buf->bo); - out_err1: - free(buf->bo); - out_err0: - free(buf); - return NULL; -} - -static int -pool_destroy(struct _DriBufferPool *driPool, void *private) -{ - struct _DriSlabBuffer *buf = - (struct _DriSlabBuffer *) private; - struct _DriSlab *slab; - struct _DriSlabSizeHeader *header; - - if (!buf->isSlabBuffer) { - struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; - int ret; - - ret = drmBOUnreference(pool->fd, buf->bo); - free(buf->bo); - free(buf); - return ret; - } - - slab = buf->parent; - header = slab->header; - - _glthread_LOCK_MUTEX(header->mutex); - buf->unFenced = 0; - buf->mapCount = 0; - - if (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)) { - DRMLISTADDTAIL(&buf->head, &header->delayedBuffers); - header->numDelayed++; - } else { - if (buf->fence) - driFenceUnReference(&buf->fence); - driSlabFreeBufferLocked(buf); - } - - _glthread_UNLOCK_MUTEX(header->mutex); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *driPool, void *private, - _glthread_Mutex *mutex, int lazy) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - while(buf->unFenced) - _glthread_COND_WAIT(buf->event, *mutex); - - if (!buf->fence) - return 0; - - driFenceFinish(buf->fence, buf->fenceType, lazy); - driFenceUnReference(&buf->fence); - - return 0; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, _glthread_Mutex *mutex, void **virtual) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - int busy; - - if (buf->isSlabBuffer) - busy = buf->unFenced || (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)); - else - busy = buf->fence && !driFenceSignaled(buf->fence, buf->fenceType); - - - if (busy) { - if (hint & DRM_BO_HINT_DONT_BLOCK) - return -EBUSY; - else { - (void) pool_waitIdle(pool, private, mutex, 0); - } - } - - ++buf->mapCount; - *virtual = (buf->isSlabBuffer) ? - (void *) ((uint8_t *) buf->parent->kbo->virtual + buf->start) : - (void *) buf->bo->virtual; - - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - --buf->mapCount; - if (buf->mapCount == 0 && buf->isSlabBuffer) - _glthread_COND_BROADCAST(buf->event); - - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - struct _DriSlab *slab; - struct _DriSlabSizeHeader *header; - - if (!buf->isSlabBuffer) { - assert(buf->bo->proposedFlags & DRM_BO_FLAG_NO_MOVE); - return buf->bo->offset; - } - - slab = buf->parent; - header = slab->header; - - (void) header; - assert(header->slabPool->proposedFlags & DRM_BO_FLAG_NO_MOVE); - return slab->kbo->bo.offset + buf->start; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - return buf->start; -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - if (!buf->isSlabBuffer) - return buf->bo->flags; - - return buf->parent->kbo->bo.flags; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - if (!buf->isSlabBuffer) - return buf->bo->size; - - return buf->parent->header->bufSize; -} - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - drmBO *bo; - - if (buf->fence) - driFenceUnReference(&buf->fence); - - buf->fence = driFenceReference(fence); - bo = (buf->isSlabBuffer) ? - &buf->parent->kbo->bo: - buf->bo; - buf->fenceType = bo->fenceFlags; - - buf->unFenced = 0; - _glthread_COND_BROADCAST(buf->event); - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - if (!buf) - return NULL; - - return (buf->isSlabBuffer) ? &buf->parent->kbo->bo : buf->bo; -} - -static int -pool_validate(struct _DriBufferPool *pool, void *private, - _glthread_Mutex *mutex) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - if (!buf->isSlabBuffer) - return 0; - - while(buf->mapCount != 0) - _glthread_COND_WAIT(buf->event, *mutex); - - buf->unFenced = 1; - return 0; -} - - -struct _DriFreeSlabManager * -driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec) -{ - struct _DriFreeSlabManager *tmp; - - tmp = calloc(1, sizeof(*tmp)); - if (!tmp) - return NULL; - - _glthread_INIT_MUTEX(tmp->mutex); - _glthread_LOCK_MUTEX(tmp->mutex); - tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000; - tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000; - tmp->slabTimeout.tv_usec -= tmp->slabTimeout.tv_sec*1000000; - - tmp->checkInterval.tv_usec = checkIntervalMsec*1000; - tmp->checkInterval.tv_sec = tmp->checkInterval.tv_usec / 1000000; - tmp->checkInterval.tv_usec -= tmp->checkInterval.tv_sec*1000000; - - gettimeofday(&tmp->nextCheck, NULL); - driTimeAdd(&tmp->nextCheck, &tmp->checkInterval); - DRMINITLISTHEAD(&tmp->timeoutList); - DRMINITLISTHEAD(&tmp->unCached); - DRMINITLISTHEAD(&tmp->cached); - _glthread_UNLOCK_MUTEX(tmp->mutex); - - return tmp; -} - -void -driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan) -{ - struct timeval time; - - time = fMan->nextCheck; - driTimeAdd(&time, &fMan->checkInterval); - - _glthread_LOCK_MUTEX(fMan->mutex); - driFreeTimeoutKBOsLocked(fMan, &time); - _glthread_UNLOCK_MUTEX(fMan->mutex); - - assert(fMan->timeoutList.next == &fMan->timeoutList); - assert(fMan->unCached.next == &fMan->unCached); - assert(fMan->cached.next == &fMan->cached); - - free(fMan); -} - -static void -driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size, - struct _DriSlabSizeHeader *header) -{ - _glthread_INIT_MUTEX(header->mutex); - _glthread_LOCK_MUTEX(header->mutex); - - DRMINITLISTHEAD(&header->slabs); - DRMINITLISTHEAD(&header->freeSlabs); - DRMINITLISTHEAD(&header->delayedBuffers); - - header->numDelayed = 0; - header->slabPool = pool; - header->bufSize = size; - - _glthread_UNLOCK_MUTEX(header->mutex); -} - -static void -driFinishSizeHeader(struct _DriSlabSizeHeader *header) -{ - drmMMListHead *list, *next; - struct _DriSlabBuffer *buf; - - _glthread_LOCK_MUTEX(header->mutex); - for (list = header->delayedBuffers.next, next = list->next; - list != &header->delayedBuffers; - list = next, next = list->next) { - - buf = DRMLISTENTRY(struct _DriSlabBuffer, list , head); - if (buf->fence) { - (void) driFenceFinish(buf->fence, buf->fenceType, 0); - driFenceUnReference(&buf->fence); - } - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - _glthread_UNLOCK_MUTEX(header->mutex); -} - -static void -pool_takedown(struct _DriBufferPool *driPool) -{ - struct _DriSlabPool *pool = driPool->data; - int i; - - for (i=0; inumBuckets; ++i) { - driFinishSizeHeader(&pool->headers[i]); - } - - free(pool->headers); - free(pool->bucketSizes); - free(pool); - free(driPool); -} - -struct _DriBufferPool * -driSlabPoolInit(int fd, uint64_t flags, - uint64_t validMask, - uint32_t smallestSize, - uint32_t numSizes, - uint32_t desiredNumBuffers, - uint32_t maxSlabSize, - uint32_t pageAlignment, - struct _DriFreeSlabManager *fMan) -{ - struct _DriBufferPool *driPool; - struct _DriSlabPool *pool; - uint32_t i; - - driPool = calloc(1, sizeof(*driPool)); - if (!driPool) - return NULL; - - pool = calloc(1, sizeof(*pool)); - if (!pool) - goto out_err0; - - pool->bucketSizes = calloc(numSizes, sizeof(*pool->bucketSizes)); - if (!pool->bucketSizes) - goto out_err1; - - pool->headers = calloc(numSizes, sizeof(*pool->headers)); - if (!pool->headers) - goto out_err2; - - pool->fMan = fMan; - pool->proposedFlags = flags; - pool->validMask = validMask; - pool->numBuckets = numSizes; - pool->pageSize = getpagesize(); - pool->fd = fd; - pool->pageAlignment = pageAlignment; - pool->maxSlabSize = maxSlabSize; - pool->desiredNumBuffers = desiredNumBuffers; - - for (i=0; inumBuckets; ++i) { - pool->bucketSizes[i] = (smallestSize << i); - driInitSizeHeader(pool, pool->bucketSizes[i], - &pool->headers[i]); - } - - driPool->data = (void *) pool; - driPool->map = &pool_map; - driPool->unmap = &pool_unmap; - driPool->destroy = &pool_destroy; - driPool->offset = &pool_offset; - driPool->poolOffset = &pool_poolOffset; - driPool->flags = &pool_flags; - driPool->size = &pool_size; - driPool->create = &pool_create; - driPool->fence = &pool_fence; - driPool->kernel = &pool_kernel; - driPool->validate = &pool_validate; - driPool->waitIdle = &pool_waitIdle; - driPool->takeDown = &pool_takedown; - - return driPool; - - out_err2: - free(pool->bucketSizes); - out_err1: - free(pool); - out_err0: - free(driPool); - - return NULL; -} -- cgit v1.2.3 From 0931b421d67b8ce471f17d43c183017f1eb92e31 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 6 Jun 2008 15:04:45 +0100 Subject: draw: make sure middle-end primitive is uptodate in vcache --- src/gallium/auxiliary/draw/draw_pt_vcache.c | 27 ++++++++++++++++++++++++++- 1 file changed, 26 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index 2eafe270bc..cda2987c9e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -58,10 +58,21 @@ struct vcache_frontend { unsigned input_prim; unsigned output_prim; + + unsigned middle_prim; + unsigned opt; }; static void vcache_flush( struct vcache_frontend *vcache ) { + if (vcache->middle_prim != vcache->output_prim) { + vcache->middle_prim = vcache->output_prim; + vcache->middle->prepare( vcache->middle, + vcache->middle_prim, + vcache->opt, + &vcache->fetch_max ); + } + if (vcache->draw_count) { vcache->middle->run( vcache->middle, vcache->fetch_elts, @@ -308,6 +319,14 @@ static void vcache_check_run( struct draw_pt_front_end *frontend, goto fail; } + if (vcache->middle_prim != vcache->input_prim) { + vcache->middle_prim = vcache->input_prim; + vcache->middle->prepare( vcache->middle, + vcache->middle_prim, + vcache->opt, + &vcache->fetch_max ); + } + if (min_index == 0 && index_size == 2) @@ -412,7 +431,13 @@ static void vcache_prepare( struct draw_pt_front_end *frontend, vcache->output_prim = draw_pt_reduced_prim(prim); vcache->middle = middle; - middle->prepare( middle, vcache->output_prim, opt, &vcache->fetch_max ); + vcache->opt = opt; + + /* Have to run prepare here, but try and guess a good prim for + * doing so: + */ + vcache->middle_prim = (opt & PT_PIPELINE) ? vcache->output_prim : vcache->input_prim; + middle->prepare( middle, vcache->middle_prim, opt, &vcache->fetch_max ); } -- cgit v1.2.3 From e4fda51404a6a05c4047a639de4ccc3ea9678c2c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 6 Jun 2008 15:57:37 +0100 Subject: mesa: turn off ffvertex prog debug --- src/mesa/main/ffvertex_prog.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 2ef3286e57..06710f405d 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -305,7 +305,7 @@ static struct state_key *make_state_key( GLcontext *ctx ) * generated program with line/function references for each * instruction back into this file: */ -#define DISASSEM 1 +#define DISASSEM 0 /* Should be tunable by the driver - do we want to do matrix * multiplications with DP4's or with MUL/MAD's? SSE works better -- cgit v1.2.3 From 0a84d327de8258104b98b176c0eaf0fb6712a982 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 6 Jun 2008 10:42:13 -0600 Subject: egl: open X display if needed --- src/gallium/winsys/egl_xlib/egl_xlib.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 07b3c0eb00..7bef46d560 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -573,6 +573,10 @@ _eglMain(_EGLDisplay *dpy, const char *args) if (!xdrv) return NULL; + if (!dpy->Xdpy) { + dpy->Xdpy = XOpenDisplay(NULL); + } + _eglInitDriverFallbacks(&xdrv->Base); xdrv->Base.API.Initialize = xlib_eglInitialize; xdrv->Base.API.Terminate = xlib_eglTerminate; -- cgit v1.2.3 From 463a47bf59398e850d5a6537da1186d855bd2919 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Fri, 6 Jun 2008 18:09:32 +0100 Subject: draw: fix intermediate buffer confusion in draw_vs_varient.c The final output buffer can't be used to hold intermediate results as the intermediate vertex size may be greater than the final vertex size, and in any case the output buffer may be uncached in hw drivers. --- .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 7 + src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h | 2 + src/gallium/auxiliary/draw/draw_vs_varient.c | 175 +++++++++++---------- 3 files changed, 101 insertions(+), 83 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 5a1d79d996..5ce3aba2a2 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -193,6 +193,13 @@ static void fse_prepare( struct draw_pt_middle_end *middle, *max_vertices = (draw->render->max_vertex_buffer_bytes / (vinfo->size * 4)); + /* Probably need to do this somewhere (or fix exec shader not to + * need it): + */ + if (1) { + struct draw_vertex_shader *vs = draw->vs.vertex_shader; + vs->prepare(vs, draw); + } //return TRUE; diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h index 5c99a47e49..ec05bbeab4 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h @@ -13,6 +13,8 @@ static void FUNC( struct draw_pt_front_end *frontend, unsigned i; ushort flags; + if (0) debug_printf("%s %d\n", __FUNCTION__, count); + switch (vcache->input_prim) { case PIPE_PRIM_POINTS: diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index 3ceca5e849..d8317d3bbe 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -55,6 +55,8 @@ struct draw_vs_varient_generic { */ struct translate *fetch; struct translate *emit; + + unsigned temp_vertex_stride; }; @@ -84,7 +86,7 @@ static void do_rhw_viewport( struct draw_vs_varient_generic *vsvg, char *ptr = (char *)output_buffer; const float *scale = vsvg->base.vs->draw->viewport.scale; const float *trans = vsvg->base.vs->draw->viewport.translate; - unsigned stride = vsvg->base.key.output_stride; + unsigned stride = vsvg->temp_vertex_stride; unsigned j; for (j = 0; j < count; j++, ptr += stride) { @@ -99,13 +101,13 @@ static void do_rhw_viewport( struct draw_vs_varient_generic *vsvg, } static void do_viewport( struct draw_vs_varient_generic *vsvg, - unsigned count, - void *output_buffer ) + unsigned count, + void *output_buffer ) { char *ptr = (char *)output_buffer; const float *scale = vsvg->base.vs->draw->viewport.scale; const float *trans = vsvg->base.vs->draw->viewport.translate; - unsigned stride = vsvg->base.key.output_stride; + unsigned stride = vsvg->temp_vertex_stride; unsigned j; for (j = 0; j < count; j++, ptr += stride) { @@ -124,7 +126,9 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, void *output_buffer) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; - + unsigned temp_vertex_stride = vsvg->temp_vertex_stride; + void *temp_buffer = MALLOC( align(count,4) * temp_vertex_stride ); + if (0) debug_printf("%s %d \n", __FUNCTION__, count); /* Want to do this in small batches for cache locality? @@ -135,44 +139,45 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, count, output_buffer ); - //if (!vsvg->base.vs->is_passthrough) - { - vsvg->base.vs->run_linear( vsvg->base.vs, - output_buffer, - output_buffer, - (const float (*)[4])vsvg->base.vs->draw->pt.user.constants, - count, - vsvg->base.key.output_stride, - vsvg->base.key.output_stride); - - - if (vsvg->base.key.clip) { - /* not really handling clipping, just do the rhw so we can - * see the results... - */ - do_rhw_viewport( vsvg, - count, - output_buffer ); - } - else if (vsvg->base.key.viewport) { - do_viewport( vsvg, - count, - output_buffer ); - } + vsvg->base.vs->run_linear( vsvg->base.vs, + temp_buffer, + temp_buffer, + (const float (*)[4])vsvg->base.vs->draw->pt.user.constants, + count, + temp_vertex_stride, + temp_vertex_stride); + + + if (vsvg->base.key.clip) { + /* not really handling clipping, just do the rhw so we can + * see the results... + */ + do_rhw_viewport( vsvg, + count, + temp_buffer ); + } + else if (vsvg->base.key.viewport) { + do_viewport( vsvg, + count, + temp_buffer ); + } - //if (!vsvg->already_in_emit_format) + vsvg->emit->set_buffer( vsvg->emit, + 0, + temp_buffer, + temp_vertex_stride ); - vsvg->emit->set_buffer( vsvg->emit, - 0, - output_buffer, - vsvg->base.key.output_stride ); + vsvg->emit->set_buffer( vsvg->emit, + 1, + &vsvg->draw->rasterizer->point_size, + 0); + vsvg->emit->run( vsvg->emit, + 0, count, + output_buffer ); - vsvg->emit->run( vsvg->emit, - 0, count, - output_buffer ); - } + FREE(temp_buffer); } @@ -182,54 +187,55 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, void *output_buffer ) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; + unsigned temp_vertex_stride = vsvg->temp_vertex_stride; + void *temp_buffer = MALLOC( align(count,4) * temp_vertex_stride ); - if (0) debug_printf("%s %d %d\n", __FUNCTION__, start, count); - - + if (0) debug_printf("%s %d %d (sz %d, %d)\n", __FUNCTION__, start, count, + vsvg->base.key.output_stride, + temp_vertex_stride); + vsvg->fetch->run( vsvg->fetch, start, count, - output_buffer ); - - //if (!vsvg->base.vs->is_passthrough) - { - vsvg->base.vs->run_linear( vsvg->base.vs, - output_buffer, - output_buffer, - (const float (*)[4])vsvg->base.vs->draw->pt.user.constants, - count, - vsvg->base.key.output_stride, - vsvg->base.key.output_stride); - - if (vsvg->base.key.clip) { - /* not really handling clipping, just do the rhw so we can - * see the results... - */ - do_rhw_viewport( vsvg, - count, - output_buffer ); - } - else if (vsvg->base.key.viewport) { - do_viewport( vsvg, - count, - output_buffer ); - } - - //if (!vsvg->already_in_emit_format) - vsvg->emit->set_buffer( vsvg->emit, - 0, - output_buffer, - vsvg->base.key.output_stride ); + temp_buffer ); + + vsvg->base.vs->run_linear( vsvg->base.vs, + temp_buffer, + temp_buffer, + (const float (*)[4])vsvg->base.vs->draw->pt.user.constants, + count, + temp_vertex_stride, + temp_vertex_stride); + + if (vsvg->base.key.clip) { + /* not really handling clipping, just do the rhw so we can + * see the results... + */ + do_rhw_viewport( vsvg, + count, + temp_buffer ); + } + else if (vsvg->base.key.viewport) { + do_viewport( vsvg, + count, + temp_buffer ); + } - vsvg->emit->set_buffer( vsvg->emit, - 1, - &vsvg->draw->rasterizer->point_size, - 0); + vsvg->emit->set_buffer( vsvg->emit, + 0, + temp_buffer, + temp_vertex_stride ); + + vsvg->emit->set_buffer( vsvg->emit, + 1, + &vsvg->draw->rasterizer->point_size, + 0); + + vsvg->emit->run( vsvg->emit, + 0, count, + output_buffer ); - vsvg->emit->run( vsvg->emit, - 0, count, - output_buffer ); - } + FREE(temp_buffer); } @@ -261,18 +267,20 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, vsvg->draw = vs->draw; + vsvg->temp_vertex_stride = MAX2(key->nr_inputs, + vsvg->base.vs->info.num_outputs) * 4 * sizeof(float); /* Build free-standing fetch and emit functions: */ fetch.nr_elements = key->nr_inputs; - fetch.output_stride = 0; + fetch.output_stride = vsvg->temp_vertex_stride; for (i = 0; i < key->nr_inputs; i++) { fetch.element[i].input_format = key->element[i].in.format; fetch.element[i].input_buffer = key->element[i].in.buffer; fetch.element[i].input_offset = key->element[i].in.offset; fetch.element[i].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fetch.element[i].output_offset = fetch.output_stride; - fetch.output_stride += 4 * sizeof(float); + fetch.element[i].output_offset = i * 4 * sizeof(float); + assert(fetch.element[i].output_offset < fetch.output_stride); } @@ -286,6 +294,7 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float); emit.element[i].output_format = draw_translate_vinfo_format(key->element[i].out.format); emit.element[i].output_offset = key->element[i].out.offset; + assert(emit.element[i].input_offset < fetch.output_stride); } else { emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT; -- cgit v1.2.3 From 53076e8b5bce9ddf6eb524b8009db67eb5d3389b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 6 Jun 2008 12:32:58 -0600 Subject: mesa: sync up with latest Makefile changes on master --- src/mesa/Makefile | 146 +++++++++++++++++-------------------- src/mesa/drivers/Makefile | 29 ++++++++ src/mesa/drivers/beos/Makefile | 14 ++-- src/mesa/drivers/directfb/Makefile | 17 ++++- 4 files changed, 119 insertions(+), 87 deletions(-) create mode 100644 src/mesa/drivers/Makefile (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 42096f8088..d39f282b33 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -6,11 +6,6 @@ include $(TOP)/configs/current include sources -GL_MAJOR = 1 -GL_MINOR = 5 -GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) - - .SUFFIXES : .cpp .c.o: @@ -23,7 +18,10 @@ GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ -default: depend subdirs libmesa.a libglapi.a + +# Default: build dependencies, then asm_subdirs, then convenience +# libs (.a) and finally the device drivers: +default: depend asm_subdirs libmesa.a libglapi.a driver_subdirs @@ -32,85 +30,64 @@ default: depend subdirs libmesa.a libglapi.a # Make archive of core mesa object files libmesa.a: $(SOLO_OBJECTS) - @ $(TOP)/bin/mklib -o mesa -static $(SOLO_OBJECTS); - @if [ "${CONFIG_NAME}" = "beos" ] ; then \ - mimeset -f "$@" ; \ - fi + @ $(TOP)/bin/mklib -o mesa -static $(SOLO_OBJECTS) # Make archive of gl* API dispatcher functions only libglapi.a: $(GLAPI_OBJECTS) @ $(TOP)/bin/mklib -o glapi -static $(GLAPI_OBJECTS) - ###################################################################### -# BeOS driver target - -beos-driver: depend subdirs libmesa.a - cd drivers/beos && $(MAKE) +# Device drivers +driver_subdirs: + (cd drivers && $(MAKE)) ###################################################################### -# DRI drivers - -dri-drivers: depend subdirs libmesa.a - cd drivers/dri && $(MAKE) +# Assembly subdirs +asm_subdirs: + @ if echo "$(ASM_FLAGS)" | grep -q USE_X86_ASM ; then \ + (cd x86 && $(MAKE)) || exit 1 ; \ + fi + @ if echo "$(ASM_FLAGS)" | grep -q USE_X86_64_ASM ; then \ + (cd x86 && $(MAKE)) || exit 1 ; \ + (cd x86-64 && $(MAKE)) || exit 1 ; \ + fi ###################################################################### -# Xlib driver (libGL.so) +# Dependency generation -xlib-driver: depend subdirs libmesa.a libglapi.a - cd drivers/x11 && $(MAKE) +depend: $(ALL_SOURCES) + @ echo "running $(MKDEP)" + @ touch depend + @$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \ + > /dev/null 2>/dev/null ###################################################################### -# osmesa driver (libOSMesa.so) - -osmesa-driver: depend subdirs libmesa.a libglapi.a - cd drivers/osmesa && $(MAKE) - - -##################################################################### -# fbdev driver (libGL.so) - -fbdev-driver: depend subdirs libmesa.a libglapi.a - cd drivers/fbdev && $(MAKE) - - -##################################################################### -# DirectFB driver (libGL.so) +# Installation rules -directfb-libgl: $(CORE_OBJECTS) - @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' -ldflags '$(LDFLAGS)' \ - -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ - -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) $(CORE_OBJECTS) \ - $(GL_LIB_DEPS) +# this isn't fleshed out yet but is probably the way to go in the future +new_install: + (cd drivers && $(MAKE) install) -directfb-driver: depend subdirs directfb-libgl - cd drivers/directfb && $(MAKE) - -###################################################################### -# Generic stuff - -depend: $(ALL_SOURCES) - @ echo "running $(MKDEP)" - @ rm -f depend # workaround oops on gutsy?!? - @ touch depend - @ $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \ - > /dev/null 2>/dev/null +# XXX replace this with new_install above someday +install: default + @for driver in $(DRIVER_DIRS) ; do \ + case "$$driver" in \ + osmesa) if [ "$(DRIVER_DIRS)" = osmesa ]; then \ + $(MAKE) install-headers install-osmesa || exit 1 ; \ + else \ + $(MAKE) install-osmesa || exit 1 ; \ + fi ;; \ + dri) $(MAKE) install-libgl install-dri || exit 1 ;; \ + *) $(MAKE) install-libgl || exit 1 ;; \ + esac ; \ + done -subdirs: - @ if echo "$(ASM_FLAGS)" | grep -q USE_X86_ASM ; then \ - (cd x86 && $(MAKE)) || exit 1 ; \ - fi - @ if echo "$(ASM_FLAGS)" | grep -q USE_X86_64_ASM ; then \ - (cd x86 && $(MAKE)) || exit 1 ; \ - (cd x86-64 && $(MAKE)) || exit 1 ; \ - fi - pcedit = sed \ -e 's,@INSTALL_DIR@,$(INSTALL_DIR),' \ -e 's,@LIB_DIR@,$(LIB_DIR),' \ @@ -119,33 +96,44 @@ pcedit = sed \ gl.pc: gl.pc.in $(pcedit) $< > $@ -install: default - $(INSTALL) -d $(INSTALL_DIR)/include/GL - $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) - $(INSTALL) -m 644 $(TOP)/include/GL/*.h $(INSTALL_DIR)/include/GL - @if [ -e $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) ]; then \ - $(INSTALL) $(TOP)/$(LIB_DIR)/libGL* $(INSTALL_DIR)/$(LIB_DIR); \ - fi - @if [ -e $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME) ]; then \ - $(INSTALL) $(TOP)/$(LIB_DIR)/libOSMesa* $(INSTALL_DIR)/$(LIB_DIR); \ - fi +install-headers: + $(INSTALL) -d $(DESTDIR)$(INSTALL_DIR)/include/GL + $(INSTALL) -m 644 $(TOP)/include/GL/*.h \ + $(DESTDIR)$(INSTALL_DIR)/include/GL + +install-libgl: default gl.pc install-headers + $(INSTALL) -d $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) -d $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR)/pkgconfig + $(INSTALL) $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME)* \ + $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) -m 644 gl.pc $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR)/pkgconfig + +install-osmesa: default + $(INSTALL) -d $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME)* \ + $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) + +install-dri: + cd drivers/dri && $(MAKE) install -## NOT INSTALLED YET: -## $(INSTALL) -d $(INSTALL_DIR)/include/GLES -## $(INSTALL) -m 644 include/GLES/*.h $(INSTALL_DIR)/include/GLES # Emacs tags tags: etags `find . -name \*.[ch]` $(TOP)/include/GL/*.h + clean: -rm -f */*.o -rm -f */*/*.o -rm -f depend depend.bak libmesa.a libglapi.a -rm -f drivers/*/*.o - (cd x86 && $(MAKE) clean) - (cd x86-64 && $(MAKE) clean) + -@cd drivers/dri && $(MAKE) clean + -@cd drivers/xorg && $(MAKE) clean + -@cd drivers/x11 && $(MAKE) clean + -@cd drivers/osmesa && $(MAKE) clean + -@cd x86 && $(MAKE) clean + -@cd x86-64 && $(MAKE) clean -include depend +-include depend diff --git a/src/mesa/drivers/Makefile b/src/mesa/drivers/Makefile new file mode 100644 index 0000000000..c5998413e8 --- /dev/null +++ b/src/mesa/drivers/Makefile @@ -0,0 +1,29 @@ +# src/mesa/drivers/Makefile + +TOP = ../../.. +include $(TOP)/configs/current + + +default: + @for dir in $(DRIVER_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1; \ + fi \ + done + + +clean: + @for dir in $(DRIVER_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) clean) || exit 1; \ + fi \ + done + + +install: + @for dir in $(DRIVER_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) install) || exit 1; \ + fi \ + done + diff --git a/src/mesa/drivers/beos/Makefile b/src/mesa/drivers/beos/Makefile index f8b7eb69a4..342d7ce024 100644 --- a/src/mesa/drivers/beos/Makefile +++ b/src/mesa/drivers/beos/Makefile @@ -169,28 +169,30 @@ OBJECTS := $(DRIVER_OBJECTS:.cpp=.o) default: depend $(TOP)/$(LIB_DIR) $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) +# XXX FIXME: mesa.a might be libmesa.a now $(MESA_MODULES): - cd $(TOP)/src/mesa; $(MAKE) mesa.a ; + cd $(TOP)/src/mesa && $(MAKE) mesa.a ; + mimeset -f "$@" $(GLU_MODULES): - cd $(GLU_DIR); $(MAKE) $(subst $(GLU_DIR)/,,$(GLU_MODULES)) ; + cd $(GLU_DIR) && $(MAKE) $(subst $(GLU_DIR)/,,$(GLU_MODULES)) ; $(TOP)/$(LIB_DIR): mkdir $(TOP)/$(LIB_DIR) $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(OBJECTS) $(MESA_MODULES) $(GLU_MODULES) - @$(TOP)/bin/mklib -o $(GL_LIB) -install $(TOP)/$(LIB_DIR) -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ + @$(TOP)/bin/mklib -o $(GL_LIB) -ldflags '$(LDFLAGS)' -install $(TOP)/$(LIB_DIR) \ + -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ $(MKLIB_OPTIONS) $(GL_LIB_DEPS) $(OBJECTS) $(MESA_MODULES) $(GLU_MODULES) # $(GLU_OBJECTS): -# cd $(GLU_DIR); $(MAKE) $< ; +# cd $(GLU_DIR) && $(MAKE) $< ; depend: $(DRIVER_SOURCES) $(GLU_SOURCES) - rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(DRIVER_SOURCES) $(GLU_SOURCES) > /dev/null clean: - rm -f depend $(OBJECTS) + -rm -f depend depend.bak $(OBJECTS) include depend diff --git a/src/mesa/drivers/directfb/Makefile b/src/mesa/drivers/directfb/Makefile index c515785b2a..945eb8b3d9 100644 --- a/src/mesa/drivers/directfb/Makefile +++ b/src/mesa/drivers/directfb/Makefile @@ -25,11 +25,24 @@ DIRECTFBGL_MESA_OBJECTS = $(DIRECTFBGL_MESA_SOURCES:.c=.o) DIRECTFBGL_MESA = libidirectfbgl_mesa.so +LIBS = $(TOP)/src/mesa/libmesa.a $(TOP)/src/mesa/libglapi.a + + .c.o: $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(DFB_CFLAGS) $< -o $@ -default: directfbgl_mesa +default: directfb-libgl directfbgl_mesa + + +# XXX this used to be in src/mesa/Makefile and is probably broken now +directfb-libgl: $(CORE_OBJECTS) + @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' -ldflags '$(LDFLAGS)' \ + -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ + -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) $(CORE_OBJECTS) \ + $(GL_LIB_DEPS) + + # Mesa DirectFBGL module directfbgl_mesa: $(DIRECTFBGL_MESA_OBJECTS) @@ -50,5 +63,5 @@ install: clean: - rm -f *.o *.so + -rm -f *.o *.so -- cgit v1.2.3 From f0fdf0c23a16b9e0338eb945c6f88c7a57a9fbd0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 6 Jun 2008 12:33:53 -0600 Subject: gallium: additional comment --- src/gallium/winsys/xlib/Makefile | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/xlib/Makefile b/src/gallium/winsys/xlib/Makefile index c443330942..ec92c79068 100644 --- a/src/gallium/winsys/xlib/Makefile +++ b/src/gallium/winsys/xlib/Makefile @@ -1,5 +1,9 @@ # src/gallium/winsys/xlib/Makefile +# This makefile produces a "stand-alone" libGL.so which is based on +# Xlib (no DRI HW acceleration) + + TOP = ../../../.. include $(TOP)/configs/current -- cgit v1.2.3 From f6e030f531f7292a373a7cd633e8af9f97726266 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 6 Jun 2008 15:10:22 -0600 Subject: egl: change default logging level to _EGL_WARNING --- src/egl/main/egllog.c | 30 ++++++++++++++++-------------- 1 file changed, 16 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/egl/main/egllog.c b/src/egl/main/egllog.c index dc1daaa996..23eb523eeb 100644 --- a/src/egl/main/egllog.c +++ b/src/egl/main/egllog.c @@ -1,5 +1,7 @@ /** * Logging facility for debug/info messages. + * _EGL_FATAL messages are printed to stderr + * The EGL_LOG_LEVEL var controls the output of other warning/info/debug msgs. */ @@ -10,37 +12,37 @@ #include "egllog.h" #define MAXSTRING 1000 -#define FALLBACK_LOG_LEVEL _EGL_DEBUG -#define FALLBACK_LOG_LEVEL_STR "debug" +#define FALLBACK_LOG_LEVEL _EGL_WARNING +#define FALLBACK_LOG_LEVEL_STR "warning" static EGLint ReportingLevel = -1; static void -log_level_initialize (void) +log_level_initialize(void) { - char *log_env = getenv ("EGL_LOG_LEVEL"); + char *log_env = getenv("EGL_LOG_LEVEL"); if (log_env == NULL) { ReportingLevel = FALLBACK_LOG_LEVEL; } - else if (strcasecmp (log_env, "fatal") == 0) { + else if (strcasecmp(log_env, "fatal") == 0) { ReportingLevel = _EGL_FATAL; } - else if (strcasecmp (log_env, "warning") == 0) { + else if (strcasecmp(log_env, "warning") == 0) { ReportingLevel = _EGL_WARNING; } - else if (strcasecmp (log_env, "info") == 0) { + else if (strcasecmp(log_env, "info") == 0) { ReportingLevel = _EGL_INFO; } - else if (strcasecmp (log_env, "debug") == 0) { + else if (strcasecmp(log_env, "debug") == 0) { ReportingLevel = _EGL_DEBUG; } else { - fprintf (stderr, "Unrecognized EGL_LOG_LEVEL environment variable value. " - "Expected one of \"fatal\", \"warning\", \"info\", \"debug\". " - "Got \"%s\". Falling back to \"%s\".\n", - log_env, FALLBACK_LOG_LEVEL_STR); + fprintf(stderr, "Unrecognized EGL_LOG_LEVEL environment variable value. " + "Expected one of \"fatal\", \"warning\", \"info\", \"debug\". " + "Got \"%s\". Falling back to \"%s\".\n", + log_env, FALLBACK_LOG_LEVEL_STR); ReportingLevel = FALLBACK_LOG_LEVEL; } } @@ -59,7 +61,7 @@ _eglLog(EGLint level, const char *fmtStr, ...) static int log_level_initialized = 0; if (!log_level_initialized) { - log_level_initialize (); + log_level_initialize(); log_level_initialized = 1; } @@ -85,7 +87,7 @@ _eglLog(EGLint level, const char *fmtStr, ...) vsnprintf(msg, MAXSTRING, fmtStr, args); va_end(args); - fprintf(stderr, "EGL %s: %s\n", levelStr, msg); + fprintf(stderr, "libEGL %s: %s\n", levelStr, msg); if (level == _EGL_FATAL) { exit(1); /* or abort()? */ -- cgit v1.2.3 From 47ddcd2a7144ea456fcd52b9470e4e41b878af48 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 6 Jun 2008 15:10:42 -0600 Subject: egl: make _eglChooseDRMDriver() non-static --- src/egl/main/egldriver.c | 4 ++-- src/egl/main/egldriver.h | 3 +++ 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 80fa49b7ab..92c529911c 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -39,7 +39,7 @@ static const char *SysFS = "/sys/class"; /** * Given a card number, use sysfs to determine the DRI driver name. */ -static const char * +const char * _eglChooseDRMDriver(int card) { #if 0 @@ -214,7 +214,7 @@ _eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy) void *handle = drv->LibHandle; EGLBoolean b; - _eglLog(_EGL_INFO, "Closing driver"); + _eglLog(_EGL_DEBUG, "Closing %s", drv->Name); /* * XXX check for currently bound context/surfaces and delete them? diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index af028eaf22..1eae6f8034 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -45,6 +45,9 @@ struct _egl_driver extern _EGLDriver *_eglMain(_EGLDisplay *dpy, const char *args); +extern const char * +_eglChooseDRMDriver(int card); + extern const char * _eglChooseDriver(_EGLDisplay *dpy); -- cgit v1.2.3 From f82d5cf76929ed2de29966117f97774b465d6b0d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 6 Jun 2008 15:41:41 -0600 Subject: mesa: disable visual warning unless LIBGL_DEBUG is defined --- src/glx/x11/glxext.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/glx/x11/glxext.c b/src/glx/x11/glxext.c index 2852217ba3..6403cbd56d 100644 --- a/src/glx/x11/glxext.c +++ b/src/glx/x11/glxext.c @@ -671,8 +671,10 @@ filter_modes( __GLcontextModes ** server_modes, if ( do_delete && (m->visualID != 0) ) { do_delete = GL_FALSE; - fprintf(stderr, "libGL warning: 3D driver claims to not support " - "visual 0x%02x\n", m->visualID); + if (getenv("LIBGL_DEBUG")) { + fprintf(stderr, "libGL warning: 3D driver claims to not support " + "visual 0x%02x\n", m->visualID); + } } if ( do_delete ) { -- cgit v1.2.3 From ab102d5a9f1db0f76eabbf22cd05b08ff7bb8448 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 6 Jun 2008 15:42:08 -0600 Subject: egl: clean-ups --- src/gallium/winsys/egl_xlib/Makefile | 20 +++++++++----------- 1 file changed, 9 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile index fe4b931e30..46666650b0 100644 --- a/src/gallium/winsys/egl_xlib/Makefile +++ b/src/gallium/winsys/egl_xlib/Makefile @@ -7,7 +7,7 @@ TOP = ../../../.. include $(TOP)/configs/current -LIBNAME = softpipe_egl.so +DRIVER_NAME = softpipe_egl.so INCLUDE_DIRS = \ @@ -36,30 +36,28 @@ LIBS = \ LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 -.SUFFIXES : .cpp - .c.o: $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $(LOCAL_CFLAGS) $< -o $@ -.cpp.o: - $(CXX) -c $(INCLUDE_DIRS) $(CXXFLAGS) $(LOCAL_CFLAGS) $< -o $@ + +.PHONY: library +default: depend library Makefile -default: $(TOP)/$(LIB_DIR)/$(LIBNAME) + +library: $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) # Make the softpipe_egl.so library -$(TOP)/$(LIB_DIR)/$(LIBNAME): $(WINSYS_OBJECTS) $(LIBS) - $(TOP)/bin/mklib -o $(LIBNAME) \ +$(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(WINSYS_OBJECTS) $(LIBS) + $(TOP)/bin/mklib -o $(DRIVER_NAME) \ -linker "$(CC)" \ -noprefix \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(WINSYS_OBJECTS) \ --start-group $(LIBS) --end-group -# $(GL_LIB_DEPS) - depend: $(ALL_SOURCES) @ echo "running $(MKDEP)" @@ -72,7 +70,7 @@ depend: $(ALL_SOURCES) install: default $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) @if [ -e $(TOP)/$(LIB_DIR) ]; then \ - $(INSTALL) $(TOP)/$(LIB_DIR)/$(LIBNAME) $(INSTALL_DIR)/$(LIB_DIR); \ + $(INSTALL) $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) $(INSTALL_DIR)/$(LIB_DIR); \ fi -- cgit v1.2.3 From 8dafe470fc29cada5f3beab114e2be19fd01dabe Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 6 Jun 2008 15:52:33 -0600 Subject: egl: EGL->DRI adaptor/driver Allows regular DRI drivers to work with libEGL. --- src/egl/drivers/xdri/Makefile | 67 ++++ src/egl/drivers/xdri/egl_xdri.c | 833 ++++++++++++++++++++++++++++++++++++++++ 2 files changed, 900 insertions(+) create mode 100644 src/egl/drivers/xdri/Makefile create mode 100644 src/egl/drivers/xdri/egl_xdri.c (limited to 'src') diff --git a/src/egl/drivers/xdri/Makefile b/src/egl/drivers/xdri/Makefile new file mode 100644 index 0000000000..477a75184a --- /dev/null +++ b/src/egl/drivers/xdri/Makefile @@ -0,0 +1,67 @@ +# src/egl/drivers/xdri/Makefile + +# Build XEGL DRI driver loader library: egl_xdri.so + + +TOP = ../../../.. +include $(TOP)/configs/current + + +DRIVER_NAME = egl_xdri.so + + +INCLUDE_DIRS = \ + -I. \ + -I/usr/include \ + -I/usr/include/drm \ + -I$(TOP)/include \ + -I$(TOP)/include/GL/internal \ + -I$(TOP)/src/mesa/glapi \ + -I$(TOP)/src/mesa/drivers/dri/common \ + -I$(TOP)/src/egl/main \ + -I$(TOP)/src/glx/x11 + +SOURCES = egl_xdri.c + +OBJECTS = $(SOURCES:.c=.o) + +DRM_LIB = `pkg-config --libs libdrm` + + +.c.o: + $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ + + +.PHONY: library + + +default: depend library Makefile + + +library: $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) + + +# Make the egl_xdri.so library +$(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(OBJECTS) + $(TOP)/bin/mklib -o $(DRIVER_NAME) \ + -noprefix \ + -major 1 -minor 0 \ + -install $(TOP)/$(LIB_DIR) \ + -ldl $(OBJECTS) $(DRM_LIB) + + +clean: + rm -f *.o + rm -f *.so + rm -f depend depend.bak + + +depend: $(SOURCES) $(HEADERS) + @ echo "running $(MKDEP)" + @ rm -f depend + @ touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) \ + $(SOURCES) $(HEADERS) >/dev/null 2>/dev/null + +include depend +# DO NOT DELETE diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c new file mode 100644 index 0000000000..4c8511fbbf --- /dev/null +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -0,0 +1,833 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Code to interface a DRI driver to libEGL. + * Note that unlike previous DRI/EGL interfaces, this one is meant to + * be used _with_ X. Applications will use eglCreateWindowSurface() + * to render into X-created windows. + * + * This is an EGL driver that, in turn, loads a regular DRI driver. + * There are some dependencies on code in libGL, but those coudl be + * removed with some effort. + * + * Authors: Brian Paul + */ + + +#include +#include +#include +#include +#include +#include "dlfcn.h" +#include +#include +#include "xf86dri.h" +#include "glxclient.h" +#include "dri_util.h" +#include "drm_sarea.h" + +#define _EGL_PLATFORM_X + +#include "eglconfig.h" +#include "eglcontext.h" +#include "egldisplay.h" +#include "egldriver.h" +#include "eglglobals.h" +#include "eglhash.h" +#include "egllog.h" +#include "eglsurface.h" + + + +#define CALLOC_STRUCT(T) (struct T *) calloc(1, sizeof(struct T)) + + +/** subclass of _EGLDriver */ +struct xdri_egl_driver +{ + _EGLDriver Base; /**< base class */ + + const char *dri_driver_name; /**< name of DRI driver to load */ + void *dri_driver_handle; /**< returned by dlopen(dri_driver_name) */ + + int chipset; + int minor; + int drmFD; + + __DRIscreen driScreen; + __DRIframebuffer framebuffer; + drm_handle_t hSAREA; + drmAddress pSAREA; + char *busID; + drm_magic_t magic; +}; + + +/** subclass of _EGLContext */ +struct xdri_egl_context +{ + _EGLContext Base; /**< base class */ + + __DRIcontext driContext; +}; + + +/** subclass of _EGLSurface */ +struct xdri_egl_surface +{ + _EGLSurface Base; /**< base class */ + + __DRIid driDrawable; /**< DRI surface */ + drm_drawable_t hDrawable; +}; + + +/** subclass of _EGLConfig */ +struct xdri_egl_config +{ + _EGLConfig Base; /**< base class */ + + const __GLcontextModes *mode; /**< corresponding GLX mode */ +}; + + +/** cast wrapper */ +static struct xdri_egl_driver * +xdri_egl_driver(_EGLDriver *drv) +{ + return (struct xdri_egl_driver *) drv; +} + + +/** Map EGLSurface handle to xdri_egl_surface object */ +static struct xdri_egl_surface * +lookup_surface(EGLSurface surf) +{ + _EGLSurface *surface = _eglLookupSurface(surf); + return (struct xdri_egl_surface *) surface; +} + + +/** Map EGLContext handle to xdri_egl_context object */ +static struct xdri_egl_context * +lookup_context(EGLContext c) +{ + _EGLContext *context = _eglLookupContext(c); + return (struct xdri_egl_context *) context; +} + + +/** Map EGLConfig handle to xdri_egl_config object */ +static struct xdri_egl_config * +lookup_config(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config) +{ + _EGLConfig *conf = _eglLookupConfig(drv, dpy, config); + return (struct xdri_egl_config *) conf; +} + + + +/** + * Produce a set of EGL configs. + * Note that we get the list of GLcontextModes from the GLX library. + * This dependency on GLX lib will be removed someday. + */ +static void +create_configs(_EGLDisplay *disp) +{ + const __GLcontextModes *m; + __GLXdisplayPrivate *priv = __glXInitialize(disp->Xdpy); + __GLXscreenConfigs *scrn = priv->screenConfigs; + int id = 1; + + for (m = scrn->configs; m; m = m->next) { + /* EGL requires double-buffered configs */ + if (m->doubleBufferMode) { + struct xdri_egl_config *config = CALLOC_STRUCT(xdri_egl_config); + + _eglInitConfig(&config->Base, id++); + + SET_CONFIG_ATTRIB(&config->Base, EGL_BUFFER_SIZE, m->rgbBits); + SET_CONFIG_ATTRIB(&config->Base, EGL_RED_SIZE, m->redBits); + SET_CONFIG_ATTRIB(&config->Base, EGL_GREEN_SIZE, m->greenBits); + SET_CONFIG_ATTRIB(&config->Base, EGL_BLUE_SIZE, m->blueBits); + SET_CONFIG_ATTRIB(&config->Base, EGL_ALPHA_SIZE, m->alphaBits); + SET_CONFIG_ATTRIB(&config->Base, EGL_DEPTH_SIZE, m->depthBits); + SET_CONFIG_ATTRIB(&config->Base, EGL_STENCIL_SIZE, m->stencilBits); + SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_ID, m->visualID); + SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_TYPE, m->visualType); + /* XXX possibly other things to init... */ + + /* Ptr from EGL config to GLcontextMode. Used in CreateContext(). */ + config->mode = m; + + _eglAddConfig(disp, &config->Base); + } + } +} + + +/** + * Called via __DRIinterfaceMethods object + */ +static __DRIfuncPtr +dri_get_proc_address(const char * proc_name) +{ + return NULL; +} + + +static void +dri_context_modes_destroy(__GLcontextModes *modes) +{ + _eglLog(_EGL_DEBUG, "%s", __FUNCTION__); + + while (modes) { + __GLcontextModes * const next = modes->next; + free(modes); + modes = next; + } +} + + +/** + * Create a linked list of 'count' GLcontextModes. + * These are used during the client/server visual negotiation phase, + * then discarded. + */ +static __GLcontextModes * +dri_context_modes_create(unsigned count, size_t minimum_size) +{ + /* This code copied from libGLX, and modified */ + const size_t size = (minimum_size > sizeof(__GLcontextModes)) + ? minimum_size : sizeof(__GLcontextModes); + __GLcontextModes * head = NULL; + __GLcontextModes ** next; + unsigned i; + + next = & head; + for (i = 0 ; i < count ; i++) { + *next = (__GLcontextModes *) calloc(1, size); + if (*next == NULL) { + dri_context_modes_destroy(head); + head = NULL; + break; + } + + (*next)->doubleBufferMode = 1; + (*next)->visualID = GLX_DONT_CARE; + (*next)->visualType = GLX_DONT_CARE; + (*next)->visualRating = GLX_NONE; + (*next)->transparentPixel = GLX_NONE; + (*next)->transparentRed = GLX_DONT_CARE; + (*next)->transparentGreen = GLX_DONT_CARE; + (*next)->transparentBlue = GLX_DONT_CARE; + (*next)->transparentAlpha = GLX_DONT_CARE; + (*next)->transparentIndex = GLX_DONT_CARE; + (*next)->xRenderable = GLX_DONT_CARE; + (*next)->fbconfigID = GLX_DONT_CARE; + (*next)->swapMethod = GLX_SWAP_UNDEFINED_OML; + (*next)->bindToTextureRgb = GLX_DONT_CARE; + (*next)->bindToTextureRgba = GLX_DONT_CARE; + (*next)->bindToMipmapTexture = GLX_DONT_CARE; + (*next)->bindToTextureTargets = 0; + (*next)->yInverted = GLX_DONT_CARE; + + next = & ((*next)->next); + } + + return head; +} + + +static __DRIscreen * +dri_find_dri_screen(__DRInativeDisplay *ndpy, int scrn) +{ + /* unused? */ + return NULL; +} + + +static GLboolean +dri_window_exists(__DRInativeDisplay *dpy, __DRIid draw) +{ + return EGL_TRUE; +} + + +static GLboolean +dri_create_context(__DRInativeDisplay *dpy, int screenNum, int configID, + void * contextID, drm_context_t * hw_context) +{ + assert(configID >= 0); + return XF86DRICreateContextWithConfig(dpy, screenNum, + configID, contextID, hw_context); +} + + +static GLboolean +dri_destroy_context(__DRInativeDisplay * ndpy, int screen, __DRIid context) +{ + return XF86DRIDestroyContext(ndpy, screen, context); +} + + +static GLboolean +dri_create_drawable(__DRInativeDisplay * ndpy, int screen, + __DRIid drawable, drm_drawable_t * hHWDrawable) +{ + _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__); + + /* Create DRI drawable for given window ID (drawable) */ + if (!XF86DRICreateDrawable(ndpy, screen, drawable, hHWDrawable)) + return EGL_FALSE; + + return EGL_TRUE; +} + + +static GLboolean +dri_destroy_drawable(__DRInativeDisplay * ndpy, int screen, __DRIid drawable) +{ + _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__); + return XF86DRIDestroyDrawable(ndpy, screen, drawable); +} + + +static GLboolean +dri_get_drawable_info(__DRInativeDisplay *dpy, int scrn, + __DRIid draw, unsigned int * index, unsigned int * stamp, + int * x, int * y, int * width, int * height, + int * numClipRects, drm_clip_rect_t ** pClipRects, + int * backX, int * backY, + int * numBackClipRects, + drm_clip_rect_t ** pBackClipRects) +{ + _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__); + + if (!XF86DRIGetDrawableInfo(dpy, scrn, draw, index, stamp, + x, y, width, height, + numClipRects, pClipRects, + backX, backY, + numBackClipRects, pBackClipRects)) { + return EGL_FALSE; + } + + return EGL_TRUE; +} + + +/** + * Table of functions exported by the loader to the driver. + */ +static const __DRIinterfaceMethods interface_methods = { + dri_get_proc_address, + + dri_context_modes_create, + dri_context_modes_destroy, + + dri_find_dri_screen, + dri_window_exists, + + dri_create_context, + dri_destroy_context, + + dri_create_drawable, + dri_destroy_drawable, + dri_get_drawable_info, + + NULL,/*__eglGetUST,*/ + NULL,/*__eglGetMSCRate,*/ +}; + + + +static EGLBoolean +init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp) +{ + static const char createNewScreenName[] = "__driCreateNewScreen_20050727"; + PFNCREATENEWSCREENFUNC createNewScreen; + int api_ver = 0;/*__glXGetInternalVersion();*/ + __DRIversion ddx_version; + __DRIversion dri_version; + __DRIversion drm_version; + drmVersionPtr version; + drm_handle_t hFB; + int newlyopened; + int status; + __GLcontextModes *modes; + int scrn = DefaultScreen(disp->Xdpy); + + createNewScreen = (PFNCREATENEWSCREENFUNC) + dlsym(xdri_drv->dri_driver_handle, createNewScreenName); + if (!createNewScreen) { + _eglLog(_EGL_WARNING, "XDRI: Couldn't find %s function in the driver.", + createNewScreenName); + return EGL_FALSE; + } + else { + _eglLog(_EGL_DEBUG, "XDRI: Found %s", createNewScreenName); + } + + /* + * Get the DRI X extension version. + */ + dri_version.major = 4; + dri_version.minor = 0; + dri_version.patch = 0; + + + if (!XF86DRIOpenConnection(disp->Xdpy, scrn, + &xdri_drv->hSAREA, &xdri_drv->busID)) { + _eglLog(_EGL_WARNING, "XF86DRIOpenConnection failed"); + } + + xdri_drv->drmFD = drmOpenOnce(NULL, xdri_drv->busID, &newlyopened); + if (xdri_drv->drmFD < 0) { + perror("drmOpenOnce failed: "); + return EGL_FALSE; + } + else { + _eglLog(_EGL_DEBUG, "XDRI: drmOpenOnce returned %d", xdri_drv->drmFD); + } + + + if (drmGetMagic(xdri_drv->drmFD, &xdri_drv->magic)) { + perror("drmGetMagic failed: "); + return EGL_FALSE; + } + + version = drmGetVersion(xdri_drv->drmFD); + if (version) { + drm_version.major = version->version_major; + drm_version.minor = version->version_minor; + drm_version.patch = version->version_patchlevel; + drmFreeVersion(version); + _eglLog(_EGL_DEBUG, "XDRI: Got DRM version %d.%d.%d", + drm_version.major, + drm_version.minor, + drm_version.patch); + } + else { + drm_version.major = -1; + drm_version.minor = -1; + drm_version.patch = -1; + _eglLog(_EGL_WARNING, "XDRI: drmGetVersion() failed"); + return EGL_FALSE; + } + + /* Authenticate w/ server. + */ + if (!XF86DRIAuthConnection(disp->Xdpy, scrn, xdri_drv->magic)) { + _eglLog(_EGL_WARNING, "XDRI: XF86DRIAuthConnection() failed"); + return EGL_FALSE; + } + else { + _eglLog(_EGL_DEBUG, "XDRI: XF86DRIAuthConnection() success"); + } + + /* Get ddx version. + */ + { + char *driverName; + + /* + * Get device name (like "tdfx") and the ddx version + * numbers. We'll check the version in each DRI driver's + * "createNewScreen" function. + */ + if (!XF86DRIGetClientDriverName(disp->Xdpy, scrn, + &ddx_version.major, + &ddx_version.minor, + &ddx_version.patch, + &driverName)) { + _eglLog(_EGL_WARNING, "XDRI: XF86DRIGetClientDriverName failed"); + return EGL_FALSE; + } + else { + _eglLog(_EGL_DEBUG, "XDRI: XF86DRIGetClientDriverName returned %s", driverName); + } + } + + /* Get framebuffer info. + */ + { + int junk; + if (!XF86DRIGetDeviceInfo(disp->Xdpy, scrn, + &hFB, + &junk, + &xdri_drv->framebuffer.size, + &xdri_drv->framebuffer.stride, + &xdri_drv->framebuffer.dev_priv_size, + &xdri_drv->framebuffer.dev_priv)) { + _eglLog(_EGL_WARNING, "XDRI: XF86DRIGetDeviceInfo() failed"); + return EGL_FALSE; + } + else { + _eglLog(_EGL_DEBUG, "XDRI: XF86DRIGetDeviceInfo() success"); + } + xdri_drv->framebuffer.width = DisplayWidth(disp->Xdpy, scrn); + xdri_drv->framebuffer.height = DisplayHeight(disp->Xdpy, scrn); + } + + /* Map the framebuffer region. (this may not be needed) + */ + status = drmMap(xdri_drv->drmFD, hFB, xdri_drv->framebuffer.size, + (drmAddressPtr) &xdri_drv->framebuffer.base); + if (status != 0) { + _eglLog(_EGL_WARNING, "XDRI: drmMap(framebuffer) failed"); + return EGL_FALSE; + } + else { + _eglLog(_EGL_DEBUG, "XDRI: drmMap(framebuffer) success"); + } + + /* Map the SAREA region. + */ + status = drmMap(xdri_drv->drmFD, xdri_drv->hSAREA, SAREA_MAX, &xdri_drv->pSAREA); + if (status != 0) { + _eglLog(_EGL_WARNING, "XDRI: drmMap(sarea) failed"); + return EGL_FALSE; + } + else { + _eglLog(_EGL_DEBUG, "XDRI: drmMap(sarea) success"); + } + + /* Create the DRI screen. + */ + xdri_drv->driScreen.private = createNewScreen(disp->Xdpy, + scrn, /* screen number */ + &xdri_drv->driScreen, + NULL, /* visuals */ + &ddx_version, + &dri_version, + &drm_version, + &xdri_drv->framebuffer, + xdri_drv->pSAREA, + xdri_drv->drmFD, + api_ver, + &interface_methods, + &modes); + if (!xdri_drv->driScreen.private) { + _eglLog(_EGL_WARNING, "XDRI: create new screen failed"); + return EGL_FALSE; + } + else { + _eglLog(_EGL_DEBUG, "XDRI: create new screen success"); + } + + create_configs(disp); + + /* print modes / debug */ + if (0) { + __GLcontextModes *m; + + for (m = modes; m; m = m->next) { + _eglLog(_EGL_DEBUG, + "mode ID 0x%x rgba %d %d %d %d z %d s %d db %d\n", m->visualID, + m->redBits, m->greenBits, m->blueBits, m->alphaBits, + m->depthBits, m->stencilBits, m->doubleBufferMode); + } + } + + return EGL_TRUE; +} + + +static EGLBoolean +load_dri_driver(struct xdri_egl_driver *xdri_drv) +{ + char filename[100]; + int flags = RTLD_NOW; + + snprintf(filename, sizeof(filename), "%s.so", xdri_drv->dri_driver_name); + _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s)", filename); + + xdri_drv->dri_driver_handle = dlopen(filename, flags); + if (!xdri_drv->dri_driver_handle) { + _eglLog(_EGL_WARNING, "XDRI Could not open %s (%s)", + filename, dlerror()); + + return EGL_FALSE; + } + return EGL_TRUE; +} + + +/** + * Called via eglInitialize(), xdri_drv->API.Initialize(). + */ +static EGLBoolean +xdri_eglInitialize(_EGLDriver *drv, EGLDisplay dpy, + EGLint *minor, EGLint *major) +{ + struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv); + _EGLDisplay *disp = _eglLookupDisplay(dpy); + static char name[100]; + + _eglLog(_EGL_DEBUG, "XDRI: eglInitialize"); + + if (!disp->Xdpy) { + disp->Xdpy = XOpenDisplay(NULL); + if (!disp->Xdpy) { + _eglLog(_EGL_WARNING, "XDRI: XOpenDisplay failed"); + return EGL_FALSE; + } + } + + /* choose the DRI driver to load */ + xdri_drv->dri_driver_name = _eglChooseDRMDriver(0); + if (!load_dri_driver(xdri_drv)) + return EGL_FALSE; + + if (!init_drm(xdri_drv, disp)) + return EGL_FALSE; + + xdri_drv->Base.Initialized = EGL_TRUE; + + snprintf(name, sizeof(name), "X/DRI:%s", xdri_drv->dri_driver_name); + xdri_drv->Base.Name = name; + + /* we're supporting EGL 1.4 */ + *minor = 1; + *major = 4; + + return EGL_TRUE; +} + + +/** + * Called via eglTerminate(), drv->API.Terminate(). + */ +static EGLBoolean +xdri_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) +{ + struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv); + + _eglLog(_EGL_DEBUG, "XDRI: eglTerminate"); + + _eglLog(_EGL_DEBUG, "XDRI: Closing %s", xdri_drv->dri_driver_name); + dlclose(xdri_drv->dri_driver_handle); + xdri_drv->dri_driver_handle = NULL; + + free((void*) xdri_drv->dri_driver_name); + + return EGL_TRUE; +} + + +/** + * Called via eglCreateContext(), drv->API.CreateContext(). + */ +static EGLContext +xdri_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, + EGLContext share_list, const EGLint *attrib_list) +{ + struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv); + _EGLDisplay *disp = _eglLookupDisplay(dpy); + struct xdri_egl_config *xdri_config = lookup_config(drv, dpy, config); + void *shared = NULL; + int renderType = GLX_RGBA_BIT; + + struct xdri_egl_context *xdri_ctx = CALLOC_STRUCT(xdri_egl_context); + if (!xdri_ctx) + return EGL_NO_CONTEXT; + + if (!_eglInitContext(drv, dpy, &xdri_ctx->Base, config, attrib_list)) { + free(xdri_ctx); + return EGL_NO_CONTEXT; + } + + assert(xdri_config); + + xdri_ctx->driContext.private = + xdri_drv->driScreen.createNewContext(disp->Xdpy, + xdri_config->mode, renderType, + shared, &xdri_ctx->driContext); + if (!xdri_ctx->driContext.private) { + free(xdri_ctx); + return EGL_NO_CONTEXT; + } + + xdri_ctx->driContext.mode = xdri_config->mode; + + return _eglGetContextHandle(&xdri_ctx->Base); +} + + +/** + * Called via eglMakeCurrent(), drv->API.MakeCurrent(). + */ +static EGLBoolean +xdri_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d, + EGLSurface r, EGLContext context) +{ + _EGLDisplay *disp = _eglLookupDisplay(dpy); + struct xdri_egl_context *xdri_ctx = lookup_context(context); + struct xdri_egl_surface *xdri_draw = lookup_surface(d); + struct xdri_egl_surface *xdri_read = lookup_surface(r); + __DRIid draw = xdri_draw->driDrawable; + __DRIid read = xdri_read->driDrawable; + int scrn = DefaultScreen(disp->Xdpy); + + if (!_eglMakeCurrent(drv, dpy, d, r, context)) + return EGL_FALSE; + + + if (!xdri_ctx->driContext.bindContext(disp->Xdpy, scrn, draw, read, + &xdri_ctx->driContext)) { + return EGL_FALSE; + } + + return EGL_TRUE; +} + + +/** + * Called via eglCreateWindowSurface(), drv->API.CreateWindowSurface(). + */ +static EGLSurface +xdri_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, + NativeWindowType window, const EGLint *attrib_list) +{ + _EGLDisplay *disp = _eglLookupDisplay(dpy); + struct xdri_egl_surface *xdri_surf; + int scrn = DefaultScreen(disp->Xdpy); + + xdri_surf = CALLOC_STRUCT(xdri_egl_surface); + if (!xdri_surf) + return EGL_NO_SURFACE; + + if (!_eglInitSurface(drv, dpy, &xdri_surf->Base, EGL_WINDOW_BIT, + config, attrib_list)) { + free(xdri_surf); + return EGL_FALSE; + } + + if (!XF86DRICreateDrawable(disp->Xdpy, scrn, window, &xdri_surf->hDrawable)) { + free(xdri_surf); + return EGL_FALSE; + } + + xdri_surf->driDrawable = window; + + _eglSaveSurface(&xdri_surf->Base); + + _eglLog(_EGL_DEBUG, + "XDRI: CreateWindowSurface win 0x%x handle %d hDrawable %d", + (int) window, _eglGetSurfaceHandle(&xdri_surf->Base), + (int) xdri_surf->hDrawable); + + return _eglGetSurfaceHandle(&xdri_surf->Base); +} + + +static EGLBoolean +xdri_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) +{ + struct xdri_egl_surface *xdri_surf = lookup_surface(surface); + if (xdri_surf) { + _eglHashRemove(_eglGlobal.Surfaces, (EGLuint) surface); + if (xdri_surf->Base.IsBound) { + xdri_surf->Base.DeletePending = EGL_TRUE; + } + else { + /* + st_unreference_framebuffer(&surf->Framebuffer); + */ + free(xdri_surf); + } + return EGL_TRUE; + } + else { + _eglError(EGL_BAD_SURFACE, "eglDestroySurface"); + return EGL_FALSE; + } +} + + +static EGLBoolean +xdri_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) +{ + _EGLDisplay *disp = _eglLookupDisplay(dpy); + + _eglLog(_EGL_DEBUG, "XDRI: EGL SwapBuffers"); + + /* error checking step: */ + if (!_eglSwapBuffers(drv, dpy, draw)) + return EGL_FALSE; + + { + struct xdri_egl_surface *xdri_surf = lookup_surface(draw); + __GLXdisplayPrivate *priv = __glXInitialize(disp->Xdpy); + __GLXscreenConfigs *scrn = priv->screenConfigs; + __DRIscreen *psc = &scrn->driScreen; + __DRIdrawable * const pdraw = psc->getDrawable(disp->Xdpy, + xdri_surf->driDrawable, + psc->private); + + if (pdraw) + pdraw->swapBuffers(disp->Xdpy, pdraw->private); + else + _eglLog(_EGL_WARNING, "pdraw is null in SwapBuffers"); + } + + return EGL_TRUE; +} + + +/** + * This is the main entrypoint into the driver, called by libEGL. + * Create a new _EGLDriver object and init its dispatch table. + */ +_EGLDriver * +_eglMain(_EGLDisplay *disp, const char *args) +{ + struct xdri_egl_driver *xdri_drv = CALLOC_STRUCT(xdri_egl_driver); + if (!xdri_drv) + return NULL; + + _eglInitDriverFallbacks(&xdri_drv->Base); + xdri_drv->Base.API.Initialize = xdri_eglInitialize; + xdri_drv->Base.API.Terminate = xdri_eglTerminate; + + xdri_drv->Base.API.CreateContext = xdri_eglCreateContext; + xdri_drv->Base.API.MakeCurrent = xdri_eglMakeCurrent; + xdri_drv->Base.API.CreateWindowSurface = xdri_eglCreateWindowSurface; + xdri_drv->Base.API.DestroySurface = xdri_eglDestroySurface; + xdri_drv->Base.API.SwapBuffers = xdri_eglSwapBuffers; + + xdri_drv->Base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; + xdri_drv->Base.Name = "X/DRI"; + + _eglLog(_EGL_DEBUG, "XDRI: main(%s)", args); + + return &xdri_drv->Base; +} -- cgit v1.2.3 From f374734d61011acc2d7b4e5f390d806603a833bd Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 6 Jun 2008 15:59:54 -0600 Subject: egl: improved Makefile --- src/egl/drivers/Makefile | 20 ++++++++++---------- 1 file changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/Makefile b/src/egl/drivers/Makefile index 837626e3d6..a2d67ca5a8 100644 --- a/src/egl/drivers/Makefile +++ b/src/egl/drivers/Makefile @@ -1,8 +1,10 @@ # src/egl/drivers/Makefile TOP = ../../.. +include $(TOP)/configs/current -SUBDIRS = demo dri + +SUBDIRS = demo dri xdri default: conditional_subdirs @@ -10,19 +12,17 @@ default: conditional_subdirs # depending on $DRIVER_DIRS... conditional_subdirs: - @echo "DRIVER_DIRS = " $(DRIVER_DIRS) - @case "$$DRIVER_DIRS" in \ - dri) $(MAKE) dri_subdir || exit 1 ;; \ - *) ;; \ - esac; - + @if [ "${DRIVER_DIRS}" = "dri" ] ; then \ + $(MAKE) dri_subdirs ; \ + fi -dri_subdir: - (cd dri ; $(MAKE)) || exit 1 ; \ +dri_subdirs: + @ (cd dri ; $(MAKE)) || exit 1 + @ (cd xdri ; $(MAKE)) || exit 1 demo_subdir: - (cd demo ; $(MAKE)) || exit 1 ; \ + @ (cd demo ; $(MAKE)) || exit 1 -- cgit v1.2.3 From 3c207936bea0f160549ff8ba76ef0bbcb7e0ed81 Mon Sep 17 00:00:00 2001 From: Jose Fonseca Date: Sat, 7 Jun 2008 12:34:45 +0900 Subject: Fix CRLF line endings. --- src/glut/os2/WarpWin.cpp | 838 ++++++------- src/glut/os2/glut_cindex.cpp | 516 ++++---- src/glut/os2/glut_gamemode.cpp | 1358 ++++++++++----------- src/glut/os2/glut_win.cpp | 2440 ++++++++++++++++++------------------- src/glut/os2/glut_winmisc.cpp | 252 ++-- src/glut/os2/os2_glx.cpp | 290 ++--- src/glut/os2/os2_menu.cpp | 1064 ++++++++--------- src/glut/os2/os2_winproc.cpp | 2592 ++++++++++++++++++++-------------------- 8 files changed, 4675 insertions(+), 4675 deletions(-) (limited to 'src') diff --git a/src/glut/os2/WarpWin.cpp b/src/glut/os2/WarpWin.cpp index ee746ecc76..bd5a6e80c7 100644 --- a/src/glut/os2/WarpWin.cpp +++ b/src/glut/os2/WarpWin.cpp @@ -1,420 +1,420 @@ -/* WarpWin.c */ -/* glut for Warp */ -#include -#include - -#include "WarpWin.h" -#include "WarpGL.h" - -#define POKA 0 - -/* global variables that must be set for some functions to operate - correctly. */ -HDC XHDC; -HWND XHWND; - - -void -XStoreColor(Display* display, Colormap colormap, XColor* color) -{ - /* KLUDGE: set XHDC to 0 if the palette should NOT be realized after - setting the color. set XHDC to the correct HDC if it should. */ - - LONG pe; - ULONG cclr; - int r,g,b; - /* X11 stores color from 0-65535, Win32 expects them to be 0-256, so - twiddle the bits ( / 256). */ - r = color->red / 256; - g = color->green / 256; - b = color->blue / 256; - pe = LONGFromRGB(r,g,b); - /* make sure we use this flag, otherwise the colors might get mapped - to another place in the colormap, and when we glIndex() that - color, it may have moved (argh!!) */ - pe |= (PC_NOCOLLAPSE<<24); -/* This function changes the entries in a palette. */ -#if POKA -OS2: - rc = GpiSetPaletteEntries(colormap,LCOLF_CONSECRGB, color->pixel, 1, &pe); - GpiSelectPalette(hps,colormap); - WinRealizePalette(hwnd,hps,&cclr); -source Win: - if (XHDC) { - UnrealizeObject(colormap); - SelectPalette(XHDC, colormap, FALSE); - RealizePalette(XHDC); - - } -#endif -} - -void -XSetWindowColormap(Display* display, Window window, Colormap colormap) -{ -#if POKA - HDC hdc = GetDC(window); - - /* if the third parameter is FALSE, the logical colormap is copied - into the device palette when the application is in the - foreground, if it is TRUE, the colors are mapped into the current - palette in the best possible way. */ - SelectPalette(hdc, colormap, FALSE); - RealizePalette(hdc); - - /* note that we don't have to release the DC, since our window class - uses the WC_OWNDC flag! */ -#endif -} - - -/* display, root and visual - don't used at all */ -Colormap -XCreateColormap(Display* display, Window root, Visual* visual, int alloc) -{ - /* KLUDGE: this function needs XHDC to be set to the HDC currently - being operated on before it is invoked! */ - - HPAL palette; - int n; -#if POKA - PIXELFORMATDESCRIPTOR pfd; - LOGPALETTE *logical; - - /* grab the pixel format */ - memset(&pfd, 0, sizeof(PIXELFORMATDESCRIPTOR)); - DescribePixelFormat(XHDC, GetPixelFormat(XHDC), - sizeof(PIXELFORMATDESCRIPTOR), &pfd); - - if (!(pfd.dwFlags & PFD_NEED_PALETTE || - pfd.iPixelType == PFD_TYPE_COLORINDEX)) - { - return 0; - } - - n = 1 << pfd.cColorBits; - - /* allocate a bunch of memory for the logical palette (assume 256 - colors in a Win32 palette */ - logical = (LOGPALETTE*)malloc(sizeof(LOGPALETTE) + - sizeof(PALETTEENTRY) * n); - memset(logical, 0, sizeof(LOGPALETTE) + sizeof(PALETTEENTRY) * n); - - /* set the entries in the logical palette */ - logical->palVersion = 0x300; - logical->palNumEntries = n; - - /* start with a copy of the current system palette */ - GetSystemPaletteEntries(XHDC, 0, 256, &logical->palPalEntry[0]); - - if (pfd.iPixelType == PFD_TYPE_RGBA) { - int redMask = (1 << pfd.cRedBits) - 1; - int greenMask = (1 << pfd.cGreenBits) - 1; - int blueMask = (1 << pfd.cBlueBits) - 1; - int i; - - /* fill in an RGBA color palette */ - for (i = 0; i < n; ++i) { - logical->palPalEntry[i].peRed = - (((i >> pfd.cRedShift) & redMask) * 255) / redMask; - logical->palPalEntry[i].peGreen = - (((i >> pfd.cGreenShift) & greenMask) * 255) / greenMask; - logical->palPalEntry[i].peBlue = - (((i >> pfd.cBlueShift) & blueMask) * 255) / blueMask; - logical->palPalEntry[i].peFlags = 0; - } - } - - palette = CreatePalette(logical); - free(logical); - - SelectPalette(XHDC, palette, FALSE); - RealizePalette(XHDC); -#endif /* POKA */ - - return palette; -} - - - -int GetSystemMetrics( int mode) -{ RECTL rect; - - switch(mode) - { case SM_CXSCREEN: - WinQueryWindowRect(HWND_DESKTOP,&rect); - return (rect.xRight-rect.xLeft); - break; - case SM_CYSCREEN: - WinQueryWindowRect(HWND_DESKTOP,&rect); - return (rect.yTop-rect.yBottom); - break; - default: ; - } - return 0; -} -/* - * XParseGeometry parses strings of the form - * "=x{+-}{+-}", where - * width, height, xoffset, and yoffset are unsigned integers. - * Example: "=80x24+300-49" - * The equal sign is optional. - * It returns a bitmask that indicates which of the four values - * were actually found in the string. For each value found, - * the corresponding argument is updated; for each value - * not found, the corresponding argument is left unchanged. - */ - -static int -ReadInteger(char *string, char **NextString) -{ - register int Result = 0; - int Sign = 1; - - if (*string == '+') - string++; - else if (*string == '-') - { - string++; - Sign = -1; - } - for (; (*string >= '0') && (*string <= '9'); string++) - { - Result = (Result * 10) + (*string - '0'); - } - *NextString = string; - if (Sign >= 0) - return (Result); - else - return (-Result); -} - -int XParseGeometry(char *string, int *x, int *y, unsigned int *width, unsigned int *height) -{ - int mask = NoValue; - register char *strind; - unsigned int tempWidth, tempHeight; - int tempX, tempY; - char *nextCharacter; - - if ( (string == NULL) || (*string == '\0')) return(mask); - if (*string == '=') - string++; /* ignore possible '=' at beg of geometry spec */ - - strind = (char *)string; - if (*strind != '+' && *strind != '-' && *strind != 'x') { - tempWidth = ReadInteger(strind, &nextCharacter); - if (strind == nextCharacter) - return (0); - strind = nextCharacter; - mask |= WidthValue; - } - - if (*strind == 'x' || *strind == 'X') { - strind++; - tempHeight = ReadInteger(strind, &nextCharacter); - if (strind == nextCharacter) - return (0); - strind = nextCharacter; - mask |= HeightValue; - } - - if ((*strind == '+') || (*strind == '-')) { - if (*strind == '-') { - strind++; - tempX = -ReadInteger(strind, &nextCharacter); - if (strind == nextCharacter) - return (0); - strind = nextCharacter; - mask |= XNegative; - - } - else - { strind++; - tempX = ReadInteger(strind, &nextCharacter); - if (strind == nextCharacter) - return(0); - strind = nextCharacter; - } - mask |= XValue; - if ((*strind == '+') || (*strind == '-')) { - if (*strind == '-') { - strind++; - tempY = -ReadInteger(strind, &nextCharacter); - if (strind == nextCharacter) - return(0); - strind = nextCharacter; - mask |= YNegative; - - } - else - { - strind++; - tempY = ReadInteger(strind, &nextCharacter); - if (strind == nextCharacter) - return(0); - strind = nextCharacter; - } - mask |= YValue; - } - } - - /* If strind isn't at the end of the string the it's an invalid - geometry specification. */ - - if (*strind != '\0') return (0); - - if (mask & XValue) - *x = tempX; - if (mask & YValue) - *y = tempY; - if (mask & WidthValue) - *width = tempWidth; - if (mask & HeightValue) - *height = tempHeight; - return (mask); -} - -int gettimeofday(struct timeval* tp, void* tzp) -{ - DATETIME DateTime; - APIRET ulrc; /* Return Code. */ - - ulrc = DosGetDateTime(&DateTime); - tp->tv_sec = 60 * (60*DateTime.hours + DateTime.minutes) + DateTime.seconds; - tp->tv_usec = DateTime.hundredths * 10000; - return 0; -} - - -int -XPending(Display* display) -{ - /* similar functionality...I don't think that it is exact, but this - will have to do. */ - QMSG msg; - extern HAB hab; /* PM anchor block handle */ - -//?? WinPeekMsg(hab - return WinPeekMsg(hab, &msg, NULLHANDLE, 0, 0, PM_NOREMOVE); -} - -void -__glutAdjustCoords(Window parent, int* x, int* y, int* width, int* height) -{ - RECTL rect; - - /* adjust the window rectangle because Win32 thinks that the x, y, - width & height are the WHOLE window (including decorations), - whereas GLUT treats the x, y, width & height as only the CLIENT - area of the window. */ - rect.xLeft = *x; rect.yTop = *y; - rect.xRight = *x + *width; rect.yBottom = *y + *height; - - /* must adjust the coordinates according to the correct style - because depending on the style, there may or may not be - borders. */ -//?? AdjustWindowRect(&rect, WS_CLIPSIBLINGS | WS_CLIPCHILDREN | -//?? (parent ? WS_CHILD : WS_OVERLAPPEDWINDOW), -//?? FALSE); - /* FALSE in the third parameter = window has no menu bar */ - - /* readjust if the x and y are offscreen */ - if(rect.xLeft < 0) { - *x = 0; - } else { - *x = rect.xLeft; - } - - if(rect.yTop < 0) { - *y = 0; - } else { - *y = rect.yTop; - } - - *width = rect.xRight - rect.xLeft; /* adjusted width */ - *height = -(rect.yBottom - rect.yTop); /* adjusted height */ -} - - -int -__glutGetTransparentPixel(Display * dpy, XVisualInfo * vinfo) -{ - /* the transparent pixel on Win32 is always index number 0. So if - we put this routine in this file, we can avoid compiling the - whole of layerutil.c which is where this routine normally comes - from. */ - return 0; -} - -/* Translate point coordinates src_x and src_y from src to dst */ - -Bool -XTranslateCoordinates(Display *display, Window src, Window dst, - int src_x, int src_y, - int* dest_x_return, int* dest_y_return, - Window* child_return) -{ - SWP swp_src,swp_dst; - - WinQueryWindowPos(src,&swp_src); - WinQueryWindowPos(dst,&swp_dst); - - *dest_x_return = src_x + swp_src.x - swp_dst.x; - *dest_y_return = src_y + swp_src.y - swp_dst.y; - - /* just to make compilers happy...we don't use the return value. */ - return True; -} - -Status -XGetGeometry(Display* display, Window window, Window* root_return, - int* x_return, int* y_return, - unsigned int* width_return, unsigned int* height_return, - unsigned int *border_width_return, unsigned int* depth_return) -{ - /* KLUDGE: doesn't return the border_width or depth or root, x & y - are in screen coordinates. */ - SWP swp_src; - WinQueryWindowPos(window,&swp_src); - - *x_return = swp_src.x; - *y_return = swp_src.y; - *width_return = swp_src.cx; - *height_return = swp_src.cy; - - /* just to make compilers happy...we don't use the return value. */ - return 1; -} - -/* Get Display Width in millimeters */ -int -DisplayWidthMM(Display* display, int screen) -{ - int width; - LONG *pVC_Caps; - pVC_Caps = GetVideoConfig(NULLHANDLE); - width = (int)( 0.001 * pVC_Caps[CAPS_WIDTH] / pVC_Caps[CAPS_HORIZONTAL_RESOLUTION]);/* mm */ - return width; -} - -/* Get Display Height in millimeters */ -int -DisplayHeightMM(Display* display, int screen) -{ - int height; - LONG *pVC_Caps; - pVC_Caps = GetVideoConfig(NULLHANDLE); - height = (int)( 0.001 * pVC_Caps[CAPS_HEIGHT] / pVC_Caps[CAPS_VERTICAL_RESOLUTION]); /* mm */ - return height; -} - -void ScreenToClient( HWND hwnd, POINTL *point) -{ - SWP swp_src; - WinQueryWindowPos(hwnd,&swp_src); - point->x -= swp_src.x; - point->y -= swp_src.y; -} - +/* WarpWin.c */ +/* glut for Warp */ +#include +#include + +#include "WarpWin.h" +#include "WarpGL.h" + +#define POKA 0 + +/* global variables that must be set for some functions to operate + correctly. */ +HDC XHDC; +HWND XHWND; + + +void +XStoreColor(Display* display, Colormap colormap, XColor* color) +{ + /* KLUDGE: set XHDC to 0 if the palette should NOT be realized after + setting the color. set XHDC to the correct HDC if it should. */ + + LONG pe; + ULONG cclr; + int r,g,b; + /* X11 stores color from 0-65535, Win32 expects them to be 0-256, so + twiddle the bits ( / 256). */ + r = color->red / 256; + g = color->green / 256; + b = color->blue / 256; + pe = LONGFromRGB(r,g,b); + /* make sure we use this flag, otherwise the colors might get mapped + to another place in the colormap, and when we glIndex() that + color, it may have moved (argh!!) */ + pe |= (PC_NOCOLLAPSE<<24); +/* This function changes the entries in a palette. */ +#if POKA +OS2: + rc = GpiSetPaletteEntries(colormap,LCOLF_CONSECRGB, color->pixel, 1, &pe); + GpiSelectPalette(hps,colormap); + WinRealizePalette(hwnd,hps,&cclr); +source Win: + if (XHDC) { + UnrealizeObject(colormap); + SelectPalette(XHDC, colormap, FALSE); + RealizePalette(XHDC); + + } +#endif +} + +void +XSetWindowColormap(Display* display, Window window, Colormap colormap) +{ +#if POKA + HDC hdc = GetDC(window); + + /* if the third parameter is FALSE, the logical colormap is copied + into the device palette when the application is in the + foreground, if it is TRUE, the colors are mapped into the current + palette in the best possible way. */ + SelectPalette(hdc, colormap, FALSE); + RealizePalette(hdc); + + /* note that we don't have to release the DC, since our window class + uses the WC_OWNDC flag! */ +#endif +} + + +/* display, root and visual - don't used at all */ +Colormap +XCreateColormap(Display* display, Window root, Visual* visual, int alloc) +{ + /* KLUDGE: this function needs XHDC to be set to the HDC currently + being operated on before it is invoked! */ + + HPAL palette; + int n; +#if POKA + PIXELFORMATDESCRIPTOR pfd; + LOGPALETTE *logical; + + /* grab the pixel format */ + memset(&pfd, 0, sizeof(PIXELFORMATDESCRIPTOR)); + DescribePixelFormat(XHDC, GetPixelFormat(XHDC), + sizeof(PIXELFORMATDESCRIPTOR), &pfd); + + if (!(pfd.dwFlags & PFD_NEED_PALETTE || + pfd.iPixelType == PFD_TYPE_COLORINDEX)) + { + return 0; + } + + n = 1 << pfd.cColorBits; + + /* allocate a bunch of memory for the logical palette (assume 256 + colors in a Win32 palette */ + logical = (LOGPALETTE*)malloc(sizeof(LOGPALETTE) + + sizeof(PALETTEENTRY) * n); + memset(logical, 0, sizeof(LOGPALETTE) + sizeof(PALETTEENTRY) * n); + + /* set the entries in the logical palette */ + logical->palVersion = 0x300; + logical->palNumEntries = n; + + /* start with a copy of the current system palette */ + GetSystemPaletteEntries(XHDC, 0, 256, &logical->palPalEntry[0]); + + if (pfd.iPixelType == PFD_TYPE_RGBA) { + int redMask = (1 << pfd.cRedBits) - 1; + int greenMask = (1 << pfd.cGreenBits) - 1; + int blueMask = (1 << pfd.cBlueBits) - 1; + int i; + + /* fill in an RGBA color palette */ + for (i = 0; i < n; ++i) { + logical->palPalEntry[i].peRed = + (((i >> pfd.cRedShift) & redMask) * 255) / redMask; + logical->palPalEntry[i].peGreen = + (((i >> pfd.cGreenShift) & greenMask) * 255) / greenMask; + logical->palPalEntry[i].peBlue = + (((i >> pfd.cBlueShift) & blueMask) * 255) / blueMask; + logical->palPalEntry[i].peFlags = 0; + } + } + + palette = CreatePalette(logical); + free(logical); + + SelectPalette(XHDC, palette, FALSE); + RealizePalette(XHDC); +#endif /* POKA */ + + return palette; +} + + + +int GetSystemMetrics( int mode) +{ RECTL rect; + + switch(mode) + { case SM_CXSCREEN: + WinQueryWindowRect(HWND_DESKTOP,&rect); + return (rect.xRight-rect.xLeft); + break; + case SM_CYSCREEN: + WinQueryWindowRect(HWND_DESKTOP,&rect); + return (rect.yTop-rect.yBottom); + break; + default: ; + } + return 0; +} +/* + * XParseGeometry parses strings of the form + * "=x{+-}{+-}", where + * width, height, xoffset, and yoffset are unsigned integers. + * Example: "=80x24+300-49" + * The equal sign is optional. + * It returns a bitmask that indicates which of the four values + * were actually found in the string. For each value found, + * the corresponding argument is updated; for each value + * not found, the corresponding argument is left unchanged. + */ + +static int +ReadInteger(char *string, char **NextString) +{ + register int Result = 0; + int Sign = 1; + + if (*string == '+') + string++; + else if (*string == '-') + { + string++; + Sign = -1; + } + for (; (*string >= '0') && (*string <= '9'); string++) + { + Result = (Result * 10) + (*string - '0'); + } + *NextString = string; + if (Sign >= 0) + return (Result); + else + return (-Result); +} + +int XParseGeometry(char *string, int *x, int *y, unsigned int *width, unsigned int *height) +{ + int mask = NoValue; + register char *strind; + unsigned int tempWidth, tempHeight; + int tempX, tempY; + char *nextCharacter; + + if ( (string == NULL) || (*string == '\0')) return(mask); + if (*string == '=') + string++; /* ignore possible '=' at beg of geometry spec */ + + strind = (char *)string; + if (*strind != '+' && *strind != '-' && *strind != 'x') { + tempWidth = ReadInteger(strind, &nextCharacter); + if (strind == nextCharacter) + return (0); + strind = nextCharacter; + mask |= WidthValue; + } + + if (*strind == 'x' || *strind == 'X') { + strind++; + tempHeight = ReadInteger(strind, &nextCharacter); + if (strind == nextCharacter) + return (0); + strind = nextCharacter; + mask |= HeightValue; + } + + if ((*strind == '+') || (*strind == '-')) { + if (*strind == '-') { + strind++; + tempX = -ReadInteger(strind, &nextCharacter); + if (strind == nextCharacter) + return (0); + strind = nextCharacter; + mask |= XNegative; + + } + else + { strind++; + tempX = ReadInteger(strind, &nextCharacter); + if (strind == nextCharacter) + return(0); + strind = nextCharacter; + } + mask |= XValue; + if ((*strind == '+') || (*strind == '-')) { + if (*strind == '-') { + strind++; + tempY = -ReadInteger(strind, &nextCharacter); + if (strind == nextCharacter) + return(0); + strind = nextCharacter; + mask |= YNegative; + + } + else + { + strind++; + tempY = ReadInteger(strind, &nextCharacter); + if (strind == nextCharacter) + return(0); + strind = nextCharacter; + } + mask |= YValue; + } + } + + /* If strind isn't at the end of the string the it's an invalid + geometry specification. */ + + if (*strind != '\0') return (0); + + if (mask & XValue) + *x = tempX; + if (mask & YValue) + *y = tempY; + if (mask & WidthValue) + *width = tempWidth; + if (mask & HeightValue) + *height = tempHeight; + return (mask); +} + +int gettimeofday(struct timeval* tp, void* tzp) +{ + DATETIME DateTime; + APIRET ulrc; /* Return Code. */ + + ulrc = DosGetDateTime(&DateTime); + tp->tv_sec = 60 * (60*DateTime.hours + DateTime.minutes) + DateTime.seconds; + tp->tv_usec = DateTime.hundredths * 10000; + return 0; +} + + +int +XPending(Display* display) +{ + /* similar functionality...I don't think that it is exact, but this + will have to do. */ + QMSG msg; + extern HAB hab; /* PM anchor block handle */ + +//?? WinPeekMsg(hab + return WinPeekMsg(hab, &msg, NULLHANDLE, 0, 0, PM_NOREMOVE); +} + +void +__glutAdjustCoords(Window parent, int* x, int* y, int* width, int* height) +{ + RECTL rect; + + /* adjust the window rectangle because Win32 thinks that the x, y, + width & height are the WHOLE window (including decorations), + whereas GLUT treats the x, y, width & height as only the CLIENT + area of the window. */ + rect.xLeft = *x; rect.yTop = *y; + rect.xRight = *x + *width; rect.yBottom = *y + *height; + + /* must adjust the coordinates according to the correct style + because depending on the style, there may or may not be + borders. */ +//?? AdjustWindowRect(&rect, WS_CLIPSIBLINGS | WS_CLIPCHILDREN | +//?? (parent ? WS_CHILD : WS_OVERLAPPEDWINDOW), +//?? FALSE); + /* FALSE in the third parameter = window has no menu bar */ + + /* readjust if the x and y are offscreen */ + if(rect.xLeft < 0) { + *x = 0; + } else { + *x = rect.xLeft; + } + + if(rect.yTop < 0) { + *y = 0; + } else { + *y = rect.yTop; + } + + *width = rect.xRight - rect.xLeft; /* adjusted width */ + *height = -(rect.yBottom - rect.yTop); /* adjusted height */ +} + + +int +__glutGetTransparentPixel(Display * dpy, XVisualInfo * vinfo) +{ + /* the transparent pixel on Win32 is always index number 0. So if + we put this routine in this file, we can avoid compiling the + whole of layerutil.c which is where this routine normally comes + from. */ + return 0; +} + +/* Translate point coordinates src_x and src_y from src to dst */ + +Bool +XTranslateCoordinates(Display *display, Window src, Window dst, + int src_x, int src_y, + int* dest_x_return, int* dest_y_return, + Window* child_return) +{ + SWP swp_src,swp_dst; + + WinQueryWindowPos(src,&swp_src); + WinQueryWindowPos(dst,&swp_dst); + + *dest_x_return = src_x + swp_src.x - swp_dst.x; + *dest_y_return = src_y + swp_src.y - swp_dst.y; + + /* just to make compilers happy...we don't use the return value. */ + return True; +} + +Status +XGetGeometry(Display* display, Window window, Window* root_return, + int* x_return, int* y_return, + unsigned int* width_return, unsigned int* height_return, + unsigned int *border_width_return, unsigned int* depth_return) +{ + /* KLUDGE: doesn't return the border_width or depth or root, x & y + are in screen coordinates. */ + SWP swp_src; + WinQueryWindowPos(window,&swp_src); + + *x_return = swp_src.x; + *y_return = swp_src.y; + *width_return = swp_src.cx; + *height_return = swp_src.cy; + + /* just to make compilers happy...we don't use the return value. */ + return 1; +} + +/* Get Display Width in millimeters */ +int +DisplayWidthMM(Display* display, int screen) +{ + int width; + LONG *pVC_Caps; + pVC_Caps = GetVideoConfig(NULLHANDLE); + width = (int)( 0.001 * pVC_Caps[CAPS_WIDTH] / pVC_Caps[CAPS_HORIZONTAL_RESOLUTION]);/* mm */ + return width; +} + +/* Get Display Height in millimeters */ +int +DisplayHeightMM(Display* display, int screen) +{ + int height; + LONG *pVC_Caps; + pVC_Caps = GetVideoConfig(NULLHANDLE); + height = (int)( 0.001 * pVC_Caps[CAPS_HEIGHT] / pVC_Caps[CAPS_VERTICAL_RESOLUTION]); /* mm */ + return height; +} + +void ScreenToClient( HWND hwnd, POINTL *point) +{ + SWP swp_src; + WinQueryWindowPos(hwnd,&swp_src); + point->x -= swp_src.x; + point->y -= swp_src.y; +} +  \ No newline at end of file diff --git a/src/glut/os2/glut_cindex.cpp b/src/glut/os2/glut_cindex.cpp index 0897a3cf85..3484185051 100644 --- a/src/glut/os2/glut_cindex.cpp +++ b/src/glut/os2/glut_cindex.cpp @@ -1,259 +1,259 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1996, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#include -#include "glutint.h" - -#if defined(__OS2PM__) - #define IsWindowVisible WinIsWindowVisible -#endif - -#define CLAMP(i) ((i) > 1.0 ? 1.0 : ((i) < 0.0 ? 0.0 : (i))) - -/* CENTRY */ -void GLUTAPIENTRY -glutSetColor(int ndx, GLfloat red, GLfloat green, GLfloat blue) -{ - GLUTcolormap *cmap, *newcmap; - XVisualInfo *vis; - XColor color; - int i; - - if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { - cmap = __glutCurrentWindow->colormap; - vis = __glutCurrentWindow->vis; - } else { - cmap = __glutCurrentWindow->overlay->colormap; - vis = __glutCurrentWindow->overlay->vis; - if (ndx == __glutCurrentWindow->overlay->transparentPixel) { - __glutWarning( - "glutSetColor: cannot set color of overlay transparent index %d\n", - ndx); - return; - } - } - - if (!cmap) { - __glutWarning("glutSetColor: current window is RGBA"); - return; - } -#if defined(_WIN32) || defined(__OS2PM__) - if (ndx >= 256 || /* always assume 256 colors on Win32 */ -#else - if (ndx >= vis->visual->map_entries || -#endif - ndx < 0) { - __glutWarning("glutSetColor: index %d out of range", ndx); - return; - } - if (cmap->refcnt > 1) { - newcmap = __glutAssociateNewColormap(vis); - cmap->refcnt--; - /* Wouldn't it be nice if XCopyColormapAndFree could be - told not to free the old colormap's entries! */ - for (i = cmap->size - 1; i >= 0; i--) { - if (i == ndx) { - /* We are going to set this cell shortly! */ - continue; - } - if (cmap->cells[i].component[GLUT_RED] >= 0.0) { - color.pixel = i; - newcmap->cells[i].component[GLUT_RED] = - cmap->cells[i].component[GLUT_RED]; - color.red = (GLfloat) 0xffff * - cmap->cells[i].component[GLUT_RED]; - newcmap->cells[i].component[GLUT_GREEN] = - cmap->cells[i].component[GLUT_GREEN]; - color.green = (GLfloat) 0xffff * - cmap->cells[i].component[GLUT_GREEN]; - newcmap->cells[i].component[GLUT_BLUE] = - cmap->cells[i].component[GLUT_BLUE]; - color.blue = (GLfloat) 0xffff * - cmap->cells[i].component[GLUT_BLUE]; - color.flags = DoRed | DoGreen | DoBlue; -#if defined(_WIN32) || defined(__OS2PM__) - if (IsWindowVisible(__glutCurrentWindow->win)) { - XHDC = __glutCurrentWindow->hdc; - } else { - XHDC = 0; - } -#endif - XStoreColor(__glutDisplay, newcmap->cmap, &color); - } else { - /* Leave unallocated entries unallocated. */ - } - } - cmap = newcmap; - if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { - __glutCurrentWindow->colormap = cmap; - __glutCurrentWindow->cmap = cmap->cmap; - } else { - __glutCurrentWindow->overlay->colormap = cmap; - __glutCurrentWindow->overlay->cmap = cmap->cmap; - } - XSetWindowColormap(__glutDisplay, - __glutCurrentWindow->renderWin, cmap->cmap); - -#if !defined(_WIN32) && !defined(__OS2PM__) - { - GLUTwindow *toplevel; - - toplevel = __glutToplevelOf(__glutCurrentWindow); - if (toplevel->cmap != cmap->cmap) { - __glutPutOnWorkList(toplevel, GLUT_COLORMAP_WORK); - } - } -#endif - } - color.pixel = ndx; - red = CLAMP(red); - cmap->cells[ndx].component[GLUT_RED] = red; - color.red = (GLfloat) 0xffff *red; - green = CLAMP(green); - cmap->cells[ndx].component[GLUT_GREEN] = green; - color.green = (GLfloat) 0xffff *green; - blue = CLAMP(blue); - cmap->cells[ndx].component[GLUT_BLUE] = blue; - color.blue = (GLfloat) 0xffff *blue; - color.flags = DoRed | DoGreen | DoBlue; -#if defined(_WIN32) || defined(__OS2PM__) - if (IsWindowVisible(__glutCurrentWindow->win)) { - XHDC = __glutCurrentWindow->hdc; - } else { - XHDC = 0; - } -#endif - XStoreColor(__glutDisplay, cmap->cmap, &color); -} - -GLfloat GLUTAPIENTRY -glutGetColor(int ndx, int comp) -{ - GLUTcolormap *colormap; - XVisualInfo *vis; - - if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { - colormap = __glutCurrentWindow->colormap; - vis = __glutCurrentWindow->vis; - } else { - colormap = __glutCurrentWindow->overlay->colormap; - vis = __glutCurrentWindow->overlay->vis; - if (ndx == __glutCurrentWindow->overlay->transparentPixel) { - __glutWarning("glutGetColor: requesting overlay transparent index %d\n", - ndx); - return -1.0; - } - } - - if (!colormap) { - __glutWarning("glutGetColor: current window is RGBA"); - return -1.0; - } -#if defined(_WIN32) || defined(__OS2PM__) -#define OUT_OF_RANGE_NDX(ndx) (ndx >= 256 || ndx < 0) -#else -#define OUT_OF_RANGE_NDX(ndx) (ndx >= vis->visual->map_entries || ndx < 0) -#endif - if (OUT_OF_RANGE_NDX(ndx)) { - __glutWarning("glutGetColor: index %d out of range", ndx); - return -1.0; - } - return colormap->cells[ndx].component[comp]; -} - -void GLUTAPIENTRY -glutCopyColormap(int winnum) -{ - GLUTwindow *window = __glutWindowList[winnum - 1]; - GLUTcolormap *oldcmap, *newcmap; - XVisualInfo *dstvis; - - if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { - oldcmap = __glutCurrentWindow->colormap; - dstvis = __glutCurrentWindow->vis; - newcmap = window->colormap; - } else { - oldcmap = __glutCurrentWindow->overlay->colormap; - dstvis = __glutCurrentWindow->overlay->vis; - if (!window->overlay) { - __glutWarning("glutCopyColormap: window %d has no overlay", winnum); - return; - } - newcmap = window->overlay->colormap; - } - - if (!oldcmap) { - __glutWarning("glutCopyColormap: destination colormap must be color index"); - return; - } - if (!newcmap) { - __glutWarning( - "glutCopyColormap: source colormap of window %d must be color index", - winnum); - return; - } - if (newcmap == oldcmap) { - /* Source and destination are the same; now copy needed. */ - return; - } -#if !defined(_WIN32) && !defined(__OS2PM__) - /* Play safe: compare visual IDs, not Visual*'s. */ - if (newcmap->visual->visualid == oldcmap->visual->visualid) { -#endif - /* Visuals match! "Copy" by reference... */ - __glutFreeColormap(oldcmap); - newcmap->refcnt++; - if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { - __glutCurrentWindow->colormap = newcmap; - __glutCurrentWindow->cmap = newcmap->cmap; - } else { - __glutCurrentWindow->overlay->colormap = newcmap; - __glutCurrentWindow->overlay->cmap = newcmap->cmap; - } - XSetWindowColormap(__glutDisplay, __glutCurrentWindow->renderWin, - newcmap->cmap); -#if !defined(_WIN32) && !defined(__OS2PM__) - __glutPutOnWorkList(__glutToplevelOf(window), GLUT_COLORMAP_WORK); -bla bla bla - - } else { - GLUTcolormap *copycmap; - XColor color; - int i, last; - - /* Visuals different - need a distinct X colormap! */ - copycmap = __glutAssociateNewColormap(dstvis); - /* Wouldn't it be nice if XCopyColormapAndFree could be - told not to free the old colormap's entries! */ - last = newcmap->size; - if (last > copycmap->size) { - last = copycmap->size; - } - for (i = last - 1; i >= 0; i--) { - if (newcmap->cells[i].component[GLUT_RED] >= 0.0) { - color.pixel = i; - copycmap->cells[i].component[GLUT_RED] = - newcmap->cells[i].component[GLUT_RED]; - color.red = (GLfloat) 0xffff * - newcmap->cells[i].component[GLUT_RED]; - copycmap->cells[i].component[GLUT_GREEN] = - newcmap->cells[i].component[GLUT_GREEN]; - color.green = (GLfloat) 0xffff * - newcmap->cells[i].component[GLUT_GREEN]; - copycmap->cells[i].component[GLUT_BLUE] = - newcmap->cells[i].component[GLUT_BLUE]; - color.blue = (GLfloat) 0xffff * - newcmap->cells[i].component[GLUT_BLUE]; - color.flags = DoRed | DoGreen | DoBlue; - XStoreColor(__glutDisplay, copycmap->cmap, &color); - } - } - } -#endif -} -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994, 1996, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#include +#include "glutint.h" + +#if defined(__OS2PM__) + #define IsWindowVisible WinIsWindowVisible +#endif + +#define CLAMP(i) ((i) > 1.0 ? 1.0 : ((i) < 0.0 ? 0.0 : (i))) + +/* CENTRY */ +void GLUTAPIENTRY +glutSetColor(int ndx, GLfloat red, GLfloat green, GLfloat blue) +{ + GLUTcolormap *cmap, *newcmap; + XVisualInfo *vis; + XColor color; + int i; + + if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { + cmap = __glutCurrentWindow->colormap; + vis = __glutCurrentWindow->vis; + } else { + cmap = __glutCurrentWindow->overlay->colormap; + vis = __glutCurrentWindow->overlay->vis; + if (ndx == __glutCurrentWindow->overlay->transparentPixel) { + __glutWarning( + "glutSetColor: cannot set color of overlay transparent index %d\n", + ndx); + return; + } + } + + if (!cmap) { + __glutWarning("glutSetColor: current window is RGBA"); + return; + } +#if defined(_WIN32) || defined(__OS2PM__) + if (ndx >= 256 || /* always assume 256 colors on Win32 */ +#else + if (ndx >= vis->visual->map_entries || +#endif + ndx < 0) { + __glutWarning("glutSetColor: index %d out of range", ndx); + return; + } + if (cmap->refcnt > 1) { + newcmap = __glutAssociateNewColormap(vis); + cmap->refcnt--; + /* Wouldn't it be nice if XCopyColormapAndFree could be + told not to free the old colormap's entries! */ + for (i = cmap->size - 1; i >= 0; i--) { + if (i == ndx) { + /* We are going to set this cell shortly! */ + continue; + } + if (cmap->cells[i].component[GLUT_RED] >= 0.0) { + color.pixel = i; + newcmap->cells[i].component[GLUT_RED] = + cmap->cells[i].component[GLUT_RED]; + color.red = (GLfloat) 0xffff * + cmap->cells[i].component[GLUT_RED]; + newcmap->cells[i].component[GLUT_GREEN] = + cmap->cells[i].component[GLUT_GREEN]; + color.green = (GLfloat) 0xffff * + cmap->cells[i].component[GLUT_GREEN]; + newcmap->cells[i].component[GLUT_BLUE] = + cmap->cells[i].component[GLUT_BLUE]; + color.blue = (GLfloat) 0xffff * + cmap->cells[i].component[GLUT_BLUE]; + color.flags = DoRed | DoGreen | DoBlue; +#if defined(_WIN32) || defined(__OS2PM__) + if (IsWindowVisible(__glutCurrentWindow->win)) { + XHDC = __glutCurrentWindow->hdc; + } else { + XHDC = 0; + } +#endif + XStoreColor(__glutDisplay, newcmap->cmap, &color); + } else { + /* Leave unallocated entries unallocated. */ + } + } + cmap = newcmap; + if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { + __glutCurrentWindow->colormap = cmap; + __glutCurrentWindow->cmap = cmap->cmap; + } else { + __glutCurrentWindow->overlay->colormap = cmap; + __glutCurrentWindow->overlay->cmap = cmap->cmap; + } + XSetWindowColormap(__glutDisplay, + __glutCurrentWindow->renderWin, cmap->cmap); + +#if !defined(_WIN32) && !defined(__OS2PM__) + { + GLUTwindow *toplevel; + + toplevel = __glutToplevelOf(__glutCurrentWindow); + if (toplevel->cmap != cmap->cmap) { + __glutPutOnWorkList(toplevel, GLUT_COLORMAP_WORK); + } + } +#endif + } + color.pixel = ndx; + red = CLAMP(red); + cmap->cells[ndx].component[GLUT_RED] = red; + color.red = (GLfloat) 0xffff *red; + green = CLAMP(green); + cmap->cells[ndx].component[GLUT_GREEN] = green; + color.green = (GLfloat) 0xffff *green; + blue = CLAMP(blue); + cmap->cells[ndx].component[GLUT_BLUE] = blue; + color.blue = (GLfloat) 0xffff *blue; + color.flags = DoRed | DoGreen | DoBlue; +#if defined(_WIN32) || defined(__OS2PM__) + if (IsWindowVisible(__glutCurrentWindow->win)) { + XHDC = __glutCurrentWindow->hdc; + } else { + XHDC = 0; + } +#endif + XStoreColor(__glutDisplay, cmap->cmap, &color); +} + +GLfloat GLUTAPIENTRY +glutGetColor(int ndx, int comp) +{ + GLUTcolormap *colormap; + XVisualInfo *vis; + + if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { + colormap = __glutCurrentWindow->colormap; + vis = __glutCurrentWindow->vis; + } else { + colormap = __glutCurrentWindow->overlay->colormap; + vis = __glutCurrentWindow->overlay->vis; + if (ndx == __glutCurrentWindow->overlay->transparentPixel) { + __glutWarning("glutGetColor: requesting overlay transparent index %d\n", + ndx); + return -1.0; + } + } + + if (!colormap) { + __glutWarning("glutGetColor: current window is RGBA"); + return -1.0; + } +#if defined(_WIN32) || defined(__OS2PM__) +#define OUT_OF_RANGE_NDX(ndx) (ndx >= 256 || ndx < 0) +#else +#define OUT_OF_RANGE_NDX(ndx) (ndx >= vis->visual->map_entries || ndx < 0) +#endif + if (OUT_OF_RANGE_NDX(ndx)) { + __glutWarning("glutGetColor: index %d out of range", ndx); + return -1.0; + } + return colormap->cells[ndx].component[comp]; +} + +void GLUTAPIENTRY +glutCopyColormap(int winnum) +{ + GLUTwindow *window = __glutWindowList[winnum - 1]; + GLUTcolormap *oldcmap, *newcmap; + XVisualInfo *dstvis; + + if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { + oldcmap = __glutCurrentWindow->colormap; + dstvis = __glutCurrentWindow->vis; + newcmap = window->colormap; + } else { + oldcmap = __glutCurrentWindow->overlay->colormap; + dstvis = __glutCurrentWindow->overlay->vis; + if (!window->overlay) { + __glutWarning("glutCopyColormap: window %d has no overlay", winnum); + return; + } + newcmap = window->overlay->colormap; + } + + if (!oldcmap) { + __glutWarning("glutCopyColormap: destination colormap must be color index"); + return; + } + if (!newcmap) { + __glutWarning( + "glutCopyColormap: source colormap of window %d must be color index", + winnum); + return; + } + if (newcmap == oldcmap) { + /* Source and destination are the same; now copy needed. */ + return; + } +#if !defined(_WIN32) && !defined(__OS2PM__) + /* Play safe: compare visual IDs, not Visual*'s. */ + if (newcmap->visual->visualid == oldcmap->visual->visualid) { +#endif + /* Visuals match! "Copy" by reference... */ + __glutFreeColormap(oldcmap); + newcmap->refcnt++; + if (__glutCurrentWindow->renderWin == __glutCurrentWindow->win) { + __glutCurrentWindow->colormap = newcmap; + __glutCurrentWindow->cmap = newcmap->cmap; + } else { + __glutCurrentWindow->overlay->colormap = newcmap; + __glutCurrentWindow->overlay->cmap = newcmap->cmap; + } + XSetWindowColormap(__glutDisplay, __glutCurrentWindow->renderWin, + newcmap->cmap); +#if !defined(_WIN32) && !defined(__OS2PM__) + __glutPutOnWorkList(__glutToplevelOf(window), GLUT_COLORMAP_WORK); +bla bla bla + + } else { + GLUTcolormap *copycmap; + XColor color; + int i, last; + + /* Visuals different - need a distinct X colormap! */ + copycmap = __glutAssociateNewColormap(dstvis); + /* Wouldn't it be nice if XCopyColormapAndFree could be + told not to free the old colormap's entries! */ + last = newcmap->size; + if (last > copycmap->size) { + last = copycmap->size; + } + for (i = last - 1; i >= 0; i--) { + if (newcmap->cells[i].component[GLUT_RED] >= 0.0) { + color.pixel = i; + copycmap->cells[i].component[GLUT_RED] = + newcmap->cells[i].component[GLUT_RED]; + color.red = (GLfloat) 0xffff * + newcmap->cells[i].component[GLUT_RED]; + copycmap->cells[i].component[GLUT_GREEN] = + newcmap->cells[i].component[GLUT_GREEN]; + color.green = (GLfloat) 0xffff * + newcmap->cells[i].component[GLUT_GREEN]; + copycmap->cells[i].component[GLUT_BLUE] = + newcmap->cells[i].component[GLUT_BLUE]; + color.blue = (GLfloat) 0xffff * + newcmap->cells[i].component[GLUT_BLUE]; + color.flags = DoRed | DoGreen | DoBlue; + XStoreColor(__glutDisplay, copycmap->cmap, &color); + } + } + } +#endif +} +/* ENDCENTRY */  \ No newline at end of file diff --git a/src/glut/os2/glut_gamemode.cpp b/src/glut/os2/glut_gamemode.cpp index 50185d7b9d..39918fdf39 100644 --- a/src/glut/os2/glut_gamemode.cpp +++ b/src/glut/os2/glut_gamemode.cpp @@ -1,680 +1,680 @@ - -/* Copyright (c) Mark J. Kilgard, 1998. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - - -#include -#include -#include -#include - -#include "glutint.h" - -#if !defined(_WIN32) && !defined(__OS2__) -#include -#include - -/* SGI optimization introduced in IRIX 6.3 to avoid X server - round trips for interning common X atoms. */ -#if defined(_SGI_EXTRA_PREDEFINES) && !defined(NO_FAST_ATOMS) -#include -#else -#define XSGIFastInternAtom(dpy,string,fast_name,how) XInternAtom(dpy,string,how) -#endif -#endif /* not _WIN32 */ - -int __glutDisplaySettingsChanged = 0; -static DisplayMode *dmodes, *currentDm = NULL; -static int ndmodes = -1; -GLUTwindow *__glutGameModeWindow = NULL; - -#ifdef TEST -static char *compstr[] = -{ - "none", "=", "!=", "<=", ">=", ">", "<", "~" -}; -static char *capstr[] = -{ - "width", "height", "bpp", "hertz", "num" -}; -#endif - -#if defined(__OS2__) -void -#else -void __cdecl -#endif -__glutCloseDownGameMode(void) -{ - if (__glutDisplaySettingsChanged) { -#ifdef _WIN32 - /* Assumes that display settings have been changed, that - is __glutDisplaySettingsChanged is true. */ - ChangeDisplaySettings(NULL, 0); -#endif - __glutDisplaySettingsChanged = 0; - } - __glutGameModeWindow = NULL; -} - -void GLUTAPIENTRY -glutLeaveGameMode(void) -{ - if (__glutGameModeWindow == NULL) { - __glutWarning("not in game mode so cannot leave game mode"); - return; - } - __glutDestroyWindow(__glutGameModeWindow, - __glutGameModeWindow); - XFlush(__glutDisplay); - __glutGameModeWindow = NULL; -} - -#ifdef _WIN32 - -/* Same values as from MSDN's SetDisp.c example. */ -#define MIN_WIDTH 400 -#define MIN_FREQUENCY 60 - -static void -initGameModeSupport(void) -{ - DEVMODE dm; - DWORD mode; - int i; - - if (ndmodes >= 0) { - /* ndmodes is initially -1 to indicate no - dmodes allocated yet. */ - return; - } - - /* Determine how many display modes there are. */ - ndmodes = 0; - mode = 0; - while (EnumDisplaySettings(NULL, mode, &dm)) { - if (dm.dmPelsWidth >= MIN_WIDTH && - (dm.dmDisplayFrequency == 0 || - dm.dmDisplayFrequency >= MIN_FREQUENCY)) { - ndmodes++; - } - mode++; - } - - /* Allocate memory for a list of all the display modes. */ - dmodes = (DisplayMode*) - malloc(ndmodes * sizeof(DisplayMode)); - - /* Now that we know how many display modes to expect, - enumerate them again and save the information in - the list we allocated above. */ - i = 0; - mode = 0; - while (EnumDisplaySettings(NULL, mode, &dm)) { - /* Try to reject any display settings that seem unplausible. */ - if (dm.dmPelsWidth >= MIN_WIDTH && - (dm.dmDisplayFrequency == 0 || - dm.dmDisplayFrequency >= MIN_FREQUENCY)) { - dmodes[i].devmode = dm; - dmodes[i].valid = 1; /* XXX Not used for now. */ - dmodes[i].cap[DM_WIDTH] = dm.dmPelsWidth; - dmodes[i].cap[DM_HEIGHT] = dm.dmPelsHeight; - dmodes[i].cap[DM_PIXEL_DEPTH] = dm.dmBitsPerPel; - if (dm.dmDisplayFrequency == 0) { - /* Guess a reasonable guess. */ - /* Lame Windows 95 version of EnumDisplaySettings. */ - dmodes[i].cap[DM_HERTZ] = 60; - } else { - dmodes[i].cap[DM_HERTZ] = dm.dmDisplayFrequency; - } - i++; - } - mode++; - } - - assert(i == ndmodes); -} - -#else - -/* X Windows version of initGameModeSupport. */ -static void -initGameModeSupport(void) -{ - if (ndmodes >= 0) { - /* ndmodes is initially -1 to indicate no - dmodes allocated yet. */ - return; - } - - /* Determine how many display modes there are. */ - ndmodes = 0; -} - -#endif - -/* This routine is based on similiar code in glut_dstr.c */ -static DisplayMode * -findMatch(DisplayMode * dmodes, int ndmodes, - Criterion * criteria, int ncriteria) -{ - DisplayMode *found; - int *bestScore, *thisScore; - int i, j, numok, result = 0, worse, better; - - found = NULL; - numok = 1; /* "num" capability is indexed from 1, - not 0. */ - - /* XXX alloca canidate. */ - bestScore = (int *) malloc(ncriteria * sizeof(int)); - if (!bestScore) { - __glutFatalError("out of memory."); - } - for (j = 0; j < ncriteria; j++) { - /* Very negative number. */ - bestScore[j] = -32768; - } - - /* XXX alloca canidate. */ - thisScore = (int *) malloc(ncriteria * sizeof(int)); - if (!thisScore) { - __glutFatalError("out of memory."); - } - - for (i = 0; i < ndmodes; i++) { - if (dmodes[i].valid) { - worse = 0; - better = 0; - - for (j = 0; j < ncriteria; j++) { - int cap, cvalue, dvalue; - - cap = criteria[j].capability; - cvalue = criteria[j].value; - if (cap == NUM) { - dvalue = numok; - } else { - dvalue = dmodes[i].cap[cap]; - } -#ifdef TEST - if (verbose) - printf(" %s %s %d to %d\n", - capstr[cap], compstr[criteria[j].comparison], cvalue, dvalue); -#endif - switch (criteria[j].comparison) { - case EQ: - result = cvalue == dvalue; - thisScore[j] = 1; - break; - case NEQ: - result = cvalue != dvalue; - thisScore[j] = 1; - break; - case LT: - result = dvalue < cvalue; - thisScore[j] = dvalue - cvalue; - break; - case GT: - result = dvalue > cvalue; - thisScore[j] = dvalue - cvalue; - break; - case LTE: - result = dvalue <= cvalue; - thisScore[j] = dvalue - cvalue; - break; - case GTE: - result = (dvalue >= cvalue); - thisScore[j] = dvalue - cvalue; - break; - case MIN: - result = dvalue >= cvalue; - thisScore[j] = cvalue - dvalue; - break; - } - -#ifdef TEST - if (verbose) - printf(" result=%d score=%d bestScore=%d\n", result, thisScore[j], bestScore[j]); -#endif - - if (result) { - if (better || thisScore[j] > bestScore[j]) { - better = 1; - } else if (thisScore[j] == bestScore[j]) { - /* Keep looking. */ - } else { - goto nextDM; - } - } else { - if (cap == NUM) { - worse = 1; - } else { - goto nextDM; - } - } - - } - - if (better && !worse) { - found = &dmodes[i]; - for (j = 0; j < ncriteria; j++) { - bestScore[j] = thisScore[j]; - } - } - numok++; - - nextDM:; - - } - } - free(bestScore); - free(thisScore); - return found; -} - -/** - * Parses strings in the form of: - * 800x600 - * 800x600:16 - * 800x600@60 - * 800x600:16@60 - * @60 - * :16 - * :16@60 - * NOTE that @ before : is not parsed. - */ -static int -specialCaseParse(char *word, Criterion * criterion, int mask) -{ - char *xstr, *response; - int got; - int width, height, bpp, hertz; - - switch(word[0]) { - case '0': - case '1': - case '2': - case '3': - case '4': - case '5': - case '6': - case '7': - case '8': - case '9': - /* The WWWxHHH case. */ - if (mask & (1 << DM_WIDTH)) { - return -1; - } - xstr = strpbrk(&word[1], "x"); - if (xstr) { - width = (int) strtol(word, &response, 0); - if (response == word || response[0] != 'x') { - /* Not a valid number OR needs to be followed by 'x'. */ - return -1; - } - height = (int) strtol(&xstr[1], &response, 0); - if (response == &xstr[1]) { - /* Not a valid number. */ - return -1; - } - criterion[0].capability = DM_WIDTH; - criterion[0].comparison = EQ; - criterion[0].value = width; - criterion[1].capability = DM_HEIGHT; - criterion[1].comparison = EQ; - criterion[1].value = height; - got = specialCaseParse(response, - &criterion[2], 1 << DM_WIDTH); - if (got >= 0) { - return got + 2; - } else { - return -1; - } - } - return -1; - case ':': - /* The :BPP case. */ - if (mask & (1 << DM_PIXEL_DEPTH)) { - return -1; - } - bpp = (int) strtol(&word[1], &response, 0); - if (response == &word[1]) { - /* Not a valid number. */ - return -1; - } - criterion[0].capability = DM_PIXEL_DEPTH; - criterion[0].comparison = EQ; - criterion[0].value = bpp; - got = specialCaseParse(response, - &criterion[1], (1 << DM_WIDTH) | (1 << DM_PIXEL_DEPTH)); - if (got >= 0) { - return got + 1; - } else { - return -1; - } - case '@': - /* The @HZ case. */ - if (mask & (1 << DM_HERTZ)) { - return -1; - } - hertz = (int) strtol(&word[1], &response, 0); - if (response == &word[1]) { - /* Not a valid number. */ - return -1; - } - criterion[0].capability = DM_HERTZ; - criterion[0].comparison = EQ; - criterion[0].value = hertz; - got = specialCaseParse(response, - &criterion[1], ~DM_HERTZ); - if (got >= 0) { - return got + 1; - } else { - return -1; - } - case '\0': - return 0; - } - return -1; -} - -/* This routine is based on similiar code in glut_dstr.c */ -static int -parseCriteria(char *word, Criterion * criterion) -{ - char *cstr, *vstr, *response; - int comparator, value = 0; - - cstr = strpbrk(word, "=>': - if (cstr[1] == '=') { - comparator = GTE; - vstr = &cstr[2]; - } else { - comparator = GT; - vstr = &cstr[1]; - } - break; - case '<': - if (cstr[1] == '=') { - comparator = LTE; - vstr = &cstr[2]; - } else { - comparator = LT; - vstr = &cstr[1]; - } - break; - case '!': - if (cstr[1] == '=') { - comparator = NEQ; - vstr = &cstr[2]; - } else { - return -1; - } - break; - default: - return -1; - } - value = (int) strtol(vstr, &response, 0); - if (response == vstr) { - /* Not a valid number. */ - return -1; - } - *cstr = '\0'; - } else { - comparator = NONE; - } - switch (word[0]) { - case 'b': - if (!strcmp(word, "bpp")) { - criterion[0].capability = DM_PIXEL_DEPTH; - if (comparator == NONE) { - return -1; - } else { - criterion[0].comparison = comparator; - criterion[0].value = value; - return 1; - } - } - return -1; - case 'h': - if (!strcmp(word, "height")) { - criterion[0].capability = DM_HEIGHT; - if (comparator == NONE) { - return -1; - } else { - criterion[0].comparison = comparator; - criterion[0].value = value; - return 1; - } - } - if (!strcmp(word, "hertz")) { - criterion[0].capability = DM_HERTZ; - if (comparator == NONE) { - return -1; - } else { - criterion[0].comparison = comparator; - criterion[0].value = value; - return 1; - } - } - return -1; - case 'n': - if (!strcmp(word, "num")) { - criterion[0].capability = DM_NUM; - if (comparator == NONE) { - return -1; - } else { - criterion[0].comparison = comparator; - criterion[0].value = value; - return 1; - } - } - return -1; - case 'w': - if (!strcmp(word, "width")) { - criterion[0].capability = DM_WIDTH; - if (comparator == NONE) { - return -1; - } else { - criterion[0].comparison = comparator; - criterion[0].value = value; - return 1; - } - } - return -1; - } - if (comparator == NONE) { - return specialCaseParse(word, criterion, 0); - } - return -1; -} - -/* This routine is based on similiar code in glut_dstr.c */ -static Criterion * -parseDisplayString(const char *display, int *ncriteria) -{ - Criterion *criteria = NULL; - int n, parsed; - char *copy, *word; - - copy = __glutStrdup(display); - /* Attempt to estimate how many criteria entries should be - needed. */ - n = 0; - word = strtok(copy, " \t"); - while (word) { - n++; - word = strtok(NULL, " \t"); - } - /* Allocate number of words of criteria. A word - could contain as many as four criteria in the - worst case. Example: 800x600:16@60 */ - criteria = (Criterion *) malloc(4 * n * sizeof(Criterion)); - if (!criteria) { - __glutFatalError("out of memory."); - } - - /* Re-copy the copy of the display string. */ - strcpy(copy, display); - - n = 0; - word = strtok(copy, " \t"); - while (word) { - parsed = parseCriteria(word, &criteria[n]); - if (parsed >= 0) { - n += parsed; - } else { - __glutWarning("Unrecognized game mode string word: %s (ignoring)\n", word); - } - word = strtok(NULL, " \t"); - } - - free(copy); - *ncriteria = n; - return criteria; -} - -void GLUTAPIENTRY -glutGameModeString(const char *string) -{ - Criterion *criteria; - int ncriteria; - - initGameModeSupport(); - criteria = parseDisplayString(string, &ncriteria); - currentDm = findMatch(dmodes, ndmodes, criteria, ncriteria); - free(criteria); -} - -int GLUTAPIENTRY -glutEnterGameMode(void) -{ - GLUTwindow *window; - int width, height; - Window win; - - if (__glutMappedMenu) { - __glutFatalUsage("entering game mode not allowed while menus in use"); - } - if (__glutGameModeWindow) { - /* Already in game mode, so blow away game mode - window so apps can change resolutions. */ - window = __glutGameModeWindow; - /* Setting the game mode window to NULL tricks - the window destroy code into not undoing the - screen display change since we plan on immediately - doing another mode change. */ - __glutGameModeWindow = NULL; - __glutDestroyWindow(window, window); - } - - /* Assume default screen size until we find out if we - can actually change the display settings. */ - width = __glutScreenWidth; - height = __glutScreenHeight; - - if (currentDm) { -#ifdef _WIN32 - LONG status; - static int registered = 0; - - status = ChangeDisplaySettings(¤tDm->devmode, - CDS_FULLSCREEN); - if (status == DISP_CHANGE_SUCCESSFUL) { - __glutDisplaySettingsChanged = 1; - width = currentDm->cap[DM_WIDTH]; - height = currentDm->cap[DM_HEIGHT]; - if (!registered) { - atexit(__glutCloseDownGameMode); - registered = 1; - } - } else { - /* Switch back to default resolution. */ - ChangeDisplaySettings(NULL, 0); - } -#endif - } - - window = __glutCreateWindow(NULL, 0, 0, - width, height, /* game mode */ 1); - win = window->win; - -#if !defined(_WIN32) && !defined(__OS2__) - if (__glutMotifHints == None) { - __glutMotifHints = XSGIFastInternAtom(__glutDisplay, "_MOTIF_WM_HINTS", - SGI_XA__MOTIF_WM_HINTS, 0); - if (__glutMotifHints == None) { - __glutWarning("Could not intern X atom for _MOTIF_WM_HINTS."); - } - } - - /* Game mode window is a toplevel window. */ - XSetWMProtocols(__glutDisplay, win, &__glutWMDeleteWindow, 1); -#endif - - /* Schedule the fullscreen property to be added and to - make sure the window is configured right. Win32 - doesn't need this. */ - window->desiredX = 0; - window->desiredY = 0; - window->desiredWidth = width; - window->desiredHeight = height; - window->desiredConfMask |= CWX | CWY | CWWidth | CWHeight; -#ifdef _WIN32 - /* Win32 does not want to use GLUT_FULL_SCREEN_WORK - for game mode because we need to be maximizing - the window in game mode, not just sizing it to - take up the full screen. The Win32-ness of game - mode happens when you pass 1 in the gameMode parameter - to __glutCreateWindow above. A gameMode of creates - a WS_POPUP window, not a standard WS_OVERLAPPEDWINDOW - window. WS_POPUP ensures the taskbar is hidden. */ - __glutPutOnWorkList(window, - GLUT_CONFIGURE_WORK); -#else - __glutPutOnWorkList(window, - GLUT_CONFIGURE_WORK | GLUT_FULL_SCREEN_WORK); -#endif - - __glutGameModeWindow = window; - return window->num + 1; -} - -int GLUTAPIENTRY -glutGameModeGet(GLenum mode) -{ - switch (mode) { - case GLUT_GAME_MODE_ACTIVE: - return __glutGameModeWindow != NULL; - case GLUT_GAME_MODE_POSSIBLE: - return currentDm != NULL; - case GLUT_GAME_MODE_WIDTH: - return currentDm ? currentDm->cap[DM_WIDTH] : -1; - case GLUT_GAME_MODE_HEIGHT: - return currentDm ? currentDm->cap[DM_HEIGHT] : -1; - case GLUT_GAME_MODE_PIXEL_DEPTH: - return currentDm ? currentDm->cap[DM_PIXEL_DEPTH] : -1; - case GLUT_GAME_MODE_REFRESH_RATE: - return currentDm ? currentDm->cap[DM_HERTZ] : -1; - case GLUT_GAME_MODE_DISPLAY_CHANGED: - return __glutDisplaySettingsChanged; - default: - return -1; - } -} + +/* Copyright (c) Mark J. Kilgard, 1998. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + + +#include +#include +#include +#include + +#include "glutint.h" + +#if !defined(_WIN32) && !defined(__OS2__) +#include +#include + +/* SGI optimization introduced in IRIX 6.3 to avoid X server + round trips for interning common X atoms. */ +#if defined(_SGI_EXTRA_PREDEFINES) && !defined(NO_FAST_ATOMS) +#include +#else +#define XSGIFastInternAtom(dpy,string,fast_name,how) XInternAtom(dpy,string,how) +#endif +#endif /* not _WIN32 */ + +int __glutDisplaySettingsChanged = 0; +static DisplayMode *dmodes, *currentDm = NULL; +static int ndmodes = -1; +GLUTwindow *__glutGameModeWindow = NULL; + +#ifdef TEST +static char *compstr[] = +{ + "none", "=", "!=", "<=", ">=", ">", "<", "~" +}; +static char *capstr[] = +{ + "width", "height", "bpp", "hertz", "num" +}; +#endif + +#if defined(__OS2__) +void +#else +void __cdecl +#endif +__glutCloseDownGameMode(void) +{ + if (__glutDisplaySettingsChanged) { +#ifdef _WIN32 + /* Assumes that display settings have been changed, that + is __glutDisplaySettingsChanged is true. */ + ChangeDisplaySettings(NULL, 0); +#endif + __glutDisplaySettingsChanged = 0; + } + __glutGameModeWindow = NULL; +} + +void GLUTAPIENTRY +glutLeaveGameMode(void) +{ + if (__glutGameModeWindow == NULL) { + __glutWarning("not in game mode so cannot leave game mode"); + return; + } + __glutDestroyWindow(__glutGameModeWindow, + __glutGameModeWindow); + XFlush(__glutDisplay); + __glutGameModeWindow = NULL; +} + +#ifdef _WIN32 + +/* Same values as from MSDN's SetDisp.c example. */ +#define MIN_WIDTH 400 +#define MIN_FREQUENCY 60 + +static void +initGameModeSupport(void) +{ + DEVMODE dm; + DWORD mode; + int i; + + if (ndmodes >= 0) { + /* ndmodes is initially -1 to indicate no + dmodes allocated yet. */ + return; + } + + /* Determine how many display modes there are. */ + ndmodes = 0; + mode = 0; + while (EnumDisplaySettings(NULL, mode, &dm)) { + if (dm.dmPelsWidth >= MIN_WIDTH && + (dm.dmDisplayFrequency == 0 || + dm.dmDisplayFrequency >= MIN_FREQUENCY)) { + ndmodes++; + } + mode++; + } + + /* Allocate memory for a list of all the display modes. */ + dmodes = (DisplayMode*) + malloc(ndmodes * sizeof(DisplayMode)); + + /* Now that we know how many display modes to expect, + enumerate them again and save the information in + the list we allocated above. */ + i = 0; + mode = 0; + while (EnumDisplaySettings(NULL, mode, &dm)) { + /* Try to reject any display settings that seem unplausible. */ + if (dm.dmPelsWidth >= MIN_WIDTH && + (dm.dmDisplayFrequency == 0 || + dm.dmDisplayFrequency >= MIN_FREQUENCY)) { + dmodes[i].devmode = dm; + dmodes[i].valid = 1; /* XXX Not used for now. */ + dmodes[i].cap[DM_WIDTH] = dm.dmPelsWidth; + dmodes[i].cap[DM_HEIGHT] = dm.dmPelsHeight; + dmodes[i].cap[DM_PIXEL_DEPTH] = dm.dmBitsPerPel; + if (dm.dmDisplayFrequency == 0) { + /* Guess a reasonable guess. */ + /* Lame Windows 95 version of EnumDisplaySettings. */ + dmodes[i].cap[DM_HERTZ] = 60; + } else { + dmodes[i].cap[DM_HERTZ] = dm.dmDisplayFrequency; + } + i++; + } + mode++; + } + + assert(i == ndmodes); +} + +#else + +/* X Windows version of initGameModeSupport. */ +static void +initGameModeSupport(void) +{ + if (ndmodes >= 0) { + /* ndmodes is initially -1 to indicate no + dmodes allocated yet. */ + return; + } + + /* Determine how many display modes there are. */ + ndmodes = 0; +} + +#endif + +/* This routine is based on similiar code in glut_dstr.c */ +static DisplayMode * +findMatch(DisplayMode * dmodes, int ndmodes, + Criterion * criteria, int ncriteria) +{ + DisplayMode *found; + int *bestScore, *thisScore; + int i, j, numok, result = 0, worse, better; + + found = NULL; + numok = 1; /* "num" capability is indexed from 1, + not 0. */ + + /* XXX alloca canidate. */ + bestScore = (int *) malloc(ncriteria * sizeof(int)); + if (!bestScore) { + __glutFatalError("out of memory."); + } + for (j = 0; j < ncriteria; j++) { + /* Very negative number. */ + bestScore[j] = -32768; + } + + /* XXX alloca canidate. */ + thisScore = (int *) malloc(ncriteria * sizeof(int)); + if (!thisScore) { + __glutFatalError("out of memory."); + } + + for (i = 0; i < ndmodes; i++) { + if (dmodes[i].valid) { + worse = 0; + better = 0; + + for (j = 0; j < ncriteria; j++) { + int cap, cvalue, dvalue; + + cap = criteria[j].capability; + cvalue = criteria[j].value; + if (cap == NUM) { + dvalue = numok; + } else { + dvalue = dmodes[i].cap[cap]; + } +#ifdef TEST + if (verbose) + printf(" %s %s %d to %d\n", + capstr[cap], compstr[criteria[j].comparison], cvalue, dvalue); +#endif + switch (criteria[j].comparison) { + case EQ: + result = cvalue == dvalue; + thisScore[j] = 1; + break; + case NEQ: + result = cvalue != dvalue; + thisScore[j] = 1; + break; + case LT: + result = dvalue < cvalue; + thisScore[j] = dvalue - cvalue; + break; + case GT: + result = dvalue > cvalue; + thisScore[j] = dvalue - cvalue; + break; + case LTE: + result = dvalue <= cvalue; + thisScore[j] = dvalue - cvalue; + break; + case GTE: + result = (dvalue >= cvalue); + thisScore[j] = dvalue - cvalue; + break; + case MIN: + result = dvalue >= cvalue; + thisScore[j] = cvalue - dvalue; + break; + } + +#ifdef TEST + if (verbose) + printf(" result=%d score=%d bestScore=%d\n", result, thisScore[j], bestScore[j]); +#endif + + if (result) { + if (better || thisScore[j] > bestScore[j]) { + better = 1; + } else if (thisScore[j] == bestScore[j]) { + /* Keep looking. */ + } else { + goto nextDM; + } + } else { + if (cap == NUM) { + worse = 1; + } else { + goto nextDM; + } + } + + } + + if (better && !worse) { + found = &dmodes[i]; + for (j = 0; j < ncriteria; j++) { + bestScore[j] = thisScore[j]; + } + } + numok++; + + nextDM:; + + } + } + free(bestScore); + free(thisScore); + return found; +} + +/** + * Parses strings in the form of: + * 800x600 + * 800x600:16 + * 800x600@60 + * 800x600:16@60 + * @60 + * :16 + * :16@60 + * NOTE that @ before : is not parsed. + */ +static int +specialCaseParse(char *word, Criterion * criterion, int mask) +{ + char *xstr, *response; + int got; + int width, height, bpp, hertz; + + switch(word[0]) { + case '0': + case '1': + case '2': + case '3': + case '4': + case '5': + case '6': + case '7': + case '8': + case '9': + /* The WWWxHHH case. */ + if (mask & (1 << DM_WIDTH)) { + return -1; + } + xstr = strpbrk(&word[1], "x"); + if (xstr) { + width = (int) strtol(word, &response, 0); + if (response == word || response[0] != 'x') { + /* Not a valid number OR needs to be followed by 'x'. */ + return -1; + } + height = (int) strtol(&xstr[1], &response, 0); + if (response == &xstr[1]) { + /* Not a valid number. */ + return -1; + } + criterion[0].capability = DM_WIDTH; + criterion[0].comparison = EQ; + criterion[0].value = width; + criterion[1].capability = DM_HEIGHT; + criterion[1].comparison = EQ; + criterion[1].value = height; + got = specialCaseParse(response, + &criterion[2], 1 << DM_WIDTH); + if (got >= 0) { + return got + 2; + } else { + return -1; + } + } + return -1; + case ':': + /* The :BPP case. */ + if (mask & (1 << DM_PIXEL_DEPTH)) { + return -1; + } + bpp = (int) strtol(&word[1], &response, 0); + if (response == &word[1]) { + /* Not a valid number. */ + return -1; + } + criterion[0].capability = DM_PIXEL_DEPTH; + criterion[0].comparison = EQ; + criterion[0].value = bpp; + got = specialCaseParse(response, + &criterion[1], (1 << DM_WIDTH) | (1 << DM_PIXEL_DEPTH)); + if (got >= 0) { + return got + 1; + } else { + return -1; + } + case '@': + /* The @HZ case. */ + if (mask & (1 << DM_HERTZ)) { + return -1; + } + hertz = (int) strtol(&word[1], &response, 0); + if (response == &word[1]) { + /* Not a valid number. */ + return -1; + } + criterion[0].capability = DM_HERTZ; + criterion[0].comparison = EQ; + criterion[0].value = hertz; + got = specialCaseParse(response, + &criterion[1], ~DM_HERTZ); + if (got >= 0) { + return got + 1; + } else { + return -1; + } + case '\0': + return 0; + } + return -1; +} + +/* This routine is based on similiar code in glut_dstr.c */ +static int +parseCriteria(char *word, Criterion * criterion) +{ + char *cstr, *vstr, *response; + int comparator, value = 0; + + cstr = strpbrk(word, "=>': + if (cstr[1] == '=') { + comparator = GTE; + vstr = &cstr[2]; + } else { + comparator = GT; + vstr = &cstr[1]; + } + break; + case '<': + if (cstr[1] == '=') { + comparator = LTE; + vstr = &cstr[2]; + } else { + comparator = LT; + vstr = &cstr[1]; + } + break; + case '!': + if (cstr[1] == '=') { + comparator = NEQ; + vstr = &cstr[2]; + } else { + return -1; + } + break; + default: + return -1; + } + value = (int) strtol(vstr, &response, 0); + if (response == vstr) { + /* Not a valid number. */ + return -1; + } + *cstr = '\0'; + } else { + comparator = NONE; + } + switch (word[0]) { + case 'b': + if (!strcmp(word, "bpp")) { + criterion[0].capability = DM_PIXEL_DEPTH; + if (comparator == NONE) { + return -1; + } else { + criterion[0].comparison = comparator; + criterion[0].value = value; + return 1; + } + } + return -1; + case 'h': + if (!strcmp(word, "height")) { + criterion[0].capability = DM_HEIGHT; + if (comparator == NONE) { + return -1; + } else { + criterion[0].comparison = comparator; + criterion[0].value = value; + return 1; + } + } + if (!strcmp(word, "hertz")) { + criterion[0].capability = DM_HERTZ; + if (comparator == NONE) { + return -1; + } else { + criterion[0].comparison = comparator; + criterion[0].value = value; + return 1; + } + } + return -1; + case 'n': + if (!strcmp(word, "num")) { + criterion[0].capability = DM_NUM; + if (comparator == NONE) { + return -1; + } else { + criterion[0].comparison = comparator; + criterion[0].value = value; + return 1; + } + } + return -1; + case 'w': + if (!strcmp(word, "width")) { + criterion[0].capability = DM_WIDTH; + if (comparator == NONE) { + return -1; + } else { + criterion[0].comparison = comparator; + criterion[0].value = value; + return 1; + } + } + return -1; + } + if (comparator == NONE) { + return specialCaseParse(word, criterion, 0); + } + return -1; +} + +/* This routine is based on similiar code in glut_dstr.c */ +static Criterion * +parseDisplayString(const char *display, int *ncriteria) +{ + Criterion *criteria = NULL; + int n, parsed; + char *copy, *word; + + copy = __glutStrdup(display); + /* Attempt to estimate how many criteria entries should be + needed. */ + n = 0; + word = strtok(copy, " \t"); + while (word) { + n++; + word = strtok(NULL, " \t"); + } + /* Allocate number of words of criteria. A word + could contain as many as four criteria in the + worst case. Example: 800x600:16@60 */ + criteria = (Criterion *) malloc(4 * n * sizeof(Criterion)); + if (!criteria) { + __glutFatalError("out of memory."); + } + + /* Re-copy the copy of the display string. */ + strcpy(copy, display); + + n = 0; + word = strtok(copy, " \t"); + while (word) { + parsed = parseCriteria(word, &criteria[n]); + if (parsed >= 0) { + n += parsed; + } else { + __glutWarning("Unrecognized game mode string word: %s (ignoring)\n", word); + } + word = strtok(NULL, " \t"); + } + + free(copy); + *ncriteria = n; + return criteria; +} + +void GLUTAPIENTRY +glutGameModeString(const char *string) +{ + Criterion *criteria; + int ncriteria; + + initGameModeSupport(); + criteria = parseDisplayString(string, &ncriteria); + currentDm = findMatch(dmodes, ndmodes, criteria, ncriteria); + free(criteria); +} + +int GLUTAPIENTRY +glutEnterGameMode(void) +{ + GLUTwindow *window; + int width, height; + Window win; + + if (__glutMappedMenu) { + __glutFatalUsage("entering game mode not allowed while menus in use"); + } + if (__glutGameModeWindow) { + /* Already in game mode, so blow away game mode + window so apps can change resolutions. */ + window = __glutGameModeWindow; + /* Setting the game mode window to NULL tricks + the window destroy code into not undoing the + screen display change since we plan on immediately + doing another mode change. */ + __glutGameModeWindow = NULL; + __glutDestroyWindow(window, window); + } + + /* Assume default screen size until we find out if we + can actually change the display settings. */ + width = __glutScreenWidth; + height = __glutScreenHeight; + + if (currentDm) { +#ifdef _WIN32 + LONG status; + static int registered = 0; + + status = ChangeDisplaySettings(¤tDm->devmode, + CDS_FULLSCREEN); + if (status == DISP_CHANGE_SUCCESSFUL) { + __glutDisplaySettingsChanged = 1; + width = currentDm->cap[DM_WIDTH]; + height = currentDm->cap[DM_HEIGHT]; + if (!registered) { + atexit(__glutCloseDownGameMode); + registered = 1; + } + } else { + /* Switch back to default resolution. */ + ChangeDisplaySettings(NULL, 0); + } +#endif + } + + window = __glutCreateWindow(NULL, 0, 0, + width, height, /* game mode */ 1); + win = window->win; + +#if !defined(_WIN32) && !defined(__OS2__) + if (__glutMotifHints == None) { + __glutMotifHints = XSGIFastInternAtom(__glutDisplay, "_MOTIF_WM_HINTS", + SGI_XA__MOTIF_WM_HINTS, 0); + if (__glutMotifHints == None) { + __glutWarning("Could not intern X atom for _MOTIF_WM_HINTS."); + } + } + + /* Game mode window is a toplevel window. */ + XSetWMProtocols(__glutDisplay, win, &__glutWMDeleteWindow, 1); +#endif + + /* Schedule the fullscreen property to be added and to + make sure the window is configured right. Win32 + doesn't need this. */ + window->desiredX = 0; + window->desiredY = 0; + window->desiredWidth = width; + window->desiredHeight = height; + window->desiredConfMask |= CWX | CWY | CWWidth | CWHeight; +#ifdef _WIN32 + /* Win32 does not want to use GLUT_FULL_SCREEN_WORK + for game mode because we need to be maximizing + the window in game mode, not just sizing it to + take up the full screen. The Win32-ness of game + mode happens when you pass 1 in the gameMode parameter + to __glutCreateWindow above. A gameMode of creates + a WS_POPUP window, not a standard WS_OVERLAPPEDWINDOW + window. WS_POPUP ensures the taskbar is hidden. */ + __glutPutOnWorkList(window, + GLUT_CONFIGURE_WORK); +#else + __glutPutOnWorkList(window, + GLUT_CONFIGURE_WORK | GLUT_FULL_SCREEN_WORK); +#endif + + __glutGameModeWindow = window; + return window->num + 1; +} + +int GLUTAPIENTRY +glutGameModeGet(GLenum mode) +{ + switch (mode) { + case GLUT_GAME_MODE_ACTIVE: + return __glutGameModeWindow != NULL; + case GLUT_GAME_MODE_POSSIBLE: + return currentDm != NULL; + case GLUT_GAME_MODE_WIDTH: + return currentDm ? currentDm->cap[DM_WIDTH] : -1; + case GLUT_GAME_MODE_HEIGHT: + return currentDm ? currentDm->cap[DM_HEIGHT] : -1; + case GLUT_GAME_MODE_PIXEL_DEPTH: + return currentDm ? currentDm->cap[DM_PIXEL_DEPTH] : -1; + case GLUT_GAME_MODE_REFRESH_RATE: + return currentDm ? currentDm->cap[DM_HERTZ] : -1; + case GLUT_GAME_MODE_DISPLAY_CHANGED: + return __glutDisplaySettingsChanged; + default: + return -1; + } +}  \ No newline at end of file diff --git a/src/glut/os2/glut_win.cpp b/src/glut/os2/glut_win.cpp index 82abba87e4..53ff5d5d95 100644 --- a/src/glut/os2/glut_win.cpp +++ b/src/glut/os2/glut_win.cpp @@ -1,1221 +1,1221 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -#ifdef __VMS -#include -#endif - -#include -#include -#include -#include -#if defined(__OS2__) -#define POKA 0 - #include "WarpGL.h" - #include "glutos2.h" - #include "glutint.h" - - #include "gl\os2mesa.h" - -// -//define for resource id for main GLUT window, in samples it is defined in GL_TEST.h - #define ID_WINDOW 256 - - int evglSetPixelFormat(int iPixelFormat); - HPS hpsCurrent; - -#elif !defined(_WIN32) - -#include -#include -#endif - -#include "glutint.h" - -GLUTwindow *__glutCurrentWindow = NULL; -GLUTwindow **__glutWindowList = NULL; -int __glutWindowListSize = 0; -#if !defined(_WIN32) && !defined(__OS2__) -GLUTstale *__glutStaleWindowList = NULL; -#endif -GLUTwindow *__glutMenuWindow = NULL; - -void (*__glutFreeOverlayFunc) (GLUToverlay *); -XVisualInfo *(*__glutDetermineVisualFromString) (char *string, Bool * treatAsSingle, - Criterion * requiredCriteria, int nRequired, int requiredMask, void** fbc) = NULL; - -static Criterion requiredWindowCriteria[] = -{ - {LEVEL, EQ, 0}, - {TRANSPARENT, EQ, 0} -}; -static int numRequiredWindowCriteria = sizeof(requiredWindowCriteria) / sizeof(Criterion); -static int requiredWindowCriteriaMask = (1 << LEVEL) | (1 << TRANSPARENT); - -static void -cleanWindowWorkList(GLUTwindow * window) -{ - GLUTwindow **pEntry = &__glutWindowWorkList; - GLUTwindow *entry = __glutWindowWorkList; - - /* Tranverse singly-linked window work list look for the - window. */ - while (entry) { - if (entry == window) { - /* Found it; delete it. */ - *pEntry = entry->prevWorkWin; - return; - } else { - pEntry = &entry->prevWorkWin; - entry = *pEntry; - } - } -} - -#if !defined(_WIN32) && !defined(__OS2PM__) - -static void -cleanStaleWindowList(GLUTwindow * window) -{ - GLUTstale **pEntry = &__glutStaleWindowList; - GLUTstale *entry = __glutStaleWindowList; - - /* Tranverse singly-linked stale window list look for the - window ID. */ - while (entry) { - if (entry->window == window) { - /* Found it; delete it. */ - *pEntry = entry->next; - free(entry); - return; - } else { - pEntry = &entry->next; - entry = *pEntry; - } - } -} - -#endif - -static GLUTwindow *__glutWindowCache = NULL; - -GLUTwindow * -__glutGetWindow(Window win) -{ - int i; - - /* Does win belong to the last window ID looked up? */ - if (__glutWindowCache && (win == __glutWindowCache->win || - (__glutWindowCache->overlay && win == - __glutWindowCache->overlay->win))) { - return - __glutWindowCache; - } - /* Otherwise scan the window list looking for the window ID. */ - for (i = 0; i < __glutWindowListSize; i++) { - if (__glutWindowList[i]) { - if (win == __glutWindowList[i]->win) { - __glutWindowCache = __glutWindowList[i]; - return __glutWindowCache; - } - if (__glutWindowList[i]->overlay) { - if (win == __glutWindowList[i]->overlay->win) { - __glutWindowCache = __glutWindowList[i]; - return __glutWindowCache; - } - } - } - } -#if !defined(_WIN32) && !defined(__OS2PM__) - { - GLUTstale *entry; - - /* Scan through destroyed overlay window IDs for which no - DestroyNotify has yet been received. */ - for (entry = __glutStaleWindowList; entry; entry = entry->next) { - if (entry->win == win) - return entry->window; - } - } -#endif - return NULL; -} - -/* CENTRY */ -int GLUTAPIENTRY -glutGetWindow(void) -{ - if (__glutCurrentWindow) { - return __glutCurrentWindow->num + 1; - } else { - return 0; - } -} -/* ENDCENTRY */ - -void -__glutSetWindow(GLUTwindow * window) -{ - /* It is tempting to try to short-circuit the call to - glXMakeCurrent if we "know" we are going to make current - to a window we are already current to. In fact, this - assumption breaks when GLUT is expected to integrated with - other OpenGL windowing APIs that also make current to - OpenGL contexts. Since glXMakeCurrent short-circuits the - "already bound" case, GLUT avoids the temptation to do so - too. */ - __glutCurrentWindow = window; - - MAKE_CURRENT_LAYER(__glutCurrentWindow); - -#if !defined(_WIN32) && !defined(__OS2__) - /* We should be careful to force a finish between each - iteration through the GLUT main loop if indirect OpenGL - contexts are in use; indirect contexts tend to have much - longer latency because lots of OpenGL extension requests - can queue up in the X protocol stream. We accomplish this - by posting GLUT_FINISH_WORK to be done. */ - if (!__glutCurrentWindow->isDirect) - __glutPutOnWorkList(__glutCurrentWindow, GLUT_FINISH_WORK); -#endif - - /* If debugging is enabled, we'll want to check this window - for any OpenGL errors every iteration through the GLUT - main loop. To accomplish this, we post the - GLUT_DEBUG_WORK to be done on this window. */ - if (__glutDebug) { - __glutPutOnWorkList(__glutCurrentWindow, GLUT_DEBUG_WORK); - } -} - -/* CENTRY */ -void GLUTAPIENTRY -glutSetWindow(int win) -{ - GLUTwindow *window; - - if (win < 1 || win > __glutWindowListSize) { - __glutWarning("glutSetWindow attempted on bogus window."); - return; - } - window = __glutWindowList[win - 1]; - if (!window) { - __glutWarning("glutSetWindow attempted on bogus window."); - return; - } - __glutSetWindow(window); -} -/* ENDCENTRY */ - -static int -getUnusedWindowSlot(void) -{ - int i; - - /* Look for allocated, unused slot. */ - for (i = 0; i < __glutWindowListSize; i++) { - if (!__glutWindowList[i]) { - return i; - } - } - /* Allocate a new slot. */ - __glutWindowListSize++; - if (__glutWindowList) { - __glutWindowList = (GLUTwindow **) - realloc(__glutWindowList, - __glutWindowListSize * sizeof(GLUTwindow *)); - } else { - /* XXX Some realloc's do not correctly perform a malloc - when asked to perform a realloc on a NULL pointer, - though the ANSI C library spec requires this. */ - __glutWindowList = (GLUTwindow **) - malloc(sizeof(GLUTwindow *)); - } - if (!__glutWindowList) - __glutFatalError("out of memory."); - __glutWindowList[__glutWindowListSize - 1] = NULL; - return __glutWindowListSize - 1; -} - -static XVisualInfo * -getVisualInfoCI(unsigned int mode) -{ -#if POKA - static int bufSizeList[] = - {16, 12, 8, 4, 2, 1, 0}; - XVisualInfo *vi; - int list[32]; - int i, n = 0; - - /* Should not be looking at display mode mask if - __glutDisplayString is non-NULL. */ - assert(!__glutDisplayString); - - list[n++] = GLX_BUFFER_SIZE; - list[n++] = 1; - if (GLUT_WIND_IS_DOUBLE(mode)) { - list[n++] = GLX_DOUBLEBUFFER; - } - if (GLUT_WIND_IS_STEREO(mode)) { - list[n++] = GLX_STEREO; - } - if (GLUT_WIND_HAS_DEPTH(mode)) { - list[n++] = GLX_DEPTH_SIZE; - list[n++] = 1; - } - if (GLUT_WIND_HAS_STENCIL(mode)) { - list[n++] = GLX_STENCIL_SIZE; - list[n++] = 1; - } - list[n] = (int) None; /* terminate list */ - - /* glXChooseVisual specify GLX_BUFFER_SIZE prefers the - "smallest index buffer of at least the specified size". - This would be reasonable if GLUT allowed the user to - specify the required buffe size, but GLUT's display mode - is too simplistic (easy to use?). GLUT should try to find - the "largest". So start with a large buffer size and - shrink until we find a matching one that exists. */ - - for (i = 0; bufSizeList[i]; i++) { - /* XXX Assumes list[1] is where GLX_BUFFER_SIZE parameter - is. */ - list[1] = bufSizeList[i]; - vi = glXChooseVisual(__glutDisplay, - __glutScreen, list); - if (vi) - return vi; - } - return NULL; -#else - return - glXChooseVisual(mode); - -#endif -} - -static XVisualInfo * -getVisualInfoRGB(unsigned int mode) -{ -#if POKA - int list[32]; - int n = 0; - - /* Should not be looking at display mode mask if - __glutDisplayString is non-NULL. */ - assert(!__glutDisplayString); - - /* XXX Would a caching mechanism to minize the calls to - glXChooseVisual? You'd have to reference count - XVisualInfo* pointers. Would also have to properly - interact with glutInitDisplayString. */ - - list[n++] = GLX_RGBA; - list[n++] = GLX_RED_SIZE; - list[n++] = 1; - list[n++] = GLX_GREEN_SIZE; - list[n++] = 1; - list[n++] = GLX_BLUE_SIZE; - list[n++] = 1; - if (GLUT_WIND_HAS_ALPHA(mode)) { - list[n++] = GLX_ALPHA_SIZE; - list[n++] = 1; - } - if (GLUT_WIND_IS_DOUBLE(mode)) { - list[n++] = GLX_DOUBLEBUFFER; - } - if (GLUT_WIND_IS_STEREO(mode)) { - list[n++] = GLX_STEREO; - } - if (GLUT_WIND_HAS_DEPTH(mode)) { - list[n++] = GLX_DEPTH_SIZE; - list[n++] = 1; - } - if (GLUT_WIND_HAS_STENCIL(mode)) { - list[n++] = GLX_STENCIL_SIZE; - list[n++] = 1; - } - if (GLUT_WIND_HAS_ACCUM(mode)) { - list[n++] = GLX_ACCUM_RED_SIZE; - list[n++] = 1; - list[n++] = GLX_ACCUM_GREEN_SIZE; - list[n++] = 1; - list[n++] = GLX_ACCUM_BLUE_SIZE; - list[n++] = 1; - if (GLUT_WIND_HAS_ALPHA(mode)) { - list[n++] = GLX_ACCUM_ALPHA_SIZE; - list[n++] = 1; - } - } -#if defined(GLX_VERSION_1_1) && (defined(GLX_SGIS_multisample) || defined(GLX_ARB_multisample)) - if (GLUT_WIND_IS_MULTISAMPLE(mode)) { - if (!__glutIsSupportedByGLX("GLX_SGIS_multisample") && - !__glutIsSupportedByGLX("GLX_ARB_multisample")) - return NULL; -#if defined(GLX_ARB_multisample) - list[n++] = GLX_SAMPLES_ARB; -#elif defined(GLX_SGIS_multisample) - list[n++] = GLX_SAMPLES_SGIS; -#endif - /* XXX Is 4 a reasonable minimum acceptable number of - samples? */ - list[n++] = 4; - } -#endif - list[n] = (int) None; /* terminate list */ - - return glXChooseVisual(__glutDisplay, - __glutScreen, list); -#else /* POKA */ - - return - glXChooseVisual(mode); - -#endif -} - -XVisualInfo * -__glutGetVisualInfo(unsigned int mode) -{ - /* XXX GLUT_LUMINANCE not implemented for GLUT 3.0. */ - if (GLUT_WIND_IS_LUMINANCE(mode)) - return NULL; - - if (GLUT_WIND_IS_RGB(mode)) - return getVisualInfoRGB(mode); - else - return getVisualInfoCI(mode); -} - -XVisualInfo * -__glutDetermineVisual( - unsigned int displayMode, - Bool * treatAsSingle, - XVisualInfo * (getVisualInfo) (unsigned int)) -{ - XVisualInfo *vis; - - /* Should not be looking at display mode mask if - __glutDisplayString is non-NULL. */ - assert(!__glutDisplayString); - - *treatAsSingle = GLUT_WIND_IS_SINGLE(displayMode); - vis = getVisualInfo(displayMode); - if (!vis) { - /* Fallback cases when can't get exactly what was asked - for... */ - if (GLUT_WIND_IS_SINGLE(displayMode)) { - /* If we can't find a single buffered visual, try looking - for a double buffered visual. We can treat a double - buffered visual as a single buffer visual by changing - the draw buffer to GL_FRONT and treating any swap - buffers as no-ops. */ - displayMode |= GLUT_DOUBLE; - vis = getVisualInfo(displayMode); - *treatAsSingle = True; - } - if (!vis && GLUT_WIND_IS_MULTISAMPLE(displayMode)) { - /* If we can't seem to get multisampling (ie, not Reality - Engine class graphics!), go without multisampling. It - is up to the application to query how many multisamples - were allocated (0 equals no multisampling) if the - application is going to use multisampling for more than - just antialiasing. */ - displayMode &= ~GLUT_MULTISAMPLE; - vis = getVisualInfo(displayMode); - } - } - return vis; -} - -static void GLUTCALLBACK -__glutDefaultDisplay(void) -{ - /* XXX Remove the warning after GLUT 3.0. */ - __glutWarning("The following is a new check for GLUT 3.0; update your code."); - __glutFatalError( - "redisplay needed for window %d, but no display callback.", - __glutCurrentWindow->num + 1); -} - -void GLUTCALLBACK -__glutDefaultReshape(int width, int height) -{ - GLUToverlay *overlay; - - /* Adjust the viewport of the window (and overlay if one - exists). */ - MAKE_CURRENT_WINDOW(__glutCurrentWindow); - glViewport(0, 0, (GLsizei) width, (GLsizei) height); - overlay = __glutCurrentWindow->overlay; - if (overlay) { - MAKE_CURRENT_OVERLAY(overlay); - glViewport(0, 0, (GLsizei) width, (GLsizei) height); - } - /* Make sure we are current to the current layer (application - should be able to count on the current layer not changing - unless the application explicitly calls glutUseLayer). */ - MAKE_CURRENT_LAYER(__glutCurrentWindow); -} - -XVisualInfo * -__glutDetermineWindowVisual(Bool * treatAsSingle, Bool * visAlloced, void **fbc) -{ - if (__glutDisplayString) { - - /* __glutDisplayString should be NULL except if - glutInitDisplayString has been called to register a - different display string. Calling glutInitDisplayString - means using a string instead of an integer mask determine - the visual to use. Using the function pointer variable - __glutDetermineVisualFromString below avoids linking in - the code for implementing glutInitDisplayString (ie, - glut_dstr.o) unless glutInitDisplayString gets called by - the application. */ - - assert(__glutDetermineVisualFromString); - *visAlloced = False; - *fbc = NULL; - return __glutDetermineVisualFromString(__glutDisplayString, treatAsSingle, - requiredWindowCriteria, numRequiredWindowCriteria, requiredWindowCriteriaMask, fbc); - } else { - *visAlloced = True; - *fbc = NULL; - return __glutDetermineVisual(__glutDisplayMode, - treatAsSingle, __glutGetVisualInfo); - } -} - -/* ARGSUSED5 */ /* Only Win32 uses gameMode parameter. */ -GLUTwindow * -__glutCreateWindow(GLUTwindow * parent, - int x, int y, int width, int height, int gameMode) -{ - GLUTwindow *window; - XSetWindowAttributes wa; - unsigned long attribMask; - int winnum; - int i; -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_fbconfig) - GLXFBConfigSGIX fbc; -#else - void *fbc; -#endif - -#if defined(__OS2PM__) - { - extern HAB hab; /* PM anchor block handle */ - CLASSINFO classinfo; - - if(!WinQueryClassInfo(hab,"GLUT", &classinfo) ) - __glutOpenOS2Connection(NULL); - } -#elif defined(_WIN32) - WNDCLASS wc; - int style; - - if (!GetClassInfo(GetModuleHandle(NULL), "GLUT", &wc)) { - __glutOpenWin32Connection(NULL); - } -#else - if (!__glutDisplay) { - __glutOpenXConnection(NULL); - } -#endif - -#ifndef __OS2PM__ - if (__glutGameModeWindow) { - __glutFatalError("cannot create windows in game mode."); - } -#endif - - winnum = getUnusedWindowSlot(); - window = (GLUTwindow *) malloc(sizeof(GLUTwindow)); - if (!window) { - __glutFatalError("out of memory."); - } - window->num = winnum; - -#if defined(__OS2PM__) - /* Add this new window to the window list. */ - __glutWindowList[winnum] = window; - window->shownState = -1; -#endif - -#if !defined(_WIN32) && !defined(__OS2PM__) - window->vis = __glutDetermineWindowVisual(&window->treatAsSingle, - &window->visAlloced, (void**) &fbc); - if (!window->vis) { - __glutFatalError( - "visual with necessary capabilities not found."); - } - __glutSetupColormap(window->vis, &window->colormap, &window->cmap); -#endif - window->eventMask = StructureNotifyMask | ExposureMask; - - attribMask = CWBackPixmap | CWBorderPixel | CWColormap | CWEventMask; - wa.background_pixmap = None; - wa.border_pixel = 0; - wa.colormap = window->cmap; - wa.event_mask = window->eventMask; - if (parent) { - if (parent->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK) - wa.event_mask |= GLUT_HACK_STOP_PROPAGATE_MASK; - attribMask |= CWDontPropagate; - wa.do_not_propagate_mask = parent->eventMask & GLUT_DONT_PROPAGATE_FILTER_MASK; - } else { - wa.do_not_propagate_mask = 0; - } - - /* Stash width and height before Win32's __glutAdjustCoords - possibly overwrites the values. */ - window->width = width; - window->height = height; - window->forceReshape = True; - window->ignoreKeyRepeat = False; - -#if defined(__OS2PM__) - - { ULONG flStyle=0; - int ii; - ERRORID erridErrorCode;/* last error id code */ - extern HAB hab; /* PM anchor block handle */ - - if (parent) { - flStyle = WS_CLIPCHILDREN|WS_VISIBLE; - } else { - if (gameMode) { - /* Game mode window should be a WS_POPUP window to - ensure that the taskbar is hidden by it. A standard - WS_OVERLAPPEDWINDOW does not hide the task bar. */ - flStyle = FCF_STANDARD | WS_MAXIMIZED; - } else { - /* A standard toplevel window with borders and such. */ - flStyle = FCF_STANDARD | WS_CLIPCHILDREN; -// flStyle = WS_OVERLAPPEDWINDOW; - } - } -{ - HWND hwnd; /* Window */ - ULONG ListBoxId; /* Window id */ - /* (supplied by application) */ - - - HWND hwndClient; /* handle to the client */ - HWND hwndFrame; /* handle to the frame */ - PFNWP GenericWndProc; - FRAMECDATA fcd; - RECTL rect; /* Boundary rectangle */ - - - -/************************************************/ -// flCreate = (FCF_STANDARD) & ~FCF_TASKLIST; -/**********************************/ - if (parent) - { window->frame = NULL; - - hwnd = WinCreateWindow(parent->win, /* Parent window */ - "GLUTCHILD", /* Class name */ - "", /* Window text */ - flStyle, /* Window style */ - x, y, /* Position (x,y) */ - width, height, /* Size (width,height) */ - parent->win, /* Owner window */ - HWND_TOP, /* Sibling window */ - 0, /* Window id */ - NULL, /* Control data */ - NULL); /* Pres parameters */ - - erridErrorCode = WinGetLastError(hab); - window->win = hwnd; - - window->hdc = WinOpenWindowDC(window->win); - window->hpsBuffer = hpsCurrent; - - - rect.xLeft = x; - rect.xRight = x+width; - rect.yBottom = y; - rect.yTop = y + height; - -/***** else parent *****************************/ - } else { - hwnd = WinCreateStdWindow(HWND_DESKTOP, - 0, /* WS_VISIBLE frame-window style */ - &flStyle, /* window style */ - "GLUT", /* class name */ - "GLUT",/* window title */ - 0L, /* default client style */ - NULLHANDLE, /* resource in executable file */ - ID_WINDOW, /* resource id */ - &hwndClient); /* receives client window handle */ - - erridErrorCode = WinGetLastError(hab); - window->win = hwndClient; - window->frame = hwnd; - window->hdc = WinOpenWindowDC(window->win); - - window->hpsBuffer = hpsCurrent; - - -/* converts a client window's boundaries into an equivalent frame rectangle */ - rect.xLeft = x; - rect.xRight = x+width; - rect.yBottom = y; - rect.yTop = y + height; - - /* calculate equivalent frame boundary from boundary data */ - WinCalcFrameRect(window->frame, &rect, FALSE); - } -/***** endof if(parent) *****************************/ - - /* Must set the XHDC for fake glXChooseVisual & fake - glXCreateContext & fake XAllocColorCells. */ - XHDC = window->hdc; - XHWND = window->win; - window->vis = __glutDetermineWindowVisual(&window->treatAsSingle, - &window->visAlloced, &fbc); - if (!window->vis) - { __glutFatalError( - "pixel format with necessary capabilities not found."); - } - { int rc; - rc = wglChoosePixelFormat(window->hdc, window->vis), - -// evglSetPixelFormat(2); /* int iPixelFormat 1 - doublebuffer/2 - single buffer ??*/ - wglSetPixelFormat(window->hdc,rc,window->vis); - } - __glutSetupColormap(window->vis, &window->colormap, &window->cmap); - - window->ctx = glXCreateContext(window->hpsBuffer, window->vis, - None, __glutTryDirect); - - WinSetWindowPos(hwnd, - HWND_TOP,rect.xLeft,rect.yBottom, - rect.xRight-rect.xLeft, rect.yTop-rect.yBottom, - SWP_ACTIVATE | SWP_MOVE | SWP_SIZE | SWP_SHOW|SWP_ZORDER); /* flags*/ - - /* Make sure subwindows get a windowStatus callback. */ - if (parent) - WinPostMsg(parent->win, WM_ACTIVATE, 0, 0); - - } -} - -#elif defined(_WIN32) - - __glutAdjustCoords(parent ? parent->win : NULL, - &x, &y, &width, &height); - if (parent) { - style = WS_CHILD; - } else { - if (gameMode) { - /* Game mode window should be a WS_POPUP window to - ensure that the taskbar is hidden by it. A standard - WS_OVERLAPPEDWINDOW does not hide the task bar. */ - style = WS_POPUP | WS_MAXIMIZE; - } else { - /* A standard toplevel window with borders and such. */ - style = WS_OVERLAPPEDWINDOW; - } - } - window->win = CreateWindow("GLUT", "GLUT", - WS_CLIPSIBLINGS | WS_CLIPCHILDREN | style, - x, y, width, height, parent ? parent->win : __glutRoot, - NULL, GetModuleHandle(NULL), 0); - window->hdc = GetDC(window->win); - /* Must set the XHDC for fake glXChooseVisual & fake - glXCreateContext & fake XAllocColorCells. */ - XHDC = window->hdc; - window->vis = __glutDetermineWindowVisual(&window->treatAsSingle, - &window->visAlloced, &fbc); - if (!window->vis) { - __glutFatalError( - "pixel format with necessary capabilities not found."); - } - if (!SetPixelFormat(window->hdc, - ChoosePixelFormat(window->hdc, window->vis), - window->vis)) { - __glutFatalError("SetPixelFormat failed during window create."); - } - __glutSetupColormap(window->vis, &window->colormap, &window->cmap); - /* Make sure subwindows get a windowStatus callback. */ - if (parent) { - PostMessage(parent->win, WM_ACTIVATE, 0, 0); - } - window->renderDc = window->hdc; -#else - window->win = XCreateWindow(__glutDisplay, - parent == NULL ? __glutRoot : parent->win, - x, y, width, height, 0, - window->vis->depth, InputOutput, window->vis->visual, - attribMask, &wa); -#endif - window->renderWin = window->win; -#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_fbconfig) - if (fbc) { - window->ctx = __glut_glXCreateContextWithConfigSGIX(__glutDisplay, fbc, - GLX_RGBA_TYPE_SGIX, None, __glutTryDirect); - } else -#endif -#if defined(__OS2PM__) -// window->ctx = glXCreateContext(window->hpsBuffer, window->vis, -// None, __glutTryDirect); -#else - window->ctx = glXCreateContext(__glutDisplay, window->vis, - None, __glutTryDirect); -#endif - if (!window->ctx) { - __glutFatalError( - "failed to create OpenGL rendering context."); - } - window->renderCtx = window->ctx; -#if !defined(_WIN32) && !defined(__OS2PM__) - window->isDirect = glXIsDirect(__glutDisplay, window->ctx); - if (__glutForceDirect) { - if (!window->isDirect) - __glutFatalError("direct rendering not possible."); - } -#endif - - window->parent = parent; - if (parent) { - window->siblings = parent->children; - parent->children = window; - } else { - window->siblings = NULL; - } - window->overlay = NULL; - window->children = NULL; - window->display = __glutDefaultDisplay; - window->reshape = __glutDefaultReshape; - window->mouse = NULL; - window->motion = NULL; - window->passive = NULL; - window->entry = NULL; - window->keyboard = NULL; - window->keyboardUp = NULL; - window->windowStatus = NULL; - window->visibility = NULL; - window->special = NULL; - window->specialUp = NULL; - window->buttonBox = NULL; - window->dials = NULL; - window->spaceMotion = NULL; - window->spaceRotate = NULL; - window->spaceButton = NULL; - window->tabletMotion = NULL; - window->tabletButton = NULL; -#ifdef _WIN32 - window->joystick = NULL; - window->joyPollInterval = 0; -#endif - -#if defined(__OS2PM__) - window->wm_command = NULL; -#endif - - window->tabletPos[0] = -1; - window->tabletPos[1] = -1; -#if defined(__OS2PM__) - if(window->shownState == -1) - window->shownState = 0; - window->visState = window->shownState; -#else - window->shownState = 0; - window->visState = -1; /* not VisibilityUnobscured, - VisibilityPartiallyObscured, or - VisibilityFullyObscured */ -#endif - window->entryState = -1; /* not EnterNotify or LeaveNotify */ - - window->desiredConfMask = 0; - window->buttonUses = 0; - window->cursor = GLUT_CURSOR_INHERIT; - - /* Setup window to be mapped when glutMainLoop starts. */ - window->workMask = GLUT_MAP_WORK; -#ifdef _WIN32 - if (gameMode) { - /* When mapping a game mode window, just show - the window. We have already created the game - mode window with a maximize flag at creation - time. Doing a ShowWindow(window->win, SW_SHOWNORMAL) - would be wrong for a game mode window since it - would unmaximize the window. */ - window->desiredMapState = GameModeState; - } else { - window->desiredMapState = NormalState; - } -#else - window->desiredMapState = NormalState; -#endif - window->prevWorkWin = __glutWindowWorkList; - __glutWindowWorkList = window; - - /* Initially, no menus attached. */ - for (i = 0; i < GLUT_MAX_MENUS; i++) { - window->menu[i] = 0; - } - - /* Add this new window to the window list. */ - __glutWindowList[winnum] = window; - - /* Make the new window the current window. */ - __glutSetWindow(window); - - __glutDetermineMesaSwapHackSupport(); - - if (window->treatAsSingle) { - /* We do this because either the window really is single - buffered (in which case this is redundant, but harmless, - because this is the initial single-buffered context - state); or we are treating a double buffered window as a - single-buffered window because the system does not appear - to export any suitable single- buffered visuals (in which - the following are necessary). */ - glDrawBuffer(GL_FRONT); - glReadBuffer(GL_FRONT); - } - return window; -} - -/* CENTRY */ -int GLUTAPIENTRY -glutCreateWindow(const char *title) -{ - static int firstWindow = 1; - GLUTwindow *window; -#if !defined(_WIN32) && !defined(__OS2__) - XWMHints *wmHints; -#endif - Window win; - XTextProperty textprop; - - if (__glutGameModeWindow) { - __glutFatalError("cannot create windows in game mode."); - } - window = __glutCreateWindow(NULL, - __glutSizeHints.x, __glutSizeHints.y, - __glutInitWidth, __glutInitHeight, - /* not game mode */ 0); - win = window->win; - /* Setup ICCCM properties. */ - textprop.value = (unsigned char *) title; - textprop.encoding = XA_STRING; - textprop.format = 8; - textprop.nitems = strlen(title); -#if defined(__OS2__) - WinSetWindowText(window->frame, (PCSZ)title); - if (__glutIconic) { - window->desiredMapState = IconicState; - } -#elif defined(_WIN32) - SetWindowText(win, title); - if (__glutIconic) { - window->desiredMapState = IconicState; - } -#else - wmHints = XAllocWMHints(); - wmHints->initial_state = - __glutIconic ? IconicState : NormalState; - wmHints->flags = StateHint; - XSetWMProperties(__glutDisplay, win, &textprop, &textprop, - /* Only put WM_COMMAND property on first window. */ - firstWindow ? __glutArgv : NULL, - firstWindow ? __glutArgc : 0, - &__glutSizeHints, wmHints, NULL); - XFree(wmHints); - XSetWMProtocols(__glutDisplay, win, &__glutWMDeleteWindow, 1); -#endif - firstWindow = 0; - return window->num + 1; -} - -#ifdef _WIN32 -int GLUTAPIENTRY -__glutCreateWindowWithExit(const char *title, void (__cdecl *exitfunc)(int)) -{ - __glutExitFunc = exitfunc; - return glutCreateWindow(title); -} -#endif - -int GLUTAPIENTRY -glutCreateSubWindow(int win, int x, int y, int width, int height) -{ - GLUTwindow *window; - - window = __glutCreateWindow(__glutWindowList[win - 1], - x, y, width, height, /* not game mode */ 0); -#if !defined(_WIN32) && !defined(__OS2__) - { - GLUTwindow *toplevel; - - toplevel = __glutToplevelOf(window); - if (toplevel->cmap != window->cmap) { - __glutPutOnWorkList(toplevel, GLUT_COLORMAP_WORK); - } - } -#endif - return window->num + 1; -} -/* ENDCENTRY */ - -void -__glutDestroyWindow(GLUTwindow * window, - GLUTwindow * initialWindow) -{ - GLUTwindow **prev, *cur, *parent, *siblings; - - /* Recursively destroy any children. */ - cur = window->children; - while (cur) { - siblings = cur->siblings; - __glutDestroyWindow(cur, initialWindow); - cur = siblings; - } - /* Remove from parent's children list (only necessary for - non-initial windows and subwindows!). */ - parent = window->parent; - if (parent && parent == initialWindow->parent) { - prev = &parent->children; - cur = parent->children; - while (cur) { - if (cur == window) { - *prev = cur->siblings; - break; - } - prev = &(cur->siblings); - cur = cur->siblings; - } - } - /* Unbind if bound to this window. */ - if (window == __glutCurrentWindow) { - UNMAKE_CURRENT(); - __glutCurrentWindow = NULL; - } - /* Begin tearing down window itself. */ - if (window->overlay) { - __glutFreeOverlayFunc(window->overlay); - } - XDestroyWindow(__glutDisplay, window->win); - glXDestroyContext(__glutDisplay, window->ctx); - if (window->colormap) { - /* Only color index windows have colormap data structure. */ - __glutFreeColormap(window->colormap); - } - /* NULLing the __glutWindowList helps detect is a window - instance has been destroyed, given a window number. */ - __glutWindowList[window->num] = NULL; - - /* Cleanup data structures that might contain window. */ - cleanWindowWorkList(window); -#if !defined(_WIN32) && !defined(__OS2__) - cleanStaleWindowList(window); -#endif - /* Remove window from the "get window cache" if it is there. */ - if (__glutWindowCache == window) - __glutWindowCache = NULL; - - if (window->visAlloced) { - /* Only free XVisualInfo* gotten from glXChooseVisual. */ - XFree(window->vis); - } - - if (window == __glutGameModeWindow) { - /* Destroying the game mode window should implicitly - have GLUT leave game mode. */ - __glutCloseDownGameMode(); - } - - free(window); -} - -/* CENTRY */ -void GLUTAPIENTRY -glutDestroyWindow(int win) -{ - GLUTwindow *window = __glutWindowList[win - 1]; - - if (__glutMappedMenu && __glutMenuWindow == window) { - __glutFatalUsage("destroying menu window not allowed while menus in use"); - } -#if !defined(_WIN32) && !defined(__OS2__) - /* If not a toplevel window... */ - if (window->parent) { - /* Destroying subwindows may change colormap requirements; - recalculate toplevel window's WM_COLORMAP_WINDOWS - property. */ - __glutPutOnWorkList(__glutToplevelOf(window->parent), - GLUT_COLORMAP_WORK); - } -#endif - __glutDestroyWindow(window, window); - XFlush(__glutDisplay); -} -/* ENDCENTRY */ - -void -__glutChangeWindowEventMask(long eventMask, Bool add) -{ - if (add) { - /* Add eventMask to window's event mask. */ - if ((__glutCurrentWindow->eventMask & eventMask) != - eventMask) { - __glutCurrentWindow->eventMask |= eventMask; - __glutPutOnWorkList(__glutCurrentWindow, - GLUT_EVENT_MASK_WORK); - } - } else { - /* Remove eventMask from window's event mask. */ - if (__glutCurrentWindow->eventMask & eventMask) { - __glutCurrentWindow->eventMask &= ~eventMask; - __glutPutOnWorkList(__glutCurrentWindow, - GLUT_EVENT_MASK_WORK); - } - } -} - -void GLUTAPIENTRY -glutDisplayFunc(GLUTdisplayCB displayFunc) -{ - /* XXX Remove the warning after GLUT 3.0. */ - if (!displayFunc) - __glutFatalError("NULL display callback not allowed in GLUT 3.0; update your code."); - __glutCurrentWindow->display = displayFunc; -} - -void GLUTAPIENTRY -glutMouseFunc(GLUTmouseCB mouseFunc) -{ - if (__glutCurrentWindow->mouse) { - if (!mouseFunc) { - /* Previous mouseFunc being disabled. */ - __glutCurrentWindow->buttonUses--; - __glutChangeWindowEventMask( - ButtonPressMask | ButtonReleaseMask, - __glutCurrentWindow->buttonUses > 0); - } - } else { - if (mouseFunc) { - /* Previously no mouseFunc, new one being installed. */ - __glutCurrentWindow->buttonUses++; - __glutChangeWindowEventMask( - ButtonPressMask | ButtonReleaseMask, True); - } - } - __glutCurrentWindow->mouse = mouseFunc; -} - -void GLUTAPIENTRY -glutMotionFunc(GLUTmotionCB motionFunc) -{ - /* Hack. Some window managers (4Dwm by default) will mask - motion events if the client is not selecting for button - press and release events. So we select for press and - release events too (being careful to use reference - counting). */ - if (__glutCurrentWindow->motion) { - if (!motionFunc) { - /* previous mouseFunc being disabled */ - __glutCurrentWindow->buttonUses--; - __glutChangeWindowEventMask( - ButtonPressMask | ButtonReleaseMask, - __glutCurrentWindow->buttonUses > 0); - } - } else { - if (motionFunc) { - /* Previously no mouseFunc, new one being installed. */ - __glutCurrentWindow->buttonUses++; - __glutChangeWindowEventMask( - ButtonPressMask | ButtonReleaseMask, True); - } - } - /* Real work of selecting for passive mouse motion. */ - __glutChangeWindowEventMask( - Button1MotionMask | Button2MotionMask | Button3MotionMask, - motionFunc != NULL); - __glutCurrentWindow->motion = motionFunc; -} - -void GLUTAPIENTRY -glutPassiveMotionFunc(GLUTpassiveCB passiveMotionFunc) -{ - __glutChangeWindowEventMask(PointerMotionMask, - passiveMotionFunc != NULL); - - /* Passive motion also requires watching enters and leaves so - that a fake passive motion event can be generated on an - enter. */ - __glutChangeWindowEventMask(EnterWindowMask | LeaveWindowMask, - __glutCurrentWindow->entry != NULL || passiveMotionFunc != NULL); - - __glutCurrentWindow->passive = passiveMotionFunc; -} - -void GLUTAPIENTRY -glutEntryFunc(GLUTentryCB entryFunc) -{ - __glutChangeWindowEventMask(EnterWindowMask | LeaveWindowMask, - entryFunc != NULL || __glutCurrentWindow->passive); - __glutCurrentWindow->entry = entryFunc; - if (!entryFunc) { - __glutCurrentWindow->entryState = -1; - } -} - -void GLUTAPIENTRY -glutWindowStatusFunc(GLUTwindowStatusCB windowStatusFunc) -{ - __glutChangeWindowEventMask(VisibilityChangeMask, - windowStatusFunc != NULL); - __glutCurrentWindow->windowStatus = windowStatusFunc; - if (!windowStatusFunc) { - /* Make state invalid. */ - __glutCurrentWindow->visState = -1; - } -} - -static void GLUTCALLBACK -visibilityHelper(int status) -{ - if (status == GLUT_HIDDEN || status == GLUT_FULLY_COVERED) - __glutCurrentWindow->visibility(GLUT_NOT_VISIBLE); - else - __glutCurrentWindow->visibility(GLUT_VISIBLE); -} - - -void GLUTAPIENTRY -glutVisibilityFunc(GLUTvisibilityCB visibilityFunc) -{ - __glutCurrentWindow->visibility = visibilityFunc; - - if (visibilityFunc) - { glutWindowStatusFunc(visibilityHelper); -#if defined(__OS2PM__) - if(__glutCurrentWindow->shownState >= 0) - { visibilityHelper(__glutCurrentWindow->shownState); - } -#endif - } - else - glutWindowStatusFunc(NULL); -} - -void GLUTAPIENTRY -glutReshapeFunc(GLUTreshapeCB reshapeFunc) -{ - if (reshapeFunc) { - __glutCurrentWindow->reshape = reshapeFunc; - } else { - __glutCurrentWindow->reshape = __glutDefaultReshape; - } -} + +/* Copyright (c) Mark J. Kilgard, 1994, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +#ifdef __VMS +#include +#endif + +#include +#include +#include +#include +#if defined(__OS2__) +#define POKA 0 + #include "WarpGL.h" + #include "glutos2.h" + #include "glutint.h" + + #include "gl\os2mesa.h" + +// +//define for resource id for main GLUT window, in samples it is defined in GL_TEST.h + #define ID_WINDOW 256 + + int evglSetPixelFormat(int iPixelFormat); + HPS hpsCurrent; + +#elif !defined(_WIN32) + +#include +#include +#endif + +#include "glutint.h" + +GLUTwindow *__glutCurrentWindow = NULL; +GLUTwindow **__glutWindowList = NULL; +int __glutWindowListSize = 0; +#if !defined(_WIN32) && !defined(__OS2__) +GLUTstale *__glutStaleWindowList = NULL; +#endif +GLUTwindow *__glutMenuWindow = NULL; + +void (*__glutFreeOverlayFunc) (GLUToverlay *); +XVisualInfo *(*__glutDetermineVisualFromString) (char *string, Bool * treatAsSingle, + Criterion * requiredCriteria, int nRequired, int requiredMask, void** fbc) = NULL; + +static Criterion requiredWindowCriteria[] = +{ + {LEVEL, EQ, 0}, + {TRANSPARENT, EQ, 0} +}; +static int numRequiredWindowCriteria = sizeof(requiredWindowCriteria) / sizeof(Criterion); +static int requiredWindowCriteriaMask = (1 << LEVEL) | (1 << TRANSPARENT); + +static void +cleanWindowWorkList(GLUTwindow * window) +{ + GLUTwindow **pEntry = &__glutWindowWorkList; + GLUTwindow *entry = __glutWindowWorkList; + + /* Tranverse singly-linked window work list look for the + window. */ + while (entry) { + if (entry == window) { + /* Found it; delete it. */ + *pEntry = entry->prevWorkWin; + return; + } else { + pEntry = &entry->prevWorkWin; + entry = *pEntry; + } + } +} + +#if !defined(_WIN32) && !defined(__OS2PM__) + +static void +cleanStaleWindowList(GLUTwindow * window) +{ + GLUTstale **pEntry = &__glutStaleWindowList; + GLUTstale *entry = __glutStaleWindowList; + + /* Tranverse singly-linked stale window list look for the + window ID. */ + while (entry) { + if (entry->window == window) { + /* Found it; delete it. */ + *pEntry = entry->next; + free(entry); + return; + } else { + pEntry = &entry->next; + entry = *pEntry; + } + } +} + +#endif + +static GLUTwindow *__glutWindowCache = NULL; + +GLUTwindow * +__glutGetWindow(Window win) +{ + int i; + + /* Does win belong to the last window ID looked up? */ + if (__glutWindowCache && (win == __glutWindowCache->win || + (__glutWindowCache->overlay && win == + __glutWindowCache->overlay->win))) { + return + __glutWindowCache; + } + /* Otherwise scan the window list looking for the window ID. */ + for (i = 0; i < __glutWindowListSize; i++) { + if (__glutWindowList[i]) { + if (win == __glutWindowList[i]->win) { + __glutWindowCache = __glutWindowList[i]; + return __glutWindowCache; + } + if (__glutWindowList[i]->overlay) { + if (win == __glutWindowList[i]->overlay->win) { + __glutWindowCache = __glutWindowList[i]; + return __glutWindowCache; + } + } + } + } +#if !defined(_WIN32) && !defined(__OS2PM__) + { + GLUTstale *entry; + + /* Scan through destroyed overlay window IDs for which no + DestroyNotify has yet been received. */ + for (entry = __glutStaleWindowList; entry; entry = entry->next) { + if (entry->win == win) + return entry->window; + } + } +#endif + return NULL; +} + +/* CENTRY */ +int GLUTAPIENTRY +glutGetWindow(void) +{ + if (__glutCurrentWindow) { + return __glutCurrentWindow->num + 1; + } else { + return 0; + } +} +/* ENDCENTRY */ + +void +__glutSetWindow(GLUTwindow * window) +{ + /* It is tempting to try to short-circuit the call to + glXMakeCurrent if we "know" we are going to make current + to a window we are already current to. In fact, this + assumption breaks when GLUT is expected to integrated with + other OpenGL windowing APIs that also make current to + OpenGL contexts. Since glXMakeCurrent short-circuits the + "already bound" case, GLUT avoids the temptation to do so + too. */ + __glutCurrentWindow = window; + + MAKE_CURRENT_LAYER(__glutCurrentWindow); + +#if !defined(_WIN32) && !defined(__OS2__) + /* We should be careful to force a finish between each + iteration through the GLUT main loop if indirect OpenGL + contexts are in use; indirect contexts tend to have much + longer latency because lots of OpenGL extension requests + can queue up in the X protocol stream. We accomplish this + by posting GLUT_FINISH_WORK to be done. */ + if (!__glutCurrentWindow->isDirect) + __glutPutOnWorkList(__glutCurrentWindow, GLUT_FINISH_WORK); +#endif + + /* If debugging is enabled, we'll want to check this window + for any OpenGL errors every iteration through the GLUT + main loop. To accomplish this, we post the + GLUT_DEBUG_WORK to be done on this window. */ + if (__glutDebug) { + __glutPutOnWorkList(__glutCurrentWindow, GLUT_DEBUG_WORK); + } +} + +/* CENTRY */ +void GLUTAPIENTRY +glutSetWindow(int win) +{ + GLUTwindow *window; + + if (win < 1 || win > __glutWindowListSize) { + __glutWarning("glutSetWindow attempted on bogus window."); + return; + } + window = __glutWindowList[win - 1]; + if (!window) { + __glutWarning("glutSetWindow attempted on bogus window."); + return; + } + __glutSetWindow(window); +} +/* ENDCENTRY */ + +static int +getUnusedWindowSlot(void) +{ + int i; + + /* Look for allocated, unused slot. */ + for (i = 0; i < __glutWindowListSize; i++) { + if (!__glutWindowList[i]) { + return i; + } + } + /* Allocate a new slot. */ + __glutWindowListSize++; + if (__glutWindowList) { + __glutWindowList = (GLUTwindow **) + realloc(__glutWindowList, + __glutWindowListSize * sizeof(GLUTwindow *)); + } else { + /* XXX Some realloc's do not correctly perform a malloc + when asked to perform a realloc on a NULL pointer, + though the ANSI C library spec requires this. */ + __glutWindowList = (GLUTwindow **) + malloc(sizeof(GLUTwindow *)); + } + if (!__glutWindowList) + __glutFatalError("out of memory."); + __glutWindowList[__glutWindowListSize - 1] = NULL; + return __glutWindowListSize - 1; +} + +static XVisualInfo * +getVisualInfoCI(unsigned int mode) +{ +#if POKA + static int bufSizeList[] = + {16, 12, 8, 4, 2, 1, 0}; + XVisualInfo *vi; + int list[32]; + int i, n = 0; + + /* Should not be looking at display mode mask if + __glutDisplayString is non-NULL. */ + assert(!__glutDisplayString); + + list[n++] = GLX_BUFFER_SIZE; + list[n++] = 1; + if (GLUT_WIND_IS_DOUBLE(mode)) { + list[n++] = GLX_DOUBLEBUFFER; + } + if (GLUT_WIND_IS_STEREO(mode)) { + list[n++] = GLX_STEREO; + } + if (GLUT_WIND_HAS_DEPTH(mode)) { + list[n++] = GLX_DEPTH_SIZE; + list[n++] = 1; + } + if (GLUT_WIND_HAS_STENCIL(mode)) { + list[n++] = GLX_STENCIL_SIZE; + list[n++] = 1; + } + list[n] = (int) None; /* terminate list */ + + /* glXChooseVisual specify GLX_BUFFER_SIZE prefers the + "smallest index buffer of at least the specified size". + This would be reasonable if GLUT allowed the user to + specify the required buffe size, but GLUT's display mode + is too simplistic (easy to use?). GLUT should try to find + the "largest". So start with a large buffer size and + shrink until we find a matching one that exists. */ + + for (i = 0; bufSizeList[i]; i++) { + /* XXX Assumes list[1] is where GLX_BUFFER_SIZE parameter + is. */ + list[1] = bufSizeList[i]; + vi = glXChooseVisual(__glutDisplay, + __glutScreen, list); + if (vi) + return vi; + } + return NULL; +#else + return + glXChooseVisual(mode); + +#endif +} + +static XVisualInfo * +getVisualInfoRGB(unsigned int mode) +{ +#if POKA + int list[32]; + int n = 0; + + /* Should not be looking at display mode mask if + __glutDisplayString is non-NULL. */ + assert(!__glutDisplayString); + + /* XXX Would a caching mechanism to minize the calls to + glXChooseVisual? You'd have to reference count + XVisualInfo* pointers. Would also have to properly + interact with glutInitDisplayString. */ + + list[n++] = GLX_RGBA; + list[n++] = GLX_RED_SIZE; + list[n++] = 1; + list[n++] = GLX_GREEN_SIZE; + list[n++] = 1; + list[n++] = GLX_BLUE_SIZE; + list[n++] = 1; + if (GLUT_WIND_HAS_ALPHA(mode)) { + list[n++] = GLX_ALPHA_SIZE; + list[n++] = 1; + } + if (GLUT_WIND_IS_DOUBLE(mode)) { + list[n++] = GLX_DOUBLEBUFFER; + } + if (GLUT_WIND_IS_STEREO(mode)) { + list[n++] = GLX_STEREO; + } + if (GLUT_WIND_HAS_DEPTH(mode)) { + list[n++] = GLX_DEPTH_SIZE; + list[n++] = 1; + } + if (GLUT_WIND_HAS_STENCIL(mode)) { + list[n++] = GLX_STENCIL_SIZE; + list[n++] = 1; + } + if (GLUT_WIND_HAS_ACCUM(mode)) { + list[n++] = GLX_ACCUM_RED_SIZE; + list[n++] = 1; + list[n++] = GLX_ACCUM_GREEN_SIZE; + list[n++] = 1; + list[n++] = GLX_ACCUM_BLUE_SIZE; + list[n++] = 1; + if (GLUT_WIND_HAS_ALPHA(mode)) { + list[n++] = GLX_ACCUM_ALPHA_SIZE; + list[n++] = 1; + } + } +#if defined(GLX_VERSION_1_1) && (defined(GLX_SGIS_multisample) || defined(GLX_ARB_multisample)) + if (GLUT_WIND_IS_MULTISAMPLE(mode)) { + if (!__glutIsSupportedByGLX("GLX_SGIS_multisample") && + !__glutIsSupportedByGLX("GLX_ARB_multisample")) + return NULL; +#if defined(GLX_ARB_multisample) + list[n++] = GLX_SAMPLES_ARB; +#elif defined(GLX_SGIS_multisample) + list[n++] = GLX_SAMPLES_SGIS; +#endif + /* XXX Is 4 a reasonable minimum acceptable number of + samples? */ + list[n++] = 4; + } +#endif + list[n] = (int) None; /* terminate list */ + + return glXChooseVisual(__glutDisplay, + __glutScreen, list); +#else /* POKA */ + + return + glXChooseVisual(mode); + +#endif +} + +XVisualInfo * +__glutGetVisualInfo(unsigned int mode) +{ + /* XXX GLUT_LUMINANCE not implemented for GLUT 3.0. */ + if (GLUT_WIND_IS_LUMINANCE(mode)) + return NULL; + + if (GLUT_WIND_IS_RGB(mode)) + return getVisualInfoRGB(mode); + else + return getVisualInfoCI(mode); +} + +XVisualInfo * +__glutDetermineVisual( + unsigned int displayMode, + Bool * treatAsSingle, + XVisualInfo * (getVisualInfo) (unsigned int)) +{ + XVisualInfo *vis; + + /* Should not be looking at display mode mask if + __glutDisplayString is non-NULL. */ + assert(!__glutDisplayString); + + *treatAsSingle = GLUT_WIND_IS_SINGLE(displayMode); + vis = getVisualInfo(displayMode); + if (!vis) { + /* Fallback cases when can't get exactly what was asked + for... */ + if (GLUT_WIND_IS_SINGLE(displayMode)) { + /* If we can't find a single buffered visual, try looking + for a double buffered visual. We can treat a double + buffered visual as a single buffer visual by changing + the draw buffer to GL_FRONT and treating any swap + buffers as no-ops. */ + displayMode |= GLUT_DOUBLE; + vis = getVisualInfo(displayMode); + *treatAsSingle = True; + } + if (!vis && GLUT_WIND_IS_MULTISAMPLE(displayMode)) { + /* If we can't seem to get multisampling (ie, not Reality + Engine class graphics!), go without multisampling. It + is up to the application to query how many multisamples + were allocated (0 equals no multisampling) if the + application is going to use multisampling for more than + just antialiasing. */ + displayMode &= ~GLUT_MULTISAMPLE; + vis = getVisualInfo(displayMode); + } + } + return vis; +} + +static void GLUTCALLBACK +__glutDefaultDisplay(void) +{ + /* XXX Remove the warning after GLUT 3.0. */ + __glutWarning("The following is a new check for GLUT 3.0; update your code."); + __glutFatalError( + "redisplay needed for window %d, but no display callback.", + __glutCurrentWindow->num + 1); +} + +void GLUTCALLBACK +__glutDefaultReshape(int width, int height) +{ + GLUToverlay *overlay; + + /* Adjust the viewport of the window (and overlay if one + exists). */ + MAKE_CURRENT_WINDOW(__glutCurrentWindow); + glViewport(0, 0, (GLsizei) width, (GLsizei) height); + overlay = __glutCurrentWindow->overlay; + if (overlay) { + MAKE_CURRENT_OVERLAY(overlay); + glViewport(0, 0, (GLsizei) width, (GLsizei) height); + } + /* Make sure we are current to the current layer (application + should be able to count on the current layer not changing + unless the application explicitly calls glutUseLayer). */ + MAKE_CURRENT_LAYER(__glutCurrentWindow); +} + +XVisualInfo * +__glutDetermineWindowVisual(Bool * treatAsSingle, Bool * visAlloced, void **fbc) +{ + if (__glutDisplayString) { + + /* __glutDisplayString should be NULL except if + glutInitDisplayString has been called to register a + different display string. Calling glutInitDisplayString + means using a string instead of an integer mask determine + the visual to use. Using the function pointer variable + __glutDetermineVisualFromString below avoids linking in + the code for implementing glutInitDisplayString (ie, + glut_dstr.o) unless glutInitDisplayString gets called by + the application. */ + + assert(__glutDetermineVisualFromString); + *visAlloced = False; + *fbc = NULL; + return __glutDetermineVisualFromString(__glutDisplayString, treatAsSingle, + requiredWindowCriteria, numRequiredWindowCriteria, requiredWindowCriteriaMask, fbc); + } else { + *visAlloced = True; + *fbc = NULL; + return __glutDetermineVisual(__glutDisplayMode, + treatAsSingle, __glutGetVisualInfo); + } +} + +/* ARGSUSED5 */ /* Only Win32 uses gameMode parameter. */ +GLUTwindow * +__glutCreateWindow(GLUTwindow * parent, + int x, int y, int width, int height, int gameMode) +{ + GLUTwindow *window; + XSetWindowAttributes wa; + unsigned long attribMask; + int winnum; + int i; +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_fbconfig) + GLXFBConfigSGIX fbc; +#else + void *fbc; +#endif + +#if defined(__OS2PM__) + { + extern HAB hab; /* PM anchor block handle */ + CLASSINFO classinfo; + + if(!WinQueryClassInfo(hab,"GLUT", &classinfo) ) + __glutOpenOS2Connection(NULL); + } +#elif defined(_WIN32) + WNDCLASS wc; + int style; + + if (!GetClassInfo(GetModuleHandle(NULL), "GLUT", &wc)) { + __glutOpenWin32Connection(NULL); + } +#else + if (!__glutDisplay) { + __glutOpenXConnection(NULL); + } +#endif + +#ifndef __OS2PM__ + if (__glutGameModeWindow) { + __glutFatalError("cannot create windows in game mode."); + } +#endif + + winnum = getUnusedWindowSlot(); + window = (GLUTwindow *) malloc(sizeof(GLUTwindow)); + if (!window) { + __glutFatalError("out of memory."); + } + window->num = winnum; + +#if defined(__OS2PM__) + /* Add this new window to the window list. */ + __glutWindowList[winnum] = window; + window->shownState = -1; +#endif + +#if !defined(_WIN32) && !defined(__OS2PM__) + window->vis = __glutDetermineWindowVisual(&window->treatAsSingle, + &window->visAlloced, (void**) &fbc); + if (!window->vis) { + __glutFatalError( + "visual with necessary capabilities not found."); + } + __glutSetupColormap(window->vis, &window->colormap, &window->cmap); +#endif + window->eventMask = StructureNotifyMask | ExposureMask; + + attribMask = CWBackPixmap | CWBorderPixel | CWColormap | CWEventMask; + wa.background_pixmap = None; + wa.border_pixel = 0; + wa.colormap = window->cmap; + wa.event_mask = window->eventMask; + if (parent) { + if (parent->eventMask & GLUT_HACK_STOP_PROPAGATE_MASK) + wa.event_mask |= GLUT_HACK_STOP_PROPAGATE_MASK; + attribMask |= CWDontPropagate; + wa.do_not_propagate_mask = parent->eventMask & GLUT_DONT_PROPAGATE_FILTER_MASK; + } else { + wa.do_not_propagate_mask = 0; + } + + /* Stash width and height before Win32's __glutAdjustCoords + possibly overwrites the values. */ + window->width = width; + window->height = height; + window->forceReshape = True; + window->ignoreKeyRepeat = False; + +#if defined(__OS2PM__) + + { ULONG flStyle=0; + int ii; + ERRORID erridErrorCode;/* last error id code */ + extern HAB hab; /* PM anchor block handle */ + + if (parent) { + flStyle = WS_CLIPCHILDREN|WS_VISIBLE; + } else { + if (gameMode) { + /* Game mode window should be a WS_POPUP window to + ensure that the taskbar is hidden by it. A standard + WS_OVERLAPPEDWINDOW does not hide the task bar. */ + flStyle = FCF_STANDARD | WS_MAXIMIZED; + } else { + /* A standard toplevel window with borders and such. */ + flStyle = FCF_STANDARD | WS_CLIPCHILDREN; +// flStyle = WS_OVERLAPPEDWINDOW; + } + } +{ + HWND hwnd; /* Window */ + ULONG ListBoxId; /* Window id */ + /* (supplied by application) */ + + + HWND hwndClient; /* handle to the client */ + HWND hwndFrame; /* handle to the frame */ + PFNWP GenericWndProc; + FRAMECDATA fcd; + RECTL rect; /* Boundary rectangle */ + + + +/************************************************/ +// flCreate = (FCF_STANDARD) & ~FCF_TASKLIST; +/**********************************/ + if (parent) + { window->frame = NULL; + + hwnd = WinCreateWindow(parent->win, /* Parent window */ + "GLUTCHILD", /* Class name */ + "", /* Window text */ + flStyle, /* Window style */ + x, y, /* Position (x,y) */ + width, height, /* Size (width,height) */ + parent->win, /* Owner window */ + HWND_TOP, /* Sibling window */ + 0, /* Window id */ + NULL, /* Control data */ + NULL); /* Pres parameters */ + + erridErrorCode = WinGetLastError(hab); + window->win = hwnd; + + window->hdc = WinOpenWindowDC(window->win); + window->hpsBuffer = hpsCurrent; + + + rect.xLeft = x; + rect.xRight = x+width; + rect.yBottom = y; + rect.yTop = y + height; + +/***** else parent *****************************/ + } else { + hwnd = WinCreateStdWindow(HWND_DESKTOP, + 0, /* WS_VISIBLE frame-window style */ + &flStyle, /* window style */ + "GLUT", /* class name */ + "GLUT",/* window title */ + 0L, /* default client style */ + NULLHANDLE, /* resource in executable file */ + ID_WINDOW, /* resource id */ + &hwndClient); /* receives client window handle */ + + erridErrorCode = WinGetLastError(hab); + window->win = hwndClient; + window->frame = hwnd; + window->hdc = WinOpenWindowDC(window->win); + + window->hpsBuffer = hpsCurrent; + + +/* converts a client window's boundaries into an equivalent frame rectangle */ + rect.xLeft = x; + rect.xRight = x+width; + rect.yBottom = y; + rect.yTop = y + height; + + /* calculate equivalent frame boundary from boundary data */ + WinCalcFrameRect(window->frame, &rect, FALSE); + } +/***** endof if(parent) *****************************/ + + /* Must set the XHDC for fake glXChooseVisual & fake + glXCreateContext & fake XAllocColorCells. */ + XHDC = window->hdc; + XHWND = window->win; + window->vis = __glutDetermineWindowVisual(&window->treatAsSingle, + &window->visAlloced, &fbc); + if (!window->vis) + { __glutFatalError( + "pixel format with necessary capabilities not found."); + } + { int rc; + rc = wglChoosePixelFormat(window->hdc, window->vis), + +// evglSetPixelFormat(2); /* int iPixelFormat 1 - doublebuffer/2 - single buffer ??*/ + wglSetPixelFormat(window->hdc,rc,window->vis); + } + __glutSetupColormap(window->vis, &window->colormap, &window->cmap); + + window->ctx = glXCreateContext(window->hpsBuffer, window->vis, + None, __glutTryDirect); + + WinSetWindowPos(hwnd, + HWND_TOP,rect.xLeft,rect.yBottom, + rect.xRight-rect.xLeft, rect.yTop-rect.yBottom, + SWP_ACTIVATE | SWP_MOVE | SWP_SIZE | SWP_SHOW|SWP_ZORDER); /* flags*/ + + /* Make sure subwindows get a windowStatus callback. */ + if (parent) + WinPostMsg(parent->win, WM_ACTIVATE, 0, 0); + + } +} + +#elif defined(_WIN32) + + __glutAdjustCoords(parent ? parent->win : NULL, + &x, &y, &width, &height); + if (parent) { + style = WS_CHILD; + } else { + if (gameMode) { + /* Game mode window should be a WS_POPUP window to + ensure that the taskbar is hidden by it. A standard + WS_OVERLAPPEDWINDOW does not hide the task bar. */ + style = WS_POPUP | WS_MAXIMIZE; + } else { + /* A standard toplevel window with borders and such. */ + style = WS_OVERLAPPEDWINDOW; + } + } + window->win = CreateWindow("GLUT", "GLUT", + WS_CLIPSIBLINGS | WS_CLIPCHILDREN | style, + x, y, width, height, parent ? parent->win : __glutRoot, + NULL, GetModuleHandle(NULL), 0); + window->hdc = GetDC(window->win); + /* Must set the XHDC for fake glXChooseVisual & fake + glXCreateContext & fake XAllocColorCells. */ + XHDC = window->hdc; + window->vis = __glutDetermineWindowVisual(&window->treatAsSingle, + &window->visAlloced, &fbc); + if (!window->vis) { + __glutFatalError( + "pixel format with necessary capabilities not found."); + } + if (!SetPixelFormat(window->hdc, + ChoosePixelFormat(window->hdc, window->vis), + window->vis)) { + __glutFatalError("SetPixelFormat failed during window create."); + } + __glutSetupColormap(window->vis, &window->colormap, &window->cmap); + /* Make sure subwindows get a windowStatus callback. */ + if (parent) { + PostMessage(parent->win, WM_ACTIVATE, 0, 0); + } + window->renderDc = window->hdc; +#else + window->win = XCreateWindow(__glutDisplay, + parent == NULL ? __glutRoot : parent->win, + x, y, width, height, 0, + window->vis->depth, InputOutput, window->vis->visual, + attribMask, &wa); +#endif + window->renderWin = window->win; +#if defined(GLX_VERSION_1_1) && defined(GLX_SGIX_fbconfig) + if (fbc) { + window->ctx = __glut_glXCreateContextWithConfigSGIX(__glutDisplay, fbc, + GLX_RGBA_TYPE_SGIX, None, __glutTryDirect); + } else +#endif +#if defined(__OS2PM__) +// window->ctx = glXCreateContext(window->hpsBuffer, window->vis, +// None, __glutTryDirect); +#else + window->ctx = glXCreateContext(__glutDisplay, window->vis, + None, __glutTryDirect); +#endif + if (!window->ctx) { + __glutFatalError( + "failed to create OpenGL rendering context."); + } + window->renderCtx = window->ctx; +#if !defined(_WIN32) && !defined(__OS2PM__) + window->isDirect = glXIsDirect(__glutDisplay, window->ctx); + if (__glutForceDirect) { + if (!window->isDirect) + __glutFatalError("direct rendering not possible."); + } +#endif + + window->parent = parent; + if (parent) { + window->siblings = parent->children; + parent->children = window; + } else { + window->siblings = NULL; + } + window->overlay = NULL; + window->children = NULL; + window->display = __glutDefaultDisplay; + window->reshape = __glutDefaultReshape; + window->mouse = NULL; + window->motion = NULL; + window->passive = NULL; + window->entry = NULL; + window->keyboard = NULL; + window->keyboardUp = NULL; + window->windowStatus = NULL; + window->visibility = NULL; + window->special = NULL; + window->specialUp = NULL; + window->buttonBox = NULL; + window->dials = NULL; + window->spaceMotion = NULL; + window->spaceRotate = NULL; + window->spaceButton = NULL; + window->tabletMotion = NULL; + window->tabletButton = NULL; +#ifdef _WIN32 + window->joystick = NULL; + window->joyPollInterval = 0; +#endif + +#if defined(__OS2PM__) + window->wm_command = NULL; +#endif + + window->tabletPos[0] = -1; + window->tabletPos[1] = -1; +#if defined(__OS2PM__) + if(window->shownState == -1) + window->shownState = 0; + window->visState = window->shownState; +#else + window->shownState = 0; + window->visState = -1; /* not VisibilityUnobscured, + VisibilityPartiallyObscured, or + VisibilityFullyObscured */ +#endif + window->entryState = -1; /* not EnterNotify or LeaveNotify */ + + window->desiredConfMask = 0; + window->buttonUses = 0; + window->cursor = GLUT_CURSOR_INHERIT; + + /* Setup window to be mapped when glutMainLoop starts. */ + window->workMask = GLUT_MAP_WORK; +#ifdef _WIN32 + if (gameMode) { + /* When mapping a game mode window, just show + the window. We have already created the game + mode window with a maximize flag at creation + time. Doing a ShowWindow(window->win, SW_SHOWNORMAL) + would be wrong for a game mode window since it + would unmaximize the window. */ + window->desiredMapState = GameModeState; + } else { + window->desiredMapState = NormalState; + } +#else + window->desiredMapState = NormalState; +#endif + window->prevWorkWin = __glutWindowWorkList; + __glutWindowWorkList = window; + + /* Initially, no menus attached. */ + for (i = 0; i < GLUT_MAX_MENUS; i++) { + window->menu[i] = 0; + } + + /* Add this new window to the window list. */ + __glutWindowList[winnum] = window; + + /* Make the new window the current window. */ + __glutSetWindow(window); + + __glutDetermineMesaSwapHackSupport(); + + if (window->treatAsSingle) { + /* We do this because either the window really is single + buffered (in which case this is redundant, but harmless, + because this is the initial single-buffered context + state); or we are treating a double buffered window as a + single-buffered window because the system does not appear + to export any suitable single- buffered visuals (in which + the following are necessary). */ + glDrawBuffer(GL_FRONT); + glReadBuffer(GL_FRONT); + } + return window; +} + +/* CENTRY */ +int GLUTAPIENTRY +glutCreateWindow(const char *title) +{ + static int firstWindow = 1; + GLUTwindow *window; +#if !defined(_WIN32) && !defined(__OS2__) + XWMHints *wmHints; +#endif + Window win; + XTextProperty textprop; + + if (__glutGameModeWindow) { + __glutFatalError("cannot create windows in game mode."); + } + window = __glutCreateWindow(NULL, + __glutSizeHints.x, __glutSizeHints.y, + __glutInitWidth, __glutInitHeight, + /* not game mode */ 0); + win = window->win; + /* Setup ICCCM properties. */ + textprop.value = (unsigned char *) title; + textprop.encoding = XA_STRING; + textprop.format = 8; + textprop.nitems = strlen(title); +#if defined(__OS2__) + WinSetWindowText(window->frame, (PCSZ)title); + if (__glutIconic) { + window->desiredMapState = IconicState; + } +#elif defined(_WIN32) + SetWindowText(win, title); + if (__glutIconic) { + window->desiredMapState = IconicState; + } +#else + wmHints = XAllocWMHints(); + wmHints->initial_state = + __glutIconic ? IconicState : NormalState; + wmHints->flags = StateHint; + XSetWMProperties(__glutDisplay, win, &textprop, &textprop, + /* Only put WM_COMMAND property on first window. */ + firstWindow ? __glutArgv : NULL, + firstWindow ? __glutArgc : 0, + &__glutSizeHints, wmHints, NULL); + XFree(wmHints); + XSetWMProtocols(__glutDisplay, win, &__glutWMDeleteWindow, 1); +#endif + firstWindow = 0; + return window->num + 1; +} + +#ifdef _WIN32 +int GLUTAPIENTRY +__glutCreateWindowWithExit(const char *title, void (__cdecl *exitfunc)(int)) +{ + __glutExitFunc = exitfunc; + return glutCreateWindow(title); +} +#endif + +int GLUTAPIENTRY +glutCreateSubWindow(int win, int x, int y, int width, int height) +{ + GLUTwindow *window; + + window = __glutCreateWindow(__glutWindowList[win - 1], + x, y, width, height, /* not game mode */ 0); +#if !defined(_WIN32) && !defined(__OS2__) + { + GLUTwindow *toplevel; + + toplevel = __glutToplevelOf(window); + if (toplevel->cmap != window->cmap) { + __glutPutOnWorkList(toplevel, GLUT_COLORMAP_WORK); + } + } +#endif + return window->num + 1; +} +/* ENDCENTRY */ + +void +__glutDestroyWindow(GLUTwindow * window, + GLUTwindow * initialWindow) +{ + GLUTwindow **prev, *cur, *parent, *siblings; + + /* Recursively destroy any children. */ + cur = window->children; + while (cur) { + siblings = cur->siblings; + __glutDestroyWindow(cur, initialWindow); + cur = siblings; + } + /* Remove from parent's children list (only necessary for + non-initial windows and subwindows!). */ + parent = window->parent; + if (parent && parent == initialWindow->parent) { + prev = &parent->children; + cur = parent->children; + while (cur) { + if (cur == window) { + *prev = cur->siblings; + break; + } + prev = &(cur->siblings); + cur = cur->siblings; + } + } + /* Unbind if bound to this window. */ + if (window == __glutCurrentWindow) { + UNMAKE_CURRENT(); + __glutCurrentWindow = NULL; + } + /* Begin tearing down window itself. */ + if (window->overlay) { + __glutFreeOverlayFunc(window->overlay); + } + XDestroyWindow(__glutDisplay, window->win); + glXDestroyContext(__glutDisplay, window->ctx); + if (window->colormap) { + /* Only color index windows have colormap data structure. */ + __glutFreeColormap(window->colormap); + } + /* NULLing the __glutWindowList helps detect is a window + instance has been destroyed, given a window number. */ + __glutWindowList[window->num] = NULL; + + /* Cleanup data structures that might contain window. */ + cleanWindowWorkList(window); +#if !defined(_WIN32) && !defined(__OS2__) + cleanStaleWindowList(window); +#endif + /* Remove window from the "get window cache" if it is there. */ + if (__glutWindowCache == window) + __glutWindowCache = NULL; + + if (window->visAlloced) { + /* Only free XVisualInfo* gotten from glXChooseVisual. */ + XFree(window->vis); + } + + if (window == __glutGameModeWindow) { + /* Destroying the game mode window should implicitly + have GLUT leave game mode. */ + __glutCloseDownGameMode(); + } + + free(window); +} + +/* CENTRY */ +void GLUTAPIENTRY +glutDestroyWindow(int win) +{ + GLUTwindow *window = __glutWindowList[win - 1]; + + if (__glutMappedMenu && __glutMenuWindow == window) { + __glutFatalUsage("destroying menu window not allowed while menus in use"); + } +#if !defined(_WIN32) && !defined(__OS2__) + /* If not a toplevel window... */ + if (window->parent) { + /* Destroying subwindows may change colormap requirements; + recalculate toplevel window's WM_COLORMAP_WINDOWS + property. */ + __glutPutOnWorkList(__glutToplevelOf(window->parent), + GLUT_COLORMAP_WORK); + } +#endif + __glutDestroyWindow(window, window); + XFlush(__glutDisplay); +} +/* ENDCENTRY */ + +void +__glutChangeWindowEventMask(long eventMask, Bool add) +{ + if (add) { + /* Add eventMask to window's event mask. */ + if ((__glutCurrentWindow->eventMask & eventMask) != + eventMask) { + __glutCurrentWindow->eventMask |= eventMask; + __glutPutOnWorkList(__glutCurrentWindow, + GLUT_EVENT_MASK_WORK); + } + } else { + /* Remove eventMask from window's event mask. */ + if (__glutCurrentWindow->eventMask & eventMask) { + __glutCurrentWindow->eventMask &= ~eventMask; + __glutPutOnWorkList(__glutCurrentWindow, + GLUT_EVENT_MASK_WORK); + } + } +} + +void GLUTAPIENTRY +glutDisplayFunc(GLUTdisplayCB displayFunc) +{ + /* XXX Remove the warning after GLUT 3.0. */ + if (!displayFunc) + __glutFatalError("NULL display callback not allowed in GLUT 3.0; update your code."); + __glutCurrentWindow->display = displayFunc; +} + +void GLUTAPIENTRY +glutMouseFunc(GLUTmouseCB mouseFunc) +{ + if (__glutCurrentWindow->mouse) { + if (!mouseFunc) { + /* Previous mouseFunc being disabled. */ + __glutCurrentWindow->buttonUses--; + __glutChangeWindowEventMask( + ButtonPressMask | ButtonReleaseMask, + __glutCurrentWindow->buttonUses > 0); + } + } else { + if (mouseFunc) { + /* Previously no mouseFunc, new one being installed. */ + __glutCurrentWindow->buttonUses++; + __glutChangeWindowEventMask( + ButtonPressMask | ButtonReleaseMask, True); + } + } + __glutCurrentWindow->mouse = mouseFunc; +} + +void GLUTAPIENTRY +glutMotionFunc(GLUTmotionCB motionFunc) +{ + /* Hack. Some window managers (4Dwm by default) will mask + motion events if the client is not selecting for button + press and release events. So we select for press and + release events too (being careful to use reference + counting). */ + if (__glutCurrentWindow->motion) { + if (!motionFunc) { + /* previous mouseFunc being disabled */ + __glutCurrentWindow->buttonUses--; + __glutChangeWindowEventMask( + ButtonPressMask | ButtonReleaseMask, + __glutCurrentWindow->buttonUses > 0); + } + } else { + if (motionFunc) { + /* Previously no mouseFunc, new one being installed. */ + __glutCurrentWindow->buttonUses++; + __glutChangeWindowEventMask( + ButtonPressMask | ButtonReleaseMask, True); + } + } + /* Real work of selecting for passive mouse motion. */ + __glutChangeWindowEventMask( + Button1MotionMask | Button2MotionMask | Button3MotionMask, + motionFunc != NULL); + __glutCurrentWindow->motion = motionFunc; +} + +void GLUTAPIENTRY +glutPassiveMotionFunc(GLUTpassiveCB passiveMotionFunc) +{ + __glutChangeWindowEventMask(PointerMotionMask, + passiveMotionFunc != NULL); + + /* Passive motion also requires watching enters and leaves so + that a fake passive motion event can be generated on an + enter. */ + __glutChangeWindowEventMask(EnterWindowMask | LeaveWindowMask, + __glutCurrentWindow->entry != NULL || passiveMotionFunc != NULL); + + __glutCurrentWindow->passive = passiveMotionFunc; +} + +void GLUTAPIENTRY +glutEntryFunc(GLUTentryCB entryFunc) +{ + __glutChangeWindowEventMask(EnterWindowMask | LeaveWindowMask, + entryFunc != NULL || __glutCurrentWindow->passive); + __glutCurrentWindow->entry = entryFunc; + if (!entryFunc) { + __glutCurrentWindow->entryState = -1; + } +} + +void GLUTAPIENTRY +glutWindowStatusFunc(GLUTwindowStatusCB windowStatusFunc) +{ + __glutChangeWindowEventMask(VisibilityChangeMask, + windowStatusFunc != NULL); + __glutCurrentWindow->windowStatus = windowStatusFunc; + if (!windowStatusFunc) { + /* Make state invalid. */ + __glutCurrentWindow->visState = -1; + } +} + +static void GLUTCALLBACK +visibilityHelper(int status) +{ + if (status == GLUT_HIDDEN || status == GLUT_FULLY_COVERED) + __glutCurrentWindow->visibility(GLUT_NOT_VISIBLE); + else + __glutCurrentWindow->visibility(GLUT_VISIBLE); +} + + +void GLUTAPIENTRY +glutVisibilityFunc(GLUTvisibilityCB visibilityFunc) +{ + __glutCurrentWindow->visibility = visibilityFunc; + + if (visibilityFunc) + { glutWindowStatusFunc(visibilityHelper); +#if defined(__OS2PM__) + if(__glutCurrentWindow->shownState >= 0) + { visibilityHelper(__glutCurrentWindow->shownState); + } +#endif + } + else + glutWindowStatusFunc(NULL); +} + +void GLUTAPIENTRY +glutReshapeFunc(GLUTreshapeCB reshapeFunc) +{ + if (reshapeFunc) { + __glutCurrentWindow->reshape = reshapeFunc; + } else { + __glutCurrentWindow->reshape = __glutDefaultReshape; + } +}  \ No newline at end of file diff --git a/src/glut/os2/glut_winmisc.cpp b/src/glut/os2/glut_winmisc.cpp index ffa31c021c..456d19a8c1 100644 --- a/src/glut/os2/glut_winmisc.cpp +++ b/src/glut/os2/glut_winmisc.cpp @@ -1,127 +1,127 @@ - -/* Copyright (c) Mark J. Kilgard, 1994. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - - -#include -#include -#include -#include - - -#include "glutint.h" - -/* CENTRY */ -void GLUTAPIENTRY -glutSetWindowTitle(const char *title) -{ -#if defined(__OS2PM__) - __glutSetWindowText(__glutCurrentWindow->win, (char *)title); - -#else - XTextProperty textprop; - - assert(!__glutCurrentWindow->parent); - IGNORE_IN_GAME_MODE(); - textprop.value = (unsigned char *) title; - textprop.encoding = XA_STRING; - textprop.format = 8; - textprop.nitems = strlen(title); - XSetWMName(__glutDisplay, - __glutCurrentWindow->win, &textprop); - XFlush(__glutDisplay); -#endif -} - -void GLUTAPIENTRY -glutSetIconTitle(const char *title) -{ -#if defined(__OS2PM__) -//todo ? -#else - - XTextProperty textprop; - - assert(!__glutCurrentWindow->parent); - IGNORE_IN_GAME_MODE(); - textprop.value = (unsigned char *) title; - textprop.encoding = XA_STRING; - textprop.format = 8; - textprop.nitems = strlen(title); - XSetWMIconName(__glutDisplay, - __glutCurrentWindow->win, &textprop); - XFlush(__glutDisplay); -#endif -} - -void GLUTAPIENTRY -glutPositionWindow(int x, int y) -{ - IGNORE_IN_GAME_MODE(); - __glutCurrentWindow->desiredX = x; - __glutCurrentWindow->desiredY = y; - __glutCurrentWindow->desiredConfMask |= CWX | CWY; - __glutPutOnWorkList(__glutCurrentWindow, GLUT_CONFIGURE_WORK); -} - -void GLUTAPIENTRY -glutReshapeWindow(int w, int h) -{ - IGNORE_IN_GAME_MODE(); - if (w <= 0 || h <= 0) - __glutWarning("glutReshapeWindow: non-positive width or height not allowed"); - - __glutCurrentWindow->desiredWidth = w; - __glutCurrentWindow->desiredHeight = h; - __glutCurrentWindow->desiredConfMask |= CWWidth | CWHeight; - __glutPutOnWorkList(__glutCurrentWindow, GLUT_CONFIGURE_WORK); -} - -void GLUTAPIENTRY -glutPopWindow(void) -{ - IGNORE_IN_GAME_MODE(); - __glutCurrentWindow->desiredStack = Above; - __glutCurrentWindow->desiredConfMask |= CWStackMode; - __glutPutOnWorkList(__glutCurrentWindow, GLUT_CONFIGURE_WORK); -} - -void GLUTAPIENTRY -glutPushWindow(void) -{ - IGNORE_IN_GAME_MODE(); - __glutCurrentWindow->desiredStack = Below; - __glutCurrentWindow->desiredConfMask |= CWStackMode; - __glutPutOnWorkList(__glutCurrentWindow, GLUT_CONFIGURE_WORK); -} - -void GLUTAPIENTRY -glutIconifyWindow(void) -{ - IGNORE_IN_GAME_MODE(); - assert(!__glutCurrentWindow->parent); - __glutCurrentWindow->desiredMapState = IconicState; - __glutPutOnWorkList(__glutCurrentWindow, GLUT_MAP_WORK); -} - -void GLUTAPIENTRY -glutShowWindow(void) -{ - IGNORE_IN_GAME_MODE(); - __glutCurrentWindow->desiredMapState = NormalState; - __glutPutOnWorkList(__glutCurrentWindow, GLUT_MAP_WORK); -} - -void GLUTAPIENTRY -glutHideWindow(void) -{ - IGNORE_IN_GAME_MODE(); - __glutCurrentWindow->desiredMapState = WithdrawnState; - __glutPutOnWorkList(__glutCurrentWindow, GLUT_MAP_WORK); -} - -/* ENDCENTRY */ + +/* Copyright (c) Mark J. Kilgard, 1994. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + + +#include +#include +#include +#include + + +#include "glutint.h" + +/* CENTRY */ +void GLUTAPIENTRY +glutSetWindowTitle(const char *title) +{ +#if defined(__OS2PM__) + __glutSetWindowText(__glutCurrentWindow->win, (char *)title); + +#else + XTextProperty textprop; + + assert(!__glutCurrentWindow->parent); + IGNORE_IN_GAME_MODE(); + textprop.value = (unsigned char *) title; + textprop.encoding = XA_STRING; + textprop.format = 8; + textprop.nitems = strlen(title); + XSetWMName(__glutDisplay, + __glutCurrentWindow->win, &textprop); + XFlush(__glutDisplay); +#endif +} + +void GLUTAPIENTRY +glutSetIconTitle(const char *title) +{ +#if defined(__OS2PM__) +//todo ? +#else + + XTextProperty textprop; + + assert(!__glutCurrentWindow->parent); + IGNORE_IN_GAME_MODE(); + textprop.value = (unsigned char *) title; + textprop.encoding = XA_STRING; + textprop.format = 8; + textprop.nitems = strlen(title); + XSetWMIconName(__glutDisplay, + __glutCurrentWindow->win, &textprop); + XFlush(__glutDisplay); +#endif +} + +void GLUTAPIENTRY +glutPositionWindow(int x, int y) +{ + IGNORE_IN_GAME_MODE(); + __glutCurrentWindow->desiredX = x; + __glutCurrentWindow->desiredY = y; + __glutCurrentWindow->desiredConfMask |= CWX | CWY; + __glutPutOnWorkList(__glutCurrentWindow, GLUT_CONFIGURE_WORK); +} + +void GLUTAPIENTRY +glutReshapeWindow(int w, int h) +{ + IGNORE_IN_GAME_MODE(); + if (w <= 0 || h <= 0) + __glutWarning("glutReshapeWindow: non-positive width or height not allowed"); + + __glutCurrentWindow->desiredWidth = w; + __glutCurrentWindow->desiredHeight = h; + __glutCurrentWindow->desiredConfMask |= CWWidth | CWHeight; + __glutPutOnWorkList(__glutCurrentWindow, GLUT_CONFIGURE_WORK); +} + +void GLUTAPIENTRY +glutPopWindow(void) +{ + IGNORE_IN_GAME_MODE(); + __glutCurrentWindow->desiredStack = Above; + __glutCurrentWindow->desiredConfMask |= CWStackMode; + __glutPutOnWorkList(__glutCurrentWindow, GLUT_CONFIGURE_WORK); +} + +void GLUTAPIENTRY +glutPushWindow(void) +{ + IGNORE_IN_GAME_MODE(); + __glutCurrentWindow->desiredStack = Below; + __glutCurrentWindow->desiredConfMask |= CWStackMode; + __glutPutOnWorkList(__glutCurrentWindow, GLUT_CONFIGURE_WORK); +} + +void GLUTAPIENTRY +glutIconifyWindow(void) +{ + IGNORE_IN_GAME_MODE(); + assert(!__glutCurrentWindow->parent); + __glutCurrentWindow->desiredMapState = IconicState; + __glutPutOnWorkList(__glutCurrentWindow, GLUT_MAP_WORK); +} + +void GLUTAPIENTRY +glutShowWindow(void) +{ + IGNORE_IN_GAME_MODE(); + __glutCurrentWindow->desiredMapState = NormalState; + __glutPutOnWorkList(__glutCurrentWindow, GLUT_MAP_WORK); +} + +void GLUTAPIENTRY +glutHideWindow(void) +{ + IGNORE_IN_GAME_MODE(); + __glutCurrentWindow->desiredMapState = WithdrawnState; + __glutPutOnWorkList(__glutCurrentWindow, GLUT_MAP_WORK); +} + +/* ENDCENTRY */  \ No newline at end of file diff --git a/src/glut/os2/os2_glx.cpp b/src/glut/os2/os2_glx.cpp index ca345ea05b..5e135bc17e 100644 --- a/src/glut/os2/os2_glx.cpp +++ b/src/glut/os2/os2_glx.cpp @@ -1,146 +1,146 @@ -/* os2_glx.c */ - -#include -#include -#include -#include "gl/gl.h" -#include "WarpGL.h" -#include "GL/os2mesa.h" - -#define POKA 0 -/* global current HDC */ - -XVisualInfo *wglDescribePixelFormat(int iPixelFormat); - -extern HDC XHDC; -extern HWND XHWND; -//extern HPS hpsCurrent; -extern HAB hab; /* PM anchor block handle */ - -GLXContext -glXCreateContext(HPS hps, XVisualInfo * visinfo, - GLXContext share, Bool direct) -{ - /* KLUDGE: GLX really expects a display pointer to be passed - in as the first parameter, but Win32 needs an HDC instead, - so BE SURE that the global XHDC is set before calling this - routine. */ - HGLRC context; - - context = wglCreateContext(XHDC,hps,hab); - - - /* Since direct rendering is implicit, the direct flag is - ignored. */ - - return context; -} - - -int -glXGetConfig(XVisualInfo * visual, int attrib, int *value) -{ - if (!visual) - return GLX_BAD_VISUAL; - - switch (attrib) { - case GLX_USE_GL: - if (visual->dwFlags & (PFD_SUPPORT_OPENGL | PFD_DRAW_TO_WINDOW)) { - /* XXX Brad's Matrix Millenium II has problems creating - color index windows in 24-bit mode (lead to GDI crash) - and 32-bit mode (lead to black window). The cColorBits - filed of the PIXELFORMATDESCRIPTOR returned claims to - have 24 and 32 bits respectively of color indices. 2^24 - and 2^32 are ridiculously huge writable colormaps. - Assume that if we get back a color index - PIXELFORMATDESCRIPTOR with 24 or more bits, the - PIXELFORMATDESCRIPTOR doesn't really work and skip it. - -mjk */ - if (visual->iPixelType == PFD_TYPE_COLORINDEX - && visual->cColorBits >= 24) { - *value = 0; - } else { - *value = 1; - } - } else { - *value = 0; - } - break; - case GLX_BUFFER_SIZE: - /* KLUDGE: if we're RGBA, return the number of bits/pixel, - otherwise, return 8 (we guessed at 256 colors in CI - mode). */ - if (visual->iPixelType == PFD_TYPE_RGBA) - *value = visual->cColorBits; - else - *value = 8; - break; - case GLX_LEVEL: - /* The bReserved flag of the pfd contains the - overlay/underlay info. */ - *value = visual->bReserved; - break; - case GLX_RGBA: - *value = visual->iPixelType == PFD_TYPE_RGBA; - break; - case GLX_DOUBLEBUFFER: - *value = visual->dwFlags & PFD_DOUBLEBUFFER; - break; - case GLX_STEREO: - *value = visual->dwFlags & PFD_STEREO; - break; - case GLX_AUX_BUFFERS: - *value = visual->cAuxBuffers; - break; - case GLX_RED_SIZE: - *value = visual->cRedBits; - break; - case GLX_GREEN_SIZE: - *value = visual->cGreenBits; - break; - case GLX_BLUE_SIZE: - *value = visual->cBlueBits; - break; - case GLX_ALPHA_SIZE: - *value = visual->cAlphaBits; - break; - case GLX_DEPTH_SIZE: - *value = visual->cDepthBits; - break; - case GLX_STENCIL_SIZE: - *value = visual->cStencilBits; - break; - case GLX_ACCUM_RED_SIZE: - *value = visual->cAccumRedBits; - break; - case GLX_ACCUM_GREEN_SIZE: - *value = visual->cAccumGreenBits; - break; - case GLX_ACCUM_BLUE_SIZE: - *value = visual->cAccumBlueBits; - break; - case GLX_ACCUM_ALPHA_SIZE: - *value = visual->cAccumAlphaBits; - break; -#if POKA == 100 -#endif /* POKA == 100 */ - default: - return GLX_BAD_ATTRIB; - } - return 0; -} - - -XVisualInfo * glXChooseVisual(int mode) -{ int imode = 2; - if(mode & GLUT_DOUBLE) - imode = 1; - return - wglDescribePixelFormat(imode); -} - - -#if POKA -#endif /* POKA */ - +/* os2_glx.c */ + +#include +#include +#include +#include "gl/gl.h" +#include "WarpGL.h" +#include "GL/os2mesa.h" + +#define POKA 0 +/* global current HDC */ + +XVisualInfo *wglDescribePixelFormat(int iPixelFormat); + +extern HDC XHDC; +extern HWND XHWND; +//extern HPS hpsCurrent; +extern HAB hab; /* PM anchor block handle */ + +GLXContext +glXCreateContext(HPS hps, XVisualInfo * visinfo, + GLXContext share, Bool direct) +{ + /* KLUDGE: GLX really expects a display pointer to be passed + in as the first parameter, but Win32 needs an HDC instead, + so BE SURE that the global XHDC is set before calling this + routine. */ + HGLRC context; + + context = wglCreateContext(XHDC,hps,hab); + + + /* Since direct rendering is implicit, the direct flag is + ignored. */ + + return context; +} + + +int +glXGetConfig(XVisualInfo * visual, int attrib, int *value) +{ + if (!visual) + return GLX_BAD_VISUAL; + + switch (attrib) { + case GLX_USE_GL: + if (visual->dwFlags & (PFD_SUPPORT_OPENGL | PFD_DRAW_TO_WINDOW)) { + /* XXX Brad's Matrix Millenium II has problems creating + color index windows in 24-bit mode (lead to GDI crash) + and 32-bit mode (lead to black window). The cColorBits + filed of the PIXELFORMATDESCRIPTOR returned claims to + have 24 and 32 bits respectively of color indices. 2^24 + and 2^32 are ridiculously huge writable colormaps. + Assume that if we get back a color index + PIXELFORMATDESCRIPTOR with 24 or more bits, the + PIXELFORMATDESCRIPTOR doesn't really work and skip it. + -mjk */ + if (visual->iPixelType == PFD_TYPE_COLORINDEX + && visual->cColorBits >= 24) { + *value = 0; + } else { + *value = 1; + } + } else { + *value = 0; + } + break; + case GLX_BUFFER_SIZE: + /* KLUDGE: if we're RGBA, return the number of bits/pixel, + otherwise, return 8 (we guessed at 256 colors in CI + mode). */ + if (visual->iPixelType == PFD_TYPE_RGBA) + *value = visual->cColorBits; + else + *value = 8; + break; + case GLX_LEVEL: + /* The bReserved flag of the pfd contains the + overlay/underlay info. */ + *value = visual->bReserved; + break; + case GLX_RGBA: + *value = visual->iPixelType == PFD_TYPE_RGBA; + break; + case GLX_DOUBLEBUFFER: + *value = visual->dwFlags & PFD_DOUBLEBUFFER; + break; + case GLX_STEREO: + *value = visual->dwFlags & PFD_STEREO; + break; + case GLX_AUX_BUFFERS: + *value = visual->cAuxBuffers; + break; + case GLX_RED_SIZE: + *value = visual->cRedBits; + break; + case GLX_GREEN_SIZE: + *value = visual->cGreenBits; + break; + case GLX_BLUE_SIZE: + *value = visual->cBlueBits; + break; + case GLX_ALPHA_SIZE: + *value = visual->cAlphaBits; + break; + case GLX_DEPTH_SIZE: + *value = visual->cDepthBits; + break; + case GLX_STENCIL_SIZE: + *value = visual->cStencilBits; + break; + case GLX_ACCUM_RED_SIZE: + *value = visual->cAccumRedBits; + break; + case GLX_ACCUM_GREEN_SIZE: + *value = visual->cAccumGreenBits; + break; + case GLX_ACCUM_BLUE_SIZE: + *value = visual->cAccumBlueBits; + break; + case GLX_ACCUM_ALPHA_SIZE: + *value = visual->cAccumAlphaBits; + break; +#if POKA == 100 +#endif /* POKA == 100 */ + default: + return GLX_BAD_ATTRIB; + } + return 0; +} + + +XVisualInfo * glXChooseVisual(int mode) +{ int imode = 2; + if(mode & GLUT_DOUBLE) + imode = 1; + return + wglDescribePixelFormat(imode); +} + + +#if POKA +#endif /* POKA */ +  \ No newline at end of file diff --git a/src/glut/os2/os2_menu.cpp b/src/glut/os2/os2_menu.cpp index 4eef308e5a..56f2cdef8b 100644 --- a/src/glut/os2/os2_menu.cpp +++ b/src/glut/os2/os2_menu.cpp @@ -1,533 +1,533 @@ - -/* Copyright (c) Mark J. Kilgard, 1994, 1997, 1998. */ -/* Copyright (c) Nate Robins, 1997. */ - -/* This program is freely distributable without licensing fees - and is provided without guarantee or warrantee expressed or - implied. This program is -not- in the public domain. */ - -/* This file completely re-implements glut_menu.c and glut_menu2.c - for Win32. Note that neither glut_menu.c nor glut_menu2.c are - compiled into Win32 GLUT. */ - -#include -#include -#include -#include -#include - -#include "glutint.h" - -void (GLUTCALLBACK *__glutMenuStatusFunc) (int, int, int); -//GLUTmenu *__glutMappedMenu; -//GLUTwindow *__glutMenuWindow; -GLUTmenuItem *__glutItemSelected; -unsigned __glutMenuButton; - -static GLUTmenu **menuList = NULL; -static int menuListSize = 0; -static UINT uniqueMenuHandler = 1; - -/* DEPRICATED, use glutMenuStatusFunc instead. */ -void GLUTAPIENTRY -glutMenuStateFunc(GLUTmenuStateCB menuStateFunc) -{ - __glutMenuStatusFunc = (GLUTmenuStatusCB) menuStateFunc; -} - -void GLUTAPIENTRY -glutMenuStatusFunc(GLUTmenuStatusCB menuStatusFunc) -{ - __glutMenuStatusFunc = menuStatusFunc; -} - -void -__glutSetMenu(GLUTmenu * menu) -{ - __glutCurrentMenu = menu; -} - -static void -unmapMenu(GLUTmenu * menu) -{ - if (menu->cascade) { - unmapMenu(menu->cascade); - menu->cascade = NULL; - } - menu->anchor = NULL; - menu->highlighted = NULL; -} - -void -__glutFinishMenu(Window win, int x, int y) -{ - - unmapMenu(__glutMappedMenu); - - /* XXX Put in a GdiFlush just in case. Probably unnecessary. -mjk */ -// GdiFlush(); - - if (__glutMenuStatusFunc) { - __glutSetWindow(__glutMenuWindow); - __glutSetMenu(__glutMappedMenu); - - /* Setting __glutMappedMenu to NULL permits operations that - change menus or destroy the menu window again. */ - __glutMappedMenu = NULL; - - __glutMenuStatusFunc(GLUT_MENU_NOT_IN_USE, x, y); - } - /* Setting __glutMappedMenu to NULL permits operations that - change menus or destroy the menu window again. */ - __glutMappedMenu = NULL; - - /* If an item is selected and it is not a submenu trigger, - generate menu callback. */ - if (__glutItemSelected && !__glutItemSelected->isTrigger) { - __glutSetWindow(__glutMenuWindow); - /* When menu callback is triggered, current menu should be - set to the callback menu. */ - __glutSetMenu(__glutItemSelected->menu); - __glutItemSelected->menu->select(__glutItemSelected->value); - } - __glutMenuWindow = NULL; -} - -static void -mapMenu(GLUTmenu * menu, int x, int y) -{ -//todo -// TrackPopupMenu((HMENU) menu->win, TPM_LEFTALIGN | -// (__glutMenuButton == TPM_RIGHTBUTTON) ? TPM_RIGHTBUTTON : TPM_LEFTBUTTON, -// x, y, 0, __glutCurrentWindow->win, NULL); -} - -void -__glutStartMenu(GLUTmenu * menu, GLUTwindow * window, - int x, int y, int x_win, int y_win) -{ - assert(__glutMappedMenu == NULL); - __glutMappedMenu = menu; - __glutMenuWindow = window; - __glutItemSelected = NULL; - if (__glutMenuStatusFunc) { - __glutSetMenu(menu); - __glutSetWindow(window); - __glutMenuStatusFunc(GLUT_MENU_IN_USE, x_win, y_win); - } - mapMenu(menu, x, y); -} - -GLUTmenuItem * -__glutGetUniqueMenuItem(GLUTmenu * menu, UINT unique) -{ - GLUTmenuItem *item; - int i; - - i = menu->num; - item = menu->list; - while (item) { - if (item->unique == unique) { - return item; - } - if (item->isTrigger) { - GLUTmenuItem *subitem; - subitem = __glutGetUniqueMenuItem(menuList[item->value], unique); - if (subitem) { - return subitem; - } - } - i--; - item = item->next; - } - return NULL; -} - -GLUTmenuItem * -__glutGetMenuItem(GLUTmenu * menu, Window win, int *which) -{ - GLUTmenuItem *item; - int i; - - i = menu->num; - item = menu->list; - while (item) { - if (item->win == win) { - *which = i; - return item; - } - if (item->isTrigger) { - GLUTmenuItem *subitem; - - subitem = __glutGetMenuItem(menuList[item->value], - win, which); - if (subitem) { - return subitem; - } - } - i--; - item = item->next; - } - return NULL; -} - -GLUTmenu * -__glutGetMenu(Window win) -{ - GLUTmenu *menu; - - menu = __glutMappedMenu; - while (menu) { - if (win == menu->win) { - return menu; - } - menu = menu->cascade; - } - return NULL; -} - -GLUTmenu * -__glutGetMenuByNum(int menunum) -{ - if (menunum < 1 || menunum > menuListSize) { - return NULL; - } - return menuList[menunum - 1]; -} - -static int -getUnusedMenuSlot(void) -{ - int i; - - /* Look for allocated, unused slot. */ - for (i = 0; i < menuListSize; i++) { - if (!menuList[i]) { - return i; - } - } - /* Allocate a new slot. */ - menuListSize++; - if (menuList) { - menuList = (GLUTmenu **) - realloc(menuList, menuListSize * sizeof(GLUTmenu *)); - } else { - /* XXX Some realloc's do not correctly perform a malloc - when asked to perform a realloc on a NULL pointer, - though the ANSI C library spec requires this. */ - menuList = (GLUTmenu **) malloc(sizeof(GLUTmenu *)); - } - if (!menuList) { - __glutFatalError("out of memory."); - } - menuList[menuListSize - 1] = NULL; - return menuListSize - 1; -} - -static void -menuModificationError(void) -{ - /* XXX Remove the warning after GLUT 3.0. */ - __glutWarning("The following is a new check for GLUT 3.0; update your code."); - __glutFatalError("menu manipulation not allowed while menus in use."); -} - -int GLUTAPIENTRY -glutCreateMenu(GLUTselectCB selectFunc) -{ - GLUTmenu *menu; - int menuid; - - if (__glutMappedMenu) { - menuModificationError(); - } - menuid = getUnusedMenuSlot(); - menu = (GLUTmenu *) malloc(sizeof(GLUTmenu)); - if (!menu) { - __glutFatalError("out of memory."); - } - menu->id = menuid; - menu->num = 0; - menu->submenus = 0; - menu->select = selectFunc; - menu->list = NULL; - menu->cascade = NULL; - menu->highlighted = NULL; - menu->anchor = NULL; -//todo -// menu->win = (HWND) CreatePopupMenu(); - menuList[menuid] = menu; - __glutSetMenu(menu); - return menuid + 1; -} - - -void GLUTAPIENTRY -glutDestroyMenu(int menunum) -{ - GLUTmenu *menu = __glutGetMenuByNum(menunum); - GLUTmenuItem *item, *next; - - if (__glutMappedMenu) { - menuModificationError(); - } - assert(menu->id == menunum - 1); -//todo DestroyMenu( (HMENU) menu->win); - menuList[menunum - 1] = NULL; - /* free all menu entries */ - item = menu->list; - while (item) { - assert(item->menu == menu); - next = item->next; - free(item->label); - free(item); - item = next; - } - if (__glutCurrentMenu == menu) { - __glutCurrentMenu = NULL; - } - free(menu); -} - -int GLUTAPIENTRY -glutGetMenu(void) -{ - if (__glutCurrentMenu) { - return __glutCurrentMenu->id + 1; - } else { - return 0; - } -} - -void GLUTAPIENTRY -glutSetMenu(int menuid) -{ - GLUTmenu *menu; - - if (menuid < 1 || menuid > menuListSize) { - __glutWarning("glutSetMenu attempted on bogus menu."); - return; - } - menu = menuList[menuid - 1]; - if (!menu) { - __glutWarning("glutSetMenu attempted on bogus menu."); - return; - } - __glutSetMenu(menu); -} - -static void -setMenuItem(GLUTmenuItem * item, const char *label, - int value, Bool isTrigger) -{ - GLUTmenu *menu; - - menu = item->menu; - item->label = __glutStrdup(label); - if (!item->label) { - __glutFatalError("out of memory."); - } - item->isTrigger = isTrigger; - item->len = (int) strlen(label); - item->value = value; - item->unique = uniqueMenuHandler++; -//todo -// if (isTrigger) { -// AppendMenu((HMENU) menu->win, MF_POPUP, (UINT)item->win, label); -// } else { -// AppendMenu((HMENU) menu->win, MF_STRING, item->unique, label); -// } -} - -void GLUTAPIENTRY -glutAddMenuEntry(const char *label, int value) -{ - GLUTmenuItem *entry; - - if (__glutMappedMenu) { - menuModificationError(); - } - entry = (GLUTmenuItem *) malloc(sizeof(GLUTmenuItem)); - if (!entry) { - __glutFatalError("out of memory."); - } - entry->menu = __glutCurrentMenu; - setMenuItem(entry, label, value, FALSE); - __glutCurrentMenu->num++; - entry->next = __glutCurrentMenu->list; - __glutCurrentMenu->list = entry; -} - -void GLUTAPIENTRY -glutAddSubMenu(const char *label, int menu) -{ - GLUTmenuItem *submenu; - GLUTmenu *popupmenu; - - if (__glutMappedMenu) { - menuModificationError(); - } - submenu = (GLUTmenuItem *) malloc(sizeof(GLUTmenuItem)); - if (!submenu) { - __glutFatalError("out of memory."); - } - __glutCurrentMenu->submenus++; - submenu->menu = __glutCurrentMenu; - popupmenu = __glutGetMenuByNum(menu); - if (popupmenu) { - submenu->win = popupmenu->win; - } - setMenuItem(submenu, label, /* base 0 */ menu - 1, TRUE); - __glutCurrentMenu->num++; - submenu->next = __glutCurrentMenu->list; - __glutCurrentMenu->list = submenu; -} - -void GLUTAPIENTRY -glutChangeToMenuEntry(int num, const char *label, int value) -{ - GLUTmenuItem *item; - int i; - - if (__glutMappedMenu) { - menuModificationError(); - } - i = __glutCurrentMenu->num; - item = __glutCurrentMenu->list; - while (item) { - if (i == num) { - if (item->isTrigger) { - /* If changing a submenu trigger to a menu entry, we - need to account for submenus. */ - item->menu->submenus--; - /* Nuke the Win32 menu. */ -//todo -// DestroyMenu((HMENU) item->win); - } - free(item->label); - - item->label = strdup(label); - if (!item->label) - __glutFatalError("out of memory"); - item->isTrigger = FALSE; - item->len = (int) strlen(label); - item->value = value; - item->unique = uniqueMenuHandler++; -//todo -// ModifyMenu((HMENU) __glutCurrentMenu->win, (UINT) i - 1, -// MF_BYPOSITION | MFT_STRING, item->unique, label); - - return; - } - i--; - item = item->next; - } - __glutWarning("Current menu has no %d item.", num); -} - -void GLUTAPIENTRY -glutChangeToSubMenu(int num, const char *label, int menu) -{ - GLUTmenu *popupmenu; - GLUTmenuItem *item; - int i; - - if (__glutMappedMenu) { - menuModificationError(); - } - i = __glutCurrentMenu->num; - item = __glutCurrentMenu->list; - while (item) { - if (i == num) { - if (!item->isTrigger) { - /* If changing a menu entry to as submenu trigger, we - need to account for submenus. */ - item->menu->submenus++; -//todo -// item->win = (HWND) CreatePopupMenu(); - } - free(item->label); - - item->label = strdup(label); - if (!item->label) - __glutFatalError("out of memory"); - item->isTrigger = TRUE; - item->len = (int) strlen(label); - item->value = menu - 1; - item->unique = uniqueMenuHandler++; - popupmenu = __glutGetMenuByNum(menu); - if (popupmenu) - item->win = popupmenu->win; -//todo -// ModifyMenu((HMENU) __glutCurrentMenu->win, (UINT) i - 1, -// MF_BYPOSITION | MF_POPUP, (UINT) item->win, label); - return; - } - i--; - item = item->next; - } - __glutWarning("Current menu has no %d item.", num); -} - -void GLUTAPIENTRY -glutRemoveMenuItem(int num) -{ - GLUTmenuItem *item, **prev; - int i; - - if (__glutMappedMenu) { - menuModificationError(); - } - i = __glutCurrentMenu->num; - prev = &__glutCurrentMenu->list; - item = __glutCurrentMenu->list; - while (item) { - if (i == num) { - /* Found the menu item in list to remove. */ - __glutCurrentMenu->num--; - - /* Patch up menu's item list. */ - *prev = item->next; -//todo -// RemoveMenu((HMENU) __glutCurrentMenu->win, (UINT) i - 1, MF_BYPOSITION); - - free(item->label); - free(item); - return; - } - i--; - prev = &item->next; - item = item->next; - } - __glutWarning("Current menu has no %d item.", num); -} - -void GLUTAPIENTRY -glutAttachMenu(int button) -{ - if (__glutCurrentWindow == __glutGameModeWindow) { - __glutWarning("cannot attach menus in game mode."); - return; - } - if (__glutMappedMenu) { - menuModificationError(); - } - if (__glutCurrentWindow->menu[button] < 1) { - __glutCurrentWindow->buttonUses++; - } - __glutCurrentWindow->menu[button] = __glutCurrentMenu->id + 1; -} - -void GLUTAPIENTRY -glutDetachMenu(int button) -{ - if (__glutMappedMenu) { - menuModificationError(); - } - if (__glutCurrentWindow->menu[button] > 0) { - __glutCurrentWindow->buttonUses--; - __glutCurrentWindow->menu[button] = 0; - } -} - + +/* Copyright (c) Mark J. Kilgard, 1994, 1997, 1998. */ +/* Copyright (c) Nate Robins, 1997. */ + +/* This program is freely distributable without licensing fees + and is provided without guarantee or warrantee expressed or + implied. This program is -not- in the public domain. */ + +/* This file completely re-implements glut_menu.c and glut_menu2.c + for Win32. Note that neither glut_menu.c nor glut_menu2.c are + compiled into Win32 GLUT. */ + +#include +#include +#include +#include +#include + +#include "glutint.h" + +void (GLUTCALLBACK *__glutMenuStatusFunc) (int, int, int); +//GLUTmenu *__glutMappedMenu; +//GLUTwindow *__glutMenuWindow; +GLUTmenuItem *__glutItemSelected; +unsigned __glutMenuButton; + +static GLUTmenu **menuList = NULL; +static int menuListSize = 0; +static UINT uniqueMenuHandler = 1; + +/* DEPRICATED, use glutMenuStatusFunc instead. */ +void GLUTAPIENTRY +glutMenuStateFunc(GLUTmenuStateCB menuStateFunc) +{ + __glutMenuStatusFunc = (GLUTmenuStatusCB) menuStateFunc; +} + +void GLUTAPIENTRY +glutMenuStatusFunc(GLUTmenuStatusCB menuStatusFunc) +{ + __glutMenuStatusFunc = menuStatusFunc; +} + +void +__glutSetMenu(GLUTmenu * menu) +{ + __glutCurrentMenu = menu; +} + +static void +unmapMenu(GLUTmenu * menu) +{ + if (menu->cascade) { + unmapMenu(menu->cascade); + menu->cascade = NULL; + } + menu->anchor = NULL; + menu->highlighted = NULL; +} + +void +__glutFinishMenu(Window win, int x, int y) +{ + + unmapMenu(__glutMappedMenu); + + /* XXX Put in a GdiFlush just in case. Probably unnecessary. -mjk */ +// GdiFlush(); + + if (__glutMenuStatusFunc) { + __glutSetWindow(__glutMenuWindow); + __glutSetMenu(__glutMappedMenu); + + /* Setting __glutMappedMenu to NULL permits operations that + change menus or destroy the menu window again. */ + __glutMappedMenu = NULL; + + __glutMenuStatusFunc(GLUT_MENU_NOT_IN_USE, x, y); + } + /* Setting __glutMappedMenu to NULL permits operations that + change menus or destroy the menu window again. */ + __glutMappedMenu = NULL; + + /* If an item is selected and it is not a submenu trigger, + generate menu callback. */ + if (__glutItemSelected && !__glutItemSelected->isTrigger) { + __glutSetWindow(__glutMenuWindow); + /* When menu callback is triggered, current menu should be + set to the callback menu. */ + __glutSetMenu(__glutItemSelected->menu); + __glutItemSelected->menu->select(__glutItemSelected->value); + } + __glutMenuWindow = NULL; +} + +static void +mapMenu(GLUTmenu * menu, int x, int y) +{ +//todo +// TrackPopupMenu((HMENU) menu->win, TPM_LEFTALIGN | +// (__glutMenuButton == TPM_RIGHTBUTTON) ? TPM_RIGHTBUTTON : TPM_LEFTBUTTON, +// x, y, 0, __glutCurrentWindow->win, NULL); +} + +void +__glutStartMenu(GLUTmenu * menu, GLUTwindow * window, + int x, int y, int x_win, int y_win) +{ + assert(__glutMappedMenu == NULL); + __glutMappedMenu = menu; + __glutMenuWindow = window; + __glutItemSelected = NULL; + if (__glutMenuStatusFunc) { + __glutSetMenu(menu); + __glutSetWindow(window); + __glutMenuStatusFunc(GLUT_MENU_IN_USE, x_win, y_win); + } + mapMenu(menu, x, y); +} + +GLUTmenuItem * +__glutGetUniqueMenuItem(GLUTmenu * menu, UINT unique) +{ + GLUTmenuItem *item; + int i; + + i = menu->num; + item = menu->list; + while (item) { + if (item->unique == unique) { + return item; + } + if (item->isTrigger) { + GLUTmenuItem *subitem; + subitem = __glutGetUniqueMenuItem(menuList[item->value], unique); + if (subitem) { + return subitem; + } + } + i--; + item = item->next; + } + return NULL; +} + +GLUTmenuItem * +__glutGetMenuItem(GLUTmenu * menu, Window win, int *which) +{ + GLUTmenuItem *item; + int i; + + i = menu->num; + item = menu->list; + while (item) { + if (item->win == win) { + *which = i; + return item; + } + if (item->isTrigger) { + GLUTmenuItem *subitem; + + subitem = __glutGetMenuItem(menuList[item->value], + win, which); + if (subitem) { + return subitem; + } + } + i--; + item = item->next; + } + return NULL; +} + +GLUTmenu * +__glutGetMenu(Window win) +{ + GLUTmenu *menu; + + menu = __glutMappedMenu; + while (menu) { + if (win == menu->win) { + return menu; + } + menu = menu->cascade; + } + return NULL; +} + +GLUTmenu * +__glutGetMenuByNum(int menunum) +{ + if (menunum < 1 || menunum > menuListSize) { + return NULL; + } + return menuList[menunum - 1]; +} + +static int +getUnusedMenuSlot(void) +{ + int i; + + /* Look for allocated, unused slot. */ + for (i = 0; i < menuListSize; i++) { + if (!menuList[i]) { + return i; + } + } + /* Allocate a new slot. */ + menuListSize++; + if (menuList) { + menuList = (GLUTmenu **) + realloc(menuList, menuListSize * sizeof(GLUTmenu *)); + } else { + /* XXX Some realloc's do not correctly perform a malloc + when asked to perform a realloc on a NULL pointer, + though the ANSI C library spec requires this. */ + menuList = (GLUTmenu **) malloc(sizeof(GLUTmenu *)); + } + if (!menuList) { + __glutFatalError("out of memory."); + } + menuList[menuListSize - 1] = NULL; + return menuListSize - 1; +} + +static void +menuModificationError(void) +{ + /* XXX Remove the warning after GLUT 3.0. */ + __glutWarning("The following is a new check for GLUT 3.0; update your code."); + __glutFatalError("menu manipulation not allowed while menus in use."); +} + +int GLUTAPIENTRY +glutCreateMenu(GLUTselectCB selectFunc) +{ + GLUTmenu *menu; + int menuid; + + if (__glutMappedMenu) { + menuModificationError(); + } + menuid = getUnusedMenuSlot(); + menu = (GLUTmenu *) malloc(sizeof(GLUTmenu)); + if (!menu) { + __glutFatalError("out of memory."); + } + menu->id = menuid; + menu->num = 0; + menu->submenus = 0; + menu->select = selectFunc; + menu->list = NULL; + menu->cascade = NULL; + menu->highlighted = NULL; + menu->anchor = NULL; +//todo +// menu->win = (HWND) CreatePopupMenu(); + menuList[menuid] = menu; + __glutSetMenu(menu); + return menuid + 1; +} + + +void GLUTAPIENTRY +glutDestroyMenu(int menunum) +{ + GLUTmenu *menu = __glutGetMenuByNum(menunum); + GLUTmenuItem *item, *next; + + if (__glutMappedMenu) { + menuModificationError(); + } + assert(menu->id == menunum - 1); +//todo DestroyMenu( (HMENU) menu->win); + menuList[menunum - 1] = NULL; + /* free all menu entries */ + item = menu->list; + while (item) { + assert(item->menu == menu); + next = item->next; + free(item->label); + free(item); + item = next; + } + if (__glutCurrentMenu == menu) { + __glutCurrentMenu = NULL; + } + free(menu); +} + +int GLUTAPIENTRY +glutGetMenu(void) +{ + if (__glutCurrentMenu) { + return __glutCurrentMenu->id + 1; + } else { + return 0; + } +} + +void GLUTAPIENTRY +glutSetMenu(int menuid) +{ + GLUTmenu *menu; + + if (menuid < 1 || menuid > menuListSize) { + __glutWarning("glutSetMenu attempted on bogus menu."); + return; + } + menu = menuList[menuid - 1]; + if (!menu) { + __glutWarning("glutSetMenu attempted on bogus menu."); + return; + } + __glutSetMenu(menu); +} + +static void +setMenuItem(GLUTmenuItem * item, const char *label, + int value, Bool isTrigger) +{ + GLUTmenu *menu; + + menu = item->menu; + item->label = __glutStrdup(label); + if (!item->label) { + __glutFatalError("out of memory."); + } + item->isTrigger = isTrigger; + item->len = (int) strlen(label); + item->value = value; + item->unique = uniqueMenuHandler++; +//todo +// if (isTrigger) { +// AppendMenu((HMENU) menu->win, MF_POPUP, (UINT)item->win, label); +// } else { +// AppendMenu((HMENU) menu->win, MF_STRING, item->unique, label); +// } +} + +void GLUTAPIENTRY +glutAddMenuEntry(const char *label, int value) +{ + GLUTmenuItem *entry; + + if (__glutMappedMenu) { + menuModificationError(); + } + entry = (GLUTmenuItem *) malloc(sizeof(GLUTmenuItem)); + if (!entry) { + __glutFatalError("out of memory."); + } + entry->menu = __glutCurrentMenu; + setMenuItem(entry, label, value, FALSE); + __glutCurrentMenu->num++; + entry->next = __glutCurrentMenu->list; + __glutCurrentMenu->list = entry; +} + +void GLUTAPIENTRY +glutAddSubMenu(const char *label, int menu) +{ + GLUTmenuItem *submenu; + GLUTmenu *popupmenu; + + if (__glutMappedMenu) { + menuModificationError(); + } + submenu = (GLUTmenuItem *) malloc(sizeof(GLUTmenuItem)); + if (!submenu) { + __glutFatalError("out of memory."); + } + __glutCurrentMenu->submenus++; + submenu->menu = __glutCurrentMenu; + popupmenu = __glutGetMenuByNum(menu); + if (popupmenu) { + submenu->win = popupmenu->win; + } + setMenuItem(submenu, label, /* base 0 */ menu - 1, TRUE); + __glutCurrentMenu->num++; + submenu->next = __glutCurrentMenu->list; + __glutCurrentMenu->list = submenu; +} + +void GLUTAPIENTRY +glutChangeToMenuEntry(int num, const char *label, int value) +{ + GLUTmenuItem *item; + int i; + + if (__glutMappedMenu) { + menuModificationError(); + } + i = __glutCurrentMenu->num; + item = __glutCurrentMenu->list; + while (item) { + if (i == num) { + if (item->isTrigger) { + /* If changing a submenu trigger to a menu entry, we + need to account for submenus. */ + item->menu->submenus--; + /* Nuke the Win32 menu. */ +//todo +// DestroyMenu((HMENU) item->win); + } + free(item->label); + + item->label = strdup(label); + if (!item->label) + __glutFatalError("out of memory"); + item->isTrigger = FALSE; + item->len = (int) strlen(label); + item->value = value; + item->unique = uniqueMenuHandler++; +//todo +// ModifyMenu((HMENU) __glutCurrentMenu->win, (UINT) i - 1, +// MF_BYPOSITION | MFT_STRING, item->unique, label); + + return; + } + i--; + item = item->next; + } + __glutWarning("Current menu has no %d item.", num); +} + +void GLUTAPIENTRY +glutChangeToSubMenu(int num, const char *label, int menu) +{ + GLUTmenu *popupmenu; + GLUTmenuItem *item; + int i; + + if (__glutMappedMenu) { + menuModificationError(); + } + i = __glutCurrentMenu->num; + item = __glutCurrentMenu->list; + while (item) { + if (i == num) { + if (!item->isTrigger) { + /* If changing a menu entry to as submenu trigger, we + need to account for submenus. */ + item->menu->submenus++; +//todo +// item->win = (HWND) CreatePopupMenu(); + } + free(item->label); + + item->label = strdup(label); + if (!item->label) + __glutFatalError("out of memory"); + item->isTrigger = TRUE; + item->len = (int) strlen(label); + item->value = menu - 1; + item->unique = uniqueMenuHandler++; + popupmenu = __glutGetMenuByNum(menu); + if (popupmenu) + item->win = popupmenu->win; +//todo +// ModifyMenu((HMENU) __glutCurrentMenu->win, (UINT) i - 1, +// MF_BYPOSITION | MF_POPUP, (UINT) item->win, label); + return; + } + i--; + item = item->next; + } + __glutWarning("Current menu has no %d item.", num); +} + +void GLUTAPIENTRY +glutRemoveMenuItem(int num) +{ + GLUTmenuItem *item, **prev; + int i; + + if (__glutMappedMenu) { + menuModificationError(); + } + i = __glutCurrentMenu->num; + prev = &__glutCurrentMenu->list; + item = __glutCurrentMenu->list; + while (item) { + if (i == num) { + /* Found the menu item in list to remove. */ + __glutCurrentMenu->num--; + + /* Patch up menu's item list. */ + *prev = item->next; +//todo +// RemoveMenu((HMENU) __glutCurrentMenu->win, (UINT) i - 1, MF_BYPOSITION); + + free(item->label); + free(item); + return; + } + i--; + prev = &item->next; + item = item->next; + } + __glutWarning("Current menu has no %d item.", num); +} + +void GLUTAPIENTRY +glutAttachMenu(int button) +{ + if (__glutCurrentWindow == __glutGameModeWindow) { + __glutWarning("cannot attach menus in game mode."); + return; + } + if (__glutMappedMenu) { + menuModificationError(); + } + if (__glutCurrentWindow->menu[button] < 1) { + __glutCurrentWindow->buttonUses++; + } + __glutCurrentWindow->menu[button] = __glutCurrentMenu->id + 1; +} + +void GLUTAPIENTRY +glutDetachMenu(int button) +{ + if (__glutMappedMenu) { + menuModificationError(); + } + if (__glutCurrentWindow->menu[button] > 0) { + __glutCurrentWindow->buttonUses--; + __glutCurrentWindow->menu[button] = 0; + } +} +  \ No newline at end of file diff --git a/src/glut/os2/os2_winproc.cpp b/src/glut/os2/os2_winproc.cpp index e2d4ba9d55..6ffe0d4624 100644 --- a/src/glut/os2/os2_winproc.cpp +++ b/src/glut/os2/os2_winproc.cpp @@ -1,1297 +1,1297 @@ -/* os2_winproc.c */ - - -#define INCL_DEV -#include "WarpGL.h" -#include "GL/os2mesa.h" - - -#define _MEERROR_H_ -#include /* It is from MMPM toolkit */ -#include -#include - - -#include "os2mesadef.h" -#include "glutint.h" - - -#define POKA 0 - -#if POKA - -extern unsigned __glutMenuButton; -extern GLUTidleCB __glutIdleFunc; -extern GLUTtimer *__glutTimerList; -extern void handleTimeouts(void); -extern GLUTmenuItem *__glutGetUniqueMenuItem(GLUTmenu * menu, int unique); -static HMENU __glutHMenu; - -#endif - -extern void _mesa_ResizeBuffersMESA( void ); - - -MRESULT EXPENTRY GlutWindowProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ); -MRESULT EXPENTRY GlutWindowChildProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ); -void updateWindowState(GLUTwindow *window, int visState); - -volatile extern HAB hab; /* PM anchor block handle */ -volatile extern HPS hpsCurrent; - -RECTL rCtls[52]; -ULONG ulNumRcls; - -MRESULT EXPENTRY GlutWindowChildProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ) -{ MRESULT rc; - rc = GlutWindowProc(hwnd, msg, mp1, mp2 ); - return rc; -} - -MRESULT EXPENTRY GlutWindowProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ) -{ - HPS hps = NULLHANDLE; /* presentation space handle */ - GLUTwindow* window; /* GLUT window associated with message. */ - GLUTmenu* menu; /* GLUT menu associated with message. */ - RECTL rclClient; - POINTL point; - int button = -1,rc,key; - - -/* Process the message. */ - - switch( msg ) - { - case WM_CREATE: - { - SIZEL sizl = { 0L, 0L }; - LONG *alCaps; - HDC hdc; - - /*+-----------------------------------------------------------------+*/ - /*| The client window is being created. Create the semaphore to |*/ - /*| control access to the presentation space. Then create the |*/ - /*| thread that will draw the lines. |*/ - /*+-----------------------------------------------------------------+*/ - // DosCreateMutexSem( (PSZ)NULL, &hmtxPS, 0UL, FALSE ); - - hdc = WinOpenWindowDC(hwnd); - - /*+-----------------------------------------------------------------+*/ - /*| Create a non-cached presentation space. We will not release |*/ - /*| this PS, as we will be Selecting a Palette to this PS and then |*/ - /*| animating the palette. Upon releasing a PS the palette is no |*/ - /*| longer selected for obvious reasons. |*/ - /*+-----------------------------------------------------------------+*/ - hpsCurrent = GpiCreatePS( hab, - hdc, - &sizl, - PU_PELS | GPIF_DEFAULT | - GPIT_MICRO | GPIA_ASSOC ); -// DevQueryCaps( hdc, lStart, lCount, alCaps ); -// fPaletteCaps = alCaps[CAPS_ADDITIONAL_GRAPHICS] & CAPS_PALETTE_MANAGER; -// PaletteInit(3); - /* ¯¥p¥¢®¤ hpsBuffer ¢ p¥¦¨¬ RGB color table */ - - GpiCreateLogColorTable(hpsCurrent,0 ,LCOLF_RGB,0,0,NULL); - GpiSetPattern(hpsCurrent,PATSYM_SOLID); - GpiSetPatternSet(hpsCurrent,LCID_DEFAULT); - - } - break; - - return 0; - case WM_CLOSE: - WinPostMsg( hwnd, WM_QUIT, NULL, NULL ); - - return 0; - - case WM_PAINT: - window = __glutGetWindow(hwnd); - if (window) - { - PWMC ctx; -// hps = WinBeginPaint(hwnd,NULLHANDLE,&rclClient); - hps = WinBeginPaint(hwnd,NULLHANDLE,&rclClient); - // blit Dive buffer to screen. - - { - SWP swp; // Window position - POINTL pointl; // Point to offset from Desktop - - // Convert the point to offset from desktop lower left. - pointl.x = 0; - pointl.y = 0; - WinMapWindowPoints ( hwnd, HWND_DESKTOP, &pointl, 1 ); - - -// ctx = window->ctx; -// ctx->xDiveScr = pointl.x; -// ctx->yDiveScr = pointl.y; - } -// rc = DiveBlitImage (ctx->hDive, -// ctx->ulDiveBufferNumber, -// DIVE_BUFFER_SCREEN ); -// - - if (window->win == hwnd) { - __glutPostRedisplay(window, GLUT_REPAIR_WORK); - } else if (window->overlay && window->overlay->win == hwnd) { - __glutPostRedisplay(window, GLUT_OVERLAY_REPAIR_WORK); - } - WinEndPaint(hps); - } else { - - hps = WinBeginPaint(hwnd,NULLHANDLE,&rclClient); - WinFillRect(hps, &rclClient, CLR_WHITE); - WinEndPaint(hps); - } - break; - - case WM_VRNDISABLED: - -// pwinData->fDataInProcess = TRUE; -// DiveSetupBlitter ( pwinData->hDive, 0 ); -// pwinData->fVrnDisabled = TRUE; - break; - - case WM_VRNENABLED: - { HRGN hrgn; /* Region handle */ - RGNRECT rgnCtl; /* Processing control structure */ -// RECTL rCtls[52]; -// ULONG ulNumRcls; - -// pwinData->fDataInProcess = TRUE; - hps = WinGetPS ( hwnd ); - if ( !hps ) - break; - hrgn = GpiCreateRegion ( hps, 0L, NULL ); - if ( hrgn ) - { /* NOTE: If mp1 is zero, then this was just a move message. - ** Illustrate the visible region on a WM_VRNENABLE. - */ - WinQueryVisibleRegion ( hwnd, hrgn ); - rgnCtl.ircStart = 0; - rgnCtl.crc = 50; - rgnCtl.ulDirection = 1; - - /* Get the all ORed rectangles */ - if ( GpiQueryRegionRects ( hps, hrgn, NULL, - &rgnCtl, rCtls) ) - { - ulNumRcls = rgnCtl.crcReturned; - - /* Now find the window position and size, relative to parent. - */ -// WinQueryWindowPos ( pwinData->hwndClient, &pwinData->swp ); - -// rcl.xLeft = 0; -// rcl.yBottom = 0; - - /* Convert the point to offset from desktop lower left. - */ -// pointl.x = pwinData->swp.x; -// pointl.y = pwinData->swp.y; - -// WinMapWindowPoints ( pwinData->hwndFrame, -// HWND_DESKTOP, &pointl, 1 ); - -// pwinData->cxWindowPos = pointl.x; -// pwinData->cyWindowPos = pointl.y; - - } - GpiDestroyRegion( hps, hrgn ); - } - WinReleasePS( hps ); - - } - break; - - case WM_SIZE: - window = __glutGetWindow(hwnd); - if (window) - { int width,height; - width = SHORT1FROMMP(mp2); - height = SHORT2FROMMP(mp2); - if (width != window->width || height != window->height) { -#if 0 /* Win32 GLUT does not support overlays for now. */ - if (window->overlay) { - XResizeWindow(__glutDisplay, window->overlay->win, width, height); - } -#endif - window->width = width; - window->height = height; - __glutSetWindow(window); - if(width <= 0 || height <= 0) - break; - _mesa_ResizeBuffersMESA(); - - /* Do not execute OpenGL out of sequence with respect - to the SetWindowPos request! */ - window->reshape(width, height); - window->forceReshape = FALSE; - /* A reshape should be considered like posting a - repair request. */ - __glutPostRedisplay(window, GLUT_REPAIR_WORK); - } - } - return 0; - case WM_SHOW: - window = __glutGetWindow(hwnd); - if (window) { - int visState; - visState = SHORT1FROMMP( mp1 ); - updateWindowState(window, visState); - } - return 0; - - case WM_ACTIVATE: - window = __glutGetWindow(hwnd); -// /* Make sure we re-select the correct palette if needed. */ -// if (LOWORD(wParam)) { -// PostMessage(hwnd, WM_PALETTECHANGED, 0, 0); -// } - if (window) { - int visState; - visState = SHORT1FROMMP( mp1 ); - updateWindowState(window, visState); - } - return 0; - - case WM_CHAR: - { USHORT fsflags; - window = __glutGetWindow(hwnd); - if (!window) { - break; - } - fsflags = SHORT1FROMMP(mp1); -/* ?? */ - if((fsflags & KC_KEYUP) ) /* ¨£­®p¨p㥬 ®â¦ â¨¥ ª­®¯ª¨, p¥ £¨p㥬 ⮫쪮 ­  ­ ¦ â¨¥ */ - break; -/////////////////////////////////////////////////// - if(!(fsflags & KC_CHAR) ) - { - if (!(fsflags & KC_VIRTUALKEY)) - break; - key = 0; - /* Get the virtual key from mp2. */ - switch (SHORT2FROMMP(mp2)) - { -/* directional keys */ - case VK_LEFT: key = GLUT_KEY_LEFT; break; - case VK_UP: key = GLUT_KEY_UP; break; - case VK_RIGHT: key = GLUT_KEY_RIGHT; break; - case VK_DOWN: key = GLUT_KEY_DOWN; break; - - case VK_PAGEUP: key = GLUT_KEY_PAGE_UP; break; - case VK_PAGEDOWN:key = GLUT_KEY_PAGE_DOWN; break; - case VK_HOME: key = GLUT_KEY_HOME;break; - case VK_END: key = GLUT_KEY_END; break; - case VK_INSERT: key = GLUT_KEY_INSERT; break; - -/* function keys */ - case VK_F1 : key = GLUT_KEY_F1; break; - case VK_F2 : key = GLUT_KEY_F2; break; - case VK_F3 : key = GLUT_KEY_F3; break; - case VK_F4 : key = GLUT_KEY_F4; break; - case VK_F5 : key = GLUT_KEY_F5; break; - case VK_F6 : key = GLUT_KEY_F6; break; - case VK_F7 : key = GLUT_KEY_F7; break; - case VK_F8 : key = GLUT_KEY_F8; break; - case VK_F9 : key = GLUT_KEY_F9; break; - case VK_F10: key = GLUT_KEY_F10;break; - case VK_F11: key = GLUT_KEY_F11; break; - case VK_F12: key = GLUT_KEY_F12; break; - case VK_ESC: key = -1; break; /* Character codes */ - case VK_SPACE: key = -1; break; - case VK_TAB: key = -1; break; - } - if(!key) - { break; /* Key Not implemented */ - } - if(key > 0) - { if (!window->special) /* ­¥ ãáâ ­®¢«¥­® ®¡à ¡®â稪  */ - break; - - WinQueryPointerPos(HWND_DESKTOP,&point); - ScreenToClient(window->win, &point); - __glutSetWindow(window); - __glutModifierMask = 0; - if(WinGetKeyState(HWND_DESKTOP,VK_SHIFT) & 0x8000) /* high order bit is on */ - __glutModifierMask |= ShiftMask; - if(WinGetKeyState(HWND_DESKTOP,VK_CTRL) & 0x8000) - __glutModifierMask |= ControlMask; - if(WinGetKeyState(HWND_DESKTOP,VK_MENU) & 0x8000) - __glutModifierMask |= Mod1Mask; - window->special(key, point.x, point.y); - __glutModifierMask = (unsigned int) ~0; - return 0; - } - - } -///////////////////////////////////////////////////// - /* If we are ignoring auto repeated key strokes for the window, bail. */ - if (window->ignoreKeyRepeat && (CHAR3FROMMP(mp1)) ) - break; - if(!((unsigned char)SHORT1FROMMP(mp2)) ) /* ¨£­®p¨p㥬 ­¥á¨¬¢®«ì­ë¥ ª®¤ë */ - break; - if (window->keyboard) { - WinQueryPointerPos(HWND_DESKTOP,&point); - - ScreenToClient(window->win, &point); - __glutSetWindow(window); - __glutModifierMask = 0; - if(WinGetKeyState(HWND_DESKTOP,VK_SHIFT) & 0x8000) /* high order bit is on */ - __glutModifierMask |= ShiftMask; - if(WinGetKeyState(HWND_DESKTOP,VK_CTRL) & 0x8000) - __glutModifierMask |= ControlMask; - if(WinGetKeyState(HWND_DESKTOP,VK_MENU) & 0x8000) - __glutModifierMask |= Mod1Mask; - window->keyboard((unsigned char)SHORT1FROMMP(mp2), point.x, point.y); - __glutModifierMask = (unsigned int) ~0; - } - return 0; - } /* endof case WM_CHAR: */ -//////////////////////////////////////////////// - case WM_BUTTON1DOWN: - button = GLUT_LEFT_BUTTON; - case WM_BUTTON3DOWN: - if (button < 0) - button = GLUT_MIDDLE_BUTTON; - case WM_BUTTON2DOWN: - if (button < 0) - button = GLUT_RIGHT_BUTTON; - { POINTS psh; - psh = *((POINTS *)&mp1); - point.x = psh.x; - point.y = psh.y; - } - /* finish the menu if we get a button down message (user must have - cancelled the menu). */ - if (__glutMappedMenu) { - /* TODO: take this out once the menu on middle mouse stuff works - properly. */ - if (button == GLUT_MIDDLE_BUTTON) - return 0; - /* get current mouse pointer position */ -// WinQueryPointerPos(HWND_DESKTOP,&point); - /* map from desktop to client window */ -// WinMapWindowPoints(HWND_DESKTOP, hwnd, &point, 1); - __glutItemSelected = NULL; - __glutFinishMenu(hwnd, point.x, point.y); - return 0; - } - window = __glutGetWindow(hwnd); - if (window) { - window->buttonDownState = button+1; - menu = __glutGetMenuByNum(window->menu[button]); - if (menu) { -//todo -// __glutMenuButton = button == GLUT_RIGHT_BUTTON ? TPM_RIGHTBUTTON : -// button == GLUT_LEFT_BUTTON ? TPM_LEFTBUTTON : -// 0x0001; -// __glutStartMenu(menu, window, point.x, point.y, x, y); - } else if (window->mouse) { - - __glutSetWindow(window); - __glutModifierMask = 0; - if (WinGetKeyState(HWND_DESKTOP,VK_SHIFT) & 0x8000) /* < 0 = high order bit is on. */ - __glutModifierMask |= ShiftMask; - if (WinGetKeyState(HWND_DESKTOP,VK_CTRL) & 0x8000) - __glutModifierMask |= ControlMask; - if (WinGetKeyState(HWND_DESKTOP,VK_MENU) & 0x8000) - __glutModifierMask |= Mod1Mask; - window->mouse(button, GLUT_DOWN, point.x, point.y); - __glutModifierMask = (unsigned int)~0; - } else { - /* Stray mouse events. Ignore. */ - } - } - return 0; - - break; -/********************************************/ - case WM_BUTTON1UP: - button = GLUT_LEFT_BUTTON; - case WM_BUTTON3UP: - if (button < 0) - button = GLUT_MIDDLE_BUTTON; - case WM_BUTTON2UP: - if (button < 0) - button = GLUT_RIGHT_BUTTON; - { POINTS psh; - psh = *((POINTS *)&mp1); - point.x = psh.x; - point.y = psh.y; - } - /* Bail out if we're processing a menu. */ - /* Bail out = ¢ë¡à®á¨âìáï á ¯ à èã⮬ */ - if (__glutMappedMenu) { - WinQueryPointerPos(HWND_DESKTOP,&point); - WinMapWindowPoints(HWND_DESKTOP, hwnd, &point, 1); - /* if we're getting the middle button up signal, then something - on the menu was selected. */ - if (button == GLUT_MIDDLE_BUTTON) { - return 0; - /* For some reason, the code below always returns -1 even - though the point IS IN THE ITEM! Therefore, just bail out if - we get a middle mouse up. The user must select using the - left mouse button. Stupid Win32. */ -#if 0 - int item = MenuItemFromPoint(hwnd, __glutHMenu, point); - if (item != -1) - __glutItemSelected = (GLUTmenuItem*)GetMenuItemID(__glutHMenu, item); - else - __glutItemSelected = NULL; - __glutFinishMenu(hwnd, point.x, point.y); -#endif - } else { - __glutItemSelected = NULL; - __glutFinishMenu(hwnd, point.x, point.y); - } - return 0; - } - - window = __glutGetWindow(hwnd); - if(window) - window->buttonDownState = 0; - - if (window && window->mouse) { - __glutSetWindow(window); - __glutModifierMask = 0; - if (WinGetKeyState(HWND_DESKTOP,VK_SHIFT) & 0x8000) /* < 0 = high order bit is on */ - __glutModifierMask |= ShiftMask; - if (WinGetKeyState(HWND_DESKTOP,VK_CTRL) & 0x8000) - __glutModifierMask |= ControlMask; - if (WinGetKeyState(HWND_DESKTOP,VK_MENU) & 0x8000) - __glutModifierMask |= Mod1Mask; - window->mouse(button, GLUT_UP, point.x, point.y); - - __glutModifierMask = (unsigned int)~0; - } else { - /* Window might have been destroyed and all the - events for the window may not yet be received. */ - } - return 0; - - - break; -////////////////////////////////////////////////// - case WM_COMMAND: - window = __glutGetWindow(hwnd); - if (window) - { if (window->wm_command) - window->wm_command(hwnd,mp1,mp2); - } - break; - - case WM_MOUSEMOVE: - if (!__glutMappedMenu) { - window = __glutGetWindow(hwnd); - if (window) { - /* If motion function registered _and_ buttons held * - down, call motion function... */ - { POINTS psh; - psh = *((POINTS *)&mp1); - point.x = psh.x; - point.y = psh.y; - } - - if (window->motion && window->buttonDownState) { - __glutSetWindow(window); - window->motion(point.x, point.y); - } - /* If passive motion function registered _and_ - buttons not held down, call passive motion - function... */ - else if (window->passive && !window->buttonDownState) { - __glutSetWindow(window); - window->passive(point.x, point.y); - } - } - } else { - /* Motion events are thrown away when a pop up menu is - active. */ - } - return 0; - - - default: - /* For all other messages, let the default window procedure process them. */ - return ( WinDefWindowProc( hwnd, msg, mp1, mp2 ) ); - - } //endof switch( msg ) - return ( WinDefWindowProc( hwnd, msg, mp1, mp2 ) ); -// return NULL; -} - -void APIENTRY glutCommandFunc(GLUTcommandCB Func) -{ -extern GLUTwindow *__glutCurrentWindow; - __glutCurrentWindow->wm_command = Func; -} - - - - -void -updateWindowState(GLUTwindow *window, int visState) -{ - GLUTwindow* child; - - /* XXX shownState and visState are the same in Win32. */ - window->shownState = visState; - if (visState != window->visState) { - if (window->windowStatus) { - window->visState = visState; - __glutSetWindow(window); - window->windowStatus(visState); - } - } - /* Since Win32 only sends an activate for the toplevel window, - update the visibility for all the child windows. */ - child = window->children; - while (child) { - updateWindowState(child, visState); - child = child->siblings; - } -} - -#if POKA - -LONG WINAPI -__glutWindowProc(HWND hwnd, UINT msg, WPARAM wParam, LPARAM lParam) -{ - POINT point; /* Point structure. */ - PAINTSTRUCT ps; /* Paint structure. */ - LPMINMAXINFO minmax; /* Minimum/maximum info structure. */ - GLUTwindow* window; /* GLUT window associated with message. */ - GLUTmenu* menu; /* GLUT menu associated with message. */ - int x, y, width, height, key; - int button = -1; - - switch(msg) { - case WM_CREATE: - return 0; - case WM_CLOSE: - PostQuitMessage(0); - return 0; -#if 0 - case WM_DESTROY: - /* XXX NVidia's NT OpenGL can have problems closing down - its OpenGL internal data structures if we just allow - the process to terminate without unbinding and deleting - the windows context. Apparently, DirectDraw unloads - before OPENGL32.DLL in the close down sequence, but - NVidia's NT OpenGL needs DirectDraw to close down its - data structures. */ - window = __glutGetWindow(hwnd); - if (window) { - if (window->ctx) { - wglMakeCurrent(NULL, NULL); - wglDeleteContext(window->ctx); - } - } - return 0; -#endif - - case WM_SYSKEYUP: - case WM_KEYUP: - window = __glutGetWindow(hwnd); - if (!window) { - break; - } - /* Win32 is dumb and sends these messages only to the parent - window. Therefore, find out if we're in a child window and - call the child windows keyboard callback if we are. */ - if (window->parent) { - GetCursorPos(&point); - ScreenToClient(hwnd, &point); - hwnd = ChildWindowFromPoint(hwnd, point); - window = __glutGetWindow(hwnd); - } - if (window->specialUp || window->keyboardUp) { - GetCursorPos(&point); - ScreenToClient(window->win, &point); - __glutSetWindow(window); - __glutModifierMask = 0; - if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ - __glutModifierMask |= ShiftMask; - if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ - __glutModifierMask |= ControlMask; - if (GetKeyState(VK_MENU) < 0) - __glutModifierMask |= Mod1Mask; - switch (wParam) { - /* *INDENT-OFF* */ - case VK_F1: key = GLUT_KEY_F1; break; - case VK_F2: key = GLUT_KEY_F2; break; - case VK_F3: key = GLUT_KEY_F3; break; - case VK_F4: key = GLUT_KEY_F4; break; - case VK_F5: key = GLUT_KEY_F5; break; - case VK_F6: key = GLUT_KEY_F6; break; - case VK_F7: key = GLUT_KEY_F7; break; - case VK_F8: key = GLUT_KEY_F8; break; - case VK_F9: key = GLUT_KEY_F9; break; - case VK_F10: key = GLUT_KEY_F10; break; - case VK_F11: key = GLUT_KEY_F11; break; - case VK_F12: key = GLUT_KEY_F12; break; - case VK_LEFT: key = GLUT_KEY_LEFT; break; - case VK_UP: key = GLUT_KEY_UP; break; - case VK_RIGHT: key = GLUT_KEY_RIGHT; break; - case VK_DOWN: key = GLUT_KEY_DOWN; break; - case VK_PRIOR: key = GLUT_KEY_PAGE_UP; break; - case VK_NEXT: key = GLUT_KEY_PAGE_DOWN; break; - case VK_HOME: key = GLUT_KEY_HOME; break; - case VK_END: key = GLUT_KEY_END; break; - case VK_INSERT: key = GLUT_KEY_INSERT; break; - case VK_DELETE: - /* Delete is an ASCII character. */ - if (window->keyboardUp) { - window->keyboardUp((unsigned char) 127, point.x, point.y); - } - return 0; - /* *INDENT-ON* */ - default: - if (window->keyboardUp) { - key = MapVirtualKey(wParam, 2); /* Map to ASCII. */ - if (isascii(key) && (key != 0)) { - - /* XXX Attempt to determine modified ASCII character - is quite incomplete. Digits, symbols, CapsLock, - Ctrl, and numeric keypad are all ignored. Fix this. */ - - if (!(__glutModifierMask & ShiftMask)) - key = tolower(key); - window->keyboardUp((unsigned char) key, point.x, point.y); - } - } - __glutModifierMask = (unsigned int) ~0; - return 0; - } - if (window->specialUp) { - window->specialUp(key, point.x, point.y); - } - __glutModifierMask = (unsigned int) ~0; - } - return 0; - - case WM_SYSCHAR: - case WM_CHAR: - window = __glutGetWindow(hwnd); - if (!window) { - break; - } - - /* Bit 30 of lParam is set if key already held down. If - we are ignoring auto repeated key strokes for the window, bail. */ - if (window->ignoreKeyRepeat && (lParam & (1 << 30)) ) { - break; - } - - /* Win32 is dumb and sends these messages only to the parent - window. Therefore, find out if we're in a child window and - call the child windows keyboard callback if we are. */ - if (window->parent) { - GetCursorPos(&point); - ScreenToClient(hwnd, &point); - hwnd = ChildWindowFromPoint(hwnd, point); - window = __glutGetWindow(hwnd); - } - if (window->keyboard) { - GetCursorPos(&point); - ScreenToClient(window->win, &point); - __glutSetWindow(window); - __glutModifierMask = 0; - if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ - __glutModifierMask |= ShiftMask; - if (GetKeyState(VK_CONTROL) < 0) - __glutModifierMask |= ControlMask; - if (GetKeyState(VK_MENU) < 0) - __glutModifierMask |= Mod1Mask; - window->keyboard((unsigned char)wParam, point.x, point.y); - __glutModifierMask = (unsigned int) ~0; - } - return 0; - - case WM_SYSKEYDOWN: - case WM_KEYDOWN: - window = __glutGetWindow(hwnd); - if (!window) { - break; - } - - /* Bit 30 of lParam is set if key already held down. If - we are ignoring auto repeated key strokes for the window, bail. */ - if (window->ignoreKeyRepeat && (lParam & (1 << 30)) ) { - break; - } - - /* Win32 is dumb and sends these messages only to the parent - window. Therefore, find out if we're in a child window and - call the child windows keyboard callback if we are. */ - if (window->parent) { - GetCursorPos(&point); - ScreenToClient(hwnd, &point); - hwnd = ChildWindowFromPoint(hwnd, point); - window = __glutGetWindow(hwnd); - } - if (window->special) { - switch (wParam) { - /* *INDENT-OFF* */ - /* function keys */ - case VK_F1: key = GLUT_KEY_F1; break; - case VK_F2: key = GLUT_KEY_F2; break; - case VK_F3: key = GLUT_KEY_F3; break; - case VK_F4: key = GLUT_KEY_F4; break; - case VK_F5: key = GLUT_KEY_F5; break; - case VK_F6: key = GLUT_KEY_F6; break; - case VK_F7: key = GLUT_KEY_F7; break; - case VK_F8: key = GLUT_KEY_F8; break; - case VK_F9: key = GLUT_KEY_F9; break; - case VK_F10: key = GLUT_KEY_F10; break; - case VK_F11: key = GLUT_KEY_F11; break; - case VK_F12: key = GLUT_KEY_F12; break; - /* directional keys */ - case VK_LEFT: key = GLUT_KEY_LEFT; break; - case VK_UP: key = GLUT_KEY_UP; break; - case VK_RIGHT: key = GLUT_KEY_RIGHT; break; - case VK_DOWN: key = GLUT_KEY_DOWN; break; - /* *INDENT-ON* */ - - case VK_PRIOR: - /* VK_PRIOR is Win32's Page Up */ - key = GLUT_KEY_PAGE_UP; - break; - case VK_NEXT: - /* VK_NEXT is Win32's Page Down */ - key = GLUT_KEY_PAGE_DOWN; - break; - case VK_HOME: - key = GLUT_KEY_HOME; - break; - case VK_END: - key = GLUT_KEY_END; - break; - case VK_INSERT: - key = GLUT_KEY_INSERT; - break; - case VK_DELETE: - goto handleDelete; - default: - goto defproc; - } - GetCursorPos(&point); - ScreenToClient(window->win, &point); - __glutSetWindow(window); - __glutModifierMask = 0; - if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ - __glutModifierMask |= ShiftMask; - if (GetKeyState(VK_CONTROL) < 0) - __glutModifierMask |= ControlMask; - if (GetKeyState(VK_MENU) < 0) - __glutModifierMask |= Mod1Mask; - window->special(key, point.x, point.y); - __glutModifierMask = (unsigned int) ~0; - } else if (window->keyboard) { - /* Specially handle any keys that match ASCII values but - do not generate Windows WM_SYSCHAR or WM_CHAR messages. */ - switch (wParam) { - case VK_DELETE: - handleDelete: - /* Delete is an ASCII character. */ - GetCursorPos(&point); - ScreenToClient(window->win, &point); - __glutSetWindow(window); - __glutModifierMask = 0; - if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ - __glutModifierMask |= ShiftMask; - if (GetKeyState(VK_CONTROL) < 0) - __glutModifierMask |= ControlMask; - if (GetKeyState(VK_MENU) < 0) - __glutModifierMask |= Mod1Mask; - window->keyboard((unsigned char) 127, point.x, point.y); - __glutModifierMask = (unsigned int) ~0; - return 0; - default: - /* Let the following WM_SYSCHAR or WM_CHAR message generate - the keyboard callback. */ - break; - } - } - return 0; - - case WM_LBUTTONDOWN: - button = GLUT_LEFT_BUTTON; - case WM_MBUTTONDOWN: - if (button < 0) - button = GLUT_MIDDLE_BUTTON; - case WM_RBUTTONDOWN: - if (button < 0) - button = GLUT_RIGHT_BUTTON; - - /* finish the menu if we get a button down message (user must have - cancelled the menu). */ - if (__glutMappedMenu) { - /* TODO: take this out once the menu on middle mouse stuff works - properly. */ - if (button == GLUT_MIDDLE_BUTTON) - return 0; - GetCursorPos(&point); - ScreenToClient(hwnd, &point); - __glutItemSelected = NULL; - __glutFinishMenu(hwnd, point.x, point.y); - return 0; - } - - /* set the capture so we can get mouse events outside the window */ - SetCapture(hwnd); - - /* Win32 doesn't return the same numbers as X does when the mouse - goes beyond the upper or left side of the window. roll the - Win32's 0..2^16 pointer co-ord range to 0 +/- 2^15. */ - x = LOWORD(lParam); - y = HIWORD(lParam); - if(x & 1 << 15) x -= (1 << 16); - if(y & 1 << 15) y -= (1 << 16); - - window = __glutGetWindow(hwnd); - if (window) { - menu = __glutGetMenuByNum(window->menu[button]); - if (menu) { - point.x = LOWORD(lParam); point.y = HIWORD(lParam); - ClientToScreen(window->win, &point); - __glutMenuButton = button == GLUT_RIGHT_BUTTON ? TPM_RIGHTBUTTON : - button == GLUT_LEFT_BUTTON ? TPM_LEFTBUTTON : - 0x0001; - __glutStartMenu(menu, window, point.x, point.y, x, y); - } else if (window->mouse) { - - __glutSetWindow(window); - __glutModifierMask = 0; - if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on. */ - __glutModifierMask |= ShiftMask; - if (GetKeyState(VK_CONTROL) < 0) - __glutModifierMask |= ControlMask; - if (GetKeyState(VK_MENU) < 0) - __glutModifierMask |= Mod1Mask; - window->mouse(button, GLUT_DOWN, x, y); - __glutModifierMask = (unsigned int)~0; - } else { - /* Stray mouse events. Ignore. */ - } - } - return 0; - - case WM_LBUTTONUP: - button = GLUT_LEFT_BUTTON; - case WM_MBUTTONUP: - if (button < 0) - button = GLUT_MIDDLE_BUTTON; - case WM_RBUTTONUP: - if (button < 0) - button = GLUT_RIGHT_BUTTON; - - /* Bail out if we're processing a menu. */ - if (__glutMappedMenu) { - GetCursorPos(&point); - ScreenToClient(hwnd, &point); - /* if we're getting the middle button up signal, then something - on the menu was selected. */ - if (button == GLUT_MIDDLE_BUTTON) { - return 0; - /* For some reason, the code below always returns -1 even - though the point IS IN THE ITEM! Therefore, just bail out if - we get a middle mouse up. The user must select using the - left mouse button. Stupid Win32. */ -#if 0 - int item = MenuItemFromPoint(hwnd, __glutHMenu, point); - if (item != -1) - __glutItemSelected = (GLUTmenuItem*)GetMenuItemID(__glutHMenu, item); - else - __glutItemSelected = NULL; - __glutFinishMenu(hwnd, point.x, point.y); -#endif - } else { - __glutItemSelected = NULL; - __glutFinishMenu(hwnd, point.x, point.y); - } - return 0; - } - - /* Release the mouse capture. */ - ReleaseCapture(); - - window = __glutGetWindow(hwnd); - if (window && window->mouse) { - /* Win32 doesn't return the same numbers as X does when the - mouse goes beyond the upper or left side of the window. roll - the Win32's 0..2^16 pointer co-ord range to 0 +/- 2^15. */ - x = LOWORD(lParam); - y = HIWORD(lParam); - if(x & 1 << 15) x -= (1 << 16); - if(y & 1 << 15) y -= (1 << 16); - - __glutSetWindow(window); - __glutModifierMask = 0; - if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ - __glutModifierMask |= ShiftMask; - if (GetKeyState(VK_CONTROL) < 0) - __glutModifierMask |= ControlMask; - if (GetKeyState(VK_MENU) < 0) - __glutModifierMask |= Mod1Mask; - window->mouse(button, GLUT_UP, x, y); - __glutModifierMask = (unsigned int)~0; - } else { - /* Window might have been destroyed and all the - events for the window may not yet be received. */ - } - return 0; - - case WM_ENTERMENULOOP: - /* KLUDGE: create a timer that fires every 100 ms when we start a - menu so that we can still process the idle & timer events (that - way, the timers will fire during a menu pick and so will the - idle func. */ - SetTimer(hwnd, 1, 1, NULL); - return 0; - - case WM_TIMER: -#if 0 - /* If the timer id is 2, then this is the timer that is set up in - the main glut message processing loop, and we don't want to do - anything but acknowledge that we got it. It is used to prevent - CPU spiking when an idle function is installed. */ - if (wParam == 2) - return 0; -#endif - - /* only worry about the idle function and the timeouts, since - these are the only events we expect to process during - processing of a menu. */ - /* we no longer process the idle functions (as outlined in the - README), since drawing can't be done until the menu has - finished...it's pretty lame when the animation goes on, but - doesn't update, so you get this weird jerkiness. */ -#if 0 - if (__glutIdleFunc) - __glutIdleFunc(); -#endif - if (__glutTimerList) - handleTimeouts(); - return 0; - - case WM_EXITMENULOOP: - /* nuke the above created timer...we don't need it anymore, since - the menu is gone now. */ - KillTimer(hwnd, 1); - return 0; - - case WM_MENUSELECT: - if (lParam != 0) - __glutHMenu = (HMENU)lParam; - return 0; - - case WM_COMMAND: - if (__glutMappedMenu) { - if (GetSubMenu(__glutHMenu, LOWORD(wParam))) - __glutItemSelected = NULL; - else - __glutItemSelected = - __glutGetUniqueMenuItem(__glutMappedMenu, LOWORD(wParam)); - GetCursorPos(&point); - ScreenToClient(hwnd, &point); - __glutFinishMenu(hwnd, point.x, point.y); - } - return 0; - - case WM_MOUSEMOVE: - if (!__glutMappedMenu) { - window = __glutGetWindow(hwnd); - if (window) { - /* If motion function registered _and_ buttons held * - down, call motion function... */ - x = LOWORD(lParam); - y = HIWORD(lParam); - - /* Win32 doesn't return the same numbers as X does when the - mouse goes beyond the upper or left side of the window. - roll the Win32's 0..2^16 pointer co-ord range to 0..+/-2^15. */ - if(x & 1 << 15) x -= (1 << 16); - if(y & 1 << 15) y -= (1 << 16); - - if (window->motion && wParam & - (MK_LBUTTON | MK_MBUTTON | MK_RBUTTON)) { - __glutSetWindow(window); - window->motion(x, y); - } - /* If passive motion function registered _and_ - buttons not held down, call passive motion - function... */ - else if (window->passive && - ((wParam & - (MK_LBUTTON | MK_MBUTTON | MK_RBUTTON)) == - 0)) { - __glutSetWindow(window); - window->passive(x, y); - } - } - } else { - /* Motion events are thrown away when a pop up menu is - active. */ - } - return 0; - - case WM_GETMINMAXINFO: - /* this voodoo is brought to you by Win32 (again). It allows the - window to be bigger than the screen, and smaller than 100x100 - (although it doesn't seem to help the y minimum). */ - minmax = (LPMINMAXINFO)lParam; - minmax->ptMaxSize.x = __glutScreenWidth; - minmax->ptMaxSize.y = __glutScreenHeight; - minmax->ptMinTrackSize.x = 0; - minmax->ptMinTrackSize.y = 0; - minmax->ptMaxTrackSize.x = __glutScreenWidth + - GetSystemMetrics(SM_CXSIZE) * 2; - minmax->ptMaxTrackSize.y = __glutScreenHeight + - GetSystemMetrics(SM_CXSIZE) * 2 + GetSystemMetrics(SM_CYCAPTION); - return 0; - - case WM_SIZE: - window = __glutGetWindow(hwnd); - if (window) { - width = LOWORD(lParam); - height = HIWORD(lParam); - if (width != window->width || height != window->height) { -#if 0 /* Win32 GLUT does not support overlays for now. */ - if (window->overlay) { - XResizeWindow(__glutDisplay, window->overlay->win, width, height); - } -#endif - window->width = width; - window->height = height; - __glutSetWindow(window); - /* Do not execute OpenGL out of sequence with respect - to the SetWindowPos request! */ - GdiFlush(); - window->reshape(width, height); - window->forceReshape = FALSE; - /* A reshape should be considered like posting a - repair request. */ - __glutPostRedisplay(window, GLUT_REPAIR_WORK); - } - } - return 0; - - case WM_SETCURSOR: - /* If the cursor is not in the client area, then we want to send - this message to the default window procedure ('cause its - probably in the border or title, and we don't handle that - cursor. otherwise, set our cursor. Win32 makes us set the - cursor every time the mouse moves (DUMB!). */ - if(LOWORD(lParam) != HTCLIENT) { - goto defproc; - } - window = __glutGetWindow(hwnd); - if (window) { - __glutSetCursor(window); - } - /* TODO: check out the info in DevStudio on WM_SETCURSOR in the - DefaultAction section. */ - return 1; - - case WM_SETFOCUS: - window = __glutGetWindow(hwnd); - if (window) { - window->entryState = WM_SETFOCUS; - if (window->entry) { - __glutSetWindow(window); - window->entry(GLUT_ENTERED); - /* XXX Generation of fake passive notify? See how much - work the X11 code does to support fake passive notify - callbacks. */ - } - if (window->joystick && __glutCurrentWindow) { - if (__glutCurrentWindow->joyPollInterval > 0) { - MMRESULT result; - - /* Because Win32 will only let one window capture the - joystick at a time, we must capture it when we get the - focus and release it when we lose the focus. */ - result = joySetCapture(__glutCurrentWindow->win, - JOYSTICKID1, 0, TRUE); - if (result != JOYERR_NOERROR) { - return 0; - } - (void) joySetThreshold(JOYSTICKID1, - __glutCurrentWindow->joyPollInterval); - } - } - } - return 0; - - case WM_KILLFOCUS: - window = __glutGetWindow(hwnd); - if (window) { - window->entryState = WM_KILLFOCUS; - if (window->entry) { - __glutSetWindow(window); - window->entry(GLUT_LEFT); - } - if (window->joystick && __glutCurrentWindow) { - if (__glutCurrentWindow->joyPollInterval > 0) { - /* Because Win32 will only let one window capture the - joystick at a time, we must capture it when we get the - focus and release it when we lose the focus. */ - (void) joyReleaseCapture(JOYSTICKID1); - } - } - } - return 0; - case WM_ACTIVATE: - window = __glutGetWindow(hwnd); - /* Make sure we re-select the correct palette if needed. */ - if (LOWORD(wParam)) { - PostMessage(hwnd, WM_PALETTECHANGED, 0, 0); - } - if (window) { - int visState; - - /* HIWORD(wParam) is the minimized flag. */ - visState = !HIWORD(wParam); - updateWindowState(window, visState); - } - return 0; - - /* Colour Palette Management */ - case WM_PALETTECHANGED: - if (hwnd == (HWND)wParam) { - /* Don't respond to the message that we sent! */ - break; - } - /* fall through to WM_QUERYNEWPALETTE */ - - case WM_QUERYNEWPALETTE: - window = __glutGetWindow(hwnd); - if (window && window->colormap) { - UnrealizeObject(window->colormap->cmap); - SelectPalette(window->hdc, window->colormap->cmap, FALSE); - RealizePalette(window->hdc); - return TRUE; - } - return FALSE; - - case MM_JOY1MOVE: - case MM_JOY1ZMOVE: - window = __glutGetWindow(hwnd); - if (window->joystick) { - JOYINFOEX jix; - int x, y, z; - - /* Because WIN32 only supports messages for X, Y, and Z - translations, we must poll for the rest */ - jix.dwSize = sizeof(jix); - jix.dwFlags = JOY_RETURNALL; - joyGetPosEx(JOYSTICKID1,&jix); - -#define SCALE(v) ((int) ((v - 32767)/32.768)) - - /* Convert to integer for scaling. */ - x = jix.dwXpos; - y = jix.dwYpos; - z = jix.dwZpos; - window->joystick(jix.dwButtons, SCALE(x), SCALE(y), SCALE(z)); - - return TRUE; - } - return FALSE; - case MM_JOY1BUTTONDOWN: - case MM_JOY1BUTTONUP: - window = __glutGetWindow(hwnd); - if (window->joystick) { - JOYINFOEX jix; - - /* Because WIN32 only supports messages for X, Y, and Z - translations, we must poll for the rest */ - jix.dwSize = sizeof(jix); - jix.dwFlags = JOY_RETURNALL; - joyGetPosEx(JOYSTICKID1,&jix); - - return TRUE; - } - return FALSE; - -#if 0 - /* Miscellaneous messages (don't really need to enumerate them, - but it's good to know what you're not getting sometimes). */ - case WM_DISPLAYCHANGE: - break; - case WM_NCHITTEST: - /* This event is generated by every mouse move event. */ - goto defproc; - case WM_NCMOUSEMOVE: - goto defproc; - case WM_NCACTIVATE: - goto defproc; - case WM_NCPAINT: - goto defproc; - case WM_NCCALCSIZE: - goto defproc; - case WM_NCCREATE: - goto defproc; - case WM_NCDESTROY: - goto defproc; - case WM_NCLBUTTONDOWN: - goto defproc; - case WM_SETTEXT: - goto defproc; - case WM_GETTEXT: - goto defproc; - case WM_ACTIVATEAPP: - goto defproc; - case WM_GETICON: - goto defproc; - case WM_ERASEBKGND: - goto defproc; - case WM_WINDOWPOSCHANGING: - goto defproc; - case WM_WINDOWPOSCHANGED: - goto defproc; - case WM_MOUSEACTIVATE: - goto defproc; - case WM_SHOWWINDOW: - goto defproc; - case WM_MOVING: - goto defproc; - case WM_MOVE: - goto defproc; - case WM_KEYUP: - goto defproc; - case WM_CAPTURECHANGED: - goto defproc; - case WM_SYSCOMMAND: - goto defproc; - case WM_ENTERSIZEMOVE: - goto defproc; - case WM_ENTERIDLE: - goto defproc; -#endif - - default: - goto defproc; - } - -defproc: - return DefWindowProc(hwnd, msg, wParam, lParam); -} - -#endif - -#if defined(__OS2PM__) -Bool __glutSetWindowText(Window window, char *text) -{ - return WinSetWindowText(window, (PCSZ)text); - -} - -#endif +/* os2_winproc.c */ + + +#define INCL_DEV +#include "WarpGL.h" +#include "GL/os2mesa.h" + + +#define _MEERROR_H_ +#include /* It is from MMPM toolkit */ +#include +#include + + +#include "os2mesadef.h" +#include "glutint.h" + + +#define POKA 0 + +#if POKA + +extern unsigned __glutMenuButton; +extern GLUTidleCB __glutIdleFunc; +extern GLUTtimer *__glutTimerList; +extern void handleTimeouts(void); +extern GLUTmenuItem *__glutGetUniqueMenuItem(GLUTmenu * menu, int unique); +static HMENU __glutHMenu; + +#endif + +extern void _mesa_ResizeBuffersMESA( void ); + + +MRESULT EXPENTRY GlutWindowProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ); +MRESULT EXPENTRY GlutWindowChildProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ); +void updateWindowState(GLUTwindow *window, int visState); + +volatile extern HAB hab; /* PM anchor block handle */ +volatile extern HPS hpsCurrent; + +RECTL rCtls[52]; +ULONG ulNumRcls; + +MRESULT EXPENTRY GlutWindowChildProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ) +{ MRESULT rc; + rc = GlutWindowProc(hwnd, msg, mp1, mp2 ); + return rc; +} + +MRESULT EXPENTRY GlutWindowProc( HWND hwnd, ULONG msg, MPARAM mp1, MPARAM mp2 ) +{ + HPS hps = NULLHANDLE; /* presentation space handle */ + GLUTwindow* window; /* GLUT window associated with message. */ + GLUTmenu* menu; /* GLUT menu associated with message. */ + RECTL rclClient; + POINTL point; + int button = -1,rc,key; + + +/* Process the message. */ + + switch( msg ) + { + case WM_CREATE: + { + SIZEL sizl = { 0L, 0L }; + LONG *alCaps; + HDC hdc; + + /*+-----------------------------------------------------------------+*/ + /*| The client window is being created. Create the semaphore to |*/ + /*| control access to the presentation space. Then create the |*/ + /*| thread that will draw the lines. |*/ + /*+-----------------------------------------------------------------+*/ + // DosCreateMutexSem( (PSZ)NULL, &hmtxPS, 0UL, FALSE ); + + hdc = WinOpenWindowDC(hwnd); + + /*+-----------------------------------------------------------------+*/ + /*| Create a non-cached presentation space. We will not release |*/ + /*| this PS, as we will be Selecting a Palette to this PS and then |*/ + /*| animating the palette. Upon releasing a PS the palette is no |*/ + /*| longer selected for obvious reasons. |*/ + /*+-----------------------------------------------------------------+*/ + hpsCurrent = GpiCreatePS( hab, + hdc, + &sizl, + PU_PELS | GPIF_DEFAULT | + GPIT_MICRO | GPIA_ASSOC ); +// DevQueryCaps( hdc, lStart, lCount, alCaps ); +// fPaletteCaps = alCaps[CAPS_ADDITIONAL_GRAPHICS] & CAPS_PALETTE_MANAGER; +// PaletteInit(3); + /* ¯¥p¥¢®¤ hpsBuffer ¢ p¥¦¨¬ RGB color table */ + + GpiCreateLogColorTable(hpsCurrent,0 ,LCOLF_RGB,0,0,NULL); + GpiSetPattern(hpsCurrent,PATSYM_SOLID); + GpiSetPatternSet(hpsCurrent,LCID_DEFAULT); + + } + break; + + return 0; + case WM_CLOSE: + WinPostMsg( hwnd, WM_QUIT, NULL, NULL ); + + return 0; + + case WM_PAINT: + window = __glutGetWindow(hwnd); + if (window) + { + PWMC ctx; +// hps = WinBeginPaint(hwnd,NULLHANDLE,&rclClient); + hps = WinBeginPaint(hwnd,NULLHANDLE,&rclClient); + // blit Dive buffer to screen. + + { + SWP swp; // Window position + POINTL pointl; // Point to offset from Desktop + + // Convert the point to offset from desktop lower left. + pointl.x = 0; + pointl.y = 0; + WinMapWindowPoints ( hwnd, HWND_DESKTOP, &pointl, 1 ); + + +// ctx = window->ctx; +// ctx->xDiveScr = pointl.x; +// ctx->yDiveScr = pointl.y; + } +// rc = DiveBlitImage (ctx->hDive, +// ctx->ulDiveBufferNumber, +// DIVE_BUFFER_SCREEN ); +// + + if (window->win == hwnd) { + __glutPostRedisplay(window, GLUT_REPAIR_WORK); + } else if (window->overlay && window->overlay->win == hwnd) { + __glutPostRedisplay(window, GLUT_OVERLAY_REPAIR_WORK); + } + WinEndPaint(hps); + } else { + + hps = WinBeginPaint(hwnd,NULLHANDLE,&rclClient); + WinFillRect(hps, &rclClient, CLR_WHITE); + WinEndPaint(hps); + } + break; + + case WM_VRNDISABLED: + +// pwinData->fDataInProcess = TRUE; +// DiveSetupBlitter ( pwinData->hDive, 0 ); +// pwinData->fVrnDisabled = TRUE; + break; + + case WM_VRNENABLED: + { HRGN hrgn; /* Region handle */ + RGNRECT rgnCtl; /* Processing control structure */ +// RECTL rCtls[52]; +// ULONG ulNumRcls; + +// pwinData->fDataInProcess = TRUE; + hps = WinGetPS ( hwnd ); + if ( !hps ) + break; + hrgn = GpiCreateRegion ( hps, 0L, NULL ); + if ( hrgn ) + { /* NOTE: If mp1 is zero, then this was just a move message. + ** Illustrate the visible region on a WM_VRNENABLE. + */ + WinQueryVisibleRegion ( hwnd, hrgn ); + rgnCtl.ircStart = 0; + rgnCtl.crc = 50; + rgnCtl.ulDirection = 1; + + /* Get the all ORed rectangles */ + if ( GpiQueryRegionRects ( hps, hrgn, NULL, + &rgnCtl, rCtls) ) + { + ulNumRcls = rgnCtl.crcReturned; + + /* Now find the window position and size, relative to parent. + */ +// WinQueryWindowPos ( pwinData->hwndClient, &pwinData->swp ); + +// rcl.xLeft = 0; +// rcl.yBottom = 0; + + /* Convert the point to offset from desktop lower left. + */ +// pointl.x = pwinData->swp.x; +// pointl.y = pwinData->swp.y; + +// WinMapWindowPoints ( pwinData->hwndFrame, +// HWND_DESKTOP, &pointl, 1 ); + +// pwinData->cxWindowPos = pointl.x; +// pwinData->cyWindowPos = pointl.y; + + } + GpiDestroyRegion( hps, hrgn ); + } + WinReleasePS( hps ); + + } + break; + + case WM_SIZE: + window = __glutGetWindow(hwnd); + if (window) + { int width,height; + width = SHORT1FROMMP(mp2); + height = SHORT2FROMMP(mp2); + if (width != window->width || height != window->height) { +#if 0 /* Win32 GLUT does not support overlays for now. */ + if (window->overlay) { + XResizeWindow(__glutDisplay, window->overlay->win, width, height); + } +#endif + window->width = width; + window->height = height; + __glutSetWindow(window); + if(width <= 0 || height <= 0) + break; + _mesa_ResizeBuffersMESA(); + + /* Do not execute OpenGL out of sequence with respect + to the SetWindowPos request! */ + window->reshape(width, height); + window->forceReshape = FALSE; + /* A reshape should be considered like posting a + repair request. */ + __glutPostRedisplay(window, GLUT_REPAIR_WORK); + } + } + return 0; + case WM_SHOW: + window = __glutGetWindow(hwnd); + if (window) { + int visState; + visState = SHORT1FROMMP( mp1 ); + updateWindowState(window, visState); + } + return 0; + + case WM_ACTIVATE: + window = __glutGetWindow(hwnd); +// /* Make sure we re-select the correct palette if needed. */ +// if (LOWORD(wParam)) { +// PostMessage(hwnd, WM_PALETTECHANGED, 0, 0); +// } + if (window) { + int visState; + visState = SHORT1FROMMP( mp1 ); + updateWindowState(window, visState); + } + return 0; + + case WM_CHAR: + { USHORT fsflags; + window = __glutGetWindow(hwnd); + if (!window) { + break; + } + fsflags = SHORT1FROMMP(mp1); +/* ?? */ + if((fsflags & KC_KEYUP) ) /* ¨£­®p¨p㥬 ®â¦ â¨¥ ª­®¯ª¨, p¥ £¨p㥬 ⮫쪮 ­  ­ ¦ â¨¥ */ + break; +/////////////////////////////////////////////////// + if(!(fsflags & KC_CHAR) ) + { + if (!(fsflags & KC_VIRTUALKEY)) + break; + key = 0; + /* Get the virtual key from mp2. */ + switch (SHORT2FROMMP(mp2)) + { +/* directional keys */ + case VK_LEFT: key = GLUT_KEY_LEFT; break; + case VK_UP: key = GLUT_KEY_UP; break; + case VK_RIGHT: key = GLUT_KEY_RIGHT; break; + case VK_DOWN: key = GLUT_KEY_DOWN; break; + + case VK_PAGEUP: key = GLUT_KEY_PAGE_UP; break; + case VK_PAGEDOWN:key = GLUT_KEY_PAGE_DOWN; break; + case VK_HOME: key = GLUT_KEY_HOME;break; + case VK_END: key = GLUT_KEY_END; break; + case VK_INSERT: key = GLUT_KEY_INSERT; break; + +/* function keys */ + case VK_F1 : key = GLUT_KEY_F1; break; + case VK_F2 : key = GLUT_KEY_F2; break; + case VK_F3 : key = GLUT_KEY_F3; break; + case VK_F4 : key = GLUT_KEY_F4; break; + case VK_F5 : key = GLUT_KEY_F5; break; + case VK_F6 : key = GLUT_KEY_F6; break; + case VK_F7 : key = GLUT_KEY_F7; break; + case VK_F8 : key = GLUT_KEY_F8; break; + case VK_F9 : key = GLUT_KEY_F9; break; + case VK_F10: key = GLUT_KEY_F10;break; + case VK_F11: key = GLUT_KEY_F11; break; + case VK_F12: key = GLUT_KEY_F12; break; + case VK_ESC: key = -1; break; /* Character codes */ + case VK_SPACE: key = -1; break; + case VK_TAB: key = -1; break; + } + if(!key) + { break; /* Key Not implemented */ + } + if(key > 0) + { if (!window->special) /* ­¥ ãáâ ­®¢«¥­® ®¡à ¡®â稪  */ + break; + + WinQueryPointerPos(HWND_DESKTOP,&point); + ScreenToClient(window->win, &point); + __glutSetWindow(window); + __glutModifierMask = 0; + if(WinGetKeyState(HWND_DESKTOP,VK_SHIFT) & 0x8000) /* high order bit is on */ + __glutModifierMask |= ShiftMask; + if(WinGetKeyState(HWND_DESKTOP,VK_CTRL) & 0x8000) + __glutModifierMask |= ControlMask; + if(WinGetKeyState(HWND_DESKTOP,VK_MENU) & 0x8000) + __glutModifierMask |= Mod1Mask; + window->special(key, point.x, point.y); + __glutModifierMask = (unsigned int) ~0; + return 0; + } + + } +///////////////////////////////////////////////////// + /* If we are ignoring auto repeated key strokes for the window, bail. */ + if (window->ignoreKeyRepeat && (CHAR3FROMMP(mp1)) ) + break; + if(!((unsigned char)SHORT1FROMMP(mp2)) ) /* ¨£­®p¨p㥬 ­¥á¨¬¢®«ì­ë¥ ª®¤ë */ + break; + if (window->keyboard) { + WinQueryPointerPos(HWND_DESKTOP,&point); + + ScreenToClient(window->win, &point); + __glutSetWindow(window); + __glutModifierMask = 0; + if(WinGetKeyState(HWND_DESKTOP,VK_SHIFT) & 0x8000) /* high order bit is on */ + __glutModifierMask |= ShiftMask; + if(WinGetKeyState(HWND_DESKTOP,VK_CTRL) & 0x8000) + __glutModifierMask |= ControlMask; + if(WinGetKeyState(HWND_DESKTOP,VK_MENU) & 0x8000) + __glutModifierMask |= Mod1Mask; + window->keyboard((unsigned char)SHORT1FROMMP(mp2), point.x, point.y); + __glutModifierMask = (unsigned int) ~0; + } + return 0; + } /* endof case WM_CHAR: */ +//////////////////////////////////////////////// + case WM_BUTTON1DOWN: + button = GLUT_LEFT_BUTTON; + case WM_BUTTON3DOWN: + if (button < 0) + button = GLUT_MIDDLE_BUTTON; + case WM_BUTTON2DOWN: + if (button < 0) + button = GLUT_RIGHT_BUTTON; + { POINTS psh; + psh = *((POINTS *)&mp1); + point.x = psh.x; + point.y = psh.y; + } + /* finish the menu if we get a button down message (user must have + cancelled the menu). */ + if (__glutMappedMenu) { + /* TODO: take this out once the menu on middle mouse stuff works + properly. */ + if (button == GLUT_MIDDLE_BUTTON) + return 0; + /* get current mouse pointer position */ +// WinQueryPointerPos(HWND_DESKTOP,&point); + /* map from desktop to client window */ +// WinMapWindowPoints(HWND_DESKTOP, hwnd, &point, 1); + __glutItemSelected = NULL; + __glutFinishMenu(hwnd, point.x, point.y); + return 0; + } + window = __glutGetWindow(hwnd); + if (window) { + window->buttonDownState = button+1; + menu = __glutGetMenuByNum(window->menu[button]); + if (menu) { +//todo +// __glutMenuButton = button == GLUT_RIGHT_BUTTON ? TPM_RIGHTBUTTON : +// button == GLUT_LEFT_BUTTON ? TPM_LEFTBUTTON : +// 0x0001; +// __glutStartMenu(menu, window, point.x, point.y, x, y); + } else if (window->mouse) { + + __glutSetWindow(window); + __glutModifierMask = 0; + if (WinGetKeyState(HWND_DESKTOP,VK_SHIFT) & 0x8000) /* < 0 = high order bit is on. */ + __glutModifierMask |= ShiftMask; + if (WinGetKeyState(HWND_DESKTOP,VK_CTRL) & 0x8000) + __glutModifierMask |= ControlMask; + if (WinGetKeyState(HWND_DESKTOP,VK_MENU) & 0x8000) + __glutModifierMask |= Mod1Mask; + window->mouse(button, GLUT_DOWN, point.x, point.y); + __glutModifierMask = (unsigned int)~0; + } else { + /* Stray mouse events. Ignore. */ + } + } + return 0; + + break; +/********************************************/ + case WM_BUTTON1UP: + button = GLUT_LEFT_BUTTON; + case WM_BUTTON3UP: + if (button < 0) + button = GLUT_MIDDLE_BUTTON; + case WM_BUTTON2UP: + if (button < 0) + button = GLUT_RIGHT_BUTTON; + { POINTS psh; + psh = *((POINTS *)&mp1); + point.x = psh.x; + point.y = psh.y; + } + /* Bail out if we're processing a menu. */ + /* Bail out = ¢ë¡à®á¨âìáï á ¯ à èã⮬ */ + if (__glutMappedMenu) { + WinQueryPointerPos(HWND_DESKTOP,&point); + WinMapWindowPoints(HWND_DESKTOP, hwnd, &point, 1); + /* if we're getting the middle button up signal, then something + on the menu was selected. */ + if (button == GLUT_MIDDLE_BUTTON) { + return 0; + /* For some reason, the code below always returns -1 even + though the point IS IN THE ITEM! Therefore, just bail out if + we get a middle mouse up. The user must select using the + left mouse button. Stupid Win32. */ +#if 0 + int item = MenuItemFromPoint(hwnd, __glutHMenu, point); + if (item != -1) + __glutItemSelected = (GLUTmenuItem*)GetMenuItemID(__glutHMenu, item); + else + __glutItemSelected = NULL; + __glutFinishMenu(hwnd, point.x, point.y); +#endif + } else { + __glutItemSelected = NULL; + __glutFinishMenu(hwnd, point.x, point.y); + } + return 0; + } + + window = __glutGetWindow(hwnd); + if(window) + window->buttonDownState = 0; + + if (window && window->mouse) { + __glutSetWindow(window); + __glutModifierMask = 0; + if (WinGetKeyState(HWND_DESKTOP,VK_SHIFT) & 0x8000) /* < 0 = high order bit is on */ + __glutModifierMask |= ShiftMask; + if (WinGetKeyState(HWND_DESKTOP,VK_CTRL) & 0x8000) + __glutModifierMask |= ControlMask; + if (WinGetKeyState(HWND_DESKTOP,VK_MENU) & 0x8000) + __glutModifierMask |= Mod1Mask; + window->mouse(button, GLUT_UP, point.x, point.y); + + __glutModifierMask = (unsigned int)~0; + } else { + /* Window might have been destroyed and all the + events for the window may not yet be received. */ + } + return 0; + + + break; +////////////////////////////////////////////////// + case WM_COMMAND: + window = __glutGetWindow(hwnd); + if (window) + { if (window->wm_command) + window->wm_command(hwnd,mp1,mp2); + } + break; + + case WM_MOUSEMOVE: + if (!__glutMappedMenu) { + window = __glutGetWindow(hwnd); + if (window) { + /* If motion function registered _and_ buttons held * + down, call motion function... */ + { POINTS psh; + psh = *((POINTS *)&mp1); + point.x = psh.x; + point.y = psh.y; + } + + if (window->motion && window->buttonDownState) { + __glutSetWindow(window); + window->motion(point.x, point.y); + } + /* If passive motion function registered _and_ + buttons not held down, call passive motion + function... */ + else if (window->passive && !window->buttonDownState) { + __glutSetWindow(window); + window->passive(point.x, point.y); + } + } + } else { + /* Motion events are thrown away when a pop up menu is + active. */ + } + return 0; + + + default: + /* For all other messages, let the default window procedure process them. */ + return ( WinDefWindowProc( hwnd, msg, mp1, mp2 ) ); + + } //endof switch( msg ) + return ( WinDefWindowProc( hwnd, msg, mp1, mp2 ) ); +// return NULL; +} + +void APIENTRY glutCommandFunc(GLUTcommandCB Func) +{ +extern GLUTwindow *__glutCurrentWindow; + __glutCurrentWindow->wm_command = Func; +} + + + + +void +updateWindowState(GLUTwindow *window, int visState) +{ + GLUTwindow* child; + + /* XXX shownState and visState are the same in Win32. */ + window->shownState = visState; + if (visState != window->visState) { + if (window->windowStatus) { + window->visState = visState; + __glutSetWindow(window); + window->windowStatus(visState); + } + } + /* Since Win32 only sends an activate for the toplevel window, + update the visibility for all the child windows. */ + child = window->children; + while (child) { + updateWindowState(child, visState); + child = child->siblings; + } +} + +#if POKA + +LONG WINAPI +__glutWindowProc(HWND hwnd, UINT msg, WPARAM wParam, LPARAM lParam) +{ + POINT point; /* Point structure. */ + PAINTSTRUCT ps; /* Paint structure. */ + LPMINMAXINFO minmax; /* Minimum/maximum info structure. */ + GLUTwindow* window; /* GLUT window associated with message. */ + GLUTmenu* menu; /* GLUT menu associated with message. */ + int x, y, width, height, key; + int button = -1; + + switch(msg) { + case WM_CREATE: + return 0; + case WM_CLOSE: + PostQuitMessage(0); + return 0; +#if 0 + case WM_DESTROY: + /* XXX NVidia's NT OpenGL can have problems closing down + its OpenGL internal data structures if we just allow + the process to terminate without unbinding and deleting + the windows context. Apparently, DirectDraw unloads + before OPENGL32.DLL in the close down sequence, but + NVidia's NT OpenGL needs DirectDraw to close down its + data structures. */ + window = __glutGetWindow(hwnd); + if (window) { + if (window->ctx) { + wglMakeCurrent(NULL, NULL); + wglDeleteContext(window->ctx); + } + } + return 0; +#endif + + case WM_SYSKEYUP: + case WM_KEYUP: + window = __glutGetWindow(hwnd); + if (!window) { + break; + } + /* Win32 is dumb and sends these messages only to the parent + window. Therefore, find out if we're in a child window and + call the child windows keyboard callback if we are. */ + if (window->parent) { + GetCursorPos(&point); + ScreenToClient(hwnd, &point); + hwnd = ChildWindowFromPoint(hwnd, point); + window = __glutGetWindow(hwnd); + } + if (window->specialUp || window->keyboardUp) { + GetCursorPos(&point); + ScreenToClient(window->win, &point); + __glutSetWindow(window); + __glutModifierMask = 0; + if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ + __glutModifierMask |= ShiftMask; + if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ + __glutModifierMask |= ControlMask; + if (GetKeyState(VK_MENU) < 0) + __glutModifierMask |= Mod1Mask; + switch (wParam) { + /* *INDENT-OFF* */ + case VK_F1: key = GLUT_KEY_F1; break; + case VK_F2: key = GLUT_KEY_F2; break; + case VK_F3: key = GLUT_KEY_F3; break; + case VK_F4: key = GLUT_KEY_F4; break; + case VK_F5: key = GLUT_KEY_F5; break; + case VK_F6: key = GLUT_KEY_F6; break; + case VK_F7: key = GLUT_KEY_F7; break; + case VK_F8: key = GLUT_KEY_F8; break; + case VK_F9: key = GLUT_KEY_F9; break; + case VK_F10: key = GLUT_KEY_F10; break; + case VK_F11: key = GLUT_KEY_F11; break; + case VK_F12: key = GLUT_KEY_F12; break; + case VK_LEFT: key = GLUT_KEY_LEFT; break; + case VK_UP: key = GLUT_KEY_UP; break; + case VK_RIGHT: key = GLUT_KEY_RIGHT; break; + case VK_DOWN: key = GLUT_KEY_DOWN; break; + case VK_PRIOR: key = GLUT_KEY_PAGE_UP; break; + case VK_NEXT: key = GLUT_KEY_PAGE_DOWN; break; + case VK_HOME: key = GLUT_KEY_HOME; break; + case VK_END: key = GLUT_KEY_END; break; + case VK_INSERT: key = GLUT_KEY_INSERT; break; + case VK_DELETE: + /* Delete is an ASCII character. */ + if (window->keyboardUp) { + window->keyboardUp((unsigned char) 127, point.x, point.y); + } + return 0; + /* *INDENT-ON* */ + default: + if (window->keyboardUp) { + key = MapVirtualKey(wParam, 2); /* Map to ASCII. */ + if (isascii(key) && (key != 0)) { + + /* XXX Attempt to determine modified ASCII character + is quite incomplete. Digits, symbols, CapsLock, + Ctrl, and numeric keypad are all ignored. Fix this. */ + + if (!(__glutModifierMask & ShiftMask)) + key = tolower(key); + window->keyboardUp((unsigned char) key, point.x, point.y); + } + } + __glutModifierMask = (unsigned int) ~0; + return 0; + } + if (window->specialUp) { + window->specialUp(key, point.x, point.y); + } + __glutModifierMask = (unsigned int) ~0; + } + return 0; + + case WM_SYSCHAR: + case WM_CHAR: + window = __glutGetWindow(hwnd); + if (!window) { + break; + } + + /* Bit 30 of lParam is set if key already held down. If + we are ignoring auto repeated key strokes for the window, bail. */ + if (window->ignoreKeyRepeat && (lParam & (1 << 30)) ) { + break; + } + + /* Win32 is dumb and sends these messages only to the parent + window. Therefore, find out if we're in a child window and + call the child windows keyboard callback if we are. */ + if (window->parent) { + GetCursorPos(&point); + ScreenToClient(hwnd, &point); + hwnd = ChildWindowFromPoint(hwnd, point); + window = __glutGetWindow(hwnd); + } + if (window->keyboard) { + GetCursorPos(&point); + ScreenToClient(window->win, &point); + __glutSetWindow(window); + __glutModifierMask = 0; + if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ + __glutModifierMask |= ShiftMask; + if (GetKeyState(VK_CONTROL) < 0) + __glutModifierMask |= ControlMask; + if (GetKeyState(VK_MENU) < 0) + __glutModifierMask |= Mod1Mask; + window->keyboard((unsigned char)wParam, point.x, point.y); + __glutModifierMask = (unsigned int) ~0; + } + return 0; + + case WM_SYSKEYDOWN: + case WM_KEYDOWN: + window = __glutGetWindow(hwnd); + if (!window) { + break; + } + + /* Bit 30 of lParam is set if key already held down. If + we are ignoring auto repeated key strokes for the window, bail. */ + if (window->ignoreKeyRepeat && (lParam & (1 << 30)) ) { + break; + } + + /* Win32 is dumb and sends these messages only to the parent + window. Therefore, find out if we're in a child window and + call the child windows keyboard callback if we are. */ + if (window->parent) { + GetCursorPos(&point); + ScreenToClient(hwnd, &point); + hwnd = ChildWindowFromPoint(hwnd, point); + window = __glutGetWindow(hwnd); + } + if (window->special) { + switch (wParam) { + /* *INDENT-OFF* */ + /* function keys */ + case VK_F1: key = GLUT_KEY_F1; break; + case VK_F2: key = GLUT_KEY_F2; break; + case VK_F3: key = GLUT_KEY_F3; break; + case VK_F4: key = GLUT_KEY_F4; break; + case VK_F5: key = GLUT_KEY_F5; break; + case VK_F6: key = GLUT_KEY_F6; break; + case VK_F7: key = GLUT_KEY_F7; break; + case VK_F8: key = GLUT_KEY_F8; break; + case VK_F9: key = GLUT_KEY_F9; break; + case VK_F10: key = GLUT_KEY_F10; break; + case VK_F11: key = GLUT_KEY_F11; break; + case VK_F12: key = GLUT_KEY_F12; break; + /* directional keys */ + case VK_LEFT: key = GLUT_KEY_LEFT; break; + case VK_UP: key = GLUT_KEY_UP; break; + case VK_RIGHT: key = GLUT_KEY_RIGHT; break; + case VK_DOWN: key = GLUT_KEY_DOWN; break; + /* *INDENT-ON* */ + + case VK_PRIOR: + /* VK_PRIOR is Win32's Page Up */ + key = GLUT_KEY_PAGE_UP; + break; + case VK_NEXT: + /* VK_NEXT is Win32's Page Down */ + key = GLUT_KEY_PAGE_DOWN; + break; + case VK_HOME: + key = GLUT_KEY_HOME; + break; + case VK_END: + key = GLUT_KEY_END; + break; + case VK_INSERT: + key = GLUT_KEY_INSERT; + break; + case VK_DELETE: + goto handleDelete; + default: + goto defproc; + } + GetCursorPos(&point); + ScreenToClient(window->win, &point); + __glutSetWindow(window); + __glutModifierMask = 0; + if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ + __glutModifierMask |= ShiftMask; + if (GetKeyState(VK_CONTROL) < 0) + __glutModifierMask |= ControlMask; + if (GetKeyState(VK_MENU) < 0) + __glutModifierMask |= Mod1Mask; + window->special(key, point.x, point.y); + __glutModifierMask = (unsigned int) ~0; + } else if (window->keyboard) { + /* Specially handle any keys that match ASCII values but + do not generate Windows WM_SYSCHAR or WM_CHAR messages. */ + switch (wParam) { + case VK_DELETE: + handleDelete: + /* Delete is an ASCII character. */ + GetCursorPos(&point); + ScreenToClient(window->win, &point); + __glutSetWindow(window); + __glutModifierMask = 0; + if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ + __glutModifierMask |= ShiftMask; + if (GetKeyState(VK_CONTROL) < 0) + __glutModifierMask |= ControlMask; + if (GetKeyState(VK_MENU) < 0) + __glutModifierMask |= Mod1Mask; + window->keyboard((unsigned char) 127, point.x, point.y); + __glutModifierMask = (unsigned int) ~0; + return 0; + default: + /* Let the following WM_SYSCHAR or WM_CHAR message generate + the keyboard callback. */ + break; + } + } + return 0; + + case WM_LBUTTONDOWN: + button = GLUT_LEFT_BUTTON; + case WM_MBUTTONDOWN: + if (button < 0) + button = GLUT_MIDDLE_BUTTON; + case WM_RBUTTONDOWN: + if (button < 0) + button = GLUT_RIGHT_BUTTON; + + /* finish the menu if we get a button down message (user must have + cancelled the menu). */ + if (__glutMappedMenu) { + /* TODO: take this out once the menu on middle mouse stuff works + properly. */ + if (button == GLUT_MIDDLE_BUTTON) + return 0; + GetCursorPos(&point); + ScreenToClient(hwnd, &point); + __glutItemSelected = NULL; + __glutFinishMenu(hwnd, point.x, point.y); + return 0; + } + + /* set the capture so we can get mouse events outside the window */ + SetCapture(hwnd); + + /* Win32 doesn't return the same numbers as X does when the mouse + goes beyond the upper or left side of the window. roll the + Win32's 0..2^16 pointer co-ord range to 0 +/- 2^15. */ + x = LOWORD(lParam); + y = HIWORD(lParam); + if(x & 1 << 15) x -= (1 << 16); + if(y & 1 << 15) y -= (1 << 16); + + window = __glutGetWindow(hwnd); + if (window) { + menu = __glutGetMenuByNum(window->menu[button]); + if (menu) { + point.x = LOWORD(lParam); point.y = HIWORD(lParam); + ClientToScreen(window->win, &point); + __glutMenuButton = button == GLUT_RIGHT_BUTTON ? TPM_RIGHTBUTTON : + button == GLUT_LEFT_BUTTON ? TPM_LEFTBUTTON : + 0x0001; + __glutStartMenu(menu, window, point.x, point.y, x, y); + } else if (window->mouse) { + + __glutSetWindow(window); + __glutModifierMask = 0; + if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on. */ + __glutModifierMask |= ShiftMask; + if (GetKeyState(VK_CONTROL) < 0) + __glutModifierMask |= ControlMask; + if (GetKeyState(VK_MENU) < 0) + __glutModifierMask |= Mod1Mask; + window->mouse(button, GLUT_DOWN, x, y); + __glutModifierMask = (unsigned int)~0; + } else { + /* Stray mouse events. Ignore. */ + } + } + return 0; + + case WM_LBUTTONUP: + button = GLUT_LEFT_BUTTON; + case WM_MBUTTONUP: + if (button < 0) + button = GLUT_MIDDLE_BUTTON; + case WM_RBUTTONUP: + if (button < 0) + button = GLUT_RIGHT_BUTTON; + + /* Bail out if we're processing a menu. */ + if (__glutMappedMenu) { + GetCursorPos(&point); + ScreenToClient(hwnd, &point); + /* if we're getting the middle button up signal, then something + on the menu was selected. */ + if (button == GLUT_MIDDLE_BUTTON) { + return 0; + /* For some reason, the code below always returns -1 even + though the point IS IN THE ITEM! Therefore, just bail out if + we get a middle mouse up. The user must select using the + left mouse button. Stupid Win32. */ +#if 0 + int item = MenuItemFromPoint(hwnd, __glutHMenu, point); + if (item != -1) + __glutItemSelected = (GLUTmenuItem*)GetMenuItemID(__glutHMenu, item); + else + __glutItemSelected = NULL; + __glutFinishMenu(hwnd, point.x, point.y); +#endif + } else { + __glutItemSelected = NULL; + __glutFinishMenu(hwnd, point.x, point.y); + } + return 0; + } + + /* Release the mouse capture. */ + ReleaseCapture(); + + window = __glutGetWindow(hwnd); + if (window && window->mouse) { + /* Win32 doesn't return the same numbers as X does when the + mouse goes beyond the upper or left side of the window. roll + the Win32's 0..2^16 pointer co-ord range to 0 +/- 2^15. */ + x = LOWORD(lParam); + y = HIWORD(lParam); + if(x & 1 << 15) x -= (1 << 16); + if(y & 1 << 15) y -= (1 << 16); + + __glutSetWindow(window); + __glutModifierMask = 0; + if (GetKeyState(VK_SHIFT) < 0) /* < 0 = high order bit is on */ + __glutModifierMask |= ShiftMask; + if (GetKeyState(VK_CONTROL) < 0) + __glutModifierMask |= ControlMask; + if (GetKeyState(VK_MENU) < 0) + __glutModifierMask |= Mod1Mask; + window->mouse(button, GLUT_UP, x, y); + __glutModifierMask = (unsigned int)~0; + } else { + /* Window might have been destroyed and all the + events for the window may not yet be received. */ + } + return 0; + + case WM_ENTERMENULOOP: + /* KLUDGE: create a timer that fires every 100 ms when we start a + menu so that we can still process the idle & timer events (that + way, the timers will fire during a menu pick and so will the + idle func. */ + SetTimer(hwnd, 1, 1, NULL); + return 0; + + case WM_TIMER: +#if 0 + /* If the timer id is 2, then this is the timer that is set up in + the main glut message processing loop, and we don't want to do + anything but acknowledge that we got it. It is used to prevent + CPU spiking when an idle function is installed. */ + if (wParam == 2) + return 0; +#endif + + /* only worry about the idle function and the timeouts, since + these are the only events we expect to process during + processing of a menu. */ + /* we no longer process the idle functions (as outlined in the + README), since drawing can't be done until the menu has + finished...it's pretty lame when the animation goes on, but + doesn't update, so you get this weird jerkiness. */ +#if 0 + if (__glutIdleFunc) + __glutIdleFunc(); +#endif + if (__glutTimerList) + handleTimeouts(); + return 0; + + case WM_EXITMENULOOP: + /* nuke the above created timer...we don't need it anymore, since + the menu is gone now. */ + KillTimer(hwnd, 1); + return 0; + + case WM_MENUSELECT: + if (lParam != 0) + __glutHMenu = (HMENU)lParam; + return 0; + + case WM_COMMAND: + if (__glutMappedMenu) { + if (GetSubMenu(__glutHMenu, LOWORD(wParam))) + __glutItemSelected = NULL; + else + __glutItemSelected = + __glutGetUniqueMenuItem(__glutMappedMenu, LOWORD(wParam)); + GetCursorPos(&point); + ScreenToClient(hwnd, &point); + __glutFinishMenu(hwnd, point.x, point.y); + } + return 0; + + case WM_MOUSEMOVE: + if (!__glutMappedMenu) { + window = __glutGetWindow(hwnd); + if (window) { + /* If motion function registered _and_ buttons held * + down, call motion function... */ + x = LOWORD(lParam); + y = HIWORD(lParam); + + /* Win32 doesn't return the same numbers as X does when the + mouse goes beyond the upper or left side of the window. + roll the Win32's 0..2^16 pointer co-ord range to 0..+/-2^15. */ + if(x & 1 << 15) x -= (1 << 16); + if(y & 1 << 15) y -= (1 << 16); + + if (window->motion && wParam & + (MK_LBUTTON | MK_MBUTTON | MK_RBUTTON)) { + __glutSetWindow(window); + window->motion(x, y); + } + /* If passive motion function registered _and_ + buttons not held down, call passive motion + function... */ + else if (window->passive && + ((wParam & + (MK_LBUTTON | MK_MBUTTON | MK_RBUTTON)) == + 0)) { + __glutSetWindow(window); + window->passive(x, y); + } + } + } else { + /* Motion events are thrown away when a pop up menu is + active. */ + } + return 0; + + case WM_GETMINMAXINFO: + /* this voodoo is brought to you by Win32 (again). It allows the + window to be bigger than the screen, and smaller than 100x100 + (although it doesn't seem to help the y minimum). */ + minmax = (LPMINMAXINFO)lParam; + minmax->ptMaxSize.x = __glutScreenWidth; + minmax->ptMaxSize.y = __glutScreenHeight; + minmax->ptMinTrackSize.x = 0; + minmax->ptMinTrackSize.y = 0; + minmax->ptMaxTrackSize.x = __glutScreenWidth + + GetSystemMetrics(SM_CXSIZE) * 2; + minmax->ptMaxTrackSize.y = __glutScreenHeight + + GetSystemMetrics(SM_CXSIZE) * 2 + GetSystemMetrics(SM_CYCAPTION); + return 0; + + case WM_SIZE: + window = __glutGetWindow(hwnd); + if (window) { + width = LOWORD(lParam); + height = HIWORD(lParam); + if (width != window->width || height != window->height) { +#if 0 /* Win32 GLUT does not support overlays for now. */ + if (window->overlay) { + XResizeWindow(__glutDisplay, window->overlay->win, width, height); + } +#endif + window->width = width; + window->height = height; + __glutSetWindow(window); + /* Do not execute OpenGL out of sequence with respect + to the SetWindowPos request! */ + GdiFlush(); + window->reshape(width, height); + window->forceReshape = FALSE; + /* A reshape should be considered like posting a + repair request. */ + __glutPostRedisplay(window, GLUT_REPAIR_WORK); + } + } + return 0; + + case WM_SETCURSOR: + /* If the cursor is not in the client area, then we want to send + this message to the default window procedure ('cause its + probably in the border or title, and we don't handle that + cursor. otherwise, set our cursor. Win32 makes us set the + cursor every time the mouse moves (DUMB!). */ + if(LOWORD(lParam) != HTCLIENT) { + goto defproc; + } + window = __glutGetWindow(hwnd); + if (window) { + __glutSetCursor(window); + } + /* TODO: check out the info in DevStudio on WM_SETCURSOR in the + DefaultAction section. */ + return 1; + + case WM_SETFOCUS: + window = __glutGetWindow(hwnd); + if (window) { + window->entryState = WM_SETFOCUS; + if (window->entry) { + __glutSetWindow(window); + window->entry(GLUT_ENTERED); + /* XXX Generation of fake passive notify? See how much + work the X11 code does to support fake passive notify + callbacks. */ + } + if (window->joystick && __glutCurrentWindow) { + if (__glutCurrentWindow->joyPollInterval > 0) { + MMRESULT result; + + /* Because Win32 will only let one window capture the + joystick at a time, we must capture it when we get the + focus and release it when we lose the focus. */ + result = joySetCapture(__glutCurrentWindow->win, + JOYSTICKID1, 0, TRUE); + if (result != JOYERR_NOERROR) { + return 0; + } + (void) joySetThreshold(JOYSTICKID1, + __glutCurrentWindow->joyPollInterval); + } + } + } + return 0; + + case WM_KILLFOCUS: + window = __glutGetWindow(hwnd); + if (window) { + window->entryState = WM_KILLFOCUS; + if (window->entry) { + __glutSetWindow(window); + window->entry(GLUT_LEFT); + } + if (window->joystick && __glutCurrentWindow) { + if (__glutCurrentWindow->joyPollInterval > 0) { + /* Because Win32 will only let one window capture the + joystick at a time, we must capture it when we get the + focus and release it when we lose the focus. */ + (void) joyReleaseCapture(JOYSTICKID1); + } + } + } + return 0; + case WM_ACTIVATE: + window = __glutGetWindow(hwnd); + /* Make sure we re-select the correct palette if needed. */ + if (LOWORD(wParam)) { + PostMessage(hwnd, WM_PALETTECHANGED, 0, 0); + } + if (window) { + int visState; + + /* HIWORD(wParam) is the minimized flag. */ + visState = !HIWORD(wParam); + updateWindowState(window, visState); + } + return 0; + + /* Colour Palette Management */ + case WM_PALETTECHANGED: + if (hwnd == (HWND)wParam) { + /* Don't respond to the message that we sent! */ + break; + } + /* fall through to WM_QUERYNEWPALETTE */ + + case WM_QUERYNEWPALETTE: + window = __glutGetWindow(hwnd); + if (window && window->colormap) { + UnrealizeObject(window->colormap->cmap); + SelectPalette(window->hdc, window->colormap->cmap, FALSE); + RealizePalette(window->hdc); + return TRUE; + } + return FALSE; + + case MM_JOY1MOVE: + case MM_JOY1ZMOVE: + window = __glutGetWindow(hwnd); + if (window->joystick) { + JOYINFOEX jix; + int x, y, z; + + /* Because WIN32 only supports messages for X, Y, and Z + translations, we must poll for the rest */ + jix.dwSize = sizeof(jix); + jix.dwFlags = JOY_RETURNALL; + joyGetPosEx(JOYSTICKID1,&jix); + +#define SCALE(v) ((int) ((v - 32767)/32.768)) + + /* Convert to integer for scaling. */ + x = jix.dwXpos; + y = jix.dwYpos; + z = jix.dwZpos; + window->joystick(jix.dwButtons, SCALE(x), SCALE(y), SCALE(z)); + + return TRUE; + } + return FALSE; + case MM_JOY1BUTTONDOWN: + case MM_JOY1BUTTONUP: + window = __glutGetWindow(hwnd); + if (window->joystick) { + JOYINFOEX jix; + + /* Because WIN32 only supports messages for X, Y, and Z + translations, we must poll for the rest */ + jix.dwSize = sizeof(jix); + jix.dwFlags = JOY_RETURNALL; + joyGetPosEx(JOYSTICKID1,&jix); + + return TRUE; + } + return FALSE; + +#if 0 + /* Miscellaneous messages (don't really need to enumerate them, + but it's good to know what you're not getting sometimes). */ + case WM_DISPLAYCHANGE: + break; + case WM_NCHITTEST: + /* This event is generated by every mouse move event. */ + goto defproc; + case WM_NCMOUSEMOVE: + goto defproc; + case WM_NCACTIVATE: + goto defproc; + case WM_NCPAINT: + goto defproc; + case WM_NCCALCSIZE: + goto defproc; + case WM_NCCREATE: + goto defproc; + case WM_NCDESTROY: + goto defproc; + case WM_NCLBUTTONDOWN: + goto defproc; + case WM_SETTEXT: + goto defproc; + case WM_GETTEXT: + goto defproc; + case WM_ACTIVATEAPP: + goto defproc; + case WM_GETICON: + goto defproc; + case WM_ERASEBKGND: + goto defproc; + case WM_WINDOWPOSCHANGING: + goto defproc; + case WM_WINDOWPOSCHANGED: + goto defproc; + case WM_MOUSEACTIVATE: + goto defproc; + case WM_SHOWWINDOW: + goto defproc; + case WM_MOVING: + goto defproc; + case WM_MOVE: + goto defproc; + case WM_KEYUP: + goto defproc; + case WM_CAPTURECHANGED: + goto defproc; + case WM_SYSCOMMAND: + goto defproc; + case WM_ENTERSIZEMOVE: + goto defproc; + case WM_ENTERIDLE: + goto defproc; +#endif + + default: + goto defproc; + } + +defproc: + return DefWindowProc(hwnd, msg, wParam, lParam); +} + +#endif + +#if defined(__OS2PM__) +Bool __glutSetWindowText(Window window, char *text) +{ + return WinSetWindowText(window, (PCSZ)text); + +} + +#endif  \ No newline at end of file -- cgit v1.2.3 From a47c222803483c208f720e3fb5015050d4d0eaf4 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sat, 7 Jun 2008 13:27:53 +0100 Subject: draw: fix temp vs output buffer typo in vsvg_run_elts --- src/gallium/auxiliary/draw/draw_vs_varient.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index d8317d3bbe..abe8c5ec2d 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -137,7 +137,7 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, vsvg->fetch->run_elts( vsvg->fetch, elts, count, - output_buffer ); + temp_buffer ); vsvg->base.vs->run_linear( vsvg->base.vs, temp_buffer, -- cgit v1.2.3 From f4535f6e5ae63d8c59428cf190a95e0eb4ae233d Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 6 Jun 2008 16:12:55 +0200 Subject: mesa: Add MESA_FORMAT_S8_Z24 texture format None of the fetch and store functions implemented. This atleast stops shadowtex from locking the GPU on i915 with the linux-dri-x86 target. It most of it looks okay, with the exception of actually displaying the texture. --- src/mesa/main/texformat.c | 35 +++++++++++++++++++++++++++++++++++ src/mesa/main/texformat.h | 2 ++ src/mesa/state_tracker/st_format.c | 6 ++++-- 3 files changed, 41 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c index 88fbd8f07c..d479bf510e 100644 --- a/src/mesa/main/texformat.c +++ b/src/mesa/main/texformat.c @@ -1207,6 +1207,41 @@ const struct gl_texture_format _mesa_texformat_z24_s8 = { store_texel_z24_s8 /* StoreTexel */ }; +const struct gl_texture_format _mesa_texformat_s8_z24 = { + MESA_FORMAT_S8_Z24, /* MesaFormat */ + GL_DEPTH_STENCIL_EXT, /* BaseFormat */ + GL_UNSIGNED_NORMALIZED_ARB, /* DataType */ + 0, /* RedBits */ + 0, /* GreenBits */ + 0, /* BlueBits */ + 0, /* AlphaBits */ + 0, /* LuminanceBits */ + 0, /* IntensityBits */ + 0, /* IndexBits */ + 24, /* DepthBits */ + 8, /* StencilBits */ + 4, /* TexelBytes */ +#if 0 + _mesa_texstore_s8_z24, /* StoreTexImageFunc */ +#else + _mesa_texstore_z24_s8, /* StoreTexImageFunc */ +#endif + NULL, /* FetchTexel1D */ + NULL, /* FetchTexel2D */ + NULL, /* FetchTexel3D */ +#if 0 + fetch_texel_1d_f_s8_z24, /* FetchTexel1Df */ + fetch_texel_2d_f_s8_z24, /* FetchTexel2Df */ + fetch_texel_3d_f_s8_z24, /* FetchTexel3Df */ + store_texel_s8_z24 /* StoreTexel */ +#else + fetch_texel_1d_f_z24_s8, /* FetchTexel1Df */ + fetch_texel_2d_f_z24_s8, /* FetchTexel2Df */ + fetch_texel_3d_f_z24_s8, /* FetchTexel3Df */ + store_texel_z24_s8 /* StoreTexel */ +#endif +}; + const struct gl_texture_format _mesa_texformat_z16 = { MESA_FORMAT_Z16, /* MesaFormat */ GL_DEPTH_COMPONENT, /* BaseFormat */ diff --git a/src/mesa/main/texformat.h b/src/mesa/main/texformat.h index 48f0fe99f2..8f4e2feb48 100644 --- a/src/mesa/main/texformat.h +++ b/src/mesa/main/texformat.h @@ -84,6 +84,7 @@ enum _format { MESA_FORMAT_YCBCR, /* YYYY YYYY UorV UorV */ MESA_FORMAT_YCBCR_REV, /* UorV UorV YYYY YYYY */ MESA_FORMAT_Z24_S8, /* ZZZZ ZZZZ ZZZZ ZZZZ ZZZZ ZZZZ SSSS SSSS */ + MESA_FORMAT_S8_Z24, /* SSSS SSSS ZZZZ ZZZZ ZZZZ ZZZZ ZZZZ ZZZZ */ MESA_FORMAT_Z16, /* ZZZZ ZZZZ ZZZZ ZZZZ */ MESA_FORMAT_Z32, /* ZZZZ ZZZZ ZZZZ ZZZZ ZZZZ ZZZZ ZZZZ ZZZZ */ /*@}*/ @@ -209,6 +210,7 @@ extern const struct gl_texture_format _mesa_texformat_l8; extern const struct gl_texture_format _mesa_texformat_i8; extern const struct gl_texture_format _mesa_texformat_ci8; extern const struct gl_texture_format _mesa_texformat_z24_s8; +extern const struct gl_texture_format _mesa_texformat_s8_z24; extern const struct gl_texture_format _mesa_texformat_z16; extern const struct gl_texture_format _mesa_texformat_z32; /*@}*/ diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 17a3cfd5a4..1b9325c159 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -274,6 +274,8 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) return PIPE_FORMAT_Z32_UNORM; case MESA_FORMAT_Z24_S8: return PIPE_FORMAT_Z24S8_UNORM; + case MESA_FORMAT_S8_Z24: + return PIPE_FORMAT_S8Z24_UNORM; case MESA_FORMAT_YCBCR: return PIPE_FORMAT_YCBCR; case MESA_FORMAT_RGB_DXT1: @@ -559,10 +561,10 @@ translate_gallium_format_to_mesa_format(enum pipe_format format) return &_mesa_texformat_z16; case PIPE_FORMAT_Z32_UNORM: return &_mesa_texformat_z32; - case PIPE_FORMAT_S8Z24_UNORM: - /* XXX fallthrough OK? */ case PIPE_FORMAT_Z24S8_UNORM: return &_mesa_texformat_z24_s8; + case PIPE_FORMAT_S8Z24_UNORM: + return &_mesa_texformat_s8_z24; case PIPE_FORMAT_YCBCR: return &_mesa_texformat_ycbcr; case PIPE_FORMAT_YCBCR_REV: -- cgit v1.2.3 From 2e3e5184176debb66bdd7f5f606cf95b7fee91bb Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 9 Jun 2008 16:29:57 +0200 Subject: mesa: Most of the functions of MESA_TEXTURE_S8_Z24 are now supported --- src/mesa/main/texformat.c | 12 +++++------- src/mesa/main/texformat_tmp.h | 26 ++++++++++++++++++++++++++ 2 files changed, 31 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c index d479bf510e..17859a9fec 100644 --- a/src/mesa/main/texformat.c +++ b/src/mesa/main/texformat.c @@ -1229,17 +1229,10 @@ const struct gl_texture_format _mesa_texformat_s8_z24 = { NULL, /* FetchTexel1D */ NULL, /* FetchTexel2D */ NULL, /* FetchTexel3D */ -#if 0 fetch_texel_1d_f_s8_z24, /* FetchTexel1Df */ fetch_texel_2d_f_s8_z24, /* FetchTexel2Df */ fetch_texel_3d_f_s8_z24, /* FetchTexel3Df */ store_texel_s8_z24 /* StoreTexel */ -#else - fetch_texel_1d_f_z24_s8, /* FetchTexel1Df */ - fetch_texel_2d_f_z24_s8, /* FetchTexel2Df */ - fetch_texel_3d_f_z24_s8, /* FetchTexel3Df */ - store_texel_z24_s8 /* StoreTexel */ -#endif }; const struct gl_texture_format _mesa_texformat_z16 = { @@ -1675,6 +1668,11 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format, *comps = 1; /* XXX OK? */ return; + case MESA_FORMAT_S8_Z24: + *datatype = GL_UNSIGNED_INT; + *comps = 1; /* XXX OK? */ + return; + case MESA_FORMAT_Z16: *datatype = GL_UNSIGNED_SHORT; *comps = 1; diff --git a/src/mesa/main/texformat_tmp.h b/src/mesa/main/texformat_tmp.h index 99785da1a0..63939f4011 100644 --- a/src/mesa/main/texformat_tmp.h +++ b/src/mesa/main/texformat_tmp.h @@ -1363,6 +1363,32 @@ static void store_texel_z24_s8(struct gl_texture_image *texImage, #endif +/* MESA_TEXFORMAT_S8_Z24 ***************************************************/ + +static void FETCH(f_s8_z24)( const struct gl_texture_image *texImage, + GLint i, GLint j, GLint k, GLfloat *texel ) +{ + /* only return Z, not stencil data */ + const GLuint *src = TEXEL_ADDR(GLuint, texImage, i, j, k, 1); + const GLfloat scale = 1.0F / (GLfloat) 0xffffff; + texel[0] = ((*src) & 0x00ffffff) * scale; + ASSERT(texImage->TexFormat->MesaFormat == MESA_FORMAT_S8_Z24); + ASSERT(texel[0] >= 0.0F); + ASSERT(texel[0] <= 1.0F); +} + +#if DIM == 3 +static void store_texel_s8_z24(struct gl_texture_image *texImage, + GLint i, GLint j, GLint k, const void *texel) +{ + /* only store Z, not stencil */ + GLuint *dst = TEXEL_ADDR(GLuint, texImage, i, j, k, 1); + GLfloat depth = *((GLfloat *) texel); + GLuint zi = (GLuint) (depth * 0xffffff); + *dst = zi | (*dst & 0xff000000); +} +#endif + #undef TEXEL_ADDR #undef DIM -- cgit v1.2.3 From cfc23bc54c8cae2615d447bc199ff87ef7e9298e Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 9 Jun 2008 16:17:35 +0200 Subject: i915: Disable color buffer writes if no color buffer is attached --- src/gallium/drivers/i915simple/i915_state_immediate.c | 10 +++++++--- 1 file changed, 7 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state_immediate.c b/src/gallium/drivers/i915simple/i915_state_immediate.c index 704ea4d838..2501f2d7cb 100644 --- a/src/gallium/drivers/i915simple/i915_state_immediate.c +++ b/src/gallium/drivers/i915simple/i915_state_immediate.c @@ -156,8 +156,12 @@ const struct i915_tracked_state i915_upload_S5 = { */ static void upload_S6( struct i915_context *i915 ) { - unsigned LIS6 = (S6_COLOR_WRITE_ENABLE | - (2 << S6_TRISTRIP_PV_SHIFT)); + unsigned LIS6 = (2 << S6_TRISTRIP_PV_SHIFT); + + /* I915_NEW_FRAMEBUFFER + */ + if (i915->framebuffer.cbufs[0]) + LIS6 |= S6_COLOR_WRITE_ENABLE; /* I915_NEW_BLEND */ @@ -174,7 +178,7 @@ static void upload_S6( struct i915_context *i915 ) } const struct i915_tracked_state i915_upload_S6 = { - I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL, + I915_NEW_BLEND | I915_NEW_DEPTH_STENCIL | I915_NEW_FRAMEBUFFER, upload_S6 }; -- cgit v1.2.3 From 1c316f1e824b094977059145a1abcdb50a391f1c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 09:27:52 -0600 Subject: gallium: disable a tgsi_dump() call --- src/gallium/auxiliary/draw/draw_vs_aos.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 45f36cff2a..458bbbe376 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -2066,7 +2066,9 @@ static struct draw_vs_varient *varient_aos_sse( struct draw_vertex_shader *vs, if (!vaos->attrib) goto fail; +#if 0 tgsi_dump(vs->state.tokens, 0); +#endif if (!build_vertex_program( vaos, TRUE )) goto fail; -- cgit v1.2.3 From 1954f49a171036c633614450e591307833421aca Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Sun, 8 Jun 2008 23:05:15 -0400 Subject: glsl: fix array size initialiazers using const variables e.g. const int kernelSize = 9; uniform vec2 kernel[kernelSize]; --- src/mesa/shader/slang/slang_compile.c | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 46b5c54bbe..f2e8aab8b0 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -258,9 +258,33 @@ parse_array_len(slang_parse_ctx * C, slang_output_ctx * O, GLuint * len) /* evaluate compile-time expression which is array size */ _slang_simplify(&array_size, &space, C->atoms); - result = (array_size.type == SLANG_OPER_LITERAL_INT); - *len = (GLint) array_size.literal[0]; + if (array_size.type == SLANG_OPER_LITERAL_INT) { + result = GL_TRUE; + *len = (GLint) array_size.literal[0]; + } else if (array_size.type == SLANG_OPER_IDENTIFIER) { + slang_variable *var = _slang_locate_variable(array_size.locals, array_size.a_id, GL_TRUE); + if (!var) { + slang_info_log_error(C->L, "undefined variable '%s'", + (char *) array_size.a_id); + result = GL_FALSE; + } else if (var->type.qualifier == SLANG_QUAL_CONST && + var->type.specifier.type == SLANG_SPEC_INT) { + if (var->initializer && + var->initializer->type == SLANG_OPER_LITERAL_INT) { + *len = (GLint) var->initializer->literal[0]; + result = GL_TRUE; + } else { + slang_info_log_error(C->L, "unable to parse array size declaration"); + result = GL_FALSE; + } + } else { + slang_info_log_error(C->L, "unable to parse array size declaration"); + result = GL_FALSE; + } + } else { + result = GL_FALSE; + } slang_operation_destruct(&array_size); return result; -- cgit v1.2.3 From 50d59c8be1e409f0f96ba5c59d74eb5ff8c35b6a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 11:47:18 -0600 Subject: egl: choose between DRI and softpipe drivers --- src/egl/main/eglx.c | 58 ++++++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 55 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglx.c b/src/egl/main/eglx.c index 085da33e7d..63ad5a7074 100644 --- a/src/egl/main/eglx.c +++ b/src/egl/main/eglx.c @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + /** * X-specific EGL code. @@ -8,22 +35,34 @@ #include #include +#include #include + +#include "egldriver.h" +#include "egllog.h" +#include "eglstring.h" #include "eglx.h" -static const char *DefaultXDriver = "softpipe_egl"; +static const char *DefaultDRIDriver = "egl_xdri"; +static const char *DefaultSoftDriver = "softpipe_egl"; /** * Given an X Display ptr (at dpy->Xdpy) try to determine the appropriate * device driver. Return its name. + * + * This boils down to whether to use the egl_xdri.so driver which will + * load a DRI driver or the egl_softpipe.so driver that'll do software + * rendering on Xlib. */ const char * _xeglChooseDriver(_EGLDisplay *dpy) { #ifdef _EGL_PLATFORM_X _XPrivDisplay xdpy; + int screen; + const char *driverName; assert(dpy); @@ -38,9 +77,22 @@ _xeglChooseDriver(_EGLDisplay *dpy) assert(dpy->Xdpy); - printf("%s\n", xdpy->display_name); + screen = DefaultScreen(dpy->Xdpy); + + /* See if we can choose a DRI/DRM driver */ + driverName = _eglChooseDRMDriver(screen); + if (driverName) { + /* DRI is available */ + free((void *) driverName); + driverName = _eglstrdup(DefaultDRIDriver); + } + else { + driverName = _eglstrdup(DefaultSoftDriver); + } + + _eglLog(_EGL_DEBUG, "_xeglChooseDriver: %s", driverName); - return DefaultXDriver; /* XXX temporary */ + return driverName; #else return NULL; #endif -- cgit v1.2.3 From 6420a62cf0f5812708a4c649691ba62c573eaced Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 12:26:20 -0600 Subject: egl: s/softpipe_egl.so/egl_softpipe.so/ --- src/egl/main/eglx.c | 2 +- src/gallium/winsys/egl_xlib/Makefile | 6 +++--- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglx.c b/src/egl/main/eglx.c index 63ad5a7074..1bba149f70 100644 --- a/src/egl/main/eglx.c +++ b/src/egl/main/eglx.c @@ -45,7 +45,7 @@ static const char *DefaultDRIDriver = "egl_xdri"; -static const char *DefaultSoftDriver = "softpipe_egl"; +static const char *DefaultSoftDriver = "egl_softpipe"; /** diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile index 46666650b0..d5595b819a 100644 --- a/src/gallium/winsys/egl_xlib/Makefile +++ b/src/gallium/winsys/egl_xlib/Makefile @@ -1,13 +1,13 @@ # src/gallium/winsys/egl_xlib/Makefile -# Build softpipe/xlib/EGL driver library/object: "softpipe_egl.so" +# Build softpipe/xlib/EGL driver library/object: "egl_softpipe.so" TOP = ../../../.. include $(TOP)/configs/current -DRIVER_NAME = softpipe_egl.so +DRIVER_NAME = egl_softpipe.so INCLUDE_DIRS = \ @@ -49,7 +49,7 @@ default: depend library Makefile library: $(TOP)/$(LIB_DIR)/$(DRIVER_NAME) -# Make the softpipe_egl.so library +# Make the egl_softpipe.so library $(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(WINSYS_OBJECTS) $(LIBS) $(TOP)/bin/mklib -o $(DRIVER_NAME) \ -linker "$(CC)" \ -- cgit v1.2.3 From 44c347c49e5b7b4be01055762da4727de2a987ff Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 13:33:00 -0600 Subject: mesa: get rid of FBDEV/SVGA/etc bits --- src/mesa/sources | 16 ++-------------- 1 file changed, 2 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index c9ecd7dbf2..41b57b07c6 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -288,16 +288,6 @@ SPARC_API = \ __COMMON_DRIVER_SOURCES = \ drivers/common/driverfuncs.c -SVGA_DRIVER_SOURCES = \ - drivers/svga/svgamesa.c \ - drivers/svga/svgamesa8.c \ - drivers/svga/svgamesa15.c \ - drivers/svga/svgamesa16.c \ - drivers/svga/svgamesa24.c \ - drivers/svga/svgamesa32.c - -FBDEV_DRIVER_SOURCES = \ - drivers/fbdev/glfbdev.c ### All the core C sources @@ -306,8 +296,8 @@ ALL_SOURCES = \ $(GLAPI_SOURCES) \ $(SOLO_SOURCES) \ $(MESA_ASM_SOURCES) \ - $(COMMON_DRIVER_SOURCES)\ - $(FBDEV_DRIVER_SOURCES) + $(COMMON_DRIVER_SOURCES) + SOLO_SOURCES = \ $(MAIN_SOURCES) \ @@ -346,8 +336,6 @@ CORE_OBJECTS = $(SOLO_OBJECTS) $(GLAPI_OBJECTS) COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o) -FBDEV_DRIVER_OBJECTS = $(FBDEV_DRIVER_SOURCES:.c=.o) - ### Include directories -- cgit v1.2.3 From 5f621f68dd449f6e5e141a85718e2a2c5d2a56fb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 13:35:34 -0600 Subject: mesa: remove some obsolete gallium bits --- src/mesa/sources | 5 ----- 1 file changed, 5 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 41b57b07c6..8a809c6d35 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -304,11 +304,6 @@ SOLO_SOURCES = \ $(MATH_SOURCES) \ $(VBO_SOURCES) \ $(VF_SOURCES) \ - $(DRAW_SOURCES) \ - $(TGSIEXEC_SOURCES) \ - $(TGSIUTIL_SOURCES) \ - $(PIPEUTIL_SOURCES) \ - $(STATECACHE_SOURCES) \ $(STATETRACKER_SOURCES) \ $(TNL_SOURCES) \ $(SHADER_SOURCES) \ -- cgit v1.2.3 From 137dbe7eb1f2b955552a06ea6b9179419947981b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 13:38:23 -0600 Subject: mesa: disable VF_SOURCES --- src/mesa/sources | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 8a809c6d35..5d02583f15 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -151,7 +151,7 @@ VBO_SOURCES = \ vbo/vbo_save_draw.c \ vbo/vbo_save_loopback.c -VF_SOURCES = \ +__VF_SOURCES = \ vf/vf.c \ vf/vf_generic.c \ vf/vf_sse.c @@ -303,7 +303,6 @@ SOLO_SOURCES = \ $(MAIN_SOURCES) \ $(MATH_SOURCES) \ $(VBO_SOURCES) \ - $(VF_SOURCES) \ $(STATETRACKER_SOURCES) \ $(TNL_SOURCES) \ $(SHADER_SOURCES) \ -- cgit v1.2.3 From ac7f416bda4506583ef577d62fee4a4d0ce890c4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 13:51:21 -0600 Subject: more old stuff, rename SOLO->MESA, etc --- src/mesa/sources | 20 +++++++------------- 1 file changed, 7 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 5d02583f15..ac0b1c1950 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -292,14 +292,7 @@ __COMMON_DRIVER_SOURCES = \ ### All the core C sources -ALL_SOURCES = \ - $(GLAPI_SOURCES) \ - $(SOLO_SOURCES) \ - $(MESA_ASM_SOURCES) \ - $(COMMON_DRIVER_SOURCES) - - -SOLO_SOURCES = \ +MESA_SOURCES = \ $(MAIN_SOURCES) \ $(MATH_SOURCES) \ $(VBO_SOURCES) \ @@ -311,22 +304,23 @@ SOLO_SOURCES = \ $(ASM_C_SOURCES) \ $(SLANG_SOURCES) -CORE_SOURCES = \ +ALL_SOURCES = \ + $(MESA_SOURCES) \ $(GLAPI_SOURCES) \ - $(SOLO_SOURCES) + $(MESA_ASM_SOURCES) \ + $(COMMON_DRIVER_SOURCES) ### Object files -SOLO_OBJECTS = \ - $(SOLO_SOURCES:.c=.o) \ +MESA_OBJECTS = \ + $(MESA_SOURCES:.c=.o) \ $(MESA_ASM_SOURCES:.S=.o) GLAPI_OBJECTS = \ $(GLAPI_SOURCES:.c=.o) \ $(GLAPI_ASM_SOURCES:.S=.o) -CORE_OBJECTS = $(SOLO_OBJECTS) $(GLAPI_OBJECTS) COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o) -- cgit v1.2.3 From d7205ba971a348a57d50c331cb49afbd81067736 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 13:51:36 -0600 Subject: Makefile clean-ups --- src/mesa/drivers/directfb/Makefile | 4 ++-- src/mesa/drivers/fbdev/Makefile | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/directfb/Makefile b/src/mesa/drivers/directfb/Makefile index 945eb8b3d9..ece0457cff 100644 --- a/src/mesa/drivers/directfb/Makefile +++ b/src/mesa/drivers/directfb/Makefile @@ -36,10 +36,10 @@ default: directfb-libgl directfbgl_mesa # XXX this used to be in src/mesa/Makefile and is probably broken now -directfb-libgl: $(CORE_OBJECTS) +directfb-libgl: $(LIBS) @ $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' -ldflags '$(LDFLAGS)' \ -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ - -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) $(CORE_OBJECTS) \ + -install $(TOP)/$(LIB_DIR) $(MKLIB_OPTIONS) $(LIBS) \ $(GL_LIB_DEPS) diff --git a/src/mesa/drivers/fbdev/Makefile b/src/mesa/drivers/fbdev/Makefile index ac43dfdb67..c0ef54f604 100644 --- a/src/mesa/drivers/fbdev/Makefile +++ b/src/mesa/drivers/fbdev/Makefile @@ -24,7 +24,7 @@ CORE_MESA = $(TOP)/src/mesa/libmesa.a $(TOP)/src/mesa/libglapi.a default: $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) -$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(CORE_OBJECTS) $(OBJECTS) +$(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(CORE_MESA) $(OBJECTS) @ $(TOP)/bin/mklib -o $(GL_LIB) \ -linker '$(CC)' -ldflags '$(LDFLAGS)' \ -major $(MESA_MAJOR) -minor $(MESA_MINOR) -patch $(MESA_TINY) \ -- cgit v1.2.3 From 4f15e3eefbd31b4fdcf90c2798d0cb41c893b049 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 14:03:01 -0600 Subject: mesa: chmod a-x context.c --- src/mesa/main/context.c | 0 1 file changed, 0 insertions(+), 0 deletions(-) mode change 100755 => 100644 src/mesa/main/context.c (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c old mode 100755 new mode 100644 -- cgit v1.2.3 From ffeb90169fd54d6e364a308a9b997437426e1b83 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 14:09:28 -0600 Subject: s/SOLO/MESA/ --- src/mesa/Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index d39f282b33..a301625179 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -29,8 +29,8 @@ default: depend asm_subdirs libmesa.a libglapi.a driver_subdirs # Helper libraries used by many drivers: # Make archive of core mesa object files -libmesa.a: $(SOLO_OBJECTS) - @ $(TOP)/bin/mklib -o mesa -static $(SOLO_OBJECTS) +libmesa.a: $(MESA_OBJECTS) + @ $(TOP)/bin/mklib -o mesa -static $(MESA_OBJECTS) # Make archive of gl* API dispatcher functions only libglapi.a: $(GLAPI_OBJECTS) -- cgit v1.2.3 From f26baad2e1e8cb3c24fa64cc31869ec7b27d71ff Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 14:14:34 -0600 Subject: mesa: refactor: move glPixelStore function into new pixelstore.c file --- src/mesa/main/pixel.c | 196 --------------------------------------- src/mesa/main/pixel.h | 7 -- src/mesa/main/pixelstore.c | 226 +++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/pixelstore.h | 46 +++++++++ src/mesa/main/sources | 2 + src/mesa/main/state.c | 1 + src/mesa/sources | 1 + 7 files changed, 276 insertions(+), 203 deletions(-) create mode 100644 src/mesa/main/pixelstore.c create mode 100644 src/mesa/main/pixelstore.h (limited to 'src') diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c index 7eeae05dbd..1e2d31e172 100644 --- a/src/mesa/main/pixel.c +++ b/src/mesa/main/pixel.c @@ -36,8 +36,6 @@ /***** glPixelZoom *****/ /**********************************************************************/ - - void GLAPIENTRY _mesa_PixelZoom( GLfloat xfactor, GLfloat yfactor ) { @@ -54,200 +52,6 @@ _mesa_PixelZoom( GLfloat xfactor, GLfloat yfactor ) -/**********************************************************************/ -/***** glPixelStore *****/ -/**********************************************************************/ - - -void GLAPIENTRY -_mesa_PixelStorei( GLenum pname, GLint param ) -{ - /* NOTE: this call can't be compiled into the display list */ - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - switch (pname) { - case GL_PACK_SWAP_BYTES: - if (param == (GLint)ctx->Pack.SwapBytes) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Pack.SwapBytes = param ? GL_TRUE : GL_FALSE; - break; - case GL_PACK_LSB_FIRST: - if (param == (GLint)ctx->Pack.LsbFirst) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Pack.LsbFirst = param ? GL_TRUE : GL_FALSE; - break; - case GL_PACK_ROW_LENGTH: - if (param<0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Pack.RowLength == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Pack.RowLength = param; - break; - case GL_PACK_IMAGE_HEIGHT: - if (param<0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Pack.ImageHeight == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Pack.ImageHeight = param; - break; - case GL_PACK_SKIP_PIXELS: - if (param<0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Pack.SkipPixels == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Pack.SkipPixels = param; - break; - case GL_PACK_SKIP_ROWS: - if (param<0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Pack.SkipRows == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Pack.SkipRows = param; - break; - case GL_PACK_SKIP_IMAGES: - if (param<0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Pack.SkipImages == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Pack.SkipImages = param; - break; - case GL_PACK_ALIGNMENT: - if (param!=1 && param!=2 && param!=4 && param!=8) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Pack.Alignment == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Pack.Alignment = param; - break; - case GL_PACK_INVERT_MESA: - if (!ctx->Extensions.MESA_pack_invert) { - _mesa_error( ctx, GL_INVALID_ENUM, "glPixelstore(pname)" ); - return; - } - if (ctx->Pack.Invert == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Pack.Invert = param; - break; - - case GL_UNPACK_SWAP_BYTES: - if (param == (GLint)ctx->Unpack.SwapBytes) - return; - if ((GLint)ctx->Unpack.SwapBytes == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Unpack.SwapBytes = param ? GL_TRUE : GL_FALSE; - break; - case GL_UNPACK_LSB_FIRST: - if (param == (GLint)ctx->Unpack.LsbFirst) - return; - if ((GLint)ctx->Unpack.LsbFirst == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Unpack.LsbFirst = param ? GL_TRUE : GL_FALSE; - break; - case GL_UNPACK_ROW_LENGTH: - if (param<0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Unpack.RowLength == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Unpack.RowLength = param; - break; - case GL_UNPACK_IMAGE_HEIGHT: - if (param<0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Unpack.ImageHeight == param) - return; - - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Unpack.ImageHeight = param; - break; - case GL_UNPACK_SKIP_PIXELS: - if (param<0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Unpack.SkipPixels == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Unpack.SkipPixels = param; - break; - case GL_UNPACK_SKIP_ROWS: - if (param<0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Unpack.SkipRows == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Unpack.SkipRows = param; - break; - case GL_UNPACK_SKIP_IMAGES: - if (param < 0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); - return; - } - if (ctx->Unpack.SkipImages == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Unpack.SkipImages = param; - break; - case GL_UNPACK_ALIGNMENT: - if (param!=1 && param!=2 && param!=4 && param!=8) { - _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore" ); - return; - } - if (ctx->Unpack.Alignment == param) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Unpack.Alignment = param; - break; - case GL_UNPACK_CLIENT_STORAGE_APPLE: - if (param == (GLint)ctx->Unpack.ClientStorage) - return; - FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); - ctx->Unpack.ClientStorage = param ? GL_TRUE : GL_FALSE; - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glPixelStore" ); - return; - } -} - - -void GLAPIENTRY -_mesa_PixelStoref( GLenum pname, GLfloat param ) -{ - _mesa_PixelStorei( pname, (GLint) param ); -} - - - /**********************************************************************/ /***** glPixelMap *****/ /**********************************************************************/ diff --git a/src/mesa/main/pixel.h b/src/mesa/main/pixel.h index 3ba5b6689a..05bddf5ccf 100644 --- a/src/mesa/main/pixel.h +++ b/src/mesa/main/pixel.h @@ -56,13 +56,6 @@ _mesa_PixelMapuiv(GLenum map, GLsizei mapsize, const GLuint *values ); extern void GLAPIENTRY _mesa_PixelMapusv(GLenum map, GLsizei mapsize, const GLushort *values ); -extern void GLAPIENTRY -_mesa_PixelStoref( GLenum pname, GLfloat param ); - - -extern void GLAPIENTRY -_mesa_PixelStorei( GLenum pname, GLint param ); - extern void GLAPIENTRY _mesa_PixelTransferf( GLenum pname, GLfloat param ); diff --git a/src/mesa/main/pixelstore.c b/src/mesa/main/pixelstore.c new file mode 100644 index 0000000000..f5f054f938 --- /dev/null +++ b/src/mesa/main/pixelstore.c @@ -0,0 +1,226 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file pixelstore.c + * glPixelStore functions. + */ + + +#include "glheader.h" +#include "bufferobj.h" +#include "colormac.h" +#include "context.h" +#include "image.h" +#include "macros.h" +#include "pixelstore.h" +#include "mtypes.h" + + +void GLAPIENTRY +_mesa_PixelStorei( GLenum pname, GLint param ) +{ + /* NOTE: this call can't be compiled into the display list */ + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + switch (pname) { + case GL_PACK_SWAP_BYTES: + if (param == (GLint)ctx->Pack.SwapBytes) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Pack.SwapBytes = param ? GL_TRUE : GL_FALSE; + break; + case GL_PACK_LSB_FIRST: + if (param == (GLint)ctx->Pack.LsbFirst) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Pack.LsbFirst = param ? GL_TRUE : GL_FALSE; + break; + case GL_PACK_ROW_LENGTH: + if (param<0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Pack.RowLength == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Pack.RowLength = param; + break; + case GL_PACK_IMAGE_HEIGHT: + if (param<0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Pack.ImageHeight == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Pack.ImageHeight = param; + break; + case GL_PACK_SKIP_PIXELS: + if (param<0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Pack.SkipPixels == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Pack.SkipPixels = param; + break; + case GL_PACK_SKIP_ROWS: + if (param<0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Pack.SkipRows == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Pack.SkipRows = param; + break; + case GL_PACK_SKIP_IMAGES: + if (param<0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Pack.SkipImages == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Pack.SkipImages = param; + break; + case GL_PACK_ALIGNMENT: + if (param!=1 && param!=2 && param!=4 && param!=8) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Pack.Alignment == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Pack.Alignment = param; + break; + case GL_PACK_INVERT_MESA: + if (!ctx->Extensions.MESA_pack_invert) { + _mesa_error( ctx, GL_INVALID_ENUM, "glPixelstore(pname)" ); + return; + } + if (ctx->Pack.Invert == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Pack.Invert = param; + break; + + case GL_UNPACK_SWAP_BYTES: + if (param == (GLint)ctx->Unpack.SwapBytes) + return; + if ((GLint)ctx->Unpack.SwapBytes == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Unpack.SwapBytes = param ? GL_TRUE : GL_FALSE; + break; + case GL_UNPACK_LSB_FIRST: + if (param == (GLint)ctx->Unpack.LsbFirst) + return; + if ((GLint)ctx->Unpack.LsbFirst == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Unpack.LsbFirst = param ? GL_TRUE : GL_FALSE; + break; + case GL_UNPACK_ROW_LENGTH: + if (param<0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Unpack.RowLength == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Unpack.RowLength = param; + break; + case GL_UNPACK_IMAGE_HEIGHT: + if (param<0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Unpack.ImageHeight == param) + return; + + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Unpack.ImageHeight = param; + break; + case GL_UNPACK_SKIP_PIXELS: + if (param<0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Unpack.SkipPixels == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Unpack.SkipPixels = param; + break; + case GL_UNPACK_SKIP_ROWS: + if (param<0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Unpack.SkipRows == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Unpack.SkipRows = param; + break; + case GL_UNPACK_SKIP_IMAGES: + if (param < 0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore(param)" ); + return; + } + if (ctx->Unpack.SkipImages == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Unpack.SkipImages = param; + break; + case GL_UNPACK_ALIGNMENT: + if (param!=1 && param!=2 && param!=4 && param!=8) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPixelStore" ); + return; + } + if (ctx->Unpack.Alignment == param) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Unpack.Alignment = param; + break; + case GL_UNPACK_CLIENT_STORAGE_APPLE: + if (param == (GLint)ctx->Unpack.ClientStorage) + return; + FLUSH_VERTICES(ctx, _NEW_PACKUNPACK); + ctx->Unpack.ClientStorage = param ? GL_TRUE : GL_FALSE; + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glPixelStore" ); + return; + } +} + + +void GLAPIENTRY +_mesa_PixelStoref( GLenum pname, GLfloat param ) +{ + _mesa_PixelStorei( pname, (GLint) param ); +} diff --git a/src/mesa/main/pixelstore.h b/src/mesa/main/pixelstore.h new file mode 100644 index 0000000000..c42f304030 --- /dev/null +++ b/src/mesa/main/pixelstore.h @@ -0,0 +1,46 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file pixelstore.h + * glPixelStore functions. + */ + + +#ifndef PIXELSTORE_H +#define PIXELSTORE_H + + +#include "glheader.h" + + +extern void GLAPIENTRY +_mesa_PixelStorei( GLenum pname, GLint param ); + + +extern void GLAPIENTRY +_mesa_PixelStoref( GLenum pname, GLfloat param ); + + +#endif diff --git a/src/mesa/main/sources b/src/mesa/main/sources index dfcff89e4b..ac809b22d2 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -42,6 +42,7 @@ mipmap.c \ mm.c \ occlude.c \ pixel.c \ +pixelstore.c \ points.c \ polygon.c \ rastpos.c \ @@ -115,6 +116,7 @@ mm.h \ mtypes.h \ occlude.h \ pixel.h \ +pixelstore.h \ points.h \ polygon.h \ rastpos.h \ diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 7e8f964bb1..13a079a2dc 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -72,6 +72,7 @@ #include "macros.h" #include "matrix.h" #include "pixel.h" +#include "pixelstore.h" #include "points.h" #include "polygon.h" #if FEATURE_ARB_occlusion_query || FEATURE_EXT_timer_query diff --git a/src/mesa/sources b/src/mesa/sources index ac0b1c1950..6a1d55f1fe 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -43,6 +43,7 @@ MAIN_SOURCES = \ main/mipmap.c \ main/mm.c \ main/pixel.c \ + main/pixelstore.c \ main/points.c \ main/polygon.c \ main/queryobj.c \ -- cgit v1.2.3 From d960a0621d65ae9977efe9bbb51dce9e1571b114 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 14:22:15 -0600 Subject: mesa: refactor: move glReadPixels code into new readpix.c file --- src/mesa/main/drawpix.c | 163 +---------------------------------------- src/mesa/main/drawpix.h | 12 +-- src/mesa/main/readpix.c | 191 ++++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/readpix.h | 42 +++++++++++ src/mesa/main/sources | 2 + src/mesa/main/state.c | 5 +- src/mesa/sources | 1 + 7 files changed, 244 insertions(+), 172 deletions(-) create mode 100644 src/mesa/main/readpix.c create mode 100644 src/mesa/main/readpix.h (limited to 'src') diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c index 016ddd0a81..6db198ea3b 100644 --- a/src/mesa/main/drawpix.c +++ b/src/mesa/main/drawpix.c @@ -30,114 +30,10 @@ #include "feedback.h" #include "framebuffer.h" #include "image.h" +#include "readpix.h" #include "state.h" -/** - * Do error checking of the format/type parameters to glReadPixels and - * glDrawPixels. - * \param drawing if GL_TRUE do checking for DrawPixels, else do checking - * for ReadPixels. - * \return GL_TRUE if error detected, GL_FALSE if no errors - */ -static GLboolean -error_check_format_type(GLcontext *ctx, GLenum format, GLenum type, - GLboolean drawing) -{ - const char *readDraw = drawing ? "Draw" : "Read"; - - if (ctx->Extensions.EXT_packed_depth_stencil - && type == GL_UNSIGNED_INT_24_8_EXT - && format != GL_DEPTH_STENCIL_EXT) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "gl%sPixels(format is not GL_DEPTH_STENCIL_EXT)", readDraw); - return GL_TRUE; - } - - /* basic combinations test */ - if (!_mesa_is_legal_format_and_type(ctx, format, type)) { - _mesa_error(ctx, GL_INVALID_ENUM, - "gl%sPixels(format or type)", readDraw); - return GL_TRUE; - } - - /* additional checks */ - switch (format) { - case GL_RED: - case GL_GREEN: - case GL_BLUE: - case GL_ALPHA: - case GL_LUMINANCE: - case GL_LUMINANCE_ALPHA: - case GL_RGB: - case GL_BGR: - case GL_RGBA: - case GL_BGRA: - case GL_ABGR_EXT: - if (drawing && !ctx->Visual.rgbMode) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glDrawPixels(drawing RGB pixels into color index buffer)"); - return GL_TRUE; - } - if (!drawing && !_mesa_dest_buffer_exists(ctx, GL_COLOR)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glReadPixels(no color buffer)"); - return GL_TRUE; - } - break; - case GL_COLOR_INDEX: - if (!drawing && ctx->Visual.rgbMode) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glReadPixels(reading color index format from RGB buffer)"); - return GL_TRUE; - } - if (!drawing && !_mesa_dest_buffer_exists(ctx, GL_COLOR)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glReadPixels(no color buffer)"); - return GL_TRUE; - } - break; - case GL_STENCIL_INDEX: - if ((drawing && !_mesa_dest_buffer_exists(ctx, format)) || - (!drawing && !_mesa_source_buffer_exists(ctx, format))) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "gl%sPixels(no stencil buffer)", readDraw); - return GL_TRUE; - } - break; - case GL_DEPTH_COMPONENT: - if ((drawing && !_mesa_dest_buffer_exists(ctx, format)) || - (!drawing && !_mesa_source_buffer_exists(ctx, format))) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "gl%sPixels(no depth buffer)", readDraw); - return GL_TRUE; - } - break; - case GL_DEPTH_STENCIL_EXT: - if (!ctx->Extensions.EXT_packed_depth_stencil || - type != GL_UNSIGNED_INT_24_8_EXT) { - _mesa_error(ctx, GL_INVALID_ENUM, "gl%sPixels(type)", readDraw); - return GL_TRUE; - } - if ((drawing && !_mesa_dest_buffer_exists(ctx, format)) || - (!drawing && !_mesa_source_buffer_exists(ctx, format))) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "gl%sPixels(no depth or stencil buffer)", readDraw); - return GL_TRUE; - } - break; - default: - /* this should have been caught in _mesa_is_legal_format_type() */ - _mesa_problem(ctx, "unexpected format in _mesa_%sPixels", readDraw); - return GL_TRUE; - } - - /* no errors */ - return GL_FALSE; -} - - - #if _HAVE_FULL_GL /* @@ -165,7 +61,7 @@ _mesa_DrawPixels( GLsizei width, GLsizei height, return; } - if (error_check_format_type(ctx, format, type, GL_TRUE)) { + if (_mesa_error_check_format_type(ctx, format, type, GL_TRUE)) { /* found an error */ return; } @@ -287,61 +183,6 @@ _mesa_CopyPixels( GLint srcx, GLint srcy, GLsizei width, GLsizei height, -void GLAPIENTRY -_mesa_ReadPixels( GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, GLvoid *pixels ) -{ - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); - - FLUSH_CURRENT(ctx, 0); - - if (width < 0 || height < 0) { - _mesa_error( ctx, GL_INVALID_VALUE, - "glReadPixels(width=%d height=%d)", width, height ); - return; - } - - if (ctx->NewState) - _mesa_update_state(ctx); - - if (error_check_format_type(ctx, format, type, GL_FALSE)) { - /* found an error */ - return; - } - - if (ctx->ReadBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { - _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT, - "glReadPixels(incomplete framebuffer)" ); - return; - } - - if (!_mesa_source_buffer_exists(ctx, format)) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(no readbuffer)"); - return; - } - - if (ctx->Pack.BufferObj->Name) { - if (!_mesa_validate_pbo_access(2, &ctx->Pack, width, height, 1, - format, type, pixels)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glReadPixels(invalid PBO access)"); - return; - } - - if (ctx->Pack.BufferObj->Pointer) { - /* buffer is mapped - that's an error */ - _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(PBO is mapped)"); - return; - } - } - - ctx->Driver.ReadPixels(ctx, x, y, width, height, - format, type, &ctx->Pack, pixels); -} - - - void GLAPIENTRY _mesa_Bitmap( GLsizei width, GLsizei height, GLfloat xorig, GLfloat yorig, GLfloat xmove, GLfloat ymove, diff --git a/src/mesa/main/drawpix.h b/src/mesa/main/drawpix.h index 2a2d7de8d8..6177adad6d 100644 --- a/src/mesa/main/drawpix.h +++ b/src/mesa/main/drawpix.h @@ -1,9 +1,8 @@ - /* * Mesa 3-D graphics library - * Version: 3.5 + * Version: 7.1 * - * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -28,7 +27,7 @@ #define DRAWPIXELS_H -#include "mtypes.h" +#include "main/glheader.h" extern void GLAPIENTRY @@ -36,11 +35,6 @@ _mesa_DrawPixels( GLsizei width, GLsizei height, GLenum format, GLenum type, const GLvoid *pixels ); -extern void GLAPIENTRY -_mesa_ReadPixels( GLint x, GLint y, GLsizei width, GLsizei height, - GLenum format, GLenum type, GLvoid *pixels ); - - extern void GLAPIENTRY _mesa_CopyPixels( GLint srcx, GLint srcy, GLsizei width, GLsizei height, GLenum type ); diff --git a/src/mesa/main/readpix.c b/src/mesa/main/readpix.c new file mode 100644 index 0000000000..dc22808e5b --- /dev/null +++ b/src/mesa/main/readpix.c @@ -0,0 +1,191 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "glheader.h" +#include "imports.h" +#include "bufferobj.h" +#include "context.h" +#include "readpix.h" +#include "framebuffer.h" +#include "image.h" +#include "state.h" + + +/** + * Do error checking of the format/type parameters to glReadPixels and + * glDrawPixels. + * \param drawing if GL_TRUE do checking for DrawPixels, else do checking + * for ReadPixels. + * \return GL_TRUE if error detected, GL_FALSE if no errors + */ +GLboolean +_mesa_error_check_format_type(GLcontext *ctx, GLenum format, GLenum type, + GLboolean drawing) +{ + const char *readDraw = drawing ? "Draw" : "Read"; + + if (ctx->Extensions.EXT_packed_depth_stencil + && type == GL_UNSIGNED_INT_24_8_EXT + && format != GL_DEPTH_STENCIL_EXT) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "gl%sPixels(format is not GL_DEPTH_STENCIL_EXT)", readDraw); + return GL_TRUE; + } + + /* basic combinations test */ + if (!_mesa_is_legal_format_and_type(ctx, format, type)) { + _mesa_error(ctx, GL_INVALID_ENUM, + "gl%sPixels(format or type)", readDraw); + return GL_TRUE; + } + + /* additional checks */ + switch (format) { + case GL_RED: + case GL_GREEN: + case GL_BLUE: + case GL_ALPHA: + case GL_LUMINANCE: + case GL_LUMINANCE_ALPHA: + case GL_RGB: + case GL_BGR: + case GL_RGBA: + case GL_BGRA: + case GL_ABGR_EXT: + if (drawing && !ctx->Visual.rgbMode) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glDrawPixels(drawing RGB pixels into color index buffer)"); + return GL_TRUE; + } + if (!drawing && !_mesa_dest_buffer_exists(ctx, GL_COLOR)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glReadPixels(no color buffer)"); + return GL_TRUE; + } + break; + case GL_COLOR_INDEX: + if (!drawing && ctx->Visual.rgbMode) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glReadPixels(reading color index format from RGB buffer)"); + return GL_TRUE; + } + if (!drawing && !_mesa_dest_buffer_exists(ctx, GL_COLOR)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glReadPixels(no color buffer)"); + return GL_TRUE; + } + break; + case GL_STENCIL_INDEX: + if ((drawing && !_mesa_dest_buffer_exists(ctx, format)) || + (!drawing && !_mesa_source_buffer_exists(ctx, format))) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "gl%sPixels(no stencil buffer)", readDraw); + return GL_TRUE; + } + break; + case GL_DEPTH_COMPONENT: + if ((drawing && !_mesa_dest_buffer_exists(ctx, format)) || + (!drawing && !_mesa_source_buffer_exists(ctx, format))) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "gl%sPixels(no depth buffer)", readDraw); + return GL_TRUE; + } + break; + case GL_DEPTH_STENCIL_EXT: + if (!ctx->Extensions.EXT_packed_depth_stencil || + type != GL_UNSIGNED_INT_24_8_EXT) { + _mesa_error(ctx, GL_INVALID_ENUM, "gl%sPixels(type)", readDraw); + return GL_TRUE; + } + if ((drawing && !_mesa_dest_buffer_exists(ctx, format)) || + (!drawing && !_mesa_source_buffer_exists(ctx, format))) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "gl%sPixels(no depth or stencil buffer)", readDraw); + return GL_TRUE; + } + break; + default: + /* this should have been caught in _mesa_is_legal_format_type() */ + _mesa_problem(ctx, "unexpected format in _mesa_%sPixels", readDraw); + return GL_TRUE; + } + + /* no errors */ + return GL_FALSE; +} + + + +void GLAPIENTRY +_mesa_ReadPixels( GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, GLvoid *pixels ) +{ + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + + FLUSH_CURRENT(ctx, 0); + + if (width < 0 || height < 0) { + _mesa_error( ctx, GL_INVALID_VALUE, + "glReadPixels(width=%d height=%d)", width, height ); + return; + } + + if (ctx->NewState) + _mesa_update_state(ctx); + + if (_mesa_error_check_format_type(ctx, format, type, GL_FALSE)) { + /* found an error */ + return; + } + + if (ctx->ReadBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { + _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT, + "glReadPixels(incomplete framebuffer)" ); + return; + } + + if (!_mesa_source_buffer_exists(ctx, format)) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(no readbuffer)"); + return; + } + + if (ctx->Pack.BufferObj->Name) { + if (!_mesa_validate_pbo_access(2, &ctx->Pack, width, height, 1, + format, type, pixels)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glReadPixels(invalid PBO access)"); + return; + } + + if (ctx->Pack.BufferObj->Pointer) { + /* buffer is mapped - that's an error */ + _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(PBO is mapped)"); + return; + } + } + + ctx->Driver.ReadPixels(ctx, x, y, width, height, + format, type, &ctx->Pack, pixels); +} diff --git a/src/mesa/main/readpix.h b/src/mesa/main/readpix.h new file mode 100644 index 0000000000..1bf02fb8e4 --- /dev/null +++ b/src/mesa/main/readpix.h @@ -0,0 +1,42 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef READPIXELS_H +#define READPIXELS_H + + +#include "main/mtypes.h" + + +extern GLboolean +_mesa_error_check_format_type(GLcontext *ctx, GLenum format, GLenum type, + GLboolean drawing); + +extern void GLAPIENTRY +_mesa_ReadPixels( GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, GLvoid *pixels ); + + +#endif diff --git a/src/mesa/main/sources b/src/mesa/main/sources index ac809b22d2..8ef5bd62fa 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -45,6 +45,7 @@ pixel.c \ pixelstore.c \ points.c \ polygon.c \ +readpix.c \ rastpos.c \ rbadaptors.c \ renderbuffer.c \ @@ -121,6 +122,7 @@ points.h \ polygon.h \ rastpos.h \ rbadaptors.h \ +readpix.h \ renderbuffer.h \ simple_list.h \ state.h \ diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 13a079a2dc..64f911e699 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.1 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -79,6 +79,7 @@ #include "queryobj.h" #endif #include "rastpos.h" +#include "readpix.h" #include "state.h" #include "stencil.h" #include "teximage.h" diff --git a/src/mesa/sources b/src/mesa/sources index 6a1d55f1fe..c94f260121 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -49,6 +49,7 @@ MAIN_SOURCES = \ main/queryobj.c \ main/rastpos.c \ main/rbadaptors.c \ + main/readpix.c \ main/renderbuffer.c \ main/shaders.c \ main/state.c \ -- cgit v1.2.3 From 5f91007f996d0b7e3233f221a6b0056203e356d2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 14:25:23 -0600 Subject: mesa: refactor: new _mesa_init_pixelstore() function --- src/mesa/main/context.c | 2 ++ src/mesa/main/pixel.c | 28 ---------------------------- src/mesa/main/pixelstore.c | 37 +++++++++++++++++++++++++++++++++++++ src/mesa/main/pixelstore.h | 4 ++++ 4 files changed, 43 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 0053180000..522420592c 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -105,6 +105,7 @@ #include "macros.h" #include "matrix.h" #include "pixel.h" +#include "pixelstore.h" #include "points.h" #include "polygon.h" #include "queryobj.h" @@ -969,6 +970,7 @@ init_attrib_groups(GLcontext *ctx) _mesa_init_matrix( ctx ); _mesa_init_multisample( ctx ); _mesa_init_pixel( ctx ); + _mesa_init_pixelstore( ctx ); _mesa_init_point( ctx ); _mesa_init_polygon( ctx ); _mesa_init_program( ctx ); diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c index 1e2d31e172..f14b225c1d 100644 --- a/src/mesa/main/pixel.c +++ b/src/mesa/main/pixel.c @@ -1321,34 +1321,6 @@ _mesa_init_pixel( GLcontext *ctx ) ASSIGN_4V(ctx->Pixel.TextureColorTableScale, 1.0, 1.0, 1.0, 1.0); ASSIGN_4V(ctx->Pixel.TextureColorTableBias, 0.0, 0.0, 0.0, 0.0); - /* Pixel transfer */ - ctx->Pack.Alignment = 4; - ctx->Pack.RowLength = 0; - ctx->Pack.ImageHeight = 0; - ctx->Pack.SkipPixels = 0; - ctx->Pack.SkipRows = 0; - ctx->Pack.SkipImages = 0; - ctx->Pack.SwapBytes = GL_FALSE; - ctx->Pack.LsbFirst = GL_FALSE; - ctx->Pack.ClientStorage = GL_FALSE; - ctx->Pack.Invert = GL_FALSE; -#if FEATURE_EXT_pixel_buffer_object - ctx->Pack.BufferObj = ctx->Array.NullBufferObj; -#endif - ctx->Unpack.Alignment = 4; - ctx->Unpack.RowLength = 0; - ctx->Unpack.ImageHeight = 0; - ctx->Unpack.SkipPixels = 0; - ctx->Unpack.SkipRows = 0; - ctx->Unpack.SkipImages = 0; - ctx->Unpack.SwapBytes = GL_FALSE; - ctx->Unpack.LsbFirst = GL_FALSE; - ctx->Unpack.ClientStorage = GL_FALSE; - ctx->Unpack.Invert = GL_FALSE; -#if FEATURE_EXT_pixel_buffer_object - ctx->Unpack.BufferObj = ctx->Array.NullBufferObj; -#endif - /* * _mesa_unpack_image() returns image data in this format. When we * execute image commands (glDrawPixels(), glTexImage(), etc) from diff --git a/src/mesa/main/pixelstore.c b/src/mesa/main/pixelstore.c index f5f054f938..3bf89bd3e1 100644 --- a/src/mesa/main/pixelstore.c +++ b/src/mesa/main/pixelstore.c @@ -224,3 +224,40 @@ _mesa_PixelStoref( GLenum pname, GLfloat param ) { _mesa_PixelStorei( pname, (GLint) param ); } + + + +/** + * Initialize the context's pixel store state. + */ +void +_mesa_init_pixelstore( GLcontext *ctx ) +{ + /* Pixel transfer */ + ctx->Pack.Alignment = 4; + ctx->Pack.RowLength = 0; + ctx->Pack.ImageHeight = 0; + ctx->Pack.SkipPixels = 0; + ctx->Pack.SkipRows = 0; + ctx->Pack.SkipImages = 0; + ctx->Pack.SwapBytes = GL_FALSE; + ctx->Pack.LsbFirst = GL_FALSE; + ctx->Pack.ClientStorage = GL_FALSE; + ctx->Pack.Invert = GL_FALSE; +#if FEATURE_EXT_pixel_buffer_object + ctx->Pack.BufferObj = ctx->Array.NullBufferObj; +#endif + ctx->Unpack.Alignment = 4; + ctx->Unpack.RowLength = 0; + ctx->Unpack.ImageHeight = 0; + ctx->Unpack.SkipPixels = 0; + ctx->Unpack.SkipRows = 0; + ctx->Unpack.SkipImages = 0; + ctx->Unpack.SwapBytes = GL_FALSE; + ctx->Unpack.LsbFirst = GL_FALSE; + ctx->Unpack.ClientStorage = GL_FALSE; + ctx->Unpack.Invert = GL_FALSE; +#if FEATURE_EXT_pixel_buffer_object + ctx->Unpack.BufferObj = ctx->Array.NullBufferObj; +#endif +} diff --git a/src/mesa/main/pixelstore.h b/src/mesa/main/pixelstore.h index c42f304030..ee963f9ba3 100644 --- a/src/mesa/main/pixelstore.h +++ b/src/mesa/main/pixelstore.h @@ -43,4 +43,8 @@ extern void GLAPIENTRY _mesa_PixelStoref( GLenum pname, GLfloat param ); +extern void +_mesa_init_pixelstore( GLcontext *ctx ); + + #endif -- cgit v1.2.3 From 74c82ebbb399a274dcfb5e82d3471dee59bd5183 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 14:32:27 -0600 Subject: mesa: refactor: move pixel map/scale/bias code into image.c pixel.c is just the API-related code now. --- src/mesa/main/image.c | 422 ++++++++++++++++++++++++++++++++++++++- src/mesa/main/image.h | 47 ++++- src/mesa/main/pixel.c | 433 ++--------------------------------------- src/mesa/main/pixel.h | 60 +----- src/mesa/swrast/s_texcombine.c | 2 +- 5 files changed, 487 insertions(+), 477 deletions(-) (limited to 'src') diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c index 285c8346a5..63bcc48e68 100644 --- a/src/mesa/main/image.c +++ b/src/mesa/main/image.c @@ -2,7 +2,7 @@ * Mesa 3-D graphics library * Version: 7.1 * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -1011,6 +1011,426 @@ _mesa_pack_bitmap( GLint width, GLint height, const GLubyte *source, } +/**********************************************************************/ +/***** Pixel processing functions ******/ +/**********************************************************************/ + +/* + * Apply scale and bias factors to an array of RGBA pixels. + */ +void +_mesa_scale_and_bias_rgba(GLuint n, GLfloat rgba[][4], + GLfloat rScale, GLfloat gScale, + GLfloat bScale, GLfloat aScale, + GLfloat rBias, GLfloat gBias, + GLfloat bBias, GLfloat aBias) +{ + if (rScale != 1.0 || rBias != 0.0) { + GLuint i; + for (i = 0; i < n; i++) { + rgba[i][RCOMP] = rgba[i][RCOMP] * rScale + rBias; + } + } + if (gScale != 1.0 || gBias != 0.0) { + GLuint i; + for (i = 0; i < n; i++) { + rgba[i][GCOMP] = rgba[i][GCOMP] * gScale + gBias; + } + } + if (bScale != 1.0 || bBias != 0.0) { + GLuint i; + for (i = 0; i < n; i++) { + rgba[i][BCOMP] = rgba[i][BCOMP] * bScale + bBias; + } + } + if (aScale != 1.0 || aBias != 0.0) { + GLuint i; + for (i = 0; i < n; i++) { + rgba[i][ACOMP] = rgba[i][ACOMP] * aScale + aBias; + } + } +} + + +/* + * Apply pixel mapping to an array of floating point RGBA pixels. + */ +void +_mesa_map_rgba( const GLcontext *ctx, GLuint n, GLfloat rgba[][4] ) +{ + const GLfloat rscale = (GLfloat) (ctx->PixelMaps.RtoR.Size - 1); + const GLfloat gscale = (GLfloat) (ctx->PixelMaps.GtoG.Size - 1); + const GLfloat bscale = (GLfloat) (ctx->PixelMaps.BtoB.Size - 1); + const GLfloat ascale = (GLfloat) (ctx->PixelMaps.AtoA.Size - 1); + const GLfloat *rMap = ctx->PixelMaps.RtoR.Map; + const GLfloat *gMap = ctx->PixelMaps.GtoG.Map; + const GLfloat *bMap = ctx->PixelMaps.BtoB.Map; + const GLfloat *aMap = ctx->PixelMaps.AtoA.Map; + GLuint i; + for (i=0;iPixel.PostColorMatrixScale[0]; + const GLfloat rb = ctx->Pixel.PostColorMatrixBias[0]; + const GLfloat gs = ctx->Pixel.PostColorMatrixScale[1]; + const GLfloat gb = ctx->Pixel.PostColorMatrixBias[1]; + const GLfloat bs = ctx->Pixel.PostColorMatrixScale[2]; + const GLfloat bb = ctx->Pixel.PostColorMatrixBias[2]; + const GLfloat as = ctx->Pixel.PostColorMatrixScale[3]; + const GLfloat ab = ctx->Pixel.PostColorMatrixBias[3]; + const GLfloat *m = ctx->ColorMatrixStack.Top->m; + GLuint i; + for (i = 0; i < n; i++) { + const GLfloat r = rgba[i][RCOMP]; + const GLfloat g = rgba[i][GCOMP]; + const GLfloat b = rgba[i][BCOMP]; + const GLfloat a = rgba[i][ACOMP]; + rgba[i][RCOMP] = (m[0] * r + m[4] * g + m[ 8] * b + m[12] * a) * rs + rb; + rgba[i][GCOMP] = (m[1] * r + m[5] * g + m[ 9] * b + m[13] * a) * gs + gb; + rgba[i][BCOMP] = (m[2] * r + m[6] * g + m[10] * b + m[14] * a) * bs + bb; + rgba[i][ACOMP] = (m[3] * r + m[7] * g + m[11] * b + m[15] * a) * as + ab; + } +} + + +/** + * Apply a color table lookup to an array of floating point RGBA colors. + */ +void +_mesa_lookup_rgba_float(const struct gl_color_table *table, + GLuint n, GLfloat rgba[][4]) +{ + const GLint max = table->Size - 1; + const GLfloat scale = (GLfloat) max; + const GLfloat *lut = table->TableF; + GLuint i; + + if (!table->TableF || table->Size == 0) + return; + + switch (table->_BaseFormat) { + case GL_INTENSITY: + /* replace RGBA with I */ + for (i = 0; i < n; i++) { + GLint j = IROUND(rgba[i][RCOMP] * scale); + GLfloat c = lut[CLAMP(j, 0, max)]; + rgba[i][RCOMP] = + rgba[i][GCOMP] = + rgba[i][BCOMP] = + rgba[i][ACOMP] = c; + } + break; + case GL_LUMINANCE: + /* replace RGB with L */ + for (i = 0; i < n; i++) { + GLint j = IROUND(rgba[i][RCOMP] * scale); + GLfloat c = lut[CLAMP(j, 0, max)]; + rgba[i][RCOMP] = + rgba[i][GCOMP] = + rgba[i][BCOMP] = c; + } + break; + case GL_ALPHA: + /* replace A with A */ + for (i = 0; i < n; i++) { + GLint j = IROUND(rgba[i][ACOMP] * scale); + rgba[i][ACOMP] = lut[CLAMP(j, 0, max)]; + } + break; + case GL_LUMINANCE_ALPHA: + /* replace RGBA with LLLA */ + for (i = 0; i < n; i++) { + GLint jL = IROUND(rgba[i][RCOMP] * scale); + GLint jA = IROUND(rgba[i][ACOMP] * scale); + GLfloat luminance, alpha; + jL = CLAMP(jL, 0, max); + jA = CLAMP(jA, 0, max); + luminance = lut[jL * 2 + 0]; + alpha = lut[jA * 2 + 1]; + rgba[i][RCOMP] = + rgba[i][GCOMP] = + rgba[i][BCOMP] = luminance; + rgba[i][ACOMP] = alpha;; + } + break; + case GL_RGB: + /* replace RGB with RGB */ + for (i = 0; i < n; i++) { + GLint jR = IROUND(rgba[i][RCOMP] * scale); + GLint jG = IROUND(rgba[i][GCOMP] * scale); + GLint jB = IROUND(rgba[i][BCOMP] * scale); + jR = CLAMP(jR, 0, max); + jG = CLAMP(jG, 0, max); + jB = CLAMP(jB, 0, max); + rgba[i][RCOMP] = lut[jR * 3 + 0]; + rgba[i][GCOMP] = lut[jG * 3 + 1]; + rgba[i][BCOMP] = lut[jB * 3 + 2]; + } + break; + case GL_RGBA: + /* replace RGBA with RGBA */ + for (i = 0; i < n; i++) { + GLint jR = IROUND(rgba[i][RCOMP] * scale); + GLint jG = IROUND(rgba[i][GCOMP] * scale); + GLint jB = IROUND(rgba[i][BCOMP] * scale); + GLint jA = IROUND(rgba[i][ACOMP] * scale); + jR = CLAMP(jR, 0, max); + jG = CLAMP(jG, 0, max); + jB = CLAMP(jB, 0, max); + jA = CLAMP(jA, 0, max); + rgba[i][RCOMP] = lut[jR * 4 + 0]; + rgba[i][GCOMP] = lut[jG * 4 + 1]; + rgba[i][BCOMP] = lut[jB * 4 + 2]; + rgba[i][ACOMP] = lut[jA * 4 + 3]; + } + break; + default: + _mesa_problem(NULL, "Bad format in _mesa_lookup_rgba_float"); + return; + } +} + + + +/** + * Apply a color table lookup to an array of ubyte/RGBA colors. + */ +void +_mesa_lookup_rgba_ubyte(const struct gl_color_table *table, + GLuint n, GLubyte rgba[][4]) +{ + const GLubyte *lut = table->TableUB; + const GLfloat scale = (GLfloat) (table->Size - 1) / (GLfloat)255.0; + GLuint i; + + if (!table->TableUB || table->Size == 0) + return; + + switch (table->_BaseFormat) { + case GL_INTENSITY: + /* replace RGBA with I */ + if (table->Size == 256) { + for (i = 0; i < n; i++) { + const GLubyte c = lut[rgba[i][RCOMP]]; + rgba[i][RCOMP] = + rgba[i][GCOMP] = + rgba[i][BCOMP] = + rgba[i][ACOMP] = c; + } + } + else { + for (i = 0; i < n; i++) { + GLint j = IROUND((GLfloat) rgba[i][RCOMP] * scale); + rgba[i][RCOMP] = + rgba[i][GCOMP] = + rgba[i][BCOMP] = + rgba[i][ACOMP] = lut[j]; + } + } + break; + case GL_LUMINANCE: + /* replace RGB with L */ + if (table->Size == 256) { + for (i = 0; i < n; i++) { + const GLubyte c = lut[rgba[i][RCOMP]]; + rgba[i][RCOMP] = + rgba[i][GCOMP] = + rgba[i][BCOMP] = c; + } + } + else { + for (i = 0; i < n; i++) { + GLint j = IROUND((GLfloat) rgba[i][RCOMP] * scale); + rgba[i][RCOMP] = + rgba[i][GCOMP] = + rgba[i][BCOMP] = lut[j]; + } + } + break; + case GL_ALPHA: + /* replace A with A */ + if (table->Size == 256) { + for (i = 0; i < n; i++) { + rgba[i][ACOMP] = lut[rgba[i][ACOMP]]; + } + } + else { + for (i = 0; i < n; i++) { + GLint j = IROUND((GLfloat) rgba[i][ACOMP] * scale); + rgba[i][ACOMP] = lut[j]; + } + } + break; + case GL_LUMINANCE_ALPHA: + /* replace RGBA with LLLA */ + if (table->Size == 256) { + for (i = 0; i < n; i++) { + GLubyte l = lut[rgba[i][RCOMP] * 2 + 0]; + GLubyte a = lut[rgba[i][ACOMP] * 2 + 1];; + rgba[i][RCOMP] = + rgba[i][GCOMP] = + rgba[i][BCOMP] = l; + rgba[i][ACOMP] = a; + } + } + else { + for (i = 0; i < n; i++) { + GLint jL = IROUND((GLfloat) rgba[i][RCOMP] * scale); + GLint jA = IROUND((GLfloat) rgba[i][ACOMP] * scale); + GLubyte luminance = lut[jL * 2 + 0]; + GLubyte alpha = lut[jA * 2 + 1]; + rgba[i][RCOMP] = + rgba[i][GCOMP] = + rgba[i][BCOMP] = luminance; + rgba[i][ACOMP] = alpha; + } + } + break; + case GL_RGB: + if (table->Size == 256) { + for (i = 0; i < n; i++) { + rgba[i][RCOMP] = lut[rgba[i][RCOMP] * 3 + 0]; + rgba[i][GCOMP] = lut[rgba[i][GCOMP] * 3 + 1]; + rgba[i][BCOMP] = lut[rgba[i][BCOMP] * 3 + 2]; + } + } + else { + for (i = 0; i < n; i++) { + GLint jR = IROUND((GLfloat) rgba[i][RCOMP] * scale); + GLint jG = IROUND((GLfloat) rgba[i][GCOMP] * scale); + GLint jB = IROUND((GLfloat) rgba[i][BCOMP] * scale); + rgba[i][RCOMP] = lut[jR * 3 + 0]; + rgba[i][GCOMP] = lut[jG * 3 + 1]; + rgba[i][BCOMP] = lut[jB * 3 + 2]; + } + } + break; + case GL_RGBA: + if (table->Size == 256) { + for (i = 0; i < n; i++) { + rgba[i][RCOMP] = lut[rgba[i][RCOMP] * 4 + 0]; + rgba[i][GCOMP] = lut[rgba[i][GCOMP] * 4 + 1]; + rgba[i][BCOMP] = lut[rgba[i][BCOMP] * 4 + 2]; + rgba[i][ACOMP] = lut[rgba[i][ACOMP] * 4 + 3]; + } + } + else { + for (i = 0; i < n; i++) { + GLint jR = IROUND((GLfloat) rgba[i][RCOMP] * scale); + GLint jG = IROUND((GLfloat) rgba[i][GCOMP] * scale); + GLint jB = IROUND((GLfloat) rgba[i][BCOMP] * scale); + GLint jA = IROUND((GLfloat) rgba[i][ACOMP] * scale); + CLAMPED_FLOAT_TO_CHAN(rgba[i][RCOMP], lut[jR * 4 + 0]); + CLAMPED_FLOAT_TO_CHAN(rgba[i][GCOMP], lut[jG * 4 + 1]); + CLAMPED_FLOAT_TO_CHAN(rgba[i][BCOMP], lut[jB * 4 + 2]); + CLAMPED_FLOAT_TO_CHAN(rgba[i][ACOMP], lut[jA * 4 + 3]); + } + } + break; + default: + _mesa_problem(NULL, "Bad format in _mesa_lookup_rgba_chan"); + return; + } +} + + + +/* + * Map color indexes to float rgba values. + */ +void +_mesa_map_ci_to_rgba( const GLcontext *ctx, GLuint n, + const GLuint index[], GLfloat rgba[][4] ) +{ + GLuint rmask = ctx->PixelMaps.ItoR.Size - 1; + GLuint gmask = ctx->PixelMaps.ItoG.Size - 1; + GLuint bmask = ctx->PixelMaps.ItoB.Size - 1; + GLuint amask = ctx->PixelMaps.ItoA.Size - 1; + const GLfloat *rMap = ctx->PixelMaps.ItoR.Map; + const GLfloat *gMap = ctx->PixelMaps.ItoG.Map; + const GLfloat *bMap = ctx->PixelMaps.ItoB.Map; + const GLfloat *aMap = ctx->PixelMaps.ItoA.Map; + GLuint i; + for (i=0;iPixelMaps.ItoR.Size - 1; + GLuint gmask = ctx->PixelMaps.ItoG.Size - 1; + GLuint bmask = ctx->PixelMaps.ItoB.Size - 1; + GLuint amask = ctx->PixelMaps.ItoA.Size - 1; + const GLubyte *rMap = ctx->PixelMaps.ItoR.Map8; + const GLubyte *gMap = ctx->PixelMaps.ItoG.Map8; + const GLubyte *bMap = ctx->PixelMaps.ItoB.Map8; + const GLubyte *aMap = ctx->PixelMaps.ItoA.Map8; + GLuint i; + for (i=0;iPixel.DepthScale; + const GLfloat bias = ctx->Pixel.DepthBias; + GLuint i; + for (i = 0; i < n; i++) { + GLfloat d = depthValues[i] * scale + bias; + depthValues[i] = CLAMP(d, 0.0F, 1.0F); + } +} + + +void +_mesa_scale_and_bias_depth_uint(const GLcontext *ctx, GLuint n, + GLuint depthValues[]) +{ + const GLdouble max = (double) 0xffffffff; + const GLdouble scale = ctx->Pixel.DepthScale; + const GLdouble bias = ctx->Pixel.DepthBias * max; + GLuint i; + for (i = 0; i < n; i++) { + GLdouble d = (GLdouble) depthValues[i] * scale + bias; + d = CLAMP(d, 0.0, max); + depthValues[i] = (GLuint) d; + } +} + + /** * Apply various pixel transfer operations to an array of RGBA pixels * as indicated by the transferOps bitmask diff --git a/src/mesa/main/image.h b/src/mesa/main/image.h index c379d5fa7d..38e1374c20 100644 --- a/src/mesa/main/image.h +++ b/src/mesa/main/image.h @@ -2,7 +2,7 @@ * Mesa 3-D graphics library * Version: 7.1 * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -111,6 +111,51 @@ _mesa_pack_bitmap( GLint width, GLint height, const GLubyte *source, GLubyte *dest, const struct gl_pixelstore_attrib *packing ); +/** \name Pixel processing functions */ +/*@{*/ + +extern void +_mesa_scale_and_bias_rgba(GLuint n, GLfloat rgba[][4], + GLfloat rScale, GLfloat gScale, + GLfloat bScale, GLfloat aScale, + GLfloat rBias, GLfloat gBias, + GLfloat bBias, GLfloat aBias); + +extern void +_mesa_map_rgba(const GLcontext *ctx, GLuint n, GLfloat rgba[][4]); + + +extern void +_mesa_transform_rgba(const GLcontext *ctx, GLuint n, GLfloat rgba[][4]); + + +extern void +_mesa_lookup_rgba_float(const struct gl_color_table *table, + GLuint n, GLfloat rgba[][4]); + +extern void +_mesa_lookup_rgba_ubyte(const struct gl_color_table *table, + GLuint n, GLubyte rgba[][4]); + + +extern void +_mesa_map_ci_to_rgba(const GLcontext *ctx, + GLuint n, const GLuint index[], GLfloat rgba[][4]); + + +extern void +_mesa_map_ci8_to_rgba8(const GLcontext *ctx, GLuint n, const GLubyte index[], + GLubyte rgba[][4]); + + +extern void +_mesa_scale_and_bias_depth(const GLcontext *ctx, GLuint n, + GLfloat depthValues[]); + +extern void +_mesa_scale_and_bias_depth_uint(const GLcontext *ctx, GLuint n, + GLuint depthValues[]); + extern void _mesa_apply_rgba_transfer_ops(GLcontext *ctx, GLbitfield transferOps, GLuint n, GLfloat rgba[][4]); diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c index f14b225c1d..c9c4289f69 100644 --- a/src/mesa/main/pixel.c +++ b/src/mesa/main/pixel.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.3 + * Version: 7.1 * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -22,6 +22,12 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ + +/** + * \file pixel.c + * Pixel transfer functions (glPixelZoom, glPixelMap, glPixelTransfer) + */ + #include "glheader.h" #include "bufferobj.h" #include "colormac.h" @@ -741,426 +747,6 @@ _mesa_PixelTransferi( GLenum pname, GLint param ) -/**********************************************************************/ -/***** Pixel processing functions ******/ -/**********************************************************************/ - -/* - * Apply scale and bias factors to an array of RGBA pixels. - */ -void -_mesa_scale_and_bias_rgba(GLuint n, GLfloat rgba[][4], - GLfloat rScale, GLfloat gScale, - GLfloat bScale, GLfloat aScale, - GLfloat rBias, GLfloat gBias, - GLfloat bBias, GLfloat aBias) -{ - if (rScale != 1.0 || rBias != 0.0) { - GLuint i; - for (i = 0; i < n; i++) { - rgba[i][RCOMP] = rgba[i][RCOMP] * rScale + rBias; - } - } - if (gScale != 1.0 || gBias != 0.0) { - GLuint i; - for (i = 0; i < n; i++) { - rgba[i][GCOMP] = rgba[i][GCOMP] * gScale + gBias; - } - } - if (bScale != 1.0 || bBias != 0.0) { - GLuint i; - for (i = 0; i < n; i++) { - rgba[i][BCOMP] = rgba[i][BCOMP] * bScale + bBias; - } - } - if (aScale != 1.0 || aBias != 0.0) { - GLuint i; - for (i = 0; i < n; i++) { - rgba[i][ACOMP] = rgba[i][ACOMP] * aScale + aBias; - } - } -} - - -/* - * Apply pixel mapping to an array of floating point RGBA pixels. - */ -void -_mesa_map_rgba( const GLcontext *ctx, GLuint n, GLfloat rgba[][4] ) -{ - const GLfloat rscale = (GLfloat) (ctx->PixelMaps.RtoR.Size - 1); - const GLfloat gscale = (GLfloat) (ctx->PixelMaps.GtoG.Size - 1); - const GLfloat bscale = (GLfloat) (ctx->PixelMaps.BtoB.Size - 1); - const GLfloat ascale = (GLfloat) (ctx->PixelMaps.AtoA.Size - 1); - const GLfloat *rMap = ctx->PixelMaps.RtoR.Map; - const GLfloat *gMap = ctx->PixelMaps.GtoG.Map; - const GLfloat *bMap = ctx->PixelMaps.BtoB.Map; - const GLfloat *aMap = ctx->PixelMaps.AtoA.Map; - GLuint i; - for (i=0;iPixel.PostColorMatrixScale[0]; - const GLfloat rb = ctx->Pixel.PostColorMatrixBias[0]; - const GLfloat gs = ctx->Pixel.PostColorMatrixScale[1]; - const GLfloat gb = ctx->Pixel.PostColorMatrixBias[1]; - const GLfloat bs = ctx->Pixel.PostColorMatrixScale[2]; - const GLfloat bb = ctx->Pixel.PostColorMatrixBias[2]; - const GLfloat as = ctx->Pixel.PostColorMatrixScale[3]; - const GLfloat ab = ctx->Pixel.PostColorMatrixBias[3]; - const GLfloat *m = ctx->ColorMatrixStack.Top->m; - GLuint i; - for (i = 0; i < n; i++) { - const GLfloat r = rgba[i][RCOMP]; - const GLfloat g = rgba[i][GCOMP]; - const GLfloat b = rgba[i][BCOMP]; - const GLfloat a = rgba[i][ACOMP]; - rgba[i][RCOMP] = (m[0] * r + m[4] * g + m[ 8] * b + m[12] * a) * rs + rb; - rgba[i][GCOMP] = (m[1] * r + m[5] * g + m[ 9] * b + m[13] * a) * gs + gb; - rgba[i][BCOMP] = (m[2] * r + m[6] * g + m[10] * b + m[14] * a) * bs + bb; - rgba[i][ACOMP] = (m[3] * r + m[7] * g + m[11] * b + m[15] * a) * as + ab; - } -} - - -/** - * Apply a color table lookup to an array of floating point RGBA colors. - */ -void -_mesa_lookup_rgba_float(const struct gl_color_table *table, - GLuint n, GLfloat rgba[][4]) -{ - const GLint max = table->Size - 1; - const GLfloat scale = (GLfloat) max; - const GLfloat *lut = table->TableF; - GLuint i; - - if (!table->TableF || table->Size == 0) - return; - - switch (table->_BaseFormat) { - case GL_INTENSITY: - /* replace RGBA with I */ - for (i = 0; i < n; i++) { - GLint j = IROUND(rgba[i][RCOMP] * scale); - GLfloat c = lut[CLAMP(j, 0, max)]; - rgba[i][RCOMP] = - rgba[i][GCOMP] = - rgba[i][BCOMP] = - rgba[i][ACOMP] = c; - } - break; - case GL_LUMINANCE: - /* replace RGB with L */ - for (i = 0; i < n; i++) { - GLint j = IROUND(rgba[i][RCOMP] * scale); - GLfloat c = lut[CLAMP(j, 0, max)]; - rgba[i][RCOMP] = - rgba[i][GCOMP] = - rgba[i][BCOMP] = c; - } - break; - case GL_ALPHA: - /* replace A with A */ - for (i = 0; i < n; i++) { - GLint j = IROUND(rgba[i][ACOMP] * scale); - rgba[i][ACOMP] = lut[CLAMP(j, 0, max)]; - } - break; - case GL_LUMINANCE_ALPHA: - /* replace RGBA with LLLA */ - for (i = 0; i < n; i++) { - GLint jL = IROUND(rgba[i][RCOMP] * scale); - GLint jA = IROUND(rgba[i][ACOMP] * scale); - GLfloat luminance, alpha; - jL = CLAMP(jL, 0, max); - jA = CLAMP(jA, 0, max); - luminance = lut[jL * 2 + 0]; - alpha = lut[jA * 2 + 1]; - rgba[i][RCOMP] = - rgba[i][GCOMP] = - rgba[i][BCOMP] = luminance; - rgba[i][ACOMP] = alpha;; - } - break; - case GL_RGB: - /* replace RGB with RGB */ - for (i = 0; i < n; i++) { - GLint jR = IROUND(rgba[i][RCOMP] * scale); - GLint jG = IROUND(rgba[i][GCOMP] * scale); - GLint jB = IROUND(rgba[i][BCOMP] * scale); - jR = CLAMP(jR, 0, max); - jG = CLAMP(jG, 0, max); - jB = CLAMP(jB, 0, max); - rgba[i][RCOMP] = lut[jR * 3 + 0]; - rgba[i][GCOMP] = lut[jG * 3 + 1]; - rgba[i][BCOMP] = lut[jB * 3 + 2]; - } - break; - case GL_RGBA: - /* replace RGBA with RGBA */ - for (i = 0; i < n; i++) { - GLint jR = IROUND(rgba[i][RCOMP] * scale); - GLint jG = IROUND(rgba[i][GCOMP] * scale); - GLint jB = IROUND(rgba[i][BCOMP] * scale); - GLint jA = IROUND(rgba[i][ACOMP] * scale); - jR = CLAMP(jR, 0, max); - jG = CLAMP(jG, 0, max); - jB = CLAMP(jB, 0, max); - jA = CLAMP(jA, 0, max); - rgba[i][RCOMP] = lut[jR * 4 + 0]; - rgba[i][GCOMP] = lut[jG * 4 + 1]; - rgba[i][BCOMP] = lut[jB * 4 + 2]; - rgba[i][ACOMP] = lut[jA * 4 + 3]; - } - break; - default: - _mesa_problem(NULL, "Bad format in _mesa_lookup_rgba_float"); - return; - } -} - - - -/** - * Apply a color table lookup to an array of ubyte/RGBA colors. - */ -void -_mesa_lookup_rgba_ubyte(const struct gl_color_table *table, - GLuint n, GLubyte rgba[][4]) -{ - const GLubyte *lut = table->TableUB; - const GLfloat scale = (GLfloat) (table->Size - 1) / (GLfloat)255.0; - GLuint i; - - if (!table->TableUB || table->Size == 0) - return; - - switch (table->_BaseFormat) { - case GL_INTENSITY: - /* replace RGBA with I */ - if (table->Size == 256) { - for (i = 0; i < n; i++) { - const GLubyte c = lut[rgba[i][RCOMP]]; - rgba[i][RCOMP] = - rgba[i][GCOMP] = - rgba[i][BCOMP] = - rgba[i][ACOMP] = c; - } - } - else { - for (i = 0; i < n; i++) { - GLint j = IROUND((GLfloat) rgba[i][RCOMP] * scale); - rgba[i][RCOMP] = - rgba[i][GCOMP] = - rgba[i][BCOMP] = - rgba[i][ACOMP] = lut[j]; - } - } - break; - case GL_LUMINANCE: - /* replace RGB with L */ - if (table->Size == 256) { - for (i = 0; i < n; i++) { - const GLubyte c = lut[rgba[i][RCOMP]]; - rgba[i][RCOMP] = - rgba[i][GCOMP] = - rgba[i][BCOMP] = c; - } - } - else { - for (i = 0; i < n; i++) { - GLint j = IROUND((GLfloat) rgba[i][RCOMP] * scale); - rgba[i][RCOMP] = - rgba[i][GCOMP] = - rgba[i][BCOMP] = lut[j]; - } - } - break; - case GL_ALPHA: - /* replace A with A */ - if (table->Size == 256) { - for (i = 0; i < n; i++) { - rgba[i][ACOMP] = lut[rgba[i][ACOMP]]; - } - } - else { - for (i = 0; i < n; i++) { - GLint j = IROUND((GLfloat) rgba[i][ACOMP] * scale); - rgba[i][ACOMP] = lut[j]; - } - } - break; - case GL_LUMINANCE_ALPHA: - /* replace RGBA with LLLA */ - if (table->Size == 256) { - for (i = 0; i < n; i++) { - GLubyte l = lut[rgba[i][RCOMP] * 2 + 0]; - GLubyte a = lut[rgba[i][ACOMP] * 2 + 1];; - rgba[i][RCOMP] = - rgba[i][GCOMP] = - rgba[i][BCOMP] = l; - rgba[i][ACOMP] = a; - } - } - else { - for (i = 0; i < n; i++) { - GLint jL = IROUND((GLfloat) rgba[i][RCOMP] * scale); - GLint jA = IROUND((GLfloat) rgba[i][ACOMP] * scale); - GLubyte luminance = lut[jL * 2 + 0]; - GLubyte alpha = lut[jA * 2 + 1]; - rgba[i][RCOMP] = - rgba[i][GCOMP] = - rgba[i][BCOMP] = luminance; - rgba[i][ACOMP] = alpha; - } - } - break; - case GL_RGB: - if (table->Size == 256) { - for (i = 0; i < n; i++) { - rgba[i][RCOMP] = lut[rgba[i][RCOMP] * 3 + 0]; - rgba[i][GCOMP] = lut[rgba[i][GCOMP] * 3 + 1]; - rgba[i][BCOMP] = lut[rgba[i][BCOMP] * 3 + 2]; - } - } - else { - for (i = 0; i < n; i++) { - GLint jR = IROUND((GLfloat) rgba[i][RCOMP] * scale); - GLint jG = IROUND((GLfloat) rgba[i][GCOMP] * scale); - GLint jB = IROUND((GLfloat) rgba[i][BCOMP] * scale); - rgba[i][RCOMP] = lut[jR * 3 + 0]; - rgba[i][GCOMP] = lut[jG * 3 + 1]; - rgba[i][BCOMP] = lut[jB * 3 + 2]; - } - } - break; - case GL_RGBA: - if (table->Size == 256) { - for (i = 0; i < n; i++) { - rgba[i][RCOMP] = lut[rgba[i][RCOMP] * 4 + 0]; - rgba[i][GCOMP] = lut[rgba[i][GCOMP] * 4 + 1]; - rgba[i][BCOMP] = lut[rgba[i][BCOMP] * 4 + 2]; - rgba[i][ACOMP] = lut[rgba[i][ACOMP] * 4 + 3]; - } - } - else { - for (i = 0; i < n; i++) { - GLint jR = IROUND((GLfloat) rgba[i][RCOMP] * scale); - GLint jG = IROUND((GLfloat) rgba[i][GCOMP] * scale); - GLint jB = IROUND((GLfloat) rgba[i][BCOMP] * scale); - GLint jA = IROUND((GLfloat) rgba[i][ACOMP] * scale); - CLAMPED_FLOAT_TO_CHAN(rgba[i][RCOMP], lut[jR * 4 + 0]); - CLAMPED_FLOAT_TO_CHAN(rgba[i][GCOMP], lut[jG * 4 + 1]); - CLAMPED_FLOAT_TO_CHAN(rgba[i][BCOMP], lut[jB * 4 + 2]); - CLAMPED_FLOAT_TO_CHAN(rgba[i][ACOMP], lut[jA * 4 + 3]); - } - } - break; - default: - _mesa_problem(NULL, "Bad format in _mesa_lookup_rgba_chan"); - return; - } -} - - - -/* - * Map color indexes to float rgba values. - */ -void -_mesa_map_ci_to_rgba( const GLcontext *ctx, GLuint n, - const GLuint index[], GLfloat rgba[][4] ) -{ - GLuint rmask = ctx->PixelMaps.ItoR.Size - 1; - GLuint gmask = ctx->PixelMaps.ItoG.Size - 1; - GLuint bmask = ctx->PixelMaps.ItoB.Size - 1; - GLuint amask = ctx->PixelMaps.ItoA.Size - 1; - const GLfloat *rMap = ctx->PixelMaps.ItoR.Map; - const GLfloat *gMap = ctx->PixelMaps.ItoG.Map; - const GLfloat *bMap = ctx->PixelMaps.ItoB.Map; - const GLfloat *aMap = ctx->PixelMaps.ItoA.Map; - GLuint i; - for (i=0;iPixelMaps.ItoR.Size - 1; - GLuint gmask = ctx->PixelMaps.ItoG.Size - 1; - GLuint bmask = ctx->PixelMaps.ItoB.Size - 1; - GLuint amask = ctx->PixelMaps.ItoA.Size - 1; - const GLubyte *rMap = ctx->PixelMaps.ItoR.Map8; - const GLubyte *gMap = ctx->PixelMaps.ItoG.Map8; - const GLubyte *bMap = ctx->PixelMaps.ItoB.Map8; - const GLubyte *aMap = ctx->PixelMaps.ItoA.Map8; - GLuint i; - for (i=0;iPixel.DepthScale; - const GLfloat bias = ctx->Pixel.DepthBias; - GLuint i; - for (i = 0; i < n; i++) { - GLfloat d = depthValues[i] * scale + bias; - depthValues[i] = CLAMP(d, 0.0F, 1.0F); - } -} - - -void -_mesa_scale_and_bias_depth_uint(const GLcontext *ctx, GLuint n, - GLuint depthValues[]) -{ - const GLdouble max = (double) 0xffffffff; - const GLdouble scale = ctx->Pixel.DepthScale; - const GLdouble bias = ctx->Pixel.DepthBias * max; - GLuint i; - for (i = 0; i < n; i++) { - GLdouble d = (GLdouble) depthValues[i] * scale + bias; - d = CLAMP(d, 0.0, max); - depthValues[i] = (GLuint) d; - } -} - - /**********************************************************************/ /***** State Management *****/ /**********************************************************************/ @@ -1232,6 +818,9 @@ update_image_transfer_state(GLcontext *ctx) } +/** + * Update meas pixel transfer derived state. + */ void _mesa_update_pixel( GLcontext *ctx, GLuint new_state ) { if (new_state & _NEW_COLOR_MATRIX) diff --git a/src/mesa/main/pixel.h b/src/mesa/main/pixel.h index 05bddf5ccf..cb6c5262a3 100644 --- a/src/mesa/main/pixel.h +++ b/src/mesa/main/pixel.h @@ -1,13 +1,8 @@ -/** - * \file pixel.h - * Pixel operations. - */ - /* * Mesa 3-D graphics library - * Version: 6.5.2 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -28,6 +23,12 @@ */ +/** + * \file pixel.h + * Pixel operations. + */ + + #ifndef PIXEL_H #define PIXEL_H @@ -68,51 +69,6 @@ _mesa_PixelZoom( GLfloat xfactor, GLfloat yfactor ); /*@}*/ -/** \name Pixel processing functions */ -/*@{*/ - -extern void -_mesa_scale_and_bias_rgba(GLuint n, GLfloat rgba[][4], - GLfloat rScale, GLfloat gScale, - GLfloat bScale, GLfloat aScale, - GLfloat rBias, GLfloat gBias, - GLfloat bBias, GLfloat aBias); - -extern void -_mesa_map_rgba(const GLcontext *ctx, GLuint n, GLfloat rgba[][4]); - - -extern void -_mesa_transform_rgba(const GLcontext *ctx, GLuint n, GLfloat rgba[][4]); - - -extern void -_mesa_lookup_rgba_float(const struct gl_color_table *table, - GLuint n, GLfloat rgba[][4]); - -extern void -_mesa_lookup_rgba_ubyte(const struct gl_color_table *table, - GLuint n, GLubyte rgba[][4]); - - -extern void -_mesa_map_ci_to_rgba(const GLcontext *ctx, - GLuint n, const GLuint index[], GLfloat rgba[][4]); - - -extern void -_mesa_map_ci8_to_rgba8(const GLcontext *ctx, GLuint n, const GLubyte index[], - GLubyte rgba[][4]); - - -extern void -_mesa_scale_and_bias_depth(const GLcontext *ctx, GLuint n, - GLfloat depthValues[]); - -extern void -_mesa_scale_and_bias_depth_uint(const GLcontext *ctx, GLuint n, - GLuint depthValues[]); - extern void _mesa_update_pixel( GLcontext *ctx, GLuint newstate ); diff --git a/src/mesa/swrast/s_texcombine.c b/src/mesa/swrast/s_texcombine.c index 4ac7222daa..4c5e22618e 100644 --- a/src/mesa/swrast/s_texcombine.c +++ b/src/mesa/swrast/s_texcombine.c @@ -28,7 +28,7 @@ #include "colormac.h" #include "imports.h" #include "macros.h" -#include "pixel.h" +#include "image.h" #include "s_context.h" #include "s_texcombine.h" -- cgit v1.2.3 From b36e6f0baf64491772b8e1a1cddf68a7dcf8ee22 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 14:49:04 -0600 Subject: mesa: refactor: move _mesa_init_exec_table() into new api_exec.c file --- src/mesa/main/api_exec.c | 826 +++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/api_exec.h | 34 ++ src/mesa/main/context.c | 1 + src/mesa/main/sources | 2 + src/mesa/main/state.c | 803 +-------------------------------------------- src/mesa/main/state.h | 12 +- src/mesa/sources | 1 + 7 files changed, 873 insertions(+), 806 deletions(-) create mode 100644 src/mesa/main/api_exec.c create mode 100644 src/mesa/main/api_exec.h (limited to 'src') diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c new file mode 100644 index 0000000000..e0f081bdbc --- /dev/null +++ b/src/mesa/main/api_exec.c @@ -0,0 +1,826 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * \file api_exec.c + * Initialize dispatch table with the immidiate mode functions. + */ + + +#include "glheader.h" +#include "accum.h" +#include "api_loopback.h" +#include "api_exec.h" +#if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program +#include "shader/arbprogram.h" +#endif +#if FEATURE_ATI_fragment_shader +#include "shader/atifragshader.h" +#endif +#include "attrib.h" +#include "blend.h" +#if FEATURE_ARB_vertex_buffer_object +#include "bufferobj.h" +#endif +#include "arrayobj.h" +#include "buffers.h" +#include "clip.h" +#include "colortab.h" +#include "context.h" +#include "convolve.h" +#include "depth.h" +#include "dlist.h" +#include "drawpix.h" +#include "enable.h" +#include "eval.h" +#include "get.h" +#include "feedback.h" +#include "fog.h" +#if FEATURE_EXT_framebuffer_object +#include "fbobject.h" +#endif +#include "ffvertex_prog.h" +#include "framebuffer.h" +#include "hint.h" +#include "histogram.h" +#include "imports.h" +#include "light.h" +#include "lines.h" +#include "macros.h" +#include "matrix.h" +#include "pixel.h" +#include "pixelstore.h" +#include "points.h" +#include "polygon.h" +#if FEATURE_ARB_occlusion_query || FEATURE_EXT_timer_query +#include "queryobj.h" +#endif +#include "rastpos.h" +#include "readpix.h" +#include "state.h" +#include "stencil.h" +#include "teximage.h" +#include "texobj.h" +#include "texstate.h" +#include "mtypes.h" +#include "varray.h" +#if FEATURE_NV_vertex_program +#include "shader/nvprogram.h" +#endif +#if FEATURE_NV_fragment_program +#include "shader/nvprogram.h" +#include "shader/program.h" +#include "texenvprogram.h" +#endif +#if FEATURE_ARB_shader_objects +#include "shaders.h" +#endif +#include "debug.h" +#include "glapi/dispatch.h" + + + +/** + * Initialize a dispatch table with pointers to Mesa's immediate-mode + * commands. + * + * Pointers to glBegin()/glEnd() object commands and a few others + * are provided via the GLvertexformat interface. + * + * \param ctx GL context to which \c exec belongs. + * \param exec dispatch table. + */ +void +_mesa_init_exec_table(struct _glapi_table *exec) +{ +#if _HAVE_FULL_GL + _mesa_loopback_init_api_table( exec ); +#endif + + /* load the dispatch slots we understand */ + SET_AlphaFunc(exec, _mesa_AlphaFunc); + SET_BlendFunc(exec, _mesa_BlendFunc); + SET_Clear(exec, _mesa_Clear); + SET_ClearColor(exec, _mesa_ClearColor); + SET_ClearStencil(exec, _mesa_ClearStencil); + SET_ColorMask(exec, _mesa_ColorMask); + SET_CullFace(exec, _mesa_CullFace); + SET_Disable(exec, _mesa_Disable); + SET_DrawBuffer(exec, _mesa_DrawBuffer); + SET_Enable(exec, _mesa_Enable); + SET_Finish(exec, _mesa_Finish); + SET_Flush(exec, _mesa_Flush); + SET_FrontFace(exec, _mesa_FrontFace); + SET_Frustum(exec, _mesa_Frustum); + SET_GetError(exec, _mesa_GetError); + SET_GetFloatv(exec, _mesa_GetFloatv); + SET_GetString(exec, _mesa_GetString); + SET_InitNames(exec, _mesa_InitNames); + SET_LineStipple(exec, _mesa_LineStipple); + SET_LineWidth(exec, _mesa_LineWidth); + SET_LoadIdentity(exec, _mesa_LoadIdentity); + SET_LoadMatrixf(exec, _mesa_LoadMatrixf); + SET_LoadName(exec, _mesa_LoadName); + SET_LogicOp(exec, _mesa_LogicOp); + SET_MatrixMode(exec, _mesa_MatrixMode); + SET_MultMatrixf(exec, _mesa_MultMatrixf); + SET_Ortho(exec, _mesa_Ortho); + SET_PixelStorei(exec, _mesa_PixelStorei); + SET_PopMatrix(exec, _mesa_PopMatrix); + SET_PopName(exec, _mesa_PopName); + SET_PushMatrix(exec, _mesa_PushMatrix); + SET_PushName(exec, _mesa_PushName); + SET_RasterPos2f(exec, _mesa_RasterPos2f); + SET_RasterPos2fv(exec, _mesa_RasterPos2fv); + SET_RasterPos2i(exec, _mesa_RasterPos2i); + SET_RasterPos2iv(exec, _mesa_RasterPos2iv); + SET_ReadBuffer(exec, _mesa_ReadBuffer); + SET_RenderMode(exec, _mesa_RenderMode); + SET_Rotatef(exec, _mesa_Rotatef); + SET_Scalef(exec, _mesa_Scalef); + SET_Scissor(exec, _mesa_Scissor); + SET_SelectBuffer(exec, _mesa_SelectBuffer); + SET_ShadeModel(exec, _mesa_ShadeModel); + SET_StencilFunc(exec, _mesa_StencilFunc); + SET_StencilMask(exec, _mesa_StencilMask); + SET_StencilOp(exec, _mesa_StencilOp); + SET_TexEnvfv(exec, _mesa_TexEnvfv); + SET_TexEnvi(exec, _mesa_TexEnvi); + SET_TexImage2D(exec, _mesa_TexImage2D); + SET_TexParameteri(exec, _mesa_TexParameteri); + SET_Translatef(exec, _mesa_Translatef); + SET_Viewport(exec, _mesa_Viewport); +#if _HAVE_FULL_GL + SET_Accum(exec, _mesa_Accum); + SET_Bitmap(exec, _mesa_Bitmap); + SET_CallList(exec, _mesa_CallList); + SET_CallLists(exec, _mesa_CallLists); + SET_ClearAccum(exec, _mesa_ClearAccum); + SET_ClearDepth(exec, _mesa_ClearDepth); + SET_ClearIndex(exec, _mesa_ClearIndex); + SET_ClipPlane(exec, _mesa_ClipPlane); + SET_ColorMaterial(exec, _mesa_ColorMaterial); + SET_CopyPixels(exec, _mesa_CopyPixels); + SET_CullParameterfvEXT(exec, _mesa_CullParameterfvEXT); + SET_CullParameterdvEXT(exec, _mesa_CullParameterdvEXT); + SET_DeleteLists(exec, _mesa_DeleteLists); + SET_DepthFunc(exec, _mesa_DepthFunc); + SET_DepthMask(exec, _mesa_DepthMask); + SET_DepthRange(exec, _mesa_DepthRange); + SET_DrawPixels(exec, _mesa_DrawPixels); + SET_EndList(exec, _mesa_EndList); + SET_FeedbackBuffer(exec, _mesa_FeedbackBuffer); + SET_FogCoordPointerEXT(exec, _mesa_FogCoordPointerEXT); + SET_Fogf(exec, _mesa_Fogf); + SET_Fogfv(exec, _mesa_Fogfv); + SET_Fogi(exec, _mesa_Fogi); + SET_Fogiv(exec, _mesa_Fogiv); + SET_GenLists(exec, _mesa_GenLists); + SET_GetClipPlane(exec, _mesa_GetClipPlane); + SET_GetBooleanv(exec, _mesa_GetBooleanv); + SET_GetDoublev(exec, _mesa_GetDoublev); + SET_GetIntegerv(exec, _mesa_GetIntegerv); + SET_GetLightfv(exec, _mesa_GetLightfv); + SET_GetLightiv(exec, _mesa_GetLightiv); + SET_GetMapdv(exec, _mesa_GetMapdv); + SET_GetMapfv(exec, _mesa_GetMapfv); + SET_GetMapiv(exec, _mesa_GetMapiv); + SET_GetMaterialfv(exec, _mesa_GetMaterialfv); + SET_GetMaterialiv(exec, _mesa_GetMaterialiv); + SET_GetPixelMapfv(exec, _mesa_GetPixelMapfv); + SET_GetPixelMapuiv(exec, _mesa_GetPixelMapuiv); + SET_GetPixelMapusv(exec, _mesa_GetPixelMapusv); + SET_GetPolygonStipple(exec, _mesa_GetPolygonStipple); + SET_GetTexEnvfv(exec, _mesa_GetTexEnvfv); + SET_GetTexEnviv(exec, _mesa_GetTexEnviv); + SET_GetTexLevelParameterfv(exec, _mesa_GetTexLevelParameterfv); + SET_GetTexLevelParameteriv(exec, _mesa_GetTexLevelParameteriv); + SET_GetTexParameterfv(exec, _mesa_GetTexParameterfv); + SET_GetTexParameteriv(exec, _mesa_GetTexParameteriv); + SET_GetTexGendv(exec, _mesa_GetTexGendv); + SET_GetTexGenfv(exec, _mesa_GetTexGenfv); + SET_GetTexGeniv(exec, _mesa_GetTexGeniv); + SET_GetTexImage(exec, _mesa_GetTexImage); + SET_Hint(exec, _mesa_Hint); + SET_IndexMask(exec, _mesa_IndexMask); + SET_IsEnabled(exec, _mesa_IsEnabled); + SET_IsList(exec, _mesa_IsList); + SET_LightModelf(exec, _mesa_LightModelf); + SET_LightModelfv(exec, _mesa_LightModelfv); + SET_LightModeli(exec, _mesa_LightModeli); + SET_LightModeliv(exec, _mesa_LightModeliv); + SET_Lightf(exec, _mesa_Lightf); + SET_Lightfv(exec, _mesa_Lightfv); + SET_Lighti(exec, _mesa_Lighti); + SET_Lightiv(exec, _mesa_Lightiv); + SET_ListBase(exec, _mesa_ListBase); + SET_LoadMatrixd(exec, _mesa_LoadMatrixd); + SET_Map1d(exec, _mesa_Map1d); + SET_Map1f(exec, _mesa_Map1f); + SET_Map2d(exec, _mesa_Map2d); + SET_Map2f(exec, _mesa_Map2f); + SET_MapGrid1d(exec, _mesa_MapGrid1d); + SET_MapGrid1f(exec, _mesa_MapGrid1f); + SET_MapGrid2d(exec, _mesa_MapGrid2d); + SET_MapGrid2f(exec, _mesa_MapGrid2f); + SET_MultMatrixd(exec, _mesa_MultMatrixd); + SET_NewList(exec, _mesa_NewList); + SET_PassThrough(exec, _mesa_PassThrough); + SET_PixelMapfv(exec, _mesa_PixelMapfv); + SET_PixelMapuiv(exec, _mesa_PixelMapuiv); + SET_PixelMapusv(exec, _mesa_PixelMapusv); + SET_PixelStoref(exec, _mesa_PixelStoref); + SET_PixelTransferf(exec, _mesa_PixelTransferf); + SET_PixelTransferi(exec, _mesa_PixelTransferi); + SET_PixelZoom(exec, _mesa_PixelZoom); + SET_PointSize(exec, _mesa_PointSize); + SET_PolygonMode(exec, _mesa_PolygonMode); + SET_PolygonOffset(exec, _mesa_PolygonOffset); + SET_PolygonStipple(exec, _mesa_PolygonStipple); + SET_PopAttrib(exec, _mesa_PopAttrib); + SET_PushAttrib(exec, _mesa_PushAttrib); + SET_RasterPos2d(exec, _mesa_RasterPos2d); + SET_RasterPos2dv(exec, _mesa_RasterPos2dv); + SET_RasterPos2s(exec, _mesa_RasterPos2s); + SET_RasterPos2sv(exec, _mesa_RasterPos2sv); + SET_RasterPos3d(exec, _mesa_RasterPos3d); + SET_RasterPos3dv(exec, _mesa_RasterPos3dv); + SET_RasterPos3f(exec, _mesa_RasterPos3f); + SET_RasterPos3fv(exec, _mesa_RasterPos3fv); + SET_RasterPos3i(exec, _mesa_RasterPos3i); + SET_RasterPos3iv(exec, _mesa_RasterPos3iv); + SET_RasterPos3s(exec, _mesa_RasterPos3s); + SET_RasterPos3sv(exec, _mesa_RasterPos3sv); + SET_RasterPos4d(exec, _mesa_RasterPos4d); + SET_RasterPos4dv(exec, _mesa_RasterPos4dv); + SET_RasterPos4f(exec, _mesa_RasterPos4f); + SET_RasterPos4fv(exec, _mesa_RasterPos4fv); + SET_RasterPos4i(exec, _mesa_RasterPos4i); + SET_RasterPos4iv(exec, _mesa_RasterPos4iv); + SET_RasterPos4s(exec, _mesa_RasterPos4s); + SET_RasterPos4sv(exec, _mesa_RasterPos4sv); + SET_ReadPixels(exec, _mesa_ReadPixels); + SET_Rotated(exec, _mesa_Rotated); + SET_Scaled(exec, _mesa_Scaled); + SET_SecondaryColorPointerEXT(exec, _mesa_SecondaryColorPointerEXT); + SET_TexEnvf(exec, _mesa_TexEnvf); + SET_TexEnviv(exec, _mesa_TexEnviv); + SET_TexGend(exec, _mesa_TexGend); + SET_TexGendv(exec, _mesa_TexGendv); + SET_TexGenf(exec, _mesa_TexGenf); + SET_TexGenfv(exec, _mesa_TexGenfv); + SET_TexGeni(exec, _mesa_TexGeni); + SET_TexGeniv(exec, _mesa_TexGeniv); + SET_TexImage1D(exec, _mesa_TexImage1D); + SET_TexParameterf(exec, _mesa_TexParameterf); + SET_TexParameterfv(exec, _mesa_TexParameterfv); + SET_TexParameteriv(exec, _mesa_TexParameteriv); + SET_Translated(exec, _mesa_Translated); +#endif + + /* 1.1 */ + SET_BindTexture(exec, _mesa_BindTexture); + SET_DeleteTextures(exec, _mesa_DeleteTextures); + SET_GenTextures(exec, _mesa_GenTextures); +#if _HAVE_FULL_GL + SET_AreTexturesResident(exec, _mesa_AreTexturesResident); + SET_ColorPointer(exec, _mesa_ColorPointer); + SET_CopyTexImage1D(exec, _mesa_CopyTexImage1D); + SET_CopyTexImage2D(exec, _mesa_CopyTexImage2D); + SET_CopyTexSubImage1D(exec, _mesa_CopyTexSubImage1D); + SET_CopyTexSubImage2D(exec, _mesa_CopyTexSubImage2D); + SET_DisableClientState(exec, _mesa_DisableClientState); + SET_EdgeFlagPointer(exec, _mesa_EdgeFlagPointer); + SET_EnableClientState(exec, _mesa_EnableClientState); + SET_GetPointerv(exec, _mesa_GetPointerv); + SET_IndexPointer(exec, _mesa_IndexPointer); + SET_InterleavedArrays(exec, _mesa_InterleavedArrays); + SET_IsTexture(exec, _mesa_IsTexture); + SET_NormalPointer(exec, _mesa_NormalPointer); + SET_PopClientAttrib(exec, _mesa_PopClientAttrib); + SET_PrioritizeTextures(exec, _mesa_PrioritizeTextures); + SET_PushClientAttrib(exec, _mesa_PushClientAttrib); + SET_TexCoordPointer(exec, _mesa_TexCoordPointer); + SET_TexSubImage1D(exec, _mesa_TexSubImage1D); + SET_TexSubImage2D(exec, _mesa_TexSubImage2D); + SET_VertexPointer(exec, _mesa_VertexPointer); +#endif + + /* 1.2 */ +#if _HAVE_FULL_GL + SET_CopyTexSubImage3D(exec, _mesa_CopyTexSubImage3D); + SET_TexImage3D(exec, _mesa_TexImage3D); + SET_TexSubImage3D(exec, _mesa_TexSubImage3D); +#endif + + /* OpenGL 1.2 GL_ARB_imaging */ +#if _HAVE_FULL_GL + SET_BlendColor(exec, _mesa_BlendColor); + SET_BlendEquation(exec, _mesa_BlendEquation); + SET_BlendEquationSeparateEXT(exec, _mesa_BlendEquationSeparateEXT); + SET_ColorSubTable(exec, _mesa_ColorSubTable); + SET_ColorTable(exec, _mesa_ColorTable); + SET_ColorTableParameterfv(exec, _mesa_ColorTableParameterfv); + SET_ColorTableParameteriv(exec, _mesa_ColorTableParameteriv); + SET_ConvolutionFilter1D(exec, _mesa_ConvolutionFilter1D); + SET_ConvolutionFilter2D(exec, _mesa_ConvolutionFilter2D); + SET_ConvolutionParameterf(exec, _mesa_ConvolutionParameterf); + SET_ConvolutionParameterfv(exec, _mesa_ConvolutionParameterfv); + SET_ConvolutionParameteri(exec, _mesa_ConvolutionParameteri); + SET_ConvolutionParameteriv(exec, _mesa_ConvolutionParameteriv); + SET_CopyColorSubTable(exec, _mesa_CopyColorSubTable); + SET_CopyColorTable(exec, _mesa_CopyColorTable); + SET_CopyConvolutionFilter1D(exec, _mesa_CopyConvolutionFilter1D); + SET_CopyConvolutionFilter2D(exec, _mesa_CopyConvolutionFilter2D); + SET_GetColorTable(exec, _mesa_GetColorTable); + SET_GetColorTableParameterfv(exec, _mesa_GetColorTableParameterfv); + SET_GetColorTableParameteriv(exec, _mesa_GetColorTableParameteriv); + SET_GetConvolutionFilter(exec, _mesa_GetConvolutionFilter); + SET_GetConvolutionParameterfv(exec, _mesa_GetConvolutionParameterfv); + SET_GetConvolutionParameteriv(exec, _mesa_GetConvolutionParameteriv); + SET_GetHistogram(exec, _mesa_GetHistogram); + SET_GetHistogramParameterfv(exec, _mesa_GetHistogramParameterfv); + SET_GetHistogramParameteriv(exec, _mesa_GetHistogramParameteriv); + SET_GetMinmax(exec, _mesa_GetMinmax); + SET_GetMinmaxParameterfv(exec, _mesa_GetMinmaxParameterfv); + SET_GetMinmaxParameteriv(exec, _mesa_GetMinmaxParameteriv); + SET_GetSeparableFilter(exec, _mesa_GetSeparableFilter); + SET_Histogram(exec, _mesa_Histogram); + SET_Minmax(exec, _mesa_Minmax); + SET_ResetHistogram(exec, _mesa_ResetHistogram); + SET_ResetMinmax(exec, _mesa_ResetMinmax); + SET_SeparableFilter2D(exec, _mesa_SeparableFilter2D); +#endif + + /* OpenGL 2.0 */ + SET_StencilFuncSeparate(exec, _mesa_StencilFuncSeparate); + SET_StencilMaskSeparate(exec, _mesa_StencilMaskSeparate); + SET_StencilOpSeparate(exec, _mesa_StencilOpSeparate); +#if FEATURE_ARB_shader_objects + SET_AttachShader(exec, _mesa_AttachShader); + SET_CreateProgram(exec, _mesa_CreateProgram); + SET_CreateShader(exec, _mesa_CreateShader); + SET_DeleteProgram(exec, _mesa_DeleteProgram); + SET_DeleteShader(exec, _mesa_DeleteShader); + SET_DetachShader(exec, _mesa_DetachShader); + SET_GetAttachedShaders(exec, _mesa_GetAttachedShaders); + SET_GetProgramiv(exec, _mesa_GetProgramiv); + SET_GetProgramInfoLog(exec, _mesa_GetProgramInfoLog); + SET_GetShaderiv(exec, _mesa_GetShaderiv); + SET_GetShaderInfoLog(exec, _mesa_GetShaderInfoLog); + SET_IsProgram(exec, _mesa_IsProgram); + SET_IsShader(exec, _mesa_IsShader); +#endif + + /* OpenGL 2.1 */ +#if FEATURE_ARB_shader_objects + SET_UniformMatrix2x3fv(exec, _mesa_UniformMatrix2x3fv); + SET_UniformMatrix3x2fv(exec, _mesa_UniformMatrix3x2fv); + SET_UniformMatrix2x4fv(exec, _mesa_UniformMatrix2x4fv); + SET_UniformMatrix4x2fv(exec, _mesa_UniformMatrix4x2fv); + SET_UniformMatrix3x4fv(exec, _mesa_UniformMatrix3x4fv); + SET_UniformMatrix4x3fv(exec, _mesa_UniformMatrix4x3fv); +#endif + + + /* 2. GL_EXT_blend_color */ +#if 0 +/* SET_BlendColorEXT(exec, _mesa_BlendColorEXT); */ +#endif + + /* 3. GL_EXT_polygon_offset */ +#if _HAVE_FULL_GL + SET_PolygonOffsetEXT(exec, _mesa_PolygonOffsetEXT); +#endif + + /* 6. GL_EXT_texture3d */ +#if 0 +/* SET_CopyTexSubImage3DEXT(exec, _mesa_CopyTexSubImage3D); */ +/* SET_TexImage3DEXT(exec, _mesa_TexImage3DEXT); */ +/* SET_TexSubImage3DEXT(exec, _mesa_TexSubImage3D); */ +#endif + + /* 11. GL_EXT_histogram */ +#if 0 + SET_GetHistogramEXT(exec, _mesa_GetHistogram); + SET_GetHistogramParameterfvEXT(exec, _mesa_GetHistogramParameterfv); + SET_GetHistogramParameterivEXT(exec, _mesa_GetHistogramParameteriv); + SET_GetMinmaxEXT(exec, _mesa_GetMinmax); + SET_GetMinmaxParameterfvEXT(exec, _mesa_GetMinmaxParameterfv); + SET_GetMinmaxParameterivEXT(exec, _mesa_GetMinmaxParameteriv); +#endif + + /* 14. SGI_color_table */ +#if 0 + SET_ColorTableSGI(exec, _mesa_ColorTable); + SET_ColorSubTableSGI(exec, _mesa_ColorSubTable); + SET_GetColorTableSGI(exec, _mesa_GetColorTable); + SET_GetColorTableParameterfvSGI(exec, _mesa_GetColorTableParameterfv); + SET_GetColorTableParameterivSGI(exec, _mesa_GetColorTableParameteriv); +#endif + + /* 30. GL_EXT_vertex_array */ +#if _HAVE_FULL_GL + SET_ColorPointerEXT(exec, _mesa_ColorPointerEXT); + SET_EdgeFlagPointerEXT(exec, _mesa_EdgeFlagPointerEXT); + SET_IndexPointerEXT(exec, _mesa_IndexPointerEXT); + SET_NormalPointerEXT(exec, _mesa_NormalPointerEXT); + SET_TexCoordPointerEXT(exec, _mesa_TexCoordPointerEXT); + SET_VertexPointerEXT(exec, _mesa_VertexPointerEXT); +#endif + + /* 37. GL_EXT_blend_minmax */ +#if 0 + SET_BlendEquationEXT(exec, _mesa_BlendEquationEXT); +#endif + + /* 54. GL_EXT_point_parameters */ +#if _HAVE_FULL_GL + SET_PointParameterfEXT(exec, _mesa_PointParameterf); + SET_PointParameterfvEXT(exec, _mesa_PointParameterfv); +#endif + + /* 97. GL_EXT_compiled_vertex_array */ +#if _HAVE_FULL_GL + SET_LockArraysEXT(exec, _mesa_LockArraysEXT); + SET_UnlockArraysEXT(exec, _mesa_UnlockArraysEXT); +#endif + + /* 148. GL_EXT_multi_draw_arrays */ +#if _HAVE_FULL_GL + SET_MultiDrawArraysEXT(exec, _mesa_MultiDrawArraysEXT); + SET_MultiDrawElementsEXT(exec, _mesa_MultiDrawElementsEXT); +#endif + + /* 173. GL_INGR_blend_func_separate */ +#if _HAVE_FULL_GL + SET_BlendFuncSeparateEXT(exec, _mesa_BlendFuncSeparateEXT); +#endif + + /* 196. GL_MESA_resize_buffers */ +#if _HAVE_FULL_GL + SET_ResizeBuffersMESA(exec, _mesa_ResizeBuffersMESA); +#endif + + /* 197. GL_MESA_window_pos */ +#if _HAVE_FULL_GL + SET_WindowPos2dMESA(exec, _mesa_WindowPos2dMESA); + SET_WindowPos2dvMESA(exec, _mesa_WindowPos2dvMESA); + SET_WindowPos2fMESA(exec, _mesa_WindowPos2fMESA); + SET_WindowPos2fvMESA(exec, _mesa_WindowPos2fvMESA); + SET_WindowPos2iMESA(exec, _mesa_WindowPos2iMESA); + SET_WindowPos2ivMESA(exec, _mesa_WindowPos2ivMESA); + SET_WindowPos2sMESA(exec, _mesa_WindowPos2sMESA); + SET_WindowPos2svMESA(exec, _mesa_WindowPos2svMESA); + SET_WindowPos3dMESA(exec, _mesa_WindowPos3dMESA); + SET_WindowPos3dvMESA(exec, _mesa_WindowPos3dvMESA); + SET_WindowPos3fMESA(exec, _mesa_WindowPos3fMESA); + SET_WindowPos3fvMESA(exec, _mesa_WindowPos3fvMESA); + SET_WindowPos3iMESA(exec, _mesa_WindowPos3iMESA); + SET_WindowPos3ivMESA(exec, _mesa_WindowPos3ivMESA); + SET_WindowPos3sMESA(exec, _mesa_WindowPos3sMESA); + SET_WindowPos3svMESA(exec, _mesa_WindowPos3svMESA); + SET_WindowPos4dMESA(exec, _mesa_WindowPos4dMESA); + SET_WindowPos4dvMESA(exec, _mesa_WindowPos4dvMESA); + SET_WindowPos4fMESA(exec, _mesa_WindowPos4fMESA); + SET_WindowPos4fvMESA(exec, _mesa_WindowPos4fvMESA); + SET_WindowPos4iMESA(exec, _mesa_WindowPos4iMESA); + SET_WindowPos4ivMESA(exec, _mesa_WindowPos4ivMESA); + SET_WindowPos4sMESA(exec, _mesa_WindowPos4sMESA); + SET_WindowPos4svMESA(exec, _mesa_WindowPos4svMESA); +#endif + + /* 200. GL_IBM_multimode_draw_arrays */ +#if _HAVE_FULL_GL + SET_MultiModeDrawArraysIBM(exec, _mesa_MultiModeDrawArraysIBM); + SET_MultiModeDrawElementsIBM(exec, _mesa_MultiModeDrawElementsIBM); +#endif + + /* 233. GL_NV_vertex_program */ +#if FEATURE_NV_vertex_program + SET_BindProgramNV(exec, _mesa_BindProgram); + SET_DeleteProgramsNV(exec, _mesa_DeletePrograms); + SET_ExecuteProgramNV(exec, _mesa_ExecuteProgramNV); + SET_GenProgramsNV(exec, _mesa_GenPrograms); + SET_AreProgramsResidentNV(exec, _mesa_AreProgramsResidentNV); + SET_RequestResidentProgramsNV(exec, _mesa_RequestResidentProgramsNV); + SET_GetProgramParameterfvNV(exec, _mesa_GetProgramParameterfvNV); + SET_GetProgramParameterdvNV(exec, _mesa_GetProgramParameterdvNV); + SET_GetProgramivNV(exec, _mesa_GetProgramivNV); + SET_GetProgramStringNV(exec, _mesa_GetProgramStringNV); + SET_GetTrackMatrixivNV(exec, _mesa_GetTrackMatrixivNV); + SET_GetVertexAttribdvNV(exec, _mesa_GetVertexAttribdvNV); + SET_GetVertexAttribfvNV(exec, _mesa_GetVertexAttribfvNV); + SET_GetVertexAttribivNV(exec, _mesa_GetVertexAttribivNV); + SET_GetVertexAttribPointervNV(exec, _mesa_GetVertexAttribPointervNV); + SET_IsProgramNV(exec, _mesa_IsProgramARB); + SET_LoadProgramNV(exec, _mesa_LoadProgramNV); + SET_ProgramEnvParameter4dARB(exec, _mesa_ProgramEnvParameter4dARB); /* alias to ProgramParameter4dNV */ + SET_ProgramEnvParameter4dvARB(exec, _mesa_ProgramEnvParameter4dvARB); /* alias to ProgramParameter4dvNV */ + SET_ProgramEnvParameter4fARB(exec, _mesa_ProgramEnvParameter4fARB); /* alias to ProgramParameter4fNV */ + SET_ProgramEnvParameter4fvARB(exec, _mesa_ProgramEnvParameter4fvARB); /* alias to ProgramParameter4fvNV */ + SET_ProgramParameters4dvNV(exec, _mesa_ProgramParameters4dvNV); + SET_ProgramParameters4fvNV(exec, _mesa_ProgramParameters4fvNV); + SET_TrackMatrixNV(exec, _mesa_TrackMatrixNV); + SET_VertexAttribPointerNV(exec, _mesa_VertexAttribPointerNV); + /* glVertexAttrib*NV functions handled in api_loopback.c */ +#endif + + /* 273. GL_APPLE_vertex_array_object */ + SET_BindVertexArrayAPPLE(exec, _mesa_BindVertexArrayAPPLE); + SET_DeleteVertexArraysAPPLE(exec, _mesa_DeleteVertexArraysAPPLE); + SET_GenVertexArraysAPPLE(exec, _mesa_GenVertexArraysAPPLE); + SET_IsVertexArrayAPPLE(exec, _mesa_IsVertexArrayAPPLE); + + /* 282. GL_NV_fragment_program */ +#if FEATURE_NV_fragment_program + SET_ProgramNamedParameter4fNV(exec, _mesa_ProgramNamedParameter4fNV); + SET_ProgramNamedParameter4dNV(exec, _mesa_ProgramNamedParameter4dNV); + SET_ProgramNamedParameter4fvNV(exec, _mesa_ProgramNamedParameter4fvNV); + SET_ProgramNamedParameter4dvNV(exec, _mesa_ProgramNamedParameter4dvNV); + SET_GetProgramNamedParameterfvNV(exec, _mesa_GetProgramNamedParameterfvNV); + SET_GetProgramNamedParameterdvNV(exec, _mesa_GetProgramNamedParameterdvNV); + SET_ProgramLocalParameter4dARB(exec, _mesa_ProgramLocalParameter4dARB); + SET_ProgramLocalParameter4dvARB(exec, _mesa_ProgramLocalParameter4dvARB); + SET_ProgramLocalParameter4fARB(exec, _mesa_ProgramLocalParameter4fARB); + SET_ProgramLocalParameter4fvARB(exec, _mesa_ProgramLocalParameter4fvARB); + SET_GetProgramLocalParameterdvARB(exec, _mesa_GetProgramLocalParameterdvARB); + SET_GetProgramLocalParameterfvARB(exec, _mesa_GetProgramLocalParameterfvARB); +#endif + + /* 262. GL_NV_point_sprite */ +#if _HAVE_FULL_GL + SET_PointParameteriNV(exec, _mesa_PointParameteri); + SET_PointParameterivNV(exec, _mesa_PointParameteriv); +#endif + + /* 268. GL_EXT_stencil_two_side */ +#if _HAVE_FULL_GL + SET_ActiveStencilFaceEXT(exec, _mesa_ActiveStencilFaceEXT); +#endif + + /* ???. GL_EXT_depth_bounds_test */ + SET_DepthBoundsEXT(exec, _mesa_DepthBoundsEXT); + + /* ARB 1. GL_ARB_multitexture */ +#if _HAVE_FULL_GL + SET_ActiveTextureARB(exec, _mesa_ActiveTextureARB); + SET_ClientActiveTextureARB(exec, _mesa_ClientActiveTextureARB); +#endif + + /* ARB 3. GL_ARB_transpose_matrix */ +#if _HAVE_FULL_GL + SET_LoadTransposeMatrixdARB(exec, _mesa_LoadTransposeMatrixdARB); + SET_LoadTransposeMatrixfARB(exec, _mesa_LoadTransposeMatrixfARB); + SET_MultTransposeMatrixdARB(exec, _mesa_MultTransposeMatrixdARB); + SET_MultTransposeMatrixfARB(exec, _mesa_MultTransposeMatrixfARB); +#endif + + /* ARB 5. GL_ARB_multisample */ +#if _HAVE_FULL_GL + SET_SampleCoverageARB(exec, _mesa_SampleCoverageARB); +#endif + + /* ARB 12. GL_ARB_texture_compression */ +#if _HAVE_FULL_GL + SET_CompressedTexImage3DARB(exec, _mesa_CompressedTexImage3DARB); + SET_CompressedTexImage2DARB(exec, _mesa_CompressedTexImage2DARB); + SET_CompressedTexImage1DARB(exec, _mesa_CompressedTexImage1DARB); + SET_CompressedTexSubImage3DARB(exec, _mesa_CompressedTexSubImage3DARB); + SET_CompressedTexSubImage2DARB(exec, _mesa_CompressedTexSubImage2DARB); + SET_CompressedTexSubImage1DARB(exec, _mesa_CompressedTexSubImage1DARB); + SET_GetCompressedTexImageARB(exec, _mesa_GetCompressedTexImageARB); +#endif + + /* ARB 14. GL_ARB_point_parameters */ + /* reuse EXT_point_parameters functions */ + + /* ARB 26. GL_ARB_vertex_program */ + /* ARB 27. GL_ARB_fragment_program */ +#if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program + /* glVertexAttrib1sARB aliases glVertexAttrib1sNV */ + /* glVertexAttrib1fARB aliases glVertexAttrib1fNV */ + /* glVertexAttrib1dARB aliases glVertexAttrib1dNV */ + /* glVertexAttrib2sARB aliases glVertexAttrib2sNV */ + /* glVertexAttrib2fARB aliases glVertexAttrib2fNV */ + /* glVertexAttrib2dARB aliases glVertexAttrib2dNV */ + /* glVertexAttrib3sARB aliases glVertexAttrib3sNV */ + /* glVertexAttrib3fARB aliases glVertexAttrib3fNV */ + /* glVertexAttrib3dARB aliases glVertexAttrib3dNV */ + /* glVertexAttrib4sARB aliases glVertexAttrib4sNV */ + /* glVertexAttrib4fARB aliases glVertexAttrib4fNV */ + /* glVertexAttrib4dARB aliases glVertexAttrib4dNV */ + /* glVertexAttrib4NubARB aliases glVertexAttrib4NubNV */ + /* glVertexAttrib1svARB aliases glVertexAttrib1svNV */ + /* glVertexAttrib1fvARB aliases glVertexAttrib1fvNV */ + /* glVertexAttrib1dvARB aliases glVertexAttrib1dvNV */ + /* glVertexAttrib2svARB aliases glVertexAttrib2svNV */ + /* glVertexAttrib2fvARB aliases glVertexAttrib2fvNV */ + /* glVertexAttrib2dvARB aliases glVertexAttrib2dvNV */ + /* glVertexAttrib3svARB aliases glVertexAttrib3svNV */ + /* glVertexAttrib3fvARB aliases glVertexAttrib3fvNV */ + /* glVertexAttrib3dvARB aliases glVertexAttrib3dvNV */ + /* glVertexAttrib4svARB aliases glVertexAttrib4svNV */ + /* glVertexAttrib4fvARB aliases glVertexAttrib4fvNV */ + /* glVertexAttrib4dvARB aliases glVertexAttrib4dvNV */ + /* glVertexAttrib4NubvARB aliases glVertexAttrib4NubvNV */ + /* glVertexAttrib4bvARB handled in api_loopback.c */ + /* glVertexAttrib4ivARB handled in api_loopback.c */ + /* glVertexAttrib4ubvARB handled in api_loopback.c */ + /* glVertexAttrib4usvARB handled in api_loopback.c */ + /* glVertexAttrib4uivARB handled in api_loopback.c */ + /* glVertexAttrib4NbvARB handled in api_loopback.c */ + /* glVertexAttrib4NsvARB handled in api_loopback.c */ + /* glVertexAttrib4NivARB handled in api_loopback.c */ + /* glVertexAttrib4NusvARB handled in api_loopback.c */ + /* glVertexAttrib4NuivARB handled in api_loopback.c */ + SET_VertexAttribPointerARB(exec, _mesa_VertexAttribPointerARB); + SET_EnableVertexAttribArrayARB(exec, _mesa_EnableVertexAttribArrayARB); + SET_DisableVertexAttribArrayARB(exec, _mesa_DisableVertexAttribArrayARB); + SET_ProgramStringARB(exec, _mesa_ProgramStringARB); + /* glBindProgramARB aliases glBindProgramNV */ + /* glDeleteProgramsARB aliases glDeleteProgramsNV */ + /* glGenProgramsARB aliases glGenProgramsNV */ + /* glIsProgramARB aliases glIsProgramNV */ + SET_GetVertexAttribdvARB(exec, _mesa_GetVertexAttribdvARB); + SET_GetVertexAttribfvARB(exec, _mesa_GetVertexAttribfvARB); + SET_GetVertexAttribivARB(exec, _mesa_GetVertexAttribivARB); + /* glGetVertexAttribPointervARB aliases glGetVertexAttribPointervNV */ + SET_ProgramEnvParameter4dARB(exec, _mesa_ProgramEnvParameter4dARB); + SET_ProgramEnvParameter4dvARB(exec, _mesa_ProgramEnvParameter4dvARB); + SET_ProgramEnvParameter4fARB(exec, _mesa_ProgramEnvParameter4fARB); + SET_ProgramEnvParameter4fvARB(exec, _mesa_ProgramEnvParameter4fvARB); + SET_ProgramLocalParameter4dARB(exec, _mesa_ProgramLocalParameter4dARB); + SET_ProgramLocalParameter4dvARB(exec, _mesa_ProgramLocalParameter4dvARB); + SET_ProgramLocalParameter4fARB(exec, _mesa_ProgramLocalParameter4fARB); + SET_ProgramLocalParameter4fvARB(exec, _mesa_ProgramLocalParameter4fvARB); + SET_GetProgramEnvParameterdvARB(exec, _mesa_GetProgramEnvParameterdvARB); + SET_GetProgramEnvParameterfvARB(exec, _mesa_GetProgramEnvParameterfvARB); + SET_GetProgramLocalParameterdvARB(exec, _mesa_GetProgramLocalParameterdvARB); + SET_GetProgramLocalParameterfvARB(exec, _mesa_GetProgramLocalParameterfvARB); + SET_GetProgramivARB(exec, _mesa_GetProgramivARB); + SET_GetProgramStringARB(exec, _mesa_GetProgramStringARB); +#endif + + /* ARB 28. GL_ARB_vertex_buffer_object */ +#if FEATURE_ARB_vertex_buffer_object + SET_BindBufferARB(exec, _mesa_BindBufferARB); + SET_BufferDataARB(exec, _mesa_BufferDataARB); + SET_BufferSubDataARB(exec, _mesa_BufferSubDataARB); + SET_DeleteBuffersARB(exec, _mesa_DeleteBuffersARB); + SET_GenBuffersARB(exec, _mesa_GenBuffersARB); + SET_GetBufferParameterivARB(exec, _mesa_GetBufferParameterivARB); + SET_GetBufferPointervARB(exec, _mesa_GetBufferPointervARB); + SET_GetBufferSubDataARB(exec, _mesa_GetBufferSubDataARB); + SET_IsBufferARB(exec, _mesa_IsBufferARB); + SET_MapBufferARB(exec, _mesa_MapBufferARB); + SET_UnmapBufferARB(exec, _mesa_UnmapBufferARB); +#endif + + /* ARB 29. GL_ARB_occlusion_query */ +#if FEATURE_ARB_occlusion_query + SET_GenQueriesARB(exec, _mesa_GenQueriesARB); + SET_DeleteQueriesARB(exec, _mesa_DeleteQueriesARB); + SET_IsQueryARB(exec, _mesa_IsQueryARB); + SET_BeginQueryARB(exec, _mesa_BeginQueryARB); + SET_EndQueryARB(exec, _mesa_EndQueryARB); + SET_GetQueryivARB(exec, _mesa_GetQueryivARB); + SET_GetQueryObjectivARB(exec, _mesa_GetQueryObjectivARB); + SET_GetQueryObjectuivARB(exec, _mesa_GetQueryObjectuivARB); +#endif + + /* ARB 37. GL_ARB_draw_buffers */ + SET_DrawBuffersARB(exec, _mesa_DrawBuffersARB); + +#if FEATURE_ARB_shader_objects + SET_DeleteObjectARB(exec, _mesa_DeleteObjectARB); + SET_GetHandleARB(exec, _mesa_GetHandleARB); + SET_DetachObjectARB(exec, _mesa_DetachObjectARB); + SET_CreateShaderObjectARB(exec, _mesa_CreateShaderObjectARB); + SET_ShaderSourceARB(exec, _mesa_ShaderSourceARB); + SET_CompileShaderARB(exec, _mesa_CompileShaderARB); + SET_CreateProgramObjectARB(exec, _mesa_CreateProgramObjectARB); + SET_AttachObjectARB(exec, _mesa_AttachObjectARB); + SET_LinkProgramARB(exec, _mesa_LinkProgramARB); + SET_UseProgramObjectARB(exec, _mesa_UseProgramObjectARB); + SET_ValidateProgramARB(exec, _mesa_ValidateProgramARB); + SET_Uniform1fARB(exec, _mesa_Uniform1fARB); + SET_Uniform2fARB(exec, _mesa_Uniform2fARB); + SET_Uniform3fARB(exec, _mesa_Uniform3fARB); + SET_Uniform4fARB(exec, _mesa_Uniform4fARB); + SET_Uniform1iARB(exec, _mesa_Uniform1iARB); + SET_Uniform2iARB(exec, _mesa_Uniform2iARB); + SET_Uniform3iARB(exec, _mesa_Uniform3iARB); + SET_Uniform4iARB(exec, _mesa_Uniform4iARB); + SET_Uniform1fvARB(exec, _mesa_Uniform1fvARB); + SET_Uniform2fvARB(exec, _mesa_Uniform2fvARB); + SET_Uniform3fvARB(exec, _mesa_Uniform3fvARB); + SET_Uniform4fvARB(exec, _mesa_Uniform4fvARB); + SET_Uniform1ivARB(exec, _mesa_Uniform1ivARB); + SET_Uniform2ivARB(exec, _mesa_Uniform2ivARB); + SET_Uniform3ivARB(exec, _mesa_Uniform3ivARB); + SET_Uniform4ivARB(exec, _mesa_Uniform4ivARB); + SET_UniformMatrix2fvARB(exec, _mesa_UniformMatrix2fvARB); + SET_UniformMatrix3fvARB(exec, _mesa_UniformMatrix3fvARB); + SET_UniformMatrix4fvARB(exec, _mesa_UniformMatrix4fvARB); + SET_GetObjectParameterfvARB(exec, _mesa_GetObjectParameterfvARB); + SET_GetObjectParameterivARB(exec, _mesa_GetObjectParameterivARB); + SET_GetInfoLogARB(exec, _mesa_GetInfoLogARB); + SET_GetAttachedObjectsARB(exec, _mesa_GetAttachedObjectsARB); + SET_GetUniformLocationARB(exec, _mesa_GetUniformLocationARB); + SET_GetActiveUniformARB(exec, _mesa_GetActiveUniformARB); + SET_GetUniformfvARB(exec, _mesa_GetUniformfvARB); + SET_GetUniformivARB(exec, _mesa_GetUniformivARB); + SET_GetShaderSourceARB(exec, _mesa_GetShaderSourceARB); +#endif /* FEATURE_ARB_shader_objects */ + +#if FEATURE_ARB_vertex_shader + SET_BindAttribLocationARB(exec, _mesa_BindAttribLocationARB); + SET_GetActiveAttribARB(exec, _mesa_GetActiveAttribARB); + SET_GetAttribLocationARB(exec, _mesa_GetAttribLocationARB); +#endif /* FEATURE_ARB_vertex_shader */ + + /* GL_ATI_fragment_shader */ +#if FEATURE_ATI_fragment_shader + SET_GenFragmentShadersATI(exec, _mesa_GenFragmentShadersATI); + SET_BindFragmentShaderATI(exec, _mesa_BindFragmentShaderATI); + SET_DeleteFragmentShaderATI(exec, _mesa_DeleteFragmentShaderATI); + SET_BeginFragmentShaderATI(exec, _mesa_BeginFragmentShaderATI); + SET_EndFragmentShaderATI(exec, _mesa_EndFragmentShaderATI); + SET_PassTexCoordATI(exec, _mesa_PassTexCoordATI); + SET_SampleMapATI(exec, _mesa_SampleMapATI); + SET_ColorFragmentOp1ATI(exec, _mesa_ColorFragmentOp1ATI); + SET_ColorFragmentOp2ATI(exec, _mesa_ColorFragmentOp2ATI); + SET_ColorFragmentOp3ATI(exec, _mesa_ColorFragmentOp3ATI); + SET_AlphaFragmentOp1ATI(exec, _mesa_AlphaFragmentOp1ATI); + SET_AlphaFragmentOp2ATI(exec, _mesa_AlphaFragmentOp2ATI); + SET_AlphaFragmentOp3ATI(exec, _mesa_AlphaFragmentOp3ATI); + SET_SetFragmentShaderConstantATI(exec, _mesa_SetFragmentShaderConstantATI); +#endif + +#if FEATURE_EXT_framebuffer_object + SET_IsRenderbufferEXT(exec, _mesa_IsRenderbufferEXT); + SET_BindRenderbufferEXT(exec, _mesa_BindRenderbufferEXT); + SET_DeleteRenderbuffersEXT(exec, _mesa_DeleteRenderbuffersEXT); + SET_GenRenderbuffersEXT(exec, _mesa_GenRenderbuffersEXT); + SET_RenderbufferStorageEXT(exec, _mesa_RenderbufferStorageEXT); + SET_GetRenderbufferParameterivEXT(exec, _mesa_GetRenderbufferParameterivEXT); + SET_IsFramebufferEXT(exec, _mesa_IsFramebufferEXT); + SET_BindFramebufferEXT(exec, _mesa_BindFramebufferEXT); + SET_DeleteFramebuffersEXT(exec, _mesa_DeleteFramebuffersEXT); + SET_GenFramebuffersEXT(exec, _mesa_GenFramebuffersEXT); + SET_CheckFramebufferStatusEXT(exec, _mesa_CheckFramebufferStatusEXT); + SET_FramebufferTexture1DEXT(exec, _mesa_FramebufferTexture1DEXT); + SET_FramebufferTexture2DEXT(exec, _mesa_FramebufferTexture2DEXT); + SET_FramebufferTexture3DEXT(exec, _mesa_FramebufferTexture3DEXT); + SET_FramebufferRenderbufferEXT(exec, _mesa_FramebufferRenderbufferEXT); + SET_GetFramebufferAttachmentParameterivEXT(exec, _mesa_GetFramebufferAttachmentParameterivEXT); + SET_GenerateMipmapEXT(exec, _mesa_GenerateMipmapEXT); +#endif + +#if FEATURE_EXT_timer_query + SET_GetQueryObjecti64vEXT(exec, _mesa_GetQueryObjecti64vEXT); + SET_GetQueryObjectui64vEXT(exec, _mesa_GetQueryObjectui64vEXT); +#endif + +#if FEATURE_EXT_framebuffer_blit + SET_BlitFramebufferEXT(exec, _mesa_BlitFramebufferEXT); +#endif + + /* GL_EXT_gpu_program_parameters */ +#if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program + SET_ProgramEnvParameters4fvEXT(exec, _mesa_ProgramEnvParameters4fvEXT); + SET_ProgramLocalParameters4fvEXT(exec, _mesa_ProgramLocalParameters4fvEXT); +#endif + + /* GL_MESA_texture_array / GL_EXT_texture_array */ +#if FEATURE_EXT_framebuffer_object + SET_FramebufferTextureLayerEXT(exec, _mesa_FramebufferTextureLayerEXT); +#endif + + /* GL_ATI_separate_stencil */ + SET_StencilFuncSeparateATI(exec, _mesa_StencilFuncSeparateATI); +} + diff --git a/src/mesa/main/api_exec.h b/src/mesa/main/api_exec.h new file mode 100644 index 0000000000..7f15ea9d00 --- /dev/null +++ b/src/mesa/main/api_exec.h @@ -0,0 +1,34 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef API_EXEC_H +#define API_EXEC_H + + +void +_mesa_init_exec_table(struct _glapi_table *exec); + + +#endif diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 522420592c..465d6b3f89 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -79,6 +79,7 @@ #include "glheader.h" #include "imports.h" #include "accum.h" +#include "api_exec.h" #include "arrayobj.h" #include "attrib.h" #include "blend.h" diff --git a/src/mesa/main/sources b/src/mesa/main/sources index 8ef5bd62fa..06f32f2275 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -2,6 +2,7 @@ MESA_MAIN_SOURCES = \ accum.c \ api_arrayelt.c \ +api_exec.c \ api_loopback.c \ api_noop.c \ api_validate.c \ @@ -72,6 +73,7 @@ MESA_MAIN_HEADERS = \ accum.h \ api_arrayelt.h \ api_eval.h \ +api_exec.h \ api_loopback.h \ api_noop.h \ api_validate.h \ diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 64f911e699..89920f8443 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -27,809 +27,26 @@ * \file state.c * State management. * - * This file manages recalculation of derived values in the __GLcontextRec. - * Also, this is where we initialize the API dispatch table. + * This file manages recalculation of derived values in GLcontext. */ + #include "glheader.h" -#include "accum.h" -#include "api_loopback.h" -#if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program -#include "shader/arbprogram.h" -#endif -#if FEATURE_ATI_fragment_shader -#include "shader/atifragshader.h" -#endif -#include "attrib.h" -#include "blend.h" -#if FEATURE_ARB_vertex_buffer_object -#include "bufferobj.h" -#endif -#include "arrayobj.h" -#include "buffers.h" -#include "clip.h" -#include "colortab.h" +#include "mtypes.h" #include "context.h" -#include "convolve.h" -#include "depth.h" -#include "dlist.h" -#include "drawpix.h" -#include "enable.h" -#include "eval.h" -#include "get.h" -#include "feedback.h" -#include "fog.h" -#if FEATURE_EXT_framebuffer_object -#include "fbobject.h" -#endif +#include "debug.h" +#include "macros.h" #include "ffvertex_prog.h" #include "framebuffer.h" -#include "hint.h" -#include "histogram.h" -#include "imports.h" #include "light.h" -#include "lines.h" -#include "macros.h" #include "matrix.h" #include "pixel.h" -#include "pixelstore.h" -#include "points.h" -#include "polygon.h" -#if FEATURE_ARB_occlusion_query || FEATURE_EXT_timer_query -#include "queryobj.h" -#endif -#include "rastpos.h" -#include "readpix.h" +#include "shader/program.h" #include "state.h" #include "stencil.h" -#include "teximage.h" +#include "texenvprogram.h" #include "texobj.h" #include "texstate.h" -#include "mtypes.h" -#include "varray.h" -#if FEATURE_NV_vertex_program -#include "shader/nvprogram.h" -#endif -#if FEATURE_NV_fragment_program -#include "shader/nvprogram.h" -#include "shader/program.h" -#include "texenvprogram.h" -#endif -#if FEATURE_ARB_shader_objects -#include "shaders.h" -#endif -#include "debug.h" -#include "glapi/dispatch.h" - - - -/** - * Initialize a dispatch table with pointers to Mesa's immediate-mode - * commands. - * - * Pointers to glBegin()/glEnd() object commands and a few others - * are provided via the GLvertexformat interface. - * - * \param ctx GL context to which \c exec belongs. - * \param exec dispatch table. - */ -void -_mesa_init_exec_table(struct _glapi_table *exec) -{ -#if _HAVE_FULL_GL - _mesa_loopback_init_api_table( exec ); -#endif - - /* load the dispatch slots we understand */ - SET_AlphaFunc(exec, _mesa_AlphaFunc); - SET_BlendFunc(exec, _mesa_BlendFunc); - SET_Clear(exec, _mesa_Clear); - SET_ClearColor(exec, _mesa_ClearColor); - SET_ClearStencil(exec, _mesa_ClearStencil); - SET_ColorMask(exec, _mesa_ColorMask); - SET_CullFace(exec, _mesa_CullFace); - SET_Disable(exec, _mesa_Disable); - SET_DrawBuffer(exec, _mesa_DrawBuffer); - SET_Enable(exec, _mesa_Enable); - SET_Finish(exec, _mesa_Finish); - SET_Flush(exec, _mesa_Flush); - SET_FrontFace(exec, _mesa_FrontFace); - SET_Frustum(exec, _mesa_Frustum); - SET_GetError(exec, _mesa_GetError); - SET_GetFloatv(exec, _mesa_GetFloatv); - SET_GetString(exec, _mesa_GetString); - SET_InitNames(exec, _mesa_InitNames); - SET_LineStipple(exec, _mesa_LineStipple); - SET_LineWidth(exec, _mesa_LineWidth); - SET_LoadIdentity(exec, _mesa_LoadIdentity); - SET_LoadMatrixf(exec, _mesa_LoadMatrixf); - SET_LoadName(exec, _mesa_LoadName); - SET_LogicOp(exec, _mesa_LogicOp); - SET_MatrixMode(exec, _mesa_MatrixMode); - SET_MultMatrixf(exec, _mesa_MultMatrixf); - SET_Ortho(exec, _mesa_Ortho); - SET_PixelStorei(exec, _mesa_PixelStorei); - SET_PopMatrix(exec, _mesa_PopMatrix); - SET_PopName(exec, _mesa_PopName); - SET_PushMatrix(exec, _mesa_PushMatrix); - SET_PushName(exec, _mesa_PushName); - SET_RasterPos2f(exec, _mesa_RasterPos2f); - SET_RasterPos2fv(exec, _mesa_RasterPos2fv); - SET_RasterPos2i(exec, _mesa_RasterPos2i); - SET_RasterPos2iv(exec, _mesa_RasterPos2iv); - SET_ReadBuffer(exec, _mesa_ReadBuffer); - SET_RenderMode(exec, _mesa_RenderMode); - SET_Rotatef(exec, _mesa_Rotatef); - SET_Scalef(exec, _mesa_Scalef); - SET_Scissor(exec, _mesa_Scissor); - SET_SelectBuffer(exec, _mesa_SelectBuffer); - SET_ShadeModel(exec, _mesa_ShadeModel); - SET_StencilFunc(exec, _mesa_StencilFunc); - SET_StencilMask(exec, _mesa_StencilMask); - SET_StencilOp(exec, _mesa_StencilOp); - SET_TexEnvfv(exec, _mesa_TexEnvfv); - SET_TexEnvi(exec, _mesa_TexEnvi); - SET_TexImage2D(exec, _mesa_TexImage2D); - SET_TexParameteri(exec, _mesa_TexParameteri); - SET_Translatef(exec, _mesa_Translatef); - SET_Viewport(exec, _mesa_Viewport); -#if _HAVE_FULL_GL - SET_Accum(exec, _mesa_Accum); - SET_Bitmap(exec, _mesa_Bitmap); - SET_CallList(exec, _mesa_CallList); - SET_CallLists(exec, _mesa_CallLists); - SET_ClearAccum(exec, _mesa_ClearAccum); - SET_ClearDepth(exec, _mesa_ClearDepth); - SET_ClearIndex(exec, _mesa_ClearIndex); - SET_ClipPlane(exec, _mesa_ClipPlane); - SET_ColorMaterial(exec, _mesa_ColorMaterial); - SET_CopyPixels(exec, _mesa_CopyPixels); - SET_CullParameterfvEXT(exec, _mesa_CullParameterfvEXT); - SET_CullParameterdvEXT(exec, _mesa_CullParameterdvEXT); - SET_DeleteLists(exec, _mesa_DeleteLists); - SET_DepthFunc(exec, _mesa_DepthFunc); - SET_DepthMask(exec, _mesa_DepthMask); - SET_DepthRange(exec, _mesa_DepthRange); - SET_DrawPixels(exec, _mesa_DrawPixels); - SET_EndList(exec, _mesa_EndList); - SET_FeedbackBuffer(exec, _mesa_FeedbackBuffer); - SET_FogCoordPointerEXT(exec, _mesa_FogCoordPointerEXT); - SET_Fogf(exec, _mesa_Fogf); - SET_Fogfv(exec, _mesa_Fogfv); - SET_Fogi(exec, _mesa_Fogi); - SET_Fogiv(exec, _mesa_Fogiv); - SET_GenLists(exec, _mesa_GenLists); - SET_GetClipPlane(exec, _mesa_GetClipPlane); - SET_GetBooleanv(exec, _mesa_GetBooleanv); - SET_GetDoublev(exec, _mesa_GetDoublev); - SET_GetIntegerv(exec, _mesa_GetIntegerv); - SET_GetLightfv(exec, _mesa_GetLightfv); - SET_GetLightiv(exec, _mesa_GetLightiv); - SET_GetMapdv(exec, _mesa_GetMapdv); - SET_GetMapfv(exec, _mesa_GetMapfv); - SET_GetMapiv(exec, _mesa_GetMapiv); - SET_GetMaterialfv(exec, _mesa_GetMaterialfv); - SET_GetMaterialiv(exec, _mesa_GetMaterialiv); - SET_GetPixelMapfv(exec, _mesa_GetPixelMapfv); - SET_GetPixelMapuiv(exec, _mesa_GetPixelMapuiv); - SET_GetPixelMapusv(exec, _mesa_GetPixelMapusv); - SET_GetPolygonStipple(exec, _mesa_GetPolygonStipple); - SET_GetTexEnvfv(exec, _mesa_GetTexEnvfv); - SET_GetTexEnviv(exec, _mesa_GetTexEnviv); - SET_GetTexLevelParameterfv(exec, _mesa_GetTexLevelParameterfv); - SET_GetTexLevelParameteriv(exec, _mesa_GetTexLevelParameteriv); - SET_GetTexParameterfv(exec, _mesa_GetTexParameterfv); - SET_GetTexParameteriv(exec, _mesa_GetTexParameteriv); - SET_GetTexGendv(exec, _mesa_GetTexGendv); - SET_GetTexGenfv(exec, _mesa_GetTexGenfv); - SET_GetTexGeniv(exec, _mesa_GetTexGeniv); - SET_GetTexImage(exec, _mesa_GetTexImage); - SET_Hint(exec, _mesa_Hint); - SET_IndexMask(exec, _mesa_IndexMask); - SET_IsEnabled(exec, _mesa_IsEnabled); - SET_IsList(exec, _mesa_IsList); - SET_LightModelf(exec, _mesa_LightModelf); - SET_LightModelfv(exec, _mesa_LightModelfv); - SET_LightModeli(exec, _mesa_LightModeli); - SET_LightModeliv(exec, _mesa_LightModeliv); - SET_Lightf(exec, _mesa_Lightf); - SET_Lightfv(exec, _mesa_Lightfv); - SET_Lighti(exec, _mesa_Lighti); - SET_Lightiv(exec, _mesa_Lightiv); - SET_ListBase(exec, _mesa_ListBase); - SET_LoadMatrixd(exec, _mesa_LoadMatrixd); - SET_Map1d(exec, _mesa_Map1d); - SET_Map1f(exec, _mesa_Map1f); - SET_Map2d(exec, _mesa_Map2d); - SET_Map2f(exec, _mesa_Map2f); - SET_MapGrid1d(exec, _mesa_MapGrid1d); - SET_MapGrid1f(exec, _mesa_MapGrid1f); - SET_MapGrid2d(exec, _mesa_MapGrid2d); - SET_MapGrid2f(exec, _mesa_MapGrid2f); - SET_MultMatrixd(exec, _mesa_MultMatrixd); - SET_NewList(exec, _mesa_NewList); - SET_PassThrough(exec, _mesa_PassThrough); - SET_PixelMapfv(exec, _mesa_PixelMapfv); - SET_PixelMapuiv(exec, _mesa_PixelMapuiv); - SET_PixelMapusv(exec, _mesa_PixelMapusv); - SET_PixelStoref(exec, _mesa_PixelStoref); - SET_PixelTransferf(exec, _mesa_PixelTransferf); - SET_PixelTransferi(exec, _mesa_PixelTransferi); - SET_PixelZoom(exec, _mesa_PixelZoom); - SET_PointSize(exec, _mesa_PointSize); - SET_PolygonMode(exec, _mesa_PolygonMode); - SET_PolygonOffset(exec, _mesa_PolygonOffset); - SET_PolygonStipple(exec, _mesa_PolygonStipple); - SET_PopAttrib(exec, _mesa_PopAttrib); - SET_PushAttrib(exec, _mesa_PushAttrib); - SET_RasterPos2d(exec, _mesa_RasterPos2d); - SET_RasterPos2dv(exec, _mesa_RasterPos2dv); - SET_RasterPos2s(exec, _mesa_RasterPos2s); - SET_RasterPos2sv(exec, _mesa_RasterPos2sv); - SET_RasterPos3d(exec, _mesa_RasterPos3d); - SET_RasterPos3dv(exec, _mesa_RasterPos3dv); - SET_RasterPos3f(exec, _mesa_RasterPos3f); - SET_RasterPos3fv(exec, _mesa_RasterPos3fv); - SET_RasterPos3i(exec, _mesa_RasterPos3i); - SET_RasterPos3iv(exec, _mesa_RasterPos3iv); - SET_RasterPos3s(exec, _mesa_RasterPos3s); - SET_RasterPos3sv(exec, _mesa_RasterPos3sv); - SET_RasterPos4d(exec, _mesa_RasterPos4d); - SET_RasterPos4dv(exec, _mesa_RasterPos4dv); - SET_RasterPos4f(exec, _mesa_RasterPos4f); - SET_RasterPos4fv(exec, _mesa_RasterPos4fv); - SET_RasterPos4i(exec, _mesa_RasterPos4i); - SET_RasterPos4iv(exec, _mesa_RasterPos4iv); - SET_RasterPos4s(exec, _mesa_RasterPos4s); - SET_RasterPos4sv(exec, _mesa_RasterPos4sv); - SET_ReadPixels(exec, _mesa_ReadPixels); - SET_Rotated(exec, _mesa_Rotated); - SET_Scaled(exec, _mesa_Scaled); - SET_SecondaryColorPointerEXT(exec, _mesa_SecondaryColorPointerEXT); - SET_TexEnvf(exec, _mesa_TexEnvf); - SET_TexEnviv(exec, _mesa_TexEnviv); - SET_TexGend(exec, _mesa_TexGend); - SET_TexGendv(exec, _mesa_TexGendv); - SET_TexGenf(exec, _mesa_TexGenf); - SET_TexGenfv(exec, _mesa_TexGenfv); - SET_TexGeni(exec, _mesa_TexGeni); - SET_TexGeniv(exec, _mesa_TexGeniv); - SET_TexImage1D(exec, _mesa_TexImage1D); - SET_TexParameterf(exec, _mesa_TexParameterf); - SET_TexParameterfv(exec, _mesa_TexParameterfv); - SET_TexParameteriv(exec, _mesa_TexParameteriv); - SET_Translated(exec, _mesa_Translated); -#endif - - /* 1.1 */ - SET_BindTexture(exec, _mesa_BindTexture); - SET_DeleteTextures(exec, _mesa_DeleteTextures); - SET_GenTextures(exec, _mesa_GenTextures); -#if _HAVE_FULL_GL - SET_AreTexturesResident(exec, _mesa_AreTexturesResident); - SET_ColorPointer(exec, _mesa_ColorPointer); - SET_CopyTexImage1D(exec, _mesa_CopyTexImage1D); - SET_CopyTexImage2D(exec, _mesa_CopyTexImage2D); - SET_CopyTexSubImage1D(exec, _mesa_CopyTexSubImage1D); - SET_CopyTexSubImage2D(exec, _mesa_CopyTexSubImage2D); - SET_DisableClientState(exec, _mesa_DisableClientState); - SET_EdgeFlagPointer(exec, _mesa_EdgeFlagPointer); - SET_EnableClientState(exec, _mesa_EnableClientState); - SET_GetPointerv(exec, _mesa_GetPointerv); - SET_IndexPointer(exec, _mesa_IndexPointer); - SET_InterleavedArrays(exec, _mesa_InterleavedArrays); - SET_IsTexture(exec, _mesa_IsTexture); - SET_NormalPointer(exec, _mesa_NormalPointer); - SET_PopClientAttrib(exec, _mesa_PopClientAttrib); - SET_PrioritizeTextures(exec, _mesa_PrioritizeTextures); - SET_PushClientAttrib(exec, _mesa_PushClientAttrib); - SET_TexCoordPointer(exec, _mesa_TexCoordPointer); - SET_TexSubImage1D(exec, _mesa_TexSubImage1D); - SET_TexSubImage2D(exec, _mesa_TexSubImage2D); - SET_VertexPointer(exec, _mesa_VertexPointer); -#endif - - /* 1.2 */ -#if _HAVE_FULL_GL - SET_CopyTexSubImage3D(exec, _mesa_CopyTexSubImage3D); - SET_TexImage3D(exec, _mesa_TexImage3D); - SET_TexSubImage3D(exec, _mesa_TexSubImage3D); -#endif - - /* OpenGL 1.2 GL_ARB_imaging */ -#if _HAVE_FULL_GL - SET_BlendColor(exec, _mesa_BlendColor); - SET_BlendEquation(exec, _mesa_BlendEquation); - SET_BlendEquationSeparateEXT(exec, _mesa_BlendEquationSeparateEXT); - SET_ColorSubTable(exec, _mesa_ColorSubTable); - SET_ColorTable(exec, _mesa_ColorTable); - SET_ColorTableParameterfv(exec, _mesa_ColorTableParameterfv); - SET_ColorTableParameteriv(exec, _mesa_ColorTableParameteriv); - SET_ConvolutionFilter1D(exec, _mesa_ConvolutionFilter1D); - SET_ConvolutionFilter2D(exec, _mesa_ConvolutionFilter2D); - SET_ConvolutionParameterf(exec, _mesa_ConvolutionParameterf); - SET_ConvolutionParameterfv(exec, _mesa_ConvolutionParameterfv); - SET_ConvolutionParameteri(exec, _mesa_ConvolutionParameteri); - SET_ConvolutionParameteriv(exec, _mesa_ConvolutionParameteriv); - SET_CopyColorSubTable(exec, _mesa_CopyColorSubTable); - SET_CopyColorTable(exec, _mesa_CopyColorTable); - SET_CopyConvolutionFilter1D(exec, _mesa_CopyConvolutionFilter1D); - SET_CopyConvolutionFilter2D(exec, _mesa_CopyConvolutionFilter2D); - SET_GetColorTable(exec, _mesa_GetColorTable); - SET_GetColorTableParameterfv(exec, _mesa_GetColorTableParameterfv); - SET_GetColorTableParameteriv(exec, _mesa_GetColorTableParameteriv); - SET_GetConvolutionFilter(exec, _mesa_GetConvolutionFilter); - SET_GetConvolutionParameterfv(exec, _mesa_GetConvolutionParameterfv); - SET_GetConvolutionParameteriv(exec, _mesa_GetConvolutionParameteriv); - SET_GetHistogram(exec, _mesa_GetHistogram); - SET_GetHistogramParameterfv(exec, _mesa_GetHistogramParameterfv); - SET_GetHistogramParameteriv(exec, _mesa_GetHistogramParameteriv); - SET_GetMinmax(exec, _mesa_GetMinmax); - SET_GetMinmaxParameterfv(exec, _mesa_GetMinmaxParameterfv); - SET_GetMinmaxParameteriv(exec, _mesa_GetMinmaxParameteriv); - SET_GetSeparableFilter(exec, _mesa_GetSeparableFilter); - SET_Histogram(exec, _mesa_Histogram); - SET_Minmax(exec, _mesa_Minmax); - SET_ResetHistogram(exec, _mesa_ResetHistogram); - SET_ResetMinmax(exec, _mesa_ResetMinmax); - SET_SeparableFilter2D(exec, _mesa_SeparableFilter2D); -#endif - - /* OpenGL 2.0 */ - SET_StencilFuncSeparate(exec, _mesa_StencilFuncSeparate); - SET_StencilMaskSeparate(exec, _mesa_StencilMaskSeparate); - SET_StencilOpSeparate(exec, _mesa_StencilOpSeparate); -#if FEATURE_ARB_shader_objects - SET_AttachShader(exec, _mesa_AttachShader); - SET_CreateProgram(exec, _mesa_CreateProgram); - SET_CreateShader(exec, _mesa_CreateShader); - SET_DeleteProgram(exec, _mesa_DeleteProgram); - SET_DeleteShader(exec, _mesa_DeleteShader); - SET_DetachShader(exec, _mesa_DetachShader); - SET_GetAttachedShaders(exec, _mesa_GetAttachedShaders); - SET_GetProgramiv(exec, _mesa_GetProgramiv); - SET_GetProgramInfoLog(exec, _mesa_GetProgramInfoLog); - SET_GetShaderiv(exec, _mesa_GetShaderiv); - SET_GetShaderInfoLog(exec, _mesa_GetShaderInfoLog); - SET_IsProgram(exec, _mesa_IsProgram); - SET_IsShader(exec, _mesa_IsShader); -#endif - - /* OpenGL 2.1 */ -#if FEATURE_ARB_shader_objects - SET_UniformMatrix2x3fv(exec, _mesa_UniformMatrix2x3fv); - SET_UniformMatrix3x2fv(exec, _mesa_UniformMatrix3x2fv); - SET_UniformMatrix2x4fv(exec, _mesa_UniformMatrix2x4fv); - SET_UniformMatrix4x2fv(exec, _mesa_UniformMatrix4x2fv); - SET_UniformMatrix3x4fv(exec, _mesa_UniformMatrix3x4fv); - SET_UniformMatrix4x3fv(exec, _mesa_UniformMatrix4x3fv); -#endif - - - /* 2. GL_EXT_blend_color */ -#if 0 -/* SET_BlendColorEXT(exec, _mesa_BlendColorEXT); */ -#endif - - /* 3. GL_EXT_polygon_offset */ -#if _HAVE_FULL_GL - SET_PolygonOffsetEXT(exec, _mesa_PolygonOffsetEXT); -#endif - - /* 6. GL_EXT_texture3d */ -#if 0 -/* SET_CopyTexSubImage3DEXT(exec, _mesa_CopyTexSubImage3D); */ -/* SET_TexImage3DEXT(exec, _mesa_TexImage3DEXT); */ -/* SET_TexSubImage3DEXT(exec, _mesa_TexSubImage3D); */ -#endif - - /* 11. GL_EXT_histogram */ -#if 0 - SET_GetHistogramEXT(exec, _mesa_GetHistogram); - SET_GetHistogramParameterfvEXT(exec, _mesa_GetHistogramParameterfv); - SET_GetHistogramParameterivEXT(exec, _mesa_GetHistogramParameteriv); - SET_GetMinmaxEXT(exec, _mesa_GetMinmax); - SET_GetMinmaxParameterfvEXT(exec, _mesa_GetMinmaxParameterfv); - SET_GetMinmaxParameterivEXT(exec, _mesa_GetMinmaxParameteriv); -#endif - - /* 14. SGI_color_table */ -#if 0 - SET_ColorTableSGI(exec, _mesa_ColorTable); - SET_ColorSubTableSGI(exec, _mesa_ColorSubTable); - SET_GetColorTableSGI(exec, _mesa_GetColorTable); - SET_GetColorTableParameterfvSGI(exec, _mesa_GetColorTableParameterfv); - SET_GetColorTableParameterivSGI(exec, _mesa_GetColorTableParameteriv); -#endif - - /* 30. GL_EXT_vertex_array */ -#if _HAVE_FULL_GL - SET_ColorPointerEXT(exec, _mesa_ColorPointerEXT); - SET_EdgeFlagPointerEXT(exec, _mesa_EdgeFlagPointerEXT); - SET_IndexPointerEXT(exec, _mesa_IndexPointerEXT); - SET_NormalPointerEXT(exec, _mesa_NormalPointerEXT); - SET_TexCoordPointerEXT(exec, _mesa_TexCoordPointerEXT); - SET_VertexPointerEXT(exec, _mesa_VertexPointerEXT); -#endif - - /* 37. GL_EXT_blend_minmax */ -#if 0 - SET_BlendEquationEXT(exec, _mesa_BlendEquationEXT); -#endif - - /* 54. GL_EXT_point_parameters */ -#if _HAVE_FULL_GL - SET_PointParameterfEXT(exec, _mesa_PointParameterf); - SET_PointParameterfvEXT(exec, _mesa_PointParameterfv); -#endif - - /* 97. GL_EXT_compiled_vertex_array */ -#if _HAVE_FULL_GL - SET_LockArraysEXT(exec, _mesa_LockArraysEXT); - SET_UnlockArraysEXT(exec, _mesa_UnlockArraysEXT); -#endif - - /* 148. GL_EXT_multi_draw_arrays */ -#if _HAVE_FULL_GL - SET_MultiDrawArraysEXT(exec, _mesa_MultiDrawArraysEXT); - SET_MultiDrawElementsEXT(exec, _mesa_MultiDrawElementsEXT); -#endif - - /* 173. GL_INGR_blend_func_separate */ -#if _HAVE_FULL_GL - SET_BlendFuncSeparateEXT(exec, _mesa_BlendFuncSeparateEXT); -#endif - - /* 196. GL_MESA_resize_buffers */ -#if _HAVE_FULL_GL - SET_ResizeBuffersMESA(exec, _mesa_ResizeBuffersMESA); -#endif - - /* 197. GL_MESA_window_pos */ -#if _HAVE_FULL_GL - SET_WindowPos2dMESA(exec, _mesa_WindowPos2dMESA); - SET_WindowPos2dvMESA(exec, _mesa_WindowPos2dvMESA); - SET_WindowPos2fMESA(exec, _mesa_WindowPos2fMESA); - SET_WindowPos2fvMESA(exec, _mesa_WindowPos2fvMESA); - SET_WindowPos2iMESA(exec, _mesa_WindowPos2iMESA); - SET_WindowPos2ivMESA(exec, _mesa_WindowPos2ivMESA); - SET_WindowPos2sMESA(exec, _mesa_WindowPos2sMESA); - SET_WindowPos2svMESA(exec, _mesa_WindowPos2svMESA); - SET_WindowPos3dMESA(exec, _mesa_WindowPos3dMESA); - SET_WindowPos3dvMESA(exec, _mesa_WindowPos3dvMESA); - SET_WindowPos3fMESA(exec, _mesa_WindowPos3fMESA); - SET_WindowPos3fvMESA(exec, _mesa_WindowPos3fvMESA); - SET_WindowPos3iMESA(exec, _mesa_WindowPos3iMESA); - SET_WindowPos3ivMESA(exec, _mesa_WindowPos3ivMESA); - SET_WindowPos3sMESA(exec, _mesa_WindowPos3sMESA); - SET_WindowPos3svMESA(exec, _mesa_WindowPos3svMESA); - SET_WindowPos4dMESA(exec, _mesa_WindowPos4dMESA); - SET_WindowPos4dvMESA(exec, _mesa_WindowPos4dvMESA); - SET_WindowPos4fMESA(exec, _mesa_WindowPos4fMESA); - SET_WindowPos4fvMESA(exec, _mesa_WindowPos4fvMESA); - SET_WindowPos4iMESA(exec, _mesa_WindowPos4iMESA); - SET_WindowPos4ivMESA(exec, _mesa_WindowPos4ivMESA); - SET_WindowPos4sMESA(exec, _mesa_WindowPos4sMESA); - SET_WindowPos4svMESA(exec, _mesa_WindowPos4svMESA); -#endif - - /* 200. GL_IBM_multimode_draw_arrays */ -#if _HAVE_FULL_GL - SET_MultiModeDrawArraysIBM(exec, _mesa_MultiModeDrawArraysIBM); - SET_MultiModeDrawElementsIBM(exec, _mesa_MultiModeDrawElementsIBM); -#endif - - /* 233. GL_NV_vertex_program */ -#if FEATURE_NV_vertex_program - SET_BindProgramNV(exec, _mesa_BindProgram); - SET_DeleteProgramsNV(exec, _mesa_DeletePrograms); - SET_ExecuteProgramNV(exec, _mesa_ExecuteProgramNV); - SET_GenProgramsNV(exec, _mesa_GenPrograms); - SET_AreProgramsResidentNV(exec, _mesa_AreProgramsResidentNV); - SET_RequestResidentProgramsNV(exec, _mesa_RequestResidentProgramsNV); - SET_GetProgramParameterfvNV(exec, _mesa_GetProgramParameterfvNV); - SET_GetProgramParameterdvNV(exec, _mesa_GetProgramParameterdvNV); - SET_GetProgramivNV(exec, _mesa_GetProgramivNV); - SET_GetProgramStringNV(exec, _mesa_GetProgramStringNV); - SET_GetTrackMatrixivNV(exec, _mesa_GetTrackMatrixivNV); - SET_GetVertexAttribdvNV(exec, _mesa_GetVertexAttribdvNV); - SET_GetVertexAttribfvNV(exec, _mesa_GetVertexAttribfvNV); - SET_GetVertexAttribivNV(exec, _mesa_GetVertexAttribivNV); - SET_GetVertexAttribPointervNV(exec, _mesa_GetVertexAttribPointervNV); - SET_IsProgramNV(exec, _mesa_IsProgramARB); - SET_LoadProgramNV(exec, _mesa_LoadProgramNV); - SET_ProgramEnvParameter4dARB(exec, _mesa_ProgramEnvParameter4dARB); /* alias to ProgramParameter4dNV */ - SET_ProgramEnvParameter4dvARB(exec, _mesa_ProgramEnvParameter4dvARB); /* alias to ProgramParameter4dvNV */ - SET_ProgramEnvParameter4fARB(exec, _mesa_ProgramEnvParameter4fARB); /* alias to ProgramParameter4fNV */ - SET_ProgramEnvParameter4fvARB(exec, _mesa_ProgramEnvParameter4fvARB); /* alias to ProgramParameter4fvNV */ - SET_ProgramParameters4dvNV(exec, _mesa_ProgramParameters4dvNV); - SET_ProgramParameters4fvNV(exec, _mesa_ProgramParameters4fvNV); - SET_TrackMatrixNV(exec, _mesa_TrackMatrixNV); - SET_VertexAttribPointerNV(exec, _mesa_VertexAttribPointerNV); - /* glVertexAttrib*NV functions handled in api_loopback.c */ -#endif - - /* 273. GL_APPLE_vertex_array_object */ - SET_BindVertexArrayAPPLE(exec, _mesa_BindVertexArrayAPPLE); - SET_DeleteVertexArraysAPPLE(exec, _mesa_DeleteVertexArraysAPPLE); - SET_GenVertexArraysAPPLE(exec, _mesa_GenVertexArraysAPPLE); - SET_IsVertexArrayAPPLE(exec, _mesa_IsVertexArrayAPPLE); - - /* 282. GL_NV_fragment_program */ -#if FEATURE_NV_fragment_program - SET_ProgramNamedParameter4fNV(exec, _mesa_ProgramNamedParameter4fNV); - SET_ProgramNamedParameter4dNV(exec, _mesa_ProgramNamedParameter4dNV); - SET_ProgramNamedParameter4fvNV(exec, _mesa_ProgramNamedParameter4fvNV); - SET_ProgramNamedParameter4dvNV(exec, _mesa_ProgramNamedParameter4dvNV); - SET_GetProgramNamedParameterfvNV(exec, _mesa_GetProgramNamedParameterfvNV); - SET_GetProgramNamedParameterdvNV(exec, _mesa_GetProgramNamedParameterdvNV); - SET_ProgramLocalParameter4dARB(exec, _mesa_ProgramLocalParameter4dARB); - SET_ProgramLocalParameter4dvARB(exec, _mesa_ProgramLocalParameter4dvARB); - SET_ProgramLocalParameter4fARB(exec, _mesa_ProgramLocalParameter4fARB); - SET_ProgramLocalParameter4fvARB(exec, _mesa_ProgramLocalParameter4fvARB); - SET_GetProgramLocalParameterdvARB(exec, _mesa_GetProgramLocalParameterdvARB); - SET_GetProgramLocalParameterfvARB(exec, _mesa_GetProgramLocalParameterfvARB); -#endif - - /* 262. GL_NV_point_sprite */ -#if _HAVE_FULL_GL - SET_PointParameteriNV(exec, _mesa_PointParameteri); - SET_PointParameterivNV(exec, _mesa_PointParameteriv); -#endif - - /* 268. GL_EXT_stencil_two_side */ -#if _HAVE_FULL_GL - SET_ActiveStencilFaceEXT(exec, _mesa_ActiveStencilFaceEXT); -#endif - - /* ???. GL_EXT_depth_bounds_test */ - SET_DepthBoundsEXT(exec, _mesa_DepthBoundsEXT); - - /* ARB 1. GL_ARB_multitexture */ -#if _HAVE_FULL_GL - SET_ActiveTextureARB(exec, _mesa_ActiveTextureARB); - SET_ClientActiveTextureARB(exec, _mesa_ClientActiveTextureARB); -#endif - - /* ARB 3. GL_ARB_transpose_matrix */ -#if _HAVE_FULL_GL - SET_LoadTransposeMatrixdARB(exec, _mesa_LoadTransposeMatrixdARB); - SET_LoadTransposeMatrixfARB(exec, _mesa_LoadTransposeMatrixfARB); - SET_MultTransposeMatrixdARB(exec, _mesa_MultTransposeMatrixdARB); - SET_MultTransposeMatrixfARB(exec, _mesa_MultTransposeMatrixfARB); -#endif - - /* ARB 5. GL_ARB_multisample */ -#if _HAVE_FULL_GL - SET_SampleCoverageARB(exec, _mesa_SampleCoverageARB); -#endif - - /* ARB 12. GL_ARB_texture_compression */ -#if _HAVE_FULL_GL - SET_CompressedTexImage3DARB(exec, _mesa_CompressedTexImage3DARB); - SET_CompressedTexImage2DARB(exec, _mesa_CompressedTexImage2DARB); - SET_CompressedTexImage1DARB(exec, _mesa_CompressedTexImage1DARB); - SET_CompressedTexSubImage3DARB(exec, _mesa_CompressedTexSubImage3DARB); - SET_CompressedTexSubImage2DARB(exec, _mesa_CompressedTexSubImage2DARB); - SET_CompressedTexSubImage1DARB(exec, _mesa_CompressedTexSubImage1DARB); - SET_GetCompressedTexImageARB(exec, _mesa_GetCompressedTexImageARB); -#endif - - /* ARB 14. GL_ARB_point_parameters */ - /* reuse EXT_point_parameters functions */ - - /* ARB 26. GL_ARB_vertex_program */ - /* ARB 27. GL_ARB_fragment_program */ -#if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program - /* glVertexAttrib1sARB aliases glVertexAttrib1sNV */ - /* glVertexAttrib1fARB aliases glVertexAttrib1fNV */ - /* glVertexAttrib1dARB aliases glVertexAttrib1dNV */ - /* glVertexAttrib2sARB aliases glVertexAttrib2sNV */ - /* glVertexAttrib2fARB aliases glVertexAttrib2fNV */ - /* glVertexAttrib2dARB aliases glVertexAttrib2dNV */ - /* glVertexAttrib3sARB aliases glVertexAttrib3sNV */ - /* glVertexAttrib3fARB aliases glVertexAttrib3fNV */ - /* glVertexAttrib3dARB aliases glVertexAttrib3dNV */ - /* glVertexAttrib4sARB aliases glVertexAttrib4sNV */ - /* glVertexAttrib4fARB aliases glVertexAttrib4fNV */ - /* glVertexAttrib4dARB aliases glVertexAttrib4dNV */ - /* glVertexAttrib4NubARB aliases glVertexAttrib4NubNV */ - /* glVertexAttrib1svARB aliases glVertexAttrib1svNV */ - /* glVertexAttrib1fvARB aliases glVertexAttrib1fvNV */ - /* glVertexAttrib1dvARB aliases glVertexAttrib1dvNV */ - /* glVertexAttrib2svARB aliases glVertexAttrib2svNV */ - /* glVertexAttrib2fvARB aliases glVertexAttrib2fvNV */ - /* glVertexAttrib2dvARB aliases glVertexAttrib2dvNV */ - /* glVertexAttrib3svARB aliases glVertexAttrib3svNV */ - /* glVertexAttrib3fvARB aliases glVertexAttrib3fvNV */ - /* glVertexAttrib3dvARB aliases glVertexAttrib3dvNV */ - /* glVertexAttrib4svARB aliases glVertexAttrib4svNV */ - /* glVertexAttrib4fvARB aliases glVertexAttrib4fvNV */ - /* glVertexAttrib4dvARB aliases glVertexAttrib4dvNV */ - /* glVertexAttrib4NubvARB aliases glVertexAttrib4NubvNV */ - /* glVertexAttrib4bvARB handled in api_loopback.c */ - /* glVertexAttrib4ivARB handled in api_loopback.c */ - /* glVertexAttrib4ubvARB handled in api_loopback.c */ - /* glVertexAttrib4usvARB handled in api_loopback.c */ - /* glVertexAttrib4uivARB handled in api_loopback.c */ - /* glVertexAttrib4NbvARB handled in api_loopback.c */ - /* glVertexAttrib4NsvARB handled in api_loopback.c */ - /* glVertexAttrib4NivARB handled in api_loopback.c */ - /* glVertexAttrib4NusvARB handled in api_loopback.c */ - /* glVertexAttrib4NuivARB handled in api_loopback.c */ - SET_VertexAttribPointerARB(exec, _mesa_VertexAttribPointerARB); - SET_EnableVertexAttribArrayARB(exec, _mesa_EnableVertexAttribArrayARB); - SET_DisableVertexAttribArrayARB(exec, _mesa_DisableVertexAttribArrayARB); - SET_ProgramStringARB(exec, _mesa_ProgramStringARB); - /* glBindProgramARB aliases glBindProgramNV */ - /* glDeleteProgramsARB aliases glDeleteProgramsNV */ - /* glGenProgramsARB aliases glGenProgramsNV */ - /* glIsProgramARB aliases glIsProgramNV */ - SET_GetVertexAttribdvARB(exec, _mesa_GetVertexAttribdvARB); - SET_GetVertexAttribfvARB(exec, _mesa_GetVertexAttribfvARB); - SET_GetVertexAttribivARB(exec, _mesa_GetVertexAttribivARB); - /* glGetVertexAttribPointervARB aliases glGetVertexAttribPointervNV */ - SET_ProgramEnvParameter4dARB(exec, _mesa_ProgramEnvParameter4dARB); - SET_ProgramEnvParameter4dvARB(exec, _mesa_ProgramEnvParameter4dvARB); - SET_ProgramEnvParameter4fARB(exec, _mesa_ProgramEnvParameter4fARB); - SET_ProgramEnvParameter4fvARB(exec, _mesa_ProgramEnvParameter4fvARB); - SET_ProgramLocalParameter4dARB(exec, _mesa_ProgramLocalParameter4dARB); - SET_ProgramLocalParameter4dvARB(exec, _mesa_ProgramLocalParameter4dvARB); - SET_ProgramLocalParameter4fARB(exec, _mesa_ProgramLocalParameter4fARB); - SET_ProgramLocalParameter4fvARB(exec, _mesa_ProgramLocalParameter4fvARB); - SET_GetProgramEnvParameterdvARB(exec, _mesa_GetProgramEnvParameterdvARB); - SET_GetProgramEnvParameterfvARB(exec, _mesa_GetProgramEnvParameterfvARB); - SET_GetProgramLocalParameterdvARB(exec, _mesa_GetProgramLocalParameterdvARB); - SET_GetProgramLocalParameterfvARB(exec, _mesa_GetProgramLocalParameterfvARB); - SET_GetProgramivARB(exec, _mesa_GetProgramivARB); - SET_GetProgramStringARB(exec, _mesa_GetProgramStringARB); -#endif - - /* ARB 28. GL_ARB_vertex_buffer_object */ -#if FEATURE_ARB_vertex_buffer_object - SET_BindBufferARB(exec, _mesa_BindBufferARB); - SET_BufferDataARB(exec, _mesa_BufferDataARB); - SET_BufferSubDataARB(exec, _mesa_BufferSubDataARB); - SET_DeleteBuffersARB(exec, _mesa_DeleteBuffersARB); - SET_GenBuffersARB(exec, _mesa_GenBuffersARB); - SET_GetBufferParameterivARB(exec, _mesa_GetBufferParameterivARB); - SET_GetBufferPointervARB(exec, _mesa_GetBufferPointervARB); - SET_GetBufferSubDataARB(exec, _mesa_GetBufferSubDataARB); - SET_IsBufferARB(exec, _mesa_IsBufferARB); - SET_MapBufferARB(exec, _mesa_MapBufferARB); - SET_UnmapBufferARB(exec, _mesa_UnmapBufferARB); -#endif - - /* ARB 29. GL_ARB_occlusion_query */ -#if FEATURE_ARB_occlusion_query - SET_GenQueriesARB(exec, _mesa_GenQueriesARB); - SET_DeleteQueriesARB(exec, _mesa_DeleteQueriesARB); - SET_IsQueryARB(exec, _mesa_IsQueryARB); - SET_BeginQueryARB(exec, _mesa_BeginQueryARB); - SET_EndQueryARB(exec, _mesa_EndQueryARB); - SET_GetQueryivARB(exec, _mesa_GetQueryivARB); - SET_GetQueryObjectivARB(exec, _mesa_GetQueryObjectivARB); - SET_GetQueryObjectuivARB(exec, _mesa_GetQueryObjectuivARB); -#endif - - /* ARB 37. GL_ARB_draw_buffers */ - SET_DrawBuffersARB(exec, _mesa_DrawBuffersARB); - -#if FEATURE_ARB_shader_objects - SET_DeleteObjectARB(exec, _mesa_DeleteObjectARB); - SET_GetHandleARB(exec, _mesa_GetHandleARB); - SET_DetachObjectARB(exec, _mesa_DetachObjectARB); - SET_CreateShaderObjectARB(exec, _mesa_CreateShaderObjectARB); - SET_ShaderSourceARB(exec, _mesa_ShaderSourceARB); - SET_CompileShaderARB(exec, _mesa_CompileShaderARB); - SET_CreateProgramObjectARB(exec, _mesa_CreateProgramObjectARB); - SET_AttachObjectARB(exec, _mesa_AttachObjectARB); - SET_LinkProgramARB(exec, _mesa_LinkProgramARB); - SET_UseProgramObjectARB(exec, _mesa_UseProgramObjectARB); - SET_ValidateProgramARB(exec, _mesa_ValidateProgramARB); - SET_Uniform1fARB(exec, _mesa_Uniform1fARB); - SET_Uniform2fARB(exec, _mesa_Uniform2fARB); - SET_Uniform3fARB(exec, _mesa_Uniform3fARB); - SET_Uniform4fARB(exec, _mesa_Uniform4fARB); - SET_Uniform1iARB(exec, _mesa_Uniform1iARB); - SET_Uniform2iARB(exec, _mesa_Uniform2iARB); - SET_Uniform3iARB(exec, _mesa_Uniform3iARB); - SET_Uniform4iARB(exec, _mesa_Uniform4iARB); - SET_Uniform1fvARB(exec, _mesa_Uniform1fvARB); - SET_Uniform2fvARB(exec, _mesa_Uniform2fvARB); - SET_Uniform3fvARB(exec, _mesa_Uniform3fvARB); - SET_Uniform4fvARB(exec, _mesa_Uniform4fvARB); - SET_Uniform1ivARB(exec, _mesa_Uniform1ivARB); - SET_Uniform2ivARB(exec, _mesa_Uniform2ivARB); - SET_Uniform3ivARB(exec, _mesa_Uniform3ivARB); - SET_Uniform4ivARB(exec, _mesa_Uniform4ivARB); - SET_UniformMatrix2fvARB(exec, _mesa_UniformMatrix2fvARB); - SET_UniformMatrix3fvARB(exec, _mesa_UniformMatrix3fvARB); - SET_UniformMatrix4fvARB(exec, _mesa_UniformMatrix4fvARB); - SET_GetObjectParameterfvARB(exec, _mesa_GetObjectParameterfvARB); - SET_GetObjectParameterivARB(exec, _mesa_GetObjectParameterivARB); - SET_GetInfoLogARB(exec, _mesa_GetInfoLogARB); - SET_GetAttachedObjectsARB(exec, _mesa_GetAttachedObjectsARB); - SET_GetUniformLocationARB(exec, _mesa_GetUniformLocationARB); - SET_GetActiveUniformARB(exec, _mesa_GetActiveUniformARB); - SET_GetUniformfvARB(exec, _mesa_GetUniformfvARB); - SET_GetUniformivARB(exec, _mesa_GetUniformivARB); - SET_GetShaderSourceARB(exec, _mesa_GetShaderSourceARB); -#endif /* FEATURE_ARB_shader_objects */ - -#if FEATURE_ARB_vertex_shader - SET_BindAttribLocationARB(exec, _mesa_BindAttribLocationARB); - SET_GetActiveAttribARB(exec, _mesa_GetActiveAttribARB); - SET_GetAttribLocationARB(exec, _mesa_GetAttribLocationARB); -#endif /* FEATURE_ARB_vertex_shader */ - - /* GL_ATI_fragment_shader */ -#if FEATURE_ATI_fragment_shader - SET_GenFragmentShadersATI(exec, _mesa_GenFragmentShadersATI); - SET_BindFragmentShaderATI(exec, _mesa_BindFragmentShaderATI); - SET_DeleteFragmentShaderATI(exec, _mesa_DeleteFragmentShaderATI); - SET_BeginFragmentShaderATI(exec, _mesa_BeginFragmentShaderATI); - SET_EndFragmentShaderATI(exec, _mesa_EndFragmentShaderATI); - SET_PassTexCoordATI(exec, _mesa_PassTexCoordATI); - SET_SampleMapATI(exec, _mesa_SampleMapATI); - SET_ColorFragmentOp1ATI(exec, _mesa_ColorFragmentOp1ATI); - SET_ColorFragmentOp2ATI(exec, _mesa_ColorFragmentOp2ATI); - SET_ColorFragmentOp3ATI(exec, _mesa_ColorFragmentOp3ATI); - SET_AlphaFragmentOp1ATI(exec, _mesa_AlphaFragmentOp1ATI); - SET_AlphaFragmentOp2ATI(exec, _mesa_AlphaFragmentOp2ATI); - SET_AlphaFragmentOp3ATI(exec, _mesa_AlphaFragmentOp3ATI); - SET_SetFragmentShaderConstantATI(exec, _mesa_SetFragmentShaderConstantATI); -#endif - -#if FEATURE_EXT_framebuffer_object - SET_IsRenderbufferEXT(exec, _mesa_IsRenderbufferEXT); - SET_BindRenderbufferEXT(exec, _mesa_BindRenderbufferEXT); - SET_DeleteRenderbuffersEXT(exec, _mesa_DeleteRenderbuffersEXT); - SET_GenRenderbuffersEXT(exec, _mesa_GenRenderbuffersEXT); - SET_RenderbufferStorageEXT(exec, _mesa_RenderbufferStorageEXT); - SET_GetRenderbufferParameterivEXT(exec, _mesa_GetRenderbufferParameterivEXT); - SET_IsFramebufferEXT(exec, _mesa_IsFramebufferEXT); - SET_BindFramebufferEXT(exec, _mesa_BindFramebufferEXT); - SET_DeleteFramebuffersEXT(exec, _mesa_DeleteFramebuffersEXT); - SET_GenFramebuffersEXT(exec, _mesa_GenFramebuffersEXT); - SET_CheckFramebufferStatusEXT(exec, _mesa_CheckFramebufferStatusEXT); - SET_FramebufferTexture1DEXT(exec, _mesa_FramebufferTexture1DEXT); - SET_FramebufferTexture2DEXT(exec, _mesa_FramebufferTexture2DEXT); - SET_FramebufferTexture3DEXT(exec, _mesa_FramebufferTexture3DEXT); - SET_FramebufferRenderbufferEXT(exec, _mesa_FramebufferRenderbufferEXT); - SET_GetFramebufferAttachmentParameterivEXT(exec, _mesa_GetFramebufferAttachmentParameterivEXT); - SET_GenerateMipmapEXT(exec, _mesa_GenerateMipmapEXT); -#endif - -#if FEATURE_EXT_timer_query - SET_GetQueryObjecti64vEXT(exec, _mesa_GetQueryObjecti64vEXT); - SET_GetQueryObjectui64vEXT(exec, _mesa_GetQueryObjectui64vEXT); -#endif - -#if FEATURE_EXT_framebuffer_blit - SET_BlitFramebufferEXT(exec, _mesa_BlitFramebufferEXT); -#endif - - /* GL_EXT_gpu_program_parameters */ -#if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program - SET_ProgramEnvParameters4fvEXT(exec, _mesa_ProgramEnvParameters4fvEXT); - SET_ProgramLocalParameters4fvEXT(exec, _mesa_ProgramLocalParameters4fvEXT); -#endif - - /* GL_MESA_texture_array / GL_EXT_texture_array */ -#if FEATURE_EXT_framebuffer_object - SET_FramebufferTextureLayerEXT(exec, _mesa_FramebufferTextureLayerEXT); -#endif - - /* GL_ATI_separate_stencil */ - SET_StencilFuncSeparateATI(exec, _mesa_StencilFuncSeparateATI); -} - - - -/**********************************************************************/ -/** \name State update logic */ -/*@{*/ /** @@ -1262,9 +479,3 @@ _mesa_update_state( GLcontext *ctx ) _mesa_update_state_locked(ctx); _mesa_unlock_context_textures(ctx); } - - - -/*@}*/ - - diff --git a/src/mesa/main/state.h b/src/mesa/main/state.h index 5240d4bf93..bb7cb8f32a 100644 --- a/src/mesa/main/state.h +++ b/src/mesa/main/state.h @@ -1,13 +1,8 @@ -/** - * \file state.h - * State management. - */ - /* * Mesa 3-D graphics library - * Version: 6.3 + * Version: 7.1 * - * Copyright (C) 1999-2004 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -33,9 +28,6 @@ #include "mtypes.h" -extern void -_mesa_init_exec_table(struct _glapi_table *exec); - extern void _mesa_update_state( GLcontext *ctx ); diff --git a/src/mesa/sources b/src/mesa/sources index c94f260121..21386ef4db 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -2,6 +2,7 @@ MAIN_SOURCES = \ main/api_arrayelt.c \ + main/api_exec.c \ main/api_loopback.c \ main/api_noop.c \ main/api_validate.c \ -- cgit v1.2.3 From 4be7296bfcba22a849f949d105ea385e6964cc25 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 14:55:24 -0600 Subject: mesa: refactor: move scissor functions into new scissor.c file --- src/mesa/main/api_exec.c | 1 + src/mesa/main/attrib.c | 1 + src/mesa/main/buffers.c | 75 ------------------------------------ src/mesa/main/buffers.h | 10 ----- src/mesa/main/context.c | 1 + src/mesa/main/scissor.c | 99 ++++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/scissor.h | 46 ++++++++++++++++++++++ src/mesa/main/sources | 2 + src/mesa/sources | 1 + 9 files changed, 151 insertions(+), 85 deletions(-) create mode 100644 src/mesa/main/scissor.c create mode 100644 src/mesa/main/scissor.h (limited to 'src') diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c index e0f081bdbc..a9696a6842 100644 --- a/src/mesa/main/api_exec.c +++ b/src/mesa/main/api_exec.c @@ -79,6 +79,7 @@ #endif #include "rastpos.h" #include "readpix.h" +#include "scissor.h" #include "state.h" #include "stencil.h" #include "teximage.h" diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index 39666d175a..70b5121a9f 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -43,6 +43,7 @@ #include "matrix.h" #include "points.h" #include "polygon.h" +#include "scissor.h" #include "simple_list.h" #include "stencil.h" #include "texobj.h" diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index 3cbd671bab..5424f442fe 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -699,79 +699,6 @@ _mesa_SampleCoverageARB(GLclampf value, GLboolean invert) -/** - * Define the scissor box. - * - * \param x, y coordinates of the scissor box lower-left corner. - * \param width width of the scissor box. - * \param height height of the scissor box. - * - * \sa glScissor(). - * - * Verifies the parameters and updates __GLcontextRec::Scissor. On a - * change flushes the vertices and notifies the driver via - * the dd_function_table::Scissor callback. - */ -void -_mesa_set_scissor(GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height) -{ - if (x == ctx->Scissor.X && - y == ctx->Scissor.Y && - width == ctx->Scissor.Width && - height == ctx->Scissor.Height) - return; - - FLUSH_VERTICES(ctx, _NEW_SCISSOR); - ctx->Scissor.X = x; - ctx->Scissor.Y = y; - ctx->Scissor.Width = width; - ctx->Scissor.Height = height; - - if (ctx->Driver.Scissor) - ctx->Driver.Scissor( ctx, x, y, width, height ); -} - - -void GLAPIENTRY -_mesa_Scissor( GLint x, GLint y, GLsizei width, GLsizei height ) -{ - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (width < 0 || height < 0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glScissor" ); - return; - } - - if (MESA_VERBOSE & VERBOSE_API) - _mesa_debug(ctx, "glScissor %d %d %d %d\n", x, y, width, height); - - _mesa_set_scissor(ctx, x, y, width, height); -} - - - -/**********************************************************************/ -/** \name Initialization */ -/*@{*/ - -/** - * Initialize the context's scissor state. - * \param ctx the GL context. - */ -void -_mesa_init_scissor(GLcontext *ctx) -{ - /* Scissor group */ - ctx->Scissor.Enabled = GL_FALSE; - ctx->Scissor.X = 0; - ctx->Scissor.Y = 0; - ctx->Scissor.Width = 0; - ctx->Scissor.Height = 0; -} - - /** * Initialize the context's multisample state. * \param ctx the GL context. @@ -786,5 +713,3 @@ _mesa_init_multisample(GLcontext *ctx) ctx->Multisample.SampleCoverageValue = 1.0; ctx->Multisample.SampleCoverageInvert = GL_FALSE; } - -/*@}*/ diff --git a/src/mesa/main/buffers.h b/src/mesa/main/buffers.h index 208e7af2b9..50f2f47a04 100644 --- a/src/mesa/main/buffers.h +++ b/src/mesa/main/buffers.h @@ -65,22 +65,12 @@ _mesa_ReadBuffer( GLenum mode ); extern void GLAPIENTRY _mesa_ResizeBuffersMESA( void ); -extern void GLAPIENTRY -_mesa_Scissor( GLint x, GLint y, GLsizei width, GLsizei height ); - extern void GLAPIENTRY _mesa_SampleCoverageARB(GLclampf value, GLboolean invert); -extern void -_mesa_init_scissor(GLcontext *ctx); - extern void _mesa_init_multisample(GLcontext *ctx); -extern void -_mesa_set_scissor(GLcontext *ctx, - GLint x, GLint y, GLsizei width, GLsizei height); - extern void _mesa_resizebuffers( GLcontext *ctx ); #endif diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 465d6b3f89..7ee6b2ec54 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -111,6 +111,7 @@ #include "polygon.h" #include "queryobj.h" #include "rastpos.h" +#include "scissor.h" #include "simple_list.h" #include "state.h" #include "stencil.h" diff --git a/src/mesa/main/scissor.c b/src/mesa/main/scissor.c new file mode 100644 index 0000000000..b5f4cde789 --- /dev/null +++ b/src/mesa/main/scissor.c @@ -0,0 +1,99 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#include "main/glheader.h" +#include "main/context.h" +#include "main/scissor.h" + + +/** + * Called via glScissor + */ +void GLAPIENTRY +_mesa_Scissor( GLint x, GLint y, GLsizei width, GLsizei height ) +{ + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (width < 0 || height < 0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glScissor" ); + return; + } + + if (MESA_VERBOSE & VERBOSE_API) + _mesa_debug(ctx, "glScissor %d %d %d %d\n", x, y, width, height); + + _mesa_set_scissor(ctx, x, y, width, height); +} + + +/** + * Define the scissor box. + * + * \param x, y coordinates of the scissor box lower-left corner. + * \param width width of the scissor box. + * \param height height of the scissor box. + * + * \sa glScissor(). + * + * Verifies the parameters and updates __GLcontextRec::Scissor. On a + * change flushes the vertices and notifies the driver via + * the dd_function_table::Scissor callback. + */ +void +_mesa_set_scissor(GLcontext *ctx, + GLint x, GLint y, GLsizei width, GLsizei height) +{ + if (x == ctx->Scissor.X && + y == ctx->Scissor.Y && + width == ctx->Scissor.Width && + height == ctx->Scissor.Height) + return; + + FLUSH_VERTICES(ctx, _NEW_SCISSOR); + ctx->Scissor.X = x; + ctx->Scissor.Y = y; + ctx->Scissor.Width = width; + ctx->Scissor.Height = height; + + if (ctx->Driver.Scissor) + ctx->Driver.Scissor( ctx, x, y, width, height ); +} + + +/** + * Initialize the context's scissor state. + * \param ctx the GL context. + */ +void +_mesa_init_scissor(GLcontext *ctx) +{ + /* Scissor group */ + ctx->Scissor.Enabled = GL_FALSE; + ctx->Scissor.X = 0; + ctx->Scissor.Y = 0; + ctx->Scissor.Width = 0; + ctx->Scissor.Height = 0; +} diff --git a/src/mesa/main/scissor.h b/src/mesa/main/scissor.h new file mode 100644 index 0000000000..b852a2122d --- /dev/null +++ b/src/mesa/main/scissor.h @@ -0,0 +1,46 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef SCISSOR_H +#define SCISSOR_H + + +#include "main/mtypes.h" + + +extern void GLAPIENTRY +_mesa_Scissor( GLint x, GLint y, GLsizei width, GLsizei height ); + + +extern void +_mesa_set_scissor(GLcontext *ctx, + GLint x, GLint y, GLsizei width, GLsizei height); + + +extern void +_mesa_init_scissor(GLcontext *ctx); + + +#endif diff --git a/src/mesa/main/sources b/src/mesa/main/sources index 06f32f2275..60b6b5dc3f 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -50,6 +50,7 @@ readpix.c \ rastpos.c \ rbadaptors.c \ renderbuffer.c \ +scissor.c \ state.c \ stencil.c \ texcompress.c \ @@ -127,6 +128,7 @@ rbadaptors.h \ readpix.h \ renderbuffer.h \ simple_list.h \ +scissor.h \ state.h \ stencil.h \ texcompress.h \ diff --git a/src/mesa/sources b/src/mesa/sources index 21386ef4db..f18bbff436 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -52,6 +52,7 @@ MAIN_SOURCES = \ main/rbadaptors.c \ main/readpix.c \ main/renderbuffer.c \ + main/scissor.c \ main/shaders.c \ main/state.c \ main/stencil.c \ -- cgit v1.2.3 From 8cad3189d3346bce3eee1976f1b5cf32da9dd9a7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 14:56:29 -0600 Subject: gallium: include scissor.h --- src/mesa/state_tracker/st_context.c | 1 + src/mesa/state_tracker/st_framebuffer.c | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 4081722bc7..e20e0c08cb 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -30,6 +30,7 @@ #include "main/extensions.h" #include "main/matrix.h" #include "main/buffers.h" +#include "main/scissor.h" #include "vbo/vbo.h" #include "shader/shader_api.h" #include "st_public.h" diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 47466c97d8..7099d78eb8 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -32,6 +32,7 @@ #include "main/framebuffer.h" #include "main/matrix.h" #include "main/renderbuffer.h" +#include "main/scissor.h" #include "st_public.h" #include "st_context.h" #include "st_cb_fbo.h" -- cgit v1.2.3 From eade430682516a445a2bf765165362dad19594f0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 15:01:02 -0600 Subject: mesa: refactor: move glClear, glClearColor into new clear.c file. --- src/mesa/main/api_exec.c | 1 + src/mesa/main/buffers.c | 144 +------------------------------------- src/mesa/main/buffers.h | 10 --- src/mesa/main/clear.c | 179 +++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/clear.h | 44 ++++++++++++ src/mesa/main/sources | 2 + src/mesa/sources | 1 + 7 files changed, 228 insertions(+), 153 deletions(-) create mode 100644 src/mesa/main/clear.c create mode 100644 src/mesa/main/clear.h (limited to 'src') diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c index a9696a6842..382091f580 100644 --- a/src/mesa/main/api_exec.c +++ b/src/mesa/main/api_exec.c @@ -46,6 +46,7 @@ #endif #include "arrayobj.h" #include "buffers.h" +#include "clear.h" #include "clip.h" #include "colortab.h" #include "context.h" diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index 5424f442fe..d617ef9a32 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -25,7 +25,7 @@ /** * \file buffers.c - * General framebuffer-related functions, like glClear, glScissor, etc. + * glReadBuffer, DrawBuffer functions. */ @@ -42,148 +42,6 @@ #define BAD_MASK ~0u -#if _HAVE_FULL_GL -void GLAPIENTRY -_mesa_ClearIndex( GLfloat c ) -{ - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (ctx->Color.ClearIndex == (GLuint) c) - return; - - FLUSH_VERTICES(ctx, _NEW_COLOR); - ctx->Color.ClearIndex = (GLuint) c; - - if (!ctx->Visual.rgbMode && ctx->Driver.ClearIndex) { - /* it's OK to call glClearIndex in RGBA mode but it should be a NOP */ - (*ctx->Driver.ClearIndex)( ctx, ctx->Color.ClearIndex ); - } -} -#endif - - -/** - * Specify the clear values for the color buffers. - * - * \param red red color component. - * \param green green color component. - * \param blue blue color component. - * \param alpha alpha component. - * - * \sa glClearColor(). - * - * Clamps the parameters and updates gl_colorbuffer_attrib::ClearColor. On a - * change, flushes the vertices and notifies the driver via the - * dd_function_table::ClearColor callback. - */ -void GLAPIENTRY -_mesa_ClearColor( GLclampf red, GLclampf green, GLclampf blue, GLclampf alpha ) -{ - GLfloat tmp[4]; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - tmp[0] = CLAMP(red, 0.0F, 1.0F); - tmp[1] = CLAMP(green, 0.0F, 1.0F); - tmp[2] = CLAMP(blue, 0.0F, 1.0F); - tmp[3] = CLAMP(alpha, 0.0F, 1.0F); - - if (TEST_EQ_4V(tmp, ctx->Color.ClearColor)) - return; /* no change */ - - FLUSH_VERTICES(ctx, _NEW_COLOR); - COPY_4V(ctx->Color.ClearColor, tmp); - - if (ctx->Visual.rgbMode && ctx->Driver.ClearColor) { - /* it's OK to call glClearColor in CI mode but it should be a NOP */ - (*ctx->Driver.ClearColor)(ctx, ctx->Color.ClearColor); - } -} - - -/** - * Clear buffers. - * - * \param mask bit-mask indicating the buffers to be cleared. - * - * Flushes the vertices and verifies the parameter. If __GLcontextRec::NewState - * is set then calls _mesa_update_state() to update gl_frame_buffer::_Xmin, - * etc. If the rasterization mode is set to GL_RENDER then requests the driver - * to clear the buffers, via the dd_function_table::Clear callback. - */ -void GLAPIENTRY -_mesa_Clear( GLbitfield mask ) -{ - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); - - FLUSH_CURRENT(ctx, 0); - - if (MESA_VERBOSE & VERBOSE_API) - _mesa_debug(ctx, "glClear 0x%x\n", mask); - - if (mask & ~(GL_COLOR_BUFFER_BIT | - GL_DEPTH_BUFFER_BIT | - GL_STENCIL_BUFFER_BIT | - GL_ACCUM_BUFFER_BIT)) { - /* invalid bit set */ - _mesa_error( ctx, GL_INVALID_VALUE, "glClear(0x%x)", mask); - return; - } - - if (ctx->NewState) { - _mesa_update_state( ctx ); /* update _Xmin, etc */ - } - - if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { - _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT, - "glClear(incomplete framebuffer)"); - return; - } - - if (ctx->DrawBuffer->Width == 0 || ctx->DrawBuffer->Height == 0) - return; - - if (ctx->RenderMode == GL_RENDER) { - GLbitfield bufferMask; - - /* don't clear depth buffer if depth writing disabled */ - if (!ctx->Depth.Mask) - mask &= ~GL_DEPTH_BUFFER_BIT; - - /* Build the bitmask to send to device driver's Clear function. - * Note that the GL_COLOR_BUFFER_BIT flag will expand to 0, 1, 2 or 4 - * of the BUFFER_BIT_FRONT/BACK_LEFT/RIGHT flags, or one of the - * BUFFER_BIT_COLORn flags. - */ - bufferMask = 0; - if (mask & GL_COLOR_BUFFER_BIT) { - bufferMask |= ctx->DrawBuffer->_ColorDrawBufferMask[0]; - } - - if ((mask & GL_DEPTH_BUFFER_BIT) - && ctx->DrawBuffer->Visual.haveDepthBuffer) { - bufferMask |= BUFFER_BIT_DEPTH; - } - - if ((mask & GL_STENCIL_BUFFER_BIT) - && ctx->DrawBuffer->Visual.haveStencilBuffer) { - bufferMask |= BUFFER_BIT_STENCIL; - } - - if ((mask & GL_ACCUM_BUFFER_BIT) - && ctx->DrawBuffer->Visual.haveAccumBuffer) { - bufferMask |= BUFFER_BIT_ACCUM; - } - - ASSERT(ctx->Driver.Clear); - ctx->Driver.Clear(ctx, bufferMask); - } -} - - - /** * Return bitmask of BUFFER_BIT_* flags indicating which color buffers are * available to the rendering context (for drawing or reading). diff --git a/src/mesa/main/buffers.h b/src/mesa/main/buffers.h index 50f2f47a04..ca0e7c411c 100644 --- a/src/mesa/main/buffers.h +++ b/src/mesa/main/buffers.h @@ -36,16 +36,6 @@ #include "mtypes.h" -extern void GLAPIENTRY -_mesa_ClearIndex( GLfloat c ); - -extern void GLAPIENTRY -_mesa_ClearColor( GLclampf red, GLclampf green, - GLclampf blue, GLclampf alpha ); - -extern void GLAPIENTRY -_mesa_Clear( GLbitfield mask ); - extern void GLAPIENTRY _mesa_DrawBuffer( GLenum mode ); diff --git a/src/mesa/main/clear.c b/src/mesa/main/clear.c new file mode 100644 index 0000000000..434685984d --- /dev/null +++ b/src/mesa/main/clear.c @@ -0,0 +1,179 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * \file clear.c + * glClearColor, glClearIndex, glClear() functions. + */ + + + +#include "glheader.h" +#include "clear.h" +#include "context.h" +#include "colormac.h" +#include "state.h" + + + +#if _HAVE_FULL_GL +void GLAPIENTRY +_mesa_ClearIndex( GLfloat c ) +{ + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (ctx->Color.ClearIndex == (GLuint) c) + return; + + FLUSH_VERTICES(ctx, _NEW_COLOR); + ctx->Color.ClearIndex = (GLuint) c; + + if (!ctx->Visual.rgbMode && ctx->Driver.ClearIndex) { + /* it's OK to call glClearIndex in RGBA mode but it should be a NOP */ + (*ctx->Driver.ClearIndex)( ctx, ctx->Color.ClearIndex ); + } +} +#endif + + +/** + * Specify the clear values for the color buffers. + * + * \param red red color component. + * \param green green color component. + * \param blue blue color component. + * \param alpha alpha component. + * + * \sa glClearColor(). + * + * Clamps the parameters and updates gl_colorbuffer_attrib::ClearColor. On a + * change, flushes the vertices and notifies the driver via the + * dd_function_table::ClearColor callback. + */ +void GLAPIENTRY +_mesa_ClearColor( GLclampf red, GLclampf green, GLclampf blue, GLclampf alpha ) +{ + GLfloat tmp[4]; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + tmp[0] = CLAMP(red, 0.0F, 1.0F); + tmp[1] = CLAMP(green, 0.0F, 1.0F); + tmp[2] = CLAMP(blue, 0.0F, 1.0F); + tmp[3] = CLAMP(alpha, 0.0F, 1.0F); + + if (TEST_EQ_4V(tmp, ctx->Color.ClearColor)) + return; /* no change */ + + FLUSH_VERTICES(ctx, _NEW_COLOR); + COPY_4V(ctx->Color.ClearColor, tmp); + + if (ctx->Visual.rgbMode && ctx->Driver.ClearColor) { + /* it's OK to call glClearColor in CI mode but it should be a NOP */ + (*ctx->Driver.ClearColor)(ctx, ctx->Color.ClearColor); + } +} + + +/** + * Clear buffers. + * + * \param mask bit-mask indicating the buffers to be cleared. + * + * Flushes the vertices and verifies the parameter. If __GLcontextRec::NewState + * is set then calls _mesa_update_state() to update gl_frame_buffer::_Xmin, + * etc. If the rasterization mode is set to GL_RENDER then requests the driver + * to clear the buffers, via the dd_function_table::Clear callback. + */ +void GLAPIENTRY +_mesa_Clear( GLbitfield mask ) +{ + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + + FLUSH_CURRENT(ctx, 0); + + if (MESA_VERBOSE & VERBOSE_API) + _mesa_debug(ctx, "glClear 0x%x\n", mask); + + if (mask & ~(GL_COLOR_BUFFER_BIT | + GL_DEPTH_BUFFER_BIT | + GL_STENCIL_BUFFER_BIT | + GL_ACCUM_BUFFER_BIT)) { + /* invalid bit set */ + _mesa_error( ctx, GL_INVALID_VALUE, "glClear(0x%x)", mask); + return; + } + + if (ctx->NewState) { + _mesa_update_state( ctx ); /* update _Xmin, etc */ + } + + if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) { + _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT, + "glClear(incomplete framebuffer)"); + return; + } + + if (ctx->DrawBuffer->Width == 0 || ctx->DrawBuffer->Height == 0) + return; + + if (ctx->RenderMode == GL_RENDER) { + GLbitfield bufferMask; + + /* don't clear depth buffer if depth writing disabled */ + if (!ctx->Depth.Mask) + mask &= ~GL_DEPTH_BUFFER_BIT; + + /* Build the bitmask to send to device driver's Clear function. + * Note that the GL_COLOR_BUFFER_BIT flag will expand to 0, 1, 2 or 4 + * of the BUFFER_BIT_FRONT/BACK_LEFT/RIGHT flags, or one of the + * BUFFER_BIT_COLORn flags. + */ + bufferMask = 0; + if (mask & GL_COLOR_BUFFER_BIT) { + bufferMask |= ctx->DrawBuffer->_ColorDrawBufferMask[0]; + } + + if ((mask & GL_DEPTH_BUFFER_BIT) + && ctx->DrawBuffer->Visual.haveDepthBuffer) { + bufferMask |= BUFFER_BIT_DEPTH; + } + + if ((mask & GL_STENCIL_BUFFER_BIT) + && ctx->DrawBuffer->Visual.haveStencilBuffer) { + bufferMask |= BUFFER_BIT_STENCIL; + } + + if ((mask & GL_ACCUM_BUFFER_BIT) + && ctx->DrawBuffer->Visual.haveAccumBuffer) { + bufferMask |= BUFFER_BIT_ACCUM; + } + + ASSERT(ctx->Driver.Clear); + ctx->Driver.Clear(ctx, bufferMask); + } +} diff --git a/src/mesa/main/clear.h b/src/mesa/main/clear.h new file mode 100644 index 0000000000..9a54ba14bc --- /dev/null +++ b/src/mesa/main/clear.h @@ -0,0 +1,44 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef CLEAR_H +#define CLEAR_H + + +#include "main/mtypes.h" + + +extern void GLAPIENTRY +_mesa_ClearIndex( GLfloat c ); + +extern void GLAPIENTRY +_mesa_ClearColor( GLclampf red, GLclampf green, + GLclampf blue, GLclampf alpha ); + +extern void GLAPIENTRY +_mesa_Clear( GLbitfield mask ); + + +#endif diff --git a/src/mesa/main/sources b/src/mesa/main/sources index 60b6b5dc3f..8043f4369e 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -11,6 +11,7 @@ attrib.c \ blend.c \ bufferobj.c \ buffers.c \ +clear.c \ clip.c \ colortab.c \ context.c \ @@ -84,6 +85,7 @@ bitset.h \ blend.h \ bufferobj.h \ buffers.h \ +clear.h \ clip.h \ colormac.h \ colortab.h \ diff --git a/src/mesa/sources b/src/mesa/sources index f18bbff436..95d665e3a1 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -12,6 +12,7 @@ MAIN_SOURCES = \ main/blend.c \ main/bufferobj.c \ main/buffers.c \ + main/clear.c \ main/clip.c \ main/colortab.c \ main/context.c \ -- cgit v1.2.3 From 9091015a9782ad15e58540a8fd61df83ea2bfe31 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 15:04:31 -0600 Subject: mesa: refactor: move _mesa_resizebuffers(), _mesa_ResizeBuffersMESA() to framebuffer.c --- src/mesa/main/attrib.c | 1 + src/mesa/main/buffers.c | 79 --------------------------------------------- src/mesa/main/buffers.h | 4 --- src/mesa/main/dlist.c | 1 + src/mesa/main/framebuffer.c | 78 ++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/framebuffer.h | 8 +++++ 6 files changed, 88 insertions(+), 83 deletions(-) (limited to 'src') diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index 70b5121a9f..63754569e6 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -30,6 +30,7 @@ #include "blend.h" #include "buffers.h" #include "bufferobj.h" +#include "clear.h" #include "colormac.h" #include "colortab.h" #include "context.h" diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index d617ef9a32..c767e62286 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -456,83 +456,6 @@ _mesa_ReadBuffer(GLenum buffer) } -#if _HAVE_FULL_GL - -/** - * XXX THIS IS OBSOLETE - drivers should take care of detecting window - * size changes and act accordingly, likely calling _mesa_resize_framebuffer(). - * - * GL_MESA_resize_buffers extension. - * - * When this function is called, we'll ask the window system how large - * the current window is. If it's a new size, we'll call the driver's - * ResizeBuffers function. The driver will then resize its color buffers - * as needed, and maybe call the swrast's routine for reallocating - * swrast-managed depth/stencil/accum/etc buffers. - * \note This function should only be called through the GL API, not - * from device drivers (as was done in the past). - */ - -void _mesa_resizebuffers( GLcontext *ctx ) -{ - ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH( ctx ); - - if (MESA_VERBOSE & VERBOSE_API) - _mesa_debug(ctx, "glResizeBuffersMESA\n"); - - if (!ctx->Driver.GetBufferSize) { - return; - } - - if (ctx->WinSysDrawBuffer) { - GLuint newWidth, newHeight; - GLframebuffer *buffer = ctx->WinSysDrawBuffer; - - assert(buffer->Name == 0); - - /* ask device driver for size of output buffer */ - ctx->Driver.GetBufferSize( buffer, &newWidth, &newHeight ); - - /* see if size of device driver's color buffer (window) has changed */ - if (buffer->Width != newWidth || buffer->Height != newHeight) { - if (ctx->Driver.ResizeBuffers) - ctx->Driver.ResizeBuffers(ctx, buffer, newWidth, newHeight ); - } - } - - if (ctx->WinSysReadBuffer - && ctx->WinSysReadBuffer != ctx->WinSysDrawBuffer) { - GLuint newWidth, newHeight; - GLframebuffer *buffer = ctx->WinSysReadBuffer; - - assert(buffer->Name == 0); - - /* ask device driver for size of read buffer */ - ctx->Driver.GetBufferSize( buffer, &newWidth, &newHeight ); - - /* see if size of device driver's color buffer (window) has changed */ - if (buffer->Width != newWidth || buffer->Height != newHeight) { - if (ctx->Driver.ResizeBuffers) - ctx->Driver.ResizeBuffers(ctx, buffer, newWidth, newHeight ); - } - } - - ctx->NewState |= _NEW_BUFFERS; /* to update scissor / window bounds */ -} - - -/* - * XXX THIS IS OBSOLETE - */ -void GLAPIENTRY -_mesa_ResizeBuffersMESA( void ) -{ - GET_CURRENT_CONTEXT(ctx); - - if (ctx->Extensions.MESA_resize_buffers) - _mesa_resizebuffers( ctx ); -} - /* * XXX move somewhere else someday? @@ -553,8 +476,6 @@ _mesa_SampleCoverageARB(GLclampf value, GLboolean invert) ctx->NewState |= _NEW_MULTISAMPLE; } -#endif /* _HAVE_FULL_GL */ - /** diff --git a/src/mesa/main/buffers.h b/src/mesa/main/buffers.h index ca0e7c411c..ba79db6871 100644 --- a/src/mesa/main/buffers.h +++ b/src/mesa/main/buffers.h @@ -52,15 +52,11 @@ _mesa_readbuffer_update_fields(GLcontext *ctx, GLenum buffer); extern void GLAPIENTRY _mesa_ReadBuffer( GLenum mode ); -extern void GLAPIENTRY -_mesa_ResizeBuffersMESA( void ); - extern void GLAPIENTRY _mesa_SampleCoverageARB(GLclampf value, GLboolean invert); extern void _mesa_init_multisample(GLcontext *ctx); -extern void _mesa_resizebuffers( GLcontext *ctx ); #endif diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c index f933580b2d..07d279da30 100644 --- a/src/mesa/main/dlist.c +++ b/src/mesa/main/dlist.c @@ -52,6 +52,7 @@ #include "eval.h" #include "extensions.h" #include "feedback.h" +#include "framebuffer.h" #include "get.h" #include "glapi/glapi.h" #include "hash.h" diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c index 894d99afd2..96f1b30c9b 100644 --- a/src/mesa/main/framebuffer.c +++ b/src/mesa/main/framebuffer.c @@ -338,6 +338,84 @@ _mesa_resize_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb, } + +/** + * XXX THIS IS OBSOLETE - drivers should take care of detecting window + * size changes and act accordingly, likely calling _mesa_resize_framebuffer(). + * + * GL_MESA_resize_buffers extension. + * + * When this function is called, we'll ask the window system how large + * the current window is. If it's a new size, we'll call the driver's + * ResizeBuffers function. The driver will then resize its color buffers + * as needed, and maybe call the swrast's routine for reallocating + * swrast-managed depth/stencil/accum/etc buffers. + * \note This function should only be called through the GL API, not + * from device drivers (as was done in the past). + */ +void +_mesa_resizebuffers( GLcontext *ctx ) +{ + ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH( ctx ); + + if (MESA_VERBOSE & VERBOSE_API) + _mesa_debug(ctx, "glResizeBuffersMESA\n"); + + if (!ctx->Driver.GetBufferSize) { + return; + } + + if (ctx->WinSysDrawBuffer) { + GLuint newWidth, newHeight; + GLframebuffer *buffer = ctx->WinSysDrawBuffer; + + assert(buffer->Name == 0); + + /* ask device driver for size of output buffer */ + ctx->Driver.GetBufferSize( buffer, &newWidth, &newHeight ); + + /* see if size of device driver's color buffer (window) has changed */ + if (buffer->Width != newWidth || buffer->Height != newHeight) { + if (ctx->Driver.ResizeBuffers) + ctx->Driver.ResizeBuffers(ctx, buffer, newWidth, newHeight ); + } + } + + if (ctx->WinSysReadBuffer + && ctx->WinSysReadBuffer != ctx->WinSysDrawBuffer) { + GLuint newWidth, newHeight; + GLframebuffer *buffer = ctx->WinSysReadBuffer; + + assert(buffer->Name == 0); + + /* ask device driver for size of read buffer */ + ctx->Driver.GetBufferSize( buffer, &newWidth, &newHeight ); + + /* see if size of device driver's color buffer (window) has changed */ + if (buffer->Width != newWidth || buffer->Height != newHeight) { + if (ctx->Driver.ResizeBuffers) + ctx->Driver.ResizeBuffers(ctx, buffer, newWidth, newHeight ); + } + } + + ctx->NewState |= _NEW_BUFFERS; /* to update scissor / window bounds */ +} + + +/* + * XXX THIS IS OBSOLETE + */ +void GLAPIENTRY +_mesa_ResizeBuffersMESA( void ) +{ + GET_CURRENT_CONTEXT(ctx); + + if (ctx->Extensions.MESA_resize_buffers) + _mesa_resizebuffers( ctx ); +} + + + /** * Examine all the framebuffer's renderbuffers to update the Width/Height * fields of the framebuffer. If we have renderbuffers with different diff --git a/src/mesa/main/framebuffer.h b/src/mesa/main/framebuffer.h index 4d76f3a90f..e9eeed28cb 100644 --- a/src/mesa/main/framebuffer.h +++ b/src/mesa/main/framebuffer.h @@ -53,6 +53,14 @@ extern void _mesa_resize_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb, GLuint width, GLuint height); + +extern void +_mesa_resizebuffers( GLcontext *ctx ); + +extern void GLAPIENTRY +_mesa_ResizeBuffersMESA( void ); + + extern void _mesa_update_draw_buffer_bounds(GLcontext *ctx); -- cgit v1.2.3 From bce428c4a65fdcb890ea18bf4a1dfb42ed109006 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 15:09:21 -0600 Subject: mesa: refactor: move multisample-related functions into new multisample.c file --- src/mesa/main/api_exec.c | 1 + src/mesa/main/attrib.c | 1 + src/mesa/main/buffers.c | 38 -------------------------- src/mesa/main/buffers.h | 6 ----- src/mesa/main/context.c | 1 + src/mesa/main/multisample.c | 66 +++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/multisample.h | 38 ++++++++++++++++++++++++++ src/mesa/main/sources | 2 ++ src/mesa/sources | 1 + 9 files changed, 110 insertions(+), 44 deletions(-) create mode 100644 src/mesa/main/multisample.c create mode 100644 src/mesa/main/multisample.h (limited to 'src') diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c index 382091f580..8cdbaada9b 100644 --- a/src/mesa/main/api_exec.c +++ b/src/mesa/main/api_exec.c @@ -71,6 +71,7 @@ #include "lines.h" #include "macros.h" #include "matrix.h" +#include "multisample.h" #include "pixel.h" #include "pixelstore.h" #include "points.h" diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index 63754569e6..6ed82da203 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -42,6 +42,7 @@ #include "light.h" #include "lines.h" #include "matrix.h" +#include "multisample.h" #include "points.h" #include "polygon.h" #include "scissor.h" diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c index c767e62286..1d07c68633 100644 --- a/src/mesa/main/buffers.c +++ b/src/mesa/main/buffers.c @@ -454,41 +454,3 @@ _mesa_ReadBuffer(GLenum buffer) if (ctx->Driver.ReadBuffer) (*ctx->Driver.ReadBuffer)(ctx, buffer); } - - - -/* - * XXX move somewhere else someday? - */ -void GLAPIENTRY -_mesa_SampleCoverageARB(GLclampf value, GLboolean invert) -{ - GET_CURRENT_CONTEXT(ctx); - - if (!ctx->Extensions.ARB_multisample) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glSampleCoverageARB"); - return; - } - - ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH( ctx ); - ctx->Multisample.SampleCoverageValue = (GLfloat) CLAMP(value, 0.0, 1.0); - ctx->Multisample.SampleCoverageInvert = invert; - ctx->NewState |= _NEW_MULTISAMPLE; -} - - - -/** - * Initialize the context's multisample state. - * \param ctx the GL context. - */ -void -_mesa_init_multisample(GLcontext *ctx) -{ - ctx->Multisample.Enabled = GL_FALSE; - ctx->Multisample.SampleAlphaToCoverage = GL_FALSE; - ctx->Multisample.SampleAlphaToOne = GL_FALSE; - ctx->Multisample.SampleCoverage = GL_FALSE; - ctx->Multisample.SampleCoverageValue = 1.0; - ctx->Multisample.SampleCoverageInvert = GL_FALSE; -} diff --git a/src/mesa/main/buffers.h b/src/mesa/main/buffers.h index ba79db6871..53d5fb80d4 100644 --- a/src/mesa/main/buffers.h +++ b/src/mesa/main/buffers.h @@ -52,11 +52,5 @@ _mesa_readbuffer_update_fields(GLcontext *ctx, GLenum buffer); extern void GLAPIENTRY _mesa_ReadBuffer( GLenum mode ); -extern void GLAPIENTRY -_mesa_SampleCoverageARB(GLclampf value, GLboolean invert); - -extern void -_mesa_init_multisample(GLcontext *ctx); - #endif diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 7ee6b2ec54..1b357ae6c2 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -105,6 +105,7 @@ #include "lines.h" #include "macros.h" #include "matrix.h" +#include "multisample.h" #include "pixel.h" #include "pixelstore.h" #include "points.h" diff --git a/src/mesa/main/multisample.c b/src/mesa/main/multisample.c new file mode 100644 index 0000000000..e138087436 --- /dev/null +++ b/src/mesa/main/multisample.c @@ -0,0 +1,66 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#include "main/glheader.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/multisample.h" + + +/** + * Called via glSampleCoverageARB + */ +void GLAPIENTRY +_mesa_SampleCoverageARB(GLclampf value, GLboolean invert) +{ + GET_CURRENT_CONTEXT(ctx); + + if (!ctx->Extensions.ARB_multisample) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glSampleCoverageARB"); + return; + } + + ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH( ctx ); + + ctx->Multisample.SampleCoverageValue = (GLfloat) CLAMP(value, 0.0, 1.0); + ctx->Multisample.SampleCoverageInvert = invert; + ctx->NewState |= _NEW_MULTISAMPLE; +} + + +/** + * Initialize the context's multisample state. + * \param ctx the GL context. + */ +void +_mesa_init_multisample(GLcontext *ctx) +{ + ctx->Multisample.Enabled = GL_FALSE; + ctx->Multisample.SampleAlphaToCoverage = GL_FALSE; + ctx->Multisample.SampleAlphaToOne = GL_FALSE; + ctx->Multisample.SampleCoverage = GL_FALSE; + ctx->Multisample.SampleCoverageValue = 1.0; + ctx->Multisample.SampleCoverageInvert = GL_FALSE; +} diff --git a/src/mesa/main/multisample.h b/src/mesa/main/multisample.h new file mode 100644 index 0000000000..4305900cc4 --- /dev/null +++ b/src/mesa/main/multisample.h @@ -0,0 +1,38 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef MULTISAMPLE_H +#define MULTISAMPLE_H + + +extern void GLAPIENTRY +_mesa_SampleCoverageARB(GLclampf value, GLboolean invert); + + +extern void +_mesa_init_multisample(GLcontext *ctx); + + +#endif diff --git a/src/mesa/main/sources b/src/mesa/main/sources index 8043f4369e..5379fa08f4 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -42,6 +42,7 @@ lines.c \ matrix.c \ mipmap.c \ mm.c \ +multisample.c \ occlude.c \ pixel.c \ pixelstore.c \ @@ -120,6 +121,7 @@ matrix.h \ mipmap.h \ mm.h \ mtypes.h \ +multisample.h \ occlude.h \ pixel.h \ pixelstore.h \ diff --git a/src/mesa/sources b/src/mesa/sources index 95d665e3a1..23b03932be 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -44,6 +44,7 @@ MAIN_SOURCES = \ main/matrix.c \ main/mipmap.c \ main/mm.c \ + main/multisample.c \ main/pixel.c \ main/pixelstore.c \ main/points.c \ -- cgit v1.2.3 From eded7f010d344a909cf9c403eb3bdad91804d174 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 15:14:02 -0600 Subject: mesa: refactor: move _mesa_update_minmax/histogram() into image.c --- src/mesa/main/histogram.c | 65 --------------------------------------------- src/mesa/main/histogram.h | 6 ----- src/mesa/main/image.c | 67 ++++++++++++++++++++++++++++++++++++++++++++--- 3 files changed, 64 insertions(+), 74 deletions(-) (limited to 'src') diff --git a/src/mesa/main/histogram.c b/src/mesa/main/histogram.c index 6a7f09489c..77c458d540 100644 --- a/src/mesa/main/histogram.c +++ b/src/mesa/main/histogram.c @@ -31,71 +31,6 @@ #include "histogram.h" -/********************************************************************** - * Internal functions - */ - - -/* - * Update the min/max values from an array of fragment colors. - */ -void -_mesa_update_minmax(GLcontext *ctx, GLuint n, const GLfloat rgba[][4]) -{ - GLuint i; - for (i = 0; i < n; i++) { - /* update mins */ - if (rgba[i][RCOMP] < ctx->MinMax.Min[RCOMP]) - ctx->MinMax.Min[RCOMP] = rgba[i][RCOMP]; - if (rgba[i][GCOMP] < ctx->MinMax.Min[GCOMP]) - ctx->MinMax.Min[GCOMP] = rgba[i][GCOMP]; - if (rgba[i][BCOMP] < ctx->MinMax.Min[BCOMP]) - ctx->MinMax.Min[BCOMP] = rgba[i][BCOMP]; - if (rgba[i][ACOMP] < ctx->MinMax.Min[ACOMP]) - ctx->MinMax.Min[ACOMP] = rgba[i][ACOMP]; - - /* update maxs */ - if (rgba[i][RCOMP] > ctx->MinMax.Max[RCOMP]) - ctx->MinMax.Max[RCOMP] = rgba[i][RCOMP]; - if (rgba[i][GCOMP] > ctx->MinMax.Max[GCOMP]) - ctx->MinMax.Max[GCOMP] = rgba[i][GCOMP]; - if (rgba[i][BCOMP] > ctx->MinMax.Max[BCOMP]) - ctx->MinMax.Max[BCOMP] = rgba[i][BCOMP]; - if (rgba[i][ACOMP] > ctx->MinMax.Max[ACOMP]) - ctx->MinMax.Max[ACOMP] = rgba[i][ACOMP]; - } -} - - -/* - * Update the histogram values from an array of fragment colors. - */ -void -_mesa_update_histogram(GLcontext *ctx, GLuint n, const GLfloat rgba[][4]) -{ - const GLint max = ctx->Histogram.Width - 1; - GLfloat w = (GLfloat) max; - GLuint i; - - if (ctx->Histogram.Width == 0) - return; - - for (i = 0; i < n; i++) { - GLint ri = IROUND(rgba[i][RCOMP] * w); - GLint gi = IROUND(rgba[i][GCOMP] * w); - GLint bi = IROUND(rgba[i][BCOMP] * w); - GLint ai = IROUND(rgba[i][ACOMP] * w); - ri = CLAMP(ri, 0, max); - gi = CLAMP(gi, 0, max); - bi = CLAMP(bi, 0, max); - ai = CLAMP(ai, 0, max); - ctx->Histogram.Count[ri][RCOMP]++; - ctx->Histogram.Count[gi][GCOMP]++; - ctx->Histogram.Count[bi][BCOMP]++; - ctx->Histogram.Count[ai][ACOMP]++; - } -} - /* * XXX the packed pixel formats haven't been tested. diff --git a/src/mesa/main/histogram.h b/src/mesa/main/histogram.h index 974447231d..367e9b11ba 100644 --- a/src/mesa/main/histogram.h +++ b/src/mesa/main/histogram.h @@ -71,12 +71,6 @@ _mesa_ResetHistogram(GLenum target); extern void GLAPIENTRY _mesa_ResetMinmax(GLenum target); -extern void -_mesa_update_minmax(GLcontext *ctx, GLuint n, const GLfloat rgba[][4]); - -extern void -_mesa_update_histogram(GLcontext *ctx, GLuint n, const GLfloat rgba[][4]); - extern void _mesa_init_histogram( GLcontext * ctx ); #else diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c index 63bcc48e68..a6ac821251 100644 --- a/src/mesa/main/image.c +++ b/src/mesa/main/image.c @@ -34,7 +34,6 @@ #include "context.h" #include "image.h" #include "imports.h" -#include "histogram.h" #include "macros.h" #include "pixel.h" @@ -1431,6 +1430,68 @@ _mesa_scale_and_bias_depth_uint(const GLcontext *ctx, GLuint n, } + +/* + * Update the min/max values from an array of fragment colors. + */ +static void +update_minmax(GLcontext *ctx, GLuint n, const GLfloat rgba[][4]) +{ + GLuint i; + for (i = 0; i < n; i++) { + /* update mins */ + if (rgba[i][RCOMP] < ctx->MinMax.Min[RCOMP]) + ctx->MinMax.Min[RCOMP] = rgba[i][RCOMP]; + if (rgba[i][GCOMP] < ctx->MinMax.Min[GCOMP]) + ctx->MinMax.Min[GCOMP] = rgba[i][GCOMP]; + if (rgba[i][BCOMP] < ctx->MinMax.Min[BCOMP]) + ctx->MinMax.Min[BCOMP] = rgba[i][BCOMP]; + if (rgba[i][ACOMP] < ctx->MinMax.Min[ACOMP]) + ctx->MinMax.Min[ACOMP] = rgba[i][ACOMP]; + + /* update maxs */ + if (rgba[i][RCOMP] > ctx->MinMax.Max[RCOMP]) + ctx->MinMax.Max[RCOMP] = rgba[i][RCOMP]; + if (rgba[i][GCOMP] > ctx->MinMax.Max[GCOMP]) + ctx->MinMax.Max[GCOMP] = rgba[i][GCOMP]; + if (rgba[i][BCOMP] > ctx->MinMax.Max[BCOMP]) + ctx->MinMax.Max[BCOMP] = rgba[i][BCOMP]; + if (rgba[i][ACOMP] > ctx->MinMax.Max[ACOMP]) + ctx->MinMax.Max[ACOMP] = rgba[i][ACOMP]; + } +} + + +/* + * Update the histogram values from an array of fragment colors. + */ +static void +update_histogram(GLcontext *ctx, GLuint n, const GLfloat rgba[][4]) +{ + const GLint max = ctx->Histogram.Width - 1; + GLfloat w = (GLfloat) max; + GLuint i; + + if (ctx->Histogram.Width == 0) + return; + + for (i = 0; i < n; i++) { + GLint ri = IROUND(rgba[i][RCOMP] * w); + GLint gi = IROUND(rgba[i][GCOMP] * w); + GLint bi = IROUND(rgba[i][BCOMP] * w); + GLint ai = IROUND(rgba[i][ACOMP] * w); + ri = CLAMP(ri, 0, max); + gi = CLAMP(gi, 0, max); + bi = CLAMP(bi, 0, max); + ai = CLAMP(ai, 0, max); + ctx->Histogram.Count[ri][RCOMP]++; + ctx->Histogram.Count[gi][GCOMP]++; + ctx->Histogram.Count[bi][BCOMP]++; + ctx->Histogram.Count[ai][ACOMP]++; + } +} + + /** * Apply various pixel transfer operations to an array of RGBA pixels * as indicated by the transferOps bitmask @@ -1486,11 +1547,11 @@ _mesa_apply_rgba_transfer_ops(GLcontext *ctx, GLbitfield transferOps, } /* update histogram count */ if (transferOps & IMAGE_HISTOGRAM_BIT) { - _mesa_update_histogram(ctx, n, (CONST GLfloat (*)[4]) rgba); + update_histogram(ctx, n, (CONST GLfloat (*)[4]) rgba); } /* update min/max values */ if (transferOps & IMAGE_MIN_MAX_BIT) { - _mesa_update_minmax(ctx, n, (CONST GLfloat (*)[4]) rgba); + update_minmax(ctx, n, (CONST GLfloat (*)[4]) rgba); } /* clamping to [0,1] */ if (transferOps & IMAGE_CLAMP_BIT) { -- cgit v1.2.3 From 0116ec1af36356c0ee845b3d1384e73316052497 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 15:19:08 -0600 Subject: mesa: remove unused api_eval.h header file --- src/mesa/main/api_eval.h | 42 ------------------------------------------ src/mesa/main/sources | 1 - src/mesa/vbo/vbo_exec_eval.c | 1 - 3 files changed, 44 deletions(-) delete mode 100644 src/mesa/main/api_eval.h (limited to 'src') diff --git a/src/mesa/main/api_eval.h b/src/mesa/main/api_eval.h deleted file mode 100644 index d6cb38c9f8..0000000000 --- a/src/mesa/main/api_eval.h +++ /dev/null @@ -1,42 +0,0 @@ - -/* - * Mesa 3-D graphics library - * Version: 3.5 - * - * Copyright (C) 1999-2001 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -#ifndef API_EVAL_H -#define API_EVAL_H - -#include "mtypes.h" - -extern void _mesa_EvalPoint1( GLint i ); -extern void _mesa_EvalPoint2( GLint i, GLint j ); -extern void _mesa_EvalCoord1f( GLfloat u ); -extern void _mesa_EvalCoord2f( GLfloat u, GLfloat v ); -extern void _mesa_EvalCoord1fv( const GLfloat *u ); -extern void _mesa_EvalCoord2fv( const GLfloat *u ); - -#endif - - - diff --git a/src/mesa/main/sources b/src/mesa/main/sources index 5379fa08f4..5a6cddf4bd 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -75,7 +75,6 @@ vsnprintf.c MESA_MAIN_HEADERS = \ accum.h \ api_arrayelt.h \ -api_eval.h \ api_exec.h \ api_loopback.h \ api_noop.h \ diff --git a/src/mesa/vbo/vbo_exec_eval.c b/src/mesa/vbo/vbo_exec_eval.c index 0ba5585d24..0c691b3a5c 100644 --- a/src/mesa/vbo/vbo_exec_eval.c +++ b/src/mesa/vbo/vbo_exec_eval.c @@ -26,7 +26,6 @@ */ #include "main/glheader.h" -#include "main/api_eval.h" #include "main/context.h" #include "main/macros.h" #include "math/m_eval.h" -- cgit v1.2.3 From 5f46bf77af5f49e63976ad51c5b4a7da8490be3e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 9 Jun 2008 16:18:41 -0600 Subject: mesa: Xlib libGL.so fixes from master --- src/mesa/drivers/x11/Makefile | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/x11/Makefile b/src/mesa/drivers/x11/Makefile index 8ee9b61808..51226eeae1 100644 --- a/src/mesa/drivers/x11/Makefile +++ b/src/mesa/drivers/x11/Makefile @@ -18,7 +18,6 @@ GL_TINY = 0$(MESA_MAJOR)0$(MESA_MINOR)0$(MESA_TINY) HEADERS = \ glxapi.h \ glxheader.h \ - realglx.h \ xfonts.h \ xmesaP.h \ xm_glide.h \ @@ -27,7 +26,6 @@ HEADERS = \ SOURCES = \ fakeglx.c \ glxapi.c \ - realglx.c \ xfonts.c \ xm_api.c \ xm_buffer.c \ @@ -61,7 +59,7 @@ $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(OBJECTS) $(CORE_MESA) -linker '$(CC)' -ldflags '$(LDFLAGS)' \ -major $(GL_MAJOR) -minor $(GL_MINOR) -patch $(GL_TINY) \ -install $(TOP)/$(LIB_DIR) \ - $(MKLIB_OPTIONS) $(GL_LIB_DEPS) $(OBJECTS) + $(MKLIB_OPTIONS) $(GL_LIB_DEPS) $(OBJECTS) $(CORE_MESA) -- cgit v1.2.3 From a70684bf256c3d5bc3a729bf9e9cf1a64cb2064a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 10 Jun 2008 08:32:52 +0900 Subject: gallium: Deprecate GETENV. Replace by debug_get_bool_option. debug_get_bool_option will interpret "n", "no", "0", "f", or "false" as FALSE; and everything else as TRUE. The default value (used when the variable is not set) is received as a parameter. --- src/gallium/auxiliary/draw/draw_pt.c | 4 ++-- src/gallium/drivers/i915simple/i915_context.c | 2 +- src/gallium/drivers/softpipe/sp_context.c | 4 ++-- src/gallium/include/pipe/p_util.h | 2 -- 4 files changed, 5 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index f0d7b51ad7..9140faeea9 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -119,8 +119,8 @@ draw_pt_arrays(struct draw_context *draw, boolean draw_pt_init( struct draw_context *draw ) { - draw->pt.test_fse = GETENV("DRAW_FSE") != NULL; - draw->pt.no_fse = GETENV("DRAW_NO_FSE") != NULL; + draw->pt.test_fse = debug_get_bool_option("DRAW_FSE", FALSE); + draw->pt.no_fse = debug_get_bool_option("DRAW_NO_FSE", FALSE); draw->pt.front.vcache = draw_pt_vcache( draw ); if (!draw->pt.front.vcache) diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 243b4a6852..4c01b8d5b1 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -161,7 +161,7 @@ struct pipe_context *i915_create_context( struct pipe_screen *screen, */ i915->draw = draw_create(); assert(i915->draw); - if (!GETENV("I915_NO_VBUF")) { + if (!debug_get_bool_option("I915_NO_VBUF", FALSE)) { draw_set_rasterize_stage(i915->draw, i915_draw_vbuf_stage(i915)); } else { diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 1e0106b86c..6e14f684f1 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -129,12 +129,12 @@ softpipe_create( struct pipe_screen *screen, uint i; #ifdef PIPE_ARCH_X86 - softpipe->use_sse = GETENV( "GALLIUM_NOSSE" ) == NULL; + softpipe->use_sse = !debug_get_bool_option( "GALLIUM_NOSSE", FALSE ); #else softpipe->use_sse = FALSE; #endif - softpipe->dump_fs = GETENV( "GALLIUM_DUMP_FS" ) != NULL; + softpipe->dump_fs = debug_get_bool_option( "GALLIUM_DUMP_FS", FALSE ); softpipe->pipe.winsys = pipe_winsys; softpipe->pipe.screen = screen; diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 5547e57833..8b3003bcef 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -140,8 +140,6 @@ REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) #define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) -#define GETENV( X ) debug_get_option( X, NULL ) - /** * Return memory on given byte alignment -- cgit v1.2.3 From 4b52f4df1b37918a363d05e0b3db22125e801367 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 9 Jun 2008 18:49:13 +0900 Subject: pipebuffer: Be more lenient when matching cached buffer sizes. Reuse cached buffers up to twice as big a requested. --- src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c index 4bd3f94a6c..f1a457dde4 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c @@ -207,8 +207,11 @@ pb_cache_is_buffer_compat(struct pb_cache_buffer *buf, size_t size, const struct pb_desc *desc) { - /* TODO: be more lenient with size */ - if(buf->base.base.size != size) + if(buf->base.base.size < size) + return FALSE; + + /* be lenient with size */ + if(buf->base.base.size >= 2*size) return FALSE; if(!pb_check_alignment(desc->alignment, buf->base.base.alignment)) -- cgit v1.2.3 From 0c5b1a8ffb21f72fcde64a7daa13d5dab5b90425 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 10 Jun 2008 08:38:24 +0900 Subject: pipebuffer: Alternative buffer manager. For situations where one has a reserve memory pool, or a faster/slower pool. --- src/gallium/auxiliary/pipebuffer/Makefile | 1 + src/gallium/auxiliary/pipebuffer/SConscript | 1 + src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 5 ++ src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c | 101 +++++++++++++++++++++++ 4 files changed, 108 insertions(+) create mode 100644 src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile index ff09011b66..1d9b036c07 100644 --- a/src/gallium/auxiliary/pipebuffer/Makefile +++ b/src/gallium/auxiliary/pipebuffer/Makefile @@ -6,6 +6,7 @@ LIBNAME = pipebuffer C_SOURCES = \ pb_buffer_fenced.c \ pb_buffer_malloc.c \ + pb_bufmgr_alt.c \ pb_bufmgr_cache.c \ pb_bufmgr_fenced.c \ pb_bufmgr_mm.c \ diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript index 9db0c0eae3..e52177bc79 100644 --- a/src/gallium/auxiliary/pipebuffer/SConscript +++ b/src/gallium/auxiliary/pipebuffer/SConscript @@ -5,6 +5,7 @@ pipebuffer = env.ConvenienceLibrary( source = [ 'pb_buffer_fenced.c', 'pb_buffer_malloc.c', + 'pb_bufmgr_alt.c', 'pb_bufmgr_cache.c', 'pb_bufmgr_fenced.c', 'pb_bufmgr_mm.c', diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index 4a922d16c1..00279f7010 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -171,6 +171,11 @@ fenced_bufmgr_create(struct pb_manager *provider, struct pipe_winsys *winsys); +struct pb_manager * +pb_alt_manager_create(struct pb_manager *provider1, + struct pb_manager *provider2); + + #ifdef __cplusplus } #endif diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c new file mode 100644 index 0000000000..702bef1c04 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c @@ -0,0 +1,101 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Allocate buffers from two alternative buffer providers. + * + * \author Jose Fonseca + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_util.h" + +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +struct pb_alt_manager +{ + struct pb_manager base; + + struct pb_manager *provider1; + struct pb_manager *provider2; +}; + + +static INLINE struct pb_alt_manager * +pb_alt_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct pb_alt_manager *)mgr; +} + + +static struct pb_buffer * +pb_alt_manager_create_buffer(struct pb_manager *_mgr, + size_t size, + const struct pb_desc *desc) +{ + struct pb_alt_manager *mgr = pb_alt_manager(_mgr); + struct pb_buffer *buf; + + buf = mgr->provider1->create_buffer(mgr->provider1, size, desc); + if(buf) + return buf; + + buf = mgr->provider2->create_buffer(mgr->provider2, size, desc); + return buf; +} + + +static void +pb_alt_manager_destroy(struct pb_manager *mgr) +{ + FREE(mgr); +} + + +struct pb_manager * +pb_alt_manager_create(struct pb_manager *provider1, + struct pb_manager *provider2) +{ + struct pb_alt_manager *mgr; + + mgr = CALLOC_STRUCT(pb_alt_manager); + if (!mgr) + return NULL; + + mgr->base.destroy = pb_alt_manager_destroy; + mgr->base.create_buffer = pb_alt_manager_create_buffer; + mgr->provider1 = provider1; + mgr->provider2 = provider2; + + return &mgr->base; +} -- cgit v1.2.3 From 3531c5284bb8dc772cd97c6be5bf589c160f9ae8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 9 Jun 2008 19:56:52 +0900 Subject: gallium: Detect buffer overflows in the homegrown memory debugger. --- src/gallium/auxiliary/util/p_debug_mem.c | 49 ++++++++++++++++++++++++++++++-- 1 file changed, 46 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c index 78497c5f6a..51dbd0a710 100644 --- a/src/gallium/auxiliary/util/p_debug_mem.c +++ b/src/gallium/auxiliary/util/p_debug_mem.c @@ -75,6 +75,12 @@ struct debug_memory_header unsigned magic; }; +struct debug_memory_footer +{ + unsigned magic; +}; + + static struct list_head list = { &list, &list }; static unsigned long last_no = 0; @@ -98,14 +104,24 @@ data_from_header(struct debug_memory_header *hdr) return NULL; } +static INLINE struct debug_memory_footer * +footer_from_header(struct debug_memory_header *hdr) +{ + if(hdr) + return (struct debug_memory_footer *)((char *)hdr + sizeof(struct debug_memory_header) + hdr->size); + else + return NULL; +} + void * debug_malloc(const char *file, unsigned line, const char *function, size_t size) { struct debug_memory_header *hdr; + struct debug_memory_footer *ftr; - hdr = real_malloc(sizeof(*hdr) + size); + hdr = real_malloc(sizeof(*hdr) + size + sizeof(*ftr)); if(!hdr) return NULL; @@ -116,6 +132,9 @@ debug_malloc(const char *file, unsigned line, const char *function, hdr->size = size; hdr->magic = DEBUG_MEMORY_MAGIC; + ftr = footer_from_header(hdr); + ftr->magic = DEBUG_MEMORY_MAGIC; + LIST_ADDTAIL(&hdr->head, &list); return data_from_header(hdr); @@ -126,6 +145,7 @@ debug_free(const char *file, unsigned line, const char *function, void *ptr) { struct debug_memory_header *hdr; + struct debug_memory_footer *ftr; if(!ptr) return; @@ -139,8 +159,17 @@ debug_free(const char *file, unsigned line, const char *function, return; } + ftr = footer_from_header(hdr); + if(ftr->magic != DEBUG_MEMORY_MAGIC) { + debug_printf("%s:%u:%s: buffer overflow %p\n", + hdr->file, hdr->line, hdr->function, + ptr); + debug_assert(0); + } + LIST_DEL(&hdr->head); hdr->magic = 0; + ftr->magic = 0; real_free(hdr); } @@ -160,6 +189,7 @@ debug_realloc(const char *file, unsigned line, const char *function, void *old_ptr, size_t old_size, size_t new_size ) { struct debug_memory_header *old_hdr, *new_hdr; + struct debug_memory_footer *old_ftr, *new_ftr; void *new_ptr; if(!old_ptr) @@ -179,8 +209,16 @@ debug_realloc(const char *file, unsigned line, const char *function, return NULL; } + old_ftr = footer_from_header(old_hdr); + if(old_ftr->magic != DEBUG_MEMORY_MAGIC) { + debug_printf("%s:%u:%s: buffer overflow %p\n", + old_hdr->file, old_hdr->line, old_hdr->function, + old_ptr); + debug_assert(0); + } + /* alloc new */ - new_hdr = real_malloc(sizeof(*new_hdr) + new_size); + new_hdr = real_malloc(sizeof(*new_hdr) + new_size + sizeof(*new_ftr)); if(!new_hdr) return NULL; new_hdr->no = old_hdr->no; @@ -189,14 +227,19 @@ debug_realloc(const char *file, unsigned line, const char *function, new_hdr->function = old_hdr->function; new_hdr->size = new_size; new_hdr->magic = DEBUG_MEMORY_MAGIC; - LIST_REPLACE(&old_hdr->head, &new_hdr->head); + new_ftr = footer_from_header(new_hdr); + new_ftr->magic = DEBUG_MEMORY_MAGIC; + + LIST_REPLACE(&old_hdr->head, &new_hdr->head); + /* copy data */ new_ptr = data_from_header(new_hdr); memcpy( new_ptr, old_ptr, old_size < new_size ? old_size : new_size ); /* free old */ old_hdr->magic = 0; + old_ftr->magic = 0; real_free(old_hdr); return new_ptr; -- cgit v1.2.3 From 7d3bab537c7559df7bdc7a4fae42f8218185b9f7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 10 Jun 2008 08:52:10 +0900 Subject: softpipe: Replace GETENV by debug_get_bool_option. --- src/gallium/drivers/softpipe/sp_context.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 6e14f684f1..626c3a9d4e 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -225,10 +225,10 @@ softpipe_create( struct pipe_screen *screen, if (!softpipe->setup) goto fail; - if (GETENV( "SP_NO_RAST" ) != NULL) + if (debug_get_bool_option( "SP_NO_RAST", FALSE )) softpipe->no_rast = TRUE; - if (GETENV( "SP_NO_VBUF" ) != NULL) { + if (debug_get_bool_option( "SP_NO_VBUF", FALSE )) { /* Deprecated path -- vbuf is the intended interface to the draw module: */ draw_set_rasterize_stage(softpipe->draw, softpipe->setup); -- cgit v1.2.3 From 061e1c6c57703a92ac17b553f592c0c6114cb227 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 10 Jun 2008 09:16:34 +0100 Subject: draw: rework splitting of fan/loop prims in varray.c, fix flatshade issues --- src/gallium/auxiliary/draw/draw_pt_varray.c | 138 ++++++++------------- .../auxiliary/draw/draw_pt_varray_tmp_linear.h | 54 ++++---- 2 files changed, 76 insertions(+), 116 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index f5495a80c7..4479963db1 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -31,7 +31,7 @@ #include "draw/draw_pt.h" #define FETCH_MAX 256 -#define DRAW_MAX (16*FETCH_MAX) +#define DRAW_MAX (FETCH_MAX+8) struct varray_frontend { struct draw_pt_front_end base; @@ -40,11 +40,6 @@ struct varray_frontend { ushort draw_elts[DRAW_MAX]; unsigned fetch_elts[FETCH_MAX]; - unsigned draw_count; - unsigned fetch_count; - - unsigned fetch_start; - unsigned driver_fetch_max; unsigned fetch_max; @@ -54,121 +49,86 @@ struct varray_frontend { unsigned output_prim; }; -static void varray_flush(struct varray_frontend *varray) -{ - if (varray->draw_count) { -#if 0 - debug_printf("FLUSH fc = %d, dc = %d\n", - varray->fetch_count, - varray->draw_count); - debug_printf("\telt0 = %d, eltx = %d, draw0 = %d, drawx = %d\n", - varray->fetch_elts[0], - varray->fetch_elts[varray->fetch_count-1], - varray->draw_elts[0], - varray->draw_elts[varray->draw_count-1]); -#endif - varray->middle->run(varray->middle, - varray->fetch_elts, - varray->fetch_count, - varray->draw_elts, - varray->draw_count); - } - - varray->fetch_count = 0; - varray->draw_count = 0; -} static void varray_flush_linear(struct varray_frontend *varray, unsigned start, unsigned count) { if (count) { -#if 0 - debug_printf("FLUSH LINEAR start = %d, count = %d\n", - start, - count); -#endif assert(varray->middle->run_linear); varray->middle->run_linear(varray->middle, start, count); } } -static INLINE void fetch_init(struct varray_frontend *varray, - unsigned count) +static void varray_line_loop_segment(struct varray_frontend *varray, + unsigned start, + unsigned segment_start, + unsigned segment_count, + boolean end ) { - unsigned idx; -#if 0 - debug_printf("FETCH INIT c = %d, fs = %d\n", - count, - varray->fetch_start); -#endif - for (idx = 0; idx < count; ++idx) { - varray->fetch_elts[idx] = varray->fetch_start + idx; - } - varray->fetch_start += idx; - varray->fetch_count = idx; -} + assert(segment_count+1 < varray->fetch_max); + if (segment_count >= 1) { + unsigned nr = 0, i; + for (i = 0; i < segment_count; i++) + varray->fetch_elts[nr++] = start + segment_start + i; + if (end) + varray->fetch_elts[nr++] = start; + assert(nr < FETCH_MAX); -static INLINE void add_draw_el(struct varray_frontend *varray, - unsigned idx) -{ - varray->draw_elts[varray->draw_count++] = (ushort)idx; + varray->middle->run(varray->middle, + varray->fetch_elts, + nr, + varray->draw_elts, /* ie. linear */ + nr); + } } -static INLINE void varray_triangle( struct varray_frontend *varray, - unsigned i0, - unsigned i1, - unsigned i2 ) -{ - add_draw_el(varray, i0); - add_draw_el(varray, i1); - add_draw_el(varray, i2); -} -static INLINE void varray_line( struct varray_frontend *varray, - unsigned i0, - unsigned i1 ) +static void varray_fan_segment(struct varray_frontend *varray, + unsigned start, + unsigned segment_start, + unsigned segment_count ) { - add_draw_el(varray, i0); - add_draw_el(varray, i1); -} + assert(segment_count+1 < varray->fetch_max); + if (segment_count >= 2) { + unsigned nr = 0, i; + if (segment_start != 0) + varray->fetch_elts[nr++] = start; -static INLINE void varray_point( struct varray_frontend *varray, - unsigned i0 ) -{ - add_draw_el(varray, i0); + for (i = 0 ; i < segment_count; i++) + varray->fetch_elts[nr++] = start + segment_start + i; + + assert(nr < FETCH_MAX); + + varray->middle->run(varray->middle, + varray->fetch_elts, + nr, + varray->draw_elts, /* ie. linear */ + nr); + } } -#if 0 -#define TRIANGLE(flags,i0,i1,i2) varray_triangle(varray,i0,i1,i2) -#define LINE(flags,i0,i1) varray_line(varray,i0,i1) -#define POINT(i0) varray_point(varray,i0) -#define FUNC varray_decompose -#include "draw_pt_decompose.h" -#else -#define TRIANGLE(vc,i0,i1,i2) varray_triangle(vc,i0,i1,i2) -#define LINE(vc,i0,i1) varray_line(vc,i0,i1) -#define POINT(vc,i0) varray_point(vc,i0) + + #define FUNC varray_run #include "draw_pt_varray_tmp_linear.h" -#endif static unsigned decompose_prim[PIPE_PRIM_POLYGON + 1] = { PIPE_PRIM_POINTS, PIPE_PRIM_LINES, - PIPE_PRIM_LINES, /* decomposed LINELOOP */ + PIPE_PRIM_LINE_STRIP, /* decomposed LINELOOP */ PIPE_PRIM_LINE_STRIP, PIPE_PRIM_TRIANGLES, PIPE_PRIM_TRIANGLE_STRIP, - PIPE_PRIM_TRIANGLES, /* decomposed TRI_FAN */ + PIPE_PRIM_TRIANGLE_FAN, PIPE_PRIM_QUADS, PIPE_PRIM_QUAD_STRIP, - PIPE_PRIM_TRIANGLES /* decomposed POLYGON */ + PIPE_PRIM_POLYGON }; @@ -186,7 +146,8 @@ static void varray_prepare(struct draw_pt_front_end *frontend, varray->output_prim = decompose_prim[prim]; varray->middle = middle; - middle->prepare(middle, varray->output_prim, opt, &varray->fetch_max ); + middle->prepare(middle, varray->output_prim, opt, &varray->driver_fetch_max ); + varray->fetch_max = MIN2(FETCH_MAX, varray->driver_fetch_max); } @@ -207,6 +168,7 @@ static void varray_destroy(struct draw_pt_front_end *frontend) struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw) { + unsigned i; struct varray_frontend *varray = CALLOC_STRUCT(varray_frontend); if (varray == NULL) return NULL; @@ -217,5 +179,9 @@ struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw) varray->base.destroy = varray_destroy; varray->draw = draw; + for (i = 0; i < DRAW_MAX; i++) { + varray->draw_elts[i] = i; + } + return &varray->base; } diff --git a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h index 6e5e30f38e..55a8e6521d 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h +++ b/src/gallium/auxiliary/draw/draw_pt_varray_tmp_linear.h @@ -11,11 +11,9 @@ static void FUNC(struct draw_pt_front_end *frontend, struct varray_frontend *varray = (struct varray_frontend *)frontend; unsigned start = (unsigned)elts; - unsigned i, j; + unsigned j; unsigned first, incr; - varray->fetch_start = start; - draw_pt_split_prim(varray->input_prim, &first, &incr); /* Sanitize primitive length: @@ -40,7 +38,7 @@ static void FUNC(struct draw_pt_front_end *frontend, case PIPE_PRIM_QUAD_STRIP: for (j = 0; j < count;) { unsigned remaining = count - j; - unsigned nr = trim( MIN2(varray->fetch_max, remaining), first, incr ); + unsigned nr = trim( MIN2(varray->driver_fetch_max, remaining), first, incr ); varray_flush_linear(varray, start + j, nr); j += nr; if (nr != remaining) @@ -49,45 +47,41 @@ static void FUNC(struct draw_pt_front_end *frontend, break; case PIPE_PRIM_LINE_LOOP: - if (count >= 2) { - unsigned fetch_max = MIN2(FETCH_MAX, varray->fetch_max); - for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(fetch_max, count - j); - end -= (end % incr); - for (i = 1; i < end; i++) { - LINE(varray, i - 1, i); - } - LINE(varray, i - 1, 0); - i = end; - fetch_init(varray, end); - varray_flush(varray); - } + /* Always have to decompose as we've stated that this will be + * emitted as a line-strip. + */ + for (j = 0; j < count;) { + unsigned remaining = count - j; + unsigned nr = trim( MIN2(varray->fetch_max-1, remaining), first, incr ); + varray_line_loop_segment(varray, start, j, nr, nr == remaining); + j += nr; + if (nr != remaining) + j -= (first - incr); } break; case PIPE_PRIM_POLYGON: - case PIPE_PRIM_TRIANGLE_FAN: { - unsigned fetch_max = MIN2(FETCH_MAX, varray->fetch_max); - for (j = 0; j + first <= count; j += i) { - unsigned end = MIN2(fetch_max, count - j); - end -= (end % incr); - for (i = 2; i < end; i++) { - TRIANGLE(varray, 0, i - 1, i); + case PIPE_PRIM_TRIANGLE_FAN: + if (count < varray->driver_fetch_max) { + varray_flush_linear(varray, start, count); + } + else { + for ( j = 0; j < count;) { + unsigned remaining = count - j; + unsigned nr = trim( MIN2(varray->fetch_max-1, remaining), first, incr ); + varray_fan_segment(varray, start, j, nr); + j += nr; + if (nr != remaining) + j -= (first - incr); } - i = end; - fetch_init(varray, end); - varray_flush(varray); } break; - } default: assert(0); break; } - - varray_flush(varray); } #undef TRIANGLE -- cgit v1.2.3 From d85f1e95d32777e8b5e5fa818caff4bb5afe46eb Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Tue, 10 Jun 2008 15:43:33 +0200 Subject: copy msaa visuals capability --- src/glx/x11/glcontextmodes.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/glx/x11/glcontextmodes.c b/src/glx/x11/glcontextmodes.c index c165184903..788ecf6a3a 100644 --- a/src/glx/x11/glcontextmodes.c +++ b/src/glx/x11/glcontextmodes.c @@ -118,7 +118,7 @@ _gl_convert_to_x_visual_type( int visualType ) * of the fields in \c config are copied to \c mode. Additional fields in * \c mode that can be derrived from the fields of \c config (i.e., * \c haveDepthBuffer) are also filled in. The remaining fields in \c mode - * that cannot be derrived are set to default values. + * that cannot be derived are set to default values. * * \param mode Destination GL context mode. * \param config Source GLX visual config. @@ -184,6 +184,9 @@ _gl_copy_visual_to_context_mode( __GLcontextModes * mode, mode->transparentBlue = config->transparentBlue; mode->transparentAlpha = config->transparentAlpha; mode->transparentIndex = config->transparentIndex; + mode->samples = config->multiSampleSize; + mode->sampleBuffers = config->nMultiSampleBuffers; + /* mode->visualSelectGroup = config->visualSelectGroup; ? */ mode->swapMethod = GLX_SWAP_UNDEFINED_OML; -- cgit v1.2.3 From 14a13e3767f080a48a4ae01f803dd0bc8754f441 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 10 Jun 2008 14:45:34 +0100 Subject: draw: fix edgeflags on clipped poly emit --- src/gallium/auxiliary/draw/draw_pipe_clip.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c index c11ed934a4..77ccddac4a 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_clip.c +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -171,7 +171,7 @@ static void emit_poly( struct draw_stage *stage, header.flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; header.pad = 0; - for (i = 2; i < n; i++, header.flags = 0) { + for (i = 2; i < n; i++, header.flags = edge_middle) { header.v[0] = inlist[i-1]; header.v[1] = inlist[i]; header.v[2] = inlist[0]; /* keep in v[2] for flatshading */ -- cgit v1.2.3 From ad44e68706877ab06929747e7a82c718c1c27e02 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 10 Jun 2008 14:46:25 +0100 Subject: draw: no need to rearrange most primitives in vcache for flatshade-first The driver/pipeline will still be applying flatshade-first state to the triangles emitted from vcache, so there's no need to rotate the vertices of most primitives. The only exception is POLYGON, as explained in the code. --- src/gallium/auxiliary/draw/draw_pt_fetch.c | 5 +- src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h | 123 +++++++++++++----------- 2 files changed, 72 insertions(+), 56 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 07f4c99164..b8fd3aa1d8 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -114,7 +114,10 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, fetch->translate = translate_cache_find(fetch->cache, &key); { - static struct vertex_header vh = { 0, 1, 0, 0xffff }; + static struct vertex_header vh = { 0, + 1, /* edgeflag */ + 0, + 0xffff }; fetch->translate->set_buffer(fetch->translate, draw->pt.nr_vertex_buffers, &vh, diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h index ec05bbeab4..f2334a84f6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h @@ -7,15 +7,18 @@ static void FUNC( struct draw_pt_front_end *frontend, { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; struct draw_context *draw = vcache->draw; - - boolean flatfirst = (draw->rasterizer->flatshade && - draw->rasterizer->flatshade_first); unsigned i; ushort flags; if (0) debug_printf("%s %d\n", __FUNCTION__, count); - + /* Note that no adjustment is made here for flatshade provoking + * vertex. This was previously the case, but was incorrect as the + * same logic would be applied twice both here & in the pipeline + * code or driver. The rule is now that we just preserve the + * original order in this code, and leave identification of the PV + * to the pipeline and driver. + */ switch (vcache->input_prim) { case PIPE_PRIM_POINTS: for (i = 0; i < count; i ++) { @@ -72,45 +75,23 @@ static void FUNC( struct draw_pt_front_end *frontend, break; case PIPE_PRIM_TRIANGLE_STRIP: - if (flatfirst) { - for (i = 0; i+2 < count; i++) { - TRIANGLE( vcache, - DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - get_elt(elts, i + 0), - get_elt(elts, i + 1 + (i&1)), - get_elt(elts, i + 2 - (i&1))); - } - } - else { - for (i = 0; i+2 < count; i++) { - TRIANGLE( vcache, - DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - get_elt(elts, i + 0 + (i&1)), - get_elt(elts, i + 1 - (i&1)), - get_elt(elts, i + 2 )); - } + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, i + 0 + (i&1)), + get_elt(elts, i + 1 - (i&1)), + get_elt(elts, i + 2 )); } break; case PIPE_PRIM_TRIANGLE_FAN: if (count >= 3) { - if (flatfirst) { - for (i = 0; i+2 < count; i++) { - TRIANGLE( vcache, - DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0 )); - } - } - else { - for (i = 0; i+2 < count; i++) { - TRIANGLE( vcache, - DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - get_elt(elts, 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2 )); - } + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0 )); } } break; @@ -138,26 +119,58 @@ static void FUNC( struct draw_pt_front_end *frontend, case PIPE_PRIM_POLYGON: { + boolean flatfirst = (draw->rasterizer->flatshade && + draw->rasterizer->flatshade_first); + /* These bitflags look a little odd because we submit the * vertices as (1,2,0) to satisfy flatshade requirements. + * + * Polygon is defined has having vertex 0 be the provoking + * flatshade vertex and all known API's match this usage. + * However, the PV's for the triangles we emit from this + * decomposition vary according to the API, and hence we have + * to choose between two ways of emitting the triangles. */ - const ushort edge_first = DRAW_PIPE_EDGE_FLAG_2; - const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0; - const ushort edge_last = DRAW_PIPE_EDGE_FLAG_1; - - flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; - - for (i = 0; i+2 < count; i++, flags = edge_middle) { - - if (i + 3 == count) - flags |= edge_last; - - TRIANGLE( vcache, - flags, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0)); - } + if (flatfirst) { + const ushort edge_first = DRAW_PIPE_EDGE_FLAG_0; + const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_1; + const ushort edge_last = DRAW_PIPE_EDGE_FLAG_2; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + + for (i = 0; i+2 < count; i++, flags = edge_middle) { + + if (i + 3 == count) + flags |= edge_last; + + /* PV is first vertex */ + TRIANGLE( vcache, + flags, + get_elt(elts, 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2)); + } + } + else { + const ushort edge_first = DRAW_PIPE_EDGE_FLAG_2; + const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const ushort edge_last = DRAW_PIPE_EDGE_FLAG_1; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + + for (i = 0; i+2 < count; i++, flags = edge_middle) { + + if (i + 3 == count) + flags |= edge_last; + + /* PV is third vertex */ + TRIANGLE( vcache, + flags, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0)); + } + } } break; -- cgit v1.2.3 From aa7c21a45b90e6e502c4967f892f1691411d761f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Tue, 10 Jun 2008 16:56:32 +0100 Subject: Revert "draw: no need to rearrange most primitives in vcache for flatshade-first" This reverts commit ad44e68706877ab06929747e7a82c718c1c27e02. --- src/gallium/auxiliary/draw/draw_pt_fetch.c | 5 +- src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h | 123 +++++++++++------------- 2 files changed, 56 insertions(+), 72 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index b8fd3aa1d8..07f4c99164 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -114,10 +114,7 @@ void draw_pt_fetch_prepare( struct pt_fetch *fetch, fetch->translate = translate_cache_find(fetch->cache, &key); { - static struct vertex_header vh = { 0, - 1, /* edgeflag */ - 0, - 0xffff }; + static struct vertex_header vh = { 0, 1, 0, 0xffff }; fetch->translate->set_buffer(fetch->translate, draw->pt.nr_vertex_buffers, &vh, diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h index f2334a84f6..ec05bbeab4 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h +++ b/src/gallium/auxiliary/draw/draw_pt_vcache_tmp.h @@ -7,18 +7,15 @@ static void FUNC( struct draw_pt_front_end *frontend, { struct vcache_frontend *vcache = (struct vcache_frontend *)frontend; struct draw_context *draw = vcache->draw; + + boolean flatfirst = (draw->rasterizer->flatshade && + draw->rasterizer->flatshade_first); unsigned i; ushort flags; if (0) debug_printf("%s %d\n", __FUNCTION__, count); - /* Note that no adjustment is made here for flatshade provoking - * vertex. This was previously the case, but was incorrect as the - * same logic would be applied twice both here & in the pipeline - * code or driver. The rule is now that we just preserve the - * original order in this code, and leave identification of the PV - * to the pipeline and driver. - */ + switch (vcache->input_prim) { case PIPE_PRIM_POINTS: for (i = 0; i < count; i ++) { @@ -75,23 +72,45 @@ static void FUNC( struct draw_pt_front_end *frontend, break; case PIPE_PRIM_TRIANGLE_STRIP: - for (i = 0; i+2 < count; i++) { - TRIANGLE( vcache, - DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - get_elt(elts, i + 0 + (i&1)), - get_elt(elts, i + 1 - (i&1)), - get_elt(elts, i + 2 )); + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, i + 0), + get_elt(elts, i + 1 + (i&1)), + get_elt(elts, i + 2 - (i&1))); + } + } + else { + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, i + 0 + (i&1)), + get_elt(elts, i + 1 - (i&1)), + get_elt(elts, i + 2 )); + } } break; case PIPE_PRIM_TRIANGLE_FAN: if (count >= 3) { - for (i = 0; i+2 < count; i++) { - TRIANGLE( vcache, - DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0 )); + if (flatfirst) { + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0 )); + } + } + else { + for (i = 0; i+2 < count; i++) { + TRIANGLE( vcache, + DRAW_PIPE_RESET_STIPPLE | DRAW_PIPE_EDGE_FLAG_ALL, + get_elt(elts, 0), + get_elt(elts, i + 1), + get_elt(elts, i + 2 )); + } } } break; @@ -119,58 +138,26 @@ static void FUNC( struct draw_pt_front_end *frontend, case PIPE_PRIM_POLYGON: { - boolean flatfirst = (draw->rasterizer->flatshade && - draw->rasterizer->flatshade_first); - /* These bitflags look a little odd because we submit the * vertices as (1,2,0) to satisfy flatshade requirements. - * - * Polygon is defined has having vertex 0 be the provoking - * flatshade vertex and all known API's match this usage. - * However, the PV's for the triangles we emit from this - * decomposition vary according to the API, and hence we have - * to choose between two ways of emitting the triangles. */ - if (flatfirst) { - const ushort edge_first = DRAW_PIPE_EDGE_FLAG_0; - const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_1; - const ushort edge_last = DRAW_PIPE_EDGE_FLAG_2; - - flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; - - for (i = 0; i+2 < count; i++, flags = edge_middle) { - - if (i + 3 == count) - flags |= edge_last; - - /* PV is first vertex */ - TRIANGLE( vcache, - flags, - get_elt(elts, 0), - get_elt(elts, i + 1), - get_elt(elts, i + 2)); - } - } - else { - const ushort edge_first = DRAW_PIPE_EDGE_FLAG_2; - const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0; - const ushort edge_last = DRAW_PIPE_EDGE_FLAG_1; - - flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; - - for (i = 0; i+2 < count; i++, flags = edge_middle) { - - if (i + 3 == count) - flags |= edge_last; - - /* PV is third vertex */ - TRIANGLE( vcache, - flags, - get_elt(elts, i + 1), - get_elt(elts, i + 2), - get_elt(elts, 0)); - } - } + const ushort edge_first = DRAW_PIPE_EDGE_FLAG_2; + const ushort edge_middle = DRAW_PIPE_EDGE_FLAG_0; + const ushort edge_last = DRAW_PIPE_EDGE_FLAG_1; + + flags = DRAW_PIPE_RESET_STIPPLE | edge_first | edge_middle; + + for (i = 0; i+2 < count; i++, flags = edge_middle) { + + if (i + 3 == count) + flags |= edge_last; + + TRIANGLE( vcache, + flags, + get_elt(elts, i + 1), + get_elt(elts, i + 2), + get_elt(elts, 0)); + } } break; -- cgit v1.2.3 From 5493c775b7dcf899b14efeaa27916c2b470475b0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 10 Jun 2008 10:09:14 -0600 Subject: add some missing rule dependencies --- src/mesa/Makefile | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index a301625179..5f84f72bb2 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -29,17 +29,17 @@ default: depend asm_subdirs libmesa.a libglapi.a driver_subdirs # Helper libraries used by many drivers: # Make archive of core mesa object files -libmesa.a: $(MESA_OBJECTS) +libmesa.a: $(MESA_OBJECTS) asm_subdirs @ $(TOP)/bin/mklib -o mesa -static $(MESA_OBJECTS) # Make archive of gl* API dispatcher functions only -libglapi.a: $(GLAPI_OBJECTS) +libglapi.a: $(GLAPI_OBJECTS) asm_subdirs @ $(TOP)/bin/mklib -o glapi -static $(GLAPI_OBJECTS) ###################################################################### # Device drivers -driver_subdirs: +driver_subdirs: libmesa.a libglapi.a (cd drivers && $(MAKE)) -- cgit v1.2.3 From 00d90fe845de46649528e3599c0eb42203e78309 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 10 Jun 2008 12:07:38 -0600 Subject: gallium: remove stray include of st_context.h --- src/mesa/main/renderbuffer.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/renderbuffer.c b/src/mesa/main/renderbuffer.c index d576a7a121..3c37d05b40 100644 --- a/src/mesa/main/renderbuffer.c +++ b/src/mesa/main/renderbuffer.c @@ -49,8 +49,6 @@ #include "rbadaptors.h" -#include "state_tracker/st_context.h" - /* 32-bit color index format. Not a public format. */ #define COLOR_INDEX32 0x424243 -- cgit v1.2.3 From 4c1892ca89b549899b9f3080373919c9036747fe Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 10 Jun 2008 12:08:40 -0600 Subject: mesa: remove unnecessary swrast, tnl includes --- src/mesa/vbo/vbo_save_loopback.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo_save_loopback.c b/src/mesa/vbo/vbo_save_loopback.c index f2cef698fb..92ca4ea95d 100644 --- a/src/mesa/vbo/vbo_save_loopback.c +++ b/src/mesa/vbo/vbo_save_loopback.c @@ -25,9 +25,6 @@ * **************************************************************************/ -#include "swrast_setup/swrast_setup.h" -#include "swrast/swrast.h" -#include "tnl/tnl.h" #include "main/context.h" #include "main/glheader.h" #include "main/enums.h" -- cgit v1.2.3 From ef27bdbfa8fd4b7cdd7f1ab0e65bc30f17150d2e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 10 Jun 2008 15:23:51 -0600 Subject: mesa: include mtypes.h --- src/mesa/shader/atifragshader.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/atifragshader.h b/src/mesa/shader/atifragshader.h index 32fb3a8019..e1dc20e606 100644 --- a/src/mesa/shader/atifragshader.h +++ b/src/mesa/shader/atifragshader.h @@ -8,6 +8,8 @@ #ifndef ATIFRAGSHADER_H #define ATIFRAGSHADER_H +#include "main/mtypes.h" + #define MAX_NUM_INSTRUCTIONS_PER_PASS_ATI 8 #define MAX_NUM_PASSES_ATI 2 #define MAX_NUM_FRAGMENT_REGISTERS_ATI 6 -- cgit v1.2.3 From 40d1a40f294f1ed2dacfad6f5498322fc08cc2d1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 10 Jun 2008 16:13:42 -0600 Subject: mesa: refactor: move #define FEATURE flags into new mfeatures.h file Also, check the FEATURE flags in many places. --- src/mesa/main/api_exec.c | 140 ++++++++++++++++++++++++++++++-------------- src/mesa/main/api_exec.h | 5 +- src/mesa/main/api_noop.c | 9 +++ src/mesa/main/config.h | 38 ++---------- src/mesa/main/context.c | 50 +++++++++++++++- src/mesa/main/mfeatures.h | 77 ++++++++++++++++++++++++ src/mesa/main/state.c | 4 ++ src/mesa/main/texformat.c | 26 +++++--- src/mesa/main/teximage.c | 21 ++++++- src/mesa/main/texobj.c | 4 ++ src/mesa/main/texstate.c | 15 +++-- src/mesa/main/texstore.c | 17 +++++- src/mesa/vbo/vbo_context.c | 5 +- src/mesa/vbo/vbo_context.h | 6 +- src/mesa/vbo/vbo_exec.c | 1 - src/mesa/vbo/vbo_exec_api.c | 4 ++ 16 files changed, 323 insertions(+), 99 deletions(-) create mode 100644 src/mesa/main/mfeatures.h (limited to 'src') diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c index 8cdbaada9b..f7cf42f600 100644 --- a/src/mesa/main/api_exec.c +++ b/src/mesa/main/api_exec.c @@ -30,7 +30,9 @@ #include "glheader.h" +#if FEATURE_accum #include "accum.h" +#endif #include "api_loopback.h" #include "api_exec.h" #if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program @@ -39,25 +41,42 @@ #if FEATURE_ATI_fragment_shader #include "shader/atifragshader.h" #endif +#if FEATURE_attrib_stack #include "attrib.h" +#endif #include "blend.h" #if FEATURE_ARB_vertex_buffer_object #include "bufferobj.h" #endif #include "arrayobj.h" +#if FEATURE_draw_read_buffer #include "buffers.h" +#endif #include "clear.h" #include "clip.h" +#if FEATURE_colortable #include "colortab.h" +#endif #include "context.h" +#if FEATURE_convolution #include "convolve.h" +#endif #include "depth.h" +#if FEATURE_dlist #include "dlist.h" +#endif +#if FEATURE_drawpix #include "drawpix.h" +#include "rastpos.h" +#endif #include "enable.h" +#if FEATURE_evaluators #include "eval.h" +#endif #include "get.h" +#if FEATURE_feadback #include "feedback.h" +#endif #include "fog.h" #if FEATURE_EXT_framebuffer_object #include "fbobject.h" @@ -65,21 +84,24 @@ #include "ffvertex_prog.h" #include "framebuffer.h" #include "hint.h" +#if FEATURE_histogram #include "histogram.h" +#endif #include "imports.h" #include "light.h" #include "lines.h" #include "macros.h" #include "matrix.h" #include "multisample.h" +#if FEATURE_pixel_transfer #include "pixel.h" +#endif #include "pixelstore.h" #include "points.h" #include "polygon.h" #if FEATURE_ARB_occlusion_query || FEATURE_EXT_timer_query #include "queryobj.h" #endif -#include "rastpos.h" #include "readpix.h" #include "scissor.h" #include "state.h" @@ -131,7 +153,10 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_ColorMask(exec, _mesa_ColorMask); SET_CullFace(exec, _mesa_CullFace); SET_Disable(exec, _mesa_Disable); +#if FEATURE_draw_read_buffer SET_DrawBuffer(exec, _mesa_DrawBuffer); + SET_ReadBuffer(exec, _mesa_ReadBuffer); +#endif SET_Enable(exec, _mesa_Enable); SET_Finish(exec, _mesa_Finish); SET_Flush(exec, _mesa_Flush); @@ -140,31 +165,20 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_GetError(exec, _mesa_GetError); SET_GetFloatv(exec, _mesa_GetFloatv); SET_GetString(exec, _mesa_GetString); - SET_InitNames(exec, _mesa_InitNames); SET_LineStipple(exec, _mesa_LineStipple); SET_LineWidth(exec, _mesa_LineWidth); SET_LoadIdentity(exec, _mesa_LoadIdentity); SET_LoadMatrixf(exec, _mesa_LoadMatrixf); - SET_LoadName(exec, _mesa_LoadName); SET_LogicOp(exec, _mesa_LogicOp); SET_MatrixMode(exec, _mesa_MatrixMode); SET_MultMatrixf(exec, _mesa_MultMatrixf); SET_Ortho(exec, _mesa_Ortho); SET_PixelStorei(exec, _mesa_PixelStorei); SET_PopMatrix(exec, _mesa_PopMatrix); - SET_PopName(exec, _mesa_PopName); SET_PushMatrix(exec, _mesa_PushMatrix); - SET_PushName(exec, _mesa_PushName); - SET_RasterPos2f(exec, _mesa_RasterPos2f); - SET_RasterPos2fv(exec, _mesa_RasterPos2fv); - SET_RasterPos2i(exec, _mesa_RasterPos2i); - SET_RasterPos2iv(exec, _mesa_RasterPos2iv); - SET_ReadBuffer(exec, _mesa_ReadBuffer); - SET_RenderMode(exec, _mesa_RenderMode); SET_Rotatef(exec, _mesa_Rotatef); SET_Scalef(exec, _mesa_Scalef); SET_Scissor(exec, _mesa_Scissor); - SET_SelectBuffer(exec, _mesa_SelectBuffer); SET_ShadeModel(exec, _mesa_ShadeModel); SET_StencilFunc(exec, _mesa_StencilFunc); SET_StencilMask(exec, _mesa_StencilMask); @@ -175,46 +189,57 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_TexParameteri(exec, _mesa_TexParameteri); SET_Translatef(exec, _mesa_Translatef); SET_Viewport(exec, _mesa_Viewport); -#if _HAVE_FULL_GL +#if FEATURE_accum SET_Accum(exec, _mesa_Accum); - SET_Bitmap(exec, _mesa_Bitmap); + SET_ClearAccum(exec, _mesa_ClearAccum); +#endif +#if FEATURE_dlist SET_CallList(exec, _mesa_CallList); SET_CallLists(exec, _mesa_CallLists); - SET_ClearAccum(exec, _mesa_ClearAccum); + SET_DeleteLists(exec, _mesa_DeleteLists); + SET_EndList(exec, _mesa_EndList); + SET_GenLists(exec, _mesa_GenLists); + SET_IsList(exec, _mesa_IsList); + SET_ListBase(exec, _mesa_ListBase); + SET_NewList(exec, _mesa_NewList); +#endif SET_ClearDepth(exec, _mesa_ClearDepth); SET_ClearIndex(exec, _mesa_ClearIndex); SET_ClipPlane(exec, _mesa_ClipPlane); SET_ColorMaterial(exec, _mesa_ColorMaterial); - SET_CopyPixels(exec, _mesa_CopyPixels); SET_CullParameterfvEXT(exec, _mesa_CullParameterfvEXT); SET_CullParameterdvEXT(exec, _mesa_CullParameterdvEXT); - SET_DeleteLists(exec, _mesa_DeleteLists); SET_DepthFunc(exec, _mesa_DepthFunc); SET_DepthMask(exec, _mesa_DepthMask); SET_DepthRange(exec, _mesa_DepthRange); +#if FEATURE_drawpix + SET_Bitmap(exec, _mesa_Bitmap); + SET_CopyPixels(exec, _mesa_CopyPixels); SET_DrawPixels(exec, _mesa_DrawPixels); - SET_EndList(exec, _mesa_EndList); +#endif +#if FEATURE_feadback + SET_InitNames(exec, _mesa_InitNames); SET_FeedbackBuffer(exec, _mesa_FeedbackBuffer); + SET_LoadName(exec, _mesa_LoadName); + SET_PassThrough(exec, _mesa_PassThrough); + SET_PopName(exec, _mesa_PopName); + SET_PushName(exec, _mesa_PushName); + SET_SelectBuffer(exec, _mesa_SelectBuffer); + SET_RenderMode(exec, _mesa_RenderMode); +#endif SET_FogCoordPointerEXT(exec, _mesa_FogCoordPointerEXT); SET_Fogf(exec, _mesa_Fogf); SET_Fogfv(exec, _mesa_Fogfv); SET_Fogi(exec, _mesa_Fogi); SET_Fogiv(exec, _mesa_Fogiv); - SET_GenLists(exec, _mesa_GenLists); SET_GetClipPlane(exec, _mesa_GetClipPlane); SET_GetBooleanv(exec, _mesa_GetBooleanv); SET_GetDoublev(exec, _mesa_GetDoublev); SET_GetIntegerv(exec, _mesa_GetIntegerv); SET_GetLightfv(exec, _mesa_GetLightfv); SET_GetLightiv(exec, _mesa_GetLightiv); - SET_GetMapdv(exec, _mesa_GetMapdv); - SET_GetMapfv(exec, _mesa_GetMapfv); - SET_GetMapiv(exec, _mesa_GetMapiv); SET_GetMaterialfv(exec, _mesa_GetMaterialfv); SET_GetMaterialiv(exec, _mesa_GetMaterialiv); - SET_GetPixelMapfv(exec, _mesa_GetPixelMapfv); - SET_GetPixelMapuiv(exec, _mesa_GetPixelMapuiv); - SET_GetPixelMapusv(exec, _mesa_GetPixelMapusv); SET_GetPolygonStipple(exec, _mesa_GetPolygonStipple); SET_GetTexEnvfv(exec, _mesa_GetTexEnvfv); SET_GetTexEnviv(exec, _mesa_GetTexEnviv); @@ -222,14 +247,10 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_GetTexLevelParameteriv(exec, _mesa_GetTexLevelParameteriv); SET_GetTexParameterfv(exec, _mesa_GetTexParameterfv); SET_GetTexParameteriv(exec, _mesa_GetTexParameteriv); - SET_GetTexGendv(exec, _mesa_GetTexGendv); - SET_GetTexGenfv(exec, _mesa_GetTexGenfv); - SET_GetTexGeniv(exec, _mesa_GetTexGeniv); SET_GetTexImage(exec, _mesa_GetTexImage); SET_Hint(exec, _mesa_Hint); SET_IndexMask(exec, _mesa_IndexMask); SET_IsEnabled(exec, _mesa_IsEnabled); - SET_IsList(exec, _mesa_IsList); SET_LightModelf(exec, _mesa_LightModelf); SET_LightModelfv(exec, _mesa_LightModelfv); SET_LightModeli(exec, _mesa_LightModeli); @@ -238,8 +259,11 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_Lightfv(exec, _mesa_Lightfv); SET_Lighti(exec, _mesa_Lighti); SET_Lightiv(exec, _mesa_Lightiv); - SET_ListBase(exec, _mesa_ListBase); SET_LoadMatrixd(exec, _mesa_LoadMatrixd); +#if FEATURE_evaluators + SET_GetMapdv(exec, _mesa_GetMapdv); + SET_GetMapfv(exec, _mesa_GetMapfv); + SET_GetMapiv(exec, _mesa_GetMapiv); SET_Map1d(exec, _mesa_Map1d); SET_Map1f(exec, _mesa_Map1f); SET_Map2d(exec, _mesa_Map2d); @@ -248,22 +272,35 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_MapGrid1f(exec, _mesa_MapGrid1f); SET_MapGrid2d(exec, _mesa_MapGrid2d); SET_MapGrid2f(exec, _mesa_MapGrid2f); +#endif SET_MultMatrixd(exec, _mesa_MultMatrixd); - SET_NewList(exec, _mesa_NewList); - SET_PassThrough(exec, _mesa_PassThrough); +#if FEATURE_pixel_transfer + SET_GetPixelMapfv(exec, _mesa_GetPixelMapfv); + SET_GetPixelMapuiv(exec, _mesa_GetPixelMapuiv); + SET_GetPixelMapusv(exec, _mesa_GetPixelMapusv); SET_PixelMapfv(exec, _mesa_PixelMapfv); SET_PixelMapuiv(exec, _mesa_PixelMapuiv); SET_PixelMapusv(exec, _mesa_PixelMapusv); - SET_PixelStoref(exec, _mesa_PixelStoref); SET_PixelTransferf(exec, _mesa_PixelTransferf); SET_PixelTransferi(exec, _mesa_PixelTransferi); SET_PixelZoom(exec, _mesa_PixelZoom); +#endif + SET_PixelStoref(exec, _mesa_PixelStoref); SET_PointSize(exec, _mesa_PointSize); SET_PolygonMode(exec, _mesa_PolygonMode); SET_PolygonOffset(exec, _mesa_PolygonOffset); SET_PolygonStipple(exec, _mesa_PolygonStipple); +#if FEATURE_attrib_stack SET_PopAttrib(exec, _mesa_PopAttrib); SET_PushAttrib(exec, _mesa_PushAttrib); + SET_PopClientAttrib(exec, _mesa_PopClientAttrib); + SET_PushClientAttrib(exec, _mesa_PushClientAttrib); +#endif +#if FEATURE_drawpix + SET_RasterPos2f(exec, _mesa_RasterPos2f); + SET_RasterPos2fv(exec, _mesa_RasterPos2fv); + SET_RasterPos2i(exec, _mesa_RasterPos2i); + SET_RasterPos2iv(exec, _mesa_RasterPos2iv); SET_RasterPos2d(exec, _mesa_RasterPos2d); SET_RasterPos2dv(exec, _mesa_RasterPos2dv); SET_RasterPos2s(exec, _mesa_RasterPos2s); @@ -284,24 +321,31 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_RasterPos4iv(exec, _mesa_RasterPos4iv); SET_RasterPos4s(exec, _mesa_RasterPos4s); SET_RasterPos4sv(exec, _mesa_RasterPos4sv); +#endif SET_ReadPixels(exec, _mesa_ReadPixels); SET_Rotated(exec, _mesa_Rotated); SET_Scaled(exec, _mesa_Scaled); SET_SecondaryColorPointerEXT(exec, _mesa_SecondaryColorPointerEXT); SET_TexEnvf(exec, _mesa_TexEnvf); SET_TexEnviv(exec, _mesa_TexEnviv); + +#if FEATURE_texgen + SET_GetTexGendv(exec, _mesa_GetTexGendv); + SET_GetTexGenfv(exec, _mesa_GetTexGenfv); + SET_GetTexGeniv(exec, _mesa_GetTexGeniv); SET_TexGend(exec, _mesa_TexGend); SET_TexGendv(exec, _mesa_TexGendv); SET_TexGenf(exec, _mesa_TexGenf); SET_TexGenfv(exec, _mesa_TexGenfv); SET_TexGeni(exec, _mesa_TexGeni); SET_TexGeniv(exec, _mesa_TexGeniv); +#endif + SET_TexImage1D(exec, _mesa_TexImage1D); SET_TexParameterf(exec, _mesa_TexParameterf); SET_TexParameterfv(exec, _mesa_TexParameterfv); SET_TexParameteriv(exec, _mesa_TexParameteriv); SET_Translated(exec, _mesa_Translated); -#endif /* 1.1 */ SET_BindTexture(exec, _mesa_BindTexture); @@ -322,9 +366,7 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_InterleavedArrays(exec, _mesa_InterleavedArrays); SET_IsTexture(exec, _mesa_IsTexture); SET_NormalPointer(exec, _mesa_NormalPointer); - SET_PopClientAttrib(exec, _mesa_PopClientAttrib); SET_PrioritizeTextures(exec, _mesa_PrioritizeTextures); - SET_PushClientAttrib(exec, _mesa_PushClientAttrib); SET_TexCoordPointer(exec, _mesa_TexCoordPointer); SET_TexSubImage1D(exec, _mesa_TexSubImage1D); SET_TexSubImage2D(exec, _mesa_TexSubImage2D); @@ -339,30 +381,37 @@ _mesa_init_exec_table(struct _glapi_table *exec) #endif /* OpenGL 1.2 GL_ARB_imaging */ -#if _HAVE_FULL_GL SET_BlendColor(exec, _mesa_BlendColor); SET_BlendEquation(exec, _mesa_BlendEquation); SET_BlendEquationSeparateEXT(exec, _mesa_BlendEquationSeparateEXT); + +#if FEATURE_colortable SET_ColorSubTable(exec, _mesa_ColorSubTable); SET_ColorTable(exec, _mesa_ColorTable); SET_ColorTableParameterfv(exec, _mesa_ColorTableParameterfv); SET_ColorTableParameteriv(exec, _mesa_ColorTableParameteriv); + SET_CopyColorSubTable(exec, _mesa_CopyColorSubTable); + SET_CopyColorTable(exec, _mesa_CopyColorTable); + SET_GetColorTable(exec, _mesa_GetColorTable); + SET_GetColorTableParameterfv(exec, _mesa_GetColorTableParameterfv); + SET_GetColorTableParameteriv(exec, _mesa_GetColorTableParameteriv); +#endif + +#if FEATURE_convolution SET_ConvolutionFilter1D(exec, _mesa_ConvolutionFilter1D); SET_ConvolutionFilter2D(exec, _mesa_ConvolutionFilter2D); SET_ConvolutionParameterf(exec, _mesa_ConvolutionParameterf); SET_ConvolutionParameterfv(exec, _mesa_ConvolutionParameterfv); SET_ConvolutionParameteri(exec, _mesa_ConvolutionParameteri); SET_ConvolutionParameteriv(exec, _mesa_ConvolutionParameteriv); - SET_CopyColorSubTable(exec, _mesa_CopyColorSubTable); - SET_CopyColorTable(exec, _mesa_CopyColorTable); SET_CopyConvolutionFilter1D(exec, _mesa_CopyConvolutionFilter1D); SET_CopyConvolutionFilter2D(exec, _mesa_CopyConvolutionFilter2D); - SET_GetColorTable(exec, _mesa_GetColorTable); - SET_GetColorTableParameterfv(exec, _mesa_GetColorTableParameterfv); - SET_GetColorTableParameteriv(exec, _mesa_GetColorTableParameteriv); SET_GetConvolutionFilter(exec, _mesa_GetConvolutionFilter); SET_GetConvolutionParameterfv(exec, _mesa_GetConvolutionParameterfv); SET_GetConvolutionParameteriv(exec, _mesa_GetConvolutionParameteriv); + SET_SeparableFilter2D(exec, _mesa_SeparableFilter2D); +#endif +#if FEATURE_histogram SET_GetHistogram(exec, _mesa_GetHistogram); SET_GetHistogramParameterfv(exec, _mesa_GetHistogramParameterfv); SET_GetHistogramParameteriv(exec, _mesa_GetHistogramParameteriv); @@ -374,7 +423,6 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_Minmax(exec, _mesa_Minmax); SET_ResetHistogram(exec, _mesa_ResetHistogram); SET_ResetMinmax(exec, _mesa_ResetMinmax); - SET_SeparableFilter2D(exec, _mesa_SeparableFilter2D); #endif /* OpenGL 2.0 */ @@ -488,7 +536,7 @@ _mesa_init_exec_table(struct _glapi_table *exec) #endif /* 197. GL_MESA_window_pos */ -#if _HAVE_FULL_GL +#if FEATURE_drawpix SET_WindowPos2dMESA(exec, _mesa_WindowPos2dMESA); SET_WindowPos2dvMESA(exec, _mesa_WindowPos2dvMESA); SET_WindowPos2fMESA(exec, _mesa_WindowPos2fMESA); @@ -715,8 +763,10 @@ _mesa_init_exec_table(struct _glapi_table *exec) #endif /* ARB 37. GL_ARB_draw_buffers */ +#if FEATURE_draw_read_buffer SET_DrawBuffersARB(exec, _mesa_DrawBuffersARB); - +#endif + #if FEATURE_ARB_shader_objects SET_DeleteObjectARB(exec, _mesa_DeleteObjectARB); SET_GetHandleARB(exec, _mesa_GetHandleARB); diff --git a/src/mesa/main/api_exec.h b/src/mesa/main/api_exec.h index 7f15ea9d00..4bd715053a 100644 --- a/src/mesa/main/api_exec.h +++ b/src/mesa/main/api_exec.h @@ -27,7 +27,10 @@ #define API_EXEC_H -void +struct _glapi_table; + + +extern void _mesa_init_exec_table(struct _glapi_table *exec); diff --git a/src/mesa/main/api_noop.c b/src/mesa/main/api_noop.c index 3df64362ea..a1cc3a2a4b 100644 --- a/src/mesa/main/api_noop.c +++ b/src/mesa/main/api_noop.c @@ -30,7 +30,9 @@ #include "context.h" #include "light.h" #include "macros.h" +#if FEATURE_dlist #include "dlist.h" +#endif #include "glapi/dispatch.h" @@ -621,6 +623,8 @@ static void GLAPIENTRY _mesa_noop_Vertex4f( GLfloat a, GLfloat b, GLfloat c, GLf (void) a; (void) b; (void) c; (void) d; } + +#if FEATURE_evaluators /* Similarly, these have no effect outside begin/end: */ static void GLAPIENTRY _mesa_noop_EvalCoord1f( GLfloat a ) @@ -652,6 +656,7 @@ static void GLAPIENTRY _mesa_noop_EvalPoint2( GLint a, GLint b ) { (void) a; (void) b; } +#endif /* FEATURE_evaluators */ /* Begin -- call into driver, should result in the vtxfmt being @@ -904,20 +909,24 @@ _mesa_noop_vtxfmt_init( GLvertexformat *vfmt ) { vfmt->ArrayElement = _ae_loopback_array_elt; /* generic helper */ vfmt->Begin = _mesa_noop_Begin; +#if FEATURE_dlist vfmt->CallList = _mesa_CallList; vfmt->CallLists = _mesa_CallLists; +#endif vfmt->Color3f = _mesa_noop_Color3f; vfmt->Color3fv = _mesa_noop_Color3fv; vfmt->Color4f = _mesa_noop_Color4f; vfmt->Color4fv = _mesa_noop_Color4fv; vfmt->EdgeFlag = _mesa_noop_EdgeFlag; vfmt->End = _mesa_noop_End; +#if FEATURE_evaluators vfmt->EvalCoord1f = _mesa_noop_EvalCoord1f; vfmt->EvalCoord1fv = _mesa_noop_EvalCoord1fv; vfmt->EvalCoord2f = _mesa_noop_EvalCoord2f; vfmt->EvalCoord2fv = _mesa_noop_EvalCoord2fv; vfmt->EvalPoint1 = _mesa_noop_EvalPoint1; vfmt->EvalPoint2 = _mesa_noop_EvalPoint2; +#endif vfmt->FogCoordfEXT = _mesa_noop_FogCoordfEXT; vfmt->FogCoordfvEXT = _mesa_noop_FogCoordfvEXT; vfmt->Indexf = _mesa_noop_Indexf; diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h index 232e698f50..9ff0b708dd 100644 --- a/src/mesa/main/config.h +++ b/src/mesa/main/config.h @@ -31,6 +31,10 @@ #ifndef MESA_CONFIG_H_INCLUDED #define MESA_CONFIG_H_INCLUDED + +#include "main/mfeatures.h" + + /** * \name OpenGL implementation limits */ @@ -283,40 +287,6 @@ #define ACOMP 3 -/* - * Enable/disable features (blocks of code) by setting FEATURE_xyz to 0 or 1. - */ -#ifndef _HAVE_FULL_GL -#define _HAVE_FULL_GL 1 -#endif - -#define FEATURE_userclip _HAVE_FULL_GL -#define FEATURE_texgen _HAVE_FULL_GL -#define FEATURE_windowpos _HAVE_FULL_GL -#define FEATURE_ARB_occlusion_query _HAVE_FULL_GL -#define FEATURE_ARB_fragment_program _HAVE_FULL_GL -#define FEATURE_ARB_vertex_buffer_object _HAVE_FULL_GL -#define FEATURE_ARB_vertex_program _HAVE_FULL_GL - -#define FEATURE_ARB_vertex_shader _HAVE_FULL_GL -#define FEATURE_ARB_fragment_shader _HAVE_FULL_GL -#define FEATURE_ARB_shader_objects (FEATURE_ARB_vertex_shader || FEATURE_ARB_fragment_shader) -#define FEATURE_ARB_shading_language_100 FEATURE_ARB_shader_objects -#define FEATURE_ARB_shading_language_120 FEATURE_ARB_shader_objects - -#define FEATURE_EXT_framebuffer_blit _HAVE_FULL_GL -#define FEATURE_EXT_framebuffer_object _HAVE_FULL_GL -#define FEATURE_EXT_pixel_buffer_object _HAVE_FULL_GL -#define FEATURE_EXT_texture_sRGB _HAVE_FULL_GL -#define FEATURE_EXT_timer_query _HAVE_FULL_GL -#define FEATURE_ATI_fragment_shader _HAVE_FULL_GL -#define FEATURE_MESA_program_debug _HAVE_FULL_GL -#define FEATURE_NV_fence _HAVE_FULL_GL -#define FEATURE_NV_fragment_program _HAVE_FULL_GL -#define FEATURE_NV_vertex_program _HAVE_FULL_GL -/*@}*/ - - /** * Maximum number of temporary vertices required for clipping. * diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 1b357ae6c2..d975814592 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -78,27 +78,41 @@ #include "glheader.h" #include "imports.h" +#if FEATURE_accum #include "accum.h" +#endif #include "api_exec.h" #include "arrayobj.h" +#if FEATURE_attrib_stack #include "attrib.h" +#endif #include "blend.h" #include "buffers.h" #include "bufferobj.h" +#if FEATURE_colortable #include "colortab.h" +#endif #include "context.h" #include "debug.h" #include "depth.h" +#if FEATURE_dlist #include "dlist.h" +#endif +#if FEATURE_evaluators #include "eval.h" +#endif #include "enums.h" #include "extensions.h" #include "fbobject.h" +#if FEATURE_feedback #include "feedback.h" +#endif #include "fog.h" #include "framebuffer.h" #include "get.h" +#if FEATURE_histogram #include "histogram.h" +#endif #include "hint.h" #include "hash.h" #include "light.h" @@ -106,12 +120,16 @@ #include "macros.h" #include "matrix.h" #include "multisample.h" +#if FEATURE_pixel_transfer #include "pixel.h" +#endif #include "pixelstore.h" #include "points.h" #include "polygon.h" #include "queryobj.h" +#if FEATURE_drawpix #include "rastpos.h" +#endif #include "scissor.h" #include "simple_list.h" #include "state.h" @@ -573,9 +591,11 @@ alloc_shared_state( GLcontext *ctx ) static void delete_displaylist_cb(GLuint id, void *data, void *userData) { +#if FEATURE_dlist struct mesa_display_list *list = (struct mesa_display_list *) data; GLcontext *ctx = (GLcontext *) userData; _mesa_delete_list(ctx, list); +#endif } /** @@ -954,31 +974,49 @@ init_attrib_groups(GLcontext *ctx) _mesa_init_extensions( ctx ); /* Attribute Groups */ +#if FEATURE_accum _mesa_init_accum( ctx ); +#endif +#if FEATURE_attrib_stack _mesa_init_attrib( ctx ); +#endif _mesa_init_buffer_objects( ctx ); _mesa_init_color( ctx ); +#if FEATURE_colortable _mesa_init_colortables( ctx ); +#endif _mesa_init_current( ctx ); _mesa_init_depth( ctx ); _mesa_init_debug( ctx ); +#if FEATURE_dlist _mesa_init_display_list( ctx ); +#endif +#if FEATURE_evaluators _mesa_init_eval( ctx ); +#endif +#if FEATURE_feedback _mesa_init_feedback( ctx ); +#endif _mesa_init_fog( ctx ); +#if FEATURE_histogram _mesa_init_histogram( ctx ); +#endif _mesa_init_hint( ctx ); _mesa_init_line( ctx ); _mesa_init_lighting( ctx ); _mesa_init_matrix( ctx ); _mesa_init_multisample( ctx ); +#if FEATURE_pixel_transfer _mesa_init_pixel( ctx ); +#endif _mesa_init_pixelstore( ctx ); _mesa_init_point( ctx ); _mesa_init_polygon( ctx ); _mesa_init_program( ctx ); _mesa_init_query( ctx ); +#if FEATURE_drawpix _mesa_init_rastpos( ctx ); +#endif _mesa_init_scissor( ctx ); _mesa_init_shader_state( ctx ); _mesa_init_stencil( ctx ); @@ -989,8 +1027,12 @@ init_attrib_groups(GLcontext *ctx) if (!_mesa_init_texture( ctx )) return GL_FALSE; +#if FEATURE_texture_s3tc _mesa_init_texture_s3tc( ctx ); +#endif +#if FEATURE_texture_fxt1 _mesa_init_texture_fxt1( ctx ); +#endif /* Miscellaneous */ ctx->NewState = _NEW_ALL; @@ -1123,14 +1165,14 @@ _mesa_initialize_context(GLcontext *ctx, } _mesa_init_exec_table(ctx->Exec); ctx->CurrentDispatch = ctx->Exec; -#if _HAVE_FULL_GL +#if FEATURE_dlist _mesa_init_dlist_table(ctx->Save); _mesa_install_save_vtxfmt( ctx, &ctx->ListState.ListVtxfmt ); +#endif /* Neutral tnl module stuff */ _mesa_init_exec_vtxfmt( ctx ); ctx->TnlModule.Current = NULL; ctx->TnlModule.SwapCount = 0; -#endif ctx->FragmentProgram._MaintainTexEnvProgram = (_mesa_getenv("MESA_TEX_PROG") != NULL); @@ -1220,11 +1262,15 @@ _mesa_free_context_data( GLcontext *ctx ) _mesa_reference_fragprog(ctx, &ctx->FragmentProgram._TexEnvProgram, NULL); _mesa_free_lighting_data( ctx ); +#if FEATURE_evaluators _mesa_free_eval_data( ctx ); +#endif _mesa_free_texture_data( ctx ); _mesa_free_matrix_data( ctx ); _mesa_free_viewport_data( ctx ); +#if FEATURE_colortable _mesa_free_colortables_data( ctx ); +#endif _mesa_free_program_data(ctx); _mesa_free_shader_state(ctx); _mesa_free_query_data(ctx); diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h new file mode 100644 index 0000000000..228de76af1 --- /dev/null +++ b/src/mesa/main/mfeatures.h @@ -0,0 +1,77 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/** + * \file mfeatures.h + * Flags to enable/disable specific parts of the API. + */ + +#ifndef FEATURES_H +#define FEATURES_H + + +#ifndef _HAVE_FULL_GL +#define _HAVE_FULL_GL 1 +#endif + +#define FEATURE_accum _HAVE_FULL_GL +#define FEATURE_attrib_stacks _HAVE_FULL_GL +#define FEATURE_colortable _HAVE_FULL_GL +#define FEATURE_convolution _HAVE_FULL_GL +#define FEATURE_dlist _HAVE_FULL_GL +#define FEATURE_draw_read_buffer _HAVE_FULL_GL +#define FEATURE_drawpix _HAVE_FULL_GL +#define FEATURE_evaluators _HAVE_FULL_GL +#define FEATURE_feedback _HAVE_FULL_GL +#define FEATURE_histogram _HAVE_FULL_GL +#define FEATURE_pixeltransfer _HAVE_FULL_GL +#define FEATURE_texgen _HAVE_FULL_GL +#define FEATURE_texture_fxt1 _HAVE_FULL_GL +#define FEATURE_texture_s3tc _HAVE_FULL_GL +#define FEATURE_userclip _HAVE_FULL_GL + +#define FEATURE_ARB_occlusion_query _HAVE_FULL_GL +#define FEATURE_ARB_fragment_program _HAVE_FULL_GL +#define FEATURE_ARB_vertex_buffer_object _HAVE_FULL_GL +#define FEATURE_ARB_vertex_program _HAVE_FULL_GL +#define FEATURE_ARB_vertex_shader _HAVE_FULL_GL +#define FEATURE_ARB_fragment_shader _HAVE_FULL_GL +#define FEATURE_ARB_shader_objects (FEATURE_ARB_vertex_shader || FEATURE_ARB_fragment_shader) +#define FEATURE_ARB_shading_language_100 FEATURE_ARB_shader_objects +#define FEATURE_ARB_shading_language_120 FEATURE_ARB_shader_objects + +#define FEATURE_EXT_framebuffer_blit _HAVE_FULL_GL +#define FEATURE_EXT_framebuffer_object _HAVE_FULL_GL +#define FEATURE_EXT_pixel_buffer_object _HAVE_FULL_GL +#define FEATURE_EXT_texture_sRGB _HAVE_FULL_GL +#define FEATURE_EXT_timer_query _HAVE_FULL_GL +#define FEATURE_ATI_fragment_shader _HAVE_FULL_GL +#define FEATURE_MESA_program_debug _HAVE_FULL_GL +#define FEATURE_NV_fence _HAVE_FULL_GL +#define FEATURE_NV_fragment_program _HAVE_FULL_GL +#define FEATURE_NV_vertex_program _HAVE_FULL_GL + + +#endif /* FEATURES_H */ diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 89920f8443..4d1fdbf47c 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -40,7 +40,9 @@ #include "framebuffer.h" #include "light.h" #include "matrix.h" +#if FEATURE_pixel_transfer #include "pixel.h" +#endif #include "shader/program.h" #include "state.h" #include "stencil.h" @@ -412,8 +414,10 @@ _mesa_update_state_locked( GLcontext *ctx ) if (new_state & _NEW_STENCIL) _mesa_update_stencil( ctx ); +#if FEATURE_pixel_transfer if (new_state & _IMAGE_NEW_TRANSFER_STATE) _mesa_update_pixel( ctx, new_state ); +#endif if (new_state & (_NEW_ARRAY | _NEW_PROGRAM)) update_arrays( ctx ); diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c index 17859a9fec..a35195a695 100644 --- a/src/mesa/main/texformat.c +++ b/src/mesa/main/texformat.c @@ -1449,21 +1449,27 @@ _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat, case GL_COMPRESSED_INTENSITY_ARB: return &_mesa_texformat_intensity; case GL_COMPRESSED_RGB_ARB: +#if FEATURE_texture_fxt1 if (ctx->Extensions.TDFX_texture_compression_FXT1) return &_mesa_texformat_rgb_fxt1; - else if (ctx->Extensions.EXT_texture_compression_s3tc || - ctx->Extensions.S3_s3tc) +#endif +#if FEATURE_texture_s3tc + if (ctx->Extensions.EXT_texture_compression_s3tc || + ctx->Extensions.S3_s3tc) return &_mesa_texformat_rgb_dxt1; - else - return &_mesa_texformat_rgb; +#endif + return &_mesa_texformat_rgb; case GL_COMPRESSED_RGBA_ARB: +#if FEATURE_texture_fxt1 if (ctx->Extensions.TDFX_texture_compression_FXT1) return &_mesa_texformat_rgba_fxt1; - else if (ctx->Extensions.EXT_texture_compression_s3tc || - ctx->Extensions.S3_s3tc) +#endif +#if FEATURE_texture_s3tc + if (ctx->Extensions.EXT_texture_compression_s3tc || + ctx->Extensions.S3_s3tc) return &_mesa_texformat_rgba_dxt3; /* Not rgba_dxt1, see spec */ - else - return &_mesa_texformat_rgba; +#endif + return &_mesa_texformat_rgba; default: ; /* fallthrough */ } @@ -1478,6 +1484,7 @@ _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat, } } +#if FEATURE_texture_fxt1 if (ctx->Extensions.TDFX_texture_compression_FXT1) { switch (internalFormat) { case GL_COMPRESSED_RGB_FXT1_3DFX: @@ -1488,7 +1495,9 @@ _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat, ; /* fallthrough */ } } +#endif +#if FEATURE_texture_s3tc if (ctx->Extensions.EXT_texture_compression_s3tc) { switch (internalFormat) { case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: @@ -1516,6 +1525,7 @@ _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat, ; /* fallthrough */ } } +#endif if (ctx->Extensions.ARB_texture_float) { switch (internalFormat) { diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c index 384e145520..eed1937517 100644 --- a/src/mesa/main/teximage.c +++ b/src/mesa/main/teximage.c @@ -32,7 +32,9 @@ #include "glheader.h" #include "bufferobj.h" #include "context.h" +#if FEATURE_convolve #include "convolve.h" +#endif #include "fbobject.h" #include "framebuffer.h" #include "image.h" @@ -2429,9 +2431,11 @@ _mesa_TexImage1D( GLenum target, GLint level, GLint internalFormat, GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); +#if FEATURE_convolve if (is_color_format(internalFormat)) { _mesa_adjust_image_for_convolution(ctx, 1, &postConvWidth, NULL); } +#endif if (target == GL_TEXTURE_1D) { /* non-proxy target */ @@ -2524,10 +2528,12 @@ _mesa_TexImage2D( GLenum target, GLint level, GLint internalFormat, GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); +#if FEATURE_convolve if (is_color_format(internalFormat)) { _mesa_adjust_image_for_convolution(ctx, 2, &postConvWidth, &postConvHeight); } +#endif if (target == GL_TEXTURE_2D || (ctx->Extensions.ARB_texture_cube_map && @@ -2746,10 +2752,12 @@ _mesa_TexSubImage1D( GLenum target, GLint level, if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE) _mesa_update_state(ctx); +#if FEATURE_convolve /* XXX should test internal format */ if (is_color_format(format)) { _mesa_adjust_image_for_convolution(ctx, 1, &postConvWidth, NULL); } +#endif if (subtexture_error_check(ctx, 1, target, level, xoffset, 0, 0, postConvWidth, 1, 1, format, type)) { @@ -2804,11 +2812,13 @@ _mesa_TexSubImage2D( GLenum target, GLint level, if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE) _mesa_update_state(ctx); +#if FEATURE_convolve /* XXX should test internal format */ if (is_color_format(format)) { _mesa_adjust_image_for_convolution(ctx, 2, &postConvWidth, &postConvHeight); } +#endif if (subtexture_error_check(ctx, 2, target, level, xoffset, yoffset, 0, postConvWidth, postConvHeight, 1, format, type)) { @@ -2918,9 +2928,11 @@ _mesa_CopyTexImage1D( GLenum target, GLint level, if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE) _mesa_update_state(ctx); +#if FEATURE_convolve if (is_color_format(internalFormat)) { _mesa_adjust_image_for_convolution(ctx, 1, &postConvWidth, NULL); } +#endif if (copytexture_error_check(ctx, 1, target, level, internalFormat, postConvWidth, 1, border)) @@ -2981,11 +2993,12 @@ _mesa_CopyTexImage2D( GLenum target, GLint level, GLenum internalFormat, if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE) _mesa_update_state(ctx); +#if FEATURE_convolve if (is_color_format(internalFormat)) { _mesa_adjust_image_for_convolution(ctx, 2, &postConvWidth, &postConvHeight); } - +#endif if (copytexture_error_check(ctx, 2, target, level, internalFormat, postConvWidth, postConvHeight, border)) return; @@ -3047,8 +3060,10 @@ _mesa_CopyTexSubImage1D( GLenum target, GLint level, if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE) _mesa_update_state(ctx); +#if FEATURE_convolve /* XXX should test internal format */ _mesa_adjust_image_for_convolution(ctx, 1, &postConvWidth, NULL); +#endif if (copytexsubimage_error_check(ctx, 1, target, level, xoffset, 0, 0, postConvWidth, 1)) @@ -3100,8 +3115,10 @@ _mesa_CopyTexSubImage2D( GLenum target, GLint level, if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE) _mesa_update_state(ctx); +#if FEATURE_convolve /* XXX should test internal format */ _mesa_adjust_image_for_convolution(ctx, 2, &postConvWidth, &postConvHeight); +#endif if (copytexsubimage_error_check(ctx, 2, target, level, xoffset, yoffset, 0, postConvWidth, postConvHeight)) @@ -3152,8 +3169,10 @@ _mesa_CopyTexSubImage3D( GLenum target, GLint level, if (ctx->NewState & _IMAGE_NEW_TRANSFER_STATE) _mesa_update_state(ctx); +#if FEATURE_convolve /* XXX should test internal format */ _mesa_adjust_image_for_convolution(ctx, 2, &postConvWidth, &postConvHeight); +#endif if (copytexsubimage_error_check(ctx, 3, target, level, xoffset, yoffset, zoffset, postConvWidth, postConvHeight)) diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c index df64002f99..606e62d7a0 100644 --- a/src/mesa/main/texobj.c +++ b/src/mesa/main/texobj.c @@ -29,7 +29,9 @@ #include "glheader.h" +#if FEATURE_colortable #include "colortab.h" +#endif #include "context.h" #include "enums.h" #include "fbobject.h" @@ -153,7 +155,9 @@ _mesa_delete_texture_object( GLcontext *ctx, struct gl_texture_object *texObj ) (void) ctx; +#if FEATURE_colortable _mesa_free_colortable_data(&texObj->Palette); +#endif /* free the texture images */ for (face = 0; face < 6; face++) { diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c index 84acfbd92c..2a83d6df4a 100644 --- a/src/mesa/main/texstate.c +++ b/src/mesa/main/texstate.c @@ -30,7 +30,9 @@ #include "glheader.h" #include "colormac.h" +#if FEATURE_colortable #include "colortab.h" +#endif #include "context.h" #include "enums.h" #include "macros.h" @@ -3213,7 +3215,9 @@ _mesa_init_texture(GLcontext *ctx) for (i=0; iTexture.SharedPalette = GL_FALSE; +#if FEATURE_colortable _mesa_init_colortable(&ctx->Texture.Palette); +#endif /* Allocate proxy textures */ if (!alloc_proxy_textures( ctx )) @@ -3229,8 +3233,6 @@ _mesa_init_texture(GLcontext *ctx) void _mesa_free_texture_data(GLcontext *ctx) { - GLuint i; - /* Free proxy texture objects */ (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy1D ); (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy2D ); @@ -3240,6 +3242,11 @@ _mesa_free_texture_data(GLcontext *ctx) (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy1DArray ); (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy2DArray ); - for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++) - _mesa_free_colortable_data( &ctx->Texture.Unit[i].ColorTable ); +#if FEATURE_colortable + { + GLuint i; + for (i = 0; i < MAX_TEXTURE_IMAGE_UNITS; i++) + _mesa_free_colortable_data( &ctx->Texture.Unit[i].ColorTable ); + } +#endif } diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index a6a18910fc..5ac6116950 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -55,7 +55,9 @@ #include "bufferobj.h" #include "colormac.h" #include "context.h" +#if FEATURE_convolve #include "convolve.h" +#endif #include "image.h" #include "macros.h" #include "mipmap.h" @@ -269,6 +271,16 @@ compute_component_mapping(GLenum inFormat, GLenum outFormat, } +#if !FEATURE_convolve +static void +_mesa_adjust_image_for_convolution(GLcontext *ctx, GLuint dims, + GLsizei *srcWidth, GLsizei *srcHeight) +{ + /* no-op */ +} +#endif + + /** * Make a temporary (color) texture image with GLfloat components. * Apply all needed pixel unpacking and pixel transfer operations. @@ -368,6 +380,7 @@ make_temp_float_image(GLcontext *ctx, GLuint dims, dst += srcWidth * 4; } +#if FEATURE_convolve /* do convolution */ { GLfloat *src = tempImage + img * (srcWidth * srcHeight * 4); @@ -389,7 +402,7 @@ make_temp_float_image(GLcontext *ctx, GLuint dims, } } } - +#endif /* do post-convolution transfer and pack into tempImage */ { const GLint logComponents @@ -546,6 +559,7 @@ _mesa_make_temp_chan_image(GLcontext *ctx, GLuint dims, textureBaseFormat == GL_ALPHA || textureBaseFormat == GL_INTENSITY); +#if FEATURE_convolve if ((dims == 1 && ctx->Pixel.Convolution1DEnabled) || (dims >= 2 && ctx->Pixel.Convolution2DEnabled) || (dims >= 2 && ctx->Pixel.Separable2DEnabled)) { @@ -567,6 +581,7 @@ _mesa_make_temp_chan_image(GLcontext *ctx, GLuint dims, transferOps = 0; freeSrcImage = GL_TRUE; } +#endif /* unpack and transfer the source image */ tempImage = (GLchan *) _mesa_malloc(srcWidth * srcHeight * srcDepth diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c index 235cee2429..dc7c534251 100644 --- a/src/mesa/vbo/vbo_context.c +++ b/src/mesa/vbo/vbo_context.c @@ -225,9 +225,10 @@ GLboolean _vbo_CreateContext( GLcontext *ctx ) * vtxfmt mechanism can be removed now. */ vbo_exec_init( ctx ); +#if FEATURE_dlist vbo_save_init( ctx ); +#endif - return GL_TRUE; } @@ -246,7 +247,9 @@ void _vbo_DestroyContext( GLcontext *ctx ) } vbo_exec_destroy(ctx); +#if FEATURE_dlist vbo_save_destroy(ctx); +#endif FREE(vbo_context(ctx)); ctx->swtnl_im = NULL; } diff --git a/src/mesa/vbo/vbo_context.h b/src/mesa/vbo/vbo_context.h index 013f81bdd5..bf405eb693 100644 --- a/src/mesa/vbo/vbo_context.h +++ b/src/mesa/vbo/vbo_context.h @@ -53,8 +53,10 @@ #include "vbo.h" #include "vbo_attrib.h" -#include "vbo_save.h" #include "vbo_exec.h" +#if FEATURE_dlist +#include "vbo_save.h" +#endif struct vbo_context { @@ -74,7 +76,9 @@ struct vbo_context { struct vbo_exec_context exec; +#if FEATURE_dlist struct vbo_save_context save; +#endif /* Callback into the driver. This must always succeed, the driver * is responsible for initiating any fallback actions required: diff --git a/src/mesa/vbo/vbo_exec.c b/src/mesa/vbo/vbo_exec.c index 1efa74945d..635f239acc 100644 --- a/src/mesa/vbo/vbo_exec.c +++ b/src/mesa/vbo/vbo_exec.c @@ -32,7 +32,6 @@ #include "main/context.h" #include "main/macros.h" #include "main/mtypes.h" -#include "main/dlist.h" #include "main/vtxfmt.h" #include "vbo_context.h" diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index 98580170e3..5dd774d839 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -34,7 +34,9 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "main/context.h" #include "main/macros.h" #include "main/vtxfmt.h" +#if FEATURE_dlist #include "main/dlist.h" +#endif #include "main/state.h" #include "main/light.h" #include "main/api_arrayelt.h" @@ -569,8 +571,10 @@ static void vbo_exec_vtxfmt_init( struct vbo_exec_context *exec ) vfmt->ArrayElement = _ae_loopback_array_elt; /* generic helper */ vfmt->Begin = vbo_exec_Begin; +#if FEATURE_dlist vfmt->CallList = _mesa_CallList; vfmt->CallLists = _mesa_CallLists; +#endif vfmt->End = vbo_exec_End; vfmt->EvalCoord1f = vbo_exec_EvalCoord1f; vfmt->EvalCoord1fv = vbo_exec_EvalCoord1fv; -- cgit v1.2.3 From b9c5b0bc4b0abbe12770824371c173ad1e8d1046 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 10 Jun 2008 16:14:09 -0600 Subject: egl: some prototype API detection code --- src/egl/main/eglmisc.c | 52 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 52 insertions(+) (limited to 'src') diff --git a/src/egl/main/eglmisc.c b/src/egl/main/eglmisc.c index b5bdc3ea4b..e79223657d 100644 --- a/src/egl/main/eglmisc.c +++ b/src/egl/main/eglmisc.c @@ -54,6 +54,58 @@ _eglUpdateExtensionsString(_EGLDriver *drv) } +#if 0 /* prototype code */ + +#include + +static EGLint +_eglFindAPIs_by_sym(void) +{ + EGLint mask = 0x0; + + if (dlsym(NULL, "glBegin")) + mask |= EGL_OPENGL_BIT; + if (dlsym(NULL, "glGetFixedv")) + mask |= EGL_OPENGL_ES_BIT; + if (dlsym(NULL, "vgSetf")) + mask |= EGL_OPENVG_BIT; + + return mask; +} + +static EGLint +_eglFindAPIs_by_lib(void) +{ + EGLint mask = 0x0; + int flag = RTLD_NOW; + void *h; + + if ((h = dlopen("libGLESv1_CM.so", flag))) { + dlclose(h); + mask |= EGL_OPENGL_ES_BIT; + } + + if ((h = dlopen("libGLESv2.so", flag))) { + dlclose(h); + mask |= EGL_OPENGL_ES2_BIT; + } + + if ((h = dlopen("libGL.so", flag))) { + dlclose(h); + mask |= EGL_OPENGL_BIT; + } + + if ((h = dlopen("libOpenVG.so", flag))) { + dlclose(h); + mask |= EGL_OPENVG_BIT; + } + + return mask; +} + +#endif + + static void _eglUpdateAPIsString(_EGLDriver *drv) { -- cgit v1.2.3 From e9b6ed395a69be90cbf1e2481a9da6fa4079e88e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 10 Jun 2008 16:14:35 -0600 Subject: egl: only windows are renderable at this time --- src/egl/drivers/xdri/egl_xdri.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c index 4c8511fbbf..60753a1c1f 100644 --- a/src/egl/drivers/xdri/egl_xdri.c +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -183,6 +183,9 @@ create_configs(_EGLDisplay *disp) SET_CONFIG_ATTRIB(&config->Base, EGL_STENCIL_SIZE, m->stencilBits); SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_ID, m->visualID); SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_TYPE, m->visualType); + /* XXX only window rendering allowed ATM */ + SET_CONFIG_ATTRIB(&config->Base, EGL_SURFACE_TYPE, EGL_WINDOW_BIT); + /* XXX possibly other things to init... */ /* Ptr from EGL config to GLcontextMode. Used in CreateContext(). */ -- cgit v1.2.3 From 0f36a29d5e6a50240e2c1469af68b7ac556afdc9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 10 Jun 2008 16:26:23 -0600 Subject: gallium: work-around glapi.c build problem for dri configs --- src/mesa/Makefile | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 5f84f72bb2..affd81b828 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -19,9 +19,17 @@ include sources +# XXX work-around glapi.c build failure for DRI builds: +ifeq ($(WINDOW_SYSTEM), dri) +GLAPI_LIB = +else +GLAPI_LIB = libglapi.a +endif + + # Default: build dependencies, then asm_subdirs, then convenience # libs (.a) and finally the device drivers: -default: depend asm_subdirs libmesa.a libglapi.a driver_subdirs +default: depend asm_subdirs libmesa.a $(GLAPI_LIB) driver_subdirs @@ -33,13 +41,13 @@ libmesa.a: $(MESA_OBJECTS) asm_subdirs @ $(TOP)/bin/mklib -o mesa -static $(MESA_OBJECTS) # Make archive of gl* API dispatcher functions only -libglapi.a: $(GLAPI_OBJECTS) asm_subdirs +$(GLAPI_LIB): $(GLAPI_OBJECTS) asm_subdirs @ $(TOP)/bin/mklib -o glapi -static $(GLAPI_OBJECTS) ###################################################################### # Device drivers -driver_subdirs: libmesa.a libglapi.a +driver_subdirs: libmesa.a $(GLAPI_LIB) (cd drivers && $(MAKE)) @@ -126,7 +134,7 @@ tags: clean: -rm -f */*.o -rm -f */*/*.o - -rm -f depend depend.bak libmesa.a libglapi.a + -rm -f depend depend.bak libmesa.a $(GLAPI_LIB) -rm -f drivers/*/*.o -@cd drivers/dri && $(MAKE) clean -@cd drivers/xorg && $(MAKE) clean -- cgit v1.2.3 From e4cfe0854ad968193106048179b9b52ec1768f41 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 10 Jun 2008 16:43:49 -0600 Subject: mesa: refactor: fix some FEATURE_ typos, mistakes --- src/mesa/main/mfeatures.h | 2 +- src/mesa/main/rastpos.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h index 228de76af1..162bd2fff9 100644 --- a/src/mesa/main/mfeatures.h +++ b/src/mesa/main/mfeatures.h @@ -46,7 +46,7 @@ #define FEATURE_evaluators _HAVE_FULL_GL #define FEATURE_feedback _HAVE_FULL_GL #define FEATURE_histogram _HAVE_FULL_GL -#define FEATURE_pixeltransfer _HAVE_FULL_GL +#define FEATURE_pixel_transfer _HAVE_FULL_GL #define FEATURE_texgen _HAVE_FULL_GL #define FEATURE_texture_fxt1 _HAVE_FULL_GL #define FEATURE_texture_s3tc _HAVE_FULL_GL diff --git a/src/mesa/main/rastpos.c b/src/mesa/main/rastpos.c index f0500083ec..155140f3cc 100644 --- a/src/mesa/main/rastpos.c +++ b/src/mesa/main/rastpos.c @@ -211,7 +211,7 @@ _mesa_RasterPos4sv(const GLshort *v) /*** GL_ARB_window_pos / GL_MESA_window_pos ***/ /**********************************************************************/ -#if FEATURE_windowpos +#if FEATURE_drawpix /** * All glWindowPosMESA and glWindowPosARB commands call this function to * update the current raster position. -- cgit v1.2.3 From ab399b555c99c46958c421d900109f78901ddc99 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 11 Jun 2008 09:15:54 +0100 Subject: draw: remove debug assert on failover to generic vs varient --- src/gallium/auxiliary/draw/draw_vs_aos.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 458bbbe376..c47647ea72 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -2108,7 +2108,6 @@ struct draw_vs_varient *draw_vs_varient_aos_sse( struct draw_vertex_shader *vs, struct draw_vs_varient *varient = varient_aos_sse( vs, key ); if (varient == NULL) { - assert(0); varient = draw_vs_varient_generic( vs, key ); } -- cgit v1.2.3 From bd9264210097d08073a4ea3619ca25db56245280 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 11 Jun 2008 09:36:00 +0100 Subject: draw: remove another debug assert on failover to generic vs varient --- src/gallium/auxiliary/draw/draw_vs_aos.h | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index f6f6af2dd9..66944a4e33 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -202,9 +202,8 @@ struct x86_reg aos_get_internal_xmm( struct aos_compilation *cp, #define ERROR(cp, msg) \ do { \ - debug_printf("%s: x86 translation failed: %s\n", __FUNCTION__, msg); \ + if (0) debug_printf("%s: x86 translation failed: %s\n", __FUNCTION__, msg); \ cp->error = 1; \ - assert(0); \ } while (0) -- cgit v1.2.3 From 4566b006f1a6bbdb96871e511e10e16f18bad23e Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Tue, 10 Jun 2008 20:17:16 +0100 Subject: Bring in DRI2 changes --- src/gallium/winsys/dri/intel/intel_context.c | 83 +- src/gallium/winsys/dri/intel/intel_screen.c | 449 ++++++--- src/glx/x11/Makefile | 26 +- src/glx/x11/XF86dri.c | 39 +- src/glx/x11/dri_glx.c | 885 +++++++++++------- src/glx/x11/glcontextmodes.c | 30 +- src/glx/x11/glcontextmodes.h | 6 +- src/glx/x11/glx_pbuffer.c | 100 ++- src/glx/x11/glx_texture_compression.c | 347 ------- src/glx/x11/glxclient.h | 171 ++-- src/glx/x11/glxcmds.c | 707 +++++++-------- src/glx/x11/glxext.c | 1198 ++++--------------------- src/glx/x11/glxextensions.c | 25 +- src/glx/x11/glxextensions.h | 4 +- src/glx/x11/indirect.c | 353 +++++++- src/glx/x11/indirect.h | 2 + src/glx/x11/indirect_init.c | 2 + src/glx/x11/indirect_vertex_array.c | 24 +- src/glx/x11/singlepix.c | 2 +- src/glx/x11/xf86dri.h | 14 +- src/glx/x11/xfont.c | 6 +- src/mesa/drivers/dri/common/dri_util.c | 1036 +++++++++++---------- src/mesa/drivers/dri/common/dri_util.h | 319 ++++--- src/mesa/drivers/dri/common/drirenderbuffer.c | 2 - src/mesa/drivers/dri/common/spantmp2.h | 2 +- src/mesa/drivers/dri/common/utils.c | 309 +++++-- src/mesa/drivers/dri/common/utils.h | 41 +- src/mesa/drivers/dri/common/vblank.c | 159 ++-- src/mesa/drivers/dri/common/vblank.h | 16 +- src/mesa/drivers/dri/common/xmlconfig.c | 26 +- 30 files changed, 3101 insertions(+), 3282 deletions(-) delete mode 100644 src/glx/x11/glx_texture_compression.c (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index 8284e0edbb..ecc4b0aa53 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -67,7 +67,6 @@ int __intel_debug = 0; #define need_GL_NV_vertex_program #include "extension_helper.h" - /** * Extension strings exported by the intel driver. * @@ -75,7 +74,7 @@ int __intel_debug = 0; * It appears that ARB_texture_env_crossbar has "disappeared" compared to the * old i830-specific driver. */ -const struct dri_extension card_extensions[] = { +static const struct dri_extension card_extensions[] = { {"GL_ARB_multisample", GL_ARB_multisample_functions}, {"GL_ARB_multitexture", NULL}, {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, @@ -86,22 +85,27 @@ const struct dri_extension card_extensions[] = { {"GL_ARB_texture_env_combine", NULL}, {"GL_ARB_texture_env_dot3", NULL}, {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_non_power_of_two", NULL }, {"GL_ARB_texture_rectangle", NULL}, + {"GL_NV_texture_rectangle", NULL}, + {"GL_EXT_texture_rectangle", NULL}, + {"GL_ARB_point_parameters", NULL}, {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_equation_separate", + GL_EXT_blend_equation_separate_functions}, {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_logic_op", NULL}, {"GL_EXT_blend_subtract", NULL}, {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, +#if 1 /* XXX FBO temporary? */ {"GL_EXT_packed_depth_stencil", NULL}, - {"GL_EXT_pixel_buffer_object", NULL}, +#endif {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, {"GL_EXT_stencil_wrap", NULL}, {"GL_EXT_texture_edge_clamp", NULL}, @@ -116,10 +120,67 @@ const struct dri_extension card_extensions[] = { {"GL_NV_blend_square", NULL}, {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, {"GL_NV_vertex_program1_1", NULL}, - {"GL_SGIS_generate_mipmap", NULL }, + { "GL_SGIS_generate_mipmap", NULL }, {NULL, NULL} }; +#if 0 +static const struct dri_extension brw_extensions[] = { + { "GL_ARB_shading_language_100", GL_VERSION_2_0_functions}, + { "GL_ARB_shading_language_120", GL_VERSION_2_1_functions}, + { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions}, + { "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions}, + { "GL_ARB_point_sprite", NULL}, + { "GL_ARB_fragment_shader", NULL }, + { "GL_ARB_draw_buffers", NULL }, + { "GL_ARB_depth_texture", NULL }, + { "GL_ARB_fragment_program", NULL }, + { "GL_ARB_shadow", NULL }, + { "GL_EXT_shadow_funcs", NULL }, + /* ARB extn won't work if not enabled */ + { "GL_SGIX_depth_texture", NULL }, + { "GL_ARB_texture_env_crossbar", NULL }, + { "GL_EXT_texture_sRGB", NULL}, + { NULL, NULL } +}; + +static const struct dri_extension arb_oc_extensions[] = { + {"GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions}, + {NULL, NULL} +}; + +static const struct dri_extension ttm_extensions[] = { + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {NULL, NULL} +}; +#endif + +/** + * Initializes potential list of extensions if ctx == NULL, or actually enables + * extensions for a context. + */ +void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging) +{ + /* Disable imaging extension until convolution is working in teximage paths. + */ + enable_imaging = GL_FALSE; + + driInitExtensions(ctx, card_extensions, enable_imaging); + +#if 0 + if (intel == NULL || intel->ttm) + driInitExtensions(ctx, ttm_extensions, GL_FALSE); + + if (intel == NULL || + (IS_965(intel->intelScreen->deviceID) && + intel->intelScreen->drmMinor >= 8)) + driInitExtensions(ctx, arb_oc_extensions, GL_FALSE); + + if (intel == NULL || IS_965(intel->intelScreen->deviceID)) + driInitExtensions(ctx, brw_extensions, GL_FALSE); +#endif +} #ifdef DEBUG @@ -204,10 +265,10 @@ intelCreateContext(const __GLcontextModes * visual, /* * memory pools */ - DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + DRM_LIGHT_LOCK(sPriv->fd, &sPriv->lock, driContextPriv->hHWContext); // ZZZ JB should be per screen and not be done per context havePools = intelCreatePools(sPriv); - DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + DRM_UNLOCK(sPriv->fd, &sPriv->lock, driContextPriv->hHWContext); if (!havePools) return GL_FALSE; @@ -215,7 +276,7 @@ intelCreateContext(const __GLcontextModes * visual, /* Dri stuff */ intel->hHWContext = driContextPriv->hHWContext; intel->driFd = sPriv->fd; - intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; + intel->driHwLock = (drmLock *) & sPriv->lock; fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); intel->iw.irq_seq = -1; @@ -265,7 +326,7 @@ intelCreateContext(const __GLcontextModes * visual, intel->st = st_create_context(pipe, visual, st_share); - driInitExtensions( intel->st->ctx, card_extensions, GL_TRUE ); + intelInitExtensions( intel->st->ctx, GL_TRUE ); return GL_TRUE; } diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 8817a80a5a..23889c80b8 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -60,7 +60,190 @@ static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; extern const struct dri_extension card_extensions[]; +static GLboolean +intel_get_param(__DRIscreenPrivate *psp, int param, int *value) +{ + int ret; + struct drm_i915_getparam gp; + + gp.param = param; + gp.value = value; + + ret = drmCommandWriteRead(psp->fd, DRM_I915_GETPARAM, &gp, sizeof(gp)); + if (ret) { + fprintf(stderr, "drm_i915_getparam: %d\n", ret); + return GL_FALSE; + } + + return GL_TRUE; +} + +static void +intelSetTexOffset(__DRIcontext *pDRICtx, int texname, + unsigned long long offset, int depth, uint pitch) +{ + abort(); +#if 0 + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct st_texture_object *stObj = st_texture_object(tObj); + + if (!stObj) + return; + + if (stObj->pt) + st->pipe->texture_release(intel->st->pipe, &stObj->pt); + + stObj->imageOverride = GL_TRUE; + stObj->depthOverride = depth; + stObj->pitchOverride = pitch; + + if (offset) + stObj->textureOffset = offset; +#endif +} + + +static void +intelHandleDrawableConfig(__DRIdrawablePrivate *dPriv, + __DRIcontextPrivate *pcp, + __DRIDrawableConfigEvent *event) +{ + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + struct intel_region *region = NULL; + struct intel_renderbuffer *rb, *depth_rb, *stencil_rb; + struct intel_context *intel = pcp->driverPrivate; + struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); + int cpp, pitch; + +#if 0 + cpp = intelScreen->front.cpp; + pitch = ((cpp * dPriv->w + 63) & ~63) / cpp; + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + rb = intel_fb->color_rb[1]; + if (rb) { + region = intel_region_alloc(intel, cpp, pitch, dPriv->h); + intel_renderbuffer_set_region(rb, region); + } + + rb = intel_fb->color_rb[2]; + if (rb) { + region = intel_region_alloc(intel, cpp, pitch, dPriv->h); + intel_renderbuffer_set_region(rb, region); + } + + depth_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH); + stencil_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL); + if (depth_rb || stencil_rb) + region = intel_region_alloc(intel, cpp, pitch, dPriv->h); + if (depth_rb) + intel_renderbuffer_set_region(depth_rb, region); + if (stencil_rb) + intel_renderbuffer_set_region(stencil_rb, region); + + /* FIXME: Tell the X server about the regions we just allocated and + * attached. */ +#endif + +} + +#define BUFFER_FLAG_TILED 0x0100 + +static void +intelHandleBufferAttach(__DRIdrawablePrivate *dPriv, + __DRIcontextPrivate *pcp, + __DRIBufferAttachEvent *ba) +{ + struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); + struct intel_framebuffer *intel_fb = dPriv->driverPrivate; + struct intel_renderbuffer *rb; + struct intel_region *region; + struct intel_context *intel = pcp->driverPrivate; + struct pipe_surface *surf; + GLuint tiled; + + switch (ba->buffer.attachment) { + case DRI_DRAWABLE_BUFFER_FRONT_LEFT: + #if 0 + intelScreen->front.width = ba->width; + intelScreen->front.height = ba->height; + intelScreen->front.offset = sarea->front_offset; + #endif + intelScreen->front.pitch = ba->buffer.pitch * ba->buffer.cpp; + #if 0 + intelScreen->front.size = sarea->front_size; + #endif + driGenBuffers(intelScreen->base.staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); + driBOSetReferenced(intelScreen->front.buffer, ba->buffer.handle); + + break; + +#if 0 + case DRI_DRAWABLE_BUFFER_BACK_LEFT: + rb = intel_fb->color_rb[0]; + break; + + case DRI_DRAWABLE_BUFFER_DEPTH: + rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH); + break; + + case DRI_DRAWABLE_BUFFER_STENCIL: + rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL); + break; +#endif + + case DRI_DRAWABLE_BUFFER_ACCUM: + default: + fprintf(stderr, "unhandled buffer attach event, attacment type %d\n", + ba->buffer.attachment); + return; + } + +#if 0 + /* FIXME: Add this so we can filter out when the X server sends us + * attachment events for the buffers we just allocated. Need to + * get the BO handle for a render buffer. */ + if (intel_renderbuffer_get_region_handle(rb) == ba->buffer.handle) + return; +#endif + +#if 0 + tiled = (ba->buffer.flags & BUFFER_FLAG_TILED) > 0; + + region = intel_region_alloc_for_handle(intel, ba->buffer.cpp, + ba->buffer.pitch / ba->buffer.cpp, + dPriv->h, tiled, + ba->buffer.handle); + + intel_renderbuffer_set_region(rb, region); +#endif +} + +static const __DRItexOffsetExtension intelTexOffsetExtension = { + { __DRI_TEX_OFFSET }, + intelSetTexOffset, +}; + +#if 0 +static const __DRItexBufferExtension intelTexBufferExtension = { + { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION }, + intelSetTexBuffer, +}; +#endif +static const __DRIextension *intelScreenExtensions[] = { + &driReadDrawableExtension, + &driCopySubBufferExtension.base, + &driSwapControlExtension.base, + &driFrameTrackingExtension.base, + &driMediaStreamCounterExtension.base, + &intelTexOffsetExtension.base, +// &intelTexBufferExtension.base, + NULL +}; static void @@ -177,7 +360,8 @@ intelCreatePools(__DRIscreenPrivate * sPriv) intelScreen->havePools = GL_TRUE; - intelUpdateScreenRotation(sPriv, intelScreen->sarea); + if (intelScreen->sarea) + intelUpdateScreenRotation(sPriv, intelScreen->sarea); return GL_TRUE; } @@ -210,11 +394,6 @@ intelInitDriver(__DRIscreenPrivate * sPriv) struct intel_screen *intelScreen; I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; - PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = - (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> - getProcAddress("glxEnableExtension")); - void *const psc = sPriv->psc->screenConfigs; - if (sPriv->devPrivSize != sizeof(I830DRIRec)) { fprintf(stderr, "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); @@ -231,28 +410,17 @@ intelInitDriver(__DRIscreenPrivate * sPriv) __driConfigOptions, __driNConfigOptions); sPriv->private = (void *) intelScreen; - intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + - gDRIPriv->sarea_priv_offset); - intelScreen->deviceID = gDRIPriv->deviceID; - intelScreen->front.cpp = gDRIPriv->cpp; - intelScreen->drmMinor = sPriv->drmMinor; + gDRIPriv->sarea_priv_offset); - assert(gDRIPriv->bitsPerPixel == 16 || - gDRIPriv->bitsPerPixel == 32); + intelScreen->deviceID = gDRIPriv->deviceID; intelUpdateScreenRotation(sPriv, intelScreen->sarea); if (0) intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); - if (glx_enable_extension != NULL) { - (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); - (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); - (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); - } + sPriv->extensions = intelScreenExtensions; intel_be_init_device(&intelScreen->base, sPriv->fd); intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; @@ -351,65 +519,19 @@ intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) return 0; } - -static void -intelSetTexOffset(__DRIcontext *pDRICtx, int texname, - unsigned long long offset, int depth, uint pitch) +static __DRIconfig ** +intelFillInModes(__DRIscreenPrivate *psp, + unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, GLboolean have_back_buffer) { - abort(); -#if 0 - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct st_texture_object *stObj = st_texture_object(tObj); - - if (!stObj) - return; - - if (stObj->pt) - st->pipe->texture_release(intel->st->pipe, &stObj->pt); - - stObj->imageOverride = GL_TRUE; - stObj->depthOverride = depth; - stObj->pitchOverride = pitch; - - if (offset) - stObj->textureOffset = offset; -#endif -} - - -static const struct __DriverAPIRec intelAPI = { - .InitDriver = intelInitDriver, - .DestroyScreen = intelDestroyScreen, - .CreateContext = intelCreateContext, - .DestroyContext = intelDestroyContext, - .CreateBuffer = intelCreateBuffer, - .DestroyBuffer = intelDestroyBuffer, - .SwapBuffers = intelSwapBuffers, - .MakeCurrent = intelMakeCurrent, - .UnbindContext = intelUnbindContext, - .GetSwapInfo = intelGetSwapInfo, - .GetMSC = driGetMSC32, - .WaitForMSC = driWaitForMSC32, - .WaitForSBC = NULL, - .SwapBuffersMSC = NULL, - .CopySubBuffer = intelCopySubBuffer, - .setTexOffset = intelSetTexOffset, -}; - - -static __GLcontextModes * -intelFillInModes(unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, boolean have_back_buffer) -{ - __GLcontextModes *modes; + __DRIconfig **configs; __GLcontextModes *m; unsigned num_modes; unsigned depth_buffer_factor; unsigned back_buffer_factor; GLenum fb_format; GLenum fb_type; + int i; /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't * support pageflipping at all. @@ -451,100 +573,143 @@ intelFillInModes(unsigned pixel_bits, unsigned depth_bits, fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; } - modes = - (*dri_interface->createContextModes) (num_modes, - sizeof(__GLcontextModes)); - m = modes; - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_TRUE_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_DIRECT_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + configs = driCreateConfigs(fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor); + if (configs == NULL) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, __LINE__); return NULL; } /* Mark the visual as slow if there are "fake" stencil bits. */ - for (m = modes; m != NULL; m = m->next) { + for (i = 0; configs[i]; i++) { + m = &configs[i]->modes; if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { m->visualRating = GLX_SLOW_CONFIG; } } - return modes; + return configs; } - /** - * This is the bootstrap function for the driver. libGL supplies all of the - * requisite information about the system, and the driver initializes itself. - * This routine also fills in the linked list pointed to by \c driver_modes - * with the \c __GLcontextModes that the driver can support for windows or - * pbuffers. + * This is the driver specific part of the createNewScreen entry point. + * + * \todo maybe fold this into intelInitDriver * - * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on - * failure. + * \return the __GLcontextModes supported by this driver */ -PUBLIC void * -__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, - __DRIscreen * psc, - const __GLcontextModes * modes, - const __DRIversion * ddx_version, - const __DRIversion * dri_version, - const __DRIversion * drm_version, - const __DRIframebuffer * frame_buffer, - drmAddress pSAREA, int fd, - int internal_api_version, - const __DRIinterfaceMethods * interface, - __GLcontextModes ** driver_modes) +static const __DRIconfig **intelInitScreen(__DRIscreenPrivate *psp) { - __DRIscreenPrivate *psp; - static const __DRIversion ddx_expected = { 1, 7, 0 }; +#ifdef I915 + static const __DRIversion ddx_expected = { 1, 5, 0 }; +#else + static const __DRIversion ddx_expected = { 1, 6, 0 }; +#endif static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 7, 0 }; - - dri_interface = interface; + static const __DRIversion drm_expected = { 1, 5, 0 }; + I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; if (!driCheckDriDdxDrmVersions2("i915", - dri_version, &dri_expected, - ddx_version, &ddx_expected, - drm_version, &drm_expected)) { + &psp->dri_version, &dri_expected, + &psp->ddx_version, &ddx_expected, + &psp->drm_version, &drm_expected)) { return NULL; } - psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, - ddx_version, dri_version, drm_version, - frame_buffer, pSAREA, fd, - internal_api_version, &intelAPI); - - if (psp != NULL) { - I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; - *driver_modes = intelFillInModes(dri_priv->cpp * 8, - (dri_priv->cpp == 2) ? 16 : 24, - (dri_priv->cpp == 2) ? 0 : 8, 1); - - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create - * is called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions(NULL, card_extensions, GL_FALSE); + /* Calling driInitExtensions here, with a NULL context pointer, + * does not actually enable the extensions. It just makes sure + * that all the dispatch offsets for all the extensions that + * *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create is + * called, but we can't enable the extensions until we have a + * context pointer. + * + * Hello chicken. Hello egg. How are you two today? + */ + intelInitExtensions(NULL, GL_TRUE); + + if (!intelInitDriver(psp)) + return NULL; + + psp->extensions = intelScreenExtensions; + + return (const __DRIconfig **) + intelFillInModes(psp, dri_priv->cpp * 8, + (dri_priv->cpp == 2) ? 16 : 24, + (dri_priv->cpp == 2) ? 0 : 8, 1); +} + +/** + * This is the driver specific part of the createNewScreen entry point. + * + * \return the __GLcontextModes supported by this driver + */ +static const +__DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp) +{ + struct intel_screen *intelScreen; + + /* Calling driInitExtensions here, with a NULL context pointer, + * does not actually enable the extensions. It just makes sure + * that all the dispatch offsets for all the extensions that + * *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create is + * called, but we can't enable the extensions until we have a + * context pointer. + * + * Hello chicken. Hello egg. How are you two today? + */ + intelInitExtensions(NULL, GL_TRUE); + + /* Allocate the private area */ + intelScreen = CALLOC_STRUCT(intel_screen); + if (!intelScreen) { + fprintf(stderr, "\nERROR! Allocating private area failed\n"); + return GL_FALSE; } + /* parse information in __driConfigOptions */ + driParseOptionInfo(&intelScreen->optionCache, + __driConfigOptions, __driNConfigOptions); - return (void *) psp; + psp->private = (void *) intelScreen; + + intelScreen->drmMinor = psp->drm_version.minor; + + /* Determine chipset ID? */ + if (!intel_get_param(psp, I915_PARAM_CHIPSET_ID, + &intelScreen->deviceID)) + return GL_FALSE; + + psp->extensions = intelScreenExtensions; + + intel_be_init_device(&intelScreen->base, psp->fd); + intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; + intelScreen->base.base.get_name = intel_get_name; + + return driConcatConfigs(intelFillInModes(psp, 16, 16, 0, 1), + intelFillInModes(psp, 32, 24, 8, 1)); } +const struct __DriverAPIRec driDriverAPI = { + .InitScreen = intelInitScreen, + .DestroyScreen = intelDestroyScreen, + .CreateContext = intelCreateContext, + .DestroyContext = intelDestroyContext, + .CreateBuffer = intelCreateBuffer, + .DestroyBuffer = intelDestroyBuffer, + .SwapBuffers = intelSwapBuffers, + .MakeCurrent = intelMakeCurrent, + .UnbindContext = intelUnbindContext, + .GetSwapInfo = intelGetSwapInfo, + .GetDrawableMSC = driDrawableGetMSC32, + .WaitForMSC = driWaitForMSC32, + .CopySubBuffer = intelCopySubBuffer, + + .InitScreen2 = intelInitScreen2, + .HandleDrawableConfig = intelHandleDrawableConfig, + .HandleBufferAttach = intelHandleBufferAttach, +}; diff --git a/src/glx/x11/Makefile b/src/glx/x11/Makefile index b404727f08..1304311794 100644 --- a/src/glx/x11/Makefile +++ b/src/glx/x11/Makefile @@ -10,12 +10,14 @@ SOURCES = \ compsize.c \ eval.c \ glxcmds.c \ + glxcurrent.c \ glxext.c \ glxextensions.c \ indirect.c \ indirect_init.c \ indirect_size.c \ indirect_window_pos.c \ + indirect_texture_compression.c \ indirect_transpose_matrix.c \ indirect_vertex_array.c \ indirect_vertex_program.c \ @@ -29,13 +31,16 @@ SOURCES = \ xfont.c \ glx_pbuffer.c \ glx_query.c \ - glx_texture_compression.c \ + dri_common.c \ dri_glx.c \ - XF86dri.c + XF86dri.c \ + glxhash.c \ + dri2_glx.c \ + dri2.c include $(TOP)/src/mesa/sources -MESA_GLAPI_ASM_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_ASM_SOURCES)) +MESA_ASM_API = $(addprefix $(TOP)/src/mesa/, $(ASM_API)) MESA_GLAPI_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_SOURCES)) MESA_GLAPI_OBJECTS = $(addprefix $(TOP)/src/mesa/, $(GLAPI_OBJECTS)) @@ -46,8 +51,8 @@ INCLUDES = -I. \ -I$(TOP)/include/GL/internal \ -I$(TOP)/src/mesa \ -I$(TOP)/src/mesa/main \ - -I$(TOP)/src/mesa/glapi \ $(LIBDRM_CFLAGS) \ + $(DRI2PROTO_CFLAGS) \ $(X11_INCLUDES) @@ -65,29 +70,28 @@ default: depend $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) # Make libGL $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(OBJECTS) Makefile - $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ + $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' -ldflags '$(LDFLAGS)' \ -major 1 -minor 2 $(MKLIB_OPTIONS) \ -install $(TOP)/$(LIB_DIR) $(GL_LIB_DEPS) $(OBJECTS) -depend: $(SOURCES) $(MESA_GLAPI_SOURCES) $(MESA_GLAPI_ASM_SOURCES) Makefile - rm -f depend +depend: $(SOURCES) $(MESA_GLAPI_SOURCES) $(MESA_ASM_API) Makefile touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) \ - $(MESA_GLAPI_SOURCES) $(MESA_GLAPI_ASM_SOURCES) + $(MESA_GLAPI_SOURCES) $(MESA_ASM_API) # Emacs tags tags: etags `find . -name \*.[ch]` `find $(TOP)/include` -# Dummy install target -install: +install: $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) + $(MAKE) -C $(TOP)/src/mesa install-libgl # Remove .o and backup files clean: -rm -f $(TOP)/$(LIB_DIR)/libGL.so* -rm -f *.o *~ - -rm -f depend + -rm -f depend depend.bak include depend diff --git a/src/glx/x11/XF86dri.c b/src/glx/x11/XF86dri.c index 9919a40977..ba38949c0b 100644 --- a/src/glx/x11/XF86dri.c +++ b/src/glx/x11/XF86dri.c @@ -374,10 +374,9 @@ PUBLIC Bool XF86DRICreateContext(dpy, screen, visual, context, hHWContext) context, hHWContext ); } -PUBLIC GLboolean XF86DRIDestroyContext( __DRInativeDisplay * ndpy, int screen, - __DRIid context ) +PUBLIC GLboolean XF86DRIDestroyContext(Display *dpy, int screen, + XID context ) { - Display * const dpy = (Display *) ndpy; XExtDisplayInfo *info = find_display (dpy); xXF86DRIDestroyContextReq *req; @@ -396,10 +395,9 @@ PUBLIC GLboolean XF86DRIDestroyContext( __DRInativeDisplay * ndpy, int screen, return True; } -PUBLIC GLboolean XF86DRICreateDrawable( __DRInativeDisplay * ndpy, int screen, - __DRIid drawable, drm_drawable_t * hHWDrawable ) +PUBLIC GLboolean XF86DRICreateDrawable(Display *dpy, int screen, + XID drawable, drm_drawable_t * hHWDrawable ) { - Display * const dpy = (Display *) ndpy; XExtDisplayInfo *info = find_display (dpy); xXF86DRICreateDrawableReply rep; xXF86DRICreateDrawableReq *req; @@ -426,16 +424,36 @@ PUBLIC GLboolean XF86DRICreateDrawable( __DRInativeDisplay * ndpy, int screen, return True; } -PUBLIC GLboolean XF86DRIDestroyDrawable( __DRInativeDisplay * ndpy, int screen, - __DRIid drawable ) +static int noopErrorHandler(Display *dpy, XErrorEvent *xerr) +{ + return 0; +} + +PUBLIC GLboolean XF86DRIDestroyDrawable(Display *dpy, int screen, + XID drawable ) { - Display * const dpy = (Display *) ndpy; XExtDisplayInfo *info = find_display (dpy); xXF86DRIDestroyDrawableReq *req; + int (*oldXErrorHandler)(Display *, XErrorEvent *); TRACE("DestroyDrawable..."); XF86DRICheckExtension (dpy, info, False); + /* This is called from the DRI driver, which used call it like this + * + * if (windowExists(drawable)) + * destroyDrawable(drawable); + * + * which is a textbook race condition - the window may disappear + * from the server between checking for its existance and + * destroying it. Instead we change the semantics of + * __DRIinterfaceMethodsRec::destroyDrawable() to succeed even if + * the windows is gone, by wrapping the destroy call in an error + * handler. */ + + XSync(dpy, GL_FALSE); + oldXErrorHandler = XSetErrorHandler(noopErrorHandler); + LockDisplay(dpy); GetReq(XF86DRIDestroyDrawable, req); req->reqType = info->codes->major_opcode; @@ -444,6 +462,9 @@ PUBLIC GLboolean XF86DRIDestroyDrawable( __DRInativeDisplay * ndpy, int screen, req->drawable = drawable; UnlockDisplay(dpy); SyncHandle(); + + XSetErrorHandler(oldXErrorHandler); + TRACE("DestroyDrawable... return True"); return True; } diff --git a/src/glx/x11/dri_glx.c b/src/glx/x11/dri_glx.c index 21e07c1935..1cb3204d4c 100644 --- a/src/glx/x11/dri_glx.c +++ b/src/glx/x11/dri_glx.c @@ -34,260 +34,47 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #ifdef GLX_DIRECT_RENDERING -#include -#include -#include -#include +#include +#include +#include #include "glheader.h" #include "glxclient.h" +#include "glcontextmodes.h" #include "xf86dri.h" #include "sarea.h" -#include #include -#include "dri_glx.h" #include -#include - -#ifndef RTLD_NOW -#define RTLD_NOW 0 -#endif -#ifndef RTLD_GLOBAL -#define RTLD_GLOBAL 0 -#endif - - -#ifndef DEFAULT_DRIVER_DIR -/* this is normally defined in Mesa/configs/default with DRI_DRIVER_SEARCH_PATH */ -#define DEFAULT_DRIVER_DIR "/usr/X11R6/lib/modules/dri" -#endif - -static __DRIdriver *Drivers = NULL; - - -/* - * printf wrappers - */ - -static void InfoMessageF(const char *f, ...) -{ - va_list args; - const char *env; - - if ((env = getenv("LIBGL_DEBUG")) && strstr(env, "verbose")) { - fprintf(stderr, "libGL: "); - va_start(args, f); - vfprintf(stderr, f, args); - va_end(args); - } -} - -/** - * Print error to stderr, unless LIBGL_DEBUG=="quiet". - */ -static void ErrorMessageF(const char *f, ...) -{ - va_list args; - const char *env; - - if ((env = getenv("LIBGL_DEBUG")) && !strstr(env, "quiet")) { - fprintf(stderr, "libGL error: "); - va_start(args, f); - vfprintf(stderr, f, args); - va_end(args); - } -} - - -/** - * Extract the ith directory path out of a colon-separated list of paths. No - * more than \c dirLen characters, including the terminating \c NUL, will be - * written to \c dir. - * - * \param index Index of path to extract (starting at zero) - * \param paths The colon-separated list of paths - * \param dirLen Maximum length of result to store in \c dir - * \param dir Buffer to hold the extracted directory path - * - * \returns - * The number of characters that would have been written to \c dir had there - * been enough room. This does not include the terminating \c NUL. When - * extraction fails, zero will be returned. - * - * \todo - * It seems like this function could be rewritten to use \c strchr. - */ -static size_t -ExtractDir(int index, const char *paths, int dirLen, char *dir) -{ - int i, len; - const char *start, *end; - - /* find ith colon */ - start = paths; - i = 0; - while (i < index) { - if (*start == ':') { - i++; - start++; - } - else if (*start == 0) { - /* end of string and couldn't find ith colon */ - dir[0] = 0; - return 0; - } - else { - start++; - } - } - - while (*start == ':') - start++; - - /* find next colon, or end of string */ - end = start + 1; - while (*end != ':' && *end != 0) { - end++; - } - - /* copy string between and into result string */ - len = end - start; - if (len > dirLen - 1) - len = dirLen - 1; - strncpy(dir, start, len); - dir[len] = 0; - - return( end - start ); -} - - -/** - * Versioned name of the expected \c __driCreateNewScreen function. - * - * The version of the last incompatible loader/driver inteface change is - * appended to the name of the \c __driCreateNewScreen function. This - * prevents loaders from trying to load drivers that are too old. - * - * \todo - * Create a macro or something so that this is automatically updated. - */ -static const char createNewScreenName[] = "__driCreateNewScreen_20050727"; - - -/** - * Try to \c dlopen the named driver. - * - * This function adds the "_dri.so" suffix to the driver name and searches the - * directories specified by the \c LIBGL_DRIVERS_PATH environment variable in - * order to find the driver. - * - * \param driverName - a name like "tdfx", "i810", "mga", etc. - * - * \returns - * A handle from \c dlopen, or \c NULL if driver file not found. - */ -static __DRIdriver *OpenDriver(const char *driverName) -{ - void *glhandle = NULL; - char *libPaths = NULL; - char libDir[1000]; - int i; - __DRIdriver *driver; - - /* First, search Drivers list to see if we've already opened this driver */ - for (driver = Drivers; driver; driver = driver->next) { - if (strcmp(driver->name, driverName) == 0) { - /* found it */ - return driver; - } - } - - /* Attempt to make sure libGL symbols will be visible to the driver */ - glhandle = dlopen("libGL.so.1", RTLD_NOW | RTLD_GLOBAL); - - if (geteuid() == getuid()) { - /* don't allow setuid apps to use LIBGL_DRIVERS_PATH */ - libPaths = getenv("LIBGL_DRIVERS_PATH"); - if (!libPaths) - libPaths = getenv("LIBGL_DRIVERS_DIR"); /* deprecated */ - } - if (!libPaths) - libPaths = DEFAULT_DRIVER_DIR; - - for ( i = 0 ; ExtractDir(i, libPaths, 1000, libDir) != 0 ; i++ ) { - char realDriverName[200]; - void *handle = NULL; - - - /* If TLS support is enabled, try to open the TLS version of the driver - * binary first. If that fails, try the non-TLS version. - */ -#ifdef GLX_USE_TLS - snprintf(realDriverName, 200, "%s/tls/%s_dri.so", libDir, driverName); - InfoMessageF("OpenDriver: trying %s\n", realDriverName); - handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); -#endif - - if ( handle == NULL ) { - snprintf(realDriverName, 200, "%s/%s_dri.so", libDir, driverName); - InfoMessageF("OpenDriver: trying %s\n", realDriverName); - handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); - } - - if ( handle != NULL ) { - /* allocate __DRIdriver struct */ - driver = (__DRIdriver *) Xmalloc(sizeof(__DRIdriver)); - if (!driver) - break; /* out of memory! */ - /* init the struct */ - driver->name = __glXstrdup(driverName); - if (!driver->name) { - Xfree(driver); - driver = NULL; - break; /* out of memory! */ - } - - driver->createNewScreenFunc = (PFNCREATENEWSCREENFUNC) - dlsym(handle, createNewScreenName); - - if ( driver->createNewScreenFunc == NULL ) { - /* If the driver doesn't have this symbol then something's - * really, really wrong. - */ - ErrorMessageF("%s not defined in %s_dri.so!\n" - "Your driver may be too old for this libGL.\n", - createNewScreenName, driverName); - Xfree(driver); - driver = NULL; - dlclose(handle); - continue; - } - driver->handle = handle; - /* put at head of linked list */ - driver->next = Drivers; - Drivers = driver; - break; - } - else { - ErrorMessageF("dlopen %s failed (%s)\n", realDriverName, dlerror()); - } - } - - if (!driver) - ErrorMessageF("unable to load driver: %s_dri.so\n", driverName); - - if (glhandle) - dlclose(glhandle); - - return driver; -} - +#include +#include "xf86drm.h" +#include "dri_common.h" + +typedef struct __GLXDRIdisplayPrivateRec __GLXDRIdisplayPrivate; +typedef struct __GLXDRIcontextPrivateRec __GLXDRIcontextPrivate; + +struct __GLXDRIdisplayPrivateRec { + __GLXDRIdisplay base; + + /* + ** XFree86-DRI version information + */ + int driMajor; + int driMinor; + int driPatch; +}; + +struct __GLXDRIcontextPrivateRec { + __GLXDRIcontext base; + __DRIcontext *driContext; + XID hwContextID; + __GLXscreenConfigs *psc; +}; /* * Given a display pointer and screen number, determine the name of * the DRI driver for the screen. (I.e. "r128", "tdfx", etc). * Return True for success, False for failure. */ -static Bool GetDriverName(Display *dpy, int scrNum, char **driverName) +static Bool driGetDriverName(Display *dpy, int scrNum, char **driverName) { int directCapable; Bool b; @@ -317,25 +104,6 @@ static Bool GetDriverName(Display *dpy, int scrNum, char **driverName) return True; } - -/* - * Given a display pointer and screen number, return a __DRIdriver handle. - * Return NULL if anything goes wrong. - */ -__DRIdriver *driGetDriver(Display *dpy, int scrNum) -{ - char *driverName; - if (GetDriverName(dpy, scrNum, &driverName)) { - __DRIdriver *ret; - ret = OpenDriver(driverName); - if (driverName) - Xfree(driverName); - return ret; - } - return NULL; -} - - /* * Exported function for querying the DRI driver for a given screen. * @@ -345,7 +113,7 @@ __DRIdriver *driGetDriver(Display *dpy, int scrNum) PUBLIC const char *glXGetScreenDriver (Display *dpy, int scrNum) { static char ret[32]; char *driverName; - if (GetDriverName(dpy, scrNum, &driverName)) { + if (driGetDriverName(dpy, scrNum, &driverName)) { int len; if (!driverName) return NULL; @@ -359,7 +127,6 @@ PUBLIC const char *glXGetScreenDriver (Display *dpy, int scrNum) { return NULL; } - /* * Exported function for obtaining a driver's option list (UTF-8 encoded XML). * @@ -371,71 +138,532 @@ PUBLIC const char *glXGetScreenDriver (Display *dpy, int scrNum) { * * Note: The driver remains opened after this function returns. */ -PUBLIC const char *glXGetDriverConfig (const char *driverName) { - __DRIdriver *driver = OpenDriver (driverName); - if (driver) - return dlsym (driver->handle, "__driConfigOptions"); +PUBLIC const char *glXGetDriverConfig (const char *driverName) +{ + void *handle = driOpenDriver (driverName); + if (handle) + return dlsym (handle, "__driConfigOptions"); else return NULL; } +#ifdef XDAMAGE_1_1_INTERFACE -/* Called from __glXFreeDisplayPrivate. +static GLboolean has_damage_post(Display *dpy) +{ + static GLboolean inited = GL_FALSE; + static GLboolean has_damage; + + if (!inited) { + int major, minor; + + if (XDamageQueryVersion(dpy, &major, &minor) && + major == 1 && minor >= 1) + { + has_damage = GL_TRUE; + } else { + has_damage = GL_FALSE; + } + inited = GL_TRUE; + } + + return has_damage; +} + +static void __glXReportDamage(__DRIdrawable *driDraw, + int x, int y, + drm_clip_rect_t *rects, int num_rects, + GLboolean front_buffer, + void *loaderPrivate) +{ + XRectangle *xrects; + XserverRegion region; + int i; + int x_off, y_off; + __GLXDRIdrawable *glxDraw = loaderPrivate; + __GLXscreenConfigs *psc = glxDraw->psc; + Display *dpy = psc->dpy; + Drawable drawable; + + if (!has_damage_post(dpy)) + return; + + if (front_buffer) { + x_off = x; + y_off = y; + drawable = RootWindow(dpy, psc->scr); + } else{ + x_off = 0; + y_off = 0; + drawable = glxDraw->xDrawable; + } + + xrects = malloc(sizeof(XRectangle) * num_rects); + if (xrects == NULL) + return; + + for (i = 0; i < num_rects; i++) { + xrects[i].x = rects[i].x1 + x_off; + xrects[i].y = rects[i].y1 + y_off; + xrects[i].width = rects[i].x2 - rects[i].x1; + xrects[i].height = rects[i].y2 - rects[i].y1; + } + region = XFixesCreateRegion(dpy, xrects, num_rects); + free(xrects); + XDamageAdd(dpy, drawable, region); + XFixesDestroyRegion(dpy, region); +} + +static const __DRIdamageExtension damageExtension = { + { __DRI_DAMAGE, __DRI_DAMAGE_VERSION }, + __glXReportDamage, +}; + +#endif + +static GLboolean +__glXDRIGetDrawableInfo(__DRIdrawable *drawable, + unsigned int *index, unsigned int *stamp, + int *X, int *Y, int *W, int *H, + int *numClipRects, drm_clip_rect_t ** pClipRects, + int *backX, int *backY, + int *numBackClipRects, drm_clip_rect_t **pBackClipRects, + void *loaderPrivate) +{ + __GLXDRIdrawable *glxDraw = loaderPrivate; + __GLXscreenConfigs *psc = glxDraw->psc; + Display *dpy = psc->dpy; + + return XF86DRIGetDrawableInfo(dpy, psc->scr, glxDraw->drawable, + index, stamp, X, Y, W, H, + numClipRects, pClipRects, + backX, backY, + numBackClipRects, pBackClipRects); +} + +static const __DRIgetDrawableInfoExtension getDrawableInfoExtension = { + { __DRI_GET_DRAWABLE_INFO, __DRI_GET_DRAWABLE_INFO_VERSION }, + __glXDRIGetDrawableInfo +}; + +static const __DRIextension *loader_extensions[] = { + &systemTimeExtension.base, + &getDrawableInfoExtension.base, +#ifdef XDAMAGE_1_1_INTERFACE + &damageExtension.base, +#endif + NULL +}; + +#ifndef GLX_USE_APPLEGL + +/** + * Perform the required libGL-side initialization and call the client-side + * driver's \c __driCreateNewScreen function. + * + * \param dpy Display pointer. + * \param scrn Screen number on the display. + * \param psc DRI screen information. + * \param driDpy DRI display information. + * \param createNewScreen Pointer to the client-side driver's + * \c __driCreateNewScreen function. + * \returns A pointer to the \c __DRIscreenPrivate structure returned by + * the client-side driver on success, or \c NULL on failure. */ -static void driDestroyDisplay(Display *dpy, void *private) +static void * +CallCreateNewScreen(Display *dpy, int scrn, __GLXscreenConfigs *psc, + __GLXDRIdisplayPrivate * driDpy) +{ + void *psp = NULL; + drm_handle_t hSAREA; + drmAddress pSAREA = MAP_FAILED; + char *BusID; + __DRIversion ddx_version; + __DRIversion dri_version; + __DRIversion drm_version; + __DRIframebuffer framebuffer; + int fd = -1; + int status; + + drm_magic_t magic; + drmVersionPtr version; + int newlyopened; + char *driverName; + drm_handle_t hFB; + int junk; + const __DRIconfig **driver_configs; + + /* DRI protocol version. */ + dri_version.major = driDpy->driMajor; + dri_version.minor = driDpy->driMinor; + dri_version.patch = driDpy->driPatch; + + framebuffer.base = MAP_FAILED; + framebuffer.dev_priv = NULL; + + if (!XF86DRIOpenConnection(dpy, scrn, &hSAREA, &BusID)) { + fprintf(stderr, "libGL error: XF86DRIOpenConnection failed\n"); + goto handle_error; + } + + fd = drmOpenOnce(NULL, BusID, &newlyopened); + + Xfree(BusID); /* No longer needed */ + + if (fd < 0) { + fprintf(stderr, "libGL error: drmOpenOnce failed (%s)\n", + strerror(-fd)); + goto handle_error; + } + + if (drmGetMagic(fd, &magic)) { + fprintf(stderr, "libGL error: drmGetMagic failed\n"); + goto handle_error; + } + + version = drmGetVersion(fd); + if (version) { + drm_version.major = version->version_major; + drm_version.minor = version->version_minor; + drm_version.patch = version->version_patchlevel; + drmFreeVersion(version); + } + else { + drm_version.major = -1; + drm_version.minor = -1; + drm_version.patch = -1; + } + + if (newlyopened && !XF86DRIAuthConnection(dpy, scrn, magic)) { + fprintf(stderr, "libGL error: XF86DRIAuthConnection failed\n"); + goto handle_error; + } + + /* Get device name (like "tdfx") and the ddx version numbers. + * We'll check the version in each DRI driver's "createNewScreen" + * function. */ + if (!XF86DRIGetClientDriverName(dpy, scrn, + &ddx_version.major, + &ddx_version.minor, + &ddx_version.patch, + &driverName)) { + fprintf(stderr, "libGL error: XF86DRIGetClientDriverName failed\n"); + goto handle_error; + } + + Xfree(driverName); /* No longer needed. */ + + /* + * Get device-specific info. pDevPriv will point to a struct + * (such as DRIRADEONRec in xfree86/driver/ati/radeon_dri.h) that + * has information about the screen size, depth, pitch, ancilliary + * buffers, DRM mmap handles, etc. + */ + if (!XF86DRIGetDeviceInfo(dpy, scrn, &hFB, &junk, + &framebuffer.size, &framebuffer.stride, + &framebuffer.dev_priv_size, &framebuffer.dev_priv)) { + fprintf(stderr, "libGL error: XF86DRIGetDeviceInfo failed"); + goto handle_error; + } + + framebuffer.width = DisplayWidth(dpy, scrn); + framebuffer.height = DisplayHeight(dpy, scrn); + + /* Map the framebuffer region. */ + status = drmMap(fd, hFB, framebuffer.size, + (drmAddressPtr)&framebuffer.base); + if (status != 0) { + fprintf(stderr, "libGL error: drmMap of framebuffer failed (%s)", + strerror(-status)); + goto handle_error; + } + + /* Map the SAREA region. Further mmap regions may be setup in + * each DRI driver's "createNewScreen" function. + */ + status = drmMap(fd, hSAREA, SAREA_MAX, &pSAREA); + if (status != 0) { + fprintf(stderr, "libGL error: drmMap of SAREA failed (%s)", + strerror(-status)); + goto handle_error; + } + + psp = (*psc->legacy->createNewScreen)(scrn, + &ddx_version, + &dri_version, + &drm_version, + &framebuffer, + pSAREA, + fd, + loader_extensions, + &driver_configs, + psc); + + if (psp == NULL) { + fprintf(stderr, "libGL error: Calling driver entry point failed"); + goto handle_error; + } + + psc->configs = driConvertConfigs(psc->core, psc->configs, driver_configs); + psc->visuals = driConvertConfigs(psc->core, psc->visuals, driver_configs); + + return psp; + + handle_error: + if (pSAREA != MAP_FAILED) + drmUnmap(pSAREA, SAREA_MAX); + + if (framebuffer.base != MAP_FAILED) + drmUnmap((drmAddress)framebuffer.base, framebuffer.size); + + if (framebuffer.dev_priv != NULL) + Xfree(framebuffer.dev_priv); + + if (fd >= 0) + drmCloseOnce(fd); + + XF86DRICloseConnection(dpy, scrn); + + fprintf(stderr, "libGL error: reverting to (slow) indirect rendering\n"); + + return NULL; +} + +#else /* !GLX_USE_APPLEGL */ + +static void * +CallCreateNewScreen(Display *dpy, int scrn, __GLXscreenConfigs *psc, + __GLXDRIdisplayPrivate * driDpy) { - __DRIdisplayPrivate *pdpyp = (__DRIdisplayPrivate *)private; - - if (pdpyp) { - const int numScreens = ScreenCount(dpy); - int i; - for (i = 0; i < numScreens; i++) { - if (pdpyp->libraryHandles[i]) { - __DRIdriver *driver, *prev; - - /* Remove driver from Drivers list */ - for (prev = NULL, driver = Drivers; driver; - prev = driver, driver = driver->next) { - if (driver->handle == pdpyp->libraryHandles[i]) { - if (prev) - prev->next = driver->next; - else - Drivers = driver->next; - - Xfree(driver->name); - Xfree(driver); - break; - } - } - - dlclose(pdpyp->libraryHandles[i]); - } - } - Xfree(pdpyp->libraryHandles); - Xfree(pdpyp); + return NULL; +} + +#endif /* !GLX_USE_APPLEGL */ + +static void driDestroyContext(__GLXDRIcontext *context, + __GLXscreenConfigs *psc, Display *dpy) +{ + __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; + + (*psc->core->destroyContext)(pcp->driContext); + + XF86DRIDestroyContext(psc->dpy, psc->scr, pcp->hwContextID); +} + +static Bool driBindContext(__GLXDRIcontext *context, + __GLXDRIdrawable *draw, __GLXDRIdrawable *read) +{ + __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; + const __DRIcoreExtension *core = pcp->psc->core; + + return (*core->bindContext)(pcp->driContext, + draw->driDrawable, + read->driDrawable); +} + +static void driUnbindContext(__GLXDRIcontext *context) +{ + __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; + const __DRIcoreExtension *core = pcp->psc->core; + + (*core->unbindContext)(pcp->driContext); +} + +static __GLXDRIcontext *driCreateContext(__GLXscreenConfigs *psc, + const __GLcontextModes *mode, + GLXContext gc, + GLXContext shareList, int renderType) +{ + __GLXDRIcontextPrivate *pcp, *pcp_shared; + drm_context_t hwContext; + __DRIcontext *shared = NULL; + __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode; + + if (!psc || !psc->driScreen) + return NULL; + + if (shareList) { + pcp_shared = (__GLXDRIcontextPrivate *) shareList->driContext; + shared = pcp_shared->driContext; + } + + pcp = Xmalloc(sizeof *pcp); + if (pcp == NULL) + return NULL; + + pcp->psc = psc; + if (!XF86DRICreateContextWithConfig(psc->dpy, psc->scr, + mode->visualID, + &pcp->hwContextID, &hwContext)) { + Xfree(pcp); + return NULL; } + + pcp->driContext = + (*psc->legacy->createNewContext)(psc->__driScreen, + config->driConfig, + renderType, + shared, + hwContext, + pcp); + if (pcp->driContext == NULL) { + XF86DRIDestroyContext(psc->dpy, psc->scr, pcp->hwContextID); + Xfree(pcp); + return NULL; + } + + pcp->base.destroyContext = driDestroyContext; + pcp->base.bindContext = driBindContext; + pcp->base.unbindContext = driUnbindContext; + + return &pcp->base; } +static void driDestroyDrawable(__GLXDRIdrawable *pdraw) +{ + __GLXscreenConfigs *psc = pdraw->psc; + + (*psc->core->destroyDrawable)(pdraw->driDrawable); + XF86DRIDestroyDrawable(psc->dpy, psc->scr, pdraw->drawable); + Xfree(pdraw); +} + +static __GLXDRIdrawable *driCreateDrawable(__GLXscreenConfigs *psc, + XID xDrawable, + GLXDrawable drawable, + const __GLcontextModes *modes) +{ + __GLXDRIdrawable *pdraw; + drm_drawable_t hwDrawable; + void *empty_attribute_list = NULL; + __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes; + + /* Old dri can't handle GLX 1.3+ drawable constructors. */ + if (xDrawable != drawable) + return NULL; + + pdraw = Xmalloc(sizeof(*pdraw)); + if (!pdraw) + return NULL; + + pdraw->drawable = drawable; + pdraw->psc = psc; + + if (!XF86DRICreateDrawable(psc->dpy, psc->scr, drawable, &hwDrawable)) + return NULL; + + /* Create a new drawable */ + pdraw->driDrawable = + (*psc->legacy->createNewDrawable)(psc->__driScreen, + config->driConfig, + hwDrawable, + GLX_WINDOW_BIT, + empty_attribute_list, + pdraw); + + if (!pdraw->driDrawable) { + XF86DRIDestroyDrawable(psc->dpy, psc->scr, drawable); + Xfree(pdraw); + return NULL; + } + + pdraw->destroyDrawable = driDestroyDrawable; + + return pdraw; +} + +static void driDestroyScreen(__GLXscreenConfigs *psc) +{ + /* Free the direct rendering per screen data */ + if (psc->__driScreen) + (*psc->core->destroyScreen)(psc->__driScreen); + psc->__driScreen = NULL; + if (psc->driver) + dlclose(psc->driver); +} + +static __GLXDRIscreen *driCreateScreen(__GLXscreenConfigs *psc, int screen, + __GLXdisplayPrivate *priv) +{ + __GLXDRIdisplayPrivate *pdp; + __GLXDRIscreen *psp; + const __DRIextension **extensions; + char *driverName; + int i; + + psp = Xmalloc(sizeof *psp); + if (psp == NULL) + return NULL; + + /* Initialize per screen dynamic client GLX extensions */ + psc->ext_list_first_time = GL_TRUE; + + if (!driGetDriverName(priv->dpy, screen, &driverName)) { + Xfree(psp); + return NULL; + } + + psc->driver = driOpenDriver(driverName); + Xfree(driverName); + if (psc->driver == NULL) { + Xfree(psp); + return NULL; + } + + extensions = dlsym(psc->driver, __DRI_DRIVER_EXTENSIONS); + if (extensions == NULL) { + ErrorMessageF("driver exports no extensions (%s)\n", dlerror()); + Xfree(psp); + return NULL; + } + + for (i = 0; extensions[i]; i++) { + if (strcmp(extensions[i]->name, __DRI_CORE) == 0) + psc->core = (__DRIcoreExtension *) extensions[i]; + if (strcmp(extensions[i]->name, __DRI_LEGACY) == 0) + psc->legacy = (__DRIlegacyExtension *) extensions[i]; + } + + if (psc->core == NULL || psc->legacy == NULL) { + Xfree(psp); + return NULL; + } + + pdp = (__GLXDRIdisplayPrivate *) priv->driDisplay; + psc->__driScreen = + CallCreateNewScreen(psc->dpy, screen, psc, pdp); + if (psc->__driScreen == NULL) { + dlclose(psc->driver); + Xfree(psp); + return NULL; + } + + driBindExtensions(psc); + + psp->destroyScreen = driDestroyScreen; + psp->createContext = driCreateContext; + psp->createDrawable = driCreateDrawable; + + return psp; +} + +/* Called from __glXFreeDisplayPrivate. + */ +static void driDestroyDisplay(__GLXDRIdisplay *dpy) +{ + Xfree(dpy); +} /* * Allocate, initialize and return a __DRIdisplayPrivate object. * This is called from __glXInitialize() when we are given a new * display pointer. */ -void *driCreateDisplay(Display *dpy, __DRIdisplay *pdisp) +_X_HIDDEN __GLXDRIdisplay *driCreateDisplay(Display *dpy) { - const int numScreens = ScreenCount(dpy); - __DRIdisplayPrivate *pdpyp; + __GLXDRIdisplayPrivate *pdpyp; int eventBase, errorBase; int major, minor, patch; - int scrn; - - /* Initialize these fields to NULL in case we fail. - * If we don't do this we may later get segfaults trying to free random - * addresses when the display is closed. - */ - pdisp->private = NULL; - pdisp->destroyDisplay = NULL; if (!XF86DRIQueryExtension(dpy, &eventBase, &errorBase)) { return NULL; @@ -445,7 +673,7 @@ void *driCreateDisplay(Display *dpy, __DRIdisplay *pdisp) return NULL; } - pdpyp = (__DRIdisplayPrivate *)Xmalloc(sizeof(__DRIdisplayPrivate)); + pdpyp = Xmalloc(sizeof *pdpyp); if (!pdpyp) { return NULL; } @@ -454,41 +682,10 @@ void *driCreateDisplay(Display *dpy, __DRIdisplay *pdisp) pdpyp->driMinor = minor; pdpyp->driPatch = patch; - pdisp->destroyDisplay = driDestroyDisplay; - - /* allocate array of pointers to createNewScreen funcs */ - pdisp->createNewScreen = (PFNCREATENEWSCREENFUNC *) - Xmalloc(numScreens * sizeof(void *)); - if (!pdisp->createNewScreen) { - Xfree(pdpyp); - return NULL; - } - - /* allocate array of library handles */ - pdpyp->libraryHandles = (void **) Xmalloc(numScreens * sizeof(void*)); - if (!pdpyp->libraryHandles) { - Xfree(pdisp->createNewScreen); - Xfree(pdpyp); - return NULL; - } - - /* dynamically discover DRI drivers for all screens, saving each - * driver's "__driCreateScreen" function pointer. That's the bootstrap - * entrypoint for all DRI drivers. - */ - for (scrn = 0; scrn < numScreens; scrn++) { - __DRIdriver *driver = driGetDriver(dpy, scrn); - if (driver) { - pdisp->createNewScreen[scrn] = driver->createNewScreenFunc; - pdpyp->libraryHandles[scrn] = driver->handle; - } - else { - pdisp->createNewScreen[scrn] = NULL; - pdpyp->libraryHandles[scrn] = NULL; - } - } + pdpyp->base.destroyDisplay = driDestroyDisplay; + pdpyp->base.createScreen = driCreateScreen; - return (void *)pdpyp; + return &pdpyp->base; } #endif /* GLX_DIRECT_RENDERING */ diff --git a/src/glx/x11/glcontextmodes.c b/src/glx/x11/glcontextmodes.c index 788ecf6a3a..326c8b2357 100644 --- a/src/glx/x11/glcontextmodes.c +++ b/src/glx/x11/glcontextmodes.c @@ -336,7 +336,8 @@ _gl_get_context_mode_data(const __GLcontextModes *mode, int attribute, *value_return = mode->bindToTextureRgba; return 0; case GLX_BIND_TO_MIPMAP_TEXTURE_EXT: - *value_return = mode->bindToMipmapTexture; + *value_return = mode->bindToMipmapTexture == GL_TRUE ? GL_TRUE : + GL_FALSE; return 0; case GLX_BIND_TO_TEXTURE_TARGETS_EXT: *value_return = mode->bindToTextureTargets; @@ -417,7 +418,7 @@ _gl_context_modes_create( unsigned count, size_t minimum_size ) (*next)->bindToTextureRgb = GLX_DONT_CARE; (*next)->bindToTextureRgba = GLX_DONT_CARE; (*next)->bindToMipmapTexture = GLX_DONT_CARE; - (*next)->bindToTextureTargets = 0; + (*next)->bindToTextureTargets = GLX_DONT_CARE; (*next)->yInverted = GLX_DONT_CARE; next = & ((*next)->next); @@ -456,19 +457,28 @@ _gl_context_modes_destroy( __GLcontextModes * modes ) */ __GLcontextModes * -_gl_context_modes_find_visual( __GLcontextModes * modes, int vid ) +_gl_context_modes_find_visual(__GLcontextModes *modes, int vid) { - while ( modes != NULL ) { - if ( modes->visualID == vid ) { - break; - } + __GLcontextModes *m; - modes = modes->next; - } + for (m = modes; m != NULL; m = m->next) + if (m->visualID == vid) + return m; - return modes; + return NULL; } +__GLcontextModes * +_gl_context_modes_find_fbconfig(__GLcontextModes *modes, int fbid) +{ + __GLcontextModes *m; + + for (m = modes; m != NULL; m = m->next) + if (m->fbconfigID == fbid) + return m; + + return NULL; +} /** * Determine if two context-modes are the same. This is intended to be used diff --git a/src/glx/x11/glcontextmodes.h b/src/glx/x11/glcontextmodes.h index 4b5c6f68b8..afd09cd7fb 100644 --- a/src/glx/x11/glcontextmodes.h +++ b/src/glx/x11/glcontextmodes.h @@ -44,8 +44,10 @@ extern int _gl_get_context_mode_data( const __GLcontextModes *mode, extern __GLcontextModes * _gl_context_modes_create( unsigned count, size_t minimum_size ); extern void _gl_context_modes_destroy( __GLcontextModes * modes ); -extern __GLcontextModes * _gl_context_modes_find_visual( - __GLcontextModes * modes, int vid ); +extern __GLcontextModes * + _gl_context_modes_find_visual(__GLcontextModes *modes, int vid); +extern __GLcontextModes * + _gl_context_modes_find_fbconfig(__GLcontextModes *modes, int fbid); extern GLboolean _gl_context_modes_are_same( const __GLcontextModes * a, const __GLcontextModes * b ); diff --git a/src/glx/x11/glx_pbuffer.c b/src/glx/x11/glx_pbuffer.c index 1df2d0f342..0f878f223f 100644 --- a/src/glx/x11/glx_pbuffer.c +++ b/src/glx/x11/glx_pbuffer.c @@ -164,6 +164,33 @@ DestroyPbuffer( Display * dpy, GLXDrawable drawable ) } +#ifdef GLX_DIRECT_RENDERING +extern __GLXDRIdrawable * +GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num); + +static GLenum +determineTextureTarget(const int *attribs, int numAttribs) +{ + GLenum target = 0; + int i; + + for (i = 0; i < numAttribs; i++) { + if (attribs[2 * i] == GLX_TEXTURE_TARGET_EXT) { + switch (attribs[2 * i + 1]) { + case GLX_TEXTURE_2D_EXT: + target = GL_TEXTURE_2D; + break; + case GLX_TEXTURE_RECTANGLE_EXT: + target = GL_TEXTURE_RECTANGLE_ARB; + break; + } + } + } + + return target; +} +#endif + /** * Get a drawable's attribute. * @@ -261,6 +288,16 @@ GetDrawableAttribute( Display *dpy, GLXDrawable drawable, } } +#ifdef GLX_DIRECT_RENDERING + { + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL); + + if (pdraw != NULL && !pdraw->textureTarget) + pdraw->textureTarget = determineTextureTarget((const int *)data, + num_attributes); + } +#endif + Xfree( data ); } } @@ -271,7 +308,6 @@ GetDrawableAttribute( Display *dpy, GLXDrawable drawable, return 0; } - /** * Create a non-pbuffer GLX drawable. * @@ -306,7 +342,7 @@ CreateDrawable( Display *dpy, const __GLcontextModes * fbconfig, req->glxCode = glxCode; req->screen = (CARD32) fbconfig->screen; req->fbconfig = fbconfig->fbconfigID; - req->window = (GLXPbuffer) drawable; + req->window = (CARD32) drawable; req->glxwindow = (GLXWindow) XAllocID(dpy); req->numAttribs = (CARD32) i; @@ -315,6 +351,34 @@ CreateDrawable( Display *dpy, const __GLcontextModes * fbconfig, UnlockDisplay(dpy); SyncHandle(); +#ifdef GLX_DIRECT_RENDERING + do { + /* FIXME: Maybe delay __DRIdrawable creation until the drawable + * is actually bound to a context... */ + + __GLXdisplayPrivate * const priv = __glXInitialize(dpy); + __GLXDRIdrawable *pdraw; + __GLXscreenConfigs *psc; + + psc = &priv->screenConfigs[fbconfig->screen]; + if (psc->driScreen == NULL) + break; + pdraw = psc->driScreen->createDrawable(psc, drawable, + req->glxwindow, fbconfig); + if (pdraw == NULL) { + fprintf(stderr, "failed to create drawable\n"); + break; + } + + if (__glxHashInsert(psc->drawHash, req->glxwindow, pdraw)) { + (*pdraw->destroyDrawable)(pdraw); + return None; /* FIXME: Check what we're supposed to do here... */ + } + + pdraw->textureTarget = determineTextureTarget(attrib_list, i); + } while (0); +#endif + return (GLXDrawable)req->glxwindow; } @@ -350,6 +414,20 @@ DestroyDrawable( Display * dpy, GLXDrawable drawable, CARD32 glxCode ) UnlockDisplay(dpy); SyncHandle(); +#ifdef GLX_DIRECT_RENDERING + { + int screen; + __GLXdisplayPrivate * const priv = __glXInitialize(dpy); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); + __GLXscreenConfigs *psc = &priv->screenConfigs[screen]; + + if (pdraw != NULL) { + (*pdraw->destroyDrawable)(pdraw); + __glxHashDelete(psc->drawHash, drawable); + } + } +#endif + return; } @@ -460,8 +538,24 @@ glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, PUBLIC GLXPbuffer glXCreatePbuffer(Display *dpy, GLXFBConfig config, const int *attrib_list) { + int i, width, height; + + width = 0; + height = 0; + + for (i = 0; attrib_list[i * 2]; i++) { + switch (attrib_list[i * 2]) { + case GLX_PBUFFER_WIDTH: + width = attrib_list[i * 2 + 1]; + break; + case GLX_PBUFFER_HEIGHT: + height = attrib_list[i * 2 + 1]; + break; + } + } + return (GLXPbuffer) CreatePbuffer( dpy, (__GLcontextModes *) config, - 0, 0, + width, height, attrib_list, GL_TRUE ); } diff --git a/src/glx/x11/glx_texture_compression.c b/src/glx/x11/glx_texture_compression.c deleted file mode 100644 index 5676858017..0000000000 --- a/src/glx/x11/glx_texture_compression.c +++ /dev/null @@ -1,347 +0,0 @@ -/* - * (C) Copyright IBM Corporation 2004 - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file glx_texture_compression.c - * Contains the routines required to implement GLX protocol for - * ARB_texture_compression and related extensions. - * - * \sa http://oss.sgi.com/projects/ogl-sample/registry/ARB/texture_compression.txt - * - * \author Ian Romanick - */ - -#include "packrender.h" -#include "packsingle.h" -#include "indirect.h" - -#include - - -void -__indirect_glGetCompressedTexImageARB( GLenum target, GLint level, - GLvoid * img ) -{ - __GLX_SINGLE_DECLARE_VARIABLES(); - xGLXGetTexImageReply reply; - size_t image_bytes; - - __GLX_SINGLE_LOAD_VARIABLES(); - __GLX_SINGLE_BEGIN( X_GLsop_GetCompressedTexImage, 8 ); - __GLX_SINGLE_PUT_LONG( 0, target ); - __GLX_SINGLE_PUT_LONG( 4, level ); - __GLX_SINGLE_READ_XREPLY(); - - image_bytes = reply.width; - assert( image_bytes <= ((4 * reply.length) - 0) ); - assert( image_bytes >= ((4 * reply.length) - 3) ); - - if ( image_bytes != 0 ) { - _XRead( dpy, (char *) img, image_bytes ); - if ( image_bytes < (4 * reply.length) ) { - _XEatData( dpy, (4 * reply.length) - image_bytes ); - } - } - - __GLX_SINGLE_END(); -} - - -/** - * Internal function used for \c glCompressedTexImage1D and - * \c glCompressedTexImage2D. - */ -static void -CompressedTexImage1D2D( GLenum target, GLint level, - GLenum internal_format, - GLsizei width, GLsizei height, - GLint border, GLsizei image_size, - const GLvoid *data, CARD32 rop ) -{ - __GLX_DECLARE_VARIABLES(); - - __GLX_LOAD_VARIABLES(); - if ( gc->currentDpy == NULL ) { - return; - } - - if ( (target == GL_PROXY_TEXTURE_1D) - || (target == GL_PROXY_TEXTURE_2D) - || (target == GL_PROXY_TEXTURE_CUBE_MAP) ) { - compsize = 0; - } - else { - compsize = image_size; - } - - cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE - + compsize ); - if ( cmdlen <= gc->maxSmallRenderCommandSize ) { - __GLX_BEGIN_VARIABLE( rop, cmdlen ); - __GLX_PUT_LONG( 4, target ); - __GLX_PUT_LONG( 8, level ); - __GLX_PUT_LONG( 12, internal_format ); - __GLX_PUT_LONG( 16, width ); - __GLX_PUT_LONG( 20, height ); - __GLX_PUT_LONG( 24, border ); - __GLX_PUT_LONG( 28, image_size ); - if ( compsize != 0 ) { - __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE, - data, image_size ); - } - __GLX_END( cmdlen ); - } - else { - assert( compsize != 0 ); - - __GLX_BEGIN_VARIABLE_LARGE( rop, cmdlen + 4 ); - __GLX_PUT_LONG( 8, target ); - __GLX_PUT_LONG( 12, level ); - __GLX_PUT_LONG( 16, internal_format ); - __GLX_PUT_LONG( 20, width ); - __GLX_PUT_LONG( 24, height ); - __GLX_PUT_LONG( 28, border ); - __GLX_PUT_LONG( 32, image_size ); - __glXSendLargeCommand( gc, gc->pc, - __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE + 4, - data, image_size ); - } -} - - -/** - * Internal function used for \c glCompressedTexSubImage1D and - * \c glCompressedTexSubImage2D. - */ -static void -CompressedTexSubImage1D2D( GLenum target, GLint level, - GLsizei xoffset, GLsizei yoffset, - GLsizei width, GLsizei height, - GLenum format, GLsizei image_size, - const GLvoid *data, CARD32 rop ) -{ - __GLX_DECLARE_VARIABLES(); - - __GLX_LOAD_VARIABLES(); - if ( gc->currentDpy == NULL ) { - return; - } - - if ( target == GL_PROXY_TEXTURE_3D ) { - compsize = 0; - } - else { - compsize = image_size; - } - - cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE - + compsize ); - if ( cmdlen <= gc->maxSmallRenderCommandSize ) { - __GLX_BEGIN_VARIABLE( rop, cmdlen ); - __GLX_PUT_LONG( 4, target ); - __GLX_PUT_LONG( 8, level ); - __GLX_PUT_LONG( 12, xoffset ); - __GLX_PUT_LONG( 16, yoffset ); - __GLX_PUT_LONG( 20, width ); - __GLX_PUT_LONG( 24, height ); - __GLX_PUT_LONG( 28, format ); - __GLX_PUT_LONG( 32, image_size ); - if ( compsize != 0 ) { - __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE, - data, image_size ); - } - __GLX_END( cmdlen ); - } - else { - assert( compsize != 0 ); - - __GLX_BEGIN_VARIABLE_LARGE( rop, cmdlen + 4 ); - __GLX_PUT_LONG( 8, target ); - __GLX_PUT_LONG( 12, level ); - __GLX_PUT_LONG( 16, xoffset ); - __GLX_PUT_LONG( 20, yoffset ); - __GLX_PUT_LONG( 24, width ); - __GLX_PUT_LONG( 28, height ); - __GLX_PUT_LONG( 32, format ); - __GLX_PUT_LONG( 36, image_size ); - __glXSendLargeCommand( gc, gc->pc, - __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE + 4, - data, image_size ); - } -} - - -void -__indirect_glCompressedTexImage1DARB( GLenum target, GLint level, - GLenum internal_format, GLsizei width, - GLint border, GLsizei image_size, - const GLvoid *data ) -{ - CompressedTexImage1D2D( target, level, internal_format, width, 0, - border, image_size, data, - X_GLrop_CompressedTexImage1D ); -} - - -void -__indirect_glCompressedTexImage2DARB( GLenum target, GLint level, - GLenum internal_format, - GLsizei width, GLsizei height, - GLint border, GLsizei image_size, - const GLvoid *data ) -{ - CompressedTexImage1D2D( target, level, internal_format, width, height, - border, image_size, data, - X_GLrop_CompressedTexImage2D ); -} - - -void -__indirect_glCompressedTexImage3DARB( GLenum target, GLint level, - GLenum internal_format, - GLsizei width, GLsizei height, GLsizei depth, - GLint border, GLsizei image_size, - const GLvoid *data ) -{ - __GLX_DECLARE_VARIABLES(); - - __GLX_LOAD_VARIABLES(); - if ( gc->currentDpy == NULL ) { - return; - } - - cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE - + image_size ); - if ( cmdlen <= gc->maxSmallRenderCommandSize ) { - __GLX_BEGIN_VARIABLE( X_GLrop_CompressedTexImage3D, cmdlen ); - __GLX_PUT_LONG( 4, target ); - __GLX_PUT_LONG( 8, level ); - __GLX_PUT_LONG( 12, internal_format ); - __GLX_PUT_LONG( 16, width ); - __GLX_PUT_LONG( 20, height ); - __GLX_PUT_LONG( 24, depth ); - __GLX_PUT_LONG( 28, border ); - __GLX_PUT_LONG( 32, image_size ); - if ( image_size != 0 ) { - __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE, - data, image_size ); - } - __GLX_END( cmdlen ); - } - else { - __GLX_BEGIN_VARIABLE_LARGE( X_GLrop_CompressedTexImage3D, - cmdlen + 4 ); - __GLX_PUT_LONG( 8, target ); - __GLX_PUT_LONG( 12, level ); - __GLX_PUT_LONG( 16, internal_format ); - __GLX_PUT_LONG( 20, width ); - __GLX_PUT_LONG( 24, height ); - __GLX_PUT_LONG( 28, depth ); - __GLX_PUT_LONG( 32, border ); - __GLX_PUT_LONG( 36, image_size ); - __glXSendLargeCommand( gc, gc->pc, - __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE + 4, - data, image_size ); - } -} - - -void -__indirect_glCompressedTexSubImage1DARB( GLenum target, GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLsizei image_size, - const GLvoid *data ) -{ - CompressedTexSubImage1D2D( target, level, xoffset, 0, width, 0, - format, image_size, data, - X_GLrop_CompressedTexSubImage1D ); -} - - -void -__indirect_glCompressedTexSubImage2DARB( GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLsizei image_size, - const GLvoid *data ) -{ - CompressedTexSubImage1D2D( target, level, xoffset, yoffset, width, height, - format, image_size, data, - X_GLrop_CompressedTexSubImage2D ); -} - - -void -__indirect_glCompressedTexSubImage3DARB( GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLsizei image_size, - const GLvoid *data ) -{ - __GLX_DECLARE_VARIABLES(); - - __GLX_LOAD_VARIABLES(); - if ( gc->currentDpy == NULL ) { - return; - } - - cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE - + image_size ); - if ( cmdlen <= gc->maxSmallRenderCommandSize ) { - __GLX_BEGIN_VARIABLE( X_GLrop_CompressedTexSubImage3D, cmdlen ); - __GLX_PUT_LONG( 4, target ); - __GLX_PUT_LONG( 8, level ); - __GLX_PUT_LONG( 12, xoffset ); - __GLX_PUT_LONG( 16, yoffset ); - __GLX_PUT_LONG( 20, zoffset ); - __GLX_PUT_LONG( 24, width ); - __GLX_PUT_LONG( 28, height ); - __GLX_PUT_LONG( 32, depth ); - __GLX_PUT_LONG( 36, format ); - __GLX_PUT_LONG( 40, image_size ); - if ( image_size != 0 ) { - __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE, - data, image_size ); - } - __GLX_END( cmdlen ); - } - else { - __GLX_BEGIN_VARIABLE_LARGE( X_GLrop_CompressedTexSubImage3D, - cmdlen + 4 ); - __GLX_PUT_LONG( 8, target ); - __GLX_PUT_LONG( 12, level ); - __GLX_PUT_LONG( 16, xoffset ); - __GLX_PUT_LONG( 20, yoffset ); - __GLX_PUT_LONG( 24, zoffset ); - __GLX_PUT_LONG( 28, width ); - __GLX_PUT_LONG( 32, height ); - __GLX_PUT_LONG( 36, depth ); - __GLX_PUT_LONG( 40, format ); - __GLX_PUT_LONG( 44, image_size ); - __glXSendLargeCommand( gc, gc->pc, - __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE + 4, - data, image_size ); - } -} diff --git a/src/glx/x11/glxclient.h b/src/glx/x11/glxclient.h index 03e44e5d04..73c278ee38 100644 --- a/src/glx/x11/glxclient.h +++ b/src/glx/x11/glxclient.h @@ -58,7 +58,7 @@ #include "GL/glxproto.h" #include "GL/internal/glcore.h" #include "glapitable.h" -#include "glxextensions.h" +#include "glxhash.h" #if defined( USE_XTHREADS ) # include #elif defined( PTHREADS ) @@ -70,7 +70,9 @@ #define __GLX_MAX_TEXTURE_UNITS 32 +typedef struct __GLXscreenConfigsRec __GLXscreenConfigs; typedef struct __GLXcontextRec __GLXcontext; +typedef struct __GLXdrawableRec __GLXdrawable; typedef struct __GLXdisplayPrivateRec __GLXdisplayPrivate; typedef struct _glapi_table __GLapi; @@ -78,6 +80,9 @@ typedef struct _glapi_table __GLapi; #ifdef GLX_DIRECT_RENDERING +#define containerOf(ptr, type, member) \ + (type *)( (char *)ptr - offsetof(type,member) ) + #include @@ -85,43 +90,64 @@ typedef struct _glapi_table __GLapi; * Display dependent methods. This structure is initialized during the * \c driCreateDisplay call. */ -struct __DRIdisplayRec { +typedef struct __GLXDRIdisplayRec __GLXDRIdisplay; +typedef struct __GLXDRIscreenRec __GLXDRIscreen; +typedef struct __GLXDRIdrawableRec __GLXDRIdrawable; +typedef struct __GLXDRIcontextRec __GLXDRIcontext; + +#include "glxextensions.h" + +struct __GLXDRIdisplayRec { /** * Method to destroy the private DRI display data. */ - void (*destroyDisplay)(Display *dpy, void *displayPrivate); + void (*destroyDisplay)(__GLXDRIdisplay *display); - /** - * Opaque pointer to private per display direct rendering data. - * \c NULL if direct rendering is not supported on this display. - */ - struct __DRIdisplayPrivateRec *private; + __GLXDRIscreen *(*createScreen)(__GLXscreenConfigs *psc, int screen, + __GLXdisplayPrivate *priv); +}; - /** - * Array of pointers to methods to create and initialize the private DRI - * screen data. - */ - PFNCREATENEWSCREENFUNC * createNewScreen; +struct __GLXDRIscreenRec { + + void (*destroyScreen)(__GLXscreenConfigs *psc); + + __GLXDRIcontext *(*createContext)(__GLXscreenConfigs *psc, + const __GLcontextModes *mode, + GLXContext gc, + GLXContext shareList, int renderType); + + __GLXDRIdrawable *(*createDrawable)(__GLXscreenConfigs *psc, + XID drawable, + GLXDrawable glxDrawable, + const __GLcontextModes *modes); }; +struct __GLXDRIcontextRec { + void (*destroyContext)(__GLXDRIcontext *context, __GLXscreenConfigs *psc, + Display *dpy); + Bool (*bindContext)(__GLXDRIcontext *context, + __GLXDRIdrawable *pdraw, + __GLXDRIdrawable *pread); + + void (*unbindContext)(__GLXDRIcontext *context); +}; -/* -** We keep a linked list of these structures, one per DRI device driver. -*/ -struct __DRIdriverRec { - const char *name; - void *handle; - PFNCREATENEWSCREENFUNC createNewScreenFunc; - struct __DRIdriverRec *next; +struct __GLXDRIdrawableRec { + void (*destroyDrawable)(__GLXDRIdrawable *drawable); + + XID xDrawable; + XID drawable; + __GLXscreenConfigs *psc; + __DRIdrawable *driDrawable; + GLenum textureTarget; }; /* ** Function to create and DRI display data and initialize the display ** dependent methods. */ -extern void *driCreateDisplay(Display *dpy, __DRIdisplay *pdisp); - -extern __DRIdriver *driGetDriver(Display *dpy, int scrNum); +extern __GLXDRIdisplay *driCreateDisplay(Display *dpy); +extern __GLXDRIdisplay *dri2CreateDisplay(Display *dpy); extern void DRI_glXUseXFont( Font font, int first, int count, int listbase ); @@ -133,8 +159,6 @@ extern const char *glXGetScreenDriver (Display *dpy, int scrNum); extern const char *glXGetDriverConfig (const char *driverName); -extern Bool __glXWindowExists(Display *dpy, GLXDrawable draw); - #endif /************************************************************************/ @@ -225,19 +249,11 @@ struct __GLXcontextRec { */ XID share_xid; - /** - * Visual id. - * - * \deprecated - * This filed has been largely been replaced by the \c mode field, but - * the work is not quite done. - */ - VisualID vid; - /** * Screen number. */ GLint screen; + __GLXscreenConfigs *psc; /** * \c GL_TRUE if the context was created with ImportContext, which @@ -343,24 +359,15 @@ struct __GLXcontextRec { */ GLint majorOpcode; -#ifdef GLX_DIRECT_RENDERING /** - * Per context direct rendering interface functions and data. + * Pointer to the mode used to create this context. */ - __DRIcontext driContext; + const __GLcontextModes * mode; + +#ifdef GLX_DIRECT_RENDERING + __GLXDRIcontext *driContext; + __DRIcontext *__driContext; #endif - - /** - * \c GLXFBConfigID used to create this context. May be \c None. This - * field has been replaced by the \c mode field. - * - * \since Internal API version 20030317. - * - * \deprecated - * This filed has been largely been replaced by the \c mode field, but - * the work is not quite done. - */ - GLXFBConfigID fbconfigID; /** * The current read-drawable for this context. Will be None if this @@ -438,7 +445,7 @@ extern void __glFreeAttributeState(__GLXcontext *); * One of these records exists per screen of the display. It contains * a pointer to the config data for that screen (if the screen supports GL). */ -typedef struct __GLXscreenConfigsRec { +struct __GLXscreenConfigsRec { /** * GLX extension string reported by the X-server. */ @@ -454,13 +461,46 @@ typedef struct __GLXscreenConfigsRec { /** * Per screen direct rendering interface functions and data. */ - __DRIscreen driScreen; + __DRIscreen *__driScreen; + const __DRIcoreExtension *core; + const __DRIlegacyExtension *legacy; + __glxHashTable *drawHash; + Display *dpy; + int scr, fd; + void *driver; + + __GLXDRIscreen *driScreen; + +#ifdef __DRI_COPY_SUB_BUFFER + const __DRIcopySubBufferExtension *copySubBuffer; +#endif + +#ifdef __DRI_SWAP_CONTROL + const __DRIswapControlExtension *swapControl; +#endif + +#ifdef __DRI_ALLOCATE + const __DRIallocateExtension *allocate; +#endif + +#ifdef __DRI_FRAME_TRACKING + const __DRIframeTrackingExtension *frameTracking; +#endif + +#ifdef __DRI_MEDIA_STREAM_COUNTER + const __DRImediaStreamCounterExtension *msc; +#endif + +#ifdef __DRI_TEX_BUFFER + const __DRItexBufferExtension *texBuffer; +#endif + #endif /** - * Linked list of configurations for this screen. + * Linked list of glx visuals and fbconfigs for this screen. */ - __GLcontextModes *configs; + __GLcontextModes *visuals, *configs; /** * Per-screen dynamic GLX extension tracking. The \c direct_support @@ -474,7 +514,7 @@ typedef struct __GLXscreenConfigsRec { GLboolean ext_list_first_time; /*@}*/ -} __GLXscreenConfigs; +}; /** * Per display private data. One of these records exists for each display @@ -523,11 +563,11 @@ struct __GLXdisplayPrivateRec { /** * Per display direct rendering interface functions and data. */ - __DRIdisplay driDisplay; + __GLXDRIdisplay *driDisplay; + __GLXDRIdisplay *dri2Display; #endif }; -void __glXFreeContext(__GLXcontext*); extern GLubyte *__glXFlushRenderBuffer(__GLXcontext*, GLubyte*); @@ -571,6 +611,10 @@ extern __GLXcontext *__glXcurrentContext; #endif /* defined( USE_XTHREADS ) || defined( PTHREADS ) */ +extern void __glXSetCurrentContextNull(void); + +extern void __glXFreeContext(__GLXcontext*); + /* ** Global lock for all threads in this address space using the GLX @@ -680,13 +724,16 @@ extern char *__glXstrdup(const char *str); extern const char __glXGLClientVersion[]; extern const char __glXGLClientExtensions[]; -/* Determine the internal API version */ -extern int __glXGetInternalVersion(void); - /* Get the unadjusted system time */ extern int __glXGetUST( int64_t * ust ); -extern Bool __glXGetMscRateOML(Display * dpy, GLXDrawable drawable, - int32_t * numerator, int32_t * denominator); +extern GLboolean __glXGetMscRateOML(Display * dpy, GLXDrawable drawable, + int32_t * numerator, int32_t * denominator); + +#ifdef GLX_DIRECT_RENDERING +GLboolean +__driGetMscRateOML(__DRIdrawable *draw, + int32_t *numerator, int32_t *denominator, void *private); +#endif #endif /* !__GLX_client_h__ */ diff --git a/src/glx/x11/glxcmds.c b/src/glx/x11/glxcmds.c index 80281896f6..4345678a98 100644 --- a/src/glx/x11/glxcmds.c +++ b/src/glx/x11/glxcmds.c @@ -38,63 +38,110 @@ * Client-side GLX interface. */ -#include #include "glxclient.h" -#include -#include -#include -#include #include "glapi.h" -#ifdef GLX_DIRECT_RENDERING -#include "indirect_init.h" -#include -#include "xf86dri.h" -#endif #include "glxextensions.h" #include "glcontextmodes.h" #include "glheader.h" + +#ifdef GLX_DIRECT_RENDERING #include +#include +#include "xf86dri.h" +#endif static const char __glXGLXClientVendorName[] = "SGI"; static const char __glXGLXClientVersion[] = "1.4"; /****************************************************************************/ + +#ifdef GLX_DIRECT_RENDERING + +static Bool windowExistsFlag; +static int windowExistsErrorHandler(Display *dpy, XErrorEvent *xerr) +{ + if (xerr->error_code == BadWindow) { + windowExistsFlag = GL_FALSE; + } + return 0; +} + +/** + * Find drawables in the local hash that have been destroyed on the + * server. + * + * \param dpy Display to destroy drawables for + * \param screen Screen number to destroy drawables for + */ +static void GarbageCollectDRIDrawables(Display *dpy, __GLXscreenConfigs *sc) +{ + XID draw; + __GLXDRIdrawable *pdraw; + XWindowAttributes xwa; + int (*oldXErrorHandler)(Display *, XErrorEvent *); + + /* Set no-op error handler so Xlib doesn't bail out if the windows + * has alreay been destroyed on the server. */ + XSync(dpy, GL_FALSE); + oldXErrorHandler = XSetErrorHandler(windowExistsErrorHandler); + + if (__glxHashFirst(sc->drawHash, &draw, (void *)&pdraw) == 1) { + do { + windowExistsFlag = GL_TRUE; + XGetWindowAttributes(dpy, draw, &xwa); /* dummy request */ + if (!windowExistsFlag) { + /* Destroy the local drawable data, if the drawable no + longer exists in the Xserver */ + (*pdraw->destroyDrawable)(pdraw); + __glxHashDelete(sc->drawHash, draw); + } + } while (__glxHashNext(sc->drawHash, &draw, (void *)&pdraw) == 1); + } + + XSync(dpy, GL_FALSE); + XSetErrorHandler(oldXErrorHandler); +} + +extern __GLXDRIdrawable * +GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num); + /** * Get the __DRIdrawable for the drawable associated with a GLXContext * * \param dpy The display associated with \c drawable. * \param drawable GLXDrawable whose __DRIdrawable part is to be retrieved. + * \param scrn_num If non-NULL, the drawables screen is stored there * \returns A pointer to the context's __DRIdrawable on success, or NULL if * the drawable is not associated with a direct-rendering context. */ - -#ifdef GLX_DIRECT_RENDERING -static __DRIdrawable * -GetDRIDrawable( Display *dpy, GLXDrawable drawable, int * const scrn_num ) +_X_HIDDEN __GLXDRIdrawable * +GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num) { - __GLXdisplayPrivate * const priv = __glXInitialize(dpy); - - if ( (priv != NULL) && (priv->driDisplay.private != NULL) ) { - const unsigned screen_count = ScreenCount(dpy); - unsigned i; - - for ( i = 0 ; i < screen_count ; i++ ) { - __DRIscreen * const psc = &priv->screenConfigs[i].driScreen; - __DRIdrawable * const pdraw = (psc->private != NULL) - ? (*psc->getDrawable)(dpy, drawable, psc->private) : NULL; + __GLXdisplayPrivate *priv = __glXInitialize(dpy); + __GLXDRIdrawable *pdraw; + const unsigned screen_count = ScreenCount(dpy); + unsigned i; + __GLXscreenConfigs *psc; - if ( pdraw != NULL ) { - if ( scrn_num != NULL ) { - *scrn_num = i; - } - return pdraw; - } + if (priv == NULL) + return NULL; + + for (i = 0; i < screen_count; i++) { + psc = &priv->screenConfigs[i]; + if (psc->drawHash == NULL) + continue; + + if (__glxHashLookup(psc->drawHash, drawable, (void *) &pdraw) == 0) { + if (scrn_num != NULL) + *scrn_num = i; + return pdraw; } } return NULL; } + #endif @@ -264,9 +311,9 @@ GLXContext AllocateGLXContext( Display *dpy ) */ gc->fastImageUnpack = GL_FALSE; gc->fillImage = __glFillImage; - gc->isDirect = GL_FALSE; gc->pc = gc->buf; gc->bufEnd = gc->buf + bufSize; + gc->isDirect = GL_FALSE; if (__glXDebug) { /* ** Set limit register so that there will be one command per packet @@ -312,6 +359,10 @@ CreateContext(Display *dpy, XVisualInfo *vis, Bool use_glx_1_3, int renderType) { GLXContext gc; +#ifdef GLX_DIRECT_RENDERING + int screen = (fbconfig == NULL) ? vis->screen : fbconfig->screen; + __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); +#endif if ( dpy == NULL ) return NULL; @@ -325,41 +376,36 @@ CreateContext(Display *dpy, XVisualInfo *vis, return NULL; #ifdef GLX_DIRECT_RENDERING - if (allowDirect) { - int screen = (fbconfig == NULL) ? vis->screen : fbconfig->screen; - __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); + if (allowDirect && psc->driScreen) { const __GLcontextModes * mode; - /* The value of fbconfig cannot change because it is tested - * later in the function. - */ - if ( fbconfig == NULL ) { - /* FIXME: Is it possible for the __GLcontextModes structure - * FIXME: to not be found? - */ - mode = _gl_context_modes_find_visual( psc->configs, - vis->visualid ); - assert( mode != NULL ); - assert( mode->screen == screen ); + if (fbconfig == NULL) { + mode = _gl_context_modes_find_visual(psc->visuals, vis->visualid); + if (mode == NULL) { + xError error; + + error.errorCode = BadValue; + error.resourceID = vis->visualid; + error.sequenceNumber = dpy->request; + error.type = X_Error; + error.majorCode = gc->majorOpcode; + error.minorCode = X_GLXCreateContext; + _XError(dpy, &error); + return None; + } } else { mode = fbconfig; } - if (psc && psc->driScreen.private) { - void * const shared = (shareList != NULL) - ? shareList->driContext.private : NULL; - gc->driContext.private = - (*psc->driScreen.createNewContext)( dpy, mode, renderType, - shared, - &gc->driContext ); - if (gc->driContext.private) { - gc->isDirect = GL_TRUE; - gc->screen = mode->screen; - gc->vid = mode->visualID; - gc->fbconfigID = mode->fbconfigID; - gc->driContext.mode = mode; - } + gc->driContext = psc->driScreen->createContext(psc, mode, gc, + shareList, + renderType); + if (gc->driContext != NULL) { + gc->screen = mode->screen; + gc->psc = psc; + gc->mode = mode; + gc->isDirect = GL_TRUE; } } #endif @@ -376,7 +422,7 @@ CreateContext(Display *dpy, XVisualInfo *vis, req->visual = vis->visualid; req->screen = vis->screen; req->shareList = shareList ? shareList->xid : None; - req->isDirect = gc->isDirect; + req->isDirect = gc->driContext != NULL; } else if ( use_glx_1_3 ) { xGLXCreateNewContextReq *req; @@ -390,7 +436,7 @@ CreateContext(Display *dpy, XVisualInfo *vis, req->screen = fbconfig->screen; req->renderType = renderType; req->shareList = shareList ? shareList->xid : None; - req->isDirect = gc->isDirect; + req->isDirect = gc->driContext != NULL; } else { xGLXVendorPrivateWithReplyReq *vpreq; @@ -408,7 +454,7 @@ CreateContext(Display *dpy, XVisualInfo *vis, req->screen = fbconfig->screen; req->renderType = renderType; req->shareList = shareList ? shareList->xid : None; - req->isDirect = gc->isDirect; + req->isDirect = gc->driContext != NULL; } UnlockDisplay(dpy); @@ -430,7 +476,7 @@ PUBLIC GLXContext glXCreateContext(Display *dpy, XVisualInfo *vis, False, 0); } -void __glXFreeContext(__GLXcontext *gc) +_X_HIDDEN void __glXFreeContext(__GLXcontext *gc) { if (gc->vendor) XFree((char *) gc->vendor); if (gc->renderer) XFree((char *) gc->renderer); @@ -466,12 +512,10 @@ DestroyContext(Display *dpy, GLXContext gc) #ifdef GLX_DIRECT_RENDERING /* Destroy the direct rendering context */ - if (gc->isDirect) { - if (gc->driContext.private) { - (*gc->driContext.destroyContext)(dpy, gc->screen, - gc->driContext.private); - gc->driContext.private = NULL; - } + if (gc->driContext) { + (*gc->driContext->destroyContext)(gc->driContext, gc->psc, dpy); + gc->driContext = NULL; + GarbageCollectDRIDrawables(dpy, gc->psc); } #endif @@ -552,7 +596,7 @@ PUBLIC void glXWaitGL(void) __glXFlushRenderBuffer(gc, gc->pc); #ifdef GLX_DIRECT_RENDERING - if (gc->isDirect) { + if (gc->driContext) { /* This bit of ugliness unwraps the glFinish function */ #ifdef glFinish #undef glFinish @@ -588,7 +632,7 @@ PUBLIC void glXWaitX(void) __glXFlushRenderBuffer(gc, gc->pc); #ifdef GLX_DIRECT_RENDERING - if (gc->isDirect) { + if (gc->driContext) { XSync(dpy, False); return; } @@ -618,7 +662,7 @@ PUBLIC void glXUseXFont(Font font, int first, int count, int listBase) (void) __glXFlushRenderBuffer(gc, gc->pc); #ifdef GLX_DIRECT_RENDERING - if (gc->isDirect) { + if (gc->driContext) { DRI_glXUseXFont(font, first, count, listBase); return; } @@ -658,7 +702,7 @@ PUBLIC void glXCopyContext(Display *dpy, GLXContext source, } #ifdef GLX_DIRECT_RENDERING - if (gc->isDirect) { + if (gc->driContext) { /* NOT_DONE: This does not work yet */ } #endif @@ -730,7 +774,7 @@ PUBLIC Bool glXIsDirect(Display *dpy, GLXContext gc) if (!gc) { return GL_FALSE; #ifdef GLX_DIRECT_RENDERING - } else if (gc->isDirect) { + } else if (gc->driContext) { return GL_TRUE; #endif } @@ -793,10 +837,10 @@ PUBLIC void glXSwapBuffers(Display *dpy, GLXDrawable drawable) GLXContextTag tag; CARD8 opcode; #ifdef GLX_DIRECT_RENDERING - __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, NULL ); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL); - if ( pdraw != NULL ) { - (*pdraw->swapBuffers)(dpy, pdraw->private); + if (pdraw != NULL) { + (*pdraw->psc->core->swapBuffers)(pdraw->driDrawable); return; } #endif @@ -840,12 +884,12 @@ PUBLIC int glXGetConfig(Display *dpy, XVisualInfo *vis, int attribute, { __GLXdisplayPrivate *priv; __GLXscreenConfigs *psc; + __GLcontextModes *modes; int status; status = GetGLXPrivScreenConfig( dpy, vis->screen, & priv, & psc ); if ( status == Success ) { - const __GLcontextModes * const modes = _gl_context_modes_find_visual( - psc->configs, vis->visualid ); + modes = _gl_context_modes_find_visual(psc->visuals, vis->visualid); /* Lookup attribute after first finding a match on the visual */ if ( modes != NULL ) { @@ -1223,7 +1267,7 @@ PUBLIC XVisualInfo *glXChooseVisual(Display *dpy, int screen, int *attribList) ** Compute a score for those that do ** Remember which visual, if any, got the highest score */ - for ( modes = psc->configs ; modes != NULL ; modes = modes->next ) { + for ( modes = psc->visuals ; modes != NULL ; modes = modes->next ) { if ( fbconfigs_compatible( & test_config, modes ) && ((best_config == NULL) || (fbconfig_compare( (const __GLcontextModes * const * const)&modes, &best_config ) < 0)) ) { @@ -1268,7 +1312,7 @@ PUBLIC const char *glXQueryExtensionsString( Display *dpy, int screen ) __glXCalculateUsableExtensions(psc, #ifdef GLX_DIRECT_RENDERING - (psc->driScreen.private != NULL), + (psc->driScreen != NULL), #else GL_FALSE, #endif @@ -1447,13 +1491,15 @@ static int __glXQueryContextInfo(Display *dpy, GLXContext ctx) ctx->share_xid = *pProp++; break; case GLX_VISUAL_ID_EXT: - ctx->vid = *pProp++; + ctx->mode = + _gl_context_modes_find_visual(ctx->psc->visuals, *pProp++); break; case GLX_SCREEN: ctx->screen = *pProp++; break; case GLX_FBCONFIG_ID: - ctx->fbconfigID = *pProp++; + ctx->mode = + _gl_context_modes_find_fbconfig(ctx->psc->configs, *pProp++); break; case GLX_RENDER_TYPE: ctx->renderType = *pProp++; @@ -1478,7 +1524,7 @@ glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value) int retVal; /* get the information from the server if we don't have it already */ - if (!ctx->isDirect && (ctx->vid == None)) { + if (!ctx->driContext && (ctx->mode == NULL)) { retVal = __glXQueryContextInfo(dpy, ctx); if (Success != retVal) return retVal; } @@ -1487,13 +1533,13 @@ glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value) *value = (int)(ctx->share_xid); break; case GLX_VISUAL_ID_EXT: - *value = (int)(ctx->vid); + *value = ctx->mode ? ctx->mode->visualID : None; break; case GLX_SCREEN: *value = (int)(ctx->screen); break; case GLX_FBCONFIG_ID: - *value = (int)(ctx->fbconfigID); + *value = ctx->mode ? ctx->mode->fbconfigID : None; break; case GLX_RENDER_TYPE: *value = (int)(ctx->renderType); @@ -1591,6 +1637,7 @@ PUBLIC GLXFBConfig *glXGetFBConfigs(Display *dpy, int screen, int *nelements) __GLcontextModes ** config = NULL; int i; + *nelements = 0; if ( (priv->screenConfigs != NULL) && (screen >= 0) && (screen <= ScreenCount(dpy)) && (priv->screenConfigs[screen].configs != NULL) @@ -1615,8 +1662,10 @@ PUBLIC GLXFBConfig *glXGetFBConfigs(Display *dpy, int screen, int *nelements) for ( modes = priv->screenConfigs[screen].configs ; modes != NULL ; modes = modes->next ) { - config[i] = modes; - i++; + if ( modes->fbconfigID != GLX_DONT_CARE ) { + config[i] = modes; + i++; + } } } } @@ -1668,16 +1717,15 @@ static int __glXSwapIntervalSGI(int interval) return GLX_BAD_VALUE; } -#ifdef GLX_DIRECT_RENDERING - if ( gc->isDirect ) { +#ifdef __DRI_SWAP_CONTROL + if (gc->driContext) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - __DRIdrawable * const pdraw = GetDRIDrawable( gc->currentDpy, - gc->currentDrawable, - NULL ); - if ( __glXExtensionBitIsEnabled( psc, SGI_swap_control_bit ) - && (pdraw != NULL) ) { - pdraw->swap_interval = interval; + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(gc->currentDpy, + gc->currentDrawable, + NULL); + if (psc->swapControl != NULL && pdraw != NULL) { + psc->swapControl->setSwapInterval(pdraw->driDrawable, interval); return 0; } else { @@ -1715,25 +1763,22 @@ static int __glXSwapIntervalSGI(int interval) */ static int __glXSwapIntervalMESA(unsigned int interval) { -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_SWAP_CONTROL GLXContext gc = __glXGetCurrentContext(); if ( interval < 0 ) { return GLX_BAD_VALUE; } - if ( (gc != NULL) && gc->isDirect ) { + if (gc != NULL && gc->driContext) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - if ( (psc != NULL) && (psc->driScreen.private != NULL) - && __glXExtensionBitIsEnabled( psc, MESA_swap_control_bit ) ) { - __DRIdrawable * const pdraw = - (*psc->driScreen.getDrawable)(gc->currentDpy, - gc->currentDrawable, - psc->driScreen.private); - if ( pdraw != NULL ) { - pdraw->swap_interval = interval; + if ( (psc != NULL) && (psc->driScreen != NULL) ) { + __GLXDRIdrawable *pdraw = + GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL); + if (psc->swapControl != NULL && pdraw != NULL) { + psc->swapControl->setSwapInterval(pdraw->driDrawable, interval); return 0; } } @@ -1748,21 +1793,18 @@ static int __glXSwapIntervalMESA(unsigned int interval) static int __glXGetSwapIntervalMESA(void) { -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_SWAP_CONTROL GLXContext gc = __glXGetCurrentContext(); - if ( (gc != NULL) && gc->isDirect ) { + if (gc != NULL && gc->driContext) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - if ( (psc != NULL) && (psc->driScreen.private != NULL) - && __glXExtensionBitIsEnabled( psc, MESA_swap_control_bit ) ) { - __DRIdrawable * const pdraw = - (*psc->driScreen.getDrawable)(gc->currentDpy, - gc->currentDrawable, - psc->driScreen.private); - if ( pdraw != NULL ) { - return pdraw->swap_interval; + if ( (psc != NULL) && (psc->driScreen != NULL) ) { + __GLXDRIdrawable *pdraw = + GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL); + if (psc->swapControl != NULL && pdraw != NULL) { + return psc->swapControl->getSwapInterval(pdraw->driDrawable); } } } @@ -1779,15 +1821,13 @@ static int __glXGetSwapIntervalMESA(void) static GLint __glXBeginFrameTrackingMESA(Display *dpy, GLXDrawable drawable) { int status = GLX_BAD_CONTEXT; -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_FRAME_TRACKING int screen; - __DRIdrawable * const pdraw = GetDRIDrawable(dpy, drawable, & screen); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); - if ( (pdraw != NULL) && (pdraw->frameTracking != NULL) - && __glXExtensionBitIsEnabled( psc, MESA_swap_frame_usage_bit ) ) { - status = pdraw->frameTracking( dpy, pdraw->private, GL_TRUE ); - } + if (pdraw != NULL && psc->frameTracking != NULL) + status = psc->frameTracking->frameTracking(pdraw->driDrawable, GL_TRUE); #else (void) dpy; (void) drawable; @@ -1799,15 +1839,14 @@ static GLint __glXBeginFrameTrackingMESA(Display *dpy, GLXDrawable drawable) static GLint __glXEndFrameTrackingMESA(Display *dpy, GLXDrawable drawable) { int status = GLX_BAD_CONTEXT; -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_FRAME_TRACKING int screen; - __DRIdrawable * const pdraw = GetDRIDrawable(dpy, drawable, & screen); - __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, & screen); + __GLXscreenConfigs *psc = GetGLXScreenConfigs(dpy, screen); - if ( (pdraw != NULL) && (pdraw->frameTracking != NULL) - && __glXExtensionBitIsEnabled( psc, MESA_swap_frame_usage_bit ) ) { - status = pdraw->frameTracking( dpy, pdraw->private, GL_FALSE ); - } + if (pdraw != NULL && psc->frameTracking != NULL) + status = psc->frameTracking->frameTracking(pdraw->driDrawable, + GL_FALSE); #else (void) dpy; (void) drawable; @@ -1820,19 +1859,20 @@ static GLint __glXGetFrameUsageMESA(Display *dpy, GLXDrawable drawable, GLfloat *usage) { int status = GLX_BAD_CONTEXT; -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_FRAME_TRACKING int screen; - __DRIdrawable * const pdraw = GetDRIDrawable(dpy, drawable, & screen); + __GLXDRIdrawable * const pdraw = GetGLXDRIDrawable(dpy, drawable, & screen); __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); - if ( (pdraw != NULL ) && (pdraw->queryFrameTracking != NULL) - && __glXExtensionBitIsEnabled( psc, MESA_swap_frame_usage_bit ) ) { - int64_t sbc, missedFrames; - float lastMissedUsage; + if (pdraw != NULL && psc->frameTracking != NULL) { + int64_t sbc, missedFrames; + float lastMissedUsage; - status = pdraw->queryFrameTracking( dpy, pdraw->private, &sbc, - &missedFrames, &lastMissedUsage, - usage ); + status = psc->frameTracking->queryFrameTracking(pdraw->driDrawable, + &sbc, + &missedFrames, + &lastMissedUsage, + usage); } #else (void) dpy; @@ -1848,18 +1888,17 @@ static GLint __glXQueryFrameTrackingMESA(Display *dpy, GLXDrawable drawable, GLfloat *lastMissedUsage) { int status = GLX_BAD_CONTEXT; -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_FRAME_TRACKING int screen; - __DRIdrawable * const pdraw = GetDRIDrawable(dpy, drawable, & screen); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, & screen); __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); - if ( (pdraw != NULL ) && (pdraw->queryFrameTracking != NULL) - && __glXExtensionBitIsEnabled( psc, MESA_swap_frame_usage_bit ) ) { + if (pdraw != NULL && psc->frameTracking != NULL) { float usage; - status = pdraw->queryFrameTracking( dpy, pdraw->private, sbc, - missedFrames, lastMissedUsage, - & usage ); + status = psc->frameTracking->queryFrameTracking(pdraw->driDrawable, + sbc, missedFrames, + lastMissedUsage, &usage); } #else (void) dpy; @@ -1881,21 +1920,24 @@ static int __glXGetVideoSyncSGI(unsigned int *count) * FIXME: there should be a GLX encoding for this call. I can find no * FIXME: documentation for the GLX encoding. */ -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_MEDIA_STREAM_COUNTER GLXContext gc = __glXGetCurrentContext(); - if ( (gc != NULL) && gc->isDirect ) { + if (gc != NULL && gc->driContext) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - if ( __glXExtensionBitIsEnabled( psc, SGI_video_sync_bit ) - && psc->driScreen.private && psc->driScreen.getMSC) { - int ret; - int64_t temp; + if ( psc->msc && psc->driScreen ) { + __GLXDRIdrawable *pdraw = + GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL); + int64_t temp; + int ret; + + ret = (*psc->msc->getDrawableMSC)(psc->__driScreen, + pdraw->driDrawable, &temp); + *count = (unsigned) temp; - ret = psc->driScreen.getMSC( psc->driScreen.private, & temp ); - *count = (unsigned) temp; - return (ret == 0) ? 0 : GLX_BAD_CONTEXT; + return (ret == 0) ? 0 : GLX_BAD_CONTEXT; } } #else @@ -1906,32 +1948,26 @@ static int __glXGetVideoSyncSGI(unsigned int *count) static int __glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) { -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_MEDIA_STREAM_COUNTER GLXContext gc = __glXGetCurrentContext(); if ( divisor <= 0 || remainder < 0 ) return GLX_BAD_VALUE; - if ( (gc != NULL) && gc->isDirect ) { + if (gc != NULL && gc->driContext) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - if ( __glXExtensionBitIsEnabled( psc, SGI_video_sync_bit ) - && psc->driScreen.private ) { - __DRIdrawable * const pdraw = - (*psc->driScreen.getDrawable)(gc->currentDpy, - gc->currentDrawable, - psc->driScreen.private); - if ( (pdraw != NULL) && (pdraw->waitForMSC != NULL) ) { - int ret; - int64_t msc; - int64_t sbc; - - ret = (*pdraw->waitForMSC)( gc->currentDpy, pdraw->private, - 0, divisor, remainder, - & msc, & sbc ); - *count = (unsigned) msc; - return (ret == 0) ? 0 : GLX_BAD_CONTEXT; - } + if (psc->msc != NULL && psc->driScreen ) { + __GLXDRIdrawable *pdraw = + GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL); + int ret; + int64_t msc; + int64_t sbc; + + ret = (*psc->msc->waitForMSC)(pdraw->driDrawable, 0, + divisor, remainder, &msc, &sbc); + *count = (unsigned) msc; + return (ret == 0) ? 0 : GLX_BAD_CONTEXT; } } #else @@ -2083,20 +2119,19 @@ static Bool __glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) static Bool __glXGetSyncValuesOML(Display *dpy, GLXDrawable drawable, int64_t *ust, int64_t *msc, int64_t *sbc) { -#ifdef GLX_DIRECT_RENDERING +#if defined(__DRI_SWAP_BUFFER_COUNTER) && defined(__DRI_MEDIA_STREAM_COUNTER) __GLXdisplayPrivate * const priv = __glXInitialize(dpy); if ( priv != NULL ) { int i; - __DRIdrawable * const pdraw = GetDRIDrawable( dpy, drawable, & i ); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &i); __GLXscreenConfigs * const psc = &priv->screenConfigs[i]; assert( (pdraw == NULL) || (i != -1) ); - return ( (pdraw && pdraw->getSBC && psc->driScreen.getMSC) - && __glXExtensionBitIsEnabled( psc, OML_sync_control_bit ) - && ((*psc->driScreen.getMSC)( psc->driScreen.private, msc ) == 0) - && ((*pdraw->getSBC)( dpy, psc->driScreen.private, sbc ) == 0) - && (__glXGetUST( ust ) == 0) ); + return ( (pdraw && psc->sbc && psc->msc) + && ((*psc->msc->getMSC)(psc->driScreen, msc) == 0) + && ((*psc->sbc->getSBC)(pdraw->driDrawable, sbc) == 0) + && (__glXGetUST(ust) == 0) ); } #else (void) dpy; @@ -2108,6 +2143,68 @@ static Bool __glXGetSyncValuesOML(Display *dpy, GLXDrawable drawable, return False; } +#ifdef GLX_DIRECT_RENDERING +_X_HIDDEN GLboolean +__driGetMscRateOML(__DRIdrawable *draw, + int32_t *numerator, int32_t *denominator, void *private) +{ +#ifdef XF86VIDMODE + __GLXscreenConfigs *psc; + XF86VidModeModeLine mode_line; + int dot_clock; + int i; + __GLXDRIdrawable *glxDraw = private; + + psc = glxDraw->psc; + if (XF86VidModeQueryVersion(psc->dpy, &i, &i) && + XF86VidModeGetModeLine(psc->dpy, psc->scr, &dot_clock, &mode_line) ) { + unsigned n = dot_clock * 1000; + unsigned d = mode_line.vtotal * mode_line.htotal; + +# define V_INTERLACE 0x010 +# define V_DBLSCAN 0x020 + + if (mode_line.flags & V_INTERLACE) + n *= 2; + else if (mode_line.flags & V_DBLSCAN) + d *= 2; + + /* The OML_sync_control spec requires that if the refresh rate is a + * whole number, that the returned numerator be equal to the refresh + * rate and the denominator be 1. + */ + + if (n % d == 0) { + n /= d; + d = 1; + } + else { + static const unsigned f[] = { 13, 11, 7, 5, 3, 2, 0 }; + + /* This is a poor man's way to reduce a fraction. It's far from + * perfect, but it will work well enough for this situation. + */ + + for (i = 0; f[i] != 0; i++) { + while (n % f[i] == 0 && d % f[i] == 0) { + d /= f[i]; + n /= f[i]; + } + } + } + + *numerator = n; + *denominator = d; + + return True; + } + else + return False; +#else + return False; +#endif +} +#endif /** * Determine the refresh rate of the specified drawable and display. @@ -2125,70 +2222,17 @@ static Bool __glXGetSyncValuesOML(Display *dpy, GLXDrawable drawable, * when GLX_OML_sync_control appears in the client extension string. */ -Bool __glXGetMscRateOML(Display * dpy, GLXDrawable drawable, - int32_t * numerator, int32_t * denominator) +_X_HIDDEN GLboolean __glXGetMscRateOML(Display * dpy, GLXDrawable drawable, + int32_t * numerator, + int32_t * denominator) { #if defined( GLX_DIRECT_RENDERING ) && defined( XF86VIDMODE ) - __GLXdisplayPrivate * const priv = __glXInitialize(dpy); - - - if ( priv != NULL ) { - XF86VidModeModeLine mode_line; - int dot_clock; - int screen_num; - int i; - - - GetDRIDrawable( dpy, drawable, & screen_num ); - if ( (screen_num != -1) - && XF86VidModeQueryVersion( dpy, & i, & i ) - && XF86VidModeGetModeLine( dpy, screen_num, & dot_clock, - & mode_line ) ) { - unsigned n = dot_clock * 1000; - unsigned d = mode_line.vtotal * mode_line.htotal; - -# define V_INTERLACE 0x010 -# define V_DBLSCAN 0x020 - - if ( (mode_line.flags & V_INTERLACE) ) { - n *= 2; - } - else if ( (mode_line.flags & V_DBLSCAN) ) { - d *= 2; - } - - /* The OML_sync_control spec requires that if the refresh rate is a - * whole number, that the returned numerator be equal to the refresh - * rate and the denominator be 1. - */ + __GLXDRIdrawable *draw = GetGLXDRIDrawable(dpy, drawable, NULL); - if ( (n % d) == 0 ) { - n /= d; - d = 1; - } - else { - static const unsigned f[] = { 13, 11, 7, 5, 3, 2, 0 }; - - - /* This is a poor man's way to reduce a fraction. It's far from - * perfect, but it will work well enough for this situation. - */ - - for ( i = 0 ; f[i] != 0 ; i++ ) { - while ( ((n % f[i]) == 0) && ((d % f[i]) == 0) ) { - d /= f[i]; - n /= f[i]; - } - } - } - - *numerator = n; - *denominator = d; + if (draw == NULL) + return False; - (void) drawable; - return True; - } - } + return __driGetMscRateOML(draw->driDrawable, numerator, denominator, draw); #else (void) dpy; (void) drawable; @@ -2203,9 +2247,9 @@ static int64_t __glXSwapBuffersMscOML(Display *dpy, GLXDrawable drawable, int64_t target_msc, int64_t divisor, int64_t remainder) { -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_SWAP_BUFFER_COUNTER int screen; - __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, & screen ); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen ); /* The OML_sync_control spec says these should "generate a GLX_BAD_VALUE @@ -2218,11 +2262,10 @@ static int64_t __glXSwapBuffersMscOML(Display *dpy, GLXDrawable drawable, if ( divisor > 0 && remainder >= divisor ) return -1; - if ( (pdraw != NULL) && (pdraw->swapBuffersMSC != NULL) - && __glXExtensionBitIsEnabled( psc, OML_sync_control_bit ) ) { - return (*pdraw->swapBuffersMSC)(dpy, pdraw->private, target_msc, - divisor, remainder); - } + if (pdraw != NULL && psc->counters != NULL) + return (*psc->sbc->swapBuffersMSC)(pdraw->driDrawable, target_msc, + divisor, remainder); + #else (void) dpy; (void) drawable; @@ -2239,9 +2282,9 @@ static Bool __glXWaitForMscOML(Display * dpy, GLXDrawable drawable, int64_t remainder, int64_t *ust, int64_t *msc, int64_t *sbc) { -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_MEDIA_STREAM_COUNTER int screen; - __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, & screen ); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen ); int ret; @@ -2253,10 +2296,9 @@ static Bool __glXWaitForMscOML(Display * dpy, GLXDrawable drawable, if ( divisor > 0 && remainder >= divisor ) return False; - if ( (pdraw != NULL) && (pdraw->waitForMSC != NULL) - && __glXExtensionBitIsEnabled( psc, OML_sync_control_bit ) ) { - ret = (*pdraw->waitForMSC)( dpy, pdraw->private, target_msc, - divisor, remainder, msc, sbc ); + if (pdraw != NULL && psc->msc != NULL) { + ret = (*psc->msc->waitForMSC)(pdraw->driDrawable, target_msc, + divisor, remainder, msc, sbc); /* __glXGetUST returns zero on success and non-zero on failure. * This function returns True on success and False on failure. @@ -2281,9 +2323,9 @@ static Bool __glXWaitForSbcOML(Display * dpy, GLXDrawable drawable, int64_t target_sbc, int64_t *ust, int64_t *msc, int64_t *sbc ) { -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_SWAP_BUFFER_COUNTER int screen; - __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, & screen ); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen ); int ret; @@ -2293,9 +2335,8 @@ static Bool __glXWaitForSbcOML(Display * dpy, GLXDrawable drawable, if ( target_sbc < 0 ) return False; - if ( (pdraw != NULL) && (pdraw->waitForSBC != NULL) - && __glXExtensionBitIsEnabled( psc, OML_sync_control_bit )) { - ret = (*pdraw->waitForSBC)( dpy, pdraw->private, target_sbc, msc, sbc ); + if (pdraw != NULL && psc->sbc != NULL) { + ret = (*psc->sbc->waitForSBC)(pdraw->driDrawable, target_sbc, msc, sbc); /* __glXGetUST returns zero on success and non-zero on failure. * This function returns True on success and False on failure. @@ -2323,16 +2364,13 @@ PUBLIC void *glXAllocateMemoryMESA(Display *dpy, int scrn, size_t size, float readFreq, float writeFreq, float priority) { -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_ALLOCATE __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, scrn ); - if ( __glXExtensionBitIsEnabled( psc, MESA_allocate_memory_bit ) ) { - if (psc && psc->driScreen.private && psc->driScreen.allocateMemory) { - return (*psc->driScreen.allocateMemory)( dpy, scrn, size, - readFreq, writeFreq, - priority ); - } - } + if (psc && psc->allocate) + return (*psc->allocate->allocateMemory)(psc->__driScreen, size, + readFreq, writeFreq, priority); + #else (void) dpy; (void) scrn; @@ -2348,14 +2386,12 @@ PUBLIC void *glXAllocateMemoryMESA(Display *dpy, int scrn, PUBLIC void glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer) { -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_ALLOCATE __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, scrn ); - if ( __glXExtensionBitIsEnabled( psc, MESA_allocate_memory_bit ) ) { - if (psc && psc->driScreen.private && psc->driScreen.freeMemory) { - (*psc->driScreen.freeMemory)( dpy, scrn, pointer ); - } - } + if (psc && psc->allocate) + (*psc->allocate->freeMemory)(psc->__driScreen, pointer); + #else (void) dpy; (void) scrn; @@ -2367,14 +2403,12 @@ PUBLIC void glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer) PUBLIC GLuint glXGetMemoryOffsetMESA( Display *dpy, int scrn, const void *pointer ) { -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_ALLOCATE __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, scrn ); - if ( __glXExtensionBitIsEnabled( psc, MESA_allocate_memory_bit ) ) { - if (psc && psc->driScreen.private && psc->driScreen.memoryOffset) { - return (*psc->driScreen.memoryOffset)( dpy, scrn, pointer ); - } - } + if (psc && psc->allocate) + return (*psc->allocate->memoryOffset)(psc->__driScreen, pointer); + #else (void) dpy; (void) scrn; @@ -2447,13 +2481,14 @@ static void __glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable, INT32 *x_ptr, *y_ptr, *w_ptr, *h_ptr; CARD8 opcode; -#ifdef GLX_DIRECT_RENDERING +#ifdef __DRI_COPY_SUB_BUFFER int screen; - __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, & screen ); + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); if ( pdraw != NULL ) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen ); - if ( __glXExtensionBitIsEnabled( psc, MESA_copy_sub_buffer_bit ) ) { - (*pdraw->copySubBuffer)(dpy, pdraw->private, x, y, width, height); + if (psc->copySubBuffer != NULL) { + (*psc->copySubBuffer->copySubBuffer)(pdraw->driDrawable, + x, y, width, height); } return; @@ -2529,8 +2564,16 @@ static void __glXBindTexImageEXT(Display *dpy, } #ifdef GLX_DIRECT_RENDERING - if (gc->isDirect) + if (gc->driContext) { + __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL); + + if (pdraw != NULL) + (*pdraw->psc->texBuffer->setTexBuffer)(gc->__driContext, + pdraw->textureTarget, + pdraw->driDrawable); + return; + } #endif opcode = __glXSetupForCommand(dpy); @@ -2581,7 +2624,7 @@ static void __glXReleaseTexImageEXT(Display *dpy, return; #ifdef GLX_DIRECT_RENDERING - if (gc->isDirect) + if (gc->driContext) return; #endif @@ -2613,7 +2656,7 @@ static void __glXReleaseTexImageEXT(Display *dpy, * * \sa strdup */ -char * +_X_HIDDEN char * __glXstrdup(const char *str) { char *copy; @@ -2840,98 +2883,6 @@ PUBLIC void (*glXGetProcAddress(const GLubyte *procName))( void ) #ifdef GLX_DIRECT_RENDERING -/** - * Retrieves the verion of the internal libGL API in YYYYMMDD format. This - * might be used by the DRI drivers to determine how new libGL is at runtime. - * Drivers should not call this function directly. They should instead use - * \c glXGetProcAddress to obtain a pointer to the function. - * - * \returns An 8-digit decimal number representing the internal libGL API in - * YYYYMMDD format. - * - * \sa glXGetProcAddress, PFNGLXGETINTERNALVERSIONPROC - * - * \since Internal API version 20021121. - */ -int __glXGetInternalVersion(void) -{ - /* History: - * 20021121 - Initial version - * 20021128 - Added __glXWindowExists() function - * 20021207 - Added support for dynamic GLX extensions, - * GLX_SGI_swap_control, GLX_SGI_video_sync, - * GLX_OML_sync_control, and GLX_MESA_swap_control. - * Never officially released. Do NOT test against - * this version. Use 20030317 instead. - * 20030317 - Added support GLX_SGIX_fbconfig, - * GLX_MESA_swap_frame_usage, GLX_OML_swap_method, - * GLX_{ARB,SGIS}_multisample, and - * GLX_SGIX_visual_select_group. - * 20030606 - Added support for GLX_SGI_make_current_read. - * 20030813 - Made support for dynamic extensions multi-head aware. - * 20030818 - Added support for GLX_MESA_allocate_memory in place of the - * deprecated GLX_NV_vertex_array_range & GLX_MESA_agp_offset - * interfaces. - * 20031201 - Added support for the first round of DRI interface changes. - * Do NOT test against this version! It has binary - * compatibility bugs, use 20040317 instead. - * 20040317 - Added the 'mode' field to __DRIcontextRec. - * 20040415 - Added support for bindContext3 and unbindContext3. - * 20040602 - Add __glXGetDrawableInfo. I though that was there - * months ago. :( - * 20050727 - Gut all the old interfaces. This breaks compatability with - * any DRI driver built to any previous version. - * 20060314 - Added support for GLX_MESA_copy_sub_buffer. - * 20070105 - Added support for damage reporting. - */ - return 20070105; -} - - - -static Bool windowExistsFlag; - -static int windowExistsErrorHandler(Display *dpy, XErrorEvent *xerr) -{ - if (xerr->error_code == BadWindow) { - windowExistsFlag = GL_FALSE; - } - return 0; -} - -/** - * Determine if a window associated with a \c GLXDrawable exists on the - * X-server. This function is not used internally by libGL. It is provided - * as a utility function for DRI drivers. - * Drivers should not call this function directly. They should instead use - * \c glXGetProcAddress to obtain a pointer to the function. - * - * \param dpy Display associated with the drawable to be queried. - * \param draw \c GLXDrawable to test. - * - * \returns \c GL_TRUE if a window exists that is associated with \c draw, - * otherwise \c GL_FALSE is returned. - * - * \warning This function is not currently thread-safe. - * - * \sa glXGetProcAddress - * - * \since Internal API version 20021128. - */ -Bool __glXWindowExists(Display *dpy, GLXDrawable draw) -{ - XWindowAttributes xwa; - int (*oldXErrorHandler)(Display *, XErrorEvent *); - - XSync(dpy, GL_FALSE); - windowExistsFlag = GL_TRUE; - oldXErrorHandler = XSetErrorHandler(windowExistsErrorHandler); - XGetWindowAttributes(dpy, draw, &xwa); /* dummy request */ - XSetErrorHandler(oldXErrorHandler); - return windowExistsFlag; -} - - /** * Get the unadjusted system time (UST). Currently, the UST is measured in * microseconds since Epoc. The actual resolution of the UST may vary from @@ -2946,7 +2897,7 @@ Bool __glXWindowExists(Display *dpy, GLXDrawable draw) * * \since Internal API version 20030317. */ -int __glXGetUST( int64_t * ust ) +_X_HIDDEN int __glXGetUST( int64_t * ust ) { struct timeval tv; diff --git a/src/glx/x11/glxext.c b/src/glx/x11/glxext.c index 6403cbd56d..4d814744cd 100644 --- a/src/glx/x11/glxext.c +++ b/src/glx/x11/glxext.c @@ -44,57 +44,25 @@ */ #include "glxclient.h" -#include #include #include -#include -#include -#include -#include "indirect_init.h" #include "glapi.h" #include "glxextensions.h" #include "glcontextmodes.h" #include "glheader.h" -#ifdef GLX_DIRECT_RENDERING -#include -#include -#include "xf86dri.h" -#include "sarea.h" -#include "dri_glx.h" -#endif - #ifdef USE_XCB #include #include #include #endif -#include #ifdef DEBUG void __glXDumpDrawBuffer(__GLXcontext *ctx); #endif #ifdef USE_SPARC_ASM -/* - * This is where our dispatch table's bounds are. - * And the static mesa_init is taken directly from - * Mesa's 'sparc.c' initializer. - * - * We need something like this here, because this version - * of openGL/glx never initializes a Mesa context, and so - * the address of the dispatch table pointer never gets stuffed - * into the dispatch jump table otherwise. - * - * It matters only on SPARC, and only if you are using assembler - * code instead of C-code indirect dispatch. - * - * -- FEM, 04.xii.03 - */ -extern unsigned int _mesa_sparc_glapi_begin; -extern unsigned int _mesa_sparc_glapi_end; -extern void __glapi_sparc_icache_flush(unsigned int *); static void _glx_mesa_init_sparc_glapi_relocs(void); static int _mesa_sparc_needs_init = 1; #define INIT_MESA_SPARC { \ @@ -107,173 +75,11 @@ static int _mesa_sparc_needs_init = 1; #define INIT_MESA_SPARC #endif -#ifdef GLX_DIRECT_RENDERING -static __DRIscreen *__glXFindDRIScreen(__DRInativeDisplay *dpy, int scrn); -#endif /* GLX_DIRECT_RENDERING */ - -static Bool MakeContextCurrent(Display *dpy, GLXDrawable draw, - GLXDrawable read, GLXContext gc); - -/* -** We setup some dummy structures here so that the API can be used -** even if no context is current. -*/ - -static GLubyte dummyBuffer[__GLX_BUFFER_LIMIT_SIZE]; - -/* -** Dummy context used by small commands when there is no current context. -** All the -** gl and glx entry points are designed to operate as nop's when using -** the dummy context structure. -*/ -static __GLXcontext dummyContext = { - &dummyBuffer[0], - &dummyBuffer[0], - &dummyBuffer[0], - &dummyBuffer[__GLX_BUFFER_LIMIT_SIZE], - sizeof(dummyBuffer), -}; - - -/* -** All indirect rendering contexts will share the same indirect dispatch table. -*/ -static __GLapi *IndirectAPI = NULL; - - -/* - * Current context management and locking - */ - -#if defined( USE_XTHREADS ) - -/* thread safe */ -static GLboolean TSDinitialized = GL_FALSE; -static xthread_key_t ContextTSD; - -__GLXcontext *__glXGetCurrentContext(void) -{ - if (!TSDinitialized) { - xthread_key_create(&ContextTSD, NULL); - TSDinitialized = GL_TRUE; - return &dummyContext; - } - else { - void *p; - xthread_get_specific(ContextTSD, &p); - if (!p) - return &dummyContext; - else - return (__GLXcontext *) p; - } -} - -void __glXSetCurrentContext(__GLXcontext *c) -{ - if (!TSDinitialized) { - xthread_key_create(&ContextTSD, NULL); - TSDinitialized = GL_TRUE; - } - xthread_set_specific(ContextTSD, c); -} - - -/* Used by the __glXLock() and __glXUnlock() macros */ -xmutex_rec __glXmutex; - -#elif defined( PTHREADS ) - -pthread_mutex_t __glXmutex = PTHREAD_MUTEX_INITIALIZER; - -# if defined( GLX_USE_TLS ) - -/** - * Per-thread GLX context pointer. - * - * \c __glXSetCurrentContext is written is such a way that this pointer can - * \b never be \c NULL. This is important! Because of this - * \c __glXGetCurrentContext can be implemented as trivial macro. - */ -__thread void * __glX_tls_Context __attribute__((tls_model("initial-exec"))) - = &dummyContext; - -void __glXSetCurrentContext( __GLXcontext * c ) -{ - __glX_tls_Context = (c != NULL) ? c : &dummyContext; -} - -# else - -static pthread_once_t once_control = PTHREAD_ONCE_INIT; - -/** - * Per-thread data key. - * - * Once \c init_thread_data has been called, the per-thread data key will - * take a value of \c NULL. As each new thread is created the default - * value, in that thread, will be \c NULL. - */ -static pthread_key_t ContextTSD; - -/** - * Initialize the per-thread data key. - * - * This function is called \b exactly once per-process (not per-thread!) to - * initialize the per-thread data key. This is ideally done using the - * \c pthread_once mechanism. - */ -static void init_thread_data( void ) -{ - if ( pthread_key_create( & ContextTSD, NULL ) != 0 ) { - perror( "pthread_key_create" ); - exit( -1 ); - } -} - -void __glXSetCurrentContext( __GLXcontext * c ) -{ - pthread_once( & once_control, init_thread_data ); - pthread_setspecific( ContextTSD, c ); -} - -__GLXcontext * __glXGetCurrentContext( void ) -{ - void * v; - - pthread_once( & once_control, init_thread_data ); - - v = pthread_getspecific( ContextTSD ); - return (v == NULL) ? & dummyContext : (__GLXcontext *) v; -} - -# endif /* defined( GLX_USE_TLS ) */ - -#elif defined( THREADS ) - -#error Unknown threading method specified. - -#else - -/* not thread safe */ -__GLXcontext *__glXcurrentContext = &dummyContext; - -#endif - - /* ** You can set this cell to 1 to force the gl drawing stuff to be ** one command per packet */ -int __glXDebug = 0; - -/* -** forward prototype declarations -*/ -int __glXCloseDisplay(Display *dpy, XExtCodes *codes); - - -/************************************************************************/ +_X_HIDDEN int __glXDebug = 0; /* Extension required boiler plate */ @@ -296,16 +102,13 @@ static /* const */ char *error_list[] = { "GLXBadWindow", }; -int __glXCloseDisplay(Display *dpy, XExtCodes *codes) +static int __glXCloseDisplay(Display *dpy, XExtCodes *codes) { GLXContext gc; gc = __glXGetCurrentContext(); if (dpy == gc->currentDpy) { - __glXSetCurrentContext(&dummyContext); -#ifdef GLX_DIRECT_RENDERING - _glapi_set_dispatch(NULL); /* no-op functions */ -#endif + __glXSetCurrentContextNull(); __glXFreeContext(gc); } @@ -360,11 +163,10 @@ static void FreeScreenConfigs(__GLXdisplayPrivate *priv) Xfree((char*) psc->serverGLXexts); #ifdef GLX_DIRECT_RENDERING - /* Free the direct rendering per screen data */ - if (psc->driScreen.private) - (*psc->driScreen.destroyScreen)(priv->dpy, i, - psc->driScreen.private); - psc->driScreen.private = NULL; + if (psc->driScreen) { + psc->driScreen->destroyScreen(psc); + __glxHashDestroy(psc->drawHash); + } #endif } XFree((char*) priv->screenConfigs); @@ -391,14 +193,12 @@ static int __glXFreeDisplayPrivate(XExtData *extension) #ifdef GLX_DIRECT_RENDERING /* Free the direct rendering per display data */ - if (priv->driDisplay.private) - (*priv->driDisplay.destroyDisplay)(priv->dpy, - priv->driDisplay.private); - priv->driDisplay.private = NULL; - if (priv->driDisplay.createNewScreen) { - Xfree(priv->driDisplay.createNewScreen); /* free array of ptrs */ - priv->driDisplay.createNewScreen = NULL; - } + if (priv->driDisplay) + (*priv->driDisplay->destroyDisplay)(priv->driDisplay); + priv->driDisplay = NULL; + if (priv->dri2Display) + (*priv->dri2Display->destroyDisplay)(priv->dri2Display); + priv->dri2Display = NULL; #endif Xfree((char*) priv); @@ -440,7 +240,7 @@ static Bool QueryVersion(Display *dpy, int opcode, int *major, int *minor) } -void +_X_HIDDEN void __glXInitializeVisualConfigFromTags( __GLcontextModes *config, int count, const INT32 *bp, Bool tagged_only, Bool fbconfig_style_tags ) @@ -634,371 +434,128 @@ __glXInitializeVisualConfigFromTags( __GLcontextModes *config, int count, config->haveStencilBuffer = (config->stencilBits > 0); } - -#ifdef GLX_DIRECT_RENDERING -static unsigned -filter_modes( __GLcontextModes ** server_modes, - const __GLcontextModes * driver_modes ) +static __GLcontextModes * +createConfigsFromProperties(Display *dpy, int nvisuals, int nprops, + int screen, GLboolean tagged_only) { - __GLcontextModes * m; - __GLcontextModes ** prev_next; - const __GLcontextModes * check; - unsigned modes_count = 0; - - if ( driver_modes == NULL ) { - fprintf(stderr, "libGL warning: 3D driver returned no fbconfigs.\n"); - return 0; - } - - /* For each mode in server_modes, check to see if a matching mode exists - * in driver_modes. If not, then the mode is not available. - */ + INT32 buf[__GLX_TOTAL_CONFIG], *props; + unsigned prop_size; + __GLcontextModes *modes, *m; + int i; - prev_next = server_modes; - for ( m = *prev_next ; m != NULL ; m = *prev_next ) { - GLboolean do_delete = GL_TRUE; + if (nprops == 0) + return NULL; - for ( check = driver_modes ; check != NULL ; check = check->next ) { - if ( _gl_context_modes_are_same( m, check ) ) { - do_delete = GL_FALSE; - break; - } - } + /* FIXME: Is the __GLX_MIN_CONFIG_PROPS test correct for FBconfigs? */ - /* The 3D has to support all the modes that match the GLX visuals - * sent from the X server. - */ - if ( do_delete && (m->visualID != 0) ) { - do_delete = GL_FALSE; + /* Check number of properties */ + if (nprops < __GLX_MIN_CONFIG_PROPS || nprops > __GLX_MAX_CONFIG_PROPS) + return NULL; - if (getenv("LIBGL_DEBUG")) { - fprintf(stderr, "libGL warning: 3D driver claims to not support " - "visual 0x%02x\n", m->visualID); - } - } + /* Allocate memory for our config structure */ + modes = _gl_context_modes_create(nvisuals, sizeof(__GLcontextModes)); + if (!modes) + return NULL; - if ( do_delete ) { - *prev_next = m->next; + prop_size = nprops * __GLX_SIZE_INT32; + if (prop_size <= sizeof(buf)) + props = buf; + else + props = Xmalloc(prop_size); - m->next = NULL; - _gl_context_modes_destroy( m ); - } - else { - modes_count++; - prev_next = & m->next; - } + /* Read each config structure and convert it into our format */ + m = modes; + for (i = 0; i < nvisuals; i++) { + _XRead(dpy, (char *)props, prop_size); + /* Older X servers don't send this so we default it here. */ + m->drawableType = GLX_WINDOW_BIT; + __glXInitializeVisualConfigFromTags(m, nprops, props, + tagged_only, GL_TRUE); + m->screen = screen; + m = m->next; } - return modes_count; -} + if (props != buf) + Xfree(props); - -/** - * Implement \c __DRIinterfaceMethods::getProcAddress. - */ -static __DRIfuncPtr get_proc_address( const char * proc_name ) -{ - if (strcmp( proc_name, "glxEnableExtension" ) == 0) { - return (__DRIfuncPtr) __glXScrEnableExtension; - } - - return NULL; + return modes; } -#ifdef XDAMAGE_1_1_INTERFACE -static GLboolean has_damage_post(__DRInativeDisplay *dpy) +static GLboolean +getVisualConfigs(Display *dpy, __GLXdisplayPrivate *priv, int screen) { - static GLboolean inited = GL_FALSE; - static GLboolean has_damage; - - if (!inited) { - int major, minor; - - if (XDamageQueryVersion(dpy, &major, &minor) && - major == 1 && minor >= 1) - { - has_damage = GL_TRUE; - } else { - has_damage = GL_FALSE; - } - inited = GL_TRUE; - } + xGLXGetVisualConfigsReq *req; + __GLXscreenConfigs *psc; + xGLXGetVisualConfigsReply reply; + + LockDisplay(dpy); - return has_damage; -} -#endif /* XDAMAGE_1_1_INTERFACE */ + psc = priv->screenConfigs + screen; + psc->visuals = NULL; + GetReq(GLXGetVisualConfigs, req); + req->reqType = priv->majorOpcode; + req->glxCode = X_GLXGetVisualConfigs; + req->screen = screen; -static void __glXReportDamage(__DRInativeDisplay *dpy, int screen, - __DRIid drawable, - int x, int y, - drm_clip_rect_t *rects, int num_rects, - GLboolean front_buffer) -{ -#ifdef XDAMAGE_1_1_INTERFACE - XRectangle *xrects; - XserverRegion region; - int i; - int x_off, y_off; - - if (!has_damage_post(dpy)) - return; - - if (front_buffer) { - x_off = x; - y_off = y; - drawable = RootWindow(dpy, screen); - } else{ - x_off = 0; - y_off = 0; - } + if (!_XReply(dpy, (xReply*) &reply, 0, False)) + goto out; - xrects = malloc(sizeof(XRectangle) * num_rects); - if (xrects == NULL) - return; + psc->visuals = createConfigsFromProperties(dpy, + reply.numVisuals, + reply.numProps, + screen, GL_FALSE); - for (i = 0; i < num_rects; i++) { - xrects[i].x = rects[i].x1 + x_off; - xrects[i].y = rects[i].y1 + y_off; - xrects[i].width = rects[i].x2 - rects[i].x1; - xrects[i].height = rects[i].y2 - rects[i].y1; - } - region = XFixesCreateRegion(dpy, xrects, num_rects); - free(xrects); - XDamageAdd(dpy, drawable, region); - XFixesDestroyRegion(dpy, region); -#endif + out: + UnlockDisplay(dpy); + return psc->visuals != NULL; } -/** - * Table of functions exported by the loader to the driver. - */ -static const __DRIinterfaceMethods interface_methods = { - get_proc_address, - - _gl_context_modes_create, - _gl_context_modes_destroy, - - __glXFindDRIScreen, - __glXWindowExists, - - XF86DRICreateContextWithConfig, - XF86DRIDestroyContext, - - XF86DRICreateDrawable, - XF86DRIDestroyDrawable, - XF86DRIGetDrawableInfo, - - __glXGetUST, - __glXGetMscRateOML, - - __glXReportDamage, -}; - - - -/** - * Perform the required libGL-side initialization and call the client-side - * driver's \c __driCreateNewScreen function. - * - * \param dpy Display pointer. - * \param scrn Screen number on the display. - * \param psc DRI screen information. - * \param driDpy DRI display information. - * \param createNewScreen Pointer to the client-side driver's - * \c __driCreateNewScreen function. - * \returns A pointer to the \c __DRIscreenPrivate structure returned by - * the client-side driver on success, or \c NULL on failure. - * - * \todo This function needs to be modified to remove context-modes from the - * list stored in the \c __GLXscreenConfigsRec to match the list - * returned by the client-side driver. - */ -static void * -CallCreateNewScreen(Display *dpy, int scrn, __DRIscreen *psc, - __DRIdisplay * driDpy, - PFNCREATENEWSCREENFUNC createNewScreen) +static GLboolean +getFBConfigs(Display *dpy, __GLXdisplayPrivate *priv, int screen) { - __DRIscreenPrivate *psp = NULL; -#ifndef GLX_USE_APPLEGL - drm_handle_t hSAREA; - drmAddress pSAREA = MAP_FAILED; - char *BusID; - __DRIversion ddx_version; - __DRIversion dri_version; - __DRIversion drm_version; - __DRIframebuffer framebuffer; - int fd = -1; - int status; - const char * err_msg; - const char * err_extra; - int api_ver = __glXGetInternalVersion(); - - - dri_version.major = driDpy->private->driMajor; - dri_version.minor = driDpy->private->driMinor; - dri_version.patch = driDpy->private->driPatch; - - - err_msg = "XF86DRIOpenConnection"; - err_extra = NULL; - - framebuffer.base = MAP_FAILED; - framebuffer.dev_priv = NULL; - - if (XF86DRIOpenConnection(dpy, scrn, &hSAREA, &BusID)) { - int newlyopened; - fd = drmOpenOnce(NULL,BusID, &newlyopened); - Xfree(BusID); /* No longer needed */ - - err_msg = "open DRM"; - err_extra = strerror( -fd ); - - if (fd >= 0) { - drm_magic_t magic; - - err_msg = "drmGetMagic"; - err_extra = NULL; - - if (!drmGetMagic(fd, &magic)) { - drmVersionPtr version = drmGetVersion(fd); - if (version) { - drm_version.major = version->version_major; - drm_version.minor = version->version_minor; - drm_version.patch = version->version_patchlevel; - drmFreeVersion(version); - } - else { - drm_version.major = -1; - drm_version.minor = -1; - drm_version.patch = -1; - } - - err_msg = "XF86DRIAuthConnection"; - if (!newlyopened || XF86DRIAuthConnection(dpy, scrn, magic)) { - char *driverName; - - /* - * Get device name (like "tdfx") and the ddx version - * numbers. We'll check the version in each DRI driver's - * "createNewScreen" function. - */ - err_msg = "XF86DRIGetClientDriverName"; - if (XF86DRIGetClientDriverName(dpy, scrn, - &ddx_version.major, - &ddx_version.minor, - &ddx_version.patch, - &driverName)) { - drm_handle_t hFB; - int junk; - - /* No longer needed. */ - Xfree( driverName ); - - - /* - * Get device-specific info. pDevPriv will point to a struct - * (such as DRIRADEONRec in xfree86/driver/ati/radeon_dri.h) - * that has information about the screen size, depth, pitch, - * ancilliary buffers, DRM mmap handles, etc. - */ - err_msg = "XF86DRIGetDeviceInfo"; - if (XF86DRIGetDeviceInfo(dpy, scrn, - &hFB, - &junk, - &framebuffer.size, - &framebuffer.stride, - &framebuffer.dev_priv_size, - &framebuffer.dev_priv)) { - framebuffer.width = DisplayWidth(dpy, scrn); - framebuffer.height = DisplayHeight(dpy, scrn); - - /* - * Map the framebuffer region. - */ - status = drmMap(fd, hFB, framebuffer.size, - (drmAddressPtr)&framebuffer.base); - - err_msg = "drmMap of framebuffer"; - err_extra = strerror( -status ); - - if ( status == 0 ) { - /* - * Map the SAREA region. Further mmap regions - * may be setup in each DRI driver's - * "createNewScreen" function. - */ - status = drmMap(fd, hSAREA, SAREA_MAX, - &pSAREA); - - err_msg = "drmMap of sarea"; - err_extra = strerror( -status ); - - if ( status == 0 ) { - __GLcontextModes * driver_modes = NULL; - __GLXscreenConfigs *configs = psc->screenConfigs; - - err_msg = "InitDriver"; - err_extra = NULL; - psp = (*createNewScreen)(dpy, scrn, - psc, - configs->configs, - & ddx_version, - & dri_version, - & drm_version, - & framebuffer, - pSAREA, - fd, - api_ver, - & interface_methods, - & driver_modes ); - - filter_modes( & configs->configs, - driver_modes ); - _gl_context_modes_destroy( driver_modes ); - } - } - } - } - } - } - } - } - - if ( psp == NULL ) { - if ( pSAREA != MAP_FAILED ) { - (void)drmUnmap(pSAREA, SAREA_MAX); - } - - if ( framebuffer.base != MAP_FAILED ) { - (void)drmUnmap((drmAddress)framebuffer.base, framebuffer.size); - } + xGLXGetFBConfigsReq *fb_req; + xGLXGetFBConfigsSGIXReq *sgi_req; + xGLXVendorPrivateWithReplyReq *vpreq; + xGLXGetFBConfigsReply reply; + __GLXscreenConfigs *psc; - if ( framebuffer.dev_priv != NULL ) { - Xfree(framebuffer.dev_priv); - } + psc = priv->screenConfigs + screen; + psc->serverGLXexts = __glXGetStringFromServer(dpy, priv->majorOpcode, + X_GLXQueryServerString, + screen, GLX_EXTENSIONS); - if ( fd >= 0 ) { - (void)drmCloseOnce(fd); - } + LockDisplay(dpy); - (void)XF86DRICloseConnection(dpy, scrn); + psc->configs = NULL; + if (atof(priv->serverGLXversion) >= 1.3) { + GetReq(GLXGetFBConfigs, fb_req); + fb_req->reqType = priv->majorOpcode; + fb_req->glxCode = X_GLXGetFBConfigs; + fb_req->screen = screen; + } else if (strstr(psc->serverGLXexts, "GLX_SGIX_fbconfig") != NULL) { + GetReqExtra(GLXVendorPrivateWithReply, + sz_xGLXGetFBConfigsSGIXReq + + sz_xGLXVendorPrivateWithReplyReq, vpreq); + sgi_req = (xGLXGetFBConfigsSGIXReq *) vpreq; + sgi_req->reqType = priv->majorOpcode; + sgi_req->glxCode = X_GLXVendorPrivateWithReply; + sgi_req->vendorCode = X_GLXvop_GetFBConfigsSGIX; + sgi_req->screen = screen; + } else + goto out; - if ( err_extra != NULL ) { - fprintf(stderr, "libGL error: %s failed (%s)\n", err_msg, - err_extra); - } - else { - fprintf(stderr, "libGL error: %s failed\n", err_msg ); - } + if (!_XReply(dpy, (xReply*) &reply, 0, False)) + goto out; - fprintf(stderr, "libGL error: reverting to (slow) indirect rendering\n"); - } -#endif /* !GLX_USE_APPLEGL */ + psc->configs = createConfigsFromProperties(dpy, + reply.numFBConfigs, + reply.numAttribs * 2, + screen, GL_TRUE); - return psp; + out: + UnlockDisplay(dpy); + return psc->configs != NULL; } -#endif /* GLX_DIRECT_RENDERING */ - /* ** Allocate the memory for the per screen configs for each screen. @@ -1006,17 +563,8 @@ CallCreateNewScreen(Display *dpy, int scrn, __DRIscreen *psc, */ static Bool AllocAndFetchScreenConfigs(Display *dpy, __GLXdisplayPrivate *priv) { - xGLXGetVisualConfigsReq *req; - xGLXGetFBConfigsReq *fb_req; - xGLXVendorPrivateWithReplyReq *vpreq; - xGLXGetFBConfigsSGIXReq *sgi_req; - xGLXGetVisualConfigsReply reply; __GLXscreenConfigs *psc; - __GLcontextModes *config; - GLint i, j, nprops, screens; - INT32 buf[__GLX_TOTAL_CONFIG], *props; - unsigned supported_request = 0; - unsigned prop_size; + GLint i, screens; /* ** First allocate memory for the array of per screen configs. @@ -1030,159 +578,30 @@ static Bool AllocAndFetchScreenConfigs(Display *dpy, __GLXdisplayPrivate *priv) priv->screenConfigs = psc; priv->serverGLXversion = __glXGetStringFromServer(dpy, priv->majorOpcode, - X_GLXQueryServerString, - 0, GLX_VERSION); + X_GLXQueryServerString, + 0, GLX_VERSION); if ( priv->serverGLXversion == NULL ) { FreeScreenConfigs(priv); return GL_FALSE; } - if ( atof( priv->serverGLXversion ) >= 1.3 ) { - supported_request = 1; - } - - /* - ** Now fetch each screens configs structures. If a screen supports - ** GL (by returning a numVisuals > 0) then allocate memory for our - ** config structure and then fill it in. - */ for (i = 0; i < screens; i++, psc++) { - if ( supported_request != 1 ) { - psc->serverGLXexts = __glXGetStringFromServer(dpy, priv->majorOpcode, - X_GLXQueryServerString, - i, GLX_EXTENSIONS); - if ( strstr( psc->serverGLXexts, "GLX_SGIX_fbconfig" ) != NULL ) { - supported_request = 2; - } - else { - supported_request = 3; - } - } - - - LockDisplay(dpy); - switch( supported_request ) { - case 1: - GetReq(GLXGetFBConfigs,fb_req); - fb_req->reqType = priv->majorOpcode; - fb_req->glxCode = X_GLXGetFBConfigs; - fb_req->screen = i; - break; - - case 2: - GetReqExtra(GLXVendorPrivateWithReply, - sz_xGLXGetFBConfigsSGIXReq-sz_xGLXVendorPrivateWithReplyReq,vpreq); - sgi_req = (xGLXGetFBConfigsSGIXReq *) vpreq; - sgi_req->reqType = priv->majorOpcode; - sgi_req->glxCode = X_GLXVendorPrivateWithReply; - sgi_req->vendorCode = X_GLXvop_GetFBConfigsSGIX; - sgi_req->screen = i; - break; - - case 3: - GetReq(GLXGetVisualConfigs,req); - req->reqType = priv->majorOpcode; - req->glxCode = X_GLXGetVisualConfigs; - req->screen = i; - break; - } - - if (!_XReply(dpy, (xReply*) &reply, 0, False)) { - /* Something is busted. Punt. */ - UnlockDisplay(dpy); - SyncHandle(); - FreeScreenConfigs(priv); - return GL_FALSE; - } - - if (!reply.numVisuals) { - /* This screen does not support GL rendering */ - UnlockDisplay(dpy); - continue; - } - - /* FIXME: Is the __GLX_MIN_CONFIG_PROPS test correct for - * FIXME: FBconfigs? - */ - /* Check number of properties */ - nprops = reply.numProps; - if ((nprops < __GLX_MIN_CONFIG_PROPS) || - (nprops > __GLX_MAX_CONFIG_PROPS)) { - /* Huh? Not in protocol defined limits. Punt */ - UnlockDisplay(dpy); - SyncHandle(); - FreeScreenConfigs(priv); - return GL_FALSE; - } - - /* Allocate memory for our config structure */ - psc->configs = _gl_context_modes_create(reply.numVisuals, - sizeof(__GLcontextModes)); - if (!psc->configs) { - UnlockDisplay(dpy); - SyncHandle(); - FreeScreenConfigs(priv); - return GL_FALSE; - } - - /* Allocate memory for the properties, if needed */ - if ( supported_request != 3 ) { - nprops *= 2; - } - - prop_size = nprops * __GLX_SIZE_INT32; - - if (prop_size <= sizeof(buf)) { - props = buf; - } else { - props = (INT32 *) Xmalloc(prop_size); - } - - /* Read each config structure and convert it into our format */ - config = psc->configs; - for (j = 0; j < reply.numVisuals; j++) { - assert( config != NULL ); - _XRead(dpy, (char *)props, prop_size); - - if ( supported_request != 3 ) { - config->rgbMode = GL_TRUE; - config->drawableType = GLX_WINDOW_BIT; - } - else { - config->drawableType = GLX_WINDOW_BIT | GLX_PIXMAP_BIT; - } - - __glXInitializeVisualConfigFromTags( config, nprops, props, - (supported_request != 3), - GL_TRUE ); - if ( config->fbconfigID == GLX_DONT_CARE ) { - config->fbconfigID = config->visualID; - } - config->screen = i; - config = config->next; - } - if (props != buf) { - Xfree((char *)props); - } - UnlockDisplay(dpy); + getVisualConfigs(dpy, priv, i); + getFBConfigs(dpy, priv, i); + psc->scr = i; + psc->dpy = dpy; #ifdef GLX_DIRECT_RENDERING - /* Initialize per screen dynamic client GLX extensions */ - psc->ext_list_first_time = GL_TRUE; - /* Initialize the direct rendering per screen data and functions */ - if (priv->driDisplay.private != NULL) { - /* FIXME: Should it be some sort of an error if createNewScreen[i] - * FIXME: is NULL? - */ - if (priv->driDisplay.createNewScreen && - priv->driDisplay.createNewScreen[i]) { - - psc->driScreen.screenConfigs = (void *)psc; - psc->driScreen.private = - CallCreateNewScreen(dpy, i, & psc->driScreen, - & priv->driDisplay, - priv->driDisplay.createNewScreen[i] ); - } + psc->drawHash = __glxHashCreate(); + if (psc->drawHash == NULL) + continue; + if (priv->dri2Display) + psc->driScreen = (*priv->dri2Display->createScreen)(psc, i, priv); + if (psc->driScreen == NULL && priv->driDisplay) + psc->driScreen = (*priv->driDisplay->createScreen)(psc, i, priv); + if (psc->driScreen == NULL) { + __glxHashDestroy(psc->drawHash); + psc->drawHash = NULL; } #endif } @@ -1193,7 +612,7 @@ static Bool AllocAndFetchScreenConfigs(Display *dpy, __GLXdisplayPrivate *priv) /* ** Initialize the client side extension code. */ -__GLXdisplayPrivate *__glXInitialize(Display* dpy) +_X_HIDDEN __GLXdisplayPrivate *__glXInitialize(Display* dpy) { XExtDisplayInfo *info = __glXFindDisplay(dpy); XExtData **privList, *private, *found; @@ -1273,8 +692,8 @@ __GLXdisplayPrivate *__glXInitialize(Display* dpy) ** (e.g., those called in AllocAndFetchScreenConfigs). */ if (getenv("LIBGL_ALWAYS_INDIRECT") == NULL) { - dpyPriv->driDisplay.private = - driCreateDisplay(dpy, &dpyPriv->driDisplay); + dpyPriv->dri2Display = dri2CreateDisplay(dpy); + dpyPriv->driDisplay = driCreateDisplay(dpy); } #endif @@ -1308,7 +727,7 @@ __GLXdisplayPrivate *__glXInitialize(Display* dpy) ** Setup for sending a GLX command on dpy. Make sure the extension is ** initialized. Try to avoid calling __glXInitialize as its kinda slow. */ -CARD8 __glXSetupForCommand(Display *dpy) +_X_HIDDEN CARD8 __glXSetupForCommand(Display *dpy) { GLXContext gc; __GLXdisplayPrivate *priv; @@ -1349,7 +768,7 @@ CARD8 __glXSetupForCommand(Display *dpy) * Modify this function to use \c ctx->pc instead of the explicit * \c pc parameter. */ -GLubyte *__glXFlushRenderBuffer(__GLXcontext *ctx, GLubyte *pc) +_X_HIDDEN GLubyte *__glXFlushRenderBuffer(__GLXcontext *ctx, GLubyte *pc) { Display * const dpy = ctx->currentDpy; #ifdef USE_XCB @@ -1361,7 +780,8 @@ GLubyte *__glXFlushRenderBuffer(__GLXcontext *ctx, GLubyte *pc) if ( (dpy != NULL) && (size > 0) ) { #ifdef USE_XCB - xcb_glx_render(c, ctx->currentContextTag, size, (char *)ctx->buf); + xcb_glx_render(c, ctx->currentContextTag, size, + (const uint8_t *)ctx->buf); #else /* Send the entire buffer as an X request */ LockDisplay(dpy); @@ -1398,9 +818,9 @@ GLubyte *__glXFlushRenderBuffer(__GLXcontext *ctx, GLubyte *pc) * \param data Command data. * \param dataLen Size, in bytes, of the command data. */ -void __glXSendLargeChunk(__GLXcontext *gc, GLint requestNumber, - GLint totalRequests, - const GLvoid * data, GLint dataLen) +_X_HIDDEN void __glXSendLargeChunk(__GLXcontext *gc, GLint requestNumber, + GLint totalRequests, + const GLvoid * data, GLint dataLen) { Display *dpy = gc->currentDpy; #ifdef USE_XCB @@ -1446,9 +866,9 @@ void __glXSendLargeChunk(__GLXcontext *gc, GLint requestNumber, * \param data Command data. * \param dataLen Size, in bytes, of the command data. */ -void __glXSendLargeCommand(__GLXcontext *ctx, - const GLvoid *header, GLint headerLen, - const GLvoid *data, GLint dataLen) +_X_HIDDEN void __glXSendLargeCommand(__GLXcontext *ctx, + const GLvoid *header, GLint headerLen, + const GLvoid *data, GLint dataLen) { GLint maxSize; GLint totalRequests, requestNumber; @@ -1484,330 +904,8 @@ void __glXSendLargeCommand(__GLXcontext *ctx, /************************************************************************/ -PUBLIC GLXContext glXGetCurrentContext(void) -{ - GLXContext cx = __glXGetCurrentContext(); - - if (cx == &dummyContext) { - return NULL; - } else { - return cx; - } -} - -PUBLIC GLXDrawable glXGetCurrentDrawable(void) -{ - GLXContext gc = __glXGetCurrentContext(); - return gc->currentDrawable; -} - - -/************************************************************************/ - -#ifdef GLX_DIRECT_RENDERING -/* Return the DRI per screen structure */ -__DRIscreen *__glXFindDRIScreen(__DRInativeDisplay *dpy, int scrn) -{ - __DRIscreen *pDRIScreen = NULL; - XExtDisplayInfo *info = __glXFindDisplay(dpy); - XExtData **privList, *found; - __GLXdisplayPrivate *dpyPriv; - XEDataObject dataObj; - - __glXLock(); - dataObj.display = dpy; - privList = XEHeadOfExtensionList(dataObj); - found = XFindOnExtensionList(privList, info->codes->extension); - __glXUnlock(); - - if (found) { - dpyPriv = (__GLXdisplayPrivate *)found->private_data; - pDRIScreen = &dpyPriv->screenConfigs[scrn].driScreen; - } - - return pDRIScreen; -} -#endif - -/************************************************************************/ - -static Bool SendMakeCurrentRequest( Display *dpy, CARD8 opcode, - GLXContextID gc, GLXContextTag old_gc, GLXDrawable draw, GLXDrawable read, - xGLXMakeCurrentReply * reply ); - -/** - * Sends a GLX protocol message to the specified display to make the context - * and the drawables current. - * - * \param dpy Display to send the message to. - * \param opcode Major opcode value for the display. - * \param gc_id Context tag for the context to be made current. - * \param draw Drawable ID for the "draw" drawable. - * \param read Drawable ID for the "read" drawable. - * \param reply Space to store the X-server's reply. - * - * \warning - * This function assumes that \c dpy is locked with \c LockDisplay on entry. - */ -static Bool SendMakeCurrentRequest(Display *dpy, CARD8 opcode, - GLXContextID gc_id, GLXContextTag gc_tag, - GLXDrawable draw, GLXDrawable read, - xGLXMakeCurrentReply *reply) -{ - Bool ret; - - - LockDisplay(dpy); - - if (draw == read) { - xGLXMakeCurrentReq *req; - - GetReq(GLXMakeCurrent,req); - req->reqType = opcode; - req->glxCode = X_GLXMakeCurrent; - req->drawable = draw; - req->context = gc_id; - req->oldContextTag = gc_tag; - } - else { - __GLXdisplayPrivate *priv = __glXInitialize(dpy); - - /* If the server can support the GLX 1.3 version, we should - * perfer that. Not only that, some servers support GLX 1.3 but - * not the SGI extension. - */ - - if ((priv->majorVersion > 1) || (priv->minorVersion >= 3)) { - xGLXMakeContextCurrentReq *req; - - GetReq(GLXMakeContextCurrent,req); - req->reqType = opcode; - req->glxCode = X_GLXMakeContextCurrent; - req->drawable = draw; - req->readdrawable = read; - req->context = gc_id; - req->oldContextTag = gc_tag; - } - else { - xGLXVendorPrivateWithReplyReq *vpreq; - xGLXMakeCurrentReadSGIReq *req; - - GetReqExtra(GLXVendorPrivateWithReply, - sz_xGLXMakeCurrentReadSGIReq-sz_xGLXVendorPrivateWithReplyReq,vpreq); - req = (xGLXMakeCurrentReadSGIReq *)vpreq; - req->reqType = opcode; - req->glxCode = X_GLXVendorPrivateWithReply; - req->vendorCode = X_GLXvop_MakeCurrentReadSGI; - req->drawable = draw; - req->readable = read; - req->context = gc_id; - req->oldContextTag = gc_tag; - } - } - - ret = _XReply(dpy, (xReply*) reply, 0, False); - - UnlockDisplay(dpy); - SyncHandle(); - - return ret; -} - - -#ifdef GLX_DIRECT_RENDERING -static Bool BindContextWrapper( Display *dpy, GLXContext gc, - GLXDrawable draw, GLXDrawable read ) -{ - return (*gc->driContext.bindContext)(dpy, gc->screen, draw, read, - & gc->driContext); -} - - -static Bool UnbindContextWrapper( GLXContext gc ) -{ - return (*gc->driContext.unbindContext)(gc->currentDpy, gc->screen, - gc->currentDrawable, - gc->currentReadable, - & gc->driContext ); -} -#endif /* GLX_DIRECT_RENDERING */ - - -/** - * Make a particular context current. - * - * \note This is in this file so that it can access dummyContext. - */ -USED static Bool MakeContextCurrent(Display *dpy, GLXDrawable draw, - GLXDrawable read, GLXContext gc) -{ - xGLXMakeCurrentReply reply; - const GLXContext oldGC = __glXGetCurrentContext(); - const CARD8 opcode = __glXSetupForCommand(dpy); - const CARD8 oldOpcode = ((gc == oldGC) || (oldGC == &dummyContext)) - ? opcode : __glXSetupForCommand(oldGC->currentDpy); - Bool bindReturnValue; - - - if (!opcode || !oldOpcode) { - return GL_FALSE; - } - - /* Make sure that the new context has a nonzero ID. In the request, - * a zero context ID is used only to mean that we bind to no current - * context. - */ - if ((gc != NULL) && (gc->xid == None)) { - return GL_FALSE; - } - -#ifndef GLX_DIRECT_RENDERING - if (gc && gc->isDirect) { - return GL_FALSE; - } -#endif - - _glapi_check_multithread(); - -#ifdef GLX_DIRECT_RENDERING - /* Bind the direct rendering context to the drawable */ - if (gc && gc->isDirect) { - bindReturnValue = (gc->driContext.private) - ? BindContextWrapper(dpy, gc, draw, read) - : False; - } else -#endif - { - /* Send a glXMakeCurrent request to bind the new context. */ - bindReturnValue = - SendMakeCurrentRequest(dpy, opcode, gc ? gc->xid : None, - ((dpy != oldGC->currentDpy) || oldGC->isDirect) - ? None : oldGC->currentContextTag, - draw, read, &reply); - } - - - if (!bindReturnValue) { - return False; - } - - if ((dpy != oldGC->currentDpy || (gc && gc->isDirect)) && - !oldGC->isDirect && oldGC != &dummyContext) { - xGLXMakeCurrentReply dummy_reply; - - /* We are either switching from one dpy to another and have to - * send a request to the previous dpy to unbind the previous - * context, or we are switching away from a indirect context to - * a direct context and have to send a request to the dpy to - * unbind the previous context. - */ - (void) SendMakeCurrentRequest(oldGC->currentDpy, oldOpcode, None, - oldGC->currentContextTag, None, None, - & dummy_reply); - } -#ifdef GLX_DIRECT_RENDERING - else if (oldGC->isDirect && oldGC->driContext.private) { - (void) UnbindContextWrapper(oldGC); - } -#endif - - - /* Update our notion of what is current */ - __glXLock(); - if (gc == oldGC) { - /* Even though the contexts are the same the drawable might have - * changed. Note that gc cannot be the dummy, and that oldGC - * cannot be NULL, therefore if they are the same, gc is not - * NULL and not the dummy. - */ - gc->currentDrawable = draw; - gc->currentReadable = read; - } else { - if (oldGC != &dummyContext) { - /* Old current context is no longer current to anybody */ - oldGC->currentDpy = 0; - oldGC->currentDrawable = None; - oldGC->currentReadable = None; - oldGC->currentContextTag = 0; - - if (oldGC->xid == None) { - /* We are switching away from a context that was - * previously destroyed, so we need to free the memory - * for the old handle. - */ -#ifdef GLX_DIRECT_RENDERING - /* Destroy the old direct rendering context */ - if (oldGC->isDirect) { - if (oldGC->driContext.private) { - (*oldGC->driContext.destroyContext) - (dpy, oldGC->screen, oldGC->driContext.private); - oldGC->driContext.private = NULL; - } - } -#endif - __glXFreeContext(oldGC); - } - } - if (gc) { - __glXSetCurrentContext(gc); - - gc->currentDpy = dpy; - gc->currentDrawable = draw; - gc->currentReadable = read; - - if (!gc->isDirect) { - if (!IndirectAPI) - IndirectAPI = __glXNewIndirectAPI(); - _glapi_set_dispatch(IndirectAPI); - -#ifdef GLX_USE_APPLEGL - do { - extern void XAppleDRIUseIndirectDispatch(void); - XAppleDRIUseIndirectDispatch(); - } while (0); -#endif - - __GLXattribute *state = - (__GLXattribute *)(gc->client_state_private); - - gc->currentContextTag = reply.contextTag; - if (state->array_state == NULL) { - (void) glGetString(GL_EXTENSIONS); - (void) glGetString(GL_VERSION); - __glXInitVertexArrayState(gc); - } - } - else { - gc->currentContextTag = -1; - } - } else { - __glXSetCurrentContext(&dummyContext); -#ifdef GLX_DIRECT_RENDERING - _glapi_set_dispatch(NULL); /* no-op functions */ -#endif - } - } - __glXUnlock(); - return GL_TRUE; -} - - -PUBLIC Bool glXMakeCurrent(Display *dpy, GLXDrawable draw, GLXContext gc) -{ - return MakeContextCurrent( dpy, draw, draw, gc ); -} - -PUBLIC GLX_ALIAS(Bool, glXMakeCurrentReadSGI, - (Display *dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx), - (dpy, d, r, ctx), MakeContextCurrent) - -PUBLIC GLX_ALIAS(Bool, glXMakeContextCurrent, - (Display *dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx), - (dpy, d, r, ctx), MakeContextCurrent) - - #ifdef DEBUG -void __glXDumpDrawBuffer(__GLXcontext *ctx) +_X_HIDDEN void __glXDumpDrawBuffer(__GLXcontext *ctx) { GLubyte *p = ctx->buf; GLubyte *end = ctx->pc; @@ -1832,9 +930,23 @@ void __glXDumpDrawBuffer(__GLXcontext *ctx) #ifdef USE_SPARC_ASM /* - * Used only when we are sparc, using sparc assembler. + * This is where our dispatch table's bounds are. + * And the static mesa_init is taken directly from + * Mesa's 'sparc.c' initializer. * + * We need something like this here, because this version + * of openGL/glx never initializes a Mesa context, and so + * the address of the dispatch table pointer never gets stuffed + * into the dispatch jump table otherwise. + * + * It matters only on SPARC, and only if you are using assembler + * code instead of C-code indirect dispatch. + * + * -- FEM, 04.xii.03 */ +extern unsigned int _mesa_sparc_glapi_begin; +extern unsigned int _mesa_sparc_glapi_end; +extern void __glapi_sparc_icache_flush(unsigned int *); static void _glx_mesa_init_sparc_glapi_relocs(void) diff --git a/src/glx/x11/glxextensions.c b/src/glx/x11/glxextensions.c index 1d99b61db0..e843718472 100644 --- a/src/glx/x11/glxextensions.c +++ b/src/glx/x11/glxextensions.c @@ -356,28 +356,15 @@ __glXProcessServerString( const struct extension_info * ext, } } - -/** - * Enable a named GLX extension on a given screen. - * Drivers should not call this function directly. They should instead use - * \c glXGetProcAddress to obtain a pointer to the function. - * - * \param psc Pointer to GLX per-screen record. - * \param name Name of the extension to enable. - * - * \sa glXGetProcAddress - * - * \since Internal API version 20030813. - */ void -__glXScrEnableExtension( __GLXscreenConfigs *psc, const char * name ) +__glXEnableDirectExtension(__GLXscreenConfigs *psc, const char *name) { - __glXExtensionsCtr(); - __glXExtensionsCtrScreen(psc); - set_glx_extension( known_glx_extensions, name, strlen( name ), GL_TRUE, - psc->direct_support ); -} + __glXExtensionsCtr(); + __glXExtensionsCtrScreen(psc); + set_glx_extension(known_glx_extensions, + name, strlen(name), GL_TRUE, psc->direct_support); +} /** * Initialize global extension support tables. diff --git a/src/glx/x11/glxextensions.h b/src/glx/x11/glxextensions.h index a4241b6b7f..9cdd05ed76 100644 --- a/src/glx/x11/glxextensions.h +++ b/src/glx/x11/glxextensions.h @@ -234,7 +234,7 @@ extern GLboolean __glXExtensionBitIsEnabled( struct __GLXscreenConfigsRec *psc, extern const char * __glXGetClientExtensions( void ); extern void __glXCalculateUsableExtensions( struct __GLXscreenConfigsRec *psc, GLboolean display_is_direct_capable, int server_minor_version ); -extern void __glXScrEnableExtension( struct __GLXscreenConfigsRec *psc, const char * name ); + extern void __glXCalculateUsableGLExtensions( struct __GLXcontextRec * gc, const char * server_string, int major_version, int minor_version ); extern void __glXGetGLVersion( int * major_version, int * minor_version ); @@ -243,6 +243,8 @@ extern char * __glXGetClientGLExtensionString( void ); extern GLboolean __glExtensionBitIsEnabled( const struct __GLXcontextRec * gc, unsigned bit ); +extern void +__glXEnableDirectExtension(__GLXscreenConfigs *psc, const char *name); /* Source-level backwards compatibility with old drivers. They won't * find the respective functions, though. diff --git a/src/glx/x11/indirect.c b/src/glx/x11/indirect.c index fbb2a91956..871adddb95 100644 --- a/src/glx/x11/indirect.c +++ b/src/glx/x11/indirect.c @@ -300,7 +300,9 @@ __indirect_glNewList(GLuint list, GLenum mode) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -324,7 +326,9 @@ __indirect_glEndList(void) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 0; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -360,6 +364,10 @@ __indirect_glCallLists(GLsizei n, GLenum type, const GLvoid * lists) __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint compsize = __glCallLists_size(type); const GLuint cmdlen = 12 + __GLX_PAD((compsize * n)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -393,7 +401,9 @@ __indirect_glDeleteLists(GLuint list, GLsizei range) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -418,7 +428,9 @@ __indirect_glGenLists(GLsizei range) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLuint retval = (GLuint) 0; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3582,6 +3594,10 @@ __indirect_glPixelMapfv(GLenum map, GLsizei mapsize, const GLfloat * values) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((mapsize * 4)); + if (mapsize < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((mapsize >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -3615,6 +3631,10 @@ __indirect_glPixelMapuiv(GLenum map, GLsizei mapsize, const GLuint * values) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((mapsize * 4)); + if (mapsize < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((mapsize >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -3648,6 +3668,10 @@ __indirect_glPixelMapusv(GLenum map, GLsizei mapsize, const GLushort * values) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((mapsize * 2)); + if (mapsize < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((mapsize >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -3716,7 +3740,9 @@ __indirect_glReadPixels(GLint x, GLint y, GLsizei width, GLsizei height, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 28; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3808,7 +3834,9 @@ __indirect_glGetClipPlane(GLenum plane, GLdouble * equation) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3841,7 +3869,9 @@ __indirect_glGetLightfv(GLenum light, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3879,7 +3909,9 @@ __indirect_glGetLightiv(GLenum light, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3917,7 +3949,9 @@ __indirect_glGetMapdv(GLenum target, GLenum query, GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3953,7 +3987,9 @@ __indirect_glGetMapfv(GLenum target, GLenum query, GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3989,7 +4025,9 @@ __indirect_glGetMapiv(GLenum target, GLenum query, GLint * v) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4025,7 +4063,9 @@ __indirect_glGetMaterialfv(GLenum face, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4063,7 +4103,9 @@ __indirect_glGetMaterialiv(GLenum face, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4101,7 +4143,9 @@ __indirect_glGetPixelMapfv(GLenum map, GLfloat * values) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4137,7 +4181,9 @@ __indirect_glGetPixelMapuiv(GLenum map, GLuint * values) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4174,7 +4220,9 @@ __indirect_glGetPixelMapusv(GLenum map, GLushort * values) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4210,9 +4258,10 @@ void __indirect_glGetPolygonStipple(GLubyte *mask) { __GLXcontext *const gc = __glXGetCurrentContext(); - const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4247,7 +4296,9 @@ __indirect_glGetTexEnvfv(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4285,7 +4336,9 @@ __indirect_glGetTexEnviv(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4323,7 +4376,9 @@ __indirect_glGetTexGendv(GLenum coord, GLenum pname, GLdouble * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4361,7 +4416,9 @@ __indirect_glGetTexGenfv(GLenum coord, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4399,7 +4456,9 @@ __indirect_glGetTexGeniv(GLenum coord, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4439,7 +4498,9 @@ __indirect_glGetTexImage(GLenum target, GLint level, GLenum format, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 20; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4483,7 +4544,9 @@ __indirect_glGetTexParameterfv(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4522,7 +4585,9 @@ __indirect_glGetTexParameteriv(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4562,7 +4627,9 @@ __indirect_glGetTexLevelParameterfv(GLenum target, GLint level, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 12; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4603,7 +4670,9 @@ __indirect_glGetTexLevelParameteriv(GLenum target, GLint level, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 12; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4644,7 +4713,9 @@ __indirect_glIsList(GLuint list) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5012,7 +5083,13 @@ __indirect_glAreTexturesResident(GLsizei n, const GLuint * textures, __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; +#ifndef USE_XCB const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); +#endif + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return 0; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5047,7 +5124,7 @@ glAreTexturesResidentEXT(GLsizei n, const GLuint * textures, { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { return CALL_AreTexturesResident(GET_DISPATCH(), (n, textures, residences)); } else { @@ -5055,6 +5132,10 @@ glAreTexturesResidentEXT(GLsizei n, const GLuint * textures, Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return 0; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -5163,7 +5244,13 @@ __indirect_glDeleteTextures(GLsizei n, const GLuint * textures) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); +#endif + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5187,12 +5274,16 @@ glDeleteTexturesEXT(GLsizei n, const GLuint * textures) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_DeleteTextures(GET_DISPATCH(), (n, textures)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivate, @@ -5212,7 +5303,13 @@ __indirect_glGenTextures(GLsizei n, GLuint * textures) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5245,12 +5342,16 @@ glGenTexturesEXT(GLsizei n, GLuint * textures) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GenTextures(GET_DISPATCH(), (n, textures)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4; + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -5271,7 +5372,9 @@ __indirect_glIsTexture(GLuint texture) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5301,7 +5404,7 @@ glIsTextureEXT(GLuint texture) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { return CALL_IsTexture(GET_DISPATCH(), (texture)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -5328,6 +5431,10 @@ __indirect_glPrioritizeTextures(GLsizei n, const GLuint * textures, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)) + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_PrioritizeTextures, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&n), 4); @@ -5565,7 +5672,9 @@ __indirect_glGetColorTable(GLenum target, GLenum format, GLenum type, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 16; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5609,7 +5718,7 @@ glGetColorTableEXT(GLenum target, GLenum format, GLenum type, GLvoid * table) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetColorTable(GET_DISPATCH(), (target, format, type, table)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -5641,7 +5750,9 @@ __indirect_glGetColorTableParameterfv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5680,7 +5791,7 @@ glGetColorTableParameterfvEXT(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetColorTableParameterfv(GET_DISPATCH(), (target, pname, params)); } else { @@ -5709,7 +5820,9 @@ __indirect_glGetColorTableParameteriv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5748,7 +5861,7 @@ glGetColorTableParameterivEXT(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetColorTableParameteriv(GET_DISPATCH(), (target, pname, params)); } else { @@ -6029,7 +6142,9 @@ __indirect_glGetConvolutionFilter(GLenum target, GLenum format, GLenum type, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 16; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6069,7 +6184,7 @@ gl_dispatch_stub_356(GLenum target, GLenum format, GLenum type, { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetConvolutionFilter(GET_DISPATCH(), (target, format, type, image)); } else { @@ -6103,7 +6218,9 @@ __indirect_glGetConvolutionParameterfv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6142,7 +6259,7 @@ gl_dispatch_stub_357(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetConvolutionParameterfv(GET_DISPATCH(), (target, pname, params)); } else { @@ -6171,7 +6288,9 @@ __indirect_glGetConvolutionParameteriv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6210,7 +6329,7 @@ gl_dispatch_stub_358(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetConvolutionParameteriv(GET_DISPATCH(), (target, pname, params)); } else { @@ -6240,7 +6359,9 @@ __indirect_glGetHistogram(GLenum target, GLboolean reset, GLenum format, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 16; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6285,7 +6406,7 @@ gl_dispatch_stub_361(GLenum target, GLboolean reset, GLenum format, { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetHistogram(GET_DISPATCH(), (target, reset, format, type, values)); } else { @@ -6319,7 +6440,9 @@ __indirect_glGetHistogramParameterfv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6357,7 +6480,7 @@ gl_dispatch_stub_362(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetHistogramParameterfv(GET_DISPATCH(), (target, pname, params)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -6385,7 +6508,9 @@ __indirect_glGetHistogramParameteriv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6423,7 +6548,7 @@ gl_dispatch_stub_363(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetHistogramParameteriv(GET_DISPATCH(), (target, pname, params)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -6452,7 +6577,9 @@ __indirect_glGetMinmax(GLenum target, GLboolean reset, GLenum format, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 16; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6493,7 +6620,7 @@ gl_dispatch_stub_364(GLenum target, GLboolean reset, GLenum format, { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetMinmax(GET_DISPATCH(), (target, reset, format, type, values)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -6526,7 +6653,9 @@ __indirect_glGetMinmaxParameterfv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6562,7 +6691,7 @@ gl_dispatch_stub_365(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetMinmaxParameterfv(GET_DISPATCH(), (target, pname, params)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -6589,7 +6718,9 @@ __indirect_glGetMinmaxParameteriv(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6625,7 +6756,7 @@ gl_dispatch_stub_366(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetMinmaxParameteriv(GET_DISPATCH(), (target, pname, params)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -7468,6 +7599,26 @@ __indirect_glGetProgramivARB(GLenum target, GLenum pname, GLint * params) return; } +#define X_GLrop_ProgramEnvParameter4dvARB 4185 +void +__indirect_glProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x, + GLdouble y, GLdouble z, GLdouble w) +{ + __GLXcontext *const gc = __glXGetCurrentContext(); + const GLuint cmdlen = 44; + emit_header(gc->pc, X_GLrop_ProgramEnvParameter4dvARB, cmdlen); + (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); + (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); + (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 8); + (void) memcpy((void *) (gc->pc + 20), (void *) (&y), 8); + (void) memcpy((void *) (gc->pc + 28), (void *) (&z), 8); + (void) memcpy((void *) (gc->pc + 36), (void *) (&w), 8); + gc->pc += cmdlen; + if (__builtin_expect(gc->pc > gc->limit, 0)) { + (void) __glXFlushRenderBuffer(gc, gc->pc); + } +} + #define X_GLrop_ProgramEnvParameter4dvARB 4185 void __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, @@ -7485,6 +7636,26 @@ __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, } } +#define X_GLrop_ProgramEnvParameter4fvARB 4184 +void +__indirect_glProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x, + GLfloat y, GLfloat z, GLfloat w) +{ + __GLXcontext *const gc = __glXGetCurrentContext(); + const GLuint cmdlen = 28; + emit_header(gc->pc, X_GLrop_ProgramEnvParameter4fvARB, cmdlen); + (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); + (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); + (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 4); + (void) memcpy((void *) (gc->pc + 16), (void *) (&y), 4); + (void) memcpy((void *) (gc->pc + 20), (void *) (&z), 4); + (void) memcpy((void *) (gc->pc + 24), (void *) (&w), 4); + gc->pc += cmdlen; + if (__builtin_expect(gc->pc > gc->limit, 0)) { + (void) __glXFlushRenderBuffer(gc, gc->pc); + } +} + #define X_GLrop_ProgramEnvParameter4fvARB 4184 void __indirect_glProgramEnvParameter4fvARB(GLenum target, GLuint index, @@ -7585,6 +7756,10 @@ __indirect_glProgramStringARB(GLenum target, GLenum format, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 16 + __GLX_PAD(len); + if (len < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((len >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -8200,7 +8375,13 @@ __indirect_glDeleteQueriesARB(GLsizei n, const GLuint * ids) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); +#endif + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8238,7 +8419,13 @@ __indirect_glGenQueriesARB(GLsizei n, GLuint * ids) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8271,7 +8458,9 @@ __indirect_glGetQueryObjectivARB(GLuint id, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8307,7 +8496,9 @@ __indirect_glGetQueryObjectuivARB(GLuint id, GLenum pname, GLuint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8343,7 +8534,9 @@ __indirect_glGetQueryivARB(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; +#ifndef USE_XCB const GLuint cmdlen = 8; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8383,7 +8576,9 @@ __indirect_glIsQueryARB(GLuint id) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; +#ifndef USE_XCB const GLuint cmdlen = 4; +#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8414,6 +8609,10 @@ __indirect_glDrawBuffersARB(GLsizei n, const GLenum * bufs) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -8774,6 +8973,10 @@ __indirect_glAreProgramsResidentNV(GLsizei n, const GLuint * ids, Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return 0; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -8809,6 +9012,10 @@ __indirect_glDeleteProgramsNV(GLsizei n, const GLuint * programs) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivate, @@ -8845,6 +9052,10 @@ __indirect_glGenProgramsNV(GLsizei n, GLuint * programs) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4; + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -9051,6 +9262,10 @@ __indirect_glLoadProgramNV(GLenum target, GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 16 + __GLX_PAD(len); + if (len < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_LoadProgramNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); @@ -9071,6 +9286,10 @@ __indirect_glProgramParameters4dvNV(GLenum target, GLuint index, GLuint num, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 16 + __GLX_PAD((num * 32)); + if (num < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(num >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramParameters4dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); @@ -9091,6 +9310,10 @@ __indirect_glProgramParameters4fvNV(GLenum target, GLuint index, GLuint num, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 16 + __GLX_PAD((num * 16)); + if (num < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(num >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramParameters4fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); @@ -9110,6 +9333,10 @@ __indirect_glRequestResidentProgramsNV(GLsizei n, const GLuint * ids) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_RequestResidentProgramsNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&n), 4); @@ -9561,6 +9788,10 @@ __indirect_glVertexAttribs1dvNV(GLuint index, GLsizei n, const GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 8)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs1dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9579,6 +9810,10 @@ __indirect_glVertexAttribs1fvNV(GLuint index, GLsizei n, const GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs1fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9597,6 +9832,10 @@ __indirect_glVertexAttribs1svNV(GLuint index, GLsizei n, const GLshort * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 2)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs1svNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9615,6 +9854,10 @@ __indirect_glVertexAttribs2dvNV(GLuint index, GLsizei n, const GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 16)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs2dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9633,6 +9876,10 @@ __indirect_glVertexAttribs2fvNV(GLuint index, GLsizei n, const GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 8)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs2fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9651,6 +9898,10 @@ __indirect_glVertexAttribs2svNV(GLuint index, GLsizei n, const GLshort * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs2svNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9669,6 +9920,10 @@ __indirect_glVertexAttribs3dvNV(GLuint index, GLsizei n, const GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 24)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs3dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9687,6 +9942,10 @@ __indirect_glVertexAttribs3fvNV(GLuint index, GLsizei n, const GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 12)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs3fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9705,6 +9964,10 @@ __indirect_glVertexAttribs3svNV(GLuint index, GLsizei n, const GLshort * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 6)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs3svNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9723,6 +9986,10 @@ __indirect_glVertexAttribs4dvNV(GLuint index, GLsizei n, const GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 32)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs4dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9741,6 +10008,10 @@ __indirect_glVertexAttribs4fvNV(GLuint index, GLsizei n, const GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 16)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs4fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9759,6 +10030,10 @@ __indirect_glVertexAttribs4svNV(GLuint index, GLsizei n, const GLshort * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 8)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs4svNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9777,6 +10052,10 @@ __indirect_glVertexAttribs4ubvNV(GLuint index, GLsizei n, const GLubyte *v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs4ubvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9843,6 +10122,10 @@ __indirect_glGetProgramNamedParameterdvNV(GLuint id, GLsizei len, __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 8 + __GLX_PAD(len); + if (len < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((len >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -9867,6 +10150,10 @@ __indirect_glGetProgramNamedParameterfvNV(GLuint id, GLsizei len, __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 8 + __GLX_PAD(len); + if (len < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((len >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -9890,6 +10177,10 @@ __indirect_glProgramNamedParameter4dNV(GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 44 + __GLX_PAD(len); + if (len < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramNamedParameter4dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8); @@ -9914,6 +10205,10 @@ __indirect_glProgramNamedParameter4dvNV(GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 44 + __GLX_PAD(len); + if (len < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramNamedParameter4dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (v), 32); @@ -9935,6 +10230,10 @@ __indirect_glProgramNamedParameter4fNV(GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 28 + __GLX_PAD(len); + if (len < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramNamedParameter4fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&id), 4); @@ -9959,6 +10258,10 @@ __indirect_glProgramNamedParameter4fvNV(GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 28 + __GLX_PAD(len); + if (len < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramNamedParameter4fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&id), 4); @@ -10044,6 +10347,10 @@ __indirect_glDeleteFramebuffersEXT(GLsizei n, const GLuint * framebuffers) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_DeleteFramebuffersEXT, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&n), 4); @@ -10062,6 +10369,10 @@ __indirect_glDeleteRenderbuffersEXT(GLsizei n, const GLuint * renderbuffers) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)); + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_DeleteRenderbuffersEXT, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&n), 4); @@ -10161,6 +10472,10 @@ __indirect_glGenFramebuffersEXT(GLsizei n, GLuint * framebuffers) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4; + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -10180,6 +10495,10 @@ __indirect_glGenRenderbuffersEXT(GLsizei n, GLuint * renderbuffers) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4; + if (n < 0) { + __glXSetError(gc, GL_INVALID_VALUE); + return; + } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, diff --git a/src/glx/x11/indirect.h b/src/glx/x11/indirect.h index f8c88b36bb..0719a1b302 100644 --- a/src/glx/x11/indirect.h +++ b/src/glx/x11/indirect.h @@ -517,7 +517,9 @@ extern HIDDEN void __indirect_glGetProgramivARB(GLenum target, GLenum pname, GLi extern HIDDEN void __indirect_glGetVertexAttribdvARB(GLuint index, GLenum pname, GLdouble * params); extern HIDDEN void __indirect_glGetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat * params); extern HIDDEN void __indirect_glGetVertexAttribivARB(GLuint index, GLenum pname, GLint * params); +extern HIDDEN void __indirect_glProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); extern HIDDEN void __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, const GLdouble * params); +extern HIDDEN void __indirect_glProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); extern HIDDEN void __indirect_glProgramEnvParameter4fvARB(GLenum target, GLuint index, const GLfloat * params); extern HIDDEN void __indirect_glProgramLocalParameter4dARB(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); extern HIDDEN void __indirect_glProgramLocalParameter4dvARB(GLenum target, GLuint index, const GLdouble * params); diff --git a/src/glx/x11/indirect_init.c b/src/glx/x11/indirect_init.c index 479184337c..852fe712c6 100644 --- a/src/glx/x11/indirect_init.c +++ b/src/glx/x11/indirect_init.c @@ -526,7 +526,9 @@ __GLapi * __glXNewIndirectAPI( void ) glAPI->GetVertexAttribdvARB = __indirect_glGetVertexAttribdvARB; glAPI->GetVertexAttribfvARB = __indirect_glGetVertexAttribfvARB; glAPI->GetVertexAttribivARB = __indirect_glGetVertexAttribivARB; + glAPI->ProgramEnvParameter4dARB = __indirect_glProgramEnvParameter4dARB; glAPI->ProgramEnvParameter4dvARB = __indirect_glProgramEnvParameter4dvARB; + glAPI->ProgramEnvParameter4fARB = __indirect_glProgramEnvParameter4fARB; glAPI->ProgramEnvParameter4fvARB = __indirect_glProgramEnvParameter4fvARB; glAPI->ProgramLocalParameter4dARB = __indirect_glProgramLocalParameter4dARB; glAPI->ProgramLocalParameter4dvARB = __indirect_glProgramLocalParameter4dvARB; diff --git a/src/glx/x11/indirect_vertex_array.c b/src/glx/x11/indirect_vertex_array.c index 90ec277c41..09d7244ba9 100644 --- a/src/glx/x11/indirect_vertex_array.c +++ b/src/glx/x11/indirect_vertex_array.c @@ -32,7 +32,7 @@ #include #include "glxextensions.h" #include "indirect_vertex_array.h" -#include "indirect_va_private.h" +#include "indirect_vertex_array_priv.h" #define __GLX_PAD(n) (((n)+3) & ~3) @@ -485,14 +485,14 @@ emit_DrawArrays_none( GLenum mode, GLint first, GLsizei count ) for ( i = 0 ; i < count ; i++ ) { if ( (pc + single_vertex_size) >= gc->bufEnd ) { - pc = __glXFlushRenderBuffer(gc, gc->pc); + pc = __glXFlushRenderBuffer(gc, pc); } pc = emit_element_none( pc, arrays, first + i ); } if ( (pc + 4) >= gc->bufEnd ) { - pc = __glXFlushRenderBuffer(gc, gc->pc); + pc = __glXFlushRenderBuffer(gc, pc); } (void) memcpy( pc, end_cmd, 4 ); @@ -527,7 +527,7 @@ static GLubyte * emit_DrawArrays_header_old( __GLXcontext * gc, struct array_state_vector * arrays, size_t * elements_per_request, - size_t * total_requests, + unsigned int * total_requests, GLenum mode, GLsizei count ) { size_t command_size; @@ -640,7 +640,7 @@ emit_DrawArrays_old( GLenum mode, GLint first, GLsizei count ) GLubyte * pc; size_t elements_per_request; - size_t total_requests = 0; + unsigned total_requests = 0; unsigned i; size_t total_sent = 0; @@ -726,7 +726,7 @@ emit_DrawElements_none( GLenum mode, GLsizei count, GLenum type, unsigned index = 0; if ( (pc + single_vertex_size) >= gc->bufEnd ) { - pc = __glXFlushRenderBuffer(gc, gc->pc); + pc = __glXFlushRenderBuffer(gc, pc); } switch( type ) { @@ -744,7 +744,7 @@ emit_DrawElements_none( GLenum mode, GLsizei count, GLenum type, } if ( (pc + 4) >= gc->bufEnd ) { - pc = __glXFlushRenderBuffer(gc, gc->pc); + pc = __glXFlushRenderBuffer(gc, pc); } (void) memcpy( pc, end_cmd, 4 ); @@ -770,9 +770,10 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, GLubyte * pc; size_t elements_per_request; - size_t total_requests = 0; + unsigned total_requests = 0; unsigned i; unsigned req; + unsigned req_element=0; pc = emit_DrawArrays_header_old( gc, arrays, & elements_per_request, @@ -790,7 +791,7 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, switch( type ) { case GL_UNSIGNED_INT: { - const GLuint * ui_ptr = (const GLuint *) indices; + const GLuint * ui_ptr = (const GLuint *) indices + req_element; for ( i = 0 ; i < elements_per_request ; i++ ) { const GLint index = (GLint) *(ui_ptr++); @@ -799,7 +800,7 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, break; } case GL_UNSIGNED_SHORT: { - const GLushort * us_ptr = (const GLushort *) indices; + const GLushort * us_ptr = (const GLushort *) indices + req_element; for ( i = 0 ; i < elements_per_request ; i++ ) { const GLint index = (GLint) *(us_ptr++); @@ -808,7 +809,7 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, break; } case GL_UNSIGNED_BYTE: { - const GLubyte * ub_ptr = (const GLubyte *) indices; + const GLubyte * ub_ptr = (const GLubyte *) indices + req_element; for ( i = 0 ; i < elements_per_request ; i++ ) { const GLint index = (GLint) *(ub_ptr++); @@ -826,6 +827,7 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, } count -= elements_per_request; + req_element += elements_per_request; } diff --git a/src/glx/x11/singlepix.c b/src/glx/x11/singlepix.c index cd88684f70..144e5df743 100644 --- a/src/glx/x11/singlepix.c +++ b/src/glx/x11/singlepix.c @@ -117,7 +117,7 @@ void NAME(_gloffset_GetSeparableFilter)(GLenum target, GLenum format, GLenum typ { __GLXcontext * const gc = __glXGetCurrentContext(); - if (gc->isDirect) { + if (gc->driContext) { CALL_GetSeparableFilter(GET_DISPATCH(), (target, format, type, row, column, span)); return; diff --git a/src/glx/x11/xf86dri.h b/src/glx/x11/xf86dri.h index c8c878f127..a6a57c3135 100644 --- a/src/glx/x11/xf86dri.h +++ b/src/glx/x11/xf86dri.h @@ -64,8 +64,6 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #ifndef _XF86DRI_SERVER_ -#include - _XFUNCPROTOBEGIN Bool XF86DRIQueryExtension( Display *dpy, int *event_base, int *error_base ); @@ -93,14 +91,14 @@ Bool XF86DRICreateContext( Display *dpy, int screen, Visual *visual, Bool XF86DRICreateContextWithConfig( Display *dpy, int screen, int configID, XID *ptr_to_returned_context_id, drm_context_t *hHWContext ); -extern GLboolean XF86DRIDestroyContext( __DRInativeDisplay *dpy, int screen, - __DRIid context_id ); +extern GLboolean XF86DRIDestroyContext( Display *dpy, int screen, + XID context_id ); -extern GLboolean XF86DRICreateDrawable( __DRInativeDisplay *dpy, int screen, - __DRIid drawable, drm_drawable_t *hHWDrawable ); +extern GLboolean XF86DRICreateDrawable( Display *dpy, int screen, + XID drawable, drm_drawable_t *hHWDrawable ); -extern GLboolean XF86DRIDestroyDrawable( __DRInativeDisplay *dpy, int screen, - __DRIid drawable); +extern GLboolean XF86DRIDestroyDrawable( Display *dpy, int screen, + XID drawable); Bool XF86DRIGetDrawableInfo( Display *dpy, int screen, Drawable drawable, unsigned int *index, unsigned int *stamp, diff --git a/src/glx/x11/xfont.c b/src/glx/x11/xfont.c index f3e3da3e79..6ec8c2d6bf 100644 --- a/src/glx/x11/xfont.c +++ b/src/glx/x11/xfont.c @@ -33,6 +33,7 @@ called by that routine when direct rendering is enabled. */ +#ifdef GLX_DIRECT_RENDERING #include "glxclient.h" @@ -208,8 +209,7 @@ static XCharStruct *isvalid(XFontStruct *fs, int which) return(NULL); } - -void DRI_glXUseXFont( Font font, int first, int count, int listbase ) +_X_HIDDEN void DRI_glXUseXFont( Font font, int first, int count, int listbase ) { GLXContext CC; Display *dpy; @@ -373,4 +373,4 @@ bm_height); glPixelStorei(GL_UNPACK_ALIGNMENT, alignment); } -/* The End. */ +#endif diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c index fff79c36ad..2013125efe 100644 --- a/src/mesa/drivers/dri/common/dri_util.c +++ b/src/mesa/drivers/dri/common/dri_util.c @@ -55,39 +55,19 @@ #include "dri_util.h" #include "drm_sarea.h" +#include "utils.h" #ifndef GLX_OML_sync_control -typedef GLboolean ( * PFNGLXGETMSCRATEOMLPROC) (__DRInativeDisplay *dpy, __DRIid drawable, int32_t *numerator, int32_t *denominator); +typedef GLboolean ( * PFNGLXGETMSCRATEOMLPROC) (__DRIdrawable *drawable, int32_t *numerator, int32_t *denominator); #endif -/* This pointer *must* be set by the driver's __driCreateNewScreen funciton! - */ -const __DRIinterfaceMethods * dri_interface = NULL; - /** - * This is used in a couple of places that call \c driCreateNewDrawable. + * This is just a token extension used to signal that the driver + * supports setting a read drawable. */ -static const int empty_attribute_list[1] = { None }; - - -/** - * Cached copy of the internal API version used by libGL and the client-side - * DRI driver. - */ -static int api_ver = 0; - -/* forward declarations */ -static int driQueryFrameTracking( __DRInativeDisplay *dpy, void *priv, - int64_t *sbc, int64_t *missedFrames, - float *lastMissedUsage, float *usage ); - -static void *driCreateNewDrawable(__DRInativeDisplay *dpy, - const __GLcontextModes *modes, - __DRIid draw, __DRIdrawable *pdraw, - int renderType, const int *attrs); - -static void driDestroyDrawable(__DRInativeDisplay *dpy, void *drawablePrivate); - +const __DRIextension driReadDrawableExtension = { + __DRI_READ_DRAWABLE, __DRI_READ_DRAWABLE_VERSION +}; /** * Print message to \c stderr if the \c LIBGL_DEBUG environment variable @@ -111,64 +91,19 @@ __driUtilMessage(const char *f, ...) } } - -/*****************************************************************/ -/** \name Drawable list management */ -/*****************************************************************/ -/*@{*/ - -static GLboolean __driAddDrawable(void *drawHash, __DRIdrawable *pdraw) +GLint +driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ) { - __DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *)pdraw->private; + if (rect2.x1 > rect1.x1) rect1.x1 = rect2.x1; + if (rect2.x2 < rect1.x2) rect1.x2 = rect2.x2; + if (rect2.y1 > rect1.y1) rect1.y1 = rect2.y1; + if (rect2.y2 < rect1.y2) rect1.y2 = rect2.y2; - if (drmHashInsert(drawHash, pdp->draw, pdraw)) - return GL_FALSE; + if (rect1.x1 > rect1.x2 || rect1.y1 > rect1.y2) return 0; - return GL_TRUE; -} - -static __DRIdrawable *__driFindDrawable(void *drawHash, __DRIid draw) -{ - int retcode; - __DRIdrawable *pdraw; - - retcode = drmHashLookup(drawHash, draw, (void *)&pdraw); - if (retcode) - return NULL; - - return pdraw; -} - - -/** - * Find drawables in the local hash that have been destroyed on the - * server. - * - * \param drawHash Hash-table containing all known drawables. - */ -static void __driGarbageCollectDrawables(void *drawHash) -{ - __DRIid draw; - __DRInativeDisplay *dpy; - __DRIdrawable *pdraw; - - if (drmHashFirst(drawHash, &draw, (void *)&pdraw) == 1) { - do { - __DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *)pdraw->private; - dpy = pdp->driScreenPriv->display; - if (! (*dri_interface->windowExists)(dpy, draw)) { - /* Destroy the local drawable data, if the drawable no - longer exists in the Xserver */ - (*pdraw->destroyDrawable)(dpy, pdraw->private); - _mesa_free(pdraw); - } - } while (drmHashNext(drawHash, &draw, (void *)&pdraw) == 1); - } + return (rect1.x2 - rect1.x1) * (rect1.y2 - rect1.y1); } -/*@}*/ - - /*****************************************************************/ /** \name Context (un)binding functions */ /*****************************************************************/ @@ -177,10 +112,7 @@ static void __driGarbageCollectDrawables(void *drawHash) /** * Unbind context. * - * \param dpy the display handle. - * \param scrn the screen number. - * \param draw drawable. - * \param read Current reading drawable. + * \param scrn the screen. * \param gc context. * * \return \c GL_TRUE on success, or \c GL_FALSE on failure. @@ -193,56 +125,27 @@ static void __driGarbageCollectDrawables(void *drawHash) * While casting the opaque private pointers associated with the parameters * into their respective real types it also assures they are not \c NULL. */ -static GLboolean driUnbindContext(__DRInativeDisplay *dpy, int scrn, - __DRIid draw, __DRIid read, - __DRIcontext *ctx) +static int driUnbindContext(__DRIcontext *pcp) { - __DRIscreen *pDRIScreen; - __DRIdrawable *pdraw; - __DRIdrawable *pread; - __DRIcontextPrivate *pcp; - __DRIscreenPrivate *psp; - __DRIdrawablePrivate *pdp; - __DRIdrawablePrivate *prp; + __DRIscreen *psp; + __DRIdrawable *pdp; + __DRIdrawable *prp; /* ** Assume error checking is done properly in glXMakeCurrent before ** calling driUnbindContext. */ - if (ctx == NULL || draw == None || read == None) { - /* ERROR!!! */ - return GL_FALSE; - } - - pDRIScreen = (*dri_interface->getScreen)(dpy, scrn); - if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) { - /* ERROR!!! */ - return GL_FALSE; - } - - psp = (__DRIscreenPrivate *)pDRIScreen->private; - pcp = (__DRIcontextPrivate *)ctx->private; - - pdraw = __driFindDrawable(psp->drawHash, draw); - if (!pdraw) { - /* ERROR!!! */ - return GL_FALSE; - } - pdp = (__DRIdrawablePrivate *)pdraw->private; - - pread = __driFindDrawable(psp->drawHash, read); - if (!pread) { - /* ERROR!!! */ - return GL_FALSE; - } - prp = (__DRIdrawablePrivate *)pread->private; + if (pcp == NULL) + return GL_FALSE; + psp = pcp->driScreenPriv; + pdp = pcp->driDrawablePriv; + prp = pcp->driReadablePriv; /* Let driver unbind drawable from context */ (*psp->DriverAPI.UnbindContext)(pcp); - if (pdp->refcount == 0) { /* ERROR!!! */ return GL_FALSE; @@ -259,12 +162,6 @@ static GLboolean driUnbindContext(__DRInativeDisplay *dpy, int scrn, prp->refcount--; } - /* destroy the drawables if they no longer exist on the server */ - if ((pdp->refcount == 0) || (prp->refcount == 0)) { - /* probably shouldn't need the collector here, - as we know the affected drawables (or could there be others?) */ - __driGarbageCollectDrawables(pdp->driScreenPriv->drawHash); - } /* XXX this is disabled so that if we call SwapBuffers on an unbound * window we can determine the last context bound to the window and @@ -284,72 +181,20 @@ static GLboolean driUnbindContext(__DRInativeDisplay *dpy, int scrn, * This function takes both a read buffer and a draw buffer. This is needed * for \c glXMakeCurrentReadSGI or GLX 1.3's \c glXMakeContextCurrent * function. - * - * \bug This function calls \c driCreateNewDrawable in two places with the - * \c renderType hard-coded to \c GLX_WINDOW_BIT. Some checking might - * be needed in those places when support for pbuffers and / or pixmaps - * is added. Is it safe to assume that the drawable is a window? */ -static GLboolean DoBindContext(__DRInativeDisplay *dpy, - __DRIid draw, __DRIid read, - __DRIcontext *ctx, const __GLcontextModes * modes, - __DRIscreenPrivate *psp) +static int driBindContext(__DRIcontext *pcp, + __DRIdrawable *pdp, + __DRIdrawable *prp) { - __DRIdrawable *pdraw; - __DRIdrawablePrivate *pdp; - __DRIdrawable *pread; - __DRIdrawablePrivate *prp; - __DRIcontextPrivate * const pcp = ctx->private; - - - /* Find the _DRIdrawable which corresponds to the writing drawable. */ - pdraw = __driFindDrawable(psp->drawHash, draw); - if (!pdraw) { - /* Allocate a new drawable */ - pdraw = (__DRIdrawable *)_mesa_malloc(sizeof(__DRIdrawable)); - if (!pdraw) { - /* ERROR!!! */ - return GL_FALSE; - } - - /* Create a new drawable */ - driCreateNewDrawable(dpy, modes, draw, pdraw, GLX_WINDOW_BIT, - empty_attribute_list); - if (!pdraw->private) { - /* ERROR!!! */ - _mesa_free(pdraw); - return GL_FALSE; - } + __DRIscreenPrivate *psp = pcp->driScreenPriv; - } - pdp = (__DRIdrawablePrivate *) pdraw->private; + /* + ** Assume error checking is done properly in glXMakeCurrent before + ** calling driBindContext. + */ - /* Find the _DRIdrawable which corresponds to the reading drawable. */ - if (read == draw) { - /* read buffer == draw buffer */ - prp = pdp; - } - else { - pread = __driFindDrawable(psp->drawHash, read); - if (!pread) { - /* Allocate a new drawable */ - pread = (__DRIdrawable *)_mesa_malloc(sizeof(__DRIdrawable)); - if (!pread) { - /* ERROR!!! */ - return GL_FALSE; - } - - /* Create a new drawable */ - driCreateNewDrawable(dpy, modes, read, pread, GLX_WINDOW_BIT, - empty_attribute_list); - if (!pread->private) { - /* ERROR!!! */ - _mesa_free(pread); - return GL_FALSE; - } - } - prp = (__DRIdrawablePrivate *) pread->private; - } + if (pcp == NULL || pdp == None || prp == None) + return GL_FALSE; /* Bind the drawable to the context */ pcp->driDrawablePriv = pdp; @@ -364,16 +209,22 @@ static GLboolean DoBindContext(__DRInativeDisplay *dpy, ** Now that we have a context associated with this drawable, we can ** initialize the drawable information if has not been done before. */ - if (!pdp->pStamp || *pdp->pStamp != pdp->lastStamp) { - DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); - __driUtilUpdateDrawableInfo(pdp); - DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); - } - if ((pdp != prp) && (!prp->pStamp || *prp->pStamp != prp->lastStamp)) { - DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); - __driUtilUpdateDrawableInfo(prp); - DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); + if (psp->dri2.enabled) { + __driParseEvents(pcp, pdp); + __driParseEvents(pcp, prp); + } else { + if (!pdp->pStamp || *pdp->pStamp != pdp->lastStamp) { + DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); + __driUtilUpdateDrawableInfo(pdp); + DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); + } + + if ((pdp != prp) && (!prp->pStamp || *prp->pStamp != prp->lastStamp)) { + DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); + __driUtilUpdateDrawableInfo(prp); + DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); + } } /* Call device-specific MakeCurrent */ @@ -382,37 +233,6 @@ static GLboolean DoBindContext(__DRInativeDisplay *dpy, return GL_TRUE; } - -/** - * This function takes both a read buffer and a draw buffer. This is needed - * for \c glXMakeCurrentReadSGI or GLX 1.3's \c glXMakeContextCurrent - * function. - */ -static GLboolean driBindContext(__DRInativeDisplay *dpy, int scrn, - __DRIid draw, __DRIid read, - __DRIcontext * ctx) -{ - __DRIscreen *pDRIScreen; - - /* - ** Assume error checking is done properly in glXMakeCurrent before - ** calling driBindContext. - */ - - if (ctx == NULL || draw == None || read == None) { - /* ERROR!!! */ - return GL_FALSE; - } - - pDRIScreen = (*dri_interface->getScreen)(dpy, scrn); - if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) { - /* ERROR!!! */ - return GL_FALSE; - } - - return DoBindContext( dpy, draw, read, ctx, ctx->mode, - (__DRIscreenPrivate *)pDRIScreen->private ); -} /*@}*/ @@ -436,7 +256,7 @@ static GLboolean driBindContext(__DRInativeDisplay *dpy, int scrn, void __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) { - __DRIscreenPrivate *psp; + __DRIscreenPrivate *psp = pdp->driScreenPriv; __DRIcontextPrivate *pcp = pdp->driContextPriv; if (!pcp @@ -447,15 +267,6 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) */ } - psp = pdp->driScreenPriv; - if (!psp) { - /* ERROR!!! */ - _mesa_problem(NULL, "Warning! Possible infinite loop due to bug " - "in file %s, line %d\n", - __FILE__, __LINE__); - return; - } - if (pdp->pClipRects) { _mesa_free(pdp->pClipRects); pdp->pClipRects = NULL; @@ -468,15 +279,15 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); - if (!__driFindDrawable(psp->drawHash, pdp->draw) || - ! (*dri_interface->getDrawableInfo)(pdp->display, pdp->screen, pdp->draw, + if (! (*psp->getDrawableInfo->getDrawableInfo)(pdp, &pdp->index, &pdp->lastStamp, &pdp->x, &pdp->y, &pdp->w, &pdp->h, &pdp->numClipRects, &pdp->pClipRects, &pdp->backX, &pdp->backY, &pdp->numBackClipRects, - &pdp->pBackClipRects )) { + &pdp->pBackClipRects, + pdp->loaderPrivate)) { /* Error -- eg the window may have been destroyed. Keep going * with no cliprects. */ @@ -493,6 +304,138 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) } +int +__driParseEvents(__DRIcontextPrivate *pcp, __DRIdrawablePrivate *pdp) +{ + __DRIscreenPrivate *psp = pcp->driScreenPriv; + __DRIDrawableConfigEvent *dc, *last_dc; + __DRIBufferAttachEvent *ba, *last_ba; + unsigned int tail, mask, *p, end, total, size, changed; + unsigned char *data; + size_t rect_size; + + /* Check for wraparound. */ + if (psp->dri2.buffer->prealloc - pdp->dri2.tail > psp->dri2.buffer->size) { + /* If prealloc overlaps into what we just parsed, the + * server overwrote it and we have to reset our tail + * pointer. */ + DRM_UNLOCK(psp->fd, psp->lock, pcp->hHWContext); + (*psp->dri2.loader->reemitDrawableInfo)(pdp, &pdp->dri2.tail, + pdp->loaderPrivate); + DRM_LIGHT_LOCK(psp->fd, psp->lock, pcp->hHWContext); + } + + total = psp->dri2.buffer->head - pdp->dri2.tail; + mask = psp->dri2.buffer->size - 1; + end = psp->dri2.buffer->head; + data = psp->dri2.buffer->data; + + changed = 0; + last_dc = NULL; + last_ba = NULL; + + for (tail = pdp->dri2.tail; tail != end; tail += size) { + p = (unsigned int *) (data + (tail & mask)); + size = DRI2_EVENT_SIZE(*p); + if (size > total || (tail & mask) + size > psp->dri2.buffer->size) { + /* illegal data, bail out. */ + fprintf(stderr, "illegal event size\n"); + break; + } + + switch (DRI2_EVENT_TYPE(*p)) { + case DRI2_EVENT_DRAWABLE_CONFIG: + dc = (__DRIDrawableConfigEvent *) p; + if (dc->drawable == pdp->dri2.drawable_id) + last_dc = dc; + break; + + case DRI2_EVENT_BUFFER_ATTACH: + ba = (__DRIBufferAttachEvent *) p; + if (ba->drawable == pdp->dri2.drawable_id && + ba->buffer.attachment == DRI_DRAWABLE_BUFFER_FRONT_LEFT) + last_ba = ba; + break; + } + } + + if (last_dc) { + if (pdp->w != last_dc->width || pdp->h != last_dc->height) + changed = 1; + + pdp->x = last_dc->x; + pdp->y = last_dc->y; + pdp->w = last_dc->width; + pdp->h = last_dc->height; + + pdp->backX = 0; + pdp->backY = 0; + pdp->numBackClipRects = 1; + pdp->pBackClipRects[0].x1 = 0; + pdp->pBackClipRects[0].y1 = 0; + pdp->pBackClipRects[0].x2 = pdp->w; + pdp->pBackClipRects[0].y2 = pdp->h; + + pdp->numClipRects = last_dc->num_rects; + _mesa_free(pdp->pClipRects); + rect_size = last_dc->num_rects * sizeof last_dc->rects[0]; + pdp->pClipRects = _mesa_malloc(rect_size); + memcpy(pdp->pClipRects, last_dc->rects, rect_size); + } + + /* We only care about the most recent drawable config. */ + if (last_dc && changed) + (*psp->DriverAPI.HandleDrawableConfig)(pdp, pcp, last_dc); + + /* Front buffer attachments are special, they typically mean that + * we're rendering to a redirected window (or a child window of a + * redirected window) and that it got resized. Resizing the root + * window on randr events is a special case of this. Other causes + * may be a window transitioning between redirected and + * non-redirected, or a window getting reparented between parents + * with different window pixmaps (eg two redirected windows). + * These events are special in that the X server allocates the + * buffer and that the buffer may be shared by other child + * windows. When our window share the window pixmap with its + * parent, drawable config events doesn't affect the front buffer. + * We only care about the last such event in the buffer; in fact, + * older events will refer to invalid buffer objects.*/ + if (last_ba) + (*psp->DriverAPI.HandleBufferAttach)(pdp, pcp, last_ba); + + /* If there was a drawable config event in the buffer and it + * changed the size of the window, all buffer auxillary buffer + * attachments prior to that are invalid (as opposed to the front + * buffer case discussed above). In that case we can start + * looking for buffer attachment after the last drawable config + * event. If there is no drawable config event in this batch of + * events, we have to assume that the last batch might have had + * one and process all buffer attach events.*/ + if (last_dc && changed) + tail = (unsigned char *) last_dc - data; + else + tail = pdp->dri2.tail; + + for ( ; tail != end; tail += size) { + ba = (__DRIBufferAttachEvent *) (data + (tail & mask)); + size = DRI2_EVENT_SIZE(ba->event_header); + + if (DRI2_EVENT_TYPE(ba->event_header) != DRI2_EVENT_BUFFER_ATTACH) + continue; + if (ba->drawable != pdp->dri2.drawable_id) + continue; + if (last_ba == ba) + continue; + + (*psp->DriverAPI.HandleBufferAttach)(pdp, pcp, ba); + changed = 1; + } + + pdp->dri2.tail = tail; + + return changed || last_ba; +} + /*@}*/ /*****************************************************************/ @@ -500,10 +443,33 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) /*****************************************************************/ /*@{*/ +static void driReportDamage(__DRIdrawable *pdp, + struct drm_clip_rect *pClipRects, int numClipRects) +{ + __DRIscreen *psp = pdp->driScreenPriv; + + /* Check that we actually have the new damage report method */ + if (psp->dri2.enabled) { + (*psp->dri2.loader->postDamage)(pdp, + pClipRects, + numClipRects, + pdp->loaderPrivate); + } else if (psp->damage) { + /* Report the damage. Currently, all our drivers draw + * directly to the front buffer, so we report the damage there + * rather than to the backing storein (if any). + */ + (*psp->damage->reportDamage)(pdp, + pdp->x, pdp->y, + pClipRects, numClipRects, + GL_TRUE, pdp->loaderPrivate); + } +} + + /** * Swap buffers. * - * \param dpy the display handle. * \param drawablePrivate opaque pointer to the per-drawable private info. * * \internal @@ -511,74 +477,28 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) * * Is called directly from glXSwapBuffers(). */ -static void driSwapBuffers( __DRInativeDisplay *dpy, void *drawablePrivate ) -{ - __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate; - drm_clip_rect_t rect; - - dPriv->swapBuffers(dPriv); - - /* Check that we actually have the new damage report method */ - if (api_ver < 20070105 || dri_interface->reportDamage == NULL) - return; - - /* Assume it's affecting the whole drawable for now */ - rect.x1 = 0; - rect.y1 = 0; - rect.x2 = rect.x1 + dPriv->w; - rect.y2 = rect.y1 + dPriv->h; - - /* Report the damage. Currently, all our drivers draw directly to the - * front buffer, so we report the damage there rather than to the backing - * store (if any). - */ - (*dri_interface->reportDamage)(dpy, dPriv->screen, dPriv->draw, - dPriv->x, dPriv->y, - &rect, 1, GL_TRUE); -} - -/** - * Called directly from a number of higher-level GLX functions. - */ -static int driGetMSC( void *screenPrivate, int64_t *msc ) -{ - __DRIscreenPrivate *sPriv = (__DRIscreenPrivate *) screenPrivate; - - return sPriv->DriverAPI.GetMSC( sPriv, msc ); -} - -/** - * Called directly from a number of higher-level GLX functions. - */ -static int driGetSBC( __DRInativeDisplay *dpy, void *drawablePrivate, int64_t *sbc ) +static void driSwapBuffers(__DRIdrawable *dPriv) { - __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate; - __DRIswapInfo sInfo; - int status; + __DRIscreen *psp = dPriv->driScreenPriv; + if (!dPriv->numClipRects) + return; - status = dPriv->driScreenPriv->DriverAPI.GetSwapInfo( dPriv, & sInfo ); - *sbc = sInfo.swap_count; + psp->DriverAPI.SwapBuffers(dPriv); - return status; + driReportDamage(dPriv, dPriv->pClipRects, dPriv->numClipRects); } -static int driWaitForSBC( __DRInativeDisplay * dpy, void *drawablePriv, - int64_t target_sbc, - int64_t * msc, int64_t * sbc ) +static int driDrawableGetMSC( __DRIscreen *sPriv, __DRIdrawable *dPriv, + int64_t *msc ) { - __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePriv; - - return dPriv->driScreenPriv->DriverAPI.WaitForSBC( dPriv, target_sbc, - msc, sbc ); + return sPriv->DriverAPI.GetDrawableMSC(sPriv, dPriv, msc); } -static int driWaitForMSC( __DRInativeDisplay * dpy, void *drawablePriv, - int64_t target_msc, - int64_t divisor, int64_t remainder, - int64_t * msc, int64_t * sbc ) +static int driWaitForMSC(__DRIdrawable *dPriv, int64_t target_msc, + int64_t divisor, int64_t remainder, + int64_t * msc, int64_t * sbc) { - __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePriv; __DRIswapInfo sInfo; int status; @@ -600,63 +520,70 @@ static int driWaitForMSC( __DRInativeDisplay * dpy, void *drawablePriv, return status; } -static int64_t driSwapBuffersMSC( __DRInativeDisplay * dpy, void *drawablePriv, - int64_t target_msc, - int64_t divisor, int64_t remainder ) +const __DRImediaStreamCounterExtension driMediaStreamCounterExtension = { + { __DRI_MEDIA_STREAM_COUNTER, __DRI_MEDIA_STREAM_COUNTER_VERSION }, + driWaitForMSC, + driDrawableGetMSC, +}; + +static void driCopySubBuffer(__DRIdrawable *dPriv, + int x, int y, int w, int h) { - __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePriv; + drm_clip_rect_t rect; - return dPriv->driScreenPriv->DriverAPI.SwapBuffersMSC( dPriv, target_msc, - divisor, - remainder ); + dPriv->driScreenPriv->DriverAPI.CopySubBuffer(dPriv, x, y, w, h); + + rect.x1 = x; + rect.y1 = dPriv->h - y - h; + rect.x2 = x + w; + rect.y2 = rect.y1 + h; + driReportDamage(dPriv, &rect, 1); } -static void driCopySubBuffer( __DRInativeDisplay *dpy, void *drawablePrivate, - int x, int y, int w, int h) +const __DRIcopySubBufferExtension driCopySubBufferExtension = { + { __DRI_COPY_SUB_BUFFER, __DRI_COPY_SUB_BUFFER_VERSION }, + driCopySubBuffer +}; + +static void driSetSwapInterval(__DRIdrawable *dPriv, unsigned int interval) { - __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate; - dPriv->driScreenPriv->DriverAPI.CopySubBuffer(dPriv, x, y, w, h); - (void) dpy; + dPriv->swap_interval = interval; } +static unsigned int driGetSwapInterval(__DRIdrawable *dPriv) +{ + return dPriv->swap_interval; +} + +const __DRIswapControlExtension driSwapControlExtension = { + { __DRI_SWAP_CONTROL, __DRI_SWAP_CONTROL_VERSION }, + driSetSwapInterval, + driGetSwapInterval +}; + + /** * This is called via __DRIscreenRec's createNewDrawable pointer. */ -static void *driCreateNewDrawable(__DRInativeDisplay *dpy, - const __GLcontextModes *modes, - __DRIid draw, - __DRIdrawable *pdraw, - int renderType, - const int *attrs) +static __DRIdrawable * +driCreateNewDrawable(__DRIscreen *psp, const __DRIconfig *config, + drm_drawable_t hwDrawable, int renderType, + const int *attrs, void *data) { - __DRIscreen * const pDRIScreen = (*dri_interface->getScreen)(dpy, modes->screen); - __DRIscreenPrivate *psp; - __DRIdrawablePrivate *pdp; - - - pdraw->private = NULL; + __DRIdrawable *pdp; /* Since pbuffers are not yet supported, no drawable attributes are * supported either. */ (void) attrs; - if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) { - return NULL; - } - - pdp = (__DRIdrawablePrivate *)_mesa_malloc(sizeof(__DRIdrawablePrivate)); + pdp = _mesa_malloc(sizeof *pdp); if (!pdp) { return NULL; } - if (!(*dri_interface->createDrawable)(dpy, modes->screen, draw, &pdp->hHWDrawable)) { - _mesa_free(pdp); - return NULL; - } - - pdp->draw = draw; - pdp->pdraw = pdraw; + pdp->loaderPrivate = data; + pdp->hHWDrawable = hwDrawable; pdp->refcount = 0; pdp->pStamp = NULL; pdp->lastStamp = 0; @@ -669,80 +596,55 @@ static void *driCreateNewDrawable(__DRInativeDisplay *dpy, pdp->numBackClipRects = 0; pdp->pClipRects = NULL; pdp->pBackClipRects = NULL; - pdp->display = dpy; - pdp->screen = modes->screen; + pdp->vblSeq = 0; + pdp->vblFlags = 0; - psp = (__DRIscreenPrivate *)pDRIScreen->private; pdp->driScreenPriv = psp; pdp->driContextPriv = &psp->dummyContextPriv; - if (!(*psp->DriverAPI.CreateBuffer)(psp, pdp, modes, + if (!(*psp->DriverAPI.CreateBuffer)(psp, pdp, &config->modes, renderType == GLX_PIXMAP_BIT)) { - (void)(*dri_interface->destroyDrawable)(dpy, modes->screen, pdp->draw); _mesa_free(pdp); return NULL; } - pdraw->private = pdp; - pdraw->destroyDrawable = driDestroyDrawable; - pdraw->swapBuffers = driSwapBuffers; /* called by glXSwapBuffers() */ - - pdraw->getSBC = driGetSBC; - pdraw->waitForSBC = driWaitForSBC; - pdraw->waitForMSC = driWaitForMSC; - pdraw->swapBuffersMSC = driSwapBuffersMSC; - pdraw->frameTracking = NULL; - pdraw->queryFrameTracking = driQueryFrameTracking; - - if (driCompareGLXAPIVersion (20060314) >= 0) - pdraw->copySubBuffer = driCopySubBuffer; + pdp->msc_base = 0; /* This special default value is replaced with the configured * default value when the drawable is first bound to a direct * rendering context. */ - pdraw->swap_interval = (unsigned)-1; - - pdp->swapBuffers = psp->DriverAPI.SwapBuffers; + pdp->swap_interval = (unsigned)-1; - /* Add pdraw to drawable list */ - if (!__driAddDrawable(psp->drawHash, pdraw)) { - /* ERROR!!! */ - (*pdraw->destroyDrawable)(dpy, pdp); - _mesa_free(pdp); - pdp = NULL; - pdraw->private = NULL; - } - - return (void *) pdp; + return pdp; } static __DRIdrawable * -driGetDrawable(__DRInativeDisplay *dpy, __DRIid draw, void *screenPrivate) +dri2CreateNewDrawable(__DRIscreen *screen, const __DRIconfig *config, + unsigned int drawable_id, unsigned int head, void *data) { - __DRIscreenPrivate *psp = (__DRIscreenPrivate *) screenPrivate; + __DRIdrawable *pdraw; - /* - ** Make sure this routine returns NULL if the drawable is not bound - ** to a direct rendering context! - */ - return __driFindDrawable(psp->drawHash, draw); + pdraw = driCreateNewDrawable(screen, config, 0, 0, NULL, data); + if (!pdraw) + return NULL; + + pdraw->dri2.drawable_id = drawable_id; + pdraw->dri2.tail = head; + pdraw->pBackClipRects = _mesa_malloc(sizeof *pdraw->pBackClipRects); + + return pdraw; } + static void -driDestroyDrawable(__DRInativeDisplay *dpy, void *drawablePrivate) +driDestroyDrawable(__DRIdrawable *pdp) { - __DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *) drawablePrivate; __DRIscreenPrivate *psp; - int scrn; if (pdp) { psp = pdp->driScreenPriv; - scrn = psp->myNum; (*psp->DriverAPI.DestroyBuffer)(pdp); - if ((*dri_interface->windowExists)(dpy, pdp->draw)) - (void)(*dri_interface->destroyDrawable)(dpy, scrn, pdp->draw); - drmHashDelete(psp->drawHash, pdp->draw); if (pdp->pClipRects) { _mesa_free(pdp->pClipRects); pdp->pClipRects = NULL; @@ -766,8 +668,6 @@ driDestroyDrawable(__DRInativeDisplay *dpy, void *drawablePrivate) /** * Destroy the per-context private information. * - * \param dpy the display handle. - * \param scrn the screen number. * \param contextPrivate opaque pointer to the per-drawable private info. * * \internal @@ -775,14 +675,10 @@ driDestroyDrawable(__DRInativeDisplay *dpy, void *drawablePrivate) * drmDestroyContext(), and finally frees \p contextPrivate. */ static void -driDestroyContext(__DRInativeDisplay *dpy, int scrn, void *contextPrivate) +driDestroyContext(__DRIcontext *pcp) { - __DRIcontextPrivate *pcp = (__DRIcontextPrivate *) contextPrivate; - if (pcp) { (*pcp->driScreenPriv->DriverAPI.DestroyContext)(pcp); - __driGarbageCollectDrawables(pcp->driScreenPriv->drawHash); - (void) (*dri_interface->destroyContext)(dpy, scrn, pcp->contextID); _mesa_free(pcp); } } @@ -795,7 +691,7 @@ driDestroyContext(__DRInativeDisplay *dpy, int scrn, void *contextPrivate) * \param modes Mode used to create the new context. * \param render_type Type of rendering target. \c GLX_RGBA is the only * type likely to ever be supported for direct-rendering. - * \param sharedPrivate The shared context dependent methods or \c NULL if + * \param shared The shared context dependent methods or \c NULL if * non-existent. * \param pctx DRI context to receive the context dependent methods. * @@ -809,36 +705,18 @@ driDestroyContext(__DRInativeDisplay *dpy, int scrn, void *contextPrivate) * context. * */ -static void * -driCreateNewContext(__DRInativeDisplay *dpy, const __GLcontextModes *modes, - int render_type, void *sharedPrivate, __DRIcontext *pctx) +static __DRIcontext * +driCreateNewContext(__DRIscreen *psp, const __DRIconfig *config, + int render_type, __DRIcontext *shared, + drm_context_t hwContext, void *data) { - __DRIscreen *pDRIScreen; - __DRIcontextPrivate *pcp; - __DRIcontextPrivate *pshare = (__DRIcontextPrivate *) sharedPrivate; - __DRIscreenPrivate *psp; - void * const shareCtx = (pshare != NULL) ? pshare->driverPrivate : NULL; + __DRIcontext *pcp; + void * const shareCtx = (shared != NULL) ? shared->driverPrivate : NULL; - pDRIScreen = (*dri_interface->getScreen)(dpy, modes->screen); - if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) { - /* ERROR!!! */ + pcp = _mesa_malloc(sizeof *pcp); + if (!pcp) return NULL; - } - - psp = (__DRIscreenPrivate *)pDRIScreen->private; - pcp = (__DRIcontextPrivate *)_mesa_malloc(sizeof(__DRIcontextPrivate)); - if (!pcp) { - return NULL; - } - - if (! (*dri_interface->createContext)(dpy, modes->screen, modes->fbconfigID, - &pcp->contextID, &pcp->hHWContext)) { - _mesa_free(pcp); - return NULL; - } - - pcp->display = dpy; pcp->driScreenPriv = psp; pcp->driDrawablePriv = NULL; @@ -846,8 +724,7 @@ driCreateNewContext(__DRInativeDisplay *dpy, const __GLcontextModes *modes, * context. */ - if (!psp->dummyContextPriv.driScreenPriv) { - psp->dummyContextPriv.contextID = 0; + if (!psp->dri2.enabled && !psp->dummyContextPriv.driScreenPriv) { psp->dummyContextPriv.hHWContext = psp->pSAREA->dummy_context; psp->dummyContextPriv.driScreenPriv = psp; psp->dummyContextPriv.driDrawablePriv = NULL; @@ -855,21 +732,40 @@ driCreateNewContext(__DRInativeDisplay *dpy, const __GLcontextModes *modes, /* No other fields should be used! */ } - pctx->destroyContext = driDestroyContext; - pctx->bindContext = driBindContext; - pctx->unbindContext = driUnbindContext; + pcp->hHWContext = hwContext; - if ( !(*psp->DriverAPI.CreateContext)(modes, pcp, shareCtx) ) { - (void) (*dri_interface->destroyContext)(dpy, modes->screen, - pcp->contextID); + if ( !(*psp->DriverAPI.CreateContext)(&config->modes, pcp, shareCtx) ) { _mesa_free(pcp); return NULL; } - __driGarbageCollectDrawables(pcp->driScreenPriv->drawHash); - return pcp; } + +static __DRIcontext * +dri2CreateNewContext(__DRIscreen *screen, const __DRIconfig *config, + __DRIcontext *shared, void *data) +{ + drm_context_t hwContext; + DRM_CAS_RESULT(ret); + + /* DRI2 doesn't use kernel with context IDs, we just need an ID that's + * different from the kernel context ID to make drmLock() happy. */ + + do { + hwContext = screen->dri2.lock->next_id; + DRM_CAS(&screen->dri2.lock->next_id, hwContext, hwContext + 1, ret); + } while (ret); + + return driCreateNewContext(screen, config, 0, shared, hwContext, data); +} + +static int +driCopyContext(__DRIcontext *dest, __DRIcontext *src, unsigned long mask) +{ + return GL_FALSE; +} + /*@}*/ @@ -889,10 +785,8 @@ driCreateNewContext(__DRInativeDisplay *dpy, const __GLcontextModes *modes, * This function calls __DriverAPIRec::DestroyScreen on \p screenPrivate, calls * drmClose(), and finally frees \p screenPrivate. */ -static void driDestroyScreen(__DRInativeDisplay *dpy, int scrn, void *screenPrivate) +static void driDestroyScreen(__DRIscreen *psp) { - __DRIscreenPrivate *psp = (__DRIscreenPrivate *) screenPrivate; - if (psp) { /* No interaction with the X-server is possible at this point. This * routine is called after XCloseDisplay, so there is no protocol @@ -902,26 +796,44 @@ static void driDestroyScreen(__DRInativeDisplay *dpy, int scrn, void *screenPriv if (psp->DriverAPI.DestroyScreen) (*psp->DriverAPI.DestroyScreen)(psp); - (void)drmUnmap((drmAddress)psp->pSAREA, SAREA_MAX); - (void)drmUnmap((drmAddress)psp->pFB, psp->fbSize); - _mesa_free(psp->pDevPriv); - (void)drmCloseOnce(psp->fd); - if ( psp->modes != NULL ) { - (*dri_interface->destroyContextModes)( psp->modes ); + if (psp->dri2.enabled) { + drmBOUnmap(psp->fd, &psp->dri2.sareaBO); + drmBOUnreference(psp->fd, &psp->dri2.sareaBO); + } else { + (void)drmUnmap((drmAddress)psp->pSAREA, SAREA_MAX); + (void)drmUnmap((drmAddress)psp->pFB, psp->fbSize); + (void)drmCloseOnce(psp->fd); } - assert(psp->drawHash); - drmHashDestroy(psp->drawHash); - _mesa_free(psp); } } +static void +setupLoaderExtensions(__DRIscreen *psp, + const __DRIextension **extensions) +{ + int i; + + for (i = 0; extensions[i]; i++) { + if (strcmp(extensions[i]->name, __DRI_GET_DRAWABLE_INFO) == 0) + psp->getDrawableInfo = (__DRIgetDrawableInfoExtension *) extensions[i]; + if (strcmp(extensions[i]->name, __DRI_DAMAGE) == 0) + psp->damage = (__DRIdamageExtension *) extensions[i]; + if (strcmp(extensions[i]->name, __DRI_SYSTEM_TIME) == 0) + psp->systemTime = (__DRIsystemTimeExtension *) extensions[i]; + if (strcmp(extensions[i]->name, __DRI_LOADER) == 0) + psp->dri2.loader = (__DRIloaderExtension *) extensions[i]; + } +} /** - * Utility function used to create a new driver-private screen structure. + * This is the bootstrap function for the driver. libGL supplies all of the + * requisite information about the system, and the driver initializes itself. + * This routine also fills in the linked list pointed to by \c driver_modes + * with the \c __GLcontextModes that the driver can support for windows or + * pbuffers. * - * \param dpy Display pointer * \param scrn Index of the screen * \param psc DRI screen data (not driver private) * \param modes Linked list of known display modes. This list is, at a @@ -942,44 +854,29 @@ static void driDestroyScreen(__DRInativeDisplay *dpy, int scrn, void *screenPriv * driver and libGL. * \param driverAPI Driver API functions used by other routines in dri_util.c. * - * \note - * There is no need to check the minimum API version in this function. Since - * the \c __driCreateNewScreen function is versioned, it is impossible for a - * loader that is too old to even load this driver. + * \note There is no need to check the minimum API version in this + * function. Since the name of this function is versioned, it is + * impossible for a loader that is too old to even load this driver. */ -__DRIscreenPrivate * -__driUtilCreateNewScreen(__DRInativeDisplay *dpy, int scrn, __DRIscreen *psc, - __GLcontextModes * modes, - const __DRIversion * ddx_version, - const __DRIversion * dri_version, - const __DRIversion * drm_version, - const __DRIframebuffer * frame_buffer, - drm_sarea_t *pSAREA, - int fd, - int internal_api_version, - const struct __DriverAPIRec *driverAPI) +static __DRIscreen * +driCreateNewScreen(int scrn, + const __DRIversion *ddx_version, + const __DRIversion *dri_version, + const __DRIversion *drm_version, + const __DRIframebuffer *frame_buffer, + drmAddress pSAREA, int fd, + const __DRIextension **extensions, + const __DRIconfig ***driver_modes, + void *loaderPrivate) { - __DRIscreenPrivate *psp; - - - api_ver = internal_api_version; - - psp = (__DRIscreenPrivate *)_mesa_malloc(sizeof(__DRIscreenPrivate)); - if (!psp) { - return NULL; - } + static const __DRIextension *emptyExtensionList[] = { NULL }; + __DRIscreen *psp; - /* Create the hash table */ - psp->drawHash = drmHashCreate(); - if ( psp->drawHash == NULL ) { - _mesa_free( psp ); + psp = _mesa_malloc(sizeof *psp); + if (!psp) return NULL; - } - psp->display = dpy; - psp->myNum = scrn; - psp->psc = psc; - psp->modes = modes; + setupLoaderExtensions(psp, extensions); /* ** NOT_DONE: This is used by the X server to detect when the client @@ -988,20 +885,12 @@ __driUtilCreateNewScreen(__DRInativeDisplay *dpy, int scrn, __DRIscreen *psc, */ psp->drawLockID = 1; - psp->drmMajor = drm_version->major; - psp->drmMinor = drm_version->minor; - psp->drmPatch = drm_version->patch; - psp->ddxMajor = ddx_version->major; - psp->ddxMinor = ddx_version->minor; - psp->ddxPatch = ddx_version->patch; - psp->driMajor = dri_version->major; - psp->driMinor = dri_version->minor; - psp->driPatch = dri_version->patch; - - /* install driver's callback functions */ - memcpy( &psp->DriverAPI, driverAPI, sizeof(struct __DriverAPIRec) ); + psp->drm_version = *drm_version; + psp->ddx_version = *ddx_version; + psp->dri_version = *dri_version; psp->pSAREA = pSAREA; + psp->lock = (drmLock *) &psp->pSAREA->lock; psp->pFB = frame_buffer->base; psp->fbSize = frame_buffer->size; @@ -1012,7 +901,10 @@ __driUtilCreateNewScreen(__DRInativeDisplay *dpy, int scrn, __DRIscreen *psc, psp->pDevPriv = frame_buffer->dev_priv; psp->fbBPP = psp->fbStride * 8 / frame_buffer->width; + psp->extensions = emptyExtensionList; psp->fd = fd; + psp->myNum = scrn; + psp->dri2.enabled = GL_FALSE; /* ** Do not init dummy context here; actual initialization will be @@ -1021,63 +913,143 @@ __driUtilCreateNewScreen(__DRInativeDisplay *dpy, int scrn, __DRIscreen *psc, */ psp->dummyContextPriv.driScreenPriv = NULL; - psc->destroyScreen = driDestroyScreen; - psc->createNewDrawable = driCreateNewDrawable; - psc->getDrawable = driGetDrawable; - psc->getMSC = driGetMSC; - psc->createNewContext = driCreateNewContext; - - if (internal_api_version >= 20070121) - psc->setTexOffset = psp->DriverAPI.setTexOffset; + psp->DriverAPI = driDriverAPI; - if ( (psp->DriverAPI.InitDriver != NULL) - && !(*psp->DriverAPI.InitDriver)(psp) ) { - _mesa_free( psp ); + *driver_modes = driDriverAPI.InitScreen(psp); + if (*driver_modes == NULL) { + _mesa_free(psp); return NULL; } - return psp; } -/** - * Compare the current GLX API version with a driver supplied required version. - * - * The minimum required version is compared with the API version exported by - * the \c __glXGetInternalVersion function (in libGL.so). - * - * \param required_version Minimum required internal GLX API version. - * \return A tri-value return, as from strcmp is returned. A value less - * than, equal to, or greater than zero will be returned if the - * internal GLX API version is less than, equal to, or greater - * than \c required_version. - * - * \sa __glXGetInternalVersion(). - */ -int driCompareGLXAPIVersion( GLint required_version ) +static __DRIscreen * +dri2CreateNewScreen(int scrn, int fd, unsigned int sarea_handle, + const __DRIextension **extensions, + const __DRIconfig ***driver_configs, void *data) { - if ( api_ver > required_version ) { - return 1; - } - else if ( api_ver == required_version ) { - return 0; - } + static const __DRIextension *emptyExtensionList[] = { NULL }; + __DRIscreen *psp; + unsigned int *p; + drmVersionPtr version; + + if (driDriverAPI.InitScreen2 == NULL) + return NULL; + + psp = _mesa_malloc(sizeof(*psp)); + if (!psp) + return NULL; + + setupLoaderExtensions(psp, extensions); + + version = drmGetVersion(fd); + if (version) { + psp->drm_version.major = version->version_major; + psp->drm_version.minor = version->version_minor; + psp->drm_version.patch = version->version_patchlevel; + drmFreeVersion(version); + } + + psp->extensions = emptyExtensionList; + psp->fd = fd; + psp->myNum = scrn; + psp->dri2.enabled = GL_TRUE; + + if (drmBOReference(psp->fd, sarea_handle, &psp->dri2.sareaBO)) { + fprintf(stderr, "Failed to reference DRI2 sarea BO\n"); + _mesa_free(psp); + return NULL; + } + + if (drmBOMap(psp->fd, &psp->dri2.sareaBO, + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, &psp->dri2.sarea)) { + drmBOUnreference(psp->fd, &psp->dri2.sareaBO); + _mesa_free(psp); + return NULL; + } + + p = psp->dri2.sarea; + while (DRI2_SAREA_BLOCK_TYPE(*p)) { + switch (DRI2_SAREA_BLOCK_TYPE(*p)) { + case DRI2_SAREA_BLOCK_LOCK: + psp->dri2.lock = (__DRILock *) p; + break; + case DRI2_SAREA_BLOCK_EVENT_BUFFER: + psp->dri2.buffer = (__DRIEventBuffer *) p; + break; + } + p = DRI2_SAREA_BLOCK_NEXT(p); + } + + psp->lock = (drmLock *) &psp->dri2.lock->lock; - return -1; + psp->DriverAPI = driDriverAPI; + *driver_configs = driDriverAPI.InitScreen2(psp); + if (*driver_configs == NULL) { + drmBOUnmap(psp->fd, &psp->dri2.sareaBO); + drmBOUnreference(psp->fd, &psp->dri2.sareaBO); + _mesa_free(psp); + return NULL; + } + + psp->DriverAPI = driDriverAPI; + + return psp; } +static const __DRIextension **driGetExtensions(__DRIscreen *psp) +{ + return psp->extensions; +} + +const __DRIlegacyExtension driLegacyExtension = { + { __DRI_LEGACY, __DRI_LEGACY_VERSION }, + driCreateNewScreen, + driCreateNewDrawable, + driCreateNewContext +}; + +const __DRIcoreExtension driCoreExtension = { + { __DRI_CORE, __DRI_CORE_VERSION }, + dri2CreateNewScreen, + driDestroyScreen, + driGetExtensions, + driGetConfigAttrib, + driIndexConfigAttrib, + dri2CreateNewDrawable, + driDestroyDrawable, + driSwapBuffers, + dri2CreateNewContext, + driCopyContext, + driDestroyContext, + driBindContext, + driUnbindContext +}; + +/* This is the table of extensions that the loader will dlsym() for. */ +PUBLIC const __DRIextension *__driDriverExtensions[] = { + &driCoreExtension.base, + &driLegacyExtension.base, + NULL +}; + +static int +driFrameTracking(__DRIdrawable *drawable, GLboolean enable) +{ + return GLX_BAD_CONTEXT; +} static int -driQueryFrameTracking( __DRInativeDisplay * dpy, void * priv, - int64_t * sbc, int64_t * missedFrames, - float * lastMissedUsage, float * usage ) +driQueryFrameTracking(__DRIdrawable *dpriv, + int64_t * sbc, int64_t * missedFrames, + float * lastMissedUsage, float * usage) { __DRIswapInfo sInfo; int status; int64_t ust; - __DRIdrawablePrivate * dpriv = (__DRIdrawablePrivate *) priv; - + __DRIscreenPrivate *psp = dpriv->driScreenPriv; status = dpriv->driScreenPriv->DriverAPI.GetSwapInfo( dpriv, & sInfo ); if ( status == 0 ) { @@ -1085,13 +1057,18 @@ driQueryFrameTracking( __DRInativeDisplay * dpy, void * priv, *missedFrames = sInfo.swap_missed_count; *lastMissedUsage = sInfo.swap_missed_usage; - (*dri_interface->getUST)( & ust ); + (*psp->systemTime->getUST)( & ust ); *usage = driCalculateSwapUsage( dpriv, sInfo.swap_ust, ust ); } return status; } +const __DRIframeTrackingExtension driFrameTrackingExtension = { + { __DRI_FRAME_TRACKING, __DRI_FRAME_TRACKING_VERSION }, + driFrameTracking, + driQueryFrameTracking +}; /** * Calculate amount of swap interval used between GLX buffer swaps. @@ -1129,11 +1106,10 @@ driCalculateSwapUsage( __DRIdrawablePrivate *dPriv, int64_t last_swap_ust, int32_t d; int interval; float usage = 1.0; + __DRIscreenPrivate *psp = dPriv->driScreenPriv; - - if ( (*dri_interface->getMSCRate)( dPriv->display, dPriv->draw, &n, &d ) ) { - interval = (dPriv->pdraw->swap_interval != 0) - ? dPriv->pdraw->swap_interval : 1; + if ( (*psp->systemTime->getMSCRate)(dPriv, &n, &d, dPriv->loaderPrivate) ) { + interval = (dPriv->swap_interval != 0) ? dPriv->swap_interval : 1; /* We want to calculate diff --git a/src/mesa/drivers/dri/common/dri_util.h b/src/mesa/drivers/dri/common/dri_util.h index 027cb7f461..f2bc456307 100644 --- a/src/mesa/drivers/dri/common/dri_util.h +++ b/src/mesa/drivers/dri/common/dri_util.h @@ -1,3 +1,24 @@ +/** + * \file dri_util.h + * DRI utility functions definitions. + * + * This module acts as glue between GLX and the actual hardware driver. A DRI + * driver doesn't really \e have to use any of this - it's optional. But, some + * useful stuff is done here that otherwise would have to be duplicated in most + * drivers. + * + * Basically, these utility functions take care of some of the dirty details of + * screen initialization, context creation, context binding, DRM setup, etc. + * + * These functions are compiled into each DRI driver so libGL.so knows nothing + * about them. + * + * \sa dri_util.c. + * + * \author Kevin E. Martin + * \author Brian Paul + */ + /* * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. * All Rights Reserved. @@ -23,46 +44,37 @@ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -/** - * \file dri_util.h - * DRI utility functions definitions. - * - * This module acts as glue between GLX and the actual hardware driver. A DRI - * driver doesn't really \e have to use any of this - it's optional. But, some - * useful stuff is done here that otherwise would have to be duplicated in most - * drivers. - * - * Basically, these utility functions take care of some of the dirty details of - * screen initialization, context creation, context binding, DRM setup, etc. - * - * These functions are compiled into each DRI driver so libGL.so knows nothing - * about them. - * - * \sa dri_util.c. - * - * \author Kevin E. Martin - * \author Brian Paul - */ #ifndef _DRI_UTIL_H_ #define _DRI_UTIL_H_ #include -#include "drm.h" -#include "drm_sarea.h" -#include "xf86drm.h" +#include +#include +#include #include "GL/internal/glcore.h" #include "GL/internal/dri_interface.h" +#include "GL/internal/dri_sarea.h" #define GLX_BAD_CONTEXT 5 -typedef struct __DRIdisplayPrivateRec __DRIdisplayPrivate; -typedef struct __DRIscreenPrivateRec __DRIscreenPrivate; -typedef struct __DRIcontextPrivateRec __DRIcontextPrivate; -typedef struct __DRIdrawablePrivateRec __DRIdrawablePrivate; typedef struct __DRIswapInfoRec __DRIswapInfo; -typedef struct __DRIutilversionRec2 __DRIutilversion2; +/* Typedefs to avoid rewriting the world. */ +typedef struct __DRIscreenRec __DRIscreenPrivate; +typedef struct __DRIdrawableRec __DRIdrawablePrivate; +typedef struct __DRIcontextRec __DRIcontextPrivate; + +/** + * Extensions. + */ +extern const __DRIlegacyExtension driLegacyExtension; +extern const __DRIcoreExtension driCoreExtension; +extern const __DRIextension driReadDrawableExtension; +extern const __DRIcopySubBufferExtension driCopySubBufferExtension; +extern const __DRIswapControlExtension driSwapControlExtension; +extern const __DRIframeTrackingExtension driFrameTrackingExtension; +extern const __DRImediaStreamCounterExtension driMediaStreamCounterExtension; /** * Used by DRI_VALIDATE_DRAWABLE_INFO @@ -78,7 +90,7 @@ typedef struct __DRIutilversionRec2 __DRIutilversion2; /** * Utility macro to validate the drawable information. * - * See __DRIdrawablePrivate::pStamp and __DRIdrawablePrivate::lastStamp. + * See __DRIdrawable::pStamp and __DRIdrawable::lastStamp. */ #define DRI_VALIDATE_DRAWABLE_INFO(psp, pdp) \ do { \ @@ -107,94 +119,103 @@ do { \ * this structure. */ struct __DriverAPIRec { - /** - * Driver initialization callback - */ - GLboolean (*InitDriver)(__DRIscreenPrivate *driScrnPriv); - + const __DRIconfig **(*InitScreen) (__DRIscreen * priv); + /** * Screen destruction callback */ - void (*DestroyScreen)(__DRIscreenPrivate *driScrnPriv); + void (*DestroyScreen)(__DRIscreen *driScrnPriv); /** * Context creation callback */ GLboolean (*CreateContext)(const __GLcontextModes *glVis, - __DRIcontextPrivate *driContextPriv, + __DRIcontext *driContextPriv, void *sharedContextPrivate); /** * Context destruction callback */ - void (*DestroyContext)(__DRIcontextPrivate *driContextPriv); + void (*DestroyContext)(__DRIcontext *driContextPriv); /** * Buffer (drawable) creation callback */ - GLboolean (*CreateBuffer)(__DRIscreenPrivate *driScrnPriv, - __DRIdrawablePrivate *driDrawPriv, + GLboolean (*CreateBuffer)(__DRIscreen *driScrnPriv, + __DRIdrawable *driDrawPriv, const __GLcontextModes *glVis, GLboolean pixmapBuffer); /** * Buffer (drawable) destruction callback */ - void (*DestroyBuffer)(__DRIdrawablePrivate *driDrawPriv); + void (*DestroyBuffer)(__DRIdrawable *driDrawPriv); /** * Buffer swapping callback */ - void (*SwapBuffers)(__DRIdrawablePrivate *driDrawPriv); + void (*SwapBuffers)(__DRIdrawable *driDrawPriv); /** * Context activation callback */ - GLboolean (*MakeCurrent)(__DRIcontextPrivate *driContextPriv, - __DRIdrawablePrivate *driDrawPriv, - __DRIdrawablePrivate *driReadPriv); + GLboolean (*MakeCurrent)(__DRIcontext *driContextPriv, + __DRIdrawable *driDrawPriv, + __DRIdrawable *driReadPriv); /** * Context unbinding callback */ - GLboolean (*UnbindContext)(__DRIcontextPrivate *driContextPriv); + GLboolean (*UnbindContext)(__DRIcontext *driContextPriv); /** * Retrieves statistics about buffer swap operations. Required if * GLX_OML_sync_control or GLX_MESA_swap_frame_usage is supported. */ - int (*GetSwapInfo)( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo ); - + int (*GetSwapInfo)( __DRIdrawable *dPriv, __DRIswapInfo * sInfo ); - /** - * Required if GLX_SGI_video_sync or GLX_OML_sync_control is - * supported. - */ - int (*GetMSC)( __DRIscreenPrivate * priv, int64_t * count ); /** * These are required if GLX_OML_sync_control is supported. */ /*@{*/ - int (*WaitForMSC)( __DRIdrawablePrivate *priv, int64_t target_msc, + int (*WaitForMSC)( __DRIdrawable *priv, int64_t target_msc, int64_t divisor, int64_t remainder, int64_t * msc ); - int (*WaitForSBC)( __DRIdrawablePrivate *priv, int64_t target_sbc, + int (*WaitForSBC)( __DRIdrawable *priv, int64_t target_sbc, int64_t * msc, int64_t * sbc ); - int64_t (*SwapBuffersMSC)( __DRIdrawablePrivate *priv, int64_t target_msc, + int64_t (*SwapBuffersMSC)( __DRIdrawable *priv, int64_t target_msc, int64_t divisor, int64_t remainder ); /*@}*/ - void (*CopySubBuffer)(__DRIdrawablePrivate *driDrawPriv, + void (*CopySubBuffer)(__DRIdrawable *driDrawPriv, int x, int y, int w, int h); /** - * See corresponding field in \c __DRIscreenRec. + * New version of GetMSC so we can pass drawable data to the low + * level DRM driver (e.g. pipe info). Required if + * GLX_SGI_video_sync or GLX_OML_sync_control is supported. */ - void (*setTexOffset)(__DRIcontext *pDRICtx, GLint texname, - unsigned long long offset, GLint depth, GLuint pitch); + int (*GetDrawableMSC) ( __DRIscreen * priv, + __DRIdrawable *drawablePrivate, + int64_t *count); + + + + /* DRI2 Entry points */ + const __DRIconfig **(*InitScreen2) (__DRIscreen * priv); + void (*HandleDrawableConfig)(__DRIdrawable *dPriv, + __DRIcontext *pcp, + __DRIDrawableConfigEvent *event); + + void (*HandleBufferAttach)(__DRIdrawable *dPriv, + __DRIcontext *pcp, + __DRIBufferAttachEvent *ba); + }; +extern const struct __DriverAPIRec driDriverAPI; + struct __DRIswapInfoRec { /** @@ -230,7 +251,7 @@ struct __DRIswapInfoRec { /** * Per-drawable private DRI driver information. */ -struct __DRIdrawablePrivateRec { +struct __DRIdrawableRec { /** * Kernel drawable handle */ @@ -244,10 +265,10 @@ struct __DRIdrawablePrivateRec { void *driverPrivate; /** - * X's drawable ID associated with this private drawable. + * Private data from the loader. We just hold on to it and pass + * it back when calling into loader provided functions. */ - __DRIid draw; - __DRIdrawable *pdraw; + void *loaderPrivate; /** * Reference count for number of context's currently bound to this @@ -272,7 +293,7 @@ struct __DRIdrawablePrivateRec { /** * Last value of the stamp. * - * If this differs from the value stored at __DRIdrawablePrivate::pStamp, + * If this differs from the value stored at __DRIdrawable::pStamp, * then the drawable information has been modified by the X server, and the * drawable information (below) should be retrieved from the X server. */ @@ -306,41 +327,56 @@ struct __DRIdrawablePrivateRec { /*@}*/ /** - * Pointer to context to which this drawable is currently bound. + * \name Vertical blank tracking information + * Used for waiting on vertical blank events. */ - __DRIcontextPrivate *driContextPriv; + /*@{*/ + unsigned int vblSeq; + unsigned int vblFlags; + /*@}*/ /** - * Pointer to screen on which this drawable was created. + * \name Monotonic MSC tracking + * + * Low level driver is responsible for updating msc_base and + * vblSeq values so that higher level code can calculate + * a new msc value or msc target for a WaitMSC call. The new value + * will be: + * msc = msc_base + get_vblank_count() - vblank_base; + * + * And for waiting on a value, core code will use: + * actual_target = target_msc - msc_base + vblank_base; + */ + /*@{*/ + int64_t vblank_base; + int64_t msc_base; + /*@}*/ + + /** + * Pointer to context to which this drawable is currently bound. */ - __DRIscreenPrivate *driScreenPriv; + __DRIcontext *driContextPriv; /** - * \name Display and screen information. - * - * Basically just need these for when the locking code needs to call - * \c __driUtilUpdateDrawableInfo. + * Pointer to screen on which this drawable was created. */ - /*@{*/ - __DRInativeDisplay *display; - int screen; - /*@}*/ + __DRIscreen *driScreenPriv; /** - * Called via glXSwapBuffers(). + * Controls swap interval as used by GLX_SGI_swap_control and + * GLX_MESA_swap_control. */ - void (*swapBuffers)( __DRIdrawablePrivate *dPriv ); + unsigned int swap_interval; + struct { + unsigned int tail; + unsigned int drawable_id; + } dri2; }; /** * Per-context private driver information. */ -struct __DRIcontextPrivateRec { - /** - * Kernel context handle used to access the device lock. - */ - __DRIid contextID; - +struct __DRIcontextRec { /** * Kernel context handle used to access the device lock. */ @@ -352,35 +388,30 @@ struct __DRIcontextPrivateRec { void *driverPrivate; /** - * This context's display pointer. + * Pointer back to the \c __DRIcontext that contains this structure. */ - __DRInativeDisplay *display; + __DRIcontext *pctx; /** * Pointer to drawable currently bound to this context for drawing. */ - __DRIdrawablePrivate *driDrawablePriv; + __DRIdrawable *driDrawablePriv; /** * Pointer to drawable currently bound to this context for reading. */ - __DRIdrawablePrivate *driReadablePriv; + __DRIdrawable *driReadablePriv; /** * Pointer to screen on which this context was created. */ - __DRIscreenPrivate *driScreenPriv; + __DRIscreen *driScreenPriv; }; /** * Per-screen private driver information. */ -struct __DRIscreenPrivateRec { - /** - * Display for this screen - */ - __DRInativeDisplay *display; - +struct __DRIscreenRec { /** * Current screen's number */ @@ -391,38 +422,21 @@ struct __DRIscreenPrivateRec { */ struct __DriverAPIRec DriverAPI; + const __DRIextension **extensions; /** - * \name DDX version * DDX / 2D driver version information. - * \todo Replace these fields with a \c __DRIversionRec. */ - /*@{*/ - int ddxMajor; - int ddxMinor; - int ddxPatch; - /*@}*/ + __DRIversion ddx_version; /** - * \name DRI version * DRI X extension version information. - * \todo Replace these fields with a \c __DRIversionRec. */ - /*@{*/ - int driMajor; - int driMinor; - int driPatch; - /*@}*/ + __DRIversion dri_version; /** - * \name DRM version * DRM (kernel module) version information. - * \todo Replace these fields with a \c __DRIversionRec. */ - /*@{*/ - int drmMajor; - int drmMinor; - int drmPatch; - /*@}*/ + __DRIversion drm_version; /** * ID used when the client sets the drawable lock. @@ -485,12 +499,7 @@ struct __DRIscreenPrivateRec { * context is created when the first "real" context is created on this * screen. */ - __DRIcontextPrivate dummyContextPriv; - - /** - * Hash table to hold the drawable information for this screen. - */ - void *drawHash; + __DRIcontext dummyContextPriv; /** * Device-dependent private information (not stored in the SAREA). @@ -499,66 +508,46 @@ struct __DRIscreenPrivateRec { */ void *private; - /** - * GLX visuals / FBConfigs for this screen. These are stored as a - * linked list. - * - * \note - * This field is \b only used in conjunction with the old interfaces. If - * the new interfaces are used, this field will be set to \c NULL and will - * not be dereferenced. - */ - __GLcontextModes *modes; - /** * Pointer back to the \c __DRIscreen that contains this structure. */ - __DRIscreen *psc; -}; - -/** - * Used to store a version which includes a major range instead of a single - * major version number. - */ -struct __DRIutilversionRec2 { - int major_min; /** min allowed Major version number. */ - int major_max; /** max allowed Major version number. */ - int minor; /**< Minor version number. */ - int patch; /**< Patch-level. */ + /* Extensions provided by the loader. */ + const __DRIgetDrawableInfoExtension *getDrawableInfo; + const __DRIsystemTimeExtension *systemTime; + const __DRIdamageExtension *damage; + + struct { + /* Flag to indicate that this is a DRI2 screen. Many of the above + * fields will not be valid or initializaed in that case. */ + int enabled; + drmBO sareaBO; + void *sarea; + __DRIEventBuffer *buffer; + __DRILock *lock; + __DRIloaderExtension *loader; + } dri2; + + /* The lock actually in use, old sarea or DRI2 */ + drmLock *lock; }; - extern void __driUtilMessage(const char *f, ...); extern void -__driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp); - - -extern __DRIscreenPrivate * __driUtilCreateNewScreen( __DRInativeDisplay *dpy, - int scrn, __DRIscreen *psc, __GLcontextModes * modes, - const __DRIversion * ddx_version, const __DRIversion * dri_version, - const __DRIversion * drm_version, const __DRIframebuffer * frame_buffer, - drm_sarea_t *pSAREA, int fd, int internal_api_version, - const struct __DriverAPIRec *driverAPI ); +__driUtilUpdateDrawableInfo(__DRIdrawable *pdp); -/* Test the version of the internal GLX API. Returns a value like strcmp. */ extern int -driCompareGLXAPIVersion( GLint required_version ); +__driParseEvents(__DRIcontext *psp, __DRIdrawable *pdp); extern float -driCalculateSwapUsage( __DRIdrawablePrivate *dPriv, +driCalculateSwapUsage( __DRIdrawable *dPriv, int64_t last_swap_ust, int64_t current_ust ); -/** - * Pointer to the \c __DRIinterfaceMethods passed to the driver by the loader. - * - * This pointer is set in the driver's \c __driCreateNewScreen function and - * is defined in dri_util.c. - */ -extern const __DRIinterfaceMethods * dri_interface; +extern GLint +driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ); #endif /* _DRI_UTIL_H_ */ diff --git a/src/mesa/drivers/dri/common/drirenderbuffer.c b/src/mesa/drivers/dri/common/drirenderbuffer.c index d34da53479..d36af3e5be 100644 --- a/src/mesa/drivers/dri/common/drirenderbuffer.c +++ b/src/mesa/drivers/dri/common/drirenderbuffer.c @@ -209,8 +209,6 @@ driUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv) struct gl_framebuffer *fb = (struct gl_framebuffer *) dPriv->driverPrivate; if (fb && (dPriv->w != fb->Width || dPriv->h != fb->Height)) { ctx->Driver.ResizeBuffers(ctx, fb, dPriv->w, dPriv->h); - /* if the driver needs the hw lock for ResizeBuffers, the drawable - might have changed again by now */ assert(fb->Width == dPriv->w); assert(fb->Height == dPriv->h); } diff --git a/src/mesa/drivers/dri/common/spantmp2.h b/src/mesa/drivers/dri/common/spantmp2.h index 50f3cf5581..53f5f846a0 100644 --- a/src/mesa/drivers/dri/common/spantmp2.h +++ b/src/mesa/drivers/dri/common/spantmp2.h @@ -114,7 +114,7 @@ do { \ GLuint p = *(volatile GLuint *) GET_PTR(_x, _y); \ __asm__ __volatile__( "bswap %0; rorl $8, %0" \ - : "=r" (p) : "r" (p) ); \ + : "=r" (p) : "0" (p) ); \ ((GLuint *)rgba)[0] = p; \ } while (0) # elif defined( MESA_BIG_ENDIAN ) diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c index 6a189e7285..7fbe0d855d 100644 --- a/src/mesa/drivers/dri/common/utils.c +++ b/src/mesa/drivers/dri/common/utils.c @@ -419,21 +419,6 @@ driCheckDriDdxDrmVersions2(const char * driver_name, drmActual, drmExpected); } - - -GLint -driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ) -{ - if (rect2.x1 > rect1.x1) rect1.x1 = rect2.x1; - if (rect2.x2 < rect1.x2) rect1.x2 = rect2.x2; - if (rect2.y1 > rect1.y1) rect1.y1 = rect2.y1; - if (rect2.y2 < rect1.y2) rect1.y2 = rect2.y2; - - if (rect1.x1 > rect1.x2 || rect1.y1 > rect1.y2) return 0; - - return (rect1.x2 - rect1.x1) * (rect1.y2 - rect1.y1); -} - GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, GLint *x, GLint *y, GLsizei *width, GLsizei *height ) @@ -467,8 +452,6 @@ GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, return GL_TRUE; } - - /** * Creates a set of \c __GLcontextModes that a driver will expose. * @@ -536,86 +519,99 @@ GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, * \c GL_UNSIGNED_3BYTE_8_8_8, \c GL_4FLOAT_32_32_32_32, * \c GL_4HALF_16_16_16_16, etc. We can cross that bridge when we come to it. */ -GLboolean -driFillInModes( __GLcontextModes ** ptr_to_modes, - GLenum fb_format, GLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, - unsigned num_depth_stencil_bits, - const GLenum * db_modes, unsigned num_db_modes, - int visType ) +__DRIconfig ** +driCreateConfigs(GLenum fb_format, GLenum fb_type, + const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + unsigned num_depth_stencil_bits, + const GLenum * db_modes, unsigned num_db_modes) { - static const u_int8_t bits_table[3][4] = { + static const u_int8_t bits_table[4][4] = { /* R G B A */ + { 3, 3, 2, 0 }, /* Any GL_UNSIGNED_BYTE_3_3_2 */ { 5, 6, 5, 0 }, /* Any GL_UNSIGNED_SHORT_5_6_5 */ { 8, 8, 8, 0 }, /* Any RGB with any GL_UNSIGNED_INT_8_8_8_8 */ { 8, 8, 8, 8 } /* Any RGBA with any GL_UNSIGNED_INT_8_8_8_8 */ }; - /* The following arrays are all indexed by the fb_type masked with 0x07. - * Given the four supported fb_type values, this results in valid array - * indices of 3, 4, 5, and 7. - */ - static const u_int32_t masks_table_rgb[8][4] = { - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + static const u_int32_t masks_table_rgb[6][4] = { + { 0x000000E0, 0x0000001C, 0x00000003, 0x00000000 }, /* 3_3_2 */ + { 0x00000007, 0x00000038, 0x000000C0, 0x00000000 }, /* 2_3_3_REV */ { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 }, /* 5_6_5 */ { 0x0000001F, 0x000007E0, 0x0000F800, 0x00000000 }, /* 5_6_5_REV */ { 0xFF000000, 0x00FF0000, 0x0000FF00, 0x00000000 }, /* 8_8_8_8 */ - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x000000FF, 0x0000FF00, 0x00FF0000, 0x00000000 } /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_rgba[8][4] = { - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + static const u_int32_t masks_table_rgba[6][4] = { + { 0x000000E0, 0x0000001C, 0x00000003, 0x00000000 }, /* 3_3_2 */ + { 0x00000007, 0x00000038, 0x000000C0, 0x00000000 }, /* 2_3_3_REV */ { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 }, /* 5_6_5 */ { 0x0000001F, 0x000007E0, 0x0000F800, 0x00000000 }, /* 5_6_5_REV */ { 0xFF000000, 0x00FF0000, 0x0000FF00, 0x000000FF }, /* 8_8_8_8 */ - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x000000FF, 0x0000FF00, 0x00FF0000, 0xFF000000 }, /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_bgr[8][4] = { - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + static const u_int32_t masks_table_bgr[6][4] = { + { 0x00000007, 0x00000038, 0x000000C0, 0x00000000 }, /* 3_3_2 */ + { 0x000000E0, 0x0000001C, 0x00000003, 0x00000000 }, /* 2_3_3_REV */ { 0x0000001F, 0x000007E0, 0x0000F800, 0x00000000 }, /* 5_6_5 */ { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 }, /* 5_6_5_REV */ { 0x0000FF00, 0x00FF0000, 0xFF000000, 0x00000000 }, /* 8_8_8_8 */ - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00FF0000, 0x0000FF00, 0x000000FF, 0x00000000 }, /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_bgra[8][4] = { - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + static const u_int32_t masks_table_bgra[6][4] = { + { 0x00000007, 0x00000038, 0x000000C0, 0x00000000 }, /* 3_3_2 */ + { 0x000000E0, 0x0000001C, 0x00000003, 0x00000000 }, /* 2_3_3_REV */ { 0x0000001F, 0x000007E0, 0x0000F800, 0x00000000 }, /* 5_6_5 */ { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 }, /* 5_6_5_REV */ { 0x0000FF00, 0x00FF0000, 0xFF000000, 0x000000FF }, /* 8_8_8_8 */ - { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000 }, /* 8_8_8_8_REV */ }; - static const u_int8_t bytes_per_pixel[8] = { - 0, 0, 0, 2, 2, 4, 0, 4 + static const u_int8_t bytes_per_pixel[6] = { + 1, /* 3_3_2 */ + 1, /* 2_3_3_REV */ + 2, /* 5_6_5 */ + 2, /* 5_6_5_REV */ + 4, /* 8_8_8_8 */ + 4 /* 8_8_8_8_REV */ }; const u_int8_t * bits; const u_int32_t * masks; - const int index = fb_type & 0x07; - __GLcontextModes * modes = *ptr_to_modes; + int index; + __DRIconfig **configs, **c; + __GLcontextModes *modes; unsigned i; unsigned j; unsigned k; - - - if ( bytes_per_pixel[ index ] == 0 ) { - fprintf( stderr, "[%s:%u] Framebuffer type 0x%04x has 0 bytes per pixel.\n", - __FUNCTION__, __LINE__, fb_type ); - return GL_FALSE; + unsigned num_modes; + unsigned num_accum_bits = 2; + + switch ( fb_type ) { + case GL_UNSIGNED_BYTE_3_3_2: + index = 0; + break; + case GL_UNSIGNED_BYTE_2_3_3_REV: + index = 1; + break; + case GL_UNSIGNED_SHORT_5_6_5: + index = 2; + break; + case GL_UNSIGNED_SHORT_5_6_5_REV: + index = 3; + break; + case GL_UNSIGNED_INT_8_8_8_8: + index = 4; + break; + case GL_UNSIGNED_INT_8_8_8_8_REV: + index = 5; + break; + default: + fprintf( stderr, "[%s:%u] Unknown framebuffer type 0x%04x.\n", + __FUNCTION__, __LINE__, fb_type ); + return NULL; } @@ -627,40 +623,55 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, switch ( fb_format ) { case GL_RGB: - bits = (bytes_per_pixel[ index ] == 2) - ? bits_table[0] : bits_table[1]; masks = masks_table_rgb[ index ]; break; case GL_RGBA: - bits = (bytes_per_pixel[ index ] == 2) - ? bits_table[0] : bits_table[2]; masks = masks_table_rgba[ index ]; break; case GL_BGR: - bits = (bytes_per_pixel[ index ] == 2) - ? bits_table[0] : bits_table[1]; masks = masks_table_bgr[ index ]; break; case GL_BGRA: - bits = (bytes_per_pixel[ index ] == 2) - ? bits_table[0] : bits_table[2]; masks = masks_table_bgra[ index ]; break; default: - fprintf( stderr, "[%s:%u] Framebuffer format 0x%04x is not GL_RGB, GL_RGBA, GL_BGR, or GL_BGRA.\n", - __FUNCTION__, __LINE__, fb_format ); - return GL_FALSE; + fprintf( stderr, "[%s:%u] Unknown framebuffer format 0x%04x.\n", + __FUNCTION__, __LINE__, fb_format ); + return NULL; + } + + switch ( bytes_per_pixel[ index ] ) { + case 1: + bits = bits_table[0]; + break; + case 2: + bits = bits_table[1]; + break; + default: + bits = ((fb_format == GL_RGB) || (fb_format == GL_BGR)) + ? bits_table[2] + : bits_table[3]; + break; } + num_modes = num_depth_stencil_bits * num_db_modes * num_accum_bits; + configs = _mesa_calloc((num_modes + 1) * sizeof *configs); + if (configs == NULL) + return NULL; + c = configs; for ( k = 0 ; k < num_depth_stencil_bits ; k++ ) { for ( i = 0 ; i < num_db_modes ; i++ ) { - for ( j = 0 ; j < 2 ; j++ ) { + for ( j = 0 ; j < num_accum_bits ; j++ ) { + *c = _mesa_malloc (sizeof **c); + modes = &(*c)->modes; + c++; + memset(modes, 0, sizeof *modes); modes->redBits = bits[0]; modes->greenBits = bits[1]; modes->blueBits = bits[2]; @@ -681,7 +692,13 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, modes->stencilBits = stencil_bits[k]; modes->depthBits = depth_bits[k]; - modes->visualType = visType; + modes->transparentPixel = GLX_NONE; + modes->transparentRed = GLX_DONT_CARE; + modes->transparentGreen = GLX_DONT_CARE; + modes->transparentBlue = GLX_DONT_CARE; + modes->transparentAlpha = GLX_DONT_CARE; + modes->transparentIndex = GLX_DONT_CARE; + modes->visualType = GLX_DONT_CARE; modes->renderType = GLX_RGBA_BIT; modes->drawableType = GLX_WINDOW_BIT; modes->rgbMode = GL_TRUE; @@ -701,11 +718,155 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, modes->haveDepthBuffer = (modes->depthBits > 0); modes->haveStencilBuffer = (modes->stencilBits > 0); - modes = modes->next; + modes->bindToTextureRgb = GL_TRUE; + modes->bindToTextureRgba = GL_TRUE; + modes->bindToMipmapTexture = GL_FALSE; + modes->bindToTextureTargets = modes->rgbMode ? + __DRI_ATTRIB_TEXTURE_1D_BIT | + __DRI_ATTRIB_TEXTURE_2D_BIT | + __DRI_ATTRIB_TEXTURE_RECTANGLE_BIT : + 0; } } } + *c = NULL; + + return configs; +} + +const __DRIconfig **driConcatConfigs(__DRIconfig **a, __DRIconfig **b) +{ + const __DRIconfig **all; + int i, j, index; + + i = 0; + while (a[i] != NULL) + i++; + j = 0; + while (b[j] != NULL) + j++; + + all = _mesa_malloc((i + j + 1) * sizeof *all); + index = 0; + for (i = 0; a[i] != NULL; i++) + all[index++] = a[i]; + for (j = 0; b[j] != NULL; j++) + all[index++] = b[j]; + all[index++] = NULL; + + _mesa_free(a); + _mesa_free(b); + + return all; +} + +#define __ATTRIB(attrib, field) \ + { attrib, offsetof(__GLcontextModes, field) } + +static const struct { unsigned int attrib, offset; } attribMap[] = { + __ATTRIB(__DRI_ATTRIB_BUFFER_SIZE, rgbBits), + __ATTRIB(__DRI_ATTRIB_LEVEL, level), + __ATTRIB(__DRI_ATTRIB_RED_SIZE, redBits), + __ATTRIB(__DRI_ATTRIB_GREEN_SIZE, greenBits), + __ATTRIB(__DRI_ATTRIB_BLUE_SIZE, blueBits), + __ATTRIB(__DRI_ATTRIB_ALPHA_SIZE, alphaBits), + __ATTRIB(__DRI_ATTRIB_DEPTH_SIZE, depthBits), + __ATTRIB(__DRI_ATTRIB_STENCIL_SIZE, stencilBits), + __ATTRIB(__DRI_ATTRIB_ACCUM_RED_SIZE, accumRedBits), + __ATTRIB(__DRI_ATTRIB_ACCUM_GREEN_SIZE, accumGreenBits), + __ATTRIB(__DRI_ATTRIB_ACCUM_BLUE_SIZE, accumBlueBits), + __ATTRIB(__DRI_ATTRIB_ACCUM_ALPHA_SIZE, accumAlphaBits), + __ATTRIB(__DRI_ATTRIB_SAMPLE_BUFFERS, sampleBuffers), + __ATTRIB(__DRI_ATTRIB_SAMPLES, samples), + __ATTRIB(__DRI_ATTRIB_DOUBLE_BUFFER, doubleBufferMode), + __ATTRIB(__DRI_ATTRIB_STEREO, stereoMode), + __ATTRIB(__DRI_ATTRIB_AUX_BUFFERS, numAuxBuffers), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_TYPE, transparentPixel), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_INDEX_VALUE, transparentPixel), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_RED_VALUE, transparentRed), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_GREEN_VALUE, transparentGreen), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_BLUE_VALUE, transparentBlue), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_ALPHA_VALUE, transparentAlpha), + __ATTRIB(__DRI_ATTRIB_FLOAT_MODE, floatMode), + __ATTRIB(__DRI_ATTRIB_RED_MASK, redMask), + __ATTRIB(__DRI_ATTRIB_GREEN_MASK, greenMask), + __ATTRIB(__DRI_ATTRIB_BLUE_MASK, blueMask), + __ATTRIB(__DRI_ATTRIB_ALPHA_MASK, alphaMask), + __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_WIDTH, maxPbufferWidth), + __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_HEIGHT, maxPbufferHeight), + __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_PIXELS, maxPbufferPixels), + __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_WIDTH, optimalPbufferWidth), + __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_HEIGHT, optimalPbufferHeight), + __ATTRIB(__DRI_ATTRIB_SWAP_METHOD, swapMethod), + __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGB, bindToTextureRgb), + __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGBA, bindToTextureRgba), + __ATTRIB(__DRI_ATTRIB_BIND_TO_MIPMAP_TEXTURE, bindToMipmapTexture), + __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_TARGETS, bindToTextureTargets), + __ATTRIB(__DRI_ATTRIB_YINVERTED, yInverted), + + /* The struct field doesn't matter here, these are handled by the + * switch in driGetConfigAttribIndex. We need them in the array + * so the iterator includes them though.*/ + __ATTRIB(__DRI_ATTRIB_RENDER_TYPE, level), + __ATTRIB(__DRI_ATTRIB_CONFIG_CAVEAT, level), + __ATTRIB(__DRI_ATTRIB_SWAP_METHOD, level) +}; + +#define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0])) + +static int +driGetConfigAttribIndex(const __DRIconfig *config, + unsigned int index, unsigned int *value) +{ + switch (attribMap[index].attrib) { + case __DRI_ATTRIB_RENDER_TYPE: + if (config->modes.rgbMode) + *value = __DRI_ATTRIB_RGBA_BIT; + else + *value = __DRI_ATTRIB_COLOR_INDEX_BIT; + break; + case __DRI_ATTRIB_CONFIG_CAVEAT: + if (config->modes.visualRating == GLX_NON_CONFORMANT_CONFIG) + *value = __DRI_ATTRIB_NON_CONFORMANT_CONFIG; + else if (config->modes.visualRating == GLX_SLOW_CONFIG) + *value = __DRI_ATTRIB_SLOW_BIT; + else + *value = 0; + break; + case __DRI_ATTRIB_SWAP_METHOD: + break; + + default: + *value = *(unsigned int *) + ((char *) &config->modes + attribMap[index].offset); + + break; + } - *ptr_to_modes = modes; return GL_TRUE; } + +int +driGetConfigAttrib(const __DRIconfig *config, + unsigned int attrib, unsigned int *value) +{ + int i; + + for (i = 0; i < ARRAY_SIZE(attribMap); i++) + if (attribMap[i].attrib == attrib) + return driGetConfigAttribIndex(config, i, value); + + return GL_FALSE; +} + +int +driIndexConfigAttrib(const __DRIconfig *config, int index, + unsigned int *attrib, unsigned int *value) +{ + if (index >= 0 && index < ARRAY_SIZE(attribMap)) { + *attrib = attribMap[index].attrib; + return driGetConfigAttribIndex(config, index, value); + } + + return GL_FALSE; +} diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h index b28b895627..9ac3b51447 100644 --- a/src/mesa/drivers/dri/common/utils.h +++ b/src/mesa/drivers/dri/common/utils.h @@ -28,8 +28,11 @@ #ifndef DRI_DEBUG_H #define DRI_DEBUG_H +#include +#include #include "context.h" -#include "dri_util.h" + +typedef struct __DRIutilversionRec2 __DRIutilversion2; struct dri_debug_control { const char * string; @@ -83,6 +86,17 @@ struct dri_extension { const struct dri_extension_function * functions; }; +/** + * Used to store a version which includes a major range instead of a single + * major version number. + */ +struct __DRIutilversionRec2 { + int major_min; /** min allowed Major version number. */ + int major_max; /** max allowed Major version number. */ + int minor; /**< Minor version number. */ + int patch; /**< Patch-level. */ +}; + extern unsigned driParseDebugString( const char * debug, const struct dri_debug_control * control ); @@ -105,16 +119,27 @@ extern GLboolean driCheckDriDdxDrmVersions3(const char * driver_name, const __DRIversion * ddxActual, const __DRIutilversion2 * ddxExpected, const __DRIversion * drmActual, const __DRIversion * drmExpected); -extern GLint driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ); - extern GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, GLint *x, GLint *y, GLsizei *width, GLsizei *height ); -extern GLboolean driFillInModes( __GLcontextModes ** modes, - GLenum fb_format, GLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, - unsigned num_depth_stencil_bits, - const GLenum * db_modes, unsigned num_db_modes, int visType ); +struct __DRIconfigRec { + __GLcontextModes modes; +}; + +extern __DRIconfig ** +driCreateConfigs(GLenum fb_format, GLenum fb_type, + const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + unsigned num_depth_stencil_bits, + const GLenum * db_modes, unsigned num_db_modes); + +const __DRIconfig **driConcatConfigs(__DRIconfig **a, __DRIconfig **b); + +int +driGetConfigAttrib(const __DRIconfig *config, + unsigned int attrib, unsigned int *value); +int +driIndexConfigAttrib(const __DRIconfig *config, int index, + unsigned int *attrib, unsigned int *value); #endif /* DRI_DEBUG_H */ diff --git a/src/mesa/drivers/dri/common/vblank.c b/src/mesa/drivers/dri/common/vblank.c index 094950d362..0008ab1c34 100644 --- a/src/mesa/drivers/dri/common/vblank.c +++ b/src/mesa/drivers/dri/common/vblank.c @@ -34,6 +34,16 @@ #include "vblank.h" #include "xmlpool.h" +static unsigned int msc_to_vblank(__DRIdrawablePrivate * dPriv, int64_t msc) +{ + return (unsigned int)(msc - dPriv->msc_base + dPriv->vblank_base); +} + +static int64_t vblank_to_msc(__DRIdrawablePrivate * dPriv, unsigned int vblank) +{ + return (int64_t)(vblank - dPriv->vblank_base + dPriv->msc_base); +} + /****************************************************************************/ /** @@ -41,7 +51,7 @@ * * Stores the 64-bit count of vertical refreshes since some (arbitrary) * point in time in \c count. Unless the value wraps around, which it - * may, it will never decrease. + * may, it will never decrease for a given drawable. * * \warning This function is called from \c glXGetVideoSyncSGI, which expects * a \c count of type \c unsigned (32-bit), and \c glXGetSyncValuesOML, which @@ -49,11 +59,14 @@ * currently always returns a \c sequence of type \c unsigned. * * \param priv Pointer to the DRI screen private struct. + * \param dPriv Pointer to the DRI drawable private struct * \param count Storage to hold MSC counter. * \return Zero is returned on success. A negative errno value * is returned on failure. */ -int driGetMSC32( __DRIscreenPrivate * priv, int64_t * count ) +int driDrawableGetMSC32( __DRIscreenPrivate * priv, + __DRIdrawablePrivate * dPriv, + int64_t * count) { drmVBlank vbl; int ret; @@ -62,14 +75,21 @@ int driGetMSC32( __DRIscreenPrivate * priv, int64_t * count ) vbl.request.type = DRM_VBLANK_RELATIVE; vbl.request.sequence = 0; + if ( dPriv && dPriv->vblFlags & VBLANK_FLAG_SECONDARY ) + vbl.request.type |= DRM_VBLANK_SECONDARY; ret = drmWaitVBlank( priv->fd, &vbl ); - *count = (int64_t)vbl.reply.sequence; + + if (dPriv) { + *count = vblank_to_msc(dPriv, vbl.reply.sequence); + } else { + /* Old driver (no knowledge of drawable MSC callback) */ + *count = vbl.reply.sequence; + } return ret; } - /****************************************************************************/ /** * Wait for a specified refresh count. This implements most of the @@ -122,7 +142,9 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, */ vbl.request.type = dont_wait ? DRM_VBLANK_RELATIVE : DRM_VBLANK_ABSOLUTE; - vbl.request.sequence = next; + vbl.request.sequence = next ? msc_to_vblank(priv, next) : 0; + if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) + vbl.request.type |= DRM_VBLANK_SECONDARY; if ( drmWaitVBlank( priv->driScreenPriv->fd, &vbl ) != 0 ) { /* FIXME: This doesn't seem like the right thing to return here. @@ -130,8 +152,10 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, return GLX_BAD_CONTEXT; } + *msc = vblank_to_msc(priv, vbl.reply.sequence); + dont_wait = 0; - if (target_msc != 0 && vbl.reply.sequence == target) + if (target_msc != 0 && *msc == target) break; /* Assuming the wait-done test fails, the next refresh to wait for @@ -141,9 +165,9 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, * If this refresh has already happened, we add divisor to obtain * the next refresh after the current one that will satisfy it. */ - r = (vbl.reply.sequence % (unsigned int)divisor); - next = (vbl.reply.sequence - r + (unsigned int)remainder); - if (next <= vbl.reply.sequence) next += (unsigned int)divisor; + r = (*msc % (unsigned int)divisor); + next = (*msc - r + (unsigned int)remainder); + if (next <= *msc) next += (unsigned int)divisor; } while ( r != (unsigned int)remainder ); } @@ -153,7 +177,10 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, */ vbl.request.type = DRM_VBLANK_ABSOLUTE; - vbl.request.sequence = target_msc; + vbl.request.sequence = target_msc ? msc_to_vblank(priv, target_msc) : 0; + + if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) + vbl.request.type |= DRM_VBLANK_SECONDARY; if ( drmWaitVBlank( priv->driScreenPriv->fd, &vbl ) != 0 ) { /* FIXME: This doesn't seem like the right thing to return here. @@ -162,8 +189,8 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, } } - *msc = (target_msc & 0xffffffff00000000LL); - *msc |= vbl.reply.sequence; + *msc = vblank_to_msc(priv, vbl.reply.sequence); + if ( *msc < target_msc ) { *msc += 0x0000000100000000LL; } @@ -232,8 +259,8 @@ static int do_wait( drmVBlank * vbl, GLuint * vbl_seq, int fd ) if ( first_time ) { fprintf(stderr, "%s: drmWaitVBlank returned %d, IRQs don't seem to be" - " working correctly.\nTry running with LIBGL_THROTTLE_REFRESH" - " and LIBL_SYNC_REFRESH unset.\n", __FUNCTION__, ret); + " working correctly.\nTry adjusting the vblank_mode" + " configuration parameter.\n", __FUNCTION__, ret); first_time = GL_FALSE; } @@ -245,22 +272,44 @@ static int do_wait( drmVBlank * vbl, GLuint * vbl_seq, int fd ) } +/****************************************************************************/ +/** + * Returns the default swap interval of the given drawable. + */ + +static unsigned +driGetDefaultVBlankInterval( const __DRIdrawablePrivate *priv ) +{ + if ( (priv->vblFlags & (VBLANK_FLAG_THROTTLE | VBLANK_FLAG_SYNC)) != 0 ) { + return 1; + } + else { + return 0; + } +} + + /****************************************************************************/ /** * Sets the default swap interval when the drawable is first bound to a * direct rendering context. */ -void driDrawableInitVBlank( __DRIdrawablePrivate *priv, GLuint flags, - GLuint *vbl_seq ) +void driDrawableInitVBlank( __DRIdrawablePrivate *priv ) { - if ( priv->pdraw->swap_interval == (unsigned)-1 ) { + if ( priv->swap_interval == (unsigned)-1 && + !( priv->vblFlags & VBLANK_FLAG_NO_IRQ ) ) { /* Get current vertical blank sequence */ - drmVBlank vbl = { .request={ .type = DRM_VBLANK_RELATIVE, .sequence = 0 } }; - do_wait( &vbl, vbl_seq, priv->driScreenPriv->fd ); - - priv->pdraw->swap_interval = (flags & (VBLANK_FLAG_THROTTLE | - VBLANK_FLAG_SYNC)) != 0 ? 1 : 0; + drmVBlank vbl; + + vbl.request.type = DRM_VBLANK_RELATIVE; + if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) + vbl.request.type |= DRM_VBLANK_SECONDARY; + vbl.request.sequence = 0; + do_wait( &vbl, &priv->vblSeq, priv->driScreenPriv->fd ); + priv->vblank_base = priv->vblSeq; + + priv->swap_interval = driGetDefaultVBlankInterval( priv ); } } @@ -271,21 +320,17 @@ void driDrawableInitVBlank( __DRIdrawablePrivate *priv, GLuint flags, */ unsigned -driGetVBlankInterval( const __DRIdrawablePrivate *priv, GLuint flags ) +driGetVBlankInterval( const __DRIdrawablePrivate *priv ) { - if ( (flags & VBLANK_FLAG_INTERVAL) != 0 ) { + if ( (priv->vblFlags & VBLANK_FLAG_INTERVAL) != 0 ) { /* this must have been initialized when the drawable was first bound * to a direct rendering context. */ - assert ( priv->pdraw->swap_interval != (unsigned)-1 ); + assert ( priv->swap_interval != (unsigned)-1 ); - return priv->pdraw->swap_interval; - } - else if ( (flags & (VBLANK_FLAG_THROTTLE | VBLANK_FLAG_SYNC)) != 0 ) { - return 1; - } - else { - return 0; + return priv->swap_interval; } + else + return driGetDefaultVBlankInterval( priv ); } @@ -295,18 +340,17 @@ driGetVBlankInterval( const __DRIdrawablePrivate *priv, GLuint flags ) */ void -driGetCurrentVBlank( const __DRIdrawablePrivate *priv, GLuint flags, - GLuint *vbl_seq ) +driGetCurrentVBlank( __DRIdrawablePrivate *priv ) { drmVBlank vbl; vbl.request.type = DRM_VBLANK_RELATIVE; - if ( flags & VBLANK_FLAG_SECONDARY ) { + if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) { vbl.request.type |= DRM_VBLANK_SECONDARY; } vbl.request.sequence = 0; - (void) do_wait( &vbl, vbl_seq, priv->driScreenPriv->fd ); + (void) do_wait( &vbl, &priv->vblSeq, priv->driScreenPriv->fd ); } @@ -314,19 +358,15 @@ driGetCurrentVBlank( const __DRIdrawablePrivate *priv, GLuint flags, /** * Waits for the vertical blank for use with glXSwapBuffers. * - * \param vbl_seq Vertical blank sequence number (MSC) after the last buffer - * swap. Updated after this wait. - * \param flags \c VBLANK_FLAG bits that control how long to wait. * \param missed_deadline Set to \c GL_TRUE if the MSC after waiting is later - * than the "target" based on \c flags. The idea is that if - * \c missed_deadline is set, then the application is not - * achieving its desired framerate. + * than the "target" based on \c priv->vblFlags. The idea is + * that if \c missed_deadline is set, then the application is + * not achieving its desired framerate. * \return Zero on success, -1 on error. */ int -driWaitForVBlank( const __DRIdrawablePrivate *priv, GLuint * vbl_seq, - GLuint flags, GLboolean * missed_deadline ) +driWaitForVBlank( __DRIdrawablePrivate *priv, GLboolean * missed_deadline ) { drmVBlank vbl; unsigned original_seq; @@ -335,10 +375,10 @@ driWaitForVBlank( const __DRIdrawablePrivate *priv, GLuint * vbl_seq, unsigned diff; *missed_deadline = GL_FALSE; - if ( (flags & (VBLANK_FLAG_INTERVAL | - VBLANK_FLAG_THROTTLE | - VBLANK_FLAG_SYNC)) == 0 || - (flags & VBLANK_FLAG_NO_IRQ) != 0 ) { + if ( (priv->vblFlags & (VBLANK_FLAG_INTERVAL | + VBLANK_FLAG_THROTTLE | + VBLANK_FLAG_SYNC)) == 0 || + (priv->vblFlags & VBLANK_FLAG_NO_IRQ) != 0 ) { return 0; } @@ -349,44 +389,45 @@ driWaitForVBlank( const __DRIdrawablePrivate *priv, GLuint * vbl_seq, * * VBLANK_FLAG_INTERVAL and VBLANK_FLAG_THROTTLE mean to wait for at * least one vertical blank since the last wait. Since do_wait modifies - * vbl_seq, we have to save the original value of vbl_seq for the + * priv->vblSeq, we have to save the original value of priv->vblSeq for the * VBLANK_FLAG_INTERVAL / VBLANK_FLAG_THROTTLE calculation later. */ - original_seq = *vbl_seq; - interval = driGetVBlankInterval(priv, flags); + original_seq = priv->vblSeq; + interval = driGetVBlankInterval(priv); deadline = original_seq + interval; vbl.request.type = DRM_VBLANK_RELATIVE; - if ( flags & VBLANK_FLAG_SECONDARY ) { + if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) { vbl.request.type |= DRM_VBLANK_SECONDARY; } - vbl.request.sequence = ((flags & VBLANK_FLAG_SYNC) != 0) ? 1 : 0; + vbl.request.sequence = ((priv->vblFlags & VBLANK_FLAG_SYNC) != 0) ? 1 : 0; - if ( do_wait( & vbl, vbl_seq, priv->driScreenPriv->fd ) != 0 ) { + if ( do_wait( & vbl, &priv->vblSeq, priv->driScreenPriv->fd ) != 0 ) { return -1; } - diff = *vbl_seq - deadline; + diff = priv->vblSeq - deadline; /* No need to wait again if we've already reached the target */ if (diff <= (1 << 23)) { - *missed_deadline = (flags & VBLANK_FLAG_SYNC) ? (diff > 0) : GL_TRUE; + *missed_deadline = (priv->vblFlags & VBLANK_FLAG_SYNC) ? (diff > 0) : + GL_TRUE; return 0; } /* Wait until the target vertical blank. */ vbl.request.type = DRM_VBLANK_ABSOLUTE; - if ( flags & VBLANK_FLAG_SECONDARY ) { + if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) { vbl.request.type |= DRM_VBLANK_SECONDARY; } vbl.request.sequence = deadline; - if ( do_wait( & vbl, vbl_seq, priv->driScreenPriv->fd ) != 0 ) { + if ( do_wait( & vbl, &priv->vblSeq, priv->driScreenPriv->fd ) != 0 ) { return -1; } - diff = *vbl_seq - deadline; + diff = priv->vblSeq - deadline; *missed_deadline = diff > 0 && diff <= (1 << 23); return 0; diff --git a/src/mesa/drivers/dri/common/vblank.h b/src/mesa/drivers/dri/common/vblank.h index 52c1933ca5..b3a0dadab1 100644 --- a/src/mesa/drivers/dri/common/vblank.h +++ b/src/mesa/drivers/dri/common/vblank.h @@ -45,17 +45,17 @@ */ extern int driGetMSC32( __DRIscreenPrivate * priv, int64_t * count ); +extern int driDrawableGetMSC32( __DRIscreenPrivate * priv, + __DRIdrawablePrivate * drawablePrivate, + int64_t * count); extern int driWaitForMSC32( __DRIdrawablePrivate *priv, int64_t target_msc, int64_t divisor, int64_t remainder, int64_t * msc ); extern GLuint driGetDefaultVBlankFlags( const driOptionCache *optionCache ); -extern void driDrawableInitVBlank ( __DRIdrawablePrivate *priv, GLuint flags, - GLuint *vbl_seq ); -extern unsigned driGetVBlankInterval( const __DRIdrawablePrivate *priv, - GLuint flags ); -extern void driGetCurrentVBlank( const __DRIdrawablePrivate *priv, - GLuint flags, GLuint *vbl_seq ); -extern int driWaitForVBlank( const __DRIdrawablePrivate *priv, - GLuint * vbl_seq, GLuint flags, GLboolean * missed_deadline ); +extern void driDrawableInitVBlank ( __DRIdrawablePrivate *priv ); +extern unsigned driGetVBlankInterval( const __DRIdrawablePrivate *priv ); +extern void driGetCurrentVBlank( __DRIdrawablePrivate *priv ); +extern int driWaitForVBlank( __DRIdrawablePrivate *priv, + GLboolean * missed_deadline ); #undef usleep #include /* for usleep() */ diff --git a/src/mesa/drivers/dri/common/xmlconfig.c b/src/mesa/drivers/dri/common/xmlconfig.c index b635894fe5..8602d47cf9 100644 --- a/src/mesa/drivers/dri/common/xmlconfig.c +++ b/src/mesa/drivers/dri/common/xmlconfig.c @@ -279,7 +279,7 @@ static GLfloat strToF (const XML_Char *string, const XML_Char **tail) { /** \brief Parse a value of a given type. */ static GLboolean parseValue (driOptionValue *v, driOptionType type, const XML_Char *string) { - const XML_Char *tail; + const XML_Char *tail = NULL; /* skip leading white-space */ string += strspn (string, " \f\n\r\t\v"); switch (type) { @@ -403,40 +403,40 @@ static GLboolean checkValue (const driOptionValue *v, const driOptionInfo *info) /** \brief Output a warning message. */ #define XML_WARNING1(msg) do {\ __driUtilMessage ("Warning in %s line %d, column %d: "msg, data->name, \ - XML_GetCurrentLineNumber(data->parser), \ - XML_GetCurrentColumnNumber(data->parser)); \ + (int) XML_GetCurrentLineNumber(data->parser), \ + (int) XML_GetCurrentColumnNumber(data->parser)); \ } while (0) #define XML_WARNING(msg,args...) do { \ __driUtilMessage ("Warning in %s line %d, column %d: "msg, data->name, \ - XML_GetCurrentLineNumber(data->parser), \ - XML_GetCurrentColumnNumber(data->parser), \ + (int) XML_GetCurrentLineNumber(data->parser), \ + (int) XML_GetCurrentColumnNumber(data->parser), \ args); \ } while (0) /** \brief Output an error message. */ #define XML_ERROR1(msg) do { \ __driUtilMessage ("Error in %s line %d, column %d: "msg, data->name, \ - XML_GetCurrentLineNumber(data->parser), \ - XML_GetCurrentColumnNumber(data->parser)); \ + (int) XML_GetCurrentLineNumber(data->parser), \ + (int) XML_GetCurrentColumnNumber(data->parser)); \ } while (0) #define XML_ERROR(msg,args...) do { \ __driUtilMessage ("Error in %s line %d, column %d: "msg, data->name, \ - XML_GetCurrentLineNumber(data->parser), \ - XML_GetCurrentColumnNumber(data->parser), \ + (int) XML_GetCurrentLineNumber(data->parser), \ + (int) XML_GetCurrentColumnNumber(data->parser), \ args); \ } while (0) /** \brief Output a fatal error message and abort. */ #define XML_FATAL1(msg) do { \ fprintf (stderr, "Fatal error in %s line %d, column %d: "msg"\n", \ data->name, \ - XML_GetCurrentLineNumber(data->parser), \ - XML_GetCurrentColumnNumber(data->parser)); \ + (int) XML_GetCurrentLineNumber(data->parser), \ + (int) XML_GetCurrentColumnNumber(data->parser)); \ abort();\ } while (0) #define XML_FATAL(msg,args...) do { \ fprintf (stderr, "Fatal error in %s line %d, column %d: "msg"\n", \ data->name, \ - XML_GetCurrentLineNumber(data->parser), \ - XML_GetCurrentColumnNumber(data->parser), \ + (int) XML_GetCurrentLineNumber(data->parser), \ + (int) XML_GetCurrentColumnNumber(data->parser), \ args); \ abort();\ } while (0) -- cgit v1.2.3 From 4dd1917e4be3ae48b436ed333bd2fcd37603d1ed Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Wed, 11 Jun 2008 14:51:41 +0100 Subject: small optimization --- src/gallium/auxiliary/util/u_pack_color.h | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 655e2c8259..06abb34d5a 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -255,20 +255,28 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) static INLINE uint util_pack_z(enum pipe_format format, double z) { + if (z == 0.0) + return 0; + switch (format) { case PIPE_FORMAT_Z16_UNORM: + if (z == 1.0) + return 0xffff; return (uint) (z * 0xffff); case PIPE_FORMAT_Z32_UNORM: /* special-case to avoid overflow */ if (z == 1.0) return 0xffffffff; - else - return (uint) (z * 0xffffffff); + return (uint) (z * 0xffffffff); case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_X8Z24_UNORM: + if (z == 1.0) + return 0xffffff; return (uint) (z * 0xffffff); case PIPE_FORMAT_Z24S8_UNORM: case PIPE_FORMAT_Z24X8_UNORM: + if (z == 1.0) + return 0xffffff00; return ((uint) (z * 0xffffff)) << 8; default: debug_print_format("gallium: unhandled format in util_pack_z()", format); -- cgit v1.2.3 From 1bcb817167773d6a148dd4b2cd63777d0f072c08 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Wed, 11 Jun 2008 14:52:55 +0100 Subject: Hook gallium i915 up to DRI2 --- src/gallium/winsys/dri/intel/intel_context.c | 9 ++- src/gallium/winsys/dri/intel/intel_lock.c | 7 +- src/gallium/winsys/dri/intel/intel_screen.c | 98 +++--------------------- src/gallium/winsys/dri/intel/intel_screen.h | 4 +- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 3 +- src/glx/x11/Makefile | 1 + 6 files changed, 27 insertions(+), 95 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index ecc4b0aa53..af44018053 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -28,6 +28,8 @@ #include "i830_dri.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" #include "intel_screen.h" #include "intel_context.h" #include "intel_swapbuffers.h" @@ -36,8 +38,6 @@ #include "i915simple/i915_screen.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" #include "pipe/p_defines.h" #include "pipe/p_context.h" @@ -160,8 +160,9 @@ static const struct dri_extension ttm_extensions[] = { * Initializes potential list of extensions if ctx == NULL, or actually enables * extensions for a context. */ -void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging) +void intelInitExtensions(struct st_context *st, GLboolean enable_imaging) { + GLcontext *ctx = st ? st->ctx : NULL; /* Disable imaging extension until convolution is working in teximage paths. */ enable_imaging = GL_FALSE; @@ -326,7 +327,7 @@ intelCreateContext(const __GLcontextModes * visual, intel->st = st_create_context(pipe, visual, st_share); - intelInitExtensions( intel->st->ctx, GL_TRUE ); + intelInitExtensions( intel->st, GL_TRUE ); return GL_TRUE; } diff --git a/src/gallium/winsys/dri/intel/intel_lock.c b/src/gallium/winsys/dri/intel/intel_lock.c index 406284c98f..469090c0e1 100644 --- a/src/gallium/winsys/dri/intel/intel_lock.c +++ b/src/gallium/winsys/dri/intel/intel_lock.c @@ -50,6 +50,9 @@ intelContendedLock(struct intel_context *intel, uint flags) DBG(LOCK, "%s - got contended lock\n", __progname); + if (sPriv->dri2.enabled) + return; + /* If the window moved, may need to set a new cliprect now. * * NOTE: This releases and regains the hw lock, so all state @@ -58,8 +61,8 @@ intelContendedLock(struct intel_context *intel, uint flags) if (dPriv) DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - if (sarea->width != intelScreen->front.width || - sarea->height != intelScreen->front.height) { + if (sarea && (sarea->width != intelScreen->front.width || + sarea->height != intelScreen->front.height)) { intelUpdateScreenRotation(sPriv, sarea); } diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 23889c80b8..a36c9407d4 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -110,116 +110,40 @@ intelHandleDrawableConfig(__DRIdrawablePrivate *dPriv, __DRIcontextPrivate *pcp, __DRIDrawableConfigEvent *event) { - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - struct intel_region *region = NULL; - struct intel_renderbuffer *rb, *depth_rb, *stencil_rb; - struct intel_context *intel = pcp->driverPrivate; - struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - int cpp, pitch; - -#if 0 - cpp = intelScreen->front.cpp; - pitch = ((cpp * dPriv->w + 63) & ~63) / cpp; - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - rb = intel_fb->color_rb[1]; - if (rb) { - region = intel_region_alloc(intel, cpp, pitch, dPriv->h); - intel_renderbuffer_set_region(rb, region); - } - - rb = intel_fb->color_rb[2]; - if (rb) { - region = intel_region_alloc(intel, cpp, pitch, dPriv->h); - intel_renderbuffer_set_region(rb, region); - } - - depth_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH); - stencil_rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL); - if (depth_rb || stencil_rb) - region = intel_region_alloc(intel, cpp, pitch, dPriv->h); - if (depth_rb) - intel_renderbuffer_set_region(depth_rb, region); - if (stencil_rb) - intel_renderbuffer_set_region(stencil_rb, region); - - /* FIXME: Tell the X server about the regions we just allocated and - * attached. */ -#endif - + (void) dPriv; + (void) pcp; + (void) event; } -#define BUFFER_FLAG_TILED 0x0100 - static void intelHandleBufferAttach(__DRIdrawablePrivate *dPriv, __DRIcontextPrivate *pcp, __DRIBufferAttachEvent *ba) { struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - struct intel_framebuffer *intel_fb = dPriv->driverPrivate; - struct intel_renderbuffer *rb; - struct intel_region *region; - struct intel_context *intel = pcp->driverPrivate; - struct pipe_surface *surf; - GLuint tiled; switch (ba->buffer.attachment) { case DRI_DRAWABLE_BUFFER_FRONT_LEFT: - #if 0 - intelScreen->front.width = ba->width; - intelScreen->front.height = ba->height; - intelScreen->front.offset = sarea->front_offset; - #endif - intelScreen->front.pitch = ba->buffer.pitch * ba->buffer.cpp; - #if 0 - intelScreen->front.size = sarea->front_size; - #endif + intelScreen->front.width = dPriv->w; + intelScreen->front.height = dPriv->h; + intelScreen->front.cpp = ba->buffer.cpp; + intelScreen->front.pitch = ba->buffer.pitch; driGenBuffers(intelScreen->base.staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); driBOSetReferenced(intelScreen->front.buffer, ba->buffer.handle); - break; -#if 0 case DRI_DRAWABLE_BUFFER_BACK_LEFT: - rb = intel_fb->color_rb[0]; - break; - case DRI_DRAWABLE_BUFFER_DEPTH: - rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_DEPTH); - break; - case DRI_DRAWABLE_BUFFER_STENCIL: - rb = intel_get_renderbuffer(&intel_fb->Base, BUFFER_STENCIL); - break; -#endif - case DRI_DRAWABLE_BUFFER_ACCUM: + /* anything ?? */ + break; + default: - fprintf(stderr, "unhandled buffer attach event, attacment type %d\n", + fprintf(stderr, "unhandled buffer attach event, attachment type %d\n", ba->buffer.attachment); return; } - -#if 0 - /* FIXME: Add this so we can filter out when the X server sends us - * attachment events for the buffers we just allocated. Need to - * get the BO handle for a render buffer. */ - if (intel_renderbuffer_get_region_handle(rb) == ba->buffer.handle) - return; -#endif - -#if 0 - tiled = (ba->buffer.flags & BUFFER_FLAG_TILED) > 0; - - region = intel_region_alloc_for_handle(intel, ba->buffer.cpp, - ba->buffer.pitch / ba->buffer.cpp, - dPriv->h, tiled, - ba->buffer.handle); - - intel_renderbuffer_set_region(rb, region); -#endif } static const __DRItexOffsetExtension intelTexOffsetExtension = { diff --git a/src/gallium/winsys/dri/intel/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h index 8036917903..d1389a31fb 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.h +++ b/src/gallium/winsys/dri/intel/intel_screen.h @@ -32,7 +32,7 @@ #include "i830_common.h" #include "xmlconfig.h" #include "intel_drm/ws_dri_bufpool.h" - +#include "state_tracker/st_context.h" #include "pipe/p_compiler.h" #include "intel_drm/intel_be_device.h" @@ -116,5 +116,7 @@ intelCreateContext(const __GLcontextModes * visual, __DRIcontextPrivate * driContextPriv, void *sharedContextPrivate); +extern void +intelInitExtensions(struct st_context *st, GLboolean enable_imaging); #endif diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index 7f3babd98e..8b17154cf9 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -78,7 +78,8 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, * current context (which is what intelScreenContext should return) might * not get a contended lock and thus cliprects not updated (tests/manywin) */ - if (intel_context(dPriv->driContextPriv) != intel) + if (!intel->driScreen->dri2.enabled && + intel_context(dPriv->driContextPriv) != intel) DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); diff --git a/src/glx/x11/Makefile b/src/glx/x11/Makefile index 1304311794..bb4d3cc5ee 100644 --- a/src/glx/x11/Makefile +++ b/src/glx/x11/Makefile @@ -50,6 +50,7 @@ INCLUDES = -I. \ -I$(TOP)/include \ -I$(TOP)/include/GL/internal \ -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/glapi \ -I$(TOP)/src/mesa/main \ $(LIBDRM_CFLAGS) \ $(DRI2PROTO_CFLAGS) \ -- cgit v1.2.3 From 5e9fe62c7090768b82a3cc905d1da6b623304b50 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Wed, 11 Jun 2008 15:25:35 +0100 Subject: add missed files for DRI2 merge --- src/glx/x11/dri2.c | 252 ++++++++++++++ src/glx/x11/dri2.h | 53 +++ src/glx/x11/dri2_glx.c | 371 +++++++++++++++++++++ src/glx/x11/dri_common.c | 399 ++++++++++++++++++++++ src/glx/x11/dri_common.h | 60 ++++ src/glx/x11/glxcurrent.c | 510 +++++++++++++++++++++++++++++ src/glx/x11/glxhash.c | 416 +++++++++++++++++++++++ src/glx/x11/glxhash.h | 16 + src/glx/x11/indirect_texture_compression.c | 347 ++++++++++++++++++++ src/glx/x11/indirect_vertex_array_priv.h | 308 +++++++++++++++++ 10 files changed, 2732 insertions(+) create mode 100644 src/glx/x11/dri2.c create mode 100644 src/glx/x11/dri2.h create mode 100644 src/glx/x11/dri2_glx.c create mode 100644 src/glx/x11/dri_common.c create mode 100644 src/glx/x11/dri_common.h create mode 100644 src/glx/x11/glxcurrent.c create mode 100644 src/glx/x11/glxhash.c create mode 100644 src/glx/x11/glxhash.h create mode 100644 src/glx/x11/indirect_texture_compression.c create mode 100644 src/glx/x11/indirect_vertex_array_priv.h (limited to 'src') diff --git a/src/glx/x11/dri2.c b/src/glx/x11/dri2.c new file mode 100644 index 0000000000..e7044ab424 --- /dev/null +++ b/src/glx/x11/dri2.c @@ -0,0 +1,252 @@ +/* + * Copyright © 2008 Red Hat, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Soft- + * ware"), to deal in the Software without restriction, including without + * limitation the rights to use, copy, modify, merge, publish, distribute, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, provided that the above copyright + * notice(s) and this permission notice appear in all copies of the Soft- + * ware and that both the above copyright notice(s) and this permission + * notice appear in supporting documentation. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- + * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY + * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN + * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- + * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, + * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER + * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- + * MANCE OF THIS SOFTWARE. + * + * Except as contained in this notice, the name of a copyright holder shall + * not be used in advertising or otherwise to promote the sale, use or + * other dealings in this Software without prior written authorization of + * the copyright holder. + * + * Authors: + * Kristian Høgsberg (krh@redhat.com) + */ + + +#define NEED_REPLIES +#include +#include +#include +#include +#include "glheader.h" +#include "xf86drm.h" +#include "dri2.h" + +static char dri2ExtensionName[] = DRI2_NAME; +static XExtensionInfo *dri2Info; +static XEXT_GENERATE_CLOSE_DISPLAY (DRI2CloseDisplay, dri2Info) +static /* const */ XExtensionHooks dri2ExtensionHooks = { + NULL, /* create_gc */ + NULL, /* copy_gc */ + NULL, /* flush_gc */ + NULL, /* free_gc */ + NULL, /* create_font */ + NULL, /* free_font */ + DRI2CloseDisplay, /* close_display */ + NULL, /* wire_to_event */ + NULL, /* event_to_wire */ + NULL, /* error */ + NULL, /* error_string */ +}; + +static XEXT_GENERATE_FIND_DISPLAY (DRI2FindDisplay, dri2Info, + dri2ExtensionName, + &dri2ExtensionHooks, + 0, NULL) + +Bool DRI2QueryExtension(Display *dpy, int *eventBase, int *errorBase) +{ + XExtDisplayInfo *info = DRI2FindDisplay(dpy); + + if (XextHasExtension(info)) { + *eventBase = info->codes->first_event; + *errorBase = info->codes->first_error; + return True; + } + + return False; +} + +Bool DRI2QueryVersion(Display *dpy, int *major, int *minor) +{ + XExtDisplayInfo *info = DRI2FindDisplay (dpy); + xDRI2QueryVersionReply rep; + xDRI2QueryVersionReq *req; + + XextCheckExtension (dpy, info, dri2ExtensionName, False); + + LockDisplay(dpy); + GetReq(DRI2QueryVersion, req); + req->reqType = info->codes->major_opcode; + req->dri2ReqType = X_DRI2QueryVersion; + req->majorVersion = DRI2_MAJOR; + req->minorVersion = DRI2_MINOR; + if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { + UnlockDisplay(dpy); + SyncHandle(); + return False; + } + *major = rep.majorVersion; + *minor = rep.minorVersion; + UnlockDisplay(dpy); + SyncHandle(); + + return True; +} + +Bool DRI2Connect(Display *dpy, int screen, + char **driverName, char **busId, unsigned int *sareaHandle) +{ + XExtDisplayInfo *info = DRI2FindDisplay(dpy); + xDRI2ConnectReply rep; + xDRI2ConnectReq *req; + + XextCheckExtension (dpy, info, dri2ExtensionName, False); + + LockDisplay(dpy); + GetReq(DRI2Connect, req); + req->reqType = info->codes->major_opcode; + req->dri2ReqType = X_DRI2Connect; + req->screen = screen; + if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { + UnlockDisplay(dpy); + SyncHandle(); + return False; + } + + *sareaHandle = rep.sareaHandle; + + *driverName = Xmalloc(rep.driverNameLength + 1); + if (*driverName == NULL) { + _XEatData(dpy, + ((rep.driverNameLength + 3) & ~3) + + ((rep.busIdLength + 3) & ~3)); + UnlockDisplay(dpy); + SyncHandle(); + return False; + } + _XReadPad(dpy, *driverName, rep.driverNameLength); + (*driverName)[rep.driverNameLength] = '\0'; + + *busId = Xmalloc(rep.busIdLength + 1); + if (*busId == NULL) { + Xfree(*driverName); + _XEatData(dpy, ((rep.busIdLength + 3) & ~3)); + UnlockDisplay(dpy); + SyncHandle(); + return False; + } + _XReadPad(dpy, *busId, rep.busIdLength); + (*busId)[rep.busIdLength] = '\0'; + + UnlockDisplay(dpy); + SyncHandle(); + + return rep.sareaHandle != 0; +} + +Bool DRI2AuthConnection(Display *dpy, int screen, drm_magic_t magic) +{ + XExtDisplayInfo *info = DRI2FindDisplay(dpy); + xDRI2AuthConnectionReq *req; + xDRI2AuthConnectionReply rep; + + XextCheckExtension (dpy, info, dri2ExtensionName, False); + + LockDisplay(dpy); + GetReq(DRI2AuthConnection, req); + req->reqType = info->codes->major_opcode; + req->dri2ReqType = X_DRI2AuthConnection; + req->screen = screen; + req->magic = magic; + rep.authenticated = 0; + if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { + UnlockDisplay(dpy); + SyncHandle(); + return False; + } + UnlockDisplay(dpy); + SyncHandle(); + + return rep.authenticated; +} + +Bool DRI2CreateDrawable(Display *dpy, XID drawable, + unsigned int *handle, unsigned int *head) +{ + XExtDisplayInfo *info = DRI2FindDisplay(dpy); + xDRI2CreateDrawableReply rep; + xDRI2CreateDrawableReq *req; + + XextCheckExtension (dpy, info, dri2ExtensionName, False); + + LockDisplay(dpy); + GetReq(DRI2CreateDrawable, req); + req->reqType = info->codes->major_opcode; + req->dri2ReqType = X_DRI2CreateDrawable; + req->drawable = drawable; + if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { + UnlockDisplay(dpy); + SyncHandle(); + return False; + } + UnlockDisplay(dpy); + SyncHandle(); + + *handle = rep.handle; + *head = rep.head; + + return True; +} + +void DRI2DestroyDrawable(Display *dpy, XID drawable) +{ + XExtDisplayInfo *info = DRI2FindDisplay(dpy); + xDRI2DestroyDrawableReq *req; + + XextSimpleCheckExtension (dpy, info, dri2ExtensionName); + + XSync(dpy, GL_FALSE); + + LockDisplay(dpy); + GetReq(DRI2DestroyDrawable, req); + req->reqType = info->codes->major_opcode; + req->dri2ReqType = X_DRI2DestroyDrawable; + req->drawable = drawable; + UnlockDisplay(dpy); + SyncHandle(); +} + +Bool DRI2ReemitDrawableInfo(Display *dpy, XID drawable, unsigned int *head) +{ + XExtDisplayInfo *info = DRI2FindDisplay(dpy); + xDRI2ReemitDrawableInfoReply rep; + xDRI2ReemitDrawableInfoReq *req; + + XextCheckExtension (dpy, info, dri2ExtensionName, False); + + LockDisplay(dpy); + GetReq(DRI2ReemitDrawableInfo, req); + req->reqType = info->codes->major_opcode; + req->dri2ReqType = X_DRI2ReemitDrawableInfo; + req->drawable = drawable; + if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { + UnlockDisplay(dpy); + SyncHandle(); + return False; + } + UnlockDisplay(dpy); + SyncHandle(); + + *head = rep.head; + + return True; +} diff --git a/src/glx/x11/dri2.h b/src/glx/x11/dri2.h new file mode 100644 index 0000000000..1dfd0448b2 --- /dev/null +++ b/src/glx/x11/dri2.h @@ -0,0 +1,53 @@ +/* + * Copyright © 2007,2008 Red Hat, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Soft- + * ware"), to deal in the Software without restriction, including without + * limitation the rights to use, copy, modify, merge, publish, distribute, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, provided that the above copyright + * notice(s) and this permission notice appear in all copies of the Soft- + * ware and that both the above copyright notice(s) and this permission + * notice appear in supporting documentation. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- + * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY + * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN + * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- + * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, + * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER + * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- + * MANCE OF THIS SOFTWARE. + * + * Except as contained in this notice, the name of a copyright holder shall + * not be used in advertising or otherwise to promote the sale, use or + * other dealings in this Software without prior written authorization of + * the copyright holder. + * + * Authors: + * Kristian Høgsberg (krh@redhat.com) + */ + +#ifndef _DRI2_H_ +#define _DRI2_H_ + +extern Bool +DRI2QueryExtension(Display *display, int *eventBase, int *errorBase); +extern Bool +DRI2QueryVersion(Display *display, int *major, int *minor); +extern Bool +DRI2Connect(Display *display, int screen, + char **driverName, char **busId, unsigned int *sareaHandle); +extern Bool +DRI2AuthConnection(Display *display, int screen, drm_magic_t magic); +extern Bool +DRI2CreateDrawable(Display *display, XID drawable, + unsigned int *handle, unsigned int *head); +extern void +DRI2DestroyDrawable(Display *display, XID handle); +extern Bool +DRI2ReemitDrawableInfo(Display *dpy, XID handle, unsigned int *head); + +#endif diff --git a/src/glx/x11/dri2_glx.c b/src/glx/x11/dri2_glx.c new file mode 100644 index 0000000000..b679c72c10 --- /dev/null +++ b/src/glx/x11/dri2_glx.c @@ -0,0 +1,371 @@ +/* + * Copyright © 2008 Red Hat, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Soft- + * ware"), to deal in the Software without restriction, including without + * limitation the rights to use, copy, modify, merge, publish, distribute, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, provided that the above copyright + * notice(s) and this permission notice appear in all copies of the Soft- + * ware and that both the above copyright notice(s) and this permission + * notice appear in supporting documentation. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- + * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY + * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN + * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- + * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, + * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER + * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- + * MANCE OF THIS SOFTWARE. + * + * Except as contained in this notice, the name of a copyright holder shall + * not be used in advertising or otherwise to promote the sale, use or + * other dealings in this Software without prior written authorization of + * the copyright holder. + * + * Authors: + * Kristian Høgsberg (krh@redhat.com) + */ + +#ifdef GLX_DIRECT_RENDERING + +#include +#include +#include +#include "glheader.h" +#include "glxclient.h" +#include "glcontextmodes.h" +#include "xf86dri.h" +#include "sarea.h" +#include +#include +#include +#include "xf86drm.h" +#include "dri2.h" +#include "dri_common.h" + +typedef struct __GLXDRIdisplayPrivateRec __GLXDRIdisplayPrivate; +typedef struct __GLXDRIcontextPrivateRec __GLXDRIcontextPrivate; + +struct __GLXDRIdisplayPrivateRec { + __GLXDRIdisplay base; + + /* + ** XFree86-DRI version information + */ + int driMajor; + int driMinor; + int driPatch; +}; + +struct __GLXDRIcontextPrivateRec { + __GLXDRIcontext base; + __DRIcontext *driContext; + __GLXscreenConfigs *psc; +}; + +static void dri2DestroyContext(__GLXDRIcontext *context, + __GLXscreenConfigs *psc, Display *dpy) +{ + __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; + const __DRIcoreExtension *core = pcp->psc->core; + + (*core->destroyContext)(pcp->driContext); + + Xfree(pcp); +} + +static Bool dri2BindContext(__GLXDRIcontext *context, + __GLXDRIdrawable *draw, __GLXDRIdrawable *read) +{ + __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; + const __DRIcoreExtension *core = pcp->psc->core; + + return (*core->bindContext)(pcp->driContext, + draw->driDrawable, + read->driDrawable); +} + +static void dri2UnbindContext(__GLXDRIcontext *context) +{ + __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; + const __DRIcoreExtension *core = pcp->psc->core; + + (*core->unbindContext)(pcp->driContext); +} + +static __GLXDRIcontext *dri2CreateContext(__GLXscreenConfigs *psc, + const __GLcontextModes *mode, + GLXContext gc, + GLXContext shareList, int renderType) +{ + __GLXDRIcontextPrivate *pcp, *pcp_shared; + __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode; + const __DRIcoreExtension *core = psc->core; + __DRIcontext *shared = NULL; + + if (shareList) { + pcp_shared = (__GLXDRIcontextPrivate *) shareList->driContext; + shared = pcp_shared->driContext; + } + + pcp = Xmalloc(sizeof *pcp); + if (pcp == NULL) + return NULL; + + pcp->psc = psc; + pcp->driContext = + (*core->createNewContext)(psc->__driScreen, + config->driConfig, shared, pcp); + gc->__driContext = pcp->driContext; + + if (pcp->driContext == NULL) { + Xfree(pcp); + return NULL; + } + + pcp->base.destroyContext = dri2DestroyContext; + pcp->base.bindContext = dri2BindContext; + pcp->base.unbindContext = dri2UnbindContext; + + return &pcp->base; +} + +static void dri2DestroyDrawable(__GLXDRIdrawable *pdraw) +{ + const __DRIcoreExtension *core = pdraw->psc->core; + + (*core->destroyDrawable)(pdraw->driDrawable); + DRI2DestroyDrawable(pdraw->psc->dpy, pdraw->drawable); + Xfree(pdraw); +} + +static __GLXDRIdrawable *dri2CreateDrawable(__GLXscreenConfigs *psc, + XID xDrawable, + GLXDrawable drawable, + const __GLcontextModes *modes) +{ + __GLXDRIdrawable *pdraw; + __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes; + unsigned int handle, head; + const __DRIcoreExtension *core = psc->core; + + pdraw = Xmalloc(sizeof(*pdraw)); + if (!pdraw) + return NULL; + + pdraw->destroyDrawable = dri2DestroyDrawable; + pdraw->xDrawable = xDrawable; + pdraw->drawable = drawable; + pdraw->psc = psc; + + fprintf(stderr, "calling DRI2CreateDrawable, XID 0x%lx, GLX ID 0x%lx\n", + xDrawable, drawable); + + if (!DRI2CreateDrawable(psc->dpy, xDrawable, &handle, &head)) { + Xfree(pdraw); + return NULL; + } + + fprintf(stderr, "success, head 0x%x, handle 0x%x\n", head, handle); + + /* Create a new drawable */ + pdraw->driDrawable = + (*core->createNewDrawable)(psc->__driScreen, + config->driConfig, + handle, + head, + pdraw); + + if (!pdraw->driDrawable) { + DRI2DestroyDrawable(psc->dpy, drawable); + Xfree(pdraw); + return NULL; + } + + return pdraw; +} + +static void dri2DestroyScreen(__GLXscreenConfigs *psc) +{ + /* Free the direct rendering per screen data */ + (*psc->core->destroyScreen)(psc->__driScreen); + drmClose(psc->fd); + psc->__driScreen = NULL; +} + + +static void dri2ReemitDrawableInfo(__DRIdrawable *draw, unsigned int *tail, + void *loaderPrivate) +{ + __GLXDRIdrawable *pdraw = loaderPrivate; + + DRI2ReemitDrawableInfo(pdraw->psc->dpy, pdraw->drawable, tail); +} + +static void dri2PostDamage(__DRIdrawable *draw, + struct drm_clip_rect *rects, + int numRects, void *loaderPrivate) +{ + XRectangle *xrects; + XserverRegion region; + __GLXDRIdrawable *glxDraw = loaderPrivate; + __GLXscreenConfigs *psc = glxDraw->psc; + Display *dpy = psc->dpy; + int i; + + xrects = malloc(sizeof(XRectangle) * numRects); + if (xrects == NULL) + return; + + for (i = 0; i < numRects; i++) { + xrects[i].x = rects[i].x1; + xrects[i].y = rects[i].y1; + xrects[i].width = rects[i].x2 - rects[i].x1; + xrects[i].height = rects[i].y2 - rects[i].y1; + } + region = XFixesCreateRegion(dpy, xrects, numRects); + free(xrects); + XDamageAdd(dpy, glxDraw->xDrawable, region); + XFixesDestroyRegion(dpy, region); +} + +static const __DRIloaderExtension dri2LoaderExtension = { + { __DRI_LOADER, __DRI_LOADER_VERSION }, + dri2ReemitDrawableInfo, + dri2PostDamage +}; + +static const __DRIextension *loader_extensions[] = { + &dri2LoaderExtension.base, + &systemTimeExtension.base, + NULL +}; + +static __GLXDRIscreen *dri2CreateScreen(__GLXscreenConfigs *psc, int screen, + __GLXdisplayPrivate *priv) +{ + const __DRIconfig **driver_configs; + const __DRIextension **extensions; + __GLXDRIscreen *psp; + unsigned int sareaHandle; + char *driverName, *busID; + drm_magic_t magic; + int i; + + psp = Xmalloc(sizeof *psp); + if (psp == NULL) + return NULL; + + /* Initialize per screen dynamic client GLX extensions */ + psc->ext_list_first_time = GL_TRUE; + + if (!DRI2Connect(psc->dpy, screen, &driverName, &busID, &sareaHandle)) + return NULL; + + psc->driver = driOpenDriver(driverName); + if (psc->driver == NULL) + goto handle_error; + + extensions = dlsym(psc->driver, __DRI_DRIVER_EXTENSIONS); + if (extensions == NULL) { + ErrorMessageF("driver exports no extensions (%s)\n", dlerror()); + goto handle_error; + } + + for (i = 0; extensions[i]; i++) { + if (strcmp(extensions[i]->name, __DRI_CORE) == 0) + psc->core = (__DRIcoreExtension *) extensions[i]; + } + + if (psc->core == NULL) { + ErrorMessageF("core dri extension not found\n"); + goto handle_error; + } + + psc->fd = drmOpen(NULL, busID); + if (psc->fd < 0) { + ErrorMessageF("failed to open drm device: %s\n", strerror(errno)); + return NULL; + } + + if (drmGetMagic(psc->fd, &magic)) + return NULL; + + if (!DRI2AuthConnection(psc->dpy, screen, magic)) { + ErrorMessageF("failed to authenticate drm access\n"); + return NULL; + } + + psc->__driScreen = + psc->core->createNewScreen(screen, psc->fd, sareaHandle, + loader_extensions, &driver_configs, psc); + if (psc->__driScreen == NULL) { + ErrorMessageF("failed to create dri screen\n"); + return NULL; + } + + driBindExtensions(psc); + + psc->configs = driConvertConfigs(psc->core, psc->configs, driver_configs); + psc->visuals = driConvertConfigs(psc->core, psc->visuals, driver_configs); + + psp->destroyScreen = dri2DestroyScreen; + psp->createContext = dri2CreateContext; + psp->createDrawable = dri2CreateDrawable; + + Xfree(driverName); + Xfree(busID); + + return psp; + + handle_error: + Xfree(driverName); + Xfree(busID); + + /* FIXME: clean up here */ + + return NULL; +} + +/* Called from __glXFreeDisplayPrivate. + */ +static void dri2DestroyDisplay(__GLXDRIdisplay *dpy) +{ + Xfree(dpy); +} + +/* + * Allocate, initialize and return a __DRIdisplayPrivate object. + * This is called from __glXInitialize() when we are given a new + * display pointer. + */ +_X_HIDDEN __GLXDRIdisplay *dri2CreateDisplay(Display *dpy) +{ + __GLXDRIdisplayPrivate *pdp; + int eventBase, errorBase; + + if (!DRI2QueryExtension(dpy, &eventBase, &errorBase)) + return NULL; + + pdp = Xmalloc(sizeof *pdp); + if (pdp == NULL) + return NULL; + + if (!DRI2QueryVersion(dpy, &pdp->driMajor, &pdp->driMinor)) { + Xfree(pdp); + return NULL; + } + + pdp->driPatch = 0; + + pdp->base.destroyDisplay = dri2DestroyDisplay; + pdp->base.createScreen = dri2CreateScreen; + + return &pdp->base; +} + +#endif /* GLX_DIRECT_RENDERING */ diff --git a/src/glx/x11/dri_common.c b/src/glx/x11/dri_common.c new file mode 100644 index 0000000000..b159d193a5 --- /dev/null +++ b/src/glx/x11/dri_common.c @@ -0,0 +1,399 @@ +/* + * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. + * Copyright © 2008 Red Hat, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Soft- + * ware"), to deal in the Software without restriction, including without + * limitation the rights to use, copy, modify, merge, publish, distribute, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, provided that the above copyright + * notice(s) and this permission notice appear in all copies of the Soft- + * ware and that both the above copyright notice(s) and this permission + * notice appear in supporting documentation. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- + * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY + * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN + * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- + * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, + * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER + * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- + * MANCE OF THIS SOFTWARE. + * + * Except as contained in this notice, the name of a copyright holder shall + * not be used in advertising or otherwise to promote the sale, use or + * other dealings in this Software without prior written authorization of + * the copyright holder. + * + * Authors: + * Kevin E. Martin + * Brian Paul + * Kristian Høgsberg (krh@redhat.com) + */ + +#ifdef GLX_DIRECT_RENDERING + +#include +#include +#include "glheader.h" +#include "glxclient.h" +#include "glcontextmodes.h" +#include "dri_common.h" + +#ifndef RTLD_NOW +#define RTLD_NOW 0 +#endif +#ifndef RTLD_GLOBAL +#define RTLD_GLOBAL 0 +#endif + +_X_HIDDEN void InfoMessageF(const char *f, ...) +{ + va_list args; + const char *env; + + if ((env = getenv("LIBGL_DEBUG")) && strstr(env, "verbose")) { + fprintf(stderr, "libGL: "); + va_start(args, f); + vfprintf(stderr, f, args); + va_end(args); + } +} + +/** + * Print error to stderr, unless LIBGL_DEBUG=="quiet". + */ +_X_HIDDEN void ErrorMessageF(const char *f, ...) +{ + va_list args; + const char *env; + + if ((env = getenv("LIBGL_DEBUG")) && !strstr(env, "quiet")) { + fprintf(stderr, "libGL error: "); + va_start(args, f); + vfprintf(stderr, f, args); + va_end(args); + } +} + +#ifndef DEFAULT_DRIVER_DIR +/* this is normally defined in Mesa/configs/default with DRI_DRIVER_SEARCH_PATH */ +#define DEFAULT_DRIVER_DIR "/usr/local/lib/dri" +#endif + +/** + * Try to \c dlopen the named driver. + * + * This function adds the "_dri.so" suffix to the driver name and searches the + * directories specified by the \c LIBGL_DRIVERS_PATH environment variable in + * order to find the driver. + * + * \param driverName - a name like "tdfx", "i810", "mga", etc. + * + * \returns + * A handle from \c dlopen, or \c NULL if driver file not found. + */ +_X_HIDDEN void *driOpenDriver(const char *driverName) +{ + void *glhandle, *handle; + const char *libPaths, *p, *next; + char realDriverName[200]; + int len; + + /* Attempt to make sure libGL symbols will be visible to the driver */ + glhandle = dlopen("libGL.so.1", RTLD_NOW | RTLD_GLOBAL); + + libPaths = NULL; + if (geteuid() == getuid()) { + /* don't allow setuid apps to use LIBGL_DRIVERS_PATH */ + libPaths = getenv("LIBGL_DRIVERS_PATH"); + if (!libPaths) + libPaths = getenv("LIBGL_DRIVERS_DIR"); /* deprecated */ + } + if (libPaths == NULL) + libPaths = DEFAULT_DRIVER_DIR; + + handle = NULL; + for (p = libPaths; *p; p = next) { + next = strchr(p, ':'); + if (next == NULL) { + len = strlen(p); + next = p + len; + } else { + len = next - p; + next++; + } + +#ifdef GLX_USE_TLS + snprintf(realDriverName, sizeof realDriverName, + "%.*s/tls/%s_dri.so", len, p, driverName); + InfoMessageF("OpenDriver: trying %s\n", realDriverName); + handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); +#endif + + if ( handle == NULL ) { + snprintf(realDriverName, sizeof realDriverName, + "%.*s/%s_dri.so", len, p, driverName); + InfoMessageF("OpenDriver: trying %s\n", realDriverName); + handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); + } + + if ( handle != NULL ) + break; + else + ErrorMessageF("dlopen %s failed (%s)\n", realDriverName, dlerror()); + } + + if (!handle) + ErrorMessageF("unable to load driver: %s_dri.so\n", driverName); + + if (glhandle) + dlclose(glhandle); + + return handle; +} + +_X_HIDDEN const __DRIsystemTimeExtension systemTimeExtension = { + { __DRI_SYSTEM_TIME, __DRI_SYSTEM_TIME_VERSION }, + __glXGetUST, + __driGetMscRateOML +}; + +#define __ATTRIB(attrib, field) \ + { attrib, offsetof(__GLcontextModes, field) } + +static const struct { unsigned int attrib, offset; } attribMap[] = { + __ATTRIB(__DRI_ATTRIB_BUFFER_SIZE, rgbBits), + __ATTRIB(__DRI_ATTRIB_LEVEL, level), + __ATTRIB(__DRI_ATTRIB_RED_SIZE, redBits), + __ATTRIB(__DRI_ATTRIB_GREEN_SIZE, greenBits), + __ATTRIB(__DRI_ATTRIB_BLUE_SIZE, blueBits), + __ATTRIB(__DRI_ATTRIB_ALPHA_SIZE, alphaBits), + __ATTRIB(__DRI_ATTRIB_DEPTH_SIZE, depthBits), + __ATTRIB(__DRI_ATTRIB_STENCIL_SIZE, stencilBits), + __ATTRIB(__DRI_ATTRIB_ACCUM_RED_SIZE, accumRedBits), + __ATTRIB(__DRI_ATTRIB_ACCUM_GREEN_SIZE, accumGreenBits), + __ATTRIB(__DRI_ATTRIB_ACCUM_BLUE_SIZE, accumBlueBits), + __ATTRIB(__DRI_ATTRIB_ACCUM_ALPHA_SIZE, accumAlphaBits), + __ATTRIB(__DRI_ATTRIB_SAMPLE_BUFFERS, sampleBuffers), + __ATTRIB(__DRI_ATTRIB_SAMPLES, samples), + __ATTRIB(__DRI_ATTRIB_DOUBLE_BUFFER, doubleBufferMode), + __ATTRIB(__DRI_ATTRIB_STEREO, stereoMode), + __ATTRIB(__DRI_ATTRIB_AUX_BUFFERS, numAuxBuffers), +#if 0 + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_TYPE, transparentPixel), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_INDEX_VALUE, transparentIndex), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_RED_VALUE, transparentRed), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_GREEN_VALUE, transparentGreen), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_BLUE_VALUE, transparentBlue), + __ATTRIB(__DRI_ATTRIB_TRANSPARENT_ALPHA_VALUE, transparentAlpha), + __ATTRIB(__DRI_ATTRIB_RED_MASK, redMask), + __ATTRIB(__DRI_ATTRIB_GREEN_MASK, greenMask), + __ATTRIB(__DRI_ATTRIB_BLUE_MASK, blueMask), + __ATTRIB(__DRI_ATTRIB_ALPHA_MASK, alphaMask), +#endif + __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_WIDTH, maxPbufferWidth), + __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_HEIGHT, maxPbufferHeight), + __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_PIXELS, maxPbufferPixels), + __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_WIDTH, optimalPbufferWidth), + __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_HEIGHT, optimalPbufferHeight), +#if 0 + __ATTRIB(__DRI_ATTRIB_SWAP_METHOD, swapMethod), +#endif + __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGB, bindToTextureRgb), + __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGBA, bindToTextureRgba), + __ATTRIB(__DRI_ATTRIB_BIND_TO_MIPMAP_TEXTURE, bindToMipmapTexture), + __ATTRIB(__DRI_ATTRIB_YINVERTED, yInverted), +}; + +#define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0])) + +static int +scalarEqual(__GLcontextModes *mode, unsigned int attrib, unsigned int value) +{ + unsigned int glxValue; + int i; + + for (i = 0; i < ARRAY_SIZE(attribMap); i++) + if (attribMap[i].attrib == attrib) { + glxValue = *(unsigned int *) ((char *) mode + attribMap[i].offset); + return glxValue == GLX_DONT_CARE || glxValue == value; + } + + return GL_TRUE; /* Is a non-existing attribute equal to value? */ +} + +static int +driConfigEqual(const __DRIcoreExtension *core, + __GLcontextModes *modes, const __DRIconfig *driConfig) +{ + unsigned int attrib, value, glxValue; + int i; + + i = 0; + while (core->indexConfigAttrib(driConfig, i++, &attrib, &value)) { + switch (attrib) { + case __DRI_ATTRIB_RENDER_TYPE: + glxValue = 0; + if (value & __DRI_ATTRIB_RGBA_BIT) { + glxValue |= GLX_RGBA_BIT; + } else if (value & __DRI_ATTRIB_COLOR_INDEX_BIT) { + glxValue |= GLX_COLOR_INDEX_BIT; + } + if (glxValue != modes->renderType) + return GL_FALSE; + break; + + case __DRI_ATTRIB_CONFIG_CAVEAT: + if (value & __DRI_ATTRIB_NON_CONFORMANT_CONFIG) + glxValue = GLX_NON_CONFORMANT_CONFIG; + else if (value & __DRI_ATTRIB_SLOW_BIT) + glxValue = GLX_SLOW_CONFIG; + else + glxValue = GLX_NONE; + if (glxValue != modes->visualRating) + return GL_FALSE; + break; + + case __DRI_ATTRIB_BIND_TO_TEXTURE_TARGETS: + glxValue = 0; + if (value & __DRI_ATTRIB_TEXTURE_1D_BIT) + glxValue |= GLX_TEXTURE_1D_BIT_EXT; + if (value & __DRI_ATTRIB_TEXTURE_2D_BIT) + glxValue |= GLX_TEXTURE_2D_BIT_EXT; + if (value & __DRI_ATTRIB_TEXTURE_RECTANGLE_BIT) + glxValue |= GLX_TEXTURE_RECTANGLE_BIT_EXT; + if (modes->bindToTextureTargets != GLX_DONT_CARE && + glxValue != modes->bindToTextureTargets) + return GL_FALSE; + break; + + default: + if (!scalarEqual(modes, attrib, value)) + return GL_FALSE; + } + } + + return GL_TRUE; +} + +static __GLcontextModes * +createDriMode(const __DRIcoreExtension *core, + __GLcontextModes *modes, const __DRIconfig **driConfigs) +{ + __GLXDRIconfigPrivate *config; + int i; + + for (i = 0; driConfigs[i]; i++) { + if (driConfigEqual(core, modes, driConfigs[i])) + break; + } + + if (driConfigs[i] == NULL) + return NULL; + + config = Xmalloc(sizeof *config); + if (config == NULL) + return NULL; + + config->modes = *modes; + config->driConfig = driConfigs[i]; + + return &config->modes; +} + +_X_HIDDEN __GLcontextModes * +driConvertConfigs(const __DRIcoreExtension *core, + __GLcontextModes *modes, const __DRIconfig **configs) +{ + __GLcontextModes head, *tail, *m; + + tail = &head; + head.next = NULL; + for (m = modes; m; m = m->next) { + tail->next = createDriMode(core, m, configs); + if (tail->next == NULL) { + /* no matching dri config for m */ + continue; + } + + + tail = tail->next; + } + + _gl_context_modes_destroy(modes); + + return head.next; +} + +_X_HIDDEN void +driBindExtensions(__GLXscreenConfigs *psc) +{ + const __DRIextension **extensions; + int i; + + extensions = psc->core->getExtensions(psc->__driScreen); + + for (i = 0; extensions[i]; i++) { +#ifdef __DRI_COPY_SUB_BUFFER + if (strcmp(extensions[i]->name, __DRI_COPY_SUB_BUFFER) == 0) { + psc->copySubBuffer = (__DRIcopySubBufferExtension *) extensions[i]; + __glXEnableDirectExtension(psc, "GLX_MESA_copy_sub_buffer_bit"); + } +#endif + +#ifdef __DRI_SWAP_CONTROL + if (strcmp(extensions[i]->name, __DRI_SWAP_CONTROL) == 0) { + psc->swapControl = (__DRIswapControlExtension *) extensions[i]; + __glXEnableDirectExtension(psc, "GLX_SGI_swap_control"); + __glXEnableDirectExtension(psc, "GLX_MESA_swap_control"); + } +#endif + +#ifdef __DRI_ALLOCATE + if (strcmp(extensions[i]->name, __DRI_ALLOCATE) == 0) { + psc->allocate = (__DRIallocateExtension *) extensions[i]; + __glXEnableDirectExtension(psc, "GLX_MESA_allocate_memory"); + } +#endif + +#ifdef __DRI_FRAME_TRACKING + if (strcmp(extensions[i]->name, __DRI_FRAME_TRACKING) == 0) { + psc->frameTracking = (__DRIframeTrackingExtension *) extensions[i]; + __glXEnableDirectExtension(psc, "GLX_MESA_swap_frame_usage"); + } +#endif + +#ifdef __DRI_MEDIA_STREAM_COUNTER + if (strcmp(extensions[i]->name, __DRI_MEDIA_STREAM_COUNTER) == 0) { + psc->msc = (__DRImediaStreamCounterExtension *) extensions[i]; + __glXEnableDirectExtension(psc, "GLX_SGI_video_sync"); + } +#endif + +#ifdef __DRI_SWAP_BUFFER_COUNTER + /* No driver supports this at this time and the extension is + * not defined in dri_interface.h. Will enable + * GLX_OML_sync_control if implemented. */ +#endif + +#ifdef __DRI_READ_DRAWABLE + if (strcmp(extensions[i]->name, __DRI_READ_DRAWABLE) == 0) { + __glXEnableDirectExtension(psc, "GLX_SGI_make_current_read"); + } +#endif + +#ifdef __DRI_TEX_BUFFER + if (strcmp(extensions[i]->name, __DRI_TEX_BUFFER) == 0) { + psc->texBuffer = (__DRItexBufferExtension *) extensions[i]; + __glXEnableDirectExtension(psc, "GLX_EXT_texture_from_pixmap"); + } +#endif + + /* Ignore unknown extensions */ + } +} + +#endif /* GLX_DIRECT_RENDERING */ diff --git a/src/glx/x11/dri_common.h b/src/glx/x11/dri_common.h new file mode 100644 index 0000000000..3556510335 --- /dev/null +++ b/src/glx/x11/dri_common.h @@ -0,0 +1,60 @@ +/* + * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. + * Copyright © 2008 Red Hat, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Soft- + * ware"), to deal in the Software without restriction, including without + * limitation the rights to use, copy, modify, merge, publish, distribute, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, provided that the above copyright + * notice(s) and this permission notice appear in all copies of the Soft- + * ware and that both the above copyright notice(s) and this permission + * notice appear in supporting documentation. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- + * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY + * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN + * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- + * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, + * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER + * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- + * MANCE OF THIS SOFTWARE. + * + * Except as contained in this notice, the name of a copyright holder shall + * not be used in advertising or otherwise to promote the sale, use or + * other dealings in this Software without prior written authorization of + * the copyright holder. + * + * Authors: + * Kevin E. Martin + * Brian Paul + * Kristian Høgsberg (krh@redhat.com) + */ + +#ifndef _DRI_COMMON_H +#define _DRI_COMMON_H + +typedef struct __GLXDRIconfigPrivateRec __GLXDRIconfigPrivate; + +struct __GLXDRIconfigPrivateRec { + __GLcontextModes modes; + const __DRIconfig *driConfig; +}; + +extern __GLcontextModes * +driConvertConfigs(const __DRIcoreExtension *core, + __GLcontextModes *modes, const __DRIconfig **configs); + +extern const __DRIsystemTimeExtension systemTimeExtension; + +extern void InfoMessageF(const char *f, ...); + +extern void ErrorMessageF(const char *f, ...); + +extern void *driOpenDriver(const char *driverName); + +extern void driBindExtensions(__GLXscreenConfigs *psc); + +#endif /* _DRI_COMMON_H */ diff --git a/src/glx/x11/glxcurrent.c b/src/glx/x11/glxcurrent.c new file mode 100644 index 0000000000..ad648fd438 --- /dev/null +++ b/src/glx/x11/glxcurrent.c @@ -0,0 +1,510 @@ +/* +** License Applicability. Except to the extent portions of this file are +** made subject to an alternative license as permitted in the SGI Free +** Software License B, Version 1.1 (the "License"), the contents of this +** file are subject only to the provisions of the License. You may not use +** this file except in compliance with the License. You may obtain a copy +** of the License at Silicon Graphics, Inc., attn: Legal Services, 1600 +** Amphitheatre Parkway, Mountain View, CA 94043-1351, or at: +** +** http://oss.sgi.com/projects/FreeB +** +** Note that, as provided in the License, the Software is distributed on an +** "AS IS" basis, with ALL EXPRESS AND IMPLIED WARRANTIES AND CONDITIONS +** DISCLAIMED, INCLUDING, WITHOUT LIMITATION, ANY IMPLIED WARRANTIES AND +** CONDITIONS OF MERCHANTABILITY, SATISFACTORY QUALITY, FITNESS FOR A +** PARTICULAR PURPOSE, AND NON-INFRINGEMENT. +** +** Original Code. The Original Code is: OpenGL Sample Implementation, +** Version 1.2.1, released January 26, 2000, developed by Silicon Graphics, +** Inc. The Original Code is Copyright (c) 1991-2000 Silicon Graphics, Inc. +** Copyright in any portions created by third parties is as indicated +** elsewhere herein. All Rights Reserved. +** +** Additional Notice Provisions: The application programming interfaces +** established by SGI in conjunction with the Original Code are The +** OpenGL(R) Graphics System: A Specification (Version 1.2.1), released +** April 1, 1999; The OpenGL(R) Graphics System Utility Library (Version +** 1.3), released November 4, 1998; and OpenGL(R) Graphics with the X +** Window System(R) (Version 1.3), released October 19, 1998. This software +** was created using the OpenGL(R) version 1.2.1 Sample Implementation +** published by SGI, but has not been independently verified as being +** compliant with the OpenGL(R) version 1.2.1 Specification. +** +*/ + +/** + * \file glxcurrent.c + * Client-side GLX interface for current context management. + */ + +#include "glxclient.h" +#include "glapi.h" +#include "glheader.h" +#include "indirect_init.h" + +#ifdef GLX_DIRECT_RENDERING +#include "xf86dri.h" +#endif + +/* +** We setup some dummy structures here so that the API can be used +** even if no context is current. +*/ + +static GLubyte dummyBuffer[__GLX_BUFFER_LIMIT_SIZE]; + +/* +** Dummy context used by small commands when there is no current context. +** All the +** gl and glx entry points are designed to operate as nop's when using +** the dummy context structure. +*/ +static __GLXcontext dummyContext = { + &dummyBuffer[0], + &dummyBuffer[0], + &dummyBuffer[0], + &dummyBuffer[__GLX_BUFFER_LIMIT_SIZE], + sizeof(dummyBuffer), +}; + + +/* +** All indirect rendering contexts will share the same indirect dispatch table. +*/ +static __GLapi *IndirectAPI = NULL; + + +/* + * Current context management and locking + */ + +#if defined( USE_XTHREADS ) + +/* thread safe */ +static GLboolean TSDinitialized = GL_FALSE; +static xthread_key_t ContextTSD; + +_X_HIDDEN __GLXcontext *__glXGetCurrentContext(void) +{ + if (!TSDinitialized) { + xthread_key_create(&ContextTSD, NULL); + TSDinitialized = GL_TRUE; + return &dummyContext; + } + else { + void *p; + xthread_get_specific(ContextTSD, &p); + if (!p) + return &dummyContext; + else + return (__GLXcontext *) p; + } +} + +_X_HIDDEN void __glXSetCurrentContext(__GLXcontext *c) +{ + if (!TSDinitialized) { + xthread_key_create(&ContextTSD, NULL); + TSDinitialized = GL_TRUE; + } + xthread_set_specific(ContextTSD, c); +} + + +/* Used by the __glXLock() and __glXUnlock() macros */ +_X_HIDDEN xmutex_rec __glXmutex; + +#elif defined( PTHREADS ) + +_X_HIDDEN pthread_mutex_t __glXmutex = PTHREAD_MUTEX_INITIALIZER; + +# if defined( GLX_USE_TLS ) + +/** + * Per-thread GLX context pointer. + * + * \c __glXSetCurrentContext is written is such a way that this pointer can + * \b never be \c NULL. This is important! Because of this + * \c __glXGetCurrentContext can be implemented as trivial macro. + */ +__thread void * __glX_tls_Context __attribute__((tls_model("initial-exec"))) + = &dummyContext; + +_X_HIDDEN void __glXSetCurrentContext( __GLXcontext * c ) +{ + __glX_tls_Context = (c != NULL) ? c : &dummyContext; +} + +# else + +static pthread_once_t once_control = PTHREAD_ONCE_INIT; + +/** + * Per-thread data key. + * + * Once \c init_thread_data has been called, the per-thread data key will + * take a value of \c NULL. As each new thread is created the default + * value, in that thread, will be \c NULL. + */ +static pthread_key_t ContextTSD; + +/** + * Initialize the per-thread data key. + * + * This function is called \b exactly once per-process (not per-thread!) to + * initialize the per-thread data key. This is ideally done using the + * \c pthread_once mechanism. + */ +static void init_thread_data( void ) +{ + if ( pthread_key_create( & ContextTSD, NULL ) != 0 ) { + perror( "pthread_key_create" ); + exit( -1 ); + } +} + +_X_HIDDEN void __glXSetCurrentContext( __GLXcontext * c ) +{ + pthread_once( & once_control, init_thread_data ); + pthread_setspecific( ContextTSD, c ); +} + +_X_HIDDEN __GLXcontext * __glXGetCurrentContext( void ) +{ + void * v; + + pthread_once( & once_control, init_thread_data ); + + v = pthread_getspecific( ContextTSD ); + return (v == NULL) ? & dummyContext : (__GLXcontext *) v; +} + +# endif /* defined( GLX_USE_TLS ) */ + +#elif defined( THREADS ) + +#error Unknown threading method specified. + +#else + +/* not thread safe */ +_X_HIDDEN __GLXcontext *__glXcurrentContext = &dummyContext; + +#endif + + +_X_HIDDEN void __glXSetCurrentContextNull(void) +{ + __glXSetCurrentContext(&dummyContext); +#ifdef GLX_DIRECT_RENDERING + _glapi_set_dispatch(NULL); /* no-op functions */ +#endif +} + + +/************************************************************************/ + +PUBLIC GLXContext glXGetCurrentContext(void) +{ + GLXContext cx = __glXGetCurrentContext(); + + if (cx == &dummyContext) { + return NULL; + } else { + return cx; + } +} + +PUBLIC GLXDrawable glXGetCurrentDrawable(void) +{ + GLXContext gc = __glXGetCurrentContext(); + return gc->currentDrawable; +} + + +/************************************************************************/ + +/** + * Sends a GLX protocol message to the specified display to make the context + * and the drawables current. + * + * \param dpy Display to send the message to. + * \param opcode Major opcode value for the display. + * \param gc_id Context tag for the context to be made current. + * \param draw Drawable ID for the "draw" drawable. + * \param read Drawable ID for the "read" drawable. + * \param reply Space to store the X-server's reply. + * + * \warning + * This function assumes that \c dpy is locked with \c LockDisplay on entry. + */ +static Bool SendMakeCurrentRequest(Display *dpy, CARD8 opcode, + GLXContextID gc_id, GLXContextTag gc_tag, + GLXDrawable draw, GLXDrawable read, + xGLXMakeCurrentReply *reply) +{ + Bool ret; + + + LockDisplay(dpy); + + if (draw == read) { + xGLXMakeCurrentReq *req; + + GetReq(GLXMakeCurrent,req); + req->reqType = opcode; + req->glxCode = X_GLXMakeCurrent; + req->drawable = draw; + req->context = gc_id; + req->oldContextTag = gc_tag; + } + else { + __GLXdisplayPrivate *priv = __glXInitialize(dpy); + + /* If the server can support the GLX 1.3 version, we should + * perfer that. Not only that, some servers support GLX 1.3 but + * not the SGI extension. + */ + + if ((priv->majorVersion > 1) || (priv->minorVersion >= 3)) { + xGLXMakeContextCurrentReq *req; + + GetReq(GLXMakeContextCurrent,req); + req->reqType = opcode; + req->glxCode = X_GLXMakeContextCurrent; + req->drawable = draw; + req->readdrawable = read; + req->context = gc_id; + req->oldContextTag = gc_tag; + } + else { + xGLXVendorPrivateWithReplyReq *vpreq; + xGLXMakeCurrentReadSGIReq *req; + + GetReqExtra(GLXVendorPrivateWithReply, + sz_xGLXMakeCurrentReadSGIReq-sz_xGLXVendorPrivateWithReplyReq,vpreq); + req = (xGLXMakeCurrentReadSGIReq *)vpreq; + req->reqType = opcode; + req->glxCode = X_GLXVendorPrivateWithReply; + req->vendorCode = X_GLXvop_MakeCurrentReadSGI; + req->drawable = draw; + req->readable = read; + req->context = gc_id; + req->oldContextTag = gc_tag; + } + } + + ret = _XReply(dpy, (xReply*) reply, 0, False); + + UnlockDisplay(dpy); + SyncHandle(); + + return ret; +} + + +#ifdef GLX_DIRECT_RENDERING +static __GLXDRIdrawable * +FetchDRIDrawable(Display *dpy, + GLXDrawable glxDrawable, GLXContext gc, Bool pre13) +{ + __GLXdisplayPrivate * const priv = __glXInitialize(dpy); + __GLXDRIdrawable *pdraw; + __GLXscreenConfigs *psc; + XID drawable; + + if (priv == NULL) + return NULL; + + psc = &priv->screenConfigs[gc->screen]; + if (psc->drawHash == NULL) + return NULL; + + if (__glxHashLookup(psc->drawHash, glxDrawable, (void *) &pdraw) == 0) + return pdraw; + + /* If this is glXMakeCurrent (pre GLX 1.3) we allow creating the + * GLX drawable on the fly. Otherwise we pass None as the X + * drawable */ + if (pre13) + drawable = glxDrawable; + else + drawable = None; + + pdraw = psc->driScreen->createDrawable(psc, drawable, + glxDrawable, gc->mode); + if (__glxHashInsert(psc->drawHash, glxDrawable, pdraw)) { + (*pdraw->destroyDrawable)(pdraw); + return NULL; + } + + return pdraw; +} +#endif /* GLX_DIRECT_RENDERING */ + + +/** + * Make a particular context current. + * + * \note This is in this file so that it can access dummyContext. + */ +static Bool MakeContextCurrent(Display *dpy, GLXDrawable draw, + GLXDrawable read, GLXContext gc, + Bool pre13) +{ + xGLXMakeCurrentReply reply; + const GLXContext oldGC = __glXGetCurrentContext(); + const CARD8 opcode = __glXSetupForCommand(dpy); + const CARD8 oldOpcode = ((gc == oldGC) || (oldGC == &dummyContext)) + ? opcode : __glXSetupForCommand(oldGC->currentDpy); + Bool bindReturnValue; + + + if (!opcode || !oldOpcode) { + return GL_FALSE; + } + + /* Make sure that the new context has a nonzero ID. In the request, + * a zero context ID is used only to mean that we bind to no current + * context. + */ + if ((gc != NULL) && (gc->xid == None)) { + return GL_FALSE; + } + + _glapi_check_multithread(); + +#ifdef GLX_DIRECT_RENDERING + /* Bind the direct rendering context to the drawable */ + if (gc && gc->driContext) { + __GLXDRIdrawable *pdraw = FetchDRIDrawable(dpy, draw, gc, pre13); + __GLXDRIdrawable *pread = FetchDRIDrawable(dpy, read, gc, pre13); + + bindReturnValue = + (gc->driContext->bindContext) (gc->driContext, pdraw, pread); + } else +#endif + { + /* Send a glXMakeCurrent request to bind the new context. */ + bindReturnValue = + SendMakeCurrentRequest(dpy, opcode, gc ? gc->xid : None, + ((dpy != oldGC->currentDpy) || oldGC->isDirect) + ? None : oldGC->currentContextTag, + draw, read, &reply); + } + + + if (!bindReturnValue) { + return False; + } + + if ((dpy != oldGC->currentDpy || (gc && gc->driContext)) && + !oldGC->isDirect && oldGC != &dummyContext) { + xGLXMakeCurrentReply dummy_reply; + + /* We are either switching from one dpy to another and have to + * send a request to the previous dpy to unbind the previous + * context, or we are switching away from a indirect context to + * a direct context and have to send a request to the dpy to + * unbind the previous context. + */ + (void) SendMakeCurrentRequest(oldGC->currentDpy, oldOpcode, None, + oldGC->currentContextTag, None, None, + & dummy_reply); + } +#ifdef GLX_DIRECT_RENDERING + else if (oldGC->driContext) { + oldGC->driContext->unbindContext(oldGC->driContext); + } +#endif + + + /* Update our notion of what is current */ + __glXLock(); + if (gc == oldGC) { + /* Even though the contexts are the same the drawable might have + * changed. Note that gc cannot be the dummy, and that oldGC + * cannot be NULL, therefore if they are the same, gc is not + * NULL and not the dummy. + */ + gc->currentDrawable = draw; + gc->currentReadable = read; + } else { + if (oldGC != &dummyContext) { + /* Old current context is no longer current to anybody */ + oldGC->currentDpy = 0; + oldGC->currentDrawable = None; + oldGC->currentReadable = None; + oldGC->currentContextTag = 0; + + if (oldGC->xid == None) { + /* We are switching away from a context that was + * previously destroyed, so we need to free the memory + * for the old handle. + */ +#ifdef GLX_DIRECT_RENDERING + /* Destroy the old direct rendering context */ + if (oldGC->driContext) { + oldGC->driContext->destroyContext(oldGC->driContext, + oldGC->psc, + oldGC->createDpy); + oldGC->driContext = NULL; + } +#endif + __glXFreeContext(oldGC); + } + } + if (gc) { + __glXSetCurrentContext(gc); + + gc->currentDpy = dpy; + gc->currentDrawable = draw; + gc->currentReadable = read; + + if (!gc->driContext) { + if (!IndirectAPI) + IndirectAPI = __glXNewIndirectAPI(); + _glapi_set_dispatch(IndirectAPI); + +#ifdef GLX_USE_APPLEGL + do { + extern void XAppleDRIUseIndirectDispatch(void); + XAppleDRIUseIndirectDispatch(); + } while (0); +#endif + + __GLXattribute *state = + (__GLXattribute *)(gc->client_state_private); + + gc->currentContextTag = reply.contextTag; + if (state->array_state == NULL) { + (void) glGetString(GL_EXTENSIONS); + (void) glGetString(GL_VERSION); + __glXInitVertexArrayState(gc); + } + } + else { + gc->currentContextTag = -1; + } + } else { + __glXSetCurrentContextNull(); + } + } + __glXUnlock(); + return GL_TRUE; +} + + +PUBLIC Bool glXMakeCurrent(Display *dpy, GLXDrawable draw, GLXContext gc) +{ + return MakeContextCurrent(dpy, draw, draw, gc, True); +} + +PUBLIC GLX_ALIAS(Bool, glXMakeCurrentReadSGI, + (Display *dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx), + (dpy, d, r, ctx, False), MakeContextCurrent) + +PUBLIC GLX_ALIAS(Bool, glXMakeContextCurrent, + (Display *dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx), + (dpy, d, r, ctx, False), MakeContextCurrent) diff --git a/src/glx/x11/glxhash.c b/src/glx/x11/glxhash.c new file mode 100644 index 0000000000..9ed0429032 --- /dev/null +++ b/src/glx/x11/glxhash.c @@ -0,0 +1,416 @@ +/* glxhash.c -- Small hash table support for integer -> integer mapping + * Taken from libdrm. + * + * Created: Sun Apr 18 09:35:45 1999 by faith@precisioninsight.com + * + * Copyright 1999 Precision Insight, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER + * DEALINGS IN THE SOFTWARE. + * + * Authors: Rickard E. (Rik) Faith + * + * DESCRIPTION + * + * This file contains a straightforward implementation of a fixed-sized + * hash table using self-organizing linked lists [Knuth73, pp. 398-399] for + * collision resolution. There are two potentially interesting things + * about this implementation: + * + * 1) The table is power-of-two sized. Prime sized tables are more + * traditional, but do not have a significant advantage over power-of-two + * sized table, especially when double hashing is not used for collision + * resolution. + * + * 2) The hash computation uses a table of random integers [Hanson97, + * pp. 39-41]. + * + * FUTURE ENHANCEMENTS + * + * With a table size of 512, the current implementation is sufficient for a + * few hundred keys. Since this is well above the expected size of the + * tables for which this implementation was designed, the implementation of + * dynamic hash tables was postponed until the need arises. A common (and + * naive) approach to dynamic hash table implementation simply creates a + * new hash table when necessary, rehashes all the data into the new table, + * and destroys the old table. The approach in [Larson88] is superior in + * two ways: 1) only a portion of the table is expanded when needed, + * distributing the expansion cost over several insertions, and 2) portions + * of the table can be locked, enabling a scalable thread-safe + * implementation. + * + * REFERENCES + * + * [Hanson97] David R. Hanson. C Interfaces and Implementations: + * Techniques for Creating Reusable Software. Reading, Massachusetts: + * Addison-Wesley, 1997. + * + * [Knuth73] Donald E. Knuth. The Art of Computer Programming. Volume 3: + * Sorting and Searching. Reading, Massachusetts: Addison-Wesley, 1973. + * + * [Larson88] Per-Ake Larson. "Dynamic Hash Tables". CACM 31(4), April + * 1988, pp. 446-457. + * + */ + +#include "glxhash.h" +#include + +#define HASH_MAIN 0 + +#include +#include + +#define HASH_MAGIC 0xdeadbeef +#define HASH_DEBUG 0 +#define HASH_SIZE 512 /* Good for about 100 entries */ + /* If you change this value, you probably + have to change the HashHash hashing + function! */ + +#define HASH_ALLOC malloc +#define HASH_FREE free +#define HASH_RANDOM_DECL +#define HASH_RANDOM_INIT(seed) srandom(seed) +#define HASH_RANDOM random() +#define HASH_RANDOM_DESTROY + +typedef struct __glxHashBucket { + unsigned long key; + void *value; + struct __glxHashBucket *next; +} __glxHashBucket, *__glxHashBucketPtr; + +typedef struct __glxHashTable *__glxHashTablePtr; +struct __glxHashTable { + unsigned long magic; + unsigned long entries; + unsigned long hits; /* At top of linked list */ + unsigned long partials; /* Not at top of linked list */ + unsigned long misses; /* Not in table */ + __glxHashBucketPtr buckets[HASH_SIZE]; + int p0; + __glxHashBucketPtr p1; +}; + +static unsigned long HashHash(unsigned long key) +{ + unsigned long hash = 0; + unsigned long tmp = key; + static int init = 0; + static unsigned long scatter[256]; + int i; + + if (!init) { + HASH_RANDOM_DECL; + HASH_RANDOM_INIT(37); + for (i = 0; i < 256; i++) scatter[i] = HASH_RANDOM; + HASH_RANDOM_DESTROY; + ++init; + } + + while (tmp) { + hash = (hash << 1) + scatter[tmp & 0xff]; + tmp >>= 8; + } + + hash %= HASH_SIZE; +#if HASH_DEBUG + printf( "Hash(%d) = %d\n", key, hash); +#endif + return hash; +} + +_X_HIDDEN __glxHashTable *__glxHashCreate(void) +{ + __glxHashTablePtr table; + int i; + + table = HASH_ALLOC(sizeof(*table)); + if (!table) return NULL; + table->magic = HASH_MAGIC; + table->entries = 0; + table->hits = 0; + table->partials = 0; + table->misses = 0; + + for (i = 0; i < HASH_SIZE; i++) table->buckets[i] = NULL; + return table; +} + +_X_HIDDEN int __glxHashDestroy(__glxHashTable *t) +{ + __glxHashTablePtr table = (__glxHashTablePtr)t; + __glxHashBucketPtr bucket; + __glxHashBucketPtr next; + int i; + + if (table->magic != HASH_MAGIC) return -1; /* Bad magic */ + + for (i = 0; i < HASH_SIZE; i++) { + for (bucket = table->buckets[i]; bucket;) { + next = bucket->next; + HASH_FREE(bucket); + bucket = next; + } + } + HASH_FREE(table); + return 0; +} + +/* Find the bucket and organize the list so that this bucket is at the + top. */ + +static __glxHashBucketPtr HashFind(__glxHashTablePtr table, + unsigned long key, unsigned long *h) +{ + unsigned long hash = HashHash(key); + __glxHashBucketPtr prev = NULL; + __glxHashBucketPtr bucket; + + if (h) *h = hash; + + for (bucket = table->buckets[hash]; bucket; bucket = bucket->next) { + if (bucket->key == key) { + if (prev) { + /* Organize */ + prev->next = bucket->next; + bucket->next = table->buckets[hash]; + table->buckets[hash] = bucket; + ++table->partials; + } else { + ++table->hits; + } + return bucket; + } + prev = bucket; + } + ++table->misses; + return NULL; +} + +_X_HIDDEN int __glxHashLookup(__glxHashTable *t, + unsigned long key, void **value) +{ + __glxHashTablePtr table = (__glxHashTablePtr)t; + __glxHashBucketPtr bucket; + + if (!table || table->magic != HASH_MAGIC) return -1; /* Bad magic */ + + bucket = HashFind(table, key, NULL); + if (!bucket) return 1; /* Not found */ + *value = bucket->value; + return 0; /* Found */ +} + +_X_HIDDEN int __glxHashInsert(__glxHashTable *t, + unsigned long key, void *value) +{ + __glxHashTablePtr table = (__glxHashTablePtr)t; + __glxHashBucketPtr bucket; + unsigned long hash; + + if (table->magic != HASH_MAGIC) return -1; /* Bad magic */ + + if (HashFind(table, key, &hash)) return 1; /* Already in table */ + + bucket = HASH_ALLOC(sizeof(*bucket)); + if (!bucket) return -1; /* Error */ + bucket->key = key; + bucket->value = value; + bucket->next = table->buckets[hash]; + table->buckets[hash] = bucket; +#if HASH_DEBUG + printf("Inserted %d at %d/%p\n", key, hash, bucket); +#endif + return 0; /* Added to table */ +} + +_X_HIDDEN int __glxHashDelete(__glxHashTable *t, unsigned long key) +{ + __glxHashTablePtr table = (__glxHashTablePtr)t; + unsigned long hash; + __glxHashBucketPtr bucket; + + if (table->magic != HASH_MAGIC) return -1; /* Bad magic */ + + bucket = HashFind(table, key, &hash); + + if (!bucket) return 1; /* Not found */ + + table->buckets[hash] = bucket->next; + HASH_FREE(bucket); + return 0; +} + +_X_HIDDEN int __glxHashNext(__glxHashTable *t, + unsigned long *key, void **value) +{ + __glxHashTablePtr table = (__glxHashTablePtr)t; + + while (table->p0 < HASH_SIZE) { + if (table->p1) { + *key = table->p1->key; + *value = table->p1->value; + table->p1 = table->p1->next; + return 1; + } + table->p1 = table->buckets[table->p0]; + ++table->p0; + } + return 0; +} + +_X_HIDDEN int __glxHashFirst(__glxHashTable *t, + unsigned long *key, void **value) +{ + __glxHashTablePtr table = (__glxHashTablePtr)t; + + if (table->magic != HASH_MAGIC) return -1; /* Bad magic */ + + table->p0 = 0; + table->p1 = table->buckets[0]; + return __glxHashNext(table, key, value); +} + +#if HASH_MAIN +#define DIST_LIMIT 10 +static int dist[DIST_LIMIT]; + +static void clear_dist(void) { + int i; + + for (i = 0; i < DIST_LIMIT; i++) dist[i] = 0; +} + +static int count_entries(__glxHashBucketPtr bucket) +{ + int count = 0; + + for (; bucket; bucket = bucket->next) ++count; + return count; +} + +static void update_dist(int count) +{ + if (count >= DIST_LIMIT) ++dist[DIST_LIMIT-1]; + else ++dist[count]; +} + +static void compute_dist(__glxHashTablePtr table) +{ + int i; + __glxHashBucketPtr bucket; + + printf("Entries = %ld, hits = %ld, partials = %ld, misses = %ld\n", + table->entries, table->hits, table->partials, table->misses); + clear_dist(); + for (i = 0; i < HASH_SIZE; i++) { + bucket = table->buckets[i]; + update_dist(count_entries(bucket)); + } + for (i = 0; i < DIST_LIMIT; i++) { + if (i != DIST_LIMIT-1) printf("%5d %10d\n", i, dist[i]); + else printf("other %10d\n", dist[i]); + } +} + +static void check_table(__glxHashTablePtr table, + unsigned long key, unsigned long value) +{ + unsigned long retval = 0; + int retcode = __glxHashLookup(table, key, &retval); + + switch (retcode) { + case -1: + printf("Bad magic = 0x%08lx:" + " key = %lu, expected = %lu, returned = %lu\n", + table->magic, key, value, retval); + break; + case 1: + printf("Not found: key = %lu, expected = %lu returned = %lu\n", + key, value, retval); + break; + case 0: + if (value != retval) + printf("Bad value: key = %lu, expected = %lu, returned = %lu\n", + key, value, retval); + break; + default: + printf("Bad retcode = %d: key = %lu, expected = %lu, returned = %lu\n", + retcode, key, value, retval); + break; + } +} + +int main(void) +{ + __glxHashTablePtr table; + int i; + + printf("\n***** 256 consecutive integers ****\n"); + table = __glxHashCreate(); + for (i = 0; i < 256; i++) __glxHashInsert(table, i, i); + for (i = 0; i < 256; i++) check_table(table, i, i); + for (i = 256; i >= 0; i--) check_table(table, i, i); + compute_dist(table); + __glxHashDestroy(table); + + printf("\n***** 1024 consecutive integers ****\n"); + table = __glxHashCreate(); + for (i = 0; i < 1024; i++) __glxHashInsert(table, i, i); + for (i = 0; i < 1024; i++) check_table(table, i, i); + for (i = 1024; i >= 0; i--) check_table(table, i, i); + compute_dist(table); + __glxHashDestroy(table); + + printf("\n***** 1024 consecutive page addresses (4k pages) ****\n"); + table = __glxHashCreate(); + for (i = 0; i < 1024; i++) __glxHashInsert(table, i*4096, i); + for (i = 0; i < 1024; i++) check_table(table, i*4096, i); + for (i = 1024; i >= 0; i--) check_table(table, i*4096, i); + compute_dist(table); + __glxHashDestroy(table); + + printf("\n***** 1024 random integers ****\n"); + table = __glxHashCreate(); + srandom(0xbeefbeef); + for (i = 0; i < 1024; i++) __glxHashInsert(table, random(), i); + srandom(0xbeefbeef); + for (i = 0; i < 1024; i++) check_table(table, random(), i); + srandom(0xbeefbeef); + for (i = 0; i < 1024; i++) check_table(table, random(), i); + compute_dist(table); + __glxHashDestroy(table); + + printf("\n***** 5000 random integers ****\n"); + table = __glxHashCreate(); + srandom(0xbeefbeef); + for (i = 0; i < 5000; i++) __glxHashInsert(table, random(), i); + srandom(0xbeefbeef); + for (i = 0; i < 5000; i++) check_table(table, random(), i); + srandom(0xbeefbeef); + for (i = 0; i < 5000; i++) check_table(table, random(), i); + compute_dist(table); + __glxHashDestroy(table); + + return 0; +} +#endif diff --git a/src/glx/x11/glxhash.h b/src/glx/x11/glxhash.h new file mode 100644 index 0000000000..66012fb889 --- /dev/null +++ b/src/glx/x11/glxhash.h @@ -0,0 +1,16 @@ +#ifndef _GLX_HASH_H_ +#define _GLX_HASH_H_ + + +typedef struct __glxHashTable __glxHashTable; + +/* Hash table routines */ +extern __glxHashTable *__glxHashCreate(void); +extern int __glxHashDestroy(__glxHashTable *t); +extern int __glxHashLookup(__glxHashTable *t, unsigned long key, void **value); +extern int __glxHashInsert(__glxHashTable *t, unsigned long key, void *value); +extern int __glxHashDelete(__glxHashTable *t, unsigned long key); +extern int __glxHashFirst(__glxHashTable *t, unsigned long *key, void **value); +extern int __glxHashNext(__glxHashTable *t, unsigned long *key, void **value); + +#endif /* _GLX_HASH_H_ */ diff --git a/src/glx/x11/indirect_texture_compression.c b/src/glx/x11/indirect_texture_compression.c new file mode 100644 index 0000000000..5676858017 --- /dev/null +++ b/src/glx/x11/indirect_texture_compression.c @@ -0,0 +1,347 @@ +/* + * (C) Copyright IBM Corporation 2004 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file glx_texture_compression.c + * Contains the routines required to implement GLX protocol for + * ARB_texture_compression and related extensions. + * + * \sa http://oss.sgi.com/projects/ogl-sample/registry/ARB/texture_compression.txt + * + * \author Ian Romanick + */ + +#include "packrender.h" +#include "packsingle.h" +#include "indirect.h" + +#include + + +void +__indirect_glGetCompressedTexImageARB( GLenum target, GLint level, + GLvoid * img ) +{ + __GLX_SINGLE_DECLARE_VARIABLES(); + xGLXGetTexImageReply reply; + size_t image_bytes; + + __GLX_SINGLE_LOAD_VARIABLES(); + __GLX_SINGLE_BEGIN( X_GLsop_GetCompressedTexImage, 8 ); + __GLX_SINGLE_PUT_LONG( 0, target ); + __GLX_SINGLE_PUT_LONG( 4, level ); + __GLX_SINGLE_READ_XREPLY(); + + image_bytes = reply.width; + assert( image_bytes <= ((4 * reply.length) - 0) ); + assert( image_bytes >= ((4 * reply.length) - 3) ); + + if ( image_bytes != 0 ) { + _XRead( dpy, (char *) img, image_bytes ); + if ( image_bytes < (4 * reply.length) ) { + _XEatData( dpy, (4 * reply.length) - image_bytes ); + } + } + + __GLX_SINGLE_END(); +} + + +/** + * Internal function used for \c glCompressedTexImage1D and + * \c glCompressedTexImage2D. + */ +static void +CompressedTexImage1D2D( GLenum target, GLint level, + GLenum internal_format, + GLsizei width, GLsizei height, + GLint border, GLsizei image_size, + const GLvoid *data, CARD32 rop ) +{ + __GLX_DECLARE_VARIABLES(); + + __GLX_LOAD_VARIABLES(); + if ( gc->currentDpy == NULL ) { + return; + } + + if ( (target == GL_PROXY_TEXTURE_1D) + || (target == GL_PROXY_TEXTURE_2D) + || (target == GL_PROXY_TEXTURE_CUBE_MAP) ) { + compsize = 0; + } + else { + compsize = image_size; + } + + cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE + + compsize ); + if ( cmdlen <= gc->maxSmallRenderCommandSize ) { + __GLX_BEGIN_VARIABLE( rop, cmdlen ); + __GLX_PUT_LONG( 4, target ); + __GLX_PUT_LONG( 8, level ); + __GLX_PUT_LONG( 12, internal_format ); + __GLX_PUT_LONG( 16, width ); + __GLX_PUT_LONG( 20, height ); + __GLX_PUT_LONG( 24, border ); + __GLX_PUT_LONG( 28, image_size ); + if ( compsize != 0 ) { + __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE, + data, image_size ); + } + __GLX_END( cmdlen ); + } + else { + assert( compsize != 0 ); + + __GLX_BEGIN_VARIABLE_LARGE( rop, cmdlen + 4 ); + __GLX_PUT_LONG( 8, target ); + __GLX_PUT_LONG( 12, level ); + __GLX_PUT_LONG( 16, internal_format ); + __GLX_PUT_LONG( 20, width ); + __GLX_PUT_LONG( 24, height ); + __GLX_PUT_LONG( 28, border ); + __GLX_PUT_LONG( 32, image_size ); + __glXSendLargeCommand( gc, gc->pc, + __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE + 4, + data, image_size ); + } +} + + +/** + * Internal function used for \c glCompressedTexSubImage1D and + * \c glCompressedTexSubImage2D. + */ +static void +CompressedTexSubImage1D2D( GLenum target, GLint level, + GLsizei xoffset, GLsizei yoffset, + GLsizei width, GLsizei height, + GLenum format, GLsizei image_size, + const GLvoid *data, CARD32 rop ) +{ + __GLX_DECLARE_VARIABLES(); + + __GLX_LOAD_VARIABLES(); + if ( gc->currentDpy == NULL ) { + return; + } + + if ( target == GL_PROXY_TEXTURE_3D ) { + compsize = 0; + } + else { + compsize = image_size; + } + + cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE + + compsize ); + if ( cmdlen <= gc->maxSmallRenderCommandSize ) { + __GLX_BEGIN_VARIABLE( rop, cmdlen ); + __GLX_PUT_LONG( 4, target ); + __GLX_PUT_LONG( 8, level ); + __GLX_PUT_LONG( 12, xoffset ); + __GLX_PUT_LONG( 16, yoffset ); + __GLX_PUT_LONG( 20, width ); + __GLX_PUT_LONG( 24, height ); + __GLX_PUT_LONG( 28, format ); + __GLX_PUT_LONG( 32, image_size ); + if ( compsize != 0 ) { + __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE, + data, image_size ); + } + __GLX_END( cmdlen ); + } + else { + assert( compsize != 0 ); + + __GLX_BEGIN_VARIABLE_LARGE( rop, cmdlen + 4 ); + __GLX_PUT_LONG( 8, target ); + __GLX_PUT_LONG( 12, level ); + __GLX_PUT_LONG( 16, xoffset ); + __GLX_PUT_LONG( 20, yoffset ); + __GLX_PUT_LONG( 24, width ); + __GLX_PUT_LONG( 28, height ); + __GLX_PUT_LONG( 32, format ); + __GLX_PUT_LONG( 36, image_size ); + __glXSendLargeCommand( gc, gc->pc, + __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE + 4, + data, image_size ); + } +} + + +void +__indirect_glCompressedTexImage1DARB( GLenum target, GLint level, + GLenum internal_format, GLsizei width, + GLint border, GLsizei image_size, + const GLvoid *data ) +{ + CompressedTexImage1D2D( target, level, internal_format, width, 0, + border, image_size, data, + X_GLrop_CompressedTexImage1D ); +} + + +void +__indirect_glCompressedTexImage2DARB( GLenum target, GLint level, + GLenum internal_format, + GLsizei width, GLsizei height, + GLint border, GLsizei image_size, + const GLvoid *data ) +{ + CompressedTexImage1D2D( target, level, internal_format, width, height, + border, image_size, data, + X_GLrop_CompressedTexImage2D ); +} + + +void +__indirect_glCompressedTexImage3DARB( GLenum target, GLint level, + GLenum internal_format, + GLsizei width, GLsizei height, GLsizei depth, + GLint border, GLsizei image_size, + const GLvoid *data ) +{ + __GLX_DECLARE_VARIABLES(); + + __GLX_LOAD_VARIABLES(); + if ( gc->currentDpy == NULL ) { + return; + } + + cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE + + image_size ); + if ( cmdlen <= gc->maxSmallRenderCommandSize ) { + __GLX_BEGIN_VARIABLE( X_GLrop_CompressedTexImage3D, cmdlen ); + __GLX_PUT_LONG( 4, target ); + __GLX_PUT_LONG( 8, level ); + __GLX_PUT_LONG( 12, internal_format ); + __GLX_PUT_LONG( 16, width ); + __GLX_PUT_LONG( 20, height ); + __GLX_PUT_LONG( 24, depth ); + __GLX_PUT_LONG( 28, border ); + __GLX_PUT_LONG( 32, image_size ); + if ( image_size != 0 ) { + __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE, + data, image_size ); + } + __GLX_END( cmdlen ); + } + else { + __GLX_BEGIN_VARIABLE_LARGE( X_GLrop_CompressedTexImage3D, + cmdlen + 4 ); + __GLX_PUT_LONG( 8, target ); + __GLX_PUT_LONG( 12, level ); + __GLX_PUT_LONG( 16, internal_format ); + __GLX_PUT_LONG( 20, width ); + __GLX_PUT_LONG( 24, height ); + __GLX_PUT_LONG( 28, depth ); + __GLX_PUT_LONG( 32, border ); + __GLX_PUT_LONG( 36, image_size ); + __glXSendLargeCommand( gc, gc->pc, + __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE + 4, + data, image_size ); + } +} + + +void +__indirect_glCompressedTexSubImage1DARB( GLenum target, GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLsizei image_size, + const GLvoid *data ) +{ + CompressedTexSubImage1D2D( target, level, xoffset, 0, width, 0, + format, image_size, data, + X_GLrop_CompressedTexSubImage1D ); +} + + +void +__indirect_glCompressedTexSubImage2DARB( GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLsizei image_size, + const GLvoid *data ) +{ + CompressedTexSubImage1D2D( target, level, xoffset, yoffset, width, height, + format, image_size, data, + X_GLrop_CompressedTexSubImage2D ); +} + + +void +__indirect_glCompressedTexSubImage3DARB( GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLsizei image_size, + const GLvoid *data ) +{ + __GLX_DECLARE_VARIABLES(); + + __GLX_LOAD_VARIABLES(); + if ( gc->currentDpy == NULL ) { + return; + } + + cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE + + image_size ); + if ( cmdlen <= gc->maxSmallRenderCommandSize ) { + __GLX_BEGIN_VARIABLE( X_GLrop_CompressedTexSubImage3D, cmdlen ); + __GLX_PUT_LONG( 4, target ); + __GLX_PUT_LONG( 8, level ); + __GLX_PUT_LONG( 12, xoffset ); + __GLX_PUT_LONG( 16, yoffset ); + __GLX_PUT_LONG( 20, zoffset ); + __GLX_PUT_LONG( 24, width ); + __GLX_PUT_LONG( 28, height ); + __GLX_PUT_LONG( 32, depth ); + __GLX_PUT_LONG( 36, format ); + __GLX_PUT_LONG( 40, image_size ); + if ( image_size != 0 ) { + __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE, + data, image_size ); + } + __GLX_END( cmdlen ); + } + else { + __GLX_BEGIN_VARIABLE_LARGE( X_GLrop_CompressedTexSubImage3D, + cmdlen + 4 ); + __GLX_PUT_LONG( 8, target ); + __GLX_PUT_LONG( 12, level ); + __GLX_PUT_LONG( 16, xoffset ); + __GLX_PUT_LONG( 20, yoffset ); + __GLX_PUT_LONG( 24, zoffset ); + __GLX_PUT_LONG( 28, width ); + __GLX_PUT_LONG( 32, height ); + __GLX_PUT_LONG( 36, depth ); + __GLX_PUT_LONG( 40, format ); + __GLX_PUT_LONG( 44, image_size ); + __glXSendLargeCommand( gc, gc->pc, + __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE + 4, + data, image_size ); + } +} diff --git a/src/glx/x11/indirect_vertex_array_priv.h b/src/glx/x11/indirect_vertex_array_priv.h new file mode 100644 index 0000000000..ab97dc645f --- /dev/null +++ b/src/glx/x11/indirect_vertex_array_priv.h @@ -0,0 +1,308 @@ +/* + * (C) Copyright IBM Corporation 2004, 2005 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sub license, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * IBM, + * AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, + * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF + * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ + +#ifndef _INDIRECT_VA_PRIVATE_ +#define _INDIRECT_VA_PRIVATE_ + +/** + * \file indirect_va_private.h + * + * \author Ian Romanick + */ + +#include + +#include "glxclient.h" +#include "indirect.h" +#include + + +/** + * State descriptor for a single array of vertex data. + */ +struct array_state { + /** + * Pointer to the application supplied data. + */ + const void * data; + + /** + * Enum representing the type of the application supplied data. + */ + GLenum data_type; + + /** + * Stride value supplied by the application. This value is not used + * internally. It is only kept so that it can be queried by the + * application using glGet*v. + */ + GLsizei user_stride; + + /** + * Calculated size, in bytes, of a single element in the array. This + * is calculated based on \c count and the size of the data type + * represented by \c data_type. + */ + GLsizei element_size; + + /** + * Actual byte-stride from one element to the next. This value will + * be equal to either \c user_stride or \c element_stride. + */ + GLsizei true_stride; + + /** + * Number of data values in each element. + */ + GLint count; + + /** + * "Normalized" data is on the range [0,1] (unsigned) or [-1,1] (signed). + * This is used for mapping integral types to floating point types. + */ + GLboolean normalized; + + /** + * Pre-calculated GLX protocol command header. + */ + uint32_t header[2]; + + /** + * Size of the header data. For simple data, like glColorPointerfv, + * this is 4. For complex data that requires either a count (e.g., + * glWeightfvARB), an index (e.g., glVertexAttrib1fvARB), or a + * selector enum (e.g., glMultiTexCoord2fv) this is 8. + */ + unsigned header_size; + + /** + * Set to \c GL_TRUE if this array is enabled. Otherwise, it is set + * to \c GL_FALSE. + */ + GLboolean enabled; + + /** + * For multi-arrayed data (e.g., texture coordinates, generic vertex + * program attributes, etc.), this specifies which array this is. + */ + unsigned index; + + /** + * Per-array-type key. For most arrays, this will be the GL enum for + * that array (e.g., GL_VERTEX_ARRAY for vertex data, GL_NORMAL_ARRAY + * for normal data, GL_TEXTURE_COORD_ARRAY for texture coordinate data, + * etc.). + */ + GLenum key; + + /** + * If this array can be used with the "classic" \c glDrawArrays protocol, + * this is set to \c GL_TRUE. Otherwise, it is set to \c GL_FALSE. + */ + GLboolean old_DrawArrays_possible; +}; + + +/** + * Array state that is pushed / poped by \c glPushClientAttrib and + * \c glPopClientAttrib. + */ +struct array_stack_state { + /** + * Pointer to the application supplied data. + */ + const void * data; + + /** + * Enum representing the type of the application supplied data. + */ + GLenum data_type; + + /** + * Stride value supplied by the application. This value is not used + * internally. It is only kept so that it can be queried by the + * application using glGet*v. + */ + GLsizei user_stride; + + /** + * Number of data values in each element. + */ + GLint count; + + /** + * Per-array-type key. For most arrays, this will be the GL enum for + * that array (e.g., GL_VERTEX_ARRAY for vertex data, GL_NORMAL_ARRAY + * for normal data, GL_TEXTURE_COORD_ARRAY for texture coordinate data, + * etc.). + */ + GLenum key; + + /** + * For multi-arrayed data (e.g., texture coordinates, generic vertex + * program attributes, etc.), this specifies which array this is. + */ + unsigned index; + + /** + * Set to \c GL_TRUE if this array is enabled. Otherwise, it is set + * to \c GL_FALSE. + */ + GLboolean enabled; +}; + + +/** + * Collection of all the vertex array state. + */ +struct array_state_vector { + /** + * Number of arrays tracked by \c ::arrays. + */ + size_t num_arrays; + + /** + * Array of vertex array state. This array contains all of the valid + * vertex arrays. If a vertex array isn't in this array, then it isn't + * valid. For example, if an implementation does not support + * EXT_fog_coord, there won't be a GL_FOG_COORD_ARRAY entry in this + * array. + */ + struct array_state * arrays; + + /** + * Number of currently enabled client-side arrays. The value of this + * field is only valid if \c array_info_cache_valid is true. + */ + size_t enabled_client_array_count; + + /** + * \name ARRAY_INFO cache. + * + * These fields track the state of the ARRAY_INFO cache. The + * \c array_info_cache_size is the size of the actual data stored in + * \c array_info_cache. \c array_info_cache_buffer_size is the size of + * the buffer. This will always be greater than or equal to + * \c array_info_cache_size. + * + * \note + * There are some bytes of extra data before \c array_info_cache that is + * used to hold the header for RenderLarge commands. This is + * \b not included in \c array_info_cache_size or + * \c array_info_cache_buffer_size. \c array_info_cache_base stores a + * pointer to the true start of the buffer (i.e., what malloc returned). + */ + /*@{*/ + size_t array_info_cache_size; + size_t array_info_cache_buffer_size; + void * array_info_cache; + void * array_info_cache_base; + /*@}*/ + + + /** + * Is the cache of ARRAY_INFO data valid? The cache can become invalid + * when one of several state changes occur. Among these chages are + * modifying the array settings for an enabled array and enabling / + * disabling an array. + */ + GLboolean array_info_cache_valid; + + /** + * Is it possible to use the GL 1.1 / EXT_vertex_arrays protocol? Use + * of this protocol is disabled with really old servers (i.e., servers + * that don't support GL 1.1 or EXT_vertex_arrays) or when an environment + * variable is set. + * + * \todo + * GL 1.1 and EXT_vertex_arrays use identical protocol, but have different + * opcodes for \c glDrawArrays. For servers that advertise one or the + * other, there should be a way to select which opcode to use. + */ + GLboolean old_DrawArrays_possible; + + /** + * Is it possible to use the new GL X.X / ARB_vertex_buffer_object + * protocol? + * + * \todo + * This protocol has not yet been defined by the ARB, but is currently a + * work in progress. This field is a place-holder. + */ + GLboolean new_DrawArrays_possible; + + /** + * Active texture unit set by \c glClientActiveTexture. + * + * \sa __glXGetActiveTextureUnit + */ + unsigned active_texture_unit; + + /** + * Number of supported texture units. Even if ARB_multitexture / + * GL 1.3 are not supported, this will be at least 1. When multitexture + * is supported, this will be the value queried by calling + * \c glGetIntegerv with \c GL_MAX_TEXTURE_UNITS. + * + * \todo + * Investigate if this should be the value of \c GL_MAX_TEXTURE_COORDS + * instead (if GL 2.0 / ARB_fragment_shader / ARB_fragment_program / + * NV_fragment_program are supported). + */ + unsigned num_texture_units; + + /** + * Number of generic vertex program attribs. If GL_ARB_vertex_program + * is not supported, this will be zero. Otherwise it will be the value + * queries by calling \c glGetProgramiv with \c GL_VERTEX_PROGRAM_ARB + * and \c GL_MAX_PROGRAM_ATTRIBS_ARB. + */ + unsigned num_vertex_program_attribs; + + /** + * \n Methods for implementing various GL functions. + * + * These method pointers are only valid \c array_info_cache_valid is set. + * When each function starts, it much check \c array_info_cache_valid. + * If it is not set, it must call \c fill_array_info_cache and call + * the new method. + * + * \sa fill_array_info_cache + * + * \todo + * Write code to plug these functions directly into the dispatch table. + */ + /*@{*/ + void (*DrawArrays)( GLenum, GLint, GLsizei ); + void (*DrawElements)( GLenum mode, GLsizei count, GLenum type, + const GLvoid *indices ); + /*@}*/ + + struct array_stack_state * stack; + unsigned active_texture_unit_stack[ __GL_CLIENT_ATTRIB_STACK_DEPTH ]; + unsigned stack_index; +}; + +#endif /* _INDIRECT_VA_PRIVATE_ */ -- cgit v1.2.3 From f851ba705ae878f62149afa377c755286cf53c85 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Wed, 11 Jun 2008 15:35:12 +0100 Subject: fix legacy DRI --- src/gallium/winsys/dri/intel/intel_screen.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index a36c9407d4..e456c28d0d 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -339,6 +339,8 @@ intelInitDriver(__DRIscreenPrivate * sPriv) intelScreen->deviceID = gDRIPriv->deviceID; + intelScreen->front.cpp = gDRIPriv->cpp; + intelScreen->drmMinor = sPriv->drm_version.minor; intelUpdateScreenRotation(sPriv, intelScreen->sarea); if (0) -- cgit v1.2.3 From d45e99060d531d3a3889bb59fb55875bde05811f Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Wed, 11 Jun 2008 15:50:07 +0100 Subject: fix card_extensions --- src/gallium/winsys/dri/intel/intel_context.c | 17 +++++------------ src/gallium/winsys/dri/intel/intel_screen.h | 2 +- 2 files changed, 6 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index af44018053..9d78e9ebf8 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -141,6 +141,8 @@ static const struct dri_extension brw_extensions[] = { { "GL_SGIX_depth_texture", NULL }, { "GL_ARB_texture_env_crossbar", NULL }, { "GL_EXT_texture_sRGB", NULL}, + { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + { "GL_ARB_pixel_buffer_object", NULL}, { NULL, NULL } }; @@ -148,21 +150,15 @@ static const struct dri_extension arb_oc_extensions[] = { {"GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions}, {NULL, NULL} }; - -static const struct dri_extension ttm_extensions[] = { - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, - {NULL, NULL} -}; #endif /** * Initializes potential list of extensions if ctx == NULL, or actually enables * extensions for a context. */ -void intelInitExtensions(struct st_context *st, GLboolean enable_imaging) +void intelInitExtensions(struct intel_context *intel, GLboolean enable_imaging) { - GLcontext *ctx = st ? st->ctx : NULL; + GLcontext *ctx = intel ? intel->st->ctx : NULL; /* Disable imaging extension until convolution is working in teximage paths. */ enable_imaging = GL_FALSE; @@ -170,9 +166,6 @@ void intelInitExtensions(struct st_context *st, GLboolean enable_imaging) driInitExtensions(ctx, card_extensions, enable_imaging); #if 0 - if (intel == NULL || intel->ttm) - driInitExtensions(ctx, ttm_extensions, GL_FALSE); - if (intel == NULL || (IS_965(intel->intelScreen->deviceID) && intel->intelScreen->drmMinor >= 8)) @@ -327,7 +320,7 @@ intelCreateContext(const __GLcontextModes * visual, intel->st = st_create_context(pipe, visual, st_share); - intelInitExtensions( intel->st, GL_TRUE ); + intelInitExtensions( intel, GL_TRUE ); return GL_TRUE; } diff --git a/src/gallium/winsys/dri/intel/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h index d1389a31fb..5a3ae1dc0f 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.h +++ b/src/gallium/winsys/dri/intel/intel_screen.h @@ -117,6 +117,6 @@ intelCreateContext(const __GLcontextModes * visual, void *sharedContextPrivate); extern void -intelInitExtensions(struct st_context *st, GLboolean enable_imaging); +intelInitExtensions(struct intel_context *intel, GLboolean enable_imaging); #endif -- cgit v1.2.3 From 807f8f177b3a2833806d84a70e71019f8849239f Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 11 Jun 2008 18:46:26 +0100 Subject: draw: preserve specular alpha when flatshading -- may be FOGC --- src/gallium/auxiliary/draw/draw_pipe_flatshade.c | 30 ++++++++++++++++++++++-- 1 file changed, 28 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c index 21a9c3b77f..4741b22d02 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c @@ -40,9 +40,19 @@ struct flat_stage struct draw_stage stage; uint num_color_attribs; - uint color_attribs[4]; /* front/back primary/secondary colors */ + uint color_attribs[2]; /* front/back primary colors */ + + uint num_spec_attribs; + uint spec_attribs[2]; /* front/back secondary colors */ }; +#define COPY_3FV( DST, SRC ) \ +do { \ + (DST)[0] = (SRC)[0]; \ + (DST)[1] = (SRC)[1]; \ + (DST)[2] = (SRC)[2]; \ +} while (0) + static INLINE struct flat_stage * flat_stage(struct draw_stage *stage) @@ -58,10 +68,16 @@ static INLINE void copy_colors( struct draw_stage *stage, { const struct flat_stage *flat = flat_stage(stage); uint i; + for (i = 0; i < flat->num_color_attribs; i++) { const uint attr = flat->color_attribs[i]; COPY_4FV(dst->data[attr], src->data[attr]); } + + for (i = 0; i < flat->num_spec_attribs; i++) { + const uint attr = flat->spec_attribs[i]; + COPY_3FV(dst->data[attr], src->data[attr]); + } } @@ -78,6 +94,12 @@ static INLINE void copy_colors2( struct draw_stage *stage, COPY_4FV(dst0->data[attr], src->data[attr]); COPY_4FV(dst1->data[attr], src->data[attr]); } + + for (i = 0; i < flat->num_spec_attribs; i++) { + const uint attr = flat->spec_attribs[i]; + COPY_3FV(dst0->data[attr], src->data[attr]); + COPY_3FV(dst1->data[attr], src->data[attr]); + } } @@ -164,10 +186,14 @@ static void flatshade_init_state( struct draw_stage *stage ) /* Find which vertex shader outputs are colors, make a list */ flat->num_color_attribs = 0; + flat->num_spec_attribs = 0; for (i = 0; i < vs->info.num_outputs; i++) { if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_COLOR || vs->info.output_semantic_name[i] == TGSI_SEMANTIC_BCOLOR) { - flat->color_attribs[flat->num_color_attribs++] = i; + if (vs->info.output_semantic_index[i] == 0) + flat->color_attribs[flat->num_color_attribs++] = i; + else + flat->spec_attribs[flat->num_spec_attribs++] = i; } } -- cgit v1.2.3 From 2161b0fafcdc16703162dd489d2ec1e7114cce4c Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Wed, 11 Jun 2008 23:48:13 +0100 Subject: draw: don't assume vertex position is in data[0] --- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 25 ++++++++------- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 12 ++++--- src/gallium/auxiliary/draw/draw_pipe_clip.c | 16 +++++----- src/gallium/auxiliary/draw/draw_pipe_cull.c | 8 +++-- src/gallium/auxiliary/draw/draw_pipe_offset.c | 7 +++-- src/gallium/auxiliary/draw/draw_pipe_stipple.c | 5 +-- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 2 ++ src/gallium/auxiliary/draw/draw_pipe_wide_line.c | 9 +++--- src/gallium/auxiliary/draw/draw_pipe_wide_point.c | 9 +++--- src/gallium/auxiliary/draw/draw_private.h | 2 +- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 38 +++++++++++++---------- src/gallium/auxiliary/draw/draw_vs.c | 11 +++++++ src/gallium/auxiliary/draw/draw_vs.h | 1 + src/gallium/auxiliary/draw/draw_vs_aos.c | 8 ++--- src/gallium/auxiliary/draw/draw_vs_varient.c | 4 +++ 15 files changed, 97 insertions(+), 60 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index 634bf067f1..ecdebca5f1 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -78,6 +78,8 @@ struct aaline_stage /** For AA lines, this is the vertex attrib slot for the new texcoords */ uint tex_slot; + /** position, not necessarily output zero */ + uint pos_slot; void *sampler_cso; struct pipe_texture *texture; @@ -520,9 +522,10 @@ aaline_line(struct draw_stage *stage, struct prim_header *header) struct prim_header tri; struct vertex_header *v[8]; uint texPos = aaline->tex_slot; + uint posPos = aaline->pos_slot; float *pos, *tex; - float dx = header->v[1]->data[0][0] - header->v[0]->data[0][0]; - float dy = header->v[1]->data[0][1] - header->v[0]->data[0][1]; + float dx = header->v[1]->data[posPos][0] - header->v[0]->data[posPos][0]; + float dy = header->v[1]->data[posPos][1] - header->v[0]->data[posPos][1]; double a = atan2(dy, dx); float c_a = (float) cos(a), s_a = (float) sin(a); uint i; @@ -549,35 +552,35 @@ aaline_line(struct draw_stage *stage, struct prim_header *header) */ /* new verts */ - pos = v[0]->data[0]; + pos = v[0]->data[posPos]; pos[0] += (-dx * c_a - dy * s_a); pos[1] += (-dx * s_a + dy * c_a); - pos = v[1]->data[0]; + pos = v[1]->data[posPos]; pos[0] += (-dx * c_a - -dy * s_a); pos[1] += (-dx * s_a + -dy * c_a); - pos = v[2]->data[0]; + pos = v[2]->data[posPos]; pos[0] += ( dx * c_a - dy * s_a); pos[1] += ( dx * s_a + dy * c_a); - pos = v[3]->data[0]; + pos = v[3]->data[posPos]; pos[0] += ( dx * c_a - -dy * s_a); pos[1] += ( dx * s_a + -dy * c_a); - pos = v[4]->data[0]; + pos = v[4]->data[posPos]; pos[0] += (-dx * c_a - dy * s_a); pos[1] += (-dx * s_a + dy * c_a); - pos = v[5]->data[0]; + pos = v[5]->data[posPos]; pos[0] += (-dx * c_a - -dy * s_a); pos[1] += (-dx * s_a + -dy * c_a); - pos = v[6]->data[0]; + pos = v[6]->data[posPos]; pos[0] += ( dx * c_a - dy * s_a); pos[1] += ( dx * s_a + dy * c_a); - pos = v[7]->data[0]; + pos = v[7]->data[posPos]; pos[0] += ( dx * c_a - -dy * s_a); pos[1] += ( dx * s_a + -dy * c_a); @@ -653,7 +656,7 @@ aaline_first_line(struct draw_stage *stage, struct prim_header *header) /* update vertex attrib info */ aaline->tex_slot = draw->vs.num_vs_outputs; - assert(aaline->tex_slot > 0); /* output[0] is vertex pos */ + aaline->pos_slot = draw->vs.position_output; /* advertise the extra post-transformed vertex attribute */ draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 96dcdb43d5..87fd303649 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -85,6 +85,7 @@ struct aapoint_stage /** this is the vertex attrib slot for the new texcoords */ uint tex_slot; + uint pos_slot; /* * Currently bound state @@ -570,6 +571,7 @@ aapoint_point(struct draw_stage *stage, struct prim_header *header) struct prim_header tri; struct vertex_header *v[4]; uint texPos = aapoint->tex_slot; + uint pos_slot = aapoint->pos_slot; float radius, *pos, *tex; uint i; float k; @@ -619,19 +621,19 @@ aapoint_point(struct draw_stage *stage, struct prim_header *header) } /* new verts */ - pos = v[0]->data[0]; + pos = v[0]->data[pos_slot]; pos[0] -= radius; pos[1] -= radius; - pos = v[1]->data[0]; + pos = v[1]->data[pos_slot]; pos[0] += radius; pos[1] -= radius; - pos = v[2]->data[0]; + pos = v[2]->data[pos_slot]; pos[0] += radius; pos[1] += radius; - pos = v[3]->data[0]; + pos = v[3]->data[pos_slot]; pos[0] -= radius; pos[1] += radius; @@ -683,6 +685,8 @@ aapoint_first_point(struct draw_stage *stage, struct prim_header *header) aapoint->tex_slot = draw->vs.num_vs_outputs; assert(aapoint->tex_slot > 0); /* output[0] is vertex pos */ + aapoint->pos_slot = draw->vs.position_output; + draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_GENERIC; draw->extra_vp_outputs.semantic_index = aapoint->fs->generic_attrib; draw->extra_vp_outputs.slot = aapoint->tex_slot; diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c index 77ccddac4a..fa10f8efca 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_clip.c +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -113,6 +113,7 @@ static void interp( const struct clipper *clip, const struct vertex_header *in ) { const unsigned nr_attrs = clip->stage.draw->vs.num_vs_outputs; + const unsigned pos_attr = clip->stage.draw->vs.position_output; unsigned j; /* Vertex header. @@ -138,18 +139,17 @@ static void interp( const struct clipper *clip, const float *trans = clip->stage.draw->viewport.translate; const float oow = 1.0f / pos[3]; - dst->data[0][0] = pos[0] * oow * scale[0] + trans[0]; - dst->data[0][1] = pos[1] * oow * scale[1] + trans[1]; - dst->data[0][2] = pos[2] * oow * scale[2] + trans[2]; - dst->data[0][3] = oow; + dst->data[pos_attr][0] = pos[0] * oow * scale[0] + trans[0]; + dst->data[pos_attr][1] = pos[1] * oow * scale[1] + trans[1]; + dst->data[pos_attr][2] = pos[2] * oow * scale[2] + trans[2]; + dst->data[pos_attr][3] = oow; } /* Other attributes - * Note: start at 1 to skip winpos (data[0]) since we just computed - * it above. */ - for (j = 1; j < nr_attrs; j++) { - interp_attr(dst->data[j], t, in->data[j], out->data[j]); + for (j = 0; j < nr_attrs; j++) { + if (j != pos_attr) + interp_attr(dst->data[j], t, in->data[j], out->data[j]); } } diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c index 87aaf1f85b..d0d22a38e0 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_cull.c +++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c @@ -55,10 +55,12 @@ static INLINE struct cull_stage *cull_stage( struct draw_stage *stage ) static void cull_tri( struct draw_stage *stage, struct prim_header *header ) { + const unsigned pos = stage->draw->vs.position_output; + /* Window coords: */ - const float *v0 = header->v[0]->data[0]; - const float *v1 = header->v[1]->data[0]; - const float *v2 = header->v[2]->data[0]; + const float *v0 = header->v[0]->data[pos]; + const float *v1 = header->v[1]->data[pos]; + const float *v2 = header->v[2]->data[pos]; /* edge vectors e = v0 - v2, f = v1 - v2 */ const float ex = v0[0] - v2[0]; diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c index ea6de8c571..8f1650e55c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_offset.c +++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c @@ -62,14 +62,15 @@ static INLINE struct offset_stage *offset_stage( struct draw_stage *stage ) static void do_offset_tri( struct draw_stage *stage, struct prim_header *header ) { + const unsigned pos = stage->draw->vs.position_output; struct offset_stage *offset = offset_stage(stage); float inv_det = 1.0f / header->det; /* Window coords: */ - float *v0 = header->v[0]->data[0]; - float *v1 = header->v[1]->data[0]; - float *v2 = header->v[2]->data[0]; + float *v0 = header->v[0]->data[pos]; + float *v1 = header->v[1]->data[pos]; + float *v2 = header->v[2]->data[pos]; /* edge vectors e = v0 - v2, f = v1 - v2 */ float ex = v0[0] - v2[0]; diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c index 9522b79582..bf0db18a68 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c @@ -119,8 +119,9 @@ stipple_line(struct draw_stage *stage, struct prim_header *header) struct stipple_stage *stipple = stipple_stage(stage); struct vertex_header *v0 = header->v[0]; struct vertex_header *v1 = header->v[1]; - const float *pos0 = v0->data[0]; - const float *pos1 = v1->data[0]; + const unsigned pos = stage->draw->vs.position_output; + const float *pos0 = v0->data[pos]; + const float *pos1 = v1->data[pos]; float start = 0; int state = 0; diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 9e5597c32c..10a1f7df79 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -136,6 +136,8 @@ emit_vertex( struct vbuf_stage *vbuf, * set_buffer is efficient. Consider a special one-shot mode for * translate. */ + /* Note: we really do want data[0] here, not data[pos]: + */ vbuf->translate->set_buffer(vbuf->translate, 0, vertex->data[0], 0); vbuf->translate->run(vbuf->translate, 0, 1, vbuf->vertex_ptr); diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c index 878c9c7169..29649f5787 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c @@ -58,6 +58,7 @@ static void wideline_line( struct draw_stage *stage, struct prim_header *header ) { /*const struct wideline_stage *wide = wideline_stage(stage);*/ + const unsigned pos = stage->draw->vs.position_output; const float half_width = 0.5f * stage->draw->rasterizer->line_width; struct prim_header tri; @@ -67,10 +68,10 @@ static void wideline_line( struct draw_stage *stage, struct vertex_header *v2 = dup_vert(stage, header->v[1], 2); struct vertex_header *v3 = dup_vert(stage, header->v[1], 3); - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; + float *pos0 = v0->data[pos]; + float *pos1 = v1->data[pos]; + float *pos2 = v2->data[pos]; + float *pos3 = v3->data[pos]; const float dx = FABSF(pos0[0] - pos2[0]); const float dy = FABSF(pos0[1] - pos2[1]); diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index df92e3f2d0..d40a07f4ae 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -96,6 +96,7 @@ static void widepoint_point( struct draw_stage *stage, struct prim_header *header ) { const struct widepoint_stage *wide = widepoint_stage(stage); + const unsigned pos = stage->draw->vs.position_output; const boolean sprite = (boolean) stage->draw->rasterizer->point_sprite; float half_size; float left_adj, right_adj, bot_adj, top_adj; @@ -108,10 +109,10 @@ static void widepoint_point( struct draw_stage *stage, struct vertex_header *v2 = dup_vert(stage, header->v[0], 2); struct vertex_header *v3 = dup_vert(stage, header->v[0], 3); - float *pos0 = v0->data[0]; - float *pos1 = v1->data[0]; - float *pos2 = v2->data[0]; - float *pos3 = v3->data[0]; + float *pos0 = v0->data[pos]; + float *pos1 = v1->data[pos]; + float *pos2 = v2->data[pos]; + float *pos3 = v3->data[pos]; /* point size is either per-vertex or fixed size */ if (wide->psize_slot >= 0) { diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 1865601cc0..7bd1e670b4 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -177,7 +177,7 @@ struct draw_context struct { struct draw_vertex_shader *vertex_shader; uint num_vs_outputs; /**< convenience, from vertex_shader */ - + uint position_output; /** TGSI program interpreter runtime state */ struct tgsi_exec_machine machine; diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index c4a67c8289..af6306b1c6 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -100,16 +100,19 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs, struct vertex_header *out = vertices; const float *scale = pvs->draw->viewport.scale; const float *trans = pvs->draw->viewport.translate; + const unsigned pos = pvs->draw->vs.position_output; unsigned clipped = 0; unsigned j; if (0) debug_printf("%s\n"); for (j = 0; j < count; j++) { - out->clip[0] = out->data[0][0]; - out->clip[1] = out->data[0][1]; - out->clip[2] = out->data[0][2]; - out->clip[3] = out->data[0][3]; + float *position = out->data[pos]; + + out->clip[0] = position[0]; + out->clip[1] = position[1]; + out->clip[2] = position[2]; + out->clip[3] = position[3]; out->vertex_id = 0xffff; out->clipmask = compute_clipmask_gl(out->clip, @@ -120,19 +123,19 @@ static boolean post_vs_cliptest_viewport_gl( struct pt_post_vs *pvs, if (out->clipmask == 0) { /* divide by w */ - float w = 1.0f / out->data[0][3]; + float w = 1.0f / position[3]; /* Viewport mapping */ - out->data[0][0] = out->data[0][0] * w * scale[0] + trans[0]; - out->data[0][1] = out->data[0][1] * w * scale[1] + trans[1]; - out->data[0][2] = out->data[0][2] * w * scale[2] + trans[2]; - out->data[0][3] = w; + position[0] = position[0] * w * scale[0] + trans[0]; + position[1] = position[1] * w * scale[1] + trans[1]; + position[2] = position[2] * w * scale[2] + trans[2]; + position[3] = w; #if 0 debug_printf("post viewport: %f %f %f %f\n", - out->data[0][0], - out->data[0][1], - out->data[0][2], - out->data[0][3]); + position[0], + position[1], + position[2], + position[3]); #endif } @@ -154,15 +157,18 @@ static boolean post_vs_viewport( struct pt_post_vs *pvs, struct vertex_header *out = vertices; const float *scale = pvs->draw->viewport.scale; const float *trans = pvs->draw->viewport.translate; + const unsigned pos = pvs->draw->vs.position_output; unsigned j; if (0) debug_printf("%s\n", __FUNCTION__); for (j = 0; j < count; j++) { + float *position = out->data[pos]; + /* Viewport mapping only, no cliptest/rhw divide */ - out->data[0][0] = out->data[0][0] * scale[0] + trans[0]; - out->data[0][1] = out->data[0][1] * scale[1] + trans[1]; - out->data[0][2] = out->data[0][2] * scale[2] + trans[2]; + position[0] = position[0] * scale[0] + trans[0]; + position[1] = position[1] * scale[1] + trans[1]; + position[2] = position[2] * scale[2] + trans[2]; out = (struct vertex_header *)((char *)out + stride); } diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index 979f9864fd..978954e91c 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -85,6 +85,16 @@ draw_create_vertex_shader(struct draw_context *draw, } } + if (vs) + { + uint i; + for (i = 0; i < vs->info.num_outputs; i++) { + if (vs->info.output_semantic_name[i] == TGSI_SEMANTIC_POSITION && + vs->info.output_semantic_index[i] == 0) + vs->position_output = i; + } + } + assert(vs); return vs; } @@ -100,6 +110,7 @@ draw_bind_vertex_shader(struct draw_context *draw, { draw->vs.vertex_shader = dvs; draw->vs.num_vs_outputs = dvs->info.num_outputs; + draw->vs.position_output = dvs->position_output; dvs->prepare( dvs, draw ); } else { diff --git a/src/gallium/auxiliary/draw/draw_vs.h b/src/gallium/auxiliary/draw/draw_vs.h index 08c6de8ba8..45992d1986 100644 --- a/src/gallium/auxiliary/draw/draw_vs.h +++ b/src/gallium/auxiliary/draw/draw_vs.h @@ -106,6 +106,7 @@ struct draw_vertex_shader { struct pipe_shader_state state; struct tgsi_shader_info info; + unsigned position_output; /* Extracted from shader: */ diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index c47647ea72..5d4a8b38c8 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1686,7 +1686,7 @@ static boolean emit_viewport( struct aos_compilation *cp ) { struct x86_reg pos = aos_get_shader_reg_xmm(cp, TGSI_FILE_OUTPUT, - 0); + cp->vaos->draw->vs.position_output ); struct x86_reg scale = x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, scale)); @@ -1700,7 +1700,7 @@ static boolean emit_viewport( struct aos_compilation *cp ) aos_adopt_xmm_reg( cp, pos, TGSI_FILE_OUTPUT, - 0, + cp->vaos->draw->vs.position_output, TRUE ); return TRUE; } @@ -1715,7 +1715,7 @@ static boolean emit_rhw_viewport( struct aos_compilation *cp ) struct x86_reg tmp = aos_get_xmm_reg(cp); struct x86_reg pos = aos_get_shader_reg_xmm(cp, TGSI_FILE_OUTPUT, - 0); + cp->vaos->draw->vs.position_output); struct x86_reg scale = x86_make_disp(cp->machine_EDX, Offset(struct aos_machine, scale)); @@ -1740,7 +1740,7 @@ static boolean emit_rhw_viewport( struct aos_compilation *cp ) aos_adopt_xmm_reg( cp, pos, TGSI_FILE_OUTPUT, - 0, + cp->vaos->draw->vs.position_output, TRUE ); return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index abe8c5ec2d..ad0b829afa 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -89,6 +89,8 @@ static void do_rhw_viewport( struct draw_vs_varient_generic *vsvg, unsigned stride = vsvg->temp_vertex_stride; unsigned j; + ptr += vsvg->base.vs->position_output * 4 * sizeof(float); + for (j = 0; j < count; j++, ptr += stride) { float *data = (float *)ptr; float w = 1.0f / data[3]; @@ -110,6 +112,8 @@ static void do_viewport( struct draw_vs_varient_generic *vsvg, unsigned stride = vsvg->temp_vertex_stride; unsigned j; + ptr += vsvg->base.vs->position_output * 4 * sizeof(float); + for (j = 0; j < count; j++, ptr += stride) { float *data = (float *)ptr; -- cgit v1.2.3 From 6f7ed99b306990f284f9d57c0b66efaa7f2277e1 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Thu, 12 Jun 2008 00:11:54 +0100 Subject: revert the DRI2 commits --- src/gallium/winsys/dri/intel/intel_context.c | 81 +- src/gallium/winsys/dri/intel/intel_lock.c | 7 +- src/gallium/winsys/dri/intel/intel_screen.c | 375 +++---- src/gallium/winsys/dri/intel/intel_screen.h | 4 +- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 3 +- src/glx/x11/Makefile | 27 +- src/glx/x11/XF86dri.c | 39 +- src/glx/x11/dri2.c | 252 ----- src/glx/x11/dri2.h | 53 - src/glx/x11/dri2_glx.c | 371 ------- src/glx/x11/dri_common.c | 399 ------- src/glx/x11/dri_common.h | 60 -- src/glx/x11/dri_glx.c | 885 +++++++--------- src/glx/x11/glcontextmodes.c | 30 +- src/glx/x11/glcontextmodes.h | 6 +- src/glx/x11/glx_pbuffer.c | 100 +- src/glx/x11/glx_texture_compression.c | 347 +++++++ src/glx/x11/glxclient.h | 171 ++- src/glx/x11/glxcmds.c | 707 +++++++------ src/glx/x11/glxcurrent.c | 510 --------- src/glx/x11/glxext.c | 1198 +++++++++++++++++++--- src/glx/x11/glxextensions.c | 25 +- src/glx/x11/glxextensions.h | 4 +- src/glx/x11/glxhash.c | 416 -------- src/glx/x11/glxhash.h | 16 - src/glx/x11/indirect.c | 353 +------ src/glx/x11/indirect.h | 2 - src/glx/x11/indirect_init.c | 2 - src/glx/x11/indirect_texture_compression.c | 347 ------- src/glx/x11/indirect_vertex_array.c | 24 +- src/glx/x11/indirect_vertex_array_priv.h | 308 ------ src/glx/x11/singlepix.c | 2 +- src/glx/x11/xf86dri.h | 14 +- src/glx/x11/xfont.c | 6 +- src/mesa/drivers/dri/common/dri_util.c | 1036 ++++++++++--------- src/mesa/drivers/dri/common/dri_util.h | 319 +++--- src/mesa/drivers/dri/common/drirenderbuffer.c | 2 + src/mesa/drivers/dri/common/spantmp2.h | 2 +- src/mesa/drivers/dri/common/utils.c | 309 ++---- src/mesa/drivers/dri/common/utils.h | 41 +- src/mesa/drivers/dri/common/vblank.c | 159 ++- src/mesa/drivers/dri/common/vblank.h | 16 +- src/mesa/drivers/dri/common/xmlconfig.c | 26 +- 43 files changed, 3288 insertions(+), 5766 deletions(-) delete mode 100644 src/glx/x11/dri2.c delete mode 100644 src/glx/x11/dri2.h delete mode 100644 src/glx/x11/dri2_glx.c delete mode 100644 src/glx/x11/dri_common.c delete mode 100644 src/glx/x11/dri_common.h create mode 100644 src/glx/x11/glx_texture_compression.c delete mode 100644 src/glx/x11/glxcurrent.c delete mode 100644 src/glx/x11/glxhash.c delete mode 100644 src/glx/x11/glxhash.h delete mode 100644 src/glx/x11/indirect_texture_compression.c delete mode 100644 src/glx/x11/indirect_vertex_array_priv.h (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index 9d78e9ebf8..8284e0edbb 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -28,8 +28,6 @@ #include "i830_dri.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" #include "intel_screen.h" #include "intel_context.h" #include "intel_swapbuffers.h" @@ -38,6 +36,8 @@ #include "i915simple/i915_screen.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" #include "pipe/p_defines.h" #include "pipe/p_context.h" @@ -67,6 +67,7 @@ int __intel_debug = 0; #define need_GL_NV_vertex_program #include "extension_helper.h" + /** * Extension strings exported by the intel driver. * @@ -74,7 +75,7 @@ int __intel_debug = 0; * It appears that ARB_texture_env_crossbar has "disappeared" compared to the * old i830-specific driver. */ -static const struct dri_extension card_extensions[] = { +const struct dri_extension card_extensions[] = { {"GL_ARB_multisample", GL_ARB_multisample_functions}, {"GL_ARB_multitexture", NULL}, {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, @@ -85,27 +86,22 @@ static const struct dri_extension card_extensions[] = { {"GL_ARB_texture_env_combine", NULL}, {"GL_ARB_texture_env_dot3", NULL}, {"GL_ARB_texture_mirrored_repeat", NULL}, - {"GL_ARB_texture_non_power_of_two", NULL }, {"GL_ARB_texture_rectangle", NULL}, - {"GL_NV_texture_rectangle", NULL}, - {"GL_EXT_texture_rectangle", NULL}, - {"GL_ARB_point_parameters", NULL}, {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", - GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_logic_op", NULL}, {"GL_EXT_blend_subtract", NULL}, {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, -#if 1 /* XXX FBO temporary? */ {"GL_EXT_packed_depth_stencil", NULL}, -#endif + {"GL_EXT_pixel_buffer_object", NULL}, {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, {"GL_EXT_stencil_wrap", NULL}, {"GL_EXT_texture_edge_clamp", NULL}, @@ -120,61 +116,10 @@ static const struct dri_extension card_extensions[] = { {"GL_NV_blend_square", NULL}, {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, {"GL_NV_vertex_program1_1", NULL}, - { "GL_SGIS_generate_mipmap", NULL }, + {"GL_SGIS_generate_mipmap", NULL }, {NULL, NULL} }; -#if 0 -static const struct dri_extension brw_extensions[] = { - { "GL_ARB_shading_language_100", GL_VERSION_2_0_functions}, - { "GL_ARB_shading_language_120", GL_VERSION_2_1_functions}, - { "GL_ARB_shader_objects", GL_ARB_shader_objects_functions}, - { "GL_ARB_vertex_shader", GL_ARB_vertex_shader_functions}, - { "GL_ARB_point_sprite", NULL}, - { "GL_ARB_fragment_shader", NULL }, - { "GL_ARB_draw_buffers", NULL }, - { "GL_ARB_depth_texture", NULL }, - { "GL_ARB_fragment_program", NULL }, - { "GL_ARB_shadow", NULL }, - { "GL_EXT_shadow_funcs", NULL }, - /* ARB extn won't work if not enabled */ - { "GL_SGIX_depth_texture", NULL }, - { "GL_ARB_texture_env_crossbar", NULL }, - { "GL_EXT_texture_sRGB", NULL}, - { "GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - { "GL_ARB_pixel_buffer_object", NULL}, - { NULL, NULL } -}; - -static const struct dri_extension arb_oc_extensions[] = { - {"GL_ARB_occlusion_query", GL_ARB_occlusion_query_functions}, - {NULL, NULL} -}; -#endif - -/** - * Initializes potential list of extensions if ctx == NULL, or actually enables - * extensions for a context. - */ -void intelInitExtensions(struct intel_context *intel, GLboolean enable_imaging) -{ - GLcontext *ctx = intel ? intel->st->ctx : NULL; - /* Disable imaging extension until convolution is working in teximage paths. - */ - enable_imaging = GL_FALSE; - - driInitExtensions(ctx, card_extensions, enable_imaging); - -#if 0 - if (intel == NULL || - (IS_965(intel->intelScreen->deviceID) && - intel->intelScreen->drmMinor >= 8)) - driInitExtensions(ctx, arb_oc_extensions, GL_FALSE); - - if (intel == NULL || IS_965(intel->intelScreen->deviceID)) - driInitExtensions(ctx, brw_extensions, GL_FALSE); -#endif -} #ifdef DEBUG @@ -259,10 +204,10 @@ intelCreateContext(const __GLcontextModes * visual, /* * memory pools */ - DRM_LIGHT_LOCK(sPriv->fd, &sPriv->lock, driContextPriv->hHWContext); + DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); // ZZZ JB should be per screen and not be done per context havePools = intelCreatePools(sPriv); - DRM_UNLOCK(sPriv->fd, &sPriv->lock, driContextPriv->hHWContext); + DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); if (!havePools) return GL_FALSE; @@ -270,7 +215,7 @@ intelCreateContext(const __GLcontextModes * visual, /* Dri stuff */ intel->hHWContext = driContextPriv->hHWContext; intel->driFd = sPriv->fd; - intel->driHwLock = (drmLock *) & sPriv->lock; + intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); intel->iw.irq_seq = -1; @@ -320,7 +265,7 @@ intelCreateContext(const __GLcontextModes * visual, intel->st = st_create_context(pipe, visual, st_share); - intelInitExtensions( intel, GL_TRUE ); + driInitExtensions( intel->st->ctx, card_extensions, GL_TRUE ); return GL_TRUE; } diff --git a/src/gallium/winsys/dri/intel/intel_lock.c b/src/gallium/winsys/dri/intel/intel_lock.c index 469090c0e1..406284c98f 100644 --- a/src/gallium/winsys/dri/intel/intel_lock.c +++ b/src/gallium/winsys/dri/intel/intel_lock.c @@ -50,9 +50,6 @@ intelContendedLock(struct intel_context *intel, uint flags) DBG(LOCK, "%s - got contended lock\n", __progname); - if (sPriv->dri2.enabled) - return; - /* If the window moved, may need to set a new cliprect now. * * NOTE: This releases and regains the hw lock, so all state @@ -61,8 +58,8 @@ intelContendedLock(struct intel_context *intel, uint flags) if (dPriv) DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - if (sarea && (sarea->width != intelScreen->front.width || - sarea->height != intelScreen->front.height)) { + if (sarea->width != intelScreen->front.width || + sarea->height != intelScreen->front.height) { intelUpdateScreenRotation(sPriv, sarea); } diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index e456c28d0d..8817a80a5a 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -60,114 +60,7 @@ static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; extern const struct dri_extension card_extensions[]; -static GLboolean -intel_get_param(__DRIscreenPrivate *psp, int param, int *value) -{ - int ret; - struct drm_i915_getparam gp; - - gp.param = param; - gp.value = value; - ret = drmCommandWriteRead(psp->fd, DRM_I915_GETPARAM, &gp, sizeof(gp)); - if (ret) { - fprintf(stderr, "drm_i915_getparam: %d\n", ret); - return GL_FALSE; - } - - return GL_TRUE; -} - -static void -intelSetTexOffset(__DRIcontext *pDRICtx, int texname, - unsigned long long offset, int depth, uint pitch) -{ - abort(); -#if 0 - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct st_texture_object *stObj = st_texture_object(tObj); - - if (!stObj) - return; - - if (stObj->pt) - st->pipe->texture_release(intel->st->pipe, &stObj->pt); - - stObj->imageOverride = GL_TRUE; - stObj->depthOverride = depth; - stObj->pitchOverride = pitch; - - if (offset) - stObj->textureOffset = offset; -#endif -} - - -static void -intelHandleDrawableConfig(__DRIdrawablePrivate *dPriv, - __DRIcontextPrivate *pcp, - __DRIDrawableConfigEvent *event) -{ - (void) dPriv; - (void) pcp; - (void) event; -} - -static void -intelHandleBufferAttach(__DRIdrawablePrivate *dPriv, - __DRIcontextPrivate *pcp, - __DRIBufferAttachEvent *ba) -{ - struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - - switch (ba->buffer.attachment) { - case DRI_DRAWABLE_BUFFER_FRONT_LEFT: - intelScreen->front.width = dPriv->w; - intelScreen->front.height = dPriv->h; - intelScreen->front.cpp = ba->buffer.cpp; - intelScreen->front.pitch = ba->buffer.pitch; - driGenBuffers(intelScreen->base.staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); - driBOSetReferenced(intelScreen->front.buffer, ba->buffer.handle); - break; - - case DRI_DRAWABLE_BUFFER_BACK_LEFT: - case DRI_DRAWABLE_BUFFER_DEPTH: - case DRI_DRAWABLE_BUFFER_STENCIL: - case DRI_DRAWABLE_BUFFER_ACCUM: - /* anything ?? */ - break; - - default: - fprintf(stderr, "unhandled buffer attach event, attachment type %d\n", - ba->buffer.attachment); - return; - } -} - -static const __DRItexOffsetExtension intelTexOffsetExtension = { - { __DRI_TEX_OFFSET }, - intelSetTexOffset, -}; - -#if 0 -static const __DRItexBufferExtension intelTexBufferExtension = { - { __DRI_TEX_BUFFER, __DRI_TEX_BUFFER_VERSION }, - intelSetTexBuffer, -}; -#endif - -static const __DRIextension *intelScreenExtensions[] = { - &driReadDrawableExtension, - &driCopySubBufferExtension.base, - &driSwapControlExtension.base, - &driFrameTrackingExtension.base, - &driMediaStreamCounterExtension.base, - &intelTexOffsetExtension.base, -// &intelTexBufferExtension.base, - NULL -}; static void @@ -284,8 +177,7 @@ intelCreatePools(__DRIscreenPrivate * sPriv) intelScreen->havePools = GL_TRUE; - if (intelScreen->sarea) - intelUpdateScreenRotation(sPriv, intelScreen->sarea); + intelUpdateScreenRotation(sPriv, intelScreen->sarea); return GL_TRUE; } @@ -318,6 +210,11 @@ intelInitDriver(__DRIscreenPrivate * sPriv) struct intel_screen *intelScreen; I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; + PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = + (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> + getProcAddress("glxEnableExtension")); + void *const psc = sPriv->psc->screenConfigs; + if (sPriv->devPrivSize != sizeof(I830DRIRec)) { fprintf(stderr, "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); @@ -334,19 +231,28 @@ intelInitDriver(__DRIscreenPrivate * sPriv) __driConfigOptions, __driNConfigOptions); sPriv->private = (void *) intelScreen; - intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + - gDRIPriv->sarea_priv_offset); + intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + + gDRIPriv->sarea_priv_offset); intelScreen->deviceID = gDRIPriv->deviceID; - intelScreen->front.cpp = gDRIPriv->cpp; - intelScreen->drmMinor = sPriv->drm_version.minor; + intelScreen->drmMinor = sPriv->drmMinor; + + assert(gDRIPriv->bitsPerPixel == 16 || + gDRIPriv->bitsPerPixel == 32); + intelUpdateScreenRotation(sPriv, intelScreen->sarea); if (0) intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); - sPriv->extensions = intelScreenExtensions; + if (glx_enable_extension != NULL) { + (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); + (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); + (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); + } intel_be_init_device(&intelScreen->base, sPriv->fd); intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; @@ -445,19 +351,65 @@ intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) return 0; } -static __DRIconfig ** -intelFillInModes(__DRIscreenPrivate *psp, - unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, GLboolean have_back_buffer) + +static void +intelSetTexOffset(__DRIcontext *pDRICtx, int texname, + unsigned long long offset, int depth, uint pitch) +{ + abort(); +#if 0 + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct st_texture_object *stObj = st_texture_object(tObj); + + if (!stObj) + return; + + if (stObj->pt) + st->pipe->texture_release(intel->st->pipe, &stObj->pt); + + stObj->imageOverride = GL_TRUE; + stObj->depthOverride = depth; + stObj->pitchOverride = pitch; + + if (offset) + stObj->textureOffset = offset; +#endif +} + + +static const struct __DriverAPIRec intelAPI = { + .InitDriver = intelInitDriver, + .DestroyScreen = intelDestroyScreen, + .CreateContext = intelCreateContext, + .DestroyContext = intelDestroyContext, + .CreateBuffer = intelCreateBuffer, + .DestroyBuffer = intelDestroyBuffer, + .SwapBuffers = intelSwapBuffers, + .MakeCurrent = intelMakeCurrent, + .UnbindContext = intelUnbindContext, + .GetSwapInfo = intelGetSwapInfo, + .GetMSC = driGetMSC32, + .WaitForMSC = driWaitForMSC32, + .WaitForSBC = NULL, + .SwapBuffersMSC = NULL, + .CopySubBuffer = intelCopySubBuffer, + .setTexOffset = intelSetTexOffset, +}; + + +static __GLcontextModes * +intelFillInModes(unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, boolean have_back_buffer) { - __DRIconfig **configs; + __GLcontextModes *modes; __GLcontextModes *m; unsigned num_modes; unsigned depth_buffer_factor; unsigned back_buffer_factor; GLenum fb_format; GLenum fb_type; - int i; /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't * support pageflipping at all. @@ -499,143 +451,100 @@ intelFillInModes(__DRIscreenPrivate *psp, fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; } - configs = driCreateConfigs(fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor); - if (configs == NULL) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + modes = + (*dri_interface->createContextModes) (num_modes, + sizeof(__GLcontextModes)); + m = modes; + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_TRUE_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, GLX_DIRECT_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, __LINE__); return NULL; } /* Mark the visual as slow if there are "fake" stencil bits. */ - for (i = 0; configs[i]; i++) { - m = &configs[i]->modes; + for (m = modes; m != NULL; m = m->next) { if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { m->visualRating = GLX_SLOW_CONFIG; } } - return configs; + return modes; } + /** - * This is the driver specific part of the createNewScreen entry point. - * - * \todo maybe fold this into intelInitDriver + * This is the bootstrap function for the driver. libGL supplies all of the + * requisite information about the system, and the driver initializes itself. + * This routine also fills in the linked list pointed to by \c driver_modes + * with the \c __GLcontextModes that the driver can support for windows or + * pbuffers. * - * \return the __GLcontextModes supported by this driver + * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on + * failure. */ -static const __DRIconfig **intelInitScreen(__DRIscreenPrivate *psp) +PUBLIC void * +__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, + __DRIscreen * psc, + const __GLcontextModes * modes, + const __DRIversion * ddx_version, + const __DRIversion * dri_version, + const __DRIversion * drm_version, + const __DRIframebuffer * frame_buffer, + drmAddress pSAREA, int fd, + int internal_api_version, + const __DRIinterfaceMethods * interface, + __GLcontextModes ** driver_modes) { -#ifdef I915 - static const __DRIversion ddx_expected = { 1, 5, 0 }; -#else - static const __DRIversion ddx_expected = { 1, 6, 0 }; -#endif + __DRIscreenPrivate *psp; + static const __DRIversion ddx_expected = { 1, 7, 0 }; static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 5, 0 }; - I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; + static const __DRIversion drm_expected = { 1, 7, 0 }; + + dri_interface = interface; if (!driCheckDriDdxDrmVersions2("i915", - &psp->dri_version, &dri_expected, - &psp->ddx_version, &ddx_expected, - &psp->drm_version, &drm_expected)) { + dri_version, &dri_expected, + ddx_version, &ddx_expected, + drm_version, &drm_expected)) { return NULL; } - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - intelInitExtensions(NULL, GL_TRUE); - - if (!intelInitDriver(psp)) - return NULL; - - psp->extensions = intelScreenExtensions; - - return (const __DRIconfig **) - intelFillInModes(psp, dri_priv->cpp * 8, - (dri_priv->cpp == 2) ? 16 : 24, - (dri_priv->cpp == 2) ? 0 : 8, 1); -} - -/** - * This is the driver specific part of the createNewScreen entry point. - * - * \return the __GLcontextModes supported by this driver - */ -static const -__DRIconfig **intelInitScreen2(__DRIscreenPrivate *psp) -{ - struct intel_screen *intelScreen; - - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create is - * called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - intelInitExtensions(NULL, GL_TRUE); - - /* Allocate the private area */ - intelScreen = CALLOC_STRUCT(intel_screen); - if (!intelScreen) { - fprintf(stderr, "\nERROR! Allocating private area failed\n"); - return GL_FALSE; + psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, + ddx_version, dri_version, drm_version, + frame_buffer, pSAREA, fd, + internal_api_version, &intelAPI); + + if (psp != NULL) { + I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; + *driver_modes = intelFillInModes(dri_priv->cpp * 8, + (dri_priv->cpp == 2) ? 16 : 24, + (dri_priv->cpp == 2) ? 0 : 8, 1); + + /* Calling driInitExtensions here, with a NULL context pointer, + * does not actually enable the extensions. It just makes sure + * that all the dispatch offsets for all the extensions that + * *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create + * is called, but we can't enable the extensions until we have a + * context pointer. + * + * Hello chicken. Hello egg. How are you two today? + */ + driInitExtensions(NULL, card_extensions, GL_FALSE); } - /* parse information in __driConfigOptions */ - driParseOptionInfo(&intelScreen->optionCache, - __driConfigOptions, __driNConfigOptions); - - psp->private = (void *) intelScreen; - - intelScreen->drmMinor = psp->drm_version.minor; - /* Determine chipset ID? */ - if (!intel_get_param(psp, I915_PARAM_CHIPSET_ID, - &intelScreen->deviceID)) - return GL_FALSE; - - psp->extensions = intelScreenExtensions; - - intel_be_init_device(&intelScreen->base, psp->fd); - intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; - intelScreen->base.base.get_name = intel_get_name; - - return driConcatConfigs(intelFillInModes(psp, 16, 16, 0, 1), - intelFillInModes(psp, 32, 24, 8, 1)); + return (void *) psp; } -const struct __DriverAPIRec driDriverAPI = { - .InitScreen = intelInitScreen, - .DestroyScreen = intelDestroyScreen, - .CreateContext = intelCreateContext, - .DestroyContext = intelDestroyContext, - .CreateBuffer = intelCreateBuffer, - .DestroyBuffer = intelDestroyBuffer, - .SwapBuffers = intelSwapBuffers, - .MakeCurrent = intelMakeCurrent, - .UnbindContext = intelUnbindContext, - .GetSwapInfo = intelGetSwapInfo, - .GetDrawableMSC = driDrawableGetMSC32, - .WaitForMSC = driWaitForMSC32, - .CopySubBuffer = intelCopySubBuffer, - - .InitScreen2 = intelInitScreen2, - .HandleDrawableConfig = intelHandleDrawableConfig, - .HandleBufferAttach = intelHandleBufferAttach, -}; diff --git a/src/gallium/winsys/dri/intel/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h index 5a3ae1dc0f..8036917903 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.h +++ b/src/gallium/winsys/dri/intel/intel_screen.h @@ -32,7 +32,7 @@ #include "i830_common.h" #include "xmlconfig.h" #include "intel_drm/ws_dri_bufpool.h" -#include "state_tracker/st_context.h" + #include "pipe/p_compiler.h" #include "intel_drm/intel_be_device.h" @@ -116,7 +116,5 @@ intelCreateContext(const __GLcontextModes * visual, __DRIcontextPrivate * driContextPriv, void *sharedContextPrivate); -extern void -intelInitExtensions(struct intel_context *intel, GLboolean enable_imaging); #endif diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index 8b17154cf9..7f3babd98e 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -78,8 +78,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, * current context (which is what intelScreenContext should return) might * not get a contended lock and thus cliprects not updated (tests/manywin) */ - if (!intel->driScreen->dri2.enabled && - intel_context(dPriv->driContextPriv) != intel) + if (intel_context(dPriv->driContextPriv) != intel) DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); diff --git a/src/glx/x11/Makefile b/src/glx/x11/Makefile index bb4d3cc5ee..b404727f08 100644 --- a/src/glx/x11/Makefile +++ b/src/glx/x11/Makefile @@ -10,14 +10,12 @@ SOURCES = \ compsize.c \ eval.c \ glxcmds.c \ - glxcurrent.c \ glxext.c \ glxextensions.c \ indirect.c \ indirect_init.c \ indirect_size.c \ indirect_window_pos.c \ - indirect_texture_compression.c \ indirect_transpose_matrix.c \ indirect_vertex_array.c \ indirect_vertex_program.c \ @@ -31,16 +29,13 @@ SOURCES = \ xfont.c \ glx_pbuffer.c \ glx_query.c \ - dri_common.c \ + glx_texture_compression.c \ dri_glx.c \ - XF86dri.c \ - glxhash.c \ - dri2_glx.c \ - dri2.c + XF86dri.c include $(TOP)/src/mesa/sources -MESA_ASM_API = $(addprefix $(TOP)/src/mesa/, $(ASM_API)) +MESA_GLAPI_ASM_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_ASM_SOURCES)) MESA_GLAPI_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_SOURCES)) MESA_GLAPI_OBJECTS = $(addprefix $(TOP)/src/mesa/, $(GLAPI_OBJECTS)) @@ -50,10 +45,9 @@ INCLUDES = -I. \ -I$(TOP)/include \ -I$(TOP)/include/GL/internal \ -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/glapi \ -I$(TOP)/src/mesa/main \ + -I$(TOP)/src/mesa/glapi \ $(LIBDRM_CFLAGS) \ - $(DRI2PROTO_CFLAGS) \ $(X11_INCLUDES) @@ -71,28 +65,29 @@ default: depend $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) # Make libGL $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME): $(OBJECTS) Makefile - $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' -ldflags '$(LDFLAGS)' \ + $(TOP)/bin/mklib -o $(GL_LIB) -linker '$(CC)' \ -major 1 -minor 2 $(MKLIB_OPTIONS) \ -install $(TOP)/$(LIB_DIR) $(GL_LIB_DEPS) $(OBJECTS) -depend: $(SOURCES) $(MESA_GLAPI_SOURCES) $(MESA_ASM_API) Makefile +depend: $(SOURCES) $(MESA_GLAPI_SOURCES) $(MESA_GLAPI_ASM_SOURCES) Makefile + rm -f depend touch depend $(MKDEP) $(MKDEP_OPTIONS) $(INCLUDES) $(SOURCES) \ - $(MESA_GLAPI_SOURCES) $(MESA_ASM_API) + $(MESA_GLAPI_SOURCES) $(MESA_GLAPI_ASM_SOURCES) # Emacs tags tags: etags `find . -name \*.[ch]` `find $(TOP)/include` -install: $(TOP)/$(LIB_DIR)/$(GL_LIB_NAME) - $(MAKE) -C $(TOP)/src/mesa install-libgl +# Dummy install target +install: # Remove .o and backup files clean: -rm -f $(TOP)/$(LIB_DIR)/libGL.so* -rm -f *.o *~ - -rm -f depend depend.bak + -rm -f depend include depend diff --git a/src/glx/x11/XF86dri.c b/src/glx/x11/XF86dri.c index ba38949c0b..9919a40977 100644 --- a/src/glx/x11/XF86dri.c +++ b/src/glx/x11/XF86dri.c @@ -374,9 +374,10 @@ PUBLIC Bool XF86DRICreateContext(dpy, screen, visual, context, hHWContext) context, hHWContext ); } -PUBLIC GLboolean XF86DRIDestroyContext(Display *dpy, int screen, - XID context ) +PUBLIC GLboolean XF86DRIDestroyContext( __DRInativeDisplay * ndpy, int screen, + __DRIid context ) { + Display * const dpy = (Display *) ndpy; XExtDisplayInfo *info = find_display (dpy); xXF86DRIDestroyContextReq *req; @@ -395,9 +396,10 @@ PUBLIC GLboolean XF86DRIDestroyContext(Display *dpy, int screen, return True; } -PUBLIC GLboolean XF86DRICreateDrawable(Display *dpy, int screen, - XID drawable, drm_drawable_t * hHWDrawable ) +PUBLIC GLboolean XF86DRICreateDrawable( __DRInativeDisplay * ndpy, int screen, + __DRIid drawable, drm_drawable_t * hHWDrawable ) { + Display * const dpy = (Display *) ndpy; XExtDisplayInfo *info = find_display (dpy); xXF86DRICreateDrawableReply rep; xXF86DRICreateDrawableReq *req; @@ -424,36 +426,16 @@ PUBLIC GLboolean XF86DRICreateDrawable(Display *dpy, int screen, return True; } -static int noopErrorHandler(Display *dpy, XErrorEvent *xerr) -{ - return 0; -} - -PUBLIC GLboolean XF86DRIDestroyDrawable(Display *dpy, int screen, - XID drawable ) +PUBLIC GLboolean XF86DRIDestroyDrawable( __DRInativeDisplay * ndpy, int screen, + __DRIid drawable ) { + Display * const dpy = (Display *) ndpy; XExtDisplayInfo *info = find_display (dpy); xXF86DRIDestroyDrawableReq *req; - int (*oldXErrorHandler)(Display *, XErrorEvent *); TRACE("DestroyDrawable..."); XF86DRICheckExtension (dpy, info, False); - /* This is called from the DRI driver, which used call it like this - * - * if (windowExists(drawable)) - * destroyDrawable(drawable); - * - * which is a textbook race condition - the window may disappear - * from the server between checking for its existance and - * destroying it. Instead we change the semantics of - * __DRIinterfaceMethodsRec::destroyDrawable() to succeed even if - * the windows is gone, by wrapping the destroy call in an error - * handler. */ - - XSync(dpy, GL_FALSE); - oldXErrorHandler = XSetErrorHandler(noopErrorHandler); - LockDisplay(dpy); GetReq(XF86DRIDestroyDrawable, req); req->reqType = info->codes->major_opcode; @@ -462,9 +444,6 @@ PUBLIC GLboolean XF86DRIDestroyDrawable(Display *dpy, int screen, req->drawable = drawable; UnlockDisplay(dpy); SyncHandle(); - - XSetErrorHandler(oldXErrorHandler); - TRACE("DestroyDrawable... return True"); return True; } diff --git a/src/glx/x11/dri2.c b/src/glx/x11/dri2.c deleted file mode 100644 index e7044ab424..0000000000 --- a/src/glx/x11/dri2.c +++ /dev/null @@ -1,252 +0,0 @@ -/* - * Copyright © 2008 Red Hat, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Soft- - * ware"), to deal in the Software without restriction, including without - * limitation the rights to use, copy, modify, merge, publish, distribute, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, provided that the above copyright - * notice(s) and this permission notice appear in all copies of the Soft- - * ware and that both the above copyright notice(s) and this permission - * notice appear in supporting documentation. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- - * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY - * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN - * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- - * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- - * MANCE OF THIS SOFTWARE. - * - * Except as contained in this notice, the name of a copyright holder shall - * not be used in advertising or otherwise to promote the sale, use or - * other dealings in this Software without prior written authorization of - * the copyright holder. - * - * Authors: - * Kristian Høgsberg (krh@redhat.com) - */ - - -#define NEED_REPLIES -#include -#include -#include -#include -#include "glheader.h" -#include "xf86drm.h" -#include "dri2.h" - -static char dri2ExtensionName[] = DRI2_NAME; -static XExtensionInfo *dri2Info; -static XEXT_GENERATE_CLOSE_DISPLAY (DRI2CloseDisplay, dri2Info) -static /* const */ XExtensionHooks dri2ExtensionHooks = { - NULL, /* create_gc */ - NULL, /* copy_gc */ - NULL, /* flush_gc */ - NULL, /* free_gc */ - NULL, /* create_font */ - NULL, /* free_font */ - DRI2CloseDisplay, /* close_display */ - NULL, /* wire_to_event */ - NULL, /* event_to_wire */ - NULL, /* error */ - NULL, /* error_string */ -}; - -static XEXT_GENERATE_FIND_DISPLAY (DRI2FindDisplay, dri2Info, - dri2ExtensionName, - &dri2ExtensionHooks, - 0, NULL) - -Bool DRI2QueryExtension(Display *dpy, int *eventBase, int *errorBase) -{ - XExtDisplayInfo *info = DRI2FindDisplay(dpy); - - if (XextHasExtension(info)) { - *eventBase = info->codes->first_event; - *errorBase = info->codes->first_error; - return True; - } - - return False; -} - -Bool DRI2QueryVersion(Display *dpy, int *major, int *minor) -{ - XExtDisplayInfo *info = DRI2FindDisplay (dpy); - xDRI2QueryVersionReply rep; - xDRI2QueryVersionReq *req; - - XextCheckExtension (dpy, info, dri2ExtensionName, False); - - LockDisplay(dpy); - GetReq(DRI2QueryVersion, req); - req->reqType = info->codes->major_opcode; - req->dri2ReqType = X_DRI2QueryVersion; - req->majorVersion = DRI2_MAJOR; - req->minorVersion = DRI2_MINOR; - if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { - UnlockDisplay(dpy); - SyncHandle(); - return False; - } - *major = rep.majorVersion; - *minor = rep.minorVersion; - UnlockDisplay(dpy); - SyncHandle(); - - return True; -} - -Bool DRI2Connect(Display *dpy, int screen, - char **driverName, char **busId, unsigned int *sareaHandle) -{ - XExtDisplayInfo *info = DRI2FindDisplay(dpy); - xDRI2ConnectReply rep; - xDRI2ConnectReq *req; - - XextCheckExtension (dpy, info, dri2ExtensionName, False); - - LockDisplay(dpy); - GetReq(DRI2Connect, req); - req->reqType = info->codes->major_opcode; - req->dri2ReqType = X_DRI2Connect; - req->screen = screen; - if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { - UnlockDisplay(dpy); - SyncHandle(); - return False; - } - - *sareaHandle = rep.sareaHandle; - - *driverName = Xmalloc(rep.driverNameLength + 1); - if (*driverName == NULL) { - _XEatData(dpy, - ((rep.driverNameLength + 3) & ~3) + - ((rep.busIdLength + 3) & ~3)); - UnlockDisplay(dpy); - SyncHandle(); - return False; - } - _XReadPad(dpy, *driverName, rep.driverNameLength); - (*driverName)[rep.driverNameLength] = '\0'; - - *busId = Xmalloc(rep.busIdLength + 1); - if (*busId == NULL) { - Xfree(*driverName); - _XEatData(dpy, ((rep.busIdLength + 3) & ~3)); - UnlockDisplay(dpy); - SyncHandle(); - return False; - } - _XReadPad(dpy, *busId, rep.busIdLength); - (*busId)[rep.busIdLength] = '\0'; - - UnlockDisplay(dpy); - SyncHandle(); - - return rep.sareaHandle != 0; -} - -Bool DRI2AuthConnection(Display *dpy, int screen, drm_magic_t magic) -{ - XExtDisplayInfo *info = DRI2FindDisplay(dpy); - xDRI2AuthConnectionReq *req; - xDRI2AuthConnectionReply rep; - - XextCheckExtension (dpy, info, dri2ExtensionName, False); - - LockDisplay(dpy); - GetReq(DRI2AuthConnection, req); - req->reqType = info->codes->major_opcode; - req->dri2ReqType = X_DRI2AuthConnection; - req->screen = screen; - req->magic = magic; - rep.authenticated = 0; - if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { - UnlockDisplay(dpy); - SyncHandle(); - return False; - } - UnlockDisplay(dpy); - SyncHandle(); - - return rep.authenticated; -} - -Bool DRI2CreateDrawable(Display *dpy, XID drawable, - unsigned int *handle, unsigned int *head) -{ - XExtDisplayInfo *info = DRI2FindDisplay(dpy); - xDRI2CreateDrawableReply rep; - xDRI2CreateDrawableReq *req; - - XextCheckExtension (dpy, info, dri2ExtensionName, False); - - LockDisplay(dpy); - GetReq(DRI2CreateDrawable, req); - req->reqType = info->codes->major_opcode; - req->dri2ReqType = X_DRI2CreateDrawable; - req->drawable = drawable; - if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { - UnlockDisplay(dpy); - SyncHandle(); - return False; - } - UnlockDisplay(dpy); - SyncHandle(); - - *handle = rep.handle; - *head = rep.head; - - return True; -} - -void DRI2DestroyDrawable(Display *dpy, XID drawable) -{ - XExtDisplayInfo *info = DRI2FindDisplay(dpy); - xDRI2DestroyDrawableReq *req; - - XextSimpleCheckExtension (dpy, info, dri2ExtensionName); - - XSync(dpy, GL_FALSE); - - LockDisplay(dpy); - GetReq(DRI2DestroyDrawable, req); - req->reqType = info->codes->major_opcode; - req->dri2ReqType = X_DRI2DestroyDrawable; - req->drawable = drawable; - UnlockDisplay(dpy); - SyncHandle(); -} - -Bool DRI2ReemitDrawableInfo(Display *dpy, XID drawable, unsigned int *head) -{ - XExtDisplayInfo *info = DRI2FindDisplay(dpy); - xDRI2ReemitDrawableInfoReply rep; - xDRI2ReemitDrawableInfoReq *req; - - XextCheckExtension (dpy, info, dri2ExtensionName, False); - - LockDisplay(dpy); - GetReq(DRI2ReemitDrawableInfo, req); - req->reqType = info->codes->major_opcode; - req->dri2ReqType = X_DRI2ReemitDrawableInfo; - req->drawable = drawable; - if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) { - UnlockDisplay(dpy); - SyncHandle(); - return False; - } - UnlockDisplay(dpy); - SyncHandle(); - - *head = rep.head; - - return True; -} diff --git a/src/glx/x11/dri2.h b/src/glx/x11/dri2.h deleted file mode 100644 index 1dfd0448b2..0000000000 --- a/src/glx/x11/dri2.h +++ /dev/null @@ -1,53 +0,0 @@ -/* - * Copyright © 2007,2008 Red Hat, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Soft- - * ware"), to deal in the Software without restriction, including without - * limitation the rights to use, copy, modify, merge, publish, distribute, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, provided that the above copyright - * notice(s) and this permission notice appear in all copies of the Soft- - * ware and that both the above copyright notice(s) and this permission - * notice appear in supporting documentation. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- - * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY - * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN - * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- - * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- - * MANCE OF THIS SOFTWARE. - * - * Except as contained in this notice, the name of a copyright holder shall - * not be used in advertising or otherwise to promote the sale, use or - * other dealings in this Software without prior written authorization of - * the copyright holder. - * - * Authors: - * Kristian Høgsberg (krh@redhat.com) - */ - -#ifndef _DRI2_H_ -#define _DRI2_H_ - -extern Bool -DRI2QueryExtension(Display *display, int *eventBase, int *errorBase); -extern Bool -DRI2QueryVersion(Display *display, int *major, int *minor); -extern Bool -DRI2Connect(Display *display, int screen, - char **driverName, char **busId, unsigned int *sareaHandle); -extern Bool -DRI2AuthConnection(Display *display, int screen, drm_magic_t magic); -extern Bool -DRI2CreateDrawable(Display *display, XID drawable, - unsigned int *handle, unsigned int *head); -extern void -DRI2DestroyDrawable(Display *display, XID handle); -extern Bool -DRI2ReemitDrawableInfo(Display *dpy, XID handle, unsigned int *head); - -#endif diff --git a/src/glx/x11/dri2_glx.c b/src/glx/x11/dri2_glx.c deleted file mode 100644 index b679c72c10..0000000000 --- a/src/glx/x11/dri2_glx.c +++ /dev/null @@ -1,371 +0,0 @@ -/* - * Copyright © 2008 Red Hat, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Soft- - * ware"), to deal in the Software without restriction, including without - * limitation the rights to use, copy, modify, merge, publish, distribute, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, provided that the above copyright - * notice(s) and this permission notice appear in all copies of the Soft- - * ware and that both the above copyright notice(s) and this permission - * notice appear in supporting documentation. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- - * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY - * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN - * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- - * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- - * MANCE OF THIS SOFTWARE. - * - * Except as contained in this notice, the name of a copyright holder shall - * not be used in advertising or otherwise to promote the sale, use or - * other dealings in this Software without prior written authorization of - * the copyright holder. - * - * Authors: - * Kristian Høgsberg (krh@redhat.com) - */ - -#ifdef GLX_DIRECT_RENDERING - -#include -#include -#include -#include "glheader.h" -#include "glxclient.h" -#include "glcontextmodes.h" -#include "xf86dri.h" -#include "sarea.h" -#include -#include -#include -#include "xf86drm.h" -#include "dri2.h" -#include "dri_common.h" - -typedef struct __GLXDRIdisplayPrivateRec __GLXDRIdisplayPrivate; -typedef struct __GLXDRIcontextPrivateRec __GLXDRIcontextPrivate; - -struct __GLXDRIdisplayPrivateRec { - __GLXDRIdisplay base; - - /* - ** XFree86-DRI version information - */ - int driMajor; - int driMinor; - int driPatch; -}; - -struct __GLXDRIcontextPrivateRec { - __GLXDRIcontext base; - __DRIcontext *driContext; - __GLXscreenConfigs *psc; -}; - -static void dri2DestroyContext(__GLXDRIcontext *context, - __GLXscreenConfigs *psc, Display *dpy) -{ - __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; - const __DRIcoreExtension *core = pcp->psc->core; - - (*core->destroyContext)(pcp->driContext); - - Xfree(pcp); -} - -static Bool dri2BindContext(__GLXDRIcontext *context, - __GLXDRIdrawable *draw, __GLXDRIdrawable *read) -{ - __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; - const __DRIcoreExtension *core = pcp->psc->core; - - return (*core->bindContext)(pcp->driContext, - draw->driDrawable, - read->driDrawable); -} - -static void dri2UnbindContext(__GLXDRIcontext *context) -{ - __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; - const __DRIcoreExtension *core = pcp->psc->core; - - (*core->unbindContext)(pcp->driContext); -} - -static __GLXDRIcontext *dri2CreateContext(__GLXscreenConfigs *psc, - const __GLcontextModes *mode, - GLXContext gc, - GLXContext shareList, int renderType) -{ - __GLXDRIcontextPrivate *pcp, *pcp_shared; - __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode; - const __DRIcoreExtension *core = psc->core; - __DRIcontext *shared = NULL; - - if (shareList) { - pcp_shared = (__GLXDRIcontextPrivate *) shareList->driContext; - shared = pcp_shared->driContext; - } - - pcp = Xmalloc(sizeof *pcp); - if (pcp == NULL) - return NULL; - - pcp->psc = psc; - pcp->driContext = - (*core->createNewContext)(psc->__driScreen, - config->driConfig, shared, pcp); - gc->__driContext = pcp->driContext; - - if (pcp->driContext == NULL) { - Xfree(pcp); - return NULL; - } - - pcp->base.destroyContext = dri2DestroyContext; - pcp->base.bindContext = dri2BindContext; - pcp->base.unbindContext = dri2UnbindContext; - - return &pcp->base; -} - -static void dri2DestroyDrawable(__GLXDRIdrawable *pdraw) -{ - const __DRIcoreExtension *core = pdraw->psc->core; - - (*core->destroyDrawable)(pdraw->driDrawable); - DRI2DestroyDrawable(pdraw->psc->dpy, pdraw->drawable); - Xfree(pdraw); -} - -static __GLXDRIdrawable *dri2CreateDrawable(__GLXscreenConfigs *psc, - XID xDrawable, - GLXDrawable drawable, - const __GLcontextModes *modes) -{ - __GLXDRIdrawable *pdraw; - __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes; - unsigned int handle, head; - const __DRIcoreExtension *core = psc->core; - - pdraw = Xmalloc(sizeof(*pdraw)); - if (!pdraw) - return NULL; - - pdraw->destroyDrawable = dri2DestroyDrawable; - pdraw->xDrawable = xDrawable; - pdraw->drawable = drawable; - pdraw->psc = psc; - - fprintf(stderr, "calling DRI2CreateDrawable, XID 0x%lx, GLX ID 0x%lx\n", - xDrawable, drawable); - - if (!DRI2CreateDrawable(psc->dpy, xDrawable, &handle, &head)) { - Xfree(pdraw); - return NULL; - } - - fprintf(stderr, "success, head 0x%x, handle 0x%x\n", head, handle); - - /* Create a new drawable */ - pdraw->driDrawable = - (*core->createNewDrawable)(psc->__driScreen, - config->driConfig, - handle, - head, - pdraw); - - if (!pdraw->driDrawable) { - DRI2DestroyDrawable(psc->dpy, drawable); - Xfree(pdraw); - return NULL; - } - - return pdraw; -} - -static void dri2DestroyScreen(__GLXscreenConfigs *psc) -{ - /* Free the direct rendering per screen data */ - (*psc->core->destroyScreen)(psc->__driScreen); - drmClose(psc->fd); - psc->__driScreen = NULL; -} - - -static void dri2ReemitDrawableInfo(__DRIdrawable *draw, unsigned int *tail, - void *loaderPrivate) -{ - __GLXDRIdrawable *pdraw = loaderPrivate; - - DRI2ReemitDrawableInfo(pdraw->psc->dpy, pdraw->drawable, tail); -} - -static void dri2PostDamage(__DRIdrawable *draw, - struct drm_clip_rect *rects, - int numRects, void *loaderPrivate) -{ - XRectangle *xrects; - XserverRegion region; - __GLXDRIdrawable *glxDraw = loaderPrivate; - __GLXscreenConfigs *psc = glxDraw->psc; - Display *dpy = psc->dpy; - int i; - - xrects = malloc(sizeof(XRectangle) * numRects); - if (xrects == NULL) - return; - - for (i = 0; i < numRects; i++) { - xrects[i].x = rects[i].x1; - xrects[i].y = rects[i].y1; - xrects[i].width = rects[i].x2 - rects[i].x1; - xrects[i].height = rects[i].y2 - rects[i].y1; - } - region = XFixesCreateRegion(dpy, xrects, numRects); - free(xrects); - XDamageAdd(dpy, glxDraw->xDrawable, region); - XFixesDestroyRegion(dpy, region); -} - -static const __DRIloaderExtension dri2LoaderExtension = { - { __DRI_LOADER, __DRI_LOADER_VERSION }, - dri2ReemitDrawableInfo, - dri2PostDamage -}; - -static const __DRIextension *loader_extensions[] = { - &dri2LoaderExtension.base, - &systemTimeExtension.base, - NULL -}; - -static __GLXDRIscreen *dri2CreateScreen(__GLXscreenConfigs *psc, int screen, - __GLXdisplayPrivate *priv) -{ - const __DRIconfig **driver_configs; - const __DRIextension **extensions; - __GLXDRIscreen *psp; - unsigned int sareaHandle; - char *driverName, *busID; - drm_magic_t magic; - int i; - - psp = Xmalloc(sizeof *psp); - if (psp == NULL) - return NULL; - - /* Initialize per screen dynamic client GLX extensions */ - psc->ext_list_first_time = GL_TRUE; - - if (!DRI2Connect(psc->dpy, screen, &driverName, &busID, &sareaHandle)) - return NULL; - - psc->driver = driOpenDriver(driverName); - if (psc->driver == NULL) - goto handle_error; - - extensions = dlsym(psc->driver, __DRI_DRIVER_EXTENSIONS); - if (extensions == NULL) { - ErrorMessageF("driver exports no extensions (%s)\n", dlerror()); - goto handle_error; - } - - for (i = 0; extensions[i]; i++) { - if (strcmp(extensions[i]->name, __DRI_CORE) == 0) - psc->core = (__DRIcoreExtension *) extensions[i]; - } - - if (psc->core == NULL) { - ErrorMessageF("core dri extension not found\n"); - goto handle_error; - } - - psc->fd = drmOpen(NULL, busID); - if (psc->fd < 0) { - ErrorMessageF("failed to open drm device: %s\n", strerror(errno)); - return NULL; - } - - if (drmGetMagic(psc->fd, &magic)) - return NULL; - - if (!DRI2AuthConnection(psc->dpy, screen, magic)) { - ErrorMessageF("failed to authenticate drm access\n"); - return NULL; - } - - psc->__driScreen = - psc->core->createNewScreen(screen, psc->fd, sareaHandle, - loader_extensions, &driver_configs, psc); - if (psc->__driScreen == NULL) { - ErrorMessageF("failed to create dri screen\n"); - return NULL; - } - - driBindExtensions(psc); - - psc->configs = driConvertConfigs(psc->core, psc->configs, driver_configs); - psc->visuals = driConvertConfigs(psc->core, psc->visuals, driver_configs); - - psp->destroyScreen = dri2DestroyScreen; - psp->createContext = dri2CreateContext; - psp->createDrawable = dri2CreateDrawable; - - Xfree(driverName); - Xfree(busID); - - return psp; - - handle_error: - Xfree(driverName); - Xfree(busID); - - /* FIXME: clean up here */ - - return NULL; -} - -/* Called from __glXFreeDisplayPrivate. - */ -static void dri2DestroyDisplay(__GLXDRIdisplay *dpy) -{ - Xfree(dpy); -} - -/* - * Allocate, initialize and return a __DRIdisplayPrivate object. - * This is called from __glXInitialize() when we are given a new - * display pointer. - */ -_X_HIDDEN __GLXDRIdisplay *dri2CreateDisplay(Display *dpy) -{ - __GLXDRIdisplayPrivate *pdp; - int eventBase, errorBase; - - if (!DRI2QueryExtension(dpy, &eventBase, &errorBase)) - return NULL; - - pdp = Xmalloc(sizeof *pdp); - if (pdp == NULL) - return NULL; - - if (!DRI2QueryVersion(dpy, &pdp->driMajor, &pdp->driMinor)) { - Xfree(pdp); - return NULL; - } - - pdp->driPatch = 0; - - pdp->base.destroyDisplay = dri2DestroyDisplay; - pdp->base.createScreen = dri2CreateScreen; - - return &pdp->base; -} - -#endif /* GLX_DIRECT_RENDERING */ diff --git a/src/glx/x11/dri_common.c b/src/glx/x11/dri_common.c deleted file mode 100644 index b159d193a5..0000000000 --- a/src/glx/x11/dri_common.c +++ /dev/null @@ -1,399 +0,0 @@ -/* - * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. - * Copyright © 2008 Red Hat, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Soft- - * ware"), to deal in the Software without restriction, including without - * limitation the rights to use, copy, modify, merge, publish, distribute, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, provided that the above copyright - * notice(s) and this permission notice appear in all copies of the Soft- - * ware and that both the above copyright notice(s) and this permission - * notice appear in supporting documentation. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- - * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY - * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN - * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- - * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- - * MANCE OF THIS SOFTWARE. - * - * Except as contained in this notice, the name of a copyright holder shall - * not be used in advertising or otherwise to promote the sale, use or - * other dealings in this Software without prior written authorization of - * the copyright holder. - * - * Authors: - * Kevin E. Martin - * Brian Paul - * Kristian Høgsberg (krh@redhat.com) - */ - -#ifdef GLX_DIRECT_RENDERING - -#include -#include -#include "glheader.h" -#include "glxclient.h" -#include "glcontextmodes.h" -#include "dri_common.h" - -#ifndef RTLD_NOW -#define RTLD_NOW 0 -#endif -#ifndef RTLD_GLOBAL -#define RTLD_GLOBAL 0 -#endif - -_X_HIDDEN void InfoMessageF(const char *f, ...) -{ - va_list args; - const char *env; - - if ((env = getenv("LIBGL_DEBUG")) && strstr(env, "verbose")) { - fprintf(stderr, "libGL: "); - va_start(args, f); - vfprintf(stderr, f, args); - va_end(args); - } -} - -/** - * Print error to stderr, unless LIBGL_DEBUG=="quiet". - */ -_X_HIDDEN void ErrorMessageF(const char *f, ...) -{ - va_list args; - const char *env; - - if ((env = getenv("LIBGL_DEBUG")) && !strstr(env, "quiet")) { - fprintf(stderr, "libGL error: "); - va_start(args, f); - vfprintf(stderr, f, args); - va_end(args); - } -} - -#ifndef DEFAULT_DRIVER_DIR -/* this is normally defined in Mesa/configs/default with DRI_DRIVER_SEARCH_PATH */ -#define DEFAULT_DRIVER_DIR "/usr/local/lib/dri" -#endif - -/** - * Try to \c dlopen the named driver. - * - * This function adds the "_dri.so" suffix to the driver name and searches the - * directories specified by the \c LIBGL_DRIVERS_PATH environment variable in - * order to find the driver. - * - * \param driverName - a name like "tdfx", "i810", "mga", etc. - * - * \returns - * A handle from \c dlopen, or \c NULL if driver file not found. - */ -_X_HIDDEN void *driOpenDriver(const char *driverName) -{ - void *glhandle, *handle; - const char *libPaths, *p, *next; - char realDriverName[200]; - int len; - - /* Attempt to make sure libGL symbols will be visible to the driver */ - glhandle = dlopen("libGL.so.1", RTLD_NOW | RTLD_GLOBAL); - - libPaths = NULL; - if (geteuid() == getuid()) { - /* don't allow setuid apps to use LIBGL_DRIVERS_PATH */ - libPaths = getenv("LIBGL_DRIVERS_PATH"); - if (!libPaths) - libPaths = getenv("LIBGL_DRIVERS_DIR"); /* deprecated */ - } - if (libPaths == NULL) - libPaths = DEFAULT_DRIVER_DIR; - - handle = NULL; - for (p = libPaths; *p; p = next) { - next = strchr(p, ':'); - if (next == NULL) { - len = strlen(p); - next = p + len; - } else { - len = next - p; - next++; - } - -#ifdef GLX_USE_TLS - snprintf(realDriverName, sizeof realDriverName, - "%.*s/tls/%s_dri.so", len, p, driverName); - InfoMessageF("OpenDriver: trying %s\n", realDriverName); - handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); -#endif - - if ( handle == NULL ) { - snprintf(realDriverName, sizeof realDriverName, - "%.*s/%s_dri.so", len, p, driverName); - InfoMessageF("OpenDriver: trying %s\n", realDriverName); - handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); - } - - if ( handle != NULL ) - break; - else - ErrorMessageF("dlopen %s failed (%s)\n", realDriverName, dlerror()); - } - - if (!handle) - ErrorMessageF("unable to load driver: %s_dri.so\n", driverName); - - if (glhandle) - dlclose(glhandle); - - return handle; -} - -_X_HIDDEN const __DRIsystemTimeExtension systemTimeExtension = { - { __DRI_SYSTEM_TIME, __DRI_SYSTEM_TIME_VERSION }, - __glXGetUST, - __driGetMscRateOML -}; - -#define __ATTRIB(attrib, field) \ - { attrib, offsetof(__GLcontextModes, field) } - -static const struct { unsigned int attrib, offset; } attribMap[] = { - __ATTRIB(__DRI_ATTRIB_BUFFER_SIZE, rgbBits), - __ATTRIB(__DRI_ATTRIB_LEVEL, level), - __ATTRIB(__DRI_ATTRIB_RED_SIZE, redBits), - __ATTRIB(__DRI_ATTRIB_GREEN_SIZE, greenBits), - __ATTRIB(__DRI_ATTRIB_BLUE_SIZE, blueBits), - __ATTRIB(__DRI_ATTRIB_ALPHA_SIZE, alphaBits), - __ATTRIB(__DRI_ATTRIB_DEPTH_SIZE, depthBits), - __ATTRIB(__DRI_ATTRIB_STENCIL_SIZE, stencilBits), - __ATTRIB(__DRI_ATTRIB_ACCUM_RED_SIZE, accumRedBits), - __ATTRIB(__DRI_ATTRIB_ACCUM_GREEN_SIZE, accumGreenBits), - __ATTRIB(__DRI_ATTRIB_ACCUM_BLUE_SIZE, accumBlueBits), - __ATTRIB(__DRI_ATTRIB_ACCUM_ALPHA_SIZE, accumAlphaBits), - __ATTRIB(__DRI_ATTRIB_SAMPLE_BUFFERS, sampleBuffers), - __ATTRIB(__DRI_ATTRIB_SAMPLES, samples), - __ATTRIB(__DRI_ATTRIB_DOUBLE_BUFFER, doubleBufferMode), - __ATTRIB(__DRI_ATTRIB_STEREO, stereoMode), - __ATTRIB(__DRI_ATTRIB_AUX_BUFFERS, numAuxBuffers), -#if 0 - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_TYPE, transparentPixel), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_INDEX_VALUE, transparentIndex), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_RED_VALUE, transparentRed), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_GREEN_VALUE, transparentGreen), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_BLUE_VALUE, transparentBlue), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_ALPHA_VALUE, transparentAlpha), - __ATTRIB(__DRI_ATTRIB_RED_MASK, redMask), - __ATTRIB(__DRI_ATTRIB_GREEN_MASK, greenMask), - __ATTRIB(__DRI_ATTRIB_BLUE_MASK, blueMask), - __ATTRIB(__DRI_ATTRIB_ALPHA_MASK, alphaMask), -#endif - __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_WIDTH, maxPbufferWidth), - __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_HEIGHT, maxPbufferHeight), - __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_PIXELS, maxPbufferPixels), - __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_WIDTH, optimalPbufferWidth), - __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_HEIGHT, optimalPbufferHeight), -#if 0 - __ATTRIB(__DRI_ATTRIB_SWAP_METHOD, swapMethod), -#endif - __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGB, bindToTextureRgb), - __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGBA, bindToTextureRgba), - __ATTRIB(__DRI_ATTRIB_BIND_TO_MIPMAP_TEXTURE, bindToMipmapTexture), - __ATTRIB(__DRI_ATTRIB_YINVERTED, yInverted), -}; - -#define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0])) - -static int -scalarEqual(__GLcontextModes *mode, unsigned int attrib, unsigned int value) -{ - unsigned int glxValue; - int i; - - for (i = 0; i < ARRAY_SIZE(attribMap); i++) - if (attribMap[i].attrib == attrib) { - glxValue = *(unsigned int *) ((char *) mode + attribMap[i].offset); - return glxValue == GLX_DONT_CARE || glxValue == value; - } - - return GL_TRUE; /* Is a non-existing attribute equal to value? */ -} - -static int -driConfigEqual(const __DRIcoreExtension *core, - __GLcontextModes *modes, const __DRIconfig *driConfig) -{ - unsigned int attrib, value, glxValue; - int i; - - i = 0; - while (core->indexConfigAttrib(driConfig, i++, &attrib, &value)) { - switch (attrib) { - case __DRI_ATTRIB_RENDER_TYPE: - glxValue = 0; - if (value & __DRI_ATTRIB_RGBA_BIT) { - glxValue |= GLX_RGBA_BIT; - } else if (value & __DRI_ATTRIB_COLOR_INDEX_BIT) { - glxValue |= GLX_COLOR_INDEX_BIT; - } - if (glxValue != modes->renderType) - return GL_FALSE; - break; - - case __DRI_ATTRIB_CONFIG_CAVEAT: - if (value & __DRI_ATTRIB_NON_CONFORMANT_CONFIG) - glxValue = GLX_NON_CONFORMANT_CONFIG; - else if (value & __DRI_ATTRIB_SLOW_BIT) - glxValue = GLX_SLOW_CONFIG; - else - glxValue = GLX_NONE; - if (glxValue != modes->visualRating) - return GL_FALSE; - break; - - case __DRI_ATTRIB_BIND_TO_TEXTURE_TARGETS: - glxValue = 0; - if (value & __DRI_ATTRIB_TEXTURE_1D_BIT) - glxValue |= GLX_TEXTURE_1D_BIT_EXT; - if (value & __DRI_ATTRIB_TEXTURE_2D_BIT) - glxValue |= GLX_TEXTURE_2D_BIT_EXT; - if (value & __DRI_ATTRIB_TEXTURE_RECTANGLE_BIT) - glxValue |= GLX_TEXTURE_RECTANGLE_BIT_EXT; - if (modes->bindToTextureTargets != GLX_DONT_CARE && - glxValue != modes->bindToTextureTargets) - return GL_FALSE; - break; - - default: - if (!scalarEqual(modes, attrib, value)) - return GL_FALSE; - } - } - - return GL_TRUE; -} - -static __GLcontextModes * -createDriMode(const __DRIcoreExtension *core, - __GLcontextModes *modes, const __DRIconfig **driConfigs) -{ - __GLXDRIconfigPrivate *config; - int i; - - for (i = 0; driConfigs[i]; i++) { - if (driConfigEqual(core, modes, driConfigs[i])) - break; - } - - if (driConfigs[i] == NULL) - return NULL; - - config = Xmalloc(sizeof *config); - if (config == NULL) - return NULL; - - config->modes = *modes; - config->driConfig = driConfigs[i]; - - return &config->modes; -} - -_X_HIDDEN __GLcontextModes * -driConvertConfigs(const __DRIcoreExtension *core, - __GLcontextModes *modes, const __DRIconfig **configs) -{ - __GLcontextModes head, *tail, *m; - - tail = &head; - head.next = NULL; - for (m = modes; m; m = m->next) { - tail->next = createDriMode(core, m, configs); - if (tail->next == NULL) { - /* no matching dri config for m */ - continue; - } - - - tail = tail->next; - } - - _gl_context_modes_destroy(modes); - - return head.next; -} - -_X_HIDDEN void -driBindExtensions(__GLXscreenConfigs *psc) -{ - const __DRIextension **extensions; - int i; - - extensions = psc->core->getExtensions(psc->__driScreen); - - for (i = 0; extensions[i]; i++) { -#ifdef __DRI_COPY_SUB_BUFFER - if (strcmp(extensions[i]->name, __DRI_COPY_SUB_BUFFER) == 0) { - psc->copySubBuffer = (__DRIcopySubBufferExtension *) extensions[i]; - __glXEnableDirectExtension(psc, "GLX_MESA_copy_sub_buffer_bit"); - } -#endif - -#ifdef __DRI_SWAP_CONTROL - if (strcmp(extensions[i]->name, __DRI_SWAP_CONTROL) == 0) { - psc->swapControl = (__DRIswapControlExtension *) extensions[i]; - __glXEnableDirectExtension(psc, "GLX_SGI_swap_control"); - __glXEnableDirectExtension(psc, "GLX_MESA_swap_control"); - } -#endif - -#ifdef __DRI_ALLOCATE - if (strcmp(extensions[i]->name, __DRI_ALLOCATE) == 0) { - psc->allocate = (__DRIallocateExtension *) extensions[i]; - __glXEnableDirectExtension(psc, "GLX_MESA_allocate_memory"); - } -#endif - -#ifdef __DRI_FRAME_TRACKING - if (strcmp(extensions[i]->name, __DRI_FRAME_TRACKING) == 0) { - psc->frameTracking = (__DRIframeTrackingExtension *) extensions[i]; - __glXEnableDirectExtension(psc, "GLX_MESA_swap_frame_usage"); - } -#endif - -#ifdef __DRI_MEDIA_STREAM_COUNTER - if (strcmp(extensions[i]->name, __DRI_MEDIA_STREAM_COUNTER) == 0) { - psc->msc = (__DRImediaStreamCounterExtension *) extensions[i]; - __glXEnableDirectExtension(psc, "GLX_SGI_video_sync"); - } -#endif - -#ifdef __DRI_SWAP_BUFFER_COUNTER - /* No driver supports this at this time and the extension is - * not defined in dri_interface.h. Will enable - * GLX_OML_sync_control if implemented. */ -#endif - -#ifdef __DRI_READ_DRAWABLE - if (strcmp(extensions[i]->name, __DRI_READ_DRAWABLE) == 0) { - __glXEnableDirectExtension(psc, "GLX_SGI_make_current_read"); - } -#endif - -#ifdef __DRI_TEX_BUFFER - if (strcmp(extensions[i]->name, __DRI_TEX_BUFFER) == 0) { - psc->texBuffer = (__DRItexBufferExtension *) extensions[i]; - __glXEnableDirectExtension(psc, "GLX_EXT_texture_from_pixmap"); - } -#endif - - /* Ignore unknown extensions */ - } -} - -#endif /* GLX_DIRECT_RENDERING */ diff --git a/src/glx/x11/dri_common.h b/src/glx/x11/dri_common.h deleted file mode 100644 index 3556510335..0000000000 --- a/src/glx/x11/dri_common.h +++ /dev/null @@ -1,60 +0,0 @@ -/* - * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. - * Copyright © 2008 Red Hat, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Soft- - * ware"), to deal in the Software without restriction, including without - * limitation the rights to use, copy, modify, merge, publish, distribute, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, provided that the above copyright - * notice(s) and this permission notice appear in all copies of the Soft- - * ware and that both the above copyright notice(s) and this permission - * notice appear in supporting documentation. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABIL- - * ITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT OF THIRD PARTY - * RIGHTS. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR HOLDERS INCLUDED IN - * THIS NOTICE BE LIABLE FOR ANY CLAIM, OR ANY SPECIAL INDIRECT OR CONSE- - * QUENTIAL DAMAGES, OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, - * DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER - * TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFOR- - * MANCE OF THIS SOFTWARE. - * - * Except as contained in this notice, the name of a copyright holder shall - * not be used in advertising or otherwise to promote the sale, use or - * other dealings in this Software without prior written authorization of - * the copyright holder. - * - * Authors: - * Kevin E. Martin - * Brian Paul - * Kristian Høgsberg (krh@redhat.com) - */ - -#ifndef _DRI_COMMON_H -#define _DRI_COMMON_H - -typedef struct __GLXDRIconfigPrivateRec __GLXDRIconfigPrivate; - -struct __GLXDRIconfigPrivateRec { - __GLcontextModes modes; - const __DRIconfig *driConfig; -}; - -extern __GLcontextModes * -driConvertConfigs(const __DRIcoreExtension *core, - __GLcontextModes *modes, const __DRIconfig **configs); - -extern const __DRIsystemTimeExtension systemTimeExtension; - -extern void InfoMessageF(const char *f, ...); - -extern void ErrorMessageF(const char *f, ...); - -extern void *driOpenDriver(const char *driverName); - -extern void driBindExtensions(__GLXscreenConfigs *psc); - -#endif /* _DRI_COMMON_H */ diff --git a/src/glx/x11/dri_glx.c b/src/glx/x11/dri_glx.c index 1cb3204d4c..21e07c1935 100644 --- a/src/glx/x11/dri_glx.c +++ b/src/glx/x11/dri_glx.c @@ -34,47 +34,260 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #ifdef GLX_DIRECT_RENDERING -#include -#include -#include +#include +#include +#include +#include #include "glheader.h" #include "glxclient.h" -#include "glcontextmodes.h" #include "xf86dri.h" #include "sarea.h" +#include #include +#include "dri_glx.h" #include -#include -#include "xf86drm.h" -#include "dri_common.h" - -typedef struct __GLXDRIdisplayPrivateRec __GLXDRIdisplayPrivate; -typedef struct __GLXDRIcontextPrivateRec __GLXDRIcontextPrivate; - -struct __GLXDRIdisplayPrivateRec { - __GLXDRIdisplay base; - - /* - ** XFree86-DRI version information - */ - int driMajor; - int driMinor; - int driPatch; -}; - -struct __GLXDRIcontextPrivateRec { - __GLXDRIcontext base; - __DRIcontext *driContext; - XID hwContextID; - __GLXscreenConfigs *psc; -}; +#include + +#ifndef RTLD_NOW +#define RTLD_NOW 0 +#endif +#ifndef RTLD_GLOBAL +#define RTLD_GLOBAL 0 +#endif + + +#ifndef DEFAULT_DRIVER_DIR +/* this is normally defined in Mesa/configs/default with DRI_DRIVER_SEARCH_PATH */ +#define DEFAULT_DRIVER_DIR "/usr/X11R6/lib/modules/dri" +#endif + +static __DRIdriver *Drivers = NULL; + + +/* + * printf wrappers + */ + +static void InfoMessageF(const char *f, ...) +{ + va_list args; + const char *env; + + if ((env = getenv("LIBGL_DEBUG")) && strstr(env, "verbose")) { + fprintf(stderr, "libGL: "); + va_start(args, f); + vfprintf(stderr, f, args); + va_end(args); + } +} + +/** + * Print error to stderr, unless LIBGL_DEBUG=="quiet". + */ +static void ErrorMessageF(const char *f, ...) +{ + va_list args; + const char *env; + + if ((env = getenv("LIBGL_DEBUG")) && !strstr(env, "quiet")) { + fprintf(stderr, "libGL error: "); + va_start(args, f); + vfprintf(stderr, f, args); + va_end(args); + } +} + + +/** + * Extract the ith directory path out of a colon-separated list of paths. No + * more than \c dirLen characters, including the terminating \c NUL, will be + * written to \c dir. + * + * \param index Index of path to extract (starting at zero) + * \param paths The colon-separated list of paths + * \param dirLen Maximum length of result to store in \c dir + * \param dir Buffer to hold the extracted directory path + * + * \returns + * The number of characters that would have been written to \c dir had there + * been enough room. This does not include the terminating \c NUL. When + * extraction fails, zero will be returned. + * + * \todo + * It seems like this function could be rewritten to use \c strchr. + */ +static size_t +ExtractDir(int index, const char *paths, int dirLen, char *dir) +{ + int i, len; + const char *start, *end; + + /* find ith colon */ + start = paths; + i = 0; + while (i < index) { + if (*start == ':') { + i++; + start++; + } + else if (*start == 0) { + /* end of string and couldn't find ith colon */ + dir[0] = 0; + return 0; + } + else { + start++; + } + } + + while (*start == ':') + start++; + + /* find next colon, or end of string */ + end = start + 1; + while (*end != ':' && *end != 0) { + end++; + } + + /* copy string between and into result string */ + len = end - start; + if (len > dirLen - 1) + len = dirLen - 1; + strncpy(dir, start, len); + dir[len] = 0; + + return( end - start ); +} + + +/** + * Versioned name of the expected \c __driCreateNewScreen function. + * + * The version of the last incompatible loader/driver inteface change is + * appended to the name of the \c __driCreateNewScreen function. This + * prevents loaders from trying to load drivers that are too old. + * + * \todo + * Create a macro or something so that this is automatically updated. + */ +static const char createNewScreenName[] = "__driCreateNewScreen_20050727"; + + +/** + * Try to \c dlopen the named driver. + * + * This function adds the "_dri.so" suffix to the driver name and searches the + * directories specified by the \c LIBGL_DRIVERS_PATH environment variable in + * order to find the driver. + * + * \param driverName - a name like "tdfx", "i810", "mga", etc. + * + * \returns + * A handle from \c dlopen, or \c NULL if driver file not found. + */ +static __DRIdriver *OpenDriver(const char *driverName) +{ + void *glhandle = NULL; + char *libPaths = NULL; + char libDir[1000]; + int i; + __DRIdriver *driver; + + /* First, search Drivers list to see if we've already opened this driver */ + for (driver = Drivers; driver; driver = driver->next) { + if (strcmp(driver->name, driverName) == 0) { + /* found it */ + return driver; + } + } + + /* Attempt to make sure libGL symbols will be visible to the driver */ + glhandle = dlopen("libGL.so.1", RTLD_NOW | RTLD_GLOBAL); + + if (geteuid() == getuid()) { + /* don't allow setuid apps to use LIBGL_DRIVERS_PATH */ + libPaths = getenv("LIBGL_DRIVERS_PATH"); + if (!libPaths) + libPaths = getenv("LIBGL_DRIVERS_DIR"); /* deprecated */ + } + if (!libPaths) + libPaths = DEFAULT_DRIVER_DIR; + + for ( i = 0 ; ExtractDir(i, libPaths, 1000, libDir) != 0 ; i++ ) { + char realDriverName[200]; + void *handle = NULL; + + + /* If TLS support is enabled, try to open the TLS version of the driver + * binary first. If that fails, try the non-TLS version. + */ +#ifdef GLX_USE_TLS + snprintf(realDriverName, 200, "%s/tls/%s_dri.so", libDir, driverName); + InfoMessageF("OpenDriver: trying %s\n", realDriverName); + handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); +#endif + + if ( handle == NULL ) { + snprintf(realDriverName, 200, "%s/%s_dri.so", libDir, driverName); + InfoMessageF("OpenDriver: trying %s\n", realDriverName); + handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); + } + + if ( handle != NULL ) { + /* allocate __DRIdriver struct */ + driver = (__DRIdriver *) Xmalloc(sizeof(__DRIdriver)); + if (!driver) + break; /* out of memory! */ + /* init the struct */ + driver->name = __glXstrdup(driverName); + if (!driver->name) { + Xfree(driver); + driver = NULL; + break; /* out of memory! */ + } + + driver->createNewScreenFunc = (PFNCREATENEWSCREENFUNC) + dlsym(handle, createNewScreenName); + + if ( driver->createNewScreenFunc == NULL ) { + /* If the driver doesn't have this symbol then something's + * really, really wrong. + */ + ErrorMessageF("%s not defined in %s_dri.so!\n" + "Your driver may be too old for this libGL.\n", + createNewScreenName, driverName); + Xfree(driver); + driver = NULL; + dlclose(handle); + continue; + } + driver->handle = handle; + /* put at head of linked list */ + driver->next = Drivers; + Drivers = driver; + break; + } + else { + ErrorMessageF("dlopen %s failed (%s)\n", realDriverName, dlerror()); + } + } + + if (!driver) + ErrorMessageF("unable to load driver: %s_dri.so\n", driverName); + + if (glhandle) + dlclose(glhandle); + + return driver; +} + /* * Given a display pointer and screen number, determine the name of * the DRI driver for the screen. (I.e. "r128", "tdfx", etc). * Return True for success, False for failure. */ -static Bool driGetDriverName(Display *dpy, int scrNum, char **driverName) +static Bool GetDriverName(Display *dpy, int scrNum, char **driverName) { int directCapable; Bool b; @@ -104,6 +317,25 @@ static Bool driGetDriverName(Display *dpy, int scrNum, char **driverName) return True; } + +/* + * Given a display pointer and screen number, return a __DRIdriver handle. + * Return NULL if anything goes wrong. + */ +__DRIdriver *driGetDriver(Display *dpy, int scrNum) +{ + char *driverName; + if (GetDriverName(dpy, scrNum, &driverName)) { + __DRIdriver *ret; + ret = OpenDriver(driverName); + if (driverName) + Xfree(driverName); + return ret; + } + return NULL; +} + + /* * Exported function for querying the DRI driver for a given screen. * @@ -113,7 +345,7 @@ static Bool driGetDriverName(Display *dpy, int scrNum, char **driverName) PUBLIC const char *glXGetScreenDriver (Display *dpy, int scrNum) { static char ret[32]; char *driverName; - if (driGetDriverName(dpy, scrNum, &driverName)) { + if (GetDriverName(dpy, scrNum, &driverName)) { int len; if (!driverName) return NULL; @@ -127,6 +359,7 @@ PUBLIC const char *glXGetScreenDriver (Display *dpy, int scrNum) { return NULL; } + /* * Exported function for obtaining a driver's option list (UTF-8 encoded XML). * @@ -138,532 +371,71 @@ PUBLIC const char *glXGetScreenDriver (Display *dpy, int scrNum) { * * Note: The driver remains opened after this function returns. */ -PUBLIC const char *glXGetDriverConfig (const char *driverName) -{ - void *handle = driOpenDriver (driverName); - if (handle) - return dlsym (handle, "__driConfigOptions"); +PUBLIC const char *glXGetDriverConfig (const char *driverName) { + __DRIdriver *driver = OpenDriver (driverName); + if (driver) + return dlsym (driver->handle, "__driConfigOptions"); else return NULL; } -#ifdef XDAMAGE_1_1_INTERFACE - -static GLboolean has_damage_post(Display *dpy) -{ - static GLboolean inited = GL_FALSE; - static GLboolean has_damage; - - if (!inited) { - int major, minor; - - if (XDamageQueryVersion(dpy, &major, &minor) && - major == 1 && minor >= 1) - { - has_damage = GL_TRUE; - } else { - has_damage = GL_FALSE; - } - inited = GL_TRUE; - } - - return has_damage; -} - -static void __glXReportDamage(__DRIdrawable *driDraw, - int x, int y, - drm_clip_rect_t *rects, int num_rects, - GLboolean front_buffer, - void *loaderPrivate) -{ - XRectangle *xrects; - XserverRegion region; - int i; - int x_off, y_off; - __GLXDRIdrawable *glxDraw = loaderPrivate; - __GLXscreenConfigs *psc = glxDraw->psc; - Display *dpy = psc->dpy; - Drawable drawable; - - if (!has_damage_post(dpy)) - return; - - if (front_buffer) { - x_off = x; - y_off = y; - drawable = RootWindow(dpy, psc->scr); - } else{ - x_off = 0; - y_off = 0; - drawable = glxDraw->xDrawable; - } - - xrects = malloc(sizeof(XRectangle) * num_rects); - if (xrects == NULL) - return; - - for (i = 0; i < num_rects; i++) { - xrects[i].x = rects[i].x1 + x_off; - xrects[i].y = rects[i].y1 + y_off; - xrects[i].width = rects[i].x2 - rects[i].x1; - xrects[i].height = rects[i].y2 - rects[i].y1; - } - region = XFixesCreateRegion(dpy, xrects, num_rects); - free(xrects); - XDamageAdd(dpy, drawable, region); - XFixesDestroyRegion(dpy, region); -} - -static const __DRIdamageExtension damageExtension = { - { __DRI_DAMAGE, __DRI_DAMAGE_VERSION }, - __glXReportDamage, -}; - -#endif - -static GLboolean -__glXDRIGetDrawableInfo(__DRIdrawable *drawable, - unsigned int *index, unsigned int *stamp, - int *X, int *Y, int *W, int *H, - int *numClipRects, drm_clip_rect_t ** pClipRects, - int *backX, int *backY, - int *numBackClipRects, drm_clip_rect_t **pBackClipRects, - void *loaderPrivate) -{ - __GLXDRIdrawable *glxDraw = loaderPrivate; - __GLXscreenConfigs *psc = glxDraw->psc; - Display *dpy = psc->dpy; - - return XF86DRIGetDrawableInfo(dpy, psc->scr, glxDraw->drawable, - index, stamp, X, Y, W, H, - numClipRects, pClipRects, - backX, backY, - numBackClipRects, pBackClipRects); -} - -static const __DRIgetDrawableInfoExtension getDrawableInfoExtension = { - { __DRI_GET_DRAWABLE_INFO, __DRI_GET_DRAWABLE_INFO_VERSION }, - __glXDRIGetDrawableInfo -}; - -static const __DRIextension *loader_extensions[] = { - &systemTimeExtension.base, - &getDrawableInfoExtension.base, -#ifdef XDAMAGE_1_1_INTERFACE - &damageExtension.base, -#endif - NULL -}; -#ifndef GLX_USE_APPLEGL - -/** - * Perform the required libGL-side initialization and call the client-side - * driver's \c __driCreateNewScreen function. - * - * \param dpy Display pointer. - * \param scrn Screen number on the display. - * \param psc DRI screen information. - * \param driDpy DRI display information. - * \param createNewScreen Pointer to the client-side driver's - * \c __driCreateNewScreen function. - * \returns A pointer to the \c __DRIscreenPrivate structure returned by - * the client-side driver on success, or \c NULL on failure. +/* Called from __glXFreeDisplayPrivate. */ -static void * -CallCreateNewScreen(Display *dpy, int scrn, __GLXscreenConfigs *psc, - __GLXDRIdisplayPrivate * driDpy) -{ - void *psp = NULL; - drm_handle_t hSAREA; - drmAddress pSAREA = MAP_FAILED; - char *BusID; - __DRIversion ddx_version; - __DRIversion dri_version; - __DRIversion drm_version; - __DRIframebuffer framebuffer; - int fd = -1; - int status; - - drm_magic_t magic; - drmVersionPtr version; - int newlyopened; - char *driverName; - drm_handle_t hFB; - int junk; - const __DRIconfig **driver_configs; - - /* DRI protocol version. */ - dri_version.major = driDpy->driMajor; - dri_version.minor = driDpy->driMinor; - dri_version.patch = driDpy->driPatch; - - framebuffer.base = MAP_FAILED; - framebuffer.dev_priv = NULL; - - if (!XF86DRIOpenConnection(dpy, scrn, &hSAREA, &BusID)) { - fprintf(stderr, "libGL error: XF86DRIOpenConnection failed\n"); - goto handle_error; - } - - fd = drmOpenOnce(NULL, BusID, &newlyopened); - - Xfree(BusID); /* No longer needed */ - - if (fd < 0) { - fprintf(stderr, "libGL error: drmOpenOnce failed (%s)\n", - strerror(-fd)); - goto handle_error; - } - - if (drmGetMagic(fd, &magic)) { - fprintf(stderr, "libGL error: drmGetMagic failed\n"); - goto handle_error; - } - - version = drmGetVersion(fd); - if (version) { - drm_version.major = version->version_major; - drm_version.minor = version->version_minor; - drm_version.patch = version->version_patchlevel; - drmFreeVersion(version); - } - else { - drm_version.major = -1; - drm_version.minor = -1; - drm_version.patch = -1; - } - - if (newlyopened && !XF86DRIAuthConnection(dpy, scrn, magic)) { - fprintf(stderr, "libGL error: XF86DRIAuthConnection failed\n"); - goto handle_error; - } - - /* Get device name (like "tdfx") and the ddx version numbers. - * We'll check the version in each DRI driver's "createNewScreen" - * function. */ - if (!XF86DRIGetClientDriverName(dpy, scrn, - &ddx_version.major, - &ddx_version.minor, - &ddx_version.patch, - &driverName)) { - fprintf(stderr, "libGL error: XF86DRIGetClientDriverName failed\n"); - goto handle_error; - } - - Xfree(driverName); /* No longer needed. */ - - /* - * Get device-specific info. pDevPriv will point to a struct - * (such as DRIRADEONRec in xfree86/driver/ati/radeon_dri.h) that - * has information about the screen size, depth, pitch, ancilliary - * buffers, DRM mmap handles, etc. - */ - if (!XF86DRIGetDeviceInfo(dpy, scrn, &hFB, &junk, - &framebuffer.size, &framebuffer.stride, - &framebuffer.dev_priv_size, &framebuffer.dev_priv)) { - fprintf(stderr, "libGL error: XF86DRIGetDeviceInfo failed"); - goto handle_error; - } - - framebuffer.width = DisplayWidth(dpy, scrn); - framebuffer.height = DisplayHeight(dpy, scrn); - - /* Map the framebuffer region. */ - status = drmMap(fd, hFB, framebuffer.size, - (drmAddressPtr)&framebuffer.base); - if (status != 0) { - fprintf(stderr, "libGL error: drmMap of framebuffer failed (%s)", - strerror(-status)); - goto handle_error; - } - - /* Map the SAREA region. Further mmap regions may be setup in - * each DRI driver's "createNewScreen" function. - */ - status = drmMap(fd, hSAREA, SAREA_MAX, &pSAREA); - if (status != 0) { - fprintf(stderr, "libGL error: drmMap of SAREA failed (%s)", - strerror(-status)); - goto handle_error; - } - - psp = (*psc->legacy->createNewScreen)(scrn, - &ddx_version, - &dri_version, - &drm_version, - &framebuffer, - pSAREA, - fd, - loader_extensions, - &driver_configs, - psc); - - if (psp == NULL) { - fprintf(stderr, "libGL error: Calling driver entry point failed"); - goto handle_error; - } - - psc->configs = driConvertConfigs(psc->core, psc->configs, driver_configs); - psc->visuals = driConvertConfigs(psc->core, psc->visuals, driver_configs); - - return psp; - - handle_error: - if (pSAREA != MAP_FAILED) - drmUnmap(pSAREA, SAREA_MAX); - - if (framebuffer.base != MAP_FAILED) - drmUnmap((drmAddress)framebuffer.base, framebuffer.size); - - if (framebuffer.dev_priv != NULL) - Xfree(framebuffer.dev_priv); - - if (fd >= 0) - drmCloseOnce(fd); - - XF86DRICloseConnection(dpy, scrn); - - fprintf(stderr, "libGL error: reverting to (slow) indirect rendering\n"); - - return NULL; -} - -#else /* !GLX_USE_APPLEGL */ - -static void * -CallCreateNewScreen(Display *dpy, int scrn, __GLXscreenConfigs *psc, - __GLXDRIdisplayPrivate * driDpy) +static void driDestroyDisplay(Display *dpy, void *private) { - return NULL; -} - -#endif /* !GLX_USE_APPLEGL */ - -static void driDestroyContext(__GLXDRIcontext *context, - __GLXscreenConfigs *psc, Display *dpy) -{ - __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; - - (*psc->core->destroyContext)(pcp->driContext); - - XF86DRIDestroyContext(psc->dpy, psc->scr, pcp->hwContextID); -} - -static Bool driBindContext(__GLXDRIcontext *context, - __GLXDRIdrawable *draw, __GLXDRIdrawable *read) -{ - __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; - const __DRIcoreExtension *core = pcp->psc->core; - - return (*core->bindContext)(pcp->driContext, - draw->driDrawable, - read->driDrawable); -} - -static void driUnbindContext(__GLXDRIcontext *context) -{ - __GLXDRIcontextPrivate *pcp = (__GLXDRIcontextPrivate *) context; - const __DRIcoreExtension *core = pcp->psc->core; - - (*core->unbindContext)(pcp->driContext); -} - -static __GLXDRIcontext *driCreateContext(__GLXscreenConfigs *psc, - const __GLcontextModes *mode, - GLXContext gc, - GLXContext shareList, int renderType) -{ - __GLXDRIcontextPrivate *pcp, *pcp_shared; - drm_context_t hwContext; - __DRIcontext *shared = NULL; - __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) mode; - - if (!psc || !psc->driScreen) - return NULL; - - if (shareList) { - pcp_shared = (__GLXDRIcontextPrivate *) shareList->driContext; - shared = pcp_shared->driContext; - } - - pcp = Xmalloc(sizeof *pcp); - if (pcp == NULL) - return NULL; - - pcp->psc = psc; - if (!XF86DRICreateContextWithConfig(psc->dpy, psc->scr, - mode->visualID, - &pcp->hwContextID, &hwContext)) { - Xfree(pcp); - return NULL; + __DRIdisplayPrivate *pdpyp = (__DRIdisplayPrivate *)private; + + if (pdpyp) { + const int numScreens = ScreenCount(dpy); + int i; + for (i = 0; i < numScreens; i++) { + if (pdpyp->libraryHandles[i]) { + __DRIdriver *driver, *prev; + + /* Remove driver from Drivers list */ + for (prev = NULL, driver = Drivers; driver; + prev = driver, driver = driver->next) { + if (driver->handle == pdpyp->libraryHandles[i]) { + if (prev) + prev->next = driver->next; + else + Drivers = driver->next; + + Xfree(driver->name); + Xfree(driver); + break; + } + } + + dlclose(pdpyp->libraryHandles[i]); + } + } + Xfree(pdpyp->libraryHandles); + Xfree(pdpyp); } - - pcp->driContext = - (*psc->legacy->createNewContext)(psc->__driScreen, - config->driConfig, - renderType, - shared, - hwContext, - pcp); - if (pcp->driContext == NULL) { - XF86DRIDestroyContext(psc->dpy, psc->scr, pcp->hwContextID); - Xfree(pcp); - return NULL; - } - - pcp->base.destroyContext = driDestroyContext; - pcp->base.bindContext = driBindContext; - pcp->base.unbindContext = driUnbindContext; - - return &pcp->base; } -static void driDestroyDrawable(__GLXDRIdrawable *pdraw) -{ - __GLXscreenConfigs *psc = pdraw->psc; - - (*psc->core->destroyDrawable)(pdraw->driDrawable); - XF86DRIDestroyDrawable(psc->dpy, psc->scr, pdraw->drawable); - Xfree(pdraw); -} - -static __GLXDRIdrawable *driCreateDrawable(__GLXscreenConfigs *psc, - XID xDrawable, - GLXDrawable drawable, - const __GLcontextModes *modes) -{ - __GLXDRIdrawable *pdraw; - drm_drawable_t hwDrawable; - void *empty_attribute_list = NULL; - __GLXDRIconfigPrivate *config = (__GLXDRIconfigPrivate *) modes; - - /* Old dri can't handle GLX 1.3+ drawable constructors. */ - if (xDrawable != drawable) - return NULL; - - pdraw = Xmalloc(sizeof(*pdraw)); - if (!pdraw) - return NULL; - - pdraw->drawable = drawable; - pdraw->psc = psc; - - if (!XF86DRICreateDrawable(psc->dpy, psc->scr, drawable, &hwDrawable)) - return NULL; - - /* Create a new drawable */ - pdraw->driDrawable = - (*psc->legacy->createNewDrawable)(psc->__driScreen, - config->driConfig, - hwDrawable, - GLX_WINDOW_BIT, - empty_attribute_list, - pdraw); - - if (!pdraw->driDrawable) { - XF86DRIDestroyDrawable(psc->dpy, psc->scr, drawable); - Xfree(pdraw); - return NULL; - } - - pdraw->destroyDrawable = driDestroyDrawable; - - return pdraw; -} - -static void driDestroyScreen(__GLXscreenConfigs *psc) -{ - /* Free the direct rendering per screen data */ - if (psc->__driScreen) - (*psc->core->destroyScreen)(psc->__driScreen); - psc->__driScreen = NULL; - if (psc->driver) - dlclose(psc->driver); -} - -static __GLXDRIscreen *driCreateScreen(__GLXscreenConfigs *psc, int screen, - __GLXdisplayPrivate *priv) -{ - __GLXDRIdisplayPrivate *pdp; - __GLXDRIscreen *psp; - const __DRIextension **extensions; - char *driverName; - int i; - - psp = Xmalloc(sizeof *psp); - if (psp == NULL) - return NULL; - - /* Initialize per screen dynamic client GLX extensions */ - psc->ext_list_first_time = GL_TRUE; - - if (!driGetDriverName(priv->dpy, screen, &driverName)) { - Xfree(psp); - return NULL; - } - - psc->driver = driOpenDriver(driverName); - Xfree(driverName); - if (psc->driver == NULL) { - Xfree(psp); - return NULL; - } - - extensions = dlsym(psc->driver, __DRI_DRIVER_EXTENSIONS); - if (extensions == NULL) { - ErrorMessageF("driver exports no extensions (%s)\n", dlerror()); - Xfree(psp); - return NULL; - } - - for (i = 0; extensions[i]; i++) { - if (strcmp(extensions[i]->name, __DRI_CORE) == 0) - psc->core = (__DRIcoreExtension *) extensions[i]; - if (strcmp(extensions[i]->name, __DRI_LEGACY) == 0) - psc->legacy = (__DRIlegacyExtension *) extensions[i]; - } - - if (psc->core == NULL || psc->legacy == NULL) { - Xfree(psp); - return NULL; - } - - pdp = (__GLXDRIdisplayPrivate *) priv->driDisplay; - psc->__driScreen = - CallCreateNewScreen(psc->dpy, screen, psc, pdp); - if (psc->__driScreen == NULL) { - dlclose(psc->driver); - Xfree(psp); - return NULL; - } - - driBindExtensions(psc); - - psp->destroyScreen = driDestroyScreen; - psp->createContext = driCreateContext; - psp->createDrawable = driCreateDrawable; - - return psp; -} - -/* Called from __glXFreeDisplayPrivate. - */ -static void driDestroyDisplay(__GLXDRIdisplay *dpy) -{ - Xfree(dpy); -} /* * Allocate, initialize and return a __DRIdisplayPrivate object. * This is called from __glXInitialize() when we are given a new * display pointer. */ -_X_HIDDEN __GLXDRIdisplay *driCreateDisplay(Display *dpy) +void *driCreateDisplay(Display *dpy, __DRIdisplay *pdisp) { - __GLXDRIdisplayPrivate *pdpyp; + const int numScreens = ScreenCount(dpy); + __DRIdisplayPrivate *pdpyp; int eventBase, errorBase; int major, minor, patch; + int scrn; + + /* Initialize these fields to NULL in case we fail. + * If we don't do this we may later get segfaults trying to free random + * addresses when the display is closed. + */ + pdisp->private = NULL; + pdisp->destroyDisplay = NULL; if (!XF86DRIQueryExtension(dpy, &eventBase, &errorBase)) { return NULL; @@ -673,7 +445,7 @@ _X_HIDDEN __GLXDRIdisplay *driCreateDisplay(Display *dpy) return NULL; } - pdpyp = Xmalloc(sizeof *pdpyp); + pdpyp = (__DRIdisplayPrivate *)Xmalloc(sizeof(__DRIdisplayPrivate)); if (!pdpyp) { return NULL; } @@ -682,10 +454,41 @@ _X_HIDDEN __GLXDRIdisplay *driCreateDisplay(Display *dpy) pdpyp->driMinor = minor; pdpyp->driPatch = patch; - pdpyp->base.destroyDisplay = driDestroyDisplay; - pdpyp->base.createScreen = driCreateScreen; + pdisp->destroyDisplay = driDestroyDisplay; + + /* allocate array of pointers to createNewScreen funcs */ + pdisp->createNewScreen = (PFNCREATENEWSCREENFUNC *) + Xmalloc(numScreens * sizeof(void *)); + if (!pdisp->createNewScreen) { + Xfree(pdpyp); + return NULL; + } + + /* allocate array of library handles */ + pdpyp->libraryHandles = (void **) Xmalloc(numScreens * sizeof(void*)); + if (!pdpyp->libraryHandles) { + Xfree(pdisp->createNewScreen); + Xfree(pdpyp); + return NULL; + } + + /* dynamically discover DRI drivers for all screens, saving each + * driver's "__driCreateScreen" function pointer. That's the bootstrap + * entrypoint for all DRI drivers. + */ + for (scrn = 0; scrn < numScreens; scrn++) { + __DRIdriver *driver = driGetDriver(dpy, scrn); + if (driver) { + pdisp->createNewScreen[scrn] = driver->createNewScreenFunc; + pdpyp->libraryHandles[scrn] = driver->handle; + } + else { + pdisp->createNewScreen[scrn] = NULL; + pdpyp->libraryHandles[scrn] = NULL; + } + } - return &pdpyp->base; + return (void *)pdpyp; } #endif /* GLX_DIRECT_RENDERING */ diff --git a/src/glx/x11/glcontextmodes.c b/src/glx/x11/glcontextmodes.c index 326c8b2357..788ecf6a3a 100644 --- a/src/glx/x11/glcontextmodes.c +++ b/src/glx/x11/glcontextmodes.c @@ -336,8 +336,7 @@ _gl_get_context_mode_data(const __GLcontextModes *mode, int attribute, *value_return = mode->bindToTextureRgba; return 0; case GLX_BIND_TO_MIPMAP_TEXTURE_EXT: - *value_return = mode->bindToMipmapTexture == GL_TRUE ? GL_TRUE : - GL_FALSE; + *value_return = mode->bindToMipmapTexture; return 0; case GLX_BIND_TO_TEXTURE_TARGETS_EXT: *value_return = mode->bindToTextureTargets; @@ -418,7 +417,7 @@ _gl_context_modes_create( unsigned count, size_t minimum_size ) (*next)->bindToTextureRgb = GLX_DONT_CARE; (*next)->bindToTextureRgba = GLX_DONT_CARE; (*next)->bindToMipmapTexture = GLX_DONT_CARE; - (*next)->bindToTextureTargets = GLX_DONT_CARE; + (*next)->bindToTextureTargets = 0; (*next)->yInverted = GLX_DONT_CARE; next = & ((*next)->next); @@ -457,28 +456,19 @@ _gl_context_modes_destroy( __GLcontextModes * modes ) */ __GLcontextModes * -_gl_context_modes_find_visual(__GLcontextModes *modes, int vid) +_gl_context_modes_find_visual( __GLcontextModes * modes, int vid ) { - __GLcontextModes *m; + while ( modes != NULL ) { + if ( modes->visualID == vid ) { + break; + } - for (m = modes; m != NULL; m = m->next) - if (m->visualID == vid) - return m; + modes = modes->next; + } - return NULL; + return modes; } -__GLcontextModes * -_gl_context_modes_find_fbconfig(__GLcontextModes *modes, int fbid) -{ - __GLcontextModes *m; - - for (m = modes; m != NULL; m = m->next) - if (m->fbconfigID == fbid) - return m; - - return NULL; -} /** * Determine if two context-modes are the same. This is intended to be used diff --git a/src/glx/x11/glcontextmodes.h b/src/glx/x11/glcontextmodes.h index afd09cd7fb..4b5c6f68b8 100644 --- a/src/glx/x11/glcontextmodes.h +++ b/src/glx/x11/glcontextmodes.h @@ -44,10 +44,8 @@ extern int _gl_get_context_mode_data( const __GLcontextModes *mode, extern __GLcontextModes * _gl_context_modes_create( unsigned count, size_t minimum_size ); extern void _gl_context_modes_destroy( __GLcontextModes * modes ); -extern __GLcontextModes * - _gl_context_modes_find_visual(__GLcontextModes *modes, int vid); -extern __GLcontextModes * - _gl_context_modes_find_fbconfig(__GLcontextModes *modes, int fbid); +extern __GLcontextModes * _gl_context_modes_find_visual( + __GLcontextModes * modes, int vid ); extern GLboolean _gl_context_modes_are_same( const __GLcontextModes * a, const __GLcontextModes * b ); diff --git a/src/glx/x11/glx_pbuffer.c b/src/glx/x11/glx_pbuffer.c index 0f878f223f..1df2d0f342 100644 --- a/src/glx/x11/glx_pbuffer.c +++ b/src/glx/x11/glx_pbuffer.c @@ -164,33 +164,6 @@ DestroyPbuffer( Display * dpy, GLXDrawable drawable ) } -#ifdef GLX_DIRECT_RENDERING -extern __GLXDRIdrawable * -GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num); - -static GLenum -determineTextureTarget(const int *attribs, int numAttribs) -{ - GLenum target = 0; - int i; - - for (i = 0; i < numAttribs; i++) { - if (attribs[2 * i] == GLX_TEXTURE_TARGET_EXT) { - switch (attribs[2 * i + 1]) { - case GLX_TEXTURE_2D_EXT: - target = GL_TEXTURE_2D; - break; - case GLX_TEXTURE_RECTANGLE_EXT: - target = GL_TEXTURE_RECTANGLE_ARB; - break; - } - } - } - - return target; -} -#endif - /** * Get a drawable's attribute. * @@ -288,16 +261,6 @@ GetDrawableAttribute( Display *dpy, GLXDrawable drawable, } } -#ifdef GLX_DIRECT_RENDERING - { - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL); - - if (pdraw != NULL && !pdraw->textureTarget) - pdraw->textureTarget = determineTextureTarget((const int *)data, - num_attributes); - } -#endif - Xfree( data ); } } @@ -308,6 +271,7 @@ GetDrawableAttribute( Display *dpy, GLXDrawable drawable, return 0; } + /** * Create a non-pbuffer GLX drawable. * @@ -342,7 +306,7 @@ CreateDrawable( Display *dpy, const __GLcontextModes * fbconfig, req->glxCode = glxCode; req->screen = (CARD32) fbconfig->screen; req->fbconfig = fbconfig->fbconfigID; - req->window = (CARD32) drawable; + req->window = (GLXPbuffer) drawable; req->glxwindow = (GLXWindow) XAllocID(dpy); req->numAttribs = (CARD32) i; @@ -351,34 +315,6 @@ CreateDrawable( Display *dpy, const __GLcontextModes * fbconfig, UnlockDisplay(dpy); SyncHandle(); -#ifdef GLX_DIRECT_RENDERING - do { - /* FIXME: Maybe delay __DRIdrawable creation until the drawable - * is actually bound to a context... */ - - __GLXdisplayPrivate * const priv = __glXInitialize(dpy); - __GLXDRIdrawable *pdraw; - __GLXscreenConfigs *psc; - - psc = &priv->screenConfigs[fbconfig->screen]; - if (psc->driScreen == NULL) - break; - pdraw = psc->driScreen->createDrawable(psc, drawable, - req->glxwindow, fbconfig); - if (pdraw == NULL) { - fprintf(stderr, "failed to create drawable\n"); - break; - } - - if (__glxHashInsert(psc->drawHash, req->glxwindow, pdraw)) { - (*pdraw->destroyDrawable)(pdraw); - return None; /* FIXME: Check what we're supposed to do here... */ - } - - pdraw->textureTarget = determineTextureTarget(attrib_list, i); - } while (0); -#endif - return (GLXDrawable)req->glxwindow; } @@ -414,20 +350,6 @@ DestroyDrawable( Display * dpy, GLXDrawable drawable, CARD32 glxCode ) UnlockDisplay(dpy); SyncHandle(); -#ifdef GLX_DIRECT_RENDERING - { - int screen; - __GLXdisplayPrivate * const priv = __glXInitialize(dpy); - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); - __GLXscreenConfigs *psc = &priv->screenConfigs[screen]; - - if (pdraw != NULL) { - (*pdraw->destroyDrawable)(pdraw); - __glxHashDelete(psc->drawHash, drawable); - } - } -#endif - return; } @@ -538,24 +460,8 @@ glXCreateGLXPbufferSGIX(Display *dpy, GLXFBConfigSGIX config, PUBLIC GLXPbuffer glXCreatePbuffer(Display *dpy, GLXFBConfig config, const int *attrib_list) { - int i, width, height; - - width = 0; - height = 0; - - for (i = 0; attrib_list[i * 2]; i++) { - switch (attrib_list[i * 2]) { - case GLX_PBUFFER_WIDTH: - width = attrib_list[i * 2 + 1]; - break; - case GLX_PBUFFER_HEIGHT: - height = attrib_list[i * 2 + 1]; - break; - } - } - return (GLXPbuffer) CreatePbuffer( dpy, (__GLcontextModes *) config, - width, height, + 0, 0, attrib_list, GL_TRUE ); } diff --git a/src/glx/x11/glx_texture_compression.c b/src/glx/x11/glx_texture_compression.c new file mode 100644 index 0000000000..5676858017 --- /dev/null +++ b/src/glx/x11/glx_texture_compression.c @@ -0,0 +1,347 @@ +/* + * (C) Copyright IBM Corporation 2004 + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file glx_texture_compression.c + * Contains the routines required to implement GLX protocol for + * ARB_texture_compression and related extensions. + * + * \sa http://oss.sgi.com/projects/ogl-sample/registry/ARB/texture_compression.txt + * + * \author Ian Romanick + */ + +#include "packrender.h" +#include "packsingle.h" +#include "indirect.h" + +#include + + +void +__indirect_glGetCompressedTexImageARB( GLenum target, GLint level, + GLvoid * img ) +{ + __GLX_SINGLE_DECLARE_VARIABLES(); + xGLXGetTexImageReply reply; + size_t image_bytes; + + __GLX_SINGLE_LOAD_VARIABLES(); + __GLX_SINGLE_BEGIN( X_GLsop_GetCompressedTexImage, 8 ); + __GLX_SINGLE_PUT_LONG( 0, target ); + __GLX_SINGLE_PUT_LONG( 4, level ); + __GLX_SINGLE_READ_XREPLY(); + + image_bytes = reply.width; + assert( image_bytes <= ((4 * reply.length) - 0) ); + assert( image_bytes >= ((4 * reply.length) - 3) ); + + if ( image_bytes != 0 ) { + _XRead( dpy, (char *) img, image_bytes ); + if ( image_bytes < (4 * reply.length) ) { + _XEatData( dpy, (4 * reply.length) - image_bytes ); + } + } + + __GLX_SINGLE_END(); +} + + +/** + * Internal function used for \c glCompressedTexImage1D and + * \c glCompressedTexImage2D. + */ +static void +CompressedTexImage1D2D( GLenum target, GLint level, + GLenum internal_format, + GLsizei width, GLsizei height, + GLint border, GLsizei image_size, + const GLvoid *data, CARD32 rop ) +{ + __GLX_DECLARE_VARIABLES(); + + __GLX_LOAD_VARIABLES(); + if ( gc->currentDpy == NULL ) { + return; + } + + if ( (target == GL_PROXY_TEXTURE_1D) + || (target == GL_PROXY_TEXTURE_2D) + || (target == GL_PROXY_TEXTURE_CUBE_MAP) ) { + compsize = 0; + } + else { + compsize = image_size; + } + + cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE + + compsize ); + if ( cmdlen <= gc->maxSmallRenderCommandSize ) { + __GLX_BEGIN_VARIABLE( rop, cmdlen ); + __GLX_PUT_LONG( 4, target ); + __GLX_PUT_LONG( 8, level ); + __GLX_PUT_LONG( 12, internal_format ); + __GLX_PUT_LONG( 16, width ); + __GLX_PUT_LONG( 20, height ); + __GLX_PUT_LONG( 24, border ); + __GLX_PUT_LONG( 28, image_size ); + if ( compsize != 0 ) { + __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE, + data, image_size ); + } + __GLX_END( cmdlen ); + } + else { + assert( compsize != 0 ); + + __GLX_BEGIN_VARIABLE_LARGE( rop, cmdlen + 4 ); + __GLX_PUT_LONG( 8, target ); + __GLX_PUT_LONG( 12, level ); + __GLX_PUT_LONG( 16, internal_format ); + __GLX_PUT_LONG( 20, width ); + __GLX_PUT_LONG( 24, height ); + __GLX_PUT_LONG( 28, border ); + __GLX_PUT_LONG( 32, image_size ); + __glXSendLargeCommand( gc, gc->pc, + __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE + 4, + data, image_size ); + } +} + + +/** + * Internal function used for \c glCompressedTexSubImage1D and + * \c glCompressedTexSubImage2D. + */ +static void +CompressedTexSubImage1D2D( GLenum target, GLint level, + GLsizei xoffset, GLsizei yoffset, + GLsizei width, GLsizei height, + GLenum format, GLsizei image_size, + const GLvoid *data, CARD32 rop ) +{ + __GLX_DECLARE_VARIABLES(); + + __GLX_LOAD_VARIABLES(); + if ( gc->currentDpy == NULL ) { + return; + } + + if ( target == GL_PROXY_TEXTURE_3D ) { + compsize = 0; + } + else { + compsize = image_size; + } + + cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE + + compsize ); + if ( cmdlen <= gc->maxSmallRenderCommandSize ) { + __GLX_BEGIN_VARIABLE( rop, cmdlen ); + __GLX_PUT_LONG( 4, target ); + __GLX_PUT_LONG( 8, level ); + __GLX_PUT_LONG( 12, xoffset ); + __GLX_PUT_LONG( 16, yoffset ); + __GLX_PUT_LONG( 20, width ); + __GLX_PUT_LONG( 24, height ); + __GLX_PUT_LONG( 28, format ); + __GLX_PUT_LONG( 32, image_size ); + if ( compsize != 0 ) { + __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE, + data, image_size ); + } + __GLX_END( cmdlen ); + } + else { + assert( compsize != 0 ); + + __GLX_BEGIN_VARIABLE_LARGE( rop, cmdlen + 4 ); + __GLX_PUT_LONG( 8, target ); + __GLX_PUT_LONG( 12, level ); + __GLX_PUT_LONG( 16, xoffset ); + __GLX_PUT_LONG( 20, yoffset ); + __GLX_PUT_LONG( 24, width ); + __GLX_PUT_LONG( 28, height ); + __GLX_PUT_LONG( 32, format ); + __GLX_PUT_LONG( 36, image_size ); + __glXSendLargeCommand( gc, gc->pc, + __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE + 4, + data, image_size ); + } +} + + +void +__indirect_glCompressedTexImage1DARB( GLenum target, GLint level, + GLenum internal_format, GLsizei width, + GLint border, GLsizei image_size, + const GLvoid *data ) +{ + CompressedTexImage1D2D( target, level, internal_format, width, 0, + border, image_size, data, + X_GLrop_CompressedTexImage1D ); +} + + +void +__indirect_glCompressedTexImage2DARB( GLenum target, GLint level, + GLenum internal_format, + GLsizei width, GLsizei height, + GLint border, GLsizei image_size, + const GLvoid *data ) +{ + CompressedTexImage1D2D( target, level, internal_format, width, height, + border, image_size, data, + X_GLrop_CompressedTexImage2D ); +} + + +void +__indirect_glCompressedTexImage3DARB( GLenum target, GLint level, + GLenum internal_format, + GLsizei width, GLsizei height, GLsizei depth, + GLint border, GLsizei image_size, + const GLvoid *data ) +{ + __GLX_DECLARE_VARIABLES(); + + __GLX_LOAD_VARIABLES(); + if ( gc->currentDpy == NULL ) { + return; + } + + cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE + + image_size ); + if ( cmdlen <= gc->maxSmallRenderCommandSize ) { + __GLX_BEGIN_VARIABLE( X_GLrop_CompressedTexImage3D, cmdlen ); + __GLX_PUT_LONG( 4, target ); + __GLX_PUT_LONG( 8, level ); + __GLX_PUT_LONG( 12, internal_format ); + __GLX_PUT_LONG( 16, width ); + __GLX_PUT_LONG( 20, height ); + __GLX_PUT_LONG( 24, depth ); + __GLX_PUT_LONG( 28, border ); + __GLX_PUT_LONG( 32, image_size ); + if ( image_size != 0 ) { + __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE, + data, image_size ); + } + __GLX_END( cmdlen ); + } + else { + __GLX_BEGIN_VARIABLE_LARGE( X_GLrop_CompressedTexImage3D, + cmdlen + 4 ); + __GLX_PUT_LONG( 8, target ); + __GLX_PUT_LONG( 12, level ); + __GLX_PUT_LONG( 16, internal_format ); + __GLX_PUT_LONG( 20, width ); + __GLX_PUT_LONG( 24, height ); + __GLX_PUT_LONG( 28, depth ); + __GLX_PUT_LONG( 32, border ); + __GLX_PUT_LONG( 36, image_size ); + __glXSendLargeCommand( gc, gc->pc, + __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE + 4, + data, image_size ); + } +} + + +void +__indirect_glCompressedTexSubImage1DARB( GLenum target, GLint level, + GLint xoffset, + GLsizei width, + GLenum format, GLsizei image_size, + const GLvoid *data ) +{ + CompressedTexSubImage1D2D( target, level, xoffset, 0, width, 0, + format, image_size, data, + X_GLrop_CompressedTexSubImage1D ); +} + + +void +__indirect_glCompressedTexSubImage2DARB( GLenum target, GLint level, + GLint xoffset, GLint yoffset, + GLsizei width, GLsizei height, + GLenum format, GLsizei image_size, + const GLvoid *data ) +{ + CompressedTexSubImage1D2D( target, level, xoffset, yoffset, width, height, + format, image_size, data, + X_GLrop_CompressedTexSubImage2D ); +} + + +void +__indirect_glCompressedTexSubImage3DARB( GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLsizei width, GLsizei height, GLsizei depth, + GLenum format, GLsizei image_size, + const GLvoid *data ) +{ + __GLX_DECLARE_VARIABLES(); + + __GLX_LOAD_VARIABLES(); + if ( gc->currentDpy == NULL ) { + return; + } + + cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE + + image_size ); + if ( cmdlen <= gc->maxSmallRenderCommandSize ) { + __GLX_BEGIN_VARIABLE( X_GLrop_CompressedTexSubImage3D, cmdlen ); + __GLX_PUT_LONG( 4, target ); + __GLX_PUT_LONG( 8, level ); + __GLX_PUT_LONG( 12, xoffset ); + __GLX_PUT_LONG( 16, yoffset ); + __GLX_PUT_LONG( 20, zoffset ); + __GLX_PUT_LONG( 24, width ); + __GLX_PUT_LONG( 28, height ); + __GLX_PUT_LONG( 32, depth ); + __GLX_PUT_LONG( 36, format ); + __GLX_PUT_LONG( 40, image_size ); + if ( image_size != 0 ) { + __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE, + data, image_size ); + } + __GLX_END( cmdlen ); + } + else { + __GLX_BEGIN_VARIABLE_LARGE( X_GLrop_CompressedTexSubImage3D, + cmdlen + 4 ); + __GLX_PUT_LONG( 8, target ); + __GLX_PUT_LONG( 12, level ); + __GLX_PUT_LONG( 16, xoffset ); + __GLX_PUT_LONG( 20, yoffset ); + __GLX_PUT_LONG( 24, zoffset ); + __GLX_PUT_LONG( 28, width ); + __GLX_PUT_LONG( 32, height ); + __GLX_PUT_LONG( 36, depth ); + __GLX_PUT_LONG( 40, format ); + __GLX_PUT_LONG( 44, image_size ); + __glXSendLargeCommand( gc, gc->pc, + __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE + 4, + data, image_size ); + } +} diff --git a/src/glx/x11/glxclient.h b/src/glx/x11/glxclient.h index 73c278ee38..03e44e5d04 100644 --- a/src/glx/x11/glxclient.h +++ b/src/glx/x11/glxclient.h @@ -58,7 +58,7 @@ #include "GL/glxproto.h" #include "GL/internal/glcore.h" #include "glapitable.h" -#include "glxhash.h" +#include "glxextensions.h" #if defined( USE_XTHREADS ) # include #elif defined( PTHREADS ) @@ -70,9 +70,7 @@ #define __GLX_MAX_TEXTURE_UNITS 32 -typedef struct __GLXscreenConfigsRec __GLXscreenConfigs; typedef struct __GLXcontextRec __GLXcontext; -typedef struct __GLXdrawableRec __GLXdrawable; typedef struct __GLXdisplayPrivateRec __GLXdisplayPrivate; typedef struct _glapi_table __GLapi; @@ -80,9 +78,6 @@ typedef struct _glapi_table __GLapi; #ifdef GLX_DIRECT_RENDERING -#define containerOf(ptr, type, member) \ - (type *)( (char *)ptr - offsetof(type,member) ) - #include @@ -90,64 +85,43 @@ typedef struct _glapi_table __GLapi; * Display dependent methods. This structure is initialized during the * \c driCreateDisplay call. */ -typedef struct __GLXDRIdisplayRec __GLXDRIdisplay; -typedef struct __GLXDRIscreenRec __GLXDRIscreen; -typedef struct __GLXDRIdrawableRec __GLXDRIdrawable; -typedef struct __GLXDRIcontextRec __GLXDRIcontext; - -#include "glxextensions.h" - -struct __GLXDRIdisplayRec { +struct __DRIdisplayRec { /** * Method to destroy the private DRI display data. */ - void (*destroyDisplay)(__GLXDRIdisplay *display); - - __GLXDRIscreen *(*createScreen)(__GLXscreenConfigs *psc, int screen, - __GLXdisplayPrivate *priv); -}; + void (*destroyDisplay)(Display *dpy, void *displayPrivate); -struct __GLXDRIscreenRec { - - void (*destroyScreen)(__GLXscreenConfigs *psc); - - __GLXDRIcontext *(*createContext)(__GLXscreenConfigs *psc, - const __GLcontextModes *mode, - GLXContext gc, - GLXContext shareList, int renderType); - - __GLXDRIdrawable *(*createDrawable)(__GLXscreenConfigs *psc, - XID drawable, - GLXDrawable glxDrawable, - const __GLcontextModes *modes); -}; + /** + * Opaque pointer to private per display direct rendering data. + * \c NULL if direct rendering is not supported on this display. + */ + struct __DRIdisplayPrivateRec *private; -struct __GLXDRIcontextRec { - void (*destroyContext)(__GLXDRIcontext *context, __GLXscreenConfigs *psc, - Display *dpy); - Bool (*bindContext)(__GLXDRIcontext *context, - __GLXDRIdrawable *pdraw, - __GLXDRIdrawable *pread); - - void (*unbindContext)(__GLXDRIcontext *context); + /** + * Array of pointers to methods to create and initialize the private DRI + * screen data. + */ + PFNCREATENEWSCREENFUNC * createNewScreen; }; -struct __GLXDRIdrawableRec { - void (*destroyDrawable)(__GLXDRIdrawable *drawable); - XID xDrawable; - XID drawable; - __GLXscreenConfigs *psc; - __DRIdrawable *driDrawable; - GLenum textureTarget; +/* +** We keep a linked list of these structures, one per DRI device driver. +*/ +struct __DRIdriverRec { + const char *name; + void *handle; + PFNCREATENEWSCREENFUNC createNewScreenFunc; + struct __DRIdriverRec *next; }; /* ** Function to create and DRI display data and initialize the display ** dependent methods. */ -extern __GLXDRIdisplay *driCreateDisplay(Display *dpy); -extern __GLXDRIdisplay *dri2CreateDisplay(Display *dpy); +extern void *driCreateDisplay(Display *dpy, __DRIdisplay *pdisp); + +extern __DRIdriver *driGetDriver(Display *dpy, int scrNum); extern void DRI_glXUseXFont( Font font, int first, int count, int listbase ); @@ -159,6 +133,8 @@ extern const char *glXGetScreenDriver (Display *dpy, int scrNum); extern const char *glXGetDriverConfig (const char *driverName); +extern Bool __glXWindowExists(Display *dpy, GLXDrawable draw); + #endif /************************************************************************/ @@ -249,11 +225,19 @@ struct __GLXcontextRec { */ XID share_xid; + /** + * Visual id. + * + * \deprecated + * This filed has been largely been replaced by the \c mode field, but + * the work is not quite done. + */ + VisualID vid; + /** * Screen number. */ GLint screen; - __GLXscreenConfigs *psc; /** * \c GL_TRUE if the context was created with ImportContext, which @@ -359,15 +343,24 @@ struct __GLXcontextRec { */ GLint majorOpcode; +#ifdef GLX_DIRECT_RENDERING /** - * Pointer to the mode used to create this context. + * Per context direct rendering interface functions and data. */ - const __GLcontextModes * mode; - -#ifdef GLX_DIRECT_RENDERING - __GLXDRIcontext *driContext; - __DRIcontext *__driContext; + __DRIcontext driContext; #endif + + /** + * \c GLXFBConfigID used to create this context. May be \c None. This + * field has been replaced by the \c mode field. + * + * \since Internal API version 20030317. + * + * \deprecated + * This filed has been largely been replaced by the \c mode field, but + * the work is not quite done. + */ + GLXFBConfigID fbconfigID; /** * The current read-drawable for this context. Will be None if this @@ -445,7 +438,7 @@ extern void __glFreeAttributeState(__GLXcontext *); * One of these records exists per screen of the display. It contains * a pointer to the config data for that screen (if the screen supports GL). */ -struct __GLXscreenConfigsRec { +typedef struct __GLXscreenConfigsRec { /** * GLX extension string reported by the X-server. */ @@ -461,46 +454,13 @@ struct __GLXscreenConfigsRec { /** * Per screen direct rendering interface functions and data. */ - __DRIscreen *__driScreen; - const __DRIcoreExtension *core; - const __DRIlegacyExtension *legacy; - __glxHashTable *drawHash; - Display *dpy; - int scr, fd; - void *driver; - - __GLXDRIscreen *driScreen; - -#ifdef __DRI_COPY_SUB_BUFFER - const __DRIcopySubBufferExtension *copySubBuffer; -#endif - -#ifdef __DRI_SWAP_CONTROL - const __DRIswapControlExtension *swapControl; -#endif - -#ifdef __DRI_ALLOCATE - const __DRIallocateExtension *allocate; -#endif - -#ifdef __DRI_FRAME_TRACKING - const __DRIframeTrackingExtension *frameTracking; -#endif - -#ifdef __DRI_MEDIA_STREAM_COUNTER - const __DRImediaStreamCounterExtension *msc; -#endif - -#ifdef __DRI_TEX_BUFFER - const __DRItexBufferExtension *texBuffer; -#endif - + __DRIscreen driScreen; #endif /** - * Linked list of glx visuals and fbconfigs for this screen. + * Linked list of configurations for this screen. */ - __GLcontextModes *visuals, *configs; + __GLcontextModes *configs; /** * Per-screen dynamic GLX extension tracking. The \c direct_support @@ -514,7 +474,7 @@ struct __GLXscreenConfigsRec { GLboolean ext_list_first_time; /*@}*/ -}; +} __GLXscreenConfigs; /** * Per display private data. One of these records exists for each display @@ -563,11 +523,11 @@ struct __GLXdisplayPrivateRec { /** * Per display direct rendering interface functions and data. */ - __GLXDRIdisplay *driDisplay; - __GLXDRIdisplay *dri2Display; + __DRIdisplay driDisplay; #endif }; +void __glXFreeContext(__GLXcontext*); extern GLubyte *__glXFlushRenderBuffer(__GLXcontext*, GLubyte*); @@ -611,10 +571,6 @@ extern __GLXcontext *__glXcurrentContext; #endif /* defined( USE_XTHREADS ) || defined( PTHREADS ) */ -extern void __glXSetCurrentContextNull(void); - -extern void __glXFreeContext(__GLXcontext*); - /* ** Global lock for all threads in this address space using the GLX @@ -724,16 +680,13 @@ extern char *__glXstrdup(const char *str); extern const char __glXGLClientVersion[]; extern const char __glXGLClientExtensions[]; +/* Determine the internal API version */ +extern int __glXGetInternalVersion(void); + /* Get the unadjusted system time */ extern int __glXGetUST( int64_t * ust ); -extern GLboolean __glXGetMscRateOML(Display * dpy, GLXDrawable drawable, - int32_t * numerator, int32_t * denominator); - -#ifdef GLX_DIRECT_RENDERING -GLboolean -__driGetMscRateOML(__DRIdrawable *draw, - int32_t *numerator, int32_t *denominator, void *private); -#endif +extern Bool __glXGetMscRateOML(Display * dpy, GLXDrawable drawable, + int32_t * numerator, int32_t * denominator); #endif /* !__GLX_client_h__ */ diff --git a/src/glx/x11/glxcmds.c b/src/glx/x11/glxcmds.c index 4345678a98..80281896f6 100644 --- a/src/glx/x11/glxcmds.c +++ b/src/glx/x11/glxcmds.c @@ -38,110 +38,63 @@ * Client-side GLX interface. */ +#include #include "glxclient.h" +#include +#include +#include +#include #include "glapi.h" -#include "glxextensions.h" -#include "glcontextmodes.h" -#include "glheader.h" - #ifdef GLX_DIRECT_RENDERING -#include +#include "indirect_init.h" #include #include "xf86dri.h" #endif +#include "glxextensions.h" +#include "glcontextmodes.h" +#include "glheader.h" +#include static const char __glXGLXClientVendorName[] = "SGI"; static const char __glXGLXClientVersion[] = "1.4"; /****************************************************************************/ - -#ifdef GLX_DIRECT_RENDERING - -static Bool windowExistsFlag; -static int windowExistsErrorHandler(Display *dpy, XErrorEvent *xerr) -{ - if (xerr->error_code == BadWindow) { - windowExistsFlag = GL_FALSE; - } - return 0; -} - -/** - * Find drawables in the local hash that have been destroyed on the - * server. - * - * \param dpy Display to destroy drawables for - * \param screen Screen number to destroy drawables for - */ -static void GarbageCollectDRIDrawables(Display *dpy, __GLXscreenConfigs *sc) -{ - XID draw; - __GLXDRIdrawable *pdraw; - XWindowAttributes xwa; - int (*oldXErrorHandler)(Display *, XErrorEvent *); - - /* Set no-op error handler so Xlib doesn't bail out if the windows - * has alreay been destroyed on the server. */ - XSync(dpy, GL_FALSE); - oldXErrorHandler = XSetErrorHandler(windowExistsErrorHandler); - - if (__glxHashFirst(sc->drawHash, &draw, (void *)&pdraw) == 1) { - do { - windowExistsFlag = GL_TRUE; - XGetWindowAttributes(dpy, draw, &xwa); /* dummy request */ - if (!windowExistsFlag) { - /* Destroy the local drawable data, if the drawable no - longer exists in the Xserver */ - (*pdraw->destroyDrawable)(pdraw); - __glxHashDelete(sc->drawHash, draw); - } - } while (__glxHashNext(sc->drawHash, &draw, (void *)&pdraw) == 1); - } - - XSync(dpy, GL_FALSE); - XSetErrorHandler(oldXErrorHandler); -} - -extern __GLXDRIdrawable * -GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num); - /** * Get the __DRIdrawable for the drawable associated with a GLXContext * * \param dpy The display associated with \c drawable. * \param drawable GLXDrawable whose __DRIdrawable part is to be retrieved. - * \param scrn_num If non-NULL, the drawables screen is stored there * \returns A pointer to the context's __DRIdrawable on success, or NULL if * the drawable is not associated with a direct-rendering context. */ -_X_HIDDEN __GLXDRIdrawable * -GetGLXDRIDrawable(Display *dpy, GLXDrawable drawable, int * const scrn_num) + +#ifdef GLX_DIRECT_RENDERING +static __DRIdrawable * +GetDRIDrawable( Display *dpy, GLXDrawable drawable, int * const scrn_num ) { - __GLXdisplayPrivate *priv = __glXInitialize(dpy); - __GLXDRIdrawable *pdraw; - const unsigned screen_count = ScreenCount(dpy); - unsigned i; - __GLXscreenConfigs *psc; + __GLXdisplayPrivate * const priv = __glXInitialize(dpy); - if (priv == NULL) - return NULL; - - for (i = 0; i < screen_count; i++) { - psc = &priv->screenConfigs[i]; - if (psc->drawHash == NULL) - continue; - - if (__glxHashLookup(psc->drawHash, drawable, (void *) &pdraw) == 0) { - if (scrn_num != NULL) - *scrn_num = i; - return pdraw; + if ( (priv != NULL) && (priv->driDisplay.private != NULL) ) { + const unsigned screen_count = ScreenCount(dpy); + unsigned i; + + for ( i = 0 ; i < screen_count ; i++ ) { + __DRIscreen * const psc = &priv->screenConfigs[i].driScreen; + __DRIdrawable * const pdraw = (psc->private != NULL) + ? (*psc->getDrawable)(dpy, drawable, psc->private) : NULL; + + if ( pdraw != NULL ) { + if ( scrn_num != NULL ) { + *scrn_num = i; + } + return pdraw; + } } } return NULL; } - #endif @@ -311,9 +264,9 @@ GLXContext AllocateGLXContext( Display *dpy ) */ gc->fastImageUnpack = GL_FALSE; gc->fillImage = __glFillImage; + gc->isDirect = GL_FALSE; gc->pc = gc->buf; gc->bufEnd = gc->buf + bufSize; - gc->isDirect = GL_FALSE; if (__glXDebug) { /* ** Set limit register so that there will be one command per packet @@ -359,10 +312,6 @@ CreateContext(Display *dpy, XVisualInfo *vis, Bool use_glx_1_3, int renderType) { GLXContext gc; -#ifdef GLX_DIRECT_RENDERING - int screen = (fbconfig == NULL) ? vis->screen : fbconfig->screen; - __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); -#endif if ( dpy == NULL ) return NULL; @@ -376,36 +325,41 @@ CreateContext(Display *dpy, XVisualInfo *vis, return NULL; #ifdef GLX_DIRECT_RENDERING - if (allowDirect && psc->driScreen) { + if (allowDirect) { + int screen = (fbconfig == NULL) ? vis->screen : fbconfig->screen; + __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); const __GLcontextModes * mode; - if (fbconfig == NULL) { - mode = _gl_context_modes_find_visual(psc->visuals, vis->visualid); - if (mode == NULL) { - xError error; - - error.errorCode = BadValue; - error.resourceID = vis->visualid; - error.sequenceNumber = dpy->request; - error.type = X_Error; - error.majorCode = gc->majorOpcode; - error.minorCode = X_GLXCreateContext; - _XError(dpy, &error); - return None; - } + /* The value of fbconfig cannot change because it is tested + * later in the function. + */ + if ( fbconfig == NULL ) { + /* FIXME: Is it possible for the __GLcontextModes structure + * FIXME: to not be found? + */ + mode = _gl_context_modes_find_visual( psc->configs, + vis->visualid ); + assert( mode != NULL ); + assert( mode->screen == screen ); } else { mode = fbconfig; } - gc->driContext = psc->driScreen->createContext(psc, mode, gc, - shareList, - renderType); - if (gc->driContext != NULL) { - gc->screen = mode->screen; - gc->psc = psc; - gc->mode = mode; - gc->isDirect = GL_TRUE; + if (psc && psc->driScreen.private) { + void * const shared = (shareList != NULL) + ? shareList->driContext.private : NULL; + gc->driContext.private = + (*psc->driScreen.createNewContext)( dpy, mode, renderType, + shared, + &gc->driContext ); + if (gc->driContext.private) { + gc->isDirect = GL_TRUE; + gc->screen = mode->screen; + gc->vid = mode->visualID; + gc->fbconfigID = mode->fbconfigID; + gc->driContext.mode = mode; + } } } #endif @@ -422,7 +376,7 @@ CreateContext(Display *dpy, XVisualInfo *vis, req->visual = vis->visualid; req->screen = vis->screen; req->shareList = shareList ? shareList->xid : None; - req->isDirect = gc->driContext != NULL; + req->isDirect = gc->isDirect; } else if ( use_glx_1_3 ) { xGLXCreateNewContextReq *req; @@ -436,7 +390,7 @@ CreateContext(Display *dpy, XVisualInfo *vis, req->screen = fbconfig->screen; req->renderType = renderType; req->shareList = shareList ? shareList->xid : None; - req->isDirect = gc->driContext != NULL; + req->isDirect = gc->isDirect; } else { xGLXVendorPrivateWithReplyReq *vpreq; @@ -454,7 +408,7 @@ CreateContext(Display *dpy, XVisualInfo *vis, req->screen = fbconfig->screen; req->renderType = renderType; req->shareList = shareList ? shareList->xid : None; - req->isDirect = gc->driContext != NULL; + req->isDirect = gc->isDirect; } UnlockDisplay(dpy); @@ -476,7 +430,7 @@ PUBLIC GLXContext glXCreateContext(Display *dpy, XVisualInfo *vis, False, 0); } -_X_HIDDEN void __glXFreeContext(__GLXcontext *gc) +void __glXFreeContext(__GLXcontext *gc) { if (gc->vendor) XFree((char *) gc->vendor); if (gc->renderer) XFree((char *) gc->renderer); @@ -512,10 +466,12 @@ DestroyContext(Display *dpy, GLXContext gc) #ifdef GLX_DIRECT_RENDERING /* Destroy the direct rendering context */ - if (gc->driContext) { - (*gc->driContext->destroyContext)(gc->driContext, gc->psc, dpy); - gc->driContext = NULL; - GarbageCollectDRIDrawables(dpy, gc->psc); + if (gc->isDirect) { + if (gc->driContext.private) { + (*gc->driContext.destroyContext)(dpy, gc->screen, + gc->driContext.private); + gc->driContext.private = NULL; + } } #endif @@ -596,7 +552,7 @@ PUBLIC void glXWaitGL(void) __glXFlushRenderBuffer(gc, gc->pc); #ifdef GLX_DIRECT_RENDERING - if (gc->driContext) { + if (gc->isDirect) { /* This bit of ugliness unwraps the glFinish function */ #ifdef glFinish #undef glFinish @@ -632,7 +588,7 @@ PUBLIC void glXWaitX(void) __glXFlushRenderBuffer(gc, gc->pc); #ifdef GLX_DIRECT_RENDERING - if (gc->driContext) { + if (gc->isDirect) { XSync(dpy, False); return; } @@ -662,7 +618,7 @@ PUBLIC void glXUseXFont(Font font, int first, int count, int listBase) (void) __glXFlushRenderBuffer(gc, gc->pc); #ifdef GLX_DIRECT_RENDERING - if (gc->driContext) { + if (gc->isDirect) { DRI_glXUseXFont(font, first, count, listBase); return; } @@ -702,7 +658,7 @@ PUBLIC void glXCopyContext(Display *dpy, GLXContext source, } #ifdef GLX_DIRECT_RENDERING - if (gc->driContext) { + if (gc->isDirect) { /* NOT_DONE: This does not work yet */ } #endif @@ -774,7 +730,7 @@ PUBLIC Bool glXIsDirect(Display *dpy, GLXContext gc) if (!gc) { return GL_FALSE; #ifdef GLX_DIRECT_RENDERING - } else if (gc->driContext) { + } else if (gc->isDirect) { return GL_TRUE; #endif } @@ -837,10 +793,10 @@ PUBLIC void glXSwapBuffers(Display *dpy, GLXDrawable drawable) GLXContextTag tag; CARD8 opcode; #ifdef GLX_DIRECT_RENDERING - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL); + __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, NULL ); - if (pdraw != NULL) { - (*pdraw->psc->core->swapBuffers)(pdraw->driDrawable); + if ( pdraw != NULL ) { + (*pdraw->swapBuffers)(dpy, pdraw->private); return; } #endif @@ -884,12 +840,12 @@ PUBLIC int glXGetConfig(Display *dpy, XVisualInfo *vis, int attribute, { __GLXdisplayPrivate *priv; __GLXscreenConfigs *psc; - __GLcontextModes *modes; int status; status = GetGLXPrivScreenConfig( dpy, vis->screen, & priv, & psc ); if ( status == Success ) { - modes = _gl_context_modes_find_visual(psc->visuals, vis->visualid); + const __GLcontextModes * const modes = _gl_context_modes_find_visual( + psc->configs, vis->visualid ); /* Lookup attribute after first finding a match on the visual */ if ( modes != NULL ) { @@ -1267,7 +1223,7 @@ PUBLIC XVisualInfo *glXChooseVisual(Display *dpy, int screen, int *attribList) ** Compute a score for those that do ** Remember which visual, if any, got the highest score */ - for ( modes = psc->visuals ; modes != NULL ; modes = modes->next ) { + for ( modes = psc->configs ; modes != NULL ; modes = modes->next ) { if ( fbconfigs_compatible( & test_config, modes ) && ((best_config == NULL) || (fbconfig_compare( (const __GLcontextModes * const * const)&modes, &best_config ) < 0)) ) { @@ -1312,7 +1268,7 @@ PUBLIC const char *glXQueryExtensionsString( Display *dpy, int screen ) __glXCalculateUsableExtensions(psc, #ifdef GLX_DIRECT_RENDERING - (psc->driScreen != NULL), + (psc->driScreen.private != NULL), #else GL_FALSE, #endif @@ -1491,15 +1447,13 @@ static int __glXQueryContextInfo(Display *dpy, GLXContext ctx) ctx->share_xid = *pProp++; break; case GLX_VISUAL_ID_EXT: - ctx->mode = - _gl_context_modes_find_visual(ctx->psc->visuals, *pProp++); + ctx->vid = *pProp++; break; case GLX_SCREEN: ctx->screen = *pProp++; break; case GLX_FBCONFIG_ID: - ctx->mode = - _gl_context_modes_find_fbconfig(ctx->psc->configs, *pProp++); + ctx->fbconfigID = *pProp++; break; case GLX_RENDER_TYPE: ctx->renderType = *pProp++; @@ -1524,7 +1478,7 @@ glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value) int retVal; /* get the information from the server if we don't have it already */ - if (!ctx->driContext && (ctx->mode == NULL)) { + if (!ctx->isDirect && (ctx->vid == None)) { retVal = __glXQueryContextInfo(dpy, ctx); if (Success != retVal) return retVal; } @@ -1533,13 +1487,13 @@ glXQueryContext(Display *dpy, GLXContext ctx, int attribute, int *value) *value = (int)(ctx->share_xid); break; case GLX_VISUAL_ID_EXT: - *value = ctx->mode ? ctx->mode->visualID : None; + *value = (int)(ctx->vid); break; case GLX_SCREEN: *value = (int)(ctx->screen); break; case GLX_FBCONFIG_ID: - *value = ctx->mode ? ctx->mode->fbconfigID : None; + *value = (int)(ctx->fbconfigID); break; case GLX_RENDER_TYPE: *value = (int)(ctx->renderType); @@ -1637,7 +1591,6 @@ PUBLIC GLXFBConfig *glXGetFBConfigs(Display *dpy, int screen, int *nelements) __GLcontextModes ** config = NULL; int i; - *nelements = 0; if ( (priv->screenConfigs != NULL) && (screen >= 0) && (screen <= ScreenCount(dpy)) && (priv->screenConfigs[screen].configs != NULL) @@ -1662,10 +1615,8 @@ PUBLIC GLXFBConfig *glXGetFBConfigs(Display *dpy, int screen, int *nelements) for ( modes = priv->screenConfigs[screen].configs ; modes != NULL ; modes = modes->next ) { - if ( modes->fbconfigID != GLX_DONT_CARE ) { - config[i] = modes; - i++; - } + config[i] = modes; + i++; } } } @@ -1717,15 +1668,16 @@ static int __glXSwapIntervalSGI(int interval) return GLX_BAD_VALUE; } -#ifdef __DRI_SWAP_CONTROL - if (gc->driContext) { +#ifdef GLX_DIRECT_RENDERING + if ( gc->isDirect ) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(gc->currentDpy, - gc->currentDrawable, - NULL); - if (psc->swapControl != NULL && pdraw != NULL) { - psc->swapControl->setSwapInterval(pdraw->driDrawable, interval); + __DRIdrawable * const pdraw = GetDRIDrawable( gc->currentDpy, + gc->currentDrawable, + NULL ); + if ( __glXExtensionBitIsEnabled( psc, SGI_swap_control_bit ) + && (pdraw != NULL) ) { + pdraw->swap_interval = interval; return 0; } else { @@ -1763,22 +1715,25 @@ static int __glXSwapIntervalSGI(int interval) */ static int __glXSwapIntervalMESA(unsigned int interval) { -#ifdef __DRI_SWAP_CONTROL +#ifdef GLX_DIRECT_RENDERING GLXContext gc = __glXGetCurrentContext(); if ( interval < 0 ) { return GLX_BAD_VALUE; } - if (gc != NULL && gc->driContext) { + if ( (gc != NULL) && gc->isDirect ) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - if ( (psc != NULL) && (psc->driScreen != NULL) ) { - __GLXDRIdrawable *pdraw = - GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL); - if (psc->swapControl != NULL && pdraw != NULL) { - psc->swapControl->setSwapInterval(pdraw->driDrawable, interval); + if ( (psc != NULL) && (psc->driScreen.private != NULL) + && __glXExtensionBitIsEnabled( psc, MESA_swap_control_bit ) ) { + __DRIdrawable * const pdraw = + (*psc->driScreen.getDrawable)(gc->currentDpy, + gc->currentDrawable, + psc->driScreen.private); + if ( pdraw != NULL ) { + pdraw->swap_interval = interval; return 0; } } @@ -1793,18 +1748,21 @@ static int __glXSwapIntervalMESA(unsigned int interval) static int __glXGetSwapIntervalMESA(void) { -#ifdef __DRI_SWAP_CONTROL +#ifdef GLX_DIRECT_RENDERING GLXContext gc = __glXGetCurrentContext(); - if (gc != NULL && gc->driContext) { + if ( (gc != NULL) && gc->isDirect ) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - if ( (psc != NULL) && (psc->driScreen != NULL) ) { - __GLXDRIdrawable *pdraw = - GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL); - if (psc->swapControl != NULL && pdraw != NULL) { - return psc->swapControl->getSwapInterval(pdraw->driDrawable); + if ( (psc != NULL) && (psc->driScreen.private != NULL) + && __glXExtensionBitIsEnabled( psc, MESA_swap_control_bit ) ) { + __DRIdrawable * const pdraw = + (*psc->driScreen.getDrawable)(gc->currentDpy, + gc->currentDrawable, + psc->driScreen.private); + if ( pdraw != NULL ) { + return pdraw->swap_interval; } } } @@ -1821,13 +1779,15 @@ static int __glXGetSwapIntervalMESA(void) static GLint __glXBeginFrameTrackingMESA(Display *dpy, GLXDrawable drawable) { int status = GLX_BAD_CONTEXT; -#ifdef __DRI_FRAME_TRACKING +#ifdef GLX_DIRECT_RENDERING int screen; - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); + __DRIdrawable * const pdraw = GetDRIDrawable(dpy, drawable, & screen); __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); - if (pdraw != NULL && psc->frameTracking != NULL) - status = psc->frameTracking->frameTracking(pdraw->driDrawable, GL_TRUE); + if ( (pdraw != NULL) && (pdraw->frameTracking != NULL) + && __glXExtensionBitIsEnabled( psc, MESA_swap_frame_usage_bit ) ) { + status = pdraw->frameTracking( dpy, pdraw->private, GL_TRUE ); + } #else (void) dpy; (void) drawable; @@ -1839,14 +1799,15 @@ static GLint __glXBeginFrameTrackingMESA(Display *dpy, GLXDrawable drawable) static GLint __glXEndFrameTrackingMESA(Display *dpy, GLXDrawable drawable) { int status = GLX_BAD_CONTEXT; -#ifdef __DRI_FRAME_TRACKING +#ifdef GLX_DIRECT_RENDERING int screen; - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, & screen); - __GLXscreenConfigs *psc = GetGLXScreenConfigs(dpy, screen); + __DRIdrawable * const pdraw = GetDRIDrawable(dpy, drawable, & screen); + __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); - if (pdraw != NULL && psc->frameTracking != NULL) - status = psc->frameTracking->frameTracking(pdraw->driDrawable, - GL_FALSE); + if ( (pdraw != NULL) && (pdraw->frameTracking != NULL) + && __glXExtensionBitIsEnabled( psc, MESA_swap_frame_usage_bit ) ) { + status = pdraw->frameTracking( dpy, pdraw->private, GL_FALSE ); + } #else (void) dpy; (void) drawable; @@ -1859,20 +1820,19 @@ static GLint __glXGetFrameUsageMESA(Display *dpy, GLXDrawable drawable, GLfloat *usage) { int status = GLX_BAD_CONTEXT; -#ifdef __DRI_FRAME_TRACKING +#ifdef GLX_DIRECT_RENDERING int screen; - __GLXDRIdrawable * const pdraw = GetGLXDRIDrawable(dpy, drawable, & screen); + __DRIdrawable * const pdraw = GetDRIDrawable(dpy, drawable, & screen); __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); - if (pdraw != NULL && psc->frameTracking != NULL) { - int64_t sbc, missedFrames; - float lastMissedUsage; + if ( (pdraw != NULL ) && (pdraw->queryFrameTracking != NULL) + && __glXExtensionBitIsEnabled( psc, MESA_swap_frame_usage_bit ) ) { + int64_t sbc, missedFrames; + float lastMissedUsage; - status = psc->frameTracking->queryFrameTracking(pdraw->driDrawable, - &sbc, - &missedFrames, - &lastMissedUsage, - usage); + status = pdraw->queryFrameTracking( dpy, pdraw->private, &sbc, + &missedFrames, &lastMissedUsage, + usage ); } #else (void) dpy; @@ -1888,17 +1848,18 @@ static GLint __glXQueryFrameTrackingMESA(Display *dpy, GLXDrawable drawable, GLfloat *lastMissedUsage) { int status = GLX_BAD_CONTEXT; -#ifdef __DRI_FRAME_TRACKING +#ifdef GLX_DIRECT_RENDERING int screen; - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, & screen); + __DRIdrawable * const pdraw = GetDRIDrawable(dpy, drawable, & screen); __GLXscreenConfigs * const psc = GetGLXScreenConfigs(dpy, screen); - if (pdraw != NULL && psc->frameTracking != NULL) { + if ( (pdraw != NULL ) && (pdraw->queryFrameTracking != NULL) + && __glXExtensionBitIsEnabled( psc, MESA_swap_frame_usage_bit ) ) { float usage; - status = psc->frameTracking->queryFrameTracking(pdraw->driDrawable, - sbc, missedFrames, - lastMissedUsage, &usage); + status = pdraw->queryFrameTracking( dpy, pdraw->private, sbc, + missedFrames, lastMissedUsage, + & usage ); } #else (void) dpy; @@ -1920,24 +1881,21 @@ static int __glXGetVideoSyncSGI(unsigned int *count) * FIXME: there should be a GLX encoding for this call. I can find no * FIXME: documentation for the GLX encoding. */ -#ifdef __DRI_MEDIA_STREAM_COUNTER +#ifdef GLX_DIRECT_RENDERING GLXContext gc = __glXGetCurrentContext(); - if (gc != NULL && gc->driContext) { + if ( (gc != NULL) && gc->isDirect ) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - if ( psc->msc && psc->driScreen ) { - __GLXDRIdrawable *pdraw = - GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL); - int64_t temp; - int ret; - - ret = (*psc->msc->getDrawableMSC)(psc->__driScreen, - pdraw->driDrawable, &temp); - *count = (unsigned) temp; + if ( __glXExtensionBitIsEnabled( psc, SGI_video_sync_bit ) + && psc->driScreen.private && psc->driScreen.getMSC) { + int ret; + int64_t temp; - return (ret == 0) ? 0 : GLX_BAD_CONTEXT; + ret = psc->driScreen.getMSC( psc->driScreen.private, & temp ); + *count = (unsigned) temp; + return (ret == 0) ? 0 : GLX_BAD_CONTEXT; } } #else @@ -1948,26 +1906,32 @@ static int __glXGetVideoSyncSGI(unsigned int *count) static int __glXWaitVideoSyncSGI(int divisor, int remainder, unsigned int *count) { -#ifdef __DRI_MEDIA_STREAM_COUNTER +#ifdef GLX_DIRECT_RENDERING GLXContext gc = __glXGetCurrentContext(); if ( divisor <= 0 || remainder < 0 ) return GLX_BAD_VALUE; - if (gc != NULL && gc->driContext) { + if ( (gc != NULL) && gc->isDirect ) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( gc->currentDpy, gc->screen ); - if (psc->msc != NULL && psc->driScreen ) { - __GLXDRIdrawable *pdraw = - GetGLXDRIDrawable(gc->currentDpy, gc->currentDrawable, NULL); - int ret; - int64_t msc; - int64_t sbc; - - ret = (*psc->msc->waitForMSC)(pdraw->driDrawable, 0, - divisor, remainder, &msc, &sbc); - *count = (unsigned) msc; - return (ret == 0) ? 0 : GLX_BAD_CONTEXT; + if ( __glXExtensionBitIsEnabled( psc, SGI_video_sync_bit ) + && psc->driScreen.private ) { + __DRIdrawable * const pdraw = + (*psc->driScreen.getDrawable)(gc->currentDpy, + gc->currentDrawable, + psc->driScreen.private); + if ( (pdraw != NULL) && (pdraw->waitForMSC != NULL) ) { + int ret; + int64_t msc; + int64_t sbc; + + ret = (*pdraw->waitForMSC)( gc->currentDpy, pdraw->private, + 0, divisor, remainder, + & msc, & sbc ); + *count = (unsigned) msc; + return (ret == 0) ? 0 : GLX_BAD_CONTEXT; + } } } #else @@ -2119,19 +2083,20 @@ static Bool __glXQueryMaxSwapBarriersSGIX(Display *dpy, int screen, int *max) static Bool __glXGetSyncValuesOML(Display *dpy, GLXDrawable drawable, int64_t *ust, int64_t *msc, int64_t *sbc) { -#if defined(__DRI_SWAP_BUFFER_COUNTER) && defined(__DRI_MEDIA_STREAM_COUNTER) +#ifdef GLX_DIRECT_RENDERING __GLXdisplayPrivate * const priv = __glXInitialize(dpy); if ( priv != NULL ) { int i; - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &i); + __DRIdrawable * const pdraw = GetDRIDrawable( dpy, drawable, & i ); __GLXscreenConfigs * const psc = &priv->screenConfigs[i]; assert( (pdraw == NULL) || (i != -1) ); - return ( (pdraw && psc->sbc && psc->msc) - && ((*psc->msc->getMSC)(psc->driScreen, msc) == 0) - && ((*psc->sbc->getSBC)(pdraw->driDrawable, sbc) == 0) - && (__glXGetUST(ust) == 0) ); + return ( (pdraw && pdraw->getSBC && psc->driScreen.getMSC) + && __glXExtensionBitIsEnabled( psc, OML_sync_control_bit ) + && ((*psc->driScreen.getMSC)( psc->driScreen.private, msc ) == 0) + && ((*pdraw->getSBC)( dpy, psc->driScreen.private, sbc ) == 0) + && (__glXGetUST( ust ) == 0) ); } #else (void) dpy; @@ -2143,68 +2108,6 @@ static Bool __glXGetSyncValuesOML(Display *dpy, GLXDrawable drawable, return False; } -#ifdef GLX_DIRECT_RENDERING -_X_HIDDEN GLboolean -__driGetMscRateOML(__DRIdrawable *draw, - int32_t *numerator, int32_t *denominator, void *private) -{ -#ifdef XF86VIDMODE - __GLXscreenConfigs *psc; - XF86VidModeModeLine mode_line; - int dot_clock; - int i; - __GLXDRIdrawable *glxDraw = private; - - psc = glxDraw->psc; - if (XF86VidModeQueryVersion(psc->dpy, &i, &i) && - XF86VidModeGetModeLine(psc->dpy, psc->scr, &dot_clock, &mode_line) ) { - unsigned n = dot_clock * 1000; - unsigned d = mode_line.vtotal * mode_line.htotal; - -# define V_INTERLACE 0x010 -# define V_DBLSCAN 0x020 - - if (mode_line.flags & V_INTERLACE) - n *= 2; - else if (mode_line.flags & V_DBLSCAN) - d *= 2; - - /* The OML_sync_control spec requires that if the refresh rate is a - * whole number, that the returned numerator be equal to the refresh - * rate and the denominator be 1. - */ - - if (n % d == 0) { - n /= d; - d = 1; - } - else { - static const unsigned f[] = { 13, 11, 7, 5, 3, 2, 0 }; - - /* This is a poor man's way to reduce a fraction. It's far from - * perfect, but it will work well enough for this situation. - */ - - for (i = 0; f[i] != 0; i++) { - while (n % f[i] == 0 && d % f[i] == 0) { - d /= f[i]; - n /= f[i]; - } - } - } - - *numerator = n; - *denominator = d; - - return True; - } - else - return False; -#else - return False; -#endif -} -#endif /** * Determine the refresh rate of the specified drawable and display. @@ -2222,17 +2125,70 @@ __driGetMscRateOML(__DRIdrawable *draw, * when GLX_OML_sync_control appears in the client extension string. */ -_X_HIDDEN GLboolean __glXGetMscRateOML(Display * dpy, GLXDrawable drawable, - int32_t * numerator, - int32_t * denominator) +Bool __glXGetMscRateOML(Display * dpy, GLXDrawable drawable, + int32_t * numerator, int32_t * denominator) { #if defined( GLX_DIRECT_RENDERING ) && defined( XF86VIDMODE ) - __GLXDRIdrawable *draw = GetGLXDRIDrawable(dpy, drawable, NULL); + __GLXdisplayPrivate * const priv = __glXInitialize(dpy); - if (draw == NULL) - return False; - return __driGetMscRateOML(draw->driDrawable, numerator, denominator, draw); + if ( priv != NULL ) { + XF86VidModeModeLine mode_line; + int dot_clock; + int screen_num; + int i; + + + GetDRIDrawable( dpy, drawable, & screen_num ); + if ( (screen_num != -1) + && XF86VidModeQueryVersion( dpy, & i, & i ) + && XF86VidModeGetModeLine( dpy, screen_num, & dot_clock, + & mode_line ) ) { + unsigned n = dot_clock * 1000; + unsigned d = mode_line.vtotal * mode_line.htotal; + +# define V_INTERLACE 0x010 +# define V_DBLSCAN 0x020 + + if ( (mode_line.flags & V_INTERLACE) ) { + n *= 2; + } + else if ( (mode_line.flags & V_DBLSCAN) ) { + d *= 2; + } + + /* The OML_sync_control spec requires that if the refresh rate is a + * whole number, that the returned numerator be equal to the refresh + * rate and the denominator be 1. + */ + + if ( (n % d) == 0 ) { + n /= d; + d = 1; + } + else { + static const unsigned f[] = { 13, 11, 7, 5, 3, 2, 0 }; + + + /* This is a poor man's way to reduce a fraction. It's far from + * perfect, but it will work well enough for this situation. + */ + + for ( i = 0 ; f[i] != 0 ; i++ ) { + while ( ((n % f[i]) == 0) && ((d % f[i]) == 0) ) { + d /= f[i]; + n /= f[i]; + } + } + } + + *numerator = n; + *denominator = d; + + (void) drawable; + return True; + } + } #else (void) dpy; (void) drawable; @@ -2247,9 +2203,9 @@ static int64_t __glXSwapBuffersMscOML(Display *dpy, GLXDrawable drawable, int64_t target_msc, int64_t divisor, int64_t remainder) { -#ifdef __DRI_SWAP_BUFFER_COUNTER +#ifdef GLX_DIRECT_RENDERING int screen; - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); + __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, & screen ); __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen ); /* The OML_sync_control spec says these should "generate a GLX_BAD_VALUE @@ -2262,10 +2218,11 @@ static int64_t __glXSwapBuffersMscOML(Display *dpy, GLXDrawable drawable, if ( divisor > 0 && remainder >= divisor ) return -1; - if (pdraw != NULL && psc->counters != NULL) - return (*psc->sbc->swapBuffersMSC)(pdraw->driDrawable, target_msc, - divisor, remainder); - + if ( (pdraw != NULL) && (pdraw->swapBuffersMSC != NULL) + && __glXExtensionBitIsEnabled( psc, OML_sync_control_bit ) ) { + return (*pdraw->swapBuffersMSC)(dpy, pdraw->private, target_msc, + divisor, remainder); + } #else (void) dpy; (void) drawable; @@ -2282,9 +2239,9 @@ static Bool __glXWaitForMscOML(Display * dpy, GLXDrawable drawable, int64_t remainder, int64_t *ust, int64_t *msc, int64_t *sbc) { -#ifdef __DRI_MEDIA_STREAM_COUNTER +#ifdef GLX_DIRECT_RENDERING int screen; - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); + __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, & screen ); __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen ); int ret; @@ -2296,9 +2253,10 @@ static Bool __glXWaitForMscOML(Display * dpy, GLXDrawable drawable, if ( divisor > 0 && remainder >= divisor ) return False; - if (pdraw != NULL && psc->msc != NULL) { - ret = (*psc->msc->waitForMSC)(pdraw->driDrawable, target_msc, - divisor, remainder, msc, sbc); + if ( (pdraw != NULL) && (pdraw->waitForMSC != NULL) + && __glXExtensionBitIsEnabled( psc, OML_sync_control_bit ) ) { + ret = (*pdraw->waitForMSC)( dpy, pdraw->private, target_msc, + divisor, remainder, msc, sbc ); /* __glXGetUST returns zero on success and non-zero on failure. * This function returns True on success and False on failure. @@ -2323,9 +2281,9 @@ static Bool __glXWaitForSbcOML(Display * dpy, GLXDrawable drawable, int64_t target_sbc, int64_t *ust, int64_t *msc, int64_t *sbc ) { -#ifdef __DRI_SWAP_BUFFER_COUNTER +#ifdef GLX_DIRECT_RENDERING int screen; - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); + __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, & screen ); __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen ); int ret; @@ -2335,8 +2293,9 @@ static Bool __glXWaitForSbcOML(Display * dpy, GLXDrawable drawable, if ( target_sbc < 0 ) return False; - if (pdraw != NULL && psc->sbc != NULL) { - ret = (*psc->sbc->waitForSBC)(pdraw->driDrawable, target_sbc, msc, sbc); + if ( (pdraw != NULL) && (pdraw->waitForSBC != NULL) + && __glXExtensionBitIsEnabled( psc, OML_sync_control_bit )) { + ret = (*pdraw->waitForSBC)( dpy, pdraw->private, target_sbc, msc, sbc ); /* __glXGetUST returns zero on success and non-zero on failure. * This function returns True on success and False on failure. @@ -2364,13 +2323,16 @@ PUBLIC void *glXAllocateMemoryMESA(Display *dpy, int scrn, size_t size, float readFreq, float writeFreq, float priority) { -#ifdef __DRI_ALLOCATE +#ifdef GLX_DIRECT_RENDERING __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, scrn ); - if (psc && psc->allocate) - return (*psc->allocate->allocateMemory)(psc->__driScreen, size, - readFreq, writeFreq, priority); - + if ( __glXExtensionBitIsEnabled( psc, MESA_allocate_memory_bit ) ) { + if (psc && psc->driScreen.private && psc->driScreen.allocateMemory) { + return (*psc->driScreen.allocateMemory)( dpy, scrn, size, + readFreq, writeFreq, + priority ); + } + } #else (void) dpy; (void) scrn; @@ -2386,12 +2348,14 @@ PUBLIC void *glXAllocateMemoryMESA(Display *dpy, int scrn, PUBLIC void glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer) { -#ifdef __DRI_ALLOCATE +#ifdef GLX_DIRECT_RENDERING __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, scrn ); - if (psc && psc->allocate) - (*psc->allocate->freeMemory)(psc->__driScreen, pointer); - + if ( __glXExtensionBitIsEnabled( psc, MESA_allocate_memory_bit ) ) { + if (psc && psc->driScreen.private && psc->driScreen.freeMemory) { + (*psc->driScreen.freeMemory)( dpy, scrn, pointer ); + } + } #else (void) dpy; (void) scrn; @@ -2403,12 +2367,14 @@ PUBLIC void glXFreeMemoryMESA(Display *dpy, int scrn, void *pointer) PUBLIC GLuint glXGetMemoryOffsetMESA( Display *dpy, int scrn, const void *pointer ) { -#ifdef __DRI_ALLOCATE +#ifdef GLX_DIRECT_RENDERING __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, scrn ); - if (psc && psc->allocate) - return (*psc->allocate->memoryOffset)(psc->__driScreen, pointer); - + if ( __glXExtensionBitIsEnabled( psc, MESA_allocate_memory_bit ) ) { + if (psc && psc->driScreen.private && psc->driScreen.memoryOffset) { + return (*psc->driScreen.memoryOffset)( dpy, scrn, pointer ); + } + } #else (void) dpy; (void) scrn; @@ -2481,14 +2447,13 @@ static void __glXCopySubBufferMESA(Display *dpy, GLXDrawable drawable, INT32 *x_ptr, *y_ptr, *w_ptr, *h_ptr; CARD8 opcode; -#ifdef __DRI_COPY_SUB_BUFFER +#ifdef GLX_DIRECT_RENDERING int screen; - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, &screen); + __DRIdrawable *pdraw = GetDRIDrawable( dpy, drawable, & screen ); if ( pdraw != NULL ) { __GLXscreenConfigs * const psc = GetGLXScreenConfigs( dpy, screen ); - if (psc->copySubBuffer != NULL) { - (*psc->copySubBuffer->copySubBuffer)(pdraw->driDrawable, - x, y, width, height); + if ( __glXExtensionBitIsEnabled( psc, MESA_copy_sub_buffer_bit ) ) { + (*pdraw->copySubBuffer)(dpy, pdraw->private, x, y, width, height); } return; @@ -2564,16 +2529,8 @@ static void __glXBindTexImageEXT(Display *dpy, } #ifdef GLX_DIRECT_RENDERING - if (gc->driContext) { - __GLXDRIdrawable *pdraw = GetGLXDRIDrawable(dpy, drawable, NULL); - - if (pdraw != NULL) - (*pdraw->psc->texBuffer->setTexBuffer)(gc->__driContext, - pdraw->textureTarget, - pdraw->driDrawable); - + if (gc->isDirect) return; - } #endif opcode = __glXSetupForCommand(dpy); @@ -2624,7 +2581,7 @@ static void __glXReleaseTexImageEXT(Display *dpy, return; #ifdef GLX_DIRECT_RENDERING - if (gc->driContext) + if (gc->isDirect) return; #endif @@ -2656,7 +2613,7 @@ static void __glXReleaseTexImageEXT(Display *dpy, * * \sa strdup */ -_X_HIDDEN char * +char * __glXstrdup(const char *str) { char *copy; @@ -2883,6 +2840,98 @@ PUBLIC void (*glXGetProcAddress(const GLubyte *procName))( void ) #ifdef GLX_DIRECT_RENDERING +/** + * Retrieves the verion of the internal libGL API in YYYYMMDD format. This + * might be used by the DRI drivers to determine how new libGL is at runtime. + * Drivers should not call this function directly. They should instead use + * \c glXGetProcAddress to obtain a pointer to the function. + * + * \returns An 8-digit decimal number representing the internal libGL API in + * YYYYMMDD format. + * + * \sa glXGetProcAddress, PFNGLXGETINTERNALVERSIONPROC + * + * \since Internal API version 20021121. + */ +int __glXGetInternalVersion(void) +{ + /* History: + * 20021121 - Initial version + * 20021128 - Added __glXWindowExists() function + * 20021207 - Added support for dynamic GLX extensions, + * GLX_SGI_swap_control, GLX_SGI_video_sync, + * GLX_OML_sync_control, and GLX_MESA_swap_control. + * Never officially released. Do NOT test against + * this version. Use 20030317 instead. + * 20030317 - Added support GLX_SGIX_fbconfig, + * GLX_MESA_swap_frame_usage, GLX_OML_swap_method, + * GLX_{ARB,SGIS}_multisample, and + * GLX_SGIX_visual_select_group. + * 20030606 - Added support for GLX_SGI_make_current_read. + * 20030813 - Made support for dynamic extensions multi-head aware. + * 20030818 - Added support for GLX_MESA_allocate_memory in place of the + * deprecated GLX_NV_vertex_array_range & GLX_MESA_agp_offset + * interfaces. + * 20031201 - Added support for the first round of DRI interface changes. + * Do NOT test against this version! It has binary + * compatibility bugs, use 20040317 instead. + * 20040317 - Added the 'mode' field to __DRIcontextRec. + * 20040415 - Added support for bindContext3 and unbindContext3. + * 20040602 - Add __glXGetDrawableInfo. I though that was there + * months ago. :( + * 20050727 - Gut all the old interfaces. This breaks compatability with + * any DRI driver built to any previous version. + * 20060314 - Added support for GLX_MESA_copy_sub_buffer. + * 20070105 - Added support for damage reporting. + */ + return 20070105; +} + + + +static Bool windowExistsFlag; + +static int windowExistsErrorHandler(Display *dpy, XErrorEvent *xerr) +{ + if (xerr->error_code == BadWindow) { + windowExistsFlag = GL_FALSE; + } + return 0; +} + +/** + * Determine if a window associated with a \c GLXDrawable exists on the + * X-server. This function is not used internally by libGL. It is provided + * as a utility function for DRI drivers. + * Drivers should not call this function directly. They should instead use + * \c glXGetProcAddress to obtain a pointer to the function. + * + * \param dpy Display associated with the drawable to be queried. + * \param draw \c GLXDrawable to test. + * + * \returns \c GL_TRUE if a window exists that is associated with \c draw, + * otherwise \c GL_FALSE is returned. + * + * \warning This function is not currently thread-safe. + * + * \sa glXGetProcAddress + * + * \since Internal API version 20021128. + */ +Bool __glXWindowExists(Display *dpy, GLXDrawable draw) +{ + XWindowAttributes xwa; + int (*oldXErrorHandler)(Display *, XErrorEvent *); + + XSync(dpy, GL_FALSE); + windowExistsFlag = GL_TRUE; + oldXErrorHandler = XSetErrorHandler(windowExistsErrorHandler); + XGetWindowAttributes(dpy, draw, &xwa); /* dummy request */ + XSetErrorHandler(oldXErrorHandler); + return windowExistsFlag; +} + + /** * Get the unadjusted system time (UST). Currently, the UST is measured in * microseconds since Epoc. The actual resolution of the UST may vary from @@ -2897,7 +2946,7 @@ PUBLIC void (*glXGetProcAddress(const GLubyte *procName))( void ) * * \since Internal API version 20030317. */ -_X_HIDDEN int __glXGetUST( int64_t * ust ) +int __glXGetUST( int64_t * ust ) { struct timeval tv; diff --git a/src/glx/x11/glxcurrent.c b/src/glx/x11/glxcurrent.c deleted file mode 100644 index ad648fd438..0000000000 --- a/src/glx/x11/glxcurrent.c +++ /dev/null @@ -1,510 +0,0 @@ -/* -** License Applicability. Except to the extent portions of this file are -** made subject to an alternative license as permitted in the SGI Free -** Software License B, Version 1.1 (the "License"), the contents of this -** file are subject only to the provisions of the License. You may not use -** this file except in compliance with the License. You may obtain a copy -** of the License at Silicon Graphics, Inc., attn: Legal Services, 1600 -** Amphitheatre Parkway, Mountain View, CA 94043-1351, or at: -** -** http://oss.sgi.com/projects/FreeB -** -** Note that, as provided in the License, the Software is distributed on an -** "AS IS" basis, with ALL EXPRESS AND IMPLIED WARRANTIES AND CONDITIONS -** DISCLAIMED, INCLUDING, WITHOUT LIMITATION, ANY IMPLIED WARRANTIES AND -** CONDITIONS OF MERCHANTABILITY, SATISFACTORY QUALITY, FITNESS FOR A -** PARTICULAR PURPOSE, AND NON-INFRINGEMENT. -** -** Original Code. The Original Code is: OpenGL Sample Implementation, -** Version 1.2.1, released January 26, 2000, developed by Silicon Graphics, -** Inc. The Original Code is Copyright (c) 1991-2000 Silicon Graphics, Inc. -** Copyright in any portions created by third parties is as indicated -** elsewhere herein. All Rights Reserved. -** -** Additional Notice Provisions: The application programming interfaces -** established by SGI in conjunction with the Original Code are The -** OpenGL(R) Graphics System: A Specification (Version 1.2.1), released -** April 1, 1999; The OpenGL(R) Graphics System Utility Library (Version -** 1.3), released November 4, 1998; and OpenGL(R) Graphics with the X -** Window System(R) (Version 1.3), released October 19, 1998. This software -** was created using the OpenGL(R) version 1.2.1 Sample Implementation -** published by SGI, but has not been independently verified as being -** compliant with the OpenGL(R) version 1.2.1 Specification. -** -*/ - -/** - * \file glxcurrent.c - * Client-side GLX interface for current context management. - */ - -#include "glxclient.h" -#include "glapi.h" -#include "glheader.h" -#include "indirect_init.h" - -#ifdef GLX_DIRECT_RENDERING -#include "xf86dri.h" -#endif - -/* -** We setup some dummy structures here so that the API can be used -** even if no context is current. -*/ - -static GLubyte dummyBuffer[__GLX_BUFFER_LIMIT_SIZE]; - -/* -** Dummy context used by small commands when there is no current context. -** All the -** gl and glx entry points are designed to operate as nop's when using -** the dummy context structure. -*/ -static __GLXcontext dummyContext = { - &dummyBuffer[0], - &dummyBuffer[0], - &dummyBuffer[0], - &dummyBuffer[__GLX_BUFFER_LIMIT_SIZE], - sizeof(dummyBuffer), -}; - - -/* -** All indirect rendering contexts will share the same indirect dispatch table. -*/ -static __GLapi *IndirectAPI = NULL; - - -/* - * Current context management and locking - */ - -#if defined( USE_XTHREADS ) - -/* thread safe */ -static GLboolean TSDinitialized = GL_FALSE; -static xthread_key_t ContextTSD; - -_X_HIDDEN __GLXcontext *__glXGetCurrentContext(void) -{ - if (!TSDinitialized) { - xthread_key_create(&ContextTSD, NULL); - TSDinitialized = GL_TRUE; - return &dummyContext; - } - else { - void *p; - xthread_get_specific(ContextTSD, &p); - if (!p) - return &dummyContext; - else - return (__GLXcontext *) p; - } -} - -_X_HIDDEN void __glXSetCurrentContext(__GLXcontext *c) -{ - if (!TSDinitialized) { - xthread_key_create(&ContextTSD, NULL); - TSDinitialized = GL_TRUE; - } - xthread_set_specific(ContextTSD, c); -} - - -/* Used by the __glXLock() and __glXUnlock() macros */ -_X_HIDDEN xmutex_rec __glXmutex; - -#elif defined( PTHREADS ) - -_X_HIDDEN pthread_mutex_t __glXmutex = PTHREAD_MUTEX_INITIALIZER; - -# if defined( GLX_USE_TLS ) - -/** - * Per-thread GLX context pointer. - * - * \c __glXSetCurrentContext is written is such a way that this pointer can - * \b never be \c NULL. This is important! Because of this - * \c __glXGetCurrentContext can be implemented as trivial macro. - */ -__thread void * __glX_tls_Context __attribute__((tls_model("initial-exec"))) - = &dummyContext; - -_X_HIDDEN void __glXSetCurrentContext( __GLXcontext * c ) -{ - __glX_tls_Context = (c != NULL) ? c : &dummyContext; -} - -# else - -static pthread_once_t once_control = PTHREAD_ONCE_INIT; - -/** - * Per-thread data key. - * - * Once \c init_thread_data has been called, the per-thread data key will - * take a value of \c NULL. As each new thread is created the default - * value, in that thread, will be \c NULL. - */ -static pthread_key_t ContextTSD; - -/** - * Initialize the per-thread data key. - * - * This function is called \b exactly once per-process (not per-thread!) to - * initialize the per-thread data key. This is ideally done using the - * \c pthread_once mechanism. - */ -static void init_thread_data( void ) -{ - if ( pthread_key_create( & ContextTSD, NULL ) != 0 ) { - perror( "pthread_key_create" ); - exit( -1 ); - } -} - -_X_HIDDEN void __glXSetCurrentContext( __GLXcontext * c ) -{ - pthread_once( & once_control, init_thread_data ); - pthread_setspecific( ContextTSD, c ); -} - -_X_HIDDEN __GLXcontext * __glXGetCurrentContext( void ) -{ - void * v; - - pthread_once( & once_control, init_thread_data ); - - v = pthread_getspecific( ContextTSD ); - return (v == NULL) ? & dummyContext : (__GLXcontext *) v; -} - -# endif /* defined( GLX_USE_TLS ) */ - -#elif defined( THREADS ) - -#error Unknown threading method specified. - -#else - -/* not thread safe */ -_X_HIDDEN __GLXcontext *__glXcurrentContext = &dummyContext; - -#endif - - -_X_HIDDEN void __glXSetCurrentContextNull(void) -{ - __glXSetCurrentContext(&dummyContext); -#ifdef GLX_DIRECT_RENDERING - _glapi_set_dispatch(NULL); /* no-op functions */ -#endif -} - - -/************************************************************************/ - -PUBLIC GLXContext glXGetCurrentContext(void) -{ - GLXContext cx = __glXGetCurrentContext(); - - if (cx == &dummyContext) { - return NULL; - } else { - return cx; - } -} - -PUBLIC GLXDrawable glXGetCurrentDrawable(void) -{ - GLXContext gc = __glXGetCurrentContext(); - return gc->currentDrawable; -} - - -/************************************************************************/ - -/** - * Sends a GLX protocol message to the specified display to make the context - * and the drawables current. - * - * \param dpy Display to send the message to. - * \param opcode Major opcode value for the display. - * \param gc_id Context tag for the context to be made current. - * \param draw Drawable ID for the "draw" drawable. - * \param read Drawable ID for the "read" drawable. - * \param reply Space to store the X-server's reply. - * - * \warning - * This function assumes that \c dpy is locked with \c LockDisplay on entry. - */ -static Bool SendMakeCurrentRequest(Display *dpy, CARD8 opcode, - GLXContextID gc_id, GLXContextTag gc_tag, - GLXDrawable draw, GLXDrawable read, - xGLXMakeCurrentReply *reply) -{ - Bool ret; - - - LockDisplay(dpy); - - if (draw == read) { - xGLXMakeCurrentReq *req; - - GetReq(GLXMakeCurrent,req); - req->reqType = opcode; - req->glxCode = X_GLXMakeCurrent; - req->drawable = draw; - req->context = gc_id; - req->oldContextTag = gc_tag; - } - else { - __GLXdisplayPrivate *priv = __glXInitialize(dpy); - - /* If the server can support the GLX 1.3 version, we should - * perfer that. Not only that, some servers support GLX 1.3 but - * not the SGI extension. - */ - - if ((priv->majorVersion > 1) || (priv->minorVersion >= 3)) { - xGLXMakeContextCurrentReq *req; - - GetReq(GLXMakeContextCurrent,req); - req->reqType = opcode; - req->glxCode = X_GLXMakeContextCurrent; - req->drawable = draw; - req->readdrawable = read; - req->context = gc_id; - req->oldContextTag = gc_tag; - } - else { - xGLXVendorPrivateWithReplyReq *vpreq; - xGLXMakeCurrentReadSGIReq *req; - - GetReqExtra(GLXVendorPrivateWithReply, - sz_xGLXMakeCurrentReadSGIReq-sz_xGLXVendorPrivateWithReplyReq,vpreq); - req = (xGLXMakeCurrentReadSGIReq *)vpreq; - req->reqType = opcode; - req->glxCode = X_GLXVendorPrivateWithReply; - req->vendorCode = X_GLXvop_MakeCurrentReadSGI; - req->drawable = draw; - req->readable = read; - req->context = gc_id; - req->oldContextTag = gc_tag; - } - } - - ret = _XReply(dpy, (xReply*) reply, 0, False); - - UnlockDisplay(dpy); - SyncHandle(); - - return ret; -} - - -#ifdef GLX_DIRECT_RENDERING -static __GLXDRIdrawable * -FetchDRIDrawable(Display *dpy, - GLXDrawable glxDrawable, GLXContext gc, Bool pre13) -{ - __GLXdisplayPrivate * const priv = __glXInitialize(dpy); - __GLXDRIdrawable *pdraw; - __GLXscreenConfigs *psc; - XID drawable; - - if (priv == NULL) - return NULL; - - psc = &priv->screenConfigs[gc->screen]; - if (psc->drawHash == NULL) - return NULL; - - if (__glxHashLookup(psc->drawHash, glxDrawable, (void *) &pdraw) == 0) - return pdraw; - - /* If this is glXMakeCurrent (pre GLX 1.3) we allow creating the - * GLX drawable on the fly. Otherwise we pass None as the X - * drawable */ - if (pre13) - drawable = glxDrawable; - else - drawable = None; - - pdraw = psc->driScreen->createDrawable(psc, drawable, - glxDrawable, gc->mode); - if (__glxHashInsert(psc->drawHash, glxDrawable, pdraw)) { - (*pdraw->destroyDrawable)(pdraw); - return NULL; - } - - return pdraw; -} -#endif /* GLX_DIRECT_RENDERING */ - - -/** - * Make a particular context current. - * - * \note This is in this file so that it can access dummyContext. - */ -static Bool MakeContextCurrent(Display *dpy, GLXDrawable draw, - GLXDrawable read, GLXContext gc, - Bool pre13) -{ - xGLXMakeCurrentReply reply; - const GLXContext oldGC = __glXGetCurrentContext(); - const CARD8 opcode = __glXSetupForCommand(dpy); - const CARD8 oldOpcode = ((gc == oldGC) || (oldGC == &dummyContext)) - ? opcode : __glXSetupForCommand(oldGC->currentDpy); - Bool bindReturnValue; - - - if (!opcode || !oldOpcode) { - return GL_FALSE; - } - - /* Make sure that the new context has a nonzero ID. In the request, - * a zero context ID is used only to mean that we bind to no current - * context. - */ - if ((gc != NULL) && (gc->xid == None)) { - return GL_FALSE; - } - - _glapi_check_multithread(); - -#ifdef GLX_DIRECT_RENDERING - /* Bind the direct rendering context to the drawable */ - if (gc && gc->driContext) { - __GLXDRIdrawable *pdraw = FetchDRIDrawable(dpy, draw, gc, pre13); - __GLXDRIdrawable *pread = FetchDRIDrawable(dpy, read, gc, pre13); - - bindReturnValue = - (gc->driContext->bindContext) (gc->driContext, pdraw, pread); - } else -#endif - { - /* Send a glXMakeCurrent request to bind the new context. */ - bindReturnValue = - SendMakeCurrentRequest(dpy, opcode, gc ? gc->xid : None, - ((dpy != oldGC->currentDpy) || oldGC->isDirect) - ? None : oldGC->currentContextTag, - draw, read, &reply); - } - - - if (!bindReturnValue) { - return False; - } - - if ((dpy != oldGC->currentDpy || (gc && gc->driContext)) && - !oldGC->isDirect && oldGC != &dummyContext) { - xGLXMakeCurrentReply dummy_reply; - - /* We are either switching from one dpy to another and have to - * send a request to the previous dpy to unbind the previous - * context, or we are switching away from a indirect context to - * a direct context and have to send a request to the dpy to - * unbind the previous context. - */ - (void) SendMakeCurrentRequest(oldGC->currentDpy, oldOpcode, None, - oldGC->currentContextTag, None, None, - & dummy_reply); - } -#ifdef GLX_DIRECT_RENDERING - else if (oldGC->driContext) { - oldGC->driContext->unbindContext(oldGC->driContext); - } -#endif - - - /* Update our notion of what is current */ - __glXLock(); - if (gc == oldGC) { - /* Even though the contexts are the same the drawable might have - * changed. Note that gc cannot be the dummy, and that oldGC - * cannot be NULL, therefore if they are the same, gc is not - * NULL and not the dummy. - */ - gc->currentDrawable = draw; - gc->currentReadable = read; - } else { - if (oldGC != &dummyContext) { - /* Old current context is no longer current to anybody */ - oldGC->currentDpy = 0; - oldGC->currentDrawable = None; - oldGC->currentReadable = None; - oldGC->currentContextTag = 0; - - if (oldGC->xid == None) { - /* We are switching away from a context that was - * previously destroyed, so we need to free the memory - * for the old handle. - */ -#ifdef GLX_DIRECT_RENDERING - /* Destroy the old direct rendering context */ - if (oldGC->driContext) { - oldGC->driContext->destroyContext(oldGC->driContext, - oldGC->psc, - oldGC->createDpy); - oldGC->driContext = NULL; - } -#endif - __glXFreeContext(oldGC); - } - } - if (gc) { - __glXSetCurrentContext(gc); - - gc->currentDpy = dpy; - gc->currentDrawable = draw; - gc->currentReadable = read; - - if (!gc->driContext) { - if (!IndirectAPI) - IndirectAPI = __glXNewIndirectAPI(); - _glapi_set_dispatch(IndirectAPI); - -#ifdef GLX_USE_APPLEGL - do { - extern void XAppleDRIUseIndirectDispatch(void); - XAppleDRIUseIndirectDispatch(); - } while (0); -#endif - - __GLXattribute *state = - (__GLXattribute *)(gc->client_state_private); - - gc->currentContextTag = reply.contextTag; - if (state->array_state == NULL) { - (void) glGetString(GL_EXTENSIONS); - (void) glGetString(GL_VERSION); - __glXInitVertexArrayState(gc); - } - } - else { - gc->currentContextTag = -1; - } - } else { - __glXSetCurrentContextNull(); - } - } - __glXUnlock(); - return GL_TRUE; -} - - -PUBLIC Bool glXMakeCurrent(Display *dpy, GLXDrawable draw, GLXContext gc) -{ - return MakeContextCurrent(dpy, draw, draw, gc, True); -} - -PUBLIC GLX_ALIAS(Bool, glXMakeCurrentReadSGI, - (Display *dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx), - (dpy, d, r, ctx, False), MakeContextCurrent) - -PUBLIC GLX_ALIAS(Bool, glXMakeContextCurrent, - (Display *dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx), - (dpy, d, r, ctx, False), MakeContextCurrent) diff --git a/src/glx/x11/glxext.c b/src/glx/x11/glxext.c index 4d814744cd..6403cbd56d 100644 --- a/src/glx/x11/glxext.c +++ b/src/glx/x11/glxext.c @@ -44,25 +44,57 @@ */ #include "glxclient.h" +#include #include #include +#include +#include +#include +#include "indirect_init.h" #include "glapi.h" #include "glxextensions.h" #include "glcontextmodes.h" #include "glheader.h" +#ifdef GLX_DIRECT_RENDERING +#include +#include +#include "xf86dri.h" +#include "sarea.h" +#include "dri_glx.h" +#endif + #ifdef USE_XCB #include #include #include #endif +#include #ifdef DEBUG void __glXDumpDrawBuffer(__GLXcontext *ctx); #endif #ifdef USE_SPARC_ASM +/* + * This is where our dispatch table's bounds are. + * And the static mesa_init is taken directly from + * Mesa's 'sparc.c' initializer. + * + * We need something like this here, because this version + * of openGL/glx never initializes a Mesa context, and so + * the address of the dispatch table pointer never gets stuffed + * into the dispatch jump table otherwise. + * + * It matters only on SPARC, and only if you are using assembler + * code instead of C-code indirect dispatch. + * + * -- FEM, 04.xii.03 + */ +extern unsigned int _mesa_sparc_glapi_begin; +extern unsigned int _mesa_sparc_glapi_end; +extern void __glapi_sparc_icache_flush(unsigned int *); static void _glx_mesa_init_sparc_glapi_relocs(void); static int _mesa_sparc_needs_init = 1; #define INIT_MESA_SPARC { \ @@ -75,11 +107,173 @@ static int _mesa_sparc_needs_init = 1; #define INIT_MESA_SPARC #endif +#ifdef GLX_DIRECT_RENDERING +static __DRIscreen *__glXFindDRIScreen(__DRInativeDisplay *dpy, int scrn); +#endif /* GLX_DIRECT_RENDERING */ + +static Bool MakeContextCurrent(Display *dpy, GLXDrawable draw, + GLXDrawable read, GLXContext gc); + +/* +** We setup some dummy structures here so that the API can be used +** even if no context is current. +*/ + +static GLubyte dummyBuffer[__GLX_BUFFER_LIMIT_SIZE]; + +/* +** Dummy context used by small commands when there is no current context. +** All the +** gl and glx entry points are designed to operate as nop's when using +** the dummy context structure. +*/ +static __GLXcontext dummyContext = { + &dummyBuffer[0], + &dummyBuffer[0], + &dummyBuffer[0], + &dummyBuffer[__GLX_BUFFER_LIMIT_SIZE], + sizeof(dummyBuffer), +}; + + +/* +** All indirect rendering contexts will share the same indirect dispatch table. +*/ +static __GLapi *IndirectAPI = NULL; + + +/* + * Current context management and locking + */ + +#if defined( USE_XTHREADS ) + +/* thread safe */ +static GLboolean TSDinitialized = GL_FALSE; +static xthread_key_t ContextTSD; + +__GLXcontext *__glXGetCurrentContext(void) +{ + if (!TSDinitialized) { + xthread_key_create(&ContextTSD, NULL); + TSDinitialized = GL_TRUE; + return &dummyContext; + } + else { + void *p; + xthread_get_specific(ContextTSD, &p); + if (!p) + return &dummyContext; + else + return (__GLXcontext *) p; + } +} + +void __glXSetCurrentContext(__GLXcontext *c) +{ + if (!TSDinitialized) { + xthread_key_create(&ContextTSD, NULL); + TSDinitialized = GL_TRUE; + } + xthread_set_specific(ContextTSD, c); +} + + +/* Used by the __glXLock() and __glXUnlock() macros */ +xmutex_rec __glXmutex; + +#elif defined( PTHREADS ) + +pthread_mutex_t __glXmutex = PTHREAD_MUTEX_INITIALIZER; + +# if defined( GLX_USE_TLS ) + +/** + * Per-thread GLX context pointer. + * + * \c __glXSetCurrentContext is written is such a way that this pointer can + * \b never be \c NULL. This is important! Because of this + * \c __glXGetCurrentContext can be implemented as trivial macro. + */ +__thread void * __glX_tls_Context __attribute__((tls_model("initial-exec"))) + = &dummyContext; + +void __glXSetCurrentContext( __GLXcontext * c ) +{ + __glX_tls_Context = (c != NULL) ? c : &dummyContext; +} + +# else + +static pthread_once_t once_control = PTHREAD_ONCE_INIT; + +/** + * Per-thread data key. + * + * Once \c init_thread_data has been called, the per-thread data key will + * take a value of \c NULL. As each new thread is created the default + * value, in that thread, will be \c NULL. + */ +static pthread_key_t ContextTSD; + +/** + * Initialize the per-thread data key. + * + * This function is called \b exactly once per-process (not per-thread!) to + * initialize the per-thread data key. This is ideally done using the + * \c pthread_once mechanism. + */ +static void init_thread_data( void ) +{ + if ( pthread_key_create( & ContextTSD, NULL ) != 0 ) { + perror( "pthread_key_create" ); + exit( -1 ); + } +} + +void __glXSetCurrentContext( __GLXcontext * c ) +{ + pthread_once( & once_control, init_thread_data ); + pthread_setspecific( ContextTSD, c ); +} + +__GLXcontext * __glXGetCurrentContext( void ) +{ + void * v; + + pthread_once( & once_control, init_thread_data ); + + v = pthread_getspecific( ContextTSD ); + return (v == NULL) ? & dummyContext : (__GLXcontext *) v; +} + +# endif /* defined( GLX_USE_TLS ) */ + +#elif defined( THREADS ) + +#error Unknown threading method specified. + +#else + +/* not thread safe */ +__GLXcontext *__glXcurrentContext = &dummyContext; + +#endif + + /* ** You can set this cell to 1 to force the gl drawing stuff to be ** one command per packet */ -_X_HIDDEN int __glXDebug = 0; +int __glXDebug = 0; + +/* +** forward prototype declarations +*/ +int __glXCloseDisplay(Display *dpy, XExtCodes *codes); + + +/************************************************************************/ /* Extension required boiler plate */ @@ -102,13 +296,16 @@ static /* const */ char *error_list[] = { "GLXBadWindow", }; -static int __glXCloseDisplay(Display *dpy, XExtCodes *codes) +int __glXCloseDisplay(Display *dpy, XExtCodes *codes) { GLXContext gc; gc = __glXGetCurrentContext(); if (dpy == gc->currentDpy) { - __glXSetCurrentContextNull(); + __glXSetCurrentContext(&dummyContext); +#ifdef GLX_DIRECT_RENDERING + _glapi_set_dispatch(NULL); /* no-op functions */ +#endif __glXFreeContext(gc); } @@ -163,10 +360,11 @@ static void FreeScreenConfigs(__GLXdisplayPrivate *priv) Xfree((char*) psc->serverGLXexts); #ifdef GLX_DIRECT_RENDERING - if (psc->driScreen) { - psc->driScreen->destroyScreen(psc); - __glxHashDestroy(psc->drawHash); - } + /* Free the direct rendering per screen data */ + if (psc->driScreen.private) + (*psc->driScreen.destroyScreen)(priv->dpy, i, + psc->driScreen.private); + psc->driScreen.private = NULL; #endif } XFree((char*) priv->screenConfigs); @@ -193,12 +391,14 @@ static int __glXFreeDisplayPrivate(XExtData *extension) #ifdef GLX_DIRECT_RENDERING /* Free the direct rendering per display data */ - if (priv->driDisplay) - (*priv->driDisplay->destroyDisplay)(priv->driDisplay); - priv->driDisplay = NULL; - if (priv->dri2Display) - (*priv->dri2Display->destroyDisplay)(priv->dri2Display); - priv->dri2Display = NULL; + if (priv->driDisplay.private) + (*priv->driDisplay.destroyDisplay)(priv->dpy, + priv->driDisplay.private); + priv->driDisplay.private = NULL; + if (priv->driDisplay.createNewScreen) { + Xfree(priv->driDisplay.createNewScreen); /* free array of ptrs */ + priv->driDisplay.createNewScreen = NULL; + } #endif Xfree((char*) priv); @@ -240,7 +440,7 @@ static Bool QueryVersion(Display *dpy, int opcode, int *major, int *minor) } -_X_HIDDEN void +void __glXInitializeVisualConfigFromTags( __GLcontextModes *config, int count, const INT32 *bp, Bool tagged_only, Bool fbconfig_style_tags ) @@ -434,128 +634,371 @@ __glXInitializeVisualConfigFromTags( __GLcontextModes *config, int count, config->haveStencilBuffer = (config->stencilBits > 0); } -static __GLcontextModes * -createConfigsFromProperties(Display *dpy, int nvisuals, int nprops, - int screen, GLboolean tagged_only) + +#ifdef GLX_DIRECT_RENDERING +static unsigned +filter_modes( __GLcontextModes ** server_modes, + const __GLcontextModes * driver_modes ) { - INT32 buf[__GLX_TOTAL_CONFIG], *props; - unsigned prop_size; - __GLcontextModes *modes, *m; - int i; + __GLcontextModes * m; + __GLcontextModes ** prev_next; + const __GLcontextModes * check; + unsigned modes_count = 0; - if (nprops == 0) - return NULL; + if ( driver_modes == NULL ) { + fprintf(stderr, "libGL warning: 3D driver returned no fbconfigs.\n"); + return 0; + } - /* FIXME: Is the __GLX_MIN_CONFIG_PROPS test correct for FBconfigs? */ + /* For each mode in server_modes, check to see if a matching mode exists + * in driver_modes. If not, then the mode is not available. + */ - /* Check number of properties */ - if (nprops < __GLX_MIN_CONFIG_PROPS || nprops > __GLX_MAX_CONFIG_PROPS) - return NULL; + prev_next = server_modes; + for ( m = *prev_next ; m != NULL ; m = *prev_next ) { + GLboolean do_delete = GL_TRUE; - /* Allocate memory for our config structure */ - modes = _gl_context_modes_create(nvisuals, sizeof(__GLcontextModes)); - if (!modes) - return NULL; + for ( check = driver_modes ; check != NULL ; check = check->next ) { + if ( _gl_context_modes_are_same( m, check ) ) { + do_delete = GL_FALSE; + break; + } + } - prop_size = nprops * __GLX_SIZE_INT32; - if (prop_size <= sizeof(buf)) - props = buf; - else - props = Xmalloc(prop_size); + /* The 3D has to support all the modes that match the GLX visuals + * sent from the X server. + */ + if ( do_delete && (m->visualID != 0) ) { + do_delete = GL_FALSE; - /* Read each config structure and convert it into our format */ - m = modes; - for (i = 0; i < nvisuals; i++) { - _XRead(dpy, (char *)props, prop_size); - /* Older X servers don't send this so we default it here. */ - m->drawableType = GLX_WINDOW_BIT; - __glXInitializeVisualConfigFromTags(m, nprops, props, - tagged_only, GL_TRUE); - m->screen = screen; - m = m->next; - } + if (getenv("LIBGL_DEBUG")) { + fprintf(stderr, "libGL warning: 3D driver claims to not support " + "visual 0x%02x\n", m->visualID); + } + } - if (props != buf) - Xfree(props); + if ( do_delete ) { + *prev_next = m->next; - return modes; + m->next = NULL; + _gl_context_modes_destroy( m ); + } + else { + modes_count++; + prev_next = & m->next; + } + } + + return modes_count; } -static GLboolean -getVisualConfigs(Display *dpy, __GLXdisplayPrivate *priv, int screen) + +/** + * Implement \c __DRIinterfaceMethods::getProcAddress. + */ +static __DRIfuncPtr get_proc_address( const char * proc_name ) { - xGLXGetVisualConfigsReq *req; - __GLXscreenConfigs *psc; - xGLXGetVisualConfigsReply reply; + if (strcmp( proc_name, "glxEnableExtension" ) == 0) { + return (__DRIfuncPtr) __glXScrEnableExtension; + } - LockDisplay(dpy); + return NULL; +} + +#ifdef XDAMAGE_1_1_INTERFACE +static GLboolean has_damage_post(__DRInativeDisplay *dpy) +{ + static GLboolean inited = GL_FALSE; + static GLboolean has_damage; + + if (!inited) { + int major, minor; - psc = priv->screenConfigs + screen; - psc->visuals = NULL; - GetReq(GLXGetVisualConfigs, req); - req->reqType = priv->majorOpcode; - req->glxCode = X_GLXGetVisualConfigs; - req->screen = screen; + if (XDamageQueryVersion(dpy, &major, &minor) && + major == 1 && minor >= 1) + { + has_damage = GL_TRUE; + } else { + has_damage = GL_FALSE; + } + inited = GL_TRUE; + } - if (!_XReply(dpy, (xReply*) &reply, 0, False)) - goto out; + return has_damage; +} +#endif /* XDAMAGE_1_1_INTERFACE */ - psc->visuals = createConfigsFromProperties(dpy, - reply.numVisuals, - reply.numProps, - screen, GL_FALSE); +static void __glXReportDamage(__DRInativeDisplay *dpy, int screen, + __DRIid drawable, + int x, int y, + drm_clip_rect_t *rects, int num_rects, + GLboolean front_buffer) +{ +#ifdef XDAMAGE_1_1_INTERFACE + XRectangle *xrects; + XserverRegion region; + int i; + int x_off, y_off; + + if (!has_damage_post(dpy)) + return; + + if (front_buffer) { + x_off = x; + y_off = y; + drawable = RootWindow(dpy, screen); + } else{ + x_off = 0; + y_off = 0; + } - out: - UnlockDisplay(dpy); - return psc->visuals != NULL; + xrects = malloc(sizeof(XRectangle) * num_rects); + if (xrects == NULL) + return; + + for (i = 0; i < num_rects; i++) { + xrects[i].x = rects[i].x1 + x_off; + xrects[i].y = rects[i].y1 + y_off; + xrects[i].width = rects[i].x2 - rects[i].x1; + xrects[i].height = rects[i].y2 - rects[i].y1; + } + region = XFixesCreateRegion(dpy, xrects, num_rects); + free(xrects); + XDamageAdd(dpy, drawable, region); + XFixesDestroyRegion(dpy, region); +#endif } -static GLboolean -getFBConfigs(Display *dpy, __GLXdisplayPrivate *priv, int screen) +/** + * Table of functions exported by the loader to the driver. + */ +static const __DRIinterfaceMethods interface_methods = { + get_proc_address, + + _gl_context_modes_create, + _gl_context_modes_destroy, + + __glXFindDRIScreen, + __glXWindowExists, + + XF86DRICreateContextWithConfig, + XF86DRIDestroyContext, + + XF86DRICreateDrawable, + XF86DRIDestroyDrawable, + XF86DRIGetDrawableInfo, + + __glXGetUST, + __glXGetMscRateOML, + + __glXReportDamage, +}; + + + +/** + * Perform the required libGL-side initialization and call the client-side + * driver's \c __driCreateNewScreen function. + * + * \param dpy Display pointer. + * \param scrn Screen number on the display. + * \param psc DRI screen information. + * \param driDpy DRI display information. + * \param createNewScreen Pointer to the client-side driver's + * \c __driCreateNewScreen function. + * \returns A pointer to the \c __DRIscreenPrivate structure returned by + * the client-side driver on success, or \c NULL on failure. + * + * \todo This function needs to be modified to remove context-modes from the + * list stored in the \c __GLXscreenConfigsRec to match the list + * returned by the client-side driver. + */ +static void * +CallCreateNewScreen(Display *dpy, int scrn, __DRIscreen *psc, + __DRIdisplay * driDpy, + PFNCREATENEWSCREENFUNC createNewScreen) { - xGLXGetFBConfigsReq *fb_req; - xGLXGetFBConfigsSGIXReq *sgi_req; - xGLXVendorPrivateWithReplyReq *vpreq; - xGLXGetFBConfigsReply reply; - __GLXscreenConfigs *psc; + __DRIscreenPrivate *psp = NULL; +#ifndef GLX_USE_APPLEGL + drm_handle_t hSAREA; + drmAddress pSAREA = MAP_FAILED; + char *BusID; + __DRIversion ddx_version; + __DRIversion dri_version; + __DRIversion drm_version; + __DRIframebuffer framebuffer; + int fd = -1; + int status; + const char * err_msg; + const char * err_extra; + int api_ver = __glXGetInternalVersion(); + + + dri_version.major = driDpy->private->driMajor; + dri_version.minor = driDpy->private->driMinor; + dri_version.patch = driDpy->private->driPatch; + + + err_msg = "XF86DRIOpenConnection"; + err_extra = NULL; + + framebuffer.base = MAP_FAILED; + framebuffer.dev_priv = NULL; + + if (XF86DRIOpenConnection(dpy, scrn, &hSAREA, &BusID)) { + int newlyopened; + fd = drmOpenOnce(NULL,BusID, &newlyopened); + Xfree(BusID); /* No longer needed */ + + err_msg = "open DRM"; + err_extra = strerror( -fd ); + + if (fd >= 0) { + drm_magic_t magic; + + err_msg = "drmGetMagic"; + err_extra = NULL; + + if (!drmGetMagic(fd, &magic)) { + drmVersionPtr version = drmGetVersion(fd); + if (version) { + drm_version.major = version->version_major; + drm_version.minor = version->version_minor; + drm_version.patch = version->version_patchlevel; + drmFreeVersion(version); + } + else { + drm_version.major = -1; + drm_version.minor = -1; + drm_version.patch = -1; + } + + err_msg = "XF86DRIAuthConnection"; + if (!newlyopened || XF86DRIAuthConnection(dpy, scrn, magic)) { + char *driverName; + + /* + * Get device name (like "tdfx") and the ddx version + * numbers. We'll check the version in each DRI driver's + * "createNewScreen" function. + */ + err_msg = "XF86DRIGetClientDriverName"; + if (XF86DRIGetClientDriverName(dpy, scrn, + &ddx_version.major, + &ddx_version.minor, + &ddx_version.patch, + &driverName)) { + drm_handle_t hFB; + int junk; + + /* No longer needed. */ + Xfree( driverName ); + + + /* + * Get device-specific info. pDevPriv will point to a struct + * (such as DRIRADEONRec in xfree86/driver/ati/radeon_dri.h) + * that has information about the screen size, depth, pitch, + * ancilliary buffers, DRM mmap handles, etc. + */ + err_msg = "XF86DRIGetDeviceInfo"; + if (XF86DRIGetDeviceInfo(dpy, scrn, + &hFB, + &junk, + &framebuffer.size, + &framebuffer.stride, + &framebuffer.dev_priv_size, + &framebuffer.dev_priv)) { + framebuffer.width = DisplayWidth(dpy, scrn); + framebuffer.height = DisplayHeight(dpy, scrn); + + /* + * Map the framebuffer region. + */ + status = drmMap(fd, hFB, framebuffer.size, + (drmAddressPtr)&framebuffer.base); + + err_msg = "drmMap of framebuffer"; + err_extra = strerror( -status ); + + if ( status == 0 ) { + /* + * Map the SAREA region. Further mmap regions + * may be setup in each DRI driver's + * "createNewScreen" function. + */ + status = drmMap(fd, hSAREA, SAREA_MAX, + &pSAREA); + + err_msg = "drmMap of sarea"; + err_extra = strerror( -status ); + + if ( status == 0 ) { + __GLcontextModes * driver_modes = NULL; + __GLXscreenConfigs *configs = psc->screenConfigs; + + err_msg = "InitDriver"; + err_extra = NULL; + psp = (*createNewScreen)(dpy, scrn, + psc, + configs->configs, + & ddx_version, + & dri_version, + & drm_version, + & framebuffer, + pSAREA, + fd, + api_ver, + & interface_methods, + & driver_modes ); + + filter_modes( & configs->configs, + driver_modes ); + _gl_context_modes_destroy( driver_modes ); + } + } + } + } + } + } + } + } - psc = priv->screenConfigs + screen; - psc->serverGLXexts = __glXGetStringFromServer(dpy, priv->majorOpcode, - X_GLXQueryServerString, - screen, GLX_EXTENSIONS); + if ( psp == NULL ) { + if ( pSAREA != MAP_FAILED ) { + (void)drmUnmap(pSAREA, SAREA_MAX); + } - LockDisplay(dpy); + if ( framebuffer.base != MAP_FAILED ) { + (void)drmUnmap((drmAddress)framebuffer.base, framebuffer.size); + } - psc->configs = NULL; - if (atof(priv->serverGLXversion) >= 1.3) { - GetReq(GLXGetFBConfigs, fb_req); - fb_req->reqType = priv->majorOpcode; - fb_req->glxCode = X_GLXGetFBConfigs; - fb_req->screen = screen; - } else if (strstr(psc->serverGLXexts, "GLX_SGIX_fbconfig") != NULL) { - GetReqExtra(GLXVendorPrivateWithReply, - sz_xGLXGetFBConfigsSGIXReq + - sz_xGLXVendorPrivateWithReplyReq, vpreq); - sgi_req = (xGLXGetFBConfigsSGIXReq *) vpreq; - sgi_req->reqType = priv->majorOpcode; - sgi_req->glxCode = X_GLXVendorPrivateWithReply; - sgi_req->vendorCode = X_GLXvop_GetFBConfigsSGIX; - sgi_req->screen = screen; - } else - goto out; + if ( framebuffer.dev_priv != NULL ) { + Xfree(framebuffer.dev_priv); + } - if (!_XReply(dpy, (xReply*) &reply, 0, False)) - goto out; + if ( fd >= 0 ) { + (void)drmCloseOnce(fd); + } - psc->configs = createConfigsFromProperties(dpy, - reply.numFBConfigs, - reply.numAttribs * 2, - screen, GL_TRUE); + (void)XF86DRICloseConnection(dpy, scrn); - out: - UnlockDisplay(dpy); - return psc->configs != NULL; + if ( err_extra != NULL ) { + fprintf(stderr, "libGL error: %s failed (%s)\n", err_msg, + err_extra); + } + else { + fprintf(stderr, "libGL error: %s failed\n", err_msg ); + } + + fprintf(stderr, "libGL error: reverting to (slow) indirect rendering\n"); + } +#endif /* !GLX_USE_APPLEGL */ + + return psp; } +#endif /* GLX_DIRECT_RENDERING */ + /* ** Allocate the memory for the per screen configs for each screen. @@ -563,8 +1006,17 @@ getFBConfigs(Display *dpy, __GLXdisplayPrivate *priv, int screen) */ static Bool AllocAndFetchScreenConfigs(Display *dpy, __GLXdisplayPrivate *priv) { + xGLXGetVisualConfigsReq *req; + xGLXGetFBConfigsReq *fb_req; + xGLXVendorPrivateWithReplyReq *vpreq; + xGLXGetFBConfigsSGIXReq *sgi_req; + xGLXGetVisualConfigsReply reply; __GLXscreenConfigs *psc; - GLint i, screens; + __GLcontextModes *config; + GLint i, j, nprops, screens; + INT32 buf[__GLX_TOTAL_CONFIG], *props; + unsigned supported_request = 0; + unsigned prop_size; /* ** First allocate memory for the array of per screen configs. @@ -578,30 +1030,159 @@ static Bool AllocAndFetchScreenConfigs(Display *dpy, __GLXdisplayPrivate *priv) priv->screenConfigs = psc; priv->serverGLXversion = __glXGetStringFromServer(dpy, priv->majorOpcode, - X_GLXQueryServerString, - 0, GLX_VERSION); + X_GLXQueryServerString, + 0, GLX_VERSION); if ( priv->serverGLXversion == NULL ) { FreeScreenConfigs(priv); return GL_FALSE; } + if ( atof( priv->serverGLXversion ) >= 1.3 ) { + supported_request = 1; + } + + /* + ** Now fetch each screens configs structures. If a screen supports + ** GL (by returning a numVisuals > 0) then allocate memory for our + ** config structure and then fill it in. + */ for (i = 0; i < screens; i++, psc++) { - getVisualConfigs(dpy, priv, i); - getFBConfigs(dpy, priv, i); + if ( supported_request != 1 ) { + psc->serverGLXexts = __glXGetStringFromServer(dpy, priv->majorOpcode, + X_GLXQueryServerString, + i, GLX_EXTENSIONS); + if ( strstr( psc->serverGLXexts, "GLX_SGIX_fbconfig" ) != NULL ) { + supported_request = 2; + } + else { + supported_request = 3; + } + } - psc->scr = i; - psc->dpy = dpy; -#ifdef GLX_DIRECT_RENDERING - psc->drawHash = __glxHashCreate(); - if (psc->drawHash == NULL) + + LockDisplay(dpy); + switch( supported_request ) { + case 1: + GetReq(GLXGetFBConfigs,fb_req); + fb_req->reqType = priv->majorOpcode; + fb_req->glxCode = X_GLXGetFBConfigs; + fb_req->screen = i; + break; + + case 2: + GetReqExtra(GLXVendorPrivateWithReply, + sz_xGLXGetFBConfigsSGIXReq-sz_xGLXVendorPrivateWithReplyReq,vpreq); + sgi_req = (xGLXGetFBConfigsSGIXReq *) vpreq; + sgi_req->reqType = priv->majorOpcode; + sgi_req->glxCode = X_GLXVendorPrivateWithReply; + sgi_req->vendorCode = X_GLXvop_GetFBConfigsSGIX; + sgi_req->screen = i; + break; + + case 3: + GetReq(GLXGetVisualConfigs,req); + req->reqType = priv->majorOpcode; + req->glxCode = X_GLXGetVisualConfigs; + req->screen = i; + break; + } + + if (!_XReply(dpy, (xReply*) &reply, 0, False)) { + /* Something is busted. Punt. */ + UnlockDisplay(dpy); + SyncHandle(); + FreeScreenConfigs(priv); + return GL_FALSE; + } + + if (!reply.numVisuals) { + /* This screen does not support GL rendering */ + UnlockDisplay(dpy); continue; - if (priv->dri2Display) - psc->driScreen = (*priv->dri2Display->createScreen)(psc, i, priv); - if (psc->driScreen == NULL && priv->driDisplay) - psc->driScreen = (*priv->driDisplay->createScreen)(psc, i, priv); - if (psc->driScreen == NULL) { - __glxHashDestroy(psc->drawHash); - psc->drawHash = NULL; + } + + /* FIXME: Is the __GLX_MIN_CONFIG_PROPS test correct for + * FIXME: FBconfigs? + */ + /* Check number of properties */ + nprops = reply.numProps; + if ((nprops < __GLX_MIN_CONFIG_PROPS) || + (nprops > __GLX_MAX_CONFIG_PROPS)) { + /* Huh? Not in protocol defined limits. Punt */ + UnlockDisplay(dpy); + SyncHandle(); + FreeScreenConfigs(priv); + return GL_FALSE; + } + + /* Allocate memory for our config structure */ + psc->configs = _gl_context_modes_create(reply.numVisuals, + sizeof(__GLcontextModes)); + if (!psc->configs) { + UnlockDisplay(dpy); + SyncHandle(); + FreeScreenConfigs(priv); + return GL_FALSE; + } + + /* Allocate memory for the properties, if needed */ + if ( supported_request != 3 ) { + nprops *= 2; + } + + prop_size = nprops * __GLX_SIZE_INT32; + + if (prop_size <= sizeof(buf)) { + props = buf; + } else { + props = (INT32 *) Xmalloc(prop_size); + } + + /* Read each config structure and convert it into our format */ + config = psc->configs; + for (j = 0; j < reply.numVisuals; j++) { + assert( config != NULL ); + _XRead(dpy, (char *)props, prop_size); + + if ( supported_request != 3 ) { + config->rgbMode = GL_TRUE; + config->drawableType = GLX_WINDOW_BIT; + } + else { + config->drawableType = GLX_WINDOW_BIT | GLX_PIXMAP_BIT; + } + + __glXInitializeVisualConfigFromTags( config, nprops, props, + (supported_request != 3), + GL_TRUE ); + if ( config->fbconfigID == GLX_DONT_CARE ) { + config->fbconfigID = config->visualID; + } + config->screen = i; + config = config->next; + } + if (props != buf) { + Xfree((char *)props); + } + UnlockDisplay(dpy); + +#ifdef GLX_DIRECT_RENDERING + /* Initialize per screen dynamic client GLX extensions */ + psc->ext_list_first_time = GL_TRUE; + /* Initialize the direct rendering per screen data and functions */ + if (priv->driDisplay.private != NULL) { + /* FIXME: Should it be some sort of an error if createNewScreen[i] + * FIXME: is NULL? + */ + if (priv->driDisplay.createNewScreen && + priv->driDisplay.createNewScreen[i]) { + + psc->driScreen.screenConfigs = (void *)psc; + psc->driScreen.private = + CallCreateNewScreen(dpy, i, & psc->driScreen, + & priv->driDisplay, + priv->driDisplay.createNewScreen[i] ); + } } #endif } @@ -612,7 +1193,7 @@ static Bool AllocAndFetchScreenConfigs(Display *dpy, __GLXdisplayPrivate *priv) /* ** Initialize the client side extension code. */ -_X_HIDDEN __GLXdisplayPrivate *__glXInitialize(Display* dpy) +__GLXdisplayPrivate *__glXInitialize(Display* dpy) { XExtDisplayInfo *info = __glXFindDisplay(dpy); XExtData **privList, *private, *found; @@ -692,8 +1273,8 @@ _X_HIDDEN __GLXdisplayPrivate *__glXInitialize(Display* dpy) ** (e.g., those called in AllocAndFetchScreenConfigs). */ if (getenv("LIBGL_ALWAYS_INDIRECT") == NULL) { - dpyPriv->dri2Display = dri2CreateDisplay(dpy); - dpyPriv->driDisplay = driCreateDisplay(dpy); + dpyPriv->driDisplay.private = + driCreateDisplay(dpy, &dpyPriv->driDisplay); } #endif @@ -727,7 +1308,7 @@ _X_HIDDEN __GLXdisplayPrivate *__glXInitialize(Display* dpy) ** Setup for sending a GLX command on dpy. Make sure the extension is ** initialized. Try to avoid calling __glXInitialize as its kinda slow. */ -_X_HIDDEN CARD8 __glXSetupForCommand(Display *dpy) +CARD8 __glXSetupForCommand(Display *dpy) { GLXContext gc; __GLXdisplayPrivate *priv; @@ -768,7 +1349,7 @@ _X_HIDDEN CARD8 __glXSetupForCommand(Display *dpy) * Modify this function to use \c ctx->pc instead of the explicit * \c pc parameter. */ -_X_HIDDEN GLubyte *__glXFlushRenderBuffer(__GLXcontext *ctx, GLubyte *pc) +GLubyte *__glXFlushRenderBuffer(__GLXcontext *ctx, GLubyte *pc) { Display * const dpy = ctx->currentDpy; #ifdef USE_XCB @@ -780,8 +1361,7 @@ _X_HIDDEN GLubyte *__glXFlushRenderBuffer(__GLXcontext *ctx, GLubyte *pc) if ( (dpy != NULL) && (size > 0) ) { #ifdef USE_XCB - xcb_glx_render(c, ctx->currentContextTag, size, - (const uint8_t *)ctx->buf); + xcb_glx_render(c, ctx->currentContextTag, size, (char *)ctx->buf); #else /* Send the entire buffer as an X request */ LockDisplay(dpy); @@ -818,9 +1398,9 @@ _X_HIDDEN GLubyte *__glXFlushRenderBuffer(__GLXcontext *ctx, GLubyte *pc) * \param data Command data. * \param dataLen Size, in bytes, of the command data. */ -_X_HIDDEN void __glXSendLargeChunk(__GLXcontext *gc, GLint requestNumber, - GLint totalRequests, - const GLvoid * data, GLint dataLen) +void __glXSendLargeChunk(__GLXcontext *gc, GLint requestNumber, + GLint totalRequests, + const GLvoid * data, GLint dataLen) { Display *dpy = gc->currentDpy; #ifdef USE_XCB @@ -866,9 +1446,9 @@ _X_HIDDEN void __glXSendLargeChunk(__GLXcontext *gc, GLint requestNumber, * \param data Command data. * \param dataLen Size, in bytes, of the command data. */ -_X_HIDDEN void __glXSendLargeCommand(__GLXcontext *ctx, - const GLvoid *header, GLint headerLen, - const GLvoid *data, GLint dataLen) +void __glXSendLargeCommand(__GLXcontext *ctx, + const GLvoid *header, GLint headerLen, + const GLvoid *data, GLint dataLen) { GLint maxSize; GLint totalRequests, requestNumber; @@ -904,8 +1484,330 @@ _X_HIDDEN void __glXSendLargeCommand(__GLXcontext *ctx, /************************************************************************/ +PUBLIC GLXContext glXGetCurrentContext(void) +{ + GLXContext cx = __glXGetCurrentContext(); + + if (cx == &dummyContext) { + return NULL; + } else { + return cx; + } +} + +PUBLIC GLXDrawable glXGetCurrentDrawable(void) +{ + GLXContext gc = __glXGetCurrentContext(); + return gc->currentDrawable; +} + + +/************************************************************************/ + +#ifdef GLX_DIRECT_RENDERING +/* Return the DRI per screen structure */ +__DRIscreen *__glXFindDRIScreen(__DRInativeDisplay *dpy, int scrn) +{ + __DRIscreen *pDRIScreen = NULL; + XExtDisplayInfo *info = __glXFindDisplay(dpy); + XExtData **privList, *found; + __GLXdisplayPrivate *dpyPriv; + XEDataObject dataObj; + + __glXLock(); + dataObj.display = dpy; + privList = XEHeadOfExtensionList(dataObj); + found = XFindOnExtensionList(privList, info->codes->extension); + __glXUnlock(); + + if (found) { + dpyPriv = (__GLXdisplayPrivate *)found->private_data; + pDRIScreen = &dpyPriv->screenConfigs[scrn].driScreen; + } + + return pDRIScreen; +} +#endif + +/************************************************************************/ + +static Bool SendMakeCurrentRequest( Display *dpy, CARD8 opcode, + GLXContextID gc, GLXContextTag old_gc, GLXDrawable draw, GLXDrawable read, + xGLXMakeCurrentReply * reply ); + +/** + * Sends a GLX protocol message to the specified display to make the context + * and the drawables current. + * + * \param dpy Display to send the message to. + * \param opcode Major opcode value for the display. + * \param gc_id Context tag for the context to be made current. + * \param draw Drawable ID for the "draw" drawable. + * \param read Drawable ID for the "read" drawable. + * \param reply Space to store the X-server's reply. + * + * \warning + * This function assumes that \c dpy is locked with \c LockDisplay on entry. + */ +static Bool SendMakeCurrentRequest(Display *dpy, CARD8 opcode, + GLXContextID gc_id, GLXContextTag gc_tag, + GLXDrawable draw, GLXDrawable read, + xGLXMakeCurrentReply *reply) +{ + Bool ret; + + + LockDisplay(dpy); + + if (draw == read) { + xGLXMakeCurrentReq *req; + + GetReq(GLXMakeCurrent,req); + req->reqType = opcode; + req->glxCode = X_GLXMakeCurrent; + req->drawable = draw; + req->context = gc_id; + req->oldContextTag = gc_tag; + } + else { + __GLXdisplayPrivate *priv = __glXInitialize(dpy); + + /* If the server can support the GLX 1.3 version, we should + * perfer that. Not only that, some servers support GLX 1.3 but + * not the SGI extension. + */ + + if ((priv->majorVersion > 1) || (priv->minorVersion >= 3)) { + xGLXMakeContextCurrentReq *req; + + GetReq(GLXMakeContextCurrent,req); + req->reqType = opcode; + req->glxCode = X_GLXMakeContextCurrent; + req->drawable = draw; + req->readdrawable = read; + req->context = gc_id; + req->oldContextTag = gc_tag; + } + else { + xGLXVendorPrivateWithReplyReq *vpreq; + xGLXMakeCurrentReadSGIReq *req; + + GetReqExtra(GLXVendorPrivateWithReply, + sz_xGLXMakeCurrentReadSGIReq-sz_xGLXVendorPrivateWithReplyReq,vpreq); + req = (xGLXMakeCurrentReadSGIReq *)vpreq; + req->reqType = opcode; + req->glxCode = X_GLXVendorPrivateWithReply; + req->vendorCode = X_GLXvop_MakeCurrentReadSGI; + req->drawable = draw; + req->readable = read; + req->context = gc_id; + req->oldContextTag = gc_tag; + } + } + + ret = _XReply(dpy, (xReply*) reply, 0, False); + + UnlockDisplay(dpy); + SyncHandle(); + + return ret; +} + + +#ifdef GLX_DIRECT_RENDERING +static Bool BindContextWrapper( Display *dpy, GLXContext gc, + GLXDrawable draw, GLXDrawable read ) +{ + return (*gc->driContext.bindContext)(dpy, gc->screen, draw, read, + & gc->driContext); +} + + +static Bool UnbindContextWrapper( GLXContext gc ) +{ + return (*gc->driContext.unbindContext)(gc->currentDpy, gc->screen, + gc->currentDrawable, + gc->currentReadable, + & gc->driContext ); +} +#endif /* GLX_DIRECT_RENDERING */ + + +/** + * Make a particular context current. + * + * \note This is in this file so that it can access dummyContext. + */ +USED static Bool MakeContextCurrent(Display *dpy, GLXDrawable draw, + GLXDrawable read, GLXContext gc) +{ + xGLXMakeCurrentReply reply; + const GLXContext oldGC = __glXGetCurrentContext(); + const CARD8 opcode = __glXSetupForCommand(dpy); + const CARD8 oldOpcode = ((gc == oldGC) || (oldGC == &dummyContext)) + ? opcode : __glXSetupForCommand(oldGC->currentDpy); + Bool bindReturnValue; + + + if (!opcode || !oldOpcode) { + return GL_FALSE; + } + + /* Make sure that the new context has a nonzero ID. In the request, + * a zero context ID is used only to mean that we bind to no current + * context. + */ + if ((gc != NULL) && (gc->xid == None)) { + return GL_FALSE; + } + +#ifndef GLX_DIRECT_RENDERING + if (gc && gc->isDirect) { + return GL_FALSE; + } +#endif + + _glapi_check_multithread(); + +#ifdef GLX_DIRECT_RENDERING + /* Bind the direct rendering context to the drawable */ + if (gc && gc->isDirect) { + bindReturnValue = (gc->driContext.private) + ? BindContextWrapper(dpy, gc, draw, read) + : False; + } else +#endif + { + /* Send a glXMakeCurrent request to bind the new context. */ + bindReturnValue = + SendMakeCurrentRequest(dpy, opcode, gc ? gc->xid : None, + ((dpy != oldGC->currentDpy) || oldGC->isDirect) + ? None : oldGC->currentContextTag, + draw, read, &reply); + } + + + if (!bindReturnValue) { + return False; + } + + if ((dpy != oldGC->currentDpy || (gc && gc->isDirect)) && + !oldGC->isDirect && oldGC != &dummyContext) { + xGLXMakeCurrentReply dummy_reply; + + /* We are either switching from one dpy to another and have to + * send a request to the previous dpy to unbind the previous + * context, or we are switching away from a indirect context to + * a direct context and have to send a request to the dpy to + * unbind the previous context. + */ + (void) SendMakeCurrentRequest(oldGC->currentDpy, oldOpcode, None, + oldGC->currentContextTag, None, None, + & dummy_reply); + } +#ifdef GLX_DIRECT_RENDERING + else if (oldGC->isDirect && oldGC->driContext.private) { + (void) UnbindContextWrapper(oldGC); + } +#endif + + + /* Update our notion of what is current */ + __glXLock(); + if (gc == oldGC) { + /* Even though the contexts are the same the drawable might have + * changed. Note that gc cannot be the dummy, and that oldGC + * cannot be NULL, therefore if they are the same, gc is not + * NULL and not the dummy. + */ + gc->currentDrawable = draw; + gc->currentReadable = read; + } else { + if (oldGC != &dummyContext) { + /* Old current context is no longer current to anybody */ + oldGC->currentDpy = 0; + oldGC->currentDrawable = None; + oldGC->currentReadable = None; + oldGC->currentContextTag = 0; + + if (oldGC->xid == None) { + /* We are switching away from a context that was + * previously destroyed, so we need to free the memory + * for the old handle. + */ +#ifdef GLX_DIRECT_RENDERING + /* Destroy the old direct rendering context */ + if (oldGC->isDirect) { + if (oldGC->driContext.private) { + (*oldGC->driContext.destroyContext) + (dpy, oldGC->screen, oldGC->driContext.private); + oldGC->driContext.private = NULL; + } + } +#endif + __glXFreeContext(oldGC); + } + } + if (gc) { + __glXSetCurrentContext(gc); + + gc->currentDpy = dpy; + gc->currentDrawable = draw; + gc->currentReadable = read; + + if (!gc->isDirect) { + if (!IndirectAPI) + IndirectAPI = __glXNewIndirectAPI(); + _glapi_set_dispatch(IndirectAPI); + +#ifdef GLX_USE_APPLEGL + do { + extern void XAppleDRIUseIndirectDispatch(void); + XAppleDRIUseIndirectDispatch(); + } while (0); +#endif + + __GLXattribute *state = + (__GLXattribute *)(gc->client_state_private); + + gc->currentContextTag = reply.contextTag; + if (state->array_state == NULL) { + (void) glGetString(GL_EXTENSIONS); + (void) glGetString(GL_VERSION); + __glXInitVertexArrayState(gc); + } + } + else { + gc->currentContextTag = -1; + } + } else { + __glXSetCurrentContext(&dummyContext); +#ifdef GLX_DIRECT_RENDERING + _glapi_set_dispatch(NULL); /* no-op functions */ +#endif + } + } + __glXUnlock(); + return GL_TRUE; +} + + +PUBLIC Bool glXMakeCurrent(Display *dpy, GLXDrawable draw, GLXContext gc) +{ + return MakeContextCurrent( dpy, draw, draw, gc ); +} + +PUBLIC GLX_ALIAS(Bool, glXMakeCurrentReadSGI, + (Display *dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx), + (dpy, d, r, ctx), MakeContextCurrent) + +PUBLIC GLX_ALIAS(Bool, glXMakeContextCurrent, + (Display *dpy, GLXDrawable d, GLXDrawable r, GLXContext ctx), + (dpy, d, r, ctx), MakeContextCurrent) + + #ifdef DEBUG -_X_HIDDEN void __glXDumpDrawBuffer(__GLXcontext *ctx) +void __glXDumpDrawBuffer(__GLXcontext *ctx) { GLubyte *p = ctx->buf; GLubyte *end = ctx->pc; @@ -930,23 +1832,9 @@ _X_HIDDEN void __glXDumpDrawBuffer(__GLXcontext *ctx) #ifdef USE_SPARC_ASM /* - * This is where our dispatch table's bounds are. - * And the static mesa_init is taken directly from - * Mesa's 'sparc.c' initializer. + * Used only when we are sparc, using sparc assembler. * - * We need something like this here, because this version - * of openGL/glx never initializes a Mesa context, and so - * the address of the dispatch table pointer never gets stuffed - * into the dispatch jump table otherwise. - * - * It matters only on SPARC, and only if you are using assembler - * code instead of C-code indirect dispatch. - * - * -- FEM, 04.xii.03 */ -extern unsigned int _mesa_sparc_glapi_begin; -extern unsigned int _mesa_sparc_glapi_end; -extern void __glapi_sparc_icache_flush(unsigned int *); static void _glx_mesa_init_sparc_glapi_relocs(void) diff --git a/src/glx/x11/glxextensions.c b/src/glx/x11/glxextensions.c index e843718472..1d99b61db0 100644 --- a/src/glx/x11/glxextensions.c +++ b/src/glx/x11/glxextensions.c @@ -356,16 +356,29 @@ __glXProcessServerString( const struct extension_info * ext, } } + +/** + * Enable a named GLX extension on a given screen. + * Drivers should not call this function directly. They should instead use + * \c glXGetProcAddress to obtain a pointer to the function. + * + * \param psc Pointer to GLX per-screen record. + * \param name Name of the extension to enable. + * + * \sa glXGetProcAddress + * + * \since Internal API version 20030813. + */ void -__glXEnableDirectExtension(__GLXscreenConfigs *psc, const char *name) +__glXScrEnableExtension( __GLXscreenConfigs *psc, const char * name ) { - __glXExtensionsCtr(); - __glXExtensionsCtrScreen(psc); - - set_glx_extension(known_glx_extensions, - name, strlen(name), GL_TRUE, psc->direct_support); + __glXExtensionsCtr(); + __glXExtensionsCtrScreen(psc); + set_glx_extension( known_glx_extensions, name, strlen( name ), GL_TRUE, + psc->direct_support ); } + /** * Initialize global extension support tables. */ diff --git a/src/glx/x11/glxextensions.h b/src/glx/x11/glxextensions.h index 9cdd05ed76..a4241b6b7f 100644 --- a/src/glx/x11/glxextensions.h +++ b/src/glx/x11/glxextensions.h @@ -234,7 +234,7 @@ extern GLboolean __glXExtensionBitIsEnabled( struct __GLXscreenConfigsRec *psc, extern const char * __glXGetClientExtensions( void ); extern void __glXCalculateUsableExtensions( struct __GLXscreenConfigsRec *psc, GLboolean display_is_direct_capable, int server_minor_version ); - +extern void __glXScrEnableExtension( struct __GLXscreenConfigsRec *psc, const char * name ); extern void __glXCalculateUsableGLExtensions( struct __GLXcontextRec * gc, const char * server_string, int major_version, int minor_version ); extern void __glXGetGLVersion( int * major_version, int * minor_version ); @@ -243,8 +243,6 @@ extern char * __glXGetClientGLExtensionString( void ); extern GLboolean __glExtensionBitIsEnabled( const struct __GLXcontextRec * gc, unsigned bit ); -extern void -__glXEnableDirectExtension(__GLXscreenConfigs *psc, const char *name); /* Source-level backwards compatibility with old drivers. They won't * find the respective functions, though. diff --git a/src/glx/x11/glxhash.c b/src/glx/x11/glxhash.c deleted file mode 100644 index 9ed0429032..0000000000 --- a/src/glx/x11/glxhash.c +++ /dev/null @@ -1,416 +0,0 @@ -/* glxhash.c -- Small hash table support for integer -> integer mapping - * Taken from libdrm. - * - * Created: Sun Apr 18 09:35:45 1999 by faith@precisioninsight.com - * - * Copyright 1999 Precision Insight, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - * - * Authors: Rickard E. (Rik) Faith - * - * DESCRIPTION - * - * This file contains a straightforward implementation of a fixed-sized - * hash table using self-organizing linked lists [Knuth73, pp. 398-399] for - * collision resolution. There are two potentially interesting things - * about this implementation: - * - * 1) The table is power-of-two sized. Prime sized tables are more - * traditional, but do not have a significant advantage over power-of-two - * sized table, especially when double hashing is not used for collision - * resolution. - * - * 2) The hash computation uses a table of random integers [Hanson97, - * pp. 39-41]. - * - * FUTURE ENHANCEMENTS - * - * With a table size of 512, the current implementation is sufficient for a - * few hundred keys. Since this is well above the expected size of the - * tables for which this implementation was designed, the implementation of - * dynamic hash tables was postponed until the need arises. A common (and - * naive) approach to dynamic hash table implementation simply creates a - * new hash table when necessary, rehashes all the data into the new table, - * and destroys the old table. The approach in [Larson88] is superior in - * two ways: 1) only a portion of the table is expanded when needed, - * distributing the expansion cost over several insertions, and 2) portions - * of the table can be locked, enabling a scalable thread-safe - * implementation. - * - * REFERENCES - * - * [Hanson97] David R. Hanson. C Interfaces and Implementations: - * Techniques for Creating Reusable Software. Reading, Massachusetts: - * Addison-Wesley, 1997. - * - * [Knuth73] Donald E. Knuth. The Art of Computer Programming. Volume 3: - * Sorting and Searching. Reading, Massachusetts: Addison-Wesley, 1973. - * - * [Larson88] Per-Ake Larson. "Dynamic Hash Tables". CACM 31(4), April - * 1988, pp. 446-457. - * - */ - -#include "glxhash.h" -#include - -#define HASH_MAIN 0 - -#include -#include - -#define HASH_MAGIC 0xdeadbeef -#define HASH_DEBUG 0 -#define HASH_SIZE 512 /* Good for about 100 entries */ - /* If you change this value, you probably - have to change the HashHash hashing - function! */ - -#define HASH_ALLOC malloc -#define HASH_FREE free -#define HASH_RANDOM_DECL -#define HASH_RANDOM_INIT(seed) srandom(seed) -#define HASH_RANDOM random() -#define HASH_RANDOM_DESTROY - -typedef struct __glxHashBucket { - unsigned long key; - void *value; - struct __glxHashBucket *next; -} __glxHashBucket, *__glxHashBucketPtr; - -typedef struct __glxHashTable *__glxHashTablePtr; -struct __glxHashTable { - unsigned long magic; - unsigned long entries; - unsigned long hits; /* At top of linked list */ - unsigned long partials; /* Not at top of linked list */ - unsigned long misses; /* Not in table */ - __glxHashBucketPtr buckets[HASH_SIZE]; - int p0; - __glxHashBucketPtr p1; -}; - -static unsigned long HashHash(unsigned long key) -{ - unsigned long hash = 0; - unsigned long tmp = key; - static int init = 0; - static unsigned long scatter[256]; - int i; - - if (!init) { - HASH_RANDOM_DECL; - HASH_RANDOM_INIT(37); - for (i = 0; i < 256; i++) scatter[i] = HASH_RANDOM; - HASH_RANDOM_DESTROY; - ++init; - } - - while (tmp) { - hash = (hash << 1) + scatter[tmp & 0xff]; - tmp >>= 8; - } - - hash %= HASH_SIZE; -#if HASH_DEBUG - printf( "Hash(%d) = %d\n", key, hash); -#endif - return hash; -} - -_X_HIDDEN __glxHashTable *__glxHashCreate(void) -{ - __glxHashTablePtr table; - int i; - - table = HASH_ALLOC(sizeof(*table)); - if (!table) return NULL; - table->magic = HASH_MAGIC; - table->entries = 0; - table->hits = 0; - table->partials = 0; - table->misses = 0; - - for (i = 0; i < HASH_SIZE; i++) table->buckets[i] = NULL; - return table; -} - -_X_HIDDEN int __glxHashDestroy(__glxHashTable *t) -{ - __glxHashTablePtr table = (__glxHashTablePtr)t; - __glxHashBucketPtr bucket; - __glxHashBucketPtr next; - int i; - - if (table->magic != HASH_MAGIC) return -1; /* Bad magic */ - - for (i = 0; i < HASH_SIZE; i++) { - for (bucket = table->buckets[i]; bucket;) { - next = bucket->next; - HASH_FREE(bucket); - bucket = next; - } - } - HASH_FREE(table); - return 0; -} - -/* Find the bucket and organize the list so that this bucket is at the - top. */ - -static __glxHashBucketPtr HashFind(__glxHashTablePtr table, - unsigned long key, unsigned long *h) -{ - unsigned long hash = HashHash(key); - __glxHashBucketPtr prev = NULL; - __glxHashBucketPtr bucket; - - if (h) *h = hash; - - for (bucket = table->buckets[hash]; bucket; bucket = bucket->next) { - if (bucket->key == key) { - if (prev) { - /* Organize */ - prev->next = bucket->next; - bucket->next = table->buckets[hash]; - table->buckets[hash] = bucket; - ++table->partials; - } else { - ++table->hits; - } - return bucket; - } - prev = bucket; - } - ++table->misses; - return NULL; -} - -_X_HIDDEN int __glxHashLookup(__glxHashTable *t, - unsigned long key, void **value) -{ - __glxHashTablePtr table = (__glxHashTablePtr)t; - __glxHashBucketPtr bucket; - - if (!table || table->magic != HASH_MAGIC) return -1; /* Bad magic */ - - bucket = HashFind(table, key, NULL); - if (!bucket) return 1; /* Not found */ - *value = bucket->value; - return 0; /* Found */ -} - -_X_HIDDEN int __glxHashInsert(__glxHashTable *t, - unsigned long key, void *value) -{ - __glxHashTablePtr table = (__glxHashTablePtr)t; - __glxHashBucketPtr bucket; - unsigned long hash; - - if (table->magic != HASH_MAGIC) return -1; /* Bad magic */ - - if (HashFind(table, key, &hash)) return 1; /* Already in table */ - - bucket = HASH_ALLOC(sizeof(*bucket)); - if (!bucket) return -1; /* Error */ - bucket->key = key; - bucket->value = value; - bucket->next = table->buckets[hash]; - table->buckets[hash] = bucket; -#if HASH_DEBUG - printf("Inserted %d at %d/%p\n", key, hash, bucket); -#endif - return 0; /* Added to table */ -} - -_X_HIDDEN int __glxHashDelete(__glxHashTable *t, unsigned long key) -{ - __glxHashTablePtr table = (__glxHashTablePtr)t; - unsigned long hash; - __glxHashBucketPtr bucket; - - if (table->magic != HASH_MAGIC) return -1; /* Bad magic */ - - bucket = HashFind(table, key, &hash); - - if (!bucket) return 1; /* Not found */ - - table->buckets[hash] = bucket->next; - HASH_FREE(bucket); - return 0; -} - -_X_HIDDEN int __glxHashNext(__glxHashTable *t, - unsigned long *key, void **value) -{ - __glxHashTablePtr table = (__glxHashTablePtr)t; - - while (table->p0 < HASH_SIZE) { - if (table->p1) { - *key = table->p1->key; - *value = table->p1->value; - table->p1 = table->p1->next; - return 1; - } - table->p1 = table->buckets[table->p0]; - ++table->p0; - } - return 0; -} - -_X_HIDDEN int __glxHashFirst(__glxHashTable *t, - unsigned long *key, void **value) -{ - __glxHashTablePtr table = (__glxHashTablePtr)t; - - if (table->magic != HASH_MAGIC) return -1; /* Bad magic */ - - table->p0 = 0; - table->p1 = table->buckets[0]; - return __glxHashNext(table, key, value); -} - -#if HASH_MAIN -#define DIST_LIMIT 10 -static int dist[DIST_LIMIT]; - -static void clear_dist(void) { - int i; - - for (i = 0; i < DIST_LIMIT; i++) dist[i] = 0; -} - -static int count_entries(__glxHashBucketPtr bucket) -{ - int count = 0; - - for (; bucket; bucket = bucket->next) ++count; - return count; -} - -static void update_dist(int count) -{ - if (count >= DIST_LIMIT) ++dist[DIST_LIMIT-1]; - else ++dist[count]; -} - -static void compute_dist(__glxHashTablePtr table) -{ - int i; - __glxHashBucketPtr bucket; - - printf("Entries = %ld, hits = %ld, partials = %ld, misses = %ld\n", - table->entries, table->hits, table->partials, table->misses); - clear_dist(); - for (i = 0; i < HASH_SIZE; i++) { - bucket = table->buckets[i]; - update_dist(count_entries(bucket)); - } - for (i = 0; i < DIST_LIMIT; i++) { - if (i != DIST_LIMIT-1) printf("%5d %10d\n", i, dist[i]); - else printf("other %10d\n", dist[i]); - } -} - -static void check_table(__glxHashTablePtr table, - unsigned long key, unsigned long value) -{ - unsigned long retval = 0; - int retcode = __glxHashLookup(table, key, &retval); - - switch (retcode) { - case -1: - printf("Bad magic = 0x%08lx:" - " key = %lu, expected = %lu, returned = %lu\n", - table->magic, key, value, retval); - break; - case 1: - printf("Not found: key = %lu, expected = %lu returned = %lu\n", - key, value, retval); - break; - case 0: - if (value != retval) - printf("Bad value: key = %lu, expected = %lu, returned = %lu\n", - key, value, retval); - break; - default: - printf("Bad retcode = %d: key = %lu, expected = %lu, returned = %lu\n", - retcode, key, value, retval); - break; - } -} - -int main(void) -{ - __glxHashTablePtr table; - int i; - - printf("\n***** 256 consecutive integers ****\n"); - table = __glxHashCreate(); - for (i = 0; i < 256; i++) __glxHashInsert(table, i, i); - for (i = 0; i < 256; i++) check_table(table, i, i); - for (i = 256; i >= 0; i--) check_table(table, i, i); - compute_dist(table); - __glxHashDestroy(table); - - printf("\n***** 1024 consecutive integers ****\n"); - table = __glxHashCreate(); - for (i = 0; i < 1024; i++) __glxHashInsert(table, i, i); - for (i = 0; i < 1024; i++) check_table(table, i, i); - for (i = 1024; i >= 0; i--) check_table(table, i, i); - compute_dist(table); - __glxHashDestroy(table); - - printf("\n***** 1024 consecutive page addresses (4k pages) ****\n"); - table = __glxHashCreate(); - for (i = 0; i < 1024; i++) __glxHashInsert(table, i*4096, i); - for (i = 0; i < 1024; i++) check_table(table, i*4096, i); - for (i = 1024; i >= 0; i--) check_table(table, i*4096, i); - compute_dist(table); - __glxHashDestroy(table); - - printf("\n***** 1024 random integers ****\n"); - table = __glxHashCreate(); - srandom(0xbeefbeef); - for (i = 0; i < 1024; i++) __glxHashInsert(table, random(), i); - srandom(0xbeefbeef); - for (i = 0; i < 1024; i++) check_table(table, random(), i); - srandom(0xbeefbeef); - for (i = 0; i < 1024; i++) check_table(table, random(), i); - compute_dist(table); - __glxHashDestroy(table); - - printf("\n***** 5000 random integers ****\n"); - table = __glxHashCreate(); - srandom(0xbeefbeef); - for (i = 0; i < 5000; i++) __glxHashInsert(table, random(), i); - srandom(0xbeefbeef); - for (i = 0; i < 5000; i++) check_table(table, random(), i); - srandom(0xbeefbeef); - for (i = 0; i < 5000; i++) check_table(table, random(), i); - compute_dist(table); - __glxHashDestroy(table); - - return 0; -} -#endif diff --git a/src/glx/x11/glxhash.h b/src/glx/x11/glxhash.h deleted file mode 100644 index 66012fb889..0000000000 --- a/src/glx/x11/glxhash.h +++ /dev/null @@ -1,16 +0,0 @@ -#ifndef _GLX_HASH_H_ -#define _GLX_HASH_H_ - - -typedef struct __glxHashTable __glxHashTable; - -/* Hash table routines */ -extern __glxHashTable *__glxHashCreate(void); -extern int __glxHashDestroy(__glxHashTable *t); -extern int __glxHashLookup(__glxHashTable *t, unsigned long key, void **value); -extern int __glxHashInsert(__glxHashTable *t, unsigned long key, void *value); -extern int __glxHashDelete(__glxHashTable *t, unsigned long key); -extern int __glxHashFirst(__glxHashTable *t, unsigned long *key, void **value); -extern int __glxHashNext(__glxHashTable *t, unsigned long *key, void **value); - -#endif /* _GLX_HASH_H_ */ diff --git a/src/glx/x11/indirect.c b/src/glx/x11/indirect.c index 871adddb95..fbb2a91956 100644 --- a/src/glx/x11/indirect.c +++ b/src/glx/x11/indirect.c @@ -300,9 +300,7 @@ __indirect_glNewList(GLuint list, GLenum mode) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -326,9 +324,7 @@ __indirect_glEndList(void) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 0; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -364,10 +360,6 @@ __indirect_glCallLists(GLsizei n, GLenum type, const GLvoid * lists) __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint compsize = __glCallLists_size(type); const GLuint cmdlen = 12 + __GLX_PAD((compsize * n)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -401,9 +393,7 @@ __indirect_glDeleteLists(GLuint list, GLsizei range) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -428,9 +418,7 @@ __indirect_glGenLists(GLsizei range) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLuint retval = (GLuint) 0; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3594,10 +3582,6 @@ __indirect_glPixelMapfv(GLenum map, GLsizei mapsize, const GLfloat * values) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((mapsize * 4)); - if (mapsize < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((mapsize >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -3631,10 +3615,6 @@ __indirect_glPixelMapuiv(GLenum map, GLsizei mapsize, const GLuint * values) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((mapsize * 4)); - if (mapsize < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((mapsize >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -3668,10 +3648,6 @@ __indirect_glPixelMapusv(GLenum map, GLsizei mapsize, const GLushort * values) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((mapsize * 2)); - if (mapsize < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((mapsize >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -3740,9 +3716,7 @@ __indirect_glReadPixels(GLint x, GLint y, GLsizei width, GLsizei height, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 28; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3834,9 +3808,7 @@ __indirect_glGetClipPlane(GLenum plane, GLdouble * equation) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3869,9 +3841,7 @@ __indirect_glGetLightfv(GLenum light, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3909,9 +3879,7 @@ __indirect_glGetLightiv(GLenum light, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3949,9 +3917,7 @@ __indirect_glGetMapdv(GLenum target, GLenum query, GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -3987,9 +3953,7 @@ __indirect_glGetMapfv(GLenum target, GLenum query, GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4025,9 +3989,7 @@ __indirect_glGetMapiv(GLenum target, GLenum query, GLint * v) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4063,9 +4025,7 @@ __indirect_glGetMaterialfv(GLenum face, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4103,9 +4063,7 @@ __indirect_glGetMaterialiv(GLenum face, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4143,9 +4101,7 @@ __indirect_glGetPixelMapfv(GLenum map, GLfloat * values) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4181,9 +4137,7 @@ __indirect_glGetPixelMapuiv(GLenum map, GLuint * values) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4220,9 +4174,7 @@ __indirect_glGetPixelMapusv(GLenum map, GLushort * values) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4258,10 +4210,9 @@ void __indirect_glGetPolygonStipple(GLubyte *mask) { __GLXcontext *const gc = __glXGetCurrentContext(); + const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4296,9 +4247,7 @@ __indirect_glGetTexEnvfv(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4336,9 +4285,7 @@ __indirect_glGetTexEnviv(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4376,9 +4323,7 @@ __indirect_glGetTexGendv(GLenum coord, GLenum pname, GLdouble * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4416,9 +4361,7 @@ __indirect_glGetTexGenfv(GLenum coord, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4456,9 +4399,7 @@ __indirect_glGetTexGeniv(GLenum coord, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4498,9 +4439,7 @@ __indirect_glGetTexImage(GLenum target, GLint level, GLenum format, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 20; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4544,9 +4483,7 @@ __indirect_glGetTexParameterfv(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4585,9 +4522,7 @@ __indirect_glGetTexParameteriv(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4627,9 +4562,7 @@ __indirect_glGetTexLevelParameterfv(GLenum target, GLint level, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 12; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4670,9 +4603,7 @@ __indirect_glGetTexLevelParameteriv(GLenum target, GLint level, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 12; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -4713,9 +4644,7 @@ __indirect_glIsList(GLuint list) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5083,13 +5012,7 @@ __indirect_glAreTexturesResident(GLsizei n, const GLuint * textures, __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; -#ifndef USE_XCB const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); -#endif - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return 0; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5124,7 +5047,7 @@ glAreTexturesResidentEXT(GLsizei n, const GLuint * textures, { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { return CALL_AreTexturesResident(GET_DISPATCH(), (n, textures, residences)); } else { @@ -5132,10 +5055,6 @@ glAreTexturesResidentEXT(GLsizei n, const GLuint * textures, Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return 0; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -5244,13 +5163,7 @@ __indirect_glDeleteTextures(GLsizei n, const GLuint * textures) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); -#endif - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5274,16 +5187,12 @@ glDeleteTexturesEXT(GLsizei n, const GLuint * textures) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_DeleteTextures(GET_DISPATCH(), (n, textures)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivate, @@ -5303,13 +5212,7 @@ __indirect_glGenTextures(GLsizei n, GLuint * textures) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5342,16 +5245,12 @@ glGenTexturesEXT(GLsizei n, GLuint * textures) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GenTextures(GET_DISPATCH(), (n, textures)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4; - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -5372,9 +5271,7 @@ __indirect_glIsTexture(GLuint texture) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5404,7 +5301,7 @@ glIsTextureEXT(GLuint texture) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { return CALL_IsTexture(GET_DISPATCH(), (texture)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -5431,10 +5328,6 @@ __indirect_glPrioritizeTextures(GLsizei n, const GLuint * textures, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)) + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_PrioritizeTextures, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&n), 4); @@ -5672,9 +5565,7 @@ __indirect_glGetColorTable(GLenum target, GLenum format, GLenum type, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 16; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5718,7 +5609,7 @@ glGetColorTableEXT(GLenum target, GLenum format, GLenum type, GLvoid * table) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetColorTable(GET_DISPATCH(), (target, format, type, table)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -5750,9 +5641,7 @@ __indirect_glGetColorTableParameterfv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5791,7 +5680,7 @@ glGetColorTableParameterfvEXT(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetColorTableParameterfv(GET_DISPATCH(), (target, pname, params)); } else { @@ -5820,9 +5709,7 @@ __indirect_glGetColorTableParameteriv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -5861,7 +5748,7 @@ glGetColorTableParameterivEXT(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetColorTableParameteriv(GET_DISPATCH(), (target, pname, params)); } else { @@ -6142,9 +6029,7 @@ __indirect_glGetConvolutionFilter(GLenum target, GLenum format, GLenum type, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 16; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6184,7 +6069,7 @@ gl_dispatch_stub_356(GLenum target, GLenum format, GLenum type, { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetConvolutionFilter(GET_DISPATCH(), (target, format, type, image)); } else { @@ -6218,9 +6103,7 @@ __indirect_glGetConvolutionParameterfv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6259,7 +6142,7 @@ gl_dispatch_stub_357(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetConvolutionParameterfv(GET_DISPATCH(), (target, pname, params)); } else { @@ -6288,9 +6171,7 @@ __indirect_glGetConvolutionParameteriv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6329,7 +6210,7 @@ gl_dispatch_stub_358(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetConvolutionParameteriv(GET_DISPATCH(), (target, pname, params)); } else { @@ -6359,9 +6240,7 @@ __indirect_glGetHistogram(GLenum target, GLboolean reset, GLenum format, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 16; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6406,7 +6285,7 @@ gl_dispatch_stub_361(GLenum target, GLboolean reset, GLenum format, { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetHistogram(GET_DISPATCH(), (target, reset, format, type, values)); } else { @@ -6440,9 +6319,7 @@ __indirect_glGetHistogramParameterfv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6480,7 +6357,7 @@ gl_dispatch_stub_362(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetHistogramParameterfv(GET_DISPATCH(), (target, pname, params)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -6508,9 +6385,7 @@ __indirect_glGetHistogramParameteriv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6548,7 +6423,7 @@ gl_dispatch_stub_363(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetHistogramParameteriv(GET_DISPATCH(), (target, pname, params)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -6577,9 +6452,7 @@ __indirect_glGetMinmax(GLenum target, GLboolean reset, GLenum format, __GLXcontext *const gc = __glXGetCurrentContext(); const __GLXattribute *const state = gc->client_state_private; Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 16; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6620,7 +6493,7 @@ gl_dispatch_stub_364(GLenum target, GLboolean reset, GLenum format, { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetMinmax(GET_DISPATCH(), (target, reset, format, type, values)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -6653,9 +6526,7 @@ __indirect_glGetMinmaxParameterfv(GLenum target, GLenum pname, { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6691,7 +6562,7 @@ gl_dispatch_stub_365(GLenum target, GLenum pname, GLfloat * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetMinmaxParameterfv(GET_DISPATCH(), (target, pname, params)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -6718,9 +6589,7 @@ __indirect_glGetMinmaxParameteriv(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -6756,7 +6625,7 @@ gl_dispatch_stub_366(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetMinmaxParameteriv(GET_DISPATCH(), (target, pname, params)); } else { __GLXcontext *const gc = __glXGetCurrentContext(); @@ -7599,26 +7468,6 @@ __indirect_glGetProgramivARB(GLenum target, GLenum pname, GLint * params) return; } -#define X_GLrop_ProgramEnvParameter4dvARB 4185 -void -__indirect_glProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x, - GLdouble y, GLdouble z, GLdouble w) -{ - __GLXcontext *const gc = __glXGetCurrentContext(); - const GLuint cmdlen = 44; - emit_header(gc->pc, X_GLrop_ProgramEnvParameter4dvARB, cmdlen); - (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); - (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); - (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 8); - (void) memcpy((void *) (gc->pc + 20), (void *) (&y), 8); - (void) memcpy((void *) (gc->pc + 28), (void *) (&z), 8); - (void) memcpy((void *) (gc->pc + 36), (void *) (&w), 8); - gc->pc += cmdlen; - if (__builtin_expect(gc->pc > gc->limit, 0)) { - (void) __glXFlushRenderBuffer(gc, gc->pc); - } -} - #define X_GLrop_ProgramEnvParameter4dvARB 4185 void __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, @@ -7636,26 +7485,6 @@ __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, } } -#define X_GLrop_ProgramEnvParameter4fvARB 4184 -void -__indirect_glProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x, - GLfloat y, GLfloat z, GLfloat w) -{ - __GLXcontext *const gc = __glXGetCurrentContext(); - const GLuint cmdlen = 28; - emit_header(gc->pc, X_GLrop_ProgramEnvParameter4fvARB, cmdlen); - (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); - (void) memcpy((void *) (gc->pc + 8), (void *) (&index), 4); - (void) memcpy((void *) (gc->pc + 12), (void *) (&x), 4); - (void) memcpy((void *) (gc->pc + 16), (void *) (&y), 4); - (void) memcpy((void *) (gc->pc + 20), (void *) (&z), 4); - (void) memcpy((void *) (gc->pc + 24), (void *) (&w), 4); - gc->pc += cmdlen; - if (__builtin_expect(gc->pc > gc->limit, 0)) { - (void) __glXFlushRenderBuffer(gc, gc->pc); - } -} - #define X_GLrop_ProgramEnvParameter4fvARB 4184 void __indirect_glProgramEnvParameter4fvARB(GLenum target, GLuint index, @@ -7756,10 +7585,6 @@ __indirect_glProgramStringARB(GLenum target, GLenum format, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 16 + __GLX_PAD(len); - if (len < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((len >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -8375,13 +8200,7 @@ __indirect_glDeleteQueriesARB(GLsizei n, const GLuint * ids) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); -#endif - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8419,13 +8238,7 @@ __indirect_glGenQueriesARB(GLsizei n, GLuint * ids) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8458,9 +8271,7 @@ __indirect_glGetQueryObjectivARB(GLuint id, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8496,9 +8307,7 @@ __indirect_glGetQueryObjectuivARB(GLuint id, GLenum pname, GLuint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8534,9 +8343,7 @@ __indirect_glGetQueryivARB(GLenum target, GLenum pname, GLint * params) { __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; -#ifndef USE_XCB const GLuint cmdlen = 8; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8576,9 +8383,7 @@ __indirect_glIsQueryARB(GLuint id) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; -#ifndef USE_XCB const GLuint cmdlen = 4; -#endif if (__builtin_expect(dpy != NULL, 1)) { #ifdef USE_XCB xcb_connection_t *c = XGetXCBConnection(dpy); @@ -8609,10 +8414,6 @@ __indirect_glDrawBuffersARB(GLsizei n, const GLenum * bufs) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (gc->currentDpy != NULL), 1)) { if (cmdlen <= gc->maxSmallRenderCommandSize) { if ((gc->pc + cmdlen) > gc->bufEnd) { @@ -8973,10 +8774,6 @@ __indirect_glAreProgramsResidentNV(GLsizei n, const GLuint * ids, Display *const dpy = gc->currentDpy; GLboolean retval = (GLboolean) 0; const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return 0; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -9012,10 +8809,6 @@ __indirect_glDeleteProgramsNV(GLsizei n, const GLuint * programs) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivate, @@ -9052,10 +8845,6 @@ __indirect_glGenProgramsNV(GLsizei n, GLuint * programs) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4; - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -9262,10 +9051,6 @@ __indirect_glLoadProgramNV(GLenum target, GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 16 + __GLX_PAD(len); - if (len < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_LoadProgramNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); @@ -9286,10 +9071,6 @@ __indirect_glProgramParameters4dvNV(GLenum target, GLuint index, GLuint num, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 16 + __GLX_PAD((num * 32)); - if (num < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(num >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramParameters4dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); @@ -9310,10 +9091,6 @@ __indirect_glProgramParameters4fvNV(GLenum target, GLuint index, GLuint num, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 16 + __GLX_PAD((num * 16)); - if (num < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(num >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramParameters4fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&target), 4); @@ -9333,10 +9110,6 @@ __indirect_glRequestResidentProgramsNV(GLsizei n, const GLuint * ids) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_RequestResidentProgramsNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&n), 4); @@ -9788,10 +9561,6 @@ __indirect_glVertexAttribs1dvNV(GLuint index, GLsizei n, const GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 8)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs1dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9810,10 +9579,6 @@ __indirect_glVertexAttribs1fvNV(GLuint index, GLsizei n, const GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs1fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9832,10 +9597,6 @@ __indirect_glVertexAttribs1svNV(GLuint index, GLsizei n, const GLshort * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 2)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs1svNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9854,10 +9615,6 @@ __indirect_glVertexAttribs2dvNV(GLuint index, GLsizei n, const GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 16)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs2dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9876,10 +9633,6 @@ __indirect_glVertexAttribs2fvNV(GLuint index, GLsizei n, const GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 8)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs2fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9898,10 +9651,6 @@ __indirect_glVertexAttribs2svNV(GLuint index, GLsizei n, const GLshort * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs2svNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9920,10 +9669,6 @@ __indirect_glVertexAttribs3dvNV(GLuint index, GLsizei n, const GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 24)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs3dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9942,10 +9687,6 @@ __indirect_glVertexAttribs3fvNV(GLuint index, GLsizei n, const GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 12)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs3fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9964,10 +9705,6 @@ __indirect_glVertexAttribs3svNV(GLuint index, GLsizei n, const GLshort * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 6)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs3svNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -9986,10 +9723,6 @@ __indirect_glVertexAttribs4dvNV(GLuint index, GLsizei n, const GLdouble * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 32)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs4dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -10008,10 +9741,6 @@ __indirect_glVertexAttribs4fvNV(GLuint index, GLsizei n, const GLfloat * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 16)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs4fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -10030,10 +9759,6 @@ __indirect_glVertexAttribs4svNV(GLuint index, GLsizei n, const GLshort * v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 8)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs4svNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -10052,10 +9777,6 @@ __indirect_glVertexAttribs4ubvNV(GLuint index, GLsizei n, const GLubyte *v) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 12 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_VertexAttribs4ubvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&index), 4); @@ -10122,10 +9843,6 @@ __indirect_glGetProgramNamedParameterdvNV(GLuint id, GLsizei len, __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 8 + __GLX_PAD(len); - if (len < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((len >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -10150,10 +9867,6 @@ __indirect_glGetProgramNamedParameterfvNV(GLuint id, GLsizei len, __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 8 + __GLX_PAD(len); - if (len < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((len >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -10177,10 +9890,6 @@ __indirect_glProgramNamedParameter4dNV(GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 44 + __GLX_PAD(len); - if (len < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramNamedParameter4dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&x), 8); @@ -10205,10 +9914,6 @@ __indirect_glProgramNamedParameter4dvNV(GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 44 + __GLX_PAD(len); - if (len < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramNamedParameter4dvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (v), 32); @@ -10230,10 +9935,6 @@ __indirect_glProgramNamedParameter4fNV(GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 28 + __GLX_PAD(len); - if (len < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramNamedParameter4fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&id), 4); @@ -10258,10 +9959,6 @@ __indirect_glProgramNamedParameter4fvNV(GLuint id, GLsizei len, { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 28 + __GLX_PAD(len); - if (len < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(len >= 0, 1)) { emit_header(gc->pc, X_GLrop_ProgramNamedParameter4fvNV, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&id), 4); @@ -10347,10 +10044,6 @@ __indirect_glDeleteFramebuffersEXT(GLsizei n, const GLuint * framebuffers) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_DeleteFramebuffersEXT, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&n), 4); @@ -10369,10 +10062,6 @@ __indirect_glDeleteRenderbuffersEXT(GLsizei n, const GLuint * renderbuffers) { __GLXcontext *const gc = __glXGetCurrentContext(); const GLuint cmdlen = 8 + __GLX_PAD((n * 4)); - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect(n >= 0, 1)) { emit_header(gc->pc, X_GLrop_DeleteRenderbuffersEXT, cmdlen); (void) memcpy((void *) (gc->pc + 4), (void *) (&n), 4); @@ -10472,10 +10161,6 @@ __indirect_glGenFramebuffersEXT(GLsizei n, GLuint * framebuffers) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4; - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, @@ -10495,10 +10180,6 @@ __indirect_glGenRenderbuffersEXT(GLsizei n, GLuint * renderbuffers) __GLXcontext *const gc = __glXGetCurrentContext(); Display *const dpy = gc->currentDpy; const GLuint cmdlen = 4; - if (n < 0) { - __glXSetError(gc, GL_INVALID_VALUE); - return; - } if (__builtin_expect((n >= 0) && (dpy != NULL), 1)) { GLubyte const *pc = __glXSetupVendorRequest(gc, X_GLXVendorPrivateWithReply, diff --git a/src/glx/x11/indirect.h b/src/glx/x11/indirect.h index 0719a1b302..f8c88b36bb 100644 --- a/src/glx/x11/indirect.h +++ b/src/glx/x11/indirect.h @@ -517,9 +517,7 @@ extern HIDDEN void __indirect_glGetProgramivARB(GLenum target, GLenum pname, GLi extern HIDDEN void __indirect_glGetVertexAttribdvARB(GLuint index, GLenum pname, GLdouble * params); extern HIDDEN void __indirect_glGetVertexAttribfvARB(GLuint index, GLenum pname, GLfloat * params); extern HIDDEN void __indirect_glGetVertexAttribivARB(GLuint index, GLenum pname, GLint * params); -extern HIDDEN void __indirect_glProgramEnvParameter4dARB(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); extern HIDDEN void __indirect_glProgramEnvParameter4dvARB(GLenum target, GLuint index, const GLdouble * params); -extern HIDDEN void __indirect_glProgramEnvParameter4fARB(GLenum target, GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); extern HIDDEN void __indirect_glProgramEnvParameter4fvARB(GLenum target, GLuint index, const GLfloat * params); extern HIDDEN void __indirect_glProgramLocalParameter4dARB(GLenum target, GLuint index, GLdouble x, GLdouble y, GLdouble z, GLdouble w); extern HIDDEN void __indirect_glProgramLocalParameter4dvARB(GLenum target, GLuint index, const GLdouble * params); diff --git a/src/glx/x11/indirect_init.c b/src/glx/x11/indirect_init.c index 852fe712c6..479184337c 100644 --- a/src/glx/x11/indirect_init.c +++ b/src/glx/x11/indirect_init.c @@ -526,9 +526,7 @@ __GLapi * __glXNewIndirectAPI( void ) glAPI->GetVertexAttribdvARB = __indirect_glGetVertexAttribdvARB; glAPI->GetVertexAttribfvARB = __indirect_glGetVertexAttribfvARB; glAPI->GetVertexAttribivARB = __indirect_glGetVertexAttribivARB; - glAPI->ProgramEnvParameter4dARB = __indirect_glProgramEnvParameter4dARB; glAPI->ProgramEnvParameter4dvARB = __indirect_glProgramEnvParameter4dvARB; - glAPI->ProgramEnvParameter4fARB = __indirect_glProgramEnvParameter4fARB; glAPI->ProgramEnvParameter4fvARB = __indirect_glProgramEnvParameter4fvARB; glAPI->ProgramLocalParameter4dARB = __indirect_glProgramLocalParameter4dARB; glAPI->ProgramLocalParameter4dvARB = __indirect_glProgramLocalParameter4dvARB; diff --git a/src/glx/x11/indirect_texture_compression.c b/src/glx/x11/indirect_texture_compression.c deleted file mode 100644 index 5676858017..0000000000 --- a/src/glx/x11/indirect_texture_compression.c +++ /dev/null @@ -1,347 +0,0 @@ -/* - * (C) Copyright IBM Corporation 2004 - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * on the rights to use, copy, modify, merge, publish, distribute, sub - * license, and/or sell copies of the Software, and to permit persons to whom - * the Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - -/** - * \file glx_texture_compression.c - * Contains the routines required to implement GLX protocol for - * ARB_texture_compression and related extensions. - * - * \sa http://oss.sgi.com/projects/ogl-sample/registry/ARB/texture_compression.txt - * - * \author Ian Romanick - */ - -#include "packrender.h" -#include "packsingle.h" -#include "indirect.h" - -#include - - -void -__indirect_glGetCompressedTexImageARB( GLenum target, GLint level, - GLvoid * img ) -{ - __GLX_SINGLE_DECLARE_VARIABLES(); - xGLXGetTexImageReply reply; - size_t image_bytes; - - __GLX_SINGLE_LOAD_VARIABLES(); - __GLX_SINGLE_BEGIN( X_GLsop_GetCompressedTexImage, 8 ); - __GLX_SINGLE_PUT_LONG( 0, target ); - __GLX_SINGLE_PUT_LONG( 4, level ); - __GLX_SINGLE_READ_XREPLY(); - - image_bytes = reply.width; - assert( image_bytes <= ((4 * reply.length) - 0) ); - assert( image_bytes >= ((4 * reply.length) - 3) ); - - if ( image_bytes != 0 ) { - _XRead( dpy, (char *) img, image_bytes ); - if ( image_bytes < (4 * reply.length) ) { - _XEatData( dpy, (4 * reply.length) - image_bytes ); - } - } - - __GLX_SINGLE_END(); -} - - -/** - * Internal function used for \c glCompressedTexImage1D and - * \c glCompressedTexImage2D. - */ -static void -CompressedTexImage1D2D( GLenum target, GLint level, - GLenum internal_format, - GLsizei width, GLsizei height, - GLint border, GLsizei image_size, - const GLvoid *data, CARD32 rop ) -{ - __GLX_DECLARE_VARIABLES(); - - __GLX_LOAD_VARIABLES(); - if ( gc->currentDpy == NULL ) { - return; - } - - if ( (target == GL_PROXY_TEXTURE_1D) - || (target == GL_PROXY_TEXTURE_2D) - || (target == GL_PROXY_TEXTURE_CUBE_MAP) ) { - compsize = 0; - } - else { - compsize = image_size; - } - - cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE - + compsize ); - if ( cmdlen <= gc->maxSmallRenderCommandSize ) { - __GLX_BEGIN_VARIABLE( rop, cmdlen ); - __GLX_PUT_LONG( 4, target ); - __GLX_PUT_LONG( 8, level ); - __GLX_PUT_LONG( 12, internal_format ); - __GLX_PUT_LONG( 16, width ); - __GLX_PUT_LONG( 20, height ); - __GLX_PUT_LONG( 24, border ); - __GLX_PUT_LONG( 28, image_size ); - if ( compsize != 0 ) { - __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE, - data, image_size ); - } - __GLX_END( cmdlen ); - } - else { - assert( compsize != 0 ); - - __GLX_BEGIN_VARIABLE_LARGE( rop, cmdlen + 4 ); - __GLX_PUT_LONG( 8, target ); - __GLX_PUT_LONG( 12, level ); - __GLX_PUT_LONG( 16, internal_format ); - __GLX_PUT_LONG( 20, width ); - __GLX_PUT_LONG( 24, height ); - __GLX_PUT_LONG( 28, border ); - __GLX_PUT_LONG( 32, image_size ); - __glXSendLargeCommand( gc, gc->pc, - __GLX_COMPRESSED_TEXIMAGE_CMD_HDR_SIZE + 4, - data, image_size ); - } -} - - -/** - * Internal function used for \c glCompressedTexSubImage1D and - * \c glCompressedTexSubImage2D. - */ -static void -CompressedTexSubImage1D2D( GLenum target, GLint level, - GLsizei xoffset, GLsizei yoffset, - GLsizei width, GLsizei height, - GLenum format, GLsizei image_size, - const GLvoid *data, CARD32 rop ) -{ - __GLX_DECLARE_VARIABLES(); - - __GLX_LOAD_VARIABLES(); - if ( gc->currentDpy == NULL ) { - return; - } - - if ( target == GL_PROXY_TEXTURE_3D ) { - compsize = 0; - } - else { - compsize = image_size; - } - - cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE - + compsize ); - if ( cmdlen <= gc->maxSmallRenderCommandSize ) { - __GLX_BEGIN_VARIABLE( rop, cmdlen ); - __GLX_PUT_LONG( 4, target ); - __GLX_PUT_LONG( 8, level ); - __GLX_PUT_LONG( 12, xoffset ); - __GLX_PUT_LONG( 16, yoffset ); - __GLX_PUT_LONG( 20, width ); - __GLX_PUT_LONG( 24, height ); - __GLX_PUT_LONG( 28, format ); - __GLX_PUT_LONG( 32, image_size ); - if ( compsize != 0 ) { - __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE, - data, image_size ); - } - __GLX_END( cmdlen ); - } - else { - assert( compsize != 0 ); - - __GLX_BEGIN_VARIABLE_LARGE( rop, cmdlen + 4 ); - __GLX_PUT_LONG( 8, target ); - __GLX_PUT_LONG( 12, level ); - __GLX_PUT_LONG( 16, xoffset ); - __GLX_PUT_LONG( 20, yoffset ); - __GLX_PUT_LONG( 24, width ); - __GLX_PUT_LONG( 28, height ); - __GLX_PUT_LONG( 32, format ); - __GLX_PUT_LONG( 36, image_size ); - __glXSendLargeCommand( gc, gc->pc, - __GLX_COMPRESSED_TEXSUBIMAGE_CMD_HDR_SIZE + 4, - data, image_size ); - } -} - - -void -__indirect_glCompressedTexImage1DARB( GLenum target, GLint level, - GLenum internal_format, GLsizei width, - GLint border, GLsizei image_size, - const GLvoid *data ) -{ - CompressedTexImage1D2D( target, level, internal_format, width, 0, - border, image_size, data, - X_GLrop_CompressedTexImage1D ); -} - - -void -__indirect_glCompressedTexImage2DARB( GLenum target, GLint level, - GLenum internal_format, - GLsizei width, GLsizei height, - GLint border, GLsizei image_size, - const GLvoid *data ) -{ - CompressedTexImage1D2D( target, level, internal_format, width, height, - border, image_size, data, - X_GLrop_CompressedTexImage2D ); -} - - -void -__indirect_glCompressedTexImage3DARB( GLenum target, GLint level, - GLenum internal_format, - GLsizei width, GLsizei height, GLsizei depth, - GLint border, GLsizei image_size, - const GLvoid *data ) -{ - __GLX_DECLARE_VARIABLES(); - - __GLX_LOAD_VARIABLES(); - if ( gc->currentDpy == NULL ) { - return; - } - - cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE - + image_size ); - if ( cmdlen <= gc->maxSmallRenderCommandSize ) { - __GLX_BEGIN_VARIABLE( X_GLrop_CompressedTexImage3D, cmdlen ); - __GLX_PUT_LONG( 4, target ); - __GLX_PUT_LONG( 8, level ); - __GLX_PUT_LONG( 12, internal_format ); - __GLX_PUT_LONG( 16, width ); - __GLX_PUT_LONG( 20, height ); - __GLX_PUT_LONG( 24, depth ); - __GLX_PUT_LONG( 28, border ); - __GLX_PUT_LONG( 32, image_size ); - if ( image_size != 0 ) { - __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE, - data, image_size ); - } - __GLX_END( cmdlen ); - } - else { - __GLX_BEGIN_VARIABLE_LARGE( X_GLrop_CompressedTexImage3D, - cmdlen + 4 ); - __GLX_PUT_LONG( 8, target ); - __GLX_PUT_LONG( 12, level ); - __GLX_PUT_LONG( 16, internal_format ); - __GLX_PUT_LONG( 20, width ); - __GLX_PUT_LONG( 24, height ); - __GLX_PUT_LONG( 28, depth ); - __GLX_PUT_LONG( 32, border ); - __GLX_PUT_LONG( 36, image_size ); - __glXSendLargeCommand( gc, gc->pc, - __GLX_COMPRESSED_TEXIMAGE_3D_CMD_HDR_SIZE + 4, - data, image_size ); - } -} - - -void -__indirect_glCompressedTexSubImage1DARB( GLenum target, GLint level, - GLint xoffset, - GLsizei width, - GLenum format, GLsizei image_size, - const GLvoid *data ) -{ - CompressedTexSubImage1D2D( target, level, xoffset, 0, width, 0, - format, image_size, data, - X_GLrop_CompressedTexSubImage1D ); -} - - -void -__indirect_glCompressedTexSubImage2DARB( GLenum target, GLint level, - GLint xoffset, GLint yoffset, - GLsizei width, GLsizei height, - GLenum format, GLsizei image_size, - const GLvoid *data ) -{ - CompressedTexSubImage1D2D( target, level, xoffset, yoffset, width, height, - format, image_size, data, - X_GLrop_CompressedTexSubImage2D ); -} - - -void -__indirect_glCompressedTexSubImage3DARB( GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLsizei width, GLsizei height, GLsizei depth, - GLenum format, GLsizei image_size, - const GLvoid *data ) -{ - __GLX_DECLARE_VARIABLES(); - - __GLX_LOAD_VARIABLES(); - if ( gc->currentDpy == NULL ) { - return; - } - - cmdlen = __GLX_PAD( __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE - + image_size ); - if ( cmdlen <= gc->maxSmallRenderCommandSize ) { - __GLX_BEGIN_VARIABLE( X_GLrop_CompressedTexSubImage3D, cmdlen ); - __GLX_PUT_LONG( 4, target ); - __GLX_PUT_LONG( 8, level ); - __GLX_PUT_LONG( 12, xoffset ); - __GLX_PUT_LONG( 16, yoffset ); - __GLX_PUT_LONG( 20, zoffset ); - __GLX_PUT_LONG( 24, width ); - __GLX_PUT_LONG( 28, height ); - __GLX_PUT_LONG( 32, depth ); - __GLX_PUT_LONG( 36, format ); - __GLX_PUT_LONG( 40, image_size ); - if ( image_size != 0 ) { - __GLX_PUT_CHAR_ARRAY( __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE, - data, image_size ); - } - __GLX_END( cmdlen ); - } - else { - __GLX_BEGIN_VARIABLE_LARGE( X_GLrop_CompressedTexSubImage3D, - cmdlen + 4 ); - __GLX_PUT_LONG( 8, target ); - __GLX_PUT_LONG( 12, level ); - __GLX_PUT_LONG( 16, xoffset ); - __GLX_PUT_LONG( 20, yoffset ); - __GLX_PUT_LONG( 24, zoffset ); - __GLX_PUT_LONG( 28, width ); - __GLX_PUT_LONG( 32, height ); - __GLX_PUT_LONG( 36, depth ); - __GLX_PUT_LONG( 40, format ); - __GLX_PUT_LONG( 44, image_size ); - __glXSendLargeCommand( gc, gc->pc, - __GLX_COMPRESSED_TEXSUBIMAGE_3D_CMD_HDR_SIZE + 4, - data, image_size ); - } -} diff --git a/src/glx/x11/indirect_vertex_array.c b/src/glx/x11/indirect_vertex_array.c index 09d7244ba9..90ec277c41 100644 --- a/src/glx/x11/indirect_vertex_array.c +++ b/src/glx/x11/indirect_vertex_array.c @@ -32,7 +32,7 @@ #include #include "glxextensions.h" #include "indirect_vertex_array.h" -#include "indirect_vertex_array_priv.h" +#include "indirect_va_private.h" #define __GLX_PAD(n) (((n)+3) & ~3) @@ -485,14 +485,14 @@ emit_DrawArrays_none( GLenum mode, GLint first, GLsizei count ) for ( i = 0 ; i < count ; i++ ) { if ( (pc + single_vertex_size) >= gc->bufEnd ) { - pc = __glXFlushRenderBuffer(gc, pc); + pc = __glXFlushRenderBuffer(gc, gc->pc); } pc = emit_element_none( pc, arrays, first + i ); } if ( (pc + 4) >= gc->bufEnd ) { - pc = __glXFlushRenderBuffer(gc, pc); + pc = __glXFlushRenderBuffer(gc, gc->pc); } (void) memcpy( pc, end_cmd, 4 ); @@ -527,7 +527,7 @@ static GLubyte * emit_DrawArrays_header_old( __GLXcontext * gc, struct array_state_vector * arrays, size_t * elements_per_request, - unsigned int * total_requests, + size_t * total_requests, GLenum mode, GLsizei count ) { size_t command_size; @@ -640,7 +640,7 @@ emit_DrawArrays_old( GLenum mode, GLint first, GLsizei count ) GLubyte * pc; size_t elements_per_request; - unsigned total_requests = 0; + size_t total_requests = 0; unsigned i; size_t total_sent = 0; @@ -726,7 +726,7 @@ emit_DrawElements_none( GLenum mode, GLsizei count, GLenum type, unsigned index = 0; if ( (pc + single_vertex_size) >= gc->bufEnd ) { - pc = __glXFlushRenderBuffer(gc, pc); + pc = __glXFlushRenderBuffer(gc, gc->pc); } switch( type ) { @@ -744,7 +744,7 @@ emit_DrawElements_none( GLenum mode, GLsizei count, GLenum type, } if ( (pc + 4) >= gc->bufEnd ) { - pc = __glXFlushRenderBuffer(gc, pc); + pc = __glXFlushRenderBuffer(gc, gc->pc); } (void) memcpy( pc, end_cmd, 4 ); @@ -770,10 +770,9 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, GLubyte * pc; size_t elements_per_request; - unsigned total_requests = 0; + size_t total_requests = 0; unsigned i; unsigned req; - unsigned req_element=0; pc = emit_DrawArrays_header_old( gc, arrays, & elements_per_request, @@ -791,7 +790,7 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, switch( type ) { case GL_UNSIGNED_INT: { - const GLuint * ui_ptr = (const GLuint *) indices + req_element; + const GLuint * ui_ptr = (const GLuint *) indices; for ( i = 0 ; i < elements_per_request ; i++ ) { const GLint index = (GLint) *(ui_ptr++); @@ -800,7 +799,7 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, break; } case GL_UNSIGNED_SHORT: { - const GLushort * us_ptr = (const GLushort *) indices + req_element; + const GLushort * us_ptr = (const GLushort *) indices; for ( i = 0 ; i < elements_per_request ; i++ ) { const GLint index = (GLint) *(us_ptr++); @@ -809,7 +808,7 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, break; } case GL_UNSIGNED_BYTE: { - const GLubyte * ub_ptr = (const GLubyte *) indices + req_element; + const GLubyte * ub_ptr = (const GLubyte *) indices; for ( i = 0 ; i < elements_per_request ; i++ ) { const GLint index = (GLint) *(ub_ptr++); @@ -827,7 +826,6 @@ emit_DrawElements_old( GLenum mode, GLsizei count, GLenum type, } count -= elements_per_request; - req_element += elements_per_request; } diff --git a/src/glx/x11/indirect_vertex_array_priv.h b/src/glx/x11/indirect_vertex_array_priv.h deleted file mode 100644 index ab97dc645f..0000000000 --- a/src/glx/x11/indirect_vertex_array_priv.h +++ /dev/null @@ -1,308 +0,0 @@ -/* - * (C) Copyright IBM Corporation 2004, 2005 - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sub license, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the next - * paragraph) shall be included in all copies or substantial portions of the - * Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * IBM, - * AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, - * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF - * OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE - * SOFTWARE. - */ - -#ifndef _INDIRECT_VA_PRIVATE_ -#define _INDIRECT_VA_PRIVATE_ - -/** - * \file indirect_va_private.h - * - * \author Ian Romanick - */ - -#include - -#include "glxclient.h" -#include "indirect.h" -#include - - -/** - * State descriptor for a single array of vertex data. - */ -struct array_state { - /** - * Pointer to the application supplied data. - */ - const void * data; - - /** - * Enum representing the type of the application supplied data. - */ - GLenum data_type; - - /** - * Stride value supplied by the application. This value is not used - * internally. It is only kept so that it can be queried by the - * application using glGet*v. - */ - GLsizei user_stride; - - /** - * Calculated size, in bytes, of a single element in the array. This - * is calculated based on \c count and the size of the data type - * represented by \c data_type. - */ - GLsizei element_size; - - /** - * Actual byte-stride from one element to the next. This value will - * be equal to either \c user_stride or \c element_stride. - */ - GLsizei true_stride; - - /** - * Number of data values in each element. - */ - GLint count; - - /** - * "Normalized" data is on the range [0,1] (unsigned) or [-1,1] (signed). - * This is used for mapping integral types to floating point types. - */ - GLboolean normalized; - - /** - * Pre-calculated GLX protocol command header. - */ - uint32_t header[2]; - - /** - * Size of the header data. For simple data, like glColorPointerfv, - * this is 4. For complex data that requires either a count (e.g., - * glWeightfvARB), an index (e.g., glVertexAttrib1fvARB), or a - * selector enum (e.g., glMultiTexCoord2fv) this is 8. - */ - unsigned header_size; - - /** - * Set to \c GL_TRUE if this array is enabled. Otherwise, it is set - * to \c GL_FALSE. - */ - GLboolean enabled; - - /** - * For multi-arrayed data (e.g., texture coordinates, generic vertex - * program attributes, etc.), this specifies which array this is. - */ - unsigned index; - - /** - * Per-array-type key. For most arrays, this will be the GL enum for - * that array (e.g., GL_VERTEX_ARRAY for vertex data, GL_NORMAL_ARRAY - * for normal data, GL_TEXTURE_COORD_ARRAY for texture coordinate data, - * etc.). - */ - GLenum key; - - /** - * If this array can be used with the "classic" \c glDrawArrays protocol, - * this is set to \c GL_TRUE. Otherwise, it is set to \c GL_FALSE. - */ - GLboolean old_DrawArrays_possible; -}; - - -/** - * Array state that is pushed / poped by \c glPushClientAttrib and - * \c glPopClientAttrib. - */ -struct array_stack_state { - /** - * Pointer to the application supplied data. - */ - const void * data; - - /** - * Enum representing the type of the application supplied data. - */ - GLenum data_type; - - /** - * Stride value supplied by the application. This value is not used - * internally. It is only kept so that it can be queried by the - * application using glGet*v. - */ - GLsizei user_stride; - - /** - * Number of data values in each element. - */ - GLint count; - - /** - * Per-array-type key. For most arrays, this will be the GL enum for - * that array (e.g., GL_VERTEX_ARRAY for vertex data, GL_NORMAL_ARRAY - * for normal data, GL_TEXTURE_COORD_ARRAY for texture coordinate data, - * etc.). - */ - GLenum key; - - /** - * For multi-arrayed data (e.g., texture coordinates, generic vertex - * program attributes, etc.), this specifies which array this is. - */ - unsigned index; - - /** - * Set to \c GL_TRUE if this array is enabled. Otherwise, it is set - * to \c GL_FALSE. - */ - GLboolean enabled; -}; - - -/** - * Collection of all the vertex array state. - */ -struct array_state_vector { - /** - * Number of arrays tracked by \c ::arrays. - */ - size_t num_arrays; - - /** - * Array of vertex array state. This array contains all of the valid - * vertex arrays. If a vertex array isn't in this array, then it isn't - * valid. For example, if an implementation does not support - * EXT_fog_coord, there won't be a GL_FOG_COORD_ARRAY entry in this - * array. - */ - struct array_state * arrays; - - /** - * Number of currently enabled client-side arrays. The value of this - * field is only valid if \c array_info_cache_valid is true. - */ - size_t enabled_client_array_count; - - /** - * \name ARRAY_INFO cache. - * - * These fields track the state of the ARRAY_INFO cache. The - * \c array_info_cache_size is the size of the actual data stored in - * \c array_info_cache. \c array_info_cache_buffer_size is the size of - * the buffer. This will always be greater than or equal to - * \c array_info_cache_size. - * - * \note - * There are some bytes of extra data before \c array_info_cache that is - * used to hold the header for RenderLarge commands. This is - * \b not included in \c array_info_cache_size or - * \c array_info_cache_buffer_size. \c array_info_cache_base stores a - * pointer to the true start of the buffer (i.e., what malloc returned). - */ - /*@{*/ - size_t array_info_cache_size; - size_t array_info_cache_buffer_size; - void * array_info_cache; - void * array_info_cache_base; - /*@}*/ - - - /** - * Is the cache of ARRAY_INFO data valid? The cache can become invalid - * when one of several state changes occur. Among these chages are - * modifying the array settings for an enabled array and enabling / - * disabling an array. - */ - GLboolean array_info_cache_valid; - - /** - * Is it possible to use the GL 1.1 / EXT_vertex_arrays protocol? Use - * of this protocol is disabled with really old servers (i.e., servers - * that don't support GL 1.1 or EXT_vertex_arrays) or when an environment - * variable is set. - * - * \todo - * GL 1.1 and EXT_vertex_arrays use identical protocol, but have different - * opcodes for \c glDrawArrays. For servers that advertise one or the - * other, there should be a way to select which opcode to use. - */ - GLboolean old_DrawArrays_possible; - - /** - * Is it possible to use the new GL X.X / ARB_vertex_buffer_object - * protocol? - * - * \todo - * This protocol has not yet been defined by the ARB, but is currently a - * work in progress. This field is a place-holder. - */ - GLboolean new_DrawArrays_possible; - - /** - * Active texture unit set by \c glClientActiveTexture. - * - * \sa __glXGetActiveTextureUnit - */ - unsigned active_texture_unit; - - /** - * Number of supported texture units. Even if ARB_multitexture / - * GL 1.3 are not supported, this will be at least 1. When multitexture - * is supported, this will be the value queried by calling - * \c glGetIntegerv with \c GL_MAX_TEXTURE_UNITS. - * - * \todo - * Investigate if this should be the value of \c GL_MAX_TEXTURE_COORDS - * instead (if GL 2.0 / ARB_fragment_shader / ARB_fragment_program / - * NV_fragment_program are supported). - */ - unsigned num_texture_units; - - /** - * Number of generic vertex program attribs. If GL_ARB_vertex_program - * is not supported, this will be zero. Otherwise it will be the value - * queries by calling \c glGetProgramiv with \c GL_VERTEX_PROGRAM_ARB - * and \c GL_MAX_PROGRAM_ATTRIBS_ARB. - */ - unsigned num_vertex_program_attribs; - - /** - * \n Methods for implementing various GL functions. - * - * These method pointers are only valid \c array_info_cache_valid is set. - * When each function starts, it much check \c array_info_cache_valid. - * If it is not set, it must call \c fill_array_info_cache and call - * the new method. - * - * \sa fill_array_info_cache - * - * \todo - * Write code to plug these functions directly into the dispatch table. - */ - /*@{*/ - void (*DrawArrays)( GLenum, GLint, GLsizei ); - void (*DrawElements)( GLenum mode, GLsizei count, GLenum type, - const GLvoid *indices ); - /*@}*/ - - struct array_stack_state * stack; - unsigned active_texture_unit_stack[ __GL_CLIENT_ATTRIB_STACK_DEPTH ]; - unsigned stack_index; -}; - -#endif /* _INDIRECT_VA_PRIVATE_ */ diff --git a/src/glx/x11/singlepix.c b/src/glx/x11/singlepix.c index 144e5df743..cd88684f70 100644 --- a/src/glx/x11/singlepix.c +++ b/src/glx/x11/singlepix.c @@ -117,7 +117,7 @@ void NAME(_gloffset_GetSeparableFilter)(GLenum target, GLenum format, GLenum typ { __GLXcontext * const gc = __glXGetCurrentContext(); - if (gc->driContext) { + if (gc->isDirect) { CALL_GetSeparableFilter(GET_DISPATCH(), (target, format, type, row, column, span)); return; diff --git a/src/glx/x11/xf86dri.h b/src/glx/x11/xf86dri.h index a6a57c3135..c8c878f127 100644 --- a/src/glx/x11/xf86dri.h +++ b/src/glx/x11/xf86dri.h @@ -64,6 +64,8 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. #ifndef _XF86DRI_SERVER_ +#include + _XFUNCPROTOBEGIN Bool XF86DRIQueryExtension( Display *dpy, int *event_base, int *error_base ); @@ -91,14 +93,14 @@ Bool XF86DRICreateContext( Display *dpy, int screen, Visual *visual, Bool XF86DRICreateContextWithConfig( Display *dpy, int screen, int configID, XID *ptr_to_returned_context_id, drm_context_t *hHWContext ); -extern GLboolean XF86DRIDestroyContext( Display *dpy, int screen, - XID context_id ); +extern GLboolean XF86DRIDestroyContext( __DRInativeDisplay *dpy, int screen, + __DRIid context_id ); -extern GLboolean XF86DRICreateDrawable( Display *dpy, int screen, - XID drawable, drm_drawable_t *hHWDrawable ); +extern GLboolean XF86DRICreateDrawable( __DRInativeDisplay *dpy, int screen, + __DRIid drawable, drm_drawable_t *hHWDrawable ); -extern GLboolean XF86DRIDestroyDrawable( Display *dpy, int screen, - XID drawable); +extern GLboolean XF86DRIDestroyDrawable( __DRInativeDisplay *dpy, int screen, + __DRIid drawable); Bool XF86DRIGetDrawableInfo( Display *dpy, int screen, Drawable drawable, unsigned int *index, unsigned int *stamp, diff --git a/src/glx/x11/xfont.c b/src/glx/x11/xfont.c index 6ec8c2d6bf..f3e3da3e79 100644 --- a/src/glx/x11/xfont.c +++ b/src/glx/x11/xfont.c @@ -33,7 +33,6 @@ called by that routine when direct rendering is enabled. */ -#ifdef GLX_DIRECT_RENDERING #include "glxclient.h" @@ -209,7 +208,8 @@ static XCharStruct *isvalid(XFontStruct *fs, int which) return(NULL); } -_X_HIDDEN void DRI_glXUseXFont( Font font, int first, int count, int listbase ) + +void DRI_glXUseXFont( Font font, int first, int count, int listbase ) { GLXContext CC; Display *dpy; @@ -373,4 +373,4 @@ bm_height); glPixelStorei(GL_UNPACK_ALIGNMENT, alignment); } -#endif +/* The End. */ diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c index 2013125efe..fff79c36ad 100644 --- a/src/mesa/drivers/dri/common/dri_util.c +++ b/src/mesa/drivers/dri/common/dri_util.c @@ -55,19 +55,39 @@ #include "dri_util.h" #include "drm_sarea.h" -#include "utils.h" #ifndef GLX_OML_sync_control -typedef GLboolean ( * PFNGLXGETMSCRATEOMLPROC) (__DRIdrawable *drawable, int32_t *numerator, int32_t *denominator); +typedef GLboolean ( * PFNGLXGETMSCRATEOMLPROC) (__DRInativeDisplay *dpy, __DRIid drawable, int32_t *numerator, int32_t *denominator); #endif +/* This pointer *must* be set by the driver's __driCreateNewScreen funciton! + */ +const __DRIinterfaceMethods * dri_interface = NULL; + /** - * This is just a token extension used to signal that the driver - * supports setting a read drawable. + * This is used in a couple of places that call \c driCreateNewDrawable. */ -const __DRIextension driReadDrawableExtension = { - __DRI_READ_DRAWABLE, __DRI_READ_DRAWABLE_VERSION -}; +static const int empty_attribute_list[1] = { None }; + + +/** + * Cached copy of the internal API version used by libGL and the client-side + * DRI driver. + */ +static int api_ver = 0; + +/* forward declarations */ +static int driQueryFrameTracking( __DRInativeDisplay *dpy, void *priv, + int64_t *sbc, int64_t *missedFrames, + float *lastMissedUsage, float *usage ); + +static void *driCreateNewDrawable(__DRInativeDisplay *dpy, + const __GLcontextModes *modes, + __DRIid draw, __DRIdrawable *pdraw, + int renderType, const int *attrs); + +static void driDestroyDrawable(__DRInativeDisplay *dpy, void *drawablePrivate); + /** * Print message to \c stderr if the \c LIBGL_DEBUG environment variable @@ -91,19 +111,64 @@ __driUtilMessage(const char *f, ...) } } -GLint -driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ) + +/*****************************************************************/ +/** \name Drawable list management */ +/*****************************************************************/ +/*@{*/ + +static GLboolean __driAddDrawable(void *drawHash, __DRIdrawable *pdraw) { - if (rect2.x1 > rect1.x1) rect1.x1 = rect2.x1; - if (rect2.x2 < rect1.x2) rect1.x2 = rect2.x2; - if (rect2.y1 > rect1.y1) rect1.y1 = rect2.y1; - if (rect2.y2 < rect1.y2) rect1.y2 = rect2.y2; + __DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *)pdraw->private; - if (rect1.x1 > rect1.x2 || rect1.y1 > rect1.y2) return 0; + if (drmHashInsert(drawHash, pdp->draw, pdraw)) + return GL_FALSE; - return (rect1.x2 - rect1.x1) * (rect1.y2 - rect1.y1); + return GL_TRUE; +} + +static __DRIdrawable *__driFindDrawable(void *drawHash, __DRIid draw) +{ + int retcode; + __DRIdrawable *pdraw; + + retcode = drmHashLookup(drawHash, draw, (void *)&pdraw); + if (retcode) + return NULL; + + return pdraw; +} + + +/** + * Find drawables in the local hash that have been destroyed on the + * server. + * + * \param drawHash Hash-table containing all known drawables. + */ +static void __driGarbageCollectDrawables(void *drawHash) +{ + __DRIid draw; + __DRInativeDisplay *dpy; + __DRIdrawable *pdraw; + + if (drmHashFirst(drawHash, &draw, (void *)&pdraw) == 1) { + do { + __DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *)pdraw->private; + dpy = pdp->driScreenPriv->display; + if (! (*dri_interface->windowExists)(dpy, draw)) { + /* Destroy the local drawable data, if the drawable no + longer exists in the Xserver */ + (*pdraw->destroyDrawable)(dpy, pdraw->private); + _mesa_free(pdraw); + } + } while (drmHashNext(drawHash, &draw, (void *)&pdraw) == 1); + } } +/*@}*/ + + /*****************************************************************/ /** \name Context (un)binding functions */ /*****************************************************************/ @@ -112,7 +177,10 @@ driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ) /** * Unbind context. * - * \param scrn the screen. + * \param dpy the display handle. + * \param scrn the screen number. + * \param draw drawable. + * \param read Current reading drawable. * \param gc context. * * \return \c GL_TRUE on success, or \c GL_FALSE on failure. @@ -125,27 +193,56 @@ driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ) * While casting the opaque private pointers associated with the parameters * into their respective real types it also assures they are not \c NULL. */ -static int driUnbindContext(__DRIcontext *pcp) +static GLboolean driUnbindContext(__DRInativeDisplay *dpy, int scrn, + __DRIid draw, __DRIid read, + __DRIcontext *ctx) { - __DRIscreen *psp; - __DRIdrawable *pdp; - __DRIdrawable *prp; + __DRIscreen *pDRIScreen; + __DRIdrawable *pdraw; + __DRIdrawable *pread; + __DRIcontextPrivate *pcp; + __DRIscreenPrivate *psp; + __DRIdrawablePrivate *pdp; + __DRIdrawablePrivate *prp; /* ** Assume error checking is done properly in glXMakeCurrent before ** calling driUnbindContext. */ - if (pcp == NULL) - return GL_FALSE; + if (ctx == NULL || draw == None || read == None) { + /* ERROR!!! */ + return GL_FALSE; + } + + pDRIScreen = (*dri_interface->getScreen)(dpy, scrn); + if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) { + /* ERROR!!! */ + return GL_FALSE; + } + + psp = (__DRIscreenPrivate *)pDRIScreen->private; + pcp = (__DRIcontextPrivate *)ctx->private; + + pdraw = __driFindDrawable(psp->drawHash, draw); + if (!pdraw) { + /* ERROR!!! */ + return GL_FALSE; + } + pdp = (__DRIdrawablePrivate *)pdraw->private; + + pread = __driFindDrawable(psp->drawHash, read); + if (!pread) { + /* ERROR!!! */ + return GL_FALSE; + } + prp = (__DRIdrawablePrivate *)pread->private; - psp = pcp->driScreenPriv; - pdp = pcp->driDrawablePriv; - prp = pcp->driReadablePriv; /* Let driver unbind drawable from context */ (*psp->DriverAPI.UnbindContext)(pcp); + if (pdp->refcount == 0) { /* ERROR!!! */ return GL_FALSE; @@ -162,6 +259,12 @@ static int driUnbindContext(__DRIcontext *pcp) prp->refcount--; } + /* destroy the drawables if they no longer exist on the server */ + if ((pdp->refcount == 0) || (prp->refcount == 0)) { + /* probably shouldn't need the collector here, + as we know the affected drawables (or could there be others?) */ + __driGarbageCollectDrawables(pdp->driScreenPriv->drawHash); + } /* XXX this is disabled so that if we call SwapBuffers on an unbound * window we can determine the last context bound to the window and @@ -181,20 +284,72 @@ static int driUnbindContext(__DRIcontext *pcp) * This function takes both a read buffer and a draw buffer. This is needed * for \c glXMakeCurrentReadSGI or GLX 1.3's \c glXMakeContextCurrent * function. + * + * \bug This function calls \c driCreateNewDrawable in two places with the + * \c renderType hard-coded to \c GLX_WINDOW_BIT. Some checking might + * be needed in those places when support for pbuffers and / or pixmaps + * is added. Is it safe to assume that the drawable is a window? */ -static int driBindContext(__DRIcontext *pcp, - __DRIdrawable *pdp, - __DRIdrawable *prp) +static GLboolean DoBindContext(__DRInativeDisplay *dpy, + __DRIid draw, __DRIid read, + __DRIcontext *ctx, const __GLcontextModes * modes, + __DRIscreenPrivate *psp) { - __DRIscreenPrivate *psp = pcp->driScreenPriv; + __DRIdrawable *pdraw; + __DRIdrawablePrivate *pdp; + __DRIdrawable *pread; + __DRIdrawablePrivate *prp; + __DRIcontextPrivate * const pcp = ctx->private; + + + /* Find the _DRIdrawable which corresponds to the writing drawable. */ + pdraw = __driFindDrawable(psp->drawHash, draw); + if (!pdraw) { + /* Allocate a new drawable */ + pdraw = (__DRIdrawable *)_mesa_malloc(sizeof(__DRIdrawable)); + if (!pdraw) { + /* ERROR!!! */ + return GL_FALSE; + } - /* - ** Assume error checking is done properly in glXMakeCurrent before - ** calling driBindContext. - */ + /* Create a new drawable */ + driCreateNewDrawable(dpy, modes, draw, pdraw, GLX_WINDOW_BIT, + empty_attribute_list); + if (!pdraw->private) { + /* ERROR!!! */ + _mesa_free(pdraw); + return GL_FALSE; + } - if (pcp == NULL || pdp == None || prp == None) - return GL_FALSE; + } + pdp = (__DRIdrawablePrivate *) pdraw->private; + + /* Find the _DRIdrawable which corresponds to the reading drawable. */ + if (read == draw) { + /* read buffer == draw buffer */ + prp = pdp; + } + else { + pread = __driFindDrawable(psp->drawHash, read); + if (!pread) { + /* Allocate a new drawable */ + pread = (__DRIdrawable *)_mesa_malloc(sizeof(__DRIdrawable)); + if (!pread) { + /* ERROR!!! */ + return GL_FALSE; + } + + /* Create a new drawable */ + driCreateNewDrawable(dpy, modes, read, pread, GLX_WINDOW_BIT, + empty_attribute_list); + if (!pread->private) { + /* ERROR!!! */ + _mesa_free(pread); + return GL_FALSE; + } + } + prp = (__DRIdrawablePrivate *) pread->private; + } /* Bind the drawable to the context */ pcp->driDrawablePriv = pdp; @@ -209,22 +364,16 @@ static int driBindContext(__DRIcontext *pcp, ** Now that we have a context associated with this drawable, we can ** initialize the drawable information if has not been done before. */ + if (!pdp->pStamp || *pdp->pStamp != pdp->lastStamp) { + DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); + __driUtilUpdateDrawableInfo(pdp); + DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); + } - if (psp->dri2.enabled) { - __driParseEvents(pcp, pdp); - __driParseEvents(pcp, prp); - } else { - if (!pdp->pStamp || *pdp->pStamp != pdp->lastStamp) { - DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); - __driUtilUpdateDrawableInfo(pdp); - DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); - } - - if ((pdp != prp) && (!prp->pStamp || *prp->pStamp != prp->lastStamp)) { - DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); - __driUtilUpdateDrawableInfo(prp); - DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); - } + if ((pdp != prp) && (!prp->pStamp || *prp->pStamp != prp->lastStamp)) { + DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); + __driUtilUpdateDrawableInfo(prp); + DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); } /* Call device-specific MakeCurrent */ @@ -233,6 +382,37 @@ static int driBindContext(__DRIcontext *pcp, return GL_TRUE; } + +/** + * This function takes both a read buffer and a draw buffer. This is needed + * for \c glXMakeCurrentReadSGI or GLX 1.3's \c glXMakeContextCurrent + * function. + */ +static GLboolean driBindContext(__DRInativeDisplay *dpy, int scrn, + __DRIid draw, __DRIid read, + __DRIcontext * ctx) +{ + __DRIscreen *pDRIScreen; + + /* + ** Assume error checking is done properly in glXMakeCurrent before + ** calling driBindContext. + */ + + if (ctx == NULL || draw == None || read == None) { + /* ERROR!!! */ + return GL_FALSE; + } + + pDRIScreen = (*dri_interface->getScreen)(dpy, scrn); + if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) { + /* ERROR!!! */ + return GL_FALSE; + } + + return DoBindContext( dpy, draw, read, ctx, ctx->mode, + (__DRIscreenPrivate *)pDRIScreen->private ); +} /*@}*/ @@ -256,7 +436,7 @@ static int driBindContext(__DRIcontext *pcp, void __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) { - __DRIscreenPrivate *psp = pdp->driScreenPriv; + __DRIscreenPrivate *psp; __DRIcontextPrivate *pcp = pdp->driContextPriv; if (!pcp @@ -267,6 +447,15 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) */ } + psp = pdp->driScreenPriv; + if (!psp) { + /* ERROR!!! */ + _mesa_problem(NULL, "Warning! Possible infinite loop due to bug " + "in file %s, line %d\n", + __FILE__, __LINE__); + return; + } + if (pdp->pClipRects) { _mesa_free(pdp->pClipRects); pdp->pClipRects = NULL; @@ -279,15 +468,15 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID); - if (! (*psp->getDrawableInfo->getDrawableInfo)(pdp, + if (!__driFindDrawable(psp->drawHash, pdp->draw) || + ! (*dri_interface->getDrawableInfo)(pdp->display, pdp->screen, pdp->draw, &pdp->index, &pdp->lastStamp, &pdp->x, &pdp->y, &pdp->w, &pdp->h, &pdp->numClipRects, &pdp->pClipRects, &pdp->backX, &pdp->backY, &pdp->numBackClipRects, - &pdp->pBackClipRects, - pdp->loaderPrivate)) { + &pdp->pBackClipRects )) { /* Error -- eg the window may have been destroyed. Keep going * with no cliprects. */ @@ -304,138 +493,6 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp) } -int -__driParseEvents(__DRIcontextPrivate *pcp, __DRIdrawablePrivate *pdp) -{ - __DRIscreenPrivate *psp = pcp->driScreenPriv; - __DRIDrawableConfigEvent *dc, *last_dc; - __DRIBufferAttachEvent *ba, *last_ba; - unsigned int tail, mask, *p, end, total, size, changed; - unsigned char *data; - size_t rect_size; - - /* Check for wraparound. */ - if (psp->dri2.buffer->prealloc - pdp->dri2.tail > psp->dri2.buffer->size) { - /* If prealloc overlaps into what we just parsed, the - * server overwrote it and we have to reset our tail - * pointer. */ - DRM_UNLOCK(psp->fd, psp->lock, pcp->hHWContext); - (*psp->dri2.loader->reemitDrawableInfo)(pdp, &pdp->dri2.tail, - pdp->loaderPrivate); - DRM_LIGHT_LOCK(psp->fd, psp->lock, pcp->hHWContext); - } - - total = psp->dri2.buffer->head - pdp->dri2.tail; - mask = psp->dri2.buffer->size - 1; - end = psp->dri2.buffer->head; - data = psp->dri2.buffer->data; - - changed = 0; - last_dc = NULL; - last_ba = NULL; - - for (tail = pdp->dri2.tail; tail != end; tail += size) { - p = (unsigned int *) (data + (tail & mask)); - size = DRI2_EVENT_SIZE(*p); - if (size > total || (tail & mask) + size > psp->dri2.buffer->size) { - /* illegal data, bail out. */ - fprintf(stderr, "illegal event size\n"); - break; - } - - switch (DRI2_EVENT_TYPE(*p)) { - case DRI2_EVENT_DRAWABLE_CONFIG: - dc = (__DRIDrawableConfigEvent *) p; - if (dc->drawable == pdp->dri2.drawable_id) - last_dc = dc; - break; - - case DRI2_EVENT_BUFFER_ATTACH: - ba = (__DRIBufferAttachEvent *) p; - if (ba->drawable == pdp->dri2.drawable_id && - ba->buffer.attachment == DRI_DRAWABLE_BUFFER_FRONT_LEFT) - last_ba = ba; - break; - } - } - - if (last_dc) { - if (pdp->w != last_dc->width || pdp->h != last_dc->height) - changed = 1; - - pdp->x = last_dc->x; - pdp->y = last_dc->y; - pdp->w = last_dc->width; - pdp->h = last_dc->height; - - pdp->backX = 0; - pdp->backY = 0; - pdp->numBackClipRects = 1; - pdp->pBackClipRects[0].x1 = 0; - pdp->pBackClipRects[0].y1 = 0; - pdp->pBackClipRects[0].x2 = pdp->w; - pdp->pBackClipRects[0].y2 = pdp->h; - - pdp->numClipRects = last_dc->num_rects; - _mesa_free(pdp->pClipRects); - rect_size = last_dc->num_rects * sizeof last_dc->rects[0]; - pdp->pClipRects = _mesa_malloc(rect_size); - memcpy(pdp->pClipRects, last_dc->rects, rect_size); - } - - /* We only care about the most recent drawable config. */ - if (last_dc && changed) - (*psp->DriverAPI.HandleDrawableConfig)(pdp, pcp, last_dc); - - /* Front buffer attachments are special, they typically mean that - * we're rendering to a redirected window (or a child window of a - * redirected window) and that it got resized. Resizing the root - * window on randr events is a special case of this. Other causes - * may be a window transitioning between redirected and - * non-redirected, or a window getting reparented between parents - * with different window pixmaps (eg two redirected windows). - * These events are special in that the X server allocates the - * buffer and that the buffer may be shared by other child - * windows. When our window share the window pixmap with its - * parent, drawable config events doesn't affect the front buffer. - * We only care about the last such event in the buffer; in fact, - * older events will refer to invalid buffer objects.*/ - if (last_ba) - (*psp->DriverAPI.HandleBufferAttach)(pdp, pcp, last_ba); - - /* If there was a drawable config event in the buffer and it - * changed the size of the window, all buffer auxillary buffer - * attachments prior to that are invalid (as opposed to the front - * buffer case discussed above). In that case we can start - * looking for buffer attachment after the last drawable config - * event. If there is no drawable config event in this batch of - * events, we have to assume that the last batch might have had - * one and process all buffer attach events.*/ - if (last_dc && changed) - tail = (unsigned char *) last_dc - data; - else - tail = pdp->dri2.tail; - - for ( ; tail != end; tail += size) { - ba = (__DRIBufferAttachEvent *) (data + (tail & mask)); - size = DRI2_EVENT_SIZE(ba->event_header); - - if (DRI2_EVENT_TYPE(ba->event_header) != DRI2_EVENT_BUFFER_ATTACH) - continue; - if (ba->drawable != pdp->dri2.drawable_id) - continue; - if (last_ba == ba) - continue; - - (*psp->DriverAPI.HandleBufferAttach)(pdp, pcp, ba); - changed = 1; - } - - pdp->dri2.tail = tail; - - return changed || last_ba; -} - /*@}*/ /*****************************************************************/ @@ -443,33 +500,10 @@ __driParseEvents(__DRIcontextPrivate *pcp, __DRIdrawablePrivate *pdp) /*****************************************************************/ /*@{*/ -static void driReportDamage(__DRIdrawable *pdp, - struct drm_clip_rect *pClipRects, int numClipRects) -{ - __DRIscreen *psp = pdp->driScreenPriv; - - /* Check that we actually have the new damage report method */ - if (psp->dri2.enabled) { - (*psp->dri2.loader->postDamage)(pdp, - pClipRects, - numClipRects, - pdp->loaderPrivate); - } else if (psp->damage) { - /* Report the damage. Currently, all our drivers draw - * directly to the front buffer, so we report the damage there - * rather than to the backing storein (if any). - */ - (*psp->damage->reportDamage)(pdp, - pdp->x, pdp->y, - pClipRects, numClipRects, - GL_TRUE, pdp->loaderPrivate); - } -} - - /** * Swap buffers. * + * \param dpy the display handle. * \param drawablePrivate opaque pointer to the per-drawable private info. * * \internal @@ -477,28 +511,74 @@ static void driReportDamage(__DRIdrawable *pdp, * * Is called directly from glXSwapBuffers(). */ -static void driSwapBuffers(__DRIdrawable *dPriv) +static void driSwapBuffers( __DRInativeDisplay *dpy, void *drawablePrivate ) +{ + __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate; + drm_clip_rect_t rect; + + dPriv->swapBuffers(dPriv); + + /* Check that we actually have the new damage report method */ + if (api_ver < 20070105 || dri_interface->reportDamage == NULL) + return; + + /* Assume it's affecting the whole drawable for now */ + rect.x1 = 0; + rect.y1 = 0; + rect.x2 = rect.x1 + dPriv->w; + rect.y2 = rect.y1 + dPriv->h; + + /* Report the damage. Currently, all our drivers draw directly to the + * front buffer, so we report the damage there rather than to the backing + * store (if any). + */ + (*dri_interface->reportDamage)(dpy, dPriv->screen, dPriv->draw, + dPriv->x, dPriv->y, + &rect, 1, GL_TRUE); +} + +/** + * Called directly from a number of higher-level GLX functions. + */ +static int driGetMSC( void *screenPrivate, int64_t *msc ) +{ + __DRIscreenPrivate *sPriv = (__DRIscreenPrivate *) screenPrivate; + + return sPriv->DriverAPI.GetMSC( sPriv, msc ); +} + +/** + * Called directly from a number of higher-level GLX functions. + */ +static int driGetSBC( __DRInativeDisplay *dpy, void *drawablePrivate, int64_t *sbc ) { - __DRIscreen *psp = dPriv->driScreenPriv; + __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate; + __DRIswapInfo sInfo; + int status; - if (!dPriv->numClipRects) - return; - psp->DriverAPI.SwapBuffers(dPriv); + status = dPriv->driScreenPriv->DriverAPI.GetSwapInfo( dPriv, & sInfo ); + *sbc = sInfo.swap_count; - driReportDamage(dPriv, dPriv->pClipRects, dPriv->numClipRects); + return status; } -static int driDrawableGetMSC( __DRIscreen *sPriv, __DRIdrawable *dPriv, - int64_t *msc ) +static int driWaitForSBC( __DRInativeDisplay * dpy, void *drawablePriv, + int64_t target_sbc, + int64_t * msc, int64_t * sbc ) { - return sPriv->DriverAPI.GetDrawableMSC(sPriv, dPriv, msc); + __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePriv; + + return dPriv->driScreenPriv->DriverAPI.WaitForSBC( dPriv, target_sbc, + msc, sbc ); } -static int driWaitForMSC(__DRIdrawable *dPriv, int64_t target_msc, - int64_t divisor, int64_t remainder, - int64_t * msc, int64_t * sbc) +static int driWaitForMSC( __DRInativeDisplay * dpy, void *drawablePriv, + int64_t target_msc, + int64_t divisor, int64_t remainder, + int64_t * msc, int64_t * sbc ) { + __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePriv; __DRIswapInfo sInfo; int status; @@ -520,70 +600,63 @@ static int driWaitForMSC(__DRIdrawable *dPriv, int64_t target_msc, return status; } -const __DRImediaStreamCounterExtension driMediaStreamCounterExtension = { - { __DRI_MEDIA_STREAM_COUNTER, __DRI_MEDIA_STREAM_COUNTER_VERSION }, - driWaitForMSC, - driDrawableGetMSC, -}; - -static void driCopySubBuffer(__DRIdrawable *dPriv, - int x, int y, int w, int h) +static int64_t driSwapBuffersMSC( __DRInativeDisplay * dpy, void *drawablePriv, + int64_t target_msc, + int64_t divisor, int64_t remainder ) { - drm_clip_rect_t rect; + __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePriv; - dPriv->driScreenPriv->DriverAPI.CopySubBuffer(dPriv, x, y, w, h); - - rect.x1 = x; - rect.y1 = dPriv->h - y - h; - rect.x2 = x + w; - rect.y2 = rect.y1 + h; - driReportDamage(dPriv, &rect, 1); + return dPriv->driScreenPriv->DriverAPI.SwapBuffersMSC( dPriv, target_msc, + divisor, + remainder ); } -const __DRIcopySubBufferExtension driCopySubBufferExtension = { - { __DRI_COPY_SUB_BUFFER, __DRI_COPY_SUB_BUFFER_VERSION }, - driCopySubBuffer -}; - -static void driSetSwapInterval(__DRIdrawable *dPriv, unsigned int interval) -{ - dPriv->swap_interval = interval; -} - -static unsigned int driGetSwapInterval(__DRIdrawable *dPriv) +static void driCopySubBuffer( __DRInativeDisplay *dpy, void *drawablePrivate, + int x, int y, int w, int h) { - return dPriv->swap_interval; + __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate; + dPriv->driScreenPriv->DriverAPI.CopySubBuffer(dPriv, x, y, w, h); + (void) dpy; } -const __DRIswapControlExtension driSwapControlExtension = { - { __DRI_SWAP_CONTROL, __DRI_SWAP_CONTROL_VERSION }, - driSetSwapInterval, - driGetSwapInterval -}; - - /** * This is called via __DRIscreenRec's createNewDrawable pointer. */ -static __DRIdrawable * -driCreateNewDrawable(__DRIscreen *psp, const __DRIconfig *config, - drm_drawable_t hwDrawable, int renderType, - const int *attrs, void *data) +static void *driCreateNewDrawable(__DRInativeDisplay *dpy, + const __GLcontextModes *modes, + __DRIid draw, + __DRIdrawable *pdraw, + int renderType, + const int *attrs) { - __DRIdrawable *pdp; + __DRIscreen * const pDRIScreen = (*dri_interface->getScreen)(dpy, modes->screen); + __DRIscreenPrivate *psp; + __DRIdrawablePrivate *pdp; + + + pdraw->private = NULL; /* Since pbuffers are not yet supported, no drawable attributes are * supported either. */ (void) attrs; - pdp = _mesa_malloc(sizeof *pdp); + if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) { + return NULL; + } + + pdp = (__DRIdrawablePrivate *)_mesa_malloc(sizeof(__DRIdrawablePrivate)); if (!pdp) { return NULL; } - pdp->loaderPrivate = data; - pdp->hHWDrawable = hwDrawable; + if (!(*dri_interface->createDrawable)(dpy, modes->screen, draw, &pdp->hHWDrawable)) { + _mesa_free(pdp); + return NULL; + } + + pdp->draw = draw; + pdp->pdraw = pdraw; pdp->refcount = 0; pdp->pStamp = NULL; pdp->lastStamp = 0; @@ -596,55 +669,80 @@ driCreateNewDrawable(__DRIscreen *psp, const __DRIconfig *config, pdp->numBackClipRects = 0; pdp->pClipRects = NULL; pdp->pBackClipRects = NULL; - pdp->vblSeq = 0; - pdp->vblFlags = 0; + pdp->display = dpy; + pdp->screen = modes->screen; + psp = (__DRIscreenPrivate *)pDRIScreen->private; pdp->driScreenPriv = psp; pdp->driContextPriv = &psp->dummyContextPriv; - if (!(*psp->DriverAPI.CreateBuffer)(psp, pdp, &config->modes, + if (!(*psp->DriverAPI.CreateBuffer)(psp, pdp, modes, renderType == GLX_PIXMAP_BIT)) { + (void)(*dri_interface->destroyDrawable)(dpy, modes->screen, pdp->draw); _mesa_free(pdp); return NULL; } - pdp->msc_base = 0; + pdraw->private = pdp; + pdraw->destroyDrawable = driDestroyDrawable; + pdraw->swapBuffers = driSwapBuffers; /* called by glXSwapBuffers() */ + + pdraw->getSBC = driGetSBC; + pdraw->waitForSBC = driWaitForSBC; + pdraw->waitForMSC = driWaitForMSC; + pdraw->swapBuffersMSC = driSwapBuffersMSC; + pdraw->frameTracking = NULL; + pdraw->queryFrameTracking = driQueryFrameTracking; + + if (driCompareGLXAPIVersion (20060314) >= 0) + pdraw->copySubBuffer = driCopySubBuffer; /* This special default value is replaced with the configured * default value when the drawable is first bound to a direct * rendering context. */ - pdp->swap_interval = (unsigned)-1; + pdraw->swap_interval = (unsigned)-1; + + pdp->swapBuffers = psp->DriverAPI.SwapBuffers; - return pdp; + /* Add pdraw to drawable list */ + if (!__driAddDrawable(psp->drawHash, pdraw)) { + /* ERROR!!! */ + (*pdraw->destroyDrawable)(dpy, pdp); + _mesa_free(pdp); + pdp = NULL; + pdraw->private = NULL; + } + + return (void *) pdp; } static __DRIdrawable * -dri2CreateNewDrawable(__DRIscreen *screen, const __DRIconfig *config, - unsigned int drawable_id, unsigned int head, void *data) +driGetDrawable(__DRInativeDisplay *dpy, __DRIid draw, void *screenPrivate) { - __DRIdrawable *pdraw; - - pdraw = driCreateNewDrawable(screen, config, 0, 0, NULL, data); - if (!pdraw) - return NULL; + __DRIscreenPrivate *psp = (__DRIscreenPrivate *) screenPrivate; - pdraw->dri2.drawable_id = drawable_id; - pdraw->dri2.tail = head; - pdraw->pBackClipRects = _mesa_malloc(sizeof *pdraw->pBackClipRects); - - return pdraw; + /* + ** Make sure this routine returns NULL if the drawable is not bound + ** to a direct rendering context! + */ + return __driFindDrawable(psp->drawHash, draw); } - static void -driDestroyDrawable(__DRIdrawable *pdp) +driDestroyDrawable(__DRInativeDisplay *dpy, void *drawablePrivate) { + __DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *) drawablePrivate; __DRIscreenPrivate *psp; + int scrn; if (pdp) { psp = pdp->driScreenPriv; + scrn = psp->myNum; (*psp->DriverAPI.DestroyBuffer)(pdp); + if ((*dri_interface->windowExists)(dpy, pdp->draw)) + (void)(*dri_interface->destroyDrawable)(dpy, scrn, pdp->draw); + drmHashDelete(psp->drawHash, pdp->draw); if (pdp->pClipRects) { _mesa_free(pdp->pClipRects); pdp->pClipRects = NULL; @@ -668,6 +766,8 @@ driDestroyDrawable(__DRIdrawable *pdp) /** * Destroy the per-context private information. * + * \param dpy the display handle. + * \param scrn the screen number. * \param contextPrivate opaque pointer to the per-drawable private info. * * \internal @@ -675,10 +775,14 @@ driDestroyDrawable(__DRIdrawable *pdp) * drmDestroyContext(), and finally frees \p contextPrivate. */ static void -driDestroyContext(__DRIcontext *pcp) +driDestroyContext(__DRInativeDisplay *dpy, int scrn, void *contextPrivate) { + __DRIcontextPrivate *pcp = (__DRIcontextPrivate *) contextPrivate; + if (pcp) { (*pcp->driScreenPriv->DriverAPI.DestroyContext)(pcp); + __driGarbageCollectDrawables(pcp->driScreenPriv->drawHash); + (void) (*dri_interface->destroyContext)(dpy, scrn, pcp->contextID); _mesa_free(pcp); } } @@ -691,7 +795,7 @@ driDestroyContext(__DRIcontext *pcp) * \param modes Mode used to create the new context. * \param render_type Type of rendering target. \c GLX_RGBA is the only * type likely to ever be supported for direct-rendering. - * \param shared The shared context dependent methods or \c NULL if + * \param sharedPrivate The shared context dependent methods or \c NULL if * non-existent. * \param pctx DRI context to receive the context dependent methods. * @@ -705,18 +809,36 @@ driDestroyContext(__DRIcontext *pcp) * context. * */ -static __DRIcontext * -driCreateNewContext(__DRIscreen *psp, const __DRIconfig *config, - int render_type, __DRIcontext *shared, - drm_context_t hwContext, void *data) +static void * +driCreateNewContext(__DRInativeDisplay *dpy, const __GLcontextModes *modes, + int render_type, void *sharedPrivate, __DRIcontext *pctx) { - __DRIcontext *pcp; - void * const shareCtx = (shared != NULL) ? shared->driverPrivate : NULL; + __DRIscreen *pDRIScreen; + __DRIcontextPrivate *pcp; + __DRIcontextPrivate *pshare = (__DRIcontextPrivate *) sharedPrivate; + __DRIscreenPrivate *psp; + void * const shareCtx = (pshare != NULL) ? pshare->driverPrivate : NULL; - pcp = _mesa_malloc(sizeof *pcp); - if (!pcp) + pDRIScreen = (*dri_interface->getScreen)(dpy, modes->screen); + if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) { + /* ERROR!!! */ return NULL; + } + + psp = (__DRIscreenPrivate *)pDRIScreen->private; + pcp = (__DRIcontextPrivate *)_mesa_malloc(sizeof(__DRIcontextPrivate)); + if (!pcp) { + return NULL; + } + + if (! (*dri_interface->createContext)(dpy, modes->screen, modes->fbconfigID, + &pcp->contextID, &pcp->hHWContext)) { + _mesa_free(pcp); + return NULL; + } + + pcp->display = dpy; pcp->driScreenPriv = psp; pcp->driDrawablePriv = NULL; @@ -724,7 +846,8 @@ driCreateNewContext(__DRIscreen *psp, const __DRIconfig *config, * context. */ - if (!psp->dri2.enabled && !psp->dummyContextPriv.driScreenPriv) { + if (!psp->dummyContextPriv.driScreenPriv) { + psp->dummyContextPriv.contextID = 0; psp->dummyContextPriv.hHWContext = psp->pSAREA->dummy_context; psp->dummyContextPriv.driScreenPriv = psp; psp->dummyContextPriv.driDrawablePriv = NULL; @@ -732,40 +855,21 @@ driCreateNewContext(__DRIscreen *psp, const __DRIconfig *config, /* No other fields should be used! */ } - pcp->hHWContext = hwContext; + pctx->destroyContext = driDestroyContext; + pctx->bindContext = driBindContext; + pctx->unbindContext = driUnbindContext; - if ( !(*psp->DriverAPI.CreateContext)(&config->modes, pcp, shareCtx) ) { + if ( !(*psp->DriverAPI.CreateContext)(modes, pcp, shareCtx) ) { + (void) (*dri_interface->destroyContext)(dpy, modes->screen, + pcp->contextID); _mesa_free(pcp); return NULL; } - return pcp; -} - -static __DRIcontext * -dri2CreateNewContext(__DRIscreen *screen, const __DRIconfig *config, - __DRIcontext *shared, void *data) -{ - drm_context_t hwContext; - DRM_CAS_RESULT(ret); - - /* DRI2 doesn't use kernel with context IDs, we just need an ID that's - * different from the kernel context ID to make drmLock() happy. */ + __driGarbageCollectDrawables(pcp->driScreenPriv->drawHash); - do { - hwContext = screen->dri2.lock->next_id; - DRM_CAS(&screen->dri2.lock->next_id, hwContext, hwContext + 1, ret); - } while (ret); - - return driCreateNewContext(screen, config, 0, shared, hwContext, data); -} - -static int -driCopyContext(__DRIcontext *dest, __DRIcontext *src, unsigned long mask) -{ - return GL_FALSE; + return pcp; } - /*@}*/ @@ -785,8 +889,10 @@ driCopyContext(__DRIcontext *dest, __DRIcontext *src, unsigned long mask) * This function calls __DriverAPIRec::DestroyScreen on \p screenPrivate, calls * drmClose(), and finally frees \p screenPrivate. */ -static void driDestroyScreen(__DRIscreen *psp) +static void driDestroyScreen(__DRInativeDisplay *dpy, int scrn, void *screenPrivate) { + __DRIscreenPrivate *psp = (__DRIscreenPrivate *) screenPrivate; + if (psp) { /* No interaction with the X-server is possible at this point. This * routine is called after XCloseDisplay, so there is no protocol @@ -796,44 +902,26 @@ static void driDestroyScreen(__DRIscreen *psp) if (psp->DriverAPI.DestroyScreen) (*psp->DriverAPI.DestroyScreen)(psp); - if (psp->dri2.enabled) { - drmBOUnmap(psp->fd, &psp->dri2.sareaBO); - drmBOUnreference(psp->fd, &psp->dri2.sareaBO); - } else { - (void)drmUnmap((drmAddress)psp->pSAREA, SAREA_MAX); - (void)drmUnmap((drmAddress)psp->pFB, psp->fbSize); - (void)drmCloseOnce(psp->fd); + (void)drmUnmap((drmAddress)psp->pSAREA, SAREA_MAX); + (void)drmUnmap((drmAddress)psp->pFB, psp->fbSize); + _mesa_free(psp->pDevPriv); + (void)drmCloseOnce(psp->fd); + if ( psp->modes != NULL ) { + (*dri_interface->destroyContextModes)( psp->modes ); } + assert(psp->drawHash); + drmHashDestroy(psp->drawHash); + _mesa_free(psp); } } -static void -setupLoaderExtensions(__DRIscreen *psp, - const __DRIextension **extensions) -{ - int i; - - for (i = 0; extensions[i]; i++) { - if (strcmp(extensions[i]->name, __DRI_GET_DRAWABLE_INFO) == 0) - psp->getDrawableInfo = (__DRIgetDrawableInfoExtension *) extensions[i]; - if (strcmp(extensions[i]->name, __DRI_DAMAGE) == 0) - psp->damage = (__DRIdamageExtension *) extensions[i]; - if (strcmp(extensions[i]->name, __DRI_SYSTEM_TIME) == 0) - psp->systemTime = (__DRIsystemTimeExtension *) extensions[i]; - if (strcmp(extensions[i]->name, __DRI_LOADER) == 0) - psp->dri2.loader = (__DRIloaderExtension *) extensions[i]; - } -} /** - * This is the bootstrap function for the driver. libGL supplies all of the - * requisite information about the system, and the driver initializes itself. - * This routine also fills in the linked list pointed to by \c driver_modes - * with the \c __GLcontextModes that the driver can support for windows or - * pbuffers. + * Utility function used to create a new driver-private screen structure. * + * \param dpy Display pointer * \param scrn Index of the screen * \param psc DRI screen data (not driver private) * \param modes Linked list of known display modes. This list is, at a @@ -854,29 +942,44 @@ setupLoaderExtensions(__DRIscreen *psp, * driver and libGL. * \param driverAPI Driver API functions used by other routines in dri_util.c. * - * \note There is no need to check the minimum API version in this - * function. Since the name of this function is versioned, it is - * impossible for a loader that is too old to even load this driver. + * \note + * There is no need to check the minimum API version in this function. Since + * the \c __driCreateNewScreen function is versioned, it is impossible for a + * loader that is too old to even load this driver. */ -static __DRIscreen * -driCreateNewScreen(int scrn, - const __DRIversion *ddx_version, - const __DRIversion *dri_version, - const __DRIversion *drm_version, - const __DRIframebuffer *frame_buffer, - drmAddress pSAREA, int fd, - const __DRIextension **extensions, - const __DRIconfig ***driver_modes, - void *loaderPrivate) +__DRIscreenPrivate * +__driUtilCreateNewScreen(__DRInativeDisplay *dpy, int scrn, __DRIscreen *psc, + __GLcontextModes * modes, + const __DRIversion * ddx_version, + const __DRIversion * dri_version, + const __DRIversion * drm_version, + const __DRIframebuffer * frame_buffer, + drm_sarea_t *pSAREA, + int fd, + int internal_api_version, + const struct __DriverAPIRec *driverAPI) { - static const __DRIextension *emptyExtensionList[] = { NULL }; - __DRIscreen *psp; + __DRIscreenPrivate *psp; + + + api_ver = internal_api_version; + + psp = (__DRIscreenPrivate *)_mesa_malloc(sizeof(__DRIscreenPrivate)); + if (!psp) { + return NULL; + } - psp = _mesa_malloc(sizeof *psp); - if (!psp) + /* Create the hash table */ + psp->drawHash = drmHashCreate(); + if ( psp->drawHash == NULL ) { + _mesa_free( psp ); return NULL; + } - setupLoaderExtensions(psp, extensions); + psp->display = dpy; + psp->myNum = scrn; + psp->psc = psc; + psp->modes = modes; /* ** NOT_DONE: This is used by the X server to detect when the client @@ -885,12 +988,20 @@ driCreateNewScreen(int scrn, */ psp->drawLockID = 1; - psp->drm_version = *drm_version; - psp->ddx_version = *ddx_version; - psp->dri_version = *dri_version; + psp->drmMajor = drm_version->major; + psp->drmMinor = drm_version->minor; + psp->drmPatch = drm_version->patch; + psp->ddxMajor = ddx_version->major; + psp->ddxMinor = ddx_version->minor; + psp->ddxPatch = ddx_version->patch; + psp->driMajor = dri_version->major; + psp->driMinor = dri_version->minor; + psp->driPatch = dri_version->patch; + + /* install driver's callback functions */ + memcpy( &psp->DriverAPI, driverAPI, sizeof(struct __DriverAPIRec) ); psp->pSAREA = pSAREA; - psp->lock = (drmLock *) &psp->pSAREA->lock; psp->pFB = frame_buffer->base; psp->fbSize = frame_buffer->size; @@ -901,10 +1012,7 @@ driCreateNewScreen(int scrn, psp->pDevPriv = frame_buffer->dev_priv; psp->fbBPP = psp->fbStride * 8 / frame_buffer->width; - psp->extensions = emptyExtensionList; psp->fd = fd; - psp->myNum = scrn; - psp->dri2.enabled = GL_FALSE; /* ** Do not init dummy context here; actual initialization will be @@ -913,143 +1021,63 @@ driCreateNewScreen(int scrn, */ psp->dummyContextPriv.driScreenPriv = NULL; - psp->DriverAPI = driDriverAPI; - - *driver_modes = driDriverAPI.InitScreen(psp); - if (*driver_modes == NULL) { - _mesa_free(psp); - return NULL; - } - - return psp; -} - - -static __DRIscreen * -dri2CreateNewScreen(int scrn, int fd, unsigned int sarea_handle, - const __DRIextension **extensions, - const __DRIconfig ***driver_configs, void *data) -{ - static const __DRIextension *emptyExtensionList[] = { NULL }; - __DRIscreen *psp; - unsigned int *p; - drmVersionPtr version; - - if (driDriverAPI.InitScreen2 == NULL) - return NULL; - - psp = _mesa_malloc(sizeof(*psp)); - if (!psp) - return NULL; - - setupLoaderExtensions(psp, extensions); - - version = drmGetVersion(fd); - if (version) { - psp->drm_version.major = version->version_major; - psp->drm_version.minor = version->version_minor; - psp->drm_version.patch = version->version_patchlevel; - drmFreeVersion(version); - } - - psp->extensions = emptyExtensionList; - psp->fd = fd; - psp->myNum = scrn; - psp->dri2.enabled = GL_TRUE; - - if (drmBOReference(psp->fd, sarea_handle, &psp->dri2.sareaBO)) { - fprintf(stderr, "Failed to reference DRI2 sarea BO\n"); - _mesa_free(psp); - return NULL; - } - - if (drmBOMap(psp->fd, &psp->dri2.sareaBO, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, &psp->dri2.sarea)) { - drmBOUnreference(psp->fd, &psp->dri2.sareaBO); - _mesa_free(psp); - return NULL; - } - - p = psp->dri2.sarea; - while (DRI2_SAREA_BLOCK_TYPE(*p)) { - switch (DRI2_SAREA_BLOCK_TYPE(*p)) { - case DRI2_SAREA_BLOCK_LOCK: - psp->dri2.lock = (__DRILock *) p; - break; - case DRI2_SAREA_BLOCK_EVENT_BUFFER: - psp->dri2.buffer = (__DRIEventBuffer *) p; - break; - } - p = DRI2_SAREA_BLOCK_NEXT(p); - } + psc->destroyScreen = driDestroyScreen; + psc->createNewDrawable = driCreateNewDrawable; + psc->getDrawable = driGetDrawable; + psc->getMSC = driGetMSC; + psc->createNewContext = driCreateNewContext; - psp->lock = (drmLock *) &psp->dri2.lock->lock; + if (internal_api_version >= 20070121) + psc->setTexOffset = psp->DriverAPI.setTexOffset; - psp->DriverAPI = driDriverAPI; - *driver_configs = driDriverAPI.InitScreen2(psp); - if (*driver_configs == NULL) { - drmBOUnmap(psp->fd, &psp->dri2.sareaBO); - drmBOUnreference(psp->fd, &psp->dri2.sareaBO); - _mesa_free(psp); + if ( (psp->DriverAPI.InitDriver != NULL) + && !(*psp->DriverAPI.InitDriver)(psp) ) { + _mesa_free( psp ); return NULL; } - psp->DriverAPI = driDriverAPI; return psp; } -static const __DRIextension **driGetExtensions(__DRIscreen *psp) -{ - return psp->extensions; -} - -const __DRIlegacyExtension driLegacyExtension = { - { __DRI_LEGACY, __DRI_LEGACY_VERSION }, - driCreateNewScreen, - driCreateNewDrawable, - driCreateNewContext -}; - -const __DRIcoreExtension driCoreExtension = { - { __DRI_CORE, __DRI_CORE_VERSION }, - dri2CreateNewScreen, - driDestroyScreen, - driGetExtensions, - driGetConfigAttrib, - driIndexConfigAttrib, - dri2CreateNewDrawable, - driDestroyDrawable, - driSwapBuffers, - dri2CreateNewContext, - driCopyContext, - driDestroyContext, - driBindContext, - driUnbindContext -}; - -/* This is the table of extensions that the loader will dlsym() for. */ -PUBLIC const __DRIextension *__driDriverExtensions[] = { - &driCoreExtension.base, - &driLegacyExtension.base, - NULL -}; -static int -driFrameTracking(__DRIdrawable *drawable, GLboolean enable) +/** + * Compare the current GLX API version with a driver supplied required version. + * + * The minimum required version is compared with the API version exported by + * the \c __glXGetInternalVersion function (in libGL.so). + * + * \param required_version Minimum required internal GLX API version. + * \return A tri-value return, as from strcmp is returned. A value less + * than, equal to, or greater than zero will be returned if the + * internal GLX API version is less than, equal to, or greater + * than \c required_version. + * + * \sa __glXGetInternalVersion(). + */ +int driCompareGLXAPIVersion( GLint required_version ) { - return GLX_BAD_CONTEXT; + if ( api_ver > required_version ) { + return 1; + } + else if ( api_ver == required_version ) { + return 0; + } + + return -1; } + static int -driQueryFrameTracking(__DRIdrawable *dpriv, - int64_t * sbc, int64_t * missedFrames, - float * lastMissedUsage, float * usage) +driQueryFrameTracking( __DRInativeDisplay * dpy, void * priv, + int64_t * sbc, int64_t * missedFrames, + float * lastMissedUsage, float * usage ) { __DRIswapInfo sInfo; int status; int64_t ust; - __DRIscreenPrivate *psp = dpriv->driScreenPriv; + __DRIdrawablePrivate * dpriv = (__DRIdrawablePrivate *) priv; + status = dpriv->driScreenPriv->DriverAPI.GetSwapInfo( dpriv, & sInfo ); if ( status == 0 ) { @@ -1057,18 +1085,13 @@ driQueryFrameTracking(__DRIdrawable *dpriv, *missedFrames = sInfo.swap_missed_count; *lastMissedUsage = sInfo.swap_missed_usage; - (*psp->systemTime->getUST)( & ust ); + (*dri_interface->getUST)( & ust ); *usage = driCalculateSwapUsage( dpriv, sInfo.swap_ust, ust ); } return status; } -const __DRIframeTrackingExtension driFrameTrackingExtension = { - { __DRI_FRAME_TRACKING, __DRI_FRAME_TRACKING_VERSION }, - driFrameTracking, - driQueryFrameTracking -}; /** * Calculate amount of swap interval used between GLX buffer swaps. @@ -1106,10 +1129,11 @@ driCalculateSwapUsage( __DRIdrawablePrivate *dPriv, int64_t last_swap_ust, int32_t d; int interval; float usage = 1.0; - __DRIscreenPrivate *psp = dPriv->driScreenPriv; - if ( (*psp->systemTime->getMSCRate)(dPriv, &n, &d, dPriv->loaderPrivate) ) { - interval = (dPriv->swap_interval != 0) ? dPriv->swap_interval : 1; + + if ( (*dri_interface->getMSCRate)( dPriv->display, dPriv->draw, &n, &d ) ) { + interval = (dPriv->pdraw->swap_interval != 0) + ? dPriv->pdraw->swap_interval : 1; /* We want to calculate diff --git a/src/mesa/drivers/dri/common/dri_util.h b/src/mesa/drivers/dri/common/dri_util.h index f2bc456307..027cb7f461 100644 --- a/src/mesa/drivers/dri/common/dri_util.h +++ b/src/mesa/drivers/dri/common/dri_util.h @@ -1,24 +1,3 @@ -/** - * \file dri_util.h - * DRI utility functions definitions. - * - * This module acts as glue between GLX and the actual hardware driver. A DRI - * driver doesn't really \e have to use any of this - it's optional. But, some - * useful stuff is done here that otherwise would have to be duplicated in most - * drivers. - * - * Basically, these utility functions take care of some of the dirty details of - * screen initialization, context creation, context binding, DRM setup, etc. - * - * These functions are compiled into each DRI driver so libGL.so knows nothing - * about them. - * - * \sa dri_util.c. - * - * \author Kevin E. Martin - * \author Brian Paul - */ - /* * Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas. * All Rights Reserved. @@ -44,37 +23,46 @@ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ +/** + * \file dri_util.h + * DRI utility functions definitions. + * + * This module acts as glue between GLX and the actual hardware driver. A DRI + * driver doesn't really \e have to use any of this - it's optional. But, some + * useful stuff is done here that otherwise would have to be duplicated in most + * drivers. + * + * Basically, these utility functions take care of some of the dirty details of + * screen initialization, context creation, context binding, DRM setup, etc. + * + * These functions are compiled into each DRI driver so libGL.so knows nothing + * about them. + * + * \sa dri_util.c. + * + * \author Kevin E. Martin + * \author Brian Paul + */ #ifndef _DRI_UTIL_H_ #define _DRI_UTIL_H_ #include -#include -#include -#include +#include "drm.h" +#include "drm_sarea.h" +#include "xf86drm.h" #include "GL/internal/glcore.h" #include "GL/internal/dri_interface.h" -#include "GL/internal/dri_sarea.h" #define GLX_BAD_CONTEXT 5 +typedef struct __DRIdisplayPrivateRec __DRIdisplayPrivate; +typedef struct __DRIscreenPrivateRec __DRIscreenPrivate; +typedef struct __DRIcontextPrivateRec __DRIcontextPrivate; +typedef struct __DRIdrawablePrivateRec __DRIdrawablePrivate; typedef struct __DRIswapInfoRec __DRIswapInfo; +typedef struct __DRIutilversionRec2 __DRIutilversion2; -/* Typedefs to avoid rewriting the world. */ -typedef struct __DRIscreenRec __DRIscreenPrivate; -typedef struct __DRIdrawableRec __DRIdrawablePrivate; -typedef struct __DRIcontextRec __DRIcontextPrivate; - -/** - * Extensions. - */ -extern const __DRIlegacyExtension driLegacyExtension; -extern const __DRIcoreExtension driCoreExtension; -extern const __DRIextension driReadDrawableExtension; -extern const __DRIcopySubBufferExtension driCopySubBufferExtension; -extern const __DRIswapControlExtension driSwapControlExtension; -extern const __DRIframeTrackingExtension driFrameTrackingExtension; -extern const __DRImediaStreamCounterExtension driMediaStreamCounterExtension; /** * Used by DRI_VALIDATE_DRAWABLE_INFO @@ -90,7 +78,7 @@ extern const __DRImediaStreamCounterExtension driMediaStreamCounterExtension; /** * Utility macro to validate the drawable information. * - * See __DRIdrawable::pStamp and __DRIdrawable::lastStamp. + * See __DRIdrawablePrivate::pStamp and __DRIdrawablePrivate::lastStamp. */ #define DRI_VALIDATE_DRAWABLE_INFO(psp, pdp) \ do { \ @@ -119,103 +107,94 @@ do { \ * this structure. */ struct __DriverAPIRec { - const __DRIconfig **(*InitScreen) (__DRIscreen * priv); - + /** + * Driver initialization callback + */ + GLboolean (*InitDriver)(__DRIscreenPrivate *driScrnPriv); + /** * Screen destruction callback */ - void (*DestroyScreen)(__DRIscreen *driScrnPriv); + void (*DestroyScreen)(__DRIscreenPrivate *driScrnPriv); /** * Context creation callback */ GLboolean (*CreateContext)(const __GLcontextModes *glVis, - __DRIcontext *driContextPriv, + __DRIcontextPrivate *driContextPriv, void *sharedContextPrivate); /** * Context destruction callback */ - void (*DestroyContext)(__DRIcontext *driContextPriv); + void (*DestroyContext)(__DRIcontextPrivate *driContextPriv); /** * Buffer (drawable) creation callback */ - GLboolean (*CreateBuffer)(__DRIscreen *driScrnPriv, - __DRIdrawable *driDrawPriv, + GLboolean (*CreateBuffer)(__DRIscreenPrivate *driScrnPriv, + __DRIdrawablePrivate *driDrawPriv, const __GLcontextModes *glVis, GLboolean pixmapBuffer); /** * Buffer (drawable) destruction callback */ - void (*DestroyBuffer)(__DRIdrawable *driDrawPriv); + void (*DestroyBuffer)(__DRIdrawablePrivate *driDrawPriv); /** * Buffer swapping callback */ - void (*SwapBuffers)(__DRIdrawable *driDrawPriv); + void (*SwapBuffers)(__DRIdrawablePrivate *driDrawPriv); /** * Context activation callback */ - GLboolean (*MakeCurrent)(__DRIcontext *driContextPriv, - __DRIdrawable *driDrawPriv, - __DRIdrawable *driReadPriv); + GLboolean (*MakeCurrent)(__DRIcontextPrivate *driContextPriv, + __DRIdrawablePrivate *driDrawPriv, + __DRIdrawablePrivate *driReadPriv); /** * Context unbinding callback */ - GLboolean (*UnbindContext)(__DRIcontext *driContextPriv); + GLboolean (*UnbindContext)(__DRIcontextPrivate *driContextPriv); /** * Retrieves statistics about buffer swap operations. Required if * GLX_OML_sync_control or GLX_MESA_swap_frame_usage is supported. */ - int (*GetSwapInfo)( __DRIdrawable *dPriv, __DRIswapInfo * sInfo ); + int (*GetSwapInfo)( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo ); + + /** + * Required if GLX_SGI_video_sync or GLX_OML_sync_control is + * supported. + */ + int (*GetMSC)( __DRIscreenPrivate * priv, int64_t * count ); /** * These are required if GLX_OML_sync_control is supported. */ /*@{*/ - int (*WaitForMSC)( __DRIdrawable *priv, int64_t target_msc, + int (*WaitForMSC)( __DRIdrawablePrivate *priv, int64_t target_msc, int64_t divisor, int64_t remainder, int64_t * msc ); - int (*WaitForSBC)( __DRIdrawable *priv, int64_t target_sbc, + int (*WaitForSBC)( __DRIdrawablePrivate *priv, int64_t target_sbc, int64_t * msc, int64_t * sbc ); - int64_t (*SwapBuffersMSC)( __DRIdrawable *priv, int64_t target_msc, + int64_t (*SwapBuffersMSC)( __DRIdrawablePrivate *priv, int64_t target_msc, int64_t divisor, int64_t remainder ); /*@}*/ - void (*CopySubBuffer)(__DRIdrawable *driDrawPriv, + void (*CopySubBuffer)(__DRIdrawablePrivate *driDrawPriv, int x, int y, int w, int h); /** - * New version of GetMSC so we can pass drawable data to the low - * level DRM driver (e.g. pipe info). Required if - * GLX_SGI_video_sync or GLX_OML_sync_control is supported. + * See corresponding field in \c __DRIscreenRec. */ - int (*GetDrawableMSC) ( __DRIscreen * priv, - __DRIdrawable *drawablePrivate, - int64_t *count); - - - - /* DRI2 Entry points */ - const __DRIconfig **(*InitScreen2) (__DRIscreen * priv); - void (*HandleDrawableConfig)(__DRIdrawable *dPriv, - __DRIcontext *pcp, - __DRIDrawableConfigEvent *event); - - void (*HandleBufferAttach)(__DRIdrawable *dPriv, - __DRIcontext *pcp, - __DRIBufferAttachEvent *ba); - + void (*setTexOffset)(__DRIcontext *pDRICtx, GLint texname, + unsigned long long offset, GLint depth, GLuint pitch); }; -extern const struct __DriverAPIRec driDriverAPI; - struct __DRIswapInfoRec { /** @@ -251,7 +230,7 @@ struct __DRIswapInfoRec { /** * Per-drawable private DRI driver information. */ -struct __DRIdrawableRec { +struct __DRIdrawablePrivateRec { /** * Kernel drawable handle */ @@ -265,10 +244,10 @@ struct __DRIdrawableRec { void *driverPrivate; /** - * Private data from the loader. We just hold on to it and pass - * it back when calling into loader provided functions. + * X's drawable ID associated with this private drawable. */ - void *loaderPrivate; + __DRIid draw; + __DRIdrawable *pdraw; /** * Reference count for number of context's currently bound to this @@ -293,7 +272,7 @@ struct __DRIdrawableRec { /** * Last value of the stamp. * - * If this differs from the value stored at __DRIdrawable::pStamp, + * If this differs from the value stored at __DRIdrawablePrivate::pStamp, * then the drawable information has been modified by the X server, and the * drawable information (below) should be retrieved from the X server. */ @@ -327,56 +306,41 @@ struct __DRIdrawableRec { /*@}*/ /** - * \name Vertical blank tracking information - * Used for waiting on vertical blank events. - */ - /*@{*/ - unsigned int vblSeq; - unsigned int vblFlags; - /*@}*/ - - /** - * \name Monotonic MSC tracking - * - * Low level driver is responsible for updating msc_base and - * vblSeq values so that higher level code can calculate - * a new msc value or msc target for a WaitMSC call. The new value - * will be: - * msc = msc_base + get_vblank_count() - vblank_base; - * - * And for waiting on a value, core code will use: - * actual_target = target_msc - msc_base + vblank_base; + * Pointer to context to which this drawable is currently bound. */ - /*@{*/ - int64_t vblank_base; - int64_t msc_base; - /*@}*/ + __DRIcontextPrivate *driContextPriv; /** - * Pointer to context to which this drawable is currently bound. + * Pointer to screen on which this drawable was created. */ - __DRIcontext *driContextPriv; + __DRIscreenPrivate *driScreenPriv; /** - * Pointer to screen on which this drawable was created. + * \name Display and screen information. + * + * Basically just need these for when the locking code needs to call + * \c __driUtilUpdateDrawableInfo. */ - __DRIscreen *driScreenPriv; + /*@{*/ + __DRInativeDisplay *display; + int screen; + /*@}*/ /** - * Controls swap interval as used by GLX_SGI_swap_control and - * GLX_MESA_swap_control. + * Called via glXSwapBuffers(). */ - unsigned int swap_interval; - struct { - unsigned int tail; - unsigned int drawable_id; - } dri2; + void (*swapBuffers)( __DRIdrawablePrivate *dPriv ); }; /** * Per-context private driver information. */ -struct __DRIcontextRec { +struct __DRIcontextPrivateRec { + /** + * Kernel context handle used to access the device lock. + */ + __DRIid contextID; + /** * Kernel context handle used to access the device lock. */ @@ -388,30 +352,35 @@ struct __DRIcontextRec { void *driverPrivate; /** - * Pointer back to the \c __DRIcontext that contains this structure. + * This context's display pointer. */ - __DRIcontext *pctx; + __DRInativeDisplay *display; /** * Pointer to drawable currently bound to this context for drawing. */ - __DRIdrawable *driDrawablePriv; + __DRIdrawablePrivate *driDrawablePriv; /** * Pointer to drawable currently bound to this context for reading. */ - __DRIdrawable *driReadablePriv; + __DRIdrawablePrivate *driReadablePriv; /** * Pointer to screen on which this context was created. */ - __DRIscreen *driScreenPriv; + __DRIscreenPrivate *driScreenPriv; }; /** * Per-screen private driver information. */ -struct __DRIscreenRec { +struct __DRIscreenPrivateRec { + /** + * Display for this screen + */ + __DRInativeDisplay *display; + /** * Current screen's number */ @@ -422,21 +391,38 @@ struct __DRIscreenRec { */ struct __DriverAPIRec DriverAPI; - const __DRIextension **extensions; /** + * \name DDX version * DDX / 2D driver version information. + * \todo Replace these fields with a \c __DRIversionRec. */ - __DRIversion ddx_version; + /*@{*/ + int ddxMajor; + int ddxMinor; + int ddxPatch; + /*@}*/ /** + * \name DRI version * DRI X extension version information. + * \todo Replace these fields with a \c __DRIversionRec. */ - __DRIversion dri_version; + /*@{*/ + int driMajor; + int driMinor; + int driPatch; + /*@}*/ /** + * \name DRM version * DRM (kernel module) version information. + * \todo Replace these fields with a \c __DRIversionRec. */ - __DRIversion drm_version; + /*@{*/ + int drmMajor; + int drmMinor; + int drmPatch; + /*@}*/ /** * ID used when the client sets the drawable lock. @@ -499,7 +485,12 @@ struct __DRIscreenRec { * context is created when the first "real" context is created on this * screen. */ - __DRIcontext dummyContextPriv; + __DRIcontextPrivate dummyContextPriv; + + /** + * Hash table to hold the drawable information for this screen. + */ + void *drawHash; /** * Device-dependent private information (not stored in the SAREA). @@ -508,46 +499,66 @@ struct __DRIscreenRec { */ void *private; + /** + * GLX visuals / FBConfigs for this screen. These are stored as a + * linked list. + * + * \note + * This field is \b only used in conjunction with the old interfaces. If + * the new interfaces are used, this field will be set to \c NULL and will + * not be dereferenced. + */ + __GLcontextModes *modes; + /** * Pointer back to the \c __DRIscreen that contains this structure. */ + __DRIscreen *psc; +}; - /* Extensions provided by the loader. */ - const __DRIgetDrawableInfoExtension *getDrawableInfo; - const __DRIsystemTimeExtension *systemTime; - const __DRIdamageExtension *damage; - - struct { - /* Flag to indicate that this is a DRI2 screen. Many of the above - * fields will not be valid or initializaed in that case. */ - int enabled; - drmBO sareaBO; - void *sarea; - __DRIEventBuffer *buffer; - __DRILock *lock; - __DRIloaderExtension *loader; - } dri2; - - /* The lock actually in use, old sarea or DRI2 */ - drmLock *lock; + +/** + * Used to store a version which includes a major range instead of a single + * major version number. + */ +struct __DRIutilversionRec2 { + int major_min; /** min allowed Major version number. */ + int major_max; /** max allowed Major version number. */ + int minor; /**< Minor version number. */ + int patch; /**< Patch-level. */ }; + extern void __driUtilMessage(const char *f, ...); extern void -__driUtilUpdateDrawableInfo(__DRIdrawable *pdp); +__driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp); + + +extern __DRIscreenPrivate * __driUtilCreateNewScreen( __DRInativeDisplay *dpy, + int scrn, __DRIscreen *psc, __GLcontextModes * modes, + const __DRIversion * ddx_version, const __DRIversion * dri_version, + const __DRIversion * drm_version, const __DRIframebuffer * frame_buffer, + drm_sarea_t *pSAREA, int fd, int internal_api_version, + const struct __DriverAPIRec *driverAPI ); +/* Test the version of the internal GLX API. Returns a value like strcmp. */ extern int -__driParseEvents(__DRIcontext *psp, __DRIdrawable *pdp); +driCompareGLXAPIVersion( GLint required_version ); extern float -driCalculateSwapUsage( __DRIdrawable *dPriv, +driCalculateSwapUsage( __DRIdrawablePrivate *dPriv, int64_t last_swap_ust, int64_t current_ust ); -extern GLint -driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ); +/** + * Pointer to the \c __DRIinterfaceMethods passed to the driver by the loader. + * + * This pointer is set in the driver's \c __driCreateNewScreen function and + * is defined in dri_util.c. + */ +extern const __DRIinterfaceMethods * dri_interface; #endif /* _DRI_UTIL_H_ */ diff --git a/src/mesa/drivers/dri/common/drirenderbuffer.c b/src/mesa/drivers/dri/common/drirenderbuffer.c index d36af3e5be..d34da53479 100644 --- a/src/mesa/drivers/dri/common/drirenderbuffer.c +++ b/src/mesa/drivers/dri/common/drirenderbuffer.c @@ -209,6 +209,8 @@ driUpdateFramebufferSize(GLcontext *ctx, const __DRIdrawablePrivate *dPriv) struct gl_framebuffer *fb = (struct gl_framebuffer *) dPriv->driverPrivate; if (fb && (dPriv->w != fb->Width || dPriv->h != fb->Height)) { ctx->Driver.ResizeBuffers(ctx, fb, dPriv->w, dPriv->h); + /* if the driver needs the hw lock for ResizeBuffers, the drawable + might have changed again by now */ assert(fb->Width == dPriv->w); assert(fb->Height == dPriv->h); } diff --git a/src/mesa/drivers/dri/common/spantmp2.h b/src/mesa/drivers/dri/common/spantmp2.h index 53f5f846a0..50f3cf5581 100644 --- a/src/mesa/drivers/dri/common/spantmp2.h +++ b/src/mesa/drivers/dri/common/spantmp2.h @@ -114,7 +114,7 @@ do { \ GLuint p = *(volatile GLuint *) GET_PTR(_x, _y); \ __asm__ __volatile__( "bswap %0; rorl $8, %0" \ - : "=r" (p) : "0" (p) ); \ + : "=r" (p) : "r" (p) ); \ ((GLuint *)rgba)[0] = p; \ } while (0) # elif defined( MESA_BIG_ENDIAN ) diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c index 7fbe0d855d..6a189e7285 100644 --- a/src/mesa/drivers/dri/common/utils.c +++ b/src/mesa/drivers/dri/common/utils.c @@ -419,6 +419,21 @@ driCheckDriDdxDrmVersions2(const char * driver_name, drmActual, drmExpected); } + + +GLint +driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ) +{ + if (rect2.x1 > rect1.x1) rect1.x1 = rect2.x1; + if (rect2.x2 < rect1.x2) rect1.x2 = rect2.x2; + if (rect2.y1 > rect1.y1) rect1.y1 = rect2.y1; + if (rect2.y2 < rect1.y2) rect1.y2 = rect2.y2; + + if (rect1.x1 > rect1.x2 || rect1.y1 > rect1.y2) return 0; + + return (rect1.x2 - rect1.x1) * (rect1.y2 - rect1.y1); +} + GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, GLint *x, GLint *y, GLsizei *width, GLsizei *height ) @@ -452,6 +467,8 @@ GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, return GL_TRUE; } + + /** * Creates a set of \c __GLcontextModes that a driver will expose. * @@ -519,99 +536,86 @@ GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, * \c GL_UNSIGNED_3BYTE_8_8_8, \c GL_4FLOAT_32_32_32_32, * \c GL_4HALF_16_16_16_16, etc. We can cross that bridge when we come to it. */ -__DRIconfig ** -driCreateConfigs(GLenum fb_format, GLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, - unsigned num_depth_stencil_bits, - const GLenum * db_modes, unsigned num_db_modes) +GLboolean +driFillInModes( __GLcontextModes ** ptr_to_modes, + GLenum fb_format, GLenum fb_type, + const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + unsigned num_depth_stencil_bits, + const GLenum * db_modes, unsigned num_db_modes, + int visType ) { - static const u_int8_t bits_table[4][4] = { + static const u_int8_t bits_table[3][4] = { /* R G B A */ - { 3, 3, 2, 0 }, /* Any GL_UNSIGNED_BYTE_3_3_2 */ { 5, 6, 5, 0 }, /* Any GL_UNSIGNED_SHORT_5_6_5 */ { 8, 8, 8, 0 }, /* Any RGB with any GL_UNSIGNED_INT_8_8_8_8 */ { 8, 8, 8, 8 } /* Any RGBA with any GL_UNSIGNED_INT_8_8_8_8 */ }; - static const u_int32_t masks_table_rgb[6][4] = { - { 0x000000E0, 0x0000001C, 0x00000003, 0x00000000 }, /* 3_3_2 */ - { 0x00000007, 0x00000038, 0x000000C0, 0x00000000 }, /* 2_3_3_REV */ + /* The following arrays are all indexed by the fb_type masked with 0x07. + * Given the four supported fb_type values, this results in valid array + * indices of 3, 4, 5, and 7. + */ + static const u_int32_t masks_table_rgb[8][4] = { + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 }, /* 5_6_5 */ { 0x0000001F, 0x000007E0, 0x0000F800, 0x00000000 }, /* 5_6_5_REV */ { 0xFF000000, 0x00FF0000, 0x0000FF00, 0x00000000 }, /* 8_8_8_8 */ + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x000000FF, 0x0000FF00, 0x00FF0000, 0x00000000 } /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_rgba[6][4] = { - { 0x000000E0, 0x0000001C, 0x00000003, 0x00000000 }, /* 3_3_2 */ - { 0x00000007, 0x00000038, 0x000000C0, 0x00000000 }, /* 2_3_3_REV */ + static const u_int32_t masks_table_rgba[8][4] = { + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 }, /* 5_6_5 */ { 0x0000001F, 0x000007E0, 0x0000F800, 0x00000000 }, /* 5_6_5_REV */ { 0xFF000000, 0x00FF0000, 0x0000FF00, 0x000000FF }, /* 8_8_8_8 */ + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x000000FF, 0x0000FF00, 0x00FF0000, 0xFF000000 }, /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_bgr[6][4] = { - { 0x00000007, 0x00000038, 0x000000C0, 0x00000000 }, /* 3_3_2 */ - { 0x000000E0, 0x0000001C, 0x00000003, 0x00000000 }, /* 2_3_3_REV */ + static const u_int32_t masks_table_bgr[8][4] = { + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x0000001F, 0x000007E0, 0x0000F800, 0x00000000 }, /* 5_6_5 */ { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 }, /* 5_6_5_REV */ { 0x0000FF00, 0x00FF0000, 0xFF000000, 0x00000000 }, /* 8_8_8_8 */ + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00FF0000, 0x0000FF00, 0x000000FF, 0x00000000 }, /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_bgra[6][4] = { - { 0x00000007, 0x00000038, 0x000000C0, 0x00000000 }, /* 3_3_2 */ - { 0x000000E0, 0x0000001C, 0x00000003, 0x00000000 }, /* 2_3_3_REV */ + static const u_int32_t masks_table_bgra[8][4] = { + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x0000001F, 0x000007E0, 0x0000F800, 0x00000000 }, /* 5_6_5 */ { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 }, /* 5_6_5_REV */ { 0x0000FF00, 0x00FF0000, 0xFF000000, 0x000000FF }, /* 8_8_8_8 */ + { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000 }, /* 8_8_8_8_REV */ }; - static const u_int8_t bytes_per_pixel[6] = { - 1, /* 3_3_2 */ - 1, /* 2_3_3_REV */ - 2, /* 5_6_5 */ - 2, /* 5_6_5_REV */ - 4, /* 8_8_8_8 */ - 4 /* 8_8_8_8_REV */ + static const u_int8_t bytes_per_pixel[8] = { + 0, 0, 0, 2, 2, 4, 0, 4 }; const u_int8_t * bits; const u_int32_t * masks; - int index; - __DRIconfig **configs, **c; - __GLcontextModes *modes; + const int index = fb_type & 0x07; + __GLcontextModes * modes = *ptr_to_modes; unsigned i; unsigned j; unsigned k; - unsigned num_modes; - unsigned num_accum_bits = 2; - - switch ( fb_type ) { - case GL_UNSIGNED_BYTE_3_3_2: - index = 0; - break; - case GL_UNSIGNED_BYTE_2_3_3_REV: - index = 1; - break; - case GL_UNSIGNED_SHORT_5_6_5: - index = 2; - break; - case GL_UNSIGNED_SHORT_5_6_5_REV: - index = 3; - break; - case GL_UNSIGNED_INT_8_8_8_8: - index = 4; - break; - case GL_UNSIGNED_INT_8_8_8_8_REV: - index = 5; - break; - default: - fprintf( stderr, "[%s:%u] Unknown framebuffer type 0x%04x.\n", - __FUNCTION__, __LINE__, fb_type ); - return NULL; + + + if ( bytes_per_pixel[ index ] == 0 ) { + fprintf( stderr, "[%s:%u] Framebuffer type 0x%04x has 0 bytes per pixel.\n", + __FUNCTION__, __LINE__, fb_type ); + return GL_FALSE; } @@ -623,55 +627,40 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type, switch ( fb_format ) { case GL_RGB: + bits = (bytes_per_pixel[ index ] == 2) + ? bits_table[0] : bits_table[1]; masks = masks_table_rgb[ index ]; break; case GL_RGBA: + bits = (bytes_per_pixel[ index ] == 2) + ? bits_table[0] : bits_table[2]; masks = masks_table_rgba[ index ]; break; case GL_BGR: + bits = (bytes_per_pixel[ index ] == 2) + ? bits_table[0] : bits_table[1]; masks = masks_table_bgr[ index ]; break; case GL_BGRA: + bits = (bytes_per_pixel[ index ] == 2) + ? bits_table[0] : bits_table[2]; masks = masks_table_bgra[ index ]; break; default: - fprintf( stderr, "[%s:%u] Unknown framebuffer format 0x%04x.\n", - __FUNCTION__, __LINE__, fb_format ); - return NULL; - } - - switch ( bytes_per_pixel[ index ] ) { - case 1: - bits = bits_table[0]; - break; - case 2: - bits = bits_table[1]; - break; - default: - bits = ((fb_format == GL_RGB) || (fb_format == GL_BGR)) - ? bits_table[2] - : bits_table[3]; - break; + fprintf( stderr, "[%s:%u] Framebuffer format 0x%04x is not GL_RGB, GL_RGBA, GL_BGR, or GL_BGRA.\n", + __FUNCTION__, __LINE__, fb_format ); + return GL_FALSE; } - num_modes = num_depth_stencil_bits * num_db_modes * num_accum_bits; - configs = _mesa_calloc((num_modes + 1) * sizeof *configs); - if (configs == NULL) - return NULL; - c = configs; for ( k = 0 ; k < num_depth_stencil_bits ; k++ ) { for ( i = 0 ; i < num_db_modes ; i++ ) { - for ( j = 0 ; j < num_accum_bits ; j++ ) { - *c = _mesa_malloc (sizeof **c); - modes = &(*c)->modes; - c++; + for ( j = 0 ; j < 2 ; j++ ) { - memset(modes, 0, sizeof *modes); modes->redBits = bits[0]; modes->greenBits = bits[1]; modes->blueBits = bits[2]; @@ -692,13 +681,7 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type, modes->stencilBits = stencil_bits[k]; modes->depthBits = depth_bits[k]; - modes->transparentPixel = GLX_NONE; - modes->transparentRed = GLX_DONT_CARE; - modes->transparentGreen = GLX_DONT_CARE; - modes->transparentBlue = GLX_DONT_CARE; - modes->transparentAlpha = GLX_DONT_CARE; - modes->transparentIndex = GLX_DONT_CARE; - modes->visualType = GLX_DONT_CARE; + modes->visualType = visType; modes->renderType = GLX_RGBA_BIT; modes->drawableType = GLX_WINDOW_BIT; modes->rgbMode = GL_TRUE; @@ -718,155 +701,11 @@ driCreateConfigs(GLenum fb_format, GLenum fb_type, modes->haveDepthBuffer = (modes->depthBits > 0); modes->haveStencilBuffer = (modes->stencilBits > 0); - modes->bindToTextureRgb = GL_TRUE; - modes->bindToTextureRgba = GL_TRUE; - modes->bindToMipmapTexture = GL_FALSE; - modes->bindToTextureTargets = modes->rgbMode ? - __DRI_ATTRIB_TEXTURE_1D_BIT | - __DRI_ATTRIB_TEXTURE_2D_BIT | - __DRI_ATTRIB_TEXTURE_RECTANGLE_BIT : - 0; + modes = modes->next; } } } - *c = NULL; - - return configs; -} - -const __DRIconfig **driConcatConfigs(__DRIconfig **a, __DRIconfig **b) -{ - const __DRIconfig **all; - int i, j, index; - - i = 0; - while (a[i] != NULL) - i++; - j = 0; - while (b[j] != NULL) - j++; - - all = _mesa_malloc((i + j + 1) * sizeof *all); - index = 0; - for (i = 0; a[i] != NULL; i++) - all[index++] = a[i]; - for (j = 0; b[j] != NULL; j++) - all[index++] = b[j]; - all[index++] = NULL; - - _mesa_free(a); - _mesa_free(b); - - return all; -} - -#define __ATTRIB(attrib, field) \ - { attrib, offsetof(__GLcontextModes, field) } - -static const struct { unsigned int attrib, offset; } attribMap[] = { - __ATTRIB(__DRI_ATTRIB_BUFFER_SIZE, rgbBits), - __ATTRIB(__DRI_ATTRIB_LEVEL, level), - __ATTRIB(__DRI_ATTRIB_RED_SIZE, redBits), - __ATTRIB(__DRI_ATTRIB_GREEN_SIZE, greenBits), - __ATTRIB(__DRI_ATTRIB_BLUE_SIZE, blueBits), - __ATTRIB(__DRI_ATTRIB_ALPHA_SIZE, alphaBits), - __ATTRIB(__DRI_ATTRIB_DEPTH_SIZE, depthBits), - __ATTRIB(__DRI_ATTRIB_STENCIL_SIZE, stencilBits), - __ATTRIB(__DRI_ATTRIB_ACCUM_RED_SIZE, accumRedBits), - __ATTRIB(__DRI_ATTRIB_ACCUM_GREEN_SIZE, accumGreenBits), - __ATTRIB(__DRI_ATTRIB_ACCUM_BLUE_SIZE, accumBlueBits), - __ATTRIB(__DRI_ATTRIB_ACCUM_ALPHA_SIZE, accumAlphaBits), - __ATTRIB(__DRI_ATTRIB_SAMPLE_BUFFERS, sampleBuffers), - __ATTRIB(__DRI_ATTRIB_SAMPLES, samples), - __ATTRIB(__DRI_ATTRIB_DOUBLE_BUFFER, doubleBufferMode), - __ATTRIB(__DRI_ATTRIB_STEREO, stereoMode), - __ATTRIB(__DRI_ATTRIB_AUX_BUFFERS, numAuxBuffers), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_TYPE, transparentPixel), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_INDEX_VALUE, transparentPixel), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_RED_VALUE, transparentRed), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_GREEN_VALUE, transparentGreen), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_BLUE_VALUE, transparentBlue), - __ATTRIB(__DRI_ATTRIB_TRANSPARENT_ALPHA_VALUE, transparentAlpha), - __ATTRIB(__DRI_ATTRIB_FLOAT_MODE, floatMode), - __ATTRIB(__DRI_ATTRIB_RED_MASK, redMask), - __ATTRIB(__DRI_ATTRIB_GREEN_MASK, greenMask), - __ATTRIB(__DRI_ATTRIB_BLUE_MASK, blueMask), - __ATTRIB(__DRI_ATTRIB_ALPHA_MASK, alphaMask), - __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_WIDTH, maxPbufferWidth), - __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_HEIGHT, maxPbufferHeight), - __ATTRIB(__DRI_ATTRIB_MAX_PBUFFER_PIXELS, maxPbufferPixels), - __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_WIDTH, optimalPbufferWidth), - __ATTRIB(__DRI_ATTRIB_OPTIMAL_PBUFFER_HEIGHT, optimalPbufferHeight), - __ATTRIB(__DRI_ATTRIB_SWAP_METHOD, swapMethod), - __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGB, bindToTextureRgb), - __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_RGBA, bindToTextureRgba), - __ATTRIB(__DRI_ATTRIB_BIND_TO_MIPMAP_TEXTURE, bindToMipmapTexture), - __ATTRIB(__DRI_ATTRIB_BIND_TO_TEXTURE_TARGETS, bindToTextureTargets), - __ATTRIB(__DRI_ATTRIB_YINVERTED, yInverted), - - /* The struct field doesn't matter here, these are handled by the - * switch in driGetConfigAttribIndex. We need them in the array - * so the iterator includes them though.*/ - __ATTRIB(__DRI_ATTRIB_RENDER_TYPE, level), - __ATTRIB(__DRI_ATTRIB_CONFIG_CAVEAT, level), - __ATTRIB(__DRI_ATTRIB_SWAP_METHOD, level) -}; - -#define ARRAY_SIZE(a) (sizeof (a) / sizeof ((a)[0])) - -static int -driGetConfigAttribIndex(const __DRIconfig *config, - unsigned int index, unsigned int *value) -{ - switch (attribMap[index].attrib) { - case __DRI_ATTRIB_RENDER_TYPE: - if (config->modes.rgbMode) - *value = __DRI_ATTRIB_RGBA_BIT; - else - *value = __DRI_ATTRIB_COLOR_INDEX_BIT; - break; - case __DRI_ATTRIB_CONFIG_CAVEAT: - if (config->modes.visualRating == GLX_NON_CONFORMANT_CONFIG) - *value = __DRI_ATTRIB_NON_CONFORMANT_CONFIG; - else if (config->modes.visualRating == GLX_SLOW_CONFIG) - *value = __DRI_ATTRIB_SLOW_BIT; - else - *value = 0; - break; - case __DRI_ATTRIB_SWAP_METHOD: - break; - - default: - *value = *(unsigned int *) - ((char *) &config->modes + attribMap[index].offset); - - break; - } + *ptr_to_modes = modes; return GL_TRUE; } - -int -driGetConfigAttrib(const __DRIconfig *config, - unsigned int attrib, unsigned int *value) -{ - int i; - - for (i = 0; i < ARRAY_SIZE(attribMap); i++) - if (attribMap[i].attrib == attrib) - return driGetConfigAttribIndex(config, i, value); - - return GL_FALSE; -} - -int -driIndexConfigAttrib(const __DRIconfig *config, int index, - unsigned int *attrib, unsigned int *value) -{ - if (index >= 0 && index < ARRAY_SIZE(attribMap)) { - *attrib = attribMap[index].attrib; - return driGetConfigAttribIndex(config, index, value); - } - - return GL_FALSE; -} diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h index 9ac3b51447..b28b895627 100644 --- a/src/mesa/drivers/dri/common/utils.h +++ b/src/mesa/drivers/dri/common/utils.h @@ -28,11 +28,8 @@ #ifndef DRI_DEBUG_H #define DRI_DEBUG_H -#include -#include #include "context.h" - -typedef struct __DRIutilversionRec2 __DRIutilversion2; +#include "dri_util.h" struct dri_debug_control { const char * string; @@ -86,17 +83,6 @@ struct dri_extension { const struct dri_extension_function * functions; }; -/** - * Used to store a version which includes a major range instead of a single - * major version number. - */ -struct __DRIutilversionRec2 { - int major_min; /** min allowed Major version number. */ - int major_max; /** max allowed Major version number. */ - int minor; /**< Minor version number. */ - int patch; /**< Patch-level. */ -}; - extern unsigned driParseDebugString( const char * debug, const struct dri_debug_control * control ); @@ -119,27 +105,16 @@ extern GLboolean driCheckDriDdxDrmVersions3(const char * driver_name, const __DRIversion * ddxActual, const __DRIutilversion2 * ddxExpected, const __DRIversion * drmActual, const __DRIversion * drmExpected); +extern GLint driIntersectArea( drm_clip_rect_t rect1, drm_clip_rect_t rect2 ); + extern GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, GLint *x, GLint *y, GLsizei *width, GLsizei *height ); -struct __DRIconfigRec { - __GLcontextModes modes; -}; - -extern __DRIconfig ** -driCreateConfigs(GLenum fb_format, GLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, - unsigned num_depth_stencil_bits, - const GLenum * db_modes, unsigned num_db_modes); - -const __DRIconfig **driConcatConfigs(__DRIconfig **a, __DRIconfig **b); - -int -driGetConfigAttrib(const __DRIconfig *config, - unsigned int attrib, unsigned int *value); -int -driIndexConfigAttrib(const __DRIconfig *config, int index, - unsigned int *attrib, unsigned int *value); +extern GLboolean driFillInModes( __GLcontextModes ** modes, + GLenum fb_format, GLenum fb_type, + const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + unsigned num_depth_stencil_bits, + const GLenum * db_modes, unsigned num_db_modes, int visType ); #endif /* DRI_DEBUG_H */ diff --git a/src/mesa/drivers/dri/common/vblank.c b/src/mesa/drivers/dri/common/vblank.c index 0008ab1c34..094950d362 100644 --- a/src/mesa/drivers/dri/common/vblank.c +++ b/src/mesa/drivers/dri/common/vblank.c @@ -34,16 +34,6 @@ #include "vblank.h" #include "xmlpool.h" -static unsigned int msc_to_vblank(__DRIdrawablePrivate * dPriv, int64_t msc) -{ - return (unsigned int)(msc - dPriv->msc_base + dPriv->vblank_base); -} - -static int64_t vblank_to_msc(__DRIdrawablePrivate * dPriv, unsigned int vblank) -{ - return (int64_t)(vblank - dPriv->vblank_base + dPriv->msc_base); -} - /****************************************************************************/ /** @@ -51,7 +41,7 @@ static int64_t vblank_to_msc(__DRIdrawablePrivate * dPriv, unsigned int vblank) * * Stores the 64-bit count of vertical refreshes since some (arbitrary) * point in time in \c count. Unless the value wraps around, which it - * may, it will never decrease for a given drawable. + * may, it will never decrease. * * \warning This function is called from \c glXGetVideoSyncSGI, which expects * a \c count of type \c unsigned (32-bit), and \c glXGetSyncValuesOML, which @@ -59,14 +49,11 @@ static int64_t vblank_to_msc(__DRIdrawablePrivate * dPriv, unsigned int vblank) * currently always returns a \c sequence of type \c unsigned. * * \param priv Pointer to the DRI screen private struct. - * \param dPriv Pointer to the DRI drawable private struct * \param count Storage to hold MSC counter. * \return Zero is returned on success. A negative errno value * is returned on failure. */ -int driDrawableGetMSC32( __DRIscreenPrivate * priv, - __DRIdrawablePrivate * dPriv, - int64_t * count) +int driGetMSC32( __DRIscreenPrivate * priv, int64_t * count ) { drmVBlank vbl; int ret; @@ -75,21 +62,14 @@ int driDrawableGetMSC32( __DRIscreenPrivate * priv, vbl.request.type = DRM_VBLANK_RELATIVE; vbl.request.sequence = 0; - if ( dPriv && dPriv->vblFlags & VBLANK_FLAG_SECONDARY ) - vbl.request.type |= DRM_VBLANK_SECONDARY; ret = drmWaitVBlank( priv->fd, &vbl ); - - if (dPriv) { - *count = vblank_to_msc(dPriv, vbl.reply.sequence); - } else { - /* Old driver (no knowledge of drawable MSC callback) */ - *count = vbl.reply.sequence; - } + *count = (int64_t)vbl.reply.sequence; return ret; } + /****************************************************************************/ /** * Wait for a specified refresh count. This implements most of the @@ -142,9 +122,7 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, */ vbl.request.type = dont_wait ? DRM_VBLANK_RELATIVE : DRM_VBLANK_ABSOLUTE; - vbl.request.sequence = next ? msc_to_vblank(priv, next) : 0; - if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) - vbl.request.type |= DRM_VBLANK_SECONDARY; + vbl.request.sequence = next; if ( drmWaitVBlank( priv->driScreenPriv->fd, &vbl ) != 0 ) { /* FIXME: This doesn't seem like the right thing to return here. @@ -152,10 +130,8 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, return GLX_BAD_CONTEXT; } - *msc = vblank_to_msc(priv, vbl.reply.sequence); - dont_wait = 0; - if (target_msc != 0 && *msc == target) + if (target_msc != 0 && vbl.reply.sequence == target) break; /* Assuming the wait-done test fails, the next refresh to wait for @@ -165,9 +141,9 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, * If this refresh has already happened, we add divisor to obtain * the next refresh after the current one that will satisfy it. */ - r = (*msc % (unsigned int)divisor); - next = (*msc - r + (unsigned int)remainder); - if (next <= *msc) next += (unsigned int)divisor; + r = (vbl.reply.sequence % (unsigned int)divisor); + next = (vbl.reply.sequence - r + (unsigned int)remainder); + if (next <= vbl.reply.sequence) next += (unsigned int)divisor; } while ( r != (unsigned int)remainder ); } @@ -177,10 +153,7 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, */ vbl.request.type = DRM_VBLANK_ABSOLUTE; - vbl.request.sequence = target_msc ? msc_to_vblank(priv, target_msc) : 0; - - if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) - vbl.request.type |= DRM_VBLANK_SECONDARY; + vbl.request.sequence = target_msc; if ( drmWaitVBlank( priv->driScreenPriv->fd, &vbl ) != 0 ) { /* FIXME: This doesn't seem like the right thing to return here. @@ -189,8 +162,8 @@ int driWaitForMSC32( __DRIdrawablePrivate *priv, } } - *msc = vblank_to_msc(priv, vbl.reply.sequence); - + *msc = (target_msc & 0xffffffff00000000LL); + *msc |= vbl.reply.sequence; if ( *msc < target_msc ) { *msc += 0x0000000100000000LL; } @@ -259,8 +232,8 @@ static int do_wait( drmVBlank * vbl, GLuint * vbl_seq, int fd ) if ( first_time ) { fprintf(stderr, "%s: drmWaitVBlank returned %d, IRQs don't seem to be" - " working correctly.\nTry adjusting the vblank_mode" - " configuration parameter.\n", __FUNCTION__, ret); + " working correctly.\nTry running with LIBGL_THROTTLE_REFRESH" + " and LIBL_SYNC_REFRESH unset.\n", __FUNCTION__, ret); first_time = GL_FALSE; } @@ -272,44 +245,22 @@ static int do_wait( drmVBlank * vbl, GLuint * vbl_seq, int fd ) } -/****************************************************************************/ -/** - * Returns the default swap interval of the given drawable. - */ - -static unsigned -driGetDefaultVBlankInterval( const __DRIdrawablePrivate *priv ) -{ - if ( (priv->vblFlags & (VBLANK_FLAG_THROTTLE | VBLANK_FLAG_SYNC)) != 0 ) { - return 1; - } - else { - return 0; - } -} - - /****************************************************************************/ /** * Sets the default swap interval when the drawable is first bound to a * direct rendering context. */ -void driDrawableInitVBlank( __DRIdrawablePrivate *priv ) +void driDrawableInitVBlank( __DRIdrawablePrivate *priv, GLuint flags, + GLuint *vbl_seq ) { - if ( priv->swap_interval == (unsigned)-1 && - !( priv->vblFlags & VBLANK_FLAG_NO_IRQ ) ) { + if ( priv->pdraw->swap_interval == (unsigned)-1 ) { /* Get current vertical blank sequence */ - drmVBlank vbl; - - vbl.request.type = DRM_VBLANK_RELATIVE; - if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) - vbl.request.type |= DRM_VBLANK_SECONDARY; - vbl.request.sequence = 0; - do_wait( &vbl, &priv->vblSeq, priv->driScreenPriv->fd ); - priv->vblank_base = priv->vblSeq; - - priv->swap_interval = driGetDefaultVBlankInterval( priv ); + drmVBlank vbl = { .request={ .type = DRM_VBLANK_RELATIVE, .sequence = 0 } }; + do_wait( &vbl, vbl_seq, priv->driScreenPriv->fd ); + + priv->pdraw->swap_interval = (flags & (VBLANK_FLAG_THROTTLE | + VBLANK_FLAG_SYNC)) != 0 ? 1 : 0; } } @@ -320,17 +271,21 @@ void driDrawableInitVBlank( __DRIdrawablePrivate *priv ) */ unsigned -driGetVBlankInterval( const __DRIdrawablePrivate *priv ) +driGetVBlankInterval( const __DRIdrawablePrivate *priv, GLuint flags ) { - if ( (priv->vblFlags & VBLANK_FLAG_INTERVAL) != 0 ) { + if ( (flags & VBLANK_FLAG_INTERVAL) != 0 ) { /* this must have been initialized when the drawable was first bound * to a direct rendering context. */ - assert ( priv->swap_interval != (unsigned)-1 ); + assert ( priv->pdraw->swap_interval != (unsigned)-1 ); - return priv->swap_interval; + return priv->pdraw->swap_interval; + } + else if ( (flags & (VBLANK_FLAG_THROTTLE | VBLANK_FLAG_SYNC)) != 0 ) { + return 1; + } + else { + return 0; } - else - return driGetDefaultVBlankInterval( priv ); } @@ -340,17 +295,18 @@ driGetVBlankInterval( const __DRIdrawablePrivate *priv ) */ void -driGetCurrentVBlank( __DRIdrawablePrivate *priv ) +driGetCurrentVBlank( const __DRIdrawablePrivate *priv, GLuint flags, + GLuint *vbl_seq ) { drmVBlank vbl; vbl.request.type = DRM_VBLANK_RELATIVE; - if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) { + if ( flags & VBLANK_FLAG_SECONDARY ) { vbl.request.type |= DRM_VBLANK_SECONDARY; } vbl.request.sequence = 0; - (void) do_wait( &vbl, &priv->vblSeq, priv->driScreenPriv->fd ); + (void) do_wait( &vbl, vbl_seq, priv->driScreenPriv->fd ); } @@ -358,15 +314,19 @@ driGetCurrentVBlank( __DRIdrawablePrivate *priv ) /** * Waits for the vertical blank for use with glXSwapBuffers. * + * \param vbl_seq Vertical blank sequence number (MSC) after the last buffer + * swap. Updated after this wait. + * \param flags \c VBLANK_FLAG bits that control how long to wait. * \param missed_deadline Set to \c GL_TRUE if the MSC after waiting is later - * than the "target" based on \c priv->vblFlags. The idea is - * that if \c missed_deadline is set, then the application is - * not achieving its desired framerate. + * than the "target" based on \c flags. The idea is that if + * \c missed_deadline is set, then the application is not + * achieving its desired framerate. * \return Zero on success, -1 on error. */ int -driWaitForVBlank( __DRIdrawablePrivate *priv, GLboolean * missed_deadline ) +driWaitForVBlank( const __DRIdrawablePrivate *priv, GLuint * vbl_seq, + GLuint flags, GLboolean * missed_deadline ) { drmVBlank vbl; unsigned original_seq; @@ -375,10 +335,10 @@ driWaitForVBlank( __DRIdrawablePrivate *priv, GLboolean * missed_deadline ) unsigned diff; *missed_deadline = GL_FALSE; - if ( (priv->vblFlags & (VBLANK_FLAG_INTERVAL | - VBLANK_FLAG_THROTTLE | - VBLANK_FLAG_SYNC)) == 0 || - (priv->vblFlags & VBLANK_FLAG_NO_IRQ) != 0 ) { + if ( (flags & (VBLANK_FLAG_INTERVAL | + VBLANK_FLAG_THROTTLE | + VBLANK_FLAG_SYNC)) == 0 || + (flags & VBLANK_FLAG_NO_IRQ) != 0 ) { return 0; } @@ -389,45 +349,44 @@ driWaitForVBlank( __DRIdrawablePrivate *priv, GLboolean * missed_deadline ) * * VBLANK_FLAG_INTERVAL and VBLANK_FLAG_THROTTLE mean to wait for at * least one vertical blank since the last wait. Since do_wait modifies - * priv->vblSeq, we have to save the original value of priv->vblSeq for the + * vbl_seq, we have to save the original value of vbl_seq for the * VBLANK_FLAG_INTERVAL / VBLANK_FLAG_THROTTLE calculation later. */ - original_seq = priv->vblSeq; - interval = driGetVBlankInterval(priv); + original_seq = *vbl_seq; + interval = driGetVBlankInterval(priv, flags); deadline = original_seq + interval; vbl.request.type = DRM_VBLANK_RELATIVE; - if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) { + if ( flags & VBLANK_FLAG_SECONDARY ) { vbl.request.type |= DRM_VBLANK_SECONDARY; } - vbl.request.sequence = ((priv->vblFlags & VBLANK_FLAG_SYNC) != 0) ? 1 : 0; + vbl.request.sequence = ((flags & VBLANK_FLAG_SYNC) != 0) ? 1 : 0; - if ( do_wait( & vbl, &priv->vblSeq, priv->driScreenPriv->fd ) != 0 ) { + if ( do_wait( & vbl, vbl_seq, priv->driScreenPriv->fd ) != 0 ) { return -1; } - diff = priv->vblSeq - deadline; + diff = *vbl_seq - deadline; /* No need to wait again if we've already reached the target */ if (diff <= (1 << 23)) { - *missed_deadline = (priv->vblFlags & VBLANK_FLAG_SYNC) ? (diff > 0) : - GL_TRUE; + *missed_deadline = (flags & VBLANK_FLAG_SYNC) ? (diff > 0) : GL_TRUE; return 0; } /* Wait until the target vertical blank. */ vbl.request.type = DRM_VBLANK_ABSOLUTE; - if ( priv->vblFlags & VBLANK_FLAG_SECONDARY ) { + if ( flags & VBLANK_FLAG_SECONDARY ) { vbl.request.type |= DRM_VBLANK_SECONDARY; } vbl.request.sequence = deadline; - if ( do_wait( & vbl, &priv->vblSeq, priv->driScreenPriv->fd ) != 0 ) { + if ( do_wait( & vbl, vbl_seq, priv->driScreenPriv->fd ) != 0 ) { return -1; } - diff = priv->vblSeq - deadline; + diff = *vbl_seq - deadline; *missed_deadline = diff > 0 && diff <= (1 << 23); return 0; diff --git a/src/mesa/drivers/dri/common/vblank.h b/src/mesa/drivers/dri/common/vblank.h index b3a0dadab1..52c1933ca5 100644 --- a/src/mesa/drivers/dri/common/vblank.h +++ b/src/mesa/drivers/dri/common/vblank.h @@ -45,17 +45,17 @@ */ extern int driGetMSC32( __DRIscreenPrivate * priv, int64_t * count ); -extern int driDrawableGetMSC32( __DRIscreenPrivate * priv, - __DRIdrawablePrivate * drawablePrivate, - int64_t * count); extern int driWaitForMSC32( __DRIdrawablePrivate *priv, int64_t target_msc, int64_t divisor, int64_t remainder, int64_t * msc ); extern GLuint driGetDefaultVBlankFlags( const driOptionCache *optionCache ); -extern void driDrawableInitVBlank ( __DRIdrawablePrivate *priv ); -extern unsigned driGetVBlankInterval( const __DRIdrawablePrivate *priv ); -extern void driGetCurrentVBlank( __DRIdrawablePrivate *priv ); -extern int driWaitForVBlank( __DRIdrawablePrivate *priv, - GLboolean * missed_deadline ); +extern void driDrawableInitVBlank ( __DRIdrawablePrivate *priv, GLuint flags, + GLuint *vbl_seq ); +extern unsigned driGetVBlankInterval( const __DRIdrawablePrivate *priv, + GLuint flags ); +extern void driGetCurrentVBlank( const __DRIdrawablePrivate *priv, + GLuint flags, GLuint *vbl_seq ); +extern int driWaitForVBlank( const __DRIdrawablePrivate *priv, + GLuint * vbl_seq, GLuint flags, GLboolean * missed_deadline ); #undef usleep #include /* for usleep() */ diff --git a/src/mesa/drivers/dri/common/xmlconfig.c b/src/mesa/drivers/dri/common/xmlconfig.c index 8602d47cf9..b635894fe5 100644 --- a/src/mesa/drivers/dri/common/xmlconfig.c +++ b/src/mesa/drivers/dri/common/xmlconfig.c @@ -279,7 +279,7 @@ static GLfloat strToF (const XML_Char *string, const XML_Char **tail) { /** \brief Parse a value of a given type. */ static GLboolean parseValue (driOptionValue *v, driOptionType type, const XML_Char *string) { - const XML_Char *tail = NULL; + const XML_Char *tail; /* skip leading white-space */ string += strspn (string, " \f\n\r\t\v"); switch (type) { @@ -403,40 +403,40 @@ static GLboolean checkValue (const driOptionValue *v, const driOptionInfo *info) /** \brief Output a warning message. */ #define XML_WARNING1(msg) do {\ __driUtilMessage ("Warning in %s line %d, column %d: "msg, data->name, \ - (int) XML_GetCurrentLineNumber(data->parser), \ - (int) XML_GetCurrentColumnNumber(data->parser)); \ + XML_GetCurrentLineNumber(data->parser), \ + XML_GetCurrentColumnNumber(data->parser)); \ } while (0) #define XML_WARNING(msg,args...) do { \ __driUtilMessage ("Warning in %s line %d, column %d: "msg, data->name, \ - (int) XML_GetCurrentLineNumber(data->parser), \ - (int) XML_GetCurrentColumnNumber(data->parser), \ + XML_GetCurrentLineNumber(data->parser), \ + XML_GetCurrentColumnNumber(data->parser), \ args); \ } while (0) /** \brief Output an error message. */ #define XML_ERROR1(msg) do { \ __driUtilMessage ("Error in %s line %d, column %d: "msg, data->name, \ - (int) XML_GetCurrentLineNumber(data->parser), \ - (int) XML_GetCurrentColumnNumber(data->parser)); \ + XML_GetCurrentLineNumber(data->parser), \ + XML_GetCurrentColumnNumber(data->parser)); \ } while (0) #define XML_ERROR(msg,args...) do { \ __driUtilMessage ("Error in %s line %d, column %d: "msg, data->name, \ - (int) XML_GetCurrentLineNumber(data->parser), \ - (int) XML_GetCurrentColumnNumber(data->parser), \ + XML_GetCurrentLineNumber(data->parser), \ + XML_GetCurrentColumnNumber(data->parser), \ args); \ } while (0) /** \brief Output a fatal error message and abort. */ #define XML_FATAL1(msg) do { \ fprintf (stderr, "Fatal error in %s line %d, column %d: "msg"\n", \ data->name, \ - (int) XML_GetCurrentLineNumber(data->parser), \ - (int) XML_GetCurrentColumnNumber(data->parser)); \ + XML_GetCurrentLineNumber(data->parser), \ + XML_GetCurrentColumnNumber(data->parser)); \ abort();\ } while (0) #define XML_FATAL(msg,args...) do { \ fprintf (stderr, "Fatal error in %s line %d, column %d: "msg"\n", \ data->name, \ - (int) XML_GetCurrentLineNumber(data->parser), \ - (int) XML_GetCurrentColumnNumber(data->parser), \ + XML_GetCurrentLineNumber(data->parser), \ + XML_GetCurrentColumnNumber(data->parser), \ args); \ abort();\ } while (0) -- cgit v1.2.3 From 6b389b5c2facc431af0ffb21e3a9bcd532765367 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 11 Jun 2008 11:37:46 -0600 Subject: egl: some prototype Windows code (eq for dlopen,dlsym) --- src/egl/main/egldriver.c | 45 +++++++++++++++++++++++++++++++++++++++++---- 1 file changed, 41 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 92c529911c..5ecfb9eb30 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -5,7 +5,6 @@ #include #include -#include #include #include #include "eglconfig.h" @@ -22,6 +21,7 @@ #include "eglsurface.h" #if defined(_EGL_PLATFORM_X) +#include #include "eglx.h" #elif defined(_EGL_PLATFORM_WINDOWS) /* XXX to do */ @@ -165,39 +165,70 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) { _EGLDriver *drv; _EGLMain_t mainFunc; +#if defined(_EGL_PLATFORM_WINDOWS) + HMODULE lib; +#elif defined(_EGL_PLATFORM_X) void *lib; +#endif char driverFilename[1000]; assert(driverName); +#if defined(_EGL_PLATFORM_WINDOWS) + /* XXX untested */ + sprintf(driverFilename, "%s.dll", driverName); + _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename); + lib = LoadLibrary(driverFilename); +#elif defined(_EGL_PLATFORM_X) /* XXX also prepend a directory path??? */ sprintf(driverFilename, "%s.so", driverName); - _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename); lib = dlopen(driverFilename, RTLD_NOW); +#endif + if (!lib) { _eglLog(_EGL_WARNING, "Could not open %s (%s)", driverFilename, dlerror()); return NULL; } +#if defined(_EGL_PLATFORM_WINDOWS) + mainFunc = (_EGLMain_t) GetProcAddress(lib, "_eglMain"); +#elif defined(_EGL_PLATFORM_X) mainFunc = (_EGLMain_t) dlsym(lib, "_eglMain"); +#endif + if (!mainFunc) { _eglLog(_EGL_WARNING, "_eglMain not found in %s", driverFilename); +#if defined(_EGL_PLATFORM_WINDOWS) + FreeLibrary(lib); +#elif defined(_EGL_PLATFORM_X) dlclose(lib); +#endif return NULL; } drv = mainFunc(dpy, args); if (!drv) { +#if defined(_EGL_PLATFORM_WINDOWS) + FreeLibrary(lib); +#elif defined(_EGL_PLATFORM_X) dlclose(lib); +#endif return NULL; } + /* with a recurvise open you want the inner most handle */ - if (!drv->LibHandle) + if (!drv->LibHandle) { drv->LibHandle = lib; - else + } + else { +#if defined(_EGL_PLATFORM_WINDOWS) + FreeLibrary(lib); +#elif defined(_EGL_PLATFORM_X) dlclose(lib); +#endif + } /* update the global notion of supported APIs */ _eglGlobal.ClientAPIsMask |= drv->ClientAPIsMask; @@ -221,7 +252,13 @@ _eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy) */ b = drv->API.Terminate(drv, dpy); + +#if defined(_EGL_PLATFORM_WINDOWS) + FreeLibrary(handle); +#elif defined(_EGL_PLATFORM_X) dlclose(handle); +#endif + return b; } -- cgit v1.2.3 From f61923441f94439b9ca3c4304ce74fe4e13a14b4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 11 Jun 2008 19:35:52 -0600 Subject: egl: some basic docs for libEGL --- src/egl/main/README.txt | 71 +++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 71 insertions(+) create mode 100644 src/egl/main/README.txt (limited to 'src') diff --git a/src/egl/main/README.txt b/src/egl/main/README.txt new file mode 100644 index 0000000000..b3d253dd13 --- /dev/null +++ b/src/egl/main/README.txt @@ -0,0 +1,71 @@ + + +Notes about the EGL library: + + +The EGL code here basically consists of two things: + +1. An EGL API dispatcher. This directly routes all the eglFooBar() API + calls into driver-specific functions. + +2. Fallbacks for EGL API functions. A driver _could_ implement all the + EGL API calls from scratch. But in many cases, the fallbacks provided + in libEGL (such as eglChooseConfig()) will do the job. + + + +Bootstrapping: + +When the apps calls eglOpenDisplay() a device driver is selected and loaded +(look for dlsym() or LoadLibrary() in egldriver.c). + +The driver's _eglMain() function is then called. This driver function +allocates, initializes and returns a new _EGLDriver object (usually a +subclass of that type). + +As part of initialization, the dispatch table in _EGLDriver->API must be +populated with all the EGL entrypoints. Typically, _eglInitDriverFallbacks() +can be used to plug in default/fallback functions. Some functions like +driver->API.Initialize and driver->API.Terminate _must_ be implemented +with driver-specific code (no default/fallback function is possible). + + +A bit later, the app will call eglInitialize(). This will get routed +to the driver->API.Initialize() function. Any additional driver +initialization that wasn't done in _eglMain() should be done at this +point. Typically, this will involve setting up visual configs, etc. + + + +Special Functions: + +Certain EGL functions _must_ be implemented by the driver. This includes: + +eglCreateContext +eglCreateWindowSurface +eglCreatePixmapSurface +eglCreatePBufferSurface +eglMakeCurrent +eglSwapBuffers + +Most of the EGLConfig-related functions can be implemented with the +defaults/fallbacks. Same thing for the eglGet/Query functions. + + + + +Teardown: + +When eglTerminate() is called, the driver->API.Terminate() function is +called. The driver should clean up after itself. eglTerminate() will +then close/unload the driver (shared library). + + + + +Subclassing: + +The internal libEGL data structures such as _EGLDisplay, _EGLContext, +_EGLSurface, etc should be considered base classes from which drivers +will derive subclasses. + -- cgit v1.2.3 From 23422d603ae002a1f368e20cd0f158e057876cb8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 10 Jun 2008 23:22:12 +0900 Subject: gallium: Implement debug_get_num_option. For numeric options. --- src/gallium/auxiliary/util/p_debug.c | 30 ++++++++++++++++++++++++++++-- src/gallium/include/pipe/p_debug.h | 2 +- 2 files changed, 29 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index d1dfc377f8..a0ffb024e4 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -230,8 +230,34 @@ debug_get_bool_option(const char *name, boolean dfault) long debug_get_num_option(const char *name, long dfault) { - /* FIXME */ - return dfault; + long result; + const char *str; + + str = debug_get_option(name, NULL); + if(!str) + result = dfault; + else { + long sign; + char c; + c = *str++; + if(c == '-') { + sign = -1; + c = *str++; + } + else { + sign = 1; + } + result = 0; + while('0' <= c && c <= '9') { + result = result*10 + (c - '0'); + c = *str++; + } + result *= sign; + } + + debug_printf("%s: %s = %li\n", __FUNCTION__, name, result); + + return result; } diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 05eca75201..9011557006 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -289,7 +289,7 @@ boolean debug_get_bool_option(const char *name, boolean dfault); long -debug_get_unsigned_option(const char *name, long dfault); +debug_get_num_option(const char *name, long dfault); unsigned long debug_get_flags_option(const char *name, -- cgit v1.2.3 From 0f552f500c05d041eda751867c779a8ecc11849c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 11 Jun 2008 10:33:41 +0900 Subject: pipebuffer: Fix improper memory free. --- src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c index 6c3502eea7..e90d2e5623 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c @@ -120,7 +120,7 @@ pb_malloc_buffer_create(size_t size, buf->data = align_malloc(size, desc->alignment < sizeof(void*) ? sizeof(void*) : desc->alignment); if(!buf->data) { - align_free(buf); + FREE(buf); return NULL; } -- cgit v1.2.3 From 2fed8d8496af4aa9105de298c6f320a85eb01623 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 11 Jun 2008 12:25:47 +0900 Subject: gallium: Support L16 pixel format. --- src/gallium/auxiliary/util/p_debug.c | 1 + src/gallium/include/pipe/p_format.h | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index a0ffb024e4..e4de12167a 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -370,6 +370,7 @@ static const struct debug_named_value pipe_format_names[] = { DEBUG_NAMED_VALUE(PIPE_FORMAT_A8_UNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_I8_UNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_A8L8_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_L16_UNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_YCBCR), DEBUG_NAMED_VALUE(PIPE_FORMAT_YCBCR_REV), DEBUG_NAMED_VALUE(PIPE_FORMAT_Z16_UNORM), diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 9ba00f8d7b..3aad463049 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -242,6 +242,7 @@ enum pipe_format { PIPE_FORMAT_A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 0, 0, 0, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ PIPE_FORMAT_I8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ PIPE_FORMAT_A8L8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha, luminance */ + PIPE_FORMAT_L16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 2, 2, 2, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ushort luminance */ PIPE_FORMAT_YCBCR = _PIPE_FORMAT_YCBCR( 0 ), PIPE_FORMAT_YCBCR_REV = _PIPE_FORMAT_YCBCR( 1 ), PIPE_FORMAT_Z16_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_Z000, 2, 0, 0, 0, PIPE_FORMAT_TYPE_UNORM ), -- cgit v1.2.3 From 27049189d6221fefe43eb55846efaa51742dcdf4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 11 Jun 2008 19:48:01 -0600 Subject: mesa: refactor: move glTexGen-related functions into new texgen.c file --- src/mesa/main/api_exec.c | 3 + src/mesa/main/attrib.c | 1 + src/mesa/main/sources | 2 + src/mesa/main/texgen.c | 605 +++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/texgen.h | 62 +++++ src/mesa/main/texstate.c | 568 -------------------------------------------- src/mesa/main/texstate.h | 28 --- src/mesa/sources | 1 + 8 files changed, 674 insertions(+), 596 deletions(-) create mode 100644 src/mesa/main/texgen.c create mode 100644 src/mesa/main/texgen.h (limited to 'src') diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c index f7cf42f600..e27c607307 100644 --- a/src/mesa/main/api_exec.c +++ b/src/mesa/main/api_exec.c @@ -107,6 +107,9 @@ #include "state.h" #include "stencil.h" #include "teximage.h" +#if FEATURE_texgen +#include "texgen.h" +#endif #include "texobj.h" #include "texstate.h" #include "mtypes.h" diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index 6ed82da203..009d71a67c 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -48,6 +48,7 @@ #include "scissor.h" #include "simple_list.h" #include "stencil.h" +#include "texgen.h" #include "texobj.h" #include "texstate.h" #include "mtypes.h" diff --git a/src/mesa/main/sources b/src/mesa/main/sources index 5a6cddf4bd..372423096a 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -60,6 +60,7 @@ texcompress_fxt1.c \ texcompress_s3tc.c \ texenvprogram.c \ texformat.c \ +texgen.c \ teximage.c \ texobj.c \ texrender.c \ @@ -138,6 +139,7 @@ texcompress.h \ texenvprogram.h \ texformat.h \ texformat_tmp.h \ +texgen.h \ teximage.h \ texobj.h \ texrender.h \ diff --git a/src/mesa/main/texgen.c b/src/mesa/main/texgen.c new file mode 100644 index 0000000000..1018c3037b --- /dev/null +++ b/src/mesa/main/texgen.c @@ -0,0 +1,605 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file texgen.c + * + * glTexGen-related functions + */ + + +#include "main/glheader.h" +#include "main/context.h" +#include "main/enums.h" +#include "main/macros.h" +#include "main/texgen.h" +#include "math/m_xform.h" + + +#define ENUM_TO_FLOAT(X) ((GLfloat)(GLint)(X)) +#define ENUM_TO_DOUBLE(X) ((GLdouble)(GLint)(X)) + + + +void GLAPIENTRY +_mesa_TexGenfv( GLenum coord, GLenum pname, const GLfloat *params ) +{ + GET_CURRENT_CONTEXT(ctx); + struct gl_texture_unit *texUnit; + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (MESA_VERBOSE&(VERBOSE_API|VERBOSE_TEXTURE)) + _mesa_debug(ctx, "glTexGen %s %s %.1f(%s)...\n", + _mesa_lookup_enum_by_nr(coord), + _mesa_lookup_enum_by_nr(pname), + *params, + _mesa_lookup_enum_by_nr((GLenum) (GLint) *params)); + + if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glTexGen(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + switch (coord) { + case GL_S: + if (pname==GL_TEXTURE_GEN_MODE) { + GLenum mode = (GLenum) (GLint) *params; + GLbitfield bits; + switch (mode) { + case GL_OBJECT_LINEAR: + bits = TEXGEN_OBJ_LINEAR; + break; + case GL_EYE_LINEAR: + bits = TEXGEN_EYE_LINEAR; + break; + case GL_REFLECTION_MAP_NV: + bits = TEXGEN_REFLECTION_MAP_NV; + break; + case GL_NORMAL_MAP_NV: + bits = TEXGEN_NORMAL_MAP_NV; + break; + case GL_SPHERE_MAP: + bits = TEXGEN_SPHERE_MAP; + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(param)" ); + return; + } + if (texUnit->GenModeS == mode) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->GenModeS = mode; + texUnit->_GenBitS = bits; + } + else if (pname==GL_OBJECT_PLANE) { + if (TEST_EQ_4V(texUnit->ObjectPlaneS, params)) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + COPY_4FV(texUnit->ObjectPlaneS, params); + } + else if (pname==GL_EYE_PLANE) { + GLfloat tmp[4]; + /* Transform plane equation by the inverse modelview matrix */ + if (_math_matrix_is_dirty(ctx->ModelviewMatrixStack.Top)) { + _math_matrix_analyse( ctx->ModelviewMatrixStack.Top ); + } + _mesa_transform_vector( tmp, params, ctx->ModelviewMatrixStack.Top->inv ); + if (TEST_EQ_4V(texUnit->EyePlaneS, tmp)) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + COPY_4FV(texUnit->EyePlaneS, tmp); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(pname)" ); + return; + } + break; + case GL_T: + if (pname==GL_TEXTURE_GEN_MODE) { + GLenum mode = (GLenum) (GLint) *params; + GLbitfield bitt; + switch (mode) { + case GL_OBJECT_LINEAR: + bitt = TEXGEN_OBJ_LINEAR; + break; + case GL_EYE_LINEAR: + bitt = TEXGEN_EYE_LINEAR; + break; + case GL_REFLECTION_MAP_NV: + bitt = TEXGEN_REFLECTION_MAP_NV; + break; + case GL_NORMAL_MAP_NV: + bitt = TEXGEN_NORMAL_MAP_NV; + break; + case GL_SPHERE_MAP: + bitt = TEXGEN_SPHERE_MAP; + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(param)" ); + return; + } + if (texUnit->GenModeT == mode) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->GenModeT = mode; + texUnit->_GenBitT = bitt; + } + else if (pname==GL_OBJECT_PLANE) { + if (TEST_EQ_4V(texUnit->ObjectPlaneT, params)) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + COPY_4FV(texUnit->ObjectPlaneT, params); + } + else if (pname==GL_EYE_PLANE) { + GLfloat tmp[4]; + /* Transform plane equation by the inverse modelview matrix */ + if (_math_matrix_is_dirty(ctx->ModelviewMatrixStack.Top)) { + _math_matrix_analyse( ctx->ModelviewMatrixStack.Top ); + } + _mesa_transform_vector( tmp, params, ctx->ModelviewMatrixStack.Top->inv ); + if (TEST_EQ_4V(texUnit->EyePlaneT, tmp)) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + COPY_4FV(texUnit->EyePlaneT, tmp); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(pname)" ); + return; + } + break; + case GL_R: + if (pname==GL_TEXTURE_GEN_MODE) { + GLenum mode = (GLenum) (GLint) *params; + GLbitfield bitr; + switch (mode) { + case GL_OBJECT_LINEAR: + bitr = TEXGEN_OBJ_LINEAR; + break; + case GL_REFLECTION_MAP_NV: + bitr = TEXGEN_REFLECTION_MAP_NV; + break; + case GL_NORMAL_MAP_NV: + bitr = TEXGEN_NORMAL_MAP_NV; + break; + case GL_EYE_LINEAR: + bitr = TEXGEN_EYE_LINEAR; + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(param)" ); + return; + } + if (texUnit->GenModeR == mode) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->GenModeR = mode; + texUnit->_GenBitR = bitr; + } + else if (pname==GL_OBJECT_PLANE) { + if (TEST_EQ_4V(texUnit->ObjectPlaneR, params)) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + COPY_4FV(texUnit->ObjectPlaneR, params); + } + else if (pname==GL_EYE_PLANE) { + GLfloat tmp[4]; + /* Transform plane equation by the inverse modelview matrix */ + if (_math_matrix_is_dirty(ctx->ModelviewMatrixStack.Top)) { + _math_matrix_analyse( ctx->ModelviewMatrixStack.Top ); + } + _mesa_transform_vector( tmp, params, ctx->ModelviewMatrixStack.Top->inv ); + if (TEST_EQ_4V(texUnit->EyePlaneR, tmp)) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + COPY_4FV(texUnit->EyePlaneR, tmp); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(pname)" ); + return; + } + break; + case GL_Q: + if (pname==GL_TEXTURE_GEN_MODE) { + GLenum mode = (GLenum) (GLint) *params; + GLbitfield bitq; + switch (mode) { + case GL_OBJECT_LINEAR: + bitq = TEXGEN_OBJ_LINEAR; + break; + case GL_EYE_LINEAR: + bitq = TEXGEN_EYE_LINEAR; + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(param)" ); + return; + } + if (texUnit->GenModeQ == mode) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->GenModeQ = mode; + texUnit->_GenBitQ = bitq; + } + else if (pname==GL_OBJECT_PLANE) { + if (TEST_EQ_4V(texUnit->ObjectPlaneQ, params)) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + COPY_4FV(texUnit->ObjectPlaneQ, params); + } + else if (pname==GL_EYE_PLANE) { + GLfloat tmp[4]; + /* Transform plane equation by the inverse modelview matrix */ + if (_math_matrix_is_dirty(ctx->ModelviewMatrixStack.Top)) { + _math_matrix_analyse( ctx->ModelviewMatrixStack.Top ); + } + _mesa_transform_vector( tmp, params, ctx->ModelviewMatrixStack.Top->inv ); + if (TEST_EQ_4V(texUnit->EyePlaneQ, tmp)) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + COPY_4FV(texUnit->EyePlaneQ, tmp); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(pname)" ); + return; + } + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(coord)" ); + return; + } + + if (ctx->Driver.TexGen) + ctx->Driver.TexGen( ctx, coord, pname, params ); +} + + +void GLAPIENTRY +_mesa_TexGeniv(GLenum coord, GLenum pname, const GLint *params ) +{ + GLfloat p[4]; + p[0] = (GLfloat) params[0]; + if (pname == GL_TEXTURE_GEN_MODE) { + p[1] = p[2] = p[3] = 0.0F; + } + else { + p[1] = (GLfloat) params[1]; + p[2] = (GLfloat) params[2]; + p[3] = (GLfloat) params[3]; + } + _mesa_TexGenfv(coord, pname, p); +} + + +void GLAPIENTRY +_mesa_TexGend(GLenum coord, GLenum pname, GLdouble param ) +{ + GLfloat p = (GLfloat) param; + _mesa_TexGenfv( coord, pname, &p ); +} + + +void GLAPIENTRY +_mesa_TexGendv(GLenum coord, GLenum pname, const GLdouble *params ) +{ + GLfloat p[4]; + p[0] = (GLfloat) params[0]; + if (pname == GL_TEXTURE_GEN_MODE) { + p[1] = p[2] = p[3] = 0.0F; + } + else { + p[1] = (GLfloat) params[1]; + p[2] = (GLfloat) params[2]; + p[3] = (GLfloat) params[3]; + } + _mesa_TexGenfv( coord, pname, p ); +} + + +void GLAPIENTRY +_mesa_TexGenf( GLenum coord, GLenum pname, GLfloat param ) +{ + _mesa_TexGenfv(coord, pname, ¶m); +} + + +void GLAPIENTRY +_mesa_TexGeni( GLenum coord, GLenum pname, GLint param ) +{ + _mesa_TexGeniv( coord, pname, ¶m ); +} + + + +void GLAPIENTRY +_mesa_GetTexGendv( GLenum coord, GLenum pname, GLdouble *params ) +{ + const struct gl_texture_unit *texUnit; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexGendv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + switch (coord) { + case GL_S: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = ENUM_TO_DOUBLE(texUnit->GenModeS); + } + else if (pname==GL_OBJECT_PLANE) { + COPY_4V( params, texUnit->ObjectPlaneS ); + } + else if (pname==GL_EYE_PLANE) { + COPY_4V( params, texUnit->EyePlaneS ); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(pname)" ); + return; + } + break; + case GL_T: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = ENUM_TO_DOUBLE(texUnit->GenModeT); + } + else if (pname==GL_OBJECT_PLANE) { + COPY_4V( params, texUnit->ObjectPlaneT ); + } + else if (pname==GL_EYE_PLANE) { + COPY_4V( params, texUnit->EyePlaneT ); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(pname)" ); + return; + } + break; + case GL_R: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = ENUM_TO_DOUBLE(texUnit->GenModeR); + } + else if (pname==GL_OBJECT_PLANE) { + COPY_4V( params, texUnit->ObjectPlaneR ); + } + else if (pname==GL_EYE_PLANE) { + COPY_4V( params, texUnit->EyePlaneR ); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(pname)" ); + return; + } + break; + case GL_Q: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = ENUM_TO_DOUBLE(texUnit->GenModeQ); + } + else if (pname==GL_OBJECT_PLANE) { + COPY_4V( params, texUnit->ObjectPlaneQ ); + } + else if (pname==GL_EYE_PLANE) { + COPY_4V( params, texUnit->EyePlaneQ ); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(pname)" ); + return; + } + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(coord)" ); + return; + } +} + + + +void GLAPIENTRY +_mesa_GetTexGenfv( GLenum coord, GLenum pname, GLfloat *params ) +{ + const struct gl_texture_unit *texUnit; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexGenfv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + switch (coord) { + case GL_S: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = ENUM_TO_FLOAT(texUnit->GenModeS); + } + else if (pname==GL_OBJECT_PLANE) { + COPY_4V( params, texUnit->ObjectPlaneS ); + } + else if (pname==GL_EYE_PLANE) { + COPY_4V( params, texUnit->EyePlaneS ); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(pname)" ); + return; + } + break; + case GL_T: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = ENUM_TO_FLOAT(texUnit->GenModeT); + } + else if (pname==GL_OBJECT_PLANE) { + COPY_4V( params, texUnit->ObjectPlaneT ); + } + else if (pname==GL_EYE_PLANE) { + COPY_4V( params, texUnit->EyePlaneT ); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(pname)" ); + return; + } + break; + case GL_R: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = ENUM_TO_FLOAT(texUnit->GenModeR); + } + else if (pname==GL_OBJECT_PLANE) { + COPY_4V( params, texUnit->ObjectPlaneR ); + } + else if (pname==GL_EYE_PLANE) { + COPY_4V( params, texUnit->EyePlaneR ); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(pname)" ); + return; + } + break; + case GL_Q: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = ENUM_TO_FLOAT(texUnit->GenModeQ); + } + else if (pname==GL_OBJECT_PLANE) { + COPY_4V( params, texUnit->ObjectPlaneQ ); + } + else if (pname==GL_EYE_PLANE) { + COPY_4V( params, texUnit->EyePlaneQ ); + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(pname)" ); + return; + } + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(coord)" ); + return; + } +} + + + +void GLAPIENTRY +_mesa_GetTexGeniv( GLenum coord, GLenum pname, GLint *params ) +{ + const struct gl_texture_unit *texUnit; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexGeniv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + switch (coord) { + case GL_S: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = texUnit->GenModeS; + } + else if (pname==GL_OBJECT_PLANE) { + params[0] = (GLint) texUnit->ObjectPlaneS[0]; + params[1] = (GLint) texUnit->ObjectPlaneS[1]; + params[2] = (GLint) texUnit->ObjectPlaneS[2]; + params[3] = (GLint) texUnit->ObjectPlaneS[3]; + } + else if (pname==GL_EYE_PLANE) { + params[0] = (GLint) texUnit->EyePlaneS[0]; + params[1] = (GLint) texUnit->EyePlaneS[1]; + params[2] = (GLint) texUnit->EyePlaneS[2]; + params[3] = (GLint) texUnit->EyePlaneS[3]; + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(pname)" ); + return; + } + break; + case GL_T: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = texUnit->GenModeT; + } + else if (pname==GL_OBJECT_PLANE) { + params[0] = (GLint) texUnit->ObjectPlaneT[0]; + params[1] = (GLint) texUnit->ObjectPlaneT[1]; + params[2] = (GLint) texUnit->ObjectPlaneT[2]; + params[3] = (GLint) texUnit->ObjectPlaneT[3]; + } + else if (pname==GL_EYE_PLANE) { + params[0] = (GLint) texUnit->EyePlaneT[0]; + params[1] = (GLint) texUnit->EyePlaneT[1]; + params[2] = (GLint) texUnit->EyePlaneT[2]; + params[3] = (GLint) texUnit->EyePlaneT[3]; + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(pname)" ); + return; + } + break; + case GL_R: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = texUnit->GenModeR; + } + else if (pname==GL_OBJECT_PLANE) { + params[0] = (GLint) texUnit->ObjectPlaneR[0]; + params[1] = (GLint) texUnit->ObjectPlaneR[1]; + params[2] = (GLint) texUnit->ObjectPlaneR[2]; + params[3] = (GLint) texUnit->ObjectPlaneR[3]; + } + else if (pname==GL_EYE_PLANE) { + params[0] = (GLint) texUnit->EyePlaneR[0]; + params[1] = (GLint) texUnit->EyePlaneR[1]; + params[2] = (GLint) texUnit->EyePlaneR[2]; + params[3] = (GLint) texUnit->EyePlaneR[3]; + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(pname)" ); + return; + } + break; + case GL_Q: + if (pname==GL_TEXTURE_GEN_MODE) { + params[0] = texUnit->GenModeQ; + } + else if (pname==GL_OBJECT_PLANE) { + params[0] = (GLint) texUnit->ObjectPlaneQ[0]; + params[1] = (GLint) texUnit->ObjectPlaneQ[1]; + params[2] = (GLint) texUnit->ObjectPlaneQ[2]; + params[3] = (GLint) texUnit->ObjectPlaneQ[3]; + } + else if (pname==GL_EYE_PLANE) { + params[0] = (GLint) texUnit->EyePlaneQ[0]; + params[1] = (GLint) texUnit->EyePlaneQ[1]; + params[2] = (GLint) texUnit->EyePlaneQ[2]; + params[3] = (GLint) texUnit->EyePlaneQ[3]; + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(pname)" ); + return; + } + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(coord)" ); + return; + } +} + + diff --git a/src/mesa/main/texgen.h b/src/mesa/main/texgen.h new file mode 100644 index 0000000000..073588efcd --- /dev/null +++ b/src/mesa/main/texgen.h @@ -0,0 +1,62 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef TEXGEN_H +#define TEXGEN_H + + +#include "main/glheader.h" + + +extern void GLAPIENTRY +_mesa_GetTexGendv( GLenum coord, GLenum pname, GLdouble *params ); + +extern void GLAPIENTRY +_mesa_GetTexGenfv( GLenum coord, GLenum pname, GLfloat *params ); + +extern void GLAPIENTRY +_mesa_GetTexGeniv( GLenum coord, GLenum pname, GLint *params ); + +extern void GLAPIENTRY +_mesa_TexGend( GLenum coord, GLenum pname, GLdouble param ); + +extern void GLAPIENTRY +_mesa_TexGendv( GLenum coord, GLenum pname, const GLdouble *params ); + +extern void GLAPIENTRY +_mesa_TexGenf( GLenum coord, GLenum pname, GLfloat param ); + +extern void GLAPIENTRY +_mesa_TexGenfv( GLenum coord, GLenum pname, const GLfloat *params ); + +extern void GLAPIENTRY +_mesa_TexGeni( GLenum coord, GLenum pname, GLint param ); + +extern void GLAPIENTRY +_mesa_TexGeniv( GLenum coord, GLenum pname, const GLint *params ); + + + +#endif /* TEXGEN_H */ diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c index 2a83d6df4a..8140787a7b 100644 --- a/src/mesa/main/texstate.c +++ b/src/mesa/main/texstate.c @@ -2145,574 +2145,6 @@ _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params ) - -/**********************************************************************/ -/* Texture Coord Generation */ -/**********************************************************************/ - -#if FEATURE_texgen -void GLAPIENTRY -_mesa_TexGenfv( GLenum coord, GLenum pname, const GLfloat *params ) -{ - GET_CURRENT_CONTEXT(ctx); - struct gl_texture_unit *texUnit; - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (MESA_VERBOSE&(VERBOSE_API|VERBOSE_TEXTURE)) - _mesa_debug(ctx, "glTexGen %s %s %.1f(%s)...\n", - _mesa_lookup_enum_by_nr(coord), - _mesa_lookup_enum_by_nr(pname), - *params, - _mesa_lookup_enum_by_nr((GLenum) (GLint) *params)); - - if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glTexGen(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - switch (coord) { - case GL_S: - if (pname==GL_TEXTURE_GEN_MODE) { - GLenum mode = (GLenum) (GLint) *params; - GLbitfield bits; - switch (mode) { - case GL_OBJECT_LINEAR: - bits = TEXGEN_OBJ_LINEAR; - break; - case GL_EYE_LINEAR: - bits = TEXGEN_EYE_LINEAR; - break; - case GL_REFLECTION_MAP_NV: - bits = TEXGEN_REFLECTION_MAP_NV; - break; - case GL_NORMAL_MAP_NV: - bits = TEXGEN_NORMAL_MAP_NV; - break; - case GL_SPHERE_MAP: - bits = TEXGEN_SPHERE_MAP; - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(param)" ); - return; - } - if (texUnit->GenModeS == mode) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->GenModeS = mode; - texUnit->_GenBitS = bits; - } - else if (pname==GL_OBJECT_PLANE) { - if (TEST_EQ_4V(texUnit->ObjectPlaneS, params)) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - COPY_4FV(texUnit->ObjectPlaneS, params); - } - else if (pname==GL_EYE_PLANE) { - GLfloat tmp[4]; - /* Transform plane equation by the inverse modelview matrix */ - if (_math_matrix_is_dirty(ctx->ModelviewMatrixStack.Top)) { - _math_matrix_analyse( ctx->ModelviewMatrixStack.Top ); - } - _mesa_transform_vector( tmp, params, ctx->ModelviewMatrixStack.Top->inv ); - if (TEST_EQ_4V(texUnit->EyePlaneS, tmp)) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - COPY_4FV(texUnit->EyePlaneS, tmp); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(pname)" ); - return; - } - break; - case GL_T: - if (pname==GL_TEXTURE_GEN_MODE) { - GLenum mode = (GLenum) (GLint) *params; - GLbitfield bitt; - switch (mode) { - case GL_OBJECT_LINEAR: - bitt = TEXGEN_OBJ_LINEAR; - break; - case GL_EYE_LINEAR: - bitt = TEXGEN_EYE_LINEAR; - break; - case GL_REFLECTION_MAP_NV: - bitt = TEXGEN_REFLECTION_MAP_NV; - break; - case GL_NORMAL_MAP_NV: - bitt = TEXGEN_NORMAL_MAP_NV; - break; - case GL_SPHERE_MAP: - bitt = TEXGEN_SPHERE_MAP; - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(param)" ); - return; - } - if (texUnit->GenModeT == mode) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->GenModeT = mode; - texUnit->_GenBitT = bitt; - } - else if (pname==GL_OBJECT_PLANE) { - if (TEST_EQ_4V(texUnit->ObjectPlaneT, params)) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - COPY_4FV(texUnit->ObjectPlaneT, params); - } - else if (pname==GL_EYE_PLANE) { - GLfloat tmp[4]; - /* Transform plane equation by the inverse modelview matrix */ - if (_math_matrix_is_dirty(ctx->ModelviewMatrixStack.Top)) { - _math_matrix_analyse( ctx->ModelviewMatrixStack.Top ); - } - _mesa_transform_vector( tmp, params, ctx->ModelviewMatrixStack.Top->inv ); - if (TEST_EQ_4V(texUnit->EyePlaneT, tmp)) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - COPY_4FV(texUnit->EyePlaneT, tmp); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(pname)" ); - return; - } - break; - case GL_R: - if (pname==GL_TEXTURE_GEN_MODE) { - GLenum mode = (GLenum) (GLint) *params; - GLbitfield bitr; - switch (mode) { - case GL_OBJECT_LINEAR: - bitr = TEXGEN_OBJ_LINEAR; - break; - case GL_REFLECTION_MAP_NV: - bitr = TEXGEN_REFLECTION_MAP_NV; - break; - case GL_NORMAL_MAP_NV: - bitr = TEXGEN_NORMAL_MAP_NV; - break; - case GL_EYE_LINEAR: - bitr = TEXGEN_EYE_LINEAR; - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(param)" ); - return; - } - if (texUnit->GenModeR == mode) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->GenModeR = mode; - texUnit->_GenBitR = bitr; - } - else if (pname==GL_OBJECT_PLANE) { - if (TEST_EQ_4V(texUnit->ObjectPlaneR, params)) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - COPY_4FV(texUnit->ObjectPlaneR, params); - } - else if (pname==GL_EYE_PLANE) { - GLfloat tmp[4]; - /* Transform plane equation by the inverse modelview matrix */ - if (_math_matrix_is_dirty(ctx->ModelviewMatrixStack.Top)) { - _math_matrix_analyse( ctx->ModelviewMatrixStack.Top ); - } - _mesa_transform_vector( tmp, params, ctx->ModelviewMatrixStack.Top->inv ); - if (TEST_EQ_4V(texUnit->EyePlaneR, tmp)) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - COPY_4FV(texUnit->EyePlaneR, tmp); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(pname)" ); - return; - } - break; - case GL_Q: - if (pname==GL_TEXTURE_GEN_MODE) { - GLenum mode = (GLenum) (GLint) *params; - GLbitfield bitq; - switch (mode) { - case GL_OBJECT_LINEAR: - bitq = TEXGEN_OBJ_LINEAR; - break; - case GL_EYE_LINEAR: - bitq = TEXGEN_EYE_LINEAR; - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(param)" ); - return; - } - if (texUnit->GenModeQ == mode) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->GenModeQ = mode; - texUnit->_GenBitQ = bitq; - } - else if (pname==GL_OBJECT_PLANE) { - if (TEST_EQ_4V(texUnit->ObjectPlaneQ, params)) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - COPY_4FV(texUnit->ObjectPlaneQ, params); - } - else if (pname==GL_EYE_PLANE) { - GLfloat tmp[4]; - /* Transform plane equation by the inverse modelview matrix */ - if (_math_matrix_is_dirty(ctx->ModelviewMatrixStack.Top)) { - _math_matrix_analyse( ctx->ModelviewMatrixStack.Top ); - } - _mesa_transform_vector( tmp, params, ctx->ModelviewMatrixStack.Top->inv ); - if (TEST_EQ_4V(texUnit->EyePlaneQ, tmp)) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - COPY_4FV(texUnit->EyePlaneQ, tmp); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(pname)" ); - return; - } - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glTexGenfv(coord)" ); - return; - } - - if (ctx->Driver.TexGen) - ctx->Driver.TexGen( ctx, coord, pname, params ); -} - - -void GLAPIENTRY -_mesa_TexGeniv(GLenum coord, GLenum pname, const GLint *params ) -{ - GLfloat p[4]; - p[0] = (GLfloat) params[0]; - if (pname == GL_TEXTURE_GEN_MODE) { - p[1] = p[2] = p[3] = 0.0F; - } - else { - p[1] = (GLfloat) params[1]; - p[2] = (GLfloat) params[2]; - p[3] = (GLfloat) params[3]; - } - _mesa_TexGenfv(coord, pname, p); -} - - -void GLAPIENTRY -_mesa_TexGend(GLenum coord, GLenum pname, GLdouble param ) -{ - GLfloat p = (GLfloat) param; - _mesa_TexGenfv( coord, pname, &p ); -} - - -void GLAPIENTRY -_mesa_TexGendv(GLenum coord, GLenum pname, const GLdouble *params ) -{ - GLfloat p[4]; - p[0] = (GLfloat) params[0]; - if (pname == GL_TEXTURE_GEN_MODE) { - p[1] = p[2] = p[3] = 0.0F; - } - else { - p[1] = (GLfloat) params[1]; - p[2] = (GLfloat) params[2]; - p[3] = (GLfloat) params[3]; - } - _mesa_TexGenfv( coord, pname, p ); -} - - -void GLAPIENTRY -_mesa_TexGenf( GLenum coord, GLenum pname, GLfloat param ) -{ - _mesa_TexGenfv(coord, pname, ¶m); -} - - -void GLAPIENTRY -_mesa_TexGeni( GLenum coord, GLenum pname, GLint param ) -{ - _mesa_TexGeniv( coord, pname, ¶m ); -} - - - -void GLAPIENTRY -_mesa_GetTexGendv( GLenum coord, GLenum pname, GLdouble *params ) -{ - const struct gl_texture_unit *texUnit; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexGendv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - switch (coord) { - case GL_S: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = ENUM_TO_DOUBLE(texUnit->GenModeS); - } - else if (pname==GL_OBJECT_PLANE) { - COPY_4V( params, texUnit->ObjectPlaneS ); - } - else if (pname==GL_EYE_PLANE) { - COPY_4V( params, texUnit->EyePlaneS ); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(pname)" ); - return; - } - break; - case GL_T: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = ENUM_TO_DOUBLE(texUnit->GenModeT); - } - else if (pname==GL_OBJECT_PLANE) { - COPY_4V( params, texUnit->ObjectPlaneT ); - } - else if (pname==GL_EYE_PLANE) { - COPY_4V( params, texUnit->EyePlaneT ); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(pname)" ); - return; - } - break; - case GL_R: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = ENUM_TO_DOUBLE(texUnit->GenModeR); - } - else if (pname==GL_OBJECT_PLANE) { - COPY_4V( params, texUnit->ObjectPlaneR ); - } - else if (pname==GL_EYE_PLANE) { - COPY_4V( params, texUnit->EyePlaneR ); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(pname)" ); - return; - } - break; - case GL_Q: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = ENUM_TO_DOUBLE(texUnit->GenModeQ); - } - else if (pname==GL_OBJECT_PLANE) { - COPY_4V( params, texUnit->ObjectPlaneQ ); - } - else if (pname==GL_EYE_PLANE) { - COPY_4V( params, texUnit->EyePlaneQ ); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(pname)" ); - return; - } - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGendv(coord)" ); - return; - } -} - - - -void GLAPIENTRY -_mesa_GetTexGenfv( GLenum coord, GLenum pname, GLfloat *params ) -{ - const struct gl_texture_unit *texUnit; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexGenfv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - switch (coord) { - case GL_S: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = ENUM_TO_FLOAT(texUnit->GenModeS); - } - else if (pname==GL_OBJECT_PLANE) { - COPY_4V( params, texUnit->ObjectPlaneS ); - } - else if (pname==GL_EYE_PLANE) { - COPY_4V( params, texUnit->EyePlaneS ); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(pname)" ); - return; - } - break; - case GL_T: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = ENUM_TO_FLOAT(texUnit->GenModeT); - } - else if (pname==GL_OBJECT_PLANE) { - COPY_4V( params, texUnit->ObjectPlaneT ); - } - else if (pname==GL_EYE_PLANE) { - COPY_4V( params, texUnit->EyePlaneT ); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(pname)" ); - return; - } - break; - case GL_R: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = ENUM_TO_FLOAT(texUnit->GenModeR); - } - else if (pname==GL_OBJECT_PLANE) { - COPY_4V( params, texUnit->ObjectPlaneR ); - } - else if (pname==GL_EYE_PLANE) { - COPY_4V( params, texUnit->EyePlaneR ); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(pname)" ); - return; - } - break; - case GL_Q: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = ENUM_TO_FLOAT(texUnit->GenModeQ); - } - else if (pname==GL_OBJECT_PLANE) { - COPY_4V( params, texUnit->ObjectPlaneQ ); - } - else if (pname==GL_EYE_PLANE) { - COPY_4V( params, texUnit->EyePlaneQ ); - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(pname)" ); - return; - } - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGenfv(coord)" ); - return; - } -} - - - -void GLAPIENTRY -_mesa_GetTexGeniv( GLenum coord, GLenum pname, GLint *params ) -{ - const struct gl_texture_unit *texUnit; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureCoordUnits) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexGeniv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - switch (coord) { - case GL_S: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = texUnit->GenModeS; - } - else if (pname==GL_OBJECT_PLANE) { - params[0] = (GLint) texUnit->ObjectPlaneS[0]; - params[1] = (GLint) texUnit->ObjectPlaneS[1]; - params[2] = (GLint) texUnit->ObjectPlaneS[2]; - params[3] = (GLint) texUnit->ObjectPlaneS[3]; - } - else if (pname==GL_EYE_PLANE) { - params[0] = (GLint) texUnit->EyePlaneS[0]; - params[1] = (GLint) texUnit->EyePlaneS[1]; - params[2] = (GLint) texUnit->EyePlaneS[2]; - params[3] = (GLint) texUnit->EyePlaneS[3]; - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(pname)" ); - return; - } - break; - case GL_T: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = texUnit->GenModeT; - } - else if (pname==GL_OBJECT_PLANE) { - params[0] = (GLint) texUnit->ObjectPlaneT[0]; - params[1] = (GLint) texUnit->ObjectPlaneT[1]; - params[2] = (GLint) texUnit->ObjectPlaneT[2]; - params[3] = (GLint) texUnit->ObjectPlaneT[3]; - } - else if (pname==GL_EYE_PLANE) { - params[0] = (GLint) texUnit->EyePlaneT[0]; - params[1] = (GLint) texUnit->EyePlaneT[1]; - params[2] = (GLint) texUnit->EyePlaneT[2]; - params[3] = (GLint) texUnit->EyePlaneT[3]; - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(pname)" ); - return; - } - break; - case GL_R: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = texUnit->GenModeR; - } - else if (pname==GL_OBJECT_PLANE) { - params[0] = (GLint) texUnit->ObjectPlaneR[0]; - params[1] = (GLint) texUnit->ObjectPlaneR[1]; - params[2] = (GLint) texUnit->ObjectPlaneR[2]; - params[3] = (GLint) texUnit->ObjectPlaneR[3]; - } - else if (pname==GL_EYE_PLANE) { - params[0] = (GLint) texUnit->EyePlaneR[0]; - params[1] = (GLint) texUnit->EyePlaneR[1]; - params[2] = (GLint) texUnit->EyePlaneR[2]; - params[3] = (GLint) texUnit->EyePlaneR[3]; - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(pname)" ); - return; - } - break; - case GL_Q: - if (pname==GL_TEXTURE_GEN_MODE) { - params[0] = texUnit->GenModeQ; - } - else if (pname==GL_OBJECT_PLANE) { - params[0] = (GLint) texUnit->ObjectPlaneQ[0]; - params[1] = (GLint) texUnit->ObjectPlaneQ[1]; - params[2] = (GLint) texUnit->ObjectPlaneQ[2]; - params[3] = (GLint) texUnit->ObjectPlaneQ[3]; - } - else if (pname==GL_EYE_PLANE) { - params[0] = (GLint) texUnit->EyePlaneQ[0]; - params[1] = (GLint) texUnit->EyePlaneQ[1]; - params[2] = (GLint) texUnit->EyePlaneQ[2]; - params[3] = (GLint) texUnit->EyePlaneQ[3]; - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(pname)" ); - return; - } - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexGeniv(coord)" ); - return; - } -} -#endif - - /* GL_ARB_multitexture */ void GLAPIENTRY _mesa_ActiveTextureARB(GLenum texture) diff --git a/src/mesa/main/texstate.h b/src/mesa/main/texstate.h index 60145691b8..67bf487dfc 100644 --- a/src/mesa/main/texstate.h +++ b/src/mesa/main/texstate.h @@ -54,15 +54,6 @@ _mesa_GetTexEnvfv( GLenum target, GLenum pname, GLfloat *params ); extern void GLAPIENTRY _mesa_GetTexEnviv( GLenum target, GLenum pname, GLint *params ); -extern void GLAPIENTRY -_mesa_GetTexGendv( GLenum coord, GLenum pname, GLdouble *params ); - -extern void GLAPIENTRY -_mesa_GetTexGenfv( GLenum coord, GLenum pname, GLfloat *params ); - -extern void GLAPIENTRY -_mesa_GetTexGeniv( GLenum coord, GLenum pname, GLint *params ); - extern void GLAPIENTRY _mesa_GetTexLevelParameterfv( GLenum target, GLint level, GLenum pname, GLfloat *params ); @@ -105,25 +96,6 @@ extern void GLAPIENTRY _mesa_TexParameteriv( GLenum target, GLenum pname, const GLint *params ); -extern void GLAPIENTRY -_mesa_TexGend( GLenum coord, GLenum pname, GLdouble param ); - -extern void GLAPIENTRY -_mesa_TexGendv( GLenum coord, GLenum pname, const GLdouble *params ); - -extern void GLAPIENTRY -_mesa_TexGenf( GLenum coord, GLenum pname, GLfloat param ); - -extern void GLAPIENTRY -_mesa_TexGenfv( GLenum coord, GLenum pname, const GLfloat *params ); - -extern void GLAPIENTRY -_mesa_TexGeni( GLenum coord, GLenum pname, GLint param ); - -extern void GLAPIENTRY -_mesa_TexGeniv( GLenum coord, GLenum pname, const GLint *params ); - - /* * GL_ARB_multitexture */ diff --git a/src/mesa/sources b/src/mesa/sources index 23b03932be..d5a9103fb3 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -63,6 +63,7 @@ MAIN_SOURCES = \ main/texcompress_fxt1.c \ main/texenvprogram.c \ main/texformat.c \ + main/texgen.c \ main/teximage.c \ main/texobj.c \ main/texrender.c \ -- cgit v1.2.3 From 7ecac78ab53016ae3db3dd601b187cb050037463 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 11 Jun 2008 19:58:30 -0600 Subject: mesa: refactor: move glTexEnv-related functions into new texenv.c file --- src/mesa/main/api_exec.c | 1 + src/mesa/main/attrib.c | 1 + src/mesa/main/sources | 2 + src/mesa/main/texenv.c | 863 +++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/texenv.h | 52 +++ src/mesa/main/texstate.c | 823 -------------------------------------------- src/mesa/main/texstate.h | 18 - src/mesa/sources | 1 + 8 files changed, 920 insertions(+), 841 deletions(-) create mode 100644 src/mesa/main/texenv.c create mode 100644 src/mesa/main/texenv.h (limited to 'src') diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c index e27c607307..10b3f9276f 100644 --- a/src/mesa/main/api_exec.c +++ b/src/mesa/main/api_exec.c @@ -106,6 +106,7 @@ #include "scissor.h" #include "state.h" #include "stencil.h" +#include "texenv.h" #include "teximage.h" #if FEATURE_texgen #include "texgen.h" diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index 009d71a67c..48356c75fe 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -48,6 +48,7 @@ #include "scissor.h" #include "simple_list.h" #include "stencil.h" +#include "texenv.h" #include "texgen.h" #include "texobj.h" #include "texstate.h" diff --git a/src/mesa/main/sources b/src/mesa/main/sources index 372423096a..04a2101bb9 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -58,6 +58,7 @@ stencil.c \ texcompress.c \ texcompress_fxt1.c \ texcompress_s3tc.c \ +texenv.c \ texenvprogram.c \ texformat.c \ texgen.c \ @@ -136,6 +137,7 @@ scissor.h \ state.h \ stencil.h \ texcompress.h \ +texenv.h \ texenvprogram.h \ texformat.h \ texformat_tmp.h \ diff --git a/src/mesa/main/texenv.c b/src/mesa/main/texenv.c new file mode 100644 index 0000000000..2b5b15524e --- /dev/null +++ b/src/mesa/main/texenv.c @@ -0,0 +1,863 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file texenv.c + * + * glTexEnv-related functions + */ + + +#include "main/glheader.h" +#include "main/context.h" +#include "main/enums.h" +#include "main/macros.h" +#include "main/texenv.h" +#include "math/m_xform.h" + + +#define ENUM_TO_FLOAT(X) ((GLfloat)(GLint)(X)) + + +void GLAPIENTRY +_mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param ) +{ + GLuint maxUnit; + GET_CURRENT_CONTEXT(ctx); + struct gl_texture_unit *texUnit; + ASSERT_OUTSIDE_BEGIN_END(ctx); + + maxUnit = (target == GL_POINT_SPRITE_NV && pname == GL_COORD_REPLACE_NV) + ? ctx->Const.MaxTextureCoordUnits : ctx->Const.MaxTextureImageUnits; + if (ctx->Texture.CurrentUnit >= maxUnit) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glTexEnvfv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + +#define TE_ERROR(errCode, msg, value) \ + _mesa_error(ctx, errCode, msg, _mesa_lookup_enum_by_nr(value)); + + if (target == GL_TEXTURE_ENV) { + switch (pname) { + case GL_TEXTURE_ENV_MODE: + { + GLenum mode = (GLenum) (GLint) *param; + if (mode == GL_REPLACE_EXT) + mode = GL_REPLACE; + if (texUnit->EnvMode == mode) + return; + if (mode == GL_MODULATE || + mode == GL_BLEND || + mode == GL_DECAL || + mode == GL_REPLACE || + (mode == GL_ADD && ctx->Extensions.EXT_texture_env_add) || + (mode == GL_COMBINE && + (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine))) { + /* legal */ + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->EnvMode = mode; + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); + return; + } + } + break; + case GL_TEXTURE_ENV_COLOR: + { + GLfloat tmp[4]; + tmp[0] = CLAMP( param[0], 0.0F, 1.0F ); + tmp[1] = CLAMP( param[1], 0.0F, 1.0F ); + tmp[2] = CLAMP( param[2], 0.0F, 1.0F ); + tmp[3] = CLAMP( param[3], 0.0F, 1.0F ); + if (TEST_EQ_4V(tmp, texUnit->EnvColor)) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + COPY_4FV(texUnit->EnvColor, tmp); + } + break; + case GL_COMBINE_RGB: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const GLenum mode = (GLenum) (GLint) *param; + if (texUnit->Combine.ModeRGB == mode) + return; + switch (mode) { + case GL_REPLACE: + case GL_MODULATE: + case GL_ADD: + case GL_ADD_SIGNED: + case GL_INTERPOLATE: + /* OK */ + break; + case GL_SUBTRACT: + if (!ctx->Extensions.ARB_texture_env_combine) { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); + return; + } + break; + case GL_DOT3_RGB_EXT: + case GL_DOT3_RGBA_EXT: + if (!ctx->Extensions.EXT_texture_env_dot3) { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); + return; + } + break; + case GL_DOT3_RGB: + case GL_DOT3_RGBA: + if (!ctx->Extensions.ARB_texture_env_dot3) { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); + return; + } + break; + case GL_MODULATE_ADD_ATI: + case GL_MODULATE_SIGNED_ADD_ATI: + case GL_MODULATE_SUBTRACT_ATI: + if (!ctx->Extensions.ATI_texture_env_combine3) { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); + return; + } + break; + default: + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); + return; + } + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.ModeRGB = mode; + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + case GL_COMBINE_ALPHA: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const GLenum mode = (GLenum) (GLint) *param; + if (texUnit->Combine.ModeA == mode) + return; + switch (mode) { + case GL_REPLACE: + case GL_MODULATE: + case GL_ADD: + case GL_ADD_SIGNED: + case GL_INTERPOLATE: + /* OK */ + break; + case GL_SUBTRACT: + if (!ctx->Extensions.ARB_texture_env_combine) { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); + return; + } + break; + case GL_MODULATE_ADD_ATI: + case GL_MODULATE_SIGNED_ADD_ATI: + case GL_MODULATE_SUBTRACT_ATI: + if (!ctx->Extensions.ATI_texture_env_combine3) { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); + return; + } + break; + default: + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); + return; + } + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.ModeA = mode; + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + case GL_SOURCE0_RGB: + case GL_SOURCE1_RGB: + case GL_SOURCE2_RGB: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const GLenum source = (GLenum) (GLint) *param; + const GLuint s = pname - GL_SOURCE0_RGB; + if (texUnit->Combine.SourceRGB[s] == source) + return; + if (source == GL_TEXTURE || + source == GL_CONSTANT || + source == GL_PRIMARY_COLOR || + source == GL_PREVIOUS || + (ctx->Extensions.ARB_texture_env_crossbar && + source >= GL_TEXTURE0 && + source < GL_TEXTURE0 + ctx->Const.MaxTextureUnits) || + (ctx->Extensions.ATI_texture_env_combine3 && + (source == GL_ZERO || source == GL_ONE))) { + /* legal */ + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.SourceRGB[s] = source; + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", source); + return; + } + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + case GL_SOURCE0_ALPHA: + case GL_SOURCE1_ALPHA: + case GL_SOURCE2_ALPHA: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const GLenum source = (GLenum) (GLint) *param; + const GLuint s = pname - GL_SOURCE0_ALPHA; + if (texUnit->Combine.SourceA[s] == source) + return; + if (source == GL_TEXTURE || + source == GL_CONSTANT || + source == GL_PRIMARY_COLOR || + source == GL_PREVIOUS || + (ctx->Extensions.ARB_texture_env_crossbar && + source >= GL_TEXTURE0 && + source < GL_TEXTURE0 + ctx->Const.MaxTextureUnits) || + (ctx->Extensions.ATI_texture_env_combine3 && + (source == GL_ZERO || source == GL_ONE))) { + /* legal */ + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.SourceA[s] = source; + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", source); + return; + } + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + case GL_OPERAND0_RGB: + case GL_OPERAND1_RGB: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const GLenum operand = (GLenum) (GLint) *param; + const GLuint s = pname - GL_OPERAND0_RGB; + if (texUnit->Combine.OperandRGB[s] == operand) + return; + switch (operand) { + case GL_SRC_COLOR: + case GL_ONE_MINUS_SRC_COLOR: + case GL_SRC_ALPHA: + case GL_ONE_MINUS_SRC_ALPHA: + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.OperandRGB[s] = operand; + break; + default: + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); + return; + } + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + case GL_OPERAND0_ALPHA: + case GL_OPERAND1_ALPHA: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const GLenum operand = (GLenum) (GLint) *param; + if (texUnit->Combine.OperandA[pname-GL_OPERAND0_ALPHA] == operand) + return; + switch (operand) { + case GL_SRC_ALPHA: + case GL_ONE_MINUS_SRC_ALPHA: + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.OperandA[pname-GL_OPERAND0_ALPHA] = operand; + break; + default: + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); + return; + } + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + case GL_OPERAND2_RGB: + if (ctx->Extensions.ARB_texture_env_combine) { + const GLenum operand = (GLenum) (GLint) *param; + if (texUnit->Combine.OperandRGB[2] == operand) + return; + switch (operand) { + case GL_SRC_COLOR: /* ARB combine only */ + case GL_ONE_MINUS_SRC_COLOR: /* ARB combine only */ + case GL_SRC_ALPHA: + case GL_ONE_MINUS_SRC_ALPHA: /* ARB combine only */ + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.OperandRGB[2] = operand; + break; + default: + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); + return; + } + } + else if (ctx->Extensions.EXT_texture_env_combine) { + const GLenum operand = (GLenum) (GLint) *param; + if (texUnit->Combine.OperandRGB[2] == operand) + return; + /* operand must be GL_SRC_ALPHA which is the initial value - thus + don't need to actually compare the operand to the possible value */ + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); + return; + } + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + case GL_OPERAND2_ALPHA: + if (ctx->Extensions.ARB_texture_env_combine) { + const GLenum operand = (GLenum) (GLint) *param; + if (texUnit->Combine.OperandA[2] == operand) + return; + switch (operand) { + case GL_SRC_ALPHA: + case GL_ONE_MINUS_SRC_ALPHA: /* ARB combine only */ + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.OperandA[2] = operand; + break; + default: + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); + return; + } + } + else if (ctx->Extensions.EXT_texture_env_combine) { + const GLenum operand = (GLenum) (GLint) *param; + if (texUnit->Combine.OperandA[2] == operand) + return; + /* operand must be GL_SRC_ALPHA which is the initial value - thus + don't need to actually compare the operand to the possible value */ + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); + return; + } + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + case GL_RGB_SCALE: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + GLuint newshift; + if (*param == 1.0) { + newshift = 0; + } + else if (*param == 2.0) { + newshift = 1; + } + else if (*param == 4.0) { + newshift = 2; + } + else { + _mesa_error( ctx, GL_INVALID_VALUE, + "glTexEnv(GL_RGB_SCALE not 1, 2 or 4)" ); + return; + } + if (texUnit->Combine.ScaleShiftRGB == newshift) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.ScaleShiftRGB = newshift; + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + case GL_ALPHA_SCALE: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + GLuint newshift; + if (*param == 1.0) { + newshift = 0; + } + else if (*param == 2.0) { + newshift = 1; + } + else if (*param == 4.0) { + newshift = 2; + } + else { + _mesa_error( ctx, GL_INVALID_VALUE, + "glTexEnv(GL_ALPHA_SCALE not 1, 2 or 4)" ); + return; + } + if (texUnit->Combine.ScaleShiftA == newshift) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->Combine.ScaleShiftA = newshift; + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(pname)" ); + return; + } + } + else if (target == GL_TEXTURE_FILTER_CONTROL_EXT) { + /* GL_EXT_texture_lod_bias */ + if (!ctx->Extensions.EXT_texture_lod_bias) { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(target=0x%x)", target ); + return; + } + if (pname == GL_TEXTURE_LOD_BIAS_EXT) { + if (texUnit->LodBias == param[0]) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texUnit->LodBias = param[0]; + } + else { + TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); + return; + } + } + else if (target == GL_POINT_SPRITE_NV) { + /* GL_ARB_point_sprite / GL_NV_point_sprite */ + if (!ctx->Extensions.NV_point_sprite + && !ctx->Extensions.ARB_point_sprite) { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(target=0x%x)", target ); + return; + } + if (pname == GL_COORD_REPLACE_NV) { + const GLenum value = (GLenum) param[0]; + if (value == GL_TRUE || value == GL_FALSE) { + /* It's kind of weird to set point state via glTexEnv, + * but that's what the spec calls for. + */ + const GLboolean state = (GLboolean) value; + if (ctx->Point.CoordReplace[ctx->Texture.CurrentUnit] == state) + return; + FLUSH_VERTICES(ctx, _NEW_POINT); + ctx->Point.CoordReplace[ctx->Texture.CurrentUnit] = state; + } + else { + _mesa_error( ctx, GL_INVALID_VALUE, "glTexEnv(param=0x%x)", value); + return; + } + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(pname=0x%x)", pname ); + return; + } + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(target=0x%x)",target ); + return; + } + + if (MESA_VERBOSE&(VERBOSE_API|VERBOSE_TEXTURE)) + _mesa_debug(ctx, "glTexEnv %s %s %.1f(%s) ...\n", + _mesa_lookup_enum_by_nr(target), + _mesa_lookup_enum_by_nr(pname), + *param, + _mesa_lookup_enum_by_nr((GLenum) (GLint) *param)); + + /* Tell device driver about the new texture environment */ + if (ctx->Driver.TexEnv) { + (*ctx->Driver.TexEnv)( ctx, target, pname, param ); + } +} + + +void GLAPIENTRY +_mesa_TexEnvf( GLenum target, GLenum pname, GLfloat param ) +{ + _mesa_TexEnvfv( target, pname, ¶m ); +} + + + +void GLAPIENTRY +_mesa_TexEnvi( GLenum target, GLenum pname, GLint param ) +{ + GLfloat p[4]; + p[0] = (GLfloat) param; + p[1] = p[2] = p[3] = 0.0; + _mesa_TexEnvfv( target, pname, p ); +} + + +void GLAPIENTRY +_mesa_TexEnviv( GLenum target, GLenum pname, const GLint *param ) +{ + GLfloat p[4]; + if (pname == GL_TEXTURE_ENV_COLOR) { + p[0] = INT_TO_FLOAT( param[0] ); + p[1] = INT_TO_FLOAT( param[1] ); + p[2] = INT_TO_FLOAT( param[2] ); + p[3] = INT_TO_FLOAT( param[3] ); + } + else { + p[0] = (GLfloat) param[0]; + p[1] = p[2] = p[3] = 0; /* init to zero, just to be safe */ + } + _mesa_TexEnvfv( target, pname, p ); +} + + +void GLAPIENTRY +_mesa_GetTexEnvfv( GLenum target, GLenum pname, GLfloat *params ) +{ + GLuint maxUnit; + const struct gl_texture_unit *texUnit; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + maxUnit = (target == GL_POINT_SPRITE_NV && pname == GL_COORD_REPLACE_NV) + ? ctx->Const.MaxTextureCoordUnits : ctx->Const.MaxTextureImageUnits; + if (ctx->Texture.CurrentUnit >= maxUnit) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexEnvfv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + if (target == GL_TEXTURE_ENV) { + switch (pname) { + case GL_TEXTURE_ENV_MODE: + *params = ENUM_TO_FLOAT(texUnit->EnvMode); + break; + case GL_TEXTURE_ENV_COLOR: + COPY_4FV( params, texUnit->EnvColor ); + break; + case GL_COMBINE_RGB: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + *params = (GLfloat) texUnit->Combine.ModeRGB; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); + } + break; + case GL_COMBINE_ALPHA: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + *params = (GLfloat) texUnit->Combine.ModeA; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); + } + break; + case GL_SOURCE0_RGB: + case GL_SOURCE1_RGB: + case GL_SOURCE2_RGB: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const unsigned rgb_idx = pname - GL_SOURCE0_RGB; + *params = (GLfloat) texUnit->Combine.SourceRGB[rgb_idx]; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); + } + break; + case GL_SOURCE0_ALPHA: + case GL_SOURCE1_ALPHA: + case GL_SOURCE2_ALPHA: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const unsigned alpha_idx = pname - GL_SOURCE0_ALPHA; + *params = (GLfloat) texUnit->Combine.SourceA[alpha_idx]; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); + } + break; + case GL_OPERAND0_RGB: + case GL_OPERAND1_RGB: + case GL_OPERAND2_RGB: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const unsigned op_rgb = pname - GL_OPERAND0_RGB; + *params = (GLfloat) texUnit->Combine.OperandRGB[op_rgb]; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); + } + break; + case GL_OPERAND0_ALPHA: + case GL_OPERAND1_ALPHA: + case GL_OPERAND2_ALPHA: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const unsigned op_alpha = pname - GL_OPERAND0_ALPHA; + *params = (GLfloat) texUnit->Combine.OperandA[op_alpha]; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); + } + break; + case GL_RGB_SCALE: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + if (texUnit->Combine.ScaleShiftRGB == 0) + *params = 1.0; + else if (texUnit->Combine.ScaleShiftRGB == 1) + *params = 2.0; + else + *params = 4.0; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); + return; + } + break; + case GL_ALPHA_SCALE: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + if (texUnit->Combine.ScaleShiftA == 0) + *params = 1.0; + else if (texUnit->Combine.ScaleShiftA == 1) + *params = 2.0; + else + *params = 4.0; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); + return; + } + break; + default: + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname=0x%x)", pname); + } + } + else if (target == GL_TEXTURE_FILTER_CONTROL_EXT) { + /* GL_EXT_texture_lod_bias */ + if (!ctx->Extensions.EXT_texture_lod_bias) { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(target)" ); + return; + } + if (pname == GL_TEXTURE_LOD_BIAS_EXT) { + *params = texUnit->LodBias; + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)" ); + return; + } + } + else if (target == GL_POINT_SPRITE_NV) { + /* GL_ARB_point_sprite / GL_NV_point_sprite */ + if (!ctx->Extensions.NV_point_sprite + && !ctx->Extensions.ARB_point_sprite) { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(target)" ); + return; + } + if (pname == GL_COORD_REPLACE_NV) { + *params = (GLfloat) ctx->Point.CoordReplace[ctx->Texture.CurrentUnit]; + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)" ); + return; + } + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(target)" ); + return; + } +} + + +void GLAPIENTRY +_mesa_GetTexEnviv( GLenum target, GLenum pname, GLint *params ) +{ + GLuint maxUnit; + const struct gl_texture_unit *texUnit; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + maxUnit = (target == GL_POINT_SPRITE_NV && pname == GL_COORD_REPLACE_NV) + ? ctx->Const.MaxTextureCoordUnits : ctx->Const.MaxTextureImageUnits; + if (ctx->Texture.CurrentUnit >= maxUnit) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexEnviv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + if (target == GL_TEXTURE_ENV) { + switch (pname) { + case GL_TEXTURE_ENV_MODE: + *params = (GLint) texUnit->EnvMode; + break; + case GL_TEXTURE_ENV_COLOR: + params[0] = FLOAT_TO_INT( texUnit->EnvColor[0] ); + params[1] = FLOAT_TO_INT( texUnit->EnvColor[1] ); + params[2] = FLOAT_TO_INT( texUnit->EnvColor[2] ); + params[3] = FLOAT_TO_INT( texUnit->EnvColor[3] ); + break; + case GL_COMBINE_RGB: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + *params = (GLint) texUnit->Combine.ModeRGB; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); + } + break; + case GL_COMBINE_ALPHA: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + *params = (GLint) texUnit->Combine.ModeA; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); + } + break; + case GL_SOURCE0_RGB: + case GL_SOURCE1_RGB: + case GL_SOURCE2_RGB: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const unsigned rgb_idx = pname - GL_SOURCE0_RGB; + *params = (GLint) texUnit->Combine.SourceRGB[rgb_idx]; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); + } + break; + case GL_SOURCE0_ALPHA: + case GL_SOURCE1_ALPHA: + case GL_SOURCE2_ALPHA: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const unsigned alpha_idx = pname - GL_SOURCE0_ALPHA; + *params = (GLint) texUnit->Combine.SourceA[alpha_idx]; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); + } + break; + case GL_OPERAND0_RGB: + case GL_OPERAND1_RGB: + case GL_OPERAND2_RGB: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const unsigned op_rgb = pname - GL_OPERAND0_RGB; + *params = (GLint) texUnit->Combine.OperandRGB[op_rgb]; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); + } + break; + case GL_OPERAND0_ALPHA: + case GL_OPERAND1_ALPHA: + case GL_OPERAND2_ALPHA: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + const unsigned op_alpha = pname - GL_OPERAND0_ALPHA; + *params = (GLint) texUnit->Combine.OperandA[op_alpha]; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); + } + break; + case GL_RGB_SCALE: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + if (texUnit->Combine.ScaleShiftRGB == 0) + *params = 1; + else if (texUnit->Combine.ScaleShiftRGB == 1) + *params = 2; + else + *params = 4; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); + return; + } + break; + case GL_ALPHA_SCALE: + if (ctx->Extensions.EXT_texture_env_combine || + ctx->Extensions.ARB_texture_env_combine) { + if (texUnit->Combine.ScaleShiftA == 0) + *params = 1; + else if (texUnit->Combine.ScaleShiftA == 1) + *params = 2; + else + *params = 4; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); + return; + } + break; + default: + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname=0x%x)", + pname); + } + } + else if (target == GL_TEXTURE_FILTER_CONTROL_EXT) { + /* GL_EXT_texture_lod_bias */ + if (!ctx->Extensions.EXT_texture_lod_bias) { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(target)" ); + return; + } + if (pname == GL_TEXTURE_LOD_BIAS_EXT) { + *params = (GLint) texUnit->LodBias; + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)" ); + return; + } + } + else if (target == GL_POINT_SPRITE_NV) { + /* GL_ARB_point_sprite / GL_NV_point_sprite */ + if (!ctx->Extensions.NV_point_sprite + && !ctx->Extensions.ARB_point_sprite) { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(target)" ); + return; + } + if (pname == GL_COORD_REPLACE_NV) { + *params = (GLint) ctx->Point.CoordReplace[ctx->Texture.CurrentUnit]; + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)" ); + return; + } + } + else { + _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(target)" ); + return; + } +} + + diff --git a/src/mesa/main/texenv.h b/src/mesa/main/texenv.h new file mode 100644 index 0000000000..bdff7fdb82 --- /dev/null +++ b/src/mesa/main/texenv.h @@ -0,0 +1,52 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef TEXENV_H +#define TEXENV_H + + +#include "main/glheader.h" + + +extern void GLAPIENTRY +_mesa_GetTexEnvfv( GLenum target, GLenum pname, GLfloat *params ); + +extern void GLAPIENTRY +_mesa_GetTexEnviv( GLenum target, GLenum pname, GLint *params ); + +extern void GLAPIENTRY +_mesa_TexEnvf( GLenum target, GLenum pname, GLfloat param ); + +extern void GLAPIENTRY +_mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param ); + +extern void GLAPIENTRY +_mesa_TexEnvi( GLenum target, GLenum pname, GLint param ); + +extern void GLAPIENTRY +_mesa_TexEnviv( GLenum target, GLenum pname, const GLint *param ); + + +#endif /* TEXENV_H */ diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c index 8140787a7b..448fc53912 100644 --- a/src/mesa/main/texstate.c +++ b/src/mesa/main/texstate.c @@ -321,829 +321,6 @@ calculate_derived_texenv( struct gl_tex_env_combine_state *state, } -void GLAPIENTRY -_mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param ) -{ - GLuint maxUnit; - GET_CURRENT_CONTEXT(ctx); - struct gl_texture_unit *texUnit; - ASSERT_OUTSIDE_BEGIN_END(ctx); - - maxUnit = (target == GL_POINT_SPRITE_NV && pname == GL_COORD_REPLACE_NV) - ? ctx->Const.MaxTextureCoordUnits : ctx->Const.MaxTextureImageUnits; - if (ctx->Texture.CurrentUnit >= maxUnit) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glTexEnvfv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - -#define TE_ERROR(errCode, msg, value) \ - _mesa_error(ctx, errCode, msg, _mesa_lookup_enum_by_nr(value)); - - if (target == GL_TEXTURE_ENV) { - switch (pname) { - case GL_TEXTURE_ENV_MODE: - { - GLenum mode = (GLenum) (GLint) *param; - if (mode == GL_REPLACE_EXT) - mode = GL_REPLACE; - if (texUnit->EnvMode == mode) - return; - if (mode == GL_MODULATE || - mode == GL_BLEND || - mode == GL_DECAL || - mode == GL_REPLACE || - (mode == GL_ADD && ctx->Extensions.EXT_texture_env_add) || - (mode == GL_COMBINE && - (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine))) { - /* legal */ - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->EnvMode = mode; - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); - return; - } - } - break; - case GL_TEXTURE_ENV_COLOR: - { - GLfloat tmp[4]; - tmp[0] = CLAMP( param[0], 0.0F, 1.0F ); - tmp[1] = CLAMP( param[1], 0.0F, 1.0F ); - tmp[2] = CLAMP( param[2], 0.0F, 1.0F ); - tmp[3] = CLAMP( param[3], 0.0F, 1.0F ); - if (TEST_EQ_4V(tmp, texUnit->EnvColor)) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - COPY_4FV(texUnit->EnvColor, tmp); - } - break; - case GL_COMBINE_RGB: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const GLenum mode = (GLenum) (GLint) *param; - if (texUnit->Combine.ModeRGB == mode) - return; - switch (mode) { - case GL_REPLACE: - case GL_MODULATE: - case GL_ADD: - case GL_ADD_SIGNED: - case GL_INTERPOLATE: - /* OK */ - break; - case GL_SUBTRACT: - if (!ctx->Extensions.ARB_texture_env_combine) { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); - return; - } - break; - case GL_DOT3_RGB_EXT: - case GL_DOT3_RGBA_EXT: - if (!ctx->Extensions.EXT_texture_env_dot3) { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); - return; - } - break; - case GL_DOT3_RGB: - case GL_DOT3_RGBA: - if (!ctx->Extensions.ARB_texture_env_dot3) { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); - return; - } - break; - case GL_MODULATE_ADD_ATI: - case GL_MODULATE_SIGNED_ADD_ATI: - case GL_MODULATE_SUBTRACT_ATI: - if (!ctx->Extensions.ATI_texture_env_combine3) { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); - return; - } - break; - default: - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); - return; - } - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.ModeRGB = mode; - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - case GL_COMBINE_ALPHA: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const GLenum mode = (GLenum) (GLint) *param; - if (texUnit->Combine.ModeA == mode) - return; - switch (mode) { - case GL_REPLACE: - case GL_MODULATE: - case GL_ADD: - case GL_ADD_SIGNED: - case GL_INTERPOLATE: - /* OK */ - break; - case GL_SUBTRACT: - if (!ctx->Extensions.ARB_texture_env_combine) { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); - return; - } - break; - case GL_MODULATE_ADD_ATI: - case GL_MODULATE_SIGNED_ADD_ATI: - case GL_MODULATE_SUBTRACT_ATI: - if (!ctx->Extensions.ATI_texture_env_combine3) { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); - return; - } - break; - default: - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", mode); - return; - } - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.ModeA = mode; - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - case GL_SOURCE0_RGB: - case GL_SOURCE1_RGB: - case GL_SOURCE2_RGB: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const GLenum source = (GLenum) (GLint) *param; - const GLuint s = pname - GL_SOURCE0_RGB; - if (texUnit->Combine.SourceRGB[s] == source) - return; - if (source == GL_TEXTURE || - source == GL_CONSTANT || - source == GL_PRIMARY_COLOR || - source == GL_PREVIOUS || - (ctx->Extensions.ARB_texture_env_crossbar && - source >= GL_TEXTURE0 && - source < GL_TEXTURE0 + ctx->Const.MaxTextureUnits) || - (ctx->Extensions.ATI_texture_env_combine3 && - (source == GL_ZERO || source == GL_ONE))) { - /* legal */ - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.SourceRGB[s] = source; - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", source); - return; - } - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - case GL_SOURCE0_ALPHA: - case GL_SOURCE1_ALPHA: - case GL_SOURCE2_ALPHA: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const GLenum source = (GLenum) (GLint) *param; - const GLuint s = pname - GL_SOURCE0_ALPHA; - if (texUnit->Combine.SourceA[s] == source) - return; - if (source == GL_TEXTURE || - source == GL_CONSTANT || - source == GL_PRIMARY_COLOR || - source == GL_PREVIOUS || - (ctx->Extensions.ARB_texture_env_crossbar && - source >= GL_TEXTURE0 && - source < GL_TEXTURE0 + ctx->Const.MaxTextureUnits) || - (ctx->Extensions.ATI_texture_env_combine3 && - (source == GL_ZERO || source == GL_ONE))) { - /* legal */ - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.SourceA[s] = source; - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", source); - return; - } - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - case GL_OPERAND0_RGB: - case GL_OPERAND1_RGB: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const GLenum operand = (GLenum) (GLint) *param; - const GLuint s = pname - GL_OPERAND0_RGB; - if (texUnit->Combine.OperandRGB[s] == operand) - return; - switch (operand) { - case GL_SRC_COLOR: - case GL_ONE_MINUS_SRC_COLOR: - case GL_SRC_ALPHA: - case GL_ONE_MINUS_SRC_ALPHA: - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.OperandRGB[s] = operand; - break; - default: - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); - return; - } - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - case GL_OPERAND0_ALPHA: - case GL_OPERAND1_ALPHA: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const GLenum operand = (GLenum) (GLint) *param; - if (texUnit->Combine.OperandA[pname-GL_OPERAND0_ALPHA] == operand) - return; - switch (operand) { - case GL_SRC_ALPHA: - case GL_ONE_MINUS_SRC_ALPHA: - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.OperandA[pname-GL_OPERAND0_ALPHA] = operand; - break; - default: - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); - return; - } - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - case GL_OPERAND2_RGB: - if (ctx->Extensions.ARB_texture_env_combine) { - const GLenum operand = (GLenum) (GLint) *param; - if (texUnit->Combine.OperandRGB[2] == operand) - return; - switch (operand) { - case GL_SRC_COLOR: /* ARB combine only */ - case GL_ONE_MINUS_SRC_COLOR: /* ARB combine only */ - case GL_SRC_ALPHA: - case GL_ONE_MINUS_SRC_ALPHA: /* ARB combine only */ - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.OperandRGB[2] = operand; - break; - default: - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); - return; - } - } - else if (ctx->Extensions.EXT_texture_env_combine) { - const GLenum operand = (GLenum) (GLint) *param; - if (texUnit->Combine.OperandRGB[2] == operand) - return; - /* operand must be GL_SRC_ALPHA which is the initial value - thus - don't need to actually compare the operand to the possible value */ - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); - return; - } - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - case GL_OPERAND2_ALPHA: - if (ctx->Extensions.ARB_texture_env_combine) { - const GLenum operand = (GLenum) (GLint) *param; - if (texUnit->Combine.OperandA[2] == operand) - return; - switch (operand) { - case GL_SRC_ALPHA: - case GL_ONE_MINUS_SRC_ALPHA: /* ARB combine only */ - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.OperandA[2] = operand; - break; - default: - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); - return; - } - } - else if (ctx->Extensions.EXT_texture_env_combine) { - const GLenum operand = (GLenum) (GLint) *param; - if (texUnit->Combine.OperandA[2] == operand) - return; - /* operand must be GL_SRC_ALPHA which is the initial value - thus - don't need to actually compare the operand to the possible value */ - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(param=%s)", operand); - return; - } - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - case GL_RGB_SCALE: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - GLuint newshift; - if (*param == 1.0) { - newshift = 0; - } - else if (*param == 2.0) { - newshift = 1; - } - else if (*param == 4.0) { - newshift = 2; - } - else { - _mesa_error( ctx, GL_INVALID_VALUE, - "glTexEnv(GL_RGB_SCALE not 1, 2 or 4)" ); - return; - } - if (texUnit->Combine.ScaleShiftRGB == newshift) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.ScaleShiftRGB = newshift; - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - case GL_ALPHA_SCALE: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - GLuint newshift; - if (*param == 1.0) { - newshift = 0; - } - else if (*param == 2.0) { - newshift = 1; - } - else if (*param == 4.0) { - newshift = 2; - } - else { - _mesa_error( ctx, GL_INVALID_VALUE, - "glTexEnv(GL_ALPHA_SCALE not 1, 2 or 4)" ); - return; - } - if (texUnit->Combine.ScaleShiftA == newshift) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->Combine.ScaleShiftA = newshift; - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(pname)" ); - return; - } - } - else if (target == GL_TEXTURE_FILTER_CONTROL_EXT) { - /* GL_EXT_texture_lod_bias */ - if (!ctx->Extensions.EXT_texture_lod_bias) { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(target=0x%x)", target ); - return; - } - if (pname == GL_TEXTURE_LOD_BIAS_EXT) { - if (texUnit->LodBias == param[0]) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texUnit->LodBias = param[0]; - } - else { - TE_ERROR(GL_INVALID_ENUM, "glTexEnv(pname=%s)", pname); - return; - } - } - else if (target == GL_POINT_SPRITE_NV) { - /* GL_ARB_point_sprite / GL_NV_point_sprite */ - if (!ctx->Extensions.NV_point_sprite - && !ctx->Extensions.ARB_point_sprite) { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(target=0x%x)", target ); - return; - } - if (pname == GL_COORD_REPLACE_NV) { - const GLenum value = (GLenum) param[0]; - if (value == GL_TRUE || value == GL_FALSE) { - /* It's kind of weird to set point state via glTexEnv, - * but that's what the spec calls for. - */ - const GLboolean state = (GLboolean) value; - if (ctx->Point.CoordReplace[ctx->Texture.CurrentUnit] == state) - return; - FLUSH_VERTICES(ctx, _NEW_POINT); - ctx->Point.CoordReplace[ctx->Texture.CurrentUnit] = state; - } - else { - _mesa_error( ctx, GL_INVALID_VALUE, "glTexEnv(param=0x%x)", value); - return; - } - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(pname=0x%x)", pname ); - return; - } - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexEnv(target=0x%x)",target ); - return; - } - - if (MESA_VERBOSE&(VERBOSE_API|VERBOSE_TEXTURE)) - _mesa_debug(ctx, "glTexEnv %s %s %.1f(%s) ...\n", - _mesa_lookup_enum_by_nr(target), - _mesa_lookup_enum_by_nr(pname), - *param, - _mesa_lookup_enum_by_nr((GLenum) (GLint) *param)); - - /* Tell device driver about the new texture environment */ - if (ctx->Driver.TexEnv) { - (*ctx->Driver.TexEnv)( ctx, target, pname, param ); - } -} - - -void GLAPIENTRY -_mesa_TexEnvf( GLenum target, GLenum pname, GLfloat param ) -{ - _mesa_TexEnvfv( target, pname, ¶m ); -} - - - -void GLAPIENTRY -_mesa_TexEnvi( GLenum target, GLenum pname, GLint param ) -{ - GLfloat p[4]; - p[0] = (GLfloat) param; - p[1] = p[2] = p[3] = 0.0; - _mesa_TexEnvfv( target, pname, p ); -} - - -void GLAPIENTRY -_mesa_TexEnviv( GLenum target, GLenum pname, const GLint *param ) -{ - GLfloat p[4]; - if (pname == GL_TEXTURE_ENV_COLOR) { - p[0] = INT_TO_FLOAT( param[0] ); - p[1] = INT_TO_FLOAT( param[1] ); - p[2] = INT_TO_FLOAT( param[2] ); - p[3] = INT_TO_FLOAT( param[3] ); - } - else { - p[0] = (GLfloat) param[0]; - p[1] = p[2] = p[3] = 0; /* init to zero, just to be safe */ - } - _mesa_TexEnvfv( target, pname, p ); -} - - -void GLAPIENTRY -_mesa_GetTexEnvfv( GLenum target, GLenum pname, GLfloat *params ) -{ - GLuint maxUnit; - const struct gl_texture_unit *texUnit; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - maxUnit = (target == GL_POINT_SPRITE_NV && pname == GL_COORD_REPLACE_NV) - ? ctx->Const.MaxTextureCoordUnits : ctx->Const.MaxTextureImageUnits; - if (ctx->Texture.CurrentUnit >= maxUnit) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexEnvfv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - if (target == GL_TEXTURE_ENV) { - switch (pname) { - case GL_TEXTURE_ENV_MODE: - *params = ENUM_TO_FLOAT(texUnit->EnvMode); - break; - case GL_TEXTURE_ENV_COLOR: - COPY_4FV( params, texUnit->EnvColor ); - break; - case GL_COMBINE_RGB: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - *params = (GLfloat) texUnit->Combine.ModeRGB; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); - } - break; - case GL_COMBINE_ALPHA: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - *params = (GLfloat) texUnit->Combine.ModeA; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); - } - break; - case GL_SOURCE0_RGB: - case GL_SOURCE1_RGB: - case GL_SOURCE2_RGB: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const unsigned rgb_idx = pname - GL_SOURCE0_RGB; - *params = (GLfloat) texUnit->Combine.SourceRGB[rgb_idx]; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); - } - break; - case GL_SOURCE0_ALPHA: - case GL_SOURCE1_ALPHA: - case GL_SOURCE2_ALPHA: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const unsigned alpha_idx = pname - GL_SOURCE0_ALPHA; - *params = (GLfloat) texUnit->Combine.SourceA[alpha_idx]; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); - } - break; - case GL_OPERAND0_RGB: - case GL_OPERAND1_RGB: - case GL_OPERAND2_RGB: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const unsigned op_rgb = pname - GL_OPERAND0_RGB; - *params = (GLfloat) texUnit->Combine.OperandRGB[op_rgb]; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); - } - break; - case GL_OPERAND0_ALPHA: - case GL_OPERAND1_ALPHA: - case GL_OPERAND2_ALPHA: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const unsigned op_alpha = pname - GL_OPERAND0_ALPHA; - *params = (GLfloat) texUnit->Combine.OperandA[op_alpha]; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); - } - break; - case GL_RGB_SCALE: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - if (texUnit->Combine.ScaleShiftRGB == 0) - *params = 1.0; - else if (texUnit->Combine.ScaleShiftRGB == 1) - *params = 2.0; - else - *params = 4.0; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); - return; - } - break; - case GL_ALPHA_SCALE: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - if (texUnit->Combine.ScaleShiftA == 0) - *params = 1.0; - else if (texUnit->Combine.ScaleShiftA == 1) - *params = 2.0; - else - *params = 4.0; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)"); - return; - } - break; - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname=0x%x)", pname); - } - } - else if (target == GL_TEXTURE_FILTER_CONTROL_EXT) { - /* GL_EXT_texture_lod_bias */ - if (!ctx->Extensions.EXT_texture_lod_bias) { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(target)" ); - return; - } - if (pname == GL_TEXTURE_LOD_BIAS_EXT) { - *params = texUnit->LodBias; - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)" ); - return; - } - } - else if (target == GL_POINT_SPRITE_NV) { - /* GL_ARB_point_sprite / GL_NV_point_sprite */ - if (!ctx->Extensions.NV_point_sprite - && !ctx->Extensions.ARB_point_sprite) { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(target)" ); - return; - } - if (pname == GL_COORD_REPLACE_NV) { - *params = (GLfloat) ctx->Point.CoordReplace[ctx->Texture.CurrentUnit]; - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(pname)" ); - return; - } - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnvfv(target)" ); - return; - } -} - - -void GLAPIENTRY -_mesa_GetTexEnviv( GLenum target, GLenum pname, GLint *params ) -{ - GLuint maxUnit; - const struct gl_texture_unit *texUnit; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - maxUnit = (target == GL_POINT_SPRITE_NV && pname == GL_COORD_REPLACE_NV) - ? ctx->Const.MaxTextureCoordUnits : ctx->Const.MaxTextureImageUnits; - if (ctx->Texture.CurrentUnit >= maxUnit) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glGetTexEnviv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - if (target == GL_TEXTURE_ENV) { - switch (pname) { - case GL_TEXTURE_ENV_MODE: - *params = (GLint) texUnit->EnvMode; - break; - case GL_TEXTURE_ENV_COLOR: - params[0] = FLOAT_TO_INT( texUnit->EnvColor[0] ); - params[1] = FLOAT_TO_INT( texUnit->EnvColor[1] ); - params[2] = FLOAT_TO_INT( texUnit->EnvColor[2] ); - params[3] = FLOAT_TO_INT( texUnit->EnvColor[3] ); - break; - case GL_COMBINE_RGB: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - *params = (GLint) texUnit->Combine.ModeRGB; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); - } - break; - case GL_COMBINE_ALPHA: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - *params = (GLint) texUnit->Combine.ModeA; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); - } - break; - case GL_SOURCE0_RGB: - case GL_SOURCE1_RGB: - case GL_SOURCE2_RGB: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const unsigned rgb_idx = pname - GL_SOURCE0_RGB; - *params = (GLint) texUnit->Combine.SourceRGB[rgb_idx]; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); - } - break; - case GL_SOURCE0_ALPHA: - case GL_SOURCE1_ALPHA: - case GL_SOURCE2_ALPHA: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const unsigned alpha_idx = pname - GL_SOURCE0_ALPHA; - *params = (GLint) texUnit->Combine.SourceA[alpha_idx]; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); - } - break; - case GL_OPERAND0_RGB: - case GL_OPERAND1_RGB: - case GL_OPERAND2_RGB: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const unsigned op_rgb = pname - GL_OPERAND0_RGB; - *params = (GLint) texUnit->Combine.OperandRGB[op_rgb]; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); - } - break; - case GL_OPERAND0_ALPHA: - case GL_OPERAND1_ALPHA: - case GL_OPERAND2_ALPHA: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - const unsigned op_alpha = pname - GL_OPERAND0_ALPHA; - *params = (GLint) texUnit->Combine.OperandA[op_alpha]; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); - } - break; - case GL_RGB_SCALE: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - if (texUnit->Combine.ScaleShiftRGB == 0) - *params = 1; - else if (texUnit->Combine.ScaleShiftRGB == 1) - *params = 2; - else - *params = 4; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); - return; - } - break; - case GL_ALPHA_SCALE: - if (ctx->Extensions.EXT_texture_env_combine || - ctx->Extensions.ARB_texture_env_combine) { - if (texUnit->Combine.ScaleShiftA == 0) - *params = 1; - else if (texUnit->Combine.ScaleShiftA == 1) - *params = 2; - else - *params = 4; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)"); - return; - } - break; - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname=0x%x)", - pname); - } - } - else if (target == GL_TEXTURE_FILTER_CONTROL_EXT) { - /* GL_EXT_texture_lod_bias */ - if (!ctx->Extensions.EXT_texture_lod_bias) { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(target)" ); - return; - } - if (pname == GL_TEXTURE_LOD_BIAS_EXT) { - *params = (GLint) texUnit->LodBias; - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)" ); - return; - } - } - else if (target == GL_POINT_SPRITE_NV) { - /* GL_ARB_point_sprite / GL_NV_point_sprite */ - if (!ctx->Extensions.NV_point_sprite - && !ctx->Extensions.ARB_point_sprite) { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(target)" ); - return; - } - if (pname == GL_COORD_REPLACE_NV) { - *params = (GLint) ctx->Point.CoordReplace[ctx->Texture.CurrentUnit]; - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(pname)" ); - return; - } - } - else { - _mesa_error( ctx, GL_INVALID_ENUM, "glGetTexEnviv(target)" ); - return; - } -} - - - - /**********************************************************************/ /* Texture Parameters */ /**********************************************************************/ diff --git a/src/mesa/main/texstate.h b/src/mesa/main/texstate.h index 67bf487dfc..1f1c2267aa 100644 --- a/src/mesa/main/texstate.h +++ b/src/mesa/main/texstate.h @@ -48,12 +48,6 @@ _mesa_print_texunit_state( GLcontext *ctx, GLuint unit ); */ /*@{*/ -extern void GLAPIENTRY -_mesa_GetTexEnvfv( GLenum target, GLenum pname, GLfloat *params ); - -extern void GLAPIENTRY -_mesa_GetTexEnviv( GLenum target, GLenum pname, GLint *params ); - extern void GLAPIENTRY _mesa_GetTexLevelParameterfv( GLenum target, GLint level, GLenum pname, GLfloat *params ); @@ -69,18 +63,6 @@ extern void GLAPIENTRY _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params ); -extern void GLAPIENTRY -_mesa_TexEnvf( GLenum target, GLenum pname, GLfloat param ); - -extern void GLAPIENTRY -_mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param ); - -extern void GLAPIENTRY -_mesa_TexEnvi( GLenum target, GLenum pname, GLint param ); - -extern void GLAPIENTRY -_mesa_TexEnviv( GLenum target, GLenum pname, const GLint *param ); - extern void GLAPIENTRY _mesa_TexParameterfv( GLenum target, GLenum pname, const GLfloat *params ); diff --git a/src/mesa/sources b/src/mesa/sources index d5a9103fb3..d225d7d15b 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -61,6 +61,7 @@ MAIN_SOURCES = \ main/texcompress.c \ main/texcompress_s3tc.c \ main/texcompress_fxt1.c \ + main/texenv.c \ main/texenvprogram.c \ main/texformat.c \ main/texgen.c \ -- cgit v1.2.3 From 77b794201a96300af4473307a7663500d62296e8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 11 Jun 2008 20:05:53 -0600 Subject: mesa: refactor: move glTexParameter-related functions into new texparam.c file --- src/mesa/main/api_exec.c | 1 + src/mesa/main/attrib.c | 1 + src/mesa/main/sources | 2 + src/mesa/main/texparam.c | 1038 ++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/texparam.h | 63 +++ src/mesa/main/texstate.c | 999 -------------------------------------------- src/mesa/main/texstate.h | 29 -- src/mesa/sources | 1 + 8 files changed, 1106 insertions(+), 1028 deletions(-) create mode 100644 src/mesa/main/texparam.c create mode 100644 src/mesa/main/texparam.h (limited to 'src') diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c index 10b3f9276f..8ebe4a3e4a 100644 --- a/src/mesa/main/api_exec.c +++ b/src/mesa/main/api_exec.c @@ -112,6 +112,7 @@ #include "texgen.h" #endif #include "texobj.h" +#include "texparam.h" #include "texstate.h" #include "mtypes.h" #include "varray.h" diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index 48356c75fe..2e6bb76586 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -51,6 +51,7 @@ #include "texenv.h" #include "texgen.h" #include "texobj.h" +#include "texparam.h" #include "texstate.h" #include "mtypes.h" #include "math/m_xform.h" diff --git a/src/mesa/main/sources b/src/mesa/main/sources index 04a2101bb9..468121bd1d 100644 --- a/src/mesa/main/sources +++ b/src/mesa/main/sources @@ -64,6 +64,7 @@ texformat.c \ texgen.c \ teximage.c \ texobj.c \ +texparam.c \ texrender.c \ texstate.c \ texstore.c \ @@ -144,6 +145,7 @@ texformat_tmp.h \ texgen.h \ teximage.h \ texobj.h \ +texparam.h \ texrender.h \ texstate.h \ texstore.h \ diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c new file mode 100644 index 0000000000..10ad0ef1d8 --- /dev/null +++ b/src/mesa/main/texparam.c @@ -0,0 +1,1038 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file texparam.c + * + * glTexParameter-related functions + */ + + +#include "main/glheader.h" +#include "main/context.h" +#include "main/enums.h" +#include "main/colormac.h" +#include "main/macros.h" +#include "main/texcompress.h" +#include "main/texparam.h" +#include "main/teximage.h" + + +#define ENUM_TO_FLOAT(X) ((GLfloat)(GLint)(X)) + + +/** + * Check if a coordinate wrap mode is supported for the texture target. + * \return GL_TRUE if legal, GL_FALSE otherwise + */ +static GLboolean +validate_texture_wrap_mode(GLcontext * ctx, GLenum target, GLenum wrap) +{ + const struct gl_extensions * const e = & ctx->Extensions; + + if (wrap == GL_CLAMP || wrap == GL_CLAMP_TO_EDGE || + (wrap == GL_CLAMP_TO_BORDER && e->ARB_texture_border_clamp)) { + /* any texture target */ + return GL_TRUE; + } + else if (target != GL_TEXTURE_RECTANGLE_NV && + (wrap == GL_REPEAT || + (wrap == GL_MIRRORED_REPEAT && + e->ARB_texture_mirrored_repeat) || + (wrap == GL_MIRROR_CLAMP_EXT && + (e->ATI_texture_mirror_once || e->EXT_texture_mirror_clamp)) || + (wrap == GL_MIRROR_CLAMP_TO_EDGE_EXT && + (e->ATI_texture_mirror_once || e->EXT_texture_mirror_clamp)) || + (wrap == GL_MIRROR_CLAMP_TO_BORDER_EXT && + (e->EXT_texture_mirror_clamp)))) { + /* non-rectangle texture */ + return GL_TRUE; + } + + _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" ); + return GL_FALSE; +} + + +void GLAPIENTRY +_mesa_TexParameterf( GLenum target, GLenum pname, GLfloat param ) +{ + _mesa_TexParameterfv(target, pname, ¶m); +} + + +void GLAPIENTRY +_mesa_TexParameterfv( GLenum target, GLenum pname, const GLfloat *params ) +{ + const GLenum eparam = (GLenum) (GLint) params[0]; + struct gl_texture_unit *texUnit; + struct gl_texture_object *texObj; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (MESA_VERBOSE&(VERBOSE_API|VERBOSE_TEXTURE)) + _mesa_debug(ctx, "glTexParameter %s %s %.1f(%s)...\n", + _mesa_lookup_enum_by_nr(target), + _mesa_lookup_enum_by_nr(pname), + *params, + _mesa_lookup_enum_by_nr(eparam)); + + if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glTexParameterfv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + switch (target) { + case GL_TEXTURE_1D: + texObj = texUnit->Current1D; + break; + case GL_TEXTURE_2D: + texObj = texUnit->Current2D; + break; + case GL_TEXTURE_3D: + texObj = texUnit->Current3D; + break; + case GL_TEXTURE_CUBE_MAP: + if (!ctx->Extensions.ARB_texture_cube_map) { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); + return; + } + texObj = texUnit->CurrentCubeMap; + break; + case GL_TEXTURE_RECTANGLE_NV: + if (!ctx->Extensions.NV_texture_rectangle) { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); + return; + } + texObj = texUnit->CurrentRect; + break; + case GL_TEXTURE_1D_ARRAY_EXT: + if (!ctx->Extensions.MESA_texture_array) { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); + return; + } + texObj = texUnit->Current1DArray; + break; + case GL_TEXTURE_2D_ARRAY_EXT: + if (!ctx->Extensions.MESA_texture_array) { + _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); + return; + } + texObj = texUnit->Current2DArray; + break; + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); + return; + } + + switch (pname) { + case GL_TEXTURE_MIN_FILTER: + /* A small optimization */ + if (texObj->MinFilter == eparam) + return; + if (eparam==GL_NEAREST || eparam==GL_LINEAR) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->MinFilter = eparam; + } + else if ((eparam==GL_NEAREST_MIPMAP_NEAREST || + eparam==GL_LINEAR_MIPMAP_NEAREST || + eparam==GL_NEAREST_MIPMAP_LINEAR || + eparam==GL_LINEAR_MIPMAP_LINEAR) && + texObj->Target != GL_TEXTURE_RECTANGLE_NV) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->MinFilter = eparam; + } + else { + _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" ); + return; + } + break; + case GL_TEXTURE_MAG_FILTER: + /* A small optimization */ + if (texObj->MagFilter == eparam) + return; + + if (eparam==GL_NEAREST || eparam==GL_LINEAR) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->MagFilter = eparam; + } + else { + _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" ); + return; + } + break; + case GL_TEXTURE_WRAP_S: + if (texObj->WrapS == eparam) + return; + if (validate_texture_wrap_mode(ctx, texObj->Target, eparam)) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->WrapS = eparam; + } + else { + return; + } + break; + case GL_TEXTURE_WRAP_T: + if (texObj->WrapT == eparam) + return; + if (validate_texture_wrap_mode(ctx, texObj->Target, eparam)) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->WrapT = eparam; + } + else { + return; + } + break; + case GL_TEXTURE_WRAP_R: + if (texObj->WrapR == eparam) + return; + if (validate_texture_wrap_mode(ctx, texObj->Target, eparam)) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->WrapR = eparam; + } + else { + return; + } + break; + case GL_TEXTURE_BORDER_COLOR: + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->BorderColor[RCOMP] = params[0]; + texObj->BorderColor[GCOMP] = params[1]; + texObj->BorderColor[BCOMP] = params[2]; + texObj->BorderColor[ACOMP] = params[3]; + UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[RCOMP], params[0]); + UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[GCOMP], params[1]); + UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[BCOMP], params[2]); + UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[ACOMP], params[3]); + break; + case GL_TEXTURE_MIN_LOD: + if (texObj->MinLod == params[0]) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->MinLod = params[0]; + break; + case GL_TEXTURE_MAX_LOD: + if (texObj->MaxLod == params[0]) + return; + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->MaxLod = params[0]; + break; + case GL_TEXTURE_BASE_LEVEL: + if (params[0] < 0.0) { + _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)"); + return; + } + if (target == GL_TEXTURE_RECTANGLE_ARB && params[0] != 0.0) { + _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)"); + return; + } + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->BaseLevel = (GLint) params[0]; + break; + case GL_TEXTURE_MAX_LEVEL: + if (params[0] < 0.0) { + _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)"); + return; + } + if (target == GL_TEXTURE_RECTANGLE_ARB) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glTexParameter(param)"); + return; + } + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->MaxLevel = (GLint) params[0]; + break; + case GL_TEXTURE_PRIORITY: + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->Priority = CLAMP( params[0], 0.0F, 1.0F ); + break; + case GL_TEXTURE_MAX_ANISOTROPY_EXT: + if (ctx->Extensions.EXT_texture_filter_anisotropic) { + if (params[0] < 1.0) { + _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)" ); + return; + } + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + /* clamp to max, that's what NVIDIA does */ + texObj->MaxAnisotropy = MIN2(params[0], + ctx->Const.MaxTextureMaxAnisotropy); + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)"); + return; + } + break; + case GL_TEXTURE_COMPARE_SGIX: + if (ctx->Extensions.SGIX_shadow) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->CompareFlag = params[0] ? GL_TRUE : GL_FALSE; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(pname=GL_TEXTURE_COMPARE_SGIX)"); + return; + } + break; + case GL_TEXTURE_COMPARE_OPERATOR_SGIX: + if (ctx->Extensions.SGIX_shadow) { + GLenum op = (GLenum) params[0]; + if (op == GL_TEXTURE_LEQUAL_R_SGIX || + op == GL_TEXTURE_GEQUAL_R_SGIX) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->CompareOperator = op; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(param)"); + } + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(pname=GL_TEXTURE_COMPARE_OPERATOR_SGIX)"); + return; + } + break; + case GL_SHADOW_AMBIENT_SGIX: /* aka GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ + if (ctx->Extensions.SGIX_shadow_ambient) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->ShadowAmbient = CLAMP(params[0], 0.0F, 1.0F); + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(pname=GL_SHADOW_AMBIENT_SGIX)"); + return; + } + break; + case GL_GENERATE_MIPMAP_SGIS: + if (ctx->Extensions.SGIS_generate_mipmap) { + texObj->GenerateMipmap = params[0] ? GL_TRUE : GL_FALSE; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(pname=GL_GENERATE_MIPMAP_SGIS)"); + return; + } + break; + case GL_TEXTURE_COMPARE_MODE_ARB: + if (ctx->Extensions.ARB_shadow) { + const GLenum mode = (GLenum) params[0]; + if (mode == GL_NONE || mode == GL_COMPARE_R_TO_TEXTURE_ARB) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->CompareMode = mode; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(bad GL_TEXTURE_COMPARE_MODE_ARB: 0x%x)", mode); + return; + } + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(pname=GL_TEXTURE_COMPARE_MODE_ARB)"); + return; + } + break; + case GL_TEXTURE_COMPARE_FUNC_ARB: + if (ctx->Extensions.ARB_shadow) { + const GLenum func = (GLenum) params[0]; + if (func == GL_LEQUAL || func == GL_GEQUAL) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->CompareFunc = func; + } + else if (ctx->Extensions.EXT_shadow_funcs && + (func == GL_EQUAL || + func == GL_NOTEQUAL || + func == GL_LESS || + func == GL_GREATER || + func == GL_ALWAYS || + func == GL_NEVER)) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->CompareFunc = func; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(bad GL_TEXTURE_COMPARE_FUNC_ARB)"); + return; + } + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(pname=GL_TEXTURE_COMPARE_FUNC_ARB)"); + return; + } + break; + case GL_DEPTH_TEXTURE_MODE_ARB: + if (ctx->Extensions.ARB_depth_texture) { + const GLenum result = (GLenum) params[0]; + if (result == GL_LUMINANCE || result == GL_INTENSITY + || result == GL_ALPHA) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->DepthMode = result; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(bad GL_DEPTH_TEXTURE_MODE_ARB)"); + return; + } + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(pname=GL_DEPTH_TEXTURE_MODE_ARB)"); + return; + } + break; + case GL_TEXTURE_LOD_BIAS: + /* NOTE: this is really part of OpenGL 1.4, not EXT_texture_lod_bias*/ + if (ctx->Extensions.EXT_texture_lod_bias) { + if (texObj->LodBias != params[0]) { + FLUSH_VERTICES(ctx, _NEW_TEXTURE); + texObj->LodBias = params[0]; + } + } + break; + + default: + _mesa_error(ctx, GL_INVALID_ENUM, + "glTexParameter(pname=0x%x)", pname); + return; + } + + texObj->_Complete = GL_FALSE; + + if (ctx->Driver.TexParameter) { + (*ctx->Driver.TexParameter)( ctx, target, texObj, pname, params ); + } +} + + +void GLAPIENTRY +_mesa_TexParameteri( GLenum target, GLenum pname, GLint param ) +{ + GLfloat fparam[4]; + if (pname == GL_TEXTURE_PRIORITY) + fparam[0] = INT_TO_FLOAT(param); + else + fparam[0] = (GLfloat) param; + fparam[1] = fparam[2] = fparam[3] = 0.0; + _mesa_TexParameterfv(target, pname, fparam); +} + + +void GLAPIENTRY +_mesa_TexParameteriv( GLenum target, GLenum pname, const GLint *params ) +{ + GLfloat fparam[4]; + if (pname == GL_TEXTURE_BORDER_COLOR) { + fparam[0] = INT_TO_FLOAT(params[0]); + fparam[1] = INT_TO_FLOAT(params[1]); + fparam[2] = INT_TO_FLOAT(params[2]); + fparam[3] = INT_TO_FLOAT(params[3]); + } + else { + if (pname == GL_TEXTURE_PRIORITY) + fparam[0] = INT_TO_FLOAT(params[0]); + else + fparam[0] = (GLfloat) params[0]; + fparam[1] = fparam[2] = fparam[3] = 0.0F; + } + _mesa_TexParameterfv(target, pname, fparam); +} + + +void GLAPIENTRY +_mesa_GetTexLevelParameterfv( GLenum target, GLint level, + GLenum pname, GLfloat *params ) +{ + GLint iparam; + _mesa_GetTexLevelParameteriv( target, level, pname, &iparam ); + *params = (GLfloat) iparam; +} + + +static GLuint +tex_image_dimensions(GLcontext *ctx, GLenum target) +{ + switch (target) { + case GL_TEXTURE_1D: + case GL_PROXY_TEXTURE_1D: + return 1; + case GL_TEXTURE_2D: + case GL_PROXY_TEXTURE_2D: + return 2; + case GL_TEXTURE_3D: + case GL_PROXY_TEXTURE_3D: + return 3; + case GL_TEXTURE_CUBE_MAP: + case GL_PROXY_TEXTURE_CUBE_MAP: + case GL_TEXTURE_CUBE_MAP_POSITIVE_X: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: + case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: + case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: + return ctx->Extensions.ARB_texture_cube_map ? 2 : 0; + case GL_TEXTURE_RECTANGLE_NV: + case GL_PROXY_TEXTURE_RECTANGLE_NV: + return ctx->Extensions.NV_texture_rectangle ? 2 : 0; + case GL_TEXTURE_1D_ARRAY_EXT: + case GL_PROXY_TEXTURE_1D_ARRAY_EXT: + return ctx->Extensions.MESA_texture_array ? 2 : 0; + case GL_TEXTURE_2D_ARRAY_EXT: + case GL_PROXY_TEXTURE_2D_ARRAY_EXT: + return ctx->Extensions.MESA_texture_array ? 3 : 0; + default: + _mesa_problem(ctx, "bad target in _mesa_tex_target_dimensions()"); + return 0; + } +} + + +void GLAPIENTRY +_mesa_GetTexLevelParameteriv( GLenum target, GLint level, + GLenum pname, GLint *params ) +{ + const struct gl_texture_unit *texUnit; + struct gl_texture_object *texObj; + const struct gl_texture_image *img = NULL; + GLuint dimensions; + GLboolean isProxy; + GLint maxLevels; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glGetTexLevelParameteriv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + /* this will catch bad target values */ + dimensions = tex_image_dimensions(ctx, target); /* 1, 2 or 3 */ + if (dimensions == 0) { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexLevelParameter[if]v(target)"); + return; + } + + maxLevels = _mesa_max_texture_levels(ctx, target); + if (maxLevels == 0) { + /* should not happen since was just checked above */ + _mesa_problem(ctx, "maxLevels=0 in _mesa_GetTexLevelParameter"); + return; + } + + if (level < 0 || level >= maxLevels) { + _mesa_error( ctx, GL_INVALID_VALUE, "glGetTexLevelParameter[if]v" ); + return; + } + + texObj = _mesa_select_tex_object(ctx, texUnit, target); + _mesa_lock_texture(ctx, texObj); + + img = _mesa_select_tex_image(ctx, texObj, target, level); + if (!img || !img->TexFormat) { + /* undefined texture image */ + if (pname == GL_TEXTURE_COMPONENTS) + *params = 1; + else + *params = 0; + goto out; + } + + isProxy = _mesa_is_proxy_texture(target); + + switch (pname) { + case GL_TEXTURE_WIDTH: + *params = img->Width; + break; + case GL_TEXTURE_HEIGHT: + *params = img->Height; + break; + case GL_TEXTURE_DEPTH: + *params = img->Depth; + break; + case GL_TEXTURE_INTERNAL_FORMAT: + *params = img->InternalFormat; + break; + case GL_TEXTURE_BORDER: + *params = img->Border; + break; + case GL_TEXTURE_RED_SIZE: + if (img->_BaseFormat == GL_RGB || img->_BaseFormat == GL_RGBA) + *params = img->TexFormat->RedBits; + else + *params = 0; + break; + case GL_TEXTURE_GREEN_SIZE: + if (img->_BaseFormat == GL_RGB || img->_BaseFormat == GL_RGBA) + *params = img->TexFormat->GreenBits; + else + *params = 0; + break; + case GL_TEXTURE_BLUE_SIZE: + if (img->_BaseFormat == GL_RGB || img->_BaseFormat == GL_RGBA) + *params = img->TexFormat->BlueBits; + else + *params = 0; + break; + case GL_TEXTURE_ALPHA_SIZE: + if (img->_BaseFormat == GL_ALPHA || + img->_BaseFormat == GL_LUMINANCE_ALPHA || + img->_BaseFormat == GL_RGBA) + *params = img->TexFormat->AlphaBits; + else + *params = 0; + break; + case GL_TEXTURE_INTENSITY_SIZE: + if (img->_BaseFormat != GL_INTENSITY) + *params = 0; + else if (img->TexFormat->IntensityBits > 0) + *params = img->TexFormat->IntensityBits; + else /* intensity probably stored as rgb texture */ + *params = MIN2(img->TexFormat->RedBits, img->TexFormat->GreenBits); + break; + case GL_TEXTURE_LUMINANCE_SIZE: + if (img->_BaseFormat != GL_LUMINANCE && + img->_BaseFormat != GL_LUMINANCE_ALPHA) + *params = 0; + else if (img->TexFormat->LuminanceBits > 0) + *params = img->TexFormat->LuminanceBits; + else /* luminance probably stored as rgb texture */ + *params = MIN2(img->TexFormat->RedBits, img->TexFormat->GreenBits); + break; + case GL_TEXTURE_INDEX_SIZE_EXT: + if (img->_BaseFormat == GL_COLOR_INDEX) + *params = img->TexFormat->IndexBits; + else + *params = 0; + break; + case GL_TEXTURE_DEPTH_SIZE_ARB: + if (ctx->Extensions.SGIX_depth_texture || + ctx->Extensions.ARB_depth_texture) + *params = img->TexFormat->DepthBits; + else + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + break; + case GL_TEXTURE_STENCIL_SIZE_EXT: + if (ctx->Extensions.EXT_packed_depth_stencil) { + *params = img->TexFormat->StencilBits; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + + /* GL_ARB_texture_compression */ + case GL_TEXTURE_COMPRESSED_IMAGE_SIZE: + if (ctx->Extensions.ARB_texture_compression) { + if (img->IsCompressed && !isProxy) { + /* Don't use ctx->Driver.CompressedTextureSize() since that + * may returned a padded hardware size. + */ + *params = _mesa_compressed_texture_size(ctx, img->Width, + img->Height, img->Depth, + img->TexFormat->MesaFormat); + } + else { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glGetTexLevelParameter[if]v(pname)"); + } + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + case GL_TEXTURE_COMPRESSED: + if (ctx->Extensions.ARB_texture_compression) { + *params = (GLint) img->IsCompressed; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + + /* GL_ARB_texture_float */ + case GL_TEXTURE_RED_TYPE_ARB: + if (ctx->Extensions.ARB_texture_float) { + *params = img->TexFormat->RedBits ? img->TexFormat->DataType : GL_NONE; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + case GL_TEXTURE_GREEN_TYPE_ARB: + if (ctx->Extensions.ARB_texture_float) { + *params = img->TexFormat->GreenBits ? img->TexFormat->DataType : GL_NONE; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + case GL_TEXTURE_BLUE_TYPE_ARB: + if (ctx->Extensions.ARB_texture_float) { + *params = img->TexFormat->BlueBits ? img->TexFormat->DataType : GL_NONE; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + case GL_TEXTURE_ALPHA_TYPE_ARB: + if (ctx->Extensions.ARB_texture_float) { + *params = img->TexFormat->AlphaBits ? img->TexFormat->DataType : GL_NONE; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + case GL_TEXTURE_LUMINANCE_TYPE_ARB: + if (ctx->Extensions.ARB_texture_float) { + *params = img->TexFormat->LuminanceBits ? img->TexFormat->DataType : GL_NONE; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + case GL_TEXTURE_INTENSITY_TYPE_ARB: + if (ctx->Extensions.ARB_texture_float) { + *params = img->TexFormat->IntensityBits ? img->TexFormat->DataType : GL_NONE; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + case GL_TEXTURE_DEPTH_TYPE_ARB: + if (ctx->Extensions.ARB_texture_float) { + *params = img->TexFormat->DepthBits ? img->TexFormat->DataType : GL_NONE; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + break; + + default: + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetTexLevelParameter[if]v(pname)"); + } + + out: + _mesa_unlock_texture(ctx, texObj); +} + + + +void GLAPIENTRY +_mesa_GetTexParameterfv( GLenum target, GLenum pname, GLfloat *params ) +{ + struct gl_texture_unit *texUnit; + struct gl_texture_object *obj; + GLboolean error = GL_FALSE; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glGetTexParameterfv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + obj = _mesa_select_tex_object(ctx, texUnit, target); + if (!obj) { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameterfv(target)"); + return; + } + + _mesa_lock_texture(ctx, obj); + switch (pname) { + case GL_TEXTURE_MAG_FILTER: + *params = ENUM_TO_FLOAT(obj->MagFilter); + break; + case GL_TEXTURE_MIN_FILTER: + *params = ENUM_TO_FLOAT(obj->MinFilter); + break; + case GL_TEXTURE_WRAP_S: + *params = ENUM_TO_FLOAT(obj->WrapS); + break; + case GL_TEXTURE_WRAP_T: + *params = ENUM_TO_FLOAT(obj->WrapT); + break; + case GL_TEXTURE_WRAP_R: + *params = ENUM_TO_FLOAT(obj->WrapR); + break; + case GL_TEXTURE_BORDER_COLOR: + params[0] = CLAMP(obj->BorderColor[0], 0.0F, 1.0F); + params[1] = CLAMP(obj->BorderColor[1], 0.0F, 1.0F); + params[2] = CLAMP(obj->BorderColor[2], 0.0F, 1.0F); + params[3] = CLAMP(obj->BorderColor[3], 0.0F, 1.0F); + break; + case GL_TEXTURE_RESIDENT: + { + GLboolean resident; + if (ctx->Driver.IsTextureResident) + resident = ctx->Driver.IsTextureResident(ctx, obj); + else + resident = GL_TRUE; + *params = ENUM_TO_FLOAT(resident); + } + break; + case GL_TEXTURE_PRIORITY: + *params = obj->Priority; + break; + case GL_TEXTURE_MIN_LOD: + *params = obj->MinLod; + break; + case GL_TEXTURE_MAX_LOD: + *params = obj->MaxLod; + break; + case GL_TEXTURE_BASE_LEVEL: + *params = (GLfloat) obj->BaseLevel; + break; + case GL_TEXTURE_MAX_LEVEL: + *params = (GLfloat) obj->MaxLevel; + break; + case GL_TEXTURE_MAX_ANISOTROPY_EXT: + if (ctx->Extensions.EXT_texture_filter_anisotropic) { + *params = obj->MaxAnisotropy; + } + else + error = 1; + break; + case GL_TEXTURE_COMPARE_SGIX: + if (ctx->Extensions.SGIX_shadow) { + *params = (GLfloat) obj->CompareFlag; + } + else + error = 1; + break; + case GL_TEXTURE_COMPARE_OPERATOR_SGIX: + if (ctx->Extensions.SGIX_shadow) { + *params = (GLfloat) obj->CompareOperator; + } + else + error = 1; + break; + case GL_SHADOW_AMBIENT_SGIX: /* aka GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ + if (ctx->Extensions.SGIX_shadow_ambient) { + *params = obj->ShadowAmbient; + } + else + error = 1; + break; + case GL_GENERATE_MIPMAP_SGIS: + if (ctx->Extensions.SGIS_generate_mipmap) { + *params = (GLfloat) obj->GenerateMipmap; + } + else + error = 1; + break; + case GL_TEXTURE_COMPARE_MODE_ARB: + if (ctx->Extensions.ARB_shadow) { + *params = (GLfloat) obj->CompareMode; + } + else + error = 1; + break; + case GL_TEXTURE_COMPARE_FUNC_ARB: + if (ctx->Extensions.ARB_shadow) { + *params = (GLfloat) obj->CompareFunc; + } + else + error = 1; + break; + case GL_DEPTH_TEXTURE_MODE_ARB: + if (ctx->Extensions.ARB_depth_texture) { + *params = (GLfloat) obj->DepthMode; + } + else + error = 1; + break; + case GL_TEXTURE_LOD_BIAS: + if (ctx->Extensions.EXT_texture_lod_bias) { + *params = obj->LodBias; + } + else + error = 1; + break; + default: + error = 1; + break; + } + if (error) + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameterfv(pname=0x%x)", + pname); + + _mesa_unlock_texture(ctx, obj); +} + + +void GLAPIENTRY +_mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params ) +{ + struct gl_texture_unit *texUnit; + struct gl_texture_object *obj; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glGetTexParameteriv(current unit)"); + return; + } + + texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; + + obj = _mesa_select_tex_object(ctx, texUnit, target); + if (!obj) { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameteriv(target)"); + return; + } + + switch (pname) { + case GL_TEXTURE_MAG_FILTER: + *params = (GLint) obj->MagFilter; + return; + case GL_TEXTURE_MIN_FILTER: + *params = (GLint) obj->MinFilter; + return; + case GL_TEXTURE_WRAP_S: + *params = (GLint) obj->WrapS; + return; + case GL_TEXTURE_WRAP_T: + *params = (GLint) obj->WrapT; + return; + case GL_TEXTURE_WRAP_R: + *params = (GLint) obj->WrapR; + return; + case GL_TEXTURE_BORDER_COLOR: + { + GLfloat b[4]; + b[0] = CLAMP(obj->BorderColor[0], 0.0F, 1.0F); + b[1] = CLAMP(obj->BorderColor[1], 0.0F, 1.0F); + b[2] = CLAMP(obj->BorderColor[2], 0.0F, 1.0F); + b[3] = CLAMP(obj->BorderColor[3], 0.0F, 1.0F); + params[0] = FLOAT_TO_INT(b[0]); + params[1] = FLOAT_TO_INT(b[1]); + params[2] = FLOAT_TO_INT(b[2]); + params[3] = FLOAT_TO_INT(b[3]); + } + return; + case GL_TEXTURE_RESIDENT: + { + GLboolean resident; + if (ctx->Driver.IsTextureResident) + resident = ctx->Driver.IsTextureResident(ctx, obj); + else + resident = GL_TRUE; + *params = (GLint) resident; + } + return; + case GL_TEXTURE_PRIORITY: + *params = FLOAT_TO_INT(obj->Priority); + return; + case GL_TEXTURE_MIN_LOD: + *params = (GLint) obj->MinLod; + return; + case GL_TEXTURE_MAX_LOD: + *params = (GLint) obj->MaxLod; + return; + case GL_TEXTURE_BASE_LEVEL: + *params = obj->BaseLevel; + return; + case GL_TEXTURE_MAX_LEVEL: + *params = obj->MaxLevel; + return; + case GL_TEXTURE_MAX_ANISOTROPY_EXT: + if (ctx->Extensions.EXT_texture_filter_anisotropic) { + *params = (GLint) obj->MaxAnisotropy; + return; + } + break; + case GL_TEXTURE_COMPARE_SGIX: + if (ctx->Extensions.SGIX_shadow) { + *params = (GLint) obj->CompareFlag; + return; + } + break; + case GL_TEXTURE_COMPARE_OPERATOR_SGIX: + if (ctx->Extensions.SGIX_shadow) { + *params = (GLint) obj->CompareOperator; + return; + } + break; + case GL_SHADOW_AMBIENT_SGIX: /* aka GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ + if (ctx->Extensions.SGIX_shadow_ambient) { + *params = (GLint) FLOAT_TO_INT(obj->ShadowAmbient); + return; + } + break; + case GL_GENERATE_MIPMAP_SGIS: + if (ctx->Extensions.SGIS_generate_mipmap) { + *params = (GLint) obj->GenerateMipmap; + return; + } + break; + case GL_TEXTURE_COMPARE_MODE_ARB: + if (ctx->Extensions.ARB_shadow) { + *params = (GLint) obj->CompareMode; + return; + } + break; + case GL_TEXTURE_COMPARE_FUNC_ARB: + if (ctx->Extensions.ARB_shadow) { + *params = (GLint) obj->CompareFunc; + return; + } + break; + case GL_DEPTH_TEXTURE_MODE_ARB: + if (ctx->Extensions.ARB_depth_texture) { + *params = (GLint) obj->DepthMode; + return; + } + break; + case GL_TEXTURE_LOD_BIAS: + if (ctx->Extensions.EXT_texture_lod_bias) { + *params = (GLint) obj->LodBias; + return; + } + break; + default: + ; /* silence warnings */ + } + /* If we get here, pname was an unrecognized enum */ + _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameteriv(pname=0x%x)", pname); +} diff --git a/src/mesa/main/texparam.h b/src/mesa/main/texparam.h new file mode 100644 index 0000000000..454b96350e --- /dev/null +++ b/src/mesa/main/texparam.h @@ -0,0 +1,63 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +#ifndef TEXPARAM_H +#define TEXPARAM_H + + +#include "main/glheader.h" + + +extern void GLAPIENTRY +_mesa_GetTexLevelParameterfv( GLenum target, GLint level, + GLenum pname, GLfloat *params ); + +extern void GLAPIENTRY +_mesa_GetTexLevelParameteriv( GLenum target, GLint level, + GLenum pname, GLint *params ); + +extern void GLAPIENTRY +_mesa_GetTexParameterfv( GLenum target, GLenum pname, GLfloat *params ); + +extern void GLAPIENTRY +_mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params ); + + + +extern void GLAPIENTRY +_mesa_TexParameterfv( GLenum target, GLenum pname, const GLfloat *params ); + +extern void GLAPIENTRY +_mesa_TexParameterf( GLenum target, GLenum pname, GLfloat param ); + + +extern void GLAPIENTRY +_mesa_TexParameteri( GLenum target, GLenum pname, GLint param ); + +extern void GLAPIENTRY +_mesa_TexParameteriv( GLenum target, GLenum pname, const GLint *params ); + + +#endif /* TEXPARAM_H */ diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c index 448fc53912..dbd769d740 100644 --- a/src/mesa/main/texstate.c +++ b/src/mesa/main/texstate.c @@ -321,1005 +321,6 @@ calculate_derived_texenv( struct gl_tex_env_combine_state *state, } -/**********************************************************************/ -/* Texture Parameters */ -/**********************************************************************/ - -/** - * Check if a coordinate wrap mode is supported for the texture target. - * \return GL_TRUE if legal, GL_FALSE otherwise - */ -static GLboolean -validate_texture_wrap_mode(GLcontext * ctx, GLenum target, GLenum wrap) -{ - const struct gl_extensions * const e = & ctx->Extensions; - - if (wrap == GL_CLAMP || wrap == GL_CLAMP_TO_EDGE || - (wrap == GL_CLAMP_TO_BORDER && e->ARB_texture_border_clamp)) { - /* any texture target */ - return GL_TRUE; - } - else if (target != GL_TEXTURE_RECTANGLE_NV && - (wrap == GL_REPEAT || - (wrap == GL_MIRRORED_REPEAT && - e->ARB_texture_mirrored_repeat) || - (wrap == GL_MIRROR_CLAMP_EXT && - (e->ATI_texture_mirror_once || e->EXT_texture_mirror_clamp)) || - (wrap == GL_MIRROR_CLAMP_TO_EDGE_EXT && - (e->ATI_texture_mirror_once || e->EXT_texture_mirror_clamp)) || - (wrap == GL_MIRROR_CLAMP_TO_BORDER_EXT && - (e->EXT_texture_mirror_clamp)))) { - /* non-rectangle texture */ - return GL_TRUE; - } - - _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" ); - return GL_FALSE; -} - - -void GLAPIENTRY -_mesa_TexParameterf( GLenum target, GLenum pname, GLfloat param ) -{ - _mesa_TexParameterfv(target, pname, ¶m); -} - - -void GLAPIENTRY -_mesa_TexParameterfv( GLenum target, GLenum pname, const GLfloat *params ) -{ - const GLenum eparam = (GLenum) (GLint) params[0]; - struct gl_texture_unit *texUnit; - struct gl_texture_object *texObj; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (MESA_VERBOSE&(VERBOSE_API|VERBOSE_TEXTURE)) - _mesa_debug(ctx, "glTexParameter %s %s %.1f(%s)...\n", - _mesa_lookup_enum_by_nr(target), - _mesa_lookup_enum_by_nr(pname), - *params, - _mesa_lookup_enum_by_nr(eparam)); - - if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glTexParameterfv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - switch (target) { - case GL_TEXTURE_1D: - texObj = texUnit->Current1D; - break; - case GL_TEXTURE_2D: - texObj = texUnit->Current2D; - break; - case GL_TEXTURE_3D: - texObj = texUnit->Current3D; - break; - case GL_TEXTURE_CUBE_MAP: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); - return; - } - texObj = texUnit->CurrentCubeMap; - break; - case GL_TEXTURE_RECTANGLE_NV: - if (!ctx->Extensions.NV_texture_rectangle) { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); - return; - } - texObj = texUnit->CurrentRect; - break; - case GL_TEXTURE_1D_ARRAY_EXT: - if (!ctx->Extensions.MESA_texture_array) { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); - return; - } - texObj = texUnit->Current1DArray; - break; - case GL_TEXTURE_2D_ARRAY_EXT: - if (!ctx->Extensions.MESA_texture_array) { - _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); - return; - } - texObj = texUnit->Current2DArray; - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glTexParameter(target)" ); - return; - } - - switch (pname) { - case GL_TEXTURE_MIN_FILTER: - /* A small optimization */ - if (texObj->MinFilter == eparam) - return; - if (eparam==GL_NEAREST || eparam==GL_LINEAR) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->MinFilter = eparam; - } - else if ((eparam==GL_NEAREST_MIPMAP_NEAREST || - eparam==GL_LINEAR_MIPMAP_NEAREST || - eparam==GL_NEAREST_MIPMAP_LINEAR || - eparam==GL_LINEAR_MIPMAP_LINEAR) && - texObj->Target != GL_TEXTURE_RECTANGLE_NV) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->MinFilter = eparam; - } - else { - _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" ); - return; - } - break; - case GL_TEXTURE_MAG_FILTER: - /* A small optimization */ - if (texObj->MagFilter == eparam) - return; - - if (eparam==GL_NEAREST || eparam==GL_LINEAR) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->MagFilter = eparam; - } - else { - _mesa_error( ctx, GL_INVALID_VALUE, "glTexParameter(param)" ); - return; - } - break; - case GL_TEXTURE_WRAP_S: - if (texObj->WrapS == eparam) - return; - if (validate_texture_wrap_mode(ctx, texObj->Target, eparam)) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->WrapS = eparam; - } - else { - return; - } - break; - case GL_TEXTURE_WRAP_T: - if (texObj->WrapT == eparam) - return; - if (validate_texture_wrap_mode(ctx, texObj->Target, eparam)) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->WrapT = eparam; - } - else { - return; - } - break; - case GL_TEXTURE_WRAP_R: - if (texObj->WrapR == eparam) - return; - if (validate_texture_wrap_mode(ctx, texObj->Target, eparam)) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->WrapR = eparam; - } - else { - return; - } - break; - case GL_TEXTURE_BORDER_COLOR: - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->BorderColor[RCOMP] = params[0]; - texObj->BorderColor[GCOMP] = params[1]; - texObj->BorderColor[BCOMP] = params[2]; - texObj->BorderColor[ACOMP] = params[3]; - UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[RCOMP], params[0]); - UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[GCOMP], params[1]); - UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[BCOMP], params[2]); - UNCLAMPED_FLOAT_TO_CHAN(texObj->_BorderChan[ACOMP], params[3]); - break; - case GL_TEXTURE_MIN_LOD: - if (texObj->MinLod == params[0]) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->MinLod = params[0]; - break; - case GL_TEXTURE_MAX_LOD: - if (texObj->MaxLod == params[0]) - return; - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->MaxLod = params[0]; - break; - case GL_TEXTURE_BASE_LEVEL: - if (params[0] < 0.0) { - _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)"); - return; - } - if (target == GL_TEXTURE_RECTANGLE_ARB && params[0] != 0.0) { - _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)"); - return; - } - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->BaseLevel = (GLint) params[0]; - break; - case GL_TEXTURE_MAX_LEVEL: - if (params[0] < 0.0) { - _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)"); - return; - } - if (target == GL_TEXTURE_RECTANGLE_ARB) { - _mesa_error(ctx, GL_INVALID_OPERATION, "glTexParameter(param)"); - return; - } - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->MaxLevel = (GLint) params[0]; - break; - case GL_TEXTURE_PRIORITY: - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->Priority = CLAMP( params[0], 0.0F, 1.0F ); - break; - case GL_TEXTURE_MAX_ANISOTROPY_EXT: - if (ctx->Extensions.EXT_texture_filter_anisotropic) { - if (params[0] < 1.0) { - _mesa_error(ctx, GL_INVALID_VALUE, "glTexParameter(param)" ); - return; - } - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - /* clamp to max, that's what NVIDIA does */ - texObj->MaxAnisotropy = MIN2(params[0], - ctx->Const.MaxTextureMaxAnisotropy); - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(pname=GL_TEXTURE_MAX_ANISOTROPY_EXT)"); - return; - } - break; - case GL_TEXTURE_COMPARE_SGIX: - if (ctx->Extensions.SGIX_shadow) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->CompareFlag = params[0] ? GL_TRUE : GL_FALSE; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(pname=GL_TEXTURE_COMPARE_SGIX)"); - return; - } - break; - case GL_TEXTURE_COMPARE_OPERATOR_SGIX: - if (ctx->Extensions.SGIX_shadow) { - GLenum op = (GLenum) params[0]; - if (op == GL_TEXTURE_LEQUAL_R_SGIX || - op == GL_TEXTURE_GEQUAL_R_SGIX) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->CompareOperator = op; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glTexParameter(param)"); - } - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(pname=GL_TEXTURE_COMPARE_OPERATOR_SGIX)"); - return; - } - break; - case GL_SHADOW_AMBIENT_SGIX: /* aka GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ - if (ctx->Extensions.SGIX_shadow_ambient) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->ShadowAmbient = CLAMP(params[0], 0.0F, 1.0F); - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(pname=GL_SHADOW_AMBIENT_SGIX)"); - return; - } - break; - case GL_GENERATE_MIPMAP_SGIS: - if (ctx->Extensions.SGIS_generate_mipmap) { - texObj->GenerateMipmap = params[0] ? GL_TRUE : GL_FALSE; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(pname=GL_GENERATE_MIPMAP_SGIS)"); - return; - } - break; - case GL_TEXTURE_COMPARE_MODE_ARB: - if (ctx->Extensions.ARB_shadow) { - const GLenum mode = (GLenum) params[0]; - if (mode == GL_NONE || mode == GL_COMPARE_R_TO_TEXTURE_ARB) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->CompareMode = mode; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(bad GL_TEXTURE_COMPARE_MODE_ARB: 0x%x)", mode); - return; - } - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(pname=GL_TEXTURE_COMPARE_MODE_ARB)"); - return; - } - break; - case GL_TEXTURE_COMPARE_FUNC_ARB: - if (ctx->Extensions.ARB_shadow) { - const GLenum func = (GLenum) params[0]; - if (func == GL_LEQUAL || func == GL_GEQUAL) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->CompareFunc = func; - } - else if (ctx->Extensions.EXT_shadow_funcs && - (func == GL_EQUAL || - func == GL_NOTEQUAL || - func == GL_LESS || - func == GL_GREATER || - func == GL_ALWAYS || - func == GL_NEVER)) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->CompareFunc = func; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(bad GL_TEXTURE_COMPARE_FUNC_ARB)"); - return; - } - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(pname=GL_TEXTURE_COMPARE_FUNC_ARB)"); - return; - } - break; - case GL_DEPTH_TEXTURE_MODE_ARB: - if (ctx->Extensions.ARB_depth_texture) { - const GLenum result = (GLenum) params[0]; - if (result == GL_LUMINANCE || result == GL_INTENSITY - || result == GL_ALPHA) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->DepthMode = result; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(bad GL_DEPTH_TEXTURE_MODE_ARB)"); - return; - } - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(pname=GL_DEPTH_TEXTURE_MODE_ARB)"); - return; - } - break; - case GL_TEXTURE_LOD_BIAS: - /* NOTE: this is really part of OpenGL 1.4, not EXT_texture_lod_bias*/ - if (ctx->Extensions.EXT_texture_lod_bias) { - if (texObj->LodBias != params[0]) { - FLUSH_VERTICES(ctx, _NEW_TEXTURE); - texObj->LodBias = params[0]; - } - } - break; - - default: - _mesa_error(ctx, GL_INVALID_ENUM, - "glTexParameter(pname=0x%x)", pname); - return; - } - - texObj->_Complete = GL_FALSE; - - if (ctx->Driver.TexParameter) { - (*ctx->Driver.TexParameter)( ctx, target, texObj, pname, params ); - } -} - - -void GLAPIENTRY -_mesa_TexParameteri( GLenum target, GLenum pname, GLint param ) -{ - GLfloat fparam[4]; - if (pname == GL_TEXTURE_PRIORITY) - fparam[0] = INT_TO_FLOAT(param); - else - fparam[0] = (GLfloat) param; - fparam[1] = fparam[2] = fparam[3] = 0.0; - _mesa_TexParameterfv(target, pname, fparam); -} - - -void GLAPIENTRY -_mesa_TexParameteriv( GLenum target, GLenum pname, const GLint *params ) -{ - GLfloat fparam[4]; - if (pname == GL_TEXTURE_BORDER_COLOR) { - fparam[0] = INT_TO_FLOAT(params[0]); - fparam[1] = INT_TO_FLOAT(params[1]); - fparam[2] = INT_TO_FLOAT(params[2]); - fparam[3] = INT_TO_FLOAT(params[3]); - } - else { - if (pname == GL_TEXTURE_PRIORITY) - fparam[0] = INT_TO_FLOAT(params[0]); - else - fparam[0] = (GLfloat) params[0]; - fparam[1] = fparam[2] = fparam[3] = 0.0F; - } - _mesa_TexParameterfv(target, pname, fparam); -} - - -void GLAPIENTRY -_mesa_GetTexLevelParameterfv( GLenum target, GLint level, - GLenum pname, GLfloat *params ) -{ - GLint iparam; - _mesa_GetTexLevelParameteriv( target, level, pname, &iparam ); - *params = (GLfloat) iparam; -} - - -static GLuint -tex_image_dimensions(GLcontext *ctx, GLenum target) -{ - switch (target) { - case GL_TEXTURE_1D: - case GL_PROXY_TEXTURE_1D: - return 1; - case GL_TEXTURE_2D: - case GL_PROXY_TEXTURE_2D: - return 2; - case GL_TEXTURE_3D: - case GL_PROXY_TEXTURE_3D: - return 3; - case GL_TEXTURE_CUBE_MAP: - case GL_PROXY_TEXTURE_CUBE_MAP: - case GL_TEXTURE_CUBE_MAP_POSITIVE_X: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_X: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Y: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y: - case GL_TEXTURE_CUBE_MAP_POSITIVE_Z: - case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z: - return ctx->Extensions.ARB_texture_cube_map ? 2 : 0; - case GL_TEXTURE_RECTANGLE_NV: - case GL_PROXY_TEXTURE_RECTANGLE_NV: - return ctx->Extensions.NV_texture_rectangle ? 2 : 0; - case GL_TEXTURE_1D_ARRAY_EXT: - case GL_PROXY_TEXTURE_1D_ARRAY_EXT: - return ctx->Extensions.MESA_texture_array ? 2 : 0; - case GL_TEXTURE_2D_ARRAY_EXT: - case GL_PROXY_TEXTURE_2D_ARRAY_EXT: - return ctx->Extensions.MESA_texture_array ? 3 : 0; - default: - _mesa_problem(ctx, "bad target in _mesa_tex_target_dimensions()"); - return 0; - } -} - - -void GLAPIENTRY -_mesa_GetTexLevelParameteriv( GLenum target, GLint level, - GLenum pname, GLint *params ) -{ - const struct gl_texture_unit *texUnit; - struct gl_texture_object *texObj; - const struct gl_texture_image *img = NULL; - GLuint dimensions; - GLboolean isProxy; - GLint maxLevels; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glGetTexLevelParameteriv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - /* this will catch bad target values */ - dimensions = tex_image_dimensions(ctx, target); /* 1, 2 or 3 */ - if (dimensions == 0) { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexLevelParameter[if]v(target)"); - return; - } - - maxLevels = _mesa_max_texture_levels(ctx, target); - if (maxLevels == 0) { - /* should not happen since was just checked above */ - _mesa_problem(ctx, "maxLevels=0 in _mesa_GetTexLevelParameter"); - return; - } - - if (level < 0 || level >= maxLevels) { - _mesa_error( ctx, GL_INVALID_VALUE, "glGetTexLevelParameter[if]v" ); - return; - } - - texObj = _mesa_select_tex_object(ctx, texUnit, target); - _mesa_lock_texture(ctx, texObj); - - img = _mesa_select_tex_image(ctx, texObj, target, level); - if (!img || !img->TexFormat) { - /* undefined texture image */ - if (pname == GL_TEXTURE_COMPONENTS) - *params = 1; - else - *params = 0; - goto out; - } - - isProxy = _mesa_is_proxy_texture(target); - - switch (pname) { - case GL_TEXTURE_WIDTH: - *params = img->Width; - break; - case GL_TEXTURE_HEIGHT: - *params = img->Height; - break; - case GL_TEXTURE_DEPTH: - *params = img->Depth; - break; - case GL_TEXTURE_INTERNAL_FORMAT: - *params = img->InternalFormat; - break; - case GL_TEXTURE_BORDER: - *params = img->Border; - break; - case GL_TEXTURE_RED_SIZE: - if (img->_BaseFormat == GL_RGB || img->_BaseFormat == GL_RGBA) - *params = img->TexFormat->RedBits; - else - *params = 0; - break; - case GL_TEXTURE_GREEN_SIZE: - if (img->_BaseFormat == GL_RGB || img->_BaseFormat == GL_RGBA) - *params = img->TexFormat->GreenBits; - else - *params = 0; - break; - case GL_TEXTURE_BLUE_SIZE: - if (img->_BaseFormat == GL_RGB || img->_BaseFormat == GL_RGBA) - *params = img->TexFormat->BlueBits; - else - *params = 0; - break; - case GL_TEXTURE_ALPHA_SIZE: - if (img->_BaseFormat == GL_ALPHA || - img->_BaseFormat == GL_LUMINANCE_ALPHA || - img->_BaseFormat == GL_RGBA) - *params = img->TexFormat->AlphaBits; - else - *params = 0; - break; - case GL_TEXTURE_INTENSITY_SIZE: - if (img->_BaseFormat != GL_INTENSITY) - *params = 0; - else if (img->TexFormat->IntensityBits > 0) - *params = img->TexFormat->IntensityBits; - else /* intensity probably stored as rgb texture */ - *params = MIN2(img->TexFormat->RedBits, img->TexFormat->GreenBits); - break; - case GL_TEXTURE_LUMINANCE_SIZE: - if (img->_BaseFormat != GL_LUMINANCE && - img->_BaseFormat != GL_LUMINANCE_ALPHA) - *params = 0; - else if (img->TexFormat->LuminanceBits > 0) - *params = img->TexFormat->LuminanceBits; - else /* luminance probably stored as rgb texture */ - *params = MIN2(img->TexFormat->RedBits, img->TexFormat->GreenBits); - break; - case GL_TEXTURE_INDEX_SIZE_EXT: - if (img->_BaseFormat == GL_COLOR_INDEX) - *params = img->TexFormat->IndexBits; - else - *params = 0; - break; - case GL_TEXTURE_DEPTH_SIZE_ARB: - if (ctx->Extensions.SGIX_depth_texture || - ctx->Extensions.ARB_depth_texture) - *params = img->TexFormat->DepthBits; - else - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - break; - case GL_TEXTURE_STENCIL_SIZE_EXT: - if (ctx->Extensions.EXT_packed_depth_stencil) { - *params = img->TexFormat->StencilBits; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - - /* GL_ARB_texture_compression */ - case GL_TEXTURE_COMPRESSED_IMAGE_SIZE: - if (ctx->Extensions.ARB_texture_compression) { - if (img->IsCompressed && !isProxy) { - /* Don't use ctx->Driver.CompressedTextureSize() since that - * may returned a padded hardware size. - */ - *params = _mesa_compressed_texture_size(ctx, img->Width, - img->Height, img->Depth, - img->TexFormat->MesaFormat); - } - else { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glGetTexLevelParameter[if]v(pname)"); - } - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - case GL_TEXTURE_COMPRESSED: - if (ctx->Extensions.ARB_texture_compression) { - *params = (GLint) img->IsCompressed; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - - /* GL_ARB_texture_float */ - case GL_TEXTURE_RED_TYPE_ARB: - if (ctx->Extensions.ARB_texture_float) { - *params = img->TexFormat->RedBits ? img->TexFormat->DataType : GL_NONE; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - case GL_TEXTURE_GREEN_TYPE_ARB: - if (ctx->Extensions.ARB_texture_float) { - *params = img->TexFormat->GreenBits ? img->TexFormat->DataType : GL_NONE; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - case GL_TEXTURE_BLUE_TYPE_ARB: - if (ctx->Extensions.ARB_texture_float) { - *params = img->TexFormat->BlueBits ? img->TexFormat->DataType : GL_NONE; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - case GL_TEXTURE_ALPHA_TYPE_ARB: - if (ctx->Extensions.ARB_texture_float) { - *params = img->TexFormat->AlphaBits ? img->TexFormat->DataType : GL_NONE; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - case GL_TEXTURE_LUMINANCE_TYPE_ARB: - if (ctx->Extensions.ARB_texture_float) { - *params = img->TexFormat->LuminanceBits ? img->TexFormat->DataType : GL_NONE; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - case GL_TEXTURE_INTENSITY_TYPE_ARB: - if (ctx->Extensions.ARB_texture_float) { - *params = img->TexFormat->IntensityBits ? img->TexFormat->DataType : GL_NONE; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - case GL_TEXTURE_DEPTH_TYPE_ARB: - if (ctx->Extensions.ARB_texture_float) { - *params = img->TexFormat->DepthBits ? img->TexFormat->DataType : GL_NONE; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - break; - - default: - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetTexLevelParameter[if]v(pname)"); - } - - out: - _mesa_unlock_texture(ctx, texObj); -} - - - -void GLAPIENTRY -_mesa_GetTexParameterfv( GLenum target, GLenum pname, GLfloat *params ) -{ - struct gl_texture_unit *texUnit; - struct gl_texture_object *obj; - GLboolean error = GL_FALSE; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glGetTexParameterfv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - obj = _mesa_select_tex_object(ctx, texUnit, target); - if (!obj) { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameterfv(target)"); - return; - } - - _mesa_lock_texture(ctx, obj); - switch (pname) { - case GL_TEXTURE_MAG_FILTER: - *params = ENUM_TO_FLOAT(obj->MagFilter); - break; - case GL_TEXTURE_MIN_FILTER: - *params = ENUM_TO_FLOAT(obj->MinFilter); - break; - case GL_TEXTURE_WRAP_S: - *params = ENUM_TO_FLOAT(obj->WrapS); - break; - case GL_TEXTURE_WRAP_T: - *params = ENUM_TO_FLOAT(obj->WrapT); - break; - case GL_TEXTURE_WRAP_R: - *params = ENUM_TO_FLOAT(obj->WrapR); - break; - case GL_TEXTURE_BORDER_COLOR: - params[0] = CLAMP(obj->BorderColor[0], 0.0F, 1.0F); - params[1] = CLAMP(obj->BorderColor[1], 0.0F, 1.0F); - params[2] = CLAMP(obj->BorderColor[2], 0.0F, 1.0F); - params[3] = CLAMP(obj->BorderColor[3], 0.0F, 1.0F); - break; - case GL_TEXTURE_RESIDENT: - { - GLboolean resident; - if (ctx->Driver.IsTextureResident) - resident = ctx->Driver.IsTextureResident(ctx, obj); - else - resident = GL_TRUE; - *params = ENUM_TO_FLOAT(resident); - } - break; - case GL_TEXTURE_PRIORITY: - *params = obj->Priority; - break; - case GL_TEXTURE_MIN_LOD: - *params = obj->MinLod; - break; - case GL_TEXTURE_MAX_LOD: - *params = obj->MaxLod; - break; - case GL_TEXTURE_BASE_LEVEL: - *params = (GLfloat) obj->BaseLevel; - break; - case GL_TEXTURE_MAX_LEVEL: - *params = (GLfloat) obj->MaxLevel; - break; - case GL_TEXTURE_MAX_ANISOTROPY_EXT: - if (ctx->Extensions.EXT_texture_filter_anisotropic) { - *params = obj->MaxAnisotropy; - } - else - error = 1; - break; - case GL_TEXTURE_COMPARE_SGIX: - if (ctx->Extensions.SGIX_shadow) { - *params = (GLfloat) obj->CompareFlag; - } - else - error = 1; - break; - case GL_TEXTURE_COMPARE_OPERATOR_SGIX: - if (ctx->Extensions.SGIX_shadow) { - *params = (GLfloat) obj->CompareOperator; - } - else - error = 1; - break; - case GL_SHADOW_AMBIENT_SGIX: /* aka GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ - if (ctx->Extensions.SGIX_shadow_ambient) { - *params = obj->ShadowAmbient; - } - else - error = 1; - break; - case GL_GENERATE_MIPMAP_SGIS: - if (ctx->Extensions.SGIS_generate_mipmap) { - *params = (GLfloat) obj->GenerateMipmap; - } - else - error = 1; - break; - case GL_TEXTURE_COMPARE_MODE_ARB: - if (ctx->Extensions.ARB_shadow) { - *params = (GLfloat) obj->CompareMode; - } - else - error = 1; - break; - case GL_TEXTURE_COMPARE_FUNC_ARB: - if (ctx->Extensions.ARB_shadow) { - *params = (GLfloat) obj->CompareFunc; - } - else - error = 1; - break; - case GL_DEPTH_TEXTURE_MODE_ARB: - if (ctx->Extensions.ARB_depth_texture) { - *params = (GLfloat) obj->DepthMode; - } - else - error = 1; - break; - case GL_TEXTURE_LOD_BIAS: - if (ctx->Extensions.EXT_texture_lod_bias) { - *params = obj->LodBias; - } - else - error = 1; - break; - default: - error = 1; - break; - } - if (error) - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameterfv(pname=0x%x)", - pname); - - _mesa_unlock_texture(ctx, obj); -} - - -void GLAPIENTRY -_mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params ) -{ - struct gl_texture_unit *texUnit; - struct gl_texture_object *obj; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (ctx->Texture.CurrentUnit >= ctx->Const.MaxTextureImageUnits) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glGetTexParameteriv(current unit)"); - return; - } - - texUnit = &ctx->Texture.Unit[ctx->Texture.CurrentUnit]; - - obj = _mesa_select_tex_object(ctx, texUnit, target); - if (!obj) { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameteriv(target)"); - return; - } - - switch (pname) { - case GL_TEXTURE_MAG_FILTER: - *params = (GLint) obj->MagFilter; - return; - case GL_TEXTURE_MIN_FILTER: - *params = (GLint) obj->MinFilter; - return; - case GL_TEXTURE_WRAP_S: - *params = (GLint) obj->WrapS; - return; - case GL_TEXTURE_WRAP_T: - *params = (GLint) obj->WrapT; - return; - case GL_TEXTURE_WRAP_R: - *params = (GLint) obj->WrapR; - return; - case GL_TEXTURE_BORDER_COLOR: - { - GLfloat b[4]; - b[0] = CLAMP(obj->BorderColor[0], 0.0F, 1.0F); - b[1] = CLAMP(obj->BorderColor[1], 0.0F, 1.0F); - b[2] = CLAMP(obj->BorderColor[2], 0.0F, 1.0F); - b[3] = CLAMP(obj->BorderColor[3], 0.0F, 1.0F); - params[0] = FLOAT_TO_INT(b[0]); - params[1] = FLOAT_TO_INT(b[1]); - params[2] = FLOAT_TO_INT(b[2]); - params[3] = FLOAT_TO_INT(b[3]); - } - return; - case GL_TEXTURE_RESIDENT: - { - GLboolean resident; - if (ctx->Driver.IsTextureResident) - resident = ctx->Driver.IsTextureResident(ctx, obj); - else - resident = GL_TRUE; - *params = (GLint) resident; - } - return; - case GL_TEXTURE_PRIORITY: - *params = FLOAT_TO_INT(obj->Priority); - return; - case GL_TEXTURE_MIN_LOD: - *params = (GLint) obj->MinLod; - return; - case GL_TEXTURE_MAX_LOD: - *params = (GLint) obj->MaxLod; - return; - case GL_TEXTURE_BASE_LEVEL: - *params = obj->BaseLevel; - return; - case GL_TEXTURE_MAX_LEVEL: - *params = obj->MaxLevel; - return; - case GL_TEXTURE_MAX_ANISOTROPY_EXT: - if (ctx->Extensions.EXT_texture_filter_anisotropic) { - *params = (GLint) obj->MaxAnisotropy; - return; - } - break; - case GL_TEXTURE_COMPARE_SGIX: - if (ctx->Extensions.SGIX_shadow) { - *params = (GLint) obj->CompareFlag; - return; - } - break; - case GL_TEXTURE_COMPARE_OPERATOR_SGIX: - if (ctx->Extensions.SGIX_shadow) { - *params = (GLint) obj->CompareOperator; - return; - } - break; - case GL_SHADOW_AMBIENT_SGIX: /* aka GL_TEXTURE_COMPARE_FAIL_VALUE_ARB */ - if (ctx->Extensions.SGIX_shadow_ambient) { - *params = (GLint) FLOAT_TO_INT(obj->ShadowAmbient); - return; - } - break; - case GL_GENERATE_MIPMAP_SGIS: - if (ctx->Extensions.SGIS_generate_mipmap) { - *params = (GLint) obj->GenerateMipmap; - return; - } - break; - case GL_TEXTURE_COMPARE_MODE_ARB: - if (ctx->Extensions.ARB_shadow) { - *params = (GLint) obj->CompareMode; - return; - } - break; - case GL_TEXTURE_COMPARE_FUNC_ARB: - if (ctx->Extensions.ARB_shadow) { - *params = (GLint) obj->CompareFunc; - return; - } - break; - case GL_DEPTH_TEXTURE_MODE_ARB: - if (ctx->Extensions.ARB_depth_texture) { - *params = (GLint) obj->DepthMode; - return; - } - break; - case GL_TEXTURE_LOD_BIAS: - if (ctx->Extensions.EXT_texture_lod_bias) { - *params = (GLint) obj->LodBias; - return; - } - break; - default: - ; /* silence warnings */ - } - /* If we get here, pname was an unrecognized enum */ - _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexParameteriv(pname=0x%x)", pname); -} - /* GL_ARB_multitexture */ diff --git a/src/mesa/main/texstate.h b/src/mesa/main/texstate.h index 1f1c2267aa..b5003d5d6e 100644 --- a/src/mesa/main/texstate.h +++ b/src/mesa/main/texstate.h @@ -48,35 +48,6 @@ _mesa_print_texunit_state( GLcontext *ctx, GLuint unit ); */ /*@{*/ -extern void GLAPIENTRY -_mesa_GetTexLevelParameterfv( GLenum target, GLint level, - GLenum pname, GLfloat *params ); - -extern void GLAPIENTRY -_mesa_GetTexLevelParameteriv( GLenum target, GLint level, - GLenum pname, GLint *params ); - -extern void GLAPIENTRY -_mesa_GetTexParameterfv( GLenum target, GLenum pname, GLfloat *params ); - -extern void GLAPIENTRY -_mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params ); - - - -extern void GLAPIENTRY -_mesa_TexParameterfv( GLenum target, GLenum pname, const GLfloat *params ); - -extern void GLAPIENTRY -_mesa_TexParameterf( GLenum target, GLenum pname, GLfloat param ); - - -extern void GLAPIENTRY -_mesa_TexParameteri( GLenum target, GLenum pname, GLint param ); - -extern void GLAPIENTRY -_mesa_TexParameteriv( GLenum target, GLenum pname, const GLint *params ); - /* * GL_ARB_multitexture diff --git a/src/mesa/sources b/src/mesa/sources index d225d7d15b..8c000087c6 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -67,6 +67,7 @@ MAIN_SOURCES = \ main/texgen.c \ main/teximage.c \ main/texobj.c \ + main/texparam.c \ main/texrender.c \ main/texstate.c \ main/texstore.c \ -- cgit v1.2.3 From 222c80893a738b964dc75796d47545d88e6a6973 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 11 Jun 2008 20:13:07 -0600 Subject: mesa: refactor: move various ENUM_TO_x macros into macros.h --- src/mesa/main/get.c | 56 ++++++++++++++++++++++++++++++++++++++++++------ src/mesa/main/get_gen.py | 4 ---- src/mesa/main/macros.h | 7 ++++++ src/mesa/main/texenv.c | 2 -- src/mesa/main/texgen.c | 4 ---- src/mesa/main/texparam.c | 3 --- src/mesa/main/texstate.c | 3 --- 7 files changed, 57 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c index ee48b78318..ac437a86e2 100644 --- a/src/mesa/main/get.c +++ b/src/mesa/main/get.c @@ -19,10 +19,6 @@ #define INT_TO_BOOLEAN(I) ( (I) ? GL_TRUE : GL_FALSE ) -#define ENUM_TO_BOOLEAN(E) ( (E) ? GL_TRUE : GL_FALSE ) -#define ENUM_TO_INT(E) ( (GLint) (E) ) -#define ENUM_TO_FLOAT(E) ( (GLfloat) (E) ) - #define BOOLEAN_TO_INT(B) ( (GLint) (B) ) #define BOOLEAN_TO_FLOAT(B) ( (B) ? 1.0F : 0.0F ) @@ -870,6 +866,14 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params ) case GL_TEXTURE_3D: params[0] = _mesa_IsEnabled(GL_TEXTURE_3D); break; + case GL_TEXTURE_1D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetBooleanv"); + params[0] = _mesa_IsEnabled(GL_TEXTURE_1D_ARRAY_EXT); + break; + case GL_TEXTURE_2D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetBooleanv"); + params[0] = _mesa_IsEnabled(GL_TEXTURE_2D_ARRAY_EXT); + break; case GL_TEXTURE_BINDING_1D: params[0] = INT_TO_BOOLEAN(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current1D->Name); break; @@ -879,6 +883,14 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params ) case GL_TEXTURE_BINDING_3D: params[0] = INT_TO_BOOLEAN(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current3D->Name); break; + case GL_TEXTURE_BINDING_1D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetBooleanv"); + params[0] = INT_TO_BOOLEAN(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current1DArray->Name); + break; + case GL_TEXTURE_BINDING_2D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetBooleanv"); + params[0] = INT_TO_BOOLEAN(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current2DArray->Name); + break; case GL_TEXTURE_ENV_COLOR: { const GLfloat *color = ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvColor; @@ -2124,7 +2136,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) params[0] = (GLfloat)(ctx->DrawBuffer->Visual.depthBits); break; case GL_DEPTH_CLEAR_VALUE: - params[0] = (GLfloat)ctx->Depth.Clear; + params[0] = ctx->Depth.Clear; break; case GL_DEPTH_FUNC: params[0] = ENUM_TO_FLOAT(ctx->Depth.Func); @@ -2701,6 +2713,14 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) case GL_TEXTURE_3D: params[0] = BOOLEAN_TO_FLOAT(_mesa_IsEnabled(GL_TEXTURE_3D)); break; + case GL_TEXTURE_1D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetFloatv"); + params[0] = BOOLEAN_TO_FLOAT(_mesa_IsEnabled(GL_TEXTURE_1D_ARRAY_EXT)); + break; + case GL_TEXTURE_2D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetFloatv"); + params[0] = BOOLEAN_TO_FLOAT(_mesa_IsEnabled(GL_TEXTURE_2D_ARRAY_EXT)); + break; case GL_TEXTURE_BINDING_1D: params[0] = (GLfloat)(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current1D->Name); break; @@ -2710,6 +2730,14 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) case GL_TEXTURE_BINDING_3D: params[0] = (GLfloat)(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current3D->Name); break; + case GL_TEXTURE_BINDING_1D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetFloatv"); + params[0] = (GLfloat)(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current1DArray->Name); + break; + case GL_TEXTURE_BINDING_2D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetFloatv"); + params[0] = (GLfloat)(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current2DArray->Name); + break; case GL_TEXTURE_ENV_COLOR: { const GLfloat *color = ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvColor; @@ -2914,7 +2942,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) GLuint i, n = _mesa_get_compressed_formats(ctx, formats, GL_FALSE); ASSERT(n <= 100); for (i = 0; i < n; i++) - params[i] = (GLfloat)(ENUM_TO_INT(formats[i])); + params[i] = ENUM_TO_INT(formats[i]); } break; case GL_ARRAY_ELEMENT_LOCK_FIRST_EXT: @@ -4532,6 +4560,14 @@ _mesa_GetIntegerv( GLenum pname, GLint *params ) case GL_TEXTURE_3D: params[0] = BOOLEAN_TO_INT(_mesa_IsEnabled(GL_TEXTURE_3D)); break; + case GL_TEXTURE_1D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetIntegerv"); + params[0] = BOOLEAN_TO_INT(_mesa_IsEnabled(GL_TEXTURE_1D_ARRAY_EXT)); + break; + case GL_TEXTURE_2D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetIntegerv"); + params[0] = BOOLEAN_TO_INT(_mesa_IsEnabled(GL_TEXTURE_2D_ARRAY_EXT)); + break; case GL_TEXTURE_BINDING_1D: params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current1D->Name; break; @@ -4541,6 +4577,14 @@ _mesa_GetIntegerv( GLenum pname, GLint *params ) case GL_TEXTURE_BINDING_3D: params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current3D->Name; break; + case GL_TEXTURE_BINDING_1D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetIntegerv"); + params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current1DArray->Name; + break; + case GL_TEXTURE_BINDING_2D_ARRAY_EXT: + CHECK_EXT1(MESA_texture_array, "GetIntegerv"); + params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current2DArray->Name; + break; case GL_TEXTURE_ENV_COLOR: { const GLfloat *color = ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvColor; diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py index 9055433281..c307058da6 100644 --- a/src/mesa/main/get_gen.py +++ b/src/mesa/main/get_gen.py @@ -1127,10 +1127,6 @@ def EmitHeader(): #define INT_TO_BOOLEAN(I) ( (I) ? GL_TRUE : GL_FALSE ) -#define ENUM_TO_BOOLEAN(E) ( (E) ? GL_TRUE : GL_FALSE ) -#define ENUM_TO_INT(E) ( (GLint) (E) ) -#define ENUM_TO_FLOAT(E) ( (GLfloat) (E) ) - #define BOOLEAN_TO_INT(B) ( (GLint) (B) ) #define BOOLEAN_TO_FLOAT(B) ( (B) ? 1.0F : 0.0F ) diff --git a/src/mesa/main/macros.h b/src/mesa/main/macros.h index fbbcd4e269..2630855a0e 100644 --- a/src/mesa/main/macros.h +++ b/src/mesa/main/macros.h @@ -657,4 +657,11 @@ do { \ #define LEN_SQUARED_2FV( V ) ((V)[0]*(V)[0]+(V)[1]*(V)[1]) +/** casts to silence warnings with some compilers */ +#define ENUM_TO_INT(E) ((GLint)(E)) +#define ENUM_TO_FLOAT(E) ((GLfloat)(GLint)(E)) +#define ENUM_TO_DOUBLE(E) ((GLdouble)(GLint)(E)) +#define ENUM_TO_BOOLEAN(E) ((E) ? GL_TRUE : GL_FALSE) + + #endif diff --git a/src/mesa/main/texenv.c b/src/mesa/main/texenv.c index 2b5b15524e..e072cea136 100644 --- a/src/mesa/main/texenv.c +++ b/src/mesa/main/texenv.c @@ -37,8 +37,6 @@ #include "math/m_xform.h" -#define ENUM_TO_FLOAT(X) ((GLfloat)(GLint)(X)) - void GLAPIENTRY _mesa_TexEnvfv( GLenum target, GLenum pname, const GLfloat *param ) diff --git a/src/mesa/main/texgen.c b/src/mesa/main/texgen.c index 1018c3037b..244c7aaafc 100644 --- a/src/mesa/main/texgen.c +++ b/src/mesa/main/texgen.c @@ -37,10 +37,6 @@ #include "math/m_xform.h" -#define ENUM_TO_FLOAT(X) ((GLfloat)(GLint)(X)) -#define ENUM_TO_DOUBLE(X) ((GLdouble)(GLint)(X)) - - void GLAPIENTRY _mesa_TexGenfv( GLenum coord, GLenum pname, const GLfloat *params ) diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c index 10ad0ef1d8..af288c4e18 100644 --- a/src/mesa/main/texparam.c +++ b/src/mesa/main/texparam.c @@ -39,9 +39,6 @@ #include "main/teximage.h" -#define ENUM_TO_FLOAT(X) ((GLfloat)(GLint)(X)) - - /** * Check if a coordinate wrap mode is supported for the texture target. * \return GL_TRUE if legal, GL_FALSE otherwise diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c index dbd769d740..421f912849 100644 --- a/src/mesa/main/texstate.c +++ b/src/mesa/main/texstate.c @@ -45,9 +45,6 @@ #include "math/m_xform.h" -#define ENUM_TO_FLOAT(X) ((GLfloat)(GLint)(X)) -#define ENUM_TO_DOUBLE(X) ((GLdouble)(GLint)(X)) - /** * Default texture combine environment state. This is used to initialize -- cgit v1.2.3 From 11461f52be35d12bfe59ce0e9271cde8f118ed8e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 11 Jun 2008 20:16:16 -0600 Subject: mesa: refactor: update SConscript with recently added new source files --- src/mesa/SConscript | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index 1beb81c595..ff5bf60215 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -24,6 +24,7 @@ if env['platform'] != 'winddk': main_sources = [ 'main/api_arrayelt.c', + 'main/api_exec.c' 'main/api_loopback.c', 'main/api_noop.c', 'main/api_validate.c', @@ -33,6 +34,7 @@ if env['platform'] != 'winddk': 'main/blend.c', 'main/bufferobj.c', 'main/buffers.c', + 'main/clear.c', 'main/clip.c', 'main/colortab.c', 'main/context.c', @@ -64,23 +66,30 @@ if env['platform'] != 'winddk': 'main/matrix.c', 'main/mipmap.c', 'main/mm.c', + 'main/multisample.c', 'main/pixel.c', + 'main/pixelstore.c', 'main/points.c', 'main/polygon.c', 'main/queryobj.c', 'main/rastpos.c', 'main/rbadaptors.c', + 'main/readpix.c', 'main/renderbuffer.c', + 'main/scissor.c', 'main/shaders.c', 'main/state.c', 'main/stencil.c', 'main/texcompress.c', 'main/texcompress_s3tc.c', 'main/texcompress_fxt1.c', + 'main/texenv.c', 'main/texenvprogram.c', 'main/texformat.c', + 'main/texgen.c', 'main/teximage.c', 'main/texobj.c', + 'main/texparam.c', 'main/texrender.c', 'main/texstate.c', 'main/texstore.c', -- cgit v1.2.3 From 34ff12ca1fe7153671eea2fe084f3991094ec3ce Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 11 Jun 2008 20:50:26 -0600 Subject: Revert "mesa: further degenerate the special case lit substitute" This reverts commit e841b92d9c8bf48085b4996df828ae745977f931. This fixes two specular lighting conform failures. --- src/mesa/main/ffvertex_prog.c | 21 +++++++++------------ 1 file changed, 9 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 06710f405d..d71e0c00fd 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -975,19 +975,19 @@ static void emit_degenerate_lit( struct tnl_program *p, { struct ureg id = get_identity_param(p); - /* Note that result.x & result.w will not be examined. Note also that - * dots.xyzw == dots.xxxx. + /* 1, 0, 0, 1 */ + emit_op1(p, OPCODE_MOV, lit, 0, swizzle(id, Z, X, X, Z)); - /* result[1] = MAX2(in, 0) + /* 1, MAX2(in[0], 0), 0, 1 */ - emit_op2(p, OPCODE_MAX, lit, 0, id, dots); + emit_op2(p, OPCODE_MAX, lit, WRITEMASK_Y, lit, swizzle1(dots, X)); - /* result[2] = (in > 0 ? 1 : 0) + /* 1, MAX2(in[0], 0), (in[0] > 0 ? 1 : 0), 1 */ emit_op2(p, OPCODE_SLT, lit, WRITEMASK_Z, lit, /* 0 */ - dots); /* in[0] */ + swizzle1(dots, X)); /* in[0] */ } @@ -1144,13 +1144,10 @@ static void build_lighting( struct tnl_program *p ) /* Calculate dot products: */ - if (p->state->material_shininess_is_zero) { - emit_op2(p, OPCODE_DP3, dots, 0, normal, VPpli); - } - else { - emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); + emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); + + if (!p->state->material_shininess_is_zero) emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half); - } /* Front face lighting: */ -- cgit v1.2.3 From 23d340c9edbbe64a99478b922c008ae3e2ec7103 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 12 Jun 2008 14:39:57 +0100 Subject: pb: don't assert(0) on failure to allocate - this is a normal condition in many drivers --- src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index 66256f3fa7..0a1e8c83b1 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -184,7 +184,6 @@ mm_bufmgr_create_buffer(struct pb_manager *mgr, mm_buf->block = mmAllocMem(mm->heap, size, mm->align2, 0); if(!mm_buf->block) { - assert(0); FREE(mm_buf); _glthread_UNLOCK_MUTEX(mm->mutex); return NULL; -- cgit v1.2.3 From 5ecb2f2d0fca0c5ea847d1968459aa0dd8138f14 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 12 Jun 2008 11:17:20 -0600 Subject: mesa: restore and fix Keith's "further degenerate the special case lit substitute" There was a bug in emit_degenerate_lit() that caused the SLT to produce unpredictable results in lit.z Plus, added a bunch of new comments. --- src/mesa/main/ffvertex_prog.c | 44 ++++++++++++++++++++++++++++++------------- 1 file changed, 31 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index d71e0c00fd..e6c7c1040f 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -969,25 +969,29 @@ static struct ureg calculate_light_attenuation( struct tnl_program *p, } +/** + * Compute: + * lit.y = MAX(0, dots.x) + * lit.z = SLT(0, dots.x) + */ static void emit_degenerate_lit( struct tnl_program *p, struct ureg lit, struct ureg dots ) { - struct ureg id = get_identity_param(p); - - /* 1, 0, 0, 1 + struct ureg id = get_identity_param(p); /* id = {0,0,0,1} */ + + /* Note that lit.x & lit.w will not be examined. Note also that + * dots.xyzw == dots.xxxx. */ - emit_op1(p, OPCODE_MOV, lit, 0, swizzle(id, Z, X, X, Z)); - /* 1, MAX2(in[0], 0), 0, 1 + /* MAX lit, id, dots; */ - emit_op2(p, OPCODE_MAX, lit, WRITEMASK_Y, lit, swizzle1(dots, X)); + emit_op2(p, OPCODE_MAX, lit, WRITEMASK_XYZW, id, dots); - /* 1, MAX2(in[0], 0), (in[0] > 0 ? 1 : 0), 1 + /* result[2] = (in > 0 ? 1 : 0) + * SLT lit.z, id.z, dots; # lit.z = (0 < dots.z) ? 1 : 0 */ - emit_op2(p, OPCODE_SLT, lit, WRITEMASK_Z, - lit, /* 0 */ - swizzle1(dots, X)); /* in[0] */ + emit_op2(p, OPCODE_SLT, lit, WRITEMASK_Z, swizzle1(id,Z), dots); } @@ -1007,6 +1011,14 @@ static void build_lighting( struct tnl_program *p ) struct ureg _bfc0 = undef, _bfc1 = undef; GLuint i; + /* + * NOTE: + * dot.x = dot(normal, VPpli) + * dot.y = dot(normal, halfAngle) + * dot.z = back.shininess + * dot.w = front.shininess + */ + for (i = 0; i < MAX_LIGHTS; i++) if (p->state->unit[i].light_enabled) nr_lights++; @@ -1144,10 +1156,13 @@ static void build_lighting( struct tnl_program *p ) /* Calculate dot products: */ - emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); - - if (!p->state->material_shininess_is_zero) + if (p->state->material_shininess_is_zero) { + emit_op2(p, OPCODE_DP3, dots, 0, normal, VPpli); + } + else { + emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli); emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half); + } /* Front face lighting: */ @@ -1181,15 +1196,18 @@ static void build_lighting( struct tnl_program *p ) if (!is_undef(att)) { + /* light is attenuated by distance */ emit_op1(p, OPCODE_LIT, lit, 0, dots); emit_op2(p, OPCODE_MUL, lit, 0, lit, att); emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0); } else if (!p->state->material_shininess_is_zero) { + /* there's a non-zero specular term */ emit_op1(p, OPCODE_LIT, lit, 0, dots); emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0); } else { + /* no attenutation, no specular */ emit_degenerate_lit(p, lit, dots); emit_op2(p, OPCODE_ADD, _col0, 0, ambient, _col0); } -- cgit v1.2.3 From 68ef8e89a5f25cd9f80e2b9088604631a28edc3c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Tue, 10 Jun 2008 16:59:44 -0400 Subject: glsl: implement variable array indexes --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 9 ++++++++- src/mesa/shader/arbprogparse.c | 2 +- src/mesa/shader/prog_print.c | 17 +++++++++++------ src/mesa/shader/slang/slang_emit.c | 16 +++++++++++----- src/mesa/shader/slang/slang_ir.h | 1 + src/mesa/state_tracker/st_mesa_to_tgsi.c | 4 ++-- 6 files changed, 34 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index d1a3dfd9c7..92aff88925 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -803,7 +803,14 @@ tgsi_dump_instruction( ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); CHR( '[' ); - SID( src->SrcRegister.Index ); + if (src->SrcRegister.Indirect) { + TXT( "addr" ); + if (src->SrcRegister.Index > 0) + CHR( '+' ); + SID( src->SrcRegister.Index ); + } + else + SID( src->SrcRegister.Index ); CHR( ']' ); if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index b60b9656c6..a6bbdc64f1 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -3880,7 +3880,7 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target, } #if DEBUG_FP - _mesa_printf("____________Fragment program %u ________\n", program->Base.ID); + _mesa_printf("____________Fragment program %u ________\n", program->Base.Id); _mesa_print_program(&program->Base); #endif } diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index 1c35ce3fec..09bf15f004 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -206,7 +206,7 @@ arb_output_attrib_string(GLint index, GLenum progType) */ static const char * reg_string(enum register_file f, GLint index, gl_prog_print_mode mode, - const struct gl_program *prog) + GLint relAddr, const struct gl_program *prog) { static char str[100]; @@ -214,7 +214,10 @@ reg_string(enum register_file f, GLint index, gl_prog_print_mode mode, switch (mode) { case PROG_PRINT_DEBUG: - sprintf(str, "%s[%d]", file_string(f, mode), index); + if (relAddr) + sprintf(str, "%s[ADDR%s%d]", file_string(f, mode), (index > 0) ? "+" : "", index); + else + sprintf(str, "%s[%d]", file_string(f, mode), index); break; case PROG_PRINT_ARB: @@ -401,7 +404,7 @@ print_dst_reg(const struct prog_dst_register *dstReg, gl_prog_print_mode mode, { _mesa_printf("%s%s", reg_string((enum register_file) dstReg->File, - dstReg->Index, mode, prog), + dstReg->Index, mode, GL_FALSE, prog), writemask_string(dstReg->WriteMask)); if (dstReg->CondMask != COND_TR) { @@ -424,9 +427,9 @@ print_src_reg(const struct prog_src_register *srcReg, gl_prog_print_mode mode, { _mesa_printf("%s%s", reg_string((enum register_file) srcReg->File, - srcReg->Index, mode, prog), + srcReg->Index, mode, srcReg->RelAddr, prog), _mesa_swizzle_string(srcReg->Swizzle, - srcReg->NegateBase, GL_FALSE)); + srcReg->NegateBase, GL_FALSE)); #if 0 _mesa_printf("%s[%d]%s", file_string((enum register_file) srcReg->File, mode), @@ -590,7 +593,9 @@ _mesa_print_instruction_opt(const struct prog_instruction *inst, GLint indent, break; case OPCODE_ARL: - _mesa_printf("ARL addr.x, "); + _mesa_printf("ARL "); + print_dst_reg(&inst->DstReg, mode, prog); + _mesa_printf(", "); print_src_reg(&inst->SrcReg[0], mode, prog); print_comment(inst); break; diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index ff63e05dd2..93256f8647 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -223,6 +223,7 @@ storage_to_src_reg(struct prog_src_register *src, const slang_ir_storage *st) assert(st->Size <= 4); src->File = st->File; src->Index = st->Index; + src->RelAddr = st->RelAddr; if (st->Swizzle != SWIZZLE_NOOP) src->Swizzle = st->Swizzle; else @@ -1488,11 +1489,16 @@ emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n) n->Store->Index = arrayAddr + index; } else { - /* Variable index - PROBLEM */ - const GLint arrayAddr = n->Children[0]->Store->Index; - const GLint index = 0; - _mesa_problem(NULL, "variable array indexes not supported yet!"); - n->Store->Index = arrayAddr + index; + /* Variable index*/ + struct prog_instruction *inst; + inst = new_instruction(emitInfo, OPCODE_ARL); + storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); + storage_to_src_reg(&inst->SrcReg[0], n->Children[1]->Store); + inst->DstReg.File = PROGRAM_ADDRESS; + inst->Comment = _mesa_strdup("ARL ADDR"); + n->Store->RelAddr = GL_TRUE; + n->Store->Index = inst->DstReg.Index;/*index of the array*/ + inst->DstReg.Index = 0; /*addr index is always 0*/ } return NULL; /* no instruction */ } diff --git a/src/mesa/shader/slang/slang_ir.h b/src/mesa/shader/slang/slang_ir.h index c7c0ddbf9a..ba0735d64d 100644 --- a/src/mesa/shader/slang/slang_ir.h +++ b/src/mesa/shader/slang/slang_ir.h @@ -146,6 +146,7 @@ struct _slang_ir_storage GLint Size; /**< number of floats */ GLuint Swizzle; GLint RefCount; /**< Used during IR tree delete */ + GLboolean RelAddr; }; typedef struct _slang_ir_storage slang_ir_storage; diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 12979de523..a8b6faad1c 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -68,8 +68,8 @@ map_register_file( case PROGRAM_STATE_VAR: case PROGRAM_NAMED_PARAM: case PROGRAM_UNIFORM: - if (immediateMapping[index] != ~0) - return TGSI_FILE_IMMEDIATE; + if (immediateMapping && immediateMapping[index] != ~0) + return TGSI_FILE_IMMEDIATE; else return TGSI_FILE_CONSTANT; case PROGRAM_CONSTANT: -- cgit v1.2.3 From 38fdf130d4016197e9ecfa6beeade0362f9622eb Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 12 Jun 2008 00:03:35 -0400 Subject: shaders: fix allocation of arrays of parameters --- src/mesa/shader/prog_parameter.c | 24 +++++++++++++++++++++++- 1 file changed, 23 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index 152bd79f69..afdd7e5489 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -62,6 +62,28 @@ _mesa_free_parameter_list(struct gl_program_parameter_list *paramList) } +static GLint +_mesa_fit_type_in_vec4(GLenum type) +{ + switch (type) { + case GL_FLOAT: + case GL_INT: + return 4; + break; + case GL_FLOAT_VEC2: + case GL_INT_VEC2: + return 2; + break; + case GL_FLOAT_VEC3: + case GL_INT_VEC3: + return 1; + break; + case GL_FLOAT_VEC4: + case GL_INT_VEC4: + default: + return 1; + } +} /** * Add a new parameter to a parameter list. @@ -272,7 +294,7 @@ _mesa_add_uniform(struct gl_program_parameter_list *paramList, } else { i = _mesa_add_parameter(paramList, PROGRAM_UNIFORM, name, - size, datatype, NULL, NULL); + size * _mesa_fit_type_in_vec4(datatype), datatype, NULL, NULL); return i; } } -- cgit v1.2.3 From 494c40dbed2bd63f43f70c1e6f7b46399d0a6377 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 12 Jun 2008 14:19:10 -0400 Subject: almost forgot - RelAddr is a boolean so use it here --- src/mesa/shader/prog_print.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index 09bf15f004..2c5e03acc2 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -206,7 +206,7 @@ arb_output_attrib_string(GLint index, GLenum progType) */ static const char * reg_string(enum register_file f, GLint index, gl_prog_print_mode mode, - GLint relAddr, const struct gl_program *prog) + GLboolean relAddr, const struct gl_program *prog) { static char str[100]; -- cgit v1.2.3 From cfe876371454074b598197582f8b01317b21eb51 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 12 Jun 2008 15:08:14 -0600 Subject: gallium: remove unused/old st_atom_fixedfunction.c file --- src/mesa/state_tracker/st_atom_fixedfunction.c | 66 -------------------------- 1 file changed, 66 deletions(-) delete mode 100644 src/mesa/state_tracker/st_atom_fixedfunction.c (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_fixedfunction.c b/src/mesa/state_tracker/st_atom_fixedfunction.c deleted file mode 100644 index 165567af70..0000000000 --- a/src/mesa/state_tracker/st_atom_fixedfunction.c +++ /dev/null @@ -1,66 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - /* - * Authors: - * Keith Whitwell - * Brian Paul - */ - -#include "tnl/t_vp_build.h" - -#include "pipe/p_context.h" -#include "pipe/p_defines.h" - -#include "st_context.h" -#include "st_atom.h" - - -#define TGSI_DEBUG 0 - - -/** - * When TnL state has changed, need to generate new vertex program. - * This should be done before updating the vertes shader (vs) state. - */ -static void update_tnl( struct st_context *st ) -{ - /* Would be good to avoid this when shaders are active: - */ - _tnl_UpdateFixedFunctionProgram( st->ctx ); -} - - -const struct st_tracked_state st_update_tnl = { - "st_update_tnl", /* name */ - { /* dirty */ - TNL_FIXED_FUNCTION_STATE_FLAGS, /* mesa */ - 0 /* st */ - }, - update_tnl /* update */ -}; - - -- cgit v1.2.3 From b19e8f720bb2f4d360f5497b64901fc48321f172 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 12 Jun 2008 15:46:10 -0600 Subject: gallium: test for FEATURE_convolution --- src/mesa/state_tracker/st_cb_texture.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index e15ccec89b..b2f4910225 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -26,7 +26,9 @@ **************************************************************************/ #include "main/imports.h" +#if FEATURE_convolution #include "main/convolve.h" +#endif #include "main/enums.h" #include "main/image.h" #include "main/macros.h" @@ -517,10 +519,12 @@ st_TexImage(GLcontext * ctx, stImage->face = _mesa_tex_target_to_face(target); stImage->level = level; +#if FEATURE_convolution if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) { _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth, &postConvHeight); } +#endif /* choose the texture format */ texImage->TexFormat = st_ChooseTextureFormat(ctx, internalFormat, -- cgit v1.2.3 From 08f1b8ac709105d42ec34f8b8a81421e3b0fbc81 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Thu, 12 Jun 2008 16:01:05 -0600 Subject: gallium: fix SSE codegen for instructions that use both a CONSTANT and IMMEDIATE Fixes codegen for instructions like MUL dst, CONST[0], IMM[0]; the two operands would up getting aliased in the x86/sse code. Fixes glean/vertProg1/fogparams test. --- src/gallium/auxiliary/draw/draw_vs_aos.c | 26 +++++++++++++++++++------- src/gallium/auxiliary/draw/draw_vs_aos.h | 3 ++- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 2 +- 3 files changed, 22 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 5d4a8b38c8..388dd3fbee 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -67,19 +67,30 @@ static INLINE boolean eq( struct x86_reg a, } struct x86_reg aos_get_x86( struct aos_compilation *cp, + unsigned which_reg, /* quick hack */ unsigned value ) { - if (cp->ebp != value) { + struct x86_reg reg; + + if (which_reg == 0) + reg = cp->temp_EBP; + else + reg = cp->tmp_EAX; + + if (cp->x86_reg[which_reg] != value) { unsigned offset; switch (value) { case X86_IMMEDIATES: + assert(which_reg == 0); offset = Offset(struct aos_machine, immediates); break; case X86_CONSTANTS: + assert(which_reg == 1); offset = Offset(struct aos_machine, constants); break; case X86_ATTRIBS: + assert(which_reg == 0); offset = Offset(struct aos_machine, attrib); break; default: @@ -87,14 +98,14 @@ struct x86_reg aos_get_x86( struct aos_compilation *cp, offset = 0; } - x86_mov(cp->func, cp->temp_EBP, + + x86_mov(cp->func, reg, x86_make_disp(cp->machine_EDX, offset)); - /* x86_deref(x86_make_disp(cp->machine_EDX, offset))); */ - cp->ebp = value; + cp->x86_reg[which_reg] = value; } - return cp->temp_EBP; + return reg; } @@ -118,10 +129,10 @@ static struct x86_reg get_reg_ptr(struct aos_compilation *cp, return x86_make_disp(ptr, Offset(struct aos_machine, internal[idx])); case TGSI_FILE_IMMEDIATE: - return x86_make_disp(aos_get_x86(cp, X86_IMMEDIATES), idx * 4 * sizeof(float)); + return x86_make_disp(aos_get_x86(cp, 0, X86_IMMEDIATES), idx * 4 * sizeof(float)); case TGSI_FILE_CONSTANT: - return x86_make_disp(aos_get_x86(cp, X86_CONSTANTS), idx * 4 * sizeof(float)); + return x86_make_disp(aos_get_x86(cp, 1, X86_CONSTANTS), idx * 4 * sizeof(float)); default: ERROR(cp, "unknown reg file"); @@ -1413,6 +1424,7 @@ static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_inst x87_fld_src( cp, &op->FullSrcRegisters[0], 0 ); x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 0 ) ); + /* tmp_EAX has been pushed & will be restored below */ x86_mov_reg_imm( cp->func, cp->tmp_EAX, (unsigned long) _powerf ); x86_call( cp->func, cp->tmp_EAX ); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.h b/src/gallium/auxiliary/draw/draw_vs_aos.h index 66944a4e33..64e021ff6b 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.h +++ b/src/gallium/auxiliary/draw/draw_vs_aos.h @@ -145,7 +145,7 @@ struct aos_compilation { unsigned last_used; } xmm[8]; - unsigned ebp; /* one of X86_* */ + unsigned x86_reg[2]; /* one of X86_* */ boolean input_fetched[PIPE_MAX_ATTRIBS]; unsigned output_last_write[PIPE_MAX_ATTRIBS]; @@ -213,6 +213,7 @@ do { \ #define X86_ATTRIBS 3 struct x86_reg aos_get_x86( struct aos_compilation *cp, + unsigned which_reg, unsigned value ); diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index b720185709..6b92811870 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -96,7 +96,7 @@ static void get_src_ptr( struct aos_compilation *cp, struct x86_reg elt, unsigned a ) { - struct x86_reg attrib = x86_make_disp(aos_get_x86( cp, X86_ATTRIBS ), + struct x86_reg attrib = x86_make_disp(aos_get_x86( cp, 0, X86_ATTRIBS ), a * sizeof(struct aos_attrib)); struct x86_reg input_ptr = x86_make_disp(attrib, -- cgit v1.2.3 From 20ee00754d432cf6c9aca2ba61e004a83795e160 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 12 Jun 2008 16:01:43 -0600 Subject: gallium: disable the codegen for TGSI_OPCODE_EXPBASE2 for now. The x86 code seems to fail for exponents of 4 or larger. See glean's vertProg1/EX2 test. --- src/gallium/auxiliary/draw/draw_vs_aos.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 388dd3fbee..1f926b3e85 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -1662,7 +1662,14 @@ emit_instruction( struct aos_compilation *cp, return emit_RND(cp, inst); case TGSI_OPCODE_EXPBASE2: +#if 0 + /* this seems to fail for "larger" exponents. + * See glean tvertProg1's EX2 test. + */ return emit_EX2(cp, inst); +#else + return FALSE; +#endif case TGSI_OPCODE_LOGBASE2: return emit_LG2(cp, inst); -- cgit v1.2.3 From 73f78dedad36a5bae6cac1845d297bb4357a2786 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 12 Jun 2008 16:37:16 -0600 Subject: gallium: add missing comma --- src/mesa/SConscript | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index ff5bf60215..af83db0b5f 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -24,7 +24,7 @@ if env['platform'] != 'winddk': main_sources = [ 'main/api_arrayelt.c', - 'main/api_exec.c' + 'main/api_exec.c', 'main/api_loopback.c', 'main/api_noop.c', 'main/api_validate.c', -- cgit v1.2.3 From 6363960db55769db73021081b60f1028723d9f58 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 12 Jun 2008 16:48:24 -0600 Subject: mesa: remove unneeded #include --- src/mesa/shader/prog_statevars.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 8f48155825..7e444de304 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -37,7 +37,6 @@ #include "mtypes.h" #include "prog_statevars.h" #include "prog_parameter.h" -#include "nvvertparse.h" /** -- cgit v1.2.3 From 5b51cc020efc5519bc7cc34f6fcb4138eab76e0a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 12 Jun 2008 16:54:44 -0600 Subject: gallium: add some #if FEATURE_x tests --- src/mesa/state_tracker/st_cb_blit.c | 2 ++ src/mesa/state_tracker/st_cb_texture.c | 4 ++++ src/mesa/state_tracker/st_format.c | 4 ++++ 3 files changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c index 0533d9460f..799ef3f3be 100644 --- a/src/mesa/state_tracker/st_cb_blit.c +++ b/src/mesa/state_tracker/st_cb_blit.c @@ -106,5 +106,7 @@ st_BlitFramebuffer(GLcontext *ctx, void st_init_blit_functions(struct dd_function_table *functions) { +#if FEATURE_EXT_framebuffer_blit functions->BlitFramebuffer = st_BlitFramebuffer; +#endif } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index b2f4910225..ed65f3b0aa 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -97,14 +97,18 @@ static int compressed_num_bytes(GLuint mesaFormat) { switch(mesaFormat) { +#if FEATURE_texture_fxt1 case MESA_FORMAT_RGB_FXT1: case MESA_FORMAT_RGBA_FXT1: +#endif +#if FEATURE_texture_s3tc case MESA_FORMAT_RGB_DXT1: case MESA_FORMAT_RGBA_DXT1: return 2; case MESA_FORMAT_RGBA_DXT3: case MESA_FORMAT_RGBA_DXT5: return 4; +#endif default: return 0; } diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 1b9325c159..5759f393af 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -278,6 +278,7 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) return PIPE_FORMAT_S8Z24_UNORM; case MESA_FORMAT_YCBCR: return PIPE_FORMAT_YCBCR; +#if FEATURE_texture_s3tc case MESA_FORMAT_RGB_DXT1: return PIPE_FORMAT_DXT1_RGB; case MESA_FORMAT_RGBA_DXT1: @@ -286,6 +287,7 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) return PIPE_FORMAT_DXT3_RGBA; case MESA_FORMAT_RGBA_DXT5: return PIPE_FORMAT_DXT5_RGBA; +#endif default: assert(0); return 0; @@ -569,6 +571,7 @@ translate_gallium_format_to_mesa_format(enum pipe_format format) return &_mesa_texformat_ycbcr; case PIPE_FORMAT_YCBCR_REV: return &_mesa_texformat_ycbcr_rev; +#if FEATURE_texture_s3tc case PIPE_FORMAT_DXT1_RGB: return &_mesa_texformat_rgb_dxt1; case PIPE_FORMAT_DXT1_RGBA: @@ -577,6 +580,7 @@ translate_gallium_format_to_mesa_format(enum pipe_format format) return &_mesa_texformat_rgba_dxt3; case PIPE_FORMAT_DXT5_RGBA: return &_mesa_texformat_rgba_dxt5; +#endif /* XXX add additional cases */ default: assert(0); -- cgit v1.2.3 From e961a5da77cbcdb0e32400ec707c16fcfe9d7083 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 12 Jun 2008 16:55:28 -0600 Subject: mesa: add some #if FEATURE_x tests --- src/mesa/main/context.c | 4 ++-- src/mesa/main/enable.c | 5 +++++ src/mesa/main/texcompress.c | 17 +++++++++++++++++ src/mesa/main/texformat.c | 6 ++++++ src/mesa/main/texformat.h | 8 ++++++++ 5 files changed, 38 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index d975814592..8f22f6bb78 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -144,9 +144,7 @@ #include "vtxfmt.h" #include "glapi/glthread.h" #include "glapi/glapioffsets.h" -#if FEATURE_NV_vertex_program || FEATURE_NV_fragment_program #include "shader/program.h" -#endif #include "shader/shader_api.h" #include "shader/atifragshader.h" #if _HAVE_FULL_GL @@ -622,6 +620,7 @@ delete_program_cb(GLuint id, void *data, void *userData) ctx->Driver.DeleteProgram(ctx, prog); } +#if FEATURE_ATI_fragment_shader /** * Callback for deleting an ATI fragment shader object. * Called by _mesa_HashDeleteAll(). @@ -633,6 +632,7 @@ delete_fragshader_cb(GLuint id, void *data, void *userData) GLcontext *ctx = (GLcontext *) userData; _mesa_delete_ati_fragment_shader(ctx, shader); } +#endif /** * Callback for deleting a buffer object. Called by _mesa_HashDeleteAll(). diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c index 52dd63f2ce..6b4ad8eea3 100644 --- a/src/mesa/main/enable.c +++ b/src/mesa/main/enable.c @@ -904,6 +904,7 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state) break; /* GL_MESA_program_debug */ +#if FEATURE_MESA_program_debug case GL_FRAGMENT_PROGRAM_CALLBACK_MESA: CHECK_EXTENSION(MESA_program_debug, cap); ctx->FragmentProgram.CallbackEnabled = state; @@ -912,6 +913,7 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state) CHECK_EXTENSION(MESA_program_debug, cap); ctx->VertexProgram.CallbackEnabled = state; break; +#endif #if FEATURE_ATI_fragment_shader case GL_FRAGMENT_SHADER_ATI: @@ -1349,12 +1351,15 @@ _mesa_IsEnabled( GLenum cap ) return ctx->Depth.BoundsTest; /* GL_MESA_program_debug */ +#if FEATURE_MESA_program_debug case GL_FRAGMENT_PROGRAM_CALLBACK_MESA: CHECK_EXTENSION(MESA_program_debug); return ctx->FragmentProgram.CallbackEnabled; case GL_VERTEX_PROGRAM_CALLBACK_MESA: CHECK_EXTENSION(MESA_program_debug); return ctx->VertexProgram.CallbackEnabled; +#endif + #if FEATURE_ATI_fragment_shader case GL_FRAGMENT_SHADER_ATI: CHECK_EXTENSION(ATI_fragment_shader); diff --git a/src/mesa/main/texcompress.c b/src/mesa/main/texcompress.c index c44d594d68..5ad936419b 100644 --- a/src/mesa/main/texcompress.c +++ b/src/mesa/main/texcompress.c @@ -137,8 +137,10 @@ _mesa_compressed_texture_size( GLcontext *ctx, ASSERT(depth == 1); (void) depth; + (void) size; switch (mesaFormat) { +#if FEATURE_texture_fxt1 case MESA_FORMAT_RGB_FXT1: case MESA_FORMAT_RGBA_FXT1: /* round up width to next multiple of 8, height to next multiple of 4 */ @@ -152,6 +154,8 @@ _mesa_compressed_texture_size( GLcontext *ctx, if (size < 16) size = 16; return size; +#endif +#if FEATURE_texture_s3tc case MESA_FORMAT_RGB_DXT1: case MESA_FORMAT_RGBA_DXT1: /* round up width, height to next multiple of 4 */ @@ -178,6 +182,7 @@ _mesa_compressed_texture_size( GLcontext *ctx, if (size < 16) size = 16; return size; +#endif default: _mesa_problem(ctx, "bad mesaFormat in _mesa_compressed_texture_size"); return 0; @@ -202,12 +207,15 @@ _mesa_compressed_texture_size_glenum(GLcontext *ctx, GLuint mesaFormat; switch (glformat) { +#if FEATURE_texture_fxt1 case GL_COMPRESSED_RGB_FXT1_3DFX: mesaFormat = MESA_FORMAT_RGB_FXT1; break; case GL_COMPRESSED_RGBA_FXT1_3DFX: mesaFormat = MESA_FORMAT_RGBA_FXT1; break; +#endif +#if FEATURE_texture_s3tc case GL_COMPRESSED_RGB_S3TC_DXT1_EXT: case GL_RGB_S3TC: mesaFormat = MESA_FORMAT_RGB_DXT1; @@ -224,6 +232,7 @@ _mesa_compressed_texture_size_glenum(GLcontext *ctx, case GL_RGBA4_S3TC: mesaFormat = MESA_FORMAT_RGBA_DXT5; break; +#endif default: return 0; } @@ -245,10 +254,13 @@ _mesa_compressed_row_stride(GLuint mesaFormat, GLsizei width) GLint stride; switch (mesaFormat) { +#if FEATURE_texture_fxt1 case MESA_FORMAT_RGB_FXT1: case MESA_FORMAT_RGBA_FXT1: stride = ((width + 7) / 8) * 16; /* 16 bytes per 8x4 tile */ break; +#endif +#if FEATURE_texture_s3tc case MESA_FORMAT_RGB_DXT1: case MESA_FORMAT_RGBA_DXT1: stride = ((width + 3) / 4) * 8; /* 8 bytes per 4x4 tile */ @@ -257,6 +269,7 @@ _mesa_compressed_row_stride(GLuint mesaFormat, GLsizei width) case MESA_FORMAT_RGBA_DXT5: stride = ((width + 3) / 4) * 16; /* 16 bytes per 4x4 tile */ break; +#endif default: _mesa_problem(NULL, "bad mesaFormat in _mesa_compressed_row_stride"); return 0; @@ -293,10 +306,13 @@ _mesa_compressed_image_address(GLint col, GLint row, GLint img, */ switch (mesaFormat) { +#if FEATURE_texture_fxt1 case MESA_FORMAT_RGB_FXT1: case MESA_FORMAT_RGBA_FXT1: addr = (GLubyte *) image + 16 * (((width + 7) / 8) * (row / 4) + col / 8); break; +#endif +#if FEATURE_texture_s3tc case MESA_FORMAT_RGB_DXT1: case MESA_FORMAT_RGBA_DXT1: addr = (GLubyte *) image + 8 * (((width + 3) / 4) * (row / 4) + col / 4); @@ -305,6 +321,7 @@ _mesa_compressed_image_address(GLint col, GLint row, GLint img, case MESA_FORMAT_RGBA_DXT5: addr = (GLubyte *) image + 16 * (((width + 3) / 4) * (row / 4) + col / 4); break; +#endif default: _mesa_problem(NULL, "bad mesaFormat in _mesa_compressed_image_address"); addr = NULL; diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c index a35195a695..ea55002f58 100644 --- a/src/mesa/main/texformat.c +++ b/src/mesa/main/texformat.c @@ -1693,6 +1693,7 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format, *comps = 1; return; +#if FEATURE_EXT_texture_sRGB case MESA_FORMAT_SRGB8: *datatype = GL_UNSIGNED_BYTE; *comps = 3; @@ -1709,9 +1710,13 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format, *datatype = GL_UNSIGNED_BYTE; *comps = 2; return; +#endif +#if FEATURE_texture_fxt1 case MESA_FORMAT_RGB_FXT1: case MESA_FORMAT_RGBA_FXT1: +#endif +#if FEATURE_texture_s3tc case MESA_FORMAT_RGB_DXT1: case MESA_FORMAT_RGBA_DXT1: case MESA_FORMAT_RGBA_DXT3: @@ -1720,6 +1725,7 @@ _mesa_format_to_type_and_comps(const struct gl_texture_format *format, *datatype = GL_UNSIGNED_BYTE; *comps = 0; return; +#endif case MESA_FORMAT_RGBA: *datatype = CHAN_TYPE; diff --git a/src/mesa/main/texformat.h b/src/mesa/main/texformat.h index 8f4e2feb48..3b6d05612d 100644 --- a/src/mesa/main/texformat.h +++ b/src/mesa/main/texformat.h @@ -105,12 +105,16 @@ enum _format { * \name Compressed texture formats. */ /*@{*/ +#if FEATURE_texture_fxt1 MESA_FORMAT_RGB_FXT1, MESA_FORMAT_RGBA_FXT1, +#endif +#if FEATURE_texture_s3tc MESA_FORMAT_RGB_DXT1, MESA_FORMAT_RGBA_DXT1, MESA_FORMAT_RGBA_DXT3, MESA_FORMAT_RGBA_DXT5, +#endif /*@}*/ /** @@ -223,12 +227,16 @@ extern const struct gl_texture_format _mesa_texformat_ycbcr_rev; /** \name Compressed formats */ /*@{*/ +#if FEATURE_texture_fxt1 extern const struct gl_texture_format _mesa_texformat_rgb_fxt1; extern const struct gl_texture_format _mesa_texformat_rgba_fxt1; +#endif +#if FEATURE_texture_s3tc extern const struct gl_texture_format _mesa_texformat_rgb_dxt1; extern const struct gl_texture_format _mesa_texformat_rgba_dxt1; extern const struct gl_texture_format _mesa_texformat_rgba_dxt3; extern const struct gl_texture_format _mesa_texformat_rgba_dxt5; +#endif /*@}*/ /** \name The null format */ -- cgit v1.2.3 From 3911ef032745b18071f48a6ad343ece14202049c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 12 Jun 2008 18:53:52 -0400 Subject: glsl: make sure we replace all output reads with temporaries test in if.glsl --- src/mesa/shader/programopt.c | 12 +++++++----- src/mesa/shader/programopt.h | 2 +- src/mesa/shader/slang/slang_compile.c | 3 ++- 3 files changed, 10 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index 7d560c74a5..da4e3fcfcf 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -367,20 +367,22 @@ _mesa_count_texture_instructions(struct gl_program *prog) /** - * Scan/rewrite program to remove reads of varying (output) registers. + * Scan/rewrite program to remove reads of custom (output) registers. + * The passed type has to be either PROGRAM_VARYING or PROGRAM_OUTPUT. * In GLSL vertex shaders, varying vars can be read and written. * Normally, vertex varying vars are implemented as output registers. * On some hardware, trying to read an output register causes trouble. * So, rewrite the program to use a temporary register in this case. */ void -_mesa_remove_varying_reads(struct gl_program *prog) +_mesa_remove_output_reads(struct gl_program *prog, enum register_file type) { GLuint i; GLint outputMap[VERT_RESULT_MAX]; GLuint numVaryingReads = 0; assert(prog->Target == GL_VERTEX_PROGRAM_ARB); + assert(type == PROGRAM_UNIFORM || type == PROGRAM_OUTPUT); for (i = 0; i < VERT_RESULT_MAX; i++) outputMap[i] = -1; @@ -391,7 +393,7 @@ _mesa_remove_varying_reads(struct gl_program *prog) const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode); GLuint j; for (j = 0; j < numSrc; j++) { - if (inst->SrcReg[j].File == PROGRAM_VARYING) { + if (inst->SrcReg[j].File == type) { /* replace the read with a temp reg */ const GLuint var = inst->SrcReg[j].Index; if (outputMap[var] == -1) { @@ -414,7 +416,7 @@ _mesa_remove_varying_reads(struct gl_program *prog) const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode); GLuint j; for (j = 0; j < numSrc; j++) { - if (inst->DstReg.File == PROGRAM_VARYING && + if (inst->DstReg.File == type && outputMap[inst->DstReg.Index] >= 0) { /* change inst to write to the temp reg, instead of the varying */ inst->DstReg.File = PROGRAM_TEMPORARY; @@ -447,7 +449,7 @@ _mesa_remove_varying_reads(struct gl_program *prog) if (outputMap[var] >= 0) { /* MOV VAR[var], TEMP[tmp]; */ inst->Opcode = OPCODE_MOV; - inst->DstReg.File = PROGRAM_VARYING; + inst->DstReg.File = type; inst->DstReg.Index = var; inst->SrcReg[0].File = PROGRAM_TEMPORARY; inst->SrcReg[0].Index = outputMap[var]; diff --git a/src/mesa/shader/programopt.h b/src/mesa/shader/programopt.h index 47ff2f0c7b..11572e64f5 100644 --- a/src/mesa/shader/programopt.h +++ b/src/mesa/shader/programopt.h @@ -40,6 +40,6 @@ extern void _mesa_count_texture_instructions(struct gl_program *prog); extern void -_mesa_remove_varying_reads(struct gl_program *prog); +_mesa_remove_output_reads(struct gl_program *prog, enum register_file type); #endif /* PROGRAMOPT_H */ diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index f2e8aab8b0..cdea1c5128 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -2212,7 +2212,8 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) printf("Pre-remove output reads:\n"); _mesa_print_program(shader->Programs[0]); #endif - _mesa_remove_varying_reads(shader->Programs[0]); + _mesa_remove_output_reads(shader->Programs[0], PROGRAM_VARYING); + _mesa_remove_output_reads(shader->Programs[0], PROGRAM_OUTPUT); #if 0 printf("Post-remove output reads:\n"); _mesa_print_program(shader->Programs[0]); -- cgit v1.2.3 From 314c46060486ae77cb78b3d8be155813145fc72e Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Thu, 12 Jun 2008 19:41:54 -0400 Subject: silly bug: it's PROGRAM_VARYING --- src/mesa/shader/programopt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index da4e3fcfcf..f3511ba00e 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -382,7 +382,7 @@ _mesa_remove_output_reads(struct gl_program *prog, enum register_file type) GLuint numVaryingReads = 0; assert(prog->Target == GL_VERTEX_PROGRAM_ARB); - assert(type == PROGRAM_UNIFORM || type == PROGRAM_OUTPUT); + assert(type == PROGRAM_VARYING || type == PROGRAM_OUTPUT); for (i = 0; i < VERT_RESULT_MAX; i++) outputMap[i] = -1; -- cgit v1.2.3 From e9b8df69b40c55953e1b3503690b54f993773223 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 13 Jun 2008 12:21:58 +0200 Subject: i915: Support all primtive types in vbuf path --- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 44 +++++++++++++++++++++++-- 1 file changed, 42 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index adfb16fbc5..c7a31fdb70 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -163,6 +163,10 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render, i915_render->hwprim = PRIM3D_LINELIST; i915_render->fallback = 0; return TRUE; + case PIPE_PRIM_LINE_LOOP: + i915_render->hwprim = PRIM3D_LINELIST; + i915_render->fallback = PIPE_PRIM_LINE_LOOP; + return TRUE; case PIPE_PRIM_LINE_STRIP: i915_render->hwprim = PRIM3D_LINESTRIP; i915_render->fallback = 0; @@ -175,6 +179,10 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render, i915_render->hwprim = PRIM3D_TRISTRIP; i915_render->fallback = 0; return TRUE; + case PIPE_PRIM_TRIANGLE_FAN: + i915_render->hwprim = PRIM3D_TRIFAN; + i915_render->fallback = 0; + return TRUE; case PIPE_PRIM_QUADS: i915_render->hwprim = PRIM3D_TRILIST; i915_render->fallback = PIPE_PRIM_QUADS; @@ -183,7 +191,12 @@ i915_vbuf_render_set_primitive( struct vbuf_render *render, i915_render->hwprim = PRIM3D_TRILIST; i915_render->fallback = PIPE_PRIM_QUAD_STRIP; return TRUE; + case PIPE_PRIM_POLYGON: + i915_render->hwprim = PRIM3D_POLY; + i915_render->fallback = 0; + return TRUE; default: + assert((int)"Error unkown primtive type" & 0); /* Actually, can handle a lot more just fine... Fixme. */ return FALSE; @@ -211,6 +224,13 @@ draw_arrays_generate_indices( struct vbuf_render *render, if (i < end) OUT_BATCH( i ); break; + case PIPE_PRIM_LINE_LOOP: + if (nr >= 2) { + for (i = start + 1; i < end; i++) + OUT_BATCH( (i-0) | (i+0) << 16 ); + OUT_BATCH( (i-0) | ( 0) << 16 ); + } + break; case PIPE_PRIM_QUADS: for (i = start; i + 3 < end; i += 4) { OUT_BATCH( (i+0) | (i+1) << 16 ); @@ -236,6 +256,11 @@ draw_arrays_calc_nr_indices( uint nr, unsigned type ) switch (type) { case 0: return nr; + case PIPE_PRIM_LINE_LOOP: + if (nr >= 2) + return nr * 2; + else + return 0; case PIPE_PRIM_QUADS: return (nr / 4) * 6; case PIPE_PRIM_QUAD_STRIP: @@ -262,6 +287,8 @@ draw_arrays_fallback( struct vbuf_render *render, i915_emit_hardware_state( i915 ); nr_indices = draw_arrays_calc_nr_indices( nr, i915_render->fallback ); + if (!nr_indices) + return; if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) { FLUSH_BATCH(NULL); @@ -328,6 +355,13 @@ draw_generate_indices( struct vbuf_render *render, OUT_BATCH( indices[i] ); } break; + case PIPE_PRIM_LINE_LOOP: + if (nr_indices >= 2) { + for (i = 1; i < nr_indices; i++) + OUT_BATCH( indices[i-0] | indices[i+0] << 16 ); + OUT_BATCH( indices[i-0] | indices[0] << 16 ); + } + break; case PIPE_PRIM_QUADS: for (i = 0; i + 3 < nr_indices; i += 4) { OUT_BATCH( indices[i+0] | indices[i+1] << 16 ); @@ -354,6 +388,11 @@ draw_calc_nr_indices( uint nr_indices, unsigned type ) switch (type) { case 0: return nr_indices; + case PIPE_PRIM_LINE_LOOP: + if (nr_indices >= 2) + return nr_indices * 2; + else + return 0; case PIPE_PRIM_QUADS: return (nr_indices / 4) * 6; case PIPE_PRIM_QUAD_STRIP: @@ -374,9 +413,10 @@ i915_vbuf_render_draw( struct vbuf_render *render, unsigned save_nr_indices; save_nr_indices = nr_indices; - nr_indices = draw_calc_nr_indices( nr_indices, i915_render->fallback ); - assert(nr_indices); + nr_indices = draw_calc_nr_indices( nr_indices, i915_render->fallback ); + if (!nr_indices) + return; if (i915->dirty) i915_update_derived( i915 ); -- cgit v1.2.3 From de35bf5bdca83f05dc919bf175588ce749009dfc Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 13 Jun 2008 14:11:20 +0200 Subject: i915: Messed up lineloop now works --- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index c7a31fdb70..aef3682bbf 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -228,7 +228,7 @@ draw_arrays_generate_indices( struct vbuf_render *render, if (nr >= 2) { for (i = start + 1; i < end; i++) OUT_BATCH( (i-0) | (i+0) << 16 ); - OUT_BATCH( (i-0) | ( 0) << 16 ); + OUT_BATCH( (i-0) | ( start) << 16 ); } break; case PIPE_PRIM_QUADS: @@ -358,8 +358,8 @@ draw_generate_indices( struct vbuf_render *render, case PIPE_PRIM_LINE_LOOP: if (nr_indices >= 2) { for (i = 1; i < nr_indices; i++) - OUT_BATCH( indices[i-0] | indices[i+0] << 16 ); - OUT_BATCH( indices[i-0] | indices[0] << 16 ); + OUT_BATCH( indices[i-1] | indices[i] << 16 ); + OUT_BATCH( indices[i-1] | indices[0] << 16 ); } break; case PIPE_PRIM_QUADS: -- cgit v1.2.3 From b8504534978ba7647838c29549d6698acb09a94f Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 13 Jun 2008 14:49:04 +0200 Subject: mesa: Remove duplicate code. --- src/mesa/state_tracker/st_format.c | 19 +------------------ 1 file changed, 1 insertion(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 5759f393af..03ec558a36 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -49,24 +49,7 @@ format_bits( pipe_format_rgbazs_t info, GLuint comp ) { - GLuint size; - - if (pf_swizzle_x(info) == comp) { - size = pf_size_x(info); - } - else if (pf_swizzle_y(info) == comp) { - size = pf_size_y(info); - } - else if (pf_swizzle_z(info) == comp) { - size = pf_size_z(info); - } - else if (pf_swizzle_w(info) == comp) { - size = pf_size_w(info); - } - else { - size = 0; - } - return size << (pf_exp8(info) * 3); + return pf_get_component_bits( (enum pipe_format) info, comp ); } static GLuint -- cgit v1.2.3 From 6f548c88e265c70cd0eb6bc148e23d8ce8d83133 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 13 Jun 2008 14:50:29 +0200 Subject: gallium: Allow pipe format component sizes to be specified with finer granularity. This will allow us to define A2R10G10B10 format. --- src/gallium/include/pipe/p_format.h | 22 ++++++++++++++-------- 1 file changed, 14 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 3aad463049..b7716363b8 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -121,21 +121,21 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) #define pf_size_z(f) pf_get(f, 20, 0x7) /**< Size of Z */ #define pf_size_w(f) pf_get(f, 23, 0x7) /**< Size of W */ #define pf_size_xyzw(f,i) pf_get(f, 14+((i)*3), 0x7) -#define pf_exp8(f) pf_get(f, 26, 0x3) /**< Scale size by 8 ^ exp8 */ -#define pf_type(f) pf_get(f, 28, 0xf) /**< PIPE_FORMAT_TYPE_ */ +#define pf_exp2(f) pf_get(f, 26, 0x7) /**< Scale size by 2 ^ exp2 */ +#define pf_type(f) pf_get(f, 29, 0x7) /**< PIPE_FORMAT_TYPE_ */ /** * Helper macro to encode the above structure into a 32-bit value. */ -#define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP8, TYPE ) (\ +#define _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, EXP2, TYPE ) (\ (PIPE_FORMAT_LAYOUT_RGBAZS << 0) |\ ((SWZ) << 2) |\ ((SIZEX) << 14) |\ ((SIZEY) << 17) |\ ((SIZEZ) << 20) |\ ((SIZEW) << 23) |\ - ((EXP8) << 26) |\ - ((TYPE) << 28) ) + ((EXP2) << 26) |\ + ((TYPE) << 29) ) /** * Helper macro to encode the swizzle part of the structure above. @@ -148,17 +148,23 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) #define _PIPE_FORMAT_RGBAZS_1( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 0, TYPE ) +/** + * Shorthand macro for RGBAZS layout with component sizes in 2-bit units. + */ +#define _PIPE_FORMAT_RGBAZS_2( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) + /** * Shorthand macro for RGBAZS layout with component sizes in 8-bit units. */ #define _PIPE_FORMAT_RGBAZS_8( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 1, TYPE ) + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 3, TYPE ) /** * Shorthand macro for RGBAZS layout with component sizes in 64-bit units. */ #define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ - _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 2, TYPE ) + _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 6, TYPE ) /** * Shorthand macro for common format swizzles. @@ -359,7 +365,7 @@ static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) else { size = 0; } - return size << (pf_exp8(format) * 3); + return size << pf_exp2(format); } /** -- cgit v1.2.3 From b03a0373a234af00e50652a6a2d75fd9ed8fc19b Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 13 Jun 2008 14:58:24 +0200 Subject: gallium: Add PIPE_FORMAT_A2B10G10R10_UNORM. --- src/gallium/include/pipe/p_format.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index b7716363b8..b4d6399c02 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -174,6 +174,7 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) #define _PIPE_FORMAT_RGB1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_1 ) #define _PIPE_FORMAT_RGBA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_A ) #define _PIPE_FORMAT_ARGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) +#define _PIPE_FORMAT_ABGR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R ) #define _PIPE_FORMAT_BGRA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_A ) #define _PIPE_FORMAT_1RGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_1, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) #define _PIPE_FORMAT_BGR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) @@ -244,6 +245,7 @@ enum pipe_format { PIPE_FORMAT_A1R5G5B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_A4R4G4B4_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB1, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_A2B10G10R10_UNORM = _PIPE_FORMAT_RGBAZS_2 ( _PIPE_FORMAT_ABGR, 2, 10,10,10,PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_L8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ PIPE_FORMAT_A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 0, 0, 0, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ PIPE_FORMAT_I8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ -- cgit v1.2.3 From 62f03a9ecc1d73ff67f5c2a9e5f9638bc6f7d458 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 13 Jun 2008 15:21:11 +0200 Subject: gallium: Fix PIPE_FORMAT_A2B10G10R10_UNORM definition. Whoops! --- src/gallium/include/pipe/p_format.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index b4d6399c02..b1bad8ab0e 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -245,7 +245,7 @@ enum pipe_format { PIPE_FORMAT_A1R5G5B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_A4R4G4B4_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_ARGB, 4, 4, 4, 4, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_R5G6B5_UNORM = _PIPE_FORMAT_RGBAZS_1 ( _PIPE_FORMAT_RGB1, 5, 6, 5, 0, PIPE_FORMAT_TYPE_UNORM ), - PIPE_FORMAT_A2B10G10R10_UNORM = _PIPE_FORMAT_RGBAZS_2 ( _PIPE_FORMAT_ABGR, 2, 10,10,10,PIPE_FORMAT_TYPE_UNORM ), + PIPE_FORMAT_A2B10G10R10_UNORM = _PIPE_FORMAT_RGBAZS_2 ( _PIPE_FORMAT_ABGR, 1, 5, 5, 5, PIPE_FORMAT_TYPE_UNORM ), PIPE_FORMAT_L8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte luminance */ PIPE_FORMAT_A8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_000R, 0, 0, 0, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte alpha */ PIPE_FORMAT_I8_UNORM = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRR, 1, 1, 1, 1, PIPE_FORMAT_TYPE_UNORM ), /**< ubyte intensity */ -- cgit v1.2.3 From 9350fd62b6b8dbee77b0c3c4f23195eaee3045d7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 13 Jun 2008 09:10:09 -0600 Subject: mesa: fix typo: s/stacks/stack/ --- src/mesa/main/mfeatures.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h index 162bd2fff9..4e60611c48 100644 --- a/src/mesa/main/mfeatures.h +++ b/src/mesa/main/mfeatures.h @@ -37,7 +37,7 @@ #endif #define FEATURE_accum _HAVE_FULL_GL -#define FEATURE_attrib_stacks _HAVE_FULL_GL +#define FEATURE_attrib_stack _HAVE_FULL_GL #define FEATURE_colortable _HAVE_FULL_GL #define FEATURE_convolution _HAVE_FULL_GL #define FEATURE_dlist _HAVE_FULL_GL -- cgit v1.2.3 From 2cc1b7c742aa1195e6f9cce2c9d0ded804af9dea Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 13 Jun 2008 09:15:07 -0600 Subject: gallium: another attempt at fixing libglapi.a compilation for DRI builds If building for DRI, make a dummy/empty libglapi.a for now. --- src/mesa/Makefile | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index affd81b828..3ce30a1e25 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -19,12 +19,7 @@ include sources -# XXX work-around glapi.c build failure for DRI builds: -ifeq ($(WINDOW_SYSTEM), dri) -GLAPI_LIB = -else GLAPI_LIB = libglapi.a -endif # Default: build dependencies, then asm_subdirs, then convenience @@ -42,7 +37,11 @@ libmesa.a: $(MESA_OBJECTS) asm_subdirs # Make archive of gl* API dispatcher functions only $(GLAPI_LIB): $(GLAPI_OBJECTS) asm_subdirs - @ $(TOP)/bin/mklib -o glapi -static $(GLAPI_OBJECTS) + if [ "${WINDOW_SYSTEM}" = "dri" ] ; then \ + $(TOP)/bin/mklib -o glapi -static ; \ + else \ + $(TOP)/bin/mklib -o glapi -static $(GLAPI_OBJECTS) ; \ + fi ###################################################################### -- cgit v1.2.3 From c341094921da7c4ff30f81706279ead39cb3b812 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 13 Jun 2008 19:03:51 +0200 Subject: i915: Fix most of the clear problems trivial/clear-undefined still fails --- src/gallium/drivers/i915simple/i915_clear.c | 1 + src/gallium/drivers/i915simple/i915_texture.c | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_clear.c b/src/gallium/drivers/i915simple/i915_clear.c index cde69daacc..8a2d3ca43f 100644 --- a/src/gallium/drivers/i915simple/i915_clear.c +++ b/src/gallium/drivers/i915simple/i915_clear.c @@ -44,4 +44,5 @@ i915_clear(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { pipe->surface_fill(pipe, ps, 0, 0, ps->width, ps->height, clearValue); + ps->status = PIPE_SURFACE_STATUS_DEFINED; } diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 16354dce50..7a1665d48b 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -635,6 +635,7 @@ i915_get_tex_surface(struct pipe_screen *screen, ps->pitch = tex->pitch; ps->offset = offset; ps->usage = flags; + ps->status = PIPE_SURFACE_STATUS_DEFINED; } return ps; } -- cgit v1.2.3 From d1397fd779ce63416bb59662a3c96f9c0a6ef7f5 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 13 Jun 2008 19:07:29 +0200 Subject: i915: Fix offsets not being used for surface targets --- src/gallium/drivers/i915simple/i915_state_emit.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index de0c1a787a..480b0b54c4 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -222,7 +222,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_RELOC(cbuf_surface->buffer, I915_BUFFER_ACCESS_WRITE, - 0); + cbuf_surface->offset); } /* What happens if no zbuf?? @@ -238,7 +238,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) OUT_RELOC(depth_surface->buffer, I915_BUFFER_ACCESS_WRITE, - 0); + depth_surface->offset); } { -- cgit v1.2.3 From 907b753ae62db8c2b266b18ed122eb00e1d66587 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 13 Jun 2008 19:13:35 +0200 Subject: mesa: Fix libglapi.a build for dri --- src/mesa/Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 3ce30a1e25..2d0a6e0167 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -37,8 +37,8 @@ libmesa.a: $(MESA_OBJECTS) asm_subdirs # Make archive of gl* API dispatcher functions only $(GLAPI_LIB): $(GLAPI_OBJECTS) asm_subdirs - if [ "${WINDOW_SYSTEM}" = "dri" ] ; then \ - $(TOP)/bin/mklib -o glapi -static ; \ + @if [ "${WINDOW_SYSTEM}" = "dri" ] ; then \ + touch libglapi.a ; \ else \ $(TOP)/bin/mklib -o glapi -static $(GLAPI_OBJECTS) ; \ fi -- cgit v1.2.3 From 3da77b33bb0093ff27c16833ed93a3a114c3e95e Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 13 Jun 2008 20:46:14 +0200 Subject: i915: Improved the not used tile code --- src/gallium/drivers/i915simple/i915_state_emit.c | 21 ++++++++++----- src/gallium/drivers/i915simple/i915_texture.c | 33 +++++++++++++++--------- 2 files changed, 36 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index 480b0b54c4..bc801a82f0 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -211,14 +211,18 @@ i915_emit_hardware_state(struct i915_context *i915 ) struct pipe_surface *depth_surface = i915->framebuffer.zsbuf; if (cbuf_surface) { - unsigned pitch = (cbuf_surface->pitch * cbuf_surface->cpp); + unsigned cpitch = (cbuf_surface->pitch * cbuf_surface->cpp); + unsigned ctile = BUF_3D_USE_FENCE; +#if 0 + if (!((cpitch - 1) & cpitch) && cpitch >= 512) + ctile = BUF_3D_TILED_SURFACE; +#endif OUT_BATCH(_3DSTATE_BUF_INFO_CMD); OUT_BATCH(BUF_3D_ID_COLOR_BACK | - BUF_3D_PITCH(pitch) | /* pitch in bytes */ -// BUF_3D_TILED_SURFACE); /* JB: Used to force tileing */ - BUF_3D_USE_FENCE); + BUF_3D_PITCH(cpitch) | /* pitch in bytes */ + ctile); OUT_RELOC(cbuf_surface->buffer, I915_BUFFER_ACCESS_WRITE, @@ -229,12 +233,17 @@ i915_emit_hardware_state(struct i915_context *i915 ) */ if (depth_surface) { unsigned zpitch = (depth_surface->pitch * depth_surface->cpp); - + unsigned ztile = BUF_3D_USE_FENCE; +#if 0 + if (!((zpitch - 1) & zpitch) && zpitch >= 512) + ztile = BUF_3D_TILED_SURFACE; +#endif + OUT_BATCH(_3DSTATE_BUF_INFO_CMD); OUT_BATCH(BUF_3D_ID_DEPTH | BUF_3D_PITCH(zpitch) | /* pitch in bytes */ - BUF_3D_USE_FENCE); + ztile); OUT_RELOC(depth_surface->buffer, I915_BUFFER_ACCESS_WRITE, diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 7a1665d48b..d591b09e3c 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -104,16 +104,14 @@ i915_miptree_set_image_offset(struct i915_texture *tex, */ } -#if 0 static unsigned power_of_two(unsigned x) { - int value = 1; + unsigned value = 1; while (value <= x) value = value << 1; return value; } -#endif static unsigned round_up(unsigned n, unsigned multiple) @@ -125,8 +123,7 @@ round_up(unsigned n, unsigned multiple) * Special case to deal with display targets. */ static boolean -i915_displaytarget_layout(struct pipe_screen *screen, - struct i915_texture *tex) +i915_displaytarget_layout(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; @@ -139,13 +136,19 @@ i915_displaytarget_layout(struct pipe_screen *screen, 1 ); i915_miptree_set_image_offset( tex, 0, 0, 0, 0 ); -#if 0 - tex->pitch = MAX2(512, power_of_two(tex->base.width[0] * pt->cpp)) / pt->cpp; - tex->total_height = round_up(tex->base.height[0], 8); -#else - tex->pitch = round_up(tex->base.width[0], 64 / pt->cpp); - tex->total_height = tex->base.height[0]; -#endif + if (tex->base.width[0] >= 128) { + tex->pitch = power_of_two(tex->base.width[0] * pt->cpp) / pt->cpp; + tex->total_height = round_up(tex->base.height[0], 8); + } else { + tex->pitch = round_up(tex->base.width[0], 64 / pt->cpp); + tex->total_height = tex->base.height[0]; + } + +/* + printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, + tex->base.width[0], tex->base.height[0], pt->cpp, + tex->pitch, tex->total_height, tex->pitch * tex->total_height * 4); +*/ return 1; } @@ -161,6 +164,12 @@ i945_miptree_layout_2d( struct i915_texture *tex ) unsigned width = pt->width[0]; unsigned height = pt->height[0]; +#if 0 /* used for tiled display targets */ + if (pt->last_level == 0 && pt->cpp == 4) + if (i915_displaytarget_layout(tex)) + return; +#endif + tex->pitch = pt->width[0]; /* May need to adjust pitch to accomodate the placement of -- cgit v1.2.3 From be16f04e76be28dd0bab039b21b63ce0752de336 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 13 Jun 2008 13:56:31 -0600 Subject: egl: added a debug msg --- src/egl/drivers/xdri/egl_xdri.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c index 60753a1c1f..df80c6a1c4 100644 --- a/src/egl/drivers/xdri/egl_xdri.c +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -675,6 +675,7 @@ xdri_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, xdri_config->mode, renderType, shared, &xdri_ctx->driContext); if (!xdri_ctx->driContext.private) { + _eglLog(_EGL_DEBUG, "driScreen.createNewContext failed"); free(xdri_ctx); return EGL_NO_CONTEXT; } -- cgit v1.2.3 From 2b4e2841a70ddba683158b4310b19c98037a2337 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 13 Jun 2008 13:56:53 -0600 Subject: mesa: check FEATURE_ARB_occlusion_query --- src/mesa/main/context.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 8f22f6bb78..e2f3ffb17b 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -126,7 +126,9 @@ #include "pixelstore.h" #include "points.h" #include "polygon.h" +#if FEATURE_ARB_occlusion_query #include "queryobj.h" +#endif #if FEATURE_drawpix #include "rastpos.h" #endif @@ -1013,7 +1015,9 @@ init_attrib_groups(GLcontext *ctx) _mesa_init_point( ctx ); _mesa_init_polygon( ctx ); _mesa_init_program( ctx ); +#if FEATURE_ARB_occlusion_query _mesa_init_query( ctx ); +#endif #if FEATURE_drawpix _mesa_init_rastpos( ctx ); #endif @@ -1213,6 +1217,7 @@ _mesa_create_context(const GLvisual *visual, { GLcontext *ctx; + printf("***** enter %s\n", __FUNCTION__); ASSERT(visual); /*ASSERT(driverContext);*/ @@ -1273,7 +1278,9 @@ _mesa_free_context_data( GLcontext *ctx ) #endif _mesa_free_program_data(ctx); _mesa_free_shader_state(ctx); +#if FEATURE_ARB_occlusion_query _mesa_free_query_data(ctx); +#endif #if FEATURE_ARB_vertex_buffer_object _mesa_delete_buffer_object(ctx, ctx->Array.NullBufferObj); -- cgit v1.2.3 From e9a6832737e17fd41d1f9e660239bd0bd2355b0b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 13 Jun 2008 14:13:25 -0600 Subject: mesa: remove some temp debug code --- src/mesa/main/context.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index e2f3ffb17b..086272ca01 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1217,7 +1217,6 @@ _mesa_create_context(const GLvisual *visual, { GLcontext *ctx; - printf("***** enter %s\n", __FUNCTION__); ASSERT(visual); /*ASSERT(driverContext);*/ -- cgit v1.2.3 From 8b11fa4d4496032246b33182b9285c1181d41f1f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 13 Jun 2008 16:45:15 -0600 Subject: mesa: move some glapi bits around Move _glapi_proc typedef from glapitable.h to glapi.h Also, don't include glapitable.h from glapi.h Before we were including the huge glapitable.h file in every .c file. --- src/mesa/glapi/dispatch.h | 3 +++ src/mesa/glapi/gl_table.py | 1 - src/mesa/glapi/glapi.h | 5 ++++- src/mesa/glapi/glapitable.h | 1 - src/mesa/main/blend.c | 1 + src/mesa/main/context.c | 1 + src/mesa/main/mtypes.h | 3 +-- 7 files changed, 10 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/glapi/dispatch.h b/src/mesa/glapi/dispatch.h index 7123156085..98f654f402 100644 --- a/src/mesa/glapi/dispatch.h +++ b/src/mesa/glapi/dispatch.h @@ -28,6 +28,9 @@ #if !defined( _DISPATCH_H_ ) # define _DISPATCH_H_ +#include "glapitable.h" + + /** * \file dispatch.h * Macros for handling GL dispatch tables. diff --git a/src/mesa/glapi/gl_table.py b/src/mesa/glapi/gl_table.py index 69f7bd7c7b..7023a4b71a 100644 --- a/src/mesa/glapi/gl_table.py +++ b/src/mesa/glapi/gl_table.py @@ -56,7 +56,6 @@ class PrintGlTable(gl_XML.gl_print_base): print '# define GLAPIENTRYP GLAPIENTRY *' print '#endif' print '' - print 'typedef void (*_glapi_proc)(void); /* generic function pointer */' print '' print 'struct _glapi_table' print '{' diff --git a/src/mesa/glapi/glapi.h b/src/mesa/glapi/glapi.h index ddfb1cffb9..20d35769cf 100644 --- a/src/mesa/glapi/glapi.h +++ b/src/mesa/glapi/glapi.h @@ -46,10 +46,13 @@ #include "GL/gl.h" -#include "glapitable.h" #include "glthread.h" +struct _glapi_table; + +typedef void (*_glapi_proc)(void); /* generic function pointer */ + typedef void (*_glapi_warning_func)(void *ctx, const char *str, ...); diff --git a/src/mesa/glapi/glapitable.h b/src/mesa/glapi/glapitable.h index 48941f5590..5d9d40a8a2 100644 --- a/src/mesa/glapi/glapitable.h +++ b/src/mesa/glapi/glapitable.h @@ -37,7 +37,6 @@ # define GLAPIENTRYP GLAPIENTRY * #endif -typedef void (*_glapi_proc)(void); /* generic function pointer */ struct _glapi_table { diff --git a/src/mesa/main/blend.c b/src/mesa/main/blend.c index 81bd4c2f32..742247f8e2 100644 --- a/src/mesa/main/blend.c +++ b/src/mesa/main/blend.c @@ -36,6 +36,7 @@ #include "enums.h" #include "macros.h" #include "mtypes.h" +#include "glapi/glapitable.h" /** diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 086272ca01..dcc518e9fd 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -146,6 +146,7 @@ #include "vtxfmt.h" #include "glapi/glthread.h" #include "glapi/glapioffsets.h" +#include "glapi/glapitable.h" #include "shader/program.h" #include "shader/shader_api.h" #include "shader/atifragshader.h" diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 463142fe39..8a6c84368a 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -38,8 +38,7 @@ #include "glheader.h" #include /* __GLcontextModes (GLvisual) */ #include "config.h" /* Hardwired parameters */ -#include "glapi/glapitable.h" -#include "glapi/glthread.h" +#include "glapi/glapi.h" #include "math/m_matrix.h" /* GLmatrix */ #include "bitset.h" -- cgit v1.2.3 From 64d854ebf7074f4a86bb9d45bb2e1003f86a86a1 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 13 Jun 2008 17:22:11 +0200 Subject: util: Use pf_get_size(). --- src/gallium/auxiliary/util/u_blit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 999a3e5099..7b9415d49a 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -281,7 +281,7 @@ util_blit_pixels(struct blit_state *ctx, texTemp.height[0] = srcH; texTemp.depth[0] = 1; texTemp.compressed = 0; - texTemp.cpp = pf_get_bits(src->format) / 8; + texTemp.cpp = pf_get_size(src->format); tex = screen->texture_create(screen, &texTemp); if (!tex) -- cgit v1.2.3 From d9d1e39d95fef4a8da15147956ff0c3e0a188b5b Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 17 Jun 2008 15:47:01 +0200 Subject: i915: Max and Min lod now works --- src/gallium/drivers/i915simple/i915_context.h | 2 + src/gallium/drivers/i915simple/i915_state.c | 10 ++- .../drivers/i915simple/i915_state_sampler.c | 84 ++++++++++++++++++---- 3 files changed, 79 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 2ee0381648..892a88fd2c 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -178,6 +178,8 @@ struct i915_rasterizer_state { struct i915_sampler_state { unsigned state[3]; const struct pipe_sampler_state *templ; + unsigned minlod; + unsigned maxlod; }; diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index 39d5d5e9d6..dbb33f2695 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -251,11 +251,17 @@ i915_create_sampler_state(struct pipe_context *pipe, if (sampler->normalized_coords) cso->state[1] |= SS3_NORMALIZED_COORDS; - if (0) /* XXX not tested yet */ { int minlod = (int) (16.0 * sampler->min_lod); + int maxlod = (int) (16.0 * sampler->max_lod); minlod = CLAMP(minlod, 0, 16 * 11); - cso->state[1] |= (minlod << SS3_MIN_LOD_SHIFT); + maxlod = CLAMP(maxlod, 0, 16 * 11); + + if (minlod > maxlod) + maxlod = minlod; + + cso->minlod = minlod; + cso->maxlod = maxlod; } { diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index 982eec4a1b..fbdc1a3b68 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -35,6 +35,35 @@ #include "i915_state.h" +/* + * A note about min_lod & max_lod. + * + * There is a circular dependancy between the sampler state + * and the map state to be submitted to hw. + * + * Two condition must be meet: + * min_lod =< max_lod == true + * max_lod =< last_level == true + * + * + * This is all fine and dandy if it where for the fact that max_lod + * is set on the map state instead of the sampler state. That is + * the max_lod we submit on map is: + * max_lod = MIN2(last_level, max_lod); + * + * So we need to update the map state when we change samplers and + * we need to be change the sampler state when map state is changed. + * The first part is done by calling i915_update_texture in + * i915_update_samplers and the second part is done else where in + * code tracking the state changes. + */ + +static void +i915_update_texture(struct i915_context *i915, + uint unit, + const struct i915_texture *tex, + const struct i915_sampler_state *sampler, + uint state[6]); /** * Compute i915 texture sampling state. * @@ -50,6 +79,7 @@ static void update_sampler(struct i915_context *i915, unsigned state[3] ) { const struct pipe_texture *pt = &tex->base; + unsigned minlod, lastlod; /* Need to do this after updating the maps, which call the * intel_finalize_mipmap_tree and hence can update firstLevel: @@ -95,6 +125,15 @@ static void update_sampler(struct i915_context *i915, } #endif + /* See note at the top of file */ + minlod = sampler->minlod; + lastlod = pt->last_level << 4; + + if (lastlod < minlod) { + minlod = lastlod; + } + + state[1] |= (sampler->minlod << SS3_MIN_LOD_SHIFT); state[1] |= (unit << SS3_TEXTUREMAP_INDEX_SHIFT); } @@ -112,18 +151,23 @@ void i915_update_samplers( struct i915_context *i915 ) /* could also examine the fragment program? */ if (i915->texture[unit]) { update_sampler( i915, - unit, - i915->sampler[unit], /* sampler state */ - i915->texture[unit], /* texture */ - i915->current.sampler[unit] /* the result */ - ); - - i915->current.sampler_enable_nr++; - i915->current.sampler_enable_flags |= (1 << unit); + unit, + i915->sampler[unit], /* sampler state */ + i915->texture[unit], /* texture */ + i915->current.sampler[unit] /* the result */ + ); + i915_update_texture( i915, + unit, + i915->texture[unit], /* texture */ + i915->sampler[unit], /* sampler state */ + i915->current.texbuffer[unit] ); + + i915->current.sampler_enable_nr++; + i915->current.sampler_enable_flags |= (1 << unit); } } - i915->hardware_dirty |= I915_HW_SAMPLER; + i915->hardware_dirty |= I915_HW_SAMPLER | I915_HW_MAP; } @@ -179,14 +223,17 @@ translate_texture_format(enum pipe_format pipeFormat) static void -i915_update_texture(struct i915_context *i915, uint unit, +i915_update_texture(struct i915_context *i915, + uint unit, + const struct i915_texture *tex, + const struct i915_sampler_state *sampler, uint state[6]) { - const struct i915_texture *tex = i915->texture[unit]; const struct pipe_texture *pt = &tex->base; uint format, pitch; const uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; const uint num_levels = pt->last_level; + unsigned max_lod = num_levels * 4; assert(tex); assert(width); @@ -207,16 +254,19 @@ i915_update_texture(struct i915_context *i915, uint unit, | MS3_USE_FENCE_REGS); /* - * XXX sampler->max_lod should be used to program the MAX_LOD field below. - * Also, when min_filter != mag_filter and there's just one mipmap level, + * XXX When min_filter != mag_filter and there's just one mipmap level, * set max_lod = 1 to make sure i915 chooses between min/mag filtering. */ + /* See note at the top of file */ + if (max_lod > (sampler->maxlod >> 2)) + max_lod = sampler->maxlod >> 2; + /* MS4 state */ state[1] = ((((pitch / 4) - 1) << MS4_PITCH_SHIFT) | MS4_CUBE_FACE_ENA_MASK - | ((num_levels * 4) << MS4_MAX_LOD_SHIFT) + | ((max_lod) << MS4_MAX_LOD_SHIFT) | ((depth - 1) << MS4_VOLUME_DEPTH_SHIFT)); } @@ -231,7 +281,11 @@ i915_update_textures(struct i915_context *i915) /* determine unit enable/disable by looking for a bound texture */ /* could also examine the fragment program? */ if (i915->texture[unit]) { - i915_update_texture(i915, unit, i915->current.texbuffer[unit]); + i915_update_texture( i915, + unit, + i915->texture[unit], /* texture */ + i915->sampler[unit], /* sampler state */ + i915->current.texbuffer[unit] ); } } -- cgit v1.2.3 From 5805a9e32881831f320bcd1a95fbfacf17aef5d5 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 17 Jun 2008 17:46:45 +0200 Subject: i915: Made region pool a slabpool --- src/gallium/winsys/common/intel_drm/intel_be_device.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/common/intel_drm/intel_be_device.c b/src/gallium/winsys/common/intel_drm/intel_be_device.c index 0fc1894eaa..3eed0fe410 100644 --- a/src/gallium/winsys/common/intel_drm/intel_be_device.c +++ b/src/gallium/winsys/common/intel_drm/intel_be_device.c @@ -220,7 +220,18 @@ intel_be_init_device(struct intel_be_device *dev, int fd) dev->mallocPool = driMallocPoolInit(); dev->staticPool = driDRMPoolInit(dev->fd); - dev->regionPool = driDRMPoolInit(dev->fd); + /* Sizes: 64 128 256 512 1024 2048 4096 8192 16384 32768 */ + dev->regionPool = driSlabPoolInit(dev->fd, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + 64, + 10, 120, 4096 * 64, 0, + dev->fMan); + dev->vertexPool = driSlabPoolInit(dev->fd, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE | -- cgit v1.2.3 From 3ccbde627edb420071b08a830dd58ed5daf82ffa Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 10:11:53 -0600 Subject: mesa: make mm.c use unsigned ints for offsets. If you have a GPU using this code and it has the offsets up in this space, this fails. cherry-picked from master --- src/mesa/main/mm.c | 8 ++++---- src/mesa/main/mm.h | 5 +++-- 2 files changed, 7 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mm.c b/src/mesa/main/mm.c index 846c329c70..fb7809ed22 100644 --- a/src/mesa/main/mm.c +++ b/src/mesa/main/mm.c @@ -53,7 +53,7 @@ mmDumpMemInfo(const struct mem_block *heap) } struct mem_block * -mmInit(int ofs, int size) +mmInit(unsigned int ofs, int size) { struct mem_block *heap, *block; @@ -91,7 +91,7 @@ mmInit(int ofs, int size) static struct mem_block * SliceBlock(struct mem_block *p, - int startofs, int size, + unsigned int startofs, int size, int reserved, int alignment) { struct mem_block *newblock; @@ -164,8 +164,8 @@ mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) { struct mem_block *p; const int mask = (1 << align2)-1; - int startofs = 0; - int endofs; + unsigned int startofs = 0; + unsigned int endofs; if (!heap || align2 < 0 || size <= 0) return NULL; diff --git a/src/mesa/main/mm.h b/src/mesa/main/mm.h index 26d59fff13..5ad3ffd6d1 100644 --- a/src/mesa/main/mm.h +++ b/src/mesa/main/mm.h @@ -39,7 +39,8 @@ struct mem_block { struct mem_block *next, *prev; struct mem_block *next_free, *prev_free; struct mem_block *heap; - int ofs,size; + unsigned int ofs; + int size; unsigned int free:1; unsigned int reserved:1; }; @@ -50,7 +51,7 @@ struct mem_block { * input: total size in bytes * return: a heap pointer if OK, NULL if error */ -extern struct mem_block *mmInit(int ofs, int size); +extern struct mem_block *mmInit(unsigned int ofs, int size); /** * Allocate 'size' bytes with 2^align2 bytes alignment, -- cgit v1.2.3 From 93deede760e5a1309f1b948d91eae3d27f59ed97 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 10:28:00 -0600 Subject: gallium: check if sampler->max_lod < sampler->min_lod --- src/mesa/state_tracker/st_atom_sampler.c | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 10283d31a1..9abd0b3531 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -153,6 +153,15 @@ update_samplers(struct st_context *st) sampler->min_lod = MAX2(0.0f, texobj->MinLod); sampler->max_lod = MIN2(texobj->MaxLevel - texobj->BaseLevel, texobj->MaxLod); + if (sampler->max_lod < sampler->min_lod) { + /* The GL spec doesn't seem to specify what to do in this case. + * Swap the values. + */ + float tmp = sampler->max_lod; + sampler->max_lod = sampler->min_lod; + sampler->min_lod = tmp; + assert(sampler->min_lod <= sampler->max_lod); + } sampler->border_color[0] = texobj->BorderColor[RCOMP]; sampler->border_color[1] = texobj->BorderColor[GCOMP]; -- cgit v1.2.3 From 318f00ae9f9bca783d1d3aa7700109402f3e52cb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 12:20:22 -0600 Subject: gallium: need to flush bitmap cache when raster color changes --- src/mesa/state_tracker/st_cb_bitmap.c | 20 ++++++++++++++------ 1 file changed, 14 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 9763eebe54..9e32ee2eb4 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -90,6 +90,8 @@ struct bitmap_cache /** Bounds of region used in window coords */ GLint xmin, ymin, xmax, ymax; + GLfloat color[4]; + struct pipe_texture *texture; struct pipe_surface *surf; @@ -429,7 +431,8 @@ setup_bitmap_vertex_data(struct st_context *st, static void draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, GLsizei width, GLsizei height, - struct pipe_texture *pt) + struct pipe_texture *pt, + const GLfloat *color) { struct st_context *st = ctx->st; struct pipe_context *pipe = ctx->st->pipe; @@ -505,7 +508,7 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* draw textured quad */ setup_bitmap_vertex_data(st, x, y, width, height, ctx->Current.RasterPos[2], - ctx->Current.RasterColor); + color); util_draw_vertex_buffer(pipe, st->bitmap.vbuf, PIPE_PRIM_TRIANGLE_FAN, @@ -592,7 +595,8 @@ st_flush_bitmap_cache(struct st_context *st) cache->ypos, st->ctx->Current.RasterPos[2], BITMAP_CACHE_WIDTH, BITMAP_CACHE_HEIGHT, - cache->texture); + cache->texture, + cache->color); } /* release/free the texture */ @@ -624,8 +628,10 @@ accum_bitmap(struct st_context *st, px = x - cache->xpos; /* pos in buffer */ py = y - cache->ypos; if (px < 0 || px + width > BITMAP_CACHE_WIDTH || - py < 0 || py + height > BITMAP_CACHE_HEIGHT) { - /* This bitmap would extend beyond cache bounds, + py < 0 || py + height > BITMAP_CACHE_HEIGHT || + !TEST_EQ_4V(st->ctx->Current.RasterColor, cache->color)) { + /* This bitmap would extend beyond cache bounds, or the bitmap + * color is changing * so flush and continue. */ st_flush_bitmap_cache(st); @@ -639,6 +645,7 @@ accum_bitmap(struct st_context *st, cache->xpos = x; cache->ypos = y - py; cache->empty = GL_FALSE; + COPY_4FV(cache->color, st->ctx->Current.RasterColor); } assert(px != -999); @@ -694,7 +701,8 @@ st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (pt) { assert(pt->target == PIPE_TEXTURE_2D); draw_bitmap_quad(ctx, x, y, ctx->Current.RasterPos[2], - width, height, pt); + width, height, pt, + st->ctx->Current.RasterColor); /* release/free the texture */ pipe_texture_reference(&pt, NULL); } -- cgit v1.2.3 From 4539410d71d5cf5c7d7344b1d3012a8a3e825a18 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 17 Jun 2008 21:43:41 +0200 Subject: gallium: Add facilities for mixed pipe formats. --- src/gallium/include/pipe/p_format.h | 48 ++++++++++++++++++++++++++++++++----- 1 file changed, 42 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index b1bad8ab0e..e7d63350ef 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -50,7 +50,7 @@ extern "C" { #define PIPE_FORMAT_LAYOUT_RGBAZS 0 #define PIPE_FORMAT_LAYOUT_YCBCR 1 #define PIPE_FORMAT_LAYOUT_DXT 2 /**< XXX temporary? */ - +#define PIPE_FORMAT_LAYOUT_MIXED 3 static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ { @@ -62,7 +62,7 @@ static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ */ /** - * Format component selectors for RGBAZS layout. + * Format component selectors for RGBAZS & MIXED layout. */ #define PIPE_FORMAT_COMP_R 0 #define PIPE_FORMAT_COMP_G 1 @@ -109,8 +109,6 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) return (f >> shift) & mask; } -/* XXX: The bit layout needs to be revised, can't currently encode 10-bit components. */ - #define pf_swizzle_x(f) pf_get(f, 2, 0x7) /**< PIPE_FORMAT_COMP_ */ #define pf_swizzle_y(f) pf_get(f, 5, 0x7) /**< PIPE_FORMAT_COMP_ */ #define pf_swizzle_z(f) pf_get(f, 8, 0x7) /**< PIPE_FORMAT_COMP_ */ @@ -166,6 +164,36 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) #define _PIPE_FORMAT_RGBAZS_64( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, TYPE )\ _PIPE_FORMAT_RGBAZS( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, 6, TYPE ) +typedef uint pipe_format_mixed_t; + +/* NOTE: Use pf_swizzle_* and pf_size_* macros for swizzles and sizes. + */ + +#define pf_mixed_sign_x(f) pf_get( f, 26, 0x1 ) /*< Sign of X */ +#define pf_mixed_sign_y(f) pf_get( f, 27, 0x1 ) /*< Sign of Y */ +#define pf_mixed_sign_z(f) pf_get( f, 28, 0x1 ) /*< Sign of Z */ +#define pf_mixed_sign_w(f) pf_get( f, 29, 0x1 ) /*< Sign of W */ +#define pf_mixed_sign_xyzw(f, i) pf_get( f, 26 + (i), 0x1 ) +#define pf_mixed_normalized(f) pf_get( f, 30, 0x1 ) /*< Type is NORM (1) or SCALED (0) */ +#define pf_mixed_scale8(f) pf_get( f, 31, 0x1 ) /*< Scale size by either one (0) or eight (1) */ + +/** + * Helper macro to encode the above structure into a 32-bit value. + */ +#define _PIPE_FORMAT_MIXED( SWZ, SIZEX, SIZEY, SIZEZ, SIZEW, SIGNX, SIGNY, SIGNZ, SIGNW, NORMALIZED, SCALE8 ) (\ + (PIPE_FORMAT_LAYOUT_MIXED << 0) |\ + ((SWZ) << 2) |\ + ((SIZEX) << 14) |\ + ((SIZEY) << 17) |\ + ((SIZEZ) << 20) |\ + ((SIZEW) << 23) |\ + ((SIGNX) << 26) |\ + ((SIGNY) << 27) |\ + ((SIGNZ) << 28) |\ + ((SIGNW) << 29) |\ + ((NORMALIZED) << 30) |\ + ((SCALE8) << 31) ) + /** * Shorthand macro for common format swizzles. */ @@ -177,6 +205,7 @@ static INLINE uint pf_get(pipe_format_rgbazs_t f, uint shift, uint mask) #define _PIPE_FORMAT_ABGR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_A, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R ) #define _PIPE_FORMAT_BGRA _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_A ) #define _PIPE_FORMAT_1RGB _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_1, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_B ) +#define _PIPE_FORMAT_1BGR _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_1, PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R ) #define _PIPE_FORMAT_BGR1 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_B, PIPE_FORMAT_COMP_G, PIPE_FORMAT_COMP_R, PIPE_FORMAT_COMP_1 ) #define _PIPE_FORMAT_0000 _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0 ) #define _PIPE_FORMAT_000R _PIPE_FORMAT_SWZ( PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_0, PIPE_FORMAT_COMP_R ) @@ -331,6 +360,11 @@ enum pipe_format { PIPE_FORMAT_R8G8B8A8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), PIPE_FORMAT_R8G8B8X8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), + /* mixed formats */ + PIPE_FORMAT_A8UB8UG8SR8S_NORM = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_ABGR, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1 ), + PIPE_FORMAT_X8UB8UG8SR8S_NORM = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_1BGR, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1 ), + PIPE_FORMAT_B6UG5SR5S_NORM = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_BGR1, 6, 5, 5, 0, 0, 1, 1, 0, 1, 0 ), + /* compressed formats */ PIPE_FORMAT_DXT1_RGB = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 0 ), PIPE_FORMAT_DXT1_RGBA = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 8 ), @@ -338,7 +372,6 @@ enum pipe_format { PIPE_FORMAT_DXT5_RGBA = _PIPE_FORMAT_DXT( 5, 8, 8, 8, 8 ) }; - /** * Builds pipe format name from format token. */ @@ -367,7 +400,9 @@ static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) else { size = 0; } - return size << pf_exp2(format); + if (pf_layout( format ) == PIPE_FORMAT_LAYOUT_RGBAZS) + return size << pf_exp2( format ); + return size << (pf_mixed_scale8( format ) * 3); } /** @@ -377,6 +412,7 @@ static INLINE uint pf_get_bits( enum pipe_format format ) { switch (pf_layout(format)) { case PIPE_FORMAT_LAYOUT_RGBAZS: + case PIPE_FORMAT_LAYOUT_MIXED: return pf_get_component_bits( format, PIPE_FORMAT_COMP_0 ) + pf_get_component_bits( format, PIPE_FORMAT_COMP_1 ) + -- cgit v1.2.3 From 3bf8fb64862b882bff8b372f3c2ce08ea8d401a7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 16:43:48 -0600 Subject: mesa: fix inconsistent use of GL_UNSIGNED_INT vs. GL_UNSIGNED_INT_24_8_EXT for Z unpacking --- src/mesa/main/image.c | 2 +- src/mesa/main/texstore.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c index a6ac821251..52c4999e16 100644 --- a/src/mesa/main/image.c +++ b/src/mesa/main/image.c @@ -4447,7 +4447,7 @@ _mesa_unpack_depth_span( const GLcontext *ctx, GLuint n, DEPTH_VALUES(GLuint, UINT_TO_FLOAT); break; case GL_UNSIGNED_INT_24_8_EXT: /* GL_EXT_packed_depth_stencil */ - if (dstType == GL_UNSIGNED_INT && + if (dstType == GL_UNSIGNED_INT_24_8_EXT && depthMax == 0xffffff && ctx->Pixel.DepthScale == 1.0 && ctx->Pixel.DepthBias == 0.0) { diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index 5ac6116950..562e07a0f5 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -2336,7 +2336,7 @@ _mesa_texstore_ycbcr(TEXSTORE_PARAMS) GLboolean _mesa_texstore_z24_s8(TEXSTORE_PARAMS) { - const GLuint depthScale = 0xffffff; + const GLfloat depthScale = (GLfloat) 0xffffff; ASSERT(dstFormat == &_mesa_texformat_z24_s8); ASSERT(srcFormat == GL_DEPTH_STENCIL_EXT); @@ -2374,7 +2374,7 @@ _mesa_texstore_z24_s8(TEXSTORE_PARAMS) GLint i; /* the 24 depth bits will be in the high position: */ _mesa_unpack_depth_span(ctx, srcWidth, - GL_UNSIGNED_INT, /* dst type */ + GL_UNSIGNED_INT_24_8_EXT, /* dst type */ dstRow, /* dst addr */ depthScale, srcType, src, srcPacking); -- cgit v1.2.3 From a1524162bf838920ad965cd44ead97da29408e50 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 18 Jun 2008 01:39:46 +0200 Subject: mesa: Added _mesa_texstore_s8_z24 --- src/mesa/main/texformat.c | 4 ---- src/mesa/main/texstore.c | 54 +++++++++++++++++++++++++++++++++++++++++++++++ src/mesa/main/texstore.h | 1 + 3 files changed, 55 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c index ea55002f58..11e7755960 100644 --- a/src/mesa/main/texformat.c +++ b/src/mesa/main/texformat.c @@ -1221,11 +1221,7 @@ const struct gl_texture_format _mesa_texformat_s8_z24 = { 24, /* DepthBits */ 8, /* StencilBits */ 4, /* TexelBytes */ -#if 0 _mesa_texstore_s8_z24, /* StoreTexImageFunc */ -#else - _mesa_texstore_z24_s8, /* StoreTexImageFunc */ -#endif NULL, /* FetchTexel1D */ NULL, /* FetchTexel2D */ NULL, /* FetchTexel3D */ diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index 562e07a0f5..f3629cb0d1 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -2397,6 +2397,60 @@ _mesa_texstore_z24_s8(TEXSTORE_PARAMS) } +/** + * Store a combined depth/stencil texture image. + */ +GLboolean +_mesa_texstore_s8_z24(TEXSTORE_PARAMS) +{ + const GLuint depthScale = 0xffffff; + + ASSERT(dstFormat == &_mesa_texformat_s8_z24); + ASSERT(srcFormat == GL_DEPTH_STENCIL_EXT); + ASSERT(srcType == GL_UNSIGNED_INT_24_8_EXT); + + + /* general path */ + const GLint srcRowStride + = _mesa_image_row_stride(srcPacking, srcWidth, srcFormat, srcType) + / sizeof(GLuint); + GLint img, row; + + for (img = 0; img < srcDepth; img++) { + GLuint *dstRow = (GLuint *) dstAddr + + dstImageOffsets[dstZoffset + img] + + dstYoffset * dstRowStride / sizeof(GLuint) + + dstXoffset; + const GLuint *src + = (const GLuint *) _mesa_image_address(dims, srcPacking, srcAddr, + srcWidth, srcHeight, + srcFormat, srcType, + img, 0, 0); + for (row = 0; row < srcHeight; row++) { + GLubyte stencil[MAX_WIDTH]; + GLint i; + /* the 24 depth bits will be in the high position: */ + _mesa_unpack_depth_span(ctx, srcWidth, + GL_UNSIGNED_INT, /* dst type */ + dstRow, /* dst addr */ + depthScale, + srcType, src, srcPacking); + /* get the 8-bit stencil values */ + _mesa_unpack_stencil_span(ctx, srcWidth, + GL_UNSIGNED_BYTE, /* dst type */ + stencil, /* dst addr */ + srcType, src, srcPacking, + ctx->_ImageTransferState); + /* merge stencil values into depth values */ + for (i = 0; i < srcWidth; i++) + dstRow[i] = stencil[i] << 24; + + src += srcRowStride; + dstRow += dstRowStride / sizeof(GLuint); + } + } + return GL_TRUE; +} /** * Store an image in any of the formats: diff --git a/src/mesa/main/texstore.h b/src/mesa/main/texstore.h index da0c7cb78e..c5813fbeee 100644 --- a/src/mesa/main/texstore.h +++ b/src/mesa/main/texstore.h @@ -57,6 +57,7 @@ extern GLboolean _mesa_texstore_a8(TEXSTORE_PARAMS); extern GLboolean _mesa_texstore_ci8(TEXSTORE_PARAMS); extern GLboolean _mesa_texstore_ycbcr(TEXSTORE_PARAMS); extern GLboolean _mesa_texstore_z24_s8(TEXSTORE_PARAMS); +extern GLboolean _mesa_texstore_s8_z24(TEXSTORE_PARAMS); extern GLboolean _mesa_texstore_z16(TEXSTORE_PARAMS); extern GLboolean _mesa_texstore_z32(TEXSTORE_PARAMS); extern GLboolean _mesa_texstore_rgba_float32(TEXSTORE_PARAMS); -- cgit v1.2.3 From 081c05605f1c308c35fcf4168aac09fbf3c0a108 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 18 Jun 2008 01:45:21 +0200 Subject: i915: Fix for s8_z24 textures not being shown --- src/gallium/drivers/i915simple/i915_state_sampler.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index fbdc1a3b68..24440843f3 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -212,7 +212,7 @@ translate_texture_format(enum pipe_format pipeFormat) return (MAPSURF_COMPRESSED | MT_COMPRESS_DXT4_5); #endif case PIPE_FORMAT_S8Z24_UNORM: - return (MAPSURF_32BIT | MT_32BIT_xL824); + return (MAPSURF_32BIT | MT_32BIT_xI824); default: debug_printf("i915: translate_texture_format() bad image format %x\n", pipeFormat); -- cgit v1.2.3 From 28ac7d37fe9576428417351bf1acd180b179502a Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 18 Jun 2008 08:51:38 +0200 Subject: gallium: Remove PIPE_FORMAT_A8UB8UG8SR8S_NORM definition. --- src/gallium/include/pipe/p_format.h | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index e7d63350ef..579fdb2957 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -361,7 +361,6 @@ enum pipe_format { PIPE_FORMAT_R8G8B8X8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), /* mixed formats */ - PIPE_FORMAT_A8UB8UG8SR8S_NORM = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_ABGR, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1 ), PIPE_FORMAT_X8UB8UG8SR8S_NORM = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_1BGR, 1, 1, 1, 1, 0, 0, 1, 1, 1, 1 ), PIPE_FORMAT_B6UG5SR5S_NORM = _PIPE_FORMAT_MIXED( _PIPE_FORMAT_BGR1, 6, 5, 5, 0, 0, 1, 1, 0, 1, 0 ), -- cgit v1.2.3 From 666b771e512a7c91fa43544afec61bda63edc240 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 18 Jun 2008 11:38:12 +0200 Subject: mesa: _mesa_texstore_s8_z24 now supports depth only uploads --- src/mesa/main/texstore.c | 96 +++++++++++++++++++++++++++++++----------------- 1 file changed, 62 insertions(+), 34 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index f3629cb0d1..d7bfb69443 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -2406,47 +2406,75 @@ _mesa_texstore_s8_z24(TEXSTORE_PARAMS) const GLuint depthScale = 0xffffff; ASSERT(dstFormat == &_mesa_texformat_s8_z24); - ASSERT(srcFormat == GL_DEPTH_STENCIL_EXT); - ASSERT(srcType == GL_UNSIGNED_INT_24_8_EXT); + ASSERT(srcFormat == GL_DEPTH_STENCIL_EXT || srcFormat == GL_DEPTH_COMPONENT); + ASSERT(srcFormat != GL_DEPTH_STENCIL_EXT || srcType == GL_UNSIGNED_INT_24_8_EXT); - - /* general path */ const GLint srcRowStride = _mesa_image_row_stride(srcPacking, srcWidth, srcFormat, srcType) / sizeof(GLuint); GLint img, row; - for (img = 0; img < srcDepth; img++) { - GLuint *dstRow = (GLuint *) dstAddr - + dstImageOffsets[dstZoffset + img] - + dstYoffset * dstRowStride / sizeof(GLuint) - + dstXoffset; - const GLuint *src - = (const GLuint *) _mesa_image_address(dims, srcPacking, srcAddr, - srcWidth, srcHeight, - srcFormat, srcType, - img, 0, 0); - for (row = 0; row < srcHeight; row++) { - GLubyte stencil[MAX_WIDTH]; - GLint i; - /* the 24 depth bits will be in the high position: */ - _mesa_unpack_depth_span(ctx, srcWidth, - GL_UNSIGNED_INT, /* dst type */ - dstRow, /* dst addr */ - depthScale, - srcType, src, srcPacking); - /* get the 8-bit stencil values */ - _mesa_unpack_stencil_span(ctx, srcWidth, - GL_UNSIGNED_BYTE, /* dst type */ - stencil, /* dst addr */ - srcType, src, srcPacking, - ctx->_ImageTransferState); - /* merge stencil values into depth values */ - for (i = 0; i < srcWidth; i++) - dstRow[i] = stencil[i] << 24; + /* Incase we only upload depth we need to preserve the stencil */ + if (srcFormat == GL_DEPTH_COMPONENT) { + for (img = 0; img < srcDepth; img++) { + GLuint *dstRow = (GLuint *) dstAddr + + dstImageOffsets[dstZoffset + img] + + dstYoffset * dstRowStride / sizeof(GLuint) + + dstXoffset; + const GLuint *src + = (const GLuint *) _mesa_image_address(dims, srcPacking, srcAddr, + srcWidth, srcHeight, + srcFormat, srcType, + img, 0, 0); + for (row = 0; row < srcHeight; row++) { + GLuint depth[MAX_WIDTH]; + GLint i; + _mesa_unpack_depth_span(ctx, srcWidth, + GL_UNSIGNED_INT, /* dst type */ + depth, /* dst addr */ + depthScale, + srcType, src, srcPacking); - src += srcRowStride; - dstRow += dstRowStride / sizeof(GLuint); + for (i = 0; i < srcWidth; i++) + dstRow[i] = depth[i] | (dstRow[i] & 0xFF000000); + + src += srcRowStride; + dstRow += dstRowStride / sizeof(GLuint); + } + } + } else { + for (img = 0; img < srcDepth; img++) { + GLuint *dstRow = (GLuint *) dstAddr + + dstImageOffsets[dstZoffset + img] + + dstYoffset * dstRowStride / sizeof(GLuint) + + dstXoffset; + const GLuint *src + = (const GLuint *) _mesa_image_address(dims, srcPacking, srcAddr, + srcWidth, srcHeight, + srcFormat, srcType, + img, 0, 0); + for (row = 0; row < srcHeight; row++) { + GLubyte stencil[MAX_WIDTH]; + GLint i; + /* the 24 depth bits will be in the high position: */ + _mesa_unpack_depth_span(ctx, srcWidth, + GL_UNSIGNED_INT, /* dst type */ + dstRow, /* dst addr */ + depthScale, + srcType, src, srcPacking); + /* get the 8-bit stencil values */ + _mesa_unpack_stencil_span(ctx, srcWidth, + GL_UNSIGNED_BYTE, /* dst type */ + stencil, /* dst addr */ + srcType, src, srcPacking, + ctx->_ImageTransferState); + /* merge stencil values into depth values */ + for (i = 0; i < srcWidth; i++) + dstRow[i] = stencil[i] << 24; + + src += srcRowStride; + dstRow += dstRowStride / sizeof(GLuint); + } } } return GL_TRUE; -- cgit v1.2.3 From c366fd83b617db6c8c064802ff4bf120d654507d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 11:29:59 -0600 Subject: mesa: add parenthesis --- src/mesa/main/dd.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index 37ef2a865b..a24021c63f 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -912,9 +912,9 @@ struct dd_function_table { void (*ValidateTnlModule)( GLcontext *ctx, GLuint new_state ); -#define PRIM_OUTSIDE_BEGIN_END GL_POLYGON+1 -#define PRIM_INSIDE_UNKNOWN_PRIM GL_POLYGON+2 -#define PRIM_UNKNOWN GL_POLYGON+3 +#define PRIM_OUTSIDE_BEGIN_END (GL_POLYGON+1) +#define PRIM_INSIDE_UNKNOWN_PRIM (GL_POLYGON+2) +#define PRIM_UNKNOWN (GL_POLYGON+3) /** * Set by the driver-supplied T&L engine. -- cgit v1.2.3 From 8a369b909a6648ae7a5a0c2dcb972a2f96f99a80 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 11:31:45 -0600 Subject: mesa: s/GL_POLYGON+1/PRIM_OUTSIDE_BEGIN_END/ --- src/mesa/vbo/vbo_exec_api.c | 10 +++++----- src/mesa/vbo/vbo_exec_draw.c | 2 +- 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index 5dd774d839..b263885388 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -67,7 +67,7 @@ static void vbo_exec_wrap_buffers( struct vbo_exec_context *exec ) GLuint last_begin = exec->vtx.prim[exec->vtx.prim_count-1].begin; GLuint last_count; - if (exec->ctx->Driver.CurrentExecPrimitive != GL_POLYGON+1) { + if (exec->ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) { GLint i = exec->vtx.prim_count - 1; assert(i >= 0); exec->vtx.prim[i].count = (exec->vtx.vert_count - @@ -89,7 +89,7 @@ static void vbo_exec_wrap_buffers( struct vbo_exec_context *exec ) */ assert(exec->vtx.prim_count == 0); - if (exec->ctx->Driver.CurrentExecPrimitive != GL_POLYGON+1) { + if (exec->ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) { exec->vtx.prim[0].mode = exec->ctx->Driver.CurrentExecPrimitive; exec->vtx.prim[0].start = 0; exec->vtx.prim[0].count = 0; @@ -503,7 +503,7 @@ static void GLAPIENTRY vbo_exec_Begin( GLenum mode ) { GET_CURRENT_CONTEXT( ctx ); - if (ctx->Driver.CurrentExecPrimitive == GL_POLYGON+1) { + if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END) { struct vbo_exec_context *exec = &vbo_context(ctx)->exec; int i; @@ -547,7 +547,7 @@ static void GLAPIENTRY vbo_exec_End( void ) { GET_CURRENT_CONTEXT( ctx ); - if (ctx->Driver.CurrentExecPrimitive != GL_POLYGON+1) { + if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) { struct vbo_exec_context *exec = &vbo_context(ctx)->exec; int idx = exec->vtx.vert_count; int i = exec->vtx.prim_count - 1; @@ -555,7 +555,7 @@ static void GLAPIENTRY vbo_exec_End( void ) exec->vtx.prim[i].end = 1; exec->vtx.prim[i].count = idx - exec->vtx.prim[i].start; - ctx->Driver.CurrentExecPrimitive = GL_POLYGON+1; + ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END; if (exec->vtx.prim_count == VBO_MAX_PRIM) vbo_exec_vtx_flush( exec ); diff --git a/src/mesa/vbo/vbo_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c index 6bb6849a6a..3609a7452a 100644 --- a/src/mesa/vbo/vbo_exec_draw.c +++ b/src/mesa/vbo/vbo_exec_draw.c @@ -128,7 +128,7 @@ static GLuint vbo_copy_vertices( struct vbo_exec_context *exec ) for (i = 0 ; i < ovf ; i++) _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) ); return i; - case GL_POLYGON+1: + case PRIM_OUTSIDE_BEGIN_END: return 0; default: assert(0); -- cgit v1.2.3 From 04996bcef6d65b81f7c2acc873befa2805296800 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 13:13:34 -0600 Subject: gallium: added FEATURE_x tests --- src/mesa/state_tracker/st_context.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index e20e0c08cb..9decca8a14 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -247,17 +247,27 @@ void st_init_driver_functions(struct dd_function_table *functions) { _mesa_init_glsl_driver_functions(functions); +#if FEATURE_accum st_init_accum_functions(functions); - st_init_bitmap_functions(functions); +#endif +#if FEATURE_EXT_framebuffer_blit st_init_blit_functions(functions); +#endif st_init_bufferobject_functions(functions); st_init_clear_functions(functions); +#if FEATURE_drawpix + st_init_bitmap_functions(functions); st_init_drawpixels_functions(functions); + st_init_rasterpos_functions(functions); +#endif st_init_fbo_functions(functions); +#if FEATURE_feedback st_init_feedback_functions(functions); +#endif st_init_program_functions(functions); +#if FEATURE_ARB_occlusion_query st_init_query_functions(functions); - st_init_rasterpos_functions(functions); +#endif st_init_readpixels_functions(functions); st_init_texture_functions(functions); st_init_flush_functions(functions); -- cgit v1.2.3 From 29d632efbc29200ed4f6f5e34dd10c857a2568aa Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 13:23:32 -0600 Subject: mesa: refactor: move GetProcAddress code from glapi.c into new glapi_getproc.c file --- src/mesa/glapi/glapi.c | 652 ++--------------------------------------- src/mesa/glapi/glapi.h | 12 +- src/mesa/glapi/glapi_getproc.c | 615 ++++++++++++++++++++++++++++++++++++++ src/mesa/sources | 1 + 4 files changed, 656 insertions(+), 624 deletions(-) create mode 100644 src/mesa/glapi/glapi_getproc.c (limited to 'src') diff --git a/src/mesa/glapi/glapi.c b/src/mesa/glapi/glapi.c index 47c5782273..ed59ddb0b7 100644 --- a/src/mesa/glapi/glapi.c +++ b/src/mesa/glapi/glapi.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -55,18 +55,12 @@ #include "glapioffsets.h" #include "glapitable.h" + /***** BEGIN NO-OP DISPATCH *****/ static GLboolean WarnFlag = GL_FALSE; static _glapi_warning_func warning_func; -#if defined(PTHREADS) || defined(GLX_USE_TLS) -static void init_glapi_relocs(void); -#endif - -static _glapi_proc generate_entrypoint(GLuint functionOffset); -static void fill_in_entrypoint_offset(_glapi_proc entrypoint, GLuint offset); - /* * Enable/disable printing of warning messages. */ @@ -205,21 +199,6 @@ PUBLIC void *_glapi_Context = NULL; /*@}*/ -/** - * strdup() is actually not a standard ANSI C or POSIX routine. - * Irix will not define it if ANSI mode is in effect. - */ -static char * -str_dup(const char *str) -{ - char *copy; - copy = (char*) malloc(strlen(str) + 1); - if (!copy) - return NULL; - strcpy(copy, str); - return copy; -} - /** @@ -297,6 +276,32 @@ _glapi_get_context(void) +#if defined(PTHREADS) || defined(GLX_USE_TLS) +/** + * Perform platform-specific GL API entry-point fixups. + */ +static void +init_glapi_relocs( void ) +{ +#if defined(USE_X86_ASM) && defined(GLX_USE_TLS) && !defined(GLX_X86_READONLY_TEXT) + extern unsigned long _x86_get_dispatch(void); + char run_time_patch[] = { + 0x65, 0xa1, 0, 0, 0, 0 /* movl %gs:0,%eax */ + }; + GLuint *offset = (GLuint *) &run_time_patch[2]; /* 32-bits for x86/32 */ + const GLubyte * const get_disp = (const GLubyte *) run_time_patch; + GLubyte * curr_func = (GLubyte *) gl_dispatch_functions_start; + + *offset = _x86_get_dispatch(); + while ( curr_func != (GLubyte *) gl_dispatch_functions_end ) { + (void) memcpy( curr_func, get_disp, sizeof(run_time_patch)); + curr_func += DISPATCH_FUNCTION_SIZE; + } +#endif +} +#endif /* defined(PTHREADS) || defined(GLX_USE_TLS) */ + + /** * Set the global or per-thread dispatch table pointer. * If the dispatch parameter is NULL we'll plug in the no-op dispatch @@ -353,132 +358,6 @@ _glapi_get_dispatch(void) -/*** - *** The rest of this file is pretty much concerned with GetProcAddress - *** functionality. - ***/ - -#if defined(USE_X64_64_ASM) && defined(GLX_USE_TLS) -# define DISPATCH_FUNCTION_SIZE 16 -#elif defined(USE_X86_ASM) -# if defined(THREADS) && !defined(GLX_USE_TLS) -# define DISPATCH_FUNCTION_SIZE 32 -# else -# define DISPATCH_FUNCTION_SIZE 16 -# endif -#endif - -#if !defined(DISPATCH_FUNCTION_SIZE) && !defined(XFree86Server) && !defined(XGLServer) -# define NEED_FUNCTION_POINTER -#endif - -/* The code in this file is auto-generated with Python */ -#include "glprocs.h" - - -/** - * Search the table of static entrypoint functions for the named function - * and return the corresponding glprocs_table_t entry. - */ -static const glprocs_table_t * -find_entry( const char * n ) -{ - GLuint i; - for (i = 0; static_functions[i].Name_offset >= 0; i++) { - const char *testName = gl_string_table + static_functions[i].Name_offset; - if (strcmp(testName, n) == 0) { - return &static_functions[i]; - } - } - return NULL; -} - - -/** - * Return dispatch table offset of the named static (built-in) function. - * Return -1 if function not found. - */ -static GLint -get_static_proc_offset(const char *funcName) -{ - const glprocs_table_t * const f = find_entry( funcName ); - if (f) { - return f->Offset; - } - return -1; -} - - -#if !defined(XFree86Server) && !defined(XGLServer) -#ifdef USE_X86_ASM - -#if defined( GLX_USE_TLS ) -extern GLubyte gl_dispatch_functions_start[]; -extern GLubyte gl_dispatch_functions_end[]; -#else -extern const GLubyte gl_dispatch_functions_start[]; -#endif - -#endif /* USE_X86_ASM */ - - -/** - * Return dispatch function address for the named static (built-in) function. - * Return NULL if function not found. - */ -static _glapi_proc -get_static_proc_address(const char *funcName) -{ - const glprocs_table_t * const f = find_entry( funcName ); - if (f) { -#if defined(DISPATCH_FUNCTION_SIZE) && defined(GLX_INDIRECT_RENDERING) - return (f->Address == NULL) - ? (_glapi_proc) (gl_dispatch_functions_start - + (DISPATCH_FUNCTION_SIZE * f->Offset)) - : f->Address; -#elif defined(DISPATCH_FUNCTION_SIZE) - return (_glapi_proc) (gl_dispatch_functions_start - + (DISPATCH_FUNCTION_SIZE * f->Offset)); -#else - return f->Address; -#endif - } - else { - return NULL; - } -} - -#endif /* !defined(XFree86Server) && !defined(XGLServer) */ - - - -/** - * Return the name of the function at the given offset in the dispatch - * table. For debugging only. - */ -static const char * -get_static_proc_name( GLuint offset ) -{ - GLuint i; - for (i = 0; static_functions[i].Name_offset >= 0; i++) { - if (static_functions[i].Offset == offset) { - return gl_string_table + static_functions[i].Name_offset; - } - } - return NULL; -} - - - -/********************************************************************** - * Extension function management. - */ - -/* - * Number of extension functions which we can dynamically add at runtime. - */ -#define MAX_EXTENSION_FUNCS 300 - /* * The dispatch table size (number of entries) is the size of the @@ -489,450 +368,6 @@ get_static_proc_name( GLuint offset ) #define DISPATCH_TABLE_SIZE (sizeof(struct _glapi_table) / sizeof(void *) + MAX_EXTENSION_FUNCS) -/** - * Track information about a function added to the GL API. - */ -struct _glapi_function { - /** - * Name of the function. - */ - const char * name; - - - /** - * Text string that describes the types of the parameters passed to the - * named function. Parameter types are converted to characters using the - * following rules: - * - 'i' for \c GLint, \c GLuint, and \c GLenum - * - 'p' for any pointer type - * - 'f' for \c GLfloat and \c GLclampf - * - 'd' for \c GLdouble and \c GLclampd - */ - const char * parameter_signature; - - - /** - * Offset in the dispatch table where the pointer to the real function is - * located. If the driver has not requested that the named function be - * added to the dispatch table, this will have the value ~0. - */ - unsigned dispatch_offset; - - - /** - * Pointer to the dispatch stub for the named function. - * - * \todo - * The semantic of this field should be changed slightly. Currently, it - * is always expected to be non-\c NULL. However, it would be better to - * only allocate the entry-point stub when the application requests the - * function via \c glXGetProcAddress. This would save memory for all the - * functions that the driver exports but that the application never wants - * to call. - */ - _glapi_proc dispatch_stub; -}; - - -static struct _glapi_function ExtEntryTable[MAX_EXTENSION_FUNCS]; -static GLuint NumExtEntryPoints = 0; - -#ifdef USE_SPARC_ASM -extern void __glapi_sparc_icache_flush(unsigned int *); -#endif - -/** - * Generate a dispatch function (entrypoint) which jumps through - * the given slot number (offset) in the current dispatch table. - * We need assembly language in order to accomplish this. - */ -static _glapi_proc -generate_entrypoint(GLuint functionOffset) -{ -#if defined(USE_X86_ASM) - /* 32 is chosen as something of a magic offset. For x86, the dispatch - * at offset 32 is the first one where the offset in the - * "jmp OFFSET*4(%eax)" can't be encoded in a single byte. - */ - const GLubyte * const template_func = gl_dispatch_functions_start - + (DISPATCH_FUNCTION_SIZE * 32); - GLubyte * const code = (GLubyte *) malloc(DISPATCH_FUNCTION_SIZE); - - - if ( code != NULL ) { - (void) memcpy(code, template_func, DISPATCH_FUNCTION_SIZE); - fill_in_entrypoint_offset( (_glapi_proc) code, functionOffset ); - } - - return (_glapi_proc) code; -#elif defined(USE_SPARC_ASM) - -#ifdef __arch64__ - static const unsigned int insn_template[] = { - 0x05000000, /* sethi %uhi(_glapi_Dispatch), %g2 */ - 0x03000000, /* sethi %hi(_glapi_Dispatch), %g1 */ - 0x8410a000, /* or %g2, %ulo(_glapi_Dispatch), %g2 */ - 0x82106000, /* or %g1, %lo(_glapi_Dispatch), %g1 */ - 0x8528b020, /* sllx %g2, 32, %g2 */ - 0xc2584002, /* ldx [%g1 + %g2], %g1 */ - 0x05000000, /* sethi %hi(8 * glapioffset), %g2 */ - 0x8410a000, /* or %g2, %lo(8 * glapioffset), %g2 */ - 0xc6584002, /* ldx [%g1 + %g2], %g3 */ - 0x81c0c000, /* jmpl %g3, %g0 */ - 0x01000000 /* nop */ - }; -#else - static const unsigned int insn_template[] = { - 0x03000000, /* sethi %hi(_glapi_Dispatch), %g1 */ - 0xc2006000, /* ld [%g1 + %lo(_glapi_Dispatch)], %g1 */ - 0xc6006000, /* ld [%g1 + %lo(4*glapioffset)], %g3 */ - 0x81c0c000, /* jmpl %g3, %g0 */ - 0x01000000 /* nop */ - }; -#endif /* __arch64__ */ - unsigned int *code = (unsigned int *) malloc(sizeof(insn_template)); - unsigned long glapi_addr = (unsigned long) &_glapi_Dispatch; - if (code) { - memcpy(code, insn_template, sizeof(insn_template)); - -#ifdef __arch64__ - code[0] |= (glapi_addr >> (32 + 10)); - code[1] |= ((glapi_addr & 0xffffffff) >> 10); - __glapi_sparc_icache_flush(&code[0]); - code[2] |= ((glapi_addr >> 32) & ((1 << 10) - 1)); - code[3] |= (glapi_addr & ((1 << 10) - 1)); - __glapi_sparc_icache_flush(&code[2]); - code[6] |= ((functionOffset * 8) >> 10); - code[7] |= ((functionOffset * 8) & ((1 << 10) - 1)); - __glapi_sparc_icache_flush(&code[6]); -#else - code[0] |= (glapi_addr >> 10); - code[1] |= (glapi_addr & ((1 << 10) - 1)); - __glapi_sparc_icache_flush(&code[0]); - code[2] |= (functionOffset * 4); - __glapi_sparc_icache_flush(&code[2]); -#endif /* __arch64__ */ - } - return (_glapi_proc) code; -#else - (void) functionOffset; - return NULL; -#endif /* USE_*_ASM */ -} - - -/** - * This function inserts a new dispatch offset into the assembly language - * stub that was generated with the preceeding function. - */ -static void -fill_in_entrypoint_offset(_glapi_proc entrypoint, GLuint offset) -{ -#if defined(USE_X86_ASM) - GLubyte * const code = (GLubyte *) entrypoint; - -#if DISPATCH_FUNCTION_SIZE == 32 - *((unsigned int *)(code + 11)) = 4 * offset; - *((unsigned int *)(code + 22)) = 4 * offset; -#elif DISPATCH_FUNCTION_SIZE == 16 && defined( GLX_USE_TLS ) - *((unsigned int *)(code + 8)) = 4 * offset; -#elif DISPATCH_FUNCTION_SIZE == 16 - *((unsigned int *)(code + 7)) = 4 * offset; -#else -# error Invalid DISPATCH_FUNCTION_SIZE! -#endif - -#elif defined(USE_SPARC_ASM) - - /* XXX this hasn't been tested! */ - unsigned int *code = (unsigned int *) entrypoint; -#ifdef __arch64__ - code[6] = 0x05000000; /* sethi %hi(8 * glapioffset), %g2 */ - code[7] = 0x8410a000; /* or %g2, %lo(8 * glapioffset), %g2 */ - code[6] |= ((offset * 8) >> 10); - code[7] |= ((offset * 8) & ((1 << 10) - 1)); - __glapi_sparc_icache_flush(&code[6]); -#else /* __arch64__ */ - code[2] = 0xc6006000; /* ld [%g1 + %lo(4*glapioffset)], %g3 */ - code[2] |= (offset * 4); - __glapi_sparc_icache_flush(&code[2]); -#endif /* __arch64__ */ - -#else - - /* an unimplemented architecture */ - (void) entrypoint; - (void) offset; - -#endif /* USE_*_ASM */ -} - - -/** - * Generate new entrypoint - * - * Use a temporary dispatch offset of ~0 (i.e. -1). Later, when the driver - * calls \c _glapi_add_dispatch we'll put in the proper offset. If that - * never happens, and the user calls this function, he'll segfault. That's - * what you get when you try calling a GL function that doesn't really exist. - * - * \param funcName Name of the function to create an entry-point for. - * - * \sa _glapi_add_entrypoint - */ - -static struct _glapi_function * -add_function_name( const char * funcName ) -{ - struct _glapi_function * entry = NULL; - - if (NumExtEntryPoints < MAX_EXTENSION_FUNCS) { - _glapi_proc entrypoint = generate_entrypoint(~0); - if (entrypoint != NULL) { - entry = & ExtEntryTable[NumExtEntryPoints]; - - ExtEntryTable[NumExtEntryPoints].name = str_dup(funcName); - ExtEntryTable[NumExtEntryPoints].parameter_signature = NULL; - ExtEntryTable[NumExtEntryPoints].dispatch_offset = ~0; - ExtEntryTable[NumExtEntryPoints].dispatch_stub = entrypoint; - NumExtEntryPoints++; - } - } - - return entry; -} - - -/** - * Fill-in the dispatch stub for the named function. - * - * This function is intended to be called by a hardware driver. When called, - * a dispatch stub may be created created for the function. A pointer to this - * dispatch function will be returned by glXGetProcAddress. - * - * \param function_names Array of pointers to function names that should - * share a common dispatch offset. - * \param parameter_signature String representing the types of the parameters - * passed to the named function. Parameter types - * are converted to characters using the following - * rules: - * - 'i' for \c GLint, \c GLuint, and \c GLenum - * - 'p' for any pointer type - * - 'f' for \c GLfloat and \c GLclampf - * - 'd' for \c GLdouble and \c GLclampd - * - * \returns - * The offset in the dispatch table of the named function. A pointer to the - * driver's implementation of the named function should be stored at - * \c dispatch_table[\c offset]. - * - * \sa glXGetProcAddress - * - * \warning - * This function can only handle up to 8 names at a time. As far as I know, - * the maximum number of names ever associated with an existing GL function is - * 4 (\c glPointParameterfSGIS, \c glPointParameterfEXT, - * \c glPointParameterfARB, and \c glPointParameterf), so this should not be - * too painful of a limitation. - * - * \todo - * Determine whether or not \c parameter_signature should be allowed to be - * \c NULL. It doesn't seem like much of a hardship for drivers to have to - * pass in an empty string. - * - * \todo - * Determine if code should be added to reject function names that start with - * 'glX'. - * - * \bug - * Add code to compare \c parameter_signature with the parameter signature of - * a static function. In order to do that, we need to find a way to \b get - * the parameter signature of a static function. - */ - -PUBLIC int -_glapi_add_dispatch( const char * const * function_names, - const char * parameter_signature ) -{ - static int next_dynamic_offset = _gloffset_FIRST_DYNAMIC; - const char * const real_sig = (parameter_signature != NULL) - ? parameter_signature : ""; - struct _glapi_function * entry[8]; - GLboolean is_static[8]; - unsigned i; - unsigned j; - int offset = ~0; - int new_offset; - - - (void) memset( is_static, 0, sizeof( is_static ) ); - (void) memset( entry, 0, sizeof( entry ) ); - - for ( i = 0 ; function_names[i] != NULL ; i++ ) { - /* Do some trivial validation on the name of the function. - */ - - if (!function_names[i] || function_names[i][0] != 'g' || function_names[i][1] != 'l') - return GL_FALSE; - - /* Determine if the named function already exists. If the function does - * exist, it must have the same parameter signature as the function - * being added. - */ - - new_offset = get_static_proc_offset(function_names[i]); - if (new_offset >= 0) { - /* FIXME: Make sure the parameter signatures match! How do we get - * FIXME: the parameter signature for static functions? - */ - - if ( (offset != ~0) && (new_offset != offset) ) { - return -1; - } - - is_static[i] = GL_TRUE; - offset = new_offset; - } - - - for ( j = 0 ; j < NumExtEntryPoints ; j++ ) { - if (strcmp(ExtEntryTable[j].name, function_names[i]) == 0) { - /* The offset may be ~0 if the function name was added by - * glXGetProcAddress but never filled in by the driver. - */ - - if (ExtEntryTable[j].dispatch_offset != ~0) { - if (strcmp(real_sig, ExtEntryTable[j].parameter_signature) - != 0) { - return -1; - } - - if ( (offset != ~0) && (ExtEntryTable[j].dispatch_offset != offset) ) { - return -1; - } - - offset = ExtEntryTable[j].dispatch_offset; - } - - entry[i] = & ExtEntryTable[j]; - break; - } - } - } - - if (offset == ~0) { - offset = next_dynamic_offset; - next_dynamic_offset++; - } - - for ( i = 0 ; function_names[i] != NULL ; i++ ) { - if (! is_static[i] ) { - if (entry[i] == NULL) { - entry[i] = add_function_name( function_names[i] ); - if (entry[i] == NULL) { - /* FIXME: Possible memory leak here. - */ - return -1; - } - } - - entry[i]->parameter_signature = str_dup(real_sig); - fill_in_entrypoint_offset(entry[i]->dispatch_stub, offset); - entry[i]->dispatch_offset = offset; - } - } - - return offset; -} - - -/** - * Return offset of entrypoint for named function within dispatch table. - */ -PUBLIC GLint -_glapi_get_proc_offset(const char *funcName) -{ - /* search extension functions first */ - GLuint i; - for (i = 0; i < NumExtEntryPoints; i++) { - if (strcmp(ExtEntryTable[i].name, funcName) == 0) { - return ExtEntryTable[i].dispatch_offset; - } - } - /* search static functions */ - return get_static_proc_offset(funcName); -} - - - -/** - * Return pointer to the named function. If the function name isn't found - * in the name of static functions, try generating a new API entrypoint on - * the fly with assembly language. - */ -_glapi_proc -_glapi_get_proc_address(const char *funcName) -{ - struct _glapi_function * entry; - GLuint i; - -#ifdef MANGLE - if (funcName[0] != 'm' || funcName[1] != 'g' || funcName[2] != 'l') - return NULL; -#else - if (funcName[0] != 'g' || funcName[1] != 'l') - return NULL; -#endif - - /* search extension functions first */ - for (i = 0; i < NumExtEntryPoints; i++) { - if (strcmp(ExtEntryTable[i].name, funcName) == 0) { - return ExtEntryTable[i].dispatch_stub; - } - } - -#if !defined( XFree86Server ) && !defined( XGLServer ) - /* search static functions */ - { - const _glapi_proc func = get_static_proc_address(funcName); - if (func) - return func; - } -#endif /* !defined( XFree86Server ) */ - - entry = add_function_name(funcName); - return (entry == NULL) ? NULL : entry->dispatch_stub; -} - - - -/** - * Return the name of the function at the given dispatch offset. - * This is only intended for debugging. - */ -const char * -_glapi_get_proc_name(GLuint offset) -{ - GLuint i; - const char * n; - - /* search built-in functions */ - n = get_static_proc_name(offset); - if ( n != NULL ) { - return n; - } - - /* search added extension functions */ - for (i = 0; i < NumExtEntryPoints; i++) { - if (ExtEntryTable[i].dispatch_offset == offset) { - return ExtEntryTable[i].name; - } - } - return NULL; -} - - - /** * Return size of dispatch table struct as number of functions (or * slots). @@ -944,7 +379,6 @@ _glapi_get_dispatch_table_size(void) } - /** * Make sure there are no NULL pointers in the given dispatch table. * Intended for debugging purposes. @@ -952,7 +386,7 @@ _glapi_get_dispatch_table_size(void) void _glapi_check_table(const struct _glapi_table *table) { -#ifdef DEBUG +#ifdef EXTRA_DEBUG const GLuint entries = _glapi_get_dispatch_table_size(); const void **tab = (const void **) table; GLuint i; @@ -1023,29 +457,3 @@ _glapi_check_table(const struct _glapi_table *table) (void) table; #endif } - - -#if defined(PTHREADS) || defined(GLX_USE_TLS) -/** - * Perform platform-specific GL API entry-point fixups. - */ -static void -init_glapi_relocs( void ) -{ -#if defined(USE_X86_ASM) && defined(GLX_USE_TLS) && !defined(GLX_X86_READONLY_TEXT) - extern unsigned long _x86_get_dispatch(void); - char run_time_patch[] = { - 0x65, 0xa1, 0, 0, 0, 0 /* movl %gs:0,%eax */ - }; - GLuint *offset = (GLuint *) &run_time_patch[2]; /* 32-bits for x86/32 */ - const GLubyte * const get_disp = (const GLubyte *) run_time_patch; - GLubyte * curr_func = (GLubyte *) gl_dispatch_functions_start; - - *offset = _x86_get_dispatch(); - while ( curr_func != (GLubyte *) gl_dispatch_functions_end ) { - (void) memcpy( curr_func, get_disp, sizeof(run_time_patch)); - curr_func += DISPATCH_FUNCTION_SIZE; - } -#endif -} -#endif /* defined(PTHREADS) || defined(GLX_USE_TLS) */ diff --git a/src/mesa/glapi/glapi.h b/src/mesa/glapi/glapi.h index 20d35769cf..8f2cf66218 100644 --- a/src/mesa/glapi/glapi.h +++ b/src/mesa/glapi/glapi.h @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -44,8 +44,10 @@ #ifndef _GLAPI_H #define _GLAPI_H +#define GL_GLEXT_PROTOTYPES #include "GL/gl.h" +#include "GL/glext.h" #include "glthread.h" @@ -66,6 +68,12 @@ typedef void (*_glapi_warning_func)(void *ctx, const char *str, ...); #endif +/* + * Number of extension functions which we can dynamically add at runtime. + */ +#define MAX_EXTENSION_FUNCS 300 + + /** ** Define the GET_CURRENT_CONTEXT() macro. ** \param C local variable which will hold the current context. diff --git a/src/mesa/glapi/glapi_getproc.c b/src/mesa/glapi/glapi_getproc.c new file mode 100644 index 0000000000..b3ecc9d25a --- /dev/null +++ b/src/mesa/glapi/glapi_getproc.c @@ -0,0 +1,615 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +/** + * \file glapi_getproc. + * + * Code for implementing glXGetProcAddress(), etc. + * This was originally in glapi.c but refactored out. + */ + + +#include +#include +#include "glheader.h" +#include "glapi.h" +#include "glapioffsets.h" +#include "glapitable.h" + + +/** + * strdup() is actually not a standard ANSI C or POSIX routine. + * Irix will not define it if ANSI mode is in effect. + */ +static char * +str_dup(const char *str) +{ + char *copy; + copy = (char*) malloc(strlen(str) + 1); + if (!copy) + return NULL; + strcpy(copy, str); + return copy; +} + + + +#if defined(USE_X64_64_ASM) && defined(GLX_USE_TLS) +# define DISPATCH_FUNCTION_SIZE 16 +#elif defined(USE_X86_ASM) +# if defined(THREADS) && !defined(GLX_USE_TLS) +# define DISPATCH_FUNCTION_SIZE 32 +# else +# define DISPATCH_FUNCTION_SIZE 16 +# endif +#endif + +#if !defined(DISPATCH_FUNCTION_SIZE) && !defined(XFree86Server) && !defined(XGLServer) +# define NEED_FUNCTION_POINTER +#endif + +/* The code in this file is auto-generated with Python */ +#include "glprocs.h" + + +/** + * Search the table of static entrypoint functions for the named function + * and return the corresponding glprocs_table_t entry. + */ +static const glprocs_table_t * +find_entry( const char * n ) +{ + GLuint i; + for (i = 0; static_functions[i].Name_offset >= 0; i++) { + const char *testName = gl_string_table + static_functions[i].Name_offset; + if (strcmp(testName, n) == 0) { + return &static_functions[i]; + } + } + return NULL; +} + + +/** + * Return dispatch table offset of the named static (built-in) function. + * Return -1 if function not found. + */ +static GLint +get_static_proc_offset(const char *funcName) +{ + const glprocs_table_t * const f = find_entry( funcName ); + if (f) { + return f->Offset; + } + return -1; +} + + +#if !defined(XFree86Server) && !defined(XGLServer) +#ifdef USE_X86_ASM + +#if defined( GLX_USE_TLS ) +extern GLubyte gl_dispatch_functions_start[]; +extern GLubyte gl_dispatch_functions_end[]; +#else +extern const GLubyte gl_dispatch_functions_start[]; +#endif + +#endif /* USE_X86_ASM */ + + +/** + * Return dispatch function address for the named static (built-in) function. + * Return NULL if function not found. + */ +static _glapi_proc +get_static_proc_address(const char *funcName) +{ + const glprocs_table_t * const f = find_entry( funcName ); + if (f) { +#if defined(DISPATCH_FUNCTION_SIZE) && defined(GLX_INDIRECT_RENDERING) + return (f->Address == NULL) + ? (_glapi_proc) (gl_dispatch_functions_start + + (DISPATCH_FUNCTION_SIZE * f->Offset)) + : f->Address; +#elif defined(DISPATCH_FUNCTION_SIZE) + return (_glapi_proc) (gl_dispatch_functions_start + + (DISPATCH_FUNCTION_SIZE * f->Offset)); +#else + return f->Address; +#endif + } + else { + return NULL; + } +} + +#endif /* !defined(XFree86Server) && !defined(XGLServer) */ + + + +/** + * Return the name of the function at the given offset in the dispatch + * table. For debugging only. + */ +static const char * +get_static_proc_name( GLuint offset ) +{ + GLuint i; + for (i = 0; static_functions[i].Name_offset >= 0; i++) { + if (static_functions[i].Offset == offset) { + return gl_string_table + static_functions[i].Name_offset; + } + } + return NULL; +} + + + +/********************************************************************** + * Extension function management. + */ + + +/** + * Track information about a function added to the GL API. + */ +struct _glapi_function { + /** + * Name of the function. + */ + const char * name; + + + /** + * Text string that describes the types of the parameters passed to the + * named function. Parameter types are converted to characters using the + * following rules: + * - 'i' for \c GLint, \c GLuint, and \c GLenum + * - 'p' for any pointer type + * - 'f' for \c GLfloat and \c GLclampf + * - 'd' for \c GLdouble and \c GLclampd + */ + const char * parameter_signature; + + + /** + * Offset in the dispatch table where the pointer to the real function is + * located. If the driver has not requested that the named function be + * added to the dispatch table, this will have the value ~0. + */ + unsigned dispatch_offset; + + + /** + * Pointer to the dispatch stub for the named function. + * + * \todo + * The semantic of this field should be changed slightly. Currently, it + * is always expected to be non-\c NULL. However, it would be better to + * only allocate the entry-point stub when the application requests the + * function via \c glXGetProcAddress. This would save memory for all the + * functions that the driver exports but that the application never wants + * to call. + */ + _glapi_proc dispatch_stub; +}; + + +static struct _glapi_function ExtEntryTable[MAX_EXTENSION_FUNCS]; +static GLuint NumExtEntryPoints = 0; + +#ifdef USE_SPARC_ASM +extern void __glapi_sparc_icache_flush(unsigned int *); +#endif + +/** + * Generate a dispatch function (entrypoint) which jumps through + * the given slot number (offset) in the current dispatch table. + * We need assembly language in order to accomplish this. + */ +static _glapi_proc +generate_entrypoint(GLuint functionOffset) +{ +#if defined(USE_X86_ASM) + /* 32 is chosen as something of a magic offset. For x86, the dispatch + * at offset 32 is the first one where the offset in the + * "jmp OFFSET*4(%eax)" can't be encoded in a single byte. + */ + const GLubyte * const template_func = gl_dispatch_functions_start + + (DISPATCH_FUNCTION_SIZE * 32); + GLubyte * const code = (GLubyte *) malloc(DISPATCH_FUNCTION_SIZE); + + + if ( code != NULL ) { + (void) memcpy(code, template_func, DISPATCH_FUNCTION_SIZE); + fill_in_entrypoint_offset( (_glapi_proc) code, functionOffset ); + } + + return (_glapi_proc) code; +#elif defined(USE_SPARC_ASM) + +#ifdef __arch64__ + static const unsigned int insn_template[] = { + 0x05000000, /* sethi %uhi(_glapi_Dispatch), %g2 */ + 0x03000000, /* sethi %hi(_glapi_Dispatch), %g1 */ + 0x8410a000, /* or %g2, %ulo(_glapi_Dispatch), %g2 */ + 0x82106000, /* or %g1, %lo(_glapi_Dispatch), %g1 */ + 0x8528b020, /* sllx %g2, 32, %g2 */ + 0xc2584002, /* ldx [%g1 + %g2], %g1 */ + 0x05000000, /* sethi %hi(8 * glapioffset), %g2 */ + 0x8410a000, /* or %g2, %lo(8 * glapioffset), %g2 */ + 0xc6584002, /* ldx [%g1 + %g2], %g3 */ + 0x81c0c000, /* jmpl %g3, %g0 */ + 0x01000000 /* nop */ + }; +#else + static const unsigned int insn_template[] = { + 0x03000000, /* sethi %hi(_glapi_Dispatch), %g1 */ + 0xc2006000, /* ld [%g1 + %lo(_glapi_Dispatch)], %g1 */ + 0xc6006000, /* ld [%g1 + %lo(4*glapioffset)], %g3 */ + 0x81c0c000, /* jmpl %g3, %g0 */ + 0x01000000 /* nop */ + }; +#endif /* __arch64__ */ + unsigned int *code = (unsigned int *) malloc(sizeof(insn_template)); + unsigned long glapi_addr = (unsigned long) &_glapi_Dispatch; + if (code) { + memcpy(code, insn_template, sizeof(insn_template)); + +#ifdef __arch64__ + code[0] |= (glapi_addr >> (32 + 10)); + code[1] |= ((glapi_addr & 0xffffffff) >> 10); + __glapi_sparc_icache_flush(&code[0]); + code[2] |= ((glapi_addr >> 32) & ((1 << 10) - 1)); + code[3] |= (glapi_addr & ((1 << 10) - 1)); + __glapi_sparc_icache_flush(&code[2]); + code[6] |= ((functionOffset * 8) >> 10); + code[7] |= ((functionOffset * 8) & ((1 << 10) - 1)); + __glapi_sparc_icache_flush(&code[6]); +#else + code[0] |= (glapi_addr >> 10); + code[1] |= (glapi_addr & ((1 << 10) - 1)); + __glapi_sparc_icache_flush(&code[0]); + code[2] |= (functionOffset * 4); + __glapi_sparc_icache_flush(&code[2]); +#endif /* __arch64__ */ + } + return (_glapi_proc) code; +#else + (void) functionOffset; + return NULL; +#endif /* USE_*_ASM */ +} + + +/** + * This function inserts a new dispatch offset into the assembly language + * stub that was generated with the preceeding function. + */ +static void +fill_in_entrypoint_offset(_glapi_proc entrypoint, GLuint offset) +{ +#if defined(USE_X86_ASM) + GLubyte * const code = (GLubyte *) entrypoint; + +#if DISPATCH_FUNCTION_SIZE == 32 + *((unsigned int *)(code + 11)) = 4 * offset; + *((unsigned int *)(code + 22)) = 4 * offset; +#elif DISPATCH_FUNCTION_SIZE == 16 && defined( GLX_USE_TLS ) + *((unsigned int *)(code + 8)) = 4 * offset; +#elif DISPATCH_FUNCTION_SIZE == 16 + *((unsigned int *)(code + 7)) = 4 * offset; +#else +# error Invalid DISPATCH_FUNCTION_SIZE! +#endif + +#elif defined(USE_SPARC_ASM) + + /* XXX this hasn't been tested! */ + unsigned int *code = (unsigned int *) entrypoint; +#ifdef __arch64__ + code[6] = 0x05000000; /* sethi %hi(8 * glapioffset), %g2 */ + code[7] = 0x8410a000; /* or %g2, %lo(8 * glapioffset), %g2 */ + code[6] |= ((offset * 8) >> 10); + code[7] |= ((offset * 8) & ((1 << 10) - 1)); + __glapi_sparc_icache_flush(&code[6]); +#else /* __arch64__ */ + code[2] = 0xc6006000; /* ld [%g1 + %lo(4*glapioffset)], %g3 */ + code[2] |= (offset * 4); + __glapi_sparc_icache_flush(&code[2]); +#endif /* __arch64__ */ + +#else + + /* an unimplemented architecture */ + (void) entrypoint; + (void) offset; + +#endif /* USE_*_ASM */ +} + + +/** + * Generate new entrypoint + * + * Use a temporary dispatch offset of ~0 (i.e. -1). Later, when the driver + * calls \c _glapi_add_dispatch we'll put in the proper offset. If that + * never happens, and the user calls this function, he'll segfault. That's + * what you get when you try calling a GL function that doesn't really exist. + * + * \param funcName Name of the function to create an entry-point for. + * + * \sa _glapi_add_entrypoint + */ + +static struct _glapi_function * +add_function_name( const char * funcName ) +{ + struct _glapi_function * entry = NULL; + + if (NumExtEntryPoints < MAX_EXTENSION_FUNCS) { + _glapi_proc entrypoint = generate_entrypoint(~0); + if (entrypoint != NULL) { + entry = & ExtEntryTable[NumExtEntryPoints]; + + ExtEntryTable[NumExtEntryPoints].name = str_dup(funcName); + ExtEntryTable[NumExtEntryPoints].parameter_signature = NULL; + ExtEntryTable[NumExtEntryPoints].dispatch_offset = ~0; + ExtEntryTable[NumExtEntryPoints].dispatch_stub = entrypoint; + NumExtEntryPoints++; + } + } + + return entry; +} + + +/** + * Fill-in the dispatch stub for the named function. + * + * This function is intended to be called by a hardware driver. When called, + * a dispatch stub may be created created for the function. A pointer to this + * dispatch function will be returned by glXGetProcAddress. + * + * \param function_names Array of pointers to function names that should + * share a common dispatch offset. + * \param parameter_signature String representing the types of the parameters + * passed to the named function. Parameter types + * are converted to characters using the following + * rules: + * - 'i' for \c GLint, \c GLuint, and \c GLenum + * - 'p' for any pointer type + * - 'f' for \c GLfloat and \c GLclampf + * - 'd' for \c GLdouble and \c GLclampd + * + * \returns + * The offset in the dispatch table of the named function. A pointer to the + * driver's implementation of the named function should be stored at + * \c dispatch_table[\c offset]. + * + * \sa glXGetProcAddress + * + * \warning + * This function can only handle up to 8 names at a time. As far as I know, + * the maximum number of names ever associated with an existing GL function is + * 4 (\c glPointParameterfSGIS, \c glPointParameterfEXT, + * \c glPointParameterfARB, and \c glPointParameterf), so this should not be + * too painful of a limitation. + * + * \todo + * Determine whether or not \c parameter_signature should be allowed to be + * \c NULL. It doesn't seem like much of a hardship for drivers to have to + * pass in an empty string. + * + * \todo + * Determine if code should be added to reject function names that start with + * 'glX'. + * + * \bug + * Add code to compare \c parameter_signature with the parameter signature of + * a static function. In order to do that, we need to find a way to \b get + * the parameter signature of a static function. + */ + +PUBLIC int +_glapi_add_dispatch( const char * const * function_names, + const char * parameter_signature ) +{ + static int next_dynamic_offset = _gloffset_FIRST_DYNAMIC; + const char * const real_sig = (parameter_signature != NULL) + ? parameter_signature : ""; + struct _glapi_function * entry[8]; + GLboolean is_static[8]; + unsigned i; + unsigned j; + int offset = ~0; + int new_offset; + + + (void) memset( is_static, 0, sizeof( is_static ) ); + (void) memset( entry, 0, sizeof( entry ) ); + + for ( i = 0 ; function_names[i] != NULL ; i++ ) { + /* Do some trivial validation on the name of the function. + */ + + if (!function_names[i] || function_names[i][0] != 'g' || function_names[i][1] != 'l') + return GL_FALSE; + + /* Determine if the named function already exists. If the function does + * exist, it must have the same parameter signature as the function + * being added. + */ + + new_offset = get_static_proc_offset(function_names[i]); + if (new_offset >= 0) { + /* FIXME: Make sure the parameter signatures match! How do we get + * FIXME: the parameter signature for static functions? + */ + + if ( (offset != ~0) && (new_offset != offset) ) { + return -1; + } + + is_static[i] = GL_TRUE; + offset = new_offset; + } + + + for ( j = 0 ; j < NumExtEntryPoints ; j++ ) { + if (strcmp(ExtEntryTable[j].name, function_names[i]) == 0) { + /* The offset may be ~0 if the function name was added by + * glXGetProcAddress but never filled in by the driver. + */ + + if (ExtEntryTable[j].dispatch_offset != ~0) { + if (strcmp(real_sig, ExtEntryTable[j].parameter_signature) + != 0) { + return -1; + } + + if ( (offset != ~0) && (ExtEntryTable[j].dispatch_offset != offset) ) { + return -1; + } + + offset = ExtEntryTable[j].dispatch_offset; + } + + entry[i] = & ExtEntryTable[j]; + break; + } + } + } + + if (offset == ~0) { + offset = next_dynamic_offset; + next_dynamic_offset++; + } + + for ( i = 0 ; function_names[i] != NULL ; i++ ) { + if (! is_static[i] ) { + if (entry[i] == NULL) { + entry[i] = add_function_name( function_names[i] ); + if (entry[i] == NULL) { + /* FIXME: Possible memory leak here. + */ + return -1; + } + } + + entry[i]->parameter_signature = str_dup(real_sig); + fill_in_entrypoint_offset(entry[i]->dispatch_stub, offset); + entry[i]->dispatch_offset = offset; + } + } + + return offset; +} + + +/** + * Return offset of entrypoint for named function within dispatch table. + */ +PUBLIC GLint +_glapi_get_proc_offset(const char *funcName) +{ + /* search extension functions first */ + GLuint i; + for (i = 0; i < NumExtEntryPoints; i++) { + if (strcmp(ExtEntryTable[i].name, funcName) == 0) { + return ExtEntryTable[i].dispatch_offset; + } + } + /* search static functions */ + return get_static_proc_offset(funcName); +} + + + +/** + * Return pointer to the named function. If the function name isn't found + * in the name of static functions, try generating a new API entrypoint on + * the fly with assembly language. + */ +_glapi_proc +_glapi_get_proc_address(const char *funcName) +{ + struct _glapi_function * entry; + GLuint i; + +#ifdef MANGLE + if (funcName[0] != 'm' || funcName[1] != 'g' || funcName[2] != 'l') + return NULL; +#else + if (funcName[0] != 'g' || funcName[1] != 'l') + return NULL; +#endif + + /* search extension functions first */ + for (i = 0; i < NumExtEntryPoints; i++) { + if (strcmp(ExtEntryTable[i].name, funcName) == 0) { + return ExtEntryTable[i].dispatch_stub; + } + } + +#if !defined( XFree86Server ) && !defined( XGLServer ) + /* search static functions */ + { + const _glapi_proc func = get_static_proc_address(funcName); + if (func) + return func; + } +#endif /* !defined( XFree86Server ) */ + + entry = add_function_name(funcName); + return (entry == NULL) ? NULL : entry->dispatch_stub; +} + + + +/** + * Return the name of the function at the given dispatch offset. + * This is only intended for debugging. + */ +const char * +_glapi_get_proc_name(GLuint offset) +{ + GLuint i; + const char * n; + + /* search built-in functions */ + n = get_static_proc_name(offset); + if ( n != NULL ) { + return n; + } + + /* search added extension functions */ + for (i = 0; i < NumExtEntryPoints; i++) { + if (ExtEntryTable[i].dispatch_offset == offset) { + return ExtEntryTable[i].name; + } + } + return NULL; +} diff --git a/src/mesa/sources b/src/mesa/sources index 8c000087c6..7f49ac6e32 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -77,6 +77,7 @@ MAIN_SOURCES = \ GLAPI_SOURCES = \ main/dispatch.c \ glapi/glapi.c \ + glapi/glapi_getproc.c \ glapi/glthread.c MATH_SOURCES = \ -- cgit v1.2.3 From 5ebd8495b31fadf7452aa336fb5aa708c1a5d753 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 14:27:36 -0600 Subject: gallium: added st_get_proc_address() --- src/mesa/state_tracker/st_context.c | 9 +++++++++ src/mesa/state_tracker/st_public.h | 7 +++++++ 2 files changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 9decca8a14..1b3b999714 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -33,6 +33,7 @@ #include "main/scissor.h" #include "vbo/vbo.h" #include "shader/shader_api.h" +#include "glapi/glapi.h" #include "st_public.h" #include "st_context.h" #include "st_cb_accum.h" @@ -243,6 +244,14 @@ void st_copy_context_state(struct st_context *dst, } + +st_proc st_get_proc_address(const char *procname) +{ + return (st_proc) _glapi_get_proc_address(procname); +} + + + void st_init_driver_functions(struct dd_function_table *functions) { _mesa_init_glsl_driver_functions(functions); diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index 9d88ce9764..a0dab33257 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -86,4 +86,11 @@ void st_finish( struct st_context *st ); void st_notify_swapbuffers(struct st_framebuffer *stfb); void st_notify_swapbuffers_complete(struct st_framebuffer *stfb); + +/** Generic function type */ +typedef void (*st_proc)(); + +st_proc st_get_proc_address(const char *procname); + + #endif -- cgit v1.2.3 From 1496bba42904cbc9b827816b7924fa930661c22a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 14:28:07 -0600 Subject: egl: call st_get_proc_address() --- src/gallium/winsys/egl_xlib/egl_xlib.c | 6 +----- 1 file changed, 1 insertion(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 7bef46d560..194bf2842a 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -210,11 +210,7 @@ xlib_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) static _EGLProc xlib_eglGetProcAddress(const char *procname) { - /* XXX for each supported API, evaluate GetProcAddress(name) */ - /* - return _glapi_get_proc_address(procname); - */ - return NULL; + return (_EGLProc) st_get_proc_address(procname); } -- cgit v1.2.3 From 03d579aa19bcb1facec5ab67b6f7123e9ec9f26e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 16:56:32 -0600 Subject: mesa: FEATURE_dispatch to control dispatch table usage --- src/mesa/main/context.c | 2 ++ src/mesa/main/mfeatures.h | 1 + 2 files changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index dcc518e9fd..439ac6aba5 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1168,7 +1168,9 @@ _mesa_initialize_context(GLcontext *ctx, if (ctx->Exec) _mesa_free(ctx->Exec); } +#if FEATURE_dispatch _mesa_init_exec_table(ctx->Exec); +#endif ctx->CurrentDispatch = ctx->Exec; #if FEATURE_dlist _mesa_init_dlist_table(ctx->Save); diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h index 4e60611c48..6a71954648 100644 --- a/src/mesa/main/mfeatures.h +++ b/src/mesa/main/mfeatures.h @@ -40,6 +40,7 @@ #define FEATURE_attrib_stack _HAVE_FULL_GL #define FEATURE_colortable _HAVE_FULL_GL #define FEATURE_convolution _HAVE_FULL_GL +#define FEATURE_dispatch _HAVE_FULL_GL #define FEATURE_dlist _HAVE_FULL_GL #define FEATURE_draw_read_buffer _HAVE_FULL_GL #define FEATURE_drawpix _HAVE_FULL_GL -- cgit v1.2.3 From 414b535134256dbe9e4d4a92068143f054e13720 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 16:57:02 -0600 Subject: gallium: more FEATURE_x tests --- src/mesa/state_tracker/st_context.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 1b3b999714..d394562feb 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -164,10 +164,14 @@ static void st_destroy_context_priv( struct st_context *st ) st_destroy_atoms( st ); st_destroy_draw( st ); st_destroy_generate_mipmap(st); - st_destroy_bitmap(st); +#if FEATURE_EXT_framebuffer_blit st_destroy_blit(st); +#endif st_destroy_clear(st); +#if FEATURE_drawpix + st_destroy_bitmap(st); st_destroy_drawpix(st); +#endif _vbo_DestroyContext(st->ctx); -- cgit v1.2.3 From 51f24ef5bae3fa1f1ae6172fd5c5e6235f3f6857 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 17 Jun 2008 16:58:22 -0600 Subject: gallium: remove duplicated u_mm.c in file list --- src/gallium/auxiliary/util/Makefile | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 05bc43131a..db3d810a2e 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -15,8 +15,7 @@ C_SOURCES = \ u_mm.c \ u_simple_shaders.c \ u_snprintf.c \ - u_time.c \ - u_mm.c + u_time.c include ../../Makefile.template -- cgit v1.2.3 From 0cf79316d076d598f143b68f61308df9ddaf5e9a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 18 Jun 2008 09:21:09 -0600 Subject: egl: omit libmesa.a glapi.c from library --- src/gallium/winsys/egl_xlib/Makefile | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/Makefile b/src/gallium/winsys/egl_xlib/Makefile index d5595b819a..76f1b56da4 100644 --- a/src/gallium/winsys/egl_xlib/Makefile +++ b/src/gallium/winsys/egl_xlib/Makefile @@ -28,9 +28,13 @@ WINSYS_OBJECTS = $(WINSYS_SOURCES:.c=.o) LIBS = \ $(GALLIUM_DRIVERS) \ + $(GALLIUM_AUXILIARIES) + +# XXX temporary (should create a separate lib with the GL API funcs and +# mesa code, as done for ES 1.x, 2.x, OpenVG, etc) +UNUSED_LIBS = \ $(TOP)/src/mesa/libglapi.a \ $(TOP)/src/mesa/libmesa.a \ - $(GALLIUM_AUXILIARIES) LOCAL_CFLAGS = -D_EGL_PLATFORM_X=1 @@ -56,7 +60,7 @@ $(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(WINSYS_OBJECTS) $(LIBS) -noprefix \ -install $(TOP)/$(LIB_DIR) \ $(MKLIB_OPTIONS) $(WINSYS_OBJECTS) \ - --start-group $(LIBS) --end-group + --whole-archive $(LIBS) --no-whole-archive depend: $(ALL_SOURCES) -- cgit v1.2.3 From a668b43568b99170e354c7e7cbb393e633765ec4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 18 Jun 2008 09:21:28 -0600 Subject: egl: use RTLD_LAZY --- src/egl/main/egldriver.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 5ecfb9eb30..4a611b9fc9 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -183,7 +183,7 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) /* XXX also prepend a directory path??? */ sprintf(driverFilename, "%s.so", driverName); _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename); - lib = dlopen(driverFilename, RTLD_NOW); + lib = dlopen(driverFilename, RTLD_LAZY); #endif if (!lib) { -- cgit v1.2.3 From 19f872f2ecb45d5e95ccd2b434a88781c9b4f451 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 18 Jun 2008 09:30:13 -0600 Subject: mesa: fix ReadBuffer initialization --- src/mesa/main/context.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 439ac6aba5..8aab1eebc8 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1011,6 +1011,8 @@ init_attrib_groups(GLcontext *ctx) _mesa_init_multisample( ctx ); #if FEATURE_pixel_transfer _mesa_init_pixel( ctx ); +#else + ctx->Pixel.ReadBuffer = ctx->Visual.doubleBufferMode ? GL_BACK : GL_FRONT; #endif _mesa_init_pixelstore( ctx ); _mesa_init_point( ctx ); -- cgit v1.2.3 From 9a2c70f0754b275e7ec370bbbff3d81b48bfd476 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 18 Jun 2008 10:20:13 -0600 Subject: gallium: added new st_set_framebuffer_surface() This allows the winsys to explicitly specify gallium surfaces for a framebuffer object. --- src/mesa/state_tracker/st_framebuffer.c | 56 +++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_public.h | 3 ++ 2 files changed, 59 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 7099d78eb8..7e6db46757 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -38,6 +38,7 @@ #include "st_cb_fbo.h" #include "pipe/p_defines.h" #include "pipe/p_context.h" +#include "pipe/p_inlines.h" struct st_framebuffer * @@ -154,6 +155,61 @@ void st_unreference_framebuffer( struct st_framebuffer **stfb ) +/** + * Set/replace a framebuffer surface. + * The user of the state tracker can use this instead of + * st_resize_framebuffer() to provide new surfaces when a window is resized. + */ +void +st_set_framebuffer_surface(struct st_framebuffer *stfb, + uint surfIndex, struct pipe_surface *surf) +{ + static const GLuint invalid_size = 9999999; + struct st_renderbuffer *strb; + GLuint width, height, i; + + assert(surfIndex < BUFFER_COUNT); + + strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer); + assert(strb); + + /* replace the renderbuffer's surface/texture pointers */ + pipe_surface_reference( &strb->surface, surf ); + pipe_texture_reference( &strb->texture, surf->texture ); + + /* update renderbuffer's width/height */ + strb->Base.Width = surf->width; + strb->Base.Height = surf->height; + + /* Try to update the framebuffer's width/height from the renderbuffer + * sizes. Before we start drawing, all the rbs _should_ be the same size. + */ + width = height = invalid_size; + for (i = 0; i < BUFFER_COUNT; i++) { + if (stfb->Base.Attachment[i].Renderbuffer) { + if (width == invalid_size) { + width = stfb->Base.Attachment[i].Renderbuffer->Width; + height = stfb->Base.Attachment[i].Renderbuffer->Height; + } + else if (width != stfb->Base.Attachment[i].Renderbuffer->Width || + height != stfb->Base.Attachment[i].Renderbuffer->Height) { + /* inconsistant renderbuffer sizes, bail out */ + return; + } + } + } + + if (width != invalid_size) { + /* OK, the renderbuffers are of a consistant size, so update the + * parent framebuffer's size. + */ + stfb->Base.Width = width; + stfb->Base.Height = height; + } +} + + + /** * Return the pipe_surface for the given renderbuffer. */ diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index a0dab33257..b99984215f 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -68,6 +68,9 @@ struct st_framebuffer *st_create_framebuffer( const __GLcontextModes *visual, void st_resize_framebuffer( struct st_framebuffer *stfb, uint width, uint height ); +void st_set_framebuffer_surface(struct st_framebuffer *stfb, + uint surfIndex, struct pipe_surface *surf); + struct pipe_surface *st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex); -- cgit v1.2.3 From b1eff0228b7199920c02e62446ebf5496c298814 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 18 Jun 2008 11:00:43 -0600 Subject: mesa: add fwd decl of fill_in_entrypoint_offset() --- src/mesa/glapi/glapi_getproc.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/glapi/glapi_getproc.c b/src/mesa/glapi/glapi_getproc.c index b3ecc9d25a..5e0af26a01 100644 --- a/src/mesa/glapi/glapi_getproc.c +++ b/src/mesa/glapi/glapi_getproc.c @@ -38,6 +38,10 @@ #include "glapitable.h" +static void +fill_in_entrypoint_offset(_glapi_proc entrypoint, GLuint offset); + + /** * strdup() is actually not a standard ANSI C or POSIX routine. * Irix will not define it if ANSI mode is in effect. -- cgit v1.2.3 From b623fa9e2d6f97f9febc978c158d790b26e175a7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 18 Jun 2008 19:43:06 +0200 Subject: mesa: Fix bug in _mesa_swizzle_ubyte_image --- src/mesa/main/texstore.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index d7bfb69443..519a73b960 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -823,7 +823,8 @@ _mesa_swizzle_ubyte_image(GLcontext *ctx, /* _mesa_printf("map %d %d %d %d\n", map[0], map[1], map[2], map[3]); */ - if (srcRowStride == dstRowStride && + if (srcComponents == dstComponents && + srcRowStride == dstRowStride && srcRowStride == srcWidth * srcComponents && dimensions < 3) { /* 1 and 2D images only */ -- cgit v1.2.3 From 7d7f3e2c9451b2233c196d82d523c50b5d2616cc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 18 Jun 2008 14:50:35 -0600 Subject: gallium: split long prims into chunks with an even number of vertices This fixes culling "parity" errors when splitting long tri strips. Splitting strips into chunks with an odd number of vertices causes front/back-face orientation to get reversed and upsets culling. --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c | 9 +++++++++ src/gallium/auxiliary/draw/draw_pt_varray.c | 4 ++++ 2 files changed, 13 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 5ce3aba2a2..fdf9b6fe6a 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -193,6 +193,15 @@ static void fse_prepare( struct draw_pt_middle_end *middle, *max_vertices = (draw->render->max_vertex_buffer_bytes / (vinfo->size * 4)); + /* Return an even number of verts. + * This prevents "parity" errors when splitting long triangle strips which + * can lead to front/back culling mix-ups. + * Every other triangle in a strip has an alternate front/back orientation + * so splitting at an odd position can cause the orientation of subsequent + * triangles to get reversed. + */ + *max_vertices = *max_vertices & ~1; + /* Probably need to do this somewhere (or fix exec shader not to * need it): */ diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 4479963db1..2cc08a9e93 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -147,6 +147,10 @@ static void varray_prepare(struct draw_pt_front_end *frontend, varray->middle = middle; middle->prepare(middle, varray->output_prim, opt, &varray->driver_fetch_max ); + + /* check that the max is even */ + assert((varray->driver_fetch_max & 1) == 0); + varray->fetch_max = MIN2(FETCH_MAX, varray->driver_fetch_max); } -- cgit v1.2.3 From f1401385587882bb9d18a5f5b01dcbb71ddf0a2f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 18 Jun 2008 15:08:19 -0600 Subject: gallium: additional fixes to ensure even number of vertices per buffer --- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 3 +++ src/gallium/auxiliary/draw/draw_pt_emit.c | 3 +++ src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 9 +++++++++ src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 3 +++ 4 files changed, 18 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index 10a1f7df79..a6fde77a0e 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -390,6 +390,9 @@ vbuf_alloc_vertices( struct vbuf_stage *vbuf ) /* Allocate a new vertex buffer */ vbuf->max_vertices = vbuf->render->max_vertex_buffer_bytes / vbuf->vertex_size; + /* even number */ + vbuf->max_vertices = vbuf->max_vertices & ~1; + /* Must always succeed -- driver gives us a * 'max_vertex_buffer_bytes' which it guarantees it can allocate, * and it will flush itself if necessary to do so. If this does diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index a02f1f46fe..40f05cb9e0 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -143,6 +143,9 @@ void draw_pt_emit_prepare( struct pt_emit *emit, *max_vertices = (draw->render->max_vertex_buffer_bytes / (vinfo->size * 4)); + + /* even number */ + *max_vertices = *max_vertices & ~1; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 083ce105bf..4a1f3b0953 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -200,6 +200,15 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle, *max_vertices = (draw->render->max_vertex_buffer_bytes / (vinfo->size * 4)); + + /* Return an even number of verts. + * This prevents "parity" errors when splitting long triangle strips which + * can lead to front/back culling mix-ups. + * Every other triangle in a strip has an alternate front/back orientation + * so splitting at an odd position can cause the orientation of subsequent + * triangles to get reversed. + */ + *max_vertices = *max_vertices & ~1; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 25118712a6..0aec4b71ba 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -99,6 +99,9 @@ static void fetch_pipeline_prepare( struct draw_pt_middle_end *middle, *max_vertices = DRAW_PIPE_MAX_VERTICES; } + /* return even number */ + *max_vertices = *max_vertices & ~1; + /* No need to prepare the shader. */ vs->prepare(vs, draw); -- cgit v1.2.3 From aa816d114ee90d0c5b861f622c0063b54f7eb612 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 12 Jun 2008 23:34:21 +0900 Subject: draw: Fix MSVC integer size conversion warning. --- src/gallium/auxiliary/draw/draw_pt_varray.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 2cc08a9e93..46e722a154 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -172,7 +172,7 @@ static void varray_destroy(struct draw_pt_front_end *frontend) struct draw_pt_front_end *draw_pt_varray(struct draw_context *draw) { - unsigned i; + ushort i; struct varray_frontend *varray = CALLOC_STRUCT(varray_frontend); if (varray == NULL) return NULL; -- cgit v1.2.3 From b440cea343b007ca97edb2cd4e1e6150c989417b Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 19 Jun 2008 14:06:28 +0200 Subject: util: Add missing format names. --- src/gallium/auxiliary/util/p_debug.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index e4de12167a..9d56f8bfab 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -366,6 +366,7 @@ static const struct debug_named_value pipe_format_names[] = { DEBUG_NAMED_VALUE(PIPE_FORMAT_A1R5G5B5_UNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_A4R4G4B4_UNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_R5G6B5_UNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_A2B10G10R10_UNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_L8_UNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_A8_UNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_I8_UNORM), @@ -436,6 +437,9 @@ static const struct debug_named_value pipe_format_names[] = { DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_B6G5R5_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_A8B8G8R8_SNORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_X8B8G8R8_SNORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_R8_SSCALED), DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8_SSCALED), DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SSCALED), @@ -446,6 +450,8 @@ static const struct debug_named_value pipe_format_names[] = { DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8_SRGB), DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8A8_SRGB), DEBUG_NAMED_VALUE(PIPE_FORMAT_R8G8B8X8_SRGB), + DEBUG_NAMED_VALUE(PIPE_FORMAT_X8UB8UG8SR8S_NORM), + DEBUG_NAMED_VALUE(PIPE_FORMAT_B6UG5SR5S_NORM), DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_RGB), DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT1_RGBA), DEBUG_NAMED_VALUE(PIPE_FORMAT_DXT3_RGBA), -- cgit v1.2.3 From c5bf215b1b955c8dafeae84a5f526052fd2e0256 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 19 Jun 2008 21:16:16 +0900 Subject: gallium: Add extra parenthesis as advised by gcc. --- src/gallium/auxiliary/util/p_debug_mem.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug_mem.c b/src/gallium/auxiliary/util/p_debug_mem.c index 51dbd0a710..ed18c6540e 100644 --- a/src/gallium/auxiliary/util/p_debug_mem.c +++ b/src/gallium/auxiliary/util/p_debug_mem.c @@ -263,8 +263,8 @@ debug_memory_end(unsigned long start_no) void *ptr; hdr = LIST_ENTRY(struct debug_memory_header, entry, head); ptr = data_from_header(hdr); - if(start_no <= hdr->no && hdr->no < last_no || - last_no < start_no && (hdr->no < last_no || start_no <= hdr->no)) { + if((start_no <= hdr->no && hdr->no < last_no) || + (last_no < start_no && (hdr->no < last_no || start_no <= hdr->no))) { debug_printf("%s:%u:%s: %u bytes at %p not freed\n", hdr->file, hdr->line, hdr->function, hdr->size, ptr); -- cgit v1.2.3 From 6cbc2734d19567afeaa6c5d93149ef87b08cd167 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 19 Jun 2008 14:42:17 +0200 Subject: i915: Refractored and clean up i915_texture.c --- src/gallium/drivers/i915simple/i915_texture.c | 269 ++++++++++++++------------ 1 file changed, 141 insertions(+), 128 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index d591b09e3c..09518fafe7 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -42,12 +42,58 @@ #include "i915_debug.h" #include "i915_screen.h" +/* + * Helper function and arrays + */ + +/** + * Initial offset for Cube map. + */ +static const int initial_offsets[6][2] = { + {0, 0}, + {0, 2}, + {1, 0}, + {1, 2}, + {1, 1}, + {1, 3} +}; + +/** + * Step offsets for Cube map. + */ +static const int step_offsets[6][2] = { + {0, 2}, + {0, 2}, + {-1, 2}, + {-1, 2}, + {-1, 1}, + {-1, 1} +}; static unsigned minify( unsigned d ) { return MAX2(1, d>>1); } +static unsigned +power_of_two(unsigned x) +{ + unsigned value = 1; + while (value <= x) + value = value << 1; + return value; +} + +static unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + + +/* + * More advanced helper funcs + */ static void @@ -86,7 +132,6 @@ i915_miptree_set_level_info(struct i915_texture *tex, tex->image_offset[level][0] = 0; } - static void i915_miptree_set_image_offset(struct i915_texture *tex, unsigned level, unsigned img, unsigned x, unsigned y) @@ -97,27 +142,17 @@ i915_miptree_set_image_offset(struct i915_texture *tex, assert(img < tex->nr_images[level]); tex->image_offset[level][img] = (x + y * tex->pitch); - /* - DBG("%s level %d img %d pos %d,%d image_offset %x\n", + printf("%s level %d img %d pos %d,%d image_offset %x\n", __FUNCTION__, level, img, x, y, tex->image_offset[level][img]); */ } -static unsigned -power_of_two(unsigned x) -{ - unsigned value = 1; - while (value <= x) - value = value << 1; - return value; -} -static unsigned -round_up(unsigned n, unsigned multiple) -{ - return (n + multiple - 1) & ~(multiple - 1); -} +/* + * Layout functions + */ + /** * Special case to deal with display targets. @@ -144,11 +179,11 @@ i915_displaytarget_layout(struct i915_texture *tex) tex->total_height = tex->base.height[0]; } -/* + /* printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, tex->base.width[0], tex->base.height[0], pt->cpp, tex->pitch, tex->total_height, tex->pitch * tex->total_height * 4); -*/ + */ return 1; } @@ -221,25 +256,93 @@ i945_miptree_layout_2d( struct i915_texture *tex ) } } +static void +i945_miptree_layout_cube(struct i915_texture *tex) +{ + struct pipe_texture *pt = &tex->base; + unsigned level; -static const int initial_offsets[6][2] = { - {0, 0}, - {0, 2}, - {1, 0}, - {1, 2}, - {1, 1}, - {1, 3} -}; + const unsigned dim = pt->width[0]; + unsigned face; + unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; -static const int step_offsets[6][2] = { - {0, 2}, - {0, 2}, - {-1, 2}, - {-1, 2}, - {-1, 1}, - {-1, 1} -}; + assert(lvlWidth == lvlHeight); /* cubemap images are square */ + + /* Depending on the size of the largest images, pitch can be + * determined either by the old-style packing of cubemap faces, + * or the final row of 4x4, 2x2 and 1x1 faces below this. + */ + if (dim > 32) + tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; + else + tex->pitch = 14 * 8; + + tex->total_height = dim * 4 + 4; + + /* Set all the levels to effectively occupy the whole rectangular region. + */ + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_level_info(tex, level, 6, 0, 0, lvlWidth, lvlHeight, 1); + lvlWidth /= 2; + lvlHeight /= 2; + } + for (face = 0; face < 6; face++) { + unsigned x = initial_offsets[face][0] * dim; + unsigned y = initial_offsets[face][1] * dim; + unsigned d = dim; + + if (dim == 4 && face >= 4) { + y = tex->total_height - 4; + x = (face - 4) * 8; + } + else if (dim < 4 && (face > 0)) { + y = tex->total_height - 4; + x = face * 8; + } + + for (level = 0; level <= pt->last_level; level++) { + i915_miptree_set_image_offset(tex, level, face, x, y); + + d >>= 1; + + switch (d) { + case 4: + switch (face) { + case PIPE_TEX_FACE_POS_X: + case PIPE_TEX_FACE_NEG_X: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + case PIPE_TEX_FACE_POS_Y: + case PIPE_TEX_FACE_NEG_Y: + y += 12; + x -= 8; + break; + case PIPE_TEX_FACE_POS_Z: + case PIPE_TEX_FACE_NEG_Z: + y = tex->total_height - 4; + x = (face - 4) * 8; + break; + } + + case 2: + y = tex->total_height - 4; + x = 16 + face * 8; + break; + + case 1: + x += 48; + break; + + default: + x += step_offsets[face][0] * d; + y += step_offsets[face][1] * d; + break; + } + } + } +} static boolean i915_miptree_layout(struct i915_texture * tex) @@ -372,99 +475,15 @@ i945_miptree_layout(struct i915_texture * tex) unsigned level; switch (pt->target) { - case PIPE_TEXTURE_CUBE:{ - const unsigned dim = pt->width[0]; - unsigned face; - unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; - - assert(lvlWidth == lvlHeight); /* cubemap images are square */ - - /* Depending on the size of the largest images, pitch can be - * determined either by the old-style packing of cubemap faces, - * or the final row of 4x4, 2x2 and 1x1 faces below this. - */ - if (dim > 32) - tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; - else - tex->pitch = 14 * 8; - - tex->total_height = dim * 4 + 4; - - /* Set all the levels to effectively occupy the whole rectangular region. - */ - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_level_info(tex, level, 6, - 0, 0, - lvlWidth, lvlHeight, 1); - lvlWidth /= 2; - lvlHeight /= 2; - } - - - for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * dim; - unsigned y = initial_offsets[face][1] * dim; - unsigned d = dim; - - if (dim == 4 && face >= 4) { - y = tex->total_height - 4; - x = (face - 4) * 8; - } - else if (dim < 4 && (face > 0)) { - y = tex->total_height - 4; - x = face * 8; - } - - for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_image_offset(tex, level, face, x, y); - - d >>= 1; - - switch (d) { - case 4: - switch (face) { - case PIPE_TEX_FACE_POS_X: - case PIPE_TEX_FACE_NEG_X: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - case PIPE_TEX_FACE_POS_Y: - case PIPE_TEX_FACE_NEG_Y: - y += 12; - x -= 8; - break; - case PIPE_TEX_FACE_POS_Z: - case PIPE_TEX_FACE_NEG_Z: - y = tex->total_height - 4; - x = (face - 4) * 8; - break; - } - - case 2: - y = tex->total_height - 4; - x = 16 + face * 8; - break; - - case 1: - x += 48; - break; - - default: - x += step_offsets[face][0] * d; - y += step_offsets[face][1] * d; - break; - } - } - } - break; - } + case PIPE_TEXTURE_CUBE: + i945_miptree_layout_cube(tex); + break; case PIPE_TEXTURE_3D:{ unsigned width = pt->width[0]; unsigned height = pt->height[0]; unsigned depth = pt->depth[0]; unsigned pack_x_pitch, pack_x_nr; unsigned pack_y_pitch; - unsigned level; tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; tex->total_height = 0; @@ -602,11 +621,6 @@ i915_texture_release(struct pipe_screen *screen, *pt = NULL; } - - -/* - * XXX note: same as code in sp_surface.c - */ static struct pipe_surface * i915_get_tex_surface(struct pipe_screen *screen, struct pipe_texture *pt, @@ -631,7 +645,7 @@ i915_get_tex_surface(struct pipe_screen *screen, assert(zslice == 0); } - ps = CALLOC_STRUCT(pipe_surface);//ws->surface_alloc(ws); + ps = CALLOC_STRUCT(pipe_surface); if (ps) { ps->refcount = 1; ps->winsys = ws; @@ -649,7 +663,6 @@ i915_get_tex_surface(struct pipe_screen *screen, return ps; } - void i915_init_texture_functions(struct i915_context *i915) { -- cgit v1.2.3 From 6fbfcf922210ddf29b73290557f9d40171b09d2e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 19 Jun 2008 22:57:33 +0900 Subject: gallium: Handle malloc failure. --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 6689c3f1fb..13b8c2e5bf 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -165,9 +165,17 @@ tgsi_exec_machine_bind_shader( declarations = (struct tgsi_full_declaration *) MALLOC( maxDeclarations * sizeof(struct tgsi_full_declaration) ); + if (!declarations) { + return; + } + instructions = (struct tgsi_full_instruction *) MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) ); + if (!instructions) { + FREE( declarations ); + return; + } while( !tgsi_parse_end_of_tokens( &parse ) ) { uint pointer = parse.Position; -- cgit v1.2.3 From 8d0329fb6af657a53cb010a3d7a8f4282e2715b8 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 19 Jun 2008 17:09:09 +0200 Subject: i915: Fixed cubemap layouts Apprently we shouldn't do all the advanced layout operation for none compressed formats. The compressed code was also broken, its currently disabled, but should be fixed once i915simple starts to support compressed formats. --- src/gallium/drivers/i915simple/i915_texture.c | 22 +++++++++++++++++++--- 1 file changed, 19 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 09518fafe7..9cd32e3919 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -142,6 +142,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex, assert(img < tex->nr_images[level]); tex->image_offset[level][img] = (x + y * tex->pitch); + /* printf("%s level %d img %d pos %d,%d image_offset %x\n", __FUNCTION__, level, img, x, y, tex->image_offset[level][img]); @@ -266,9 +267,17 @@ i945_miptree_layout_cube(struct i915_texture *tex) unsigned face; unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; + /* + printf("%s %i, %i\n", __FUNCTION__, pt->width[0], pt->height[0]); + */ + assert(lvlWidth == lvlHeight); /* cubemap images are square */ - /* Depending on the size of the largest images, pitch can be + /* + * XXX Should only be used for compressed formats. But lets + * keep this code active just in case. + * + * Depending on the size of the largest images, pitch can be * determined either by the old-style packing of cubemap faces, * or the final row of 4x4, 2x2 and 1x1 faces below this. */ @@ -277,6 +286,9 @@ i945_miptree_layout_cube(struct i915_texture *tex) else tex->pitch = 14 * 8; + /* + * XXX The 4 is only needed for compressed formats. See above. + */ tex->total_height = dim * 4 + 4; /* Set all the levels to effectively occupy the whole rectangular region. @@ -292,6 +304,7 @@ i945_miptree_layout_cube(struct i915_texture *tex) unsigned y = initial_offsets[face][1] * dim; unsigned d = dim; +#if 0 /* Fix and enable this code for compressed formats */ if (dim == 4 && face >= 4) { y = tex->total_height - 4; x = (face - 4) * 8; @@ -300,12 +313,14 @@ i945_miptree_layout_cube(struct i915_texture *tex) y = tex->total_height - 4; x = face * 8; } +#endif for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_image_offset(tex, level, face, x, y); d >>= 1; +#if 0 /* Fix and enable this code for compressed formats */ switch (d) { case 4: switch (face) { @@ -325,7 +340,6 @@ i945_miptree_layout_cube(struct i915_texture *tex) x = (face - 4) * 8; break; } - case 2: y = tex->total_height - 4; x = 16 + face * 8; @@ -334,12 +348,14 @@ i945_miptree_layout_cube(struct i915_texture *tex) case 1: x += 48; break; - default: +#endif x += step_offsets[face][0] * d; y += step_offsets[face][1] * d; +#if 0 break; } +#endif } } } -- cgit v1.2.3 From a8533d54930f8fa989036c197ad20b0778ec0cac Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 19 Jun 2008 16:06:56 -0600 Subject: egl: clean up prototype code, new _eglFindAPIs() function. --- src/egl/main/eglcontext.c | 2 + src/egl/main/egldriver.c | 118 +++++++++++++++++++++++++++++++++++----------- src/egl/main/egldriver.h | 4 ++ src/egl/main/eglmisc.c | 52 -------------------- 4 files changed, 97 insertions(+), 79 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglcontext.c b/src/egl/main/eglcontext.c index bf1addf241..15847ef2da 100644 --- a/src/egl/main/eglcontext.c +++ b/src/egl/main/eglcontext.c @@ -33,6 +33,8 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, return EGL_FALSE; } + ctx->ClientVersion = 1; /* the default, per EGL spec */ + for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { switch (attrib_list[i]) { case EGL_CONTEXT_CLIENT_VERSION: diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index 4a611b9fc9..edf85abe01 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -35,6 +35,44 @@ static const char *SysFS = "/sys/class"; +/** + * Wrappers for dlopen/dlclose() + */ +#if defined(_EGL_PLATFORM_WINDOWS) + + typedef HMODULE lib_handle; + + static HMODULE + open_library(const char *filename) + { + return LoadLibrary(filename); + } + + static void + close_library(HMODULE lib) + { + FreeLibrary(lib); + } + +#elif defined(_EGL_PLATFORM_X) + + typedef void * lib_handle; + + static void * + open_library(const char *filename) + { + return dlopen(filename, RTLD_LAZY); + } + + static void + close_library(void *lib) + { + dlclose(lib); + } + +#endif + + /** * Given a card number, use sysfs to determine the DRI driver name. @@ -165,11 +203,7 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) { _EGLDriver *drv; _EGLMain_t mainFunc; -#if defined(_EGL_PLATFORM_WINDOWS) - HMODULE lib; -#elif defined(_EGL_PLATFORM_X) - void *lib; -#endif + lib_handle lib; char driverFilename[1000]; assert(driverName); @@ -178,13 +212,12 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) /* XXX untested */ sprintf(driverFilename, "%s.dll", driverName); _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename); - lib = LoadLibrary(driverFilename); #elif defined(_EGL_PLATFORM_X) /* XXX also prepend a directory path??? */ sprintf(driverFilename, "%s.so", driverName); _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename); - lib = dlopen(driverFilename, RTLD_LAZY); #endif + lib = open_library(driverFilename); if (!lib) { _eglLog(_EGL_WARNING, "Could not open %s (%s)", @@ -200,21 +233,13 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) if (!mainFunc) { _eglLog(_EGL_WARNING, "_eglMain not found in %s", driverFilename); -#if defined(_EGL_PLATFORM_WINDOWS) - FreeLibrary(lib); -#elif defined(_EGL_PLATFORM_X) - dlclose(lib); -#endif + close_library(lib); return NULL; } drv = mainFunc(dpy, args); if (!drv) { -#if defined(_EGL_PLATFORM_WINDOWS) - FreeLibrary(lib); -#elif defined(_EGL_PLATFORM_X) - dlclose(lib); -#endif + close_library(lib); return NULL; } @@ -223,11 +248,7 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) drv->LibHandle = lib; } else { -#if defined(_EGL_PLATFORM_WINDOWS) - FreeLibrary(lib); -#elif defined(_EGL_PLATFORM_X) - dlclose(lib); -#endif + close_library(lib); } /* update the global notion of supported APIs */ @@ -253,11 +274,7 @@ _eglCloseDriver(_EGLDriver *drv, EGLDisplay dpy) b = drv->API.Terminate(drv, dpy); -#if defined(_EGL_PLATFORM_WINDOWS) - FreeLibrary(handle); -#elif defined(_EGL_PLATFORM_X) - dlclose(handle); -#endif + close_library(handle); return b; } @@ -343,3 +360,50 @@ _eglInitDriverFallbacks(_EGLDriver *drv) drv->API.CreatePbufferFromClientBuffer = _eglCreatePbufferFromClientBuffer; #endif /* EGL_VERSION_1_2 */ } + + + +/** + * Try to determine which EGL APIs (OpenGL, OpenGL ES, OpenVG, etc) + * are supported on the system by looking for standard library names. + */ +EGLint +_eglFindAPIs(void) +{ + EGLint mask = 0x0; + lib_handle lib; +#if defined(_EGL_PLATFORM_WINDOWS) + /* XXX not sure about these names */ + const char *es1_libname = "libGLESv1_CM.dll"; + const char *es2_libname = "libGLESv2.dll"; + const char *gl_libname = "OpenGL32.dll"; + const char *vg_libname = "libOpenVG.dll"; +#elif defined(_EGL_PLATFORM_X) + const char *es1_libname = "libGLESv1_CM.so"; + const char *es2_libname = "libGLESv2.so"; + const char *gl_libname = "libGL.so"; + const char *vg_libname = "libOpenVG.so"; +#endif + + if ((lib = open_library(es1_libname))) { + close_library(lib); + mask |= EGL_OPENGL_ES_BIT; + } + + if ((lib = open_library(es2_libname))) { + close_library(lib); + mask |= EGL_OPENGL_ES2_BIT; + } + + if ((lib = open_library(gl_libname))) { + close_library(lib); + mask |= EGL_OPENGL_BIT; + } + + if ((lib = open_library(vg_libname))) { + close_library(lib); + mask |= EGL_OPENVG_BIT; + } + + return mask; +} diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index 1eae6f8034..7827ca3ae3 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -72,4 +72,8 @@ extern void _eglInitDriverFallbacks(_EGLDriver *drv); +extern EGLint +_eglFindAPIs(void); + + #endif /* EGLDRIVER_INCLUDED */ diff --git a/src/egl/main/eglmisc.c b/src/egl/main/eglmisc.c index e79223657d..b5bdc3ea4b 100644 --- a/src/egl/main/eglmisc.c +++ b/src/egl/main/eglmisc.c @@ -54,58 +54,6 @@ _eglUpdateExtensionsString(_EGLDriver *drv) } -#if 0 /* prototype code */ - -#include - -static EGLint -_eglFindAPIs_by_sym(void) -{ - EGLint mask = 0x0; - - if (dlsym(NULL, "glBegin")) - mask |= EGL_OPENGL_BIT; - if (dlsym(NULL, "glGetFixedv")) - mask |= EGL_OPENGL_ES_BIT; - if (dlsym(NULL, "vgSetf")) - mask |= EGL_OPENVG_BIT; - - return mask; -} - -static EGLint -_eglFindAPIs_by_lib(void) -{ - EGLint mask = 0x0; - int flag = RTLD_NOW; - void *h; - - if ((h = dlopen("libGLESv1_CM.so", flag))) { - dlclose(h); - mask |= EGL_OPENGL_ES_BIT; - } - - if ((h = dlopen("libGLESv2.so", flag))) { - dlclose(h); - mask |= EGL_OPENGL_ES2_BIT; - } - - if ((h = dlopen("libGL.so", flag))) { - dlclose(h); - mask |= EGL_OPENGL_BIT; - } - - if ((h = dlopen("libOpenVG.so", flag))) { - dlclose(h); - mask |= EGL_OPENVG_BIT; - } - - return mask; -} - -#endif - - static void _eglUpdateAPIsString(_EGLDriver *drv) { -- cgit v1.2.3 From 064001dbe7fd6cbb6bd5c91bdf7f0831b9c351a9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 19 Jun 2008 16:12:17 -0600 Subject: egl: use dlsym() to try to identify APIs --- src/gallium/winsys/egl_xlib/egl_xlib.c | 47 ++++++++++++++++++++++++++++++++-- 1 file changed, 45 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 194bf2842a..85fc83a4dc 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -32,6 +32,7 @@ */ +#include #include #include "pipe/p_compiler.h" @@ -202,7 +203,6 @@ xlib_eglInitialize(_EGLDriver *drv, EGLDisplay dpy, static EGLBoolean xlib_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) { - return EGL_TRUE; } @@ -342,6 +342,10 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, /* API-dependent context creation */ switch (ctx->Base.ClientAPI) { + case EGL_OPENGL_ES_API: + _eglLog(_EGL_DEBUG, "Create Context for ES version %d\n", + ctx->Base.ClientVersion); + /* fall-through */ case EGL_OPENGL_API: /* create a softpipe context */ ctx->pipe = softpipe_create(xdrv->screen, xdrv->winsys, NULL); @@ -372,6 +376,8 @@ xlib_eglDestroyContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx) else { /* API-dependent clean-up */ switch (context->Base.ClientAPI) { + case EGL_OPENGL_ES_API: + /* fall-through */ case EGL_OPENGL_API: st_destroy_context(context->Context); break; @@ -554,6 +560,36 @@ xlib_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) } +/** + * Determine which API(s) is(are) present by looking for some specific + * global symbols. + */ +static EGLint +find_supported_apis(void) +{ + EGLint mask = 0; + void *handle; + + handle = dlopen(NULL, 0); + + if (dlsym(handle, "st_api_OpenGL_ES1")) + mask |= EGL_OPENGL_ES_BIT; + + if (dlsym(handle, "st_api_OpenGL_ES2")) + mask |= EGL_OPENGL_ES2_BIT; + + if (dlsym(handle, "st_api_OpenGL")) + mask |= EGL_OPENGL_BIT; + + if (dlsym(handle, "st_api_OpenVG")) + mask |= EGL_OPENVG_BIT; + + dlclose(handle); + + return mask; +} + + /** * This is the main entrypoint into the driver. * Called by libEGL to instantiate an _EGLDriver object. @@ -584,7 +620,14 @@ _eglMain(_EGLDisplay *dpy, const char *args) xdrv->Base.API.MakeCurrent = xlib_eglMakeCurrent; xdrv->Base.API.SwapBuffers = xlib_eglSwapBuffers; - xdrv->Base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; + xdrv->Base.ClientAPIsMask = find_supported_apis(); + if (xdrv->Base.ClientAPIsMask == 0x0) { + /* the app isn't directly linked with any EGL-supprted APIs + * (such as libGLESv2.so) so use an EGL utility to see what + * APIs might be loaded dynamically on this system. + */ + xdrv->Base.ClientAPIsMask = _eglFindAPIs(); + } xdrv->Base.Name = "Xlib/softpipe"; -- cgit v1.2.3 From 158298eea075e87e4e3b4da9f6a606550f975b5d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 19 Jun 2008 16:15:09 -0600 Subject: gallium: remove unused vars --- src/mesa/state_tracker/st_cb_fbo.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 76145a4d5c..db25ddd615 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -347,11 +347,8 @@ st_render_texture(GLcontext *ctx, struct gl_framebuffer *fb, struct gl_renderbuffer_attachment *att) { - struct st_context *st = ctx->st; struct st_renderbuffer *strb; struct gl_renderbuffer *rb; - struct pipe_context *pipe = st->pipe; - struct pipe_screen *screen = pipe->screen; struct pipe_texture *pt; struct st_texture_object *stObj; const struct gl_texture_image *texImage = -- cgit v1.2.3 From 78d05bcc27ba02282711174a8b3aa5f1ebea9151 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 19 Jun 2008 16:15:56 -0600 Subject: gallium: new st_api.c file with token symbol for identifying the API --- src/mesa/sources | 1 + src/mesa/state_tracker/st_api.c | 33 +++++++++++++++++++++++++++++++++ 2 files changed, 34 insertions(+) create mode 100644 src/mesa/state_tracker/st_api.c (limited to 'src') diff --git a/src/mesa/sources b/src/mesa/sources index 7f49ac6e32..e6b050c3f2 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -196,6 +196,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_readpixels.c \ state_tracker/st_cb_strings.c \ state_tracker/st_cb_texture.c \ + state_tracker/st_api.c \ state_tracker/st_context.c \ state_tracker/st_debug.c \ state_tracker/st_draw.c \ diff --git a/src/mesa/state_tracker/st_api.c b/src/mesa/state_tracker/st_api.c new file mode 100644 index 0000000000..fc0e9a2316 --- /dev/null +++ b/src/mesa/state_tracker/st_api.c @@ -0,0 +1,33 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Just a global symbol for EGL to look for to identify the supported + * graphics API. + */ +int st_api_OpenGL = 1; -- cgit v1.2.3 From 9f6a4e2a65ff872bc27e7081df7d6205393a1180 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 19 Jun 2008 21:19:33 -0600 Subject: egl: fix assorted context-related bugs Move memset() to proper place. Added EGL_CONTEXT_CLIENT_VERSION query. Fix bad return EGL_FALSE -> EGL_TRUE. --- src/egl/main/eglcontext.c | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglcontext.c b/src/egl/main/eglcontext.c index 15847ef2da..461679db09 100644 --- a/src/egl/main/eglcontext.c +++ b/src/egl/main/eglcontext.c @@ -33,6 +33,8 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, return EGL_FALSE; } + memset(ctx, 0, sizeof(_EGLContext)); + ctx->ClientVersion = 1; /* the default, per EGL spec */ for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { @@ -47,7 +49,6 @@ _eglInitContext(_EGLDriver *drv, EGLDisplay dpy, _EGLContext *ctx, } } - memset(ctx, 0, sizeof(_EGLContext)); ctx->Display = display; ctx->Config = conf; ctx->DrawSurface = EGL_NO_SURFACE; @@ -188,8 +189,11 @@ _eglQueryContext(_EGLDriver *drv, EGLDisplay dpy, EGLContext ctx, #ifdef EGL_VERSION_1_2 case EGL_CONTEXT_CLIENT_TYPE: *value = c->ClientAPI; - return EGL_FALSE; + return EGL_TRUE; #endif /* EGL_VERSION_1_2 */ + case EGL_CONTEXT_CLIENT_VERSION: + *value = c->ClientVersion; + return EGL_TRUE; default: _eglError(EGL_BAD_ATTRIBUTE, "eglQueryContext"); return EGL_FALSE; -- cgit v1.2.3 From 907c0978affb6bc7f8cb077f568829ecfaa89b04 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 08:07:38 -0600 Subject: mesa: test for FEATURE_ATI_fragment_shader --- src/mesa/main/context.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 8aab1eebc8..bd90027c8a 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -149,7 +149,9 @@ #include "glapi/glapitable.h" #include "shader/program.h" #include "shader/shader_api.h" +#if FEATURE_ATI_fragment_shader #include "shader/atifragshader.h" +#endif #if _HAVE_FULL_GL #include "math/m_translate.h" #include "math/m_matrix.h" -- cgit v1.2.3 From 2dcd4ce4b62122d1088d130bade83bd051fa6d0c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 08:09:46 -0600 Subject: gallium: remove unneeded #include --- src/mesa/state_tracker/st_gen_mipmap.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index f51eff05dc..851f17c3b4 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -46,7 +46,6 @@ #include "st_gen_mipmap.h" #include "st_program.h" #include "st_texture.h" -#include "st_cb_drawpixels.h" #include "st_cb_texture.h" -- cgit v1.2.3 From d7a4d498e4fe2ba8036626bc97913370ad991fac Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 08:09:59 -0600 Subject: gallium: assorted FEATURE tests --- src/mesa/state_tracker/st_context.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index d394562feb..5da5bcfd37 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -38,15 +38,21 @@ #include "st_context.h" #include "st_cb_accum.h" #include "st_cb_bitmap.h" +#if FEATURE_EXT_framebuffer_blit #include "st_cb_blit.h" +#endif #include "st_cb_bufferobjects.h" #include "st_cb_clear.h" +#if FEATURE_drawpix #include "st_cb_drawpixels.h" +#include "st_cb_rasterpos.h" +#endif #include "st_cb_fbo.h" +#if FEATURE_feedback #include "st_cb_feedback.h" +#endif #include "st_cb_program.h" #include "st_cb_queryobj.h" -#include "st_cb_rasterpos.h" #include "st_cb_readpixels.h" #include "st_cb_texture.h" #include "st_cb_flush.h" @@ -114,7 +120,9 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st_init_clear(st); st_init_draw( st ); st_init_generate_mipmap(st); +#if FEATURE_EXT_framebuffer_blit st_init_blit(st); +#endif for (i = 0; i < PIPE_MAX_SAMPLERS; i++) st->state.sampler_list[i] = &st->state.samplers[i]; -- cgit v1.2.3 From cae4444fa179c9ed556f1d1d207967b4e72736e2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 10:19:51 -0600 Subject: egl: default API should be ES --- src/egl/main/eglglobals.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/eglglobals.c b/src/egl/main/eglglobals.c index 9a077edbd7..5da59af91d 100644 --- a/src/egl/main/eglglobals.c +++ b/src/egl/main/eglglobals.c @@ -49,7 +49,7 @@ _eglNewThreadInfo(void) if (t) { t->CurrentContext = EGL_NO_CONTEXT; t->LastError = EGL_SUCCESS; - t->CurrentAPI = EGL_NONE; + t->CurrentAPI = EGL_OPENGL_ES_API; /* default, per EGL spec */ } return t; } -- cgit v1.2.3 From 8a75d109531755358708de5381eb912b7e2bf69c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 10:20:06 -0600 Subject: egl: added null ptr check --- src/egl/main/eglconfig.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c index 794a783b9e..3ef0564a54 100644 --- a/src/egl/main/eglconfig.c +++ b/src/egl/main/eglconfig.c @@ -340,8 +340,10 @@ _eglChooseConfig(_EGLDriver *drv, EGLDisplay dpy, const EGLint *attrib_list, qsort(configList, count, sizeof(_EGLConfig *), _eglCompareConfigs); /* copy config handles to output array */ - for (i = 0; i < count; i++) { - configs[i] = configList[i]->Handle; + if (configs) { + for (i = 0; i < count; i++) { + configs[i] = configList[i]->Handle; + } } free(configList); -- cgit v1.2.3 From fbd6e86b8f3c9bdeae7581d8d852b4df66e4b42c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 10:20:25 -0600 Subject: egl: implment EGL_LARGEST_PBUFFER query --- src/egl/main/egldriver.h | 2 ++ src/egl/main/eglsurface.c | 5 ++++- 2 files changed, 6 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/egldriver.h b/src/egl/main/egldriver.h index 7827ca3ae3..4066c6ec1d 100644 --- a/src/egl/main/egldriver.h +++ b/src/egl/main/egldriver.h @@ -39,6 +39,8 @@ struct _egl_driver _EGLAPI API; /**< EGL API dispatch table */ _EGLExtensions Extensions; + + int LargestPbuffer; }; diff --git a/src/egl/main/eglsurface.c b/src/egl/main/eglsurface.c index 796d62f2b9..3777049ca6 100644 --- a/src/egl/main/eglsurface.c +++ b/src/egl/main/eglsurface.c @@ -8,6 +8,7 @@ #include #include "eglcontext.h" #include "eglconfig.h" +#include "egldriver.h" #include "eglglobals.h" #include "eglhash.h" #include "egllog.h" @@ -319,7 +320,9 @@ _eglQuerySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surf, case EGL_CONFIG_ID: *value = GET_CONFIG_ATTRIB(surface->Config, EGL_CONFIG_ID); return EGL_TRUE; - /*XXX case EGL_LARGEST_PBUFFER:*/ + case EGL_LARGEST_PBUFFER: + *value = drv->LargestPbuffer; + return EGL_TRUE; case EGL_SURFACE_TYPE: *value = surface->Type; return EGL_TRUE; -- cgit v1.2.3 From a87e717d0d302bc376eff29d17d35f1314ff1a57 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 10:20:47 -0600 Subject: egl: added null ptr checks --- src/gallium/winsys/egl_xlib/egl_xlib.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 85fc83a4dc..eeb15e30a9 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -413,7 +413,10 @@ xlib_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, (draw_surf ? draw_surf->Framebuffer : NULL), (read_surf ? read_surf->Framebuffer : NULL)); - check_and_update_buffer_size(draw_surf); + if (draw_surf) + check_and_update_buffer_size(draw_surf); + if (read_surf && read_surf != draw_surf) + check_and_update_buffer_size(draw_surf); return EGL_TRUE; } -- cgit v1.2.3 From 95c9fc82f58a8f38d25b3e405891566c8f8a51f6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 10:47:38 -0600 Subject: mesa: fix some FEATURE_x tests --- src/mesa/main/context.c | 7 +------ 1 file changed, 1 insertion(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index bd90027c8a..e5264bcaf9 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -441,9 +441,7 @@ alloc_shared_state( GLcontext *ctx ) ss->DisplayList = _mesa_NewHashTable(); ss->TexObjects = _mesa_NewHashTable(); -#if FEATURE_NV_vertex_program || FEATURE_NV_fragment_program ss->Programs = _mesa_NewHashTable(); -#endif #if FEATURE_ARB_vertex_program ss->DefaultVertexProgram = (struct gl_vertex_program *) @@ -533,10 +531,8 @@ alloc_shared_state( GLcontext *ctx ) _mesa_DeleteHashTable(ss->DisplayList); if (ss->TexObjects) _mesa_DeleteHashTable(ss->TexObjects); -#if FEATURE_NV_vertex_program if (ss->Programs) _mesa_DeleteHashTable(ss->Programs); -#endif #if FEATURE_ARB_vertex_program _mesa_reference_vertprog(ctx, &ss->DefaultVertexProgram, NULL); #endif @@ -723,10 +719,9 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) _mesa_DeleteHashTable(ss->ShaderObjects); #endif -#if defined(FEATURE_NV_vertex_program) || defined(FEATURE_NV_fragment_program) _mesa_HashDeleteAll(ss->Programs, delete_program_cb, ctx); _mesa_DeleteHashTable(ss->Programs); -#endif + #if FEATURE_ARB_vertex_program _mesa_reference_vertprog(ctx, &ss->DefaultVertexProgram, NULL); #endif -- cgit v1.2.3 From 71f67dde34546f40c7617948f8fa945288e55e17 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 10:48:14 -0600 Subject: mesa: _vbo_Color4f, _vbo_Normal3f, _vbo_MultiTexCoord4f functions --- src/mesa/vbo/vbo.h | 10 ++++++++++ src/mesa/vbo/vbo_exec_api.c | 21 +++++++++++++++++++++ 2 files changed, 31 insertions(+) (limited to 'src') diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h index 79d33d09c1..1a737babd3 100644 --- a/src/mesa/vbo/vbo.h +++ b/src/mesa/vbo/vbo.h @@ -120,4 +120,14 @@ void vbo_use_buffer_objects(GLcontext *ctx); void vbo_set_draw_func(GLcontext *ctx, vbo_draw_func func); +void +_vbo_Color4f(GLfloat r, GLfloat g, GLfloat b, GLfloat a); + +void +_vbo_Normal3f(GLfloat x, GLfloat y, GLfloat z); + +void +_vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q); + + #endif diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index b263885388..cf9c860263 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -769,3 +769,24 @@ static void reset_attrfv( struct vbo_exec_context *exec ) exec->vtx.vertex_size = 0; } + +void +_vbo_Color4f(GLfloat r, GLfloat g, GLfloat b, GLfloat a) +{ + vbo_Color4f(r, g, b, a); +} + + +void +_vbo_Normal3f(GLfloat x, GLfloat y, GLfloat z) +{ + vbo_Normal3f(x, y, z); +} + + +void +_vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q) +{ + vbo_MultiTexCoord4f(target, s, t, r, q); +} + -- cgit v1.2.3 From 1b241a4369fcba5972a7ab38b9a5595a434e0b56 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 11:04:37 -0600 Subject: mesa: added _vbo_Materialfv() --- src/mesa/vbo/vbo.h | 3 +++ src/mesa/vbo/vbo_exec_api.c | 5 +++++ 2 files changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h index 1a737babd3..bd10956ee4 100644 --- a/src/mesa/vbo/vbo.h +++ b/src/mesa/vbo/vbo.h @@ -129,5 +129,8 @@ _vbo_Normal3f(GLfloat x, GLfloat y, GLfloat z); void _vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q); +void +_vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params); + #endif diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index cf9c860263..1e2a08bc8e 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -790,3 +790,8 @@ _vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q) vbo_MultiTexCoord4f(target, s, t, r, q); } +void +_vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params) +{ + vbo_Materialfv(face, pname, params); +} -- cgit v1.2.3 From 9cae37870e66550d8cceac4b4a8765c1936d4ddc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 11:05:00 -0600 Subject: mesa: revamp glBlendFunc loopback --- src/mesa/main/blend.c | 10 +--------- src/mesa/main/dlist.c | 9 ++++++++- 2 files changed, 9 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/main/blend.c b/src/mesa/main/blend.c index 742247f8e2..4d4a897141 100644 --- a/src/mesa/main/blend.c +++ b/src/mesa/main/blend.c @@ -46,19 +46,11 @@ * \param dfactor destination factor operator. * * \sa glBlendFunc, glBlendFuncSeparateEXT - * - * Swizzles the inputs and calls \c glBlendFuncSeparateEXT. This is done - * using the \c CurrentDispatch table in the context, so this same function - * can be used while compiling display lists. Therefore, there is no need - * for the display list code to save and restore this function. */ void GLAPIENTRY _mesa_BlendFunc( GLenum sfactor, GLenum dfactor ) { - GET_CURRENT_CONTEXT(ctx); - - (*ctx->CurrentDispatch->BlendFuncSeparateEXT)( sfactor, dfactor, - sfactor, dfactor ); + _mesa_BlendFuncSeparateEXT(sfactor, dfactor, sfactor, dfactor); } diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c index 07d279da30..c672c0f8b8 100644 --- a/src/mesa/main/dlist.c +++ b/src/mesa/main/dlist.c @@ -925,6 +925,13 @@ save_BlendFuncSeparateEXT(GLenum sfactorRGB, GLenum dfactorRGB, } +static void GLAPIENTRY +save_BlendFunc(GLenum srcfactor, GLenum dstfactor) +{ + save_BlendFuncSeparate(srcfactor, dstfactor, srcfactor, dstfactor); +} + + static void GLAPIENTRY save_BlendColor(GLfloat red, GLfloat green, GLfloat blue, GLfloat alpha) { @@ -7602,7 +7609,7 @@ _mesa_init_dlist_table(struct _glapi_table *table) SET_Accum(table, save_Accum); SET_AlphaFunc(table, save_AlphaFunc); SET_Bitmap(table, save_Bitmap); - SET_BlendFunc(table, _mesa_BlendFunc); /* loops-back to BlendFuncSeparate */ + SET_BlendFunc(table, _save_BlendFunc); SET_CallList(table, _mesa_save_CallList); SET_CallLists(table, _mesa_save_CallLists); SET_Clear(table, save_Clear); -- cgit v1.2.3 From 39ce3940828fb816d7c1cf7f061d85f7db9a0c28 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 11:15:30 -0600 Subject: mesa: fix errors in prev commit --- src/mesa/main/dlist.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c index c672c0f8b8..b4ed300b2e 100644 --- a/src/mesa/main/dlist.c +++ b/src/mesa/main/dlist.c @@ -928,7 +928,7 @@ save_BlendFuncSeparateEXT(GLenum sfactorRGB, GLenum dfactorRGB, static void GLAPIENTRY save_BlendFunc(GLenum srcfactor, GLenum dstfactor) { - save_BlendFuncSeparate(srcfactor, dstfactor, srcfactor, dstfactor); + save_BlendFuncSeparateEXT(srcfactor, dstfactor, srcfactor, dstfactor); } @@ -7609,7 +7609,7 @@ _mesa_init_dlist_table(struct _glapi_table *table) SET_Accum(table, save_Accum); SET_AlphaFunc(table, save_AlphaFunc); SET_Bitmap(table, save_Bitmap); - SET_BlendFunc(table, _save_BlendFunc); + SET_BlendFunc(table, save_BlendFunc); SET_CallList(table, _mesa_save_CallList); SET_CallLists(table, _mesa_save_CallLists); SET_Clear(table, save_Clear); -- cgit v1.2.3 From 62d1d591d5c4d6efdbb10b45e3a38116f95794b6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 11:30:21 -0600 Subject: gallium: remove unneeded #include --- src/mesa/state_tracker/st_cb_blit.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c index 799ef3f3be..327bafeb98 100644 --- a/src/mesa/state_tracker/st_cb_blit.c +++ b/src/mesa/state_tracker/st_cb_blit.c @@ -40,7 +40,6 @@ #include "st_context.h" #include "st_program.h" -#include "st_cb_drawpixels.h" #include "st_cb_blit.h" #include "st_cb_fbo.h" -- cgit v1.2.3 From 722db809e64cab61846c3f086fcba4847caad216 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 11:30:43 -0600 Subject: gallium: always need st_init-blit() --- src/mesa/state_tracker/st_context.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 5da5bcfd37..bf78cacb8e 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -38,9 +38,7 @@ #include "st_context.h" #include "st_cb_accum.h" #include "st_cb_bitmap.h" -#if FEATURE_EXT_framebuffer_blit #include "st_cb_blit.h" -#endif #include "st_cb_bufferobjects.h" #include "st_cb_clear.h" #if FEATURE_drawpix @@ -120,9 +118,7 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) st_init_clear(st); st_init_draw( st ); st_init_generate_mipmap(st); -#if FEATURE_EXT_framebuffer_blit st_init_blit(st); -#endif for (i = 0; i < PIPE_MAX_SAMPLERS; i++) st->state.sampler_list[i] = &st->state.samplers[i]; -- cgit v1.2.3 From 8e0f166eb5edb8537af573d8d33a26ffaf8e66c9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 11:31:05 -0600 Subject: mesa: refactor: move initialization of DefaultPacking state. --- src/mesa/main/pixel.c | 20 -------------------- src/mesa/main/pixelstore.c | 20 ++++++++++++++++++++ 2 files changed, 20 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c index c9c4289f69..c98506b2bb 100644 --- a/src/mesa/main/pixel.c +++ b/src/mesa/main/pixel.c @@ -910,26 +910,6 @@ _mesa_init_pixel( GLcontext *ctx ) ASSIGN_4V(ctx->Pixel.TextureColorTableScale, 1.0, 1.0, 1.0, 1.0); ASSIGN_4V(ctx->Pixel.TextureColorTableBias, 0.0, 0.0, 0.0, 0.0); - /* - * _mesa_unpack_image() returns image data in this format. When we - * execute image commands (glDrawPixels(), glTexImage(), etc) from - * within display lists we have to be sure to set the current - * unpacking parameters to these values! - */ - ctx->DefaultPacking.Alignment = 1; - ctx->DefaultPacking.RowLength = 0; - ctx->DefaultPacking.SkipPixels = 0; - ctx->DefaultPacking.SkipRows = 0; - ctx->DefaultPacking.ImageHeight = 0; - ctx->DefaultPacking.SkipImages = 0; - ctx->DefaultPacking.SwapBytes = GL_FALSE; - ctx->DefaultPacking.LsbFirst = GL_FALSE; - ctx->DefaultPacking.ClientStorage = GL_FALSE; - ctx->DefaultPacking.Invert = GL_FALSE; -#if FEATURE_EXT_pixel_buffer_object - ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj; -#endif - if (ctx->Visual.doubleBufferMode) { ctx->Pixel.ReadBuffer = GL_BACK; } diff --git a/src/mesa/main/pixelstore.c b/src/mesa/main/pixelstore.c index 3bf89bd3e1..ff1a6344cc 100644 --- a/src/mesa/main/pixelstore.c +++ b/src/mesa/main/pixelstore.c @@ -260,4 +260,24 @@ _mesa_init_pixelstore( GLcontext *ctx ) #if FEATURE_EXT_pixel_buffer_object ctx->Unpack.BufferObj = ctx->Array.NullBufferObj; #endif + + /* + * _mesa_unpack_image() returns image data in this format. When we + * execute image commands (glDrawPixels(), glTexImage(), etc) from + * within display lists we have to be sure to set the current + * unpacking parameters to these values! + */ + ctx->DefaultPacking.Alignment = 1; + ctx->DefaultPacking.RowLength = 0; + ctx->DefaultPacking.SkipPixels = 0; + ctx->DefaultPacking.SkipRows = 0; + ctx->DefaultPacking.ImageHeight = 0; + ctx->DefaultPacking.SkipImages = 0; + ctx->DefaultPacking.SwapBytes = GL_FALSE; + ctx->DefaultPacking.LsbFirst = GL_FALSE; + ctx->DefaultPacking.ClientStorage = GL_FALSE; + ctx->DefaultPacking.Invert = GL_FALSE; +#if FEATURE_EXT_pixel_buffer_object + ctx->DefaultPacking.BufferObj = ctx->Array.NullBufferObj; +#endif } -- cgit v1.2.3 From 42c468a5dea30d1428205d81798eaf3a0c42ef3a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 11:32:22 -0600 Subject: mesa: initial support for fixed-pt vertex arrays --- src/mesa/main/glheader.h | 5 +++++ src/mesa/main/mfeatures.h | 1 + src/mesa/main/varray.c | 20 ++++++++++++++++++++ 3 files changed, 26 insertions(+) (limited to 'src') diff --git a/src/mesa/main/glheader.h b/src/mesa/main/glheader.h index fd4127558a..0f74bc83cc 100644 --- a/src/mesa/main/glheader.h +++ b/src/mesa/main/glheader.h @@ -145,6 +145,11 @@ #include "GL/glext.h" +#ifndef GL_FIXED +#define GL_FIXED 0x140C +#endif + + #if !defined(CAPI) && defined(WIN32) && !defined(BUILD_FOR_SNAP) #define CAPI _cdecl #endif diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h index 6a71954648..13397e97ba 100644 --- a/src/mesa/main/mfeatures.h +++ b/src/mesa/main/mfeatures.h @@ -46,6 +46,7 @@ #define FEATURE_drawpix _HAVE_FULL_GL #define FEATURE_evaluators _HAVE_FULL_GL #define FEATURE_feedback _HAVE_FULL_GL +#define FEATURE_fixedpt 0 #define FEATURE_histogram _HAVE_FULL_GL #define FEATURE_pixel_transfer _HAVE_FULL_GL #define FEATURE_texgen _HAVE_FULL_GL diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c index fe4a7c684f..9474724c33 100644 --- a/src/mesa/main/varray.c +++ b/src/mesa/main/varray.c @@ -121,6 +121,11 @@ _mesa_VertexPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr) case GL_DOUBLE: elementSize = size * sizeof(GLdouble); break; +#if FEATURE_fixedpt + case GL_FIXED: + elementSize = size * sizeof(GLfixed); + break; +#endif default: _mesa_error( ctx, GL_INVALID_ENUM, "glVertexPointer(type)" ); return; @@ -166,6 +171,11 @@ _mesa_NormalPointer(GLenum type, GLsizei stride, const GLvoid *ptr ) case GL_DOUBLE: elementSize = 3 * sizeof(GLdouble); break; +#if FEATURE_fixedpt + case GL_FIXED: + elementSize = 3 * sizeof(GLfixed); + break; +#endif default: _mesa_error( ctx, GL_INVALID_ENUM, "glNormalPointer(type)" ); return; @@ -224,6 +234,11 @@ _mesa_ColorPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr) case GL_DOUBLE: elementSize = size * sizeof(GLdouble); break; +#if FEATURE_fixedpt + case GL_FIXED: + elementSize = size * sizeof(GLfixed); + break; +#endif default: _mesa_error( ctx, GL_INVALID_ENUM, "glColorPointer(type)" ); return; @@ -405,6 +420,11 @@ _mesa_TexCoordPointer(GLint size, GLenum type, GLsizei stride, case GL_DOUBLE: elementSize = size * sizeof(GLdouble); break; +#if FEATURE_fixedpt + case GL_FIXED: + elementSize = size * sizeof(GLfixed); + break; +#endif default: _mesa_error( ctx, GL_INVALID_ENUM, "glTexCoordPointer(type)" ); return; -- cgit v1.2.3 From a9b46b9e4c79665febb21180150ba54731aa4bc9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 11:49:25 -0600 Subject: mesa: GL_BYTE vertex/texcoord arrays --- src/mesa/main/mfeatures.h | 1 + src/mesa/main/varray.c | 10 ++++++++++ 2 files changed, 11 insertions(+) (limited to 'src') diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h index 13397e97ba..a305bbd605 100644 --- a/src/mesa/main/mfeatures.h +++ b/src/mesa/main/mfeatures.h @@ -53,6 +53,7 @@ #define FEATURE_texture_fxt1 _HAVE_FULL_GL #define FEATURE_texture_s3tc _HAVE_FULL_GL #define FEATURE_userclip _HAVE_FULL_GL +#define FEATURE_vertex_array_byte 0 #define FEATURE_ARB_occlusion_query _HAVE_FULL_GL #define FEATURE_ARB_fragment_program _HAVE_FULL_GL diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c index 9474724c33..220db35855 100644 --- a/src/mesa/main/varray.c +++ b/src/mesa/main/varray.c @@ -125,6 +125,11 @@ _mesa_VertexPointer(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr) case GL_FIXED: elementSize = size * sizeof(GLfixed); break; +#endif +#if FEATURE_vertex_array_byte + case GL_BYTE: + elementSize = size * sizeof(GLbyte); + break; #endif default: _mesa_error( ctx, GL_INVALID_ENUM, "glVertexPointer(type)" ); @@ -424,6 +429,11 @@ _mesa_TexCoordPointer(GLint size, GLenum type, GLsizei stride, case GL_FIXED: elementSize = size * sizeof(GLfixed); break; +#endif +#if FEATURE_vertex_array_byte + case GL_BYTE: + elementSize = size * sizeof(GLbyte); + break; #endif default: _mesa_error( ctx, GL_INVALID_ENUM, "glTexCoordPointer(type)" ); -- cgit v1.2.3 From 677151ad71d7f0792fb79597e972e2cad2cfc7d4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 14:28:59 -0600 Subject: egl: added null ptr check --- src/egl/main/egldisplay.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/main/egldisplay.c b/src/egl/main/egldisplay.c index 540efd4fee..47a2323eaf 100644 --- a/src/egl/main/egldisplay.c +++ b/src/egl/main/egldisplay.c @@ -68,8 +68,9 @@ _EGLDisplay * _eglLookupDisplay(EGLDisplay dpy) { EGLuint key = (EGLuint) dpy; - _EGLDisplay *d = (_EGLDisplay *) _eglHashLookup(_eglGlobal.Displays, key); - return d; + if (!_eglGlobal.Displays) + return NULL; + return (_EGLDisplay *) _eglHashLookup(_eglGlobal.Displays, key); } -- cgit v1.2.3 From 36aae1868345567975ce4fa449b547ae3e01dbc3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 14:29:49 -0600 Subject: mesa: init ctx->RenderMode --- src/mesa/main/context.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index e5264bcaf9..be93d844e0 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -996,6 +996,8 @@ init_attrib_groups(GLcontext *ctx) #endif #if FEATURE_feedback _mesa_init_feedback( ctx ); +#else + ctx->RenderMode = GL_RENDER; #endif _mesa_init_fog( ctx ); #if FEATURE_histogram -- cgit v1.2.3 From 78791d1065c93694a105d4c2cdaee7678a69213a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 14:30:21 -0600 Subject: egl: added cpp assertions/sanity checks --- src/gallium/winsys/egl_xlib/sw_winsys.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c index fd0cb9fb12..28cca9e581 100644 --- a/src/gallium/winsys/egl_xlib/sw_winsys.c +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -194,6 +194,8 @@ surface_alloc_storage(struct pipe_winsys *winsys, surf->pitch = round_up(width, alignment / surf->cpp); surf->usage = flags; + assert(surf->cpp >= 1); + assert(surf->cpp <= 16); assert(!surf->buffer); surf->buffer = winsys->buffer_create(winsys, alignment, PIPE_BUFFER_USAGE_PIXEL, -- cgit v1.2.3 From f38bb109694f2879036c54c97c1c69ea2fecd6c8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 15:58:19 -0600 Subject: gallium: fix some surface usage bugs When a surface is created with GPU_WRITE that really means "GPU render" and that can involve reads (blending). Set surface usage to PIPE_BUFFER_USAGE_CPU_READ + WRITE. Fixes progs/demos/lodbias demo. Also, mark texture as 'modified' when mapped for writing so that the tile cache can know when to freshen a cached tile. Fixes glTexSubImage2D(). --- src/gallium/drivers/softpipe/sp_texture.c | 11 +++++++++-- src/gallium/drivers/softpipe/sp_texture.h | 2 ++ src/gallium/drivers/softpipe/sp_tile_cache.c | 10 ++++++++++ 3 files changed, 21 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 1d7a1fffe4..ef8c5bd6b0 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -207,12 +207,19 @@ softpipe_get_tex_surface(struct pipe_screen *screen, * done with the CPU. Let's adjust the flags to take that into * account. */ - if (ps->usage & PIPE_BUFFER_USAGE_GPU_WRITE) - ps->usage |= PIPE_BUFFER_USAGE_CPU_WRITE; + if (ps->usage & PIPE_BUFFER_USAGE_GPU_WRITE) { + /* GPU_WRITE means "render" and that can involve reads (blending) */ + ps->usage |= PIPE_BUFFER_USAGE_CPU_WRITE | PIPE_BUFFER_USAGE_CPU_READ; + } if (ps->usage & PIPE_BUFFER_USAGE_GPU_READ) ps->usage |= PIPE_BUFFER_USAGE_CPU_READ; + if (ps->usage & (PIPE_BUFFER_USAGE_CPU_WRITE | + PIPE_BUFFER_USAGE_GPU_WRITE)) { + /* Mark the surface as dirty. The tile cache will look for this. */ + spt->modified = TRUE; + } pipe_texture_reference(&ps->texture, pt); ps->face = face; diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h index 779a9d8fc9..0e1017632c 100644 --- a/src/gallium/drivers/softpipe/sp_texture.h +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -47,6 +47,8 @@ struct softpipe_texture /* The data is held here: */ struct pipe_buffer *buffer; + + boolean modified; }; diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 0e4c8c41ee..2d5d2b50f5 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -37,6 +37,7 @@ #include "util/p_tile.h" #include "sp_context.h" #include "sp_surface.h" +#include "sp_texture.h" #include "sp_tile_cache.h" #define NUM_ENTRIES 32 @@ -506,6 +507,15 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, face, level); struct softpipe_cached_tile *tile = tc->entries + pos; + if (tc->texture) { + struct softpipe_texture *spt = softpipe_texture(tc->texture); + if (spt->modified) { + /* texture was modified, force a cache reload */ + tile->x = -1; + spt->modified = FALSE; + } + } + if (tile_x != tile->x || tile_y != tile->y || z != tile->z || -- cgit v1.2.3 From be4259b06cbb2b4c1d8a2dacc19313a30fa909d8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 20 Jun 2008 17:28:42 -0600 Subject: gallium: fix invalid call to draw_set_mapped_constant_buffer() We were indexing sp_constants[i] outside the loop so i was 2. Replace i with PIPE_SHADER_VERTEX. Also, replace magic '2' with PIPE_SHADER_TYPES in a few places. --- src/gallium/drivers/softpipe/sp_context.h | 2 +- src/gallium/drivers/softpipe/sp_draw_arrays.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h index 62eabfb30e..078886f93c 100644 --- a/src/gallium/drivers/softpipe/sp_context.h +++ b/src/gallium/drivers/softpipe/sp_context.h @@ -69,7 +69,7 @@ struct softpipe_context { struct pipe_blend_color blend_color; struct pipe_clip_state clip; - struct pipe_constant_buffer constants[2]; + struct pipe_constant_buffer constants[PIPE_SHADER_TYPES]; struct pipe_framebuffer_state framebuffer; struct pipe_poly_stipple poly_stipple; struct pipe_scissor_state scissor; diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c index d4d5fa744f..12b44a8211 100644 --- a/src/gallium/drivers/softpipe/sp_draw_arrays.c +++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c @@ -47,7 +47,7 @@ softpipe_map_constant_buffers(struct softpipe_context *sp) { struct pipe_winsys *ws = sp->pipe.winsys; uint i; - for (i = 0; i < 2; i++) { + for (i = 0; i < PIPE_SHADER_TYPES; i++) { if (sp->constants[i].size) sp->mapped_constants[i] = ws->buffer_map(ws, sp->constants[i].buffer, PIPE_BUFFER_USAGE_CPU_READ); @@ -55,7 +55,7 @@ softpipe_map_constant_buffers(struct softpipe_context *sp) draw_set_mapped_constant_buffer(sp->draw, sp->mapped_constants[PIPE_SHADER_VERTEX], - sp->constants[i].size); + sp->constants[PIPE_SHADER_VERTEX].size); } static void -- cgit v1.2.3 From 2a5a95d0c08fe07b50385028b6972a4383f2b095 Mon Sep 17 00:00:00 2001 From: Brian Date: Fri, 20 Jun 2008 18:29:23 -0600 Subject: gallium: s/feadback/feedback/, duh --- src/mesa/main/api_exec.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/api_exec.c b/src/mesa/main/api_exec.c index 8ebe4a3e4a..0c3c9c4de4 100644 --- a/src/mesa/main/api_exec.c +++ b/src/mesa/main/api_exec.c @@ -74,7 +74,7 @@ #include "eval.h" #endif #include "get.h" -#if FEATURE_feadback +#if FEATURE_feedback #include "feedback.h" #endif #include "fog.h" @@ -222,7 +222,7 @@ _mesa_init_exec_table(struct _glapi_table *exec) SET_CopyPixels(exec, _mesa_CopyPixels); SET_DrawPixels(exec, _mesa_DrawPixels); #endif -#if FEATURE_feadback +#if FEATURE_feedback SET_InitNames(exec, _mesa_InitNames); SET_FeedbackBuffer(exec, _mesa_FeedbackBuffer); SET_LoadName(exec, _mesa_LoadName); -- cgit v1.2.3 From e2c3f06e9649b5b87fc9adbca7d1f07841bba895 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Sun, 22 Jun 2008 13:14:29 +0100 Subject: draw: fix non-i386 builds --- src/gallium/auxiliary/draw/draw_vs_aos_machine.c | 22 +++++++++++++++++++++- 1 file changed, 21 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c index dcb6c2c8d4..6a54917ae3 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c @@ -28,7 +28,6 @@ #include "pipe/p_config.h" -#ifdef PIPE_ARCH_X86 #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" @@ -39,6 +38,8 @@ #include "draw_vs_aos.h" #include "draw_vertex.h" +#ifdef PIPE_ARCH_X86 + #include "rtasm/rtasm_x86sse.h" @@ -298,6 +299,25 @@ struct aos_machine *draw_vs_aos_machine( void ) return machine; } +#else + +void draw_vs_aos_machine_viewport( struct aos_machine *machine, + const struct pipe_viewport_state *viewport ) +{ +} + +void draw_vs_aos_machine_constants( struct aos_machine *machine, + const float (*constants)[4] ) +{ +} + +void draw_vs_aos_machine_destroy( struct aos_machine *machine ) +{ +} +struct aos_machine *draw_vs_aos_machine( void ) +{ + return NULL; +} #endif -- cgit v1.2.3 From d40ff294510236faff4bb9a58b81282705b98562 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 19 Jun 2008 20:28:20 +0200 Subject: i915: Removed level_offset from i915_texture All offsets are now on image_offset. --- src/gallium/drivers/i915simple/i915_context.h | 4 ---- src/gallium/drivers/i915simple/i915_texture.c | 27 ++++++++++++--------------- 2 files changed, 12 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 892a88fd2c..9e02f78714 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -202,10 +202,6 @@ struct i915_texture { */ unsigned *image_offset[PIPE_MAX_TEXTURE_LEVELS]; /**< array [depth] of offsets */ - /* Includes image offset tables: - */ - unsigned level_offset[PIPE_MAX_TEXTURE_LEVELS]; - /* The data is held here: */ struct pipe_buffer *buffer; diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 9cd32e3919..d9b33df1d5 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -100,7 +100,7 @@ static void i915_miptree_set_level_info(struct i915_texture *tex, unsigned level, unsigned nr_images, - unsigned x, unsigned y, unsigned w, unsigned h, unsigned d) + unsigned w, unsigned h, unsigned d) { struct pipe_texture *pt = &tex->base; @@ -110,7 +110,6 @@ i915_miptree_set_level_info(struct i915_texture *tex, pt->height[level] = h; pt->depth[level] = d; - tex->level_offset[level] = (x + y * tex->pitch) * pt->cpp; tex->nr_images[level] = nr_images; /* @@ -166,7 +165,7 @@ i915_displaytarget_layout(struct i915_texture *tex) if (pt->last_level > 0 || pt->cpp != 4) return 0; - i915_miptree_set_level_info( tex, 0, 1, 0, 0, + i915_miptree_set_level_info( tex, 0, 1, tex->base.width[0], tex->base.height[0], 1 ); @@ -230,7 +229,8 @@ i945_miptree_layout_2d( struct i915_texture *tex ) for (level = 0; level <= pt->last_level; level++) { unsigned img_height; - i915_miptree_set_level_info(tex, level, 1, x, y, width, height, 1); + i915_miptree_set_level_info(tex, level, 1, width, height, 1); + i915_miptree_set_image_offset(tex, level, 0, x, y); if (pt->compressed) img_height = MAX2(1, height/4); @@ -294,7 +294,7 @@ i945_miptree_layout_cube(struct i915_texture *tex) /* Set all the levels to effectively occupy the whole rectangular region. */ for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_level_info(tex, level, 6, 0, 0, lvlWidth, lvlHeight, 1); + i915_miptree_set_level_info(tex, level, 6, lvlWidth, lvlHeight, 1); lvlWidth /= 2; lvlHeight /= 2; } @@ -380,8 +380,6 @@ i915_miptree_layout(struct i915_texture * tex) for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_level_info(tex, level, 6, - 0, 0, - /*OLD: tex->pitch, tex->total_height,*/ lvlWidth, lvlHeight, 1); lvlWidth /= 2; @@ -416,7 +414,7 @@ i915_miptree_layout(struct i915_texture * tex) */ for (level = 0; level <= MAX2(8, pt->last_level); level++) { - i915_miptree_set_level_info(tex, level, depth, 0, tex->total_height, + i915_miptree_set_level_info(tex, level, depth, width, height, depth); @@ -458,8 +456,9 @@ i915_miptree_layout(struct i915_texture * tex) for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_level_info(tex, level, 1, - 0, tex->total_height, width, height, 1); + i915_miptree_set_image_offset(tex, level, 0, + 0, tex->total_height); if (pt->compressed) img_height = MAX2(1, height / 4); @@ -515,12 +514,11 @@ i945_miptree_layout(struct i915_texture * tex) unsigned q, j; i915_miptree_set_level_info(tex, level, nr_images, - 0, tex->total_height, width, height, depth); for (q = 0; q < nr_images;) { for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - i915_miptree_set_image_offset(tex, level, q, x, y); + i915_miptree_set_image_offset(tex, level, q, x, y + tex->total_height); x += pack_x_pitch; } @@ -648,15 +646,14 @@ i915_get_tex_surface(struct pipe_screen *screen, struct pipe_surface *ps; unsigned offset; /* in bytes */ - offset = tex->level_offset[level]; - if (pt->target == PIPE_TEXTURE_CUBE) { - offset += tex->image_offset[level][face] * pt->cpp; + offset = tex->image_offset[level][face] * pt->cpp; } else if (pt->target == PIPE_TEXTURE_3D) { - offset += tex->image_offset[level][zslice] * pt->cpp; + offset = tex->image_offset[level][zslice] * pt->cpp; } else { + offset = tex->image_offset[level][0] * pt->cpp; assert(face == 0); assert(zslice == 0); } -- cgit v1.2.3 From f08da6b8214f1bf1d4a33e19dac4eeb155302481 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 23 Jun 2008 12:31:46 +0200 Subject: gallium: Fix warning in u_draw_quad.h --- src/gallium/auxiliary/util/u_draw_quad.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_draw_quad.h b/src/gallium/auxiliary/util/u_draw_quad.h index 5b6539a99c..ec4862ead3 100644 --- a/src/gallium/auxiliary/util/u_draw_quad.h +++ b/src/gallium/auxiliary/util/u_draw_quad.h @@ -33,6 +33,7 @@ extern "C" { #endif +struct pipe_buffer; extern void util_draw_vertex_buffer(struct pipe_context *pipe, -- cgit v1.2.3 From 52501801c7107614387717e7b97e66a667f9e5b8 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 23 Jun 2008 15:54:25 +0200 Subject: gallium: Add accessor functions to get textures from a st_framebuffer --- src/mesa/state_tracker/st_framebuffer.c | 16 ++++++++++++++++ src/mesa/state_tracker/st_public.h | 3 +++ 2 files changed, 19 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 7e6db46757..1b6e68c2a1 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -230,6 +230,22 @@ st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex) return NULL; } +struct pipe_texture * +st_get_framebuffer_texture(struct st_framebuffer *stfb, uint surfIndex) +{ + struct st_renderbuffer *strb; + + assert(surfIndex <= ST_SURFACE_DEPTH); + + /* sanity checks, ST tokens should match Mesa tokens */ + assert(ST_SURFACE_FRONT_LEFT == BUFFER_FRONT_LEFT); + assert(ST_SURFACE_BACK_RIGHT == BUFFER_BACK_RIGHT); + + strb = st_renderbuffer(stfb->Base.Attachment[surfIndex].Renderbuffer); + if (strb) + return strb->texture; + return NULL; +} /** * This function is to be called prior to SwapBuffers on the given diff --git a/src/mesa/state_tracker/st_public.h b/src/mesa/state_tracker/st_public.h index b99984215f..ca4e9577b1 100644 --- a/src/mesa/state_tracker/st_public.h +++ b/src/mesa/state_tracker/st_public.h @@ -74,6 +74,9 @@ void st_set_framebuffer_surface(struct st_framebuffer *stfb, struct pipe_surface *st_get_framebuffer_surface(struct st_framebuffer *stfb, uint surfIndex); +struct pipe_texture *st_get_framebuffer_texture(struct st_framebuffer *stfb, + uint surfIndex); + void *st_framebuffer_private( struct st_framebuffer *stfb ); void st_unreference_framebuffer( struct st_framebuffer **stfb ); -- cgit v1.2.3 From 708bb35194e16bf7aaf5cd676f572b7200a156d1 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 23 Jun 2008 16:01:41 +0200 Subject: util: Blit can now copy from texture to surface --- src/gallium/auxiliary/util/u_blit.c | 101 ++++++++++++++++++++++++++++++++++++ src/gallium/auxiliary/util/u_blit.h | 14 +++-- 2 files changed, 112 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 7b9415d49a..28513f54e6 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -362,3 +362,104 @@ util_blit_pixels(struct blit_state *ctx, screen->texture_release(screen, &tex); } +/** + * Copy pixel block from src texture to dst surface. + * Overlapping regions are acceptable. + * + * XXX Should support selection of level. + * XXX need some control over blitting Z and/or stencil. + */ +void +util_blit_pixels_tex(struct blit_state *ctx, + struct pipe_texture *tex, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter) +{ + struct pipe_context *pipe = ctx->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_framebuffer_state fb; + const int srcLeft = MIN2(srcX0, srcX1); + const int srcTop = MIN2(srcY0, srcY1); + + assert(filter == PIPE_TEX_MIPFILTER_NEAREST || + filter == PIPE_TEX_MIPFILTER_LINEAR); + + if (srcLeft != srcX0) { + /* left-right flip */ + int tmp = dstX0; + dstX0 = dstX1; + dstX1 = tmp; + } + + if (srcTop != srcY0) { + /* up-down flip */ + int tmp = dstY0; + dstY0 = dstY1; + dstY1 = tmp; + } + + assert(screen->is_format_supported(screen, dst->format, PIPE_SURFACE)); + + /* save state (restored below) */ + cso_save_blend(ctx->cso); + cso_save_depth_stencil_alpha(ctx->cso); + cso_save_rasterizer(ctx->cso); + cso_save_samplers(ctx->cso); + cso_save_sampler_textures(ctx->cso); + cso_save_framebuffer(ctx->cso); + cso_save_fragment_shader(ctx->cso); + cso_save_vertex_shader(ctx->cso); + cso_save_viewport(ctx->cso); + + /* set misc state we care about */ + cso_set_blend(ctx->cso, &ctx->blend); + cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil); + cso_set_rasterizer(ctx->cso, &ctx->rasterizer); + cso_set_viewport(ctx->cso, &ctx->viewport); + + /* sampler */ + ctx->sampler.min_img_filter = filter; + ctx->sampler.mag_img_filter = filter; + cso_single_sampler(ctx->cso, 0, &ctx->sampler); + cso_single_sampler_done(ctx->cso); + + /* texture */ + cso_set_sampler_textures(ctx->cso, 1, &tex); + + /* shaders */ + cso_set_fragment_shader_handle(ctx->cso, ctx->fs); + cso_set_vertex_shader_handle(ctx->cso, ctx->vs); + + /* drawing dest */ + memset(&fb, 0, sizeof(fb)); + fb.width = dst->width; + fb.height = dst->height; + fb.num_cbufs = 1; + fb.cbufs[0] = dst; + cso_set_framebuffer(ctx->cso, &fb); + + /* draw quad */ + setup_vertex_data(ctx, + (float) dstX0, (float) dstY0, + (float) dstX1, (float) dstY1, z); + + util_draw_vertex_buffer(ctx->pipe, ctx->vbuf, + PIPE_PRIM_TRIANGLE_FAN, + 4, /* verts */ + 2); /* attribs/vert */ + + /* restore state we changed */ + cso_restore_blend(ctx->cso); + cso_restore_depth_stencil_alpha(ctx->cso); + cso_restore_rasterizer(ctx->cso); + cso_restore_samplers(ctx->cso); + cso_restore_sampler_textures(ctx->cso); + cso_restore_framebuffer(ctx->cso); + cso_restore_fragment_shader(ctx->cso); + cso_restore_vertex_shader(ctx->cso); + cso_restore_viewport(ctx->cso); +} diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h index 0ce9732e62..7d0406c104 100644 --- a/src/gallium/auxiliary/util/u_blit.h +++ b/src/gallium/auxiliary/util/u_blit.h @@ -42,15 +42,13 @@ struct cso_context; struct blit_state; + extern struct blit_state * util_create_blit(struct pipe_context *pipe, struct cso_context *cso); - extern void util_destroy_blit(struct blit_state *ctx); - - extern void util_blit_pixels(struct blit_state *ctx, struct pipe_surface *src, @@ -61,6 +59,16 @@ util_blit_pixels(struct blit_state *ctx, int dstX1, int dstY1, float z, uint filter); +extern void +util_blit_pixels_tex(struct blit_state *ctx, + struct pipe_texture *tex, + int srcX0, int srcY0, + int srcX1, int srcY1, + struct pipe_surface *dst, + int dstX0, int dstY0, + int dstX1, int dstY1, + float z, uint filter); + #ifdef __cplusplus } -- cgit v1.2.3 From 464c585681a647542e4ae8188dc2883c0c8a142e Mon Sep 17 00:00:00 2001 From: Dan Nicholson Date: Sun, 22 Jun 2008 20:29:45 -0700 Subject: Ensure all objects are built when installing DRI --- src/mesa/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 2d0a6e0167..4e0cea240a 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -120,7 +120,7 @@ install-osmesa: default $(INSTALL) $(TOP)/$(LIB_DIR)/$(OSMESA_LIB_NAME)* \ $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) -install-dri: +install-dri: default cd drivers/dri && $(MAKE) install -- cgit v1.2.3 From 8319f2f326f74a93b51eb3db1e9f619a2a74e14e Mon Sep 17 00:00:00 2001 From: Dan Nicholson Date: Sun, 22 Jun 2008 20:19:35 -0700 Subject: Create $(TOP)/$(LIB_DIR) for install, too If `make install' is run without running `make' first, the $(LIB_DIR) will not be created. This also changes the mkdir a little bit so that it isn't run if necessary and added `-p' so that it is immune to races. --- src/Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/Makefile b/src/Makefile index f93987e688..fab0069a71 100644 --- a/src/Makefile +++ b/src/Makefile @@ -21,7 +21,7 @@ subdirs: fi \ done -install: +install: $(TOP)/$(LIB_DIR) @for dir in $(SUBDIRS) ; do \ if [ -d $$dir ] ; then \ (cd $$dir && $(MAKE) install) || exit 1 ; \ @@ -29,7 +29,7 @@ install: done $(TOP)/$(LIB_DIR): - -mkdir $(TOP)/$(LIB_DIR) + -@test -d $(TOP)/$(LIB_DIR) || mkdir -p $(TOP)/$(LIB_DIR) clean: -- cgit v1.2.3 From 76b94a636eea0ffb642a90cf9ff4a00bfe6e064e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 08:14:14 -0600 Subject: Don't make libmesa.a or libglapi.a depend on asm_subdirs Since the asm_subdirs target does not actually create a file, make will always consider that it needs to be rebuilt. If libmesa.a and libglapi.a have asm_subdirs as a prerequisite, then they will always need to be rebuilt, too. The correct order will be preserved by the default target, though. This should fix #16358. Conflicts: src/mesa/Makefile cherry-picked from master --- src/mesa/Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index 4e0cea240a..cf75baa706 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -32,11 +32,11 @@ default: depend asm_subdirs libmesa.a $(GLAPI_LIB) driver_subdirs # Helper libraries used by many drivers: # Make archive of core mesa object files -libmesa.a: $(MESA_OBJECTS) asm_subdirs +libmesa.a: $(MESA_OBJECTS) @ $(TOP)/bin/mklib -o mesa -static $(MESA_OBJECTS) # Make archive of gl* API dispatcher functions only -$(GLAPI_LIB): $(GLAPI_OBJECTS) asm_subdirs +$(GLAPI_LIB): $(GLAPI_OBJECTS) @if [ "${WINDOW_SYSTEM}" = "dri" ] ; then \ touch libglapi.a ; \ else \ -- cgit v1.2.3 From 016dbb0cf395702cfad046f827e3cc4541ae5818 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 08:35:41 -0600 Subject: gallium: added some assertions --- src/gallium/auxiliary/util/p_util.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_util.c b/src/gallium/auxiliary/util/p_util.c index 2a92f8e408..0e3f082b3e 100644 --- a/src/gallium/auxiliary/util/p_util.c +++ b/src/gallium/auxiliary/util/p_util.c @@ -53,6 +53,12 @@ pipe_copy_rect(ubyte * dst, { unsigned i; + assert(cpp > 0); + assert(src_x >= 0); + assert(src_y >= 0); + assert(dst_x >= 0); + assert(dst_y >= 0); + dst_pitch *= cpp; src_pitch *= cpp; dst += dst_x * cpp; -- cgit v1.2.3 From 92d48a4cb9481adc7abd7cdf9550fbf5a9c9613d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 08:44:30 -0600 Subject: gallium: fix bad srcy coord if do_flip --- src/gallium/drivers/softpipe/sp_surface.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index 29a1e92416..acedebfcc5 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -35,7 +35,11 @@ -/* Assumes all values are within bounds -- no checking at this level - +/** + * Copy a rectangular region from one surface to another. + * Surfaces must have same bpp. + * + * Assumes all values are within bounds -- no checking at this level - * do it higher up if required. */ static void @@ -54,9 +58,11 @@ sp_surface_copy(struct pipe_context *pipe, src, PIPE_BUFFER_USAGE_CPU_READ ); - assert( dst->cpp == src->cpp ); - assert(src_map && dst_map); + assert(dst->cpp == src->cpp); + assert(src_map); + assert(dst_map); + /* If do_flip, invert src_y position and pass negative src stride */ pipe_copy_rect(dst_map, dst->cpp, dst->pitch, @@ -64,7 +70,7 @@ sp_surface_copy(struct pipe_context *pipe, width, height, src_map, do_flip ? -(int) src->pitch : src->pitch, - srcx, do_flip ? 1 - srcy - height : srcy); + srcx, do_flip ? src->height - 1 - srcy : srcy); pipe->screen->surface_unmap(pipe->screen, src); pipe->screen->surface_unmap(pipe->screen, dst); -- cgit v1.2.3 From f1601c2b75ca4c2223079f676cf796843b284df2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 08:51:00 -0600 Subject: gallium: fix bad srcy coord if do_flip --- src/gallium/drivers/cell/ppu/cell_surface.c | 2 +- src/gallium/drivers/i915simple/i915_surface.c | 2 +- src/gallium/drivers/i965simple/brw_surface.c | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c index a35db0ef99..4a098101e1 100644 --- a/src/gallium/drivers/cell/ppu/cell_surface.c +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -76,7 +76,7 @@ cell_surface_copy(struct pipe_context *pipe, width, height, pipe_surface_map(src), do_flip ? -src->pitch : src->pitch, - srcx, do_flip ? 1 - srcy - height : srcy); + srcx, do_flip ? height - 1 - srcy : srcy); pipe_surface_unmap(src); pipe_surface_unmap(dst); diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index 98367ac073..1729ea26f5 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -66,7 +66,7 @@ i915_surface_copy(struct pipe_context *pipe, width, height, src_map, do_flip ? -(int) src->pitch : src->pitch, - srcx, do_flip ? 1 - srcy - height : srcy); + srcx, do_flip ? height - 1 - srcy : srcy); pipe->screen->surface_unmap(pipe->screen, src); pipe->screen->surface_unmap(pipe->screen, dst); diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index 3e3736b280..62d75bc251 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -66,7 +66,7 @@ brw_surface_copy(struct pipe_context *pipe, width, height, src_map, do_flip ? -(int) src->pitch : src->pitch, - srcx, do_flip ? 1 - srcy - height : srcy); + srcx, do_flip ? height - 1 - srcy : srcy); pipe->screen->surface_unmap(pipe->screen, src); pipe->screen->surface_unmap(pipe->screen, dst); -- cgit v1.2.3 From e14126ec811e4f37cf085be27cac4f750d9e011a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 08:54:16 -0600 Subject: gallium: change surface_copy()'s do_flip to boolean --- src/gallium/drivers/cell/ppu/cell_surface.c | 2 +- src/gallium/drivers/i915simple/i915_surface.c | 2 +- src/gallium/drivers/i965simple/brw_surface.c | 2 +- src/gallium/drivers/softpipe/sp_surface.c | 2 +- src/gallium/include/pipe/p_context.h | 2 +- 5 files changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c index 4a098101e1..18f3791924 100644 --- a/src/gallium/drivers/cell/ppu/cell_surface.c +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -60,7 +60,7 @@ cell_surface_data(struct pipe_context *pipe, static void cell_surface_copy(struct pipe_context *pipe, - unsigned do_flip, + boolean do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index 1729ea26f5..cc55a0910e 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -41,7 +41,7 @@ */ static void i915_surface_copy(struct pipe_context *pipe, - unsigned do_flip, + boolean do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index 62d75bc251..3d98a2bf19 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -41,7 +41,7 @@ */ static void brw_surface_copy(struct pipe_context *pipe, - unsigned do_flip, + boolean do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index acedebfcc5..9fd48aeccc 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -44,7 +44,7 @@ */ static void sp_surface_copy(struct pipe_context *pipe, - unsigned do_flip, + boolean do_flip, struct pipe_surface *dst, unsigned dstx, unsigned dsty, struct pipe_surface *src, diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h index faf112c6d6..2646706ff2 100644 --- a/src/gallium/include/pipe/p_context.h +++ b/src/gallium/include/pipe/p_context.h @@ -192,7 +192,7 @@ struct pipe_context { */ /*@{*/ void (*surface_copy)(struct pipe_context *pipe, - unsigned do_flip, /*<< flip surface contents vertically */ + boolean do_flip,/**< flip surface contents vertically */ struct pipe_surface *dest, unsigned destx, unsigned desty, struct pipe_surface *src, /* don't make this const - -- cgit v1.2.3 From 81b1a4224de1992d25ada006f54ff9147aa82da3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 09:07:56 -0600 Subject: gallium: remove dead/prototype code --- src/mesa/state_tracker/st_cb_texture.c | 22 ---------------------- 1 file changed, 22 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index ed65f3b0aa..40d56c5edd 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1188,27 +1188,6 @@ do_copy_texsubimage(GLcontext *ctx, if (ctx->_ImageTransferState == 0x0) { /* do blit-style copy */ - - /* XXX may need to invert image depending on window - * vs. user-created FBO - */ - -#if 0 - /* A bit of fiddling to get the blitter to work with -ve - * pitches. But we get a nice inverted blit this way, so it's - * worth it: - */ - intelEmitCopyBlit(intel, - stImage->pt->cpp, - -src->pitch, - src->buffer, - src->height * src->pitch * src->cpp, - stImage->pt->pitch, - stImage->pt->region->buffer, - dest_offset, - x, y + height, dstx, dsty, width, height, - GL_COPY); /* ? */ -#else struct pipe_surface *dest_surface; dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, @@ -1246,7 +1225,6 @@ do_copy_texsubimage(GLcontext *ctx, } pipe_surface_reference(&dest_surface, NULL); -#endif } if (use_fallback) { -- cgit v1.2.3 From f738c3acaca235c68a26c7b7d41903c64a36ae9f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 09:47:12 -0600 Subject: gallium: fix Y-inverted copies Don't require the caller to pass a non-intuitive negative src_y coord anymore when doing a src-inverted copy. --- src/gallium/auxiliary/util/p_util.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_util.c b/src/gallium/auxiliary/util/p_util.c index 0e3f082b3e..4e60b1b841 100644 --- a/src/gallium/auxiliary/util/p_util.c +++ b/src/gallium/auxiliary/util/p_util.c @@ -37,6 +37,7 @@ /** * Copy 2D rect from one place to another. * Position and sizes are in pixels. + * src_pitch may be negative to do vertical flip of pixels from source. */ void pipe_copy_rect(ubyte * dst, @@ -52,6 +53,7 @@ pipe_copy_rect(ubyte * dst, int src_y) { unsigned i; + int src_pitch_pos = src_pitch < 0 ? -src_pitch : src_pitch; assert(cpp > 0); assert(src_x >= 0); @@ -61,10 +63,11 @@ pipe_copy_rect(ubyte * dst, dst_pitch *= cpp; src_pitch *= cpp; + src_pitch_pos *= cpp; dst += dst_x * cpp; src += src_x * cpp; dst += dst_y * dst_pitch; - src += src_y * src_pitch; + src += src_y * src_pitch_pos; width *= cpp; if (width == dst_pitch && width == src_pitch) -- cgit v1.2.3 From a479bf62353b6517841e620122112e7565d2391c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 09:50:09 -0600 Subject: gallium: fix-up inverted do_copy_texsubimage() The logic/arithmetic for inverting the src image is a bit simpler now. --- src/mesa/state_tracker/st_cb_texture.c | 51 ++++++++++++++++++++-------------- 1 file changed, 30 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 40d56c5edd..da45a5e321 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1157,11 +1157,12 @@ do_copy_texsubimage(GLcontext *ctx, struct pipe_context *pipe = ctx->st->pipe; struct pipe_screen *screen = pipe->screen; uint dest_format, src_format; - uint do_flip = FALSE; GLboolean use_fallback = GL_TRUE; (void) texImage; + /* XX need this? st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);*/ + /* determine if copying depth or color data */ if (baseFormat == GL_DEPTH_COMPONENT) { strb = st_renderbuffer(fb->_DepthBuffer); @@ -1178,11 +1179,6 @@ do_copy_texsubimage(GLcontext *ctx, assert(strb->surface); assert(stImage->pt); - if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { - srcY = strb->Base.Height - srcY - height; - do_flip = TRUE; - } - src_format = strb->surface->format; dest_format = stImage->pt->format; @@ -1190,7 +1186,8 @@ do_copy_texsubimage(GLcontext *ctx, /* do blit-style copy */ struct pipe_surface *dest_surface; - dest_surface = screen->get_tex_surface(screen, stImage->pt, stImage->face, + dest_surface = screen->get_tex_surface(screen, stImage->pt, + stImage->face, stImage->level, destZ, PIPE_BUFFER_USAGE_GPU_WRITE); @@ -1198,28 +1195,40 @@ do_copy_texsubimage(GLcontext *ctx, assert(dest_surface->buffer); if (src_format == dest_format) { - pipe->surface_copy(pipe, - do_flip, - /* dest */ - dest_surface, - destX, destY, - /* src */ - strb->surface, - srcX, srcY, - /* size */ - width, height); - use_fallback = GL_FALSE; + boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP); + pipe->surface_copy(pipe, + do_flip, + /* dest */ + dest_surface, + destX, destY, + /* src */ + strb->surface, + srcX, srcY, + /* size */ + width, height); + use_fallback = GL_FALSE; } else if (screen->is_format_supported(screen, strb->surface->format, PIPE_TEXTURE) && screen->is_format_supported(screen, dest_surface->format, PIPE_SURFACE)) { + boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP); + int srcY0, srcY1; + if (do_flip) { + srcY1 = strb->Base.Height - srcY - height; + srcY0 = srcY1 + height; + } + else { + srcY0 = srcY; + srcY1 = srcY0 + height; + } util_blit_pixels(ctx->st->blit, strb->surface, - srcX, do_flip ? srcY + height : srcY, - srcX + width, do_flip ? srcY : srcY + height, + srcX, srcY0, + srcX + width, srcY1, dest_surface, - destX, destY, destX + width, destY + height, + destX, destY, + destX + width, destY + height, 0.0, PIPE_TEX_MIPFILTER_NEAREST); use_fallback = GL_FALSE; } -- cgit v1.2.3 From f52ab4cc22bfb6708724f3c3966ce734d605cddd Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 23 Jun 2008 17:57:45 +0200 Subject: i915: Add render and texture support for tiled texture and buffers This is step towards tiled textures and buffer support for i915. But the tiled attribute is never set. --- src/gallium/drivers/i915simple/i915_context.h | 2 ++ src/gallium/drivers/i915simple/i915_state_emit.c | 12 ++++++------ src/gallium/drivers/i915simple/i915_state_sampler.c | 8 +++++++- src/gallium/drivers/i915simple/i915_texture.c | 1 + 4 files changed, 16 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 9e02f78714..5d411a6648 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -192,6 +192,8 @@ struct i915_texture { unsigned depth_pitch; /* per-image on i945? */ unsigned total_height; + unsigned tiled; + unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; /* Explicitly store the offset of each image for each cube face or diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index bc801a82f0..19d968fd8b 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -213,10 +213,10 @@ i915_emit_hardware_state(struct i915_context *i915 ) if (cbuf_surface) { unsigned cpitch = (cbuf_surface->pitch * cbuf_surface->cpp); unsigned ctile = BUF_3D_USE_FENCE; -#if 0 - if (!((cpitch - 1) & cpitch) && cpitch >= 512) + if (cbuf_surface->texture && + ((struct i915_texture*)(cbuf_surface->texture))->tiled) { ctile = BUF_3D_TILED_SURFACE; -#endif + } OUT_BATCH(_3DSTATE_BUF_INFO_CMD); @@ -234,10 +234,10 @@ i915_emit_hardware_state(struct i915_context *i915 ) if (depth_surface) { unsigned zpitch = (depth_surface->pitch * depth_surface->cpp); unsigned ztile = BUF_3D_USE_FENCE; -#if 0 - if (!((zpitch - 1) & zpitch) && zpitch >= 512) + if (depth_surface->texture && + ((struct i915_texture*)(depth_surface->texture))->tiled) { ztile = BUF_3D_TILED_SURFACE; -#endif + } OUT_BATCH(_3DSTATE_BUF_INFO_CMD); diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index 24440843f3..379aff3846 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -234,6 +234,7 @@ i915_update_texture(struct i915_context *i915, const uint width = pt->width[0], height = pt->height[0], depth = pt->depth[0]; const uint num_levels = pt->last_level; unsigned max_lod = num_levels * 4; + unsigned tiled = MS3_USE_FENCE_REGS; assert(tex); assert(width); @@ -246,12 +247,17 @@ i915_update_texture(struct i915_context *i915, assert(format); assert(pitch); + if (tex->tiled) { + assert(!((pitch - 1) & pitch)); + tiled = MS3_TILED_SURFACE; + } + /* MS3 state */ state[0] = (((height - 1) << MS3_HEIGHT_SHIFT) | ((width - 1) << MS3_WIDTH_SHIFT) | format - | MS3_USE_FENCE_REGS); + | tiled); /* * XXX When min_filter != mag_filter and there's just one mipmap level, diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index d9b33df1d5..ae107c6676 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -174,6 +174,7 @@ i915_displaytarget_layout(struct i915_texture *tex) if (tex->base.width[0] >= 128) { tex->pitch = power_of_two(tex->base.width[0] * pt->cpp) / pt->cpp; tex->total_height = round_up(tex->base.height[0], 8); + tex->tiled = 1; } else { tex->pitch = round_up(tex->base.width[0], 64 / pt->cpp); tex->total_height = tex->base.height[0]; -- cgit v1.2.3 From 25da42a650048cd960c81af56744e5fdadd923ad Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 11:27:44 -0600 Subject: gallium: in softpipe_get_tex_surface() use the pitch specified in the softpipe_texture object. Fixes a pitch/width mix-up. --- src/gallium/drivers/softpipe/sp_texture.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index ef8c5bd6b0..2ef17a220b 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -198,7 +198,7 @@ softpipe_get_tex_surface(struct pipe_screen *screen, ps->cpp = pt->cpp; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->pitch = ps->width; + ps->pitch = spt->pitch[level]; ps->offset = spt->level_offset[level]; ps->usage = usage; -- cgit v1.2.3 From 72b6fddefb15b98499eda422a30a1e92da4c2850 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 01:25:20 +0900 Subject: gallium: WinCE does not have cosf, sinf, etc. --- src/gallium/include/pipe/p_util.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 8b3003bcef..f62faf616a 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -409,7 +409,7 @@ extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, #if defined(_MSC_VER) -#if _MSC_VER < 1400 && !defined(__cplusplus) +#if _MSC_VER < 1400 && !defined(__cplusplus) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) static INLINE float cosf( float f ) { -- cgit v1.2.3 From d9f38a2ad1477cc6aeb94c083ab439b8c704be2e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 01:26:22 +0900 Subject: gallium: Use the more portable PIPE_ARCH_* PIPE_CC_* macros instead of ad-hoc ones. --- src/gallium/auxiliary/util/p_debug.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 9d56f8bfab..b1106a5e55 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -100,9 +100,9 @@ void debug_print_blob( const char *name, void _debug_break(void) { -#if (defined(__i386__) || defined(__386__)) && defined(__GNUC__) +#if defined(PIPE_ARCH_X86) && defined(PIPE_CC_GCC) __asm("int3"); -#elif (defined(__i386__) || defined(__386__)) && defined(__MSC__) +#elif defined(PIPE_ARCH_X86) && defined(PIPE_CC_MSVC) _asm {int 3}; #elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) EngDebugBreak(); -- cgit v1.2.3 From ab7ad60d47fdd6fbbbbdb8968676083356e8692d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 02:35:50 +0900 Subject: mesa: No getenv on WinCE. --- src/mesa/glapi/glapi.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/glapi/glapi.c b/src/mesa/glapi/glapi.c index ed59ddb0b7..c236656d9a 100644 --- a/src/mesa/glapi/glapi.c +++ b/src/mesa/glapi/glapi.c @@ -82,6 +82,7 @@ _glapi_set_warning_func( _glapi_warning_func func ) static GLboolean warn(void) { +#if !defined(_WIN32_WCE) if ((WarnFlag || getenv("MESA_DEBUG") || getenv("LIBGL_DEBUG")) && warning_func) { return GL_TRUE; @@ -89,6 +90,9 @@ warn(void) else { return GL_FALSE; } +#else + return GL_FALSE; +#endif } -- cgit v1.2.3 From 8db7ef544c4f1be702e34b97882441df31274f10 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 02:37:21 +0900 Subject: mesa: ASSERT macro is already defined by WinCE headers. Even when just the standard headers are used.... --- src/mesa/main/glheader.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/main/glheader.h b/src/mesa/main/glheader.h index 0f74bc83cc..9e39a8370e 100644 --- a/src/mesa/main/glheader.h +++ b/src/mesa/main/glheader.h @@ -233,6 +233,7 @@ #endif +#if !defined(_WIN32_WCE) #if defined(BUILD_FOR_SNAP) && defined(CHECKED) # define ASSERT(X) _CHECK(X) #elif defined(DEBUG) @@ -240,6 +241,7 @@ #else # define ASSERT(X) #endif +#endif #if !defined __GNUC__ || __GNUC__ < 3 -- cgit v1.2.3 From dc73d15a9a1cc830781f8f3ef81507bfff79b7f9 Mon Sep 17 00:00:00 2001 From: Jonathan White Date: Mon, 23 Jun 2008 16:24:58 -0600 Subject: gallium: code for PIPE_SUBSYSTEM_WINDOWS_USER --- src/gallium/auxiliary/util/p_debug.c | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index b1106a5e55..7b28900a25 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -30,14 +30,28 @@ #include + #ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY + #include #include + +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) + +#ifndef WIN32_LEAN_AND_MEAN +#define WIN32_LEAN_AND_MEAN // Exclude rarely-used stuff from Windows headers +#endif +#include + #else + #include #include + #endif + + #include "pipe/p_compiler.h" #include "pipe/p_util.h" #include "pipe/p_debug.h" @@ -74,6 +88,17 @@ void _debug_vprintf(const char *format, va_list ap) #else /* TODO: Implement debug print for WINCE */ #endif +#elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) + /* EngDebugPrint does not handle float point arguments, so we need to use + * our own vsnprintf implementation. It is also very slow, so buffer until + * we find a newline. */ + static char buf[512 + 1] = {'\0'}; + size_t len = strlen(buf); + int ret = util_vsnprintf(buf + len, sizeof(buf) - len, format, ap); + if(ret > (int)(sizeof(buf) - len - 1) || util_strchr(buf + len, '\n')) { + OutputDebugStringA(buf); + buf[0] = '\0'; + } #else vfprintf(stderr, format, ap); #endif -- cgit v1.2.3 From 89e9d6b6db933c870443714c3d7c9539d117cddf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Jun 2008 17:13:14 -0600 Subject: gallium: added support for fixed-point formats, drawing --- src/gallium/auxiliary/translate/translate_generic.c | 18 ++++++++++++++++++ src/gallium/include/pipe/p_format.h | 5 +++++ src/mesa/state_tracker/st_cb_texture.c | 3 ++- src/mesa/state_tracker/st_draw.c | 14 ++++++++++++-- 4 files changed, 37 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index 3fec89b36e..17e37d1745 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -121,6 +121,8 @@ emit_##NAME(const float *attrib, void *ptr) \ #define FROM_16_SNORM(i) ((float) ((short *) ptr)[i] / 32767.0f) #define FROM_32_SNORM(i) ((float) ((int *) ptr)[i] / 2147483647.0f) +#define FROM_32_FIXED(i) (((int *) ptr)[i] / 65536.0) + #define TO_64_FLOAT(x) ((double) x) #define TO_32_FLOAT(x) (x) @@ -140,6 +142,8 @@ emit_##NAME(const float *attrib, void *ptr) \ #define TO_16_SNORM(x) ((short) (x * 32767.0f)) #define TO_32_SNORM(x) ((int) (x * 2147483647.0f)) +#define TO_32_FIXED(x) ((int) (x * 65536.0f)) + ATTRIB( R64G64B64A64_FLOAT, 4, double, FROM_64_FLOAT, TO_64_FLOAT ) @@ -215,6 +219,11 @@ ATTRIB( R8_SNORM, 1, char, FROM_8_SNORM, TO_8_SNORM ) ATTRIB( A8R8G8B8_UNORM, 4, ubyte, FROM_8_UNORM, TO_8_UNORM ) //ATTRIB( R8G8B8A8_UNORM, 4, ubyte, FROM_8_UNORM, TO_8_UNORM ) +ATTRIB( R32G32B32A32_FIXED, 4, int, FROM_32_FIXED, TO_32_FIXED ) +ATTRIB( R32G32B32_FIXED, 3, int, FROM_32_FIXED, TO_32_FIXED ) +ATTRIB( R32G32_FIXED, 2, int, FROM_32_FIXED, TO_32_FIXED ) +ATTRIB( R32_FIXED, 1, int, FROM_32_FIXED, TO_32_FIXED ) + static void @@ -386,6 +395,15 @@ static fetch_func get_fetch_func( enum pipe_format format ) case PIPE_FORMAT_B8G8R8A8_UNORM: return &fetch_B8G8R8A8_UNORM; + case PIPE_FORMAT_R32_FIXED: + return &fetch_R32_FIXED; + case PIPE_FORMAT_R32G32_FIXED: + return &fetch_R32G32_FIXED; + case PIPE_FORMAT_R32G32B32_FIXED: + return &fetch_R32G32B32_FIXED; + case PIPE_FORMAT_R32G32B32A32_FIXED: + return &fetch_R32G32B32A32_FIXED; + default: assert(0); return &fetch_NULL; diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 579fdb2957..00aa02311c 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -83,6 +83,7 @@ static INLINE uint pf_layout(uint f) /**< PIPE_FORMAT_LAYOUT_ */ #define PIPE_FORMAT_TYPE_USCALED 4 /**< uints, not normalized */ #define PIPE_FORMAT_TYPE_SSCALED 5 /**< ints, not normalized */ #define PIPE_FORMAT_TYPE_SRGB 6 /**< sRGB colorspace */ +#define PIPE_FORMAT_TYPE_FIXED 7 /**< 16.16 fixed point */ /** @@ -353,6 +354,10 @@ enum pipe_format { PIPE_FORMAT_R8G8B8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R8G8B8A8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ), PIPE_FORMAT_R8G8B8X8_SSCALED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ), + PIPE_FORMAT_R32_FIXED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R001, 4, 0, 0, 0, PIPE_FORMAT_TYPE_FIXED ), + PIPE_FORMAT_R32G32_FIXED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG01, 4, 4, 0, 0, PIPE_FORMAT_TYPE_FIXED ), + PIPE_FORMAT_R32G32B32_FIXED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB1, 4, 4, 4, 0, PIPE_FORMAT_TYPE_FIXED ), + PIPE_FORMAT_R32G32B32A32_FIXED = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 4, 4, 4, 4, PIPE_FORMAT_TYPE_FIXED ), /* sRGB formats */ PIPE_FORMAT_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ), PIPE_FORMAT_A8_L8_SRGB = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ), diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index da45a5e321..7d52d1da1b 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1161,7 +1161,8 @@ do_copy_texsubimage(GLcontext *ctx, (void) texImage; - /* XX need this? st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL);*/ + /* XX need this?*/ + st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); /* determine if copying depth or color data */ if (baseFormat == GL_DEPTH_COMPONENT) { diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 5300848ef6..6867d50c87 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -147,6 +147,14 @@ static GLuint byte_types_scale[4] = { PIPE_FORMAT_R8G8B8A8_SSCALED }; +static GLuint fixed_types[4] = { + PIPE_FORMAT_R32_FIXED, + PIPE_FORMAT_R32G32_FIXED, + PIPE_FORMAT_R32G32B32_FIXED, + PIPE_FORMAT_R32G32B32A32_FIXED +}; + + /** * Return a PIPE_FORMAT_x for the given GL datatype and size. @@ -154,8 +162,8 @@ static GLuint byte_types_scale[4] = { static GLuint pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized) { - assert(type >= GL_BYTE); - assert(type <= GL_DOUBLE); + assert((type >= GL_BYTE && type <= GL_DOUBLE) || + type == GL_FIXED); assert(size >= 1); assert(size <= 4); @@ -169,6 +177,7 @@ pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized) case GL_UNSIGNED_INT: return uint_types_norm[size-1]; case GL_UNSIGNED_SHORT: return ushort_types_norm[size-1]; case GL_UNSIGNED_BYTE: return ubyte_types_norm[size-1]; + case GL_FIXED: return fixed_types[size-1]; default: assert(0); return 0; } } @@ -182,6 +191,7 @@ pipe_vertex_format(GLenum type, GLuint size, GLboolean normalized) case GL_UNSIGNED_INT: return uint_types_scale[size-1]; case GL_UNSIGNED_SHORT: return ushort_types_scale[size-1]; case GL_UNSIGNED_BYTE: return ubyte_types_scale[size-1]; + case GL_FIXED: return fixed_types[size-1]; default: assert(0); return 0; } } -- cgit v1.2.3 From c47248bdf8d55f985b199fc6e15b0177305cb6fd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 10:18:08 +0900 Subject: mesa: Move variable declarations to the scope top. --- src/mesa/main/texstore.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index 519a73b960..113eef69f4 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -2405,16 +2405,15 @@ GLboolean _mesa_texstore_s8_z24(TEXSTORE_PARAMS) { const GLuint depthScale = 0xffffff; - - ASSERT(dstFormat == &_mesa_texformat_s8_z24); - ASSERT(srcFormat == GL_DEPTH_STENCIL_EXT || srcFormat == GL_DEPTH_COMPONENT); - ASSERT(srcFormat != GL_DEPTH_STENCIL_EXT || srcType == GL_UNSIGNED_INT_24_8_EXT); - const GLint srcRowStride = _mesa_image_row_stride(srcPacking, srcWidth, srcFormat, srcType) / sizeof(GLuint); GLint img, row; + ASSERT(dstFormat == &_mesa_texformat_s8_z24); + ASSERT(srcFormat == GL_DEPTH_STENCIL_EXT || srcFormat == GL_DEPTH_COMPONENT); + ASSERT(srcFormat != GL_DEPTH_STENCIL_EXT || srcType == GL_UNSIGNED_INT_24_8_EXT); + /* Incase we only upload depth we need to preserve the stencil */ if (srcFormat == GL_DEPTH_COMPONENT) { for (img = 0; img < srcDepth; img++) { -- cgit v1.2.3 From c921a29b3a696f6fa99b648d6cd975dbcf106429 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 10:56:36 +0900 Subject: mesa: Replace abort by asserts. --- src/mesa/glapi/glthread.h | 10 +++++----- src/mesa/math/m_debug_xform.c | 3 ++- src/mesa/state_tracker/st_atom_sampler.c | 2 +- src/mesa/state_tracker/st_mesa_to_tgsi.c | 3 ++- 4 files changed, 10 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/glapi/glthread.h b/src/mesa/glapi/glthread.h index b8e9d5f59b..ad2a63584b 100644 --- a/src/mesa/glapi/glthread.h +++ b/src/mesa/glapi/glthread.h @@ -144,19 +144,19 @@ typedef unsigned int _glthread_Cond; // #warning Condition variables not implemented. #define _glthread_INIT_COND(cond) \ - abort(); + ASSERT(0); #define _glthread_DESTROY_COND(name) \ - abort(); + ASSERT(0); #define _glthread_COND_WAIT(cond, mutex) \ - abort(); + ASSERT(0); #define _glthread_COND_SIGNAL(cond) \ - abort(); + ASSERT(0); #define _glthread_COND_BROADCAST(cond) \ - abort(); + ASSERT(0); #endif diff --git a/src/mesa/math/m_debug_xform.c b/src/mesa/math/m_debug_xform.c index 2af837cf64..0b07b4fd68 100644 --- a/src/mesa/math/m_debug_xform.c +++ b/src/mesa/math/m_debug_xform.c @@ -206,7 +206,8 @@ static int test_transform_function( transform_func func, int psize, case VAR: break; default: - abort(); + ASSERT(0); + return 0; } } } diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 9abd0b3531..1e182eef1c 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -67,7 +67,7 @@ gl_wrap_to_sp(GLenum wrap) case GL_MIRROR_CLAMP_TO_BORDER_EXT: return PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER; default: - abort(); + assert(0); return 0; } } diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index a8b6faad1c..150545dfad 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -772,7 +772,8 @@ tgsi_translate_mesa_program( TGSI_WRITEMASK_XYZW ); break; default: - abort(); + assert(0); + return 0; } ti += tgsi_build_full_declaration(&fulldecl, &tokens[ti], -- cgit v1.2.3 From 182b644c71462d86be2f2c79c9ff3fa636b7470a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 10:58:55 +0900 Subject: mesa: bsearch implementation for WinCE. --- src/mesa/main/imports.c | 19 ++++++++++++++++++- 1 file changed, 18 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c index d798f80e25..c46e5beb91 100644 --- a/src/mesa/main/imports.c +++ b/src/mesa/main/imports.c @@ -765,7 +765,24 @@ void * _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size, int (*compar)(const void *, const void *) ) { +#if defined(_WIN32_WCE) + void *mid; + int cmp; + while (nmemb) { + nmemb >>= 1; + mid = (char *)base + nmemb * size; + cmp = (*compar)(key, mid); + if (cmp == 0) + return mid; + if (cmp > 0) { + base = (char *)mid + size; + --nmemb; + } + } + return NULL; +#else return bsearch(key, base, nmemb, size, compar); +#endif } /*@}*/ @@ -781,7 +798,7 @@ _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size, char * _mesa_getenv( const char *var ) { -#if defined(_XBOX) +#if defined(_XBOX) || defined(_WIN32_WCE) return NULL; #else return getenv(var); -- cgit v1.2.3 From 80b359f574cd8565af46e0900d2da9dd0faf4291 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 11:33:03 +0900 Subject: mesa: Use _mesa_bsearch. --- src/mesa/main/enums.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/main/enums.c b/src/mesa/main/enums.c index 6aeb18fa27..8ce6b51d17 100644 --- a/src/mesa/main/enums.c +++ b/src/mesa/main/enums.c @@ -4870,8 +4870,8 @@ const char *_mesa_lookup_enum_by_nr( int nr ) { unsigned * i; - i = (unsigned *)bsearch( & nr, reduced_enums, Elements(reduced_enums), - sizeof(reduced_enums[0]), (cfunc) compar_nr ); + i = (unsigned *)_mesa_bsearch( & nr, reduced_enums, Elements(reduced_enums), + sizeof(reduced_enums[0]), (cfunc) compar_nr ); if ( i != NULL ) { return & enum_string_table[ all_enums[ *i ].offset ]; @@ -4888,8 +4888,8 @@ int _mesa_lookup_enum_by_name( const char *symbol ) enum_elt * f = NULL; if ( symbol != NULL ) { - f = (enum_elt *)bsearch( symbol, all_enums, Elements(all_enums), - sizeof( enum_elt ), (cfunc) compar_name ); + f = (enum_elt *)_mesa_bsearch( symbol, all_enums, Elements(all_enums), + sizeof( enum_elt ), (cfunc) compar_name ); } return (f != NULL) ? f->n : -1; -- cgit v1.2.3 From 18ec140ef27b6488bea9d54e21b08b0a3afbcafe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 11:34:46 +0900 Subject: mesa: Use appropriate unsigned/signed, float/integer types. --- src/mesa/main/mm.c | 4 ++-- src/mesa/shader/program.c | 2 +- src/mesa/shader/shader_api.c | 4 ++-- src/mesa/shader/slang/slang_compile.c | 2 +- src/mesa/shader/slang/slang_print.c | 6 +++--- src/mesa/state_tracker/st_cb_drawpixels.c | 18 +++++++++--------- src/mesa/state_tracker/st_cb_readpixels.c | 9 +++++---- src/mesa/state_tracker/st_draw.c | 2 +- src/mesa/state_tracker/st_extensions.c | 12 ++++++------ 9 files changed, 30 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mm.c b/src/mesa/main/mm.c index fb7809ed22..9f3aa00aff 100644 --- a/src/mesa/main/mm.c +++ b/src/mesa/main/mm.c @@ -164,8 +164,8 @@ mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) { struct mem_block *p; const int mask = (1 << align2)-1; - unsigned int startofs = 0; - unsigned int endofs; + int startofs = 0; + int endofs; if (!heap || align2 < 0 || size <= 0) return NULL; diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index a0817a91ec..b27ed6b7d3 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -467,7 +467,7 @@ _mesa_insert_instructions(struct gl_program *prog, GLuint start, GLuint count) for (i = 0; i < prog->NumInstructions; i++) { struct prog_instruction *inst = prog->Instructions + i; if (inst->BranchTarget > 0) { - if (inst->BranchTarget >= start) { + if ((GLuint)inst->BranchTarget >= start) { inst->BranchTarget += count; } } diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 856179e1d5..97edb25400 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -1079,7 +1079,7 @@ update_textures_used(struct gl_program *prog) */ static void set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, - GLenum type, GLint count, GLint elems, const void *values) + GLenum type, GLsizei count, GLint elems, const void *values) { if (program->Parameters->Parameters[location].Type == PROGRAM_SAMPLER) { /* This controls which texture unit which is used by a sampler */ @@ -1111,7 +1111,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, } else { /* ordinary uniform variable */ - GLuint k, i; + GLsizei k, i; if (count * elems > program->Parameters->Parameters[location].Size) { _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(count too large)"); diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index cdea1c5128..8485103129 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -1934,7 +1934,7 @@ compile_with_grammar(grammar id, const char *source, slang_code_unit * unit, byte *prod; GLuint size, start, version; slang_string preprocessed; - int maxVersion; + GLuint maxVersion; #if FEATURE_ARB_shading_language_120 maxVersion = 120; diff --git a/src/mesa/shader/slang/slang_print.c b/src/mesa/shader/slang/slang_print.c index f3e127cb13..4a7d4bbbc7 100644 --- a/src/mesa/shader/slang/slang_print.c +++ b/src/mesa/shader/slang/slang_print.c @@ -182,14 +182,14 @@ static void print_generic2(const slang_operation *op, const char *oper, const char *s, int indent) { - int i; + GLuint i; if (oper) { spaces(indent); printf("[%p locals %p] %s %s\n", (void*) op, (void*) op->locals, oper, s); } for (i = 0; i < op->num_children; i++) { spaces(indent); - printf("//child %d:\n", i); + printf("//child %u:\n", i); slang_print_tree(&op->children[i], indent); } } @@ -804,7 +804,7 @@ int slang_checksum_tree(const slang_operation *op) { int s = op->num_children; - int i; + GLuint i; for (i = 0; i < op->num_children; i++) { s += slang_checksum_tree(&op->children[i]); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index a486581989..8fdeb5c380 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -517,7 +517,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, struct pipe_context *pipe = ctx->st->pipe; struct cso_context *cso = ctx->st->cso_context; GLfloat x0, y0, x1, y1; - GLuint maxSize; + GLsizei maxSize; /* limit checks */ /* XXX if DrawPixels image is larger than max texture size, break @@ -574,14 +574,14 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, const float width = ctx->DrawBuffer->Width; const float height = ctx->DrawBuffer->Height; struct pipe_viewport_state vp; - vp.scale[0] = 0.5 * width; - vp.scale[1] = -0.5 * height; - vp.scale[2] = 1.0; - vp.scale[3] = 1.0; - vp.translate[0] = 0.5 * width; - vp.translate[1] = 0.5 * height; - vp.translate[2] = 0.0; - vp.translate[3] = 0.0; + vp.scale[0] = 0.5f * width; + vp.scale[1] = -0.5f * height; + vp.scale[2] = 1.0f; + vp.scale[3] = 1.0f; + vp.translate[0] = 0.5f * width; + vp.translate[1] = 0.5f * height; + vp.translate[2] = 0.0f; + vp.translate[3] = 0.0f; cso_set_viewport(cso, &vp); } diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 08934af319..c5193631a7 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -177,7 +177,8 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, struct pipe_screen *screen = pipe->screen; GLfloat temp[MAX_WIDTH][4]; const GLbitfield transferOps = ctx->_ImageTransferState; - GLint i, yStep, dfStride; + GLsizei i, j; + GLint yStep, dfStride; GLfloat *df; struct st_renderbuffer *strb; struct gl_pixelstore_attrib clippedPacking = *pack; @@ -258,7 +259,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, surf->format == PIPE_FORMAT_X8Z24_UNORM) { if (format == GL_DEPTH_COMPONENT) { for (i = 0; i < height; i++) { - GLuint ztemp[MAX_WIDTH], j; + GLuint ztemp[MAX_WIDTH]; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / ((1 << 24) - 1); pipe_get_tile_raw(pipe, surf, x, y, width, 1, ztemp, 0); @@ -283,7 +284,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } else if (surf->format == PIPE_FORMAT_Z16_UNORM) { for (i = 0; i < height; i++) { - GLushort ztemp[MAX_WIDTH], j; + GLushort ztemp[MAX_WIDTH]; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / 0xffff; pipe_get_tile_raw(pipe, surf, x, y, width, 1, ztemp, 0); @@ -298,7 +299,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, } else if (surf->format == PIPE_FORMAT_Z32_UNORM) { for (i = 0; i < height; i++) { - GLuint ztemp[MAX_WIDTH], j; + GLuint ztemp[MAX_WIDTH]; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / 0xffffffff; pipe_get_tile_raw(pipe, surf, x, y, width, 1, ztemp, 0); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 6867d50c87..21911774d7 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -216,7 +216,7 @@ setup_edgeflags(GLcontext *ctx, GLenum primMode, GLint start, GLint count, (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL)) { /* need edge flags */ - GLuint i; + GLint i; unsigned *vec; struct st_buffer_object *stobj = st_buffer_object(array->BufferObj); ubyte *map; diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index 6f94ba39ae..d804d2b453 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -43,7 +43,7 @@ static int _min(int a, int b) return (a < b) ? a : b; } -static int _max(int a, int b) +static float _maxf(float a, float b) { return (a > b) ? a : b; } @@ -94,17 +94,17 @@ void st_init_limits(struct st_context *st) 1, MAX_DRAW_BUFFERS); c->MaxLineWidth - = _max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH)); + = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH)); c->MaxLineWidthAA - = _max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH_AA)); + = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_LINE_WIDTH_AA)); c->MaxPointSize - = _max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH)); + = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH)); c->MaxPointSizeAA - = _max(1.0, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH_AA)); + = _maxf(1.0f, screen->get_paramf(screen, PIPE_CAP_MAX_POINT_WIDTH_AA)); c->MaxTextureMaxAnisotropy - = _max(2.0, screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_ANISOTROPY)); + = _maxf(2.0f, screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_ANISOTROPY)); c->MaxTextureLodBias = screen->get_paramf(screen, PIPE_CAP_MAX_TEXTURE_LOD_BIAS); -- cgit v1.2.3 From b750b5326df0a067f9baf237bee50d31ca69729d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 11:40:25 +0900 Subject: gdi: Update for surface_alloc_storage changes. --- src/gallium/winsys/gdi/wmesa.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/gdi/wmesa.c b/src/gallium/winsys/gdi/wmesa.c index 74a8292352..0b93f8c4c3 100644 --- a/src/gallium/winsys/gdi/wmesa.c +++ b/src/gallium/winsys/gdi/wmesa.c @@ -469,7 +469,8 @@ wm_surface_alloc_storage(struct pipe_winsys *winsys, struct pipe_surface *surf, unsigned width, unsigned height, enum pipe_format format, - unsigned flags) + unsigned flags, + unsigned tex_usage) { const unsigned alignment = 64; -- cgit v1.2.3 From 5c1a78b7a85d23ad1358b34d03a0002a19483655 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 13:12:41 +0900 Subject: mesa: More signed/unsigned float/integer fixes. --- src/mesa/main/mm.c | 20 ++++++++++---------- src/mesa/main/mm.h | 16 ++++++++-------- src/mesa/state_tracker/st_cb_bitmap.c | 16 ++++++++-------- src/mesa/state_tracker/st_cb_drawpixels.c | 24 ++++++++++++------------ 4 files changed, 38 insertions(+), 38 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mm.c b/src/mesa/main/mm.c index 9f3aa00aff..6f381b02a7 100644 --- a/src/mesa/main/mm.c +++ b/src/mesa/main/mm.c @@ -53,11 +53,11 @@ mmDumpMemInfo(const struct mem_block *heap) } struct mem_block * -mmInit(unsigned int ofs, int size) +mmInit(unsigned ofs, unsigned size) { struct mem_block *heap, *block; - if (size <= 0) + if (!size) return NULL; heap = (struct mem_block *) _mesa_calloc(sizeof(struct mem_block)); @@ -91,8 +91,8 @@ mmInit(unsigned int ofs, int size) static struct mem_block * SliceBlock(struct mem_block *p, - unsigned int startofs, int size, - int reserved, int alignment) + unsigned startofs, unsigned size, + unsigned reserved, unsigned alignment) { struct mem_block *newblock; @@ -160,14 +160,14 @@ SliceBlock(struct mem_block *p, struct mem_block * -mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) +mmAllocMem(struct mem_block *heap, unsigned size, unsigned align2, unsigned startSearch) { struct mem_block *p; - const int mask = (1 << align2)-1; - int startofs = 0; - int endofs; + const unsigned mask = (1 << align2)-1; + unsigned startofs = 0; + unsigned endofs; - if (!heap || align2 < 0 || size <= 0) + if (!heap || !align2 || !size) return NULL; for (p = heap->next_free; p != heap; p = p->next_free) { @@ -193,7 +193,7 @@ mmAllocMem(struct mem_block *heap, int size, int align2, int startSearch) struct mem_block * -mmFindBlock(struct mem_block *heap, int start) +mmFindBlock(struct mem_block *heap, unsigned start) { struct mem_block *p; diff --git a/src/mesa/main/mm.h b/src/mesa/main/mm.h index 5ad3ffd6d1..df340808ac 100644 --- a/src/mesa/main/mm.h +++ b/src/mesa/main/mm.h @@ -39,10 +39,10 @@ struct mem_block { struct mem_block *next, *prev; struct mem_block *next_free, *prev_free; struct mem_block *heap; - unsigned int ofs; - int size; - unsigned int free:1; - unsigned int reserved:1; + unsigned ofs; + unsigned size; + unsigned free:1; + unsigned reserved:1; }; @@ -51,7 +51,7 @@ struct mem_block { * input: total size in bytes * return: a heap pointer if OK, NULL if error */ -extern struct mem_block *mmInit(unsigned int ofs, int size); +extern struct mem_block *mmInit(unsigned ofs, unsigned size); /** * Allocate 'size' bytes with 2^align2 bytes alignment, @@ -63,8 +63,8 @@ extern struct mem_block *mmInit(unsigned int ofs, int size); * startSearch = linear offset from start of heap to begin search * return: pointer to the allocated block, 0 if error */ -extern struct mem_block *mmAllocMem(struct mem_block *heap, int size, int align2, - int startSearch); +extern struct mem_block *mmAllocMem(struct mem_block *heap, unsigned size, + unsigned align2, unsigned startSearch); /** * Free block starts at offset @@ -78,7 +78,7 @@ extern int mmFreeMem(struct mem_block *b); * input: pointer to a heap, start offset * return: pointer to a block */ -extern struct mem_block *mmFindBlock(struct mem_block *heap, int start); +extern struct mem_block *mmFindBlock(struct mem_block *heap, unsigned start); /** * destroy MM diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 9e32ee2eb4..6fa3cbd533 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -494,14 +494,14 @@ draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, const GLfloat width = (GLfloat)fb->Width; const GLfloat height = (GLfloat)fb->Height; struct pipe_viewport_state vp; - vp.scale[0] = 0.5 * width; - vp.scale[1] = (GLfloat)(height * (invert ? -0.5 : 0.5)); - vp.scale[2] = 1.0; - vp.scale[3] = 1.0; - vp.translate[0] = (GLfloat)(0.5 * width); - vp.translate[1] = (GLfloat)(0.5 * height); - vp.translate[2] = 0.0; - vp.translate[3] = 0.0; + vp.scale[0] = 0.5f * width; + vp.scale[1] = height * (invert ? -0.5f : 0.5f); + vp.scale[2] = 1.0f; + vp.scale[3] = 1.0f; + vp.translate[0] = 0.5f * width; + vp.translate[1] = 0.5f * height; + vp.translate[2] = 0.0f; + vp.translate[3] = 0.0f; cso_set_viewport(cso, &vp); } diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 8fdeb5c380..d8f1d2367c 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -425,12 +425,12 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, const struct gl_framebuffer *fb = st->ctx->DrawBuffer; const GLfloat fb_width = fb->Width; const GLfloat fb_height = fb->Height; - const GLfloat clip_x0 = x0 / fb_width * 2.0 - 1.0; - const GLfloat clip_y0 = y0 / fb_height * 2.0 - 1.0; - const GLfloat clip_x1 = x1 / fb_width * 2.0 - 1.0; - const GLfloat clip_y1 = y1 / fb_height * 2.0 - 1.0; - const GLfloat sLeft = 0.0F, sRight = 1.0F; - const GLfloat tTop = invertTex, tBot = 1.0 - tTop; + const GLfloat clip_x0 = x0 / fb_width * 2.0f - 1.0f; + const GLfloat clip_y0 = y0 / fb_height * 2.0f - 1.0f; + const GLfloat clip_x1 = x1 / fb_width * 2.0f - 1.0f; + const GLfloat clip_y1 = y1 / fb_height * 2.0f - 1.0f; + const GLfloat sLeft = 0.0f, sRight = 1.0f; + const GLfloat tTop = invertTex, tBot = 1.0f - tTop; GLuint tex, i; /* upper-left */ @@ -463,21 +463,21 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, if (color) { for (i = 0; i < 4; i++) { verts[i][0][2] = z; /*Z*/ - verts[i][0][3] = 1.0; /*W*/ + verts[i][0][3] = 1.0f; /*W*/ verts[i][1][0] = color[0]; verts[i][1][1] = color[1]; verts[i][1][2] = color[2]; verts[i][1][3] = color[3]; - verts[i][2][2] = 0.0; /*R*/ - verts[i][2][3] = 1.0; /*Q*/ + verts[i][2][2] = 0.0f; /*R*/ + verts[i][2][3] = 1.0f; /*Q*/ } } else { for (i = 0; i < 4; i++) { verts[i][0][2] = z; /*Z*/ - verts[i][0][3] = 1.0; /*W*/ - verts[i][1][2] = 0.0; /*R*/ - verts[i][1][3] = 1.0; /*Q*/ + verts[i][0][3] = 1.0f; /*W*/ + verts[i][1][2] = 0.0f; /*R*/ + verts[i][1][3] = 1.0f; /*Q*/ } } } -- cgit v1.2.3 From 2e328dcde2be8174b2a78936206ced71b03b2b57 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 13:51:52 +0900 Subject: mesa/dri: User standard integer types. --- src/mesa/drivers/dri/common/dri_util.h | 4 +- src/mesa/drivers/dri/common/mmio.h | 8 +- src/mesa/drivers/dri/common/utils.c | 18 +-- src/mesa/drivers/dri/common/utils.h | 2 +- src/mesa/drivers/dri/fb/fb_dri.c | 4 +- src/mesa/drivers/dri/fb/fb_egl.c | 4 +- src/mesa/drivers/dri/ffb/ffb_xmesa.c | 4 +- src/mesa/drivers/dri/gamma/gamma_context.h | 38 +++--- src/mesa/drivers/dri/gamma/gamma_render.c | 6 +- src/mesa/drivers/dri/gamma/gamma_span.c | 8 +- src/mesa/drivers/dri/gamma/gamma_state.c | 34 +++--- src/mesa/drivers/dri/gamma/gamma_tex.c | 8 +- src/mesa/drivers/dri/gamma/gamma_tritmp.h | 16 +-- src/mesa/drivers/dri/i810/i810screen.c | 4 +- src/mesa/drivers/dri/i915/intel_ioctl.c | 4 +- src/mesa/drivers/dri/i915/intel_ioctl.h | 2 +- src/mesa/drivers/dri/i915/intel_screen.c | 4 +- src/mesa/drivers/dri/i965/intel_screen.c | 4 +- src/mesa/drivers/dri/mach64/mach64_screen.c | 4 +- src/mesa/drivers/dri/mga/mga_xmesa.c | 4 +- src/mesa/drivers/dri/mga/mga_xmesa.h | 2 +- src/mesa/drivers/dri/nouveau/nouveau_context.h | 14 +-- src/mesa/drivers/dri/nouveau/nouveau_fifo.c | 4 +- src/mesa/drivers/dri/nouveau/nouveau_fifo.h | 8 +- src/mesa/drivers/dri/nouveau/nouveau_screen.c | 4 +- src/mesa/drivers/dri/nouveau/nouveau_screen.h | 4 +- src/mesa/drivers/dri/r128/r128_context.h | 6 +- src/mesa/drivers/dri/r128/r128_ioctl.c | 2 +- src/mesa/drivers/dri/r128/r128_ioctl.h | 6 +- src/mesa/drivers/dri/r128/r128_screen.c | 4 +- src/mesa/drivers/dri/r128/r128_tex.c | 6 +- src/mesa/drivers/dri/r128/r128_tex.h | 2 +- src/mesa/drivers/dri/r128/r128_texmem.c | 6 +- src/mesa/drivers/dri/r128/r128_texobj.h | 4 +- src/mesa/drivers/dri/r128/server/r128.h | 128 ++++++++++----------- src/mesa/drivers/dri/r128/server/r128_dri.c | 2 +- src/mesa/drivers/dri/r200/r200_ioctl.c | 4 +- src/mesa/drivers/dri/r200/r200_state.c | 2 +- src/mesa/drivers/dri/radeon/radeon_ioctl.c | 6 +- src/mesa/drivers/dri/radeon/radeon_screen.c | 6 +- src/mesa/drivers/dri/radeon/radeon_screen.h | 2 +- src/mesa/drivers/dri/radeon/radeon_state.c | 2 +- src/mesa/drivers/dri/radeon/server/radeon_egl.c | 4 +- src/mesa/drivers/dri/savage/savage_3d_reg.h | 70 +++++------ src/mesa/drivers/dri/savage/savage_xmesa.c | 6 +- src/mesa/drivers/dri/savage/savagecontext.h | 2 +- src/mesa/drivers/dri/savage/savageioctl.c | 52 ++++----- src/mesa/drivers/dri/savage/savageioctl.h | 14 +-- src/mesa/drivers/dri/savage/savagestate.c | 46 ++++---- src/mesa/drivers/dri/savage/savagetex.c | 6 +- src/mesa/drivers/dri/savage/savagetris.c | 14 +-- src/mesa/drivers/dri/sis/sis_screen.c | 4 +- src/mesa/drivers/dri/unichrome/server/via_dri.c | 34 +++--- src/mesa/drivers/dri/unichrome/server/via_driver.h | 26 ++--- src/mesa/drivers/dri/unichrome/via_screen.c | 4 +- src/mesa/drivers/dri/unichrome/via_state.c | 8 +- 56 files changed, 347 insertions(+), 347 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/common/dri_util.h b/src/mesa/drivers/dri/common/dri_util.h index 027cb7f461..c0cda5d25f 100644 --- a/src/mesa/drivers/dri/common/dri_util.h +++ b/src/mesa/drivers/dri/common/dri_util.h @@ -200,7 +200,7 @@ struct __DRIswapInfoRec { /** * Number of swapBuffers operations that have been *completed*. */ - u_int64_t swap_count; + uint64_t swap_count; /** * Unadjusted system time of the last buffer swap. This is the time @@ -214,7 +214,7 @@ struct __DRIswapInfoRec { * swap, it has missed its deadline. If swap_interval is 0, then the * swap deadline is 1 frame after the previous swap. */ - u_int64_t swap_missed_count; + uint64_t swap_missed_count; /** * Amount of time used by the last swap that missed its deadline. This diff --git a/src/mesa/drivers/dri/common/mmio.h b/src/mesa/drivers/dri/common/mmio.h index 2b7b4123e5..891056e170 100644 --- a/src/mesa/drivers/dri/common/mmio.h +++ b/src/mesa/drivers/dri/common/mmio.h @@ -37,10 +37,10 @@ #if defined( __powerpc__ ) -static __inline__ u_int32_t +static __inline__ uint32_t read_MMIO_LE32( volatile void * base, unsigned long offset ) { - u_int32_t val; + uint32_t val; __asm__ __volatile__( "lwbrx %0, %1, %2 ; eieio" : "=r" (val) @@ -50,10 +50,10 @@ read_MMIO_LE32( volatile void * base, unsigned long offset ) #else -static __inline__ u_int32_t +static __inline__ uint32_t read_MMIO_LE32( volatile void * base, unsigned long offset ) { - volatile u_int32_t * p = (volatile u_int32_t *) (((volatile char *) base) + offset); + volatile uint32_t * p = (volatile uint32_t *) (((volatile char *) base) + offset); return LE32_TO_CPU( p[0] ); } diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c index 6a189e7285..94db319928 100644 --- a/src/mesa/drivers/dri/common/utils.c +++ b/src/mesa/drivers/dri/common/utils.c @@ -539,12 +539,12 @@ GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, GLboolean driFillInModes( __GLcontextModes ** ptr_to_modes, GLenum fb_format, GLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + const uint8_t * depth_bits, const uint8_t * stencil_bits, unsigned num_depth_stencil_bits, const GLenum * db_modes, unsigned num_db_modes, int visType ) { - static const u_int8_t bits_table[3][4] = { + static const uint8_t bits_table[3][4] = { /* R G B A */ { 5, 6, 5, 0 }, /* Any GL_UNSIGNED_SHORT_5_6_5 */ { 8, 8, 8, 0 }, /* Any RGB with any GL_UNSIGNED_INT_8_8_8_8 */ @@ -555,7 +555,7 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, * Given the four supported fb_type values, this results in valid array * indices of 3, 4, 5, and 7. */ - static const u_int32_t masks_table_rgb[8][4] = { + static const uint32_t masks_table_rgb[8][4] = { { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, @@ -566,7 +566,7 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, { 0x000000FF, 0x0000FF00, 0x00FF0000, 0x00000000 } /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_rgba[8][4] = { + static const uint32_t masks_table_rgba[8][4] = { { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, @@ -577,7 +577,7 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, { 0x000000FF, 0x0000FF00, 0x00FF0000, 0xFF000000 }, /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_bgr[8][4] = { + static const uint32_t masks_table_bgr[8][4] = { { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, @@ -588,7 +588,7 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, { 0x00FF0000, 0x0000FF00, 0x000000FF, 0x00000000 }, /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_bgra[8][4] = { + static const uint32_t masks_table_bgra[8][4] = { { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, { 0x00000000, 0x00000000, 0x00000000, 0x00000000 }, @@ -599,12 +599,12 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, { 0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000 }, /* 8_8_8_8_REV */ }; - static const u_int8_t bytes_per_pixel[8] = { + static const uint8_t bytes_per_pixel[8] = { 0, 0, 0, 2, 2, 4, 0, 4 }; - const u_int8_t * bits; - const u_int32_t * masks; + const uint8_t * bits; + const uint32_t * masks; const int index = fb_type & 0x07; __GLcontextModes * modes = *ptr_to_modes; unsigned i; diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h index b28b895627..1067d0a8bf 100644 --- a/src/mesa/drivers/dri/common/utils.h +++ b/src/mesa/drivers/dri/common/utils.h @@ -113,7 +113,7 @@ extern GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, extern GLboolean driFillInModes( __GLcontextModes ** modes, GLenum fb_format, GLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + const uint8_t * depth_bits, const uint8_t * stencil_bits, unsigned num_depth_stencil_bits, const GLenum * db_modes, unsigned num_db_modes, int visType ); diff --git a/src/mesa/drivers/dri/fb/fb_dri.c b/src/mesa/drivers/dri/fb/fb_dri.c index a6d7590eff..523a0c0380 100644 --- a/src/mesa/drivers/dri/fb/fb_dri.c +++ b/src/mesa/drivers/dri/fb/fb_dri.c @@ -677,8 +677,8 @@ fbFillInModes( unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML /*, GLX_SWAP_COPY_OML */ }; - u_int8_t depth_bits_array[2]; - u_int8_t stencil_bits_array[2]; + uint8_t depth_bits_array[2]; + uint8_t stencil_bits_array[2]; depth_bits_array[0] = depth_bits; diff --git a/src/mesa/drivers/dri/fb/fb_egl.c b/src/mesa/drivers/dri/fb/fb_egl.c index 517e71f888..cc6a266df3 100644 --- a/src/mesa/drivers/dri/fb/fb_egl.c +++ b/src/mesa/drivers/dri/fb/fb_egl.c @@ -113,8 +113,8 @@ fbFillInConfigs(_EGLDisplay *disp, unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML /*, GLX_SWAP_COPY_OML */ }; - u_int8_t depth_bits_array[2]; - u_int8_t stencil_bits_array[2]; + uint8_t depth_bits_array[2]; + uint8_t stencil_bits_array[2]; depth_bits_array[0] = 0; depth_bits_array[1] = depth_bits; diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.c b/src/mesa/drivers/dri/ffb/ffb_xmesa.c index f521de63c0..a31cd57e88 100644 --- a/src/mesa/drivers/dri/ffb/ffb_xmesa.c +++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.c @@ -641,8 +641,8 @@ ffbFillInModes( unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML }; - u_int8_t depth_bits_array[3]; - u_int8_t stencil_bits_array[3]; + uint8_t depth_bits_array[3]; + uint8_t stencil_bits_array[3]; depth_bits_array[0] = 0; diff --git a/src/mesa/drivers/dri/gamma/gamma_context.h b/src/mesa/drivers/dri/gamma/gamma_context.h index fb70df6c37..86af674ae7 100644 --- a/src/mesa/drivers/dri/gamma/gamma_context.h +++ b/src/mesa/drivers/dri/gamma/gamma_context.h @@ -159,13 +159,13 @@ struct gamma_texture_object_t { int internalFormat; } image[GAMMA_TEX_MAXLEVELS]; - u_int32_t TextureBaseAddr[GAMMA_TEX_MAXLEVELS]; - u_int32_t TextureAddressMode; - u_int32_t TextureColorMode; - u_int32_t TextureFilterMode; - u_int32_t TextureFormat; - u_int32_t TextureReadMode; - u_int32_t TextureBorderColor; + uint32_t TextureBaseAddr[GAMMA_TEX_MAXLEVELS]; + uint32_t TextureAddressMode; + uint32_t TextureColorMode; + uint32_t TextureFilterMode; + uint32_t TextureFormat; + uint32_t TextureReadMode; + uint32_t TextureBorderColor; }; #define GAMMA_NO_PALETTE 0x0 @@ -299,18 +299,18 @@ struct gamma_context { unsigned int lastStamp; - u_int32_t ClearColor; - u_int32_t Color; - u_int32_t DitherMode; - u_int32_t ClearDepth; - u_int32_t FogMode; - u_int32_t AreaStippleMode; - u_int32_t LBReadFormat; - u_int32_t LBWriteFormat; - u_int32_t LineMode; - u_int32_t PointMode; - u_int32_t TriangleMode; - u_int32_t AntialiasMode; + uint32_t ClearColor; + uint32_t Color; + uint32_t DitherMode; + uint32_t ClearDepth; + uint32_t FogMode; + uint32_t AreaStippleMode; + uint32_t LBReadFormat; + uint32_t LBWriteFormat; + uint32_t LineMode; + uint32_t PointMode; + uint32_t TriangleMode; + uint32_t AntialiasMode; GLfloat ViewportScaleX; GLfloat ViewportScaleY; GLfloat ViewportScaleZ; diff --git a/src/mesa/drivers/dri/gamma/gamma_render.c b/src/mesa/drivers/dri/gamma/gamma_render.c index 4b462f2252..9180ef925d 100644 --- a/src/mesa/drivers/dri/gamma/gamma_render.c +++ b/src/mesa/drivers/dri/gamma/gamma_render.c @@ -74,7 +74,7 @@ static void gamma_emit( GLcontext *ctx, GLuint start, GLuint end) WRITEF(gmesa->buf, Tr4, tc0[i][2]); WRITEF(gmesa->buf, Tt4, tc0[i][0]); WRITEF(gmesa->buf, Ts4, tc0[i][1]); - WRITE(gmesa->buf, PackedColor4, *(u_int32_t*)col[i]); + WRITE(gmesa->buf, PackedColor4, *(uint32_t*)col[i]); WRITEF(gmesa->buf, Vw, coord[i][3]); WRITEF(gmesa->buf, Vz, coord[i][2]); WRITEF(gmesa->buf, Vy, coord[i][1]); @@ -85,7 +85,7 @@ static void gamma_emit( GLcontext *ctx, GLuint start, GLuint end) CHECK_DMA_BUFFER(gmesa, 7); WRITEF(gmesa->buf, Tt2, tc0[i][0]); WRITEF(gmesa->buf, Ts2, tc0[i][1]); - WRITE(gmesa->buf, PackedColor4, *(u_int32_t*)col[i]); + WRITE(gmesa->buf, PackedColor4, *(uint32_t*)col[i]); WRITEF(gmesa->buf, Vw, coord[i][3]); WRITEF(gmesa->buf, Vz, coord[i][2]); WRITEF(gmesa->buf, Vy, coord[i][1]); @@ -94,7 +94,7 @@ static void gamma_emit( GLcontext *ctx, GLuint start, GLuint end) } else { for (i=start; i < end; i++) { CHECK_DMA_BUFFER(gmesa, 4); - WRITE(gmesa->buf, PackedColor4, *(u_int32_t*)col[i]); + WRITE(gmesa->buf, PackedColor4, *(uint32_t*)col[i]); WRITEF(gmesa->buf, Vz, coord[i][2]); WRITEF(gmesa->buf, Vy, coord[i][1]); WRITEF(gmesa->buf, Vx3, coord[i][0]); diff --git a/src/mesa/drivers/dri/gamma/gamma_span.c b/src/mesa/drivers/dri/gamma/gamma_span.c index 012d77782b..ea9a20595c 100644 --- a/src/mesa/drivers/dri/gamma/gamma_span.c +++ b/src/mesa/drivers/dri/gamma/gamma_span.c @@ -171,10 +171,10 @@ static void gammaReadRGBASpan8888( const GLcontext *ctx, { gammaContextPtr gmesa = GAMMA_CONTEXT(ctx); gammaScreenPtr gammascrn = gmesa->gammaScreen; - u_int32_t dwords1, dwords2, i = 0; + uint32_t dwords1, dwords2, i = 0; char *src = (char *)rgba[0]; GLuint read = n * gammascrn->cpp; /* Number of bytes we are expecting */ - u_int32_t data; + uint32_t data; FLUSH_DMA_BUFFER(gmesa); CHECK_DMA_BUFFER(gmesa, 16); @@ -196,8 +196,8 @@ static void gammaReadRGBASpan8888( const GLcontext *ctx, moredata: - dwords1 = *(volatile u_int32_t*)(void *)(((u_int8_t*)gammascrn->regions[0].map) + (GlintOutFIFOWords)); - dwords2 = *(volatile u_int32_t*)(void *)(((u_int8_t*)gammascrn->regions[2].map) + (GlintOutFIFOWords)); + dwords1 = *(volatile uint32_t*)(void *)(((uint8_t*)gammascrn->regions[0].map) + (GlintOutFIFOWords)); + dwords2 = *(volatile uint32_t*)(void *)(((uint8_t*)gammascrn->regions[2].map) + (GlintOutFIFOWords)); if (dwords1) { memcpy(src, (char*)gammascrn->regions[1].map + 0x1000, dwords1 << 2); diff --git a/src/mesa/drivers/dri/gamma/gamma_state.c b/src/mesa/drivers/dri/gamma/gamma_state.c index a0690f64d0..9e7626a6fd 100644 --- a/src/mesa/drivers/dri/gamma/gamma_state.c +++ b/src/mesa/drivers/dri/gamma/gamma_state.c @@ -44,9 +44,9 @@ static void gammaUpdateAlphaMode( GLcontext *ctx ) { gammaContextPtr gmesa = GAMMA_CONTEXT(ctx); - u_int32_t a = gmesa->AlphaTestMode; - u_int32_t b = gmesa->AlphaBlendMode; - u_int32_t f = gmesa->AB_FBReadMode_Save = 0; + uint32_t a = gmesa->AlphaTestMode; + uint32_t b = gmesa->AlphaBlendMode; + uint32_t f = gmesa->AB_FBReadMode_Save = 0; GLubyte refByte = (GLint) (ctx->Color.AlphaRef * 255.0); a &= ~(AT_CompareMask | AT_RefValueMask); @@ -418,10 +418,10 @@ static void gammaDDClear( GLcontext *ctx, GLbitfield mask ) static void gammaUpdateZMode( GLcontext *ctx ) { gammaContextPtr gmesa = GAMMA_CONTEXT(ctx); - u_int32_t z = gmesa->DepthMode; - u_int32_t delta = gmesa->DeltaMode; - u_int32_t window = gmesa->Window; - u_int32_t lbread = gmesa->LBReadMode; + uint32_t z = gmesa->DepthMode; + uint32_t delta = gmesa->DeltaMode; + uint32_t window = gmesa->Window; + uint32_t lbread = gmesa->LBReadMode; z &= ~DM_CompareMask; @@ -537,9 +537,9 @@ static void gammaDDFlush( GLcontext *ctx ) static void gammaUpdateFogAttrib( GLcontext *ctx ) { gammaContextPtr gmesa = GAMMA_CONTEXT(ctx); - u_int32_t f = gmesa->FogMode; - u_int32_t g = gmesa->GeometryMode; - u_int32_t d = gmesa->DeltaMode; + uint32_t f = gmesa->FogMode; + uint32_t g = gmesa->GeometryMode; + uint32_t d = gmesa->DeltaMode; if (ctx->Fog.Enabled) { f |= FogModeEnable; @@ -635,7 +635,7 @@ static void gammaDDPointSize( GLcontext *ctx, GLfloat size ) static void gammaUpdatePolygon( GLcontext *ctx ) { gammaContextPtr gmesa = GAMMA_CONTEXT(ctx); - u_int32_t g = gmesa->GeometryMode; + uint32_t g = gmesa->GeometryMode; g &= ~(GM_PolyOffsetFillEnable | GM_PolyOffsetPointEnable | GM_PolyOffsetLineEnable); @@ -753,7 +753,7 @@ static void gammaDDScissor( GLcontext *ctx, static void gammaUpdateCull( GLcontext *ctx ) { gammaContextPtr gmesa = GAMMA_CONTEXT(ctx); - u_int32_t g = gmesa->GeometryMode; + uint32_t g = gmesa->GeometryMode; g &= ~(GM_PolyCullMask | GM_FFMask); @@ -973,8 +973,8 @@ static void gammaDDLightModelfv( GLcontext *ctx, GLenum pname, static void gammaDDShadeModel( GLcontext *ctx, GLenum mode ) { gammaContextPtr gmesa = GAMMA_CONTEXT(ctx); - u_int32_t g = gmesa->GeometryMode; - u_int32_t c = gmesa->ColorDDAMode; + uint32_t g = gmesa->GeometryMode; + uint32_t c = gmesa->ColorDDAMode; g &= ~GM_ShadingMask; c &= ~ColorDDAShadingMask; @@ -1241,7 +1241,7 @@ static void gammaDDEnable( GLcontext *ctx, GLenum cap, GLboolean state ) case GL_DITHER: do { - u_int32_t d = gmesa->DitherMode; + uint32_t d = gmesa->DitherMode; FLUSH_BATCH( gmesa ); if ( state ) { @@ -1273,7 +1273,7 @@ static void gammaDDEnable( GLcontext *ctx, GLenum cap, GLboolean state ) #if ENABLELIGHTING case GL_LIGHTING: do { - u_int32_t l = gmesa->LightingMode; + uint32_t l = gmesa->LightingMode; FLUSH_BATCH( gmesa ); if ( state ) { @@ -1291,7 +1291,7 @@ static void gammaDDEnable( GLcontext *ctx, GLenum cap, GLboolean state ) case GL_COLOR_MATERIAL: do { - u_int32_t m = gmesa->MaterialMode; + uint32_t m = gmesa->MaterialMode; FLUSH_BATCH( gmesa ); if ( state ) { diff --git a/src/mesa/drivers/dri/gamma/gamma_tex.c b/src/mesa/drivers/dri/gamma/gamma_tex.c index 0770cbf694..3b4ee4e581 100644 --- a/src/mesa/drivers/dri/gamma/gamma_tex.c +++ b/src/mesa/drivers/dri/gamma/gamma_tex.c @@ -31,8 +31,8 @@ static GLuint gammaComputeLodBias(GLfloat bias) static void gammaSetTexWrapping(gammaTextureObjectPtr t, GLenum wraps, GLenum wrapt) { - u_int32_t t1 = t->TextureAddressMode; - u_int32_t t2 = t->TextureReadMode; + uint32_t t1 = t->TextureAddressMode; + uint32_t t2 = t->TextureReadMode; t1 &= ~(TAM_SWrap_Mask | TAM_TWrap_Mask); t2 &= ~(TRM_UWrap_Mask | TRM_VWrap_Mask); @@ -57,8 +57,8 @@ static void gammaSetTexFilter(gammaContextPtr gmesa, GLenum minf, GLenum magf, GLfloat bias) { - u_int32_t t1 = t->TextureAddressMode; - u_int32_t t2 = t->TextureReadMode; + uint32_t t1 = t->TextureAddressMode; + uint32_t t2 = t->TextureReadMode; t2 &= ~(TRM_Mag_Mask | TRM_Min_Mask); diff --git a/src/mesa/drivers/dri/gamma/gamma_tritmp.h b/src/mesa/drivers/dri/gamma/gamma_tritmp.h index 56e0a850c8..bea2508d4a 100644 --- a/src/mesa/drivers/dri/gamma/gamma_tritmp.h +++ b/src/mesa/drivers/dri/gamma/gamma_tritmp.h @@ -2,8 +2,8 @@ static void TAG(gamma_point)( gammaContextPtr gmesa, const gammaVertex *v0 ) { - u_int32_t vColor; - u_int32_t vBegin; + uint32_t vColor; + uint32_t vBegin; vBegin = gmesa->Begin | B_PrimType_Points; @@ -56,8 +56,8 @@ static void TAG(gamma_line)( gammaContextPtr gmesa, const gammaVertex *v0, const gammaVertex *v1 ) { - u_int32_t vColor; - u_int32_t vBegin; + uint32_t vColor; + uint32_t vBegin; vBegin = gmesa->Begin | B_PrimType_Lines; @@ -165,8 +165,8 @@ static void TAG(gamma_triangle)( gammaContextPtr gmesa, const gammaVertex *v1, const gammaVertex *v2 ) { - u_int32_t vColor; - u_int32_t vBegin; + uint32_t vColor; + uint32_t vBegin; vBegin = gmesa->Begin | B_PrimType_Triangles; @@ -310,8 +310,8 @@ static void TAG(gamma_quad)( gammaContextPtr gmesa, const gammaVertex *v2, const gammaVertex *v3 ) { - u_int32_t vColor; - u_int32_t vBegin; + uint32_t vColor; + uint32_t vBegin; vBegin = gmesa->Begin | B_PrimType_Quads; diff --git a/src/mesa/drivers/dri/i810/i810screen.c b/src/mesa/drivers/dri/i810/i810screen.c index 695b996319..733c97b2ad 100644 --- a/src/mesa/drivers/dri/i810/i810screen.c +++ b/src/mesa/drivers/dri/i810/i810screen.c @@ -63,11 +63,11 @@ static __GLcontextModes *fill_in_modes( __GLcontextModes *modes, unsigned num_db_modes, int visType ) { - static const u_int8_t bits[1][4] = { + static const uint8_t bits[1][4] = { { 5, 6, 5, 0 } }; - static const u_int32_t masks[1][4] = { + static const uint32_t masks[1][4] = { { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 } }; diff --git a/src/mesa/drivers/dri/i915/intel_ioctl.c b/src/mesa/drivers/dri/i915/intel_ioctl.c index ede3b6378f..aa7c6d106c 100644 --- a/src/mesa/drivers/dri/i915/intel_ioctl.c +++ b/src/mesa/drivers/dri/i915/intel_ioctl.c @@ -40,10 +40,10 @@ #include "intel_batchbuffer.h" #include "drm.h" -u_int32_t intelGetLastFrame (intelContextPtr intel) +uint32_t intelGetLastFrame (intelContextPtr intel) { int ret; - u_int32_t frame; + uint32_t frame; drm_i915_getparam_t gp; gp.param = I915_PARAM_LAST_DISPATCH; diff --git a/src/mesa/drivers/dri/i915/intel_ioctl.h b/src/mesa/drivers/dri/i915/intel_ioctl.h index 6ea47e462e..8d79e9d73b 100644 --- a/src/mesa/drivers/dri/i915/intel_ioctl.h +++ b/src/mesa/drivers/dri/i915/intel_ioctl.h @@ -67,6 +67,6 @@ extern GLboolean intelIsAgpMemory( intelContextPtr intel, const GLvoid *pointer, extern GLuint intelAgpOffsetFromVirtual( intelContextPtr intel, const GLvoid *p ); extern void intelWaitIrq( intelContextPtr intel, int seq ); -extern u_int32_t intelGetLastFrame (intelContextPtr intel); +extern uint32_t intelGetLastFrame (intelContextPtr intel); extern int intelEmitIrqLocked( intelContextPtr intel ); #endif diff --git a/src/mesa/drivers/dri/i915/intel_screen.c b/src/mesa/drivers/dri/i915/intel_screen.c index ca8610b496..7e7935510a 100644 --- a/src/mesa/drivers/dri/i915/intel_screen.c +++ b/src/mesa/drivers/dri/i915/intel_screen.c @@ -566,8 +566,8 @@ intelFillInModes( unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML }; - u_int8_t depth_bits_array[3]; - u_int8_t stencil_bits_array[3]; + uint8_t depth_bits_array[3]; + uint8_t stencil_bits_array[3]; depth_bits_array[0] = 0; diff --git a/src/mesa/drivers/dri/i965/intel_screen.c b/src/mesa/drivers/dri/i965/intel_screen.c index 5dac50df32..bc3dd30b7e 100644 --- a/src/mesa/drivers/dri/i965/intel_screen.c +++ b/src/mesa/drivers/dri/i965/intel_screen.c @@ -578,8 +578,8 @@ intelFillInModes( unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML }; - u_int8_t depth_bits_array[3]; - u_int8_t stencil_bits_array[3]; + uint8_t depth_bits_array[3]; + uint8_t stencil_bits_array[3]; depth_bits_array[0] = 0; diff --git a/src/mesa/drivers/dri/mach64/mach64_screen.c b/src/mesa/drivers/dri/mach64/mach64_screen.c index b780ba65ea..300a3deaaf 100644 --- a/src/mesa/drivers/dri/mach64/mach64_screen.c +++ b/src/mesa/drivers/dri/mach64/mach64_screen.c @@ -77,12 +77,12 @@ static __GLcontextModes * fill_in_modes( __GLcontextModes * modes, unsigned num_db_modes, int visType ) { - static const u_int8_t bits[2][4] = { + static const uint8_t bits[2][4] = { { 5, 6, 5, 0 }, { 8, 8, 8, 0 } }; - static const u_int32_t masks[2][4] = { + static const uint32_t masks[2][4] = { { 0x0000F800, 0x000007E0, 0x0000001F, 0x00000000 }, { 0x00FF0000, 0x0000FF00, 0x000000FF, 0x00000000 } }; diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.c b/src/mesa/drivers/dri/mga/mga_xmesa.c index 6148f6b488..d0cf06fea5 100644 --- a/src/mesa/drivers/dri/mga/mga_xmesa.c +++ b/src/mesa/drivers/dri/mga/mga_xmesa.c @@ -130,8 +130,8 @@ mgaFillInModes( unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML }; - u_int8_t depth_bits_array[3]; - u_int8_t stencil_bits_array[3]; + uint8_t depth_bits_array[3]; + uint8_t stencil_bits_array[3]; depth_bits_array[0] = 0; diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.h b/src/mesa/drivers/dri/mga/mga_xmesa.h index 0f81c9cbec..7de7bb06e8 100644 --- a/src/mesa/drivers/dri/mga/mga_xmesa.h +++ b/src/mesa/drivers/dri/mga/mga_xmesa.h @@ -148,7 +148,7 @@ do { \ #define MGA_BASE( reg ) ((unsigned long)(mmesa->mgaScreen->mmio.map)) #define MGA_ADDR( reg ) (MGA_BASE(reg) + reg) -#define MGA_DEREF( reg ) *(volatile u_int32_t *)MGA_ADDR( reg ) +#define MGA_DEREF( reg ) *(volatile uint32_t *)MGA_ADDR( reg ) #define MGA_READ( reg ) MGA_DEREF( reg ) #endif diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.h b/src/mesa/drivers/dri/nouveau/nouveau_context.h index 9a0be2cb2a..db4d4cb6b7 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_context.h +++ b/src/mesa/drivers/dri/nouveau/nouveau_context.h @@ -46,13 +46,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. typedef struct nouveau_fifo_t{ int channel; - u_int32_t* buffer; - u_int32_t* mmio; - u_int32_t put_base; - u_int32_t current; - u_int32_t put; - u_int32_t free; - u_int32_t max; + uint32_t* buffer; + uint32_t* mmio; + uint32_t put_base; + uint32_t current; + uint32_t put; + uint32_t free; + uint32_t max; } nouveau_fifo; diff --git a/src/mesa/drivers/dri/nouveau/nouveau_fifo.c b/src/mesa/drivers/dri/nouveau/nouveau_fifo.c index 7b5e96b4c2..5eb53aa46c 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_fifo.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_fifo.c @@ -45,12 +45,12 @@ int nouveau_fifo_remaining=0; #define RING_SKIPS 8 -void WAIT_RING(nouveauContextPtr nmesa,u_int32_t size) +void WAIT_RING(nouveauContextPtr nmesa,uint32_t size) { #ifdef NOUVEAU_RING_DEBUG return; #endif - u_int32_t fifo_get; + uint32_t fifo_get; while(nmesa->fifo.free < size+1) { fifo_get = NV_FIFO_READ_GET(); diff --git a/src/mesa/drivers/dri/nouveau/nouveau_fifo.h b/src/mesa/drivers/dri/nouveau/nouveau_fifo.h index 23325dcea5..67f9cd4fc8 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_fifo.h +++ b/src/mesa/drivers/dri/nouveau/nouveau_fifo.h @@ -44,10 +44,10 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define NOUVEAU_RING_TRACE 1 #endif -#define NV_READ(reg) *(volatile u_int32_t *)(nmesa->mmio + (reg)) +#define NV_READ(reg) *(volatile uint32_t *)(nmesa->mmio + (reg)) -#define NV_FIFO_READ(reg) *(volatile u_int32_t *)(nmesa->fifo.mmio + (reg/4)) -#define NV_FIFO_WRITE(reg,value) *(volatile u_int32_t *)(nmesa->fifo.mmio + (reg/4)) = value; +#define NV_FIFO_READ(reg) *(volatile uint32_t *)(nmesa->fifo.mmio + (reg/4)) +#define NV_FIFO_WRITE(reg,value) *(volatile uint32_t *)(nmesa->fifo.mmio + (reg/4)) = value; #define NV_FIFO_READ_GET() ((NV_FIFO_READ(NV03_FIFO_REGS_DMAGET) - nmesa->fifo.put_base) >> 2) #define NV_FIFO_WRITE_PUT(val) do { \ if (NOUVEAU_RING_TRACE) {\ @@ -136,7 +136,7 @@ if (NOUVEAU_RING_TRACE) \ #endif -extern void WAIT_RING(nouveauContextPtr nmesa,u_int32_t size); +extern void WAIT_RING(nouveauContextPtr nmesa,uint32_t size); extern void nouveau_state_cache_flush(nouveauContextPtr nmesa); extern void nouveau_state_cache_init(nouveauContextPtr nmesa); diff --git a/src/mesa/drivers/dri/nouveau/nouveau_screen.c b/src/mesa/drivers/dri/nouveau/nouveau_screen.c index 69b0691bb7..2cf6f979e4 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_screen.c +++ b/src/mesa/drivers/dri/nouveau/nouveau_screen.c @@ -257,8 +257,8 @@ nouveauFillInModes( unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML }; - u_int8_t depth_bits_array[4] = { 0, 16, 24, 24 }; - u_int8_t stencil_bits_array[4] = { 0, 0, 0, 8 }; + uint8_t depth_bits_array[4] = { 0, 16, 24, 24 }; + uint8_t stencil_bits_array[4] = { 0, 0, 0, 8 }; depth_buffer_factor = 4; back_buffer_factor = (have_back_buffer) ? 3 : 1; diff --git a/src/mesa/drivers/dri/nouveau/nouveau_screen.h b/src/mesa/drivers/dri/nouveau/nouveau_screen.h index decdafa86d..bbe5810128 100644 --- a/src/mesa/drivers/dri/nouveau/nouveau_screen.h +++ b/src/mesa/drivers/dri/nouveau/nouveau_screen.h @@ -35,8 +35,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. typedef struct { nouveau_card* card; - u_int32_t bus_type; - u_int32_t agp_mode; + uint32_t bus_type; + uint32_t agp_mode; GLint fbFormat; diff --git a/src/mesa/drivers/dri/r128/r128_context.h b/src/mesa/drivers/dri/r128/r128_context.h index 3f96836df1..1d10cb0b4f 100644 --- a/src/mesa/drivers/dri/r128/r128_context.h +++ b/src/mesa/drivers/dri/r128/r128_context.h @@ -137,9 +137,9 @@ struct r128_context { GLfloat hw_viewport[16]; GLfloat depth_scale; - u_int32_t ClearColor; /* Color used to clear color buffer */ - u_int32_t ClearDepth; /* Value used to clear depth buffer */ - u_int32_t ClearStencil; /* Value used to clear stencil */ + uint32_t ClearColor; /* Color used to clear color buffer */ + uint32_t ClearDepth; /* Value used to clear depth buffer */ + uint32_t ClearStencil; /* Value used to clear stencil */ /* Map GL texture units onto hardware */ diff --git a/src/mesa/drivers/dri/r128/r128_ioctl.c b/src/mesa/drivers/dri/r128/r128_ioctl.c index 25188061a0..dcb5d10459 100644 --- a/src/mesa/drivers/dri/r128/r128_ioctl.c +++ b/src/mesa/drivers/dri/r128/r128_ioctl.c @@ -230,7 +230,7 @@ static int r128WaitForFrameCompletion( r128ContextPtr rmesa ) int wait = 0; while ( 1 ) { - u_int32_t frame = read_MMIO_LE32( R128MMIO, R128_LAST_FRAME_REG ); + uint32_t frame = read_MMIO_LE32( R128MMIO, R128_LAST_FRAME_REG ); if ( rmesa->sarea->last_frame - frame <= R128_MAX_OUTSTANDING ) { break; diff --git a/src/mesa/drivers/dri/r128/r128_ioctl.h b/src/mesa/drivers/dri/r128/r128_ioctl.h index 57063c41f5..52ae3a293f 100644 --- a/src/mesa/drivers/dri/r128/r128_ioctl.h +++ b/src/mesa/drivers/dri/r128/r128_ioctl.h @@ -38,7 +38,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #include "r128_reg.h" #include "r128_lock.h" -#define R128_BUFFER_MAX_DWORDS (R128_BUFFER_SIZE / sizeof(u_int32_t)) +#define R128_BUFFER_MAX_DWORDS (R128_BUFFER_SIZE / sizeof(uint32_t)) extern drmBufPtr r128GetBufferLocked( r128ContextPtr rmesa ); @@ -47,7 +47,7 @@ extern void r128FlushVerticesLocked( r128ContextPtr rmesa ); static __inline void *r128AllocDmaLow( r128ContextPtr rmesa, int count, int vert_size ) { - u_int32_t *head; + uint32_t *head; int bytes = count * vert_size; if ( !rmesa->vert_buf ) { @@ -61,7 +61,7 @@ static __inline void *r128AllocDmaLow( r128ContextPtr rmesa, int count, UNLOCK_HARDWARE( rmesa ); } - head = (u_int32_t *)((char *)rmesa->vert_buf->address + rmesa->vert_buf->used); + head = (uint32_t *)((char *)rmesa->vert_buf->address + rmesa->vert_buf->used); rmesa->vert_buf->used += bytes; rmesa->num_verts += count; diff --git a/src/mesa/drivers/dri/r128/r128_screen.c b/src/mesa/drivers/dri/r128/r128_screen.c index 0722b80ee5..1e9616272f 100644 --- a/src/mesa/drivers/dri/r128/r128_screen.c +++ b/src/mesa/drivers/dri/r128/r128_screen.c @@ -442,8 +442,8 @@ r128FillInModes( unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML /*, GLX_SWAP_COPY_OML */ }; - u_int8_t depth_bits_array[2]; - u_int8_t stencil_bits_array[2]; + uint8_t depth_bits_array[2]; + uint8_t stencil_bits_array[2]; depth_bits_array[0] = depth_bits; diff --git a/src/mesa/drivers/dri/r128/r128_tex.c b/src/mesa/drivers/dri/r128/r128_tex.c index 554a92287f..5712351b03 100644 --- a/src/mesa/drivers/dri/r128/r128_tex.c +++ b/src/mesa/drivers/dri/r128/r128_tex.c @@ -461,9 +461,9 @@ static void r128TexEnv( GLcontext *ctx, GLenum target, case GL_TEXTURE_LOD_BIAS: { - u_int32_t t = rmesa->setup.tex_cntl_c; + uint32_t t = rmesa->setup.tex_cntl_c; GLint bias; - u_int32_t b; + uint32_t b; /* GTH: This isn't exactly correct, but gives good results up to a * certain point. It is better than completely ignoring the LOD @@ -483,7 +483,7 @@ static void r128TexEnv( GLcontext *ctx, GLenum target, bias = 127; } - b = (u_int32_t)bias & 0xff; + b = (uint32_t)bias & 0xff; t &= ~R128_LOD_BIAS_MASK; t |= (b << R128_LOD_BIAS_SHIFT); diff --git a/src/mesa/drivers/dri/r128/r128_tex.h b/src/mesa/drivers/dri/r128/r128_tex.h index 994dffb5a9..e863436f4b 100644 --- a/src/mesa/drivers/dri/r128/r128_tex.h +++ b/src/mesa/drivers/dri/r128/r128_tex.h @@ -67,7 +67,7 @@ extern void r128InitTextureFuncs( struct dd_function_table *functions ); #define R128PACKCOLOR4444( r, g, b, a ) \ ((((a) & 0xf0) << 8) | (((r) & 0xf0) << 4) | ((g) & 0xf0) | ((b) >> 4)) -static __inline__ u_int32_t r128PackColor( GLuint cpp, +static __inline__ uint32_t r128PackColor( GLuint cpp, GLubyte r, GLubyte g, GLubyte b, GLubyte a ) { diff --git a/src/mesa/drivers/dri/r128/r128_texmem.c b/src/mesa/drivers/dri/r128/r128_texmem.c index a7d0280636..af510a38c9 100644 --- a/src/mesa/drivers/dri/r128/r128_texmem.c +++ b/src/mesa/drivers/dri/r128/r128_texmem.c @@ -84,7 +84,7 @@ static void uploadSubImage( r128ContextPtr rmesa, r128TexObjPtr t, int imageWidth, imageHeight; int remaining, rows; int format, dwords; - u_int32_t pitch, offset; + uint32_t pitch, offset; int i; /* Ensure we have a valid texture to upload */ @@ -200,7 +200,7 @@ static void uploadSubImage( r128ContextPtr rmesa, r128TexObjPtr t, remaining > 0 ; remaining -= rows, y += rows, i++ ) { - u_int32_t *dst; + uint32_t *dst; drmBufPtr buffer; assert(image->Data); @@ -211,7 +211,7 @@ static void uploadSubImage( r128ContextPtr rmesa, r128TexObjPtr t, LOCK_HARDWARE( rmesa ); buffer = r128GetBufferLocked( rmesa ); - dst = (u_int32_t *)((char *)buffer->address + R128_HOSTDATA_BLIT_OFFSET); + dst = (uint32_t *)((char *)buffer->address + R128_HOSTDATA_BLIT_OFFSET); /* Copy the next chunck of the texture image into the blit buffer */ { diff --git a/src/mesa/drivers/dri/r128/r128_texobj.h b/src/mesa/drivers/dri/r128/r128_texobj.h index 08eac87758..8a0596eea8 100644 --- a/src/mesa/drivers/dri/r128/r128_texobj.h +++ b/src/mesa/drivers/dri/r128/r128_texobj.h @@ -52,14 +52,14 @@ typedef struct r128_tex_obj r128TexObj, *r128TexObjPtr; struct r128_tex_obj { driTextureObject base; - u_int32_t bufAddr; /* Offset to start of locally + uint32_t bufAddr; /* Offset to start of locally shared texture block */ GLuint age; r128TexImage image[R128_MAX_TEXTURE_LEVELS]; /* Image data for all mipmap levels */ - u_int32_t textureFormat; /* Actual hardware format */ + uint32_t textureFormat; /* Actual hardware format */ drm_r128_texture_regs_t setup; /* Setup regs for texture */ }; diff --git a/src/mesa/drivers/dri/r128/server/r128.h b/src/mesa/drivers/dri/r128/server/r128.h index ca08d7c86a..413c7c6690 100644 --- a/src/mesa/drivers/dri/r128/server/r128.h +++ b/src/mesa/drivers/dri/r128/server/r128.h @@ -86,74 +86,74 @@ typedef struct { /* All values in XCLKS */ typedef struct { /* Common registers */ - u_int32_t ovr_clr; - u_int32_t ovr_wid_left_right; - u_int32_t ovr_wid_top_bottom; - u_int32_t ov0_scale_cntl; - u_int32_t mpp_tb_config; - u_int32_t mpp_gp_config; - u_int32_t subpic_cntl; - u_int32_t viph_control; - u_int32_t i2c_cntl_1; - u_int32_t gen_int_cntl; - u_int32_t cap0_trig_cntl; - u_int32_t cap1_trig_cntl; - u_int32_t bus_cntl; - u_int32_t config_cntl; + uint32_t ovr_clr; + uint32_t ovr_wid_left_right; + uint32_t ovr_wid_top_bottom; + uint32_t ov0_scale_cntl; + uint32_t mpp_tb_config; + uint32_t mpp_gp_config; + uint32_t subpic_cntl; + uint32_t viph_control; + uint32_t i2c_cntl_1; + uint32_t gen_int_cntl; + uint32_t cap0_trig_cntl; + uint32_t cap1_trig_cntl; + uint32_t bus_cntl; + uint32_t config_cntl; /* Other registers to save for VT switches */ - u_int32_t dp_datatype; - u_int32_t gen_reset_cntl; - u_int32_t clock_cntl_index; - u_int32_t amcgpio_en_reg; - u_int32_t amcgpio_mask; + uint32_t dp_datatype; + uint32_t gen_reset_cntl; + uint32_t clock_cntl_index; + uint32_t amcgpio_en_reg; + uint32_t amcgpio_mask; /* CRTC registers */ - u_int32_t crtc_gen_cntl; - u_int32_t crtc_ext_cntl; - u_int32_t dac_cntl; - u_int32_t crtc_h_total_disp; - u_int32_t crtc_h_sync_strt_wid; - u_int32_t crtc_v_total_disp; - u_int32_t crtc_v_sync_strt_wid; - u_int32_t crtc_offset; - u_int32_t crtc_offset_cntl; - u_int32_t crtc_pitch; + uint32_t crtc_gen_cntl; + uint32_t crtc_ext_cntl; + uint32_t dac_cntl; + uint32_t crtc_h_total_disp; + uint32_t crtc_h_sync_strt_wid; + uint32_t crtc_v_total_disp; + uint32_t crtc_v_sync_strt_wid; + uint32_t crtc_offset; + uint32_t crtc_offset_cntl; + uint32_t crtc_pitch; /* CRTC2 registers */ - u_int32_t crtc2_gen_cntl; + uint32_t crtc2_gen_cntl; /* Flat panel registers */ - u_int32_t fp_crtc_h_total_disp; - u_int32_t fp_crtc_v_total_disp; - u_int32_t fp_gen_cntl; - u_int32_t fp_h_sync_strt_wid; - u_int32_t fp_horz_stretch; - u_int32_t fp_panel_cntl; - u_int32_t fp_v_sync_strt_wid; - u_int32_t fp_vert_stretch; - u_int32_t lvds_gen_cntl; - u_int32_t tmds_crc; - u_int32_t tmds_transmitter_cntl; + uint32_t fp_crtc_h_total_disp; + uint32_t fp_crtc_v_total_disp; + uint32_t fp_gen_cntl; + uint32_t fp_h_sync_strt_wid; + uint32_t fp_horz_stretch; + uint32_t fp_panel_cntl; + uint32_t fp_v_sync_strt_wid; + uint32_t fp_vert_stretch; + uint32_t lvds_gen_cntl; + uint32_t tmds_crc; + uint32_t tmds_transmitter_cntl; /* Computed values for PLL */ - u_int32_t dot_clock_freq; - u_int32_t pll_output_freq; + uint32_t dot_clock_freq; + uint32_t pll_output_freq; int feedback_div; int post_div; /* PLL registers */ - u_int32_t ppll_ref_div; - u_int32_t ppll_div_3; - u_int32_t htotal_cntl; + uint32_t ppll_ref_div; + uint32_t ppll_div_3; + uint32_t htotal_cntl; /* DDA register */ - u_int32_t dda_config; - u_int32_t dda_on_off; + uint32_t dda_config; + uint32_t dda_on_off; /* Pallet */ GLboolean palette_valid; - u_int32_t palette[256]; + uint32_t palette[256]; } R128SaveRec, *R128SavePtr; typedef struct { @@ -168,8 +168,8 @@ typedef struct { unsigned char *MMIO; /* Map of MMIO region */ unsigned char *FB; /* Map of frame buffer */ - u_int32_t MemCntl; - u_int32_t BusCntl; + uint32_t MemCntl; + uint32_t BusCntl; unsigned long FbMapSize; /* Size of frame buffer, in bytes */ int Flags; /* Saved copy of mode flags */ @@ -200,7 +200,7 @@ typedef struct { /* Computed values for Rage 128 */ int pitch; int datatype; - u_int32_t dp_gui_master_cntl; + uint32_t dp_gui_master_cntl; /* Saved values for ScreenToScreenCopy */ int xdir; @@ -305,18 +305,18 @@ typedef struct { int log2TexGran; /* Saved scissor values */ - u_int32_t sc_left; - u_int32_t sc_right; - u_int32_t sc_top; - u_int32_t sc_bottom; + uint32_t sc_left; + uint32_t sc_right; + uint32_t sc_top; + uint32_t sc_bottom; - u_int32_t re_top_left; - u_int32_t re_width_height; + uint32_t re_top_left; + uint32_t re_width_height; - u_int32_t aux_sc_cntl; + uint32_t aux_sc_cntl; int irq; - u_int32_t gen_int_cntl; + uint32_t gen_int_cntl; GLboolean DMAForXv; @@ -387,7 +387,7 @@ do { \ #define R128_VERBOSE 0 -#define RING_LOCALS u_int32_t *__head; int __count; +#define RING_LOCALS uint32_t *__head; int __count; #define R128CCE_REFRESH(pScrn, info) \ do { \ @@ -428,12 +428,12 @@ do { \ fprintf(stderr, \ "ADVANCE_RING() used: %d+%d=%d/%d\n", \ info->indirectBuffer->used - info->indirectStart, \ - __count * sizeof(u_int32_t), \ + __count * sizeof(uint32_t), \ info->indirectBuffer->used - info->indirectStart + \ - __count * sizeof(u_int32_t), \ + __count * sizeof(uint32_t), \ info->indirectBuffer->total - info->indirectStart ); \ } \ - info->indirectBuffer->used += __count * (int)sizeof(u_int32_t); \ + info->indirectBuffer->used += __count * (int)sizeof(uint32_t); \ } while (0) #define OUT_RING( x ) do { \ diff --git a/src/mesa/drivers/dri/r128/server/r128_dri.c b/src/mesa/drivers/dri/r128/server/r128_dri.c index efe9232dc2..6e3db948af 100644 --- a/src/mesa/drivers/dri/r128/server/r128_dri.c +++ b/src/mesa/drivers/dri/r128/server/r128_dri.c @@ -271,7 +271,7 @@ static GLboolean R128DRIPciInit(const DRIDriverContext *ctx) { R128InfoPtr info = ctx->driverPrivate; unsigned char *R128MMIO = ctx->MMIOAddress; - u_int32_t chunk; + uint32_t chunk; int ret; int flags; diff --git a/src/mesa/drivers/dri/r200/r200_ioctl.c b/src/mesa/drivers/dri/r200/r200_ioctl.c index 2366bde525..ed57defb49 100644 --- a/src/mesa/drivers/dri/r200/r200_ioctl.c +++ b/src/mesa/drivers/dri/r200/r200_ioctl.c @@ -333,11 +333,11 @@ void r200AllocDmaRegion( r200ContextPtr rmesa, * SwapBuffers with client-side throttling */ -static u_int32_t r200GetLastFrame(r200ContextPtr rmesa) +static uint32_t r200GetLastFrame(r200ContextPtr rmesa) { drm_radeon_getparam_t gp; int ret; - u_int32_t frame; + uint32_t frame; gp.param = RADEON_PARAM_LAST_FRAME; gp.value = (int *)&frame; diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c index 1d975ecd57..e9d2d9a3cd 100644 --- a/src/mesa/drivers/dri/r200/r200_state.c +++ b/src/mesa/drivers/dri/r200/r200_state.c @@ -902,7 +902,7 @@ static void r200PolygonMode( GLcontext *ctx, GLenum face, GLenum mode ) static void r200UpdateSpecular( GLcontext *ctx ) { r200ContextPtr rmesa = R200_CONTEXT(ctx); - u_int32_t p = rmesa->hw.ctx.cmd[CTX_PP_CNTL]; + uint32_t p = rmesa->hw.ctx.cmd[CTX_PP_CNTL]; R200_STATECHANGE( rmesa, tcl ); R200_STATECHANGE( rmesa, vtx ); diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c index f7e461239e..a0c563c7cb 100644 --- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c +++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c @@ -777,11 +777,11 @@ void radeonAllocDmaRegion( radeonContextPtr rmesa, * SwapBuffers with client-side throttling */ -static u_int32_t radeonGetLastFrame (radeonContextPtr rmesa) +static uint32_t radeonGetLastFrame (radeonContextPtr rmesa) { drm_radeon_getparam_t gp; int ret; - u_int32_t frame; + uint32_t frame; gp.param = RADEON_PARAM_LAST_FRAME; gp.value = (int *)&frame; @@ -1025,7 +1025,7 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask ) radeonContextPtr rmesa = RADEON_CONTEXT(ctx); __DRIdrawablePrivate *dPriv = rmesa->dri.drawable; drm_radeon_sarea_t *sarea = rmesa->sarea; - u_int32_t clear; + uint32_t clear; GLuint flags = 0; GLuint color_mask = 0; GLint ret, i; diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c index 4a45948608..762848fc26 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.c +++ b/src/mesa/drivers/dri/radeon/radeon_screen.c @@ -271,8 +271,8 @@ radeonFillInModes( unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML /*, GLX_SWAP_COPY_OML */ }; - u_int8_t depth_bits_array[2]; - u_int8_t stencil_bits_array[2]; + uint8_t depth_bits_array[2]; + uint8_t stencil_bits_array[2]; depth_bits_array[0] = depth_bits; @@ -427,7 +427,7 @@ radeonCreateScreen( __DRIscreenPrivate *sPriv ) __driUtilMessage("%s: drmMap (2) failed\n", __FUNCTION__ ); return NULL; } - screen->scratch = (__volatile__ u_int32_t *) + screen->scratch = (__volatile__ uint32_t *) ((GLubyte *)screen->status.map + RADEON_SCRATCH_REG_OFFSET); screen->buffers = drmMapBufs( sPriv->fd ); diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.h b/src/mesa/drivers/dri/radeon/radeon_screen.h index f8c0cc96df..c9b0c3af12 100644 --- a/src/mesa/drivers/dri/radeon/radeon_screen.h +++ b/src/mesa/drivers/dri/radeon/radeon_screen.h @@ -83,7 +83,7 @@ typedef struct { drmBufMapPtr buffers; - __volatile__ u_int32_t *scratch; + __volatile__ uint32_t *scratch; __DRIscreenPrivate *driScreen; unsigned int sarea_priv_offset; diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c index 856d27df75..1fe5c244cd 100644 --- a/src/mesa/drivers/dri/radeon/radeon_state.c +++ b/src/mesa/drivers/dri/radeon/radeon_state.c @@ -687,7 +687,7 @@ static void radeonPolygonMode( GLcontext *ctx, GLenum face, GLenum mode ) static void radeonUpdateSpecular( GLcontext *ctx ) { radeonContextPtr rmesa = RADEON_CONTEXT(ctx); - u_int32_t p = rmesa->hw.ctx.cmd[CTX_PP_CNTL]; + uint32_t p = rmesa->hw.ctx.cmd[CTX_PP_CNTL]; GLuint flag = 0; RADEON_STATECHANGE( rmesa, tcl ); diff --git a/src/mesa/drivers/dri/radeon/server/radeon_egl.c b/src/mesa/drivers/dri/radeon/server/radeon_egl.c index 2f6ea55903..c16d66e4ec 100644 --- a/src/mesa/drivers/dri/radeon/server/radeon_egl.c +++ b/src/mesa/drivers/dri/radeon/server/radeon_egl.c @@ -935,8 +935,8 @@ radeonFillInConfigs(_EGLDisplay *disp, unsigned pixel_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML /*, GLX_SWAP_COPY_OML */ }; - u_int8_t depth_bits_array[2]; - u_int8_t stencil_bits_array[2]; + uint8_t depth_bits_array[2]; + uint8_t stencil_bits_array[2]; depth_bits_array[0] = depth_bits; depth_bits_array[1] = depth_bits; diff --git a/src/mesa/drivers/dri/savage/savage_3d_reg.h b/src/mesa/drivers/dri/savage/savage_3d_reg.h index bc81d73296..45733cdf92 100644 --- a/src/mesa/drivers/dri/savage/savage_3d_reg.h +++ b/src/mesa/drivers/dri/savage/savage_3d_reg.h @@ -43,7 +43,7 @@ typedef union unsigned reserved : 4; unsigned ofs : 28; }ni; - u_int32_t ui; + uint32_t ui; } savageRegZPixelOffset; /* This reg exists only on Savage4. */ @@ -60,7 +60,7 @@ typedef union unsigned passZpassOp : 3; unsigned reserved : 3; }ni; - u_int32_t ui; + uint32_t ui; } savageRegStencilCtrl; /************************** @@ -85,7 +85,7 @@ typedef union unsigned palSize : 2; unsigned newPal : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegTexDescr_s4; typedef union { @@ -102,7 +102,7 @@ typedef union unsigned reserved3 : 10; unsigned newPal : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegTexDescr_s3d; /* The layout of this reg is the same on Savage4 and Savage3D, @@ -116,7 +116,7 @@ typedef union unsigned reserved : 1; unsigned addr : 29; }ni; - u_int32_t ui; + uint32_t ui; } savageRegTexAddr; /* The layout of this reg is the same on Savage4 and Savage3D. */ @@ -127,7 +127,7 @@ typedef union unsigned reserved : 3; unsigned addr : 29; }ni; - u_int32_t ui; + uint32_t ui; } savageRegTexPalAddr; /* The layout of this reg on Savage4 and Savage3D are very similar. */ @@ -138,7 +138,7 @@ typedef union unsigned xprClr0 : 16; unsigned xprClr1 : 16; /* this is reserved on Savage3D */ }ni; - u_int32_t ui; + uint32_t ui; } savageRegTexXprClr; /* transparency color in RGB565 format*/ /* The layout of this reg differs between Savage4 and Savage3D. @@ -164,7 +164,7 @@ typedef union unsigned alphaArg1Invert : 1; unsigned alphaArg2Invert : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegTexCtrl_s4; typedef union { @@ -187,7 +187,7 @@ typedef union unsigned texXprEn : 1; unsigned reserved2 : 11; }ni; - u_int32_t ui; + uint32_t ui; } savageRegTexCtrl_s3d; /* This reg exists only on Savage4. */ @@ -218,7 +218,7 @@ typedef union unsigned colorDoDiffMul : 1; unsigned LeftShiftVal : 2; }ni; - u_int32_t ui; + uint32_t ui; } savageRegTexBlendCtrl; /* This reg exists only on Savage4. */ @@ -231,7 +231,7 @@ typedef union unsigned red : 8; unsigned alpha : 8; }ni; - u_int32_t ui; + uint32_t ui; } savageRegTexBlendColor; /******************************** @@ -247,7 +247,7 @@ typedef union unsigned widthInTile : 6; unsigned bitPerPixel : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegTiledSurface; /******************************** @@ -264,7 +264,7 @@ typedef union unsigned scissorYStart : 12; unsigned alphaRefVal : 8; }ni; - u_int32_t ui; + uint32_t ui; } savageRegDrawCtrl0; /* This reg exists only on Savage4. */ @@ -281,7 +281,7 @@ typedef union unsigned alphaTestCmpFunc : 3; unsigned alphaTestEn : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegDrawCtrl1; /* This reg exists only on Savage4. */ @@ -312,7 +312,7 @@ typedef union unsigned flushPdDestWrites : 1; unsigned flushPdZbufWrites : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegDrawLocalCtrl; /* This reg exists only on Savage3D. */ @@ -342,7 +342,7 @@ typedef union */ unsigned interpMode : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegDrawCtrl; #define SAVAGETBC_DECAL_S3D 0 @@ -364,7 +364,7 @@ typedef union unsigned scissorYStart : 11; unsigned reserved2 : 5; } ni; - u_int32_t ui; + uint32_t ui; } savageRegScissorsStart; /* This reg exists only on Savage3D. */ @@ -377,7 +377,7 @@ typedef union unsigned scissorYEnd : 11; unsigned reserved2 : 5; } ni; - u_int32_t ui; + uint32_t ui; } savageRegScissorsEnd; /******************************** @@ -396,7 +396,7 @@ typedef union unsigned reserved : 1; unsigned addr : 29; /*quad word aligned*/ }ni; - u_int32_t ui; + uint32_t ui; } savageRegVertBufAddr; /* I havn't found a Savage3D equivalent of this reg in the Utah-driver. @@ -411,7 +411,7 @@ typedef union unsigned reserved : 1; unsigned addr : 29; /*4-quad word aligned*/ }ni; - u_int32_t ui; + uint32_t ui; } savageRegDMABufAddr; /******************************** @@ -439,7 +439,7 @@ typedef union unsigned reserved : 17; unsigned kickOff : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegFlag; /******************************** @@ -464,7 +464,7 @@ typedef union unsigned floatZEn : 1; unsigned wToZEn : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegZBufCtrl_s4; typedef union { @@ -486,7 +486,7 @@ typedef union unsigned wrZafterAlphaTst : 1; unsigned reserved2 : 15; }ni; - u_int32_t ui; + uint32_t ui; } savageRegZBufCtrl_s3d; /* The layout of this reg on Savage4 and Savage3D is very similar. */ @@ -507,7 +507,7 @@ typedef union */ unsigned zDepthSelect : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegZBufOffset; /* The layout of this reg is the same on Savage4 and Savage3D. */ @@ -524,7 +524,7 @@ typedef union unsigned wHigh : 6; unsigned reserved4 : 2; }ni; - u_int32_t ui; + uint32_t ui; } savageRegZWatermarks; /******************************** @@ -542,7 +542,7 @@ typedef union unsigned fogMode : 1; unsigned fogEndShift : 2; }ni; - u_int32_t ui; + uint32_t ui; } savageRegFogCtrl; /*not in spec, but tempo for pp and driver*/ @@ -553,7 +553,7 @@ typedef union unsigned fogDensity : 16; unsigned fogStart : 16; }ni; - u_int32_t ui; + uint32_t ui; } savageRegFogParam; /************************************** @@ -577,7 +577,7 @@ typedef union unsigned antiAliasMode : 2; unsigned dstPixFmt : 1; }ni; - u_int32_t ui; + uint32_t ui; } savageRegDestCtrl; /* The layout of this reg on Savage4 and Savage3D are very similar. */ @@ -596,7 +596,7 @@ typedef union * However, it is not used in either driver. */ unsigned destFlush : 2; }ni; - u_int32_t ui; + uint32_t ui; } savageRegDestTexWatermarks; /* Savage4/Twister/ProSavage register BCI addresses */ @@ -641,7 +641,7 @@ typedef union #define SAVAGE_FIRST_REG 0x18 #define SAVAGE_NR_REGS 34 typedef struct savage_registers_s4_t { - u_int32_t unused1[6]; /* 0x18-0x1d */ + uint32_t unused1[6]; /* 0x18-0x1d */ savageRegDrawLocalCtrl drawLocalCtrl; /* 0x1e */ savageRegTexPalAddr texPalAddr; /* 0x1f */ savageRegTexCtrl_s4 texCtrl[2]; /* 0x20, 0x21 */ @@ -649,7 +649,7 @@ typedef struct savage_registers_s4_t { savageRegTexBlendCtrl texBlendCtrl[2]; /* 0x24, 0x25 */ savageRegTexXprClr texXprClr; /* 0x26 */ savageRegTexDescr_s4 texDescr; /* 0x27 */ - u_int8_t fogTable[32]; /* 0x28-0x2f (8dwords) */ + uint8_t fogTable[32]; /* 0x28-0x2f (8dwords) */ savageRegFogCtrl fogCtrl; /* 0x30 */ savageRegStencilCtrl stencilCtrl; /* 0x31 */ savageRegZBufCtrl_s4 zBufCtrl; /* 0x32 */ @@ -667,8 +667,8 @@ typedef struct savage_registers_s3d_t { savageRegTexAddr texAddr; /* 0x1a */ savageRegTexDescr_s3d texDescr; /* 0x1b */ savageRegTexCtrl_s3d texCtrl; /* 0x1c */ - u_int32_t unused1[3]; /* 0x1d-0x1f */ - u_int8_t fogTable[64]; /* 0x20-0x2f (16dwords) */ + uint32_t unused1[3]; /* 0x1d-0x1f */ + uint8_t fogTable[64]; /* 0x20-0x2f (16dwords) */ savageRegFogCtrl fogCtrl; /* 0x30 */ savageRegDrawCtrl drawCtrl; /* 0x31 */ savageRegZBufCtrl_s3d zBufCtrl; /* 0x32 */ @@ -678,12 +678,12 @@ typedef struct savage_registers_s3d_t { savageRegScissorsEnd scissorsEnd; /* 0x36 */ savageRegZWatermarks zWatermarks; /* 0x37 */ savageRegDestTexWatermarks destTexWatermarks; /* 0x38 */ - u_int32_t unused2; /* 0x39 */ + uint32_t unused2; /* 0x39 */ } savageRegistersS3D; typedef union savage_registers_t { savageRegistersS4 s4; savageRegistersS3D s3d; - u_int32_t ui[SAVAGE_NR_REGS]; + uint32_t ui[SAVAGE_NR_REGS]; } savageRegisters; diff --git a/src/mesa/drivers/dri/savage/savage_xmesa.c b/src/mesa/drivers/dri/savage/savage_xmesa.c index 43422db9a8..0017125329 100644 --- a/src/mesa/drivers/dri/savage/savage_xmesa.c +++ b/src/mesa/drivers/dri/savage/savage_xmesa.c @@ -487,7 +487,7 @@ savageCreateContext( const __GLcontextModes *mesaVis, imesa->clientVtxBuf.total = imesa->bufferSize / 4; imesa->clientVtxBuf.used = 0; imesa->clientVtxBuf.flushed = 0; - imesa->clientVtxBuf.buf = (u_int32_t *)malloc(imesa->bufferSize); + imesa->clientVtxBuf.buf = (uint32_t *)malloc(imesa->bufferSize); imesa->vtxBuf = &imesa->clientVtxBuf; @@ -956,8 +956,8 @@ savageFillInModes( unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML /*, GLX_SWAP_COPY_OML */ }; - u_int8_t depth_bits_array[2]; - u_int8_t stencil_bits_array[2]; + uint8_t depth_bits_array[2]; + uint8_t stencil_bits_array[2]; depth_bits_array[0] = depth_bits; diff --git a/src/mesa/drivers/dri/savage/savagecontext.h b/src/mesa/drivers/dri/savage/savagecontext.h index c4573b0da4..3fe3d71d4e 100644 --- a/src/mesa/drivers/dri/savage/savagecontext.h +++ b/src/mesa/drivers/dri/savage/savagecontext.h @@ -130,7 +130,7 @@ typedef void (*savage_point_func)( savageContextPtr, savageVertex * ); struct savage_vtxbuf_t { GLuint total, used, flushed; /* in 32 bit units */ GLuint idx; /* for DMA buffers */ - u_int32_t *buf; + uint32_t *buf; }; struct savage_cmdbuf_t { diff --git a/src/mesa/drivers/dri/savage/savageioctl.c b/src/mesa/drivers/dri/savage/savageioctl.c index 4eac1fb349..bea9a3ea53 100644 --- a/src/mesa/drivers/dri/savage/savageioctl.c +++ b/src/mesa/drivers/dri/savage/savageioctl.c @@ -112,7 +112,7 @@ void savageGetDMABuffer( savageContextPtr imesa ) imesa->dmaVtxBuf.used = 0; imesa->dmaVtxBuf.flushed = 0; imesa->dmaVtxBuf.idx = buf->idx; - imesa->dmaVtxBuf.buf = (u_int32_t *)buf->address; + imesa->dmaVtxBuf.buf = (uint32_t *)buf->address; if (SAVAGE_DEBUG & DEBUG_DMA) fprintf(stderr, "finished getbuffer\n"); @@ -137,7 +137,7 @@ static void savage_BCI_clear(GLcontext *ctx, drm_savage_clear_t *pclear) unsigned int y = pbox->y1; unsigned int width = pbox->x2 - x; unsigned int height = pbox->y2 - y; - u_int32_t *bciptr; + uint32_t *bciptr; if (pbox->x1 > pbox->x2 || pbox->y1 > pbox->y2 || @@ -147,27 +147,27 @@ static void savage_BCI_clear(GLcontext *ctx, drm_savage_clear_t *pclear) if ( pclear->flags & SAVAGE_FRONT ) { bciptr = savageDMAAlloc (imesa, 8); - WRITE_CMD((bciptr) , 0x4BCC8C00,u_int32_t); - WRITE_CMD((bciptr) , imesa->savageScreen->frontOffset,u_int32_t); - WRITE_CMD((bciptr) , imesa->savageScreen->frontBitmapDesc,u_int32_t); - WRITE_CMD((bciptr) , pclear->clear_color,u_int32_t); - WRITE_CMD((bciptr) , (y <<16) | x,u_int32_t); - WRITE_CMD((bciptr) , (height << 16) | width,u_int32_t); + WRITE_CMD((bciptr) , 0x4BCC8C00,uint32_t); + WRITE_CMD((bciptr) , imesa->savageScreen->frontOffset,uint32_t); + WRITE_CMD((bciptr) , imesa->savageScreen->frontBitmapDesc,uint32_t); + WRITE_CMD((bciptr) , pclear->clear_color,uint32_t); + WRITE_CMD((bciptr) , (y <<16) | x,uint32_t); + WRITE_CMD((bciptr) , (height << 16) | width,uint32_t); savageDMACommit (imesa, bciptr); } if ( pclear->flags & SAVAGE_BACK ) { bciptr = savageDMAAlloc (imesa, 8); - WRITE_CMD((bciptr) , 0x4BCC8C00,u_int32_t); - WRITE_CMD((bciptr) , imesa->savageScreen->backOffset,u_int32_t); - WRITE_CMD((bciptr) , imesa->savageScreen->backBitmapDesc,u_int32_t); - WRITE_CMD((bciptr) , pclear->clear_color,u_int32_t); - WRITE_CMD((bciptr) , (y <<16) | x,u_int32_t); - WRITE_CMD((bciptr) , (height << 16) | width,u_int32_t); + WRITE_CMD((bciptr) , 0x4BCC8C00,uint32_t); + WRITE_CMD((bciptr) , imesa->savageScreen->backOffset,uint32_t); + WRITE_CMD((bciptr) , imesa->savageScreen->backBitmapDesc,uint32_t); + WRITE_CMD((bciptr) , pclear->clear_color,uint32_t); + WRITE_CMD((bciptr) , (y <<16) | x,uint32_t); + WRITE_CMD((bciptr) , (height << 16) | width,uint32_t); savageDMACommit (imesa, bciptr); } if ( pclear->flags & (SAVAGE_DEPTH |SAVAGE_STENCIL) ) { - u_int32_t writeMask = 0x0; + uint32_t writeMask = 0x0; if(imesa->hw_stencil) { if(pclear->flags & SAVAGE_STENCIL) @@ -199,8 +199,8 @@ static void savage_BCI_clear(GLcontext *ctx, drm_savage_clear_t *pclear) bciptr = savageDMAAlloc (imesa, 10); if(writeMask != 0xFFFFFFFF) { - WRITE_CMD((bciptr) , 0x960100D7,u_int32_t); - WRITE_CMD((bciptr) , writeMask,u_int32_t); + WRITE_CMD((bciptr) , 0x960100D7,uint32_t); + WRITE_CMD((bciptr) , writeMask,uint32_t); } } else @@ -208,18 +208,18 @@ static void savage_BCI_clear(GLcontext *ctx, drm_savage_clear_t *pclear) bciptr = savageDMAAlloc (imesa, 6); } - WRITE_CMD((bciptr) , 0x4BCC8C00,u_int32_t); - WRITE_CMD((bciptr) , imesa->savageScreen->depthOffset,u_int32_t); - WRITE_CMD((bciptr) , imesa->savageScreen->depthBitmapDesc,u_int32_t); - WRITE_CMD((bciptr) , pclear->clear_depth,u_int32_t); - WRITE_CMD((bciptr) , (y <<16) | x,u_int32_t); - WRITE_CMD((bciptr) , (height << 16) | width,u_int32_t); + WRITE_CMD((bciptr) , 0x4BCC8C00,uint32_t); + WRITE_CMD((bciptr) , imesa->savageScreen->depthOffset,uint32_t); + WRITE_CMD((bciptr) , imesa->savageScreen->depthBitmapDesc,uint32_t); + WRITE_CMD((bciptr) , pclear->clear_depth,uint32_t); + WRITE_CMD((bciptr) , (y <<16) | x,uint32_t); + WRITE_CMD((bciptr) , (height << 16) | width,uint32_t); if(imesa->hw_stencil) { if(writeMask != 0xFFFFFFFF) { - WRITE_CMD((bciptr) , 0x960100D7,u_int32_t); - WRITE_CMD((bciptr) , 0xFFFFFFFF,u_int32_t); + WRITE_CMD((bciptr) , 0x960100D7,uint32_t); + WRITE_CMD((bciptr) , 0xFFFFFFFF,uint32_t); } } savageDMACommit (imesa, bciptr); @@ -236,7 +236,7 @@ static void savage_BCI_swap(savageContextPtr imesa) int nbox = imesa->sarea->nbox; drm_clip_rect_t *pbox = imesa->sarea->boxes; int i; - volatile u_int32_t *bciptr; + volatile uint32_t *bciptr; if (nbox > SAVAGE_NR_SAREA_CLIPRECTS) nbox = SAVAGE_NR_SAREA_CLIPRECTS; diff --git a/src/mesa/drivers/dri/savage/savageioctl.h b/src/mesa/drivers/dri/savage/savageioctl.h index c7ff0017fd..98629048ae 100644 --- a/src/mesa/drivers/dri/savage/savageioctl.h +++ b/src/mesa/drivers/dri/savage/savageioctl.h @@ -77,10 +77,10 @@ GLboolean savageHaveIndexedVerts( savageContextPtr imesa ) } static __inline -u_int32_t *savageAllocVtxBuf( savageContextPtr imesa, GLuint words ) +uint32_t *savageAllocVtxBuf( savageContextPtr imesa, GLuint words ) { struct savage_vtxbuf_t *buffer = imesa->vtxBuf; - u_int32_t *head; + uint32_t *head; if (buffer == &imesa->dmaVtxBuf) { if (!buffer->total) { @@ -116,9 +116,9 @@ u_int32_t *savageAllocVtxBuf( savageContextPtr imesa, GLuint words ) } static __inline -u_int32_t *savageAllocIndexedVerts( savageContextPtr imesa, GLuint n ) +uint32_t *savageAllocIndexedVerts( savageContextPtr imesa, GLuint n ) { - u_int32_t *ret; + uint32_t *ret; savageFlushVertices(imesa); ret = savageAllocVtxBuf(imesa, n*imesa->HwVertexSize); imesa->firstElt = imesa->vtxBuf->flushed / imesa->HwVertexSize; @@ -172,9 +172,9 @@ drm_savage_cmd_header_t *savageAllocCmdBuf( savageContextPtr imesa, GLuint bytes * - increments the number of elts. Final allocation is done in savageFlushElts */ static __inline -u_int16_t *savageAllocElts( savageContextPtr imesa, GLuint n ) +uint16_t *savageAllocElts( savageContextPtr imesa, GLuint n ) { - u_int16_t *ret; + uint16_t *ret; GLuint qwords; assert (savageHaveIndexedVerts(imesa)); @@ -195,7 +195,7 @@ u_int16_t *savageAllocElts( savageContextPtr imesa, GLuint n ) imesa->elts.n = 0; } - ret = (u_int16_t *)(imesa->elts.cmd+1) + imesa->elts.n; + ret = (uint16_t *)(imesa->elts.cmd+1) + imesa->elts.n; imesa->elts.n += n; return ret; } diff --git a/src/mesa/drivers/dri/savage/savagestate.c b/src/mesa/drivers/dri/savage/savagestate.c index 741a9dda4f..84fd3157ed 100644 --- a/src/mesa/drivers/dri/savage/savagestate.c +++ b/src/mesa/drivers/dri/savage/savagestate.c @@ -122,9 +122,9 @@ static void savageDDBlendEquationSeparate(GLcontext *ctx, static void savageBlendFunc_s4(GLcontext *ctx) { savageContextPtr imesa = SAVAGE_CONTEXT(ctx); - u_int32_t drawLocalCtrl = imesa->regs.s4.drawLocalCtrl.ui; - u_int32_t drawCtrl0 = imesa->regs.s4.drawCtrl0.ui; - u_int32_t drawCtrl1 = imesa->regs.s4.drawCtrl1.ui; + uint32_t drawLocalCtrl = imesa->regs.s4.drawLocalCtrl.ui; + uint32_t drawCtrl0 = imesa->regs.s4.drawCtrl0.ui; + uint32_t drawCtrl1 = imesa->regs.s4.drawCtrl1.ui; /* set up draw control register (including blending, alpha * test, and shading model) @@ -297,8 +297,8 @@ static void savageBlendFunc_s4(GLcontext *ctx) static void savageBlendFunc_s3d(GLcontext *ctx) { savageContextPtr imesa = SAVAGE_CONTEXT(ctx); - u_int32_t drawCtrl = imesa->regs.s3d.drawCtrl.ui; - u_int32_t zBufCtrl = imesa->regs.s3d.zBufCtrl.ui; + uint32_t drawCtrl = imesa->regs.s3d.drawCtrl.ui; + uint32_t zBufCtrl = imesa->regs.s3d.zBufCtrl.ui; /* set up draw control register (including blending, alpha * test, dithering, and shading model) @@ -486,9 +486,9 @@ static void savageDDDepthFunc_s4(GLcontext *ctx, GLenum func) { savageContextPtr imesa = SAVAGE_CONTEXT(ctx); ZCmpFunc zmode; - u_int32_t drawLocalCtrl = imesa->regs.s4.drawLocalCtrl.ui; - u_int32_t zBufCtrl = imesa->regs.s4.zBufCtrl.ui; - u_int32_t zWatermarks = imesa->regs.s4.zWatermarks.ui; /* FIXME: in DRM */ + uint32_t drawLocalCtrl = imesa->regs.s4.drawLocalCtrl.ui; + uint32_t zBufCtrl = imesa->regs.s4.zBufCtrl.ui; + uint32_t zWatermarks = imesa->regs.s4.zWatermarks.ui; /* FIXME: in DRM */ /* set up z-buffer control register (global) * set up z-buffer offset register (global) @@ -550,9 +550,9 @@ static void savageDDDepthFunc_s3d(GLcontext *ctx, GLenum func) { savageContextPtr imesa = SAVAGE_CONTEXT(ctx); ZCmpFunc zmode; - u_int32_t drawCtrl = imesa->regs.s3d.drawCtrl.ui; - u_int32_t zBufCtrl = imesa->regs.s3d.zBufCtrl.ui; - u_int32_t zWatermarks = imesa->regs.s3d.zWatermarks.ui; /* FIXME: in DRM */ + uint32_t drawCtrl = imesa->regs.s3d.drawCtrl.ui; + uint32_t zBufCtrl = imesa->regs.s3d.zBufCtrl.ui; + uint32_t zWatermarks = imesa->regs.s3d.zWatermarks.ui; /* FIXME: in DRM */ /* set up z-buffer control register (global) * set up z-buffer offset register (global) @@ -638,7 +638,7 @@ static void savageDDScissor( GLcontext *ctx, GLint x, GLint y, static void savageDDDrawBuffer(GLcontext *ctx, GLenum mode ) { savageContextPtr imesa = SAVAGE_CONTEXT(ctx); - u_int32_t destCtrl = imesa->regs.s4.destCtrl.ui; + uint32_t destCtrl = imesa->regs.s4.destCtrl.ui; /* * _DrawDestMask is easier to cope with than . @@ -865,7 +865,7 @@ static void savageDDColorMask_s3d(GLcontext *ctx, static void savageUpdateSpecular_s4(GLcontext *ctx) { savageContextPtr imesa = SAVAGE_CONTEXT( ctx ); - u_int32_t drawLocalCtrl = imesa->regs.s4.drawLocalCtrl.ui; + uint32_t drawLocalCtrl = imesa->regs.s4.drawLocalCtrl.ui; if (NEED_SECONDARY_COLOR(ctx)) { imesa->regs.s4.drawLocalCtrl.ni.specShadeEn = GL_TRUE; @@ -879,7 +879,7 @@ static void savageUpdateSpecular_s4(GLcontext *ctx) { static void savageUpdateSpecular_s3d(GLcontext *ctx) { savageContextPtr imesa = SAVAGE_CONTEXT( ctx ); - u_int32_t drawCtrl = imesa->regs.s3d.drawCtrl.ui; + uint32_t drawCtrl = imesa->regs.s3d.drawCtrl.ui; if (NEED_SECONDARY_COLOR(ctx)) { imesa->regs.s3d.drawCtrl.ni.specShadeEn = GL_TRUE; @@ -905,7 +905,7 @@ static void savageDDLightModelfv_s3d(GLcontext *ctx, GLenum pname, static void savageDDShadeModel_s4(GLcontext *ctx, GLuint mod) { savageContextPtr imesa = SAVAGE_CONTEXT( ctx ); - u_int32_t drawLocalCtrl = imesa->regs.s4.drawLocalCtrl.ui; + uint32_t drawLocalCtrl = imesa->regs.s4.drawLocalCtrl.ui; if (mod == GL_SMOOTH) { @@ -922,7 +922,7 @@ static void savageDDShadeModel_s4(GLcontext *ctx, GLuint mod) static void savageDDShadeModel_s3d(GLcontext *ctx, GLuint mod) { savageContextPtr imesa = SAVAGE_CONTEXT( ctx ); - u_int32_t drawCtrl = imesa->regs.s3d.drawCtrl.ui; + uint32_t drawCtrl = imesa->regs.s3d.drawCtrl.ui; if (mod == GL_SMOOTH) { @@ -948,7 +948,7 @@ static void savageDDFogfv(GLcontext *ctx, GLenum pname, const GLfloat *param) { savageContextPtr imesa = SAVAGE_CONTEXT(ctx); GLuint fogClr; - u_int32_t fogCtrl = imesa->regs.s4.fogCtrl.ui; + uint32_t fogCtrl = imesa->regs.s4.fogCtrl.ui; /*if ((ctx->Fog.Enabled) &&(pname == GL_FOG_COLOR))*/ if (ctx->Fog.Enabled) @@ -980,8 +980,8 @@ savageDDStencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func, { savageContextPtr imesa = SAVAGE_CONTEXT(ctx); unsigned a=0; - const u_int32_t zBufCtrl = imesa->regs.s4.zBufCtrl.ui; - const u_int32_t stencilCtrl = imesa->regs.s4.stencilCtrl.ui; + const uint32_t zBufCtrl = imesa->regs.s4.zBufCtrl.ui; + const uint32_t stencilCtrl = imesa->regs.s4.stencilCtrl.ui; imesa->regs.s4.zBufCtrl.ni.stencilRefVal = ctx->Stencil.Ref[0] & 0xff; imesa->regs.s4.stencilCtrl.ni.readMask = ctx->Stencil.ValueMask[0] & 0xff; @@ -1041,7 +1041,7 @@ savageDDStencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail, GLenum zfail, GLenum zpass) { savageContextPtr imesa = SAVAGE_CONTEXT(ctx); - const u_int32_t stencilCtrl = imesa->regs.s4.stencilCtrl.ui; + const uint32_t stencilCtrl = imesa->regs.s4.stencilCtrl.ui; imesa->regs.s4.stencilCtrl.ni.failOp = get_stencil_op_value( ctx->Stencil.FailFunc[0] ); imesa->regs.s4.stencilCtrl.ni.passZfailOp = get_stencil_op_value( ctx->Stencil.ZFailFunc[0] ); @@ -1613,8 +1613,8 @@ static void savageDDInitState_s3d( savageContextPtr imesa ) imesa->globalRegMask.s3d.zBufCtrl.ni.zBufEn = 0x1; } void savageDDInitState( savageContextPtr imesa ) { - memset (imesa->regs.ui, 0, SAVAGE_NR_REGS*sizeof(u_int32_t)); - memset (imesa->globalRegMask.ui, 0xff, SAVAGE_NR_REGS*sizeof(u_int32_t)); + memset (imesa->regs.ui, 0, SAVAGE_NR_REGS*sizeof(uint32_t)); + memset (imesa->globalRegMask.ui, 0xff, SAVAGE_NR_REGS*sizeof(uint32_t)); if (imesa->savageScreen->chipset >= S3_SAVAGE4) savageDDInitState_s4 (imesa); else @@ -1656,7 +1656,7 @@ void savageDDInitState( savageContextPtr imesa ) { imesa->regs.s4.zBufOffset.ni.zDepthSelect = 1; } - memcpy (imesa->oldRegs.ui, imesa->regs.ui, SAVAGE_NR_REGS*sizeof(u_int32_t)); + memcpy (imesa->oldRegs.ui, imesa->regs.ui, SAVAGE_NR_REGS*sizeof(uint32_t)); /* Emit the initial state to the (empty) command buffer. */ assert (imesa->cmdBuf.write == imesa->cmdBuf.base); diff --git a/src/mesa/drivers/dri/savage/savagetex.c b/src/mesa/drivers/dri/savage/savagetex.c index 719e50f964..dbe30d4c94 100644 --- a/src/mesa/drivers/dri/savage/savagetex.c +++ b/src/mesa/drivers/dri/savage/savagetex.c @@ -1365,7 +1365,7 @@ static void savageUpdateTex0State_s4( GLcontext *ctx ) if (imesa->regs.s4.texDescr.ni.tex1En) imesa->regs.s4.texDescr.ni.texBLoopEn = GL_TRUE; - imesa->regs.s4.texAddr[0].ui = (u_int32_t) t->setup.physAddr | 0x2; + imesa->regs.s4.texAddr[0].ui = (uint32_t) t->setup.physAddr | 0x2; if(t->base.heap->heapId == SAVAGE_AGP_HEAP) imesa->regs.s4.texAddr[0].ui |= 0x1; @@ -1550,7 +1550,7 @@ static void savageUpdateTex1State_s4( GLcontext *ctx ) imesa->regs.s4.texCtrl[1].ni.dMax = t->base.lastLevel - t->base.firstLevel; imesa->regs.s4.texDescr.ni.texBLoopEn = GL_TRUE; - imesa->regs.s4.texAddr[1].ui = (u_int32_t) t->setup.physAddr | 2; + imesa->regs.s4.texAddr[1].ui = (uint32_t) t->setup.physAddr | 2; if(t->base.heap->heapId == SAVAGE_AGP_HEAP) imesa->regs.s4.texAddr[1].ui |= 0x1; } @@ -1701,7 +1701,7 @@ static void savageUpdateTexState_s3d( GLcontext *ctx ) assert (t->hwFormat <= 7); imesa->regs.s3d.texDescr.ni.texFmt = t->hwFormat; - imesa->regs.s3d.texAddr.ui = (u_int32_t) t->setup.physAddr | 2; + imesa->regs.s3d.texAddr.ui = (uint32_t) t->setup.physAddr | 2; if(t->base.heap->heapId == SAVAGE_AGP_HEAP) imesa->regs.s3d.texAddr.ui |= 0x1; } diff --git a/src/mesa/drivers/dri/savage/savagetris.c b/src/mesa/drivers/dri/savage/savagetris.c index 52c7f5fa76..a1c7bb167e 100644 --- a/src/mesa/drivers/dri/savage/savagetris.c +++ b/src/mesa/drivers/dri/savage/savagetris.c @@ -100,7 +100,7 @@ static void __inline__ savage_draw_triangle (savageContextPtr imesa, savageVertexPtr v1, savageVertexPtr v2) { GLuint vertsize = imesa->HwVertexSize; - u_int32_t *vb = savageAllocVtxBuf (imesa, 3*vertsize); + uint32_t *vb = savageAllocVtxBuf (imesa, 3*vertsize); GLuint j; EMIT_VERT (j, vb, vertsize, 0, v0); @@ -114,7 +114,7 @@ static void __inline__ savage_draw_quad (savageContextPtr imesa, savageVertexPtr v2, savageVertexPtr v3) { GLuint vertsize = imesa->HwVertexSize; - u_int32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); + uint32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); GLuint j; EMIT_VERT (j, vb, vertsize, 0, v0); @@ -128,7 +128,7 @@ static void __inline__ savage_draw_quad (savageContextPtr imesa, static __inline__ void savage_draw_point (savageContextPtr imesa, savageVertexPtr tmp) { GLuint vertsize = imesa->HwVertexSize; - u_int32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); + uint32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); const GLfloat x = tmp->v.x; const GLfloat y = tmp->v.y; const GLfloat sz = 0.5 * CLAMP(imesa->glCtx->Point.Size, @@ -165,7 +165,7 @@ static __inline__ void savage_draw_line (savageContextPtr imesa, savageVertexPtr v0, savageVertexPtr v1 ) { GLuint vertsize = imesa->HwVertexSize; - u_int32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); + uint32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); const GLfloat width = CLAMP(imesa->glCtx->Line.Width, imesa->glCtx->Const.MinLineWidth, imesa->glCtx->Const.MaxLineWidth); @@ -224,7 +224,7 @@ static void __inline__ savage_ptex_tri (savageContextPtr imesa, savageVertexPtr v1, savageVertexPtr v2) { GLuint vertsize = imesa->HwVertexSize; - u_int32_t *vb = savageAllocVtxBuf (imesa, 3*vertsize); + uint32_t *vb = savageAllocVtxBuf (imesa, 3*vertsize); savageVertex tmp; GLuint j; @@ -237,7 +237,7 @@ static __inline__ void savage_ptex_line (savageContextPtr imesa, savageVertexPtr v0, savageVertexPtr v1 ) { GLuint vertsize = imesa->HwVertexSize; - u_int32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); + uint32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); const GLfloat width = CLAMP(imesa->glCtx->Line.Width, imesa->glCtx->Const.MinLineWidth, imesa->glCtx->Const.MaxLineWidth); @@ -284,7 +284,7 @@ static __inline__ void savage_ptex_line (savageContextPtr imesa, static __inline__ void savage_ptex_point (savageContextPtr imesa, savageVertexPtr v0) { GLuint vertsize = imesa->HwVertexSize; - u_int32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); + uint32_t *vb = savageAllocVtxBuf (imesa, 6*vertsize); const GLfloat x = v0->v.x; const GLfloat y = v0->v.y; const GLfloat sz = 0.5 * CLAMP(imesa->glCtx->Point.Size, diff --git a/src/mesa/drivers/dri/sis/sis_screen.c b/src/mesa/drivers/dri/sis/sis_screen.c index d90482f3d7..f55027e094 100644 --- a/src/mesa/drivers/dri/sis/sis_screen.c +++ b/src/mesa/drivers/dri/sis/sis_screen.c @@ -77,8 +77,8 @@ sisFillInModes(int bpp) static const GLenum back_buffer_modes[] = { GLX_NONE, GLX_SWAP_UNDEFINED_OML }; - u_int8_t depth_bits_array[4]; - u_int8_t stencil_bits_array[4]; + uint8_t depth_bits_array[4]; + uint8_t stencil_bits_array[4]; depth_bits_array[0] = 0; stencil_bits_array[0] = 0; diff --git a/src/mesa/drivers/dri/unichrome/server/via_dri.c b/src/mesa/drivers/dri/unichrome/server/via_dri.c index 9833145940..74034485e2 100644 --- a/src/mesa/drivers/dri/unichrome/server/via_dri.c +++ b/src/mesa/drivers/dri/unichrome/server/via_dri.c @@ -52,9 +52,9 @@ static int VIADRIFinishScreenInit(DRIDriverContext * ctx); /* _SOLO : missing macros normally defined by X code */ #define xf86DrvMsg(a, b, ...) fprintf(stderr, __VA_ARGS__) -#define MMIO_IN8(base, addr) ((*(((volatile u_int8_t*)base)+(addr)))+0) -#define MMIO_OUT8(base, addr, val) ((*(((volatile u_int8_t*)base)+(addr)))=((u_int8_t)val)) -#define MMIO_OUT16(base, addr, val) ((*(volatile u_int16_t*)(((u_int8_t*)base)+(addr)))=((u_int16_t)val)) +#define MMIO_IN8(base, addr) ((*(((volatile uint8_t*)base)+(addr)))+0) +#define MMIO_OUT8(base, addr, val) ((*(((volatile uint8_t*)base)+(addr)))=((uint8_t)val)) +#define MMIO_OUT16(base, addr, val) ((*(volatile uint16_t*)(((uint8_t*)base)+(addr)))=((uint16_t)val)) #define VIDEO 0 #define AGP 1 @@ -523,9 +523,9 @@ static int VIADRIKernelInit(DRIDriverContext * ctx, VIAPtr pVia) drmInfo.fb_offset = pVia->FrameBufferBase; drmInfo.mmio_offset = pVia->registerHandle; if (pVia->IsPCI) - drmInfo.agpAddr = (u_int32_t)NULL; + drmInfo.agpAddr = (uint32_t)NULL; else - drmInfo.agpAddr = (u_int32_t)pVia->agpAddr; + drmInfo.agpAddr = (uint32_t)pVia->agpAddr; if ((drmCommandWrite(pVia->drmFD, DRM_VIA_MAP_INIT,&drmInfo, sizeof(drm_via_init_t))) < 0) @@ -631,7 +631,7 @@ static void VIADisableMMIO(DRIDriverContext * ctx) static void VIADisableExtendedFIFO(DRIDriverContext *ctx) { VIAPtr pVia = VIAPTR(ctx); - u_int32_t dwGE230, dwGE298; + uint32_t dwGE230, dwGE298; /* Cause of exit XWindow will dump back register value, others chipset no * need to set extended fifo value */ @@ -653,8 +653,8 @@ static void VIADisableExtendedFIFO(DRIDriverContext *ctx) static void VIAEnableExtendedFIFO(DRIDriverContext *ctx) { VIAPtr pVia = VIAPTR(ctx); - u_int8_t bRegTemp; - u_int32_t dwGE230, dwGE298; + uint8_t bRegTemp; + uint32_t dwGE230, dwGE298; switch (pVia->Chipset) { case VIA_CLE266: @@ -849,7 +849,7 @@ static void VIAEnableExtendedFIFO(DRIDriverContext *ctx) SR1C[7:0], SR1D[1:0] (10bits) *=*/ wRegTemp = (pBIOSInfo->offsetWidthByQWord >> 1) + 4; VGAOUT8(0x3c4, 0x1c); - VGAOUT8(0x3c5, (u_int8_t)(wRegTemp & 0xFF)); + VGAOUT8(0x3c5, (uint8_t)(wRegTemp & 0xFF)); VGAOUT8(0x3c4, 0x1d); bRegTemp = VGAIN8(0x3c5) & ~0x03; VGAOUT8(0x3c5, bRegTemp | ((wRegTemp & 0x300) >> 8)); @@ -895,7 +895,7 @@ static void VIAEnableExtendedFIFO(DRIDriverContext *ctx) SR1C[7:0], SR1D[1:0] (10bits) *=*/ wRegTemp = (pBIOSInfo->offsetWidthByQWord >> 1) + 4; VGAOUT8(0x3c4, 0x1c); - VGAOUT8(0x3c5, (u_int8_t)(wRegTemp & 0xFF)); + VGAOUT8(0x3c5, (uint8_t)(wRegTemp & 0xFF)); VGAOUT8(0x3c4, 0x1d); bRegTemp = VGAIN8(0x3c5) & ~0x03; VGAOUT8(0x3c5, bRegTemp | ((wRegTemp & 0x300) >> 8)); @@ -923,9 +923,9 @@ static void VIAEnableExtendedFIFO(DRIDriverContext *ctx) static void VIAInitialize2DEngine(DRIDriverContext *ctx) { VIAPtr pVia = VIAPTR(ctx); - u_int32_t dwVQStartAddr, dwVQEndAddr; - u_int32_t dwVQLen, dwVQStartL, dwVQEndL, dwVQStartEndH; - u_int32_t dwGEMode; + uint32_t dwVQStartAddr, dwVQEndAddr; + uint32_t dwVQLen, dwVQStartL, dwVQEndL, dwVQStartEndH; + uint32_t dwGEMode; /* init 2D engine regs to reset 2D engine */ VIASETREG(0x04, 0x0); @@ -1068,14 +1068,14 @@ static void VIAInitialize3DEngine(DRIDriverContext *ctx) for (i = 0; i <= 0x7D; i++) { - VIASETREG(0x440, (u_int32_t) i << 24); + VIASETREG(0x440, (uint32_t) i << 24); } VIASETREG(0x43C, 0x00020000); for (i = 0; i <= 0x94; i++) { - VIASETREG(0x440, (u_int32_t) i << 24); + VIASETREG(0x440, (uint32_t) i << 24); } VIASETREG(0x440, 0x82400000); @@ -1085,7 +1085,7 @@ static void VIAInitialize3DEngine(DRIDriverContext *ctx) for (i = 0; i <= 0x94; i++) { - VIASETREG(0x440, (u_int32_t) i << 24); + VIASETREG(0x440, (uint32_t) i << 24); } VIASETREG(0x440, 0x82400000); @@ -1093,7 +1093,7 @@ static void VIAInitialize3DEngine(DRIDriverContext *ctx) for (i = 0; i <= 0x03; i++) { - VIASETREG(0x440, (u_int32_t) i << 24); + VIASETREG(0x440, (uint32_t) i << 24); } VIASETREG(0x43C, 0x00030000); diff --git a/src/mesa/drivers/dri/unichrome/server/via_driver.h b/src/mesa/drivers/dri/unichrome/server/via_driver.h index a643fd9fbb..cd3b038bfd 100644 --- a/src/mesa/drivers/dri/unichrome/server/via_driver.h +++ b/src/mesa/drivers/dri/unichrome/server/via_driver.h @@ -211,12 +211,12 @@ typedef struct _VIA { XAAInfoRecPtr AccelInfoRec; xRectangle Rect; #endif - u_int32_t SavedCmd; - u_int32_t SavedFgColor; - u_int32_t SavedBgColor; - u_int32_t SavedPattern0; - u_int32_t SavedPattern1; - u_int32_t SavedPatternAddr; + uint32_t SavedCmd; + uint32_t SavedFgColor; + uint32_t SavedBgColor; + uint32_t SavedPattern0; + uint32_t SavedPattern1; + uint32_t SavedPatternAddr; #if 0 /* Support for Int10 processing */ @@ -254,8 +254,8 @@ typedef struct _VIA { #endif /* - u_int32_t Cap0_Deinterlace; - u_int32_t Cap1_Deinterlace; + uint32_t Cap0_Deinterlace; + uint32_t Cap1_Deinterlace; int Cap0_FieldSwap; int NoCap0_HFilter; @@ -271,7 +271,7 @@ typedef struct _VIA { VIAConfigPrivPtr pVisualConfigsPriv; unsigned long agpHandle; unsigned long registerHandle; - u_int32_t agpAddr; + uint32_t agpAddr; unsigned char *agpBase; unsigned int agpSize; int IsPCI; @@ -287,11 +287,11 @@ typedef struct _VIA { #endif int V4LEnabled; - u_int16_t ActiveDevice; /* if SAMM, non-equal pBIOSInfo->ActiveDevice */ + uint16_t ActiveDevice; /* if SAMM, non-equal pBIOSInfo->ActiveDevice */ unsigned char *CursorImage; - u_int32_t CursorFG; - u_int32_t CursorBG; - u_int32_t CursorMC; + uint32_t CursorFG; + uint32_t CursorBG; + uint32_t CursorMC; unsigned char MemClk; int EnableExtendedFIFO; diff --git a/src/mesa/drivers/dri/unichrome/via_screen.c b/src/mesa/drivers/dri/unichrome/via_screen.c index 90f76be44d..b81549df0e 100644 --- a/src/mesa/drivers/dri/unichrome/via_screen.c +++ b/src/mesa/drivers/dri/unichrome/via_screen.c @@ -365,8 +365,8 @@ viaFillInModes( unsigned pixel_bits, GLboolean have_back_buffer ) /* The 32-bit depth-buffer mode isn't supported yet, so don't actually * enable it. */ - static const u_int8_t depth_bits_array[4] = { 0, 16, 24, 32 }; - static const u_int8_t stencil_bits_array[4] = { 0, 0, 8, 0 }; + static const uint8_t depth_bits_array[4] = { 0, 16, 24, 32 }; + static const uint8_t stencil_bits_array[4] = { 0, 0, 8, 0 }; const unsigned depth_buffer_factor = 3; diff --git a/src/mesa/drivers/dri/unichrome/via_state.c b/src/mesa/drivers/dri/unichrome/via_state.c index bccdbe9326..5b0453423e 100644 --- a/src/mesa/drivers/dri/unichrome/via_state.c +++ b/src/mesa/drivers/dri/unichrome/via_state.c @@ -773,10 +773,10 @@ void viaInitState(GLcontext *ctx) /** * Convert S and T texture coordinate wrap modes to hardware bits. */ -static u_int32_t +static uint32_t get_wrap_mode( GLenum sWrap, GLenum tWrap ) { - u_int32_t v = 0; + uint32_t v = 0; switch( sWrap ) { @@ -808,10 +808,10 @@ get_wrap_mode( GLenum sWrap, GLenum tWrap ) return v; } -static u_int32_t +static uint32_t get_minmag_filter( GLenum min, GLenum mag ) { - u_int32_t v = 0; + uint32_t v = 0; switch (min) { case GL_NEAREST: -- cgit v1.2.3 From 0d947674305a349662b3ab57733c54f9f4c1a8b3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 13:55:38 +0900 Subject: egl: Use standard integer types. --- src/egl/main/eglconfigutil.c | 18 +++++++++--------- src/egl/main/eglconfigutil.h | 2 +- 2 files changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglconfigutil.c b/src/egl/main/eglconfigutil.c index b6585619a5..db2ff1bfe8 100644 --- a/src/egl/main/eglconfigutil.c +++ b/src/egl/main/eglconfigutil.c @@ -118,13 +118,13 @@ _eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode) EGLBoolean _eglFillInConfigs(_EGLConfig * configs, GLenum fb_format, GLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + const uint8_t * depth_bits, const uint8_t * stencil_bits, unsigned num_depth_stencil_bits, const GLenum * db_modes, unsigned num_db_modes, int visType) { #if 0 - static const u_int8_t bits_table[3][4] = { + static const uint8_t bits_table[3][4] = { /* R G B A */ { 5, 6, 5, 0 }, /* Any GL_UNSIGNED_SHORT_5_6_5 */ { 8, 8, 8, 0 }, /* Any RGB with any GL_UNSIGNED_INT_8_8_8_8 */ @@ -135,7 +135,7 @@ _eglFillInConfigs(_EGLConfig * configs, * Given the four supported fb_type values, this results in valid array * indices of 3, 4, 5, and 7. */ - static const u_int32_t masks_table_rgb[8][4] = { + static const uint32_t masks_table_rgb[8][4] = { {0x00000000, 0x00000000, 0x00000000, 0x00000000}, {0x00000000, 0x00000000, 0x00000000, 0x00000000}, {0x00000000, 0x00000000, 0x00000000, 0x00000000}, @@ -146,7 +146,7 @@ _eglFillInConfigs(_EGLConfig * configs, {0x000000FF, 0x0000FF00, 0x00FF0000, 0x00000000} /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_rgba[8][4] = { + static const uint32_t masks_table_rgba[8][4] = { {0x00000000, 0x00000000, 0x00000000, 0x00000000}, {0x00000000, 0x00000000, 0x00000000, 0x00000000}, {0x00000000, 0x00000000, 0x00000000, 0x00000000}, @@ -157,7 +157,7 @@ _eglFillInConfigs(_EGLConfig * configs, {0x000000FF, 0x0000FF00, 0x00FF0000, 0xFF000000}, /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_bgr[8][4] = { + static const uint32_t masks_table_bgr[8][4] = { {0x00000000, 0x00000000, 0x00000000, 0x00000000}, {0x00000000, 0x00000000, 0x00000000, 0x00000000}, {0x00000000, 0x00000000, 0x00000000, 0x00000000}, @@ -168,7 +168,7 @@ _eglFillInConfigs(_EGLConfig * configs, {0x00FF0000, 0x0000FF00, 0x000000FF, 0x00000000}, /* 8_8_8_8_REV */ }; - static const u_int32_t masks_table_bgra[8][4] = { + static const uint32_t masks_table_bgra[8][4] = { {0x00000000, 0x00000000, 0x00000000, 0x00000000}, {0x00000000, 0x00000000, 0x00000000, 0x00000000}, {0x00000000, 0x00000000, 0x00000000, 0x00000000}, @@ -179,12 +179,12 @@ _eglFillInConfigs(_EGLConfig * configs, {0x00FF0000, 0x0000FF00, 0x000000FF, 0xFF000000}, /* 8_8_8_8_REV */ }; - static const u_int8_t bytes_per_pixel[8] = { + static const uint8_t bytes_per_pixel[8] = { 0, 0, 0, 2, 2, 4, 0, 4 }; - const u_int8_t * bits; - const u_int32_t * masks; + const uint8_t * bits; + const uint32_t * masks; const int index = fb_type & 0x07; _EGLConfig *config; unsigned i; diff --git a/src/egl/main/eglconfigutil.h b/src/egl/main/eglconfigutil.h index 5db906db65..dff407231f 100644 --- a/src/egl/main/eglconfigutil.h +++ b/src/egl/main/eglconfigutil.h @@ -13,7 +13,7 @@ _eglConfigToContextModesRec(const _EGLConfig *config, __GLcontextModes *mode); extern EGLBoolean _eglFillInConfigs( _EGLConfig *configs, EGLenum fb_format, EGLenum fb_type, - const u_int8_t * depth_bits, const u_int8_t * stencil_bits, + const uint8_t * depth_bits, const uint8_t * stencil_bits, unsigned num_depth_stencil_bits, const EGLenum * db_modes, unsigned num_db_modes, int visType ); -- cgit v1.2.3 From e8b52b3f5682c969e58077d42f5aebdad5d32e89 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 13:56:41 +0900 Subject: gallium: Drop deprecated __MSC__ macro. --- src/gallium/include/pipe/p_compiler.h | 15 +++++---------- src/gallium/include/pipe/p_debug.h | 4 ++-- src/gallium/include/pipe/p_util.h | 6 +++--- 3 files changed, 10 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index 521ef2d189..d2b8c41be3 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -39,20 +39,15 @@ #define __WIN32__ #endif -#if defined(_MSC_VER) && !defined(__MSC__) -#define __MSC__ -#endif - - -#if defined(__MSC__) +#if defined(_MSC_VER) /* Avoid 'expression is always true' warning */ #pragma warning(disable: 4296) -#endif /* __MSC__ */ +#endif /* _MSC_VER */ -#if defined(__MSC__) +#if defined(_MSC_VER) typedef __int8 int8_t; typedef unsigned __int8 uint8_t; @@ -114,7 +109,7 @@ typedef unsigned char boolean; /* Function inlining */ #if defined(__GNUC__) # define INLINE __inline__ -#elif defined(__MSC__) +#elif defined(_MSC_VER) # define INLINE __inline #elif defined(__ICL) # define INLINE __inline @@ -138,7 +133,7 @@ typedef unsigned char boolean; -#if defined __GNUC__ +#if defined(__GNUC__) #define ALIGN16_DECL(TYPE, NAME, SIZE) TYPE NAME##___aligned[SIZE] __attribute__(( aligned( 16 ) )) #define ALIGN16_ASSIGN(NAME) NAME##___aligned #define ALIGN16_ATTRIB __attribute__(( aligned( 16 ) )) diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index 9011557006..a5816c3cbb 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -132,9 +132,9 @@ void _debug_break(void); * Hard-coded breakpoint. */ #ifdef DEBUG -#if (defined(__i386__) || defined(__386__)) && defined(__GNUC__) +#if defined(PIPE_ARCH_X86) && defined(PIPE_CC_GCC) #define debug_break() __asm("int3") -#elif defined(_M_IX86) && defined(_MSC_VER) +#elif defined(PIPE_ARCH_X86) && defined(PIPE_CC_MSVC) #define debug_break() do { _asm {int 3} } while(0) #else #define debug_break() _debug_break() diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index f62faf616a..cf2447822a 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -224,7 +224,7 @@ static INLINE int align_int(int x, int align) -#if defined(__MSC__) && defined(__WIN32__) +#if defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86) static INLINE unsigned ffs( unsigned u ) { unsigned i; @@ -341,14 +341,14 @@ static INLINE int ifloor(float f) } -#if defined(__GNUC__) && defined(__i386__) +#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86) static INLINE int iround(float f) { int r; __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st"); return r; } -#elif defined(__MSC__) && defined(__WIN32__) +#elif defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86) static INLINE int iround(float f) { int r; -- cgit v1.2.3 From 4802a687a7050b7b5dd3683ae43fd4f26c4a718e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 14:00:10 +0900 Subject: dri/intel: Use standard integer types. --- src/gallium/winsys/dri/intel/intel_screen.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 8817a80a5a..1d61617ab5 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -418,8 +418,8 @@ intelFillInModes(unsigned pixel_bits, unsigned depth_bits, GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML }; - u_int8_t depth_bits_array[3]; - u_int8_t stencil_bits_array[3]; + uint8_t depth_bits_array[3]; + uint8_t stencil_bits_array[3]; depth_bits_array[0] = 0; -- cgit v1.2.3 From b6f053739f66c1c88db12df4690051c0a54ff0f7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 14:02:24 +0900 Subject: mesa: Replace deprecated __MSC__ macro. --- src/mesa/main/imports.h | 2 +- src/mesa/math/m_debug_util.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/imports.h b/src/mesa/main/imports.h index ebdfc452a7..813be52e4e 100644 --- a/src/mesa/main/imports.h +++ b/src/mesa/main/imports.h @@ -332,7 +332,7 @@ static INLINE int iround(float f) return r; } #define IROUND(x) iround(x) -#elif defined(USE_X86_ASM) && defined(__MSC__) && defined(__WIN32__) +#elif defined(USE_X86_ASM) && defined(_MSC_VER) static INLINE int iround(float f) { int r; diff --git a/src/mesa/math/m_debug_util.h b/src/mesa/math/m_debug_util.h index f7ce4679b6..776c26475e 100644 --- a/src/mesa/math/m_debug_util.h +++ b/src/mesa/math/m_debug_util.h @@ -303,7 +303,7 @@ enum { NIL = 0, ONE = 1, NEG = -1, VAR = 2 }; */ #if defined(__GNUC__) # define ALIGN16(type, array) type array __attribute__ ((aligned (16))) -#elif defined(__MSC__) +#elif defined(_MSC_VER) # define ALIGN16(type, array) type array __declspec(align(16)) /* GH: Does this work? */ #elif defined(__WATCOMC__) # define ALIGN16(type, array) /* Watcom does not support this */ -- cgit v1.2.3 From a148025d94505bca08f9baa1689048032bb60e2c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 14:18:07 +0900 Subject: mesa: Use standard integer types. Especially get rid of the non-portable long long. --- src/mesa/main/glheader.h | 35 ++++++++++++++++++++++++++--------- src/mesa/main/imports.c | 14 +++++--------- src/mesa/main/imports.h | 8 ++------ src/mesa/main/texcompress_fxt1.c | 13 ++++--------- 4 files changed, 37 insertions(+), 33 deletions(-) (limited to 'src') diff --git a/src/mesa/main/glheader.h b/src/mesa/main/glheader.h index 9e39a8370e..57d7e60ad3 100644 --- a/src/mesa/main/glheader.h +++ b/src/mesa/main/glheader.h @@ -69,16 +69,33 @@ #include -/* Get typedefs for uintptr_t and friends */ -#if defined(__MINGW32__) || defined(__NetBSD__) -# include -#elif defined(_WIN32) -# include -# if _MSC_VER == 1200 - typedef UINT_PTR uintptr_t; -# endif +/* Get standard integer types */ +#if defined(_MSC_VER) + + typedef __int8 int8_t; + typedef unsigned __int8 uint8_t; + typedef __int16 int16_t; + typedef unsigned __int16 uint16_t; +# ifndef __eglplatform_h_ + typedef __int32 int32_t; +# endif + typedef unsigned __int32 uint32_t; + typedef __int64 int64_t; + typedef unsigned __int64 uint64_t; + +# if defined(_WIN64) + typedef __int64 intptr_t; + typedef unsigned __int64 uintptr_t; +# else + typedef __int32 intptr_t; + typedef unsigned __int32 uintptr_t; +# endif + +# define INT64_C(__val) __val##i64 +# define UINT64_C(__val) __val##ui64 + #else -# include +# include #endif #if defined(_WIN32) && !defined(__WIN32__) && !defined(__CYGWIN__) && !defined(BUILD_FOR_SNAP) diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c index c46e5beb91..7231fa699c 100644 --- a/src/mesa/main/imports.c +++ b/src/mesa/main/imports.c @@ -539,10 +539,10 @@ _mesa_pow(double x, double y) * Find the first bit set in a word. */ int -_mesa_ffs(int i) +_mesa_ffs(int32_t i) { #if (defined(_WIN32) && !defined(__MINGW32__) ) || defined(__IBMC__) || defined(__IBMCPP__) - register int bit = 1; + register int32_t bit = 1; if ((i & 0xffff) == 0) { bit += 16; i >>= 16; @@ -573,11 +573,7 @@ _mesa_ffs(int i) * if no bits set. */ int -#ifdef __MINGW32__ -_mesa_ffsll(long val) -#else -_mesa_ffsll(long long val) -#endif +_mesa_ffsll(int64_t val) { #ifdef ffsll return ffsll(val); @@ -586,11 +582,11 @@ _mesa_ffsll(long long val) assert(sizeof(val) == 8); - bit = _mesa_ffs(val); + bit = _mesa_ffs((int32_t)val); if (bit != 0) return bit; - bit = _mesa_ffs(val >> 32); + bit = _mesa_ffs((int32_t)(val >> 32)); if (bit != 0) return 32 + bit; diff --git a/src/mesa/main/imports.h b/src/mesa/main/imports.h index 813be52e4e..d91a1366b0 100644 --- a/src/mesa/main/imports.h +++ b/src/mesa/main/imports.h @@ -697,14 +697,10 @@ extern double _mesa_pow(double x, double y); extern int -_mesa_ffs(int i); +_mesa_ffs(int32_t i); extern int -#ifdef __MINGW32__ -_mesa_ffsll(long i); -#else -_mesa_ffsll(long long i); -#endif +_mesa_ffsll(int64_t i); extern unsigned int _mesa_bitcount(unsigned int n); diff --git a/src/mesa/main/texcompress_fxt1.c b/src/mesa/main/texcompress_fxt1.c index b6991f45ed..16a3ba076f 100644 --- a/src/mesa/main/texcompress_fxt1.c +++ b/src/mesa/main/texcompress_fxt1.c @@ -298,22 +298,17 @@ const struct gl_texture_format _mesa_texformat_rgba_fxt1 = { /* * Define a 64-bit unsigned integer type and macros */ -#if defined(__GNUC__) && !defined(__cplusplus) +#if 1 #define FX64_NATIVE 1 -#ifdef __MINGW32__ -typedef unsigned long Fx64; -#else -typedef unsigned long long Fx64; -#endif - +typedef uint64_t Fx64; #define FX64_MOV32(a, b) a = b #define FX64_OR32(a, b) a |= b #define FX64_SHL(a, c) a <<= c -#else /* !__GNUC__ */ +#else #define FX64_NATIVE 0 @@ -335,7 +330,7 @@ typedef struct { } \ } while (0) -#endif /* !__GNUC__ */ +#endif #define F(i) (GLfloat)1 /* can be used to obtain an oblong metric: 0.30 / 0.59 / 0.11 */ -- cgit v1.2.3 From 9dfa6063be5ea506cc5a43d352fa2f7dcfa68dff Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 14:22:09 +0900 Subject: gallium: Avoid double arithmetic. --- src/gallium/auxiliary/translate/translate_generic.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index 17e37d1745..4c8179ffa8 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -121,7 +121,7 @@ emit_##NAME(const float *attrib, void *ptr) \ #define FROM_16_SNORM(i) ((float) ((short *) ptr)[i] / 32767.0f) #define FROM_32_SNORM(i) ((float) ((int *) ptr)[i] / 2147483647.0f) -#define FROM_32_FIXED(i) (((int *) ptr)[i] / 65536.0) +#define FROM_32_FIXED(i) (((int *) ptr)[i] / 65536.0f) #define TO_64_FLOAT(x) ((double) x) #define TO_32_FLOAT(x) (x) -- cgit v1.2.3 From 8c90148cb90c6e5947068e5f7dc9be4e46614dfe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 24 Jun 2008 20:44:47 +0900 Subject: mesa: Fix Windows GDI winsys build. --- src/mesa/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index af83db0b5f..d2de189f69 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -231,6 +231,7 @@ if env['platform'] != 'winddk': glapi_sources = [ 'main/dispatch.c', 'glapi/glapi.c', + 'glapi/glapi_getproc.c', 'glapi/glthread.c', ] -- cgit v1.2.3 From 72edcd171190f1d358abeb485582c87c104f5add Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 24 Jun 2008 14:30:38 +0200 Subject: egl: Stdint was missing from eglconfigutil.h --- src/egl/main/eglconfigutil.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/egl/main/eglconfigutil.h b/src/egl/main/eglconfigutil.h index dff407231f..7fe875a7d0 100644 --- a/src/egl/main/eglconfigutil.h +++ b/src/egl/main/eglconfigutil.h @@ -4,6 +4,7 @@ #include "eglconfig.h" #include "GL/internal/glcore.h" +#include "stdint.h" extern void -- cgit v1.2.3 From e8af160b247a4cf60c82a3d0c63ec952aef22d96 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 24 Jun 2008 13:00:42 +0200 Subject: gallium: Fix warning in u_blit.h --- src/gallium/auxiliary/util/u_blit.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h index 7d0406c104..308075698f 100644 --- a/src/gallium/auxiliary/util/u_blit.h +++ b/src/gallium/auxiliary/util/u_blit.h @@ -37,6 +37,7 @@ extern "C" { struct pipe_context; struct pipe_surface; +struct pipe_texture; struct cso_context; -- cgit v1.2.3 From 2d11411b19f725591e096b8a75dec94831d7bce7 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 24 Jun 2008 13:09:43 +0200 Subject: i915: Moved screen creation to intel_be_device --- .../winsys/common/intel_drm/intel_be_device.c | 9 ++++++++- .../winsys/common/intel_drm/intel_be_device.h | 7 ++++++- src/gallium/winsys/dri/intel/Makefile | 5 +++-- src/gallium/winsys/dri/intel/intel_context.c | 23 +++------------------- src/gallium/winsys/dri/intel/intel_screen.c | 10 ++++++---- 5 files changed, 26 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/common/intel_drm/intel_be_device.c b/src/gallium/winsys/common/intel_drm/intel_be_device.c index 3eed0fe410..efb57a394e 100644 --- a/src/gallium/winsys/common/intel_drm/intel_be_device.c +++ b/src/gallium/winsys/common/intel_drm/intel_be_device.c @@ -16,6 +16,8 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "i915simple/i915_screen.h" + /* Turn a pipe winsys into an intel/pipe winsys: */ static INLINE struct intel_be_device * @@ -192,7 +194,7 @@ intel_be_fence_finish( struct pipe_winsys *sws, */ boolean -intel_be_init_device(struct intel_be_device *dev, int fd) +intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) { dev->fd = fd; dev->max_batch_size = 16 * 4096; @@ -252,6 +254,11 @@ intel_be_init_device(struct intel_be_device *dev, int fd) 1, 40, dev->max_batch_size * 16, 0, dev->fMan); + /* Fill in this struct with callbacks that i915simple will need to + * communicate with the window system, buffer manager, etc. + */ + dev->screen = i915_create_screen(&dev->base, id); + return true; } diff --git a/src/gallium/winsys/common/intel_drm/intel_be_device.h b/src/gallium/winsys/common/intel_drm/intel_be_device.h index ec5cace71c..abf0253917 100644 --- a/src/gallium/winsys/common/intel_drm/intel_be_device.h +++ b/src/gallium/winsys/common/intel_drm/intel_be_device.h @@ -12,6 +12,11 @@ struct intel_be_device { struct pipe_winsys base; + /** + * Hw level screen + */ + struct pipe_screen *screen; + int fd; /**< Drm file discriptor */ size_t max_batch_size; @@ -28,7 +33,7 @@ struct intel_be_device }; boolean -intel_be_init_device(struct intel_be_device *device, int fd); +intel_be_init_device(struct intel_be_device *device, int fd, unsigned id); void intel_be_destroy_device(struct intel_be_device *dev); diff --git a/src/gallium/winsys/dri/intel/Makefile b/src/gallium/winsys/dri/intel/Makefile index 00ce21f690..5b51f0815d 100644 --- a/src/gallium/winsys/dri/intel/Makefile +++ b/src/gallium/winsys/dri/intel/Makefile @@ -6,8 +6,9 @@ LIBNAME = i915_dri.so PIPE_DRIVERS = \ $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ - $(TOP)/src/gallium/drivers/i915simple/libi915simple.a \ - $(TOP)/src/gallium/winsys/common/intel_drm/libinteldrm.a + $(TOP)/src/gallium/winsys/common/intel_drm/libinteldrm.a \ + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a + DRIVER_SOURCES = \ intel_winsys_softpipe.c \ diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c index 8284e0edbb..97ef731aaa 100644 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ b/src/gallium/winsys/dri/intel/intel_context.c @@ -134,25 +134,6 @@ static const struct dri_debug_control debug_control[] = { -/** - * Create i915 hardware rendering context. - */ -static struct pipe_context * -intel_create_i915simple(struct intel_context *intel, - struct pipe_winsys *winsys) -{ - struct pipe_screen *screen; - - /* Fill in this struct with callbacks that i915simple will need to - * communicate with the window system, buffer manager, etc. - */ - screen = i915_create_screen(winsys, intel->intelScreen->deviceID); - - /* Create the i915simple context: - */ - return i915_create_context(screen, winsys, &intel->base.base ); -} - static void intel_lock_hardware(struct intel_be_context *context) { @@ -250,7 +231,9 @@ intelCreateContext(const __GLcontextModes * visual, case PCI_CHIP_Q35_G: case PCI_CHIP_I915_G: case PCI_CHIP_I915_GM: - pipe = intel_create_i915simple( intel, &intelScreen->base.base ); + pipe = i915_create_context(intelScreen->base.screen, + &intelScreen->base.base, + &intel->base.base); break; default: fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 1d61617ab5..809af2e553 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -198,10 +198,12 @@ intel_flush_frontbuffer( struct pipe_winsys *winsys, struct pipe_surface *surf, void *context_private) { - struct intel_context *intel = (struct intel_context *) context_private; - __DRIdrawablePrivate *dPriv = intel->driDrawable; + //struct intel_context *intel = (struct intel_context *) context_private; + //__DRIdrawablePrivate *dPriv = intel->driDrawable; - intelDisplaySurface(dPriv, surf, NULL); + assert((int)"Doesn't work currently" & 0); + + //intelDisplaySurface(dPriv, surf, NULL); } static boolean @@ -254,9 +256,9 @@ intelInitDriver(__DRIscreenPrivate * sPriv) (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); } - intel_be_init_device(&intelScreen->base, sPriv->fd); intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; intelScreen->base.base.get_name = intel_get_name; + intel_be_init_device(&intelScreen->base, sPriv->fd, intelScreen->deviceID); return GL_TRUE; } -- cgit v1.2.3 From 19dad109bb7b271e0bb4b55e1374c299770b107e Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 24 Jun 2008 14:19:30 +0200 Subject: i915: Added texture_blanket function --- src/gallium/drivers/i915simple/i915_texture.c | 34 +++++++++++++++++++++++++++ 1 file changed, 34 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index ae107c6676..9d8a6d6de4 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -677,6 +677,39 @@ i915_get_tex_surface(struct pipe_screen *screen, return ps; } +static struct pipe_texture * +i915_texture_blanket(struct pipe_screen * screen, + const struct pipe_texture *base, + const unsigned *pitch, + struct pipe_buffer *buffer) +{ + struct i915_texture *tex; + assert(screen); + assert(templat); + + /* Only supports one type */ + if (base->target != PIPE_TEXTURE_2D || + base->last_level != 0 || + base->depth[0] != 1) { + return NULL; + } + + tex = CALLOC_STRUCT(i915_texture); + if (!tex) + return NULL; + + tex->base = *base; + + tex->pitch = pitch[0]; + + i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1); + i915_miptree_set_image_offset(tex, 0, 0, 0, 0); + + pipe_buffer_reference(screen->winsys, &tex->buffer, buffer); + + return &tex->base; +} + void i915_init_texture_functions(struct i915_context *i915) { @@ -713,5 +746,6 @@ i915_init_screen_texture_functions(struct pipe_screen *screen) screen->texture_create = i915_texture_create; screen->texture_release = i915_texture_release; screen->get_tex_surface = i915_get_tex_surface; + screen->texture_blanket = i915_texture_blanket; screen->tex_surface_release = i915_tex_surface_release; } -- cgit v1.2.3 From e95697758eda9b92a01a504bce823acd5201b48d Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 24 Jun 2008 14:42:12 +0200 Subject: i915: Create a texture and surface for shared frontbuffer --- src/gallium/winsys/dri/intel/intel_screen.c | 67 +++++++++++++++++++++++++++-- src/gallium/winsys/dri/intel/intel_screen.h | 2 + 2 files changed, 66 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 809af2e553..18427a4586 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -38,10 +38,69 @@ #include "intel_drm/ws_dri_bufpool.h" #include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_inlines.h" #include "state_tracker/st_public.h" #include "state_tracker/st_cb_fbo.h" +static void +intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle); +static void +intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle) +{ + struct intel_be_buffer *be_buf = malloc(sizeof(*be_buf)); + struct pipe_screen *screen = intelScreen->base.screen; + struct pipe_texture *texture; + struct pipe_texture templat; + struct pipe_surface *surface; + struct pipe_buffer *buffer = &be_buf->base; + unsigned pitch; + + assert(intelScreen->front.cpp == 4); + + /* XXX create a intel_be function for this */ + { + driGenBuffers(intelScreen->base.staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); + driBOSetReferenced(intelScreen->front.buffer, handle); + + memset(be_buf, 0, sizeof(*be_buf)); + buffer->refcount = 1; + buffer->alignment = 0; + buffer->usage = 0; + buffer->size = driBOSize(intelScreen->front.buffer); + be_buf->driBO = intelScreen->front.buffer; + } + + memset(&templat, 0, sizeof(templat)); + templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET; + templat.target = PIPE_TEXTURE_2D; + templat.last_level = 0; + templat.depth[0] = 1; + templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; + templat.cpp = intelScreen->front.cpp; + templat.width[0] = intelScreen->front.width; + templat.height[0] = intelScreen->front.height; + pitch = intelScreen->front.pitch / intelScreen->front.cpp; + + texture = screen->texture_blanket(screen, + &templat, + &pitch, + buffer); + + /* Unref the buffer we don't need it anyways */ + pipe_buffer_reference(screen->winsys, &buffer, NULL); + + surface = screen->get_tex_surface(screen, + texture, + 0, + 0, + 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + + intelScreen->front.texture = texture; + intelScreen->front.surface = surface; +} PUBLIC const char __driConfigOptions[] = DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE @@ -157,10 +216,12 @@ intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) } #else if (intelScreen->base.staticPool) { - if (intelScreen->front.buffer) + if (intelScreen->front.buffer) { driBOUnReference(intelScreen->front.buffer); - driGenBuffers(intelScreen->base.staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); - driBOSetReferenced(intelScreen->front.buffer, sarea->front_bo_handle); + pipe_surface_reference(&intelScreen->front.surface, NULL); + pipe_texture_reference(&intelScreen->front.texture, NULL); + } + intelCreateSurface(intelScreen, &intelScreen->base.base, sarea->front_bo_handle); } #endif } diff --git a/src/gallium/winsys/dri/intel/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h index 8036917903..e62f9e71ec 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.h +++ b/src/gallium/winsys/dri/intel/intel_screen.h @@ -47,6 +47,8 @@ struct intel_screen /* We create a static dri buffer for the frontbuffer. */ struct _DriBufferObject *buffer; + struct pipe_surface *surface; + struct pipe_texture *texture; char *map; /* memory map */ int offset; /* from start of video mem, in bytes */ -- cgit v1.2.3 From 62fd280c524b282f2448995a353d94f973d904d4 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 24 Jun 2008 16:06:59 +0200 Subject: gallium: Fix whole source being used in u_blit --- src/gallium/auxiliary/util/u_blit.c | 70 ++++++++++++++++++++++++++++--------- 1 file changed, 54 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 28513f54e6..6555dcd588 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -222,6 +222,49 @@ setup_vertex_data(struct blit_state *ctx, } +/** + * Setup vertex data for the textured quad we'll draw. + * Note: y=0=top + */ +static void +setup_vertex_data_tex(struct blit_state *ctx, + float x0, float y0, float x1, float y1, + float s0, float t0, float s1, float t1, + float z) +{ + void *buf; + + ctx->vertices[0][0][0] = x0; + ctx->vertices[0][0][1] = y0; + ctx->vertices[0][0][2] = z; + ctx->vertices[0][1][0] = s0; /*s*/ + ctx->vertices[0][1][1] = t0; /*t*/ + + ctx->vertices[1][0][0] = x1; + ctx->vertices[1][0][1] = y0; + ctx->vertices[1][0][2] = z; + ctx->vertices[1][1][0] = s1; /*s*/ + ctx->vertices[1][1][1] = t0; /*t*/ + + ctx->vertices[2][0][0] = x1; + ctx->vertices[2][0][1] = y1; + ctx->vertices[2][0][2] = z; + ctx->vertices[2][1][0] = s1; + ctx->vertices[2][1][1] = t1; + + ctx->vertices[3][0][0] = x0; + ctx->vertices[3][0][1] = y1; + ctx->vertices[3][0][2] = z; + ctx->vertices[3][1][0] = s0; + ctx->vertices[3][1][1] = t1; + + buf = ctx->pipe->winsys->buffer_map(ctx->pipe->winsys, ctx->vbuf, + PIPE_BUFFER_USAGE_CPU_WRITE); + + memcpy(buf, ctx->vertices, sizeof(ctx->vertices)); + + ctx->pipe->winsys->buffer_unmap(ctx->pipe->winsys, ctx->vbuf); +} /** * Copy pixel block from src surface to dst surface. * Overlapping regions are acceptable. @@ -382,25 +425,18 @@ util_blit_pixels_tex(struct blit_state *ctx, struct pipe_context *pipe = ctx->pipe; struct pipe_screen *screen = pipe->screen; struct pipe_framebuffer_state fb; - const int srcLeft = MIN2(srcX0, srcX1); - const int srcTop = MIN2(srcY0, srcY1); + float s0, t0, s1, t1; assert(filter == PIPE_TEX_MIPFILTER_NEAREST || filter == PIPE_TEX_MIPFILTER_LINEAR); - if (srcLeft != srcX0) { - /* left-right flip */ - int tmp = dstX0; - dstX0 = dstX1; - dstX1 = tmp; - } + assert(tex->width[0] != 0); + assert(tex->height[0] != 0); - if (srcTop != srcY0) { - /* up-down flip */ - int tmp = dstY0; - dstY0 = dstY1; - dstY1 = tmp; - } + s0 = srcX0 / (float)tex->width[0]; + s1 = srcX1 / (float)tex->width[0]; + t0 = srcY0 / (float)tex->height[0]; + t1 = srcY1 / (float)tex->height[0]; assert(screen->is_format_supported(screen, dst->format, PIPE_SURFACE)); @@ -443,9 +479,11 @@ util_blit_pixels_tex(struct blit_state *ctx, cso_set_framebuffer(ctx->cso, &fb); /* draw quad */ - setup_vertex_data(ctx, + setup_vertex_data_tex(ctx, (float) dstX0, (float) dstY0, - (float) dstX1, (float) dstY1, z); + (float) dstX1, (float) dstY1, + s0, t0, s1, t1, + z); util_draw_vertex_buffer(ctx->pipe, ctx->vbuf, PIPE_PRIM_TRIANGLE_FAN, -- cgit v1.2.3 From 2b692335101d248106b4e1c2d116a1f95712fd25 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 24 Jun 2008 08:45:59 -0600 Subject: gallium: remove bad assertion --- src/gallium/drivers/i915simple/i915_texture.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 9d8a6d6de4..b2e490c7db 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -685,7 +685,6 @@ i915_texture_blanket(struct pipe_screen * screen, { struct i915_texture *tex; assert(screen); - assert(templat); /* Only supports one type */ if (base->target != PIPE_TEXTURE_2D || -- cgit v1.2.3 From 0561a293b6596641c0200df7e6580599ecb8b111 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 24 Jun 2008 08:47:15 -0600 Subject: gallium: remove some old dispatch code --- src/gallium/include/pipe/p_thread.h | 22 ---------------------- 1 file changed, 22 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_thread.h b/src/gallium/include/pipe/p_thread.h index af9150aafb..6e526b7aa8 100644 --- a/src/gallium/include/pipe/p_thread.h +++ b/src/gallium/include/pipe/p_thread.h @@ -62,12 +62,6 @@ #define _P_THREAD_H_ -#if defined(USE_MGL_NAMESPACE) -#define _glapi_Dispatch _mglapi_Dispatch -#endif - - - #if (defined(PTHREADS) || defined(SOLARIS_THREADS) ||\ defined(WIN32_THREADS) || defined(USE_XTHREADS) || defined(BEOS_THREADS)) \ && !defined(THREADS) @@ -330,22 +324,6 @@ _glthread_GetTSD(_glthread_TSD *); extern void _glthread_SetTSD(_glthread_TSD *, void *); -#if defined(GLX_USE_TLS) - -extern __thread struct _glapi_table * _glapi_tls_Dispatch - __attribute__((tls_model("initial-exec"))); - -#define GET_DISPATCH() _glapi_tls_Dispatch - -#elif !defined(GL_CALL) -# if defined(THREADS) -# define GET_DISPATCH() \ - ((__builtin_expect( _glapi_Dispatch != NULL, 1 )) \ - ? _glapi_Dispatch : _glapi_get_dispatch()) -# else -# define GET_DISPATCH() _glapi_Dispatch -# endif /* defined(THREADS) */ -#endif /* ndef GL_CALL */ #endif /* _P_THREAD_H_ */ -- cgit v1.2.3 From d378f7b3dfda3b549e4b02380e492671cc34bb59 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 25 Jun 2008 08:45:14 -0600 Subject: mesa: point size arrays --- src/mesa/main/arrayobj.c | 9 +++++++++ src/mesa/main/enable.c | 12 ++++++++++++ src/mesa/main/ffvertex_prog.c | 34 ++++++++++++++++++++++++++++++---- src/mesa/main/glheader.h | 10 ++++++++++ src/mesa/main/mfeatures.h | 1 + src/mesa/main/mtypes.h | 4 ++++ src/mesa/main/state.c | 2 +- src/mesa/main/varray.c | 31 +++++++++++++++++++++++++++++++ src/mesa/main/varray.h | 4 ++++ src/mesa/state_tracker/st_program.c | 4 ++++ src/mesa/vbo/vbo_exec_array.c | 4 ++++ 11 files changed, 110 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/main/arrayobj.c b/src/mesa/main/arrayobj.c index f08f99d8e1..d62661e2b5 100644 --- a/src/mesa/main/arrayobj.c +++ b/src/mesa/main/arrayobj.c @@ -164,6 +164,15 @@ _mesa_initialize_array_object( GLcontext *ctx, obj->VertexAttrib[i].Normalized = GL_FALSE; } +#if FEATURE_point_size_array + obj->PointSize.Type = GL_FLOAT; + obj->PointSize.Stride = 0; + obj->PointSize.StrideB = 0; + obj->PointSize.Ptr = NULL; + obj->PointSize.Enabled = GL_FALSE; + obj->PointSize.BufferObj = ctx->Array.NullBufferObj; +#endif + #if FEATURE_ARB_vertex_buffer_object /* Vertex array buffers */ obj->Vertex.BufferObj = ctx->Array.NullBufferObj; diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c index 6b4ad8eea3..9dc55d4e69 100644 --- a/src/mesa/main/enable.c +++ b/src/mesa/main/enable.c @@ -92,6 +92,13 @@ client_state(GLcontext *ctx, GLenum cap, GLboolean state) flag = _NEW_ARRAY_COLOR1; break; +#if FEATURE_point_size_array + case GL_POINT_SIZE_ARRAY_OES: + var = &ctx->Array.ArrayObj->PointSize.Enabled; + flag = _NEW_ARRAY_POINT_SIZE; + break; +#endif + #if FEATURE_NV_vertex_program case GL_VERTEX_ATTRIB_ARRAY0_NV: case GL_VERTEX_ATTRIB_ARRAY1_NV: @@ -652,6 +659,7 @@ _mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state) case GL_EDGE_FLAG_ARRAY: case GL_FOG_COORDINATE_ARRAY_EXT: case GL_SECONDARY_COLOR_ARRAY_EXT: + case GL_POINT_SIZE_ARRAY_OES: client_state( ctx, cap, state ); return; @@ -1174,6 +1182,10 @@ _mesa_IsEnabled( GLenum cap ) case GL_SECONDARY_COLOR_ARRAY_EXT: CHECK_EXTENSION(EXT_secondary_color); return (ctx->Array.ArrayObj->SecondaryColor.Enabled != 0); +#if FEATURE_point_size_array + case GL_POINT_SIZE_ARRAY_OES: + return (ctx->Array.ArrayObj->PointSize.Enabled != 0); +#endif /* GL_EXT_histogram */ case GL_HISTOGRAM: diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index e6c7c1040f..2baae77570 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -63,6 +63,7 @@ struct state_key { unsigned separate_specular:1; unsigned fog_mode:2; unsigned point_attenuated:1; + unsigned point_array:1; unsigned texture_enabled_global:1; unsigned fragprog_inputs_read:12; @@ -264,6 +265,9 @@ static struct state_key *make_state_key( GLcontext *ctx ) if (ctx->Point._Attenuated) key->point_attenuated = 1; + if (ctx->Array.ArrayObj->PointSize.Enabled) + key->point_array = 1; + if (ctx->Texture._TexGenEnabled || ctx->Texture._TexMatEnabled || ctx->Texture._EnabledUnits) @@ -444,12 +448,18 @@ static void release_temps( struct tnl_program *p ) +/** + * \param input one of VERT_ATTRIB_x tokens. + */ static struct ureg register_input( struct tnl_program *p, GLuint input ) { p->program->Base.InputsRead |= (1<program->Base.OutputsWritten |= (1<state->point_attenuated) - build_pointsize(p); + build_atten_pointsize(p); + else if (p->state->point_array) + build_array_pointsize(p); #if 0 else - constant_pointsize(p); + build_constant_pointsize(p); #endif /* Finish up: diff --git a/src/mesa/main/glheader.h b/src/mesa/main/glheader.h index 57d7e60ad3..3131a356b8 100644 --- a/src/mesa/main/glheader.h +++ b/src/mesa/main/glheader.h @@ -167,6 +167,16 @@ #endif +#ifndef GL_OES_point_size_array +#define GL_POINT_SIZE_ARRAY_OES 0x8B9C +#define GL_POINT_SIZE_ARRAY_TYPE_OES 0x898A +#define GL_POINT_SIZE_ARRAY_STRIDE_OES 0x898B +#define GL_POINT_SIZE_ARRAY_POINTER_OES 0x898C +#define GL_POINT_SIZE_ARRAY_BUFFER_BINDING_OES 0x8B9F +#endif + + + #if !defined(CAPI) && defined(WIN32) && !defined(BUILD_FOR_SNAP) #define CAPI _cdecl #endif diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h index a305bbd605..c3c337ea90 100644 --- a/src/mesa/main/mfeatures.h +++ b/src/mesa/main/mfeatures.h @@ -49,6 +49,7 @@ #define FEATURE_fixedpt 0 #define FEATURE_histogram _HAVE_FULL_GL #define FEATURE_pixel_transfer _HAVE_FULL_GL +#define FEATURE_point_size_array 0 #define FEATURE_texgen _HAVE_FULL_GL #define FEATURE_texture_fxt1 _HAVE_FULL_GL #define FEATURE_texture_s3tc _HAVE_FULL_GL diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 8a6c84368a..dceb7611f2 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -151,6 +151,7 @@ enum VERT_ATTRIB_COLOR1 = 4, VERT_ATTRIB_FOG = 5, VERT_ATTRIB_COLOR_INDEX = 6, + VERT_ATTRIB_POINT_SIZE = 6, /*alias*/ VERT_ATTRIB_EDGEFLAG = 7, VERT_ATTRIB_TEX0 = 8, VERT_ATTRIB_TEX1 = 9, @@ -1707,6 +1708,7 @@ struct gl_array_object struct gl_client_array Index; struct gl_client_array EdgeFlag; struct gl_client_array TexCoord[MAX_TEXTURE_COORD_UNITS]; + struct gl_client_array PointSize; /*@}*/ /** Generic arrays for vertex programs/shaders */ @@ -2748,6 +2750,7 @@ struct gl_matrix_stack #define _NEW_ARRAY_FOGCOORD VERT_BIT_FOG #define _NEW_ARRAY_INDEX VERT_BIT_COLOR_INDEX #define _NEW_ARRAY_EDGEFLAG VERT_BIT_EDGEFLAG +#define _NEW_ARRAY_POINT_SIZE VERT_BIT_COLOR_INDEX /* aliased */ #define _NEW_ARRAY_TEXCOORD_0 VERT_BIT_TEX0 #define _NEW_ARRAY_TEXCOORD_1 VERT_BIT_TEX1 #define _NEW_ARRAY_TEXCOORD_2 VERT_BIT_TEX2 @@ -2765,6 +2768,7 @@ struct gl_matrix_stack /*@}*/ + /** * \name A bunch of flags that we think might be useful to drivers. * diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 4d1fdbf47c..315253d90a 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -448,7 +448,7 @@ _mesa_update_state_locked( GLcontext *ctx ) prog_flags |= (_NEW_TEXTURE | _NEW_FOG | _DD_NEW_SEPARATE_SPECULAR); } if (ctx->VertexProgram._MaintainTnlProgram) { - prog_flags |= (_NEW_TEXTURE | _NEW_TEXTURE_MATRIX | + prog_flags |= (_NEW_ARRAY | _NEW_TEXTURE | _NEW_TEXTURE_MATRIX | _NEW_TRANSFORM | _NEW_POINT | _NEW_FOG | _NEW_LIGHT | _MESA_NEW_NEED_EYE_COORDS); diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c index 220db35855..50fe874556 100644 --- a/src/mesa/main/varray.c +++ b/src/mesa/main/varray.c @@ -468,6 +468,37 @@ _mesa_EdgeFlagPointer(GLsizei stride, const GLvoid *ptr) } +void GLAPIENTRY +_mesa_PointSizePointer(GLenum type, GLsizei stride, const GLvoid *ptr) +{ + GLsizei elementSize; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + + if (stride < 0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glPointSizePointer(stride)" ); + return; + } + + switch (type) { + case GL_FLOAT: + elementSize = sizeof(GLfloat); + break; +#if FEATURE_fixedpt + case GL_FIXED: + elementSize = sizeof(GLfixed); + break; +#endif + default: + _mesa_error( ctx, GL_INVALID_ENUM, "glPointSizePointer(type)" ); + return; + } + + update_array(ctx, &ctx->Array.ArrayObj->PointSize, _NEW_ARRAY_POINT_SIZE, + elementSize, 1, type, stride, GL_FALSE, ptr); +} + + #if FEATURE_NV_vertex_program void GLAPIENTRY _mesa_VertexAttribPointerNV(GLuint index, GLint size, GLenum type, diff --git a/src/mesa/main/varray.h b/src/mesa/main/varray.h index ba91ecf5f6..f557940738 100644 --- a/src/mesa/main/varray.h +++ b/src/mesa/main/varray.h @@ -111,6 +111,10 @@ _mesa_SecondaryColorPointerEXT(GLint size, GLenum type, GLsizei stride, const GLvoid *ptr); +extern void GLAPIENTRY +_mesa_PointSizePointer(GLenum type, GLsizei stride, const GLvoid *ptr); + + extern void GLAPIENTRY _mesa_VertexAttribPointerNV(GLuint index, GLint size, GLenum type, GLsizei stride, const GLvoid *pointer); diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index d450c30694..958096f68e 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -136,6 +136,10 @@ st_translate_vertex_program(struct st_context *st, vs_input_semantic_name[slot] = TGSI_SEMANTIC_FOG; vs_input_semantic_index[slot] = 0; break; + case VERT_ATTRIB_POINT_SIZE: + vs_input_semantic_name[slot] = TGSI_SEMANTIC_PSIZE; + vs_input_semantic_index[slot] = 0; + break; case VERT_ATTRIB_TEX0: case VERT_ATTRIB_TEX1: case VERT_ATTRIB_TEX2: diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c index e3d2fc51cb..bf97956286 100644 --- a/src/mesa/vbo/vbo_exec_array.c +++ b/src/mesa/vbo/vbo_exec_array.c @@ -107,6 +107,10 @@ static void bind_array_obj( GLcontext *ctx ) exec->array.legacy_array[VERT_ATTRIB_COLOR1] = &ctx->Array.ArrayObj->SecondaryColor; exec->array.legacy_array[VERT_ATTRIB_FOG] = &ctx->Array.ArrayObj->FogCoord; exec->array.legacy_array[VERT_ATTRIB_COLOR_INDEX] = &ctx->Array.ArrayObj->Index; + if (ctx->Array.ArrayObj->PointSize.Enabled) { + /* this aliases COLOR_INDEX */ + exec->array.legacy_array[VERT_ATTRIB_POINT_SIZE] = &ctx->Array.ArrayObj->PointSize; + } exec->array.legacy_array[VERT_ATTRIB_EDGEFLAG] = &ctx->Array.ArrayObj->EdgeFlag; for (i = 0; i < 8; i++) -- cgit v1.2.3 From 38d779a3e68bb0ed78135c868e5bd435e1d91fb5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 26 Jun 2008 23:39:13 +0900 Subject: gallium: Describe pixel block. Chars-per-pixel paradigm is not enough to represent compressed and yuv pixel formats. --- src/gallium/include/pipe/p_format.h | 56 +++++++++++++++++++++++++++++++++++++ 1 file changed, 56 insertions(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 00aa02311c..d973fb357b 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -445,6 +445,62 @@ static INLINE uint pf_get_size( enum pipe_format format ) return pf_get_bits(format) / 8; } +/** + * Describe accurately the pixel format. + * + * The chars-per-pixel concept falls apart with compressed and yuv images, where + * more than one pixel are coded in a single data block. This structure + * describes that block. + * + * Simple pixel formats are effectively a 1x1xcpp block. + */ +struct pipe_format_block +{ + /** Block size in bytes */ + unsigned size; + + /** Block width in pixels */ + unsigned width; + + /** Block height in pixels */ + unsigned height; +}; + +/** + * Describe pixel format's block. + * + * @sa http://msdn2.microsoft.com/en-us/library/ms796147.aspx + */ +static INLINE void +pf_get_block(enum pipe_format format, struct pipe_format_block *block) +{ + switch(format) { + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT1_RGB: + block->size = 8; + block->width = 4; + block->height = 4; + break; + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: + block->size = 16; + block->width = 4; + block->height = 4; + break; + case PIPE_FORMAT_YCBCR: + case PIPE_FORMAT_YCBCR_REV: + block->size = 4; /* 2*cpp */ + block->width = 2; + block->height = 1; + break; + default: + block->size = pf_get_size(format); + block->width = 1; + block->height = 1; + break; + } +} + #ifdef __cplusplus } #endif -- cgit v1.2.3 From cb09d1ef60dae3b3c44f1370fd2f726c7044bc17 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 27 Jun 2008 12:34:57 +0900 Subject: mesa: Use the appropriate relocation. I don't fully understand this, but this matches the assembly code gcc generates glapi.c and fixes following error on x86-64 with assembly: relocation R_X86_64_PC32 against `_gl_DispatchTSD' can not be used when making a shared object --- src/mesa/x86-64/glapi_x86-64.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/x86-64/glapi_x86-64.S b/src/mesa/x86-64/glapi_x86-64.S index cb34061b36..f8337ff93e 100644 --- a/src/mesa/x86-64/glapi_x86-64.S +++ b/src/mesa/x86-64/glapi_x86-64.S @@ -73,7 +73,7 @@ _x86_64_get_dispatch: .p2align 4,,15 _x86_64_get_dispatch: - movq _gl_DispatchTSD(%rip), %rdi + movq _gl_DispatchTSD@GOTPCREL(%rip), %rdi jmp pthread_getspecific@PLT #elif defined(THREADS) -- cgit v1.2.3 From 6106db4c5da6fc5ae9ef157c939ce0834cdc5b92 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 27 Jun 2008 13:01:37 +0900 Subject: scons: Fix i965/xlib build. --- src/gallium/drivers/i965simple/SConscript | 1 + src/gallium/winsys/xlib/SConscript | 7 ++++++- 2 files changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/SConscript b/src/gallium/drivers/i965simple/SConscript index c0825c4de3..43fc2a4005 100644 --- a/src/gallium/drivers/i965simple/SConscript +++ b/src/gallium/drivers/i965simple/SConscript @@ -26,6 +26,7 @@ i965simple = env.ConvenienceLibrary( 'brw_gs_emit.c', 'brw_gs_state.c', 'brw_misc_state.c', + 'brw_screen.c', 'brw_sf.c', 'brw_sf_emit.c', 'brw_sf_state.c', diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript index 5e98a36abc..14a85ae0f2 100644 --- a/src/gallium/winsys/xlib/SConscript +++ b/src/gallium/winsys/xlib/SConscript @@ -3,7 +3,11 @@ Import('*') -if env['platform'] == 'linux' and 'mesa' in env['statetrackers'] and not env['dri']: +if env['platform'] == 'linux' \ + and 'mesa' in env['statetrackers'] \ + and 'softpipe' in env['drivers'] \ + and 'i965simple' in env['drivers'] \ + and not env['dri']: env = env.Clone() @@ -24,6 +28,7 @@ if env['platform'] == 'linux' and 'mesa' in env['statetrackers'] and not env['dr drivers = [ softpipe, + i965simple ] # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions -- cgit v1.2.3 From 4ddd65967915ca4846f2831bc676c878a29dae4a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 27 Jun 2008 19:37:56 +0900 Subject: gallium: Drop pipe_texture->cpp and pipe_surface->cpp. The chars-per-pixel concept falls apart with compressed and yuv images, where more than one pixel are coded in a single data block. --- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 4 +- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 6 +- src/gallium/auxiliary/util/p_tile.c | 98 ++++------- src/gallium/auxiliary/util/p_util.c | 100 +++++++++-- src/gallium/auxiliary/util/u_blit.c | 2 +- src/gallium/auxiliary/util/u_gen_mipmap.c | 8 +- src/gallium/drivers/i915simple/i915_blit.c | 2 - src/gallium/drivers/i915simple/i915_context.h | 6 +- src/gallium/drivers/i915simple/i915_state_emit.c | 4 +- .../drivers/i915simple/i915_state_sampler.c | 2 +- src/gallium/drivers/i915simple/i915_surface.c | 68 ++----- src/gallium/drivers/i915simple/i915_texture.c | 196 +++++++++++---------- src/gallium/drivers/i965simple/brw_context.h | 4 +- src/gallium/drivers/i965simple/brw_misc_state.c | 8 +- src/gallium/drivers/i965simple/brw_surface.c | 69 ++------ src/gallium/drivers/i965simple/brw_tex_layout.c | 117 ++++++------ .../drivers/i965simple/brw_wm_surface_state.c | 6 +- src/gallium/drivers/softpipe/sp_surface.c | 45 ++--- src/gallium/drivers/softpipe/sp_texture.c | 20 ++- src/gallium/drivers/softpipe/sp_texture.h | 2 +- src/gallium/include/pipe/p_format.h | 41 +++++ src/gallium/include/pipe/p_state.h | 11 +- src/gallium/include/pipe/p_util.h | 15 +- src/gallium/winsys/xlib/brw_aub.c | 9 +- src/gallium/winsys/xlib/xm_winsys.c | 24 +-- src/gallium/winsys/xlib/xm_winsys_aub.c | 31 ++-- src/mesa/state_tracker/st_cb_accum.c | 24 +-- src/mesa/state_tracker/st_cb_bitmap.c | 4 +- src/mesa/state_tracker/st_cb_drawpixels.c | 11 +- src/mesa/state_tracker/st_cb_fbo.c | 8 +- src/mesa/state_tracker/st_cb_readpixels.c | 6 +- src/mesa/state_tracker/st_cb_texture.c | 17 +- src/mesa/state_tracker/st_gen_mipmap.c | 4 +- src/mesa/state_tracker/st_texture.c | 26 ++- 34 files changed, 513 insertions(+), 485 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index ecdebca5f1..3dd7ee19fd 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -398,7 +398,7 @@ aaline_create_texture(struct aaline_stage *aaline) texTemp.width[0] = 1 << MAX_TEXTURE_LEVEL; texTemp.height[0] = 1 << MAX_TEXTURE_LEVEL; texTemp.depth[0] = 1; - texTemp.cpp = 1; + pf_get_block(texTemp.format, &texTemp.block); aaline->texture = screen->texture_create(screen, &texTemp); if (!aaline->texture) @@ -439,7 +439,7 @@ aaline_create_texture(struct aaline_stage *aaline) else { d = 255; } - data[i * surface->pitch + j] = d; + data[i * surface->stride + j] = d; } } diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 4087cf7a49..1f63f94365 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -394,11 +394,11 @@ pstip_update_texture(struct pstip_stage *pstip) for (j = 0; j < 32; j++) { if (stipple[i] & (bit31 >> j)) { /* fragment "on" */ - data[i * surface->pitch + j] = 0; + data[i * surface->stride + j] = 0; } else { /* fragment "off" */ - data[i * surface->pitch + j] = 255; + data[i * surface->stride + j] = 255; } } } @@ -426,7 +426,7 @@ pstip_create_texture(struct pstip_stage *pstip) texTemp.width[0] = 32; texTemp.height[0] = 32; texTemp.depth[0] = 1; - texTemp.cpp = 1; + pf_get_block(texTemp.format, &texTemp.block); pstip->texture = screen->texture_create(screen, &texTemp); if (pstip->texture == NULL) diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 5728757d2f..ab603ff6e4 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -48,34 +48,23 @@ void pipe_get_tile_raw(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, - void *p, int dst_stride) + void *dst, int dst_stride) { struct pipe_screen *screen = pipe->screen; - const uint cpp = ps->cpp; - const ubyte *pSrc; - const uint src_stride = ps->pitch * cpp; - ubyte *pDest; - uint i; - - if (dst_stride == 0) { - dst_stride = w * cpp; - } + const void *src; if (pipe_clip_tile(x, y, &w, &h, ps)) return; - pSrc = (const ubyte *) screen->surface_map(screen, ps, - PIPE_BUFFER_USAGE_CPU_READ); - assert(pSrc); /* XXX: proper error handling! */ + if (dst_stride == 0) + dst_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; - pSrc += (y * ps->pitch + x) * cpp; - pDest = (ubyte *) p; + src = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ); + assert(src); + if(!src) + return; - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w * cpp); - pDest += dst_stride; - pSrc += src_stride; - } + pipe_copy_rect(dst, &ps->block, dst_stride, 0, 0, w, h, src, ps->stride, x, y); screen->surface_unmap(screen, ps); } @@ -89,34 +78,23 @@ void pipe_put_tile_raw(struct pipe_context *pipe, struct pipe_surface *ps, uint x, uint y, uint w, uint h, - const void *p, int src_stride) + const void *src, int src_stride) { struct pipe_screen *screen = pipe->screen; - const uint cpp = ps->cpp; - const ubyte *pSrc; - const uint dst_stride = ps->pitch * cpp; - ubyte *pDest; - uint i; - - if (src_stride == 0) { - src_stride = w * cpp; - } + void *dst; if (pipe_clip_tile(x, y, &w, &h, ps)) return; - pSrc = (const ubyte *) p; + if (src_stride == 0) + src_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; - pDest = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE); - assert(pDest); /* XXX: proper error handling */ - - pDest += (y * ps->pitch + x) * cpp; + dst = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE); + assert(dst); + if(!dst) + return; - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, w * cpp); - pDest += dst_stride; - pSrc += src_stride; - } + pipe_copy_rect(dst, &ps->block, ps->stride, x, y, w, h, src, src_stride, 0, 0); screen->surface_unmap(screen, ps); } @@ -692,12 +670,12 @@ pipe_get_tile_rgba(struct pipe_context *pipe, if (pipe_clip_tile(x, y, &w, &h, ps)) return; - packed = MALLOC(h * w * ps->cpp); + packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); if (!packed) return; - pipe_get_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); + pipe_get_tile_raw(pipe, ps, x, y, w, h, packed, 0); switch (ps->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: @@ -774,7 +752,7 @@ pipe_put_tile_rgba(struct pipe_context *pipe, if (pipe_clip_tile(x, y, &w, &h, ps)) return; - packed = MALLOC(h * w * ps->cpp); + packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); if (!packed) return; @@ -829,7 +807,7 @@ pipe_put_tile_rgba(struct pipe_context *pipe, assert(0); } - pipe_put_tile_raw(pipe, ps, x, y, w, h, packed, w * ps->cpp); + pipe_put_tile_raw(pipe, ps, x, y, w, h, packed, 0); FREE(packed); } @@ -846,14 +824,14 @@ pipe_get_tile_z(struct pipe_context *pipe, { struct pipe_screen *screen = pipe->screen; const uint dstStride = w; - void *map; + ubyte *map; uint *pDest = z; uint i, j; if (pipe_clip_tile(x, y, &w, &h, ps)) return; - map = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ); + map = (ubyte *)screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ); if (!map) { assert(0); return; @@ -863,11 +841,11 @@ pipe_get_tile_z(struct pipe_context *pipe, case PIPE_FORMAT_Z32_UNORM: { const uint *pSrc - = (const uint *)map + (y * ps->pitch + x); + = (const uint *)(map + y * ps->stride + x*4); for (i = 0; i < h; i++) { memcpy(pDest, pSrc, 4 * w); pDest += dstStride; - pSrc += ps->pitch; + pSrc += ps->stride/4; } } break; @@ -875,28 +853,28 @@ pipe_get_tile_z(struct pipe_context *pipe, case PIPE_FORMAT_X8Z24_UNORM: { const uint *pSrc - = (const uint *)map + (y * ps->pitch + x); + = (const uint *)(map + y * ps->stride + x*4); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 24-bit Z to 32-bit Z */ pDest[j] = (pSrc[j] << 8) | (pSrc[j] & 0xff); } pDest += dstStride; - pSrc += ps->pitch; + pSrc += ps->stride/4; } } break; case PIPE_FORMAT_Z16_UNORM: { const ushort *pSrc - = (const ushort *)map + (y * ps->pitch + x); + = (const ushort *)(map + y * ps->stride + x*2); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 16-bit Z to 32-bit Z */ pDest[j] = (pSrc[j] << 16) | pSrc[j]; } pDest += dstStride; - pSrc += ps->pitch; + pSrc += ps->stride/2; } } break; @@ -917,13 +895,13 @@ pipe_put_tile_z(struct pipe_context *pipe, struct pipe_screen *screen = pipe->screen; const uint srcStride = w; const uint *pSrc = zSrc; - void *map; + ubyte *map; uint i, j; if (pipe_clip_tile(x, y, &w, &h, ps)) return; - map = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE); + map = (ubyte *)screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE); if (!map) { assert(0); return; @@ -932,10 +910,10 @@ pipe_put_tile_z(struct pipe_context *pipe, switch (ps->format) { case PIPE_FORMAT_Z32_UNORM: { - uint *pDest = (uint *) map + (y * ps->pitch + x); + uint *pDest = (uint *) (map + y * ps->stride + x*4); for (i = 0; i < h; i++) { memcpy(pDest, pSrc, 4 * w); - pDest += ps->pitch; + pDest += ps->stride/4; pSrc += srcStride; } } @@ -943,26 +921,26 @@ pipe_put_tile_z(struct pipe_context *pipe, case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_X8Z24_UNORM: { - uint *pDest = (uint *) map + (y * ps->pitch + x); + uint *pDest = (uint *) (map + y * ps->stride + x*4); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 32-bit Z to 24-bit Z (0 stencil) */ pDest[j] = pSrc[j] >> 8; } - pDest += ps->pitch; + pDest += ps->stride/4; pSrc += srcStride; } } break; case PIPE_FORMAT_Z16_UNORM: { - ushort *pDest = (ushort *) map + (y * ps->pitch + x); + ushort *pDest = (ushort *) (map + y * ps->stride + x*2); for (i = 0; i < h; i++) { for (j = 0; j < w; j++) { /* convert 32-bit Z to 16-bit Z */ pDest[j] = pSrc[j] >> 16; } - pDest += ps->pitch; + pDest += ps->stride/2; pSrc += srcStride; } } diff --git a/src/gallium/auxiliary/util/p_util.c b/src/gallium/auxiliary/util/p_util.c index 4e60b1b841..271be4edf1 100644 --- a/src/gallium/auxiliary/util/p_util.c +++ b/src/gallium/auxiliary/util/p_util.c @@ -32,6 +32,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" +#include "pipe/p_format.h" /** @@ -41,42 +42,109 @@ */ void pipe_copy_rect(ubyte * dst, - unsigned cpp, - unsigned dst_pitch, + const struct pipe_format_block *block, + unsigned dst_stride, unsigned dst_x, unsigned dst_y, unsigned width, unsigned height, const ubyte * src, - int src_pitch, + int src_stride, unsigned src_x, int src_y) { unsigned i; - int src_pitch_pos = src_pitch < 0 ? -src_pitch : src_pitch; + int src_stride_pos = src_stride < 0 ? -src_stride : src_stride; - assert(cpp > 0); + assert(block->size > 0); + assert(block->width > 0); + assert(block->height > 0); assert(src_x >= 0); assert(src_y >= 0); assert(dst_x >= 0); assert(dst_y >= 0); - dst_pitch *= cpp; - src_pitch *= cpp; - src_pitch_pos *= cpp; - dst += dst_x * cpp; - src += src_x * cpp; - dst += dst_y * dst_pitch; - src += src_y * src_pitch_pos; - width *= cpp; + dst_x /= block->width; + dst_y /= block->height; + width = (width + block->width - 1)/block->width; + height = (height + block->height - 1)/block->height; + src_x /= block->width; + src_y /= block->height; + + dst += dst_x * block->size; + src += src_x * block->size; + dst += dst_y * dst_stride; + src += src_y * src_stride_pos; + width *= block->size; - if (width == dst_pitch && width == src_pitch) + if (width == dst_stride && width == src_stride) memcpy(dst, src, height * width); else { for (i = 0; i < height; i++) { memcpy(dst, src, width); - dst += dst_pitch; - src += src_pitch; + dst += dst_stride; + src += src_stride; } } } + +void +pipe_fill_rect(ubyte * dst, + const struct pipe_format_block *block, + unsigned dst_stride, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + uint32_t value) +{ + unsigned i, j; + unsigned width_size; + + assert(block->size > 0); + assert(block->width > 0); + assert(block->height > 0); + assert(dst_x >= 0); + assert(dst_y >= 0); + + dst_x /= block->width; + dst_y /= block->height; + width = (width + block->width - 1)/block->width; + height = (height + block->height - 1)/block->height; + + dst += dst_x * block->size; + dst += dst_y * dst_stride; + width_size = width * block->size; + + switch (block->size) { + case 1: + if(dst_stride == width_size) + memset(dst, (ubyte) value, height * width_size); + else { + for (i = 0; i < height; i++) { + memset(dst, (ubyte) value, width_size); + dst += dst_stride; + } + } + break; + case 2: + for (i = 0; i < height; i++) { + uint16_t *row = (uint16_t *)dst; + for (j = 0; j < width; j++) + *row++ = (uint16_t) value; + dst += dst_stride; + } + break; + case 4: + for (i = 0; i < height; i++) { + uint32_t *row = (uint32_t *)dst; + for (j = 0; j < width; j++) + *row++ = value; + dst += dst_stride; + } + break; + default: + assert(0); + break; + } +} diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 6555dcd588..ae779335dc 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -324,7 +324,7 @@ util_blit_pixels(struct blit_state *ctx, texTemp.height[0] = srcH; texTemp.depth[0] = 1; texTemp.compressed = 0; - texTemp.cpp = pf_get_size(src->format); + pf_get_block(src->format, &texTemp.block); tex = screen->texture_create(screen, &texTemp); if (!tex) diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 7d71aefda9..5313a8008a 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -625,7 +625,9 @@ make_2d_mipmap(struct gen_mipmap_state *ctx, struct pipe_winsys *winsys = pipe->winsys; const uint zslice = 0; uint dstLevel; - const int bpt = pf_get_size(pt->format); + + assert(pt->block.width == 1); + assert(pt->block.height == 1); for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; @@ -646,9 +648,9 @@ make_2d_mipmap(struct gen_mipmap_state *ctx, reduce_2d(pt->format, srcSurf->width, srcSurf->height, - srcSurf->pitch * bpt, srcMap, + srcSurf->stride, srcMap, dstSurf->width, dstSurf->height, - dstSurf->pitch * bpt, dstMap); + dstSurf->stride, dstMap); winsys->buffer_unmap(winsys, srcSurf->buffer); winsys->buffer_unmap(winsys, dstSurf->buffer); diff --git a/src/gallium/drivers/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c index 22f91fab92..45fae4c999 100644 --- a/src/gallium/drivers/i915simple/i915_blit.c +++ b/src/gallium/drivers/i915simple/i915_blit.c @@ -47,8 +47,6 @@ i915_fill_blit(struct i915_context *i915, { unsigned BR13, CMD; - dst_pitch *= (short) cpp; - switch (cpp) { case 1: case 2: diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index 5d411a6648..c8db4f608c 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -188,9 +188,9 @@ struct i915_texture { /* Derived from the above: */ - unsigned pitch; - unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; + unsigned stride; + unsigned depth_stride; /* per-image on i945? */ + unsigned total_nblocksy; unsigned tiled; diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c index 19d968fd8b..9bd6f92323 100644 --- a/src/gallium/drivers/i915simple/i915_state_emit.c +++ b/src/gallium/drivers/i915simple/i915_state_emit.c @@ -211,7 +211,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) struct pipe_surface *depth_surface = i915->framebuffer.zsbuf; if (cbuf_surface) { - unsigned cpitch = (cbuf_surface->pitch * cbuf_surface->cpp); + unsigned cpitch = cbuf_surface->stride; unsigned ctile = BUF_3D_USE_FENCE; if (cbuf_surface->texture && ((struct i915_texture*)(cbuf_surface->texture))->tiled) { @@ -232,7 +232,7 @@ i915_emit_hardware_state(struct i915_context *i915 ) /* What happens if no zbuf?? */ if (depth_surface) { - unsigned zpitch = (depth_surface->pitch * depth_surface->cpp); + unsigned zpitch = depth_surface->stride; unsigned ztile = BUF_3D_USE_FENCE; if (depth_surface->texture && ((struct i915_texture*)(depth_surface->texture))->tiled) { diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index 379aff3846..7868f21ca6 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -242,7 +242,7 @@ i915_update_texture(struct i915_context *i915, assert(depth); format = translate_texture_format(pt->format); - pitch = tex->pitch * pt->cpp; + pitch = tex->stride; assert(format); assert(pitch); diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index cc55a0910e..0061b22f26 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -48,7 +48,9 @@ i915_surface_copy(struct pipe_context *pipe, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { assert( dst != src ); - assert( dst->cpp == src->cpp ); + assert( dst->block.size == src->block.size ); + assert( dst->block.width == src->block.height ); + assert( dst->block.height == src->block.height ); if (0) { void *dst_map = pipe->screen->surface_map( pipe->screen, @@ -60,38 +62,30 @@ i915_surface_copy(struct pipe_context *pipe, PIPE_BUFFER_USAGE_CPU_READ ); pipe_copy_rect(dst_map, - dst->cpp, - dst->pitch, + &dst->block, + dst->stride, dstx, dsty, width, height, src_map, - do_flip ? -(int) src->pitch : src->pitch, + do_flip ? -(int) src->stride : src->stride, srcx, do_flip ? height - 1 - srcy : srcy); pipe->screen->surface_unmap(pipe->screen, src); pipe->screen->surface_unmap(pipe->screen, dst); } else { + assert(dst->block.width == 1); + assert(dst->block.height == 1); i915_copy_blit( i915_context(pipe), do_flip, - dst->cpp, - (short) src->pitch, src->buffer, src->offset, - (short) dst->pitch, dst->buffer, dst->offset, + dst->block.size, + (short) src->stride/src->block.size, src->buffer, src->offset, + (short) dst->stride/dst->block.size, dst->buffer, dst->offset, (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); } } -/* Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static void * -get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) -{ - return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; -} - - static void i915_surface_fill(struct pipe_context *pipe, struct pipe_surface *dst, @@ -99,50 +93,20 @@ i915_surface_fill(struct pipe_context *pipe, unsigned width, unsigned height, unsigned value) { if (0) { - unsigned i, j; void *dst_map = pipe->screen->surface_map( pipe->screen, dst, PIPE_BUFFER_USAGE_CPU_WRITE ); - - switch (dst->cpp) { - case 1: { - ubyte *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; - case 2: { - ushort *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; - case 4: { - unsigned *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - break; - default: - assert(0); - break; - } + pipe_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value); pipe->screen->surface_unmap(pipe->screen, dst); } else { + assert(dst->block.width == 1); + assert(dst->block.height == 1); i915_fill_blit( i915_context(pipe), - dst->cpp, - (short) dst->pitch, + dst->block.size, + (short) dst->stride/dst->block.size, dst->buffer, dst->offset, (short) dstx, (short) dsty, (short) width, (short) height, diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index b2e490c7db..2815e61345 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -109,6 +109,9 @@ i915_miptree_set_level_info(struct i915_texture *tex, pt->width[level] = w; pt->height[level] = h; pt->depth[level] = d; + + pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w); + pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h); tex->nr_images[level] = nr_images; @@ -140,7 +143,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex, assert(img < tex->nr_images[level]); - tex->image_offset[level][img] = (x + y * tex->pitch); + tex->image_offset[level][img] = y * tex->stride + x * tex->base.block.size; /* printf("%s level %d img %d pos %d,%d image_offset %x\n", @@ -162,7 +165,7 @@ i915_displaytarget_layout(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; - if (pt->last_level > 0 || pt->cpp != 4) + if (pt->last_level > 0 || pt->block.size != 4) return 0; i915_miptree_set_level_info( tex, 0, 1, @@ -172,18 +175,18 @@ i915_displaytarget_layout(struct i915_texture *tex) i915_miptree_set_image_offset( tex, 0, 0, 0, 0 ); if (tex->base.width[0] >= 128) { - tex->pitch = power_of_two(tex->base.width[0] * pt->cpp) / pt->cpp; - tex->total_height = round_up(tex->base.height[0], 8); + tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); + tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); tex->tiled = 1; } else { - tex->pitch = round_up(tex->base.width[0], 64 / pt->cpp); - tex->total_height = tex->base.height[0]; + tex->stride = round_up(tex->base.nblocksx[0] * pt->block.size, 64); + tex->total_nblocksy = tex->base.nblocksy[0]; } /* printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - tex->base.width[0], tex->base.height[0], pt->cpp, - tex->pitch, tex->total_height, tex->pitch * tex->total_height * 4); + tex->base.width[0], tex->base.height[0], pt->block.size, + tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); */ return 1; @@ -193,12 +196,14 @@ static void i945_miptree_layout_2d( struct i915_texture *tex ) { struct pipe_texture *pt = &tex->base; - int align_h = 2, align_w = 4; + const int align_x = 2, align_y = 4; unsigned level; unsigned x = 0; unsigned y = 0; unsigned width = pt->width[0]; unsigned height = pt->height[0]; + unsigned nblocksx = pt->nblocksx[0]; + unsigned nblocksy = pt->nblocksy[0]; #if 0 /* used for tiled display targets */ if (pt->last_level == 0 && pt->cpp == 4) @@ -206,7 +211,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) return; #endif - tex->pitch = pt->width[0]; + tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); /* May need to adjust pitch to accomodate the placement of * the 2nd mipmap level. This occurs when the alignment @@ -214,47 +219,43 @@ i945_miptree_layout_2d( struct i915_texture *tex ) * 2nd mipmap level out past the width of its parent. */ if (pt->last_level > 0) { - unsigned mip1_width = align_int(minify(pt->width[0]), align_w) - + minify(minify(pt->width[0])); + unsigned mip1_nblocksx + = align_int(pf_get_nblocksx(&pt->block, minify(width)), align_x) + + pf_get_nblocksx(&pt->block, minify(minify(width))); - if (mip1_width > pt->width[0]) - tex->pitch = mip1_width; + if (mip1_nblocksx > nblocksx) + tex->stride = mip1_nblocksx * pt->block.size; } - /* Pitch must be a whole number of dwords, even though we - * express it in texels. + /* Pitch must be a whole number of dwords */ - tex->pitch = align_int(tex->pitch * pt->cpp, 64) / pt->cpp; - tex->total_height = 0; + tex->stride = align_int(tex->stride, 64); + tex->total_nblocksy = 0; for (level = 0; level <= pt->last_level; level++) { - unsigned img_height; - i915_miptree_set_level_info(tex, level, 1, width, height, 1); i915_miptree_set_image_offset(tex, level, 0, x, y); - if (pt->compressed) - img_height = MAX2(1, height/4); - else - img_height = align_int(height, align_h); - + nblocksy = align_int(nblocksy, align_y); /* Because the images are packed better, the final offset * might not be the maximal one: */ - tex->total_height = MAX2(tex->total_height, y + img_height); + tex->total_nblocksy = MAX2(tex->total_nblocksy, y + nblocksy); /* Layout_below: step right after second mipmap level. */ if (level == 1) { - x += align_int(width, align_w); + x += align_int(nblocksx, align_x); } else { - y += img_height; + y += nblocksy; } width = minify(width); height = minify(height); + nblocksx = pf_get_nblocksx(&pt->block, width); + nblocksy = pf_get_nblocksy(&pt->block, height); } } @@ -264,15 +265,16 @@ i945_miptree_layout_cube(struct i915_texture *tex) struct pipe_texture *pt = &tex->base; unsigned level; - const unsigned dim = pt->width[0]; + const unsigned nblocks = pt->nblocksx[0]; unsigned face; - unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; + unsigned width = pt->width[0]; + unsigned height = pt->height[0]; /* printf("%s %i, %i\n", __FUNCTION__, pt->width[0], pt->height[0]); */ - assert(lvlWidth == lvlHeight); /* cubemap images are square */ + assert(width == height); /* cubemap images are square */ /* * XXX Should only be used for compressed formats. But lets @@ -282,35 +284,32 @@ i945_miptree_layout_cube(struct i915_texture *tex) * determined either by the old-style packing of cubemap faces, * or the final row of 4x4, 2x2 and 1x1 faces below this. */ - if (dim > 32) - tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; + if (nblocks > 32) + tex->stride = round_up(nblocks * pt->block.size * 2, 4); else - tex->pitch = 14 * 8; + tex->stride = 14 * 8 * pt->block.size; - /* - * XXX The 4 is only needed for compressed formats. See above. - */ - tex->total_height = dim * 4 + 4; + tex->total_nblocksy = nblocks * 4; /* Set all the levels to effectively occupy the whole rectangular region. */ for (level = 0; level <= pt->last_level; level++) { - i915_miptree_set_level_info(tex, level, 6, lvlWidth, lvlHeight, 1); - lvlWidth /= 2; - lvlHeight /= 2; + i915_miptree_set_level_info(tex, level, 6, width, height, 1); + width /= 2; + height /= 2; } for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * dim; - unsigned y = initial_offsets[face][1] * dim; - unsigned d = dim; + unsigned x = initial_offsets[face][0] * nblocks; + unsigned y = initial_offsets[face][1] * nblocks; + unsigned d = nblocks; #if 0 /* Fix and enable this code for compressed formats */ - if (dim == 4 && face >= 4) { + if (nblocks == 4 && face >= 4) { y = tex->total_height - 4; x = (face - 4) * 8; } - else if (dim < 4 && (face > 0)) { + else if (nblocks < 4 && (face > 0)) { y = tex->total_height - 4; x = face * 8; } @@ -369,28 +368,28 @@ i915_miptree_layout(struct i915_texture * tex) switch (pt->target) { case PIPE_TEXTURE_CUBE: { - const unsigned dim = pt->width[0]; + const unsigned nblocks = pt->nblocksx[0]; unsigned face; - unsigned lvlWidth = pt->width[0], lvlHeight = pt->height[0]; + unsigned width = pt->width[0], height = pt->height[0]; - assert(lvlWidth == lvlHeight); /* cubemap images are square */ + assert(width == height); /* cubemap images are square */ /* double pitch for cube layouts */ - tex->pitch = ((dim * pt->cpp * 2 + 3) & ~3) / pt->cpp; - tex->total_height = dim * 4; + tex->stride = round_up(nblocks * pt->block.size * 2, 4); + tex->total_nblocksy = nblocks * 4; for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_level_info(tex, level, 6, - lvlWidth, lvlHeight, + width, height, 1); - lvlWidth /= 2; - lvlHeight /= 2; + width /= 2; + height /= 2; } for (face = 0; face < 6; face++) { - unsigned x = initial_offsets[face][0] * dim; - unsigned y = initial_offsets[face][1] * dim; - unsigned d = dim; + unsigned x = initial_offsets[face][0] * nblocks; + unsigned y = initial_offsets[face][1] * nblocks; + unsigned d = nblocks; for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_image_offset(tex, level, face, x, y); @@ -405,25 +404,29 @@ i915_miptree_layout(struct i915_texture * tex) unsigned width = pt->width[0]; unsigned height = pt->height[0]; unsigned depth = pt->depth[0]; - unsigned stack_height = 0; + unsigned nblocksx = pt->nblocksx[0]; + unsigned nblocksy = pt->nblocksy[0]; + unsigned stack_nblocksy = 0; /* Calculate the size of a single slice. */ - tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; + tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); /* XXX: hardware expects/requires 9 levels at minimum. */ for (level = 0; level <= MAX2(8, pt->last_level); level++) { i915_miptree_set_level_info(tex, level, depth, - width, height, depth); + width, height, depth); - stack_height += MAX2(2, height); + stack_nblocksy += MAX2(2, nblocksy); width = minify(width); height = minify(height); depth = minify(depth); + nblocksx = pf_get_nblocksx(&pt->block, width); + nblocksy = pf_get_nblocksy(&pt->block, height); } /* Fixup depth image_offsets: @@ -433,7 +436,7 @@ i915_miptree_layout(struct i915_texture * tex) unsigned i; for (i = 0; i < depth; i++) i915_miptree_set_image_offset(tex, level, i, - 0, i * stack_height); + 0, i * stack_nblocksy); depth = minify(depth); } @@ -443,33 +446,33 @@ i915_miptree_layout(struct i915_texture * tex) * remarkable how wasteful of memory the i915 texture layouts * are. They are largely fixed in the i945. */ - tex->total_height = stack_height * pt->depth[0]; + tex->total_nblocksy = stack_nblocksy * pt->depth[0]; break; } default:{ unsigned width = pt->width[0]; unsigned height = pt->height[0]; - unsigned img_height; + unsigned nblocksx = pt->nblocksx[0]; + unsigned nblocksy = pt->nblocksy[0]; - tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; - tex->total_height = 0; + tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); + tex->total_nblocksy = 0; for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_level_info(tex, level, 1, - width, height, 1); + width, height, 1); i915_miptree_set_image_offset(tex, level, 0, - 0, tex->total_height); + 0, tex->total_nblocksy); - if (pt->compressed) - img_height = MAX2(1, height / 4); - else - img_height = (MAX2(2, height) + 1) & ~1; + nblocksy = round_up(MAX2(2, nblocksy), 2); - tex->total_height += img_height; + tex->total_nblocksy += nblocksy; width = minify(width); height = minify(height); + nblocksx = pf_get_nblocksx(&pt->block, width); + nblocksy = pf_get_nblocksy(&pt->block, height); } break; } @@ -477,7 +480,7 @@ i915_miptree_layout(struct i915_texture * tex) /* DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, tex->pitch, - tex->total_height, pt->cpp, tex->pitch * tex->total_height * pt->cpp); + tex->total_nblocksy, pt->block.size, tex->stride * tex->total_nblocksy); */ return TRUE; @@ -498,14 +501,16 @@ i945_miptree_layout(struct i915_texture * tex) unsigned width = pt->width[0]; unsigned height = pt->height[0]; unsigned depth = pt->depth[0]; + unsigned nblocksx = pt->nblocksx[0]; + unsigned nblocksy = pt->nblocksy[0]; unsigned pack_x_pitch, pack_x_nr; unsigned pack_y_pitch; - tex->pitch = ((pt->width[0] * pt->cpp + 3) & ~3) / pt->cpp; - tex->total_height = 0; + tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); + tex->total_nblocksy = 0; - pack_y_pitch = MAX2(pt->height[0], 2); - pack_x_pitch = tex->pitch; + pack_y_pitch = MAX2(pt->nblocksy[0], 2); + pack_x_pitch = tex->stride / pt->block.size; pack_x_nr = 1; for (level = 0; level <= pt->last_level; level++) { @@ -515,11 +520,11 @@ i945_miptree_layout(struct i915_texture * tex) unsigned q, j; i915_miptree_set_level_info(tex, level, nr_images, - width, height, depth); + width, height, depth); for (q = 0; q < nr_images;) { for (j = 0; j < pack_x_nr && q < nr_images; j++, q++) { - i915_miptree_set_image_offset(tex, level, q, x, y + tex->total_height); + i915_miptree_set_image_offset(tex, level, q, x, y + tex->total_nblocksy); x += pack_x_pitch; } @@ -528,12 +533,12 @@ i945_miptree_layout(struct i915_texture * tex) } - tex->total_height += y; + tex->total_nblocksy += y; if (pack_x_pitch > 4) { pack_x_pitch >>= 1; pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= tex->pitch); + assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride); } if (pack_y_pitch > 2) { @@ -543,6 +548,8 @@ i945_miptree_layout(struct i915_texture * tex) width = minify(width); height = minify(height); depth = minify(depth); + nblocksx = pf_get_nblocksx(&pt->block, width); + nblocksy = pf_get_nblocksy(&pt->block, height); } break; } @@ -560,7 +567,7 @@ i945_miptree_layout(struct i915_texture * tex) /* DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, tex->pitch, - tex->total_height, pt->cpp, tex->pitch * tex->total_height * pt->cpp); + tex->total_nblocksy, pt->block.size, tex->stride * tex->total_nblocksy); */ return TRUE; @@ -582,6 +589,9 @@ i915_texture_create(struct pipe_screen *screen, tex->base.refcount = 1; tex->base.screen = screen; + tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width[0]); + tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height[0]); + if (i915screen->is_i945) { if (!i945_miptree_layout(tex)) goto fail; @@ -592,8 +602,8 @@ i915_texture_create(struct pipe_screen *screen, tex->buffer = ws->buffer_create(ws, 64, PIPE_BUFFER_USAGE_PIXEL, - tex->pitch * tex->base.cpp * - tex->total_height); + tex->stride * + tex->total_nblocksy); if (!tex->buffer) goto fail; @@ -648,13 +658,13 @@ i915_get_tex_surface(struct pipe_screen *screen, unsigned offset; /* in bytes */ if (pt->target == PIPE_TEXTURE_CUBE) { - offset = tex->image_offset[level][face] * pt->cpp; + offset = tex->image_offset[level][face]; } else if (pt->target == PIPE_TEXTURE_3D) { - offset = tex->image_offset[level][zslice] * pt->cpp; + offset = tex->image_offset[level][zslice]; } else { - offset = tex->image_offset[level][0] * pt->cpp; + offset = tex->image_offset[level][0]; assert(face == 0); assert(zslice == 0); } @@ -666,10 +676,12 @@ i915_get_tex_surface(struct pipe_screen *screen, pipe_texture_reference(&ps->texture, pt); pipe_buffer_reference(ws, &ps->buffer, tex->buffer); ps->format = pt->format; - ps->cpp = pt->cpp; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->pitch = tex->pitch; + ps->block = pt->block; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = tex->stride; ps->offset = offset; ps->usage = flags; ps->status = PIPE_SURFACE_STATUS_DEFINED; @@ -680,7 +692,7 @@ i915_get_tex_surface(struct pipe_screen *screen, static struct pipe_texture * i915_texture_blanket(struct pipe_screen * screen, const struct pipe_texture *base, - const unsigned *pitch, + const unsigned *stride, struct pipe_buffer *buffer) { struct i915_texture *tex; @@ -699,7 +711,7 @@ i915_texture_blanket(struct pipe_screen * screen, tex->base = *base; - tex->pitch = pitch[0]; + tex->stride = stride[0]; i915_miptree_set_level_info(tex, 0, 1, base->width[0], base->height[0], 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index 8ac6b4e689..2cae7665f7 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -231,9 +231,9 @@ struct brw_texture { /* Derived from the above: */ - unsigned pitch; + unsigned stride; unsigned depth_pitch; /* per-image on i945? */ - unsigned total_height; + unsigned total_nblocksy; unsigned nr_images[PIPE_MAX_TEXTURE_LEVELS]; diff --git a/src/gallium/drivers/i965simple/brw_misc_state.c b/src/gallium/drivers/i965simple/brw_misc_state.c index 925049ecc1..be812c5da9 100644 --- a/src/gallium/drivers/i965simple/brw_misc_state.c +++ b/src/gallium/drivers/i965simple/brw_misc_state.c @@ -224,7 +224,9 @@ static void upload_depthbuffer(struct brw_context *brw) } else { unsigned int format; - switch (depth_surface->cpp) { + assert(depth_surface->block.width == 1); + assert(depth_surface->block.height == 1); + switch (depth_surface->block.size) { case 2: format = BRW_DEPTHFORMAT_D16_UNORM; break; @@ -239,7 +241,7 @@ static void upload_depthbuffer(struct brw_context *brw) return; } - OUT_BATCH(((depth_surface->pitch * depth_surface->cpp) - 1) | + OUT_BATCH((depth_surface->stride - 1) | (format << 18) | (BRW_TILEWALK_YMAJOR << 26) | // (depth_surface->region->tiled << 27) | @@ -247,7 +249,7 @@ static void upload_depthbuffer(struct brw_context *brw) OUT_RELOC(depth_surface->buffer, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE, 0); OUT_BATCH((BRW_SURFACE_MIPMAPLAYOUT_BELOW << 1) | - ((depth_surface->pitch - 1) << 6) | + ((depth_surface->stride/depth_surface->block.size - 1) << 6) | ((depth_surface->height - 1) << 19)); OUT_BATCH(0); } diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index 3d98a2bf19..0be3dfc743 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -47,8 +47,10 @@ brw_surface_copy(struct pipe_context *pipe, struct pipe_surface *src, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { - assert(dst != src); - assert(dst->cpp == src->cpp); + assert( dst != src ); + assert( dst->block.size == src->block.size ); + assert( dst->block.width == src->block.height ); + assert( dst->block.height == src->block.height ); if (0) { void *dst_map = pipe->screen->surface_map( pipe->screen, @@ -60,37 +62,30 @@ brw_surface_copy(struct pipe_context *pipe, PIPE_BUFFER_USAGE_CPU_READ ); pipe_copy_rect(dst_map, - dst->cpp, - dst->pitch, + &dst->block, + dst->stride, dstx, dsty, width, height, src_map, - do_flip ? -(int) src->pitch : src->pitch, + do_flip ? -(int) src->stride : src->stride, srcx, do_flip ? height - 1 - srcy : srcy); pipe->screen->surface_unmap(pipe->screen, src); pipe->screen->surface_unmap(pipe->screen, dst); } else { + assert(dst->block.width == 1); + assert(dst->block.height == 1); brw_copy_blit(brw_context(pipe), do_flip, - dst->cpp, - (short) src->pitch, src->buffer, src->offset, FALSE, - (short) dst->pitch, dst->buffer, dst->offset, FALSE, + dst->block.size, + (short) src->stride/src->block.size, src->buffer, src->offset, FALSE, + (short) dst->stride/dst->block.size, dst->buffer, dst->offset, FALSE, (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height, PIPE_LOGICOP_COPY); } } -/* Fill a rectangular sub-region. Need better logic about when to - * push buffers into AGP - will currently do so whenever possible. - */ -static void * -get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) -{ - return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; -} - static void brw_surface_fill(struct pipe_context *pipe, @@ -99,50 +94,20 @@ brw_surface_fill(struct pipe_context *pipe, unsigned width, unsigned height, unsigned value) { if (0) { - unsigned i, j; void *dst_map = pipe->screen->surface_map( pipe->screen, dst, PIPE_BUFFER_USAGE_CPU_WRITE ); - - switch (dst->cpp) { - case 1: { - ubyte *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; - case 2: { - ushort *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; - case 4: { - unsigned *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } - break; - default: - assert(0); - break; - } + pipe_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value); pipe->screen->surface_unmap(pipe->screen, dst); } else { + assert(dst->block.width == 1); + assert(dst->block.height == 1); brw_fill_blit(brw_context(pipe), - dst->cpp, - (short) dst->pitch, + dst->block.size, + (short) dst->stride/dst->block.size, dst->buffer, dst->offset, FALSE, (short) dstx, (short) dsty, (short) width, (short) height, diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 78ae0b1223..8c7725605b 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -81,7 +81,7 @@ static void intel_miptree_set_image_offset(struct brw_texture *tex, assert(x == 0 && y == 0); assert(img < tex->nr_images[level]); - tex->image_offset[level][img] = (x + y * tex->pitch) * pt->cpp; + tex->image_offset[level][img] = y * tex->stride + x * pt->block.size; } static void intel_miptree_set_level_info(struct brw_texture *tex, @@ -97,8 +97,11 @@ static void intel_miptree_set_level_info(struct brw_texture *tex, pt->width[level] = w; pt->height[level] = h; pt->depth[level] = d; + + pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w); + pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h); - tex->level_offset[level] = (x + y * tex->pitch) * pt->cpp; + tex->level_offset[level] = y * tex->stride + x * tex->base.block.size; tex->nr_images[level] = nr_images; /* @@ -123,77 +126,60 @@ static void intel_miptree_set_level_info(struct brw_texture *tex, static void i945_miptree_layout_2d(struct brw_texture *tex) { struct pipe_texture *pt = &tex->base; - unsigned align_h = 2, align_w = 4; + const int align_x = 2, align_y = 4; unsigned level; unsigned x = 0; unsigned y = 0; unsigned width = pt->width[0]; unsigned height = pt->height[0]; + unsigned nblocksx = pt->nblocksx[0]; + unsigned nblocksy = pt->nblocksy[0]; - tex->pitch = pt->width[0]; - -#if 0 - if (pt->compressed) { - align_w = intel_compressed_alignment(pt->internal_format); - tex->pitch = ALIGN(pt->width[0], align_w); - } -#endif + tex->stride = align(pt->nblocksx[0] * pt->block.size, 4); /* May need to adjust pitch to accomodate the placement of - * the 2nd mipmap. This occurs when the alignment + * the 2nd mipmap level. This occurs when the alignment * constraints of mipmap placement push the right edge of the - * 2nd mipmap out past the width of its parent. + * 2nd mipmap level out past the width of its parent. */ if (pt->last_level > 0) { - unsigned mip1_width; - - if (pt->compressed) { - mip1_width = align(minify(pt->width[0]), align_w) - + align(minify(minify(pt->width[0])), align_w); - } else { - mip1_width = align(minify(pt->width[0]), align_w) - + minify(minify(pt->width[0])); - } + unsigned mip1_nblocksx + = align_int(pf_get_nblocksx(&pt->block, minify(width)), align_x) + + pf_get_nblocksx(&pt->block, minify(minify(width))); - if (mip1_width > tex->pitch) { - tex->pitch = mip1_width; - } + if (mip1_nblocksx > nblocksx) + tex->stride = mip1_nblocksx * pt->block.size; } - /* Pitch must be a whole number of dwords, even though we - * express it in texels. + /* Pitch must be a whole number of dwords */ - tex->pitch = align(tex->pitch * pt->cpp, 4) / pt->cpp; - tex->total_height = 0; + tex->stride = align_int(tex->stride, 64); + tex->total_nblocksy = 0; for (level = 0; level <= pt->last_level; level++) { - unsigned img_height; - intel_miptree_set_level_info(tex, level, 1, x, y, width, height, 1); - if (pt->compressed) - img_height = MAX2(1, height/4); - else - img_height = align(height, align_h); - + nblocksy = align_int(nblocksy, align_y); /* Because the images are packed better, the final offset * might not be the maximal one: */ - tex->total_height = MAX2(tex->total_height, y + img_height); + tex->total_nblocksy = MAX2(tex->total_nblocksy, y + nblocksy); - /* Layout_below: step right after second mipmap. + /* Layout_below: step right after second mipmap level. */ if (level == 1) { - x += align(width, align_w); + x += align_int(nblocksx, align_x); } else { - y += img_height; + y += nblocksy; } width = minify(width); height = minify(height); + nblocksx = pf_get_nblocksx(&pt->block, width); + nblocksy = pf_get_nblocksy(&pt->block, height); } } @@ -210,26 +196,20 @@ static boolean brw_miptree_layout(struct brw_texture *tex) unsigned width = pt->width[0]; unsigned height = pt->height[0]; unsigned depth = pt->depth[0]; + unsigned nblocksx = pt->nblocksx[0]; + unsigned nblocksy = pt->nblocksy[0]; unsigned pack_x_pitch, pack_x_nr; unsigned pack_y_pitch; unsigned level; unsigned align_h = 2; unsigned align_w = 4; - tex->total_height = 0; -#if 0 - if (pt->compressed) { - align_w = intel_compressed_alignment(pt->internal_format); - pt->pitch = align(width, align_w); - pack_y_pitch = (height + 3) / 4; - } else -#endif - { - tex->pitch = align(pt->width[0] * pt->cpp, 4) / pt->cpp; - pack_y_pitch = align(pt->height[0], align_h); - } + tex->total_nblocksy = 0; + + tex->stride = align(pt->nblocksx[0], 4); + pack_y_pitch = align(pt->nblocksy[0], align_h); - pack_x_pitch = tex->pitch; + pack_x_pitch = tex->stride / pt->block.size; pack_x_nr = 1; for (level = 0; level <= pt->last_level; level++) { @@ -239,7 +219,7 @@ static boolean brw_miptree_layout(struct brw_texture *tex) uint q, j; intel_miptree_set_level_info(tex, level, nr_images, - 0, tex->total_height, + 0, tex->total_nblocksy, width, height, depth); for (q = 0; q < nr_images;) { @@ -253,10 +233,12 @@ static boolean brw_miptree_layout(struct brw_texture *tex) } - tex->total_height += y; + tex->total_nblocksy += y; width = minify(width); height = minify(height); depth = minify(depth); + nblocksx = pf_get_nblocksx(&pt->block, width); + nblocksy = pf_get_nblocksy(&pt->block, height); if (pt->compressed) { pack_y_pitch = (height + 3) / 4; @@ -269,7 +251,7 @@ static boolean brw_miptree_layout(struct brw_texture *tex) if (pack_x_pitch > 4) { pack_x_pitch >>= 1; pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr <= tex->pitch); + assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride); } if (pack_y_pitch > 2) { @@ -289,9 +271,9 @@ static boolean brw_miptree_layout(struct brw_texture *tex) #if 0 PRINT("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, pt->pitch, - pt->total_height, - pt->cpp, - pt->pitch * pt->total_height * pt->cpp ); + pt->total_nblocksy, + pt->block.size, + pt->stride * pt->total_nblocksy ); #endif return TRUE; @@ -309,11 +291,14 @@ brw_texture_create_screen(struct pipe_screen *screen, tex->base = *templat; tex->base.refcount = 1; + tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width[0]); + tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height[0]); + if (brw_miptree_layout(tex)) tex->buffer = ws->buffer_create(ws, 64, PIPE_BUFFER_USAGE_PIXEL, - tex->pitch * tex->base.cpp * - tex->total_height); + tex->stride * + tex->total_nblocksy); if (!tex->buffer) { FREE(tex); @@ -370,10 +355,10 @@ brw_get_tex_surface_screen(struct pipe_screen *screen, offset = tex->level_offset[level]; if (pt->target == PIPE_TEXTURE_CUBE) { - offset += tex->image_offset[level][face] * pt->cpp; + offset += tex->image_offset[level][face]; } else if (pt->target == PIPE_TEXTURE_3D) { - offset += tex->image_offset[level][zslice] * pt->cpp; + offset += tex->image_offset[level][zslice]; } else { assert(face == 0); @@ -386,10 +371,12 @@ brw_get_tex_surface_screen(struct pipe_screen *screen, assert(ps->refcount); pipe_buffer_reference(ws, &ps->buffer, tex->buffer); ps->format = pt->format; - ps->cpp = pt->cpp; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->pitch = tex->pitch; + ps->block = pt->block; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = tex->stride; ps->offset = offset; } return ps; diff --git a/src/gallium/drivers/i965simple/brw_wm_surface_state.c b/src/gallium/drivers/i965simple/brw_wm_surface_state.c index 69e56dc8bd..1a326f9918 100644 --- a/src/gallium/drivers/i965simple/brw_wm_surface_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_surface_state.c @@ -160,7 +160,7 @@ void brw_update_texture_surface( struct brw_context *brw, surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; surf.ss3.tiled_surface = 0; /* always zero */ - surf.ss3.pitch = tObj->pitch - 1; + surf.ss3.pitch = tObj->stride - 1; surf.ss3.depth = tObj->base.depth[0] - 1; surf.ss4.min_lod = 0; @@ -197,7 +197,7 @@ static void upload_wm_surfaces(struct brw_context *brw ) memset(&surf, 0, sizeof(surf)); if (pipe_surface != NULL) { - if (pipe_surface->cpp == 4) + if (pipe_surface->block.size == 4) surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; else surf.ss0.surface_format = BRW_SURFACEFORMAT_B5G6R5_UNORM; @@ -210,7 +210,7 @@ static void upload_wm_surfaces(struct brw_context *brw ) surf.ss2.height = pipe_surface->height - 1; surf.ss3.tile_walk = BRW_TILEWALK_XMAJOR; surf.ss3.tiled_surface = 0; - surf.ss3.pitch = (pipe_surface->pitch * pipe_surface->cpp) - 1; + surf.ss3.pitch = pipe_surface->stride - 1; } else { surf.ss0.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; surf.ss0.surface_type = BRW_SURFACE_NULL; diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index 9fd48aeccc..7dc15c38d1 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -58,18 +58,20 @@ sp_surface_copy(struct pipe_context *pipe, src, PIPE_BUFFER_USAGE_CPU_READ ); - assert(dst->cpp == src->cpp); + assert(dst->block.size == src->block.size); + assert(dst->block.width == src->block.width); + assert(dst->block.height == src->block.height); assert(src_map); assert(dst_map); /* If do_flip, invert src_y position and pass negative src stride */ pipe_copy_rect(dst_map, - dst->cpp, - dst->pitch, + &dst->block, + dst->stride, dstx, dsty, width, height, src_map, - do_flip ? -(int) src->pitch : src->pitch, + do_flip ? -(int) src->stride : src->stride, srcx, do_flip ? src->height - 1 - srcy : srcy); pipe->screen->surface_unmap(pipe->screen, src); @@ -80,7 +82,7 @@ sp_surface_copy(struct pipe_context *pipe, static void * get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) { - return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; + return (char *)dst_map + y / dst->block.height * dst->stride + x / dst->block.width * dst->block.size; } @@ -102,39 +104,14 @@ sp_surface_fill(struct pipe_context *pipe, dst, PIPE_BUFFER_USAGE_CPU_WRITE ); - assert(dst->pitch > 0); - assert(width <= dst->pitch); + assert(dst->stride > 0); - switch (dst->cpp) { + switch (dst->block.size) { case 1: - { - ubyte *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; case 2: - { - ushort *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; case 4: - { - unsigned *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } + pipe_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value); break; case 8: { @@ -155,7 +132,7 @@ sp_surface_fill(struct pipe_context *pipe, row[j*4+2] = val2; row[j*4+3] = val3; } - row += dst->pitch * 4; + row += dst->stride/2; } } break; diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 2ef17a220b..4db045cdc3 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -71,13 +71,15 @@ softpipe_texture_layout(struct pipe_screen *screen, pt->width[level] = width; pt->height[level] = height; pt->depth[level] = depth; - spt->pitch[level] = width; + pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width); + pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height); + spt->stride[level] = pt->nblocksx[level]*pt->block.size; spt->level_offset[level] = buffer_size; - buffer_size += (((pt->compressed) ? MAX2(1, height/4) : height) * + buffer_size += (pt->nblocksy[level] * ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * - width * pt->cpp); + spt->stride[level]); width = minify(width); height = minify(height); @@ -121,7 +123,7 @@ softpipe_displaytarget_layout(struct pipe_screen *screen, /* Now extract the goodies: */ spt->buffer = surf.buffer; - spt->pitch[0] = surf.pitch; + spt->stride[0] = surf.stride; return spt->buffer != NULL; } @@ -195,10 +197,12 @@ softpipe_get_tex_surface(struct pipe_screen *screen, assert(ps->winsys); pipe_buffer_reference(ws, &ps->buffer, spt->buffer); ps->format = pt->format; - ps->cpp = pt->cpp; + ps->block = pt->block; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->pitch = spt->pitch[level]; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = spt->stride[level]; ps->offset = spt->level_offset[level]; ps->usage = usage; @@ -228,8 +232,8 @@ softpipe_get_tex_surface(struct pipe_screen *screen, if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - (pt->compressed ? ps->height/4 : ps->height) * - ps->width * ps->cpp; + ps->nblocksy * + ps->stride; } else { assert(face == 0); diff --git a/src/gallium/drivers/softpipe/sp_texture.h b/src/gallium/drivers/softpipe/sp_texture.h index 0e1017632c..bf437a7c61 100644 --- a/src/gallium/drivers/softpipe/sp_texture.h +++ b/src/gallium/drivers/softpipe/sp_texture.h @@ -42,7 +42,7 @@ struct softpipe_texture struct pipe_texture base; unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; - unsigned long pitch[PIPE_MAX_TEXTURE_LEVELS]; + unsigned long stride[PIPE_MAX_TEXTURE_LEVELS]; /* The data is held here: */ diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index d973fb357b..a2c6155d01 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -501,6 +501,47 @@ pf_get_block(enum pipe_format format, struct pipe_format_block *block) } } +static INLINE unsigned +pf_get_nblocksx(const struct pipe_format_block *block, unsigned x) +{ + return (x + block->width - 1)/block->width; +} + +static INLINE unsigned +pf_get_nblocksy(const struct pipe_format_block *block, unsigned y) +{ + return (y + block->height - 1)/block->height; +} + +static INLINE unsigned +pf_get_nblocks(const struct pipe_format_block *block, unsigned width, unsigned height) +{ + return pf_get_nblocksx(block, width)*pf_get_nblocksy(block, height); +} + +static INLINE void +pipe_rect_to_blocks(const struct pipe_format_block *block, + unsigned *width, unsigned *height, + unsigned *src_x, unsigned *src_y, + unsigned *dst_x, unsigned *dst_y) +{ + assert(block->size > 0); + assert(block->width > 0); + assert(block->height > 0); + if(width) + *width = pf_get_nblocksx(block, *width); + if(height) + *height = pf_get_nblocksy(block, *height); + if(src_x) + *src_x /= block->width; + if(src_y) + *src_y /= block->height; + if(dst_x) + *dst_x /= block->width; + if(dst_y) + *dst_y /= block->height; +} + #ifdef __cplusplus } #endif diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index e7ee8c97ed..2992e2f3b3 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -267,10 +267,12 @@ struct pipe_surface enum pipe_format format; /**< PIPE_FORMAT_x */ unsigned status; /**< PIPE_SURFACE_STATUS_x */ unsigned clear_value; /**< XXX may be temporary */ - unsigned cpp; /**< bytes per pixel */ unsigned width; unsigned height; - unsigned pitch; /**< in pixels */ + struct pipe_format_block block; + unsigned nblocksx; + unsigned nblocksy; + unsigned stride; /**< in bytes */ unsigned layout; /**< PIPE_SURFACE_LAYOUT_x */ unsigned offset; /**< offset from start of buffer, in bytes */ unsigned refcount; @@ -303,7 +305,10 @@ struct pipe_texture unsigned height[PIPE_MAX_TEXTURE_LEVELS]; unsigned depth[PIPE_MAX_TEXTURE_LEVELS]; - unsigned cpp:8; + struct pipe_format_block block; + unsigned nblocksx[PIPE_MAX_TEXTURE_LEVELS]; + unsigned nblocksy[PIPE_MAX_TEXTURE_LEVELS]; + unsigned last_level:8; /**< Index of last mipmap level present/defined */ unsigned compressed:1; diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index cf2447822a..7dcdd28287 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -31,6 +31,7 @@ #include "p_config.h" #include "p_compiler.h" #include "p_debug.h" +#include "p_format.h" #include "p_pointer.h" #include #include @@ -401,11 +402,15 @@ static INLINE int align(int value, int alignment) /* util/p_util.c */ -extern void pipe_copy_rect(ubyte * dst, unsigned cpp, unsigned dst_pitch, - unsigned dst_x, unsigned dst_y, unsigned width, - unsigned height, const ubyte * src, - int src_pitch, unsigned src_x, int src_y); - +extern void pipe_copy_rect(ubyte * dst, const struct pipe_format_block *block, + unsigned dst_stride, unsigned dst_x, unsigned dst_y, + unsigned width, unsigned height, const ubyte * src, + int src_stride, unsigned src_x, int src_y); + +extern void +pipe_fill_rect(ubyte * dst, const struct pipe_format_block *block, + unsigned dst_stride, unsigned dst_x, unsigned dst_y, + unsigned width, unsigned height, uint32_t value); #if defined(_MSC_VER) diff --git a/src/gallium/winsys/xlib/brw_aub.c b/src/gallium/winsys/xlib/brw_aub.c index 10eedd8402..6e814ce5d1 100644 --- a/src/gallium/winsys/xlib/brw_aub.c +++ b/src/gallium/winsys/xlib/brw_aub.c @@ -322,7 +322,10 @@ void brw_aub_dump_bmp( struct brw_aubfile *aubfile, struct aub_dump_bmp db; unsigned format; - if (surface->cpp == 4) + assert(surface->block.width == 1); + assert(surface->block.height == 1); + + if (surface->block.size == 4) format = 0x7; else format = 0x3; @@ -331,8 +334,8 @@ void brw_aub_dump_bmp( struct brw_aubfile *aubfile, db.xmin = 0; db.ymin = 0; db.format = format; - db.bpp = surface->cpp * 8; - db.pitch = surface->pitch; + db.bpp = surface->block.size * 8; + db.pitch = surface->stride/surface->block.size; db.xsize = surface->width; db.ysize = surface->height; db.addr = gtt_offset; diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index b14758f333..9225ee510d 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -364,9 +364,10 @@ xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) return; } - if (XSHM_ENABLED(xm_buf) && (xm_buf->tempImage == NULL)) { - alloc_shm_ximage(xm_buf, b, surf->pitch, surf->height); + assert(surf->block.width == 1); + assert(surf->block.height == 1); + alloc_shm_ximage(xm_buf, b, surf->stride/surf->block.size, surf->height); } ximage = (XSHM_ENABLED(xm_buf)) ? xm_buf->tempImage : b->tempImage; @@ -386,7 +387,7 @@ xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) /* update XImage's fields */ ximage->width = surf->width; ximage->height = surf->height; - ximage->bytes_per_line = surf->pitch * surf->cpp; + ximage->bytes_per_line = surf->stride; XPutImage(b->xm_visual->display, b->drawable, b->gc, ximage, 0, 0, 0, 0, surf->width, surf->height); @@ -497,18 +498,21 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys, surf->width = width; surf->height = height; surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = round_up(width, alignment / surf->cpp); + pf_get_block(format, &surf->block); + surf->nblocksx = pf_get_nblocksx(&surf->block, width); + surf->nblocksy = pf_get_nblocksy(&surf->block, height); + surf->stride = round_up(surf->nblocksx * surf->block.size, alignment); surf->usage = flags; -#ifdef GALLIUM_CELL /* XXX a bit of a hack */ - height = round_up(height, TILE_SIZE); -#endif - assert(!surf->buffer); surf->buffer = winsys->buffer_create(winsys, alignment, PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * height); +#ifdef GALLIUM_CELL /* XXX a bit of a hack */ + surf->stride * round_up(surf->nblocksy, TILE_SIZE)); +#else + surf->stride * surf->nblocksy); +#endif + if(!surf->buffer) return -1; diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c index 77376099f0..7fc9debdd5 100644 --- a/src/gallium/winsys/xlib/xm_winsys_aub.c +++ b/src/gallium/winsys/xlib/xm_winsys_aub.c @@ -279,22 +279,25 @@ aub_i915_surface_alloc_storage(struct pipe_winsys *winsys, unsigned flags, unsigned tex_usage) { - const unsigned alignment = 64; - - surf->width = width; - surf->height = height; - surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = round_up(width, alignment / surf->cpp); - - assert(!surf->buffer); - surf->buffer = winsys->buffer_create(winsys, alignment, - PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * height); + const unsigned alignment = 64; + + surf->width = width; + surf->height = height; + surf->format = format; + pf_get_block(format, &surf->block); + surf->nblocksx = pf_get_nblocksx(&surf->block, width); + surf->nblocksy = pf_get_nblocksy(&surf->block, height); + surf->stride = round_up(surf->nblocksx * surf->block.size, alignment); + surf->usage = flags; + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->stride * surf->nblocksy); if(!surf->buffer) - return -1; + return -1; - return 0; + return 0; } static void diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 8098d75e18..809a906ba3 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -66,15 +66,17 @@ acc_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps, uint x, uint y, uint w, uint h, float *p) { const enum pipe_format f = acc_ps->format; - const int cpp = acc_ps->cpp; + const struct pipe_format_block b = acc_ps->block; acc_ps->format = DEFAULT_ACCUM_PIPE_FORMAT; - acc_ps->cpp = 8; + acc_ps->block.size = 8; + acc_ps->block.width = 1; + acc_ps->block.height = 1; pipe_get_tile_rgba(pipe, acc_ps, x, y, w, h, p); acc_ps->format = f; - acc_ps->cpp = cpp; + acc_ps->block = b; } @@ -88,15 +90,17 @@ acc_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps, uint x, uint y, uint w, uint h, const float *p) { enum pipe_format f = acc_ps->format; - const int cpp = acc_ps->cpp; + const struct pipe_format_block b = acc_ps->block; acc_ps->format = DEFAULT_ACCUM_PIPE_FORMAT; - acc_ps->cpp = 8; + acc_ps->block.size = 8; + acc_ps->block.width = 1; + acc_ps->block.height = 1; pipe_put_tile_rgba(pipe, acc_ps, x, y, w, h, p); acc_ps->format = f; - acc_ps->cpp = cpp; + acc_ps->block = b; } @@ -111,7 +115,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) const GLint ypos = ctx->DrawBuffer->_Ymin; const GLint width = ctx->DrawBuffer->_Xmax - xpos; const GLint height = ctx->DrawBuffer->_Ymax - ypos; - GLvoid *map; + GLubyte *map; acc_ps = screen->get_tex_surface(screen, acc_strb->texture, 0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -128,8 +132,7 @@ st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) GLshort a = FLOAT_TO_SHORT(ctx->Accum.ClearColor[3]); int i, j; for (i = 0; i < height; i++) { - GLshort *dst = ((GLshort *) map - + ((ypos + i) * acc_ps->pitch + xpos) * 4); + GLshort *dst = (GLshort *) (map + (ypos + i) * acc_ps->stride + xpos * 8); for (j = 0; j < width; j++) { dst[0] = r; dst[1] = g; @@ -168,8 +171,7 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias, { int i, j; for (i = 0; i < height; i++) { - GLshort *acc = ((GLshort *) map - + ((ypos + i) * acc_ps->pitch + xpos) * 4); + GLshort *acc = (GLshort *) (map + (ypos + i) * acc_ps->stride + xpos * 8); for (j = 0; j < width * 4; j++) { float val = SHORT_TO_FLOAT(acc[j]) * scale + bias; acc[j] = FLOAT_TO_SHORT(val); diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index 6fa3cbd533..f3bc3b8584 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -341,9 +341,9 @@ make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height, dest = screen->surface_map(screen, surface, PIPE_BUFFER_USAGE_CPU_WRITE); /* Put image into texture surface */ - memset(dest, 0xff, height * surface->pitch); + memset(dest, 0xff, height * surface->stride); unpack_bitmap(ctx->st, 0, 0, width, height, unpack, bitmap, - dest, surface->pitch); + dest, surface->stride); _mesa_unmap_bitmap_pbo(ctx, unpack); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index d8f1d2367c..a7781f3ab2 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -382,7 +382,7 @@ make_texture(struct st_context *st, mformat, /* gl_texture_format */ dest, /* dest */ 0, 0, 0, /* dstX/Y/Zoffset */ - surface->pitch * cpp, /* dstRowStride, bytes */ + surface->stride, /* dstRowStride, bytes */ &dstImageOffsets, /* dstImageOffsets */ width, height, 1, /* size */ format, type, /* src format/type */ @@ -786,13 +786,13 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y, switch (ps->format) { case PIPE_FORMAT_S8_UNORM: { - ubyte *dest = stmap + spanY * ps->pitch + spanX; + ubyte *dest = stmap + spanY * ps->stride + spanX; memcpy(dest, values, spanWidth); } break; case PIPE_FORMAT_S8Z24_UNORM: { - uint *dest = (uint *) stmap + spanY * ps->pitch + spanX; + uint *dest = (uint *) (stmap + spanY * ps->stride + spanX*4); GLint k; for (k = 0; k < spanWidth; k++) { uint p = dest[k]; @@ -903,6 +903,9 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, psDraw = screen->get_tex_surface(screen, rbDraw->texture, 0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE); + assert(psDraw->block.width == 1); + assert(psDraw->block.height == 1); + /* map the stencil buffer */ drawMap = screen->surface_map(screen, psDraw, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -919,7 +922,7 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, y = ctx->DrawBuffer->Height - y - 1; } - dst = drawMap + (y * psDraw->pitch + dstx) * psDraw->cpp; + dst = drawMap + y * psDraw->stride + dstx * psDraw->block.size; src = buffer + i * width; switch (psDraw->format) { diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index db25ddd615..1067caf9b3 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -113,7 +113,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, template.target = PIPE_TEXTURE_2D; template.compressed = 0; - template.cpp = pf_get_size(template.format); + pf_get_block(template.format, &template.block); template.width[0] = width; template.height[0] = height; template.depth[0] = 1; @@ -171,10 +171,12 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, assert(strb->surface->buffer); assert(strb->surface->format); - assert(strb->surface->cpp); + assert(strb->surface->block.size); + assert(strb->surface->block.width); + assert(strb->surface->block.height); assert(strb->surface->width == width); assert(strb->surface->height == height); - assert(strb->surface->pitch); + assert(strb->surface->stride); return strb->surface != NULL; diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index c5193631a7..09d9c29e44 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -94,13 +94,13 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, switch (ps->format) { case PIPE_FORMAT_S8_UNORM: { - const ubyte *src = stmap + srcY * ps->pitch + x; + const ubyte *src = stmap + srcY * ps->stride + x; memcpy(values, src, width); } break; case PIPE_FORMAT_S8Z24_UNORM: { - const uint *src = (uint *) stmap + srcY * ps->pitch + x; + const uint *src = (uint *) (stmap + srcY * ps->stride + x*4); GLint k; for (k = 0; k < width; k++) { values[k] = src[k] >> 24; @@ -109,7 +109,7 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y, break; case PIPE_FORMAT_Z24S8_UNORM: { - const uint *src = (uint *) stmap + srcY * ps->pitch + x; + const uint *src = (uint *) (stmap + srcY * ps->stride + x*4); GLint k; for (k = 0; k < width; k++) { values[k] = src[k] & 0xff; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 7d52d1da1b..b9aa513d72 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -658,7 +658,7 @@ st_TexImage(GLcontext * ctx, texImage->Data = st_texture_image_map(ctx->st, stImage, 0, PIPE_BUFFER_USAGE_CPU_WRITE); if (stImage->surface) - dstRowStride = stImage->surface->pitch * stImage->surface->cpp; + dstRowStride = stImage->surface->stride; } else { /* Allocate regular memory and store the image there temporarily. */ @@ -820,7 +820,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, */ texImage->Data = st_texture_image_map(ctx->st, stImage, 0, PIPE_BUFFER_USAGE_CPU_READ); - texImage->RowStride = stImage->surface->pitch; + texImage->RowStride = stImage->surface->stride / stImage->pt->block.size; } else { /* Otherwise, the image should actually be stored in @@ -925,7 +925,7 @@ st_TexSubimage(GLcontext * ctx, texImage->Data = st_texture_image_map(ctx->st, stImage, zoffset, PIPE_BUFFER_USAGE_CPU_WRITE); if (stImage->surface) - dstRowStride = stImage->surface->pitch * stImage->surface->cpp; + dstRowStride = stImage->surface->stride; } if (!texImage->Data) { @@ -1425,9 +1425,11 @@ copy_image_data_to_texture(struct st_context *st, stImage->face, dstLevel, stImage->base.Data, - stImage->base.RowStride, + stImage->base.RowStride * + stObj->pt->block.size, stImage->base.RowStride * - stImage->base.Height); + stImage->base.Height * + stObj->pt->block.size); _mesa_align_free(stImage->base.Data); stImage->base.Data = NULL; } @@ -1477,6 +1479,7 @@ st_finalize_texture(GLcontext *ctx, pipe_texture_reference(&stObj->pt, firstImage->pt); } + /* FIXME: determine format block instead of cpp */ if (firstImage->base.IsCompressed) { comp_byte = compressed_num_bytes(firstImage->base.TexFormat->MesaFormat); cpp = comp_byte; @@ -1497,7 +1500,9 @@ st_finalize_texture(GLcontext *ctx, stObj->pt->width[0] != firstImage->base.Width2 || stObj->pt->height[0] != firstImage->base.Height2 || stObj->pt->depth[0] != firstImage->base.Depth2 || - stObj->pt->cpp != cpp || + stObj->pt->block.size != cpp || + stObj->pt->block.width != 1 || + stObj->pt->block.height != 1 || stObj->pt->compressed != firstImage->base.IsCompressed) { pipe_texture_release(&stObj->pt); ctx->st->dirty.st |= ST_NEW_FRAMEBUFFER; diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 851f17c3b4..2fc00df429 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -137,10 +137,10 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, _mesa_generate_mipmap_level(target, datatype, comps, 0 /*border*/, pt->width[srcLevel], pt->height[srcLevel], pt->depth[srcLevel], - srcSurf->pitch * srcSurf->cpp, /* stride in bytes */ + srcSurf->stride, /* stride in bytes */ srcData, pt->width[dstLevel], pt->height[dstLevel], pt->depth[dstLevel], - dstSurf->pitch * dstSurf->cpp, /* stride in bytes */ + dstSurf->stride, /* stride in bytes */ dstData); pipe_buffer_unmap(pipe, srcSurf->buffer); diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 9553b34e31..8222826e7a 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -98,7 +98,7 @@ st_texture_create(struct st_context *st, pt.height[0] = height0; pt.depth[0] = depth0; pt.compressed = compress_byte ? 1 : 0; - pt.cpp = pt.compressed ? compress_byte : st_sizeof_format(format); + pf_get_block(format, &pt.block); pt.tex_usage = usage; newtex = screen->texture_create(screen, &pt); @@ -231,16 +231,19 @@ static void st_surface_data(struct pipe_context *pipe, struct pipe_surface *dst, unsigned dstx, unsigned dsty, - const void *src, unsigned src_pitch, + const void *src, unsigned src_stride, unsigned srcx, unsigned srcy, unsigned width, unsigned height) { struct pipe_screen *screen = pipe->screen; void *map = screen->surface_map(screen, dst, PIPE_BUFFER_USAGE_CPU_WRITE); pipe_copy_rect(map, - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); + &dst->block, + dst->stride, + dstx, dsty, + width, height, + src, src_stride, + srcx, srcy); screen->surface_unmap(screen, dst); } @@ -254,34 +257,29 @@ st_texture_image_data(struct pipe_context *pipe, GLuint face, GLuint level, void *src, - GLuint src_row_pitch, GLuint src_image_pitch) + GLuint src_row_stride, GLuint src_image_stride) { struct pipe_screen *screen = pipe->screen; GLuint depth = dst->depth[level]; GLuint i; - GLuint height = 0; const GLubyte *srcUB = src; struct pipe_surface *dst_surface; DBG("%s\n", __FUNCTION__); for (i = 0; i < depth; i++) { - height = dst->height[level]; - if(dst->compressed) - height /= 4; - dst_surface = screen->get_tex_surface(screen, dst, face, level, i, PIPE_BUFFER_USAGE_CPU_WRITE); st_surface_data(pipe, dst_surface, 0, 0, /* dstx, dsty */ srcUB, - src_row_pitch, + src_row_stride, 0, 0, /* source x, y */ - dst->width[level], height); /* width, height */ + dst->width[level], dst->height[level]); /* width, height */ screen->tex_surface_release(screen, &dst_surface); - srcUB += src_image_pitch * dst->cpp; + srcUB += src_image_stride; } } -- cgit v1.2.3 From 891469a3a5525a741094fd3f1e4a1270ec8b3c3b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 27 Jun 2008 20:10:04 +0900 Subject: cell: Update for cpp removal. Not tested -- just mymic the softpipe changes. --- src/gallium/drivers/cell/ppu/cell_surface.c | 65 ++++------------------------- src/gallium/drivers/cell/ppu/cell_texture.c | 16 ++++--- 2 files changed, 18 insertions(+), 63 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c index 18f3791924..5549eb496d 100644 --- a/src/gallium/drivers/cell/ppu/cell_surface.c +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -34,30 +34,6 @@ #include "cell_surface.h" -/* Upload data to a rectangular sub-region. Lots of choices how to do this: - * - * - memcpy by span to current destination - * - upload data as new buffer and blit - * - * Currently always memcpy. - */ -static void -cell_surface_data(struct pipe_context *pipe, - struct pipe_surface *dst, - unsigned dstx, unsigned dsty, - const void *src, unsigned src_pitch, - unsigned srcx, unsigned srcy, - unsigned width, unsigned height) -{ - pipe_copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, - dstx, dsty, width, height, src, src_pitch, srcx, srcy); - - pipe_surface_unmap(dst); -} - - static void cell_surface_copy(struct pipe_context *pipe, boolean do_flip, @@ -70,12 +46,12 @@ cell_surface_copy(struct pipe_context *pipe, assert( dst->cpp == src->cpp ); pipe_copy_rect(pipe_surface_map(dst), - dst->cpp, - dst->pitch, + &dst->block, + dst->stride, dstx, dsty, width, height, pipe_surface_map(src), - do_flip ? -src->pitch : src->pitch, + do_flip ? -src->stride : src->stride, srcx, do_flip ? height - 1 - srcy : srcy); pipe_surface_unmap(src); @@ -86,7 +62,7 @@ cell_surface_copy(struct pipe_context *pipe, static void * get_pointer(struct pipe_surface *dst, void *dst_map, unsigned x, unsigned y) { - return (char *)dst_map + (y * dst->pitch + x) * dst->cpp; + return (char *)dst_map + y / dst->block.height * dst->stride + x / dst->block.width * dst->block.size; } @@ -106,38 +82,13 @@ cell_surface_fill(struct pipe_context *pipe, unsigned i, j; void *dst_map = pipe_surface_map(dst); - assert(dst->pitch > 0); - assert(width <= dst->pitch); + assert(dst->stride > 0); - switch (dst->cpp) { + switch (dst->block.size) { case 1: - { - ubyte *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - memset(row, value, width); - row += dst->pitch; - } - } - break; case 2: - { - ushort *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = (ushort) value; - row += dst->pitch; - } - } - break; case 4: - { - unsigned *row = get_pointer(dst, dst_map, dstx, dsty); - for (i = 0; i < height; i++) { - for (j = 0; j < width; j++) - row[j] = value; - row += dst->pitch; - } - } + pipe_fill_rect(dst_map, &dst->block, dst->stride, dstx, dsty, width, height, value); break; case 8: { @@ -158,7 +109,7 @@ cell_surface_fill(struct pipe_context *pipe, row[j*4+2] = val2; row[j*4+3] = val3; } - row += dst->pitch * 4; + row += dst->stride/2; } } break; diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index 07717da8f6..533b64227d 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -65,12 +65,14 @@ cell_texture_layout(struct cell_texture * spt) pt->width[level] = width; pt->height[level] = height; pt->depth[level] = depth; + pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width); + pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height); spt->level_offset[level] = spt->buffer_size; - spt->buffer_size += ((pt->compressed) ? MAX2(1, height/4) : height) * + spt->buffer_size += (pt->nblocksy[level] * ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * - width * pt->cpp; + pt->nblocksx[level] * pt->block.size; width = minify(width); height = minify(height); @@ -157,16 +159,18 @@ cell_get_tex_surface_screen(struct pipe_screen *screen, assert(ps->winsys); pipe_buffer_reference(ws, &ps->buffer, spt->buffer); ps->format = pt->format; - ps->cpp = pt->cpp; + ps->block = pt->block; ps->width = pt->width[level]; ps->height = pt->height[level]; - ps->pitch = ps->width; + ps->nblocksx = pt->nblocksx[level]; + ps->nblocksy = pt->nblocksy[level]; + ps->stride = spt->stride[level]; ps->offset = spt->level_offset[level]; if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * - (pt->compressed ? ps->height/4 : ps->height) * - ps->width * ps->cpp; + ps->nblocksy * + ps->stride; } else { assert(face == 0); assert(zslice == 0); -- cgit v1.2.3 From 4f45dbc6aa9b55d96cf6aeb32117e117e333b4a8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 27 Jun 2008 20:56:29 +0900 Subject: gdi: Update for cpp removal. --- src/gallium/winsys/gdi/wmesa.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/gdi/wmesa.c b/src/gallium/winsys/gdi/wmesa.c index 0b93f8c4c3..86b085ab84 100644 --- a/src/gallium/winsys/gdi/wmesa.c +++ b/src/gallium/winsys/gdi/wmesa.c @@ -477,13 +477,15 @@ wm_surface_alloc_storage(struct pipe_winsys *winsys, surf->width = width; surf->height = height; surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = round_up(width, alignment / surf->cpp); + pf_get_block(format, &surf->block); + surf->nblocksx = pf_get_nblocksx(&surf->block, width); + surf->nblocksy = pf_get_nblocksy(&surf->block, height); + surf->stride = round_up(surf->nblocksx * surf->block.size, alignment); assert(!surf->buffer); surf->buffer = winsys->buffer_create(winsys, alignment, PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * height); + surf->nblocksy * surf->stride); if(!surf->buffer) return -1; -- cgit v1.2.3 From fb95b603efa3a70b73d28f0e3a7886fe736b9787 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 27 Jun 2008 20:56:49 +0900 Subject: mesa: Replace void * arithmetic. --- src/mesa/state_tracker/st_cb_accum.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 809a906ba3..2283905662 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -160,7 +160,7 @@ accum_mad(GLcontext *ctx, GLfloat scale, GLfloat bias, { struct pipe_screen *screen = ctx->st->pipe->screen; struct pipe_surface *acc_ps = acc_strb->surface; - GLvoid *map; + GLubyte *map; map = screen->surface_map(screen, acc_ps, PIPE_BUFFER_USAGE_CPU_WRITE); -- cgit v1.2.3 From 5b9d823545ec588ea97cc599a278626b99430d81 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 27 Jun 2008 14:16:42 +0200 Subject: i915: Fix the last of the stride/pitch changes --- src/gallium/drivers/i915simple/i915_surface.c | 18 +++++++++--------- src/gallium/winsys/dri/intel/intel_screen.c | 4 ++-- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 2 +- 3 files changed, 12 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index 0061b22f26..4430e81626 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -64,10 +64,10 @@ i915_surface_copy(struct pipe_context *pipe, pipe_copy_rect(dst_map, &dst->block, dst->stride, - dstx, dsty, - width, height, - src_map, - do_flip ? -(int) src->stride : src->stride, + dstx, dsty, + width, height, + src_map, + do_flip ? -(int) src->stride : src->stride, srcx, do_flip ? height - 1 - srcy : srcy); pipe->screen->surface_unmap(pipe->screen, src); @@ -79,8 +79,8 @@ i915_surface_copy(struct pipe_context *pipe, i915_copy_blit( i915_context(pipe), do_flip, dst->block.size, - (short) src->stride/src->block.size, src->buffer, src->offset, - (short) dst->stride/dst->block.size, dst->buffer, dst->offset, + (short) src->stride, src->buffer, src->offset, + (short) dst->stride, dst->buffer, dst->offset, (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); } } @@ -106,10 +106,10 @@ i915_surface_fill(struct pipe_context *pipe, assert(dst->block.height == 1); i915_fill_blit( i915_context(pipe), dst->block.size, - (short) dst->stride/dst->block.size, + (short) dst->stride, dst->buffer, dst->offset, - (short) dstx, (short) dsty, - (short) width, (short) height, + (short) dstx, (short) dsty, + (short) width, (short) height, value ); } } diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 18427a4586..89de188ada 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -78,10 +78,10 @@ intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, templat.last_level = 0; templat.depth[0] = 1; templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; - templat.cpp = intelScreen->front.cpp; templat.width[0] = intelScreen->front.width; templat.height[0] = intelScreen->front.height; - pitch = intelScreen->front.pitch / intelScreen->front.cpp; + pf_get_block(templat.format, &templat.block); + pitch = intelScreen->front.pitch; texture = screen->texture_blanket(screen, &templat, diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index 7f3babd98e..f58da97c64 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -89,7 +89,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, const drm_clip_rect_t *pbox = dPriv->pClipRects; const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; const int cpp = intelScreen->front.cpp; - const int srcpitch = surf->pitch; + const int srcpitch = surf->stride / cpp; int BR13, CMD; int i; -- cgit v1.2.3 From a1fb565ea7ac9f86beb4deece6a24d79e7b7860e Mon Sep 17 00:00:00 2001 From: Robert Ellison Date: Thu, 26 Jun 2008 08:55:00 -0600 Subject: egl: These changes allow an eglBindAPI(EGL_OPENGL_ES_API) to succeed, and to work correctly with GLES1 and GLES2. - egl_xdri.c just sets the EGL_OPENGL_ES_BIT as well as the EGL_OPENGL_BIT in ClientAPIsMask - eglconfig.c allows the renderable type to include EGL_OPENGL_ES2_BIT as well as EGL_OPENGL_ES_BIT. - egl_xlib.c sets the EGL_NATIVE_RENDERABLE attribute to EGL_FALSE for all softpipe configurations. (Otherwise, an eglChooseConfig() that looks for particular values of EGL_NATIVE_RENDERABLE will fail.) --- src/egl/drivers/xdri/egl_xdri.c | 2 +- src/egl/main/eglconfig.c | 2 +- src/gallium/winsys/egl_xlib/egl_xlib.c | 1 + 3 files changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c index df80c6a1c4..71d4f15371 100644 --- a/src/egl/drivers/xdri/egl_xdri.c +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -828,7 +828,7 @@ _eglMain(_EGLDisplay *disp, const char *args) xdri_drv->Base.API.DestroySurface = xdri_eglDestroySurface; xdri_drv->Base.API.SwapBuffers = xdri_eglSwapBuffers; - xdri_drv->Base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; + xdri_drv->Base.ClientAPIsMask = EGL_OPENGL_BIT | EGL_OPENGL_ES_BIT; xdri_drv->Base.Name = "X/DRI"; _eglLog(_EGL_DEBUG, "XDRI: main(%s)", args); diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c index 3ef0564a54..b6846d4928 100644 --- a/src/egl/main/eglconfig.c +++ b/src/egl/main/eglconfig.c @@ -147,7 +147,7 @@ _eglParseConfigAttribs(_EGLConfig *config, const EGLint *attrib_list) } else if (attr == EGL_RENDERABLE_TYPE) { EGLint renType = attrib_list[++i]; - if (renType & ~(EGL_OPENGL_ES_BIT | EGL_OPENVG_BIT)) { + if (renType & ~(EGL_OPENGL_ES_BIT | EGL_OPENGL_ES2_BIT | EGL_OPENVG_BIT)) { _eglError(EGL_BAD_ATTRIBUTE, "eglChooseConfig"); return EGL_FALSE; } diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index eeb15e30a9..8045c0d0ca 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -172,6 +172,7 @@ create_configs(_EGLDriver *drv, EGLDisplay dpy) SET_CONFIG_ATTRIB(config, EGL_STENCIL_SIZE, sbits); SET_CONFIG_ATTRIB(config, EGL_NATIVE_VISUAL_ID, visid); SET_CONFIG_ATTRIB(config, EGL_NATIVE_VISUAL_TYPE, vistype); + SET_CONFIG_ATTRIB(config, EGL_NATIVE_RENDERABLE, EGL_FALSE); _eglAddConfig(disp, config); } -- cgit v1.2.3 From 429a08384c2ea66d446e46beb28e33ee3b764d52 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 27 Jun 2008 16:02:43 +0200 Subject: gallium: handle msaa --- src/gallium/include/pipe/p_state.h | 6 +- src/gallium/winsys/dri/intel/intel_screen.c | 6 +- src/mesa/drivers/dri/common/utils.c | 97 ++++++++++++++++------------- src/mesa/drivers/dri/common/utils.h | 3 +- src/mesa/main/mtypes.h | 1 + src/mesa/state_tracker/st_cb_fbo.c | 4 +- src/mesa/state_tracker/st_cb_fbo.h | 2 +- src/mesa/state_tracker/st_framebuffer.c | 18 +++--- 8 files changed, 77 insertions(+), 60 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 2992e2f3b3..5546796936 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -276,7 +276,7 @@ struct pipe_surface unsigned layout; /**< PIPE_SURFACE_LAYOUT_x */ unsigned offset; /**< offset from start of buffer, in bytes */ unsigned refcount; - unsigned usage; /**< PIPE_BUFFER_USAGE_* */ + unsigned usage; /**< PIPE_BUFFER_USAGE_* */ struct pipe_winsys *winsys; /**< winsys which owns/created the surface */ @@ -311,7 +311,9 @@ struct pipe_texture unsigned last_level:8; /**< Index of last mipmap level present/defined */ unsigned compressed:1; - + + unsigned nr_samples:8; /**< for multisampled surfaces, nr of samples */ + unsigned tex_usage; /* PIPE_TEXTURE_USAGE_* */ /* These are also refcounted: diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index 89de188ada..cfecebdb8c 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -483,11 +483,13 @@ intelFillInModes(unsigned pixel_bits, unsigned depth_bits, uint8_t depth_bits_array[3]; uint8_t stencil_bits_array[3]; + uint8_t msaa_samples_array[1]; depth_bits_array[0] = 0; depth_bits_array[1] = depth_bits; depth_bits_array[2] = depth_bits; + msaa_samples_array[0] = 0; /* Just like with the accumulation buffer, always provide some modes * with a stencil buffer. It will be a sw fallback, but some apps won't @@ -521,7 +523,7 @@ intelFillInModes(unsigned pixel_bits, unsigned depth_bits, if (!driFillInModes(&m, fb_format, fb_type, depth_bits_array, stencil_bits_array, depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_TRUE_COLOR)) { + back_buffer_factor, msaa_samples_array, 1, GLX_TRUE_COLOR)) { fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, __LINE__); return NULL; @@ -529,7 +531,7 @@ intelFillInModes(unsigned pixel_bits, unsigned depth_bits, if (!driFillInModes(&m, fb_format, fb_type, depth_bits_array, stencil_bits_array, depth_buffer_factor, back_buffer_modes, - back_buffer_factor, GLX_DIRECT_COLOR)) { + back_buffer_factor, msaa_samples_array, 1, GLX_DIRECT_COLOR)) { fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, __LINE__); return NULL; diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c index 94db319928..3cf2146dce 100644 --- a/src/mesa/drivers/dri/common/utils.c +++ b/src/mesa/drivers/dri/common/utils.c @@ -521,6 +521,9 @@ GLboolean driClipRectToFramebuffer( const GLframebuffer *buffer, * \c GLX_SWAP_UNDEFINED_OML. See the * GLX_OML_swap_method extension spec for more details. * \param num_db_modes Number of entries in \c db_modes. + * \param msaa_samples Array of msaa sample count. 0 represents a visual + * without a multisample buffer. + * \param num_msaa_modes Number of entries in \c msaa_samples. * \param visType GLX visual type. Usually either \c GLX_TRUE_COLOR or * \c GLX_DIRECT_COLOR. * @@ -542,6 +545,7 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, const uint8_t * depth_bits, const uint8_t * stencil_bits, unsigned num_depth_stencil_bits, const GLenum * db_modes, unsigned num_db_modes, + const u_int8_t * msaa_samples, unsigned num_msaa_modes, int visType ) { static const uint8_t bits_table[3][4] = { @@ -607,9 +611,7 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, const uint32_t * masks; const int index = fb_type & 0x07; __GLcontextModes * modes = *ptr_to_modes; - unsigned i; - unsigned j; - unsigned k; + unsigned i, j, k, h; if ( bytes_per_pixel[ index ] == 0 ) { @@ -659,49 +661,54 @@ driFillInModes( __GLcontextModes ** ptr_to_modes, for ( k = 0 ; k < num_depth_stencil_bits ; k++ ) { for ( i = 0 ; i < num_db_modes ; i++ ) { - for ( j = 0 ; j < 2 ; j++ ) { - - modes->redBits = bits[0]; - modes->greenBits = bits[1]; - modes->blueBits = bits[2]; - modes->alphaBits = bits[3]; - modes->redMask = masks[0]; - modes->greenMask = masks[1]; - modes->blueMask = masks[2]; - modes->alphaMask = masks[3]; - modes->rgbBits = modes->redBits + modes->greenBits - + modes->blueBits + modes->alphaBits; - - modes->accumRedBits = 16 * j; - modes->accumGreenBits = 16 * j; - modes->accumBlueBits = 16 * j; - modes->accumAlphaBits = (masks[3] != 0) ? 16 * j : 0; - modes->visualRating = (j == 0) ? GLX_NONE : GLX_SLOW_CONFIG; - - modes->stencilBits = stencil_bits[k]; - modes->depthBits = depth_bits[k]; - - modes->visualType = visType; - modes->renderType = GLX_RGBA_BIT; - modes->drawableType = GLX_WINDOW_BIT; - modes->rgbMode = GL_TRUE; - - if ( db_modes[i] == GLX_NONE ) { - modes->doubleBufferMode = GL_FALSE; + for ( h = 0 ; h < num_msaa_modes; h++ ) { + for ( j = 0 ; j < 2 ; j++ ) { + + modes->redBits = bits[0]; + modes->greenBits = bits[1]; + modes->blueBits = bits[2]; + modes->alphaBits = bits[3]; + modes->redMask = masks[0]; + modes->greenMask = masks[1]; + modes->blueMask = masks[2]; + modes->alphaMask = masks[3]; + modes->rgbBits = modes->redBits + modes->greenBits + + modes->blueBits + modes->alphaBits; + + modes->accumRedBits = 16 * j; + modes->accumGreenBits = 16 * j; + modes->accumBlueBits = 16 * j; + modes->accumAlphaBits = (masks[3] != 0) ? 16 * j : 0; + modes->visualRating = (j == 0) ? GLX_NONE : GLX_SLOW_CONFIG; + + modes->stencilBits = stencil_bits[k]; + modes->depthBits = depth_bits[k]; + + modes->visualType = visType; + modes->renderType = GLX_RGBA_BIT; + modes->drawableType = GLX_WINDOW_BIT; + modes->rgbMode = GL_TRUE; + + if ( db_modes[i] == GLX_NONE ) { + modes->doubleBufferMode = GL_FALSE; + } + else { + modes->doubleBufferMode = GL_TRUE; + modes->swapMethod = db_modes[i]; + } + + modes->samples = msaa_samples[h]; + modes->sampleBuffers = modes->samples ? 1 : 0; + + modes->haveAccumBuffer = ((modes->accumRedBits + + modes->accumGreenBits + + modes->accumBlueBits + + modes->accumAlphaBits) > 0); + modes->haveDepthBuffer = (modes->depthBits > 0); + modes->haveStencilBuffer = (modes->stencilBits > 0); + + modes = modes->next; } - else { - modes->doubleBufferMode = GL_TRUE; - modes->swapMethod = db_modes[i]; - } - - modes->haveAccumBuffer = ((modes->accumRedBits + - modes->accumGreenBits + - modes->accumBlueBits + - modes->accumAlphaBits) > 0); - modes->haveDepthBuffer = (modes->depthBits > 0); - modes->haveStencilBuffer = (modes->stencilBits > 0); - - modes = modes->next; } } } diff --git a/src/mesa/drivers/dri/common/utils.h b/src/mesa/drivers/dri/common/utils.h index 1067d0a8bf..20940c21b4 100644 --- a/src/mesa/drivers/dri/common/utils.h +++ b/src/mesa/drivers/dri/common/utils.h @@ -115,6 +115,7 @@ extern GLboolean driFillInModes( __GLcontextModes ** modes, GLenum fb_format, GLenum fb_type, const uint8_t * depth_bits, const uint8_t * stencil_bits, unsigned num_depth_stencil_bits, - const GLenum * db_modes, unsigned num_db_modes, int visType ); + const GLenum * db_modes, unsigned num_db_modes, + const u_int8_t * msaa_samples, unsigned num_msaa_modes, int visType ); #endif /* DRI_DEBUG_H */ diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index dceb7611f2..0a065541e1 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2269,6 +2269,7 @@ struct gl_renderbuffer GLubyte IndexBits; GLubyte DepthBits; GLubyte StencilBits; + GLubyte Samples; /**< Number of samples - 0 if not multisampled */ GLvoid *Data; /**< This may not be used by some kinds of RBs */ /* Used to wrap one renderbuffer around another: */ diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 1067caf9b3..7245798d0d 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -118,6 +118,7 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, template.height[0] = height; template.depth[0] = 1; template.last_level = 0; + template.nr_samples = rb->Samples; if (pf_is_depth_stencil(template.format)) { template.tex_usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL; @@ -249,7 +250,7 @@ st_new_renderbuffer(GLcontext *ctx, GLuint name) * renderbuffer). The window system code determines the format. */ struct gl_renderbuffer * -st_new_renderbuffer_fb(enum pipe_format format) +st_new_renderbuffer_fb(enum pipe_format format, int samples) { struct st_renderbuffer *strb; @@ -261,6 +262,7 @@ st_new_renderbuffer_fb(enum pipe_format format) _mesa_init_renderbuffer(&strb->Base, 0); strb->Base.ClassID = 0x4242; /* just a unique value */ + strb->Base.Samples = samples; strb->format = format; switch (format) { diff --git a/src/mesa/state_tracker/st_cb_fbo.h b/src/mesa/state_tracker/st_cb_fbo.h index 87b0734a0c..ff56001a4e 100644 --- a/src/mesa/state_tracker/st_cb_fbo.h +++ b/src/mesa/state_tracker/st_cb_fbo.h @@ -58,7 +58,7 @@ st_renderbuffer(struct gl_renderbuffer *rb) extern struct gl_renderbuffer * -st_new_renderbuffer_fb(enum pipe_format format); +st_new_renderbuffer_fb(enum pipe_format format, int samples); extern void st_init_fbo_functions(struct dd_function_table *functions); diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 1b6e68c2a1..1994a1c826 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -51,27 +51,29 @@ st_create_framebuffer( const __GLcontextModes *visual, { struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer); if (stfb) { + int samples = 0; _mesa_initialize_framebuffer(&stfb->Base, visual); + if (visual->sampleBuffers) samples = visual->samples; { /* fake frontbuffer */ /* XXX allocation should only happen in the unusual case it's actually needed */ struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(colorFormat); + = st_new_renderbuffer_fb(colorFormat, samples); _mesa_add_renderbuffer(&stfb->Base, BUFFER_FRONT_LEFT, rb); } if (visual->doubleBufferMode) { struct gl_renderbuffer *rb - = st_new_renderbuffer_fb(colorFormat); + = st_new_renderbuffer_fb(colorFormat, samples); _mesa_add_renderbuffer(&stfb->Base, BUFFER_BACK_LEFT, rb); } if (visual->depthBits == 24 && visual->stencilBits == 8) { /* combined depth/stencil buffer */ struct gl_renderbuffer *depthStencilRb - = st_new_renderbuffer_fb(depthFormat); + = st_new_renderbuffer_fb(depthFormat, samples); /* note: bind RB to two attachment points */ _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthStencilRb); _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, depthStencilRb); @@ -82,26 +84,26 @@ st_create_framebuffer( const __GLcontextModes *visual, if (visual->depthBits == 32) { /* 32-bit depth buffer */ struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(depthFormat); + = st_new_renderbuffer_fb(depthFormat, samples); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } else if (visual->depthBits == 24) { /* 24-bit depth buffer, ignore stencil bits */ struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(depthFormat); + = st_new_renderbuffer_fb(depthFormat, samples); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } else if (visual->depthBits > 0) { /* 16-bit depth buffer */ struct gl_renderbuffer *depthRb - = st_new_renderbuffer_fb(depthFormat); + = st_new_renderbuffer_fb(depthFormat, samples); _mesa_add_renderbuffer(&stfb->Base, BUFFER_DEPTH, depthRb); } if (visual->stencilBits > 0) { /* 8-bit stencil */ struct gl_renderbuffer *stencilRb - = st_new_renderbuffer_fb(stencilFormat); + = st_new_renderbuffer_fb(stencilFormat, samples); _mesa_add_renderbuffer(&stfb->Base, BUFFER_STENCIL, stencilRb); } } @@ -109,7 +111,7 @@ st_create_framebuffer( const __GLcontextModes *visual, if (visual->accumRedBits > 0) { /* 16-bit/channel accum */ struct gl_renderbuffer *accumRb - = st_new_renderbuffer_fb(DEFAULT_ACCUM_PIPE_FORMAT); + = st_new_renderbuffer_fb(DEFAULT_ACCUM_PIPE_FORMAT, 0); /* XXX accum isn't multisampled right? */ _mesa_add_renderbuffer(&stfb->Base, BUFFER_ACCUM, accumRb); } -- cgit v1.2.3 From 838b0d6e480c9ae90abb067c1ba4aee413df6fd9 Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 27 Jun 2008 15:56:09 -0400 Subject: eh, we need a buildbot... fix the compilation --- src/gallium/winsys/egl_xlib/egl_xlib.c | 2 +- src/gallium/winsys/egl_xlib/sw_winsys.c | 8 +++----- 2 files changed, 4 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 8045c0d0ca..042d7a5457 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -291,7 +291,7 @@ display_surface(struct pipe_winsys *pws, ximage->data = data; ximage->width = psurf->width; ximage->height = psurf->height; - ximage->bytes_per_line = psurf->pitch * psurf->cpp; + ximage->bytes_per_line = psurf->stride; XPutImage(xsurf->Dpy, xsurf->Win, xsurf->Gc, ximage, 0, 0, 0, 0, psurf->width, psurf->height); diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c index 28cca9e581..6956fb5b6f 100644 --- a/src/gallium/winsys/egl_xlib/sw_winsys.c +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -190,16 +190,14 @@ surface_alloc_storage(struct pipe_winsys *winsys, surf->width = width; surf->height = height; surf->format = format; - surf->cpp = pf_get_size(format); - surf->pitch = round_up(width, alignment / surf->cpp); + pf_get_block(format, &surf->block); + surf->stride = round_up(surf->nblocksx * surf->block.size, alignment); surf->usage = flags; - assert(surf->cpp >= 1); - assert(surf->cpp <= 16); assert(!surf->buffer); surf->buffer = winsys->buffer_create(winsys, alignment, PIPE_BUFFER_USAGE_PIXEL, - surf->pitch * surf->cpp * height); + surf->stride * height); if(!surf->buffer) return -1; -- cgit v1.2.3 From a7499b7fc753653fd24a91b1ab6a5b68e479193b Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Fri, 27 Jun 2008 16:47:22 -0400 Subject: egl: helps if the stride is right --- src/gallium/winsys/egl_xlib/sw_winsys.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c index 6956fb5b6f..f4199e6f89 100644 --- a/src/gallium/winsys/egl_xlib/sw_winsys.c +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -190,7 +190,9 @@ surface_alloc_storage(struct pipe_winsys *winsys, surf->width = width; surf->height = height; surf->format = format; - pf_get_block(format, &surf->block); + pf_get_block(surf->format, &surf->block); + surf->nblocksx = pf_get_nblocksx(&surf->block, width); + surf->nblocksy = pf_get_nblocksy(&surf->block, height); surf->stride = round_up(surf->nblocksx * surf->block.size, alignment); surf->usage = flags; -- cgit v1.2.3 From 8f91a83669b8626678eb38fbbdd304c72243b6fe Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 28 Jun 2008 16:03:28 -0600 Subject: egl: fix default value of EGL_SURFACE_TYPE, added some sanity check assertions --- src/egl/main/eglconfig.c | 12 +++++++++--- 1 file changed, 9 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglconfig.c b/src/egl/main/eglconfig.c index b6846d4928..b19988f49a 100644 --- a/src/egl/main/eglconfig.c +++ b/src/egl/main/eglconfig.c @@ -43,9 +43,7 @@ _eglInitConfig(_EGLConfig *config, EGLint id) _eglSetConfigAttrib(config, EGL_NATIVE_VISUAL_TYPE, EGL_DONT_CARE); _eglSetConfigAttrib(config, EGL_MIN_SWAP_INTERVAL, EGL_DONT_CARE); _eglSetConfigAttrib(config, EGL_MAX_SWAP_INTERVAL, EGL_DONT_CARE); - _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, - EGL_SCREEN_BIT_MESA | EGL_PBUFFER_BIT | - EGL_PIXMAP_BIT | EGL_WINDOW_BIT); + _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_WINDOW_BIT); _eglSetConfigAttrib(config, EGL_TRANSPARENT_TYPE, EGL_NONE); _eglSetConfigAttrib(config, EGL_TRANSPARENT_RED_VALUE, EGL_DONT_CARE); _eglSetConfigAttrib(config, EGL_TRANSPARENT_GREEN_VALUE, EGL_DONT_CARE); @@ -96,6 +94,14 @@ _eglAddConfig(_EGLDisplay *display, _EGLConfig *config) _EGLConfig **newConfigs; EGLint n; + /* do some sanity checks on the config's attribs */ + assert(GET_CONFIG_ATTRIB(config, EGL_CONFIG_ID) > 0); + assert(GET_CONFIG_ATTRIB(config, EGL_RENDERABLE_TYPE) != 0x0); + assert(GET_CONFIG_ATTRIB(config, EGL_SURFACE_TYPE) != 0x0); + assert(GET_CONFIG_ATTRIB(config, EGL_RED_SIZE) > 0); + assert(GET_CONFIG_ATTRIB(config, EGL_GREEN_SIZE) > 0); + assert(GET_CONFIG_ATTRIB(config, EGL_BLUE_SIZE) > 0); + n = display->NumConfigs; /* realloc array of ptrs */ -- cgit v1.2.3 From d4b100a6a1f13ceaedf603bdce5ce95c2ce7e12c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 28 Jun 2008 16:04:01 -0600 Subject: egl: set config's EGL_CONFORMANT, EGL_RENDERABLE_TYPE, EGL_SURFACE_TYPE attributes --- src/gallium/winsys/egl_xlib/egl_xlib.c | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 042d7a5457..83b8bb95b1 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -133,6 +133,10 @@ bitcount(unsigned int n) static void create_configs(_EGLDriver *drv, EGLDisplay dpy) { + static const EGLint all_apis = (EGL_OPENGL_ES_BIT | + EGL_OPENGL_ES2_BIT | + EGL_OPENVG_BIT | + EGL_OPENGL_BIT); _EGLDisplay *disp = _eglLookupDisplay(dpy); XVisualInfo *visInfo, visTemplate; int num_visuals, i; @@ -173,6 +177,9 @@ create_configs(_EGLDriver *drv, EGLDisplay dpy) SET_CONFIG_ATTRIB(config, EGL_NATIVE_VISUAL_ID, visid); SET_CONFIG_ATTRIB(config, EGL_NATIVE_VISUAL_TYPE, vistype); SET_CONFIG_ATTRIB(config, EGL_NATIVE_RENDERABLE, EGL_FALSE); + SET_CONFIG_ATTRIB(config, EGL_CONFORMANT, all_apis); + SET_CONFIG_ATTRIB(config, EGL_RENDERABLE_TYPE, all_apis); + SET_CONFIG_ATTRIB(config, EGL_SURFACE_TYPE, EGL_WINDOW_BIT); _eglAddConfig(disp, config); } -- cgit v1.2.3 From a1ec6efce0f614dfc2fb7af2cab68eca3be43850 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 28 Jun 2008 16:15:03 -0600 Subject: mesa: check FEATURE_point_size_array --- src/mesa/main/ffvertex_prog.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c index 2baae77570..5f3def257d 100644 --- a/src/mesa/main/ffvertex_prog.c +++ b/src/mesa/main/ffvertex_prog.c @@ -265,8 +265,10 @@ static struct state_key *make_state_key( GLcontext *ctx ) if (ctx->Point._Attenuated) key->point_attenuated = 1; +#if FEATURE_point_size_array if (ctx->Array.ArrayObj->PointSize.Enabled) key->point_array = 1; +#endif if (ctx->Texture._TexGenEnabled || ctx->Texture._TexMatEnabled || -- cgit v1.2.3 From 2242769a13de228df10fb55d50aa4d96c99d5975 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 28 Jun 2008 16:47:22 -0600 Subject: s/GL_INVALID_VALUE/GL_INVALID_OPERATION/ in _mesa_get_uniformfv() --- src/mesa/shader/shader_api.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 97edb25400..a005c902e1 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -892,8 +892,10 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, program); if (shProg) { - if (location < shProg->Uniforms->NumUniforms) { - GLuint progPos, i; + if (shProg->Uniforms && + location >= 0 && location < shProg->Uniforms->NumUniforms) { + GLint progPos; + GLuint i; const struct gl_program *prog = NULL; progPos = shProg->Uniforms->Uniforms[location].VertPos; @@ -915,11 +917,11 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, } } else { - _mesa_error(ctx, GL_INVALID_VALUE, "glGetUniformfv(location)"); + _mesa_error(ctx, GL_INVALID_OPERATION, "glGetUniformfv(location)"); } } else { - _mesa_error(ctx, GL_INVALID_VALUE, "glGetUniformfv(program)"); + _mesa_error(ctx, GL_INVALID_OPERATION, "glGetUniformfv(program)"); } } -- cgit v1.2.3 From 9d94d133b019ef23ee03cc691fcb5602451604ae Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sat, 28 Jun 2008 16:47:39 -0600 Subject: mesa: added null ptr checks --- src/mesa/shader/prog_uniform.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_uniform.c b/src/mesa/shader/prog_uniform.c index d96a916533..f57df3d86d 100644 --- a/src/mesa/shader/prog_uniform.c +++ b/src/mesa/shader/prog_uniform.c @@ -119,7 +119,7 @@ GLint _mesa_lookup_uniform(const struct gl_uniform_list *list, const char *name) { GLuint i; - for (i = 0; i < list->NumUniforms; i++) { + for (i = 0; list && i < list->NumUniforms; i++) { if (!_mesa_strcmp(list->Uniforms[i].Name, name)) { return i; } @@ -133,7 +133,7 @@ _mesa_longest_uniform_name(const struct gl_uniform_list *list) { GLint max = 0; GLuint i; - for (i = 0; i < list->NumUniforms; i++) { + for (i = 0; list && i < list->NumUniforms; i++) { GLuint len = _mesa_strlen(list->Uniforms[i].Name); if (len > (GLuint)max) max = len; -- cgit v1.2.3 From b3da2a9524b47838da33d8f7fd63c5d90a659cfa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 1 Jul 2008 22:04:58 +0900 Subject: gallium: Use the inline keyword on C++. --- src/gallium/include/pipe/p_compiler.h | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index d2b8c41be3..2afb8464c7 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -107,7 +107,9 @@ typedef unsigned char boolean; /* Function inlining */ -#if defined(__GNUC__) +#ifdef __cplusplus +# define INLINE inline +#elif defined(__GNUC__) # define INLINE __inline__ #elif defined(_MSC_VER) # define INLINE __inline -- cgit v1.2.3 From e99ce4af8a6a6f9b257bfceb0f9133a29f06f2c7 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 1 Jul 2008 15:52:37 +0200 Subject: i915: Last reference to surface -> pitch --- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index f58da97c64..f751f97524 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -97,7 +97,7 @@ intelDisplaySurface(__DRIdrawablePrivate *dPriv, ASSERT(surf->cpp == cpp); DBG(SWAP, "screen pitch %d src surface pitch %d\n", - pitch, surf->pitch); + pitch, surf->stride); if (cpp == 2) { BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); -- cgit v1.2.3 From 7d4f01413f16c15b55e99aba6da18b5c979c880c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 1 Jul 2008 11:48:27 -0600 Subject: mesa: add/fix some IrInfo entries for debugging purposes cherry-picked from master --- src/mesa/shader/slang/slang_ir.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_ir.c b/src/mesa/shader/slang/slang_ir.c index 92e8d0345e..31ad94183c 100644 --- a/src/mesa/shader/slang/slang_ir.c +++ b/src/mesa/shader/slang/slang_ir.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.3 + * Version: 7.1 * - * Copyright (C) 2005-2007 Brian Paul All Rights Reserved. + * Copyright (C) 2005-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -50,6 +50,7 @@ static const slang_ir_info IrInfo[] = { { IR_SLE, "IR_SLE", OPCODE_SLE, 4, 2 }, { IR_SLT, "IR_SLT", OPCODE_SLT, 4, 2 }, { IR_POW, "IR_POW", OPCODE_POW, 1, 2 }, + { IR_EQUAL, "IR_EQUAL", OPCODE_NOP, 1, 2 }, /* unary ops */ { IR_I_TO_F, "IR_I_TO_F", OPCODE_NOP, 1, 1 }, { IR_F_TO_I, "IR_F_TO_I", OPCODE_INT, 4, 1 }, /* 4 floats to 4 ints */ @@ -90,7 +91,8 @@ static const slang_ir_info IrInfo[] = { { IR_FIELD, "IR_FIELD", OPCODE_NOP, 0, 0 }, { IR_ELEMENT, "IR_ELEMENT", OPCODE_NOP, 0, 0 }, { IR_SWIZZLE, "IR_SWIZZLE", OPCODE_NOP, 0, 0 }, - { IR_NOP, NULL, OPCODE_NOP, 0, 0 } + { IR_NOP, "IR_NOP", OPCODE_NOP, 0, 0 }, + { 0, NULL, 0, 0, 0 } }; -- cgit v1.2.3 From 44c99ad23655204fd2e567fc38512f12e5f262af Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 1 Jul 2008 11:48:57 -0600 Subject: mesa: better function inlining in the presence of 'return' statements Before, the presence of a 'return' statement always prevented inlining a function. This was because we didn't want to accidentally return from the _calling_ function. We still need the semantic of 'return' when inlining but we can't always use unconditional branches/jumps (GPUs don't always support arbitrary branching). Now, we allow inlining functions w/ return if the return is the last statement in the function. This fixes the common case of a function that returns a value, such as: vec4 square(const in vec4 x) { return x * x; } which effectively compiles into: vec4 square(const in vec4 x) { __retVal = x * x; return; } The 'return' can be no-op'd now and we can inline the function. cherry-picked from master --- src/mesa/shader/slang/slang_codegen.c | 128 ++++++++++++++++++++++++++++------ 1 file changed, 105 insertions(+), 23 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index d19d5a0abb..821ed11901 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -707,6 +707,64 @@ _slang_find_node_type(slang_operation *oper, slang_operation_type type) } +/** + * Count the number of operations of the given time rooted at 'oper'. + */ +static GLuint +_slang_count_node_type(slang_operation *oper, slang_operation_type type) +{ + GLuint i, count = 0; + if (oper->type == type) { + return 1; + } + for (i = 0; i < oper->num_children; i++) { + count += _slang_count_node_type(&oper->children[i], type); + } + return count; +} + + +/** + * Check if the 'return' statement found under 'oper' is a "tail return" + * that can be no-op'd. For example: + * + * void func(void) + * { + * .. do something .. + * return; // this is a no-op + * } + * + * This is used when determining if a function can be inlined. If the + * 'return' is not the last statement, we can't inline the function since + * we still need the semantic behaviour of the 'return' but we don't want + * to accidentally return from the _calling_ function. We'd need to use an + * unconditional branch, but we don't have such a GPU instruction (not + * always, at least). + */ +static GLboolean +_slang_is_tail_return(const slang_operation *oper) +{ + GLuint k = oper->num_children; + + while (k > 0) { + const slang_operation *last = &oper->children[k - 1]; + if (last->type == SLANG_OPER_RETURN) + return GL_TRUE; + else if (last->type == SLANG_OPER_IDENTIFIER || + last->type == SLANG_OPER_LABEL) + k--; /* try prev child */ + else if (last->type == SLANG_OPER_BLOCK_NO_NEW_SCOPE || + last->type == SLANG_OPER_BLOCK_NEW_SCOPE) + /* try sub-children */ + return _slang_is_tail_return(last); + else + break; + } + + return GL_FALSE; +} + + static void slang_resolve_variable(slang_operation *oper) { @@ -1207,38 +1265,62 @@ _slang_gen_function_call(slang_assemble_ctx *A, slang_function *fun, } else { /* non-assembly function */ + /* We always generate an "inline-able" block of code here. + * We may either: + * 1. insert the inline code + * 2. Generate a call to the "inline" code as a subroutine + */ + + + slang_operation *ret = NULL; + inlined = slang_inline_function_call(A, fun, oper, dest); - if (inlined && _slang_find_node_type(inlined, SLANG_OPER_RETURN)) { - slang_operation *callOper; - /* The function we're calling has one or more 'return' statements. - * So, we can't truly inline this function because we need to - * implement 'return' with RET (and CAL). - * Nevertheless, we performed "inlining" to make a new instance - * of the function body to deal with static register allocation. - * - * XXX check if there's one 'return' and if it's the very last - * statement in the function - we can optimize that case. - */ - assert(inlined->type == SLANG_OPER_BLOCK_NEW_SCOPE || - inlined->type == SLANG_OPER_SEQUENCE); - if (_slang_function_has_return_value(fun) && !dest) { - assert(inlined->children[0].type == SLANG_OPER_VARIABLE_DECL); - assert(inlined->children[2].type == SLANG_OPER_IDENTIFIER); - callOper = &inlined->children[1]; + if (!inlined) + return NULL; + + ret = _slang_find_node_type(inlined, SLANG_OPER_RETURN); + if (ret) { + /* check if this is a "tail" return */ + if (_slang_count_node_type(inlined, SLANG_OPER_RETURN) == 1 && + _slang_is_tail_return(inlined)) { + /* The only RETURN is the last stmt in the function, no-op it + * and inline the function body. + */ + ret->type = SLANG_OPER_NONE; } else { - callOper = inlined; + slang_operation *callOper; + /* The function we're calling has one or more 'return' statements. + * So, we can't truly inline this function because we need to + * implement 'return' with RET (and CAL). + * Nevertheless, we performed "inlining" to make a new instance + * of the function body to deal with static register allocation. + * + * XXX check if there's one 'return' and if it's the very last + * statement in the function - we can optimize that case. + */ + assert(inlined->type == SLANG_OPER_BLOCK_NEW_SCOPE || + inlined->type == SLANG_OPER_SEQUENCE); + + if (_slang_function_has_return_value(fun) && !dest) { + assert(inlined->children[0].type == SLANG_OPER_VARIABLE_DECL); + assert(inlined->children[2].type == SLANG_OPER_IDENTIFIER); + callOper = &inlined->children[1]; + } + else { + callOper = inlined; + } + callOper->type = SLANG_OPER_NON_INLINED_CALL; + callOper->fun = fun; + callOper->label = _slang_label_new_unique((char*) fun->header.a_name); } - callOper->type = SLANG_OPER_NON_INLINED_CALL; - callOper->fun = fun; - callOper->label = _slang_label_new_unique((char*) fun->header.a_name); } } if (!inlined) return NULL; - /* Replace the function call with the inlined block */ + /* Replace the function call with the inlined block (or new CALL stmt) */ slang_operation_destruct(oper); *oper = *inlined; _slang_free(inlined); @@ -3024,7 +3106,7 @@ _slang_codegen_function(slang_assemble_ctx * A, slang_function * fun) if (_mesa_strcmp((char *) fun->header.a_name, "main") != 0) { /* we only really generate code for main, all other functions get - * inlined. + * inlined or codegen'd upon an actual call. */ #if 0 /* do some basic error checking though */ -- cgit v1.2.3 From 81a0acca31aa8f96ac9488ff81345a6cd51403a4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 1 Jul 2008 18:39:22 -0600 Subject: mesa: make _slang_swizzle_swizzle() non-private cherry-picked from master --- src/mesa/shader/slang/slang_emit.c | 17 ++++++++++------- src/mesa/shader/slang/slang_emit.h | 8 ++++++-- 2 files changed, 16 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 93256f8647..6d79380ab5 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 7.0.3 + * Version: 7.1 * - * Copyright (C) 2005-2007 Brian Paul All Rights Reserved. + * Copyright (C) 2005-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -110,13 +110,16 @@ writemask_to_swizzle(GLuint writemask) /** * Swizzle a swizzle. That is, return swz2(swz1) */ -static GLuint -swizzle_swizzle(GLuint swz1, GLuint swz2) +GLuint +_slang_swizzle_swizzle(GLuint swz1, GLuint swz2) { GLuint i, swz, s[4]; for (i = 0; i < 4; i++) { GLuint c = GET_SWZ(swz2, i); - s[i] = GET_SWZ(swz1, c); + if (c <= SWIZZLE_W) + s[i] = GET_SWZ(swz1, c); + else + s[i] = c; } swz = MAKE_SWIZZLE4(s[0], s[1], s[2], s[3]); return swz; @@ -1459,8 +1462,8 @@ emit_swizzle(slang_emit_info *emitInfo, slang_ir_node *n) /* apply this swizzle to child's swizzle to get composed swizzle */ swizzle = fix_swizzle(n->Store->Swizzle); /* remove the don't care terms */ - n->Store->Swizzle = swizzle_swizzle(n->Children[0]->Store->Swizzle, - swizzle); + n->Store->Swizzle = _slang_swizzle_swizzle(n->Children[0]->Store->Swizzle, + swizzle); return inst; } diff --git a/src/mesa/shader/slang/slang_emit.h b/src/mesa/shader/slang/slang_emit.h index 13ba6d7d6c..7d70b3311b 100644 --- a/src/mesa/shader/slang/slang_emit.h +++ b/src/mesa/shader/slang/slang_emit.h @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.3 + * Version: 7.1 * - * Copyright (C) 2005-2007 Brian Paul All Rights Reserved. + * Copyright (C) 2005-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -36,6 +36,10 @@ extern void slang_print_ir(const slang_ir_node *n, int indent); +extern GLuint +_slang_swizzle_swizzle(GLuint swz1, GLuint swz2); + + extern slang_ir_storage * _slang_new_ir_storage(enum register_file file, GLint index, GLint size); -- cgit v1.2.3 From eeefe175de8bae93b9ed2204afa9e748b205e868 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 1 Jul 2008 18:39:46 -0600 Subject: mesa: move some functions cherry-picked from master --- src/mesa/shader/slang/slang_codegen.c | 198 +++++++++++++++++----------------- 1 file changed, 99 insertions(+), 99 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 821ed11901..efeae283fc 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1394,6 +1394,105 @@ make_writemask(const char *field) } +/** + * Some write-masked assignments are simple, but others are hard. + * Simple example: + * vec3 v; + * v.xy = vec2(a, b); + * Hard example: + * vec3 v; + * v.zy = vec2(a, b); + * this gets transformed/swizzled into: + * v.zy = vec2(a, b).*yx* (* = don't care) + * This function helps to determine simple vs. non-simple. + */ +static GLboolean +_slang_simple_writemask(GLuint writemask, GLuint swizzle) +{ + switch (writemask) { + case WRITEMASK_X: + return GET_SWZ(swizzle, 0) == SWIZZLE_X; + case WRITEMASK_Y: + return GET_SWZ(swizzle, 1) == SWIZZLE_Y; + case WRITEMASK_Z: + return GET_SWZ(swizzle, 2) == SWIZZLE_Z; + case WRITEMASK_W: + return GET_SWZ(swizzle, 3) == SWIZZLE_W; + case WRITEMASK_XY: + return (GET_SWZ(swizzle, 0) == SWIZZLE_X) + && (GET_SWZ(swizzle, 1) == SWIZZLE_Y); + case WRITEMASK_XYZ: + return (GET_SWZ(swizzle, 0) == SWIZZLE_X) + && (GET_SWZ(swizzle, 1) == SWIZZLE_Y) + && (GET_SWZ(swizzle, 2) == SWIZZLE_Z); + case WRITEMASK_XYZW: + return swizzle == SWIZZLE_NOOP; + default: + return GL_FALSE; + } +} + + +/** + * Convert the given swizzle into a writemask. In some cases this + * is trivial, in other cases, we'll need to also swizzle the right + * hand side to put components in the right places. + * \param swizzle the incoming swizzle + * \param writemaskOut returns the writemask + * \param swizzleOut swizzle to apply to the right-hand-side + * \return GL_FALSE for simple writemasks, GL_TRUE for non-simple + */ +static GLboolean +swizzle_to_writemask(GLuint swizzle, + GLuint *writemaskOut, GLuint *swizzleOut) +{ + GLuint mask = 0x0, newSwizzle[4]; + GLint i, size; + + /* make new dst writemask, compute size */ + for (i = 0; i < 4; i++) { + const GLuint swz = GET_SWZ(swizzle, i); + if (swz == SWIZZLE_NIL) { + /* end */ + break; + } + assert(swz >= 0 && swz <= 3); + mask |= (1 << swz); + } + assert(mask <= 0xf); + size = i; /* number of components in mask/swizzle */ + + *writemaskOut = mask; + + /* make new src swizzle, by inversion */ + for (i = 0; i < 4; i++) { + newSwizzle[i] = i; /*identity*/ + } + for (i = 0; i < size; i++) { + const GLuint swz = GET_SWZ(swizzle, i); + newSwizzle[swz] = i; + } + *swizzleOut = MAKE_SWIZZLE4(newSwizzle[0], + newSwizzle[1], + newSwizzle[2], + newSwizzle[3]); + + if (_slang_simple_writemask(mask, *swizzleOut)) { + if (size >= 1) + assert(GET_SWZ(*swizzleOut, 0) == SWIZZLE_X); + if (size >= 2) + assert(GET_SWZ(*swizzleOut, 1) == SWIZZLE_Y); + if (size >= 3) + assert(GET_SWZ(*swizzleOut, 2) == SWIZZLE_Z); + if (size >= 4) + assert(GET_SWZ(*swizzleOut, 3) == SWIZZLE_W); + return GL_TRUE; + } + else + return GL_FALSE; +} + + /** * Generate IR tree for an asm instruction/operation such as: * __asm vec4_dot __retVal.x, v1, v2; @@ -2208,105 +2307,6 @@ _slang_gen_variable(slang_assemble_ctx * A, slang_operation *oper) } -/** - * Some write-masked assignments are simple, but others are hard. - * Simple example: - * vec3 v; - * v.xy = vec2(a, b); - * Hard example: - * vec3 v; - * v.zy = vec2(a, b); - * this gets transformed/swizzled into: - * v.zy = vec2(a, b).*yx* (* = don't care) - * This function helps to determine simple vs. non-simple. - */ -static GLboolean -_slang_simple_writemask(GLuint writemask, GLuint swizzle) -{ - switch (writemask) { - case WRITEMASK_X: - return GET_SWZ(swizzle, 0) == SWIZZLE_X; - case WRITEMASK_Y: - return GET_SWZ(swizzle, 1) == SWIZZLE_Y; - case WRITEMASK_Z: - return GET_SWZ(swizzle, 2) == SWIZZLE_Z; - case WRITEMASK_W: - return GET_SWZ(swizzle, 3) == SWIZZLE_W; - case WRITEMASK_XY: - return (GET_SWZ(swizzle, 0) == SWIZZLE_X) - && (GET_SWZ(swizzle, 1) == SWIZZLE_Y); - case WRITEMASK_XYZ: - return (GET_SWZ(swizzle, 0) == SWIZZLE_X) - && (GET_SWZ(swizzle, 1) == SWIZZLE_Y) - && (GET_SWZ(swizzle, 2) == SWIZZLE_Z); - case WRITEMASK_XYZW: - return swizzle == SWIZZLE_NOOP; - default: - return GL_FALSE; - } -} - - -/** - * Convert the given swizzle into a writemask. In some cases this - * is trivial, in other cases, we'll need to also swizzle the right - * hand side to put components in the right places. - * \param swizzle the incoming swizzle - * \param writemaskOut returns the writemask - * \param swizzleOut swizzle to apply to the right-hand-side - * \return GL_FALSE for simple writemasks, GL_TRUE for non-simple - */ -static GLboolean -swizzle_to_writemask(GLuint swizzle, - GLuint *writemaskOut, GLuint *swizzleOut) -{ - GLuint mask = 0x0, newSwizzle[4]; - GLint i, size; - - /* make new dst writemask, compute size */ - for (i = 0; i < 4; i++) { - const GLuint swz = GET_SWZ(swizzle, i); - if (swz == SWIZZLE_NIL) { - /* end */ - break; - } - assert(swz >= 0 && swz <= 3); - mask |= (1 << swz); - } - assert(mask <= 0xf); - size = i; /* number of components in mask/swizzle */ - - *writemaskOut = mask; - - /* make new src swizzle, by inversion */ - for (i = 0; i < 4; i++) { - newSwizzle[i] = i; /*identity*/ - } - for (i = 0; i < size; i++) { - const GLuint swz = GET_SWZ(swizzle, i); - newSwizzle[swz] = i; - } - *swizzleOut = MAKE_SWIZZLE4(newSwizzle[0], - newSwizzle[1], - newSwizzle[2], - newSwizzle[3]); - - if (_slang_simple_writemask(mask, *swizzleOut)) { - if (size >= 1) - assert(GET_SWZ(*swizzleOut, 0) == SWIZZLE_X); - if (size >= 2) - assert(GET_SWZ(*swizzleOut, 1) == SWIZZLE_Y); - if (size >= 3) - assert(GET_SWZ(*swizzleOut, 2) == SWIZZLE_Z); - if (size >= 4) - assert(GET_SWZ(*swizzleOut, 3) == SWIZZLE_W); - return GL_TRUE; - } - else - return GL_FALSE; -} - - static slang_ir_node * _slang_gen_swizzle(slang_ir_node *child, GLuint swizzle) { -- cgit v1.2.3 From 66b48202c221a25f3980df8f443ce63c2fb4119f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 1 Jul 2008 18:40:23 -0600 Subject: mesa: fix a GLSL vector subscript/writemask bug This fixes a failure for cases like: vec4 v; v[1] *= 2.0; The v[1] actually acts like a writemask, equivalent to v.y The fix is a bit convoluted, but will do for now. cherry-picked from master --- src/mesa/shader/slang/slang_codegen.c | 137 ++++++++++++++++++++++------------ 1 file changed, 90 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index efeae283fc..4846c257f1 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1355,45 +1355,6 @@ slang_find_asm_info(const char *name) } -static GLuint -make_writemask(const char *field) -{ - GLuint mask = 0x0; - while (*field) { - switch (*field) { - case 'x': - case 's': - case 'r': - mask |= WRITEMASK_X; - break; - case 'y': - case 't': - case 'g': - mask |= WRITEMASK_Y; - break; - case 'z': - case 'p': - case 'b': - mask |= WRITEMASK_Z; - break; - case 'w': - case 'q': - case 'a': - mask |= WRITEMASK_W; - break; - default: - _mesa_problem(NULL, "invalid writemask in make_writemask()"); - return 0; - } - field++; - } - if (mask == 0x0) - return WRITEMASK_XYZW; - else - return mask; -} - - /** * Some write-masked assignments are simple, but others are hard. * Simple example: @@ -1493,6 +1454,88 @@ swizzle_to_writemask(GLuint swizzle, } +/** + * Recursively traverse 'oper' to produce a swizzle mask in the event + * of any vector subscripts and swizzle suffixes. + * Ex: for "vec4 v", "v[2].x" resolves to v.z + */ +static GLuint +resolve_swizzle(const slang_operation *oper) +{ + if (oper->type == SLANG_OPER_FIELD) { + /* writemask from .xyzw suffix */ + slang_swizzle swz; + if (_slang_is_swizzle((char*) oper->a_id, 4, &swz)) { + GLuint swizzle = MAKE_SWIZZLE4(swz.swizzle[0], + swz.swizzle[1], + swz.swizzle[2], + swz.swizzle[3]); + GLuint child_swizzle = resolve_swizzle(&oper->children[0]); + GLuint s = _slang_swizzle_swizzle(child_swizzle, swizzle); + return s; + } + else + return SWIZZLE_XYZW; + } + else if (oper->type == SLANG_OPER_SUBSCRIPT && + oper->children[1].type == SLANG_OPER_LITERAL_INT) { + /* writemask from [index] */ + GLuint child_swizzle = resolve_swizzle(&oper->children[0]); + GLuint i = (GLuint) oper->children[1].literal[0]; + GLuint swizzle; + GLuint s; + switch (i) { + case 0: + swizzle = SWIZZLE_XXXX; + break; + case 1: + swizzle = SWIZZLE_YYYY; + break; + case 2: + swizzle = SWIZZLE_ZZZZ; + break; + case 3: + swizzle = SWIZZLE_WWWW; + break; + default: + swizzle = SWIZZLE_XYZW; + } + s = _slang_swizzle_swizzle(child_swizzle, swizzle); + return s; + } + else { + return SWIZZLE_XYZW; + } +} + + +/** + * As above, but produce a writemask. + */ +static GLuint +resolve_writemask(const slang_operation *oper) +{ + GLuint swizzle = resolve_swizzle(oper); + GLuint writemask, swizzleOut; + swizzle_to_writemask(swizzle, &writemask, &swizzleOut); + return writemask; +} + + +/** + * Recursively descend through swizzle nodes to find the node's storage info. + */ +static slang_ir_storage * +get_store(const slang_ir_node *n) +{ + if (n->Opcode == IR_SWIZZLE) { + return get_store(n->Children[0]); + } + return n->Store; +} + + + /** * Generate IR tree for an asm instruction/operation such as: * __asm vec4_dot __retVal.x, v1, v2; @@ -1547,19 +1590,19 @@ _slang_gen_asm(slang_assemble_ctx *A, slang_operation *oper, slang_ir_node *n0; dest_oper = &oper->children[0]; - while (dest_oper->type == SLANG_OPER_FIELD) { - /* writemask */ - writemask &= make_writemask((char*) dest_oper->a_id); - dest_oper = &dest_oper->children[0]; - } + + writemask = resolve_writemask(dest_oper); n0 = _slang_gen_operation(A, dest_oper); - assert(n0->Var); - assert(n0->Store); + if (!n0) + return NULL; + assert(!n->Store); - n->Store = n0->Store; + n->Store = get_store(n0); n->Writemask = writemask; + assert(n->Store->File != PROGRAM_UNDEFINED); + _slang_free(n0); } -- cgit v1.2.3 From ea4ca10b1bec67c8a60db0e4e5581318ce9f62f9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 2 Jul 2008 12:22:51 +0900 Subject: pipebuffer: Verify usage flag consistency. Minor cleanups. --- src/gallium/auxiliary/pipebuffer/pb_buffer.h | 19 +++++++++++++++---- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c | 3 +++ src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c | 10 +++++++--- src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c | 6 +++--- src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c | 9 ++++++--- src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c | 9 +++++++-- 8 files changed, 43 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h index 857ea53c78..83ea0be74b 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h @@ -204,13 +204,24 @@ pb_reference(struct pb_buffer **dst, /** - * Utility function to check whether a requested alignment is consistent with - * the provided alignment or not. + * Utility function to check whether the provided alignment is consistent with + * the requested or not. */ -static INLINE int +static INLINE boolean pb_check_alignment(size_t requested, size_t provided) { - return requested <= provided && (provided % requested) == 0; + return requested <= provided && (provided % requested) == 0 ? TRUE : FALSE; +} + + +/** + * Utility function to check whether the provided alignment is consistent with + * the requested or not. + */ +static INLINE boolean +pb_check_usage(unsigned requested, unsigned provided) +{ + return (requested & provided) == provided ? TRUE : FALSE; } diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 7f236887a9..0ffca298cd 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -406,7 +406,7 @@ fenced_buffer_list_create(struct pipe_winsys *winsys) { struct fenced_buffer_list *fenced_list; - fenced_list = (struct fenced_buffer_list *)CALLOC(1, sizeof(*fenced_list)); + fenced_list = CALLOC_STRUCT(fenced_buffer_list); if (!fenced_list) return NULL; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c index 702bef1c04..0d2d6c0c1b 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c @@ -88,6 +88,9 @@ pb_alt_manager_create(struct pb_manager *provider1, { struct pb_alt_manager *mgr; + if(!provider1 || !provider2) + return NULL; + mgr = CALLOC_STRUCT(pb_alt_manager); if (!mgr) return NULL; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c index f1a457dde4..bed4bec4fe 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c @@ -217,7 +217,8 @@ pb_cache_is_buffer_compat(struct pb_cache_buffer *buf, if(!pb_check_alignment(desc->alignment, buf->base.base.alignment)) return FALSE; - /* XXX: check usage too? */ + if(!pb_check_usage(desc->usage, buf->base.base.usage)) + return FALSE; return TRUE; } @@ -282,7 +283,7 @@ pb_cache_manager_create_buffer(struct pb_manager *_mgr, assert(buf->buffer->base.refcount >= 1); assert(pb_check_alignment(desc->alignment, buf->buffer->base.alignment)); - assert((buf->buffer->base.usage & desc->usage) == desc->usage); + assert(pb_check_usage(desc->usage, buf->buffer->base.usage)); assert(buf->buffer->base.size >= size); buf->base.base.refcount = 1; @@ -331,7 +332,10 @@ pb_cache_manager_create(struct pb_manager *provider, { struct pb_cache_manager *mgr; - mgr = (struct pb_cache_manager *)CALLOC(1, sizeof(*mgr)); + if(!provider) + return NULL; + + mgr = CALLOC_STRUCT(pb_cache_manager); if (!mgr) return NULL; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c index 9d809e2f9b..05efd8ce41 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c @@ -117,7 +117,7 @@ fenced_bufmgr_create(struct pb_manager *provider, if(!provider) return NULL; - fenced_mgr = (struct fenced_pb_manager *)CALLOC(1, sizeof(*fenced_mgr)); + fenced_mgr = CALLOC_STRUCT(fenced_pb_manager); if (!fenced_mgr) return NULL; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index 0a1e8c83b1..c51e582611 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -29,7 +29,7 @@ * \file * Buffer manager using the old texture memory manager. * - * \author José Fonseca + * \author José Fonseca */ @@ -271,8 +271,8 @@ mm_bufmgr_create(struct pb_manager *provider, struct pb_manager *mgr; struct pb_desc desc; - assert(provider); - assert(provider->create_buffer); + if(!provider) + return NULL; memset(&desc, 0, sizeof(desc)); desc.alignment = 1 << align2; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c index 528e9528f6..95af08929a 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c @@ -30,8 +30,8 @@ * \file * Batch buffer pool management. * - * \author José Fonseca - * \author Thomas Hellström + * \author José Fonseca + * \author Thomas Hellström */ @@ -229,7 +229,10 @@ pool_bufmgr_create(struct pb_manager *provider, struct pool_buffer *pool_buf; size_t i; - pool = (struct pool_pb_manager *)CALLOC(1, sizeof(*pool)); + if(!provider) + return NULL; + + pool = CALLOC_STRUCT(pool_pb_manager); if (!pool) return NULL; diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c index b9dff09804..598d9ce310 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c @@ -324,8 +324,10 @@ pb_slab_manager_create_buffer(struct pb_manager *_mgr, if(!pb_check_alignment(desc->alignment, mgr->bufSize)) return NULL; - /* XXX: check for compatible buffer usage too? */ - + assert(pb_check_usage(desc->usage, mgr->desc.usage)); + if(!pb_check_usage(desc->usage, mgr->desc.usage)) + return NULL; + _glthread_LOCK_MUTEX(mgr->mutex); if (mgr->slabs.next == &mgr->slabs) { (void) pb_slab_create(mgr); @@ -438,6 +440,9 @@ pb_slab_range_manager_create(struct pb_manager *provider, size_t bufSize; unsigned i; + if(!provider) + return NULL; + mgr = CALLOC_STRUCT(pb_slab_range_manager); if (!mgr) goto out_err0; -- cgit v1.2.3 From d16fcd07f876fe7fb29f5f4e3df4e83ff7de3422 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 2 Jul 2008 12:24:19 +0900 Subject: pipebuffer: Debug buffer manager to detect buffer under- and overflows. It should detect both cpu and gpu buffer overflows. --- src/gallium/auxiliary/pipebuffer/Makefile | 1 + src/gallium/auxiliary/pipebuffer/SConscript | 1 + src/gallium/auxiliary/pipebuffer/pb_bufmgr.h | 11 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c | 297 +++++++++++++++++++++ 4 files changed, 309 insertions(+), 1 deletion(-) create mode 100644 src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile index 1d9b036c07..f9b39d9ce0 100644 --- a/src/gallium/auxiliary/pipebuffer/Makefile +++ b/src/gallium/auxiliary/pipebuffer/Makefile @@ -8,6 +8,7 @@ C_SOURCES = \ pb_buffer_malloc.c \ pb_bufmgr_alt.c \ pb_bufmgr_cache.c \ + pb_bufmgr_debug.c \ pb_bufmgr_fenced.c \ pb_bufmgr_mm.c \ pb_bufmgr_pool.c \ diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript index e52177bc79..56a40dda0d 100644 --- a/src/gallium/auxiliary/pipebuffer/SConscript +++ b/src/gallium/auxiliary/pipebuffer/SConscript @@ -7,6 +7,7 @@ pipebuffer = env.ConvenienceLibrary( 'pb_buffer_malloc.c', 'pb_bufmgr_alt.c', 'pb_bufmgr_cache.c', + 'pb_bufmgr_debug.c', 'pb_bufmgr_fenced.c', 'pb_bufmgr_mm.c', 'pb_bufmgr_pool.c', diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h index 00279f7010..32867029ee 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h @@ -175,7 +175,16 @@ struct pb_manager * pb_alt_manager_create(struct pb_manager *provider1, struct pb_manager *provider2); - + +/** + * Debug buffer manager to detect buffer under- and overflows. + * + * Band size should be a multiple of the largest alignment + */ +struct pb_manager * +pb_debug_manager_create(struct pb_manager *provider, size_t band_size); + + #ifdef __cplusplus } #endif diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c new file mode 100644 index 0000000000..acb9d7ad14 --- /dev/null +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c @@ -0,0 +1,297 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * \file + * Debug buffer manager to detect buffer under- and overflows. + * + * \author José Fonseca + */ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_winsys.h" +#include "pipe/p_thread.h" +#include "pipe/p_util.h" +#include "util/u_double_list.h" +#include "util/u_time.h" + +#include "pb_buffer.h" +#include "pb_bufmgr.h" + + +#ifdef DEBUG + + +/** + * Convenience macro (type safe). + */ +#define SUPER(__derived) (&(__derived)->base) + + +struct pb_debug_manager; + + +/** + * Wrapper around a pipe buffer which adds delayed destruction. + */ +struct pb_debug_buffer +{ + struct pb_buffer base; + + struct pb_buffer *buffer; + struct pb_debug_manager *mgr; + + size_t underflow_size; + size_t overflow_size; +}; + + +struct pb_debug_manager +{ + struct pb_manager base; + + struct pb_manager *provider; + + size_t band_size; +}; + + +static INLINE struct pb_debug_buffer * +pb_debug_buffer(struct pb_buffer *buf) +{ + assert(buf); + return (struct pb_debug_buffer *)buf; +} + + +static INLINE struct pb_debug_manager * +pb_debug_manager(struct pb_manager *mgr) +{ + assert(mgr); + return (struct pb_debug_manager *)mgr; +} + + +static const uint8_t random_pattern[32] = { + 0xaf, 0xcf, 0xa5, 0xa2, 0xc2, 0x63, 0x15, 0x1a, + 0x7e, 0xe2, 0x7e, 0x84, 0x15, 0x49, 0xa2, 0x1e, + 0x49, 0x63, 0xf5, 0x52, 0x74, 0x66, 0x9e, 0xc4, + 0x6d, 0xcf, 0x2c, 0x4a, 0x74, 0xe6, 0xfd, 0x94 +}; + + +static INLINE void +fill_random_pattern(uint8_t *dst, size_t size) +{ + unsigned i = 0; + while(size--) { + *dst++ = random_pattern[i++]; + i &= sizeof(random_pattern) - 1; + } +} + + +static INLINE boolean +check_random_pattern(const uint8_t *dst, size_t size) +{ + unsigned i = 0; + while(size--) { + if(*dst++ != random_pattern[i++]) + return FALSE; + i &= sizeof(random_pattern) - 1; + } + return TRUE; +} + + +static void +pb_debug_buffer_destroy(struct pb_buffer *_buf) +{ + struct pb_debug_buffer *buf = pb_debug_buffer(_buf); + uint8_t *map; + + assert(!buf->base.base.refcount); + + map = pb_map(buf->buffer, PIPE_BUFFER_USAGE_CPU_READ); + assert(map); + if(map) { + if(!check_random_pattern(map, buf->underflow_size)) { + debug_printf("buffer underflow\n"); + debug_assert(0); + } + if(!check_random_pattern(map + buf->underflow_size + buf->base.base.size, + buf->overflow_size)) { + debug_printf("buffer overflow\n"); + debug_assert(0); + } + pb_unmap(buf->buffer); + } + + pb_reference(&buf->buffer, NULL); + FREE(buf); +} + + +static void * +pb_debug_buffer_map(struct pb_buffer *_buf, + unsigned flags) +{ + struct pb_debug_buffer *buf = pb_debug_buffer(_buf); + void *map = pb_map(buf->buffer, flags); + if(!map) + return NULL; + return (uint8_t *)map + buf->underflow_size; +} + + +static void +pb_debug_buffer_unmap(struct pb_buffer *_buf) +{ + struct pb_debug_buffer *buf = pb_debug_buffer(_buf); + pb_unmap(buf->buffer); +} + + +static void +pb_debug_buffer_get_base_buffer(struct pb_buffer *_buf, + struct pb_buffer **base_buf, + unsigned *offset) +{ + struct pb_debug_buffer *buf = pb_debug_buffer(_buf); + pb_get_base_buffer(buf->buffer, base_buf, offset); + *offset += buf->underflow_size; +} + + +const struct pb_vtbl +pb_debug_buffer_vtbl = { + pb_debug_buffer_destroy, + pb_debug_buffer_map, + pb_debug_buffer_unmap, + pb_debug_buffer_get_base_buffer +}; + + +static struct pb_buffer * +pb_debug_manager_create_buffer(struct pb_manager *_mgr, + size_t size, + const struct pb_desc *desc) +{ + struct pb_debug_manager *mgr = pb_debug_manager(_mgr); + struct pb_debug_buffer *buf; + struct pb_desc real_desc; + size_t real_size; + uint8_t *map; + + buf = CALLOC_STRUCT(pb_debug_buffer); + if(!buf) + return NULL; + + real_size = size + 2*mgr->band_size; + real_desc = *desc; + real_desc.usage |= PIPE_BUFFER_USAGE_CPU_WRITE; + real_desc.usage |= PIPE_BUFFER_USAGE_CPU_READ; + + buf->buffer = mgr->provider->create_buffer(mgr->provider, + real_size, + &real_desc); + if(!buf->buffer) { + FREE(buf); + return NULL; + } + + assert(buf->buffer->base.refcount >= 1); + assert(pb_check_alignment(real_desc.alignment, buf->buffer->base.alignment)); + assert(pb_check_usage(real_desc.usage, buf->buffer->base.usage)); + assert(buf->buffer->base.size >= real_size); + + buf->base.base.refcount = 1; + buf->base.base.alignment = desc->alignment; + buf->base.base.usage = desc->usage; + buf->base.base.size = size; + + buf->base.vtbl = &pb_debug_buffer_vtbl; + buf->mgr = mgr; + + buf->underflow_size = mgr->band_size; + buf->overflow_size = buf->buffer->base.size - buf->underflow_size - size; + + map = pb_map(buf->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); + assert(map); + if(map) { + fill_random_pattern(map, buf->underflow_size); + fill_random_pattern(map + buf->underflow_size + size, buf->overflow_size); + pb_unmap(buf->buffer); + } + + return &buf->base; +} + + +static void +pb_debug_manager_destroy(struct pb_manager *_mgr) +{ + struct pb_debug_manager *mgr = pb_debug_manager(_mgr); + mgr->provider->destroy(mgr->provider); + FREE(mgr); +} + + +struct pb_manager * +pb_debug_manager_create(struct pb_manager *provider, size_t band_size) +{ + struct pb_debug_manager *mgr; + + if(!provider) + return NULL; + + mgr = CALLOC_STRUCT(pb_debug_manager); + if (!mgr) + return NULL; + + mgr->base.destroy = pb_debug_manager_destroy; + mgr->base.create_buffer = pb_debug_manager_create_buffer; + mgr->provider = provider; + mgr->band_size = band_size; + + return &mgr->base; +} + + +#else /* !DEBUG */ + + +struct pb_manager * +pb_debug_manager_create(struct pb_manager *provider, size_t band_size) +{ + return provider; +} + + +#endif /* !DEBUG */ -- cgit v1.2.3 From cc31eecbcb90dabacabac3e6be7c01cfe3a7a2a6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Wed, 2 Jul 2008 12:10:15 +0200 Subject: gallium: Allow draw module to work on non-x86 platforms again. --- src/gallium/auxiliary/draw/draw_vs.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index 978954e91c..f798b20492 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -160,8 +160,10 @@ draw_vs_init( struct draw_context *draw ) return FALSE; draw->vs.aos_machine = draw_vs_aos_machine(); +#ifdef PIPE_ARCH_X86 if (!draw->vs.aos_machine) return FALSE; +#endif return TRUE; } -- cgit v1.2.3 From 489fc4d10a57538de59a89e19ce752e4b7253d22 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 2 Jul 2008 20:08:27 +0200 Subject: mesa: fix issues around multisample enable multisample enable is enabled by default, however gl mandates multisample rendering rules only apply if there's also a multisampled buffer. --- src/mesa/drivers/dri/r300/r300_render.c | 2 +- src/mesa/main/mtypes.h | 1 + src/mesa/main/multisample.c | 2 +- src/mesa/main/state.c | 17 +++++++++++++++++ src/mesa/state_tracker/st_atom_rasterizer.c | 2 +- src/mesa/swrast/s_points.c | 2 +- 6 files changed, 22 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c index eee1e803a0..c809679e6c 100644 --- a/src/mesa/drivers/dri/r300/r300_render.c +++ b/src/mesa/drivers/dri/r300/r300_render.c @@ -359,7 +359,7 @@ static int r300Fallback(GLcontext * ctx) if (!r300->disable_lowimpact_fallback) { FALLBACK_IF(ctx->Polygon.StippleFlag); - FALLBACK_IF(ctx->Multisample.Enabled); + FALLBACK_IF(ctx->Multisample._Enabled); FALLBACK_IF(ctx->Line.StippleFlag); FALLBACK_IF(ctx->Line.SmoothFlag); FALLBACK_IF(ctx->Point.SmoothFlag); diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 0a065541e1..00e7d5d395 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -961,6 +961,7 @@ struct gl_list_extensions struct gl_multisample_attrib { GLboolean Enabled; + GLboolean _Enabled; /**< true if Enabled and multisample buffer */ GLboolean SampleAlphaToCoverage; GLboolean SampleAlphaToOne; GLboolean SampleCoverage; diff --git a/src/mesa/main/multisample.c b/src/mesa/main/multisample.c index e138087436..b9cfad9216 100644 --- a/src/mesa/main/multisample.c +++ b/src/mesa/main/multisample.c @@ -57,7 +57,7 @@ _mesa_SampleCoverageARB(GLclampf value, GLboolean invert) void _mesa_init_multisample(GLcontext *ctx) { - ctx->Multisample.Enabled = GL_FALSE; + ctx->Multisample.Enabled = GL_TRUE; ctx->Multisample.SampleAlphaToCoverage = GL_FALSE; ctx->Multisample.SampleAlphaToOne = GL_FALSE; ctx->Multisample.SampleCoverage = GL_FALSE; diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c index 315253d90a..344af91e17 100644 --- a/src/mesa/main/state.c +++ b/src/mesa/main/state.c @@ -288,6 +288,20 @@ update_viewport_matrix(GLcontext *ctx) } +/** + * Update derived multisample state. + */ +static void +update_multisample(GLcontext *ctx) +{ + ctx->Multisample._Enabled = GL_FALSE; + if (ctx->DrawBuffer) { + if (ctx->DrawBuffer->Visual.sampleBuffers) + ctx->Multisample._Enabled = GL_TRUE; + } +} + + /** * Update derived color/blend/logicop state. */ @@ -425,6 +439,9 @@ _mesa_update_state_locked( GLcontext *ctx ) if (new_state & (_NEW_BUFFERS | _NEW_VIEWPORT)) update_viewport_matrix(ctx); + if (new_state & _NEW_MULTISAMPLE) + update_multisample( ctx ); + if (new_state & _NEW_COLOR) update_color( ctx ); diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 87a91d56d0..ff40bb4312 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -254,7 +254,7 @@ static void update_raster_state( struct st_context *st ) raster->line_stipple_factor = ctx->Line.StippleFactor - 1; /* _NEW_MULTISAMPLE */ - if (ctx->Multisample.Enabled) + if (ctx->Multisample._Enabled) raster->multisample = 1; /* _NEW_SCISSOR */ diff --git a/src/mesa/swrast/s_points.c b/src/mesa/swrast/s_points.c index c0f32e9c91..f4b3650210 100644 --- a/src/mesa/swrast/s_points.c +++ b/src/mesa/swrast/s_points.c @@ -250,7 +250,7 @@ smooth_point(GLcontext *ctx, const SWvertex *vert) size = CLAMP(size, ctx->Const.MinPointSizeAA, ctx->Const.MaxPointSizeAA); /* alpha attenuation / fade factor */ - if (ctx->Multisample.Enabled) { + if (ctx->Multisample._Enabled) { if (vert->pointSize >= ctx->Point.Threshold) { alphaAtten = 1.0F; } -- cgit v1.2.3 From 511733b862d2cb6836beb0965bdc5d699e6a2233 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 12:40:03 -0600 Subject: mesa: added _mesa_print_swizzle() debugging helper --- src/mesa/shader/prog_print.c | 13 +++++++++++++ src/mesa/shader/prog_print.h | 3 +++ 2 files changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index 2c5e03acc2..11f82c1fc1 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -356,6 +356,19 @@ _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) } +void +_mesa_print_swizzle(GLuint swizzle) +{ + if (swizzle == SWIZZLE_XYZW) { + _mesa_printf(".xyzw\n"); + } + else { + char *s = _mesa_swizzle_string(swizzle, 0, 0); + _mesa_printf("%s\n", s); + } +} + + static const char * writemask_string(GLuint writeMask) { diff --git a/src/mesa/shader/prog_print.h b/src/mesa/shader/prog_print.h index 36c47e0dff..3cdb1b195e 100644 --- a/src/mesa/shader/prog_print.h +++ b/src/mesa/shader/prog_print.h @@ -43,6 +43,9 @@ _mesa_condcode_string(GLuint condcode); extern const char * _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended); +extern void +_mesa_print_swizzle(GLuint swizzle); + extern void _mesa_print_alu_instruction(const struct prog_instruction *inst, const char *opcode_string, GLuint numRegs); -- cgit v1.2.3 From 019ad5e284db08b46b484b409c1a54c302afd50b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 12:41:18 -0600 Subject: gallium: replace 128 with MAX_LABELS --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 2 +- src/gallium/auxiliary/tgsi/exec/tgsi_exec.h | 4 +++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 13b8c2e5bf..9649396c9b 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -214,7 +214,7 @@ tgsi_exec_machine_bind_shader( break; case TGSI_TOKEN_TYPE_INSTRUCTION: - assert( labels->count < 128 ); + assert( labels->count < MAX_LABELS ); labels->labels[labels->count][0] = instno; labels->labels[labels->count][1] = pointer; diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h index 19bd78df3d..ea182bc454 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h @@ -34,6 +34,8 @@ extern "C" { #endif +#define MAX_LABELS 1024 + #define NUM_CHANNELS 4 /* R,G,B,A */ #define QUAD_SIZE 4 /* 4 pixel/quad */ @@ -93,7 +95,7 @@ struct tgsi_sampler */ struct tgsi_exec_labels { - unsigned labels[128][2]; + unsigned labels[MAX_LABELS][2]; unsigned count; }; -- cgit v1.2.3 From 789fae163dbdbd7669b0442bd1a3ca4fea4ec259 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 12:45:55 -0600 Subject: mesa: fix all(bvec2) function typo, add missing bvec2/3/4() constuctors cherry-picked from master --- .../shader/slang/library/slang_common_builtin.gc | 2 +- src/mesa/shader/slang/library/slang_core.gc | 38 ++++++++++++++++++++++ 2 files changed, 39 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index 26080cf26c..356f5f9d8e 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -1559,7 +1559,7 @@ bool any(const bvec4 v) //// all -bool all (const vec2 v) +bool all (const bvec2 v) { float prod; __asm vec4_multiply prod.x, v.x, v.y; diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index ffa57f74c2..78e3ea9fda 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -350,6 +350,19 @@ bvec2 __constructor(const int i) __asm vec4_seq __retVal.xy, i.xx, zero; } +bvec2 __constructor(const vec2 v) +{ + const vec2 zero = vec2(0.0, 0.0); + __asm vec4_seq __retVal.xy, v, zero; +} + +bvec2 __constructor(const ivec2 v) +{ + const ivec2 zero = ivec2(0, 0); + __asm vec4_seq __retVal.xy, v, zero; +} + + //// bvec3 constructors @@ -377,6 +390,19 @@ bvec3 __constructor(const int i) __asm vec4_seq __retVal.xyz, i.xxx, zero; } +bvec3 __constructor(const vec3 v) +{ + const vec3 zero = vec3(0.0, 0.0, 0.0); + __asm vec4_seq __retVal.xyz, v, zero; +} + +bvec3 __constructor(const ivec3 v) +{ + const ivec3 zero = ivec3(0, 0, 0); + __asm vec4_seq __retVal.xyz, v, zero; +} + + //// bvec4 constructors @@ -405,6 +431,18 @@ bvec4 __constructor(const int i) __asm vec4_seq __retVal, i.xxxx, zero; } +bvec4 __constructor(const vec4 v) +{ + const vec4 zero = vec4(0.0, 0.0, 0.0, 0.0); + __asm vec4_seq __retVal, v, zero; +} + +bvec4 __constructor(const ivec4 v) +{ + const ivec4 zero = ivec4(0, 0, 0, 0); + __asm vec4_seq __retVal, v, zero; +} + //// mat2 constructors -- cgit v1.2.3 From eab88236b3ef5a4f861c1b99dc7220abd5c9f6c7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 12:46:18 -0600 Subject: mesa: regenerated files --- .../shader/slang/library/slang_common_builtin_gc.h | 2 +- src/mesa/shader/slang/library/slang_core_gc.h | 53 ++++++++++++++-------- 2 files changed, 35 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index c8d1ffc174..42f997db8d 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -671,7 +671,7 @@ 101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122, 0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18, 118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,10,118,0,0,0,1,3,2,0,9,1, +115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1, 112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59, 120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114, 101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,8,18,118,0,59, diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index 1604d5a034..b8e6d1a990 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -61,26 +61,41 @@ 59,120,121,0,0,18,102,0,59,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,3,2, 1,6,1,122,101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95, 115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,59,120,120,0,0,18,122,101, -114,111,0,0,0,0,1,0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9, -102,0,0,0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17, -48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,102,0,59,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,5,105,0,0,0,1,3,2,1,7, -1,122,101,114,111,0,2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101, -99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,59,120,120,120, -0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,1,1,0, -1,98,52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,4,1,1,1,0,1,98,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,119,0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,4,1,1,1,0,9, -102,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17, +114,111,0,0,0,0,1,0,2,1,1,1,0,10,118,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17, 48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, -108,0,0,18,102,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5,105,0,0,0,1,3,2, -1,8,1,122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0, -0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,59,120,120,120, -120,0,0,18,122,101,114,111,0,0,0,0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9, +108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,3,2,1,6,1, +122,101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115, +101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1, +0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,3,2,1, +11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0, +0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,59, +120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,5,105,0,0,0,1,3,2,1,7,1,122,101,114,111,0, +2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,59,120,120,120,0,0,18,122,101,114, +111,0,0,0,0,1,0,3,1,1,1,0,11,118,0,0,0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0, +48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,7,118,0,0,0, +1,3,2,1,7,1,122,101,114,111,0,2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4, +118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18, +122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,1,1,0,1,98, +52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,4,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,119,0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,4,1,1,1,0,9,102,0, +0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0, +48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0, +0,18,102,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5,105,0,0,0,1,3,2,1,8,1, +122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4, +118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,59,120,120,120,120,0,0, +18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,12,118,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101, +99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52, +95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1, +1,1,0,8,118,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0, +16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0, +18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9, 109,48,49,0,0,1,1,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0, 18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9, 18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116, -- cgit v1.2.3 From 8f98242d40b58c517181bc4777f4910ad24e6866 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 12:47:18 -0600 Subject: mesa: disable some debug assertions We can sometimes fail these assertions because of how swizzled storage works. Will revisit someday. --- src/mesa/shader/slang/slang_vartable.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_vartable.c b/src/mesa/shader/slang/slang_vartable.c index 1d817000c6..26411255df 100644 --- a/src/mesa/shader/slang/slang_vartable.c +++ b/src/mesa/shader/slang/slang_vartable.c @@ -298,10 +298,15 @@ _slang_free_temp(slang_var_table *vt, slang_ir_storage *store) if (dbg) printf("Free temp sz %d at %d (level %d)\n", store->Size, r, t->Level); if (store->Size == 1) { const GLuint comp = GET_SWZ(store->Swizzle, 0); + /* we can actually fail some of these assertions because of the + * troublesome IR_SWIZZLE handling. + */ +#if 0 assert(store->Swizzle == MAKE_SWIZZLE4(comp, comp, comp, comp)); assert(comp < 4); assert(t->ValSize[r * 4 + comp] == 1); assert(t->Temps[r * 4 + comp] == TEMP); +#endif t->Temps[r * 4 + comp] = FREE; } else { -- cgit v1.2.3 From adc709e9960966bc03e510bc793b9d2703b3e863 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 17:05:56 -0600 Subject: mesa: added some missing equal() notEqual() intrinsics cherry-picked from master --- .../shader/slang/library/slang_common_builtin.gc | 33 ++++++++++++++++++++++ 1 file changed, 33 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index 356f5f9d8e..f741c657e4 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -1496,6 +1496,23 @@ bvec4 equal(const ivec4 u, const ivec4 v) __asm vec4_seq __retVal, u, v; } +bvec2 equal(const bvec2 u, const bvec2 v) +{ + __asm vec4_seq __retVal.xy, u, v; +} + +bvec3 equal(const bvec3 u, const bvec3 v) +{ + __asm vec4_seq __retVal.xyz, u, v; +} + +bvec4 equal(const bvec4 u, const bvec4 v) +{ + __asm vec4_seq __retVal, u, v; +} + + + //// notEqual @@ -1529,6 +1546,22 @@ bvec4 notEqual(const ivec4 u, const ivec4 v) __asm vec4_sne __retVal, u, v; } +bvec2 notEqual(const bvec2 u, const bvec2 v) +{ + __asm vec4_sne __retVal.xy, u, v; +} + +bvec3 notEqual(const bvec3 u, const bvec3 v) +{ + __asm vec4_sne __retVal.xyz, u, v; +} + +bvec4 notEqual(const bvec4 u, const bvec4 v) +{ + __asm vec4_sne __retVal, u, v; +} + + //// any -- cgit v1.2.3 From 16caeeab4b495702b54963816ae45976b2563423 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 17:06:26 -0600 Subject: mesa: regenerated file --- .../shader/slang/library/slang_common_builtin_gc.h | 315 +++++++++++---------- 1 file changed, 163 insertions(+), 152 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index 42f997db8d..260a768de6 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -648,162 +648,173 @@ 1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121, 122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1, 4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0, -2,0,110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110, -111,116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116, -69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, -95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1, -0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1, -1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, -0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118, -0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0, -0,0,1,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100, -100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0, -0,0,1,0,1,0,97,110,121,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100, -100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97, -100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101, +2,0,101,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108, +0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1, +1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, +18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118, +101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4, +0,110,111,116,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117, +97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0, +7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0, +1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, +18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118, +101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4, +0,110,111,116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,2, +118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0, +0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0, +3,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120, +0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, +120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97, +110,121,0,1,1,0,4,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115, +117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18, +115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97, +100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101, 99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17, -48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,4,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101, -99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118, -101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122, -0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18, -118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1, -112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59, -120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,8,18,118,0,59, -120,0,18,118,0,59,121,0,34,0,0,1,0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,112,114,111,100, -0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118, -0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114, -111,100,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0, -0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,4,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0, -59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0, -18,112,114,111,100,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,119,0,0,0,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59, -120,0,0,17,48,0,48,0,0,0,0,0,1,0,2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101, -113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,0,110, -111,116,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118, -101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0, -12,0,116,101,120,116,117,114,101,49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111, -111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18, -115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101, -49,68,80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1, -4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, -108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,121,121,0,0,0,0,1,0,12,0,116,101,120,116,117, -114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100, +48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0, +18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, +18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,8,18,118,0,59,120,0,18,118,0,59,121,0,34,0,0,1, +0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0, +0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114, +111,100,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1, +1,0,4,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0, +18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59, +120,0,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0, +2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4, +118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17, +48,0,48,0,0,0,0,0,1,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95, +95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101, +49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52, +95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, +111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115, +97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112, +49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, +100,0,59,120,121,121,121,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0, +16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, +120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, +111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,0,1,1,0,17,115,97,109,112,108,101, +114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114, +101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116, +101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,11,99, +111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108, +0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,12,0, +116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0, +12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86, +97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120, +116,117,114,101,51,68,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0, +1,4,118,101,99,52,95,116,101,120,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, +101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111, +106,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, +95,116,101,120,112,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, +18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,67,117,98,101,0,1,1,0,19,115,97, +109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,99,117, +98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, +100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,49,68,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0, +11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97, +108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100, +111,119,49,68,80,114,111,106,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100, 0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97, -109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,0, -1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116, -101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, -111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97, -109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50, -100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0, -59,120,121,122,122,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17, +109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,0,1,1, +0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, +120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, +114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,21,115,97,109,112,108, +101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95, +95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0, +12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1, +0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114, +101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116, +101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101,114,0, +0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95, +95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121, +122,122,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22, 115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120, -112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, -114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,51,68,0,1,1,0,18,115,97,109,112,108,101,114,0, -0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,51,100,0,18,95,95,114,101,116, -86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120, -116,117,114,101,51,68,80,114,111,106,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, -114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18, -115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101, -67,117,98,101,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118, -101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, -101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,49,68,0,1,1,0,20,115,97, -109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100, -0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0, -0,1,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1, -1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116, -86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97, -100,111,119,50,68,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4, -118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0, -21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, -120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, -111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,0,1,1,0,22,115,97,109, -112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101, -99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, -100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115, -97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112, +112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, +99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,0,1,1,0,23,115,97, +109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114, +101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, +114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,80,114,111,106,0,1,1,0,23,115, +97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112, 95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, -111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116, -80,114,111,106,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118, -101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109, -112,108,101,114,0,0,18,99,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,82,101, -99,116,0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99, -52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,80,114, -111,106,0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99, -52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, -101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,9,0,110,111,105,115,101,49,0,1,1,0,9,120,0,0,0,1,4, -102,108,111,97,116,95,110,111,105,115,101,49,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1, -0,9,0,110,111,105,115,101,49,0,1,1,0,10,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,50, -0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,0,110,111,105,115,101,49,0,1,1,0,11,120, -0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,51,0,18,95,95,114,101,116,86,97,108,0,0,18,120, -0,0,0,0,1,0,9,0,110,111,105,115,101,49,0,1,1,0,12,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105, -115,101,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,10,0,110,111,105,115,101,50,0,1, -1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0, -0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57, -0,51,52,0,0,46,0,0,20,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114, +111,114,100,0,0,0,0,1,0,9,0,110,111,105,115,101,49,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95, +110,111,105,115,101,49,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,0,110,111,105,115, +101,49,0,1,1,0,10,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,50,0,18,95,95,114,101, +116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,0,110,111,105,115,101,49,0,1,1,0,11,120,0,0,0,1,4,102,108, +111,97,116,95,110,111,105,115,101,51,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,0, +110,111,105,115,101,49,0,1,1,0,12,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,52,0,18, +95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,9,120,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0, +46,0,0,20,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, +121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54, +0,0,0,0,46,0,0,20,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0, +17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,12, +120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, +99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0, +0,20,0,0,1,0,11,0,110,111,105,115,101,51,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0, +58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,0,1,0,11,0,110, +111,105,115,101,51,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111, +105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115, +101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17, +53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,0,11,0,110,111,105,115,101,51,0,1,1,0,11, +120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, +99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0, +52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,0,1,0,11,0,110,111,105,115, +101,51,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101, +49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18, +120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55, +55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18, +120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49, +51,0,49,57,0,0,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1,1,0,9,120,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0, +46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,17,50, +51,0,53,52,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114, 101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116, 86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0, -0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,11,120,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57, -0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,0,10,0,110,111,105,115,101, -50,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0, -18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0, -58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0, -0,0,0,46,0,0,20,0,0,1,0,11,0,110,111,105,115,101,51,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20, -0,0,1,0,11,0,110,111,105,115,101,51,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, -110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0, +0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115, +101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17, +50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1,1,0, +11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0, +0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, +99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0, +52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,50,51,0,53,52,0,0,0, +17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1, +1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120, +0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118, +101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0, 46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58, -118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,0,11,0,110,111,105, -115,101,51,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115, -101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0, -18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, -99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,0,1,0,11,0, -110,111,105,115,101,51,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110, -111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105, -115,101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51, -0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105, -115,101,49,0,18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48, -52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1,1,0,9,120,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0, -46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53, -0,52,55,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18, -120,0,17,50,51,0,53,52,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1,1,0,10,120,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57, -0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110, -111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, -99,50,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101, -52,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0, -18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0, -58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0, -17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,50,51,0,53,52, -0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101, -52,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0, -18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0, -58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0, -0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0, -58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49, -57,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18, -120,0,58,118,101,99,52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,17, -51,55,0,52,56,0,0,0,0,46,0,0,20,0,0,0 +118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57, +0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120, +0,58,118,101,99,52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,17,51,55, +0,52,56,0,0,0,0,46,0,0,20,0,0,0 -- cgit v1.2.3 From 088c42c5c3cb34bac98fce95e4f3d08642834112 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 17:08:09 -0600 Subject: mesa: fix error codes in _mesa_shader_source(), _mesa_get_shader_source() If the 'shader' parameter is wrong, need to either generate GL_INVALID_VALUE or GL_INVALID_OPERATION. It depends on whether 'shader' actually names a 'program' or is a totally unknown ID. There might be other cases to fix... cherry-picked from master --- src/mesa/shader/shader_api.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index a005c902e1..26d0b3d476 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -875,7 +875,12 @@ _mesa_get_shader_source(GLcontext *ctx, GLuint shader, GLsizei maxLength, { struct gl_shader *sh = _mesa_lookup_shader(ctx, shader); if (!sh) { - _mesa_error(ctx, GL_INVALID_VALUE, "glGetShaderSource(shader)"); + GLenum err; + if (_mesa_lookup_shader_program(ctx, shader)) + err = GL_INVALID_OPERATION; + else + err = GL_INVALID_VALUE; + _mesa_error(ctx, err, "glGetShaderSource(shader)"); return; } copy_string(sourceOut, maxLength, length, sh->Source); @@ -966,7 +971,12 @@ _mesa_shader_source(GLcontext *ctx, GLuint shader, const GLchar *source) { struct gl_shader *sh = _mesa_lookup_shader(ctx, shader); if (!sh) { - _mesa_error(ctx, GL_INVALID_VALUE, "glShaderSource(shaderObj)"); + GLenum err; + if (_mesa_lookup_shader_program(ctx, shader)) + err = GL_INVALID_OPERATION; + else + err = GL_INVALID_VALUE; + _mesa_error(ctx, err, "glShaderSource(shaderObj)"); return; } -- cgit v1.2.3 From 40739d4ae95054d45fa86e9668ad29fcbafd67e9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 17:08:28 -0600 Subject: mesa: added some debug code (disabled) cherry-picked from master --- src/mesa/shader/slang/slang_compile.c | 6 ++++++ src/mesa/shader/slang/slang_log.c | 3 +++ 2 files changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 8485103129..e2117c03e6 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -2129,6 +2129,12 @@ compile_shader(GLcontext *ctx, slang_code_object * object, GLboolean success; grammar id = 0; +#if 0 /* for debug */ + _mesa_printf("********* COMPILE SHADER ***********\n"); + _mesa_printf("%s\n", shader->Source); + _mesa_printf("************************************\n"); +#endif + assert(program); _slang_code_object_dtr(object); diff --git a/src/mesa/shader/slang/slang_log.c b/src/mesa/shader/slang/slang_log.c index e9234bafec..c963914fe3 100644 --- a/src/mesa/shader/slang/slang_log.c +++ b/src/mesa/shader/slang/slang_log.c @@ -86,6 +86,9 @@ slang_info_log_message(slang_info_log * log, const char *prefix, } slang_string_concat(log->text, msg); slang_string_concat(log->text, "\n"); +#if 0 /* debug */ + _mesa_printf("Mesa GLSL error/warning: %s\n", log->text); +#endif return 1; } -- cgit v1.2.3 From dff477a5e27c39aa61aac8e1555c345211e832b8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 17:08:47 -0600 Subject: mesa: when linking a shader program, make sure all the shaders compiled OK cherry-picked from master --- src/mesa/shader/slang/slang_link.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index dd4990ba02..f08b53e8b7 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -388,6 +388,14 @@ _slang_link(GLcontext *ctx, _mesa_clear_shader_program_data(ctx, shProg); + /* check that all programs compiled successfully */ + for (i = 0; i < shProg->NumShaders; i++) { + if (!shProg->Shaders[i]->CompileStatus) { + link_error(shProg, "linking with uncompiled shader\n"); + return; + } + } + shProg->Uniforms = _mesa_new_uniform_list(); shProg->Varying = _mesa_new_parameter_list(); -- cgit v1.2.3 From 39b9b05313c8b8fce9b80e96819aded479e382c9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 17:10:42 -0600 Subject: mesa: additional GLSL built-in constants --- src/mesa/main/mfeatures.h | 1 + src/mesa/shader/slang/slang_simplify.c | 32 +++++++++++++++++++------------- 2 files changed, 20 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mfeatures.h b/src/mesa/main/mfeatures.h index c3c337ea90..b08c017ec8 100644 --- a/src/mesa/main/mfeatures.h +++ b/src/mesa/main/mfeatures.h @@ -44,6 +44,7 @@ #define FEATURE_dlist _HAVE_FULL_GL #define FEATURE_draw_read_buffer _HAVE_FULL_GL #define FEATURE_drawpix _HAVE_FULL_GL +#define FEATURE_es2_glsl 0 #define FEATURE_evaluators _HAVE_FULL_GL #define FEATURE_feedback _HAVE_FULL_GL #define FEATURE_fixedpt 0 diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c index 21d004db88..158d6bc8cf 100644 --- a/src/mesa/shader/slang/slang_simplify.c +++ b/src/mesa/shader/slang/slang_simplify.c @@ -49,20 +49,26 @@ _slang_lookup_constant(const char *name) struct constant_info { const char *Name; const GLenum Token; + GLint Divisor; }; static const struct constant_info info[] = { - { "gl_MaxClipPlanes", GL_MAX_CLIP_PLANES }, - { "gl_MaxCombinedTextureImageUnits", GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS }, - { "gl_MaxDrawBuffers", GL_MAX_DRAW_BUFFERS }, - { "gl_MaxFragmentUniformComponents", GL_MAX_FRAGMENT_UNIFORM_COMPONENTS }, - { "gl_MaxLights", GL_MAX_LIGHTS }, - { "gl_MaxTextureUnits", GL_MAX_TEXTURE_UNITS }, - { "gl_MaxTextureCoords", GL_MAX_TEXTURE_COORDS }, - { "gl_MaxVertexAttribs", GL_MAX_VERTEX_ATTRIBS }, - { "gl_MaxVertexUniformComponents", GL_MAX_VERTEX_UNIFORM_COMPONENTS }, - { "gl_MaxVaryingFloats", GL_MAX_VARYING_FLOATS }, - { "gl_MaxVertexTextureImageUnits", GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS }, - { "gl_MaxTextureImageUnits", GL_MAX_TEXTURE_IMAGE_UNITS }, + { "gl_MaxClipPlanes", GL_MAX_CLIP_PLANES, 1 }, + { "gl_MaxCombinedTextureImageUnits", GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS, 1 }, + { "gl_MaxDrawBuffers", GL_MAX_DRAW_BUFFERS, 1 }, + { "gl_MaxFragmentUniformComponents", GL_MAX_FRAGMENT_UNIFORM_COMPONENTS, 1 }, + { "gl_MaxLights", GL_MAX_LIGHTS, 1 }, + { "gl_MaxTextureUnits", GL_MAX_TEXTURE_UNITS, 1 }, + { "gl_MaxTextureCoords", GL_MAX_TEXTURE_COORDS, 1 }, + { "gl_MaxVertexAttribs", GL_MAX_VERTEX_ATTRIBS, 1 }, + { "gl_MaxVertexUniformComponents", GL_MAX_VERTEX_UNIFORM_COMPONENTS, 1 }, + { "gl_MaxVaryingFloats", GL_MAX_VARYING_FLOATS, 1 }, + { "gl_MaxVertexTextureImageUnits", GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS, 1 }, + { "gl_MaxTextureImageUnits", GL_MAX_TEXTURE_IMAGE_UNITS, 1 }, +#if FEATURE_es2_glsl + { "gl_MaxVertexUniformVectors", GL_MAX_VERTEX_UNIFORM_COMPONENTS, 4 }, + { "gl_MaxVaryingVectors", GL_MAX_VARYING_FLOATS, 4 }, + { "gl_MaxFragmentUniformVectors", GL_MAX_FRAGMENT_UNIFORM_COMPONENTS, 4 }, +#endif { NULL, 0 } }; GLuint i; @@ -73,7 +79,7 @@ _slang_lookup_constant(const char *name) GLint value = -1.0; _mesa_GetIntegerv(info[i].Token, &value); ASSERT(value >= 0); /* sanity check that glGetFloatv worked */ - return value; + return value / info[i].Divisor; } } return -1; -- cgit v1.2.3 From 8fb4d602db48d425380e5508e3fd71cbdc2e7095 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 19:05:18 -0600 Subject: gallium: nr_attrs was off by one, updated comments, minor code movement --- src/gallium/drivers/softpipe/sp_setup.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 543d86a5cb..e99df9d018 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -1266,12 +1266,14 @@ void setup_prepare( struct setup_context *setup ) sp->framebuffer.zsbuf->status = PIPE_SURFACE_STATUS_DEFINED; } + /* Note: nr_attrs is only used for debugging (vertex printing) */ { const struct sp_fragment_shader *fs = setup->softpipe->fs; - setup->quad.nr_attrs = fs->info.num_inputs; - sp->quad.first->begin(sp->quad.first); + setup->quad.nr_attrs = fs->info.num_inputs + 1; /* +1 for vert pos */ } + sp->quad.first->begin(sp->quad.first); + if (sp->reduced_api_prim == PIPE_PRIM_TRIANGLES && sp->rasterizer->fill_cw == PIPE_POLYGON_MODE_FILL && sp->rasterizer->fill_ccw == PIPE_POLYGON_MODE_FILL) { -- cgit v1.2.3 From dad8a7c90d96f37fedfe912a10ba3754c1666161 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 19:07:11 -0600 Subject: gallium: fix a bug in vertex program output mapping Need to translate VERT_RESULT_PSIZ, BFC0, BFC1 to TGSI shader output slots after all other attributes have been handled. This fixes a bug where generic vertex program outputs (varying vars) could get mapped to the same slot at point size or back-face colors. --- src/mesa/state_tracker/st_atom_shader.c | 33 ++++++++++++++++++++++++--------- 1 file changed, 24 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index a62ea8161c..4fa304ede0 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -175,6 +175,7 @@ find_translated_vp(struct st_context *st, GLuint outAttr, dummySlot; const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten; GLuint numVpOuts = 0; + GLboolean emitPntSize = GL_FALSE, emitBFC0 = GL_FALSE, emitBFC1 = GL_FALSE; /* Compute mapping of vertex program outputs to slots, which depends * on the fragment program's input->slot mapping. @@ -199,19 +200,28 @@ find_translated_vp(struct st_context *st, numVpOuts++; } } - else if (outAttr == VERT_RESULT_PSIZ || - outAttr == VERT_RESULT_BFC0 || - outAttr == VERT_RESULT_BFC1) { - /* backface colors go into last slots */ - xvp->output_to_slot[outAttr] = numVpOuts++; - } + else if (outAttr == VERT_RESULT_PSIZ) + emitPntSize = GL_TRUE; + else if (outAttr == VERT_RESULT_BFC0) + emitBFC0 = GL_TRUE; + else if (outAttr == VERT_RESULT_BFC1) + emitBFC1 = GL_TRUE; } - /* - printf("output_to_slot[%d] = %d\n", outAttr, +#if 0 /*debug*/ + printf("assign output_to_slot[%d] = %d\n", outAttr, xvp->output_to_slot[outAttr]); - */ +#endif } + /* must do these last */ + if (emitPntSize) + xvp->output_to_slot[VERT_RESULT_PSIZ] = numVpOuts++; + if (emitBFC0) + xvp->output_to_slot[VERT_RESULT_BFC0] = numVpOuts++; + if (emitBFC1) + xvp->output_to_slot[VERT_RESULT_BFC1] = numVpOuts++; + + /* Unneeded vertex program outputs will go to this slot. * We could use this info to do dead code elimination in the * vertex program. @@ -224,6 +234,11 @@ find_translated_vp(struct st_context *st, if (xvp->output_to_slot[outAttr] == UNUSED) xvp->output_to_slot[outAttr] = dummySlot; } +#if 0 /*debug*/ + printf("output_to_slot[%d] = %d\n", outAttr, + xvp->output_to_slot[outAttr]); +#endif + } assert(stvp->Base.Base.NumInstructions > 1); -- cgit v1.2.3 From 98b7174ad6f6a74361e58136bdd235a092b07a40 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 19:11:49 -0600 Subject: gallium: replace an assertion with "if (!texobj) continue" It's possible to call update_samplers() between the time a fragment shader is bound and when a texture image is defined (such as glClear). This fixes the case where we don't have a complete texture object yet. --- src/mesa/state_tracker/st_atom_sampler.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 1e182eef1c..3ba6a971f6 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -136,7 +136,8 @@ update_samplers(struct st_context *st) const struct gl_texture_object *texobj = st->ctx->Texture.Unit[texUnit]._Current; - assert(texobj); + if (!texobj) + continue; sampler->wrap_s = gl_wrap_to_sp(texobj->WrapS); sampler->wrap_t = gl_wrap_to_sp(texobj->WrapT); -- cgit v1.2.3 From 1ca23061478868d61b9b2e6a30367e8e1de4a456 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 2 Jul 2008 19:18:10 -0600 Subject: mesa: fix vertex array validation test for attribute 0 (vert pos) We don't actually need vertex array[0] enabled when using a vertex program/shader. cherry-picked from master --- src/mesa/main/api_validate.c | 20 ++++++++++++-------- 1 file changed, 12 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c index 9144b4bc66..2eb96ae3f4 100644 --- a/src/mesa/main/api_validate.c +++ b/src/mesa/main/api_validate.c @@ -59,9 +59,10 @@ _mesa_validate_DrawElements(GLcontext *ctx, if (ctx->NewState) _mesa_update_state(ctx); - /* Always need vertex positions */ - if (!ctx->Array.ArrayObj->Vertex.Enabled - && !(ctx->VertexProgram._Enabled && ctx->Array.ArrayObj->VertexAttrib[0].Enabled)) + /* Always need vertex positions, unless a vertex program is in use */ + if (!ctx->VertexProgram._Current && + !ctx->Array.ArrayObj->Vertex.Enabled && + !ctx->Array.ArrayObj->VertexAttrib[0].Enabled) return GL_FALSE; /* Vertex buffer object tests */ @@ -177,9 +178,10 @@ _mesa_validate_DrawRangeElements(GLcontext *ctx, GLenum mode, if (ctx->NewState) _mesa_update_state(ctx); - /* Always need vertex positions */ - if (!ctx->Array.ArrayObj->Vertex.Enabled - && !(ctx->VertexProgram._Enabled && ctx->Array.ArrayObj->VertexAttrib[0].Enabled)) + /* Always need vertex positions, unless a vertex program is in use */ + if (!ctx->VertexProgram._Current && + !ctx->Array.ArrayObj->Vertex.Enabled && + !ctx->Array.ArrayObj->VertexAttrib[0].Enabled) return GL_FALSE; /* Vertex buffer object tests */ @@ -247,8 +249,10 @@ _mesa_validate_DrawArrays(GLcontext *ctx, if (ctx->NewState) _mesa_update_state(ctx); - /* Always need vertex positions */ - if (!ctx->Array.ArrayObj->Vertex.Enabled && !ctx->Array.ArrayObj->VertexAttrib[0].Enabled) + /* Always need vertex positions, unless a vertex program is in use */ + if (!ctx->VertexProgram._Current && + !ctx->Array.ArrayObj->Vertex.Enabled && + !ctx->Array.ArrayObj->VertexAttrib[0].Enabled) return GL_FALSE; if (ctx->Const.CheckArrayBounds) { -- cgit v1.2.3 From 8ccab313561d8cca3caf8d76504119bf5eb2e9ae Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 3 Jul 2008 15:02:56 +0900 Subject: gallium: Use surface_copy for 1:1 blits. --- src/gallium/auxiliary/util/u_blit.c | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index ae779335dc..c9e3abc88b 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -308,6 +308,16 @@ util_blit_pixels(struct blit_state *ctx, } assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE)); + assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE)); + + if(dst->format == src->format && (dstX1 - dstX0) == srcW && (dstY1 - dstY0) == srcH) { + pipe->surface_copy(pipe, FALSE, + dst, dstX0, dstY0, /* dest */ + src, srcX0, srcY0, /* src */ + srcW, srcH); /* size */ + return; + } + assert(screen->is_format_supported(screen, dst->format, PIPE_SURFACE)); /* -- cgit v1.2.3 From 28f3d390ddf87f796a9a9f3dc28b0b05c19e40fa Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 3 Jul 2008 12:10:36 +0200 Subject: egl: Fix demodriver --- src/egl/drivers/demo/demo.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/demo/demo.c b/src/egl/drivers/demo/demo.c index 6b8b71d16b..1750e976b8 100644 --- a/src/egl/drivers/demo/demo.c +++ b/src/egl/drivers/demo/demo.c @@ -67,22 +67,22 @@ demoInitialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) /* Create the display's visual configs - silly example */ for (i = 0; i < 4; i++) { - _EGLConfig config; - _eglInitConfig(&config, i + 1); - _eglSetConfigAttrib(&config, EGL_RED_SIZE, 8); - _eglSetConfigAttrib(&config, EGL_GREEN_SIZE, 8); - _eglSetConfigAttrib(&config, EGL_BLUE_SIZE, 8); - _eglSetConfigAttrib(&config, EGL_ALPHA_SIZE, 8); - _eglSetConfigAttrib(&config, EGL_BUFFER_SIZE, 32); + _EGLConfig *config = calloc(1, sizeof(_EGLConfig)); + _eglInitConfig(config, i + 1); + _eglSetConfigAttrib(config, EGL_RED_SIZE, 8); + _eglSetConfigAttrib(config, EGL_GREEN_SIZE, 8); + _eglSetConfigAttrib(config, EGL_BLUE_SIZE, 8); + _eglSetConfigAttrib(config, EGL_ALPHA_SIZE, 8); + _eglSetConfigAttrib(config, EGL_BUFFER_SIZE, 32); if (i & 1) { - _eglSetConfigAttrib(&config, EGL_DEPTH_SIZE, 32); + _eglSetConfigAttrib(config, EGL_DEPTH_SIZE, 32); } if (i & 2) { - _eglSetConfigAttrib(&config, EGL_STENCIL_SIZE, 8); + _eglSetConfigAttrib(config, EGL_STENCIL_SIZE, 8); } - _eglSetConfigAttrib(&config, EGL_SURFACE_TYPE, + _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, (EGL_WINDOW_BIT | EGL_PIXMAP_BIT | EGL_PBUFFER_BIT)); - _eglAddConfig(disp, &config); + _eglAddConfig(disp, config); } drv->Initialized = EGL_TRUE; -- cgit v1.2.3 From d94c7063ec9e87dbefea386606f0f21cdbd1d6f3 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 3 Jul 2008 12:12:25 +0200 Subject: i915: EGL almost works again --- src/gallium/winsys/egl_drm/intel/intel_context.c | 3 + src/gallium/winsys/egl_drm/intel/intel_context.h | 3 +- src/gallium/winsys/egl_drm/intel/intel_egl.c | 85 +++++++++++++--------- src/gallium/winsys/egl_drm/intel/intel_screen.c | 4 +- src/gallium/winsys/egl_drm/intel/intel_screen.h | 3 + .../winsys/egl_drm/intel/intel_swapbuffers.c | 9 ++- 6 files changed, 67 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.c b/src/gallium/winsys/egl_drm/intel/intel_context.c index fdbaa230e5..7205fa4483 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_context.c +++ b/src/gallium/winsys/egl_drm/intel/intel_context.c @@ -172,6 +172,9 @@ intel_create_context(struct egl_drm_context *egl_context, const __GLcontextModes pipe->priv = intel; intel->st = st_create_context(pipe, visual, st_share); + + screen->dummy = intel; + return TRUE; } diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.h b/src/gallium/winsys/egl_drm/intel/intel_context.h index ccf8120761..f05a213ff0 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_context.h +++ b/src/gallium/winsys/egl_drm/intel/intel_context.h @@ -64,8 +64,7 @@ struct intel_framebuffer { struct st_framebuffer *stfb; - /* other fields TBD */ - int other; + struct intel_screen *screen; struct _DriBufferObject *front_buffer; struct egl_drm_frontbuffer *front; }; diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 3b4ab330f5..6865bfa3d4 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -74,9 +74,11 @@ struct drm_screen uint32_t fbID; drmModeCrtcPtr crtc; - /* currently only support one output */ - drmModeOutputPtr output; - uint32_t outputID; + /* currently only support one connector */ + drmModeConnectorPtr connector; + drmModeEncoderPtr encoder; + uint32_t connectorID; + uint32_t encoderID; struct drm_mode_modeinfo *mode; @@ -92,26 +94,39 @@ drm_update_res(struct drm_driver *drm_drv) } static void -drm_add_modes_from_output(_EGLScreen *screen, drmModeOutputPtr output) +drm_add_modes_from_connector(_EGLScreen *screen, drmModeConnectorPtr connector) { struct drm_mode_modeinfo *m; int i; - for (i = 0; i < output->count_modes; i++) { - m = &output->modes[i]; + for (i = 0; i < connector->count_modes; i++) { + m = &connector->modes[i]; _eglAddNewMode(screen, m->hdisplay, m->vdisplay, m->vrefresh, m->name); } } +static void +print_modes(drmModeConnectorPtr connector) +{ + struct drm_mode_modeinfo *m; + int i; + + for (i = 0; i < connector->count_modes; i++) { + m = &connector->modes[i]; + printf("dfm %p %i %i %i\n", m, m->hdisplay, m->vdisplay, m->vrefresh); + } +} + static EGLBoolean drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) { + printf("%s enter\n", __FUNCTION__); _EGLDisplay *disp = _eglLookupDisplay(dpy); struct drm_driver *drm_drv = (struct drm_driver *)drv; struct drm_screen *screen = NULL; - drmModeOutputPtr output = NULL; + drmModeConnectorPtr connector = NULL; drmModeResPtr res = NULL; - unsigned count_outputs = 0; + unsigned count_connectors = 0; EGLint i; int fd; @@ -130,40 +145,41 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) drm_update_res(drm_drv); res = drm_drv->res; if (res) - count_outputs = res->count_outputs; + count_connectors = res->count_connectors; - for(i = 0; i < count_outputs; i++) { - output = drmModeGetOutput(fd, res->outputs[i]); + for(i = 0; i < count_connectors; i++) { + connector = drmModeGetConnector(fd, res->connectors[i]); - if (!output) + if (!connector) continue; - if (output->connection == DRM_MODE_DISCONNECTED) { - drmModeFreeOutput(output); + if (connector->connection == DRM_MODE_DISCONNECTED) { + drmModeFreeConnector(connector); continue; } screen = malloc(sizeof(struct drm_screen)); memset(screen, 0, sizeof(*screen)); - screen->outputID = res->outputs[i]; - screen->output = output; + screen->connectorID = res->connectors[i]; + screen->connector = connector; _eglInitScreen(&screen->base); _eglAddScreen(disp, &screen->base); - drm_add_modes_from_output(&screen->base, output); + drm_add_modes_from_connector(&screen->base, connector); } /* for now we only have one config */ - _EGLConfig config; - _eglInitConfig(&config, i + 1); - _eglSetConfigAttrib(&config, EGL_RED_SIZE, 8); - _eglSetConfigAttrib(&config, EGL_GREEN_SIZE, 8); - _eglSetConfigAttrib(&config, EGL_BLUE_SIZE, 8); - _eglSetConfigAttrib(&config, EGL_ALPHA_SIZE, 8); - _eglSetConfigAttrib(&config, EGL_BUFFER_SIZE, 32); - _eglSetConfigAttrib(&config, EGL_DEPTH_SIZE, 24); - _eglSetConfigAttrib(&config, EGL_STENCIL_SIZE, 8); - _eglSetConfigAttrib(&config, EGL_SURFACE_TYPE, EGL_PBUFFER_BIT); - _eglAddConfig(disp, &config); + _EGLConfig *config = calloc(1, sizeof(*config)); + memset(config, 1, sizeof(*config)); + _eglInitConfig(config, 1); + _eglSetConfigAttrib(config, EGL_RED_SIZE, 8); + _eglSetConfigAttrib(config, EGL_GREEN_SIZE, 8); + _eglSetConfigAttrib(config, EGL_BLUE_SIZE, 8); + _eglSetConfigAttrib(config, EGL_ALPHA_SIZE, 8); + _eglSetConfigAttrib(config, EGL_BUFFER_SIZE, 32); + _eglSetConfigAttrib(config, EGL_DEPTH_SIZE, 24); + _eglSetConfigAttrib(config, EGL_STENCIL_SIZE, 8); + _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_PBUFFER_BIT); + _eglAddConfig(disp, config); drv->Initialized = EGL_TRUE; @@ -401,13 +417,13 @@ drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, } static struct drm_mode_modeinfo * -drm_find_mode(drmModeOutputPtr output, _EGLMode *mode) +drm_find_mode(drmModeConnectorPtr connector, _EGLMode *mode) { int i; struct drm_mode_modeinfo *m; - for (i = 0; i < output->count_modes; i++) { - m = &output->modes[i]; + for (i = 0; i < connector->count_modes; i++) { + m = &connector->modes[i]; if (m->hdisplay == mode->Width && m->vdisplay == mode->Height && m->vrefresh == mode->RefreshRate) break; m = NULL; @@ -493,7 +509,7 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, if (!scrn->fb) goto err_bo; - scrn->mode = drm_find_mode(scrn->output, mode); + scrn->mode = drm_find_mode(scrn->connector, mode); if (!scrn->mode) { printf("oh noes, no matching mode found\n"); goto err_fb; @@ -504,7 +520,7 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, drm_drv->res->crtcs[1], scrn->fbID, 0, 0, - &scrn->outputID, 1, + &scrn->connectorID, 1, scrn->mode); @@ -549,7 +565,6 @@ drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface re struct drm_context *ctx = lookup_drm_context(context); EGLBoolean b; - printf("drm_make_current\n"); b = _eglMakeCurrent(drv, dpy, draw, read, context); if (!b) return EGL_FALSE; @@ -559,7 +574,6 @@ drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface re (void) readSurf; (void) ctx; - printf("enter intel_make_current\n"); intel_make_current(ctx->context, drawSurf->drawable, readSurf->drawable); return EGL_TRUE; } @@ -588,6 +602,7 @@ _EGLDriver * _eglMain(_EGLDisplay *dpy, const char *args) { struct drm_driver *drm; + printf("%s enter\n", __FUNCTION__); drm = (struct drm_driver *) calloc(1, sizeof(struct drm_driver)); if (!drm) { diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.c b/src/gallium/winsys/egl_drm/intel/intel_screen.c index 96b0bf1b85..2818b9676c 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_screen.c +++ b/src/gallium/winsys/egl_drm/intel/intel_screen.c @@ -56,7 +56,7 @@ intel_init_driver(struct egl_drm_device *device) /** TODO JB: ugly hack */ intel_screen->deviceID = PCI_CHIP_I945_GM; - intel_be_init_device(&intel_screen->base, device->drmFD); + intel_be_init_device(&intel_screen->base, device->drmFD, PCI_CHIP_I945_GM); intel_screen->pipe = i915_create_screen(&intel_screen->base.base, intel_screen->deviceID); @@ -76,6 +76,8 @@ intel_create_drawable(struct egl_drm_drawable *drawable, if (!intelfb) return GL_FALSE; + intelfb->screen = drawable->device->priv; + if (visual->redBits == 5) colorFormat = PIPE_FORMAT_R5G6B5_UNORM; else diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.h b/src/gallium/winsys/egl_drm/intel/intel_screen.h index 87c4406f54..9b990b48e7 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_screen.h +++ b/src/gallium/winsys/egl_drm/intel/intel_screen.h @@ -34,6 +34,7 @@ struct pipe_screen; struct egl_drm_device; +struct intel_context; struct intel_screen { @@ -42,6 +43,8 @@ struct intel_screen int deviceID; struct egl_drm_device *device; + + struct intel_context *dummy; }; #endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c index 24e55f1568..bfff6f935f 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c @@ -68,18 +68,23 @@ intel_display_surface(struct egl_drm_drawable *draw, //const int srcWidth = surf->width; //const int srcHeight = surf->height; - const int srcPitch = surf->pitch; + + intel = intel_fb->screen->dummy; + if (!intel) { + printf("No dummy context\n"); + return; + } const int dstWidth = intel_fb->front->width; const int dstHeight = intel_fb->front->height; const int dstPitch = intel_fb->front->pitch / 4;//draw->front.cpp; const int cpp = 4;//intel_fb->front->cpp; + const int srcPitch = surf->stride / cpp; int BR13, CMD; //int i; - BR13 = (dstPitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | XY_SRC_COPY_BLT_WRITE_RGB); -- cgit v1.2.3 From c193cc506f74443c805a0df42e3364b39245d265 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 3 Jul 2008 19:52:05 +0900 Subject: gallium: Don't forget to get overllaping blits working again. --- src/gallium/auxiliary/util/u_blit.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index c9e3abc88b..3dc9fdd11e 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -311,6 +311,7 @@ util_blit_pixels(struct blit_state *ctx, assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE)); if(dst->format == src->format && (dstX1 - dstX0) == srcW && (dstY1 - dstY0) == srcH) { + /* FIXME: this will most surely fail for overlapping rectangles */ pipe->surface_copy(pipe, FALSE, dst, dstX0, dstY0, /* dest */ src, srcX0, srcY0, /* src */ -- cgit v1.2.3 From ade03755bcaec2dedb5cd4d13160ba366ee804cd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 3 Jul 2008 21:28:56 +0900 Subject: pipebuffer: Silent warnings. Although rarely hit in normal apps, they are too noisy with test suites. --- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 0ffca298cd..f599bee07e 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -174,7 +174,9 @@ _fenced_buffer_finish(struct fenced_buffer *fenced_buf) struct fenced_buffer_list *fenced_list = fenced_buf->list; struct pipe_winsys *winsys = fenced_list->winsys; +#if 0 debug_warning("waiting for GPU"); +#endif assert(fenced_buf->fence); if(fenced_buf->fence) { @@ -278,11 +280,13 @@ fenced_buffer_map(struct pb_buffer *buf, _fenced_buffer_finish(fenced_buf); } +#if 0 /* Check for CPU write access (read is OK) */ if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) { /* this is legal -- just for debugging */ debug_warning("concurrent CPU writes"); } +#endif map = pb_map(fenced_buf->buffer, flags); if(map) { -- cgit v1.2.3 From 2212c214f1890091c102d4cb3c69ec2df2132322 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 10:12:00 -0600 Subject: gallium: fix surface memleak in bitmap code Found w/ tunnel2.c demo. --- src/mesa/state_tracker/st_cb_bitmap.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index f3bc3b8584..de86832342 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -541,6 +541,9 @@ reset_cache(struct st_context *st) cache->ymin = 1000000; cache->ymax = -1000000; + if (cache->surf) + screen->tex_surface_release(screen, &cache->surf); + assert(!cache->texture); /* allocate a new texture */ @@ -588,6 +591,8 @@ st_flush_bitmap_cache(struct st_context *st) * So unmap and release the texture surface before drawing. */ screen->surface_unmap(screen, cache->surf); + cache->buffer = NULL; + screen->tex_surface_release(screen, &cache->surf); draw_bitmap_quad(st->ctx, -- cgit v1.2.3 From 3c128748579c637ba7c777ba91ff4287a03190f6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 10:42:31 -0600 Subject: gallium: increase TGSI_EXEC_MAX_COND_NESTING, etc --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.h | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h index ea182bc454..c89dfcb167 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h @@ -147,9 +147,9 @@ struct tgsi_exec_labels #define TGSI_EXEC_NUM_ADDRS 1 #define TGSI_EXEC_NUM_IMMEDIATES 256 -#define TGSI_EXEC_MAX_COND_NESTING 10 -#define TGSI_EXEC_MAX_LOOP_NESTING 10 -#define TGSI_EXEC_MAX_CALL_NESTING 10 +#define TGSI_EXEC_MAX_COND_NESTING 20 +#define TGSI_EXEC_MAX_LOOP_NESTING 20 +#define TGSI_EXEC_MAX_CALL_NESTING 20 /** * Run-time virtual machine state for executing TGSI shader. -- cgit v1.2.3 From 00eb309c31692d29b01568b0b45f5066928b9107 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 10:43:14 -0600 Subject: gallium: move, increase ST_MAX_SHADER_TOKENS --- src/mesa/state_tracker/st_program.c | 7 +++++++ src/mesa/state_tracker/st_program.h | 3 --- 2 files changed, 7 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 958096f68e..5966bbadae 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -49,6 +49,9 @@ #include "cso_cache/cso_context.h" +#define ST_MAX_SHADER_TOKENS 4096 + + #define TGSI_DEBUG 0 @@ -296,6 +299,8 @@ st_translate_vertex_program(struct st_context *st, /* tokenized result */ tokens, ST_MAX_SHADER_TOKENS); + assert(num_tokens < ST_MAX_SHADER_TOKENS); + vs.tokens = (struct tgsi_token *) mem_dup(tokens, num_tokens * sizeof(tokens[0])); @@ -467,6 +472,8 @@ st_translate_fragment_program(struct st_context *st, /* tokenized result */ tokens, ST_MAX_SHADER_TOKENS); + assert(num_tokens < ST_MAX_SHADER_TOKENS); + fs.tokens = (struct tgsi_token *) mem_dup(tokens, num_tokens * sizeof(tokens[0])); diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index bf07a50789..086e9391aa 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -39,9 +39,6 @@ #include "pipe/p_shader_tokens.h" -#define ST_MAX_SHADER_TOKENS 1024 - - struct cso_fragment_shader; struct cso_vertex_shader; struct translated_vertex_program; -- cgit v1.2.3 From f042d662e2cec4315ddaae1ee536f593139f703d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 12:56:33 -0600 Subject: gallium: increase TGSI interpreter's number of temp registers to 64 Also, clean up the definitions of the misc/extra temp regs. A few new assertions too. --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 4 ++- src/gallium/auxiliary/tgsi/exec/tgsi_exec.h | 48 +++++++++++++++-------------- 2 files changed, 28 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 9649396c9b..46949661af 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -262,7 +262,7 @@ tgsi_exec_machine_init( uint i; mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); - mach->Addrs = &mach->Temps[TGSI_EXEC_NUM_TEMPS]; + mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR]; /* Setup constants. */ for( i = 0; i < 4; i++ ) { @@ -941,6 +941,7 @@ fetch_src_file_channel( break; case TGSI_FILE_TEMPORARY: + assert(index->i[0] < TGSI_EXEC_NUM_TEMPS); chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; @@ -1127,6 +1128,7 @@ store_dest( break; case TGSI_FILE_TEMPORARY: + assert(reg->DstRegister.Index < TGSI_EXEC_NUM_TEMPS); dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; break; diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h index c89dfcb167..18abdd9ac0 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h @@ -99,53 +99,57 @@ struct tgsi_exec_labels unsigned count; }; + +#define TGSI_EXEC_NUM_TEMPS 64 +#define TGSI_EXEC_NUM_TEMP_EXTRAS 6 +#define TGSI_EXEC_NUM_IMMEDIATES 256 + /* * Locations of various utility registers (_I = Index, _C = Channel) */ -#define TGSI_EXEC_TEMP_00000000_I 32 +#define TGSI_EXEC_TEMP_00000000_I (TGSI_EXEC_NUM_TEMPS + 0) #define TGSI_EXEC_TEMP_00000000_C 0 -#define TGSI_EXEC_TEMP_7FFFFFFF_I 32 +#define TGSI_EXEC_TEMP_7FFFFFFF_I (TGSI_EXEC_NUM_TEMPS + 0) #define TGSI_EXEC_TEMP_7FFFFFFF_C 1 -#define TGSI_EXEC_TEMP_80000000_I 32 +#define TGSI_EXEC_TEMP_80000000_I (TGSI_EXEC_NUM_TEMPS + 0) #define TGSI_EXEC_TEMP_80000000_C 2 -#define TGSI_EXEC_TEMP_FFFFFFFF_I 32 +#define TGSI_EXEC_TEMP_FFFFFFFF_I (TGSI_EXEC_NUM_TEMPS + 0) #define TGSI_EXEC_TEMP_FFFFFFFF_C 3 -#define TGSI_EXEC_TEMP_ONE_I 33 +#define TGSI_EXEC_TEMP_ONE_I (TGSI_EXEC_NUM_TEMPS + 1) #define TGSI_EXEC_TEMP_ONE_C 0 -#define TGSI_EXEC_TEMP_TWO_I 33 +#define TGSI_EXEC_TEMP_TWO_I (TGSI_EXEC_NUM_TEMPS + 1) #define TGSI_EXEC_TEMP_TWO_C 1 -#define TGSI_EXEC_TEMP_128_I 33 +#define TGSI_EXEC_TEMP_128_I (TGSI_EXEC_NUM_TEMPS + 1) #define TGSI_EXEC_TEMP_128_C 2 -#define TGSI_EXEC_TEMP_MINUS_128_I 33 +#define TGSI_EXEC_TEMP_MINUS_128_I (TGSI_EXEC_NUM_TEMPS + 1) #define TGSI_EXEC_TEMP_MINUS_128_C 3 -#define TGSI_EXEC_TEMP_KILMASK_I 34 +#define TGSI_EXEC_TEMP_KILMASK_I (TGSI_EXEC_NUM_TEMPS + 2) #define TGSI_EXEC_TEMP_KILMASK_C 0 -#define TGSI_EXEC_TEMP_OUTPUT_I 34 +#define TGSI_EXEC_TEMP_OUTPUT_I (TGSI_EXEC_NUM_TEMPS + 2) #define TGSI_EXEC_TEMP_OUTPUT_C 1 -#define TGSI_EXEC_TEMP_PRIMITIVE_I 34 +#define TGSI_EXEC_TEMP_PRIMITIVE_I (TGSI_EXEC_NUM_TEMPS + 2) #define TGSI_EXEC_TEMP_PRIMITIVE_C 2 -#define TGSI_EXEC_TEMP_THREE_I 34 +#define TGSI_EXEC_TEMP_THREE_I (TGSI_EXEC_NUM_TEMPS + 2) #define TGSI_EXEC_TEMP_THREE_C 3 -#define TGSI_EXEC_TEMP_HALF_I 35 +#define TGSI_EXEC_TEMP_HALF_I (TGSI_EXEC_NUM_TEMPS + 3) #define TGSI_EXEC_TEMP_HALF_C 0 -#define TGSI_EXEC_TEMP_R0 36 +#define TGSI_EXEC_TEMP_R0 (TGSI_EXEC_NUM_TEMPS + 4) + +#define TGSI_EXEC_TEMP_ADDR (TGSI_EXEC_NUM_TEMPS + 5) -#define TGSI_EXEC_NUM_TEMPS (32 + 5) -#define TGSI_EXEC_NUM_ADDRS 1 -#define TGSI_EXEC_NUM_IMMEDIATES 256 #define TGSI_EXEC_MAX_COND_NESTING 20 #define TGSI_EXEC_MAX_LOOP_NESTING 20 @@ -156,13 +160,11 @@ struct tgsi_exec_labels */ struct tgsi_exec_machine { - /* - * 32 program temporaries - * 4 internal temporaries - * 1 address - * 1 temporary of padding to align to 16 bytes + /* Total = program temporaries + internal temporaries + * + 1 padding to align to 16 bytes */ - struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + TGSI_EXEC_NUM_ADDRS + 1]; + struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + + TGSI_EXEC_NUM_TEMP_EXTRAS + 1]; /* * This will point to _Temps after aligning to 16B boundary. -- cgit v1.2.3 From 78388c0e374349810fe3fb38cd4103a9544df4bf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 14:12:27 -0600 Subject: mesa: added case for fixed pt --- src/mesa/main/varray.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c index 50fe874556..a4ef948b58 100644 --- a/src/mesa/main/varray.c +++ b/src/mesa/main/varray.c @@ -615,6 +615,11 @@ _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type, case GL_DOUBLE: elementSize = size * sizeof(GLdouble); break; +#if FEATURE_fixedpt + case GL_FIXED: + elementSize = size * sizeof(GLfixed); + break; +#endif default: _mesa_error( ctx, GL_INVALID_ENUM, "glVertexAttribPointerARB(type)" ); return; -- cgit v1.2.3 From 51abbdd227224c596ae2232ff3137dc3f346d563 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 14:55:14 -0600 Subject: gallium: added a4r4g4b4_put_tile_rgba() --- src/gallium/auxiliary/util/p_tile.c | 31 +++++++++++++++++++++++++++++++ 1 file changed, 31 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index ab603ff6e4..93abef9879 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -298,6 +298,33 @@ a4r4g4b4_get_tile_rgba(ushort *src, } +static void +a4r4g4b4_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + r >>= 4; + g >>= 4; + b >>= 4; + a >>= 4; + *dst++ = (a << 12) | (r << 16) | (g << 4) | b; + } + p += src_stride; + } +} + + /*** PIPE_FORMAT_R5G6B5_UNORM ***/ static void @@ -774,6 +801,10 @@ pipe_put_tile_rgba(struct pipe_context *pipe, r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_R8G8B8A8_UNORM: + assert(0); + break; + case PIPE_FORMAT_A4R4G4B4_UNORM: + a4r4g4b4_put_tile_rgba((ushort *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_L8_UNORM: /*l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ -- cgit v1.2.3 From d015ffa6ea20f00f23513e63b5c27e5e6d0d3627 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 15:41:45 -0600 Subject: mesa: fix problem freeing framebuffer/renderbuffer objects Basically, set up no-op Delete() methods for the DummyFrame/Renderbuffer obj --- src/mesa/main/context.c | 34 ++++++++++++++++++++++++++++++++++ src/mesa/main/fbobject.c | 25 +++++++++++++++++++++++-- src/mesa/main/fbobject.h | 7 +++++-- 3 files changed, 62 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index be93d844e0..33f6d2c0d0 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -691,6 +691,37 @@ delete_shader_cb(GLuint id, void *data, void *userData) } } +/** + * Callback for deleting a framebuffer object. Called by _mesa_HashDeleteAll() + */ +static void +delete_framebuffer_cb(GLuint id, void *data, void *userData) +{ + struct gl_framebuffer *fb = (struct gl_framebuffer *) data; + /* The fact that the framebuffer is in the hashtable means its refcount + * is one, but we're removing from the hashtable now. So clear refcount. + */ + /*assert(fb->RefCount == 1);*/ + fb->RefCount = 0; + + /* NOTE: Delete should always be defined but there are two reports + * of it being NULL (bugs 13507, 14293). Work-around for now. + */ + if (fb->Delete) + fb->Delete(fb); +} + +/** + * Callback for deleting a renderbuffer object. Called by _mesa_HashDeleteAll() + */ +static void +delete_renderbuffer_cb(GLuint id, void *data, void *userData) +{ + struct gl_renderbuffer *rb = (struct gl_renderbuffer *) data; + rb->RefCount = 0; /* see comment for FBOs above */ + rb->Delete(rb); +} + /** * Deallocate a shared state object and all children structures. @@ -744,7 +775,9 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss ) _mesa_DeleteHashTable(ss->ArrayObjects); #if FEATURE_EXT_framebuffer_object + _mesa_HashDeleteAll(ss->FrameBuffers, delete_framebuffer_cb, ctx); _mesa_DeleteHashTable(ss->FrameBuffers); + _mesa_HashDeleteAll(ss->RenderBuffers, delete_renderbuffer_cb, ctx); _mesa_DeleteHashTable(ss->RenderBuffers); #endif @@ -994,6 +1027,7 @@ init_attrib_groups(GLcontext *ctx) #if FEATURE_evaluators _mesa_init_eval( ctx ); #endif + _mesa_init_fbobjects( ctx ); #if FEATURE_feedback _mesa_init_feedback( ctx ); #else diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 800f6ee9a3..e4ff575e18 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.1 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -66,6 +66,27 @@ static struct gl_renderbuffer DummyRenderbuffer; (TARGET) <= GL_TEXTURE_CUBE_MAP_NEGATIVE_Z) +static void +delete_dummy_renderbuffer(struct gl_renderbuffer *rb) +{ + /* no op */ +} + +static void +delete_dummy_framebuffer(struct gl_framebuffer *fb) +{ + /* no op */ +} + + +void +_mesa_init_fbobjects(GLcontext *ctx) +{ + DummyFramebuffer.Delete = delete_dummy_framebuffer; + DummyRenderbuffer.Delete = delete_dummy_renderbuffer; +} + + /** * Helper routine for getting a gl_renderbuffer. */ diff --git a/src/mesa/main/fbobject.h b/src/mesa/main/fbobject.h index 782ad8cb18..b6154719ab 100644 --- a/src/mesa/main/fbobject.h +++ b/src/mesa/main/fbobject.h @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -27,6 +27,9 @@ #define FBOBJECT_H +extern void +_mesa_init_fbobjects(GLcontext *ctx); + extern struct gl_renderbuffer * _mesa_lookup_renderbuffer(GLcontext *ctx, GLuint id); -- cgit v1.2.3 From bb18a8b2eaaf8eea02c3255ed774751378fc7aa6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 16:02:13 -0600 Subject: mesa: fix some error codes in _mesa_ShaderSourceARB() --- src/mesa/main/shaders.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/main/shaders.c b/src/mesa/main/shaders.c index 7bf8808767..b7b2f791a5 100644 --- a/src/mesa/main/shaders.c +++ b/src/mesa/main/shaders.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.3 + * Version: 7.1 * - * Copyright (C) 2004-2007 Brian Paul All Rights Reserved. + * Copyright (C) 2004-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -382,7 +382,7 @@ _mesa_ShaderSourceARB(GLhandleARB shaderObj, GLsizei count, GLsizei i, totalLength; GLcharARB *source; - if (string == NULL) { + if (!shaderObj || string == NULL) { _mesa_error(ctx, GL_INVALID_VALUE, "glShaderSourceARB"); return; } @@ -400,7 +400,7 @@ _mesa_ShaderSourceARB(GLhandleARB shaderObj, GLsizei count, for (i = 0; i < count; i++) { if (string[i] == NULL) { _mesa_free((GLvoid *) offsets); - _mesa_error(ctx, GL_INVALID_VALUE, "glShaderSourceARB(null string)"); + _mesa_error(ctx, GL_INVALID_OPERATION, "glShaderSourceARB(null string)"); return; } if (length == NULL || length[i] < 0) -- cgit v1.2.3 From d4172263e57fdbd2204fd8669e6b6f460e08ca3b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 16:21:15 -0600 Subject: mesa: fix various error codes --- src/mesa/shader/shader_api.c | 238 ++++++++++++++++++++++++++----------------- 1 file changed, 144 insertions(+), 94 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 26d0b3d476..3e28d92fe1 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -212,6 +212,35 @@ _mesa_lookup_shader_program(GLcontext *ctx, GLuint name) } +/** + * As above, but record an error if program is not found. + */ +static struct gl_shader_program * +_mesa_lookup_shader_program_err(GLcontext *ctx, GLuint name, + const char *caller) +{ + if (!name) { + _mesa_error(ctx, GL_INVALID_VALUE, caller); + return NULL; + } + else { + struct gl_shader_program *shProg = (struct gl_shader_program *) + _mesa_HashLookup(ctx->Shared->ShaderObjects, name); + if (!shProg) { + _mesa_error(ctx, GL_INVALID_VALUE, caller); + return NULL; + } + if (shProg->Type != GL_SHADER_PROGRAM_MESA) { + _mesa_error(ctx, GL_INVALID_OPERATION, caller); + return NULL; + } + return shProg; + } +} + + + + /** * Allocate a new gl_shader object, initialize it. */ @@ -314,6 +343,33 @@ _mesa_lookup_shader(GLcontext *ctx, GLuint name) } +/** + * As above, but record an error if shader is not found. + */ +static struct gl_shader * +_mesa_lookup_shader_err(GLcontext *ctx, GLuint name, const char *caller) +{ + if (!name) { + _mesa_error(ctx, GL_INVALID_VALUE, caller); + return NULL; + } + else { + struct gl_shader *sh = (struct gl_shader *) + _mesa_HashLookup(ctx->Shared->ShaderObjects, name); + if (!sh) { + _mesa_error(ctx, GL_INVALID_VALUE, caller); + return NULL; + } + if (sh->Type == GL_SHADER_PROGRAM_MESA) { + _mesa_error(ctx, GL_INVALID_OPERATION, caller); + return NULL; + } + return sh; + } +} + + + /** * Initialize context's shader state. */ @@ -360,26 +416,42 @@ copy_string(GLchar *dst, GLsizei maxLength, GLsizei *length, const GLchar *src) } +static GLboolean +_mesa_is_program(GLcontext *ctx, GLuint name) +{ + struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, name); + return shProg ? GL_TRUE : GL_FALSE; +} + + +static GLboolean +_mesa_is_shader(GLcontext *ctx, GLuint name) +{ + struct gl_shader *shader = _mesa_lookup_shader(ctx, name); + return shader ? GL_TRUE : GL_FALSE; +} + + /** * Called via ctx->Driver.AttachShader() */ static void _mesa_attach_shader(GLcontext *ctx, GLuint program, GLuint shader) { - struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); - struct gl_shader *sh = _mesa_lookup_shader(ctx, shader); - GLuint n; - GLuint i; + struct gl_shader_program *shProg; + struct gl_shader *sh; + GLuint i, n; - if (!shProg || !sh) { - _mesa_error(ctx, GL_INVALID_VALUE, - "glAttachShader(bad program or shader name)"); + shProg = _mesa_lookup_shader_program_err(ctx, program, "glAttachShader"); + if (!shProg) + return; + + sh = _mesa_lookup_shader_err(ctx, shader, "glAttachShader"); + if (!sh) { return; } n = shProg->NumShaders; - for (i = 0; i < n; i++) { if (shProg->Shaders[i] == sh) { /* already attached */ @@ -409,10 +481,9 @@ _mesa_get_attrib_location(GLcontext *ctx, GLuint program, const GLchar *name) { struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); + = _mesa_lookup_shader_program_err(ctx, program, "glGetAttribLocation"); if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, "glGetAttribLocation"); return -1; } @@ -439,13 +510,13 @@ static void _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, const GLchar *name) { - struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); + struct gl_shader_program *shProg; const GLint size = -1; /* unknown size */ GLint i, oldIndex; + shProg = _mesa_lookup_shader_program_err(ctx, program, + "glBindAttribLocation"); if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, "glBindAttribLocation(program)"); return; } @@ -458,6 +529,11 @@ _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, return; } + if (index >= ctx->Const.VertexProgram.MaxAttribs) { + _mesa_error(ctx, GL_INVALID_VALUE, "glBindAttribLocation(index)"); + return; + } + if (shProg->LinkStatus) { /* get current index/location for the attribute */ oldIndex = _mesa_get_attrib_location(ctx, program, name); @@ -539,11 +615,9 @@ _mesa_delete_program2(GLcontext *ctx, GLuint name) */ struct gl_shader_program *shProg; - shProg = _mesa_lookup_shader_program(ctx, name); - if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, "glDeleteProgram(name)"); + shProg = _mesa_lookup_shader_program_err(ctx, name, "glDeleteProgram"); + if (!shProg) return; - } shProg->DeletePending = GL_TRUE; @@ -555,10 +629,11 @@ _mesa_delete_program2(GLcontext *ctx, GLuint name) static void _mesa_delete_shader(GLcontext *ctx, GLuint shader) { - struct gl_shader *sh = _mesa_lookup_shader(ctx, shader); - if (!sh) { + struct gl_shader *sh; + + sh = _mesa_lookup_shader_err(ctx, shader, "glDeleteShader"); + if (!sh) return; - } sh->DeletePending = GL_TRUE; @@ -570,16 +645,13 @@ _mesa_delete_shader(GLcontext *ctx, GLuint shader) static void _mesa_detach_shader(GLcontext *ctx, GLuint program, GLuint shader) { - struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); + struct gl_shader_program *shProg; GLuint n; GLuint i, j; - if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, - "glDetachShader(bad program or shader name)"); + shProg = _mesa_lookup_shader_program_err(ctx, program, "glDetachShader"); + if (!shProg) return; - } n = shProg->NumShaders; @@ -588,7 +660,7 @@ _mesa_detach_shader(GLcontext *ctx, GLuint program, GLuint shader) /* found it */ struct gl_shader **newList; - /* derefernce */ + /* release */ _mesa_reference_shader(ctx, &shProg->Shaders[i], NULL); /* alloc new, smaller array */ @@ -624,8 +696,17 @@ _mesa_detach_shader(GLcontext *ctx, GLuint program, GLuint shader) } /* not found */ - _mesa_error(ctx, GL_INVALID_VALUE, - "glDetachShader(shader not found)"); + { + GLenum err; + if (_mesa_is_shader(ctx, shader)) + err = GL_INVALID_OPERATION; + else if (_mesa_is_program(ctx, shader)) + err = GL_INVALID_OPERATION; + else + err = GL_INVALID_VALUE; + _mesa_error(ctx, err, "glDetachProgram(shader)"); + return; + } } @@ -637,14 +718,12 @@ _mesa_get_active_attrib(GLcontext *ctx, GLuint program, GLuint index, static const GLenum vec_types[] = { GL_FLOAT, GL_FLOAT_VEC2, GL_FLOAT_VEC3, GL_FLOAT_VEC4 }; - struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); + struct gl_shader_program *shProg; GLint sz; - if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, "glGetActiveAttrib"); + shProg = _mesa_lookup_shader_program_err(ctx, program, "glGetActiveAttrib"); + if (!shProg) return; - } if (!shProg->Attributes || index >= shProg->Attributes->NumParameters) { _mesa_error(ctx, GL_INVALID_VALUE, "glGetActiveAttrib(index)"); @@ -669,15 +748,13 @@ _mesa_get_active_uniform(GLcontext *ctx, GLuint program, GLuint index, GLsizei maxLength, GLsizei *length, GLint *size, GLenum *type, GLchar *nameOut) { - const struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); + const struct gl_shader_program *shProg; const struct gl_program *prog; GLint progPos; - if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, "glGetActiveUniform"); + shProg = _mesa_lookup_shader_program_err(ctx, program, "glGetActiveUniform"); + if (!shProg) return; - } if (!shProg->Uniforms || index >= shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glGetActiveUniform(index)"); @@ -716,19 +793,16 @@ static void _mesa_get_attached_shaders(GLcontext *ctx, GLuint program, GLsizei maxCount, GLsizei *count, GLuint *obj) { - struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); + struct gl_shader_program *shProg = + _mesa_lookup_shader_program_err(ctx, program, "glGetAttachedShaders"); if (shProg) { GLuint i; - for (i = 0; i < maxCount && i < shProg->NumShaders; i++) { + for (i = 0; i < (GLuint) maxCount && i < shProg->NumShaders; i++) { obj[i] = shProg->Shaders[i]->Name; } if (count) *count = i; } - else { - _mesa_error(ctx, GL_INVALID_VALUE, "glGetAttachedShaders"); - } } @@ -809,10 +883,9 @@ _mesa_get_programiv(GLcontext *ctx, GLuint program, static void _mesa_get_shaderiv(GLcontext *ctx, GLuint name, GLenum pname, GLint *params) { - struct gl_shader *shader = _mesa_lookup_shader(ctx, name); + struct gl_shader *shader = _mesa_lookup_shader_err(ctx, name, "glGetShaderiv"); if (!shader) { - _mesa_error(ctx, GL_INVALID_VALUE, "glGetShaderiv(shader)"); return; } @@ -873,14 +946,9 @@ static void _mesa_get_shader_source(GLcontext *ctx, GLuint shader, GLsizei maxLength, GLsizei *length, GLchar *sourceOut) { - struct gl_shader *sh = _mesa_lookup_shader(ctx, shader); + struct gl_shader *sh; + sh = _mesa_lookup_shader_err(ctx, shader, "glGetShaderSource"); if (!sh) { - GLenum err; - if (_mesa_lookup_shader_program(ctx, shader)) - err = GL_INVALID_OPERATION; - else - err = GL_INVALID_VALUE; - _mesa_error(ctx, err, "glGetShaderSource(shader)"); return; } copy_string(sourceOut, maxLength, length, sh->Source); @@ -937,28 +1005,17 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, static GLint _mesa_get_uniform_location(GLcontext *ctx, GLuint program, const GLchar *name) { - struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); + struct gl_shader_program *shProg = + _mesa_lookup_shader_program_err(ctx, program, "glGetUniformLocation"); + if (!shProg) return -1; - return _mesa_lookup_uniform(shProg->Uniforms, name); -} - - -static GLboolean -_mesa_is_program(GLcontext *ctx, GLuint name) -{ - struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, name); - return shProg ? GL_TRUE : GL_FALSE; -} - + /* XXX we should return -1 if the uniform was declared, but not + * actually used. + */ -static GLboolean -_mesa_is_shader(GLcontext *ctx, GLuint name) -{ - struct gl_shader *shader = _mesa_lookup_shader(ctx, name); - return shader ? GL_TRUE : GL_FALSE; + return _mesa_lookup_uniform(shProg->Uniforms, name); } @@ -969,16 +1026,11 @@ _mesa_is_shader(GLcontext *ctx, GLuint name) static void _mesa_shader_source(GLcontext *ctx, GLuint shader, const GLchar *source) { - struct gl_shader *sh = _mesa_lookup_shader(ctx, shader); - if (!sh) { - GLenum err; - if (_mesa_lookup_shader_program(ctx, shader)) - err = GL_INVALID_OPERATION; - else - err = GL_INVALID_VALUE; - _mesa_error(ctx, err, "glShaderSource(shaderObj)"); + struct gl_shader *sh; + + sh = _mesa_lookup_shader_err(ctx, shader, "glShaderSource"); + if (!sh) return; - } /* free old shader source string and install new one */ if (sh->Source) { @@ -995,12 +1047,11 @@ _mesa_shader_source(GLcontext *ctx, GLuint shader, const GLchar *source) static void _mesa_compile_shader(GLcontext *ctx, GLuint shaderObj) { - struct gl_shader *sh = _mesa_lookup_shader(ctx, shaderObj); + struct gl_shader *sh; - if (!sh) { - _mesa_error(ctx, GL_INVALID_VALUE, "glCompileShader(shaderObj)"); + sh = _mesa_lookup_shader_err(ctx, shaderObj, "glCompileShader"); + if (!sh) return; - } sh->CompileStatus = _slang_compile(ctx, sh); } @@ -1014,11 +1065,9 @@ _mesa_link_program(GLcontext *ctx, GLuint program) { struct gl_shader_program *shProg; - shProg = _mesa_lookup_shader_program(ctx, program); - if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, "glLinkProgram(program)"); + shProg = _mesa_lookup_shader_program_err(ctx, program, "glLinkProgram"); + if (!shProg) return; - } FLUSH_VERTICES(ctx, _NEW_PROGRAM); @@ -1043,10 +1092,12 @@ _mesa_use_program(GLcontext *ctx, GLuint program) FLUSH_VERTICES(ctx, _NEW_PROGRAM); if (program) { - shProg = _mesa_lookup_shader_program(ctx, program); + shProg = _mesa_lookup_shader_program_err(ctx, program, "glUseProgram"); if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, - "glUseProgramObjectARB(programObj)"); + return; + } + if (!shProg->LinkStatus) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glUseProgram"); return; } } @@ -1170,7 +1221,6 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, if (location == -1) return; /* The standard specifies this as a no-op */ - if (location < 0 || location >= (GLint) shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniform(location)"); return; -- cgit v1.2.3 From 4d1d6ed1590428172816933efce3303fd7c815b2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 16:22:36 -0600 Subject: mesa: additional vec4 constructor --- src/mesa/shader/slang/library/slang_core.gc | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 78e3ea9fda..15a0792f43 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -249,6 +249,14 @@ vec4 __constructor(const vec3 v3, const float f) __retVal.w = f; } +vec4 __constructor(const vec2 v2, const float f1, const float f2) +{ + // XXX this constructor shouldn't be needed anymore + __retVal.xy = v2; + __retVal.z = f1; + __retVal.w = f2; +} + //// ivec2 constructors -- cgit v1.2.3 From 22ac107971151a02cbe7a03a05bc99a53bc80eb1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 16:22:49 -0600 Subject: mesa: regenerated file --- src/mesa/shader/slang/library/slang_core_gc.h | 1524 +++++++++++++------------ 1 file changed, 763 insertions(+), 761 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index b8e6d1a990..5213b693b3 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -35,785 +35,787 @@ 120,120,120,0,20,0,0,1,0,12,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,59,120, 120,120,120,0,20,0,0,1,0,12,1,1,1,0,11,118,51,0,0,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97, 108,0,59,120,121,122,0,18,118,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,0,20,0, -0,1,0,6,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,18,105,0,59,120,120,0,20,0,0,1,0,6,1,1,1,0,9,102,0,0,0, -1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, -18,102,0,59,120,120,0,0,0,0,1,0,6,1,1,1,0,1,98,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,7,1,1,1,0,5, -105,0,0,1,1,0,5,106,0,0,1,1,0,5,107,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,122,0,18,107,0,20,0,0,1,0,7,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,18,105,0,59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,18,102,0,59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,8,1,1,1,0,5,120,0,0,1,1,0, -5,121,0,0,1,1,0,5,122,0,0,1,1,0,5,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,8,1,1,1,0,5, -105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,8,1,1,1,0, -9,102,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0, -18,102,0,59,120,120,120,120,0,0,0,0,1,0,8,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -18,98,0,59,120,120,120,120,0,20,0,0,1,0,2,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50, -0,20,0,0,1,0,2,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,59,120, -120,0,20,0,0,1,0,2,1,1,1,0,9,102,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0, -48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,0,0,18,102,0,59,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,3,2, -1,6,1,122,101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95, -115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,59,120,120,0,0,18,122,101, -114,111,0,0,0,0,1,0,2,1,1,1,0,10,118,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17, -48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,3,2,1,6,1, -122,101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115, -101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1, -0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,3,2,1, -11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0, -0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,59, -120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,5,105,0,0,0,1,3,2,1,7,1,122,101,114,111,0, -2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,59,120,120,120,0,0,18,122,101,114, -111,0,0,0,0,1,0,3,1,1,1,0,11,118,0,0,0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0, -48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,7,118,0,0,0, -1,3,2,1,7,1,122,101,114,111,0,2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4, -118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18, -122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,1,1,0,1,98, -52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,4,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,119,0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,4,1,1,1,0,9,102,0, -0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0, -48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0, -0,18,102,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5,105,0,0,0,1,3,2,1,8,1, -122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4, -118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,59,120,120,120,120,0,0, -18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,12,118,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101, -99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52, -95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1, -1,1,0,8,118,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0, -16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0, -18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9, -109,48,49,0,0,1,1,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0, -18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,0,13,1,1,1,0,9,102,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0, -17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,0,1,0, -13,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,13, -1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,13,1,1,1, -0,10,99,48,0,0,1,1,0,10,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,14,1,1,1,0,9,109,48,48, -0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0, -9,109,50,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, -0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0, -18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0, -0,1,0,14,1,1,1,0,9,102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0, -0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,118,0,59,121,121,120,0,20,0,0,1,0,14,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,51,0, -58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,14,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,51,0,58, -102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,14,1,1,1,0,11,99,48,0,0,1,1,0,11,99,49,0,0,1,1,0,11, -99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -99,50,0,20,0,0,1,0,15,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9, -109,51,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,51,49, -0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,1,1,0,9,109,51,50,0,0,1,1,0, -9,109,48,51,0,0,1,1,0,9,109,49,51,0,0,1,1,0,9,109,50,51,0,0,1,1,0,9,109,51,51,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, -0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,109,51,48,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0, -18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0, -18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,109,49,51,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0,0,1,0,15,1,1,1,0,9,102,0,0,0,1,3,2,0,10, -1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59, -121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0,59,121,121,121, -120,0,20,0,0,1,0,15,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,105,0,0, -0,0,0,0,0,1,0,15,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0, -0,0,1,0,15,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,1,1,0,12,99,51,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,5,2,26,1,1,0,5,97,0,0,1,1,0,5,98, -0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102, -108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5, -2,27,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114, -97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0, -9,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0, -0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, -0,1,0,5,2,22,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105, -110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,26,1,1,0,6,97,0,0,1,1,0,6,98,0, -0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102, -108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6, -2,27,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114, -97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,21,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0, -10,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0, -0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, -0,1,0,6,2,22,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116, -111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,26,1,1,0,7,97,0,0,1, -1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0, -0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0, -0,0,1,0,7,2,27,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,115,117, -98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110, -116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,21,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0, -1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0, +0,1,0,12,1,1,1,0,10,118,50,0,0,1,1,0,9,102,49,0,0,1,1,0,9,102,50,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,18,118,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,102,49,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,50,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,1,1,0,5, +106,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,121,0,18,106,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,18,105,0,59,120,120,0,20,0,0,1,0,6,1,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,116,111, +95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,59,120,120,0,0,0,0,1,0,6, +1,1,1,0,1,98,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,7,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,1,1,0,5, +107,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,121,0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,107,0,20,0,0,1,0,7,1,1, +1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,105,0,59,120,120,120,0,20, +0,0,1,0,7,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,102,0,59, +120,120,120,0,20,0,0,1,0,7,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122, +0,18,98,0,59,120,120,120,0,20,0,0,1,0,8,1,1,1,0,5,120,0,0,1,1,0,5,121,0,0,1,1,0,5,122,0,0,1,1,0,5, +119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,8,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,8,1,1,1,0,9,102,0,0,0,1,4,102,108,111,97, +116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,59,120,120,120,120,0,0, +0,0,1,0,8,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,59,120,120,120,120,0,20, +0,0,1,0,2,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18, +98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,0,2,1,1,1,0,1,98,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,59,120,120,0,20,0,0,1,0,2,1,1,1,0,9, +102,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0, +0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,59, +120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,3,2,1,6,1,122,101,114,111,0,2, +58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,0,0,18,105,0,59,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1, +0,10,118,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0,48,0,0,0,17,48,0,48,0,0, +0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0, +0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,3,2,1,6,1,122,101,114,111,0,2,58,105,118, +101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98, +50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18, +98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18, +98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101, +99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,59,120,120,120,0,0,18,122,101,114,111, +0,0,0,0,1,0,3,1,1,1,0,5,105,0,0,0,1,3,2,1,7,1,122,101,114,111,0,2,58,105,118,101,99,51,0,16,8,48,0, +0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0, +59,120,121,122,0,0,18,105,0,59,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,11,118,0,0, +0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48, +0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, +18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,7,118,0,0,0,1,3,2,1,7,1,122,101,114,111,0,2,58, +105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1, +0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,1,1,0,1,98,52,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0, +18,98,52,0,20,0,0,1,0,4,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119, +0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,4,1,1,1,0,9,102,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2, +58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118, +101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,59,120,120,120,120,0,0,18, +122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5,105,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101, +99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18, +95,95,114,101,116,86,97,108,0,0,18,105,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1, +1,1,0,12,118,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48, +0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101, +116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,8,118,0,0,0,1,3,2,1,8,1,122, +101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118, +101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0, +0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121, +0,18,109,49,49,0,20,0,0,1,0,13,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,59,120,0,18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,17,48,0,48,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,17,48,0,48,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,0,1,0,13,1,1,1,0,5,105,0,0,0,1,8,58,109, +97,116,50,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,13,1,1,1,0,1,98,0,0,0,1,8,58,109,97, +116,50,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,13,1,1,1,0,10,99,48,0,0,1,1,0,10,99,49,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,14,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9, +109,50,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,48,50, +0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49, +48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, +59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48, +50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,0,1,0,14,1,1,1,0,9,102,0,0,0,1, +3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, +57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59,121, +121,120,0,20,0,0,1,0,14,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,51,0,58,102,108,111,97,116,0,18,105, +0,0,0,0,0,0,0,1,0,14,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,51,0,58,102,108,111,97,116,0,18,98,0,0,0, +0,0,0,0,1,0,14,1,1,1,0,11,99,48,0,0,1,1,0,11,99,49,0,0,1,1,0,11,99,50,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,0,15,1,1,1,0,9, +109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,51,48,0,0,1,1,0,9,109,48,49, +0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,51,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0, +9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,1,1,0,9,109,51,50,0,0,1,1,0,9,109,48,51,0,0,1,1,0,9,109,49, +51,0,0,1,1,0,9,109,50,51,0,0,1,1,0,9,109,51,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109, +49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,109,51,48,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, +59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50, +49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0,18,109,51,49,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,119,0,18,109,51, +50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0,18,109,48,51,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,109,49,51,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57, +59,119,0,18,109,51,51,0,20,0,0,1,0,15,1,1,1,0,9,102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0, +18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120, +121,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,121,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59,121,121,120,121,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0,59,121,121,121,120,0,20,0,0,1,0,15,1,1,1,0,5,105, +0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,15,1,1,1,0,1,98,0,0, +0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,15,1,1,1,0,12,99,48,0,0,1, +1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,1,1,0,12,99,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +51,0,57,18,99,51,0,20,0,0,1,0,5,2,26,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118, +101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105, +110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5,2,27,1,1,0,5,97,0,0,1,1,0,5,98,0, +0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0, 18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,0,0,1,0,7,2,22,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,98,73,110,118,0,0,1,1,120,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116, -111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,26,1,1,0,8,97,0,0,1, -1,0,8,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0, -0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0, -0,0,1,0,8,2,27,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,115,117, -98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110, -116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,21,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0, -1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0, +120,0,0,0,0,1,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, +95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5,2,22,1,1,0,5,97,0,0,1,1,0, +5,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98, +73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0, +0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86, +97,108,0,0,18,120,0,0,0,0,1,0,6,2,26,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118, +101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105, +110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,27,1,1,0,6,97,0,0,1,1,0,6,98,0, +0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0, 18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,0,0,1,0,8,2,22,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,98,73,110,118,0,0,1,1,120,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,98,73,110,118,0,59,119,0,0,18,98,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105, -110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,2,26,1,1,0,9,97,0,0,1,1,0,9,98,0, -0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98, -0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,21,1,1,0,9,97,0, -0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86, -97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,22,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1, -98,73,110,118,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59, -120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0, -0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,21,1,1,0, -10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1,0,10, -117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0, -59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118, +120,0,0,0,0,1,0,6,2,21,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, +95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,22,1,1,0,6,97,0,0,1,1,0, +6,98,0,0,0,1,3,2,0,10,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98, +73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118, +0,59,121,0,0,18,98,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0, +18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, +116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,26,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0, +4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, +95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,27,1,1,0,7,97,0,0,1,1,0, +7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18, +97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108, +0,0,18,120,0,0,0,0,1,0,7,2,21,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99, +52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95, +116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,22,1,1,0,7,97,0, +0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112, +0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73, +110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0, +59,122,0,0,18,98,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18, +97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, +116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,26,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,11,1,120,0,0,0, +4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, +95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,27,1,1,0,8,97,0,0,1,1,0, +8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18, +97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108, +0,0,18,120,0,0,0,0,1,0,8,2,21,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118,101,99, +52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95, +116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,22,1,1,0,8,97,0, +0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112, +0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73, +110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0, +59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,119,0,0, +18,98,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18, +98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97, +108,0,0,18,120,0,0,0,0,1,0,9,2,26,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100, +0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,1, +1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108, +0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,21,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52, +95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0, +0,0,0,1,0,9,2,22,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,73,110,118,0, +0,0,0,1,0,10,2,26,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1, +0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108, +0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118, 101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -118,0,0,18,119,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100, -100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,27,1, -1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1, -0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0, -1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0, -0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,119,0,0,0, -0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114, -101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1, -4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, -117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,22,1,1,0, -12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0, -59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0, -59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0,1,0,10, -2,26,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,26,1,1,0,10, +118,0,0,18,117,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,119,0,0,0,0,1,0,11,2, +26,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,1,1,0,11,117,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, +118,0,0,18,117,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,119,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,119,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0, +0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0, +0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0, +12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, +101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1, +3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0,1,0,10,2,26,1,1,0,9,97,0,0,1,1,0,10,117, +0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59, +120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101, +99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18, +98,0,59,120,120,0,0,0,0,1,0,10,2,27,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117, +98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18, +117,0,59,120,121,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0, +18,98,0,59,120,120,0,0,0,0,1,0,10,2,21,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0, +0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120, +121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,22,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,105, +110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59, +120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0, +0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +0,0,18,97,0,59,120,120,0,0,18,105,110,118,85,0,59,120,121,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1, +0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, +66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,66,0,59,120,120,0,0,0,0,1,0,11,2,26,1, +1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0, +59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,26,1,1,0,11, 118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,27,1,1,0,9,97,0,0,1,1,0,10,117, -0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,9,98, -0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,21,1,1,0,9,97,0,0,1,1,0,10,117, -0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0, -9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,22,1,1,0,9,97,0,0,1,1,0, -10,117,0,0,0,1,3,2,0,10,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, -85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59, -121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,105,110,118,85,0,59,120,121,0,0,0,0,1, -0,10,2,22,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,66,0, -59,120,120,0,0,0,0,1,0,11,2,26,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120,121, -122,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0, -0,0,0,1,0,11,2,27,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59, -120,121,122,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98, -116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0, -0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,21,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0, -59,120,120,120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1, -4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,22,1,1,0,9,97,0,0,1,1, -0,11,117,0,0,0,1,3,2,0,11,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, +121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,27,1,1,0,9,97,0,0, +1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,27,1, +1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0, +1,0,11,2,21,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120, +121,122,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0, +18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,22,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,105,110, +118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, +97,0,59,120,120,120,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1, +1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, +118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,66,0,59,120,120,120,0,0, +0,0,1,0,12,2,26,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114, +101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1, +1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, +98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,27,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95, +115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0, +18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116, +114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1, +0,12,2,21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, +18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0, +12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, +101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,22,1,1,0,9,97,0,0,1,1, +0,12,117,0,0,0,1,3,2,0,12,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, 118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0, 59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0, -0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0, -0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105, -110,118,66,0,59,120,120,120,0,0,0,0,1,0,12,2,26,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99, -52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0, -0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101, -116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,27,1,1,0,9,97,0,0,1,1,0, -12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0, -0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4, -118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, -98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0, -0,18,117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0, -0,0,1,0,12,2,22,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,105,110,118,85,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,105,110,118,85,0,0, -0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,59,120,120,120,120, -0,0,0,0,1,0,6,2,26,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105, -118,101,99,50,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,6,2,27,1,1, -0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0, -0,18,117,0,47,20,0,0,1,0,6,2,27,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,47,20,0,0,1,0,6,2,21,1,1,0,5,97,0,0,1,1,0,6,117, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,48,20,0,0,1, -0,6,2,21,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118, -101,99,50,0,18,98,0,0,0,48,20,0,0,1,0,6,2,22,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,6,2,22,1,1,0,6,118,0,0,1, -1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,49, -20,0,0,1,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, -101,99,51,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,46,20,0,0,1,0,7,2,27,1,1,0, -5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0, -18,117,0,47,20,0,0,1,0,7,2,27,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,47,20,0,0,1,0,7,2,21,1,1,0,5,97,0,0,1,1,0,7,117,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,7, -2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, -99,51,0,18,98,0,0,0,48,20,0,0,1,0,7,2,22,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,7,2,22,1,1,0,7,118,0,0,1,1,0, -5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,49,20, -0,0,1,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, -101,99,52,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,46,20,0,0,1,0,8,2,27,1,1,0, -5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0, -18,117,0,47,20,0,0,1,0,8,2,27,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,47,20,0,0,1,0,8,2,21,1,1,0,5,97,0,0,1,1,0,8,117,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,8, -2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, -99,52,0,18,98,0,0,0,48,20,0,0,1,0,8,2,22,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,8,2,22,1,1,0,8,118,0,0,1,1,0, -5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,49,20, -0,0,1,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,6,2,27,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,110,101,103, -97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,7,2,27,1,1,0,7,118,0,0,0,1,4, -118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,8, -2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97, -108,0,0,18,118,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,0,1,4,118, -101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59, -120,121,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,0,0,1,0,12,2,27,1,1,0,12, -118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118, -0,0,0,0,1,0,13,2,27,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20, -0,0,1,0,14,2,27,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8, -48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,0,1,0,15,2,27,1,1,0, -15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,0,9,0,100,111,116,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,48,20,0,0,1,0,9,0,100,111,116,0,1,1,0,10,97,0,0,1, -1,0,10,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0, -59,121,0,18,98,0,59,121,0,48,46,20,0,0,1,0,9,0,100,111,116,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4, -118,101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0, -100,111,116,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,18,95,95,114, -101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,1,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,5,97,0,0,1,1,0,5,98,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2, -3,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0, -18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,105,110,118,66,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0, -1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2, -1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0, -18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,6, -117,0,0,0,1,3,2,0,6,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105, -110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0, -59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0, -18,118,0,0,18,105,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, -116,86,97,108,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95, -97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4, +0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,119,0,0,18,117, +0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, +0,0,18,97,0,59,120,120,120,120,0,0,18,105,110,118,85,0,0,0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9, +98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0, +0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, +0,0,18,118,0,0,18,105,110,118,66,0,59,120,120,120,120,0,0,0,0,1,0,6,2,26,1,1,0,5,97,0,0,1,1,0,6, +117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,46,20,0, +0,1,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105, +118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,6,2,27,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,6,2,27,1,1,0,6,118,0, +0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0, +0,47,20,0,0,1,0,6,2,21,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, +105,118,101,99,50,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,6,2,21,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,48,20,0,0,1,0,6,2,22, +1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97, +0,0,0,18,117,0,49,20,0,0,1,0,6,2,22,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,49,20,0,0,1,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7, +117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,46,20,0, +0,1,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105, +118,101,99,51,0,18,98,0,0,0,46,20,0,0,1,0,7,2,27,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,7,2,27,1,1,0,7,118,0, +0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0, +0,47,20,0,0,1,0,7,2,21,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, +105,118,101,99,51,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,7,2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,48,20,0,0,1,0,7,2,22, +1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97, +0,0,0,18,117,0,49,20,0,0,1,0,7,2,22,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,49,20,0,0,1,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8, +117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,46,20,0, +0,1,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105, +118,101,99,52,0,18,98,0,0,0,46,20,0,0,1,0,8,2,27,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,8,2,27,1,1,0,8,118,0, +0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0, +0,47,20,0,0,1,0,8,2,21,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, +105,118,101,99,52,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,8,2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,48,20,0,0,1,0,8,2,22, +1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97, +0,0,0,18,117,0,49,20,0,0,1,0,8,2,22,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,49,20,0,0,1,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118, +101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1, +0,6,2,27,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86, +97,108,0,0,18,118,0,0,0,0,1,0,7,2,27,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101, +0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,8,2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52, +95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,9,2,27,1,1,0,9, +97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, +18,97,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0, +0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,118,0,59,120,121,122,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,110,101,103, +97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,13,2,27,1,1,0,13,109,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, +0,57,18,109,0,16,10,50,0,57,54,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, +57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10, +50,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,54,20,0,0, +1,0,9,0,100,111,116,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0, +18,98,0,48,20,0,0,1,0,9,0,100,111,116,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0,59,121,0,18,98,0,59,121,0,48,46,20,0,0,1, +0,9,0,100,111,116,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,51,95,100,111,116,0,18,95,95, +114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0,100,111,116,0,1,1,0,12,97,0,0,1,1,0,12, +98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0, +0,0,1,0,0,2,1,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,97,0,0,18,97,0, +0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114, +97,99,116,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5, +97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18, +97,0,0,18,105,110,118,66,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95, +97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4, 118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0, -2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18, -118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,3,2,0,7,1,105,110,118,0,0,1,1, +2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18, +118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,3,2,0,6,1,105,110,118,0,0,1,1, 122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4, 102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99, 52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,102,108,111, 97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0, -2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0, -0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, -18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0, -0,1,1,0,8,117,0,0,0,1,3,2,0,8,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0, +0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, +18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52, +95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0, +0,1,1,0,7,117,0,0,0,1,3,2,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0, 18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, 118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122, 0,0,18,118,0,0,18,105,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114, -101,116,86,97,108,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52, -95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,9,97,0,0,1,1,0, -9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,59,120,0,0,18,97,0,0,18,98, -0,0,0,0,1,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0, -9,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2, -10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0, -59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118, -101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117, -0,59,120,121,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0, -0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119, -0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, -59,120,121,0,0,18,118,0,59,120,121,0,0,18,119,0,59,120,121,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1, -0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0, -0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, -116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,11, -117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18, -118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,119,0,0,0,4,102, +101,116,86,97,108,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52, +95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1, +4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1, +0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0, +18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,3,2,0,8,1,105,110,118,0,0, +1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,102,108, +111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,122,0,0,0,0,1,0,0,2,1, +1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59, +120,0,0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98, +0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0, +0,0,0,1,0,0,2,4,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,119,0,0,0,4,102,108,111,97,116,95,114, +99,112,0,18,119,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +97,0,59,120,0,0,18,97,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101, +99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0, +1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, +18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,3,1,0,2,10,118, +0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121, +0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117,0, +0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59, +120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,119, +0,59,120,121,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100, +0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,11,117,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18, +117,0,0,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1, +0,11,117,0,0,0,1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18, +117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0, +4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,119,0, +59,120,121,122,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100, +100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118, +101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2, +12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18, +118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102, 108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95, 114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119, -0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, -59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,119,0,59,120,121,122,0,0,0,0,1,0,0,2,1,1,0,2,12, -118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0, -1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, -18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99, -52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,12, -118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59, -120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59, -121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,5, -97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0, -59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116, -114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2, -3,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, -59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,9,18,118,0,59,120,0,18,97,0,23,0,9, -18,118,0,59,121,0,18,97,0,23,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0, -18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,59,120, -120,120,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114, -97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0, -1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, -18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1, -0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0, -9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95, -97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1, -1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0, -59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2, -8,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9, -18,118,0,59,122,0,18,97,0,24,0,9,18,118,0,59,119,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0, -9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120, -0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, -116,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1, -0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118, -0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118, -65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95, +0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117, +0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0, +0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120, +121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,5,97,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120, +121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95, 109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120, -120,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0, -59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,9,97, -0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0, -18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0, -0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,0, -0,1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118, -0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99, -52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0, -0,2,3,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0, -1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97, -0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120, -120,120,120,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0, -13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0, -48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0, -57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10, -49,0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27, -1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59, -120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16, -10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122, -122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110, -0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0, -48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1, -0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, -16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0, -57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10, -51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2, -21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48, -0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0, -48,46,18,109,0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120, -0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, -57,18,110,0,16,10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57, -59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48, -0,57,18,110,0,16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57, -59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48, -46,18,109,0,16,10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120, -0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, -57,18,110,0,16,10,51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, -59,119,119,119,119,0,48,46,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49, -20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, -0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0, -13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13, -2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97, +120,0,0,0,9,18,118,0,59,120,0,18,97,0,23,0,9,18,118,0,59,121,0,18,97,0,23,0,0,1,0,0,2,4,1,0,2,6, +118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,0,1,0, +0,2,1,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0, +0,18,118,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,5,97, +0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,59, +120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122, +0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120, +0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2, +8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120, +120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116, +114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0, +1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18, +97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0, +18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,9,18,118,0,59,119,0, +18,97,0,24,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, +118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,9,97,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0, +59,120,120,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,4,1,0, +2,10,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112, +0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, +59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0, +9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120, +120,120,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116, +114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,3,1, +0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59, +120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,9,97,0, +0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18, +97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0, +59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4, +118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,2,1, +0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18, +118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0, +0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1, +1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0, +16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2, +21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59, +121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110, +0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46, +20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1, +1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0, +16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0, +57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20, +0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0, +16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0, +48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10, +49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,0,48,46, +18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120,120,0,48,18,109, +0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16, +10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49, +20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15, +109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, +18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, +57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10, +50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, +16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120, +120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,121,0,48,46,18,109,0, +16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16, +8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, +0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16, +10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122, +122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120, +120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,121,0,48,46,18,109,0,16, +10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10, +50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, +16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10, +51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,51,0,57,59,122,122,122, +122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59,119,119,119,119,0,48,46,20,0,0,1,0,15, +2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, +0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, +57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0, +13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13, +2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, +57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97, 0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48, -0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20, -0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2,26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0, -1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, -47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2, -27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47, -20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, -0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0, -16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9, +0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20, +0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, +16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9, 98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9, -97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8, -48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1, -1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2, -27,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, -110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, -49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0, -15,2,27,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, -0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, -0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0, -0,1,0,15,2,21,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, -18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18, -110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16, -10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57, -48,20,0,0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16, -10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57, -18,98,0,48,20,0,0,1,0,15,2,22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0, -18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0, -16,10,51,0,57,49,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, -16,10,51,0,57,18,98,0,49,20,0,0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0, -59,121,121,0,48,46,20,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21, -1,1,0,14,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18, -118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50, -0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, -57,0,0,20,0,0,1,0,12,2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121, -121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57, -18,118,0,59,119,119,119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, -57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10, -51,0,57,0,0,20,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110, -0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109, -0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0, -57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109, -0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14, -109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2, -2,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22, -0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2, -4,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24, -0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, -50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1, -0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51, -0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18, -109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10, -50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2, -1,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, -57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22, -0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0, -9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2, -1,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, -57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0, -0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50, -0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, -23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2, -14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97, -0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18, -109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18, -97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9, -18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57, -18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1, -9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0, -57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2, +26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, +110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, +49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46, +20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, +0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50, +0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14, +110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0, +14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, +48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0, +1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, +97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110, +0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50, +0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, +57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109, +0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0, +57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2,27,1,1,0,9,97,0,0,1,1,0,15,110,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,9, +98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0,0,1,0,15,2,21,1,1,0,9,97,0,0, +1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, +48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,15,2,21,1,1,0, +15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, +48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,0,15,2, +22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, +110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, +49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,49,20,0,0,1,0, +15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, +0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, +0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,49,20,0, +0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16, +8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,0,48,46,20,0,0,1,0, +10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100, +111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, +100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21,1,1,0,14,109,0,0,1,1,0,11, +118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,0,48, +18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122, +122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,12, +2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0, +57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,121,0,48,46,18, +109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,118,0,59,119,119, +119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0, +0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21, +0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,13,110,0, +0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10, +49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20, +0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0,14, +110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0, +16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2,2,1,0,2,14,109,0, +0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,0,0,2,3,1,0,2, +14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,14,109,0,0, +1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,0,0,2,1,1,0,2, +15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16, +10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18, +109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1, +9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, +22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10, +51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20, +0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, +50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1, +0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,0,0, +2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0, +57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23, +0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, +16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0, +9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0, +16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57, +18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,0,0,2, +3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0, +57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,14,109,0,0,1,1,0,9,97,0,0, 0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50, -0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109, -0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9, -18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18, -118,0,18,109,0,48,20,0,0,1,0,5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58, -105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, -7,2,25,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0, -58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1, -0,9,2,25,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,18,97,0,20,0,0,1,0,10,2,25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49, -0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0, -0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,118,0,20,0,0,1,0,12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0, -17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13, -109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47, -20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0, -9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16, -8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49, -0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,18,109,0,20,0,0,1,0,15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, -118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118, -101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2, -24,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -97,0,20,0,0,1,0,6,2,24,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18, -118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,0,1,0,8,2,24,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0, -0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18, -97,0,18,97,0,17,49,0,48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24, -1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95, -114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, -12,2,24,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9, -18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,109,0,20,0,0,1,0,14,2,24,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, -15,2,24,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, -49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, -48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101, -99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16, -10,49,0,47,20,0,0,1,0,6,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20, -0,0,1,0,7,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95, -95,112,111,115,116,68,101,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0, -20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115, -116,68,101,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0, -18,97,0,17,49,0,48,0,0,47,20,0,0,1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0, -48,0,0,0,0,47,20,0,0,1,0,11,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, -47,20,0,0,1,0,12,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0, -1,0,13,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0, -0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0, -0,47,20,0,0,1,0,14,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17, -49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0, -48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0, -0,0,0,47,20,0,0,1,0,15,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0, -17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, -49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, -48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115, -116,73,110,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118, -0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110, -99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0, -2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101, -99,52,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0, -95,95,112,111,115,116,73,110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112, -111,115,116,73,110,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, -18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73, -110,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, -118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114, -0,1,0,2,13,109,0,0,0,1,3,2,0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110, -99,114,0,1,0,2,14,109,0,0,0,1,3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110, -99,114,0,1,0,2,15,109,0,0,0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, -58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58, -118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0, -0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0, -0,0,0,1,0,1,2,15,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102, -108,111,97,116,0,18,98,0,0,0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0, -0,4,102,108,111,97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2, -16,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0, -18,98,0,0,0,41,0,0,1,0,1,2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4, -102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95, -101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0, -5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0, -0,0,43,0,0,1,0,1,2,17,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108, -111,97,116,95,108,101,115,115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113, -117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0, -1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0, -0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114, -105,110,116,0,18,102,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105, -110,116,95,112,114,105,110,116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1, -98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77, -69,83,65,0,1,1,0,10,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, -58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, -83,65,0,1,1,0,11,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58, -112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, -18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112, -114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1, -0,6,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105, -110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7, -118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0, -0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77, -69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0, -9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83, -65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58, -112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, -18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112, -114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112, -114,105,110,116,77,69,83,65,0,1,1,0,4,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, +0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10, +51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97, +0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16, +10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18, +97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,9,18,109,0, +16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57, +18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,9,18,109, +0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,118,0,18,118,0, +18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48, +20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0, +5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0, +18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49, +0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,25,1,0,2,7,118,0,0,0,1,9, +18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0, +18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49, +0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,25,1,0,2,9,97,0,0,0,1,9, +18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2, +25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0, +57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, +15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, +0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2,24,1,0,2,5,97,0,0,0,1,9,18,97, +0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,24,1,0,2,6, +118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101, +99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,24,1,0,2, +8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24,1,0,2,10,118,0,0,0,1,9,18,118,0, +18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0, +20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, +46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,12,2,24,1,0,2,12,118,0,0,0,1,9,18, +118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118, +101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101, +99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,24, +1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0, +0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0, +0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,15,2,24,1,0,2,15,109,0,0,0,1,9,18,109, +0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10, +50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0, +57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,47,20,0,0,1,0,6,0,95,95,112, +111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, +18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,0,1,0,7,0,95,95,112,111,115,116,68, +101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, +118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95,95,112,111,115,116,68,101,99,114, +0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105, +118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,9,97, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,0, +1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,11,0, +95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,12,0,95,95,112, +111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0, +9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,13,0,95,95,112,111,115,116, +68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0, +16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10, +49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,14,0,95,95,112, +111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0, +9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18, +109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109, +0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,15,0,95, +95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109, +0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0, +9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18, +109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109, +0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,9,0,95, +95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0, +20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2, +10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99, +50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,11,118,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49, +0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0, +46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0,95,95,112,111,115,116,73, +110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, +118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112,111,115,116,73,110,99,114, +0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105, +118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,8,118, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51, +0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,13,109,0,0,0,1,3,2, +0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49, +0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48, +0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,14,109,0,0,0,1, +3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0, +17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17, +49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0, +48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,15,109,0,0, +0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52, +0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, +103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0,0,0,0,1,0,1,2,15,1,1,0,5, +97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0, +0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0,0,4,102,108,111,97,116,95, +108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2,16,1,1,0,5,97,0,0,1,1,0,5, +98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,41,0,0,1,0,1, +2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108, +101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101, +0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1, +8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,43,0,0,1,0,1,2,17,1,1,0, +9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115, +115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18, +97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58, +102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0,0,1,0,0,0,112,114,105, +110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114,105,110,116,0,18,102,0,0, +0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105,110,116,95,112,114,105,110, +116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1,98,0,0,0,1,4,98,111,111,108, +95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,10,118,0,0, +0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69, +83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,11,118,0,0,0,1,9, +58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65, +0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0, +112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114, +105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, +59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,6,118,0,0,0,1,9,58,112,114,105, +110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59, +121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7,118,0,0,0,1,9,58,112,114,105,110, +116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0, +0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116, +77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, +58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65, +0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0, +112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112, +114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, 59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0, -0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77, -69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0, -57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116, -77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49, -0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114, -105,110,116,77,69,83,65,0,1,1,0,15,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16, -8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105, -110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17, -101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77, -69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112, -114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0, -0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105, -110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110, -116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 +116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,4,118, +0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77, +69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0, +9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, +83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9, +58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77, +69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0, +9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69, +83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,15,109,0,0, +0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77, +69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0, +57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105, +110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1, +0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116, +0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95, +112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0, +1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0, +1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110, +116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 -- cgit v1.2.3 From 0ed3f6f575ebdb345ad9891380686eafce63c75e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 16:23:18 -0600 Subject: mesa: fix array storage allocation bug --- src/mesa/shader/slang/slang_codegen.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 4846c257f1..14fa38c3cd 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2071,6 +2071,14 @@ _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) n->Store->File = PROGRAM_TEMPORARY; n->Store->Size = _slang_sizeof_type_specifier(&n->Var->type.specifier); + if (var->array_len > 0) { + /* this is an array */ + /* round up element size to mult of 4 */ + GLint sz = (n->Store->Size + 3) & ~3; + /* mult by array size */ + sz *= var->array_len; + n->Store->Size = sz; + } A->program->NumTemporaries++; assert(n->Store->Size > 0); } -- cgit v1.2.3 From 7366f3fc54be280969f3adef16ed67fddfe55e67 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 16:24:26 -0600 Subject: mesa: fix incorrect array size, added assertion --- src/mesa/shader/slang/slang_vartable.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_vartable.c b/src/mesa/shader/slang/slang_vartable.c index 26411255df..bb4d2d656c 100644 --- a/src/mesa/shader/slang/slang_vartable.c +++ b/src/mesa/shader/slang/slang_vartable.c @@ -28,7 +28,7 @@ struct table slang_variable **Vars; /* array [NumVars] */ TempState Temps[MAX_PROGRAM_TEMPS * 4]; /* per-component state */ - int ValSize[MAX_PROGRAM_TEMPS]; /* For debug only */ + int ValSize[MAX_PROGRAM_TEMPS * 4]; /**< For debug only */ struct table *Parent; /** Parent scope table */ }; @@ -220,6 +220,7 @@ alloc_reg(slang_var_table *vt, GLint size, GLboolean isTemp) assert(i % 4 == 0); for (j = 0; j < size; j++) t->Temps[i + j] = isTemp ? TEMP : VAR; + assert(i < MAX_PROGRAM_TEMPS * 4); t->ValSize[i] = size; return i; } -- cgit v1.2.3 From dd3311aa89035aa2e624f54b6914cfbcf57b33d7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 3 Jul 2008 16:59:24 -0600 Subject: mesa: added _vbo_VertexAttrib4f() --- src/mesa/vbo/vbo.h | 2 ++ src/mesa/vbo/vbo_exec_api.c | 7 +++++++ 2 files changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h index bd10956ee4..59cd7f6984 100644 --- a/src/mesa/vbo/vbo.h +++ b/src/mesa/vbo/vbo.h @@ -132,5 +132,7 @@ _vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q); void _vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params); +void +_vbo_VertexAttrib4f(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); #endif diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index 1e2a08bc8e..2c2e66dd0b 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -795,3 +795,10 @@ _vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params) { vbo_Materialfv(face, pname, params); } + + +void +_vbo_VertexAttrib4f(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w) +{ + vbo_VertexAttrib4fARB(index, x, y, z, w); +} -- cgit v1.2.3 From 1942e29bf7e4df34164ed815ccd77b08cd28ed56 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 5 Jul 2008 00:55:18 +0900 Subject: softpipe: Implement texture blankets. --- src/gallium/drivers/softpipe/sp_texture.c | 34 +++++++++++++++++++++++++++++++ 1 file changed, 34 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 4db045cdc3..20ad336b4f 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -162,6 +162,39 @@ softpipe_texture_create(struct pipe_screen *screen, } +static struct pipe_texture * +softpipe_texture_blanket(struct pipe_screen * screen, + const struct pipe_texture *base, + const unsigned *stride, + struct pipe_buffer *buffer) +{ + struct softpipe_texture *spt; + assert(screen); + + /* Only supports one type */ + if (base->target != PIPE_TEXTURE_2D || + base->last_level != 0 || + base->depth[0] != 1) { + return NULL; + } + + spt = CALLOC_STRUCT(softpipe_texture); + if (!spt) + return NULL; + + spt->base = *base; + spt->base.refcount = 1; + spt->base.screen = screen; + spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]); + spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]); + spt->stride[0] = stride[0]; + + pipe_buffer_reference(screen->winsys, &spt->buffer, buffer); + + return &spt->base; +} + + static void softpipe_texture_release(struct pipe_screen *screen, struct pipe_texture **pt) @@ -309,6 +342,7 @@ void softpipe_init_screen_texture_funcs(struct pipe_screen *screen) { screen->texture_create = softpipe_texture_create; + screen->texture_blanket = softpipe_texture_blanket; screen->texture_release = softpipe_texture_release; screen->get_tex_surface = softpipe_get_tex_surface; -- cgit v1.2.3 From 4a18324c0bca4ae806b1c62cad44e859f0924a9b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 5 Jul 2008 00:56:36 +0900 Subject: psb: Fill all texture fields when creating texture blanket. --- src/gallium/drivers/i915simple/i915_texture.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 2815e61345..5d6769924c 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -710,6 +710,8 @@ i915_texture_blanket(struct pipe_screen * screen, return NULL; tex->base = *base; + tex->base.refcount = 1; + tex->base.screen = screen; tex->stride = stride[0]; -- cgit v1.2.3 From c4b6941e62f5e30a7c479b7b8d6521cc4dd92e3e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Jul 2008 09:56:08 -0600 Subject: mesa: move assertion --- src/mesa/main/fbobject.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index e4ff575e18..56a3131016 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -185,9 +185,9 @@ _mesa_remove_attachment(GLcontext *ctx, struct gl_renderbuffer_attachment *att) att->Texture = NULL; } if (att->Type == GL_TEXTURE || att->Type == GL_RENDERBUFFER_EXT) { - ASSERT(att->Renderbuffer); ASSERT(!att->Texture); - _mesa_reference_renderbuffer(&att->Renderbuffer, NULL); + _mesa_reference_renderbuffer(&att->Renderbuffer, NULL); /* unbind */ + ASSERT(!att->Renderbuffer); } att->Type = GL_NONE; att->Complete = GL_TRUE; -- cgit v1.2.3 From 294b061256220960e7e29fbc8ecbd2ffce75de40 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Jul 2008 09:58:14 -0600 Subject: mesa: generate GL_INVALID_OPERATION in _mesa_get_uniform_location() if program isn't linked --- src/mesa/shader/shader_api.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 3e28d92fe1..182de37b50 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -1011,6 +1011,11 @@ _mesa_get_uniform_location(GLcontext *ctx, GLuint program, const GLchar *name) if (!shProg) return -1; + if (shProg->LinkStatus == GL_FALSE) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glGetUniformfv(program)"); + return -1; + } + /* XXX we should return -1 if the uniform was declared, but not * actually used. */ -- cgit v1.2.3 From ba9e6339028c36269cb50bb8535e415fa8e6e4c9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Jul 2008 09:59:43 -0600 Subject: gallium: fix trim() function bug when count < first If the user called glDrawArrays(GL_TRIANGLES, count=1), trim() returned a very large integer because of the unsigned arithmetic. --- src/gallium/auxiliary/draw/draw_pt.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 9140faeea9..85a75525c8 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -37,6 +37,8 @@ static unsigned trim( unsigned count, unsigned first, unsigned incr ) { + if (count < first) + return 0; return count - (count - first) % incr; } -- cgit v1.2.3 From 9ca1c62a963ca7024c4bccf83af3f90955cd5068 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Jul 2008 10:02:16 -0600 Subject: gallium: replace assertion with conditional --- src/mesa/state_tracker/st_cb_fbo.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 7245798d0d..8406bf247f 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -422,7 +422,8 @@ st_finish_render_texture(GLcontext *ctx, struct pipe_screen *screen = ctx->st->pipe->screen; struct st_renderbuffer *strb = st_renderbuffer(att->Renderbuffer); - assert(strb); + if (!strb) + return; ctx->st->pipe->flush(ctx->st->pipe, PIPE_FLUSH_RENDER_CACHE, NULL); -- cgit v1.2.3 From 2fa7b3f78639114aec42fcbbfc29d3645832708b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Jul 2008 10:29:15 -0600 Subject: mesa: Implement mutex/locking around texture object reference counting. Use new _mesa_reference_texobj() function for referencing/unreferencing textures. Add new assertions/tests to try to detect invalid usage of deleted textures. cherry-picked from master (9e01b915f1243a3f551cb795b7124bd1e52ca15f) --- src/mesa/main/attrib.c | 1 + src/mesa/main/context.c | 14 +-- src/mesa/main/fbobject.c | 17 ++- src/mesa/main/framebuffer.c | 13 +-- src/mesa/main/mtypes.h | 1 + src/mesa/main/texobj.c | 253 ++++++++++++++++++++++---------------------- src/mesa/main/texobj.h | 4 + src/mesa/main/texstate.c | 104 ++++++++---------- 8 files changed, 194 insertions(+), 213 deletions(-) (limited to 'src') diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c index 2e6bb76586..b990369a9e 100644 --- a/src/mesa/main/attrib.c +++ b/src/mesa/main/attrib.c @@ -361,6 +361,7 @@ _mesa_PushAttrib(GLbitfield mask) ctx->Texture.Unit[u].Current1DArray->RefCount++; ctx->Texture.Unit[u].Current2DArray->RefCount++; } + attr = MALLOC_STRUCT( gl_texture_attrib ); MEMCPY( attr, &ctx->Texture, sizeof(struct gl_texture_attrib) ); /* copy state of the currently bound texture objects */ diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 33f6d2c0d0..279880cf40 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -500,19 +500,12 @@ alloc_shared_state( GLcontext *ctx ) if (!ss->Default2DArray) goto cleanup; - /* Effectively bind the default textures to all texture units */ - ss->Default1D->RefCount += MAX_TEXTURE_IMAGE_UNITS; - ss->Default2D->RefCount += MAX_TEXTURE_IMAGE_UNITS; - ss->Default3D->RefCount += MAX_TEXTURE_IMAGE_UNITS; - ss->DefaultCubeMap->RefCount += MAX_TEXTURE_IMAGE_UNITS; - ss->DefaultRect->RefCount += MAX_TEXTURE_IMAGE_UNITS; - ss->Default1DArray->RefCount += MAX_TEXTURE_IMAGE_UNITS; - ss->Default2DArray->RefCount += MAX_TEXTURE_IMAGE_UNITS; + /* sanity check */ + assert(ss->Default1D->RefCount == 1); _glthread_INIT_MUTEX(ss->TexMutex); ss->TextureStateStamp = 0; - #if FEATURE_EXT_framebuffer_object ss->FrameBuffers = _mesa_NewHashTable(); if (!ss->FrameBuffers) @@ -522,10 +515,9 @@ alloc_shared_state( GLcontext *ctx ) goto cleanup; #endif - return GL_TRUE; - cleanup: +cleanup: /* Ran out of memory at some point. Free everything and return NULL */ if (ss->DisplayList) _mesa_DeleteHashTable(ss->DisplayList); diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 56a3131016..0ae69bdce7 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -172,17 +172,12 @@ _mesa_remove_attachment(GLcontext *ctx, struct gl_renderbuffer_attachment *att) { if (att->Type == GL_TEXTURE) { ASSERT(att->Texture); - att->Texture->RefCount--; - if (att->Texture->RefCount == 0) { - ctx->Driver.DeleteTexture(ctx, att->Texture); - } - else { + if (ctx->Driver.FinishRenderTexture) { /* tell driver that we're done rendering to this texture. */ - if (ctx->Driver.FinishRenderTexture) { - ctx->Driver.FinishRenderTexture(ctx, att); - } + ctx->Driver.FinishRenderTexture(ctx, att); } - att->Texture = NULL; + _mesa_reference_texobj(&att->Texture, NULL); /* unbind */ + ASSERT(!att->Texture); } if (att->Type == GL_TEXTURE || att->Type == GL_RENDERBUFFER_EXT) { ASSERT(!att->Texture); @@ -213,8 +208,8 @@ _mesa_set_texture_attachment(GLcontext *ctx, /* new attachment */ _mesa_remove_attachment(ctx, att); att->Type = GL_TEXTURE; - att->Texture = texObj; - texObj->RefCount++; + assert(!att->Texture); + _mesa_reference_texobj(&att->Texture, texObj); } /* always update these fields */ diff --git a/src/mesa/main/framebuffer.c b/src/mesa/main/framebuffer.c index 96f1b30c9b..dab449fc09 100644 --- a/src/mesa/main/framebuffer.c +++ b/src/mesa/main/framebuffer.c @@ -38,6 +38,7 @@ #include "fbobject.h" #include "framebuffer.h" #include "renderbuffer.h" +#include "texobj.h" @@ -192,17 +193,11 @@ _mesa_free_framebuffer_data(struct gl_framebuffer *fb) _mesa_reference_renderbuffer(&att->Renderbuffer, NULL); } if (att->Texture) { - /* render to texture */ - att->Texture->RefCount--; - if (att->Texture->RefCount == 0) { - GET_CURRENT_CONTEXT(ctx); - if (ctx) { - ctx->Driver.DeleteTexture(ctx, att->Texture); - } - } + _mesa_reference_texobj(&att->Texture, NULL); } + ASSERT(!att->Renderbuffer); + ASSERT(!att->Texture); att->Type = GL_NONE; - att->Texture = NULL; } /* unbind _Depth/_StencilBuffer to decr ref counts */ diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 00e7d5d395..a38ec02852 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -1404,6 +1404,7 @@ struct gl_texture_image */ struct gl_texture_object { + _glthread_Mutex Mutex; /**< for thread safety */ GLint RefCount; /**< reference count */ GLuint Name; /**< the user-visible texture object ID */ GLenum Target; /**< GL_TEXTURE_1D, GL_TEXTURE_2D, etc. */ diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c index 606e62d7a0..b77a00dd15 100644 --- a/src/mesa/main/texobj.c +++ b/src/mesa/main/texobj.c @@ -5,9 +5,9 @@ /* * Mesa 3-D graphics library - * Version: 6.5 + * Version: 7.1 * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -108,6 +108,7 @@ _mesa_initialize_texture_object( struct gl_texture_object *obj, _mesa_bzero(obj, sizeof(*obj)); /* init the non-zero fields */ + _glthread_INIT_MUTEX(obj->Mutex); obj->RefCount = 1; obj->Name = name; obj->Target = target; @@ -155,6 +156,11 @@ _mesa_delete_texture_object( GLcontext *ctx, struct gl_texture_object *texObj ) (void) ctx; + /* Set Target to an invalid value. With some assertions elsewhere + * we can try to detect possible use of deleted textures. + */ + texObj->Target = 0x99; + #if FEATURE_colortable _mesa_free_colortable_data(&texObj->Palette); #endif @@ -168,6 +174,9 @@ _mesa_delete_texture_object( GLcontext *ctx, struct gl_texture_object *texObj ) } } + /* destroy the mutex -- it may have allocated memory (eg on bsd) */ + _glthread_DESTROY_MUTEX(texObj->Mutex); + /* free this object */ _mesa_free(texObj); } @@ -186,6 +195,7 @@ void _mesa_copy_texture_object( struct gl_texture_object *dest, const struct gl_texture_object *src ) { + dest->Target = src->Target; dest->Name = src->Name; dest->Priority = src->Priority; dest->BorderColor[0] = src->BorderColor[0]; @@ -217,6 +227,94 @@ _mesa_copy_texture_object( struct gl_texture_object *dest, } +/** + * Check if the given texture object is valid by examining its Target field. + * For debugging only. + */ +static GLboolean +valid_texture_object(const struct gl_texture_object *tex) +{ + switch (tex->Target) { + case 0: + case GL_TEXTURE_1D: + case GL_TEXTURE_2D: + case GL_TEXTURE_3D: + case GL_TEXTURE_CUBE_MAP_ARB: + case GL_TEXTURE_RECTANGLE_NV: + case GL_TEXTURE_1D_ARRAY_EXT: + case GL_TEXTURE_2D_ARRAY_EXT: + return GL_TRUE; + case 0x99: + _mesa_problem(NULL, "invalid reference to a deleted texture object"); + return GL_FALSE; + default: + _mesa_problem(NULL, "invalid texture object Target value"); + return GL_FALSE; + } +} + + +/** + * Reference (or unreference) a texture object. + * If '*ptr', decrement *ptr's refcount (and delete if it becomes zero). + * If 'tex' is non-null, increment its refcount. + */ +void +_mesa_reference_texobj(struct gl_texture_object **ptr, + struct gl_texture_object *tex) +{ + assert(ptr); + if (*ptr == tex) { + /* no change */ + return; + } + + if (*ptr) { + /* Unreference the old texture */ + GLboolean deleteFlag = GL_FALSE; + struct gl_texture_object *oldTex = *ptr; + + assert(valid_texture_object(oldTex)); + + _glthread_LOCK_MUTEX(oldTex->Mutex); + ASSERT(oldTex->RefCount > 0); + oldTex->RefCount--; + + deleteFlag = (oldTex->RefCount == 0); + _glthread_UNLOCK_MUTEX(oldTex->Mutex); + + if (deleteFlag) { + GET_CURRENT_CONTEXT(ctx); + if (ctx) + ctx->Driver.DeleteTexture(ctx, oldTex); + else + _mesa_problem(NULL, "Unable to delete texture, no context"); + } + + *ptr = NULL; + } + assert(!*ptr); + + if (tex) { + /* reference new texture */ + assert(valid_texture_object(tex)); + _glthread_LOCK_MUTEX(tex->Mutex); + if (tex->RefCount == 0) { + /* this texture's being deleted (look just above) */ + /* Not sure this can every really happen. Warn if it does. */ + _mesa_problem(NULL, "referencing deleted texture object"); + *ptr = NULL; + } + else { + tex->RefCount++; + *ptr = tex; + } + _glthread_UNLOCK_MUTEX(tex->Mutex); + } +} + + + /** * Report why a texture object is incomplete. * @@ -613,8 +711,7 @@ unbind_texobj_from_fbo(GLcontext *ctx, struct gl_texture_object *texObj) /** * Check if the given texture object is bound to any texture image units and - * unbind it if so. - * XXX all RefCount accesses should be protected by a mutex. + * unbind it if so (revert to default textures). */ static void unbind_texobj_from_texunits(GLcontext *ctx, struct gl_texture_object *texObj) @@ -623,42 +720,26 @@ unbind_texobj_from_texunits(GLcontext *ctx, struct gl_texture_object *texObj) for (u = 0; u < MAX_TEXTURE_IMAGE_UNITS; u++) { struct gl_texture_unit *unit = &ctx->Texture.Unit[u]; - struct gl_texture_object **curr = NULL; - if (texObj == unit->Current1D) { - curr = &unit->Current1D; - unit->Current1D = ctx->Shared->Default1D; + _mesa_reference_texobj(&unit->Current1D, ctx->Shared->Default1D); } else if (texObj == unit->Current2D) { - curr = &unit->Current2D; - unit->Current2D = ctx->Shared->Default2D; + _mesa_reference_texobj(&unit->Current2D, ctx->Shared->Default2D); } else if (texObj == unit->Current3D) { - curr = &unit->Current3D; - unit->Current3D = ctx->Shared->Default3D; + _mesa_reference_texobj(&unit->Current3D, ctx->Shared->Default3D); } else if (texObj == unit->CurrentCubeMap) { - curr = &unit->CurrentCubeMap; - unit->CurrentCubeMap = ctx->Shared->DefaultCubeMap; + _mesa_reference_texobj(&unit->CurrentCubeMap, ctx->Shared->DefaultCubeMap); } else if (texObj == unit->CurrentRect) { - curr = &unit->CurrentRect; - unit->CurrentRect = ctx->Shared->DefaultRect; + _mesa_reference_texobj(&unit->CurrentRect, ctx->Shared->DefaultRect); } else if (texObj == unit->Current1DArray) { - curr = &unit->Current1DArray; - unit->CurrentRect = ctx->Shared->Default1DArray; + _mesa_reference_texobj(&unit->Current1DArray, ctx->Shared->Default1DArray); } else if (texObj == unit->Current2DArray) { - curr = &unit->Current1DArray; - unit->CurrentRect = ctx->Shared->Default2DArray; - } - - if (curr) { - (*curr)->RefCount++; - texObj->RefCount--; - if (texObj == unit->_Current) - unit->_Current = *curr; + _mesa_reference_texobj(&unit->Current2DArray, ctx->Shared->Default2DArray); } } } @@ -694,8 +775,6 @@ _mesa_DeleteTextures( GLsizei n, const GLuint *textures) = _mesa_lookup_texture(ctx, textures[i]); if (delObj) { - GLboolean deleted; - _mesa_lock_texture(ctx, delObj); /* Check if texture is bound to any framebuffer objects. @@ -705,10 +784,12 @@ _mesa_DeleteTextures( GLsizei n, const GLuint *textures) unbind_texobj_from_fbo(ctx, delObj); /* Check if this texture is currently bound to any texture units. - * If so, unbind it and decrement the reference count. + * If so, unbind it. */ unbind_texobj_from_texunits(ctx, delObj); + _mesa_unlock_texture(ctx, delObj); + ctx->NewState |= _NEW_TEXTURE; /* The texture _name_ is now free for re-use. @@ -718,23 +799,10 @@ _mesa_DeleteTextures( GLsizei n, const GLuint *textures) _mesa_HashRemove(ctx->Shared->TexObjects, delObj->Name); _glthread_UNLOCK_MUTEX(ctx->Shared->Mutex); - /* The actual texture object will not be freed until it's no - * longer bound in any context. - * XXX all RefCount accesses should be protected by a mutex. + /* Unreference the texobj. If refcount hits zero, the texture + * will be deleted. */ - delObj->RefCount--; - deleted = (delObj->RefCount == 0); - _mesa_unlock_texture(ctx, delObj); - - /* We know that refcount went to zero above, so this is - * the only pointer left to delObj, so we don't have to - * worry about locking any more: - */ - if (deleted) { - ASSERT(delObj->Name != 0); /* Never delete default tex objs */ - ASSERT(ctx->Driver.DeleteTexture); - (*ctx->Driver.DeleteTexture)(ctx, delObj); - } + _mesa_reference_texobj(&delObj, NULL); } } } @@ -762,7 +830,6 @@ _mesa_BindTexture( GLenum target, GLuint texName ) GET_CURRENT_CONTEXT(ctx); const GLuint unit = ctx->Texture.CurrentUnit; struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *oldTexObj; struct gl_texture_object *newTexObj = NULL; ASSERT_OUTSIDE_BEGIN_END(ctx); @@ -770,62 +837,6 @@ _mesa_BindTexture( GLenum target, GLuint texName ) _mesa_debug(ctx, "glBindTexture %s %d\n", _mesa_lookup_enum_by_nr(target), (GLint) texName); - /* - * Get pointer to currently bound texture object (oldTexObj) - */ - switch (target) { - case GL_TEXTURE_1D: - oldTexObj = texUnit->Current1D; - break; - case GL_TEXTURE_2D: - oldTexObj = texUnit->Current2D; - break; - case GL_TEXTURE_3D: - oldTexObj = texUnit->Current3D; - break; - case GL_TEXTURE_CUBE_MAP_ARB: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error( ctx, GL_INVALID_ENUM, "glBindTexture(target)" ); - return; - } - oldTexObj = texUnit->CurrentCubeMap; - break; - case GL_TEXTURE_RECTANGLE_NV: - if (!ctx->Extensions.NV_texture_rectangle) { - _mesa_error( ctx, GL_INVALID_ENUM, "glBindTexture(target)" ); - return; - } - oldTexObj = texUnit->CurrentRect; - break; - case GL_TEXTURE_1D_ARRAY_EXT: - if (!ctx->Extensions.MESA_texture_array) { - _mesa_error( ctx, GL_INVALID_ENUM, "glBindTexture(target)" ); - return; - } - oldTexObj = texUnit->Current1DArray; - break; - case GL_TEXTURE_2D_ARRAY_EXT: - if (!ctx->Extensions.MESA_texture_array) { - _mesa_error( ctx, GL_INVALID_ENUM, "glBindTexture(target)" ); - return; - } - oldTexObj = texUnit->Current2DArray; - break; - default: - _mesa_error( ctx, GL_INVALID_ENUM, "glBindTexture(target)" ); - return; - } - - if (oldTexObj->Name == texName) { - /* XXX this might be wrong. If the texobj is in use by another - * context and a texobj parameter was changed, this might be our - * only chance to update this context's hardware state. - * Note that some applications re-bind the same texture a lot so we - * want to handle that case quickly. - */ - return; /* rebinding the same texture- no change */ - } - /* * Get pointer to new texture object (newTexObj) */ @@ -854,7 +865,8 @@ _mesa_BindTexture( GLenum target, GLuint texName ) newTexObj = ctx->Shared->Default2DArray; break; default: - ; /* Bad targets are caught above */ + _mesa_error(ctx, GL_INVALID_ENUM, "glBindTexture(target)"); + return; } } else { @@ -900,28 +912,30 @@ _mesa_BindTexture( GLenum target, GLuint texName ) newTexObj->Target = target; } - /* XXX all RefCount accesses should be protected by a mutex. */ - newTexObj->RefCount++; + assert(valid_texture_object(newTexObj)); - /* do the actual binding, but first flush outstanding vertices: - */ + /* flush before changing binding */ FLUSH_VERTICES(ctx, _NEW_TEXTURE); + /* Do the actual binding. The refcount on the previously bound + * texture object will be decremented. It'll be deleted if the + * count hits zero. + */ switch (target) { case GL_TEXTURE_1D: - texUnit->Current1D = newTexObj; + _mesa_reference_texobj(&texUnit->Current1D, newTexObj); break; case GL_TEXTURE_2D: - texUnit->Current2D = newTexObj; + _mesa_reference_texobj(&texUnit->Current2D, newTexObj); break; case GL_TEXTURE_3D: - texUnit->Current3D = newTexObj; + _mesa_reference_texobj(&texUnit->Current3D, newTexObj); break; case GL_TEXTURE_CUBE_MAP_ARB: - texUnit->CurrentCubeMap = newTexObj; + _mesa_reference_texobj(&texUnit->CurrentCubeMap, newTexObj); break; case GL_TEXTURE_RECTANGLE_NV: - texUnit->CurrentRect = newTexObj; + _mesa_reference_texobj(&texUnit->CurrentRect, newTexObj); break; case GL_TEXTURE_1D_ARRAY_EXT: texUnit->Current1DArray = newTexObj; @@ -930,6 +944,7 @@ _mesa_BindTexture( GLenum target, GLuint texName ) texUnit->Current2DArray = newTexObj; break; default: + /* Bad target should be caught above */ _mesa_problem(ctx, "bad target in BindTexture"); return; } @@ -937,18 +952,6 @@ _mesa_BindTexture( GLenum target, GLuint texName ) /* Pass BindTexture call to device driver */ if (ctx->Driver.BindTexture) (*ctx->Driver.BindTexture)( ctx, target, newTexObj ); - - /* Decrement the reference count on the old texture and check if it's - * time to delete it. - */ - /* XXX all RefCount accesses should be protected by a mutex. */ - oldTexObj->RefCount--; - ASSERT(oldTexObj->RefCount >= 0); - if (oldTexObj->RefCount == 0) { - ASSERT(oldTexObj->Name != 0); - ASSERT(ctx->Driver.DeleteTexture); - (*ctx->Driver.DeleteTexture)( ctx, oldTexObj ); - } } diff --git a/src/mesa/main/texobj.h b/src/mesa/main/texobj.h index 2a2bde3601..d5374c5d6c 100644 --- a/src/mesa/main/texobj.h +++ b/src/mesa/main/texobj.h @@ -57,6 +57,10 @@ extern void _mesa_copy_texture_object( struct gl_texture_object *dest, const struct gl_texture_object *src ); +extern void +_mesa_reference_texobj(struct gl_texture_object **ptr, + struct gl_texture_object *tex); + extern void _mesa_test_texobj_completeness( const GLcontext *ctx, struct gl_texture_object *obj ); diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c index 421f912849..3bdb55257f 100644 --- a/src/mesa/main/texstate.c +++ b/src/mesa/main/texstate.c @@ -62,31 +62,6 @@ static const struct gl_tex_env_combine_state default_combine_state = { }; -/** - * Copy a texture binding. Helper used by _mesa_copy_texture_state(). - */ -static void -copy_texture_binding(const GLcontext *ctx, - struct gl_texture_object **dst, - struct gl_texture_object *src) -{ - /* only copy if names differ (per OpenGL SI) */ - if ((*dst)->Name != src->Name) { - /* unbind/delete dest binding which we're changing */ - (*dst)->RefCount--; - if ((*dst)->RefCount == 0) { - /* time to delete this texture object */ - ASSERT((*dst)->Name != 0); - ASSERT(ctx->Driver.DeleteTexture); - /* XXX cast-away const, unfortunately */ - (*ctx->Driver.DeleteTexture)((GLcontext *) ctx, *dst); - } - /* make new binding, incrementing ref count */ - *dst = src; - src->RefCount++; - } -} - /** * Used by glXCopyContext to copy texture state from one context to another. @@ -143,20 +118,20 @@ _mesa_copy_texture_state( const GLcontext *src, GLcontext *dst ) /* copy texture object bindings, not contents of texture objects */ _mesa_lock_context_textures(dst); - copy_texture_binding(src, &dst->Texture.Unit[i].Current1D, - src->Texture.Unit[i].Current1D); - copy_texture_binding(src, &dst->Texture.Unit[i].Current2D, - src->Texture.Unit[i].Current2D); - copy_texture_binding(src, &dst->Texture.Unit[i].Current3D, - src->Texture.Unit[i].Current3D); - copy_texture_binding(src, &dst->Texture.Unit[i].CurrentCubeMap, - src->Texture.Unit[i].CurrentCubeMap); - copy_texture_binding(src, &dst->Texture.Unit[i].CurrentRect, - src->Texture.Unit[i].CurrentRect); - copy_texture_binding(src, &dst->Texture.Unit[i].Current1DArray, - src->Texture.Unit[i].Current1DArray); - copy_texture_binding(src, &dst->Texture.Unit[i].Current2DArray, - src->Texture.Unit[i].Current2DArray); + _mesa_reference_texobj(&dst->Texture.Unit[i].Current1D, + src->Texture.Unit[i].Current1D); + _mesa_reference_texobj(&dst->Texture.Unit[i].Current2D, + src->Texture.Unit[i].Current2D); + _mesa_reference_texobj(&dst->Texture.Unit[i].Current3D, + src->Texture.Unit[i].Current3D); + _mesa_reference_texobj(&dst->Texture.Unit[i].CurrentCubeMap, + src->Texture.Unit[i].CurrentCubeMap); + _mesa_reference_texobj(&dst->Texture.Unit[i].CurrentRect, + src->Texture.Unit[i].CurrentRect); + _mesa_reference_texobj(&dst->Texture.Unit[i].Current1DArray, + src->Texture.Unit[i].Current1DArray); + _mesa_reference_texobj(&dst->Texture.Unit[i].Current2DArray, + src->Texture.Unit[i].Current2DArray); _mesa_unlock_context_textures(dst); } @@ -727,6 +702,8 @@ alloc_proxy_textures( GLcontext *ctx ) if (!ctx->Texture.Proxy2DArray) goto cleanup; + assert(ctx->Texture.Proxy1D->RefCount == 1); + return GL_TRUE; cleanup: @@ -786,13 +763,14 @@ init_texture_unit( GLcontext *ctx, GLuint unit ) ASSIGN_4V( texUnit->EyePlaneR, 0.0, 0.0, 0.0, 0.0 ); ASSIGN_4V( texUnit->EyePlaneQ, 0.0, 0.0, 0.0, 0.0 ); - texUnit->Current1D = ctx->Shared->Default1D; - texUnit->Current2D = ctx->Shared->Default2D; - texUnit->Current3D = ctx->Shared->Default3D; - texUnit->CurrentCubeMap = ctx->Shared->DefaultCubeMap; - texUnit->CurrentRect = ctx->Shared->DefaultRect; - texUnit->Current1DArray = ctx->Shared->Default1DArray; - texUnit->Current2DArray = ctx->Shared->Default2DArray; + /* initialize current texture object ptrs to the shared default objects */ + _mesa_reference_texobj(&texUnit->Current1D, ctx->Shared->Default1D); + _mesa_reference_texobj(&texUnit->Current2D, ctx->Shared->Default2D); + _mesa_reference_texobj(&texUnit->Current3D, ctx->Shared->Default3D); + _mesa_reference_texobj(&texUnit->CurrentCubeMap, ctx->Shared->DefaultCubeMap); + _mesa_reference_texobj(&texUnit->CurrentRect, ctx->Shared->DefaultRect); + _mesa_reference_texobj(&texUnit->Current1DArray, ctx->Shared->Default1DArray); + _mesa_reference_texobj(&texUnit->Current2DArray, ctx->Shared->Default2DArray); } @@ -807,25 +785,22 @@ _mesa_init_texture(GLcontext *ctx) assert(MAX_TEXTURE_LEVELS >= MAX_3D_TEXTURE_LEVELS); assert(MAX_TEXTURE_LEVELS >= MAX_CUBE_TEXTURE_LEVELS); - /* Effectively bind the default textures to all texture units */ - ctx->Shared->Default1D->RefCount += MAX_TEXTURE_UNITS; - ctx->Shared->Default2D->RefCount += MAX_TEXTURE_UNITS; - ctx->Shared->Default3D->RefCount += MAX_TEXTURE_UNITS; - ctx->Shared->DefaultCubeMap->RefCount += MAX_TEXTURE_UNITS; - ctx->Shared->DefaultRect->RefCount += MAX_TEXTURE_UNITS; - ctx->Shared->Default1DArray->RefCount += MAX_TEXTURE_UNITS; - ctx->Shared->Default2DArray->RefCount += MAX_TEXTURE_UNITS; - /* Texture group */ ctx->Texture.CurrentUnit = 0; /* multitexture */ ctx->Texture._EnabledUnits = 0; - for (i=0; iTexture.SharedPalette = GL_FALSE; #if FEATURE_colortable _mesa_init_colortable(&ctx->Texture.Palette); #endif + for (i = 0; i < MAX_TEXTURE_UNITS; i++) + init_texture_unit( ctx, i ); + + /* After we're done initializing the context's texture state the default + * texture objects' refcounts should be at least MAX_TEXTURE_UNITS + 1. + */ + assert(ctx->Shared->Default1D->RefCount >= MAX_TEXTURE_UNITS + 1); + /* Allocate proxy textures */ if (!alloc_proxy_textures( ctx )) return GL_FALSE; @@ -840,6 +815,20 @@ _mesa_init_texture(GLcontext *ctx) void _mesa_free_texture_data(GLcontext *ctx) { + GLuint u; + + /* unreference current textures */ + for (u = 0; u < MAX_TEXTURE_IMAGE_UNITS; u++) { + struct gl_texture_unit *unit = ctx->Texture.Unit + u; + _mesa_reference_texobj(&unit->Current1D, NULL); + _mesa_reference_texobj(&unit->Current2D, NULL); + _mesa_reference_texobj(&unit->Current3D, NULL); + _mesa_reference_texobj(&unit->CurrentCubeMap, NULL); + _mesa_reference_texobj(&unit->CurrentRect, NULL); + _mesa_reference_texobj(&unit->Current1DArray, NULL); + _mesa_reference_texobj(&unit->Current2DArray, NULL); + } + /* Free proxy texture objects */ (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy1D ); (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy2D ); @@ -849,6 +838,7 @@ _mesa_free_texture_data(GLcontext *ctx) (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy1DArray ); (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy2DArray ); + #if FEATURE_colortable { GLuint i; -- cgit v1.2.3 From a3de65659cf07420363c91ae38dd61468f122ef0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Jul 2008 10:35:32 -0600 Subject: gallium: fix st_choose_format(internalFormat=GL_RGBA16) Need to check if we're choosing a surface or texture format. --- src/mesa/state_tracker/st_format.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 03ec558a36..50a06868df 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -362,7 +362,10 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_RGBA12: return default_rgba_format( screen, surfType ); case GL_RGBA16: - return default_deep_rgba_format( screen, surfType ); + if (surfType == PIPE_SURFACE) + return default_deep_rgba_format( screen, surfType ); + else + return default_rgba_format( screen, surfType ); case GL_RGBA4: case GL_RGBA2: -- cgit v1.2.3 From 0bc2409e3808df54bf7d228475320e2ec4fe80a1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 4 Jul 2008 10:37:07 -0600 Subject: mesa: Replace Proxy1D/2D/etc with ProxyTex[] indexed by TEXTURE_x_INDEX. Simplification in colortab.c too. cherry-picked from master (fe469007037d9d5cdbe1677d8ff7368b276e9e7c) --- src/mesa/drivers/dri/tdfx/tdfx_tex.c | 2 +- src/mesa/main/colortab.c | 306 +++++++++++------------------------ src/mesa/main/mtypes.h | 4 + src/mesa/main/teximage.c | 58 +++---- src/mesa/main/texstate.c | 81 ++++------ 5 files changed, 154 insertions(+), 297 deletions(-) (limited to 'src') diff --git a/src/mesa/drivers/dri/tdfx/tdfx_tex.c b/src/mesa/drivers/dri/tdfx/tdfx_tex.c index 65e665ee39..70ef726437 100644 --- a/src/mesa/drivers/dri/tdfx/tdfx_tex.c +++ b/src/mesa/drivers/dri/tdfx/tdfx_tex.c @@ -1821,7 +1821,7 @@ tdfxTestProxyTexImage(GLcontext *ctx, GLenum target, tdfxTexInfo *ti; int memNeeded; - tObj = ctx->Texture.Proxy2D; + tObj = ctx->Texture.ProxyTex[TEXTURE_2D_INDEX]; if (!tObj->DriverData) tObj->DriverData = fxAllocTexObjData(fxMesa); ti = TDFX_TEXTURE_DATA(tObj); diff --git a/src/mesa/main/colortab.c b/src/mesa/main/colortab.c index 610acba306..97120398f9 100644 --- a/src/mesa/main/colortab.c +++ b/src/mesa/main/colortab.c @@ -1,6 +1,6 @@ /* * Mesa 3-D graphics library - * Version: 6.5.3 + * Version: 7.1 * * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. * @@ -30,6 +30,7 @@ #include "image.h" #include "macros.h" #include "state.h" +#include "teximage.h" /** @@ -305,49 +306,6 @@ _mesa_ColorTable( GLenum target, GLenum internalFormat, ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); /* too complex */ switch (target) { - case GL_TEXTURE_1D: - texObj = texUnit->Current1D; - table = &texObj->Palette; - break; - case GL_TEXTURE_2D: - texObj = texUnit->Current2D; - table = &texObj->Palette; - break; - case GL_TEXTURE_3D: - texObj = texUnit->Current3D; - table = &texObj->Palette; - break; - case GL_TEXTURE_CUBE_MAP_ARB: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error(ctx, GL_INVALID_ENUM, "glColorTable(target)"); - return; - } - texObj = texUnit->CurrentCubeMap; - table = &texObj->Palette; - break; - case GL_PROXY_TEXTURE_1D: - texObj = ctx->Texture.Proxy1D; - table = &texObj->Palette; - proxy = GL_TRUE; - break; - case GL_PROXY_TEXTURE_2D: - texObj = ctx->Texture.Proxy2D; - table = &texObj->Palette; - proxy = GL_TRUE; - break; - case GL_PROXY_TEXTURE_3D: - texObj = ctx->Texture.Proxy3D; - table = &texObj->Palette; - proxy = GL_TRUE; - break; - case GL_PROXY_TEXTURE_CUBE_MAP_ARB: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error(ctx, GL_INVALID_ENUM, "glColorTable(target)"); - return; - } - texObj = ctx->Texture.ProxyCubeMap; - table = &texObj->Palette; - break; case GL_SHARED_TEXTURE_PALETTE_EXT: table = &ctx->Texture.Palette; break; @@ -396,8 +354,19 @@ _mesa_ColorTable( GLenum target, GLenum internalFormat, proxy = GL_TRUE; break; default: - _mesa_error(ctx, GL_INVALID_ENUM, "glColorTable(target)"); - return; + /* try texture targets */ + { + struct gl_texture_object *texobj + = _mesa_select_tex_object(ctx, texUnit, target); + if (texobj) { + table = &texobj->Palette; + proxy = _mesa_is_proxy_texture(target); + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glColorTable(target)"); + return; + } + } } assert(table); @@ -499,26 +468,6 @@ _mesa_ColorSubTable( GLenum target, GLsizei start, ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); switch (target) { - case GL_TEXTURE_1D: - texObj = texUnit->Current1D; - table = &texObj->Palette; - break; - case GL_TEXTURE_2D: - texObj = texUnit->Current2D; - table = &texObj->Palette; - break; - case GL_TEXTURE_3D: - texObj = texUnit->Current3D; - table = &texObj->Palette; - break; - case GL_TEXTURE_CUBE_MAP_ARB: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error(ctx, GL_INVALID_ENUM, "glColorSubTable(target)"); - return; - } - texObj = texUnit->CurrentCubeMap; - table = &texObj->Palette; - break; case GL_SHARED_TEXTURE_PALETTE_EXT: table = &ctx->Texture.Palette; break; @@ -547,8 +496,15 @@ _mesa_ColorSubTable( GLenum target, GLsizei start, bias = ctx->Pixel.ColorTableBias[COLORTABLE_POSTCOLORMATRIX]; break; default: - _mesa_error(ctx, GL_INVALID_ENUM, "glColorSubTable(target)"); - return; + /* try texture targets */ + texObj = _mesa_select_tex_object(ctx, texUnit, target); + if (texObj && !_mesa_is_proxy_texture(target)) { + table = &texObj->Palette; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glColorSubTable(target)"); + return; + } } assert(table); @@ -636,22 +592,6 @@ _mesa_GetColorTable( GLenum target, GLenum format, } switch (target) { - case GL_TEXTURE_1D: - table = &texUnit->Current1D->Palette; - break; - case GL_TEXTURE_2D: - table = &texUnit->Current2D->Palette; - break; - case GL_TEXTURE_3D: - table = &texUnit->Current3D->Palette; - break; - case GL_TEXTURE_CUBE_MAP_ARB: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTable(target)"); - return; - } - table = &texUnit->CurrentCubeMap->Palette; - break; case GL_SHARED_TEXTURE_PALETTE_EXT: table = &ctx->Texture.Palette; break; @@ -672,8 +612,18 @@ _mesa_GetColorTable( GLenum target, GLenum format, table = &ctx->ColorTable[COLORTABLE_POSTCOLORMATRIX]; break; default: - _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTable(target)"); - return; + /* try texture targets */ + { + struct gl_texture_object *texobj + = _mesa_select_tex_object(ctx, texUnit, target); + if (texobj && !_mesa_is_proxy_texture(target)) { + table = &texobj->Palette; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTable(target)"); + return; + } + } } ASSERT(table); @@ -781,65 +731,41 @@ _mesa_GetColorTable( GLenum target, GLenum format, void GLAPIENTRY _mesa_ColorTableParameterfv(GLenum target, GLenum pname, const GLfloat *params) { + GLfloat *scale, *bias; GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); switch (target) { - case GL_COLOR_TABLE_SGI: - if (pname == GL_COLOR_TABLE_SCALE_SGI) { - COPY_4V(ctx->Pixel.ColorTableScale[COLORTABLE_PRECONVOLUTION], params); - } - else if (pname == GL_COLOR_TABLE_BIAS_SGI) { - COPY_4V(ctx->Pixel.ColorTableBias[COLORTABLE_PRECONVOLUTION], params); - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameterfv(pname)"); - return; - } - break; - case GL_TEXTURE_COLOR_TABLE_SGI: - if (!ctx->Extensions.SGI_texture_color_table) { - _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameter(target)"); - return; - } - if (pname == GL_COLOR_TABLE_SCALE_SGI) { - COPY_4V(ctx->Pixel.TextureColorTableScale, params); - } - else if (pname == GL_COLOR_TABLE_BIAS_SGI) { - COPY_4V(ctx->Pixel.TextureColorTableBias, params); - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameterfv(pname)"); - return; - } - break; - case GL_POST_CONVOLUTION_COLOR_TABLE_SGI: - if (pname == GL_COLOR_TABLE_SCALE_SGI) { - COPY_4V(ctx->Pixel.ColorTableScale[COLORTABLE_POSTCONVOLUTION], params); - } - else if (pname == GL_COLOR_TABLE_BIAS_SGI) { - COPY_4V(ctx->Pixel.ColorTableBias[COLORTABLE_POSTCONVOLUTION], params); - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameterfv(pname)"); - return; - } - break; - case GL_POST_COLOR_MATRIX_COLOR_TABLE_SGI: - if (pname == GL_COLOR_TABLE_SCALE_SGI) { - COPY_4V(ctx->Pixel.ColorTableScale[COLORTABLE_POSTCOLORMATRIX], params); - } - else if (pname == GL_COLOR_TABLE_BIAS_SGI) { - COPY_4V(ctx->Pixel.ColorTableBias[COLORTABLE_POSTCOLORMATRIX], params); - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameterfv(pname)"); - return; - } - break; - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameter(target)"); - return; + case GL_COLOR_TABLE_SGI: + scale = ctx->Pixel.ColorTableScale[COLORTABLE_PRECONVOLUTION]; + bias = ctx->Pixel.ColorTableBias[COLORTABLE_PRECONVOLUTION]; + break; + case GL_TEXTURE_COLOR_TABLE_SGI: + scale = ctx->Pixel.TextureColorTableScale; + bias = ctx->Pixel.TextureColorTableBias; + break; + case GL_POST_CONVOLUTION_COLOR_TABLE_SGI: + scale = ctx->Pixel.ColorTableScale[COLORTABLE_POSTCONVOLUTION]; + bias = ctx->Pixel.ColorTableBias[COLORTABLE_POSTCONVOLUTION]; + break; + case GL_POST_COLOR_MATRIX_COLOR_TABLE_SGI: + scale = ctx->Pixel.ColorTableScale[COLORTABLE_POSTCOLORMATRIX]; + bias = ctx->Pixel.ColorTableBias[COLORTABLE_POSTCOLORMATRIX]; + break; + default: + _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameter(target)"); + return; + } + + if (pname == GL_COLOR_TABLE_SCALE_SGI) { + COPY_4V(scale, params); + } + else if (pname == GL_COLOR_TABLE_BIAS_SGI) { + COPY_4V(bias, params); + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glColorTableParameterfv(pname)"); + return; } ctx->NewState |= _NEW_PIXEL; @@ -879,40 +805,6 @@ _mesa_GetColorTableParameterfv( GLenum target, GLenum pname, GLfloat *params ) ASSERT_OUTSIDE_BEGIN_END(ctx); switch (target) { - case GL_TEXTURE_1D: - table = &texUnit->Current1D->Palette; - break; - case GL_TEXTURE_2D: - table = &texUnit->Current2D->Palette; - break; - case GL_TEXTURE_3D: - table = &texUnit->Current3D->Palette; - break; - case GL_TEXTURE_CUBE_MAP_ARB: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetColorTableParameterfv(target)"); - return; - } - table = &texUnit->CurrentCubeMap->Palette; - break; - case GL_PROXY_TEXTURE_1D: - table = &ctx->Texture.Proxy1D->Palette; - break; - case GL_PROXY_TEXTURE_2D: - table = &ctx->Texture.Proxy2D->Palette; - break; - case GL_PROXY_TEXTURE_3D: - table = &ctx->Texture.Proxy3D->Palette; - break; - case GL_PROXY_TEXTURE_CUBE_MAP_ARB: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetColorTableParameterfv(target)"); - return; - } - table = &ctx->Texture.ProxyCubeMap->Palette; - break; case GL_SHARED_TEXTURE_PALETTE_EXT: table = &ctx->Texture.Palette; break; @@ -981,8 +873,19 @@ _mesa_GetColorTableParameterfv( GLenum target, GLenum pname, GLfloat *params ) table = &ctx->ProxyColorTable[COLORTABLE_POSTCOLORMATRIX]; break; default: - _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTableParameterfv(target)"); - return; + /* try texture targets */ + { + struct gl_texture_object *texobj + = _mesa_select_tex_object(ctx, texUnit, target); + if (texobj) { + table = &texobj->Palette; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetColorTableParameterfv(target)"); + return; + } + } } assert(table); @@ -1029,40 +932,6 @@ _mesa_GetColorTableParameteriv( GLenum target, GLenum pname, GLint *params ) ASSERT_OUTSIDE_BEGIN_END(ctx); switch (target) { - case GL_TEXTURE_1D: - table = &texUnit->Current1D->Palette; - break; - case GL_TEXTURE_2D: - table = &texUnit->Current2D->Palette; - break; - case GL_TEXTURE_3D: - table = &texUnit->Current3D->Palette; - break; - case GL_TEXTURE_CUBE_MAP_ARB: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetColorTableParameteriv(target)"); - return; - } - table = &texUnit->CurrentCubeMap->Palette; - break; - case GL_PROXY_TEXTURE_1D: - table = &ctx->Texture.Proxy1D->Palette; - break; - case GL_PROXY_TEXTURE_2D: - table = &ctx->Texture.Proxy2D->Palette; - break; - case GL_PROXY_TEXTURE_3D: - table = &ctx->Texture.Proxy3D->Palette; - break; - case GL_PROXY_TEXTURE_CUBE_MAP_ARB: - if (!ctx->Extensions.ARB_texture_cube_map) { - _mesa_error(ctx, GL_INVALID_ENUM, - "glGetColorTableParameteriv(target)"); - return; - } - table = &ctx->Texture.ProxyCubeMap->Palette; - break; case GL_SHARED_TEXTURE_PALETTE_EXT: table = &ctx->Texture.Palette; break; @@ -1161,8 +1030,19 @@ _mesa_GetColorTableParameteriv( GLenum target, GLenum pname, GLint *params ) table = &ctx->ProxyColorTable[COLORTABLE_POSTCOLORMATRIX]; break; default: - _mesa_error(ctx, GL_INVALID_ENUM, "glGetColorTableParameteriv(target)"); - return; + /* Try texture targets */ + { + struct gl_texture_object *texobj + = _mesa_select_tex_object(ctx, texUnit, target); + if (texobj) { + table = &texobj->Palette; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, + "glGetColorTableParameteriv(target)"); + return; + } + } } assert(table); diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index a38ec02852..2d6be1983c 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -1579,6 +1579,7 @@ struct gl_texture_attrib struct gl_texture_unit Unit[MAX_TEXTURE_UNITS]; +#if 0 struct gl_texture_object *Proxy1D; struct gl_texture_object *Proxy2D; struct gl_texture_object *Proxy3D; @@ -1586,6 +1587,9 @@ struct gl_texture_attrib struct gl_texture_object *ProxyRect; struct gl_texture_object *Proxy1DArray; struct gl_texture_object *Proxy2DArray; +#else + struct gl_texture_object *ProxyTex[NUM_TEXTURE_TARGETS]; +#endif /** GL_EXT_shared_texture_palette */ GLboolean SharedPalette; diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c index eed1937517..36557f19fd 100644 --- a/src/mesa/main/teximage.c +++ b/src/mesa/main/teximage.c @@ -767,15 +767,15 @@ _mesa_select_tex_object(GLcontext *ctx, const struct gl_texture_unit *texUnit, case GL_TEXTURE_1D: return texUnit->Current1D; case GL_PROXY_TEXTURE_1D: - return ctx->Texture.Proxy1D; + return ctx->Texture.ProxyTex[TEXTURE_1D_INDEX]; case GL_TEXTURE_2D: return texUnit->Current2D; case GL_PROXY_TEXTURE_2D: - return ctx->Texture.Proxy2D; + return ctx->Texture.ProxyTex[TEXTURE_2D_INDEX]; case GL_TEXTURE_3D: return texUnit->Current3D; case GL_PROXY_TEXTURE_3D: - return ctx->Texture.Proxy3D; + return ctx->Texture.ProxyTex[TEXTURE_3D_INDEX]; case GL_TEXTURE_CUBE_MAP_POSITIVE_X_ARB: case GL_TEXTURE_CUBE_MAP_NEGATIVE_X_ARB: case GL_TEXTURE_CUBE_MAP_POSITIVE_Y_ARB: @@ -787,25 +787,25 @@ _mesa_select_tex_object(GLcontext *ctx, const struct gl_texture_unit *texUnit, ? texUnit->CurrentCubeMap : NULL; case GL_PROXY_TEXTURE_CUBE_MAP_ARB: return ctx->Extensions.ARB_texture_cube_map - ? ctx->Texture.ProxyCubeMap : NULL; + ? ctx->Texture.ProxyTex[TEXTURE_CUBE_INDEX] : NULL; case GL_TEXTURE_RECTANGLE_NV: return ctx->Extensions.NV_texture_rectangle ? texUnit->CurrentRect : NULL; case GL_PROXY_TEXTURE_RECTANGLE_NV: return ctx->Extensions.NV_texture_rectangle - ? ctx->Texture.ProxyRect : NULL; + ? ctx->Texture.ProxyTex[TEXTURE_RECT_INDEX] : NULL; case GL_TEXTURE_1D_ARRAY_EXT: return ctx->Extensions.MESA_texture_array ? texUnit->Current1DArray : NULL; case GL_PROXY_TEXTURE_1D_ARRAY_EXT: return ctx->Extensions.MESA_texture_array - ? ctx->Texture.Proxy1DArray : NULL; + ? ctx->Texture.ProxyTex[TEXTURE_1D_ARRAY_INDEX] : NULL; case GL_TEXTURE_2D_ARRAY_EXT: return ctx->Extensions.MESA_texture_array ? texUnit->Current2DArray : NULL; case GL_PROXY_TEXTURE_2D_ARRAY_EXT: return ctx->Extensions.MESA_texture_array - ? ctx->Texture.Proxy2DArray : NULL; + ? ctx->Texture.ProxyTex[TEXTURE_2D_ARRAY_INDEX] : NULL; default: _mesa_problem(NULL, "bad target in _mesa_select_tex_object()"); return NULL; @@ -932,106 +932,106 @@ _mesa_get_proxy_tex_image(GLcontext *ctx, GLenum target, GLint level) case GL_PROXY_TEXTURE_1D: if (level >= ctx->Const.MaxTextureLevels) return NULL; - texImage = ctx->Texture.Proxy1D->Image[0][level]; + texImage = ctx->Texture.ProxyTex[TEXTURE_1D_INDEX]->Image[0][level]; if (!texImage) { texImage = ctx->Driver.NewTextureImage(ctx); if (!texImage) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation"); return NULL; } - ctx->Texture.Proxy1D->Image[0][level] = texImage; + ctx->Texture.ProxyTex[TEXTURE_1D_INDEX]->Image[0][level] = texImage; /* Set the 'back' pointer */ - texImage->TexObject = ctx->Texture.Proxy1D; + texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_1D_INDEX]; } return texImage; case GL_PROXY_TEXTURE_2D: if (level >= ctx->Const.MaxTextureLevels) return NULL; - texImage = ctx->Texture.Proxy2D->Image[0][level]; + texImage = ctx->Texture.ProxyTex[TEXTURE_2D_INDEX]->Image[0][level]; if (!texImage) { texImage = ctx->Driver.NewTextureImage(ctx); if (!texImage) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation"); return NULL; } - ctx->Texture.Proxy2D->Image[0][level] = texImage; + ctx->Texture.ProxyTex[TEXTURE_2D_INDEX]->Image[0][level] = texImage; /* Set the 'back' pointer */ - texImage->TexObject = ctx->Texture.Proxy2D; + texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_2D_INDEX]; } return texImage; case GL_PROXY_TEXTURE_3D: if (level >= ctx->Const.Max3DTextureLevels) return NULL; - texImage = ctx->Texture.Proxy3D->Image[0][level]; + texImage = ctx->Texture.ProxyTex[TEXTURE_3D_INDEX]->Image[0][level]; if (!texImage) { texImage = ctx->Driver.NewTextureImage(ctx); if (!texImage) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation"); return NULL; } - ctx->Texture.Proxy3D->Image[0][level] = texImage; + ctx->Texture.ProxyTex[TEXTURE_3D_INDEX]->Image[0][level] = texImage; /* Set the 'back' pointer */ - texImage->TexObject = ctx->Texture.Proxy3D; + texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_3D_INDEX]; } return texImage; case GL_PROXY_TEXTURE_CUBE_MAP: if (level >= ctx->Const.MaxCubeTextureLevels) return NULL; - texImage = ctx->Texture.ProxyCubeMap->Image[0][level]; + texImage = ctx->Texture.ProxyTex[TEXTURE_CUBE_INDEX]->Image[0][level]; if (!texImage) { texImage = ctx->Driver.NewTextureImage(ctx); if (!texImage) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation"); return NULL; } - ctx->Texture.ProxyCubeMap->Image[0][level] = texImage; + ctx->Texture.ProxyTex[TEXTURE_CUBE_INDEX]->Image[0][level] = texImage; /* Set the 'back' pointer */ - texImage->TexObject = ctx->Texture.ProxyCubeMap; + texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_CUBE_INDEX]; } return texImage; case GL_PROXY_TEXTURE_RECTANGLE_NV: if (level > 0) return NULL; - texImage = ctx->Texture.ProxyRect->Image[0][level]; + texImage = ctx->Texture.ProxyTex[TEXTURE_RECT_INDEX]->Image[0][level]; if (!texImage) { texImage = ctx->Driver.NewTextureImage(ctx); if (!texImage) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation"); return NULL; } - ctx->Texture.ProxyRect->Image[0][level] = texImage; + ctx->Texture.ProxyTex[TEXTURE_RECT_INDEX]->Image[0][level] = texImage; /* Set the 'back' pointer */ - texImage->TexObject = ctx->Texture.ProxyRect; + texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_RECT_INDEX]; } return texImage; case GL_PROXY_TEXTURE_1D_ARRAY_EXT: if (level >= ctx->Const.MaxTextureLevels) return NULL; - texImage = ctx->Texture.Proxy1DArray->Image[0][level]; + texImage = ctx->Texture.ProxyTex[TEXTURE_1D_ARRAY_INDEX]->Image[0][level]; if (!texImage) { texImage = ctx->Driver.NewTextureImage(ctx); if (!texImage) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation"); return NULL; } - ctx->Texture.Proxy1DArray->Image[0][level] = texImage; + ctx->Texture.ProxyTex[TEXTURE_1D_ARRAY_INDEX]->Image[0][level] = texImage; /* Set the 'back' pointer */ - texImage->TexObject = ctx->Texture.Proxy1DArray; + texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_1D_ARRAY_INDEX]; } return texImage; case GL_PROXY_TEXTURE_2D_ARRAY_EXT: if (level >= ctx->Const.MaxTextureLevels) return NULL; - texImage = ctx->Texture.Proxy2DArray->Image[0][level]; + texImage = ctx->Texture.ProxyTex[TEXTURE_2D_ARRAY_INDEX]->Image[0][level]; if (!texImage) { texImage = ctx->Driver.NewTextureImage(ctx); if (!texImage) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "proxy texture allocation"); return NULL; } - ctx->Texture.Proxy2DArray->Image[0][level] = texImage; + ctx->Texture.ProxyTex[TEXTURE_2D_ARRAY_INDEX]->Image[0][level] = texImage; /* Set the 'back' pointer */ - texImage->TexObject = ctx->Texture.Proxy2DArray; + texImage->TexObject = ctx->Texture.ProxyTex[TEXTURE_2D_ARRAY_INDEX]; } return texImage; default: @@ -2611,7 +2611,7 @@ _mesa_TexImage2D( GLenum target, GLint level, GLint internalFormat, 1, border)) { /* when error, clear all proxy texture image parameters */ if (texImage) - clear_teximage_fields(ctx->Texture.Proxy2D->Image[0][level]); + clear_teximage_fields(ctx->Texture.ProxyTex[TEXTURE_2D_INDEX]->Image[0][level]); } else { /* no error, set the tex image parameters */ diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c index 3bdb55257f..20f9c4512c 100644 --- a/src/mesa/main/texstate.c +++ b/src/mesa/main/texstate.c @@ -674,54 +674,32 @@ _mesa_update_texture( GLcontext *ctx, GLuint new_state ) static GLboolean alloc_proxy_textures( GLcontext *ctx ) { - ctx->Texture.Proxy1D = (*ctx->Driver.NewTextureObject)(ctx, 0, GL_TEXTURE_1D); - if (!ctx->Texture.Proxy1D) - goto cleanup; - - ctx->Texture.Proxy2D = (*ctx->Driver.NewTextureObject)(ctx, 0, GL_TEXTURE_2D); - if (!ctx->Texture.Proxy2D) - goto cleanup; - - ctx->Texture.Proxy3D = (*ctx->Driver.NewTextureObject)(ctx, 0, GL_TEXTURE_3D); - if (!ctx->Texture.Proxy3D) - goto cleanup; - - ctx->Texture.ProxyCubeMap = (*ctx->Driver.NewTextureObject)(ctx, 0, GL_TEXTURE_CUBE_MAP_ARB); - if (!ctx->Texture.ProxyCubeMap) - goto cleanup; - - ctx->Texture.ProxyRect = (*ctx->Driver.NewTextureObject)(ctx, 0, GL_TEXTURE_RECTANGLE_NV); - if (!ctx->Texture.ProxyRect) - goto cleanup; - - ctx->Texture.Proxy1DArray = (*ctx->Driver.NewTextureObject)(ctx, 0, GL_TEXTURE_1D_ARRAY_EXT); - if (!ctx->Texture.Proxy1DArray) - goto cleanup; - - ctx->Texture.Proxy2DArray = (*ctx->Driver.NewTextureObject)(ctx, 0, GL_TEXTURE_2D_ARRAY_EXT); - if (!ctx->Texture.Proxy2DArray) - goto cleanup; - - assert(ctx->Texture.Proxy1D->RefCount == 1); + static const GLenum targets[] = { + GL_TEXTURE_1D, + GL_TEXTURE_2D, + GL_TEXTURE_3D, + GL_TEXTURE_CUBE_MAP_ARB, + GL_TEXTURE_RECTANGLE_NV, + GL_TEXTURE_1D_ARRAY_EXT, + GL_TEXTURE_2D_ARRAY_EXT + }; + GLint tgt; + + ASSERT(Elements(targets) == NUM_TEXTURE_TARGETS); + + for (tgt = 0; tgt < NUM_TEXTURE_TARGETS; tgt++) { + if (!(ctx->Texture.ProxyTex[tgt] + = ctx->Driver.NewTextureObject(ctx, 0, targets[tgt]))) { + /* out of memory, free what we did allocate */ + while (--tgt >= 0) { + ctx->Driver.DeleteTexture(ctx, ctx->Texture.ProxyTex[tgt]); + } + return GL_FALSE; + } + } + assert(ctx->Texture.ProxyTex[0]->RefCount == 1); /* sanity check */ return GL_TRUE; - - cleanup: - if (ctx->Texture.Proxy1D) - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy1D); - if (ctx->Texture.Proxy2D) - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy2D); - if (ctx->Texture.Proxy3D) - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy3D); - if (ctx->Texture.ProxyCubeMap) - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.ProxyCubeMap); - if (ctx->Texture.ProxyRect) - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.ProxyRect); - if (ctx->Texture.Proxy1DArray) - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy1DArray); - if (ctx->Texture.Proxy2DArray) - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy2DArray); - return GL_FALSE; } @@ -815,7 +793,7 @@ _mesa_init_texture(GLcontext *ctx) void _mesa_free_texture_data(GLcontext *ctx) { - GLuint u; + GLuint u, tgt; /* unreference current textures */ for (u = 0; u < MAX_TEXTURE_IMAGE_UNITS; u++) { @@ -830,13 +808,8 @@ _mesa_free_texture_data(GLcontext *ctx) } /* Free proxy texture objects */ - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy1D ); - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy2D ); - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy3D ); - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.ProxyCubeMap ); - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.ProxyRect ); - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy1DArray ); - (ctx->Driver.DeleteTexture)(ctx, ctx->Texture.Proxy2DArray ); + for (tgt = 0; tgt < NUM_TEXTURE_TARGETS; tgt++) + ctx->Driver.DeleteTexture(ctx, ctx->Texture.ProxyTex[tgt]); #if FEATURE_colortable -- cgit v1.2.3 From f6e8963bb152166c28dae896a0c8c83ce554f987 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 4 Jul 2008 19:20:08 +0200 Subject: i915: Clean up intel_egl.c a bit --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 24 +++--------------------- 1 file changed, 3 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 6865bfa3d4..cc4fb247d5 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -33,8 +33,8 @@ egl_drm_create_device(int drmFD) device->drmFD = drmFD; if (!intel_init_driver(device)) { - printf("EGL: failed to initalize device\n"); free(device); + return NULL; } return device; @@ -105,22 +105,9 @@ drm_add_modes_from_connector(_EGLScreen *screen, drmModeConnectorPtr connector) } } -static void -print_modes(drmModeConnectorPtr connector) -{ - struct drm_mode_modeinfo *m; - int i; - - for (i = 0; i < connector->count_modes; i++) { - m = &connector->modes[i]; - printf("dfm %p %i %i %i\n", m, m->hdisplay, m->vdisplay, m->vrefresh); - } -} - static EGLBoolean drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) { - printf("%s enter\n", __FUNCTION__); _EGLDisplay *disp = _eglLookupDisplay(dpy); struct drm_driver *drm_drv = (struct drm_driver *)drv; struct drm_screen *screen = NULL; @@ -490,10 +477,8 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, DRM_BO_FLAG_NO_EVICT, DRM_BO_HINT_DONT_FENCE, &scrn->buffer); - if (ret) { - printf("failed to create framebuffer (ret %d)\n", ret); + if (ret) return EGL_FALSE; - } prettyColors(drm_drv->device->drmFD, scrn->buffer.handle, pitch); @@ -510,10 +495,8 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, goto err_bo; scrn->mode = drm_find_mode(scrn->connector, mode); - if (!scrn->mode) { - printf("oh noes, no matching mode found\n"); + if (!scrn->mode) goto err_fb; - } ret = drmModeSetCrtc( drm_drv->device->drmFD, @@ -602,7 +585,6 @@ _EGLDriver * _eglMain(_EGLDisplay *dpy, const char *args) { struct drm_driver *drm; - printf("%s enter\n", __FUNCTION__); drm = (struct drm_driver *) calloc(1, sizeof(struct drm_driver)); if (!drm) { -- cgit v1.2.3 From e6c24539c3d482271a8c1bcba697b08e24344e05 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 4 Jul 2008 19:23:52 +0200 Subject: i915: Added debug filling code of texture, not active --- src/gallium/drivers/i915simple/i915_texture.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 5d6769924c..cf4964b26b 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -581,6 +581,7 @@ i915_texture_create(struct pipe_screen *screen, struct i915_screen *i915screen = i915_screen(screen); struct pipe_winsys *ws = screen->winsys; struct i915_texture *tex = CALLOC_STRUCT(i915_texture); + size_t tex_size; if (!tex) return NULL; @@ -600,14 +601,22 @@ i915_texture_create(struct pipe_screen *screen, goto fail; } + tex_size = tex->stride * tex->total_nblocksy; + tex->buffer = ws->buffer_create(ws, 64, PIPE_BUFFER_USAGE_PIXEL, - tex->stride * - tex->total_nblocksy); + tex_size); if (!tex->buffer) goto fail; +#if 0 + void *ptr = ws->buffer_map(ws, tex->buffer, + PIPE_BUFFER_USAGE_CPU_WRITE); + memset(ptr, 0x80, tex_size); + ws->buffer_unmap(ws, tex->buffer); +#endif + return &tex->base; fail: -- cgit v1.2.3 From c23b64f1646ac5349c395ede47707906ddff7b4c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 5 Jul 2008 03:19:56 +0900 Subject: softpipe: Compute block size for display targets. --- src/gallium/drivers/softpipe/sp_texture.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 20ad336b4f..f775591352 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -122,8 +122,10 @@ softpipe_displaytarget_layout(struct pipe_screen *screen, /* Now extract the goodies: */ - spt->buffer = surf.buffer; + spt->base.nblocksx[0] = pf_get_nblocksx(&spt->base.block, spt->base.width[0]); + spt->base.nblocksy[0] = pf_get_nblocksy(&spt->base.block, spt->base.height[0]); spt->stride[0] = surf.stride; + spt->buffer = surf.buffer; return spt->buffer != NULL; } -- cgit v1.2.3 From 152ed98b84acf500f4b5122cf3fde82c6e5206f2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 5 Jul 2008 03:21:27 +0900 Subject: softpipe: Prevent NULL ptr derreference on takedown. --- src/gallium/drivers/softpipe/sp_state_fs.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 24b91fbc79..901c8f83e7 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -154,8 +154,8 @@ softpipe_set_constant_buffer(struct pipe_context *pipe, /* note: reference counting */ pipe_buffer_reference(ws, &softpipe->constants[shader].buffer, - buf->buffer); - softpipe->constants[shader].size = buf->size; + buf ? buf->buffer : NULL); + softpipe->constants[shader].size = buf ? buf->size : 0; softpipe->dirty |= SP_NEW_CONSTANTS; } -- cgit v1.2.3 From 74db8e9b3f354b5afa30f7cd52a511504693a3c8 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Sat, 5 Jul 2008 14:04:02 +0200 Subject: i915: Fix haveDepth and haveStencil for visual in EGL --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index cc4fb247d5..a008d952bf 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -224,7 +224,9 @@ visual_from_config(_EGLConfig *conf) visual->doubleBufferMode = 1; visual->depthBits = 24; + visual->haveDepthBuffer = visual->depthBits > 0; visual->stencilBits = 8; + visual->haveStencilBuffer = visual->stencilBits > 0; return visual; } -- cgit v1.2.3 From c6d6a57424b77ff7de611333046c054d1948ae7a Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Sat, 5 Jul 2008 15:58:42 +0200 Subject: i915: Added a intel be function to wrap a drm bo handle --- .../winsys/common/intel_drm/intel_be_device.c | 33 ++++++++++++++++++++++ .../winsys/common/intel_drm/intel_be_device.h | 13 +++++++-- src/gallium/winsys/dri/intel/intel_screen.c | 22 ++++++--------- 3 files changed, 52 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/common/intel_drm/intel_be_device.c b/src/gallium/winsys/common/intel_drm/intel_be_device.c index efb57a394e..8db0329615 100644 --- a/src/gallium/winsys/common/intel_drm/intel_be_device.c +++ b/src/gallium/winsys/common/intel_drm/intel_be_device.c @@ -125,6 +125,37 @@ intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned byte return &buffer->base; } +struct pipe_buffer * +intel_be_buffer_from_handle(struct intel_be_device *device, + const char* name, unsigned handle) +{ + struct intel_be_buffer *be_buf = malloc(sizeof(*be_buf)); + struct pipe_buffer *buffer; + + if (!be_buf) + goto err; + + memset(be_buf, 0, sizeof(*be_buf)); + + driGenBuffers(device->staticPool, name, 1, &be_buf->driBO, 0, 0, 0); + driBOSetReferenced(be_buf->driBO, handle); + + if (0) /** XXX TODO check error */ + goto err_bo; + + buffer = &be_buf->base; + buffer->refcount = 1; + buffer->alignment = 0; + buffer->usage = 0; + buffer->size = driBOSize(be_buf->driBO); + + return buffer; +err_bo: + free(be_buf); +err: + return NULL; +} + /* * Surface functions. @@ -157,6 +188,7 @@ intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) assert((size_t)"intel_i915_surface_release is deprecated" & 0); } + /* * Fence functions */ @@ -189,6 +221,7 @@ intel_be_fence_finish( struct pipe_winsys *sws, return driFenceFinish((struct _DriFenceObject *)fence, flag, 0); } + /* * Misc functions */ diff --git a/src/gallium/winsys/common/intel_drm/intel_be_device.h b/src/gallium/winsys/common/intel_drm/intel_be_device.h index abf0253917..3f8b3f585c 100644 --- a/src/gallium/winsys/common/intel_drm/intel_be_device.h +++ b/src/gallium/winsys/common/intel_drm/intel_be_device.h @@ -48,14 +48,23 @@ struct intel_be_buffer { struct _DriBufferObject *driBO; }; +/** + * Create a be buffer from a drm bo handle + * + * Takes a reference + */ +struct pipe_buffer * +intel_be_buffer_from_handle(struct intel_be_device *device, + const char* name, unsigned handle); + static INLINE struct intel_be_buffer * -intel_be_buffer( struct pipe_buffer *buf ) +intel_be_buffer(struct pipe_buffer *buf) { return (struct intel_be_buffer *)buf; } static INLINE struct _DriBufferObject * -dri_bo( struct pipe_buffer *buf ) +dri_bo(struct pipe_buffer *buf) { return intel_be_buffer(buf)->driBO; } diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index cfecebdb8c..f2412217f3 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -49,28 +49,22 @@ intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, static void intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle) { - struct intel_be_buffer *be_buf = malloc(sizeof(*be_buf)); struct pipe_screen *screen = intelScreen->base.screen; struct pipe_texture *texture; struct pipe_texture templat; struct pipe_surface *surface; - struct pipe_buffer *buffer = &be_buf->base; + struct pipe_buffer *buffer; unsigned pitch; assert(intelScreen->front.cpp == 4); - /* XXX create a intel_be function for this */ - { - driGenBuffers(intelScreen->base.staticPool, "front", 1, &intelScreen->front.buffer, 0, 0, 0); - driBOSetReferenced(intelScreen->front.buffer, handle); - - memset(be_buf, 0, sizeof(*be_buf)); - buffer->refcount = 1; - buffer->alignment = 0; - buffer->usage = 0; - buffer->size = driBOSize(intelScreen->front.buffer); - be_buf->driBO = intelScreen->front.buffer; - } + buffer = intel_be_buffer_from_handle(&intelScreen->base, + "front", handle); + + if (!buffer) + return; + + intelScreen->front.buffer = dri_bo(buffer); memset(&templat, 0, sizeof(templat)); templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET; -- cgit v1.2.3 From b30fb6d54019c84174379a773121438ba943557a Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Sat, 5 Jul 2008 16:46:04 +0200 Subject: i915: Fix EGL version and name --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index a008d952bf..e078b9eda1 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -171,7 +171,7 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) drv->Initialized = EGL_TRUE; *major = 1; - *minor = 0; + *minor = 4; return EGL_TRUE; } @@ -610,6 +610,7 @@ _eglMain(_EGLDisplay *dpy, const char *args) drm->base.API.SwapBuffers = drm_swap_buffers; drm->base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; + drm->base.Name = "DRM/Gallium"; /* enable supported extensions */ drm->base.Extensions.MESA_screen_surface = EGL_TRUE; -- cgit v1.2.3 From 1315f720ba80596b866103a7e153bfc2f2bd267b Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 7 Jul 2008 13:24:09 +0200 Subject: i915: Destruction of surface, context and device in EGL --- src/gallium/winsys/egl_drm/intel/intel_context.c | 14 +++++++++++ src/gallium/winsys/egl_drm/intel/intel_egl.c | 21 ++++++++++++---- src/gallium/winsys/egl_drm/intel/intel_egl.h | 8 +++++- src/gallium/winsys/egl_drm/intel/intel_screen.c | 31 +++++++++++++++++++++--- 4 files changed, 65 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.c b/src/gallium/winsys/egl_drm/intel/intel_context.c index 7205fa4483..7513f3feef 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_context.c +++ b/src/gallium/winsys/egl_drm/intel/intel_context.c @@ -178,6 +178,20 @@ intel_create_context(struct egl_drm_context *egl_context, const __GLcontextModes return TRUE; } +int +intel_destroy_context(struct egl_drm_context *egl_context) +{ + struct intel_context *intel = egl_context->priv; + + if (intel->intel_screen->dummy == intel) + intel->intel_screen->dummy = NULL; + + st_destroy_context(intel->st); + intel_be_destroy_context(&intel->base); + free(intel); + return TRUE; +} + void intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *draw, struct egl_drm_drawable *read) { diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index e078b9eda1..9e12a024eb 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -32,7 +32,7 @@ egl_drm_create_device(int drmFD) memset(device, 0, sizeof(*device)); device->drmFD = drmFD; - if (!intel_init_driver(device)) { + if (!intel_create_device(device)) { free(device); return NULL; } @@ -180,8 +180,15 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) static EGLBoolean drm_terminate(_EGLDriver *drv, EGLDisplay dpy) { - /* TODO: clean up */ - free(drv); + struct drm_driver *drm_drv = (struct drm_driver *)drv; + + intel_destroy_device(drm_drv->device); + + drmClose(drm_drv->device->drmFD); + + free(drm_drv->device); + free(drm_drv); + return EGL_TRUE; } @@ -303,6 +310,8 @@ drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context) if (fc->base.IsBound) { fc->base.DeletePending = EGL_TRUE; } else { + intel_destroy_context(fc->context); + free(fc->context); free(fc); } return EGL_TRUE; @@ -322,6 +331,7 @@ drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, Nat return EGL_NO_SURFACE; } + static EGLSurface drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, const EGLint *attrib_list) @@ -534,14 +544,15 @@ drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) if (fs->base.IsBound) { fs->base.DeletePending = EGL_TRUE; } else { + intel_bind_frontbuffer(fs->drawable, NULL); + intel_destroy_drawable(fs->drawable); + free(fs->drawable); free(fs); } return EGL_TRUE; } - - static EGLBoolean drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context) { diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.h b/src/gallium/winsys/egl_drm/intel/intel_egl.h index 18e84e8eea..84b6c212bf 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.h +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.h @@ -32,9 +32,15 @@ struct egl_drm_frontbuffer #include "GL/internal/glcore.h" -int intel_init_driver(struct egl_drm_device *device); +int intel_create_device(struct egl_drm_device *device); +int intel_destroy_device(struct egl_drm_device *device); + int intel_create_context(struct egl_drm_context *context, const __GLcontextModes *visual, void *sharedContextPrivate); +int intel_destroy_context(struct egl_drm_context *context); + int intel_create_drawable(struct egl_drm_drawable *drawable, const __GLcontextModes * visual); +int intel_destroy_drawable(struct egl_drm_drawable *drawable); + void intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *draw, struct egl_drm_drawable *read); void intel_swap_buffers(struct egl_drm_drawable *draw); void intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer *front); diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.c b/src/gallium/winsys/egl_drm/intel/intel_screen.c index 2818b9676c..12f605e0be 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_screen.c +++ b/src/gallium/winsys/egl_drm/intel/intel_screen.c @@ -41,7 +41,7 @@ extern const struct dri_extension card_extensions[]; int -intel_init_driver(struct egl_drm_device *device) +intel_create_device(struct egl_drm_device *device) { struct intel_screen *intel_screen; @@ -66,6 +66,19 @@ intel_init_driver(struct egl_drm_device *device) return TRUE; } +int +intel_destroy_device(struct egl_drm_device *device) +{ + struct intel_screen *intel_screen = (struct intel_screen *)device->priv; + + intel_be_destroy_device(&intel_screen->base); + + free(intel_screen); + device->priv = NULL; + + return TRUE; +} + int intel_create_drawable(struct egl_drm_drawable *drawable, const __GLcontextModes * visual) @@ -108,6 +121,18 @@ intel_create_drawable(struct egl_drm_drawable *drawable, return GL_FALSE; } - drawable->priv = (void *) intelfb; - return GL_TRUE; + drawable->priv = (void *) intelfb; + return GL_TRUE; +} + +int +intel_destroy_drawable(struct egl_drm_drawable *drawable) +{ + struct intel_framebuffer *intelfb = (struct intel_framebuffer *)drawable->priv; + drawable->priv = NULL; + + assert(intelfb->stfb); + st_unreference_framebuffer(&intelfb->stfb); + free(intelfb); + return TRUE; } -- cgit v1.2.3 From 687c8d8941bb62fdb4fbdbb0ff5f359851c290b5 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 7 Jul 2008 14:29:11 +0200 Subject: i915: Fetch the real deviceID for EGL --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 30 +++++++++++++++++++++++-- src/gallium/winsys/egl_drm/intel/intel_egl.h | 5 +++++ src/gallium/winsys/egl_drm/intel/intel_screen.c | 3 +-- 3 files changed, 34 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 9e12a024eb..ed96703e37 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -13,6 +13,7 @@ #include "eglmode.h" #include "eglscreen.h" #include "eglsurface.h" +#include "egllog.h" #include "intel_egl.h" @@ -23,15 +24,38 @@ #include "state_tracker/st_public.h" -struct egl_drm_device* egl_drm_create_device(int drmFD); +static void +drm_get_device_id(struct egl_drm_device *device) +{ + char path[512]; + FILE *file; + + /* TODO get the real minor */ + int minor = 0; + + snprintf(path, sizeof(path), "/sys/class/drm/card%d/device/device", minor); + file = fopen(path, "r"); + if (!file) { + _eglLog(_EGL_WARNING, "Could not retrive device ID\n"); + return; + } -struct egl_drm_device* + fgets(path, sizeof( path ), file); + sscanf(path, "%x", &device->deviceID); + fclose(file); +} + +static struct egl_drm_device* egl_drm_create_device(int drmFD) { struct egl_drm_device *device = malloc(sizeof(*device)); memset(device, 0, sizeof(*device)); device->drmFD = drmFD; + device->version = drmGetVersion(device->drmFD); + + drm_get_device_id(device); + if (!intel_create_device(device)) { free(device); return NULL; @@ -105,6 +129,7 @@ drm_add_modes_from_connector(_EGLScreen *screen, drmModeConnectorPtr connector) } } + static EGLBoolean drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) { @@ -183,6 +208,7 @@ drm_terminate(_EGLDriver *drv, EGLDisplay dpy) struct drm_driver *drm_drv = (struct drm_driver *)drv; intel_destroy_device(drm_drv->device); + drmFreeVersion(drm_drv->device->version); drmClose(drm_drv->device->drmFD); diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.h b/src/gallium/winsys/egl_drm/intel/intel_egl.h index 84b6c212bf..1ee27e0847 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.h +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.h @@ -2,10 +2,15 @@ #ifndef _INTEL_EGL_H_ #define _INTEL_EGL_H_ +#include + struct egl_drm_device { void *priv; int drmFD; + + drmVersionPtr version; + int deviceID; }; struct egl_drm_context diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.c b/src/gallium/winsys/egl_drm/intel/intel_screen.c index 12f605e0be..8e6d184440 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_screen.c +++ b/src/gallium/winsys/egl_drm/intel/intel_screen.c @@ -53,8 +53,7 @@ intel_create_device(struct egl_drm_device *device) device->priv = (void *)intel_screen; intel_screen->device = device; - /** TODO JB: ugly hack */ - intel_screen->deviceID = PCI_CHIP_I945_GM; + intel_screen->deviceID = device->deviceID; intel_be_init_device(&intel_screen->base, device->drmFD, PCI_CHIP_I945_GM); -- cgit v1.2.3 From 88b806a0697efa9f7fde6aa59530675908f7294d Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 7 Jul 2008 15:26:47 +0200 Subject: i915: More cleanup of display --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index ed96703e37..4d7e0f4bd5 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -213,6 +213,8 @@ drm_terminate(_EGLDriver *drv, EGLDisplay dpy) drmClose(drm_drv->device->drmFD); free(drm_drv->device); + + _eglCleanupDisplay(drv); free(drm_drv); return EGL_TRUE; -- cgit v1.2.3 From 7420bc05a862db09b8d5f34a7760fa19b578b4d0 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 7 Jul 2008 15:49:09 +0200 Subject: i915: Fix EGL make current --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 4d7e0f4bd5..98928979e7 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -593,12 +593,15 @@ drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface re if (!b) return EGL_FALSE; - /* XXX this is where we'd do the hardware context switch */ - (void) drawSurf; - (void) readSurf; - (void) ctx; + if (ctx) { + if (!drawSurf || !readSurf) + return EGL_FALSE; + + intel_make_current(ctx->context, drawSurf->drawable, readSurf->drawable); + } else { + intel_make_current(NULL, NULL, NULL); + } - intel_make_current(ctx->context, drawSurf->drawable, readSurf->drawable); return EGL_TRUE; } -- cgit v1.2.3 From 3596339d31d05b7d5120290a7b866485c395b5f2 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 7 Jul 2008 15:49:59 +0200 Subject: i915: Fix wrong pointer to cleanup --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 98928979e7..a476490b90 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -214,7 +214,7 @@ drm_terminate(_EGLDriver *drv, EGLDisplay dpy) free(drm_drv->device); - _eglCleanupDisplay(drv); + _eglCleanupDisplay(_eglLookupDisplay(dpy)); free(drm_drv); return EGL_TRUE; -- cgit v1.2.3 From b7dfabcaab020e89644f47ec243b231c21fece63 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Jul 2008 10:08:43 -0600 Subject: gallium: simplify/fix buffer_offset, src_offset computation for VBO arrays Things broke when vertex position wasn't the first attribute. --- src/mesa/state_tracker/st_draw.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 21911774d7..3291387929 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -300,8 +300,8 @@ st_draw_vbo(GLcontext *ctx, vbuffer[attr].buffer = NULL; pipe_reference_buffer(pipe, &vbuffer[attr].buffer, stobj->buffer); - vbuffer[attr].buffer_offset = (unsigned) arrays[0]->Ptr;/* in bytes */ - velements[attr].src_offset = arrays[mesaAttr]->Ptr - arrays[0]->Ptr; + vbuffer[attr].buffer_offset = (unsigned) arrays[mesaAttr]->Ptr; + velements[attr].src_offset = 0; assert(velements[attr].src_offset <= 2048); /* 11-bit field */ } else { @@ -335,6 +335,25 @@ st_draw_vbo(GLcontext *ctx, assert(velements[attr].src_format); } +#if 0 + { + GLuint i; + char buf[100]; + for (i = 0; i < vp->num_inputs; i++) { + printf("buffers[%d].pitch = %u\n", i, vbuffer[i].pitch); + printf("buffers[%d].max_index = %u\n", i, vbuffer[i].max_index); + printf("buffers[%d].buffer_offset = %u\n", i, vbuffer[i].buffer_offset); + printf("buffers[%d].buffer = %p\n", i, (void*) vbuffer[i].buffer); + } + for (i = 0; i < vp->num_inputs; i++) { + printf("vlements[%d].src_offset = %u\n", i, velements[i].src_offset); + printf("vlements[%d].vbuffer_index = %u\n", i, velements[i].vertex_buffer_index); + printf("vlements[%d].nr_comps = %u\n", i, velements[i].nr_components); + printf("vlements[%d].format = %s\n", i, pf_sprint_name(buf, velements[i].src_format)); + } + } +#endif + pipe->set_vertex_buffers(pipe, vp->num_inputs, vbuffer); pipe->set_vertex_elements(pipe, vp->num_inputs, velements); -- cgit v1.2.3 From 1828b6d0315be805ec1eea1594fae5b28cfb4ca8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Jul 2008 10:12:55 -0600 Subject: gallium: move assertion (though not really significant) --- src/mesa/state_tracker/st_draw.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 3291387929..a57d7b6a7a 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -302,7 +302,6 @@ st_draw_vbo(GLcontext *ctx, pipe_reference_buffer(pipe, &vbuffer[attr].buffer, stobj->buffer); vbuffer[attr].buffer_offset = (unsigned) arrays[mesaAttr]->Ptr; velements[attr].src_offset = 0; - assert(velements[attr].src_offset <= 2048); /* 11-bit field */ } else { /* attribute data is in user-space memory, not a VBO */ @@ -323,6 +322,8 @@ st_draw_vbo(GLcontext *ctx, velements[attr].src_offset = 0; } + assert(velements[attr].src_offset <= 2048); /* 11-bit field */ + /* common-case setup */ vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ vbuffer[attr].max_index = max_index; -- cgit v1.2.3 From de71e4741d8b4e3719e702431912374f1cb90034 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 7 Jul 2008 14:34:30 -0600 Subject: egl: bump up MAX_ATTRIBS, added assertion --- src/egl/main/eglconfig.h | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglconfig.h b/src/egl/main/eglconfig.h index d12f66245c..db1c4c10e0 100644 --- a/src/egl/main/eglconfig.h +++ b/src/egl/main/eglconfig.h @@ -6,7 +6,7 @@ #include -#define MAX_ATTRIBS 100 +#define MAX_ATTRIBS 128 #define FIRST_ATTRIB EGL_BUFFER_SIZE @@ -17,7 +17,11 @@ struct _egl_config }; -#define SET_CONFIG_ATTRIB(CONF, ATTR, VAL) ((CONF)->Attrib[(ATTR) - FIRST_ATTRIB] = VAL) +#define SET_CONFIG_ATTRIB(CONF, ATTR, VAL) \ + assert((ATTR) - FIRST_ATTRIB < MAX_ATTRIBS); \ + ((CONF)->Attrib[(ATTR) - FIRST_ATTRIB] = VAL) + + #define GET_CONFIG_ATTRIB(CONF, ATTR) ((CONF)->Attrib[(ATTR) - FIRST_ATTRIB]) -- cgit v1.2.3 From 8d2400f2161efa95d2ef18fec353db4841f09637 Mon Sep 17 00:00:00 2001 From: Jonathan White Date: Mon, 7 Jul 2008 17:56:29 -0600 Subject: mesa: added GLAPIENTRY keywords --- src/mesa/vbo/vbo.h | 10 +++++----- src/mesa/vbo/vbo_exec_api.c | 10 +++++----- 2 files changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h index 59cd7f6984..5362226c2f 100644 --- a/src/mesa/vbo/vbo.h +++ b/src/mesa/vbo/vbo.h @@ -120,19 +120,19 @@ void vbo_use_buffer_objects(GLcontext *ctx); void vbo_set_draw_func(GLcontext *ctx, vbo_draw_func func); -void +void GLAPIENTRY _vbo_Color4f(GLfloat r, GLfloat g, GLfloat b, GLfloat a); -void +void GLAPIENTRY _vbo_Normal3f(GLfloat x, GLfloat y, GLfloat z); -void +void GLAPIENTRY _vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q); -void +void GLAPIENTRY _vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params); -void +void GLAPIENTRY _vbo_VertexAttrib4f(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w); #endif diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c index 2c2e66dd0b..579bbeb4f3 100644 --- a/src/mesa/vbo/vbo_exec_api.c +++ b/src/mesa/vbo/vbo_exec_api.c @@ -770,34 +770,34 @@ static void reset_attrfv( struct vbo_exec_context *exec ) } -void +void GLAPIENTRY _vbo_Color4f(GLfloat r, GLfloat g, GLfloat b, GLfloat a) { vbo_Color4f(r, g, b, a); } -void +void GLAPIENTRY _vbo_Normal3f(GLfloat x, GLfloat y, GLfloat z) { vbo_Normal3f(x, y, z); } -void +void GLAPIENTRY _vbo_MultiTexCoord4f(GLenum target, GLfloat s, GLfloat t, GLfloat r, GLfloat q) { vbo_MultiTexCoord4f(target, s, t, r, q); } -void +void GLAPIENTRY _vbo_Materialfv(GLenum face, GLenum pname, const GLfloat *params) { vbo_Materialfv(face, pname, params); } -void +void GLAPIENTRY _vbo_VertexAttrib4f(GLuint index, GLfloat x, GLfloat y, GLfloat z, GLfloat w) { vbo_VertexAttrib4fARB(index, x, y, z, w); -- cgit v1.2.3 From b13b1210c0174e4fd315bad3e6e18b6cbeeb1518 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 8 Jul 2008 14:16:15 +0200 Subject: i915: Renamed intel_screen to intel_device The renameing makes more sense, why because: In egl you have a display/driver/device as the basis, and you have screens as outputs. --- src/gallium/winsys/egl_drm/intel/Makefile | 2 +- src/gallium/winsys/egl_drm/intel/intel_context.c | 12 +- src/gallium/winsys/egl_drm/intel/intel_context.h | 4 +- src/gallium/winsys/egl_drm/intel/intel_device.c | 137 +++++++++++++++++++++ src/gallium/winsys/egl_drm/intel/intel_device.h | 50 ++++++++ src/gallium/winsys/egl_drm/intel/intel_screen.c | 137 --------------------- src/gallium/winsys/egl_drm/intel/intel_screen.h | 50 -------- .../winsys/egl_drm/intel/intel_swapbuffers.c | 2 +- 8 files changed, 197 insertions(+), 197 deletions(-) create mode 100644 src/gallium/winsys/egl_drm/intel/intel_device.c create mode 100644 src/gallium/winsys/egl_drm/intel/intel_device.h delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_screen.c delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_screen.h (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/Makefile b/src/gallium/winsys/egl_drm/intel/Makefile index 9c7ff065ff..e67b49f3ad 100644 --- a/src/gallium/winsys/egl_drm/intel/Makefile +++ b/src/gallium/winsys/egl_drm/intel/Makefile @@ -12,7 +12,7 @@ PIPE_DRIVERS = \ DRIVER_SOURCES = \ intel_swapbuffers.c \ intel_context.c \ - intel_screen.c \ + intel_device.c \ intel_egl.c C_SOURCES = \ diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.c b/src/gallium/winsys/egl_drm/intel/intel_context.c index 7513f3feef..c2003f5e0e 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_context.c +++ b/src/gallium/winsys/egl_drm/intel/intel_context.c @@ -27,7 +27,7 @@ #include "i915simple/i915_screen.h" -#include "intel_screen.h" +#include "intel_device.h" #include "intel_context.h" #include "intel_batchbuffer.h" @@ -147,13 +147,13 @@ int intel_create_context(struct egl_drm_context *egl_context, const __GLcontextModes *visual, void *sharedContextPrivate) { struct intel_context *intel = CALLOC_STRUCT(intel_context); - struct intel_screen *screen = (struct intel_screen *)egl_context->device->priv; + struct intel_device *screen = (struct intel_device *)egl_context->device->priv; struct pipe_context *pipe; struct st_context *st_share = NULL; egl_context->priv = intel; - intel->intel_screen = screen; + intel->intel_device = screen; intel->egl_context = egl_context; intel->egl_device = egl_context->device; @@ -183,8 +183,8 @@ intel_destroy_context(struct egl_drm_context *egl_context) { struct intel_context *intel = egl_context->priv; - if (intel->intel_screen->dummy == intel) - intel->intel_screen->dummy = NULL; + if (intel->intel_device->dummy == intel) + intel->intel_device->dummy = NULL; st_destroy_context(intel->st); intel_be_destroy_context(&intel->base); @@ -220,7 +220,7 @@ intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *dra void intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer *front) { - struct intel_screen *intelScreen = (struct intel_screen *)draw->device->priv; + struct intel_device *intelScreen = (struct intel_device *)draw->device->priv; struct intel_framebuffer *draw_fb = (struct intel_framebuffer *)draw->priv; if (draw_fb->front_buffer) diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.h b/src/gallium/winsys/egl_drm/intel/intel_context.h index f05a213ff0..48f9e21b5d 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_context.h +++ b/src/gallium/winsys/egl_drm/intel/intel_context.h @@ -47,7 +47,7 @@ struct intel_context struct st_context *st; - struct intel_screen *intel_screen; + struct intel_device *intel_device; /* new egl stuff */ struct egl_drm_device *egl_device; @@ -64,7 +64,7 @@ struct intel_framebuffer { struct st_framebuffer *stfb; - struct intel_screen *screen; + struct intel_device *screen; struct _DriBufferObject *front_buffer; struct egl_drm_frontbuffer *front; }; diff --git a/src/gallium/winsys/egl_drm/intel/intel_device.c b/src/gallium/winsys/egl_drm/intel/intel_device.c new file mode 100644 index 0000000000..7496c1bb30 --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_device.c @@ -0,0 +1,137 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "utils.h" + +#include "state_tracker/st_public.h" +#include "i915simple/i915_screen.h" + +#include "intel_context.h" +#include "intel_device.h" +#include "intel_batchbuffer.h" +#include "intel_egl.h" + + +extern const struct dri_extension card_extensions[]; + + +int +intel_create_device(struct egl_drm_device *device) +{ + struct intel_device *intel_device; + + /* Allocate the private area */ + intel_device = CALLOC_STRUCT(intel_device); + if (!intel_device) + return FALSE; + + device->priv = (void *)intel_device; + intel_device->device = device; + + intel_device->deviceID = device->deviceID; + + intel_be_init_device(&intel_device->base, device->drmFD, PCI_CHIP_I945_GM); + + intel_device->pipe = i915_create_screen(&intel_device->base.base, intel_device->deviceID); + + /* hack */ + driInitExtensions(NULL, card_extensions, GL_FALSE); + + return TRUE; +} + +int +intel_destroy_device(struct egl_drm_device *device) +{ + struct intel_device *intel_device = (struct intel_device *)device->priv; + + intel_be_destroy_device(&intel_device->base); + + free(intel_device); + device->priv = NULL; + + return TRUE; +} + +int +intel_create_drawable(struct egl_drm_drawable *drawable, + const __GLcontextModes * visual) +{ + enum pipe_format colorFormat, depthFormat, stencilFormat; + struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); + + if (!intelfb) + return GL_FALSE; + + intelfb->screen = drawable->device->priv; + + if (visual->redBits == 5) + colorFormat = PIPE_FORMAT_R5G6B5_UNORM; + else + colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; + + if (visual->depthBits == 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (visual->depthBits == 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_NONE; + + if (visual->stencilBits == 8) + stencilFormat = PIPE_FORMAT_S8Z24_UNORM; + else + stencilFormat = PIPE_FORMAT_NONE; + + intelfb->stfb = st_create_framebuffer(visual, + colorFormat, + depthFormat, + stencilFormat, + drawable->w, + drawable->h, + (void*) intelfb); + + if (!intelfb->stfb) { + free(intelfb); + return GL_FALSE; + } + + drawable->priv = (void *) intelfb; + return GL_TRUE; +} + +int +intel_destroy_drawable(struct egl_drm_drawable *drawable) +{ + struct intel_framebuffer *intelfb = (struct intel_framebuffer *)drawable->priv; + drawable->priv = NULL; + + assert(intelfb->stfb); + st_unreference_framebuffer(&intelfb->stfb); + free(intelfb); + return TRUE; +} diff --git a/src/gallium/winsys/egl_drm/intel/intel_device.h b/src/gallium/winsys/egl_drm/intel/intel_device.h new file mode 100644 index 0000000000..2f9d4f887e --- /dev/null +++ b/src/gallium/winsys/egl_drm/intel/intel_device.h @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _INTEL_SCREEN_H_ +#define _INTEL_SCREEN_H_ + +#include "intel_drm/intel_be_device.h" + +#include "pipe/p_compiler.h" + +struct pipe_screen; +struct egl_drm_device; +struct intel_context; + +struct intel_device +{ + struct intel_be_device base; + struct pipe_screen *pipe; + + int deviceID; + struct egl_drm_device *device; + + struct intel_context *dummy; +}; + +#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.c b/src/gallium/winsys/egl_drm/intel/intel_screen.c deleted file mode 100644 index 8e6d184440..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_screen.c +++ /dev/null @@ -1,137 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "utils.h" - -#include "state_tracker/st_public.h" -#include "i915simple/i915_screen.h" - -#include "intel_context.h" -#include "intel_screen.h" -#include "intel_batchbuffer.h" -#include "intel_egl.h" - - -extern const struct dri_extension card_extensions[]; - - -int -intel_create_device(struct egl_drm_device *device) -{ - struct intel_screen *intel_screen; - - /* Allocate the private area */ - intel_screen = CALLOC_STRUCT(intel_screen); - if (!intel_screen) - return FALSE; - - device->priv = (void *)intel_screen; - intel_screen->device = device; - - intel_screen->deviceID = device->deviceID; - - intel_be_init_device(&intel_screen->base, device->drmFD, PCI_CHIP_I945_GM); - - intel_screen->pipe = i915_create_screen(&intel_screen->base.base, intel_screen->deviceID); - - /* hack */ - driInitExtensions(NULL, card_extensions, GL_FALSE); - - return TRUE; -} - -int -intel_destroy_device(struct egl_drm_device *device) -{ - struct intel_screen *intel_screen = (struct intel_screen *)device->priv; - - intel_be_destroy_device(&intel_screen->base); - - free(intel_screen); - device->priv = NULL; - - return TRUE; -} - -int -intel_create_drawable(struct egl_drm_drawable *drawable, - const __GLcontextModes * visual) -{ - enum pipe_format colorFormat, depthFormat, stencilFormat; - struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); - - if (!intelfb) - return GL_FALSE; - - intelfb->screen = drawable->device->priv; - - if (visual->redBits == 5) - colorFormat = PIPE_FORMAT_R5G6B5_UNORM; - else - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (visual->depthBits == 16) - depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (visual->depthBits == 24) - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - else - depthFormat = PIPE_FORMAT_NONE; - - if (visual->stencilBits == 8) - stencilFormat = PIPE_FORMAT_S8Z24_UNORM; - else - stencilFormat = PIPE_FORMAT_NONE; - - intelfb->stfb = st_create_framebuffer(visual, - colorFormat, - depthFormat, - stencilFormat, - drawable->w, - drawable->h, - (void*) intelfb); - - if (!intelfb->stfb) { - free(intelfb); - return GL_FALSE; - } - - drawable->priv = (void *) intelfb; - return GL_TRUE; -} - -int -intel_destroy_drawable(struct egl_drm_drawable *drawable) -{ - struct intel_framebuffer *intelfb = (struct intel_framebuffer *)drawable->priv; - drawable->priv = NULL; - - assert(intelfb->stfb); - st_unreference_framebuffer(&intelfb->stfb); - free(intelfb); - return TRUE; -} diff --git a/src/gallium/winsys/egl_drm/intel/intel_screen.h b/src/gallium/winsys/egl_drm/intel/intel_screen.h deleted file mode 100644 index 9b990b48e7..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_screen.h +++ /dev/null @@ -1,50 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef _INTEL_SCREEN_H_ -#define _INTEL_SCREEN_H_ - -#include "intel_drm/intel_be_device.h" - -#include "pipe/p_compiler.h" - -struct pipe_screen; -struct egl_drm_device; -struct intel_context; - -struct intel_screen -{ - struct intel_be_device base; - struct pipe_screen *pipe; - - int deviceID; - struct egl_drm_device *device; - - struct intel_context *dummy; -}; - -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c index bfff6f935f..86e4a0f92a 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "intel_screen.h" +#include "intel_device.h" #include "intel_context.h" #include "intel_batchbuffer.h" #include "intel_reg.h" -- cgit v1.2.3 From 0c6efeb02beba32a35dda7c5d5a3086ca2d8cbc4 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 8 Jul 2008 14:41:17 +0200 Subject: i915: Complete the screen -> device renaming --- src/gallium/winsys/egl_drm/intel/intel_context.c | 14 +++++++------- src/gallium/winsys/egl_drm/intel/intel_context.h | 2 +- src/gallium/winsys/egl_drm/intel/intel_device.c | 2 +- src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c | 2 +- 4 files changed, 10 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.c b/src/gallium/winsys/egl_drm/intel/intel_context.c index c2003f5e0e..927addb834 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_context.c +++ b/src/gallium/winsys/egl_drm/intel/intel_context.c @@ -147,13 +147,13 @@ int intel_create_context(struct egl_drm_context *egl_context, const __GLcontextModes *visual, void *sharedContextPrivate) { struct intel_context *intel = CALLOC_STRUCT(intel_context); - struct intel_device *screen = (struct intel_device *)egl_context->device->priv; + struct intel_device *device = (struct intel_device *)egl_context->device->priv; struct pipe_context *pipe; struct st_context *st_share = NULL; egl_context->priv = intel; - intel->intel_device = screen; + intel->intel_device = device; intel->egl_context = egl_context; intel->egl_device = egl_context->device; @@ -161,19 +161,19 @@ intel_create_context(struct egl_drm_context *egl_context, const __GLcontextModes intel->base.hardware_unlock = intel_unlock_hardware; intel->base.hardware_locked = intel_locked_hardware; - intel_be_init_context(&intel->base, &screen->base); + intel_be_init_context(&intel->base, &device->base); #if 0 pipe = intel_create_softpipe(intel, screen->winsys); #else - pipe = i915_create_context(screen->pipe, &screen->base.base, &intel->base.base); + pipe = i915_create_context(device->pipe, &device->base.base, &intel->base.base); #endif pipe->priv = intel; intel->st = st_create_context(pipe, visual, st_share); - screen->dummy = intel; + device->dummy = intel; return TRUE; } @@ -220,7 +220,7 @@ intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *dra void intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer *front) { - struct intel_device *intelScreen = (struct intel_device *)draw->device->priv; + struct intel_device *device = (struct intel_device *)draw->device->priv; struct intel_framebuffer *draw_fb = (struct intel_framebuffer *)draw->priv; if (draw_fb->front_buffer) @@ -235,7 +235,7 @@ intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer draw_fb->front = front; - driGenBuffers(intelScreen->base.staticPool, "front", 1, &draw_fb->front_buffer, 0, 0, 0); + driGenBuffers(device->base.staticPool, "front", 1, &draw_fb->front_buffer, 0, 0, 0); driBOSetReferenced(draw_fb->front_buffer, front->handle); st_resize_framebuffer(draw_fb->stfb, draw->w, draw->h); diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.h b/src/gallium/winsys/egl_drm/intel/intel_context.h index 48f9e21b5d..dfa4720b08 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_context.h +++ b/src/gallium/winsys/egl_drm/intel/intel_context.h @@ -64,7 +64,7 @@ struct intel_framebuffer { struct st_framebuffer *stfb; - struct intel_device *screen; + struct intel_device *device; struct _DriBufferObject *front_buffer; struct egl_drm_frontbuffer *front; }; diff --git a/src/gallium/winsys/egl_drm/intel/intel_device.c b/src/gallium/winsys/egl_drm/intel/intel_device.c index 7496c1bb30..a601904ba4 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_device.c +++ b/src/gallium/winsys/egl_drm/intel/intel_device.c @@ -88,7 +88,7 @@ intel_create_drawable(struct egl_drm_drawable *drawable, if (!intelfb) return GL_FALSE; - intelfb->screen = drawable->device->priv; + intelfb->device = drawable->device->priv; if (visual->redBits == 5) colorFormat = PIPE_FORMAT_R5G6B5_UNORM; diff --git a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c index 86e4a0f92a..2edcbc79ff 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c @@ -69,7 +69,7 @@ intel_display_surface(struct egl_drm_drawable *draw, //const int srcWidth = surf->width; //const int srcHeight = surf->height; - intel = intel_fb->screen->dummy; + intel = intel_fb->device->dummy; if (!intel) { printf("No dummy context\n"); return; -- cgit v1.2.3 From 1257bb9b130c9487a32e422a8ce8ddba2f14a568 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 8 Jul 2008 14:44:23 +0200 Subject: i915: Use deviceID in EGL device create --- src/gallium/winsys/egl_drm/intel/intel_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_device.c b/src/gallium/winsys/egl_drm/intel/intel_device.c index a601904ba4..b9649cbec7 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_device.c +++ b/src/gallium/winsys/egl_drm/intel/intel_device.c @@ -55,7 +55,7 @@ intel_create_device(struct egl_drm_device *device) intel_device->deviceID = device->deviceID; - intel_be_init_device(&intel_device->base, device->drmFD, PCI_CHIP_I945_GM); + intel_be_init_device(&intel_device->base, device->drmFD, intel_device->deviceID); intel_device->pipe = i915_create_screen(&intel_device->base.base, intel_device->deviceID); -- cgit v1.2.3 From 2e7e1837512c1959a97f51903a770cf182fc9820 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Tue, 8 Jul 2008 16:54:22 +0200 Subject: i915: Takedown EGL screen --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 81 ++++++++++++++++++++++++---- 1 file changed, 71 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index a476490b90..55f4d92afb 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -24,6 +24,8 @@ #include "state_tracker/st_public.h" +#define MAX_SCREENS 16 + static void drm_get_device_id(struct egl_drm_device *device) { @@ -66,11 +68,17 @@ egl_drm_create_device(int drmFD) __GLcontextModes* _gl_context_modes_create( unsigned count, size_t minimum_size ); +struct drm_screen; + struct drm_driver { _EGLDriver base; /* base class/object */ drmModeResPtr res; + + struct drm_screen *screens[MAX_SCREENS]; + size_t count_screens; + struct egl_drm_device *device; }; @@ -92,17 +100,26 @@ struct drm_screen { _EGLScreen base; - /* backing buffer and crtc */ + /* currently only support one connector */ + drmModeConnectorPtr connector; + uint32_t connectorID; + + /* Has this screen been shown */ + int shown; + + /* Surface that is currently attached to this screen */ + struct drm_surface *surf; + + /* backing buffer */ drmBO buffer; + + /* framebuffer */ drmModeFBPtr fb; uint32_t fbID; - drmModeCrtcPtr crtc; - /* currently only support one connector */ - drmModeConnectorPtr connector; - drmModeEncoderPtr encoder; - uint32_t connectorID; - uint32_t encoderID; + /* crtc and mode used */ + drmModeCrtcPtr crtc; + uint32_t crtcID; struct drm_mode_modeinfo *mode; @@ -139,6 +156,7 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) drmModeConnectorPtr connector = NULL; drmModeResPtr res = NULL; unsigned count_connectors = 0; + int num_screens = 0; EGLint i; int fd; @@ -159,13 +177,13 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) if (res) count_connectors = res->count_connectors; - for(i = 0; i < count_connectors; i++) { + for(i = 0; i < count_connectors && i < MAX_SCREENS; i++) { connector = drmModeGetConnector(fd, res->connectors[i]); if (!connector) continue; - if (connector->connection == DRM_MODE_DISCONNECTED) { + if (connector->connection != DRM_MODE_CONNECTED) { drmModeFreeConnector(connector); continue; } @@ -177,7 +195,9 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) _eglInitScreen(&screen->base); _eglAddScreen(disp, &screen->base); drm_add_modes_from_connector(&screen->base, connector); + drm_drv->screens[num_screens++] = screen; } + drm_drv->count_screens = num_screens; /* for now we only have one config */ _EGLConfig *config = calloc(1, sizeof(*config)); @@ -201,15 +221,52 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) return EGL_TRUE; } +static void +drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen) +{ + struct drm_driver *drm_drv = (struct drm_driver *)drv; + + intel_bind_frontbuffer(screen->surf->drawable, NULL); + screen->surf = NULL; + + drmModeSetCrtc( + drm_drv->device->drmFD, + drm_drv->res->crtcs[1], + 0, // FD + 0, 0, + NULL, 0, // List of output ids + NULL); + + drmModeRmFB(drm_drv->device->drmFD, screen->fbID); + drmModeFreeFB(screen->fb); + screen->fb = NULL; + + drmBOUnreference(drm_drv->device->drmFD, &screen->buffer); + + screen->shown = 0; +} static EGLBoolean drm_terminate(_EGLDriver *drv, EGLDisplay dpy) { struct drm_driver *drm_drv = (struct drm_driver *)drv; + struct drm_screen *screen; + int i = 0; intel_destroy_device(drm_drv->device); drmFreeVersion(drm_drv->device->version); + for (i = 0; i < drm_drv->count_screens; i++) { + screen = drm_drv->screens[i]; + + if (screen->shown) + drm_takedown_shown_screen(drv, screen); + + drmModeFreeConnector(screen->connector); + _eglDestroyScreen(&screen->base); + drm_drv->screens[i] = NULL; + } + drmClose(drm_drv->device->drmFD); free(drm_drv->device); @@ -507,7 +564,8 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, size_t size = mode->Height * pitch; int ret; - /* TODO if allready shown take down */ + if (scrn->shown) + drm_takedown_shown_screen(drv, scrn); ret = drmBOCreate(drm_drv->device->drmFD, size, 0, 0, DRM_BO_FLAG_READ | @@ -552,8 +610,11 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, scrn->front.width = mode->Width; scrn->front.height = mode->Height; + scrn->surf = surf; intel_bind_frontbuffer(surf->drawable, &scrn->front); + scrn->shown = 1; + return EGL_TRUE; err_fb: -- cgit v1.2.3 From 2e638cfe3c3a8ab639d9cdb13abced6dc8ed0755 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 8 Jul 2008 13:02:33 -0600 Subject: mesa: add missing VARYING case to storage_string() --- src/mesa/shader/slang/slang_ir.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_ir.c b/src/mesa/shader/slang/slang_ir.c index 31ad94183c..16c36aad0c 100644 --- a/src/mesa/shader/slang/slang_ir.c +++ b/src/mesa/shader/slang/slang_ir.c @@ -218,12 +218,14 @@ storage_string(const slang_ir_storage *st) "NAMED_PARAM", "CONSTANT", "UNIFORM", + "VARYING", "WRITE_ONLY", "ADDRESS", "SAMPLER", "UNDEFINED" }; static char s[100]; + assert(Elements(files) == PROGRAM_FILE_MAX); #if 0 if (st->Size == 1) sprintf(s, "%s[%d]", files[st->File], st->Index); -- cgit v1.2.3 From 0d1f54db6b4e66b164ea4193e7aa41a98b6fce0d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 8 Jul 2008 13:05:43 -0600 Subject: mesa: compute global var size before doing codegen --- src/mesa/shader/slang/slang_compile.c | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index e2117c03e6..4acbe802d2 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -1634,6 +1634,15 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O, return 0; } + /* allocate global address space for a variable with a known size */ + if (C->global_scope + && !(var->type.specifier.type == SLANG_SPEC_ARRAY + && var->array_len == 0)) { + if (!calculate_var_size(C, O, var)) + return GL_FALSE; + var->address = slang_var_pool_alloc(O->global_pool, var->size); + } + /* emit code for global var decl */ if (C->global_scope) { slang_assemble_ctx A; @@ -1648,15 +1657,6 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O, return 0; } - /* allocate global address space for a variable with a known size */ - if (C->global_scope - && !(var->type.specifier.type == SLANG_SPEC_ARRAY - && var->array_len == 0)) { - if (!calculate_var_size(C, O, var)) - return GL_FALSE; - var->address = slang_var_pool_alloc(O->global_pool, var->size); - } - /* initialize global variable */ if (C->global_scope) { if (var->initializer != NULL) { -- cgit v1.2.3 From fe1c0e44aadc033075797804b3275ac91f72c1ba Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 8 Jul 2008 14:59:31 -0600 Subject: mesa: more debug output --- src/mesa/shader/slang/slang_print.c | 80 +++++++++++++++++++++++++++---------- 1 file changed, 60 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_print.c b/src/mesa/shader/slang/slang_print.c index 4a7d4bbbc7..ff9c62c929 100644 --- a/src/mesa/shader/slang/slang_print.c +++ b/src/mesa/shader/slang/slang_print.c @@ -171,9 +171,15 @@ static void print_binary(const slang_operation *op, const char *oper, int indent) { assert(op->num_children == 2); + printf("binary at %p locals=%p outer=%p\n", + (void *) op, + (void *) op->locals, + (void *) op->locals->outer_scope); slang_print_tree(&op->children[0], indent + 3); spaces(indent); - printf("%s\n", oper); + printf("%s at %p locals=%p outer=%p\n", + oper, (void *) op, (void *) op->locals, + (void *) op->locals->outer_scope); slang_print_tree(&op->children[1], indent + 3); } @@ -185,11 +191,13 @@ print_generic2(const slang_operation *op, const char *oper, GLuint i; if (oper) { spaces(indent); - printf("[%p locals %p] %s %s\n", (void*) op, (void*) op->locals, oper, s); + printf("%s %s at %p locals=%p outer=%p\n", + oper, s, (void *) op, (void *) op->locals, + (void *) op->locals->outer_scope); } for (i = 0; i < op->num_children; i++) { spaces(indent); - printf("//child %u:\n", i); + printf("//child %u of %u:\n", i, op->num_children); slang_print_tree(&op->children[i], indent); } } @@ -244,7 +252,7 @@ slang_print_tree(const slang_operation *op, int indent) case SLANG_OPER_BLOCK_NO_NEW_SCOPE: spaces(indent); - printf("{ locals %p outer %p\n", (void*)op->locals, (void*)op->locals->outer_scope); + printf("{ locals=%p outer=%p\n", (void*)op->locals, (void*)op->locals->outer_scope); print_generic(op, NULL, indent+3); spaces(indent); printf("}\n"); @@ -252,7 +260,14 @@ slang_print_tree(const slang_operation *op, int indent) case SLANG_OPER_BLOCK_NEW_SCOPE: spaces(indent); - printf("{{ // new scope locals %p\n", (void*)op->locals); + printf("{{ // new scope locals=%p: ", (void*)op->locals); + { + int i; + for (i = 0; i < op->locals->num_variables; i++) { + printf("%s ", (char *) op->locals->variables[i]->a_name); + } + printf("\n"); + } print_generic(op, NULL, indent+3); spaces(indent); printf("}}\n"); @@ -264,14 +279,16 @@ slang_print_tree(const slang_operation *op, int indent) slang_variable *v; v = _slang_locate_variable(op->locals, op->a_id, GL_TRUE); if (v) { + const slang_variable_scope *scope; spaces(indent); printf("DECL (locals=%p outer=%p) ", (void*)op->locals, (void*) op->locals->outer_scope); print_type(&v->type); printf(" %s (%p)", (char *) op->a_id, (void *) find_var(op->locals, op->a_id)); - printf(" (in scope %p) ", - (void *) find_scope(op->locals, op->a_id)); + scope = find_scope(op->locals, op->a_id); + printf(" (in scope %p) ", (void *) scope); + assert(scope); if (op->num_children == 1) { printf(" :=\n"); slang_print_tree(&op->children[0], indent + 3); @@ -300,8 +317,12 @@ slang_print_tree(const slang_operation *op, int indent) case SLANG_OPER_ASM: spaces(indent); - printf("ASM: %s\n", (char*) op->a_id); - print_generic(op, NULL, indent+3); + printf("ASM: %s at %p locals=%p outer=%p\n", + (char *) op->a_id, + (void *) op, + (void *) op->locals, + (void *) op->locals->outer_scope); + print_generic(op, "ASM", indent+3); break; case SLANG_OPER_BREAK: @@ -333,7 +354,9 @@ slang_print_tree(const slang_operation *op, int indent) case SLANG_OPER_EXPRESSION: spaces(indent); - printf("EXPR: locals %p\n", (void*) op->locals); + printf("EXPR: locals=%p outer=%p\n", + (void *) op->locals, + (void *) op->locals->outer_scope); /*print_generic(op, "SLANG_OPER_EXPRESSION", indent);*/ slang_print_tree(&op->children[0], indent + 3); break; @@ -422,13 +445,25 @@ slang_print_tree(const slang_operation *op, int indent) break; case SLANG_OPER_IDENTIFIER: - spaces(indent); - if (op->var && op->var->a_name) - printf("VAR %s (in scope %p)\n", (char *) op->var->a_name, - (void *) find_scope(op->locals, op->a_id)); - else - printf("VAR' %s (in scope %p)\n", (char *) op->a_id, - (void *) find_scope(op->locals, op->a_id)); + { + const slang_variable_scope *scope; + spaces(indent); + if (op->var && op->var->a_name) { + scope = find_scope(op->locals, op->var->a_name); + printf("VAR %s (in scope %p)\n", (char *) op->var->a_name, + (void *) scope); + assert(scope); + } + else { + scope = find_scope(op->locals, op->a_id); + printf("VAR' %s (in scope %p) locals=%p outer=%p\n", + (char *) op->a_id, + (void *) scope, + (void *) op->locals, + (void *) op->locals->outer_scope); + assert(scope); + } + } break; case SLANG_OPER_SEQUENCE: @@ -437,7 +472,9 @@ slang_print_tree(const slang_operation *op, int indent) case SLANG_OPER_ASSIGN: spaces(indent); - printf("ASSIGNMENT locals %p\n", (void*)op->locals); + printf("ASSIGNMENT locals=%p outer=%p\n", + (void *) op->locals, + (void *) op->locals->outer_scope); print_binary(op, ":=", indent); break; @@ -573,7 +610,9 @@ slang_print_tree(const slang_operation *op, int indent) case SLANG_OPER_SUBSCRIPT: spaces(indent); - printf("SLANG_OPER_SUBSCRIPT\n"); + printf("SLANG_OPER_SUBSCRIPT locals=%p outer=%p\n", + (void *) op->locals, + (void *) op->locals->outer_scope); print_generic(op, NULL, indent+3); break; @@ -640,7 +679,8 @@ slang_print_function(const slang_function *f, GLboolean body) print_variable(f->parameters->variables[i], 3); } - printf(")\n"); + printf(") param scope = %p\n", (void *) f->parameters); + if (body && f->body) slang_print_tree(f->body, 0); } -- cgit v1.2.3 From 7cbc244c52610eb59b0fe1fc7275f307b560c281 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 8 Jul 2008 15:00:11 -0600 Subject: gallium: tweak printing of generic declarations --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 92aff88925..a395c630cc 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -573,7 +573,8 @@ tgsi_dump_declaration( if (decl->Declaration.Semantic) { TXT( ", " ); ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); - if (decl->Semantic.SemanticIndex != 0) { + if (decl->Semantic.SemanticIndex != 0 || + decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC) { CHR( '[' ); UID( decl->Semantic.SemanticIndex ); CHR( ']' ); -- cgit v1.2.3 From 3bdf50bab0951c5435bbf4b938d37cc1fe9a5d7c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 8 Jul 2008 15:04:47 -0600 Subject: mesa: fix bug/failure in recursive function inlining Fixes a failure for cases such as y = f(a, f(a, b)) All the usual tests still pass but regressions are possible... --- src/mesa/shader/slang/slang_codegen.c | 62 +++++++++++++++++++++++++++++++---- 1 file changed, 56 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 14fa38c3cd..af45dfb2e8 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -924,6 +924,30 @@ slang_substitute(slang_assemble_ctx *A, slang_operation *oper, } +/** + * Recursively traverse 'oper', replacing occurances of 'oldScope' with + * 'newScope' in the oper->locals->outer_scope filed. + * + * This is used after function inlining to update the scoping of + * the newly copied/inlined code so that vars are found in the new, + * inlined scope and not in the original function code. + */ +static void +slang_replace_scope(slang_operation *oper, + slang_variable_scope *oldScope, + slang_variable_scope *newScope) +{ + GLuint i; + if (oper->locals != newScope && + oper->locals->outer_scope == oldScope) { + oper->locals->outer_scope = newScope; + } + for (i = 0; i < oper->num_children; i++) { + slang_replace_scope(&oper->children[i], oldScope, newScope); + } +} + + /** * Produce inline code for a call to an assembly instruction. @@ -985,6 +1009,13 @@ slang_inline_asm_function(slang_assemble_ctx *A, _slang_free(substOld); _slang_free(substNew); +#if 0 + printf("+++++++++++++ inlined asm function %s +++++++++++++\n", + (char *) fun->header.a_name); + slang_print_tree(inlined, 3); + printf("+++++++++++++++++++++++++++++++++++++++++++++++++++\n"); +#endif + return inlined; } @@ -1012,6 +1043,7 @@ slang_inline_function_call(slang_assemble_ctx * A, slang_function *fun, slang_operation **substNew; GLuint substCount, numCopyIn, i; slang_function *prevFunction; + slang_variable_scope *newScope = NULL; /* save / push */ prevFunction = A->CurFunction; @@ -1029,7 +1061,7 @@ slang_inline_function_call(slang_assemble_ctx * A, slang_function *fun, _slang_alloc(totalArgs * sizeof(slang_operation *)); #if 0 - printf("Inline call to %s (total vars=%d nparams=%d)\n", + printf("\nInline call to %s (total vars=%d nparams=%d)\n", (char *) fun->header.a_name, fun->parameters->num_variables, numArgs); #endif @@ -1179,9 +1211,7 @@ slang_inline_function_call(slang_assemble_ctx * A, slang_function *fun, slang_operation *decl = slang_operation_insert(&inlined->num_children, &inlined->children, numCopyIn); - /* - printf("COPY_IN %s from expr\n", (char*)p->a_name); - */ + decl->type = SLANG_OPER_VARIABLE_DECL; assert(decl->locals); decl->locals->outer_scope = inlined->locals; @@ -1192,6 +1222,15 @@ slang_inline_function_call(slang_assemble_ctx * A, slang_function *fun, /* child[0] is the var's initializer */ slang_operation_copy(&decl->children[0], args + i); + /* add parameter 'p' to the local variable scope here */ + { + slang_variable *pCopy = slang_variable_scope_grow(inlined->locals); + pCopy->type = p->type; + pCopy->a_name = p->a_name; + pCopy->array_len = p->array_len; + } + + newScope = inlined->locals; numCopyIn++; } } @@ -1231,8 +1270,15 @@ slang_inline_function_call(slang_assemble_ctx * A, slang_function *fun, _slang_free(substOld); _slang_free(substNew); + /* Update scoping to use the new local vars instead of the + * original function's vars. This is especially important + * for nested inlining. + */ + if (newScope) + slang_replace_scope(inlined, fun->parameters, newScope); + #if 0 - printf("Done Inline call to %s (total vars=%d nparams=%d)\n", + printf("Done Inline call to %s (total vars=%d nparams=%d)\n\n", (char *) fun->header.a_name, fun->parameters->num_variables, numArgs); slang_print_tree(top, 0); @@ -2063,7 +2109,11 @@ _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) n = new_node0(IR_VAR_DECL); if (n) { _slang_attach_storage(n, var); - +#if 0 + printf("%s var %p %s store=%p\n", + __FUNCTION__, (void *) var, (char *) var->a_name, + (void *) n->Store); +#endif assert(var->aux); assert(n->Store == var->aux); assert(n->Store); -- cgit v1.2.3 From 072c47483674021425922132cf6d977090077f8e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 8 Jul 2008 16:12:01 -0600 Subject: mesa: implement glGetUniformiv() with new ctx->Driver function The old implementation could overwrite the caller's param buffer. --- src/mesa/main/dd.h | 2 ++ src/mesa/main/shaders.c | 7 ++----- src/mesa/shader/shader_api.c | 45 ++++++++++++++++++++++++++++++++++++++++---- 3 files changed, 45 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/main/dd.h b/src/mesa/main/dd.h index a24021c63f..7fb0a211d7 100644 --- a/src/mesa/main/dd.h +++ b/src/mesa/main/dd.h @@ -870,6 +870,8 @@ struct dd_function_table { GLsizei *length, GLcharARB *sourceOut); void (*GetUniformfv)(GLcontext *ctx, GLuint program, GLint location, GLfloat *params); + void (*GetUniformiv)(GLcontext *ctx, GLuint program, GLint location, + GLint *params); GLint (*GetUniformLocation)(GLcontext *ctx, GLuint program, const GLcharARB *name); GLboolean (*IsProgram)(GLcontext *ctx, GLuint name); diff --git a/src/mesa/main/shaders.c b/src/mesa/main/shaders.c index b7b2f791a5..a2670fda32 100644 --- a/src/mesa/main/shaders.c +++ b/src/mesa/main/shaders.c @@ -128,6 +128,7 @@ _mesa_DeleteObjectARB(GLhandleARB obj) void GLAPIENTRY _mesa_DeleteProgram(GLuint name) { + printf("%s name=%u\n", __FUNCTION__, name); if (name) { GET_CURRENT_CONTEXT(ctx); ctx->Driver.DeleteProgram2(ctx, name); @@ -309,11 +310,7 @@ void GLAPIENTRY _mesa_GetUniformivARB(GLhandleARB program, GLint location, GLint * params) { GET_CURRENT_CONTEXT(ctx); - GLfloat fparams[16]; /* XXX is 16 enough? */ - GLuint i; - ctx->Driver.GetUniformfv(ctx, program, location, fparams); - for (i = 0; i < 16; i++) - params[i] = (GLint) fparams[i]; /* XXX correct? */ + ctx->Driver.GetUniformiv(ctx, program, location, params); } diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 182de37b50..c1fbcde61e 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -955,12 +955,15 @@ _mesa_get_shader_source(GLcontext *ctx, GLuint shader, GLsizei maxLength, } +#define MAX_UNIFORM_ELEMENTS 16 + /** - * Called via ctx->Driver.GetUniformfv(). + * Helper for GetUniformfv(), GetUniformiv() + * Returns number of elements written to 'params' output. */ -static void -_mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, - GLfloat *params) +static GLuint +get_uniformfv(GLcontext *ctx, GLuint program, GLint location, + GLfloat *params) { struct gl_shader_program *shProg = _mesa_lookup_shader_program(ctx, program); @@ -984,9 +987,13 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, ASSERT(prog); if (prog) { + /* See uniformiv() below */ + assert(prog->Parameters->Parameters[progPos].Size <= MAX_UNIFORM_ELEMENTS); + for (i = 0; i < prog->Parameters->Parameters[progPos].Size; i++) { params[i] = prog->Parameters->ParameterValues[progPos][i]; } + return prog->Parameters->Parameters[progPos].Size; } } else { @@ -996,6 +1003,35 @@ _mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, else { _mesa_error(ctx, GL_INVALID_OPERATION, "glGetUniformfv(program)"); } + return 0; +} + + +/** + * Called via ctx->Driver.GetUniformfv(). + */ +static void +_mesa_get_uniformfv(GLcontext *ctx, GLuint program, GLint location, + GLfloat *params) +{ + (void) get_uniformfv(ctx, program, location, params); +} + + +/** + * Called via ctx->Driver.GetUniformiv(). + */ +static void +_mesa_get_uniformiv(GLcontext *ctx, GLuint program, GLint location, + GLint *params) +{ + GLfloat fparams[MAX_UNIFORM_ELEMENTS]; + GLuint n = get_uniformfv(ctx, program, location, fparams); + GLuint i; + assert(n <= MAX_UNIFORM_ELEMENTS); + for (i = 0; i < n; i++) { + params[i] = (GLint) fparams[i]; + } } @@ -1413,6 +1449,7 @@ _mesa_init_glsl_driver_functions(struct dd_function_table *driver) driver->GetShaderInfoLog = _mesa_get_shader_info_log; driver->GetShaderSource = _mesa_get_shader_source; driver->GetUniformfv = _mesa_get_uniformfv; + driver->GetUniformiv = _mesa_get_uniformiv; driver->GetUniformLocation = _mesa_get_uniform_location; driver->IsProgram = _mesa_is_program; driver->IsShader = _mesa_is_shader; -- cgit v1.2.3 From cdc0b6e5236591ac16ba71867818d63f9f2fd93b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 8 Jul 2008 16:58:50 -0600 Subject: mesa: remove debug code --- src/mesa/main/shaders.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/shaders.c b/src/mesa/main/shaders.c index a2670fda32..f0db0d2a81 100644 --- a/src/mesa/main/shaders.c +++ b/src/mesa/main/shaders.c @@ -128,7 +128,6 @@ _mesa_DeleteObjectARB(GLhandleARB obj) void GLAPIENTRY _mesa_DeleteProgram(GLuint name) { - printf("%s name=%u\n", __FUNCTION__, name); if (name) { GET_CURRENT_CONTEXT(ctx); ctx->Driver.DeleteProgram2(ctx, name); -- cgit v1.2.3 From d25709df1d5d9dccdeb173b33a57018004fe849c Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 9 Jul 2008 09:29:49 -0400 Subject: draw: remove some debug output --- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index 6b92811870..8e834501a4 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -184,7 +184,6 @@ boolean aos_fetch_inputs( struct aos_compilation *cp, boolean linear ) if (!load_input( cp, i, linear )) return FALSE; cp->insn_counter++; - debug_printf("\n"); } return TRUE; @@ -316,7 +315,6 @@ boolean aos_emit_outputs( struct aos_compilation *cp ) aos_release_xmm_reg( cp, data.idx ); cp->insn_counter++; - debug_printf("\n"); } return TRUE; -- cgit v1.2.3 From 48b8a32c6dcd259046958b187927ac584b1a564a Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 9 Jul 2008 10:20:28 -0400 Subject: egl: plug a small memleak --- src/gallium/winsys/egl_xlib/egl_xlib.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 83b8bb95b1..708a58b781 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -535,6 +535,7 @@ xlib_eglDestroySurface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) surf->Base.DeletePending = EGL_TRUE; } else { + XFreeGC(surf->Dpy, surf->Gc); st_unreference_framebuffer(&surf->Framebuffer); free(surf); } -- cgit v1.2.3 From 57f2d071d1503823867029e4d672c09e6417cc68 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Jul 2008 08:47:23 -0600 Subject: mesa: fix state.clip[n].plane parsing bug (bug 16611) cherry-picked from master --- src/mesa/shader/arbprogparse.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index a6bbdc64f1..ff583352ce 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -1133,7 +1133,9 @@ parse_state_single_item (GLcontext * ctx, const GLubyte ** inst, struct arb_program *Program, gl_state_index state_tokens[STATE_LENGTH]) { - switch (*(*inst)++) { + GLubyte token = *(*inst)++; + + switch (token) { case STATE_MATERIAL_PARSER: state_tokens[0] = STATE_MATERIAL; state_tokens[1] = parse_face_type (inst); @@ -1318,7 +1320,6 @@ parse_state_single_item (GLcontext * ctx, const GLubyte ** inst, case STATE_CLIP_PLANE: state_tokens[0] = STATE_CLIPPLANE; - state_tokens[1] = parse_integer (inst, Program); if (parse_clipplane_num (ctx, inst, Program, (GLint *) &state_tokens[1])) return 1; @@ -1770,7 +1771,9 @@ parse_param_elements (GLcontext * ctx, const GLubyte ** inst, gl_state_index state_tokens[STATE_LENGTH] = {0, 0, 0, 0, 0}; GLfloat const_values[4]; - switch (*(*inst)++) { + GLubyte token = *(*inst)++; + + switch (token) { case PARAM_STATE_ELEMENT: if (parse_state_single_item (ctx, inst, Program, state_tokens)) return 1; -- cgit v1.2.3 From 520dbdea22442493771763a3a895f9e9039c2a5c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Jul 2008 08:48:41 -0600 Subject: mesa: check for OpenBSD (bug 15604) cherry-picked from master --- src/mesa/main/execmem.c | 12 +++++++++++- 1 file changed, 11 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/execmem.c b/src/mesa/main/execmem.c index 40f66d7da2..aa40b02205 100644 --- a/src/mesa/main/execmem.c +++ b/src/mesa/main/execmem.c @@ -36,7 +36,7 @@ -#if defined(__linux__) +#if defined(__linux__) || defined(__OpenBSD__) /* * Allocate a large block of memory which can hold code then dole it out @@ -47,6 +47,16 @@ #include #include "mm.h" +#ifdef MESA_SELINUX +#include +#endif + + +#ifndef MAP_ANONYMOUS +#define MAP_ANONYMOUS MAP_ANON +#endif + + #define EXEC_HEAP_SIZE (10*1024*1024) _glthread_DECLARE_STATIC_MUTEX(exec_mutex); -- cgit v1.2.3 From 93ff702b4f1c6016e249c4d326e71cdc4932f57c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Jul 2008 11:46:16 -0600 Subject: gallium: fix logic in pb_check_usage() --- src/gallium/auxiliary/pipebuffer/pb_buffer.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer.h b/src/gallium/auxiliary/pipebuffer/pb_buffer.h index 83ea0be74b..8505d333bd 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer.h +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer.h @@ -221,7 +221,7 @@ pb_check_alignment(size_t requested, size_t provided) static INLINE boolean pb_check_usage(unsigned requested, unsigned provided) { - return (requested & provided) == provided ? TRUE : FALSE; + return (requested & provided) == requested ? TRUE : FALSE; } -- cgit v1.2.3 From 70b0a949b0af9954359be452febaa44f08052056 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Jul 2008 15:47:27 -0600 Subject: egl: misc updates/fixes Set surface size when created. Implement dri_find_dri_screen(). Look for "egl_xxx_dri.so" library before regular DRI driver. --- src/egl/drivers/xdri/egl_xdri.c | 93 +++++++++++++++++++++++++++++++++-------- 1 file changed, 76 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c index 71d4f15371..5e50d6034a 100644 --- a/src/egl/drivers/xdri/egl_xdri.c +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -118,6 +118,10 @@ struct xdri_egl_config }; +/* XXX temp hack */ +static struct xdri_egl_driver *TheDriver = NULL; + + /** cast wrapper */ static struct xdri_egl_driver * xdri_egl_driver(_EGLDriver *drv) @@ -154,6 +158,21 @@ lookup_config(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config) +/** Get size of given window */ +static Status +get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height) +{ + Window root; + Status stat; + int xpos, ypos; + unsigned int w, h, bw, depth; + stat = XGetGeometry(dpy, d, &root, &xpos, &ypos, &w, &h, &bw, &depth); + *width = w; + *height = h; + return stat; +} + + /** * Produce a set of EGL configs. * Note that we get the list of GLcontextModes from the GLX library. @@ -273,24 +292,25 @@ dri_context_modes_create(unsigned count, size_t minimum_size) static __DRIscreen * dri_find_dri_screen(__DRInativeDisplay *ndpy, int scrn) { - /* unused? */ - return NULL; + assert(TheDriver); + + return &TheDriver->driScreen; } static GLboolean -dri_window_exists(__DRInativeDisplay *dpy, __DRIid draw) +dri_window_exists(__DRInativeDisplay *ndpy, __DRIid draw) { return EGL_TRUE; } static GLboolean -dri_create_context(__DRInativeDisplay *dpy, int screenNum, int configID, +dri_create_context(__DRInativeDisplay *ndpy, int screenNum, int configID, void * contextID, drm_context_t * hw_context) { assert(configID >= 0); - return XF86DRICreateContextWithConfig(dpy, screenNum, + return XF86DRICreateContextWithConfig(ndpy, screenNum, configID, contextID, hw_context); } @@ -325,7 +345,7 @@ dri_destroy_drawable(__DRInativeDisplay * ndpy, int screen, __DRIid drawable) static GLboolean -dri_get_drawable_info(__DRInativeDisplay *dpy, int scrn, +dri_get_drawable_info(__DRInativeDisplay *ndpy, int scrn, __DRIid draw, unsigned int * index, unsigned int * stamp, int * x, int * y, int * width, int * height, int * numClipRects, drm_clip_rect_t ** pClipRects, @@ -335,7 +355,7 @@ dri_get_drawable_info(__DRInativeDisplay *dpy, int scrn, { _eglLog(_EGL_DEBUG, "XDRI: %s", __FUNCTION__); - if (!XF86DRIGetDrawableInfo(dpy, scrn, draw, index, stamp, + if (!XF86DRIGetDrawableInfo(ndpy, scrn, draw, index, stamp, x, y, width, height, numClipRects, pClipRects, backX, backY, @@ -570,17 +590,29 @@ load_dri_driver(struct xdri_egl_driver *xdri_drv) char filename[100]; int flags = RTLD_NOW; - snprintf(filename, sizeof(filename), "%s.so", xdri_drv->dri_driver_name); + /* try "egl_xxx_dri.so" first */ + snprintf(filename, sizeof(filename), "egl_%s.so", xdri_drv->dri_driver_name); _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s)", filename); - xdri_drv->dri_driver_handle = dlopen(filename, flags); - if (!xdri_drv->dri_driver_handle) { - _eglLog(_EGL_WARNING, "XDRI Could not open %s (%s)", - filename, dlerror()); + if (xdri_drv->dri_driver_handle) { + _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s) OK", filename); + return EGL_TRUE; + } + else { + _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s) fail (%s)", filename, dlerror()); + } - return EGL_FALSE; + /* try regular "xxx_dri.so" next */ + snprintf(filename, sizeof(filename), "%s.so", xdri_drv->dri_driver_name); + _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s)", filename); + xdri_drv->dri_driver_handle = dlopen(filename, flags); + if (xdri_drv->dri_driver_handle) { + _eglLog(_EGL_DEBUG, "XDRI: dlopen(%s) OK", filename); + return EGL_TRUE; } - return EGL_TRUE; + + _eglLog(_EGL_WARNING, "XDRI Could not open %s (%s)", filename, dlerror()); + return EGL_FALSE; } @@ -646,6 +678,24 @@ xdri_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) } +static _EGLProc +xdri_eglGetProcAddress(const char *procname) +{ +#if 0 + _EGLDriver *drv = NULL; + + struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv); + /*_EGLDisplay *disp = _eglLookupDisplay(dpy);*/ + _EGLProc *proc = xdri_drv->driScreen.getProcAddress(procname); + return proc; +#elif 0 + return (_EGLProc) st_get_proc_address(procname); +#else + return NULL; +#endif +} + + /** * Called via eglCreateContext(), drv->API.CreateContext(). */ @@ -724,6 +774,7 @@ xdri_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, _EGLDisplay *disp = _eglLookupDisplay(dpy); struct xdri_egl_surface *xdri_surf; int scrn = DefaultScreen(disp->Xdpy); + uint width, height; xdri_surf = CALLOC_STRUCT(xdri_egl_surface); if (!xdri_surf) @@ -744,6 +795,10 @@ xdri_eglCreateWindowSurface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, _eglSaveSurface(&xdri_surf->Base); + get_drawable_size(disp->Xdpy, window, &width, &height); + xdri_surf->Base.Width = width; + xdri_surf->Base.Height = height; + _eglLog(_EGL_DEBUG, "XDRI: CreateWindowSurface win 0x%x handle %d hDrawable %d", (int) window, _eglGetSurfaceHandle(&xdri_surf->Base), @@ -790,9 +845,8 @@ xdri_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) { struct xdri_egl_surface *xdri_surf = lookup_surface(draw); - __GLXdisplayPrivate *priv = __glXInitialize(disp->Xdpy); - __GLXscreenConfigs *scrn = priv->screenConfigs; - __DRIscreen *psc = &scrn->driScreen; + struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv); + __DRIscreen *psc = &xdri_drv->driScreen; __DRIdrawable * const pdraw = psc->getDrawable(disp->Xdpy, xdri_surf->driDrawable, psc->private); @@ -818,10 +872,15 @@ _eglMain(_EGLDisplay *disp, const char *args) if (!xdri_drv) return NULL; + /* XXX temp hack */ + TheDriver = xdri_drv; + _eglInitDriverFallbacks(&xdri_drv->Base); xdri_drv->Base.API.Initialize = xdri_eglInitialize; xdri_drv->Base.API.Terminate = xdri_eglTerminate; + xdri_drv->Base.API.GetProcAddress = xdri_eglGetProcAddress; + xdri_drv->Base.API.CreateContext = xdri_eglCreateContext; xdri_drv->Base.API.MakeCurrent = xdri_eglMakeCurrent; xdri_drv->Base.API.CreateWindowSurface = xdri_eglCreateWindowSurface; -- cgit v1.2.3 From e3b13dba4df0522cce11c55c34a9e3e7d0f3a32c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Jul 2008 15:49:59 -0600 Subject: egl: link libEGL w/ additional libs --- src/egl/drivers/xdri/Makefile | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/drivers/xdri/Makefile b/src/egl/drivers/xdri/Makefile index 477a75184a..252d00b6c2 100644 --- a/src/egl/drivers/xdri/Makefile +++ b/src/egl/drivers/xdri/Makefile @@ -27,6 +27,8 @@ OBJECTS = $(SOURCES:.c=.o) DRM_LIB = `pkg-config --libs libdrm` +MISC_LIBS = -ldl -lXext -lGL + .c.o: $(CC) -c $(INCLUDE_DIRS) $(CFLAGS) $< -o $@ @@ -47,7 +49,7 @@ $(TOP)/$(LIB_DIR)/$(DRIVER_NAME): $(OBJECTS) -noprefix \ -major 1 -minor 0 \ -install $(TOP)/$(LIB_DIR) \ - -ldl $(OBJECTS) $(DRM_LIB) + $(OBJECTS) $(DRM_LIB) $(MISC_LIBS) clean: -- cgit v1.2.3 From eb9efc6ad5544b8c643f81c03fcc09438aa0316c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Jul 2008 15:51:18 -0600 Subject: mesa: disable fprintf() error --- src/mesa/drivers/dri/common/utils.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/drivers/dri/common/utils.c b/src/mesa/drivers/dri/common/utils.c index 3cf2146dce..67cffead00 100644 --- a/src/mesa/drivers/dri/common/utils.c +++ b/src/mesa/drivers/dri/common/utils.c @@ -310,8 +310,10 @@ void driInitSingleExtension( GLcontext * ctx, */ offset = _glapi_add_dispatch( functions, parameter_signature ); if (offset == -1) { +#if 0 /* this causes noise with egl */ fprintf(stderr, "DISPATCH ERROR! _glapi_add_dispatch failed " "to add %s!\n", functions[0]); +#endif } else if (ext->functions[i].remap_index != -1) { driDispatchRemapTable[ ext->functions[i].remap_index ] = -- cgit v1.2.3 From 811a183a274048158af5f5a05c4d3106dcffd08b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Jul 2008 15:08:39 -0600 Subject: mesa: return -1, not GL_FALSE if _glapi_add_dispatch() fails name sanity check --- src/mesa/glapi/glapi_getproc.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/glapi/glapi_getproc.c b/src/mesa/glapi/glapi_getproc.c index 5e0af26a01..3634444c85 100644 --- a/src/mesa/glapi/glapi_getproc.c +++ b/src/mesa/glapi/glapi_getproc.c @@ -411,7 +411,7 @@ add_function_name( const char * funcName ) * \returns * The offset in the dispatch table of the named function. A pointer to the * driver's implementation of the named function should be stored at - * \c dispatch_table[\c offset]. + * \c dispatch_table[\c offset]. Return -1 if error/problem. * * \sa glXGetProcAddress * @@ -460,7 +460,7 @@ _glapi_add_dispatch( const char * const * function_names, */ if (!function_names[i] || function_names[i][0] != 'g' || function_names[i][1] != 'l') - return GL_FALSE; + return -1; /* Determine if the named function already exists. If the function does * exist, it must have the same parameter signature as the function -- cgit v1.2.3 From 64f92e00c8292113f9a6372959febe903af09db6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 9 Jul 2008 15:54:21 -0600 Subject: mesa: check for null shader->Source --- src/mesa/shader/slang/slang_compile.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 4acbe802d2..ccb04494bf 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -2167,6 +2167,9 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) type = SLANG_UNIT_FRAGMENT_SHADER; } + if (!shader->Source) + return GL_FALSE; + ctx->Shader.MemPool = _slang_new_mempool(1024*1024); /* XXX temporary hack */ -- cgit v1.2.3 From fb3867aec018ba0c0aa548db541236528400d934 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 10 Jul 2008 10:48:28 -0600 Subject: egl: implement xdri_eglGetProcAddress() for gallium Plus comments, clean-ups. --- src/egl/drivers/xdri/egl_xdri.c | 28 +++++++++++++++++++++++++--- 1 file changed, 25 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c index 5e50d6034a..9d95c850d0 100644 --- a/src/egl/drivers/xdri/egl_xdri.c +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -574,7 +574,8 @@ init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp) for (m = modes; m; m = m->next) { _eglLog(_EGL_DEBUG, - "mode ID 0x%x rgba %d %d %d %d z %d s %d db %d\n", m->visualID, + "mode ID 0x%x rgba %d %d %d %d z %d s %d db %d\n", + m->visualID, m->redBits, m->greenBits, m->blueBits, m->alphaBits, m->depthBits, m->stencilBits, m->doubleBufferMode); } @@ -584,6 +585,15 @@ init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp) } +/** + * Load the DRI driver named by "xdri_drv->dri_driver_name". + * Basically, dlopen() the library to set "xdri_drv->dri_driver_handle". + * + * Later, we'll call dlsym(createNewScreenName) to get a pointer to + * the driver's createNewScreen() function which is the bootstrap function. + * + * \return EGL_TRUE for success, EGL_FALSE for failure + */ static EGLBoolean load_dri_driver(struct xdri_egl_driver *xdri_drv) { @@ -678,6 +688,9 @@ xdri_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) } +/* + * Called from eglGetProcAddress() via drv->API.GetProcAddress(). + */ static _EGLProc xdri_eglGetProcAddress(const char *procname) { @@ -688,8 +701,17 @@ xdri_eglGetProcAddress(const char *procname) /*_EGLDisplay *disp = _eglLookupDisplay(dpy);*/ _EGLProc *proc = xdri_drv->driScreen.getProcAddress(procname); return proc; -#elif 0 - return (_EGLProc) st_get_proc_address(procname); +#elif 1 + /* This is a bit of a hack to get at the gallium/Mesa state tracker + * function st_get_proc_address(). This will probably change at + * some point. + */ + _EGLProc (*st_get_proc_addr)(const char *procname); + st_get_proc_addr = dlsym(NULL, "st_get_proc_address"); + if (st_get_proc_addr) { + return st_get_proc_addr(procname); + } + return NULL; #else return NULL; #endif -- cgit v1.2.3 From b101890c150d2ba17a42f5164f0de86914b9b28b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 10 Jul 2008 13:29:42 -0600 Subject: egl: add all EGL_*_BITs to ClientAPIsMask --- src/egl/drivers/xdri/egl_xdri.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c index 9d95c850d0..0f8496c0fb 100644 --- a/src/egl/drivers/xdri/egl_xdri.c +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -909,7 +909,10 @@ _eglMain(_EGLDisplay *disp, const char *args) xdri_drv->Base.API.DestroySurface = xdri_eglDestroySurface; xdri_drv->Base.API.SwapBuffers = xdri_eglSwapBuffers; - xdri_drv->Base.ClientAPIsMask = EGL_OPENGL_BIT | EGL_OPENGL_ES_BIT; + xdri_drv->Base.ClientAPIsMask = (EGL_OPENGL_BIT | + EGL_OPENGL_ES_BIT | + EGL_OPENGL_ES2_BIT | + EGL_OPENVG_BIT); xdri_drv->Base.Name = "X/DRI"; _eglLog(_EGL_DEBUG, "XDRI: main(%s)", args); -- cgit v1.2.3 From f5cdc657b244ec7840cff1f36892f105b4aa2ef2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 10 Jul 2008 13:30:37 -0600 Subject: egl: added EGL_OPENVG_API case (allow all APIs) --- src/gallium/winsys/egl_xlib/egl_xlib.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 708a58b781..829732eea8 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -350,6 +350,7 @@ xlib_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, /* API-dependent context creation */ switch (ctx->Base.ClientAPI) { + case EGL_OPENVG_API: case EGL_OPENGL_ES_API: _eglLog(_EGL_DEBUG, "Create Context for ES version %d\n", ctx->Base.ClientVersion); -- cgit v1.2.3 From bae90edfdae0eac6c844ede21285453ef96a4f6e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 10 Jul 2008 14:23:27 -0600 Subject: gallium: check for FEATURE_feedback and FEATURE_drawpix when creating/using the aux draw module --- src/mesa/state_tracker/st_cb_program.c | 4 ++++ src/mesa/state_tracker/st_context.c | 4 ++++ src/mesa/state_tracker/st_draw.c | 2 ++ 3 files changed, 10 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c index 11ba5d56f3..ea0fa20012 100644 --- a/src/mesa/state_tracker/st_cb_program.c +++ b/src/mesa/state_tracker/st_cb_program.c @@ -144,9 +144,11 @@ st_delete_program(GLcontext *ctx, struct gl_program *prog) } if (stvp->draw_shader) { +#if FEATURE_feedback || FEATURE_drawpix /* this would only have been allocated for the RasterPos path */ draw_delete_vertex_shader(st->draw, stvp->draw_shader); stvp->draw_shader = NULL; +#endif } if (stvp->state.tokens) { @@ -232,9 +234,11 @@ static void st_program_string_notify( GLcontext *ctx, } if (stvp->draw_shader) { +#if FEATURE_feedback || FEATURE_drawpix /* this would only have been allocated for the RasterPos path */ draw_delete_vertex_shader(st->draw, stvp->draw_shader); stvp->draw_shader = NULL; +#endif } if (stvp->state.tokens) { diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index bf78cacb8e..8b30e35684 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -98,6 +98,7 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) /* state tracker needs the VBO module */ _vbo_CreateContext(ctx); +#if FEATURE_feedback || FEATURE_drawpix st->draw = draw_create(); /* for selection/feedback */ /* Disable draw options that might convert points/lines to tris, etc. @@ -107,6 +108,7 @@ st_create_context_priv( GLcontext *ctx, struct pipe_context *pipe ) draw_wide_point_threshold(st->draw, 1000.0f); draw_enable_line_stipple(st->draw, FALSE); draw_enable_point_sprites(st->draw, FALSE); +#endif st->dirty.mesa = ~0; st->dirty.st = ~0; @@ -164,7 +166,9 @@ static void st_destroy_context_priv( struct st_context *st ) { uint i; +#if FEATURE_feedback || FEATURE_drawpix draw_destroy(st->draw); +#endif st_destroy_atoms( st ); st_destroy_draw( st ); st_destroy_generate_mipmap(st); diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index a57d7b6a7a..30c2bdb723 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -447,6 +447,7 @@ st_draw_vbo(GLcontext *ctx, } +#if FEATURE_feedback || FEATURE_drawpix /** * Set the (private) draw module's post-transformed vertex format when in @@ -657,6 +658,7 @@ st_feedback_draw_vbo(GLcontext *ctx, } } +#endif /* FEATURE_feedback || FEATURE_drawpix */ void st_init_draw( struct st_context *st ) -- cgit v1.2.3 From 6e938e4f82755c20472532907cc47c2d501aaee2 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Fri, 11 Jul 2008 20:01:33 +0200 Subject: gallium: Make dri drivers create a egl_name_dri.so if supported --- src/gallium/winsys/dri/Makefile.template | 10 +++++++++- src/gallium/winsys/dri/intel/Makefile | 1 + 2 files changed, 10 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/Makefile.template b/src/gallium/winsys/dri/Makefile.template index 07abfa53f3..80e817b808 100644 --- a/src/gallium/winsys/dri/Makefile.template +++ b/src/gallium/winsys/dri/Makefile.template @@ -79,17 +79,25 @@ SHARED_INCLUDES = \ ##### TARGETS ##### -default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) +default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR)/$(LIBNAME_EGL) $(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template $(TOP)/bin/mklib -noprefix -o $@ \ $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) +$(LIBNAME_EGL): $(WINSYS_OBJECTS) $(LIBS) + $(TOP)/bin/mklib -o $(LIBNAME_EGL) \ + -linker "$(CC)" \ + -noprefix \ + $(OBJECTS) $(MKLIB_OPTIONS) $(WINSYS_OBJECTS) $(PIPE_DRIVERS) $(WINOBJ) $(DRI_LIB_DEPS) \ + --whole-archive $(LIBS) $(GALLIUM_AUXILIARIES) --no-whole-archive $(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) $(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR) +$(TOP)/$(LIB_DIR)/$(LIBNAME_EGL): $(LIBNAME_EGL) + $(INSTALL) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR) depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) rm -f depend diff --git a/src/gallium/winsys/dri/intel/Makefile b/src/gallium/winsys/dri/intel/Makefile index 5b51f0815d..e0716ea28e 100644 --- a/src/gallium/winsys/dri/intel/Makefile +++ b/src/gallium/winsys/dri/intel/Makefile @@ -3,6 +3,7 @@ TOP = ../../../../.. include $(TOP)/configs/current LIBNAME = i915_dri.so +LIBNAME_EGL = egl_i915_dri.so PIPE_DRIVERS = \ $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ -- cgit v1.2.3 From 7986baf7cf5434c0ff8210eedd0be366ab5e8f14 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 10 Jul 2008 15:11:01 -0600 Subject: build egl subdirs --- src/egl/drivers/Makefile | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/Makefile b/src/egl/drivers/Makefile index a2d67ca5a8..6ce0e6b66d 100644 --- a/src/egl/drivers/Makefile +++ b/src/egl/drivers/Makefile @@ -7,10 +7,10 @@ include $(TOP)/configs/current SUBDIRS = demo dri xdri -default: conditional_subdirs +default: dri_subdirs -# depending on $DRIVER_DIRS... +# (UNUSED) depending on $DRIVER_DIRS... conditional_subdirs: @if [ "${DRIVER_DIRS}" = "dri" ] ; then \ $(MAKE) dri_subdirs ; \ -- cgit v1.2.3 From df2c9424d3b625d6e97528c74051257bc5630a67 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 11 Jul 2008 15:43:52 -0600 Subject: glx: added __glXPreferEGL() to tell libGL to prefer "egl_" drivers over regular DRI drivers. Also, clean-up, consolidate the dlopen() code a bit. --- src/glx/x11/dri_glx.c | 66 +++++++++++++++++++++++++++++++++++++------------ src/glx/x11/glxclient.h | 2 ++ 2 files changed, 52 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/glx/x11/dri_glx.c b/src/glx/x11/dri_glx.c index 21e07c1935..e637c96ac9 100644 --- a/src/glx/x11/dri_glx.c +++ b/src/glx/x11/dri_glx.c @@ -64,6 +64,20 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. static __DRIdriver *Drivers = NULL; +/** If non-zero, prefer an "egl_foo_dri.so" driver over "foo_dri.so" */ +static int PreferEGL = 0; + + +/** + * This may be called by libEGL.so to tell libGL to prefer "egl_" drivers + * over regular DRI drivers". + */ +void __glXPreferEGL(int state) +{ + PreferEGL = state; +} + + /* * printf wrappers */ @@ -160,6 +174,33 @@ ExtractDir(int index, const char *paths, int dirLen, char *dir) } +/** + * Try to dlopen() a driver. + * \param libDir the directory to search + * \param driverName the name of the driver (such as "r300"). + * \param tls if true, try to find TLS version of driver + * \param preferEGL if true, try to find EGL-specific driver + * \return handle from dlopen(), will be NULL if fails. + */ +static void * +try_open(const char *libDir, const char *driverName, + GLboolean tls, GLboolean preferEGL) +{ + const char *tlsDir = tls ? "/tls/" : ""; + const char *eglPrefix = preferEGL ? "egl_" : ""; + char fullPathName[200]; + void *handle = NULL; + + snprintf(fullPathName, 200, "%s%s/%s%s_dri.so", + libDir, tlsDir, eglPrefix, driverName); + InfoMessageF("OpenDriver: trying %s\n", fullPathName); + handle = dlopen(fullPathName, RTLD_NOW | RTLD_GLOBAL); + + return handle; +} + + + /** * Versioned name of the expected \c __driCreateNewScreen function. * @@ -214,24 +255,16 @@ static __DRIdriver *OpenDriver(const char *driverName) libPaths = DEFAULT_DRIVER_DIR; for ( i = 0 ; ExtractDir(i, libPaths, 1000, libDir) != 0 ; i++ ) { - char realDriverName[200]; void *handle = NULL; - - /* If TLS support is enabled, try to open the TLS version of the driver - * binary first. If that fails, try the non-TLS version. - */ + if (PreferEGL) + handle = try_open(libDir, driverName, GL_FALSE, GL_TRUE); #ifdef GLX_USE_TLS - snprintf(realDriverName, 200, "%s/tls/%s_dri.so", libDir, driverName); - InfoMessageF("OpenDriver: trying %s\n", realDriverName); - handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); + if (!handle) + handle = try_open(libDir, driverName, GL_TRUE, GL_FALSE); #endif - - if ( handle == NULL ) { - snprintf(realDriverName, 200, "%s/%s_dri.so", libDir, driverName); - InfoMessageF("OpenDriver: trying %s\n", realDriverName); - handle = dlopen(realDriverName, RTLD_NOW | RTLD_GLOBAL); - } + if (!handle) + handle = try_open(libDir, driverName, GL_FALSE, GL_FALSE); if ( handle != NULL ) { /* allocate __DRIdriver struct */ @@ -268,7 +301,8 @@ static __DRIdriver *OpenDriver(const char *driverName) break; } else { - ErrorMessageF("dlopen %s failed (%s)\n", realDriverName, dlerror()); + ErrorMessageF("Unable to find/open driver for %s (%s)\n", + driverName, dlerror()); } } @@ -402,7 +436,7 @@ static void driDestroyDisplay(Display *dpy, void *private) else Drivers = driver->next; - Xfree(driver->name); + Xfree((void *) driver->name); Xfree(driver); break; } diff --git a/src/glx/x11/glxclient.h b/src/glx/x11/glxclient.h index 03e44e5d04..f9d076f56f 100644 --- a/src/glx/x11/glxclient.h +++ b/src/glx/x11/glxclient.h @@ -541,6 +541,8 @@ extern void __glXSendLargeCommand(__GLXcontext *, const GLvoid *, GLint, /* Initialize the GLX extension for dpy */ extern __GLXdisplayPrivate *__glXInitialize(Display*); +extern void __glXPreferEGL(int state); + /************************************************************************/ extern int __glXDebug; -- cgit v1.2.3 From 80b24166a51f26e7bb341a731ceeb4f85c6cc4cf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 11 Jul 2008 15:44:46 -0600 Subject: egl: a minor overhauld of egl_xdri.c Rely more on the libGL code to avoid duplicated efforts. Also fix confusion arising from multiple __DRIscreen objects. --- src/egl/drivers/xdri/egl_xdri.c | 104 ++++++++++++++++------------------------ 1 file changed, 42 insertions(+), 62 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c index 0f8496c0fb..dfb491958b 100644 --- a/src/egl/drivers/xdri/egl_xdri.c +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -33,7 +33,7 @@ * to render into X-created windows. * * This is an EGL driver that, in turn, loads a regular DRI driver. - * There are some dependencies on code in libGL, but those coudl be + * There are some dependencies on code in libGL, but those could be * removed with some effort. * * Authors: Brian Paul @@ -77,11 +77,14 @@ struct xdri_egl_driver const char *dri_driver_name; /**< name of DRI driver to load */ void *dri_driver_handle; /**< returned by dlopen(dri_driver_name) */ + __GLXdisplayPrivate *glx_priv; + + + /* XXX we're not actually using these at this time: */ int chipset; int minor; int drmFD; - __DRIscreen driScreen; __DRIframebuffer framebuffer; drm_handle_t hSAREA; drmAddress pSAREA; @@ -118,9 +121,6 @@ struct xdri_egl_config }; -/* XXX temp hack */ -static struct xdri_egl_driver *TheDriver = NULL; - /** cast wrapper */ static struct xdri_egl_driver * @@ -179,11 +179,10 @@ get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height) * This dependency on GLX lib will be removed someday. */ static void -create_configs(_EGLDisplay *disp) +create_configs(_EGLDisplay *disp, __GLXdisplayPrivate *glx_priv) { + __GLXscreenConfigs *scrn = glx_priv->screenConfigs; const __GLcontextModes *m; - __GLXdisplayPrivate *priv = __glXInitialize(disp->Xdpy); - __GLXscreenConfigs *scrn = priv->screenConfigs; int id = 1; for (m = scrn->configs; m; m = m->next) { @@ -292,9 +291,9 @@ dri_context_modes_create(unsigned count, size_t minimum_size) static __DRIscreen * dri_find_dri_screen(__DRInativeDisplay *ndpy, int scrn) { - assert(TheDriver); - - return &TheDriver->driScreen; + __GLXdisplayPrivate *priv = __glXInitialize(ndpy); + __GLXscreenConfigs *scrnConf = priv->screenConfigs; + return &scrnConf->driScreen; } @@ -395,9 +394,6 @@ static const __DRIinterfaceMethods interface_methods = { static EGLBoolean init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp) { - static const char createNewScreenName[] = "__driCreateNewScreen_20050727"; - PFNCREATENEWSCREENFUNC createNewScreen; - int api_ver = 0;/*__glXGetInternalVersion();*/ __DRIversion ddx_version; __DRIversion dri_version; __DRIversion drm_version; @@ -405,9 +401,9 @@ init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp) drm_handle_t hFB; int newlyopened; int status; - __GLcontextModes *modes; int scrn = DefaultScreen(disp->Xdpy); +#if 0 createNewScreen = (PFNCREATENEWSCREENFUNC) dlsym(xdri_drv->dri_driver_handle, createNewScreenName); if (!createNewScreen) { @@ -418,6 +414,7 @@ init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp) else { _eglLog(_EGL_DEBUG, "XDRI: Found %s", createNewScreenName); } +#endif /* * Get the DRI X extension version. @@ -426,7 +423,6 @@ init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp) dri_version.minor = 0; dri_version.patch = 0; - if (!XF86DRIOpenConnection(disp->Xdpy, scrn, &xdri_drv->hSAREA, &xdri_drv->busID)) { _eglLog(_EGL_WARNING, "XF86DRIOpenConnection failed"); @@ -543,44 +539,6 @@ init_drm(struct xdri_egl_driver *xdri_drv, _EGLDisplay *disp) _eglLog(_EGL_DEBUG, "XDRI: drmMap(sarea) success"); } - /* Create the DRI screen. - */ - xdri_drv->driScreen.private = createNewScreen(disp->Xdpy, - scrn, /* screen number */ - &xdri_drv->driScreen, - NULL, /* visuals */ - &ddx_version, - &dri_version, - &drm_version, - &xdri_drv->framebuffer, - xdri_drv->pSAREA, - xdri_drv->drmFD, - api_ver, - &interface_methods, - &modes); - if (!xdri_drv->driScreen.private) { - _eglLog(_EGL_WARNING, "XDRI: create new screen failed"); - return EGL_FALSE; - } - else { - _eglLog(_EGL_DEBUG, "XDRI: create new screen success"); - } - - create_configs(disp); - - /* print modes / debug */ - if (0) { - __GLcontextModes *m; - - for (m = modes; m; m = m->next) { - _eglLog(_EGL_DEBUG, - "mode ID 0x%x rgba %d %d %d %d z %d s %d db %d\n", - m->visualID, - m->redBits, m->greenBits, m->blueBits, m->alphaBits, - m->depthBits, m->stencilBits, m->doubleBufferMode); - } - } - return EGL_TRUE; } @@ -647,13 +605,30 @@ xdri_eglInitialize(_EGLDriver *drv, EGLDisplay dpy, } } +#if 0 /* choose the DRI driver to load */ xdri_drv->dri_driver_name = _eglChooseDRMDriver(0); if (!load_dri_driver(xdri_drv)) return EGL_FALSE; +#else + (void) load_dri_driver; +#endif +#if 0 if (!init_drm(xdri_drv, disp)) return EGL_FALSE; +#else + (void) init_drm; +#endif + + /* + * NOTE: this call to __glXInitialize() bootstraps the whole GLX/DRI + * interface, loads the DRI driver, etc. + * This replaces the load_dri_driver() and init_drm() code above. + */ + xdri_drv->glx_priv = __glXInitialize(disp->Xdpy); + + create_configs(disp, xdri_drv->glx_priv); xdri_drv->Base.Initialized = EGL_TRUE; @@ -725,7 +700,6 @@ static EGLContext xdri_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list) { - struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv); _EGLDisplay *disp = _eglLookupDisplay(dpy); struct xdri_egl_config *xdri_config = lookup_config(drv, dpy, config); void *shared = NULL; @@ -742,10 +716,15 @@ xdri_eglCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, assert(xdri_config); - xdri_ctx->driContext.private = - xdri_drv->driScreen.createNewContext(disp->Xdpy, - xdri_config->mode, renderType, - shared, &xdri_ctx->driContext); + { + struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv); + __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs; + xdri_ctx->driContext.private = + scrnConf->driScreen.createNewContext(disp->Xdpy, + xdri_config->mode, renderType, + shared, &xdri_ctx->driContext); + } + if (!xdri_ctx->driContext.private) { _eglLog(_EGL_DEBUG, "driScreen.createNewContext failed"); free(xdri_ctx); @@ -868,7 +847,8 @@ xdri_eglSwapBuffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) { struct xdri_egl_surface *xdri_surf = lookup_surface(draw); struct xdri_egl_driver *xdri_drv = xdri_egl_driver(drv); - __DRIscreen *psc = &xdri_drv->driScreen; + __GLXscreenConfigs *scrnConf = xdri_drv->glx_priv->screenConfigs; + __DRIscreen *psc = &scrnConf->driScreen; __DRIdrawable * const pdraw = psc->getDrawable(disp->Xdpy, xdri_surf->driDrawable, psc->private); @@ -894,8 +874,8 @@ _eglMain(_EGLDisplay *disp, const char *args) if (!xdri_drv) return NULL; - /* XXX temp hack */ - TheDriver = xdri_drv; + /* Tell libGL to prefer the EGL drivers over regular DRI drivers */ + __glXPreferEGL(1); _eglInitDriverFallbacks(&xdri_drv->Base); xdri_drv->Base.API.Initialize = xdri_eglInitialize; -- cgit v1.2.3 From 7279d663e984ae8a243f56c010f175fee9ffccb3 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 12 Jul 2008 11:16:01 +0200 Subject: tgsi: Add missing copyright headers. --- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h | 30 +++++++++++++++++++++++++-- src/gallium/auxiliary/tgsi/util/tgsi_build.c | 28 ++++++++++++++++++++++++- src/gallium/auxiliary/tgsi/util/tgsi_build.h | 30 +++++++++++++++++++++++++-- src/gallium/auxiliary/tgsi/util/tgsi_dump.h | 31 +++++++++++++++++++++++++--- src/gallium/auxiliary/tgsi/util/tgsi_util.c | 28 ++++++++++++++++++++++++- src/gallium/auxiliary/tgsi/util/tgsi_util.h | 30 +++++++++++++++++++++++++-- 6 files changed, 166 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h index e66d115283..af838b2a25 100755 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h @@ -1,4 +1,31 @@ -#if !defined TGSI_SSE2_H +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_SSE2_H #define TGSI_SSE2_H #if defined __cplusplus @@ -20,4 +47,3 @@ tgsi_emit_sse2( #endif #endif /* TGSI_SSE2_H */ - diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.c b/src/gallium/auxiliary/tgsi/util/tgsi_build.c index 18e44b38c2..742ef14c35 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.c @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" @@ -1295,4 +1322,3 @@ tgsi_build_dst_register_ext_modulate( return dst_register_ext_modulate; } - diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.h b/src/gallium/auxiliary/tgsi/util/tgsi_build.h index 423cf141f5..ed25830248 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_build.h @@ -1,4 +1,31 @@ -#if !defined TGSI_BUILD_H +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_BUILD_H #define TGSI_BUILD_H #if defined __cplusplus @@ -303,4 +330,3 @@ tgsi_build_dst_register_ext_modulate( #endif #endif /* TGSI_BUILD_H */ - diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h index ca83bdef20..ba7692b511 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h @@ -1,4 +1,31 @@ -#if !defined TGSI_DUMP_H +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_DUMP_H #define TGSI_DUMP_H #if defined __cplusplus @@ -31,10 +58,8 @@ void tgsi_dump_declaration( const struct tgsi_full_declaration *decl ); - #if defined __cplusplus } #endif #endif /* TGSI_DUMP_H */ - diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_util.c b/src/gallium/auxiliary/tgsi/util/tgsi_util.c index 56a50d3b21..10762b6c1a 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_util.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_util.c @@ -1,3 +1,30 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + #include "pipe/p_debug.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" @@ -271,4 +298,3 @@ tgsi_util_set_full_src_register_sign_mode( assert( 0 ); } } - diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_util.h b/src/gallium/auxiliary/tgsi/util/tgsi_util.h index 45f5f0be0e..7877f34558 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_util.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_util.h @@ -1,4 +1,31 @@ -#if !defined TGSI_UTIL_H +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_UTIL_H #define TGSI_UTIL_H #if defined __cplusplus @@ -67,4 +94,3 @@ tgsi_util_set_full_src_register_sign_mode( #endif #endif /* TGSI_UTIL_H */ - -- cgit v1.2.3 From 9ea485f8865404e3e2e10cdb3b1627e7194c27fe Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 12 Jul 2008 17:03:30 +0200 Subject: tgsi: Fix dumping of indirect addressing. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index a395c630cc..6356b6de06 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -805,10 +805,12 @@ tgsi_dump_instruction( CHR( '[' ); if (src->SrcRegister.Indirect) { - TXT( "addr" ); - if (src->SrcRegister.Index > 0) - CHR( '+' ); - SID( src->SrcRegister.Index ); + TXT( "ADDR[0]" ); + if (src->SrcRegister.Index != 0) { + if (src->SrcRegister.Index > 0) + CHR( '+' ); + SID( src->SrcRegister.Index ); + } } else SID( src->SrcRegister.Index ); -- cgit v1.2.3 From d0386d55ff257ab09475178d058ddcd9f1e37c2d Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 12 Jul 2008 17:06:37 +0200 Subject: tgsi: Add tgsi_text utility module. Translates textual shader into a binary token stream. The syntax matches the tgsi_dump module, so it's possible to simply copy-paste the shader dump and transform it back to a binary form. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 580 ++++++++++++++++++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_text.h | 47 +++ 2 files changed, 627 insertions(+) create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_text.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_text.h (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c new file mode 100644 index 0000000000..bb365d1efe --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -0,0 +1,580 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "tgsi_text.h" +#include "tgsi_build.h" +#include "tgsi_parse.h" +#include "tgsi_util.h" + +static boolean is_alpha_underscore( const char *cur ) +{ + return + (*cur >= 'a' && *cur <= 'z') || + (*cur >= 'A' && *cur <= 'Z') || + *cur == '_'; +} + +static boolean is_digit( const char *cur ) +{ + return *cur >= '0' && *cur <= '9'; +} + +static boolean is_digit_alpha_underscore( const char *cur ) +{ + return is_digit( cur ) || is_alpha_underscore( cur ); +} + +/* Eat zero or more whitespaces. + */ +static void eat_opt_white( const char **pcur ) +{ + while (**pcur == ' ' || **pcur == '\t' || **pcur == '\n') + (*pcur)++; +} + +/* Eat one or more whitespaces. + * Return TRUE if at least one whitespace eaten. + */ +static boolean eat_white( const char **pcur ) +{ + const char *cur = *pcur; + + eat_opt_white( pcur ); + return *pcur > cur; +} + +/* Parse unsigned integer. + * No checks for overflow. + */ +static boolean parse_uint( const char **pcur, uint *val ) +{ + const char *cur = *pcur; + + if (is_digit( cur )) { + *val = *cur++ - '0'; + while (is_digit( cur )) + *val = *val * 10 + *cur++ - '0'; + *pcur = cur; + return TRUE; + } + return FALSE; +} + +struct translate_ctx +{ + const char *text; + const char *cur; + struct tgsi_token *tokens; + struct tgsi_token *tokens_cur; + struct tgsi_token *tokens_end; + struct tgsi_header *header; +}; + +static void report_error( struct translate_ctx *ctx, const char *msg ) +{ + debug_printf( "\nError: %s", msg ); +} + +/* Parse shader header. + * Return TRUE for one of the following headers. + * FRAG1.1 + * GEOM1.1 + * VERT1.1 + */ +static boolean parse_header( struct translate_ctx *ctx ) +{ + uint processor; + + if (ctx->cur[0] == 'F' && ctx->cur[1] == 'R' && ctx->cur[2] == 'A' && ctx->cur[3] == 'G') { + ctx->cur += 4; + processor = TGSI_PROCESSOR_FRAGMENT; + } + else if (ctx->cur[0] == 'V' && ctx->cur[1] == 'E' && ctx->cur[2] == 'R' && ctx->cur[3] == 'T') { + ctx->cur += 4; + processor = TGSI_PROCESSOR_VERTEX; + } + else if (ctx->cur[0] == 'G' && ctx->cur[1] == 'E' && ctx->cur[2] == 'O' && ctx->cur[3] == 'M') { + ctx->cur += 4; + processor = TGSI_PROCESSOR_GEOMETRY; + } + else { + report_error( ctx, "Unknown processor type" ); + return FALSE; + } + + if (ctx->cur[0] == '1' && ctx->cur[1] == '.' && ctx->cur[2] == '1') { + ctx->cur += 3; + } + else { + report_error( ctx, "Unknown version" ); + return FALSE; + } + + if (ctx->tokens_cur >= ctx->tokens_end) + return FALSE; + *(struct tgsi_version *) ctx->tokens_cur++ = tgsi_build_version(); + + if (ctx->tokens_cur >= ctx->tokens_end) + return FALSE; + ctx->header = (struct tgsi_header *) ctx->tokens_cur++; + *ctx->header = tgsi_build_header(); + + if (ctx->tokens_cur >= ctx->tokens_end) + return FALSE; + *(struct tgsi_processor *) ctx->tokens_cur++ = tgsi_build_processor( processor, ctx->header ); + + return TRUE; +} + +static boolean parse_label( struct translate_ctx *ctx, uint *val ) +{ + const char *cur = ctx->cur; + + if (parse_uint( &cur, val )) { + eat_opt_white( &cur ); + if (*cur == ':') { + cur++; + ctx->cur = cur; + return TRUE; + } + } + return FALSE; +} + +static const char *file_names[TGSI_FILE_COUNT] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + +static boolean +parse_file( + struct translate_ctx *ctx, + uint *file ) +{ + uint i; + + for (i = 0; i < TGSI_FILE_COUNT; i++) { + const char *cur = ctx->cur; + const char *name = file_names[i]; + + while (*name != '\0' && *name == toupper( *cur )) { + name++; + cur++; + } + if (*name == '\0' && !is_digit_alpha_underscore( cur )) { + ctx->cur = cur; + *file = i; + return TRUE; + } + } + report_error( ctx, "Unknown register file" ); + return FALSE; +} + +static boolean +parse_register( + struct translate_ctx *ctx, + uint *file, + uint *index ) +{ + if (!parse_file( ctx, file )) + return FALSE; + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '[') { + report_error( ctx, "Expected `['" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + if (!parse_uint( &ctx->cur, index )) { + report_error( ctx, "Expected literal integer" ); + return FALSE; + } + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ']') { + report_error( ctx, "Expected `]'" ); + return FALSE; + } + ctx->cur++; + /* TODO: Modulate suffix */ + return TRUE; +} + +static boolean +parse_dst_operand( + struct translate_ctx *ctx, + struct tgsi_full_dst_register *dst ) +{ + const char *cur; + uint file; + uint index; + + if (!parse_register( ctx, &file, &index )) + return FALSE; + dst->DstRegister.File = file; + dst->DstRegister.Index = index; + + /* Parse optional write mask. + */ + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == '.') { + uint writemask = TGSI_WRITEMASK_NONE; + + cur++; + eat_opt_white( &cur ); + if (toupper( *cur ) == 'X') { + cur++; + writemask |= TGSI_WRITEMASK_X; + } + if (toupper( *cur ) == 'Y') { + cur++; + writemask |= TGSI_WRITEMASK_Y; + } + if (toupper( *cur ) == 'Z') { + cur++; + writemask |= TGSI_WRITEMASK_Z; + } + if (toupper( *cur ) == 'W') { + cur++; + writemask |= TGSI_WRITEMASK_W; + } + + if (writemask == TGSI_WRITEMASK_NONE) { + report_error( ctx, "Writemask expected" ); + return FALSE; + } + + dst->DstRegister.WriteMask = writemask; + ctx->cur = cur; + } + return TRUE; +} + +static boolean +parse_src_operand( + struct translate_ctx *ctx, + struct tgsi_full_src_register *src ) +{ + const char *cur; + uint file; + uint index; + + /* TODO: Extended register modifiers */ + if (*ctx->cur == '-') { + ctx->cur++; + src->SrcRegister.Negate = 1; + eat_opt_white( &ctx->cur ); + } + + if (!parse_register( ctx, &file, &index )) + return FALSE; + src->SrcRegister.File = file; + src->SrcRegister.Index = index; + + /* Parse optional swizzle + */ + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == '.') { + uint i; + + cur++; + eat_opt_white( &cur ); + for (i = 0; i < 4; i++) { + uint swizzle; + + if (toupper( *cur ) == 'X') + swizzle = TGSI_SWIZZLE_X; + else if (toupper( *cur ) == 'Y') + swizzle = TGSI_SWIZZLE_Y; + else if (toupper( *cur ) == 'Z') + swizzle = TGSI_SWIZZLE_Z; + else if (toupper( *cur ) == 'W') + swizzle = TGSI_SWIZZLE_W; + else { + report_error( ctx, "Expected register swizzle component either `x', `y', `z' or `w'" ); + return FALSE; + } + cur++; + tgsi_util_set_src_register_swizzle( &src->SrcRegister, swizzle, i ); + } + + ctx->cur = cur; + } + return TRUE; +} + +struct opcode_info +{ + uint num_dst; + uint num_src; + const char *mnemonic; +}; + +static const struct opcode_info opcode_info[TGSI_OPCODE_LAST] = +{ +{ 1, 1, "ARL" }, + { 1, 1, "MOV" }, +{ 1, 1, "LIT" }, + { 1, 1, "RCP" }, + { 1, 1, "RSQ" }, + { 1, 1, "EXP" }, + { 1, 1, "LOG" }, + { 1, 2, "MUL" }, + { 1, 2, "ADD" }, + { 1, 2, "DP3" }, + { 1, 2, "DP4" }, +{ 1, 2, "DST" }, + { 1, 2, "MIN" }, + { 1, 2, "MAX" }, +{ 1, 2, "SLT" }, +{ 1, 2, "SGE" }, + { 1, 3, "MAD" }, + { 1, 2, "SUB" }, +{ 1, 3, "LERP" }, +{ 1, 2, "CND" }, +{ 1, 2, "CND0" }, +{ 1, 2, "DOT2ADD" }, +{ 1, 2, "INDEX" }, +{ 1, 2, "NEGATE" }, +{ 1, 2, "FRAC" }, +{ 1, 2, "CLAMP" }, +{ 1, 2, "FLOOR" }, +{ 1, 2, "ROUND" }, +{ 1, 2, "EXPBASE2" }, +{ 1, 2, "LOGBASE2" }, +{ 1, 2, "POWER" }, +{ 1, 2, "CROSSPRODUCT" }, +{ 1, 2, "MULTIPLYMATRIX" }, + { 1, 2, "ABS" }, +{ 1, 2, "RCC" }, +{ 1, 2, "DPH" }, +{ 1, 2, "COS" }, +{ 1, 2, "DDX" }, +{ 1, 2, "DDY" }, +{ 1, 2, "KILP" }, +{ 1, 2, "PK2H" }, +{ 1, 2, "PK2US" }, +{ 1, 2, "PK4B" }, +{ 1, 2, "PK4UB" }, +{ 1, 2, "RFL" }, +{ 1, 2, "SEQ" }, +{ 1, 2, "SFL" }, +{ 1, 2, "SGT" }, +{ 1, 2, "SIN" }, +{ 1, 2, "SLE" }, +{ 1, 2, "SNE" }, +{ 1, 2, "STR" }, +{ 1, 2, "TEX" }, +{ 1, 2, "TXD" }, +{ 1, 2, "TXP" }, +{ 1, 2, "UP2H" }, +{ 1, 2, "UP2US" }, +{ 1, 2, "UP4B" }, +{ 1, 2, "UP4UB" }, +{ 1, 2, "X2D" }, +{ 1, 2, "ARA" }, +{ 1, 2, "ARR" }, +{ 1, 2, "BRA" }, +{ 1, 2, "CAL" }, +{ 1, 2, "RET" }, +{ 1, 2, "SSG" }, +{ 1, 2, "CMP" }, +{ 1, 2, "SCS" }, +{ 1, 2, "TXB" }, +{ 1, 2, "NRM" }, +{ 1, 2, "DIV" }, +{ 1, 2, "DP2" }, +{ 1, 2, "TXL" }, +{ 1, 2, "BRK" }, +{ 1, 2, "IF" }, +{ 1, 2, "LOOP" }, +{ 1, 2, "REP" }, +{ 1, 2, "ELSE" }, +{ 1, 2, "ENDIF" }, +{ 1, 2, "ENDLOOP" }, +{ 1, 2, "ENDREP" }, +{ 1, 2, "PUSHA" }, +{ 1, 2, "POPA" }, +{ 1, 2, "CEIL" }, +{ 1, 2, "I2F" }, +{ 1, 2, "NOT" }, +{ 1, 2, "TRUNC" }, +{ 1, 2, "SHL" }, +{ 1, 2, "SHR" }, +{ 1, 2, "AND" }, +{ 1, 2, "OR" }, +{ 1, 2, "MOD" }, +{ 1, 2, "XOR" }, +{ 1, 2, "SAD" }, +{ 1, 2, "TXF" }, +{ 1, 2, "TXQ" }, +{ 1, 2, "CONT" }, +{ 1, 2, "EMIT" }, +{ 1, 2, "ENDPRIM" }, +{ 1, 2, "BGNLOOP2" }, +{ 1, 2, "BGNSUB" }, +{ 1, 2, "ENDLOOP2" }, +{ 1, 2, "ENDSUB" }, +{ 1, 2, "NOISE1" }, +{ 1, 2, "NOISE2" }, +{ 1, 2, "NOISE3" }, +{ 1, 2, "NOISE4" }, +{ 1, 2, "NOP" }, +{ 1, 2, "M4X3" }, +{ 1, 2, "M3X4" }, +{ 1, 2, "M3X3" }, +{ 1, 2, "M3X2" }, +{ 1, 2, "NRM4" }, +{ 1, 2, "CALLNZ" }, +{ 1, 2, "IFC" }, +{ 1, 2, "BREAKC" }, +{ 1, 2, "KIL" }, + { 0, 0, "END" } +}; + +static boolean parse_instruction( struct translate_ctx *ctx ) +{ + uint i; + const struct opcode_info *info; + struct tgsi_full_instruction inst; + uint advance; + + /* Parse instruction name. + */ + eat_opt_white( &ctx->cur ); + for (i = 0; i < TGSI_OPCODE_LAST; i++) { + const char *cur = ctx->cur; + const char *op = opcode_info[i].mnemonic; + + while (*op != '\0' && *op == toupper( *cur )) { + op++; + cur++; + } + if (*op == '\0') { + /* TODO: _SAT suffix */ + if (*cur == '\0' || eat_white( &cur )) { + ctx->cur = cur; + break; + } + } + } + if (i == TGSI_OPCODE_LAST) { + report_error( ctx, "Unknown opcode" ); + return FALSE; + } + info = &opcode_info[i]; + + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = i; + inst.Instruction.NumDstRegs = info->num_dst; + inst.Instruction.NumSrcRegs = info->num_src; + + /* Parse instruction operands. + */ + for (i = 0; i < info->num_dst + info->num_src; i++) { + if (i > 0) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ',') { + report_error( ctx, "Expected `,'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + } + + if (i < info->num_dst) { + if (!parse_dst_operand( ctx, &inst.FullDstRegisters[i] )) + return FALSE; + } + else { + if (!parse_src_operand( ctx, &inst.FullSrcRegisters[i - info->num_dst] )) + return FALSE; + } + } + eat_opt_white( &ctx->cur ); + + advance = tgsi_build_full_instruction( + &inst, + ctx->tokens_cur, + ctx->header, + (uint) (ctx->tokens_end - ctx->tokens_cur) ); + if (advance == 0) + return FALSE; + ctx->tokens_cur += advance; + + return TRUE; +} + +static boolean translate( struct translate_ctx *ctx ) +{ + eat_opt_white( &ctx->cur ); + if (!parse_header( ctx )) + return FALSE; + + eat_white( &ctx->cur ); + while (*ctx->cur != '\0') { + uint label_val = 0; + + if (parse_label( ctx, &label_val )) { + if (!parse_instruction( ctx )) + return FALSE; + } + else { + report_error( ctx, "Instruction expected" ); + return FALSE; + } + } + + return TRUE; +} + +boolean +tgsi_text_translate( + const char *text, + struct tgsi_token *tokens, + uint num_tokens ) +{ + struct translate_ctx ctx; + + ctx.text = text; + ctx.cur = text; + ctx.tokens = tokens; + ctx.tokens_cur = tokens; + ctx.tokens_end = tokens + num_tokens; + + return translate( &ctx ); +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.h b/src/gallium/auxiliary/tgsi/util/tgsi_text.h new file mode 100644 index 0000000000..8eeeeef140 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_TEXT_H +#define TGSI_TEXT_H + +#include "pipe/p_shader_tokens.h" + +#if defined __cplusplus +extern "C" { +#endif + +boolean +tgsi_text_translate( + const char *text, + struct tgsi_token *tokens, + uint num_tokens ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_TEXT_H */ -- cgit v1.2.3 From c415de5e251eb4004b1ef5bc57299032d95c4842 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 12 Jul 2008 17:10:21 +0200 Subject: scons: List `util/tgsi_text.c'. --- src/gallium/auxiliary/tgsi/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript index 4632dcc072..b62c8efe02 100644 --- a/src/gallium/auxiliary/tgsi/SConscript +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -9,6 +9,7 @@ tgsi = env.ConvenienceLibrary( 'util/tgsi_dump.c', 'util/tgsi_parse.c', 'util/tgsi_scan.c', + 'util/tgsi_text.c', 'util/tgsi_transform.c', 'util/tgsi_util.c', ]) -- cgit v1.2.3 From 92d711e9e6c1934e1cec774bfa4581869530cda6 Mon Sep 17 00:00:00 2001 From: Stephane Marchesin Date: Sun, 13 Jul 2008 11:33:41 +0200 Subject: llvm: build fixes. --- src/gallium/auxiliary/draw/draw_vs_llvm.c | 10 +++++----- src/gallium/auxiliary/gallivm/instructions.h | 2 +- 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index 621472ec7c..c63bd51a10 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -119,7 +119,7 @@ draw_create_vs_llvm(struct draw_context *draw, vs->base.create_varient = draw_vs_varient_generic; vs->base.run_linear = vs_llvm_run_linear; vs->base.delete = vs_llvm_delete; - vs->machine = &draw->machine; + vs->machine = &draw->vs.machine; { struct gallivm_ir *ir = gallivm_ir_new(GALLIVM_VS); @@ -130,16 +130,16 @@ draw_create_vs_llvm(struct draw_context *draw, gallivm_ir_delete(ir); } - draw->engine = gallivm_global_cpu_engine(); + draw->vs.engine = gallivm_global_cpu_engine(); /* XXX: Why are there two versions of this? Shouldn't creating the * engine be a separate operation to compiling a shader? */ - if (!draw->engine) { - draw->engine = gallivm_cpu_engine_create(vs->llvm_prog); + if (!draw->vs.engine) { + draw->vs.engine = gallivm_cpu_engine_create(vs->llvm_prog); } else { - gallivm_cpu_jit_compile(draw->engine, vs->llvm_prog); + gallivm_cpu_jit_compile(draw->vs.engine, vs->llvm_prog); } return &vs->base; diff --git a/src/gallium/auxiliary/gallivm/instructions.h b/src/gallium/auxiliary/gallivm/instructions.h index 19ca84ddc6..3a476928b6 100644 --- a/src/gallium/auxiliary/gallivm/instructions.h +++ b/src/gallium/auxiliary/gallivm/instructions.h @@ -85,7 +85,7 @@ public: llvm::Value *lit(llvm::Value *in); llvm::Value *lg2(llvm::Value *in); llvm::Value *madd(llvm::Value *in1, llvm::Value *in2, - llvm::Value *in2); + llvm::Value *in3); llvm::Value *min(llvm::Value *in1, llvm::Value *in2); llvm::Value *max(llvm::Value *in1, llvm::Value *in2); llvm::Value *mul(llvm::Value *in1, llvm::Value *in2); -- cgit v1.2.3 From bd3b47590e2a8e91a8f5545d7c8872b26879c228 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 11:30:02 +0200 Subject: tgsi: Remove depricated ATTRIB interpolate mode. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 3 +-- src/gallium/include/pipe/p_shader_tokens.h | 1 + 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 6356b6de06..59e7aaeceb 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -119,8 +119,7 @@ static const char *TGSI_INTERPOLATES_SHORT[] = { "CONSTANT", "LINEAR", - "PERSPECTIVE", - "ATTRIB" + "PERSPECTIVE" }; static const char *TGSI_SEMANTICS[] = diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 84e5096418..33268553ff 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -75,6 +75,7 @@ enum tgsi_file_type { #define TGSI_INTERPOLATE_CONSTANT 0 #define TGSI_INTERPOLATE_LINEAR 1 #define TGSI_INTERPOLATE_PERSPECTIVE 2 +#define TGSI_INTERPOLATE_COUNT 3 struct tgsi_declaration { -- cgit v1.2.3 From ee647b9020b5e16b9b6d399edfa4a2c99f491863 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 11:37:36 +0200 Subject: tgsi: Parse DCL statements. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 330 +++++++++++++++++++++------- 1 file changed, 256 insertions(+), 74 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index bb365d1efe..03c9217243 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -49,6 +49,21 @@ static boolean is_digit_alpha_underscore( const char *cur ) return is_digit( cur ) || is_alpha_underscore( cur ); } +static boolean str_match_no_case( const char **pcur, const char *str ) +{ + const char *cur = *pcur; + + while (*str != '\0' && *str == toupper( *cur )) { + str++; + cur++; + } + if (*str == '\0') { + *pcur = cur; + return TRUE; + } + return FALSE; +} + /* Eat zero or more whitespaces. */ static void eat_opt_white( const char **pcur ) @@ -110,28 +125,14 @@ static boolean parse_header( struct translate_ctx *ctx ) { uint processor; - if (ctx->cur[0] == 'F' && ctx->cur[1] == 'R' && ctx->cur[2] == 'A' && ctx->cur[3] == 'G') { - ctx->cur += 4; + if (str_match_no_case( &ctx->cur, "FRAG1.1" )) processor = TGSI_PROCESSOR_FRAGMENT; - } - else if (ctx->cur[0] == 'V' && ctx->cur[1] == 'E' && ctx->cur[2] == 'R' && ctx->cur[3] == 'T') { - ctx->cur += 4; + else if (str_match_no_case( &ctx->cur, "VERT1.1" )) processor = TGSI_PROCESSOR_VERTEX; - } - else if (ctx->cur[0] == 'G' && ctx->cur[1] == 'E' && ctx->cur[2] == 'O' && ctx->cur[3] == 'M') { - ctx->cur += 4; + else if (str_match_no_case( &ctx->cur, "GEOM1.1" )) processor = TGSI_PROCESSOR_GEOMETRY; - } - else { - report_error( ctx, "Unknown processor type" ); - return FALSE; - } - - if (ctx->cur[0] == '1' && ctx->cur[1] == '.' && ctx->cur[2] == '1') { - ctx->cur += 3; - } else { - report_error( ctx, "Unknown version" ); + report_error( ctx, "Unknown header" ); return FALSE; } @@ -187,16 +188,13 @@ parse_file( for (i = 0; i < TGSI_FILE_COUNT; i++) { const char *cur = ctx->cur; - const char *name = file_names[i]; - while (*name != '\0' && *name == toupper( *cur )) { - name++; - cur++; - } - if (*name == '\0' && !is_digit_alpha_underscore( cur )) { - ctx->cur = cur; - *file = i; - return TRUE; + if (str_match_no_case( &cur, file_names[i] )) { + if (!is_digit_alpha_underscore( cur )) { + ctx->cur = cur; + *file = i; + return TRUE; + } } } report_error( ctx, "Unknown register file" ); @@ -204,7 +202,53 @@ parse_file( } static boolean -parse_register( +parse_opt_writemask( + struct translate_ctx *ctx, + uint *writemask ) +{ + const char *cur; + + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == '.') { + cur++; + *writemask = TGSI_WRITEMASK_NONE; + eat_opt_white( &cur ); + if (toupper( *cur ) == 'X') { + cur++; + *writemask |= TGSI_WRITEMASK_X; + } + if (toupper( *cur ) == 'Y') { + cur++; + *writemask |= TGSI_WRITEMASK_Y; + } + if (toupper( *cur ) == 'Z') { + cur++; + *writemask |= TGSI_WRITEMASK_Z; + } + if (toupper( *cur ) == 'W') { + cur++; + *writemask |= TGSI_WRITEMASK_W; + } + + if (*writemask == TGSI_WRITEMASK_NONE) { + report_error( ctx, "Writemask expected" ); + return FALSE; + } + + ctx->cur = cur; + } + else { + *writemask = TGSI_WRITEMASK_XYZW; + } + return TRUE; +} + +/* Parse register part common for decls and operands. + * ::= `[' + */ +static boolean +parse_register_prefix( struct translate_ctx *ctx, uint *file, uint *index ) @@ -222,6 +266,20 @@ parse_register( report_error( ctx, "Expected literal integer" ); return FALSE; } + return TRUE; +} + +/* Parse register operand. + * ::= `]' + */ +static boolean +parse_register( + struct translate_ctx *ctx, + uint *file, + uint *index ) +{ + if (!parse_register_prefix( ctx, file, index )) + return FALSE; eat_opt_white( &ctx->cur ); if (*ctx->cur != ']') { report_error( ctx, "Expected `]'" ); @@ -232,54 +290,55 @@ parse_register( return TRUE; } +/* Parse register declaration. + * ::= `]' | `..' `]' + */ +static boolean +parse_register_dcl( + struct translate_ctx *ctx, + uint *file, + uint *first, + uint *last ) +{ + if (!parse_register_prefix( ctx, file, first )) + return FALSE; + eat_opt_white( &ctx->cur ); + if (ctx->cur[0] == '.' && ctx->cur[1] == '.') { + ctx->cur += 2; + eat_opt_white( &ctx->cur ); + if (!parse_uint( &ctx->cur, last )) { + report_error( ctx, "Expected literal integer" ); + return FALSE; + } + eat_opt_white( &ctx->cur ); + } + else { + *last = *first; + } + if (*ctx->cur != ']') { + report_error( ctx, "Expected `]' or `..'" ); + return FALSE; + } + ctx->cur++; + return TRUE; +} + static boolean parse_dst_operand( struct translate_ctx *ctx, struct tgsi_full_dst_register *dst ) { - const char *cur; uint file; uint index; + uint writemask; if (!parse_register( ctx, &file, &index )) return FALSE; + if (!parse_opt_writemask( ctx, &writemask )) + return FALSE; dst->DstRegister.File = file; dst->DstRegister.Index = index; - - /* Parse optional write mask. - */ - cur = ctx->cur; - eat_opt_white( &cur ); - if (*cur == '.') { - uint writemask = TGSI_WRITEMASK_NONE; - - cur++; - eat_opt_white( &cur ); - if (toupper( *cur ) == 'X') { - cur++; - writemask |= TGSI_WRITEMASK_X; - } - if (toupper( *cur ) == 'Y') { - cur++; - writemask |= TGSI_WRITEMASK_Y; - } - if (toupper( *cur ) == 'Z') { - cur++; - writemask |= TGSI_WRITEMASK_Z; - } - if (toupper( *cur ) == 'W') { - cur++; - writemask |= TGSI_WRITEMASK_W; - } - - if (writemask == TGSI_WRITEMASK_NONE) { - report_error( ctx, "Writemask expected" ); - return FALSE; - } - - dst->DstRegister.WriteMask = writemask; - ctx->cur = cur; - } + dst->DstRegister.WriteMask = writemask; return TRUE; } @@ -478,15 +537,10 @@ static boolean parse_instruction( struct translate_ctx *ctx ) eat_opt_white( &ctx->cur ); for (i = 0; i < TGSI_OPCODE_LAST; i++) { const char *cur = ctx->cur; - const char *op = opcode_info[i].mnemonic; - while (*op != '\0' && *op == toupper( *cur )) { - op++; - cur++; - } - if (*op == '\0') { + if (str_match_no_case( &cur, opcode_info[i].mnemonic )) { /* TODO: _SAT suffix */ - if (*cur == '\0' || eat_white( &cur )) { + if (eat_white( &cur )) { ctx->cur = cur; break; } @@ -525,7 +579,6 @@ static boolean parse_instruction( struct translate_ctx *ctx ) return FALSE; } } - eat_opt_white( &ctx->cur ); advance = tgsi_build_full_instruction( &inst, @@ -539,22 +592,151 @@ static boolean parse_instruction( struct translate_ctx *ctx ) return TRUE; } +static const char *semantic_names[TGSI_SEMANTIC_COUNT] = +{ + "POSITION", + "COLOR", + "BCOLOR", + "FOG", + "PSIZE", + "GENERIC", + "NORMAL" +}; + +static const char *interpolate_names[TGSI_INTERPOLATE_COUNT] = +{ + "CONSTANT", + "LINEAR", + "PERSPECTIVE" +}; + +static boolean parse_declaration( struct translate_ctx *ctx ) +{ + struct tgsi_full_declaration decl; + uint file; + uint first; + uint last; + uint writemask; + const char *cur; + uint advance; + + if (!eat_white( &ctx->cur )) { + report_error( ctx, "Syntax error" ); + return FALSE; + } + if (!parse_register_dcl( ctx, &file, &first, &last )) + return FALSE; + if (!parse_opt_writemask( ctx, &writemask )) + return FALSE; + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = file; + decl.Declaration.UsageMask = writemask; + decl.DeclarationRange.First = first; + decl.DeclarationRange.Last = last; + + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == ',') { + uint i; + + cur++; + eat_opt_white( &cur ); + for (i = 0; i < TGSI_SEMANTIC_COUNT; i++) { + if (str_match_no_case( &cur, semantic_names[i] )) { + const char *cur2 = cur; + uint index; + + if (is_digit_alpha_underscore( cur )) + continue; + eat_opt_white( &cur2 ); + if (*cur2 == '[') { + cur2++; + eat_opt_white( &cur2 ); + if (!parse_uint( &cur2, &index )) { + report_error( ctx, "Expected literal integer" ); + return FALSE; + } + eat_opt_white( &cur2 ); + if (*cur2 != ']') { + report_error( ctx, "Expected `]'" ); + return FALSE; + } + cur2++; + + decl.Semantic.SemanticIndex = index; + + cur = cur2; + } + + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = i; + + ctx->cur = cur; + break; + } + } + } + + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == ',') { + uint i; + + cur++; + eat_opt_white( &cur ); + for (i = 0; i < TGSI_INTERPOLATE_COUNT; i++) { + if (str_match_no_case( &cur, interpolate_names[i] )) { + if (is_digit_alpha_underscore( cur )) + continue; + decl.Declaration.Interpolate = i; + + ctx->cur = cur; + break; + } + } + if (i == TGSI_INTERPOLATE_COUNT) { + report_error( ctx, "Expected semantic or interpolate attribute" ); + return FALSE; + } + } + + advance = tgsi_build_full_declaration( + &decl, + ctx->tokens_cur, + ctx->header, + (uint) (ctx->tokens_end - ctx->tokens_cur) ); + if (advance == 0) + return FALSE; + ctx->tokens_cur += advance; + + return TRUE; +} + static boolean translate( struct translate_ctx *ctx ) { eat_opt_white( &ctx->cur ); if (!parse_header( ctx )) return FALSE; - eat_white( &ctx->cur ); while (*ctx->cur != '\0') { uint label_val = 0; + if (!eat_white( &ctx->cur )) { + report_error( ctx, "Syntax error" ); + return FALSE; + } + if (parse_label( ctx, &label_val )) { if (!parse_instruction( ctx )) return FALSE; } + else if (str_match_no_case( &ctx->cur, "DCL" )) { + if (!parse_declaration( ctx )) + return FALSE; + } else { - report_error( ctx, "Instruction expected" ); + report_error( ctx, "Expected `DCL' or a label" ); return FALSE; } } -- cgit v1.2.3 From 625034104aacaca793ff373414eff4ee53afc1fe Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 11:42:33 +0200 Subject: tgsi: Add missing SWZ opcode. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 3 ++- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 59e7aaeceb..cae74a9267 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -395,7 +395,8 @@ static const char *TGSI_OPCODES_SHORT[TGSI_OPCODE_LAST] = "IFC", "BREAKC", "KIL", - "END" + "END", + "SWZ" }; static const char *TGSI_SATS[] = diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 03c9217243..3415b38f9c 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -522,7 +522,8 @@ static const struct opcode_info opcode_info[TGSI_OPCODE_LAST] = { 1, 2, "IFC" }, { 1, 2, "BREAKC" }, { 1, 2, "KIL" }, - { 0, 0, "END" } + { 0, 0, "END" }, + { 1, 1, "SWZ" } }; static boolean parse_instruction( struct translate_ctx *ctx ) -- cgit v1.2.3 From cfd2bf9fa127a7f0b1a89650fde34a23f318f90c Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 11:43:30 +0200 Subject: tgsi: Fix instruction opcode parsing. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 3415b38f9c..7848f3d55f 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -541,7 +541,7 @@ static boolean parse_instruction( struct translate_ctx *ctx ) if (str_match_no_case( &cur, opcode_info[i].mnemonic )) { /* TODO: _SAT suffix */ - if (eat_white( &cur )) { + if (*cur == '\0' || eat_white( &cur )) { ctx->cur = cur; break; } -- cgit v1.2.3 From 46a7843099f02b6dcff56a52c9247c11d5c4aa8b Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 12:16:16 +0200 Subject: tgsi: Fix instruction operand counts. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 208 ++++++++++++++-------------- 1 file changed, 104 insertions(+), 104 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 7848f3d55f..3c7a4688c9 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -405,9 +405,9 @@ struct opcode_info static const struct opcode_info opcode_info[TGSI_OPCODE_LAST] = { -{ 1, 1, "ARL" }, + { 1, 1, "ARL" }, { 1, 1, "MOV" }, -{ 1, 1, "LIT" }, + { 1, 1, "LIT" }, { 1, 1, "RCP" }, { 1, 1, "RSQ" }, { 1, 1, "EXP" }, @@ -416,112 +416,112 @@ static const struct opcode_info opcode_info[TGSI_OPCODE_LAST] = { 1, 2, "ADD" }, { 1, 2, "DP3" }, { 1, 2, "DP4" }, -{ 1, 2, "DST" }, + { 1, 2, "DST" }, { 1, 2, "MIN" }, { 1, 2, "MAX" }, -{ 1, 2, "SLT" }, -{ 1, 2, "SGE" }, + { 1, 2, "SLT" }, + { 1, 2, "SGE" }, { 1, 3, "MAD" }, { 1, 2, "SUB" }, -{ 1, 3, "LERP" }, -{ 1, 2, "CND" }, -{ 1, 2, "CND0" }, -{ 1, 2, "DOT2ADD" }, -{ 1, 2, "INDEX" }, -{ 1, 2, "NEGATE" }, -{ 1, 2, "FRAC" }, -{ 1, 2, "CLAMP" }, -{ 1, 2, "FLOOR" }, -{ 1, 2, "ROUND" }, -{ 1, 2, "EXPBASE2" }, -{ 1, 2, "LOGBASE2" }, -{ 1, 2, "POWER" }, -{ 1, 2, "CROSSPRODUCT" }, -{ 1, 2, "MULTIPLYMATRIX" }, - { 1, 2, "ABS" }, -{ 1, 2, "RCC" }, -{ 1, 2, "DPH" }, -{ 1, 2, "COS" }, -{ 1, 2, "DDX" }, -{ 1, 2, "DDY" }, -{ 1, 2, "KILP" }, -{ 1, 2, "PK2H" }, -{ 1, 2, "PK2US" }, -{ 1, 2, "PK4B" }, -{ 1, 2, "PK4UB" }, -{ 1, 2, "RFL" }, -{ 1, 2, "SEQ" }, -{ 1, 2, "SFL" }, -{ 1, 2, "SGT" }, -{ 1, 2, "SIN" }, -{ 1, 2, "SLE" }, -{ 1, 2, "SNE" }, -{ 1, 2, "STR" }, -{ 1, 2, "TEX" }, -{ 1, 2, "TXD" }, -{ 1, 2, "TXP" }, -{ 1, 2, "UP2H" }, -{ 1, 2, "UP2US" }, -{ 1, 2, "UP4B" }, -{ 1, 2, "UP4UB" }, -{ 1, 2, "X2D" }, -{ 1, 2, "ARA" }, -{ 1, 2, "ARR" }, -{ 1, 2, "BRA" }, -{ 1, 2, "CAL" }, -{ 1, 2, "RET" }, -{ 1, 2, "SSG" }, -{ 1, 2, "CMP" }, -{ 1, 2, "SCS" }, -{ 1, 2, "TXB" }, -{ 1, 2, "NRM" }, -{ 1, 2, "DIV" }, -{ 1, 2, "DP2" }, -{ 1, 2, "TXL" }, -{ 1, 2, "BRK" }, -{ 1, 2, "IF" }, -{ 1, 2, "LOOP" }, -{ 1, 2, "REP" }, -{ 1, 2, "ELSE" }, -{ 1, 2, "ENDIF" }, -{ 1, 2, "ENDLOOP" }, -{ 1, 2, "ENDREP" }, -{ 1, 2, "PUSHA" }, -{ 1, 2, "POPA" }, -{ 1, 2, "CEIL" }, -{ 1, 2, "I2F" }, -{ 1, 2, "NOT" }, -{ 1, 2, "TRUNC" }, -{ 1, 2, "SHL" }, -{ 1, 2, "SHR" }, -{ 1, 2, "AND" }, -{ 1, 2, "OR" }, -{ 1, 2, "MOD" }, -{ 1, 2, "XOR" }, -{ 1, 2, "SAD" }, -{ 1, 2, "TXF" }, -{ 1, 2, "TXQ" }, -{ 1, 2, "CONT" }, -{ 1, 2, "EMIT" }, -{ 1, 2, "ENDPRIM" }, -{ 1, 2, "BGNLOOP2" }, -{ 1, 2, "BGNSUB" }, -{ 1, 2, "ENDLOOP2" }, -{ 1, 2, "ENDSUB" }, -{ 1, 2, "NOISE1" }, -{ 1, 2, "NOISE2" }, -{ 1, 2, "NOISE3" }, -{ 1, 2, "NOISE4" }, -{ 1, 2, "NOP" }, -{ 1, 2, "M4X3" }, -{ 1, 2, "M3X4" }, -{ 1, 2, "M3X3" }, -{ 1, 2, "M3X2" }, -{ 1, 2, "NRM4" }, -{ 1, 2, "CALLNZ" }, -{ 1, 2, "IFC" }, -{ 1, 2, "BREAKC" }, -{ 1, 2, "KIL" }, + { 1, 3, "LERP" }, + { 1, 3, "CND" }, + { 1, 3, "CND0" }, + { 1, 3, "DOT2ADD" }, + { 1, 2, "INDEX" }, + { 1, 1, "NEGATE" }, + { 1, 1, "FRAC" }, + { 1, 3, "CLAMP" }, + { 1, 1, "FLOOR" }, + { 1, 1, "ROUND" }, + { 1, 1, "EXPBASE2" }, + { 1, 1, "LOGBASE2" }, + { 1, 2, "POWER" }, + { 1, 2, "CROSSPRODUCT" }, + { 1, 2, "MULTIPLYMATRIX" }, + { 1, 1, "ABS" }, + { 1, 1, "RCC" }, + { 1, 2, "DPH" }, + { 1, 1, "COS" }, + { 1, 1, "DDX" }, + { 1, 1, "DDY" }, + { 0, 1, "KILP" }, + { 1, 1, "PK2H" }, + { 1, 1, "PK2US" }, + { 1, 1, "PK4B" }, + { 1, 1, "PK4UB" }, + { 1, 2, "RFL" }, + { 1, 2, "SEQ" }, + { 1, 2, "SFL" }, + { 1, 2, "SGT" }, + { 1, 1, "SIN" }, + { 1, 2, "SLE" }, + { 1, 2, "SNE" }, + { 1, 2, "STR" }, + { 1, 1, "TEX" }, + { 1, 3, "TXD" }, + { 1, 1, "TXP" }, + { 1, 1, "UP2H" }, + { 1, 1, "UP2US" }, + { 1, 1, "UP4B" }, + { 1, 1, "UP4UB" }, + { 1, 3, "X2D" }, + { 1, 1, "ARA" }, + { 1, 1, "ARR" }, + { 0, 1, "BRA" }, + { 0, 0, "CAL" }, + { 0, 0, "RET" }, + { 1, 1, "SSG" }, + { 1, 3, "CMP" }, + { 1, 1, "SCS" }, + { 1, 1, "TXB" }, + { 1, 1, "NRM" }, + { 1, 2, "DIV" }, + { 1, 2, "DP2" }, + { 1, 1, "TXL" }, + { 0, 0, "BRK" }, + { 0, 1, "IF" }, + { 0, 0, "LOOP" }, + { 0, 1, "REP" }, + { 0, 0, "ELSE" }, + { 0, 0, "ENDIF" }, + { 0, 0, "ENDLOOP" }, + { 0, 0, "ENDREP" }, + { 0, 1, "PUSHA" }, + { 1, 0, "POPA" }, + { 1, 1, "CEIL" }, + { 1, 1, "I2F" }, + { 1, 1, "NOT" }, + { 1, 1, "TRUNC" }, + { 1, 2, "SHL" }, + { 1, 2, "SHR" }, + { 1, 2, "AND" }, + { 1, 2, "OR" }, + { 1, 2, "MOD" }, + { 1, 2, "XOR" }, + { 1, 3, "SAD" }, + { 1, 1, "TXF" }, + { 1, 1, "TXQ" }, + { 0, 0, "CONT" }, + { 0, 0, "EMIT" }, + { 0, 0, "ENDPRIM" }, + { 0, 0, "BGNLOOP2" }, + { 0, 0, "BGNSUB" }, + { 0, 0, "ENDLOOP2" }, + { 0, 0, "ENDSUB" }, + { 1, 1, "NOISE1" }, + { 1, 1, "NOISE2" }, + { 1, 1, "NOISE3" }, + { 1, 1, "NOISE4" }, + { 0, 0, "NOP" }, + { 1, 2, "M4X3" }, + { 1, 2, "M3X4" }, + { 1, 2, "M3X3" }, + { 1, 2, "M3X2" }, + { 1, 1, "NRM4" }, + { 0, 1, "CALLNZ" }, + { 0, 1, "IFC" }, + { 0, 1, "BREAKC" }, + { 0, 0, "KIL" }, { 0, 0, "END" }, { 1, 1, "SWZ" } }; -- cgit v1.2.3 From 3d5dcc2203f5018863ad51958871542696e1ed1b Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 13:13:39 +0200 Subject: tgsi: Parse texture instructions correctly. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 272 +++++++++++++++------------- src/gallium/include/pipe/p_shader_tokens.h | 1 + 2 files changed, 152 insertions(+), 121 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 3c7a4688c9..7485002862 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -400,130 +400,144 @@ struct opcode_info { uint num_dst; uint num_src; + uint is_tex; const char *mnemonic; }; static const struct opcode_info opcode_info[TGSI_OPCODE_LAST] = { - { 1, 1, "ARL" }, - { 1, 1, "MOV" }, - { 1, 1, "LIT" }, - { 1, 1, "RCP" }, - { 1, 1, "RSQ" }, - { 1, 1, "EXP" }, - { 1, 1, "LOG" }, - { 1, 2, "MUL" }, - { 1, 2, "ADD" }, - { 1, 2, "DP3" }, - { 1, 2, "DP4" }, - { 1, 2, "DST" }, - { 1, 2, "MIN" }, - { 1, 2, "MAX" }, - { 1, 2, "SLT" }, - { 1, 2, "SGE" }, - { 1, 3, "MAD" }, - { 1, 2, "SUB" }, - { 1, 3, "LERP" }, - { 1, 3, "CND" }, - { 1, 3, "CND0" }, - { 1, 3, "DOT2ADD" }, - { 1, 2, "INDEX" }, - { 1, 1, "NEGATE" }, - { 1, 1, "FRAC" }, - { 1, 3, "CLAMP" }, - { 1, 1, "FLOOR" }, - { 1, 1, "ROUND" }, - { 1, 1, "EXPBASE2" }, - { 1, 1, "LOGBASE2" }, - { 1, 2, "POWER" }, - { 1, 2, "CROSSPRODUCT" }, - { 1, 2, "MULTIPLYMATRIX" }, - { 1, 1, "ABS" }, - { 1, 1, "RCC" }, - { 1, 2, "DPH" }, - { 1, 1, "COS" }, - { 1, 1, "DDX" }, - { 1, 1, "DDY" }, - { 0, 1, "KILP" }, - { 1, 1, "PK2H" }, - { 1, 1, "PK2US" }, - { 1, 1, "PK4B" }, - { 1, 1, "PK4UB" }, - { 1, 2, "RFL" }, - { 1, 2, "SEQ" }, - { 1, 2, "SFL" }, - { 1, 2, "SGT" }, - { 1, 1, "SIN" }, - { 1, 2, "SLE" }, - { 1, 2, "SNE" }, - { 1, 2, "STR" }, - { 1, 1, "TEX" }, - { 1, 3, "TXD" }, - { 1, 1, "TXP" }, - { 1, 1, "UP2H" }, - { 1, 1, "UP2US" }, - { 1, 1, "UP4B" }, - { 1, 1, "UP4UB" }, - { 1, 3, "X2D" }, - { 1, 1, "ARA" }, - { 1, 1, "ARR" }, - { 0, 1, "BRA" }, - { 0, 0, "CAL" }, - { 0, 0, "RET" }, - { 1, 1, "SSG" }, - { 1, 3, "CMP" }, - { 1, 1, "SCS" }, - { 1, 1, "TXB" }, - { 1, 1, "NRM" }, - { 1, 2, "DIV" }, - { 1, 2, "DP2" }, - { 1, 1, "TXL" }, - { 0, 0, "BRK" }, - { 0, 1, "IF" }, - { 0, 0, "LOOP" }, - { 0, 1, "REP" }, - { 0, 0, "ELSE" }, - { 0, 0, "ENDIF" }, - { 0, 0, "ENDLOOP" }, - { 0, 0, "ENDREP" }, - { 0, 1, "PUSHA" }, - { 1, 0, "POPA" }, - { 1, 1, "CEIL" }, - { 1, 1, "I2F" }, - { 1, 1, "NOT" }, - { 1, 1, "TRUNC" }, - { 1, 2, "SHL" }, - { 1, 2, "SHR" }, - { 1, 2, "AND" }, - { 1, 2, "OR" }, - { 1, 2, "MOD" }, - { 1, 2, "XOR" }, - { 1, 3, "SAD" }, - { 1, 1, "TXF" }, - { 1, 1, "TXQ" }, - { 0, 0, "CONT" }, - { 0, 0, "EMIT" }, - { 0, 0, "ENDPRIM" }, - { 0, 0, "BGNLOOP2" }, - { 0, 0, "BGNSUB" }, - { 0, 0, "ENDLOOP2" }, - { 0, 0, "ENDSUB" }, - { 1, 1, "NOISE1" }, - { 1, 1, "NOISE2" }, - { 1, 1, "NOISE3" }, - { 1, 1, "NOISE4" }, - { 0, 0, "NOP" }, - { 1, 2, "M4X3" }, - { 1, 2, "M3X4" }, - { 1, 2, "M3X3" }, - { 1, 2, "M3X2" }, - { 1, 1, "NRM4" }, - { 0, 1, "CALLNZ" }, - { 0, 1, "IFC" }, - { 0, 1, "BREAKC" }, - { 0, 0, "KIL" }, - { 0, 0, "END" }, - { 1, 1, "SWZ" } + { 1, 1, 0, "ARL" }, + { 1, 1, 0, "MOV" }, + { 1, 1, 0, "LIT" }, + { 1, 1, 0, "RCP" }, + { 1, 1, 0, "RSQ" }, + { 1, 1, 0, "EXP" }, + { 1, 1, 0, "LOG" }, + { 1, 2, 0, "MUL" }, + { 1, 2, 0, "ADD" }, + { 1, 2, 0, "DP3" }, + { 1, 2, 0, "DP4" }, + { 1, 2, 0, "DST" }, + { 1, 2, 0, "MIN" }, + { 1, 2, 0, "MAX" }, + { 1, 2, 0, "SLT" }, + { 1, 2, 0, "SGE" }, + { 1, 3, 0, "MAD" }, + { 1, 2, 0, "SUB" }, + { 1, 3, 0, "LERP" }, + { 1, 3, 0, "CND" }, + { 1, 3, 0, "CND0" }, + { 1, 3, 0, "DOT2ADD" }, + { 1, 2, 0, "INDEX" }, + { 1, 1, 0, "NEGATE" }, + { 1, 1, 0, "FRAC" }, + { 1, 3, 0, "CLAMP" }, + { 1, 1, 0, "FLOOR" }, + { 1, 1, 0, "ROUND" }, + { 1, 1, 0, "EXPBASE2" }, + { 1, 1, 0, "LOGBASE2" }, + { 1, 2, 0, "POWER" }, + { 1, 2, 0, "CROSSPRODUCT" }, + { 1, 2, 0, "MULTIPLYMATRIX" }, + { 1, 1, 0, "ABS" }, + { 1, 1, 0, "RCC" }, + { 1, 2, 0, "DPH" }, + { 1, 1, 0, "COS" }, + { 1, 1, 0, "DDX" }, + { 1, 1, 0, "DDY" }, + { 0, 1, 0, "KILP" }, + { 1, 1, 0, "PK2H" }, + { 1, 1, 0, "PK2US" }, + { 1, 1, 0, "PK4B" }, + { 1, 1, 0, "PK4UB" }, + { 1, 2, 0, "RFL" }, + { 1, 2, 0, "SEQ" }, + { 1, 2, 0, "SFL" }, + { 1, 2, 0, "SGT" }, + { 1, 1, 0, "SIN" }, + { 1, 2, 0, "SLE" }, + { 1, 2, 0, "SNE" }, + { 1, 2, 0, "STR" }, + { 1, 2, 1, "TEX" }, + { 1, 4, 1, "TXD" }, + { 1, 2, 1, "TXP" }, + { 1, 1, 0, "UP2H" }, + { 1, 1, 0, "UP2US" }, + { 1, 1, 0, "UP4B" }, + { 1, 1, 0, "UP4UB" }, + { 1, 3, 0, "X2D" }, + { 1, 1, 0, "ARA" }, + { 1, 1, 0, "ARR" }, + { 0, 1, 0, "BRA" }, + { 0, 0, 0, "CAL" }, + { 0, 0, 0, "RET" }, + { 1, 1, 0, "SSG" }, + { 1, 3, 0, "CMP" }, + { 1, 1, 0, "SCS" }, + { 1, 2, 1, "TXB" }, + { 1, 1, 0, "NRM" }, + { 1, 2, 0, "DIV" }, + { 1, 2, 0, "DP2" }, + { 1, 2, 1, "TXL" }, + { 0, 0, 0, "BRK" }, + { 0, 1, 0, "IF" }, + { 0, 0, 0, "LOOP" }, + { 0, 1, 0, "REP" }, + { 0, 0, 0, "ELSE" }, + { 0, 0, 0, "ENDIF" }, + { 0, 0, 0, "ENDLOOP" }, + { 0, 0, 0, "ENDREP" }, + { 0, 1, 0, "PUSHA" }, + { 1, 0, 0, "POPA" }, + { 1, 1, 0, "CEIL" }, + { 1, 1, 0, "I2F" }, + { 1, 1, 0, "NOT" }, + { 1, 1, 0, "TRUNC" }, + { 1, 2, 0, "SHL" }, + { 1, 2, 0, "SHR" }, + { 1, 2, 0, "AND" }, + { 1, 2, 0, "OR" }, + { 1, 2, 0, "MOD" }, + { 1, 2, 0, "XOR" }, + { 1, 3, 0, "SAD" }, + { 1, 2, 1, "TXF" }, + { 1, 2, 1, "TXQ" }, + { 0, 0, 0, "CONT" }, + { 0, 0, 0, "EMIT" }, + { 0, 0, 0, "ENDPRIM" }, + { 0, 0, 0, "BGNLOOP2" }, + { 0, 0, 0, "BGNSUB" }, + { 0, 0, 0, "ENDLOOP2" }, + { 0, 0, 0, "ENDSUB" }, + { 1, 1, 0, "NOISE1" }, + { 1, 1, 0, "NOISE2" }, + { 1, 1, 0, "NOISE3" }, + { 1, 1, 0, "NOISE4" }, + { 0, 0, 0, "NOP" }, + { 1, 2, 0, "M4X3" }, + { 1, 2, 0, "M3X4" }, + { 1, 2, 0, "M3X3" }, + { 1, 2, 0, "M3X2" }, + { 1, 1, 0, "NRM4" }, + { 0, 1, 0, "CALLNZ" }, + { 0, 1, 0, "IFC" }, + { 0, 1, 0, "BREAKC" }, + { 0, 0, 0, "KIL" }, + { 0, 0, 0, "END" }, + { 1, 1, 0, "SWZ" } +}; + +static const char *texture_names[TGSI_TEXTURE_COUNT] = +{ + "UNKNOWN", + "1D", + "2D", + "3D", + "CUBE", + "RECT", + "SHADOW1D", + "SHADOW2D", + "SHADOWRECT" }; static boolean parse_instruction( struct translate_ctx *ctx ) @@ -560,7 +574,7 @@ static boolean parse_instruction( struct translate_ctx *ctx ) /* Parse instruction operands. */ - for (i = 0; i < info->num_dst + info->num_src; i++) { + for (i = 0; i < info->num_dst + info->num_src + info->is_tex; i++) { if (i > 0) { eat_opt_white( &ctx->cur ); if (*ctx->cur != ',') { @@ -575,10 +589,26 @@ static boolean parse_instruction( struct translate_ctx *ctx ) if (!parse_dst_operand( ctx, &inst.FullDstRegisters[i] )) return FALSE; } - else { + else if (i < info->num_dst + info->num_src) { if (!parse_src_operand( ctx, &inst.FullSrcRegisters[i - info->num_dst] )) return FALSE; } + else { + uint j; + + for (j = 0; j < TGSI_TEXTURE_COUNT; j++) { + if (str_match_no_case( &ctx->cur, texture_names[j] )) { + if (!is_digit_alpha_underscore( ctx->cur )) { + inst.InstructionExtTexture.Texture = j; + break; + } + } + } + if (j == TGSI_TEXTURE_COUNT) { + report_error( ctx, "Expected texture target" ); + return FALSE; + } + } } advance = tgsi_build_full_instruction( diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 33268553ff..c8e2830516 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -529,6 +529,7 @@ struct tgsi_instruction_ext_label #define TGSI_TEXTURE_SHADOW1D 6 #define TGSI_TEXTURE_SHADOW2D 7 #define TGSI_TEXTURE_SHADOWRECT 8 +#define TGSI_TEXTURE_COUNT 9 struct tgsi_instruction_ext_texture { -- cgit v1.2.3 From a7d8eed61c1fd0bd3d99ebcd10c24bc70f8e3293 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 14:03:21 +0200 Subject: tgsi: Parse IMM statements. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 111 +++++++++++++++++++++++++++- 1 file changed, 110 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 7485002862..d5a9480f4a 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -100,6 +100,51 @@ static boolean parse_uint( const char **pcur, uint *val ) return FALSE; } +/* Parse floating point. + */ +static boolean parse_float( const char **pcur, float *val ) +{ + const char *cur = *pcur; + boolean integral_part = FALSE; + boolean fractional_part = FALSE; + + *val = (float) atof( cur ); + + if (*cur == '-' || *cur == '+') + cur++; + if (is_digit( cur )) { + cur++; + integral_part = TRUE; + while (is_digit( cur )) + cur++; + } + if (*cur == '.') { + cur++; + if (is_digit( cur )) { + cur++; + fractional_part = TRUE; + while (is_digit( cur )) + cur++; + } + } + if (!integral_part && !fractional_part) + return FALSE; + if (toupper( *cur ) == 'E') { + cur++; + if (*cur == '-' || *cur == '+') + cur++; + if (is_digit( cur )) { + cur++; + while (is_digit( cur )) + cur++; + } + else + return FALSE; + } + *pcur = cur; + return TRUE; +} + struct translate_ctx { const char *text; @@ -744,6 +789,66 @@ static boolean parse_declaration( struct translate_ctx *ctx ) return TRUE; } +static boolean parse_immediate( struct translate_ctx *ctx ) +{ + struct tgsi_full_immediate imm; + uint i; + float values[4]; + uint advance; + + if (!eat_white( &ctx->cur )) { + report_error( ctx, "Syntax error" ); + return FALSE; + } + if (!str_match_no_case( &ctx->cur, "FLT32" ) || is_digit_alpha_underscore( ctx->cur )) { + report_error( ctx, "Expected `FLT32'" ); + return FALSE; + } + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '{') { + report_error( ctx, "Expected `{'" ); + return FALSE; + } + ctx->cur++; + for (i = 0; i < 4; i++) { + eat_opt_white( &ctx->cur ); + if (i > 0) { + if (*ctx->cur != ',') { + report_error( ctx, "Expected `,'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + } + if (!parse_float( &ctx->cur, &values[i] )) { + report_error( ctx, "Expected literal floating point" ); + return FALSE; + } + } + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '}') { + report_error( ctx, "Expected `}'" ); + return FALSE; + } + ctx->cur++; + + imm = tgsi_default_full_immediate(); + imm.Immediate.Size += 4; + imm.Immediate.DataType = TGSI_IMM_FLOAT32; + imm.u.Pointer = values; + + advance = tgsi_build_full_immediate( + &imm, + ctx->tokens_cur, + ctx->header, + (uint) (ctx->tokens_end - ctx->tokens_cur) ); + if (advance == 0) + return FALSE; + ctx->tokens_cur += advance; + + return TRUE; +} + static boolean translate( struct translate_ctx *ctx ) { eat_opt_white( &ctx->cur ); @@ -766,8 +871,12 @@ static boolean translate( struct translate_ctx *ctx ) if (!parse_declaration( ctx )) return FALSE; } + else if (str_match_no_case( &ctx->cur, "IMM" )) { + if (!parse_immediate( ctx )) + return FALSE; + } else { - report_error( ctx, "Expected `DCL' or a label" ); + report_error( ctx, "Expected `DCL', `IMM' or a label" ); return FALSE; } } -- cgit v1.2.3 From 47a45aaa0fd1dbc0de45de2ed2995f81a0154baf Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 14:50:12 +0200 Subject: tgsi: Parse _SAT and _SAT opcode suffix. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 2 +- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 8 +++++++- 2 files changed, 8 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index cae74a9267..8220e09199 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -716,7 +716,7 @@ tgsi_dump_instruction( TXT( "_SAT" ); break; case TGSI_SAT_MINUS_PLUS_ONE: - TXT( "_SAT[-1,1]" ); + TXT( "_SATNV" ); break; default: assert( 0 ); diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index d5a9480f4a..f3fc675807 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -588,6 +588,7 @@ static const char *texture_names[TGSI_TEXTURE_COUNT] = static boolean parse_instruction( struct translate_ctx *ctx ) { uint i; + uint saturate = TGSI_SAT_NONE; const struct opcode_info *info; struct tgsi_full_instruction inst; uint advance; @@ -599,7 +600,11 @@ static boolean parse_instruction( struct translate_ctx *ctx ) const char *cur = ctx->cur; if (str_match_no_case( &cur, opcode_info[i].mnemonic )) { - /* TODO: _SAT suffix */ + if (str_match_no_case( &cur, "_SATNV" )) + saturate = TGSI_SAT_MINUS_PLUS_ONE; + else if (str_match_no_case( &cur, "_SAT" )) + saturate = TGSI_SAT_ZERO_ONE; + if (*cur == '\0' || eat_white( &cur )) { ctx->cur = cur; break; @@ -614,6 +619,7 @@ static boolean parse_instruction( struct translate_ctx *ctx ) inst = tgsi_default_full_instruction(); inst.Instruction.Opcode = i; + inst.Instruction.Saturate = saturate; inst.Instruction.NumDstRegs = info->num_dst; inst.Instruction.NumSrcRegs = info->num_src; -- cgit v1.2.3 From 94013b66b91112f31802c3d935e8d918ba12be62 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 15:14:31 +0200 Subject: tgsi: Parse extended source register modifiers. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 127 +++++++++++++++++++++++++++- 1 file changed, 126 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index f3fc675807..c618a50fb9 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -393,14 +393,78 @@ parse_src_operand( struct tgsi_full_src_register *src ) { const char *cur; + float value; uint file; uint index; - /* TODO: Extended register modifiers */ + if (*ctx->cur == '-') { + cur = ctx->cur; + cur++; + eat_opt_white( &cur ); + if (*cur == '(') { + cur++; + src->SrcRegisterExtMod.Negate = 1; + eat_opt_white( &cur ); + ctx->cur = cur; + } + } + + if (*ctx->cur == '|') { + ctx->cur++; + eat_opt_white( &ctx->cur ); + src->SrcRegisterExtMod.Absolute = 1; + } + if (*ctx->cur == '-') { ctx->cur++; + eat_opt_white( &ctx->cur ); src->SrcRegister.Negate = 1; + } + + cur = ctx->cur; + if (parse_float( &cur, &value )) { + if (value == 2.0f) { + eat_opt_white( &cur ); + if (*cur != '*') { + report_error( ctx, "Expected `*'" ); + return FALSE; + } + cur++; + if (*cur != '(') { + report_error( ctx, "Expected `('" ); + return FALSE; + } + cur++; + src->SrcRegisterExtMod.Scale2X = 1; + eat_opt_white( &cur ); + ctx->cur = cur; + } + } + + if (*ctx->cur == '(') { + ctx->cur++; eat_opt_white( &ctx->cur ); + src->SrcRegisterExtMod.Bias = 1; + } + + cur = ctx->cur; + if (parse_float( &cur, &value )) { + if (value == 1.0f) { + eat_opt_white( &cur ); + if (*cur != '-') { + report_error( ctx, "Expected `-'" ); + return FALSE; + } + cur++; + if (*cur != '(') { + report_error( ctx, "Expected `('" ); + return FALSE; + } + cur++; + src->SrcRegisterExtMod.Complement = 1; + eat_opt_white( &cur ); + ctx->cur = cur; + } } if (!parse_register( ctx, &file, &index )) @@ -438,6 +502,67 @@ parse_src_operand( ctx->cur = cur; } + + if (src->SrcRegisterExtMod.Complement) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ')') { + report_error( ctx, "Expected `)'" ); + return FALSE; + } + ctx->cur++; + } + + if (src->SrcRegisterExtMod.Bias) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ')') { + report_error( ctx, "Expected `)'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '-') { + report_error( ctx, "Expected `-'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + if (!parse_float( &ctx->cur, &value )) { + report_error( ctx, "Expected literal floating point" ); + return FALSE; + } + if (value != 0.5f) { + report_error( ctx, "Expected 0.5" ); + return FALSE; + } + } + + if (src->SrcRegisterExtMod.Scale2X) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ')') { + report_error( ctx, "Expected `)'" ); + return FALSE; + } + ctx->cur++; + } + + if (src->SrcRegisterExtMod.Absolute) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '|') { + report_error( ctx, "Expected `|'" ); + return FALSE; + } + ctx->cur++; + } + + if (src->SrcRegisterExtMod.Negate) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ')') { + report_error( ctx, "Expected `)'" ); + return FALSE; + } + ctx->cur++; + } + return TRUE; } -- cgit v1.2.3 From f5c51ebd2afdfc87de40dca115526a5e0f6ab115 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 13 Jul 2008 15:23:14 +0200 Subject: tgsi: Parse destination operand modulate modifier. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 36 ++++++++++------------------- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 30 ++++++++++++++++++++++++ src/gallium/include/pipe/p_shader_tokens.h | 1 + 3 files changed, 43 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 8220e09199..0cf4454f25 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -539,6 +539,17 @@ static const char *TGSI_MODULATES[] = "MODULATE_EIGHTH" }; +static const char *TGSI_MODULATES_SHORT[TGSI_MODULATE_COUNT] = +{ + "", + "_2X", + "_4X", + "_8X", + "_D2", + "_D4", + "_D8" +}; + void tgsi_dump_declaration( const struct tgsi_full_declaration *decl ) @@ -736,30 +747,7 @@ tgsi_dump_instruction( SID( dst->DstRegister.Index ); CHR( ']' ); - switch (dst->DstRegisterExtModulate.Modulate) { - case TGSI_MODULATE_1X: - break; - case TGSI_MODULATE_2X: - TXT( "_2X" ); - break; - case TGSI_MODULATE_4X: - TXT( "_4X" ); - break; - case TGSI_MODULATE_8X: - TXT( "_8X" ); - break; - case TGSI_MODULATE_HALF: - TXT( "_D2" ); - break; - case TGSI_MODULATE_QUARTER: - TXT( "_D4" ); - break; - case TGSI_MODULATE_EIGHTH: - TXT( "_D8" ); - break; - default: - assert( 0 ); - } + ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES_SHORT ); if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { CHR( '.' ); diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index c618a50fb9..1b283feb4c 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -368,6 +368,17 @@ parse_register_dcl( return TRUE; } +static const char *modulate_names[TGSI_MODULATE_COUNT] = +{ + "_1X", + "_2X", + "_4X", + "_8X", + "_D2", + "_D4", + "_D8" +}; + static boolean parse_dst_operand( struct translate_ctx *ctx, @@ -376,11 +387,30 @@ parse_dst_operand( uint file; uint index; uint writemask; + const char *cur; if (!parse_register( ctx, &file, &index )) return FALSE; + + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == '_') { + uint i; + + for (i = 0; i < TGSI_MODULATE_COUNT; i++) { + if (str_match_no_case( &cur, modulate_names[i] )) { + if (!is_digit_alpha_underscore( cur )) { + dst->DstRegisterExtModulate.Modulate = i; + ctx->cur = cur; + break; + } + } + } + } + if (!parse_opt_writemask( ctx, &writemask )) return FALSE; + dst->DstRegister.File = file; dst->DstRegister.Index = index; dst->DstRegister.WriteMask = writemask; diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index c8e2830516..ed931d24b9 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -743,6 +743,7 @@ struct tgsi_dst_register_ext_concode #define TGSI_MODULATE_HALF 4 #define TGSI_MODULATE_QUARTER 5 #define TGSI_MODULATE_EIGHTH 6 +#define TGSI_MODULATE_COUNT 7 struct tgsi_dst_register_ext_modulate { -- cgit v1.2.3 From 17af66fc1a141920969ddf404bd7ffb52a94fb31 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 13 Jul 2008 22:37:47 +0900 Subject: pb: buffer over/underflows are errors. --- src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c index acb9d7ad14..affa6aa85c 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c @@ -142,12 +142,12 @@ pb_debug_buffer_destroy(struct pb_buffer *_buf) assert(map); if(map) { if(!check_random_pattern(map, buf->underflow_size)) { - debug_printf("buffer underflow\n"); + debug_error("buffer underflow detected\n"); debug_assert(0); } if(!check_random_pattern(map + buf->underflow_size + buf->base.base.size, buf->overflow_size)) { - debug_printf("buffer overflow\n"); + debug_error("buffer overflow detected\n"); debug_assert(0); } pb_unmap(buf->buffer); -- cgit v1.2.3 From 36dd89c8a7f2a911e8f7f18d1edcaf982a75a438 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 13 Jul 2008 22:39:58 +0900 Subject: util: Eliminate pipe from the arguments to pipe_get/put_tile_xxx functions. You don't need a pipe_context * for this, and all other necessary info is already inside pipe_surface. --- src/gallium/auxiliary/util/p_tile.c | 42 +++++++++++----------------- src/gallium/auxiliary/util/p_tile.h | 19 ++++--------- src/gallium/drivers/softpipe/sp_tile_cache.c | 16 +++++------ src/mesa/drivers/x11/xm_surface.c | 4 +-- src/mesa/state_tracker/st_cb_accum.c | 12 ++++---- src/mesa/state_tracker/st_cb_drawpixels.c | 8 +++--- src/mesa/state_tracker/st_cb_readpixels.c | 10 +++---- src/mesa/state_tracker/st_cb_texture.c | 8 +++--- 8 files changed, 51 insertions(+), 68 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 93abef9879..1a1a2d96cc 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -45,12 +45,10 @@ * This should be usable by any hw driver that has mappable surfaces. */ void -pipe_get_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_get_tile_raw(struct pipe_surface *ps, uint x, uint y, uint w, uint h, void *dst, int dst_stride) { - struct pipe_screen *screen = pipe->screen; const void *src; if (pipe_clip_tile(x, y, &w, &h, ps)) @@ -59,14 +57,14 @@ pipe_get_tile_raw(struct pipe_context *pipe, if (dst_stride == 0) dst_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; - src = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ); + src = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ); assert(src); if(!src) return; pipe_copy_rect(dst, &ps->block, dst_stride, 0, 0, w, h, src, ps->stride, x, y); - screen->surface_unmap(screen, ps); + pipe_surface_unmap(ps); } @@ -75,12 +73,10 @@ pipe_get_tile_raw(struct pipe_context *pipe, * This should be usable by any hw driver that has mappable surfaces. */ void -pipe_put_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_put_tile_raw(struct pipe_surface *ps, uint x, uint y, uint w, uint h, const void *src, int src_stride) { - struct pipe_screen *screen = pipe->screen; void *dst; if (pipe_clip_tile(x, y, &w, &h, ps)) @@ -89,14 +85,14 @@ pipe_put_tile_raw(struct pipe_context *pipe, if (src_stride == 0) src_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; - dst = screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE); + dst = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE); assert(dst); if(!dst) return; pipe_copy_rect(dst, &ps->block, ps->stride, x, y, w, h, src, src_stride, 0, 0); - screen->surface_unmap(screen, ps); + pipe_surface_unmap(ps); } @@ -686,8 +682,7 @@ ycbcr_get_tile_rgba(ushort *src, void -pipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_get_tile_rgba(struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p) { @@ -702,7 +697,7 @@ pipe_get_tile_rgba(struct pipe_context *pipe, if (!packed) return; - pipe_get_tile_raw(pipe, ps, x, y, w, h, packed, 0); + pipe_get_tile_raw(ps, x, y, w, h, packed, 0); switch (ps->format) { case PIPE_FORMAT_A8R8G8B8_UNORM: @@ -768,8 +763,7 @@ pipe_get_tile_rgba(struct pipe_context *pipe, void -pipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_put_tile_rgba(struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p) { @@ -838,7 +832,7 @@ pipe_put_tile_rgba(struct pipe_context *pipe, assert(0); } - pipe_put_tile_raw(pipe, ps, x, y, w, h, packed, 0); + pipe_put_tile_raw(ps, x, y, w, h, packed, 0); FREE(packed); } @@ -848,12 +842,10 @@ pipe_put_tile_rgba(struct pipe_context *pipe, * Get a block of Z values, converted to 32-bit range. */ void -pipe_get_tile_z(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_get_tile_z(struct pipe_surface *ps, uint x, uint y, uint w, uint h, uint *z) { - struct pipe_screen *screen = pipe->screen; const uint dstStride = w; ubyte *map; uint *pDest = z; @@ -862,7 +854,7 @@ pipe_get_tile_z(struct pipe_context *pipe, if (pipe_clip_tile(x, y, &w, &h, ps)) return; - map = (ubyte *)screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_READ); + map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ); if (!map) { assert(0); return; @@ -913,17 +905,15 @@ pipe_get_tile_z(struct pipe_context *pipe, assert(0); } - screen->surface_unmap(screen, ps); + pipe_surface_unmap(ps); } void -pipe_put_tile_z(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_put_tile_z(struct pipe_surface *ps, uint x, uint y, uint w, uint h, const uint *zSrc) { - struct pipe_screen *screen = pipe->screen; const uint srcStride = w; const uint *pSrc = zSrc; ubyte *map; @@ -932,7 +922,7 @@ pipe_put_tile_z(struct pipe_context *pipe, if (pipe_clip_tile(x, y, &w, &h, ps)) return; - map = (ubyte *)screen->surface_map(screen, ps, PIPE_BUFFER_USAGE_CPU_WRITE); + map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE); if (!map) { assert(0); return; @@ -980,7 +970,7 @@ pipe_put_tile_z(struct pipe_context *pipe, assert(0); } - screen->surface_unmap(screen, ps); + pipe_surface_unmap(ps); } diff --git a/src/gallium/auxiliary/util/p_tile.h b/src/gallium/auxiliary/util/p_tile.h index fdc80a13b3..adfec8bcee 100644 --- a/src/gallium/auxiliary/util/p_tile.h +++ b/src/gallium/auxiliary/util/p_tile.h @@ -30,7 +30,6 @@ #include "pipe/p_compiler.h" -struct pipe_context; struct pipe_surface; @@ -57,40 +56,34 @@ extern "C" { #endif void -pipe_get_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_get_tile_raw(struct pipe_surface *ps, uint x, uint y, uint w, uint h, void *p, int dst_stride); void -pipe_put_tile_raw(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_put_tile_raw(struct pipe_surface *ps, uint x, uint y, uint w, uint h, const void *p, int src_stride); void -pipe_get_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_get_tile_rgba(struct pipe_surface *ps, uint x, uint y, uint w, uint h, float *p); void -pipe_put_tile_rgba(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_put_tile_rgba(struct pipe_surface *ps, uint x, uint y, uint w, uint h, const float *p); void -pipe_get_tile_z(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_get_tile_z(struct pipe_surface *ps, uint x, uint y, uint w, uint h, uint *z); void -pipe_put_tile_z(struct pipe_context *pipe, - struct pipe_surface *ps, +pipe_put_tile_z(struct pipe_surface *ps, uint x, uint y, uint w, uint h, const uint *z); diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 2d5d2b50f5..bfdaaa6b8f 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -339,7 +339,7 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe, for (y = 0; y < h; y += TILE_SIZE) { for (x = 0; x < w; x += TILE_SIZE) { if (is_clear_flag_set(tc->clear_flags, x, y)) { - pipe_put_tile_raw(pipe, ps, + pipe_put_tile_raw(ps, x, y, TILE_SIZE, TILE_SIZE, tc->tile.data.color32, 0/*STRIDE*/); @@ -374,12 +374,12 @@ sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_cached_tile *tile = tc->entries + pos; if (tile->x >= 0) { if (tc->depth_stencil) { - pipe_put_tile_raw(pipe, ps, + pipe_put_tile_raw(ps, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { - pipe_put_tile_rgba(pipe, ps, + pipe_put_tile_rgba(ps, tile->x, tile->y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); } @@ -431,12 +431,12 @@ sp_get_cached_tile(struct softpipe_context *softpipe, if (tile->x != -1) { /* put dirty tile back in framebuffer */ if (tc->depth_stencil) { - pipe_put_tile_raw(pipe, ps, + pipe_put_tile_raw(ps, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { - pipe_put_tile_rgba(pipe, ps, + pipe_put_tile_rgba(ps, tile->x, tile->y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); } @@ -458,12 +458,12 @@ sp_get_cached_tile(struct softpipe_context *softpipe, else { /* get new tile data from surface */ if (tc->depth_stencil) { - pipe_get_tile_raw(pipe, ps, + pipe_get_tile_raw(ps, tile->x, tile->y, TILE_SIZE, TILE_SIZE, tile->data.depth32, 0/*STRIDE*/); } else { - pipe_get_tile_rgba(pipe, ps, + pipe_get_tile_rgba(ps, tile->x, tile->y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); } @@ -544,7 +544,7 @@ sp_get_cached_tile_tex(struct pipe_context *pipe, } /* get tile from the surface (view into texture) */ - pipe_get_tile_rgba(pipe, tc->tex_surf, + pipe_get_tile_rgba(tc->tex_surf, tile_x, tile_y, TILE_SIZE, TILE_SIZE, (float *) tile->data.color); tile->x = tile_x; diff --git a/src/mesa/drivers/x11/xm_surface.c b/src/mesa/drivers/x11/xm_surface.c index 81616b92d9..a3f2fe7d68 100644 --- a/src/mesa/drivers/x11/xm_surface.c +++ b/src/mesa/drivers/x11/xm_surface.c @@ -106,7 +106,7 @@ xmesa_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, } else { /* other softpipe surface */ - softpipe_get_tile_rgba(pipe, ps, x, y, w, h, p); + softpipe_get_tile_rgba(ps, x, y, w, h, p); } } @@ -142,7 +142,7 @@ xmesa_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *ps, } else { /* other softpipe surface */ - softpipe_put_tile_rgba(pipe, ps, x, y, w, h, p); + softpipe_put_tile_rgba(ps, x, y, w, h, p); } } diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index 2283905662..c0e8c6bf33 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -73,7 +73,7 @@ acc_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps, acc_ps->block.width = 1; acc_ps->block.height = 1; - pipe_get_tile_rgba(pipe, acc_ps, x, y, w, h, p); + pipe_get_tile_rgba(acc_ps, x, y, w, h, p); acc_ps->format = f; acc_ps->block = b; @@ -97,7 +97,7 @@ acc_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps, acc_ps->block.width = 1; acc_ps->block.height = 1; - pipe_put_tile_rgba(pipe, acc_ps, x, y, w, h, p); + pipe_put_tile_rgba(acc_ps, x, y, w, h, p); acc_ps->format = f; acc_ps->block = b; @@ -208,7 +208,7 @@ accum_accum(struct pipe_context *pipe, GLfloat value, colorBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe_get_tile_rgba(pipe, color_surf, xpos, ypos, width, height, colorBuf); + pipe_get_tile_rgba(color_surf, xpos, ypos, width, height, colorBuf); acc_get_tile_rgba(pipe, acc_surf, xpos, ypos, width, height, accBuf); for (i = 0; i < 4 * width * height; i++) { @@ -243,7 +243,7 @@ accum_load(struct pipe_context *pipe, GLfloat value, buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe_get_tile_rgba(pipe, color_surf, xpos, ypos, width, height, buf); + pipe_get_tile_rgba(color_surf, xpos, ypos, width, height, buf); for (i = 0; i < 4 * width * height; i++) { buf[i] = buf[i] * value; @@ -283,7 +283,7 @@ accum_return(GLcontext *ctx, GLfloat value, if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) { cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe_get_tile_rgba(pipe, color_surf, xpos, ypos, width, height, cbuf); + pipe_get_tile_rgba(color_surf, xpos, ypos, width, height, cbuf); } for (i = 0; i < width * height; i++) { @@ -298,7 +298,7 @@ accum_return(GLcontext *ctx, GLfloat value, } } - pipe_put_tile_rgba(pipe, color_surf, xpos, ypos, width, height, abuf); + pipe_put_tile_rgba(color_surf, xpos, ypos, width, height, abuf); free(abuf); if (cbuf) diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index a7781f3ab2..6b0137dcf8 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -1049,16 +1049,16 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, /* alternate path using get/put_tile() */ GLfloat *buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); - pipe_get_tile_rgba(pipe, psRead, srcx, srcy, width, height, buf); - pipe_put_tile_rgba(pipe, psTex, 0, 0, width, height, buf); + pipe_get_tile_rgba(psRead, srcx, srcy, width, height, buf); + pipe_put_tile_rgba(psTex, 0, 0, width, height, buf); free(buf); } else { /* GL_DEPTH */ GLuint *buf = (GLuint *) malloc(width * height * sizeof(GLuint)); - pipe_get_tile_z(pipe, psRead, srcx, srcy, width, height, buf); - pipe_put_tile_z(pipe, psTex, 0, 0, width, height, buf); + pipe_get_tile_z(psRead, srcx, srcy, width, height, buf); + pipe_put_tile_z(psTex, 0, 0, width, height, buf); free(buf); } pipe_surface_reference(&psRead, NULL); diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 09d9c29e44..eb71779cc9 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -262,7 +262,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLuint ztemp[MAX_WIDTH]; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / ((1 << 24) - 1); - pipe_get_tile_raw(pipe, surf, x, y, width, 1, ztemp, 0); + pipe_get_tile_raw(surf, x, y, width, 1, ztemp, 0); y += yStep; for (j = 0; j < width; j++) { zfloat[j] = (float) (scale * (ztemp[j] & 0xffffff)); @@ -276,7 +276,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, /* untested, but simple: */ assert(format == GL_DEPTH_STENCIL_EXT); for (i = 0; i < height; i++) { - pipe_get_tile_raw(pipe, surf, x, y, width, 1, dst, 0); + pipe_get_tile_raw(surf, x, y, width, 1, dst, 0); y += yStep; dst += dstStride; } @@ -287,7 +287,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLushort ztemp[MAX_WIDTH]; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / 0xffff; - pipe_get_tile_raw(pipe, surf, x, y, width, 1, ztemp, 0); + pipe_get_tile_raw(surf, x, y, width, 1, ztemp, 0); y += yStep; for (j = 0; j < width; j++) { zfloat[j] = (float) (scale * ztemp[j]); @@ -302,7 +302,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, GLuint ztemp[MAX_WIDTH]; GLfloat zfloat[MAX_WIDTH]; const double scale = 1.0 / 0xffffffff; - pipe_get_tile_raw(pipe, surf, x, y, width, 1, ztemp, 0); + pipe_get_tile_raw(surf, x, y, width, 1, ztemp, 0); y += yStep; for (j = 0; j < width; j++) { zfloat[j] = (float) (scale * ztemp[j]); @@ -316,7 +316,7 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, /* RGBA format */ /* Do a row at a time to flip image data vertically */ for (i = 0; i < height; i++) { - pipe_get_tile_rgba(pipe, surf, x, y, width, 1, df); + pipe_get_tile_rgba(surf, x, y, width, 1, df); y += yStep; df += dfStride; if (!dfStride) { diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index b9aa513d72..73f8b8f788 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1100,25 +1100,25 @@ fallback_copy_texsubimage(GLcontext *ctx, for (row = 0; row < height; row++, srcY++, destY += yStep) { uint data[MAX_WIDTH]; - pipe_get_tile_z(pipe, src_surf, srcX, srcY, width, 1, data); + pipe_get_tile_z(src_surf, srcX, srcY, width, 1, data); if (scaleOrBias) { _mesa_scale_and_bias_depth_uint(ctx, width, data); } - pipe_put_tile_z(pipe, dest_surf, destX, destY, width, 1, data); + pipe_put_tile_z(dest_surf, destX, destY, width, 1, data); } } else { /* RGBA format */ for (row = 0; row < height; row++, srcY++, destY += yStep) { float data[4 * MAX_WIDTH]; - pipe_get_tile_rgba(pipe, src_surf, srcX, srcY, width, 1, data); + pipe_get_tile_rgba(src_surf, srcX, srcY, width, 1, data); /* XXX we're ignoring convolution for now */ if (ctx->_ImageTransferState) { _mesa_apply_rgba_transfer_ops(ctx, ctx->_ImageTransferState & ~IMAGE_CONVOLUTION_BIT, width, (GLfloat (*)[4]) data); } - pipe_put_tile_rgba(pipe, dest_surf, destX, destY, width, 1, data); + pipe_put_tile_rgba(dest_surf, destX, destY, width, 1, data); } } -- cgit v1.2.3 From 6410e94b966148dde81b5121e53a250d7b530d91 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 13 Jul 2008 23:36:59 +0900 Subject: python: New state tracker which exposes the pipe driver to python scripts. Still under development. Just barely works. --- src/gallium/SConscript | 2 + src/gallium/state_trackers/python/README | 32 ++ src/gallium/state_trackers/python/SConscript | 24 ++ src/gallium/state_trackers/python/gallium.i | 448 +++++++++++++++++++++ src/gallium/state_trackers/python/p_format.i | 152 +++++++ .../state_trackers/python/samples/simple.py | 158 ++++++++ src/gallium/state_trackers/python/st_device.c | 169 ++++++++ src/gallium/state_trackers/python/st_device.h | 83 ++++ .../state_trackers/python/st_softpipe_winsys.c | 321 +++++++++++++++ src/gallium/state_trackers/python/st_winsys.h | 58 +++ 10 files changed, 1447 insertions(+) create mode 100644 src/gallium/state_trackers/python/README create mode 100644 src/gallium/state_trackers/python/SConscript create mode 100644 src/gallium/state_trackers/python/gallium.i create mode 100644 src/gallium/state_trackers/python/p_format.i create mode 100644 src/gallium/state_trackers/python/samples/simple.py create mode 100644 src/gallium/state_trackers/python/st_device.c create mode 100644 src/gallium/state_trackers/python/st_device.h create mode 100644 src/gallium/state_trackers/python/st_softpipe_winsys.c create mode 100644 src/gallium/state_trackers/python/st_winsys.h (limited to 'src') diff --git a/src/gallium/SConscript b/src/gallium/SConscript index 2653f91bd2..6a3e7e77ed 100644 --- a/src/gallium/SConscript +++ b/src/gallium/SConscript @@ -25,3 +25,5 @@ SConscript([ for driver in env['drivers']: SConscript(os.path.join('drivers', driver, 'SConscript')) + +SConscript('state_trackers/python/SConscript') diff --git a/src/gallium/state_trackers/python/README b/src/gallium/state_trackers/python/README new file mode 100644 index 0000000000..75341aa28c --- /dev/null +++ b/src/gallium/state_trackers/python/README @@ -0,0 +1,32 @@ +This directory contains Python bindings to Gallium3D. It looks like a state +tracker from the pipe driver perspective, and it looks like a pipe driver from +the python script perspective. + + +To build you'll need: +* Python (with development packages) +* SCons +* SWIG +* Python Imaging Library (for the samples) + +Invoke scons on the top dir as + + scons statetrackers=python + +To use do + + export PYTHONPATH=build/XXXX-XXXX-XXXX/gallium/state_trackers/python + +and then try running + + python src/gallium/state_trackers/python/samples/simple.py + +which should create a simple.png + + +This is still in experimental phase, and there many limitations to what you can +do with from Python. + + +-- +Jose Fonseca diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript new file mode 100644 index 0000000000..57a3fb2075 --- /dev/null +++ b/src/gallium/state_trackers/python/SConscript @@ -0,0 +1,24 @@ +Import('*') + +if 'python' in env['statetrackers']: + + env = env.Clone() + + env.Append(CPPPATH = '.') + + env.Tool('swig') + env.Append(SWIGPATH = ['#src/gallium/include', '#src/gallium/include/pipe']) + env.Append(SWIGFLAGS = ['-python', '-keyword']) + + env.ParseConfig('python-config --cflags --ldflags --libs') + + env.SharedLibrary( + target = '_gallium', + source = [ + 'gallium.i', + 'st_device.c', + 'st_softpipe_winsys.c', + ], + SHLIBPREFIX = '', + LIBS = softpipe + auxiliaries + env['LIBS'], + ) diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i new file mode 100644 index 0000000000..d6594f3a87 --- /dev/null +++ b/src/gallium/state_trackers/python/gallium.i @@ -0,0 +1,448 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca + */ + +%module gallium; + +%{ + +#include +#include + +#include "pipe/p_screen.h" +#include "pipe/p_context.h" +#include "pipe/p_inlines.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "util/u_draw_quad.h" +#include "util/p_tile.h" +#include "cso_cache/cso_context.h" + +#include "st_device.h" + +%} + +%include "carrays.i" +%array_class(int, IntArray); +%array_class(float, FloatArray); + + +%rename(Device) st_device; +%rename(Context) st_context; +%rename(Texture) pipe_texture; +%rename(Surface) pipe_surface; + +%rename(Buffer) pipe_buffer; +%rename(BlendColor) pipe_blend_color; +%rename(Blend) pipe_blend_state; +%rename(Clip) pipe_clip_state; +%rename(ConstantBuffer) pipe_constant_buffer; +%rename(DepthStencilAlpha) pipe_depth_stencil_alpha_state; +%rename(FormatBlock) pipe_format_block; +%rename(Framebuffer) pipe_framebuffer_state; +%rename(PolyStipple) pipe_poly_stipple; +%rename(Rasterizer) pipe_rasterizer_state; +%rename(Sampler) pipe_sampler_state; +%rename(Scissor) pipe_scissor_state; +%rename(Shader) pipe_shader_state; +%rename(VertexBuffer) pipe_vertex_buffer; +%rename(VertexElement) pipe_vertex_element; +%rename(Viewport) pipe_viewport_state; + +%include "p_format.i"; +%include "pipe/p_defines.h"; +%include "pipe/p_state.h"; +%include "pipe/p_shader_tokens.h"; + + +%nodefaultctor; +%nodefaultdtor; + +struct st_device { +}; + +struct st_context { +}; + + +%extend st_device { + + st_device(int hardware = 0) { + return st_device_create(hardware ? TRUE : FALSE); + } + + ~st_device() { + st_device_destroy($self); + } + + const char * get_name( void ) { + return $self->screen->get_name($self->screen); + } + + const char * get_vendor( void ) { + return $self->screen->get_vendor($self->screen); + } + + /** + * Query an integer-valued capability/parameter/limit + * \param param one of PIPE_CAP_x + */ + int get_param( int param ) { + return $self->screen->get_param($self->screen, param); + } + + /** + * Query a float-valued capability/parameter/limit + * \param param one of PIPE_CAP_x + */ + float get_paramf( int param ) { + return $self->screen->get_paramf($self->screen, param); + } + + /** + * Check if the given pipe_format is supported as a texture or + * drawing surface. + * \param type one of PIPE_TEXTURE, PIPE_SURFACE + */ + int is_format_supported( enum pipe_format format, unsigned type ) { + return $self->screen->is_format_supported( $self->screen, format, type); + } + + struct st_context * + context_create(void) { + return st_context_create($self); + } + + struct pipe_texture * + texture_create( + enum pipe_format format, + unsigned width, + unsigned height, + unsigned depth = 1, + unsigned last_level = 0, + enum pipe_texture_target target = PIPE_TEXTURE_2D, + unsigned usage = 0 + ) { + struct pipe_texture templat; + memset(&templat, 0, sizeof(templat)); + templat.format = format; + pf_get_block(templat.format, &templat.block); + templat.width[0] = width; + templat.height[0] = height; + templat.depth[0] = depth; + templat.last_level = last_level; + templat.target = target; + templat.tex_usage = usage; + return $self->screen->texture_create($self->screen, &templat); + } + + struct pipe_buffer * + buffer_create(unsigned size, unsigned alignment = 0, unsigned usage = 0) { + return $self->screen->winsys->buffer_create($self->screen->winsys, alignment, usage, size); + } + +}; + + +%extend st_context { + + ~st_context() { + st_context_destroy($self); + } + + /* + * State functions (create/bind/destroy state objects) + */ + + void set_blend( const struct pipe_blend_state *state ) { + cso_set_blend($self->cso, state); + } + + void set_sampler( unsigned index, const struct pipe_sampler_state *state ) { + cso_single_sampler($self->cso, index, state); + cso_single_sampler_done($self->cso); + } + + void set_rasterizer( const struct pipe_rasterizer_state *state ) { + cso_set_rasterizer($self->cso, state); + } + + void set_depth_stencil_alpha(const struct pipe_depth_stencil_alpha_state *state) { + cso_set_depth_stencil_alpha($self->cso, state); + } + + + void * create_fs( const struct pipe_shader_state *state ) { + return $self->pipe->create_fs_state($self->pipe, state); + } + + void bind_fs( void *state_obj ) { + $self->pipe->bind_fs_state($self->pipe, state_obj); + } + + void delete_fs( void *state_obj ) { + $self->pipe->delete_fs_state($self->pipe, state_obj); + } + + void * create_vs( const struct pipe_shader_state *state ) { + return $self->pipe->create_vs_state($self->pipe, state); + } + + void bind_vs( void *state_obj ) { + $self->pipe->bind_vs_state($self->pipe, state_obj); + } + + void delete_vs( void *state_obj ) { + $self->pipe->delete_vs_state($self->pipe, state_obj); + } + + /* + * Parameter-like state (or properties) + */ + + void set_blend_color(const struct pipe_blend_color *state ) { + cso_set_blend_color($self->cso, state); + } + + void set_clip(const struct pipe_clip_state *state ) { + $self->pipe->set_clip_state($self->pipe, state); + } + + void set_constant_buffer(unsigned shader, unsigned index, + const struct pipe_constant_buffer *buf ) { + $self->pipe->set_constant_buffer($self->pipe, shader, index, buf); + } + + void set_framebuffer(const struct pipe_framebuffer_state *state ) { + cso_set_framebuffer($self->cso, state); + } + + void set_polygon_stipple(const struct pipe_poly_stipple *state ) { + $self->pipe->set_polygon_stipple($self->pipe, state); + } + + void set_scissor(const struct pipe_scissor_state *state ) { + $self->pipe->set_scissor_state($self->pipe, state); + } + + void set_viewport(const struct pipe_viewport_state *state) { + cso_set_viewport($self->cso, state); + } + + void set_sampler_texture(unsigned index, + struct pipe_texture *texture) { + pipe_texture_reference(&$self->sampler_textures[index], texture); + $self->pipe->set_sampler_textures($self->pipe, + PIPE_MAX_SAMPLERS, + $self->sampler_textures); + } + + void set_vertex_buffer(unsigned index, + const struct pipe_vertex_buffer *buffer) { + memcpy(&$self->vertex_buffers[index], buffer, sizeof(*buffer)); + $self->pipe->set_vertex_buffers($self->pipe, PIPE_MAX_ATTRIBS, $self->vertex_buffers); + } + + void set_vertex_element(unsigned index, + const struct pipe_vertex_element *element) { + memcpy(&$self->vertex_elements[index], element, sizeof(*element)); + $self->pipe->set_vertex_elements($self->pipe, PIPE_MAX_ATTRIBS, $self->vertex_elements); + } + + /* + * Draw functions + */ + + void draw_arrays(unsigned mode, unsigned start, unsigned count) { + $self->pipe->draw_arrays($self->pipe, mode, start, count); + } + + void draw_elements( struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) { + $self->pipe->draw_elements($self->pipe, indexBuffer, indexSize, mode, start, count); + } + + void draw_vertices(unsigned prim, + unsigned num_verts, + unsigned num_attribs, + const float *vertices) + { + struct pipe_context *pipe = $self->pipe; + struct pipe_winsys *winsys = pipe->winsys; + struct pipe_buffer *vbuf; + float *map; + unsigned size; + + size = num_verts * num_attribs * 4 * sizeof(float); + + vbuf = winsys->buffer_create(winsys, + 32, + PIPE_BUFFER_USAGE_VERTEX, + size); + if(!vbuf) + goto error1; + + map = winsys->buffer_map(winsys, vbuf, PIPE_BUFFER_USAGE_CPU_WRITE); + if (!map) + goto error2; + memcpy(map, vertices, size); + pipe->winsys->buffer_unmap(pipe->winsys, vbuf); + + util_draw_vertex_buffer(pipe, vbuf, prim, num_verts, num_attribs); + +error2: + pipe_buffer_reference(pipe->winsys, &vbuf, NULL); +error1: + ; + } + + void draw_quad(float x0, float y0, float x1, float y1, float z = 0.0f) { + util_draw_texquad($self->pipe, x0, y0, x1, y1, z); + } + + void + flush(void) { + struct pipe_fence_handle *fence = NULL; + $self->pipe->flush($self->pipe, PIPE_FLUSH_RENDER_CACHE, &fence); + /* TODO: allow asynchronous operation */ + $self->pipe->winsys->fence_finish( $self->pipe->winsys, fence, 0 ); + $self->pipe->winsys->fence_reference( $self->pipe->winsys, &fence, NULL ); + } + + /* + * Surface functions + */ + + void surface_copy(int do_flip, + struct pipe_surface *dest, + unsigned destx, unsigned desty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) { + $self->pipe->surface_copy($self->pipe, do_flip, dest, destx, desty, src, srcx, srcy, width, height); + } + + void surface_fill(struct pipe_surface *dst, + unsigned x, unsigned y, + unsigned width, unsigned height, + unsigned value) { + $self->pipe->surface_fill($self->pipe, dst, x, y, width, height, value); + } + + void clear(struct pipe_surface *surface, unsigned value) { + $self->pipe->clear($self->pipe, surface, value); + } + +}; + + +%extend pipe_texture { + + ~pipe_texture() { + struct pipe_texture *ptr = $self; + pipe_texture_reference(&ptr, NULL); + } + + /** Get a surface which is a "view" into a texture */ + struct pipe_surface * + get_surface(unsigned face=0, unsigned level=0, unsigned zslice=0, unsigned usage=0 ) + { + struct pipe_screen *screen = $self->screen; + return screen->get_tex_surface(screen, $self, face, level, zslice, usage); + } + +}; + + +%extend pipe_surface { + + ~pipe_surface() { + struct pipe_surface *ptr = $self; + pipe_surface_reference(&ptr, NULL); + } + + // gets mapped to pipe_surface_map automatically + void * map( unsigned flags ); + + // gets mapped to pipe_surface_unmap automatically + void unmap( void ); + + void + get_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, float *p) { + pipe_get_tile_rgba($self, x, y, w, h, p); + } + + void + put_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *p) { + pipe_put_tile_rgba($self, x, y, w, h, p); + } + + void + get_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, unsigned *z) { + pipe_get_tile_z($self, x, y, w, h, z); + } + + void + put_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned *z) { + pipe_put_tile_z($self, x, y, w, h, z); + } + +}; + + +%extend pipe_framebuffer_state { + + pipe_framebuffer_state(void) { + return CALLOC_STRUCT(pipe_framebuffer_state); + } + + ~pipe_framebuffer_state() { + unsigned index; + for(index = 0; index < PIPE_MAX_COLOR_BUFS; ++index) + pipe_surface_reference(&$self->cbufs[index], NULL); + pipe_surface_reference(&$self->zsbuf, NULL); + FREE($self); + } + + void + set_cbuf(unsigned index, struct pipe_surface *surface) { + pipe_surface_reference(&$self->cbufs[index], surface); + } + + void + set_zsbuf(struct pipe_surface *surface) { + pipe_surface_reference(&$self->zsbuf, surface); + } + +}; diff --git a/src/gallium/state_trackers/python/p_format.i b/src/gallium/state_trackers/python/p_format.i new file mode 100644 index 0000000000..51ad4bebcd --- /dev/null +++ b/src/gallium/state_trackers/python/p_format.i @@ -0,0 +1,152 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/* + * XXX: SWIG can't parse p_format.h, so we need to duplicate the relevant + * declarations here + */ + +%{ +#include "pipe/p_format.h" +%} + +enum pipe_format { + PIPE_FORMAT_NONE, + PIPE_FORMAT_A8R8G8B8_UNORM, + PIPE_FORMAT_X8R8G8B8_UNORM, + PIPE_FORMAT_B8G8R8A8_UNORM, + PIPE_FORMAT_B8G8R8X8_UNORM, + PIPE_FORMAT_A1R5G5B5_UNORM, + PIPE_FORMAT_A4R4G4B4_UNORM, + PIPE_FORMAT_R5G6B5_UNORM, + PIPE_FORMAT_A2B10G10R10_UNORM, + PIPE_FORMAT_L8_UNORM, + PIPE_FORMAT_A8_UNORM, + PIPE_FORMAT_I8_UNORM, + PIPE_FORMAT_A8L8_UNORM, + PIPE_FORMAT_L16_UNORM, + PIPE_FORMAT_YCBCR, + PIPE_FORMAT_YCBCR_REV, + PIPE_FORMAT_Z16_UNORM, + PIPE_FORMAT_Z32_UNORM, + PIPE_FORMAT_Z32_FLOAT, + PIPE_FORMAT_S8Z24_UNORM, + PIPE_FORMAT_Z24S8_UNORM, + PIPE_FORMAT_X8Z24_UNORM, + PIPE_FORMAT_Z24X8_UNORM, + PIPE_FORMAT_S8_UNORM, + PIPE_FORMAT_R64_FLOAT, + PIPE_FORMAT_R64G64_FLOAT, + PIPE_FORMAT_R64G64B64_FLOAT, + PIPE_FORMAT_R64G64B64A64_FLOAT, + PIPE_FORMAT_R32_FLOAT, + PIPE_FORMAT_R32G32_FLOAT, + PIPE_FORMAT_R32G32B32_FLOAT, + PIPE_FORMAT_R32G32B32A32_FLOAT, + PIPE_FORMAT_R32_UNORM, + PIPE_FORMAT_R32G32_UNORM, + PIPE_FORMAT_R32G32B32_UNORM, + PIPE_FORMAT_R32G32B32A32_UNORM, + PIPE_FORMAT_R32_USCALED, + PIPE_FORMAT_R32G32_USCALED, + PIPE_FORMAT_R32G32B32_USCALED, + PIPE_FORMAT_R32G32B32A32_USCALED, + PIPE_FORMAT_R32_SNORM, + PIPE_FORMAT_R32G32_SNORM, + PIPE_FORMAT_R32G32B32_SNORM, + PIPE_FORMAT_R32G32B32A32_SNORM, + PIPE_FORMAT_R32_SSCALED, + PIPE_FORMAT_R32G32_SSCALED, + PIPE_FORMAT_R32G32B32_SSCALED, + PIPE_FORMAT_R32G32B32A32_SSCALED, + PIPE_FORMAT_R16_UNORM, + PIPE_FORMAT_R16G16_UNORM, + PIPE_FORMAT_R16G16B16_UNORM, + PIPE_FORMAT_R16G16B16A16_UNORM, + PIPE_FORMAT_R16_USCALED, + PIPE_FORMAT_R16G16_USCALED, + PIPE_FORMAT_R16G16B16_USCALED, + PIPE_FORMAT_R16G16B16A16_USCALED, + PIPE_FORMAT_R16_SNORM, + PIPE_FORMAT_R16G16_SNORM, + PIPE_FORMAT_R16G16B16_SNORM, + PIPE_FORMAT_R16G16B16A16_SNORM, + PIPE_FORMAT_R16_SSCALED, + PIPE_FORMAT_R16G16_SSCALED, + PIPE_FORMAT_R16G16B16_SSCALED, + PIPE_FORMAT_R16G16B16A16_SSCALED, + PIPE_FORMAT_R8_UNORM, + PIPE_FORMAT_R8G8_UNORM, + PIPE_FORMAT_R8G8B8_UNORM, + PIPE_FORMAT_R8G8B8A8_UNORM, + PIPE_FORMAT_R8G8B8X8_UNORM, + PIPE_FORMAT_R8_USCALED, + PIPE_FORMAT_R8G8_USCALED, + PIPE_FORMAT_R8G8B8_USCALED, + PIPE_FORMAT_R8G8B8A8_USCALED, + PIPE_FORMAT_R8G8B8X8_USCALED, + PIPE_FORMAT_R8_SNORM, + PIPE_FORMAT_R8G8_SNORM, + PIPE_FORMAT_R8G8B8_SNORM, + PIPE_FORMAT_R8G8B8A8_SNORM, + PIPE_FORMAT_R8G8B8X8_SNORM, + PIPE_FORMAT_B6G5R5_SNORM, + PIPE_FORMAT_A8B8G8R8_SNORM, + PIPE_FORMAT_X8B8G8R8_SNORM, + PIPE_FORMAT_R8_SSCALED, + PIPE_FORMAT_R8G8_SSCALED, + PIPE_FORMAT_R8G8B8_SSCALED, + PIPE_FORMAT_R8G8B8A8_SSCALED, + PIPE_FORMAT_R8G8B8X8_SSCALED, + PIPE_FORMAT_R32_FIXED, + PIPE_FORMAT_R32G32_FIXED, + PIPE_FORMAT_R32G32B32_FIXED, + PIPE_FORMAT_R32G32B32A32_FIXED, + + PIPE_FORMAT_L8_SRGB, + PIPE_FORMAT_A8_L8_SRGB, + PIPE_FORMAT_R8G8B8_SRGB, + PIPE_FORMAT_R8G8B8A8_SRGB, + PIPE_FORMAT_R8G8B8X8_SRGB, + + PIPE_FORMAT_X8UB8UG8SR8S_NORM, + PIPE_FORMAT_B6UG5SR5S_NORM, + + PIPE_FORMAT_DXT1_RGB, + PIPE_FORMAT_DXT1_RGBA, + PIPE_FORMAT_DXT3_RGBA, + PIPE_FORMAT_DXT5_RGBA, +}; + + +struct pipe_format_block +{ + unsigned size; + unsigned width; + unsigned height; +}; + diff --git a/src/gallium/state_trackers/python/samples/simple.py b/src/gallium/state_trackers/python/samples/simple.py new file mode 100644 index 0000000000..77e182b644 --- /dev/null +++ b/src/gallium/state_trackers/python/samples/simple.py @@ -0,0 +1,158 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +from gallium import * + + +def save_image(filename, surface): + pixels = FloatArray(surface.height*surface.width*4) + surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) + + import Image + outimage = Image.new( + mode='RGB', + size=(surface.width, surface.height), + color=(0,0,0)) + outpixels = outimage.load() + for y in range(0, surface.height): + for x in range(0, surface.width): + offset = (y*surface.width + x)*4 + r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] + outpixels[x, y] = r, g, b + outimage.save(filename, "PNG") + + +def test(dev): + ctx = dev.context_create() + + width = 256 + height = 256 + + # disabled blending/masking + blend = Blend() + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.colormask = PIPE_MASK_RGBA + ctx.set_blend(blend) + + # no-op depth/stencil/alpha + depth_stencil_alpha = DepthStencilAlpha() + ctx.set_depth_stencil_alpha(depth_stencil_alpha) + + # rasterizer + rasterizer = Rasterizer() + rasterizer.front_winding = PIPE_WINDING_CW + rasterizer.cull_mode = PIPE_WINDING_NONE + rasterizer.bypass_clipping = 1 + #rasterizer.bypass_vs = 1 + ctx.set_rasterizer(rasterizer) + + # viewport (identity, we setup vertices in wincoords) + viewport = Viewport() + scale = FloatArray(4) + scale[0] = 1.0 + scale[1] = 1.0 + scale[2] = 1.0 + scale[3] = 1.0 + viewport.scale = scale + translate = FloatArray(4) + translate[0] = 0.0 + translate[1] = 0.0 + translate[2] = 0.0 + translate[3] = 0.0 + viewport.translate = translate + ctx.set_viewport(viewport) + + # samplers + sampler = Sampler() + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.normalized_coords = 1 + ctx.set_sampler(0, sampler) + + # texture + texture = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, width, height, usage=PIPE_TEXTURE_USAGE_RENDER_TARGET) + ctx.set_sampler_texture(0, texture) + + # drawing dest + surface = texture.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE) + fb = Framebuffer() + fb.width = surface.width + fb.height = surface.height + fb.num_cbufs = 1 + fb.set_cbuf(0, surface) + ctx.set_framebuffer(fb) + + # vertex shader + # vs = Shader() + #ctx.set_vertex_shader(vs) + + # fragment shader + #fs = Shader() + #ctx.set_fragment_shader(fs) + + if 0: + nverts = 4 + nattrs = 1 + vertices = FloatArray(n_verts * nattrs * 4) + + # init vertex data that doesn't change + for i in range(nverts): + for j in range(nattrs): + vertices[(i*nattrs +j)*4 + 0] = 0.0 + vertices[(i*nattrs +j)*4 + 1] = 0.0 + vertices[(i*nattrs +j)*4 + 2] = 0.0 + vertices[(i*nattrs +j)*4 + 3] = 0.0 + + ctx.draw_vertices(PIPE_PRIM_TRIANGLE_FAN, + 4, # verts + 2, # attribs/vert + vertices) + else: + ctx.draw_quad(32.0, 32.0, 224.0, 224.0) + + ctx.flush() + + save_image("simple.png", surface) + + + +def main(): + dev = Device(0) + test(dev) + + +if __name__ == '__main__': + main() diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c new file mode 100644 index 0000000000..430a7af176 --- /dev/null +++ b/src/gallium/state_trackers/python/st_device.c @@ -0,0 +1,169 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_util.h" +#include "pipe/p_winsys.h" +#include "pipe/p_context.h" +#include "pipe/p_shader_tokens.h" +#include "pipe/p_inlines.h" +#include "cso_cache/cso_context.h" +#include "util/u_simple_shaders.h" +#include "st_device.h" +#include "st_winsys.h" + + +static void +st_device_really_destroy(struct st_device *st_dev) +{ + if(st_dev->screen) + st_dev->st_ws->screen_destroy(st_dev->screen); + + FREE(st_dev); +} + + +void +st_device_destroy(struct st_device *st_dev) +{ + if(!--st_dev->refcount) + st_device_really_destroy(st_dev); +} + + +static struct st_device * +st_device_create_from_st_winsys(const struct st_winsys *st_ws) +{ + struct st_device *st_dev; + + if(!st_ws->screen_create || + !st_ws->screen_destroy || + !st_ws->context_create || + !st_ws->context_destroy) + return NULL; + + st_dev = CALLOC_STRUCT(st_device); + if(!st_dev) + return NULL; + + st_dev->st_ws = st_ws; + + st_dev->screen = st_ws->screen_create(); + if(!st_dev->screen) + st_device_destroy(st_dev); + + return st_dev; +} + + +struct st_device * +st_device_create(boolean hardware) { +#if 0 + if(hardware) + return st_device_create_from_st_winsys(&st_hardware_winsys); + else +#endif + return st_device_create_from_st_winsys(&st_software_winsys); +} + + +void +st_context_destroy(struct st_context *st_ctx) +{ + unsigned i; + + if(st_ctx) { + struct st_device *st_dev = st_ctx->st_dev; + + if(st_ctx->vs) { + st_ctx->pipe->bind_vs_state(st_ctx->pipe, NULL); + st_ctx->pipe->delete_vs_state(st_ctx->pipe, st_ctx->vs); + } + + if(st_ctx->fs) { + st_ctx->pipe->bind_fs_state(st_ctx->pipe, NULL); + st_ctx->pipe->delete_fs_state(st_ctx->pipe, st_ctx->fs); + } + + if(st_ctx->cso) + cso_destroy_context(st_ctx->cso); + + if(st_ctx->pipe) + st_ctx->st_dev->st_ws->context_destroy(st_ctx->pipe); + + for(i = 0; i < PIPE_MAX_SAMPLERS; ++i) + pipe_texture_reference(&st_ctx->sampler_textures[i], NULL); + + FREE(st_ctx); + + if(!--st_dev->refcount) + st_device_really_destroy(st_dev); + } +} + + +struct st_context * +st_context_create(struct st_device *st_dev) +{ + struct st_context *st_ctx; + + st_ctx = CALLOC_STRUCT(st_context); + if(!st_ctx) + return NULL; + + st_ctx->st_dev = st_dev; + ++st_dev->refcount; + + st_ctx->pipe = st_dev->st_ws->context_create(st_dev->screen); + if(!st_ctx->pipe) + st_context_destroy(st_ctx); + + st_ctx->cso = cso_create_context(st_ctx->pipe); + if(!st_ctx->cso) + st_context_destroy(st_ctx); + + /* vertex shader */ + { + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, + TGSI_SEMANTIC_GENERIC }; + const uint semantic_indexes[] = { 0, 0 }; + st_ctx->vs = util_make_vertex_passthrough_shader(st_ctx->pipe, + 2, + semantic_names, + semantic_indexes, + &st_ctx->vert_shader); + } + + /* fragment shader */ + st_ctx->fs = util_make_fragment_passthrough_shader(st_ctx->pipe, + &st_ctx->frag_shader); + + st_ctx->pipe->bind_fs_state(st_ctx->pipe, st_ctx->fs); + st_ctx->pipe->bind_vs_state(st_ctx->pipe, st_ctx->vs); + + return st_ctx; +} diff --git a/src/gallium/state_trackers/python/st_device.h b/src/gallium/state_trackers/python/st_device.h new file mode 100644 index 0000000000..46db20acd7 --- /dev/null +++ b/src/gallium/state_trackers/python/st_device.h @@ -0,0 +1,83 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_DEVICE_H_ +#define ST_DEVICE_H_ + + +#include "pipe/p_state.h" + +struct cso_context; +struct pipe_screen; +struct pipe_context; +struct st_winsys; + + +struct st_context { + struct st_device *st_dev; + + struct pipe_context *pipe; + + struct cso_context *cso; + + struct pipe_shader_state vert_shader; + struct pipe_shader_state frag_shader; + + void *vs; + void *fs; + + struct pipe_texture *sampler_textures[PIPE_MAX_SAMPLERS]; + struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; + struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS]; +}; + + +struct st_device { + const struct st_winsys *st_ws; + + struct pipe_screen *screen; + + /* FIXME: we also need to refcount for textures and surfaces... */ + unsigned refcount; +}; + + +struct st_context * +st_context_create(struct st_device *st_dev); + +void +st_context_destroy(struct st_context *st_ctx); + +struct st_device * +st_device_create(boolean hardware); + +void +st_device_destroy(struct st_device *st_dev); + + +#endif /* ST_DEVICE_H_ */ diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c new file mode 100644 index 0000000000..964d60de1d --- /dev/null +++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c @@ -0,0 +1,321 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * @file + * Softpipe support. + * + * @author Keith Whitwell + * @author Brian Paul + * @author Jose Fonseca + */ + + +#include "pipe/p_winsys.h" +#include "pipe/p_format.h" +#include "pipe/p_context.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "softpipe/sp_winsys.h" +#include "st_winsys.h" + + +struct st_softpipe_buffer +{ + struct pipe_buffer base; + boolean userBuffer; /** Is this a user-space buffer? */ + void *data; + void *mapped; +}; + + +/** Cast wrapper */ +static INLINE struct st_softpipe_buffer * +st_softpipe_buffer( struct pipe_buffer *buf ) +{ + return (struct st_softpipe_buffer *)buf; +} + + +static void * +st_softpipe_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags) +{ + struct st_softpipe_buffer *st_softpipe_buf = st_softpipe_buffer(buf); + st_softpipe_buf->mapped = st_softpipe_buf->data; + return st_softpipe_buf->mapped; +} + + +static void +st_softpipe_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + struct st_softpipe_buffer *st_softpipe_buf = st_softpipe_buffer(buf); + st_softpipe_buf->mapped = NULL; +} + + +static void +st_softpipe_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + struct st_softpipe_buffer *oldBuf = st_softpipe_buffer(buf); + + if (oldBuf->data) { + if (!oldBuf->userBuffer) + align_free(oldBuf->data); + + oldBuf->data = NULL; + } + + FREE(oldBuf); +} + + +static void +st_softpipe_flush_frontbuffer(struct pipe_winsys *winsys, + struct pipe_surface *surf, + void *context_private) +{ +} + + + +static const char * +st_softpipe_get_name(struct pipe_winsys *winsys) +{ + return "softpipe"; +} + + +static struct pipe_buffer * +st_softpipe_buffer_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct st_softpipe_buffer *buffer = CALLOC_STRUCT(st_softpipe_buffer); + + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + + buffer->data = align_malloc(size, alignment); + + return &buffer->base; +} + + +/** + * Create buffer which wraps user-space data. + */ +static struct pipe_buffer * +st_softpipe_user_buffer_create(struct pipe_winsys *winsys, + void *ptr, + unsigned bytes) +{ + struct st_softpipe_buffer *buffer; + + buffer = CALLOC_STRUCT(st_softpipe_buffer); + if(!buffer) + return NULL; + + buffer->base.refcount = 1; + buffer->base.size = bytes; + buffer->userBuffer = TRUE; + buffer->data = ptr; + + return &buffer->base; +} + + +/** + * Round n up to next multiple. + */ +static INLINE unsigned +round_up(unsigned n, unsigned multiple) +{ + return (n + multiple - 1) & ~(multiple - 1); +} + + +static int +st_softpipe_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags, + unsigned tex_usage) +{ + const unsigned alignment = 64; + + surf->width = width; + surf->height = height; + surf->format = format; + pf_get_block(format, &surf->block); + surf->nblocksx = pf_get_nblocksx(&surf->block, width); + surf->nblocksy = pf_get_nblocksy(&surf->block, height); + surf->stride = round_up(surf->nblocksx * surf->block.size, alignment); + surf->usage = flags; + + assert(!surf->buffer); + surf->buffer = winsys->buffer_create(winsys, alignment, + PIPE_BUFFER_USAGE_PIXEL, + surf->stride * surf->nblocksy); + if(!surf->buffer) + return -1; + + return 0; +} + + +static struct pipe_surface * +st_softpipe_surface_alloc(struct pipe_winsys *winsys) +{ + struct pipe_surface *surface = CALLOC_STRUCT(pipe_surface); + + assert(winsys); + + surface->refcount = 1; + surface->winsys = winsys; + + return surface; +} + + +static void +st_softpipe_surface_release(struct pipe_winsys *winsys, + struct pipe_surface **s) +{ + struct pipe_surface *surf = *s; + assert(!surf->texture); + surf->refcount--; + if (surf->refcount == 0) { + if (surf->buffer) + pipe_buffer_reference(winsys, &surf->buffer, NULL); + free(surf); + } + *s = NULL; +} + + +static void +st_softpipe_fence_reference(struct pipe_winsys *winsys, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence) +{ +} + + +static int +st_softpipe_fence_signalled(struct pipe_winsys *winsys, + struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + + +static int +st_softpipe_fence_finish(struct pipe_winsys *winsys, + struct pipe_fence_handle *fence, + unsigned flag) +{ + return 0; +} + + +static void +st_softpipe_screen_destroy(struct pipe_screen *screen) +{ + struct pipe_winsys *winsys = screen->winsys; + + screen->destroy(screen); + + FREE(winsys); +} + + +static struct pipe_screen * +st_softpipe_screen_create(void) +{ + static struct pipe_winsys *winsys; + struct pipe_screen *screen; + + winsys = CALLOC_STRUCT(pipe_winsys); + if(!winsys) + return NULL; + + winsys->buffer_create = st_softpipe_buffer_create; + winsys->user_buffer_create = st_softpipe_user_buffer_create; + winsys->buffer_map = st_softpipe_buffer_map; + winsys->buffer_unmap = st_softpipe_buffer_unmap; + winsys->buffer_destroy = st_softpipe_buffer_destroy; + + winsys->surface_alloc = st_softpipe_surface_alloc; + winsys->surface_alloc_storage = st_softpipe_surface_alloc_storage; + winsys->surface_release = st_softpipe_surface_release; + + winsys->fence_reference = st_softpipe_fence_reference; + winsys->fence_signalled = st_softpipe_fence_signalled; + winsys->fence_finish = st_softpipe_fence_finish; + + winsys->flush_frontbuffer = st_softpipe_flush_frontbuffer; + winsys->get_name = st_softpipe_get_name; + + screen = softpipe_create_screen(winsys); + if(!screen) + FREE(winsys); + + return screen; +} + + +static void +st_softpipe_context_destroy(struct pipe_context *pipe) +{ + pipe->destroy(pipe); +} + + +static struct pipe_context * +st_softpipe_context_create(struct pipe_screen *screen) +{ + return softpipe_create(screen, screen->winsys, NULL); +} + + +const struct st_winsys st_software_winsys = { + &st_softpipe_screen_create, + &st_softpipe_screen_destroy, + &st_softpipe_context_create, + &st_softpipe_context_destroy +}; diff --git a/src/gallium/state_trackers/python/st_winsys.h b/src/gallium/state_trackers/python/st_winsys.h new file mode 100644 index 0000000000..992fc9ab4b --- /dev/null +++ b/src/gallium/state_trackers/python/st_winsys.h @@ -0,0 +1,58 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_WINSYS_H_ +#define ST_WINSYS_H_ + + +struct pipe_screen; +struct pipe_context; + + +struct st_winsys +{ + struct pipe_screen * + (*screen_create)(void); + + void + (*screen_destroy)(struct pipe_screen *screen); + + struct pipe_context * + (*context_create)(struct pipe_screen *screen); + + void + (*context_destroy)(struct pipe_context *pipe); +}; + + +extern const struct st_winsys st_software_winsys; + +extern const struct st_winsys st_hardware_winsys; + + +#endif /* ST_WINSYS_H_ */ -- cgit v1.2.3 From 930a863c4f6f11d0fd5cf396ef76054d52c69b9f Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Thu, 10 Jul 2008 20:20:46 +0200 Subject: i915: WIP swap rework --- src/gallium/drivers/i915simple/i915_texture.c | 6 +- src/gallium/winsys/dri/intel/intel_context.h | 6 + src/gallium/winsys/dri/intel/intel_swapbuffers.c | 303 ++++++++++------------- src/gallium/winsys/dri/intel/intel_swapbuffers.h | 4 +- 4 files changed, 147 insertions(+), 172 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index cf4964b26b..8427d8271a 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -79,7 +79,7 @@ static unsigned power_of_two(unsigned x) { unsigned value = 1; - while (value <= x) + while (value < x) value = value << 1; return value; } @@ -205,8 +205,8 @@ i945_miptree_layout_2d( struct i915_texture *tex ) unsigned nblocksx = pt->nblocksx[0]; unsigned nblocksy = pt->nblocksy[0]; -#if 0 /* used for tiled display targets */ - if (pt->last_level == 0 && pt->cpp == 4) +#if 1 /* used for tiled display targets */ + if (pt->last_level == 0 /*&& pt->bpp == 4*/) if (i915_displaytarget_layout(tex)) return; #endif diff --git a/src/gallium/winsys/dri/intel/intel_context.h b/src/gallium/winsys/dri/intel/intel_context.h index ced18da143..18b96217ac 100644 --- a/src/gallium/winsys/dri/intel/intel_context.h +++ b/src/gallium/winsys/dri/intel/intel_context.h @@ -32,6 +32,7 @@ #include "drm.h" #include "pipe/p_debug.h" +#include "util/u_blit.h" #include "intel_screen.h" #include "i915_drm.h" @@ -82,6 +83,11 @@ struct intel_context * Configuration cache */ driOptionCache optionCache; + + /** + * Blit state for swapbuffers + */ + struct blit_state *blit; }; diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index f751f97524..6e10235e89 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -28,16 +28,13 @@ #include "intel_screen.h" #include "intel_context.h" #include "intel_swapbuffers.h" - +#include "intel_batchbuffer.h" #include "intel_reg.h" -#include "pipe/p_context.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" -#include "state_tracker/st_cb_fbo.h" - #include "intel_drm/ws_dri_bufmgr.h" -#include "intel_batchbuffer.h" + /** * Display a colorbuffer surface in an X window. @@ -49,158 +46,130 @@ */ void intelDisplaySurface(__DRIdrawablePrivate *dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t *rect) + struct pipe_texture *tex, + const drm_clip_rect_t *rect) { - struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - struct intel_context *intel = intelScreen->dummyContext; - - DBG(SWAP, "%s\n", __FUNCTION__); - - if (!intel) { - /* XXX this is where some kind of extra/meta context could be useful */ - return; - } - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); - driFenceUnReference(&intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - * current context (which is what intelScreenContext should return) might - * not get a contended lock and thus cliprects not updated (tests/manywin) - */ - if (intel_context(dPriv->driContextPriv) != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - - if (dPriv && dPriv->numClipRects) { - const int srcWidth = surf->width; - const int srcHeight = surf->height; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; - const int cpp = intelScreen->front.cpp; - const int srcpitch = surf->stride / cpp; - int BR13, CMD; - int i; - - ASSERT(surf->buffer); - ASSERT(surf->cpp == cpp); - - DBG(SWAP, "screen pitch %d src surface pitch %d\n", - pitch, surf->stride); - - if (cpp == 2) { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - } - else { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - } - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) { - /* invalid cliprect, skip it */ - continue; - } - - box = *pbox; - - if (rect) { - /* intersect cliprect with user-provided src rect */ - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > srcWidth) - box.x2 = srcWidth + box.x1; - if (box.y2 - box.y1 > srcHeight) - box.y2 = srcHeight + box.y1; - - DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - assert(box.x1 < box.x2); - assert(box.y1 < box.y2); - - /* XXX this could be done with pipe->surface_copy() */ - /* XXX should have its own batch buffer */ - if (!BEGIN_BATCH(8, 2)) { - /* - * Since we share this batch buffer with a context - * we can't flush it since that risks a GPU lockup - */ - assert(0); - continue; - } - - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((box.y1 << 16) | box.x1); - OUT_BATCH((box.y2 << 16) | box.x2); - - OUT_RELOC(intelScreen->front.buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((sbox.y1 << 16) | sbox.x1); - OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(dri_bo(surf->buffer), - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - } - - if (intel->first_swap_fence) - driFenceUnReference(&intel->first_swap_fence); - intel->first_swap_fence = intel_be_batchbuffer_flush(intel->base.batch); - } - - UNLOCK_HARDWARE(intel); - - if (intel->lastStamp != dPriv->lastStamp) { - intelUpdateWindowSize(dPriv); - intel->lastStamp = dPriv->lastStamp; - } + struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); + struct intel_context *intel = intelScreen->dummyContext; + + DBG(SWAP, "%s\n", __FUNCTION__); + + if (!intel) { + /* XXX this is where some kind of extra/meta context could be useful */ + return; + } else if (!intel->blit) { + intel->blit = util_create_blit(intel->st->pipe, intel->st->cso_context); + } + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); + driFenceUnReference(&intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + * current context (which is what intelScreenContext should return) might + * not get a contended lock and thus cliprects not updated (tests/manywin) + */ + if (intel_context(dPriv->driContextPriv) != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + + if (dPriv && dPriv->numClipRects) { + const int srcWidth = tex->width[0]; + const int srcHeight = tex->height[0]; + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + int i; + + assert(surf->buffer); + assert(surf->cpp == cpp); + + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + drm_clip_rect_t sbox; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->front.width || + pbox->y2 > intelScreen->front.height) { + /* invalid cliprect, skip it */ + continue; + } + + box = *pbox; + + if (rect) { + /* intersect cliprect with user-provided src rect */ + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > srcWidth) + box.x2 = srcWidth + box.x1; + if (box.y2 - box.y1 > srcHeight) + box.y2 = srcHeight + box.y1; + + DBG(SWAP, "\tdbox x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + sbox.x2 = box.x2 - dPriv->x; + sbox.y2 = box.y2 - dPriv->y; + DBG(SWAP, "\tsbox x1 x2 y1 y2 %d %d %d %d\n", + sbox.x1, sbox.x2, sbox.y1, sbox.y2); + + assert(box.x1 < box.x2); + assert(box.y1 < box.y2); + + util_blit_pixels_tex(intel->blit, + tex, + sbox.x1, sbox.y1, + sbox.x2, sbox.y2, + intelScreen->front.surface, + box.x1, box.y1, + box.x2, box.y2, + 0.0, + PIPE_TEX_MIPFILTER_NEAREST); + } + + if (intel->first_swap_fence) + driFenceUnReference(&intel->first_swap_fence); + intel->first_swap_fence = intel_be_batchbuffer_flush(intel->base.batch); + } + + + UNLOCK_HARDWARE(intel); + + if (intel->lastStamp != dPriv->lastStamp) { + intelUpdateWindowSize(dPriv); + intel->lastStamp = dPriv->lastStamp; + } } - - /** * This will be called whenever the currently bound window is moved/resized. */ @@ -212,27 +181,27 @@ intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); } - - +/** + * Called from glx code + */ void intelSwapBuffers(__DRIdrawablePrivate * dPriv) { struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; + struct pipe_texture *back_tex; assert(intel_fb); assert(intel_fb->stfb); - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { + back_tex = st_get_framebuffer_texture(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_tex) { st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, NULL); + intelDisplaySurface(dPriv, back_tex, NULL); st_notify_swapbuffers_complete(intel_fb->stfb); } } - /** * Called via glXCopySubBufferMESA() to copy a subrect of the back * buffer to the front buffer/screen. @@ -241,14 +210,14 @@ void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) { struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; + struct pipe_texture *back_tex; assert(intel_fb); assert(intel_fb->stfb); - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { + back_tex = st_get_framebuffer_texture(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_tex) { drm_clip_rect_t rect; rect.x1 = x; rect.y1 = y; @@ -256,6 +225,6 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) rect.y2 = h; st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, &rect); + intelDisplaySurface(dPriv, back_tex, &rect); } } diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.h b/src/gallium/winsys/dri/intel/intel_swapbuffers.h index 46c9bab3af..88c479f81d 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.h +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.h @@ -29,11 +29,11 @@ #define INTEL_SWAPBUFFERS_H -struct pipe_surface; +struct pipe_texture; extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, - struct pipe_surface *surf, + struct pipe_texture *surf, const drm_clip_rect_t * rect); extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); -- cgit v1.2.3 From e59086ab91281cafcae094164bf4d8454df337ae Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Sun, 13 Jul 2008 18:04:49 +0200 Subject: EGL: Only build EGL dri based drivers under linux-dri based configs --- src/egl/drivers/Makefile | 21 ++------------------- 1 file changed, 2 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/Makefile b/src/egl/drivers/Makefile index 6ce0e6b66d..f0538f563f 100644 --- a/src/egl/drivers/Makefile +++ b/src/egl/drivers/Makefile @@ -3,27 +3,10 @@ TOP = ../../.. include $(TOP)/configs/current +SUBDIRS = $(EGL_DRIVERS_DIRS) -SUBDIRS = demo dri xdri - - -default: dri_subdirs - - -# (UNUSED) depending on $DRIVER_DIRS... -conditional_subdirs: - @if [ "${DRIVER_DIRS}" = "dri" ] ; then \ - $(MAKE) dri_subdirs ; \ - fi - - -dri_subdirs: - @ (cd dri ; $(MAKE)) || exit 1 - @ (cd xdri ; $(MAKE)) || exit 1 - -demo_subdir: - @ (cd demo ; $(MAKE)) || exit 1 +default: subdirs subdirs: -- cgit v1.2.3 From 16c2267d55fb14d0ffcb676540345a14ecc0f323 Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Sun, 13 Jul 2008 18:55:38 +0200 Subject: i915: Ops should not have pushed that This reverts commit 930a863c4f6f11d0fd5cf396ef76054d52c69b9f. --- src/gallium/drivers/i915simple/i915_texture.c | 6 +- src/gallium/winsys/dri/intel/intel_context.h | 6 - src/gallium/winsys/dri/intel/intel_swapbuffers.c | 303 +++++++++++++---------- src/gallium/winsys/dri/intel/intel_swapbuffers.h | 4 +- 4 files changed, 172 insertions(+), 147 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index 8427d8271a..cf4964b26b 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -79,7 +79,7 @@ static unsigned power_of_two(unsigned x) { unsigned value = 1; - while (value < x) + while (value <= x) value = value << 1; return value; } @@ -205,8 +205,8 @@ i945_miptree_layout_2d( struct i915_texture *tex ) unsigned nblocksx = pt->nblocksx[0]; unsigned nblocksy = pt->nblocksy[0]; -#if 1 /* used for tiled display targets */ - if (pt->last_level == 0 /*&& pt->bpp == 4*/) +#if 0 /* used for tiled display targets */ + if (pt->last_level == 0 && pt->cpp == 4) if (i915_displaytarget_layout(tex)) return; #endif diff --git a/src/gallium/winsys/dri/intel/intel_context.h b/src/gallium/winsys/dri/intel/intel_context.h index 18b96217ac..ced18da143 100644 --- a/src/gallium/winsys/dri/intel/intel_context.h +++ b/src/gallium/winsys/dri/intel/intel_context.h @@ -32,7 +32,6 @@ #include "drm.h" #include "pipe/p_debug.h" -#include "util/u_blit.h" #include "intel_screen.h" #include "i915_drm.h" @@ -83,11 +82,6 @@ struct intel_context * Configuration cache */ driOptionCache optionCache; - - /** - * Blit state for swapbuffers - */ - struct blit_state *blit; }; diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c index 6e10235e89..f751f97524 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.c @@ -28,13 +28,16 @@ #include "intel_screen.h" #include "intel_context.h" #include "intel_swapbuffers.h" -#include "intel_batchbuffer.h" + #include "intel_reg.h" +#include "pipe/p_context.h" #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" -#include "intel_drm/ws_dri_bufmgr.h" +#include "state_tracker/st_cb_fbo.h" +#include "intel_drm/ws_dri_bufmgr.h" +#include "intel_batchbuffer.h" /** * Display a colorbuffer surface in an X window. @@ -46,130 +49,158 @@ */ void intelDisplaySurface(__DRIdrawablePrivate *dPriv, - struct pipe_texture *tex, - const drm_clip_rect_t *rect) + struct pipe_surface *surf, + const drm_clip_rect_t *rect) { - struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - struct intel_context *intel = intelScreen->dummyContext; - - DBG(SWAP, "%s\n", __FUNCTION__); - - if (!intel) { - /* XXX this is where some kind of extra/meta context could be useful */ - return; - } else if (!intel->blit) { - intel->blit = util_create_blit(intel->st->pipe, intel->st->cso_context); - } - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); - driFenceUnReference(&intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - * current context (which is what intelScreenContext should return) might - * not get a contended lock and thus cliprects not updated (tests/manywin) - */ - if (intel_context(dPriv->driContextPriv) != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - if (dPriv && dPriv->numClipRects) { - const int srcWidth = tex->width[0]; - const int srcHeight = tex->height[0]; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - int i; - - assert(surf->buffer); - assert(surf->cpp == cpp); - - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) { - /* invalid cliprect, skip it */ - continue; - } - - box = *pbox; - - if (rect) { - /* intersect cliprect with user-provided src rect */ - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > srcWidth) - box.x2 = srcWidth + box.x1; - if (box.y2 - box.y1 > srcHeight) - box.y2 = srcHeight + box.y1; - - DBG(SWAP, "\tdbox x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - sbox.x2 = box.x2 - dPriv->x; - sbox.y2 = box.y2 - dPriv->y; - DBG(SWAP, "\tsbox x1 x2 y1 y2 %d %d %d %d\n", - sbox.x1, sbox.x2, sbox.y1, sbox.y2); - - assert(box.x1 < box.x2); - assert(box.y1 < box.y2); - - util_blit_pixels_tex(intel->blit, - tex, - sbox.x1, sbox.y1, - sbox.x2, sbox.y2, - intelScreen->front.surface, - box.x1, box.y1, - box.x2, box.y2, - 0.0, - PIPE_TEX_MIPFILTER_NEAREST); - } - - if (intel->first_swap_fence) - driFenceUnReference(&intel->first_swap_fence); - intel->first_swap_fence = intel_be_batchbuffer_flush(intel->base.batch); - } - - - UNLOCK_HARDWARE(intel); - - if (intel->lastStamp != dPriv->lastStamp) { - intelUpdateWindowSize(dPriv); - intel->lastStamp = dPriv->lastStamp; - } + struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); + struct intel_context *intel = intelScreen->dummyContext; + + DBG(SWAP, "%s\n", __FUNCTION__); + + if (!intel) { + /* XXX this is where some kind of extra/meta context could be useful */ + return; + } + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); + driFenceUnReference(&intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + * current context (which is what intelScreenContext should return) might + * not get a contended lock and thus cliprects not updated (tests/manywin) + */ + if (intel_context(dPriv->driContextPriv) != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + + + if (dPriv && dPriv->numClipRects) { + const int srcWidth = surf->width; + const int srcHeight = surf->height; + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; + const int cpp = intelScreen->front.cpp; + const int srcpitch = surf->stride / cpp; + int BR13, CMD; + int i; + + ASSERT(surf->buffer); + ASSERT(surf->cpp == cpp); + + DBG(SWAP, "screen pitch %d src surface pitch %d\n", + pitch, surf->stride); + + if (cpp == 2) { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + } + else { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + } + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + drm_clip_rect_t sbox; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->front.width || + pbox->y2 > intelScreen->front.height) { + /* invalid cliprect, skip it */ + continue; + } + + box = *pbox; + + if (rect) { + /* intersect cliprect with user-provided src rect */ + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > srcWidth) + box.x2 = srcWidth + box.x1; + if (box.y2 - box.y1 > srcHeight) + box.y2 = srcHeight + box.y1; + + DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + + assert(box.x1 < box.x2); + assert(box.y1 < box.y2); + + /* XXX this could be done with pipe->surface_copy() */ + /* XXX should have its own batch buffer */ + if (!BEGIN_BATCH(8, 2)) { + /* + * Since we share this batch buffer with a context + * we can't flush it since that risks a GPU lockup + */ + assert(0); + continue; + } + + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((box.y1 << 16) | box.x1); + OUT_BATCH((box.y2 << 16) | box.x2); + + OUT_RELOC(intelScreen->front.buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + OUT_BATCH((sbox.y1 << 16) | sbox.x1); + OUT_BATCH((srcpitch * cpp) & 0xffff); + OUT_RELOC(dri_bo(surf->buffer), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + } + + if (intel->first_swap_fence) + driFenceUnReference(&intel->first_swap_fence); + intel->first_swap_fence = intel_be_batchbuffer_flush(intel->base.batch); + } + + UNLOCK_HARDWARE(intel); + + if (intel->lastStamp != dPriv->lastStamp) { + intelUpdateWindowSize(dPriv); + intel->lastStamp = dPriv->lastStamp; + } } + + /** * This will be called whenever the currently bound window is moved/resized. */ @@ -181,27 +212,27 @@ intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); } -/** - * Called from glx code - */ + + void intelSwapBuffers(__DRIdrawablePrivate * dPriv) { struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_texture *back_tex; + struct pipe_surface *back_surf; assert(intel_fb); assert(intel_fb->stfb); - back_tex = st_get_framebuffer_texture(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_tex) { + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_tex, NULL); + intelDisplaySurface(dPriv, back_surf, NULL); st_notify_swapbuffers_complete(intel_fb->stfb); } } + /** * Called via glXCopySubBufferMESA() to copy a subrect of the back * buffer to the front buffer/screen. @@ -210,14 +241,14 @@ void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) { struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_texture *back_tex; + struct pipe_surface *back_surf; assert(intel_fb); assert(intel_fb->stfb); - back_tex = st_get_framebuffer_texture(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_tex) { + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { drm_clip_rect_t rect; rect.x1 = x; rect.y1 = y; @@ -225,6 +256,6 @@ intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) rect.y2 = h; st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_tex, &rect); + intelDisplaySurface(dPriv, back_surf, &rect); } } diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.h b/src/gallium/winsys/dri/intel/intel_swapbuffers.h index 88c479f81d..46c9bab3af 100644 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.h +++ b/src/gallium/winsys/dri/intel/intel_swapbuffers.h @@ -29,11 +29,11 @@ #define INTEL_SWAPBUFFERS_H -struct pipe_texture; +struct pipe_surface; extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, - struct pipe_texture *surf, + struct pipe_surface *surf, const drm_clip_rect_t * rect); extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); -- cgit v1.2.3 From 05a23e6c6f08d4d901dc9fc9995b3444fa6355dd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 14 Jul 2008 10:45:40 +0900 Subject: python: Allow to create/specify shaders. --- src/gallium/state_trackers/python/gallium.i | 88 ++++++++++++++++------ .../state_trackers/python/samples/simple.py | 29 +++++-- src/gallium/state_trackers/python/st_device.c | 20 ++--- 3 files changed, 97 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index d6594f3a87..53d160ef6c 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -44,9 +44,11 @@ #include "pipe/p_inlines.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" +#include "cso_cache/cso_context.h" #include "util/u_draw_quad.h" #include "util/p_tile.h" -#include "cso_cache/cso_context.h" +#include "tgsi/util/tgsi_text.h" +#include "tgsi/util/tgsi_dump.h" #include "st_device.h" @@ -201,29 +203,32 @@ struct st_context { cso_set_depth_stencil_alpha($self->cso, state); } - - void * create_fs( const struct pipe_shader_state *state ) { - return $self->pipe->create_fs_state($self->pipe, state); - } - - void bind_fs( void *state_obj ) { - $self->pipe->bind_fs_state($self->pipe, state_obj); - } - - void delete_fs( void *state_obj ) { - $self->pipe->delete_fs_state($self->pipe, state_obj); - } + void set_fragment_shader( const struct pipe_shader_state *state ) { + void *fs; + + fs = $self->pipe->create_fs_state($self->pipe, state); + if(!fs) + return; + + if(cso_set_fragment_shader_handle($self->cso, fs) != PIPE_OK) + return; - void * create_vs( const struct pipe_shader_state *state ) { - return $self->pipe->create_vs_state($self->pipe, state); - } - - void bind_vs( void *state_obj ) { - $self->pipe->bind_vs_state($self->pipe, state_obj); + cso_delete_fragment_shader($self->cso, $self->fs); + $self->fs = fs; } - - void delete_vs( void *state_obj ) { - $self->pipe->delete_vs_state($self->pipe, state_obj); + + void set_vertex_shader( const struct pipe_shader_state *state ) { + void *vs; + + vs = $self->pipe->create_vs_state($self->pipe, state); + if(!vs) + return; + + if(cso_set_vertex_shader_handle($self->cso, vs) != PIPE_OK) + return; + + cso_delete_vertex_shader($self->cso, $self->vs); + $self->vs = vs; } /* @@ -446,3 +451,42 @@ error1: } }; + + +%extend pipe_shader_state { + + pipe_shader_state(const char *text, unsigned num_tokens = 1024) { + struct tgsi_token *tokens; + struct pipe_shader_state *shader; + + tokens = MALLOC(num_tokens * sizeof(struct tgsi_token)); + if(!tokens) + goto error1; + + if(tgsi_text_translate(text, tokens, num_tokens ) != TRUE) + goto error2; + + shader = CALLOC_STRUCT(pipe_shader_state); + if(!shader) + goto error3; + + shader->tokens = tokens; + + return shader; + +error3: +error2: + FREE(tokens); +error1: + return NULL; + } + + ~pipe_shader_state() { + FREE((void*)$self->tokens); + FREE($self); + } + + void dump(unsigned flags = 0) { + tgsi_dump($self->tokens, flags); + } +} \ No newline at end of file diff --git a/src/gallium/state_trackers/python/samples/simple.py b/src/gallium/state_trackers/python/samples/simple.py index 77e182b644..6e90eec28a 100644 --- a/src/gallium/state_trackers/python/samples/simple.py +++ b/src/gallium/state_trackers/python/samples/simple.py @@ -103,7 +103,9 @@ def test(dev): ctx.set_sampler(0, sampler) # texture - texture = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, width, height, usage=PIPE_TEXTURE_USAGE_RENDER_TARGET) + texture = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, + width, height, + usage=PIPE_TEXTURE_USAGE_RENDER_TARGET) ctx.set_sampler_texture(0, texture) # drawing dest @@ -116,12 +118,29 @@ def test(dev): ctx.set_framebuffer(fb) # vertex shader - # vs = Shader() - #ctx.set_vertex_shader(vs) + vs = Shader(''' + VERT1.1 + DCL IN[0], POSITION, CONSTANT + DCL IN[1], GENERIC[0], CONSTANT + DCL OUT[0], POSITION, CONSTANT + DCL OUT[1], GENERIC[0], CONSTANT + 0:MOV OUT[0], IN[0] + 1:MOV OUT[1], IN[1] + 2:END + ''') + #vs.dump() + ctx.set_vertex_shader(vs) # fragment shader - #fs = Shader() - #ctx.set_fragment_shader(fs) + fs = Shader(''' + FRAG1.1 + DCL IN[0], COLOR, CONSTANT + DCL OUT[0], COLOR, CONSTANT + 0:MOV OUT[0], IN[0] + 1:END + ''') + #fs.dump() + ctx.set_fragment_shader(fs) if 0: nverts = 4 diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index 430a7af176..e9002b493d 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -99,18 +99,12 @@ st_context_destroy(struct st_context *st_ctx) if(st_ctx) { struct st_device *st_dev = st_ctx->st_dev; - if(st_ctx->vs) { - st_ctx->pipe->bind_vs_state(st_ctx->pipe, NULL); - st_ctx->pipe->delete_vs_state(st_ctx->pipe, st_ctx->vs); - } - - if(st_ctx->fs) { - st_ctx->pipe->bind_fs_state(st_ctx->pipe, NULL); - st_ctx->pipe->delete_fs_state(st_ctx->pipe, st_ctx->fs); - } - - if(st_ctx->cso) + if(st_ctx->cso) { + cso_delete_vertex_shader(st_ctx->cso, st_ctx->vs); + cso_delete_fragment_shader(st_ctx->cso, st_ctx->fs); + cso_destroy_context(st_ctx->cso); + } if(st_ctx->pipe) st_ctx->st_dev->st_ws->context_destroy(st_ctx->pipe); @@ -162,8 +156,8 @@ st_context_create(struct st_device *st_dev) st_ctx->fs = util_make_fragment_passthrough_shader(st_ctx->pipe, &st_ctx->frag_shader); - st_ctx->pipe->bind_fs_state(st_ctx->pipe, st_ctx->fs); - st_ctx->pipe->bind_vs_state(st_ctx->pipe, st_ctx->vs); + cso_set_fragment_shader_handle(st_ctx->cso, st_ctx->fs); + cso_set_vertex_shader_handle(st_ctx->cso, st_ctx->vs); return st_ctx; } -- cgit v1.2.3 From 3679f690230b9691157d44cfa5baf947d84e1487 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 14 Jul 2008 11:38:44 +0900 Subject: python: Set default state. --- src/gallium/state_trackers/python/gallium.i | 2 + src/gallium/state_trackers/python/st_device.c | 120 ++++++++++++++++++++++++-- src/gallium/state_trackers/python/st_device.h | 4 +- 3 files changed, 116 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 53d160ef6c..6e355b7975 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -266,6 +266,8 @@ struct st_context { void set_sampler_texture(unsigned index, struct pipe_texture *texture) { + if(!texture) + texture = $self->default_texture; pipe_texture_reference(&$self->sampler_textures[index], texture); $self->pipe->set_sampler_textures($self->pipe, PIPE_MAX_SAMPLERS, diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index e9002b493d..25b5a4fa8c 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -111,7 +111,8 @@ st_context_destroy(struct st_context *st_ctx) for(i = 0; i < PIPE_MAX_SAMPLERS; ++i) pipe_texture_reference(&st_ctx->sampler_textures[i], NULL); - + pipe_texture_reference(&st_ctx->default_texture, NULL); + FREE(st_ctx); if(!--st_dev->refcount) @@ -140,8 +141,111 @@ st_context_create(struct st_device *st_dev) if(!st_ctx->cso) st_context_destroy(st_ctx); + /* disabled blending/masking */ + { + struct pipe_blend_state blend; + memset(&blend, 0, sizeof(blend)); + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE; + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE; + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO; + blend.colormask = PIPE_MASK_RGBA; + cso_set_blend(st_ctx->cso, &blend); + } + + /* no-op depth/stencil/alpha */ + { + struct pipe_depth_stencil_alpha_state depthstencil; + memset(&depthstencil, 0, sizeof(depthstencil)); + cso_set_depth_stencil_alpha(st_ctx->cso, &depthstencil); + } + + /* rasterizer */ + { + struct pipe_rasterizer_state rasterizer; + memset(&rasterizer, 0, sizeof(rasterizer)); + rasterizer.front_winding = PIPE_WINDING_CW; + rasterizer.cull_mode = PIPE_WINDING_NONE; + rasterizer.bypass_clipping = 1; + /*rasterizer.bypass_vs = 1;*/ + cso_set_rasterizer(st_ctx->cso, &rasterizer); + } + + /* identity viewport */ + { + struct pipe_viewport_state viewport; + viewport.scale[0] = 1.0; + viewport.scale[1] = 1.0; + viewport.scale[2] = 1.0; + viewport.scale[3] = 1.0; + viewport.translate[0] = 0.0; + viewport.translate[1] = 0.0; + viewport.translate[2] = 0.0; + viewport.translate[3] = 0.0; + cso_set_viewport(st_ctx->cso, &viewport); + } + + /* samplers */ + { + struct pipe_sampler_state sampler; + unsigned i; + memset(&sampler, 0, sizeof(sampler)); + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; + sampler.normalized_coords = 1; + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + cso_single_sampler(st_ctx->cso, i, &sampler); + cso_single_sampler_done(st_ctx->cso); + } + + /* default textures */ + { + struct pipe_screen *screen = st_dev->screen; + struct pipe_texture templat; + struct pipe_surface *surface; + unsigned i; + + memset( &templat, 0, sizeof( templat ) ); + templat.target = PIPE_TEXTURE_2D; + templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; + templat.block.size = 4; + templat.block.width = 1; + templat.block.height = 1; + templat.width[0] = 1; + templat.height[0] = 1; + templat.depth[0] = 1; + templat.last_level = 0; + + st_ctx->default_texture = screen->texture_create( screen, &templat ); + if(st_ctx->default_texture) { + surface = screen->get_tex_surface( screen, + st_ctx->default_texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_WRITE ); + if(surface) { + uint32_t *map; + map = (uint32_t *) pipe_surface_map(surface, PIPE_BUFFER_USAGE_CPU_WRITE ); + if(map) { + *map = 0x00000000; + pipe_surface_unmap( surface ); + } + pipe_surface_reference(&surface, NULL); + } + } + + for (i = 0; i < PIPE_MAX_SAMPLERS; i++) + pipe_texture_reference(&st_ctx->sampler_textures[i], st_ctx->default_texture); + + cso_set_sampler_textures(st_ctx->cso, PIPE_MAX_SAMPLERS, st_ctx->sampler_textures); + } + /* vertex shader */ { + struct pipe_shader_state vert_shader; + const uint semantic_names[] = { TGSI_SEMANTIC_POSITION, TGSI_SEMANTIC_GENERIC }; const uint semantic_indexes[] = { 0, 0 }; @@ -149,15 +253,17 @@ st_context_create(struct st_device *st_dev) 2, semantic_names, semantic_indexes, - &st_ctx->vert_shader); + &vert_shader); + cso_set_vertex_shader_handle(st_ctx->cso, st_ctx->vs); } /* fragment shader */ - st_ctx->fs = util_make_fragment_passthrough_shader(st_ctx->pipe, - &st_ctx->frag_shader); - - cso_set_fragment_shader_handle(st_ctx->cso, st_ctx->fs); - cso_set_vertex_shader_handle(st_ctx->cso, st_ctx->vs); + { + struct pipe_shader_state frag_shader; + st_ctx->fs = util_make_fragment_passthrough_shader(st_ctx->pipe, + &frag_shader); + cso_set_fragment_shader_handle(st_ctx->cso, st_ctx->fs); + } return st_ctx; } diff --git a/src/gallium/state_trackers/python/st_device.h b/src/gallium/state_trackers/python/st_device.h index 46db20acd7..2d95c2da73 100644 --- a/src/gallium/state_trackers/python/st_device.h +++ b/src/gallium/state_trackers/python/st_device.h @@ -45,12 +45,10 @@ struct st_context { struct cso_context *cso; - struct pipe_shader_state vert_shader; - struct pipe_shader_state frag_shader; - void *vs; void *fs; + struct pipe_texture *default_texture; struct pipe_texture *sampler_textures[PIPE_MAX_SAMPLERS]; struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS]; -- cgit v1.2.3 From f07ad529af96b903e7b19fa26c3372d16b507bf1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 14 Jul 2008 11:44:21 +0900 Subject: python: Surface clears. --- src/gallium/state_trackers/python/gallium.i | 2 +- src/gallium/state_trackers/python/samples/simple.py | 2 ++ 2 files changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 6e355b7975..f616037ec2 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -367,7 +367,7 @@ error1: $self->pipe->surface_fill($self->pipe, dst, x, y, width, height, value); } - void clear(struct pipe_surface *surface, unsigned value) { + void surface_clear(struct pipe_surface *surface, unsigned value = 0) { $self->pipe->clear($self->pipe, surface, value); } diff --git a/src/gallium/state_trackers/python/samples/simple.py b/src/gallium/state_trackers/python/samples/simple.py index 6e90eec28a..10742db0dc 100644 --- a/src/gallium/state_trackers/python/samples/simple.py +++ b/src/gallium/state_trackers/python/samples/simple.py @@ -142,6 +142,8 @@ def test(dev): #fs.dump() ctx.set_fragment_shader(fs) + ctx.surface_clear(surface, 0x00ff0000) + if 0: nverts = 4 nattrs = 1 -- cgit v1.2.3 From 00cd96b68f21cb281aa30199f520aae8b2f93083 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 14 Jul 2008 12:40:50 +0900 Subject: python: Get the sample app to draw triangle like trivial/tri exanple. --- .../state_trackers/python/samples/simple.py | 179 ------------------- src/gallium/state_trackers/python/samples/tri.py | 193 +++++++++++++++++++++ 2 files changed, 193 insertions(+), 179 deletions(-) delete mode 100644 src/gallium/state_trackers/python/samples/simple.py create mode 100644 src/gallium/state_trackers/python/samples/tri.py (limited to 'src') diff --git a/src/gallium/state_trackers/python/samples/simple.py b/src/gallium/state_trackers/python/samples/simple.py deleted file mode 100644 index 10742db0dc..0000000000 --- a/src/gallium/state_trackers/python/samples/simple.py +++ /dev/null @@ -1,179 +0,0 @@ -#!/usr/bin/env python -########################################################################## -# -# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. -# All Rights Reserved. -# -# Permission is hereby granted, free of charge, to any person obtaining a -# copy of this software and associated documentation files (the -# "Software"), to deal in the Software without restriction, including -# without limitation the rights to use, copy, modify, merge, publish, -# distribute, sub license, and/or sell copies of the Software, and to -# permit persons to whom the Software is furnished to do so, subject to -# the following conditions: -# -# The above copyright notice and this permission notice (including the -# next paragraph) shall be included in all copies or substantial portions -# of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. -# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR -# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, -# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE -# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. -# -########################################################################## - - -from gallium import * - - -def save_image(filename, surface): - pixels = FloatArray(surface.height*surface.width*4) - surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) - - import Image - outimage = Image.new( - mode='RGB', - size=(surface.width, surface.height), - color=(0,0,0)) - outpixels = outimage.load() - for y in range(0, surface.height): - for x in range(0, surface.width): - offset = (y*surface.width + x)*4 - r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] - outpixels[x, y] = r, g, b - outimage.save(filename, "PNG") - - -def test(dev): - ctx = dev.context_create() - - width = 256 - height = 256 - - # disabled blending/masking - blend = Blend() - blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE - blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE - blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO - blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO - blend.colormask = PIPE_MASK_RGBA - ctx.set_blend(blend) - - # no-op depth/stencil/alpha - depth_stencil_alpha = DepthStencilAlpha() - ctx.set_depth_stencil_alpha(depth_stencil_alpha) - - # rasterizer - rasterizer = Rasterizer() - rasterizer.front_winding = PIPE_WINDING_CW - rasterizer.cull_mode = PIPE_WINDING_NONE - rasterizer.bypass_clipping = 1 - #rasterizer.bypass_vs = 1 - ctx.set_rasterizer(rasterizer) - - # viewport (identity, we setup vertices in wincoords) - viewport = Viewport() - scale = FloatArray(4) - scale[0] = 1.0 - scale[1] = 1.0 - scale[2] = 1.0 - scale[3] = 1.0 - viewport.scale = scale - translate = FloatArray(4) - translate[0] = 0.0 - translate[1] = 0.0 - translate[2] = 0.0 - translate[3] = 0.0 - viewport.translate = translate - ctx.set_viewport(viewport) - - # samplers - sampler = Sampler() - sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE - sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE - sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE - sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST - sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST - sampler.normalized_coords = 1 - ctx.set_sampler(0, sampler) - - # texture - texture = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, - width, height, - usage=PIPE_TEXTURE_USAGE_RENDER_TARGET) - ctx.set_sampler_texture(0, texture) - - # drawing dest - surface = texture.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE) - fb = Framebuffer() - fb.width = surface.width - fb.height = surface.height - fb.num_cbufs = 1 - fb.set_cbuf(0, surface) - ctx.set_framebuffer(fb) - - # vertex shader - vs = Shader(''' - VERT1.1 - DCL IN[0], POSITION, CONSTANT - DCL IN[1], GENERIC[0], CONSTANT - DCL OUT[0], POSITION, CONSTANT - DCL OUT[1], GENERIC[0], CONSTANT - 0:MOV OUT[0], IN[0] - 1:MOV OUT[1], IN[1] - 2:END - ''') - #vs.dump() - ctx.set_vertex_shader(vs) - - # fragment shader - fs = Shader(''' - FRAG1.1 - DCL IN[0], COLOR, CONSTANT - DCL OUT[0], COLOR, CONSTANT - 0:MOV OUT[0], IN[0] - 1:END - ''') - #fs.dump() - ctx.set_fragment_shader(fs) - - ctx.surface_clear(surface, 0x00ff0000) - - if 0: - nverts = 4 - nattrs = 1 - vertices = FloatArray(n_verts * nattrs * 4) - - # init vertex data that doesn't change - for i in range(nverts): - for j in range(nattrs): - vertices[(i*nattrs +j)*4 + 0] = 0.0 - vertices[(i*nattrs +j)*4 + 1] = 0.0 - vertices[(i*nattrs +j)*4 + 2] = 0.0 - vertices[(i*nattrs +j)*4 + 3] = 0.0 - - ctx.draw_vertices(PIPE_PRIM_TRIANGLE_FAN, - 4, # verts - 2, # attribs/vert - vertices) - else: - ctx.draw_quad(32.0, 32.0, 224.0, 224.0) - - ctx.flush() - - save_image("simple.png", surface) - - - -def main(): - dev = Device(0) - test(dev) - - -if __name__ == '__main__': - main() diff --git a/src/gallium/state_trackers/python/samples/tri.py b/src/gallium/state_trackers/python/samples/tri.py new file mode 100644 index 0000000000..7a10905ddb --- /dev/null +++ b/src/gallium/state_trackers/python/samples/tri.py @@ -0,0 +1,193 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +from gallium import * + + +def save_image(filename, surface): + pixels = FloatArray(surface.height*surface.width*4) + surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) + + import Image + outimage = Image.new( + mode='RGB', + size=(surface.width, surface.height), + color=(0,0,0)) + outpixels = outimage.load() + for y in range(0, surface.height): + for x in range(0, surface.width): + offset = (y*surface.width + x)*4 + r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] + outpixels[x, y] = r, g, b + outimage.save(filename, "PNG") + + +def test(dev): + ctx = dev.context_create() + + width = 256 + height = 256 + + # disabled blending/masking + blend = Blend() + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.colormask = PIPE_MASK_RGBA + ctx.set_blend(blend) + + # no-op depth/stencil/alpha + depth_stencil_alpha = DepthStencilAlpha() + ctx.set_depth_stencil_alpha(depth_stencil_alpha) + + # rasterizer + rasterizer = Rasterizer() + rasterizer.front_winding = PIPE_WINDING_CW + rasterizer.cull_mode = PIPE_WINDING_NONE + rasterizer.bypass_clipping = 1 + #rasterizer.bypass_vs = 1 + ctx.set_rasterizer(rasterizer) + + # viewport (identity, we setup vertices in wincoords) + viewport = Viewport() + scale = FloatArray(4) + scale[0] = 1.0 + scale[1] = 1.0 + scale[2] = 1.0 + scale[3] = 1.0 + viewport.scale = scale + translate = FloatArray(4) + translate[0] = 0.0 + translate[1] = 0.0 + translate[2] = 0.0 + translate[3] = 0.0 + viewport.translate = translate + ctx.set_viewport(viewport) + + # samplers + sampler = Sampler() + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.normalized_coords = 1 + ctx.set_sampler(0, sampler) + + # texture + texture = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, + width, height, + usage=PIPE_TEXTURE_USAGE_RENDER_TARGET) + ctx.set_sampler_texture(0, texture) + + # drawing dest + surface = texture.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE) + fb = Framebuffer() + fb.width = surface.width + fb.height = surface.height + fb.num_cbufs = 1 + fb.set_cbuf(0, surface) + ctx.set_framebuffer(fb) + + # vertex shader + vs = Shader(''' + VERT1.1 + DCL IN[0], POSITION, CONSTANT + DCL IN[1], COLOR, CONSTANT + DCL OUT[0], POSITION, CONSTANT + DCL OUT[1], COLOR, CONSTANT + 0:MOV OUT[0], IN[0] + 1:MOV OUT[1], IN[1] + 2:END + ''') + #vs.dump() + ctx.set_vertex_shader(vs) + + # fragment shader + fs = Shader(''' + FRAG1.1 + DCL IN[0], COLOR, LINEAR + DCL OUT[0], COLOR, CONSTANT + 0:MOV OUT[0], IN[0] + 1:END + ''') + #fs.dump() + ctx.set_fragment_shader(fs) + + nverts = 3 + nattrs = 2 + verts = FloatArray(nverts * nattrs * 4) + + verts[ 0] = 128.0 # x1 + verts[ 1] = 32.0 # y1 + verts[ 2] = 0.0 # z1 + verts[ 3] = 1.0 # w1 + verts[ 4] = 1.0 # r1 + verts[ 5] = 0.0 # g1 + verts[ 6] = 0.0 # b1 + verts[ 7] = 1.0 # a1 + verts[ 8] = 32.0 # x2 + verts[ 9] = 224.0 # y2 + verts[10] = 0.0 # z2 + verts[11] = 1.0 # w2 + verts[12] = 0.0 # r2 + verts[13] = 1.0 # g2 + verts[14] = 0.0 # b2 + verts[15] = 1.0 # a2 + verts[16] = 224.0 # x3 + verts[17] = 224.0 # y3 + verts[18] = 0.0 # z3 + verts[19] = 1.0 # w3 + verts[20] = 0.0 # r3 + verts[21] = 0.0 # g3 + verts[22] = 1.0 # b3 + verts[23] = 1.0 # a3 + + ctx.surface_clear(surface, 0x00000000) + + ctx.draw_vertices(PIPE_PRIM_TRIANGLES, + nverts, + nattrs, + verts) + + ctx.flush() + + save_image("tri.png", surface) + + + +def main(): + dev = Device(hardware = False) + test(dev) + + +if __name__ == '__main__': + main() -- cgit v1.2.3 From 3aea28b9b1e9100be6836d4f88444e3cdfe5c4be Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 14 Jul 2008 22:26:40 +0900 Subject: python: Raw tile access to surface. --- src/gallium/state_trackers/python/gallium.i | 11 +++++++++++ 1 file changed, 11 insertions(+) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index f616037ec2..60f8e50300 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -55,6 +55,7 @@ %} %include "carrays.i" +%array_class(unsigned char, ByteArray); %array_class(int, IntArray); %array_class(float, FloatArray); @@ -405,6 +406,16 @@ error1: // gets mapped to pipe_surface_unmap automatically void unmap( void ); + void + get_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, unsigned char *p, unsigned stride) { + pipe_get_tile_raw($self, x, y, w, h, p, stride); + } + + void + put_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned char *p, unsigned stride) { + pipe_put_tile_raw($self, x, y, w, h, p, stride); + } + void get_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, float *p) { pipe_get_tile_rgba($self, x, y, w, h, p); -- cgit v1.2.3 From 64979d618a24e8378d95944d5f1d1b5847880a48 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 14 Jul 2008 22:27:40 +0900 Subject: python: Basic test case for 2d texture. --- src/gallium/state_trackers/python/tests/base.py | 96 ++++++++ src/gallium/state_trackers/python/tests/data.py | 274 +++++++++++++++++++++ src/gallium/state_trackers/python/tests/texture.py | 236 ++++++++++++++++++ 3 files changed, 606 insertions(+) create mode 100644 src/gallium/state_trackers/python/tests/base.py create mode 100644 src/gallium/state_trackers/python/tests/data.py create mode 100644 src/gallium/state_trackers/python/tests/texture.py (limited to 'src') diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py new file mode 100644 index 0000000000..5778ad4382 --- /dev/null +++ b/src/gallium/state_trackers/python/tests/base.py @@ -0,0 +1,96 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +from gallium import * + + +# Enumerate all pixel formats +formats = {} +for name, value in globals().items(): + if name.startswith("PIPE_FORMAT_") and isinstance(value, int): + formats[value] = name + + +def save_image(filename, surface): + pixels = FloatArray(surface.height*surface.width*4) + surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) + + import Image + outimage = Image.new( + mode='RGB', + size=(surface.width, surface.height), + color=(0,0,0)) + outpixels = outimage.load() + for y in range(0, surface.height): + for x in range(0, surface.width): + offset = (y*surface.width + x)*4 + r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] + outpixels[x, y] = r, g, b + outimage.save(filename, "PNG") + + + +class Test: + + def __init__(self): + pass + + def run(self): + raise NotImplementedError + + +class TestSuite(Test): + + def __init__(self, tests = None): + Test.__init__(self) + if tests is None: + self.tests = [] + else: + self.tests = tests + + def add_test(self, test): + self.tests.append(test) + + def run(self): + for test in self.tests: + self.test.run() + + +class TextureTemplate: + + def __init__(self, format=PIPE_FORMAT_R8G8B8A8_UNORM, width=1, height=1, depth=1, last_level=0, target=PIPE_TEXTURE_2D): + self.format = format + self.width = width + self.height = height + self.depth = depth + self.last_level = last_level + self.target = target + + + diff --git a/src/gallium/state_trackers/python/tests/data.py b/src/gallium/state_trackers/python/tests/data.py new file mode 100644 index 0000000000..2287884285 --- /dev/null +++ b/src/gallium/state_trackers/python/tests/data.py @@ -0,0 +1,274 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + +"""Texture data generation for use in testing.""" + + +import random + +from gallium import * + + +dxt1_rgb_data = [ + ( + [ + [0x99, 0xb0, 0x8e, 0xff], + [0x5d, 0x62, 0x89, 0xff], + [0x99, 0xb0, 0x8e, 0xff], + [0x99, 0xb0, 0x8e, 0xff], + [0xd6, 0xff, 0x94, 0xff], + [0x5d, 0x62, 0x89, 0xff], + [0x99, 0xb0, 0x8e, 0xff], + [0xd6, 0xff, 0x94, 0xff], + [0x5d, 0x62, 0x89, 0xff], + [0x5d, 0x62, 0x89, 0xff], + [0x99, 0xb0, 0x8e, 0xff], + [0x21, 0x14, 0x84, 0xff], + [0x5d, 0x62, 0x89, 0xff], + [0x21, 0x14, 0x84, 0xff], + [0x21, 0x14, 0x84, 0xff], + [0x99, 0xb0, 0x8e, 0xff], + ], + [0xf2, 0xd7, 0xb0, 0x20, 0xae, 0x2c, 0x6f, 0x97], + ), + ( + [ + [0xb5, 0xcf, 0x9c, 0xff], + [0x83, 0x8c, 0x8b, 0xff], + [0x21, 0x08, 0x6b, 0xff], + [0x83, 0x8c, 0x8b, 0xff], + [0x52, 0x4a, 0x7b, 0xff], + [0x83, 0x8c, 0x8b, 0xff], + [0x83, 0x8c, 0x8b, 0xff], + [0xb5, 0xcf, 0x9c, 0xff], + [0x21, 0x08, 0x6b, 0xff], + [0xb5, 0xcf, 0x9c, 0xff], + [0x83, 0x8c, 0x8b, 0xff], + [0x52, 0x4a, 0x7b, 0xff], + [0xb5, 0xcf, 0x9c, 0xff], + [0x83, 0x8c, 0x8b, 0xff], + [0x52, 0x4a, 0x7b, 0xff], + [0x83, 0x8c, 0x8b, 0xff], + ], + [0x73, 0xb6, 0x4d, 0x20, 0x98, 0x2b, 0xe1, 0xb8], + ), + ( + [ + [0x00, 0x2c, 0xff, 0xff], + [0x94, 0x8d, 0x7b, 0xff], + [0x4a, 0x5c, 0xbd, 0xff], + [0x4a, 0x5c, 0xbd, 0xff], + [0x4a, 0x5c, 0xbd, 0xff], + [0x94, 0x8d, 0x7b, 0xff], + [0x94, 0x8d, 0x7b, 0xff], + [0x94, 0x8d, 0x7b, 0xff], + [0xde, 0xbe, 0x39, 0xff], + [0x94, 0x8d, 0x7b, 0xff], + [0xde, 0xbe, 0x39, 0xff], + [0xde, 0xbe, 0x39, 0xff], + [0xde, 0xbe, 0x39, 0xff], + [0xde, 0xbe, 0x39, 0xff], + [0xde, 0xbe, 0x39, 0xff], + [0x94, 0x8d, 0x7b, 0xff], + ], + [0xe7, 0xdd, 0x7f, 0x01, 0xf9, 0xab, 0x08, 0x80], + ), + ( + [ + [0x6b, 0x24, 0x21, 0xff], + [0x7b, 0x4f, 0x5d, 0xff], + [0x7b, 0x4f, 0x5d, 0xff], + [0x8b, 0x7a, 0x99, 0xff], + [0x7b, 0x4f, 0x5d, 0xff], + [0x7b, 0x4f, 0x5d, 0xff], + [0x6b, 0x24, 0x21, 0xff], + [0x8b, 0x7a, 0x99, 0xff], + [0x9c, 0xa6, 0xd6, 0xff], + [0x6b, 0x24, 0x21, 0xff], + [0x7b, 0x4f, 0x5d, 0xff], + [0x8b, 0x7a, 0x99, 0xff], + [0x6b, 0x24, 0x21, 0xff], + [0x8b, 0x7a, 0x99, 0xff], + [0x7b, 0x4f, 0x5d, 0xff], + [0x9c, 0xa6, 0xd6, 0xff], + ], + [0x3a, 0x9d, 0x24, 0x69, 0xbd, 0x9f, 0xb4, 0x39], + ), +] + +dxt1_rgba = [ + ( + [ + [0x00, 0x00, 0x00, 0x00], + [0x4e, 0xaa, 0x90, 0xff], + [0x4e, 0xaa, 0x90, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x4e, 0xaa, 0x90, 0xff], + [0x29, 0xff, 0xff, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x4e, 0xaa, 0x90, 0xff], + [0x73, 0x55, 0x21, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x4e, 0xaa, 0x90, 0xff], + [0x4e, 0xaa, 0x90, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x4e, 0xaa, 0x90, 0xff], + ], + [0xff, 0x2f, 0xa4, 0x72, 0xeb, 0xb2, 0xbd, 0xbe], + ), + ( + [ + [0xb5, 0xe3, 0x63, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x6b, 0x24, 0x84, 0xff], + [0xb5, 0xe3, 0x63, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0xb5, 0xe3, 0x63, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x6b, 0x24, 0x84, 0xff], + [0x6b, 0x24, 0x84, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0xb5, 0xe3, 0x63, 0xff], + [0x90, 0x83, 0x73, 0xff], + [0xb5, 0xe3, 0x63, 0xff], + ], + [0x30, 0x69, 0x0c, 0xb7, 0x4d, 0xf7, 0x0f, 0x67], + ), + ( + [ + [0x00, 0x00, 0x00, 0x00], + [0xc6, 0x86, 0x8c, 0xff], + [0xc6, 0x86, 0x8c, 0xff], + [0x21, 0x65, 0x42, 0xff], + [0x21, 0x65, 0x42, 0xff], + [0x21, 0x65, 0x42, 0xff], + [0x21, 0x65, 0x42, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x21, 0x65, 0x42, 0xff], + [0xc6, 0x86, 0x8c, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0xc6, 0x86, 0x8c, 0xff], + ], + [0x28, 0x23, 0x31, 0xc4, 0x17, 0xc0, 0xd3, 0x7f], + ), + ( + [ + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0xc6, 0xe3, 0x9c, 0xff], + [0x7b, 0x1c, 0x52, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x00, 0x00, 0x00, 0x00], + [0x7b, 0x1c, 0x52, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0x7b, 0x1c, 0x52, 0xff], + [0xa0, 0x7f, 0x77, 0xff], + [0xc6, 0xe3, 0x9c, 0xff], + [0x00, 0x00, 0x00, 0x00], + [0xa0, 0x7f, 0x77, 0xff], + ], + [0xea, 0x78, 0x13, 0xc7, 0x7f, 0xfc, 0x33, 0xb6], + ), +] + + + +def generate_data_compressed(surface, blocks): + pixels, block = blocks[0] + + stride = surface.nblocksx*surface.block.size + size = surface.nblocksy*stride + + raw = ByteArray(size) + rgba = FloatArray(surface.height*surface.width*4) + + for y in range(0, surface.nblocksx): + for x in range(0, surface.nblocksy): + + offset = (y*surface.nblocksx + x)*surface.block.width + for i in range(0, surface.block.size): + raw[offset + i] = block[i] + + for j in range(0, surface.block.width): + for i in range(0, surface.block.height): + offset = ((y*surface.block.height + j)*surface.width + x*surface.block.width + i)*4 + pixel = pixels[j*surface.block.width + i] + for ch in range(0, 4): + rgba[offset + ch] = float(pixel[ch])/255.0 + + surface.put_tile_raw(0, 0, surface.width, surface.height, raw, stride) + + return rgba + + +def generate_data_simple(surface): + stride = surface.nblocksx*surface.block.size + size = surface.nblocksy*stride + + raw = ByteArray(size) + rgba = FloatArray(surface.height*surface.width*4) + + for i in range(0, size): + raw[i] = random.randint(0, 255) + + surface.put_tile_raw(0, 0, surface.width, surface.height, raw, stride) + + surface.get_tile_rgba(0, 0, surface.width, surface.height, rgba) + + return rgba + + +def generate_data(surface): + width = surface.width + height = surface.height + + if surface.format == PIPE_FORMAT_DXT1_RGB: + return generate_data_compressed(surface, dxt1_rgb_data) + if surface.format == PIPE_FORMAT_DXT1_RGBA: + return generate_data_compressed(surface, dxt1_rgba_data) + if surface.format == PIPE_FORMAT_DXT3_RGBA: + assert 0 + if surface.format == PIPE_FORMAT_DXT5_RGBA: + assert 0 + + return generate_data_simple(surface) + + + diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py new file mode 100644 index 0000000000..d0b657423e --- /dev/null +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -0,0 +1,236 @@ +#!/usr/bin/env python +########################################################################## +# +# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. +# All Rights Reserved. +# +# Permission is hereby granted, free of charge, to any person obtaining a +# copy of this software and associated documentation files (the +# "Software"), to deal in the Software without restriction, including +# without limitation the rights to use, copy, modify, merge, publish, +# distribute, sub license, and/or sell copies of the Software, and to +# permit persons to whom the Software is furnished to do so, subject to +# the following conditions: +# +# The above copyright notice and this permission notice (including the +# next paragraph) shall be included in all copies or substantial portions +# of the Software. +# +# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS +# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF +# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. +# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR +# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, +# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE +# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. +# +########################################################################## + + +from gallium import * +from base import * +from data import generate_data + + +def compare_rgba(width, height, rgba1, rgba2, tol=0.01): + result = True + for y in range(0, height): + for x in range(0, width): + for ch in range(4): + offset = (y*width + x)*4 + ch + v1 = rgba1[offset] + v2 = rgba2[offset] + if abs(v1 - v2) > tol: + sys.stderr.write("x=%u, y=%u, ch=%u differ: %f vs %f\n", + x, y, ch, v1, v2) + result = False + + +class TextureTest(Test): + + def __init__(self, **kargs): + Test.__init__(self) + self.__dict__.update(kargs) + + def run(self): + dev = self.dev + + format = PIPE_FORMAT_A8R8G8B8_UNORM + #format = PIPE_FORMAT_DXT1_RGB + + if not dev.is_format_supported(format, PIPE_TEXTURE): + pass + if not dev.is_format_supported(format, PIPE_SURFACE): + pass + + ctx = dev.context_create() + + width = 256 + height = 256 + + # disabled blending/masking + blend = Blend() + blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE + blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE + blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO + blend.colormask = PIPE_MASK_RGBA + ctx.set_blend(blend) + + # no-op depth/stencil/alpha + depth_stencil_alpha = DepthStencilAlpha() + ctx.set_depth_stencil_alpha(depth_stencil_alpha) + + # rasterizer + rasterizer = Rasterizer() + rasterizer.front_winding = PIPE_WINDING_CW + rasterizer.cull_mode = PIPE_WINDING_NONE + rasterizer.bypass_clipping = 1 + #rasterizer.bypass_vs = 1 + ctx.set_rasterizer(rasterizer) + + # viewport (identity, we setup vertices in wincoords) + viewport = Viewport() + scale = FloatArray(4) + scale[0] = 1.0 + scale[1] = 1.0 + scale[2] = 1.0 + scale[3] = 1.0 + viewport.scale = scale + translate = FloatArray(4) + translate[0] = 0.0 + translate[1] = 0.0 + translate[2] = 0.0 + translate[3] = 0.0 + viewport.translate = translate + ctx.set_viewport(viewport) + + # samplers + sampler = Sampler() + sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE + sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST + sampler.normalized_coords = 1 + ctx.set_sampler(0, sampler) + + # texture + texture = dev.texture_create(format, + width, + height) + ctx.set_sampler_texture(0, texture) + + surface = texture.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE) + + expected_rgba = generate_data(surface) + + cbuf_tex = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, + width, + height) + + # drawing dest + cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE) + fb = Framebuffer() + fb.width = cbuf.width + fb.height = cbuf.height + fb.num_cbufs = 1 + fb.set_cbuf(0, cbuf) + ctx.set_framebuffer(fb) + + # vertex shader + vs = Shader(''' + VERT1.1 + DCL IN[0], POSITION, CONSTANT + DCL IN[1], GENERIC, CONSTANT + DCL OUT[0], POSITION, CONSTANT + DCL OUT[1], GENERIC, CONSTANT + 0:MOV OUT[0], IN[0] + 1:MOV OUT[1], IN[1] + 2:END + ''') + #vs.dump() + ctx.set_vertex_shader(vs) + + # fragment shader + fs = Shader(''' + FRAG1.1 + DCL IN[0], GENERIC[0], PERSPECTIVE + DCL OUT[0], COLOR, CONSTANT + DCL SAMP[0], CONSTANT + 0:TEX OUT[0], IN[0], SAMP[0], 2D + 1:END + ''') + fs.dump() + ctx.set_fragment_shader(fs) + + nverts = 4 + nattrs = 2 + verts = FloatArray(nverts * nattrs * 4) + + x = 0 + y = 0 + w = width + h = height + + verts[ 0] = x # x1 + verts[ 1] = y # y1 + verts[ 2] = 0.0 # z1 + verts[ 3] = 1.0 # w1 + verts[ 4] = 0.0 # s1 + verts[ 5] = 0.0 # t + verts[ 6] = 0.0 + verts[ 7] = 0.0 + verts[ 8] = x + w # x2 + verts[ 9] = y # y2 + verts[10] = 0.0 # z2 + verts[11] = 1.0 # w2 + verts[12] = 1.0 # s2 + verts[13] = 0.0 # t2 + verts[14] = 0.0 + verts[15] = 0.0 + verts[16] = x + w # x3 + verts[17] = y + h # y3 + verts[18] = 0.0 # z3 + verts[19] = 1.0 # w3 + verts[20] = 1.0 # s3 + verts[21] = 1.0 # t3 + verts[22] = 0.0 + verts[23] = 0.0 + verts[24] = x # x4 + verts[25] = y + h # y4 + verts[26] = 0.0 # z4 + verts[27] = 1.0 # w4 + verts[28] = 0.0 # s4 + verts[29] = 1.0 # t4 + verts[30] = 0.0 + verts[31] = 0.0 + + ctx.surface_clear(cbuf, 0x00000000) + + ctx.draw_vertices(PIPE_PRIM_TRIANGLE_FAN, + nverts, + nattrs, + verts) + + ctx.flush() + + rgba = FloatArray(surface.height*surface.width*4) + + cbuf.get_tile_rgba(x, y, w, h, rgba) + + compare_rgba(width, height, rgba, expected_rgba) + + #save_image("texture1.png", surface) + #save_image("texture2.png", cbuf) + + +def main(): + dev = Device(hardware = False) + test = TextureTest(dev = dev) + test.run() + + +if __name__ == '__main__': + main() -- cgit v1.2.3 From 2f5cd68797ee4a159fa0a5fc4d315e4d3f2f644e Mon Sep 17 00:00:00 2001 From: Blair Sadewitz Date: Mon, 14 Jul 2008 08:16:46 -0600 Subject: mesa: also check for __NetBSD__ --- src/mesa/main/execmem.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/execmem.c b/src/mesa/main/execmem.c index aa40b02205..0fe85af93e 100644 --- a/src/mesa/main/execmem.c +++ b/src/mesa/main/execmem.c @@ -36,7 +36,7 @@ -#if defined(__linux__) || defined(__OpenBSD__) +#if defined(__linux__) || defined(__OpenBSD__) || defined(_NetBSD__) /* * Allocate a large block of memory which can hold code then dole it out -- cgit v1.2.3 From 6852ef64f34e5cdcd7f2b3f16f2db74f4d35868e Mon Sep 17 00:00:00 2001 From: Julien Cristau Date: Sun, 13 Jul 2008 17:27:58 +0200 Subject: glu: only export public symbols --- src/glu/sgi/Makefile | 1 + src/glu/sgi/glu.exports | 59 +++++++++++++++++++++++++++++++++++++++++++++++++ 2 files changed, 60 insertions(+) create mode 100644 src/glu/sgi/glu.exports (limited to 'src') diff --git a/src/glu/sgi/Makefile b/src/glu/sgi/Makefile index 2ce6ac0430..cdd267a025 100644 --- a/src/glu/sgi/Makefile +++ b/src/glu/sgi/Makefile @@ -139,6 +139,7 @@ $(TOP)/$(LIB_DIR)/$(GLU_LIB_NAME): $(OBJECTS) $(TOP)/bin/mklib -o $(GLU_LIB) -linker '$(CXX)' \ -major $(GLU_MAJOR) -minor $(GLU_MINOR) -patch $(GLU_TINY) \ -cplusplus $(MKLIB_OPTIONS) -install $(TOP)/$(LIB_DIR) \ + -exports glu.exports \ $(GLU_LIB_DEPS) $(OBJECTS) diff --git a/src/glu/sgi/glu.exports b/src/glu/sgi/glu.exports new file mode 100644 index 0000000000..1d1b6da24f --- /dev/null +++ b/src/glu/sgi/glu.exports @@ -0,0 +1,59 @@ + gluBeginCurve + gluBeginPolygon + gluBeginSurface + gluBeginTrim + gluBuild1DMipmapLevels + gluBuild1DMipmaps + gluBuild2DMipmapLevels + gluBuild2DMipmaps + gluBuild3DMipmapLevels + gluBuild3DMipmaps + gluCheckExtension + gluCylinder + gluDeleteNurbsRenderer + gluDeleteQuadric + gluDeleteTess + gluDisk + gluEndCurve + gluEndPolygon + gluEndSurface + gluEndTrim + gluErrorString + gluGetNurbsProperty + gluGetString + gluGetTessProperty + gluLoadSamplingMatrices + gluLookAt + gluNewNurbsRenderer + gluNewQuadric + gluNewTess + gluNextContour + gluNurbsCallback + gluNurbsCallbackData + gluNurbsCallbackDataEXT + gluNurbsCurve + gluNurbsProperty + gluNurbsSurface + gluOrtho2D + gluPartialDisk + gluPerspective + gluPickMatrix + gluProject + gluPwlCurve + gluQuadricCallback + gluQuadricDrawStyle + gluQuadricNormals + gluQuadricOrientation + gluQuadricTexture + gluScaleImage + gluSphere + gluTessBeginContour + gluTessBeginPolygon + gluTessCallback + gluTessEndContour + gluTessEndPolygon + gluTessNormal + gluTessProperty + gluTessVertex + gluUnProject + gluUnProject4 -- cgit v1.2.3 From 2833d4e715d613384e413fe113d608b8daf0460d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 14 Jul 2008 09:55:33 -0600 Subject: mesa: fix stencil state problem when GL_ATI_separate_stencil wasn't enabled In glStencilFunc/Op/Mask() set both the front and back-face state, unless GL_EXT_stencil_two_side is enabled. Before, we only set the front+back state if GL_ATI_separate_stencil was enabled. Ultimately, we probably should remove GL_EXT_stencil_two_side since it's incompatible with GL 2.x. --- src/mesa/main/stencil.c | 96 +++++++++++++++++++++++++------------------------ 1 file changed, 49 insertions(+), 47 deletions(-) (limited to 'src') diff --git a/src/mesa/main/stencil.c b/src/mesa/main/stencil.c index c1906197de..2a4c38b1f2 100644 --- a/src/mesa/main/stencil.c +++ b/src/mesa/main/stencil.c @@ -207,7 +207,23 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask ) ref = CLAMP( ref, 0, stencilMax ); - if (ctx->Extensions.ATI_separate_stencil) { + if (ctx->Extensions.EXT_stencil_two_side) { + /* only set active face state */ + const GLint face = ctx->Stencil.ActiveFace; + if (ctx->Stencil.Function[face] == func && + ctx->Stencil.ValueMask[face] == mask && + ctx->Stencil.Ref[face] == ref) + return; + FLUSH_VERTICES(ctx, _NEW_STENCIL); + ctx->Stencil.Function[face] = func; + ctx->Stencil.Ref[face] = ref; + ctx->Stencil.ValueMask[face] = mask; + if (ctx->Driver.StencilFuncSeparate) { + ctx->Driver.StencilFuncSeparate(ctx, face ? GL_BACK : GL_FRONT, + func, ref, mask); + } + } + else { /* set both front and back state */ if (ctx->Stencil.Function[0] == func && ctx->Stencil.Function[1] == func && @@ -225,22 +241,6 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask ) func, ref, mask); } } - else { - /* only set active face state */ - const GLint face = ctx->Stencil.ActiveFace; - if (ctx->Stencil.Function[face] == func && - ctx->Stencil.ValueMask[face] == mask && - ctx->Stencil.Ref[face] == ref) - return; - FLUSH_VERTICES(ctx, _NEW_STENCIL); - ctx->Stencil.Function[face] = func; - ctx->Stencil.Ref[face] = ref; - ctx->Stencil.ValueMask[face] = mask; - if (ctx->Driver.StencilFuncSeparate) { - ctx->Driver.StencilFuncSeparate(ctx, face ? GL_BACK : GL_FRONT, - func, ref, mask); - } - } } @@ -261,26 +261,26 @@ _mesa_StencilMask( GLuint mask ) GET_CURRENT_CONTEXT(ctx); ASSERT_OUTSIDE_BEGIN_END(ctx); - if (ctx->Extensions.ATI_separate_stencil) { - /* set both front and back state */ - if (ctx->Stencil.WriteMask[0] == mask && - ctx->Stencil.WriteMask[1] == mask) + if (ctx->Extensions.EXT_stencil_two_side) { + /* only set active face state */ + const GLint face = ctx->Stencil.ActiveFace; + if (ctx->Stencil.WriteMask[face] == mask) return; FLUSH_VERTICES(ctx, _NEW_STENCIL); - ctx->Stencil.WriteMask[0] = ctx->Stencil.WriteMask[1] = mask; + ctx->Stencil.WriteMask[face] = mask; if (ctx->Driver.StencilMaskSeparate) { - ctx->Driver.StencilMaskSeparate(ctx, GL_FRONT_AND_BACK, mask); + ctx->Driver.StencilMaskSeparate(ctx, face ? GL_BACK : GL_FRONT, mask); } } else { - /* only set active face state */ - const GLint face = ctx->Stencil.ActiveFace; - if (ctx->Stencil.WriteMask[face] == mask) + /* set both front and back state */ + if (ctx->Stencil.WriteMask[0] == mask && + ctx->Stencil.WriteMask[1] == mask) return; FLUSH_VERTICES(ctx, _NEW_STENCIL); - ctx->Stencil.WriteMask[face] = mask; + ctx->Stencil.WriteMask[0] = ctx->Stencil.WriteMask[1] = mask; if (ctx->Driver.StencilMaskSeparate) { - ctx->Driver.StencilMaskSeparate(ctx, face ? GL_BACK : GL_FRONT, mask); + ctx->Driver.StencilMaskSeparate(ctx, GL_FRONT_AND_BACK, mask); } } } @@ -319,7 +319,23 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass) return; } - if (ctx->Extensions.ATI_separate_stencil) { + if (ctx->Extensions.EXT_stencil_two_side) { + /* only set active face state */ + const GLint face = ctx->Stencil.ActiveFace; + if (ctx->Stencil.ZFailFunc[face] == zfail && + ctx->Stencil.ZPassFunc[face] == zpass && + ctx->Stencil.FailFunc[face] == fail) + return; + FLUSH_VERTICES(ctx, _NEW_STENCIL); + ctx->Stencil.ZFailFunc[face] = zfail; + ctx->Stencil.ZPassFunc[face] = zpass; + ctx->Stencil.FailFunc[face] = fail; + if (ctx->Driver.StencilOpSeparate) { + ctx->Driver.StencilOpSeparate(ctx, face ? GL_BACK : GL_FRONT, + fail, zfail, zpass); + } + } + else { /* set both front and back state */ if (ctx->Stencil.ZFailFunc[0] == zfail && ctx->Stencil.ZFailFunc[1] == zfail && @@ -337,22 +353,6 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass) fail, zfail, zpass); } } - else { - /* only set active face state */ - const GLint face = ctx->Stencil.ActiveFace; - if (ctx->Stencil.ZFailFunc[face] == zfail && - ctx->Stencil.ZPassFunc[face] == zpass && - ctx->Stencil.FailFunc[face] == fail) - return; - FLUSH_VERTICES(ctx, _NEW_STENCIL); - ctx->Stencil.ZFailFunc[face] = zfail; - ctx->Stencil.ZPassFunc[face] = zpass; - ctx->Stencil.FailFunc[face] = fail; - if (ctx->Driver.StencilOpSeparate) { - ctx->Driver.StencilOpSeparate(ctx, face ? GL_BACK : GL_FRONT, - fail, zfail, zpass); - } - } } @@ -463,12 +463,14 @@ _mesa_StencilFuncSeparate(GLenum face, GLenum func, GLint ref, GLuint mask) FLUSH_VERTICES(ctx, _NEW_STENCIL); - if (face == GL_FRONT || face == GL_FRONT_AND_BACK) { + if (face != GL_BACK) { + /* set front */ ctx->Stencil.Function[0] = func; ctx->Stencil.Ref[0] = ref; ctx->Stencil.ValueMask[0] = mask; } - if (face == GL_BACK || face == GL_FRONT_AND_BACK) { + if (face != GL_FRONT) { + /* set back */ ctx->Stencil.Function[1] = func; ctx->Stencil.Ref[1] = ref; ctx->Stencil.ValueMask[1] = mask; -- cgit v1.2.3 From e7fd3de9277a0c5efe52966908da8e4e7f1995ee Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 14 Jul 2008 12:09:07 -0600 Subject: egl: set EGL_CONFORMANT, EGL_RENDERABLE_TYPE config fields, fix null ptr crashes in MakeCurrent --- src/egl/drivers/xdri/egl_xdri.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c index dfb491958b..2471c0425c 100644 --- a/src/egl/drivers/xdri/egl_xdri.c +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -181,6 +181,10 @@ get_drawable_size(Display *dpy, Drawable d, uint *width, uint *height) static void create_configs(_EGLDisplay *disp, __GLXdisplayPrivate *glx_priv) { + static const EGLint all_apis = (EGL_OPENGL_ES_BIT | + EGL_OPENGL_ES2_BIT | + EGL_OPENVG_BIT | + EGL_OPENGL_BIT); __GLXscreenConfigs *scrn = glx_priv->screenConfigs; const __GLcontextModes *m; int id = 1; @@ -201,6 +205,8 @@ create_configs(_EGLDisplay *disp, __GLXdisplayPrivate *glx_priv) SET_CONFIG_ATTRIB(&config->Base, EGL_STENCIL_SIZE, m->stencilBits); SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_ID, m->visualID); SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_TYPE, m->visualType); + SET_CONFIG_ATTRIB(&config->Base, EGL_CONFORMANT, all_apis); + SET_CONFIG_ATTRIB(&config->Base, EGL_RENDERABLE_TYPE, all_apis); /* XXX only window rendering allowed ATM */ SET_CONFIG_ATTRIB(&config->Base, EGL_SURFACE_TYPE, EGL_WINDOW_BIT); @@ -654,7 +660,9 @@ xdri_eglTerminate(_EGLDriver *drv, EGLDisplay dpy) _eglLog(_EGL_DEBUG, "XDRI: eglTerminate"); _eglLog(_EGL_DEBUG, "XDRI: Closing %s", xdri_drv->dri_driver_name); +#if 0 dlclose(xdri_drv->dri_driver_handle); +#endif xdri_drv->dri_driver_handle = NULL; free((void*) xdri_drv->dri_driver_name); @@ -748,15 +756,16 @@ xdri_eglMakeCurrent(_EGLDriver *drv, EGLDisplay dpy, EGLSurface d, struct xdri_egl_context *xdri_ctx = lookup_context(context); struct xdri_egl_surface *xdri_draw = lookup_surface(d); struct xdri_egl_surface *xdri_read = lookup_surface(r); - __DRIid draw = xdri_draw->driDrawable; - __DRIid read = xdri_read->driDrawable; + __DRIid draw = xdri_draw ? xdri_draw->driDrawable : 0; + __DRIid read = xdri_read ? xdri_read->driDrawable : 0; int scrn = DefaultScreen(disp->Xdpy); if (!_eglMakeCurrent(drv, dpy, d, r, context)) return EGL_FALSE; - if (!xdri_ctx->driContext.bindContext(disp->Xdpy, scrn, draw, read, + if (xdri_ctx && + !xdri_ctx->driContext.bindContext(disp->Xdpy, scrn, draw, read, &xdri_ctx->driContext)) { return EGL_FALSE; } -- cgit v1.2.3 From 51654783ef0aa48560f70cd3944128a94a87d86b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 14 Jul 2008 11:20:58 -0600 Subject: mesa: comments about vectors vs components --- src/mesa/main/config.h | 4 ++-- src/mesa/main/mtypes.h | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/config.h b/src/mesa/main/config.h index 9ff0b708dd..f8109ec755 100644 --- a/src/mesa/main/config.h +++ b/src/mesa/main/config.h @@ -190,8 +190,8 @@ #define MAX_PROGRAM_CALL_DEPTH 8 #define MAX_PROGRAM_TEMPS 128 #define MAX_PROGRAM_ADDRESS_REGS 2 -#define MAX_UNIFORMS 128 -#define MAX_VARYING 8 +#define MAX_UNIFORMS 128 /**< number of float components */ +#define MAX_VARYING 8 /**< number of float[4] vectors */ #define MAX_SAMPLERS 8 /*@}*/ diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 2d6be1983c..88312da7f0 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2519,7 +2519,7 @@ struct gl_constants GLuint MaxRenderbufferSize; /* GL_ARB_vertex_shader */ GLuint MaxVertexTextureImageUnits; - GLuint MaxVarying; + GLuint MaxVarying; /**< Number of float[4] vectors */ }; -- cgit v1.2.3 From ec698034d1a7cc390ce8eda8a28732fa40ff80ec Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 14 Jul 2008 15:58:57 -0600 Subject: mesa: assemble main() after all other functions Before, main() had to come after any functions it called. --- src/mesa/shader/slang/slang_compile.c | 36 +++++++++++++++++++++-------------- 1 file changed, 22 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index ccb04494bf..29c5ca2f4b 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -1788,20 +1788,6 @@ parse_function(slang_parse_ctx * C, slang_output_ctx * O, int definition, *parsed_func_ret = found_func; } - /* assemble the parsed function */ - { - slang_assemble_ctx A; - - A.atoms = C->atoms; - A.space.funcs = O->funs; - A.space.structs = O->structs; - A.space.vars = O->vars; - A.program = O->program; - A.vartable = O->vartable; - A.log = C->L; - - _slang_codegen_function(&A, *parsed_func_ret); - } return GL_TRUE; } @@ -1844,6 +1830,7 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, slang_output_ctx o; GLboolean success; GLuint maxRegs; + slang_function *mainFunc = NULL; if (unit->type == SLANG_UNIT_FRAGMENT_BUILTIN || unit->type == SLANG_UNIT_FRAGMENT_SHADER) { @@ -1871,6 +1858,11 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, { slang_function *func; success = parse_function(C, &o, 1, &func); + if (success && + _mesa_strcmp((char *) func->header.a_name, "main") == 0) { + /* found main() */ + mainFunc = func; + } } break; case EXTERNAL_DECLARATION: @@ -1888,6 +1880,22 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, } C->I++; + if (mainFunc) { + /* assemble (generate code) for main() */ + slang_assemble_ctx A; + + A.atoms = C->atoms; + A.space.funcs = o.funs; + A.space.structs = o.structs; + A.space.vars = o.vars; + A.program = o.program; + A.vartable = o.vartable; + A.log = C->L; + + _slang_codegen_function(&A, mainFunc); + + } + _slang_pop_var_table(o.vartable); _slang_delete_var_table(o.vartable); -- cgit v1.2.3 From 6eb7f763fbbbb7a32640760cd5d122020866fea1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 14 Jul 2008 18:08:52 -0600 Subject: tgsi: fix bug in execution of loops inside of conditionals. Fixes infinite loop bug. --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c index 46949661af..001a4c4b15 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c @@ -2400,7 +2400,8 @@ exec_instruction( /* Restore ContMask, but don't pop */ assert(mach->ContStackTop > 0); mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; - if (mach->LoopMask) { + UPDATE_EXEC_MASK(mach); + if (mach->ExecMask) { /* repeat loop: jump to instruction just past BGNLOOP */ *pc = inst->InstructionExtLabel.Label + 1; } -- cgit v1.2.3 From c60e009a9138a573d2219ee2ad85e3203b09a7ae Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 15 Jul 2008 10:56:30 +0200 Subject: tgsi: Numerical label before an instruction is optional. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 1b283feb4c..6cab475b88 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -740,7 +740,10 @@ static const char *texture_names[TGSI_TEXTURE_COUNT] = "SHADOWRECT" }; -static boolean parse_instruction( struct translate_ctx *ctx ) +static boolean +parse_instruction( + struct translate_ctx *ctx, + boolean has_label ) { uint i; uint saturate = TGSI_SAT_NONE; @@ -767,7 +770,10 @@ static boolean parse_instruction( struct translate_ctx *ctx ) } } if (i == TGSI_OPCODE_LAST) { - report_error( ctx, "Unknown opcode" ); + if (has_label) + report_error( ctx, "Unknown opcode" ); + else + report_error( ctx, "Expected `DCL', `IMM' or a label" ); return FALSE; } info = &opcode_info[i]; @@ -1025,7 +1031,7 @@ static boolean translate( struct translate_ctx *ctx ) } if (parse_label( ctx, &label_val )) { - if (!parse_instruction( ctx )) + if (!parse_instruction( ctx, TRUE )) return FALSE; } else if (str_match_no_case( &ctx->cur, "DCL" )) { @@ -1036,8 +1042,7 @@ static boolean translate( struct translate_ctx *ctx ) if (!parse_immediate( ctx )) return FALSE; } - else { - report_error( ctx, "Expected `DCL', `IMM' or a label" ); + else if (!parse_instruction( ctx, FALSE )) { return FALSE; } } -- cgit v1.2.3 From ead8fcd92a737c7184020768203496a08eb3b7dd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Jul 2008 00:33:54 +0900 Subject: python: Get it to build on windows too. --- src/gallium/state_trackers/python/SConscript | 36 ++++++++++++++++++++++++++-- 1 file changed, 34 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript index 57a3fb2075..854cd83b7c 100644 --- a/src/gallium/state_trackers/python/SConscript +++ b/src/gallium/state_trackers/python/SConscript @@ -1,3 +1,6 @@ +import sys +import os.path + Import('*') if 'python' in env['statetrackers']: @@ -10,8 +13,38 @@ if 'python' in env['statetrackers']: env.Append(SWIGPATH = ['#src/gallium/include', '#src/gallium/include/pipe']) env.Append(SWIGFLAGS = ['-python', '-keyword']) - env.ParseConfig('python-config --cflags --ldflags --libs') + # http://www.scons.org/wiki/PythonExtensions + #env.AppendUnique(CPPATH = [distutils.sysconfig.get_python_inc()]) + #distutils.sysconfig.get_config_vars('SO') + + env['SHLIBPREFIX'] = '' + if platform in ['linux']: + env.ParseConfig('python-config --cflags --ldflags --libs') + + if platform in ['windows']: + python_root = sys.prefix + python_version = '%u%u' % sys.version_info[:2] + python_include = os.path.join(python_root, 'include') + python_libs = os.path.join(python_root, 'libs') + python_lib = os.path.join(python_libs, 'python' + python_version + '.lib') + + env.Append(CPPPATH = [python_include]) + env.Append(LIBPATH = [python_libs]) + env.Append(LIBS = ['python' + python_version + '.lib']) + env.Replace(SHLIBSUFFIX = '.pyd') + + if env['debug']: + # XXX; python25_d.lib is not included in Python for windows, and + # we'll get missing symbols unless we undefine _DEBUG + cppdefines = env['CPPDEFINES'] + cppdefines = [define for define in cppdefines if define != '_DEBUG'] + env.Replace(CPPDEFINES = cppdefines) + + # http://www.swig.org/Doc1.3/Windows.html + env['ENV']['PYTHON_INCLUDE'] = python_include + env['ENV']['PYTHON_LIB'] = python_lib + env.SharedLibrary( target = '_gallium', source = [ @@ -19,6 +52,5 @@ if 'python' in env['statetrackers']: 'st_device.c', 'st_softpipe_winsys.c', ], - SHLIBPREFIX = '', LIBS = softpipe + auxiliaries + env['LIBS'], ) -- cgit v1.2.3 From 2c4349aa3ac9a6d5ee1f7b1aeb5eb0ee1cb54690 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Jul 2008 07:56:42 +0900 Subject: python: Move the python scons code to a separate tool module. --- src/gallium/state_trackers/python/SConscript | 44 +++++----------------------- 1 file changed, 7 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript index 854cd83b7c..687c46bfd7 100644 --- a/src/gallium/state_trackers/python/SConscript +++ b/src/gallium/state_trackers/python/SConscript @@ -7,50 +7,20 @@ if 'python' in env['statetrackers']: env = env.Clone() - env.Append(CPPPATH = '.') + env.Tool('python') env.Tool('swig') env.Append(SWIGPATH = ['#src/gallium/include', '#src/gallium/include/pipe']) env.Append(SWIGFLAGS = ['-python', '-keyword']) - # http://www.scons.org/wiki/PythonExtensions - #env.AppendUnique(CPPATH = [distutils.sysconfig.get_python_inc()]) - #distutils.sysconfig.get_config_vars('SO') - - env['SHLIBPREFIX'] = '' - - if platform in ['linux']: - env.ParseConfig('python-config --cflags --ldflags --libs') - - if platform in ['windows']: - python_root = sys.prefix - python_version = '%u%u' % sys.version_info[:2] - python_include = os.path.join(python_root, 'include') - python_libs = os.path.join(python_root, 'libs') - python_lib = os.path.join(python_libs, 'python' + python_version + '.lib') - - env.Append(CPPPATH = [python_include]) - env.Append(LIBPATH = [python_libs]) - env.Append(LIBS = ['python' + python_version + '.lib']) - env.Replace(SHLIBSUFFIX = '.pyd') - - if env['debug']: - # XXX; python25_d.lib is not included in Python for windows, and - # we'll get missing symbols unless we undefine _DEBUG - cppdefines = env['CPPDEFINES'] - cppdefines = [define for define in cppdefines if define != '_DEBUG'] - env.Replace(CPPDEFINES = cppdefines) - - # http://www.swig.org/Doc1.3/Windows.html - env['ENV']['PYTHON_INCLUDE'] = python_include - env['ENV']['PYTHON_LIB'] = python_lib - + env.Append(CPPPATH = '.') + env.SharedLibrary( - target = '_gallium', - source = [ - 'gallium.i', + target = '_gallium', + source = [ + 'gallium.i', 'st_device.c', 'st_softpipe_winsys.c', - ], + ], LIBS = softpipe + auxiliaries + env['LIBS'], ) -- cgit v1.2.3 From 1d03102b319f7bca2bf76c5d6e3996dc87f03634 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Jul 2008 08:42:16 +0900 Subject: python: Allow hardware support. --- src/gallium/state_trackers/python/SConscript | 13 +++- src/gallium/state_trackers/python/gallium.i | 2 +- src/gallium/state_trackers/python/samples/tri.py | 2 +- src/gallium/state_trackers/python/st_device.c | 6 +- .../state_trackers/python/st_hardpipe_winsys.c | 78 ++++++++++++++++++++++ .../state_trackers/python/st_softpipe_winsys.c | 2 +- src/gallium/state_trackers/python/st_winsys.h | 4 +- src/gallium/state_trackers/python/tests/texture.py | 2 +- 8 files changed, 96 insertions(+), 13 deletions(-) create mode 100644 src/gallium/state_trackers/python/st_hardpipe_winsys.c (limited to 'src') diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript index 687c46bfd7..b284d2020b 100644 --- a/src/gallium/state_trackers/python/SConscript +++ b/src/gallium/state_trackers/python/SConscript @@ -15,12 +15,19 @@ if 'python' in env['statetrackers']: env.Append(CPPPATH = '.') - env.SharedLibrary( - target = '_gallium', + pyst = env.ConvenienceLibrary( + target = 'pyst', source = [ 'gallium.i', 'st_device.c', 'st_softpipe_winsys.c', ], - LIBS = softpipe + auxiliaries + env['LIBS'], + ) + + env.SharedLibrary( + target = '_gallium', + source = [ + 'st_hardpipe_winsys.c', + ], + LIBS = [pyst, softpipe] + auxiliaries + env['LIBS'], ) diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 60f8e50300..967840a063 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -100,7 +100,7 @@ struct st_context { %extend st_device { - st_device(int hardware = 0) { + st_device(int hardware = 1) { return st_device_create(hardware ? TRUE : FALSE); } diff --git a/src/gallium/state_trackers/python/samples/tri.py b/src/gallium/state_trackers/python/samples/tri.py index 7a10905ddb..f915c26215 100644 --- a/src/gallium/state_trackers/python/samples/tri.py +++ b/src/gallium/state_trackers/python/samples/tri.py @@ -185,7 +185,7 @@ def test(dev): def main(): - dev = Device(hardware = False) + dev = Device() test(dev) diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index 25b5a4fa8c..d88d2de3c6 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -82,12 +82,10 @@ st_device_create_from_st_winsys(const struct st_winsys *st_ws) struct st_device * st_device_create(boolean hardware) { -#if 0 if(hardware) - return st_device_create_from_st_winsys(&st_hardware_winsys); + return st_device_create_from_st_winsys(&st_hardpipe_winsys); else -#endif - return st_device_create_from_st_winsys(&st_software_winsys); + return st_device_create_from_st_winsys(&st_softpipe_winsys); } diff --git a/src/gallium/state_trackers/python/st_hardpipe_winsys.c b/src/gallium/state_trackers/python/st_hardpipe_winsys.c new file mode 100644 index 0000000000..33b75637b3 --- /dev/null +++ b/src/gallium/state_trackers/python/st_hardpipe_winsys.c @@ -0,0 +1,78 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ + +/** + * @file + * Stub for hardware pipe driver support. + */ + + +#include "pipe/p_compiler.h" + +#include "st_winsys.h" + + +extern void init_gallium(void); + +void (*force_init_gallium_linkage)(void) = &init_gallium; + + +static void +st_hardpipe_screen_destroy(struct pipe_screen *screen) +{ + st_softpipe_winsys.screen_destroy(screen); +} + + +static struct pipe_screen * +st_hardpipe_screen_create(void) +{ + return st_softpipe_winsys.screen_create(); +} + + +static void +st_hardpipe_context_destroy(struct pipe_context *pipe) +{ + st_softpipe_winsys.context_destroy(pipe); +} + + +static struct pipe_context * +st_hardpipe_context_create(struct pipe_screen *screen) +{ + return st_softpipe_winsys.context_create(screen); +} + + +const struct st_winsys st_hardpipe_winsys = { + &st_hardpipe_screen_create, + &st_hardpipe_screen_destroy, + &st_hardpipe_context_create, + &st_hardpipe_context_destroy +}; diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c index 964d60de1d..1fda70ca00 100644 --- a/src/gallium/state_trackers/python/st_softpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c @@ -313,7 +313,7 @@ st_softpipe_context_create(struct pipe_screen *screen) } -const struct st_winsys st_software_winsys = { +const struct st_winsys st_softpipe_winsys = { &st_softpipe_screen_create, &st_softpipe_screen_destroy, &st_softpipe_context_create, diff --git a/src/gallium/state_trackers/python/st_winsys.h b/src/gallium/state_trackers/python/st_winsys.h index 992fc9ab4b..43db8b6bff 100644 --- a/src/gallium/state_trackers/python/st_winsys.h +++ b/src/gallium/state_trackers/python/st_winsys.h @@ -50,9 +50,9 @@ struct st_winsys }; -extern const struct st_winsys st_software_winsys; +extern const struct st_winsys st_softpipe_winsys; -extern const struct st_winsys st_hardware_winsys; +extern const struct st_winsys st_hardpipe_winsys; #endif /* ST_WINSYS_H_ */ diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index d0b657423e..ad11e8dcc9 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -227,7 +227,7 @@ class TextureTest(Test): def main(): - dev = Device(hardware = False) + dev = Device() test = TextureTest(dev = dev) test.run() -- cgit v1.2.3 From d5ed158dc87669f62bb7d3fb65c23fb2a465442b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Jul 2008 17:13:37 +0900 Subject: python: Get object ownership done correctly. --- src/gallium/state_trackers/python/gallium.i | 44 +++++++++++++++++----- src/gallium/state_trackers/python/st_device.c | 1 + .../state_trackers/python/st_hardpipe_winsys.c | 2 +- 3 files changed, 37 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 967840a063..2f1d8b22fb 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -66,6 +66,7 @@ %rename(Surface) pipe_surface; %rename(Buffer) pipe_buffer; + %rename(BlendColor) pipe_blend_color; %rename(Blend) pipe_blend_state; %rename(Clip) pipe_clip_state; @@ -82,15 +83,30 @@ %rename(VertexElement) pipe_vertex_element; %rename(Viewport) pipe_viewport_state; +%nodefaultctor st_device; +%nodefaultctor st_context; +%nodefaultctor pipe_texture; +%nodefaultctor pipe_surface; +%nodefaultctor pipe_buffer; + +%nodefaultdtor st_device; +%nodefaultdtor st_context; +%nodefaultdtor pipe_texture; +%nodefaultdtor pipe_surface; +%nodefaultdtor pipe_buffer; + +%ignore pipe_texture::screen; + +%ignore pipe_surface::winsys; +%immutable pipe_surface::texture; +%immutable pipe_surface::buffer; + %include "p_format.i"; %include "pipe/p_defines.h"; %include "pipe/p_state.h"; %include "pipe/p_shader_tokens.h"; -%nodefaultctor; -%nodefaultdtor; - struct st_device { }; @@ -98,6 +114,10 @@ struct st_context { }; +%newobject st_device::texture_create; +%newobject st_device::context_create; +%newobject st_device::buffer_create; + %extend st_device { st_device(int hardware = 1) { @@ -157,6 +177,7 @@ struct st_context { unsigned usage = 0 ) { struct pipe_texture templat; + struct pipe_texture *texture; memset(&templat, 0, sizeof(templat)); templat.format = format; pf_get_block(templat.format, &templat.block); @@ -166,7 +187,9 @@ struct st_context { templat.last_level = last_level; templat.target = target; templat.tex_usage = usage; - return $self->screen->texture_create($self->screen, &templat); + texture = $self->screen->texture_create($self->screen, &templat); + fprintf(stderr, "creating texture %p\n", texture); + return texture; } struct pipe_buffer * @@ -335,10 +358,6 @@ error1: ; } - void draw_quad(float x0, float y0, float x1, float y1, float z = 0.0f) { - util_draw_texquad($self->pipe, x0, y0, x1, y1, z); - } - void flush(void) { struct pipe_fence_handle *fence = NULL; @@ -375,10 +394,13 @@ error1: }; +%newobject pipe_texture::get_surface; + %extend pipe_texture { ~pipe_texture() { struct pipe_texture *ptr = $self; + fprintf(stderr, "destroying texture %p\n", $self); pipe_texture_reference(&ptr, NULL); } @@ -387,7 +409,10 @@ error1: get_surface(unsigned face=0, unsigned level=0, unsigned zslice=0, unsigned usage=0 ) { struct pipe_screen *screen = $self->screen; - return screen->get_tex_surface(screen, $self, face, level, zslice, usage); + struct pipe_surface *surface; + surface = screen->get_tex_surface(screen, $self, face, level, zslice, usage); + fprintf(stderr, "creating surface %p\n", surface); + return surface; } }; @@ -397,6 +422,7 @@ error1: ~pipe_surface() { struct pipe_surface *ptr = $self; + fprintf(stderr, "destroying surface %p\n", $self); pipe_surface_reference(&ptr, NULL); } diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index d88d2de3c6..9e625ae2e0 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -70,6 +70,7 @@ st_device_create_from_st_winsys(const struct st_winsys *st_ws) if(!st_dev) return NULL; + st_dev->refcount = 1; st_dev->st_ws = st_ws; st_dev->screen = st_ws->screen_create(); diff --git a/src/gallium/state_trackers/python/st_hardpipe_winsys.c b/src/gallium/state_trackers/python/st_hardpipe_winsys.c index 33b75637b3..1e04998232 100644 --- a/src/gallium/state_trackers/python/st_hardpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_hardpipe_winsys.c @@ -37,8 +37,8 @@ #include "st_winsys.h" +/* XXX: Force init_gallium symbol to be linked */ extern void init_gallium(void); - void (*force_init_gallium_linkage)(void) = &init_gallium; -- cgit v1.2.3 From ee470020e17bc4999f3540fbad49fe645a4b914e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Jul 2008 17:14:07 +0900 Subject: python: Request/respect the texture & buffer usage flags in the examples. --- src/gallium/state_trackers/python/samples/tri.py | 81 +++++++++++++++++----- src/gallium/state_trackers/python/tests/base.py | 35 +++++++++- src/gallium/state_trackers/python/tests/texture.py | 46 ++++++------ 3 files changed, 121 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/samples/tri.py b/src/gallium/state_trackers/python/samples/tri.py index f915c26215..e5044a0245 100644 --- a/src/gallium/state_trackers/python/samples/tri.py +++ b/src/gallium/state_trackers/python/samples/tri.py @@ -30,7 +30,7 @@ from gallium import * -def save_image(filename, surface): +def make_image(surface): pixels = FloatArray(surface.height*surface.width*4) surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) @@ -45,14 +45,38 @@ def save_image(filename, surface): offset = (y*surface.width + x)*4 r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] outpixels[x, y] = r, g, b + return outimage + +def save_image(filename, surface): + outimage = make_image(surface) outimage.save(filename, "PNG") +def show_image(surface): + outimage = make_image(surface) + + import Tkinter as tk + from PIL import Image, ImageTk + root = tk.Tk() + + root.title('background image') + + image1 = ImageTk.PhotoImage(outimage) + w = image1.width() + h = image1.height() + x = 100 + y = 100 + root.geometry("%dx%d+%d+%d" % (w, h, x, y)) + panel1 = tk.Label(root, image=image1) + panel1.pack(side='top', fill='both', expand='yes') + panel1.image = image1 + root.mainloop() + def test(dev): ctx = dev.context_create() - width = 256 - height = 256 + width = 255 + height = 255 # disabled blending/masking blend = Blend() @@ -72,6 +96,7 @@ def test(dev): rasterizer.front_winding = PIPE_WINDING_CW rasterizer.cull_mode = PIPE_WINDING_NONE rasterizer.bypass_clipping = 1 + rasterizer.scissor = 1 #rasterizer.bypass_vs = 1 ctx.set_rasterizer(rasterizer) @@ -102,21 +127,41 @@ def test(dev): sampler.normalized_coords = 1 ctx.set_sampler(0, sampler) - # texture - texture = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, - width, height, - usage=PIPE_TEXTURE_USAGE_RENDER_TARGET) - ctx.set_sampler_texture(0, texture) - - # drawing dest - surface = texture.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE) + # scissor + scissor = Scissor() + scissor.minx = 0 + scissor.miny = 0 + scissor.maxx = width + scissor.maxy = height + ctx.set_scissor(scissor) + + clip = Clip() + clip.nr = 0 + ctx.set_clip(clip) + + # framebuffer + cbuf = dev.texture_create(PIPE_FORMAT_X8R8G8B8_UNORM, + width, height, + usage=PIPE_TEXTURE_USAGE_DISPLAY_TARGET) + zbuf = dev.texture_create(PIPE_FORMAT_X8Z24_UNORM, + width, height, + usage=PIPE_TEXTURE_USAGE_DEPTH_STENCIL) + _cbuf = cbuf.get_surface(usage = PIPE_BUFFER_USAGE_GPU_READ|PIPE_BUFFER_USAGE_GPU_WRITE) + _zsbuf = zbuf.get_surface(usage = PIPE_BUFFER_USAGE_GPU_READ|PIPE_BUFFER_USAGE_GPU_WRITE) fb = Framebuffer() - fb.width = surface.width - fb.height = surface.height + fb.width = width + fb.height = height fb.num_cbufs = 1 - fb.set_cbuf(0, surface) + fb.set_cbuf(0, _cbuf) + fb.set_zsbuf(_zsbuf) ctx.set_framebuffer(fb) - + _cbuf.clear_value = 0x00000000 + _zsbuf.clear_value = 0x00ffffff + ctx.surface_clear(_cbuf, _cbuf.clear_value) + ctx.surface_clear(_zsbuf, _zsbuf.clear_value) + del _cbuf + del _zsbuf + # vertex shader vs = Shader(''' VERT1.1 @@ -171,16 +216,14 @@ def test(dev): verts[22] = 1.0 # b3 verts[23] = 1.0 # a3 - ctx.surface_clear(surface, 0x00000000) - ctx.draw_vertices(PIPE_PRIM_TRIANGLES, nverts, nattrs, verts) ctx.flush() - - save_image("tri.png", surface) + + show_image(cbuf.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE)) diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py index 5778ad4382..185a59996e 100644 --- a/src/gallium/state_trackers/python/tests/base.py +++ b/src/gallium/state_trackers/python/tests/base.py @@ -37,7 +37,7 @@ for name, value in globals().items(): formats[value] = name -def save_image(filename, surface): +def make_image(surface): pixels = FloatArray(surface.height*surface.width*4) surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) @@ -52,8 +52,41 @@ def save_image(filename, surface): offset = (y*surface.width + x)*4 r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] outpixels[x, y] = r, g, b + return outimage + +def save_image(filename, surface): + outimage = make_image(surface) outimage.save(filename, "PNG") +def show_image(*surfaces): + import Tkinter as tk + from PIL import Image, ImageTk + + root = tk.Tk() + + x = 64 + y = 64 + + for i in range(len(surfaces)): + surface = surfaces[i] + outimage = make_image(surface) + + if i: + window = tk.Toplevel(root) + else: + window = root + window.title('Image %u' % (i+1)) + image1 = ImageTk.PhotoImage(outimage) + w = image1.width() + h = image1.height() + window.geometry("%dx%d+%d+%d" % (w, h, x, y)) + panel1 = tk.Label(window, image=image1) + panel1.pack(side='top', fill='both', expand='yes') + panel1.image = image1 + x += w + 2 + + root.mainloop() + class Test: diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index ad11e8dcc9..84ceebb169 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -27,13 +27,14 @@ ########################################################################## +import sys from gallium import * from base import * from data import generate_data def compare_rgba(width, height, rgba1, rgba2, tol=0.01): - result = True + errors = 0 for y in range(0, height): for x in range(0, width): for ch in range(4): @@ -41,9 +42,14 @@ def compare_rgba(width, height, rgba1, rgba2, tol=0.01): v1 = rgba1[offset] v2 = rgba2[offset] if abs(v1 - v2) > tol: - sys.stderr.write("x=%u, y=%u, ch=%u differ: %f vs %f\n", - x, y, ch, v1, v2) - result = False + if errors == 0: + sys.stderr.write("x=%u, y=%u, ch=%u differ: %f vs %f\n" % (x, y, ch, v1, v2)) + if errors == 1: + sys.stderr.write("...\n") + errors += 1 + if errors: + sys.stderr.write("%u out of %u pixels differ\n" % (errors/4, height*width)) + return errors == 0 class TextureTest(Test): @@ -65,8 +71,8 @@ class TextureTest(Test): ctx = dev.context_create() - width = 256 - height = 256 + width = 64 + height = 64 # disabled blending/masking blend = Blend() @@ -122,21 +128,21 @@ class TextureTest(Test): height) ctx.set_sampler_texture(0, texture) - surface = texture.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE) - - expected_rgba = generate_data(surface) + expected_rgba = generate_data(texture.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE)) cbuf_tex = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, width, - height) + height, + usage = PIPE_TEXTURE_USAGE_RENDER_TARGET) # drawing dest - cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE) + cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE|PIPE_BUFFER_USAGE_GPU_READ) fb = Framebuffer() - fb.width = cbuf.width - fb.height = cbuf.height + fb.width = width + fb.height = height fb.num_cbufs = 1 fb.set_cbuf(0, cbuf) + ctx.surface_clear(cbuf, 0x00000000) ctx.set_framebuffer(fb) # vertex shader @@ -164,7 +170,7 @@ class TextureTest(Test): ''') fs.dump() ctx.set_fragment_shader(fs) - + nverts = 4 nattrs = 2 verts = FloatArray(nverts * nattrs * 4) @@ -207,23 +213,21 @@ class TextureTest(Test): verts[30] = 0.0 verts[31] = 0.0 - ctx.surface_clear(cbuf, 0x00000000) - ctx.draw_vertices(PIPE_PRIM_TRIANGLE_FAN, nverts, nattrs, verts) ctx.flush() - - rgba = FloatArray(surface.height*surface.width*4) - cbuf.get_tile_rgba(x, y, w, h, rgba) + rgba = FloatArray(height*width*4) + + cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ).get_tile_rgba(x, y, w, h, rgba) compare_rgba(width, height, rgba, expected_rgba) - #save_image("texture1.png", surface) - #save_image("texture2.png", cbuf) + show_image(texture.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ), + cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ)) def main(): -- cgit v1.2.3 From 3392bcaaa823bd791e8d7e4c5a7ce013831899bb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 15 Jul 2008 17:57:48 +0900 Subject: python: Cleanup. --- src/gallium/state_trackers/python/samples/tri.py | 8 ------ src/gallium/state_trackers/python/tests/base.py | 31 +++++++++++----------- src/gallium/state_trackers/python/tests/texture.py | 14 +++++----- 3 files changed, 23 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/samples/tri.py b/src/gallium/state_trackers/python/samples/tri.py index e5044a0245..3665922929 100644 --- a/src/gallium/state_trackers/python/samples/tri.py +++ b/src/gallium/state_trackers/python/samples/tri.py @@ -143,24 +143,16 @@ def test(dev): cbuf = dev.texture_create(PIPE_FORMAT_X8R8G8B8_UNORM, width, height, usage=PIPE_TEXTURE_USAGE_DISPLAY_TARGET) - zbuf = dev.texture_create(PIPE_FORMAT_X8Z24_UNORM, - width, height, - usage=PIPE_TEXTURE_USAGE_DEPTH_STENCIL) _cbuf = cbuf.get_surface(usage = PIPE_BUFFER_USAGE_GPU_READ|PIPE_BUFFER_USAGE_GPU_WRITE) - _zsbuf = zbuf.get_surface(usage = PIPE_BUFFER_USAGE_GPU_READ|PIPE_BUFFER_USAGE_GPU_WRITE) fb = Framebuffer() fb.width = width fb.height = height fb.num_cbufs = 1 fb.set_cbuf(0, _cbuf) - fb.set_zsbuf(_zsbuf) ctx.set_framebuffer(fb) _cbuf.clear_value = 0x00000000 - _zsbuf.clear_value = 0x00ffffff ctx.surface_clear(_cbuf, _cbuf.clear_value) - ctx.surface_clear(_zsbuf, _zsbuf.clear_value) del _cbuf - del _zsbuf # vertex shader vs = Shader(''' diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py index 185a59996e..d5d6114a45 100644 --- a/src/gallium/state_trackers/python/tests/base.py +++ b/src/gallium/state_trackers/python/tests/base.py @@ -37,28 +37,25 @@ for name, value in globals().items(): formats[value] = name -def make_image(surface): - pixels = FloatArray(surface.height*surface.width*4) - surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) - +def make_image(width, height, rgba): import Image outimage = Image.new( mode='RGB', - size=(surface.width, surface.height), + size=(width, height), color=(0,0,0)) outpixels = outimage.load() - for y in range(0, surface.height): - for x in range(0, surface.width): - offset = (y*surface.width + x)*4 - r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] + for y in range(0, height): + for x in range(0, width): + offset = (y*width + x)*4 + r, g, b, a = [int(rgba[offset + ch]*255) for ch in range(4)] outpixels[x, y] = r, g, b return outimage -def save_image(filename, surface): - outimage = make_image(surface) +def save_image(width, height, rgba, filename): + outimage = make_image(width, height, rgba) outimage.save(filename, "PNG") -def show_image(*surfaces): +def show_image(width, height, **rgbas): import Tkinter as tk from PIL import Image, ImageTk @@ -67,15 +64,17 @@ def show_image(*surfaces): x = 64 y = 64 - for i in range(len(surfaces)): - surface = surfaces[i] - outimage = make_image(surface) + labels = rgbas.keys() + labels.sort() + for i in range(len(labels)): + label = labels[i] + outimage = make_image(width, height, rgbas[label]) if i: window = tk.Toplevel(root) else: window = root - window.title('Image %u' % (i+1)) + window.title(label) image1 = ImageTk.PhotoImage(outimage) w = image1.width() h = image1.height() diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index 84ceebb169..6a5ffb38b9 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -61,8 +61,8 @@ class TextureTest(Test): def run(self): dev = self.dev - format = PIPE_FORMAT_A8R8G8B8_UNORM - #format = PIPE_FORMAT_DXT1_RGB + #format = PIPE_FORMAT_A8R8G8B8_UNORM + format = PIPE_FORMAT_DXT1_RGB if not dev.is_format_supported(format, PIPE_TEXTURE): pass @@ -130,20 +130,21 @@ class TextureTest(Test): expected_rgba = generate_data(texture.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE)) + # framebuffer cbuf_tex = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, width, height, usage = PIPE_TEXTURE_USAGE_RENDER_TARGET) - # drawing dest cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE|PIPE_BUFFER_USAGE_GPU_READ) fb = Framebuffer() fb.width = width fb.height = height fb.num_cbufs = 1 fb.set_cbuf(0, cbuf) - ctx.surface_clear(cbuf, 0x00000000) ctx.set_framebuffer(fb) + ctx.surface_clear(cbuf, 0x00000000) + del fb # vertex shader vs = Shader(''' @@ -220,14 +221,15 @@ class TextureTest(Test): ctx.flush() + del ctx + rgba = FloatArray(height*width*4) cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ).get_tile_rgba(x, y, w, h, rgba) compare_rgba(width, height, rgba, expected_rgba) - show_image(texture.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ), - cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ)) + show_image(width, height, Result=rgba, Expected=expected_rgba) def main(): -- cgit v1.2.3 From 090e212c0c5e54156c3c33f7eecdfe01398a7222 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 15 Jul 2008 11:15:27 +0200 Subject: mesa: Silence compiler warnings on Windows. --- src/mesa/main/texformat.c | 4 ++-- src/mesa/main/texrender.c | 8 ++++---- src/mesa/main/texstore.c | 2 +- 3 files changed, 7 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c index 11e7755960..60f36c4a87 100644 --- a/src/mesa/main/texformat.c +++ b/src/mesa/main/texformat.c @@ -55,10 +55,10 @@ nonlinear_to_linear(GLubyte cs8) for (i = 0; i < 256; i++) { const GLfloat cs = UBYTE_TO_FLOAT(i); if (cs <= 0.04045) { - table[i] = cs / 12.92; + table[i] = cs / 12.92f; } else { - table[i] = _mesa_pow((cs + 0.055) / 1.055, 2.4); + table[i] = (GLfloat) _mesa_pow((cs + 0.055) / 1.055, 2.4); } } tableReady = GL_TRUE; diff --git a/src/mesa/main/texrender.c b/src/mesa/main/texrender.c index 8d5468aff0..163bda4501 100644 --- a/src/mesa/main/texrender.c +++ b/src/mesa/main/texrender.c @@ -159,7 +159,7 @@ texture_put_row(GLcontext *ctx, struct gl_renderbuffer *rb, GLuint count, const GLuint *zValues = (const GLuint *) values; for (i = 0; i < count; i++) { if (!mask || mask[i]) { - GLfloat flt = (zValues[i] >> 8) * (1.0 / 0xffffff); + GLfloat flt = (GLfloat) ((zValues[i] >> 8) * (1.0 / 0xffffff)); trb->Store(trb->TexImage, x + i, y, z, &flt); } } @@ -199,7 +199,7 @@ texture_put_mono_row(GLcontext *ctx, struct gl_renderbuffer *rb, GLuint count, } else if (rb->DataType == GL_UNSIGNED_INT_24_8_EXT) { const GLuint zValue = *((const GLuint *) value); - const GLfloat flt = (zValue >> 8) * (1.0 / 0xffffff); + const GLfloat flt = (GLfloat) ((zValue >> 8) * (1.0 / 0xffffff)); for (i = 0; i < count; i++) { if (!mask || mask[i]) { trb->Store(trb->TexImage, x + i, y, z, &flt); @@ -244,7 +244,7 @@ texture_put_values(GLcontext *ctx, struct gl_renderbuffer *rb, GLuint count, const GLuint *zValues = (const GLuint *) values; for (i = 0; i < count; i++) { if (!mask || mask[i]) { - GLfloat flt = (zValues[i] >> 8) * (1.0 / 0xffffff); + GLfloat flt = (GLfloat) ((zValues[i] >> 8) * (1.0 / 0xffffff)); trb->Store(trb->TexImage, x[i], y[i] + trb->Yoffset, z, &flt); } } @@ -283,7 +283,7 @@ texture_put_mono_values(GLcontext *ctx, struct gl_renderbuffer *rb, } else if (rb->DataType == GL_UNSIGNED_INT_24_8_EXT) { const GLuint zValue = *((const GLuint *) value); - const GLfloat flt = (zValue >> 8) * (1.0 / 0xffffff); + const GLfloat flt = (GLfloat) ((zValue >> 8) * (1.0 / 0xffffff)); for (i = 0; i < count; i++) { if (!mask || mask[i]) { trb->Store(trb->TexImage, x[i], y[i] + trb->Yoffset, z, &flt); diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index 113eef69f4..edb1ae965d 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -2377,7 +2377,7 @@ _mesa_texstore_z24_s8(TEXSTORE_PARAMS) _mesa_unpack_depth_span(ctx, srcWidth, GL_UNSIGNED_INT_24_8_EXT, /* dst type */ dstRow, /* dst addr */ - depthScale, + (GLuint) depthScale, srcType, src, srcPacking); /* get the 8-bit stencil values */ _mesa_unpack_stencil_span(ctx, srcWidth, -- cgit v1.2.3 From 6c534b830c6f5427c391c5225c34561141c201ba Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 15 Jul 2008 11:26:38 +0200 Subject: st: Silence compiler warnings on Windows. --- src/mesa/state_tracker/st_atom_framebuffer.c | 2 +- src/mesa/state_tracker/st_atom_rasterizer.c | 2 +- src/mesa/state_tracker/st_atom_viewport.c | 6 +++--- src/mesa/state_tracker/st_cb_accum.c | 2 +- src/mesa/state_tracker/st_cb_clear.c | 14 +++++++------- src/mesa/state_tracker/st_cb_drawpixels.c | 12 ++++++------ src/mesa/state_tracker/st_cb_rasterpos.c | 2 +- src/mesa/state_tracker/st_cb_texture.c | 8 ++++---- 8 files changed, 24 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c index d46c3ee16c..80df3b0506 100644 --- a/src/mesa/state_tracker/st_atom_framebuffer.c +++ b/src/mesa/state_tracker/st_atom_framebuffer.c @@ -59,7 +59,7 @@ update_renderbuffer_surface(struct st_context *st, strb->surface->texture != texture || strb->surface->width != rtt_width || strb->surface->height != rtt_height) { - int level; + GLuint level; /* find matching mipmap level size */ for (level = 0; level <= texture->last_level; level++) { if (texture->width[level] == rtt_width && diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index ff40bb4312..9f4e1c1c69 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -188,7 +188,7 @@ static void update_raster_state( struct st_context *st ) { GLfloat mrd = (ctx->DrawBuffer ? ctx->DrawBuffer->_MRD : - 1.0); + 1.0f); raster->offset_units = ctx->Polygon.OffsetFactor * mrd; raster->offset_scale = (ctx->Polygon.OffsetUnits * mrd * diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index b105909e96..8b9f1abda4 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -62,9 +62,9 @@ update_viewport( struct st_context *st ) GLfloat x = (GLfloat)ctx->Viewport.X; GLfloat y = (GLfloat)ctx->Viewport.Y; GLfloat z = ctx->Viewport.Near; - GLfloat half_width = (GLfloat)ctx->Viewport.Width / 2.0; - GLfloat half_height = (GLfloat)ctx->Viewport.Height / 2.0; - GLfloat half_depth = (GLfloat)(ctx->Viewport.Far - ctx->Viewport.Near) / 2.0; + GLfloat half_width = (GLfloat)ctx->Viewport.Width / 2.0f; + GLfloat half_height = (GLfloat)ctx->Viewport.Height / 2.0f; + GLfloat half_depth = (GLfloat)(ctx->Viewport.Far - ctx->Viewport.Near) / 2.0f; st->state.viewport.scale[0] = half_width; st->state.viewport.scale[1] = half_height * yScale; diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index c0e8c6bf33..a992e08ff6 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -290,7 +290,7 @@ accum_return(GLcontext *ctx, GLfloat value, for (ch = 0; ch < 4; ch++) { if (colormask[ch]) { GLfloat val = abuf[i * 4 + ch] * value; - abuf[i * 4 + ch] = CLAMP(val, 0.0, 1.0); + abuf[i * 4 + ch] = CLAMP(val, 0.0f, 1.0f); } else { abuf[i * 4 + ch] = cbuf[i * 4 + ch]; diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c index cdfcdcee72..e475f022d3 100644 --- a/src/mesa/state_tracker/st_cb_clear.c +++ b/src/mesa/state_tracker/st_cb_clear.c @@ -203,17 +203,17 @@ clear_with_quad(GLcontext *ctx, GLboolean color, GLboolean depth, GLboolean stencil) { struct st_context *st = ctx->st; - const GLfloat x0 = ctx->DrawBuffer->_Xmin; - const GLfloat x1 = ctx->DrawBuffer->_Xmax; + const GLfloat x0 = (GLfloat) ctx->DrawBuffer->_Xmin; + const GLfloat x1 = (GLfloat) ctx->DrawBuffer->_Xmax; GLfloat y0, y1; if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) { - y0 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax; - y1 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin; + y0 = (GLfloat) (ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax); + y1 = (GLfloat) (ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin); } else { - y0 = ctx->DrawBuffer->_Ymin; - y1 = ctx->DrawBuffer->_Ymax; + y0 = (GLfloat) ctx->DrawBuffer->_Ymin; + y1 = (GLfloat) ctx->DrawBuffer->_Ymax; } /* @@ -286,7 +286,7 @@ clear_with_quad(GLcontext *ctx, cso_set_vertex_shader_handle(st->cso_context, st->clear.vs); /* draw quad matching scissor rect (XXX verify coord round-off) */ - draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor); + draw_quad(ctx, x0, y0, x1, y1, (GLfloat) ctx->Depth.Clear, ctx->Color.ClearColor); /* Restore pipe state */ cso_restore_blend(st->cso_context); diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 6b0137dcf8..2ebfcaf82b 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -423,8 +423,8 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z, /* setup vertex data */ { const struct gl_framebuffer *fb = st->ctx->DrawBuffer; - const GLfloat fb_width = fb->Width; - const GLfloat fb_height = fb->Height; + const GLfloat fb_width = (GLfloat) fb->Width; + const GLfloat fb_height = (GLfloat) fb->Height; const GLfloat clip_x0 = x0 / fb_width * 2.0f - 1.0f; const GLfloat clip_y0 = y0 / fb_height * 2.0f - 1.0f; const GLfloat clip_x1 = x1 / fb_width * 2.0f - 1.0f; @@ -571,8 +571,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, /* viewport state: viewport matching window dims */ { - const float width = ctx->DrawBuffer->Width; - const float height = ctx->DrawBuffer->Height; + const float width = (float) ctx->DrawBuffer->Width; + const float height = (float) ctx->DrawBuffer->Height; struct pipe_viewport_state vp; vp.scale[0] = 0.5f * width; vp.scale[1] = -0.5f * height; @@ -600,9 +600,9 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z, * Recall that these coords are transformed by the current * vertex shader and viewport transformation. */ - x0 = x; + x0 = (GLfloat) x; x1 = x + width * ctx->Pixel.ZoomX; - y0 = y; + y0 = (GLfloat) y; y1 = y + height * ctx->Pixel.ZoomY; //if(!color) draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex); diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c index 3cb7b68bea..3b30c2a61b 100644 --- a/src/mesa/state_tracker/st_cb_rasterpos.c +++ b/src/mesa/state_tracker/st_cb_rasterpos.c @@ -135,7 +135,7 @@ rastpos_point(struct draw_stage *stage, struct prim_header *prim) struct rastpos_stage *rs = rastpos_stage(stage); GLcontext *ctx = rs->ctx; struct st_context *st = ctx->st; - const GLfloat height = ctx->DrawBuffer->Height; + const GLfloat height = (GLfloat) ctx->DrawBuffer->Height; const GLuint *outputMapping = st->vertex_result_to_slot; const GLfloat *pos; GLuint i; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 73f8b8f788..de782e8232 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -285,7 +285,7 @@ guess_and_alloc_texture(struct st_context *st, assert(!stObj->pt); if (stObj->pt && - stImage->level > stObj->base.BaseLevel && + (GLint) stImage->level > stObj->base.BaseLevel && (stImage->base.Width == 1 || (stObj->base.Target != GL_TEXTURE_1D && stImage->base.Height == 1) || @@ -296,7 +296,7 @@ guess_and_alloc_texture(struct st_context *st, /* If this image disrespects BaseLevel, allocate from level zero. * Usually BaseLevel == 0, so it's unlikely to happen. */ - if (stImage->level < stObj->base.BaseLevel) + if ((GLint) stImage->level < stObj->base.BaseLevel) firstLevel = 0; else firstLevel = stObj->base.BaseLevel; @@ -810,7 +810,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, texImage->Height, format, type); GLuint depth; - int i; + GLuint i; GLubyte *dest; /* Map */ @@ -1383,7 +1383,7 @@ calculate_first_last_level(struct st_texture_object *stObj) } else { firstLevel = 0; - lastLevel = MIN2(tObj->MaxLevel, tObj->Image[0][tObj->BaseLevel]->WidthLog2); + lastLevel = MIN2(tObj->MaxLevel, (int) tObj->Image[0][tObj->BaseLevel]->WidthLog2); } break; case GL_TEXTURE_RECTANGLE_NV: -- cgit v1.2.3 From f9c574d7192d2193ff3e12629a8db1a74b6dbf55 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 15 Jul 2008 11:44:47 +0200 Subject: mesa: Silence compiler warnings on Windows. --- src/mesa/shader/prog_execute.c | 2 +- src/mesa/shader/prog_parameter.c | 1 + src/mesa/shader/prog_print.c | 2 +- src/mesa/shader/prog_statevars.c | 8 ++++---- src/mesa/shader/shader_api.c | 6 +++--- src/mesa/shader/slang/slang_codegen.c | 2 +- src/mesa/shader/slang/slang_print.c | 13 +++++-------- src/mesa/shader/slang/slang_simplify.c | 10 +++++----- src/mesa/shader/slang/slang_vartable.c | 10 +++++----- 9 files changed, 26 insertions(+), 28 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c index abb143c4b9..4745211c65 100644 --- a/src/mesa/shader/prog_execute.c +++ b/src/mesa/shader/prog_execute.c @@ -716,7 +716,7 @@ _mesa_execute_program(GLcontext * ctx, * result.z = result.x * APPX(result.y) * We do what the ARB extension says. */ - q[2] = pow(2.0, t[0]); + q[2] = (GLfloat) pow(2.0, t[0]); } q[1] = t[0] - floor_t0; q[3] = 1.0F; diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index afdd7e5489..b302442227 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -319,6 +319,7 @@ _mesa_add_sampler(struct gl_program_parameter_list *paramList, return (GLint) paramList->ParameterValues[i][0]; } else { + GLuint i; const GLint size = 1; /* a sampler is basically a texture unit number */ GLfloat value; GLint numSamplers = 0; diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index 11f82c1fc1..014d08f062 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -363,7 +363,7 @@ _mesa_print_swizzle(GLuint swizzle) _mesa_printf(".xyzw\n"); } else { - char *s = _mesa_swizzle_string(swizzle, 0, 0); + const char *s = _mesa_swizzle_string(swizzle, 0, 0); _mesa_printf("%s\n", s); } } diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 7e444de304..81bb4122f2 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -249,7 +249,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], value[1] = ctx->Fog.Start; value[2] = ctx->Fog.End; value[3] = (ctx->Fog.End == ctx->Fog.Start) - ? 1.0 : (GLfloat)(1.0 / (ctx->Fog.End - ctx->Fog.Start)); + ? 1.0f : (GLfloat)(1.0 / (ctx->Fog.End - ctx->Fog.Start)); return; case STATE_CLIPPLANE: { @@ -409,9 +409,9 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], = ctx->Texture.Unit[unit]._Current; if (texObj) { struct gl_texture_image *texImage = texObj->Image[0][0]; - ASSIGN_4V(value, 1.0 / texImage->Width, + ASSIGN_4V(value, (GLfloat) (1.0 / texImage->Width), (GLfloat)(1.0 / texImage->Height), - 0.0, 1.0); + 0.0f, 1.0f); } } return; @@ -425,7 +425,7 @@ _mesa_fetch_state(GLcontext *ctx, const gl_state_index state[], * exp2: 2^-((density/(ln(2)^2) * fogcoord)^2) */ value[0] = (ctx->Fog.End == ctx->Fog.Start) - ? 1.0 : (GLfloat)(-1.0F / (ctx->Fog.End - ctx->Fog.Start)); + ? 1.0f : (GLfloat)(-1.0F / (ctx->Fog.End - ctx->Fog.Start)); value[1] = ctx->Fog.End * -value[0]; value[2] = (GLfloat)(ctx->Fog.Density * ONE_DIV_LN2); value[3] = (GLfloat)(ctx->Fog.Density * ONE_DIV_SQRT_LN2); diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index c1fbcde61e..53bf2108aa 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -969,7 +969,7 @@ get_uniformfv(GLcontext *ctx, GLuint program, GLint location, = _mesa_lookup_shader_program(ctx, program); if (shProg) { if (shProg->Uniforms && - location >= 0 && location < shProg->Uniforms->NumUniforms) { + location >= 0 && location < (GLint) shProg->Uniforms->NumUniforms) { GLint progPos; GLuint i; const struct gl_program *prog = NULL; @@ -1217,7 +1217,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, /* ordinary uniform variable */ GLsizei k, i; - if (count * elems > program->Parameters->Parameters[location].Size) { + if (count * elems > (GLint) program->Parameters->Parameters[location].Size) { _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(count too large)"); return; } @@ -1367,7 +1367,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, if (location == -1) return; /* The standard specifies this as a no-op */ - if (location < 0 || location >= shProg->Uniforms->NumUniforms) { + if (location < 0 || location >= (GLint) shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniformMatrix(location)"); return; } diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index af45dfb2e8..425beefe85 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2623,7 +2623,7 @@ _slang_gen_subscript(slang_assemble_ctx * A, slang_operation *oper) index = (GLint) oper->children[1].literal[0]; if (oper->children[1].type != SLANG_OPER_LITERAL_INT || - index >= max) { + index >= (GLint) max) { slang_info_log_error(A->log, "Invalid array index for vector type"); return NULL; } diff --git a/src/mesa/shader/slang/slang_print.c b/src/mesa/shader/slang/slang_print.c index ff9c62c929..f08c25ffdc 100644 --- a/src/mesa/shader/slang/slang_print.c +++ b/src/mesa/shader/slang/slang_print.c @@ -241,7 +241,7 @@ find_var(const slang_variable_scope *s, slang_atom name) void slang_print_tree(const slang_operation *op, int indent) { - int i; + GLuint i; switch (op->type) { @@ -261,13 +261,10 @@ slang_print_tree(const slang_operation *op, int indent) case SLANG_OPER_BLOCK_NEW_SCOPE: spaces(indent); printf("{{ // new scope locals=%p: ", (void*)op->locals); - { - int i; - for (i = 0; i < op->locals->num_variables; i++) { - printf("%s ", (char *) op->locals->variables[i]->a_name); - } - printf("\n"); + for (i = 0; i < op->locals->num_variables; i++) { + printf("%s ", (char *) op->locals->variables[i]->a_name); } + printf("\n"); print_generic(op, NULL, indent+3); spaces(indent); printf("}}\n"); @@ -665,7 +662,7 @@ slang_print_tree(const slang_operation *op, int indent) void slang_print_function(const slang_function *f, GLboolean body) { - int i; + GLuint i; #if 0 if (_mesa_strcmp((char *) f->header.a_name, "main") != 0) diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c index 158d6bc8cf..88ca83d288 100644 --- a/src/mesa/shader/slang/slang_simplify.c +++ b/src/mesa/shader/slang/slang_simplify.c @@ -76,7 +76,7 @@ _slang_lookup_constant(const char *name) for (i = 0; info[i].Name; i++) { if (strcmp(info[i].Name, name) == 0) { /* found */ - GLint value = -1.0; + GLint value = -1; _mesa_GetIntegerv(info[i].Token, &value); ASSERT(value >= 0); /* sanity check that glGetFloatv worked */ return value / info[i].Divisor; @@ -110,7 +110,7 @@ _slang_simplify(slang_operation *oper, oper->literal[0] = oper->literal[1] = oper->literal[2] = - oper->literal[3] = value; + oper->literal[3] = (GLfloat) value; oper->type = SLANG_OPER_LITERAL_INT; return; } @@ -380,7 +380,7 @@ _slang_adapt_call(slang_operation *callOper, const slang_function *fun, &origArg); callOper->children[i + j].children[1].type = SLANG_OPER_LITERAL_INT; - callOper->children[i + j].children[1].literal[0] = j; + callOper->children[i + j].children[1].literal[0] = (GLfloat) j; } } @@ -394,11 +394,11 @@ _slang_adapt_call(slang_operation *callOper, const slang_function *fun, } } - if (callOper->num_children < numParams) { + if (callOper->num_children < (GLuint) numParams) { /* still not enough args for all params */ return GL_FALSE; } - else if (callOper->num_children > numParams) { + else if (callOper->num_children > (GLuint) numParams) { /* now too many arguments */ /* XXX this isn't always an error, see spec */ return GL_FALSE; diff --git a/src/mesa/shader/slang/slang_vartable.c b/src/mesa/shader/slang/slang_vartable.c index bb4d2d656c..7bc92ea297 100644 --- a/src/mesa/shader/slang/slang_vartable.c +++ b/src/mesa/shader/slang/slang_vartable.c @@ -129,7 +129,7 @@ _slang_pop_var_table(slang_var_table *vt) /* just verify that any remaining allocations in this scope * were for temps */ - for (i = 0; i < vt->MaxRegisters * 4; i++) { + for (i = 0; i < (int) vt->MaxRegisters * 4; i++) { if (t->Temps[i] != FREE && t->Parent->Temps[i] == FREE) { if (dbg) printf(" Free reg %d\n", i/4); assert(t->Temps[i] == TEMP); @@ -206,7 +206,7 @@ alloc_reg(slang_var_table *vt, GLint size, GLboolean isTemp) for (i = 0; i <= vt->MaxRegisters * 4 - size; i += step) { GLuint found = 0; - for (j = 0; j < size; j++) { + for (j = 0; j < (GLuint) size; j++) { if (i + j < vt->MaxRegisters * 4 && t->Temps[i + j] == FREE) { found++; } @@ -218,7 +218,7 @@ alloc_reg(slang_var_table *vt, GLint size, GLboolean isTemp) /* found block of size free regs */ if (size > 1) assert(i % 4 == 0); - for (j = 0; j < size; j++) + for (j = 0; j < (GLuint) size; j++) t->Temps[i + j] = isTemp ? TEMP : VAR; assert(i < MAX_PROGRAM_TEMPS * 4); t->ValSize[i] = size; @@ -313,7 +313,7 @@ _slang_free_temp(slang_var_table *vt, slang_ir_storage *store) else { /*assert(store->Swizzle == SWIZZLE_NOOP);*/ assert(t->ValSize[r*4] == store->Size); - for (i = 0; i < store->Size; i++) { + for (i = 0; i < (GLuint) store->Size; i++) { assert(t->Temps[r * 4 + i] == TEMP); t->Temps[r * 4 + i] = FREE; } @@ -327,7 +327,7 @@ _slang_is_temp(const slang_var_table *vt, const slang_ir_storage *store) struct table *t = vt->Top; GLuint comp; assert(store->Index >= 0); - assert(store->Index < vt->MaxRegisters); + assert(store->Index < (int) vt->MaxRegisters); if (store->Swizzle == SWIZZLE_NOOP) comp = 0; else -- cgit v1.2.3 From bc4b1ca5636e68a6e6b0e3f88e9ff6b888837656 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 15 Jul 2008 13:11:08 -0600 Subject: mesa: improved printing --- src/mesa/shader/slang/slang_print.c | 19 +++++++++++++------ 1 file changed, 13 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_print.c b/src/mesa/shader/slang/slang_print.c index f08c25ffdc..bb0da9e34d 100644 --- a/src/mesa/shader/slang/slang_print.c +++ b/src/mesa/shader/slang/slang_print.c @@ -156,7 +156,7 @@ print_variable(const slang_variable *v, int indent) spaces(indent); printf("VAR "); print_type(&v->type); - printf(" %s", (char *) v->a_name); + printf(" %s (at %p)", (char *) v->a_name, (void *) v); if (v->initializer) { printf(" :=\n"); slang_print_tree(v->initializer, indent + 3); @@ -171,10 +171,12 @@ static void print_binary(const slang_operation *op, const char *oper, int indent) { assert(op->num_children == 2); +#if 0 printf("binary at %p locals=%p outer=%p\n", (void *) op, (void *) op->locals, (void *) op->locals->outer_scope); +#endif slang_print_tree(&op->children[0], indent + 3); spaces(indent); printf("%s at %p locals=%p outer=%p\n", @@ -260,9 +262,14 @@ slang_print_tree(const slang_operation *op, int indent) case SLANG_OPER_BLOCK_NEW_SCOPE: spaces(indent); - printf("{{ // new scope locals=%p: ", (void*)op->locals); - for (i = 0; i < op->locals->num_variables; i++) { - printf("%s ", (char *) op->locals->variables[i]->a_name); + printf("{{ // new scope locals=%p outer=%p: ", + (void *) op->locals, + (void *) op->locals->outer_scope); + { + int i; + for (i = 0; i < op->locals->num_variables; i++) { + printf("%s ", (char *) op->locals->variables[i]->a_name); + } } printf("\n"); print_generic(op, NULL, indent+3); @@ -669,8 +676,8 @@ slang_print_function(const slang_function *f, GLboolean body) return; #endif - printf("FUNCTION %s (\n", - (char *) f->header.a_name); + printf("FUNCTION %s ( scope=%p\n", + (char *) f->header.a_name, (void *) f->parameters); for (i = 0; i < f->param_count; i++) { print_variable(f->parameters->variables[i], 3); -- cgit v1.2.3 From f6152c35101311c2b128eb23b25b3f38d9e8b5be Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 15 Jul 2008 13:25:57 -0600 Subject: mesa: fix some function inlining bugs Need to add local vars of original function to the new scope's variable list (though the DECLs were already present). In slang_operation_copy() call slang_replace_scope() for SLANG_OPER_BLOCK_NEW_SCOPE opers. --- src/mesa/shader/slang/slang_codegen.c | 41 ++++++++++--------------- src/mesa/shader/slang/slang_compile_operation.c | 31 +++++++++++++++++++ src/mesa/shader/slang/slang_compile_operation.h | 5 +++ src/mesa/shader/slang/slang_compile_variable.h | 1 + 4 files changed, 53 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 425beefe85..9d5721214f 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -651,6 +651,8 @@ new_var(slang_assemble_ctx *A, slang_operation *oper, slang_atom name) if (!var) return NULL; + assert(var->declared); + assert(!oper->var || oper->var == var); n = new_node0(IR_VAR); @@ -924,31 +926,6 @@ slang_substitute(slang_assemble_ctx *A, slang_operation *oper, } -/** - * Recursively traverse 'oper', replacing occurances of 'oldScope' with - * 'newScope' in the oper->locals->outer_scope filed. - * - * This is used after function inlining to update the scoping of - * the newly copied/inlined code so that vars are found in the new, - * inlined scope and not in the original function code. - */ -static void -slang_replace_scope(slang_operation *oper, - slang_variable_scope *oldScope, - slang_variable_scope *newScope) -{ - GLuint i; - if (oper->locals != newScope && - oper->locals->outer_scope == oldScope) { - oper->locals->outer_scope = newScope; - } - for (i = 0; i < oper->num_children; i++) { - slang_replace_scope(&oper->children[i], oldScope, newScope); - } -} - - - /** * Produce inline code for a call to an assembly instruction. * This is typically used to compile a call to a built-in function like this: @@ -1235,6 +1212,16 @@ slang_inline_function_call(slang_assemble_ctx * A, slang_function *fun, } } + /* Now add copies of the function's local vars to the new variable scope */ + for (i = totalArgs; i < fun->parameters->num_variables; i++) { + slang_variable *p = fun->parameters->variables[i]; + slang_variable *pCopy = slang_variable_scope_grow(inlined->locals); + pCopy->type = p->type; + pCopy->a_name = p->a_name; + pCopy->array_len = p->array_len; + } + + /* New epilog statements: * 1. Create end of function label to jump to from return statements. * 2. Copy the 'out' parameter vars @@ -2105,6 +2092,8 @@ static slang_ir_node * _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) { slang_ir_node *n; + /*assert(!var->declared);*/ + var->declared = GL_TRUE; assert(!is_sampler_type(&var->type)); n = new_node0(IR_VAR_DECL); if (n) { @@ -3190,6 +3179,8 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, if (store) var->aux = store; /* save var's storage info */ + var->declared = GL_TRUE; + return success; } diff --git a/src/mesa/shader/slang/slang_compile_operation.c b/src/mesa/shader/slang/slang_compile_operation.c index b18e08d2ae..4d2fd5b666 100644 --- a/src/mesa/shader/slang/slang_compile_operation.c +++ b/src/mesa/shader/slang/slang_compile_operation.c @@ -69,8 +69,31 @@ slang_operation_destruct(slang_operation * oper) oper->locals = NULL; } + +/** + * Recursively traverse 'oper', replacing occurances of 'oldScope' with + * 'newScope' in the oper->locals->outer_scope field. + */ +void +slang_replace_scope(slang_operation *oper, + slang_variable_scope *oldScope, + slang_variable_scope *newScope) +{ + GLuint i; + if (oper->locals != newScope && + oper->locals->outer_scope == oldScope) { + oper->locals->outer_scope = newScope; + } + for (i = 0; i < oper->num_children; i++) { + slang_replace_scope(&oper->children[i], oldScope, newScope); + } +} + + /** * Recursively copy a slang_operation node. + * \param x copy target + * \param y copy source * \return GL_TRUE for success, GL_FALSE if failure */ GLboolean @@ -121,6 +144,14 @@ slang_operation_copy(slang_operation * x, const slang_operation * y) #endif slang_operation_destruct(x); *x = z; + + /* If this operation declares a new scope, we need to make sure + * all children point to it, not the original operation's scope! + */ + if (x->type == SLANG_OPER_BLOCK_NEW_SCOPE) { + slang_replace_scope(x, y->locals, x->locals); + } + return GL_TRUE; } diff --git a/src/mesa/shader/slang/slang_compile_operation.h b/src/mesa/shader/slang/slang_compile_operation.h index d5cbe779a6..4f92aa9a08 100644 --- a/src/mesa/shader/slang/slang_compile_operation.h +++ b/src/mesa/shader/slang/slang_compile_operation.h @@ -128,6 +128,11 @@ slang_operation_construct(slang_operation *); extern void slang_operation_destruct(slang_operation *); +extern void +slang_replace_scope(slang_operation *oper, + slang_variable_scope *oldScope, + slang_variable_scope *newScope); + extern GLboolean slang_operation_copy(slang_operation *, const slang_operation *); diff --git a/src/mesa/shader/slang/slang_compile_variable.h b/src/mesa/shader/slang/slang_compile_variable.h index d3691f0f51..9b0f85859a 100644 --- a/src/mesa/shader/slang/slang_compile_variable.h +++ b/src/mesa/shader/slang/slang_compile_variable.h @@ -80,6 +80,7 @@ typedef struct slang_variable_ GLuint address; /**< Storage location */ GLuint size; /**< Variable's size in bytes */ GLboolean isTemp; /**< a named temporary (__resultTmp) */ + GLboolean declared; /**< for debug */ void *aux; /**< Used during code gen */ } slang_variable; -- cgit v1.2.3 From f006358d56e6b24c3e32665e088d5ee11877ec2c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 15 Jul 2008 13:27:44 -0600 Subject: gallium: added some sanity check assertions --- src/gallium/drivers/softpipe/sp_setup.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index e99df9d018..4321ca46f4 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -941,6 +941,11 @@ setup_line(struct setup_context *setup, print_vertex(setup, v1); #endif + assert(v0[0][0] < 1.0e9); + assert(v0[0][1] < 1.0e9); + assert(v1[0][0] < 1.0e9); + assert(v1[0][1] < 1.0e9); + if (setup->softpipe->no_rast) return; -- cgit v1.2.3 From 5bf89cf2c399e40b3d5dc8cf03f03db70f50e777 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 15 Jul 2008 14:41:47 -0600 Subject: mesa: fix storage size computation in emit_arith() --- src/mesa/shader/slang/slang_emit.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 6d79380ab5..b087483b02 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -539,8 +539,9 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) /* result storage */ if (!n->Store) { - /* XXX this size isn't correct, it depends on the operands */ - if (!alloc_temp_storage(emitInfo, n, info->ResultSize)) + GLint size = n->Children[0]->Store + ? n->Children[0]->Store->Size : info->ResultSize; + if (!alloc_temp_storage(emitInfo, n, size)) return NULL; } storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); -- cgit v1.2.3 From 5e8e532cd026fe22d1fce88299a6d4d4ddcda71d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 15 Jul 2008 15:03:42 -0600 Subject: mesa: fix some broken bool, bvec2, bvec3, bvec4 constructors --- src/mesa/shader/slang/library/slang_core.gc | 28 ++++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 15a0792f43..82828a826c 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -113,13 +113,13 @@ int __constructor(const float f) bool __constructor(const int i) { const float zero = 0.0; - __asm vec4_seq __retVal, i, zero; + __asm vec4_sne __retVal, i, zero; } bool __constructor(const float f) { const float zero = 0.0; - __asm vec4_seq __retVal, i, zero; + __asm vec4_sne __retVal, f, zero; } int __constructor(const bool b) @@ -349,25 +349,25 @@ bvec2 __constructor(const bool b) bvec2 __constructor(const float f) { const vec2 zero = vec2(0.0, 0.0); - __asm vec4_seq __retVal.xy, f.xx, zero; + __asm vec4_sne __retVal.xy, f.xx, zero; } bvec2 __constructor(const int i) { const ivec2 zero = ivec2(0, 0); - __asm vec4_seq __retVal.xy, i.xx, zero; + __asm vec4_sne __retVal.xy, i.xx, zero; } bvec2 __constructor(const vec2 v) { const vec2 zero = vec2(0.0, 0.0); - __asm vec4_seq __retVal.xy, v, zero; + __asm vec4_sne __retVal.xy, v, zero; } bvec2 __constructor(const ivec2 v) { const ivec2 zero = ivec2(0, 0); - __asm vec4_seq __retVal.xy, v, zero; + __asm vec4_sne __retVal.xy, v, zero; } @@ -389,25 +389,25 @@ bvec3 __constructor(const bool b) bvec3 __constructor(const float f) { const vec3 zero = vec3(0.0, 0.0, 0.0); - __asm vec4_seq __retVal.xyz, f.xxx, zero; + __asm vec4_sne __retVal.xyz, f.xxx, zero; } bvec3 __constructor(const int i) { const ivec3 zero = ivec3(0, 0, 0); - __asm vec4_seq __retVal.xyz, i.xxx, zero; + __asm vec4_sne __retVal.xyz, i.xxx, zero; } bvec3 __constructor(const vec3 v) { const vec3 zero = vec3(0.0, 0.0, 0.0); - __asm vec4_seq __retVal.xyz, v, zero; + __asm vec4_sne __retVal.xyz, v, zero; } bvec3 __constructor(const ivec3 v) { const ivec3 zero = ivec3(0, 0, 0); - __asm vec4_seq __retVal.xyz, v, zero; + __asm vec4_sne __retVal.xyz, v, zero; } @@ -430,25 +430,25 @@ bvec4 __constructor(const bool b) bvec4 __constructor(const float f) { const vec4 zero = vec4(0.0, 0.0, 0.0, 0.0); - __asm vec4_seq __retVal, f.xxxx, zero; + __asm vec4_sne __retVal, f.xxxx, zero; } bvec4 __constructor(const int i) { const ivec4 zero = ivec4(0, 0, 0, 0); - __asm vec4_seq __retVal, i.xxxx, zero; + __asm vec4_sne __retVal, i.xxxx, zero; } bvec4 __constructor(const vec4 v) { const vec4 zero = vec4(0.0, 0.0, 0.0, 0.0); - __asm vec4_seq __retVal, v, zero; + __asm vec4_sne __retVal, v, zero; } bvec4 __constructor(const ivec4 v) { const ivec4 zero = ivec4(0, 0, 0, 0); - __asm vec4_seq __retVal, v, zero; + __asm vec4_sne __retVal, v, zero; } -- cgit v1.2.3 From 0bafe4b771f0a3e3dbfe16573a6cd3a7cb4244cd Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 15 Jul 2008 15:04:15 -0600 Subject: mesa: regenerated file --- src/mesa/shader/slang/library/slang_core_gc.h | 28 +++++++++++++-------------- 1 file changed, 14 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index 5213b693b3..3f5e6e4ea8 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -4,9 +4,9 @@ 3,1,0,5,1,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, 116,86,97,108,0,0,18,102,0,0,0,0,1,0,1,1,1,1,0,5,105,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0, -48,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,18,122, +48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,18,122, 101,114,111,0,0,0,0,1,0,1,1,1,1,0,9,102,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4, -118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,18,122,101,114,111,0, +118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,18,122,101,114,111,0, 0,0,0,1,0,5,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,1, 98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,5,105,0,0,0,1,4,105,110, 116,95,116,111,95,102,108,111,97,116,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,0,1,1,1, @@ -60,28 +60,28 @@ 98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,0,2,1,1,1,0,1,98,0,0, 0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,59,120,120,0,20,0,0,1,0,2,1,1,1,0,9, 102,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0, -0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,59, +0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,59, 120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,3,2,1,6,1,122,101,114,111,0,2, -58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114, +58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114, 101,116,86,97,108,0,59,120,121,0,0,18,105,0,59,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1, 0,10,118,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0,48,0,0,0,17,48,0,48,0,0, -0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0, +0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0, 0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,3,2,1,6,1,122,101,114,111,0,2,58,105,118, -101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, +101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86, 97,108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98, 50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95, 95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18, 98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18, 98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101, -99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0, +99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0, 18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,59,120,120,120,0,0,18,122,101,114,111, 0,0,0,0,1,0,3,1,1,1,0,5,105,0,0,0,1,3,2,1,7,1,122,101,114,111,0,2,58,105,118,101,99,51,0,16,8,48,0, -0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0, +0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0, 59,120,121,122,0,0,18,105,0,59,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,11,118,0,0, 0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48, -0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, +0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, 18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,7,118,0,0,0,1,3,2,1,7,1,122,101,114,111,0,2,58, -105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18, +105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18, 95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1, 0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,1,1,0,1,98,52,0,0,0,1,9,18,95,95,114,101,116,86, 97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18, @@ -89,15 +89,15 @@ 18,98,52,0,20,0,0,1,0,4,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119, 0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,4,1,1,1,0,9,102,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2, 58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118, -101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,59,120,120,120,120,0,0,18, +101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,59,120,120,120,120,0,0,18, 122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5,105,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101, -99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18, +99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18, 95,95,114,101,116,86,97,108,0,0,18,105,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1, 1,1,0,12,118,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48, -0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101, +0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, 116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,8,118,0,0,0,1,3,2,1,8,1,122, 101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118, -101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0, +101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0, 0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0, 0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114, 101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -- cgit v1.2.3 From e35329c701bfc36f73aaedb84ea1b704d660bd0f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 15 Jul 2008 15:32:53 -0600 Subject: mesa: fix some broken /= operators --- src/mesa/shader/slang/library/slang_core.gc | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 82828a826c..0fd2b7e6b3 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -1226,6 +1226,7 @@ void __operator /= (inout int a, const int b) float invB; __asm float_rcp invB, b; __asm vec4_multiply a, a, invB; + __asm float_to_int __retVal, a; } @@ -1518,7 +1519,7 @@ void __operator /= (inout vec2 v, const float a) { float invA; __asm float_rcp invA, a; - __asm vec4_multiply v.xy, v.xy, a.xx; + __asm vec4_multiply v.xy, v.xy, invA.xx; } @@ -1543,7 +1544,7 @@ void __operator /= (inout vec3 v, const float a) { float invA; __asm float_rcp invA, a; - __asm vec4_multiply v.xyz, v.xyz, a.xxx; + __asm vec4_multiply v.xyz, v.xyz, invA.xxx; } @@ -1568,7 +1569,7 @@ void __operator /= (inout vec4 v, const float a) { float invA; __asm float_rcp invA, a; - __asm vec4_multiply v, v, a.xxxx; + __asm vec4_multiply v, v, invA.xxxx; } -- cgit v1.2.3 From 9d43e387a3391b091f0a22710a7414a52eafae33 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 15 Jul 2008 15:33:04 -0600 Subject: mesa: regenerated file --- src/mesa/shader/slang/library/slang_core_gc.h | 730 +++++++++++++------------- 1 file changed, 366 insertions(+), 364 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index 3f5e6e4ea8..09ceb69edd 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -364,7 +364,8 @@ 101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5, 97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, 105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18, -97,0,0,18,105,110,118,66,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95, +97,0,0,18,105,110,118,66,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, +116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95, 97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4, 118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0, 2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18, @@ -450,372 +451,373 @@ 116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,4,1,0, 2,10,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112, 0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, -59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0, -9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120, -120,120,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116, -114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,3,1, -0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59, -120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,9,97,0, -0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18, -97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0, -59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4, -118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,2,1, -0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18, -118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0, -0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1, -1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0, -16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2, -21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59, -121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110, -0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46, -20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1, -1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0, -16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0, -57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20, -0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0, -16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0, -48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10, -49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,0,48,46, -18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120,120,0,48,18,109, -0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16, -10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49, -20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15, -109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, +59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,65,0,59,120,120,0,0,0,0,1,0,0,2,1,1,0,2,11, +118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0, +18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, +115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0, +0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0, +1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, +118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121, +122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,65,0,59,120,120,120,0,0,0,0,1,0,0,2,1,1,0,2,12, +118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120, +120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,12, +118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118, +0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105, +110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,105,110,118,65,0,59,120,120, +120,120,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13, +109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, 18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10, -50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, -16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, +57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48, +18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49, +0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49, +0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1, +0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, +16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9, 18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120, -120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,121,0,48,46,18,109,0, -16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16, -8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, -0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16, -10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122, -122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120, -120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,121,0,48,46,18,109,0,16, -10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10, -50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, -16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10, -51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,51,0,57,59,122,122,122, -122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59,119,119,119,119,0,48,46,20,0,0,1,0,15, -2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, -0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0, -13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13, -2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97, +120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50, +0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, +0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122, +122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110, +0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0, +48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1, +0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, +16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0, +57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10, +51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2, +21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48, +0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0, +48,46,18,109,0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120, +0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, +57,18,110,0,16,10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57, +59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48, +0,57,18,110,0,16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57, +59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48, +46,18,109,0,16,10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120, +0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, +57,18,110,0,16,10,51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, +59,119,119,119,119,0,48,46,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49, +20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, +0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0, +13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13, +2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, +57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97, 0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48, -0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20, -0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9, +0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20, +0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, +16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2,26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0, +1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, +47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2, +27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47, +20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, +0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0, +16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9, 98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2, -26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, -110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, -49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46, -20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, -0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50, -0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14, -110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0, -14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0, -1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110, -0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50, -0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109, -0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0, -57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2,27,1,1,0,9,97,0,0,1,1,0,15,110,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,9, -98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0,0,1,0,15,2,21,1,1,0,9,97,0,0, -1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,15,2,21,1,1,0, -15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,0,15,2, -22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, -110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, -49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,49,20,0,0,1,0, -15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, -0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, -0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,49,20,0, -0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16, -8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,0,48,46,20,0,0,1,0, -10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100, -111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, -100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21,1,1,0,14,109,0,0,1,1,0,11, -118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,0,48, -18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122, -122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,12, -2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0, -57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,121,0,48,46,18, -109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,118,0,59,119,119, -119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0, -0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21, -0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,13,110,0, -0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10, -49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20, -0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0,14, -110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0, -16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2,2,1,0,2,14,109,0, -0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,0,0,2,3,1,0,2, -14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,14,109,0,0, -1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,0,0,2,1,1,0,2, -15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16, -10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18, -109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1, -9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10, -51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20, -0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, -50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1, -0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,0,0, -2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0, -57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23, -0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0, -9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0, -16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57, -18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,0,0,2, -3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0, -57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,14,109,0,0,1,1,0,9,97,0,0, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9, +97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8, +48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1, +1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98, +0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2, +27,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, +110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, +49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0, +15,2,27,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, +0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, +0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0, +0,1,0,15,2,21,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18, +110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16, +10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57, +48,20,0,0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16, +10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57, +18,98,0,48,20,0,0,1,0,15,2,22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, +57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0, +18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0, +16,10,51,0,57,49,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +18,109,0,16,10,50,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, +16,10,51,0,57,18,98,0,49,20,0,0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0, +59,121,121,0,48,46,20,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21, +1,1,0,14,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18, +118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50, +0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, +57,0,0,20,0,0,1,0,12,2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121, +121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57, +18,118,0,59,119,119,119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, +57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10, +51,0,57,0,0,20,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110, +0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109, +0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0, +57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109, +0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14, +109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2, +2,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18, +109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22, +0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2, +4,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18, +109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24, +0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, +50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1, +0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51, +0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18, +109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57, +18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10, +50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2, +1,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, +57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22, +0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, +16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0, +9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2, +1,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, +57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0, +0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50, +0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2, +14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97, +0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18, +109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18, +97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9, +18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57, +18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1, +9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0, +57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0, 0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50, -0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, -21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10, -51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97, -0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16, -10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18, -97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,9,18,109,0, -16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57, -18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,9,18,109, -0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,118,0,18,118,0, -18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48, -20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0, -5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0, -18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49, -0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,25,1,0,2,7,118,0,0,0,1,9, -18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0, -18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49, -0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,25,1,0,2,9,97,0,0,0,1,9, -18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2, -25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, -12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, -15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, -49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, -48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, -0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2,24,1,0,2,5,97,0,0,0,1,9,18,97, -0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,24,1,0,2,6, -118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101, -116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101, -99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,24,1,0,2, -8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101, -116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24,1,0,2,10,118,0,0,0,1,9,18,118,0, -18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0, -20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, -46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,12,2,24,1,0,2,12,118,0,0,0,1,9,18, -118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118, -101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101, -99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,24, -1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0, -0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0, -0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,15,2,24,1,0,2,15,109,0,0,0,1,9,18,109, -0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10, -50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0, -57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,47,20,0,0,1,0,6,0,95,95,112, -111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, -18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,0,1,0,7,0,95,95,112,111,115,116,68, -101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, -118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95,95,112,111,115,116,68,101,99,114, -0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105, -118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,9,97, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,0, -1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,11,0, -95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,12,0,95,95,112, -111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0, -9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,13,0,95,95,112,111,115,116, -68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10, -49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,14,0,95,95,112, -111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0, -9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18, -109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109, -0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,15,0,95, -95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109, -0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0, -9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18, -109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109, -0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,9,0,95, -95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0, -20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2, -10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99, -50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,11,118,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49, -0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0, -46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0,95,95,112,111,115,116,73, -110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, -118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112,111,115,116,73,110,99,114, -0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105, -118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,8,118, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51, -0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,13,109,0,0,0,1,3,2, -0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49, -0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48, -0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,14,109,0,0,0,1, -3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0, -17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17, -49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0, -48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,15,109,0,0, -0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52, -0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, -49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, -48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, -103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0,0,0,0,1,0,1,2,15,1,1,0,5, -97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0, -0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0,0,4,102,108,111,97,116,95, -108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2,16,1,1,0,5,97,0,0,1,1,0,5, -98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,41,0,0,1,0,1, -2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108, -101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101, -0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1, -8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,43,0,0,1,0,1,2,17,1,1,0, -9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115, -115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18, -97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58, -102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0,0,1,0,0,0,112,114,105, -110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114,105,110,116,0,18,102,0,0, -0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105,110,116,95,112,114,105,110, -116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1,98,0,0,0,1,4,98,111,111,108, -95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,10,118,0,0, -0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69, -83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,11,118,0,0,0,1,9, -58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65, -0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0, -112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, -59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,6,118,0,0,0,1,9,58,112,114,105, -110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59, -121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7,118,0,0,0,1,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0, -0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116, -77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, -58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65, -0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0, -112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112, -114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, +0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109, +0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9, +18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18, +118,0,18,109,0,48,20,0,0,1,0,5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18, +95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58, +105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +7,2,25,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18, +95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0, +58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1, +0,9,2,25,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,18,97,0,20,0,0,1,0,10,2,25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49, +0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0, +0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,118,0,20,0,0,1,0,12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0, +17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13, +109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47, +20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0, +9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16, +8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49, +0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57, +18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,18,109,0,20,0,0,1,0,15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, +118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118, +101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2, +24,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, +97,0,20,0,0,1,0,6,2,24,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0, +0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18, +118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,0,1,0,8,2,24,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0, +0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18, +97,0,18,97,0,17,49,0,48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24, +1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95, +114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +12,2,24,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0, +57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,109,0,20,0,0,1,0,14,2,24,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, +15,2,24,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, +0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101, +99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16, +10,49,0,47,20,0,0,1,0,6,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20, +0,0,1,0,7,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95, +95,112,111,115,116,68,101,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0, +20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115, +116,68,101,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0, +18,97,0,17,49,0,48,0,0,47,20,0,0,1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0, +48,0,0,0,0,47,20,0,0,1,0,11,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, +47,20,0,0,1,0,12,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0, +1,0,13,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0, +0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0, +0,47,20,0,0,1,0,14,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17, +49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0, +48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0, +0,0,0,47,20,0,0,1,0,15,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0, +17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115, +116,73,110,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118, +0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110, +99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0, +2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101, +99,52,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0, +95,95,112,111,115,116,73,110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112, +111,115,116,73,110,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, +18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73, +110,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, +118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114, +0,1,0,2,13,109,0,0,0,1,3,2,0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110, +99,114,0,1,0,2,14,109,0,0,0,1,3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110, +99,114,0,1,0,2,15,109,0,0,0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, +58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58, +118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0, +0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0, +0,0,0,1,0,1,2,15,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102, +108,111,97,116,0,18,98,0,0,0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0, +0,4,102,108,111,97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2, +16,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0, +18,98,0,0,0,41,0,0,1,0,1,2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4, +102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95, +101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0, +5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0, +0,0,43,0,0,1,0,1,2,17,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108, +111,97,116,95,108,101,115,115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113, +117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0, +1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0, +0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114, +105,110,116,0,18,102,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105, +110,116,95,112,114,105,110,116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1, +98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77, +69,83,65,0,1,1,0,10,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, +58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, +83,65,0,1,1,0,11,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58, +112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, +18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112, +114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114, +105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1, +0,6,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105, +110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7, +118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110, +116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0, +0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77, +69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0, +9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83, +65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58, +112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, +18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112, +114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112, +114,105,110,116,77,69,83,65,0,1,1,0,4,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, 59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,4,118, -0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77, -69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0, -9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, -83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9, -58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77, -69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0, -9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69, -83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,15,109,0,0, -0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77, -69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0, -57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105, -110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1, -0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116, -0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95, -112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0, -1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0, -1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110, -116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 +116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0, +0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77, +69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0, +57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116, +77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49, +0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114, +105,110,116,77,69,83,65,0,1,1,0,15,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16, +8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114, +105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, +109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105, +110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17, +101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77, +69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112, +114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0, +0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105, +110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110, +116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 -- cgit v1.2.3 From 178726ebced83809ce42973f7225a1a5b9794f41 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 15 Jul 2008 16:04:26 -0600 Subject: mesa: add missing IR_LOG2 case --- src/mesa/shader/slang/slang_emit.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index b087483b02..f025b15de9 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1658,6 +1658,9 @@ emit(slang_emit_info *emitInfo, slang_ir_node *n) case IR_COS: case IR_DDX: case IR_DDY: + case IR_EXP: + case IR_EXP2: + case IR_LOG2: case IR_NOISE1: case IR_NOISE2: case IR_NOISE3: @@ -1678,8 +1681,6 @@ emit(slang_emit_info *emitInfo, slang_ir_node *n) case IR_SLE: case IR_SLT: case IR_POW: - case IR_EXP: - case IR_EXP2: /* trinary operators */ case IR_LRP: return emit_arith(emitInfo, n); -- cgit v1.2.3 From 45209baf5d0234c5c70cf2b581c9fda9f73b4eac Mon Sep 17 00:00:00 2001 From: Blair Sadewitz Date: Tue, 15 Jul 2008 17:11:33 -0600 Subject: mesa: added test for __NetBSD__ --- src/mesa/x86/assyntax.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/x86/assyntax.h b/src/mesa/x86/assyntax.h index 8eb47e41a1..524944f73a 100644 --- a/src/mesa/x86/assyntax.h +++ b/src/mesa/x86/assyntax.h @@ -985,7 +985,8 @@ SECTION _DATA public align=16 class=DATA use32 flat #if defined(Lynx) || (defined(SYSV) || defined(SVR4)) \ || (defined(__linux__) || defined(__OS2ELF__)) && defined(__ELF__) \ - || defined(__FreeBSD__) && __FreeBSD__ >= 3 + || (defined(__FreeBSD__) && __FreeBSD__ >= 3) \ + || (defined(__NetBSD__) && defined(__ELF__)) #define GLNAME(a) a #else #define GLNAME(a) CONCAT(_, a) -- cgit v1.2.3 From b4adb9e1de6728febbb758d6633edf765868153e Mon Sep 17 00:00:00 2001 From: Blair Sadewitz Date: Tue, 15 Jul 2008 17:12:23 -0600 Subject: mesa: added test for __NetBSD__ --- src/mesa/x86/common_x86.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/x86/common_x86.c b/src/mesa/x86/common_x86.c index 0b2af0a370..d93241a977 100644 --- a/src/mesa/x86/common_x86.c +++ b/src/mesa/x86/common_x86.c @@ -113,6 +113,14 @@ static void check_os_sse_support( void ) if (ret || !enabled) _mesa_x86_cpu_features &= ~(X86_FEATURE_XMM); } +#elif defined (__NetBSD__) + { + int ret, enabled; + size_t len = sizeof(enabled); + ret = sysctlbyname("machdep.sse", &enabled, &len, (void *)NULL, 0); + if (ret || !enabled) + _mesa_x86_cpu_features &= ~(X86_FEATURE_XMM); + } #elif defined(WIN32) LPTOP_LEVEL_EXCEPTION_FILTER oldFilter; -- cgit v1.2.3 From 0c2c0a862c40c0ed39a9ac52344d22c8a7c8b100 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Jul 2008 14:10:10 +0900 Subject: softpipe: DXT formats not really supported. --- src/gallium/drivers/softpipe/sp_screen.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index e9926bf41f..f6193bfaf9 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -119,11 +119,16 @@ softpipe_is_format_supported( struct pipe_screen *screen, { switch (type) { case PIPE_TEXTURE: - /* softpipe supports all texture formats */ - return TRUE; case PIPE_SURFACE: - /* softpipe supports all (off-screen) surface formats */ - return TRUE; + switch(format) { + case PIPE_FORMAT_DXT1_RGB: + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: + return FALSE; + default: + return TRUE; + } default: assert(0); return FALSE; -- cgit v1.2.3 From fbf0f6bd4d6ec20ead2797042866e6e910cd0f77 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Jul 2008 14:10:35 +0900 Subject: gallium: Add pf_is_compressed utility function. --- src/gallium/include/pipe/p_format.h | 31 +++++++++++-------------------- 1 file changed, 11 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index a2c6155d01..fefb8fdf1b 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -519,27 +519,18 @@ pf_get_nblocks(const struct pipe_format_block *block, unsigned width, unsigned h return pf_get_nblocksx(block, width)*pf_get_nblocksy(block, height); } -static INLINE void -pipe_rect_to_blocks(const struct pipe_format_block *block, - unsigned *width, unsigned *height, - unsigned *src_x, unsigned *src_y, - unsigned *dst_x, unsigned *dst_y) +static INLINE boolean +pf_is_compressed( enum pipe_format format ) { - assert(block->size > 0); - assert(block->width > 0); - assert(block->height > 0); - if(width) - *width = pf_get_nblocksx(block, *width); - if(height) - *height = pf_get_nblocksy(block, *height); - if(src_x) - *src_x /= block->width; - if(src_y) - *src_y /= block->height; - if(dst_x) - *dst_x /= block->width; - if(dst_y) - *dst_y /= block->height; + switch (format) { + case PIPE_FORMAT_DXT1_RGB: + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: + return TRUE; + default: + return FALSE; + } } #ifdef __cplusplus -- cgit v1.2.3 From 61c4de53c360fd2893b0490075d760f2ca3bac33 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Jul 2008 14:10:51 +0900 Subject: python: Silent debug output. --- src/gallium/state_trackers/python/gallium.i | 12 ++---------- 1 file changed, 2 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 2f1d8b22fb..fe0afb18a9 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -177,7 +177,6 @@ struct st_context { unsigned usage = 0 ) { struct pipe_texture templat; - struct pipe_texture *texture; memset(&templat, 0, sizeof(templat)); templat.format = format; pf_get_block(templat.format, &templat.block); @@ -187,9 +186,7 @@ struct st_context { templat.last_level = last_level; templat.target = target; templat.tex_usage = usage; - texture = $self->screen->texture_create($self->screen, &templat); - fprintf(stderr, "creating texture %p\n", texture); - return texture; + return $self->screen->texture_create($self->screen, &templat); } struct pipe_buffer * @@ -400,7 +397,6 @@ error1: ~pipe_texture() { struct pipe_texture *ptr = $self; - fprintf(stderr, "destroying texture %p\n", $self); pipe_texture_reference(&ptr, NULL); } @@ -409,10 +405,7 @@ error1: get_surface(unsigned face=0, unsigned level=0, unsigned zslice=0, unsigned usage=0 ) { struct pipe_screen *screen = $self->screen; - struct pipe_surface *surface; - surface = screen->get_tex_surface(screen, $self, face, level, zslice, usage); - fprintf(stderr, "creating surface %p\n", surface); - return surface; + return screen->get_tex_surface(screen, $self, face, level, zslice, usage); } }; @@ -422,7 +415,6 @@ error1: ~pipe_surface() { struct pipe_surface *ptr = $self; - fprintf(stderr, "destroying surface %p\n", $self); pipe_surface_reference(&ptr, NULL); } -- cgit v1.2.3 From 70b1ff9ff39ca29bdbd25b31ebb183eea683d625 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Jul 2008 14:11:45 +0900 Subject: python: Expand the texture test suit to cover one YUV and one DXT format. --- src/gallium/state_trackers/python/tests/base.py | 6 +++- src/gallium/state_trackers/python/tests/data.py | 31 +++++++++++------ src/gallium/state_trackers/python/tests/texture.py | 40 +++++++++++++++------- 3 files changed, 53 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py index d5d6114a45..d83abdfa76 100644 --- a/src/gallium/state_trackers/python/tests/base.py +++ b/src/gallium/state_trackers/python/tests/base.py @@ -93,6 +93,9 @@ class Test: def __init__(self): pass + def description(self): + raise NotImplementedError + def run(self): raise NotImplementedError @@ -111,7 +114,8 @@ class TestSuite(Test): def run(self): for test in self.tests: - self.test.run() + print "Running %s..." % test.description() + test.run() class TextureTemplate: diff --git a/src/gallium/state_trackers/python/tests/data.py b/src/gallium/state_trackers/python/tests/data.py index 2287884285..3c1cda59d7 100644 --- a/src/gallium/state_trackers/python/tests/data.py +++ b/src/gallium/state_trackers/python/tests/data.py @@ -211,7 +211,6 @@ dxt1_rgba = [ def generate_data_compressed(surface, blocks): - pixels, block = blocks[0] stride = surface.nblocksx*surface.block.size size = surface.nblocksy*stride @@ -219,19 +218,23 @@ def generate_data_compressed(surface, blocks): raw = ByteArray(size) rgba = FloatArray(surface.height*surface.width*4) - for y in range(0, surface.nblocksx): - for x in range(0, surface.nblocksy): + for yj in range(0, surface.nblocksy): + for xj in range(0, surface.nblocksx): + pixels, block = blocks[random.randint(0, len(blocks) - 1)] - offset = (y*surface.nblocksx + x)*surface.block.width + offset = (yj*surface.nblocksx + xj)*surface.block.size for i in range(0, surface.block.size): raw[offset + i] = block[i] - for j in range(0, surface.block.width): - for i in range(0, surface.block.height): - offset = ((y*surface.block.height + j)*surface.width + x*surface.block.width + i)*4 - pixel = pixels[j*surface.block.width + i] - for ch in range(0, 4): - rgba[offset + ch] = float(pixel[ch])/255.0 + for yi in range(0, surface.block.height): + for xi in range(0, surface.block.width): + y = yj*surface.block.height + yi + x = xj*surface.block.width + xi + if y < surface.height and x < surface.width: + offset = (y*surface.width + x)*4 + pixel = pixels[yi*surface.block.width + xi] + for ch in range(0, 4): + rgba[offset + ch] = float(pixel[ch])/255.0 surface.put_tile_raw(0, 0, surface.width, surface.height, raw, stride) @@ -252,6 +255,14 @@ def generate_data_simple(surface): surface.get_tile_rgba(0, 0, surface.width, surface.height, rgba) + if surface.format in (PIPE_FORMAT_YCBCR, PIPE_FORMAT_YCBCR_REV): + # normalize + for y in range(0, surface.height): + for x in range(0, surface.width): + for ch in range(4): + offset = (y*surface.width + x)*4 + ch + rgba[offset] = min(max(rgba[offset], 0.0), 1.0) + return rgba diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index 6a5ffb38b9..7812d8aebf 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -33,7 +33,7 @@ from base import * from data import generate_data -def compare_rgba(width, height, rgba1, rgba2, tol=0.01): +def compare_rgba(width, height, rgba1, rgba2, tol=4.0/256): errors = 0 for y in range(0, height): for x in range(0, width): @@ -58,22 +58,25 @@ class TextureTest(Test): Test.__init__(self) self.__dict__.update(kargs) + def description(self): + return "%s %ux%u" % (formats[self.format], self.width, self.height) + def run(self): dev = self.dev - #format = PIPE_FORMAT_A8R8G8B8_UNORM - format = PIPE_FORMAT_DXT1_RGB + format = self.format + width = self.width + height = self.height if not dev.is_format_supported(format, PIPE_TEXTURE): - pass + print "SKIP" + return if not dev.is_format_supported(format, PIPE_SURFACE): - pass + print "SKIP" + return ctx = dev.context_create() - width = 64 - height = 64 - # disabled blending/masking blend = Blend() blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE @@ -169,7 +172,7 @@ class TextureTest(Test): 0:TEX OUT[0], IN[0], SAMP[0], 2D 1:END ''') - fs.dump() + #fs.dump() ctx.set_fragment_shader(fs) nverts = 4 @@ -227,15 +230,26 @@ class TextureTest(Test): cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ).get_tile_rgba(x, y, w, h, rgba) - compare_rgba(width, height, rgba, expected_rgba) + if compare_rgba(width, height, rgba, expected_rgba): + print "OK" + else: + print "FAIL" - show_image(width, height, Result=rgba, Expected=expected_rgba) + show_image(width, height, Result=rgba, Expected=expected_rgba) + #save_image(width, height, rgba, "result.png") + #save_image(width, height, expected_rgba, "expected.png") + sys.exit(0) def main(): dev = Device() - test = TextureTest(dev = dev) - test.run() + suite = TestSuite() + formats = [PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_YCBCR, PIPE_FORMAT_DXT1_RGB] + sizes = [64, 32, 16, 8, 4, 2] + for format in formats: + for size in sizes: + suite.add_test(TextureTest(dev=dev, format=format, width=size, height=size)) + suite.run() if __name__ == '__main__': -- cgit v1.2.3 From 78d18bb690d47eba6d13f8f154e64f1c33ef29fd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Jul 2008 19:36:36 +0900 Subject: gallium: ycbcr_get_tile_rgba allow reading an uneven number of pixels from yuv surfaces. --- src/gallium/auxiliary/util/p_tile.c | 32 ++++++++++++++++++++++++++++---- 1 file changed, 28 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 1a1a2d96cc..580a95568e 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -632,13 +632,10 @@ ycbcr_get_tile_rgba(ushort *src, const float scale = 1.0f / 255.0f; unsigned i, j; - /* we're assuming we're being asked for an even number of texels */ - assert((w & 1) == 0); - for (i = 0; i < h; i++) { float *pRow = p; /* do two texels at a time */ - for (j = 0; j < w; j += 2, src += 2) { + for (j = 0; j < (w & ~1); j += 2, src += 2) { const ushort t0 = src[0]; const ushort t1 = src[1]; const ubyte y0 = (t0 >> 8) & 0xff; /* luminance */ @@ -676,6 +673,33 @@ ycbcr_get_tile_rgba(ushort *src, pRow += 4; } + /* do the last texel */ + if (w & 1) { + const ushort t0 = src[0]; + const ushort t1 = src[1]; + const ubyte y0 = (t0 >> 8) & 0xff; /* luminance */ + ubyte cb, cr; + float r, g, b; + + if (rev) { + cb = t1 & 0xff; /* chroma U */ + cr = t0 & 0xff; /* chroma V */ + } + else { + cb = t0 & 0xff; /* chroma U */ + cr = t1 & 0xff; /* chroma V */ + } + + /* even pixel: y0,cr,cb */ + r = 1.164f * (y0-16) + 1.596f * (cr-128); + g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128); + b = 1.164f * (y0-16) + 2.018f * (cb-128); + pRow[0] = r * scale; + pRow[1] = g * scale; + pRow[2] = b * scale; + pRow[3] = 1.0f; + pRow += 4; + } p += dst_stride; } } -- cgit v1.2.3 From 32fe752d31a07293ad68cccb2fffbbfef32ab449 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Jul 2008 19:36:59 +0900 Subject: python: Test all miplevels. --- src/gallium/state_trackers/python/tests/base.py | 87 ++++++++++++++++++---- src/gallium/state_trackers/python/tests/texture.py | 72 +++++++++++------- 2 files changed, 116 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py index d83abdfa76..c50abf7f69 100644 --- a/src/gallium/state_trackers/python/tests/base.py +++ b/src/gallium/state_trackers/python/tests/base.py @@ -27,6 +27,12 @@ ########################################################################## +"""Base classes for tests. + +Loosely inspired on Python's unittest module. +""" + + from gallium import * @@ -87,17 +93,54 @@ def show_image(width, height, **rgbas): root.mainloop() +class TestFailure(Exception): -class Test: + pass + +class TestSkip(Exception): + pass + + +class Test: + def __init__(self): pass + + def _run(self, result): + raise NotImplementedError + def run(self): + result = TestResult() + self._run(result) + result.summary() + + +class TestCase(Test): + + def __init__(self, dev, **kargs): + Test.__init__(self) + self.dev = dev + self.__dict__.update(kargs) + def description(self): raise NotImplementedError - def run(self): + def test(self): raise NotImplementedError + + def _run(self, result): + result.test_start(self) + try: + self.test() + except KeyboardInterrupt: + raise + except TestSkip: + result.test_skipped(self) + except TestFailure: + result.test_failed(self) + else: + result.test_passed(self) class TestSuite(Test): @@ -112,21 +155,35 @@ class TestSuite(Test): def add_test(self, test): self.tests.append(test) - def run(self): + def _run(self, result): for test in self.tests: - print "Running %s..." % test.description() - test.run() + test._run(result) -class TextureTemplate: +class TestResult: - def __init__(self, format=PIPE_FORMAT_R8G8B8A8_UNORM, width=1, height=1, depth=1, last_level=0, target=PIPE_TEXTURE_2D): - self.format = format - self.width = width - self.height = height - self.depth = depth - self.last_level = last_level - self.target = target - - + def __init__(self): + self.tests = 0 + self.passed = 0 + self.skipped = 0 + self.failed = 0 + + def test_start(self, test): + self.tests += 1 + print "Running %s..." % test.description() + + def test_passed(self, test): + self.passed += 1 + print "PASS" + + def test_skipped(self, test): + self.skipped += 1 + print "SKIP" + + def test_failed(self): + self.failed += 1 + print "FAIL" + def summary(self): + print "%u tests, %u passed, %u skipped, %u failed" % (self.tests, self.passed, self.skipped, self.failed) + \ No newline at end of file diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index 7812d8aebf..7c7245e65c 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -52,30 +52,39 @@ def compare_rgba(width, height, rgba1, rgba2, tol=4.0/256): return errors == 0 -class TextureTest(Test): - - def __init__(self, **kargs): - Test.__init__(self) - self.__dict__.update(kargs) +def lods(*dims): + size = max(dims) + lods = 0 + while size: + lods += 1 + size >>= 1 + return lods + + +def minify(dims, level = 1): + return [max(dim>>level, 1) for dim in dims] + +class TextureTest(TestCase): + def description(self): - return "%s %ux%u" % (formats[self.format], self.width, self.height) - - def run(self): + return "%s %ux%u level=%u" % (formats[self.format], self.width, self.height, self.level) + + def test(self): dev = self.dev format = self.format width = self.width height = self.height + level = self.level + + levels = lods(width, height) + if not dev.is_format_supported(format, PIPE_TEXTURE): - print "SKIP" - return - if not dev.is_format_supported(format, PIPE_SURFACE): - print "SKIP" - return + raise TestSkip - ctx = dev.context_create() + ctx = self.dev.context_create() # disabled blending/masking blend = Blend() @@ -123,15 +132,20 @@ class TextureTest(Test): sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST sampler.normalized_coords = 1 + sampler.min_lod = level + sampler.max_lod = level ctx.set_sampler(0, sampler) # texture texture = dev.texture_create(format, width, - height) + height, + last_level = levels - 1) ctx.set_sampler_texture(0, texture) - expected_rgba = generate_data(texture.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE)) + expected_rgba = generate_data(texture.get_surface( + usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE, + level = level)) # framebuffer cbuf_tex = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, @@ -181,8 +195,7 @@ class TextureTest(Test): x = 0 y = 0 - w = width - h = height + w, h = minify((width, height), level) verts[ 0] = x # x1 verts[ 1] = y # y1 @@ -223,32 +236,35 @@ class TextureTest(Test): verts) ctx.flush() - - del ctx - - rgba = FloatArray(height*width*4) + + rgba = FloatArray(h*w*4) cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ).get_tile_rgba(x, y, w, h, rgba) - if compare_rgba(width, height, rgba, expected_rgba): - print "OK" - else: - print "FAIL" + if not compare_rgba(w, h, rgba, expected_rgba): - show_image(width, height, Result=rgba, Expected=expected_rgba) + show_image(w, h, Result=rgba, Expected=expected_rgba) #save_image(width, height, rgba, "result.png") #save_image(width, height, expected_rgba, "expected.png") sys.exit(0) + + raise TestFailure + + del ctx + def main(): dev = Device() suite = TestSuite() formats = [PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_YCBCR, PIPE_FORMAT_DXT1_RGB] + #formats = [PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_DXT1_RGB] sizes = [64, 32, 16, 8, 4, 2] for format in formats: for size in sizes: - suite.add_test(TextureTest(dev=dev, format=format, width=size, height=size)) + levels = lods(size) + for level in range(levels): + suite.add_test(TextureTest(dev=dev, format=format, width=size, height=size, level=level)) suite.run() -- cgit v1.2.3 From 1049e65240539228f64a91cc7b55af37bbd0417b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Jul 2008 23:26:23 +0900 Subject: python: Set PIPE_TEX_MIPFILTER_NONE by default. --- src/gallium/state_trackers/python/st_device.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index 9e625ae2e0..2e53a83eea 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -192,7 +192,7 @@ st_context_create(struct st_device *st_dev) sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE; sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE; sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE; - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE; + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST; sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST; sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST; sampler.normalized_coords = 1; -- cgit v1.2.3 From 7f32834df3c428fa283e966982926c94630b1b8e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 16 Jul 2008 23:27:15 +0900 Subject: python/tests: Extend the texture tests to cover cube maps. --- src/gallium/state_trackers/python/tests/base.py | 8 +- src/gallium/state_trackers/python/tests/texture.py | 215 +++++++++++++++------ 2 files changed, 162 insertions(+), 61 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/tests/base.py b/src/gallium/state_trackers/python/tests/base.py index c50abf7f69..8477aa5fc9 100644 --- a/src/gallium/state_trackers/python/tests/base.py +++ b/src/gallium/state_trackers/python/tests/base.py @@ -53,7 +53,7 @@ def make_image(width, height, rgba): for y in range(0, height): for x in range(0, width): offset = (y*width + x)*4 - r, g, b, a = [int(rgba[offset + ch]*255) for ch in range(4)] + r, g, b, a = [int(min(max(rgba[offset + ch], 0.0), 1.0)*255) for ch in range(4)] outpixels[x, y] = r, g, b return outimage @@ -167,6 +167,7 @@ class TestResult: self.passed = 0 self.skipped = 0 self.failed = 0 + self.failed_descriptions = [] def test_start(self, test): self.tests += 1 @@ -180,10 +181,13 @@ class TestResult: self.skipped += 1 print "SKIP" - def test_failed(self): + def test_failed(self, test): self.failed += 1 + self.failed_descriptions.append(test.description()) print "FAIL" def summary(self): print "%u tests, %u passed, %u skipped, %u failed" % (self.tests, self.passed, self.skipped, self.failed) + for description in self.failed_descriptions: + print " %s" % description \ No newline at end of file diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index 7c7245e65c..448bdbcd3e 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -33,10 +33,11 @@ from base import * from data import generate_data -def compare_rgba(width, height, rgba1, rgba2, tol=4.0/256): +def compare_rgba(width, height, rgba1, rgba2, tol=4.0/256, ratio=0.85): errors = 0 for y in range(0, height): for x in range(0, width): + differs = 0 for ch in range(4): offset = (y*width + x)*4 + ch v1 = rgba1[offset] @@ -44,12 +45,14 @@ def compare_rgba(width, height, rgba1, rgba2, tol=4.0/256): if abs(v1 - v2) > tol: if errors == 0: sys.stderr.write("x=%u, y=%u, ch=%u differ: %f vs %f\n" % (x, y, ch, v1, v2)) - if errors == 1: + if errors == 1 and ch == 0: sys.stderr.write("...\n") - errors += 1 + differs = 1 + errors += differs + total = height*width if errors: - sys.stderr.write("%u out of %u pixels differ\n" % (errors/4, height*width)) - return errors == 0 + sys.stderr.write("%u out of %u pixels differ\n" % (errors, total)) + return float(total - errors)/float(total) >= ratio def lods(*dims): @@ -65,22 +68,87 @@ def minify(dims, level = 1): return [max(dim>>level, 1) for dim in dims] +def tex_coords(texture, face, level, zslice): + st = [ + [0.0, 0.0], + [1.0, 0.0], + [1.0, 1.0], + [0.0, 1.0], + ] + + if texture.target == PIPE_TEXTURE_2D: + return [[s, t, 0.0] for s, t in st] + elif texture.target == PIPE_TEXTURE_3D: + assert 0 + elif texture.target == PIPE_TEXTURE_CUBE: + result = [] + for s, t in st: + # See http://developer.nvidia.com/object/cube_map_ogl_tutorial.html + sc = 2.0*s - 1.0 + tc = 2.0*t - 1.0 + if face == PIPE_TEX_FACE_POS_X: + rx = 1.0 + ry = -tc + rz = -sc + if face == PIPE_TEX_FACE_NEG_X: + rx = -1.0 + ry = -tc + rz = sc + if face == PIPE_TEX_FACE_POS_Y: + rx = sc + ry = 1.0 + rz = tc + if face == PIPE_TEX_FACE_NEG_Y: + rx = sc + ry = -1.0 + rz = -tc + if face == PIPE_TEX_FACE_POS_Z: + rx = sc + ry = -tc + rz = 1.0 + if face == PIPE_TEX_FACE_NEG_Z: + rx = -sc + ry = -tc + rz = -1.0 + result.append([rx, ry, rz]) + return result + + class TextureTest(TestCase): def description(self): - return "%s %ux%u level=%u" % (formats[self.format], self.width, self.height, self.level) + target = { + PIPE_TEXTURE_1D: "1d", + PIPE_TEXTURE_2D: "2d", + PIPE_TEXTURE_3D: "3d", + PIPE_TEXTURE_CUBE: "cube", + }[self.target] + format = formats[self.format] + if self.target == PIPE_TEXTURE_CUBE: + face = { + PIPE_TEX_FACE_POS_X: "+x", + PIPE_TEX_FACE_NEG_X: "-x", + PIPE_TEX_FACE_POS_Y: "+y", + PIPE_TEX_FACE_NEG_Y: "-y", + PIPE_TEX_FACE_POS_Z: "+z", + PIPE_TEX_FACE_NEG_Z: "-z", + }[self.face] + else: + face = "" + return "%s %s %ux%u %s level=%u" % (target, format, self.width, self.height, face, self.level) def test(self): dev = self.dev + target = self.target format = self.format width = self.width height = self.height level = self.level + face = self.face levels = lods(width, height) - if not dev.is_format_supported(format, PIPE_TEXTURE): raise TestSkip @@ -128,23 +196,25 @@ class TextureTest(TestCase): sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE - sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE + sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST sampler.normalized_coords = 1 - sampler.min_lod = level - sampler.max_lod = level + sampler.min_lod = 0 + sampler.max_lod = PIPE_MAX_TEXTURE_LEVELS - 1 ctx.set_sampler(0, sampler) # texture - texture = dev.texture_create(format, - width, - height, + texture = dev.texture_create(target=target, + format=format, + width=width, + height=height, last_level = levels - 1) ctx.set_sampler_texture(0, texture) expected_rgba = generate_data(texture.get_surface( usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE, + face = face, level = level)) # framebuffer @@ -178,14 +248,20 @@ class TextureTest(TestCase): ctx.set_vertex_shader(vs) # fragment shader + op = { + PIPE_TEXTURE_1D: "1D", + PIPE_TEXTURE_2D: "2D", + PIPE_TEXTURE_3D: "3D", + PIPE_TEXTURE_CUBE: "CUBE", + }[target] fs = Shader(''' FRAG1.1 - DCL IN[0], GENERIC[0], PERSPECTIVE + DCL IN[0], GENERIC[0], LINEAR DCL OUT[0], COLOR, CONSTANT DCL SAMP[0], CONSTANT - 0:TEX OUT[0], IN[0], SAMP[0], 2D + 0:TEX OUT[0], IN[0], SAMP[0], %s 1:END - ''') + ''' % op) #fs.dump() ctx.set_fragment_shader(fs) @@ -197,38 +273,25 @@ class TextureTest(TestCase): y = 0 w, h = minify((width, height), level) - verts[ 0] = x # x1 - verts[ 1] = y # y1 - verts[ 2] = 0.0 # z1 - verts[ 3] = 1.0 # w1 - verts[ 4] = 0.0 # s1 - verts[ 5] = 0.0 # t - verts[ 6] = 0.0 - verts[ 7] = 0.0 - verts[ 8] = x + w # x2 - verts[ 9] = y # y2 - verts[10] = 0.0 # z2 - verts[11] = 1.0 # w2 - verts[12] = 1.0 # s2 - verts[13] = 0.0 # t2 - verts[14] = 0.0 - verts[15] = 0.0 - verts[16] = x + w # x3 - verts[17] = y + h # y3 - verts[18] = 0.0 # z3 - verts[19] = 1.0 # w3 - verts[20] = 1.0 # s3 - verts[21] = 1.0 # t3 - verts[22] = 0.0 - verts[23] = 0.0 - verts[24] = x # x4 - verts[25] = y + h # y4 - verts[26] = 0.0 # z4 - verts[27] = 1.0 # w4 - verts[28] = 0.0 # s4 - verts[29] = 1.0 # t4 - verts[30] = 0.0 - verts[31] = 0.0 + pos = [ + [x, y], + [x+w, y], + [x+w, y+h], + [x, y+h], + ] + + tex = tex_coords(texture, face, level, zslice=0) + + for i in range(0, 4): + j = 8*i + verts[j + 0] = pos[i][0] # x + verts[j + 1] = pos[i][1] # y + verts[j + 2] = 0.0 # z + verts[j + 3] = 1.0 # w + verts[j + 4] = tex[i][0] # s + verts[j + 5] = tex[i][1] # r + verts[j + 6] = tex[i][2] # q + verts[j + 7] = 1.0 ctx.draw_vertices(PIPE_PRIM_TRIANGLE_FAN, nverts, @@ -243,10 +306,10 @@ class TextureTest(TestCase): if not compare_rgba(w, h, rgba, expected_rgba): - show_image(w, h, Result=rgba, Expected=expected_rgba) - #save_image(width, height, rgba, "result.png") - #save_image(width, height, expected_rgba, "expected.png") - sys.exit(0) + #show_image(w, h, Result=rgba, Expected=expected_rgba) + #save_image(w, h, rgba, "result.png") + #save_image(w, h, expected_rgba, "expected.png") + #sys.exit(0) raise TestFailure @@ -257,14 +320,48 @@ class TextureTest(TestCase): def main(): dev = Device() suite = TestSuite() - formats = [PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_YCBCR, PIPE_FORMAT_DXT1_RGB] - #formats = [PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_FORMAT_DXT1_RGB] + + targets = [] + targets += [PIPE_TEXTURE_2D] + targets += [PIPE_TEXTURE_CUBE] + + formats = [] + formats += [PIPE_FORMAT_A8R8G8B8_UNORM] + formats += [PIPE_FORMAT_R5G6B5_UNORM] + formats += [PIPE_FORMAT_L8_UNORM] + formats += [PIPE_FORMAT_YCBCR] + formats += [PIPE_FORMAT_DXT1_RGB] + sizes = [64, 32, 16, 8, 4, 2] - for format in formats: - for size in sizes: - levels = lods(size) - for level in range(levels): - suite.add_test(TextureTest(dev=dev, format=format, width=size, height=size, level=level)) + #sizes = [16] + + for target in targets: + for format in formats: + for size in sizes: + if target == PIPE_TEXTURE_CUBE: + faces = [ + PIPE_TEX_FACE_POS_X, + PIPE_TEX_FACE_NEG_X, + PIPE_TEX_FACE_POS_Y, + PIPE_TEX_FACE_NEG_Y, + PIPE_TEX_FACE_POS_Z, + PIPE_TEX_FACE_NEG_Z, + ] + else: + faces = [0] + for face in faces: + levels = lods(size) + for level in range(levels): + test = TextureTest( + dev=dev, + target=target, + format=format, + width=size, + height=size, + face=face, + level=level, + ) + suite.add_test(test) suite.run() -- cgit v1.2.3 From 479a807e01bc980aa2a3c9c56c20ba53cc6e7d60 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Jul 2008 10:23:28 -0600 Subject: mesa: add GL_POLYGON_OFFSET_POINT/LINE/FILL queries, remove GL_TEXTURE_ENV_COLOR, GL_TEXTURE_ENV_MODE Issues found by Bob Ellison. --- src/mesa/main/get.c | 63 +++++++++++++++++++++--------------------------- src/mesa/main/get_gen.py | 9 +++---- 2 files changed, 30 insertions(+), 42 deletions(-) (limited to 'src') diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c index ac437a86e2..4674c44a3f 100644 --- a/src/mesa/main/get.c +++ b/src/mesa/main/get.c @@ -749,6 +749,15 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params ) case GL_POLYGON_OFFSET_UNITS: params[0] = FLOAT_TO_BOOLEAN(ctx->Polygon.OffsetUnits ); break; + case GL_POLYGON_OFFSET_POINT: + params[0] = ctx->Polygon.OffsetPoint; + break; + case GL_POLYGON_OFFSET_LINE: + params[0] = ctx->Polygon.OffsetLine; + break; + case GL_POLYGON_OFFSET_FILL: + params[0] = ctx->Polygon.OffsetFill; + break; case GL_POLYGON_SMOOTH: params[0] = ctx->Polygon.SmoothFlag; break; @@ -891,18 +900,6 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params ) CHECK_EXT1(MESA_texture_array, "GetBooleanv"); params[0] = INT_TO_BOOLEAN(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current2DArray->Name); break; - case GL_TEXTURE_ENV_COLOR: - { - const GLfloat *color = ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvColor; - params[0] = FLOAT_TO_BOOLEAN(color[0]); - params[1] = FLOAT_TO_BOOLEAN(color[1]); - params[2] = FLOAT_TO_BOOLEAN(color[2]); - params[3] = FLOAT_TO_BOOLEAN(color[3]); - } - break; - case GL_TEXTURE_ENV_MODE: - params[0] = ENUM_TO_BOOLEAN(ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvMode); - break; case GL_TEXTURE_GEN_S: params[0] = ((ctx->Texture.Unit[ctx->Texture.CurrentUnit].TexGenEnabled & S_BIT) ? 1 : 0); break; @@ -2596,6 +2593,15 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) case GL_POLYGON_OFFSET_UNITS: params[0] = ctx->Polygon.OffsetUnits ; break; + case GL_POLYGON_OFFSET_POINT: + params[0] = BOOLEAN_TO_FLOAT(ctx->Polygon.OffsetPoint); + break; + case GL_POLYGON_OFFSET_LINE: + params[0] = BOOLEAN_TO_FLOAT(ctx->Polygon.OffsetLine); + break; + case GL_POLYGON_OFFSET_FILL: + params[0] = BOOLEAN_TO_FLOAT(ctx->Polygon.OffsetFill); + break; case GL_POLYGON_SMOOTH: params[0] = BOOLEAN_TO_FLOAT(ctx->Polygon.SmoothFlag); break; @@ -2738,18 +2744,6 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params ) CHECK_EXT1(MESA_texture_array, "GetFloatv"); params[0] = (GLfloat)(ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current2DArray->Name); break; - case GL_TEXTURE_ENV_COLOR: - { - const GLfloat *color = ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvColor; - params[0] = color[0]; - params[1] = color[1]; - params[2] = color[2]; - params[3] = color[3]; - } - break; - case GL_TEXTURE_ENV_MODE: - params[0] = ENUM_TO_FLOAT(ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvMode); - break; case GL_TEXTURE_GEN_S: params[0] = BOOLEAN_TO_FLOAT(((ctx->Texture.Unit[ctx->Texture.CurrentUnit].TexGenEnabled & S_BIT) ? 1 : 0)); break; @@ -4443,6 +4437,15 @@ _mesa_GetIntegerv( GLenum pname, GLint *params ) case GL_POLYGON_OFFSET_UNITS: params[0] = IROUND(ctx->Polygon.OffsetUnits ); break; + case GL_POLYGON_OFFSET_POINT: + params[0] = BOOLEAN_TO_INT(ctx->Polygon.OffsetPoint); + break; + case GL_POLYGON_OFFSET_LINE: + params[0] = BOOLEAN_TO_INT(ctx->Polygon.OffsetLine); + break; + case GL_POLYGON_OFFSET_FILL: + params[0] = BOOLEAN_TO_INT(ctx->Polygon.OffsetFill); + break; case GL_POLYGON_SMOOTH: params[0] = BOOLEAN_TO_INT(ctx->Polygon.SmoothFlag); break; @@ -4585,18 +4588,6 @@ _mesa_GetIntegerv( GLenum pname, GLint *params ) CHECK_EXT1(MESA_texture_array, "GetIntegerv"); params[0] = ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current2DArray->Name; break; - case GL_TEXTURE_ENV_COLOR: - { - const GLfloat *color = ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvColor; - params[0] = FLOAT_TO_INT(color[0]); - params[1] = FLOAT_TO_INT(color[1]); - params[2] = FLOAT_TO_INT(color[2]); - params[3] = FLOAT_TO_INT(color[3]); - } - break; - case GL_TEXTURE_ENV_MODE: - params[0] = ENUM_TO_INT(ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvMode); - break; case GL_TEXTURE_GEN_S: params[0] = BOOLEAN_TO_INT(((ctx->Texture.Unit[ctx->Texture.CurrentUnit].TexGenEnabled & S_BIT) ? 1 : 0)); break; diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py index c307058da6..653deefa72 100644 --- a/src/mesa/main/get_gen.py +++ b/src/mesa/main/get_gen.py @@ -372,6 +372,9 @@ StateVars = [ ( "GL_POLYGON_OFFSET_BIAS_EXT", GLfloat, ["ctx->Polygon.OffsetUnits"], "", None ), ( "GL_POLYGON_OFFSET_FACTOR", GLfloat, ["ctx->Polygon.OffsetFactor "], "", None ), ( "GL_POLYGON_OFFSET_UNITS", GLfloat, ["ctx->Polygon.OffsetUnits "], "", None ), + ( "GL_POLYGON_OFFSET_POINT", GLboolean, ["ctx->Polygon.OffsetPoint"], "", None ), + ( "GL_POLYGON_OFFSET_LINE", GLboolean, ["ctx->Polygon.OffsetLine"], "", None ), + ( "GL_POLYGON_OFFSET_FILL", GLboolean, ["ctx->Polygon.OffsetFill"], "", None ), ( "GL_POLYGON_SMOOTH", GLboolean, ["ctx->Polygon.SmoothFlag"], "", None ), ( "GL_POLYGON_SMOOTH_HINT", GLenum, ["ctx->Hint.PolygonSmooth"], "", None ), ( "GL_POLYGON_STIPPLE", GLboolean, ["ctx->Polygon.StippleFlag"], "", None ), @@ -437,12 +440,6 @@ StateVars = [ ["ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current1DArray->Name"], "", ["MESA_texture_array"] ), ( "GL_TEXTURE_BINDING_2D_ARRAY_EXT", GLint, ["ctx->Texture.Unit[ctx->Texture.CurrentUnit].Current2DArray->Name"], "", ["MESA_texture_array"] ), - ( "GL_TEXTURE_ENV_COLOR", GLfloatN, - ["color[0]", "color[1]", "color[2]", "color[3]"], - "const GLfloat *color = ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvColor;", - None ), - ( "GL_TEXTURE_ENV_MODE", GLenum, - ["ctx->Texture.Unit[ctx->Texture.CurrentUnit].EnvMode"], "", None ), ( "GL_TEXTURE_GEN_S", GLboolean, ["((ctx->Texture.Unit[ctx->Texture.CurrentUnit].TexGenEnabled & S_BIT) ? 1 : 0)"], "", None ), ( "GL_TEXTURE_GEN_T", GLboolean, -- cgit v1.2.3 From 26df49cd7a67aa9dc26e9cf271d6ed121f08f95d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Jul 2008 10:34:50 -0600 Subject: egl: set EGL_SAMPLES, EGL_SAMPLE_BUFFERS config attribs --- src/egl/drivers/xdri/egl_xdri.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/egl/drivers/xdri/egl_xdri.c b/src/egl/drivers/xdri/egl_xdri.c index 2471c0425c..3681d9879a 100644 --- a/src/egl/drivers/xdri/egl_xdri.c +++ b/src/egl/drivers/xdri/egl_xdri.c @@ -203,6 +203,8 @@ create_configs(_EGLDisplay *disp, __GLXdisplayPrivate *glx_priv) SET_CONFIG_ATTRIB(&config->Base, EGL_ALPHA_SIZE, m->alphaBits); SET_CONFIG_ATTRIB(&config->Base, EGL_DEPTH_SIZE, m->depthBits); SET_CONFIG_ATTRIB(&config->Base, EGL_STENCIL_SIZE, m->stencilBits); + SET_CONFIG_ATTRIB(&config->Base, EGL_SAMPLES, m->samples); + SET_CONFIG_ATTRIB(&config->Base, EGL_SAMPLE_BUFFERS, m->sampleBuffers); SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_ID, m->visualID); SET_CONFIG_ATTRIB(&config->Base, EGL_NATIVE_VISUAL_TYPE, m->visualType); SET_CONFIG_ATTRIB(&config->Base, EGL_CONFORMANT, all_apis); -- cgit v1.2.3 From 7a0b79fe3632ca45def56679868473334ef1e194 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 16 Jul 2008 20:14:42 +0200 Subject: mesa: Silence compiler warnings on Windows. --- src/mesa/shader/slang/slang_print.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_print.c b/src/mesa/shader/slang/slang_print.c index bb0da9e34d..f48762fb11 100644 --- a/src/mesa/shader/slang/slang_print.c +++ b/src/mesa/shader/slang/slang_print.c @@ -265,11 +265,8 @@ slang_print_tree(const slang_operation *op, int indent) printf("{{ // new scope locals=%p outer=%p: ", (void *) op->locals, (void *) op->locals->outer_scope); - { - int i; - for (i = 0; i < op->locals->num_variables; i++) { - printf("%s ", (char *) op->locals->variables[i]->a_name); - } + for (i = 0; i < op->locals->num_variables; i++) { + printf("%s ", (char *) op->locals->variables[i]->a_name); } printf("\n"); print_generic(op, NULL, indent+3); -- cgit v1.2.3 From ebcdbff6f1f6fa4749e88f9984b0df784a3d9a0c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Jul 2008 14:27:50 -0600 Subject: mesa: implement grammar/parsing for precision/invariant syntax Plus, fix some issues with pre-defined preprocessor symbols and version checking. --- src/mesa/shader/slang/library/slang_shader.syn | 71 ++++++++++-- src/mesa/shader/slang/library/slang_shader_syn.h | 29 +++++ src/mesa/shader/slang/slang_compile.c | 134 ++++++++++++++++++++++- src/mesa/shader/slang/slang_preprocess.c | 27 +++++ 4 files changed, 251 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_shader.syn b/src/mesa/shader/slang/library/slang_shader.syn index 1764d1ae68..52a792ad17 100644 --- a/src/mesa/shader/slang/library/slang_shader.syn +++ b/src/mesa/shader/slang/library/slang_shader.syn @@ -51,10 +51,17 @@ /* revision number - increment after each change affecting emitted output */ .emtcode REVISION 3 -/* external declaration */ +/* external declaration (or precision or invariant stmt) */ .emtcode EXTERNAL_NULL 0 .emtcode EXTERNAL_FUNCTION_DEFINITION 1 .emtcode EXTERNAL_DECLARATION 2 +.emtcode DEFAULT_PRECISION 3 +.emtcode INVARIANT_STMT 4 + +/* precision */ +.emtcode PRECISION_LOW 0 +.emtcode PRECISION_MEDIUM 1 +.emtcode PRECISION_HIGH 2 /* declaration */ .emtcode DECLARATION_FUNCTION_PROTOTYPE 1 @@ -240,6 +247,9 @@ .errtext LBRACE_EXPECTED "2003: '{' expected but '$err_token$' found." .errtext LPAREN_EXPECTED "2004: '(' expected but '$err_token$' found." .errtext RPAREN_EXPECTED "2005: ')' expected but '$err_token$' found." +.errtext INVALID_PRECISION "2006: Invalid precision specifier '$err_token$'." +.errtext INVALID_PRECISION_TYPE "2007: Invalid precision type '$err_token$'." + /* tells whether the shader that is being parsed is a built-in shader or not */ /* 0 - normal behaviour */ @@ -1226,23 +1236,70 @@ asm_argument var_with_field variable_identifier .and dot .and field_selection .emit OP_FIELD; + /* - ::= - | -*/ + * ::= + * | + */ translation_unit optional_space .emit REVISION .and external_declaration .error INVALID_EXTERNAL_DECLARATION .and .loop external_declaration .and optional_space .and '\0' .error INVALID_EXTERNAL_DECLARATION .emit EXTERNAL_NULL; + /* - ::= - | -*/ + * ::= + * | + */ external_declaration + precision_stmt .emit DEFAULT_PRECISION .or + invariant_stmt .emit INVARIANT_STMT .or function_definition .emit EXTERNAL_FUNCTION_DEFINITION .or declaration .emit EXTERNAL_DECLARATION; + +/* + * ::= "precision" + */ +precision_stmt + "precision" .and space .and precision .error INVALID_PRECISION .and space .and prectype .error INVALID_PRECISION_TYPE .and semicolon; + +/* + * ::= "lowp" + * | "mediump" + * | "highp" + */ +precision + "lowp" .emit PRECISION_LOW .or + "mediump" .emit PRECISION_MEDIUM .or + "highp" .emit PRECISION_HIGH; + +/* + * ::= "int" + * | "float" + * | "a sampler type" + */ +prectype + "int" .emit TYPE_SPECIFIER_INT .or + "float" .emit TYPE_SPECIFIER_FLOAT .or + "sampler1D" .emit TYPE_SPECIFIER_SAMPLER1D .or + "sampler2D" .emit TYPE_SPECIFIER_SAMPLER2D .or + "sampler3D" .emit TYPE_SPECIFIER_SAMPLER3D .or + "samplerCube" .emit TYPE_SPECIFIER_SAMPLERCUBE .or + "sampler1DShadow" .emit TYPE_SPECIFIER_SAMPLER1DSHADOW .or + "sampler2DShadow" .emit TYPE_SPECIFIER_SAMPLER2DSHADOW .or + "sampler2DRect" .emit TYPE_SPECIFIER_SAMPLER2DRECT .or + "sampler2DRectShadow" .emit TYPE_SPECIFIER_SAMPLER2DRECTSHADOW; + + +/* + * ::= "invariant" identifier; + */ +invariant_stmt + "invariant" .and space .and identifier .and semicolon; + + + /* :: */ diff --git a/src/mesa/shader/slang/library/slang_shader_syn.h b/src/mesa/shader/slang/library/slang_shader_syn.h index fe37adcac4..c0fef6e4e3 100644 --- a/src/mesa/shader/slang/library/slang_shader_syn.h +++ b/src/mesa/shader/slang/library/slang_shader_syn.h @@ -6,6 +6,11 @@ ".emtcode EXTERNAL_NULL 0\n" ".emtcode EXTERNAL_FUNCTION_DEFINITION 1\n" ".emtcode EXTERNAL_DECLARATION 2\n" +".emtcode DEFAULT_PRECISION 3\n" +".emtcode INVARIANT_STMT 4\n" +".emtcode PRECISION_LOW 0\n" +".emtcode PRECISION_MEDIUM 1\n" +".emtcode PRECISION_HIGH 2\n" ".emtcode DECLARATION_FUNCTION_PROTOTYPE 1\n" ".emtcode DECLARATION_INIT_DECLARATOR_LIST 2\n" ".emtcode FUNCTION_ORDINARY 0\n" @@ -41,6 +46,7 @@ ".emtcode TYPE_QUALIFIER_UNIFORM 4\n" ".emtcode TYPE_QUALIFIER_FIXEDOUTPUT 5\n" ".emtcode TYPE_QUALIFIER_FIXEDINPUT 6\n" +".emtcode TYPE_QUALIFIER_INVARIANT_VARYING 7\n" ".emtcode TYPE_SPECIFIER_VOID 0\n" ".emtcode TYPE_SPECIFIER_BOOL 1\n" ".emtcode TYPE_SPECIFIER_BVEC2 2\n" @@ -137,6 +143,8 @@ ".errtext LBRACE_EXPECTED \"2003: '{' expected but '$err_token$' found.\"\n" ".errtext LPAREN_EXPECTED \"2004: '(' expected but '$err_token$' found.\"\n" ".errtext RPAREN_EXPECTED \"2005: ')' expected but '$err_token$' found.\"\n" +".errtext INVALID_PRECISION \"2006: Invalid precision specifier '$err_token$'.\"\n" +".errtext INVALID_PRECISION_TYPE \"2007: Invalid precision type '$err_token$'.\"\n" ".regbyte parsing_builtin 0\n" ".regbyte shader_type 0\n" "variable_identifier\n" @@ -597,8 +605,29 @@ " .loop external_declaration .and optional_space .and\n" " '\\0' .error INVALID_EXTERNAL_DECLARATION .emit EXTERNAL_NULL;\n" "external_declaration\n" +" precision_stmt .emit DEFAULT_PRECISION .or\n" +" invariant_stmt .emit INVARIANT_STMT .or\n" " function_definition .emit EXTERNAL_FUNCTION_DEFINITION .or\n" " declaration .emit EXTERNAL_DECLARATION;\n" +"precision_stmt\n" +" \"precision\" .and space .and precision .error INVALID_PRECISION .and space .and prectype .error INVALID_PRECISION_TYPE .and semicolon;\n" +"precision\n" +" \"lowp\" .emit PRECISION_LOW .or\n" +" \"mediump\" .emit PRECISION_MEDIUM .or\n" +" \"highp\" .emit PRECISION_HIGH;\n" +"prectype\n" +" \"int\" .emit TYPE_SPECIFIER_INT .or\n" +" \"float\" .emit TYPE_SPECIFIER_FLOAT .or\n" +" \"sampler1D\" .emit TYPE_SPECIFIER_SAMPLER1D .or\n" +" \"sampler2D\" .emit TYPE_SPECIFIER_SAMPLER2D .or\n" +" \"sampler3D\" .emit TYPE_SPECIFIER_SAMPLER3D .or\n" +" \"samplerCube\" .emit TYPE_SPECIFIER_SAMPLERCUBE .or\n" +" \"sampler1DShadow\" .emit TYPE_SPECIFIER_SAMPLER1DSHADOW .or\n" +" \"sampler2DShadow\" .emit TYPE_SPECIFIER_SAMPLER2DSHADOW .or\n" +" \"sampler2DRect\" .emit TYPE_SPECIFIER_SAMPLER2DRECT .or\n" +" \"sampler2DRectShadow\" .emit TYPE_SPECIFIER_SAMPLER2DRECTSHADOW;\n" +"invariant_stmt\n" +" \"invariant\" .and space .and identifier .and semicolon;\n" "function_definition\n" " function_prototype .and compound_statement_no_new_scope;\n" "digit_oct\n" diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 29c5ca2f4b..fd5d4afacb 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -56,6 +56,9 @@ */ +/** re-defined below, should be the same though */ +#define TYPE_SPECIFIER_COUNT 32 + /** * Allocate storage for a variable of 'size' bytes from given pool. @@ -129,6 +132,7 @@ typedef struct slang_parse_ctx_ GLboolean global_scope; /**< Is object being declared a global? */ slang_atom_pool *atoms; slang_unit_type type; /**< Vertex vs. Fragment */ + GLuint version; /**< user-specified (or default) #version */ } slang_parse_ctx; /* slang_output_ctx */ @@ -141,6 +145,7 @@ typedef struct slang_output_ctx_ slang_var_pool *global_pool; struct gl_program *program; slang_var_table *vartable; + GLuint default_precision[TYPE_SPECIFIER_COUNT]; } slang_output_ctx; /* _slang_compile() */ @@ -525,7 +530,7 @@ parse_type_qualifier(slang_parse_ctx * C, slang_type_qualifier * qual) #define TYPE_SPECIFIER_MAT42 29 #define TYPE_SPECIFIER_MAT34 30 #define TYPE_SPECIFIER_MAT43 31 - +#define TYPE_SPECIFIER_COUNT 32 static int parse_type_specifier(slang_parse_ctx * C, slang_output_ctx * O, @@ -1817,10 +1822,111 @@ parse_declaration(slang_parse_ctx * C, slang_output_ctx * O) return 1; } -/* external declaration */ + +#define PRECISION_LOW 0 +#define PRECISION_MEDIUM 1 +#define PRECISION_HIGH 2 + +static int +parse_default_precision(slang_parse_ctx * C, slang_output_ctx * O) +{ +#if FEATURE_es2_glsl + int precision, type; + + precision = *C->I++; + switch (precision) { + case PRECISION_LOW: + case PRECISION_MEDIUM: + case PRECISION_HIGH: + /* OK */ + break; + default: + _mesa_problem(NULL, "unexpected precision %d at %s:%d\n", + precision, __FILE__, __LINE__); + return 0; + } + + type = *C->I++; + switch (type) { + case TYPE_SPECIFIER_FLOAT: + case TYPE_SPECIFIER_INT: + case TYPE_SPECIFIER_SAMPLER1D: + case TYPE_SPECIFIER_SAMPLER2D: + case TYPE_SPECIFIER_SAMPLER3D: + case TYPE_SPECIFIER_SAMPLERCUBE: + case TYPE_SPECIFIER_SAMPLER1DSHADOW: + case TYPE_SPECIFIER_SAMPLER2DSHADOW: + case TYPE_SPECIFIER_SAMPLER2DRECT: + case TYPE_SPECIFIER_SAMPLER2DRECTSHADOW: + /* OK */ + break; + default: + _mesa_problem(NULL, "unexpected type %d at %s:%d\n", + type, __FILE__, __LINE__); + return 0; + } + + assert(type < TYPE_SPECIFIER_COUNT); + O->default_precision[type] = precision; + + return 1; +#else + slang_info_log_error(C->L, "syntax error at \"precision\""); + return 0; +#endif +} + + +/** + * Initialize the default precision for all types. + * XXX this info isn't used yet. + */ +static void +init_default_precision(slang_output_ctx *O, slang_unit_type type) +{ + GLuint i; + for (i = 0; i < TYPE_SPECIFIER_COUNT; i++) { +#if FEATURE_es2_glsl + O->default_precision[i] = PRECISION_LOW; +#else + O->default_precision[i] = PRECISION_HIGH; +#endif + } +#if FEATURE_es2_glsl + if (type == SLANG_UNIT_VERTEX_SHADER) { + O->default_precision[TYPE_SPECIFIER_FLOAT] = PRECISION_HIGH; + O->default_precision[TYPE_SPECIFIER_INT] = PRECISION_HIGH; + } + else { + O->default_precision[TYPE_SPECIFIER_INT] = PRECISION_MEDIUM; + } +#endif +} + + +static int +parse_invariant(slang_parse_ctx * C, slang_output_ctx * O) +{ + if (C->version >= 120 || FEATURE_es2_glsl) { + slang_atom *a = parse_identifier(C); + /* XXX not doing anything with this var yet */ + /*printf("ID: %s\n", (char*) a);*/ + return a ? 1 : 0; + } + else { + slang_info_log_error(C->L, "syntax error at \"invariant\""); + return 0; + } +} + + +/* external declaration or default precision specifier */ #define EXTERNAL_NULL 0 #define EXTERNAL_FUNCTION_DEFINITION 1 #define EXTERNAL_DECLARATION 2 +#define DEFAULT_PRECISION 3 +#define INVARIANT_STMT 4 + static GLboolean parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, @@ -1843,6 +1949,7 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, } /* setup output context */ + init_default_precision(&o, unit->type); o.funs = &unit->funs; o.structs = &unit->structs; o.vars = &unit->vars; @@ -1868,6 +1975,12 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, case EXTERNAL_DECLARATION: success = parse_declaration(C, &o); break; + case DEFAULT_PRECISION: + success = parse_default_precision(C, &o); + break; + case INVARIANT_STMT: + success = parse_invariant(C, &o); + break; default: success = GL_FALSE; } @@ -1904,6 +2017,7 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, static GLboolean compile_binary(const byte * prod, slang_code_unit * unit, + GLuint version, slang_unit_type type, slang_info_log * infolog, slang_code_unit * builtin, slang_code_unit * downlink, struct gl_program *program) @@ -1919,6 +2033,7 @@ compile_binary(const byte * prod, slang_code_unit * unit, C.global_scope = GL_TRUE; C.atoms = &unit->object->atompool; C.type = type; + C.version = version; if (!check_revision(&C)) return GL_FALSE; @@ -1946,6 +2061,8 @@ compile_with_grammar(grammar id, const char *source, slang_code_unit * unit, #if FEATURE_ARB_shading_language_120 maxVersion = 120; +#elif FEATURE_es2_glsl + maxVersion = 100; #else maxVersion = 110; #endif @@ -1995,7 +2112,7 @@ compile_with_grammar(grammar id, const char *source, slang_code_unit * unit, slang_string_free(&preprocessed); /* Syntax is okay - translate it to internal representation. */ - if (!compile_binary(prod, unit, type, infolog, builtin, + if (!compile_binary(prod, unit, version, type, infolog, builtin, &builtin[SLANG_BUILTIN_TOTAL - 1], program)) { grammar_alloc_free(prod); @@ -2039,6 +2156,7 @@ compile_object(grammar * id, const char *source, slang_code_object * object, struct gl_program *program) { slang_code_unit *builtins = NULL; + GLuint base_version = 110; /* load GLSL grammar */ *id = grammar_load_from_text((const byte *) (slang_shader_syn)); @@ -2066,6 +2184,7 @@ compile_object(grammar * id, const char *source, slang_code_object * object, /* compile core functionality first */ if (!compile_binary(slang_core_gc, &object->builtin[SLANG_BUILTIN_CORE], + base_version, SLANG_UNIT_FRAGMENT_BUILTIN, infolog, NULL, NULL, NULL)) return GL_FALSE; @@ -2073,6 +2192,7 @@ compile_object(grammar * id, const char *source, slang_code_object * object, #if FEATURE_ARB_shading_language_120 if (!compile_binary(slang_120_core_gc, &object->builtin[SLANG_BUILTIN_120_CORE], + 120, SLANG_UNIT_FRAGMENT_BUILTIN, infolog, NULL, &object->builtin[SLANG_BUILTIN_CORE], NULL)) return GL_FALSE; @@ -2081,6 +2201,11 @@ compile_object(grammar * id, const char *source, slang_code_object * object, /* compile common functions and variables, link to core */ if (!compile_binary(slang_common_builtin_gc, &object->builtin[SLANG_BUILTIN_COMMON], +#if FEATURE_ARB_shading_language_120 + 120, +#else + base_version, +#endif SLANG_UNIT_FRAGMENT_BUILTIN, infolog, NULL, #if FEATURE_ARB_shading_language_120 &object->builtin[SLANG_BUILTIN_120_CORE], @@ -2094,12 +2219,14 @@ compile_object(grammar * id, const char *source, slang_code_object * object, if (type == SLANG_UNIT_FRAGMENT_SHADER) { if (!compile_binary(slang_fragment_builtin_gc, &object->builtin[SLANG_BUILTIN_TARGET], + base_version, SLANG_UNIT_FRAGMENT_BUILTIN, infolog, NULL, &object->builtin[SLANG_BUILTIN_COMMON], NULL)) return GL_FALSE; #if FEATURE_ARB_shading_language_120 if (!compile_binary(slang_120_fragment_gc, &object->builtin[SLANG_BUILTIN_TARGET], + 120, SLANG_UNIT_FRAGMENT_BUILTIN, infolog, NULL, &object->builtin[SLANG_BUILTIN_COMMON], NULL)) return GL_FALSE; @@ -2108,6 +2235,7 @@ compile_object(grammar * id, const char *source, slang_code_object * object, else if (type == SLANG_UNIT_VERTEX_SHADER) { if (!compile_binary(slang_vertex_builtin_gc, &object->builtin[SLANG_BUILTIN_TARGET], + base_version, SLANG_UNIT_VERTEX_BUILTIN, infolog, NULL, &object->builtin[SLANG_BUILTIN_COMMON], NULL)) return GL_FALSE; diff --git a/src/mesa/shader/slang/slang_preprocess.c b/src/mesa/shader/slang/slang_preprocess.c index 1645afcc18..d3f412c211 100644 --- a/src/mesa/shader/slang/slang_preprocess.c +++ b/src/mesa/shader/slang/slang_preprocess.c @@ -729,6 +729,14 @@ expand (expand_state *e, pp_symbols *symbols) slang_string_pushi (e->output, e->state->version); slang_string_pushc (e->output, ' '); } +#if FEATURE_es2_glsl + else if (_mesa_strcmp (id, "GL_ES") == 0 || + _mesa_strcmp (id, "GL_FRAGMENT_PRECISION_HIGH") == 0) { + slang_string_pushc (e->output, ' '); + slang_string_pushi (e->output, '1'); + slang_string_pushc (e->output, ' '); + } +#endif else { pp_symbol *symbol; @@ -829,6 +837,16 @@ static GLboolean preprocess_source (slang_string *output, const char *source, grammar pid, grammar eid, slang_info_log *elog) { + static const char *predefined[] = { + "__FILE__", + "__LINE__", + "__VERSION__", +#if FEATURE_es2_glsl + "GL_ES", + "GL_FRAGMENT_PRECISION_HIGH", +#endif + NULL + }; byte *prod; GLuint size, i; pp_state state; @@ -840,6 +858,15 @@ preprocess_source (slang_string *output, const char *source, grammar pid, gramma pp_state_init (&state, elog); + /* add the predefined symbols to the symbol table */ + for (i = 0; predefined[i]; i++) { + pp_symbol *symbol = NULL; + symbol = pp_symbols_push(&state.symbols); + assert(symbol); + slang_string_pushs(&symbol->name, + predefined[i], _mesa_strlen(predefined[i])); + } + i = 0; while (i < size) { if (prod[i] != ESCAPE_TOKEN) { -- cgit v1.2.3 From bd09676de5ad0db97159046d94e2ca72317059de Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Jul 2008 14:47:21 -0600 Subject: mesa: fix copy&paste errors in degrees() functions --- src/mesa/shader/slang/library/slang_common_builtin.gc | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index f741c657e4..b095a6434b 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -203,19 +203,19 @@ float degrees(const float rad) vec2 degrees(const vec2 rad) { - const float c = 3.1415926 / 180.0; + const float c = 180.0 / 3.1415926; __asm vec4_multiply __retVal.xy, rad.xy, c.xx; } vec3 degrees(const vec3 rad) { - const float c = 3.1415926 / 180.0; + const float c = 180.0 / 3.1415926; __asm vec4_multiply __retVal.xyz, rad.xyz, c.xxx; } vec4 degrees(const vec4 rad) { - const float c = 3.1415926 / 180.0; + const float c = 180.0 / 3.1415926; __asm vec4_multiply __retVal, rad, c.xxxx; } -- cgit v1.2.3 From a1e55dcd9c1287f9a2181b6136d8febae3a4d74c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Jul 2008 14:48:21 -0600 Subject: mesa: regenerated files --- .../shader/slang/library/slang_common_builtin_gc.h | 36 +++++++++++----------- src/mesa/shader/slang/library/slang_shader_syn.h | 1 - 2 files changed, 18 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index 260a768de6..a8698712c5 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -116,25 +116,25 @@ 101,103,114,101,101,115,0,1,1,0,9,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0, 49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, 101,116,86,97,108,0,59,120,0,0,18,114,97,100,0,0,18,99,0,0,0,0,1,0,10,0,100,101,103,114,101,101, -115,0,1,1,0,10,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0, -48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, +115,0,1,1,0,10,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50, +54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, 0,59,120,121,0,0,18,114,97,100,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,0,11,0,100,101,103, -114,101,101,115,0,1,1,0,11,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17, -49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,114,97,100,0,59,120,121,122,0,0,18,99,0,59,120,120,120,0,0,0,0,1, -0,12,0,100,101,103,114,101,101,115,0,1,1,0,12,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49, -53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, -18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,59,120,120,120,120,0,0,0,0,1,0,9,0, -115,105,110,0,1,1,0,9,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,0,10,0,115,105, -110,0,1,1,0,10,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95, -95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97, -116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115, -0,59,121,0,0,0,0,1,0,11,0,115,105,110,0,1,1,0,11,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111, -97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110, -115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,0,12,0, +114,101,101,115,0,1,1,0,11,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52, +49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,114,97,100,0,59,120,121,122,0,0,18,99,0,59,120,120,120,0,0,0, +0,1,0,12,0,100,101,103,114,101,101,115,0,1,1,0,12,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48, +0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,59,120,120,120,120,0,0,0,0,1,0, +9,0,115,105,110,0,1,1,0,9,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,0,10,0, +115,105,110,0,1,1,0,10,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0, +18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108, +111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97, +110,115,0,59,121,0,0,0,0,1,0,11,0,115,105,110,0,1,1,0,11,114,97,100,105,97,110,115,0,0,0,1,4,102, +108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105, +97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108, +0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0, +18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,0,12,0, 115,105,110,0,1,1,0,12,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0, 18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108, 111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97, diff --git a/src/mesa/shader/slang/library/slang_shader_syn.h b/src/mesa/shader/slang/library/slang_shader_syn.h index c0fef6e4e3..82909d9423 100644 --- a/src/mesa/shader/slang/library/slang_shader_syn.h +++ b/src/mesa/shader/slang/library/slang_shader_syn.h @@ -46,7 +46,6 @@ ".emtcode TYPE_QUALIFIER_UNIFORM 4\n" ".emtcode TYPE_QUALIFIER_FIXEDOUTPUT 5\n" ".emtcode TYPE_QUALIFIER_FIXEDINPUT 6\n" -".emtcode TYPE_QUALIFIER_INVARIANT_VARYING 7\n" ".emtcode TYPE_SPECIFIER_VOID 0\n" ".emtcode TYPE_SPECIFIER_BOOL 1\n" ".emtcode TYPE_SPECIFIER_BVEC2 2\n" -- cgit v1.2.3 From ab8f838060c7d30b8b18cac600c4b1d97ecf3f6c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 16 Jul 2008 16:04:43 -0600 Subject: mesa: fix temp re-use bug in emit_arith() --- src/mesa/shader/slang/slang_emit.c | 28 ++++++++++++++++++++-------- 1 file changed, 20 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index f025b15de9..c997c11a7d 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -469,6 +469,12 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) const slang_ir_info *info = _slang_ir_info(n->Opcode); char *srcAnnot[3], *dstAnnot; GLuint i; + slang_ir_node *temps[3]; + + /* we'll save pointers to nodes/storage to free in temps[] until + * the very end. + */ + temps[0] = temps[1] = temps[2] = NULL; assert(info); assert(info->InstOpcode != OPCODE_NOP); @@ -489,9 +495,9 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) storage_to_src_reg(&inst->SrcReg[0], n->Children[0]->Children[0]->Store); storage_to_src_reg(&inst->SrcReg[1], n->Children[0]->Children[1]->Store); storage_to_src_reg(&inst->SrcReg[2], n->Children[1]->Store); - free_temp_storage(emitInfo->vt, n->Children[0]->Children[0]); - free_temp_storage(emitInfo->vt, n->Children[0]->Children[1]); - free_temp_storage(emitInfo->vt, n->Children[1]); + temps[0] = n->Children[0]->Children[0]; + temps[1] = n->Children[0]->Children[1]; + temps[2] = n->Children[1]; } else if (info->NumParams == 2 && n->Opcode == IR_ADD && n->Children[1]->Opcode == IR_MUL) { @@ -505,9 +511,9 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) storage_to_src_reg(&inst->SrcReg[0], n->Children[1]->Children[0]->Store); storage_to_src_reg(&inst->SrcReg[1], n->Children[1]->Children[1]->Store); storage_to_src_reg(&inst->SrcReg[2], n->Children[0]->Store); - free_temp_storage(emitInfo->vt, n->Children[1]->Children[0]); - free_temp_storage(emitInfo->vt, n->Children[1]->Children[1]); - free_temp_storage(emitInfo->vt, n->Children[0]); + temps[0] = n->Children[1]->Children[0]; + temps[1] = n->Children[1]->Children[1]; + temps[2] = n->Children[0]; } else #endif @@ -532,9 +538,9 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) for (i = 0; i < info->NumParams; i++) srcAnnot[i] = storage_annotation(n->Children[i], emitInfo->prog); - /* free temps */ + /* record (potential) temps to free */ for (i = 0; i < info->NumParams; i++) - free_temp_storage(emitInfo->vt, n->Children[i]); + temps[i] = n->Children[i]; } /* result storage */ @@ -544,6 +550,7 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) if (!alloc_temp_storage(emitInfo, n, size)) return NULL; } + storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); dstAnnot = storage_annotation(n, emitInfo->prog); @@ -551,6 +558,11 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) inst->Comment = instruction_annotation(inst->Opcode, dstAnnot, srcAnnot[0], srcAnnot[1], srcAnnot[2]); + /* really free temps now */ + for (i = 0; i < 3; i++) + if (temps[i]) + free_temp_storage(emitInfo->vt, temps[i]); + /*_mesa_print_instruction(inst);*/ return inst; } -- cgit v1.2.3 From f610deed37dc09922eaef1c55e3d4afe8aaa7327 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 17 Jul 2008 08:21:58 +0900 Subject: python/tests: Cover all last_level combinations. --- src/gallium/state_trackers/python/tests/texture.py | 43 +++++++++++++--------- 1 file changed, 25 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index 448bdbcd3e..edaa2b8b26 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -135,7 +135,11 @@ class TextureTest(TestCase): }[self.face] else: face = "" - return "%s %s %ux%u %s level=%u" % (target, format, self.width, self.height, face, self.level) + return "%s %s %ux%u last_level=%u %s level=%u" % ( + target, format, + self.width, self.height, + self.last_level, face, self.level, + ) def test(self): dev = self.dev @@ -144,11 +148,10 @@ class TextureTest(TestCase): format = self.format width = self.width height = self.height + last_level = self.last_level level = self.level face = self.face - levels = lods(width, height) - if not dev.is_format_supported(format, PIPE_TEXTURE): raise TestSkip @@ -209,14 +212,15 @@ class TextureTest(TestCase): format=format, width=width, height=height, - last_level = levels - 1) - ctx.set_sampler_texture(0, texture) + last_level = last_level) expected_rgba = generate_data(texture.get_surface( usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE, face = face, level = level)) + ctx.set_sampler_texture(0, texture) + # framebuffer cbuf_tex = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, width, @@ -332,8 +336,8 @@ def main(): formats += [PIPE_FORMAT_YCBCR] formats += [PIPE_FORMAT_DXT1_RGB] - sizes = [64, 32, 16, 8, 4, 2] - #sizes = [16] + sizes = [64, 32, 16, 8, 4, 2, 1] + #sizes = [64] for target in targets: for format in formats: @@ -347,21 +351,24 @@ def main(): PIPE_TEX_FACE_POS_Z, PIPE_TEX_FACE_NEG_Z, ] + #faces = [PIPE_TEX_FACE_NEG_X] else: faces = [0] for face in faces: levels = lods(size) - for level in range(levels): - test = TextureTest( - dev=dev, - target=target, - format=format, - width=size, - height=size, - face=face, - level=level, - ) - suite.add_test(test) + for last_level in range(levels): + for level in range(0, last_level + 1): + test = TextureTest( + dev=dev, + target=target, + format=format, + width=size, + height=size, + face=face, + last_level = last_level, + level=level, + ) + suite.add_test(test) suite.run() -- cgit v1.2.3 From 8336fbc35c990f6bce0ee997acae8652cd44fd97 Mon Sep 17 00:00:00 2001 From: Brad Smith Date: Thu, 17 Jul 2008 08:16:19 -0600 Subject: mesa: added checks for OpenBSD --- src/mesa/x86/common_x86.c | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) (limited to 'src') diff --git a/src/mesa/x86/common_x86.c b/src/mesa/x86/common_x86.c index d93241a977..0caa36a5a0 100644 --- a/src/mesa/x86/common_x86.c +++ b/src/mesa/x86/common_x86.c @@ -42,6 +42,11 @@ #include #include #endif +#if defined(USE_SSE_ASM) && defined(__OpenBSD__) +#include +#include +#include +#endif #include "common_x86_asm.h" #include "imports.h" @@ -121,6 +126,19 @@ static void check_os_sse_support( void ) if (ret || !enabled) _mesa_x86_cpu_features &= ~(X86_FEATURE_XMM); } +#elif defined(__OpenBSD__) + { + int mib[2]; + int ret, enabled; + size_t len = sizeof(enabled); + + mib[0] = CTL_MACHDEP; + mib[1] = CPU_SSE; + + ret = sysctl(mib, 2, &enabled, &len, NULL, 0); + if (ret || !enabled) + _mesa_x86_cpu_features &= ~(X86_FEATURE_XMM); + } #elif defined(WIN32) LPTOP_LEVEL_EXCEPTION_FILTER oldFilter; -- cgit v1.2.3 From 91de7b6cd30ad5ec00dd94ea4a56ec0c9056948f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 17 Jul 2008 10:03:10 -0600 Subject: mesa: fix/improve the atan(y,x) function --- .../shader/slang/library/slang_common_builtin.gc | 21 +++++++++++---------- 1 file changed, 11 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index b095a6434b..3726335471 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -401,16 +401,17 @@ vec4 atan(const vec4 y_over_x) float atan(const float y, const float x) { - if (x == 0.0) - return 0.0; - float z = atan(y / x); - if (x < 0.0) - { - if (y < 0.0) - return z - 3.141593; - return z + 3.141593; - } - return z; + float r; + if (abs(x) > 1.0e-4) { + r = atan(y / x); + if (x < 0.0) { + r = r + sign(y) * 3.141593; + } + } + else { + r = sign(y) * 1.5707965; // pi/2 + } + return r; } vec2 atan(const vec2 u, const vec2 v) -- cgit v1.2.3 From 457bb10cee6a2f6f7b7e320f066a26e24e13c550 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 17 Jul 2008 10:03:25 -0600 Subject: mesa: regenerated file --- .../shader/slang/library/slang_common_builtin_gc.h | 965 +++++++++++---------- 1 file changed, 483 insertions(+), 482 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index a8698712c5..0a43cad2a3 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -210,500 +210,501 @@ 121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111,118, 101,114,95,120,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0, 18,121,95,111,118,101,114,95,120,0,59,119,0,0,0,20,0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,121,0,0,1,1, -0,9,120,0,0,0,1,10,18,120,0,17,48,0,48,0,0,38,0,8,17,48,0,48,0,0,0,9,14,0,3,2,0,9,1,122,0,2,58,97, -116,97,110,0,18,121,0,18,120,0,49,0,0,0,0,10,18,120,0,17,48,0,48,0,0,40,0,2,10,18,121,0,17,48,0,48, -0,0,40,0,8,18,122,0,17,51,0,49,52,49,53,57,51,0,0,47,0,9,14,0,8,18,122,0,17,51,0,49,52,49,53,57,51, -0,0,46,0,0,9,14,0,8,18,122,0,0,0,1,0,10,0,97,116,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0, -0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118, -0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121, -0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18, -118,0,59,122,0,0,0,20,0,0,1,0,12,0,97,116,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0, -59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0, -0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,18,117, -0,59,119,0,0,18,118,0,59,119,0,0,0,20,0,0,1,0,9,0,112,111,119,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1, -4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0, -0,18,98,0,0,0,0,1,0,10,0,112,111,119,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,102,108,111,97,116,95, -112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59, -120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0, -0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,0,1,0,11,0,112,111,119,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0, -0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114, -101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112, -111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0, -0,0,0,1,0,12,0,112,111,119,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,102,108,111,97,116,95,112,111, -119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0, -4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0, -59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116, -86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,112,111,119, -101,114,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,18,98,0,59,119,0,0,0,0,1, -0,9,0,101,120,112,0,1,1,0,9,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108, -111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,0,18,101,0,0,18,97,0,0,0,0,1, -0,10,0,101,120,112,0,1,1,0,10,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102, -108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18, -97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0, -59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,0,1,0,11,0,101,120,112,0,1,1,0,11,97,0,0,0,1,3,2,1,9,1, -101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119, -101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,4,102,108, -111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18,97,0, -59,122,0,0,0,0,1,0,12,0,101,120,112,0,1,1,0,12,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56, +0,9,120,0,0,0,1,3,2,0,9,1,114,0,0,0,10,58,97,98,115,0,18,120,0,0,0,17,49,0,48,0,45,52,0,41,0,2,9, +18,114,0,58,97,116,97,110,0,18,121,0,18,120,0,49,0,0,20,0,10,18,120,0,17,48,0,48,0,0,40,0,2,9,18, +114,0,18,114,0,58,115,105,103,110,0,18,121,0,0,0,17,51,0,49,52,49,53,57,51,0,0,48,46,20,0,0,9,14,0, +0,2,9,18,114,0,58,115,105,103,110,0,18,121,0,0,0,17,49,0,53,55,48,55,57,54,53,0,0,48,20,0,0,8,18, +114,0,0,0,1,0,10,0,97,116,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,0,1, +0,11,0,97,116,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,0,1, +0,12,0,97,116,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,18,117,0,59,119,0,0,18,118,0,59,119,0, +0,0,20,0,0,1,0,9,0,112,111,119,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,102,108,111,97,116,95,112,111, +119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,112,111, +119,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95, +114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95, +112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59, +121,0,0,0,0,1,0,11,0,112,111,119,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,102,108,111,97,116,95,112, +111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0, +0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18, +97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114, +101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,0,1,0,12,0,112,111,119,0,1, +1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101, +116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111, +119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0, +4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0, +59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116, +86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,18,98,0,59,119,0,0,0,0,1,0,9,0,101,120,112,0,1,1,0,9,97, +0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101, +114,0,18,95,95,114,101,116,86,97,108,0,0,18,101,0,0,18,97,0,0,0,0,1,0,10,0,101,120,112,0,1,1,0,10, +97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119, +101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4,102,108, +111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0,18,97,0, +59,121,0,0,0,0,1,0,11,0,101,120,112,0,1,1,0,11,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56, 0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, 18,101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116, 86,97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114, -0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116, -95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,101,0,0,18,97,0,59,119,0,0, -0,0,1,0,9,0,108,111,103,50,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95, -114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,0,0,1,0,10,0,108,111,103,50,0,1,1,0,10,118,0,0,0,1,4, -102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120, -0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118, -0,59,121,0,0,0,0,1,0,11,0,108,111,103,50,0,1,1,0,11,118,0,0,0,1,4,102,108,111,97,116,95,108,111, -103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95, -108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111, -97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,0, -12,0,108,111,103,50,0,1,1,0,12,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18, -95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103, -50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,108, -111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,9,0,108,111, -103,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111, -103,50,0,18,120,0,0,0,18,99,0,48,0,0,1,0,10,0,108,111,103,0,1,1,0,10,118,0,0,0,1,3,2,1,9,1,99,0,2, -17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0, -11,0,108,111,103,0,1,1,0,11,118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0, -0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,12,0,108,111,103,0,1,1,0,12,118,0,0,0,1,3, -2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99, -0,48,0,0,1,0,9,0,101,120,112,50,0,1,1,0,9,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18, -95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,0,101,120,112,50,0,1,1,0,10,97,0,0,0, -1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59, -120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18, -97,0,59,121,0,0,0,0,1,0,11,0,101,120,112,50,0,1,1,0,11,97,0,0,0,1,4,102,108,111,97,116,95,101,120, -112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95, -101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97, -116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,0,1,0,12, -0,101,120,112,50,0,1,1,0,12,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101, -116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95, -114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0, -18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120, -112,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,0,0,1,0,9,0,115,113,114, -116,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18, +0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18,97,0,59,122,0,0,0,0,1,0,12,0,101,120, +112,0,1,1,0,12,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95, +112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0, +4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0, +0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97, +108,0,59,122,0,0,18,101,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18, +95,95,114,101,116,86,97,108,0,59,119,0,0,18,101,0,0,18,97,0,59,119,0,0,0,0,1,0,9,0,108,111,103,50, +0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,120,0,0,0,0,1,0,10,0,108,111,103,50,0,1,1,0,10,118,0,0,0,1,4,102,108,111,97,116,95,108, +111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97, +116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,0,11, +0,108,111,103,50,0,1,1,0,11,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101, +116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95, +114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0, +18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,0,12,0,108,111,103,50,0,1,1, +0,12,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120, +0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108, +0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116, +86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114, +101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,9,0,108,111,103,0,1,1,0,9,120,0,0,0,1,3, +2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,120,0,0,0,18,99, +0,48,0,0,1,0,10,0,108,111,103,0,1,1,0,10,118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49, +56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,11,0,108,111,103,0,1,1,0,11, +118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18, +118,0,0,0,18,99,0,48,0,0,1,0,12,0,108,111,103,0,1,1,0,12,118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54, +57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,9,0,101,120, +112,50,0,1,1,0,9,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97, +108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,0,101,120,112,50,0,1,1,0,10,97,0,0,0,1,4,102,108,111,97,116, +95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111, +97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,0,1,0, +11,0,101,120,112,50,0,1,1,0,11,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114, +101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95, +95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50, +0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,0,1,0,12,0,101,120,112,50,0,1,1, +0,12,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0, +0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0, +59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86, +97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101, +116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,0,0,1,0,9,0,115,113,114,116,0,1,1,0,9,120,0,0,0,1,3, +2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,120,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,0,1,0,10,0,115,113, +114,116,0,1,1,0,10,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0, +0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,0,1,0,11,0, +115,113,114,116,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0, +18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97, +108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,4, +102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114, +99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,0,1,0,12,0,115,113,114,116,0,1,1, +0,12,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59, 120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114, -0,0,0,0,1,0,10,0,115,113,114,116,0,1,1,0,10,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116, -95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95, -114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0, -18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121, -0,0,18,114,0,0,0,0,1,0,11,0,115,113,114,116,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108, -111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18, -114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108, -0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,0,1,0, -12,0,115,113,114,116,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115, -113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59, -121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114, -0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,4,102,108,111,97,116,95, -114,115,113,0,18,114,0,0,18,118,0,59,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114, -101,116,86,97,108,0,59,119,0,0,18,114,0,0,0,0,1,0,9,0,105,110,118,101,114,115,101,115,113,114,116, -0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,120,0,0,0,0,1,0,10,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,10,118,0,0,0,1, -4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120, -0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0, -59,121,0,0,0,0,1,0,11,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,11,118,0,0,0,1,4,102, -108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0, -4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121, -0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0, -59,122,0,0,0,0,1,0,12,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,12,118,0,0,0,1,4,102, -108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0, -4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121, -0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0, -59,122,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18, -118,0,59,119,0,0,0,0,1,0,9,0,110,111,114,109,97,108,105,122,101,0,1,1,0,9,120,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,17,49,0,48,0,0,20,0,0,1,0,10,0,110,111,114,109,97,108,105,122,101, -0,1,1,0,10,118,0,0,0,1,3,2,1,9,1,115,0,2,58,105,110,118,101,114,115,101,115,113,114,116,0,58,100, -111,116,0,18,118,0,0,18,118,0,0,0,0,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,115,0,59,120,120,0,0,0,0,1,0,11,0,110, -111,114,109,97,108,105,122,101,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118,101,99,51, -95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0, -18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,59,120,120,120,0,0,0,0,1, -0,12,0,110,111,114,109,97,108,105,122,101,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118, -101,99,52,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114, -115,113,0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,59,120,120, -120,0,0,0,0,1,0,9,0,97,98,115,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101, -116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,0,97,98,115,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52, -95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,97,98,115,0, -1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, -0,0,18,97,0,0,0,0,1,0,12,0,97,98,115,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95, -114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,115,105,103,110,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1, -112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120,0,0,18,120,0,0,17,48,0,48,0, -0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,0,0,17,48,0,48,0,0,0,18,120,0,0,0,4,118,101, -99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,0,0,18, -110,0,0,0,0,1,0,10,0,115,105,103,110,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,112,0,0,1,1,110,0,0,0,4,118, -101,99,52,95,115,103,116,0,18,112,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52, -95,115,103,116,0,18,110,0,59,120,121,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117, -98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,0,18,110,0,0,0,0,1, -0,11,0,115,105,103,110,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95, -115,103,116,0,18,112,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103, -116,0,18,110,0,59,120,121,122,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116, -114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,0,18,110,0,0,0,0,1,0, -12,0,115,105,103,110,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95, -115,103,116,0,18,112,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0, -17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101, -116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,0,9,0,102,108,111,111,114,0,1,1,0,9,97,0,0,0,1,4,118, -101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0, -10,0,102,108,111,111,114,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,108,111,111,114,0,1,1,0,11,97,0, -0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, -18,97,0,0,0,0,1,0,12,0,102,108,111,111,114,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,108,111, -111,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,99,101,105,108,0,1,1,0,9,97,0,0, -0,1,3,2,0,9,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0, -9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,0,54,20,0,0,1,0,10,0,99,101,105,108,0,1,1,0,10, -97,0,0,0,1,3,2,0,10,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18, -98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,54,20,0,0,1,0,11,0,99,101,105,108, -0,1,1,0,11,97,0,0,0,1,3,2,0,11,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18, -98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,98,0,54,20,0,0,1,0,12,0, -99,101,105,108,0,1,1,0,12,97,0,0,0,1,3,2,0,12,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108, -111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,9,0, -102,114,97,99,116,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86, -97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,0,102,114,97,99,116,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52, -95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,114, -97,99,116,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,12,0,102,114,97,99,116,0,1,1,0,12,97,0,0,0,1,4,118,101,99, -52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,109,111,100,0,1,1,0, -9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0, -59,120,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0, -0,48,47,20,0,0,1,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79, -118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18, -98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,114,0, -18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0,1,1,0,11,97, -0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0, -48,0,0,48,47,20,0,0,1,0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110, -101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66, -0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97, -0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1, -1,0,10,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114, -66,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,120,0,0,18,98,0, -59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,121,0,0,18,98,0, -59,121,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,59,120,0,18,98,0,59,120,0,58,102, -108,111,111,114,0,18,97,0,59,120,0,18,111,110,101,79,118,101,114,66,120,0,48,0,0,48,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59,121,0,58,102,108,111,111,114,0, -18,97,0,59,121,0,18,111,110,101,79,118,101,114,66,121,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0, -1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110, -101,79,118,101,114,66,121,0,0,1,1,111,110,101,79,118,101,114,66,122,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,111,110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,111,110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,111,110,101,79,118,101,114,66,122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101,116, -86,97,108,0,59,120,0,18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0, -18,111,110,101,79,118,101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121, -0,18,97,0,59,121,0,18,98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79, -118,101,114,66,121,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122, -0,18,98,0,59,122,0,58,102,108,111,111,114,0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122, -0,48,0,0,48,47,20,0,0,1,0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,3,2,0,9,1,111,110, -101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,1,1,111,110,101,79,118,101, -114,66,122,0,0,1,1,111,110,101,79,118,101,114,66,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -111,110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -111,110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -111,110,101,79,118,101,114,66,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -111,110,101,79,118,101,114,66,119,0,0,18,98,0,59,119,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0,18,111,110,101, -79,118,101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59, -121,0,18,98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79,118,101,114,66, -121,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122,0,18,98,0,59, -122,0,58,102,108,111,111,114,0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48, -47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,97,0,59,119,0,18,98,0,59,119,0,58,102,108, -111,111,114,0,18,97,0,59,119,0,18,111,110,101,79,118,101,114,66,119,0,48,0,0,48,47,20,0,0,1,0,9,0, -109,105,110,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101, -116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,0,1,0,10,0,109,105,110,0,1,1,0, -10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,1,0,11,0,109,105,110,0,1,1,0,11,97,0, -0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0,0,1,0,12,0,109,105,110,0,1,1,0,12, -97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18, -97,0,0,18,98,0,0,0,0,1,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, -109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0, -1,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95, -95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,12,0, -109,105,110,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101, -116,86,97,108,0,0,18,97,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,9,0,109,97,120,0,1,1,0,9,97,0,0, -1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -97,0,59,120,0,0,18,98,0,59,120,0,0,0,0,1,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4, -118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0, -0,18,98,0,59,120,121,0,0,0,0,1,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101, -99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0, -0,18,98,0,59,120,121,122,0,0,0,0,1,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118, -101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,109, -97,120,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116, -86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,11,0,109,97,120,0,1,1,0,11,97, -0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0, -59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,9,98, -0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,59, -120,120,120,120,0,0,0,0,1,0,9,0,99,108,97,109,112,0,1,1,0,9,118,97,108,0,0,1,1,0,9,109,105,110,86, -97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95, -114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0, -0,0,0,1,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0, -9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97, -108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,11,0,99, -108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86, -97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97, -108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,12,0,99,108,97,109,112,0, -1,1,0,12,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4, +0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95, +114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114, +101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18, +118,0,59,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,119,0, +0,18,114,0,0,0,0,1,0,9,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,9,120,0,0,0,1,4,102, +108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,0,0,1,0,10, +0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,10,118,0,0,0,1,4,102,108,111,97,116,95,114, +115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116, +95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,0,11,0,105, +110,118,101,114,115,101,115,113,114,116,0,1,1,0,11,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113, +0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114, +115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116, +95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,0,12,0,105, +110,118,101,114,115,101,115,113,114,116,0,1,1,0,12,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113, +0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114, +115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116, +95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111, +97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,9, +0,110,111,114,109,97,108,105,122,101,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,0,17,49,0,48,0,0,20,0,0,1,0,10,0,110,111,114,109,97,108,105,122,101,0,1,1,0,10,118,0,0,0,1,3,2, +1,9,1,115,0,2,58,105,110,118,101,114,115,101,115,113,114,116,0,58,100,111,116,0,18,118,0,0,18,118, +0,0,0,0,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, +0,59,120,121,0,0,18,118,0,0,18,115,0,59,120,120,0,0,0,0,1,0,11,0,110,111,114,109,97,108,105,122, +101,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118,101,99,51,95,100,111,116,0,18,116,109, +112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0,18,116, +109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, +0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,59,120,120,120,0,0,0,0,1,0,12,0,110,111,114,109, +97,108,105,122,101,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118,101,99,52,95,100,111, +116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109, +112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,59,120,120,120,0,0,0,0,1,0,9,0,97, +98,115,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120, +0,0,18,97,0,0,0,0,1,0,10,0,97,98,115,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,97,98,115,0,1,1,0,11,97,0,0,0,1,4, +118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0, +12,0,97,98,115,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108, +0,0,18,97,0,0,0,0,1,0,9,0,115,105,103,110,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,112,0,0,1,1,110,0,0,0,4, +118,101,99,52,95,115,103,116,0,18,112,0,59,120,0,0,18,120,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52, +95,115,103,116,0,18,110,0,59,120,0,0,17,48,0,48,0,0,0,18,120,0,0,0,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,0,0,18,110,0,0,0,0,1,0,10,0, +115,105,103,110,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103, +116,0,18,112,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18, +110,0,59,120,121,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,0,18,110,0,0,0,0,1,0,11,0,115,105,103, +110,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18, +112,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0, +59,120,121,122,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,0,18,110,0,0,0,0,1,0,12,0,115,105,103, +110,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18, +112,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0, +18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0, +18,112,0,0,18,110,0,0,0,0,1,0,9,0,102,108,111,111,114,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,102, +108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,0,102,108,111, +111,114,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,108,111,111,114,0,1,1,0,11,97,0,0,0,1,4,118,101,99, +52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0, +12,0,102,108,111,111,114,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95, +114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,99,101,105,108,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,98,0, +2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101, +116,86,97,108,0,59,120,0,18,98,0,54,20,0,0,1,0,10,0,99,101,105,108,0,1,1,0,10,97,0,0,0,1,3,2,0,10, +1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,18,98,0,54,20,0,0,1,0,11,0,99,101,105,108,0,1,1,0,11,97,0,0,0, +1,3,2,0,11,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0, +9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,98,0,54,20,0,0,1,0,12,0,99,101,105,108,0,1, +1,0,12,97,0,0,0,1,3,2,0,12,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98, +0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,9,0,102,114,97,99,116,0,1, +1,0,9,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, +97,0,0,0,0,1,0,10,0,102,114,97,99,116,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,114,97,99,116,0,1,1,0,11, +97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, +18,97,0,0,0,0,1,0,12,0,102,114,97,99,116,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0, +18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,109,111,100,0,1,1,0,9,97,0,0,1,1,0,9,98,0, +0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, +110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,18, +98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0, +10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110, +101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0, +0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110, +101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,97,0, +18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1, +0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95, +95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118, +101,114,66,0,48,0,0,48,47,20,0,0,1,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,3,2,0, +9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,0,4,102,108, +111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108, +111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,9,18,95, +95,114,101,116,86,97,108,0,59,120,0,18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18, +97,0,59,120,0,18,111,110,101,79,118,101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18, +111,110,101,79,118,101,114,66,121,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1, +0,11,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66, +121,0,0,1,1,111,110,101,79,118,101,114,66,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, +110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, +110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, +110,101,79,118,101,114,66,122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0, +18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0,18,111,110,101,79,118, +101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59,121,0,18, +98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79,118,101,114,66,121,0,48, +0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122,0,18,98,0,59,122,0,58, +102,108,111,111,114,0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48,47,20,0,0, +1,0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66, +120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,1,1,111,110,101,79,118,101,114,66,122,0,0,1,1, +111,110,101,79,118,101,114,66,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, +101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, +101,114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, +101,114,66,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, +101,114,66,119,0,0,18,98,0,59,119,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,59,120, +0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0,18,111,110,101,79,118,101,114,66,120, +0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59,121,0, +58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79,118,101,114,66,121,0,48,0,0,48,47,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122,0,18,98,0,59,122,0,58,102,108,111,111, +114,0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,119,0,18,97,0,59,119,0,18,98,0,59,119,0,58,102,108,111,111,114,0,18,97,0,59,119, +0,18,111,110,101,79,118,101,114,66,119,0,48,0,0,48,47,20,0,0,1,0,9,0,109,105,110,0,1,1,0,9,97,0,0, +1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, +97,0,59,120,0,0,18,98,0,59,120,0,0,0,0,1,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1, +4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120, +121,0,0,18,98,0,59,120,121,0,0,0,0,1,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4, +118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120, +121,122,0,0,18,98,0,59,120,121,122,0,0,0,0,1,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0, +0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1, +0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95, +114,101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,11,0,109,105,110,0, +1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108, +0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,12,0,109,105,110,0,1,1,0,12,97,0, +0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, +18,98,0,59,120,120,120,120,0,0,0,0,1,0,9,0,109,97,120,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118, +101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0, +59,120,0,0,0,0,1,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,97, +120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0, +0,0,0,1,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122, +0,0,0,0,1,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0, +18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1, +1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120, +121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118, +101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59, +120,120,120,0,0,0,0,1,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, +109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,9, +0,99,108,97,109,112,0,1,1,0,9,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120, +86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118, +97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,10,0,99,108,97,109,112, +0,1,1,0,10,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4, 118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109, -105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97, -108,0,0,1,1,0,10,109,105,110,86,97,108,0,0,1,1,0,10,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52, -95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97, -108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0, -11,109,105,110,86,97,108,0,0,1,1,0,11,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97, -109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109, -97,120,86,97,108,0,0,0,0,1,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,12,109,105,110, -86,97,108,0,0,1,1,0,12,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95, -95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97, -108,0,0,0,0,1,0,9,0,109,105,120,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99, -52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,10, -0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114, -112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0,109,105,120, -0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95, -114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1,1,0,12,120, -0,0,1,1,0,12,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86, -97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10, -121,0,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18, -97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11, -97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0, -0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4, -118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0, -0,0,1,0,9,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95, -115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1, -0,10,0,115,116,101,112,0,1,1,0,10,101,100,103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115, -103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1, -0,11,0,115,116,101,112,0,1,1,0,11,101,100,103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115, -103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0, -0,1,0,12,0,115,116,101,112,0,1,1,0,12,101,100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95, -115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,10,0, -115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,0,0,0, -0,1,0,11,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, -115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0, -59,120,120,120,0,0,0,0,1,0,12,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1, -4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101, -0,59,120,120,120,120,0,0,0,0,1,0,9,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103, -101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,116,0,2,58,99,108,97,109, -112,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, -17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,10,101,100,103,101,48, -0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,116,0,2,58,99,108,97,109,112,0, -18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48, -0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48, -47,48,0,0,1,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,11,101,100,103,101,48,0,0,1,1,0, -11,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0, -18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0, -0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0, -0,1,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,12,101,100,103,101,48,0,0,1,1,0,12,101, -100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101, -100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49, -0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,10, -0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101, -49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101, -48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0, -0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,11,0,115,109, -111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1, -1,0,11,118,0,0,0,1,3,2,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47, -18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8, -18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,12,0,115,109,111,111, -116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,12, -118,0,0,0,1,3,2,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101, -100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116, -0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,9,0,108,101,110,103,116,104, -0,1,1,0,9,120,0,0,0,1,8,58,97,98,115,0,18,120,0,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,10, -118,0,0,0,1,3,2,0,9,1,114,0,0,0,3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4, -102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0, -11,118,0,0,0,1,3,2,0,9,1,114,0,0,0,3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0, -4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0, -12,118,0,0,0,1,3,2,0,9,1,114,0,0,0,3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0, -4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0, -1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,3,2,1,9,1,100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101, -116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99, -101,0,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,1,10,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95, -95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,50,0,0,0,20,0,0,1,0,9,0,100,105,115, -116,97,110,99,101,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,1,11,1,100,51,0,2,18,118,0,18,117,0, -47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0, -9,0,100,105,115,116,97,110,99,101,0,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,12,1,100,52,0,2,18, -118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0, -0,0,20,0,0,1,0,11,0,99,114,111,115,115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95, -99,114,111,115,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0, -1,0,9,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114, -101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1, -115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,59,120,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58, -109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,10,0,102,97,99,101,102,111,114,119,97, -114,100,0,1,1,0,10,78,0,0,1,1,0,10,73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100, -111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116, -0,18,115,0,59,120,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18, -115,0,0,0,0,0,1,0,11,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0, -1,1,0,11,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0, -0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,59,120,0,0,17,48,0,48,0,0,0,18, -100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,12,0,102,97,99,101,102, -111,114,119,97,114,100,0,1,1,0,12,78,0,0,1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,9,1, -100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52, -95,115,103,116,0,18,115,0,59,120,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0, -18,78,0,0,18,115,0,0,0,0,0,1,0,9,0,114,101,102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1, -8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,10,0,114, -101,102,108,101,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111, -116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11,73, -0,0,1,1,0,11,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0, -48,47,0,0,1,0,12,0,114,101,102,108,101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17, -50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,9,0,114,101,102,114,97, -99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0, -18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100, -111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0, -9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58, -115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10, -73,0,0,1,1,0,10,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0, -18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0, -0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101, -116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0, -18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,11,0,114,101,102,114,97,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78, -0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48, +105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97, +108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95, +99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108, +0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,9, +109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109, +112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97, +120,86,97,108,0,0,0,0,1,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,10,109,105,110,86, +97,108,0,0,1,1,0,10,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95, +114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0, +0,0,0,1,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,11,109,105,110,86,97,108,0,0,1,1, +0,11,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86, +97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,12,0, +99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,12,109,105,110,86,97,108,0,0,1,1,0,12,109,97,120, +86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118, +97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,9,0,109,105,120,0,1,1, +0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101, +116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1, +0,10,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0, +0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1, +0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18, +121,0,0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,9,97,0,0,0,1, +4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0, +0,0,0,1,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52, +95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0, +109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,108,114,112, +0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1, +1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114, +101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,9,0,115,116,101,112,0,1,1,0,9,101, +100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, +108,0,59,120,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,10,0,115,116,101,112,0,1,1,0,10,101,100, +103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0, +59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,11,0,115,116,101,112,0,1,1,0,11,101,100, +103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0, +59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,12,0,115,116,101,112,0,1,1,0,12,101, +100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, +108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,10,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0, +0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,0,0,0,0,1,0,11,0,115,116,101,112,0,1,1,0,9,101, +100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,120,0,0,0,0,1,0,12,0,115,116, +101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95, +95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,120,120,0,0,0,0,1,0,9,0, +115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101, +49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,116,0,2,58,99,108,97,109,112,0,18,120,0,18,101,100,103,101,48, +0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0, +0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,10,0,115,109,111, +111,116,104,115,116,101,112,0,1,1,0,10,101,100,103,101,48,0,0,1,1,0,10,101,100,103,101,49,0,0,1,1, +0,10,118,0,0,0,1,3,2,0,10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18, +101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18, +116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,11,0,115,109,111,111,116, +104,115,116,101,112,0,1,1,0,11,101,100,103,101,48,0,0,1,1,0,11,101,100,103,101,49,0,0,1,1,0,11,118, +0,0,0,1,3,2,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100, +103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18, +116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,12,0,115,109,111,111,116,104,115, +116,101,112,0,1,1,0,12,101,100,103,101,48,0,0,1,1,0,12,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1, +3,2,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101, +49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0, +48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,10,0,115,109,111,111,116,104,115,116, +101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0, +10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18, +101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51, +0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1, +1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,116,0, +2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100, +103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0, +17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9, +101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,116,0,2,58, +99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101, +48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0, +48,0,0,18,116,0,48,47,48,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,9,120,0,0,0,1,8,58,97,98,115, +0,18,120,0,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,10,118,0,0,0,1,3,2,0,9,1,114,0,0,0,3,2, +1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0, +18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,114,0,0,0, +3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113, +0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,114,0,0,0, +3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113, +0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,114,0,0,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1, +3,2,1,9,1,100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103, +116,104,0,18,100,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118,0,0,1,1,0,10, +117,0,0,0,1,3,2,1,10,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, +108,101,110,103,116,104,0,18,100,50,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,11, +118,0,0,1,1,0,11,117,0,0,0,1,3,2,1,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116, +86,97,108,0,58,108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99, +101,0,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,12,1,100,52,0,2,18,118,0,18,117,0,47,0,0,9,18,95, +95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0,11,0,99,114,111, +115,115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115,115,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,9,0,102,97,99,101,102,111, +114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0, +2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95, +115,103,116,0,18,115,0,59,120,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18, +78,0,0,18,115,0,0,0,0,0,1,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10,78,0,0,1,1,0, +10,73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0, +18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,59,120,0,0,17,48,0, +48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,11,0,102,97, +99,101,102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0,1,1,0,11,78,114,101,102,0,0,0, +1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4, +118,101,99,52,95,115,103,116,0,18,115,0,59,120,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120, +0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,12,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1, +0,12,78,0,0,1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18, +78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0, +59,120,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0, +0,1,0,9,0,114,101,102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0, +58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,10,0,114,101,102,108,101,99,116,0, +1,1,0,10,73,0,0,1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0, +0,48,18,78,0,48,47,0,0,1,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,0,1,8, +18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,12,0,114, +101,102,108,101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111, +116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,9,0,114,101,102,114,97,99,116,0,1,1,0,9,73,0, +0,1,1,0,9,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18, +101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0, +18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116, +97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18, +107,0,0,0,46,18,78,0,48,47,0,0,1,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0, +0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48, 17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47, 48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48, 18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18, -78,0,48,47,0,0,1,0,12,0,114,101,102,114,97,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,1,1,0,9,101, +78,0,48,47,0,0,1,0,11,0,114,101,102,114,97,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,1,1,0,9,101, 116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0, 58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10, 18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97, 0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0, -1,0,13,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0, -0,1,8,58,109,97,116,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,14,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0, -1,0,0,14,109,0,0,1,0,0,14,110,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48, -0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10, -50,0,57,48,0,0,0,0,1,0,15,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,15,109,0, -0,1,0,0,15,110,0,0,0,1,8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0, -18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,0,1, -1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,11, -117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0, -1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0, -0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118, -101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95, -115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4, -0,108,101,115,115,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108, -116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104, -97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115, -84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108, -101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99, -52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101, -115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95, -115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0, -108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99, -52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0, -1,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4, -118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2, -0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52, -95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0, -103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52, -95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0, -4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99, -52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114, -101,97,116,101,114,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114, -101,97,116,101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103, -114,101,97,116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115, -103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116, -101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95, -115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0, -103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1, -4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18, -118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0, -1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0, -0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117, -0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97, -108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97, -110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95, -95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0, -0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0, -1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0, -18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52, -95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117, -97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0, -1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1, -4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0, -2,0,101,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108, -0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1, -1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, -18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4, -0,110,111,116,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117, -97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0, -7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0, -1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, -18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4, -0,110,111,116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,2, -118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0, -0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0, -3,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120, -0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, -120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95, -95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97, -110,121,0,1,1,0,4,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115, -117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18, -115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97, -100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101, -99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17, -48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0, -18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, -18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,8,18,118,0,59,120,0,18,118,0,59,121,0,34,0,0,1, -0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0, -0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114, -111,100,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1, -1,0,4,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0, -18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59, -120,0,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0, -2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4, -118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17, -48,0,48,0,0,0,0,0,1,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95, -95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101, -49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52, -95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, +1,0,12,0,114,101,102,114,97,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3, +2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0, +18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0, +48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116, +0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,13,0,109,97, +116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97, +116,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49, +0,57,48,0,0,0,0,1,0,14,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1, +0,0,14,110,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0, +16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0, +1,0,15,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0,0, +0,1,8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0, +57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1,1, +0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0, +0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117, +0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4, +118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2, +0,108,101,115,115,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101, +115,115,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84, +104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101, +116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97, +108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,69,113, +117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97, +110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18, +95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,69, +113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110, +69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84, +104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,101, +0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114, +84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101, +114,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116, +101,114,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18, +95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84, +104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104, +97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104, +97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116, +86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113, +117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84, +104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103, +114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4, +118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2, +0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1, +4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118, +0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0, +7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1, +0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4, +118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0, +0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115, +101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0, +101,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95, +95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,6,117,0, +0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4, +118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18, +118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95, +115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97, +108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,3,117,0,0,1,1, +0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118, +101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0, +110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111, +116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69, +113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95, +114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0, +6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1, +0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0, +0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0, +110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111, +116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, +95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1,3, +2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59, +120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,3,118,0,0,0,1, +3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0, +59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18, +115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, +116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1, +1,0,4,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, +120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0, +59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,100,100,0,18, +115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0, +0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59, +121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,114, +111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,8,18,118,0,59,120,0,18,118,0,59,121,0,34,0,0,1,0,1,0,97, +108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114,111,100, +0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108, +0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,4,118, +0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,18,118,0, +59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0, +18,112,114,111,100,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, +114,101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,2,0, +110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108, +0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4,118, +101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0, +48,0,0,0,0,0,1,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, +114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49, +68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95, +116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, 111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115, 97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112, 49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, -- cgit v1.2.3 From d398e1360d5c1c0f44a4ba9cd167c39a29ce8254 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 17 Jul 2008 10:27:10 +0900 Subject: python: Reimplement tile comparison in C to speed up tests. --- src/gallium/state_trackers/python/gallium.i | 48 ++++++++++++++++++---- src/gallium/state_trackers/python/tests/texture.py | 20 +++++---- 2 files changed, 53 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index fe0afb18a9..1c207a41b9 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -425,23 +425,23 @@ error1: void unmap( void ); void - get_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, unsigned char *p, unsigned stride) { - pipe_get_tile_raw($self, x, y, w, h, p, stride); + get_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, unsigned char *raw, unsigned stride) { + pipe_get_tile_raw($self, x, y, w, h, raw, stride); } void - put_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned char *p, unsigned stride) { - pipe_put_tile_raw($self, x, y, w, h, p, stride); + put_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned char *raw, unsigned stride) { + pipe_put_tile_raw($self, x, y, w, h, raw, stride); } void - get_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, float *p) { - pipe_get_tile_rgba($self, x, y, w, h, p); + get_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, float *rgba) { + pipe_get_tile_rgba($self, x, y, w, h, rgba); } void - put_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *p) { - pipe_put_tile_rgba($self, x, y, w, h, p); + put_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba) { + pipe_put_tile_rgba($self, x, y, w, h, rgba); } void @@ -454,6 +454,38 @@ error1: pipe_put_tile_z($self, x, y, w, h, z); } + unsigned + compare_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba, float tol = 0.0) + { + float *rgba2; + const float *p1; + const float *p2; + unsigned i, j, n; + + rgba2 = MALLOC(h*w*4*sizeof(float)); + if(!rgba2) + return ~0; + + pipe_get_tile_rgba($self, x, y, w, h, rgba2); + + p1 = rgba; + p2 = rgba2; + n = 0; + for(i = h*w; i; --i) { + unsigned differs = 0; + for(j = 4; j; --j) { + float delta = *p2++ - *p1++; + if (delta < -tol || delta > tol) + differs = 1; + } + n += differs; + } + + FREE(rgba2); + + return n; + } + }; diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index edaa2b8b26..3d76953126 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -304,15 +304,21 @@ class TextureTest(TestCase): ctx.flush() - rgba = FloatArray(h*w*4) - - cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ).get_tile_rgba(x, y, w, h, rgba) + cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_CPU_READ) + + total = h*w + different = cbuf.compare_tile_rgba(x, y, w, h, expected_rgba, tol=4.0/256) + if different: + sys.stderr.write("%u out of %u pixels differ\n" % (different, total)) - if not compare_rgba(w, h, rgba, expected_rgba): + if float(total - different)/float(total) < 0.85: - #show_image(w, h, Result=rgba, Expected=expected_rgba) - #save_image(w, h, rgba, "result.png") - #save_image(w, h, expected_rgba, "expected.png") + if 0: + rgba = FloatArray(h*w*4) + cbuf.get_tile_rgba(x, y, w, h, rgba) + show_image(w, h, Result=rgba, Expected=expected_rgba) + save_image(w, h, rgba, "result.png") + save_image(w, h, expected_rgba, "expected.png") #sys.exit(0) raise TestFailure -- cgit v1.2.3 From 6a65a0d9efb82b11cafe5b411abddd57a4fb838a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 17 Jul 2008 11:26:53 +0900 Subject: gallium: Be less verbose with debug options messages. --- src/gallium/auxiliary/util/p_debug.c | 32 ++++++++++++++++++++------------ 1 file changed, 20 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 7b28900a25..b0240ad737 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -174,20 +174,19 @@ copy(char *dst, const char *start, const char *end, size_t n) #endif -const char * -debug_get_option(const char *name, const char *dfault) +static INLINE const char * +_debug_get_option(const char *name) { - const char *result; #if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) /* EngMapFile creates the file if it does not exists, so it must either be * disabled on release versions (or put in a less conspicuous place). */ #ifdef DEBUG + const char *result = NULL; ULONG_PTR iFile = 0; const void *pMap = NULL; const char *sol, *eol, *sep; static char output[1024]; - result = dfault; pMap = EngMapFile(L"\\??\\c:\\gallium.cfg", 0, &iFile); if(pMap) { sol = (const char *)pMap; @@ -208,18 +207,27 @@ debug_get_option(const char *name, const char *dfault) } EngUnmapFile(iFile); } + return result; #else - result = dfault; + return NULL; #endif #elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) /* TODO: implement */ - result = dfault; + return NULL; #else - result = getenv(name); - if(!result) - result = dfault; + return getenv(name); #endif +} +const char * +debug_get_option(const char *name, const char *dfault) +{ + const char *result; + + result = _debug_get_option(name); + if(!result) + result = dfault; + debug_printf("%s: %s = %s\n", __FUNCTION__, name, result ? result : "(null)"); return result; @@ -228,7 +236,7 @@ debug_get_option(const char *name, const char *dfault) boolean debug_get_bool_option(const char *name, boolean dfault) { - const char *str = debug_get_option(name, NULL); + const char *str = _debug_get_option(name); boolean result; if(str == NULL) @@ -258,7 +266,7 @@ debug_get_num_option(const char *name, long dfault) long result; const char *str; - str = debug_get_option(name, NULL); + str = _debug_get_option(name); if(!str) result = dfault; else { @@ -294,7 +302,7 @@ debug_get_flags_option(const char *name, unsigned long result; const char *str; - str = debug_get_option(name, NULL); + str = _debug_get_option(name); if(!str) result = dfault; else { -- cgit v1.2.3 From 174c6912d68a954ff5eddf8b953b813e2eac3deb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 17 Jul 2008 22:54:14 +0900 Subject: gallium: Expose the ability to get a tile outside a surface. --- src/gallium/auxiliary/util/p_tile.c | 83 +++++++++++++++++++++---------------- src/gallium/auxiliary/util/p_tile.h | 7 ++++ 2 files changed, 54 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 580a95568e..1daae6b640 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -706,81 +706,92 @@ ycbcr_get_tile_rgba(ushort *src, void -pipe_get_tile_rgba(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p) +pipe_tile_raw_to_rgba(enum pipe_format format, + void *src, + uint w, uint h, + float *dst, unsigned dst_stride) { - unsigned dst_stride = w * 4; - void *packed; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); - - if (!packed) - return; - - pipe_get_tile_raw(ps, x, y, w, h, packed, 0); - - switch (ps->format) { + switch (format) { case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + a8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_X8R8G8B8_UNORM: - x8r8g8b8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + x8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + b8g8r8a8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_A1R5G5B5_UNORM: - a1r5g5b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + a1r5g5b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_A4R4G4B4_UNORM: - a4r4g4b4_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + a4r4g4b4_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_R5G6B5_UNORM: - r5g6b5_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + r5g6b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_L8_UNORM: - l8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); + l8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_A8_UNORM: - a8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); + a8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_I8_UNORM: - i8_get_tile_rgba((ubyte *) packed, w, h, p, dst_stride); + i8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_A8L8_UNORM: - a8_l8_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + a8_l8_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_get_tile_rgba((short *) packed, w, h, p, dst_stride); + r16g16b16a16_get_tile_rgba((short *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_Z16_UNORM: - z16_get_tile_rgba((ushort *) packed, w, h, p, dst_stride); + z16_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_Z32_UNORM: - z32_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + z32_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_S8Z24_UNORM: case PIPE_FORMAT_X8Z24_UNORM: - s8z24_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + s8z24_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_Z24S8_UNORM: - z24s8_get_tile_rgba((unsigned *) packed, w, h, p, dst_stride); + z24s8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_YCBCR: - assert((x & 1) == 0); - ycbcr_get_tile_rgba((ushort *) packed, w, h, p, dst_stride, FALSE); + ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, FALSE); break; case PIPE_FORMAT_YCBCR_REV: - assert((x & 1) == 0); - ycbcr_get_tile_rgba((ushort *) packed, w, h, p, dst_stride, TRUE); + ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, TRUE); break; default: assert(0); } +} + + +void +pipe_get_tile_rgba(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p) +{ + unsigned dst_stride = w * 4; + void *packed; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); + + if (!packed) + return; + + if(ps->format == PIPE_FORMAT_YCBCR || ps->format == PIPE_FORMAT_YCBCR_REV) + assert((x & 1) == 0); + + pipe_get_tile_raw(ps, x, y, w, h, packed, 0); + + pipe_tile_raw_to_rgba(ps->format, packed, w, h, p, dst_stride); FREE(packed); } diff --git a/src/gallium/auxiliary/util/p_tile.h b/src/gallium/auxiliary/util/p_tile.h index adfec8bcee..a8ac805308 100644 --- a/src/gallium/auxiliary/util/p_tile.h +++ b/src/gallium/auxiliary/util/p_tile.h @@ -87,6 +87,13 @@ pipe_put_tile_z(struct pipe_surface *ps, uint x, uint y, uint w, uint h, const uint *z); +void +pipe_tile_raw_to_rgba(enum pipe_format format, + void *src, + uint w, uint h, + float *dst, unsigned dst_stride); + + #ifdef __cplusplus } #endif -- cgit v1.2.3 From e74f94b39ac47a319ee21c526315fb6fbc1513de Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 18 Jul 2008 01:08:06 +0900 Subject: python: Reimplement the generation of random texture samples in C. --- src/gallium/state_trackers/python/SConscript | 1 + src/gallium/state_trackers/python/gallium.i | 6 + src/gallium/state_trackers/python/st_sample.c | 548 +++++++++++++++++++++ src/gallium/state_trackers/python/st_sample.h | 47 ++ src/gallium/state_trackers/python/tests/data.py | 285 ----------- src/gallium/state_trackers/python/tests/texture.py | 29 +- 6 files changed, 606 insertions(+), 310 deletions(-) create mode 100644 src/gallium/state_trackers/python/st_sample.c create mode 100644 src/gallium/state_trackers/python/st_sample.h delete mode 100644 src/gallium/state_trackers/python/tests/data.py (limited to 'src') diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript index b284d2020b..973d96d55a 100644 --- a/src/gallium/state_trackers/python/SConscript +++ b/src/gallium/state_trackers/python/SConscript @@ -20,6 +20,7 @@ if 'python' in env['statetrackers']: source = [ 'gallium.i', 'st_device.c', + 'st_sample.c', 'st_softpipe_winsys.c', ], ) diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 1c207a41b9..d5b24b04ff 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -51,6 +51,7 @@ #include "tgsi/util/tgsi_dump.h" #include "st_device.h" +#include "st_sample.h" %} @@ -454,6 +455,11 @@ error1: pipe_put_tile_z($self, x, y, w, h, z); } + void + sample_rgba(float *rgba) { + st_sample_surface($self, rgba); + } + unsigned compare_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba, float tol = 0.0) { diff --git a/src/gallium/state_trackers/python/st_sample.c b/src/gallium/state_trackers/python/st_sample.c new file mode 100644 index 0000000000..b47c7be293 --- /dev/null +++ b/src/gallium/state_trackers/python/st_sample.c @@ -0,0 +1,548 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_compiler.h" +#include "pipe/p_format.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "util/p_tile.h" + +#include "st_sample.h" + + +/** + * Use our own pseudo random generator to ensure consistent runs among + * multiple runs and platforms. + * + * @sa http://en.wikipedia.org/wiki/Linear_congruential_generator + */ +static uint32_t st_random(void) { + static uint64_t seed = UINT64_C(0xbb9a063afb0a739d); + + seed = UINT64_C(134775813) * seed + UINT64_C(1); + + return (uint16_t)(seed >> 32); +} + + +/** + * We don't want to include the patent-encumbered DXT code here, so instead + * we store several uncompressed/compressed data pairs for hardware testing + * purposes. + */ +struct dxt_data +{ + uint8_t rgba[16*4]; + uint8_t raw[16]; +}; + + +static const struct dxt_data +dxt1_rgb_data[] = { + { + { + 0x99, 0xb0, 0x8e, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x99, 0xb0, 0x8e, 0xff, + 0x99, 0xb0, 0x8e, 0xff, + 0xd6, 0xff, 0x94, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x99, 0xb0, 0x8e, 0xff, + 0xd6, 0xff, 0x94, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x99, 0xb0, 0x8e, 0xff, + 0x21, 0x14, 0x84, 0xff, + 0x5d, 0x62, 0x89, 0xff, + 0x21, 0x14, 0x84, 0xff, + 0x21, 0x14, 0x84, 0xff, + 0x99, 0xb0, 0x8e, 0xff + }, + {0xf2, 0xd7, 0xb0, 0x20, 0xae, 0x2c, 0x6f, 0x97} + }, + { + { + 0xb5, 0xcf, 0x9c, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x21, 0x08, 0x6b, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x52, 0x4a, 0x7b, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0xb5, 0xcf, 0x9c, 0xff, + 0x21, 0x08, 0x6b, 0xff, + 0xb5, 0xcf, 0x9c, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x52, 0x4a, 0x7b, 0xff, + 0xb5, 0xcf, 0x9c, 0xff, + 0x83, 0x8c, 0x8b, 0xff, + 0x52, 0x4a, 0x7b, 0xff, + 0x83, 0x8c, 0x8b, 0xff + }, + {0x73, 0xb6, 0x4d, 0x20, 0x98, 0x2b, 0xe1, 0xb8} + }, + { + { + 0x00, 0x2c, 0xff, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0x4a, 0x5c, 0xbd, 0xff, + 0x4a, 0x5c, 0xbd, 0xff, + 0x4a, 0x5c, 0xbd, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0x94, 0x8d, 0x7b, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0xde, 0xbe, 0x39, 0xff, + 0x94, 0x8d, 0x7b, 0xff + }, + {0xe7, 0xdd, 0x7f, 0x01, 0xf9, 0xab, 0x08, 0x80} + }, + { + { + 0x6b, 0x24, 0x21, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x8b, 0x7a, 0x99, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x6b, 0x24, 0x21, 0xff, + 0x8b, 0x7a, 0x99, 0xff, + 0x9c, 0xa6, 0xd6, 0xff, + 0x6b, 0x24, 0x21, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x8b, 0x7a, 0x99, 0xff, + 0x6b, 0x24, 0x21, 0xff, + 0x8b, 0x7a, 0x99, 0xff, + 0x7b, 0x4f, 0x5d, 0xff, + 0x9c, 0xa6, 0xd6, 0xff + }, + {0x3a, 0x9d, 0x24, 0x69, 0xbd, 0x9f, 0xb4, 0x39} + } +}; + + +static const struct dxt_data +dxt1_rgba_data[] = { + { + { + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff, + 0x4e, 0xaa, 0x90, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff, + 0x29, 0xff, 0xff, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff, + 0x73, 0x55, 0x21, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff, + 0x4e, 0xaa, 0x90, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x4e, 0xaa, 0x90, 0xff + }, + {0xff, 0x2f, 0xa4, 0x72, 0xeb, 0xb2, 0xbd, 0xbe} + }, + { + { + 0xb5, 0xe3, 0x63, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x6b, 0x24, 0x84, 0xff, + 0xb5, 0xe3, 0x63, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0xb5, 0xe3, 0x63, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x6b, 0x24, 0x84, 0xff, + 0x6b, 0x24, 0x84, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0xb5, 0xe3, 0x63, 0xff, + 0x90, 0x83, 0x73, 0xff, + 0xb5, 0xe3, 0x63, 0xff + }, + {0x30, 0x69, 0x0c, 0xb7, 0x4d, 0xf7, 0x0f, 0x67} + }, + { + { + 0x00, 0x00, 0x00, 0x00, + 0xc6, 0x86, 0x8c, 0xff, + 0xc6, 0x86, 0x8c, 0xff, + 0x21, 0x65, 0x42, 0xff, + 0x21, 0x65, 0x42, 0xff, + 0x21, 0x65, 0x42, 0xff, + 0x21, 0x65, 0x42, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x21, 0x65, 0x42, 0xff, + 0xc6, 0x86, 0x8c, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0xc6, 0x86, 0x8c, 0xff + }, + {0x28, 0x23, 0x31, 0xc4, 0x17, 0xc0, 0xd3, 0x7f} + }, + { + { + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0xc6, 0xe3, 0x9c, 0xff, + 0x7b, 0x1c, 0x52, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x00, 0x00, 0x00, 0x00, + 0x7b, 0x1c, 0x52, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0x7b, 0x1c, 0x52, 0xff, + 0xa0, 0x7f, 0x77, 0xff, + 0xc6, 0xe3, 0x9c, 0xff, + 0x00, 0x00, 0x00, 0x00, + 0xa0, 0x7f, 0x77, 0xff + }, + {0xea, 0x78, 0x13, 0xc7, 0x7f, 0xfc, 0x33, 0xb6} + }, +}; + + +static const struct dxt_data +dxt3_rgba_data[] = { + { + { + 0x6d, 0xc6, 0x96, 0x77, + 0x6d, 0xc6, 0x96, 0xee, + 0x6d, 0xc6, 0x96, 0xaa, + 0x8c, 0xff, 0xb5, 0x44, + 0x6d, 0xc6, 0x96, 0xff, + 0x6d, 0xc6, 0x96, 0x88, + 0x31, 0x55, 0x5a, 0x66, + 0x6d, 0xc6, 0x96, 0x99, + 0x31, 0x55, 0x5a, 0xbb, + 0x31, 0x55, 0x5a, 0x55, + 0x31, 0x55, 0x5a, 0x11, + 0x6d, 0xc6, 0x96, 0xcc, + 0x6d, 0xc6, 0x96, 0xcc, + 0x6d, 0xc6, 0x96, 0x11, + 0x31, 0x55, 0x5a, 0x44, + 0x31, 0x55, 0x5a, 0x88 + }, + {0xe7, 0x4a, 0x8f, 0x96, 0x5b, 0xc1, 0x1c, 0x84, 0xf6, 0x8f, 0xab, 0x32, 0x2a, 0x9a, 0x95, 0x5a} + }, + { + { + 0xad, 0xeb, 0x73, 0x99, + 0x97, 0xaa, 0x86, 0x66, + 0x6b, 0x28, 0xad, 0x99, + 0xad, 0xeb, 0x73, 0x99, + 0x6b, 0x28, 0xad, 0x22, + 0xad, 0xeb, 0x73, 0xff, + 0x97, 0xaa, 0x86, 0x55, + 0x6b, 0x28, 0xad, 0x55, + 0x6b, 0x28, 0xad, 0x44, + 0xad, 0xeb, 0x73, 0x33, + 0x6b, 0x28, 0xad, 0xee, + 0x6b, 0x28, 0xad, 0x99, + 0x97, 0xaa, 0x86, 0x66, + 0xad, 0xeb, 0x73, 0xbb, + 0x97, 0xaa, 0x86, 0x99, + 0xad, 0xeb, 0x73, 0xbb + }, + {0x69, 0x99, 0xf2, 0x55, 0x34, 0x9e, 0xb6, 0xb9, 0x4e, 0xaf, 0x55, 0x69, 0x18, 0x61, 0x51, 0x22} + }, + { + { + 0x63, 0xd7, 0xd6, 0x00, + 0x57, 0x62, 0x5d, 0xdd, + 0x57, 0x62, 0x5d, 0xcc, + 0x57, 0x62, 0x5d, 0xbb, + 0x52, 0x28, 0x21, 0xaa, + 0x57, 0x62, 0x5d, 0xcc, + 0x57, 0x62, 0x5d, 0xcc, + 0x57, 0x62, 0x5d, 0x66, + 0x57, 0x62, 0x5d, 0x22, + 0x57, 0x62, 0x5d, 0xdd, + 0x63, 0xd7, 0xd6, 0xee, + 0x57, 0x62, 0x5d, 0x33, + 0x63, 0xd7, 0xd6, 0x55, + 0x52, 0x28, 0x21, 0x55, + 0x57, 0x62, 0x5d, 0x11, + 0x5d, 0x9c, 0x99, 0xee + }, + {0xd0, 0xbc, 0xca, 0x6c, 0xd2, 0x3e, 0x55, 0xe1, 0xba, 0x66, 0x44, 0x51, 0xfc, 0xfd, 0xcf, 0xb4} + }, + { + { + 0x94, 0x6f, 0x60, 0x22, + 0x94, 0x6f, 0x60, 0x22, + 0xc5, 0xab, 0x76, 0x11, + 0xc5, 0xab, 0x76, 0xee, + 0x63, 0x34, 0x4a, 0xdd, + 0x63, 0x34, 0x4a, 0x33, + 0x94, 0x6f, 0x60, 0x77, + 0xf7, 0xe7, 0x8c, 0x00, + 0x94, 0x6f, 0x60, 0x33, + 0x63, 0x34, 0x4a, 0xaa, + 0x94, 0x6f, 0x60, 0x77, + 0x63, 0x34, 0x4a, 0xcc, + 0x94, 0x6f, 0x60, 0xaa, + 0xf7, 0xe7, 0x8c, 0x99, + 0x63, 0x34, 0x4a, 0x44, + 0xc5, 0xab, 0x76, 0xaa + }, + {0x22, 0xe1, 0x3d, 0x07, 0xa3, 0xc7, 0x9a, 0xa4, 0x31, 0xf7, 0xa9, 0x61, 0xaf, 0x35, 0x77, 0x93} + }, +}; + + +static const struct dxt_data +dxt5_rgba_data[] = { + { + { + 0x6d, 0xc6, 0x96, 0x74, + 0x6d, 0xc6, 0x96, 0xf8, + 0x6d, 0xc6, 0x96, 0xb6, + 0x8c, 0xff, 0xb5, 0x53, + 0x6d, 0xc6, 0x96, 0xf8, + 0x6d, 0xc6, 0x96, 0x95, + 0x31, 0x55, 0x5a, 0x53, + 0x6d, 0xc6, 0x96, 0x95, + 0x31, 0x55, 0x5a, 0xb6, + 0x31, 0x55, 0x5a, 0x53, + 0x31, 0x55, 0x5a, 0x11, + 0x6d, 0xc6, 0x96, 0xd7, + 0x6d, 0xc6, 0x96, 0xb6, + 0x6d, 0xc6, 0x96, 0x11, + 0x31, 0x55, 0x5a, 0x32, + 0x31, 0x55, 0x5a, 0x95 + }, + {0xf8, 0x11, 0xc5, 0x0c, 0x9a, 0x73, 0xb4, 0x9c, 0xf6, 0x8f, 0xab, 0x32, 0x2a, 0x9a, 0x95, 0x5a} + }, + { + { + 0xad, 0xeb, 0x73, 0xa1, + 0x97, 0xaa, 0x86, 0x65, + 0x6b, 0x28, 0xad, 0xa1, + 0xad, 0xeb, 0x73, 0xa1, + 0x6b, 0x28, 0xad, 0x2a, + 0xad, 0xeb, 0x73, 0xfb, + 0x97, 0xaa, 0x86, 0x47, + 0x6b, 0x28, 0xad, 0x65, + 0x6b, 0x28, 0xad, 0x47, + 0xad, 0xeb, 0x73, 0x47, + 0x6b, 0x28, 0xad, 0xdd, + 0x6b, 0x28, 0xad, 0xa1, + 0x97, 0xaa, 0x86, 0x65, + 0xad, 0xeb, 0x73, 0xbf, + 0x97, 0xaa, 0x86, 0xa1, + 0xad, 0xeb, 0x73, 0xbf + }, + {0xfb, 0x2a, 0x34, 0x19, 0xdc, 0xbf, 0xe8, 0x71, 0x4e, 0xaf, 0x55, 0x69, 0x18, 0x61, 0x51, 0x22} + }, + { + { + 0x63, 0xd7, 0xd6, 0x00, + 0x57, 0x62, 0x5d, 0xf5, + 0x57, 0x62, 0x5d, 0xd2, + 0x57, 0x62, 0x5d, 0xaf, + 0x52, 0x28, 0x21, 0xaf, + 0x57, 0x62, 0x5d, 0xd2, + 0x57, 0x62, 0x5d, 0xd2, + 0x57, 0x62, 0x5d, 0x69, + 0x57, 0x62, 0x5d, 0x23, + 0x57, 0x62, 0x5d, 0xd2, + 0x63, 0xd7, 0xd6, 0xf5, + 0x57, 0x62, 0x5d, 0x46, + 0x63, 0xd7, 0xd6, 0x46, + 0x52, 0x28, 0x21, 0x69, + 0x57, 0x62, 0x5d, 0x23, + 0x5d, 0x9c, 0x99, 0xf5 + }, + {0xf5, 0x00, 0x81, 0x36, 0xa9, 0x17, 0xec, 0x1e, 0xba, 0x66, 0x44, 0x51, 0xfc, 0xfd, 0xcf, 0xb4} + }, + { + { + 0x94, 0x6f, 0x60, 0x25, + 0x94, 0x6f, 0x60, 0x25, + 0xc5, 0xab, 0x76, 0x05, + 0xc5, 0xab, 0x76, 0xe8, + 0x63, 0x34, 0x4a, 0xe8, + 0x63, 0x34, 0x4a, 0x25, + 0x94, 0x6f, 0x60, 0x86, + 0xf7, 0xe7, 0x8c, 0x05, + 0x94, 0x6f, 0x60, 0x25, + 0x63, 0x34, 0x4a, 0xa7, + 0x94, 0x6f, 0x60, 0x66, + 0x63, 0x34, 0x4a, 0xc7, + 0x94, 0x6f, 0x60, 0xa7, + 0xf7, 0xe7, 0x8c, 0xa7, + 0x63, 0x34, 0x4a, 0x45, + 0xc5, 0xab, 0x76, 0xa7 + }, + {0xe8, 0x05, 0x7f, 0x80, 0x33, 0x5f, 0xb5, 0x79, 0x31, 0xf7, 0xa9, 0x61, 0xaf, 0x35, 0x77, 0x93} + }, +}; + + +static INLINE void +st_sample_dxt_pixel_block(enum pipe_format format, + const struct pipe_format_block *block, + uint8_t *raw, + float *rgba, unsigned rgba_stride, + unsigned w, unsigned h) +{ + const struct dxt_data *data; + unsigned n; + unsigned i; + unsigned x, y, ch; + + switch(format) { + case PIPE_FORMAT_DXT1_RGB: + data = dxt1_rgb_data; + n = sizeof(dxt1_rgb_data)/sizeof(dxt1_rgb_data[0]); + break; + case PIPE_FORMAT_DXT1_RGBA: + data = dxt1_rgba_data; + n = sizeof(dxt1_rgba_data)/sizeof(dxt1_rgba_data[0]); + break; + case PIPE_FORMAT_DXT3_RGBA: + data = dxt3_rgba_data; + n = sizeof(dxt3_rgba_data)/sizeof(dxt3_rgba_data[0]); + break; + case PIPE_FORMAT_DXT5_RGBA: + data = dxt5_rgba_data; + n = sizeof(dxt5_rgba_data)/sizeof(dxt5_rgba_data[0]); + break; + default: + assert(0); + } + + i = st_random() % n; + + for(y = 0; y < h; ++y) + for(x = 0; x < w; ++x) + for(ch = 0; ch < 4; ++ch) + rgba[y*rgba_stride + x*4 + ch] = (float)(data[i].rgba[y*4*4 + x*4 + ch])/255.0f; + + memcpy(raw, data[i].raw, block->size); +} + + +static INLINE void +st_sample_generic_pixel_block(enum pipe_format format, + const struct pipe_format_block *block, + uint8_t *raw, + float *rgba, unsigned rgba_stride, + unsigned w, unsigned h) +{ + unsigned i; + unsigned x, y, ch; + + for(i = 0; i < block->size; ++i) + raw[i] = (uint8_t)st_random(); + + + pipe_tile_raw_to_rgba(format, + raw, + w, h, + rgba, rgba_stride); + + if(format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV) { + for(y = 0; y < h; ++y) { + for(x = 0; x < w; ++x) { + for(ch = 0; ch < 4; ++ch) { + unsigned offset = y*rgba_stride + x*4 + ch; + rgba[offset] = CLAMP(rgba[offset], 0.0f, 1.0f); + } + } + } + } +} + + +/** + * Randomly sample pixels. + */ +void +st_sample_pixel_block(enum pipe_format format, + const struct pipe_format_block *block, + void *raw, + float *rgba, unsigned rgba_stride, + unsigned w, unsigned h) +{ + switch(format) { + case PIPE_FORMAT_DXT1_RGB: + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: + st_sample_dxt_pixel_block(format, block, raw, rgba, rgba_stride, w, h); + break; + + default: + st_sample_generic_pixel_block(format, block, raw, rgba, rgba_stride, w, h); + break; + } +} + + +void +st_sample_surface(struct pipe_surface *surface, float *rgba) +{ + const struct pipe_format_block *block = &surface->block; + unsigned rgba_stride = surface->width*4; + void *raw; + unsigned x, y; + + raw = pipe_surface_map(surface, PIPE_BUFFER_USAGE_CPU_READ); + if(!raw) + return; + + for (y = 0; y < surface->nblocksy; ++y) { + for(x = 0; x < surface->nblocksx; ++x) { + st_sample_pixel_block(surface->format, + block, + (uint8_t*)raw + y*surface->stride + x*block->size, + rgba + y*block->height*rgba_stride + x*block->width*4, + rgba_stride, + MIN2(block->width, surface->width - x*block->width), + MIN2(block->height, surface->height - y*block->height)); + } + } + + pipe_surface_unmap(surface); +} diff --git a/src/gallium/state_trackers/python/st_sample.h b/src/gallium/state_trackers/python/st_sample.h new file mode 100644 index 0000000000..ff04a12613 --- /dev/null +++ b/src/gallium/state_trackers/python/st_sample.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_SAMPLE_H_ +#define ST_SAMPLE_H_ + + +#include "pipe/p_format.h" + + +void +st_sample_pixel_block(enum pipe_format format, + const struct pipe_format_block *block, + void *raw, + float *rgba, unsigned rgba_stride, + unsigned w, unsigned h); + +void +st_sample_surface(struct pipe_surface *surface, float *rgba); + + +#endif /* ST_SAMPLE_H_ */ diff --git a/src/gallium/state_trackers/python/tests/data.py b/src/gallium/state_trackers/python/tests/data.py deleted file mode 100644 index 3c1cda59d7..0000000000 --- a/src/gallium/state_trackers/python/tests/data.py +++ /dev/null @@ -1,285 +0,0 @@ -#!/usr/bin/env python -########################################################################## -# -# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. -# All Rights Reserved. -# -# Permission is hereby granted, free of charge, to any person obtaining a -# copy of this software and associated documentation files (the -# "Software"), to deal in the Software without restriction, including -# without limitation the rights to use, copy, modify, merge, publish, -# distribute, sub license, and/or sell copies of the Software, and to -# permit persons to whom the Software is furnished to do so, subject to -# the following conditions: -# -# The above copyright notice and this permission notice (including the -# next paragraph) shall be included in all copies or substantial portions -# of the Software. -# -# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS -# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF -# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. -# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR -# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, -# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE -# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. -# -########################################################################## - -"""Texture data generation for use in testing.""" - - -import random - -from gallium import * - - -dxt1_rgb_data = [ - ( - [ - [0x99, 0xb0, 0x8e, 0xff], - [0x5d, 0x62, 0x89, 0xff], - [0x99, 0xb0, 0x8e, 0xff], - [0x99, 0xb0, 0x8e, 0xff], - [0xd6, 0xff, 0x94, 0xff], - [0x5d, 0x62, 0x89, 0xff], - [0x99, 0xb0, 0x8e, 0xff], - [0xd6, 0xff, 0x94, 0xff], - [0x5d, 0x62, 0x89, 0xff], - [0x5d, 0x62, 0x89, 0xff], - [0x99, 0xb0, 0x8e, 0xff], - [0x21, 0x14, 0x84, 0xff], - [0x5d, 0x62, 0x89, 0xff], - [0x21, 0x14, 0x84, 0xff], - [0x21, 0x14, 0x84, 0xff], - [0x99, 0xb0, 0x8e, 0xff], - ], - [0xf2, 0xd7, 0xb0, 0x20, 0xae, 0x2c, 0x6f, 0x97], - ), - ( - [ - [0xb5, 0xcf, 0x9c, 0xff], - [0x83, 0x8c, 0x8b, 0xff], - [0x21, 0x08, 0x6b, 0xff], - [0x83, 0x8c, 0x8b, 0xff], - [0x52, 0x4a, 0x7b, 0xff], - [0x83, 0x8c, 0x8b, 0xff], - [0x83, 0x8c, 0x8b, 0xff], - [0xb5, 0xcf, 0x9c, 0xff], - [0x21, 0x08, 0x6b, 0xff], - [0xb5, 0xcf, 0x9c, 0xff], - [0x83, 0x8c, 0x8b, 0xff], - [0x52, 0x4a, 0x7b, 0xff], - [0xb5, 0xcf, 0x9c, 0xff], - [0x83, 0x8c, 0x8b, 0xff], - [0x52, 0x4a, 0x7b, 0xff], - [0x83, 0x8c, 0x8b, 0xff], - ], - [0x73, 0xb6, 0x4d, 0x20, 0x98, 0x2b, 0xe1, 0xb8], - ), - ( - [ - [0x00, 0x2c, 0xff, 0xff], - [0x94, 0x8d, 0x7b, 0xff], - [0x4a, 0x5c, 0xbd, 0xff], - [0x4a, 0x5c, 0xbd, 0xff], - [0x4a, 0x5c, 0xbd, 0xff], - [0x94, 0x8d, 0x7b, 0xff], - [0x94, 0x8d, 0x7b, 0xff], - [0x94, 0x8d, 0x7b, 0xff], - [0xde, 0xbe, 0x39, 0xff], - [0x94, 0x8d, 0x7b, 0xff], - [0xde, 0xbe, 0x39, 0xff], - [0xde, 0xbe, 0x39, 0xff], - [0xde, 0xbe, 0x39, 0xff], - [0xde, 0xbe, 0x39, 0xff], - [0xde, 0xbe, 0x39, 0xff], - [0x94, 0x8d, 0x7b, 0xff], - ], - [0xe7, 0xdd, 0x7f, 0x01, 0xf9, 0xab, 0x08, 0x80], - ), - ( - [ - [0x6b, 0x24, 0x21, 0xff], - [0x7b, 0x4f, 0x5d, 0xff], - [0x7b, 0x4f, 0x5d, 0xff], - [0x8b, 0x7a, 0x99, 0xff], - [0x7b, 0x4f, 0x5d, 0xff], - [0x7b, 0x4f, 0x5d, 0xff], - [0x6b, 0x24, 0x21, 0xff], - [0x8b, 0x7a, 0x99, 0xff], - [0x9c, 0xa6, 0xd6, 0xff], - [0x6b, 0x24, 0x21, 0xff], - [0x7b, 0x4f, 0x5d, 0xff], - [0x8b, 0x7a, 0x99, 0xff], - [0x6b, 0x24, 0x21, 0xff], - [0x8b, 0x7a, 0x99, 0xff], - [0x7b, 0x4f, 0x5d, 0xff], - [0x9c, 0xa6, 0xd6, 0xff], - ], - [0x3a, 0x9d, 0x24, 0x69, 0xbd, 0x9f, 0xb4, 0x39], - ), -] - -dxt1_rgba = [ - ( - [ - [0x00, 0x00, 0x00, 0x00], - [0x4e, 0xaa, 0x90, 0xff], - [0x4e, 0xaa, 0x90, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x4e, 0xaa, 0x90, 0xff], - [0x29, 0xff, 0xff, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x4e, 0xaa, 0x90, 0xff], - [0x73, 0x55, 0x21, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x4e, 0xaa, 0x90, 0xff], - [0x4e, 0xaa, 0x90, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x4e, 0xaa, 0x90, 0xff], - ], - [0xff, 0x2f, 0xa4, 0x72, 0xeb, 0xb2, 0xbd, 0xbe], - ), - ( - [ - [0xb5, 0xe3, 0x63, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x6b, 0x24, 0x84, 0xff], - [0xb5, 0xe3, 0x63, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0xb5, 0xe3, 0x63, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x6b, 0x24, 0x84, 0xff], - [0x6b, 0x24, 0x84, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0xb5, 0xe3, 0x63, 0xff], - [0x90, 0x83, 0x73, 0xff], - [0xb5, 0xe3, 0x63, 0xff], - ], - [0x30, 0x69, 0x0c, 0xb7, 0x4d, 0xf7, 0x0f, 0x67], - ), - ( - [ - [0x00, 0x00, 0x00, 0x00], - [0xc6, 0x86, 0x8c, 0xff], - [0xc6, 0x86, 0x8c, 0xff], - [0x21, 0x65, 0x42, 0xff], - [0x21, 0x65, 0x42, 0xff], - [0x21, 0x65, 0x42, 0xff], - [0x21, 0x65, 0x42, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x21, 0x65, 0x42, 0xff], - [0xc6, 0x86, 0x8c, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0xc6, 0x86, 0x8c, 0xff], - ], - [0x28, 0x23, 0x31, 0xc4, 0x17, 0xc0, 0xd3, 0x7f], - ), - ( - [ - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0xc6, 0xe3, 0x9c, 0xff], - [0x7b, 0x1c, 0x52, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x00, 0x00, 0x00, 0x00], - [0x7b, 0x1c, 0x52, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0x7b, 0x1c, 0x52, 0xff], - [0xa0, 0x7f, 0x77, 0xff], - [0xc6, 0xe3, 0x9c, 0xff], - [0x00, 0x00, 0x00, 0x00], - [0xa0, 0x7f, 0x77, 0xff], - ], - [0xea, 0x78, 0x13, 0xc7, 0x7f, 0xfc, 0x33, 0xb6], - ), -] - - - -def generate_data_compressed(surface, blocks): - - stride = surface.nblocksx*surface.block.size - size = surface.nblocksy*stride - - raw = ByteArray(size) - rgba = FloatArray(surface.height*surface.width*4) - - for yj in range(0, surface.nblocksy): - for xj in range(0, surface.nblocksx): - pixels, block = blocks[random.randint(0, len(blocks) - 1)] - - offset = (yj*surface.nblocksx + xj)*surface.block.size - for i in range(0, surface.block.size): - raw[offset + i] = block[i] - - for yi in range(0, surface.block.height): - for xi in range(0, surface.block.width): - y = yj*surface.block.height + yi - x = xj*surface.block.width + xi - if y < surface.height and x < surface.width: - offset = (y*surface.width + x)*4 - pixel = pixels[yi*surface.block.width + xi] - for ch in range(0, 4): - rgba[offset + ch] = float(pixel[ch])/255.0 - - surface.put_tile_raw(0, 0, surface.width, surface.height, raw, stride) - - return rgba - - -def generate_data_simple(surface): - stride = surface.nblocksx*surface.block.size - size = surface.nblocksy*stride - - raw = ByteArray(size) - rgba = FloatArray(surface.height*surface.width*4) - - for i in range(0, size): - raw[i] = random.randint(0, 255) - - surface.put_tile_raw(0, 0, surface.width, surface.height, raw, stride) - - surface.get_tile_rgba(0, 0, surface.width, surface.height, rgba) - - if surface.format in (PIPE_FORMAT_YCBCR, PIPE_FORMAT_YCBCR_REV): - # normalize - for y in range(0, surface.height): - for x in range(0, surface.width): - for ch in range(4): - offset = (y*surface.width + x)*4 + ch - rgba[offset] = min(max(rgba[offset], 0.0), 1.0) - - return rgba - - -def generate_data(surface): - width = surface.width - height = surface.height - - if surface.format == PIPE_FORMAT_DXT1_RGB: - return generate_data_compressed(surface, dxt1_rgb_data) - if surface.format == PIPE_FORMAT_DXT1_RGBA: - return generate_data_compressed(surface, dxt1_rgba_data) - if surface.format == PIPE_FORMAT_DXT3_RGBA: - assert 0 - if surface.format == PIPE_FORMAT_DXT5_RGBA: - assert 0 - - return generate_data_simple(surface) - - - diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index 3d76953126..2c9231349f 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -30,29 +30,6 @@ import sys from gallium import * from base import * -from data import generate_data - - -def compare_rgba(width, height, rgba1, rgba2, tol=4.0/256, ratio=0.85): - errors = 0 - for y in range(0, height): - for x in range(0, width): - differs = 0 - for ch in range(4): - offset = (y*width + x)*4 + ch - v1 = rgba1[offset] - v2 = rgba2[offset] - if abs(v1 - v2) > tol: - if errors == 0: - sys.stderr.write("x=%u, y=%u, ch=%u differ: %f vs %f\n" % (x, y, ch, v1, v2)) - if errors == 1 and ch == 0: - sys.stderr.write("...\n") - differs = 1 - errors += differs - total = height*width - if errors: - sys.stderr.write("%u out of %u pixels differ\n" % (errors, total)) - return float(total - errors)/float(total) >= ratio def lods(*dims): @@ -214,10 +191,12 @@ class TextureTest(TestCase): height=height, last_level = last_level) - expected_rgba = generate_data(texture.get_surface( + expected_rgba = FloatArray(height*width*4) + texture.get_surface( usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE, face = face, - level = level)) + level = level, + ).sample_rgba(expected_rgba) ctx.set_sampler_texture(0, texture) -- cgit v1.2.3 From 2727702b1731a478de8806481416080d02af5862 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 17 Jul 2008 20:19:40 +0200 Subject: tgsi: New file. --- src/gallium/auxiliary/tgsi/util/tgsi_sanity.c | 258 ++++++++++++++++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_sanity.h | 49 +++++ 2 files changed, 307 insertions(+) create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_sanity.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_sanity.h (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c new file mode 100644 index 0000000000..377309c3fd --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c @@ -0,0 +1,258 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "tgsi_sanity.h" +#include "tgsi_transform.h" + +#define MAX_REGISTERS 256 + +typedef uint reg_flag; + +#define BITS_IN_REG_FLAG (sizeof( reg_flag ) * 8) + +struct sanity_check_ctx +{ + struct tgsi_transform_context xform; + + reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REGISTERS / sizeof( uint ) / 8]; + reg_flag regs_used[TGSI_FILE_COUNT][MAX_REGISTERS / sizeof( uint ) / 8]; + uint num_imms; + uint num_instructions; + uint index_of_END; + + uint errors; + uint warnings; +}; + +static void +report_error( + struct sanity_check_ctx *ctx, + const char *msg ) +{ + debug_printf( "\nError: %s", msg ); + ctx->errors++; +} + +static void +report_warning( + struct sanity_check_ctx *ctx, + const char *msg ) +{ + debug_printf( "\nWarning: %s", msg ); + ctx->warnings++; +} + +static boolean +check_file_name( + struct sanity_check_ctx *ctx, + uint file ) +{ + if (file <= TGSI_FILE_NULL || file >= TGSI_FILE_COUNT) { + report_error( ctx, "Invalid file name" ); + return FALSE; + } + return TRUE; +} + +static boolean +is_register_declared( + struct sanity_check_ctx *ctx, + uint file, + uint index ) +{ + assert( index < MAX_REGISTERS ); + + return (ctx->regs_decl[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; +} + +static boolean +is_register_used( + struct sanity_check_ctx *ctx, + uint file, + uint index ) +{ + assert( index < MAX_REGISTERS ); + + return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; +} + +static void xform_instruction( + struct tgsi_transform_context *xform, + struct tgsi_full_instruction *inst ) +{ + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) xform; + uint i; + + /* There must be no other instructions after END. + */ + if (ctx->index_of_END != ~0) { + report_error( ctx, "Unexpected instruction after END" ); + } + else if (inst->Instruction.Opcode == TGSI_OPCODE_END) { + ctx->index_of_END = ctx->num_instructions; + } + + /* Check destination and source registers' validity. + * Mark the registers as used. + */ + for (i = 0; i < inst->Instruction.NumDstRegs; i++) { + uint file; + uint index; + + file = inst->FullDstRegisters[i].DstRegister.File; + if (!check_file_name( ctx, file )) + return; + index = inst->FullDstRegisters[i].DstRegister.Index; + if (!is_register_declared( ctx, file, index )) + report_error( ctx, "Undeclared destination register" ); + ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG)); + } + for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { + uint file; + uint index; + + file = inst->FullSrcRegisters[i].SrcRegister.File; + if (!check_file_name( ctx, file )) + return; + index = inst->FullSrcRegisters[i].SrcRegister.Index; + if (!is_register_declared( ctx, file, index )) + report_error( ctx, "Undeclared source register" ); + ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG)); + } + + ctx->num_instructions++; +} + +static void xform_declaration( + struct tgsi_transform_context *xform, + struct tgsi_full_declaration *decl ) +{ + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) xform; + uint file; + uint i; + + /* No declarations allowed after the first instruction. + */ + if (ctx->num_instructions > 0) + report_error( ctx, "Instruction expected but declaration found" ); + + /* Check registers' validity. + * Mark the registers as declared. + */ + file = decl->Declaration.File; + if (!check_file_name( ctx, file )) + return; + for (i = decl->DeclarationRange.First; i <= decl->DeclarationRange.Last; i++) { + if (is_register_declared( ctx, file, i )) + report_error( ctx, "The same register declared twice" ); + ctx->regs_decl[file][i / BITS_IN_REG_FLAG] |= (1 << (i % BITS_IN_REG_FLAG)); + } +} + +static void xform_immediate( + struct tgsi_transform_context *xform, + struct tgsi_full_immediate *imm ) +{ + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) xform; + + assert( ctx->num_imms < MAX_REGISTERS ); + + /* No immediates allowed after the first instruction. + */ + if (ctx->num_instructions > 0) + report_error( ctx, "Instruction expected but immediate found" ); + + /* Mark the register as declared. + */ + ctx->num_imms++; + ctx->regs_decl[TGSI_FILE_IMMEDIATE][ctx->num_imms / BITS_IN_REG_FLAG] |= (1 << (ctx->num_imms % BITS_IN_REG_FLAG)); + + /* Check data type validity. + */ + if (imm->Immediate.DataType != TGSI_IMM_FLOAT32) { + report_error( ctx, "Invalid immediate data type" ); + return; + } +} + +static void epilog( + struct tgsi_transform_context *xform ) +{ + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) xform; + uint file; + + /* There must be an END instruction at the end. + */ + if (ctx->index_of_END == ~0 || ctx->index_of_END != ctx->num_instructions - 1) { + report_error( ctx, "Expected END at end of instruction sequence" ); + } + + /* Check if all declared registers were used. + */ + for (file = TGSI_FILE_NULL; file < TGSI_FILE_COUNT; file++) { + uint i; + + for (i = 0; i < MAX_REGISTERS; i++) { + if (is_register_declared( ctx, file, i ) && !is_register_used( ctx, file, i )) { + report_warning( ctx, "Register never used" ); + } + } + } + + /* Print totals, if any. + */ + if (ctx->errors || ctx->warnings) + debug_printf( "\n%u errors, %u warnings", ctx->errors, ctx->warnings ); +} + +boolean +tgsi_sanity_check( + struct tgsi_token *tokens ) +{ + struct sanity_check_ctx ctx; + struct tgsi_token dummy_tokens[16]; + + ctx.xform.transform_instruction = xform_instruction; + ctx.xform.transform_declaration = xform_declaration; + ctx.xform.transform_immediate = xform_immediate; + ctx.xform.epilog = epilog; + + memset( ctx.regs_decl, 0, sizeof( ctx.regs_decl ) ); + memset( ctx.regs_used, 0, sizeof( ctx.regs_used ) ); + ctx.num_imms = 0; + ctx.num_instructions = 0; + ctx.index_of_END = ~0; + + ctx.errors = 0; + ctx.warnings = 0; + + if (tgsi_transform_shader( tokens, dummy_tokens, 16, &ctx.xform ) == -1) + return FALSE; + + return ctx.errors > 0; +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.h b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.h new file mode 100644 index 0000000000..ca45e94c7a --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.h @@ -0,0 +1,49 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_SANITY_H +#define TGSI_SANITY_H + +#include "pipe/p_shader_tokens.h" + +#if defined __cplusplus +extern "C" { +#endif + +/* Check the given token stream for errors and common mistakes. + * Diagnostic messages are printed out to the debug output. + * Returns TRUE if there are no errors, even though there could be some warnings. + */ +boolean +tgsi_sanity_check( + struct tgsi_token *tokens ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_SANITY_H */ -- cgit v1.2.3 From 3c5ec98e45880d6a896f3c094f8004000f50a149 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 17 Jul 2008 20:22:47 +0200 Subject: scons: List util/tgsi_sanity.c. --- src/gallium/auxiliary/tgsi/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript index b62c8efe02..67230f6ce0 100644 --- a/src/gallium/auxiliary/tgsi/SConscript +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -8,6 +8,7 @@ tgsi = env.ConvenienceLibrary( 'util/tgsi_build.c', 'util/tgsi_dump.c', 'util/tgsi_parse.c', + 'util/tgsi_sanity.c', 'util/tgsi_scan.c', 'util/tgsi_text.c', 'util/tgsi_transform.c', -- cgit v1.2.3 From 10d1dc68a413eaf642bf1bda2e5452835f7b7050 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 17 Jul 2008 20:23:26 +0200 Subject: tgsi: Perform a sanity check after reading a shader from text. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 6cab475b88..803f7a23b8 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -29,6 +29,7 @@ #include "tgsi_text.h" #include "tgsi_build.h" #include "tgsi_parse.h" +#include "tgsi_sanity.h" #include "tgsi_util.h" static boolean is_alpha_underscore( const char *cur ) @@ -1064,5 +1065,8 @@ tgsi_text_translate( ctx.tokens_cur = tokens; ctx.tokens_end = tokens + num_tokens; - return translate( &ctx ); + if (!translate( &ctx )) + return FALSE; + + return tgsi_sanity_check( tokens ); } -- cgit v1.2.3 From f2053cfa1ac4e4e2e0083670aac5df766adad5f9 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 17 Jul 2008 20:40:13 +0200 Subject: tgsi: Fix parsing an instruction with no operands. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 15 ++++++++++++--- 1 file changed, 12 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 803f7a23b8..9ed0f52fc7 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -758,13 +758,20 @@ parse_instruction( for (i = 0; i < TGSI_OPCODE_LAST; i++) { const char *cur = ctx->cur; - if (str_match_no_case( &cur, opcode_info[i].mnemonic )) { + info = &opcode_info[i]; + if (str_match_no_case( &cur, info->mnemonic )) { if (str_match_no_case( &cur, "_SATNV" )) saturate = TGSI_SAT_MINUS_PLUS_ONE; else if (str_match_no_case( &cur, "_SAT" )) saturate = TGSI_SAT_ZERO_ONE; - if (*cur == '\0' || eat_white( &cur )) { + if (info->num_dst + info->num_src + info->is_tex == 0) { + if (!is_digit_alpha_underscore( cur )) { + ctx->cur = cur; + break; + } + } + else if (*cur == '\0' || eat_white( &cur )) { ctx->cur = cur; break; } @@ -777,7 +784,6 @@ parse_instruction( report_error( ctx, "Expected `DCL', `IMM' or a label" ); return FALSE; } - info = &opcode_info[i]; inst = tgsi_default_full_instruction(); inst.Instruction.Opcode = i; @@ -1031,6 +1037,9 @@ static boolean translate( struct translate_ctx *ctx ) return FALSE; } + if (*ctx->cur == '\0') + break; + if (parse_label( ctx, &label_val )) { if (!parse_instruction( ctx, TRUE )) return FALSE; -- cgit v1.2.3 From 638ecbda3eed8319ce82be9c119ca1f96f21d976 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 18 Jul 2008 00:39:23 +0200 Subject: tgsi: Add tgsi_iterate_shader utility. Walks the token stream and invokes callbacks. --- src/gallium/auxiliary/tgsi/util/tgsi_iterate.c | 82 ++++++++++++++++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_iterate.h | 73 +++++++++++++++++++++++ 2 files changed, 155 insertions(+) create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_iterate.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_iterate.h (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c new file mode 100644 index 0000000000..43d7a6b1d5 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c @@ -0,0 +1,82 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "tgsi_iterate.h" + +boolean +tgsi_iterate_shader( + const struct tgsi_token *tokens, + struct tgsi_iterate_context *ctx ) +{ + struct tgsi_parse_context parse; + + if (tgsi_parse_init( &parse, tokens ) != TGSI_PARSE_OK) + return FALSE; + + if (ctx->prolog) + if (!ctx->prolog( ctx )) + goto fail; + + while (!tgsi_parse_end_of_tokens( &parse )) { + tgsi_parse_token( &parse ); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + if (ctx->iterate_instruction) + if (!ctx->iterate_instruction( ctx, &parse.FullToken.FullInstruction )) + goto fail; + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + if (ctx->iterate_declaration) + if (!ctx->iterate_declaration( ctx, &parse.FullToken.FullDeclaration )) + goto fail; + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + if (ctx->iterate_immediate) + if (!ctx->iterate_immediate( ctx, &parse.FullToken.FullImmediate )) + goto fail; + break; + + default: + assert( 0 ); + } + } + + if (ctx->epilog) + if (!ctx->epilog( ctx )) + goto fail; + + tgsi_parse_free( &parse ); + return TRUE; + +fail: + tgsi_parse_free( &parse ); + return FALSE; +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h new file mode 100644 index 0000000000..18729c2d1a --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h @@ -0,0 +1,73 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_ITERATE_H +#define TGSI_ITERATE_H + +#include "pipe/p_shader_tokens.h" +#include "tgsi/util/tgsi_parse.h" + +#if defined __cplusplus +extern "C" { +#endif + +struct tgsi_iterate_context +{ + boolean + (* prolog)( + struct tgsi_iterate_context *ctx ); + + boolean + (* iterate_instruction)( + struct tgsi_iterate_context *ctx, + struct tgsi_full_instruction *inst ); + + boolean + (* iterate_declaration)( + struct tgsi_iterate_context *ctx, + struct tgsi_full_declaration *decl ); + + boolean + (* iterate_immediate)( + struct tgsi_iterate_context *ctx, + struct tgsi_full_immediate *imm ); + + boolean + (* epilog)( + struct tgsi_iterate_context *ctx ); +}; + +boolean +tgsi_iterate_shader( + const struct tgsi_token *tokens, + struct tgsi_iterate_context *ctx ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_ITERATE_H */ -- cgit v1.2.3 From e1fd3ea415db1bc80d0bc6d94645eaac60bb4121 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 18 Jul 2008 00:40:28 +0200 Subject: scons: List util/tgsi_iterate.c. --- src/gallium/auxiliary/tgsi/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript index 67230f6ce0..54f5c75db4 100644 --- a/src/gallium/auxiliary/tgsi/SConscript +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -7,6 +7,7 @@ tgsi = env.ConvenienceLibrary( 'exec/tgsi_sse2.c', 'util/tgsi_build.c', 'util/tgsi_dump.c', + 'util/tgsi_iterate.c', 'util/tgsi_parse.c', 'util/tgsi_sanity.c', 'util/tgsi_scan.c', -- cgit v1.2.3 From 1e5419fa3061386413a98b75d0908cb3e3c6894e Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 18 Jul 2008 00:41:45 +0200 Subject: tgsi: Express tgsi_sanity in terms of tgsi_iterate. --- src/gallium/auxiliary/tgsi/util/tgsi_sanity.c | 60 ++++++++++++++++----------- 1 file changed, 36 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c index 377309c3fd..933127e661 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c @@ -27,7 +27,7 @@ #include "pipe/p_debug.h" #include "tgsi_sanity.h" -#include "tgsi_transform.h" +#include "tgsi_iterate.h" #define MAX_REGISTERS 256 @@ -37,7 +37,7 @@ typedef uint reg_flag; struct sanity_check_ctx { - struct tgsi_transform_context xform; + struct tgsi_iterate_context iter; reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REGISTERS / sizeof( uint ) / 8]; reg_flag regs_used[TGSI_FILE_COUNT][MAX_REGISTERS / sizeof( uint ) / 8]; @@ -101,11 +101,12 @@ is_register_used( return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; } -static void xform_instruction( - struct tgsi_transform_context *xform, +static boolean +iter_instruction( + struct tgsi_iterate_context *iter, struct tgsi_full_instruction *inst ) { - struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) xform; + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; uint i; /* There must be no other instructions after END. @@ -126,7 +127,7 @@ static void xform_instruction( file = inst->FullDstRegisters[i].DstRegister.File; if (!check_file_name( ctx, file )) - return; + return TRUE; index = inst->FullDstRegisters[i].DstRegister.Index; if (!is_register_declared( ctx, file, index )) report_error( ctx, "Undeclared destination register" ); @@ -138,7 +139,7 @@ static void xform_instruction( file = inst->FullSrcRegisters[i].SrcRegister.File; if (!check_file_name( ctx, file )) - return; + return TRUE; index = inst->FullSrcRegisters[i].SrcRegister.Index; if (!is_register_declared( ctx, file, index )) report_error( ctx, "Undeclared source register" ); @@ -146,13 +147,16 @@ static void xform_instruction( } ctx->num_instructions++; + + return TRUE; } -static void xform_declaration( - struct tgsi_transform_context *xform, +static boolean +iter_declaration( + struct tgsi_iterate_context *iter, struct tgsi_full_declaration *decl ) { - struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) xform; + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; uint file; uint i; @@ -166,19 +170,22 @@ static void xform_declaration( */ file = decl->Declaration.File; if (!check_file_name( ctx, file )) - return; + return TRUE; for (i = decl->DeclarationRange.First; i <= decl->DeclarationRange.Last; i++) { if (is_register_declared( ctx, file, i )) report_error( ctx, "The same register declared twice" ); ctx->regs_decl[file][i / BITS_IN_REG_FLAG] |= (1 << (i % BITS_IN_REG_FLAG)); } + + return TRUE; } -static void xform_immediate( - struct tgsi_transform_context *xform, +static boolean +iter_immediate( + struct tgsi_iterate_context *iter, struct tgsi_full_immediate *imm ) { - struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) xform; + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; assert( ctx->num_imms < MAX_REGISTERS ); @@ -196,14 +203,17 @@ static void xform_immediate( */ if (imm->Immediate.DataType != TGSI_IMM_FLOAT32) { report_error( ctx, "Invalid immediate data type" ); - return; + return TRUE; } + + return TRUE; } -static void epilog( - struct tgsi_transform_context *xform ) +static boolean +epilog( + struct tgsi_iterate_context *iter ) { - struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) xform; + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; uint file; /* There must be an END instruction at the end. @@ -228,6 +238,8 @@ static void epilog( */ if (ctx->errors || ctx->warnings) debug_printf( "\n%u errors, %u warnings", ctx->errors, ctx->warnings ); + + return TRUE; } boolean @@ -235,12 +247,12 @@ tgsi_sanity_check( struct tgsi_token *tokens ) { struct sanity_check_ctx ctx; - struct tgsi_token dummy_tokens[16]; - ctx.xform.transform_instruction = xform_instruction; - ctx.xform.transform_declaration = xform_declaration; - ctx.xform.transform_immediate = xform_immediate; - ctx.xform.epilog = epilog; + ctx.iter.prolog = NULL; + ctx.iter.iterate_instruction = iter_instruction; + ctx.iter.iterate_declaration = iter_declaration; + ctx.iter.iterate_immediate = iter_immediate; + ctx.iter.epilog = epilog; memset( ctx.regs_decl, 0, sizeof( ctx.regs_decl ) ); memset( ctx.regs_used, 0, sizeof( ctx.regs_used ) ); @@ -251,7 +263,7 @@ tgsi_sanity_check( ctx.errors = 0; ctx.warnings = 0; - if (tgsi_transform_shader( tokens, dummy_tokens, 16, &ctx.xform ) == -1) + if (tgsi_iterate_shader( tokens, &ctx.iter ) == -1) return FALSE; return ctx.errors > 0; -- cgit v1.2.3 From bfee84df8665a474fa025150f57d896bf1aa6bae Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 18 Jul 2008 10:11:29 +0900 Subject: python: Add methods to get texture width, height, etc. SWIG does a poor job with arrays inside structures. --- src/gallium/state_trackers/python/gallium.i | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index d5b24b04ff..c08ac87aca 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -401,6 +401,26 @@ error1: pipe_texture_reference(&ptr, NULL); } + unsigned get_width(unsigned level=0) { + return $self->width[level]; + } + + unsigned get_height(unsigned level=0) { + return $self->height[level]; + } + + unsigned get_depth(unsigned level=0) { + return $self->depth[level]; + } + + unsigned get_nblocksx(unsigned level=0) { + return $self->nblocksx[level]; + } + + unsigned get_nblocksy(unsigned level=0) { + return $self->nblocksy[level]; + } + /** Get a surface which is a "view" into a texture */ struct pipe_surface * get_surface(unsigned face=0, unsigned level=0, unsigned zslice=0, unsigned usage=0 ) -- cgit v1.2.3 From 3ca935d9efaef4dd9746d3b2207e329b77bb1018 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 18 Jul 2008 10:12:07 +0900 Subject: python/tests: Extend the texture tests to cover volumes. --- src/gallium/state_trackers/python/tests/texture.py | 65 ++++++++++++++-------- 1 file changed, 43 insertions(+), 22 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index 2c9231349f..16ad78c8aa 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -56,7 +56,12 @@ def tex_coords(texture, face, level, zslice): if texture.target == PIPE_TEXTURE_2D: return [[s, t, 0.0] for s, t in st] elif texture.target == PIPE_TEXTURE_3D: - assert 0 + depth = texture.get_depth(level) + if depth > 1: + r = float(zslice)/float(depth - 1) + else: + r = 0.0 + return [[s, t, r] for s, t in st] elif texture.target == PIPE_TEXTURE_CUBE: result = [] for s, t in st: @@ -112,10 +117,10 @@ class TextureTest(TestCase): }[self.face] else: face = "" - return "%s %s %ux%u last_level=%u %s level=%u" % ( + return "%s %s %ux%ux%u last_level=%u face=%s level=%u zslice=%u" % ( target, format, - self.width, self.height, - self.last_level, face, self.level, + self.width, self.height, self.depth, self.last_level, + face, self.level, self.zslice, ) def test(self): @@ -125,9 +130,11 @@ class TextureTest(TestCase): format = self.format width = self.width height = self.height + depth = self.depth last_level = self.last_level - level = self.level face = self.face + level = self.level + zslice = self.zslice if not dev.is_format_supported(format, PIPE_TEXTURE): raise TestSkip @@ -185,17 +192,21 @@ class TextureTest(TestCase): ctx.set_sampler(0, sampler) # texture - texture = dev.texture_create(target=target, - format=format, - width=width, - height=height, - last_level = last_level) + texture = dev.texture_create( + target = target, + format = format, + width = width, + height = height, + depth = depth, + last_level = last_level, + ) expected_rgba = FloatArray(height*width*4) texture.get_surface( usage = PIPE_BUFFER_USAGE_CPU_READ|PIPE_BUFFER_USAGE_CPU_WRITE, face = face, level = level, + zslice = zslice, ).sample_rgba(expected_rgba) ctx.set_sampler_texture(0, texture) @@ -263,7 +274,7 @@ class TextureTest(TestCase): [x, y+h], ] - tex = tex_coords(texture, face, level, zslice=0) + tex = tex_coords(texture, face, level, zslice) for i in range(0, 4): j = 8*i @@ -313,6 +324,7 @@ def main(): targets = [] targets += [PIPE_TEXTURE_2D] targets += [PIPE_TEXTURE_CUBE] + targets += [PIPE_TEXTURE_3D] formats = [] formats += [PIPE_FORMAT_A8R8G8B8_UNORM] @@ -322,7 +334,9 @@ def main(): formats += [PIPE_FORMAT_DXT1_RGB] sizes = [64, 32, 16, 8, 4, 2, 1] + #sizes = [1020, 508, 252, 62, 30, 14, 6, 3] #sizes = [64] + #sizes = [63] for target in targets: for format in formats: @@ -339,21 +353,28 @@ def main(): #faces = [PIPE_TEX_FACE_NEG_X] else: faces = [0] + if target == PIPE_TEXTURE_3D: + depth = size + else: + depth = 1 for face in faces: levels = lods(size) for last_level in range(levels): for level in range(0, last_level + 1): - test = TextureTest( - dev=dev, - target=target, - format=format, - width=size, - height=size, - face=face, - last_level = last_level, - level=level, - ) - suite.add_test(test) + for zslice in range(0, depth >> level): + test = TextureTest( + dev = dev, + target = target, + format = format, + width = size, + height = size, + depth = depth, + last_level = last_level, + face = face, + level = level, + zslice = zslice, + ) + suite.add_test(test) suite.run() -- cgit v1.2.3 From 583b9ccbd5961678fa7e594c3117d7c895ef6b41 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 18 Jul 2008 22:22:22 +0900 Subject: gallium: Add a pf_is_ycbcr utility function. --- src/gallium/include/pipe/p_format.h | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index fefb8fdf1b..947c445583 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -522,15 +522,13 @@ pf_get_nblocks(const struct pipe_format_block *block, unsigned width, unsigned h static INLINE boolean pf_is_compressed( enum pipe_format format ) { - switch (format) { - case PIPE_FORMAT_DXT1_RGB: - case PIPE_FORMAT_DXT1_RGBA: - case PIPE_FORMAT_DXT3_RGBA: - case PIPE_FORMAT_DXT5_RGBA: - return TRUE; - default: - return FALSE; - } + return pf_layout(format) == PIPE_FORMAT_LAYOUT_DXT ? TRUE : FALSE; +} + +static INLINE boolean +pf_is_ycbcr( enum pipe_format format ) +{ + return pf_layout(format) == PIPE_FORMAT_LAYOUT_YCBCR ? TRUE : FALSE; } #ifdef __cplusplus -- cgit v1.2.3 From b64882d0f8674df3e4e60f5786eb48027518c279 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 18 Jul 2008 12:51:39 -0600 Subject: mesa: fix set_program_uniform_matrix(): need to loop over matrix count --- src/mesa/shader/shader_api.c | 42 ++++++++++++++++++++++++------------------ 1 file changed, 24 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 53bf2108aa..ee0e369d7c 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -1319,37 +1319,43 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, static void set_program_uniform_matrix(GLcontext *ctx, struct gl_program *program, - GLuint location, GLuint rows, GLuint cols, + GLuint location, GLuint count, + GLuint rows, GLuint cols, GLboolean transpose, const GLfloat *values) { /* * Note: the _columns_ of a matrix are stored in program registers, not - * the rows. + * the rows. So, the loops below look a little funny. + * XXX could optimize this a bit... */ - /* XXXX need to test 3x3 and 2x2 matrices... */ - if (transpose) { - GLuint row, col; - for (col = 0; col < cols; col++) { - GLfloat *v = program->Parameters->ParameterValues[location + col]; - for (row = 0; row < rows; row++) { - v[row] = values[row * cols + col]; - } - } - } - else { - GLuint row, col; + GLuint mat, row, col; + GLuint dst = location, src = 0; + + /* loop over matrices */ + for (mat = 0; mat < count; mat++) { + + /* each matrix: */ for (col = 0; col < cols; col++) { - GLfloat *v = program->Parameters->ParameterValues[location + col]; + GLfloat *v = program->Parameters->ParameterValues[dst]; for (row = 0; row < rows; row++) { - v[row] = values[col * rows + row]; + if (transpose) { + v[row] = values[src + row * cols + col]; + } + else { + v[row] = values[src + col * rows + row]; + } } + dst++; } + + src += rows * cols; /* next matrix */ } } /** * Called by ctx->Driver.UniformMatrix(). + * Note: cols=2, rows=4 ==> array[2] of vec4 */ static void _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, @@ -1382,7 +1388,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, GLint loc = shProg->Uniforms->Uniforms[location].VertPos; if (loc >= 0) { set_program_uniform_matrix(ctx, &shProg->VertexProgram->Base, - loc, rows, cols, transpose, values); + loc, count, rows, cols, transpose, values); } } @@ -1390,7 +1396,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, GLint loc = shProg->Uniforms->Uniforms[location].FragPos; if (loc >= 0) { set_program_uniform_matrix(ctx, &shProg->FragmentProgram->Base, - loc, rows, cols, transpose, values); + loc, count, rows, cols, transpose, values); } } } -- cgit v1.2.3 From 2b3515c5b21a417cf32c5259b93da4deb2a951c6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 18 Jul 2008 12:52:24 -0600 Subject: mesa: fix some GLSL /= int operators plus add a few more special constructors to improve code quality. --- src/mesa/shader/slang/library/slang_core.gc | 28 +++++++++++++++++++++------- 1 file changed, 21 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 0fd2b7e6b3..2c3a5a6c45 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -176,6 +176,11 @@ vec2 __constructor(const bool b) __retVal.xy = b.xx; } +vec2 __constructor(const bvec2 b) +{ + __retVal = b; +} + vec2 __constructor(const vec3 v) { __retVal.xy = v.xy; @@ -211,6 +216,11 @@ vec3 __constructor(const bool b) __retVal.xyz = b.xxx; } +vec3 __constructor(const bvec3 b) +{ + __retVal = b; +} + vec3 __constructor(const vec4 v) { __retVal.xyz = v.xyz; @@ -242,6 +252,11 @@ vec4 __constructor(const bool b) __retVal = b.xxxx; } +vec4 __constructor(const bvec4 b) +{ + __retVal = b; +} + vec4 __constructor(const vec3 v3, const float f) { // XXX this constructor shouldn't be needed anymore @@ -360,8 +375,7 @@ bvec2 __constructor(const int i) bvec2 __constructor(const vec2 v) { - const vec2 zero = vec2(0.0, 0.0); - __asm vec4_sne __retVal.xy, v, zero; + __asm vec4_sne __retVal.xy, v, 0.0; } bvec2 __constructor(const ivec2 v) @@ -1226,7 +1240,7 @@ void __operator /= (inout int a, const int b) float invB; __asm float_rcp invB, b; __asm vec4_multiply a, a, invB; - __asm float_to_int __retVal, a; + __asm float_to_int a, a; } @@ -1253,7 +1267,7 @@ void __operator /= (inout ivec2 v, const ivec2 u) __asm float_rcp inv.x, u.x; __asm float_rcp inv.y, u.y; __asm vec4_multiply z, v, inv; - __asm float_to_int __retVal, z; + __asm float_to_int v, z; } @@ -1280,7 +1294,7 @@ void __operator /= (inout ivec3 v, const ivec3 u) __asm float_rcp inv.x, u.x; __asm float_rcp inv.y, u.y; __asm vec4_multiply z, v, inv; - __asm float_to_int __retVal, z; + __asm float_to_int v, z; } @@ -1307,7 +1321,7 @@ void __operator /= (inout ivec4 v, const ivec4 u) __asm float_rcp inv.x, u.x; __asm float_rcp inv.y, u.y; __asm vec4_multiply z, v, inv; - __asm float_to_int __retVal, z; + __asm float_to_int v, z; } @@ -1315,7 +1329,7 @@ void __operator /= (inout ivec4 v, const ivec4 u) void __operator += (inout float a, const float b) { - __asm vec4_add a.x, a.x, b; + __asm vec4_add a.x, a.x, b.x; } void __operator -= (inout float a, const float b) -- cgit v1.2.3 From a4fd90e3255ea531f2fe0693f00f342f4610cae3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 18 Jul 2008 12:53:31 -0600 Subject: mesa: regenerated file --- src/mesa/shader/slang/library/slang_core_gc.h | 1580 ++++++++++++------------- 1 file changed, 790 insertions(+), 790 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index 09ceb69edd..f71fe28cda 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -17,807 +17,807 @@ 1,0,10,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,102,0,59,120,120,0, 20,0,0,1,0,10,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,105,0,59, 120,120,0,20,0,0,1,0,10,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98, -0,59,120,120,0,20,0,0,1,0,10,1,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121, -0,18,118,0,59,120,121,0,20,0,0,1,0,10,1,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -115,116,0,18,118,0,59,120,121,0,20,0,0,1,0,11,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,0,1,0,11,1,1,1,0,9, -102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,102,0,59,120,120,120,0,20,0,0,1, -0,11,1,1,1,0,5,105,0,0,0,1,4,105,110,116,95,116,111,95,102,108,111,97,116,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,105,0,59,120,120,120,0,0,0,0,1,0,11,1,1,1,0,1,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,11,1,1,1,0,12,118, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,118,0,59,120,121,122,0,20,0,0,1,0, -12,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122,0,0,1,1,0,9,119,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18, -119,0,20,0,0,1,0,12,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,59,120,120, -120,120,0,20,0,0,1,0,12,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,59,120, -120,120,120,0,20,0,0,1,0,12,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,59,120, -120,120,120,0,20,0,0,1,0,12,1,1,1,0,11,118,51,0,0,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,18,118,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,0,20,0, -0,1,0,12,1,1,1,0,10,118,50,0,0,1,1,0,9,102,49,0,0,1,1,0,9,102,50,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,18,118,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,102,49,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,50,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,1,1,0,5, -106,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,18,106,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,18,105,0,59,120,120,0,20,0,0,1,0,6,1,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,116,111, -95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,59,120,120,0,0,0,0,1,0,6, -1,1,1,0,1,98,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,7,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,1,1,0,5, -107,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,107,0,20,0,0,1,0,7,1,1, -1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,105,0,59,120,120,120,0,20, -0,0,1,0,7,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,102,0,59, -120,120,120,0,20,0,0,1,0,7,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122, -0,18,98,0,59,120,120,120,0,20,0,0,1,0,8,1,1,1,0,5,120,0,0,1,1,0,5,121,0,0,1,1,0,5,122,0,0,1,1,0,5, -119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,8,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,8,1,1,1,0,9,102,0,0,0,1,4,102,108,111,97, -116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,59,120,120,120,120,0,0, -0,0,1,0,8,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,59,120,120,120,120,0,20, -0,0,1,0,2,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18, -98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,0,2,1,1,1,0,1,98,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,59,120,120,0,20,0,0,1,0,2,1,1,1,0,9, -102,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0, -0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,59, -120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,3,2,1,6,1,122,101,114,111,0,2, -58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,105,0,59,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1, -0,10,118,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0,48,0,0,0,17,48,0,48,0,0, -0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0, -0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,3,2,1,6,1,122,101,114,111,0,2,58,105,118, -101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98, -50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18, -98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18, -98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101, -99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,59,120,120,120,0,0,18,122,101,114,111, -0,0,0,0,1,0,3,1,1,1,0,5,105,0,0,0,1,3,2,1,7,1,122,101,114,111,0,2,58,105,118,101,99,51,0,16,8,48,0, -0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,122,0,0,18,105,0,59,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,11,118,0,0, -0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48, -0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, -18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,7,118,0,0,0,1,3,2,1,7,1,122,101,114,111,0,2,58, -105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1, -0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,1,1,0,1,98,52,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0, -18,98,52,0,20,0,0,1,0,4,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119, -0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,4,1,1,1,0,9,102,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2, -58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,59,120,120,120,120,0,0,18, -122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5,105,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101, -99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,0,18,105,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1, -1,1,0,12,118,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48, -0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, -116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,8,118,0,0,0,1,3,2,1,8,1,122, -101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0, -0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121, -0,18,109,49,49,0,20,0,0,1,0,13,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,59,120,0,18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,17,48,0,48,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,17,48,0,48,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,0,1,0,13,1,1,1,0,5,105,0,0,0,1,8,58,109, -97,116,50,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,13,1,1,1,0,1,98,0,0,0,1,8,58,109,97, -116,50,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,13,1,1,1,0,10,99,48,0,0,1,1,0,10,99,49,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,14,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9, -109,50,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,48,50, -0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49, -48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48, -50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,0,1,0,14,1,1,1,0,9,102,0,0,0,1, -3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59,121, -121,120,0,20,0,0,1,0,14,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,51,0,58,102,108,111,97,116,0,18,105, -0,0,0,0,0,0,0,1,0,14,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,51,0,58,102,108,111,97,116,0,18,98,0,0,0, -0,0,0,0,1,0,14,1,1,1,0,11,99,48,0,0,1,1,0,11,99,49,0,0,1,1,0,11,99,50,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,0,15,1,1,1,0,9, -109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,51,48,0,0,1,1,0,9,109,48,49, -0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,51,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0, -9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,1,1,0,9,109,51,50,0,0,1,1,0,9,109,48,51,0,0,1,1,0,9,109,49, -51,0,0,1,1,0,9,109,50,51,0,0,1,1,0,9,109,51,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109, -49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,109,51,48,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50, -49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0,18,109,51,49,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,119,0,18,109,51, -50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0,18,109,48,51,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,109,49,51,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57, -59,119,0,18,109,51,51,0,20,0,0,1,0,15,1,1,1,0,9,102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0, -18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120, -121,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,121,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59,121,121,120,121,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0,59,121,121,121,120,0,20,0,0,1,0,15,1,1,1,0,5,105, -0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,15,1,1,1,0,1,98,0,0, -0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,15,1,1,1,0,12,99,48,0,0,1, -1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,1,1,0,12,99,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -51,0,57,18,99,51,0,20,0,0,1,0,5,2,26,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118, -101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105, -110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5,2,27,1,1,0,5,97,0,0,1,1,0,5,98,0, -0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0, +0,59,120,120,0,20,0,0,1,0,10,1,1,1,0,2,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0, +0,1,0,10,1,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,118,0,59,120, +121,0,20,0,0,1,0,10,1,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,115,116,0,18,118, +0,59,120,121,0,20,0,0,1,0,11,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,0,1,0,11,1,1,1,0,9,102,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,18,102,0,59,120,120,120,0,20,0,0,1,0,11,1,1,1,0,5,105, +0,0,0,1,4,105,110,116,95,116,111,95,102,108,111,97,116,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,122,0,0,18,105,0,59,120,120,120,0,0,0,0,1,0,11,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,11,1,1,1,0,3,98,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,11,1,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,18,118,0,59,120,121,122,0,20,0,0,1,0,12,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1, +1,0,9,122,0,0,1,1,0,9,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0, +18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,12,1,1,1,0,9,102,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,59,120,120,120,120,0,20,0,0,1,0,12,1,1,1,0,5,105,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,12,1,1,1,0,1,98, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,12,1,1,1,0,4,98, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,12,1,1,1,0,11,118,51,0,0,1,1,0,9,102, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,118,51,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,119,0,18,102,0,20,0,0,1,0,12,1,1,1,0,10,118,50,0,0,1,1,0,9,102,49,0,0,1,1,0,9, +102,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,118,50,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,122,0,18,102,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,50,0, +20,0,0,1,0,6,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18, +105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,0,1, +9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,105,0,59,120,120,0,20,0,0,1,0,6,1,1,1,0,9,102,0, +0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +0,0,18,102,0,59,120,120,0,0,0,0,1,0,6,1,1,1,0,1,98,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105, +110,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,7,1,1,1,0, +5,105,0,0,1,1,0,5,106,0,0,1,1,0,5,107,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,122,0,18,107,0,20,0,0,1,0,7,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,18,105,0,59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,18,102,0,59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,8,1,1,1,0,5,120,0,0,1,1,0, +5,121,0,0,1,1,0,5,122,0,0,1,1,0,5,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,8,1,1,1,0,5, +105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,8,1,1,1,0, +9,102,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0, +18,102,0,59,120,120,120,120,0,0,0,0,1,0,8,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +18,98,0,59,120,120,120,120,0,20,0,0,1,0,2,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50, +0,20,0,0,1,0,2,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,59,120, +120,0,20,0,0,1,0,2,1,1,1,0,9,102,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0, +48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0, +59,120,121,0,0,18,102,0,59,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,3,2, +1,6,1,122,101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,59,120,120,0,0,18,122,101, +114,111,0,0,0,0,1,0,2,1,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,3,2,1,6,1, +122,101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1, +0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,3,2,1, +11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0, +0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,59, +120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,5,105,0,0,0,1,3,2,1,7,1,122,101,114,111,0, +2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,59,120,120,120,0,0,18,122,101,114, +111,0,0,0,0,1,0,3,1,1,1,0,11,118,0,0,0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0, +48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,7,118,0,0,0, +1,3,2,1,7,1,122,101,114,111,0,2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4, +118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18, +122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,1,1,0,1,98, +52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,4,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,119,0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,4,1,1,1,0,9,102,0, +0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0, +48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0, +0,18,102,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5,105,0,0,0,1,3,2,1,8,1, +122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4, +118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,59,120,120,120,120,0,0, +18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,12,118,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101, +99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52, +95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1, +1,1,0,8,118,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0, +16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0, +18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9, +109,48,49,0,0,1,1,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0, +18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,0,13,1,1,1,0,9,102,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0, +17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,0,1,0, +13,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,13, +1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,13,1,1,1, +0,10,99,48,0,0,1,1,0,10,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,14,1,1,1,0,9,109,48,48, +0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0, +9,109,50,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, +0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0, +18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0, +0,1,0,14,1,1,1,0,9,102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0, +0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,50,0,57,18,118,0,59,121,121,120,0,20,0,0,1,0,14,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,51,0, +58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,14,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,51,0,58, +102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,14,1,1,1,0,11,99,48,0,0,1,1,0,11,99,49,0,0,1,1,0,11, +99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, +99,50,0,20,0,0,1,0,15,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9, +109,51,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,51,49, +0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,1,1,0,9,109,51,50,0,0,1,1,0, +9,109,48,51,0,0,1,1,0,9,109,49,51,0,0,1,1,0,9,109,50,51,0,0,1,1,0,9,109,51,51,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, +0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,109,51,48,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0, +18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0, +18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,109,49,51,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0,0,1,0,15,1,1,1,0,9,102,0,0,0,1,3,2,0,10, +1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0, +16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, +18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59, +121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0,59,121,121,121, +120,0,20,0,0,1,0,15,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,105,0,0, +0,0,0,0,0,1,0,15,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0, +0,0,1,0,15,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,1,1,0,12,99,51,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,5,2,26,1,1,0,5,97,0,0,1,1,0,5,98, +0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102, +108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5, +2,27,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114, +97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18, +95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0, +9,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0, +0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, +0,1,0,5,2,22,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108, +111,97,116,95,114,99,112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105, +110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,26,1,1,0,6,97,0,0,1,1,0,6,98,0, +0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102, +108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6, +2,27,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114, +97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18, +95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,21,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0, +10,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0, +0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, +0,1,0,6,2,22,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116, +111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,26,1,1,0,7,97,0,0,1, +1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0, +0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0, +0,0,1,0,7,2,27,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,115,117, +98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110, +116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,21,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0, +1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0, 18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,0,0,1,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, -95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5,2,22,1,1,0,5,97,0,0,1,1,0, -5,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98, -73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0, -0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86, -97,108,0,0,18,120,0,0,0,0,1,0,6,2,26,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118, -101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105, -110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,27,1,1,0,6,97,0,0,1,1,0,6,98,0, -0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0, +120,0,0,0,0,1,0,7,2,22,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,98,73,110,118,0,0,1,1,120,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116, +111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,26,1,1,0,8,97,0,0,1, +1,0,8,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0, +0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0, +0,0,1,0,8,2,27,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,115,117, +98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110, +116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,21,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0, +1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0, 18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,0,0,1,0,6,2,21,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, -95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,22,1,1,0,6,97,0,0,1,1,0, -6,98,0,0,0,1,3,2,0,10,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98, -73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118, -0,59,121,0,0,18,98,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0, -18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, -116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,26,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0, -4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, -95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,27,1,1,0,7,97,0,0,1,1,0, -7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18, -97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108, -0,0,18,120,0,0,0,0,1,0,7,2,21,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99, -52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95, -116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,22,1,1,0,7,97,0, -0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112, -0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73, -110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0, -59,122,0,0,18,98,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18, -97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, -116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,26,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,11,1,120,0,0,0, -4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, -95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,27,1,1,0,8,97,0,0,1,1,0, -8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18, -97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108, -0,0,18,120,0,0,0,0,1,0,8,2,21,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118,101,99, -52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95, -116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,22,1,1,0,8,97,0, -0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112, -0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73, -110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0, -59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,119,0,0, -18,98,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18, -98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97, -108,0,0,18,120,0,0,0,0,1,0,9,2,26,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100, -0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,1, -1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108, -0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,21,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0, -0,0,0,1,0,9,2,22,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,73,110,118,0, -0,0,0,1,0,10,2,26,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1, -0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118, +120,0,0,0,0,1,0,8,2,22,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,98,73,110,118,0,0,1,1,120,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114, +99,112,0,18,98,73,110,118,0,59,119,0,0,18,98,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105, +110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,2,26,1,1,0,9,97,0,0,1,1,0,9,98,0, +0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98, +0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,21,1,1,0,9,97,0, +0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86, +97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,22,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1, +98,73,110,118,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59, +120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118, +101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0, +0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,21,1,1,0, +10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1,0,10, +117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0, +59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118, 101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -118,0,0,18,117,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,119,0,0,0,0,1,0,11,2, -26,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,1,1,0,11,117,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -118,0,0,18,117,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,119,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,119,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0, -0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0, -0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98, -116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0, -12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1, -3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0,1,0,10,2,26,1,1,0,9,97,0,0,1,1,0,10,117, -0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59, -120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101, -99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18, -98,0,59,120,120,0,0,0,0,1,0,10,2,27,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117, -98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18, -117,0,59,120,121,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0, -18,98,0,59,120,120,0,0,0,0,1,0,10,2,21,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0, -0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120, -121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,22,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,105, -110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59, -120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0, -0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -0,0,18,97,0,59,120,120,0,0,18,105,110,118,85,0,59,120,121,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1, -0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, -66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,66,0,59,120,120,0,0,0,0,1,0,11,2,26,1, -1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,26,1,1,0,11, +118,0,0,18,119,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100, +100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,27,1, +1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1, +0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0, +1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111, +97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,119,0,0,0, +0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114, +101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1, +4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, +117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,22,1,1,0, +12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0, +59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0, +59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0,1,0,10, +2,26,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,26,1,1,0,10, 118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,27,1,1,0,9,97,0,0, -1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,27,1, -1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0, -1,0,11,2,21,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120, -121,122,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0, -18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,22,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,105,110, -118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0, -0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -97,0,59,120,120,120,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1, -1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, -118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,66,0,59,120,120,120,0,0, -0,0,1,0,12,2,26,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114, -101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1, -1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, -98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,27,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95, -115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0, -18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116, -114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1, -0,12,2,21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, -18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0, -12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,22,1,1,0,9,97,0,0,1,1, -0,12,117,0,0,0,1,3,2,0,12,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, +121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,27,1,1,0,9,97,0,0,1,1,0,10,117, +0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,9,98, +0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,21,1,1,0,9,97,0,0,1,1,0,10,117, +0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0, +9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0, +59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,22,1,1,0,9,97,0,0,1,1,0, +10,117,0,0,0,1,3,2,0,10,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, +85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59, +121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,105,110,118,85,0,59,120,121,0,0,0,0,1, +0,10,2,22,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,66,0, +59,120,120,0,0,0,0,1,0,11,2,26,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120,121, +122,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0, +0,0,0,1,0,11,2,27,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59, +120,121,122,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0, +0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,21,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52, +95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0, +59,120,120,120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1, +4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,22,1,1,0,9,97,0,0,1,1, +0,11,117,0,0,0,1,3,2,0,11,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, 118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0, 59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0, -0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,119,0,0,18,117, -0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, -0,0,18,97,0,59,120,120,120,120,0,0,18,105,110,118,85,0,0,0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9, -98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0, -0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, -0,0,18,118,0,0,18,105,110,118,66,0,59,120,120,120,120,0,0,0,0,1,0,6,2,26,1,1,0,5,97,0,0,1,1,0,6, -117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,46,20,0, -0,1,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105, -118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,6,2,27,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,6,2,27,1,1,0,6,118,0, -0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0, -0,47,20,0,0,1,0,6,2,21,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, -105,118,101,99,50,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,6,2,21,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,48,20,0,0,1,0,6,2,22, -1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97, -0,0,0,18,117,0,49,20,0,0,1,0,6,2,22,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,49,20,0,0,1,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7, -117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,46,20,0, -0,1,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105, -118,101,99,51,0,18,98,0,0,0,46,20,0,0,1,0,7,2,27,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,7,2,27,1,1,0,7,118,0, -0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0, -0,47,20,0,0,1,0,7,2,21,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, -105,118,101,99,51,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,7,2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,48,20,0,0,1,0,7,2,22, -1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97, -0,0,0,18,117,0,49,20,0,0,1,0,7,2,22,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,49,20,0,0,1,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8, -117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,46,20,0, -0,1,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105, -118,101,99,52,0,18,98,0,0,0,46,20,0,0,1,0,8,2,27,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,8,2,27,1,1,0,8,118,0, -0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0, -0,47,20,0,0,1,0,8,2,21,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, -105,118,101,99,52,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,8,2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,48,20,0,0,1,0,8,2,22, -1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97, -0,0,0,18,117,0,49,20,0,0,1,0,8,2,22,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,49,20,0,0,1,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118, -101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1, -0,6,2,27,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86, -97,108,0,0,18,118,0,0,0,0,1,0,7,2,27,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101, -0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,8,2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52, -95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,9,2,27,1,1,0,9, -97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, -18,97,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0, -0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, -0,18,118,0,59,120,121,122,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,110,101,103, -97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,13,2,27,1,1,0,13,109,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, -0,57,18,109,0,16,10,50,0,57,54,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10, -50,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,54,20,0,0, -1,0,9,0,100,111,116,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0, -18,98,0,48,20,0,0,1,0,9,0,100,111,116,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0,59,121,0,18,98,0,59,121,0,48,46,20,0,0,1, -0,9,0,100,111,116,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,51,95,100,111,116,0,18,95,95, -114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0,100,111,116,0,1,1,0,12,97,0,0,1,1,0,12, -98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0, -0,0,1,0,0,2,1,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,97,0,0,18,97,0, -0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114, -97,99,116,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5, -97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18, -97,0,0,18,105,110,118,66,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, -116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95, -97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4, -118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0, -2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18, -118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,3,2,0,6,1,105,110,118,0,0,1,1, -122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99, -52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,102,108,111, -97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0, -2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0, -0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, -18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0, -0,1,1,0,7,117,0,0,0,1,3,2,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, -118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122, -0,0,18,118,0,0,18,105,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114, -101,116,86,97,108,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52, -95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1, -4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1, -0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0, -18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,3,2,0,8,1,105,110,118,0,0, -1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,102,108, -111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,122,0,0,0,0,1,0,0,2,1, -1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59, -120,0,0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98, -116,114,97,99,116,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98, -0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0, -0,0,0,1,0,0,2,4,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,119,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,119,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -97,0,59,120,0,0,18,97,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101, -99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0, -1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, -18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,3,1,0,2,10,118, -0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121, -0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117,0, -0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59, -120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,119, -0,59,120,121,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100, -0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,11,117,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18, -117,0,0,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1, -0,11,117,0,0,0,1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18, -117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,119,0, -59,120,121,122,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100, -100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118, -101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2, -12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18, -118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102, +0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0, +0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105, +110,118,66,0,59,120,120,120,0,0,0,0,1,0,12,2,26,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99, +52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0, +0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101, +116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,27,1,1,0,9,97,0,0,1,1,0, +12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0, +0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4, +118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, +98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0, +0,18,117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0, +0,0,1,0,12,2,22,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,105,110,118,85,0,0,0,4,102,108,111, +97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114, +99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,105,110,118,85,0,0, +0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111, +97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,59,120,120,120,120, +0,0,0,0,1,0,6,2,26,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105, +118,101,99,50,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,6,2,27,1,1, +0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0, +0,18,117,0,47,20,0,0,1,0,6,2,27,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,47,20,0,0,1,0,6,2,21,1,1,0,5,97,0,0,1,1,0,6,117, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,48,20,0,0,1, +0,6,2,21,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118, +101,99,50,0,18,98,0,0,0,48,20,0,0,1,0,6,2,22,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,6,2,22,1,1,0,6,118,0,0,1, +1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,49, +20,0,0,1,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, +101,99,51,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,46,20,0,0,1,0,7,2,27,1,1,0, +5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0, +18,117,0,47,20,0,0,1,0,7,2,27,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,47,20,0,0,1,0,7,2,21,1,1,0,5,97,0,0,1,1,0,7,117,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,7, +2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, +99,51,0,18,98,0,0,0,48,20,0,0,1,0,7,2,22,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,7,2,22,1,1,0,7,118,0,0,1,1,0, +5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,49,20, +0,0,1,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, +101,99,52,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,46,20,0,0,1,0,8,2,27,1,1,0, +5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0, +18,117,0,47,20,0,0,1,0,8,2,27,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,47,20,0,0,1,0,8,2,21,1,1,0,5,97,0,0,1,1,0,8,117,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,8, +2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, +99,52,0,18,98,0,0,0,48,20,0,0,1,0,8,2,22,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,8,2,22,1,1,0,8,118,0,0,1,1,0, +5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,49,20, +0,0,1,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,6,2,27,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,110,101,103, +97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,7,2,27,1,1,0,7,118,0,0,0,1,4, +118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,8, +2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97, +108,0,0,18,118,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0, +18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,0,1,4,118, +101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59, +120,121,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,0,0,1,0,12,2,27,1,1,0,12, +118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118, +0,0,0,0,1,0,13,2,27,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20, +0,0,1,0,14,2,27,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8, +48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,0,1,0,15,2,27,1,1,0, +15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,0,9,0,100,111,116,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,48,20,0,0,1,0,9,0,100,111,116,0,1,1,0,10,97,0,0,1, +1,0,10,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0, +59,121,0,18,98,0,59,121,0,48,46,20,0,0,1,0,9,0,100,111,116,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4, +118,101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0, +100,111,116,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,18,95,95,114, +101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,1,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118, +101,99,52,95,97,100,100,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,5,97,0,0,1,1,0,5,98,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2, +3,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0, +18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, +4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,105,110,118,66,0,0,0,4,102,108,111,97,116,95,116, +111,95,105,110,116,0,18,97,0,0,18,97,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118, +101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,6, +117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0, +1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, +18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,3,2,0,6,1,105, +110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117, +0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0, +0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0, +0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,7,118, +0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0, +2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0, +0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0, +7,117,0,0,0,1,3,2,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105, +110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0, +59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0, +18,118,0,0,18,105,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,118,0,0,18,122, +0,0,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0, +18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,115,117, +98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,8,117, +0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0, +1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,3,2,0,8,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108, +111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,102,108,111,97,116,95,116,111, +95,105,110,116,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101, +99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,0,1,0,0,2,2,1,0,2, +9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,59,120,0,0, +18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,9,97,0,0,1,1,0,9, +98,0,0,0,1,3,2,0,9,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,98,0,0, +0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,119,0,0,0,0, +1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120, +121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,10, +117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59, +120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117, +0,59,120,121,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102, 108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119, -0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117, -0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0, -0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120, -121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,5,97,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120, -121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120, -120,0,0,0,9,18,118,0,59,120,0,18,97,0,23,0,9,18,118,0,59,121,0,18,97,0,23,0,0,1,0,0,2,4,1,0,2,6, -118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,0,1,0, -0,2,1,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0, -0,18,118,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,5,97, -0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,59, -120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122, -0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120, -0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2, -8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120, -120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116, -114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0, -1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18, -97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0, -18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,9,18,118,0,59,119,0, -18,97,0,24,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, -118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,9,97,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0, -59,120,120,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,4,1,0, -2,10,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112, -0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, -59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,65,0,59,120,120,0,0,0,0,1,0,0,2,1,1,0,2,11, -118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0, -18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, -115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0, -0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121, -0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0, -1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, -118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121, -122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,65,0,59,120,120,120,0,0,0,0,1,0,0,2,1,1,0,2,12, -118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120, -120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98, -116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,12, -118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118, -0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105, +114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,119,0,59,120,121,0,0,0,0,1,0,0,2,1,1, +0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18, +118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117, +98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,11, +118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120, +121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1, +119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108, +111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114, +99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,119,0,59,120,121,122,0,0,0,0,1,0, +0,2,1,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0, +18,117,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116, +114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0, +1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0, +2,4,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112, +0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0, +0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52, +95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0, +0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0, +0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0, +0,0,1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,9,18,118,0,59,120,0, +18,97,0,23,0,9,18,118,0,59,121,0,18,97,0,23,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,9,18, +118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,7,118,0,0,1,1,0,5, +97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0, +18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,59,120, +120,120,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0, +0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59, +121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1, +4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,2, +1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18, +118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0, +0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121, +0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,9,18,118,0,59,119,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2, +10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,0, +18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,3, +1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, +59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1, +3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0, +0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120, +121,0,0,18,105,110,118,65,0,59,120,120,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118, +101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1, +0,0,2,2,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, +118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1, +0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18, +118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105, 110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,105,110,118,65,0,59,120,120, -120,120,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13, -109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, +99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0, +18,105,110,118,65,0,59,120,120,120,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101, +99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,12, +118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0, +18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52, +95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0, +0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95, +114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,118,0,0,18,118,0,0,18,105,110,118,65,0,59,120,120,120,120,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0, +0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,0, +13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, +109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0, +57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57, +18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,0, +48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0, +0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50, +0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20, +0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0, +16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0, +48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10, +49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,0,48,46, +18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120,120,0,48,18,109, +0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16, +10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49, +20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15, +109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, 18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48, -18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49, -0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49, -0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1, -0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, -16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9, +57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10, +50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, +16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, 18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120, -120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50, -0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122, -122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110, -0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0, -48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1, -0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, -16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0, -57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10, -51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2, -21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48, -0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0, -48,46,18,109,0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120, -0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, -57,18,110,0,16,10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57, -59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48, -0,57,18,110,0,16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57, -59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48, -46,18,109,0,16,10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120, -0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, -57,18,110,0,16,10,51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, -59,119,119,119,119,0,48,46,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49, -20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, -0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0, -13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13, -2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97, +120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,121,0,48,46,18,109,0, +16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16, +8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, +0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16, +10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122, +122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120, +120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,121,0,48,46,18,109,0,16, +10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10, +50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, +16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10, +51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,51,0,57,59,122,122,122, +122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59,119,119,119,119,0,48,46,20,0,0,1,0,15, +2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, +0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, +57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0, +13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13, +2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, +57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97, 0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48, -0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20, -0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2,26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0, -1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, -47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2, -27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47, -20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, -0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0, -16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9, +0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20, +0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, +16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9, 98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9, -97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8, -48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1, -1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2, -27,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, -110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, -49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0, -15,2,27,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, -0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, -0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0, -0,1,0,15,2,21,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, -18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18, -110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16, -10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57, -48,20,0,0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16, -10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57, -18,98,0,48,20,0,0,1,0,15,2,22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0, -18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0, -16,10,51,0,57,49,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, -16,10,51,0,57,18,98,0,49,20,0,0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0, -59,121,121,0,48,46,20,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21, -1,1,0,14,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18, -118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50, -0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, -57,0,0,20,0,0,1,0,12,2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121, -121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57, -18,118,0,59,119,119,119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, -57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10, -51,0,57,0,0,20,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110, -0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109, -0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0, -57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109, -0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14, -109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2, -2,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22, -0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2, -4,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24, -0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, -50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1, -0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51, -0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18, -109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10, -50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2, -1,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, -57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22, -0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0, -9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2, -1,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, -57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0, -0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50, -0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, -23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2, -14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97, -0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18, -109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18, -97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9, -18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57, -18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1, -9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0, -57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2, +26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, +110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, +49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46, +20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, +0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50, +0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14, +110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0, +14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, +48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0, +1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, +97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110, +0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50, +0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, +57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109, +0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0, +57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2,27,1,1,0,9,97,0,0,1,1,0,15,110,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,9, +98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0,0,1,0,15,2,21,1,1,0,9,97,0,0, +1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, +48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,15,2,21,1,1,0, +15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, +48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,0,15,2, +22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, +110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, +49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,49,20,0,0,1,0, +15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, +0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, +0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,49,20,0, +0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16, +8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,0,48,46,20,0,0,1,0, +10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100, +111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, +100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21,1,1,0,14,109,0,0,1,1,0,11, +118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,0,48, +18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122, +122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,12, +2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0, +57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,121,0,48,46,18, +109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,118,0,59,119,119, +119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0, +0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21, +0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,13,110,0, +0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10, +49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20, +0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0,14, +110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0, +16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2,2,1,0,2,14,109,0, +0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,0,0,2,3,1,0,2, +14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,14,109,0,0, +1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,0,0,2,1,1,0,2, +15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16, +10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18, +109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1, +9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, +22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10, +51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20, +0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, +50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1, +0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,0,0, +2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0, +57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23, +0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, +16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0, +9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0, +16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57, +18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,0,0,2, +3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0, +57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,14,109,0,0,1,1,0,9,97,0,0, 0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50, -0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109, -0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9, -18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18, -118,0,18,109,0,48,20,0,0,1,0,5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58, -105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, -7,2,25,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0, -58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1, -0,9,2,25,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,18,97,0,20,0,0,1,0,10,2,25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49, -0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0, -0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,118,0,20,0,0,1,0,12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0, -17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13, -109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47, -20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0, -9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16, -8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49, -0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,18,109,0,20,0,0,1,0,15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, -118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118, -101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2, -24,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -97,0,20,0,0,1,0,6,2,24,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18, -118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,0,1,0,8,2,24,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0, -0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18, -97,0,18,97,0,17,49,0,48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24, -1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95, -114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, -12,2,24,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9, -18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,109,0,20,0,0,1,0,14,2,24,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, -15,2,24,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, -49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, -48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101, -99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16, -10,49,0,47,20,0,0,1,0,6,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20, -0,0,1,0,7,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95, -95,112,111,115,116,68,101,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0, -20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115, -116,68,101,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0, -18,97,0,17,49,0,48,0,0,47,20,0,0,1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0, -48,0,0,0,0,47,20,0,0,1,0,11,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, -47,20,0,0,1,0,12,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0, -1,0,13,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0, -0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0, -0,47,20,0,0,1,0,14,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17, -49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0, -48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0, -0,0,0,47,20,0,0,1,0,15,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0, -17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, -49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, -48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115, -116,73,110,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118, -0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110, -99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0, -2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101, -99,52,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0, -95,95,112,111,115,116,73,110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112, -111,115,116,73,110,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, -18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73, -110,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, -118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114, -0,1,0,2,13,109,0,0,0,1,3,2,0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110, -99,114,0,1,0,2,14,109,0,0,0,1,3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110, -99,114,0,1,0,2,15,109,0,0,0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, -58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58, -118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0, -0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0, -0,0,0,1,0,1,2,15,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102, -108,111,97,116,0,18,98,0,0,0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0, -0,4,102,108,111,97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2, -16,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0, -18,98,0,0,0,41,0,0,1,0,1,2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4, -102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95, -101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0, -5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0, -0,0,43,0,0,1,0,1,2,17,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108, -111,97,116,95,108,101,115,115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113, -117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0, -1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0, -0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114, -105,110,116,0,18,102,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105, -110,116,95,112,114,105,110,116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1, -98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77, -69,83,65,0,1,1,0,10,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, -58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, -83,65,0,1,1,0,11,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58, -112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, -18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112, -114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1, -0,6,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105, -110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7, -118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0, -0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77, -69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0, -9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83, -65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58, -112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, -18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112, -114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112, -114,105,110,116,77,69,83,65,0,1,1,0,4,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, +0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10, +51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97, +0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16, +10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18, +97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,9,18,109,0, +16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57, +18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,9,18,109, +0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,118,0,18,118,0, +18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48, +20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0, +5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0, +18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49, +0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,25,1,0,2,7,118,0,0,0,1,9, +18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0, +18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49, +0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,25,1,0,2,9,97,0,0,0,1,9, +18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2, +25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0, +57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, +15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, +0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2,24,1,0,2,5,97,0,0,0,1,9,18,97, +0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,24,1,0,2,6, +118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101, +99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,24,1,0,2, +8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24,1,0,2,10,118,0,0,0,1,9,18,118,0, +18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0, +20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, +46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,12,2,24,1,0,2,12,118,0,0,0,1,9,18, +118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118, +101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101, +99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,24, +1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0, +0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0, +0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,15,2,24,1,0,2,15,109,0,0,0,1,9,18,109, +0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10, +50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0, +57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,47,20,0,0,1,0,6,0,95,95,112, +111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, +18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,0,1,0,7,0,95,95,112,111,115,116,68, +101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, +118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95,95,112,111,115,116,68,101,99,114, +0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105, +118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,9,97, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,0, +1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,11,0, +95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,12,0,95,95,112, +111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0, +9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,13,0,95,95,112,111,115,116, +68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0, +16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10, +49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,14,0,95,95,112, +111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0, +9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18, +109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109, +0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,15,0,95, +95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109, +0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0, +9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18, +109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109, +0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,9,0,95, +95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0, +20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2, +10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99, +50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,11,118,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49, +0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0, +46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0,95,95,112,111,115,116,73, +110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, +118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112,111,115,116,73,110,99,114, +0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105, +118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,8,118, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51, +0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,13,109,0,0,0,1,3,2, +0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49, +0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48, +0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,14,109,0,0,0,1, +3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0, +17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17, +49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0, +48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,15,109,0,0, +0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52, +0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, +103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0,0,0,0,1,0,1,2,15,1,1,0,5, +97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0, +0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0,0,4,102,108,111,97,116,95, +108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2,16,1,1,0,5,97,0,0,1,1,0,5, +98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,41,0,0,1,0,1, +2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108, +101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101, +0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1, +8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,43,0,0,1,0,1,2,17,1,1,0, +9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115, +115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18, +97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58, +102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0,0,1,0,0,0,112,114,105, +110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114,105,110,116,0,18,102,0,0, +0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105,110,116,95,112,114,105,110, +116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1,98,0,0,0,1,4,98,111,111,108, +95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,10,118,0,0, +0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69, +83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,11,118,0,0,0,1,9, +58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65, +0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0, +112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114, +105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, +59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,6,118,0,0,0,1,9,58,112,114,105, +110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59, +121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7,118,0,0,0,1,9,58,112,114,105,110, +116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0, +0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116, +77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, +58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65, +0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0, +112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112, +114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, 59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0, -0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77, -69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0, -57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116, -77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49, -0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114, -105,110,116,77,69,83,65,0,1,1,0,15,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16, -8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105, -110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17, -101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77, -69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112, -114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0, -0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105, -110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110, -116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 +116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,4,118, +0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77, +69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0, +9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, +83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9, +58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77, +69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0, +9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69, +83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,15,109,0,0, +0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77, +69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0, +57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105, +110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1, +0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116, +0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95, +112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0, +1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0, +1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110, +116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 -- cgit v1.2.3 From 7218626cf5ecc00fc7ef994253c20c9a2df35cc2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 18 Jul 2008 14:44:01 -0600 Subject: mesa: rework array/struct addressing code. The slang_ir_storage type now has a pointer to parent storage to represent storage of an array element within an array, or a field within a struct. This fixes some problems related to addressing of fields/elements in non- trivial cases. More work to follow. --- src/mesa/shader/slang/slang_builtin.c | 10 +- src/mesa/shader/slang/slang_codegen.c | 154 ++++++++++++++------- src/mesa/shader/slang/slang_emit.c | 240 ++++++++++++++++++++++----------- src/mesa/shader/slang/slang_emit.h | 4 - src/mesa/shader/slang/slang_ir.c | 63 +++++++++ src/mesa/shader/slang/slang_ir.h | 20 ++- src/mesa/shader/slang/slang_vartable.c | 23 +++- 7 files changed, 372 insertions(+), 142 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_builtin.c b/src/mesa/shader/slang/slang_builtin.c index f0f59aa87a..f99a59e3e1 100644 --- a/src/mesa/shader/slang/slang_builtin.c +++ b/src/mesa/shader/slang/slang_builtin.c @@ -35,6 +35,7 @@ #include "shader/prog_parameter.h" #include "shader/prog_statevars.h" #include "shader/slang/slang_ir.h" +#include "shader/slang/slang_emit.h" #include "shader/slang/slang_builtin.h" @@ -438,8 +439,13 @@ _slang_alloc_statevar(slang_ir_node *n, pos = lookup_statevar(var, index1, index2, field, &swizzle, paramList); assert(pos >= 0); if (pos >= 0) { - n0->Store->Index = pos; - n0->Store->Swizzle = swizzle; + /* XXX should overwrite Store's fields instead of changing pointers + * since there may be a child storage_info pointing to this one. + */ + n0->Store = _slang_new_ir_storage_swz(PROGRAM_STATE_VAR, + pos, + n0->Store->Size, + swizzle); } return pos; } diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 9d5721214f..6261aab23c 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -43,6 +43,7 @@ #include "shader/program.h" #include "shader/prog_instruction.h" #include "shader/prog_parameter.h" +#include "shader/prog_print.h" #include "shader/prog_statevars.h" #include "slang_typeinfo.h" #include "slang_codegen.h" @@ -243,7 +244,12 @@ _slang_attach_storage(slang_ir_node *n, slang_variable *var) } else { /* alloc new storage info */ - n->Store = _slang_new_ir_storage(PROGRAM_UNDEFINED, -1, -5); + n->Store = _slang_new_ir_storage(PROGRAM_UNDEFINED, -7, -5); +#if 0 + printf("%s var=%s Store=%p Size=%d\n", __FUNCTION__, + (char*) var->a_name, + (void*) n->Store, n->Store->Size); +#endif if (n->Var) n->Var->aux = n->Store; assert(n->Var->aux); @@ -658,6 +664,9 @@ new_var(slang_assemble_ctx *A, slang_operation *oper, slang_atom name) n = new_node0(IR_VAR); if (n) { _slang_attach_storage(n, var); + /* + printf("new_var %s store=%p\n", (char*)name, (void*) n->Store); + */ } return n; } @@ -935,6 +944,18 @@ slang_substitute(slang_assemble_ctx *A, slang_operation *oper, * __asm vec4_lrp __retVal, a, y, x; * } * + * + * A call to + * r = mix(p1, p2, p3); + * + * Becomes: + * + * mov + * / \ + * r vec4_lrp + * / | \ + * p3 p2 p1 + * * We basically translate a SLANG_OPER_CALL into a SLANG_OPER_ASM. */ static slang_operation * @@ -974,10 +995,10 @@ slang_inline_asm_function(slang_assemble_ctx *A, slang_operation_copy(inlined, &fun->body->children[0]); if (haveRetValue) { /* get rid of the __retVal child */ - for (i = 0; i < numArgs; i++) { + inlined->num_children--; + for (i = 0; i < inlined->num_children; i++) { inlined->children[i] = inlined->children[i + 1]; } - inlined->num_children--; } /* now do formal->actual substitutions */ @@ -1593,13 +1614,13 @@ _slang_gen_asm(slang_assemble_ctx *A, slang_operation *oper, if (info->NumParams == oper->num_children) { /* Storage for result is not specified. - * Children[0], [1] are the operands. + * Children[0], [1], [2] are the operands. */ firstOperand = 0; } else { /* Storage for result (child[0]) is specified. - * Children[1], [2] are the operands. + * Children[1], [2], [3] are the operands. */ firstOperand = 1; } @@ -1634,7 +1655,8 @@ _slang_gen_asm(slang_assemble_ctx *A, slang_operation *oper, n->Store = get_store(n0); n->Writemask = writemask; - assert(n->Store->File != PROGRAM_UNDEFINED); + assert(n->Store->File != PROGRAM_UNDEFINED || + n->Store->Parent); _slang_free(n0); } @@ -2071,7 +2093,7 @@ _slang_gen_temporary(GLint size) slang_ir_storage *store; slang_ir_node *n = NULL; - store = _slang_new_ir_storage(PROGRAM_TEMPORARY, -1, size); + store = _slang_new_ir_storage(PROGRAM_TEMPORARY, -2, size); if (store) { n = new_node0(IR_VAR_DECL); if (n) { @@ -2092,17 +2114,15 @@ static slang_ir_node * _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) { slang_ir_node *n; + /*assert(!var->declared);*/ var->declared = GL_TRUE; + assert(!is_sampler_type(&var->type)); + n = new_node0(IR_VAR_DECL); if (n) { _slang_attach_storage(n, var); -#if 0 - printf("%s var %p %s store=%p\n", - __FUNCTION__, (void *) var, (char *) var->a_name, - (void *) n->Store); -#endif assert(var->aux); assert(n->Store == var->aux); assert(n->Store); @@ -2110,6 +2130,13 @@ _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) n->Store->File = PROGRAM_TEMPORARY; n->Store->Size = _slang_sizeof_type_specifier(&n->Var->type.specifier); + +#if 0 + printf("%s var %p %s store=%p index=%d size=%d\n", + __FUNCTION__, (void *) var, (char *) var->a_name, + (void *) n->Store, n->Store->Index, n->Store->Size); +#endif + if (var->array_len > 0) { /* this is an array */ /* round up element size to mult of 4 */ @@ -2118,8 +2145,27 @@ _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) sz *= var->array_len; n->Store->Size = sz; } - A->program->NumTemporaries++; + assert(n->Store->Size > 0); + + /* setup default swizzle for storing the variable */ + switch (n->Store->Size) { + case 2: + n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, + SWIZZLE_NIL, SWIZZLE_NIL); + break; + case 3: + n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, + SWIZZLE_Z, SWIZZLE_NIL); + break; + default: + /* Note that float-sized vars may be allocated in any x/y/z/w + * slot, but that won't be determined until code emit time. + */ + n->Store->Swizzle = SWIZZLE_NOOP; + } + + A->program->NumTemporaries++; /* an approximation */ } return n; } @@ -2400,11 +2446,11 @@ _slang_gen_variable(slang_assemble_ctx * A, slang_operation *oper) static slang_ir_node * _slang_gen_swizzle(slang_ir_node *child, GLuint swizzle) { + /* XXX should rewrite this to use relative/Parent storage */ slang_ir_node *n = new_node1(IR_SWIZZLE, child); assert(child); if (n) { - n->Store = _slang_new_ir_storage(PROGRAM_UNDEFINED, -1, -1); - n->Store->Swizzle = swizzle; + n->Store = _slang_new_ir_storage_swz(PROGRAM_UNDEFINED, -1, -1, swizzle); } return n; } @@ -2495,7 +2541,7 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper) * Generate IR tree for referencing a field in a struct (or basic vector type) */ static slang_ir_node * -_slang_gen_field(slang_assemble_ctx * A, slang_operation *oper) +_slang_gen_struct_field(slang_assemble_ctx * A, slang_operation *oper) { slang_typeinfo ti; @@ -2548,7 +2594,8 @@ _slang_gen_field(slang_assemble_ctx * A, slang_operation *oper) /* oper->a_id is the field name */ slang_ir_node *base, *n; slang_typeinfo field_ti; - GLint fieldSize, fieldOffset = -1; + GLint fieldSize, fieldOffset = -1, swz; + /* type of field */ slang_typeinfo_construct(&field_ti); _slang_typeof_operation(A, oper, &field_ti); @@ -2573,20 +2620,27 @@ _slang_gen_field(slang_assemble_ctx * A, slang_operation *oper) } n = new_node1(IR_FIELD, base); - if (n) { - n->Field = (char *) oper->a_id; - n->FieldOffset = fieldOffset; - assert(n->FieldOffset >= 0); - n->Store = _slang_new_ir_storage(base->Store->File, - base->Store->Index, - fieldSize); - } - return n; + if (!n) + return NULL; -#if 0 - _mesa_problem(NULL, "glsl structs/fields not supported yet"); - return NULL; -#endif + + /* setup the storage info for this node */ + swz = fieldOffset % 4; + + n->Field = (char *) oper->a_id; + n->Store = _slang_new_ir_storage_relative(fieldOffset / 4, + fieldSize, + base->Store); + if (fieldSize == 1) + n->Store->Swizzle = MAKE_SWIZZLE4(swz, swz, swz, swz); + else if (fieldSize == 2) + n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, + SWIZZLE_NIL, SWIZZLE_NIL); + else if (fieldSize == 3) + n->Store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, + SWIZZLE_Z, SWIZZLE_NIL); + + return n; } } @@ -2595,7 +2649,7 @@ _slang_gen_field(slang_assemble_ctx * A, slang_operation *oper) * Gen code for array indexing. */ static slang_ir_node * -_slang_gen_subscript(slang_assemble_ctx * A, slang_operation *oper) +_slang_gen_array_element(slang_assemble_ctx * A, slang_operation *oper) { slang_typeinfo array_ti; @@ -2659,21 +2713,24 @@ _slang_gen_subscript(slang_assemble_ctx * A, slang_operation *oper) index = _slang_gen_operation(A, &oper->children[1]); if (array && index) { /* bounds check */ - if (index->Opcode == IR_FLOAT && - ((int) index->Value[0] < 0 || - (int) index->Value[0] >= arrayLen)) { - slang_info_log_error(A->log, + GLint constIndex = 0; + if (index->Opcode == IR_FLOAT) { + constIndex = (int) index->Value[0]; + if (constIndex < 0 || constIndex >= arrayLen) { + slang_info_log_error(A->log, "Array index out of bounds (index=%d size=%d)", - (int) index->Value[0], arrayLen); - _slang_free_ir_tree(array); - _slang_free_ir_tree(index); - return NULL; + constIndex, arrayLen); + _slang_free_ir_tree(array); + _slang_free_ir_tree(index); + return NULL; + } } elem = new_node2(IR_ELEMENT, array, index); - elem->Store = _slang_new_ir_storage(array->Store->File, - array->Store->Index, - elemSize); + elem->Store = _slang_new_ir_storage_relative(constIndex, + elemSize, + array->Store); + /* XXX try to do some array bounds checking here */ return elem; } @@ -2912,9 +2969,9 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) case SLANG_OPER_IF: return _slang_gen_if(A, oper); case SLANG_OPER_FIELD: - return _slang_gen_field(A, oper); + return _slang_gen_struct_field(A, oper); case SLANG_OPER_SUBSCRIPT: - return _slang_gen_subscript(A, oper); + return _slang_gen_array_element(A, oper); case SLANG_OPER_LITERAL_FLOAT: /* fall-through */ case SLANG_OPER_LITERAL_INT: @@ -3076,8 +3133,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, GLint index = _slang_input_index(varName, GL_FRAGMENT_PROGRAM_ARB, &swizzle); assert(index >= 0); - store = _slang_new_ir_storage(PROGRAM_INPUT, index, size); - store->Swizzle = swizzle; + store = _slang_new_ir_storage_swz(PROGRAM_INPUT, index, size, swizzle); assert(index < FRAG_ATTRIB_MAX); } else { @@ -3109,8 +3165,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, &swizzle); GLint size = 4; /* XXX? */ assert(index >= 0); - store = _slang_new_ir_storage(PROGRAM_INPUT, index, size); - store->Swizzle = swizzle; + store = _slang_new_ir_storage_swz(PROGRAM_INPUT, index, size, swizzle); } if (dbg) printf("ATTRIB "); } @@ -3119,8 +3174,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, GLint index = _slang_input_index(varName, GL_FRAGMENT_PROGRAM_ARB, &swizzle); GLint size = 4; /* XXX? */ - store = _slang_new_ir_storage(PROGRAM_INPUT, index, size); - store->Swizzle = swizzle; + store = _slang_new_ir_storage_swz(PROGRAM_INPUT, index, size, swizzle); if (dbg) printf("INPUT "); } else if (var->type.qualifier == SLANG_QUAL_FIXEDOUTPUT) { diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index c997c11a7d..ec8ed0b4e8 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -126,21 +126,6 @@ _slang_swizzle_swizzle(GLuint swz1, GLuint swz2) } -slang_ir_storage * -_slang_new_ir_storage(enum register_file file, GLint index, GLint size) -{ - slang_ir_storage *st; - st = (slang_ir_storage *) _slang_alloc(sizeof(slang_ir_storage)); - if (st) { - st->File = file; - st->Index = index; - st->Size = size; - st->Swizzle = SWIZZLE_NOOP; - } - return st; -} - - /** * Allocate temporary storage for an intermediate result (such as for * a multiply or add, etc. @@ -184,6 +169,27 @@ free_temp_storage(slang_var_table *vt, slang_ir_node *n) } + +/** + * Remove any SWIZZLE_NIL terms from given swizzle mask (smear prev term). + * Ex: fix_swizzle("zyNN") -> "zyyy" + * XXX should put in the default component for the position... + */ +static GLuint +fix_swizzle(GLuint swizzle) +{ + GLuint swz[4], i; + for (i = 0; i < 4; i++) { + swz[i] = GET_SWZ(swizzle, i); + if (swz[i] == SWIZZLE_NIL) { + swz[i] = swz[i - 1]; + } + } + return MAKE_SWIZZLE4(swz[0], swz[1], swz[2], swz[3]); +} + + + /** * Convert IR storage to an instruction dst register. */ @@ -191,14 +197,27 @@ static void storage_to_dst_reg(struct prog_dst_register *dst, const slang_ir_storage *st, GLuint writemask) { - assert(st->Index >= 0); - dst->File = st->File; - dst->Index = st->Index; + const GLint size = st->Size; + GLint index = st->Index; + GLuint swizzle = st->Swizzle; + + /* if this is storage relative to some parent storage, walk up the tree */ + while (st->Parent) { + st = st->Parent; + index += st->Index; + swizzle = _slang_swizzle_swizzle(st->Swizzle, swizzle); + } + assert(st->File != PROGRAM_UNDEFINED); - assert(st->Size >= 1); - assert(st->Size <= 4); - if (st->Size == 1) { - GLuint comp = GET_SWZ(st->Swizzle, 0); + dst->File = st->File; + + assert(index >= 0); + dst->Index = index; + + assert(size >= 1); + assert(size <= 4); + if (size == 1) { + GLuint comp = GET_SWZ(swizzle, 0); assert(comp < 4); dst->WriteMask = WRITEMASK_X << comp; } @@ -220,17 +239,35 @@ storage_to_src_reg(struct prog_src_register *src, const slang_ir_storage *st) MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W), MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W) }; + const GLint size = st->Size; + const GLboolean relAddr = st->RelAddr; + GLint index = st->Index; + GLuint swizzle = st->Swizzle; + + /* if this is storage relative to some parent storage, walk up the tree */ + while (st->Parent) { + st = st->Parent; + index += st->Index; + swizzle = _slang_swizzle_swizzle(st->Swizzle, swizzle); + } + swizzle = fix_swizzle(swizzle); + assert(st->File >= 0); assert(st->File < PROGRAM_UNDEFINED); - assert(st->Size >= 1); - assert(st->Size <= 4); src->File = st->File; - src->Index = st->Index; - src->RelAddr = st->RelAddr; - if (st->Swizzle != SWIZZLE_NOOP) - src->Swizzle = st->Swizzle; + + assert(index >= 0); + src->Index = index; + + assert(size >= 1); + assert(size <= 4); + + src->RelAddr = relAddr; + + if (swizzle != SWIZZLE_NOOP) + src->Swizzle = swizzle; else - src->Swizzle = defaultSwizzle[st->Size - 1]; /*XXX really need this?*/ + src->Swizzle = defaultSwizzle[size - 1]; /*XXX really need this?*/ assert(GET_SWZ(src->Swizzle, 0) <= 3); assert(GET_SWZ(src->Swizzle, 1) <= 3); @@ -549,6 +586,12 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) ? n->Children[0]->Store->Size : info->ResultSize; if (!alloc_temp_storage(emitInfo, n, size)) return NULL; +#if 0000 /* this should work, but doesn't yet */ + if (size == 2) + n->Writemask = WRITEMASK_XY; + else if (size == 3) + n->Writemask = WRITEMASK_XYZ; +#endif } storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); @@ -957,6 +1000,8 @@ emit_move(slang_emit_info *emitInfo, slang_ir_node *n) { struct prog_instruction *inst; + assert(n->Opcode == IR_MOVE); + /* lhs */ emit(emitInfo, n->Children[0]); if (!n->Children[0]->Store || n->Children[0]->Store->Index < 0) { @@ -995,8 +1040,15 @@ emit_move(slang_emit_info *emitInfo, slang_ir_node *n) if (n->Children[1]->Opcode != IR_SWIZZLE) _slang_free_temp(emitInfo->vt, n->Children[1]->Store); *n->Children[1]->Store = *n->Children[0]->Store; + /* fixup the previous instruction (which stored the RHS result) */ assert(n->Children[0]->Store->Index >= 0); + + /* use tighter writemask when possible */ +#if 0000 /* XXX enable this after more testing... */ + if (n->Writemask == WRITEMASK_XYZW) + n->Writemask = inst->DstReg.WriteMask; +#endif storage_to_dst_reg(&inst->DstReg, n->Children[0]->Store, n->Writemask); return inst; } @@ -1408,25 +1460,6 @@ emit_cont_break_if_true(slang_emit_info *emitInfo, slang_ir_node *n) } - -/** - * Remove any SWIZZLE_NIL terms from given swizzle mask (smear prev term). - * Ex: fix_swizzle("zyNN") -> "zyyy" - */ -static GLuint -fix_swizzle(GLuint swizzle) -{ - GLuint swz[4], i; - for (i = 0; i < 4; i++) { - swz[i] = GET_SWZ(swizzle, i); - if (swz[i] == SWIZZLE_NIL) { - swz[i] = swz[i - 1]; - } - } - return MAKE_SWIZZLE4(swz[0], swz[1], swz[2], swz[3]); -} - - /** * Return the number of components actually named by the swizzle. * Recall that swizzles may have undefined/don't-care values. @@ -1451,15 +1484,6 @@ emit_swizzle(slang_emit_info *emitInfo, slang_ir_node *n) inst = emit(emitInfo, n->Children[0]); -#ifdef DEBUG - { - GLuint s = n->Children[0]->Store->Swizzle; - assert(GET_SWZ(s, 0) != SWIZZLE_NIL); - assert(GET_SWZ(s, 1) != SWIZZLE_NIL); - assert(GET_SWZ(s, 2) != SWIZZLE_NIL); - assert(GET_SWZ(s, 3) != SWIZZLE_NIL); - } -#endif /* For debug: n->Var = n->Children[0]->Var; */ /* "pull-up" the child's storage info, applying our swizzle info */ @@ -1475,6 +1499,16 @@ emit_swizzle(slang_emit_info *emitInfo, slang_ir_node *n) /* apply this swizzle to child's swizzle to get composed swizzle */ swizzle = fix_swizzle(n->Store->Swizzle); /* remove the don't care terms */ + +#ifdef DEBUG + { + assert(GET_SWZ(swizzle, 0) != SWIZZLE_NIL); + assert(GET_SWZ(swizzle, 1) != SWIZZLE_NIL); + assert(GET_SWZ(swizzle, 2) != SWIZZLE_NIL); + assert(GET_SWZ(swizzle, 3) != SWIZZLE_NIL); + } +#endif + n->Store->Swizzle = _slang_swizzle_swizzle(n->Children[0]->Store->Swizzle, swizzle); @@ -1489,33 +1523,77 @@ emit_swizzle(slang_emit_info *emitInfo, slang_ir_node *n) static struct prog_instruction * emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n) { + slang_ir_storage *root; + + assert(n->Opcode == IR_ELEMENT); assert(n->Store); - assert(n->Store->File != PROGRAM_UNDEFINED); + assert(n->Store->File == PROGRAM_UNDEFINED); + assert(n->Store->Parent); assert(n->Store->Size > 0); - if (n->Store->File == PROGRAM_STATE_VAR) { - n->Store->Index = _slang_alloc_statevar(n, emitInfo->prog->Parameters); + root = n->Store; + while (root->Parent) + root = root->Parent; + + if (root->File == PROGRAM_STATE_VAR) { + GLint index = _slang_alloc_statevar(n, emitInfo->prog->Parameters); + assert(n->Store->Index == index); return NULL; } if (n->Children[1]->Opcode == IR_FLOAT) { - /* Constant index */ + /* Constant array index */ +#if 0 /* just debug code */ const GLint arrayAddr = n->Children[0]->Store->Index; const GLint index = (GLint) n->Children[1]->Value[0]; - n->Store->Index = arrayAddr + index; + assert(index == n->Store->Index); + assert(arrayAddr == parent->Index); + assert(n->Children[0]->Store == parent); + assert(n->Children[0]->Store->Index == parent->Index); +#endif + + GLint index = n->Store->Index; + + slang_ir_storage *p = n->Store; + index = 0; + while (p->Parent) { + int sz = (p->Size + 3) / 4; + /*printf("element [%d] of size %d (%d)\n", p->Index, p->Size, sz);*/ + index += sz * p->Index; + p = p->Parent; + } + index += p->Index; + + assert(root->File != PROGRAM_UNDEFINED); + + /* resolve new absolute storage location */ + assert(n->Store); + n->Store->File = root->File; + n->Store->Index = index; + n->Store->Parent = NULL; } else { - /* Variable index*/ + /* Variable array index */ struct prog_instruction *inst; + + /* do codegen for array */ + emit(emitInfo, n->Children[0]); + + /* do codegen for array index expression */ + emit(emitInfo, n->Children[1]); + inst = new_instruction(emitInfo, OPCODE_ARL); + storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); storage_to_src_reg(&inst->SrcReg[0], n->Children[1]->Store); + inst->DstReg.File = PROGRAM_ADDRESS; + inst->DstReg.Index = 0; /* always address register [0] */ inst->Comment = _mesa_strdup("ARL ADDR"); + n->Store->RelAddr = GL_TRUE; - n->Store->Index = inst->DstReg.Index;/*index of the array*/ - inst->DstReg.Index = 0; /*addr index is always 0*/ } + return NULL; /* no instruction */ } @@ -1526,25 +1604,27 @@ emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n) static struct prog_instruction * emit_struct_field(slang_emit_info *emitInfo, slang_ir_node *n) { - if (n->Store->File == PROGRAM_STATE_VAR) { - n->Store->Index = _slang_alloc_statevar(n, emitInfo->prog->Parameters); - if (n->Store->Index < 0) { + slang_ir_storage *root = n->Store; + + assert(n->Opcode == IR_FIELD); + + while (root->Parent) + root = root->Parent; + + /* If this is the field of a state var, allocate constant/uniform + * storage for it now if we haven't already. + * Note that we allocate storage (uniform/constant slots) for state + * variables here rather than at declaration time so we only allocate + * space for the ones that we actually use! + */ + if (root->File == PROGRAM_STATE_VAR) { + root->Index = _slang_alloc_statevar(n, emitInfo->prog->Parameters); + if (root->Index < 0) { slang_info_log_error(emitInfo->log, "Error parsing state variable"); return NULL; } } - else { - GLint offset = n->FieldOffset / 4; - assert(n->Children[0]->Store->Index >= 0); - n->Store->Index = n->Children[0]->Store->Index + offset; - if (n->Store->Size == 1) { - GLint swz = n->FieldOffset % 4; - n->Store->Swizzle = MAKE_SWIZZLE4(swz, swz, swz, swz); - } - else { - n->Store->Swizzle = SWIZZLE_XYZW; - } - } + return NULL; /* no instruction */ } diff --git a/src/mesa/shader/slang/slang_emit.h b/src/mesa/shader/slang/slang_emit.h index 7d70b3311b..153e872d74 100644 --- a/src/mesa/shader/slang/slang_emit.h +++ b/src/mesa/shader/slang/slang_emit.h @@ -40,10 +40,6 @@ extern GLuint _slang_swizzle_swizzle(GLuint swz1, GLuint swz2); -extern slang_ir_storage * -_slang_new_ir_storage(enum register_file file, GLint index, GLint size); - - extern GLboolean _slang_emit_code(slang_ir_node *n, slang_var_table *vartable, struct gl_program *prog, GLboolean withEnd, diff --git a/src/mesa/shader/slang/slang_ir.c b/src/mesa/shader/slang/slang_ir.c index 16c36aad0c..ce4a198025 100644 --- a/src/mesa/shader/slang/slang_ir.c +++ b/src/mesa/shader/slang/slang_ir.c @@ -51,6 +51,8 @@ static const slang_ir_info IrInfo[] = { { IR_SLT, "IR_SLT", OPCODE_SLT, 4, 2 }, { IR_POW, "IR_POW", OPCODE_POW, 1, 2 }, { IR_EQUAL, "IR_EQUAL", OPCODE_NOP, 1, 2 }, + { IR_NOTEQUAL, "IR_NOTEQUAL", OPCODE_NOP, 1, 2 }, + /* unary ops */ { IR_I_TO_F, "IR_I_TO_F", OPCODE_NOP, 1, 1 }, { IR_F_TO_I, "IR_F_TO_I", OPCODE_INT, 4, 1 }, /* 4 floats to 4 ints */ @@ -109,6 +111,66 @@ _slang_ir_info(slang_ir_opcode opcode) } +/** + * Return a new slang_ir_storage object. + */ +slang_ir_storage * +_slang_new_ir_storage(enum register_file file, GLint index, GLint size) +{ + slang_ir_storage *st; + st = (slang_ir_storage *) _slang_alloc(sizeof(slang_ir_storage)); + if (st) { + st->File = file; + st->Index = index; + st->Size = size; + st->Swizzle = SWIZZLE_NOOP; + st->Parent = NULL; + } + return st; +} + + +/** + * Return a new slang_ir_storage object. + */ +slang_ir_storage * +_slang_new_ir_storage_swz(enum register_file file, GLint index, GLint size, + GLuint swizzle) +{ + slang_ir_storage *st; + st = (slang_ir_storage *) _slang_alloc(sizeof(slang_ir_storage)); + if (st) { + st->File = file; + st->Index = index; + st->Size = size; + st->Swizzle = swizzle; + st->Parent = NULL; + } + return st; +} + + +/** + * Return a new slang_ir_storage object. + */ +slang_ir_storage * +_slang_new_ir_storage_relative(GLint index, GLint size, + slang_ir_storage *parent) +{ + slang_ir_storage *st; + st = (slang_ir_storage *) _slang_alloc(sizeof(slang_ir_storage)); + if (st) { + st->File = PROGRAM_UNDEFINED; + st->Index = index; + st->Size = size; + st->Swizzle = SWIZZLE_NOOP; + st->Parent = parent; + } + return st; +} + + + static const char * _slang_ir_name(slang_ir_opcode opcode) { @@ -116,6 +178,7 @@ _slang_ir_name(slang_ir_opcode opcode) } + #if 0 /* no longer needed with mempool */ /** * Since many IR nodes might point to the same IR storage info, we need diff --git a/src/mesa/shader/slang/slang_ir.h b/src/mesa/shader/slang/slang_ir.h index ba0735d64d..61ff649e5c 100644 --- a/src/mesa/shader/slang/slang_ir.h +++ b/src/mesa/shader/slang/slang_ir.h @@ -147,6 +147,11 @@ struct _slang_ir_storage GLuint Swizzle; GLint RefCount; /**< Used during IR tree delete */ GLboolean RelAddr; + + /** If Parent is non-null, Index is relative to parent. + * The other fields are ignored. + */ + struct _slang_ir_storage *Parent; }; typedef struct _slang_ir_storage slang_ir_storage; @@ -165,7 +170,6 @@ typedef struct slang_ir_node_ /** special fields depending on Opcode: */ const char *Field; /**< If Opcode == IR_FIELD */ - int FieldOffset; /**< If Opcode == IR_FIELD */ GLuint Writemask; /**< If Opcode == IR_MOVE */ GLfloat Value[4]; /**< If Opcode == IR_FLOAT */ slang_variable *Var; /**< If Opcode == IR_VAR or IR_VAR_DECL */ @@ -193,6 +197,20 @@ extern const slang_ir_info * _slang_ir_info(slang_ir_opcode opcode); +extern slang_ir_storage * +_slang_new_ir_storage(enum register_file file, GLint index, GLint size); + + +extern slang_ir_storage * +_slang_new_ir_storage_swz(enum register_file file, GLint index, GLint size, + GLuint swizzle); + +extern slang_ir_storage * +_slang_new_ir_storage_relative(GLint index, GLint size, + slang_ir_storage *parent); + + + extern void _slang_free_ir_tree(slang_ir_node *n); diff --git a/src/mesa/shader/slang/slang_vartable.c b/src/mesa/shader/slang/slang_vartable.c index 7bc92ea297..68b4e00be0 100644 --- a/src/mesa/shader/slang/slang_vartable.c +++ b/src/mesa/shader/slang/slang_vartable.c @@ -1,6 +1,7 @@ #include "main/imports.h" -#include "shader/prog_instruction.h" +#include "shader/program.h" +#include "shader/prog_print.h" #include "slang_compile.h" #include "slang_compile_variable.h" #include "slang_mem.h" @@ -247,14 +248,25 @@ _slang_alloc_var(slang_var_table *vt, slang_ir_storage *store) if (store->Size == 1) { const GLuint comp = i % 4; store->Swizzle = MAKE_SWIZZLE4(comp, comp, comp, comp); - if (dbg) printf("Alloc var sz %d at %d.%c (level %d)\n", - store->Size, store->Index, "xyzw"[comp], t->Level); + } + else if (store->Size == 2) { + store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, + SWIZZLE_NIL, SWIZZLE_NIL); + } + else if (store->Size == 3) { + store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, + SWIZZLE_Z, SWIZZLE_NIL); } else { store->Swizzle = SWIZZLE_NOOP; - if (dbg) printf("Alloc var sz %d at %d.xyzw (level %d)\n", - store->Size, store->Index, t->Level); } + + if (dbg) + printf("Alloc var sz %d at %d.%s (level %d)\n", + store->Size, store->Index, + _mesa_swizzle_string(store->Swizzle, 0, 0), + t->Level); + return GL_TRUE; } @@ -279,6 +291,7 @@ _slang_alloc_temp(slang_var_table *vt, slang_ir_storage *store) store->Size, store->Index, "xyzw"[comp], t->Level); } else { + /* XXX improve swizzled for size=2/3, use for writemask... */ store->Swizzle = SWIZZLE_NOOP; if (dbg) printf("Alloc temp sz %d at %d.xyzw (level %d)\n", store->Size, store->Index, t->Level); -- cgit v1.2.3 From 72e57b52f02db17d47dce23abedea72a47af806e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 18 Jul 2008 19:46:19 -0600 Subject: mesa: glsl: various writemask/swizzle improvements and clean-ups --- src/mesa/shader/prog_parameter.c | 2 +- src/mesa/shader/prog_print.c | 2 +- src/mesa/shader/slang/slang_codegen.c | 62 ++++++++++++++++++++++---------- src/mesa/shader/slang/slang_emit.c | 67 +++++++++++++++++------------------ src/mesa/shader/slang/slang_ir.c | 18 ++-------- 5 files changed, 81 insertions(+), 70 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index b302442227..be696385aa 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -348,7 +348,7 @@ _mesa_add_varying(struct gl_program_parameter_list *paramList, return i; } else { - assert(size == 4); + /*assert(size == 4);*/ i = _mesa_add_parameter(paramList, PROGRAM_VARYING, name, size, GL_NONE, NULL, NULL); return i; diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index 014d08f062..5368ab2187 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -313,7 +313,7 @@ reg_string(enum register_file f, GLint index, gl_prog_print_mode mode, const char * _mesa_swizzle_string(GLuint swizzle, GLuint negateBase, GLboolean extended) { - static const char swz[] = "xyzw01?!"; + static const char swz[] = "xyzw01!?"; /* See SWIZZLE_x definitions */ static char s[20]; GLuint i = 0; diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 6261aab23c..d4d2e34dee 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1409,6 +1409,27 @@ slang_find_asm_info(const char *name) } +/** + * Return the default swizzle mask for accessing a variable of the + * given size (in floats). If size = 1, comp is used to identify + * which component [0..3] of the register holds the variable. + */ +static GLuint +_slang_var_swizzle(GLint size, GLint comp) +{ + switch (size) { + case 1: + return MAKE_SWIZZLE4(comp, comp, comp, comp); + case 2: + return MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_NIL, SWIZZLE_NIL); + case 3: + return MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_NIL); + default: + return SWIZZLE_XYZW; + } +} + + /** * Some write-masked assignments are simple, but others are hard. * Simple example: @@ -3063,6 +3084,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, int dbg = 0; const GLenum datatype = _slang_gltype_from_specifier(&var->type.specifier); const GLint texIndex = sampler_to_texture_index(var->type.specifier.type); + const GLint size = _slang_sizeof_type_specifier(&var->type.specifier); if (texIndex != -1) { /* This is a texture sampler variable... @@ -3076,8 +3098,8 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, } else if (var->type.qualifier == SLANG_QUAL_UNIFORM) { /* Uniform variable */ - const GLint size = _slang_sizeof_type_specifier(&var->type.specifier) - * MAX2(var->array_len, 1); + const GLint totalSize = size * MAX2(var->array_len, 1); + const GLuint swizzle = _slang_var_swizzle(totalSize, 0); if (prog) { /* user-defined uniform */ if (datatype == GL_NONE) { @@ -3106,8 +3128,10 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, } else { GLint uniformLoc = _mesa_add_uniform(prog->Parameters, varName, - size, datatype); - store = _slang_new_ir_storage(PROGRAM_UNIFORM, uniformLoc, size); + totalSize, datatype); + store = _slang_new_ir_storage_swz(PROGRAM_UNIFORM, uniformLoc, + totalSize, swizzle); + printf("GLOBAL USER UNIFORM %s size %d\n", varName, totalSize); } } else { @@ -3115,33 +3139,40 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, /* We know it's a uniform, but don't allocate storage unless * it's really used. */ - store = _slang_new_ir_storage(PROGRAM_STATE_VAR, -1, size); + store = _slang_new_ir_storage_swz(PROGRAM_STATE_VAR, -1, + totalSize, swizzle); } - if (dbg) printf("UNIFORM (sz %d) ", size); + if (dbg) printf("UNIFORM (sz %d) ", totalSize); } else if (var->type.qualifier == SLANG_QUAL_VARYING) { - const GLint size = 4; /* XXX fix */ if (prog) { /* user-defined varying */ GLint varyingLoc = _mesa_add_varying(prog->Varying, varName, size); - store = _slang_new_ir_storage(PROGRAM_VARYING, varyingLoc, size); + GLuint swizzle = _slang_var_swizzle(size, 0); + store = _slang_new_ir_storage_swz(PROGRAM_VARYING, varyingLoc, + size, swizzle); } else { /* pre-defined varying, like gl_Color or gl_TexCoord */ if (type == SLANG_UNIT_FRAGMENT_BUILTIN) { + /* fragment program input */ GLuint swizzle; GLint index = _slang_input_index(varName, GL_FRAGMENT_PROGRAM_ARB, &swizzle); assert(index >= 0); - store = _slang_new_ir_storage_swz(PROGRAM_INPUT, index, size, swizzle); assert(index < FRAG_ATTRIB_MAX); + store = _slang_new_ir_storage_swz(PROGRAM_INPUT, index, + size, swizzle); } else { + /* vertex program output */ GLint index = _slang_output_index(varName, GL_VERTEX_PROGRAM_ARB); + GLuint swizzle = _slang_var_swizzle(size, 0); assert(index >= 0); - assert(type == SLANG_UNIT_VERTEX_BUILTIN); - store = _slang_new_ir_storage(PROGRAM_OUTPUT, index, size); assert(index < VERT_RESULT_MAX); + assert(type == SLANG_UNIT_VERTEX_BUILTIN); + store = _slang_new_ir_storage_swz(PROGRAM_OUTPUT, index, + size, swizzle); } if (dbg) printf("V/F "); } @@ -3150,7 +3181,6 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, else if (var->type.qualifier == SLANG_QUAL_ATTRIBUTE) { if (prog) { /* user-defined vertex attribute */ - const GLint size = _slang_sizeof_type_specifier(&var->type.specifier); const GLint attr = -1; /* unknown */ GLint index = _mesa_add_attribute(prog->Attributes, varName, size, attr); @@ -3163,7 +3193,6 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, GLuint swizzle; GLint index = _slang_input_index(varName, GL_VERTEX_PROGRAM_ARB, &swizzle); - GLint size = 4; /* XXX? */ assert(index >= 0); store = _slang_new_ir_storage_swz(PROGRAM_INPUT, index, size, swizzle); } @@ -3173,27 +3202,24 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, GLuint swizzle = SWIZZLE_XYZW; /* silence compiler warning */ GLint index = _slang_input_index(varName, GL_FRAGMENT_PROGRAM_ARB, &swizzle); - GLint size = 4; /* XXX? */ store = _slang_new_ir_storage_swz(PROGRAM_INPUT, index, size, swizzle); if (dbg) printf("INPUT "); } else if (var->type.qualifier == SLANG_QUAL_FIXEDOUTPUT) { if (type == SLANG_UNIT_VERTEX_BUILTIN) { GLint index = _slang_output_index(varName, GL_VERTEX_PROGRAM_ARB); - GLint size = 4; /* XXX? */ store = _slang_new_ir_storage(PROGRAM_OUTPUT, index, size); } else { GLint index = _slang_output_index(varName, GL_FRAGMENT_PROGRAM_ARB); - GLint size = 4; /* XXX? */ + GLint specialSize = 4; /* treat all fragment outputs as float[4] */ assert(type == SLANG_UNIT_FRAGMENT_BUILTIN); - store = _slang_new_ir_storage(PROGRAM_OUTPUT, index, size); + store = _slang_new_ir_storage(PROGRAM_OUTPUT, index, specialSize); } if (dbg) printf("OUTPUT "); } else if (var->type.qualifier == SLANG_QUAL_CONST && !prog) { /* pre-defined global constant, like gl_MaxLights */ - const GLint size = _slang_sizeof_type_specifier(&var->type.specifier); store = _slang_new_ir_storage(PROGRAM_CONSTANT, -1, size); if (dbg) printf("CONST "); } diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index ec8ed0b4e8..291fd7c6dd 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -171,21 +171,33 @@ free_temp_storage(slang_var_table *vt, slang_ir_node *n) /** - * Remove any SWIZZLE_NIL terms from given swizzle mask (smear prev term). - * Ex: fix_swizzle("zyNN") -> "zyyy" - * XXX should put in the default component for the position... + * Remove any SWIZZLE_NIL terms from given swizzle mask. + * For a swizzle like .z??? generate .zzzz (replicate single component). + * Else, for .wx?? generate .wxzw (insert default component for the position). */ static GLuint fix_swizzle(GLuint swizzle) { - GLuint swz[4], i; - for (i = 0; i < 4; i++) { - swz[i] = GET_SWZ(swizzle, i); - if (swz[i] == SWIZZLE_NIL) { - swz[i] = swz[i - 1]; - } + GLuint c0 = GET_SWZ(swizzle, 0), + c1 = GET_SWZ(swizzle, 1), + c2 = GET_SWZ(swizzle, 2), + c3 = GET_SWZ(swizzle, 3); + if (c1 == SWIZZLE_NIL && c2 == SWIZZLE_NIL && c3 == SWIZZLE_NIL) { + /* smear first component across all positions */ + c1 = c2 = c3 = c0; } - return MAKE_SWIZZLE4(swz[0], swz[1], swz[2], swz[3]); + else { + /* insert default swizzle components */ + if (c0 == SWIZZLE_NIL) + c0 = SWIZZLE_X; + if (c1 == SWIZZLE_NIL) + c1 = SWIZZLE_Y; + if (c2 == SWIZZLE_NIL) + c2 = SWIZZLE_Z; + if (c3 == SWIZZLE_NIL) + c3 = SWIZZLE_W; + } + return MAKE_SWIZZLE4(c0, c1, c2, c3); } @@ -216,6 +228,7 @@ storage_to_dst_reg(struct prog_dst_register *dst, const slang_ir_storage *st, assert(size >= 1); assert(size <= 4); + if (size == 1) { GLuint comp = GET_SWZ(swizzle, 0); assert(comp < 4); @@ -233,13 +246,6 @@ storage_to_dst_reg(struct prog_dst_register *dst, const slang_ir_storage *st, static void storage_to_src_reg(struct prog_src_register *src, const slang_ir_storage *st) { - static const GLuint defaultSwizzle[4] = { - MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X), - MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W), - MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W), - MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_W) - }; - const GLint size = st->Size; const GLboolean relAddr = st->RelAddr; GLint index = st->Index; GLuint swizzle = st->Swizzle; @@ -250,7 +256,6 @@ storage_to_src_reg(struct prog_src_register *src, const slang_ir_storage *st) index += st->Index; swizzle = _slang_swizzle_swizzle(st->Swizzle, swizzle); } - swizzle = fix_swizzle(swizzle); assert(st->File >= 0); assert(st->File < PROGRAM_UNDEFINED); @@ -259,20 +264,14 @@ storage_to_src_reg(struct prog_src_register *src, const slang_ir_storage *st) assert(index >= 0); src->Index = index; - assert(size >= 1); - assert(size <= 4); + swizzle = fix_swizzle(swizzle); + assert(GET_SWZ(swizzle, 0) <= SWIZZLE_W); + assert(GET_SWZ(swizzle, 1) <= SWIZZLE_W); + assert(GET_SWZ(swizzle, 2) <= SWIZZLE_W); + assert(GET_SWZ(swizzle, 3) <= SWIZZLE_W); + src->Swizzle = swizzle; src->RelAddr = relAddr; - - if (swizzle != SWIZZLE_NOOP) - src->Swizzle = swizzle; - else - src->Swizzle = defaultSwizzle[size - 1]; /*XXX really need this?*/ - - assert(GET_SWZ(src->Swizzle, 0) <= 3); - assert(GET_SWZ(src->Swizzle, 1) <= 3); - assert(GET_SWZ(src->Swizzle, 2) <= 3); - assert(GET_SWZ(src->Swizzle, 3) <= 3); } @@ -582,8 +581,7 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) /* result storage */ if (!n->Store) { - GLint size = n->Children[0]->Store - ? n->Children[0]->Store->Size : info->ResultSize; + GLint size = info->ResultSize; if (!alloc_temp_storage(emitInfo, n, size)) return NULL; #if 0000 /* this should work, but doesn't yet */ @@ -591,6 +589,8 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) n->Writemask = WRITEMASK_XY; else if (size == 3) n->Writemask = WRITEMASK_XYZ; + else if (size == 1) + n->Writemask = WRITEMASK_X << GET_SWZ(n->Store->Swizzle,0); #endif } @@ -1045,10 +1045,9 @@ emit_move(slang_emit_info *emitInfo, slang_ir_node *n) assert(n->Children[0]->Store->Index >= 0); /* use tighter writemask when possible */ -#if 0000 /* XXX enable this after more testing... */ if (n->Writemask == WRITEMASK_XYZW) n->Writemask = inst->DstReg.WriteMask; -#endif + storage_to_dst_reg(&inst->DstReg, n->Children[0]->Store, n->Writemask); return inst; } diff --git a/src/mesa/shader/slang/slang_ir.c b/src/mesa/shader/slang/slang_ir.c index ce4a198025..a414036a36 100644 --- a/src/mesa/shader/slang/slang_ir.c +++ b/src/mesa/shader/slang/slang_ir.c @@ -239,20 +239,6 @@ _slang_free_ir_tree(slang_ir_node *n) -static const char * -swizzle_string(GLuint swizzle) -{ - static char s[6]; - GLuint i; - s[0] = '.'; - for (i = 1; i < 5; i++) { - s[i] = "xyzw"[GET_SWZ(swizzle, i-1)]; - } - s[i] = 0; - return s; -} - - static const char * writemask_string(GLuint writemask) { @@ -410,7 +396,7 @@ _slang_print_ir_tree(const slang_ir_node *n, int indent) case IR_VAR: printf("VAR %s%s at %s store %p\n", (n->Var ? (char *) n->Var->a_name : "TEMP"), - swizzle_string(n->Store->Swizzle), + _mesa_swizzle_string(n->Store->Swizzle, 0, 0), storage_string(n->Store), (void*) n->Store); break; case IR_VAR_DECL: @@ -437,7 +423,7 @@ _slang_print_ir_tree(const slang_ir_node *n, int indent) break; case IR_SWIZZLE: printf("SWIZZLE %s of (store %p) \n", - swizzle_string(n->Store->Swizzle), (void*) n->Store); + _mesa_swizzle_string(n->Store->Swizzle, 0, 0), (void*) n->Store); _slang_print_ir_tree(n->Children[0], indent + 3); break; default: -- cgit v1.2.3 From 1a8ceb3828c64fb758955c85a1d3d06f53f15ed5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 19 Jul 2008 08:38:35 +0900 Subject: gallium: Move PIPE_TEXTURE_USAGE* to p_defines.h --- src/gallium/include/pipe/p_defines.h | 6 ++++++ src/gallium/include/pipe/p_state.h | 6 ------ 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index f8fadf31b6..bc4d7c845a 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -166,6 +166,12 @@ enum pipe_texture_target { #define PIPE_TEX_FACE_NEG_Z 5 #define PIPE_TEX_FACE_MAX 6 +#define PIPE_TEXTURE_USAGE_RENDER_TARGET 0x1 +#define PIPE_TEXTURE_USAGE_DISPLAY_TARGET 0x2 /* ie a backbuffer */ +#define PIPE_TEXTURE_USAGE_PRIMARY 0x4 /* ie a frontbuffer */ +#define PIPE_TEXTURE_USAGE_DEPTH_STENCIL 0x8 +#define PIPE_TEXTURE_USAGE_SAMPLER 0x10 + /** * Surfaces, textures, etc. (others may be added) */ diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 5546796936..2401305eb7 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -287,12 +287,6 @@ struct pipe_surface }; -#define PIPE_TEXTURE_USAGE_RENDER_TARGET 0x1 -#define PIPE_TEXTURE_USAGE_DISPLAY_TARGET 0x2 /* ie a backbuffer */ -#define PIPE_TEXTURE_USAGE_PRIMARY 0x4 /* ie a frontbuffer */ -#define PIPE_TEXTURE_USAGE_DEPTH_STENCIL 0x8 -#define PIPE_TEXTURE_USAGE_SAMPLER 0x10 - /** * Texture object. */ -- cgit v1.2.3 From ff26c50153b3a348b35843262ceb27062ab37214 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 19 Jul 2008 11:52:41 +0900 Subject: tgsi: Make tgsi_sanity return TRUE on success as documented. --- src/gallium/auxiliary/tgsi/util/tgsi_sanity.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c index 933127e661..7fc4c29c68 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c @@ -266,5 +266,5 @@ tgsi_sanity_check( if (tgsi_iterate_shader( tokens, &ctx.iter ) == -1) return FALSE; - return ctx.errors > 0; + return ctx.errors == 0; } -- cgit v1.2.3 From 8aafc03b260ab8923f1b373f7effa75bcdb40a72 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 19 Jul 2008 12:04:37 +0900 Subject: gallium: Finer grained is_format_supported. --- src/gallium/auxiliary/util/u_blit.c | 12 ++- src/gallium/auxiliary/util/u_gen_mipmap.c | 3 +- src/gallium/drivers/cell/ppu/cell_screen.c | 24 ++--- src/gallium/drivers/i915simple/i915_screen.c | 18 ++-- src/gallium/drivers/i965simple/brw_screen.c | 5 +- src/gallium/drivers/softpipe/sp_screen.c | 26 +++--- src/gallium/include/pipe/p_defines.h | 7 +- src/gallium/include/pipe/p_screen.h | 9 +- src/gallium/state_trackers/python/gallium.i | 15 ++- src/gallium/state_trackers/python/samples/tri.py | 8 +- src/gallium/state_trackers/python/tests/texture.py | 13 ++- src/mesa/state_tracker/st_atom_pixeltransfer.c | 2 +- src/mesa/state_tracker/st_cb_bitmap.c | 3 +- src/mesa/state_tracker/st_cb_drawpixels.c | 9 +- src/mesa/state_tracker/st_cb_texture.c | 8 +- src/mesa/state_tracker/st_extensions.c | 14 ++- src/mesa/state_tracker/st_format.c | 101 +++++++++++---------- src/mesa/state_tracker/st_format.h | 2 +- src/mesa/state_tracker/st_gen_mipmap.c | 3 +- src/mesa/state_tracker/st_texture.c | 3 +- 20 files changed, 158 insertions(+), 127 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index 3dc9fdd11e..ae087df4cf 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -307,8 +307,10 @@ util_blit_pixels(struct blit_state *ctx, dstY1 = tmp; } - assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE)); - assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE)); + assert(screen->is_format_supported(screen, src->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)); + assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)); if(dst->format == src->format && (dstX1 - dstX0) == srcW && (dstY1 - dstY0) == srcH) { /* FIXME: this will most surely fail for overlapping rectangles */ @@ -319,7 +321,8 @@ util_blit_pixels(struct blit_state *ctx, return; } - assert(screen->is_format_supported(screen, dst->format, PIPE_SURFACE)); + assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)); /* * XXX for now we're always creating a temporary texture. @@ -449,7 +452,8 @@ util_blit_pixels_tex(struct blit_state *ctx, t0 = srcY0 / (float)tex->height[0]; t1 = srcY1 / (float)tex->height[0]; - assert(screen->is_format_supported(screen, dst->format, PIPE_SURFACE)); + assert(screen->is_format_supported(screen, dst->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)); /* save state (restored below) */ cso_save_blend(ctx->cso); diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 5313a8008a..4999822068 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -858,7 +858,8 @@ util_gen_mipmap(struct gen_mipmap_state *ctx, uint zslice = 0; /* check if we can render in the texture's format */ - if (!screen->is_format_supported(screen, pt->format, PIPE_SURFACE)) { + if (!screen->is_format_supported(screen, pt->format, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) { fallback_gen_mipmap(ctx, pt, face, baseLevel, lastLevel); return; } diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c index 5198b51441..cf9b68b695 100644 --- a/src/gallium/drivers/cell/ppu/cell_screen.c +++ b/src/gallium/drivers/cell/ppu/cell_screen.c @@ -115,23 +115,17 @@ cell_get_paramf(struct pipe_screen *screen, int param) static boolean cell_is_format_supported( struct pipe_screen *screen, - enum pipe_format format, uint type ) + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags ) { - switch (type) { - case PIPE_TEXTURE: - /* cell supports most texture formats, XXX for now anyway */ - if (format == PIPE_FORMAT_DXT5_RGBA || - format == PIPE_FORMAT_R8G8B8A8_SRGB) - return FALSE; - else - return TRUE; - case PIPE_SURFACE: - /* cell supports all (off-screen) surface formats, XXX for now */ - return TRUE; - default: - assert(0); + /* cell supports most formats, XXX for now anyway */ + if (format == PIPE_FORMAT_DXT5_RGBA || + format == PIPE_FORMAT_R8G8B8A8_SRGB) return FALSE; - } + else + return TRUE; } diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index ba8f183bdf..4b1b8af7da 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -148,7 +148,10 @@ i915_get_paramf(struct pipe_screen *screen, int param) static boolean i915_is_format_supported( struct pipe_screen *screen, - enum pipe_format format, uint type ) + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags ) { static const enum pipe_format tex_supported[] = { PIPE_FORMAT_R8G8B8A8_UNORM, @@ -173,17 +176,10 @@ i915_is_format_supported( struct pipe_screen *screen, const enum pipe_format *list; uint i; - switch (type) { - case PIPE_TEXTURE: - list = tex_supported; - break; - case PIPE_SURFACE: + if(tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) list = surface_supported; - break; - default: - assert(0); - return FALSE; - } + else + list = tex_supported; for (i = 0; list[i] != PIPE_FORMAT_NONE; i++) { if (list[i] == format) diff --git a/src/gallium/drivers/i965simple/brw_screen.c b/src/gallium/drivers/i965simple/brw_screen.c index b700f7e4f5..6d8f24d1c4 100644 --- a/src/gallium/drivers/i965simple/brw_screen.c +++ b/src/gallium/drivers/i965simple/brw_screen.c @@ -136,7 +136,10 @@ brw_get_paramf(struct pipe_screen *screen, int param) static boolean brw_is_format_supported( struct pipe_screen *screen, - enum pipe_format format, uint type ) + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags ) { #if 0 /* XXX: This is broken -- rewrite if still needed. */ diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index f6193bfaf9..3f9d4b0ed3 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -115,23 +115,19 @@ softpipe_get_paramf(struct pipe_screen *screen, int param) */ static boolean softpipe_is_format_supported( struct pipe_screen *screen, - enum pipe_format format, uint type ) + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags ) { - switch (type) { - case PIPE_TEXTURE: - case PIPE_SURFACE: - switch(format) { - case PIPE_FORMAT_DXT1_RGB: - case PIPE_FORMAT_DXT1_RGBA: - case PIPE_FORMAT_DXT3_RGBA: - case PIPE_FORMAT_DXT5_RGBA: - return FALSE; - default: - return TRUE; - } - default: - assert(0); + switch(format) { + case PIPE_FORMAT_DXT1_RGB: + case PIPE_FORMAT_DXT1_RGBA: + case PIPE_FORMAT_DXT3_RGBA: + case PIPE_FORMAT_DXT5_RGBA: return FALSE; + default: + return TRUE; } } diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index bc4d7c845a..b1d100ef53 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -172,11 +172,8 @@ enum pipe_texture_target { #define PIPE_TEXTURE_USAGE_DEPTH_STENCIL 0x8 #define PIPE_TEXTURE_USAGE_SAMPLER 0x10 -/** - * Surfaces, textures, etc. (others may be added) - */ -#define PIPE_TEXTURE 1 -#define PIPE_SURFACE 2 /**< user-created surfaces */ +#define PIPE_TEXTURE_GEOM_NON_SQUARE 0x1 +#define PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO 0x2 /** diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h index cc8430dae1..b15affef7a 100644 --- a/src/gallium/include/pipe/p_screen.h +++ b/src/gallium/include/pipe/p_screen.h @@ -77,11 +77,14 @@ struct pipe_screen { /** * Check if the given pipe_format is supported as a texture or * drawing surface. - * \param type one of PIPE_TEXTURE, PIPE_SURFACE + * \param tex_usage bitmask of PIPE_TEXTURE_USAGE_* + * \param flags bitmask of PIPE_TEXTURE_GEOM_* */ boolean (*is_format_supported)( struct pipe_screen *, - enum pipe_format format, - uint type ); + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags ); /** * Create a new texture object, using the given template info. diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index c08ac87aca..8d8b762ea5 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -158,8 +158,15 @@ struct st_context { * drawing surface. * \param type one of PIPE_TEXTURE, PIPE_SURFACE */ - int is_format_supported( enum pipe_format format, unsigned type ) { - return $self->screen->is_format_supported( $self->screen, format, type); + int is_format_supported( enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags ) { + return $self->screen->is_format_supported( $self->screen, + format, + target, + tex_usage, + geom_flags ); } struct st_context * @@ -175,7 +182,7 @@ struct st_context { unsigned depth = 1, unsigned last_level = 0, enum pipe_texture_target target = PIPE_TEXTURE_2D, - unsigned usage = 0 + unsigned tex_usage = 0 ) { struct pipe_texture templat; memset(&templat, 0, sizeof(templat)); @@ -186,7 +193,7 @@ struct st_context { templat.depth[0] = depth; templat.last_level = last_level; templat.target = target; - templat.tex_usage = usage; + templat.tex_usage = tex_usage; return $self->screen->texture_create($self->screen, &templat); } diff --git a/src/gallium/state_trackers/python/samples/tri.py b/src/gallium/state_trackers/python/samples/tri.py index 3665922929..1271c67627 100644 --- a/src/gallium/state_trackers/python/samples/tri.py +++ b/src/gallium/state_trackers/python/samples/tri.py @@ -140,9 +140,11 @@ def test(dev): ctx.set_clip(clip) # framebuffer - cbuf = dev.texture_create(PIPE_FORMAT_X8R8G8B8_UNORM, - width, height, - usage=PIPE_TEXTURE_USAGE_DISPLAY_TARGET) + cbuf = dev.texture_create( + PIPE_FORMAT_X8R8G8B8_UNORM, + width, height, + tex_usage=PIPE_TEXTURE_USAGE_DISPLAY_TARGET, + ) _cbuf = cbuf.get_surface(usage = PIPE_BUFFER_USAGE_GPU_READ|PIPE_BUFFER_USAGE_GPU_WRITE) fb = Framebuffer() fb.width = width diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index 16ad78c8aa..b2ca9f416f 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -136,7 +136,7 @@ class TextureTest(TestCase): level = self.level zslice = self.zslice - if not dev.is_format_supported(format, PIPE_TEXTURE): + if not dev.is_format_supported(format, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_SAMPLER, 0): raise TestSkip ctx = self.dev.context_create() @@ -199,6 +199,7 @@ class TextureTest(TestCase): height = height, depth = depth, last_level = last_level, + tex_usage = PIPE_TEXTURE_USAGE_SAMPLER, ) expected_rgba = FloatArray(height*width*4) @@ -212,10 +213,12 @@ class TextureTest(TestCase): ctx.set_sampler_texture(0, texture) # framebuffer - cbuf_tex = dev.texture_create(PIPE_FORMAT_A8R8G8B8_UNORM, - width, - height, - usage = PIPE_TEXTURE_USAGE_RENDER_TARGET) + cbuf_tex = dev.texture_create( + PIPE_FORMAT_A8R8G8B8_UNORM, + width, + height, + tex_usage = PIPE_TEXTURE_USAGE_RENDER_TARGET, + ) cbuf = cbuf_tex.get_surface(usage = PIPE_BUFFER_USAGE_GPU_WRITE|PIPE_BUFFER_USAGE_GPU_READ) fb = Framebuffer() diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c index e4de875e8c..a357b71677 100644 --- a/src/mesa/state_tracker/st_atom_pixeltransfer.c +++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c @@ -122,7 +122,7 @@ create_color_map_texture(GLcontext *ctx) const uint texSize = 256; /* simple, and usually perfect */ /* find an RGBA texture format */ - format = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE); + format = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_SAMPLER); /* create texture for color map/table */ pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index de86832342..d5696a909f 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -746,7 +746,8 @@ st_init_bitmap(struct st_context *st) st->bitmap.rasterizer.bypass_vs = 1; /* find a usable texture format */ - if (screen->is_format_supported(screen, PIPE_FORMAT_I8_UNORM, PIPE_TEXTURE)) { + if (screen->is_format_supported(screen, PIPE_FORMAT_I8_UNORM, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { st->bitmap.tex_format = PIPE_FORMAT_I8_UNORM; } else { diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 2ebfcaf82b..db0c9fbd09 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -995,18 +995,21 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy, srcFormat = rbRead->texture->format; - if (screen->is_format_supported(screen, srcFormat, PIPE_TEXTURE)) { + if (screen->is_format_supported(screen, srcFormat, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { texFormat = srcFormat; } else { /* srcFormat can't be used as a texture format */ if (type == GL_DEPTH) { - texFormat = st_choose_format(pipe, GL_DEPTH_COMPONENT, PIPE_TEXTURE); + texFormat = st_choose_format(pipe, GL_DEPTH_COMPONENT, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL); assert(texFormat != PIPE_FORMAT_NONE); /* XXX no depth texture formats??? */ } else { /* default color format */ - texFormat = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE); + texFormat = st_choose_format(pipe, GL_RGBA, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER); assert(texFormat != PIPE_FORMAT_NONE); } } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index de782e8232..1f94a0b9ef 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1210,9 +1210,13 @@ do_copy_texsubimage(GLcontext *ctx, use_fallback = GL_FALSE; } else if (screen->is_format_supported(screen, strb->surface->format, - PIPE_TEXTURE) && + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, + 0) && screen->is_format_supported(screen, dest_surface->format, - PIPE_SURFACE)) { + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET, + 0)) { boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP); int srcY0, srcY1; if (do_flip) { diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index d804d2b453..cacf972a1b 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -213,18 +213,24 @@ void st_init_extensions(struct st_context *st) } if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_SRGB, - PIPE_TEXTURE)) { + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { ctx->Extensions.EXT_texture_sRGB = GL_TRUE; } #if 01 if (screen->is_format_supported(screen, PIPE_FORMAT_DXT5_RGBA, - PIPE_TEXTURE)) { + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE; } #endif - if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR, PIPE_TEXTURE) || - screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV, PIPE_TEXTURE)) { + if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0) || + screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { ctx->Extensions.MESA_ycbcr_texture = GL_TRUE; } diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 50a06868df..b6d97ef659 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -281,7 +281,10 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat) * Find an RGBA format supported by the context/winsys. */ static enum pipe_format -default_rgba_format(struct pipe_screen *screen, uint type) +default_rgba_format(struct pipe_screen *screen, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags) { static const enum pipe_format colorFormats[] = { PIPE_FORMAT_A8R8G8B8_UNORM, @@ -291,7 +294,7 @@ default_rgba_format(struct pipe_screen *screen, uint type) }; uint i; for (i = 0; i < Elements(colorFormats); i++) { - if (screen->is_format_supported( screen, colorFormats[i], type )) { + if (screen->is_format_supported( screen, colorFormats[i], target, tex_usage, geom_flags )) { return colorFormats[i]; } } @@ -303,13 +306,16 @@ default_rgba_format(struct pipe_screen *screen, uint type) * Search list of formats for first RGBA format with >8 bits/channel. */ static enum pipe_format -default_deep_rgba_format(struct pipe_screen *screen, uint type) +default_deep_rgba_format(struct pipe_screen *screen, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags) { - if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SNORM, type)) { + if (screen->is_format_supported(screen, PIPE_FORMAT_R16G16B16A16_SNORM, target, tex_usage, geom_flags)) { return PIPE_FORMAT_R16G16B16A16_SNORM; } - if (type == PIPE_TEXTURE) - return default_rgba_format(screen, type); + if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) + return default_rgba_format(screen, target, tex_usage, geom_flags); else return PIPE_FORMAT_NONE; } @@ -319,7 +325,10 @@ default_deep_rgba_format(struct pipe_screen *screen, uint type) * Find an Z format supported by the context/winsys. */ static enum pipe_format -default_depth_format(struct pipe_screen *screen, uint type) +default_depth_format(struct pipe_screen *screen, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags) { static const enum pipe_format zFormats[] = { PIPE_FORMAT_Z16_UNORM, @@ -329,7 +338,7 @@ default_depth_format(struct pipe_screen *screen, uint type) }; uint i; for (i = 0; i < Elements(zFormats); i++) { - if (screen->is_format_supported( screen, zFormats[i], type )) { + if (screen->is_format_supported( screen, zFormats[i], target, tex_usage, geom_flags )) { return zFormats[i]; } } @@ -343,12 +352,10 @@ default_depth_format(struct pipe_screen *screen, uint type) */ enum pipe_format st_choose_format(struct pipe_context *pipe, GLint internalFormat, - uint surfType) + enum pipe_texture_target target, unsigned tex_usage) { struct pipe_screen *screen = pipe->screen; - - assert(surfType == PIPE_SURFACE || - surfType == PIPE_TEXTURE); + unsigned geom_flags = 0; switch (internalFormat) { case 4: @@ -360,38 +367,38 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_RGBA8: case GL_RGB10_A2: case GL_RGBA12: - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case GL_RGBA16: - if (surfType == PIPE_SURFACE) - return default_deep_rgba_format( screen, surfType ); + if (tex_usage & PIPE_TEXTURE_USAGE_RENDER_TARGET) + return default_deep_rgba_format( screen, target, tex_usage, geom_flags ); else - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case GL_RGBA4: case GL_RGBA2: - if (screen->is_format_supported( screen, PIPE_FORMAT_A4R4G4B4_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_A4R4G4B4_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_A4R4G4B4_UNORM; - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case GL_RGB5_A1: - if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_A1R5G5B5_UNORM; - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case GL_RGB8: case GL_RGB10: case GL_RGB12: case GL_RGB16: - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case GL_RGB5: case GL_RGB4: case GL_R3_G3_B2: - if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_A1R5G5B5_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_A1R5G5B5_UNORM; - if (screen->is_format_supported( screen, PIPE_FORMAT_R5G6B5_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_R5G6B5_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_R5G6B5_UNORM; - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case GL_ALPHA: case GL_ALPHA4: @@ -399,9 +406,9 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_ALPHA12: case GL_ALPHA16: case GL_COMPRESSED_ALPHA: - if (screen->is_format_supported( screen, PIPE_FORMAT_A8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_A8_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_A8_UNORM; - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case 1: case GL_LUMINANCE: @@ -410,9 +417,9 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12: case GL_LUMINANCE16: case GL_COMPRESSED_LUMINANCE: - if (screen->is_format_supported( screen, PIPE_FORMAT_L8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_L8_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_L8_UNORM; - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case 2: case GL_LUMINANCE_ALPHA: @@ -423,9 +430,9 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_LUMINANCE12_ALPHA12: case GL_LUMINANCE16_ALPHA16: case GL_COMPRESSED_LUMINANCE_ALPHA: - if (screen->is_format_supported( screen, PIPE_FORMAT_A8L8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_A8L8_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_A8L8_UNORM; - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case GL_INTENSITY: case GL_INTENSITY4: @@ -433,17 +440,17 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, case GL_INTENSITY12: case GL_INTENSITY16: case GL_COMPRESSED_INTENSITY: - if (screen->is_format_supported( screen, PIPE_FORMAT_I8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_I8_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_I8_UNORM; - return default_rgba_format( screen, surfType ); + return default_rgba_format( screen, target, tex_usage, geom_flags ); case GL_YCBCR_MESA: if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR, - PIPE_TEXTURE)) { + target, tex_usage, geom_flags)) { return PIPE_FORMAT_YCBCR; } if (screen->is_format_supported(screen, PIPE_FORMAT_YCBCR_REV, - PIPE_TEXTURE)) { + target, tex_usage, geom_flags)) { return PIPE_FORMAT_YCBCR_REV; } return PIPE_FORMAT_NONE; @@ -472,40 +479,40 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, #endif case GL_DEPTH_COMPONENT16: - if (screen->is_format_supported( screen, PIPE_FORMAT_Z16_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z16_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_Z16_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT24: - if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_S8Z24_UNORM; - if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_Z24S8_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT32: - if (screen->is_format_supported( screen, PIPE_FORMAT_Z32_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z32_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_Z32_UNORM; /* fall-through */ case GL_DEPTH_COMPONENT: - return default_depth_format( screen, surfType ); + return default_depth_format( screen, target, tex_usage, geom_flags ); case GL_STENCIL_INDEX: case GL_STENCIL_INDEX1_EXT: case GL_STENCIL_INDEX4_EXT: case GL_STENCIL_INDEX8_EXT: case GL_STENCIL_INDEX16_EXT: - if (screen->is_format_supported( screen, PIPE_FORMAT_S8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_S8_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_S8_UNORM; - if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_S8Z24_UNORM; - if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_Z24S8_UNORM; return PIPE_FORMAT_NONE; case GL_DEPTH_STENCIL_EXT: case GL_DEPTH24_STENCIL8_EXT: - if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_S8Z24_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_S8Z24_UNORM; - if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, surfType )) + if (screen->is_format_supported( screen, PIPE_FORMAT_Z24S8_UNORM, target, tex_usage, geom_flags )) return PIPE_FORMAT_Z24S8_UNORM; return PIPE_FORMAT_NONE; @@ -521,7 +528,8 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, enum pipe_format st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) { - return st_choose_format(pipe, internalFormat, PIPE_SURFACE); + return st_choose_format(pipe, internalFormat, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_RENDER_TARGET); } @@ -587,7 +595,8 @@ st_ChooseTextureFormat(GLcontext *ctx, GLint internalFormat, (void) format; (void) type; - pFormat = st_choose_format(ctx->st->pipe, internalFormat, PIPE_TEXTURE); + pFormat = st_choose_format(ctx->st->pipe, internalFormat, PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER); if (pFormat == PIPE_FORMAT_NONE) return NULL; diff --git a/src/mesa/state_tracker/st_format.h b/src/mesa/state_tracker/st_format.h index ff0fd042db..3f5ac3201b 100644 --- a/src/mesa/state_tracker/st_format.h +++ b/src/mesa/state_tracker/st_format.h @@ -65,7 +65,7 @@ st_mesa_format_to_pipe_format(GLuint mesaFormat); extern enum pipe_format st_choose_format(struct pipe_context *pipe, GLint internalFormat, - uint surfType); + enum pipe_texture_target target, unsigned tex_usage); extern enum pipe_format st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat); diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 2fc00df429..6db9bc0dd5 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -86,7 +86,8 @@ st_render_mipmap(struct st_context *st, assert(target != GL_TEXTURE_3D); /* not done yet */ /* check if we can render in the texture's format */ - if (!screen->is_format_supported(screen, pt->format, PIPE_SURFACE)) { + if (!screen->is_format_supported(screen, pt->format, target, + PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) { return FALSE; } diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 8222826e7a..3e5054ecd2 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -88,7 +88,8 @@ st_texture_create(struct st_context *st, _mesa_lookup_enum_by_nr(format), last_level); assert(format); - assert(screen->is_format_supported(screen, format, PIPE_TEXTURE)); + assert(screen->is_format_supported(screen, format, target, + PIPE_TEXTURE_USAGE_SAMPLER, 0)); memset(&pt, 0, sizeof(pt)); pt.target = target; -- cgit v1.2.3 From b1d6ff1afd3ec8460f746c76764dc89f9cd70556 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 19 Jul 2008 12:30:54 +0900 Subject: python/tests: Specify the right texture target when querying formats. --- src/gallium/state_trackers/python/tests/texture.py | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index b2ca9f416f..861070c8d2 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -136,7 +136,7 @@ class TextureTest(TestCase): level = self.level zslice = self.zslice - if not dev.is_format_supported(format, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_SAMPLER, 0): + if not dev.is_format_supported(format, target, PIPE_TEXTURE_USAGE_SAMPLER, 0): raise TestSkip ctx = self.dev.context_create() @@ -364,7 +364,8 @@ def main(): levels = lods(size) for last_level in range(levels): for level in range(0, last_level + 1): - for zslice in range(0, depth >> level): + zslice = 0 + while zslice < depth >> level: test = TextureTest( dev = dev, target = target, @@ -378,6 +379,7 @@ def main(): zslice = zslice, ) suite.add_test(test) + zslice = (zslice + 1)*2 - 1 suite.run() -- cgit v1.2.3 From 43df2fe2d9fe242174aba58b5de191c3d1fff212 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 20 Jul 2008 09:27:48 +0900 Subject: gallium: Fix typo in function name. --- src/gallium/auxiliary/util/p_tile.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 1daae6b640..1bf0d72733 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -346,7 +346,7 @@ r5g6b5_get_tile_rgba(ushort *src, static void -r5g5b5_put_tile_rgba(ushort *dst, +r5g6b5_put_tile_rgba(ushort *dst, unsigned w, unsigned h, const float *p, unsigned src_stride) @@ -827,7 +827,7 @@ pipe_put_tile_rgba(struct pipe_surface *ps, /*a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_R5G6B5_UNORM: - r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); + r5g6b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_R8G8B8A8_UNORM: assert(0); -- cgit v1.2.3 From 5853b6c2fe10223738a15d294e2a1605ec076953 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 20 Jul 2008 09:28:27 +0900 Subject: python/tests: Check support for non-pot/non-square textures. --- src/gallium/state_trackers/python/tests/texture.py | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/tests/texture.py b/src/gallium/state_trackers/python/tests/texture.py index 861070c8d2..880a61306c 100644 --- a/src/gallium/state_trackers/python/tests/texture.py +++ b/src/gallium/state_trackers/python/tests/texture.py @@ -94,7 +94,10 @@ def tex_coords(texture, face, level, zslice): rz = -1.0 result.append([rx, ry, rz]) return result - + +def is_pot(n): + return n & (n - 1) == 0 + class TextureTest(TestCase): @@ -136,7 +139,14 @@ class TextureTest(TestCase): level = self.level zslice = self.zslice - if not dev.is_format_supported(format, target, PIPE_TEXTURE_USAGE_SAMPLER, 0): + tex_usage = PIPE_TEXTURE_USAGE_SAMPLER + geom_flags = 0 + if width != height: + geom_flags |= PIPE_TEXTURE_GEOM_NON_SQUARE + if not is_pot(width) or not is_pot(height) or not is_pot(depth): + geom_flags |= PIPE_TEXTURE_GEOM_NON_POWER_OF_TWO + + if not dev.is_format_supported(format, target, tex_usage, geom_flags): raise TestSkip ctx = self.dev.context_create() @@ -199,7 +209,7 @@ class TextureTest(TestCase): height = height, depth = depth, last_level = last_level, - tex_usage = PIPE_TEXTURE_USAGE_SAMPLER, + tex_usage = tex_usage, ) expected_rgba = FloatArray(height*width*4) -- cgit v1.2.3 From 2170ec9048eab751828700728c1cc8264c860229 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 20 Jul 2008 18:11:43 +0900 Subject: pipebuffer: More detailed description of bufer over-/undereflows. --- src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c | 49 +++++++++++++++------- 1 file changed, 34 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c index affa6aa85c..d02e3500ff 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c @@ -109,7 +109,7 @@ static const uint8_t random_pattern[32] = { static INLINE void fill_random_pattern(uint8_t *dst, size_t size) { - unsigned i = 0; + size_t i = 0; while(size--) { *dst++ = random_pattern[i++]; i &= sizeof(random_pattern) - 1; @@ -118,15 +118,21 @@ fill_random_pattern(uint8_t *dst, size_t size) static INLINE boolean -check_random_pattern(const uint8_t *dst, size_t size) +check_random_pattern(const uint8_t *dst, size_t size, + size_t *min_ofs, size_t *max_ofs) { - unsigned i = 0; - while(size--) { - if(*dst++ != random_pattern[i++]) - return FALSE; - i &= sizeof(random_pattern) - 1; + boolean result = TRUE; + size_t i; + *min_ofs = size; + *max_ofs = 0; + for(i = 0; i < size; ++i) { + if(*dst++ != random_pattern[i % sizeof(random_pattern)]) { + *min_ofs = MIN2(*min_ofs, i); + *max_ofs = MIN2(*max_ofs, i); + result = FALSE; + } } - return TRUE; + return result; } @@ -141,15 +147,28 @@ pb_debug_buffer_destroy(struct pb_buffer *_buf) map = pb_map(buf->buffer, PIPE_BUFFER_USAGE_CPU_READ); assert(map); if(map) { - if(!check_random_pattern(map, buf->underflow_size)) { - debug_error("buffer underflow detected\n"); - debug_assert(0); + boolean underflow, overflow; + size_t min_ofs, max_ofs; + + underflow = !check_random_pattern(map, buf->underflow_size, + &min_ofs, &max_ofs); + if(underflow) { + debug_printf("buffer underflow (%u of %u bytes) detected\n", + buf->underflow_size - min_ofs, + buf->underflow_size); } - if(!check_random_pattern(map + buf->underflow_size + buf->base.base.size, - buf->overflow_size)) { - debug_error("buffer overflow detected\n"); - debug_assert(0); + + overflow = !check_random_pattern(map + buf->underflow_size + buf->base.base.size, + buf->overflow_size, + &min_ofs, &max_ofs); + if(overflow) { + debug_printf("buffer overflow (%u of %u bytes) detected\n", + max_ofs, + buf->overflow_size); } + + debug_assert(!underflow && !overflow); + pb_unmap(buf->buffer); } -- cgit v1.2.3 From 5ded4ffc506eb051b151d3e8b1e71b13576e951a Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 16:01:50 +0200 Subject: tgsi: Split tgsi_dump into two modules. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 791 +----------------------- src/gallium/auxiliary/tgsi/util/tgsi_dump.h | 17 +- src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c | 853 ++++++++++++++++++++++++++ src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h | 49 ++ 4 files changed, 916 insertions(+), 794 deletions(-) create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c create mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 0cf4454f25..5a84b99166 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -27,7 +27,6 @@ #include "pipe/p_debug.h" #include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" #include "util/u_string.h" #include "tgsi_dump.h" #include "tgsi_parse.h" @@ -56,13 +55,6 @@ dump_enum( #define FLT(F) debug_printf( "%10.4f", F ) #define ENM(E,ENUMS) dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) -static const char *TGSI_PROCESSOR_TYPES[] = -{ - "PROCESSOR_FRAGMENT", - "PROCESSOR_VERTEX", - "PROCESSOR_GEOMETRY" -}; - static const char *TGSI_PROCESSOR_TYPES_SHORT[] = { "FRAG", @@ -70,25 +62,6 @@ static const char *TGSI_PROCESSOR_TYPES_SHORT[] = "GEOM" }; -static const char *TGSI_TOKEN_TYPES[] = -{ - "TOKEN_TYPE_DECLARATION", - "TOKEN_TYPE_IMMEDIATE", - "TOKEN_TYPE_INSTRUCTION" -}; - -static const char *TGSI_FILES[] = -{ - "FILE_NULL", - "FILE_CONSTANT", - "FILE_INPUT", - "FILE_OUTPUT", - "FILE_TEMPORARY", - "FILE_SAMPLER", - "FILE_ADDRESS", - "FILE_IMMEDIATE" -}; - static const char *TGSI_FILES_SHORT[] = { "NULL", @@ -101,20 +74,6 @@ static const char *TGSI_FILES_SHORT[] = "IMM" }; -static const char *TGSI_DECLARES[] = -{ - "DECLARE_RANGE", - "DECLARE_MASK" -}; - -static const char *TGSI_INTERPOLATES[] = -{ - "INTERPOLATE_CONSTANT", - "INTERPOLATE_LINEAR", - "INTERPOLATE_PERSPECTIVE", - "INTERPOLATE_ATTRIB" -}; - static const char *TGSI_INTERPOLATES_SHORT[] = { "CONSTANT", @@ -122,17 +81,6 @@ static const char *TGSI_INTERPOLATES_SHORT[] = "PERSPECTIVE" }; -static const char *TGSI_SEMANTICS[] = -{ - "SEMANTIC_POSITION", - "SEMANTIC_COLOR", - "SEMANTIC_BCOLOR", - "SEMANTIC_FOG", - "SEMANTIC_PSIZE", - "SEMANTIC_GENERIC", - "SEMANTIC_NORMAL" -}; - static const char *TGSI_SEMANTICS_SHORT[] = { "POSITION", @@ -144,138 +92,11 @@ static const char *TGSI_SEMANTICS_SHORT[] = "NORMAL" }; -static const char *TGSI_IMMS[] = -{ - "IMM_FLOAT32" -}; - static const char *TGSI_IMMS_SHORT[] = { "FLT32" }; -static const char *TGSI_OPCODES[TGSI_OPCODE_LAST] = -{ - "OPCODE_ARL", - "OPCODE_MOV", - "OPCODE_LIT", - "OPCODE_RCP", - "OPCODE_RSQ", - "OPCODE_EXP", - "OPCODE_LOG", - "OPCODE_MUL", - "OPCODE_ADD", - "OPCODE_DP3", - "OPCODE_DP4", - "OPCODE_DST", - "OPCODE_MIN", - "OPCODE_MAX", - "OPCODE_SLT", - "OPCODE_SGE", - "OPCODE_MAD", - "OPCODE_SUB", - "OPCODE_LERP", - "OPCODE_CND", - "OPCODE_CND0", - "OPCODE_DOT2ADD", - "OPCODE_INDEX", - "OPCODE_NEGATE", - "OPCODE_FRAC", - "OPCODE_CLAMP", - "OPCODE_FLOOR", - "OPCODE_ROUND", - "OPCODE_EXPBASE2", - "OPCODE_LOGBASE2", - "OPCODE_POWER", - "OPCODE_CROSSPRODUCT", - "OPCODE_MULTIPLYMATRIX", - "OPCODE_ABS", - "OPCODE_RCC", - "OPCODE_DPH", - "OPCODE_COS", - "OPCODE_DDX", - "OPCODE_DDY", - "OPCODE_KILP", - "OPCODE_PK2H", - "OPCODE_PK2US", - "OPCODE_PK4B", - "OPCODE_PK4UB", - "OPCODE_RFL", - "OPCODE_SEQ", - "OPCODE_SFL", - "OPCODE_SGT", - "OPCODE_SIN", - "OPCODE_SLE", - "OPCODE_SNE", - "OPCODE_STR", - "OPCODE_TEX", - "OPCODE_TXD", - "OPCODE_TXP", - "OPCODE_UP2H", - "OPCODE_UP2US", - "OPCODE_UP4B", - "OPCODE_UP4UB", - "OPCODE_X2D", - "OPCODE_ARA", - "OPCODE_ARR", - "OPCODE_BRA", - "OPCODE_CAL", - "OPCODE_RET", - "OPCODE_SSG", - "OPCODE_CMP", - "OPCODE_SCS", - "OPCODE_TXB", - "OPCODE_NRM", - "OPCODE_DIV", - "OPCODE_DP2", - "OPCODE_TXL", - "OPCODE_BRK", - "OPCODE_IF", - "OPCODE_LOOP", - "OPCODE_REP", - "OPCODE_ELSE", - "OPCODE_ENDIF", - "OPCODE_ENDLOOP", - "OPCODE_ENDREP", - "OPCODE_PUSHA", - "OPCODE_POPA", - "OPCODE_CEIL", - "OPCODE_I2F", - "OPCODE_NOT", - "OPCODE_TRUNC", - "OPCODE_SHL", - "OPCODE_SHR", - "OPCODE_AND", - "OPCODE_OR", - "OPCODE_MOD", - "OPCODE_XOR", - "OPCODE_SAD", - "OPCODE_TXF", - "OPCODE_TXQ", - "OPCODE_CONT", - "OPCODE_EMIT", - "OPCODE_ENDPRIM", - "OPCODE_BGNLOOP2", - "OPCODE_BGNSUB", - "OPCODE_ENDLOOP2", - "OPCODE_ENDSUB", - "OPCODE_NOISE1", - "OPCODE_NOISE2", - "OPCODE_NOISE3", - "OPCODE_NOISE4", - "OPCODE_NOP", - "OPCODE_M4X3", - "OPCODE_M3X4", - "OPCODE_M3X3", - "OPCODE_M3X2", - "OPCODE_NRM4", - "OPCODE_CALLNZ", - "OPCODE_IFC", - "OPCODE_BREAKC", - "OPCODE_KIL", - "OPCODE_END" -}; - static const char *TGSI_OPCODES_SHORT[TGSI_OPCODE_LAST] = { "ARL", @@ -399,49 +220,6 @@ static const char *TGSI_OPCODES_SHORT[TGSI_OPCODE_LAST] = "SWZ" }; -static const char *TGSI_SATS[] = -{ - "SAT_NONE", - "SAT_ZERO_ONE", - "SAT_MINUS_PLUS_ONE" -}; - -static const char *TGSI_INSTRUCTION_EXTS[] = -{ - "INSTRUCTION_EXT_TYPE_NV", - "INSTRUCTION_EXT_TYPE_LABEL", - "INSTRUCTION_EXT_TYPE_TEXTURE" -}; - -static const char *TGSI_PRECISIONS[] = -{ - "PRECISION_DEFAULT", - "TGSI_PRECISION_FLOAT32", - "TGSI_PRECISION_FLOAT16", - "TGSI_PRECISION_FIXED12" -}; - -static const char *TGSI_CCS[] = -{ - "CC_GT", - "CC_EQ", - "CC_LT", - "CC_UN", - "CC_GE", - "CC_LE", - "CC_NE", - "CC_TR", - "CC_FL" -}; - -static const char *TGSI_SWIZZLES[] = -{ - "SWIZZLE_X", - "SWIZZLE_Y", - "SWIZZLE_Z", - "SWIZZLE_W" -}; - static const char *TGSI_SWIZZLES_SHORT[] = { "x", @@ -450,19 +228,6 @@ static const char *TGSI_SWIZZLES_SHORT[] = "w" }; -static const char *TGSI_TEXTURES[] = -{ - "TEXTURE_UNKNOWN", - "TEXTURE_1D", - "TEXTURE_2D", - "TEXTURE_3D", - "TEXTURE_CUBE", - "TEXTURE_RECT", - "TEXTURE_SHADOW1D", - "TEXTURE_SHADOW2D", - "TEXTURE_SHADOWRECT" -}; - static const char *TGSI_TEXTURES_SHORT[] = { "UNKNOWN", @@ -476,22 +241,6 @@ static const char *TGSI_TEXTURES_SHORT[] = "SHADOWRECT" }; -static const char *TGSI_SRC_REGISTER_EXTS[] = -{ - "SRC_REGISTER_EXT_TYPE_SWZ", - "SRC_REGISTER_EXT_TYPE_MOD" -}; - -static const char *TGSI_EXTSWIZZLES[] = -{ - "EXTSWIZZLE_X", - "EXTSWIZZLE_Y", - "EXTSWIZZLE_Z", - "EXTSWIZZLE_W", - "EXTSWIZZLE_ZERO", - "EXTSWIZZLE_ONE" -}; - static const char *TGSI_EXTSWIZZLES_SHORT[] = { "x", @@ -502,43 +251,6 @@ static const char *TGSI_EXTSWIZZLES_SHORT[] = "1" }; -static const char *TGSI_WRITEMASKS[] = -{ - "0", - "WRITEMASK_X", - "WRITEMASK_Y", - "WRITEMASK_XY", - "WRITEMASK_Z", - "WRITEMASK_XZ", - "WRITEMASK_YZ", - "WRITEMASK_XYZ", - "WRITEMASK_W", - "WRITEMASK_XW", - "WRITEMASK_YW", - "WRITEMASK_XYW", - "WRITEMASK_ZW", - "WRITEMASK_XZW", - "WRITEMASK_YZW", - "WRITEMASK_XYZW" -}; - -static const char *TGSI_DST_REGISTER_EXTS[] = -{ - "DST_REGISTER_EXT_TYPE_CONDCODE", - "DST_REGISTER_EXT_TYPE_MODULATE" -}; - -static const char *TGSI_MODULATES[] = -{ - "MODULATE_1X", - "MODULATE_2X", - "MODULATE_4X", - "MODULATE_8X", - "MODULATE_HALF", - "MODULATE_QUARTER", - "MODULATE_EIGHTH" -}; - static const char *TGSI_MODULATES_SHORT[TGSI_MODULATE_COUNT] = { "", @@ -552,7 +264,7 @@ static const char *TGSI_MODULATES_SHORT[TGSI_MODULATE_COUNT] = void tgsi_dump_declaration( - const struct tgsi_full_declaration *decl ) + const struct tgsi_full_declaration *decl ) { TXT( "\nDCL " ); ENM( decl->Declaration.File, TGSI_FILES_SHORT ); @@ -596,62 +308,6 @@ tgsi_dump_declaration( ENM( decl->Declaration.Interpolate, TGSI_INTERPOLATES_SHORT ); } -static void -dump_declaration_verbose( - struct tgsi_full_declaration *decl, - unsigned ignored, - unsigned deflt, - struct tgsi_full_declaration *fd ) -{ - TXT( "\nFile : " ); - ENM( decl->Declaration.File, TGSI_FILES ); - if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { - TXT( "\nUsageMask : " ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'X' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'Y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'Z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'W' ); - } - } - if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { - TXT( "\nInterpolate: " ); - ENM( decl->Declaration.Interpolate, TGSI_INTERPOLATES ); - } - if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { - TXT( "\nSemantic : " ); - UID( decl->Declaration.Semantic ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Declaration.Padding ); - } - - EOL(); - TXT( "\nFirst: " ); - UID( decl->DeclarationRange.First ); - TXT( "\nLast : " ); - UID( decl->DeclarationRange.Last ); - - if( decl->Declaration.Semantic ) { - EOL(); - TXT( "\nSemanticName : " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); - TXT( "\nSemanticIndex: " ); - UID( decl->Semantic.SemanticIndex ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Semantic.Padding ); - } - } -} - void tgsi_dump_immediate( const struct tgsi_full_immediate *imm ) @@ -679,38 +335,10 @@ tgsi_dump_immediate( TXT( " }" ); } -static void -dump_immediate_verbose( - struct tgsi_full_immediate *imm, - unsigned ignored ) -{ - unsigned i; - - TXT( "\nDataType : " ); - ENM( imm->Immediate.DataType, TGSI_IMMS ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( imm->Immediate.Padding ); - } - - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - EOL(); - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - TXT( "\nFloat: " ); - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - } -} - void tgsi_dump_instruction( - const struct tgsi_full_instruction *inst, - unsigned instno ) + const struct tgsi_full_instruction *inst, + uint instno ) { unsigned i; boolean first_reg = TRUE; @@ -856,389 +484,28 @@ tgsi_dump_instruction( } } -static void -dump_instruction_verbose( - struct tgsi_full_instruction *inst, - unsigned ignored, - unsigned deflt, - struct tgsi_full_instruction *fi ) -{ - unsigned i; - - TXT( "\nOpcode : " ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES ); - if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { - TXT( "\nSaturate : " ); - ENM( inst->Instruction.Saturate, TGSI_SATS ); - } - if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { - TXT( "\nNumDstRegs : " ); - UID( inst->Instruction.NumDstRegs ); - } - if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { - TXT( "\nNumSrcRegs : " ); - UID( inst->Instruction.NumSrcRegs ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->Instruction.Padding ); - } - - if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { - EOL(); - TXT( "\nType : " ); - ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { - TXT( "\nPrecision : " ); - ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); - } - if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { - TXT( "\nCondDstIndex : " ); - UID( inst->InstructionExtNv.CondDstIndex ); - } - if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { - TXT( "\nCondFlowIndex : " ); - UID( inst->InstructionExtNv.CondFlowIndex ); - } - if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { - TXT( "\nCondMask : " ); - ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { - TXT( "\nCondSwizzleX : " ); - ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { - TXT( "\nCondSwizzleY : " ); - ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ : " ); - ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { - TXT( "\nCondSwizzleW : " ); - ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { - TXT( "\nCondDstUpdate : " ); - UID( inst->InstructionExtNv.CondDstUpdate ); - } - if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { - TXT( "\nCondFlowEnable: " ); - UID( inst->InstructionExtNv.CondFlowEnable ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtNv.Padding ); - if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { - TXT( "\nExtended : " ); - UID( inst->InstructionExtNv.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { - EOL(); - TXT( "\nType : " ); - ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { - TXT( "\nLabel : " ); - UID( inst->InstructionExtLabel.Label ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtLabel.Padding ); - if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtLabel.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { - EOL(); - TXT( "\nType : " ); - ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { - TXT( "\nTexture : " ); - ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtTexture.Padding ); - if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtTexture.Extended ); - } - } - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; - - EOL(); - TXT( "\nFile : " ); - ENM( dst->DstRegister.File, TGSI_FILES ); - if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { - TXT( "\nWriteMask: " ); - ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); - } - if( ignored ) { - if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( dst->DstRegister.Indirect ); - } - if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( dst->DstRegister.Dimension ); - } - } - if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { - TXT( "\nIndex : " ); - SID( dst->DstRegister.Index ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegister.Padding ); - if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegister.Extended ); - } - } - - if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { - EOL(); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { - TXT( "\nCondMask : " ); - ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { - TXT( "\nCondSwizzleX: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { - TXT( "\nCondSwizzleY: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { - TXT( "\nCondSwizzleW: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { - TXT( "\nCondSrcIndex: " ); - UID( dst->DstRegisterExtConcode.CondSrcIndex ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtConcode.Padding ); - if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegisterExtConcode.Extended ); - } - } - } - - if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { - EOL(); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { - TXT( "\nModulate: " ); - ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtModulate.Padding ); - if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { - TXT( "\nExtended: " ); - UID( dst->DstRegisterExtModulate.Extended ); - } - } - } - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; - - EOL(); - TXT( "\nFile : "); - ENM( src->SrcRegister.File, TGSI_FILES ); - if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { - TXT( "\nSwizzleX : " ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { - TXT( "\nSwizzleY : " ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { - TXT( "\nSwizzleZ : " ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { - TXT( "\nSwizzleW : " ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegister.Negate ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( src->SrcRegister.Indirect ); - } - if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( src->SrcRegister.Dimension ); - } - } - if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { - TXT( "\nIndex : " ); - SID( src->SrcRegister.Index ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegister.Extended ); - } - } - - if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { - EOL(); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { - TXT( "\nExtSwizzleX: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { - TXT( "\nExtSwizzleY: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { - TXT( "\nExtSwizzleZ: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { - TXT( "\nExtSwizzleW: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { - TXT( "\nNegateX : " ); - UID( src->SrcRegisterExtSwz.NegateX ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { - TXT( "\nNegateY : " ); - UID( src->SrcRegisterExtSwz.NegateY ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { - TXT( "\nNegateZ : " ); - UID( src->SrcRegisterExtSwz.NegateZ ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { - TXT( "\nNegateW : " ); - UID( src->SrcRegisterExtSwz.NegateW ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtSwz.Padding ); - if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtSwz.Extended ); - } - } - } - - if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { - EOL(); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { - TXT( "\nComplement: " ); - UID( src->SrcRegisterExtMod.Complement ); - } - if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { - TXT( "\nBias : " ); - UID( src->SrcRegisterExtMod.Bias ); - } - if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { - TXT( "\nScale2X : " ); - UID( src->SrcRegisterExtMod.Scale2X ); - } - if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { - TXT( "\nAbsolute : " ); - UID( src->SrcRegisterExtMod.Absolute ); - } - if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegisterExtMod.Negate ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtMod.Padding ); - if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtMod.Extended ); - } - } - } - } -} - void tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ) + const struct tgsi_token *tokens ) { struct tgsi_parse_context parse; struct tgsi_full_instruction fi; struct tgsi_full_declaration fd; - unsigned verbose = flags & TGSI_DUMP_VERBOSE; - unsigned ignored = !(flags & TGSI_DUMP_NO_IGNORED); - unsigned deflt = !(flags & TGSI_DUMP_NO_DEFAULT); unsigned instno = 0; /* sanity checks */ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); + assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); + assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "OPCODE_END") == 0); assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "END") == 0); tgsi_parse_init( &parse, tokens ); - TXT( "tgsi-dump begin -----------------" ); - EOL(); ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); UID( parse.FullVersion.Version.MajorVersion ); CHR( '.' ); UID( parse.FullVersion.Version.MinorVersion ); - if( verbose ) { - TXT( "\nMajorVersion: " ); - UID( parse.FullVersion.Version.MajorVersion ); - TXT( "\nMinorVersion: " ); - UID( parse.FullVersion.Version.MinorVersion ); - EOL(); - - TXT( "\nHeaderSize: " ); - UID( parse.FullHeader.Header.HeaderSize ); - TXT( "\nBodySize : " ); - UID( parse.FullHeader.Header.BodySize ); - TXT( "\nProcessor : " ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); - EOL(); - } - fi = tgsi_default_full_instruction(); fd = tgsi_default_full_declaration(); @@ -1266,51 +533,7 @@ tgsi_dump( default: assert( 0 ); } - - if( verbose ) { - TXT( "\nType : " ); - ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); - if( ignored ) { - TXT( "\nSize : " ); - UID( parse.FullToken.Token.Size ); - if( deflt || parse.FullToken.Token.Extended ) { - TXT( "\nExtended : " ); - UID( parse.FullToken.Token.Extended ); - } - } - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_verbose( - &parse.FullToken.FullDeclaration, - ignored, - deflt, - &fd ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_verbose( - &parse.FullToken.FullImmediate, - ignored ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_verbose( - &parse.FullToken.FullInstruction, - ignored, - deflt, - &fi ); - break; - - default: - assert( 0 ); - } - - EOL(); - } } - TXT( "\ntgsi-dump end -------------------\n" ); - tgsi_parse_free( &parse ); } diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h index ba7692b511..c130b9f1b7 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -28,18 +28,15 @@ #ifndef TGSI_DUMP_H #define TGSI_DUMP_H +#include "pipe/p_shader_tokens.h" + #if defined __cplusplus extern "C" { #endif -#define TGSI_DUMP_VERBOSE 1 -#define TGSI_DUMP_NO_IGNORED 2 -#define TGSI_DUMP_NO_DEFAULT 4 - void tgsi_dump( - const struct tgsi_token *tokens, - unsigned flags ); + const struct tgsi_token *tokens ); struct tgsi_full_immediate; struct tgsi_full_instruction; @@ -51,12 +48,12 @@ tgsi_dump_immediate( void tgsi_dump_instruction( - const struct tgsi_full_instruction *inst, - unsigned instno ); + const struct tgsi_full_instruction *inst, + uint instno ); void tgsi_dump_declaration( - const struct tgsi_full_declaration *decl ); + const struct tgsi_full_declaration *decl ); #if defined __cplusplus } diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c new file mode 100644 index 0000000000..27b085dc07 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c @@ -0,0 +1,853 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "util/u_string.h" +#include "tgsi_dump_c.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" + +static void +dump_enum( + const unsigned e, + const char **enums, + const unsigned enums_count ) +{ + if (e >= enums_count) { + debug_printf( "%u", e ); + } + else { + debug_printf( "%s", enums[e] ); + } +} + +#define EOL() debug_printf( "\n" ) +#define TXT(S) debug_printf( "%s", S ) +#define CHR(C) debug_printf( "%c", C ) +#define UIX(I) debug_printf( "0x%x", I ) +#define UID(I) debug_printf( "%u", I ) +#define SID(I) debug_printf( "%d", I ) +#define FLT(F) debug_printf( "%10.4f", F ) +#define ENM(E,ENUMS) dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) + +static const char *TGSI_PROCESSOR_TYPES[] = +{ + "PROCESSOR_FRAGMENT", + "PROCESSOR_VERTEX", + "PROCESSOR_GEOMETRY" +}; + +static const char *TGSI_TOKEN_TYPES[] = +{ + "TOKEN_TYPE_DECLARATION", + "TOKEN_TYPE_IMMEDIATE", + "TOKEN_TYPE_INSTRUCTION" +}; + +static const char *TGSI_FILES[] = +{ + "FILE_NULL", + "FILE_CONSTANT", + "FILE_INPUT", + "FILE_OUTPUT", + "FILE_TEMPORARY", + "FILE_SAMPLER", + "FILE_ADDRESS", + "FILE_IMMEDIATE" +}; + +static const char *TGSI_DECLARES[] = +{ + "DECLARE_RANGE", + "DECLARE_MASK" +}; + +static const char *TGSI_INTERPOLATES[] = +{ + "INTERPOLATE_CONSTANT", + "INTERPOLATE_LINEAR", + "INTERPOLATE_PERSPECTIVE", + "INTERPOLATE_ATTRIB" +}; + +static const char *TGSI_SEMANTICS[] = +{ + "SEMANTIC_POSITION", + "SEMANTIC_COLOR", + "SEMANTIC_BCOLOR", + "SEMANTIC_FOG", + "SEMANTIC_PSIZE", + "SEMANTIC_GENERIC", + "SEMANTIC_NORMAL" +}; + +static const char *TGSI_IMMS[] = +{ + "IMM_FLOAT32" +}; + +static const char *TGSI_OPCODES[TGSI_OPCODE_LAST] = +{ + "OPCODE_ARL", + "OPCODE_MOV", + "OPCODE_LIT", + "OPCODE_RCP", + "OPCODE_RSQ", + "OPCODE_EXP", + "OPCODE_LOG", + "OPCODE_MUL", + "OPCODE_ADD", + "OPCODE_DP3", + "OPCODE_DP4", + "OPCODE_DST", + "OPCODE_MIN", + "OPCODE_MAX", + "OPCODE_SLT", + "OPCODE_SGE", + "OPCODE_MAD", + "OPCODE_SUB", + "OPCODE_LERP", + "OPCODE_CND", + "OPCODE_CND0", + "OPCODE_DOT2ADD", + "OPCODE_INDEX", + "OPCODE_NEGATE", + "OPCODE_FRAC", + "OPCODE_CLAMP", + "OPCODE_FLOOR", + "OPCODE_ROUND", + "OPCODE_EXPBASE2", + "OPCODE_LOGBASE2", + "OPCODE_POWER", + "OPCODE_CROSSPRODUCT", + "OPCODE_MULTIPLYMATRIX", + "OPCODE_ABS", + "OPCODE_RCC", + "OPCODE_DPH", + "OPCODE_COS", + "OPCODE_DDX", + "OPCODE_DDY", + "OPCODE_KILP", + "OPCODE_PK2H", + "OPCODE_PK2US", + "OPCODE_PK4B", + "OPCODE_PK4UB", + "OPCODE_RFL", + "OPCODE_SEQ", + "OPCODE_SFL", + "OPCODE_SGT", + "OPCODE_SIN", + "OPCODE_SLE", + "OPCODE_SNE", + "OPCODE_STR", + "OPCODE_TEX", + "OPCODE_TXD", + "OPCODE_TXP", + "OPCODE_UP2H", + "OPCODE_UP2US", + "OPCODE_UP4B", + "OPCODE_UP4UB", + "OPCODE_X2D", + "OPCODE_ARA", + "OPCODE_ARR", + "OPCODE_BRA", + "OPCODE_CAL", + "OPCODE_RET", + "OPCODE_SSG", + "OPCODE_CMP", + "OPCODE_SCS", + "OPCODE_TXB", + "OPCODE_NRM", + "OPCODE_DIV", + "OPCODE_DP2", + "OPCODE_TXL", + "OPCODE_BRK", + "OPCODE_IF", + "OPCODE_LOOP", + "OPCODE_REP", + "OPCODE_ELSE", + "OPCODE_ENDIF", + "OPCODE_ENDLOOP", + "OPCODE_ENDREP", + "OPCODE_PUSHA", + "OPCODE_POPA", + "OPCODE_CEIL", + "OPCODE_I2F", + "OPCODE_NOT", + "OPCODE_TRUNC", + "OPCODE_SHL", + "OPCODE_SHR", + "OPCODE_AND", + "OPCODE_OR", + "OPCODE_MOD", + "OPCODE_XOR", + "OPCODE_SAD", + "OPCODE_TXF", + "OPCODE_TXQ", + "OPCODE_CONT", + "OPCODE_EMIT", + "OPCODE_ENDPRIM", + "OPCODE_BGNLOOP2", + "OPCODE_BGNSUB", + "OPCODE_ENDLOOP2", + "OPCODE_ENDSUB", + "OPCODE_NOISE1", + "OPCODE_NOISE2", + "OPCODE_NOISE3", + "OPCODE_NOISE4", + "OPCODE_NOP", + "OPCODE_M4X3", + "OPCODE_M3X4", + "OPCODE_M3X3", + "OPCODE_M3X2", + "OPCODE_NRM4", + "OPCODE_CALLNZ", + "OPCODE_IFC", + "OPCODE_BREAKC", + "OPCODE_KIL", + "OPCODE_END" +}; + +static const char *TGSI_SATS[] = +{ + "SAT_NONE", + "SAT_ZERO_ONE", + "SAT_MINUS_PLUS_ONE" +}; + +static const char *TGSI_INSTRUCTION_EXTS[] = +{ + "INSTRUCTION_EXT_TYPE_NV", + "INSTRUCTION_EXT_TYPE_LABEL", + "INSTRUCTION_EXT_TYPE_TEXTURE" +}; + +static const char *TGSI_PRECISIONS[] = +{ + "PRECISION_DEFAULT", + "TGSI_PRECISION_FLOAT32", + "TGSI_PRECISION_FLOAT16", + "TGSI_PRECISION_FIXED12" +}; + +static const char *TGSI_CCS[] = +{ + "CC_GT", + "CC_EQ", + "CC_LT", + "CC_UN", + "CC_GE", + "CC_LE", + "CC_NE", + "CC_TR", + "CC_FL" +}; + +static const char *TGSI_SWIZZLES[] = +{ + "SWIZZLE_X", + "SWIZZLE_Y", + "SWIZZLE_Z", + "SWIZZLE_W" +}; + +static const char *TGSI_TEXTURES[] = +{ + "TEXTURE_UNKNOWN", + "TEXTURE_1D", + "TEXTURE_2D", + "TEXTURE_3D", + "TEXTURE_CUBE", + "TEXTURE_RECT", + "TEXTURE_SHADOW1D", + "TEXTURE_SHADOW2D", + "TEXTURE_SHADOWRECT" +}; + +static const char *TGSI_SRC_REGISTER_EXTS[] = +{ + "SRC_REGISTER_EXT_TYPE_SWZ", + "SRC_REGISTER_EXT_TYPE_MOD" +}; + +static const char *TGSI_EXTSWIZZLES[] = +{ + "EXTSWIZZLE_X", + "EXTSWIZZLE_Y", + "EXTSWIZZLE_Z", + "EXTSWIZZLE_W", + "EXTSWIZZLE_ZERO", + "EXTSWIZZLE_ONE" +}; + +static const char *TGSI_WRITEMASKS[] = +{ + "0", + "WRITEMASK_X", + "WRITEMASK_Y", + "WRITEMASK_XY", + "WRITEMASK_Z", + "WRITEMASK_XZ", + "WRITEMASK_YZ", + "WRITEMASK_XYZ", + "WRITEMASK_W", + "WRITEMASK_XW", + "WRITEMASK_YW", + "WRITEMASK_XYW", + "WRITEMASK_ZW", + "WRITEMASK_XZW", + "WRITEMASK_YZW", + "WRITEMASK_XYZW" +}; + +static const char *TGSI_DST_REGISTER_EXTS[] = +{ + "DST_REGISTER_EXT_TYPE_CONDCODE", + "DST_REGISTER_EXT_TYPE_MODULATE" +}; + +static const char *TGSI_MODULATES[] = +{ + "MODULATE_1X", + "MODULATE_2X", + "MODULATE_4X", + "MODULATE_8X", + "MODULATE_HALF", + "MODULATE_QUARTER", + "MODULATE_EIGHTH" +}; + +static void +dump_declaration_verbose( + struct tgsi_full_declaration *decl, + unsigned ignored, + unsigned deflt, + struct tgsi_full_declaration *fd ) +{ + TXT( "\nFile : " ); + ENM( decl->Declaration.File, TGSI_FILES ); + if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { + TXT( "\nUsageMask : " ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'X' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'Y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'Z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'W' ); + } + } + if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { + TXT( "\nInterpolate: " ); + ENM( decl->Declaration.Interpolate, TGSI_INTERPOLATES ); + } + if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { + TXT( "\nSemantic : " ); + UID( decl->Declaration.Semantic ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Declaration.Padding ); + } + + EOL(); + TXT( "\nFirst: " ); + UID( decl->DeclarationRange.First ); + TXT( "\nLast : " ); + UID( decl->DeclarationRange.Last ); + + if( decl->Declaration.Semantic ) { + EOL(); + TXT( "\nSemanticName : " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); + TXT( "\nSemanticIndex: " ); + UID( decl->Semantic.SemanticIndex ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Semantic.Padding ); + } + } +} + +static void +dump_immediate_verbose( + struct tgsi_full_immediate *imm, + unsigned ignored ) +{ + unsigned i; + + TXT( "\nDataType : " ); + ENM( imm->Immediate.DataType, TGSI_IMMS ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( imm->Immediate.Padding ); + } + + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + EOL(); + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + TXT( "\nFloat: " ); + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + } +} + +static void +dump_instruction_verbose( + struct tgsi_full_instruction *inst, + unsigned ignored, + unsigned deflt, + struct tgsi_full_instruction *fi ) +{ + unsigned i; + + TXT( "\nOpcode : " ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES ); + if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { + TXT( "\nSaturate : " ); + ENM( inst->Instruction.Saturate, TGSI_SATS ); + } + if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { + TXT( "\nNumDstRegs : " ); + UID( inst->Instruction.NumDstRegs ); + } + if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { + TXT( "\nNumSrcRegs : " ); + UID( inst->Instruction.NumSrcRegs ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->Instruction.Padding ); + } + + if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { + EOL(); + TXT( "\nType : " ); + ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { + TXT( "\nPrecision : " ); + ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); + } + if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { + TXT( "\nCondDstIndex : " ); + UID( inst->InstructionExtNv.CondDstIndex ); + } + if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { + TXT( "\nCondFlowIndex : " ); + UID( inst->InstructionExtNv.CondFlowIndex ); + } + if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { + TXT( "\nCondMask : " ); + ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { + TXT( "\nCondSwizzleX : " ); + ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { + TXT( "\nCondSwizzleY : " ); + ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ : " ); + ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { + TXT( "\nCondSwizzleW : " ); + ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { + TXT( "\nCondDstUpdate : " ); + UID( inst->InstructionExtNv.CondDstUpdate ); + } + if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { + TXT( "\nCondFlowEnable: " ); + UID( inst->InstructionExtNv.CondFlowEnable ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtNv.Padding ); + if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { + TXT( "\nExtended : " ); + UID( inst->InstructionExtNv.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { + EOL(); + TXT( "\nType : " ); + ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { + TXT( "\nLabel : " ); + UID( inst->InstructionExtLabel.Label ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtLabel.Padding ); + if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtLabel.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { + EOL(); + TXT( "\nType : " ); + ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { + TXT( "\nTexture : " ); + ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtTexture.Padding ); + if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtTexture.Extended ); + } + } + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; + + EOL(); + TXT( "\nFile : " ); + ENM( dst->DstRegister.File, TGSI_FILES ); + if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { + TXT( "\nWriteMask: " ); + ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); + } + if( ignored ) { + if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( dst->DstRegister.Indirect ); + } + if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( dst->DstRegister.Dimension ); + } + } + if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { + TXT( "\nIndex : " ); + SID( dst->DstRegister.Index ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegister.Padding ); + if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegister.Extended ); + } + } + + if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { + EOL(); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { + TXT( "\nCondMask : " ); + ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { + TXT( "\nCondSwizzleX: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { + TXT( "\nCondSwizzleY: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { + TXT( "\nCondSwizzleW: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { + TXT( "\nCondSrcIndex: " ); + UID( dst->DstRegisterExtConcode.CondSrcIndex ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtConcode.Padding ); + if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegisterExtConcode.Extended ); + } + } + } + + if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { + EOL(); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { + TXT( "\nModulate: " ); + ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtModulate.Padding ); + if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { + TXT( "\nExtended: " ); + UID( dst->DstRegisterExtModulate.Extended ); + } + } + } + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; + + EOL(); + TXT( "\nFile : "); + ENM( src->SrcRegister.File, TGSI_FILES ); + if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { + TXT( "\nSwizzleX : " ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { + TXT( "\nSwizzleY : " ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { + TXT( "\nSwizzleZ : " ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { + TXT( "\nSwizzleW : " ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegister.Negate ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( src->SrcRegister.Indirect ); + } + if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( src->SrcRegister.Dimension ); + } + } + if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { + TXT( "\nIndex : " ); + SID( src->SrcRegister.Index ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegister.Extended ); + } + } + + if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { + EOL(); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { + TXT( "\nExtSwizzleX: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { + TXT( "\nExtSwizzleY: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { + TXT( "\nExtSwizzleZ: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { + TXT( "\nExtSwizzleW: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { + TXT( "\nNegateX : " ); + UID( src->SrcRegisterExtSwz.NegateX ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { + TXT( "\nNegateY : " ); + UID( src->SrcRegisterExtSwz.NegateY ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { + TXT( "\nNegateZ : " ); + UID( src->SrcRegisterExtSwz.NegateZ ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { + TXT( "\nNegateW : " ); + UID( src->SrcRegisterExtSwz.NegateW ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtSwz.Padding ); + if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtSwz.Extended ); + } + } + } + + if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { + EOL(); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { + TXT( "\nComplement: " ); + UID( src->SrcRegisterExtMod.Complement ); + } + if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { + TXT( "\nBias : " ); + UID( src->SrcRegisterExtMod.Bias ); + } + if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { + TXT( "\nScale2X : " ); + UID( src->SrcRegisterExtMod.Scale2X ); + } + if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { + TXT( "\nAbsolute : " ); + UID( src->SrcRegisterExtMod.Absolute ); + } + if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegisterExtMod.Negate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtMod.Padding ); + if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtMod.Extended ); + } + } + } + } +} + +void +tgsi_dump_c( + const struct tgsi_token *tokens, + uint flags ) +{ + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + uint ignored = flags & TGSI_DUMP_C_IGNORED; + uint deflt = flags & TGSI_DUMP_C_DEFAULT; + uint instno = 0; + + /* sanity checks */ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "END") == 0); + + tgsi_parse_init( &parse, tokens ); + + TXT( "tgsi-dump begin -----------------" ); + + TXT( "\nMajorVersion: " ); + UID( parse.FullVersion.Version.MajorVersion ); + TXT( "\nMinorVersion: " ); + UID( parse.FullVersion.Version.MinorVersion ); + EOL(); + + TXT( "\nHeaderSize: " ); + UID( parse.FullHeader.Header.HeaderSize ); + TXT( "\nBodySize : " ); + UID( parse.FullHeader.Header.BodySize ); + TXT( "\nProcessor : " ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + EOL(); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + TXT( "\nType : " ); + ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); + if( ignored ) { + TXT( "\nSize : " ); + UID( parse.FullToken.Token.Size ); + if( deflt || parse.FullToken.Token.Extended ) { + TXT( "\nExtended : " ); + UID( parse.FullToken.Token.Extended ); + } + } + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_verbose( + &parse.FullToken.FullDeclaration, + ignored, + deflt, + &fd ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_verbose( + &parse.FullToken.FullImmediate, + ignored ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_verbose( + &parse.FullToken.FullInstruction, + ignored, + deflt, + &fi ); + break; + + default: + assert( 0 ); + } + + EOL(); + } + + TXT( "\ntgsi-dump end -------------------\n" ); + + tgsi_parse_free( &parse ); +} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h new file mode 100644 index 0000000000..d91cd35b3b --- /dev/null +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h @@ -0,0 +1,49 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_DUMP_C_H +#define TGSI_DUMP_C_H + +#include "pipe/p_shader_tokens.h" + +#if defined __cplusplus +extern "C" { +#endif + +#define TGSI_DUMP_C_IGNORED 1 +#define TGSI_DUMP_C_DEFAULT 2 + +void +tgsi_dump_c( + const struct tgsi_token *tokens, + uint flags ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_DUMP_C_H */ -- cgit v1.2.3 From 5e9ea9d938e9734524707e46be29e243e96f32a9 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 16:07:06 +0200 Subject: scons: List util/tgsi_dump_c.c. --- src/gallium/auxiliary/tgsi/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript index 54f5c75db4..3bbfa1be54 100644 --- a/src/gallium/auxiliary/tgsi/SConscript +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -7,6 +7,7 @@ tgsi = env.ConvenienceLibrary( 'exec/tgsi_sse2.c', 'util/tgsi_build.c', 'util/tgsi_dump.c', + 'util/tgsi_dump_c.c', 'util/tgsi_iterate.c', 'util/tgsi_parse.c', 'util/tgsi_sanity.c', -- cgit v1.2.3 From 15c902455fe1b4572e614bf30912d92fe9c7bb28 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 16:08:36 +0200 Subject: tgsi: Preserve flags parameter for tgsi_dump(). --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 3 ++- src/gallium/auxiliary/tgsi/util/tgsi_dump.h | 3 ++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 5a84b99166..1056951bc8 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -486,7 +486,8 @@ tgsi_dump_instruction( void tgsi_dump( - const struct tgsi_token *tokens ) + const struct tgsi_token *tokens, + uint flags ) { struct tgsi_parse_context parse; struct tgsi_full_instruction fi; diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h index c130b9f1b7..51c230b5db 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h @@ -36,7 +36,8 @@ extern "C" { void tgsi_dump( - const struct tgsi_token *tokens ); + const struct tgsi_token *tokens, + uint flags ); struct tgsi_full_immediate; struct tgsi_full_instruction; -- cgit v1.2.3 From 307a252d1529e8e9ffa10c88cc8f5d5412f587f5 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 16:12:44 +0200 Subject: tgsi: Fix dump enums. --- src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c | 15 ++++----------- 1 file changed, 4 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c index 27b085dc07..acb1246b7a 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c @@ -81,18 +81,11 @@ static const char *TGSI_FILES[] = "FILE_IMMEDIATE" }; -static const char *TGSI_DECLARES[] = -{ - "DECLARE_RANGE", - "DECLARE_MASK" -}; - static const char *TGSI_INTERPOLATES[] = { "INTERPOLATE_CONSTANT", "INTERPOLATE_LINEAR", - "INTERPOLATE_PERSPECTIVE", - "INTERPOLATE_ATTRIB" + "INTERPOLATE_PERSPECTIVE" }; static const char *TGSI_SEMANTICS[] = @@ -250,9 +243,9 @@ static const char *TGSI_INSTRUCTION_EXTS[] = static const char *TGSI_PRECISIONS[] = { "PRECISION_DEFAULT", - "TGSI_PRECISION_FLOAT32", - "TGSI_PRECISION_FLOAT16", - "TGSI_PRECISION_FIXED12" + "PRECISION_FLOAT32", + "PRECISION_FLOAT16", + "PRECISION_FIXED12" }; static const char *TGSI_CCS[] = -- cgit v1.2.3 From 82f11f7e7c0bbe0452da65f08195c2a346f820e9 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 16:28:59 +0200 Subject: tgsi: Tidy up and fix tgsi_dump. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 215 ++++++++++++-------------- src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c | 2 +- 2 files changed, 101 insertions(+), 116 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 1056951bc8..fe8742dc02 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -34,16 +34,14 @@ static void dump_enum( - const unsigned e, - const char **enums, - const unsigned enums_count ) + uint e, + const char **enums, + uint enum_count ) { - if (e >= enums_count) { + if (e >= enum_count) debug_printf( "%u", e ); - } - else { + else debug_printf( "%s", enums[e] ); - } } #define EOL() debug_printf( "\n" ) @@ -55,14 +53,14 @@ dump_enum( #define FLT(F) debug_printf( "%10.4f", F ) #define ENM(E,ENUMS) dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) -static const char *TGSI_PROCESSOR_TYPES_SHORT[] = +static const char *processor_type_names[] = { "FRAG", "VERT", "GEOM" }; -static const char *TGSI_FILES_SHORT[] = +static const char *file_names[] = { "NULL", "CONST", @@ -74,14 +72,14 @@ static const char *TGSI_FILES_SHORT[] = "IMM" }; -static const char *TGSI_INTERPOLATES_SHORT[] = +static const char *interpolate_names[] = { "CONSTANT", "LINEAR", "PERSPECTIVE" }; -static const char *TGSI_SEMANTICS_SHORT[] = +static const char *semantic_names[] = { "POSITION", "COLOR", @@ -92,12 +90,12 @@ static const char *TGSI_SEMANTICS_SHORT[] = "NORMAL" }; -static const char *TGSI_IMMS_SHORT[] = +static const char *immediate_type_names[] = { "FLT32" }; -static const char *TGSI_OPCODES_SHORT[TGSI_OPCODE_LAST] = +static const char *opcode_names[TGSI_OPCODE_LAST] = { "ARL", "MOV", @@ -220,7 +218,7 @@ static const char *TGSI_OPCODES_SHORT[TGSI_OPCODE_LAST] = "SWZ" }; -static const char *TGSI_SWIZZLES_SHORT[] = +static const char *swizzle_names[] = { "x", "y", @@ -228,7 +226,7 @@ static const char *TGSI_SWIZZLES_SHORT[] = "w" }; -static const char *TGSI_TEXTURES_SHORT[] = +static const char *texture_names[] = { "UNKNOWN", "1D", @@ -241,7 +239,7 @@ static const char *TGSI_TEXTURES_SHORT[] = "SHADOWRECT" }; -static const char *TGSI_EXTSWIZZLES_SHORT[] = +static const char *extswizzle_names[] = { "x", "y", @@ -251,7 +249,7 @@ static const char *TGSI_EXTSWIZZLES_SHORT[] = "1" }; -static const char *TGSI_MODULATES_SHORT[TGSI_MODULATE_COUNT] = +static const char *modulate_names[TGSI_MODULATE_COUNT] = { "", "_2X", @@ -262,40 +260,55 @@ static const char *TGSI_MODULATES_SHORT[TGSI_MODULATE_COUNT] = "_D8" }; -void -tgsi_dump_declaration( - const struct tgsi_full_declaration *decl ) +static void +_dump_register( + uint file, + uint first, + uint last ) { - TXT( "\nDCL " ); - ENM( decl->Declaration.File, TGSI_FILES_SHORT ); - + ENM( file, file_names ); CHR( '[' ); - UID( decl->DeclarationRange.First ); - if (decl->DeclarationRange.First != decl->DeclarationRange.Last) { + UID( first ); + if (first != last) { TXT( ".." ); - UID( decl->DeclarationRange.Last ); + UID( last ); } CHR( ']' ); +} - if( decl->Declaration.UsageMask != TGSI_WRITEMASK_XYZW ) { +static void +_dump_writemask( + uint writemask ) +{ + if (writemask != TGSI_WRITEMASK_XYZW) { CHR( '.' ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + if (writemask & TGSI_WRITEMASK_X) CHR( 'x' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + if (writemask & TGSI_WRITEMASK_Y) CHR( 'y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + if (writemask & TGSI_WRITEMASK_Z) CHR( 'z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + if (writemask & TGSI_WRITEMASK_W) CHR( 'w' ); - } } +} + +void +tgsi_dump_declaration( + const struct tgsi_full_declaration *decl ) +{ + TXT( "\nDCL " ); + + _dump_register( + decl->Declaration.File, + decl->DeclarationRange.First, + decl->DeclarationRange.Last ); + _dump_writemask( + decl->Declaration.UsageMask ); if (decl->Declaration.Semantic) { TXT( ", " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS_SHORT ); + ENM( decl->Semantic.SemanticName, semantic_names ); if (decl->Semantic.SemanticIndex != 0 || decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC) { CHR( '[' ); @@ -305,32 +318,30 @@ tgsi_dump_declaration( } TXT( ", " ); - ENM( decl->Declaration.Interpolate, TGSI_INTERPOLATES_SHORT ); + ENM( decl->Declaration.Interpolate, interpolate_names ); } void tgsi_dump_immediate( const struct tgsi_full_immediate *imm ) { - unsigned i; + uint i; TXT( "\nIMM " ); - ENM( imm->Immediate.DataType, TGSI_IMMS_SHORT ); + ENM( imm->Immediate.DataType, immediate_type_names ); TXT( " { " ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - switch( imm->Immediate.DataType ) { + for (i = 0; i < imm->Immediate.Size - 1; i++) { + switch (imm->Immediate.DataType) { case TGSI_IMM_FLOAT32: FLT( imm->u.ImmediateFloat32[i].Float ); break; - default: assert( 0 ); } - if( i < imm->Immediate.Size - 2 ) { + if (i < imm->Immediate.Size - 2) TXT( ", " ); - } } TXT( " }" ); } @@ -340,15 +351,15 @@ tgsi_dump_instruction( const struct tgsi_full_instruction *inst, uint instno ) { - unsigned i; - boolean first_reg = TRUE; + uint i; + boolean first_reg = TRUE; EOL(); UID( instno ); CHR( ':' ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES_SHORT ); + ENM( inst->Instruction.Opcode, opcode_names ); - switch( inst->Instruction.Saturate ) { + switch (inst->Instruction.Saturate) { case TGSI_SAT_NONE: break; case TGSI_SAT_ZERO_ONE: @@ -361,47 +372,28 @@ tgsi_dump_instruction( assert( 0 ); } - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + for (i = 0; i < inst->Instruction.NumDstRegs; i++) { const struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - if( !first_reg ) { + if (!first_reg) CHR( ',' ); - } CHR( ' ' ); - ENM( dst->DstRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); - SID( dst->DstRegister.Index ); - CHR( ']' ); - - ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES_SHORT ); - - if( dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW ) { - CHR( '.' ); - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_X ) { - CHR( 'x' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Y ) { - CHR( 'y' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_Z ) { - CHR( 'z' ); - } - if( dst->DstRegister.WriteMask & TGSI_WRITEMASK_W ) { - CHR( 'w' ); - } - } + _dump_register( + dst->DstRegister.File, + dst->DstRegister.Index, + dst->DstRegister.Index ); + ENM( dst->DstRegisterExtModulate.Modulate, modulate_names ); + _dump_writemask( dst->DstRegister.WriteMask ); first_reg = FALSE; } - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { const struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - if( !first_reg ) { + if (!first_reg) CHR( ',' ); - } CHR( ' ' ); if (src->SrcRegisterExtMod.Negate) @@ -417,40 +409,45 @@ tgsi_dump_instruction( if (src->SrcRegister.Negate) CHR( '-' ); - ENM( src->SrcRegister.File, TGSI_FILES_SHORT ); - - CHR( '[' ); if (src->SrcRegister.Indirect) { + /* TODO: Tidy up + */ + ENM( src->SrcRegister.File, file_names ); + CHR( '[' ); TXT( "ADDR[0]" ); if (src->SrcRegister.Index != 0) { if (src->SrcRegister.Index > 0) CHR( '+' ); SID( src->SrcRegister.Index ); } + CHR( ']' ); + } + else { + _dump_register( + src->SrcRegister.File, + src->SrcRegister.Index, + src->SrcRegister.Index ); } - else - SID( src->SrcRegister.Index ); - CHR( ']' ); if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) { CHR( '.' ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES_SHORT ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES_SHORT ); + ENM( src->SrcRegister.SwizzleX, swizzle_names ); + ENM( src->SrcRegister.SwizzleY, swizzle_names ); + ENM( src->SrcRegister.SwizzleZ, swizzle_names ); + ENM( src->SrcRegister.SwizzleW, swizzle_names ); } if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { CHR( '.' ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES_SHORT ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES_SHORT ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, extswizzle_names ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, extswizzle_names ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, extswizzle_names ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, extswizzle_names ); } if (src->SrcRegisterExtMod.Complement) @@ -469,10 +466,10 @@ tgsi_dump_instruction( if (inst->InstructionExtTexture.Texture != TGSI_TEXTURE_UNKNOWN) { TXT( ", " ); - ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES_SHORT ); + ENM( inst->InstructionExtTexture.Texture, texture_names ); } - switch( inst->Instruction.Opcode ) { + switch (inst->Instruction.Opcode) { case TGSI_OPCODE_IF: case TGSI_OPCODE_ELSE: case TGSI_OPCODE_BGNLOOP2: @@ -490,47 +487,35 @@ tgsi_dump( uint flags ) { struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - unsigned instno = 0; + uint instno = 0; /* sanity checks */ - assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); - assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "OPCODE_END") == 0); - assert(strcmp(TGSI_OPCODES_SHORT[TGSI_OPCODE_END], "END") == 0); + assert( strcmp( opcode_names[TGSI_OPCODE_CONT], "CONT" ) == 0 ); + assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); + assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); tgsi_parse_init( &parse, tokens ); EOL(); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES_SHORT ); + ENM( parse.FullHeader.Processor.Processor, processor_type_names ); UID( parse.FullVersion.Version.MajorVersion ); CHR( '.' ); UID( parse.FullVersion.Version.MinorVersion ); - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { + while (!tgsi_parse_end_of_tokens( &parse )) { tgsi_parse_token( &parse ); - switch( parse.FullToken.Token.Type ) { + switch (parse.FullToken.Token.Type) { case TGSI_TOKEN_TYPE_DECLARATION: - tgsi_dump_declaration( - &parse.FullToken.FullDeclaration ); + tgsi_dump_declaration( &parse.FullToken.FullDeclaration ); break; - case TGSI_TOKEN_TYPE_IMMEDIATE: - tgsi_dump_immediate( - &parse.FullToken.FullImmediate ); + tgsi_dump_immediate( &parse.FullToken.FullImmediate ); break; - case TGSI_TOKEN_TYPE_INSTRUCTION: - tgsi_dump_instruction( - &parse.FullToken.FullInstruction, - instno ); + tgsi_dump_instruction( &parse.FullToken.FullInstruction, instno ); instno++; break; - default: assert( 0 ); } diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c index acb1246b7a..e7d1079588 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c @@ -773,7 +773,7 @@ tgsi_dump_c( /* sanity checks */ assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "END") == 0); + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); tgsi_parse_init( &parse, tokens ); -- cgit v1.2.3 From fef7f2b070ab797f78ebc210bb40a24685c6d4b7 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 16:42:43 +0200 Subject: tgsi: Fix error handling. --- src/gallium/auxiliary/tgsi/util/tgsi_sanity.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c index 7fc4c29c68..a4edbb6d08 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c @@ -263,7 +263,7 @@ tgsi_sanity_check( ctx.errors = 0; ctx.warnings = 0; - if (tgsi_iterate_shader( tokens, &ctx.iter ) == -1) + if (!tgsi_iterate_shader( tokens, &ctx.iter )) return FALSE; return ctx.errors == 0; -- cgit v1.2.3 From 57482e1549e131fa1aebd442a677a95909b22508 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 16:42:57 +0200 Subject: tgsi: Keep version and processor info in iterate_ctx. --- src/gallium/auxiliary/tgsi/util/tgsi_iterate.c | 3 +++ src/gallium/auxiliary/tgsi/util/tgsi_iterate.h | 3 +++ 2 files changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c index 43d7a6b1d5..5371a88b96 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c @@ -38,6 +38,9 @@ tgsi_iterate_shader( if (tgsi_parse_init( &parse, tokens ) != TGSI_PARSE_OK) return FALSE; + ctx->processor = parse.FullHeader.Processor; + ctx->version = parse.FullVersion.Version; + if (ctx->prolog) if (!ctx->prolog( ctx )) goto fail; diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h index 18729c2d1a..f5bebf89b8 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h +++ b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h @@ -59,6 +59,9 @@ struct tgsi_iterate_context boolean (* epilog)( struct tgsi_iterate_context *ctx ); + + struct tgsi_processor processor; + struct tgsi_version version; }; boolean -- cgit v1.2.3 From 83f245bd242cd2c5f59f072095dcc47aa6153b21 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 16:43:26 +0200 Subject: tgsi: Use tgsi_iterate in tgsi_dump. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 89 ++++++++++++++++++----------- 1 file changed, 57 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index fe8742dc02..08d0d5200c 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -26,11 +26,15 @@ **************************************************************************/ #include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "util/u_string.h" #include "tgsi_dump.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" +#include "tgsi_iterate.h" + +struct dump_ctx +{ + struct tgsi_iterate_context iter; + + uint instno; +}; static void dump_enum( @@ -321,6 +325,15 @@ tgsi_dump_declaration( ENM( decl->Declaration.Interpolate, interpolate_names ); } +static boolean +iter_declaration( + struct tgsi_iterate_context *iter, + struct tgsi_full_declaration *decl ) +{ + tgsi_dump_declaration( decl ); + return TRUE; +} + void tgsi_dump_immediate( const struct tgsi_full_immediate *imm ) @@ -346,6 +359,15 @@ tgsi_dump_immediate( TXT( " }" ); } +static boolean +iter_immediate( + struct tgsi_iterate_context *iter, + struct tgsi_full_immediate *imm ) +{ + tgsi_dump_immediate( imm ); + return TRUE; +} + void tgsi_dump_instruction( const struct tgsi_full_instruction *inst, @@ -481,45 +503,48 @@ tgsi_dump_instruction( } } +static boolean +iter_instruction( + struct tgsi_iterate_context *iter, + struct tgsi_full_instruction *inst ) +{ + struct dump_ctx *ctx = (struct dump_ctx *) iter; + + tgsi_dump_instruction( inst, ctx->instno++ ); + return TRUE; +} + +static boolean +prolog( + struct tgsi_iterate_context *ctx ) +{ + EOL(); + ENM( ctx->processor.Processor, processor_type_names ); + UID( ctx->version.MajorVersion ); + CHR( '.' ); + UID( ctx->version.MinorVersion ); + return TRUE; +} + void tgsi_dump( const struct tgsi_token *tokens, uint flags ) { - struct tgsi_parse_context parse; - uint instno = 0; + struct dump_ctx ctx; /* sanity checks */ assert( strcmp( opcode_names[TGSI_OPCODE_CONT], "CONT" ) == 0 ); assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); - tgsi_parse_init( &parse, tokens ); - - EOL(); - ENM( parse.FullHeader.Processor.Processor, processor_type_names ); - UID( parse.FullVersion.Version.MajorVersion ); - CHR( '.' ); - UID( parse.FullVersion.Version.MinorVersion ); - - while (!tgsi_parse_end_of_tokens( &parse )) { - tgsi_parse_token( &parse ); + ctx.iter.prolog = prolog; + ctx.iter.iterate_instruction = iter_instruction; + ctx.iter.iterate_declaration = iter_declaration; + ctx.iter.iterate_immediate = iter_immediate; + ctx.iter.epilog = NULL; - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_DECLARATION: - tgsi_dump_declaration( &parse.FullToken.FullDeclaration ); - break; - case TGSI_TOKEN_TYPE_IMMEDIATE: - tgsi_dump_immediate( &parse.FullToken.FullImmediate ); - break; - case TGSI_TOKEN_TYPE_INSTRUCTION: - tgsi_dump_instruction( &parse.FullToken.FullInstruction, instno ); - instno++; - break; - default: - assert( 0 ); - } - } + ctx.instno = 0; - tgsi_parse_free( &parse ); + tgsi_iterate_shader( tokens, &ctx.iter ); } -- cgit v1.2.3 From 613f0df64dd2c2db71dd73b595225016ae596576 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 19:29:52 +0200 Subject: tgsi: Remove redundant code. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 1 - src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c | 1 - 2 files changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 08d0d5200c..a9d500c8cf 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -536,7 +536,6 @@ tgsi_dump( /* sanity checks */ assert( strcmp( opcode_names[TGSI_OPCODE_CONT], "CONT" ) == 0 ); assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); - assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); ctx.iter.prolog = prolog; ctx.iter.iterate_instruction = iter_instruction; diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c index e7d1079588..eabd74bd6d 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c @@ -773,7 +773,6 @@ tgsi_dump_c( /* sanity checks */ assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); tgsi_parse_init( &parse, tokens ); -- cgit v1.2.3 From 73e1d0be756537376495547bc1e798805884b8ef Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 21:23:04 +0200 Subject: tgsi: Add support for indirect addressing in dump, sanity and text modules. --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 53 +++++++--- src/gallium/auxiliary/tgsi/util/tgsi_sanity.c | 86 ++++++++++----- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 147 ++++++++++++++++++++------ 3 files changed, 215 insertions(+), 71 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index a9d500c8cf..94180f7e50 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -265,17 +265,48 @@ static const char *modulate_names[TGSI_MODULATE_COUNT] = }; static void -_dump_register( +_dump_register_prefix( uint file, uint first, uint last ) { + + +} + +static void +_dump_register( + uint file, + int first, + int last ) +{ ENM( file, file_names ); CHR( '[' ); - UID( first ); + SID( first ); if (first != last) { TXT( ".." ); - UID( last ); + SID( last ); + } + CHR( ']' ); +} + +static void +_dump_register_ind( + uint file, + int index, + uint ind_file, + int ind_index ) +{ + ENM( file, file_names ); + CHR( '[' ); + ENM( ind_file, file_names ); + CHR( '[' ); + SID( ind_index ); + CHR( ']' ); + if (index != 0) { + if (index > 0) + CHR( '+' ); + SID( index ); } CHR( ']' ); } @@ -432,17 +463,11 @@ tgsi_dump_instruction( CHR( '-' ); if (src->SrcRegister.Indirect) { - /* TODO: Tidy up - */ - ENM( src->SrcRegister.File, file_names ); - CHR( '[' ); - TXT( "ADDR[0]" ); - if (src->SrcRegister.Index != 0) { - if (src->SrcRegister.Index > 0) - CHR( '+' ); - SID( src->SrcRegister.Index ); - } - CHR( ']' ); + _dump_register_ind( + src->SrcRegister.File, + src->SrcRegister.Index, + src->SrcRegisterInd.File, + src->SrcRegisterInd.Index ); } else { _dump_register( diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c index a4edbb6d08..f11de815b0 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c @@ -39,8 +39,9 @@ struct sanity_check_ctx { struct tgsi_iterate_context iter; - reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REGISTERS / sizeof( uint ) / 8]; - reg_flag regs_used[TGSI_FILE_COUNT][MAX_REGISTERS / sizeof( uint ) / 8]; + reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG]; + reg_flag regs_used[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG]; + boolean regs_ind_used[TGSI_FILE_COUNT]; uint num_imms; uint num_instructions; uint index_of_END; @@ -83,24 +84,59 @@ static boolean is_register_declared( struct sanity_check_ctx *ctx, uint file, - uint index ) + int index ) { - assert( index < MAX_REGISTERS ); + assert( index >= 0 && index < MAX_REGISTERS ); return (ctx->regs_decl[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; } +static boolean +is_any_register_declared( + struct sanity_check_ctx *ctx, + uint file ) +{ + uint i; + + for (i = 0; i < MAX_REGISTERS / BITS_IN_REG_FLAG; i++) + if (ctx->regs_decl[file][i]) + return TRUE; + return FALSE; +} + static boolean is_register_used( struct sanity_check_ctx *ctx, uint file, - uint index ) + int index ) { assert( index < MAX_REGISTERS ); return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; } +static boolean +check_register_usage( + struct sanity_check_ctx *ctx, + uint file, + int index, + boolean indirect ) +{ + if (!check_file_name( ctx, file )) + return FALSE; + if (indirect) { + if (!is_any_register_declared( ctx, file )) + report_error( ctx, "Undeclared source register" ); + ctx->regs_ind_used[file] = TRUE; + } + else { + if (!is_register_declared( ctx, file, index )) + report_error( ctx, "Undeclared destination register" ); + ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG)); + } + return TRUE; +} + static boolean iter_instruction( struct tgsi_iterate_context *iter, @@ -122,28 +158,25 @@ iter_instruction( * Mark the registers as used. */ for (i = 0; i < inst->Instruction.NumDstRegs; i++) { - uint file; - uint index; - - file = inst->FullDstRegisters[i].DstRegister.File; - if (!check_file_name( ctx, file )) - return TRUE; - index = inst->FullDstRegisters[i].DstRegister.Index; - if (!is_register_declared( ctx, file, index )) - report_error( ctx, "Undeclared destination register" ); - ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG)); + check_register_usage( + ctx, + inst->FullDstRegisters[i].DstRegister.File, + inst->FullDstRegisters[i].DstRegister.Index, + FALSE ); } for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { - uint file; - uint index; - - file = inst->FullSrcRegisters[i].SrcRegister.File; - if (!check_file_name( ctx, file )) - return TRUE; - index = inst->FullSrcRegisters[i].SrcRegister.Index; - if (!is_register_declared( ctx, file, index )) - report_error( ctx, "Undeclared source register" ); - ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG)); + check_register_usage( + ctx, + inst->FullSrcRegisters[i].SrcRegister.File, + inst->FullSrcRegisters[i].SrcRegister.Index, + inst->FullSrcRegisters[i].SrcRegister.Indirect ); + if (inst->FullSrcRegisters[i].SrcRegister.Indirect) { + check_register_usage( + ctx, + inst->FullSrcRegisters[i].SrcRegisterInd.File, + inst->FullSrcRegisters[i].SrcRegisterInd.Index, + FALSE ); + } } ctx->num_instructions++; @@ -228,7 +261,7 @@ epilog( uint i; for (i = 0; i < MAX_REGISTERS; i++) { - if (is_register_declared( ctx, file, i ) && !is_register_used( ctx, file, i )) { + if (is_register_declared( ctx, file, i ) && !is_register_used( ctx, file, i ) && !ctx->regs_ind_used[file]) { report_warning( ctx, "Register never used" ); } } @@ -256,6 +289,7 @@ tgsi_sanity_check( memset( ctx.regs_decl, 0, sizeof( ctx.regs_decl ) ); memset( ctx.regs_used, 0, sizeof( ctx.regs_used ) ); + memset( ctx.regs_ind_used, 0, sizeof( ctx.regs_ind_used ) ); ctx.num_imms = 0; ctx.num_instructions = 0; ctx.index_of_END = ~0; diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 9ed0f52fc7..2a50315463 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -226,24 +226,21 @@ static const char *file_names[TGSI_FILE_COUNT] = }; static boolean -parse_file( - struct translate_ctx *ctx, - uint *file ) +parse_file( const char **pcur, uint *file ) { uint i; for (i = 0; i < TGSI_FILE_COUNT; i++) { - const char *cur = ctx->cur; + const char *cur = *pcur; if (str_match_no_case( &cur, file_names[i] )) { if (!is_digit_alpha_underscore( cur )) { - ctx->cur = cur; + *pcur = cur; *file = i; return TRUE; } } } - report_error( ctx, "Unknown register file" ); return FALSE; } @@ -290,41 +287,57 @@ parse_opt_writemask( return TRUE; } -/* Parse register part common for decls and operands. - * ::= `[' +/* ::= `[' */ static boolean -parse_register_prefix( +parse_register_file_bracket( struct translate_ctx *ctx, - uint *file, - uint *index ) + uint *file ) { - if (!parse_file( ctx, file )) + if (!parse_file( &ctx->cur, file )) { + report_error( ctx, "Unknown register file" ); return FALSE; + } eat_opt_white( &ctx->cur ); if (*ctx->cur != '[') { report_error( ctx, "Expected `['" ); return FALSE; } ctx->cur++; + return TRUE; +} + +/* ::= + */ +static boolean +parse_register_file_bracket_index( + struct translate_ctx *ctx, + uint *file, + int *index ) +{ + uint uindex; + + if (!parse_register_file_bracket( ctx, file )) + return FALSE; eat_opt_white( &ctx->cur ); - if (!parse_uint( &ctx->cur, index )) { - report_error( ctx, "Expected literal integer" ); + if (!parse_uint( &ctx->cur, &uindex )) { + report_error( ctx, "Expected literal unsigned integer" ); return FALSE; } + *index = (int) uindex; return TRUE; } -/* Parse register operand. - * ::= `]' +/* Parse destination register operand. + * ::= `]' */ static boolean -parse_register( +parse_register_dst( struct translate_ctx *ctx, uint *file, - uint *index ) + int *index ) { - if (!parse_register_prefix( ctx, file, index )) + if (!parse_register_file_bracket_index( ctx, file, index )) return FALSE; eat_opt_white( &ctx->cur ); if (*ctx->cur != ']') { @@ -332,30 +345,95 @@ parse_register( return FALSE; } ctx->cur++; - /* TODO: Modulate suffix */ + return TRUE; +} + +/* Parse source register operand. + * ::= `]' | + * `]' | + * `+' `]' | + * `-' `]' + */ +static boolean +parse_register_src( + struct translate_ctx *ctx, + uint *file, + int *index, + uint *ind_file, + int *ind_index ) +{ + const char *cur; + uint uindex; + + if (!parse_register_file_bracket( ctx, file )) + return FALSE; + eat_opt_white( &ctx->cur ); + cur = ctx->cur; + if (parse_file( &cur, ind_file )) { + if (!parse_register_dst( ctx, ind_file, ind_index )) + return FALSE; + eat_opt_white( &ctx->cur ); + if (*ctx->cur == '+' || *ctx->cur == '-') { + boolean negate; + + negate = *ctx->cur == '-'; + ctx->cur++; + eat_opt_white( &ctx->cur ); + if (!parse_uint( &ctx->cur, &uindex )) { + report_error( ctx, "Expected literal unsigned integer" ); + return FALSE; + } + if (negate) + *index = -(int) uindex; + else + *index = (int) uindex; + } + else { + *index = 0; + } + } + else { + if (!parse_uint( &ctx->cur, &uindex )) { + report_error( ctx, "Expected literal unsigned integer" ); + return FALSE; + } + *index = (int) uindex; + *ind_file = TGSI_FILE_NULL; + *ind_index = 0; + } + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ']') { + report_error( ctx, "Expected `]'" ); + return FALSE; + } + ctx->cur++; return TRUE; } /* Parse register declaration. - * ::= `]' | `..' `]' + * ::= `]' | + * `..' `]' */ static boolean parse_register_dcl( struct translate_ctx *ctx, uint *file, - uint *first, - uint *last ) + int *first, + int *last ) { - if (!parse_register_prefix( ctx, file, first )) + if (!parse_register_file_bracket_index( ctx, file, first )) return FALSE; eat_opt_white( &ctx->cur ); if (ctx->cur[0] == '.' && ctx->cur[1] == '.') { + uint uindex; + ctx->cur += 2; eat_opt_white( &ctx->cur ); - if (!parse_uint( &ctx->cur, last )) { + if (!parse_uint( &ctx->cur, &uindex )) { report_error( ctx, "Expected literal integer" ); return FALSE; } + *last = (int) uindex; eat_opt_white( &ctx->cur ); } else { @@ -386,11 +464,11 @@ parse_dst_operand( struct tgsi_full_dst_register *dst ) { uint file; - uint index; + int index; uint writemask; const char *cur; - if (!parse_register( ctx, &file, &index )) + if (!parse_register_dst( ctx, &file, &index )) return FALSE; cur = ctx->cur; @@ -426,7 +504,9 @@ parse_src_operand( const char *cur; float value; uint file; - uint index; + int index; + uint ind_file; + int ind_index; if (*ctx->cur == '-') { cur = ctx->cur; @@ -498,10 +578,15 @@ parse_src_operand( } } - if (!parse_register( ctx, &file, &index )) + if (!parse_register_src( ctx, &file, &index, &ind_file, &ind_index )) return FALSE; src->SrcRegister.File = file; src->SrcRegister.Index = index; + if (ind_file != TGSI_FILE_NULL) { + src->SrcRegister.Indirect = 1; + src->SrcRegisterInd.File = ind_file; + src->SrcRegisterInd.Index = ind_index; + } /* Parse optional swizzle */ @@ -864,8 +949,8 @@ static boolean parse_declaration( struct translate_ctx *ctx ) { struct tgsi_full_declaration decl; uint file; - uint first; - uint last; + int first; + int last; uint writemask; const char *cur; uint advance; -- cgit v1.2.3 From 25a7f422b4e307dce966220d47794fb056d04aac Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 21:28:28 +0200 Subject: tgsi: Warn if an indirect register not ADDR[0]. --- src/gallium/auxiliary/tgsi/util/tgsi_sanity.c | 13 ++++++++----- 1 file changed, 8 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c index f11de815b0..08f3995b13 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c @@ -171,11 +171,14 @@ iter_instruction( inst->FullSrcRegisters[i].SrcRegister.Index, inst->FullSrcRegisters[i].SrcRegister.Indirect ); if (inst->FullSrcRegisters[i].SrcRegister.Indirect) { - check_register_usage( - ctx, - inst->FullSrcRegisters[i].SrcRegisterInd.File, - inst->FullSrcRegisters[i].SrcRegisterInd.Index, - FALSE ); + uint file; + int index; + + file = inst->FullSrcRegisters[i].SrcRegisterInd.File; + index = inst->FullSrcRegisters[i].SrcRegisterInd.Index; + check_register_usage( ctx, file, index, FALSE ); + if (file != TGSI_FILE_ADDRESS || index != 0) + report_warning( ctx, "Indirect register not ADDR[0]" ); } } -- cgit v1.2.3 From 9d068d4b90813d5afa243d669e8437b8922ac656 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 21:38:20 +0200 Subject: tgsi: Add support for branch instructions with target labels. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 256 +++++++++++++++------------- 1 file changed, 137 insertions(+), 119 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 2a50315463..442016104b 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -687,130 +687,131 @@ struct opcode_info uint num_dst; uint num_src; uint is_tex; + uint is_branch; const char *mnemonic; }; static const struct opcode_info opcode_info[TGSI_OPCODE_LAST] = { - { 1, 1, 0, "ARL" }, - { 1, 1, 0, "MOV" }, - { 1, 1, 0, "LIT" }, - { 1, 1, 0, "RCP" }, - { 1, 1, 0, "RSQ" }, - { 1, 1, 0, "EXP" }, - { 1, 1, 0, "LOG" }, - { 1, 2, 0, "MUL" }, - { 1, 2, 0, "ADD" }, - { 1, 2, 0, "DP3" }, - { 1, 2, 0, "DP4" }, - { 1, 2, 0, "DST" }, - { 1, 2, 0, "MIN" }, - { 1, 2, 0, "MAX" }, - { 1, 2, 0, "SLT" }, - { 1, 2, 0, "SGE" }, - { 1, 3, 0, "MAD" }, - { 1, 2, 0, "SUB" }, - { 1, 3, 0, "LERP" }, - { 1, 3, 0, "CND" }, - { 1, 3, 0, "CND0" }, - { 1, 3, 0, "DOT2ADD" }, - { 1, 2, 0, "INDEX" }, - { 1, 1, 0, "NEGATE" }, - { 1, 1, 0, "FRAC" }, - { 1, 3, 0, "CLAMP" }, - { 1, 1, 0, "FLOOR" }, - { 1, 1, 0, "ROUND" }, - { 1, 1, 0, "EXPBASE2" }, - { 1, 1, 0, "LOGBASE2" }, - { 1, 2, 0, "POWER" }, - { 1, 2, 0, "CROSSPRODUCT" }, - { 1, 2, 0, "MULTIPLYMATRIX" }, - { 1, 1, 0, "ABS" }, - { 1, 1, 0, "RCC" }, - { 1, 2, 0, "DPH" }, - { 1, 1, 0, "COS" }, - { 1, 1, 0, "DDX" }, - { 1, 1, 0, "DDY" }, - { 0, 1, 0, "KILP" }, - { 1, 1, 0, "PK2H" }, - { 1, 1, 0, "PK2US" }, - { 1, 1, 0, "PK4B" }, - { 1, 1, 0, "PK4UB" }, - { 1, 2, 0, "RFL" }, - { 1, 2, 0, "SEQ" }, - { 1, 2, 0, "SFL" }, - { 1, 2, 0, "SGT" }, - { 1, 1, 0, "SIN" }, - { 1, 2, 0, "SLE" }, - { 1, 2, 0, "SNE" }, - { 1, 2, 0, "STR" }, - { 1, 2, 1, "TEX" }, - { 1, 4, 1, "TXD" }, - { 1, 2, 1, "TXP" }, - { 1, 1, 0, "UP2H" }, - { 1, 1, 0, "UP2US" }, - { 1, 1, 0, "UP4B" }, - { 1, 1, 0, "UP4UB" }, - { 1, 3, 0, "X2D" }, - { 1, 1, 0, "ARA" }, - { 1, 1, 0, "ARR" }, - { 0, 1, 0, "BRA" }, - { 0, 0, 0, "CAL" }, - { 0, 0, 0, "RET" }, - { 1, 1, 0, "SSG" }, - { 1, 3, 0, "CMP" }, - { 1, 1, 0, "SCS" }, - { 1, 2, 1, "TXB" }, - { 1, 1, 0, "NRM" }, - { 1, 2, 0, "DIV" }, - { 1, 2, 0, "DP2" }, - { 1, 2, 1, "TXL" }, - { 0, 0, 0, "BRK" }, - { 0, 1, 0, "IF" }, - { 0, 0, 0, "LOOP" }, - { 0, 1, 0, "REP" }, - { 0, 0, 0, "ELSE" }, - { 0, 0, 0, "ENDIF" }, - { 0, 0, 0, "ENDLOOP" }, - { 0, 0, 0, "ENDREP" }, - { 0, 1, 0, "PUSHA" }, - { 1, 0, 0, "POPA" }, - { 1, 1, 0, "CEIL" }, - { 1, 1, 0, "I2F" }, - { 1, 1, 0, "NOT" }, - { 1, 1, 0, "TRUNC" }, - { 1, 2, 0, "SHL" }, - { 1, 2, 0, "SHR" }, - { 1, 2, 0, "AND" }, - { 1, 2, 0, "OR" }, - { 1, 2, 0, "MOD" }, - { 1, 2, 0, "XOR" }, - { 1, 3, 0, "SAD" }, - { 1, 2, 1, "TXF" }, - { 1, 2, 1, "TXQ" }, - { 0, 0, 0, "CONT" }, - { 0, 0, 0, "EMIT" }, - { 0, 0, 0, "ENDPRIM" }, - { 0, 0, 0, "BGNLOOP2" }, - { 0, 0, 0, "BGNSUB" }, - { 0, 0, 0, "ENDLOOP2" }, - { 0, 0, 0, "ENDSUB" }, - { 1, 1, 0, "NOISE1" }, - { 1, 1, 0, "NOISE2" }, - { 1, 1, 0, "NOISE3" }, - { 1, 1, 0, "NOISE4" }, - { 0, 0, 0, "NOP" }, - { 1, 2, 0, "M4X3" }, - { 1, 2, 0, "M3X4" }, - { 1, 2, 0, "M3X3" }, - { 1, 2, 0, "M3X2" }, - { 1, 1, 0, "NRM4" }, - { 0, 1, 0, "CALLNZ" }, - { 0, 1, 0, "IFC" }, - { 0, 1, 0, "BREAKC" }, - { 0, 0, 0, "KIL" }, - { 0, 0, 0, "END" }, - { 1, 1, 0, "SWZ" } + { 1, 1, 0, 0, "ARL" }, + { 1, 1, 0, 0, "MOV" }, + { 1, 1, 0, 0, "LIT" }, + { 1, 1, 0, 0, "RCP" }, + { 1, 1, 0, 0, "RSQ" }, + { 1, 1, 0, 0, "EXP" }, + { 1, 1, 0, 0, "LOG" }, + { 1, 2, 0, 0, "MUL" }, + { 1, 2, 0, 0, "ADD" }, + { 1, 2, 0, 0, "DP3" }, + { 1, 2, 0, 0, "DP4" }, + { 1, 2, 0, 0, "DST" }, + { 1, 2, 0, 0, "MIN" }, + { 1, 2, 0, 0, "MAX" }, + { 1, 2, 0, 0, "SLT" }, + { 1, 2, 0, 0, "SGE" }, + { 1, 3, 0, 0, "MAD" }, + { 1, 2, 0, 0, "SUB" }, + { 1, 3, 0, 0, "LERP" }, + { 1, 3, 0, 0, "CND" }, + { 1, 3, 0, 0, "CND0" }, + { 1, 3, 0, 0, "DOT2ADD" }, + { 1, 2, 0, 0, "INDEX" }, + { 1, 1, 0, 0, "NEGATE" }, + { 1, 1, 0, 0, "FRAC" }, + { 1, 3, 0, 0, "CLAMP" }, + { 1, 1, 0, 0, "FLOOR" }, + { 1, 1, 0, 0, "ROUND" }, + { 1, 1, 0, 0, "EXPBASE2" }, + { 1, 1, 0, 0, "LOGBASE2" }, + { 1, 2, 0, 0, "POWER" }, + { 1, 2, 0, 0, "CROSSPRODUCT" }, + { 1, 2, 0, 0, "MULTIPLYMATRIX" }, + { 1, 1, 0, 0, "ABS" }, + { 1, 1, 0, 0, "RCC" }, + { 1, 2, 0, 0, "DPH" }, + { 1, 1, 0, 0, "COS" }, + { 1, 1, 0, 0, "DDX" }, + { 1, 1, 0, 0, "DDY" }, + { 0, 1, 0, 0, "KILP" }, + { 1, 1, 0, 0, "PK2H" }, + { 1, 1, 0, 0, "PK2US" }, + { 1, 1, 0, 0, "PK4B" }, + { 1, 1, 0, 0, "PK4UB" }, + { 1, 2, 0, 0, "RFL" }, + { 1, 2, 0, 0, "SEQ" }, + { 1, 2, 0, 0, "SFL" }, + { 1, 2, 0, 0, "SGT" }, + { 1, 1, 0, 0, "SIN" }, + { 1, 2, 0, 0, "SLE" }, + { 1, 2, 0, 0, "SNE" }, + { 1, 2, 0, 0, "STR" }, + { 1, 2, 1, 0, "TEX" }, + { 1, 4, 1, 0, "TXD" }, + { 1, 2, 1, 0, "TXP" }, + { 1, 1, 0, 0, "UP2H" }, + { 1, 1, 0, 0, "UP2US" }, + { 1, 1, 0, 0, "UP4B" }, + { 1, 1, 0, 0, "UP4UB" }, + { 1, 3, 0, 0, "X2D" }, + { 1, 1, 0, 0, "ARA" }, + { 1, 1, 0, 0, "ARR" }, + { 0, 1, 0, 0, "BRA" }, + { 0, 0, 0, 1, "CAL" }, + { 0, 0, 0, 0, "RET" }, + { 1, 1, 0, 0, "SSG" }, + { 1, 3, 0, 0, "CMP" }, + { 1, 1, 0, 0, "SCS" }, + { 1, 2, 1, 0, "TXB" }, + { 1, 1, 0, 0, "NRM" }, + { 1, 2, 0, 0, "DIV" }, + { 1, 2, 0, 0, "DP2" }, + { 1, 2, 1, 0, "TXL" }, + { 0, 0, 0, 0, "BRK" }, + { 0, 1, 0, 1, "IF" }, + { 0, 0, 0, 0, "LOOP" }, + { 0, 1, 0, 0, "REP" }, + { 0, 0, 0, 1, "ELSE" }, + { 0, 0, 0, 0, "ENDIF" }, + { 0, 0, 0, 0, "ENDLOOP" }, + { 0, 0, 0, 0, "ENDREP" }, + { 0, 1, 0, 0, "PUSHA" }, + { 1, 0, 0, 0, "POPA" }, + { 1, 1, 0, 0, "CEIL" }, + { 1, 1, 0, 0, "I2F" }, + { 1, 1, 0, 0, "NOT" }, + { 1, 1, 0, 0, "TRUNC" }, + { 1, 2, 0, 0, "SHL" }, + { 1, 2, 0, 0, "SHR" }, + { 1, 2, 0, 0, "AND" }, + { 1, 2, 0, 0, "OR" }, + { 1, 2, 0, 0, "MOD" }, + { 1, 2, 0, 0, "XOR" }, + { 1, 3, 0, 0, "SAD" }, + { 1, 2, 1, 0, "TXF" }, + { 1, 2, 1, 0, "TXQ" }, + { 0, 0, 0, 0, "CONT" }, + { 0, 0, 0, 0, "EMIT" }, + { 0, 0, 0, 0, "ENDPRIM" }, + { 0, 0, 0, 1, "BGNLOOP2" }, + { 0, 0, 0, 0, "BGNSUB" }, + { 0, 0, 0, 1, "ENDLOOP2" }, + { 0, 0, 0, 0, "ENDSUB" }, + { 1, 1, 0, 0, "NOISE1" }, + { 1, 1, 0, 0, "NOISE2" }, + { 1, 1, 0, 0, "NOISE3" }, + { 1, 1, 0, 0, "NOISE4" }, + { 0, 0, 0, 0, "NOP" }, + { 1, 2, 0, 0, "M4X3" }, + { 1, 2, 0, 0, "M3X4" }, + { 1, 2, 0, 0, "M3X3" }, + { 1, 2, 0, 0, "M3X2" }, + { 1, 1, 0, 0, "NRM4" }, + { 0, 1, 0, 0, "CALLNZ" }, + { 0, 1, 0, 0, "IFC" }, + { 0, 1, 0, 0, "BREAKC" }, + { 0, 0, 0, 0, "KIL" }, + { 0, 0, 0, 0, "END" }, + { 1, 1, 0, 0, "SWZ" } }; static const char *texture_names[TGSI_TEXTURE_COUNT] = @@ -915,6 +916,23 @@ parse_instruction( } } + if (info->is_branch) { + uint target; + + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ':') { + report_error( ctx, "Expected `:'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + if (!parse_uint( &ctx->cur, &target )) { + report_error( ctx, "Expected a label" ); + return FALSE; + } + inst.InstructionExtLabel.Label = target; + } + advance = tgsi_build_full_instruction( &inst, ctx->tokens_cur, -- cgit v1.2.3 From b3c8d0c348c5d894e4df0314f060361893bdd38e Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sun, 20 Jul 2008 21:58:07 +0200 Subject: tgsi: Parse source register extended swizzle. --- src/gallium/auxiliary/tgsi/util/tgsi_text.c | 91 ++++++++++++++++++++--------- 1 file changed, 64 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c index 442016104b..35cb3055bb 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_text.c @@ -496,6 +496,52 @@ parse_dst_operand( return TRUE; } +static boolean +parse_optional_swizzle( + struct translate_ctx *ctx, + uint swizzle[4], + boolean *parsed_swizzle, + boolean *parsed_extswizzle ) +{ + const char *cur = ctx->cur; + + *parsed_swizzle = FALSE; + *parsed_extswizzle = FALSE; + + eat_opt_white( &cur ); + if (*cur == '.') { + uint i; + + cur++; + eat_opt_white( &cur ); + for (i = 0; i < 4; i++) { + if (toupper( *cur ) == 'X') + swizzle[i] = TGSI_SWIZZLE_X; + else if (toupper( *cur ) == 'Y') + swizzle[i] = TGSI_SWIZZLE_Y; + else if (toupper( *cur ) == 'Z') + swizzle[i] = TGSI_SWIZZLE_Z; + else if (toupper( *cur ) == 'W') + swizzle[i] = TGSI_SWIZZLE_W; + else { + if (*cur == '0') + swizzle[i] = TGSI_EXTSWIZZLE_ZERO; + else if (*cur == '1') + swizzle[i] = TGSI_EXTSWIZZLE_ONE; + else { + report_error( ctx, "Expected register swizzle component `x', `y', `z', `w', `0' or `1'" ); + return FALSE; + } + *parsed_extswizzle = TRUE; + } + cur++; + } + *parsed_swizzle = TRUE; + ctx->cur = cur; + } + return TRUE; +} + static boolean parse_src_operand( struct translate_ctx *ctx, @@ -507,6 +553,9 @@ parse_src_operand( int index; uint ind_file; int ind_index; + uint swizzle[4]; + boolean parsed_swizzle; + boolean parsed_extswizzle; if (*ctx->cur == '-') { cur = ctx->cur; @@ -588,35 +637,23 @@ parse_src_operand( src->SrcRegisterInd.Index = ind_index; } - /* Parse optional swizzle + /* Parse optional swizzle. */ - cur = ctx->cur; - eat_opt_white( &cur ); - if (*cur == '.') { - uint i; - - cur++; - eat_opt_white( &cur ); - for (i = 0; i < 4; i++) { - uint swizzle; - - if (toupper( *cur ) == 'X') - swizzle = TGSI_SWIZZLE_X; - else if (toupper( *cur ) == 'Y') - swizzle = TGSI_SWIZZLE_Y; - else if (toupper( *cur ) == 'Z') - swizzle = TGSI_SWIZZLE_Z; - else if (toupper( *cur ) == 'W') - swizzle = TGSI_SWIZZLE_W; - else { - report_error( ctx, "Expected register swizzle component either `x', `y', `z' or `w'" ); - return FALSE; - } - cur++; - tgsi_util_set_src_register_swizzle( &src->SrcRegister, swizzle, i ); + if (parse_optional_swizzle( ctx, swizzle, &parsed_swizzle, &parsed_extswizzle )) { + if (parsed_extswizzle) { + assert( parsed_swizzle ); + + src->SrcRegisterExtSwz.ExtSwizzleX = swizzle[0]; + src->SrcRegisterExtSwz.ExtSwizzleY = swizzle[1]; + src->SrcRegisterExtSwz.ExtSwizzleZ = swizzle[2]; + src->SrcRegisterExtSwz.ExtSwizzleW = swizzle[3]; + } + else if (parsed_swizzle) { + src->SrcRegister.SwizzleX = swizzle[0]; + src->SrcRegister.SwizzleY = swizzle[1]; + src->SrcRegister.SwizzleZ = swizzle[2]; + src->SrcRegister.SwizzleW = swizzle[3]; } - - ctx->cur = cur; } if (src->SrcRegisterExtMod.Complement) { -- cgit v1.2.3 From 1753ff9872562bc05aff2472fc1256539085bbfc Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Mon, 21 Jul 2008 14:38:57 +0200 Subject: tgsi: Update Makefile to include tgsi_iterate.c --- src/gallium/auxiliary/tgsi/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index 5555639b70..9c4b967651 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -6,6 +6,7 @@ LIBNAME = tgsi C_SOURCES = \ exec/tgsi_exec.c \ exec/tgsi_sse2.c \ + util/tgsi_iterate.c \ util/tgsi_build.c \ util/tgsi_dump.c \ util/tgsi_parse.c \ -- cgit v1.2.3 From d8cbac5330cfcdf66c9448ee339975869097a262 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 21 Jul 2008 10:25:02 -0600 Subject: mesa: remove debug code --- src/mesa/shader/slang/slang_codegen.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index d4d2e34dee..b8e130555d 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -3131,7 +3131,6 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, totalSize, datatype); store = _slang_new_ir_storage_swz(PROGRAM_UNIFORM, uniformLoc, totalSize, swizzle); - printf("GLOBAL USER UNIFORM %s size %d\n", varName, totalSize); } } else { -- cgit v1.2.3 From fbf26e109b1e96aa4eeca018198ff0b5a71e557b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 21 Jul 2008 13:58:50 -0600 Subject: mesa: assorted glsl uniform/attribute fixes Fix incorrect uniform/attribute size query results. Add missing error checking for glUniform, glUniformMatrix params Fix an array size/allocation error. --- src/mesa/shader/prog_parameter.c | 29 +------ src/mesa/shader/prog_parameter.h | 2 +- src/mesa/shader/shader_api.c | 152 +++++++++++++++++++++++++++++++--- src/mesa/shader/slang/slang_codegen.c | 21 ++++- src/mesa/shader/slang/slang_link.c | 5 +- 5 files changed, 166 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index be696385aa..d87e8f6e15 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -62,29 +62,6 @@ _mesa_free_parameter_list(struct gl_program_parameter_list *paramList) } -static GLint -_mesa_fit_type_in_vec4(GLenum type) -{ - switch (type) { - case GL_FLOAT: - case GL_INT: - return 4; - break; - case GL_FLOAT_VEC2: - case GL_INT_VEC2: - return 2; - break; - case GL_FLOAT_VEC3: - case GL_INT_VEC3: - return 1; - break; - case GL_FLOAT_VEC4: - case GL_INT_VEC4: - default: - return 1; - } -} - /** * Add a new parameter to a parameter list. * Note that parameter values are usually 4-element GLfloat vectors. @@ -294,7 +271,7 @@ _mesa_add_uniform(struct gl_program_parameter_list *paramList, } else { i = _mesa_add_parameter(paramList, PROGRAM_UNIFORM, name, - size * _mesa_fit_type_in_vec4(datatype), datatype, NULL, NULL); + size, datatype, NULL, NULL); return i; } } @@ -363,7 +340,7 @@ _mesa_add_varying(struct gl_program_parameter_list *paramList, */ GLint _mesa_add_attribute(struct gl_program_parameter_list *paramList, - const char *name, GLint size, GLint attrib) + const char *name, GLint size, GLenum datatype, GLint attrib) { GLint i = _mesa_lookup_parameter_index(paramList, -1, name); if (i >= 0) { @@ -379,7 +356,7 @@ _mesa_add_attribute(struct gl_program_parameter_list *paramList, if (size < 0) size = 4; i = _mesa_add_parameter(paramList, PROGRAM_INPUT, name, - size, GL_NONE, NULL, state); + size, datatype, NULL, state); } return i; } diff --git a/src/mesa/shader/prog_parameter.h b/src/mesa/shader/prog_parameter.h index 105f6f24de..dfb8c39ca4 100644 --- a/src/mesa/shader/prog_parameter.h +++ b/src/mesa/shader/prog_parameter.h @@ -122,7 +122,7 @@ _mesa_add_varying(struct gl_program_parameter_list *paramList, extern GLint _mesa_add_attribute(struct gl_program_parameter_list *paramList, - const char *name, GLint size, GLint attrib); + const char *name, GLint size, GLenum datatype, GLint attrib); extern GLint _mesa_add_state_reference(struct gl_program_parameter_list *paramList, diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index ee0e369d7c..e25cb63216 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -513,6 +513,7 @@ _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, struct gl_shader_program *shProg; const GLint size = -1; /* unknown size */ GLint i, oldIndex; + GLenum datatype; shProg = _mesa_lookup_shader_program_err(ctx, program, "glBindAttribLocation"); @@ -537,13 +538,14 @@ _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, if (shProg->LinkStatus) { /* get current index/location for the attribute */ oldIndex = _mesa_get_attrib_location(ctx, program, name); + assert(0); } else { oldIndex = -1; } /* this will replace the current value if it's already in the list */ - i = _mesa_add_attribute(shProg->Attributes, name, size, index); + i = _mesa_add_attribute(shProg->Attributes, name, size, datatype, index); if (i < 0) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBindAttribLocation"); } @@ -710,16 +712,51 @@ _mesa_detach_shader(GLcontext *ctx, GLuint program, GLuint shader) } +static GLint +sizeof_glsl_type(GLenum type) +{ + switch (type) { + case GL_FLOAT: + case GL_INT: + case GL_BOOL: + return 1; + case GL_FLOAT_VEC2: + case GL_INT_VEC2: + case GL_BOOL_VEC2: + return 2; + case GL_FLOAT_VEC3: + case GL_INT_VEC3: + case GL_BOOL_VEC3: + return 3; + case GL_FLOAT_VEC4: + case GL_INT_VEC4: + case GL_BOOL_VEC4: + return 4; + case GL_FLOAT_MAT2: + case GL_FLOAT_MAT2x3: + case GL_FLOAT_MAT2x4: + return 8; /* two float[4] vectors */ + case GL_FLOAT_MAT3: + case GL_FLOAT_MAT3x2: + case GL_FLOAT_MAT3x4: + return 12; /* three float[4] vectors */ + case GL_FLOAT_MAT4: + case GL_FLOAT_MAT4x2: + case GL_FLOAT_MAT4x3: + return 16; /* four float[4] vectors */ + default: + _mesa_problem(NULL, "Invalid type in sizeof_glsl_type()"); + return 1; + } +} + + static void _mesa_get_active_attrib(GLcontext *ctx, GLuint program, GLuint index, GLsizei maxLength, GLsizei *length, GLint *size, GLenum *type, GLchar *nameOut) { - static const GLenum vec_types[] = { - GL_FLOAT, GL_FLOAT_VEC2, GL_FLOAT_VEC3, GL_FLOAT_VEC4 - }; struct gl_shader_program *shProg; - GLint sz; shProg = _mesa_lookup_shader_program_err(ctx, program, "glGetActiveAttrib"); if (!shProg) @@ -732,11 +769,11 @@ _mesa_get_active_attrib(GLcontext *ctx, GLuint program, GLuint index, copy_string(nameOut, maxLength, length, shProg->Attributes->Parameters[index].Name); - sz = shProg->Attributes->Parameters[index].Size; if (size) - *size = sz; + *size = shProg->Attributes->Parameters[index].Size + / sizeof_glsl_type(shProg->Attributes->Parameters[index].DataType); if (type) - *type = vec_types[sz]; /* XXX this is a temporary hack */ + *type = shProg->Attributes->Parameters[index].DataType; } @@ -779,8 +816,8 @@ _mesa_get_active_uniform(GLcontext *ctx, GLuint program, GLuint index, copy_string(nameOut, maxLength, length, prog->Parameters->Parameters[progPos].Name); if (size) - *size = prog->Parameters->Parameters[progPos].Size; - + *size = prog->Parameters->Parameters[progPos].Size + / sizeof_glsl_type(prog->Parameters->Parameters[progPos].DataType); if (type) *type = prog->Parameters->Parameters[progPos].DataType; } @@ -1172,6 +1209,36 @@ update_textures_used(struct gl_program *prog) } +/** + * Check if the type given by userType is allowed to set a uniform of the + * target type. Generally, equivalence is required, but setting Boolean + * uniforms can be done with glUniformiv or glUniformfv. + */ +static GLboolean +compatible_types(GLenum userType, GLenum targetType) +{ + if (userType == targetType) + return GL_TRUE; + + if (targetType == GL_BOOL && (userType == GL_FLOAT || userType == GL_INT)) + return GL_TRUE; + + if (targetType == GL_BOOL_VEC2 && (userType == GL_FLOAT_VEC2 || + userType == GL_INT_VEC2)) + return GL_TRUE; + + if (targetType == GL_BOOL_VEC3 && (userType == GL_FLOAT_VEC3 || + userType == GL_INT_VEC3)) + return GL_TRUE; + + if (targetType == GL_BOOL_VEC4 && (userType == GL_FLOAT_VEC4 || + userType == GL_INT_VEC4)) + return GL_TRUE; + + return GL_FALSE; +} + + /** * Set the value of a program's uniform variable. * \param program the program whose uniform to update @@ -1185,6 +1252,12 @@ static void set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, GLenum type, GLsizei count, GLint elems, const void *values) { + if (!compatible_types(type, + program->Parameters->Parameters[location].DataType)) { + _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(type mismatch)"); + return; + } + if (program->Parameters->Parameters[location].Type == PROGRAM_SAMPLER) { /* This controls which texture unit which is used by a sampler */ GLuint texUnit, sampler; @@ -1317,19 +1390,74 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, } +static void +get_matrix_dims(GLenum type, GLint *rows, GLint *cols) +{ + switch (type) { + case GL_FLOAT_MAT2: + *rows = *cols = 2; + break; + case GL_FLOAT_MAT2x3: + *rows = 3; + *cols = 2; + break; + case GL_FLOAT_MAT2x4: + *rows = 4; + *cols = 2; + break; + case GL_FLOAT_MAT3: + *rows = 3; + *cols = 3; + break; + case GL_FLOAT_MAT3x2: + *rows = 2; + *cols = 3; + break; + case GL_FLOAT_MAT3x4: + *rows = 4; + *cols = 3; + break; + case GL_FLOAT_MAT4: + *rows = 4; + *cols = 4; + break; + case GL_FLOAT_MAT4x2: + *rows = 2; + *cols = 4; + break; + case GL_FLOAT_MAT4x3: + *rows = 3; + *cols = 4; + break; + default: + *rows = *cols = 0; + } +} + + static void set_program_uniform_matrix(GLcontext *ctx, struct gl_program *program, GLuint location, GLuint count, GLuint rows, GLuint cols, GLboolean transpose, const GLfloat *values) { + GLuint mat, row, col; + GLuint dst = location, src = 0; + GLint nr, nc; + + /* check that the number of rows, columns is correct */ + get_matrix_dims(program->Parameters->Parameters[location].DataType, &nr, &nc); + if (rows != nr || cols != nc) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glUniformMatrix(matrix size mismatch"); + return; + } + /* * Note: the _columns_ of a matrix are stored in program registers, not * the rows. So, the loops below look a little funny. * XXX could optimize this a bit... */ - GLuint mat, row, col; - GLuint dst = location, src = 0; /* loop over matrices */ for (mat = 0; mat < count; mat++) { diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index b8e130555d..8f5c95c06a 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -3059,6 +3059,23 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) } +/** + * Compute total size of array give size of element, number of elements. + */ +static GLint +array_size(GLint baseSize, GLint arrayLen) +{ + GLint total; + if (arrayLen > 1) { + /* round up base type to multiple of 4 */ + total = ((baseSize + 3) & ~0x3) * MAX2(arrayLen, 1); + } + else { + total = baseSize; + } + return total; +} + /** * Called by compiler when a global variable has been parsed/compiled. @@ -3098,7 +3115,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, } else if (var->type.qualifier == SLANG_QUAL_UNIFORM) { /* Uniform variable */ - const GLint totalSize = size * MAX2(var->array_len, 1); + const GLint totalSize = array_size(size, var->array_len); const GLuint swizzle = _slang_var_swizzle(totalSize, 0); if (prog) { /* user-defined uniform */ @@ -3182,7 +3199,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, /* user-defined vertex attribute */ const GLint attr = -1; /* unknown */ GLint index = _mesa_add_attribute(prog->Attributes, varName, - size, attr); + size, datatype, attr); assert(index >= 0); store = _slang_new_ir_storage(PROGRAM_INPUT, VERT_ATTRIB_GENERIC0 + index, size); diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index f08b53e8b7..e9b9bbc43f 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -211,7 +211,6 @@ _slang_resolve_attributes(struct gl_shader_program *shProg, { GLuint i, j; GLbitfield usedAttributes; - GLint size = 4; /* XXX fix */ assert(prog->Target == GL_VERTEX_PROGRAM_ARB); @@ -253,6 +252,8 @@ _slang_resolve_attributes(struct gl_shader_program *shProg, * Start at 1 since generic attribute 0 always aliases * glVertex/position. */ + GLint size = prog->Attributes->Parameters[k].Size; + GLenum datatype = prog->Attributes->Parameters[k].DataType; for (attr = 1; attr < MAX_VERTEX_ATTRIBS; attr++) { if (((1 << attr) & usedAttributes) == 0) break; @@ -261,7 +262,7 @@ _slang_resolve_attributes(struct gl_shader_program *shProg, /* too many! XXX record error log */ return GL_FALSE; } - _mesa_add_attribute(shProg->Attributes, name, size, attr); + _mesa_add_attribute(shProg->Attributes, name, size, datatype,attr); /* set the attribute as used */ usedAttributes |= 1< Date: Mon, 21 Jul 2008 14:16:07 -0600 Subject: mesa: fix some issues in _mesa_validate_program() --- src/mesa/shader/shader_api.c | 16 +++++++++++----- 1 file changed, 11 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index e25cb63216..f390e3cef4 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -1534,15 +1534,19 @@ static void _mesa_validate_program(GLcontext *ctx, GLuint program) { struct gl_shader_program *shProg; - shProg = _mesa_lookup_shader_program(ctx, program); + + shProg = _mesa_lookup_shader_program_err(ctx, program, "glValidateProgram"); if (!shProg) { - _mesa_error(ctx, GL_INVALID_VALUE, "glValidateProgram(program)"); return; } - /* XXX temporary */ - shProg->Validated = GL_TRUE; - /* From the GL spec: + if (!shProg->LinkStatus) { + shProg->Validated = GL_FALSE; + return; + } + + /* From the GL spec, a program is invalid if any of these are true: + any two active samplers in the current program object are of different types, but refer to the same texture image unit, @@ -1555,6 +1559,8 @@ _mesa_validate_program(GLcontext *ctx, GLuint program) processing exceeds the combined limit on the total number of texture image units allowed. */ + + shProg->Validated = GL_TRUE; } -- cgit v1.2.3 From 7e8f79e39dfb4ba0072e9d34e7fba958e3710f5b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 21 Jul 2008 14:23:33 -0600 Subject: mesa: remove an error check for NV_v_p that doesn't apply to ARB_v_p --- src/mesa/main/varray.c | 5 ----- 1 file changed, 5 deletions(-) (limited to 'src') diff --git a/src/mesa/main/varray.c b/src/mesa/main/varray.c index a4ef948b58..a6aa9c34b2 100644 --- a/src/mesa/main/varray.c +++ b/src/mesa/main/varray.c @@ -583,11 +583,6 @@ _mesa_VertexAttribPointerARB(GLuint index, GLint size, GLenum type, return; } - if (type == GL_UNSIGNED_BYTE && size != 4) { - _mesa_error(ctx, GL_INVALID_VALUE, "glVertexAttribPointerARB(size!=4)"); - return; - } - /* check for valid 'type' and compute StrideB right away */ /* NOTE: more types are supported here than in the NV extension */ switch (type) { -- cgit v1.2.3 From 467e6526a8210e90cf47362b1b4b87b7d49e009d Mon Sep 17 00:00:00 2001 From: Jonathan White Date: Mon, 21 Jul 2008 15:34:38 -0600 Subject: gallium: Temporary workaround for mismatched pipe create and pipe destroy --- src/mesa/state_tracker/st_context.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 8b30e35684..00447fff08 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -217,7 +217,10 @@ void st_destroy_context( struct st_context *st ) cso_destroy_context(cso); + /* Temporary workaround for mismatched pipe create and pipe destroy */ +#if !defined(PIPE_SUBSYSTEM_WINDOWS_USER) && !defined(PIPE_SUBSYSTEM_WINDOWS_CE) pipe->destroy( pipe ); +#endif free(ctx); } -- cgit v1.2.3 From 5ba32b79d8df3403212888dbdc669abf9d783359 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 21 Jul 2008 16:04:24 -0600 Subject: mesa: fix +=, -=, etc. operators --- src/mesa/shader/slang/slang_codegen.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 8f5c95c06a..cd37a33a48 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2925,14 +2925,14 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) { slang_ir_node *n; assert(oper->num_children == 2); - n = _slang_gen_function_call_name(A, "+=", oper, &oper->children[0]); + n = _slang_gen_function_call_name(A, "+=", oper, NULL); return n; } case SLANG_OPER_SUBASSIGN: { slang_ir_node *n; assert(oper->num_children == 2); - n = _slang_gen_function_call_name(A, "-=", oper, &oper->children[0]); + n = _slang_gen_function_call_name(A, "-=", oper, NULL); return n; } break; @@ -2940,14 +2940,14 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) { slang_ir_node *n; assert(oper->num_children == 2); - n = _slang_gen_function_call_name(A, "*=", oper, &oper->children[0]); + n = _slang_gen_function_call_name(A, "*=", oper, NULL); return n; } case SLANG_OPER_DIVASSIGN: { slang_ir_node *n; assert(oper->num_children == 2); - n = _slang_gen_function_call_name(A, "/=", oper, &oper->children[0]); + n = _slang_gen_function_call_name(A, "/=", oper, NULL); return n; } case SLANG_OPER_LOGICALAND: -- cgit v1.2.3 From c9a7ee2d0044c01aeb6646f0235b030ccd7ba844 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 21 Jul 2008 16:04:56 -0600 Subject: mesa: added null ptr check (error handling case) --- src/mesa/shader/slang/slang_emit.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 291fd7c6dd..c01ec3c185 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1225,6 +1225,9 @@ emit_if(slang_emit_info *emitInfo, slang_ir_node *n) } } + if (!n->Children[0]->Store) + return NULL; + #if 0 assert(n->Children[0]->Store->Size == 1); /* a bool! */ #endif -- cgit v1.2.3 From ec9c79a7dd468f3bd0483d3a9d59809bf019e54b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 21 Jul 2008 17:18:22 -0600 Subject: mesa: initial support for GLSL struct/array comparisons --- src/mesa/shader/slang/slang_emit.c | 71 ++++++++++++++++++++++++++++++-------- 1 file changed, 57 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index c01ec3c185..dc02b2f491 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -696,29 +696,72 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) } } else { - /* size > 4, struct compare */ -#if 0 + /* size > 4, struct or array compare. + * XXX this won't work reliably for structs with padding!! + */ GLint i, num = (n->Children[0]->Store->Size + 3) / 4; - /*printf("BEGIN COMPARE size %d\n", num);*/ + slang_ir_storage accTemp; + + if (!n->Store) { + if (!alloc_temp_storage(emitInfo, n, 4)) + return NULL; + } + + accTemp.Size = 4; + accTemp.File = PROGRAM_TEMPORARY; + if (!_slang_alloc_temp(emitInfo->vt, &accTemp)) { + return NULL; + /* out of temps */ + } + for (i = 0; i < num; i++) { - inst = new_instruction(emitInfo, opcode); + /* SNE t0, left[i], right[i] */ + inst = new_instruction(emitInfo, OPCODE_SNE); inst->SrcReg[0].File = n->Children[0]->Store->File; inst->SrcReg[0].Index = n->Children[0]->Store->Index + i; inst->SrcReg[1].File = n->Children[1]->Store->File; inst->SrcReg[1].Index = n->Children[1]->Store->Index + i; - inst->DstReg.File = n->Store->File; - inst->DstReg.Index = n->Store->Index; - - inst->CondUpdate = 1; /* update cond code */ + if (i == 0) { + inst->DstReg.File = accTemp.File; + inst->DstReg.Index = accTemp.Index; + inst->Comment = _mesa_strdup("Begin struct/array comparison"); + } + else { + inst->DstReg.File = n->Store->File; + inst->DstReg.Index = n->Store->Index; + } if (i > 0) { - inst->DstReg.CondMask = COND_NE; /* update if !=0 */ + /* ADD accTemp, accTemp, temp; # like logical-OR */ + inst = new_instruction(emitInfo, OPCODE_ADD); + inst->SrcReg[0].File = accTemp.File; + inst->SrcReg[0].Index = accTemp.Index; + inst->SrcReg[1].File = n->Store->File; + inst->SrcReg[1].Index = n->Store->Index; + inst->DstReg.File = accTemp.File; + inst->DstReg.Index = accTemp.Index; } - /*_mesa_print_instruction(inst);*/ } - storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); -#endif - _mesa_problem(NULL, "struct comparison not implemented yet"); - inst = NULL; + + /* compute accTemp.x || accTemp.y || accTemp.z || accTemp.w with DOT4 */ + inst = new_instruction(emitInfo, OPCODE_DP4); + inst->SrcReg[0].File = accTemp.File; + inst->SrcReg[0].Index = accTemp.Index; + inst->SrcReg[1].File = accTemp.File; + inst->SrcReg[1].Index = accTemp.Index; + inst->DstReg.File = n->Store->File; + inst->DstReg.Index = n->Store->Index; + inst->Comment = _mesa_strdup("End struct/array comparison"); + + if (n->Opcode == IR_EQUAL) { + /* compute tmp.x = !tmp.x via tmp.x = (tmp.x == 0) */ + inst = new_instruction(emitInfo, OPCODE_SEQ); + storage_to_src_reg(&inst->SrcReg[0], n->Store); + constant_to_src_reg(&inst->SrcReg[1], 0.0, emitInfo); + storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); + inst->Comment = _mesa_strdup("Invert true/false"); + } + + _slang_free_temp(emitInfo->vt, &accTemp); } /* free temps */ -- cgit v1.2.3 From 4db631a3994ed2f5b7cf6e4ac6c08c0a4c091730 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 21 Jul 2008 19:38:25 +0900 Subject: gallium: Open a new file when existing profile memory map is exhausted. --- src/gallium/auxiliary/util/p_debug_prof.c | 147 +++++++++++++++++++++++------- 1 file changed, 113 insertions(+), 34 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug_prof.c b/src/gallium/auxiliary/util/p_debug_prof.c index 958f99c327..69c914c780 100644 --- a/src/gallium/auxiliary/util/p_debug_prof.c +++ b/src/gallium/auxiliary/util/p_debug_prof.c @@ -49,12 +49,97 @@ #define PROFILE_FILE_SIZE 4*1024*1024 #define FILE_NAME_SIZE 256 -static WCHAR wFileName[FILE_NAME_SIZE]; +static WCHAR wFileName[FILE_NAME_SIZE] = L"\\??\\c:\\00000000.trace"; static ULONG_PTR iFile = 0; -static void *pMap = NULL; -static void *pMapEnd = NULL; +static BYTE *pMap = NULL; +static BYTE *pMapBegin = NULL; +static BYTE *pMapEnd = NULL; +void __declspec(naked) __cdecl +debug_profile_close(void) +{ + _asm { + push eax + push ebx + push ecx + push edx + push ebp + push edi + push esi + } + + if(iFile) { + EngUnmapFile(iFile); + /* Truncate the file */ + pMap = EngMapFile(wFileName, pMap - pMapBegin, &iFile); + if(pMap) + EngUnmapFile(iFile); + } + iFile = 0; + pMapBegin = pMapEnd = pMap = NULL; + + _asm { + pop esi + pop edi + pop ebp + pop edx + pop ecx + pop ebx + pop eax + ret + } +} + + +void __declspec(naked) __cdecl +debug_profile_open(void) +{ + WCHAR *p; + + _asm { + push eax + push ebx + push ecx + push edx + push ebp + push edi + push esi + } + + debug_profile_close(); + + // increment starting from the less significant digit + p = &wFileName[14]; + while(1) { + if(*p == '9') { + *p-- = '0'; + } + else { + *p += 1; + break; + } + } + + pMap = EngMapFile(wFileName, PROFILE_FILE_SIZE, &iFile); + if(pMap) { + pMapBegin = pMap; + pMapEnd = pMap + PROFILE_FILE_SIZE; + } + + _asm { + pop esi + pop edi + pop ebp + pop edx + pop ecx + pop ebx + pop eax + ret + } +} + + /** * Called at the start of every method or function. * @@ -64,11 +149,15 @@ void __declspec(naked) __cdecl _penter(void) { _asm { push ebx +retry: mov ebx, [pMap] test ebx, ebx jz done cmp ebx, [pMapEnd] - je done + jne ready + call debug_profile_open + jmp retry +ready: push eax push edx mov eax, [esp+12] @@ -97,11 +186,15 @@ void __declspec(naked) __cdecl _pexit(void) { _asm { push ebx +retry: mov ebx, [pMap] test ebx, ebx jz done cmp ebx, [pMapEnd] - je done + jne ready + call debug_profile_open + jmp retry +ready: push eax push edx mov eax, [esp+12] @@ -121,21 +214,13 @@ done: } +/** + * Reference function for calibration. + */ void __declspec(naked) -__debug_profile_reference1(void) { - _asm { - call _penter - call _pexit - ret - } -} - - -void __declspec(naked) -__debug_profile_reference2(void) { +__debug_profile_reference(void) { _asm { call _penter - call __debug_profile_reference1 call _pexit ret } @@ -145,31 +230,25 @@ __debug_profile_reference2(void) { void debug_profile_start(void) { - static unsigned no = 0; - char filename[FILE_NAME_SIZE]; - unsigned i; - - util_snprintf(filename, sizeof(filename), "\\??\\c:\\%03u.prof", ++no); - for(i = 0; i < FILE_NAME_SIZE; ++i) - wFileName[i] = (WCHAR)filename[i]; + debug_profile_open(); - pMap = EngMapFile(wFileName, PROFILE_FILE_SIZE, &iFile); if(pMap) { - pMapEnd = (unsigned char*)pMap + PROFILE_FILE_SIZE; - /* reference functions for calibration purposes */ - __debug_profile_reference2(); + unsigned i; + for(i = 0; i < 8; ++i) { + _asm { + call _penter + call __debug_profile_reference + call _pexit + } + } } } + void debug_profile_stop(void) { - if(iFile) { - EngUnmapFile(iFile); - /* TODO: truncate file */ - } - iFile = 0; - pMapEnd = pMap = NULL; + debug_profile_close(); } #endif /* PROFILE */ -- cgit v1.2.3 From 883097053d1d3550cab92362a07f413e110520ae Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 22 Jul 2008 09:45:10 +0900 Subject: win32kprof: Store the profile data as an caller->callee hash table, instead of a trace. --- src/gallium/auxiliary/util/p_debug_prof.c | 308 ++++++++++++++++++------------ 1 file changed, 187 insertions(+), 121 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug_prof.c b/src/gallium/auxiliary/util/p_debug_prof.c index 69c914c780..5f9772ef91 100644 --- a/src/gallium/auxiliary/util/p_debug_prof.c +++ b/src/gallium/auxiliary/util/p_debug_prof.c @@ -46,97 +46,83 @@ #include "util/u_string.h" -#define PROFILE_FILE_SIZE 4*1024*1024 +#define PROFILE_TABLE_SIZE (1024*1024) #define FILE_NAME_SIZE 256 -static WCHAR wFileName[FILE_NAME_SIZE] = L"\\??\\c:\\00000000.trace"; -static ULONG_PTR iFile = 0; -static BYTE *pMap = NULL; -static BYTE *pMapBegin = NULL; -static BYTE *pMapEnd = NULL; +struct debug_profile_entry +{ + uintptr_t caller; + uintptr_t callee; + uint64_t samples; +}; +static unsigned long enabled = 0; -void __declspec(naked) __cdecl -debug_profile_close(void) -{ - _asm { - push eax - push ebx - push ecx - push edx - push ebp - push edi - push esi - } +static WCHAR wFileName[FILE_NAME_SIZE] = L"\\??\\c:\\00000000.prof"; +static ULONG_PTR iFile = 0; - if(iFile) { - EngUnmapFile(iFile); - /* Truncate the file */ - pMap = EngMapFile(wFileName, pMap - pMapBegin, &iFile); - if(pMap) - EngUnmapFile(iFile); - } - iFile = 0; - pMapBegin = pMapEnd = pMap = NULL; - - _asm { - pop esi - pop edi - pop ebp - pop edx - pop ecx - pop ebx - pop eax - ret - } -} +static struct debug_profile_entry *table = NULL; +static unsigned long free_table_entries = 0; +static unsigned long max_table_entries = 0; +uint64_t start_stamp = 0; +uint64_t end_stamp = 0; -void __declspec(naked) __cdecl -debug_profile_open(void) -{ - WCHAR *p; - - _asm { - push eax - push ebx - push ecx - push edx - push ebp - push edi - push esi - } - debug_profile_close(); +static void +debug_profile_entry(uintptr_t caller, uintptr_t callee, uint64_t samples) +{ + unsigned hash = ( caller + callee ) & PROFILE_TABLE_SIZE - 1; - // increment starting from the less significant digit - p = &wFileName[14]; while(1) { - if(*p == '9') { - *p-- = '0'; + if(table[hash].caller == 0 && table[hash].callee == 0) { + table[hash].caller = caller; + table[hash].callee = callee; + table[hash].samples = samples; + --free_table_entries; + break; } - else { - *p += 1; + else if(table[hash].caller == caller && table[hash].callee == callee) { + table[hash].samples += samples; break; } + else { + ++hash; + } } +} - pMap = EngMapFile(wFileName, PROFILE_FILE_SIZE, &iFile); - if(pMap) { - pMapBegin = pMap; - pMapEnd = pMap + PROFILE_FILE_SIZE; - } - - _asm { - pop esi - pop edi - pop ebp - pop edx - pop ecx - pop ebx - pop eax - ret - } + +static uintptr_t caller_stack[1024]; +static unsigned last_caller = 0; + + +static int64_t delta(void) { + int64_t result = end_stamp - start_stamp; + if(result > UINT64_C(0xffffffff)) + result = 0; + return result; +} + + +static void __cdecl +debug_profile_enter(uintptr_t callee) +{ + uintptr_t caller = last_caller ? caller_stack[last_caller - 1] : 0; + + if (caller) + debug_profile_entry(caller, 0, delta()); + debug_profile_entry(caller, callee, 1); + caller_stack[last_caller++] = callee; +} + + +static void __cdecl +debug_profile_exit(uintptr_t callee) +{ + debug_profile_entry(callee, 0, delta()); + if(last_caller) + --last_caller; } @@ -148,31 +134,49 @@ debug_profile_open(void) void __declspec(naked) __cdecl _penter(void) { _asm { - push ebx -retry: - mov ebx, [pMap] - test ebx, ebx - jz done - cmp ebx, [pMapEnd] - jne ready - call debug_profile_open - jmp retry -ready: push eax + mov eax, [enabled] + test eax, eax + jz skip + push edx - mov eax, [esp+12] - and eax, 0xfffffffe - mov [ebx], eax - add ebx, 4 + rdtsc - mov [ebx], eax - add ebx, 4 - mov [pMap], ebx + mov dword ptr [end_stamp], eax + mov dword ptr [end_stamp+4], edx + + xor eax, eax + mov [enabled], eax + + mov eax, [esp+8] + + push ebx + push ecx + push ebp + push edi + push esi + + push eax + call debug_profile_enter + add esp, 4 + + pop esi + pop edi + pop ebp + pop ecx + pop ebx + + mov eax, 1 + mov [enabled], eax + + rdtsc + mov dword ptr [start_stamp], eax + mov dword ptr [start_stamp+4], edx + pop edx +skip: pop eax -done: - pop ebx - ret + ret } } @@ -185,30 +189,48 @@ done: void __declspec(naked) __cdecl _pexit(void) { _asm { - push ebx -retry: - mov ebx, [pMap] - test ebx, ebx - jz done - cmp ebx, [pMapEnd] - jne ready - call debug_profile_open - jmp retry -ready: push eax + mov eax, [enabled] + test eax, eax + jz skip + push edx - mov eax, [esp+12] - or eax, 0x00000001 - mov [ebx], eax - add ebx, 4 + rdtsc - mov [ebx], eax - add ebx, 4 - mov [pMap], ebx + mov dword ptr [end_stamp], eax + mov dword ptr [end_stamp+4], edx + + xor eax, eax + mov [enabled], eax + + mov eax, [esp+8] + + push ebx + push ecx + push ebp + push edi + push esi + + push eax + call debug_profile_exit + add esp, 4 + + pop esi + pop edi + pop ebp + pop ecx + pop ebx + + mov eax, 1 + mov [enabled], eax + + rdtsc + mov dword ptr [start_stamp], eax + mov dword ptr [start_stamp+4], edx + pop edx +skip: pop eax -done: - pop ebx ret } } @@ -230,15 +252,53 @@ __debug_profile_reference(void) { void debug_profile_start(void) { - debug_profile_open(); - - if(pMap) { + WCHAR *p; + + // increment starting from the less significant digit + p = &wFileName[14]; + while(1) { + if(*p == '9') { + *p-- = '0'; + } + else { + *p += 1; + break; + } + } + + table = EngMapFile(wFileName, + PROFILE_TABLE_SIZE*sizeof(struct debug_profile_entry), + &iFile); + if(table) { unsigned i; + + free_table_entries = max_table_entries = PROFILE_TABLE_SIZE; + memset(table, 0, PROFILE_TABLE_SIZE*sizeof(struct debug_profile_entry)); + + table[0].caller = (uintptr_t)&__debug_profile_reference; + table[0].callee = 0; + table[0].samples = 0; + --free_table_entries; + + _asm { + push edx + push eax + + rdtsc + mov dword ptr [start_stamp], eax + mov dword ptr [start_stamp+4], edx + + pop edx + pop eax + } + + last_caller = 0; + + enabled = 1; + for(i = 0; i < 8; ++i) { _asm { - call _penter call __debug_profile_reference - call _pexit } } } @@ -248,7 +308,13 @@ debug_profile_start(void) void debug_profile_stop(void) { - debug_profile_close(); + enabled = 0; + + if(iFile) + EngUnmapFile(iFile); + iFile = 0; + table = NULL; + free_table_entries = max_table_entries = 0; } #endif /* PROFILE */ -- cgit v1.2.3 From e3ca92aa28e2bf28e72cea340d9c8d5c9bb48089 Mon Sep 17 00:00:00 2001 From: Brian Date: Tue, 15 Jul 2008 19:54:25 -0600 Subject: mesa: refactor: move _mesa_Bind/Gen/DeleteProgram() to arbprogram.c No API-level functions now in program.c. --- src/mesa/shader/arbprogram.c | 202 +++++++++++++++++++++++++++++++++++++++++ src/mesa/shader/arbprogram.h | 10 +++ src/mesa/shader/program.c | 209 +------------------------------------------ src/mesa/shader/program.h | 18 ---- 4 files changed, 213 insertions(+), 226 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c index 1656dc9450..81c20a0150 100644 --- a/src/mesa/shader/arbprogram.c +++ b/src/mesa/shader/arbprogram.c @@ -33,12 +33,214 @@ #include "arbprogram.h" #include "arbprogparse.h" #include "context.h" +#include "hash.h" #include "imports.h" #include "macros.h" #include "mtypes.h" #include "program.h" + +/** + * Mixing ARB and NV vertex/fragment programs can be tricky. + * Note: GL_VERTEX_PROGRAM_ARB == GL_VERTEX_PROGRAM_NV + * but, GL_FRAGMENT_PROGRAM_ARB != GL_FRAGMENT_PROGRAM_NV + * The two different fragment program targets are supposed to be compatible + * to some extent (see GL_ARB_fragment_program spec). + * This function does the compatibility check. + */ +static GLboolean +compatible_program_targets(GLenum t1, GLenum t2) +{ + if (t1 == t2) + return GL_TRUE; + if (t1 == GL_FRAGMENT_PROGRAM_ARB && t2 == GL_FRAGMENT_PROGRAM_NV) + return GL_TRUE; + if (t1 == GL_FRAGMENT_PROGRAM_NV && t2 == GL_FRAGMENT_PROGRAM_ARB) + return GL_TRUE; + return GL_FALSE; +} + + +/** + * Bind a program (make it current) + * \note Called from the GL API dispatcher by both glBindProgramNV + * and glBindProgramARB. + */ +void GLAPIENTRY +_mesa_BindProgram(GLenum target, GLuint id) +{ + struct gl_program *curProg, *newProg; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + FLUSH_VERTICES(ctx, _NEW_PROGRAM); + + /* Error-check target and get curProg */ + if ((target == GL_VERTEX_PROGRAM_ARB) && /* == GL_VERTEX_PROGRAM_NV */ + (ctx->Extensions.NV_vertex_program || + ctx->Extensions.ARB_vertex_program)) { + curProg = &ctx->VertexProgram.Current->Base; + } + else if ((target == GL_FRAGMENT_PROGRAM_NV + && ctx->Extensions.NV_fragment_program) || + (target == GL_FRAGMENT_PROGRAM_ARB + && ctx->Extensions.ARB_fragment_program)) { + curProg = &ctx->FragmentProgram.Current->Base; + } + else { + _mesa_error(ctx, GL_INVALID_ENUM, "glBindProgramNV/ARB(target)"); + return; + } + + /* + * Get pointer to new program to bind. + * NOTE: binding to a non-existant program is not an error. + * That's supposed to be caught in glBegin. + */ + if (id == 0) { + /* Bind a default program */ + newProg = NULL; + if (target == GL_VERTEX_PROGRAM_ARB) /* == GL_VERTEX_PROGRAM_NV */ + newProg = &ctx->Shared->DefaultVertexProgram->Base; + else + newProg = &ctx->Shared->DefaultFragmentProgram->Base; + } + else { + /* Bind a user program */ + newProg = _mesa_lookup_program(ctx, id); + if (!newProg || newProg == &_mesa_DummyProgram) { + /* allocate a new program now */ + newProg = ctx->Driver.NewProgram(ctx, target, id); + if (!newProg) { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBindProgramNV/ARB"); + return; + } + _mesa_HashInsert(ctx->Shared->Programs, id, newProg); + } + else if (!compatible_program_targets(newProg->Target, target)) { + _mesa_error(ctx, GL_INVALID_OPERATION, + "glBindProgramNV/ARB(target mismatch)"); + return; + } + } + + /** All error checking is complete now **/ + + if (curProg->Id == id) { + /* binding same program - no change */ + return; + } + + /* bind newProg */ + if (target == GL_VERTEX_PROGRAM_ARB) { /* == GL_VERTEX_PROGRAM_NV */ + _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, + (struct gl_vertex_program *) newProg); + } + else if (target == GL_FRAGMENT_PROGRAM_NV || + target == GL_FRAGMENT_PROGRAM_ARB) { + _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, + (struct gl_fragment_program *) newProg); + } + + /* Never null pointers */ + ASSERT(ctx->VertexProgram.Current); + ASSERT(ctx->FragmentProgram.Current); + + if (ctx->Driver.BindProgram) + ctx->Driver.BindProgram(ctx, target, newProg); +} + + +/** + * Delete a list of programs. + * \note Not compiled into display lists. + * \note Called by both glDeleteProgramsNV and glDeleteProgramsARB. + */ +void GLAPIENTRY +_mesa_DeletePrograms(GLsizei n, const GLuint *ids) +{ + GLint i; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); + + if (n < 0) { + _mesa_error( ctx, GL_INVALID_VALUE, "glDeleteProgramsNV" ); + return; + } + + for (i = 0; i < n; i++) { + if (ids[i] != 0) { + struct gl_program *prog = _mesa_lookup_program(ctx, ids[i]); + if (prog == &_mesa_DummyProgram) { + _mesa_HashRemove(ctx->Shared->Programs, ids[i]); + } + else if (prog) { + /* Unbind program if necessary */ + if (prog->Target == GL_VERTEX_PROGRAM_ARB || /* == GL_VERTEX_PROGRAM_NV */ + prog->Target == GL_VERTEX_STATE_PROGRAM_NV) { + if (ctx->VertexProgram.Current && + ctx->VertexProgram.Current->Base.Id == ids[i]) { + /* unbind this currently bound program */ + _mesa_BindProgram(prog->Target, 0); + } + } + else if (prog->Target == GL_FRAGMENT_PROGRAM_NV || + prog->Target == GL_FRAGMENT_PROGRAM_ARB) { + if (ctx->FragmentProgram.Current && + ctx->FragmentProgram.Current->Base.Id == ids[i]) { + /* unbind this currently bound program */ + _mesa_BindProgram(prog->Target, 0); + } + } + else { + _mesa_problem(ctx, "bad target in glDeleteProgramsNV"); + return; + } + /* The ID is immediately available for re-use now */ + _mesa_HashRemove(ctx->Shared->Programs, ids[i]); + _mesa_reference_program(ctx, &prog, NULL); + } + } + } +} + + +/** + * Generate a list of new program identifiers. + * \note Not compiled into display lists. + * \note Called by both glGenProgramsNV and glGenProgramsARB. + */ +void GLAPIENTRY +_mesa_GenPrograms(GLsizei n, GLuint *ids) +{ + GLuint first; + GLuint i; + GET_CURRENT_CONTEXT(ctx); + ASSERT_OUTSIDE_BEGIN_END(ctx); + + if (n < 0) { + _mesa_error(ctx, GL_INVALID_VALUE, "glGenPrograms"); + return; + } + + if (!ids) + return; + + first = _mesa_HashFindFreeKeyBlock(ctx->Shared->Programs, n); + + /* Insert pointer to dummy program as placeholder */ + for (i = 0; i < (GLuint) n; i++) { + _mesa_HashInsert(ctx->Shared->Programs, first + i, &_mesa_DummyProgram); + } + + /* Return the program names */ + for (i = 0; i < (GLuint) n; i++) { + ids[i] = first + i; + } +} + + void GLAPIENTRY _mesa_EnableVertexAttribArrayARB(GLuint index) { diff --git a/src/mesa/shader/arbprogram.h b/src/mesa/shader/arbprogram.h index 233f662965..6fe76267be 100644 --- a/src/mesa/shader/arbprogram.h +++ b/src/mesa/shader/arbprogram.h @@ -27,6 +27,16 @@ #define ARBPROGRAM_H +extern void GLAPIENTRY +_mesa_BindProgram(GLenum target, GLuint id); + +extern void GLAPIENTRY +_mesa_DeletePrograms(GLsizei n, const GLuint *ids); + +extern void GLAPIENTRY +_mesa_GenPrograms(GLsizei n, GLuint *ids); + + extern void GLAPIENTRY _mesa_EnableVertexAttribArrayARB(GLuint index); diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index b27ed6b7d3..13d6df1ce6 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -82,7 +82,7 @@ _mesa_init_program(GLcontext *ctx) /* XXX probably move this stuff */ #if FEATURE_ATI_fragment_shader ctx->ATIFragmentShader.Enabled = GL_FALSE; - ctx->ATIFragmentShader.Current = (struct ati_fragment_shader *) ctx->Shared->DefaultFragmentShader; + ctx->ATIFragmentShader.Current = ctx->Shared->DefaultFragmentShader; assert(ctx->ATIFragmentShader.Current); ctx->ATIFragmentShader.Current->RefCount++; #endif @@ -679,210 +679,3 @@ _mesa_find_free_register(const struct gl_program *prog, GLuint regFile) return -1; } - - - -/** - * Mixing ARB and NV vertex/fragment programs can be tricky. - * Note: GL_VERTEX_PROGRAM_ARB == GL_VERTEX_PROGRAM_NV - * but, GL_FRAGMENT_PROGRAM_ARB != GL_FRAGMENT_PROGRAM_NV - * The two different fragment program targets are supposed to be compatible - * to some extent (see GL_ARB_fragment_program spec). - * This function does the compatibility check. - */ -static GLboolean -compatible_program_targets(GLenum t1, GLenum t2) -{ - if (t1 == t2) - return GL_TRUE; - if (t1 == GL_FRAGMENT_PROGRAM_ARB && t2 == GL_FRAGMENT_PROGRAM_NV) - return GL_TRUE; - if (t1 == GL_FRAGMENT_PROGRAM_NV && t2 == GL_FRAGMENT_PROGRAM_ARB) - return GL_TRUE; - return GL_FALSE; -} - - - -/**********************************************************************/ -/* API functions */ -/**********************************************************************/ - - -/** - * Bind a program (make it current) - * \note Called from the GL API dispatcher by both glBindProgramNV - * and glBindProgramARB. - */ -void GLAPIENTRY -_mesa_BindProgram(GLenum target, GLuint id) -{ - struct gl_program *curProg, *newProg; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - FLUSH_VERTICES(ctx, _NEW_PROGRAM); - - /* Error-check target and get curProg */ - if ((target == GL_VERTEX_PROGRAM_ARB) && /* == GL_VERTEX_PROGRAM_NV */ - (ctx->Extensions.NV_vertex_program || - ctx->Extensions.ARB_vertex_program)) { - curProg = &ctx->VertexProgram.Current->Base; - } - else if ((target == GL_FRAGMENT_PROGRAM_NV - && ctx->Extensions.NV_fragment_program) || - (target == GL_FRAGMENT_PROGRAM_ARB - && ctx->Extensions.ARB_fragment_program)) { - curProg = &ctx->FragmentProgram.Current->Base; - } - else { - _mesa_error(ctx, GL_INVALID_ENUM, "glBindProgramNV/ARB(target)"); - return; - } - - /* - * Get pointer to new program to bind. - * NOTE: binding to a non-existant program is not an error. - * That's supposed to be caught in glBegin. - */ - if (id == 0) { - /* Bind a default program */ - newProg = NULL; - if (target == GL_VERTEX_PROGRAM_ARB) /* == GL_VERTEX_PROGRAM_NV */ - newProg = &ctx->Shared->DefaultVertexProgram->Base; - else - newProg = &ctx->Shared->DefaultFragmentProgram->Base; - } - else { - /* Bind a user program */ - newProg = _mesa_lookup_program(ctx, id); - if (!newProg || newProg == &_mesa_DummyProgram) { - /* allocate a new program now */ - newProg = ctx->Driver.NewProgram(ctx, target, id); - if (!newProg) { - _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBindProgramNV/ARB"); - return; - } - _mesa_HashInsert(ctx->Shared->Programs, id, newProg); - } - else if (!compatible_program_targets(newProg->Target, target)) { - _mesa_error(ctx, GL_INVALID_OPERATION, - "glBindProgramNV/ARB(target mismatch)"); - return; - } - } - - /** All error checking is complete now **/ - - if (curProg->Id == id) { - /* binding same program - no change */ - return; - } - - /* bind newProg */ - if (target == GL_VERTEX_PROGRAM_ARB) { /* == GL_VERTEX_PROGRAM_NV */ - _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, - (struct gl_vertex_program *) newProg); - } - else if (target == GL_FRAGMENT_PROGRAM_NV || - target == GL_FRAGMENT_PROGRAM_ARB) { - _mesa_reference_fragprog(ctx, &ctx->FragmentProgram.Current, - (struct gl_fragment_program *) newProg); - } - - /* Never null pointers */ - ASSERT(ctx->VertexProgram.Current); - ASSERT(ctx->FragmentProgram.Current); - - if (ctx->Driver.BindProgram) - ctx->Driver.BindProgram(ctx, target, newProg); -} - - -/** - * Delete a list of programs. - * \note Not compiled into display lists. - * \note Called by both glDeleteProgramsNV and glDeleteProgramsARB. - */ -void GLAPIENTRY -_mesa_DeletePrograms(GLsizei n, const GLuint *ids) -{ - GLint i; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END_AND_FLUSH(ctx); - - if (n < 0) { - _mesa_error( ctx, GL_INVALID_VALUE, "glDeleteProgramsNV" ); - return; - } - - for (i = 0; i < n; i++) { - if (ids[i] != 0) { - struct gl_program *prog = _mesa_lookup_program(ctx, ids[i]); - if (prog == &_mesa_DummyProgram) { - _mesa_HashRemove(ctx->Shared->Programs, ids[i]); - } - else if (prog) { - /* Unbind program if necessary */ - if (prog->Target == GL_VERTEX_PROGRAM_ARB || /* == GL_VERTEX_PROGRAM_NV */ - prog->Target == GL_VERTEX_STATE_PROGRAM_NV) { - if (ctx->VertexProgram.Current && - ctx->VertexProgram.Current->Base.Id == ids[i]) { - /* unbind this currently bound program */ - _mesa_BindProgram(prog->Target, 0); - } - } - else if (prog->Target == GL_FRAGMENT_PROGRAM_NV || - prog->Target == GL_FRAGMENT_PROGRAM_ARB) { - if (ctx->FragmentProgram.Current && - ctx->FragmentProgram.Current->Base.Id == ids[i]) { - /* unbind this currently bound program */ - _mesa_BindProgram(prog->Target, 0); - } - } - else { - _mesa_problem(ctx, "bad target in glDeleteProgramsNV"); - return; - } - /* The ID is immediately available for re-use now */ - _mesa_HashRemove(ctx->Shared->Programs, ids[i]); - _mesa_reference_program(ctx, &prog, NULL); - } - } - } -} - - -/** - * Generate a list of new program identifiers. - * \note Not compiled into display lists. - * \note Called by both glGenProgramsNV and glGenProgramsARB. - */ -void GLAPIENTRY -_mesa_GenPrograms(GLsizei n, GLuint *ids) -{ - GLuint first; - GLuint i; - GET_CURRENT_CONTEXT(ctx); - ASSERT_OUTSIDE_BEGIN_END(ctx); - - if (n < 0) { - _mesa_error(ctx, GL_INVALID_VALUE, "glGenPrograms"); - return; - } - - if (!ids) - return; - - first = _mesa_HashFindFreeKeyBlock(ctx->Shared->Programs, n); - - /* Insert pointer to dummy program as placeholder */ - for (i = 0; i < (GLuint) n; i++) { - _mesa_HashInsert(ctx->Shared->Programs, first + i, &_mesa_DummyProgram); - } - - /* Return the program names */ - for (i = 0; i < (GLuint) n; i++) { - ids[i] = first + i; - } -} diff --git a/src/mesa/shader/program.h b/src/mesa/shader/program.h index 08fe576afc..7484961364 100644 --- a/src/mesa/shader/program.h +++ b/src/mesa/shader/program.h @@ -46,10 +46,6 @@ extern struct gl_program _mesa_DummyProgram; -/* - * Internal functions - */ - extern void _mesa_init_program(GLcontext *ctx); @@ -121,19 +117,5 @@ extern GLint _mesa_find_free_register(const struct gl_program *prog, GLuint regFile); -/* - * API functions common to ARB/NV_vertex/fragment_program - */ - -extern void GLAPIENTRY -_mesa_BindProgram(GLenum target, GLuint id); - -extern void GLAPIENTRY -_mesa_DeletePrograms(GLsizei n, const GLuint *ids); - -extern void GLAPIENTRY -_mesa_GenPrograms(GLsizei n, GLuint *ids); - - #endif /* PROGRAM_H */ -- cgit v1.2.3 From 63025ec726eb6d3e9a3140f3928a8fc8840c806f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 21 Jul 2008 20:42:05 -0600 Subject: mesa: fix glUniform error checking for samplers --- src/mesa/shader/shader_api.c | 24 ++++++++++++++++++++++++ 1 file changed, 24 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index f390e3cef4..99bcc0aeb4 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -1209,6 +1209,27 @@ update_textures_used(struct gl_program *prog) } +static GLboolean +is_sampler_type(GLenum type) +{ + switch (type) { + case GL_SAMPLER_1D: + case GL_SAMPLER_2D: + case GL_SAMPLER_3D: + case GL_SAMPLER_CUBE: + case GL_SAMPLER_1D_SHADOW: + case GL_SAMPLER_2D_SHADOW: + case GL_SAMPLER_2D_RECT_ARB: + case GL_SAMPLER_2D_RECT_SHADOW_ARB: + case GL_SAMPLER_1D_ARRAY_EXT: + case GL_SAMPLER_2D_ARRAY_EXT: + return GL_TRUE; + default: + return GL_FALSE; + } +} + + /** * Check if the type given by userType is allowed to set a uniform of the * target type. Generally, equivalence is required, but setting Boolean @@ -1235,6 +1256,9 @@ compatible_types(GLenum userType, GLenum targetType) userType == GL_INT_VEC4)) return GL_TRUE; + if (is_sampler_type(targetType) && userType == GL_INT) + return GL_TRUE; + return GL_FALSE; } -- cgit v1.2.3 From 7467a943fc7aeca050014424d5bb2e3ca30c1a04 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Sun, 20 Jul 2008 10:49:48 +0200 Subject: add env var to override msaa visual selection --- src/mesa/state_tracker/st_framebuffer.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 1994a1c826..039a9cfe89 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -52,8 +52,12 @@ st_create_framebuffer( const __GLcontextModes *visual, struct st_framebuffer *stfb = CALLOC_STRUCT(st_framebuffer); if (stfb) { int samples = 0; + const char *msaa_override = _mesa_getenv("__GL_FSAA_MODE"); _mesa_initialize_framebuffer(&stfb->Base, visual); if (visual->sampleBuffers) samples = visual->samples; + if (msaa_override) { + samples = _mesa_atoi(msaa_override); + } { /* fake frontbuffer */ -- cgit v1.2.3 From 51d219dbfe6852d348755574184639940af444e3 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 22 Jul 2008 17:14:54 +0200 Subject: tgsi: Fix immediate usage checks. Provide more info for register usage errors/warnings. --- src/gallium/auxiliary/tgsi/util/tgsi_sanity.c | 56 +++++++++++++++++++++------ 1 file changed, 45 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c index 08f3995b13..9673f061ce 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c @@ -53,18 +53,32 @@ struct sanity_check_ctx static void report_error( struct sanity_check_ctx *ctx, - const char *msg ) + const char *format, + ... ) { - debug_printf( "\nError: %s", msg ); + va_list args; + + debug_printf( "Error : " ); + va_start( args, format ); + _debug_vprintf( format, args ); + va_end( args ); + debug_printf( "\n" ); ctx->errors++; } static void report_warning( struct sanity_check_ctx *ctx, - const char *msg ) + const char *format, + ... ) { - debug_printf( "\nWarning: %s", msg ); + va_list args; + + debug_printf( "Warning: " ); + va_start( args, format ); + _debug_vprintf( format, args ); + va_end( args ); + debug_printf( "\n" ); ctx->warnings++; } @@ -74,7 +88,7 @@ check_file_name( uint file ) { if (file <= TGSI_FILE_NULL || file >= TGSI_FILE_COUNT) { - report_error( ctx, "Invalid file name" ); + report_error( ctx, "Invalid register file name" ); return FALSE; } return TRUE; @@ -115,23 +129,36 @@ is_register_used( return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; } +static const char *file_names[] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + static boolean check_register_usage( struct sanity_check_ctx *ctx, uint file, int index, - boolean indirect ) + const char *name, + boolean indirect_access ) { if (!check_file_name( ctx, file )) return FALSE; - if (indirect) { + if (indirect_access) { if (!is_any_register_declared( ctx, file )) - report_error( ctx, "Undeclared source register" ); + report_error( ctx, "%s: Undeclared %s register", file_names[file], name ); ctx->regs_ind_used[file] = TRUE; } else { if (!is_register_declared( ctx, file, index )) - report_error( ctx, "Undeclared destination register" ); + report_error( ctx, "%s[%d]: Undeclared %s register", file_names[file], index, name ); ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG)); } return TRUE; @@ -162,6 +189,7 @@ iter_instruction( ctx, inst->FullDstRegisters[i].DstRegister.File, inst->FullDstRegisters[i].DstRegister.Index, + "destination", FALSE ); } for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { @@ -169,6 +197,7 @@ iter_instruction( ctx, inst->FullSrcRegisters[i].SrcRegister.File, inst->FullSrcRegisters[i].SrcRegister.Index, + "source", inst->FullSrcRegisters[i].SrcRegister.Indirect ); if (inst->FullSrcRegisters[i].SrcRegister.Indirect) { uint file; @@ -176,7 +205,12 @@ iter_instruction( file = inst->FullSrcRegisters[i].SrcRegisterInd.File; index = inst->FullSrcRegisters[i].SrcRegisterInd.Index; - check_register_usage( ctx, file, index, FALSE ); + check_register_usage( + ctx, + file, + index, + "indirect", + FALSE ); if (file != TGSI_FILE_ADDRESS || index != 0) report_warning( ctx, "Indirect register not ADDR[0]" ); } @@ -232,8 +266,8 @@ iter_immediate( /* Mark the register as declared. */ - ctx->num_imms++; ctx->regs_decl[TGSI_FILE_IMMEDIATE][ctx->num_imms / BITS_IN_REG_FLAG] |= (1 << (ctx->num_imms % BITS_IN_REG_FLAG)); + ctx->num_imms++; /* Check data type validity. */ -- cgit v1.2.3 From e4139657e0ac40febef1f5e229317aa58b032261 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Jul 2008 11:54:22 -0600 Subject: mesa: remove stray debug assertion --- src/mesa/shader/shader_api.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 99bcc0aeb4..bd3745cfe6 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -538,7 +538,6 @@ _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, if (shProg->LinkStatus) { /* get current index/location for the attribute */ oldIndex = _mesa_get_attrib_location(ctx, program, name); - assert(0); } else { oldIndex = -1; -- cgit v1.2.3 From 11a5c2d4eea69122fd09f272e68afb95143873c6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Jul 2008 15:17:10 -0600 Subject: mesa: glsl: implement constructor functions for user-defined types --- src/mesa/shader/slang/slang_codegen.c | 166 +++++++++++++++++++++++++++ src/mesa/shader/slang/slang_compile_struct.h | 13 ++- 2 files changed, 174 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index cd37a33a48..6e5a54773e 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1724,6 +1724,167 @@ _slang_first_function(struct slang_function_scope_ *scope, const char *name) } +/** + * Generate a new slang_function which is a constructor for a user-defined + * struct type. + */ +static slang_function * +_slang_make_constructor(slang_assemble_ctx *A, slang_struct *str) +{ + const GLint numFields = str->fields->num_variables; + + slang_function *fun = (slang_function *) _mesa_malloc(sizeof(slang_function)); + if (!fun) + return NULL; + + slang_function_construct(fun); + + /* function header (name, return type) */ + fun->kind = SLANG_FUNC_CONSTRUCTOR; + fun->header.a_name = str->a_name; + fun->header.type.qualifier = SLANG_QUAL_NONE; + fun->header.type.specifier.type = SLANG_SPEC_STRUCT; + fun->header.type.specifier._struct = str; + + /* function parameters (= struct's fields) */ + { + GLint i; + for (i = 0; i < numFields; i++) { + /* + printf("Field %d: %s\n", i, (char*) str->fields->variables[i]->a_name); + */ + slang_variable *p = slang_variable_scope_grow(fun->parameters); + *p = *str->fields->variables[i]; + } + fun->param_count = fun->parameters->num_variables; + } + + /* Add __retVal to params */ + { + slang_variable *p = slang_variable_scope_grow(fun->parameters); + slang_atom a_retVal = slang_atom_pool_atom(A->atoms, "__retVal"); + assert(a_retVal); + p->a_name = a_retVal; + p->type = fun->header.type; + p->type.qualifier = SLANG_QUAL_OUT; + fun->param_count++; + } + + /* function body is: + * block: + * declare T; + * T.f1 = p1; + * T.f2 = p2; + * ... + * T.fn = pn; + * return T; + */ + { + slang_variable *var; + GLint i; + + fun->body = slang_operation_new(1); + fun->body->type = SLANG_OPER_BLOCK_NO_NEW_SCOPE; + fun->body->num_children = numFields + 2; + fun->body->children = slang_operation_new(numFields + 2); + + /* create local var 't' */ + var = slang_variable_scope_grow(fun->parameters); + var->a_name = slang_atom_pool_atom(A->atoms, "t"); + var->type = fun->header.type; + + /* declare t */ + { + slang_operation *decl; + + decl = &fun->body->children[0]; + decl->type = SLANG_OPER_VARIABLE_DECL; + decl->locals = _slang_variable_scope_new(fun->parameters); + decl->a_id = var->a_name; + } + + /* assign params to fields of t */ + for (i = 0; i < numFields; i++) { + slang_operation *assign = &fun->body->children[1 + i]; + + assign->type = SLANG_OPER_ASSIGN; + assign->locals = _slang_variable_scope_new(fun->parameters); + assign->num_children = 2; + assign->children = slang_operation_new(2); + + { + slang_operation *lhs = &assign->children[0]; + + lhs->type = SLANG_OPER_FIELD; + lhs->locals = _slang_variable_scope_new(fun->parameters); + lhs->num_children = 1; + lhs->children = slang_operation_new(1); + lhs->a_id = str->fields->variables[i]->a_name; + + lhs->children[0].type = SLANG_OPER_IDENTIFIER; + lhs->children[0].a_id = var->a_name; + lhs->children[0].locals = _slang_variable_scope_new(fun->parameters); + +#if 0 + lhs->children[1].num_children = 1; + lhs->children[1].children = slang_operation_new(1); + lhs->children[1].children[0].type = SLANG_OPER_IDENTIFIER; + lhs->children[1].children[0].a_id = str->fields->variables[i]->a_name; + lhs->children[1].children->locals = _slang_variable_scope_new(fun->parameters); +#endif + } + + { + slang_operation *rhs = &assign->children[1]; + + rhs->type = SLANG_OPER_IDENTIFIER; + rhs->locals = _slang_variable_scope_new(fun->parameters); + rhs->a_id = str->fields->variables[i]->a_name; + } + } + + /* return t; */ + { + slang_operation *ret = &fun->body->children[numFields + 1]; + + ret->type = SLANG_OPER_RETURN; + ret->locals = _slang_variable_scope_new(fun->parameters); + ret->num_children = 1; + ret->children = slang_operation_new(1); + ret->children[0].type = SLANG_OPER_IDENTIFIER; + ret->children[0].a_id = var->a_name; + ret->children[0].locals = _slang_variable_scope_new(fun->parameters); + + } + } + /* + slang_print_function(fun, 1); + */ + return fun; +} + + +/** + * Find/create a function (constructor) for the given structure name. + */ +static slang_function * +_slang_locate_struct_constructor(slang_assemble_ctx *A, const char *name) +{ + int i; + for (i = 0; i < A->space.structs->num_structs; i++) { + slang_struct *str = &A->space.structs->structs[i]; + if (strcmp(name, (const char *) str->a_name) == 0) { + /* found a structure type that matches the function name */ + if (!str->constructor) { + /* create the constructor function now */ + str->constructor = _slang_make_constructor(A, str); + } + return str->constructor; + } + } + return NULL; +} + /** * Assemble a function call, given a particular function name. @@ -1747,6 +1908,11 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name, */ fun = _slang_locate_function(A->space.funcs, atom, params, param_count, &A->space, A->atoms, A->log); + + if (!fun) { + fun = _slang_locate_struct_constructor(A, name); + } + if (!fun) { /* A function with exactly the right parameters/types was not found. * Try adapting the parameters. diff --git a/src/mesa/shader/slang/slang_compile_struct.h b/src/mesa/shader/slang/slang_compile_struct.h index 79e6306616..90c5512f4d 100644 --- a/src/mesa/shader/slang/slang_compile_struct.h +++ b/src/mesa/shader/slang/slang_compile_struct.h @@ -29,11 +29,13 @@ extern "C" { #endif +struct slang_function_; + typedef struct slang_struct_scope_ { - struct slang_struct_ *structs; + struct slang_struct_ *structs; GLuint num_structs; - struct slang_struct_scope_ *outer_scope; + struct slang_struct_scope_ *outer_scope; } slang_struct_scope; extern GLvoid @@ -45,9 +47,10 @@ struct slang_struct_ *slang_struct_scope_find (slang_struct_scope *, slang_atom, typedef struct slang_struct_ { - slang_atom a_name; - struct slang_variable_scope_ *fields; - slang_struct_scope *structs; + slang_atom a_name; + struct slang_variable_scope_ *fields; + slang_struct_scope *structs; + struct slang_function_ *constructor; } slang_struct; int slang_struct_construct (slang_struct *); -- cgit v1.2.3 From 0d1ed45cbfd555330c8746b0fc046bdb41d767d0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Jul 2008 17:38:55 -0600 Subject: mesa: fix uninitialized var --- src/mesa/shader/slang/slang_compile_struct.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile_struct.c b/src/mesa/shader/slang/slang_compile_struct.c index 97a34d24fe..96bdb1f491 100644 --- a/src/mesa/shader/slang/slang_compile_struct.c +++ b/src/mesa/shader/slang/slang_compile_struct.c @@ -115,6 +115,7 @@ slang_struct_construct(slang_struct * stru) return 0; } _slang_struct_scope_ctr(stru->structs); + stru->constructor = NULL; return 1; } -- cgit v1.2.3 From 7f3d6e74817e8880a0712c85f2b41fd88cf6a347 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 22 Jul 2008 18:27:56 -0600 Subject: mesa: glsl: rework swizzle storage handling Build on the heirarchal approach implemented for arrays/structs. --- src/mesa/shader/slang/slang_builtin.c | 12 +++----- src/mesa/shader/slang/slang_codegen.c | 24 +++++++++++++-- src/mesa/shader/slang/slang_emit.c | 57 ++++++++--------------------------- 3 files changed, 40 insertions(+), 53 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_builtin.c b/src/mesa/shader/slang/slang_builtin.c index f99a59e3e1..ed6de40d4b 100644 --- a/src/mesa/shader/slang/slang_builtin.c +++ b/src/mesa/shader/slang/slang_builtin.c @@ -439,13 +439,11 @@ _slang_alloc_statevar(slang_ir_node *n, pos = lookup_statevar(var, index1, index2, field, &swizzle, paramList); assert(pos >= 0); if (pos >= 0) { - /* XXX should overwrite Store's fields instead of changing pointers - * since there may be a child storage_info pointing to this one. - */ - n0->Store = _slang_new_ir_storage_swz(PROGRAM_STATE_VAR, - pos, - n0->Store->Size, - swizzle); + /* newly resolved storage for the statevar/constant/uniform */ + n0->Store->File = PROGRAM_STATE_VAR; + n0->Store->Index = pos; + n0->Store->Swizzle = swizzle; + n0->Store->Parent = NULL; } return pos; } diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 6e5a54773e..d4bf99f78f 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2630,14 +2630,34 @@ _slang_gen_variable(slang_assemble_ctx * A, slang_operation *oper) } + +/** + * Return the number of components actually named by the swizzle. + * Recall that swizzles may have undefined/don't-care values. + */ +static GLuint +swizzle_size(GLuint swizzle) +{ + GLuint size = 0, i; + for (i = 0; i < 4; i++) { + GLuint swz = GET_SWZ(swizzle, i); + size += (swz >= 0 && swz <= 3); + } + return size; +} + + static slang_ir_node * _slang_gen_swizzle(slang_ir_node *child, GLuint swizzle) { - /* XXX should rewrite this to use relative/Parent storage */ slang_ir_node *n = new_node1(IR_SWIZZLE, child); assert(child); if (n) { - n->Store = _slang_new_ir_storage_swz(PROGRAM_UNDEFINED, -1, -1, swizzle); + assert(!n->Store); + n->Store = _slang_new_ir_storage_relative(0, + swizzle_size(swizzle), + child->Store); + n->Store->Swizzle = swizzle; } return n; } diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index dc02b2f491..912c325bb2 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1505,57 +1505,18 @@ emit_cont_break_if_true(slang_emit_info *emitInfo, slang_ir_node *n) } -/** - * Return the number of components actually named by the swizzle. - * Recall that swizzles may have undefined/don't-care values. - */ -static GLuint -swizzle_size(GLuint swizzle) -{ - GLuint size = 0, i; - for (i = 0; i < 4; i++) { - GLuint swz = GET_SWZ(swizzle, i); - size += (swz >= 0 && swz <= 3); - } - return size; -} - - static struct prog_instruction * emit_swizzle(slang_emit_info *emitInfo, slang_ir_node *n) { - GLuint swizzle; struct prog_instruction *inst; inst = emit(emitInfo, n->Children[0]); - /* For debug: n->Var = n->Children[0]->Var; */ + /* setup storage info, if needed */ + if (!n->Store->Parent) + n->Store->Parent = n->Children[0]->Store; - /* "pull-up" the child's storage info, applying our swizzle info */ - n->Store->File = n->Children[0]->Store->File; - n->Store->Index = n->Children[0]->Store->Index; - n->Store->Size = swizzle_size(n->Store->Swizzle); -#if 0 - printf("Emit Swizzle %s reg %d chSize %d mySize %d\n", - _mesa_swizzle_string(n->Store->Swizzle, 0, 0), - n->Store->Index, n->Children[0]->Store->Size, - n->Store->Size); -#endif - - /* apply this swizzle to child's swizzle to get composed swizzle */ - swizzle = fix_swizzle(n->Store->Swizzle); /* remove the don't care terms */ - -#ifdef DEBUG - { - assert(GET_SWZ(swizzle, 0) != SWIZZLE_NIL); - assert(GET_SWZ(swizzle, 1) != SWIZZLE_NIL); - assert(GET_SWZ(swizzle, 2) != SWIZZLE_NIL); - assert(GET_SWZ(swizzle, 3) != SWIZZLE_NIL); - } -#endif - - n->Store->Swizzle = _slang_swizzle_swizzle(n->Children[0]->Store->Swizzle, - swizzle); + assert(n->Store->Parent); return inst; } @@ -1614,7 +1575,11 @@ emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n) /* resolve new absolute storage location */ assert(n->Store); n->Store->File = root->File; - n->Store->Index = index; + if (root->File == PROGRAM_STATE_VAR) + n->Store->Index = 0; + else + n->Store->Index = index; + n->Store->Parent = NULL; } else { @@ -1639,6 +1604,10 @@ emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n) n->Store->RelAddr = GL_TRUE; } + /* if array element size is one, make sure we only access X */ + if (n->Store->Size == 1) + n->Store->Swizzle = SWIZZLE_XXXX; + return NULL; /* no instruction */ } -- cgit v1.2.3 From 818cd9dca2861dfb9d8f2d36ff141c55472df2ba Mon Sep 17 00:00:00 2001 From: Jakob Bornecrantz Date: Wed, 23 Jul 2008 14:02:33 +0200 Subject: i915: intel_flush_frontbuffer does work --- src/gallium/winsys/dri/intel/intel_screen.c | 8 +++----- 1 file changed, 3 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c index f2412217f3..b3022fd17a 100644 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ b/src/gallium/winsys/dri/intel/intel_screen.c @@ -253,12 +253,10 @@ intel_flush_frontbuffer( struct pipe_winsys *winsys, struct pipe_surface *surf, void *context_private) { - //struct intel_context *intel = (struct intel_context *) context_private; - //__DRIdrawablePrivate *dPriv = intel->driDrawable; + struct intel_context *intel = (struct intel_context *) context_private; + __DRIdrawablePrivate *dPriv = intel->driDrawable; - assert((int)"Doesn't work currently" & 0); - - //intelDisplaySurface(dPriv, surf, NULL); + intelDisplaySurface(dPriv, surf, NULL); } static boolean -- cgit v1.2.3 From 0aa0141e0cd7b5231140805b1cda821bb3d32a86 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 23 Jul 2008 17:52:56 +0200 Subject: tgsi: Fix tgsi_util_get_full_src_register_extswizzle(). --- src/gallium/auxiliary/tgsi/util/tgsi_util.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_util.c b/src/gallium/auxiliary/tgsi/util/tgsi_util.c index 10762b6c1a..09486e649e 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_util.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_util.c @@ -119,7 +119,7 @@ tgsi_util_get_full_src_register_extswizzle( if( swizzle <= TGSI_SWIZZLE_W ) { swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegister, - component ); + swizzle ); } return swizzle; -- cgit v1.2.3 From 93f553ed4d94a13ca1bc9c1ed962bd6963cd9ec6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Jul 2008 09:56:20 -0600 Subject: gallium: fix translation of extended swizzles, per-component negation --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 38 ++++++++++++++++++-------------- 1 file changed, 21 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 150545dfad..9029f12056 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -39,8 +39,6 @@ #include "shader/prog_instruction.h" #include "shader/prog_parameter.h" -#define TGSI_DEBUG 0 - /* * Map mesa register file to TGSI register file. @@ -56,8 +54,8 @@ map_register_file( return TGSI_FILE_NULL; case PROGRAM_TEMPORARY: return TGSI_FILE_TEMPORARY; - //case PROGRAM_LOCAL_PARAM: - //case PROGRAM_ENV_PARAM: + /*case PROGRAM_LOCAL_PARAM:*/ + /*case PROGRAM_ENV_PARAM:*/ /* Because of the longstanding problem with mesa arb shaders * where constants, immediates and state variables are all @@ -230,21 +228,27 @@ compile_instruction( outputMapping, immediateMapping); - for( j = 0; j < 4; j++ ) { - GLuint swz; - - swz = GET_SWZ( inst->SrcReg[i].Swizzle, j ); - if( swz > SWIZZLE_W ) { - tgsi_util_set_src_register_extswizzle( - &fullsrc->SrcRegisterExtSwz, - swz, - j ); + /* swizzle (ext swizzle also depends on negation) */ + { + GLuint swz[4]; + GLboolean extended = (inst->SrcReg[i].NegateBase != NEGATE_NONE && + inst->SrcReg[i].NegateBase != NEGATE_XYZW); + for( j = 0; j < 4; j++ ) { + swz[j] = GET_SWZ( inst->SrcReg[i].Swizzle, j ); + if (swz[j] > SWIZZLE_W) + extended = GL_TRUE; + } + if (extended) { + for (j = 0; j < 4; j++) { + tgsi_util_set_src_register_extswizzle(&fullsrc->SrcRegisterExtSwz, + swz[j], j); + } } else { - tgsi_util_set_src_register_swizzle( - &fullsrc->SrcRegister, - swz, - j ); + for (j = 0; j < 4; j++) { + tgsi_util_set_src_register_swizzle(&fullsrc->SrcRegister, + swz[j], j); + } } } -- cgit v1.2.3 From 5f2a5f6164ace6e12c1a3ba95f1103dc8fafa68f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Jul 2008 09:56:44 -0600 Subject: gallium: print extended swizzle negation flags --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 8 ++++++++ 1 file changed, 8 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c index 94180f7e50..d2e6375212 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c @@ -491,9 +491,17 @@ tgsi_dump_instruction( src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { CHR( '.' ); + if (src->SrcRegisterExtSwz.NegateX) + TXT("-"); ENM( src->SrcRegisterExtSwz.ExtSwizzleX, extswizzle_names ); + if (src->SrcRegisterExtSwz.NegateY) + TXT("-"); ENM( src->SrcRegisterExtSwz.ExtSwizzleY, extswizzle_names ); + if (src->SrcRegisterExtSwz.NegateZ) + TXT("-"); ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, extswizzle_names ); + if (src->SrcRegisterExtSwz.NegateW) + TXT("-"); ENM( src->SrcRegisterExtSwz.ExtSwizzleW, extswizzle_names ); } -- cgit v1.2.3 From 576c5fcc61d8278c5e78f900a23f752734da3988 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Jul 2008 12:00:48 -0600 Subject: mesa: glsl: mark constructor params as const --- src/mesa/shader/slang/slang_codegen.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index d4bf99f78f..d8aa161e43 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1754,7 +1754,8 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str) printf("Field %d: %s\n", i, (char*) str->fields->variables[i]->a_name); */ slang_variable *p = slang_variable_scope_grow(fun->parameters); - *p = *str->fields->variables[i]; + *p = *str->fields->variables[i]; /* copy the type */ + p->type.qualifier = SLANG_QUAL_CONST; } fun->param_count = fun->parameters->num_variables; } -- cgit v1.2.3 From 564c97660a89f853bc6eb284bf0906ca4bad6de7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Jul 2008 15:04:25 -0600 Subject: mesa: glsl: fix/simplify array element handling Also fix bug in comparing large structs/arrays. --- src/mesa/shader/slang/slang_emit.c | 54 +++++++++++--------------------------- 1 file changed, 15 insertions(+), 39 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 912c325bb2..c28894630c 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -717,10 +717,10 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) for (i = 0; i < num; i++) { /* SNE t0, left[i], right[i] */ inst = new_instruction(emitInfo, OPCODE_SNE); - inst->SrcReg[0].File = n->Children[0]->Store->File; - inst->SrcReg[0].Index = n->Children[0]->Store->Index + i; - inst->SrcReg[1].File = n->Children[1]->Store->File; - inst->SrcReg[1].Index = n->Children[1]->Store->Index + i; + storage_to_src_reg(&inst->SrcReg[0], n->Children[0]->Store); + storage_to_src_reg(&inst->SrcReg[1], n->Children[1]->Store); + inst->SrcReg[0].Index += i; + inst->SrcReg[1].Index += i; if (i == 0) { inst->DstReg.File = accTemp.File; inst->DstReg.Index = accTemp.Index; @@ -1547,48 +1547,24 @@ emit_array_element(slang_emit_info *emitInfo, slang_ir_node *n) return NULL; } - if (n->Children[1]->Opcode == IR_FLOAT) { - /* Constant array index */ -#if 0 /* just debug code */ - const GLint arrayAddr = n->Children[0]->Store->Index; - const GLint index = (GLint) n->Children[1]->Value[0]; - assert(index == n->Store->Index); - assert(arrayAddr == parent->Index); - assert(n->Children[0]->Store == parent); - assert(n->Children[0]->Store->Index == parent->Index); -#endif - - GLint index = n->Store->Index; - - slang_ir_storage *p = n->Store; - index = 0; - while (p->Parent) { - int sz = (p->Size + 3) / 4; - /*printf("element [%d] of size %d (%d)\n", p->Index, p->Size, sz);*/ - index += sz * p->Index; - p = p->Parent; - } - index += p->Index; - - assert(root->File != PROGRAM_UNDEFINED); + /* do codegen for array */ + emit(emitInfo, n->Children[0]); - /* resolve new absolute storage location */ - assert(n->Store); - n->Store->File = root->File; - if (root->File == PROGRAM_STATE_VAR) - n->Store->Index = 0; - else - n->Store->Index = index; + if (n->Children[1]->Opcode == IR_FLOAT) { + /* Constant array index. + * Set Store's index to be the offset of the array element in + * the register file. + */ + const GLint element = (GLint) n->Children[1]->Value[0]; + const GLint sz = (n->Store->Size + 3) / 4; /* size in slots/registers */ - n->Store->Parent = NULL; + n->Store->Index = sz * element; + assert(n->Store->Parent); } else { /* Variable array index */ struct prog_instruction *inst; - /* do codegen for array */ - emit(emitInfo, n->Children[0]); - /* do codegen for array index expression */ emit(emitInfo, n->Children[1]); -- cgit v1.2.3 From d5835c1d0f7d3af9169a61e13646247c6a8b01b4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Jul 2008 15:14:22 -0600 Subject: mesa: glsl: fix/simplify built-in constant lookup --- src/mesa/shader/slang/slang_simplify.c | 42 ++++++++++++++++++++-------------- 1 file changed, 25 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c index 88ca83d288..9a4b6a88d1 100644 --- a/src/mesa/shader/slang/slang_simplify.c +++ b/src/mesa/shader/slang/slang_simplify.c @@ -37,6 +37,15 @@ #include "slang_print.h" +#ifndef GL_MAX_FRAGMENT_UNIFORM_VECTORS +#define GL_MAX_FRAGMENT_UNIFORM_VECTORS 0x8DFD +#endif +#ifndef GL_MAX_VERTEX_UNIFORM_VECTORS +#define GL_MAX_VERTEX_UNIFORM_VECTORS 0x8DFB +#endif +#ifndef GL_MAX_VARYING_VECTORS +#define GL_MAX_VARYING_VECTORS 0x8DFC +#endif /** @@ -49,25 +58,24 @@ _slang_lookup_constant(const char *name) struct constant_info { const char *Name; const GLenum Token; - GLint Divisor; }; static const struct constant_info info[] = { - { "gl_MaxClipPlanes", GL_MAX_CLIP_PLANES, 1 }, - { "gl_MaxCombinedTextureImageUnits", GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS, 1 }, - { "gl_MaxDrawBuffers", GL_MAX_DRAW_BUFFERS, 1 }, - { "gl_MaxFragmentUniformComponents", GL_MAX_FRAGMENT_UNIFORM_COMPONENTS, 1 }, - { "gl_MaxLights", GL_MAX_LIGHTS, 1 }, - { "gl_MaxTextureUnits", GL_MAX_TEXTURE_UNITS, 1 }, - { "gl_MaxTextureCoords", GL_MAX_TEXTURE_COORDS, 1 }, - { "gl_MaxVertexAttribs", GL_MAX_VERTEX_ATTRIBS, 1 }, - { "gl_MaxVertexUniformComponents", GL_MAX_VERTEX_UNIFORM_COMPONENTS, 1 }, - { "gl_MaxVaryingFloats", GL_MAX_VARYING_FLOATS, 1 }, - { "gl_MaxVertexTextureImageUnits", GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS, 1 }, - { "gl_MaxTextureImageUnits", GL_MAX_TEXTURE_IMAGE_UNITS, 1 }, + { "gl_MaxClipPlanes", GL_MAX_CLIP_PLANES }, + { "gl_MaxCombinedTextureImageUnits", GL_MAX_COMBINED_TEXTURE_IMAGE_UNITS }, + { "gl_MaxDrawBuffers", GL_MAX_DRAW_BUFFERS }, + { "gl_MaxFragmentUniformComponents", GL_MAX_FRAGMENT_UNIFORM_COMPONENTS }, + { "gl_MaxLights", GL_MAX_LIGHTS }, + { "gl_MaxTextureUnits", GL_MAX_TEXTURE_UNITS }, + { "gl_MaxTextureCoords", GL_MAX_TEXTURE_COORDS }, + { "gl_MaxVertexAttribs", GL_MAX_VERTEX_ATTRIBS }, + { "gl_MaxVertexUniformComponents", GL_MAX_VERTEX_UNIFORM_COMPONENTS }, + { "gl_MaxVaryingFloats", GL_MAX_VARYING_FLOATS }, + { "gl_MaxVertexTextureImageUnits", GL_MAX_VERTEX_TEXTURE_IMAGE_UNITS }, + { "gl_MaxTextureImageUnits", GL_MAX_TEXTURE_IMAGE_UNITS }, #if FEATURE_es2_glsl - { "gl_MaxVertexUniformVectors", GL_MAX_VERTEX_UNIFORM_COMPONENTS, 4 }, - { "gl_MaxVaryingVectors", GL_MAX_VARYING_FLOATS, 4 }, - { "gl_MaxFragmentUniformVectors", GL_MAX_FRAGMENT_UNIFORM_COMPONENTS, 4 }, + { "gl_MaxVertexUniformVectors", GL_MAX_VERTEX_UNIFORM_VECTORS }, + { "gl_MaxVaryingVectors", GL_MAX_VARYING_VECTORS }, + { "gl_MaxFragmentUniformVectors", GL_MAX_FRAGMENT_UNIFORM_VECTORS }, #endif { NULL, 0 } }; @@ -79,7 +87,7 @@ _slang_lookup_constant(const char *name) GLint value = -1; _mesa_GetIntegerv(info[i].Token, &value); ASSERT(value >= 0); /* sanity check that glGetFloatv worked */ - return value / info[i].Divisor; + return value; } } return -1; -- cgit v1.2.3 From f7be39ea105aa951d0f6e1d8ffbea63412e30801 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 23 Jul 2008 16:28:15 -0600 Subject: gallium: bump TGSI_EXEC_NUM_TEMPS to 128 --- src/gallium/auxiliary/tgsi/exec/tgsi_exec.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h index 18abdd9ac0..4f30650b07 100644 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h @@ -100,7 +100,7 @@ struct tgsi_exec_labels }; -#define TGSI_EXEC_NUM_TEMPS 64 +#define TGSI_EXEC_NUM_TEMPS 128 #define TGSI_EXEC_NUM_TEMP_EXTRAS 6 #define TGSI_EXEC_NUM_IMMEDIATES 256 -- cgit v1.2.3 From b4ed6e9b1797a1e08c51c1ad30e0f78d9c6d8b95 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 24 Jul 2008 11:03:05 +0200 Subject: mesa: Silence compiler warning on windows. --- src/mesa/shader/slang/slang_codegen.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index d8aa161e43..40f4105a64 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1871,7 +1871,7 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str) static slang_function * _slang_locate_struct_constructor(slang_assemble_ctx *A, const char *name) { - int i; + unsigned int i; for (i = 0; i < A->space.structs->num_structs; i++) { slang_struct *str = &A->space.structs->structs[i]; if (strcmp(name, (const char *) str->a_name) == 0) { -- cgit v1.2.3 From 83869ceab5b3faed8569a5ca752d4dc426db1aec Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 22 Jul 2008 22:07:22 +0900 Subject: tgsi: Silent msvc warning. Rather stupid warning: msvc is warning that converting from a 1bit structure bitfield to a unsigned char looses precision... /WX makes this an error. --- src/gallium/auxiliary/tgsi/util/tgsi_sanity.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c index 9673f061ce..2e3ec96b5b 100644 --- a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c @@ -198,7 +198,7 @@ iter_instruction( inst->FullSrcRegisters[i].SrcRegister.File, inst->FullSrcRegisters[i].SrcRegister.Index, "source", - inst->FullSrcRegisters[i].SrcRegister.Indirect ); + (boolean)inst->FullSrcRegisters[i].SrcRegister.Indirect ); if (inst->FullSrcRegisters[i].SrcRegister.Indirect) { uint file; int index; -- cgit v1.2.3 From fd6865c7e5c3c38c273b8269ff30a1acec469b6f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 23 Jul 2008 11:15:54 +0900 Subject: softpipe: Remove unused variables. --- src/gallium/drivers/softpipe/sp_tile_cache.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index bfdaaa6b8f..5d10234945 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -364,7 +364,6 @@ void sp_flush_tile_cache(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc) { - struct pipe_context *pipe = &softpipe->pipe; struct pipe_surface *ps = tc->surface; int inuse = 0, pos; @@ -414,7 +413,6 @@ struct softpipe_cached_tile * sp_get_cached_tile(struct softpipe_context *softpipe, struct softpipe_tile_cache *tc, int x, int y) { - struct pipe_context *pipe = &softpipe->pipe; struct pipe_surface *ps = tc->surface; /* tile pos in framebuffer: */ -- cgit v1.2.3 From 101d1a658a614d1e2ec02b1e697f6161291af653 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 23 Jul 2008 21:06:01 +0900 Subject: mesa: Prefix main includes with dir to avoid conflicts. Some of the headers in src/mesa/main have pretty common names which easily conflict with third-party code, e.g. config.h --- src/mesa/SConscript | 1 - src/mesa/glapi/glapi.c | 2 +- src/mesa/glapi/glapi_getproc.c | 2 +- src/mesa/glapi/glthread.c | 2 +- src/mesa/math/m_debug_clip.c | 8 ++++---- src/mesa/math/m_debug_norm.c | 8 ++++---- src/mesa/math/m_debug_xform.c | 8 ++++---- src/mesa/math/m_matrix.c | 8 ++++---- src/mesa/math/m_translate.c | 6 +++--- src/mesa/math/m_vector.c | 8 ++++---- src/mesa/math/m_xform.h | 4 ++-- src/mesa/shader/arbprogparse.c | 6 +++--- src/mesa/shader/arbprogparse.h | 2 +- src/mesa/shader/arbprogram.c | 12 ++++++------ src/mesa/shader/atifragshader.c | 14 +++++++------- src/mesa/shader/grammar/grammar_mesa.h | 2 +- src/mesa/shader/nvfragparse.c | 8 ++++---- src/mesa/shader/nvprogram.c | 10 +++++----- src/mesa/shader/nvvertparse.c | 8 ++++---- src/mesa/shader/prog_debug.c | 6 +++--- src/mesa/shader/prog_execute.c | 6 +++--- src/mesa/shader/prog_instruction.c | 6 +++--- src/mesa/shader/prog_parameter.c | 6 +++--- src/mesa/shader/prog_parameter.h | 2 +- src/mesa/shader/prog_print.c | 6 +++--- src/mesa/shader/prog_statevars.c | 12 ++++++------ src/mesa/shader/prog_statevars.h | 2 +- src/mesa/shader/program.c | 6 +++--- src/mesa/shader/program.h | 2 +- src/mesa/shader/programopt.c | 4 ++-- src/mesa/shader/shader_api.c | 8 ++++---- src/mesa/shader/shader_api.h | 4 ++-- src/mesa/shader/slang/slang_codegen.h | 2 +- src/mesa/shader/slang/slang_compile.h | 4 ++-- src/mesa/shader/slang/slang_compile_function.c | 2 +- src/mesa/shader/slang/slang_compile_operation.c | 2 +- src/mesa/shader/slang/slang_compile_struct.c | 2 +- src/mesa/shader/slang/slang_compile_variable.c | 2 +- src/mesa/shader/slang/slang_emit.h | 4 ++-- src/mesa/shader/slang/slang_ir.c | 4 ++-- src/mesa/shader/slang/slang_ir.h | 4 ++-- src/mesa/shader/slang/slang_library_noise.c | 2 +- src/mesa/shader/slang/slang_log.c | 2 +- src/mesa/shader/slang/slang_mem.c | 4 ++-- src/mesa/shader/slang/slang_mem.h | 2 +- src/mesa/shader/slang/slang_preprocess.c | 2 +- src/mesa/shader/slang/slang_print.c | 2 +- src/mesa/shader/slang/slang_simplify.c | 6 +++--- src/mesa/shader/slang/slang_storage.c | 2 +- src/mesa/shader/slang/slang_typeinfo.h | 4 ++-- src/mesa/shader/slang/slang_utility.c | 2 +- src/mesa/sources | 1 - src/mesa/sparc/sparc.c | 2 +- src/mesa/state_tracker/st_atom.c | 4 ++-- src/mesa/state_tracker/st_atom_viewport.c | 4 ++-- src/mesa/state_tracker/st_mesa_to_tgsi.h | 2 +- src/mesa/state_tracker/st_texture.c | 2 +- src/mesa/swrast/s_aaline.c | 8 ++++---- src/mesa/swrast/s_aaline.h | 2 +- src/mesa/swrast/s_aatriangle.c | 12 ++++++------ src/mesa/swrast/s_aatriangle.h | 2 +- src/mesa/swrast/s_accum.c | 8 ++++---- src/mesa/swrast/s_accum.h | 2 +- src/mesa/swrast/s_alpha.c | 8 ++++---- src/mesa/swrast/s_alpha.h | 2 +- src/mesa/swrast/s_atifragshader.c | 8 ++++---- src/mesa/swrast/s_bitmap.c | 4 ++-- src/mesa/swrast/s_blend.c | 8 ++++---- src/mesa/swrast/s_blend.h | 2 +- src/mesa/swrast/s_blit.c | 4 ++-- src/mesa/swrast/s_buffers.c | 10 +++++----- src/mesa/swrast/s_context.c | 8 ++++---- src/mesa/swrast/s_copypix.c | 10 +++++----- src/mesa/swrast/s_depth.c | 8 ++++---- src/mesa/swrast/s_depth.h | 2 +- src/mesa/swrast/s_drawpix.c | 8 ++++---- src/mesa/swrast/s_drawpix.h | 2 +- src/mesa/swrast/s_feedback.c | 10 +++++----- src/mesa/swrast/s_feedback.h | 2 +- src/mesa/swrast/s_fog.c | 8 ++++---- src/mesa/swrast/s_fog.h | 2 +- src/mesa/swrast/s_lines.c | 8 ++++---- src/mesa/swrast/s_lines.h | 2 +- src/mesa/swrast/s_logic.c | 8 ++++---- src/mesa/swrast/s_logic.h | 2 +- src/mesa/swrast/s_masking.c | 4 ++-- src/mesa/swrast/s_masking.h | 2 +- src/mesa/swrast/s_points.c | 8 ++++---- src/mesa/swrast/s_points.h | 2 +- src/mesa/swrast/s_readpix.c | 10 +++++----- src/mesa/swrast/s_span.c | 10 +++++----- src/mesa/swrast/s_span.h | 2 +- src/mesa/swrast/s_spantemp.h | 2 +- src/mesa/swrast/s_stencil.c | 6 +++--- src/mesa/swrast/s_stencil.h | 2 +- src/mesa/swrast/s_texcombine.c | 10 +++++----- src/mesa/swrast/s_texcombine.h | 2 +- src/mesa/swrast/s_texfilter.c | 8 ++++---- src/mesa/swrast/s_texfilter.h | 2 +- src/mesa/swrast/s_texstore.c | 10 +++++----- src/mesa/swrast/s_triangle.c | 10 +++++----- src/mesa/swrast/s_triangle.h | 2 +- src/mesa/swrast/s_zoom.c | 8 ++++---- src/mesa/swrast/s_zoom.h | 2 +- src/mesa/swrast_setup/ss_context.c | 6 +++--- src/mesa/swrast_setup/ss_context.h | 2 +- src/mesa/swrast_setup/ss_triangle.c | 8 ++++---- src/mesa/swrast_setup/ss_triangle.h | 2 +- src/mesa/swrast_setup/ss_vb.h | 2 +- src/mesa/tnl/t_draw.c | 12 ++++++------ src/mesa/tnl/t_pipeline.h | 2 +- src/mesa/tnl/t_rasterpos.c | 10 +++++----- src/mesa/tnl/t_vb_cull.c | 12 ++++++------ src/mesa/tnl/t_vb_fog.c | 12 ++++++------ src/mesa/tnl/t_vb_light.c | 10 +++++----- src/mesa/tnl/t_vb_normals.c | 12 ++++++------ src/mesa/tnl/t_vb_points.c | 4 ++-- src/mesa/tnl/t_vb_program.c | 10 +++++----- src/mesa/tnl/t_vb_render.c | 12 ++++++------ src/mesa/tnl/t_vb_texgen.c | 12 ++++++------ src/mesa/tnl/t_vb_texmat.c | 12 ++++++------ src/mesa/tnl/t_vb_vertex.c | 12 ++++++------ src/mesa/tnl/t_vertex.c | 6 +++--- src/mesa/tnl/t_vertex.h | 2 +- src/mesa/tnl/t_vertex_generic.c | 6 +++--- src/mesa/tnl/t_vertex_sse.c | 8 ++++---- src/mesa/tnl/t_vp_build.c | 2 +- src/mesa/tnl/t_vp_build.h | 2 +- src/mesa/vf/vf.c | 6 +++--- src/mesa/vf/vf_generic.c | 8 ++++---- src/mesa/vf/vf_sse.c | 8 ++++---- src/mesa/x86-64/x86-64.c | 4 ++-- src/mesa/x86/3dnow.c | 4 ++-- src/mesa/x86/common_x86.c | 2 +- src/mesa/x86/rtasm/x86sse.c | 2 +- src/mesa/x86/sse.c | 4 ++-- src/mesa/x86/x86.c | 4 ++-- 137 files changed, 369 insertions(+), 371 deletions(-) (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index d2de189f69..2c74dc7dd0 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -10,7 +10,6 @@ if env['platform'] != 'winddk': env.Append(CPPPATH = [ '#/src/mesa', - '#/src/mesa/main', ]) if gcc: diff --git a/src/mesa/glapi/glapi.c b/src/mesa/glapi/glapi.c index c236656d9a..53efd7eef4 100644 --- a/src/mesa/glapi/glapi.c +++ b/src/mesa/glapi/glapi.c @@ -50,7 +50,7 @@ -#include "glheader.h" +#include "main/glheader.h" #include "glapi.h" #include "glapioffsets.h" #include "glapitable.h" diff --git a/src/mesa/glapi/glapi_getproc.c b/src/mesa/glapi/glapi_getproc.c index 3634444c85..6d40b495c7 100644 --- a/src/mesa/glapi/glapi_getproc.c +++ b/src/mesa/glapi/glapi_getproc.c @@ -32,7 +32,7 @@ #include #include -#include "glheader.h" +#include "main/glheader.h" #include "glapi.h" #include "glapioffsets.h" #include "glapitable.h" diff --git a/src/mesa/glapi/glthread.c b/src/mesa/glapi/glthread.c index 4513853f5a..09cc8cfcde 100644 --- a/src/mesa/glapi/glthread.c +++ b/src/mesa/glapi/glthread.c @@ -29,7 +29,7 @@ */ -#include "glheader.h" +#include "main/glheader.h" #include "glthread.h" diff --git a/src/mesa/math/m_debug_clip.c b/src/mesa/math/m_debug_clip.c index ab28818359..460fed4a75 100644 --- a/src/mesa/math/m_debug_clip.c +++ b/src/mesa/math/m_debug_clip.c @@ -25,10 +25,10 @@ * Gareth Hughes */ -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" #include "m_matrix.h" #include "m_xform.h" diff --git a/src/mesa/math/m_debug_norm.c b/src/mesa/math/m_debug_norm.c index 11cae6bba7..89c632e7d5 100644 --- a/src/mesa/math/m_debug_norm.c +++ b/src/mesa/math/m_debug_norm.c @@ -26,10 +26,10 @@ * Gareth Hughes */ -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" #include "m_matrix.h" #include "m_xform.h" diff --git a/src/mesa/math/m_debug_xform.c b/src/mesa/math/m_debug_xform.c index 0b07b4fd68..df8cc066b6 100644 --- a/src/mesa/math/m_debug_xform.c +++ b/src/mesa/math/m_debug_xform.c @@ -26,10 +26,10 @@ * Updated for P6 architecture by Gareth Hughes. */ -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" #include "m_matrix.h" #include "m_xform.h" diff --git a/src/mesa/math/m_matrix.c b/src/mesa/math/m_matrix.c index b4ba1bc2a0..84b4cae4ad 100644 --- a/src/mesa/math/m_matrix.c +++ b/src/mesa/math/m_matrix.c @@ -34,10 +34,10 @@ */ -#include "glheader.h" -#include "imports.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/imports.h" +#include "main/macros.h" +#include "main/imports.h" #include "m_matrix.h" diff --git a/src/mesa/math/m_translate.c b/src/mesa/math/m_translate.c index c7423e9d9d..4a20f45ee4 100644 --- a/src/mesa/math/m_translate.c +++ b/src/mesa/math/m_translate.c @@ -28,9 +28,9 @@ */ -#include "glheader.h" -#include "mtypes.h" /* GLchan hack */ -#include "colormac.h" +#include "main/glheader.h" +#include "main/mtypes.h" /* GLchan hack */ +#include "main/colormac.h" #include "m_translate.h" diff --git a/src/mesa/math/m_vector.c b/src/mesa/math/m_vector.c index 3ad81d468b..c5e2fd1de1 100644 --- a/src/mesa/math/m_vector.c +++ b/src/mesa/math/m_vector.c @@ -28,10 +28,10 @@ */ -#include "glheader.h" -#include "imports.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/imports.h" +#include "main/macros.h" +#include "main/imports.h" #include "m_vector.h" diff --git a/src/mesa/math/m_xform.h b/src/mesa/math/m_xform.h index 99b071a46b..d1b974f043 100644 --- a/src/mesa/math/m_xform.h +++ b/src/mesa/math/m_xform.h @@ -27,8 +27,8 @@ #define _M_XFORM_H -#include "glheader.h" -#include "config.h" +#include "main/glheader.h" +#include "main/config.h" #include "math/m_vector.h" #include "math/m_matrix.h" diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c index ff583352ce..78cc6aa9cc 100644 --- a/src/mesa/shader/arbprogparse.c +++ b/src/mesa/shader/arbprogparse.c @@ -38,9 +38,9 @@ #include "programopt.h" #include "prog_parameter.h" #include "prog_statevars.h" -#include "context.h" -#include "macros.h" -#include "mtypes.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/mtypes.h" #include "prog_instruction.h" diff --git a/src/mesa/shader/arbprogparse.h b/src/mesa/shader/arbprogparse.h index 4574e5cd55..980d39fb9f 100644 --- a/src/mesa/shader/arbprogparse.h +++ b/src/mesa/shader/arbprogparse.h @@ -26,7 +26,7 @@ #ifndef ARBPROGPARSE_H #define ARBPROGPARSE_H -#include "mtypes.h" +#include "main/mtypes.h" extern void _mesa_parse_arb_vertex_program(GLcontext *ctx, GLenum target, diff --git a/src/mesa/shader/arbprogram.c b/src/mesa/shader/arbprogram.c index 81c20a0150..beb5deea50 100644 --- a/src/mesa/shader/arbprogram.c +++ b/src/mesa/shader/arbprogram.c @@ -29,14 +29,14 @@ */ -#include "glheader.h" +#include "main/glheader.h" #include "arbprogram.h" #include "arbprogparse.h" -#include "context.h" -#include "hash.h" -#include "imports.h" -#include "macros.h" -#include "mtypes.h" +#include "main/context.h" +#include "main/hash.h" +#include "main/imports.h" +#include "main/macros.h" +#include "main/mtypes.h" #include "program.h" diff --git a/src/mesa/shader/atifragshader.c b/src/mesa/shader/atifragshader.c index 854c911874..ac087d415c 100644 --- a/src/mesa/shader/atifragshader.c +++ b/src/mesa/shader/atifragshader.c @@ -21,13 +21,13 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "glheader.h" -#include "context.h" -#include "hash.h" -#include "imports.h" -#include "macros.h" -#include "enums.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/hash.h" +#include "main/imports.h" +#include "main/macros.h" +#include "main/enums.h" +#include "main/mtypes.h" #include "atifragshader.h" #define MESA_DEBUG_ATI_FS 0 diff --git a/src/mesa/shader/grammar/grammar_mesa.h b/src/mesa/shader/grammar/grammar_mesa.h index c14033a9d4..6c92c5812d 100644 --- a/src/mesa/shader/grammar/grammar_mesa.h +++ b/src/mesa/shader/grammar/grammar_mesa.h @@ -26,7 +26,7 @@ #define GRAMMAR_MESA_H -#include "imports.h" +#include "main/imports.h" /* NOTE: include Mesa 3-D specific headers here */ diff --git a/src/mesa/shader/nvfragparse.c b/src/mesa/shader/nvfragparse.c index 0f1a1eade4..a2a7a5f3f4 100644 --- a/src/mesa/shader/nvfragparse.c +++ b/src/mesa/shader/nvfragparse.c @@ -37,10 +37,10 @@ * including any use thereof or modifications thereto. */ -#include "glheader.h" -#include "context.h" -#include "imports.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/imports.h" +#include "main/macros.h" #include "prog_parameter.h" #include "prog_instruction.h" #include "nvfragparse.h" diff --git a/src/mesa/shader/nvprogram.c b/src/mesa/shader/nvprogram.c index 409c61cdc1..d656d4b28b 100644 --- a/src/mesa/shader/nvprogram.c +++ b/src/mesa/shader/nvprogram.c @@ -37,11 +37,11 @@ * including any use thereof or modifications thereto. */ -#include "glheader.h" -#include "context.h" -#include "hash.h" -#include "imports.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/hash.h" +#include "main/imports.h" +#include "main/macros.h" #include "prog_parameter.h" #include "prog_instruction.h" #include "nvfragparse.h" diff --git a/src/mesa/shader/nvvertparse.c b/src/mesa/shader/nvvertparse.c index ac96d4a60e..08538c0ee4 100644 --- a/src/mesa/shader/nvvertparse.c +++ b/src/mesa/shader/nvvertparse.c @@ -37,10 +37,10 @@ * including any use thereof or modifications thereto. */ -#include "glheader.h" -#include "context.h" -#include "imports.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/imports.h" +#include "main/macros.h" #include "nvprogram.h" #include "nvvertparse.h" #include "prog_instruction.h" diff --git a/src/mesa/shader/prog_debug.c b/src/mesa/shader/prog_debug.c index 57929fcbca..7bcb2ef734 100644 --- a/src/mesa/shader/prog_debug.c +++ b/src/mesa/shader/prog_debug.c @@ -23,9 +23,9 @@ */ -#include "glheader.h" -#include "context.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/macros.h" #include "nvfragparse.h" #include "nvvertparse.h" #include "program.h" diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c index 4745211c65..5afd9eb153 100644 --- a/src/mesa/shader/prog_execute.c +++ b/src/mesa/shader/prog_execute.c @@ -35,9 +35,9 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" #include "program.h" #include "prog_execute.h" #include "prog_instruction.h" diff --git a/src/mesa/shader/prog_instruction.c b/src/mesa/shader/prog_instruction.c index bea5d0551e..1033496d97 100644 --- a/src/mesa/shader/prog_instruction.c +++ b/src/mesa/shader/prog_instruction.c @@ -23,9 +23,9 @@ */ -#include "glheader.h" -#include "imports.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "prog_instruction.h" diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index d87e8f6e15..e0d2096b30 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -29,9 +29,9 @@ */ -#include "glheader.h" -#include "imports.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/imports.h" +#include "main/macros.h" #include "prog_instruction.h" #include "prog_parameter.h" #include "prog_statevars.h" diff --git a/src/mesa/shader/prog_parameter.h b/src/mesa/shader/prog_parameter.h index dfb8c39ca4..ac5c629fab 100644 --- a/src/mesa/shader/prog_parameter.h +++ b/src/mesa/shader/prog_parameter.h @@ -31,7 +31,7 @@ #ifndef PROG_PARAMETER_H #define PROG_PARAMETER_H -#include "mtypes.h" +#include "main/mtypes.h" #include "prog_statevars.h" diff --git a/src/mesa/shader/prog_print.c b/src/mesa/shader/prog_print.c index 5368ab2187..10c5afec18 100644 --- a/src/mesa/shader/prog_print.c +++ b/src/mesa/shader/prog_print.c @@ -28,9 +28,9 @@ * \author Brian Paul */ -#include "glheader.h" -#include "context.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/imports.h" #include "prog_instruction.h" #include "prog_parameter.h" #include "prog_print.h" diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c index 81bb4122f2..819db25a00 100644 --- a/src/mesa/shader/prog_statevars.c +++ b/src/mesa/shader/prog_statevars.c @@ -29,12 +29,12 @@ */ -#include "glheader.h" -#include "context.h" -#include "hash.h" -#include "imports.h" -#include "macros.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/hash.h" +#include "main/imports.h" +#include "main/macros.h" +#include "main/mtypes.h" #include "prog_statevars.h" #include "prog_parameter.h" diff --git a/src/mesa/shader/prog_statevars.h b/src/mesa/shader/prog_statevars.h index a515fda3aa..0a1c235828 100644 --- a/src/mesa/shader/prog_statevars.h +++ b/src/mesa/shader/prog_statevars.h @@ -25,7 +25,7 @@ #ifndef PROG_STATEVARS_H #define PROG_STATEVARS_H -#include "mtypes.h" +#include "main/mtypes.h" /** diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 13d6df1ce6..02e23aaa3a 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -29,9 +29,9 @@ */ -#include "glheader.h" -#include "context.h" -#include "hash.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/hash.h" #include "program.h" #include "prog_cache.h" #include "prog_parameter.h" diff --git a/src/mesa/shader/program.h b/src/mesa/shader/program.h index 7484961364..f8bd63233e 100644 --- a/src/mesa/shader/program.h +++ b/src/mesa/shader/program.h @@ -40,7 +40,7 @@ #ifndef PROGRAM_H #define PROGRAM_H -#include "mtypes.h" +#include "main/mtypes.h" extern struct gl_program _mesa_DummyProgram; diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index f3511ba00e..d6a3231055 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -31,8 +31,8 @@ */ -#include "glheader.h" -#include "context.h" +#include "main/glheader.h" +#include "main/context.h" #include "prog_parameter.h" #include "prog_statevars.h" #include "program.h" diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index bd3745cfe6..b33dfd6663 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -35,10 +35,10 @@ */ -#include "glheader.h" -#include "context.h" -#include "hash.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/hash.h" +#include "main/macros.h" #include "program.h" #include "prog_parameter.h" #include "prog_print.h" diff --git a/src/mesa/shader/shader_api.h b/src/mesa/shader/shader_api.h index 5521c585b5..e7f1266915 100644 --- a/src/mesa/shader/shader_api.h +++ b/src/mesa/shader/shader_api.h @@ -27,8 +27,8 @@ #define SHADER_API_H -#include "glheader.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/mtypes.h" /** diff --git a/src/mesa/shader/slang/slang_codegen.h b/src/mesa/shader/slang/slang_codegen.h index 821d396162..ba7ca4c142 100644 --- a/src/mesa/shader/slang/slang_codegen.h +++ b/src/mesa/shader/slang/slang_codegen.h @@ -27,7 +27,7 @@ #define SLANG_CODEGEN_H -#include "imports.h" +#include "main/imports.h" #include "slang_compile.h" diff --git a/src/mesa/shader/slang/slang_compile.h b/src/mesa/shader/slang/slang_compile.h index 0f1f820d2e..35e468b44e 100644 --- a/src/mesa/shader/slang/slang_compile.h +++ b/src/mesa/shader/slang/slang_compile.h @@ -25,8 +25,8 @@ #if !defined SLANG_COMPILE_H #define SLANG_COMPILE_H -#include "imports.h" -#include "mtypes.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "slang_typeinfo.h" #include "slang_compile_variable.h" #include "slang_compile_struct.h" diff --git a/src/mesa/shader/slang/slang_compile_function.c b/src/mesa/shader/slang/slang_compile_function.c index 80769b33ae..8405d7f778 100644 --- a/src/mesa/shader/slang/slang_compile_function.c +++ b/src/mesa/shader/slang/slang_compile_function.c @@ -28,7 +28,7 @@ * \author Michal Krol */ -#include "imports.h" +#include "main/imports.h" #include "slang_compile.h" #include "slang_mem.h" diff --git a/src/mesa/shader/slang/slang_compile_operation.c b/src/mesa/shader/slang/slang_compile_operation.c index 4d2fd5b666..c0d469c829 100644 --- a/src/mesa/shader/slang/slang_compile_operation.c +++ b/src/mesa/shader/slang/slang_compile_operation.c @@ -28,7 +28,7 @@ * \author Michal Krol */ -#include "imports.h" +#include "main/imports.h" #include "slang_compile.h" #include "slang_mem.h" diff --git a/src/mesa/shader/slang/slang_compile_struct.c b/src/mesa/shader/slang/slang_compile_struct.c index 96bdb1f491..e6c38730d7 100644 --- a/src/mesa/shader/slang/slang_compile_struct.c +++ b/src/mesa/shader/slang/slang_compile_struct.c @@ -28,7 +28,7 @@ * \author Michal Krol */ -#include "imports.h" +#include "main/imports.h" #include "slang_mem.h" #include "slang_compile.h" diff --git a/src/mesa/shader/slang/slang_compile_variable.c b/src/mesa/shader/slang/slang_compile_variable.c index d53255075f..b26c18e38d 100644 --- a/src/mesa/shader/slang/slang_compile_variable.c +++ b/src/mesa/shader/slang/slang_compile_variable.c @@ -28,7 +28,7 @@ * \author Michal Krol */ -#include "imports.h" +#include "main/imports.h" #include "slang_compile.h" #include "slang_mem.h" diff --git a/src/mesa/shader/slang/slang_emit.h b/src/mesa/shader/slang/slang_emit.h index 153e872d74..4db4bbe562 100644 --- a/src/mesa/shader/slang/slang_emit.h +++ b/src/mesa/shader/slang/slang_emit.h @@ -26,10 +26,10 @@ #define SLANG_EMIT_H -#include "imports.h" +#include "main/imports.h" #include "slang_compile.h" #include "slang_ir.h" -#include "mtypes.h" +#include "main/mtypes.h" extern void diff --git a/src/mesa/shader/slang/slang_ir.c b/src/mesa/shader/slang/slang_ir.c index a414036a36..23d554234e 100644 --- a/src/mesa/shader/slang/slang_ir.c +++ b/src/mesa/shader/slang/slang_ir.c @@ -23,8 +23,8 @@ */ -#include "imports.h" -#include "context.h" +#include "main/imports.h" +#include "main/context.h" #include "slang_ir.h" #include "slang_mem.h" #include "shader/prog_print.h" diff --git a/src/mesa/shader/slang/slang_ir.h b/src/mesa/shader/slang/slang_ir.h index 61ff649e5c..e4697ba3b4 100644 --- a/src/mesa/shader/slang/slang_ir.h +++ b/src/mesa/shader/slang/slang_ir.h @@ -33,10 +33,10 @@ #define SLANG_IR_H -#include "imports.h" +#include "main/imports.h" #include "slang_compile.h" #include "slang_label.h" -#include "mtypes.h" +#include "main/mtypes.h" /** diff --git a/src/mesa/shader/slang/slang_library_noise.c b/src/mesa/shader/slang/slang_library_noise.c index 46075c492c..25a05657cc 100644 --- a/src/mesa/shader/slang/slang_library_noise.c +++ b/src/mesa/shader/slang/slang_library_noise.c @@ -49,7 +49,7 @@ */ -#include "imports.h" +#include "main/imports.h" #include "slang_library_noise.h" #define FASTFLOOR(x) ( ((x)>0) ? ((int)x) : (((int)x)-1) ) diff --git a/src/mesa/shader/slang/slang_log.c b/src/mesa/shader/slang/slang_log.c index c963914fe3..01591ceba5 100644 --- a/src/mesa/shader/slang/slang_log.c +++ b/src/mesa/shader/slang/slang_log.c @@ -22,7 +22,7 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "imports.h" +#include "main/imports.h" #include "slang_log.h" #include "slang_utility.h" diff --git a/src/mesa/shader/slang/slang_mem.c b/src/mesa/shader/slang/slang_mem.c index e1d1e6ba14..21d6bfce88 100644 --- a/src/mesa/shader/slang/slang_mem.c +++ b/src/mesa/shader/slang/slang_mem.c @@ -32,8 +32,8 @@ * \author Brian Paul */ -#include "context.h" -#include "macros.h" +#include "main/context.h" +#include "main/macros.h" #include "slang_mem.h" diff --git a/src/mesa/shader/slang/slang_mem.h b/src/mesa/shader/slang/slang_mem.h index 49885b6c98..b5bfae2479 100644 --- a/src/mesa/shader/slang/slang_mem.h +++ b/src/mesa/shader/slang/slang_mem.h @@ -27,7 +27,7 @@ #define SLANG_MEM_H -#include "imports.h" +#include "main/imports.h" typedef struct slang_mempool_ slang_mempool; diff --git a/src/mesa/shader/slang/slang_preprocess.c b/src/mesa/shader/slang/slang_preprocess.c index d3f412c211..bfe29b6bc9 100644 --- a/src/mesa/shader/slang/slang_preprocess.c +++ b/src/mesa/shader/slang/slang_preprocess.c @@ -28,7 +28,7 @@ * \author Michal Krol */ -#include "imports.h" +#include "main/imports.h" #include "shader/grammar/grammar_mesa.h" #include "slang_preprocess.h" diff --git a/src/mesa/shader/slang/slang_print.c b/src/mesa/shader/slang/slang_print.c index f48762fb11..4422f70159 100644 --- a/src/mesa/shader/slang/slang_print.c +++ b/src/mesa/shader/slang/slang_print.c @@ -4,7 +4,7 @@ */ -#include "imports.h" +#include "main/imports.h" #include "slang_compile.h" #include "slang_print.h" diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c index 9a4b6a88d1..d5997283a8 100644 --- a/src/mesa/shader/slang/slang_simplify.c +++ b/src/mesa/shader/slang/slang_simplify.c @@ -28,9 +28,9 @@ * \author Michal Krol */ -#include "imports.h" -#include "macros.h" -#include "get.h" +#include "main/imports.h" +#include "main/macros.h" +#include "main/get.h" #include "slang_compile.h" #include "slang_codegen.h" #include "slang_simplify.h" diff --git a/src/mesa/shader/slang/slang_storage.c b/src/mesa/shader/slang/slang_storage.c index bc32aa4b02..e8b0fb7747 100644 --- a/src/mesa/shader/slang/slang_storage.c +++ b/src/mesa/shader/slang/slang_storage.c @@ -28,7 +28,7 @@ * \author Michal Krol */ -#include "imports.h" +#include "main/imports.h" #include "slang_storage.h" #include "slang_mem.h" diff --git a/src/mesa/shader/slang/slang_typeinfo.h b/src/mesa/shader/slang/slang_typeinfo.h index 587331e8b1..8a36fc3422 100644 --- a/src/mesa/shader/slang/slang_typeinfo.h +++ b/src/mesa/shader/slang/slang_typeinfo.h @@ -25,8 +25,8 @@ #ifndef SLANG_TYPEINFO_H #define SLANG_TYPEINFO_H 1 -#include "imports.h" -#include "mtypes.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "slang_log.h" #include "slang_utility.h" #include "slang_vartable.h" diff --git a/src/mesa/shader/slang/slang_utility.c b/src/mesa/shader/slang/slang_utility.c index 2a2dc8e54f..3631e32b3c 100644 --- a/src/mesa/shader/slang/slang_utility.c +++ b/src/mesa/shader/slang/slang_utility.c @@ -28,7 +28,7 @@ * \author Michal Krol */ -#include "imports.h" +#include "main/imports.h" #include "slang_utility.h" #include "slang_mem.h" diff --git a/src/mesa/sources b/src/mesa/sources index e6b050c3f2..0e0e10979b 100644 --- a/src/mesa/sources +++ b/src/mesa/sources @@ -342,7 +342,6 @@ COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o) INCLUDE_DIRS = \ -I$(TOP)/include \ -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/main \ -I$(TOP)/src/gallium/include \ -I$(TOP)/src/gallium/drivers \ -I$(TOP)/src/gallium/auxiliary diff --git a/src/mesa/sparc/sparc.c b/src/mesa/sparc/sparc.c index 1b77b0bb7b..84e8ac6723 100644 --- a/src/mesa/sparc/sparc.c +++ b/src/mesa/sparc/sparc.c @@ -31,7 +31,7 @@ #ifdef USE_SPARC_ASM -#include "context.h" +#include "main/context.h" #include "math/m_xform.h" #include "tnl/t_context.h" diff --git a/src/mesa/state_tracker/st_atom.c b/src/mesa/state_tracker/st_atom.c index ecfd117918..fc8587f459 100644 --- a/src/mesa/state_tracker/st_atom.c +++ b/src/mesa/state_tracker/st_atom.c @@ -26,8 +26,8 @@ **************************************************************************/ -#include "glheader.h" -#include "context.h" +#include "main/glheader.h" +#include "main/context.h" #include "pipe/p_defines.h" #include "st_context.h" diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c index 8b9f1abda4..27ec2eb033 100644 --- a/src/mesa/state_tracker/st_atom_viewport.c +++ b/src/mesa/state_tracker/st_atom_viewport.c @@ -26,8 +26,8 @@ **************************************************************************/ -#include "context.h" -#include "colormac.h" +#include "main/context.h" +#include "main/colormac.h" #include "st_context.h" #include "st_atom.h" #include "pipe/p_context.h" diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.h b/src/mesa/state_tracker/st_mesa_to_tgsi.h index 63fc855b53..f17f2eac96 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.h +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.h @@ -29,7 +29,7 @@ #ifndef ST_MESA_TO_TGSI_H #define ST_MESA_TO_TGSI_H -#include "mtypes.h" +#include "main/mtypes.h" #if defined __cplusplus diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 3e5054ecd2..289b78b38b 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -28,7 +28,7 @@ #include "st_context.h" #include "st_format.h" #include "st_texture.h" -#include "enums.h" +#include "main/enums.h" #undef Elements /* fix re-defined macro warning */ diff --git a/src/mesa/swrast/s_aaline.c b/src/mesa/swrast/s_aaline.c index d6a9afb421..ee65a71d7e 100644 --- a/src/mesa/swrast/s_aaline.c +++ b/src/mesa/swrast/s_aaline.c @@ -23,14 +23,14 @@ */ -#include "glheader.h" -#include "imports.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/imports.h" +#include "main/macros.h" #include "swrast/s_aaline.h" #include "swrast/s_context.h" #include "swrast/s_span.h" #include "swrast/swrast.h" -#include "mtypes.h" +#include "main/mtypes.h" #define SUB_PIXEL 4 diff --git a/src/mesa/swrast/s_aaline.h b/src/mesa/swrast/s_aaline.h index 41e7e5fd4d..9fb4959f91 100644 --- a/src/mesa/swrast/s_aaline.h +++ b/src/mesa/swrast/s_aaline.h @@ -28,7 +28,7 @@ #define S_AALINE_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_aatriangle.c b/src/mesa/swrast/s_aatriangle.c index 66891f9fec..078f16aea0 100644 --- a/src/mesa/swrast/s_aatriangle.c +++ b/src/mesa/swrast/s_aatriangle.c @@ -28,12 +28,12 @@ */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" #include "s_aatriangle.h" #include "s_context.h" #include "s_span.h" diff --git a/src/mesa/swrast/s_aatriangle.h b/src/mesa/swrast/s_aatriangle.h index ebb828eb19..734d420b62 100644 --- a/src/mesa/swrast/s_aatriangle.h +++ b/src/mesa/swrast/s_aatriangle.h @@ -28,7 +28,7 @@ #define S_AATRIANGLE_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_accum.c b/src/mesa/swrast/s_accum.c index 3c45dee399..13a42fdf53 100644 --- a/src/mesa/swrast/s_accum.c +++ b/src/mesa/swrast/s_accum.c @@ -23,10 +23,10 @@ */ -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" #include "fbobject.h" #include "s_accum.h" diff --git a/src/mesa/swrast/s_accum.h b/src/mesa/swrast/s_accum.h index 97d2bef4c3..42e38cf02b 100644 --- a/src/mesa/swrast/s_accum.h +++ b/src/mesa/swrast/s_accum.h @@ -27,7 +27,7 @@ #define S_ACCUM_H -#include "mtypes.h" +#include "main/mtypes.h" extern void diff --git a/src/mesa/swrast/s_alpha.c b/src/mesa/swrast/s_alpha.c index 3c55d3e9e3..5761bb00b4 100644 --- a/src/mesa/swrast/s_alpha.c +++ b/src/mesa/swrast/s_alpha.c @@ -27,10 +27,10 @@ * \brief Functions to apply alpha test. */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" +#include "main/macros.h" #include "s_alpha.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_alpha.h b/src/mesa/swrast/s_alpha.h index a85ef8a83a..92cb01b18a 100644 --- a/src/mesa/swrast/s_alpha.h +++ b/src/mesa/swrast/s_alpha.h @@ -28,7 +28,7 @@ #define S_ALPHA_H -#include "mtypes.h" +#include "main/mtypes.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_atifragshader.c b/src/mesa/swrast/s_atifragshader.c index 55ec757ee0..9fa352c36b 100644 --- a/src/mesa/swrast/s_atifragshader.c +++ b/src/mesa/swrast/s_atifragshader.c @@ -20,10 +20,10 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" #include "shader/program.h" #include "shader/atifragshader.h" #include "swrast/s_atifragshader.h" diff --git a/src/mesa/swrast/s_bitmap.c b/src/mesa/swrast/s_bitmap.c index f3dda12e25..2308acae1c 100644 --- a/src/mesa/swrast/s_bitmap.c +++ b/src/mesa/swrast/s_bitmap.c @@ -28,10 +28,10 @@ * \author Brian Paul */ -#include "glheader.h" +#include "main/glheader.h" #include "bufferobj.h" #include "image.h" -#include "macros.h" +#include "main/macros.h" #include "pixel.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_blend.c b/src/mesa/swrast/s_blend.c index 82c5e84294..7fd8945354 100644 --- a/src/mesa/swrast/s_blend.c +++ b/src/mesa/swrast/s_blend.c @@ -35,10 +35,10 @@ -#include "glheader.h" -#include "context.h" -#include "colormac.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" +#include "main/macros.h" #include "s_blend.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_blend.h b/src/mesa/swrast/s_blend.h index 0b8cbed1a0..c8abecc099 100644 --- a/src/mesa/swrast/s_blend.h +++ b/src/mesa/swrast/s_blend.h @@ -27,7 +27,7 @@ #define S_BLEND_H -#include "mtypes.h" +#include "main/mtypes.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_blit.c b/src/mesa/swrast/s_blit.c index 5aec4aad03..0e12dfa1da 100644 --- a/src/mesa/swrast/s_blit.c +++ b/src/mesa/swrast/s_blit.c @@ -23,8 +23,8 @@ */ -#include "glheader.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/macros.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_buffers.c b/src/mesa/swrast/s_buffers.c index 90a56284c5..a70f474587 100644 --- a/src/mesa/swrast/s_buffers.c +++ b/src/mesa/swrast/s_buffers.c @@ -24,11 +24,11 @@ /** XXX This file should be named s_clear.c */ -#include "glheader.h" -#include "colormac.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "s_accum.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_context.c b/src/mesa/swrast/s_context.c index f43fa61f8a..4d9b956f85 100644 --- a/src/mesa/swrast/s_context.c +++ b/src/mesa/swrast/s_context.c @@ -26,11 +26,11 @@ * Brian Paul */ -#include "imports.h" +#include "main/imports.h" #include "bufferobj.h" -#include "context.h" -#include "colormac.h" -#include "mtypes.h" +#include "main/context.h" +#include "main/colormac.h" +#include "main/mtypes.h" #include "teximage.h" #include "swrast.h" #include "shader/prog_statevars.h" diff --git a/src/mesa/swrast/s_copypix.c b/src/mesa/swrast/s_copypix.c index bbe1081860..7385a9942c 100644 --- a/src/mesa/swrast/s_copypix.c +++ b/src/mesa/swrast/s_copypix.c @@ -23,14 +23,14 @@ */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" #include "convolve.h" #include "histogram.h" #include "image.h" -#include "macros.h" -#include "imports.h" +#include "main/macros.h" +#include "main/imports.h" #include "pixel.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_depth.c b/src/mesa/swrast/s_depth.c index 408174c990..a9d3e9d98e 100644 --- a/src/mesa/swrast/s_depth.c +++ b/src/mesa/swrast/s_depth.c @@ -23,10 +23,10 @@ */ -#include "glheader.h" -#include "context.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" #include "fbobject.h" #include "s_depth.h" diff --git a/src/mesa/swrast/s_depth.h b/src/mesa/swrast/s_depth.h index d1ed050efd..484cc73f49 100644 --- a/src/mesa/swrast/s_depth.h +++ b/src/mesa/swrast/s_depth.h @@ -27,7 +27,7 @@ #define S_DEPTH_H -#include "mtypes.h" +#include "main/mtypes.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c index cbf6617058..d1120d2cee 100644 --- a/src/mesa/swrast/s_drawpix.c +++ b/src/mesa/swrast/s_drawpix.c @@ -23,13 +23,13 @@ */ -#include "glheader.h" +#include "main/glheader.h" #include "bufferobj.h" -#include "context.h" +#include "main/context.h" #include "convolve.h" #include "image.h" -#include "macros.h" -#include "imports.h" +#include "main/macros.h" +#include "main/imports.h" #include "pixel.h" #include "state.h" diff --git a/src/mesa/swrast/s_drawpix.h b/src/mesa/swrast/s_drawpix.h index 66067115dd..7882a79966 100644 --- a/src/mesa/swrast/s_drawpix.h +++ b/src/mesa/swrast/s_drawpix.h @@ -28,7 +28,7 @@ #define S_DRAWPIXELS_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" /* XXX kill this header? */ diff --git a/src/mesa/swrast/s_feedback.c b/src/mesa/swrast/s_feedback.c index 07b7409ab5..31cea8e41f 100644 --- a/src/mesa/swrast/s_feedback.c +++ b/src/mesa/swrast/s_feedback.c @@ -22,12 +22,12 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "enums.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/enums.h" #include "feedback.h" -#include "macros.h" +#include "main/macros.h" #include "s_context.h" #include "s_feedback.h" diff --git a/src/mesa/swrast/s_feedback.h b/src/mesa/swrast/s_feedback.h index 73f45c10be..6484f1dc75 100644 --- a/src/mesa/swrast/s_feedback.h +++ b/src/mesa/swrast/s_feedback.h @@ -28,7 +28,7 @@ #define S_FEEDBACK_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_fog.c b/src/mesa/swrast/s_fog.c index 7b143f6e5b..b9ba265db6 100644 --- a/src/mesa/swrast/s_fog.c +++ b/src/mesa/swrast/s_fog.c @@ -23,10 +23,10 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" #include "s_context.h" #include "s_fog.h" diff --git a/src/mesa/swrast/s_fog.h b/src/mesa/swrast/s_fog.h index 9639bee2cc..2346dd1734 100644 --- a/src/mesa/swrast/s_fog.h +++ b/src/mesa/swrast/s_fog.h @@ -28,7 +28,7 @@ #define S_FOG_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_lines.c b/src/mesa/swrast/s_lines.c index 3de438760b..23cb9b57ef 100644 --- a/src/mesa/swrast/s_lines.c +++ b/src/mesa/swrast/s_lines.c @@ -23,10 +23,10 @@ */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" +#include "main/macros.h" #include "s_aaline.h" #include "s_context.h" #include "s_depth.h" diff --git a/src/mesa/swrast/s_lines.h b/src/mesa/swrast/s_lines.h index 5372b99b91..6c629ca2d4 100644 --- a/src/mesa/swrast/s_lines.h +++ b/src/mesa/swrast/s_lines.h @@ -27,7 +27,7 @@ #ifndef S_LINES_H #define S_LINES_H -#include "mtypes.h" +#include "main/mtypes.h" void _swrast_choose_line( GLcontext *ctx ); diff --git a/src/mesa/swrast/s_logic.c b/src/mesa/swrast/s_logic.c index 0af9063968..f0274b4c0b 100644 --- a/src/mesa/swrast/s_logic.c +++ b/src/mesa/swrast/s_logic.c @@ -23,10 +23,10 @@ */ -#include "glheader.h" -#include "context.h" -#include "imports.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/imports.h" +#include "main/macros.h" #include "s_context.h" #include "s_logic.h" diff --git a/src/mesa/swrast/s_logic.h b/src/mesa/swrast/s_logic.h index 0bc2c3f8a8..04ef00bb99 100644 --- a/src/mesa/swrast/s_logic.h +++ b/src/mesa/swrast/s_logic.h @@ -27,7 +27,7 @@ #define S_LOGIC_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_masking.c b/src/mesa/swrast/s_masking.c index a69720e83a..df779b0739 100644 --- a/src/mesa/swrast/s_masking.c +++ b/src/mesa/swrast/s_masking.c @@ -28,8 +28,8 @@ */ -#include "glheader.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/macros.h" #include "s_context.h" #include "s_masking.h" diff --git a/src/mesa/swrast/s_masking.h b/src/mesa/swrast/s_masking.h index 0596cb3f45..688c07c7ae 100644 --- a/src/mesa/swrast/s_masking.h +++ b/src/mesa/swrast/s_masking.h @@ -27,7 +27,7 @@ #define S_MASKING_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_points.c b/src/mesa/swrast/s_points.c index f4b3650210..9f52da980c 100644 --- a/src/mesa/swrast/s_points.c +++ b/src/mesa/swrast/s_points.c @@ -23,10 +23,10 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" #include "texstate.h" #include "s_context.h" #include "s_feedback.h" diff --git a/src/mesa/swrast/s_points.h b/src/mesa/swrast/s_points.h index 40b442e951..3fda115c0d 100644 --- a/src/mesa/swrast/s_points.h +++ b/src/mesa/swrast/s_points.h @@ -27,7 +27,7 @@ #ifndef S_POINTS_H #define S_POINTS_H -#include "mtypes.h" +#include "main/mtypes.h" extern void _swrast_choose_point( GLcontext *ctx ); diff --git a/src/mesa/swrast/s_readpix.c b/src/mesa/swrast/s_readpix.c index 2f155d0b70..6186f92899 100644 --- a/src/mesa/swrast/s_readpix.c +++ b/src/mesa/swrast/s_readpix.c @@ -23,15 +23,15 @@ */ -#include "glheader.h" +#include "main/glheader.h" #include "bufferobj.h" -#include "colormac.h" +#include "main/colormac.h" #include "convolve.h" -#include "context.h" +#include "main/context.h" #include "feedback.h" #include "image.h" -#include "macros.h" -#include "imports.h" +#include "main/macros.h" +#include "main/imports.h" #include "pixel.h" #include "state.h" diff --git a/src/mesa/swrast/s_span.c b/src/mesa/swrast/s_span.c index d9c1d1bec7..457dc4a6ce 100644 --- a/src/mesa/swrast/s_span.c +++ b/src/mesa/swrast/s_span.c @@ -30,11 +30,11 @@ * \author Brian Paul */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" #include "image.h" #include "s_atifragshader.h" diff --git a/src/mesa/swrast/s_span.h b/src/mesa/swrast/s_span.h index 512134db0f..6b814fc8fb 100644 --- a/src/mesa/swrast/s_span.h +++ b/src/mesa/swrast/s_span.h @@ -27,7 +27,7 @@ #define S_SPAN_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_spantemp.h b/src/mesa/swrast/s_spantemp.h index 1eef81eb91..bab2ca7378 100644 --- a/src/mesa/swrast/s_spantemp.h +++ b/src/mesa/swrast/s_spantemp.h @@ -43,7 +43,7 @@ * ignored otherwise. */ -#include "macros.h" +#include "main/macros.h" #ifdef CI_MODE diff --git a/src/mesa/swrast/s_stencil.c b/src/mesa/swrast/s_stencil.c index d0cbdd6917..cdb7e4669c 100644 --- a/src/mesa/swrast/s_stencil.c +++ b/src/mesa/swrast/s_stencil.c @@ -23,9 +23,9 @@ */ -#include "glheader.h" -#include "context.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/imports.h" #include "s_context.h" #include "s_depth.h" diff --git a/src/mesa/swrast/s_stencil.h b/src/mesa/swrast/s_stencil.h index 1fcb538fec..742f8d4c94 100644 --- a/src/mesa/swrast/s_stencil.h +++ b/src/mesa/swrast/s_stencil.h @@ -27,7 +27,7 @@ #define S_STENCIL_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_texcombine.c b/src/mesa/swrast/s_texcombine.c index 4c5e22618e..b7724f7b20 100644 --- a/src/mesa/swrast/s_texcombine.c +++ b/src/mesa/swrast/s_texcombine.c @@ -23,11 +23,11 @@ */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" -#include "imports.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" +#include "main/imports.h" +#include "main/macros.h" #include "image.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_texcombine.h b/src/mesa/swrast/s_texcombine.h index 3bf70e0b86..20cd2bd8ad 100644 --- a/src/mesa/swrast/s_texcombine.h +++ b/src/mesa/swrast/s_texcombine.h @@ -27,7 +27,7 @@ #define S_TEXCOMBINE_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" extern void diff --git a/src/mesa/swrast/s_texfilter.c b/src/mesa/swrast/s_texfilter.c index c2a7512388..bb0fc823e7 100644 --- a/src/mesa/swrast/s_texfilter.c +++ b/src/mesa/swrast/s_texfilter.c @@ -23,10 +23,10 @@ */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" +#include "main/imports.h" #include "texformat.h" #include "s_context.h" diff --git a/src/mesa/swrast/s_texfilter.h b/src/mesa/swrast/s_texfilter.h index e4445e79a0..6267ad17eb 100644 --- a/src/mesa/swrast/s_texfilter.h +++ b/src/mesa/swrast/s_texfilter.h @@ -27,7 +27,7 @@ #define S_TEXFILTER_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_texstore.c b/src/mesa/swrast/s_texstore.c index 3f49b40d9c..f5d081d7a3 100644 --- a/src/mesa/swrast/s_texstore.c +++ b/src/mesa/swrast/s_texstore.c @@ -36,13 +36,13 @@ -#include "glheader.h" -#include "imports.h" -#include "colormac.h" -#include "context.h" +#include "main/glheader.h" +#include "main/imports.h" +#include "main/colormac.h" +#include "main/context.h" #include "convolve.h" #include "image.h" -#include "macros.h" +#include "main/macros.h" #include "mipmap.h" #include "texformat.h" #include "teximage.h" diff --git a/src/mesa/swrast/s_triangle.c b/src/mesa/swrast/s_triangle.c index c255545217..00cff6635f 100644 --- a/src/mesa/swrast/s_triangle.c +++ b/src/mesa/swrast/s_triangle.c @@ -29,11 +29,11 @@ * functions to draw triangles. */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" -#include "imports.h" -#include "macros.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" +#include "main/imports.h" +#include "main/macros.h" #include "texformat.h" #include "s_aatriangle.h" diff --git a/src/mesa/swrast/s_triangle.h b/src/mesa/swrast/s_triangle.h index 0de812500c..c3cadae2d4 100644 --- a/src/mesa/swrast/s_triangle.h +++ b/src/mesa/swrast/s_triangle.h @@ -28,7 +28,7 @@ #define S_TRIANGLES_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast/s_zoom.c b/src/mesa/swrast/s_zoom.c index 9f1a4c6f0a..48b4d1e240 100644 --- a/src/mesa/swrast/s_zoom.c +++ b/src/mesa/swrast/s_zoom.c @@ -22,10 +22,10 @@ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "glheader.h" -#include "macros.h" -#include "imports.h" -#include "colormac.h" +#include "main/glheader.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/colormac.h" #include "s_context.h" #include "s_span.h" diff --git a/src/mesa/swrast/s_zoom.h b/src/mesa/swrast/s_zoom.h index 6ca11ac211..7701515476 100644 --- a/src/mesa/swrast/s_zoom.h +++ b/src/mesa/swrast/s_zoom.h @@ -25,7 +25,7 @@ #ifndef S_ZOOM_H #define S_ZOOM_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast.h" diff --git a/src/mesa/swrast_setup/ss_context.c b/src/mesa/swrast_setup/ss_context.c index a9c7d941e5..fd6935e7be 100644 --- a/src/mesa/swrast_setup/ss_context.c +++ b/src/mesa/swrast_setup/ss_context.c @@ -25,9 +25,9 @@ * Keith Whitwell */ -#include "glheader.h" -#include "imports.h" -#include "colormac.h" +#include "main/glheader.h" +#include "main/imports.h" +#include "main/colormac.h" #include "ss_context.h" #include "ss_triangle.h" #include "swrast_setup.h" diff --git a/src/mesa/swrast_setup/ss_context.h b/src/mesa/swrast_setup/ss_context.h index 11f9ded3ff..1ec293fade 100644 --- a/src/mesa/swrast_setup/ss_context.h +++ b/src/mesa/swrast_setup/ss_context.h @@ -28,7 +28,7 @@ #ifndef SS_CONTEXT_H #define SS_CONTEXT_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast/swrast.h" #include "swrast_setup.h" #include "tnl/t_context.h" diff --git a/src/mesa/swrast_setup/ss_triangle.c b/src/mesa/swrast_setup/ss_triangle.c index b4207f2c64..9fef7a52ec 100644 --- a/src/mesa/swrast_setup/ss_triangle.c +++ b/src/mesa/swrast_setup/ss_triangle.c @@ -25,10 +25,10 @@ * Keith Whitwell */ -#include "glheader.h" -#include "colormac.h" -#include "macros.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/macros.h" +#include "main/mtypes.h" #include "tnl/t_context.h" diff --git a/src/mesa/swrast_setup/ss_triangle.h b/src/mesa/swrast_setup/ss_triangle.h index 78833269e6..863e744607 100644 --- a/src/mesa/swrast_setup/ss_triangle.h +++ b/src/mesa/swrast_setup/ss_triangle.h @@ -29,7 +29,7 @@ #ifndef SS_TRIANGLE_H #define SS_TRIANGLE_H -#include "mtypes.h" +#include "main/mtypes.h" #include "ss_context.h" diff --git a/src/mesa/swrast_setup/ss_vb.h b/src/mesa/swrast_setup/ss_vb.h index 6ea0cb1a70..2ad1f56f39 100644 --- a/src/mesa/swrast_setup/ss_vb.h +++ b/src/mesa/swrast_setup/ss_vb.h @@ -29,7 +29,7 @@ #ifndef SS_VB_H #define SS_VB_H -#include "mtypes.h" +#include "main/mtypes.h" #include "swrast_setup.h" void _swsetup_vb_init( GLcontext *ctx ); diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c index 5b2b2ae549..a1a46f6b82 100644 --- a/src/mesa/tnl/t_draw.c +++ b/src/mesa/tnl/t_draw.c @@ -26,13 +26,13 @@ * Keith Whitwell */ -#include "glheader.h" -#include "context.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/imports.h" #include "state.h" -#include "mtypes.h" -#include "macros.h" -#include "enums.h" +#include "main/mtypes.h" +#include "main/macros.h" +#include "main/enums.h" #include "t_context.h" #include "t_pipeline.h" diff --git a/src/mesa/tnl/t_pipeline.h b/src/mesa/tnl/t_pipeline.h index 0952854b85..d110010f04 100644 --- a/src/mesa/tnl/t_pipeline.h +++ b/src/mesa/tnl/t_pipeline.h @@ -30,7 +30,7 @@ #ifndef _T_PIPELINE_H_ #define _T_PIPELINE_H_ -#include "mtypes.h" +#include "main/mtypes.h" #include "t_context.h" extern void _tnl_run_pipeline( GLcontext *ctx ); diff --git a/src/mesa/tnl/t_rasterpos.c b/src/mesa/tnl/t_rasterpos.c index be963867c0..dfbe0f1806 100644 --- a/src/mesa/tnl/t_rasterpos.c +++ b/src/mesa/tnl/t_rasterpos.c @@ -23,15 +23,15 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" #include "feedback.h" #include "light.h" -#include "macros.h" +#include "main/macros.h" #include "rastpos.h" #include "simple_list.h" -#include "mtypes.h" +#include "main/mtypes.h" #include "math/m_matrix.h" #include "tnl/tnl.h" diff --git a/src/mesa/tnl/t_vb_cull.c b/src/mesa/tnl/t_vb_cull.c index 21a32e5b1d..712901acf3 100644 --- a/src/mesa/tnl/t_vb_cull.c +++ b/src/mesa/tnl/t_vb_cull.c @@ -26,12 +26,12 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "math/m_xform.h" diff --git a/src/mesa/tnl/t_vb_fog.c b/src/mesa/tnl/t_vb_fog.c index f6518500d8..00c0979f3f 100644 --- a/src/mesa/tnl/t_vb_fog.c +++ b/src/mesa/tnl/t_vb_fog.c @@ -26,12 +26,12 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "math/m_xform.h" diff --git a/src/mesa/tnl/t_vb_light.c b/src/mesa/tnl/t_vb_light.c index 12f2cc7735..ebd3412d20 100644 --- a/src/mesa/tnl/t_vb_light.c +++ b/src/mesa/tnl/t_vb_light.c @@ -24,13 +24,13 @@ -#include "glheader.h" -#include "colormac.h" +#include "main/glheader.h" +#include "main/colormac.h" #include "light.h" -#include "macros.h" -#include "imports.h" +#include "main/macros.h" +#include "main/imports.h" #include "simple_list.h" -#include "mtypes.h" +#include "main/mtypes.h" #include "math/m_translate.h" diff --git a/src/mesa/tnl/t_vb_normals.c b/src/mesa/tnl/t_vb_normals.c index 01fad0cee2..a4821cc1cc 100644 --- a/src/mesa/tnl/t_vb_normals.c +++ b/src/mesa/tnl/t_vb_normals.c @@ -26,12 +26,12 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "math/m_xform.h" diff --git a/src/mesa/tnl/t_vb_points.c b/src/mesa/tnl/t_vb_points.c index 1ac14fedf9..01d055c1dd 100644 --- a/src/mesa/tnl/t_vb_points.c +++ b/src/mesa/tnl/t_vb_points.c @@ -25,8 +25,8 @@ * Brian Paul */ -#include "mtypes.h" -#include "imports.h" +#include "main/mtypes.h" +#include "main/imports.h" #include "t_context.h" #include "t_pipeline.h" diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c index f8e561ac57..c778f5d248 100644 --- a/src/mesa/tnl/t_vb_program.c +++ b/src/mesa/tnl/t_vb_program.c @@ -30,11 +30,11 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" -#include "imports.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" #include "shader/prog_instruction.h" #include "shader/prog_statevars.h" #include "shader/prog_execute.h" diff --git a/src/mesa/tnl/t_vb_render.c b/src/mesa/tnl/t_vb_render.c index c38f0745e1..c1bebc9942 100644 --- a/src/mesa/tnl/t_vb_render.c +++ b/src/mesa/tnl/t_vb_render.c @@ -38,12 +38,12 @@ */ -#include "glheader.h" -#include "context.h" -#include "enums.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/enums.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "t_pipeline.h" diff --git a/src/mesa/tnl/t_vb_texgen.c b/src/mesa/tnl/t_vb_texgen.c index e98ab743c5..14d3876e54 100644 --- a/src/mesa/tnl/t_vb_texgen.c +++ b/src/mesa/tnl/t_vb_texgen.c @@ -35,12 +35,12 @@ * including any use thereof or modifications thereto. */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "math/m_xform.h" diff --git a/src/mesa/tnl/t_vb_texmat.c b/src/mesa/tnl/t_vb_texmat.c index 674d36f8cb..0abe8cc35d 100644 --- a/src/mesa/tnl/t_vb_texmat.c +++ b/src/mesa/tnl/t_vb_texmat.c @@ -26,12 +26,12 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "math/m_xform.h" diff --git a/src/mesa/tnl/t_vb_vertex.c b/src/mesa/tnl/t_vb_vertex.c index 276305b5e6..30aa7c4086 100644 --- a/src/mesa/tnl/t_vb_vertex.c +++ b/src/mesa/tnl/t_vb_vertex.c @@ -26,12 +26,12 @@ */ -#include "glheader.h" -#include "colormac.h" -#include "context.h" -#include "macros.h" -#include "imports.h" -#include "mtypes.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/context.h" +#include "main/macros.h" +#include "main/imports.h" +#include "main/mtypes.h" #include "math/m_xform.h" diff --git a/src/mesa/tnl/t_vertex.c b/src/mesa/tnl/t_vertex.c index a6728c318f..b661524c87 100644 --- a/src/mesa/tnl/t_vertex.c +++ b/src/mesa/tnl/t_vertex.c @@ -25,9 +25,9 @@ * Keith Whitwell */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" #include "t_context.h" #include "t_vertex.h" diff --git a/src/mesa/tnl/t_vertex.h b/src/mesa/tnl/t_vertex.h index fda8f151d3..712311a146 100644 --- a/src/mesa/tnl/t_vertex.h +++ b/src/mesa/tnl/t_vertex.h @@ -28,7 +28,7 @@ #ifndef _TNL_VERTEX_H #define _TNL_VERTEX_H -#include "mtypes.h" +#include "main/mtypes.h" #include "t_context.h" /* New mechanism to specify hardware vertices so that tnl can build diff --git a/src/mesa/tnl/t_vertex_generic.c b/src/mesa/tnl/t_vertex_generic.c index 236a5bedc8..c399423b9e 100644 --- a/src/mesa/tnl/t_vertex_generic.c +++ b/src/mesa/tnl/t_vertex_generic.c @@ -26,9 +26,9 @@ * Keith Whitwell */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" #include "t_context.h" #include "t_vertex.h" #include "simple_list.h" diff --git a/src/mesa/tnl/t_vertex_sse.c b/src/mesa/tnl/t_vertex_sse.c index a180441a5a..96def25206 100644 --- a/src/mesa/tnl/t_vertex_sse.c +++ b/src/mesa/tnl/t_vertex_sse.c @@ -25,13 +25,13 @@ * Keith Whitwell */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" #include "t_context.h" #include "t_vertex.h" #include "simple_list.h" -#include "enums.h" +#include "main/enums.h" #if defined(USE_SSE_ASM) diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c index 249bccb443..b3b63cd3e4 100644 --- a/src/mesa/tnl/t_vp_build.c +++ b/src/mesa/tnl/t_vp_build.c @@ -30,7 +30,7 @@ */ -#include "glheader.h" +#include "main/glheader.h" #include "main/ffvertex_prog.h" #include "t_vp_build.h" diff --git a/src/mesa/tnl/t_vp_build.h b/src/mesa/tnl/t_vp_build.h index efe12e41a4..d6ebc66c04 100644 --- a/src/mesa/tnl/t_vp_build.h +++ b/src/mesa/tnl/t_vp_build.h @@ -27,7 +27,7 @@ #ifndef T_VP_BUILD_H #define T_VP_BUILD_H -#include "mtypes.h" +#include "main/mtypes.h" #define TNL_FIXED_FUNCTION_STATE_FLAGS (_NEW_PROGRAM | \ _NEW_LIGHT | \ diff --git a/src/mesa/vf/vf.c b/src/mesa/vf/vf.c index cb25f2e113..82f3d2b641 100644 --- a/src/mesa/vf/vf.c +++ b/src/mesa/vf/vf.c @@ -25,9 +25,9 @@ * Keith Whitwell */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" #include "vf.h" diff --git a/src/mesa/vf/vf_generic.c b/src/mesa/vf/vf_generic.c index 68d8d0897b..baa00af29a 100644 --- a/src/mesa/vf/vf_generic.c +++ b/src/mesa/vf/vf_generic.c @@ -26,10 +26,10 @@ * Keith Whitwell */ -#include "glheader.h" -#include "context.h" -#include "colormac.h" -#include "simple_list.h" +#include "main/glheader.h" +#include "main/context.h" +#include "main/colormac.h" +#include "main/simple_list.h" #include "vf/vf.h" diff --git a/src/mesa/vf/vf_sse.c b/src/mesa/vf/vf_sse.c index c3a2166578..4d70196ffe 100644 --- a/src/mesa/vf/vf_sse.c +++ b/src/mesa/vf/vf_sse.c @@ -25,10 +25,10 @@ * Keith Whitwell */ -#include "glheader.h" -#include "colormac.h" -#include "simple_list.h" -#include "enums.h" +#include "main/glheader.h" +#include "main/colormac.h" +#include "main/simple_list.h" +#include "main/enums.h" #include "vf/vf.h" diff --git a/src/mesa/x86-64/x86-64.c b/src/mesa/x86-64/x86-64.c index dee09fd648..9ec43c841d 100644 --- a/src/mesa/x86-64/x86-64.c +++ b/src/mesa/x86-64/x86-64.c @@ -30,8 +30,8 @@ #ifdef USE_X86_64_ASM -#include "glheader.h" -#include "context.h" +#include "main/glheader.h" +#include "main/context.h" #include "math/m_xform.h" #include "tnl/t_context.h" #include "x86-64.h" diff --git a/src/mesa/x86/3dnow.c b/src/mesa/x86/3dnow.c index 4122ee4b00..c037a61761 100644 --- a/src/mesa/x86/3dnow.c +++ b/src/mesa/x86/3dnow.c @@ -28,8 +28,8 @@ * Holger Waechtler */ -#include "glheader.h" -#include "context.h" +#include "main/glheader.h" +#include "main/context.h" #include "math/m_xform.h" #include "tnl/t_context.h" diff --git a/src/mesa/x86/common_x86.c b/src/mesa/x86/common_x86.c index 0caa36a5a0..dc80d26fa9 100644 --- a/src/mesa/x86/common_x86.c +++ b/src/mesa/x86/common_x86.c @@ -49,7 +49,7 @@ #endif #include "common_x86_asm.h" -#include "imports.h" +#include "main/imports.h" int _mesa_x86_cpu_features = 0; diff --git a/src/mesa/x86/rtasm/x86sse.c b/src/mesa/x86/rtasm/x86sse.c index 7a91364ed8..5c4bab7331 100644 --- a/src/mesa/x86/rtasm/x86sse.c +++ b/src/mesa/x86/rtasm/x86sse.c @@ -1,7 +1,7 @@ #ifdef USE_X86_ASM #if defined(__i386__) || defined(__386__) -#include "imports.h" +#include "main/imports.h" #include "x86sse.h" #define DISASSEM 0 diff --git a/src/mesa/x86/sse.c b/src/mesa/x86/sse.c index 4b016a1e85..1c185387c6 100644 --- a/src/mesa/x86/sse.c +++ b/src/mesa/x86/sse.c @@ -27,8 +27,8 @@ * Andre Werthmann */ -#include "glheader.h" -#include "context.h" +#include "main/glheader.h" +#include "main/context.h" #include "math/m_xform.h" #include "tnl/t_context.h" diff --git a/src/mesa/x86/x86.c b/src/mesa/x86/x86.c index 82caa42dbd..ce649f66b0 100644 --- a/src/mesa/x86/x86.c +++ b/src/mesa/x86/x86.c @@ -27,8 +27,8 @@ * Intel x86 assembly code by Josh Vanderhoof */ -#include "glheader.h" -#include "context.h" +#include "main/glheader.h" +#include "main/context.h" #include "math/m_xform.h" #include "tnl/t_context.h" -- cgit v1.2.3 From 90a3af7d9d0849afcee0d35ed0621fe24e1008a9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Jul 2008 14:56:54 -0600 Subject: mesa: glsl: only try to link shaders defining main() --- src/mesa/main/mtypes.h | 4 +-- src/mesa/shader/shader_api.c | 6 +---- src/mesa/shader/slang/slang_compile.c | 47 ++++++++++++++++------------------- src/mesa/shader/slang/slang_link.c | 11 ++++---- 4 files changed, 31 insertions(+), 37 deletions(-) (limited to 'src') diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index 88312da7f0..a95f02b889 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -2123,9 +2123,9 @@ struct gl_shader const GLchar *Source; /**< Source code string */ GLboolean CompileStatus; - GLuint NumPrograms; /**< size of Programs[] array */ - struct gl_program **Programs; /**< Post-compile assembly code */ + struct gl_program *Program; /**< Post-compile assembly code */ GLchar *InfoLog; + GLboolean Main; /**< shader defines main() */ }; diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index b33dfd6663..e894b7b211 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -262,15 +262,11 @@ _mesa_new_shader(GLcontext *ctx, GLuint name, GLenum type) void _mesa_free_shader(GLcontext *ctx, struct gl_shader *sh) { - GLuint i; if (sh->Source) _mesa_free((void *) sh->Source); if (sh->InfoLog) _mesa_free(sh->InfoLog); - for (i = 0; i < sh->NumPrograms; i++) - _mesa_reference_program(ctx, &sh->Programs[i], NULL); - if (sh->Programs) - _mesa_free(sh->Programs); + _mesa_reference_program(ctx, &sh->Program, NULL); _mesa_free(sh); } diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index fd5d4afacb..862d0741ff 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -1930,7 +1930,7 @@ parse_invariant(slang_parse_ctx * C, slang_output_ctx * O) static GLboolean parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, - struct gl_program *program) + struct gl_shader *shader) { GET_CURRENT_CONTEXT(ctx); slang_output_ctx o; @@ -1954,7 +1954,7 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, o.structs = &unit->structs; o.vars = &unit->vars; o.global_pool = &unit->object->varpool; - o.program = program; + o.program = shader ? shader->Program : NULL; o.vartable = _slang_new_var_table(maxRegs); _slang_push_var_table(o.vartable); @@ -2007,6 +2007,7 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, _slang_codegen_function(&A, mainFunc); + shader->Main = GL_TRUE; /* this shader defines main() */ } _slang_pop_var_table(o.vartable); @@ -2020,7 +2021,7 @@ compile_binary(const byte * prod, slang_code_unit * unit, GLuint version, slang_unit_type type, slang_info_log * infolog, slang_code_unit * builtin, slang_code_unit * downlink, - struct gl_program *program) + struct gl_shader *shader) { slang_parse_ctx C; @@ -2045,14 +2046,14 @@ compile_binary(const byte * prod, slang_code_unit * unit, } /* parse translation unit */ - return parse_code_unit(&C, unit, program); + return parse_code_unit(&C, unit, shader); } static GLboolean compile_with_grammar(grammar id, const char *source, slang_code_unit * unit, slang_unit_type type, slang_info_log * infolog, slang_code_unit * builtin, - struct gl_program *program) + struct gl_shader *shader) { byte *prod; GLuint size, start, version; @@ -2114,7 +2115,7 @@ compile_with_grammar(grammar id, const char *source, slang_code_unit * unit, /* Syntax is okay - translate it to internal representation. */ if (!compile_binary(prod, unit, version, type, infolog, builtin, &builtin[SLANG_BUILTIN_TOTAL - 1], - program)) { + shader)) { grammar_alloc_free(prod); return GL_FALSE; } @@ -2153,7 +2154,7 @@ static const byte slang_vertex_builtin_gc[] = { static GLboolean compile_object(grammar * id, const char *source, slang_code_object * object, slang_unit_type type, slang_info_log * infolog, - struct gl_program *program) + struct gl_shader *shader) { slang_code_unit *builtins = NULL; GLuint base_version = 110; @@ -2252,7 +2253,7 @@ compile_object(grammar * id, const char *source, slang_code_object * object, /* compile the actual shader - pass-in built-in library for external shader */ return compile_with_grammar(*id, source, &object->unit, type, infolog, - builtins, program); + builtins, shader); } @@ -2261,7 +2262,6 @@ compile_shader(GLcontext *ctx, slang_code_object * object, slang_unit_type type, slang_info_log * infolog, struct gl_shader *shader) { - struct gl_program *program = shader->Programs[0]; GLboolean success; grammar id = 0; @@ -2271,12 +2271,12 @@ compile_shader(GLcontext *ctx, slang_code_object * object, _mesa_printf("************************************\n"); #endif - assert(program); + assert(shader->Program); _slang_code_object_dtr(object); _slang_code_object_ctr(object); - success = compile_object(&id, shader->Source, object, type, infolog, program); + success = compile_object(&id, shader->Source, object, type, infolog, shader); if (id != 0) grammar_destroy(id); if (!success) @@ -2308,21 +2308,18 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) ctx->Shader.MemPool = _slang_new_mempool(1024*1024); - /* XXX temporary hack */ - if (!shader->Programs) { + shader->Main = GL_FALSE; + + if (!shader->Program) { GLenum progTarget; if (shader->Type == GL_VERTEX_SHADER) progTarget = GL_VERTEX_PROGRAM_ARB; else progTarget = GL_FRAGMENT_PROGRAM_ARB; - shader->Programs - = (struct gl_program **) malloc(sizeof(struct gl_program*)); - shader->Programs[0] = ctx->Driver.NewProgram(ctx, progTarget, 1); - shader->NumPrograms = 1; - - shader->Programs[0]->Parameters = _mesa_new_parameter_list(); - shader->Programs[0]->Varying = _mesa_new_parameter_list(); - shader->Programs[0]->Attributes = _mesa_new_parameter_list(); + shader->Program = ctx->Driver.NewProgram(ctx, progTarget, 1); + shader->Program->Parameters = _mesa_new_parameter_list(); + shader->Program->Varying = _mesa_new_parameter_list(); + shader->Program->Attributes = _mesa_new_parameter_list(); } slang_info_log_construct(&info_log); @@ -2355,13 +2352,13 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) /* remove any reads of varying (output) registers */ #if 0 printf("Pre-remove output reads:\n"); - _mesa_print_program(shader->Programs[0]); + _mesa_print_program(shader->Programs); #endif - _mesa_remove_output_reads(shader->Programs[0], PROGRAM_VARYING); - _mesa_remove_output_reads(shader->Programs[0], PROGRAM_OUTPUT); + _mesa_remove_output_reads(shader->Program, PROGRAM_VARYING); + _mesa_remove_output_reads(shader->Program, PROGRAM_OUTPUT); #if 0 printf("Post-remove output reads:\n"); - _mesa_print_program(shader->Programs[0]); + _mesa_print_program(shader->Programs); #endif } diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index e9b9bbc43f..1de561930a 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -401,15 +401,16 @@ _slang_link(GLcontext *ctx, shProg->Varying = _mesa_new_parameter_list(); /** - * Find attached vertex shader, fragment shader + * Find attached vertex, fragment shaders defining main() */ vertProg = NULL; fragProg = NULL; for (i = 0; i < shProg->NumShaders; i++) { - if (shProg->Shaders[i]->Type == GL_VERTEX_SHADER) - vertProg = vertex_program(shProg->Shaders[i]->Programs[0]); - else if (shProg->Shaders[i]->Type == GL_FRAGMENT_SHADER) - fragProg = fragment_program(shProg->Shaders[i]->Programs[0]); + struct gl_shader *shader = shProg->Shaders[i]; + if (shader->Type == GL_VERTEX_SHADER && shader->Main) + vertProg = vertex_program(shader->Program); + else if (shader->Type == GL_FRAGMENT_SHADER && shader->Main) + fragProg = fragment_program(shader->Program); else _mesa_problem(ctx, "unexpected shader target in slang_link()"); } -- cgit v1.2.3 From 0216a5843027a679c58c65ff6445f33a34d0e729 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Jul 2008 14:28:43 -0600 Subject: mesa: don't include Mesa version in GL_SHADING_LANGUAGE_VERSION string --- src/mesa/main/getstring.c | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/main/getstring.c b/src/mesa/main/getstring.c index 973649da0d..ad99de77b9 100644 --- a/src/mesa/main/getstring.c +++ b/src/mesa/main/getstring.c @@ -57,8 +57,10 @@ _mesa_GetString( GLenum name ) static const char *version_2_0 = "2.0 Mesa " MESA_VERSION_STRING; static const char *version_2_1 = "2.1 Mesa " MESA_VERSION_STRING; -#if FEATURE_ARB_shading_language_100 - static const char *sl_version_110 = "1.10 Mesa " MESA_VERSION_STRING; +#if FEATURE_ARB_shading_language_120_foo /* support not complete! */ + static const char *sl_version = "1.20"; +#elif FEATURE_ARB_shading_language_100 + static const char *sl_version = "1.10"; #endif if (!ctx) @@ -147,7 +149,7 @@ _mesa_GetString( GLenum name ) #if FEATURE_ARB_shading_language_100 case GL_SHADING_LANGUAGE_VERSION_ARB: if (ctx->Extensions.ARB_shading_language_100) - return (const GLubyte *) sl_version_110; + return (const GLubyte *) sl_version; goto error; #endif #if FEATURE_NV_fragment_program || FEATURE_ARB_fragment_program || \ -- cgit v1.2.3 From 643228c506bde965c890f3d0604c273fc729bee7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Jul 2008 15:12:42 -0600 Subject: mesa: move extensions->version code into separate function --- src/mesa/main/getstring.c | 138 ++++++++++++++++++++++++---------------------- 1 file changed, 72 insertions(+), 66 deletions(-) (limited to 'src') diff --git a/src/mesa/main/getstring.c b/src/mesa/main/getstring.c index ad99de77b9..89daa0e21d 100644 --- a/src/mesa/main/getstring.c +++ b/src/mesa/main/getstring.c @@ -33,6 +33,76 @@ #include "extensions.h" +/** + * Examine enabled GL extensions to determine GL version. + * \return version string + */ +static const char * +compute_version(const GLcontext *ctx) +{ + static const char *version_1_2 = "1.2 Mesa " MESA_VERSION_STRING; + static const char *version_1_3 = "1.3 Mesa " MESA_VERSION_STRING; + static const char *version_1_4 = "1.4 Mesa " MESA_VERSION_STRING; + static const char *version_1_5 = "1.5 Mesa " MESA_VERSION_STRING; + static const char *version_2_0 = "2.0 Mesa " MESA_VERSION_STRING; + static const char *version_2_1 = "2.1 Mesa " MESA_VERSION_STRING; + + const GLboolean ver_1_3 = (ctx->Extensions.ARB_multisample && + ctx->Extensions.ARB_multitexture && + ctx->Extensions.ARB_texture_border_clamp && + ctx->Extensions.ARB_texture_compression && + ctx->Extensions.ARB_texture_cube_map && + ctx->Extensions.EXT_texture_env_add && + ctx->Extensions.ARB_texture_env_combine && + ctx->Extensions.ARB_texture_env_dot3); + const GLboolean ver_1_4 = (ver_1_3 && + ctx->Extensions.ARB_depth_texture && + ctx->Extensions.ARB_shadow && + ctx->Extensions.ARB_texture_env_crossbar && + ctx->Extensions.ARB_texture_mirrored_repeat && + ctx->Extensions.ARB_window_pos && + ctx->Extensions.EXT_blend_color && + ctx->Extensions.EXT_blend_func_separate && + ctx->Extensions.EXT_blend_minmax && + ctx->Extensions.EXT_blend_subtract && + ctx->Extensions.EXT_fog_coord && + ctx->Extensions.EXT_multi_draw_arrays && + ctx->Extensions.EXT_point_parameters && + ctx->Extensions.EXT_secondary_color && + ctx->Extensions.EXT_stencil_wrap && + ctx->Extensions.EXT_texture_lod_bias && + ctx->Extensions.SGIS_generate_mipmap); + const GLboolean ver_1_5 = (ver_1_4 && + ctx->Extensions.ARB_occlusion_query && + ctx->Extensions.ARB_vertex_buffer_object && + ctx->Extensions.EXT_shadow_funcs); + const GLboolean ver_2_0 = (ver_1_5 && + ctx->Extensions.ARB_draw_buffers && + ctx->Extensions.ARB_point_sprite && + ctx->Extensions.ARB_shader_objects && + ctx->Extensions.ARB_vertex_shader && + ctx->Extensions.ARB_fragment_shader && + ctx->Extensions.ARB_texture_non_power_of_two && + ctx->Extensions.EXT_blend_equation_separate); + const GLboolean ver_2_1 = (ver_2_0 && + ctx->Extensions.ARB_shading_language_120 && + ctx->Extensions.EXT_pixel_buffer_object && + ctx->Extensions.EXT_texture_sRGB); + if (ver_2_1) + return version_2_1; + if (ver_2_0) + return version_2_0; + if (ver_1_5) + return version_1_5; + if (ver_1_4) + return version_1_4; + if (ver_1_3) + return version_1_3; + return version_1_2; +} + + + /** * Query string-valued state. The return value should _not_ be freed by * the caller. @@ -50,12 +120,6 @@ _mesa_GetString( GLenum name ) GET_CURRENT_CONTEXT(ctx); static const char *vendor = "Brian Paul"; static const char *renderer = "Mesa"; - static const char *version_1_2 = "1.2 Mesa " MESA_VERSION_STRING; - static const char *version_1_3 = "1.3 Mesa " MESA_VERSION_STRING; - static const char *version_1_4 = "1.4 Mesa " MESA_VERSION_STRING; - static const char *version_1_5 = "1.5 Mesa " MESA_VERSION_STRING; - static const char *version_2_0 = "2.0 Mesa " MESA_VERSION_STRING; - static const char *version_2_1 = "2.1 Mesa " MESA_VERSION_STRING; #if FEATURE_ARB_shading_language_120_foo /* support not complete! */ static const char *sl_version = "1.20"; @@ -81,67 +145,9 @@ _mesa_GetString( GLenum name ) case GL_VENDOR: return (const GLubyte *) vendor; case GL_RENDERER: - return (const GLubyte *) renderer; + return (const GLubyte *) renderer; case GL_VERSION: - if (ctx->Extensions.ARB_multisample && - ctx->Extensions.ARB_multitexture && - ctx->Extensions.ARB_texture_border_clamp && - ctx->Extensions.ARB_texture_compression && - ctx->Extensions.ARB_texture_cube_map && - ctx->Extensions.EXT_texture_env_add && - ctx->Extensions.ARB_texture_env_combine && - ctx->Extensions.ARB_texture_env_dot3) { - if (ctx->Extensions.ARB_depth_texture && - ctx->Extensions.ARB_shadow && - ctx->Extensions.ARB_texture_env_crossbar && - ctx->Extensions.ARB_texture_mirrored_repeat && - ctx->Extensions.ARB_window_pos && - ctx->Extensions.EXT_blend_color && - ctx->Extensions.EXT_blend_func_separate && - ctx->Extensions.EXT_blend_logic_op && - ctx->Extensions.EXT_blend_minmax && - ctx->Extensions.EXT_blend_subtract && - ctx->Extensions.EXT_fog_coord && - ctx->Extensions.EXT_multi_draw_arrays && - ctx->Extensions.EXT_point_parameters && /*aka ARB*/ - ctx->Extensions.EXT_secondary_color && - ctx->Extensions.EXT_stencil_wrap && - ctx->Extensions.EXT_texture_lod_bias && - ctx->Extensions.SGIS_generate_mipmap) { - if (ctx->Extensions.ARB_occlusion_query && - ctx->Extensions.ARB_vertex_buffer_object && - ctx->Extensions.EXT_shadow_funcs) { - if (ctx->Extensions.ARB_draw_buffers && - ctx->Extensions.ARB_point_sprite && - ctx->Extensions.ARB_shader_objects && - ctx->Extensions.ARB_vertex_shader && - ctx->Extensions.ARB_fragment_shader && - ctx->Extensions.ARB_texture_non_power_of_two && - ctx->Extensions.EXT_blend_equation_separate) { - if (ctx->Extensions.ARB_shading_language_120 && - ctx->Extensions.EXT_pixel_buffer_object && - ctx->Extensions.EXT_texture_sRGB) { - return (const GLubyte *) version_2_1; - } - else { - return (const GLubyte *) version_2_0; - } - } - else { - return (const GLubyte *) version_1_5; - } - } - else { - return (const GLubyte *) version_1_4; - } - } - else { - return (const GLubyte *) version_1_3; - } - } - else { - return (const GLubyte *) version_1_2; - } + return (const GLubyte *) compute_version(ctx); case GL_EXTENSIONS: if (!ctx->Extensions.String) ctx->Extensions.String = _mesa_make_extension_string(ctx); -- cgit v1.2.3 From 7b2ef2b8849bbf400eeed9642c26c140cc7d6beb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Jul 2008 15:49:09 -0600 Subject: mesa: gls: fix broken else clause of conditional break/continue In the following case: for () { if (cond) break; // or continue; else something; } The "something" block didn't get emitted. --- src/mesa/shader/slang/slang_codegen.c | 14 ++++---------- src/mesa/shader/slang/slang_emit.c | 4 +--- 2 files changed, 5 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 40f4105a64..1b69277387 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2194,22 +2194,16 @@ _slang_gen_if(slang_assemble_ctx * A, const slang_operation *oper) cond = _slang_gen_operation(A, &oper->children[0]); cond = new_cond(cond); - if (is_operation_type(&oper->children[1], SLANG_OPER_BREAK)) { + if (is_operation_type(&oper->children[1], SLANG_OPER_BREAK) + && !haveElseClause) { /* Special case: generate a conditional break */ ifBody = new_break_if_true(A->CurLoop, cond); - if (haveElseClause) { - elseBody = _slang_gen_operation(A, &oper->children[2]); - return new_seq(ifBody, elseBody); - } return ifBody; } - else if (is_operation_type(&oper->children[1], SLANG_OPER_CONTINUE)) { + else if (is_operation_type(&oper->children[1], SLANG_OPER_CONTINUE) + && !haveElseClause) { /* Special case: generate a conditional break */ ifBody = new_cont_if_true(A->CurLoop, cond); - if (haveElseClause) { - elseBody = _slang_gen_operation(A, &oper->children[2]); - return new_seq(ifBody, elseBody); - } return ifBody; } else { diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index c28894630c..1f4d962bdf 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -161,9 +161,7 @@ free_temp_storage(slang_var_table *vt, slang_ir_node *n) if (_slang_is_temp(vt, n->Store)) { _slang_free_temp(vt, n->Store); n->Store->Index = -1; - n->Store->Size = -1; - /*_mesa_free(n->Store);*/ /* XXX leak */ - n->Store = NULL; + n->Store = NULL; /* XXX this may not be needed */ } } } -- cgit v1.2.3 From 2acf917f00b570274b58ad7e58688715730253d0 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Fri, 25 Jul 2008 10:41:53 +0200 Subject: mesa: Mark as XXX unresolved warnings on windows. --- src/mesa/main/texstore.c | 5 +++++ src/mesa/shader/shader_api.c | 4 ++++ 2 files changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index edb1ae965d..a0ea245507 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -408,6 +408,11 @@ make_temp_float_image(GLcontext *ctx, GLuint dims, const GLint logComponents = _mesa_components_in_format(logicalBaseFormat); const GLfloat *src = convImage; + + /* XXX: Both `convWidth' and `convHeight' are uninitialised -- windows compiler + * will issue warnings for the following line. + * Presumably this happens when `FEATURE_convolve' is defined to `0'. + */ GLfloat *dst = tempImage + img * (convWidth * convHeight * 4); for (row = 0; row < convHeight; row++) { _mesa_pack_rgba_span_float(ctx, convWidth, diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index e894b7b211..6423c41c12 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -539,6 +539,10 @@ _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, oldIndex = -1; } + /* XXX: `datatype' is uninitialised at this point -- windows compiler + * will issue a warning for the following line. + */ + /* this will replace the current value if it's already in the list */ i = _mesa_add_attribute(shProg->Attributes, name, size, datatype, index); if (i < 0) { -- cgit v1.2.3 From 03de81aea1c02d8b29784950a456e43b4fc9d538 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 24 Jul 2008 17:49:33 -0600 Subject: mesa: glsl: additional error detection Plus begin some fixes for vec/matrix constructors. --- src/mesa/shader/slang/slang_codegen.c | 60 ++++++++++++++++++++++++++++++---- src/mesa/shader/slang/slang_emit.c | 6 +++- src/mesa/shader/slang/slang_simplify.c | 8 +++-- 3 files changed, 64 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 1b69277387..2f9e147b87 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1702,7 +1702,7 @@ print_funcs(struct slang_function_scope_ *scope, const char *name) /** - * Return first function in the scope that has the given name. + * Find a function of the given name, taking 'numArgs' arguments. * This is the function we'll try to call when there is no exact match * between function parameters and call arguments. * @@ -1710,16 +1710,18 @@ print_funcs(struct slang_function_scope_ *scope, const char *name) * all of them... */ static slang_function * -_slang_first_function(struct slang_function_scope_ *scope, const char *name) +_slang_find_function_by_argc(struct slang_function_scope_ *scope, + const char *name, int numArgs) { GLuint i; for (i = 0; i < scope->num_functions; i++) { slang_function *f = &scope->functions[i]; - if (strcmp(name, (char*) f->header.a_name) == 0) + if (strcmp(name, (char*) f->header.a_name) == 0 + /*&& numArgs == f->param_count*/) return f; } if (scope->outer_scope) - return _slang_first_function(scope->outer_scope, name); + return _slang_find_function_by_argc(scope->outer_scope, name, numArgs); return NULL; } @@ -1918,7 +1920,8 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name, /* A function with exactly the right parameters/types was not found. * Try adapting the parameters. */ - fun = _slang_first_function(A->space.funcs, name); + int numArgs = oper->num_children; + fun = _slang_find_function_by_argc(A->space.funcs, name, numArgs); if (!fun || !_slang_adapt_call(oper, fun, &A->space, A->atoms, A->log)) { slang_info_log_error(A->log, "Function '%s' not found (check argument types)", name); return NULL; @@ -2313,6 +2316,11 @@ _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) n->Store->File = PROGRAM_TEMPORARY; n->Store->Size = _slang_sizeof_type_specifier(&n->Var->type.specifier); + if (n->Store->Size <= 0) { + slang_info_log_error(A->log, "invalid declaration for '%s'", + (char*) var->a_name); + return NULL; + } #if 0 printf("%s var %p %s store=%p index=%d size=%d\n", __FUNCTION__, (void *) var, (char *) var->a_name, @@ -2548,8 +2556,16 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) assert(oper->num_children == 0 || oper->num_children == 1); v = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE); + /*printf("Declare %s at %p\n", varName, v);*/ assert(v); +#if 0 + if (v->declared) { + slang_info_log_error(A->log, "variable '%s' redeclared", varName); + return NULL; + } +#endif + varDecl = _slang_gen_var_decl(A, v); if (oper->num_children > 0) { @@ -2706,6 +2722,11 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper) lhs = _slang_gen_operation(A, &oper->children[0]); if (lhs) { + if (!lhs->Store) { + slang_info_log_error(A->log, + "invalid left hand side for assignment"); + return NULL; + } if (!(lhs->Store->File == PROGRAM_OUTPUT || lhs->Store->File == PROGRAM_TEMPORARY || (lhs->Store->File == PROGRAM_VARYING && @@ -2759,6 +2780,7 @@ _slang_gen_struct_field(slang_assemble_ctx * A, slang_operation *oper) GLuint swizzle; if (!_slang_is_swizzle((char *) oper->a_id, rows, &swz)) { slang_info_log_error(A->log, "Bad swizzle"); + return NULL; } swizzle = MAKE_SWIZZLE4(swz.swizzle[0], swz.swizzle[1], @@ -2807,10 +2829,14 @@ _slang_gen_struct_field(slang_assemble_ctx * A, slang_operation *oper) fieldOffset = _slang_field_offset(&ti.spec, oper->a_id); if (fieldSize == 0 || fieldOffset < 0) { + const char *structName; + if (ti.spec._struct) + structName = (char *) ti.spec._struct->a_name; + else + structName = "unknown"; slang_info_log_error(A->log, "\"%s\" is not a member of struct \"%s\"", - (char *) oper->a_id, - (char *) ti.spec._struct->a_name); + (char *) oper->a_id, structName); return NULL; } assert(fieldSize >= 0); @@ -2945,6 +2971,23 @@ _slang_gen_array_element(slang_assemble_ctx * A, slang_operation *oper) } +#if 0 +static void +print_vars(slang_variable_scope *s) +{ + int i; + printf("vars: "); + for (i = 0; i < s->num_variables; i++) { + printf("%s %d, \n", + (char*) s->variables[i]->a_name, + s->variables[i]->declared); + } + + printf("\n"); +} +#endif + + /** * Generate IR tree for a slang_operation (AST node) */ @@ -2964,6 +3007,8 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) _slang_pop_var_table(A->vartable); + /*print_vars(oper->locals);*/ + if (n) n = new_node1(IR_SCOPE, n); return n; @@ -3006,6 +3051,7 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) } } #endif + /*print_vars(oper->locals);*/ return tree; } else { diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 1f4d962bdf..9bff055ad5 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -625,7 +625,11 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) emit(emitInfo, n->Children[0]); emit(emitInfo, n->Children[1]); - assert(n->Children[0]->Store->Size == n->Children[1]->Store->Size); + if (n->Children[0]->Store->Size != n->Children[1]->Store->Size) { + slang_info_log_error(emitInfo->log, "invalid operands to == or !="); + return NULL; + } + size = n->Children[0]->Store->Size; if (size == 1) { diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c index d5997283a8..c34bef2164 100644 --- a/src/mesa/shader/slang/slang_simplify.c +++ b/src/mesa/shader/slang/slang_simplify.c @@ -331,8 +331,9 @@ _slang_adapt_call(slang_operation *callOper, const slang_function *fun, int i; int dbg = 0; - if (dbg) printf("Adapt %d args to %d parameters\n", - callOper->num_children, numParams); + if (dbg) + printf("Adapt %d args to %d parameters for %s\n", + callOper->num_children, numParams, (char *) fun->header.a_name); /* Only try adapting for constructors */ if (fun->kind != SLANG_FUNC_CONSTRUCTOR) @@ -441,6 +442,9 @@ _slang_adapt_call(slang_operation *callOper, const slang_function *fun, slang_type_specifier_type_to_string(paramVar->type.specifier.type); slang_operation *child = slang_operation_new(1); + if (dbg) + printf("Need to adapt types of arg %d\n", i); + slang_operation_copy(child, &callOper->children[i]); child->locals->outer_scope = callOper->children[i].locals; -- cgit v1.2.3 From 72809f377397390fdd110d6b15e8a9a64ada05e6 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Jul 2008 08:34:54 -0600 Subject: mesa: fix issues causing warnings on Windows --- src/mesa/main/texstore.c | 11 ++++------- src/mesa/shader/shader_api.c | 7 ++----- 2 files changed, 6 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index a0ea245507..97fdfd39b2 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -380,12 +380,14 @@ make_temp_float_image(GLcontext *ctx, GLuint dims, dst += srcWidth * 4; } + /* size after optional convolution */ + convWidth = srcWidth; + convHeight = srcHeight; + #if FEATURE_convolve /* do convolution */ { GLfloat *src = tempImage + img * (srcWidth * srcHeight * 4); - convWidth = srcWidth; - convHeight = srcHeight; if (dims == 1) { ASSERT(ctx->Pixel.Convolution1DEnabled); _mesa_convolve_1d_image(ctx, &convWidth, src, convImage); @@ -408,11 +410,6 @@ make_temp_float_image(GLcontext *ctx, GLuint dims, const GLint logComponents = _mesa_components_in_format(logicalBaseFormat); const GLfloat *src = convImage; - - /* XXX: Both `convWidth' and `convHeight' are uninitialised -- windows compiler - * will issue warnings for the following line. - * Presumably this happens when `FEATURE_convolve' is defined to `0'. - */ GLfloat *dst = tempImage + img * (convWidth * convHeight * 4); for (row = 0; row < convHeight; row++) { _mesa_pack_rgba_span_float(ctx, convWidth, diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 6423c41c12..5c18e55dac 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -509,7 +509,7 @@ _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, struct gl_shader_program *shProg; const GLint size = -1; /* unknown size */ GLint i, oldIndex; - GLenum datatype; + GLenum datatype = GL_FLOAT_VEC4; shProg = _mesa_lookup_shader_program_err(ctx, program, "glBindAttribLocation"); @@ -539,14 +539,11 @@ _mesa_bind_attrib_location(GLcontext *ctx, GLuint program, GLuint index, oldIndex = -1; } - /* XXX: `datatype' is uninitialised at this point -- windows compiler - * will issue a warning for the following line. - */ - /* this will replace the current value if it's already in the list */ i = _mesa_add_attribute(shProg->Attributes, name, size, datatype, index); if (i < 0) { _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBindAttribLocation"); + return; } if (shProg->VertexProgram && oldIndex >= 0 && oldIndex != index) { -- cgit v1.2.3 From c72bc3f3722e17b30c91bb8b9d57cd0ecfd98e46 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Jul 2008 14:48:14 -0600 Subject: mesa: glsl: remove unneeded operators --- src/mesa/shader/slang/library/slang_120_core.gc | 50 ------------------------- 1 file changed, 50 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_120_core.gc b/src/mesa/shader/slang/library/slang_120_core.gc index 15ca3a81ad..04c5ec2ec5 100644 --- a/src/mesa/shader/slang/library/slang_120_core.gc +++ b/src/mesa/shader/slang/library/slang_120_core.gc @@ -1976,53 +1976,3 @@ void __operator ++ (inout mat4x3 m) { ++m[3]; } - -mat2x3 __operator -- (inout mat2x3 m, const int) { - return mat2x3 (m[0]--, m[1]--); -} - -mat2x4 __operator -- (inout mat2x4 m, const int) { - return mat2x4 (m[0]--, m[1]--); -} - -mat3x2 __operator -- (inout mat3x2 m, const int) { - return mat3x2 (m[0]--, m[1]--, m[2]--); -} - -mat3x4 __operator -- (inout mat3x4 m, const int) { - return mat3x4 (m[0]--, m[1]--, m[2]--); -} - -mat4x2 __operator -- (inout mat4x2 m, const int) { - return mat4x2 (m[0]--, m[1]--, m[2]--, m[3]--); -} - -mat4x3 __operator -- (inout mat4x3 m, const int) { - return mat4x3 (m[0]--, m[1]--, m[2]--, m[3]--); -} - - -mat2x3 __operator ++ (inout mat2x3 m, const int) { - return mat2x3 (m[0]++, m[1]++); -} - -mat2x4 __operator ++ (inout mat2x4 m, const int) { - return mat2x4 (m[0]++, m[1]++); -} - -mat3x2 __operator ++ (inout mat3x2 m, const int) { - return mat3x2 (m[0]++, m[1]++, m[2]++); -} - -mat3x4 __operator ++ (inout mat3x4 m, const int) { - return mat3x4 (m[0]++, m[1]++, m[2]++); -} - -mat4x2 __operator ++ (inout mat4x2 m, const int) { - return mat4x2 (m[0]++, m[1]++, m[2]++, m[3]++); -} - -mat4x3 __operator ++ (inout mat4x3 m, const int) { - return mat4x3 (m[0]++, m[1]++, m[2]++, m[3]++); -} - -- cgit v1.2.3 From 9d9076af5c7af6cce9ba8216c62eabbc629dda43 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Jul 2008 14:48:36 -0600 Subject: mesa: glsl: re-order some constructors --- src/mesa/shader/slang/library/slang_core.gc | 24 ++++++++++++------------ 1 file changed, 12 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 2c3a5a6c45..1f57997051 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -110,6 +110,16 @@ int __constructor(const float f) __asm float_to_int __retVal, f; } +int __constructor(const bool b) +{ + __retVal = b; +} + +int __constructor(const int i) +{ + __retVal = i; +} + bool __constructor(const int i) { const float zero = 0.0; @@ -122,12 +132,7 @@ bool __constructor(const float f) __asm vec4_sne __retVal, f, zero; } -int __constructor(const bool b) -{ - __retVal = b; -} - -float __constructor(const bool b) +bool __constructor(const bool b) { __retVal = b; } @@ -137,16 +142,11 @@ float __constructor(const int i) __asm int_to_float __retVal, i; } -bool __constructor(const bool b) +float __constructor(const bool b) { __retVal = b; } -int __constructor(const int i) -{ - __retVal = i; -} - float __constructor(const float f) { __retVal = f; -- cgit v1.2.3 From 4b3fb99b502a803bec852a35e6d50c3ff49975e3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Jul 2008 14:48:49 -0600 Subject: mesa: regenerated files --- src/mesa/shader/slang/library/slang_120_core_gc.h | 22 +--------------------- src/mesa/shader/slang/library/slang_core_gc.h | 18 +++++++++--------- 2 files changed, 10 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_120_core_gc.h b/src/mesa/shader/slang/library/slang_120_core_gc.h index b7ef5ed6c6..1bd781ffe5 100644 --- a/src/mesa/shader/slang/library/slang_120_core_gc.h +++ b/src/mesa/shader/slang/library/slang_120_core_gc.h @@ -707,24 +707,4 @@ 109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,0,1,0,0,2,24,1,0,2,29,109,0,0,0,1,9,18,109, 0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10, 51,0,57,51,0,0,1,0,0,2,24,1,0,2,31,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0, -57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10,51,0,57,51,0,0,1,0,26,2,25,1,0,2,26,109,0,0, -1,1,0,5,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,61,0,18,109,0,16,10,49,0,57,61,0, -0,0,0,1,0,28,2,25,1,0,2,28,109,0,0,1,1,0,5,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0, -57,61,0,18,109,0,16,10,49,0,57,61,0,0,0,0,1,0,27,2,25,1,0,2,27,109,0,0,1,1,0,5,0,0,0,1,8,58,109,97, -116,51,120,50,0,18,109,0,16,8,48,0,57,61,0,18,109,0,16,10,49,0,57,61,0,18,109,0,16,10,50,0,57,61,0, -0,0,0,1,0,30,2,25,1,0,2,30,109,0,0,1,1,0,5,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0, -57,61,0,18,109,0,16,10,49,0,57,61,0,18,109,0,16,10,50,0,57,61,0,0,0,0,1,0,29,2,25,1,0,2,29,109,0,0, -1,1,0,5,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,61,0,18,109,0,16,10,49,0,57,61,0, -18,109,0,16,10,50,0,57,61,0,18,109,0,16,10,51,0,57,61,0,0,0,0,1,0,31,2,25,1,0,2,31,109,0,0,1,1,0,5, -0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,61,0,18,109,0,16,10,49,0,57,61,0,18,109, -0,16,10,50,0,57,61,0,18,109,0,16,10,51,0,57,61,0,0,0,0,1,0,26,2,24,1,0,2,26,109,0,0,1,1,0,5,0,0,0, -1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,60,0,18,109,0,16,10,49,0,57,60,0,0,0,0,1,0,28, -2,24,1,0,2,28,109,0,0,1,1,0,5,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,60,0,18, -109,0,16,10,49,0,57,60,0,0,0,0,1,0,27,2,24,1,0,2,27,109,0,0,1,1,0,5,0,0,0,1,8,58,109,97,116,51,120, -50,0,18,109,0,16,8,48,0,57,60,0,18,109,0,16,10,49,0,57,60,0,18,109,0,16,10,50,0,57,60,0,0,0,0,1,0, -30,2,24,1,0,2,30,109,0,0,1,1,0,5,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,60,0,18, -109,0,16,10,49,0,57,60,0,18,109,0,16,10,50,0,57,60,0,0,0,0,1,0,29,2,24,1,0,2,29,109,0,0,1,1,0,5,0, -0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,60,0,18,109,0,16,10,49,0,57,60,0,18,109,0, -16,10,50,0,57,60,0,18,109,0,16,10,51,0,57,60,0,0,0,0,1,0,31,2,24,1,0,2,31,109,0,0,1,1,0,5,0,0,0,1, -8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,60,0,18,109,0,16,10,49,0,57,60,0,18,109,0,16,10, -50,0,57,60,0,18,109,0,16,10,51,0,57,60,0,0,0,0,0 +57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10,51,0,57,51,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index f71fe28cda..44170946ce 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -3,15 +3,15 @@ /* slang_core.gc */ 3,1,0,5,1,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, -116,86,97,108,0,0,18,102,0,0,0,0,1,0,1,1,1,1,0,5,105,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0, -48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,18,122, -101,114,111,0,0,0,0,1,0,1,1,1,1,0,9,102,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,18,122,101,114,111,0, -0,0,0,1,0,5,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,1, -98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,5,105,0,0,0,1,4,105,110, -116,95,116,111,95,102,108,111,97,116,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,0,1,1,1, -1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,5,1,1,1,0,5,105,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,18,105,0,20,0,0,1,0,9,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,0,18,102,0,0,0,0,1,0,5,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +98,0,20,0,0,1,0,5,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,20,0,0,1,0,1,1, +1,1,0,5,105,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,18,122,101,114,111,0,0,0,0,1,0,1,1,1,1,0,9,102, +0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, +114,101,116,86,97,108,0,0,18,102,0,0,18,122,101,114,111,0,0,0,0,1,0,1,1,1,1,0,1,98,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,5,105,0,0,0,1,4,105,110,116,95,116,111,95, +102,108,111,97,116,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,0,9,1,1,1,0,1,98,0,0,0,1, +9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101, 116,86,97,108,0,18,102,0,20,0,0,1,0,10,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,9,18,95,95,114,101, 116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,0, 1,0,10,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,102,0,59,120,120,0, -- cgit v1.2.3 From 40acf8cce24afe80388639b804f9ae8c0b779d81 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 25 Jul 2008 15:33:02 -0600 Subject: mesa: glsl: assorted fixes for resolving polymorphic functions Plus, - fix some issues in casting function arguments to format param types. - fix some vec/mat constructor bugs - find/report more syntax/semantic errors --- src/mesa/shader/slang/slang_codegen.c | 178 +++++++++++++------ src/mesa/shader/slang/slang_compile.c | 30 +++- src/mesa/shader/slang/slang_compile_variable.c | 22 ++- src/mesa/shader/slang/slang_emit.c | 6 + src/mesa/shader/slang/slang_simplify.c | 226 ++++++++++++++----------- src/mesa/shader/slang/slang_simplify.h | 28 +++ src/mesa/shader/slang/slang_typeinfo.c | 100 ++++++----- src/mesa/shader/slang/slang_typeinfo.h | 3 +- 8 files changed, 391 insertions(+), 202 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 2f9e147b87..ad9cf06e27 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1178,7 +1178,8 @@ slang_inline_function_call(slang_assemble_ctx * A, slang_function *fun, slang_operation_copy(inlined, fun->body); /*** XXX review this */ - assert(inlined->type == SLANG_OPER_BLOCK_NO_NEW_SCOPE); + assert(inlined->type == SLANG_OPER_BLOCK_NO_NEW_SCOPE || + inlined->type == SLANG_OPER_BLOCK_NEW_SCOPE); inlined->type = SLANG_OPER_BLOCK_NEW_SCOPE; #if 0 @@ -1710,22 +1711,51 @@ print_funcs(struct slang_function_scope_ *scope, const char *name) * all of them... */ static slang_function * -_slang_find_function_by_argc(struct slang_function_scope_ *scope, +_slang_find_function_by_argc(slang_function_scope *scope, const char *name, int numArgs) { - GLuint i; - for (i = 0; i < scope->num_functions; i++) { - slang_function *f = &scope->functions[i]; - if (strcmp(name, (char*) f->header.a_name) == 0 - /*&& numArgs == f->param_count*/) - return f; + while (scope) { + GLuint i; + for (i = 0; i < scope->num_functions; i++) { + slang_function *f = &scope->functions[i]; + if (strcmp(name, (char*) f->header.a_name) == 0) { + int haveRetValue = _slang_function_has_return_value(f); + if (numArgs == f->param_count - haveRetValue) + return f; + } + } + scope = scope->outer_scope; } - if (scope->outer_scope) - return _slang_find_function_by_argc(scope->outer_scope, name, numArgs); + return NULL; } +static slang_function * +_slang_find_function_by_max_argc(slang_function_scope *scope, + const char *name) +{ + slang_function *maxFunc = NULL; + GLuint maxArgs = 0; + + while (scope) { + GLuint i; + for (i = 0; i < scope->num_functions; i++) { + slang_function *f = &scope->functions[i]; + if (strcmp(name, (char*) f->header.a_name) == 0) { + if (f->param_count > maxArgs) { + maxArgs = f->param_count; + maxFunc = f; + } + } + } + scope = scope->outer_scope; + } + + return maxFunc; +} + + /** * Generate a new slang_function which is a constructor for a user-defined * struct type. @@ -1783,16 +1813,20 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str) * return T; */ { + slang_variable_scope *scope; slang_variable *var; GLint i; fun->body = slang_operation_new(1); - fun->body->type = SLANG_OPER_BLOCK_NO_NEW_SCOPE; + fun->body->type = SLANG_OPER_BLOCK_NEW_SCOPE; fun->body->num_children = numFields + 2; fun->body->children = slang_operation_new(numFields + 2); + scope = fun->body->locals; + scope->outer_scope = fun->parameters; + /* create local var 't' */ - var = slang_variable_scope_grow(fun->parameters); + var = slang_variable_scope_grow(scope); var->a_name = slang_atom_pool_atom(A->atoms, "t"); var->type = fun->header.type; @@ -1802,7 +1836,7 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str) decl = &fun->body->children[0]; decl->type = SLANG_OPER_VARIABLE_DECL; - decl->locals = _slang_variable_scope_new(fun->parameters); + decl->locals = _slang_variable_scope_new(scope); decl->a_id = var->a_name; } @@ -1811,7 +1845,7 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str) slang_operation *assign = &fun->body->children[1 + i]; assign->type = SLANG_OPER_ASSIGN; - assign->locals = _slang_variable_scope_new(fun->parameters); + assign->locals = _slang_variable_scope_new(scope); assign->num_children = 2; assign->children = slang_operation_new(2); @@ -1819,21 +1853,21 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str) slang_operation *lhs = &assign->children[0]; lhs->type = SLANG_OPER_FIELD; - lhs->locals = _slang_variable_scope_new(fun->parameters); + lhs->locals = _slang_variable_scope_new(scope); lhs->num_children = 1; lhs->children = slang_operation_new(1); lhs->a_id = str->fields->variables[i]->a_name; lhs->children[0].type = SLANG_OPER_IDENTIFIER; lhs->children[0].a_id = var->a_name; - lhs->children[0].locals = _slang_variable_scope_new(fun->parameters); + lhs->children[0].locals = _slang_variable_scope_new(scope); #if 0 lhs->children[1].num_children = 1; lhs->children[1].children = slang_operation_new(1); lhs->children[1].children[0].type = SLANG_OPER_IDENTIFIER; lhs->children[1].children[0].a_id = str->fields->variables[i]->a_name; - lhs->children[1].children->locals = _slang_variable_scope_new(fun->parameters); + lhs->children[1].children->locals = _slang_variable_scope_new(scope); #endif } @@ -1841,7 +1875,7 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str) slang_operation *rhs = &assign->children[1]; rhs->type = SLANG_OPER_IDENTIFIER; - rhs->locals = _slang_variable_scope_new(fun->parameters); + rhs->locals = _slang_variable_scope_new(scope); rhs->a_id = str->fields->variables[i]->a_name; } } @@ -1851,12 +1885,12 @@ _slang_make_constructor(slang_assemble_ctx *A, slang_struct *str) slang_operation *ret = &fun->body->children[numFields + 1]; ret->type = SLANG_OPER_RETURN; - ret->locals = _slang_variable_scope_new(fun->parameters); + ret->locals = _slang_variable_scope_new(scope); ret->num_children = 1; ret->children = slang_operation_new(1); ret->children[0].type = SLANG_OPER_IDENTIFIER; ret->children[0].a_id = var->a_name; - ret->children[0].locals = _slang_variable_scope_new(fun->parameters); + ret->children[0].locals = _slang_variable_scope_new(scope); } } @@ -1889,6 +1923,27 @@ _slang_locate_struct_constructor(slang_assemble_ctx *A, const char *name) } + +static GLboolean +_slang_is_vec_mat_type(const char *name) +{ + static const char *vecmat_types[] = { + "float", "int", "bool", + "vec2", "vec3", "vec4", + "ivec2", "ivec3", "ivec4", + "bvec2", "bvec3", "bvec4", + "mat2", "mat3", "mat4", + "mat2x3", "mat2x4", "mat3x2", "mat3x4", "mat4x2", "mat4x3", + NULL + }; + int i; + for (i = 0; vecmat_types[i]; i++) + if (_mesa_strcmp(name, vecmat_types[i]) == 0) + return GL_TRUE; + return GL_FALSE; +} + + /** * Assemble a function call, given a particular function name. * \param name the function's name (operators like '*' are possible). @@ -1901,29 +1956,53 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name, const GLuint param_count = oper->num_children; slang_atom atom; slang_function *fun; + GLboolean error; atom = slang_atom_pool_atom(A->atoms, name); if (atom == SLANG_ATOM_NULL) return NULL; /* - * Use 'name' to find the function to call + * First, try to find function by name and exact argument type matching. */ fun = _slang_locate_function(A->space.funcs, atom, params, param_count, - &A->space, A->atoms, A->log); + &A->space, A->atoms, A->log, &error); + + if (error) { + slang_info_log_error(A->log, + "Function '%s' not found (check argument types)", + name); + return NULL; + } if (!fun) { + /* Next, try locating a constructor function for a user-defined type */ fun = _slang_locate_struct_constructor(A, name); } - if (!fun) { - /* A function with exactly the right parameters/types was not found. - * Try adapting the parameters. + if (!fun && _slang_is_vec_mat_type(name)) { + /* Next, if this call looks like a vec() or mat() constructor call, + * try "unwinding" the args to satisfy a constructor. */ + fun = _slang_find_function_by_max_argc(A->space.funcs, name); + if (fun) { + if (!_slang_adapt_call(oper, fun, &A->space, A->atoms, A->log)) { + slang_info_log_error(A->log, + "Function '%s' not found (check argument types)", + name); + return NULL; + } + } + } + + if (!fun) { + /* Next, try casting args to the types of the formal parameters */ int numArgs = oper->num_children; fun = _slang_find_function_by_argc(A->space.funcs, name, numArgs); - if (!fun || !_slang_adapt_call(oper, fun, &A->space, A->atoms, A->log)) { - slang_info_log_error(A->log, "Function '%s' not found (check argument types)", name); + if (!fun || !_slang_cast_func_params(oper, fun, &A->space, A->atoms, A->log)) { + slang_info_log_error(A->log, + "Function '%s' not found (check argument types)", + name); return NULL; } assert(fun); @@ -2910,6 +2989,7 @@ _slang_gen_array_element(slang_assemble_ctx * A, slang_operation *oper) /*n->Store = _slang_clone_ir_storage_swz(n->Store, */ n->Writemask = WRITEMASK_X << index; } + assert(n->Store); return n; } else { @@ -2954,11 +3034,17 @@ _slang_gen_array_element(slang_assemble_ctx * A, slang_operation *oper) } } + if (!array->Store) { + slang_info_log_error(A->log, "Invalid array"); + return NULL; + } + elem = new_node2(IR_ELEMENT, array, index); elem->Store = _slang_new_ir_storage_relative(constIndex, elemSize, array->Store); + assert(elem->Store->Parent); /* XXX try to do some array bounds checking here */ return elem; } @@ -2988,6 +3074,22 @@ print_vars(slang_variable_scope *s) #endif +#if 0 +static void +_slang_undeclare_vars(slang_variable_scope *locals) +{ + if (locals->num_variables > 0) { + int i; + for (i = 0; i < locals->num_variables; i++) { + slang_variable *v = locals->variables[i]; + printf("undeclare %s at %p\n", (char*) v->a_name, v); + v->declared = GL_FALSE; + } + } +} +#endif + + /** * Generate IR tree for a slang_operation (AST node) */ @@ -3007,6 +3109,7 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) _slang_pop_var_table(A->vartable); + /*_slang_undeclare_vars(oper->locals);*/ /*print_vars(oper->locals);*/ if (n) @@ -3031,27 +3134,6 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) tree = new_seq(tree, n); } -#if 00 - if (oper->locals->num_variables > 0) { - int i; - /* - printf("\n****** Deallocate vars in scope!\n"); - */ - for (i = 0; i < oper->locals->num_variables; i++) { - slang_variable *v = oper->locals->variables + i; - if (v->aux) { - slang_ir_storage *store = (slang_ir_storage *) v->aux; - /* - printf(" Deallocate var %s\n", (char*) v->a_name); - */ - assert(store->File == PROGRAM_TEMPORARY); - assert(store->Index >= 0); - _slang_free_temp(A->vartable, store->Index, store->Size); - } - } - } -#endif - /*print_vars(oper->locals);*/ return tree; } else { diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 862d0741ff..60539f19b6 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -337,9 +337,10 @@ convert_to_array(slang_parse_ctx * C, slang_variable * var, static GLboolean parse_struct_field_var(slang_parse_ctx * C, slang_output_ctx * O, - slang_variable * var, const slang_type_specifier * sp) + slang_variable * var, slang_atom a_name, + const slang_type_specifier * sp) { - var->a_name = parse_identifier(C); + var->a_name = a_name; if (var->a_name == SLANG_ATOM_NULL) return GL_FALSE; @@ -372,12 +373,19 @@ parse_struct_field(slang_parse_ctx * C, slang_output_ctx * O, return 0; do { + slang_atom a_name; slang_variable *var = slang_variable_scope_grow(st->fields); if (!var) { slang_info_log_memory(C->L); return 0; } - if (!parse_struct_field_var(C, &o, var, sp)) + a_name = parse_identifier(C); + if (_slang_locate_variable(st->fields, a_name, GL_FALSE)) { + slang_info_log_error(C->L, "duplicate field '%s'", (char *) a_name); + return 0; + } + + if (!parse_struct_field_var(C, &o, var, a_name, sp)) return 0; } while (*C->I++ != FIELD_NONE); @@ -1579,11 +1587,22 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O, const slang_fully_specified_type * type) { slang_variable *var; + slang_atom a_name; /* empty init declatator (without name, e.g. "float ;") */ if (*C->I++ == VARIABLE_NONE) return 1; + a_name = parse_identifier(C); + + /* check if name is already in this scope */ + if (_slang_locate_variable(O->vars, a_name, GL_FALSE)) { + slang_info_log_error(C->L, + "declaration of '%s' conflicts with previous declaration", + (char *) a_name); + return 0; + } + /* make room for the new variable and initialize it */ var = slang_variable_scope_grow(O->vars); if (!var) { @@ -1593,7 +1612,7 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O, /* copy the declarator qualifier type, parse the identifier */ var->type.qualifier = type->qualifier; - var->a_name = parse_identifier(C); + var->a_name = a_name; if (var->a_name == SLANG_ATOM_NULL) return 0; @@ -1657,6 +1676,7 @@ parse_init_declarator(slang_parse_ctx * C, slang_output_ctx * O, A.space.vars = O->vars; A.program = O->program; A.vartable = O->vartable; + A.log = C->L; A.curFuncEndLabel = NULL; if (!_slang_codegen_global_variable(&A, var, C->type)) return 0; @@ -2265,7 +2285,7 @@ compile_shader(GLcontext *ctx, slang_code_object * object, GLboolean success; grammar id = 0; -#if 0 /* for debug */ +#if 1 /* for debug */ _mesa_printf("********* COMPILE SHADER ***********\n"); _mesa_printf("%s\n", shader->Source); _mesa_printf("************************************\n"); diff --git a/src/mesa/shader/slang/slang_compile_variable.c b/src/mesa/shader/slang/slang_compile_variable.c index b26c18e38d..3428b49e16 100644 --- a/src/mesa/shader/slang/slang_compile_variable.c +++ b/src/mesa/shader/slang/slang_compile_variable.c @@ -267,6 +267,7 @@ slang_variable_construct(slang_variable * var) var->size = 0; var->isTemp = GL_FALSE; var->aux = NULL; + var->declared = 0; return 1; } @@ -325,16 +326,23 @@ slang_variable_copy(slang_variable * x, const slang_variable * y) } +/** + * Search for named variable in given scope. + * \param all if true, search parent scopes too. + */ slang_variable * _slang_locate_variable(const slang_variable_scope * scope, const slang_atom a_name, GLboolean all) { - GLuint i; - - for (i = 0; i < scope->num_variables; i++) - if (a_name == scope->variables[i]->a_name) - return scope->variables[i]; - if (all && scope->outer_scope != NULL) - return _slang_locate_variable(scope->outer_scope, a_name, 1); + while (scope) { + GLuint i; + for (i = 0; i < scope->num_variables; i++) + if (a_name == scope->variables[i]->a_name) + return scope->variables[i]; + if (all) + scope = scope->outer_scope; + else + scope = NULL; + } return NULL; } diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 9bff055ad5..f1af6902ea 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -256,6 +256,12 @@ storage_to_src_reg(struct prog_src_register *src, const slang_ir_storage *st) } assert(st->File >= 0); +#if 1 /* XXX temporary */ + if (st->File == PROGRAM_UNDEFINED) { + slang_ir_storage *st0 = (slang_ir_storage *) st; + st0->File = PROGRAM_TEMPORARY; + } +#endif assert(st->File < PROGRAM_UNDEFINED); src->File = st->File; diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c index c34bef2164..c4bf0175be 100644 --- a/src/mesa/shader/slang/slang_simplify.c +++ b/src/mesa/shader/slang/slang_simplify.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 6.5.2 + * Version: 7.1 * - * Copyright (C) 2005-2006 Brian Paul All Rights Reserved. + * Copyright (C) 2005-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -23,11 +23,11 @@ */ /** - * \file slang_assemble_typeinfo.c - * slang type info - * \author Michal Krol + * Functions for constant folding, built-in constant lookup, and function + * call casting. */ + #include "main/imports.h" #include "main/macros.h" #include "main/get.h" @@ -314,17 +314,18 @@ _slang_simplify(slang_operation *oper, /** - * Adapt the arguments for a function call to match the parameters of - * the given function. - * This is for: - * 1. converting/casting argument types to match parameters - * 2. breaking up vector/matrix types into individual components to - * satisfy constructors. + * Insert casts to try to adapt actual parameters to formal parameters for a + * function call when an exact match for the parameter types is not found. + * Example: + * void foo(int i, bool b) {} + * x = foo(3.15, 9); + * Gets translated into: + * x = foo(int(3.15), bool(9)) */ GLboolean -_slang_adapt_call(slang_operation *callOper, const slang_function *fun, - const slang_name_space * space, - slang_atom_pool * atoms, slang_info_log *log) +_slang_cast_func_params(slang_operation *callOper, const slang_function *fun, + const slang_name_space * space, + slang_atom_pool * atoms, slang_info_log *log) { const GLboolean haveRetValue = _slang_function_has_return_value(fun); const int numParams = fun->param_count - haveRetValue; @@ -332,95 +333,9 @@ _slang_adapt_call(slang_operation *callOper, const slang_function *fun, int dbg = 0; if (dbg) - printf("Adapt %d args to %d parameters for %s\n", - callOper->num_children, numParams, (char *) fun->header.a_name); + printf("Adapt call of %d args to func %s (%d params)\n", + callOper->num_children, (char*) fun->header.a_name, numParams); - /* Only try adapting for constructors */ - if (fun->kind != SLANG_FUNC_CONSTRUCTOR) - return GL_FALSE; - - if (callOper->num_children != numParams) { - /* number of arguments doesn't match number of parameters */ - - if (fun->kind == SLANG_FUNC_CONSTRUCTOR) { - /* For constructor calls, we can try to unroll vector/matrix args - * into individual floats/ints and try to match the function params. - */ - for (i = 0; i < numParams; i++) { - slang_typeinfo argType; - GLint argSz, j; - - /* Get type of arg[i] */ - if (!slang_typeinfo_construct(&argType)) - return GL_FALSE; - if (!_slang_typeof_operation_(&callOper->children[i], space, - &argType, atoms, log)) { - slang_typeinfo_destruct(&argType); - return GL_FALSE; - } - - /* - paramSz = _slang_sizeof_type_specifier(¶mVar->type.specifier); - assert(paramSz == 1); - */ - argSz = _slang_sizeof_type_specifier(&argType.spec); - if (argSz > 1) { - slang_operation origArg; - /* break up arg[i] into components */ - if (dbg) - printf("Break up arg %d from 1 to %d elements\n", i, argSz); - - slang_operation_construct(&origArg); - slang_operation_copy(&origArg, - &callOper->children[i]); - - /* insert argSz-1 new children/args */ - for (j = 0; j < argSz - 1; j++) { - (void) slang_operation_insert(&callOper->num_children, - &callOper->children, i); - } - - /* replace arg[i+j] with subscript/index oper */ - for (j = 0; j < argSz; j++) { - callOper->children[i + j].type = SLANG_OPER_SUBSCRIPT; - callOper->children[i + j].num_children = 2; - callOper->children[i + j].children = slang_operation_new(2); - slang_operation_copy(&callOper->children[i + j].children[0], - &origArg); - callOper->children[i + j].children[1].type - = SLANG_OPER_LITERAL_INT; - callOper->children[i + j].children[1].literal[0] = (GLfloat) j; - } - - } - } /* for i */ - } - else { - /* non-constructor function: number of args must match number - * of function params. - */ - return GL_FALSE; /* caller will record an error msg */ - } - } - - if (callOper->num_children < (GLuint) numParams) { - /* still not enough args for all params */ - return GL_FALSE; - } - else if (callOper->num_children > (GLuint) numParams) { - /* now too many arguments */ - /* XXX this isn't always an error, see spec */ - return GL_FALSE; - } - - /* - * Second phase, argument casting. - * Example: - * void foo(int i, bool b) {} - * x = foo(3.15, 9); - * Gets translated into: - * x = foo(int(3.15), bool(9)) - */ for (i = 0; i < numParams; i++) { slang_typeinfo argType; slang_variable *paramVar = fun->parameters->variables[i]; @@ -448,6 +363,12 @@ _slang_adapt_call(slang_operation *callOper, const slang_function *fun, slang_operation_copy(child, &callOper->children[i]); child->locals->outer_scope = callOper->children[i].locals; +#if 0 + if (_slang_sizeof_type_specifier(&argType.spec) > + _slang_sizeof_type_specifier(¶mVar->type.specifier)) { + } +#endif + callOper->children[i].type = SLANG_OPER_CALL; callOper->children[i].a_id = slang_atom_pool_atom(atoms, constructorName); callOper->children[i].num_children = 1; @@ -457,6 +378,107 @@ _slang_adapt_call(slang_operation *callOper, const slang_function *fun, slang_typeinfo_destruct(&argType); } + if (dbg) { + printf("===== New call to %s with cast arguments ===============\n", + (char*) fun->header.a_name); + slang_print_tree(callOper, 5); + } + + return GL_TRUE; +} + + +/** + * Adapt the arguments for a function call to match the parameters of + * the given function. + * This is for: + * 1. converting/casting argument types to match parameters + * 2. breaking up vector/matrix types into individual components to + * satisfy constructors. + */ +GLboolean +_slang_adapt_call(slang_operation *callOper, const slang_function *fun, + const slang_name_space * space, + slang_atom_pool * atoms, slang_info_log *log) +{ + const GLboolean haveRetValue = _slang_function_has_return_value(fun); + const int numParams = fun->param_count - haveRetValue; + int i; + int dbg = 0; + + if (dbg) + printf("Adapt %d args to %d parameters for %s\n", + callOper->num_children, numParams, (char *) fun->header.a_name); + + /* Only try adapting for constructors */ + if (fun->kind != SLANG_FUNC_CONSTRUCTOR) + return GL_FALSE; + + if (callOper->num_children != numParams) { + /* number of arguments doesn't match number of parameters */ + + /* For constructor calls, we can try to unroll vector/matrix args + * into individual floats/ints and try to match the function params. + */ + for (i = 0; i < numParams; i++) { + slang_typeinfo argType; + GLint argSz, j; + + /* Get type of arg[i] */ + if (!slang_typeinfo_construct(&argType)) + return GL_FALSE; + if (!_slang_typeof_operation_(&callOper->children[i], space, + &argType, atoms, log)) { + slang_typeinfo_destruct(&argType); + return GL_FALSE; + } + + /* + paramSz = _slang_sizeof_type_specifier(¶mVar->type.specifier); + assert(paramSz == 1); + */ + argSz = _slang_sizeof_type_specifier(&argType.spec); + if (argSz > 1) { + slang_operation origArg; + /* break up arg[i] into components */ + if (dbg) + printf("Break up arg %d from 1 to %d elements\n", i, argSz); + + slang_operation_construct(&origArg); + slang_operation_copy(&origArg, &callOper->children[i]); + + /* insert argSz-1 new children/args */ + for (j = 0; j < argSz - 1; j++) { + (void) slang_operation_insert(&callOper->num_children, + &callOper->children, i); + } + + /* replace arg[i+j] with subscript/index oper */ + for (j = 0; j < argSz; j++) { + callOper->children[i + j].type = SLANG_OPER_SUBSCRIPT; + callOper->children[i + j].locals = _slang_variable_scope_new(callOper->locals); + callOper->children[i + j].num_children = 2; + callOper->children[i + j].children = slang_operation_new(2); + slang_operation_copy(&callOper->children[i + j].children[0], + &origArg); + callOper->children[i + j].children[1].type + = SLANG_OPER_LITERAL_INT; + callOper->children[i + j].children[1].literal[0] = (GLfloat) j; + } + } + } + } + + if (callOper->num_children < (GLuint) numParams) { + /* still not enough args for all params */ + return GL_FALSE; + } + else if (callOper->num_children > (GLuint) numParams) { + /* now too many arguments */ + /* just truncate */ + callOper->num_children = (GLuint) numParams; + } + if (dbg) { printf("===== New call to %s with adapted arguments ===============\n", (char*) fun->header.a_name); diff --git a/src/mesa/shader/slang/slang_simplify.h b/src/mesa/shader/slang/slang_simplify.h index b3840ee9df..8689c23b1a 100644 --- a/src/mesa/shader/slang/slang_simplify.h +++ b/src/mesa/shader/slang/slang_simplify.h @@ -1,3 +1,26 @@ +/* + * Mesa 3-D graphics library + * Version: 7.1 + * + * Copyright (C) 2005-2008 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ #ifndef SLANG_SIMPLIFY_H #define SLANG_SIMPLIFY_H @@ -13,6 +36,11 @@ _slang_simplify(slang_operation *oper, slang_atom_pool * atoms); +extern GLboolean +_slang_cast_func_params(slang_operation *callOper, const slang_function *fun, + const slang_name_space * space, + slang_atom_pool * atoms, slang_info_log *log); + extern GLboolean _slang_adapt_call(slang_operation *callOper, const slang_function *fun, const slang_name_space * space, diff --git a/src/mesa/shader/slang/slang_typeinfo.c b/src/mesa/shader/slang/slang_typeinfo.c index 8a1c3abf48..26f9f6ca85 100644 --- a/src/mesa/shader/slang/slang_typeinfo.c +++ b/src/mesa/shader/slang/slang_typeinfo.c @@ -273,6 +273,7 @@ slang_type_specifier_compatible(const slang_type_specifier * x, GLboolean slang_typeinfo_construct(slang_typeinfo * ti) { + /*_mesa_bzero(ti, sizeof(*ti));*/ slang_type_specifier_ctr(&ti->spec); ti->array_len = 0; return GL_TRUE; @@ -304,10 +305,16 @@ _slang_typeof_function(slang_atom a_name, slang_function **funFound, slang_atom_pool *atoms, slang_info_log *log) { + GLboolean error; + *funFound = _slang_locate_function(space->funcs, a_name, params, - num_params, space, atoms, log); + num_params, space, atoms, log, &error); + if (error) + return GL_FALSE; + if (!*funFound) return GL_TRUE; /* yes, not false */ + return slang_type_specifier_copy(spec, &(*funFound)->header.type.specifier); } @@ -755,51 +762,66 @@ slang_function * _slang_locate_function(const slang_function_scope * funcs, slang_atom a_name, slang_operation * args, GLuint num_args, const slang_name_space * space, slang_atom_pool * atoms, - slang_info_log *log) + slang_info_log *log, GLboolean *error) { + slang_typeinfo arg_ti[100]; GLuint i; - for (i = 0; i < funcs->num_functions; i++) { - slang_function *f = &funcs->functions[i]; - const GLuint haveRetValue = _slang_function_has_return_value(f); - GLuint j; - - if (a_name != f->header.a_name) - continue; - if (f->param_count - haveRetValue != num_args) - continue; - - /* compare parameter / argument types */ - for (j = 0; j < num_args; j++) { - slang_typeinfo ti; - - if (!slang_typeinfo_construct(&ti)) - return NULL; - if (!_slang_typeof_operation_(&args[j], space, &ti, atoms, log)) { - slang_typeinfo_destruct(&ti); - return NULL; + *error = GL_FALSE; + + /* determine type of each argument */ + assert(num_args < 100); + for (i = 0; i < num_args; i++) { + if (!slang_typeinfo_construct(&arg_ti[i])) + return NULL; + if (!_slang_typeof_operation_(&args[i], space, &arg_ti[i], atoms, log)) { + return NULL; + } + } + + /* loop over function scopes */ + while (funcs) { + + /* look for function with matching name and argument/param types */ + for (i = 0; i < funcs->num_functions; i++) { + slang_function *f = &funcs->functions[i]; + const GLuint haveRetValue = _slang_function_has_return_value(f); + GLuint j; + + if (a_name != f->header.a_name) + continue; + if (f->param_count - haveRetValue != num_args) + continue; + + /* compare parameter / argument types */ + for (j = 0; j < num_args; j++) { + if (!slang_type_specifier_compatible(&arg_ti[j].spec, + &f->parameters->variables[j]->type.specifier)) { + /* param/arg types don't match */ + break; + } + + /* "out" and "inout" formal parameter requires the actual + * argument to be an l-value. + */ + if (!arg_ti[j].can_be_referenced && + (f->parameters->variables[j]->type.qualifier == SLANG_QUAL_OUT || + f->parameters->variables[j]->type.qualifier == SLANG_QUAL_INOUT)) { + /* param is not an lvalue! */ + *error = GL_TRUE; + return NULL; + } } - if (!slang_type_specifier_compatible(&ti.spec, - &f->parameters->variables[j]->type.specifier)) { - slang_typeinfo_destruct(&ti); - break; + + if (j == num_args) { + /* name and args match! */ + return f; } - slang_typeinfo_destruct(&ti); - - /* "out" and "inout" formal parameter requires the actual - * parameter to be l-value. - */ - if (!ti.can_be_referenced && - (f->parameters->variables[j]->type.qualifier == SLANG_QUAL_OUT || - f->parameters->variables[j]->type.qualifier == SLANG_QUAL_INOUT)) - break; } - if (j == num_args) - return f; + + funcs = funcs->outer_scope; } - if (funcs->outer_scope != NULL) - return _slang_locate_function(funcs->outer_scope, a_name, args, - num_args, space, atoms, log); + return NULL; } diff --git a/src/mesa/shader/slang/slang_typeinfo.h b/src/mesa/shader/slang/slang_typeinfo.h index 8a36fc3422..960e99bda2 100644 --- a/src/mesa/shader/slang/slang_typeinfo.h +++ b/src/mesa/shader/slang/slang_typeinfo.h @@ -73,7 +73,8 @@ _slang_locate_function(const struct slang_function_scope_ *funcs, slang_atom name, struct slang_operation_ *params, GLuint num_params, const slang_name_space *space, - slang_atom_pool *atoms, slang_info_log *log); + slang_atom_pool *atoms, slang_info_log *log, + GLboolean *error); extern GLboolean -- cgit v1.2.3 From ff7a7031caa0ac592f210aca696a20c9de6dc0d4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 26 Jul 2008 09:17:10 +0900 Subject: gallium: Windows miniport portability fixes. --- src/gallium/auxiliary/util/p_debug.c | 12 +++---- src/gallium/auxiliary/util/u_snprintf.c | 10 ++++++ src/gallium/include/pipe/p_util.h | 63 +++++++++++++++++++++++++++++++++ 3 files changed, 78 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index b0240ad737..cdc7e66361 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -73,8 +73,7 @@ _EngDebugPrint(const char *format, ...) void _debug_vprintf(const char *format, va_list ap) { -#ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY -#ifndef WINCE +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) /* EngDebugPrint does not handle float point arguments, so we need to use * our own vsnprintf implementation. It is also very slow, so buffer until * we find a newline. */ @@ -85,9 +84,6 @@ void _debug_vprintf(const char *format, va_list ap) _EngDebugPrint("%s", buf); buf[0] = '\0'; } -#else - /* TODO: Implement debug print for WINCE */ -#endif #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER) /* EngDebugPrint does not handle float point arguments, so we need to use * our own vsnprintf implementation. It is also very slow, so buffer until @@ -99,7 +95,9 @@ void _debug_vprintf(const char *format, va_list ap) OutputDebugStringA(buf); buf[0] = '\0'; } -#else +#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) || defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) + /* TODO */ +#else /* !PIPE_SUBSYSTEM_WINDOWS */ vfprintf(stderr, format, ap); #endif } @@ -211,7 +209,7 @@ _debug_get_option(const char *name) #else return NULL; #endif -#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) +#elif defined(PIPE_SUBSYSTEM_WINDOWS_CE) || defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) /* TODO: implement */ return NULL; #else diff --git a/src/gallium/auxiliary/util/u_snprintf.c b/src/gallium/auxiliary/util/u_snprintf.c index c4f4bbd30c..7fa84d8bec 100644 --- a/src/gallium/auxiliary/util/u_snprintf.c +++ b/src/gallium/auxiliary/util/u_snprintf.c @@ -162,6 +162,8 @@ * . */ +#include "pipe/p_config.h" + #if HAVE_CONFIG_H #include #else @@ -1102,7 +1104,11 @@ again: * Factor of ten with the number of digits needed for the fractional * part. For example, if the precision is 3, the mask will be 1000. */ +#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) + mask = (unsigned long)mypow10(precision); +#else mask = (UINTMAX_T)mypow10(precision); +#endif /* * We "cheat" by converting the fractional part to integer by * multiplying by a factor of ten. @@ -1354,7 +1360,11 @@ cast(LDOUBLE value) if (value >= UINTMAX_MAX) return UINTMAX_MAX; +#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) + result = (unsigned long)value; +#else result = (UINTMAX_T)value; +#endif /* * At least on NetBSD/sparc64 3.0.2 and 4.99.30, casting long double to * an integer type converts e.g. 1.9 to 2 instead of 1 (which violates diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 7dcdd28287..892bd4bf8a 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -33,8 +33,71 @@ #include "p_debug.h" #include "p_format.h" #include "p_pointer.h" + +#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) +__inline double ceil(double val) +{ + double ceil_val; + + if((val - (long) val) == 0) { + ceil_val = val; + } else { + if(val > 0) { + ceil_val = (long) val + 1; + } else { + ceil_val = (long) val; + } + } + + return ceil_val; +} + +#ifndef PIPE_SUBSYSTEM_WINDOWS_CE +__inline double floor(double val) +{ + double floor_val; + + if((val - (long) val) == 0) { + floor_val = val; + } else { + if(val > 0) { + floor_val = (long) val; + } else { + floor_val = (long) val - 1; + } + } + + return floor_val; +} +#endif + +#pragma function(pow) +__inline double __cdecl pow(double val, double exponent) +{ + /* XXX */ + assert(0); + return 0; +} + +#pragma function(log) +__inline double __cdecl log(double val) +{ + /* XXX */ + assert(0); + return 0; +} + +#pragma function(atan2) +__inline double __cdecl atan2(double val) +{ + /* XXX */ + assert(0); + return 0; +} +#else #include #include +#endif #ifdef __cplusplus -- cgit v1.2.3 From c208a2c791fa24c7c5887fc496738cbddbfafc72 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 28 Jul 2008 12:42:13 +0900 Subject: Merge tgsi/exec and tgsi/util directories. --- src/gallium/auxiliary/cso_cache/cso_context.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_aaline.c | 4 +- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 4 +- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 4 +- src/gallium/auxiliary/draw/draw_private.h | 4 +- src/gallium/auxiliary/draw/draw_vs_aos.c | 8 +- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 6 +- src/gallium/auxiliary/draw/draw_vs_aos_machine.c | 6 +- src/gallium/auxiliary/draw/draw_vs_exec.c | 4 +- src/gallium/auxiliary/draw/draw_vs_llvm.c | 2 +- src/gallium/auxiliary/draw/draw_vs_sse.c | 4 +- src/gallium/auxiliary/gallivm/gallivm.cpp | 4 +- src/gallium/auxiliary/gallivm/gallivm_cpu.cpp | 4 +- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 10 +- src/gallium/auxiliary/tgsi/Makefile | 18 +- src/gallium/auxiliary/tgsi/SConscript | 24 +- src/gallium/auxiliary/tgsi/exec/Makefile | 2 - src/gallium/auxiliary/tgsi/exec/tgsi_exec.c | 2522 -------------------- src/gallium/auxiliary/tgsi/exec/tgsi_exec.h | 253 -- src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c | 2275 ------------------ src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h | 49 - src/gallium/auxiliary/tgsi/tgsi_build.c | 1324 ++++++++++ src/gallium/auxiliary/tgsi/tgsi_build.h | 332 +++ src/gallium/auxiliary/tgsi/tgsi_dump.c | 582 +++++ src/gallium/auxiliary/tgsi/tgsi_dump.h | 63 + src/gallium/auxiliary/tgsi/tgsi_dump_c.c | 845 +++++++ src/gallium/auxiliary/tgsi/tgsi_dump_c.h | 49 + src/gallium/auxiliary/tgsi/tgsi_exec.c | 2522 ++++++++++++++++++++ src/gallium/auxiliary/tgsi/tgsi_exec.h | 253 ++ src/gallium/auxiliary/tgsi/tgsi_iterate.c | 85 + src/gallium/auxiliary/tgsi/tgsi_iterate.h | 76 + src/gallium/auxiliary/tgsi/tgsi_parse.c | 332 +++ src/gallium/auxiliary/tgsi/tgsi_parse.h | 151 ++ src/gallium/auxiliary/tgsi/tgsi_sanity.c | 341 +++ src/gallium/auxiliary/tgsi/tgsi_sanity.h | 49 + src/gallium/auxiliary/tgsi/tgsi_scan.c | 226 ++ src/gallium/auxiliary/tgsi/tgsi_scan.h | 74 + src/gallium/auxiliary/tgsi/tgsi_sse2.c | 2275 ++++++++++++++++++ src/gallium/auxiliary/tgsi/tgsi_sse2.h | 49 + src/gallium/auxiliary/tgsi/tgsi_text.c | 1221 ++++++++++ src/gallium/auxiliary/tgsi/tgsi_text.h | 47 + src/gallium/auxiliary/tgsi/tgsi_transform.c | 199 ++ src/gallium/auxiliary/tgsi/tgsi_transform.h | 93 + src/gallium/auxiliary/tgsi/tgsi_util.c | 300 +++ src/gallium/auxiliary/tgsi/tgsi_util.h | 96 + src/gallium/auxiliary/tgsi/util/tgsi_build.c | 1324 ---------- src/gallium/auxiliary/tgsi/util/tgsi_build.h | 332 --- src/gallium/auxiliary/tgsi/util/tgsi_dump.c | 582 ----- src/gallium/auxiliary/tgsi/util/tgsi_dump.h | 63 - src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c | 845 ------- src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h | 49 - src/gallium/auxiliary/tgsi/util/tgsi_iterate.c | 85 - src/gallium/auxiliary/tgsi/util/tgsi_iterate.h | 76 - src/gallium/auxiliary/tgsi/util/tgsi_parse.c | 332 --- src/gallium/auxiliary/tgsi/util/tgsi_parse.h | 151 -- src/gallium/auxiliary/tgsi/util/tgsi_sanity.c | 341 --- src/gallium/auxiliary/tgsi/util/tgsi_sanity.h | 49 - src/gallium/auxiliary/tgsi/util/tgsi_scan.c | 226 -- src/gallium/auxiliary/tgsi/util/tgsi_scan.h | 74 - src/gallium/auxiliary/tgsi/util/tgsi_text.c | 1221 ---------- src/gallium/auxiliary/tgsi/util/tgsi_text.h | 47 - src/gallium/auxiliary/tgsi/util/tgsi_transform.c | 199 -- src/gallium/auxiliary/tgsi/util/tgsi_transform.h | 93 - src/gallium/auxiliary/tgsi/util/tgsi_util.c | 300 --- src/gallium/auxiliary/tgsi/util/tgsi_util.h | 96 - src/gallium/auxiliary/util/u_gen_mipmap.c | 6 +- src/gallium/auxiliary/util/u_simple_shaders.c | 6 +- src/gallium/drivers/cell/ppu/cell_context.h | 2 +- src/gallium/drivers/cell/ppu/cell_state_shader.c | 2 +- src/gallium/drivers/cell/spu/spu_exec.c | 4 +- src/gallium/drivers/cell/spu/spu_exec.h | 2 +- src/gallium/drivers/cell/spu/spu_util.c | 4 +- src/gallium/drivers/i915simple/i915_context.h | 2 +- .../drivers/i915simple/i915_fpc_translate.c | 4 +- src/gallium/drivers/i915simple/i915_state.c | 2 +- src/gallium/drivers/i965simple/brw_context.h | 2 +- src/gallium/drivers/i965simple/brw_sf.c | 2 +- src/gallium/drivers/i965simple/brw_shader_info.c | 2 +- src/gallium/drivers/i965simple/brw_state.c | 4 +- src/gallium/drivers/i965simple/brw_vs_emit.c | 2 +- src/gallium/drivers/i965simple/brw_wm_decl.c | 2 +- src/gallium/drivers/i965simple/brw_wm_glsl.c | 2 +- src/gallium/drivers/softpipe/sp_fs_exec.c | 4 +- src/gallium/drivers/softpipe/sp_fs_llvm.c | 2 +- src/gallium/drivers/softpipe/sp_fs_sse.c | 4 +- src/gallium/drivers/softpipe/sp_headers.h | 2 +- src/gallium/drivers/softpipe/sp_state.h | 2 +- src/gallium/drivers/softpipe/sp_state_fs.c | 4 +- src/gallium/drivers/softpipe/sp_tex_sample.c | 2 +- src/gallium/state_trackers/python/gallium.i | 4 +- src/mesa/state_tracker/st_debug.c | 2 +- src/mesa/state_tracker/st_mesa_to_tgsi.c | 6 +- src/mesa/state_tracker/st_program.c | 2 +- 93 files changed, 11680 insertions(+), 11682 deletions(-) delete mode 100644 src/gallium/auxiliary/tgsi/exec/Makefile delete mode 100644 src/gallium/auxiliary/tgsi/exec/tgsi_exec.c delete mode 100644 src/gallium/auxiliary/tgsi/exec/tgsi_exec.h delete mode 100755 src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c delete mode 100755 src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_build.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_build.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_dump.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_dump.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_dump_c.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_dump_c.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_exec.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_exec.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_iterate.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_iterate.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_parse.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_parse.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_sanity.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_sanity.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_scan.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_scan.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_sse2.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_sse2.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_text.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_text.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_transform.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_transform.h create mode 100644 src/gallium/auxiliary/tgsi/tgsi_util.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_util.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_build.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_build.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_dump.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_dump.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_iterate.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_iterate.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_parse.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_parse.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_sanity.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_sanity.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_scan.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_scan.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_text.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_text.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_transform.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_transform.h delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_util.c delete mode 100644 src/gallium/auxiliary/tgsi/util/tgsi_util.h (limited to 'src') diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index af4af8ac1d..86e4d46a20 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -38,7 +38,7 @@ #include "pipe/p_state.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" #include "cso_cache/cso_context.h" #include "cso_cache/cso_cache.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index 3dd7ee19fd..991304b2c8 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -38,8 +38,8 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_transform.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_transform.h" +#include "tgsi/tgsi_dump.h" #include "draw_context.h" #include "draw_private.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 87fd303649..13b4401521 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -44,8 +44,8 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_transform.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_transform.h" +#include "tgsi/tgsi_dump.h" #include "draw_context.h" #include "draw_vs.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index 1f63f94365..d3bd9baddd 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -40,8 +40,8 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_transform.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_transform.h" +#include "tgsi/tgsi_dump.h" #include "draw_context.h" #include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_private.h b/src/gallium/auxiliary/draw/draw_private.h index 7bd1e670b4..626a2e3e30 100644 --- a/src/gallium/auxiliary/draw/draw_private.h +++ b/src/gallium/auxiliary/draw/draw_private.h @@ -44,8 +44,8 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_scan.h" +#include "tgsi/tgsi_exec.h" +#include "tgsi/tgsi_scan.h" struct pipe_context; diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 1f926b3e85..441877d46f 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -31,10 +31,10 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" +#include "tgsi/tgsi_exec.h" +#include "tgsi/tgsi_dump.h" #include "draw_vs.h" #include "draw_vs_aos.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index 8e834501a4..eda677cc62 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -28,9 +28,9 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" +#include "tgsi/tgsi_exec.h" #include "draw_vs.h" #include "draw_vs_aos.h" #include "draw_vertex.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c index 6a54917ae3..e029b7b4bb 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c @@ -31,9 +31,9 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" +#include "tgsi/tgsi_exec.h" #include "draw_vs.h" #include "draw_vs_aos.h" #include "draw_vertex.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index 4501877efc..e26903d8cc 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -38,8 +38,8 @@ #include "draw_context.h" #include "draw_vs.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_scan.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_scan.h" struct exec_vertex_shader { diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index c63bd51a10..fc03473b91 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -38,7 +38,7 @@ #include "draw_context.h" #include "draw_vs.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" #ifdef MESA_LLVM diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index c3189c707d..61f0c084c3 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -45,8 +45,8 @@ #include "rtasm/rtasm_cpu.h" #include "rtasm/rtasm_x86sse.h" -#include "tgsi/exec/tgsi_sse2.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_sse2.h" +#include "tgsi/tgsi_parse.h" #define SSE_MAX_VERTICES 4 diff --git a/src/gallium/auxiliary/gallivm/gallivm.cpp b/src/gallium/auxiliary/gallivm/gallivm.cpp index 77900e342b..29adeea47d 100644 --- a/src/gallium/auxiliary/gallivm/gallivm.cpp +++ b/src/gallium/auxiliary/gallivm/gallivm.cpp @@ -42,8 +42,8 @@ #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_exec.h" +#include "tgsi/tgsi_dump.h" #include #include diff --git a/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp index 857c190f7b..cf5b978837 100644 --- a/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp +++ b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp @@ -43,8 +43,8 @@ #include "pipe/p_shader_tokens.h" #include "pipe/p_util.h" -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_exec.h" +#include "tgsi/tgsi_dump.h" #include #include diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index 98014bdaa1..b14e2affd6 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -10,11 +10,11 @@ #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi/util/tgsi_build.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_exec.h" +#include "tgsi/tgsi_util.h" +#include "tgsi/tgsi_build.h" +#include "tgsi/tgsi_dump.h" #include diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index 9c4b967651..bbeff1304d 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -4,15 +4,15 @@ include $(TOP)/configs/current LIBNAME = tgsi C_SOURCES = \ - exec/tgsi_exec.c \ - exec/tgsi_sse2.c \ - util/tgsi_iterate.c \ - util/tgsi_build.c \ - util/tgsi_dump.c \ - util/tgsi_parse.c \ - util/tgsi_scan.c \ - util/tgsi_transform.c \ - util/tgsi_util.c + tgsi_build.c \ + tgsi_dump.c \ + tgsi_exec.c \ + tgsi_iterate.c \ + tgsi_parse.c \ + tgsi_scan.c \ + tgsi_sse2.c \ + tgsi_transform.c \ + tgsi_util.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript index 3bbfa1be54..03982e2194 100644 --- a/src/gallium/auxiliary/tgsi/SConscript +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -3,18 +3,18 @@ Import('*') tgsi = env.ConvenienceLibrary( target = 'tgsi', source = [ - 'exec/tgsi_exec.c', - 'exec/tgsi_sse2.c', - 'util/tgsi_build.c', - 'util/tgsi_dump.c', - 'util/tgsi_dump_c.c', - 'util/tgsi_iterate.c', - 'util/tgsi_parse.c', - 'util/tgsi_sanity.c', - 'util/tgsi_scan.c', - 'util/tgsi_text.c', - 'util/tgsi_transform.c', - 'util/tgsi_util.c', + 'tgsi_build.c', + 'tgsi_dump.c', + 'tgsi_dump_c.c', + 'tgsi_exec.c', + 'tgsi_iterate.c', + 'tgsi_parse.c', + 'tgsi_sanity.c', + 'tgsi_scan.c', + 'tgsi_sse2.c', + 'tgsi_text.c', + 'tgsi_transform.c', + 'tgsi_util.c', ]) auxiliaries.insert(0, tgsi) diff --git a/src/gallium/auxiliary/tgsi/exec/Makefile b/src/gallium/auxiliary/tgsi/exec/Makefile deleted file mode 100644 index 451911a354..0000000000 --- a/src/gallium/auxiliary/tgsi/exec/Makefile +++ /dev/null @@ -1,2 +0,0 @@ -default: - cd .. ; make diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c deleted file mode 100644 index 001a4c4b15..0000000000 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c +++ /dev/null @@ -1,2522 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * TGSI interpretor/executor. - * - * Flow control information: - * - * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) - * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special - * care since a condition may be true for some quad components but false - * for other components. - * - * We basically execute all statements (even if they're in the part of - * an IF/ELSE clause that's "not taken") and use a special mask to - * control writing to destination registers. This is the ExecMask. - * See store_dest(). - * - * The ExecMask is computed from three other masks (CondMask, LoopMask and - * ContMask) which are controlled by the flow control instructions (namely: - * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). - * - * - * Authors: - * Michal Krol - * Brian Paul - */ - -#include "pipe/p_compiler.h" -#include "pipe/p_state.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi_exec.h" - -#define TILE_TOP_LEFT 0 -#define TILE_TOP_RIGHT 1 -#define TILE_BOTTOM_LEFT 2 -#define TILE_BOTTOM_RIGHT 3 - -/* - * Shorthand locations of various utility registers (_I = Index, _C = Channel) - */ -#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I -#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C -#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I -#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C -#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I -#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C -#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I -#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C -#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I -#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C -#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I -#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C -#define TEMP_128_I TGSI_EXEC_TEMP_128_I -#define TEMP_128_C TGSI_EXEC_TEMP_128_C -#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I -#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C -#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I -#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C -#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I -#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C -#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I -#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C -#define TEMP_3_I TGSI_EXEC_TEMP_THREE_I -#define TEMP_3_C TGSI_EXEC_TEMP_THREE_C -#define TEMP_HALF_I TGSI_EXEC_TEMP_HALF_I -#define TEMP_HALF_C TGSI_EXEC_TEMP_HALF_C -#define TEMP_R0 TGSI_EXEC_TEMP_R0 - -#define FOR_EACH_CHANNEL(CHAN)\ - for (CHAN = 0; CHAN < 4; CHAN++) - -#define IS_CHANNEL_ENABLED(INST, CHAN)\ - ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) - -#define IS_CHANNEL_ENABLED2(INST, CHAN)\ - ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) - -#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED( INST, CHAN )) - -#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ - if (IS_CHANNEL_ENABLED2( INST, CHAN )) - - -/** The execution mask depends on the conditional mask and the loop mask */ -#define UPDATE_EXEC_MASK(MACH) \ - MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask - - -#define CHAN_X 0 -#define CHAN_Y 1 -#define CHAN_Z 2 -#define CHAN_W 3 - - - -/** - * Initialize machine state by expanding tokens to full instructions, - * allocating temporary storage, setting up constants, etc. - * After this, we can call tgsi_exec_machine_run() many times. - */ -void -tgsi_exec_machine_bind_shader( - struct tgsi_exec_machine *mach, - const struct tgsi_token *tokens, - uint numSamplers, - struct tgsi_sampler *samplers) -{ - uint k; - struct tgsi_parse_context parse; - struct tgsi_exec_labels *labels = &mach->Labels; - struct tgsi_full_instruction *instructions; - struct tgsi_full_declaration *declarations; - uint maxInstructions = 10, numInstructions = 0; - uint maxDeclarations = 10, numDeclarations = 0; - uint instno = 0; - -#if 0 - tgsi_dump(tokens, 0); -#endif - - mach->Tokens = tokens; - mach->Samplers = samplers; - - k = tgsi_parse_init (&parse, mach->Tokens); - if (k != TGSI_PARSE_OK) { - debug_printf( "Problem parsing!\n" ); - return; - } - - mach->Processor = parse.FullHeader.Processor.Processor; - mach->ImmLimit = 0; - labels->count = 0; - - declarations = (struct tgsi_full_declaration *) - MALLOC( maxDeclarations * sizeof(struct tgsi_full_declaration) ); - - if (!declarations) { - return; - } - - instructions = (struct tgsi_full_instruction *) - MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) ); - - if (!instructions) { - FREE( declarations ); - return; - } - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - uint pointer = parse.Position; - uint i; - - tgsi_parse_token( &parse ); - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - /* save expanded declaration */ - if (numDeclarations == maxDeclarations) { - declarations = REALLOC(declarations, - maxDeclarations - * sizeof(struct tgsi_full_declaration), - (maxDeclarations + 10) - * sizeof(struct tgsi_full_declaration)); - maxDeclarations += 10; - } - memcpy(declarations + numDeclarations, - &parse.FullToken.FullDeclaration, - sizeof(declarations[0])); - numDeclarations++; - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; - assert( size % 4 == 0 ); - assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES ); - - for( i = 0; i < size; i++ ) { - mach->Imms[mach->ImmLimit + i / 4][i % 4] = - parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; - } - mach->ImmLimit += size / 4; - } - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - assert( labels->count < MAX_LABELS ); - - labels->labels[labels->count][0] = instno; - labels->labels[labels->count][1] = pointer; - labels->count++; - - /* save expanded instruction */ - if (numInstructions == maxInstructions) { - instructions = REALLOC(instructions, - maxInstructions - * sizeof(struct tgsi_full_instruction), - (maxInstructions + 10) - * sizeof(struct tgsi_full_instruction)); - maxInstructions += 10; - } - memcpy(instructions + numInstructions, - &parse.FullToken.FullInstruction, - sizeof(instructions[0])); - numInstructions++; - break; - - default: - assert( 0 ); - } - } - tgsi_parse_free (&parse); - - if (mach->Declarations) { - FREE( mach->Declarations ); - } - mach->Declarations = declarations; - mach->NumDeclarations = numDeclarations; - - if (mach->Instructions) { - FREE( mach->Instructions ); - } - mach->Instructions = instructions; - mach->NumInstructions = numInstructions; -} - - -void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach ) -{ - uint i; - - mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); - mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR]; - - /* Setup constants. */ - for( i = 0; i < 4; i++ ) { - mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; - mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; - mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; - mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; - mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; - mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; - mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; - mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; - mach->Temps[TEMP_3_I].xyzw[TEMP_3_C].f[i] = 3.0f; - mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C].f[i] = 0.5f; - } -} - - -void -tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach) -{ - if (mach->Instructions) { - FREE(mach->Instructions); - mach->Instructions = NULL; - mach->NumInstructions = 0; - } - if (mach->Declarations) { - FREE(mach->Declarations); - mach->Declarations = NULL; - mach->NumDeclarations = 0; - } -} - - -static void -micro_abs( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = fabsf( src->f[0] ); - dst->f[1] = fabsf( src->f[1] ); - dst->f[2] = fabsf( src->f[2] ); - dst->f[3] = fabsf( src->f[3] ); -} - -static void -micro_add( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] + src1->f[0]; - dst->f[1] = src0->f[1] + src1->f[1]; - dst->f[2] = src0->f[2] + src1->f[2]; - dst->f[3] = src0->f[3] + src1->f[3]; -} - -static void -micro_iadd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] + src1->i[0]; - dst->i[1] = src0->i[1] + src1->i[1]; - dst->i[2] = src0->i[2] + src1->i[2]; - dst->i[3] = src0->i[3] + src1->i[3]; -} - -static void -micro_and( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] & src1->u[0]; - dst->u[1] = src0->u[1] & src1->u[1]; - dst->u[2] = src0->u[2] & src1->u[2]; - dst->u[3] = src0->u[3] & src1->u[3]; -} - -static void -micro_ceil( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = ceilf( src->f[0] ); - dst->f[1] = ceilf( src->f[1] ); - dst->f[2] = ceilf( src->f[2] ); - dst->f[3] = ceilf( src->f[3] ); -} - -static void -micro_cos( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = cosf( src->f[0] ); - dst->f[1] = cosf( src->f[1] ); - dst->f[2] = cosf( src->f[2] ); - dst->f[3] = cosf( src->f[3] ); -} - -static void -micro_ddx( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_ddy( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = - dst->f[1] = - dst->f[2] = - dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; -} - -static void -micro_div( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] / src1->f[0]; - dst->f[1] = src0->f[1] / src1->f[1]; - dst->f[2] = src0->f[2] / src1->f[2]; - dst->f[3] = src0->f[3] / src1->f[3]; -} - -static void -micro_udiv( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] / src1->u[0]; - dst->u[1] = src0->u[1] / src1->u[1]; - dst->u[2] = src0->u[2] / src1->u[2]; - dst->u[3] = src0->u[3] / src1->u[3]; -} - -static void -micro_eq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_ieq( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; -} - -static void -micro_exp2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src) -{ - dst->f[0] = powf( 2.0f, src->f[0] ); - dst->f[1] = powf( 2.0f, src->f[1] ); - dst->f[2] = powf( 2.0f, src->f[2] ); - dst->f[3] = powf( 2.0f, src->f[3] ); -} - -static void -micro_f2it( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->i[0] = (int) src->f[0]; - dst->i[1] = (int) src->f[1]; - dst->i[2] = (int) src->f[2]; - dst->i[3] = (int) src->f[3]; -} - -static void -micro_f2ut( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = (uint) src->f[0]; - dst->u[1] = (uint) src->f[1]; - dst->u[2] = (uint) src->f[2]; - dst->u[3] = (uint) src->f[3]; -} - -static void -micro_flr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = floorf( src->f[0] ); - dst->f[1] = floorf( src->f[1] ); - dst->f[2] = floorf( src->f[2] ); - dst->f[3] = floorf( src->f[3] ); -} - -static void -micro_frc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = src->f[0] - floorf( src->f[0] ); - dst->f[1] = src->f[1] - floorf( src->f[1] ); - dst->f[2] = src->f[2] - floorf( src->f[2] ); - dst->f[3] = src->f[3] - floorf( src->f[3] ); -} - -static void -micro_ge( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_i2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) src->i[0]; - dst->f[1] = (float) src->i[1]; - dst->f[2] = (float) src->i[2]; - dst->f[3] = (float) src->i[3]; -} - -static void -micro_lg2( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = logf( src->f[0] ) * 1.442695f; - dst->f[1] = logf( src->f[1] ) * 1.442695f; - dst->f[2] = logf( src->f[2] ) * 1.442695f; - dst->f[3] = logf( src->f[3] ) * 1.442695f; -} - -static void -micro_le( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] <= src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] <= src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] <= src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] <= src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_lt( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; -} - -static void -micro_ilt( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; -} - -static void -micro_ult( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2, - const union tgsi_exec_channel *src3 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; -} - -static void -micro_max( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; -} - -static void -micro_imax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; -} - -static void -micro_umax( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; -} - -static void -micro_min( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; - dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; - dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; - dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; -} - -static void -micro_imin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; - dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; - dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; - dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; -} - -static void -micro_umin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; - dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; - dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; - dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; -} - -static void -micro_umod( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] % src1->u[0]; - dst->u[1] = src0->u[1] % src1->u[1]; - dst->u[2] = src0->u[2] % src1->u[2]; - dst->u[3] = src0->u[3] % src1->u[3]; -} - -static void -micro_mul( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] * src1->f[0]; - dst->f[1] = src0->f[1] * src1->f[1]; - dst->f[2] = src0->f[2] * src1->f[2]; - dst->f[3] = src0->f[3] * src1->f[3]; -} - -static void -micro_imul( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] * src1->i[0]; - dst->i[1] = src0->i[1] * src1->i[1]; - dst->i[2] = src0->i[2] * src1->i[2]; - dst->i[3] = src0->i[3] * src1->i[3]; -} - -static void -micro_imul64( - union tgsi_exec_channel *dst0, - union tgsi_exec_channel *dst1, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst1->i[0] = src0->i[0] * src1->i[0]; - dst1->i[1] = src0->i[1] * src1->i[1]; - dst1->i[2] = src0->i[2] * src1->i[2]; - dst1->i[3] = src0->i[3] * src1->i[3]; - dst0->i[0] = 0; - dst0->i[1] = 0; - dst0->i[2] = 0; - dst0->i[3] = 0; -} - -static void -micro_umul64( - union tgsi_exec_channel *dst0, - union tgsi_exec_channel *dst1, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst1->u[0] = src0->u[0] * src1->u[0]; - dst1->u[1] = src0->u[1] * src1->u[1]; - dst1->u[2] = src0->u[2] * src1->u[2]; - dst1->u[3] = src0->u[3] * src1->u[3]; - dst0->u[0] = 0; - dst0->u[1] = 0; - dst0->u[2] = 0; - dst0->u[3] = 0; -} - -static void -micro_movc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1, - const union tgsi_exec_channel *src2 ) -{ - dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; - dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; - dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; - dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; -} - -static void -micro_neg( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = -src->f[0]; - dst->f[1] = -src->f[1]; - dst->f[2] = -src->f[2]; - dst->f[3] = -src->f[3]; -} - -static void -micro_ineg( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->i[0] = -src->i[0]; - dst->i[1] = -src->i[1]; - dst->i[2] = -src->i[2]; - dst->i[3] = -src->i[3]; -} - -static void -micro_not( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->u[0] = ~src->u[0]; - dst->u[1] = ~src->u[1]; - dst->u[2] = ~src->u[2]; - dst->u[3] = ~src->u[3]; -} - -static void -micro_or( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] | src1->u[0]; - dst->u[1] = src0->u[1] | src1->u[1]; - dst->u[2] = src0->u[2] | src1->u[2]; - dst->u[3] = src0->u[3] | src1->u[3]; -} - -static void -micro_pow( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = powf( src0->f[0], src1->f[0] ); - dst->f[1] = powf( src0->f[1], src1->f[1] ); - dst->f[2] = powf( src0->f[2], src1->f[2] ); - dst->f[3] = powf( src0->f[3], src1->f[3] ); -} - -static void -micro_rnd( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = floorf( src->f[0] + 0.5f ); - dst->f[1] = floorf( src->f[1] + 0.5f ); - dst->f[2] = floorf( src->f[2] + 0.5f ); - dst->f[3] = floorf( src->f[3] + 0.5f ); -} - -static void -micro_shl( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] << src1->i[0]; - dst->i[1] = src0->i[1] << src1->i[1]; - dst->i[2] = src0->i[2] << src1->i[2]; - dst->i[3] = src0->i[3] << src1->i[3]; -} - -static void -micro_ishr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->i[0] = src0->i[0] >> src1->i[0]; - dst->i[1] = src0->i[1] >> src1->i[1]; - dst->i[2] = src0->i[2] >> src1->i[2]; - dst->i[3] = src0->i[3] >> src1->i[3]; -} - -static void -micro_trunc( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0 ) -{ - dst->f[0] = (float) (int) src0->f[0]; - dst->f[1] = (float) (int) src0->f[1]; - dst->f[2] = (float) (int) src0->f[2]; - dst->f[3] = (float) (int) src0->f[3]; -} - -static void -micro_ushr( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] >> src1->u[0]; - dst->u[1] = src0->u[1] >> src1->u[1]; - dst->u[2] = src0->u[2] >> src1->u[2]; - dst->u[3] = src0->u[3] >> src1->u[3]; -} - -static void -micro_sin( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = sinf( src->f[0] ); - dst->f[1] = sinf( src->f[1] ); - dst->f[2] = sinf( src->f[2] ); - dst->f[3] = sinf( src->f[3] ); -} - -static void -micro_sqrt( union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = sqrtf( src->f[0] ); - dst->f[1] = sqrtf( src->f[1] ); - dst->f[2] = sqrtf( src->f[2] ); - dst->f[3] = sqrtf( src->f[3] ); -} - -static void -micro_sub( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->f[0] = src0->f[0] - src1->f[0]; - dst->f[1] = src0->f[1] - src1->f[1]; - dst->f[2] = src0->f[2] - src1->f[2]; - dst->f[3] = src0->f[3] - src1->f[3]; -} - -static void -micro_u2f( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src ) -{ - dst->f[0] = (float) src->u[0]; - dst->f[1] = (float) src->u[1]; - dst->f[2] = (float) src->u[2]; - dst->f[3] = (float) src->u[3]; -} - -static void -micro_xor( - union tgsi_exec_channel *dst, - const union tgsi_exec_channel *src0, - const union tgsi_exec_channel *src1 ) -{ - dst->u[0] = src0->u[0] ^ src1->u[0]; - dst->u[1] = src0->u[1] ^ src1->u[1]; - dst->u[2] = src0->u[2] ^ src1->u[2]; - dst->u[3] = src0->u[3] ^ src1->u[3]; -} - -static void -fetch_src_file_channel( - const struct tgsi_exec_machine *mach, - const uint file, - const uint swizzle, - const union tgsi_exec_channel *index, - union tgsi_exec_channel *chan ) -{ - switch( swizzle ) { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - switch( file ) { - case TGSI_FILE_CONSTANT: - chan->f[0] = mach->Consts[index->i[0]][swizzle]; - chan->f[1] = mach->Consts[index->i[1]][swizzle]; - chan->f[2] = mach->Consts[index->i[2]][swizzle]; - chan->f[3] = mach->Consts[index->i[3]][swizzle]; - break; - - case TGSI_FILE_INPUT: - chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_TEMPORARY: - assert(index->i[0] < TGSI_EXEC_NUM_TEMPS); - chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_IMMEDIATE: - assert( index->i[0] < (int) mach->ImmLimit ); - chan->f[0] = mach->Imms[index->i[0]][swizzle]; - assert( index->i[1] < (int) mach->ImmLimit ); - chan->f[1] = mach->Imms[index->i[1]][swizzle]; - assert( index->i[2] < (int) mach->ImmLimit ); - chan->f[2] = mach->Imms[index->i[2]][swizzle]; - assert( index->i[3] < (int) mach->ImmLimit ); - chan->f[3] = mach->Imms[index->i[3]][swizzle]; - break; - - case TGSI_FILE_ADDRESS: - chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; - break; - - case TGSI_FILE_OUTPUT: - /* vertex/fragment output vars can be read too */ - chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; - chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; - chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; - chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; - break; - - default: - assert( 0 ); - } - break; - - case TGSI_EXTSWIZZLE_ZERO: - *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; - break; - - case TGSI_EXTSWIZZLE_ONE: - *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; - break; - - default: - assert( 0 ); - } -} - -static void -fetch_source( - const struct tgsi_exec_machine *mach, - union tgsi_exec_channel *chan, - const struct tgsi_full_src_register *reg, - const uint chan_index ) -{ - union tgsi_exec_channel index; - uint swizzle; - - index.i[0] = - index.i[1] = - index.i[2] = - index.i[3] = reg->SrcRegister.Index; - - if (reg->SrcRegister.Indirect) { - union tgsi_exec_channel index2; - union tgsi_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterInd, CHAN_X ); - fetch_src_file_channel( - mach, - reg->SrcRegisterInd.File, - swizzle, - &index2, - &indir_index ); - - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; - } - - if( reg->SrcRegister.Dimension ) { - switch( reg->SrcRegister.File ) { - case TGSI_FILE_INPUT: - index.i[0] *= 17; - index.i[1] *= 17; - index.i[2] *= 17; - index.i[3] *= 17; - break; - case TGSI_FILE_CONSTANT: - index.i[0] *= 4096; - index.i[1] *= 4096; - index.i[2] *= 4096; - index.i[3] *= 4096; - break; - default: - assert( 0 ); - } - - index.i[0] += reg->SrcRegisterDim.Index; - index.i[1] += reg->SrcRegisterDim.Index; - index.i[2] += reg->SrcRegisterDim.Index; - index.i[3] += reg->SrcRegisterDim.Index; - - if (reg->SrcRegisterDim.Indirect) { - union tgsi_exec_channel index2; - union tgsi_exec_channel indir_index; - - index2.i[0] = - index2.i[1] = - index2.i[2] = - index2.i[3] = reg->SrcRegisterDimInd.Index; - - swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); - fetch_src_file_channel( - mach, - reg->SrcRegisterDimInd.File, - swizzle, - &index2, - &indir_index ); - - index.i[0] += indir_index.i[0]; - index.i[1] += indir_index.i[1]; - index.i[2] += indir_index.i[2]; - index.i[3] += indir_index.i[3]; - } - } - - swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); - fetch_src_file_channel( - mach, - reg->SrcRegister.File, - swizzle, - &index, - chan ); - - switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { - case TGSI_UTIL_SIGN_CLEAR: - micro_abs( chan, chan ); - break; - - case TGSI_UTIL_SIGN_SET: - micro_abs( chan, chan ); - micro_neg( chan, chan ); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - micro_neg( chan, chan ); - break; - - case TGSI_UTIL_SIGN_KEEP: - break; - } - - if (reg->SrcRegisterExtMod.Complement) { - micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan ); - } -} - -static void -store_dest( - struct tgsi_exec_machine *mach, - const union tgsi_exec_channel *chan, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - uint chan_index ) -{ - union tgsi_exec_channel *dst; - - switch( reg->DstRegister.File ) { - case TGSI_FILE_NULL: - return; - - case TGSI_FILE_OUTPUT: - dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] - + reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_TEMPORARY: - assert(reg->DstRegister.Index < TGSI_EXEC_NUM_TEMPS); - dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; - break; - - case TGSI_FILE_ADDRESS: - dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; - break; - - default: - assert( 0 ); - return; - } - - switch (inst->Instruction.Saturate) - { - case TGSI_SAT_NONE: - if (mach->ExecMask & 0x1) - dst->i[0] = chan->i[0]; - if (mach->ExecMask & 0x2) - dst->i[1] = chan->i[1]; - if (mach->ExecMask & 0x4) - dst->i[2] = chan->i[2]; - if (mach->ExecMask & 0x8) - dst->i[3] = chan->i[3]; - break; - - case TGSI_SAT_ZERO_ONE: - /* XXX need to obey ExecMask here */ - micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); - micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); - break; - - case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); - break; - - default: - assert( 0 ); - } -} - -#define FETCH(VAL,INDEX,CHAN)\ - fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) - -#define STORE(VAL,INDEX,CHAN)\ - store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) - - -/** - * Execute ARB-style KIL which is predicated by a src register. - * Kill fragment if any of the four values is less than zero. - */ -static void -exec_kilp(struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst) -{ - uint uniquemask; - uint chan_index; - uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ - union tgsi_exec_channel r[1]; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - for (chan_index = 0; chan_index < 4; chan_index++) - { - uint swizzle; - uint i; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle ( - &inst->FullSrcRegisters[0], - chan_index); - - /* check if the component has not been already tested */ - if (uniquemask & (1 << swizzle)) - continue; - uniquemask |= 1 << swizzle; - - FETCH(&r[0], 0, chan_index); - for (i = 0; i < 4; i++) - if (r[0].f[i] < 0.0f) - kilmask |= 1 << i; - } - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; -} - - -/* - * Fetch a texel using STR texture coordinates. - */ -static void -fetch_texel( struct tgsi_sampler *sampler, - const union tgsi_exec_channel *s, - const union tgsi_exec_channel *t, - const union tgsi_exec_channel *p, - float lodbias, /* XXX should be float[4] */ - union tgsi_exec_channel *r, - union tgsi_exec_channel *g, - union tgsi_exec_channel *b, - union tgsi_exec_channel *a ) -{ - uint j; - float rgba[NUM_CHANNELS][QUAD_SIZE]; - - sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); - - for (j = 0; j < 4; j++) { - r->f[j] = rgba[0][j]; - g->f[j] = rgba[1][j]; - b->f[j] = rgba[2][j]; - a->f[j] = rgba[3][j]; - } -} - - -static void -exec_tex(struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst, - boolean biasLod, - boolean projected) -{ - const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; - union tgsi_exec_channel r[8]; - uint chan_index; - float lodBias; - - /* debug_printf("Sampler %u unit %u\n", sampler, unit); */ - - switch (inst->InstructionExtTexture.Texture) { - case TGSI_TEXTURE_1D: - - FETCH(&r[0], 0, CHAN_X); - - if (projected) { - FETCH(&r[1], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[1] ); - } - - if (biasLod) { - FETCH(&r[1], 0, CHAN_W); - lodBias = r[2].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ - &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ - break; - - case TGSI_TEXTURE_2D: - case TGSI_TEXTURE_RECT: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - if (projected) { - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, /* inputs */ - &r[0], &r[1], &r[2], &r[3]); /* outputs */ - break; - - case TGSI_TEXTURE_3D: - case TGSI_TEXTURE_CUBE: - - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 0, CHAN_Z); - - if (projected) { - FETCH(&r[3], 0, CHAN_W); - micro_div( &r[0], &r[0], &r[3] ); - micro_div( &r[1], &r[1], &r[3] ); - micro_div( &r[2], &r[2], &r[3] ); - } - - if (biasLod) { - FETCH(&r[3], 0, CHAN_W); - lodBias = r[3].f[0]; - } - else - lodBias = 0.0; - - fetch_texel(&mach->Samplers[unit], - &r[0], &r[1], &r[2], lodBias, - &r[0], &r[1], &r[2], &r[3]); - break; - - default: - assert (0); - } - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[chan_index], 0, chan_index ); - } -} - - -/** - * Evaluate a constant-valued coefficient at the position of the - * current quad. - */ -static void -eval_constant_coef( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - unsigned i; - - for( i = 0; i < QUAD_SIZE; i++ ) { - mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; - } -} - -/** - * Evaluate a linear-valued coefficient at the position of the - * current quad. - */ -static void -eval_linear_coef( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - const float x = mach->QuadPos.xyzw[0].f[0]; - const float y = mach->QuadPos.xyzw[1].f[0]; - const float dadx = mach->InterpCoefs[attrib].dadx[chan]; - const float dady = mach->InterpCoefs[attrib].dady[chan]; - const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; - mach->Inputs[attrib].xyzw[chan].f[0] = a0; - mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; - mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; - mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; -} - -/** - * Evaluate a perspective-valued coefficient at the position of the - * current quad. - */ -static void -eval_perspective_coef( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ) -{ - const float x = mach->QuadPos.xyzw[0].f[0]; - const float y = mach->QuadPos.xyzw[1].f[0]; - const float dadx = mach->InterpCoefs[attrib].dadx[chan]; - const float dady = mach->InterpCoefs[attrib].dady[chan]; - const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; - const float *w = mach->QuadPos.xyzw[3].f; - /* divide by W here */ - mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; - mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; - mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; - mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; -} - - -typedef void (* eval_coef_func)( - struct tgsi_exec_machine *mach, - unsigned attrib, - unsigned chan ); - -static void -exec_declaration( - struct tgsi_exec_machine *mach, - const struct tgsi_full_declaration *decl ) -{ - if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { - if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask; - eval_coef_func eval; - - first = decl->DeclarationRange.First; - last = decl->DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - switch( decl->Declaration.Interpolate ) { - case TGSI_INTERPOLATE_CONSTANT: - eval = eval_constant_coef; - break; - - case TGSI_INTERPOLATE_LINEAR: - eval = eval_linear_coef; - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - eval = eval_perspective_coef; - break; - - default: - assert( 0 ); - } - - if( mask == TGSI_WRITEMASK_XYZW ) { - unsigned i, j; - - for( i = first; i <= last; i++ ) { - for( j = 0; j < NUM_CHANNELS; j++ ) { - eval( mach, i, j ); - } - } - } - else { - unsigned i, j; - - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - for( i = first; i <= last; i++ ) { - eval( mach, i, j ); - } - } - } - } - } - } -} - -static void -exec_instruction( - struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst, - int *pc ) -{ - uint chan_index; - union tgsi_exec_channel r[8]; - - (*pc)++; - - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_ARL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_f2it( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOV: - case TGSI_OPCODE_SWZ: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LIT: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_X ); - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[1], 0, CHAN_Y ); - micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - - FETCH( &r[2], 0, CHAN_W ); - micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); - micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); - micro_pow( &r[1], &r[1], &r[2] ); - micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, CHAN_Z ); - } - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH( &r[0], 0, CHAN_X ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH( &r[0], 0, CHAN_X ); - micro_sqrt( &r[0], &r[0] ); - micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXP: - FETCH( &r[0], 0, CHAN_X ); - micro_flr( &r[1], &r[0] ); /* r1 = floor(r0) */ - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - micro_exp2( &r[2], &r[1] ); /* r2 = 2 ^ r1 */ - STORE( &r[2], 0, CHAN_X ); /* store r2 */ - } - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - micro_sub( &r[2], &r[0], &r[1] ); /* r2 = r0 - r1 */ - STORE( &r[2], 0, CHAN_Y ); /* store r2 */ - } - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - micro_exp2( &r[2], &r[0] ); /* r2 = 2 ^ r0 */ - STORE( &r[2], 0, CHAN_Z ); /* store r2 */ - } - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_LOG: - FETCH( &r[0], 0, CHAN_X ); - micro_abs( &r[2], &r[0] ); /* r2 = abs(r0) */ - micro_lg2( &r[1], &r[2] ); /* r1 = lg2(r2) */ - micro_flr( &r[0], &r[1] ); /* r0 = floor(r1) */ - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &r[0], 0, CHAN_X ); - } - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - micro_exp2( &r[0], &r[0] ); /* r0 = 2 ^ r0 */ - micro_div( &r[0], &r[2], &r[0] ); /* r0 = r2 / r0 */ - STORE( &r[0], 0, CHAN_Y ); - } - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - STORE( &r[1], 0, CHAN_Z ); - } - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MUL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) - { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_mul( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_ADD: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Z ); - FETCH( &r[2], 1, CHAN_Z ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_W); - FETCH(&r[2], 1, CHAN_W); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DST: - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - FETCH( &r[0], 0, CHAN_Y ); - FETCH( &r[1], 1, CHAN_Y); - micro_mul( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, CHAN_Y ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - FETCH( &r[0], 0, CHAN_Z ); - STORE( &r[0], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - FETCH( &r[0], 1, CHAN_W ); - STORE( &r[0], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MIN: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - /* XXX use micro_min()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_MAX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - /* XXX use micro_max()?? */ - micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); - - STORE(&r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_mul( &r[0], &r[0], &r[1] ); - FETCH( &r[1], 2, chan_index ); - micro_add( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SUB: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - - micro_sub( &r[0], &r[0], &r[1] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - micro_sub( &r[1], &r[1], &r[2] ); - micro_mul( &r[0], &r[0], &r[1] ); - micro_add( &r[0], &r[0], &r[2] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_CND: - assert (0); - break; - - case TGSI_OPCODE_CND0: - assert (0); - break; - - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - assert (0); - break; - - case TGSI_OPCODE_INDEX: - assert (0); - break; - - case TGSI_OPCODE_NEGATE: - assert (0); - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_frc( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CLAMP: - assert (0); - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_flr( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_ROUND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_rnd( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ - FETCH(&r[0], 0, CHAN_X); - - micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ - FETCH( &r[0], 0, CHAN_X ); - micro_lg2( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_pow( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - FETCH(&r[0], 0, CHAN_Y); - FETCH(&r[1], 1, CHAN_Z); - - micro_mul( &r[2], &r[0], &r[1] ); - - FETCH(&r[3], 0, CHAN_Z); - FETCH(&r[4], 1, CHAN_Y); - - micro_mul( &r[5], &r[3], &r[4] ); - micro_sub( &r[2], &r[2], &r[5] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &r[2], 0, CHAN_X ); - } - - FETCH(&r[2], 1, CHAN_X); - - micro_mul( &r[3], &r[3], &r[2] ); - - FETCH(&r[5], 0, CHAN_X); - - micro_mul( &r[1], &r[1], &r[5] ); - micro_sub( &r[3], &r[3], &r[1] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { - STORE( &r[3], 0, CHAN_Y ); - } - - micro_mul( &r[5], &r[5], &r[4] ); - micro_mul( &r[0], &r[0], &r[2] ); - micro_sub( &r[5], &r[5], &r[0] ); - - if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - STORE( &r[5], 0, CHAN_Z ); - } - - if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MULTIPLYMATRIX: - assert (0); - break; - - case TGSI_OPCODE_ABS: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - - micro_abs( &r[0], &r[0] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_RCC: - assert (0); - break; - - case TGSI_OPCODE_DPH: - FETCH(&r[0], 0, CHAN_X); - FETCH(&r[1], 1, CHAN_X); - - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Y); - FETCH(&r[2], 1, CHAN_Y); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 0, CHAN_Z); - FETCH(&r[2], 1, CHAN_Z); - - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FETCH(&r[1], 1, CHAN_W); - - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_COS: - FETCH(&r[0], 0, CHAN_X); - - micro_cos( &r[0], &r[0] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDX: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddx( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDY: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ddy( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_KILP: - exec_kilp (mach, inst); - break; - - case TGSI_OPCODE_KIL: - /* for enabled ExecMask bits, set the killed bit */ - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; - break; - - case TGSI_OPCODE_PK2H: - assert (0); - break; - - case TGSI_OPCODE_PK2US: - assert (0); - break; - - case TGSI_OPCODE_PK4B: - assert (0); - break; - - case TGSI_OPCODE_PK4UB: - assert (0); - break; - - case TGSI_OPCODE_RFL: - assert (0); - break; - - case TGSI_OPCODE_SEQ: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], - &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], - &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SFL: - assert (0); - break; - - case TGSI_OPCODE_SGT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SIN: - FETCH( &r[0], 0, CHAN_X ); - micro_sin( &r[0], &r[0] ); - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SNE: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_STR: - assert (0); - break; - - case TGSI_OPCODE_TEX: - /* simple texture lookup */ - /* src[0] = texcoord */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, FALSE, FALSE); - break; - - case TGSI_OPCODE_TXB: - /* Texture lookup with lod bias */ - /* src[0] = texcoord (src[0].w = LOD bias) */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE, FALSE); - break; - - case TGSI_OPCODE_TXD: - /* Texture lookup with explict partial derivatives */ - /* src[0] = texcoord */ - /* src[1] = d[strq]/dx */ - /* src[2] = d[strq]/dy */ - /* src[3] = sampler unit */ - assert (0); - break; - - case TGSI_OPCODE_TXL: - /* Texture lookup with explit LOD */ - /* src[0] = texcoord (src[0].w = LOD) */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, TRUE, FALSE); - break; - - case TGSI_OPCODE_TXP: - /* Texture lookup with projection */ - /* src[0] = texcoord (src[0].w = projection) */ - /* src[1] = sampler unit */ - exec_tex(mach, inst, FALSE, TRUE); - break; - - case TGSI_OPCODE_UP2H: - assert (0); - break; - - case TGSI_OPCODE_UP2US: - assert (0); - break; - - case TGSI_OPCODE_UP4B: - assert (0); - break; - - case TGSI_OPCODE_UP4UB: - assert (0); - break; - - case TGSI_OPCODE_X2D: - assert (0); - break; - - case TGSI_OPCODE_ARA: - assert (0); - break; - - case TGSI_OPCODE_ARR: - assert (0); - break; - - case TGSI_OPCODE_BRA: - assert (0); - break; - - case TGSI_OPCODE_CAL: - /* skip the call if no execution channels are enabled */ - if (mach->ExecMask) { - /* do the call */ - - /* push the Cond, Loop, Cont stacks */ - assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); - mach->CondStack[mach->CondStackTop++] = mach->CondMask; - assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; - assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; - - assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); - mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; - - /* note that PC was already incremented above */ - mach->CallStack[mach->CallStackTop++] = *pc; - *pc = inst->InstructionExtLabel.Label; - } - break; - - case TGSI_OPCODE_RET: - mach->FuncMask &= ~mach->ExecMask; - UPDATE_EXEC_MASK(mach); - - if (mach->ExecMask == 0x0) { - /* really return now (otherwise, keep executing */ - - if (mach->CallStackTop == 0) { - /* returning from main() */ - *pc = -1; - return; - } - *pc = mach->CallStack[--mach->CallStackTop]; - - /* pop the Cond, Loop, Cont stacks */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; - assert(mach->FuncStackTop > 0); - mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; - - UPDATE_EXEC_MASK(mach); - } - break; - - case TGSI_OPCODE_SSG: - assert (0); - break; - - case TGSI_OPCODE_CMP: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH(&r[0], 0, chan_index); - FETCH(&r[1], 1, chan_index); - FETCH(&r[2], 2, chan_index); - - micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); - - STORE(&r[0], 0, chan_index); - } - break; - - case TGSI_OPCODE_SCS: - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( &r[0], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - micro_cos( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_X ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - micro_sin( &r[1], &r[0] ); - STORE( &r[1], 0, CHAN_Y ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); - } - if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_NRM: - assert (0); - break; - - case TGSI_OPCODE_DIV: - assert( 0 ); - break; - - case TGSI_OPCODE_DP2: - FETCH( &r[0], 0, CHAN_X ); - FETCH( &r[1], 1, CHAN_X ); - micro_mul( &r[0], &r[0], &r[1] ); - - FETCH( &r[1], 0, CHAN_Y ); - FETCH( &r[2], 1, CHAN_Y ); - micro_mul( &r[1], &r[1], &r[2] ); - micro_add( &r[0], &r[0], &r[1] ); - - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_IF: - /* push CondMask */ - assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); - mach->CondStack[mach->CondStackTop++] = mach->CondMask; - FETCH( &r[0], 0, CHAN_X ); - /* update CondMask */ - if( ! r[0].u[0] ) { - mach->CondMask &= ~0x1; - } - if( ! r[0].u[1] ) { - mach->CondMask &= ~0x2; - } - if( ! r[0].u[2] ) { - mach->CondMask &= ~0x4; - } - if( ! r[0].u[3] ) { - mach->CondMask &= ~0x8; - } - UPDATE_EXEC_MASK(mach); - /* Todo: If CondMask==0, jump to ELSE */ - break; - - case TGSI_OPCODE_ELSE: - /* invert CondMask wrt previous mask */ - { - uint prevMask; - assert(mach->CondStackTop > 0); - prevMask = mach->CondStack[mach->CondStackTop - 1]; - mach->CondMask = ~mach->CondMask & prevMask; - UPDATE_EXEC_MASK(mach); - /* Todo: If CondMask==0, jump to ENDIF */ - } - break; - - case TGSI_OPCODE_ENDIF: - /* pop CondMask */ - assert(mach->CondStackTop > 0); - mach->CondMask = mach->CondStack[--mach->CondStackTop]; - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_END: - /* halt execution */ - *pc = -1; - break; - - case TGSI_OPCODE_REP: - assert (0); - break; - - case TGSI_OPCODE_ENDREP: - assert (0); - break; - - case TGSI_OPCODE_PUSHA: - assert (0); - break; - - case TGSI_OPCODE_POPA: - assert (0); - break; - - case TGSI_OPCODE_CEIL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_ceil( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_I2F: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_i2f( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_NOT: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_not( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_TRUNC: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_trunc( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SHL: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_shl( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SHR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_ishr( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_AND: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_and( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_OR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_or( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_MOD: - assert (0); - break; - - case TGSI_OPCODE_XOR: - FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - FETCH( &r[1], 1, chan_index ); - micro_xor( &r[0], &r[0], &r[1] ); - STORE( &r[0], 0, chan_index ); - } - break; - - case TGSI_OPCODE_SAD: - assert (0); - break; - - case TGSI_OPCODE_TXF: - assert (0); - break; - - case TGSI_OPCODE_TXQ: - assert (0); - break; - - case TGSI_OPCODE_EMIT: - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; - break; - - case TGSI_OPCODE_ENDPRIM: - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; - mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; - break; - - case TGSI_OPCODE_LOOP: - /* fall-through (for now) */ - case TGSI_OPCODE_BGNLOOP2: - /* push LoopMask and ContMasks */ - assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; - assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); - mach->ContStack[mach->ContStackTop++] = mach->ContMask; - break; - - case TGSI_OPCODE_ENDLOOP: - /* fall-through (for now at least) */ - case TGSI_OPCODE_ENDLOOP2: - /* Restore ContMask, but don't pop */ - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; - UPDATE_EXEC_MASK(mach); - if (mach->ExecMask) { - /* repeat loop: jump to instruction just past BGNLOOP */ - *pc = inst->InstructionExtLabel.Label + 1; - } - else { - /* exit loop: pop LoopMask */ - assert(mach->LoopStackTop > 0); - mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; - /* pop ContMask */ - assert(mach->ContStackTop > 0); - mach->ContMask = mach->ContStack[--mach->ContStackTop]; - } - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_BRK: - /* turn off loop channels for each enabled exec channel */ - mach->LoopMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_CONT: - /* turn off cont channels for each enabled exec channel */ - mach->ContMask &= ~mach->ExecMask; - /* Todo: if mach->LoopMask == 0, jump to end of loop */ - UPDATE_EXEC_MASK(mach); - break; - - case TGSI_OPCODE_BGNSUB: - /* no-op */ - break; - - case TGSI_OPCODE_ENDSUB: - /* no-op */ - break; - - case TGSI_OPCODE_NOISE1: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE2: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE3: - assert( 0 ); - break; - - case TGSI_OPCODE_NOISE4: - assert( 0 ); - break; - - case TGSI_OPCODE_NOP: - break; - - default: - assert( 0 ); - } -} - - -/** - * Run TGSI interpreter. - * \return bitmask of "alive" quad components - */ -uint -tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) -{ - uint i; - int pc = 0; - - mach->CondMask = 0xf; - mach->LoopMask = 0xf; - mach->ContMask = 0xf; - mach->FuncMask = 0xf; - mach->ExecMask = 0xf; - - mach->CondStackTop = 0; /* temporarily subvert this assertion */ - assert(mach->CondStackTop == 0); - assert(mach->LoopStackTop == 0); - assert(mach->ContStackTop == 0); - assert(mach->CallStackTop == 0); - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; - mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; - - if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { - mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; - mach->Primitives[0] = 0; - } - - - /* execute declarations (interpolants) */ - for (i = 0; i < mach->NumDeclarations; i++) { - exec_declaration( mach, mach->Declarations+i ); - } - - /* execute instructions, until pc is set to -1 */ - while (pc != -1) { - assert(pc < (int) mach->NumInstructions); - exec_instruction( mach, mach->Instructions + pc, &pc ); - } - -#if 0 - /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ - if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { - /* - * Scale back depth component. - */ - for (i = 0; i < 4; i++) - mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; - } -#endif - - return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; -} - - diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h deleted file mode 100644 index 4f30650b07..0000000000 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h +++ /dev/null @@ -1,253 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#if !defined TGSI_EXEC_H -#define TGSI_EXEC_H - -#include "pipe/p_compiler.h" - -#if defined __cplusplus -extern "C" { -#endif - -#define MAX_LABELS 1024 - -#define NUM_CHANNELS 4 /* R,G,B,A */ -#define QUAD_SIZE 4 /* 4 pixel/quad */ - -/** - * Registers may be treated as float, signed int or unsigned int. - */ -union tgsi_exec_channel -{ - float f[QUAD_SIZE]; - int i[QUAD_SIZE]; - unsigned u[QUAD_SIZE]; -}; - -/** - * A vector[RGBA] of channels[4 pixels] - */ -struct tgsi_exec_vector -{ - union tgsi_exec_channel xyzw[NUM_CHANNELS]; -}; - -/** - * For fragment programs, information for computing fragment input - * values from plane equation of the triangle/line. - */ -struct tgsi_interp_coef -{ - float a0[NUM_CHANNELS]; /* in an xyzw layout */ - float dadx[NUM_CHANNELS]; - float dady[NUM_CHANNELS]; -}; - - -struct softpipe_tile_cache; /**< Opaque to TGSI */ - -/** - * Information for sampling textures, which must be implemented - * by code outside the TGSI executor. - */ -struct tgsi_sampler -{ - const struct pipe_sampler_state *state; - struct pipe_texture *texture; - /** Get samples for four fragments in a quad */ - void (*get_samples)(struct tgsi_sampler *sampler, - const float s[QUAD_SIZE], - const float t[QUAD_SIZE], - const float p[QUAD_SIZE], - float lodbias, - float rgba[NUM_CHANNELS][QUAD_SIZE]); - void *pipe; /*XXX temporary*/ - struct softpipe_tile_cache *cache; -}; - -/** - * For branching/calling subroutines. - */ -struct tgsi_exec_labels -{ - unsigned labels[MAX_LABELS][2]; - unsigned count; -}; - - -#define TGSI_EXEC_NUM_TEMPS 128 -#define TGSI_EXEC_NUM_TEMP_EXTRAS 6 -#define TGSI_EXEC_NUM_IMMEDIATES 256 - -/* - * Locations of various utility registers (_I = Index, _C = Channel) - */ -#define TGSI_EXEC_TEMP_00000000_I (TGSI_EXEC_NUM_TEMPS + 0) -#define TGSI_EXEC_TEMP_00000000_C 0 - -#define TGSI_EXEC_TEMP_7FFFFFFF_I (TGSI_EXEC_NUM_TEMPS + 0) -#define TGSI_EXEC_TEMP_7FFFFFFF_C 1 - -#define TGSI_EXEC_TEMP_80000000_I (TGSI_EXEC_NUM_TEMPS + 0) -#define TGSI_EXEC_TEMP_80000000_C 2 - -#define TGSI_EXEC_TEMP_FFFFFFFF_I (TGSI_EXEC_NUM_TEMPS + 0) -#define TGSI_EXEC_TEMP_FFFFFFFF_C 3 - -#define TGSI_EXEC_TEMP_ONE_I (TGSI_EXEC_NUM_TEMPS + 1) -#define TGSI_EXEC_TEMP_ONE_C 0 - -#define TGSI_EXEC_TEMP_TWO_I (TGSI_EXEC_NUM_TEMPS + 1) -#define TGSI_EXEC_TEMP_TWO_C 1 - -#define TGSI_EXEC_TEMP_128_I (TGSI_EXEC_NUM_TEMPS + 1) -#define TGSI_EXEC_TEMP_128_C 2 - -#define TGSI_EXEC_TEMP_MINUS_128_I (TGSI_EXEC_NUM_TEMPS + 1) -#define TGSI_EXEC_TEMP_MINUS_128_C 3 - -#define TGSI_EXEC_TEMP_KILMASK_I (TGSI_EXEC_NUM_TEMPS + 2) -#define TGSI_EXEC_TEMP_KILMASK_C 0 - -#define TGSI_EXEC_TEMP_OUTPUT_I (TGSI_EXEC_NUM_TEMPS + 2) -#define TGSI_EXEC_TEMP_OUTPUT_C 1 - -#define TGSI_EXEC_TEMP_PRIMITIVE_I (TGSI_EXEC_NUM_TEMPS + 2) -#define TGSI_EXEC_TEMP_PRIMITIVE_C 2 - -#define TGSI_EXEC_TEMP_THREE_I (TGSI_EXEC_NUM_TEMPS + 2) -#define TGSI_EXEC_TEMP_THREE_C 3 - -#define TGSI_EXEC_TEMP_HALF_I (TGSI_EXEC_NUM_TEMPS + 3) -#define TGSI_EXEC_TEMP_HALF_C 0 - -#define TGSI_EXEC_TEMP_R0 (TGSI_EXEC_NUM_TEMPS + 4) - -#define TGSI_EXEC_TEMP_ADDR (TGSI_EXEC_NUM_TEMPS + 5) - - -#define TGSI_EXEC_MAX_COND_NESTING 20 -#define TGSI_EXEC_MAX_LOOP_NESTING 20 -#define TGSI_EXEC_MAX_CALL_NESTING 20 - -/** - * Run-time virtual machine state for executing TGSI shader. - */ -struct tgsi_exec_machine -{ - /* Total = program temporaries + internal temporaries - * + 1 padding to align to 16 bytes - */ - struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + - TGSI_EXEC_NUM_TEMP_EXTRAS + 1]; - - /* - * This will point to _Temps after aligning to 16B boundary. - */ - struct tgsi_exec_vector *Temps; - struct tgsi_exec_vector *Addrs; - - struct tgsi_sampler *Samplers; - - float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; - unsigned ImmLimit; - const float (*Consts)[4]; - struct tgsi_exec_vector *Inputs; - struct tgsi_exec_vector *Outputs; - const struct tgsi_token *Tokens; - unsigned Processor; - - /* GEOMETRY processor only. */ - unsigned *Primitives; - - /* FRAGMENT processor only. */ - const struct tgsi_interp_coef *InterpCoefs; - struct tgsi_exec_vector QuadPos; - - /* Conditional execution masks */ - uint CondMask; /**< For IF/ELSE/ENDIF */ - uint LoopMask; /**< For BGNLOOP/ENDLOOP */ - uint ContMask; /**< For loop CONT statements */ - uint FuncMask; /**< For function calls */ - uint ExecMask; /**< = CondMask & LoopMask */ - - /** Condition mask stack (for nested conditionals) */ - uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; - int CondStackTop; - - /** Loop mask stack (for nested loops) */ - uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; - int LoopStackTop; - - /** Loop continue mask stack (see comments in tgsi_exec.c) */ - uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; - int ContStackTop; - - /** Function execution mask stack (for executing subroutine code) */ - uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; - int FuncStackTop; - - /** Function call stack for saving/restoring the program counter */ - uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; - int CallStackTop; - - struct tgsi_full_instruction *Instructions; - uint NumInstructions; - - struct tgsi_full_declaration *Declarations; - uint NumDeclarations; - - struct tgsi_exec_labels Labels; -}; - -void -tgsi_exec_machine_init( - struct tgsi_exec_machine *mach ); - - -void -tgsi_exec_machine_bind_shader( - struct tgsi_exec_machine *mach, - const struct tgsi_token *tokens, - uint numSamplers, - struct tgsi_sampler *samplers); - -uint -tgsi_exec_machine_run( - struct tgsi_exec_machine *mach ); - - -void -tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach); - - -#if defined __cplusplus -} /* extern "C" */ -#endif - -#endif /* TGSI_EXEC_H */ diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c deleted file mode 100755 index cdbdf5c882..0000000000 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c +++ /dev/null @@ -1,2275 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_util.h" -#include "tgsi_exec.h" -#include "tgsi_sse2.h" - -#include "rtasm/rtasm_x86sse.h" - -#ifdef PIPE_ARCH_X86 - -/* for 1/sqrt() - * - * This costs about 100fps (close to 10%) in gears: - */ -#define HIGH_PRECISION 1 - - -#define FOR_EACH_CHANNEL( CHAN )\ - for( CHAN = 0; CHAN < 4; CHAN++ ) - -#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ - ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) - -#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ - if( IS_DST0_CHANNEL_ENABLED( INST, CHAN )) - -#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\ - FOR_EACH_CHANNEL( CHAN )\ - IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN ) - -#define CHAN_X 0 -#define CHAN_Y 1 -#define CHAN_Z 2 -#define CHAN_W 3 - -#define TEMP_R0 TGSI_EXEC_TEMP_R0 - -/** - * X86 utility functions. - */ - -static struct x86_reg -make_xmm( - unsigned xmm ) -{ - return x86_make_reg( - file_XMM, - (enum x86_reg_name) xmm ); -} - -/** - * X86 register mapping helpers. - */ - -static struct x86_reg -get_const_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_CX ); -} - -static struct x86_reg -get_input_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_AX ); -} - -static struct x86_reg -get_output_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_DX ); -} - -static struct x86_reg -get_temp_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_BX ); -} - -static struct x86_reg -get_coef_base( void ) -{ - return get_output_base(); -} - -static struct x86_reg -get_immediate_base( void ) -{ - return x86_make_reg( - file_REG32, - reg_DI ); -} - - -/** - * Data access helpers. - */ - - -static struct x86_reg -get_immediate( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_immediate_base(), - (vec * 4 + chan) * 4 ); -} - -static struct x86_reg -get_const( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_const_base(), - (vec * 4 + chan) * 4 ); -} - -static struct x86_reg -get_input( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_input_base(), - (vec * 4 + chan) * 16 ); -} - -static struct x86_reg -get_output( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_output_base(), - (vec * 4 + chan) * 16 ); -} - -static struct x86_reg -get_temp( - unsigned vec, - unsigned chan ) -{ - return x86_make_disp( - get_temp_base(), - (vec * 4 + chan) * 16 ); -} - -static struct x86_reg -get_coef( - unsigned vec, - unsigned chan, - unsigned member ) -{ - return x86_make_disp( - get_coef_base(), - ((vec * 3 + member) * 4 + chan) * 4 ); -} - - -static void -emit_ret( - struct x86_function *func ) -{ - x86_ret( func ); -} - - -/** - * Data fetch helpers. - */ - -/** - * Copy a shader constant to xmm register - * \param xmm the destination xmm register - * \param vec the src const buffer index - * \param chan src channel to fetch (X, Y, Z or W) - */ -static void -emit_const( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movss( - func, - make_xmm( xmm ), - get_const( vec, chan ) ); - sse_shufps( - func, - make_xmm( xmm ), - make_xmm( xmm ), - SHUF( 0, 0, 0, 0 ) ); -} - -static void -emit_immediate( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movss( - func, - make_xmm( xmm ), - get_immediate( vec, chan ) ); - sse_shufps( - func, - make_xmm( xmm ), - make_xmm( xmm ), - SHUF( 0, 0, 0, 0 ) ); -} - - -/** - * Copy a shader input to xmm register - * \param xmm the destination xmm register - * \param vec the src input attrib - * \param chan src channel to fetch (X, Y, Z or W) - */ -static void -emit_inputf( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movups( - func, - make_xmm( xmm ), - get_input( vec, chan ) ); -} - -/** - * Store an xmm register to a shader output - * \param xmm the source xmm register - * \param vec the dest output attrib - * \param chan src dest channel to store (X, Y, Z or W) - */ -static void -emit_output( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movups( - func, - get_output( vec, chan ), - make_xmm( xmm ) ); -} - -/** - * Copy a shader temporary to xmm register - * \param xmm the destination xmm register - * \param vec the src temp register - * \param chan src channel to fetch (X, Y, Z or W) - */ -static void -emit_tempf( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movaps( - func, - make_xmm( xmm ), - get_temp( vec, chan ) ); -} - -/** - * Load an xmm register with an input attrib coefficient (a0, dadx or dady) - * \param xmm the destination xmm register - * \param vec the src input/attribute coefficient index - * \param chan src channel to fetch (X, Y, Z or W) - * \param member 0=a0, 1=dadx, 2=dady - */ -static void -emit_coef( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan, - unsigned member ) -{ - sse_movss( - func, - make_xmm( xmm ), - get_coef( vec, chan, member ) ); - sse_shufps( - func, - make_xmm( xmm ), - make_xmm( xmm ), - SHUF( 0, 0, 0, 0 ) ); -} - -/** - * Data store helpers. - */ - -static void -emit_inputs( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movups( - func, - get_input( vec, chan ), - make_xmm( xmm ) ); -} - -static void -emit_temps( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movaps( - func, - get_temp( vec, chan ), - make_xmm( xmm ) ); -} - -static void -emit_addrs( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_temps( - func, - xmm, - vec + TGSI_EXEC_NUM_TEMPS, - chan ); -} - -/** - * Coefficent fetch helpers. - */ - -static void -emit_coef_a0( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 0 ); -} - -static void -emit_coef_dadx( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 1 ); -} - -static void -emit_coef_dady( - struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - emit_coef( - func, - xmm, - vec, - chan, - 2 ); -} - -/** - * Function call helpers. - */ - -static void -emit_push_gp( - struct x86_function *func ) -{ - x86_push( - func, - x86_make_reg( file_REG32, reg_AX) ); - x86_push( - func, - x86_make_reg( file_REG32, reg_CX) ); - x86_push( - func, - x86_make_reg( file_REG32, reg_DX) ); -} - -static void -x86_pop_gp( - struct x86_function *func ) -{ - /* Restore GP registers in a reverse order. - */ - x86_pop( - func, - x86_make_reg( file_REG32, reg_DX) ); - x86_pop( - func, - x86_make_reg( file_REG32, reg_CX) ); - x86_pop( - func, - x86_make_reg( file_REG32, reg_AX) ); -} - -static void -emit_func_call_dst( - struct x86_function *func, - unsigned xmm_dst, - void (PIPE_CDECL *code)() ) -{ - sse_movaps( - func, - get_temp( TEMP_R0, 0 ), - make_xmm( xmm_dst ) ); - - emit_push_gp( - func ); - - { - struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); - - x86_lea( - func, - ecx, - get_temp( TEMP_R0, 0 ) ); - - x86_push( func, ecx ); - x86_mov_reg_imm( func, ecx, (unsigned long) code ); - x86_call( func, ecx ); - x86_pop(func, ecx ); - } - - - x86_pop_gp( - func ); - - sse_movaps( - func, - make_xmm( xmm_dst ), - get_temp( TEMP_R0, 0 ) ); -} - -static void -emit_func_call_dst_src( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src, - void (PIPE_CDECL *code)() ) -{ - sse_movaps( - func, - get_temp( TEMP_R0, 1 ), - make_xmm( xmm_src ) ); - - emit_func_call_dst( - func, - xmm_dst, - code ); -} - -/** - * Low-level instruction translators. - */ - -static void -emit_abs( - struct x86_function *func, - unsigned xmm ) -{ - sse_andps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_7FFFFFFF_I, - TGSI_EXEC_TEMP_7FFFFFFF_C ) ); -} - -static void -emit_add( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - sse_addps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void PIPE_CDECL -cos4f( - float *store ) -{ - const unsigned X = 0; - - store[X + 0] = cosf( store[X + 0] ); - store[X + 1] = cosf( store[X + 1] ); - store[X + 2] = cosf( store[X + 2] ); - store[X + 3] = cosf( store[X + 3] ); -} - -static void -emit_cos( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - cos4f ); -} - -static void PIPE_CDECL -ex24f( - float *store ) -{ - const unsigned X = 0; - - store[X + 0] = powf( 2.0f, store[X + 0] ); - store[X + 1] = powf( 2.0f, store[X + 1] ); - store[X + 2] = powf( 2.0f, store[X + 2] ); - store[X + 3] = powf( 2.0f, store[X + 3] ); -} - -static void -emit_ex2( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - ex24f ); -} - -static void -emit_f2it( - struct x86_function *func, - unsigned xmm ) -{ - sse2_cvttps2dq( - func, - make_xmm( xmm ), - make_xmm( xmm ) ); -} - -static void PIPE_CDECL -flr4f( - float *store ) -{ - const unsigned X = 0; - - store[X + 0] = floorf( store[X + 0] ); - store[X + 1] = floorf( store[X + 1] ); - store[X + 2] = floorf( store[X + 2] ); - store[X + 3] = floorf( store[X + 3] ); -} - -static void -emit_flr( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - flr4f ); -} - -static void PIPE_CDECL -frc4f( - float *store ) -{ - const unsigned X = 0; - - store[X + 0] -= floorf( store[X + 0] ); - store[X + 1] -= floorf( store[X + 1] ); - store[X + 2] -= floorf( store[X + 2] ); - store[X + 3] -= floorf( store[X + 3] ); -} - -static void -emit_frc( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - frc4f ); -} - -static void PIPE_CDECL -lg24f( - float *store ) -{ - const unsigned X = 0; - - store[X + 0] = LOG2( store[X + 0] ); - store[X + 1] = LOG2( store[X + 1] ); - store[X + 2] = LOG2( store[X + 2] ); - store[X + 3] = LOG2( store[X + 3] ); -} - -static void -emit_lg2( - struct x86_function *func, - unsigned xmm_dst ) -{ - emit_func_call_dst( - func, - xmm_dst, - lg24f ); -} - -static void -emit_MOV( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - sse_movups( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_mul (struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src) -{ - sse_mulps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_neg( - struct x86_function *func, - unsigned xmm ) -{ - sse_xorps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C ) ); -} - -static void PIPE_CDECL -pow4f( - float *store ) -{ - const unsigned X = 0; - - store[X + 0] = powf( store[X + 0], store[X + 4] ); - store[X + 1] = powf( store[X + 1], store[X + 5] ); - store[X + 2] = powf( store[X + 2], store[X + 6] ); - store[X + 3] = powf( store[X + 3], store[X + 7] ); -} - -static void -emit_pow( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - emit_func_call_dst_src( - func, - xmm_dst, - xmm_src, - pow4f ); -} - -static void -emit_rcp ( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - /* On Intel CPUs at least, this is only accurate to 12 bits -- not - * good enough. Need to either emit a proper divide or use the - * iterative technique described below in emit_rsqrt(). - */ - sse2_rcpps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -static void -emit_rsqrt( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ -#if HIGH_PRECISION - /* Although rsqrtps() and rcpps() are low precision on some/all SSE - * implementations, it is possible to improve its precision at - * fairly low cost, using a newton/raphson step, as below: - * - * x1 = 2 * rcpps(a) - a * rcpps(a) * rcpps(a) - * x1 = 0.5 * rsqrtps(a) * [3.0 - (a * rsqrtps(a))* rsqrtps(a)] - * - * See: http://softwarecommunity.intel.com/articles/eng/1818.htm - */ - { - struct x86_reg dst = make_xmm( xmm_dst ); - struct x86_reg src = make_xmm( xmm_src ); - struct x86_reg tmp0 = make_xmm( 2 ); - struct x86_reg tmp1 = make_xmm( 3 ); - - assert( xmm_dst != xmm_src ); - assert( xmm_dst != 2 && xmm_dst != 3 ); - assert( xmm_src != 2 && xmm_src != 3 ); - - sse_movaps( func, dst, get_temp( TGSI_EXEC_TEMP_HALF_I, TGSI_EXEC_TEMP_HALF_C ) ); - sse_movaps( func, tmp0, get_temp( TGSI_EXEC_TEMP_THREE_I, TGSI_EXEC_TEMP_THREE_C ) ); - sse_rsqrtps( func, tmp1, src ); - sse_mulps( func, src, tmp1 ); - sse_mulps( func, dst, tmp1 ); - sse_mulps( func, src, tmp1 ); - sse_subps( func, tmp0, src ); - sse_mulps( func, dst, tmp0 ); - } -#else - /* On Intel CPUs at least, this is only accurate to 12 bits -- not - * good enough. - */ - sse_rsqrtps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -#endif -} - -static void -emit_setsign( - struct x86_function *func, - unsigned xmm ) -{ - sse_orps( - func, - make_xmm( xmm ), - get_temp( - TGSI_EXEC_TEMP_80000000_I, - TGSI_EXEC_TEMP_80000000_C ) ); -} - -static void PIPE_CDECL -sin4f( - float *store ) -{ - const unsigned X = 0; - - store[X + 0] = sinf( store[X + 0] ); - store[X + 1] = sinf( store[X + 1] ); - store[X + 2] = sinf( store[X + 2] ); - store[X + 3] = sinf( store[X + 3] ); -} - -static void -emit_sin (struct x86_function *func, - unsigned xmm_dst) -{ - emit_func_call_dst( - func, - xmm_dst, - sin4f ); -} - -static void -emit_sub( - struct x86_function *func, - unsigned xmm_dst, - unsigned xmm_src ) -{ - sse_subps( - func, - make_xmm( xmm_dst ), - make_xmm( xmm_src ) ); -} - -/** - * Register fetch. - */ - -static void -emit_fetch( - struct x86_function *func, - unsigned xmm, - const struct tgsi_full_src_register *reg, - const unsigned chan_index ) -{ - unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); - - switch( swizzle ) { - case TGSI_EXTSWIZZLE_X: - case TGSI_EXTSWIZZLE_Y: - case TGSI_EXTSWIZZLE_Z: - case TGSI_EXTSWIZZLE_W: - switch( reg->SrcRegister.File ) { - case TGSI_FILE_CONSTANT: - emit_const( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - case TGSI_FILE_IMMEDIATE: - emit_immediate( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - case TGSI_FILE_INPUT: - emit_inputf( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - case TGSI_FILE_TEMPORARY: - emit_tempf( - func, - xmm, - reg->SrcRegister.Index, - swizzle ); - break; - - default: - assert( 0 ); - } - break; - - case TGSI_EXTSWIZZLE_ZERO: - emit_tempf( - func, - xmm, - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ); - break; - - case TGSI_EXTSWIZZLE_ONE: - emit_tempf( - func, - xmm, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - break; - - default: - assert( 0 ); - } - - switch( tgsi_util_get_full_src_register_sign_mode( reg, chan_index ) ) { - case TGSI_UTIL_SIGN_CLEAR: - emit_abs( func, xmm ); - break; - - case TGSI_UTIL_SIGN_SET: - emit_setsign( func, xmm ); - break; - - case TGSI_UTIL_SIGN_TOGGLE: - emit_neg( func, xmm ); - break; - - case TGSI_UTIL_SIGN_KEEP: - break; - } -} - -#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\ - emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN ) - -/** - * Register store. - */ - -static void -emit_store( - struct x86_function *func, - unsigned xmm, - const struct tgsi_full_dst_register *reg, - const struct tgsi_full_instruction *inst, - unsigned chan_index ) -{ - switch( reg->DstRegister.File ) { - case TGSI_FILE_OUTPUT: - emit_output( - func, - xmm, - reg->DstRegister.Index, - chan_index ); - break; - - case TGSI_FILE_TEMPORARY: - emit_temps( - func, - xmm, - reg->DstRegister.Index, - chan_index ); - break; - - case TGSI_FILE_ADDRESS: - emit_addrs( - func, - xmm, - reg->DstRegister.Index, - chan_index ); - break; - - default: - assert( 0 ); - } - - switch( inst->Instruction.Saturate ) { - case TGSI_SAT_NONE: - break; - - case TGSI_SAT_ZERO_ONE: - /* assert( 0 ); */ - break; - - case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); - break; - } -} - -#define STORE( FUNC, INST, XMM, INDEX, CHAN )\ - emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) - -/** - * High-level instruction translators. - */ - -static void -emit_kil( - struct x86_function *func, - const struct tgsi_full_src_register *reg ) -{ - unsigned uniquemask; - unsigned registers[4]; - unsigned nextregister = 0; - unsigned firstchan = ~0; - unsigned chan_index; - - /* This mask stores component bits that were already tested. Note that - * we test if the value is less than zero, so 1.0 and 0.0 need not to be - * tested. */ - uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); - - FOR_EACH_CHANNEL( chan_index ) { - unsigned swizzle; - - /* unswizzle channel */ - swizzle = tgsi_util_get_full_src_register_extswizzle( - reg, - chan_index ); - - /* check if the component has not been already tested */ - if( !(uniquemask & (1 << swizzle)) ) { - uniquemask |= 1 << swizzle; - - /* allocate register */ - registers[chan_index] = nextregister; - emit_fetch( - func, - nextregister, - reg, - chan_index ); - nextregister++; - - /* mark the first channel used */ - if( firstchan == ~0 ) { - firstchan = chan_index; - } - } - } - - x86_push( - func, - x86_make_reg( file_REG32, reg_AX ) ); - x86_push( - func, - x86_make_reg( file_REG32, reg_DX ) ); - - FOR_EACH_CHANNEL( chan_index ) { - if( uniquemask & (1 << chan_index) ) { - sse_cmpps( - func, - make_xmm( registers[chan_index] ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ), - cc_LessThan ); - - if( chan_index == firstchan ) { - sse_pmovmskb( - func, - x86_make_reg( file_REG32, reg_AX ), - make_xmm( registers[chan_index] ) ); - } - else { - sse_pmovmskb( - func, - x86_make_reg( file_REG32, reg_DX ), - make_xmm( registers[chan_index] ) ); - x86_or( - func, - x86_make_reg( file_REG32, reg_AX ), - x86_make_reg( file_REG32, reg_DX ) ); - } - } - } - - x86_or( - func, - get_temp( - TGSI_EXEC_TEMP_KILMASK_I, - TGSI_EXEC_TEMP_KILMASK_C ), - x86_make_reg( file_REG32, reg_AX ) ); - - x86_pop( - func, - x86_make_reg( file_REG32, reg_DX ) ); - x86_pop( - func, - x86_make_reg( file_REG32, reg_AX ) ); -} - -static void -emit_setcc( - struct x86_function *func, - struct tgsi_full_instruction *inst, - enum sse_cc cc ) -{ - unsigned chan_index; - - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - sse_cmpps( - func, - make_xmm( 0 ), - make_xmm( 1 ), - cc ); - sse_andps( - func, - make_xmm( 0 ), - get_temp( - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ) ); - STORE( func, *inst, 0, 0, chan_index ); - } -} - -static void -emit_cmp( - struct x86_function *func, - struct tgsi_full_instruction *inst ) -{ - unsigned chan_index; - - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - FETCH( func, *inst, 2, 2, chan_index ); - sse_cmpps( - func, - make_xmm( 0 ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ), - cc_LessThan ); - sse_andps( - func, - make_xmm( 1 ), - make_xmm( 0 ) ); - sse_andnps( - func, - make_xmm( 0 ), - make_xmm( 2 ) ); - sse_orps( - func, - make_xmm( 0 ), - make_xmm( 1 ) ); - STORE( func, *inst, 0, 0, chan_index ); - } -} - -static int -emit_instruction( - struct x86_function *func, - struct tgsi_full_instruction *inst ) -{ - unsigned chan_index; - - switch( inst->Instruction.Opcode ) { - case TGSI_OPCODE_ARL: -#if 0 - /* XXX this isn't working properly (see glean vertProg1 test) */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_f2it( func, 0 ); - STORE( func, *inst, 0, 0, chan_index ); - } -#else - return 0; -#endif - break; - - case TGSI_OPCODE_MOV: - case TGSI_OPCODE_SWZ: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_LIT: - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C); - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ) { - STORE( func, *inst, 0, 0, CHAN_X ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { - STORE( func, *inst, 0, 0, CHAN_W ); - } - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( func, *inst, 0, 0, CHAN_X ); - sse_maxps( - func, - make_xmm( 0 ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ) ); - STORE( func, *inst, 0, 0, CHAN_Y ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - /* XMM[1] = SrcReg[0].yyyy */ - FETCH( func, *inst, 1, 0, CHAN_Y ); - /* XMM[1] = max(XMM[1], 0) */ - sse_maxps( - func, - make_xmm( 1 ), - get_temp( - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ) ); - /* XMM[2] = SrcReg[0].wwww */ - FETCH( func, *inst, 2, 0, CHAN_W ); - /* XMM[2] = min(XMM[2], 128.0) */ - sse_minps( - func, - make_xmm( 2 ), - get_temp( - TGSI_EXEC_TEMP_128_I, - TGSI_EXEC_TEMP_128_C ) ); - /* XMM[2] = max(XMM[2], -128.0) */ - sse_maxps( - func, - make_xmm( 2 ), - get_temp( - TGSI_EXEC_TEMP_MINUS_128_I, - TGSI_EXEC_TEMP_MINUS_128_C ) ); - emit_pow( func, 1, 2 ); - FETCH( func, *inst, 0, 0, CHAN_X ); - sse_xorps( - func, - make_xmm( 2 ), - make_xmm( 2 ) ); - sse_cmpps( - func, - make_xmm( 2 ), - make_xmm( 0 ), - cc_LessThanEqual ); - sse_andps( - func, - make_xmm( 2 ), - make_xmm( 1 ) ); - STORE( func, *inst, 2, 0, CHAN_Z ); - } - } - break; - - case TGSI_OPCODE_RCP: - /* TGSI_OPCODE_RECIP */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_rcp( func, 0, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_RSQ: - /* TGSI_OPCODE_RECIPSQRT */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_rsqrt( func, 1, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 1, 0, chan_index ); - } - break; - - case TGSI_OPCODE_EXP: - return 0; - break; - - case TGSI_OPCODE_LOG: - return 0; - break; - - case TGSI_OPCODE_MUL: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_mul( func, 0, 1 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_ADD: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_add( func, 0, 1 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP3: - /* TGSI_OPCODE_DOT3 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_mul( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Y ); - FETCH( func, *inst, 2, 1, CHAN_Y ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Z ); - FETCH( func, *inst, 2, 1, CHAN_Z ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DP4: - /* TGSI_OPCODE_DOT4 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_mul( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Y ); - FETCH( func, *inst, 2, 1, CHAN_Y ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Z ); - FETCH( func, *inst, 2, 1, CHAN_Z ); - emit_mul(func, 1, 2 ); - emit_add(func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_W ); - FETCH( func, *inst, 2, 1, CHAN_W ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DST: - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - STORE( func, *inst, 0, 0, CHAN_X ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - FETCH( func, *inst, 1, 1, CHAN_Y ); - emit_mul( func, 0, 1 ); - STORE( func, *inst, 0, 0, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - FETCH( func, *inst, 0, 0, CHAN_Z ); - STORE( func, *inst, 0, 0, CHAN_Z ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - FETCH( func, *inst, 0, 1, CHAN_W ); - STORE( func, *inst, 0, 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MIN: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - sse_minps( - func, - make_xmm( 0 ), - make_xmm( 1 ) ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_MAX: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - sse_maxps( - func, - make_xmm( 0 ), - make_xmm( 1 ) ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLT: - /* TGSI_OPCODE_SETLT */ - emit_setcc( func, inst, cc_LessThan ); - break; - - case TGSI_OPCODE_SGE: - /* TGSI_OPCODE_SETGE */ - emit_setcc( func, inst, cc_NotLessThan ); - break; - - case TGSI_OPCODE_MAD: - /* TGSI_OPCODE_MADD */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - FETCH( func, *inst, 2, 2, chan_index ); - emit_mul( func, 0, 1 ); - emit_add( func, 0, 2 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_SUB: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - emit_sub( func, 0, 1 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_LERP: - /* TGSI_OPCODE_LRP */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - FETCH( func, *inst, 1, 1, chan_index ); - FETCH( func, *inst, 2, 2, chan_index ); - emit_sub( func, 1, 2 ); - emit_mul( func, 0, 1 ); - emit_add( func, 0, 2 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_CND: - return 0; - break; - - case TGSI_OPCODE_CND0: - return 0; - break; - - case TGSI_OPCODE_DOT2ADD: - /* TGSI_OPCODE_DP2A */ - return 0; - break; - - case TGSI_OPCODE_INDEX: - return 0; - break; - - case TGSI_OPCODE_NEGATE: - return 0; - break; - - case TGSI_OPCODE_FRAC: - /* TGSI_OPCODE_FRC */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_frc( func, 0 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_CLAMP: - return 0; - break; - - case TGSI_OPCODE_FLOOR: - /* TGSI_OPCODE_FLR */ - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_flr( func, 0 ); - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_ROUND: - return 0; - break; - - case TGSI_OPCODE_EXPBASE2: - /* TGSI_OPCODE_EX2 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_ex2( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_LOGBASE2: - /* TGSI_OPCODE_LG2 */ - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_lg2( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_POWER: - /* TGSI_OPCODE_POW */ - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_pow( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_CROSSPRODUCT: - /* TGSI_OPCODE_XPD */ - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { - FETCH( func, *inst, 1, 1, CHAN_Z ); - FETCH( func, *inst, 3, 0, CHAN_Z ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 0, 0, CHAN_Y ); - FETCH( func, *inst, 4, 1, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - emit_MOV( func, 2, 0 ); - emit_mul( func, 2, 1 ); - emit_MOV( func, 5, 3 ); - emit_mul( func, 5, 4 ); - emit_sub( func, 2, 5 ); - STORE( func, *inst, 2, 0, CHAN_X ); - } - if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || - IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { - FETCH( func, *inst, 2, 1, CHAN_X ); - FETCH( func, *inst, 5, 0, CHAN_X ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - emit_mul( func, 3, 2 ); - emit_mul( func, 1, 5 ); - emit_sub( func, 3, 1 ); - STORE( func, *inst, 3, 0, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - emit_mul( func, 5, 4 ); - emit_mul( func, 0, 2 ); - emit_sub( func, 5, 0 ); - STORE( func, *inst, 5, 0, CHAN_Z ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - STORE( func, *inst, 0, 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_MULTIPLYMATRIX: - return 0; - break; - - case TGSI_OPCODE_ABS: - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( func, *inst, 0, 0, chan_index ); - emit_abs( func, 0) ; - - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_RCC: - return 0; - break; - - case TGSI_OPCODE_DPH: - FETCH( func, *inst, 0, 0, CHAN_X ); - FETCH( func, *inst, 1, 1, CHAN_X ); - emit_mul( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Y ); - FETCH( func, *inst, 2, 1, CHAN_Y ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 0, CHAN_Z ); - FETCH( func, *inst, 2, 1, CHAN_Z ); - emit_mul( func, 1, 2 ); - emit_add( func, 0, 1 ); - FETCH( func, *inst, 1, 1, CHAN_W ); - emit_add( func, 0, 1 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_COS: - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_cos( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_DDX: - return 0; - break; - - case TGSI_OPCODE_DDY: - return 0; - break; - - case TGSI_OPCODE_KIL: - emit_kil( func, &inst->FullSrcRegisters[0] ); - break; - - case TGSI_OPCODE_PK2H: - return 0; - break; - - case TGSI_OPCODE_PK2US: - return 0; - break; - - case TGSI_OPCODE_PK4B: - return 0; - break; - - case TGSI_OPCODE_PK4UB: - return 0; - break; - - case TGSI_OPCODE_RFL: - return 0; - break; - - case TGSI_OPCODE_SEQ: - return 0; - break; - - case TGSI_OPCODE_SFL: - return 0; - break; - - case TGSI_OPCODE_SGT: - return 0; - break; - - case TGSI_OPCODE_SIN: - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_sin( func, 0 ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - break; - - case TGSI_OPCODE_SLE: - return 0; - break; - - case TGSI_OPCODE_SNE: - return 0; - break; - - case TGSI_OPCODE_STR: - return 0; - break; - - case TGSI_OPCODE_TEX: - if (0) { - /* Disable dummy texture code: - */ - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { - STORE( func, *inst, 0, 0, chan_index ); - } - } - else { - return 0; - } - break; - - case TGSI_OPCODE_TXD: - return 0; - break; - - case TGSI_OPCODE_UP2H: - return 0; - break; - - case TGSI_OPCODE_UP2US: - return 0; - break; - - case TGSI_OPCODE_UP4B: - return 0; - break; - - case TGSI_OPCODE_UP4UB: - return 0; - break; - - case TGSI_OPCODE_X2D: - return 0; - break; - - case TGSI_OPCODE_ARA: - return 0; - break; - - case TGSI_OPCODE_ARR: - return 0; - break; - - case TGSI_OPCODE_BRA: - return 0; - break; - - case TGSI_OPCODE_CAL: - return 0; - break; - - case TGSI_OPCODE_RET: - emit_ret( func ); - break; - - case TGSI_OPCODE_END: - break; - - case TGSI_OPCODE_SSG: - return 0; - break; - - case TGSI_OPCODE_CMP: - emit_cmp (func, inst); - break; - - case TGSI_OPCODE_SCS: - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_cos( func, 0 ); - STORE( func, *inst, 0, 0, CHAN_X ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { - FETCH( func, *inst, 0, 0, CHAN_X ); - emit_sin( func, 0 ); - STORE( func, *inst, 0, 0, CHAN_Y ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_00000000_I, - TGSI_EXEC_TEMP_00000000_C ); - STORE( func, *inst, 0, 0, CHAN_Z ); - } - IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { - emit_tempf( - func, - 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); - STORE( func, *inst, 0, 0, CHAN_W ); - } - break; - - case TGSI_OPCODE_TXB: - return 0; - break; - - case TGSI_OPCODE_NRM: - return 0; - break; - - case TGSI_OPCODE_DIV: - return 0; - break; - - case TGSI_OPCODE_DP2: - return 0; - break; - - case TGSI_OPCODE_TXL: - return 0; - break; - - case TGSI_OPCODE_BRK: - return 0; - break; - - case TGSI_OPCODE_IF: - return 0; - break; - - case TGSI_OPCODE_LOOP: - return 0; - break; - - case TGSI_OPCODE_REP: - return 0; - break; - - case TGSI_OPCODE_ELSE: - return 0; - break; - - case TGSI_OPCODE_ENDIF: - return 0; - break; - - case TGSI_OPCODE_ENDLOOP: - return 0; - break; - - case TGSI_OPCODE_ENDREP: - return 0; - break; - - case TGSI_OPCODE_PUSHA: - return 0; - break; - - case TGSI_OPCODE_POPA: - return 0; - break; - - case TGSI_OPCODE_CEIL: - return 0; - break; - - case TGSI_OPCODE_I2F: - return 0; - break; - - case TGSI_OPCODE_NOT: - return 0; - break; - - case TGSI_OPCODE_TRUNC: - return 0; - break; - - case TGSI_OPCODE_SHL: - return 0; - break; - - case TGSI_OPCODE_SHR: - return 0; - break; - - case TGSI_OPCODE_AND: - return 0; - break; - - case TGSI_OPCODE_OR: - return 0; - break; - - case TGSI_OPCODE_MOD: - return 0; - break; - - case TGSI_OPCODE_XOR: - return 0; - break; - - case TGSI_OPCODE_SAD: - return 0; - break; - - case TGSI_OPCODE_TXF: - return 0; - break; - - case TGSI_OPCODE_TXQ: - return 0; - break; - - case TGSI_OPCODE_CONT: - return 0; - break; - - case TGSI_OPCODE_EMIT: - return 0; - break; - - case TGSI_OPCODE_ENDPRIM: - return 0; - break; - - default: - return 0; - } - - return 1; -} - -static void -emit_declaration( - struct x86_function *func, - struct tgsi_full_declaration *decl ) -{ - if( decl->Declaration.File == TGSI_FILE_INPUT ) { - unsigned first, last, mask; - unsigned i, j; - - first = decl->DeclarationRange.First; - last = decl->DeclarationRange.Last; - mask = decl->Declaration.UsageMask; - - for( i = first; i <= last; i++ ) { - for( j = 0; j < NUM_CHANNELS; j++ ) { - if( mask & (1 << j) ) { - switch( decl->Declaration.Interpolate ) { - case TGSI_INTERPOLATE_CONSTANT: - emit_coef_a0( func, 0, i, j ); - emit_inputs( func, 0, i, j ); - break; - - case TGSI_INTERPOLATE_LINEAR: - emit_tempf( func, 0, 0, TGSI_SWIZZLE_X ); - emit_coef_dadx( func, 1, i, j ); - emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y ); - emit_coef_dady( func, 3, i, j ); - emit_mul( func, 0, 1 ); /* x * dadx */ - emit_coef_a0( func, 4, i, j ); - emit_mul( func, 2, 3 ); /* y * dady */ - emit_add( func, 0, 4 ); /* x * dadx + a0 */ - emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ - emit_inputs( func, 0, i, j ); - break; - - case TGSI_INTERPOLATE_PERSPECTIVE: - emit_tempf( func, 0, 0, TGSI_SWIZZLE_X ); - emit_coef_dadx( func, 1, i, j ); - emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y ); - emit_coef_dady( func, 3, i, j ); - emit_mul( func, 0, 1 ); /* x * dadx */ - emit_tempf( func, 4, 0, TGSI_SWIZZLE_W ); - emit_coef_a0( func, 5, i, j ); - emit_rcp( func, 4, 4 ); /* 1.0 / w */ - emit_mul( func, 2, 3 ); /* y * dady */ - emit_add( func, 0, 5 ); /* x * dadx + a0 */ - emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ - emit_mul( func, 0, 4 ); /* (x * dadx + y * dady + a0) / w */ - emit_inputs( func, 0, i, j ); - break; - - default: - assert( 0 ); - break; - } - } - } - } - } -} - -static void aos_to_soa( struct x86_function *func, - uint arg_aos, - uint arg_soa, - uint arg_num, - uint arg_stride ) -{ - struct x86_reg soa_input = x86_make_reg( file_REG32, reg_AX ); - struct x86_reg aos_input = x86_make_reg( file_REG32, reg_BX ); - struct x86_reg num_inputs = x86_make_reg( file_REG32, reg_CX ); - struct x86_reg stride = x86_make_reg( file_REG32, reg_DX ); - int inner_loop; - - - /* Save EBX */ - x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); - - x86_mov( func, aos_input, x86_fn_arg( func, arg_aos ) ); - x86_mov( func, soa_input, x86_fn_arg( func, arg_soa ) ); - x86_mov( func, num_inputs, x86_fn_arg( func, arg_num ) ); - x86_mov( func, stride, x86_fn_arg( func, arg_stride ) ); - - /* do */ - inner_loop = x86_get_label( func ); - { - x86_push( func, aos_input ); - sse_movlps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); - sse_movlps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); - x86_add( func, aos_input, stride ); - sse_movhps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); - sse_movhps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); - x86_add( func, aos_input, stride ); - sse_movlps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); - sse_movlps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); - x86_add( func, aos_input, stride ); - sse_movhps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); - sse_movhps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); - x86_pop( func, aos_input ); - - sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); - sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); - sse_shufps( func, make_xmm( 0 ), make_xmm( 1 ), 0x88 ); - sse_shufps( func, make_xmm( 2 ), make_xmm( 1 ), 0xdd ); - sse_shufps( func, make_xmm( 3 ), make_xmm( 4 ), 0x88 ); - sse_shufps( func, make_xmm( 5 ), make_xmm( 4 ), 0xdd ); - - sse_movups( func, x86_make_disp( soa_input, 0 ), make_xmm( 0 ) ); - sse_movups( func, x86_make_disp( soa_input, 16 ), make_xmm( 2 ) ); - sse_movups( func, x86_make_disp( soa_input, 32 ), make_xmm( 3 ) ); - sse_movups( func, x86_make_disp( soa_input, 48 ), make_xmm( 5 ) ); - - /* Advance to next input */ - x86_lea( func, aos_input, x86_make_disp(aos_input, 16) ); - x86_lea( func, soa_input, x86_make_disp(soa_input, 64) ); - } - /* while --num_inputs */ - x86_dec( func, num_inputs ); - x86_jcc( func, cc_NE, inner_loop ); - - /* Restore EBX */ - x86_pop( func, aos_input ); -} - -static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, uint stride ) -{ - struct x86_reg soa_output; - struct x86_reg aos_output; - struct x86_reg num_outputs; - struct x86_reg temp; - int inner_loop; - - soa_output = x86_make_reg( file_REG32, reg_AX ); - aos_output = x86_make_reg( file_REG32, reg_BX ); - num_outputs = x86_make_reg( file_REG32, reg_CX ); - temp = x86_make_reg( file_REG32, reg_DX ); - - /* Save EBX */ - x86_push( func, aos_output ); - - x86_mov( func, soa_output, x86_fn_arg( func, soa ) ); - x86_mov( func, aos_output, x86_fn_arg( func, aos ) ); - x86_mov( func, num_outputs, x86_fn_arg( func, num ) ); - - /* do */ - inner_loop = x86_get_label( func ); - { - sse_movups( func, make_xmm( 0 ), x86_make_disp( soa_output, 0 ) ); - sse_movups( func, make_xmm( 1 ), x86_make_disp( soa_output, 16 ) ); - sse_movups( func, make_xmm( 3 ), x86_make_disp( soa_output, 32 ) ); - sse_movups( func, make_xmm( 4 ), x86_make_disp( soa_output, 48 ) ); - - sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); - sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); - sse_unpcklps( func, make_xmm( 0 ), make_xmm( 1 ) ); - sse_unpckhps( func, make_xmm( 2 ), make_xmm( 1 ) ); - sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) ); - sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) ); - - x86_mov( func, temp, x86_fn_arg( func, stride ) ); - x86_push( func, aos_output ); - sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); - sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); - x86_add( func, aos_output, temp ); - sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); - sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); - x86_add( func, aos_output, temp ); - sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); - sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); - x86_add( func, aos_output, temp ); - sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); - sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); - x86_pop( func, aos_output ); - - /* Advance to next output */ - x86_lea( func, aos_output, x86_make_disp(aos_output, 16) ); - x86_lea( func, soa_output, x86_make_disp(soa_output, 64) ); - } - /* while --num_outputs */ - x86_dec( func, num_outputs ); - x86_jcc( func, cc_NE, inner_loop ); - - /* Restore EBX */ - x86_pop( func, aos_output ); -} - -/** - * Translate a TGSI vertex/fragment shader to SSE2 code. - * Slightly different things are done for vertex vs. fragment shaders. - * - * Note that fragment shaders are responsible for interpolating shader - * inputs. Because on x86 we have only 4 GP registers, and here we - * have 5 shader arguments (input, output, const, temp and coef), the - * code is split into two phases -- DECLARATION and INSTRUCTION phase. - * GP register holding the output argument is aliased with the coeff - * argument, as outputs are not needed in the DECLARATION phase. - * - * \param tokens the TGSI input shader - * \param func the output SSE code/function - * \param immediates buffer to place immediates, later passed to SSE func - * \param return 1 for success, 0 if translation failed - */ -unsigned -tgsi_emit_sse2( - const struct tgsi_token *tokens, - struct x86_function *func, - float (*immediates)[4], - boolean do_swizzles ) -{ - struct tgsi_parse_context parse; - boolean instruction_phase = FALSE; - unsigned ok = 1; - uint num_immediates = 0; - - func->csr = func->store; - - tgsi_parse_init( &parse, tokens ); - - /* Can't just use EDI, EBX without save/restoring them: - */ - x86_push( - func, - get_immediate_base() ); - - x86_push( - func, - get_temp_base() ); - - - /* - * Different function args for vertex/fragment shaders: - */ - if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { - /* DECLARATION phase, do not load output argument. */ - x86_mov( - func, - get_input_base(), - x86_fn_arg( func, 1 ) ); - /* skipping outputs argument here */ - x86_mov( - func, - get_const_base(), - x86_fn_arg( func, 3 ) ); - x86_mov( - func, - get_temp_base(), - x86_fn_arg( func, 4 ) ); - x86_mov( - func, - get_coef_base(), - x86_fn_arg( func, 5 ) ); - x86_mov( - func, - get_immediate_base(), - x86_fn_arg( func, 6 ) ); - } - else { - assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX); - - if (do_swizzles) - aos_to_soa( func, - 6, /* aos_input */ - 1, /* machine->input */ - 7, /* num_inputs */ - 8 ); /* input_stride */ - - x86_mov( - func, - get_input_base(), - x86_fn_arg( func, 1 ) ); - x86_mov( - func, - get_output_base(), - x86_fn_arg( func, 2 ) ); - x86_mov( - func, - get_const_base(), - x86_fn_arg( func, 3 ) ); - x86_mov( - func, - get_temp_base(), - x86_fn_arg( func, 4 ) ); - x86_mov( - func, - get_immediate_base(), - x86_fn_arg( func, 5 ) ); - } - - while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { - emit_declaration( - func, - &parse.FullToken.FullDeclaration ); - } - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { - if( !instruction_phase ) { - /* INSTRUCTION phase, overwrite coeff with output. */ - instruction_phase = TRUE; - x86_mov( - func, - get_output_base(), - x86_fn_arg( func, 2 ) ); - } - } - - ok = emit_instruction( - func, - &parse.FullToken.FullInstruction ); - - if (!ok) { - debug_printf("failed to translate tgsi opcode %d to SSE (%s)\n", - parse.FullToken.FullInstruction.Instruction.Opcode, - parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ? - "vertex shader" : "fragment shader"); - } - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - /* simply copy the immediate values into the next immediates[] slot */ - { - const uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; - uint i; - assert(size <= 4); - assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES); - for( i = 0; i < size; i++ ) { - immediates[num_immediates][i] = - parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; - } -#if 0 - debug_printf("SSE FS immediate[%d] = %f %f %f %f\n", - num_immediates, - immediates[num_immediates][0], - immediates[num_immediates][1], - immediates[num_immediates][2], - immediates[num_immediates][3]); -#endif - num_immediates++; - } - break; - - default: - ok = 0; - assert( 0 ); - } - } - - if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) { - if (do_swizzles) - soa_to_aos( func, 9, 2, 10, 11 ); - } - - /* Can't just use EBX, EDI without save/restoring them: - */ - x86_pop( - func, - get_temp_base() ); - - x86_pop( - func, - get_immediate_base() ); - - emit_ret( func ); - - tgsi_parse_free( &parse ); - - return ok; -} - -#endif /* PIPE_ARCH_X86 */ diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h deleted file mode 100755 index af838b2a25..0000000000 --- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.h +++ /dev/null @@ -1,49 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_SSE2_H -#define TGSI_SSE2_H - -#if defined __cplusplus -extern "C" { -#endif - -struct tgsi_token; -struct x86_function; - -unsigned -tgsi_emit_sse2( - const struct tgsi_token *tokens, - struct x86_function *function, - float (*immediates)[4], - boolean do_swizzles ); - -#if defined __cplusplus -} -#endif - -#endif /* TGSI_SSE2_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c new file mode 100644 index 0000000000..742ef14c35 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_build.c @@ -0,0 +1,1324 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_build.h" +#include "tgsi_parse.h" + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ) +{ + struct tgsi_version version; + + version.MajorVersion = 1; + version.MinorVersion = 1; + version.Padding = 0; + + return version; +} + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ) +{ + struct tgsi_header header; + + header.HeaderSize = 1; + header.BodySize = 0; + + return header; +} + +static void +header_headersize_grow( struct tgsi_header *header ) +{ + assert( header->HeaderSize < 0xFF ); + assert( header->BodySize == 0 ); + + header->HeaderSize++; +} + +static void +header_bodysize_grow( struct tgsi_header *header ) +{ + assert( header->BodySize < 0xFFFFFF ); + + header->BodySize++; +} + +struct tgsi_processor +tgsi_default_processor( void ) +{ + struct tgsi_processor processor; + + processor.Processor = TGSI_PROCESSOR_FRAGMENT; + processor.Padding = 0; + + return processor; +} + +struct tgsi_processor +tgsi_build_processor( + unsigned type, + struct tgsi_header *header ) +{ + struct tgsi_processor processor; + + processor = tgsi_default_processor(); + processor.Processor = type; + + header_headersize_grow( header ); + + return processor; +} + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ) +{ + struct tgsi_declaration declaration; + + declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; + declaration.Size = 1; + declaration.File = TGSI_FILE_NULL; + declaration.UsageMask = TGSI_WRITEMASK_XYZW; + declaration.Interpolate = TGSI_INTERPOLATE_CONSTANT; + declaration.Semantic = 0; + declaration.Padding = 0; + declaration.Extended = 0; + + return declaration; +} + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ) +{ + struct tgsi_declaration declaration; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); + + declaration = tgsi_default_declaration(); + declaration.File = file; + declaration.UsageMask = usage_mask; + declaration.Interpolate = interpolate; + declaration.Semantic = semantic; + + header_bodysize_grow( header ); + + return declaration; +} + +static void +declaration_grow( + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + assert( declaration->Size < 0xFF ); + + declaration->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ) +{ + struct tgsi_full_declaration full_declaration; + + full_declaration.Declaration = tgsi_default_declaration(); + full_declaration.DeclarationRange = tgsi_default_declaration_range(); + full_declaration.Semantic = tgsi_default_declaration_semantic(); + + return full_declaration; +} + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0; + struct tgsi_declaration *declaration; + struct tgsi_declaration_range *dr; + + if( maxsize <= size ) + return 0; + declaration = (struct tgsi_declaration *) &tokens[size]; + size++; + + *declaration = tgsi_build_declaration( + full_decl->Declaration.File, + full_decl->Declaration.UsageMask, + full_decl->Declaration.Interpolate, + full_decl->Declaration.Semantic, + header ); + + if (maxsize <= size) + return 0; + dr = (struct tgsi_declaration_range *) &tokens[size]; + size++; + + *dr = tgsi_build_declaration_range( + full_decl->DeclarationRange.First, + full_decl->DeclarationRange.Last, + declaration, + header ); + + if( full_decl->Declaration.Semantic ) { + struct tgsi_declaration_semantic *ds; + + if( maxsize <= size ) + return 0; + ds = (struct tgsi_declaration_semantic *) &tokens[size]; + size++; + + *ds = tgsi_build_declaration_semantic( + full_decl->Semantic.SemanticName, + full_decl->Semantic.SemanticIndex, + declaration, + header ); + } + + return size; +} + +struct tgsi_declaration_range +tgsi_default_declaration_range( void ) +{ + struct tgsi_declaration_range dr; + + dr.First = 0; + dr.Last = 0; + + return dr; +} + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_range declaration_range; + + assert( last >= first ); + assert( last <= 0xFFFF ); + + declaration_range = tgsi_default_declaration_range(); + declaration_range.First = first; + declaration_range.Last = last; + + declaration_grow( declaration, header ); + + return declaration_range; +} + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ) +{ + struct tgsi_declaration_semantic ds; + + ds.SemanticName = TGSI_SEMANTIC_POSITION; + ds.SemanticIndex = 0; + ds.Padding = 0; + + return ds; +} + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ) +{ + struct tgsi_declaration_semantic ds; + + assert( semantic_name <= TGSI_SEMANTIC_COUNT ); + assert( semantic_index <= 0xFFFF ); + + ds = tgsi_default_declaration_semantic(); + ds.SemanticName = semantic_name; + ds.SemanticIndex = semantic_index; + + declaration_grow( declaration, header ); + + return ds; +} + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ) +{ + struct tgsi_immediate immediate; + + immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; + immediate.Size = 1; + immediate.DataType = TGSI_IMM_FLOAT32; + immediate.Padding = 0; + immediate.Extended = 0; + + return immediate; +} + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ) +{ + struct tgsi_immediate immediate; + + immediate = tgsi_default_immediate(); + + header_bodysize_grow( header ); + + return immediate; +} + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ) +{ + struct tgsi_full_immediate fullimm; + + fullimm.Immediate = tgsi_default_immediate(); + fullimm.u.Pointer = (void *) 0; + + return fullimm; +} + +static void +immediate_grow( + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + assert( immediate->Size < 0xFF ); + + immediate->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + float value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ) +{ + struct tgsi_immediate_float32 immediate_float32; + + immediate_float32.Float = value; + + immediate_grow( immediate, header ); + + return immediate_float32; +} + +unsigned +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0, i; + struct tgsi_immediate *immediate; + + if( maxsize <= size ) + return 0; + immediate = (struct tgsi_immediate *) &tokens[size]; + size++; + + *immediate = tgsi_build_immediate( header ); + + for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { + struct tgsi_immediate_float32 *if32; + + if( maxsize <= size ) + return 0; + if32 = (struct tgsi_immediate_float32 *) &tokens[size]; + size++; + + *if32 = tgsi_build_immediate_float32( + full_imm->u.ImmediateFloat32[i].Float, + immediate, + header ); + } + + return size; +} + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ) +{ + struct tgsi_instruction instruction; + + instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; + instruction.Size = 1; + instruction.Opcode = TGSI_OPCODE_MOV; + instruction.Saturate = TGSI_SAT_NONE; + instruction.NumDstRegs = 1; + instruction.NumSrcRegs = 1; + instruction.Padding = 0; + instruction.Extended = 0; + + return instruction; +} + +struct tgsi_instruction +tgsi_build_instruction( + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header ) +{ + struct tgsi_instruction instruction; + + assert (opcode <= TGSI_OPCODE_LAST); + assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); + assert (num_dst_regs <= 3); + assert (num_src_regs <= 15); + + instruction = tgsi_default_instruction(); + instruction.Opcode = opcode; + instruction.Saturate = saturate; + instruction.NumDstRegs = num_dst_regs; + instruction.NumSrcRegs = num_src_regs; + + header_bodysize_grow( header ); + + return instruction; +} + +static void +instruction_grow( + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + assert (instruction->Size < 0xFF); + + instruction->Size++; + + header_bodysize_grow( header ); +} + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ) +{ + struct tgsi_full_instruction full_instruction; + unsigned i; + + full_instruction.Instruction = tgsi_default_instruction(); + full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); + full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); + full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); + for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { + full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); + } + for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { + full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); + } + + return full_instruction; +} + +unsigned +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ) +{ + unsigned size = 0; + unsigned i; + struct tgsi_instruction *instruction; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + instruction = (struct tgsi_instruction *) &tokens[size]; + size++; + + *instruction = tgsi_build_instruction( + full_inst->Instruction.Opcode, + full_inst->Instruction.Saturate, + full_inst->Instruction.NumDstRegs, + full_inst->Instruction.NumSrcRegs, + header ); + prev_token = (struct tgsi_token *) instruction; + + if( tgsi_compare_instruction_ext_nv( + full_inst->InstructionExtNv, + tgsi_default_instruction_ext_nv() ) ) { + struct tgsi_instruction_ext_nv *instruction_ext_nv; + + if( maxsize <= size ) + return 0; + instruction_ext_nv = + (struct tgsi_instruction_ext_nv *) &tokens[size]; + size++; + + *instruction_ext_nv = tgsi_build_instruction_ext_nv( + full_inst->InstructionExtNv.Precision, + full_inst->InstructionExtNv.CondDstIndex, + full_inst->InstructionExtNv.CondFlowIndex, + full_inst->InstructionExtNv.CondMask, + full_inst->InstructionExtNv.CondSwizzleX, + full_inst->InstructionExtNv.CondSwizzleY, + full_inst->InstructionExtNv.CondSwizzleZ, + full_inst->InstructionExtNv.CondSwizzleW, + full_inst->InstructionExtNv.CondDstUpdate, + full_inst->InstructionExtNv.CondFlowEnable, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_nv; + } + + if( tgsi_compare_instruction_ext_label( + full_inst->InstructionExtLabel, + tgsi_default_instruction_ext_label() ) ) { + struct tgsi_instruction_ext_label *instruction_ext_label; + + if( maxsize <= size ) + return 0; + instruction_ext_label = + (struct tgsi_instruction_ext_label *) &tokens[size]; + size++; + + *instruction_ext_label = tgsi_build_instruction_ext_label( + full_inst->InstructionExtLabel.Label, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_label; + } + + if( tgsi_compare_instruction_ext_texture( + full_inst->InstructionExtTexture, + tgsi_default_instruction_ext_texture() ) ) { + struct tgsi_instruction_ext_texture *instruction_ext_texture; + + if( maxsize <= size ) + return 0; + instruction_ext_texture = + (struct tgsi_instruction_ext_texture *) &tokens[size]; + size++; + + *instruction_ext_texture = tgsi_build_instruction_ext_texture( + full_inst->InstructionExtTexture.Texture, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) instruction_ext_texture; + } + + for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { + const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; + struct tgsi_dst_register *dst_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + dst_register = (struct tgsi_dst_register *) &tokens[size]; + size++; + + *dst_register = tgsi_build_dst_register( + reg->DstRegister.File, + reg->DstRegister.WriteMask, + reg->DstRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register; + + if( tgsi_compare_dst_register_ext_concode( + reg->DstRegisterExtConcode, + tgsi_default_dst_register_ext_concode() ) ) { + struct tgsi_dst_register_ext_concode *dst_register_ext_concode; + + if( maxsize <= size ) + return 0; + dst_register_ext_concode = + (struct tgsi_dst_register_ext_concode *) &tokens[size]; + size++; + + *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( + reg->DstRegisterExtConcode.CondMask, + reg->DstRegisterExtConcode.CondSwizzleX, + reg->DstRegisterExtConcode.CondSwizzleY, + reg->DstRegisterExtConcode.CondSwizzleZ, + reg->DstRegisterExtConcode.CondSwizzleW, + reg->DstRegisterExtConcode.CondSrcIndex, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_concode; + } + + if( tgsi_compare_dst_register_ext_modulate( + reg->DstRegisterExtModulate, + tgsi_default_dst_register_ext_modulate() ) ) { + struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; + + if( maxsize <= size ) + return 0; + dst_register_ext_modulate = + (struct tgsi_dst_register_ext_modulate *) &tokens[size]; + size++; + + *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( + reg->DstRegisterExtModulate.Modulate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) dst_register_ext_modulate; + } + } + + for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { + const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; + struct tgsi_src_register *src_register; + struct tgsi_token *prev_token; + + if( maxsize <= size ) + return 0; + src_register = (struct tgsi_src_register *) &tokens[size]; + size++; + + *src_register = tgsi_build_src_register( + reg->SrcRegister.File, + reg->SrcRegister.SwizzleX, + reg->SrcRegister.SwizzleY, + reg->SrcRegister.SwizzleZ, + reg->SrcRegister.SwizzleW, + reg->SrcRegister.Negate, + reg->SrcRegister.Indirect, + reg->SrcRegister.Dimension, + reg->SrcRegister.Index, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register; + + if( tgsi_compare_src_register_ext_swz( + reg->SrcRegisterExtSwz, + tgsi_default_src_register_ext_swz() ) ) { + struct tgsi_src_register_ext_swz *src_register_ext_swz; + + /* Use of the extended swizzle requires the simple swizzle to be identity. + */ + assert( reg->SrcRegister.SwizzleX == TGSI_SWIZZLE_X ); + assert( reg->SrcRegister.SwizzleY == TGSI_SWIZZLE_Y ); + assert( reg->SrcRegister.SwizzleZ == TGSI_SWIZZLE_Z ); + assert( reg->SrcRegister.SwizzleW == TGSI_SWIZZLE_W ); + assert( reg->SrcRegister.Negate == FALSE ); + + if( maxsize <= size ) + return 0; + src_register_ext_swz = + (struct tgsi_src_register_ext_swz *) &tokens[size]; + size++; + + *src_register_ext_swz = tgsi_build_src_register_ext_swz( + reg->SrcRegisterExtSwz.ExtSwizzleX, + reg->SrcRegisterExtSwz.ExtSwizzleY, + reg->SrcRegisterExtSwz.ExtSwizzleZ, + reg->SrcRegisterExtSwz.ExtSwizzleW, + reg->SrcRegisterExtSwz.NegateX, + reg->SrcRegisterExtSwz.NegateY, + reg->SrcRegisterExtSwz.NegateZ, + reg->SrcRegisterExtSwz.NegateW, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_swz; + } + + if( tgsi_compare_src_register_ext_mod( + reg->SrcRegisterExtMod, + tgsi_default_src_register_ext_mod() ) ) { + struct tgsi_src_register_ext_mod *src_register_ext_mod; + + if( maxsize <= size ) + return 0; + src_register_ext_mod = + (struct tgsi_src_register_ext_mod *) &tokens[size]; + size++; + + *src_register_ext_mod = tgsi_build_src_register_ext_mod( + reg->SrcRegisterExtMod.Complement, + reg->SrcRegisterExtMod.Bias, + reg->SrcRegisterExtMod.Scale2X, + reg->SrcRegisterExtMod.Absolute, + reg->SrcRegisterExtMod.Negate, + prev_token, + instruction, + header ); + prev_token = (struct tgsi_token *) src_register_ext_mod; + } + + if( reg->SrcRegister.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterInd.File, + reg->SrcRegisterInd.SwizzleX, + reg->SrcRegisterInd.SwizzleY, + reg->SrcRegisterInd.SwizzleZ, + reg->SrcRegisterInd.SwizzleW, + reg->SrcRegisterInd.Negate, + reg->SrcRegisterInd.Indirect, + reg->SrcRegisterInd.Dimension, + reg->SrcRegisterInd.Index, + instruction, + header ); + } + + if( reg->SrcRegister.Dimension ) { + struct tgsi_dimension *dim; + + assert( !reg->SrcRegisterDim.Dimension ); + + if( maxsize <= size ) + return 0; + dim = (struct tgsi_dimension *) &tokens[size]; + size++; + + *dim = tgsi_build_dimension( + reg->SrcRegisterDim.Indirect, + reg->SrcRegisterDim.Index, + instruction, + header ); + + if( reg->SrcRegisterDim.Indirect ) { + struct tgsi_src_register *ind; + + if( maxsize <= size ) + return 0; + ind = (struct tgsi_src_register *) &tokens[size]; + size++; + + *ind = tgsi_build_src_register( + reg->SrcRegisterDimInd.File, + reg->SrcRegisterDimInd.SwizzleX, + reg->SrcRegisterDimInd.SwizzleY, + reg->SrcRegisterDimInd.SwizzleZ, + reg->SrcRegisterDimInd.SwizzleW, + reg->SrcRegisterDimInd.Negate, + reg->SrcRegisterDimInd.Indirect, + reg->SrcRegisterDimInd.Dimension, + reg->SrcRegisterDimInd.Index, + instruction, + header ); + } + } + } + + return size; +} + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; + instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; + instruction_ext_nv.CondDstIndex = 0; + instruction_ext_nv.CondFlowIndex = 0; + instruction_ext_nv.CondMask = TGSI_CC_TR; + instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; + instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; + instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; + instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; + instruction_ext_nv.CondDstUpdate = 0; + instruction_ext_nv.CondFlowEnable = 0; + instruction_ext_nv.Padding = 0; + instruction_ext_nv.Extended = 0; + + return instruction_ext_nv; +} + +union token_u32 +{ + unsigned u32; +}; + +unsigned +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_nv instruction_ext_nv; + + instruction_ext_nv = tgsi_default_instruction_ext_nv(); + instruction_ext_nv.Precision = precision; + instruction_ext_nv.CondDstIndex = cond_dst_index; + instruction_ext_nv.CondFlowIndex = cond_flow_index; + instruction_ext_nv.CondMask = cond_mask; + instruction_ext_nv.CondSwizzleX = cond_swizzle_x; + instruction_ext_nv.CondSwizzleY = cond_swizzle_y; + instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; + instruction_ext_nv.CondSwizzleW = cond_swizzle_w; + instruction_ext_nv.CondDstUpdate = cond_dst_update; + instruction_ext_nv.CondFlowEnable = cond_flow_update; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_nv; +} + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; + instruction_ext_label.Label = 0; + instruction_ext_label.Padding = 0; + instruction_ext_label.Extended = 0; + + return instruction_ext_label; +} + +unsigned +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + unsigned label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_label instruction_ext_label; + + instruction_ext_label = tgsi_default_instruction_ext_label(); + instruction_ext_label.Label = label; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_label; +} + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; + instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; + instruction_ext_texture.Padding = 0; + instruction_ext_texture.Extended = 0; + + return instruction_ext_texture; +} + +unsigned +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + unsigned texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_instruction_ext_texture instruction_ext_texture; + + instruction_ext_texture = tgsi_default_instruction_ext_texture(); + instruction_ext_texture.Texture = texture; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return instruction_ext_texture; +} + +struct tgsi_src_register +tgsi_default_src_register( void ) +{ + struct tgsi_src_register src_register; + + src_register.File = TGSI_FILE_NULL; + src_register.SwizzleX = TGSI_SWIZZLE_X; + src_register.SwizzleY = TGSI_SWIZZLE_Y; + src_register.SwizzleZ = TGSI_SWIZZLE_Z; + src_register.SwizzleW = TGSI_SWIZZLE_W; + src_register.Negate = 0; + src_register.Indirect = 0; + src_register.Dimension = 0; + src_register.Index = 0; + src_register.Extended = 0; + + return src_register; +} + +struct tgsi_src_register +tgsi_build_src_register( + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register src_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( negate <= 1 ); + assert( index >= -0x8000 && index <= 0x7FFF ); + + src_register = tgsi_default_src_register(); + src_register.File = file; + src_register.SwizzleX = swizzle_x; + src_register.SwizzleY = swizzle_y; + src_register.SwizzleZ = swizzle_z; + src_register.SwizzleW = swizzle_w; + src_register.Negate = negate; + src_register.Indirect = indirect; + src_register.Dimension = dimension; + src_register.Index = index; + + instruction_grow( instruction, header ); + + return src_register; +} + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ) +{ + struct tgsi_full_src_register full_src_register; + + full_src_register.SrcRegister = tgsi_default_src_register(); + full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); + full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); + full_src_register.SrcRegisterInd = tgsi_default_src_register(); + full_src_register.SrcRegisterDim = tgsi_default_dimension(); + full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); + + return full_src_register; +} + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; + src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; + src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; + src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; + src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; + src_register_ext_swz.NegateX = 0; + src_register_ext_swz.NegateY = 0; + src_register_ext_swz.NegateZ = 0; + src_register_ext_swz.NegateW = 0; + src_register_ext_swz.Padding = 0; + src_register_ext_swz.Extended = 0; + + return src_register_ext_swz; +} + +unsigned +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_swz src_register_ext_swz; + + assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE ); + assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE ); + assert( negate_x <= 1 ); + assert( negate_y <= 1 ); + assert( negate_z <= 1 ); + assert( negate_w <= 1 ); + + src_register_ext_swz = tgsi_default_src_register_ext_swz(); + src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; + src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; + src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; + src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; + src_register_ext_swz.NegateX = negate_x; + src_register_ext_swz.NegateY = negate_y; + src_register_ext_swz.NegateZ = negate_z; + src_register_ext_swz.NegateW = negate_w; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_swz; +} + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; + src_register_ext_mod.Complement = 0; + src_register_ext_mod.Bias = 0; + src_register_ext_mod.Scale2X = 0; + src_register_ext_mod.Absolute = 0; + src_register_ext_mod.Negate = 0; + src_register_ext_mod.Padding = 0; + src_register_ext_mod.Extended = 0; + + return src_register_ext_mod; +} + +unsigned +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_src_register_ext_mod src_register_ext_mod; + + assert( complement <= 1 ); + assert( bias <= 1 ); + assert( scale_2x <= 1 ); + assert( absolute <= 1 ); + assert( negate <= 1 ); + + src_register_ext_mod = tgsi_default_src_register_ext_mod(); + src_register_ext_mod.Complement = complement; + src_register_ext_mod.Bias = bias; + src_register_ext_mod.Scale2X = scale_2x; + src_register_ext_mod.Absolute = absolute; + src_register_ext_mod.Negate = negate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return src_register_ext_mod; +} + +struct tgsi_dimension +tgsi_default_dimension( void ) +{ + struct tgsi_dimension dimension; + + dimension.Indirect = 0; + dimension.Dimension = 0; + dimension.Padding = 0; + dimension.Index = 0; + dimension.Extended = 0; + + return dimension; +} + +struct tgsi_dimension +tgsi_build_dimension( + unsigned indirect, + unsigned index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dimension dimension; + + dimension = tgsi_default_dimension(); + dimension.Indirect = indirect; + dimension.Index = index; + + instruction_grow( instruction, header ); + + return dimension; +} + +struct tgsi_dst_register +tgsi_default_dst_register( void ) +{ + struct tgsi_dst_register dst_register; + + dst_register.File = TGSI_FILE_NULL; + dst_register.WriteMask = TGSI_WRITEMASK_XYZW; + dst_register.Indirect = 0; + dst_register.Dimension = 0; + dst_register.Index = 0; + dst_register.Padding = 0; + dst_register.Extended = 0; + + return dst_register; +} + +struct tgsi_dst_register +tgsi_build_dst_register( + unsigned file, + unsigned mask, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register dst_register; + + assert( file <= TGSI_FILE_IMMEDIATE ); + assert( mask <= TGSI_WRITEMASK_XYZW ); + assert( index >= -32768 && index <= 32767 ); + + dst_register = tgsi_default_dst_register(); + dst_register.File = file; + dst_register.WriteMask = mask; + dst_register.Index = index; + + instruction_grow( instruction, header ); + + return dst_register; +} + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ) +{ + struct tgsi_full_dst_register full_dst_register; + + full_dst_register.DstRegister = tgsi_default_dst_register(); + full_dst_register.DstRegisterExtConcode = + tgsi_default_dst_register_ext_concode(); + full_dst_register.DstRegisterExtModulate = + tgsi_default_dst_register_ext_modulate(); + + return full_dst_register; +} + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; + dst_register_ext_concode.CondMask = TGSI_CC_TR; + dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; + dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; + dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; + dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; + dst_register_ext_concode.CondSrcIndex = 0; + dst_register_ext_concode.Padding = 0; + dst_register_ext_concode.Extended = 0; + + return dst_register_ext_concode; +} + +unsigned +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_concode dst_register_ext_concode; + + assert( cc <= TGSI_CC_FL ); + assert( swizzle_x <= TGSI_SWIZZLE_W ); + assert( swizzle_y <= TGSI_SWIZZLE_W ); + assert( swizzle_z <= TGSI_SWIZZLE_W ); + assert( swizzle_w <= TGSI_SWIZZLE_W ); + assert( index >= -32768 && index <= 32767 ); + + dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); + dst_register_ext_concode.CondMask = cc; + dst_register_ext_concode.CondSwizzleX = swizzle_x; + dst_register_ext_concode.CondSwizzleY = swizzle_y; + dst_register_ext_concode.CondSwizzleZ = swizzle_z; + dst_register_ext_concode.CondSwizzleW = swizzle_w; + dst_register_ext_concode.CondSrcIndex = index; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_concode; +} + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; + dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; + dst_register_ext_modulate.Padding = 0; + dst_register_ext_modulate.Extended = 0; + + return dst_register_ext_modulate; +} + +unsigned +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ) +{ + a.Padding = b.Padding = 0; + a.Extended = b.Extended = 0; + return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; +} + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + unsigned modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ) +{ + struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; + + assert( modulate <= TGSI_MODULATE_EIGHTH ); + + dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); + dst_register_ext_modulate.Modulate = modulate; + + prev_token->Extended = 1; + instruction_grow( instruction, header ); + + return dst_register_ext_modulate; +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h new file mode 100644 index 0000000000..ed25830248 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_build.h @@ -0,0 +1,332 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_BUILD_H +#define TGSI_BUILD_H + +#if defined __cplusplus +extern "C" { +#endif + +/* + * version + */ + +struct tgsi_version +tgsi_build_version( void ); + +/* + * header + */ + +struct tgsi_header +tgsi_build_header( void ); + +struct tgsi_processor +tgsi_default_processor( void ); + +struct tgsi_processor +tgsi_build_processor( + unsigned processor, + struct tgsi_header *header ); + +/* + * declaration + */ + +struct tgsi_declaration +tgsi_default_declaration( void ); + +struct tgsi_declaration +tgsi_build_declaration( + unsigned file, + unsigned usage_mask, + unsigned interpolate, + unsigned semantic, + struct tgsi_header *header ); + +struct tgsi_full_declaration +tgsi_default_full_declaration( void ); + +unsigned +tgsi_build_full_declaration( + const struct tgsi_full_declaration *full_decl, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_declaration_range +tgsi_default_declaration_range( void ); + +struct tgsi_declaration_range +tgsi_build_declaration_range( + unsigned first, + unsigned last, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +struct tgsi_declaration_semantic +tgsi_default_declaration_semantic( void ); + +struct tgsi_declaration_semantic +tgsi_build_declaration_semantic( + unsigned semantic_name, + unsigned semantic_index, + struct tgsi_declaration *declaration, + struct tgsi_header *header ); + +/* + * immediate + */ + +struct tgsi_immediate +tgsi_default_immediate( void ); + +struct tgsi_immediate +tgsi_build_immediate( + struct tgsi_header *header ); + +struct tgsi_full_immediate +tgsi_default_full_immediate( void ); + +struct tgsi_immediate_float32 +tgsi_build_immediate_float32( + float value, + struct tgsi_immediate *immediate, + struct tgsi_header *header ); + +unsigned +tgsi_build_full_immediate( + const struct tgsi_full_immediate *full_imm, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +/* + * instruction + */ + +struct tgsi_instruction +tgsi_default_instruction( void ); + +struct tgsi_instruction +tgsi_build_instruction( + unsigned opcode, + unsigned saturate, + unsigned num_dst_regs, + unsigned num_src_regs, + struct tgsi_header *header ); + +struct tgsi_full_instruction +tgsi_default_full_instruction( void ); + +unsigned +tgsi_build_full_instruction( + const struct tgsi_full_instruction *full_inst, + struct tgsi_token *tokens, + struct tgsi_header *header, + unsigned maxsize ); + +struct tgsi_instruction_ext_nv +tgsi_default_instruction_ext_nv( void ); + +unsigned +tgsi_compare_instruction_ext_nv( + struct tgsi_instruction_ext_nv a, + struct tgsi_instruction_ext_nv b ); + +struct tgsi_instruction_ext_nv +tgsi_build_instruction_ext_nv( + unsigned precision, + unsigned cond_dst_index, + unsigned cond_flow_index, + unsigned cond_mask, + unsigned cond_swizzle_x, + unsigned cond_swizzle_y, + unsigned cond_swizzle_z, + unsigned cond_swizzle_w, + unsigned cond_dst_update, + unsigned cond_flow_update, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_label +tgsi_default_instruction_ext_label( void ); + +unsigned +tgsi_compare_instruction_ext_label( + struct tgsi_instruction_ext_label a, + struct tgsi_instruction_ext_label b ); + +struct tgsi_instruction_ext_label +tgsi_build_instruction_ext_label( + unsigned label, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_instruction_ext_texture +tgsi_default_instruction_ext_texture( void ); + +unsigned +tgsi_compare_instruction_ext_texture( + struct tgsi_instruction_ext_texture a, + struct tgsi_instruction_ext_texture b ); + +struct tgsi_instruction_ext_texture +tgsi_build_instruction_ext_texture( + unsigned texture, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register +tgsi_default_src_register( void ); + +struct tgsi_src_register +tgsi_build_src_register( + unsigned file, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + unsigned negate, + unsigned indirect, + unsigned dimension, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_src_register +tgsi_default_full_src_register( void ); + +struct tgsi_src_register_ext_swz +tgsi_default_src_register_ext_swz( void ); + +unsigned +tgsi_compare_src_register_ext_swz( + struct tgsi_src_register_ext_swz a, + struct tgsi_src_register_ext_swz b ); + +struct tgsi_src_register_ext_swz +tgsi_build_src_register_ext_swz( + unsigned ext_swizzle_x, + unsigned ext_swizzle_y, + unsigned ext_swizzle_z, + unsigned ext_swizzle_w, + unsigned negate_x, + unsigned negate_y, + unsigned negate_z, + unsigned negate_w, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_src_register_ext_mod +tgsi_default_src_register_ext_mod( void ); + +unsigned +tgsi_compare_src_register_ext_mod( + struct tgsi_src_register_ext_mod a, + struct tgsi_src_register_ext_mod b ); + +struct tgsi_src_register_ext_mod +tgsi_build_src_register_ext_mod( + unsigned complement, + unsigned bias, + unsigned scale_2x, + unsigned absolute, + unsigned negate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dimension +tgsi_default_dimension( void ); + +struct tgsi_dimension +tgsi_build_dimension( + unsigned indirect, + unsigned index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register +tgsi_default_dst_register( void ); + +struct tgsi_dst_register +tgsi_build_dst_register( + unsigned file, + unsigned mask, + int index, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_full_dst_register +tgsi_default_full_dst_register( void ); + +struct tgsi_dst_register_ext_concode +tgsi_default_dst_register_ext_concode( void ); + +unsigned +tgsi_compare_dst_register_ext_concode( + struct tgsi_dst_register_ext_concode a, + struct tgsi_dst_register_ext_concode b ); + +struct tgsi_dst_register_ext_concode +tgsi_build_dst_register_ext_concode( + unsigned cc, + unsigned swizzle_x, + unsigned swizzle_y, + unsigned swizzle_z, + unsigned swizzle_w, + int index, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +struct tgsi_dst_register_ext_modulate +tgsi_default_dst_register_ext_modulate( void ); + +unsigned +tgsi_compare_dst_register_ext_modulate( + struct tgsi_dst_register_ext_modulate a, + struct tgsi_dst_register_ext_modulate b ); + +struct tgsi_dst_register_ext_modulate +tgsi_build_dst_register_ext_modulate( + unsigned modulate, + struct tgsi_token *prev_token, + struct tgsi_instruction *instruction, + struct tgsi_header *header ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_BUILD_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c new file mode 100644 index 0000000000..d2e6375212 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -0,0 +1,582 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "tgsi_dump.h" +#include "tgsi_iterate.h" + +struct dump_ctx +{ + struct tgsi_iterate_context iter; + + uint instno; +}; + +static void +dump_enum( + uint e, + const char **enums, + uint enum_count ) +{ + if (e >= enum_count) + debug_printf( "%u", e ); + else + debug_printf( "%s", enums[e] ); +} + +#define EOL() debug_printf( "\n" ) +#define TXT(S) debug_printf( "%s", S ) +#define CHR(C) debug_printf( "%c", C ) +#define UIX(I) debug_printf( "0x%x", I ) +#define UID(I) debug_printf( "%u", I ) +#define SID(I) debug_printf( "%d", I ) +#define FLT(F) debug_printf( "%10.4f", F ) +#define ENM(E,ENUMS) dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) + +static const char *processor_type_names[] = +{ + "FRAG", + "VERT", + "GEOM" +}; + +static const char *file_names[] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + +static const char *interpolate_names[] = +{ + "CONSTANT", + "LINEAR", + "PERSPECTIVE" +}; + +static const char *semantic_names[] = +{ + "POSITION", + "COLOR", + "BCOLOR", + "FOG", + "PSIZE", + "GENERIC", + "NORMAL" +}; + +static const char *immediate_type_names[] = +{ + "FLT32" +}; + +static const char *opcode_names[TGSI_OPCODE_LAST] = +{ + "ARL", + "MOV", + "LIT", + "RCP", + "RSQ", + "EXP", + "LOG", + "MUL", + "ADD", + "DP3", + "DP4", + "DST", + "MIN", + "MAX", + "SLT", + "SGE", + "MAD", + "SUB", + "LERP", + "CND", + "CND0", + "DOT2ADD", + "INDEX", + "NEGATE", + "FRAC", + "CLAMP", + "FLOOR", + "ROUND", + "EXPBASE2", + "LOGBASE2", + "POWER", + "CROSSPRODUCT", + "MULTIPLYMATRIX", + "ABS", + "RCC", + "DPH", + "COS", + "DDX", + "DDY", + "KILP", + "PK2H", + "PK2US", + "PK4B", + "PK4UB", + "RFL", + "SEQ", + "SFL", + "SGT", + "SIN", + "SLE", + "SNE", + "STR", + "TEX", + "TXD", + "TXP", + "UP2H", + "UP2US", + "UP4B", + "UP4UB", + "X2D", + "ARA", + "ARR", + "BRA", + "CAL", + "RET", + "SSG", + "CMP", + "SCS", + "TXB", + "NRM", + "DIV", + "DP2", + "TXL", + "BRK", + "IF", + "LOOP", + "REP", + "ELSE", + "ENDIF", + "ENDLOOP", + "ENDREP", + "PUSHA", + "POPA", + "CEIL", + "I2F", + "NOT", + "TRUNC", + "SHL", + "SHR", + "AND", + "OR", + "MOD", + "XOR", + "SAD", + "TXF", + "TXQ", + "CONT", + "EMIT", + "ENDPRIM", + "BGNLOOP2", + "BGNSUB", + "ENDLOOP2", + "ENDSUB", + "NOISE1", + "NOISE2", + "NOISE3", + "NOISE4", + "NOP", + "M4X3", + "M3X4", + "M3X3", + "M3X2", + "NRM4", + "CALLNZ", + "IFC", + "BREAKC", + "KIL", + "END", + "SWZ" +}; + +static const char *swizzle_names[] = +{ + "x", + "y", + "z", + "w" +}; + +static const char *texture_names[] = +{ + "UNKNOWN", + "1D", + "2D", + "3D", + "CUBE", + "RECT", + "SHADOW1D", + "SHADOW2D", + "SHADOWRECT" +}; + +static const char *extswizzle_names[] = +{ + "x", + "y", + "z", + "w", + "0", + "1" +}; + +static const char *modulate_names[TGSI_MODULATE_COUNT] = +{ + "", + "_2X", + "_4X", + "_8X", + "_D2", + "_D4", + "_D8" +}; + +static void +_dump_register_prefix( + uint file, + uint first, + uint last ) +{ + + +} + +static void +_dump_register( + uint file, + int first, + int last ) +{ + ENM( file, file_names ); + CHR( '[' ); + SID( first ); + if (first != last) { + TXT( ".." ); + SID( last ); + } + CHR( ']' ); +} + +static void +_dump_register_ind( + uint file, + int index, + uint ind_file, + int ind_index ) +{ + ENM( file, file_names ); + CHR( '[' ); + ENM( ind_file, file_names ); + CHR( '[' ); + SID( ind_index ); + CHR( ']' ); + if (index != 0) { + if (index > 0) + CHR( '+' ); + SID( index ); + } + CHR( ']' ); +} + +static void +_dump_writemask( + uint writemask ) +{ + if (writemask != TGSI_WRITEMASK_XYZW) { + CHR( '.' ); + if (writemask & TGSI_WRITEMASK_X) + CHR( 'x' ); + if (writemask & TGSI_WRITEMASK_Y) + CHR( 'y' ); + if (writemask & TGSI_WRITEMASK_Z) + CHR( 'z' ); + if (writemask & TGSI_WRITEMASK_W) + CHR( 'w' ); + } +} + +void +tgsi_dump_declaration( + const struct tgsi_full_declaration *decl ) +{ + TXT( "\nDCL " ); + + _dump_register( + decl->Declaration.File, + decl->DeclarationRange.First, + decl->DeclarationRange.Last ); + _dump_writemask( + decl->Declaration.UsageMask ); + + if (decl->Declaration.Semantic) { + TXT( ", " ); + ENM( decl->Semantic.SemanticName, semantic_names ); + if (decl->Semantic.SemanticIndex != 0 || + decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC) { + CHR( '[' ); + UID( decl->Semantic.SemanticIndex ); + CHR( ']' ); + } + } + + TXT( ", " ); + ENM( decl->Declaration.Interpolate, interpolate_names ); +} + +static boolean +iter_declaration( + struct tgsi_iterate_context *iter, + struct tgsi_full_declaration *decl ) +{ + tgsi_dump_declaration( decl ); + return TRUE; +} + +void +tgsi_dump_immediate( + const struct tgsi_full_immediate *imm ) +{ + uint i; + + TXT( "\nIMM " ); + ENM( imm->Immediate.DataType, immediate_type_names ); + + TXT( " { " ); + for (i = 0; i < imm->Immediate.Size - 1; i++) { + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + default: + assert( 0 ); + } + + if (i < imm->Immediate.Size - 2) + TXT( ", " ); + } + TXT( " }" ); +} + +static boolean +iter_immediate( + struct tgsi_iterate_context *iter, + struct tgsi_full_immediate *imm ) +{ + tgsi_dump_immediate( imm ); + return TRUE; +} + +void +tgsi_dump_instruction( + const struct tgsi_full_instruction *inst, + uint instno ) +{ + uint i; + boolean first_reg = TRUE; + + EOL(); + UID( instno ); + CHR( ':' ); + ENM( inst->Instruction.Opcode, opcode_names ); + + switch (inst->Instruction.Saturate) { + case TGSI_SAT_NONE: + break; + case TGSI_SAT_ZERO_ONE: + TXT( "_SAT" ); + break; + case TGSI_SAT_MINUS_PLUS_ONE: + TXT( "_SATNV" ); + break; + default: + assert( 0 ); + } + + for (i = 0; i < inst->Instruction.NumDstRegs; i++) { + const struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + + if (!first_reg) + CHR( ',' ); + CHR( ' ' ); + + _dump_register( + dst->DstRegister.File, + dst->DstRegister.Index, + dst->DstRegister.Index ); + ENM( dst->DstRegisterExtModulate.Modulate, modulate_names ); + _dump_writemask( dst->DstRegister.WriteMask ); + + first_reg = FALSE; + } + + for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { + const struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + + if (!first_reg) + CHR( ',' ); + CHR( ' ' ); + + if (src->SrcRegisterExtMod.Negate) + TXT( "-(" ); + if (src->SrcRegisterExtMod.Absolute) + CHR( '|' ); + if (src->SrcRegisterExtMod.Scale2X) + TXT( "2*(" ); + if (src->SrcRegisterExtMod.Bias) + CHR( '(' ); + if (src->SrcRegisterExtMod.Complement) + TXT( "1-(" ); + if (src->SrcRegister.Negate) + CHR( '-' ); + + if (src->SrcRegister.Indirect) { + _dump_register_ind( + src->SrcRegister.File, + src->SrcRegister.Index, + src->SrcRegisterInd.File, + src->SrcRegisterInd.Index ); + } + else { + _dump_register( + src->SrcRegister.File, + src->SrcRegister.Index, + src->SrcRegister.Index ); + } + + if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) { + CHR( '.' ); + ENM( src->SrcRegister.SwizzleX, swizzle_names ); + ENM( src->SrcRegister.SwizzleY, swizzle_names ); + ENM( src->SrcRegister.SwizzleZ, swizzle_names ); + ENM( src->SrcRegister.SwizzleW, swizzle_names ); + } + if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { + CHR( '.' ); + if (src->SrcRegisterExtSwz.NegateX) + TXT("-"); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, extswizzle_names ); + if (src->SrcRegisterExtSwz.NegateY) + TXT("-"); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, extswizzle_names ); + if (src->SrcRegisterExtSwz.NegateZ) + TXT("-"); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, extswizzle_names ); + if (src->SrcRegisterExtSwz.NegateW) + TXT("-"); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, extswizzle_names ); + } + + if (src->SrcRegisterExtMod.Complement) + CHR( ')' ); + if (src->SrcRegisterExtMod.Bias) + TXT( ")-.5" ); + if (src->SrcRegisterExtMod.Scale2X) + CHR( ')' ); + if (src->SrcRegisterExtMod.Absolute) + CHR( '|' ); + if (src->SrcRegisterExtMod.Negate) + CHR( ')' ); + + first_reg = FALSE; + } + + if (inst->InstructionExtTexture.Texture != TGSI_TEXTURE_UNKNOWN) { + TXT( ", " ); + ENM( inst->InstructionExtTexture.Texture, texture_names ); + } + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_IF: + case TGSI_OPCODE_ELSE: + case TGSI_OPCODE_BGNLOOP2: + case TGSI_OPCODE_ENDLOOP2: + case TGSI_OPCODE_CAL: + TXT( " :" ); + UID( inst->InstructionExtLabel.Label ); + break; + } +} + +static boolean +iter_instruction( + struct tgsi_iterate_context *iter, + struct tgsi_full_instruction *inst ) +{ + struct dump_ctx *ctx = (struct dump_ctx *) iter; + + tgsi_dump_instruction( inst, ctx->instno++ ); + return TRUE; +} + +static boolean +prolog( + struct tgsi_iterate_context *ctx ) +{ + EOL(); + ENM( ctx->processor.Processor, processor_type_names ); + UID( ctx->version.MajorVersion ); + CHR( '.' ); + UID( ctx->version.MinorVersion ); + return TRUE; +} + +void +tgsi_dump( + const struct tgsi_token *tokens, + uint flags ) +{ + struct dump_ctx ctx; + + /* sanity checks */ + assert( strcmp( opcode_names[TGSI_OPCODE_CONT], "CONT" ) == 0 ); + assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); + + ctx.iter.prolog = prolog; + ctx.iter.iterate_instruction = iter_instruction; + ctx.iter.iterate_declaration = iter_declaration; + ctx.iter.iterate_immediate = iter_immediate; + ctx.iter.epilog = NULL; + + ctx.instno = 0; + + tgsi_iterate_shader( tokens, &ctx.iter ); +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.h b/src/gallium/auxiliary/tgsi/tgsi_dump.h new file mode 100644 index 0000000000..51c230b5db --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.h @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_DUMP_H +#define TGSI_DUMP_H + +#include "pipe/p_shader_tokens.h" + +#if defined __cplusplus +extern "C" { +#endif + +void +tgsi_dump( + const struct tgsi_token *tokens, + uint flags ); + +struct tgsi_full_immediate; +struct tgsi_full_instruction; +struct tgsi_full_declaration; + +void +tgsi_dump_immediate( + const struct tgsi_full_immediate *imm ); + +void +tgsi_dump_instruction( + const struct tgsi_full_instruction *inst, + uint instno ); + +void +tgsi_dump_declaration( + const struct tgsi_full_declaration *decl ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_DUMP_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c new file mode 100644 index 0000000000..eabd74bd6d --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c @@ -0,0 +1,845 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "util/u_string.h" +#include "tgsi_dump_c.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" + +static void +dump_enum( + const unsigned e, + const char **enums, + const unsigned enums_count ) +{ + if (e >= enums_count) { + debug_printf( "%u", e ); + } + else { + debug_printf( "%s", enums[e] ); + } +} + +#define EOL() debug_printf( "\n" ) +#define TXT(S) debug_printf( "%s", S ) +#define CHR(C) debug_printf( "%c", C ) +#define UIX(I) debug_printf( "0x%x", I ) +#define UID(I) debug_printf( "%u", I ) +#define SID(I) debug_printf( "%d", I ) +#define FLT(F) debug_printf( "%10.4f", F ) +#define ENM(E,ENUMS) dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) + +static const char *TGSI_PROCESSOR_TYPES[] = +{ + "PROCESSOR_FRAGMENT", + "PROCESSOR_VERTEX", + "PROCESSOR_GEOMETRY" +}; + +static const char *TGSI_TOKEN_TYPES[] = +{ + "TOKEN_TYPE_DECLARATION", + "TOKEN_TYPE_IMMEDIATE", + "TOKEN_TYPE_INSTRUCTION" +}; + +static const char *TGSI_FILES[] = +{ + "FILE_NULL", + "FILE_CONSTANT", + "FILE_INPUT", + "FILE_OUTPUT", + "FILE_TEMPORARY", + "FILE_SAMPLER", + "FILE_ADDRESS", + "FILE_IMMEDIATE" +}; + +static const char *TGSI_INTERPOLATES[] = +{ + "INTERPOLATE_CONSTANT", + "INTERPOLATE_LINEAR", + "INTERPOLATE_PERSPECTIVE" +}; + +static const char *TGSI_SEMANTICS[] = +{ + "SEMANTIC_POSITION", + "SEMANTIC_COLOR", + "SEMANTIC_BCOLOR", + "SEMANTIC_FOG", + "SEMANTIC_PSIZE", + "SEMANTIC_GENERIC", + "SEMANTIC_NORMAL" +}; + +static const char *TGSI_IMMS[] = +{ + "IMM_FLOAT32" +}; + +static const char *TGSI_OPCODES[TGSI_OPCODE_LAST] = +{ + "OPCODE_ARL", + "OPCODE_MOV", + "OPCODE_LIT", + "OPCODE_RCP", + "OPCODE_RSQ", + "OPCODE_EXP", + "OPCODE_LOG", + "OPCODE_MUL", + "OPCODE_ADD", + "OPCODE_DP3", + "OPCODE_DP4", + "OPCODE_DST", + "OPCODE_MIN", + "OPCODE_MAX", + "OPCODE_SLT", + "OPCODE_SGE", + "OPCODE_MAD", + "OPCODE_SUB", + "OPCODE_LERP", + "OPCODE_CND", + "OPCODE_CND0", + "OPCODE_DOT2ADD", + "OPCODE_INDEX", + "OPCODE_NEGATE", + "OPCODE_FRAC", + "OPCODE_CLAMP", + "OPCODE_FLOOR", + "OPCODE_ROUND", + "OPCODE_EXPBASE2", + "OPCODE_LOGBASE2", + "OPCODE_POWER", + "OPCODE_CROSSPRODUCT", + "OPCODE_MULTIPLYMATRIX", + "OPCODE_ABS", + "OPCODE_RCC", + "OPCODE_DPH", + "OPCODE_COS", + "OPCODE_DDX", + "OPCODE_DDY", + "OPCODE_KILP", + "OPCODE_PK2H", + "OPCODE_PK2US", + "OPCODE_PK4B", + "OPCODE_PK4UB", + "OPCODE_RFL", + "OPCODE_SEQ", + "OPCODE_SFL", + "OPCODE_SGT", + "OPCODE_SIN", + "OPCODE_SLE", + "OPCODE_SNE", + "OPCODE_STR", + "OPCODE_TEX", + "OPCODE_TXD", + "OPCODE_TXP", + "OPCODE_UP2H", + "OPCODE_UP2US", + "OPCODE_UP4B", + "OPCODE_UP4UB", + "OPCODE_X2D", + "OPCODE_ARA", + "OPCODE_ARR", + "OPCODE_BRA", + "OPCODE_CAL", + "OPCODE_RET", + "OPCODE_SSG", + "OPCODE_CMP", + "OPCODE_SCS", + "OPCODE_TXB", + "OPCODE_NRM", + "OPCODE_DIV", + "OPCODE_DP2", + "OPCODE_TXL", + "OPCODE_BRK", + "OPCODE_IF", + "OPCODE_LOOP", + "OPCODE_REP", + "OPCODE_ELSE", + "OPCODE_ENDIF", + "OPCODE_ENDLOOP", + "OPCODE_ENDREP", + "OPCODE_PUSHA", + "OPCODE_POPA", + "OPCODE_CEIL", + "OPCODE_I2F", + "OPCODE_NOT", + "OPCODE_TRUNC", + "OPCODE_SHL", + "OPCODE_SHR", + "OPCODE_AND", + "OPCODE_OR", + "OPCODE_MOD", + "OPCODE_XOR", + "OPCODE_SAD", + "OPCODE_TXF", + "OPCODE_TXQ", + "OPCODE_CONT", + "OPCODE_EMIT", + "OPCODE_ENDPRIM", + "OPCODE_BGNLOOP2", + "OPCODE_BGNSUB", + "OPCODE_ENDLOOP2", + "OPCODE_ENDSUB", + "OPCODE_NOISE1", + "OPCODE_NOISE2", + "OPCODE_NOISE3", + "OPCODE_NOISE4", + "OPCODE_NOP", + "OPCODE_M4X3", + "OPCODE_M3X4", + "OPCODE_M3X3", + "OPCODE_M3X2", + "OPCODE_NRM4", + "OPCODE_CALLNZ", + "OPCODE_IFC", + "OPCODE_BREAKC", + "OPCODE_KIL", + "OPCODE_END" +}; + +static const char *TGSI_SATS[] = +{ + "SAT_NONE", + "SAT_ZERO_ONE", + "SAT_MINUS_PLUS_ONE" +}; + +static const char *TGSI_INSTRUCTION_EXTS[] = +{ + "INSTRUCTION_EXT_TYPE_NV", + "INSTRUCTION_EXT_TYPE_LABEL", + "INSTRUCTION_EXT_TYPE_TEXTURE" +}; + +static const char *TGSI_PRECISIONS[] = +{ + "PRECISION_DEFAULT", + "PRECISION_FLOAT32", + "PRECISION_FLOAT16", + "PRECISION_FIXED12" +}; + +static const char *TGSI_CCS[] = +{ + "CC_GT", + "CC_EQ", + "CC_LT", + "CC_UN", + "CC_GE", + "CC_LE", + "CC_NE", + "CC_TR", + "CC_FL" +}; + +static const char *TGSI_SWIZZLES[] = +{ + "SWIZZLE_X", + "SWIZZLE_Y", + "SWIZZLE_Z", + "SWIZZLE_W" +}; + +static const char *TGSI_TEXTURES[] = +{ + "TEXTURE_UNKNOWN", + "TEXTURE_1D", + "TEXTURE_2D", + "TEXTURE_3D", + "TEXTURE_CUBE", + "TEXTURE_RECT", + "TEXTURE_SHADOW1D", + "TEXTURE_SHADOW2D", + "TEXTURE_SHADOWRECT" +}; + +static const char *TGSI_SRC_REGISTER_EXTS[] = +{ + "SRC_REGISTER_EXT_TYPE_SWZ", + "SRC_REGISTER_EXT_TYPE_MOD" +}; + +static const char *TGSI_EXTSWIZZLES[] = +{ + "EXTSWIZZLE_X", + "EXTSWIZZLE_Y", + "EXTSWIZZLE_Z", + "EXTSWIZZLE_W", + "EXTSWIZZLE_ZERO", + "EXTSWIZZLE_ONE" +}; + +static const char *TGSI_WRITEMASKS[] = +{ + "0", + "WRITEMASK_X", + "WRITEMASK_Y", + "WRITEMASK_XY", + "WRITEMASK_Z", + "WRITEMASK_XZ", + "WRITEMASK_YZ", + "WRITEMASK_XYZ", + "WRITEMASK_W", + "WRITEMASK_XW", + "WRITEMASK_YW", + "WRITEMASK_XYW", + "WRITEMASK_ZW", + "WRITEMASK_XZW", + "WRITEMASK_YZW", + "WRITEMASK_XYZW" +}; + +static const char *TGSI_DST_REGISTER_EXTS[] = +{ + "DST_REGISTER_EXT_TYPE_CONDCODE", + "DST_REGISTER_EXT_TYPE_MODULATE" +}; + +static const char *TGSI_MODULATES[] = +{ + "MODULATE_1X", + "MODULATE_2X", + "MODULATE_4X", + "MODULATE_8X", + "MODULATE_HALF", + "MODULATE_QUARTER", + "MODULATE_EIGHTH" +}; + +static void +dump_declaration_verbose( + struct tgsi_full_declaration *decl, + unsigned ignored, + unsigned deflt, + struct tgsi_full_declaration *fd ) +{ + TXT( "\nFile : " ); + ENM( decl->Declaration.File, TGSI_FILES ); + if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { + TXT( "\nUsageMask : " ); + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { + CHR( 'X' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { + CHR( 'Y' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { + CHR( 'Z' ); + } + if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { + CHR( 'W' ); + } + } + if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { + TXT( "\nInterpolate: " ); + ENM( decl->Declaration.Interpolate, TGSI_INTERPOLATES ); + } + if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { + TXT( "\nSemantic : " ); + UID( decl->Declaration.Semantic ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Declaration.Padding ); + } + + EOL(); + TXT( "\nFirst: " ); + UID( decl->DeclarationRange.First ); + TXT( "\nLast : " ); + UID( decl->DeclarationRange.Last ); + + if( decl->Declaration.Semantic ) { + EOL(); + TXT( "\nSemanticName : " ); + ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); + TXT( "\nSemanticIndex: " ); + UID( decl->Semantic.SemanticIndex ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( decl->Semantic.Padding ); + } + } +} + +static void +dump_immediate_verbose( + struct tgsi_full_immediate *imm, + unsigned ignored ) +{ + unsigned i; + + TXT( "\nDataType : " ); + ENM( imm->Immediate.DataType, TGSI_IMMS ); + if( ignored ) { + TXT( "\nPadding : " ); + UIX( imm->Immediate.Padding ); + } + + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + EOL(); + switch( imm->Immediate.DataType ) { + case TGSI_IMM_FLOAT32: + TXT( "\nFloat: " ); + FLT( imm->u.ImmediateFloat32[i].Float ); + break; + + default: + assert( 0 ); + } + } +} + +static void +dump_instruction_verbose( + struct tgsi_full_instruction *inst, + unsigned ignored, + unsigned deflt, + struct tgsi_full_instruction *fi ) +{ + unsigned i; + + TXT( "\nOpcode : " ); + ENM( inst->Instruction.Opcode, TGSI_OPCODES ); + if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { + TXT( "\nSaturate : " ); + ENM( inst->Instruction.Saturate, TGSI_SATS ); + } + if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { + TXT( "\nNumDstRegs : " ); + UID( inst->Instruction.NumDstRegs ); + } + if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { + TXT( "\nNumSrcRegs : " ); + UID( inst->Instruction.NumSrcRegs ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->Instruction.Padding ); + } + + if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { + EOL(); + TXT( "\nType : " ); + ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { + TXT( "\nPrecision : " ); + ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); + } + if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { + TXT( "\nCondDstIndex : " ); + UID( inst->InstructionExtNv.CondDstIndex ); + } + if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { + TXT( "\nCondFlowIndex : " ); + UID( inst->InstructionExtNv.CondFlowIndex ); + } + if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { + TXT( "\nCondMask : " ); + ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { + TXT( "\nCondSwizzleX : " ); + ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { + TXT( "\nCondSwizzleY : " ); + ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ : " ); + ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { + TXT( "\nCondSwizzleW : " ); + ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { + TXT( "\nCondDstUpdate : " ); + UID( inst->InstructionExtNv.CondDstUpdate ); + } + if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { + TXT( "\nCondFlowEnable: " ); + UID( inst->InstructionExtNv.CondFlowEnable ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtNv.Padding ); + if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { + TXT( "\nExtended : " ); + UID( inst->InstructionExtNv.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { + EOL(); + TXT( "\nType : " ); + ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { + TXT( "\nLabel : " ); + UID( inst->InstructionExtLabel.Label ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtLabel.Padding ); + if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtLabel.Extended ); + } + } + } + + if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { + EOL(); + TXT( "\nType : " ); + ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); + if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { + TXT( "\nTexture : " ); + ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( inst->InstructionExtTexture.Padding ); + if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { + TXT( "\nExtended: " ); + UID( inst->InstructionExtTexture.Extended ); + } + } + } + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; + struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; + + EOL(); + TXT( "\nFile : " ); + ENM( dst->DstRegister.File, TGSI_FILES ); + if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { + TXT( "\nWriteMask: " ); + ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); + } + if( ignored ) { + if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( dst->DstRegister.Indirect ); + } + if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( dst->DstRegister.Dimension ); + } + } + if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { + TXT( "\nIndex : " ); + SID( dst->DstRegister.Index ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegister.Padding ); + if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegister.Extended ); + } + } + + if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { + EOL(); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { + TXT( "\nCondMask : " ); + ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { + TXT( "\nCondSwizzleX: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { + TXT( "\nCondSwizzleY: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { + TXT( "\nCondSwizzleZ: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { + TXT( "\nCondSwizzleW: " ); + ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { + TXT( "\nCondSrcIndex: " ); + UID( dst->DstRegisterExtConcode.CondSrcIndex ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtConcode.Padding ); + if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { + TXT( "\nExtended : " ); + UID( dst->DstRegisterExtConcode.Extended ); + } + } + } + + if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { + EOL(); + TXT( "\nType : " ); + ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); + if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { + TXT( "\nModulate: " ); + ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( dst->DstRegisterExtModulate.Padding ); + if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { + TXT( "\nExtended: " ); + UID( dst->DstRegisterExtModulate.Extended ); + } + } + } + } + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; + struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; + + EOL(); + TXT( "\nFile : "); + ENM( src->SrcRegister.File, TGSI_FILES ); + if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { + TXT( "\nSwizzleX : " ); + ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { + TXT( "\nSwizzleY : " ); + ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { + TXT( "\nSwizzleZ : " ); + ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { + TXT( "\nSwizzleW : " ); + ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); + } + if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegister.Negate ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { + TXT( "\nIndirect : " ); + UID( src->SrcRegister.Indirect ); + } + if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { + TXT( "\nDimension: " ); + UID( src->SrcRegister.Dimension ); + } + } + if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { + TXT( "\nIndex : " ); + SID( src->SrcRegister.Index ); + } + if( ignored ) { + if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegister.Extended ); + } + } + + if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { + EOL(); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { + TXT( "\nExtSwizzleX: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { + TXT( "\nExtSwizzleY: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { + TXT( "\nExtSwizzleZ: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { + TXT( "\nExtSwizzleW: " ); + ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { + TXT( "\nNegateX : " ); + UID( src->SrcRegisterExtSwz.NegateX ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { + TXT( "\nNegateY : " ); + UID( src->SrcRegisterExtSwz.NegateY ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { + TXT( "\nNegateZ : " ); + UID( src->SrcRegisterExtSwz.NegateZ ); + } + if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { + TXT( "\nNegateW : " ); + UID( src->SrcRegisterExtSwz.NegateW ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtSwz.Padding ); + if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtSwz.Extended ); + } + } + } + + if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { + EOL(); + TXT( "\nType : " ); + ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); + if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { + TXT( "\nComplement: " ); + UID( src->SrcRegisterExtMod.Complement ); + } + if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { + TXT( "\nBias : " ); + UID( src->SrcRegisterExtMod.Bias ); + } + if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { + TXT( "\nScale2X : " ); + UID( src->SrcRegisterExtMod.Scale2X ); + } + if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { + TXT( "\nAbsolute : " ); + UID( src->SrcRegisterExtMod.Absolute ); + } + if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { + TXT( "\nNegate : " ); + UID( src->SrcRegisterExtMod.Negate ); + } + if( ignored ) { + TXT( "\nPadding : " ); + UIX( src->SrcRegisterExtMod.Padding ); + if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { + TXT( "\nExtended : " ); + UID( src->SrcRegisterExtMod.Extended ); + } + } + } + } +} + +void +tgsi_dump_c( + const struct tgsi_token *tokens, + uint flags ) +{ + struct tgsi_parse_context parse; + struct tgsi_full_instruction fi; + struct tgsi_full_declaration fd; + uint ignored = flags & TGSI_DUMP_C_IGNORED; + uint deflt = flags & TGSI_DUMP_C_DEFAULT; + uint instno = 0; + + /* sanity checks */ + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); + assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); + + tgsi_parse_init( &parse, tokens ); + + TXT( "tgsi-dump begin -----------------" ); + + TXT( "\nMajorVersion: " ); + UID( parse.FullVersion.Version.MajorVersion ); + TXT( "\nMinorVersion: " ); + UID( parse.FullVersion.Version.MinorVersion ); + EOL(); + + TXT( "\nHeaderSize: " ); + UID( parse.FullHeader.Header.HeaderSize ); + TXT( "\nBodySize : " ); + UID( parse.FullHeader.Header.BodySize ); + TXT( "\nProcessor : " ); + ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); + EOL(); + + fi = tgsi_default_full_instruction(); + fd = tgsi_default_full_declaration(); + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + tgsi_parse_token( &parse ); + + TXT( "\nType : " ); + ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); + if( ignored ) { + TXT( "\nSize : " ); + UID( parse.FullToken.Token.Size ); + if( deflt || parse.FullToken.Token.Extended ) { + TXT( "\nExtended : " ); + UID( parse.FullToken.Token.Extended ); + } + } + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + dump_declaration_verbose( + &parse.FullToken.FullDeclaration, + ignored, + deflt, + &fd ); + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + dump_immediate_verbose( + &parse.FullToken.FullImmediate, + ignored ); + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + dump_instruction_verbose( + &parse.FullToken.FullInstruction, + ignored, + deflt, + &fi ); + break; + + default: + assert( 0 ); + } + + EOL(); + } + + TXT( "\ntgsi-dump end -------------------\n" ); + + tgsi_parse_free( &parse ); +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.h b/src/gallium/auxiliary/tgsi/tgsi_dump_c.h new file mode 100644 index 0000000000..d91cd35b3b --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.h @@ -0,0 +1,49 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_DUMP_C_H +#define TGSI_DUMP_C_H + +#include "pipe/p_shader_tokens.h" + +#if defined __cplusplus +extern "C" { +#endif + +#define TGSI_DUMP_C_IGNORED 1 +#define TGSI_DUMP_C_DEFAULT 2 + +void +tgsi_dump_c( + const struct tgsi_token *tokens, + uint flags ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_DUMP_C_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c new file mode 100644 index 0000000000..8b430548bc --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -0,0 +1,2522 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI interpretor/executor. + * + * Flow control information: + * + * Since we operate on 'quads' (4 pixels or 4 vertices in parallel) + * flow control statements (IF/ELSE/ENDIF, LOOP/ENDLOOP) require special + * care since a condition may be true for some quad components but false + * for other components. + * + * We basically execute all statements (even if they're in the part of + * an IF/ELSE clause that's "not taken") and use a special mask to + * control writing to destination registers. This is the ExecMask. + * See store_dest(). + * + * The ExecMask is computed from three other masks (CondMask, LoopMask and + * ContMask) which are controlled by the flow control instructions (namely: + * (IF/ELSE/ENDIF, LOOP/ENDLOOP and CONT). + * + * + * Authors: + * Michal Krol + * Brian Paul + */ + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" +#include "tgsi_exec.h" + +#define TILE_TOP_LEFT 0 +#define TILE_TOP_RIGHT 1 +#define TILE_BOTTOM_LEFT 2 +#define TILE_BOTTOM_RIGHT 3 + +/* + * Shorthand locations of various utility registers (_I = Index, _C = Channel) + */ +#define TEMP_0_I TGSI_EXEC_TEMP_00000000_I +#define TEMP_0_C TGSI_EXEC_TEMP_00000000_C +#define TEMP_7F_I TGSI_EXEC_TEMP_7FFFFFFF_I +#define TEMP_7F_C TGSI_EXEC_TEMP_7FFFFFFF_C +#define TEMP_80_I TGSI_EXEC_TEMP_80000000_I +#define TEMP_80_C TGSI_EXEC_TEMP_80000000_C +#define TEMP_FF_I TGSI_EXEC_TEMP_FFFFFFFF_I +#define TEMP_FF_C TGSI_EXEC_TEMP_FFFFFFFF_C +#define TEMP_1_I TGSI_EXEC_TEMP_ONE_I +#define TEMP_1_C TGSI_EXEC_TEMP_ONE_C +#define TEMP_2_I TGSI_EXEC_TEMP_TWO_I +#define TEMP_2_C TGSI_EXEC_TEMP_TWO_C +#define TEMP_128_I TGSI_EXEC_TEMP_128_I +#define TEMP_128_C TGSI_EXEC_TEMP_128_C +#define TEMP_M128_I TGSI_EXEC_TEMP_MINUS_128_I +#define TEMP_M128_C TGSI_EXEC_TEMP_MINUS_128_C +#define TEMP_KILMASK_I TGSI_EXEC_TEMP_KILMASK_I +#define TEMP_KILMASK_C TGSI_EXEC_TEMP_KILMASK_C +#define TEMP_OUTPUT_I TGSI_EXEC_TEMP_OUTPUT_I +#define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C +#define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I +#define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C +#define TEMP_3_I TGSI_EXEC_TEMP_THREE_I +#define TEMP_3_C TGSI_EXEC_TEMP_THREE_C +#define TEMP_HALF_I TGSI_EXEC_TEMP_HALF_I +#define TEMP_HALF_C TGSI_EXEC_TEMP_HALF_C +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +#define FOR_EACH_CHANNEL(CHAN)\ + for (CHAN = 0; CHAN < 4; CHAN++) + +#define IS_CHANNEL_ENABLED(INST, CHAN)\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IS_CHANNEL_ENABLED2(INST, CHAN)\ + ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) + +#define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ + FOR_EACH_CHANNEL( CHAN )\ + if (IS_CHANNEL_ENABLED2( INST, CHAN )) + + +/** The execution mask depends on the conditional mask and the loop mask */ +#define UPDATE_EXEC_MASK(MACH) \ + MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask + + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + + + +/** + * Initialize machine state by expanding tokens to full instructions, + * allocating temporary storage, setting up constants, etc. + * After this, we can call tgsi_exec_machine_run() many times. + */ +void +tgsi_exec_machine_bind_shader( + struct tgsi_exec_machine *mach, + const struct tgsi_token *tokens, + uint numSamplers, + struct tgsi_sampler *samplers) +{ + uint k; + struct tgsi_parse_context parse; + struct tgsi_exec_labels *labels = &mach->Labels; + struct tgsi_full_instruction *instructions; + struct tgsi_full_declaration *declarations; + uint maxInstructions = 10, numInstructions = 0; + uint maxDeclarations = 10, numDeclarations = 0; + uint instno = 0; + +#if 0 + tgsi_dump(tokens, 0); +#endif + + mach->Tokens = tokens; + mach->Samplers = samplers; + + k = tgsi_parse_init (&parse, mach->Tokens); + if (k != TGSI_PARSE_OK) { + debug_printf( "Problem parsing!\n" ); + return; + } + + mach->Processor = parse.FullHeader.Processor.Processor; + mach->ImmLimit = 0; + labels->count = 0; + + declarations = (struct tgsi_full_declaration *) + MALLOC( maxDeclarations * sizeof(struct tgsi_full_declaration) ); + + if (!declarations) { + return; + } + + instructions = (struct tgsi_full_instruction *) + MALLOC( maxInstructions * sizeof(struct tgsi_full_instruction) ); + + if (!instructions) { + FREE( declarations ); + return; + } + + while( !tgsi_parse_end_of_tokens( &parse ) ) { + uint pointer = parse.Position; + uint i; + + tgsi_parse_token( &parse ); + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + /* save expanded declaration */ + if (numDeclarations == maxDeclarations) { + declarations = REALLOC(declarations, + maxDeclarations + * sizeof(struct tgsi_full_declaration), + (maxDeclarations + 10) + * sizeof(struct tgsi_full_declaration)); + maxDeclarations += 10; + } + memcpy(declarations + numDeclarations, + &parse.FullToken.FullDeclaration, + sizeof(declarations[0])); + numDeclarations++; + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; + assert( size % 4 == 0 ); + assert( mach->ImmLimit + size / 4 <= TGSI_EXEC_NUM_IMMEDIATES ); + + for( i = 0; i < size; i++ ) { + mach->Imms[mach->ImmLimit + i / 4][i % 4] = + parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } + mach->ImmLimit += size / 4; + } + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + assert( labels->count < MAX_LABELS ); + + labels->labels[labels->count][0] = instno; + labels->labels[labels->count][1] = pointer; + labels->count++; + + /* save expanded instruction */ + if (numInstructions == maxInstructions) { + instructions = REALLOC(instructions, + maxInstructions + * sizeof(struct tgsi_full_instruction), + (maxInstructions + 10) + * sizeof(struct tgsi_full_instruction)); + maxInstructions += 10; + } + memcpy(instructions + numInstructions, + &parse.FullToken.FullInstruction, + sizeof(instructions[0])); + numInstructions++; + break; + + default: + assert( 0 ); + } + } + tgsi_parse_free (&parse); + + if (mach->Declarations) { + FREE( mach->Declarations ); + } + mach->Declarations = declarations; + mach->NumDeclarations = numDeclarations; + + if (mach->Instructions) { + FREE( mach->Instructions ); + } + mach->Instructions = instructions; + mach->NumInstructions = numInstructions; +} + + +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach ) +{ + uint i; + + mach->Temps = (struct tgsi_exec_vector *) tgsi_align_128bit( mach->_Temps); + mach->Addrs = &mach->Temps[TGSI_EXEC_TEMP_ADDR]; + + /* Setup constants. */ + for( i = 0; i < 4; i++ ) { + mach->Temps[TEMP_0_I].xyzw[TEMP_0_C].u[i] = 0x00000000; + mach->Temps[TEMP_7F_I].xyzw[TEMP_7F_C].u[i] = 0x7FFFFFFF; + mach->Temps[TEMP_80_I].xyzw[TEMP_80_C].u[i] = 0x80000000; + mach->Temps[TEMP_FF_I].xyzw[TEMP_FF_C].u[i] = 0xFFFFFFFF; + mach->Temps[TEMP_1_I].xyzw[TEMP_1_C].f[i] = 1.0f; + mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f; + mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f; + mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f; + mach->Temps[TEMP_3_I].xyzw[TEMP_3_C].f[i] = 3.0f; + mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C].f[i] = 0.5f; + } +} + + +void +tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach) +{ + if (mach->Instructions) { + FREE(mach->Instructions); + mach->Instructions = NULL; + mach->NumInstructions = 0; + } + if (mach->Declarations) { + FREE(mach->Declarations); + mach->Declarations = NULL; + mach->NumDeclarations = 0; + } +} + + +static void +micro_abs( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = fabsf( src->f[0] ); + dst->f[1] = fabsf( src->f[1] ); + dst->f[2] = fabsf( src->f[2] ); + dst->f[3] = fabsf( src->f[3] ); +} + +static void +micro_add( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] + src1->f[0]; + dst->f[1] = src0->f[1] + src1->f[1]; + dst->f[2] = src0->f[2] + src1->f[2]; + dst->f[3] = src0->f[3] + src1->f[3]; +} + +static void +micro_iadd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] + src1->i[0]; + dst->i[1] = src0->i[1] + src1->i[1]; + dst->i[2] = src0->i[2] + src1->i[2]; + dst->i[3] = src0->i[3] + src1->i[3]; +} + +static void +micro_and( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] & src1->u[0]; + dst->u[1] = src0->u[1] & src1->u[1]; + dst->u[2] = src0->u[2] & src1->u[2]; + dst->u[3] = src0->u[3] & src1->u[3]; +} + +static void +micro_ceil( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = ceilf( src->f[0] ); + dst->f[1] = ceilf( src->f[1] ); + dst->f[2] = ceilf( src->f[2] ); + dst->f[3] = ceilf( src->f[3] ); +} + +static void +micro_cos( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = cosf( src->f[0] ); + dst->f[1] = cosf( src->f[1] ); + dst->f[2] = cosf( src->f[2] ); + dst->f[3] = cosf( src->f[3] ); +} + +static void +micro_ddx( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_BOTTOM_RIGHT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_ddy( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = + dst->f[1] = + dst->f[2] = + dst->f[3] = src->f[TILE_TOP_LEFT] - src->f[TILE_BOTTOM_LEFT]; +} + +static void +micro_div( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] / src1->f[0]; + dst->f[1] = src0->f[1] / src1->f[1]; + dst->f[2] = src0->f[2] / src1->f[2]; + dst->f[3] = src0->f[3] / src1->f[3]; +} + +static void +micro_udiv( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] / src1->u[0]; + dst->u[1] = src0->u[1] / src1->u[1]; + dst->u[2] = src0->u[2] / src1->u[2]; + dst->u[3] = src0->u[3] / src1->u[3]; +} + +static void +micro_eq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] == src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] == src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] == src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] == src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ieq( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] == src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] == src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] == src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] == src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_exp2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src) +{ + dst->f[0] = powf( 2.0f, src->f[0] ); + dst->f[1] = powf( 2.0f, src->f[1] ); + dst->f[2] = powf( 2.0f, src->f[2] ); + dst->f[3] = powf( 2.0f, src->f[3] ); +} + +static void +micro_f2it( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = (int) src->f[0]; + dst->i[1] = (int) src->f[1]; + dst->i[2] = (int) src->f[2]; + dst->i[3] = (int) src->f[3]; +} + +static void +micro_f2ut( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = (uint) src->f[0]; + dst->u[1] = (uint) src->f[1]; + dst->u[2] = (uint) src->f[2]; + dst->u[3] = (uint) src->f[3]; +} + +static void +micro_flr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = floorf( src->f[0] ); + dst->f[1] = floorf( src->f[1] ); + dst->f[2] = floorf( src->f[2] ); + dst->f[3] = floorf( src->f[3] ); +} + +static void +micro_frc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = src->f[0] - floorf( src->f[0] ); + dst->f[1] = src->f[1] - floorf( src->f[1] ); + dst->f[2] = src->f[2] - floorf( src->f[2] ); + dst->f[3] = src->f[3] - floorf( src->f[3] ); +} + +static void +micro_ge( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] >= src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] >= src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] >= src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] >= src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_i2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) src->i[0]; + dst->f[1] = (float) src->i[1]; + dst->f[2] = (float) src->i[2]; + dst->f[3] = (float) src->i[3]; +} + +static void +micro_lg2( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = logf( src->f[0] ) * 1.442695f; + dst->f[1] = logf( src->f[1] ) * 1.442695f; + dst->f[2] = logf( src->f[2] ) * 1.442695f; + dst->f[3] = logf( src->f[3] ) * 1.442695f; +} + +static void +micro_le( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] <= src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] <= src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] <= src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] <= src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_lt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src2->f[0] : src3->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src2->f[1] : src3->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src2->f[2] : src3->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src2->f[3] : src3->f[3]; +} + +static void +micro_ilt( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src2->i[0] : src3->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src2->i[1] : src3->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src2->i[2] : src3->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src2->i[3] : src3->i[3]; +} + +static void +micro_ult( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2, + const union tgsi_exec_channel *src3 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src2->u[0] : src3->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src2->u[1] : src3->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src2->u[2] : src3->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src2->u[3] : src3->u[3]; +} + +static void +micro_max( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] > src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] > src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] > src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] > src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] > src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] > src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] > src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] > src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umax( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] > src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] > src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] > src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] > src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_min( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] < src1->f[0] ? src0->f[0] : src1->f[0]; + dst->f[1] = src0->f[1] < src1->f[1] ? src0->f[1] : src1->f[1]; + dst->f[2] = src0->f[2] < src1->f[2] ? src0->f[2] : src1->f[2]; + dst->f[3] = src0->f[3] < src1->f[3] ? src0->f[3] : src1->f[3]; +} + +static void +micro_imin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] < src1->i[0] ? src0->i[0] : src1->i[0]; + dst->i[1] = src0->i[1] < src1->i[1] ? src0->i[1] : src1->i[1]; + dst->i[2] = src0->i[2] < src1->i[2] ? src0->i[2] : src1->i[2]; + dst->i[3] = src0->i[3] < src1->i[3] ? src0->i[3] : src1->i[3]; +} + +static void +micro_umin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] < src1->u[0] ? src0->u[0] : src1->u[0]; + dst->u[1] = src0->u[1] < src1->u[1] ? src0->u[1] : src1->u[1]; + dst->u[2] = src0->u[2] < src1->u[2] ? src0->u[2] : src1->u[2]; + dst->u[3] = src0->u[3] < src1->u[3] ? src0->u[3] : src1->u[3]; +} + +static void +micro_umod( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] % src1->u[0]; + dst->u[1] = src0->u[1] % src1->u[1]; + dst->u[2] = src0->u[2] % src1->u[2]; + dst->u[3] = src0->u[3] % src1->u[3]; +} + +static void +micro_mul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] * src1->f[0]; + dst->f[1] = src0->f[1] * src1->f[1]; + dst->f[2] = src0->f[2] * src1->f[2]; + dst->f[3] = src0->f[3] * src1->f[3]; +} + +static void +micro_imul( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] * src1->i[0]; + dst->i[1] = src0->i[1] * src1->i[1]; + dst->i[2] = src0->i[2] * src1->i[2]; + dst->i[3] = src0->i[3] * src1->i[3]; +} + +static void +micro_imul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->i[0] = src0->i[0] * src1->i[0]; + dst1->i[1] = src0->i[1] * src1->i[1]; + dst1->i[2] = src0->i[2] * src1->i[2]; + dst1->i[3] = src0->i[3] * src1->i[3]; + dst0->i[0] = 0; + dst0->i[1] = 0; + dst0->i[2] = 0; + dst0->i[3] = 0; +} + +static void +micro_umul64( + union tgsi_exec_channel *dst0, + union tgsi_exec_channel *dst1, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst1->u[0] = src0->u[0] * src1->u[0]; + dst1->u[1] = src0->u[1] * src1->u[1]; + dst1->u[2] = src0->u[2] * src1->u[2]; + dst1->u[3] = src0->u[3] * src1->u[3]; + dst0->u[0] = 0; + dst0->u[1] = 0; + dst0->u[2] = 0; + dst0->u[3] = 0; +} + +static void +micro_movc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1, + const union tgsi_exec_channel *src2 ) +{ + dst->u[0] = src0->u[0] ? src1->u[0] : src2->u[0]; + dst->u[1] = src0->u[1] ? src1->u[1] : src2->u[1]; + dst->u[2] = src0->u[2] ? src1->u[2] : src2->u[2]; + dst->u[3] = src0->u[3] ? src1->u[3] : src2->u[3]; +} + +static void +micro_neg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = -src->f[0]; + dst->f[1] = -src->f[1]; + dst->f[2] = -src->f[2]; + dst->f[3] = -src->f[3]; +} + +static void +micro_ineg( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->i[0] = -src->i[0]; + dst->i[1] = -src->i[1]; + dst->i[2] = -src->i[2]; + dst->i[3] = -src->i[3]; +} + +static void +micro_not( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->u[0] = ~src->u[0]; + dst->u[1] = ~src->u[1]; + dst->u[2] = ~src->u[2]; + dst->u[3] = ~src->u[3]; +} + +static void +micro_or( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] | src1->u[0]; + dst->u[1] = src0->u[1] | src1->u[1]; + dst->u[2] = src0->u[2] | src1->u[2]; + dst->u[3] = src0->u[3] | src1->u[3]; +} + +static void +micro_pow( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = powf( src0->f[0], src1->f[0] ); + dst->f[1] = powf( src0->f[1], src1->f[1] ); + dst->f[2] = powf( src0->f[2], src1->f[2] ); + dst->f[3] = powf( src0->f[3], src1->f[3] ); +} + +static void +micro_rnd( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = floorf( src->f[0] + 0.5f ); + dst->f[1] = floorf( src->f[1] + 0.5f ); + dst->f[2] = floorf( src->f[2] + 0.5f ); + dst->f[3] = floorf( src->f[3] + 0.5f ); +} + +static void +micro_shl( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] << src1->i[0]; + dst->i[1] = src0->i[1] << src1->i[1]; + dst->i[2] = src0->i[2] << src1->i[2]; + dst->i[3] = src0->i[3] << src1->i[3]; +} + +static void +micro_ishr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->i[0] = src0->i[0] >> src1->i[0]; + dst->i[1] = src0->i[1] >> src1->i[1]; + dst->i[2] = src0->i[2] >> src1->i[2]; + dst->i[3] = src0->i[3] >> src1->i[3]; +} + +static void +micro_trunc( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0 ) +{ + dst->f[0] = (float) (int) src0->f[0]; + dst->f[1] = (float) (int) src0->f[1]; + dst->f[2] = (float) (int) src0->f[2]; + dst->f[3] = (float) (int) src0->f[3]; +} + +static void +micro_ushr( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] >> src1->u[0]; + dst->u[1] = src0->u[1] >> src1->u[1]; + dst->u[2] = src0->u[2] >> src1->u[2]; + dst->u[3] = src0->u[3] >> src1->u[3]; +} + +static void +micro_sin( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = sinf( src->f[0] ); + dst->f[1] = sinf( src->f[1] ); + dst->f[2] = sinf( src->f[2] ); + dst->f[3] = sinf( src->f[3] ); +} + +static void +micro_sqrt( union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = sqrtf( src->f[0] ); + dst->f[1] = sqrtf( src->f[1] ); + dst->f[2] = sqrtf( src->f[2] ); + dst->f[3] = sqrtf( src->f[3] ); +} + +static void +micro_sub( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->f[0] = src0->f[0] - src1->f[0]; + dst->f[1] = src0->f[1] - src1->f[1]; + dst->f[2] = src0->f[2] - src1->f[2]; + dst->f[3] = src0->f[3] - src1->f[3]; +} + +static void +micro_u2f( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src ) +{ + dst->f[0] = (float) src->u[0]; + dst->f[1] = (float) src->u[1]; + dst->f[2] = (float) src->u[2]; + dst->f[3] = (float) src->u[3]; +} + +static void +micro_xor( + union tgsi_exec_channel *dst, + const union tgsi_exec_channel *src0, + const union tgsi_exec_channel *src1 ) +{ + dst->u[0] = src0->u[0] ^ src1->u[0]; + dst->u[1] = src0->u[1] ^ src1->u[1]; + dst->u[2] = src0->u[2] ^ src1->u[2]; + dst->u[3] = src0->u[3] ^ src1->u[3]; +} + +static void +fetch_src_file_channel( + const struct tgsi_exec_machine *mach, + const uint file, + const uint swizzle, + const union tgsi_exec_channel *index, + union tgsi_exec_channel *chan ) +{ + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( file ) { + case TGSI_FILE_CONSTANT: + chan->f[0] = mach->Consts[index->i[0]][swizzle]; + chan->f[1] = mach->Consts[index->i[1]][swizzle]; + chan->f[2] = mach->Consts[index->i[2]][swizzle]; + chan->f[3] = mach->Consts[index->i[3]][swizzle]; + break; + + case TGSI_FILE_INPUT: + chan->u[0] = mach->Inputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Inputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Inputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Inputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_TEMPORARY: + assert(index->i[0] < TGSI_EXEC_NUM_TEMPS); + chan->u[0] = mach->Temps[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Temps[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Temps[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Temps[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_IMMEDIATE: + assert( index->i[0] < (int) mach->ImmLimit ); + chan->f[0] = mach->Imms[index->i[0]][swizzle]; + assert( index->i[1] < (int) mach->ImmLimit ); + chan->f[1] = mach->Imms[index->i[1]][swizzle]; + assert( index->i[2] < (int) mach->ImmLimit ); + chan->f[2] = mach->Imms[index->i[2]][swizzle]; + assert( index->i[3] < (int) mach->ImmLimit ); + chan->f[3] = mach->Imms[index->i[3]][swizzle]; + break; + + case TGSI_FILE_ADDRESS: + chan->u[0] = mach->Addrs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Addrs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Addrs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Addrs[index->i[3]].xyzw[swizzle].u[3]; + break; + + case TGSI_FILE_OUTPUT: + /* vertex/fragment output vars can be read too */ + chan->u[0] = mach->Outputs[index->i[0]].xyzw[swizzle].u[0]; + chan->u[1] = mach->Outputs[index->i[1]].xyzw[swizzle].u[1]; + chan->u[2] = mach->Outputs[index->i[2]].xyzw[swizzle].u[2]; + chan->u[3] = mach->Outputs[index->i[3]].xyzw[swizzle].u[3]; + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + *chan = mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]; + break; + + case TGSI_EXTSWIZZLE_ONE: + *chan = mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]; + break; + + default: + assert( 0 ); + } +} + +static void +fetch_source( + const struct tgsi_exec_machine *mach, + union tgsi_exec_channel *chan, + const struct tgsi_full_src_register *reg, + const uint chan_index ) +{ + union tgsi_exec_channel index; + uint swizzle; + + index.i[0] = + index.i[1] = + index.i[2] = + index.i[3] = reg->SrcRegister.Index; + + if (reg->SrcRegister.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + + if( reg->SrcRegister.Dimension ) { + switch( reg->SrcRegister.File ) { + case TGSI_FILE_INPUT: + index.i[0] *= 17; + index.i[1] *= 17; + index.i[2] *= 17; + index.i[3] *= 17; + break; + case TGSI_FILE_CONSTANT: + index.i[0] *= 4096; + index.i[1] *= 4096; + index.i[2] *= 4096; + index.i[3] *= 4096; + break; + default: + assert( 0 ); + } + + index.i[0] += reg->SrcRegisterDim.Index; + index.i[1] += reg->SrcRegisterDim.Index; + index.i[2] += reg->SrcRegisterDim.Index; + index.i[3] += reg->SrcRegisterDim.Index; + + if (reg->SrcRegisterDim.Indirect) { + union tgsi_exec_channel index2; + union tgsi_exec_channel indir_index; + + index2.i[0] = + index2.i[1] = + index2.i[2] = + index2.i[3] = reg->SrcRegisterDimInd.Index; + + swizzle = tgsi_util_get_src_register_swizzle( ®->SrcRegisterDimInd, CHAN_X ); + fetch_src_file_channel( + mach, + reg->SrcRegisterDimInd.File, + swizzle, + &index2, + &indir_index ); + + index.i[0] += indir_index.i[0]; + index.i[1] += indir_index.i[1]; + index.i[2] += indir_index.i[2]; + index.i[3] += indir_index.i[3]; + } + } + + swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + fetch_src_file_channel( + mach, + reg->SrcRegister.File, + swizzle, + &index, + chan ); + + switch (tgsi_util_get_full_src_register_sign_mode( reg, chan_index )) { + case TGSI_UTIL_SIGN_CLEAR: + micro_abs( chan, chan ); + break; + + case TGSI_UTIL_SIGN_SET: + micro_abs( chan, chan ); + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + micro_neg( chan, chan ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } + + if (reg->SrcRegisterExtMod.Complement) { + micro_sub( chan, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], chan ); + } +} + +static void +store_dest( + struct tgsi_exec_machine *mach, + const union tgsi_exec_channel *chan, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + uint chan_index ) +{ + union tgsi_exec_channel *dst; + + switch( reg->DstRegister.File ) { + case TGSI_FILE_NULL: + return; + + case TGSI_FILE_OUTPUT: + dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] + + reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_TEMPORARY: + assert(reg->DstRegister.Index < TGSI_EXEC_NUM_TEMPS); + dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; + break; + + case TGSI_FILE_ADDRESS: + dst = &mach->Addrs[reg->DstRegister.Index].xyzw[chan_index]; + break; + + default: + assert( 0 ); + return; + } + + switch (inst->Instruction.Saturate) + { + case TGSI_SAT_NONE: + if (mach->ExecMask & 0x1) + dst->i[0] = chan->i[0]; + if (mach->ExecMask & 0x2) + dst->i[1] = chan->i[1]; + if (mach->ExecMask & 0x4) + dst->i[2] = chan->i[2]; + if (mach->ExecMask & 0x8) + dst->i[3] = chan->i[3]; + break; + + case TGSI_SAT_ZERO_ONE: + /* XXX need to obey ExecMask here */ + micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); + micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + + default: + assert( 0 ); + } +} + +#define FETCH(VAL,INDEX,CHAN)\ + fetch_source (mach, VAL, &inst->FullSrcRegisters[INDEX], CHAN) + +#define STORE(VAL,INDEX,CHAN)\ + store_dest (mach, VAL, &inst->FullDstRegisters[INDEX], inst, CHAN ) + + +/** + * Execute ARB-style KIL which is predicated by a src register. + * Kill fragment if any of the four values is less than zero. + */ +static void +exec_kilp(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + uint uniquemask; + uint chan_index; + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + union tgsi_exec_channel r[1]; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + for (chan_index = 0; chan_index < 4; chan_index++) + { + uint swizzle; + uint i; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle ( + &inst->FullSrcRegisters[0], + chan_index); + + /* check if the component has not been already tested */ + if (uniquemask & (1 << swizzle)) + continue; + uniquemask |= 1 << swizzle; + + FETCH(&r[0], 0, chan_index); + for (i = 0; i < 4; i++) + if (r[0].f[i] < 0.0f) + kilmask |= 1 << i; + } + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; +} + + +/* + * Fetch a texel using STR texture coordinates. + */ +static void +fetch_texel( struct tgsi_sampler *sampler, + const union tgsi_exec_channel *s, + const union tgsi_exec_channel *t, + const union tgsi_exec_channel *p, + float lodbias, /* XXX should be float[4] */ + union tgsi_exec_channel *r, + union tgsi_exec_channel *g, + union tgsi_exec_channel *b, + union tgsi_exec_channel *a ) +{ + uint j; + float rgba[NUM_CHANNELS][QUAD_SIZE]; + + sampler->get_samples(sampler, s->f, t->f, p->f, lodbias, rgba); + + for (j = 0; j < 4; j++) { + r->f[j] = rgba[0][j]; + g->f[j] = rgba[1][j]; + b->f[j] = rgba[2][j]; + a->f[j] = rgba[3][j]; + } +} + + +static void +exec_tex(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + boolean biasLod, + boolean projected) +{ + const uint unit = inst->FullSrcRegisters[1].SrcRegister.Index; + union tgsi_exec_channel r[8]; + uint chan_index; + float lodBias; + + /* debug_printf("Sampler %u unit %u\n", sampler, unit); */ + + switch (inst->InstructionExtTexture.Texture) { + case TGSI_TEXTURE_1D: + + FETCH(&r[0], 0, CHAN_X); + + if (projected) { + FETCH(&r[1], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[1] ); + } + + if (biasLod) { + FETCH(&r[1], 0, CHAN_W); + lodBias = r[2].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], NULL, NULL, lodBias, /* S, T, P, BIAS */ + &r[0], &r[1], &r[2], &r[3]); /* R, G, B, A */ + break; + + case TGSI_TEXTURE_2D: + case TGSI_TEXTURE_RECT: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + if (projected) { + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, /* inputs */ + &r[0], &r[1], &r[2], &r[3]); /* outputs */ + break; + + case TGSI_TEXTURE_3D: + case TGSI_TEXTURE_CUBE: + + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 0, CHAN_Z); + + if (projected) { + FETCH(&r[3], 0, CHAN_W); + micro_div( &r[0], &r[0], &r[3] ); + micro_div( &r[1], &r[1], &r[3] ); + micro_div( &r[2], &r[2], &r[3] ); + } + + if (biasLod) { + FETCH(&r[3], 0, CHAN_W); + lodBias = r[3].f[0]; + } + else + lodBias = 0.0; + + fetch_texel(&mach->Samplers[unit], + &r[0], &r[1], &r[2], lodBias, + &r[0], &r[1], &r[2], &r[3]); + break; + + default: + assert (0); + } + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[chan_index], 0, chan_index ); + } +} + + +/** + * Evaluate a constant-valued coefficient at the position of the + * current quad. + */ +static void +eval_constant_coef( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + unsigned i; + + for( i = 0; i < QUAD_SIZE; i++ ) { + mach->Inputs[attrib].xyzw[chan].f[i] = mach->InterpCoefs[attrib].a0[chan]; + } +} + +/** + * Evaluate a linear-valued coefficient at the position of the + * current quad. + */ +static void +eval_linear_coef( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + mach->Inputs[attrib].xyzw[chan].f[0] = a0; + mach->Inputs[attrib].xyzw[chan].f[1] = a0 + dadx; + mach->Inputs[attrib].xyzw[chan].f[2] = a0 + dady; + mach->Inputs[attrib].xyzw[chan].f[3] = a0 + dadx + dady; +} + +/** + * Evaluate a perspective-valued coefficient at the position of the + * current quad. + */ +static void +eval_perspective_coef( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ) +{ + const float x = mach->QuadPos.xyzw[0].f[0]; + const float y = mach->QuadPos.xyzw[1].f[0]; + const float dadx = mach->InterpCoefs[attrib].dadx[chan]; + const float dady = mach->InterpCoefs[attrib].dady[chan]; + const float a0 = mach->InterpCoefs[attrib].a0[chan] + dadx * x + dady * y; + const float *w = mach->QuadPos.xyzw[3].f; + /* divide by W here */ + mach->Inputs[attrib].xyzw[chan].f[0] = a0 / w[0]; + mach->Inputs[attrib].xyzw[chan].f[1] = (a0 + dadx) / w[1]; + mach->Inputs[attrib].xyzw[chan].f[2] = (a0 + dady) / w[2]; + mach->Inputs[attrib].xyzw[chan].f[3] = (a0 + dadx + dady) / w[3]; +} + + +typedef void (* eval_coef_func)( + struct tgsi_exec_machine *mach, + unsigned attrib, + unsigned chan ); + +static void +exec_declaration( + struct tgsi_exec_machine *mach, + const struct tgsi_full_declaration *decl ) +{ + if( mach->Processor == TGSI_PROCESSOR_FRAGMENT ) { + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + eval_coef_func eval; + + first = decl->DeclarationRange.First; + last = decl->DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + switch( decl->Declaration.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + eval = eval_constant_coef; + break; + + case TGSI_INTERPOLATE_LINEAR: + eval = eval_linear_coef; + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + eval = eval_perspective_coef; + break; + + default: + assert( 0 ); + } + + if( mask == TGSI_WRITEMASK_XYZW ) { + unsigned i, j; + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + eval( mach, i, j ); + } + } + } + else { + unsigned i, j; + + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + for( i = first; i <= last; i++ ) { + eval( mach, i, j ); + } + } + } + } + } + } +} + +static void +exec_instruction( + struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst, + int *pc ) +{ + uint chan_index; + union tgsi_exec_channel r[8]; + + (*pc)++; + + switch (inst->Instruction.Opcode) { + case TGSI_OPCODE_ARL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_f2it( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOV: + case TGSI_OPCODE_SWZ: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y ) || IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_X ); + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_max( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[1], 0, CHAN_Y ); + micro_max( &r[1], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + + FETCH( &r[2], 0, CHAN_W ); + micro_min( &r[2], &r[2], &mach->Temps[TEMP_128_I].xyzw[TEMP_128_C] ); + micro_max( &r[2], &r[2], &mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C] ); + micro_pow( &r[1], &r[1], &r[2] ); + micro_lt( &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, CHAN_Z ); + } + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( &r[0], 0, CHAN_X ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( &r[0], 0, CHAN_X ); + micro_sqrt( &r[0], &r[0] ); + micro_div( &r[0], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + FETCH( &r[0], 0, CHAN_X ); + micro_flr( &r[1], &r[0] ); /* r1 = floor(r0) */ + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + micro_exp2( &r[2], &r[1] ); /* r2 = 2 ^ r1 */ + STORE( &r[2], 0, CHAN_X ); /* store r2 */ + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_sub( &r[2], &r[0], &r[1] ); /* r2 = r0 - r1 */ + STORE( &r[2], 0, CHAN_Y ); /* store r2 */ + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + micro_exp2( &r[2], &r[0] ); /* r2 = 2 ^ r0 */ + STORE( &r[2], 0, CHAN_Z ); /* store r2 */ + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_LOG: + FETCH( &r[0], 0, CHAN_X ); + micro_abs( &r[2], &r[0] ); /* r2 = abs(r0) */ + micro_lg2( &r[1], &r[2] ); /* r1 = lg2(r2) */ + micro_flr( &r[0], &r[1] ); /* r0 = floor(r1) */ + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &r[0], 0, CHAN_X ); + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + micro_exp2( &r[0], &r[0] ); /* r0 = 2 ^ r0 */ + micro_div( &r[0], &r[2], &r[0] ); /* r0 = r2 / r0 */ + STORE( &r[0], 0, CHAN_Y ); + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( &r[1], 0, CHAN_Z ); + } + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) + { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_mul( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Z ); + FETCH( &r[2], 1, CHAN_Z ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_W); + FETCH(&r[2], 1, CHAN_W); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_X ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + FETCH( &r[0], 0, CHAN_Y ); + FETCH( &r[1], 1, CHAN_Y); + micro_mul( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, CHAN_Y ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( &r[0], 0, CHAN_Z ); + STORE( &r[0], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + FETCH( &r[0], 1, CHAN_W ); + STORE( &r[0], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + /* XXX use micro_min()?? */ + micro_lt( &r[0], &r[0], &r[1], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + /* XXX use micro_max()?? */ + micro_lt( &r[0], &r[0], &r[1], &r[1], &r[0] ); + + STORE(&r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_lt( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ge( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_mul( &r[0], &r[0], &r[1] ); + FETCH( &r[1], 2, chan_index ); + micro_add( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + + micro_sub( &r[0], &r[0], &r[1] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_sub( &r[1], &r[1], &r[2] ); + micro_mul( &r[0], &r[0], &r[1] ); + micro_add( &r[0], &r[0], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_CND: + assert (0); + break; + + case TGSI_OPCODE_CND0: + assert (0); + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + assert (0); + break; + + case TGSI_OPCODE_INDEX: + assert (0); + break; + + case TGSI_OPCODE_NEGATE: + assert (0); + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_frc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + assert (0); + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_flr( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_rnd( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH(&r[0], 0, CHAN_X); + + micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( &r[0], 0, CHAN_X ); + micro_lg2( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_pow( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + FETCH(&r[0], 0, CHAN_Y); + FETCH(&r[1], 1, CHAN_Z); + + micro_mul( &r[2], &r[0], &r[1] ); + + FETCH(&r[3], 0, CHAN_Z); + FETCH(&r[4], 1, CHAN_Y); + + micro_mul( &r[5], &r[3], &r[4] ); + micro_sub( &r[2], &r[2], &r[5] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( &r[2], 0, CHAN_X ); + } + + FETCH(&r[2], 1, CHAN_X); + + micro_mul( &r[3], &r[3], &r[2] ); + + FETCH(&r[5], 0, CHAN_X); + + micro_mul( &r[1], &r[1], &r[5] ); + micro_sub( &r[3], &r[3], &r[1] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { + STORE( &r[3], 0, CHAN_Y ); + } + + micro_mul( &r[5], &r[5], &r[4] ); + micro_mul( &r[0], &r[0], &r[2] ); + micro_sub( &r[5], &r[5], &r[0] ); + + if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( &r[5], 0, CHAN_Z ); + } + + if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + assert (0); + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + + micro_abs( &r[0], &r[0] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_RCC: + assert (0); + break; + + case TGSI_OPCODE_DPH: + FETCH(&r[0], 0, CHAN_X); + FETCH(&r[1], 1, CHAN_X); + + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Y); + FETCH(&r[2], 1, CHAN_Y); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 0, CHAN_Z); + FETCH(&r[2], 1, CHAN_Z); + + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FETCH(&r[1], 1, CHAN_W); + + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH(&r[0], 0, CHAN_X); + + micro_cos( &r[0], &r[0] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddx( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDY: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ddy( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_KILP: + exec_kilp (mach, inst); + break; + + case TGSI_OPCODE_KIL: + /* for enabled ExecMask bits, set the killed bit */ + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; + break; + + case TGSI_OPCODE_PK2H: + assert (0); + break; + + case TGSI_OPCODE_PK2US: + assert (0); + break; + + case TGSI_OPCODE_PK4B: + assert (0); + break; + + case TGSI_OPCODE_PK4UB: + assert (0); + break; + + case TGSI_OPCODE_RFL: + assert (0); + break; + + case TGSI_OPCODE_SEQ: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_eq( &r[0], &r[0], &r[1], + &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], + &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SFL: + assert (0); + break; + + case TGSI_OPCODE_SGT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SIN: + FETCH( &r[0], 0, CHAN_X ); + micro_sin( &r[0], &r[0] ); + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_le( &r[0], &r[0], &r[1], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SNE: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_eq( &r[0], &r[0], &r[1], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_STR: + assert (0); + break; + + case TGSI_OPCODE_TEX: + /* simple texture lookup */ + /* src[0] = texcoord */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, FALSE, FALSE); + break; + + case TGSI_OPCODE_TXB: + /* Texture lookup with lod bias */ + /* src[0] = texcoord (src[0].w = LOD bias) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE, FALSE); + break; + + case TGSI_OPCODE_TXD: + /* Texture lookup with explict partial derivatives */ + /* src[0] = texcoord */ + /* src[1] = d[strq]/dx */ + /* src[2] = d[strq]/dy */ + /* src[3] = sampler unit */ + assert (0); + break; + + case TGSI_OPCODE_TXL: + /* Texture lookup with explit LOD */ + /* src[0] = texcoord (src[0].w = LOD) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, TRUE, FALSE); + break; + + case TGSI_OPCODE_TXP: + /* Texture lookup with projection */ + /* src[0] = texcoord (src[0].w = projection) */ + /* src[1] = sampler unit */ + exec_tex(mach, inst, FALSE, TRUE); + break; + + case TGSI_OPCODE_UP2H: + assert (0); + break; + + case TGSI_OPCODE_UP2US: + assert (0); + break; + + case TGSI_OPCODE_UP4B: + assert (0); + break; + + case TGSI_OPCODE_UP4UB: + assert (0); + break; + + case TGSI_OPCODE_X2D: + assert (0); + break; + + case TGSI_OPCODE_ARA: + assert (0); + break; + + case TGSI_OPCODE_ARR: + assert (0); + break; + + case TGSI_OPCODE_BRA: + assert (0); + break; + + case TGSI_OPCODE_CAL: + /* skip the call if no execution channels are enabled */ + if (mach->ExecMask) { + /* do the call */ + + /* push the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + + assert(mach->FuncStackTop < TGSI_EXEC_MAX_CALL_NESTING); + mach->FuncStack[mach->FuncStackTop++] = mach->FuncMask; + + /* note that PC was already incremented above */ + mach->CallStack[mach->CallStackTop++] = *pc; + *pc = inst->InstructionExtLabel.Label; + } + break; + + case TGSI_OPCODE_RET: + mach->FuncMask &= ~mach->ExecMask; + UPDATE_EXEC_MASK(mach); + + if (mach->ExecMask == 0x0) { + /* really return now (otherwise, keep executing */ + + if (mach->CallStackTop == 0) { + /* returning from main() */ + *pc = -1; + return; + } + *pc = mach->CallStack[--mach->CallStackTop]; + + /* pop the Cond, Loop, Cont stacks */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + assert(mach->FuncStackTop > 0); + mach->FuncMask = mach->FuncStack[--mach->FuncStackTop]; + + UPDATE_EXEC_MASK(mach); + } + break; + + case TGSI_OPCODE_SSG: + assert (0); + break; + + case TGSI_OPCODE_CMP: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH(&r[0], 0, chan_index); + FETCH(&r[1], 1, chan_index); + FETCH(&r[2], 2, chan_index); + + micro_lt( &r[0], &r[0], &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], &r[1], &r[2] ); + + STORE(&r[0], 0, chan_index); + } + break; + + case TGSI_OPCODE_SCS: + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) || IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( &r[0], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + micro_cos( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_X ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + micro_sin( &r[1], &r[0] ); + STORE( &r[1], 0, CHAN_Y ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + STORE( &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C], 0, CHAN_Z ); + } + if( IS_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_NRM: + assert (0); + break; + + case TGSI_OPCODE_DIV: + assert( 0 ); + break; + + case TGSI_OPCODE_DP2: + FETCH( &r[0], 0, CHAN_X ); + FETCH( &r[1], 1, CHAN_X ); + micro_mul( &r[0], &r[0], &r[1] ); + + FETCH( &r[1], 0, CHAN_Y ); + FETCH( &r[2], 1, CHAN_Y ); + micro_mul( &r[1], &r[1], &r[2] ); + micro_add( &r[0], &r[0], &r[1] ); + + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_IF: + /* push CondMask */ + assert(mach->CondStackTop < TGSI_EXEC_MAX_COND_NESTING); + mach->CondStack[mach->CondStackTop++] = mach->CondMask; + FETCH( &r[0], 0, CHAN_X ); + /* update CondMask */ + if( ! r[0].u[0] ) { + mach->CondMask &= ~0x1; + } + if( ! r[0].u[1] ) { + mach->CondMask &= ~0x2; + } + if( ! r[0].u[2] ) { + mach->CondMask &= ~0x4; + } + if( ! r[0].u[3] ) { + mach->CondMask &= ~0x8; + } + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ELSE */ + break; + + case TGSI_OPCODE_ELSE: + /* invert CondMask wrt previous mask */ + { + uint prevMask; + assert(mach->CondStackTop > 0); + prevMask = mach->CondStack[mach->CondStackTop - 1]; + mach->CondMask = ~mach->CondMask & prevMask; + UPDATE_EXEC_MASK(mach); + /* Todo: If CondMask==0, jump to ENDIF */ + } + break; + + case TGSI_OPCODE_ENDIF: + /* pop CondMask */ + assert(mach->CondStackTop > 0); + mach->CondMask = mach->CondStack[--mach->CondStackTop]; + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_END: + /* halt execution */ + *pc = -1; + break; + + case TGSI_OPCODE_REP: + assert (0); + break; + + case TGSI_OPCODE_ENDREP: + assert (0); + break; + + case TGSI_OPCODE_PUSHA: + assert (0); + break; + + case TGSI_OPCODE_POPA: + assert (0); + break; + + case TGSI_OPCODE_CEIL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_ceil( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_I2F: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_i2f( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_NOT: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_not( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_TRUNC: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + micro_trunc( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHL: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_shl( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SHR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_ishr( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_AND: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_and( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_OR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_or( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_MOD: + assert (0); + break; + + case TGSI_OPCODE_XOR: + FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( &r[0], 0, chan_index ); + FETCH( &r[1], 1, chan_index ); + micro_xor( &r[0], &r[0], &r[1] ); + STORE( &r[0], 0, chan_index ); + } + break; + + case TGSI_OPCODE_SAD: + assert (0); + break; + + case TGSI_OPCODE_TXF: + assert (0); + break; + + case TGSI_OPCODE_TXQ: + assert (0); + break; + + case TGSI_OPCODE_EMIT: + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] += 16; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]]++; + break; + + case TGSI_OPCODE_ENDPRIM: + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]++; + mach->Primitives[mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0]] = 0; + break; + + case TGSI_OPCODE_LOOP: + /* fall-through (for now) */ + case TGSI_OPCODE_BGNLOOP2: + /* push LoopMask and ContMasks */ + assert(mach->LoopStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->LoopStack[mach->LoopStackTop++] = mach->LoopMask; + assert(mach->ContStackTop < TGSI_EXEC_MAX_LOOP_NESTING); + mach->ContStack[mach->ContStackTop++] = mach->ContMask; + break; + + case TGSI_OPCODE_ENDLOOP: + /* fall-through (for now at least) */ + case TGSI_OPCODE_ENDLOOP2: + /* Restore ContMask, but don't pop */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[mach->ContStackTop - 1]; + UPDATE_EXEC_MASK(mach); + if (mach->ExecMask) { + /* repeat loop: jump to instruction just past BGNLOOP */ + *pc = inst->InstructionExtLabel.Label + 1; + } + else { + /* exit loop: pop LoopMask */ + assert(mach->LoopStackTop > 0); + mach->LoopMask = mach->LoopStack[--mach->LoopStackTop]; + /* pop ContMask */ + assert(mach->ContStackTop > 0); + mach->ContMask = mach->ContStack[--mach->ContStackTop]; + } + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BRK: + /* turn off loop channels for each enabled exec channel */ + mach->LoopMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_CONT: + /* turn off cont channels for each enabled exec channel */ + mach->ContMask &= ~mach->ExecMask; + /* Todo: if mach->LoopMask == 0, jump to end of loop */ + UPDATE_EXEC_MASK(mach); + break; + + case TGSI_OPCODE_BGNSUB: + /* no-op */ + break; + + case TGSI_OPCODE_ENDSUB: + /* no-op */ + break; + + case TGSI_OPCODE_NOISE1: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE2: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE3: + assert( 0 ); + break; + + case TGSI_OPCODE_NOISE4: + assert( 0 ); + break; + + case TGSI_OPCODE_NOP: + break; + + default: + assert( 0 ); + } +} + + +/** + * Run TGSI interpreter. + * \return bitmask of "alive" quad components + */ +uint +tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) +{ + uint i; + int pc = 0; + + mach->CondMask = 0xf; + mach->LoopMask = 0xf; + mach->ContMask = 0xf; + mach->FuncMask = 0xf; + mach->ExecMask = 0xf; + + mach->CondStackTop = 0; /* temporarily subvert this assertion */ + assert(mach->CondStackTop == 0); + assert(mach->LoopStackTop == 0); + assert(mach->ContStackTop == 0); + assert(mach->CallStackTop == 0); + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] = 0; + mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] = 0; + + if( mach->Processor == TGSI_PROCESSOR_GEOMETRY ) { + mach->Temps[TEMP_PRIMITIVE_I].xyzw[TEMP_PRIMITIVE_C].u[0] = 0; + mach->Primitives[0] = 0; + } + + + /* execute declarations (interpolants) */ + for (i = 0; i < mach->NumDeclarations; i++) { + exec_declaration( mach, mach->Declarations+i ); + } + + /* execute instructions, until pc is set to -1 */ + while (pc != -1) { + assert(pc < (int) mach->NumInstructions); + exec_instruction( mach, mach->Instructions + pc, &pc ); + } + +#if 0 + /* we scale from floats in [0,1] to Zbuffer ints in sp_quad_depth_test.c */ + if (mach->Processor == TGSI_PROCESSOR_FRAGMENT) { + /* + * Scale back depth component. + */ + for (i = 0; i < 4; i++) + mach->Outputs[0].xyzw[2].f[i] *= ctx->DrawBuffer->_DepthMaxF; + } +#endif + + return ~mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0]; +} + + diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h new file mode 100644 index 0000000000..4f30650b07 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h @@ -0,0 +1,253 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#if !defined TGSI_EXEC_H +#define TGSI_EXEC_H + +#include "pipe/p_compiler.h" + +#if defined __cplusplus +extern "C" { +#endif + +#define MAX_LABELS 1024 + +#define NUM_CHANNELS 4 /* R,G,B,A */ +#define QUAD_SIZE 4 /* 4 pixel/quad */ + +/** + * Registers may be treated as float, signed int or unsigned int. + */ +union tgsi_exec_channel +{ + float f[QUAD_SIZE]; + int i[QUAD_SIZE]; + unsigned u[QUAD_SIZE]; +}; + +/** + * A vector[RGBA] of channels[4 pixels] + */ +struct tgsi_exec_vector +{ + union tgsi_exec_channel xyzw[NUM_CHANNELS]; +}; + +/** + * For fragment programs, information for computing fragment input + * values from plane equation of the triangle/line. + */ +struct tgsi_interp_coef +{ + float a0[NUM_CHANNELS]; /* in an xyzw layout */ + float dadx[NUM_CHANNELS]; + float dady[NUM_CHANNELS]; +}; + + +struct softpipe_tile_cache; /**< Opaque to TGSI */ + +/** + * Information for sampling textures, which must be implemented + * by code outside the TGSI executor. + */ +struct tgsi_sampler +{ + const struct pipe_sampler_state *state; + struct pipe_texture *texture; + /** Get samples for four fragments in a quad */ + void (*get_samples)(struct tgsi_sampler *sampler, + const float s[QUAD_SIZE], + const float t[QUAD_SIZE], + const float p[QUAD_SIZE], + float lodbias, + float rgba[NUM_CHANNELS][QUAD_SIZE]); + void *pipe; /*XXX temporary*/ + struct softpipe_tile_cache *cache; +}; + +/** + * For branching/calling subroutines. + */ +struct tgsi_exec_labels +{ + unsigned labels[MAX_LABELS][2]; + unsigned count; +}; + + +#define TGSI_EXEC_NUM_TEMPS 128 +#define TGSI_EXEC_NUM_TEMP_EXTRAS 6 +#define TGSI_EXEC_NUM_IMMEDIATES 256 + +/* + * Locations of various utility registers (_I = Index, _C = Channel) + */ +#define TGSI_EXEC_TEMP_00000000_I (TGSI_EXEC_NUM_TEMPS + 0) +#define TGSI_EXEC_TEMP_00000000_C 0 + +#define TGSI_EXEC_TEMP_7FFFFFFF_I (TGSI_EXEC_NUM_TEMPS + 0) +#define TGSI_EXEC_TEMP_7FFFFFFF_C 1 + +#define TGSI_EXEC_TEMP_80000000_I (TGSI_EXEC_NUM_TEMPS + 0) +#define TGSI_EXEC_TEMP_80000000_C 2 + +#define TGSI_EXEC_TEMP_FFFFFFFF_I (TGSI_EXEC_NUM_TEMPS + 0) +#define TGSI_EXEC_TEMP_FFFFFFFF_C 3 + +#define TGSI_EXEC_TEMP_ONE_I (TGSI_EXEC_NUM_TEMPS + 1) +#define TGSI_EXEC_TEMP_ONE_C 0 + +#define TGSI_EXEC_TEMP_TWO_I (TGSI_EXEC_NUM_TEMPS + 1) +#define TGSI_EXEC_TEMP_TWO_C 1 + +#define TGSI_EXEC_TEMP_128_I (TGSI_EXEC_NUM_TEMPS + 1) +#define TGSI_EXEC_TEMP_128_C 2 + +#define TGSI_EXEC_TEMP_MINUS_128_I (TGSI_EXEC_NUM_TEMPS + 1) +#define TGSI_EXEC_TEMP_MINUS_128_C 3 + +#define TGSI_EXEC_TEMP_KILMASK_I (TGSI_EXEC_NUM_TEMPS + 2) +#define TGSI_EXEC_TEMP_KILMASK_C 0 + +#define TGSI_EXEC_TEMP_OUTPUT_I (TGSI_EXEC_NUM_TEMPS + 2) +#define TGSI_EXEC_TEMP_OUTPUT_C 1 + +#define TGSI_EXEC_TEMP_PRIMITIVE_I (TGSI_EXEC_NUM_TEMPS + 2) +#define TGSI_EXEC_TEMP_PRIMITIVE_C 2 + +#define TGSI_EXEC_TEMP_THREE_I (TGSI_EXEC_NUM_TEMPS + 2) +#define TGSI_EXEC_TEMP_THREE_C 3 + +#define TGSI_EXEC_TEMP_HALF_I (TGSI_EXEC_NUM_TEMPS + 3) +#define TGSI_EXEC_TEMP_HALF_C 0 + +#define TGSI_EXEC_TEMP_R0 (TGSI_EXEC_NUM_TEMPS + 4) + +#define TGSI_EXEC_TEMP_ADDR (TGSI_EXEC_NUM_TEMPS + 5) + + +#define TGSI_EXEC_MAX_COND_NESTING 20 +#define TGSI_EXEC_MAX_LOOP_NESTING 20 +#define TGSI_EXEC_MAX_CALL_NESTING 20 + +/** + * Run-time virtual machine state for executing TGSI shader. + */ +struct tgsi_exec_machine +{ + /* Total = program temporaries + internal temporaries + * + 1 padding to align to 16 bytes + */ + struct tgsi_exec_vector _Temps[TGSI_EXEC_NUM_TEMPS + + TGSI_EXEC_NUM_TEMP_EXTRAS + 1]; + + /* + * This will point to _Temps after aligning to 16B boundary. + */ + struct tgsi_exec_vector *Temps; + struct tgsi_exec_vector *Addrs; + + struct tgsi_sampler *Samplers; + + float Imms[TGSI_EXEC_NUM_IMMEDIATES][4]; + unsigned ImmLimit; + const float (*Consts)[4]; + struct tgsi_exec_vector *Inputs; + struct tgsi_exec_vector *Outputs; + const struct tgsi_token *Tokens; + unsigned Processor; + + /* GEOMETRY processor only. */ + unsigned *Primitives; + + /* FRAGMENT processor only. */ + const struct tgsi_interp_coef *InterpCoefs; + struct tgsi_exec_vector QuadPos; + + /* Conditional execution masks */ + uint CondMask; /**< For IF/ELSE/ENDIF */ + uint LoopMask; /**< For BGNLOOP/ENDLOOP */ + uint ContMask; /**< For loop CONT statements */ + uint FuncMask; /**< For function calls */ + uint ExecMask; /**< = CondMask & LoopMask */ + + /** Condition mask stack (for nested conditionals) */ + uint CondStack[TGSI_EXEC_MAX_COND_NESTING]; + int CondStackTop; + + /** Loop mask stack (for nested loops) */ + uint LoopStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int LoopStackTop; + + /** Loop continue mask stack (see comments in tgsi_exec.c) */ + uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING]; + int ContStackTop; + + /** Function execution mask stack (for executing subroutine code) */ + uint FuncStack[TGSI_EXEC_MAX_CALL_NESTING]; + int FuncStackTop; + + /** Function call stack for saving/restoring the program counter */ + uint CallStack[TGSI_EXEC_MAX_CALL_NESTING]; + int CallStackTop; + + struct tgsi_full_instruction *Instructions; + uint NumInstructions; + + struct tgsi_full_declaration *Declarations; + uint NumDeclarations; + + struct tgsi_exec_labels Labels; +}; + +void +tgsi_exec_machine_init( + struct tgsi_exec_machine *mach ); + + +void +tgsi_exec_machine_bind_shader( + struct tgsi_exec_machine *mach, + const struct tgsi_token *tokens, + uint numSamplers, + struct tgsi_sampler *samplers); + +uint +tgsi_exec_machine_run( + struct tgsi_exec_machine *mach ); + + +void +tgsi_exec_machine_free_data(struct tgsi_exec_machine *mach); + + +#if defined __cplusplus +} /* extern "C" */ +#endif + +#endif /* TGSI_EXEC_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_iterate.c b/src/gallium/auxiliary/tgsi/tgsi_iterate.c new file mode 100644 index 0000000000..5371a88b96 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_iterate.c @@ -0,0 +1,85 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "tgsi_iterate.h" + +boolean +tgsi_iterate_shader( + const struct tgsi_token *tokens, + struct tgsi_iterate_context *ctx ) +{ + struct tgsi_parse_context parse; + + if (tgsi_parse_init( &parse, tokens ) != TGSI_PARSE_OK) + return FALSE; + + ctx->processor = parse.FullHeader.Processor; + ctx->version = parse.FullVersion.Version; + + if (ctx->prolog) + if (!ctx->prolog( ctx )) + goto fail; + + while (!tgsi_parse_end_of_tokens( &parse )) { + tgsi_parse_token( &parse ); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + if (ctx->iterate_instruction) + if (!ctx->iterate_instruction( ctx, &parse.FullToken.FullInstruction )) + goto fail; + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + if (ctx->iterate_declaration) + if (!ctx->iterate_declaration( ctx, &parse.FullToken.FullDeclaration )) + goto fail; + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + if (ctx->iterate_immediate) + if (!ctx->iterate_immediate( ctx, &parse.FullToken.FullImmediate )) + goto fail; + break; + + default: + assert( 0 ); + } + } + + if (ctx->epilog) + if (!ctx->epilog( ctx )) + goto fail; + + tgsi_parse_free( &parse ); + return TRUE; + +fail: + tgsi_parse_free( &parse ); + return FALSE; +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_iterate.h b/src/gallium/auxiliary/tgsi/tgsi_iterate.h new file mode 100644 index 0000000000..ec7b85bf63 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_iterate.h @@ -0,0 +1,76 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_ITERATE_H +#define TGSI_ITERATE_H + +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" + +#if defined __cplusplus +extern "C" { +#endif + +struct tgsi_iterate_context +{ + boolean + (* prolog)( + struct tgsi_iterate_context *ctx ); + + boolean + (* iterate_instruction)( + struct tgsi_iterate_context *ctx, + struct tgsi_full_instruction *inst ); + + boolean + (* iterate_declaration)( + struct tgsi_iterate_context *ctx, + struct tgsi_full_declaration *decl ); + + boolean + (* iterate_immediate)( + struct tgsi_iterate_context *ctx, + struct tgsi_full_immediate *imm ); + + boolean + (* epilog)( + struct tgsi_iterate_context *ctx ); + + struct tgsi_processor processor; + struct tgsi_version version; +}; + +boolean +tgsi_iterate_shader( + const struct tgsi_token *tokens, + struct tgsi_iterate_context *ctx ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_ITERATE_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c new file mode 100644 index 0000000000..d16f0cdcad --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c @@ -0,0 +1,332 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ) +{ + full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; +} + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ) +{ + if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { + FREE( (void *) full_token->FullImmediate.u.Pointer ); + } +} + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ) +{ + ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; + if( ctx->FullVersion.Version.MajorVersion > 1 ) { + return TGSI_PARSE_ERROR; + } + + ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; + if( ctx->FullHeader.Header.HeaderSize >= 2 ) { + ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; + } + else { + ctx->FullHeader.Processor = tgsi_default_processor(); + } + + ctx->Tokens = tokens; + ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; + + tgsi_full_token_init( &ctx->FullToken ); + + return TGSI_PARSE_OK; +} + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ) +{ + tgsi_full_token_free( &ctx->FullToken ); +} + +boolean +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ) +{ + return ctx->Position >= + 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; +} + +static void +next_token( + struct tgsi_parse_context *ctx, + void *token ) +{ + assert( !tgsi_parse_end_of_tokens( ctx ) ); + + *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; +} + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ) +{ + struct tgsi_token token; + unsigned i; + + tgsi_full_token_free( &ctx->FullToken ); + tgsi_full_token_init( &ctx->FullToken ); + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; + + *decl = tgsi_default_full_declaration(); + decl->Declaration = *(struct tgsi_declaration *) &token; + + next_token( ctx, &decl->DeclarationRange ); + + if( decl->Declaration.Semantic ) { + next_token( ctx, &decl->Semantic ); + } + + break; + } + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; + + *imm = tgsi_default_full_immediate(); + imm->Immediate = *(struct tgsi_immediate *) &token; + + assert( !imm->Immediate.Extended ); + + switch (imm->Immediate.DataType) { + case TGSI_IMM_FLOAT32: + imm->u.Pointer = MALLOC( + sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); + for( i = 0; i < imm->Immediate.Size - 1; i++ ) { + next_token( ctx, (struct tgsi_immediate_float32 *) &imm->u.ImmediateFloat32[i] ); + } + break; + + default: + assert( 0 ); + } + + break; + } + + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; + unsigned extended; + + *inst = tgsi_default_full_instruction(); + inst->Instruction = *(struct tgsi_instruction *) &token; + + extended = inst->Instruction.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_INSTRUCTION_EXT_TYPE_NV: + inst->InstructionExtNv = + *(struct tgsi_instruction_ext_nv *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_LABEL: + inst->InstructionExtLabel = + *(struct tgsi_instruction_ext_label *) &token; + break; + + case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: + inst->InstructionExtTexture = + *(struct tgsi_instruction_ext_texture *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); + assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); + + extended = inst->FullDstRegisters[i].DstRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: + inst->FullDstRegisters[i].DstRegisterExtConcode = + *(struct tgsi_dst_register_ext_concode *) &token; + break; + + case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: + inst->FullDstRegisters[i].DstRegisterExtModulate = + *(struct tgsi_dst_register_ext_modulate *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + } + + assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); + + for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { + unsigned extended; + + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); + + extended = inst->FullSrcRegisters[i].SrcRegister.Extended; + + while( extended ) { + struct tgsi_src_register_ext token; + + next_token( ctx, &token ); + + switch( token.Type ) { + case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: + inst->FullSrcRegisters[i].SrcRegisterExtSwz = + *(struct tgsi_src_register_ext_swz *) &token; + break; + + case TGSI_SRC_REGISTER_EXT_TYPE_MOD: + inst->FullSrcRegisters[i].SrcRegisterExtMod = + *(struct tgsi_src_register_ext_mod *) &token; + break; + + default: + assert( 0 ); + } + + extended = token.Extended; + } + + if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + + if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); + + /* + * No support for multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); + + if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { + next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); + + /* + * No support for indirect or multi-dimensional addressing. + */ + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); + assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); + } + } + } + + break; + } + + default: + assert( 0 ); + } +} + + +unsigned +tgsi_num_tokens(const struct tgsi_token *tokens) +{ + struct tgsi_parse_context ctx; + if (tgsi_parse_init(&ctx, tokens) == TGSI_PARSE_OK) { + unsigned len = (ctx.FullHeader.Header.HeaderSize + + ctx.FullHeader.Header.BodySize + + 1); + return len; + } + return 0; +} + + +/** + * Make a new copy of a token array. + */ +struct tgsi_token * +tgsi_dup_tokens(const struct tgsi_token *tokens) +{ + unsigned n = tgsi_num_tokens(tokens); + unsigned bytes = n * sizeof(struct tgsi_token); + struct tgsi_token *new_tokens = (struct tgsi_token *) MALLOC(bytes); + if (new_tokens) + memcpy(new_tokens, tokens, bytes); + return new_tokens; +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.h b/src/gallium/auxiliary/tgsi/tgsi_parse.h new file mode 100644 index 0000000000..054350712d --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.h @@ -0,0 +1,151 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_PARSE_H +#define TGSI_PARSE_H + +#include "pipe/p_shader_tokens.h" + +#if defined __cplusplus +extern "C" { +#endif + +struct tgsi_full_version +{ + struct tgsi_version Version; +}; + +struct tgsi_full_header +{ + struct tgsi_header Header; + struct tgsi_processor Processor; +}; + +struct tgsi_full_dst_register +{ + struct tgsi_dst_register DstRegister; + struct tgsi_dst_register_ext_concode DstRegisterExtConcode; + struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; +}; + +struct tgsi_full_src_register +{ + struct tgsi_src_register SrcRegister; + struct tgsi_src_register_ext_swz SrcRegisterExtSwz; + struct tgsi_src_register_ext_mod SrcRegisterExtMod; + struct tgsi_src_register SrcRegisterInd; + struct tgsi_dimension SrcRegisterDim; + struct tgsi_src_register SrcRegisterDimInd; +}; + +struct tgsi_full_declaration +{ + struct tgsi_declaration Declaration; + struct tgsi_declaration_range DeclarationRange; + struct tgsi_declaration_semantic Semantic; +}; + +struct tgsi_full_immediate +{ + struct tgsi_immediate Immediate; + union + { + const void *Pointer; + const struct tgsi_immediate_float32 *ImmediateFloat32; + } u; +}; + +#define TGSI_FULL_MAX_DST_REGISTERS 2 +#define TGSI_FULL_MAX_SRC_REGISTERS 4 /* TXD has 4 */ + +struct tgsi_full_instruction +{ + struct tgsi_instruction Instruction; + struct tgsi_instruction_ext_nv InstructionExtNv; + struct tgsi_instruction_ext_label InstructionExtLabel; + struct tgsi_instruction_ext_texture InstructionExtTexture; + struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; + struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; +}; + +union tgsi_full_token +{ + struct tgsi_token Token; + struct tgsi_full_declaration FullDeclaration; + struct tgsi_full_immediate FullImmediate; + struct tgsi_full_instruction FullInstruction; +}; + +void +tgsi_full_token_init( + union tgsi_full_token *full_token ); + +void +tgsi_full_token_free( + union tgsi_full_token *full_token ); + +struct tgsi_parse_context +{ + const struct tgsi_token *Tokens; + unsigned Position; + struct tgsi_full_version FullVersion; + struct tgsi_full_header FullHeader; + union tgsi_full_token FullToken; +}; + +#define TGSI_PARSE_OK 0 +#define TGSI_PARSE_ERROR 1 + +unsigned +tgsi_parse_init( + struct tgsi_parse_context *ctx, + const struct tgsi_token *tokens ); + +void +tgsi_parse_free( + struct tgsi_parse_context *ctx ); + +boolean +tgsi_parse_end_of_tokens( + struct tgsi_parse_context *ctx ); + +void +tgsi_parse_token( + struct tgsi_parse_context *ctx ); + +unsigned +tgsi_num_tokens(const struct tgsi_token *tokens); + +struct tgsi_token * +tgsi_dup_tokens(const struct tgsi_token *tokens); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_PARSE_H */ + diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c new file mode 100644 index 0000000000..2e3ec96b5b --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -0,0 +1,341 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "tgsi_sanity.h" +#include "tgsi_iterate.h" + +#define MAX_REGISTERS 256 + +typedef uint reg_flag; + +#define BITS_IN_REG_FLAG (sizeof( reg_flag ) * 8) + +struct sanity_check_ctx +{ + struct tgsi_iterate_context iter; + + reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG]; + reg_flag regs_used[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG]; + boolean regs_ind_used[TGSI_FILE_COUNT]; + uint num_imms; + uint num_instructions; + uint index_of_END; + + uint errors; + uint warnings; +}; + +static void +report_error( + struct sanity_check_ctx *ctx, + const char *format, + ... ) +{ + va_list args; + + debug_printf( "Error : " ); + va_start( args, format ); + _debug_vprintf( format, args ); + va_end( args ); + debug_printf( "\n" ); + ctx->errors++; +} + +static void +report_warning( + struct sanity_check_ctx *ctx, + const char *format, + ... ) +{ + va_list args; + + debug_printf( "Warning: " ); + va_start( args, format ); + _debug_vprintf( format, args ); + va_end( args ); + debug_printf( "\n" ); + ctx->warnings++; +} + +static boolean +check_file_name( + struct sanity_check_ctx *ctx, + uint file ) +{ + if (file <= TGSI_FILE_NULL || file >= TGSI_FILE_COUNT) { + report_error( ctx, "Invalid register file name" ); + return FALSE; + } + return TRUE; +} + +static boolean +is_register_declared( + struct sanity_check_ctx *ctx, + uint file, + int index ) +{ + assert( index >= 0 && index < MAX_REGISTERS ); + + return (ctx->regs_decl[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; +} + +static boolean +is_any_register_declared( + struct sanity_check_ctx *ctx, + uint file ) +{ + uint i; + + for (i = 0; i < MAX_REGISTERS / BITS_IN_REG_FLAG; i++) + if (ctx->regs_decl[file][i]) + return TRUE; + return FALSE; +} + +static boolean +is_register_used( + struct sanity_check_ctx *ctx, + uint file, + int index ) +{ + assert( index < MAX_REGISTERS ); + + return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; +} + +static const char *file_names[] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + +static boolean +check_register_usage( + struct sanity_check_ctx *ctx, + uint file, + int index, + const char *name, + boolean indirect_access ) +{ + if (!check_file_name( ctx, file )) + return FALSE; + if (indirect_access) { + if (!is_any_register_declared( ctx, file )) + report_error( ctx, "%s: Undeclared %s register", file_names[file], name ); + ctx->regs_ind_used[file] = TRUE; + } + else { + if (!is_register_declared( ctx, file, index )) + report_error( ctx, "%s[%d]: Undeclared %s register", file_names[file], index, name ); + ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG)); + } + return TRUE; +} + +static boolean +iter_instruction( + struct tgsi_iterate_context *iter, + struct tgsi_full_instruction *inst ) +{ + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; + uint i; + + /* There must be no other instructions after END. + */ + if (ctx->index_of_END != ~0) { + report_error( ctx, "Unexpected instruction after END" ); + } + else if (inst->Instruction.Opcode == TGSI_OPCODE_END) { + ctx->index_of_END = ctx->num_instructions; + } + + /* Check destination and source registers' validity. + * Mark the registers as used. + */ + for (i = 0; i < inst->Instruction.NumDstRegs; i++) { + check_register_usage( + ctx, + inst->FullDstRegisters[i].DstRegister.File, + inst->FullDstRegisters[i].DstRegister.Index, + "destination", + FALSE ); + } + for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { + check_register_usage( + ctx, + inst->FullSrcRegisters[i].SrcRegister.File, + inst->FullSrcRegisters[i].SrcRegister.Index, + "source", + (boolean)inst->FullSrcRegisters[i].SrcRegister.Indirect ); + if (inst->FullSrcRegisters[i].SrcRegister.Indirect) { + uint file; + int index; + + file = inst->FullSrcRegisters[i].SrcRegisterInd.File; + index = inst->FullSrcRegisters[i].SrcRegisterInd.Index; + check_register_usage( + ctx, + file, + index, + "indirect", + FALSE ); + if (file != TGSI_FILE_ADDRESS || index != 0) + report_warning( ctx, "Indirect register not ADDR[0]" ); + } + } + + ctx->num_instructions++; + + return TRUE; +} + +static boolean +iter_declaration( + struct tgsi_iterate_context *iter, + struct tgsi_full_declaration *decl ) +{ + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; + uint file; + uint i; + + /* No declarations allowed after the first instruction. + */ + if (ctx->num_instructions > 0) + report_error( ctx, "Instruction expected but declaration found" ); + + /* Check registers' validity. + * Mark the registers as declared. + */ + file = decl->Declaration.File; + if (!check_file_name( ctx, file )) + return TRUE; + for (i = decl->DeclarationRange.First; i <= decl->DeclarationRange.Last; i++) { + if (is_register_declared( ctx, file, i )) + report_error( ctx, "The same register declared twice" ); + ctx->regs_decl[file][i / BITS_IN_REG_FLAG] |= (1 << (i % BITS_IN_REG_FLAG)); + } + + return TRUE; +} + +static boolean +iter_immediate( + struct tgsi_iterate_context *iter, + struct tgsi_full_immediate *imm ) +{ + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; + + assert( ctx->num_imms < MAX_REGISTERS ); + + /* No immediates allowed after the first instruction. + */ + if (ctx->num_instructions > 0) + report_error( ctx, "Instruction expected but immediate found" ); + + /* Mark the register as declared. + */ + ctx->regs_decl[TGSI_FILE_IMMEDIATE][ctx->num_imms / BITS_IN_REG_FLAG] |= (1 << (ctx->num_imms % BITS_IN_REG_FLAG)); + ctx->num_imms++; + + /* Check data type validity. + */ + if (imm->Immediate.DataType != TGSI_IMM_FLOAT32) { + report_error( ctx, "Invalid immediate data type" ); + return TRUE; + } + + return TRUE; +} + +static boolean +epilog( + struct tgsi_iterate_context *iter ) +{ + struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; + uint file; + + /* There must be an END instruction at the end. + */ + if (ctx->index_of_END == ~0 || ctx->index_of_END != ctx->num_instructions - 1) { + report_error( ctx, "Expected END at end of instruction sequence" ); + } + + /* Check if all declared registers were used. + */ + for (file = TGSI_FILE_NULL; file < TGSI_FILE_COUNT; file++) { + uint i; + + for (i = 0; i < MAX_REGISTERS; i++) { + if (is_register_declared( ctx, file, i ) && !is_register_used( ctx, file, i ) && !ctx->regs_ind_used[file]) { + report_warning( ctx, "Register never used" ); + } + } + } + + /* Print totals, if any. + */ + if (ctx->errors || ctx->warnings) + debug_printf( "\n%u errors, %u warnings", ctx->errors, ctx->warnings ); + + return TRUE; +} + +boolean +tgsi_sanity_check( + struct tgsi_token *tokens ) +{ + struct sanity_check_ctx ctx; + + ctx.iter.prolog = NULL; + ctx.iter.iterate_instruction = iter_instruction; + ctx.iter.iterate_declaration = iter_declaration; + ctx.iter.iterate_immediate = iter_immediate; + ctx.iter.epilog = epilog; + + memset( ctx.regs_decl, 0, sizeof( ctx.regs_decl ) ); + memset( ctx.regs_used, 0, sizeof( ctx.regs_used ) ); + memset( ctx.regs_ind_used, 0, sizeof( ctx.regs_ind_used ) ); + ctx.num_imms = 0; + ctx.num_instructions = 0; + ctx.index_of_END = ~0; + + ctx.errors = 0; + ctx.warnings = 0; + + if (!tgsi_iterate_shader( tokens, &ctx.iter )) + return FALSE; + + return ctx.errors == 0; +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.h b/src/gallium/auxiliary/tgsi/tgsi_sanity.h new file mode 100644 index 0000000000..ca45e94c7a --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.h @@ -0,0 +1,49 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_SANITY_H +#define TGSI_SANITY_H + +#include "pipe/p_shader_tokens.h" + +#if defined __cplusplus +extern "C" { +#endif + +/* Check the given token stream for errors and common mistakes. + * Diagnostic messages are printed out to the debug output. + * Returns TRUE if there are no errors, even though there could be some warnings. + */ +boolean +tgsi_sanity_check( + struct tgsi_token *tokens ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_SANITY_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c new file mode 100644 index 0000000000..59bcf10b53 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c @@ -0,0 +1,226 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI program scan utility. + * Used to determine which registers and instructions are used by a shader. + * + * Authors: Brian Paul + */ + + +#include "tgsi_scan.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_build.h" + +#include "pipe/p_util.h" + + + +/** + */ +void +tgsi_scan_shader(const struct tgsi_token *tokens, + struct tgsi_shader_info *info) +{ + uint procType, i; + struct tgsi_parse_context parse; + + memset(info, 0, sizeof(*info)); + for (i = 0; i < TGSI_FILE_COUNT; i++) + info->file_max[i] = -1; + + /** + ** Setup to begin parsing input shader + **/ + if (tgsi_parse_init( &parse, tokens ) != TGSI_PARSE_OK) { + debug_printf("tgsi_parse_init() failed in tgsi_scan_shader()!\n"); + return; + } + procType = parse.FullHeader.Processor.Processor; + assert(procType == TGSI_PROCESSOR_FRAGMENT || + procType == TGSI_PROCESSOR_VERTEX || + procType == TGSI_PROCESSOR_GEOMETRY); + + + /** + ** Loop over incoming program tokens/instructions + */ + while( !tgsi_parse_end_of_tokens( &parse ) ) { + + info->num_tokens++; + + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *fullinst + = &parse.FullToken.FullInstruction; + + assert(fullinst->Instruction.Opcode < TGSI_OPCODE_LAST); + info->opcode_count[fullinst->Instruction.Opcode]++; + } + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *fulldecl + = &parse.FullToken.FullDeclaration; + uint file = fulldecl->Declaration.File; + uint i; + for (i = fulldecl->DeclarationRange.First; + i <= fulldecl->DeclarationRange.Last; + i++) { + + /* only first 32 regs will appear in this bitfield */ + info->file_mask[file] |= (1 << i); + info->file_count[file]++; + info->file_max[file] = MAX2(info->file_max[file], (int)i); + + if (file == TGSI_FILE_INPUT) { + info->input_semantic_name[i] = (ubyte)fulldecl->Semantic.SemanticName; + info->input_semantic_index[i] = (ubyte)fulldecl->Semantic.SemanticIndex; + info->num_inputs++; + } + + if (file == TGSI_FILE_OUTPUT) { + info->output_semantic_name[i] = (ubyte)fulldecl->Semantic.SemanticName; + info->output_semantic_index[i] = (ubyte)fulldecl->Semantic.SemanticIndex; + info->num_outputs++; + } + + /* special case */ + if (procType == TGSI_PROCESSOR_FRAGMENT && + file == TGSI_FILE_OUTPUT && + fulldecl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) { + info->writes_z = TRUE; + } + } + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + info->immediate_count++; + break; + + default: + assert( 0 ); + } + } + + assert( info->file_max[TGSI_FILE_INPUT] + 1 == info->num_inputs ); + assert( info->file_max[TGSI_FILE_OUTPUT] + 1 == info->num_outputs ); + + info->uses_kill = (info->opcode_count[TGSI_OPCODE_KIL] || + info->opcode_count[TGSI_OPCODE_KILP]); + + tgsi_parse_free (&parse); +} + + + +/** + * Check if the given shader is a "passthrough" shader consisting of only + * MOV instructions of the form: MOV OUT[n], IN[n] + * + */ +boolean +tgsi_is_passthrough_shader(const struct tgsi_token *tokens) +{ + struct tgsi_parse_context parse; + + /** + ** Setup to begin parsing input shader + **/ + if (tgsi_parse_init(&parse, tokens) != TGSI_PARSE_OK) { + debug_printf("tgsi_parse_init() failed in tgsi_is_passthrough_shader()!\n"); + return FALSE; + } + + /** + ** Loop over incoming program tokens/instructions + */ + while (!tgsi_parse_end_of_tokens(&parse)) { + + tgsi_parse_token(&parse); + + switch (parse.FullToken.Token.Type) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *fullinst = + &parse.FullToken.FullInstruction; + const struct tgsi_full_src_register *src = + &fullinst->FullSrcRegisters[0]; + const struct tgsi_full_dst_register *dst = + &fullinst->FullDstRegisters[0]; + + /* Do a whole bunch of checks for a simple move */ + if (fullinst->Instruction.Opcode != TGSI_OPCODE_MOV || + src->SrcRegister.File != TGSI_FILE_INPUT || + dst->DstRegister.File != TGSI_FILE_OUTPUT || + src->SrcRegister.Index != dst->DstRegister.Index || + + src->SrcRegister.Negate || + src->SrcRegisterExtMod.Negate || + src->SrcRegisterExtMod.Absolute || + src->SrcRegisterExtMod.Scale2X || + src->SrcRegisterExtMod.Bias || + src->SrcRegisterExtMod.Complement || + + src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || + src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || + src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || + src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W || + + src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || + src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || + src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || + src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W || + + dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW) + { + tgsi_parse_free(&parse); + return FALSE; + } + } + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + /* fall-through */ + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* fall-through */ + default: + ; /* no-op */ + } + } + + tgsi_parse_free(&parse); + + /* if we get here, it's a pass-through shader */ + return TRUE; +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.h b/src/gallium/auxiliary/tgsi/tgsi_scan.h new file mode 100644 index 0000000000..5cb6efb343 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.h @@ -0,0 +1,74 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_SCAN_H +#define TGSI_SCAN_H + + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" +#include "pipe/p_shader_tokens.h" + + +/** + * Shader summary info + */ +struct tgsi_shader_info +{ + uint num_tokens; + + /* XXX eventually remove the corresponding fields from pipe_shader_state: */ + ubyte num_inputs; + ubyte num_outputs; + ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ + ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ + ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; + + uint file_mask[TGSI_FILE_COUNT]; /**< bitmask of declared registers */ + uint file_count[TGSI_FILE_COUNT]; /**< number of declared registers */ + int file_max[TGSI_FILE_COUNT]; /**< highest index of declared registers */ + + uint immediate_count; /**< number of immediates declared */ + + uint opcode_count[TGSI_OPCODE_LAST]; /**< opcode histogram */ + + boolean writes_z; /**< does fragment shader write Z value? */ + boolean uses_kill; /**< KIL or KILP instruction used? */ +}; + + +extern void +tgsi_scan_shader(const struct tgsi_token *tokens, + struct tgsi_shader_info *info); + + +extern boolean +tgsi_is_passthrough_shader(const struct tgsi_token *tokens); + + +#endif /* TGSI_SCAN_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c new file mode 100644 index 0000000000..0cb1f11ef2 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -0,0 +1,2275 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" +#include "tgsi_exec.h" +#include "tgsi_sse2.h" + +#include "rtasm/rtasm_x86sse.h" + +#ifdef PIPE_ARCH_X86 + +/* for 1/sqrt() + * + * This costs about 100fps (close to 10%) in gears: + */ +#define HIGH_PRECISION 1 + + +#define FOR_EACH_CHANNEL( CHAN )\ + for( CHAN = 0; CHAN < 4; CHAN++ ) + +#define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ + ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) + +#define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ + if( IS_DST0_CHANNEL_ENABLED( INST, CHAN )) + +#define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\ + FOR_EACH_CHANNEL( CHAN )\ + IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN ) + +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + +#define TEMP_R0 TGSI_EXEC_TEMP_R0 + +/** + * X86 utility functions. + */ + +static struct x86_reg +make_xmm( + unsigned xmm ) +{ + return x86_make_reg( + file_XMM, + (enum x86_reg_name) xmm ); +} + +/** + * X86 register mapping helpers. + */ + +static struct x86_reg +get_const_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_CX ); +} + +static struct x86_reg +get_input_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_AX ); +} + +static struct x86_reg +get_output_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_DX ); +} + +static struct x86_reg +get_temp_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_BX ); +} + +static struct x86_reg +get_coef_base( void ) +{ + return get_output_base(); +} + +static struct x86_reg +get_immediate_base( void ) +{ + return x86_make_reg( + file_REG32, + reg_DI ); +} + + +/** + * Data access helpers. + */ + + +static struct x86_reg +get_immediate( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_immediate_base(), + (vec * 4 + chan) * 4 ); +} + +static struct x86_reg +get_const( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_const_base(), + (vec * 4 + chan) * 4 ); +} + +static struct x86_reg +get_input( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_input_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_output( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_output_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_temp( + unsigned vec, + unsigned chan ) +{ + return x86_make_disp( + get_temp_base(), + (vec * 4 + chan) * 16 ); +} + +static struct x86_reg +get_coef( + unsigned vec, + unsigned chan, + unsigned member ) +{ + return x86_make_disp( + get_coef_base(), + ((vec * 3 + member) * 4 + chan) * 4 ); +} + + +static void +emit_ret( + struct x86_function *func ) +{ + x86_ret( func ); +} + + +/** + * Data fetch helpers. + */ + +/** + * Copy a shader constant to xmm register + * \param xmm the destination xmm register + * \param vec the src const buffer index + * \param chan src channel to fetch (X, Y, Z or W) + */ +static void +emit_const( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movss( + func, + make_xmm( xmm ), + get_const( vec, chan ) ); + sse_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + +static void +emit_immediate( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movss( + func, + make_xmm( xmm ), + get_immediate( vec, chan ) ); + sse_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + + +/** + * Copy a shader input to xmm register + * \param xmm the destination xmm register + * \param vec the src input attrib + * \param chan src channel to fetch (X, Y, Z or W) + */ +static void +emit_inputf( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movups( + func, + make_xmm( xmm ), + get_input( vec, chan ) ); +} + +/** + * Store an xmm register to a shader output + * \param xmm the source xmm register + * \param vec the dest output attrib + * \param chan src dest channel to store (X, Y, Z or W) + */ +static void +emit_output( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movups( + func, + get_output( vec, chan ), + make_xmm( xmm ) ); +} + +/** + * Copy a shader temporary to xmm register + * \param xmm the destination xmm register + * \param vec the src temp register + * \param chan src channel to fetch (X, Y, Z or W) + */ +static void +emit_tempf( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movaps( + func, + make_xmm( xmm ), + get_temp( vec, chan ) ); +} + +/** + * Load an xmm register with an input attrib coefficient (a0, dadx or dady) + * \param xmm the destination xmm register + * \param vec the src input/attribute coefficient index + * \param chan src channel to fetch (X, Y, Z or W) + * \param member 0=a0, 1=dadx, 2=dady + */ +static void +emit_coef( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan, + unsigned member ) +{ + sse_movss( + func, + make_xmm( xmm ), + get_coef( vec, chan, member ) ); + sse_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); +} + +/** + * Data store helpers. + */ + +static void +emit_inputs( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movups( + func, + get_input( vec, chan ), + make_xmm( xmm ) ); +} + +static void +emit_temps( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + sse_movaps( + func, + get_temp( vec, chan ), + make_xmm( xmm ) ); +} + +static void +emit_addrs( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_temps( + func, + xmm, + vec + TGSI_EXEC_NUM_TEMPS, + chan ); +} + +/** + * Coefficent fetch helpers. + */ + +static void +emit_coef_a0( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 0 ); +} + +static void +emit_coef_dadx( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 1 ); +} + +static void +emit_coef_dady( + struct x86_function *func, + unsigned xmm, + unsigned vec, + unsigned chan ) +{ + emit_coef( + func, + xmm, + vec, + chan, + 2 ); +} + +/** + * Function call helpers. + */ + +static void +emit_push_gp( + struct x86_function *func ) +{ + x86_push( + func, + x86_make_reg( file_REG32, reg_AX) ); + x86_push( + func, + x86_make_reg( file_REG32, reg_CX) ); + x86_push( + func, + x86_make_reg( file_REG32, reg_DX) ); +} + +static void +x86_pop_gp( + struct x86_function *func ) +{ + /* Restore GP registers in a reverse order. + */ + x86_pop( + func, + x86_make_reg( file_REG32, reg_DX) ); + x86_pop( + func, + x86_make_reg( file_REG32, reg_CX) ); + x86_pop( + func, + x86_make_reg( file_REG32, reg_AX) ); +} + +static void +emit_func_call_dst( + struct x86_function *func, + unsigned xmm_dst, + void (PIPE_CDECL *code)() ) +{ + sse_movaps( + func, + get_temp( TEMP_R0, 0 ), + make_xmm( xmm_dst ) ); + + emit_push_gp( + func ); + + { + struct x86_reg ecx = x86_make_reg( file_REG32, reg_CX ); + + x86_lea( + func, + ecx, + get_temp( TEMP_R0, 0 ) ); + + x86_push( func, ecx ); + x86_mov_reg_imm( func, ecx, (unsigned long) code ); + x86_call( func, ecx ); + x86_pop(func, ecx ); + } + + + x86_pop_gp( + func ); + + sse_movaps( + func, + make_xmm( xmm_dst ), + get_temp( TEMP_R0, 0 ) ); +} + +static void +emit_func_call_dst_src( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src, + void (PIPE_CDECL *code)() ) +{ + sse_movaps( + func, + get_temp( TEMP_R0, 1 ), + make_xmm( xmm_src ) ); + + emit_func_call_dst( + func, + xmm_dst, + code ); +} + +/** + * Low-level instruction translators. + */ + +static void +emit_abs( + struct x86_function *func, + unsigned xmm ) +{ + sse_andps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_7FFFFFFF_I, + TGSI_EXEC_TEMP_7FFFFFFF_C ) ); +} + +static void +emit_add( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + sse_addps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void PIPE_CDECL +cos4f( + float *store ) +{ + const unsigned X = 0; + + store[X + 0] = cosf( store[X + 0] ); + store[X + 1] = cosf( store[X + 1] ); + store[X + 2] = cosf( store[X + 2] ); + store[X + 3] = cosf( store[X + 3] ); +} + +static void +emit_cos( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + cos4f ); +} + +static void PIPE_CDECL +ex24f( + float *store ) +{ + const unsigned X = 0; + + store[X + 0] = powf( 2.0f, store[X + 0] ); + store[X + 1] = powf( 2.0f, store[X + 1] ); + store[X + 2] = powf( 2.0f, store[X + 2] ); + store[X + 3] = powf( 2.0f, store[X + 3] ); +} + +static void +emit_ex2( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + ex24f ); +} + +static void +emit_f2it( + struct x86_function *func, + unsigned xmm ) +{ + sse2_cvttps2dq( + func, + make_xmm( xmm ), + make_xmm( xmm ) ); +} + +static void PIPE_CDECL +flr4f( + float *store ) +{ + const unsigned X = 0; + + store[X + 0] = floorf( store[X + 0] ); + store[X + 1] = floorf( store[X + 1] ); + store[X + 2] = floorf( store[X + 2] ); + store[X + 3] = floorf( store[X + 3] ); +} + +static void +emit_flr( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + flr4f ); +} + +static void PIPE_CDECL +frc4f( + float *store ) +{ + const unsigned X = 0; + + store[X + 0] -= floorf( store[X + 0] ); + store[X + 1] -= floorf( store[X + 1] ); + store[X + 2] -= floorf( store[X + 2] ); + store[X + 3] -= floorf( store[X + 3] ); +} + +static void +emit_frc( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + frc4f ); +} + +static void PIPE_CDECL +lg24f( + float *store ) +{ + const unsigned X = 0; + + store[X + 0] = LOG2( store[X + 0] ); + store[X + 1] = LOG2( store[X + 1] ); + store[X + 2] = LOG2( store[X + 2] ); + store[X + 3] = LOG2( store[X + 3] ); +} + +static void +emit_lg2( + struct x86_function *func, + unsigned xmm_dst ) +{ + emit_func_call_dst( + func, + xmm_dst, + lg24f ); +} + +static void +emit_MOV( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + sse_movups( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_mul (struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src) +{ + sse_mulps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_neg( + struct x86_function *func, + unsigned xmm ) +{ + sse_xorps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); +} + +static void PIPE_CDECL +pow4f( + float *store ) +{ + const unsigned X = 0; + + store[X + 0] = powf( store[X + 0], store[X + 4] ); + store[X + 1] = powf( store[X + 1], store[X + 5] ); + store[X + 2] = powf( store[X + 2], store[X + 6] ); + store[X + 3] = powf( store[X + 3], store[X + 7] ); +} + +static void +emit_pow( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + emit_func_call_dst_src( + func, + xmm_dst, + xmm_src, + pow4f ); +} + +static void +emit_rcp ( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + /* On Intel CPUs at least, this is only accurate to 12 bits -- not + * good enough. Need to either emit a proper divide or use the + * iterative technique described below in emit_rsqrt(). + */ + sse2_rcpps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +static void +emit_rsqrt( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ +#if HIGH_PRECISION + /* Although rsqrtps() and rcpps() are low precision on some/all SSE + * implementations, it is possible to improve its precision at + * fairly low cost, using a newton/raphson step, as below: + * + * x1 = 2 * rcpps(a) - a * rcpps(a) * rcpps(a) + * x1 = 0.5 * rsqrtps(a) * [3.0 - (a * rsqrtps(a))* rsqrtps(a)] + * + * See: http://softwarecommunity.intel.com/articles/eng/1818.htm + */ + { + struct x86_reg dst = make_xmm( xmm_dst ); + struct x86_reg src = make_xmm( xmm_src ); + struct x86_reg tmp0 = make_xmm( 2 ); + struct x86_reg tmp1 = make_xmm( 3 ); + + assert( xmm_dst != xmm_src ); + assert( xmm_dst != 2 && xmm_dst != 3 ); + assert( xmm_src != 2 && xmm_src != 3 ); + + sse_movaps( func, dst, get_temp( TGSI_EXEC_TEMP_HALF_I, TGSI_EXEC_TEMP_HALF_C ) ); + sse_movaps( func, tmp0, get_temp( TGSI_EXEC_TEMP_THREE_I, TGSI_EXEC_TEMP_THREE_C ) ); + sse_rsqrtps( func, tmp1, src ); + sse_mulps( func, src, tmp1 ); + sse_mulps( func, dst, tmp1 ); + sse_mulps( func, src, tmp1 ); + sse_subps( func, tmp0, src ); + sse_mulps( func, dst, tmp0 ); + } +#else + /* On Intel CPUs at least, this is only accurate to 12 bits -- not + * good enough. + */ + sse_rsqrtps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +#endif +} + +static void +emit_setsign( + struct x86_function *func, + unsigned xmm ) +{ + sse_orps( + func, + make_xmm( xmm ), + get_temp( + TGSI_EXEC_TEMP_80000000_I, + TGSI_EXEC_TEMP_80000000_C ) ); +} + +static void PIPE_CDECL +sin4f( + float *store ) +{ + const unsigned X = 0; + + store[X + 0] = sinf( store[X + 0] ); + store[X + 1] = sinf( store[X + 1] ); + store[X + 2] = sinf( store[X + 2] ); + store[X + 3] = sinf( store[X + 3] ); +} + +static void +emit_sin (struct x86_function *func, + unsigned xmm_dst) +{ + emit_func_call_dst( + func, + xmm_dst, + sin4f ); +} + +static void +emit_sub( + struct x86_function *func, + unsigned xmm_dst, + unsigned xmm_src ) +{ + sse_subps( + func, + make_xmm( xmm_dst ), + make_xmm( xmm_src ) ); +} + +/** + * Register fetch. + */ + +static void +emit_fetch( + struct x86_function *func, + unsigned xmm, + const struct tgsi_full_src_register *reg, + const unsigned chan_index ) +{ + unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); + + switch( swizzle ) { + case TGSI_EXTSWIZZLE_X: + case TGSI_EXTSWIZZLE_Y: + case TGSI_EXTSWIZZLE_Z: + case TGSI_EXTSWIZZLE_W: + switch( reg->SrcRegister.File ) { + case TGSI_FILE_CONSTANT: + emit_const( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + case TGSI_FILE_IMMEDIATE: + emit_immediate( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + case TGSI_FILE_INPUT: + emit_inputf( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + case TGSI_FILE_TEMPORARY: + emit_tempf( + func, + xmm, + reg->SrcRegister.Index, + swizzle ); + break; + + default: + assert( 0 ); + } + break; + + case TGSI_EXTSWIZZLE_ZERO: + emit_tempf( + func, + xmm, + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ); + break; + + case TGSI_EXTSWIZZLE_ONE: + emit_tempf( + func, + xmm, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + break; + + default: + assert( 0 ); + } + + switch( tgsi_util_get_full_src_register_sign_mode( reg, chan_index ) ) { + case TGSI_UTIL_SIGN_CLEAR: + emit_abs( func, xmm ); + break; + + case TGSI_UTIL_SIGN_SET: + emit_setsign( func, xmm ); + break; + + case TGSI_UTIL_SIGN_TOGGLE: + emit_neg( func, xmm ); + break; + + case TGSI_UTIL_SIGN_KEEP: + break; + } +} + +#define FETCH( FUNC, INST, XMM, INDEX, CHAN )\ + emit_fetch( FUNC, XMM, &(INST).FullSrcRegisters[INDEX], CHAN ) + +/** + * Register store. + */ + +static void +emit_store( + struct x86_function *func, + unsigned xmm, + const struct tgsi_full_dst_register *reg, + const struct tgsi_full_instruction *inst, + unsigned chan_index ) +{ + switch( reg->DstRegister.File ) { + case TGSI_FILE_OUTPUT: + emit_output( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; + + case TGSI_FILE_TEMPORARY: + emit_temps( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; + + case TGSI_FILE_ADDRESS: + emit_addrs( + func, + xmm, + reg->DstRegister.Index, + chan_index ); + break; + + default: + assert( 0 ); + } + + switch( inst->Instruction.Saturate ) { + case TGSI_SAT_NONE: + break; + + case TGSI_SAT_ZERO_ONE: + /* assert( 0 ); */ + break; + + case TGSI_SAT_MINUS_PLUS_ONE: + assert( 0 ); + break; + } +} + +#define STORE( FUNC, INST, XMM, INDEX, CHAN )\ + emit_store( FUNC, XMM, &(INST).FullDstRegisters[INDEX], &(INST), CHAN ) + +/** + * High-level instruction translators. + */ + +static void +emit_kil( + struct x86_function *func, + const struct tgsi_full_src_register *reg ) +{ + unsigned uniquemask; + unsigned registers[4]; + unsigned nextregister = 0; + unsigned firstchan = ~0; + unsigned chan_index; + + /* This mask stores component bits that were already tested. Note that + * we test if the value is less than zero, so 1.0 and 0.0 need not to be + * tested. */ + uniquemask = (1 << TGSI_EXTSWIZZLE_ZERO) | (1 << TGSI_EXTSWIZZLE_ONE); + + FOR_EACH_CHANNEL( chan_index ) { + unsigned swizzle; + + /* unswizzle channel */ + swizzle = tgsi_util_get_full_src_register_extswizzle( + reg, + chan_index ); + + /* check if the component has not been already tested */ + if( !(uniquemask & (1 << swizzle)) ) { + uniquemask |= 1 << swizzle; + + /* allocate register */ + registers[chan_index] = nextregister; + emit_fetch( + func, + nextregister, + reg, + chan_index ); + nextregister++; + + /* mark the first channel used */ + if( firstchan == ~0 ) { + firstchan = chan_index; + } + } + } + + x86_push( + func, + x86_make_reg( file_REG32, reg_AX ) ); + x86_push( + func, + x86_make_reg( file_REG32, reg_DX ) ); + + FOR_EACH_CHANNEL( chan_index ) { + if( uniquemask & (1 << chan_index) ) { + sse_cmpps( + func, + make_xmm( registers[chan_index] ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ), + cc_LessThan ); + + if( chan_index == firstchan ) { + sse_pmovmskb( + func, + x86_make_reg( file_REG32, reg_AX ), + make_xmm( registers[chan_index] ) ); + } + else { + sse_pmovmskb( + func, + x86_make_reg( file_REG32, reg_DX ), + make_xmm( registers[chan_index] ) ); + x86_or( + func, + x86_make_reg( file_REG32, reg_AX ), + x86_make_reg( file_REG32, reg_DX ) ); + } + } + } + + x86_or( + func, + get_temp( + TGSI_EXEC_TEMP_KILMASK_I, + TGSI_EXEC_TEMP_KILMASK_C ), + x86_make_reg( file_REG32, reg_AX ) ); + + x86_pop( + func, + x86_make_reg( file_REG32, reg_DX ) ); + x86_pop( + func, + x86_make_reg( file_REG32, reg_AX ) ); +} + +static void +emit_setcc( + struct x86_function *func, + struct tgsi_full_instruction *inst, + enum sse_cc cc ) +{ + unsigned chan_index; + + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + sse_cmpps( + func, + make_xmm( 0 ), + make_xmm( 1 ), + cc ); + sse_andps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ) ); + STORE( func, *inst, 0, 0, chan_index ); + } +} + +static void +emit_cmp( + struct x86_function *func, + struct tgsi_full_instruction *inst ) +{ + unsigned chan_index; + + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + sse_cmpps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ), + cc_LessThan ); + sse_andps( + func, + make_xmm( 1 ), + make_xmm( 0 ) ); + sse_andnps( + func, + make_xmm( 0 ), + make_xmm( 2 ) ); + sse_orps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } +} + +static int +emit_instruction( + struct x86_function *func, + struct tgsi_full_instruction *inst ) +{ + unsigned chan_index; + + switch( inst->Instruction.Opcode ) { + case TGSI_OPCODE_ARL: +#if 0 + /* XXX this isn't working properly (see glean vertProg1 test) */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_f2it( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } +#else + return 0; +#endif + break; + + case TGSI_OPCODE_MOV: + case TGSI_OPCODE_SWZ: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LIT: + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C); + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ) { + STORE( func, *inst, 0, 0, CHAN_X ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) ) { + STORE( func, *inst, 0, 0, CHAN_W ); + } + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( func, *inst, 0, 0, CHAN_X ); + sse_maxps( + func, + make_xmm( 0 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ) ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + /* XMM[1] = SrcReg[0].yyyy */ + FETCH( func, *inst, 1, 0, CHAN_Y ); + /* XMM[1] = max(XMM[1], 0) */ + sse_maxps( + func, + make_xmm( 1 ), + get_temp( + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ) ); + /* XMM[2] = SrcReg[0].wwww */ + FETCH( func, *inst, 2, 0, CHAN_W ); + /* XMM[2] = min(XMM[2], 128.0) */ + sse_minps( + func, + make_xmm( 2 ), + get_temp( + TGSI_EXEC_TEMP_128_I, + TGSI_EXEC_TEMP_128_C ) ); + /* XMM[2] = max(XMM[2], -128.0) */ + sse_maxps( + func, + make_xmm( 2 ), + get_temp( + TGSI_EXEC_TEMP_MINUS_128_I, + TGSI_EXEC_TEMP_MINUS_128_C ) ); + emit_pow( func, 1, 2 ); + FETCH( func, *inst, 0, 0, CHAN_X ); + sse_xorps( + func, + make_xmm( 2 ), + make_xmm( 2 ) ); + sse_cmpps( + func, + make_xmm( 2 ), + make_xmm( 0 ), + cc_LessThanEqual ); + sse_andps( + func, + make_xmm( 2 ), + make_xmm( 1 ) ); + STORE( func, *inst, 2, 0, CHAN_Z ); + } + } + break; + + case TGSI_OPCODE_RCP: + /* TGSI_OPCODE_RECIP */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_rcp( func, 0, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_RSQ: + /* TGSI_OPCODE_RECIPSQRT */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_rsqrt( func, 1, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 1, 0, chan_index ); + } + break; + + case TGSI_OPCODE_EXP: + return 0; + break; + + case TGSI_OPCODE_LOG: + return 0; + break; + + case TGSI_OPCODE_MUL: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_mul( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_ADD: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_add( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP3: + /* TGSI_OPCODE_DOT3 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DP4: + /* TGSI_OPCODE_DOT4 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul(func, 1, 2 ); + emit_add(func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_W ); + FETCH( func, *inst, 2, 1, CHAN_W ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DST: + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + FETCH( func, *inst, 1, 1, CHAN_Y ); + emit_mul( func, 0, 1 ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + FETCH( func, *inst, 0, 0, CHAN_Z ); + STORE( func, *inst, 0, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + FETCH( func, *inst, 0, 1, CHAN_W ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MIN: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + sse_minps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_MAX: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + sse_maxps( + func, + make_xmm( 0 ), + make_xmm( 1 ) ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLT: + /* TGSI_OPCODE_SETLT */ + emit_setcc( func, inst, cc_LessThan ); + break; + + case TGSI_OPCODE_SGE: + /* TGSI_OPCODE_SETGE */ + emit_setcc( func, inst, cc_NotLessThan ); + break; + + case TGSI_OPCODE_MAD: + /* TGSI_OPCODE_MADD */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_mul( func, 0, 1 ); + emit_add( func, 0, 2 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SUB: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + emit_sub( func, 0, 1 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LERP: + /* TGSI_OPCODE_LRP */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + FETCH( func, *inst, 1, 1, chan_index ); + FETCH( func, *inst, 2, 2, chan_index ); + emit_sub( func, 1, 2 ); + emit_mul( func, 0, 1 ); + emit_add( func, 0, 2 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CND: + return 0; + break; + + case TGSI_OPCODE_CND0: + return 0; + break; + + case TGSI_OPCODE_DOT2ADD: + /* TGSI_OPCODE_DP2A */ + return 0; + break; + + case TGSI_OPCODE_INDEX: + return 0; + break; + + case TGSI_OPCODE_NEGATE: + return 0; + break; + + case TGSI_OPCODE_FRAC: + /* TGSI_OPCODE_FRC */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_frc( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CLAMP: + return 0; + break; + + case TGSI_OPCODE_FLOOR: + /* TGSI_OPCODE_FLR */ + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_flr( func, 0 ); + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_ROUND: + return 0; + break; + + case TGSI_OPCODE_EXPBASE2: + /* TGSI_OPCODE_EX2 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_ex2( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_LOGBASE2: + /* TGSI_OPCODE_LG2 */ + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_lg2( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_POWER: + /* TGSI_OPCODE_POW */ + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_pow( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_CROSSPRODUCT: + /* TGSI_OPCODE_XPD */ + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) ) { + FETCH( func, *inst, 1, 1, CHAN_Z ); + FETCH( func, *inst, 3, 0, CHAN_Z ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 0, 0, CHAN_Y ); + FETCH( func, *inst, 4, 1, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + emit_MOV( func, 2, 0 ); + emit_mul( func, 2, 1 ); + emit_MOV( func, 5, 3 ); + emit_mul( func, 5, 4 ); + emit_sub( func, 2, 5 ); + STORE( func, *inst, 2, 0, CHAN_X ); + } + if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) ) { + FETCH( func, *inst, 2, 1, CHAN_X ); + FETCH( func, *inst, 5, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + emit_mul( func, 3, 2 ); + emit_mul( func, 1, 5 ); + emit_sub( func, 3, 1 ); + STORE( func, *inst, 3, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + emit_mul( func, 5, 4 ); + emit_mul( func, 0, 2 ); + emit_sub( func, 5, 0 ); + STORE( func, *inst, 5, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_MULTIPLYMATRIX: + return 0; + break; + + case TGSI_OPCODE_ABS: + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + FETCH( func, *inst, 0, 0, chan_index ); + emit_abs( func, 0) ; + + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_RCC: + return 0; + break; + + case TGSI_OPCODE_DPH: + FETCH( func, *inst, 0, 0, CHAN_X ); + FETCH( func, *inst, 1, 1, CHAN_X ); + emit_mul( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Y ); + FETCH( func, *inst, 2, 1, CHAN_Y ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 0, CHAN_Z ); + FETCH( func, *inst, 2, 1, CHAN_Z ); + emit_mul( func, 1, 2 ); + emit_add( func, 0, 1 ); + FETCH( func, *inst, 1, 1, CHAN_W ); + emit_add( func, 0, 1 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_COS: + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_cos( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_DDX: + return 0; + break; + + case TGSI_OPCODE_DDY: + return 0; + break; + + case TGSI_OPCODE_KIL: + emit_kil( func, &inst->FullSrcRegisters[0] ); + break; + + case TGSI_OPCODE_PK2H: + return 0; + break; + + case TGSI_OPCODE_PK2US: + return 0; + break; + + case TGSI_OPCODE_PK4B: + return 0; + break; + + case TGSI_OPCODE_PK4UB: + return 0; + break; + + case TGSI_OPCODE_RFL: + return 0; + break; + + case TGSI_OPCODE_SEQ: + return 0; + break; + + case TGSI_OPCODE_SFL: + return 0; + break; + + case TGSI_OPCODE_SGT: + return 0; + break; + + case TGSI_OPCODE_SIN: + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_sin( func, 0 ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + break; + + case TGSI_OPCODE_SLE: + return 0; + break; + + case TGSI_OPCODE_SNE: + return 0; + break; + + case TGSI_OPCODE_STR: + return 0; + break; + + case TGSI_OPCODE_TEX: + if (0) { + /* Disable dummy texture code: + */ + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { + STORE( func, *inst, 0, 0, chan_index ); + } + } + else { + return 0; + } + break; + + case TGSI_OPCODE_TXD: + return 0; + break; + + case TGSI_OPCODE_UP2H: + return 0; + break; + + case TGSI_OPCODE_UP2US: + return 0; + break; + + case TGSI_OPCODE_UP4B: + return 0; + break; + + case TGSI_OPCODE_UP4UB: + return 0; + break; + + case TGSI_OPCODE_X2D: + return 0; + break; + + case TGSI_OPCODE_ARA: + return 0; + break; + + case TGSI_OPCODE_ARR: + return 0; + break; + + case TGSI_OPCODE_BRA: + return 0; + break; + + case TGSI_OPCODE_CAL: + return 0; + break; + + case TGSI_OPCODE_RET: + emit_ret( func ); + break; + + case TGSI_OPCODE_END: + break; + + case TGSI_OPCODE_SSG: + return 0; + break; + + case TGSI_OPCODE_CMP: + emit_cmp (func, inst); + break; + + case TGSI_OPCODE_SCS: + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) { + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_cos( func, 0 ); + STORE( func, *inst, 0, 0, CHAN_X ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_sin( func, 0 ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_00000000_I, + TGSI_EXEC_TEMP_00000000_C ); + STORE( func, *inst, 0, 0, CHAN_Z ); + } + IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W ) { + emit_tempf( + func, + 0, + TGSI_EXEC_TEMP_ONE_I, + TGSI_EXEC_TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } + break; + + case TGSI_OPCODE_TXB: + return 0; + break; + + case TGSI_OPCODE_NRM: + return 0; + break; + + case TGSI_OPCODE_DIV: + return 0; + break; + + case TGSI_OPCODE_DP2: + return 0; + break; + + case TGSI_OPCODE_TXL: + return 0; + break; + + case TGSI_OPCODE_BRK: + return 0; + break; + + case TGSI_OPCODE_IF: + return 0; + break; + + case TGSI_OPCODE_LOOP: + return 0; + break; + + case TGSI_OPCODE_REP: + return 0; + break; + + case TGSI_OPCODE_ELSE: + return 0; + break; + + case TGSI_OPCODE_ENDIF: + return 0; + break; + + case TGSI_OPCODE_ENDLOOP: + return 0; + break; + + case TGSI_OPCODE_ENDREP: + return 0; + break; + + case TGSI_OPCODE_PUSHA: + return 0; + break; + + case TGSI_OPCODE_POPA: + return 0; + break; + + case TGSI_OPCODE_CEIL: + return 0; + break; + + case TGSI_OPCODE_I2F: + return 0; + break; + + case TGSI_OPCODE_NOT: + return 0; + break; + + case TGSI_OPCODE_TRUNC: + return 0; + break; + + case TGSI_OPCODE_SHL: + return 0; + break; + + case TGSI_OPCODE_SHR: + return 0; + break; + + case TGSI_OPCODE_AND: + return 0; + break; + + case TGSI_OPCODE_OR: + return 0; + break; + + case TGSI_OPCODE_MOD: + return 0; + break; + + case TGSI_OPCODE_XOR: + return 0; + break; + + case TGSI_OPCODE_SAD: + return 0; + break; + + case TGSI_OPCODE_TXF: + return 0; + break; + + case TGSI_OPCODE_TXQ: + return 0; + break; + + case TGSI_OPCODE_CONT: + return 0; + break; + + case TGSI_OPCODE_EMIT: + return 0; + break; + + case TGSI_OPCODE_ENDPRIM: + return 0; + break; + + default: + return 0; + } + + return 1; +} + +static void +emit_declaration( + struct x86_function *func, + struct tgsi_full_declaration *decl ) +{ + if( decl->Declaration.File == TGSI_FILE_INPUT ) { + unsigned first, last, mask; + unsigned i, j; + + first = decl->DeclarationRange.First; + last = decl->DeclarationRange.Last; + mask = decl->Declaration.UsageMask; + + for( i = first; i <= last; i++ ) { + for( j = 0; j < NUM_CHANNELS; j++ ) { + if( mask & (1 << j) ) { + switch( decl->Declaration.Interpolate ) { + case TGSI_INTERPOLATE_CONSTANT: + emit_coef_a0( func, 0, i, j ); + emit_inputs( func, 0, i, j ); + break; + + case TGSI_INTERPOLATE_LINEAR: + emit_tempf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_coef_dadx( func, 1, i, j ); + emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_coef_dady( func, 3, i, j ); + emit_mul( func, 0, 1 ); /* x * dadx */ + emit_coef_a0( func, 4, i, j ); + emit_mul( func, 2, 3 ); /* y * dady */ + emit_add( func, 0, 4 ); /* x * dadx + a0 */ + emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ + emit_inputs( func, 0, i, j ); + break; + + case TGSI_INTERPOLATE_PERSPECTIVE: + emit_tempf( func, 0, 0, TGSI_SWIZZLE_X ); + emit_coef_dadx( func, 1, i, j ); + emit_tempf( func, 2, 0, TGSI_SWIZZLE_Y ); + emit_coef_dady( func, 3, i, j ); + emit_mul( func, 0, 1 ); /* x * dadx */ + emit_tempf( func, 4, 0, TGSI_SWIZZLE_W ); + emit_coef_a0( func, 5, i, j ); + emit_rcp( func, 4, 4 ); /* 1.0 / w */ + emit_mul( func, 2, 3 ); /* y * dady */ + emit_add( func, 0, 5 ); /* x * dadx + a0 */ + emit_add( func, 0, 2 ); /* x * dadx + y * dady + a0 */ + emit_mul( func, 0, 4 ); /* (x * dadx + y * dady + a0) / w */ + emit_inputs( func, 0, i, j ); + break; + + default: + assert( 0 ); + break; + } + } + } + } + } +} + +static void aos_to_soa( struct x86_function *func, + uint arg_aos, + uint arg_soa, + uint arg_num, + uint arg_stride ) +{ + struct x86_reg soa_input = x86_make_reg( file_REG32, reg_AX ); + struct x86_reg aos_input = x86_make_reg( file_REG32, reg_BX ); + struct x86_reg num_inputs = x86_make_reg( file_REG32, reg_CX ); + struct x86_reg stride = x86_make_reg( file_REG32, reg_DX ); + int inner_loop; + + + /* Save EBX */ + x86_push( func, x86_make_reg( file_REG32, reg_BX ) ); + + x86_mov( func, aos_input, x86_fn_arg( func, arg_aos ) ); + x86_mov( func, soa_input, x86_fn_arg( func, arg_soa ) ); + x86_mov( func, num_inputs, x86_fn_arg( func, arg_num ) ); + x86_mov( func, stride, x86_fn_arg( func, arg_stride ) ); + + /* do */ + inner_loop = x86_get_label( func ); + { + x86_push( func, aos_input ); + sse_movlps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); + sse_movlps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); + x86_add( func, aos_input, stride ); + sse_movhps( func, make_xmm( 0 ), x86_make_disp( aos_input, 0 ) ); + sse_movhps( func, make_xmm( 3 ), x86_make_disp( aos_input, 8 ) ); + x86_add( func, aos_input, stride ); + sse_movlps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); + sse_movlps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); + x86_add( func, aos_input, stride ); + sse_movhps( func, make_xmm( 1 ), x86_make_disp( aos_input, 0 ) ); + sse_movhps( func, make_xmm( 4 ), x86_make_disp( aos_input, 8 ) ); + x86_pop( func, aos_input ); + + sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); + sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); + sse_shufps( func, make_xmm( 0 ), make_xmm( 1 ), 0x88 ); + sse_shufps( func, make_xmm( 2 ), make_xmm( 1 ), 0xdd ); + sse_shufps( func, make_xmm( 3 ), make_xmm( 4 ), 0x88 ); + sse_shufps( func, make_xmm( 5 ), make_xmm( 4 ), 0xdd ); + + sse_movups( func, x86_make_disp( soa_input, 0 ), make_xmm( 0 ) ); + sse_movups( func, x86_make_disp( soa_input, 16 ), make_xmm( 2 ) ); + sse_movups( func, x86_make_disp( soa_input, 32 ), make_xmm( 3 ) ); + sse_movups( func, x86_make_disp( soa_input, 48 ), make_xmm( 5 ) ); + + /* Advance to next input */ + x86_lea( func, aos_input, x86_make_disp(aos_input, 16) ); + x86_lea( func, soa_input, x86_make_disp(soa_input, 64) ); + } + /* while --num_inputs */ + x86_dec( func, num_inputs ); + x86_jcc( func, cc_NE, inner_loop ); + + /* Restore EBX */ + x86_pop( func, aos_input ); +} + +static void soa_to_aos( struct x86_function *func, uint aos, uint soa, uint num, uint stride ) +{ + struct x86_reg soa_output; + struct x86_reg aos_output; + struct x86_reg num_outputs; + struct x86_reg temp; + int inner_loop; + + soa_output = x86_make_reg( file_REG32, reg_AX ); + aos_output = x86_make_reg( file_REG32, reg_BX ); + num_outputs = x86_make_reg( file_REG32, reg_CX ); + temp = x86_make_reg( file_REG32, reg_DX ); + + /* Save EBX */ + x86_push( func, aos_output ); + + x86_mov( func, soa_output, x86_fn_arg( func, soa ) ); + x86_mov( func, aos_output, x86_fn_arg( func, aos ) ); + x86_mov( func, num_outputs, x86_fn_arg( func, num ) ); + + /* do */ + inner_loop = x86_get_label( func ); + { + sse_movups( func, make_xmm( 0 ), x86_make_disp( soa_output, 0 ) ); + sse_movups( func, make_xmm( 1 ), x86_make_disp( soa_output, 16 ) ); + sse_movups( func, make_xmm( 3 ), x86_make_disp( soa_output, 32 ) ); + sse_movups( func, make_xmm( 4 ), x86_make_disp( soa_output, 48 ) ); + + sse_movaps( func, make_xmm( 2 ), make_xmm( 0 ) ); + sse_movaps( func, make_xmm( 5 ), make_xmm( 3 ) ); + sse_unpcklps( func, make_xmm( 0 ), make_xmm( 1 ) ); + sse_unpckhps( func, make_xmm( 2 ), make_xmm( 1 ) ); + sse_unpcklps( func, make_xmm( 3 ), make_xmm( 4 ) ); + sse_unpckhps( func, make_xmm( 5 ), make_xmm( 4 ) ); + + x86_mov( func, temp, x86_fn_arg( func, stride ) ); + x86_push( func, aos_output ); + sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); + sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); + x86_add( func, aos_output, temp ); + sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 0 ) ); + sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 3 ) ); + x86_add( func, aos_output, temp ); + sse_movlps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); + sse_movlps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); + x86_add( func, aos_output, temp ); + sse_movhps( func, x86_make_disp( aos_output, 0 ), make_xmm( 2 ) ); + sse_movhps( func, x86_make_disp( aos_output, 8 ), make_xmm( 5 ) ); + x86_pop( func, aos_output ); + + /* Advance to next output */ + x86_lea( func, aos_output, x86_make_disp(aos_output, 16) ); + x86_lea( func, soa_output, x86_make_disp(soa_output, 64) ); + } + /* while --num_outputs */ + x86_dec( func, num_outputs ); + x86_jcc( func, cc_NE, inner_loop ); + + /* Restore EBX */ + x86_pop( func, aos_output ); +} + +/** + * Translate a TGSI vertex/fragment shader to SSE2 code. + * Slightly different things are done for vertex vs. fragment shaders. + * + * Note that fragment shaders are responsible for interpolating shader + * inputs. Because on x86 we have only 4 GP registers, and here we + * have 5 shader arguments (input, output, const, temp and coef), the + * code is split into two phases -- DECLARATION and INSTRUCTION phase. + * GP register holding the output argument is aliased with the coeff + * argument, as outputs are not needed in the DECLARATION phase. + * + * \param tokens the TGSI input shader + * \param func the output SSE code/function + * \param immediates buffer to place immediates, later passed to SSE func + * \param return 1 for success, 0 if translation failed + */ +unsigned +tgsi_emit_sse2( + const struct tgsi_token *tokens, + struct x86_function *func, + float (*immediates)[4], + boolean do_swizzles ) +{ + struct tgsi_parse_context parse; + boolean instruction_phase = FALSE; + unsigned ok = 1; + uint num_immediates = 0; + + func->csr = func->store; + + tgsi_parse_init( &parse, tokens ); + + /* Can't just use EDI, EBX without save/restoring them: + */ + x86_push( + func, + get_immediate_base() ); + + x86_push( + func, + get_temp_base() ); + + + /* + * Different function args for vertex/fragment shaders: + */ + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { + /* DECLARATION phase, do not load output argument. */ + x86_mov( + func, + get_input_base(), + x86_fn_arg( func, 1 ) ); + /* skipping outputs argument here */ + x86_mov( + func, + get_const_base(), + x86_fn_arg( func, 3 ) ); + x86_mov( + func, + get_temp_base(), + x86_fn_arg( func, 4 ) ); + x86_mov( + func, + get_coef_base(), + x86_fn_arg( func, 5 ) ); + x86_mov( + func, + get_immediate_base(), + x86_fn_arg( func, 6 ) ); + } + else { + assert(parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX); + + if (do_swizzles) + aos_to_soa( func, + 6, /* aos_input */ + 1, /* machine->input */ + 7, /* num_inputs */ + 8 ); /* input_stride */ + + x86_mov( + func, + get_input_base(), + x86_fn_arg( func, 1 ) ); + x86_mov( + func, + get_output_base(), + x86_fn_arg( func, 2 ) ); + x86_mov( + func, + get_const_base(), + x86_fn_arg( func, 3 ) ); + x86_mov( + func, + get_temp_base(), + x86_fn_arg( func, 4 ) ); + x86_mov( + func, + get_immediate_base(), + x86_fn_arg( func, 5 ) ); + } + + while( !tgsi_parse_end_of_tokens( &parse ) && ok ) { + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_DECLARATION: + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { + emit_declaration( + func, + &parse.FullToken.FullDeclaration ); + } + break; + + case TGSI_TOKEN_TYPE_INSTRUCTION: + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_FRAGMENT) { + if( !instruction_phase ) { + /* INSTRUCTION phase, overwrite coeff with output. */ + instruction_phase = TRUE; + x86_mov( + func, + get_output_base(), + x86_fn_arg( func, 2 ) ); + } + } + + ok = emit_instruction( + func, + &parse.FullToken.FullInstruction ); + + if (!ok) { + debug_printf("failed to translate tgsi opcode %d to SSE (%s)\n", + parse.FullToken.FullInstruction.Instruction.Opcode, + parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX ? + "vertex shader" : "fragment shader"); + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + /* simply copy the immediate values into the next immediates[] slot */ + { + const uint size = parse.FullToken.FullImmediate.Immediate.Size - 1; + uint i; + assert(size <= 4); + assert(num_immediates < TGSI_EXEC_NUM_IMMEDIATES); + for( i = 0; i < size; i++ ) { + immediates[num_immediates][i] = + parse.FullToken.FullImmediate.u.ImmediateFloat32[i].Float; + } +#if 0 + debug_printf("SSE FS immediate[%d] = %f %f %f %f\n", + num_immediates, + immediates[num_immediates][0], + immediates[num_immediates][1], + immediates[num_immediates][2], + immediates[num_immediates][3]); +#endif + num_immediates++; + } + break; + + default: + ok = 0; + assert( 0 ); + } + } + + if (parse.FullHeader.Processor.Processor == TGSI_PROCESSOR_VERTEX) { + if (do_swizzles) + soa_to_aos( func, 9, 2, 10, 11 ); + } + + /* Can't just use EBX, EDI without save/restoring them: + */ + x86_pop( + func, + get_temp_base() ); + + x86_pop( + func, + get_immediate_base() ); + + emit_ret( func ); + + tgsi_parse_free( &parse ); + + return ok; +} + +#endif /* PIPE_ARCH_X86 */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.h b/src/gallium/auxiliary/tgsi/tgsi_sse2.h new file mode 100644 index 0000000000..af838b2a25 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.h @@ -0,0 +1,49 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_SSE2_H +#define TGSI_SSE2_H + +#if defined __cplusplus +extern "C" { +#endif + +struct tgsi_token; +struct x86_function; + +unsigned +tgsi_emit_sse2( + const struct tgsi_token *tokens, + struct x86_function *function, + float (*immediates)[4], + boolean do_swizzles ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_SSE2_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c new file mode 100644 index 0000000000..35cb3055bb --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_text.c @@ -0,0 +1,1221 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "tgsi_text.h" +#include "tgsi_build.h" +#include "tgsi_parse.h" +#include "tgsi_sanity.h" +#include "tgsi_util.h" + +static boolean is_alpha_underscore( const char *cur ) +{ + return + (*cur >= 'a' && *cur <= 'z') || + (*cur >= 'A' && *cur <= 'Z') || + *cur == '_'; +} + +static boolean is_digit( const char *cur ) +{ + return *cur >= '0' && *cur <= '9'; +} + +static boolean is_digit_alpha_underscore( const char *cur ) +{ + return is_digit( cur ) || is_alpha_underscore( cur ); +} + +static boolean str_match_no_case( const char **pcur, const char *str ) +{ + const char *cur = *pcur; + + while (*str != '\0' && *str == toupper( *cur )) { + str++; + cur++; + } + if (*str == '\0') { + *pcur = cur; + return TRUE; + } + return FALSE; +} + +/* Eat zero or more whitespaces. + */ +static void eat_opt_white( const char **pcur ) +{ + while (**pcur == ' ' || **pcur == '\t' || **pcur == '\n') + (*pcur)++; +} + +/* Eat one or more whitespaces. + * Return TRUE if at least one whitespace eaten. + */ +static boolean eat_white( const char **pcur ) +{ + const char *cur = *pcur; + + eat_opt_white( pcur ); + return *pcur > cur; +} + +/* Parse unsigned integer. + * No checks for overflow. + */ +static boolean parse_uint( const char **pcur, uint *val ) +{ + const char *cur = *pcur; + + if (is_digit( cur )) { + *val = *cur++ - '0'; + while (is_digit( cur )) + *val = *val * 10 + *cur++ - '0'; + *pcur = cur; + return TRUE; + } + return FALSE; +} + +/* Parse floating point. + */ +static boolean parse_float( const char **pcur, float *val ) +{ + const char *cur = *pcur; + boolean integral_part = FALSE; + boolean fractional_part = FALSE; + + *val = (float) atof( cur ); + + if (*cur == '-' || *cur == '+') + cur++; + if (is_digit( cur )) { + cur++; + integral_part = TRUE; + while (is_digit( cur )) + cur++; + } + if (*cur == '.') { + cur++; + if (is_digit( cur )) { + cur++; + fractional_part = TRUE; + while (is_digit( cur )) + cur++; + } + } + if (!integral_part && !fractional_part) + return FALSE; + if (toupper( *cur ) == 'E') { + cur++; + if (*cur == '-' || *cur == '+') + cur++; + if (is_digit( cur )) { + cur++; + while (is_digit( cur )) + cur++; + } + else + return FALSE; + } + *pcur = cur; + return TRUE; +} + +struct translate_ctx +{ + const char *text; + const char *cur; + struct tgsi_token *tokens; + struct tgsi_token *tokens_cur; + struct tgsi_token *tokens_end; + struct tgsi_header *header; +}; + +static void report_error( struct translate_ctx *ctx, const char *msg ) +{ + debug_printf( "\nError: %s", msg ); +} + +/* Parse shader header. + * Return TRUE for one of the following headers. + * FRAG1.1 + * GEOM1.1 + * VERT1.1 + */ +static boolean parse_header( struct translate_ctx *ctx ) +{ + uint processor; + + if (str_match_no_case( &ctx->cur, "FRAG1.1" )) + processor = TGSI_PROCESSOR_FRAGMENT; + else if (str_match_no_case( &ctx->cur, "VERT1.1" )) + processor = TGSI_PROCESSOR_VERTEX; + else if (str_match_no_case( &ctx->cur, "GEOM1.1" )) + processor = TGSI_PROCESSOR_GEOMETRY; + else { + report_error( ctx, "Unknown header" ); + return FALSE; + } + + if (ctx->tokens_cur >= ctx->tokens_end) + return FALSE; + *(struct tgsi_version *) ctx->tokens_cur++ = tgsi_build_version(); + + if (ctx->tokens_cur >= ctx->tokens_end) + return FALSE; + ctx->header = (struct tgsi_header *) ctx->tokens_cur++; + *ctx->header = tgsi_build_header(); + + if (ctx->tokens_cur >= ctx->tokens_end) + return FALSE; + *(struct tgsi_processor *) ctx->tokens_cur++ = tgsi_build_processor( processor, ctx->header ); + + return TRUE; +} + +static boolean parse_label( struct translate_ctx *ctx, uint *val ) +{ + const char *cur = ctx->cur; + + if (parse_uint( &cur, val )) { + eat_opt_white( &cur ); + if (*cur == ':') { + cur++; + ctx->cur = cur; + return TRUE; + } + } + return FALSE; +} + +static const char *file_names[TGSI_FILE_COUNT] = +{ + "NULL", + "CONST", + "IN", + "OUT", + "TEMP", + "SAMP", + "ADDR", + "IMM" +}; + +static boolean +parse_file( const char **pcur, uint *file ) +{ + uint i; + + for (i = 0; i < TGSI_FILE_COUNT; i++) { + const char *cur = *pcur; + + if (str_match_no_case( &cur, file_names[i] )) { + if (!is_digit_alpha_underscore( cur )) { + *pcur = cur; + *file = i; + return TRUE; + } + } + } + return FALSE; +} + +static boolean +parse_opt_writemask( + struct translate_ctx *ctx, + uint *writemask ) +{ + const char *cur; + + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == '.') { + cur++; + *writemask = TGSI_WRITEMASK_NONE; + eat_opt_white( &cur ); + if (toupper( *cur ) == 'X') { + cur++; + *writemask |= TGSI_WRITEMASK_X; + } + if (toupper( *cur ) == 'Y') { + cur++; + *writemask |= TGSI_WRITEMASK_Y; + } + if (toupper( *cur ) == 'Z') { + cur++; + *writemask |= TGSI_WRITEMASK_Z; + } + if (toupper( *cur ) == 'W') { + cur++; + *writemask |= TGSI_WRITEMASK_W; + } + + if (*writemask == TGSI_WRITEMASK_NONE) { + report_error( ctx, "Writemask expected" ); + return FALSE; + } + + ctx->cur = cur; + } + else { + *writemask = TGSI_WRITEMASK_XYZW; + } + return TRUE; +} + +/* ::= `[' + */ +static boolean +parse_register_file_bracket( + struct translate_ctx *ctx, + uint *file ) +{ + if (!parse_file( &ctx->cur, file )) { + report_error( ctx, "Unknown register file" ); + return FALSE; + } + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '[') { + report_error( ctx, "Expected `['" ); + return FALSE; + } + ctx->cur++; + return TRUE; +} + +/* ::= + */ +static boolean +parse_register_file_bracket_index( + struct translate_ctx *ctx, + uint *file, + int *index ) +{ + uint uindex; + + if (!parse_register_file_bracket( ctx, file )) + return FALSE; + eat_opt_white( &ctx->cur ); + if (!parse_uint( &ctx->cur, &uindex )) { + report_error( ctx, "Expected literal unsigned integer" ); + return FALSE; + } + *index = (int) uindex; + return TRUE; +} + +/* Parse destination register operand. + * ::= `]' + */ +static boolean +parse_register_dst( + struct translate_ctx *ctx, + uint *file, + int *index ) +{ + if (!parse_register_file_bracket_index( ctx, file, index )) + return FALSE; + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ']') { + report_error( ctx, "Expected `]'" ); + return FALSE; + } + ctx->cur++; + return TRUE; +} + +/* Parse source register operand. + * ::= `]' | + * `]' | + * `+' `]' | + * `-' `]' + */ +static boolean +parse_register_src( + struct translate_ctx *ctx, + uint *file, + int *index, + uint *ind_file, + int *ind_index ) +{ + const char *cur; + uint uindex; + + if (!parse_register_file_bracket( ctx, file )) + return FALSE; + eat_opt_white( &ctx->cur ); + cur = ctx->cur; + if (parse_file( &cur, ind_file )) { + if (!parse_register_dst( ctx, ind_file, ind_index )) + return FALSE; + eat_opt_white( &ctx->cur ); + if (*ctx->cur == '+' || *ctx->cur == '-') { + boolean negate; + + negate = *ctx->cur == '-'; + ctx->cur++; + eat_opt_white( &ctx->cur ); + if (!parse_uint( &ctx->cur, &uindex )) { + report_error( ctx, "Expected literal unsigned integer" ); + return FALSE; + } + if (negate) + *index = -(int) uindex; + else + *index = (int) uindex; + } + else { + *index = 0; + } + } + else { + if (!parse_uint( &ctx->cur, &uindex )) { + report_error( ctx, "Expected literal unsigned integer" ); + return FALSE; + } + *index = (int) uindex; + *ind_file = TGSI_FILE_NULL; + *ind_index = 0; + } + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ']') { + report_error( ctx, "Expected `]'" ); + return FALSE; + } + ctx->cur++; + return TRUE; +} + +/* Parse register declaration. + * ::= `]' | + * `..' `]' + */ +static boolean +parse_register_dcl( + struct translate_ctx *ctx, + uint *file, + int *first, + int *last ) +{ + if (!parse_register_file_bracket_index( ctx, file, first )) + return FALSE; + eat_opt_white( &ctx->cur ); + if (ctx->cur[0] == '.' && ctx->cur[1] == '.') { + uint uindex; + + ctx->cur += 2; + eat_opt_white( &ctx->cur ); + if (!parse_uint( &ctx->cur, &uindex )) { + report_error( ctx, "Expected literal integer" ); + return FALSE; + } + *last = (int) uindex; + eat_opt_white( &ctx->cur ); + } + else { + *last = *first; + } + if (*ctx->cur != ']') { + report_error( ctx, "Expected `]' or `..'" ); + return FALSE; + } + ctx->cur++; + return TRUE; +} + +static const char *modulate_names[TGSI_MODULATE_COUNT] = +{ + "_1X", + "_2X", + "_4X", + "_8X", + "_D2", + "_D4", + "_D8" +}; + +static boolean +parse_dst_operand( + struct translate_ctx *ctx, + struct tgsi_full_dst_register *dst ) +{ + uint file; + int index; + uint writemask; + const char *cur; + + if (!parse_register_dst( ctx, &file, &index )) + return FALSE; + + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == '_') { + uint i; + + for (i = 0; i < TGSI_MODULATE_COUNT; i++) { + if (str_match_no_case( &cur, modulate_names[i] )) { + if (!is_digit_alpha_underscore( cur )) { + dst->DstRegisterExtModulate.Modulate = i; + ctx->cur = cur; + break; + } + } + } + } + + if (!parse_opt_writemask( ctx, &writemask )) + return FALSE; + + dst->DstRegister.File = file; + dst->DstRegister.Index = index; + dst->DstRegister.WriteMask = writemask; + return TRUE; +} + +static boolean +parse_optional_swizzle( + struct translate_ctx *ctx, + uint swizzle[4], + boolean *parsed_swizzle, + boolean *parsed_extswizzle ) +{ + const char *cur = ctx->cur; + + *parsed_swizzle = FALSE; + *parsed_extswizzle = FALSE; + + eat_opt_white( &cur ); + if (*cur == '.') { + uint i; + + cur++; + eat_opt_white( &cur ); + for (i = 0; i < 4; i++) { + if (toupper( *cur ) == 'X') + swizzle[i] = TGSI_SWIZZLE_X; + else if (toupper( *cur ) == 'Y') + swizzle[i] = TGSI_SWIZZLE_Y; + else if (toupper( *cur ) == 'Z') + swizzle[i] = TGSI_SWIZZLE_Z; + else if (toupper( *cur ) == 'W') + swizzle[i] = TGSI_SWIZZLE_W; + else { + if (*cur == '0') + swizzle[i] = TGSI_EXTSWIZZLE_ZERO; + else if (*cur == '1') + swizzle[i] = TGSI_EXTSWIZZLE_ONE; + else { + report_error( ctx, "Expected register swizzle component `x', `y', `z', `w', `0' or `1'" ); + return FALSE; + } + *parsed_extswizzle = TRUE; + } + cur++; + } + *parsed_swizzle = TRUE; + ctx->cur = cur; + } + return TRUE; +} + +static boolean +parse_src_operand( + struct translate_ctx *ctx, + struct tgsi_full_src_register *src ) +{ + const char *cur; + float value; + uint file; + int index; + uint ind_file; + int ind_index; + uint swizzle[4]; + boolean parsed_swizzle; + boolean parsed_extswizzle; + + if (*ctx->cur == '-') { + cur = ctx->cur; + cur++; + eat_opt_white( &cur ); + if (*cur == '(') { + cur++; + src->SrcRegisterExtMod.Negate = 1; + eat_opt_white( &cur ); + ctx->cur = cur; + } + } + + if (*ctx->cur == '|') { + ctx->cur++; + eat_opt_white( &ctx->cur ); + src->SrcRegisterExtMod.Absolute = 1; + } + + if (*ctx->cur == '-') { + ctx->cur++; + eat_opt_white( &ctx->cur ); + src->SrcRegister.Negate = 1; + } + + cur = ctx->cur; + if (parse_float( &cur, &value )) { + if (value == 2.0f) { + eat_opt_white( &cur ); + if (*cur != '*') { + report_error( ctx, "Expected `*'" ); + return FALSE; + } + cur++; + if (*cur != '(') { + report_error( ctx, "Expected `('" ); + return FALSE; + } + cur++; + src->SrcRegisterExtMod.Scale2X = 1; + eat_opt_white( &cur ); + ctx->cur = cur; + } + } + + if (*ctx->cur == '(') { + ctx->cur++; + eat_opt_white( &ctx->cur ); + src->SrcRegisterExtMod.Bias = 1; + } + + cur = ctx->cur; + if (parse_float( &cur, &value )) { + if (value == 1.0f) { + eat_opt_white( &cur ); + if (*cur != '-') { + report_error( ctx, "Expected `-'" ); + return FALSE; + } + cur++; + if (*cur != '(') { + report_error( ctx, "Expected `('" ); + return FALSE; + } + cur++; + src->SrcRegisterExtMod.Complement = 1; + eat_opt_white( &cur ); + ctx->cur = cur; + } + } + + if (!parse_register_src( ctx, &file, &index, &ind_file, &ind_index )) + return FALSE; + src->SrcRegister.File = file; + src->SrcRegister.Index = index; + if (ind_file != TGSI_FILE_NULL) { + src->SrcRegister.Indirect = 1; + src->SrcRegisterInd.File = ind_file; + src->SrcRegisterInd.Index = ind_index; + } + + /* Parse optional swizzle. + */ + if (parse_optional_swizzle( ctx, swizzle, &parsed_swizzle, &parsed_extswizzle )) { + if (parsed_extswizzle) { + assert( parsed_swizzle ); + + src->SrcRegisterExtSwz.ExtSwizzleX = swizzle[0]; + src->SrcRegisterExtSwz.ExtSwizzleY = swizzle[1]; + src->SrcRegisterExtSwz.ExtSwizzleZ = swizzle[2]; + src->SrcRegisterExtSwz.ExtSwizzleW = swizzle[3]; + } + else if (parsed_swizzle) { + src->SrcRegister.SwizzleX = swizzle[0]; + src->SrcRegister.SwizzleY = swizzle[1]; + src->SrcRegister.SwizzleZ = swizzle[2]; + src->SrcRegister.SwizzleW = swizzle[3]; + } + } + + if (src->SrcRegisterExtMod.Complement) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ')') { + report_error( ctx, "Expected `)'" ); + return FALSE; + } + ctx->cur++; + } + + if (src->SrcRegisterExtMod.Bias) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ')') { + report_error( ctx, "Expected `)'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '-') { + report_error( ctx, "Expected `-'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + if (!parse_float( &ctx->cur, &value )) { + report_error( ctx, "Expected literal floating point" ); + return FALSE; + } + if (value != 0.5f) { + report_error( ctx, "Expected 0.5" ); + return FALSE; + } + } + + if (src->SrcRegisterExtMod.Scale2X) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ')') { + report_error( ctx, "Expected `)'" ); + return FALSE; + } + ctx->cur++; + } + + if (src->SrcRegisterExtMod.Absolute) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '|') { + report_error( ctx, "Expected `|'" ); + return FALSE; + } + ctx->cur++; + } + + if (src->SrcRegisterExtMod.Negate) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ')') { + report_error( ctx, "Expected `)'" ); + return FALSE; + } + ctx->cur++; + } + + return TRUE; +} + +struct opcode_info +{ + uint num_dst; + uint num_src; + uint is_tex; + uint is_branch; + const char *mnemonic; +}; + +static const struct opcode_info opcode_info[TGSI_OPCODE_LAST] = +{ + { 1, 1, 0, 0, "ARL" }, + { 1, 1, 0, 0, "MOV" }, + { 1, 1, 0, 0, "LIT" }, + { 1, 1, 0, 0, "RCP" }, + { 1, 1, 0, 0, "RSQ" }, + { 1, 1, 0, 0, "EXP" }, + { 1, 1, 0, 0, "LOG" }, + { 1, 2, 0, 0, "MUL" }, + { 1, 2, 0, 0, "ADD" }, + { 1, 2, 0, 0, "DP3" }, + { 1, 2, 0, 0, "DP4" }, + { 1, 2, 0, 0, "DST" }, + { 1, 2, 0, 0, "MIN" }, + { 1, 2, 0, 0, "MAX" }, + { 1, 2, 0, 0, "SLT" }, + { 1, 2, 0, 0, "SGE" }, + { 1, 3, 0, 0, "MAD" }, + { 1, 2, 0, 0, "SUB" }, + { 1, 3, 0, 0, "LERP" }, + { 1, 3, 0, 0, "CND" }, + { 1, 3, 0, 0, "CND0" }, + { 1, 3, 0, 0, "DOT2ADD" }, + { 1, 2, 0, 0, "INDEX" }, + { 1, 1, 0, 0, "NEGATE" }, + { 1, 1, 0, 0, "FRAC" }, + { 1, 3, 0, 0, "CLAMP" }, + { 1, 1, 0, 0, "FLOOR" }, + { 1, 1, 0, 0, "ROUND" }, + { 1, 1, 0, 0, "EXPBASE2" }, + { 1, 1, 0, 0, "LOGBASE2" }, + { 1, 2, 0, 0, "POWER" }, + { 1, 2, 0, 0, "CROSSPRODUCT" }, + { 1, 2, 0, 0, "MULTIPLYMATRIX" }, + { 1, 1, 0, 0, "ABS" }, + { 1, 1, 0, 0, "RCC" }, + { 1, 2, 0, 0, "DPH" }, + { 1, 1, 0, 0, "COS" }, + { 1, 1, 0, 0, "DDX" }, + { 1, 1, 0, 0, "DDY" }, + { 0, 1, 0, 0, "KILP" }, + { 1, 1, 0, 0, "PK2H" }, + { 1, 1, 0, 0, "PK2US" }, + { 1, 1, 0, 0, "PK4B" }, + { 1, 1, 0, 0, "PK4UB" }, + { 1, 2, 0, 0, "RFL" }, + { 1, 2, 0, 0, "SEQ" }, + { 1, 2, 0, 0, "SFL" }, + { 1, 2, 0, 0, "SGT" }, + { 1, 1, 0, 0, "SIN" }, + { 1, 2, 0, 0, "SLE" }, + { 1, 2, 0, 0, "SNE" }, + { 1, 2, 0, 0, "STR" }, + { 1, 2, 1, 0, "TEX" }, + { 1, 4, 1, 0, "TXD" }, + { 1, 2, 1, 0, "TXP" }, + { 1, 1, 0, 0, "UP2H" }, + { 1, 1, 0, 0, "UP2US" }, + { 1, 1, 0, 0, "UP4B" }, + { 1, 1, 0, 0, "UP4UB" }, + { 1, 3, 0, 0, "X2D" }, + { 1, 1, 0, 0, "ARA" }, + { 1, 1, 0, 0, "ARR" }, + { 0, 1, 0, 0, "BRA" }, + { 0, 0, 0, 1, "CAL" }, + { 0, 0, 0, 0, "RET" }, + { 1, 1, 0, 0, "SSG" }, + { 1, 3, 0, 0, "CMP" }, + { 1, 1, 0, 0, "SCS" }, + { 1, 2, 1, 0, "TXB" }, + { 1, 1, 0, 0, "NRM" }, + { 1, 2, 0, 0, "DIV" }, + { 1, 2, 0, 0, "DP2" }, + { 1, 2, 1, 0, "TXL" }, + { 0, 0, 0, 0, "BRK" }, + { 0, 1, 0, 1, "IF" }, + { 0, 0, 0, 0, "LOOP" }, + { 0, 1, 0, 0, "REP" }, + { 0, 0, 0, 1, "ELSE" }, + { 0, 0, 0, 0, "ENDIF" }, + { 0, 0, 0, 0, "ENDLOOP" }, + { 0, 0, 0, 0, "ENDREP" }, + { 0, 1, 0, 0, "PUSHA" }, + { 1, 0, 0, 0, "POPA" }, + { 1, 1, 0, 0, "CEIL" }, + { 1, 1, 0, 0, "I2F" }, + { 1, 1, 0, 0, "NOT" }, + { 1, 1, 0, 0, "TRUNC" }, + { 1, 2, 0, 0, "SHL" }, + { 1, 2, 0, 0, "SHR" }, + { 1, 2, 0, 0, "AND" }, + { 1, 2, 0, 0, "OR" }, + { 1, 2, 0, 0, "MOD" }, + { 1, 2, 0, 0, "XOR" }, + { 1, 3, 0, 0, "SAD" }, + { 1, 2, 1, 0, "TXF" }, + { 1, 2, 1, 0, "TXQ" }, + { 0, 0, 0, 0, "CONT" }, + { 0, 0, 0, 0, "EMIT" }, + { 0, 0, 0, 0, "ENDPRIM" }, + { 0, 0, 0, 1, "BGNLOOP2" }, + { 0, 0, 0, 0, "BGNSUB" }, + { 0, 0, 0, 1, "ENDLOOP2" }, + { 0, 0, 0, 0, "ENDSUB" }, + { 1, 1, 0, 0, "NOISE1" }, + { 1, 1, 0, 0, "NOISE2" }, + { 1, 1, 0, 0, "NOISE3" }, + { 1, 1, 0, 0, "NOISE4" }, + { 0, 0, 0, 0, "NOP" }, + { 1, 2, 0, 0, "M4X3" }, + { 1, 2, 0, 0, "M3X4" }, + { 1, 2, 0, 0, "M3X3" }, + { 1, 2, 0, 0, "M3X2" }, + { 1, 1, 0, 0, "NRM4" }, + { 0, 1, 0, 0, "CALLNZ" }, + { 0, 1, 0, 0, "IFC" }, + { 0, 1, 0, 0, "BREAKC" }, + { 0, 0, 0, 0, "KIL" }, + { 0, 0, 0, 0, "END" }, + { 1, 1, 0, 0, "SWZ" } +}; + +static const char *texture_names[TGSI_TEXTURE_COUNT] = +{ + "UNKNOWN", + "1D", + "2D", + "3D", + "CUBE", + "RECT", + "SHADOW1D", + "SHADOW2D", + "SHADOWRECT" +}; + +static boolean +parse_instruction( + struct translate_ctx *ctx, + boolean has_label ) +{ + uint i; + uint saturate = TGSI_SAT_NONE; + const struct opcode_info *info; + struct tgsi_full_instruction inst; + uint advance; + + /* Parse instruction name. + */ + eat_opt_white( &ctx->cur ); + for (i = 0; i < TGSI_OPCODE_LAST; i++) { + const char *cur = ctx->cur; + + info = &opcode_info[i]; + if (str_match_no_case( &cur, info->mnemonic )) { + if (str_match_no_case( &cur, "_SATNV" )) + saturate = TGSI_SAT_MINUS_PLUS_ONE; + else if (str_match_no_case( &cur, "_SAT" )) + saturate = TGSI_SAT_ZERO_ONE; + + if (info->num_dst + info->num_src + info->is_tex == 0) { + if (!is_digit_alpha_underscore( cur )) { + ctx->cur = cur; + break; + } + } + else if (*cur == '\0' || eat_white( &cur )) { + ctx->cur = cur; + break; + } + } + } + if (i == TGSI_OPCODE_LAST) { + if (has_label) + report_error( ctx, "Unknown opcode" ); + else + report_error( ctx, "Expected `DCL', `IMM' or a label" ); + return FALSE; + } + + inst = tgsi_default_full_instruction(); + inst.Instruction.Opcode = i; + inst.Instruction.Saturate = saturate; + inst.Instruction.NumDstRegs = info->num_dst; + inst.Instruction.NumSrcRegs = info->num_src; + + /* Parse instruction operands. + */ + for (i = 0; i < info->num_dst + info->num_src + info->is_tex; i++) { + if (i > 0) { + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ',') { + report_error( ctx, "Expected `,'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + } + + if (i < info->num_dst) { + if (!parse_dst_operand( ctx, &inst.FullDstRegisters[i] )) + return FALSE; + } + else if (i < info->num_dst + info->num_src) { + if (!parse_src_operand( ctx, &inst.FullSrcRegisters[i - info->num_dst] )) + return FALSE; + } + else { + uint j; + + for (j = 0; j < TGSI_TEXTURE_COUNT; j++) { + if (str_match_no_case( &ctx->cur, texture_names[j] )) { + if (!is_digit_alpha_underscore( ctx->cur )) { + inst.InstructionExtTexture.Texture = j; + break; + } + } + } + if (j == TGSI_TEXTURE_COUNT) { + report_error( ctx, "Expected texture target" ); + return FALSE; + } + } + } + + if (info->is_branch) { + uint target; + + eat_opt_white( &ctx->cur ); + if (*ctx->cur != ':') { + report_error( ctx, "Expected `:'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + if (!parse_uint( &ctx->cur, &target )) { + report_error( ctx, "Expected a label" ); + return FALSE; + } + inst.InstructionExtLabel.Label = target; + } + + advance = tgsi_build_full_instruction( + &inst, + ctx->tokens_cur, + ctx->header, + (uint) (ctx->tokens_end - ctx->tokens_cur) ); + if (advance == 0) + return FALSE; + ctx->tokens_cur += advance; + + return TRUE; +} + +static const char *semantic_names[TGSI_SEMANTIC_COUNT] = +{ + "POSITION", + "COLOR", + "BCOLOR", + "FOG", + "PSIZE", + "GENERIC", + "NORMAL" +}; + +static const char *interpolate_names[TGSI_INTERPOLATE_COUNT] = +{ + "CONSTANT", + "LINEAR", + "PERSPECTIVE" +}; + +static boolean parse_declaration( struct translate_ctx *ctx ) +{ + struct tgsi_full_declaration decl; + uint file; + int first; + int last; + uint writemask; + const char *cur; + uint advance; + + if (!eat_white( &ctx->cur )) { + report_error( ctx, "Syntax error" ); + return FALSE; + } + if (!parse_register_dcl( ctx, &file, &first, &last )) + return FALSE; + if (!parse_opt_writemask( ctx, &writemask )) + return FALSE; + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = file; + decl.Declaration.UsageMask = writemask; + decl.DeclarationRange.First = first; + decl.DeclarationRange.Last = last; + + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == ',') { + uint i; + + cur++; + eat_opt_white( &cur ); + for (i = 0; i < TGSI_SEMANTIC_COUNT; i++) { + if (str_match_no_case( &cur, semantic_names[i] )) { + const char *cur2 = cur; + uint index; + + if (is_digit_alpha_underscore( cur )) + continue; + eat_opt_white( &cur2 ); + if (*cur2 == '[') { + cur2++; + eat_opt_white( &cur2 ); + if (!parse_uint( &cur2, &index )) { + report_error( ctx, "Expected literal integer" ); + return FALSE; + } + eat_opt_white( &cur2 ); + if (*cur2 != ']') { + report_error( ctx, "Expected `]'" ); + return FALSE; + } + cur2++; + + decl.Semantic.SemanticIndex = index; + + cur = cur2; + } + + decl.Declaration.Semantic = 1; + decl.Semantic.SemanticName = i; + + ctx->cur = cur; + break; + } + } + } + + cur = ctx->cur; + eat_opt_white( &cur ); + if (*cur == ',') { + uint i; + + cur++; + eat_opt_white( &cur ); + for (i = 0; i < TGSI_INTERPOLATE_COUNT; i++) { + if (str_match_no_case( &cur, interpolate_names[i] )) { + if (is_digit_alpha_underscore( cur )) + continue; + decl.Declaration.Interpolate = i; + + ctx->cur = cur; + break; + } + } + if (i == TGSI_INTERPOLATE_COUNT) { + report_error( ctx, "Expected semantic or interpolate attribute" ); + return FALSE; + } + } + + advance = tgsi_build_full_declaration( + &decl, + ctx->tokens_cur, + ctx->header, + (uint) (ctx->tokens_end - ctx->tokens_cur) ); + if (advance == 0) + return FALSE; + ctx->tokens_cur += advance; + + return TRUE; +} + +static boolean parse_immediate( struct translate_ctx *ctx ) +{ + struct tgsi_full_immediate imm; + uint i; + float values[4]; + uint advance; + + if (!eat_white( &ctx->cur )) { + report_error( ctx, "Syntax error" ); + return FALSE; + } + if (!str_match_no_case( &ctx->cur, "FLT32" ) || is_digit_alpha_underscore( ctx->cur )) { + report_error( ctx, "Expected `FLT32'" ); + return FALSE; + } + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '{') { + report_error( ctx, "Expected `{'" ); + return FALSE; + } + ctx->cur++; + for (i = 0; i < 4; i++) { + eat_opt_white( &ctx->cur ); + if (i > 0) { + if (*ctx->cur != ',') { + report_error( ctx, "Expected `,'" ); + return FALSE; + } + ctx->cur++; + eat_opt_white( &ctx->cur ); + } + if (!parse_float( &ctx->cur, &values[i] )) { + report_error( ctx, "Expected literal floating point" ); + return FALSE; + } + } + eat_opt_white( &ctx->cur ); + if (*ctx->cur != '}') { + report_error( ctx, "Expected `}'" ); + return FALSE; + } + ctx->cur++; + + imm = tgsi_default_full_immediate(); + imm.Immediate.Size += 4; + imm.Immediate.DataType = TGSI_IMM_FLOAT32; + imm.u.Pointer = values; + + advance = tgsi_build_full_immediate( + &imm, + ctx->tokens_cur, + ctx->header, + (uint) (ctx->tokens_end - ctx->tokens_cur) ); + if (advance == 0) + return FALSE; + ctx->tokens_cur += advance; + + return TRUE; +} + +static boolean translate( struct translate_ctx *ctx ) +{ + eat_opt_white( &ctx->cur ); + if (!parse_header( ctx )) + return FALSE; + + while (*ctx->cur != '\0') { + uint label_val = 0; + + if (!eat_white( &ctx->cur )) { + report_error( ctx, "Syntax error" ); + return FALSE; + } + + if (*ctx->cur == '\0') + break; + + if (parse_label( ctx, &label_val )) { + if (!parse_instruction( ctx, TRUE )) + return FALSE; + } + else if (str_match_no_case( &ctx->cur, "DCL" )) { + if (!parse_declaration( ctx )) + return FALSE; + } + else if (str_match_no_case( &ctx->cur, "IMM" )) { + if (!parse_immediate( ctx )) + return FALSE; + } + else if (!parse_instruction( ctx, FALSE )) { + return FALSE; + } + } + + return TRUE; +} + +boolean +tgsi_text_translate( + const char *text, + struct tgsi_token *tokens, + uint num_tokens ) +{ + struct translate_ctx ctx; + + ctx.text = text; + ctx.cur = text; + ctx.tokens = tokens; + ctx.tokens_cur = tokens; + ctx.tokens_end = tokens + num_tokens; + + if (!translate( &ctx )) + return FALSE; + + return tgsi_sanity_check( tokens ); +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.h b/src/gallium/auxiliary/tgsi/tgsi_text.h new file mode 100644 index 0000000000..8eeeeef140 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_text.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_TEXT_H +#define TGSI_TEXT_H + +#include "pipe/p_shader_tokens.h" + +#if defined __cplusplus +extern "C" { +#endif + +boolean +tgsi_text_translate( + const char *text, + struct tgsi_token *tokens, + uint num_tokens ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_TEXT_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.c b/src/gallium/auxiliary/tgsi/tgsi_transform.c new file mode 100644 index 0000000000..357f77b05a --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_transform.c @@ -0,0 +1,199 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * TGSI program transformation utility. + * + * Authors: Brian Paul + */ + + +#include "tgsi_transform.h" + + + +static void +emit_instruction(struct tgsi_transform_context *ctx, + const struct tgsi_full_instruction *inst) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_instruction(inst, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + +static void +emit_declaration(struct tgsi_transform_context *ctx, + const struct tgsi_full_declaration *decl) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_declaration(decl, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + +static void +emit_immediate(struct tgsi_transform_context *ctx, + const struct tgsi_full_immediate *imm) +{ + uint ti = ctx->ti; + + ti += tgsi_build_full_immediate(imm, + ctx->tokens_out + ti, + ctx->header, + ctx->max_tokens_out - ti); + ctx->ti = ti; +} + + + +/** + * Apply user-defined transformations to the input shader to produce + * the output shader. + * For example, a register search-and-replace operation could be applied + * by defining a transform_instruction() callback that examined and changed + * the instruction src/dest regs. + * + * \return number of tokens emitted + */ +int +tgsi_transform_shader(const struct tgsi_token *tokens_in, + struct tgsi_token *tokens_out, + uint max_tokens_out, + struct tgsi_transform_context *ctx) +{ + uint procType; + + /* input shader */ + struct tgsi_parse_context parse; + + /* output shader */ + struct tgsi_processor *processor; + + + /** + ** callback context init + **/ + ctx->emit_instruction = emit_instruction; + ctx->emit_declaration = emit_declaration; + ctx->emit_immediate = emit_immediate; + ctx->tokens_out = tokens_out; + ctx->max_tokens_out = max_tokens_out; + + + /** + ** Setup to begin parsing input shader + **/ + if (tgsi_parse_init( &parse, tokens_in ) != TGSI_PARSE_OK) { + debug_printf("tgsi_parse_init() failed in tgsi_transform_shader()!\n"); + return -1; + } + procType = parse.FullHeader.Processor.Processor; + assert(procType == TGSI_PROCESSOR_FRAGMENT || + procType == TGSI_PROCESSOR_VERTEX || + procType == TGSI_PROCESSOR_GEOMETRY); + + + /** + ** Setup output shader + **/ + *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version(); + + ctx->header = (struct tgsi_header *) (tokens_out + 1); + *ctx->header = tgsi_build_header(); + + processor = (struct tgsi_processor *) (tokens_out + 2); + *processor = tgsi_build_processor( procType, ctx->header ); + + ctx->ti = 3; + + + /** + ** Loop over incoming program tokens/instructions + */ + while( !tgsi_parse_end_of_tokens( &parse ) ) { + + tgsi_parse_token( &parse ); + + switch( parse.FullToken.Token.Type ) { + case TGSI_TOKEN_TYPE_INSTRUCTION: + { + struct tgsi_full_instruction *fullinst + = &parse.FullToken.FullInstruction; + + if (ctx->transform_instruction) + ctx->transform_instruction(ctx, fullinst); + else + ctx->emit_instruction(ctx, fullinst); + } + break; + + case TGSI_TOKEN_TYPE_DECLARATION: + { + struct tgsi_full_declaration *fulldecl + = &parse.FullToken.FullDeclaration; + + if (ctx->transform_declaration) + ctx->transform_declaration(ctx, fulldecl); + else + ctx->emit_declaration(ctx, fulldecl); + } + break; + + case TGSI_TOKEN_TYPE_IMMEDIATE: + { + struct tgsi_full_immediate *fullimm + = &parse.FullToken.FullImmediate; + + if (ctx->transform_immediate) + ctx->transform_immediate(ctx, fullimm); + else + ctx->emit_immediate(ctx, fullimm); + } + break; + + default: + assert( 0 ); + } + } + + if (ctx->epilog) { + ctx->epilog(ctx); + } + + tgsi_parse_free (&parse); + + return ctx->ti; +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.h b/src/gallium/auxiliary/tgsi/tgsi_transform.h new file mode 100644 index 0000000000..3da0b38271 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_transform.h @@ -0,0 +1,93 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_TRANSFORM_H +#define TGSI_TRANSFORM_H + + +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_build.h" + + + +/** + * Subclass this to add caller-specific data + */ +struct tgsi_transform_context +{ +/**** PUBLIC ***/ + + /** + * User-defined callbacks invoked per instruction. + */ + void (*transform_instruction)(struct tgsi_transform_context *ctx, + struct tgsi_full_instruction *inst); + + void (*transform_declaration)(struct tgsi_transform_context *ctx, + struct tgsi_full_declaration *decl); + + void (*transform_immediate)(struct tgsi_transform_context *ctx, + struct tgsi_full_immediate *imm); + + /** + * Called at end of input program to allow caller to append extra + * instructions. Return number of tokens emitted. + */ + void (*epilog)(struct tgsi_transform_context *ctx); + + +/*** PRIVATE ***/ + + /** + * These are setup by tgsi_transform_shader() and cannot be overridden. + * Meant to be called from in the above user callback functions. + */ + void (*emit_instruction)(struct tgsi_transform_context *ctx, + const struct tgsi_full_instruction *inst); + void (*emit_declaration)(struct tgsi_transform_context *ctx, + const struct tgsi_full_declaration *decl); + void (*emit_immediate)(struct tgsi_transform_context *ctx, + const struct tgsi_full_immediate *imm); + + struct tgsi_header *header; + uint max_tokens_out; + struct tgsi_token *tokens_out; + uint ti; +}; + + + +extern int +tgsi_transform_shader(const struct tgsi_token *tokens_in, + struct tgsi_token *tokens_out, + uint max_tokens_out, + struct tgsi_transform_context *ctx); + + +#endif /* TGSI_TRANSFORM_H */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.c b/src/gallium/auxiliary/tgsi/tgsi_util.c new file mode 100644 index 0000000000..09486e649e --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_util.c @@ -0,0 +1,300 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "pipe/p_util.h" +#include "pipe/p_shader_tokens.h" +#include "tgsi_parse.h" +#include "tgsi_build.h" +#include "tgsi_util.h" + +union pointer_hack +{ + void *pointer; + uint64_t uint64; +}; + +void * +tgsi_align_128bit( + void *unaligned ) +{ + union pointer_hack ph; + + ph.uint64 = 0; + ph.pointer = unaligned; + ph.uint64 = (ph.uint64 + 15) & ~15; + return ph.pointer; +} + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->SwizzleX; + case 1: + return reg->SwizzleY; + case 2: + return reg->SwizzleZ; + case 3: + return reg->SwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->ExtSwizzleX; + case 1: + return reg->ExtSwizzleY; + case 2: + return reg->ExtSwizzleZ; + case 3: + return reg->ExtSwizzleW; + default: + assert( 0 ); + } + return 0; +} + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned swizzle; + + /* + * First, calculate the extended swizzle for a given channel. This will give + * us either a channel index into the simple swizzle or a constant 1 or 0. + */ + swizzle = tgsi_util_get_src_register_extswizzle( + ®->SrcRegisterExtSwz, + component ); + + assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); + assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); + assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); + assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); + assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); + + /* + * Second, calculate the simple swizzle for the unswizzled channel index. + * Leave the constants intact, they are not affected by the simple swizzle. + */ + if( swizzle <= TGSI_SWIZZLE_W ) { + swizzle = tgsi_util_get_src_register_swizzle( + ®->SrcRegister, + swizzle ); + } + + return swizzle; +} + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + unsigned swizzle, + unsigned component ) +{ + switch( component ) { + case 0: + reg->SwizzleX = swizzle; + break; + case 1: + reg->SwizzleY = swizzle; + break; + case 2: + reg->SwizzleZ = swizzle; + break; + case 3: + reg->SwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + unsigned swizzle, + unsigned component ) +{ + switch( component ) { + case 0: + reg->ExtSwizzleX = swizzle; + break; + case 1: + reg->ExtSwizzleY = swizzle; + break; + case 2: + reg->ExtSwizzleZ = swizzle; + break; + case 3: + reg->ExtSwizzleW = swizzle; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ) +{ + switch( component ) { + case 0: + return reg->NegateX; + case 1: + return reg->NegateY; + case 2: + return reg->NegateZ; + case 3: + return reg->NegateW; + default: + assert( 0 ); + } + return 0; +} + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ) +{ + switch( component ) { + case 0: + reg->NegateX = negate; + break; + case 1: + reg->NegateY = negate; + break; + case 2: + reg->NegateZ = negate; + break; + case 3: + reg->NegateW = negate; + break; + default: + assert( 0 ); + } +} + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ) +{ + unsigned sign_mode; + + if( reg->SrcRegisterExtMod.Absolute ) { + /* Consider only the post-abs negation. */ + + if( reg->SrcRegisterExtMod.Negate ) { + sign_mode = TGSI_UTIL_SIGN_SET; + } + else { + sign_mode = TGSI_UTIL_SIGN_CLEAR; + } + } + else { + /* Accumulate the three negations. */ + + unsigned negate; + + negate = reg->SrcRegister.Negate; + if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { + negate = !negate; + } + if( reg->SrcRegisterExtMod.Negate ) { + negate = !negate; + } + + if( negate ) { + sign_mode = TGSI_UTIL_SIGN_TOGGLE; + } + else { + sign_mode = TGSI_UTIL_SIGN_KEEP; + } + } + + return sign_mode; +} + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + unsigned sign_mode ) +{ + reg->SrcRegisterExtSwz.NegateX = 0; + reg->SrcRegisterExtSwz.NegateY = 0; + reg->SrcRegisterExtSwz.NegateZ = 0; + reg->SrcRegisterExtSwz.NegateW = 0; + + switch (sign_mode) + { + case TGSI_UTIL_SIGN_CLEAR: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_SET: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 1; + reg->SrcRegisterExtMod.Negate = 1; + break; + + case TGSI_UTIL_SIGN_TOGGLE: + reg->SrcRegister.Negate = 1; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + case TGSI_UTIL_SIGN_KEEP: + reg->SrcRegister.Negate = 0; + reg->SrcRegisterExtMod.Absolute = 0; + reg->SrcRegisterExtMod.Negate = 0; + break; + + default: + assert( 0 ); + } +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.h b/src/gallium/auxiliary/tgsi/tgsi_util.h new file mode 100644 index 0000000000..7877f34558 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_util.h @@ -0,0 +1,96 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_UTIL_H +#define TGSI_UTIL_H + +#if defined __cplusplus +extern "C" { +#endif + +void * +tgsi_align_128bit( + void *unaligned ); + +unsigned +tgsi_util_get_src_register_swizzle( + const struct tgsi_src_register *reg, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extswizzle( + const struct tgsi_src_register_ext_swz *reg, + unsigned component); + +unsigned +tgsi_util_get_full_src_register_extswizzle( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_src_register_swizzle( + struct tgsi_src_register *reg, + unsigned swizzle, + unsigned component ); + +void +tgsi_util_set_src_register_extswizzle( + struct tgsi_src_register_ext_swz *reg, + unsigned swizzle, + unsigned component ); + +unsigned +tgsi_util_get_src_register_extnegate( + const struct tgsi_src_register_ext_swz *reg, + unsigned component ); + +void +tgsi_util_set_src_register_extnegate( + struct tgsi_src_register_ext_swz *reg, + unsigned negate, + unsigned component ); + +#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ +#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ +#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ +#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ + +unsigned +tgsi_util_get_full_src_register_sign_mode( + const struct tgsi_full_src_register *reg, + unsigned component ); + +void +tgsi_util_set_full_src_register_sign_mode( + struct tgsi_full_src_register *reg, + unsigned sign_mode ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_UTIL_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.c b/src/gallium/auxiliary/tgsi/util/tgsi_build.c deleted file mode 100644 index 742ef14c35..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.c +++ /dev/null @@ -1,1324 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_build.h" -#include "tgsi_parse.h" - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ) -{ - struct tgsi_version version; - - version.MajorVersion = 1; - version.MinorVersion = 1; - version.Padding = 0; - - return version; -} - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ) -{ - struct tgsi_header header; - - header.HeaderSize = 1; - header.BodySize = 0; - - return header; -} - -static void -header_headersize_grow( struct tgsi_header *header ) -{ - assert( header->HeaderSize < 0xFF ); - assert( header->BodySize == 0 ); - - header->HeaderSize++; -} - -static void -header_bodysize_grow( struct tgsi_header *header ) -{ - assert( header->BodySize < 0xFFFFFF ); - - header->BodySize++; -} - -struct tgsi_processor -tgsi_default_processor( void ) -{ - struct tgsi_processor processor; - - processor.Processor = TGSI_PROCESSOR_FRAGMENT; - processor.Padding = 0; - - return processor; -} - -struct tgsi_processor -tgsi_build_processor( - unsigned type, - struct tgsi_header *header ) -{ - struct tgsi_processor processor; - - processor = tgsi_default_processor(); - processor.Processor = type; - - header_headersize_grow( header ); - - return processor; -} - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ) -{ - struct tgsi_declaration declaration; - - declaration.Type = TGSI_TOKEN_TYPE_DECLARATION; - declaration.Size = 1; - declaration.File = TGSI_FILE_NULL; - declaration.UsageMask = TGSI_WRITEMASK_XYZW; - declaration.Interpolate = TGSI_INTERPOLATE_CONSTANT; - declaration.Semantic = 0; - declaration.Padding = 0; - declaration.Extended = 0; - - return declaration; -} - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ) -{ - struct tgsi_declaration declaration; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( interpolate <= TGSI_INTERPOLATE_PERSPECTIVE ); - - declaration = tgsi_default_declaration(); - declaration.File = file; - declaration.UsageMask = usage_mask; - declaration.Interpolate = interpolate; - declaration.Semantic = semantic; - - header_bodysize_grow( header ); - - return declaration; -} - -static void -declaration_grow( - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - assert( declaration->Size < 0xFF ); - - declaration->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ) -{ - struct tgsi_full_declaration full_declaration; - - full_declaration.Declaration = tgsi_default_declaration(); - full_declaration.DeclarationRange = tgsi_default_declaration_range(); - full_declaration.Semantic = tgsi_default_declaration_semantic(); - - return full_declaration; -} - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0; - struct tgsi_declaration *declaration; - struct tgsi_declaration_range *dr; - - if( maxsize <= size ) - return 0; - declaration = (struct tgsi_declaration *) &tokens[size]; - size++; - - *declaration = tgsi_build_declaration( - full_decl->Declaration.File, - full_decl->Declaration.UsageMask, - full_decl->Declaration.Interpolate, - full_decl->Declaration.Semantic, - header ); - - if (maxsize <= size) - return 0; - dr = (struct tgsi_declaration_range *) &tokens[size]; - size++; - - *dr = tgsi_build_declaration_range( - full_decl->DeclarationRange.First, - full_decl->DeclarationRange.Last, - declaration, - header ); - - if( full_decl->Declaration.Semantic ) { - struct tgsi_declaration_semantic *ds; - - if( maxsize <= size ) - return 0; - ds = (struct tgsi_declaration_semantic *) &tokens[size]; - size++; - - *ds = tgsi_build_declaration_semantic( - full_decl->Semantic.SemanticName, - full_decl->Semantic.SemanticIndex, - declaration, - header ); - } - - return size; -} - -struct tgsi_declaration_range -tgsi_default_declaration_range( void ) -{ - struct tgsi_declaration_range dr; - - dr.First = 0; - dr.Last = 0; - - return dr; -} - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_range declaration_range; - - assert( last >= first ); - assert( last <= 0xFFFF ); - - declaration_range = tgsi_default_declaration_range(); - declaration_range.First = first; - declaration_range.Last = last; - - declaration_grow( declaration, header ); - - return declaration_range; -} - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ) -{ - struct tgsi_declaration_semantic ds; - - ds.SemanticName = TGSI_SEMANTIC_POSITION; - ds.SemanticIndex = 0; - ds.Padding = 0; - - return ds; -} - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ) -{ - struct tgsi_declaration_semantic ds; - - assert( semantic_name <= TGSI_SEMANTIC_COUNT ); - assert( semantic_index <= 0xFFFF ); - - ds = tgsi_default_declaration_semantic(); - ds.SemanticName = semantic_name; - ds.SemanticIndex = semantic_index; - - declaration_grow( declaration, header ); - - return ds; -} - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ) -{ - struct tgsi_immediate immediate; - - immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; - immediate.Size = 1; - immediate.DataType = TGSI_IMM_FLOAT32; - immediate.Padding = 0; - immediate.Extended = 0; - - return immediate; -} - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ) -{ - struct tgsi_immediate immediate; - - immediate = tgsi_default_immediate(); - - header_bodysize_grow( header ); - - return immediate; -} - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ) -{ - struct tgsi_full_immediate fullimm; - - fullimm.Immediate = tgsi_default_immediate(); - fullimm.u.Pointer = (void *) 0; - - return fullimm; -} - -static void -immediate_grow( - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - assert( immediate->Size < 0xFF ); - - immediate->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - float value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ) -{ - struct tgsi_immediate_float32 immediate_float32; - - immediate_float32.Float = value; - - immediate_grow( immediate, header ); - - return immediate_float32; -} - -unsigned -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0, i; - struct tgsi_immediate *immediate; - - if( maxsize <= size ) - return 0; - immediate = (struct tgsi_immediate *) &tokens[size]; - size++; - - *immediate = tgsi_build_immediate( header ); - - for( i = 0; i < full_imm->Immediate.Size - 1; i++ ) { - struct tgsi_immediate_float32 *if32; - - if( maxsize <= size ) - return 0; - if32 = (struct tgsi_immediate_float32 *) &tokens[size]; - size++; - - *if32 = tgsi_build_immediate_float32( - full_imm->u.ImmediateFloat32[i].Float, - immediate, - header ); - } - - return size; -} - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ) -{ - struct tgsi_instruction instruction; - - instruction.Type = TGSI_TOKEN_TYPE_INSTRUCTION; - instruction.Size = 1; - instruction.Opcode = TGSI_OPCODE_MOV; - instruction.Saturate = TGSI_SAT_NONE; - instruction.NumDstRegs = 1; - instruction.NumSrcRegs = 1; - instruction.Padding = 0; - instruction.Extended = 0; - - return instruction; -} - -struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ) -{ - struct tgsi_instruction instruction; - - assert (opcode <= TGSI_OPCODE_LAST); - assert (saturate <= TGSI_SAT_MINUS_PLUS_ONE); - assert (num_dst_regs <= 3); - assert (num_src_regs <= 15); - - instruction = tgsi_default_instruction(); - instruction.Opcode = opcode; - instruction.Saturate = saturate; - instruction.NumDstRegs = num_dst_regs; - instruction.NumSrcRegs = num_src_regs; - - header_bodysize_grow( header ); - - return instruction; -} - -static void -instruction_grow( - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - assert (instruction->Size < 0xFF); - - instruction->Size++; - - header_bodysize_grow( header ); -} - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ) -{ - struct tgsi_full_instruction full_instruction; - unsigned i; - - full_instruction.Instruction = tgsi_default_instruction(); - full_instruction.InstructionExtNv = tgsi_default_instruction_ext_nv(); - full_instruction.InstructionExtLabel = tgsi_default_instruction_ext_label(); - full_instruction.InstructionExtTexture = tgsi_default_instruction_ext_texture(); - for( i = 0; i < TGSI_FULL_MAX_DST_REGISTERS; i++ ) { - full_instruction.FullDstRegisters[i] = tgsi_default_full_dst_register(); - } - for( i = 0; i < TGSI_FULL_MAX_SRC_REGISTERS; i++ ) { - full_instruction.FullSrcRegisters[i] = tgsi_default_full_src_register(); - } - - return full_instruction; -} - -unsigned -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ) -{ - unsigned size = 0; - unsigned i; - struct tgsi_instruction *instruction; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - instruction = (struct tgsi_instruction *) &tokens[size]; - size++; - - *instruction = tgsi_build_instruction( - full_inst->Instruction.Opcode, - full_inst->Instruction.Saturate, - full_inst->Instruction.NumDstRegs, - full_inst->Instruction.NumSrcRegs, - header ); - prev_token = (struct tgsi_token *) instruction; - - if( tgsi_compare_instruction_ext_nv( - full_inst->InstructionExtNv, - tgsi_default_instruction_ext_nv() ) ) { - struct tgsi_instruction_ext_nv *instruction_ext_nv; - - if( maxsize <= size ) - return 0; - instruction_ext_nv = - (struct tgsi_instruction_ext_nv *) &tokens[size]; - size++; - - *instruction_ext_nv = tgsi_build_instruction_ext_nv( - full_inst->InstructionExtNv.Precision, - full_inst->InstructionExtNv.CondDstIndex, - full_inst->InstructionExtNv.CondFlowIndex, - full_inst->InstructionExtNv.CondMask, - full_inst->InstructionExtNv.CondSwizzleX, - full_inst->InstructionExtNv.CondSwizzleY, - full_inst->InstructionExtNv.CondSwizzleZ, - full_inst->InstructionExtNv.CondSwizzleW, - full_inst->InstructionExtNv.CondDstUpdate, - full_inst->InstructionExtNv.CondFlowEnable, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_nv; - } - - if( tgsi_compare_instruction_ext_label( - full_inst->InstructionExtLabel, - tgsi_default_instruction_ext_label() ) ) { - struct tgsi_instruction_ext_label *instruction_ext_label; - - if( maxsize <= size ) - return 0; - instruction_ext_label = - (struct tgsi_instruction_ext_label *) &tokens[size]; - size++; - - *instruction_ext_label = tgsi_build_instruction_ext_label( - full_inst->InstructionExtLabel.Label, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_label; - } - - if( tgsi_compare_instruction_ext_texture( - full_inst->InstructionExtTexture, - tgsi_default_instruction_ext_texture() ) ) { - struct tgsi_instruction_ext_texture *instruction_ext_texture; - - if( maxsize <= size ) - return 0; - instruction_ext_texture = - (struct tgsi_instruction_ext_texture *) &tokens[size]; - size++; - - *instruction_ext_texture = tgsi_build_instruction_ext_texture( - full_inst->InstructionExtTexture.Texture, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) instruction_ext_texture; - } - - for( i = 0; i < full_inst->Instruction.NumDstRegs; i++ ) { - const struct tgsi_full_dst_register *reg = &full_inst->FullDstRegisters[i]; - struct tgsi_dst_register *dst_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - dst_register = (struct tgsi_dst_register *) &tokens[size]; - size++; - - *dst_register = tgsi_build_dst_register( - reg->DstRegister.File, - reg->DstRegister.WriteMask, - reg->DstRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register; - - if( tgsi_compare_dst_register_ext_concode( - reg->DstRegisterExtConcode, - tgsi_default_dst_register_ext_concode() ) ) { - struct tgsi_dst_register_ext_concode *dst_register_ext_concode; - - if( maxsize <= size ) - return 0; - dst_register_ext_concode = - (struct tgsi_dst_register_ext_concode *) &tokens[size]; - size++; - - *dst_register_ext_concode = tgsi_build_dst_register_ext_concode( - reg->DstRegisterExtConcode.CondMask, - reg->DstRegisterExtConcode.CondSwizzleX, - reg->DstRegisterExtConcode.CondSwizzleY, - reg->DstRegisterExtConcode.CondSwizzleZ, - reg->DstRegisterExtConcode.CondSwizzleW, - reg->DstRegisterExtConcode.CondSrcIndex, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_concode; - } - - if( tgsi_compare_dst_register_ext_modulate( - reg->DstRegisterExtModulate, - tgsi_default_dst_register_ext_modulate() ) ) { - struct tgsi_dst_register_ext_modulate *dst_register_ext_modulate; - - if( maxsize <= size ) - return 0; - dst_register_ext_modulate = - (struct tgsi_dst_register_ext_modulate *) &tokens[size]; - size++; - - *dst_register_ext_modulate = tgsi_build_dst_register_ext_modulate( - reg->DstRegisterExtModulate.Modulate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) dst_register_ext_modulate; - } - } - - for( i = 0; i < full_inst->Instruction.NumSrcRegs; i++ ) { - const struct tgsi_full_src_register *reg = &full_inst->FullSrcRegisters[i]; - struct tgsi_src_register *src_register; - struct tgsi_token *prev_token; - - if( maxsize <= size ) - return 0; - src_register = (struct tgsi_src_register *) &tokens[size]; - size++; - - *src_register = tgsi_build_src_register( - reg->SrcRegister.File, - reg->SrcRegister.SwizzleX, - reg->SrcRegister.SwizzleY, - reg->SrcRegister.SwizzleZ, - reg->SrcRegister.SwizzleW, - reg->SrcRegister.Negate, - reg->SrcRegister.Indirect, - reg->SrcRegister.Dimension, - reg->SrcRegister.Index, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register; - - if( tgsi_compare_src_register_ext_swz( - reg->SrcRegisterExtSwz, - tgsi_default_src_register_ext_swz() ) ) { - struct tgsi_src_register_ext_swz *src_register_ext_swz; - - /* Use of the extended swizzle requires the simple swizzle to be identity. - */ - assert( reg->SrcRegister.SwizzleX == TGSI_SWIZZLE_X ); - assert( reg->SrcRegister.SwizzleY == TGSI_SWIZZLE_Y ); - assert( reg->SrcRegister.SwizzleZ == TGSI_SWIZZLE_Z ); - assert( reg->SrcRegister.SwizzleW == TGSI_SWIZZLE_W ); - assert( reg->SrcRegister.Negate == FALSE ); - - if( maxsize <= size ) - return 0; - src_register_ext_swz = - (struct tgsi_src_register_ext_swz *) &tokens[size]; - size++; - - *src_register_ext_swz = tgsi_build_src_register_ext_swz( - reg->SrcRegisterExtSwz.ExtSwizzleX, - reg->SrcRegisterExtSwz.ExtSwizzleY, - reg->SrcRegisterExtSwz.ExtSwizzleZ, - reg->SrcRegisterExtSwz.ExtSwizzleW, - reg->SrcRegisterExtSwz.NegateX, - reg->SrcRegisterExtSwz.NegateY, - reg->SrcRegisterExtSwz.NegateZ, - reg->SrcRegisterExtSwz.NegateW, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_swz; - } - - if( tgsi_compare_src_register_ext_mod( - reg->SrcRegisterExtMod, - tgsi_default_src_register_ext_mod() ) ) { - struct tgsi_src_register_ext_mod *src_register_ext_mod; - - if( maxsize <= size ) - return 0; - src_register_ext_mod = - (struct tgsi_src_register_ext_mod *) &tokens[size]; - size++; - - *src_register_ext_mod = tgsi_build_src_register_ext_mod( - reg->SrcRegisterExtMod.Complement, - reg->SrcRegisterExtMod.Bias, - reg->SrcRegisterExtMod.Scale2X, - reg->SrcRegisterExtMod.Absolute, - reg->SrcRegisterExtMod.Negate, - prev_token, - instruction, - header ); - prev_token = (struct tgsi_token *) src_register_ext_mod; - } - - if( reg->SrcRegister.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterInd.File, - reg->SrcRegisterInd.SwizzleX, - reg->SrcRegisterInd.SwizzleY, - reg->SrcRegisterInd.SwizzleZ, - reg->SrcRegisterInd.SwizzleW, - reg->SrcRegisterInd.Negate, - reg->SrcRegisterInd.Indirect, - reg->SrcRegisterInd.Dimension, - reg->SrcRegisterInd.Index, - instruction, - header ); - } - - if( reg->SrcRegister.Dimension ) { - struct tgsi_dimension *dim; - - assert( !reg->SrcRegisterDim.Dimension ); - - if( maxsize <= size ) - return 0; - dim = (struct tgsi_dimension *) &tokens[size]; - size++; - - *dim = tgsi_build_dimension( - reg->SrcRegisterDim.Indirect, - reg->SrcRegisterDim.Index, - instruction, - header ); - - if( reg->SrcRegisterDim.Indirect ) { - struct tgsi_src_register *ind; - - if( maxsize <= size ) - return 0; - ind = (struct tgsi_src_register *) &tokens[size]; - size++; - - *ind = tgsi_build_src_register( - reg->SrcRegisterDimInd.File, - reg->SrcRegisterDimInd.SwizzleX, - reg->SrcRegisterDimInd.SwizzleY, - reg->SrcRegisterDimInd.SwizzleZ, - reg->SrcRegisterDimInd.SwizzleW, - reg->SrcRegisterDimInd.Negate, - reg->SrcRegisterDimInd.Indirect, - reg->SrcRegisterDimInd.Dimension, - reg->SrcRegisterDimInd.Index, - instruction, - header ); - } - } - } - - return size; -} - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv.Type = TGSI_INSTRUCTION_EXT_TYPE_NV; - instruction_ext_nv.Precision = TGSI_PRECISION_DEFAULT; - instruction_ext_nv.CondDstIndex = 0; - instruction_ext_nv.CondFlowIndex = 0; - instruction_ext_nv.CondMask = TGSI_CC_TR; - instruction_ext_nv.CondSwizzleX = TGSI_SWIZZLE_X; - instruction_ext_nv.CondSwizzleY = TGSI_SWIZZLE_Y; - instruction_ext_nv.CondSwizzleZ = TGSI_SWIZZLE_Z; - instruction_ext_nv.CondSwizzleW = TGSI_SWIZZLE_W; - instruction_ext_nv.CondDstUpdate = 0; - instruction_ext_nv.CondFlowEnable = 0; - instruction_ext_nv.Padding = 0; - instruction_ext_nv.Extended = 0; - - return instruction_ext_nv; -} - -union token_u32 -{ - unsigned u32; -}; - -unsigned -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - unsigned precision, - unsigned cond_dst_index, - unsigned cond_flow_index, - unsigned cond_mask, - unsigned cond_swizzle_x, - unsigned cond_swizzle_y, - unsigned cond_swizzle_z, - unsigned cond_swizzle_w, - unsigned cond_dst_update, - unsigned cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_nv instruction_ext_nv; - - instruction_ext_nv = tgsi_default_instruction_ext_nv(); - instruction_ext_nv.Precision = precision; - instruction_ext_nv.CondDstIndex = cond_dst_index; - instruction_ext_nv.CondFlowIndex = cond_flow_index; - instruction_ext_nv.CondMask = cond_mask; - instruction_ext_nv.CondSwizzleX = cond_swizzle_x; - instruction_ext_nv.CondSwizzleY = cond_swizzle_y; - instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; - instruction_ext_nv.CondSwizzleW = cond_swizzle_w; - instruction_ext_nv.CondDstUpdate = cond_dst_update; - instruction_ext_nv.CondFlowEnable = cond_flow_update; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_nv; -} - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label.Type = TGSI_INSTRUCTION_EXT_TYPE_LABEL; - instruction_ext_label.Label = 0; - instruction_ext_label.Padding = 0; - instruction_ext_label.Extended = 0; - - return instruction_ext_label; -} - -unsigned -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - unsigned label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_label instruction_ext_label; - - instruction_ext_label = tgsi_default_instruction_ext_label(); - instruction_ext_label.Label = label; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_label; -} - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture.Type = TGSI_INSTRUCTION_EXT_TYPE_TEXTURE; - instruction_ext_texture.Texture = TGSI_TEXTURE_UNKNOWN; - instruction_ext_texture.Padding = 0; - instruction_ext_texture.Extended = 0; - - return instruction_ext_texture; -} - -unsigned -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - unsigned texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_instruction_ext_texture instruction_ext_texture; - - instruction_ext_texture = tgsi_default_instruction_ext_texture(); - instruction_ext_texture.Texture = texture; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return instruction_ext_texture; -} - -struct tgsi_src_register -tgsi_default_src_register( void ) -{ - struct tgsi_src_register src_register; - - src_register.File = TGSI_FILE_NULL; - src_register.SwizzleX = TGSI_SWIZZLE_X; - src_register.SwizzleY = TGSI_SWIZZLE_Y; - src_register.SwizzleZ = TGSI_SWIZZLE_Z; - src_register.SwizzleW = TGSI_SWIZZLE_W; - src_register.Negate = 0; - src_register.Indirect = 0; - src_register.Dimension = 0; - src_register.Index = 0; - src_register.Extended = 0; - - return src_register; -} - -struct tgsi_src_register -tgsi_build_src_register( - unsigned file, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - unsigned negate, - unsigned indirect, - unsigned dimension, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register src_register; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( swizzle_x <= TGSI_SWIZZLE_W ); - assert( swizzle_y <= TGSI_SWIZZLE_W ); - assert( swizzle_z <= TGSI_SWIZZLE_W ); - assert( swizzle_w <= TGSI_SWIZZLE_W ); - assert( negate <= 1 ); - assert( index >= -0x8000 && index <= 0x7FFF ); - - src_register = tgsi_default_src_register(); - src_register.File = file; - src_register.SwizzleX = swizzle_x; - src_register.SwizzleY = swizzle_y; - src_register.SwizzleZ = swizzle_z; - src_register.SwizzleW = swizzle_w; - src_register.Negate = negate; - src_register.Indirect = indirect; - src_register.Dimension = dimension; - src_register.Index = index; - - instruction_grow( instruction, header ); - - return src_register; -} - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ) -{ - struct tgsi_full_src_register full_src_register; - - full_src_register.SrcRegister = tgsi_default_src_register(); - full_src_register.SrcRegisterExtSwz = tgsi_default_src_register_ext_swz(); - full_src_register.SrcRegisterExtMod = tgsi_default_src_register_ext_mod(); - full_src_register.SrcRegisterInd = tgsi_default_src_register(); - full_src_register.SrcRegisterDim = tgsi_default_dimension(); - full_src_register.SrcRegisterDimInd = tgsi_default_src_register(); - - return full_src_register; -} - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - src_register_ext_swz.Type = TGSI_SRC_REGISTER_EXT_TYPE_SWZ; - src_register_ext_swz.ExtSwizzleX = TGSI_EXTSWIZZLE_X; - src_register_ext_swz.ExtSwizzleY = TGSI_EXTSWIZZLE_Y; - src_register_ext_swz.ExtSwizzleZ = TGSI_EXTSWIZZLE_Z; - src_register_ext_swz.ExtSwizzleW = TGSI_EXTSWIZZLE_W; - src_register_ext_swz.NegateX = 0; - src_register_ext_swz.NegateY = 0; - src_register_ext_swz.NegateZ = 0; - src_register_ext_swz.NegateW = 0; - src_register_ext_swz.Padding = 0; - src_register_ext_swz.Extended = 0; - - return src_register_ext_swz; -} - -unsigned -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - unsigned ext_swizzle_x, - unsigned ext_swizzle_y, - unsigned ext_swizzle_z, - unsigned ext_swizzle_w, - unsigned negate_x, - unsigned negate_y, - unsigned negate_z, - unsigned negate_w, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_swz src_register_ext_swz; - - assert( ext_swizzle_x <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_y <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_z <= TGSI_EXTSWIZZLE_ONE ); - assert( ext_swizzle_w <= TGSI_EXTSWIZZLE_ONE ); - assert( negate_x <= 1 ); - assert( negate_y <= 1 ); - assert( negate_z <= 1 ); - assert( negate_w <= 1 ); - - src_register_ext_swz = tgsi_default_src_register_ext_swz(); - src_register_ext_swz.ExtSwizzleX = ext_swizzle_x; - src_register_ext_swz.ExtSwizzleY = ext_swizzle_y; - src_register_ext_swz.ExtSwizzleZ = ext_swizzle_z; - src_register_ext_swz.ExtSwizzleW = ext_swizzle_w; - src_register_ext_swz.NegateX = negate_x; - src_register_ext_swz.NegateY = negate_y; - src_register_ext_swz.NegateZ = negate_z; - src_register_ext_swz.NegateW = negate_w; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_swz; -} - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - src_register_ext_mod.Type = TGSI_SRC_REGISTER_EXT_TYPE_MOD; - src_register_ext_mod.Complement = 0; - src_register_ext_mod.Bias = 0; - src_register_ext_mod.Scale2X = 0; - src_register_ext_mod.Absolute = 0; - src_register_ext_mod.Negate = 0; - src_register_ext_mod.Padding = 0; - src_register_ext_mod.Extended = 0; - - return src_register_ext_mod; -} - -unsigned -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - unsigned complement, - unsigned bias, - unsigned scale_2x, - unsigned absolute, - unsigned negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_src_register_ext_mod src_register_ext_mod; - - assert( complement <= 1 ); - assert( bias <= 1 ); - assert( scale_2x <= 1 ); - assert( absolute <= 1 ); - assert( negate <= 1 ); - - src_register_ext_mod = tgsi_default_src_register_ext_mod(); - src_register_ext_mod.Complement = complement; - src_register_ext_mod.Bias = bias; - src_register_ext_mod.Scale2X = scale_2x; - src_register_ext_mod.Absolute = absolute; - src_register_ext_mod.Negate = negate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return src_register_ext_mod; -} - -struct tgsi_dimension -tgsi_default_dimension( void ) -{ - struct tgsi_dimension dimension; - - dimension.Indirect = 0; - dimension.Dimension = 0; - dimension.Padding = 0; - dimension.Index = 0; - dimension.Extended = 0; - - return dimension; -} - -struct tgsi_dimension -tgsi_build_dimension( - unsigned indirect, - unsigned index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dimension dimension; - - dimension = tgsi_default_dimension(); - dimension.Indirect = indirect; - dimension.Index = index; - - instruction_grow( instruction, header ); - - return dimension; -} - -struct tgsi_dst_register -tgsi_default_dst_register( void ) -{ - struct tgsi_dst_register dst_register; - - dst_register.File = TGSI_FILE_NULL; - dst_register.WriteMask = TGSI_WRITEMASK_XYZW; - dst_register.Indirect = 0; - dst_register.Dimension = 0; - dst_register.Index = 0; - dst_register.Padding = 0; - dst_register.Extended = 0; - - return dst_register; -} - -struct tgsi_dst_register -tgsi_build_dst_register( - unsigned file, - unsigned mask, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register dst_register; - - assert( file <= TGSI_FILE_IMMEDIATE ); - assert( mask <= TGSI_WRITEMASK_XYZW ); - assert( index >= -32768 && index <= 32767 ); - - dst_register = tgsi_default_dst_register(); - dst_register.File = file; - dst_register.WriteMask = mask; - dst_register.Index = index; - - instruction_grow( instruction, header ); - - return dst_register; -} - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ) -{ - struct tgsi_full_dst_register full_dst_register; - - full_dst_register.DstRegister = tgsi_default_dst_register(); - full_dst_register.DstRegisterExtConcode = - tgsi_default_dst_register_ext_concode(); - full_dst_register.DstRegisterExtModulate = - tgsi_default_dst_register_ext_modulate(); - - return full_dst_register; -} - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - dst_register_ext_concode.Type = TGSI_DST_REGISTER_EXT_TYPE_CONDCODE; - dst_register_ext_concode.CondMask = TGSI_CC_TR; - dst_register_ext_concode.CondSwizzleX = TGSI_SWIZZLE_X; - dst_register_ext_concode.CondSwizzleY = TGSI_SWIZZLE_Y; - dst_register_ext_concode.CondSwizzleZ = TGSI_SWIZZLE_Z; - dst_register_ext_concode.CondSwizzleW = TGSI_SWIZZLE_W; - dst_register_ext_concode.CondSrcIndex = 0; - dst_register_ext_concode.Padding = 0; - dst_register_ext_concode.Extended = 0; - - return dst_register_ext_concode; -} - -unsigned -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - unsigned cc, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - int index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_concode dst_register_ext_concode; - - assert( cc <= TGSI_CC_FL ); - assert( swizzle_x <= TGSI_SWIZZLE_W ); - assert( swizzle_y <= TGSI_SWIZZLE_W ); - assert( swizzle_z <= TGSI_SWIZZLE_W ); - assert( swizzle_w <= TGSI_SWIZZLE_W ); - assert( index >= -32768 && index <= 32767 ); - - dst_register_ext_concode = tgsi_default_dst_register_ext_concode(); - dst_register_ext_concode.CondMask = cc; - dst_register_ext_concode.CondSwizzleX = swizzle_x; - dst_register_ext_concode.CondSwizzleY = swizzle_y; - dst_register_ext_concode.CondSwizzleZ = swizzle_z; - dst_register_ext_concode.CondSwizzleW = swizzle_w; - dst_register_ext_concode.CondSrcIndex = index; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_concode; -} - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - dst_register_ext_modulate.Type = TGSI_DST_REGISTER_EXT_TYPE_MODULATE; - dst_register_ext_modulate.Modulate = TGSI_MODULATE_1X; - dst_register_ext_modulate.Padding = 0; - dst_register_ext_modulate.Extended = 0; - - return dst_register_ext_modulate; -} - -unsigned -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ) -{ - a.Padding = b.Padding = 0; - a.Extended = b.Extended = 0; - return ((union token_u32 *) &a)->u32 != ((union token_u32 *) &b)->u32; -} - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - unsigned modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ) -{ - struct tgsi_dst_register_ext_modulate dst_register_ext_modulate; - - assert( modulate <= TGSI_MODULATE_EIGHTH ); - - dst_register_ext_modulate = tgsi_default_dst_register_ext_modulate(); - dst_register_ext_modulate.Modulate = modulate; - - prev_token->Extended = 1; - instruction_grow( instruction, header ); - - return dst_register_ext_modulate; -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_build.h b/src/gallium/auxiliary/tgsi/util/tgsi_build.h deleted file mode 100644 index ed25830248..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_build.h +++ /dev/null @@ -1,332 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_BUILD_H -#define TGSI_BUILD_H - -#if defined __cplusplus -extern "C" { -#endif - -/* - * version - */ - -struct tgsi_version -tgsi_build_version( void ); - -/* - * header - */ - -struct tgsi_header -tgsi_build_header( void ); - -struct tgsi_processor -tgsi_default_processor( void ); - -struct tgsi_processor -tgsi_build_processor( - unsigned processor, - struct tgsi_header *header ); - -/* - * declaration - */ - -struct tgsi_declaration -tgsi_default_declaration( void ); - -struct tgsi_declaration -tgsi_build_declaration( - unsigned file, - unsigned usage_mask, - unsigned interpolate, - unsigned semantic, - struct tgsi_header *header ); - -struct tgsi_full_declaration -tgsi_default_full_declaration( void ); - -unsigned -tgsi_build_full_declaration( - const struct tgsi_full_declaration *full_decl, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_declaration_range -tgsi_default_declaration_range( void ); - -struct tgsi_declaration_range -tgsi_build_declaration_range( - unsigned first, - unsigned last, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -struct tgsi_declaration_semantic -tgsi_default_declaration_semantic( void ); - -struct tgsi_declaration_semantic -tgsi_build_declaration_semantic( - unsigned semantic_name, - unsigned semantic_index, - struct tgsi_declaration *declaration, - struct tgsi_header *header ); - -/* - * immediate - */ - -struct tgsi_immediate -tgsi_default_immediate( void ); - -struct tgsi_immediate -tgsi_build_immediate( - struct tgsi_header *header ); - -struct tgsi_full_immediate -tgsi_default_full_immediate( void ); - -struct tgsi_immediate_float32 -tgsi_build_immediate_float32( - float value, - struct tgsi_immediate *immediate, - struct tgsi_header *header ); - -unsigned -tgsi_build_full_immediate( - const struct tgsi_full_immediate *full_imm, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -/* - * instruction - */ - -struct tgsi_instruction -tgsi_default_instruction( void ); - -struct tgsi_instruction -tgsi_build_instruction( - unsigned opcode, - unsigned saturate, - unsigned num_dst_regs, - unsigned num_src_regs, - struct tgsi_header *header ); - -struct tgsi_full_instruction -tgsi_default_full_instruction( void ); - -unsigned -tgsi_build_full_instruction( - const struct tgsi_full_instruction *full_inst, - struct tgsi_token *tokens, - struct tgsi_header *header, - unsigned maxsize ); - -struct tgsi_instruction_ext_nv -tgsi_default_instruction_ext_nv( void ); - -unsigned -tgsi_compare_instruction_ext_nv( - struct tgsi_instruction_ext_nv a, - struct tgsi_instruction_ext_nv b ); - -struct tgsi_instruction_ext_nv -tgsi_build_instruction_ext_nv( - unsigned precision, - unsigned cond_dst_index, - unsigned cond_flow_index, - unsigned cond_mask, - unsigned cond_swizzle_x, - unsigned cond_swizzle_y, - unsigned cond_swizzle_z, - unsigned cond_swizzle_w, - unsigned cond_dst_update, - unsigned cond_flow_update, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_label -tgsi_default_instruction_ext_label( void ); - -unsigned -tgsi_compare_instruction_ext_label( - struct tgsi_instruction_ext_label a, - struct tgsi_instruction_ext_label b ); - -struct tgsi_instruction_ext_label -tgsi_build_instruction_ext_label( - unsigned label, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_instruction_ext_texture -tgsi_default_instruction_ext_texture( void ); - -unsigned -tgsi_compare_instruction_ext_texture( - struct tgsi_instruction_ext_texture a, - struct tgsi_instruction_ext_texture b ); - -struct tgsi_instruction_ext_texture -tgsi_build_instruction_ext_texture( - unsigned texture, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register -tgsi_default_src_register( void ); - -struct tgsi_src_register -tgsi_build_src_register( - unsigned file, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - unsigned negate, - unsigned indirect, - unsigned dimension, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_src_register -tgsi_default_full_src_register( void ); - -struct tgsi_src_register_ext_swz -tgsi_default_src_register_ext_swz( void ); - -unsigned -tgsi_compare_src_register_ext_swz( - struct tgsi_src_register_ext_swz a, - struct tgsi_src_register_ext_swz b ); - -struct tgsi_src_register_ext_swz -tgsi_build_src_register_ext_swz( - unsigned ext_swizzle_x, - unsigned ext_swizzle_y, - unsigned ext_swizzle_z, - unsigned ext_swizzle_w, - unsigned negate_x, - unsigned negate_y, - unsigned negate_z, - unsigned negate_w, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_src_register_ext_mod -tgsi_default_src_register_ext_mod( void ); - -unsigned -tgsi_compare_src_register_ext_mod( - struct tgsi_src_register_ext_mod a, - struct tgsi_src_register_ext_mod b ); - -struct tgsi_src_register_ext_mod -tgsi_build_src_register_ext_mod( - unsigned complement, - unsigned bias, - unsigned scale_2x, - unsigned absolute, - unsigned negate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dimension -tgsi_default_dimension( void ); - -struct tgsi_dimension -tgsi_build_dimension( - unsigned indirect, - unsigned index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register -tgsi_default_dst_register( void ); - -struct tgsi_dst_register -tgsi_build_dst_register( - unsigned file, - unsigned mask, - int index, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_full_dst_register -tgsi_default_full_dst_register( void ); - -struct tgsi_dst_register_ext_concode -tgsi_default_dst_register_ext_concode( void ); - -unsigned -tgsi_compare_dst_register_ext_concode( - struct tgsi_dst_register_ext_concode a, - struct tgsi_dst_register_ext_concode b ); - -struct tgsi_dst_register_ext_concode -tgsi_build_dst_register_ext_concode( - unsigned cc, - unsigned swizzle_x, - unsigned swizzle_y, - unsigned swizzle_z, - unsigned swizzle_w, - int index, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -struct tgsi_dst_register_ext_modulate -tgsi_default_dst_register_ext_modulate( void ); - -unsigned -tgsi_compare_dst_register_ext_modulate( - struct tgsi_dst_register_ext_modulate a, - struct tgsi_dst_register_ext_modulate b ); - -struct tgsi_dst_register_ext_modulate -tgsi_build_dst_register_ext_modulate( - unsigned modulate, - struct tgsi_token *prev_token, - struct tgsi_instruction *instruction, - struct tgsi_header *header ); - -#if defined __cplusplus -} -#endif - -#endif /* TGSI_BUILD_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump.c deleted file mode 100644 index d2e6375212..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.c +++ /dev/null @@ -1,582 +0,0 @@ -/************************************************************************** - * - * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "tgsi_dump.h" -#include "tgsi_iterate.h" - -struct dump_ctx -{ - struct tgsi_iterate_context iter; - - uint instno; -}; - -static void -dump_enum( - uint e, - const char **enums, - uint enum_count ) -{ - if (e >= enum_count) - debug_printf( "%u", e ); - else - debug_printf( "%s", enums[e] ); -} - -#define EOL() debug_printf( "\n" ) -#define TXT(S) debug_printf( "%s", S ) -#define CHR(C) debug_printf( "%c", C ) -#define UIX(I) debug_printf( "0x%x", I ) -#define UID(I) debug_printf( "%u", I ) -#define SID(I) debug_printf( "%d", I ) -#define FLT(F) debug_printf( "%10.4f", F ) -#define ENM(E,ENUMS) dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) - -static const char *processor_type_names[] = -{ - "FRAG", - "VERT", - "GEOM" -}; - -static const char *file_names[] = -{ - "NULL", - "CONST", - "IN", - "OUT", - "TEMP", - "SAMP", - "ADDR", - "IMM" -}; - -static const char *interpolate_names[] = -{ - "CONSTANT", - "LINEAR", - "PERSPECTIVE" -}; - -static const char *semantic_names[] = -{ - "POSITION", - "COLOR", - "BCOLOR", - "FOG", - "PSIZE", - "GENERIC", - "NORMAL" -}; - -static const char *immediate_type_names[] = -{ - "FLT32" -}; - -static const char *opcode_names[TGSI_OPCODE_LAST] = -{ - "ARL", - "MOV", - "LIT", - "RCP", - "RSQ", - "EXP", - "LOG", - "MUL", - "ADD", - "DP3", - "DP4", - "DST", - "MIN", - "MAX", - "SLT", - "SGE", - "MAD", - "SUB", - "LERP", - "CND", - "CND0", - "DOT2ADD", - "INDEX", - "NEGATE", - "FRAC", - "CLAMP", - "FLOOR", - "ROUND", - "EXPBASE2", - "LOGBASE2", - "POWER", - "CROSSPRODUCT", - "MULTIPLYMATRIX", - "ABS", - "RCC", - "DPH", - "COS", - "DDX", - "DDY", - "KILP", - "PK2H", - "PK2US", - "PK4B", - "PK4UB", - "RFL", - "SEQ", - "SFL", - "SGT", - "SIN", - "SLE", - "SNE", - "STR", - "TEX", - "TXD", - "TXP", - "UP2H", - "UP2US", - "UP4B", - "UP4UB", - "X2D", - "ARA", - "ARR", - "BRA", - "CAL", - "RET", - "SSG", - "CMP", - "SCS", - "TXB", - "NRM", - "DIV", - "DP2", - "TXL", - "BRK", - "IF", - "LOOP", - "REP", - "ELSE", - "ENDIF", - "ENDLOOP", - "ENDREP", - "PUSHA", - "POPA", - "CEIL", - "I2F", - "NOT", - "TRUNC", - "SHL", - "SHR", - "AND", - "OR", - "MOD", - "XOR", - "SAD", - "TXF", - "TXQ", - "CONT", - "EMIT", - "ENDPRIM", - "BGNLOOP2", - "BGNSUB", - "ENDLOOP2", - "ENDSUB", - "NOISE1", - "NOISE2", - "NOISE3", - "NOISE4", - "NOP", - "M4X3", - "M3X4", - "M3X3", - "M3X2", - "NRM4", - "CALLNZ", - "IFC", - "BREAKC", - "KIL", - "END", - "SWZ" -}; - -static const char *swizzle_names[] = -{ - "x", - "y", - "z", - "w" -}; - -static const char *texture_names[] = -{ - "UNKNOWN", - "1D", - "2D", - "3D", - "CUBE", - "RECT", - "SHADOW1D", - "SHADOW2D", - "SHADOWRECT" -}; - -static const char *extswizzle_names[] = -{ - "x", - "y", - "z", - "w", - "0", - "1" -}; - -static const char *modulate_names[TGSI_MODULATE_COUNT] = -{ - "", - "_2X", - "_4X", - "_8X", - "_D2", - "_D4", - "_D8" -}; - -static void -_dump_register_prefix( - uint file, - uint first, - uint last ) -{ - - -} - -static void -_dump_register( - uint file, - int first, - int last ) -{ - ENM( file, file_names ); - CHR( '[' ); - SID( first ); - if (first != last) { - TXT( ".." ); - SID( last ); - } - CHR( ']' ); -} - -static void -_dump_register_ind( - uint file, - int index, - uint ind_file, - int ind_index ) -{ - ENM( file, file_names ); - CHR( '[' ); - ENM( ind_file, file_names ); - CHR( '[' ); - SID( ind_index ); - CHR( ']' ); - if (index != 0) { - if (index > 0) - CHR( '+' ); - SID( index ); - } - CHR( ']' ); -} - -static void -_dump_writemask( - uint writemask ) -{ - if (writemask != TGSI_WRITEMASK_XYZW) { - CHR( '.' ); - if (writemask & TGSI_WRITEMASK_X) - CHR( 'x' ); - if (writemask & TGSI_WRITEMASK_Y) - CHR( 'y' ); - if (writemask & TGSI_WRITEMASK_Z) - CHR( 'z' ); - if (writemask & TGSI_WRITEMASK_W) - CHR( 'w' ); - } -} - -void -tgsi_dump_declaration( - const struct tgsi_full_declaration *decl ) -{ - TXT( "\nDCL " ); - - _dump_register( - decl->Declaration.File, - decl->DeclarationRange.First, - decl->DeclarationRange.Last ); - _dump_writemask( - decl->Declaration.UsageMask ); - - if (decl->Declaration.Semantic) { - TXT( ", " ); - ENM( decl->Semantic.SemanticName, semantic_names ); - if (decl->Semantic.SemanticIndex != 0 || - decl->Semantic.SemanticName == TGSI_SEMANTIC_GENERIC) { - CHR( '[' ); - UID( decl->Semantic.SemanticIndex ); - CHR( ']' ); - } - } - - TXT( ", " ); - ENM( decl->Declaration.Interpolate, interpolate_names ); -} - -static boolean -iter_declaration( - struct tgsi_iterate_context *iter, - struct tgsi_full_declaration *decl ) -{ - tgsi_dump_declaration( decl ); - return TRUE; -} - -void -tgsi_dump_immediate( - const struct tgsi_full_immediate *imm ) -{ - uint i; - - TXT( "\nIMM " ); - ENM( imm->Immediate.DataType, immediate_type_names ); - - TXT( " { " ); - for (i = 0; i < imm->Immediate.Size - 1; i++) { - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - default: - assert( 0 ); - } - - if (i < imm->Immediate.Size - 2) - TXT( ", " ); - } - TXT( " }" ); -} - -static boolean -iter_immediate( - struct tgsi_iterate_context *iter, - struct tgsi_full_immediate *imm ) -{ - tgsi_dump_immediate( imm ); - return TRUE; -} - -void -tgsi_dump_instruction( - const struct tgsi_full_instruction *inst, - uint instno ) -{ - uint i; - boolean first_reg = TRUE; - - EOL(); - UID( instno ); - CHR( ':' ); - ENM( inst->Instruction.Opcode, opcode_names ); - - switch (inst->Instruction.Saturate) { - case TGSI_SAT_NONE: - break; - case TGSI_SAT_ZERO_ONE: - TXT( "_SAT" ); - break; - case TGSI_SAT_MINUS_PLUS_ONE: - TXT( "_SATNV" ); - break; - default: - assert( 0 ); - } - - for (i = 0; i < inst->Instruction.NumDstRegs; i++) { - const struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - - if (!first_reg) - CHR( ',' ); - CHR( ' ' ); - - _dump_register( - dst->DstRegister.File, - dst->DstRegister.Index, - dst->DstRegister.Index ); - ENM( dst->DstRegisterExtModulate.Modulate, modulate_names ); - _dump_writemask( dst->DstRegister.WriteMask ); - - first_reg = FALSE; - } - - for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { - const struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - - if (!first_reg) - CHR( ',' ); - CHR( ' ' ); - - if (src->SrcRegisterExtMod.Negate) - TXT( "-(" ); - if (src->SrcRegisterExtMod.Absolute) - CHR( '|' ); - if (src->SrcRegisterExtMod.Scale2X) - TXT( "2*(" ); - if (src->SrcRegisterExtMod.Bias) - CHR( '(' ); - if (src->SrcRegisterExtMod.Complement) - TXT( "1-(" ); - if (src->SrcRegister.Negate) - CHR( '-' ); - - if (src->SrcRegister.Indirect) { - _dump_register_ind( - src->SrcRegister.File, - src->SrcRegister.Index, - src->SrcRegisterInd.File, - src->SrcRegisterInd.Index ); - } - else { - _dump_register( - src->SrcRegister.File, - src->SrcRegister.Index, - src->SrcRegister.Index ); - } - - if (src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || - src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || - src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || - src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W) { - CHR( '.' ); - ENM( src->SrcRegister.SwizzleX, swizzle_names ); - ENM( src->SrcRegister.SwizzleY, swizzle_names ); - ENM( src->SrcRegister.SwizzleZ, swizzle_names ); - ENM( src->SrcRegister.SwizzleW, swizzle_names ); - } - if (src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || - src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || - src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || - src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W) { - CHR( '.' ); - if (src->SrcRegisterExtSwz.NegateX) - TXT("-"); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, extswizzle_names ); - if (src->SrcRegisterExtSwz.NegateY) - TXT("-"); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, extswizzle_names ); - if (src->SrcRegisterExtSwz.NegateZ) - TXT("-"); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, extswizzle_names ); - if (src->SrcRegisterExtSwz.NegateW) - TXT("-"); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, extswizzle_names ); - } - - if (src->SrcRegisterExtMod.Complement) - CHR( ')' ); - if (src->SrcRegisterExtMod.Bias) - TXT( ")-.5" ); - if (src->SrcRegisterExtMod.Scale2X) - CHR( ')' ); - if (src->SrcRegisterExtMod.Absolute) - CHR( '|' ); - if (src->SrcRegisterExtMod.Negate) - CHR( ')' ); - - first_reg = FALSE; - } - - if (inst->InstructionExtTexture.Texture != TGSI_TEXTURE_UNKNOWN) { - TXT( ", " ); - ENM( inst->InstructionExtTexture.Texture, texture_names ); - } - - switch (inst->Instruction.Opcode) { - case TGSI_OPCODE_IF: - case TGSI_OPCODE_ELSE: - case TGSI_OPCODE_BGNLOOP2: - case TGSI_OPCODE_ENDLOOP2: - case TGSI_OPCODE_CAL: - TXT( " :" ); - UID( inst->InstructionExtLabel.Label ); - break; - } -} - -static boolean -iter_instruction( - struct tgsi_iterate_context *iter, - struct tgsi_full_instruction *inst ) -{ - struct dump_ctx *ctx = (struct dump_ctx *) iter; - - tgsi_dump_instruction( inst, ctx->instno++ ); - return TRUE; -} - -static boolean -prolog( - struct tgsi_iterate_context *ctx ) -{ - EOL(); - ENM( ctx->processor.Processor, processor_type_names ); - UID( ctx->version.MajorVersion ); - CHR( '.' ); - UID( ctx->version.MinorVersion ); - return TRUE; -} - -void -tgsi_dump( - const struct tgsi_token *tokens, - uint flags ) -{ - struct dump_ctx ctx; - - /* sanity checks */ - assert( strcmp( opcode_names[TGSI_OPCODE_CONT], "CONT" ) == 0 ); - assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); - - ctx.iter.prolog = prolog; - ctx.iter.iterate_instruction = iter_instruction; - ctx.iter.iterate_declaration = iter_declaration; - ctx.iter.iterate_immediate = iter_immediate; - ctx.iter.epilog = NULL; - - ctx.instno = 0; - - tgsi_iterate_shader( tokens, &ctx.iter ); -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump.h deleted file mode 100644 index 51c230b5db..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump.h +++ /dev/null @@ -1,63 +0,0 @@ -/************************************************************************** - * - * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_DUMP_H -#define TGSI_DUMP_H - -#include "pipe/p_shader_tokens.h" - -#if defined __cplusplus -extern "C" { -#endif - -void -tgsi_dump( - const struct tgsi_token *tokens, - uint flags ); - -struct tgsi_full_immediate; -struct tgsi_full_instruction; -struct tgsi_full_declaration; - -void -tgsi_dump_immediate( - const struct tgsi_full_immediate *imm ); - -void -tgsi_dump_instruction( - const struct tgsi_full_instruction *inst, - uint instno ); - -void -tgsi_dump_declaration( - const struct tgsi_full_declaration *decl ); - -#if defined __cplusplus -} -#endif - -#endif /* TGSI_DUMP_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c deleted file mode 100644 index eabd74bd6d..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.c +++ /dev/null @@ -1,845 +0,0 @@ -/************************************************************************** - * - * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "util/u_string.h" -#include "tgsi_dump_c.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" - -static void -dump_enum( - const unsigned e, - const char **enums, - const unsigned enums_count ) -{ - if (e >= enums_count) { - debug_printf( "%u", e ); - } - else { - debug_printf( "%s", enums[e] ); - } -} - -#define EOL() debug_printf( "\n" ) -#define TXT(S) debug_printf( "%s", S ) -#define CHR(C) debug_printf( "%c", C ) -#define UIX(I) debug_printf( "0x%x", I ) -#define UID(I) debug_printf( "%u", I ) -#define SID(I) debug_printf( "%d", I ) -#define FLT(F) debug_printf( "%10.4f", F ) -#define ENM(E,ENUMS) dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) - -static const char *TGSI_PROCESSOR_TYPES[] = -{ - "PROCESSOR_FRAGMENT", - "PROCESSOR_VERTEX", - "PROCESSOR_GEOMETRY" -}; - -static const char *TGSI_TOKEN_TYPES[] = -{ - "TOKEN_TYPE_DECLARATION", - "TOKEN_TYPE_IMMEDIATE", - "TOKEN_TYPE_INSTRUCTION" -}; - -static const char *TGSI_FILES[] = -{ - "FILE_NULL", - "FILE_CONSTANT", - "FILE_INPUT", - "FILE_OUTPUT", - "FILE_TEMPORARY", - "FILE_SAMPLER", - "FILE_ADDRESS", - "FILE_IMMEDIATE" -}; - -static const char *TGSI_INTERPOLATES[] = -{ - "INTERPOLATE_CONSTANT", - "INTERPOLATE_LINEAR", - "INTERPOLATE_PERSPECTIVE" -}; - -static const char *TGSI_SEMANTICS[] = -{ - "SEMANTIC_POSITION", - "SEMANTIC_COLOR", - "SEMANTIC_BCOLOR", - "SEMANTIC_FOG", - "SEMANTIC_PSIZE", - "SEMANTIC_GENERIC", - "SEMANTIC_NORMAL" -}; - -static const char *TGSI_IMMS[] = -{ - "IMM_FLOAT32" -}; - -static const char *TGSI_OPCODES[TGSI_OPCODE_LAST] = -{ - "OPCODE_ARL", - "OPCODE_MOV", - "OPCODE_LIT", - "OPCODE_RCP", - "OPCODE_RSQ", - "OPCODE_EXP", - "OPCODE_LOG", - "OPCODE_MUL", - "OPCODE_ADD", - "OPCODE_DP3", - "OPCODE_DP4", - "OPCODE_DST", - "OPCODE_MIN", - "OPCODE_MAX", - "OPCODE_SLT", - "OPCODE_SGE", - "OPCODE_MAD", - "OPCODE_SUB", - "OPCODE_LERP", - "OPCODE_CND", - "OPCODE_CND0", - "OPCODE_DOT2ADD", - "OPCODE_INDEX", - "OPCODE_NEGATE", - "OPCODE_FRAC", - "OPCODE_CLAMP", - "OPCODE_FLOOR", - "OPCODE_ROUND", - "OPCODE_EXPBASE2", - "OPCODE_LOGBASE2", - "OPCODE_POWER", - "OPCODE_CROSSPRODUCT", - "OPCODE_MULTIPLYMATRIX", - "OPCODE_ABS", - "OPCODE_RCC", - "OPCODE_DPH", - "OPCODE_COS", - "OPCODE_DDX", - "OPCODE_DDY", - "OPCODE_KILP", - "OPCODE_PK2H", - "OPCODE_PK2US", - "OPCODE_PK4B", - "OPCODE_PK4UB", - "OPCODE_RFL", - "OPCODE_SEQ", - "OPCODE_SFL", - "OPCODE_SGT", - "OPCODE_SIN", - "OPCODE_SLE", - "OPCODE_SNE", - "OPCODE_STR", - "OPCODE_TEX", - "OPCODE_TXD", - "OPCODE_TXP", - "OPCODE_UP2H", - "OPCODE_UP2US", - "OPCODE_UP4B", - "OPCODE_UP4UB", - "OPCODE_X2D", - "OPCODE_ARA", - "OPCODE_ARR", - "OPCODE_BRA", - "OPCODE_CAL", - "OPCODE_RET", - "OPCODE_SSG", - "OPCODE_CMP", - "OPCODE_SCS", - "OPCODE_TXB", - "OPCODE_NRM", - "OPCODE_DIV", - "OPCODE_DP2", - "OPCODE_TXL", - "OPCODE_BRK", - "OPCODE_IF", - "OPCODE_LOOP", - "OPCODE_REP", - "OPCODE_ELSE", - "OPCODE_ENDIF", - "OPCODE_ENDLOOP", - "OPCODE_ENDREP", - "OPCODE_PUSHA", - "OPCODE_POPA", - "OPCODE_CEIL", - "OPCODE_I2F", - "OPCODE_NOT", - "OPCODE_TRUNC", - "OPCODE_SHL", - "OPCODE_SHR", - "OPCODE_AND", - "OPCODE_OR", - "OPCODE_MOD", - "OPCODE_XOR", - "OPCODE_SAD", - "OPCODE_TXF", - "OPCODE_TXQ", - "OPCODE_CONT", - "OPCODE_EMIT", - "OPCODE_ENDPRIM", - "OPCODE_BGNLOOP2", - "OPCODE_BGNSUB", - "OPCODE_ENDLOOP2", - "OPCODE_ENDSUB", - "OPCODE_NOISE1", - "OPCODE_NOISE2", - "OPCODE_NOISE3", - "OPCODE_NOISE4", - "OPCODE_NOP", - "OPCODE_M4X3", - "OPCODE_M3X4", - "OPCODE_M3X3", - "OPCODE_M3X2", - "OPCODE_NRM4", - "OPCODE_CALLNZ", - "OPCODE_IFC", - "OPCODE_BREAKC", - "OPCODE_KIL", - "OPCODE_END" -}; - -static const char *TGSI_SATS[] = -{ - "SAT_NONE", - "SAT_ZERO_ONE", - "SAT_MINUS_PLUS_ONE" -}; - -static const char *TGSI_INSTRUCTION_EXTS[] = -{ - "INSTRUCTION_EXT_TYPE_NV", - "INSTRUCTION_EXT_TYPE_LABEL", - "INSTRUCTION_EXT_TYPE_TEXTURE" -}; - -static const char *TGSI_PRECISIONS[] = -{ - "PRECISION_DEFAULT", - "PRECISION_FLOAT32", - "PRECISION_FLOAT16", - "PRECISION_FIXED12" -}; - -static const char *TGSI_CCS[] = -{ - "CC_GT", - "CC_EQ", - "CC_LT", - "CC_UN", - "CC_GE", - "CC_LE", - "CC_NE", - "CC_TR", - "CC_FL" -}; - -static const char *TGSI_SWIZZLES[] = -{ - "SWIZZLE_X", - "SWIZZLE_Y", - "SWIZZLE_Z", - "SWIZZLE_W" -}; - -static const char *TGSI_TEXTURES[] = -{ - "TEXTURE_UNKNOWN", - "TEXTURE_1D", - "TEXTURE_2D", - "TEXTURE_3D", - "TEXTURE_CUBE", - "TEXTURE_RECT", - "TEXTURE_SHADOW1D", - "TEXTURE_SHADOW2D", - "TEXTURE_SHADOWRECT" -}; - -static const char *TGSI_SRC_REGISTER_EXTS[] = -{ - "SRC_REGISTER_EXT_TYPE_SWZ", - "SRC_REGISTER_EXT_TYPE_MOD" -}; - -static const char *TGSI_EXTSWIZZLES[] = -{ - "EXTSWIZZLE_X", - "EXTSWIZZLE_Y", - "EXTSWIZZLE_Z", - "EXTSWIZZLE_W", - "EXTSWIZZLE_ZERO", - "EXTSWIZZLE_ONE" -}; - -static const char *TGSI_WRITEMASKS[] = -{ - "0", - "WRITEMASK_X", - "WRITEMASK_Y", - "WRITEMASK_XY", - "WRITEMASK_Z", - "WRITEMASK_XZ", - "WRITEMASK_YZ", - "WRITEMASK_XYZ", - "WRITEMASK_W", - "WRITEMASK_XW", - "WRITEMASK_YW", - "WRITEMASK_XYW", - "WRITEMASK_ZW", - "WRITEMASK_XZW", - "WRITEMASK_YZW", - "WRITEMASK_XYZW" -}; - -static const char *TGSI_DST_REGISTER_EXTS[] = -{ - "DST_REGISTER_EXT_TYPE_CONDCODE", - "DST_REGISTER_EXT_TYPE_MODULATE" -}; - -static const char *TGSI_MODULATES[] = -{ - "MODULATE_1X", - "MODULATE_2X", - "MODULATE_4X", - "MODULATE_8X", - "MODULATE_HALF", - "MODULATE_QUARTER", - "MODULATE_EIGHTH" -}; - -static void -dump_declaration_verbose( - struct tgsi_full_declaration *decl, - unsigned ignored, - unsigned deflt, - struct tgsi_full_declaration *fd ) -{ - TXT( "\nFile : " ); - ENM( decl->Declaration.File, TGSI_FILES ); - if( deflt || fd->Declaration.UsageMask != decl->Declaration.UsageMask ) { - TXT( "\nUsageMask : " ); - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_X ) { - CHR( 'X' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Y ) { - CHR( 'Y' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_Z ) { - CHR( 'Z' ); - } - if( decl->Declaration.UsageMask & TGSI_WRITEMASK_W ) { - CHR( 'W' ); - } - } - if( deflt || fd->Declaration.Interpolate != decl->Declaration.Interpolate ) { - TXT( "\nInterpolate: " ); - ENM( decl->Declaration.Interpolate, TGSI_INTERPOLATES ); - } - if( deflt || fd->Declaration.Semantic != decl->Declaration.Semantic ) { - TXT( "\nSemantic : " ); - UID( decl->Declaration.Semantic ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Declaration.Padding ); - } - - EOL(); - TXT( "\nFirst: " ); - UID( decl->DeclarationRange.First ); - TXT( "\nLast : " ); - UID( decl->DeclarationRange.Last ); - - if( decl->Declaration.Semantic ) { - EOL(); - TXT( "\nSemanticName : " ); - ENM( decl->Semantic.SemanticName, TGSI_SEMANTICS ); - TXT( "\nSemanticIndex: " ); - UID( decl->Semantic.SemanticIndex ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( decl->Semantic.Padding ); - } - } -} - -static void -dump_immediate_verbose( - struct tgsi_full_immediate *imm, - unsigned ignored ) -{ - unsigned i; - - TXT( "\nDataType : " ); - ENM( imm->Immediate.DataType, TGSI_IMMS ); - if( ignored ) { - TXT( "\nPadding : " ); - UIX( imm->Immediate.Padding ); - } - - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - EOL(); - switch( imm->Immediate.DataType ) { - case TGSI_IMM_FLOAT32: - TXT( "\nFloat: " ); - FLT( imm->u.ImmediateFloat32[i].Float ); - break; - - default: - assert( 0 ); - } - } -} - -static void -dump_instruction_verbose( - struct tgsi_full_instruction *inst, - unsigned ignored, - unsigned deflt, - struct tgsi_full_instruction *fi ) -{ - unsigned i; - - TXT( "\nOpcode : " ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES ); - if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { - TXT( "\nSaturate : " ); - ENM( inst->Instruction.Saturate, TGSI_SATS ); - } - if( deflt || fi->Instruction.NumDstRegs != inst->Instruction.NumDstRegs ) { - TXT( "\nNumDstRegs : " ); - UID( inst->Instruction.NumDstRegs ); - } - if( deflt || fi->Instruction.NumSrcRegs != inst->Instruction.NumSrcRegs ) { - TXT( "\nNumSrcRegs : " ); - UID( inst->Instruction.NumSrcRegs ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->Instruction.Padding ); - } - - if( deflt || tgsi_compare_instruction_ext_nv( inst->InstructionExtNv, fi->InstructionExtNv ) ) { - EOL(); - TXT( "\nType : " ); - ENM( inst->InstructionExtNv.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtNv.Precision != inst->InstructionExtNv.Precision ) { - TXT( "\nPrecision : " ); - ENM( inst->InstructionExtNv.Precision, TGSI_PRECISIONS ); - } - if( deflt || fi->InstructionExtNv.CondDstIndex != inst->InstructionExtNv.CondDstIndex ) { - TXT( "\nCondDstIndex : " ); - UID( inst->InstructionExtNv.CondDstIndex ); - } - if( deflt || fi->InstructionExtNv.CondFlowIndex != inst->InstructionExtNv.CondFlowIndex ) { - TXT( "\nCondFlowIndex : " ); - UID( inst->InstructionExtNv.CondFlowIndex ); - } - if( deflt || fi->InstructionExtNv.CondMask != inst->InstructionExtNv.CondMask ) { - TXT( "\nCondMask : " ); - ENM( inst->InstructionExtNv.CondMask, TGSI_CCS ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleX != inst->InstructionExtNv.CondSwizzleX ) { - TXT( "\nCondSwizzleX : " ); - ENM( inst->InstructionExtNv.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleY != inst->InstructionExtNv.CondSwizzleY ) { - TXT( "\nCondSwizzleY : " ); - ENM( inst->InstructionExtNv.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleZ != inst->InstructionExtNv.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ : " ); - ENM( inst->InstructionExtNv.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondSwizzleW != inst->InstructionExtNv.CondSwizzleW ) { - TXT( "\nCondSwizzleW : " ); - ENM( inst->InstructionExtNv.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fi->InstructionExtNv.CondDstUpdate != inst->InstructionExtNv.CondDstUpdate ) { - TXT( "\nCondDstUpdate : " ); - UID( inst->InstructionExtNv.CondDstUpdate ); - } - if( deflt || fi->InstructionExtNv.CondFlowEnable != inst->InstructionExtNv.CondFlowEnable ) { - TXT( "\nCondFlowEnable: " ); - UID( inst->InstructionExtNv.CondFlowEnable ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtNv.Padding ); - if( deflt || fi->InstructionExtNv.Extended != inst->InstructionExtNv.Extended ) { - TXT( "\nExtended : " ); - UID( inst->InstructionExtNv.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_label( inst->InstructionExtLabel, fi->InstructionExtLabel ) ) { - EOL(); - TXT( "\nType : " ); - ENM( inst->InstructionExtLabel.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtLabel.Label != inst->InstructionExtLabel.Label ) { - TXT( "\nLabel : " ); - UID( inst->InstructionExtLabel.Label ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtLabel.Padding ); - if( deflt || fi->InstructionExtLabel.Extended != inst->InstructionExtLabel.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtLabel.Extended ); - } - } - } - - if( deflt || tgsi_compare_instruction_ext_texture( inst->InstructionExtTexture, fi->InstructionExtTexture ) ) { - EOL(); - TXT( "\nType : " ); - ENM( inst->InstructionExtTexture.Type, TGSI_INSTRUCTION_EXTS ); - if( deflt || fi->InstructionExtTexture.Texture != inst->InstructionExtTexture.Texture ) { - TXT( "\nTexture : " ); - ENM( inst->InstructionExtTexture.Texture, TGSI_TEXTURES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( inst->InstructionExtTexture.Padding ); - if( deflt || fi->InstructionExtTexture.Extended != inst->InstructionExtTexture.Extended ) { - TXT( "\nExtended: " ); - UID( inst->InstructionExtTexture.Extended ); - } - } - } - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - struct tgsi_full_dst_register *dst = &inst->FullDstRegisters[i]; - struct tgsi_full_dst_register *fd = &fi->FullDstRegisters[i]; - - EOL(); - TXT( "\nFile : " ); - ENM( dst->DstRegister.File, TGSI_FILES ); - if( deflt || fd->DstRegister.WriteMask != dst->DstRegister.WriteMask ) { - TXT( "\nWriteMask: " ); - ENM( dst->DstRegister.WriteMask, TGSI_WRITEMASKS ); - } - if( ignored ) { - if( deflt || fd->DstRegister.Indirect != dst->DstRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( dst->DstRegister.Indirect ); - } - if( deflt || fd->DstRegister.Dimension != dst->DstRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( dst->DstRegister.Dimension ); - } - } - if( deflt || fd->DstRegister.Index != dst->DstRegister.Index ) { - TXT( "\nIndex : " ); - SID( dst->DstRegister.Index ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegister.Padding ); - if( deflt || fd->DstRegister.Extended != dst->DstRegister.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegister.Extended ); - } - } - - if( deflt || tgsi_compare_dst_register_ext_concode( dst->DstRegisterExtConcode, fd->DstRegisterExtConcode ) ) { - EOL(); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtConcode.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtConcode.CondMask != dst->DstRegisterExtConcode.CondMask ) { - TXT( "\nCondMask : " ); - ENM( dst->DstRegisterExtConcode.CondMask, TGSI_CCS ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleX != dst->DstRegisterExtConcode.CondSwizzleX ) { - TXT( "\nCondSwizzleX: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleY != dst->DstRegisterExtConcode.CondSwizzleY ) { - TXT( "\nCondSwizzleY: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleZ != dst->DstRegisterExtConcode.CondSwizzleZ ) { - TXT( "\nCondSwizzleZ: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSwizzleW != dst->DstRegisterExtConcode.CondSwizzleW ) { - TXT( "\nCondSwizzleW: " ); - ENM( dst->DstRegisterExtConcode.CondSwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fd->DstRegisterExtConcode.CondSrcIndex != dst->DstRegisterExtConcode.CondSrcIndex ) { - TXT( "\nCondSrcIndex: " ); - UID( dst->DstRegisterExtConcode.CondSrcIndex ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtConcode.Padding ); - if( deflt || fd->DstRegisterExtConcode.Extended != dst->DstRegisterExtConcode.Extended ) { - TXT( "\nExtended : " ); - UID( dst->DstRegisterExtConcode.Extended ); - } - } - } - - if( deflt || tgsi_compare_dst_register_ext_modulate( dst->DstRegisterExtModulate, fd->DstRegisterExtModulate ) ) { - EOL(); - TXT( "\nType : " ); - ENM( dst->DstRegisterExtModulate.Type, TGSI_DST_REGISTER_EXTS ); - if( deflt || fd->DstRegisterExtModulate.Modulate != dst->DstRegisterExtModulate.Modulate ) { - TXT( "\nModulate: " ); - ENM( dst->DstRegisterExtModulate.Modulate, TGSI_MODULATES ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( dst->DstRegisterExtModulate.Padding ); - if( deflt || fd->DstRegisterExtModulate.Extended != dst->DstRegisterExtModulate.Extended ) { - TXT( "\nExtended: " ); - UID( dst->DstRegisterExtModulate.Extended ); - } - } - } - } - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - struct tgsi_full_src_register *src = &inst->FullSrcRegisters[i]; - struct tgsi_full_src_register *fs = &fi->FullSrcRegisters[i]; - - EOL(); - TXT( "\nFile : "); - ENM( src->SrcRegister.File, TGSI_FILES ); - if( deflt || fs->SrcRegister.SwizzleX != src->SrcRegister.SwizzleX ) { - TXT( "\nSwizzleX : " ); - ENM( src->SrcRegister.SwizzleX, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleY != src->SrcRegister.SwizzleY ) { - TXT( "\nSwizzleY : " ); - ENM( src->SrcRegister.SwizzleY, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleZ != src->SrcRegister.SwizzleZ ) { - TXT( "\nSwizzleZ : " ); - ENM( src->SrcRegister.SwizzleZ, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.SwizzleW != src->SrcRegister.SwizzleW ) { - TXT( "\nSwizzleW : " ); - ENM( src->SrcRegister.SwizzleW, TGSI_SWIZZLES ); - } - if( deflt || fs->SrcRegister.Negate != src->SrcRegister.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegister.Negate ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Indirect != src->SrcRegister.Indirect ) { - TXT( "\nIndirect : " ); - UID( src->SrcRegister.Indirect ); - } - if( deflt || fs->SrcRegister.Dimension != src->SrcRegister.Dimension ) { - TXT( "\nDimension: " ); - UID( src->SrcRegister.Dimension ); - } - } - if( deflt || fs->SrcRegister.Index != src->SrcRegister.Index ) { - TXT( "\nIndex : " ); - SID( src->SrcRegister.Index ); - } - if( ignored ) { - if( deflt || fs->SrcRegister.Extended != src->SrcRegister.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegister.Extended ); - } - } - - if( deflt || tgsi_compare_src_register_ext_swz( src->SrcRegisterExtSwz, fs->SrcRegisterExtSwz ) ) { - EOL(); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtSwz.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleX != src->SrcRegisterExtSwz.ExtSwizzleX ) { - TXT( "\nExtSwizzleX: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleX, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleY != src->SrcRegisterExtSwz.ExtSwizzleY ) { - TXT( "\nExtSwizzleY: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleY, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleZ != src->SrcRegisterExtSwz.ExtSwizzleZ ) { - TXT( "\nExtSwizzleZ: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleZ, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.ExtSwizzleW != src->SrcRegisterExtSwz.ExtSwizzleW ) { - TXT( "\nExtSwizzleW: " ); - ENM( src->SrcRegisterExtSwz.ExtSwizzleW, TGSI_EXTSWIZZLES ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateX != src->SrcRegisterExtSwz.NegateX ) { - TXT( "\nNegateX : " ); - UID( src->SrcRegisterExtSwz.NegateX ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateY != src->SrcRegisterExtSwz.NegateY ) { - TXT( "\nNegateY : " ); - UID( src->SrcRegisterExtSwz.NegateY ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateZ != src->SrcRegisterExtSwz.NegateZ ) { - TXT( "\nNegateZ : " ); - UID( src->SrcRegisterExtSwz.NegateZ ); - } - if( deflt || fs->SrcRegisterExtSwz.NegateW != src->SrcRegisterExtSwz.NegateW ) { - TXT( "\nNegateW : " ); - UID( src->SrcRegisterExtSwz.NegateW ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtSwz.Padding ); - if( deflt || fs->SrcRegisterExtSwz.Extended != src->SrcRegisterExtSwz.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtSwz.Extended ); - } - } - } - - if( deflt || tgsi_compare_src_register_ext_mod( src->SrcRegisterExtMod, fs->SrcRegisterExtMod ) ) { - EOL(); - TXT( "\nType : " ); - ENM( src->SrcRegisterExtMod.Type, TGSI_SRC_REGISTER_EXTS ); - if( deflt || fs->SrcRegisterExtMod.Complement != src->SrcRegisterExtMod.Complement ) { - TXT( "\nComplement: " ); - UID( src->SrcRegisterExtMod.Complement ); - } - if( deflt || fs->SrcRegisterExtMod.Bias != src->SrcRegisterExtMod.Bias ) { - TXT( "\nBias : " ); - UID( src->SrcRegisterExtMod.Bias ); - } - if( deflt || fs->SrcRegisterExtMod.Scale2X != src->SrcRegisterExtMod.Scale2X ) { - TXT( "\nScale2X : " ); - UID( src->SrcRegisterExtMod.Scale2X ); - } - if( deflt || fs->SrcRegisterExtMod.Absolute != src->SrcRegisterExtMod.Absolute ) { - TXT( "\nAbsolute : " ); - UID( src->SrcRegisterExtMod.Absolute ); - } - if( deflt || fs->SrcRegisterExtMod.Negate != src->SrcRegisterExtMod.Negate ) { - TXT( "\nNegate : " ); - UID( src->SrcRegisterExtMod.Negate ); - } - if( ignored ) { - TXT( "\nPadding : " ); - UIX( src->SrcRegisterExtMod.Padding ); - if( deflt || fs->SrcRegisterExtMod.Extended != src->SrcRegisterExtMod.Extended ) { - TXT( "\nExtended : " ); - UID( src->SrcRegisterExtMod.Extended ); - } - } - } - } -} - -void -tgsi_dump_c( - const struct tgsi_token *tokens, - uint flags ) -{ - struct tgsi_parse_context parse; - struct tgsi_full_instruction fi; - struct tgsi_full_declaration fd; - uint ignored = flags & TGSI_DUMP_C_IGNORED; - uint deflt = flags & TGSI_DUMP_C_DEFAULT; - uint instno = 0; - - /* sanity checks */ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); - - tgsi_parse_init( &parse, tokens ); - - TXT( "tgsi-dump begin -----------------" ); - - TXT( "\nMajorVersion: " ); - UID( parse.FullVersion.Version.MajorVersion ); - TXT( "\nMinorVersion: " ); - UID( parse.FullVersion.Version.MinorVersion ); - EOL(); - - TXT( "\nHeaderSize: " ); - UID( parse.FullHeader.Header.HeaderSize ); - TXT( "\nBodySize : " ); - UID( parse.FullHeader.Header.BodySize ); - TXT( "\nProcessor : " ); - ENM( parse.FullHeader.Processor.Processor, TGSI_PROCESSOR_TYPES ); - EOL(); - - fi = tgsi_default_full_instruction(); - fd = tgsi_default_full_declaration(); - - while( !tgsi_parse_end_of_tokens( &parse ) ) { - tgsi_parse_token( &parse ); - - TXT( "\nType : " ); - ENM( parse.FullToken.Token.Type, TGSI_TOKEN_TYPES ); - if( ignored ) { - TXT( "\nSize : " ); - UID( parse.FullToken.Token.Size ); - if( deflt || parse.FullToken.Token.Extended ) { - TXT( "\nExtended : " ); - UID( parse.FullToken.Token.Extended ); - } - } - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - dump_declaration_verbose( - &parse.FullToken.FullDeclaration, - ignored, - deflt, - &fd ); - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - dump_immediate_verbose( - &parse.FullToken.FullImmediate, - ignored ); - break; - - case TGSI_TOKEN_TYPE_INSTRUCTION: - dump_instruction_verbose( - &parse.FullToken.FullInstruction, - ignored, - deflt, - &fi ); - break; - - default: - assert( 0 ); - } - - EOL(); - } - - TXT( "\ntgsi-dump end -------------------\n" ); - - tgsi_parse_free( &parse ); -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h b/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h deleted file mode 100644 index d91cd35b3b..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_dump_c.h +++ /dev/null @@ -1,49 +0,0 @@ -/************************************************************************** - * - * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_DUMP_C_H -#define TGSI_DUMP_C_H - -#include "pipe/p_shader_tokens.h" - -#if defined __cplusplus -extern "C" { -#endif - -#define TGSI_DUMP_C_IGNORED 1 -#define TGSI_DUMP_C_DEFAULT 2 - -void -tgsi_dump_c( - const struct tgsi_token *tokens, - uint flags ); - -#if defined __cplusplus -} -#endif - -#endif /* TGSI_DUMP_C_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c deleted file mode 100644 index 5371a88b96..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.c +++ /dev/null @@ -1,85 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "tgsi_iterate.h" - -boolean -tgsi_iterate_shader( - const struct tgsi_token *tokens, - struct tgsi_iterate_context *ctx ) -{ - struct tgsi_parse_context parse; - - if (tgsi_parse_init( &parse, tokens ) != TGSI_PARSE_OK) - return FALSE; - - ctx->processor = parse.FullHeader.Processor; - ctx->version = parse.FullVersion.Version; - - if (ctx->prolog) - if (!ctx->prolog( ctx )) - goto fail; - - while (!tgsi_parse_end_of_tokens( &parse )) { - tgsi_parse_token( &parse ); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_INSTRUCTION: - if (ctx->iterate_instruction) - if (!ctx->iterate_instruction( ctx, &parse.FullToken.FullInstruction )) - goto fail; - break; - - case TGSI_TOKEN_TYPE_DECLARATION: - if (ctx->iterate_declaration) - if (!ctx->iterate_declaration( ctx, &parse.FullToken.FullDeclaration )) - goto fail; - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - if (ctx->iterate_immediate) - if (!ctx->iterate_immediate( ctx, &parse.FullToken.FullImmediate )) - goto fail; - break; - - default: - assert( 0 ); - } - } - - if (ctx->epilog) - if (!ctx->epilog( ctx )) - goto fail; - - tgsi_parse_free( &parse ); - return TRUE; - -fail: - tgsi_parse_free( &parse ); - return FALSE; -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h b/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h deleted file mode 100644 index f5bebf89b8..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_iterate.h +++ /dev/null @@ -1,76 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_ITERATE_H -#define TGSI_ITERATE_H - -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" - -#if defined __cplusplus -extern "C" { -#endif - -struct tgsi_iterate_context -{ - boolean - (* prolog)( - struct tgsi_iterate_context *ctx ); - - boolean - (* iterate_instruction)( - struct tgsi_iterate_context *ctx, - struct tgsi_full_instruction *inst ); - - boolean - (* iterate_declaration)( - struct tgsi_iterate_context *ctx, - struct tgsi_full_declaration *decl ); - - boolean - (* iterate_immediate)( - struct tgsi_iterate_context *ctx, - struct tgsi_full_immediate *imm ); - - boolean - (* epilog)( - struct tgsi_iterate_context *ctx ); - - struct tgsi_processor processor; - struct tgsi_version version; -}; - -boolean -tgsi_iterate_shader( - const struct tgsi_token *tokens, - struct tgsi_iterate_context *ctx ); - -#if defined __cplusplus -} -#endif - -#endif /* TGSI_ITERATE_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c b/src/gallium/auxiliary/tgsi/util/tgsi_parse.c deleted file mode 100644 index d16f0cdcad..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.c +++ /dev/null @@ -1,332 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ) -{ - full_token->Token.Type = TGSI_TOKEN_TYPE_DECLARATION; -} - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ) -{ - if( full_token->Token.Type == TGSI_TOKEN_TYPE_IMMEDIATE ) { - FREE( (void *) full_token->FullImmediate.u.Pointer ); - } -} - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ) -{ - ctx->FullVersion.Version = *(struct tgsi_version *) &tokens[0]; - if( ctx->FullVersion.Version.MajorVersion > 1 ) { - return TGSI_PARSE_ERROR; - } - - ctx->FullHeader.Header = *(struct tgsi_header *) &tokens[1]; - if( ctx->FullHeader.Header.HeaderSize >= 2 ) { - ctx->FullHeader.Processor = *(struct tgsi_processor *) &tokens[2]; - } - else { - ctx->FullHeader.Processor = tgsi_default_processor(); - } - - ctx->Tokens = tokens; - ctx->Position = 1 + ctx->FullHeader.Header.HeaderSize; - - tgsi_full_token_init( &ctx->FullToken ); - - return TGSI_PARSE_OK; -} - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ) -{ - tgsi_full_token_free( &ctx->FullToken ); -} - -boolean -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ) -{ - return ctx->Position >= - 1 + ctx->FullHeader.Header.HeaderSize + ctx->FullHeader.Header.BodySize; -} - -static void -next_token( - struct tgsi_parse_context *ctx, - void *token ) -{ - assert( !tgsi_parse_end_of_tokens( ctx ) ); - - *(struct tgsi_token *) token = ctx->Tokens[ctx->Position++]; -} - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ) -{ - struct tgsi_token token; - unsigned i; - - tgsi_full_token_free( &ctx->FullToken ); - tgsi_full_token_init( &ctx->FullToken ); - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *decl = &ctx->FullToken.FullDeclaration; - - *decl = tgsi_default_full_declaration(); - decl->Declaration = *(struct tgsi_declaration *) &token; - - next_token( ctx, &decl->DeclarationRange ); - - if( decl->Declaration.Semantic ) { - next_token( ctx, &decl->Semantic ); - } - - break; - } - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - struct tgsi_full_immediate *imm = &ctx->FullToken.FullImmediate; - - *imm = tgsi_default_full_immediate(); - imm->Immediate = *(struct tgsi_immediate *) &token; - - assert( !imm->Immediate.Extended ); - - switch (imm->Immediate.DataType) { - case TGSI_IMM_FLOAT32: - imm->u.Pointer = MALLOC( - sizeof( struct tgsi_immediate_float32 ) * (imm->Immediate.Size - 1) ); - for( i = 0; i < imm->Immediate.Size - 1; i++ ) { - next_token( ctx, (struct tgsi_immediate_float32 *) &imm->u.ImmediateFloat32[i] ); - } - break; - - default: - assert( 0 ); - } - - break; - } - - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *inst = &ctx->FullToken.FullInstruction; - unsigned extended; - - *inst = tgsi_default_full_instruction(); - inst->Instruction = *(struct tgsi_instruction *) &token; - - extended = inst->Instruction.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_INSTRUCTION_EXT_TYPE_NV: - inst->InstructionExtNv = - *(struct tgsi_instruction_ext_nv *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_LABEL: - inst->InstructionExtLabel = - *(struct tgsi_instruction_ext_label *) &token; - break; - - case TGSI_INSTRUCTION_EXT_TYPE_TEXTURE: - inst->InstructionExtTexture = - *(struct tgsi_instruction_ext_texture *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - assert( inst->Instruction.NumDstRegs <= TGSI_FULL_MAX_DST_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumDstRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullDstRegisters[i].DstRegister ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullDstRegisters[i].DstRegister.Indirect ); - assert( !inst->FullDstRegisters[i].DstRegister.Dimension ); - - extended = inst->FullDstRegisters[i].DstRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_DST_REGISTER_EXT_TYPE_CONDCODE: - inst->FullDstRegisters[i].DstRegisterExtConcode = - *(struct tgsi_dst_register_ext_concode *) &token; - break; - - case TGSI_DST_REGISTER_EXT_TYPE_MODULATE: - inst->FullDstRegisters[i].DstRegisterExtModulate = - *(struct tgsi_dst_register_ext_modulate *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - } - - assert( inst->Instruction.NumSrcRegs <= TGSI_FULL_MAX_SRC_REGISTERS ); - - for( i = 0; i < inst->Instruction.NumSrcRegs; i++ ) { - unsigned extended; - - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegister ); - - extended = inst->FullSrcRegisters[i].SrcRegister.Extended; - - while( extended ) { - struct tgsi_src_register_ext token; - - next_token( ctx, &token ); - - switch( token.Type ) { - case TGSI_SRC_REGISTER_EXT_TYPE_SWZ: - inst->FullSrcRegisters[i].SrcRegisterExtSwz = - *(struct tgsi_src_register_ext_swz *) &token; - break; - - case TGSI_SRC_REGISTER_EXT_TYPE_MOD: - inst->FullSrcRegisters[i].SrcRegisterExtMod = - *(struct tgsi_src_register_ext_mod *) &token; - break; - - default: - assert( 0 ); - } - - extended = token.Extended; - } - - if( inst->FullSrcRegisters[i].SrcRegister.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - - if( inst->FullSrcRegisters[i].SrcRegister.Dimension ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDim ); - - /* - * No support for multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterDim.Extended ); - - if( inst->FullSrcRegisters[i].SrcRegisterDim.Indirect ) { - next_token( ctx, &inst->FullSrcRegisters[i].SrcRegisterDimInd ); - - /* - * No support for indirect or multi-dimensional addressing. - */ - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Indirect ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Dimension ); - assert( !inst->FullSrcRegisters[i].SrcRegisterInd.Extended ); - } - } - } - - break; - } - - default: - assert( 0 ); - } -} - - -unsigned -tgsi_num_tokens(const struct tgsi_token *tokens) -{ - struct tgsi_parse_context ctx; - if (tgsi_parse_init(&ctx, tokens) == TGSI_PARSE_OK) { - unsigned len = (ctx.FullHeader.Header.HeaderSize + - ctx.FullHeader.Header.BodySize + - 1); - return len; - } - return 0; -} - - -/** - * Make a new copy of a token array. - */ -struct tgsi_token * -tgsi_dup_tokens(const struct tgsi_token *tokens) -{ - unsigned n = tgsi_num_tokens(tokens); - unsigned bytes = n * sizeof(struct tgsi_token); - struct tgsi_token *new_tokens = (struct tgsi_token *) MALLOC(bytes); - if (new_tokens) - memcpy(new_tokens, tokens, bytes); - return new_tokens; -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h b/src/gallium/auxiliary/tgsi/util/tgsi_parse.h deleted file mode 100644 index 054350712d..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_parse.h +++ /dev/null @@ -1,151 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_PARSE_H -#define TGSI_PARSE_H - -#include "pipe/p_shader_tokens.h" - -#if defined __cplusplus -extern "C" { -#endif - -struct tgsi_full_version -{ - struct tgsi_version Version; -}; - -struct tgsi_full_header -{ - struct tgsi_header Header; - struct tgsi_processor Processor; -}; - -struct tgsi_full_dst_register -{ - struct tgsi_dst_register DstRegister; - struct tgsi_dst_register_ext_concode DstRegisterExtConcode; - struct tgsi_dst_register_ext_modulate DstRegisterExtModulate; -}; - -struct tgsi_full_src_register -{ - struct tgsi_src_register SrcRegister; - struct tgsi_src_register_ext_swz SrcRegisterExtSwz; - struct tgsi_src_register_ext_mod SrcRegisterExtMod; - struct tgsi_src_register SrcRegisterInd; - struct tgsi_dimension SrcRegisterDim; - struct tgsi_src_register SrcRegisterDimInd; -}; - -struct tgsi_full_declaration -{ - struct tgsi_declaration Declaration; - struct tgsi_declaration_range DeclarationRange; - struct tgsi_declaration_semantic Semantic; -}; - -struct tgsi_full_immediate -{ - struct tgsi_immediate Immediate; - union - { - const void *Pointer; - const struct tgsi_immediate_float32 *ImmediateFloat32; - } u; -}; - -#define TGSI_FULL_MAX_DST_REGISTERS 2 -#define TGSI_FULL_MAX_SRC_REGISTERS 4 /* TXD has 4 */ - -struct tgsi_full_instruction -{ - struct tgsi_instruction Instruction; - struct tgsi_instruction_ext_nv InstructionExtNv; - struct tgsi_instruction_ext_label InstructionExtLabel; - struct tgsi_instruction_ext_texture InstructionExtTexture; - struct tgsi_full_dst_register FullDstRegisters[TGSI_FULL_MAX_DST_REGISTERS]; - struct tgsi_full_src_register FullSrcRegisters[TGSI_FULL_MAX_SRC_REGISTERS]; -}; - -union tgsi_full_token -{ - struct tgsi_token Token; - struct tgsi_full_declaration FullDeclaration; - struct tgsi_full_immediate FullImmediate; - struct tgsi_full_instruction FullInstruction; -}; - -void -tgsi_full_token_init( - union tgsi_full_token *full_token ); - -void -tgsi_full_token_free( - union tgsi_full_token *full_token ); - -struct tgsi_parse_context -{ - const struct tgsi_token *Tokens; - unsigned Position; - struct tgsi_full_version FullVersion; - struct tgsi_full_header FullHeader; - union tgsi_full_token FullToken; -}; - -#define TGSI_PARSE_OK 0 -#define TGSI_PARSE_ERROR 1 - -unsigned -tgsi_parse_init( - struct tgsi_parse_context *ctx, - const struct tgsi_token *tokens ); - -void -tgsi_parse_free( - struct tgsi_parse_context *ctx ); - -boolean -tgsi_parse_end_of_tokens( - struct tgsi_parse_context *ctx ); - -void -tgsi_parse_token( - struct tgsi_parse_context *ctx ); - -unsigned -tgsi_num_tokens(const struct tgsi_token *tokens); - -struct tgsi_token * -tgsi_dup_tokens(const struct tgsi_token *tokens); - -#if defined __cplusplus -} -#endif - -#endif /* TGSI_PARSE_H */ - diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c deleted file mode 100644 index 2e3ec96b5b..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.c +++ /dev/null @@ -1,341 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "tgsi_sanity.h" -#include "tgsi_iterate.h" - -#define MAX_REGISTERS 256 - -typedef uint reg_flag; - -#define BITS_IN_REG_FLAG (sizeof( reg_flag ) * 8) - -struct sanity_check_ctx -{ - struct tgsi_iterate_context iter; - - reg_flag regs_decl[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG]; - reg_flag regs_used[TGSI_FILE_COUNT][MAX_REGISTERS / BITS_IN_REG_FLAG]; - boolean regs_ind_used[TGSI_FILE_COUNT]; - uint num_imms; - uint num_instructions; - uint index_of_END; - - uint errors; - uint warnings; -}; - -static void -report_error( - struct sanity_check_ctx *ctx, - const char *format, - ... ) -{ - va_list args; - - debug_printf( "Error : " ); - va_start( args, format ); - _debug_vprintf( format, args ); - va_end( args ); - debug_printf( "\n" ); - ctx->errors++; -} - -static void -report_warning( - struct sanity_check_ctx *ctx, - const char *format, - ... ) -{ - va_list args; - - debug_printf( "Warning: " ); - va_start( args, format ); - _debug_vprintf( format, args ); - va_end( args ); - debug_printf( "\n" ); - ctx->warnings++; -} - -static boolean -check_file_name( - struct sanity_check_ctx *ctx, - uint file ) -{ - if (file <= TGSI_FILE_NULL || file >= TGSI_FILE_COUNT) { - report_error( ctx, "Invalid register file name" ); - return FALSE; - } - return TRUE; -} - -static boolean -is_register_declared( - struct sanity_check_ctx *ctx, - uint file, - int index ) -{ - assert( index >= 0 && index < MAX_REGISTERS ); - - return (ctx->regs_decl[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; -} - -static boolean -is_any_register_declared( - struct sanity_check_ctx *ctx, - uint file ) -{ - uint i; - - for (i = 0; i < MAX_REGISTERS / BITS_IN_REG_FLAG; i++) - if (ctx->regs_decl[file][i]) - return TRUE; - return FALSE; -} - -static boolean -is_register_used( - struct sanity_check_ctx *ctx, - uint file, - int index ) -{ - assert( index < MAX_REGISTERS ); - - return (ctx->regs_used[file][index / BITS_IN_REG_FLAG] & (1 << (index % BITS_IN_REG_FLAG))) ? TRUE : FALSE; -} - -static const char *file_names[] = -{ - "NULL", - "CONST", - "IN", - "OUT", - "TEMP", - "SAMP", - "ADDR", - "IMM" -}; - -static boolean -check_register_usage( - struct sanity_check_ctx *ctx, - uint file, - int index, - const char *name, - boolean indirect_access ) -{ - if (!check_file_name( ctx, file )) - return FALSE; - if (indirect_access) { - if (!is_any_register_declared( ctx, file )) - report_error( ctx, "%s: Undeclared %s register", file_names[file], name ); - ctx->regs_ind_used[file] = TRUE; - } - else { - if (!is_register_declared( ctx, file, index )) - report_error( ctx, "%s[%d]: Undeclared %s register", file_names[file], index, name ); - ctx->regs_used[file][index / BITS_IN_REG_FLAG] |= (1 << (index % BITS_IN_REG_FLAG)); - } - return TRUE; -} - -static boolean -iter_instruction( - struct tgsi_iterate_context *iter, - struct tgsi_full_instruction *inst ) -{ - struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; - uint i; - - /* There must be no other instructions after END. - */ - if (ctx->index_of_END != ~0) { - report_error( ctx, "Unexpected instruction after END" ); - } - else if (inst->Instruction.Opcode == TGSI_OPCODE_END) { - ctx->index_of_END = ctx->num_instructions; - } - - /* Check destination and source registers' validity. - * Mark the registers as used. - */ - for (i = 0; i < inst->Instruction.NumDstRegs; i++) { - check_register_usage( - ctx, - inst->FullDstRegisters[i].DstRegister.File, - inst->FullDstRegisters[i].DstRegister.Index, - "destination", - FALSE ); - } - for (i = 0; i < inst->Instruction.NumSrcRegs; i++) { - check_register_usage( - ctx, - inst->FullSrcRegisters[i].SrcRegister.File, - inst->FullSrcRegisters[i].SrcRegister.Index, - "source", - (boolean)inst->FullSrcRegisters[i].SrcRegister.Indirect ); - if (inst->FullSrcRegisters[i].SrcRegister.Indirect) { - uint file; - int index; - - file = inst->FullSrcRegisters[i].SrcRegisterInd.File; - index = inst->FullSrcRegisters[i].SrcRegisterInd.Index; - check_register_usage( - ctx, - file, - index, - "indirect", - FALSE ); - if (file != TGSI_FILE_ADDRESS || index != 0) - report_warning( ctx, "Indirect register not ADDR[0]" ); - } - } - - ctx->num_instructions++; - - return TRUE; -} - -static boolean -iter_declaration( - struct tgsi_iterate_context *iter, - struct tgsi_full_declaration *decl ) -{ - struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; - uint file; - uint i; - - /* No declarations allowed after the first instruction. - */ - if (ctx->num_instructions > 0) - report_error( ctx, "Instruction expected but declaration found" ); - - /* Check registers' validity. - * Mark the registers as declared. - */ - file = decl->Declaration.File; - if (!check_file_name( ctx, file )) - return TRUE; - for (i = decl->DeclarationRange.First; i <= decl->DeclarationRange.Last; i++) { - if (is_register_declared( ctx, file, i )) - report_error( ctx, "The same register declared twice" ); - ctx->regs_decl[file][i / BITS_IN_REG_FLAG] |= (1 << (i % BITS_IN_REG_FLAG)); - } - - return TRUE; -} - -static boolean -iter_immediate( - struct tgsi_iterate_context *iter, - struct tgsi_full_immediate *imm ) -{ - struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; - - assert( ctx->num_imms < MAX_REGISTERS ); - - /* No immediates allowed after the first instruction. - */ - if (ctx->num_instructions > 0) - report_error( ctx, "Instruction expected but immediate found" ); - - /* Mark the register as declared. - */ - ctx->regs_decl[TGSI_FILE_IMMEDIATE][ctx->num_imms / BITS_IN_REG_FLAG] |= (1 << (ctx->num_imms % BITS_IN_REG_FLAG)); - ctx->num_imms++; - - /* Check data type validity. - */ - if (imm->Immediate.DataType != TGSI_IMM_FLOAT32) { - report_error( ctx, "Invalid immediate data type" ); - return TRUE; - } - - return TRUE; -} - -static boolean -epilog( - struct tgsi_iterate_context *iter ) -{ - struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; - uint file; - - /* There must be an END instruction at the end. - */ - if (ctx->index_of_END == ~0 || ctx->index_of_END != ctx->num_instructions - 1) { - report_error( ctx, "Expected END at end of instruction sequence" ); - } - - /* Check if all declared registers were used. - */ - for (file = TGSI_FILE_NULL; file < TGSI_FILE_COUNT; file++) { - uint i; - - for (i = 0; i < MAX_REGISTERS; i++) { - if (is_register_declared( ctx, file, i ) && !is_register_used( ctx, file, i ) && !ctx->regs_ind_used[file]) { - report_warning( ctx, "Register never used" ); - } - } - } - - /* Print totals, if any. - */ - if (ctx->errors || ctx->warnings) - debug_printf( "\n%u errors, %u warnings", ctx->errors, ctx->warnings ); - - return TRUE; -} - -boolean -tgsi_sanity_check( - struct tgsi_token *tokens ) -{ - struct sanity_check_ctx ctx; - - ctx.iter.prolog = NULL; - ctx.iter.iterate_instruction = iter_instruction; - ctx.iter.iterate_declaration = iter_declaration; - ctx.iter.iterate_immediate = iter_immediate; - ctx.iter.epilog = epilog; - - memset( ctx.regs_decl, 0, sizeof( ctx.regs_decl ) ); - memset( ctx.regs_used, 0, sizeof( ctx.regs_used ) ); - memset( ctx.regs_ind_used, 0, sizeof( ctx.regs_ind_used ) ); - ctx.num_imms = 0; - ctx.num_instructions = 0; - ctx.index_of_END = ~0; - - ctx.errors = 0; - ctx.warnings = 0; - - if (!tgsi_iterate_shader( tokens, &ctx.iter )) - return FALSE; - - return ctx.errors == 0; -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.h b/src/gallium/auxiliary/tgsi/util/tgsi_sanity.h deleted file mode 100644 index ca45e94c7a..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_sanity.h +++ /dev/null @@ -1,49 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_SANITY_H -#define TGSI_SANITY_H - -#include "pipe/p_shader_tokens.h" - -#if defined __cplusplus -extern "C" { -#endif - -/* Check the given token stream for errors and common mistakes. - * Diagnostic messages are printed out to the debug output. - * Returns TRUE if there are no errors, even though there could be some warnings. - */ -boolean -tgsi_sanity_check( - struct tgsi_token *tokens ); - -#if defined __cplusplus -} -#endif - -#endif /* TGSI_SANITY_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c b/src/gallium/auxiliary/tgsi/util/tgsi_scan.c deleted file mode 100644 index 240aaaf362..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.c +++ /dev/null @@ -1,226 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * TGSI program scan utility. - * Used to determine which registers and instructions are used by a shader. - * - * Authors: Brian Paul - */ - - -#include "tgsi_scan.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_build.h" - -#include "pipe/p_util.h" - - - -/** - */ -void -tgsi_scan_shader(const struct tgsi_token *tokens, - struct tgsi_shader_info *info) -{ - uint procType, i; - struct tgsi_parse_context parse; - - memset(info, 0, sizeof(*info)); - for (i = 0; i < TGSI_FILE_COUNT; i++) - info->file_max[i] = -1; - - /** - ** Setup to begin parsing input shader - **/ - if (tgsi_parse_init( &parse, tokens ) != TGSI_PARSE_OK) { - debug_printf("tgsi_parse_init() failed in tgsi_scan_shader()!\n"); - return; - } - procType = parse.FullHeader.Processor.Processor; - assert(procType == TGSI_PROCESSOR_FRAGMENT || - procType == TGSI_PROCESSOR_VERTEX || - procType == TGSI_PROCESSOR_GEOMETRY); - - - /** - ** Loop over incoming program tokens/instructions - */ - while( !tgsi_parse_end_of_tokens( &parse ) ) { - - info->num_tokens++; - - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *fullinst - = &parse.FullToken.FullInstruction; - - assert(fullinst->Instruction.Opcode < TGSI_OPCODE_LAST); - info->opcode_count[fullinst->Instruction.Opcode]++; - } - break; - - case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *fulldecl - = &parse.FullToken.FullDeclaration; - uint file = fulldecl->Declaration.File; - uint i; - for (i = fulldecl->DeclarationRange.First; - i <= fulldecl->DeclarationRange.Last; - i++) { - - /* only first 32 regs will appear in this bitfield */ - info->file_mask[file] |= (1 << i); - info->file_count[file]++; - info->file_max[file] = MAX2(info->file_max[file], (int)i); - - if (file == TGSI_FILE_INPUT) { - info->input_semantic_name[i] = (ubyte)fulldecl->Semantic.SemanticName; - info->input_semantic_index[i] = (ubyte)fulldecl->Semantic.SemanticIndex; - info->num_inputs++; - } - - if (file == TGSI_FILE_OUTPUT) { - info->output_semantic_name[i] = (ubyte)fulldecl->Semantic.SemanticName; - info->output_semantic_index[i] = (ubyte)fulldecl->Semantic.SemanticIndex; - info->num_outputs++; - } - - /* special case */ - if (procType == TGSI_PROCESSOR_FRAGMENT && - file == TGSI_FILE_OUTPUT && - fulldecl->Semantic.SemanticName == TGSI_SEMANTIC_POSITION) { - info->writes_z = TRUE; - } - } - } - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - info->immediate_count++; - break; - - default: - assert( 0 ); - } - } - - assert( info->file_max[TGSI_FILE_INPUT] + 1 == info->num_inputs ); - assert( info->file_max[TGSI_FILE_OUTPUT] + 1 == info->num_outputs ); - - info->uses_kill = (info->opcode_count[TGSI_OPCODE_KIL] || - info->opcode_count[TGSI_OPCODE_KILP]); - - tgsi_parse_free (&parse); -} - - - -/** - * Check if the given shader is a "passthrough" shader consisting of only - * MOV instructions of the form: MOV OUT[n], IN[n] - * - */ -boolean -tgsi_is_passthrough_shader(const struct tgsi_token *tokens) -{ - struct tgsi_parse_context parse; - - /** - ** Setup to begin parsing input shader - **/ - if (tgsi_parse_init(&parse, tokens) != TGSI_PARSE_OK) { - debug_printf("tgsi_parse_init() failed in tgsi_is_passthrough_shader()!\n"); - return FALSE; - } - - /** - ** Loop over incoming program tokens/instructions - */ - while (!tgsi_parse_end_of_tokens(&parse)) { - - tgsi_parse_token(&parse); - - switch (parse.FullToken.Token.Type) { - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *fullinst = - &parse.FullToken.FullInstruction; - const struct tgsi_full_src_register *src = - &fullinst->FullSrcRegisters[0]; - const struct tgsi_full_dst_register *dst = - &fullinst->FullDstRegisters[0]; - - /* Do a whole bunch of checks for a simple move */ - if (fullinst->Instruction.Opcode != TGSI_OPCODE_MOV || - src->SrcRegister.File != TGSI_FILE_INPUT || - dst->DstRegister.File != TGSI_FILE_OUTPUT || - src->SrcRegister.Index != dst->DstRegister.Index || - - src->SrcRegister.Negate || - src->SrcRegisterExtMod.Negate || - src->SrcRegisterExtMod.Absolute || - src->SrcRegisterExtMod.Scale2X || - src->SrcRegisterExtMod.Bias || - src->SrcRegisterExtMod.Complement || - - src->SrcRegister.SwizzleX != TGSI_SWIZZLE_X || - src->SrcRegister.SwizzleY != TGSI_SWIZZLE_Y || - src->SrcRegister.SwizzleZ != TGSI_SWIZZLE_Z || - src->SrcRegister.SwizzleW != TGSI_SWIZZLE_W || - - src->SrcRegisterExtSwz.ExtSwizzleX != TGSI_EXTSWIZZLE_X || - src->SrcRegisterExtSwz.ExtSwizzleY != TGSI_EXTSWIZZLE_Y || - src->SrcRegisterExtSwz.ExtSwizzleZ != TGSI_EXTSWIZZLE_Z || - src->SrcRegisterExtSwz.ExtSwizzleW != TGSI_EXTSWIZZLE_W || - - dst->DstRegister.WriteMask != TGSI_WRITEMASK_XYZW) - { - tgsi_parse_free(&parse); - return FALSE; - } - } - break; - - case TGSI_TOKEN_TYPE_DECLARATION: - /* fall-through */ - case TGSI_TOKEN_TYPE_IMMEDIATE: - /* fall-through */ - default: - ; /* no-op */ - } - } - - tgsi_parse_free(&parse); - - /* if we get here, it's a pass-through shader */ - return TRUE; -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h b/src/gallium/auxiliary/tgsi/util/tgsi_scan.h deleted file mode 100644 index 5cb6efb343..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_scan.h +++ /dev/null @@ -1,74 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_SCAN_H -#define TGSI_SCAN_H - - -#include "pipe/p_compiler.h" -#include "pipe/p_state.h" -#include "pipe/p_shader_tokens.h" - - -/** - * Shader summary info - */ -struct tgsi_shader_info -{ - uint num_tokens; - - /* XXX eventually remove the corresponding fields from pipe_shader_state: */ - ubyte num_inputs; - ubyte num_outputs; - ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; /**< TGSI_SEMANTIC_x */ - ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; - ubyte output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; /**< TGSI_SEMANTIC_x */ - ubyte output_semantic_index[PIPE_MAX_SHADER_OUTPUTS]; - - uint file_mask[TGSI_FILE_COUNT]; /**< bitmask of declared registers */ - uint file_count[TGSI_FILE_COUNT]; /**< number of declared registers */ - int file_max[TGSI_FILE_COUNT]; /**< highest index of declared registers */ - - uint immediate_count; /**< number of immediates declared */ - - uint opcode_count[TGSI_OPCODE_LAST]; /**< opcode histogram */ - - boolean writes_z; /**< does fragment shader write Z value? */ - boolean uses_kill; /**< KIL or KILP instruction used? */ -}; - - -extern void -tgsi_scan_shader(const struct tgsi_token *tokens, - struct tgsi_shader_info *info); - - -extern boolean -tgsi_is_passthrough_shader(const struct tgsi_token *tokens); - - -#endif /* TGSI_SCAN_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.c b/src/gallium/auxiliary/tgsi/util/tgsi_text.c deleted file mode 100644 index 35cb3055bb..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.c +++ /dev/null @@ -1,1221 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "tgsi_text.h" -#include "tgsi_build.h" -#include "tgsi_parse.h" -#include "tgsi_sanity.h" -#include "tgsi_util.h" - -static boolean is_alpha_underscore( const char *cur ) -{ - return - (*cur >= 'a' && *cur <= 'z') || - (*cur >= 'A' && *cur <= 'Z') || - *cur == '_'; -} - -static boolean is_digit( const char *cur ) -{ - return *cur >= '0' && *cur <= '9'; -} - -static boolean is_digit_alpha_underscore( const char *cur ) -{ - return is_digit( cur ) || is_alpha_underscore( cur ); -} - -static boolean str_match_no_case( const char **pcur, const char *str ) -{ - const char *cur = *pcur; - - while (*str != '\0' && *str == toupper( *cur )) { - str++; - cur++; - } - if (*str == '\0') { - *pcur = cur; - return TRUE; - } - return FALSE; -} - -/* Eat zero or more whitespaces. - */ -static void eat_opt_white( const char **pcur ) -{ - while (**pcur == ' ' || **pcur == '\t' || **pcur == '\n') - (*pcur)++; -} - -/* Eat one or more whitespaces. - * Return TRUE if at least one whitespace eaten. - */ -static boolean eat_white( const char **pcur ) -{ - const char *cur = *pcur; - - eat_opt_white( pcur ); - return *pcur > cur; -} - -/* Parse unsigned integer. - * No checks for overflow. - */ -static boolean parse_uint( const char **pcur, uint *val ) -{ - const char *cur = *pcur; - - if (is_digit( cur )) { - *val = *cur++ - '0'; - while (is_digit( cur )) - *val = *val * 10 + *cur++ - '0'; - *pcur = cur; - return TRUE; - } - return FALSE; -} - -/* Parse floating point. - */ -static boolean parse_float( const char **pcur, float *val ) -{ - const char *cur = *pcur; - boolean integral_part = FALSE; - boolean fractional_part = FALSE; - - *val = (float) atof( cur ); - - if (*cur == '-' || *cur == '+') - cur++; - if (is_digit( cur )) { - cur++; - integral_part = TRUE; - while (is_digit( cur )) - cur++; - } - if (*cur == '.') { - cur++; - if (is_digit( cur )) { - cur++; - fractional_part = TRUE; - while (is_digit( cur )) - cur++; - } - } - if (!integral_part && !fractional_part) - return FALSE; - if (toupper( *cur ) == 'E') { - cur++; - if (*cur == '-' || *cur == '+') - cur++; - if (is_digit( cur )) { - cur++; - while (is_digit( cur )) - cur++; - } - else - return FALSE; - } - *pcur = cur; - return TRUE; -} - -struct translate_ctx -{ - const char *text; - const char *cur; - struct tgsi_token *tokens; - struct tgsi_token *tokens_cur; - struct tgsi_token *tokens_end; - struct tgsi_header *header; -}; - -static void report_error( struct translate_ctx *ctx, const char *msg ) -{ - debug_printf( "\nError: %s", msg ); -} - -/* Parse shader header. - * Return TRUE for one of the following headers. - * FRAG1.1 - * GEOM1.1 - * VERT1.1 - */ -static boolean parse_header( struct translate_ctx *ctx ) -{ - uint processor; - - if (str_match_no_case( &ctx->cur, "FRAG1.1" )) - processor = TGSI_PROCESSOR_FRAGMENT; - else if (str_match_no_case( &ctx->cur, "VERT1.1" )) - processor = TGSI_PROCESSOR_VERTEX; - else if (str_match_no_case( &ctx->cur, "GEOM1.1" )) - processor = TGSI_PROCESSOR_GEOMETRY; - else { - report_error( ctx, "Unknown header" ); - return FALSE; - } - - if (ctx->tokens_cur >= ctx->tokens_end) - return FALSE; - *(struct tgsi_version *) ctx->tokens_cur++ = tgsi_build_version(); - - if (ctx->tokens_cur >= ctx->tokens_end) - return FALSE; - ctx->header = (struct tgsi_header *) ctx->tokens_cur++; - *ctx->header = tgsi_build_header(); - - if (ctx->tokens_cur >= ctx->tokens_end) - return FALSE; - *(struct tgsi_processor *) ctx->tokens_cur++ = tgsi_build_processor( processor, ctx->header ); - - return TRUE; -} - -static boolean parse_label( struct translate_ctx *ctx, uint *val ) -{ - const char *cur = ctx->cur; - - if (parse_uint( &cur, val )) { - eat_opt_white( &cur ); - if (*cur == ':') { - cur++; - ctx->cur = cur; - return TRUE; - } - } - return FALSE; -} - -static const char *file_names[TGSI_FILE_COUNT] = -{ - "NULL", - "CONST", - "IN", - "OUT", - "TEMP", - "SAMP", - "ADDR", - "IMM" -}; - -static boolean -parse_file( const char **pcur, uint *file ) -{ - uint i; - - for (i = 0; i < TGSI_FILE_COUNT; i++) { - const char *cur = *pcur; - - if (str_match_no_case( &cur, file_names[i] )) { - if (!is_digit_alpha_underscore( cur )) { - *pcur = cur; - *file = i; - return TRUE; - } - } - } - return FALSE; -} - -static boolean -parse_opt_writemask( - struct translate_ctx *ctx, - uint *writemask ) -{ - const char *cur; - - cur = ctx->cur; - eat_opt_white( &cur ); - if (*cur == '.') { - cur++; - *writemask = TGSI_WRITEMASK_NONE; - eat_opt_white( &cur ); - if (toupper( *cur ) == 'X') { - cur++; - *writemask |= TGSI_WRITEMASK_X; - } - if (toupper( *cur ) == 'Y') { - cur++; - *writemask |= TGSI_WRITEMASK_Y; - } - if (toupper( *cur ) == 'Z') { - cur++; - *writemask |= TGSI_WRITEMASK_Z; - } - if (toupper( *cur ) == 'W') { - cur++; - *writemask |= TGSI_WRITEMASK_W; - } - - if (*writemask == TGSI_WRITEMASK_NONE) { - report_error( ctx, "Writemask expected" ); - return FALSE; - } - - ctx->cur = cur; - } - else { - *writemask = TGSI_WRITEMASK_XYZW; - } - return TRUE; -} - -/* ::= `[' - */ -static boolean -parse_register_file_bracket( - struct translate_ctx *ctx, - uint *file ) -{ - if (!parse_file( &ctx->cur, file )) { - report_error( ctx, "Unknown register file" ); - return FALSE; - } - eat_opt_white( &ctx->cur ); - if (*ctx->cur != '[') { - report_error( ctx, "Expected `['" ); - return FALSE; - } - ctx->cur++; - return TRUE; -} - -/* ::= - */ -static boolean -parse_register_file_bracket_index( - struct translate_ctx *ctx, - uint *file, - int *index ) -{ - uint uindex; - - if (!parse_register_file_bracket( ctx, file )) - return FALSE; - eat_opt_white( &ctx->cur ); - if (!parse_uint( &ctx->cur, &uindex )) { - report_error( ctx, "Expected literal unsigned integer" ); - return FALSE; - } - *index = (int) uindex; - return TRUE; -} - -/* Parse destination register operand. - * ::= `]' - */ -static boolean -parse_register_dst( - struct translate_ctx *ctx, - uint *file, - int *index ) -{ - if (!parse_register_file_bracket_index( ctx, file, index )) - return FALSE; - eat_opt_white( &ctx->cur ); - if (*ctx->cur != ']') { - report_error( ctx, "Expected `]'" ); - return FALSE; - } - ctx->cur++; - return TRUE; -} - -/* Parse source register operand. - * ::= `]' | - * `]' | - * `+' `]' | - * `-' `]' - */ -static boolean -parse_register_src( - struct translate_ctx *ctx, - uint *file, - int *index, - uint *ind_file, - int *ind_index ) -{ - const char *cur; - uint uindex; - - if (!parse_register_file_bracket( ctx, file )) - return FALSE; - eat_opt_white( &ctx->cur ); - cur = ctx->cur; - if (parse_file( &cur, ind_file )) { - if (!parse_register_dst( ctx, ind_file, ind_index )) - return FALSE; - eat_opt_white( &ctx->cur ); - if (*ctx->cur == '+' || *ctx->cur == '-') { - boolean negate; - - negate = *ctx->cur == '-'; - ctx->cur++; - eat_opt_white( &ctx->cur ); - if (!parse_uint( &ctx->cur, &uindex )) { - report_error( ctx, "Expected literal unsigned integer" ); - return FALSE; - } - if (negate) - *index = -(int) uindex; - else - *index = (int) uindex; - } - else { - *index = 0; - } - } - else { - if (!parse_uint( &ctx->cur, &uindex )) { - report_error( ctx, "Expected literal unsigned integer" ); - return FALSE; - } - *index = (int) uindex; - *ind_file = TGSI_FILE_NULL; - *ind_index = 0; - } - eat_opt_white( &ctx->cur ); - if (*ctx->cur != ']') { - report_error( ctx, "Expected `]'" ); - return FALSE; - } - ctx->cur++; - return TRUE; -} - -/* Parse register declaration. - * ::= `]' | - * `..' `]' - */ -static boolean -parse_register_dcl( - struct translate_ctx *ctx, - uint *file, - int *first, - int *last ) -{ - if (!parse_register_file_bracket_index( ctx, file, first )) - return FALSE; - eat_opt_white( &ctx->cur ); - if (ctx->cur[0] == '.' && ctx->cur[1] == '.') { - uint uindex; - - ctx->cur += 2; - eat_opt_white( &ctx->cur ); - if (!parse_uint( &ctx->cur, &uindex )) { - report_error( ctx, "Expected literal integer" ); - return FALSE; - } - *last = (int) uindex; - eat_opt_white( &ctx->cur ); - } - else { - *last = *first; - } - if (*ctx->cur != ']') { - report_error( ctx, "Expected `]' or `..'" ); - return FALSE; - } - ctx->cur++; - return TRUE; -} - -static const char *modulate_names[TGSI_MODULATE_COUNT] = -{ - "_1X", - "_2X", - "_4X", - "_8X", - "_D2", - "_D4", - "_D8" -}; - -static boolean -parse_dst_operand( - struct translate_ctx *ctx, - struct tgsi_full_dst_register *dst ) -{ - uint file; - int index; - uint writemask; - const char *cur; - - if (!parse_register_dst( ctx, &file, &index )) - return FALSE; - - cur = ctx->cur; - eat_opt_white( &cur ); - if (*cur == '_') { - uint i; - - for (i = 0; i < TGSI_MODULATE_COUNT; i++) { - if (str_match_no_case( &cur, modulate_names[i] )) { - if (!is_digit_alpha_underscore( cur )) { - dst->DstRegisterExtModulate.Modulate = i; - ctx->cur = cur; - break; - } - } - } - } - - if (!parse_opt_writemask( ctx, &writemask )) - return FALSE; - - dst->DstRegister.File = file; - dst->DstRegister.Index = index; - dst->DstRegister.WriteMask = writemask; - return TRUE; -} - -static boolean -parse_optional_swizzle( - struct translate_ctx *ctx, - uint swizzle[4], - boolean *parsed_swizzle, - boolean *parsed_extswizzle ) -{ - const char *cur = ctx->cur; - - *parsed_swizzle = FALSE; - *parsed_extswizzle = FALSE; - - eat_opt_white( &cur ); - if (*cur == '.') { - uint i; - - cur++; - eat_opt_white( &cur ); - for (i = 0; i < 4; i++) { - if (toupper( *cur ) == 'X') - swizzle[i] = TGSI_SWIZZLE_X; - else if (toupper( *cur ) == 'Y') - swizzle[i] = TGSI_SWIZZLE_Y; - else if (toupper( *cur ) == 'Z') - swizzle[i] = TGSI_SWIZZLE_Z; - else if (toupper( *cur ) == 'W') - swizzle[i] = TGSI_SWIZZLE_W; - else { - if (*cur == '0') - swizzle[i] = TGSI_EXTSWIZZLE_ZERO; - else if (*cur == '1') - swizzle[i] = TGSI_EXTSWIZZLE_ONE; - else { - report_error( ctx, "Expected register swizzle component `x', `y', `z', `w', `0' or `1'" ); - return FALSE; - } - *parsed_extswizzle = TRUE; - } - cur++; - } - *parsed_swizzle = TRUE; - ctx->cur = cur; - } - return TRUE; -} - -static boolean -parse_src_operand( - struct translate_ctx *ctx, - struct tgsi_full_src_register *src ) -{ - const char *cur; - float value; - uint file; - int index; - uint ind_file; - int ind_index; - uint swizzle[4]; - boolean parsed_swizzle; - boolean parsed_extswizzle; - - if (*ctx->cur == '-') { - cur = ctx->cur; - cur++; - eat_opt_white( &cur ); - if (*cur == '(') { - cur++; - src->SrcRegisterExtMod.Negate = 1; - eat_opt_white( &cur ); - ctx->cur = cur; - } - } - - if (*ctx->cur == '|') { - ctx->cur++; - eat_opt_white( &ctx->cur ); - src->SrcRegisterExtMod.Absolute = 1; - } - - if (*ctx->cur == '-') { - ctx->cur++; - eat_opt_white( &ctx->cur ); - src->SrcRegister.Negate = 1; - } - - cur = ctx->cur; - if (parse_float( &cur, &value )) { - if (value == 2.0f) { - eat_opt_white( &cur ); - if (*cur != '*') { - report_error( ctx, "Expected `*'" ); - return FALSE; - } - cur++; - if (*cur != '(') { - report_error( ctx, "Expected `('" ); - return FALSE; - } - cur++; - src->SrcRegisterExtMod.Scale2X = 1; - eat_opt_white( &cur ); - ctx->cur = cur; - } - } - - if (*ctx->cur == '(') { - ctx->cur++; - eat_opt_white( &ctx->cur ); - src->SrcRegisterExtMod.Bias = 1; - } - - cur = ctx->cur; - if (parse_float( &cur, &value )) { - if (value == 1.0f) { - eat_opt_white( &cur ); - if (*cur != '-') { - report_error( ctx, "Expected `-'" ); - return FALSE; - } - cur++; - if (*cur != '(') { - report_error( ctx, "Expected `('" ); - return FALSE; - } - cur++; - src->SrcRegisterExtMod.Complement = 1; - eat_opt_white( &cur ); - ctx->cur = cur; - } - } - - if (!parse_register_src( ctx, &file, &index, &ind_file, &ind_index )) - return FALSE; - src->SrcRegister.File = file; - src->SrcRegister.Index = index; - if (ind_file != TGSI_FILE_NULL) { - src->SrcRegister.Indirect = 1; - src->SrcRegisterInd.File = ind_file; - src->SrcRegisterInd.Index = ind_index; - } - - /* Parse optional swizzle. - */ - if (parse_optional_swizzle( ctx, swizzle, &parsed_swizzle, &parsed_extswizzle )) { - if (parsed_extswizzle) { - assert( parsed_swizzle ); - - src->SrcRegisterExtSwz.ExtSwizzleX = swizzle[0]; - src->SrcRegisterExtSwz.ExtSwizzleY = swizzle[1]; - src->SrcRegisterExtSwz.ExtSwizzleZ = swizzle[2]; - src->SrcRegisterExtSwz.ExtSwizzleW = swizzle[3]; - } - else if (parsed_swizzle) { - src->SrcRegister.SwizzleX = swizzle[0]; - src->SrcRegister.SwizzleY = swizzle[1]; - src->SrcRegister.SwizzleZ = swizzle[2]; - src->SrcRegister.SwizzleW = swizzle[3]; - } - } - - if (src->SrcRegisterExtMod.Complement) { - eat_opt_white( &ctx->cur ); - if (*ctx->cur != ')') { - report_error( ctx, "Expected `)'" ); - return FALSE; - } - ctx->cur++; - } - - if (src->SrcRegisterExtMod.Bias) { - eat_opt_white( &ctx->cur ); - if (*ctx->cur != ')') { - report_error( ctx, "Expected `)'" ); - return FALSE; - } - ctx->cur++; - eat_opt_white( &ctx->cur ); - if (*ctx->cur != '-') { - report_error( ctx, "Expected `-'" ); - return FALSE; - } - ctx->cur++; - eat_opt_white( &ctx->cur ); - if (!parse_float( &ctx->cur, &value )) { - report_error( ctx, "Expected literal floating point" ); - return FALSE; - } - if (value != 0.5f) { - report_error( ctx, "Expected 0.5" ); - return FALSE; - } - } - - if (src->SrcRegisterExtMod.Scale2X) { - eat_opt_white( &ctx->cur ); - if (*ctx->cur != ')') { - report_error( ctx, "Expected `)'" ); - return FALSE; - } - ctx->cur++; - } - - if (src->SrcRegisterExtMod.Absolute) { - eat_opt_white( &ctx->cur ); - if (*ctx->cur != '|') { - report_error( ctx, "Expected `|'" ); - return FALSE; - } - ctx->cur++; - } - - if (src->SrcRegisterExtMod.Negate) { - eat_opt_white( &ctx->cur ); - if (*ctx->cur != ')') { - report_error( ctx, "Expected `)'" ); - return FALSE; - } - ctx->cur++; - } - - return TRUE; -} - -struct opcode_info -{ - uint num_dst; - uint num_src; - uint is_tex; - uint is_branch; - const char *mnemonic; -}; - -static const struct opcode_info opcode_info[TGSI_OPCODE_LAST] = -{ - { 1, 1, 0, 0, "ARL" }, - { 1, 1, 0, 0, "MOV" }, - { 1, 1, 0, 0, "LIT" }, - { 1, 1, 0, 0, "RCP" }, - { 1, 1, 0, 0, "RSQ" }, - { 1, 1, 0, 0, "EXP" }, - { 1, 1, 0, 0, "LOG" }, - { 1, 2, 0, 0, "MUL" }, - { 1, 2, 0, 0, "ADD" }, - { 1, 2, 0, 0, "DP3" }, - { 1, 2, 0, 0, "DP4" }, - { 1, 2, 0, 0, "DST" }, - { 1, 2, 0, 0, "MIN" }, - { 1, 2, 0, 0, "MAX" }, - { 1, 2, 0, 0, "SLT" }, - { 1, 2, 0, 0, "SGE" }, - { 1, 3, 0, 0, "MAD" }, - { 1, 2, 0, 0, "SUB" }, - { 1, 3, 0, 0, "LERP" }, - { 1, 3, 0, 0, "CND" }, - { 1, 3, 0, 0, "CND0" }, - { 1, 3, 0, 0, "DOT2ADD" }, - { 1, 2, 0, 0, "INDEX" }, - { 1, 1, 0, 0, "NEGATE" }, - { 1, 1, 0, 0, "FRAC" }, - { 1, 3, 0, 0, "CLAMP" }, - { 1, 1, 0, 0, "FLOOR" }, - { 1, 1, 0, 0, "ROUND" }, - { 1, 1, 0, 0, "EXPBASE2" }, - { 1, 1, 0, 0, "LOGBASE2" }, - { 1, 2, 0, 0, "POWER" }, - { 1, 2, 0, 0, "CROSSPRODUCT" }, - { 1, 2, 0, 0, "MULTIPLYMATRIX" }, - { 1, 1, 0, 0, "ABS" }, - { 1, 1, 0, 0, "RCC" }, - { 1, 2, 0, 0, "DPH" }, - { 1, 1, 0, 0, "COS" }, - { 1, 1, 0, 0, "DDX" }, - { 1, 1, 0, 0, "DDY" }, - { 0, 1, 0, 0, "KILP" }, - { 1, 1, 0, 0, "PK2H" }, - { 1, 1, 0, 0, "PK2US" }, - { 1, 1, 0, 0, "PK4B" }, - { 1, 1, 0, 0, "PK4UB" }, - { 1, 2, 0, 0, "RFL" }, - { 1, 2, 0, 0, "SEQ" }, - { 1, 2, 0, 0, "SFL" }, - { 1, 2, 0, 0, "SGT" }, - { 1, 1, 0, 0, "SIN" }, - { 1, 2, 0, 0, "SLE" }, - { 1, 2, 0, 0, "SNE" }, - { 1, 2, 0, 0, "STR" }, - { 1, 2, 1, 0, "TEX" }, - { 1, 4, 1, 0, "TXD" }, - { 1, 2, 1, 0, "TXP" }, - { 1, 1, 0, 0, "UP2H" }, - { 1, 1, 0, 0, "UP2US" }, - { 1, 1, 0, 0, "UP4B" }, - { 1, 1, 0, 0, "UP4UB" }, - { 1, 3, 0, 0, "X2D" }, - { 1, 1, 0, 0, "ARA" }, - { 1, 1, 0, 0, "ARR" }, - { 0, 1, 0, 0, "BRA" }, - { 0, 0, 0, 1, "CAL" }, - { 0, 0, 0, 0, "RET" }, - { 1, 1, 0, 0, "SSG" }, - { 1, 3, 0, 0, "CMP" }, - { 1, 1, 0, 0, "SCS" }, - { 1, 2, 1, 0, "TXB" }, - { 1, 1, 0, 0, "NRM" }, - { 1, 2, 0, 0, "DIV" }, - { 1, 2, 0, 0, "DP2" }, - { 1, 2, 1, 0, "TXL" }, - { 0, 0, 0, 0, "BRK" }, - { 0, 1, 0, 1, "IF" }, - { 0, 0, 0, 0, "LOOP" }, - { 0, 1, 0, 0, "REP" }, - { 0, 0, 0, 1, "ELSE" }, - { 0, 0, 0, 0, "ENDIF" }, - { 0, 0, 0, 0, "ENDLOOP" }, - { 0, 0, 0, 0, "ENDREP" }, - { 0, 1, 0, 0, "PUSHA" }, - { 1, 0, 0, 0, "POPA" }, - { 1, 1, 0, 0, "CEIL" }, - { 1, 1, 0, 0, "I2F" }, - { 1, 1, 0, 0, "NOT" }, - { 1, 1, 0, 0, "TRUNC" }, - { 1, 2, 0, 0, "SHL" }, - { 1, 2, 0, 0, "SHR" }, - { 1, 2, 0, 0, "AND" }, - { 1, 2, 0, 0, "OR" }, - { 1, 2, 0, 0, "MOD" }, - { 1, 2, 0, 0, "XOR" }, - { 1, 3, 0, 0, "SAD" }, - { 1, 2, 1, 0, "TXF" }, - { 1, 2, 1, 0, "TXQ" }, - { 0, 0, 0, 0, "CONT" }, - { 0, 0, 0, 0, "EMIT" }, - { 0, 0, 0, 0, "ENDPRIM" }, - { 0, 0, 0, 1, "BGNLOOP2" }, - { 0, 0, 0, 0, "BGNSUB" }, - { 0, 0, 0, 1, "ENDLOOP2" }, - { 0, 0, 0, 0, "ENDSUB" }, - { 1, 1, 0, 0, "NOISE1" }, - { 1, 1, 0, 0, "NOISE2" }, - { 1, 1, 0, 0, "NOISE3" }, - { 1, 1, 0, 0, "NOISE4" }, - { 0, 0, 0, 0, "NOP" }, - { 1, 2, 0, 0, "M4X3" }, - { 1, 2, 0, 0, "M3X4" }, - { 1, 2, 0, 0, "M3X3" }, - { 1, 2, 0, 0, "M3X2" }, - { 1, 1, 0, 0, "NRM4" }, - { 0, 1, 0, 0, "CALLNZ" }, - { 0, 1, 0, 0, "IFC" }, - { 0, 1, 0, 0, "BREAKC" }, - { 0, 0, 0, 0, "KIL" }, - { 0, 0, 0, 0, "END" }, - { 1, 1, 0, 0, "SWZ" } -}; - -static const char *texture_names[TGSI_TEXTURE_COUNT] = -{ - "UNKNOWN", - "1D", - "2D", - "3D", - "CUBE", - "RECT", - "SHADOW1D", - "SHADOW2D", - "SHADOWRECT" -}; - -static boolean -parse_instruction( - struct translate_ctx *ctx, - boolean has_label ) -{ - uint i; - uint saturate = TGSI_SAT_NONE; - const struct opcode_info *info; - struct tgsi_full_instruction inst; - uint advance; - - /* Parse instruction name. - */ - eat_opt_white( &ctx->cur ); - for (i = 0; i < TGSI_OPCODE_LAST; i++) { - const char *cur = ctx->cur; - - info = &opcode_info[i]; - if (str_match_no_case( &cur, info->mnemonic )) { - if (str_match_no_case( &cur, "_SATNV" )) - saturate = TGSI_SAT_MINUS_PLUS_ONE; - else if (str_match_no_case( &cur, "_SAT" )) - saturate = TGSI_SAT_ZERO_ONE; - - if (info->num_dst + info->num_src + info->is_tex == 0) { - if (!is_digit_alpha_underscore( cur )) { - ctx->cur = cur; - break; - } - } - else if (*cur == '\0' || eat_white( &cur )) { - ctx->cur = cur; - break; - } - } - } - if (i == TGSI_OPCODE_LAST) { - if (has_label) - report_error( ctx, "Unknown opcode" ); - else - report_error( ctx, "Expected `DCL', `IMM' or a label" ); - return FALSE; - } - - inst = tgsi_default_full_instruction(); - inst.Instruction.Opcode = i; - inst.Instruction.Saturate = saturate; - inst.Instruction.NumDstRegs = info->num_dst; - inst.Instruction.NumSrcRegs = info->num_src; - - /* Parse instruction operands. - */ - for (i = 0; i < info->num_dst + info->num_src + info->is_tex; i++) { - if (i > 0) { - eat_opt_white( &ctx->cur ); - if (*ctx->cur != ',') { - report_error( ctx, "Expected `,'" ); - return FALSE; - } - ctx->cur++; - eat_opt_white( &ctx->cur ); - } - - if (i < info->num_dst) { - if (!parse_dst_operand( ctx, &inst.FullDstRegisters[i] )) - return FALSE; - } - else if (i < info->num_dst + info->num_src) { - if (!parse_src_operand( ctx, &inst.FullSrcRegisters[i - info->num_dst] )) - return FALSE; - } - else { - uint j; - - for (j = 0; j < TGSI_TEXTURE_COUNT; j++) { - if (str_match_no_case( &ctx->cur, texture_names[j] )) { - if (!is_digit_alpha_underscore( ctx->cur )) { - inst.InstructionExtTexture.Texture = j; - break; - } - } - } - if (j == TGSI_TEXTURE_COUNT) { - report_error( ctx, "Expected texture target" ); - return FALSE; - } - } - } - - if (info->is_branch) { - uint target; - - eat_opt_white( &ctx->cur ); - if (*ctx->cur != ':') { - report_error( ctx, "Expected `:'" ); - return FALSE; - } - ctx->cur++; - eat_opt_white( &ctx->cur ); - if (!parse_uint( &ctx->cur, &target )) { - report_error( ctx, "Expected a label" ); - return FALSE; - } - inst.InstructionExtLabel.Label = target; - } - - advance = tgsi_build_full_instruction( - &inst, - ctx->tokens_cur, - ctx->header, - (uint) (ctx->tokens_end - ctx->tokens_cur) ); - if (advance == 0) - return FALSE; - ctx->tokens_cur += advance; - - return TRUE; -} - -static const char *semantic_names[TGSI_SEMANTIC_COUNT] = -{ - "POSITION", - "COLOR", - "BCOLOR", - "FOG", - "PSIZE", - "GENERIC", - "NORMAL" -}; - -static const char *interpolate_names[TGSI_INTERPOLATE_COUNT] = -{ - "CONSTANT", - "LINEAR", - "PERSPECTIVE" -}; - -static boolean parse_declaration( struct translate_ctx *ctx ) -{ - struct tgsi_full_declaration decl; - uint file; - int first; - int last; - uint writemask; - const char *cur; - uint advance; - - if (!eat_white( &ctx->cur )) { - report_error( ctx, "Syntax error" ); - return FALSE; - } - if (!parse_register_dcl( ctx, &file, &first, &last )) - return FALSE; - if (!parse_opt_writemask( ctx, &writemask )) - return FALSE; - - decl = tgsi_default_full_declaration(); - decl.Declaration.File = file; - decl.Declaration.UsageMask = writemask; - decl.DeclarationRange.First = first; - decl.DeclarationRange.Last = last; - - cur = ctx->cur; - eat_opt_white( &cur ); - if (*cur == ',') { - uint i; - - cur++; - eat_opt_white( &cur ); - for (i = 0; i < TGSI_SEMANTIC_COUNT; i++) { - if (str_match_no_case( &cur, semantic_names[i] )) { - const char *cur2 = cur; - uint index; - - if (is_digit_alpha_underscore( cur )) - continue; - eat_opt_white( &cur2 ); - if (*cur2 == '[') { - cur2++; - eat_opt_white( &cur2 ); - if (!parse_uint( &cur2, &index )) { - report_error( ctx, "Expected literal integer" ); - return FALSE; - } - eat_opt_white( &cur2 ); - if (*cur2 != ']') { - report_error( ctx, "Expected `]'" ); - return FALSE; - } - cur2++; - - decl.Semantic.SemanticIndex = index; - - cur = cur2; - } - - decl.Declaration.Semantic = 1; - decl.Semantic.SemanticName = i; - - ctx->cur = cur; - break; - } - } - } - - cur = ctx->cur; - eat_opt_white( &cur ); - if (*cur == ',') { - uint i; - - cur++; - eat_opt_white( &cur ); - for (i = 0; i < TGSI_INTERPOLATE_COUNT; i++) { - if (str_match_no_case( &cur, interpolate_names[i] )) { - if (is_digit_alpha_underscore( cur )) - continue; - decl.Declaration.Interpolate = i; - - ctx->cur = cur; - break; - } - } - if (i == TGSI_INTERPOLATE_COUNT) { - report_error( ctx, "Expected semantic or interpolate attribute" ); - return FALSE; - } - } - - advance = tgsi_build_full_declaration( - &decl, - ctx->tokens_cur, - ctx->header, - (uint) (ctx->tokens_end - ctx->tokens_cur) ); - if (advance == 0) - return FALSE; - ctx->tokens_cur += advance; - - return TRUE; -} - -static boolean parse_immediate( struct translate_ctx *ctx ) -{ - struct tgsi_full_immediate imm; - uint i; - float values[4]; - uint advance; - - if (!eat_white( &ctx->cur )) { - report_error( ctx, "Syntax error" ); - return FALSE; - } - if (!str_match_no_case( &ctx->cur, "FLT32" ) || is_digit_alpha_underscore( ctx->cur )) { - report_error( ctx, "Expected `FLT32'" ); - return FALSE; - } - eat_opt_white( &ctx->cur ); - if (*ctx->cur != '{') { - report_error( ctx, "Expected `{'" ); - return FALSE; - } - ctx->cur++; - for (i = 0; i < 4; i++) { - eat_opt_white( &ctx->cur ); - if (i > 0) { - if (*ctx->cur != ',') { - report_error( ctx, "Expected `,'" ); - return FALSE; - } - ctx->cur++; - eat_opt_white( &ctx->cur ); - } - if (!parse_float( &ctx->cur, &values[i] )) { - report_error( ctx, "Expected literal floating point" ); - return FALSE; - } - } - eat_opt_white( &ctx->cur ); - if (*ctx->cur != '}') { - report_error( ctx, "Expected `}'" ); - return FALSE; - } - ctx->cur++; - - imm = tgsi_default_full_immediate(); - imm.Immediate.Size += 4; - imm.Immediate.DataType = TGSI_IMM_FLOAT32; - imm.u.Pointer = values; - - advance = tgsi_build_full_immediate( - &imm, - ctx->tokens_cur, - ctx->header, - (uint) (ctx->tokens_end - ctx->tokens_cur) ); - if (advance == 0) - return FALSE; - ctx->tokens_cur += advance; - - return TRUE; -} - -static boolean translate( struct translate_ctx *ctx ) -{ - eat_opt_white( &ctx->cur ); - if (!parse_header( ctx )) - return FALSE; - - while (*ctx->cur != '\0') { - uint label_val = 0; - - if (!eat_white( &ctx->cur )) { - report_error( ctx, "Syntax error" ); - return FALSE; - } - - if (*ctx->cur == '\0') - break; - - if (parse_label( ctx, &label_val )) { - if (!parse_instruction( ctx, TRUE )) - return FALSE; - } - else if (str_match_no_case( &ctx->cur, "DCL" )) { - if (!parse_declaration( ctx )) - return FALSE; - } - else if (str_match_no_case( &ctx->cur, "IMM" )) { - if (!parse_immediate( ctx )) - return FALSE; - } - else if (!parse_instruction( ctx, FALSE )) { - return FALSE; - } - } - - return TRUE; -} - -boolean -tgsi_text_translate( - const char *text, - struct tgsi_token *tokens, - uint num_tokens ) -{ - struct translate_ctx ctx; - - ctx.text = text; - ctx.cur = text; - ctx.tokens = tokens; - ctx.tokens_cur = tokens; - ctx.tokens_end = tokens + num_tokens; - - if (!translate( &ctx )) - return FALSE; - - return tgsi_sanity_check( tokens ); -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_text.h b/src/gallium/auxiliary/tgsi/util/tgsi_text.h deleted file mode 100644 index 8eeeeef140..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_text.h +++ /dev/null @@ -1,47 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_TEXT_H -#define TGSI_TEXT_H - -#include "pipe/p_shader_tokens.h" - -#if defined __cplusplus -extern "C" { -#endif - -boolean -tgsi_text_translate( - const char *text, - struct tgsi_token *tokens, - uint num_tokens ); - -#if defined __cplusplus -} -#endif - -#endif /* TGSI_TEXT_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_transform.c b/src/gallium/auxiliary/tgsi/util/tgsi_transform.c deleted file mode 100644 index 357f77b05a..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_transform.c +++ /dev/null @@ -1,199 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * TGSI program transformation utility. - * - * Authors: Brian Paul - */ - - -#include "tgsi_transform.h" - - - -static void -emit_instruction(struct tgsi_transform_context *ctx, - const struct tgsi_full_instruction *inst) -{ - uint ti = ctx->ti; - - ti += tgsi_build_full_instruction(inst, - ctx->tokens_out + ti, - ctx->header, - ctx->max_tokens_out - ti); - ctx->ti = ti; -} - - -static void -emit_declaration(struct tgsi_transform_context *ctx, - const struct tgsi_full_declaration *decl) -{ - uint ti = ctx->ti; - - ti += tgsi_build_full_declaration(decl, - ctx->tokens_out + ti, - ctx->header, - ctx->max_tokens_out - ti); - ctx->ti = ti; -} - - -static void -emit_immediate(struct tgsi_transform_context *ctx, - const struct tgsi_full_immediate *imm) -{ - uint ti = ctx->ti; - - ti += tgsi_build_full_immediate(imm, - ctx->tokens_out + ti, - ctx->header, - ctx->max_tokens_out - ti); - ctx->ti = ti; -} - - - -/** - * Apply user-defined transformations to the input shader to produce - * the output shader. - * For example, a register search-and-replace operation could be applied - * by defining a transform_instruction() callback that examined and changed - * the instruction src/dest regs. - * - * \return number of tokens emitted - */ -int -tgsi_transform_shader(const struct tgsi_token *tokens_in, - struct tgsi_token *tokens_out, - uint max_tokens_out, - struct tgsi_transform_context *ctx) -{ - uint procType; - - /* input shader */ - struct tgsi_parse_context parse; - - /* output shader */ - struct tgsi_processor *processor; - - - /** - ** callback context init - **/ - ctx->emit_instruction = emit_instruction; - ctx->emit_declaration = emit_declaration; - ctx->emit_immediate = emit_immediate; - ctx->tokens_out = tokens_out; - ctx->max_tokens_out = max_tokens_out; - - - /** - ** Setup to begin parsing input shader - **/ - if (tgsi_parse_init( &parse, tokens_in ) != TGSI_PARSE_OK) { - debug_printf("tgsi_parse_init() failed in tgsi_transform_shader()!\n"); - return -1; - } - procType = parse.FullHeader.Processor.Processor; - assert(procType == TGSI_PROCESSOR_FRAGMENT || - procType == TGSI_PROCESSOR_VERTEX || - procType == TGSI_PROCESSOR_GEOMETRY); - - - /** - ** Setup output shader - **/ - *(struct tgsi_version *) &tokens_out[0] = tgsi_build_version(); - - ctx->header = (struct tgsi_header *) (tokens_out + 1); - *ctx->header = tgsi_build_header(); - - processor = (struct tgsi_processor *) (tokens_out + 2); - *processor = tgsi_build_processor( procType, ctx->header ); - - ctx->ti = 3; - - - /** - ** Loop over incoming program tokens/instructions - */ - while( !tgsi_parse_end_of_tokens( &parse ) ) { - - tgsi_parse_token( &parse ); - - switch( parse.FullToken.Token.Type ) { - case TGSI_TOKEN_TYPE_INSTRUCTION: - { - struct tgsi_full_instruction *fullinst - = &parse.FullToken.FullInstruction; - - if (ctx->transform_instruction) - ctx->transform_instruction(ctx, fullinst); - else - ctx->emit_instruction(ctx, fullinst); - } - break; - - case TGSI_TOKEN_TYPE_DECLARATION: - { - struct tgsi_full_declaration *fulldecl - = &parse.FullToken.FullDeclaration; - - if (ctx->transform_declaration) - ctx->transform_declaration(ctx, fulldecl); - else - ctx->emit_declaration(ctx, fulldecl); - } - break; - - case TGSI_TOKEN_TYPE_IMMEDIATE: - { - struct tgsi_full_immediate *fullimm - = &parse.FullToken.FullImmediate; - - if (ctx->transform_immediate) - ctx->transform_immediate(ctx, fullimm); - else - ctx->emit_immediate(ctx, fullimm); - } - break; - - default: - assert( 0 ); - } - } - - if (ctx->epilog) { - ctx->epilog(ctx); - } - - tgsi_parse_free (&parse); - - return ctx->ti; -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_transform.h b/src/gallium/auxiliary/tgsi/util/tgsi_transform.h deleted file mode 100644 index fcf85d603b..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_transform.h +++ /dev/null @@ -1,93 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_TRANSFORM_H -#define TGSI_TRANSFORM_H - - -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_build.h" - - - -/** - * Subclass this to add caller-specific data - */ -struct tgsi_transform_context -{ -/**** PUBLIC ***/ - - /** - * User-defined callbacks invoked per instruction. - */ - void (*transform_instruction)(struct tgsi_transform_context *ctx, - struct tgsi_full_instruction *inst); - - void (*transform_declaration)(struct tgsi_transform_context *ctx, - struct tgsi_full_declaration *decl); - - void (*transform_immediate)(struct tgsi_transform_context *ctx, - struct tgsi_full_immediate *imm); - - /** - * Called at end of input program to allow caller to append extra - * instructions. Return number of tokens emitted. - */ - void (*epilog)(struct tgsi_transform_context *ctx); - - -/*** PRIVATE ***/ - - /** - * These are setup by tgsi_transform_shader() and cannot be overridden. - * Meant to be called from in the above user callback functions. - */ - void (*emit_instruction)(struct tgsi_transform_context *ctx, - const struct tgsi_full_instruction *inst); - void (*emit_declaration)(struct tgsi_transform_context *ctx, - const struct tgsi_full_declaration *decl); - void (*emit_immediate)(struct tgsi_transform_context *ctx, - const struct tgsi_full_immediate *imm); - - struct tgsi_header *header; - uint max_tokens_out; - struct tgsi_token *tokens_out; - uint ti; -}; - - - -extern int -tgsi_transform_shader(const struct tgsi_token *tokens_in, - struct tgsi_token *tokens_out, - uint max_tokens_out, - struct tgsi_transform_context *ctx); - - -#endif /* TGSI_TRANSFORM_H */ diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_util.c b/src/gallium/auxiliary/tgsi/util/tgsi_util.c deleted file mode 100644 index 09486e649e..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_util.c +++ /dev/null @@ -1,300 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" -#include "tgsi_parse.h" -#include "tgsi_build.h" -#include "tgsi_util.h" - -union pointer_hack -{ - void *pointer; - uint64_t uint64; -}; - -void * -tgsi_align_128bit( - void *unaligned ) -{ - union pointer_hack ph; - - ph.uint64 = 0; - ph.pointer = unaligned; - ph.uint64 = (ph.uint64 + 15) & ~15; - return ph.pointer; -} - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->SwizzleX; - case 1: - return reg->SwizzleY; - case 2: - return reg->SwizzleZ; - case 3: - return reg->SwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->ExtSwizzleX; - case 1: - return reg->ExtSwizzleY; - case 2: - return reg->ExtSwizzleZ; - case 3: - return reg->ExtSwizzleW; - default: - assert( 0 ); - } - return 0; -} - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned swizzle; - - /* - * First, calculate the extended swizzle for a given channel. This will give - * us either a channel index into the simple swizzle or a constant 1 or 0. - */ - swizzle = tgsi_util_get_src_register_extswizzle( - ®->SrcRegisterExtSwz, - component ); - - assert (TGSI_SWIZZLE_X == TGSI_EXTSWIZZLE_X); - assert (TGSI_SWIZZLE_Y == TGSI_EXTSWIZZLE_Y); - assert (TGSI_SWIZZLE_Z == TGSI_EXTSWIZZLE_Z); - assert (TGSI_SWIZZLE_W == TGSI_EXTSWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ZERO > TGSI_SWIZZLE_W); - assert (TGSI_EXTSWIZZLE_ONE > TGSI_SWIZZLE_W); - - /* - * Second, calculate the simple swizzle for the unswizzled channel index. - * Leave the constants intact, they are not affected by the simple swizzle. - */ - if( swizzle <= TGSI_SWIZZLE_W ) { - swizzle = tgsi_util_get_src_register_swizzle( - ®->SrcRegister, - swizzle ); - } - - return swizzle; -} - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - unsigned swizzle, - unsigned component ) -{ - switch( component ) { - case 0: - reg->SwizzleX = swizzle; - break; - case 1: - reg->SwizzleY = swizzle; - break; - case 2: - reg->SwizzleZ = swizzle; - break; - case 3: - reg->SwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - unsigned swizzle, - unsigned component ) -{ - switch( component ) { - case 0: - reg->ExtSwizzleX = swizzle; - break; - case 1: - reg->ExtSwizzleY = swizzle; - break; - case 2: - reg->ExtSwizzleZ = swizzle; - break; - case 3: - reg->ExtSwizzleW = swizzle; - break; - default: - assert( 0 ); - } -} - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ) -{ - switch( component ) { - case 0: - return reg->NegateX; - case 1: - return reg->NegateY; - case 2: - return reg->NegateZ; - case 3: - return reg->NegateW; - default: - assert( 0 ); - } - return 0; -} - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ) -{ - switch( component ) { - case 0: - reg->NegateX = negate; - break; - case 1: - reg->NegateY = negate; - break; - case 2: - reg->NegateZ = negate; - break; - case 3: - reg->NegateW = negate; - break; - default: - assert( 0 ); - } -} - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ) -{ - unsigned sign_mode; - - if( reg->SrcRegisterExtMod.Absolute ) { - /* Consider only the post-abs negation. */ - - if( reg->SrcRegisterExtMod.Negate ) { - sign_mode = TGSI_UTIL_SIGN_SET; - } - else { - sign_mode = TGSI_UTIL_SIGN_CLEAR; - } - } - else { - /* Accumulate the three negations. */ - - unsigned negate; - - negate = reg->SrcRegister.Negate; - if( tgsi_util_get_src_register_extnegate( ®->SrcRegisterExtSwz, component ) ) { - negate = !negate; - } - if( reg->SrcRegisterExtMod.Negate ) { - negate = !negate; - } - - if( negate ) { - sign_mode = TGSI_UTIL_SIGN_TOGGLE; - } - else { - sign_mode = TGSI_UTIL_SIGN_KEEP; - } - } - - return sign_mode; -} - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - unsigned sign_mode ) -{ - reg->SrcRegisterExtSwz.NegateX = 0; - reg->SrcRegisterExtSwz.NegateY = 0; - reg->SrcRegisterExtSwz.NegateZ = 0; - reg->SrcRegisterExtSwz.NegateW = 0; - - switch (sign_mode) - { - case TGSI_UTIL_SIGN_CLEAR: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_SET: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 1; - reg->SrcRegisterExtMod.Negate = 1; - break; - - case TGSI_UTIL_SIGN_TOGGLE: - reg->SrcRegister.Negate = 1; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - case TGSI_UTIL_SIGN_KEEP: - reg->SrcRegister.Negate = 0; - reg->SrcRegisterExtMod.Absolute = 0; - reg->SrcRegisterExtMod.Negate = 0; - break; - - default: - assert( 0 ); - } -} diff --git a/src/gallium/auxiliary/tgsi/util/tgsi_util.h b/src/gallium/auxiliary/tgsi/util/tgsi_util.h deleted file mode 100644 index 7877f34558..0000000000 --- a/src/gallium/auxiliary/tgsi/util/tgsi_util.h +++ /dev/null @@ -1,96 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef TGSI_UTIL_H -#define TGSI_UTIL_H - -#if defined __cplusplus -extern "C" { -#endif - -void * -tgsi_align_128bit( - void *unaligned ); - -unsigned -tgsi_util_get_src_register_swizzle( - const struct tgsi_src_register *reg, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extswizzle( - const struct tgsi_src_register_ext_swz *reg, - unsigned component); - -unsigned -tgsi_util_get_full_src_register_extswizzle( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_src_register_swizzle( - struct tgsi_src_register *reg, - unsigned swizzle, - unsigned component ); - -void -tgsi_util_set_src_register_extswizzle( - struct tgsi_src_register_ext_swz *reg, - unsigned swizzle, - unsigned component ); - -unsigned -tgsi_util_get_src_register_extnegate( - const struct tgsi_src_register_ext_swz *reg, - unsigned component ); - -void -tgsi_util_set_src_register_extnegate( - struct tgsi_src_register_ext_swz *reg, - unsigned negate, - unsigned component ); - -#define TGSI_UTIL_SIGN_CLEAR 0 /* Force positive */ -#define TGSI_UTIL_SIGN_SET 1 /* Force negative */ -#define TGSI_UTIL_SIGN_TOGGLE 2 /* Negate */ -#define TGSI_UTIL_SIGN_KEEP 3 /* No change */ - -unsigned -tgsi_util_get_full_src_register_sign_mode( - const struct tgsi_full_src_register *reg, - unsigned component ); - -void -tgsi_util_set_full_src_register_sign_mode( - struct tgsi_full_src_register *reg, - unsigned sign_mode ); - -#if defined __cplusplus -} -#endif - -#endif /* TGSI_UTIL_H */ diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 4999822068..8713ff5d58 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -45,9 +45,9 @@ #include "util/u_gen_mipmap.h" #include "util/u_simple_shaders.h" -#include "tgsi/util/tgsi_build.h" -#include "tgsi/util/tgsi_dump.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_build.h" +#include "tgsi/tgsi_dump.h" +#include "tgsi/tgsi_parse.h" #include "cso_cache/cso_context.h" diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index 505d93d727..c34fb6ee33 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -43,9 +43,9 @@ #include "util/u_simple_shaders.h" -#include "tgsi/util/tgsi_build.h" -#include "tgsi/util/tgsi_dump.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_build.h" +#include "tgsi/tgsi_dump.h" +#include "tgsi/tgsi_parse.h" diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h index 8df41c1d4c..f1d1ca89a9 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.h +++ b/src/gallium/drivers/cell/ppu/cell_context.h @@ -37,7 +37,7 @@ #include "cell_winsys.h" #include "cell/common.h" #include "rtasm/rtasm_ppc_spe.h" -#include "tgsi/util/tgsi_scan.h" +#include "tgsi/tgsi_scan.h" struct cell_vbuf_render; diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c index c3a3fbd066..f5707f2bb8 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -30,7 +30,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "draw/draw_context.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" #include "cell_context.h" #include "cell_state.h" diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 3a80df427d..96393732ed 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -65,8 +65,8 @@ #include "pipe/p_state.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_util.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_util.h" #include "spu_exec.h" #include "spu_main.h" #include "spu_vertex_shader.h" diff --git a/src/gallium/drivers/cell/spu/spu_exec.h b/src/gallium/drivers/cell/spu/spu_exec.h index 3e17c490d2..c68f78f59b 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.h +++ b/src/gallium/drivers/cell/spu/spu_exec.h @@ -29,7 +29,7 @@ #define SPU_EXEC_H #include "pipe/p_compiler.h" -#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/tgsi_exec.h" #if defined __cplusplus extern "C" { diff --git a/src/gallium/drivers/cell/spu/spu_util.c b/src/gallium/drivers/cell/spu/spu_util.c index ea4274a0a7..74ab2bbd1f 100644 --- a/src/gallium/drivers/cell/spu/spu_util.c +++ b/src/gallium/drivers/cell/spu/spu_util.c @@ -1,8 +1,8 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" //#include "tgsi_build.h" -#include "tgsi/util/tgsi_util.h" +#include "tgsi/tgsi_util.h" unsigned tgsi_util_get_src_register_swizzle( diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h index c8db4f608c..3cdabe45f9 100644 --- a/src/gallium/drivers/i915simple/i915_context.h +++ b/src/gallium/drivers/i915simple/i915_context.h @@ -35,7 +35,7 @@ #include "draw/draw_vertex.h" -#include "tgsi/util/tgsi_scan.h" +#include "tgsi/tgsi_scan.h" #define I915_TEX_UNITS 8 diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 23cd909337..04507ab8ad 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -34,8 +34,8 @@ #include "pipe/p_shader_tokens.h" #include "util/u_string.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_dump.h" #include "draw/draw_vertex.h" diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index dbb33f2695..e8521b385e 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -33,7 +33,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" #include "i915_context.h" #include "i915_reg.h" diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index 2cae7665f7..f00eb34f92 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -38,7 +38,7 @@ #include "pipe/p_defines.h" #include "pipe/p_state.h" -#include "tgsi/util/tgsi_scan.h" +#include "tgsi/tgsi_scan.h" #include "brw_structs.h" #include "brw_winsys.h" diff --git a/src/gallium/drivers/i965simple/brw_sf.c b/src/gallium/drivers/i965simple/brw_sf.c index 96f8fb87a3..b82a2e143b 100644 --- a/src/gallium/drivers/i965simple/brw_sf.c +++ b/src/gallium/drivers/i965simple/brw_sf.c @@ -36,7 +36,7 @@ #include "brw_util.h" #include "brw_sf.h" #include "brw_state.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" static void compile_sf_prog( struct brw_context *brw, diff --git a/src/gallium/drivers/i965simple/brw_shader_info.c b/src/gallium/drivers/i965simple/brw_shader_info.c index fb3da92421..30f37a99d4 100644 --- a/src/gallium/drivers/i965simple/brw_shader_info.c +++ b/src/gallium/drivers/i965simple/brw_shader_info.c @@ -3,7 +3,7 @@ #include "brw_state.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" /** diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index caeeba4630..27ca32843d 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -34,8 +34,8 @@ #include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_dump.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_dump.h" +#include "tgsi/tgsi_parse.h" #include "brw_context.h" #include "brw_defines.h" diff --git a/src/gallium/drivers/i965simple/brw_vs_emit.c b/src/gallium/drivers/i965simple/brw_vs_emit.c index 81423e2d7d..34dbc0624d 100644 --- a/src/gallium/drivers/i965simple/brw_vs_emit.c +++ b/src/gallium/drivers/i965simple/brw_vs_emit.c @@ -33,7 +33,7 @@ #include "brw_vs.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" struct brw_prog_info { unsigned num_temps; diff --git a/src/gallium/drivers/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c index bf1b4d961a..e6f1a44817 100644 --- a/src/gallium/drivers/i965simple/brw_wm_decl.c +++ b/src/gallium/drivers/i965simple/brw_wm_decl.c @@ -4,7 +4,7 @@ #include "brw_wm.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" static struct brw_reg alloc_tmp(struct brw_wm_compile *c) { diff --git a/src/gallium/drivers/i965simple/brw_wm_glsl.c b/src/gallium/drivers/i965simple/brw_wm_glsl.c index 5c90583824..6a4a5aef09 100644 --- a/src/gallium/drivers/i965simple/brw_wm_glsl.c +++ b/src/gallium/drivers/i965simple/brw_wm_glsl.c @@ -4,7 +4,7 @@ #include "brw_wm.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_parse.h" diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index 0b199a2193..cc171bbc39 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -36,8 +36,8 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/util/tgsi_parse.h" +#include "tgsi/tgsi_exec.h" +#include "tgsi/tgsi_parse.h" struct sp_exec_fragment_shader { struct sp_fragment_shader base; diff --git a/src/gallium/drivers/softpipe/sp_fs_llvm.c b/src/gallium/drivers/softpipe/sp_fs_llvm.c index 6e1d9280bb..20226da78c 100644 --- a/src/gallium/drivers/softpipe/sp_fs_llvm.c +++ b/src/gallium/drivers/softpipe/sp_fs_llvm.c @@ -38,7 +38,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "tgsi/exec/tgsi_sse2.h" +#include "tgsi/tgsi_sse2.h" #if 0 diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 69f7f960aa..8b7da7c747 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -36,8 +36,8 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "tgsi/exec/tgsi_exec.h" -#include "tgsi/exec/tgsi_sse2.h" +#include "tgsi/tgsi_exec.h" +#include "tgsi/tgsi_sse2.h" #ifdef PIPE_ARCH_X86 diff --git a/src/gallium/drivers/softpipe/sp_headers.h b/src/gallium/drivers/softpipe/sp_headers.h index 3d9ede69bb..ae2ee210fc 100644 --- a/src/gallium/drivers/softpipe/sp_headers.h +++ b/src/gallium/drivers/softpipe/sp_headers.h @@ -32,7 +32,7 @@ #define SP_HEADERS_H #include "pipe/p_state.h" -#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/tgsi_exec.h" #define PRIM_POINT 1 #define PRIM_LINE 2 diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h index 701e02b295..476ef3dc8f 100644 --- a/src/gallium/drivers/softpipe/sp_state.h +++ b/src/gallium/drivers/softpipe/sp_state.h @@ -32,7 +32,7 @@ #define SP_STATE_H #include "pipe/p_state.h" -#include "tgsi/util/tgsi_scan.h" +#include "tgsi/tgsi_scan.h" #define SP_NEW_VIEWPORT 0x1 diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 901c8f83e7..76fe6bfef9 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -35,8 +35,8 @@ #include "pipe/p_winsys.h" #include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" -#include "tgsi/util/tgsi_dump.h" -#include "tgsi/util/tgsi_scan.h" +#include "tgsi/tgsi_dump.h" +#include "tgsi/tgsi_scan.h" void * diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index be0b57d9fa..63b3b91110 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -40,7 +40,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_util.h" -#include "tgsi/exec/tgsi_exec.h" +#include "tgsi/tgsi_exec.h" /* diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 8d8b762ea5..284ecb827d 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -47,8 +47,8 @@ #include "cso_cache/cso_context.h" #include "util/u_draw_quad.h" #include "util/p_tile.h" -#include "tgsi/util/tgsi_text.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_text.h" +#include "tgsi/tgsi_dump.h" #include "st_device.h" #include "st_sample.h" diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c index 23ecfff0aa..c7d26ce33c 100644 --- a/src/mesa/state_tracker/st_debug.c +++ b/src/mesa/state_tracker/st_debug.c @@ -31,7 +31,7 @@ #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_dump.h" #include "cso_cache/cso_cache.h" diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 9029f12056..6565107b10 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -32,9 +32,9 @@ #include "pipe/p_compiler.h" #include "pipe/p_shader_tokens.h" -#include "tgsi/util/tgsi_parse.h" -#include "tgsi/util/tgsi_build.h" -#include "tgsi/util/tgsi_util.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_build.h" +#include "tgsi/tgsi_util.h" #include "st_mesa_to_tgsi.h" #include "shader/prog_instruction.h" #include "shader/prog_parameter.h" diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 5966bbadae..c25c668329 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -40,7 +40,7 @@ #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" -#include "tgsi/util/tgsi_dump.h" +#include "tgsi/tgsi_dump.h" #include "st_context.h" #include "st_atom.h" -- cgit v1.2.3 From 28454a512a9b1c1e9f70a04b43c34f239abd6ba7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 28 Jul 2008 22:42:18 +0900 Subject: util: Don't define replacement math functions for CE. It appears to be working without this before, and it is probably necessary. --- src/gallium/include/pipe/p_util.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 892bd4bf8a..eec413842f 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -34,7 +34,7 @@ #include "p_format.h" #include "p_pointer.h" -#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) +#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) __inline double ceil(double val) { double ceil_val; -- cgit v1.2.3 From 76164bf7a20ef6dabc3204a766f604becfba9997 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 30 Jul 2008 00:44:56 +0900 Subject: tgsi: Insert newlines after the statements, instead of before. Prevents shader dumps from getting concatenated with the next debug message. --- src/gallium/auxiliary/tgsi/tgsi_dump.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index d2e6375212..290d78bb14 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -332,7 +332,7 @@ void tgsi_dump_declaration( const struct tgsi_full_declaration *decl ) { - TXT( "\nDCL " ); + TXT( "DCL " ); _dump_register( decl->Declaration.File, @@ -354,6 +354,8 @@ tgsi_dump_declaration( TXT( ", " ); ENM( decl->Declaration.Interpolate, interpolate_names ); + + EOL(); } static boolean @@ -407,7 +409,6 @@ tgsi_dump_instruction( uint i; boolean first_reg = TRUE; - EOL(); UID( instno ); CHR( ':' ); ENM( inst->Instruction.Opcode, opcode_names ); @@ -534,6 +535,8 @@ tgsi_dump_instruction( UID( inst->InstructionExtLabel.Label ); break; } + + EOL(); } static boolean @@ -551,11 +554,11 @@ static boolean prolog( struct tgsi_iterate_context *ctx ) { - EOL(); ENM( ctx->processor.Processor, processor_type_names ); UID( ctx->version.MajorVersion ); CHR( '.' ); UID( ctx->version.MinorVersion ); + EOL(); return TRUE; } -- cgit v1.2.3 From 43ac397fb8d26c335160287294a25ffb5b4bcec5 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 29 Jul 2008 17:47:39 -0600 Subject: mesa: disable debug code --- src/mesa/shader/slang/slang_compile.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 60539f19b6..82df9b95d3 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -2285,7 +2285,7 @@ compile_shader(GLcontext *ctx, slang_code_object * object, GLboolean success; grammar id = 0; -#if 1 /* for debug */ +#if 0 /* for debug */ _mesa_printf("********* COMPILE SHADER ***********\n"); _mesa_printf("%s\n", shader->Source); _mesa_printf("************************************\n"); -- cgit v1.2.3 From ab945fc38c81adc7e033c544213ebb929edbe1f8 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 29 Jul 2008 17:49:47 -0600 Subject: mesa: glsl: remove old assertion (fixes glsl/bitmap.c) --- src/mesa/shader/slang/slang_emit.c | 1 - 1 file changed, 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index f1af6902ea..3d76c80b58 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1021,7 +1021,6 @@ emit_tex(slang_emit_info *emitInfo, slang_ir_node *n) storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); /* Child[1] is the coord */ - assert(n->Children[1]->Store->File != PROGRAM_UNDEFINED); assert(n->Children[1]->Store->Index >= 0); storage_to_src_reg(&inst->SrcReg[0], n->Children[1]->Store); -- cgit v1.2.3 From 4a8b908a46375ea2543a81bdd8d5b313d807f140 Mon Sep 17 00:00:00 2001 From: Younes Manton Date: Sat, 2 Aug 2008 19:37:16 +0200 Subject: softpipe: support PIPE_FORMAT_R16_SNORM. --- src/gallium/auxiliary/util/p_tile.c | 46 +++++++++++++++++++++++++++++++++++++ 1 file changed, 46 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 1bf0d72733..2b6db43bee 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -441,6 +441,46 @@ a8_get_tile_rgba(ubyte *src, } } +/*** PIPE_FORMAT_R16_SNORM ***/ + +static void +r16_get_tile_rgba(short *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = SHORT_TO_FLOAT(src[0]); + pRow[1] = + pRow[2] = 0.0; + pRow[3] = 1.0; + } + p += dst_stride; + } +} + + +static void +r16_put_tile_rgba(short *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, dst++, pRow += 4) { + UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); + } + p += src_stride; + } +} + /*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ @@ -742,6 +782,9 @@ pipe_tile_raw_to_rgba(enum pipe_format format, case PIPE_FORMAT_A8L8_UNORM: a8_l8_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); break; + case PIPE_FORMAT_R16_SNORM: + r16_get_tile_rgba((short *) src, w, h, dst, dst_stride); + break; case PIPE_FORMAT_R16G16B16A16_SNORM: r16g16b16a16_get_tile_rgba((short *) src, w, h, dst, dst_stride); break; @@ -847,6 +890,9 @@ pipe_put_tile_rgba(struct pipe_surface *ps, case PIPE_FORMAT_A8L8_UNORM: /*a8_l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ break; + case PIPE_FORMAT_R16_SNORM: + /*r16_put_tile_rgba((short *) packed, w, h, p, src_stride);*/ + break; case PIPE_FORMAT_R16G16B16A16_SNORM: r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); break; -- cgit v1.2.3 From eb80ed0d2eac693012e7e4c2ad772f7f57f74977 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sun, 3 Aug 2008 11:14:47 -0600 Subject: added null ptr check (fix bug 16959) --- src/mesa/main/context.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 279880cf40..27e5e2fcce 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -711,7 +711,8 @@ delete_renderbuffer_cb(GLuint id, void *data, void *userData) { struct gl_renderbuffer *rb = (struct gl_renderbuffer *) data; rb->RefCount = 0; /* see comment for FBOs above */ - rb->Delete(rb); + if (rb->Delete) + rb->Delete(rb); } -- cgit v1.2.3 From 2783f3bfabd6e316f7e221e950499c3631c041ce Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 4 Aug 2008 12:18:09 +0200 Subject: tgsi: Put a newline after IMM. --- src/gallium/auxiliary/tgsi/tgsi_dump.c | 18 +++++------------- 1 file changed, 5 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index 290d78bb14..0eedb1c91e 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -264,16 +264,6 @@ static const char *modulate_names[TGSI_MODULATE_COUNT] = "_D8" }; -static void -_dump_register_prefix( - uint file, - uint first, - uint last ) -{ - - -} - static void _dump_register( uint file, @@ -354,7 +344,7 @@ tgsi_dump_declaration( TXT( ", " ); ENM( decl->Declaration.Interpolate, interpolate_names ); - + EOL(); } @@ -373,7 +363,7 @@ tgsi_dump_immediate( { uint i; - TXT( "\nIMM " ); + TXT( "IMM " ); ENM( imm->Immediate.DataType, immediate_type_names ); TXT( " { " ); @@ -390,6 +380,8 @@ tgsi_dump_immediate( TXT( ", " ); } TXT( " }" ); + + EOL(); } static boolean @@ -535,7 +527,7 @@ tgsi_dump_instruction( UID( inst->InstructionExtLabel.Label ); break; } - + EOL(); } -- cgit v1.2.3 From 655ca5afad5f5d79f0ad60b5aa7fcafd35e97b62 Mon Sep 17 00:00:00 2001 From: Jonathan White Date: Mon, 4 Aug 2008 16:30:04 -0600 Subject: egl: fix _eglGlobal initialization for Windows --- src/egl/main/eglglobals.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglglobals.c b/src/egl/main/eglglobals.c index 5da59af91d..b770e55dbd 100644 --- a/src/egl/main/eglglobals.c +++ b/src/egl/main/eglglobals.c @@ -2,9 +2,10 @@ #include #include "eglglobals.h" - -struct _egl_global _eglGlobal = { .Initialized = EGL_FALSE }; - +struct _egl_global _eglGlobal = +{ + EGL_FALSE +}; /** * Init the fields in the _eglGlobal struct -- cgit v1.2.3 From 012a1f1cf6bc46b2ac32b456a233cf1ecd25e241 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 09:32:54 -0600 Subject: mesa: glsl: new bvec4 constructor --- src/mesa/shader/slang/library/slang_core.gc | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 1f57997051..0dfc03883f 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -436,6 +436,15 @@ bvec4 __constructor(const bool b1, const bool b2, const bool b3, const bool b4) __retVal.w = b4; } +bvec4 __constructor(const float f1, const float f2, const float f3, const float f4) +{ + const float zero = 0.0; + __asm vec4_sne __retVal.x, f1, zero; + __asm vec4_sne __retVal.y, f2, zero; + __asm vec4_sne __retVal.z, f3, zero; + __asm vec4_sne __retVal.w, f4, zero; +} + bvec4 __constructor(const bool b) { __retVal.xyzw = b.xxxx; -- cgit v1.2.3 From d95bb24ddc9fa4bfb32d07fb96d19c024f0e52cf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 10:04:53 -0600 Subject: mesa: glsl: regenerated file --- src/mesa/shader/slang/library/slang_core_gc.h | 1435 +++++++++++++------------ 1 file changed, 721 insertions(+), 714 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index 44170946ce..15a1d1776e 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -87,737 +87,744 @@ 122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,1,1,0,1,98, 52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86, 97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,4,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,119,0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,4,1,1,1,0,9,102,0, -0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0, -48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0, -0,18,102,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5,105,0,0,0,1,3,2,1,8,1, -122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,59,120,120,120,120,0,0, -18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,12,118,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101, -99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52, -95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1, -1,1,0,8,118,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0, -16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0, -18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9, -109,48,49,0,0,1,1,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0, -18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,0,13,1,1,1,0,9,102,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0, -17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,0,1,0, -13,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,13, -1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,13,1,1,1, -0,10,99,48,0,0,1,1,0,10,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,14,1,1,1,0,9,109,48,48, -0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0, -9,109,50,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, -0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20, +95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,4,1,1,1,0,9,102,49,0,0,1,1,0,9,102,50, +0,0,1,1,0,9,102,51,0,0,1,1,0,9,102,52,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4, +118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,102,49,0,0,18,122, +101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18, +102,50,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, +108,0,59,122,0,0,18,102,51,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, +114,101,116,86,97,108,0,59,119,0,0,18,102,52,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,1,98,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,18,98,0,59,120,120,120,120,0,20,0,0,1, +0,4,1,1,1,0,9,102,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48, +0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114, +101,116,86,97,108,0,0,18,102,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5, +105,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0, +0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0, +59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,12,118,0,0,0,1,3,2,1,12,1,122,101, +114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0, +0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101, +114,111,0,0,0,0,1,0,4,1,1,1,0,8,118,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101,99,52,0, +16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, +114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1, +0,9,109,49,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121, +0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,0,13,1,1,1,0,9, +102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,49,0,57,59,120,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121, +0,18,102,0,20,0,0,1,0,13,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,105, +0,0,0,0,0,0,0,1,0,13,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,98,0,0,0, +0,0,0,0,1,0,13,1,1,1,0,10,99,48,0,0,1,1,0,10,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,14, +1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9, +109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59, +120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122, +0,18,109,50,50,0,20,0,0,1,0,14,1,1,1,0,9,102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102, +0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121, +0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59,121,121,120,0,20,0,0,1,0,14,1,1,1,0,5,105,0,0,0, +1,8,58,109,97,116,51,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,14,1,1,1,0,1,98,0,0,0,1,8, +58,109,97,116,51,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,14,1,1,1,0,11,99,48,0,0,1,1,0, +11,99,49,0,0,1,1,0,11,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,0,15,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9, +109,50,48,0,0,1,1,0,9,109,51,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49, +0,0,1,1,0,9,109,51,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,1,1,0, +9,109,51,50,0,0,1,1,0,9,109,48,51,0,0,1,1,0,9,109,49,51,0,0,1,1,0,9,109,50,51,0,0,1,1,0,9,109,51, +51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119, +0,18,109,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20, 0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101, 116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0, -18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0, -0,1,0,14,1,1,1,0,9,102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0, -0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,118,0,59,121,121,120,0,20,0,0,1,0,14,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,51,0, -58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,14,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,51,0,58, -102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,14,1,1,1,0,11,99,48,0,0,1,1,0,11,99,49,0,0,1,1,0,11, -99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101, +10,49,0,57,59,119,0,18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0, +18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,51,0,57,59,120,0,18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0, +18,109,49,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0,0,1,0,15,1,1,1,0,9, +102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, +0,57,18,118,0,59,121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0, +59,121,121,121,120,0,20,0,0,1,0,15,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97, +116,0,18,105,0,0,0,0,0,0,0,1,0,15,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116, +0,18,98,0,0,0,0,0,0,0,1,0,15,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,1,1,0,12, +99,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101, 116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -99,50,0,20,0,0,1,0,15,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9, -109,51,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,51,49, -0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,1,1,0,9,109,51,50,0,0,1,1,0, -9,109,48,51,0,0,1,1,0,9,109,49,51,0,0,1,1,0,9,109,50,51,0,0,1,1,0,9,109,51,51,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, -0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,109,51,48,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0, -18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0, -18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,109,49,51,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0,0,1,0,15,1,1,1,0,9,102,0,0,0,1,3,2,0,10, -1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59, -121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0,59,121,121,121, -120,0,20,0,0,1,0,15,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,105,0,0, -0,0,0,0,0,1,0,15,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0, -0,0,1,0,15,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,1,1,0,12,99,51,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,5,2,26,1,1,0,5,97,0,0,1,1,0,5,98, -0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102, -108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5, -2,27,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114, -97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0, -9,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0, -0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, -0,1,0,5,2,22,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105, -110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,26,1,1,0,6,97,0,0,1,1,0,6,98,0, -0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102, -108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6, -2,27,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114, -97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,21,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0, -10,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,0,0, -0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, -0,1,0,6,2,22,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116, -111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,26,1,1,0,7,97,0,0,1, -1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0, -0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0, -0,0,1,0,7,2,27,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,115,117, -98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110, -116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,21,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0, -1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0, -18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,0,0,1,0,7,2,22,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,98,73,110,118,0,0,1,1,120,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116, -111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,26,1,1,0,8,97,0,0,1, -1,0,8,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0,18,98,0, -0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0, -0,0,1,0,8,2,27,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,115,117, -98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110, -116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,21,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0, -1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0, +99,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,5,2,26,1,1,0,5, +97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0, 18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,0,0,1,0,8,2,22,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,98,73,110,118,0,0,1,1,120,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,98,73,110,118,0,59,119,0,0,18,98,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105, -110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,2,26,1,1,0,9,97,0,0,1,1,0,9,98,0, -0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98, -0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,21,1,1,0,9,97,0, -0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86, -97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,22,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1, -98,73,110,118,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59, -120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0, -0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,21,1,1,0, -10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1,0,10, -117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0, -59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -118,0,0,18,119,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100, -100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,27,1, -1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1, -0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0, -1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0, -0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,119,0,0,0, -0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114, -101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1, -4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, -117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,22,1,1,0, -12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0, -59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0, -59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0,1,0,10, -2,26,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,26,1,1,0,10, -118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,27,1,1,0,9,97,0,0,1,1,0,10,117, -0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,9,98, -0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,21,1,1,0,9,97,0,0,1,1,0,10,117, -0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0, -9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,22,1,1,0,9,97,0,0,1,1,0, -10,117,0,0,0,1,3,2,0,10,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, -85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59, -121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,105,110,118,85,0,59,120,121,0,0,0,0,1, -0,10,2,22,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,66,0, -59,120,120,0,0,0,0,1,0,11,2,26,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120,121, -122,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0, -0,0,0,1,0,11,2,27,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59, -120,121,122,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98, -116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0, -0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,21,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0, -59,120,120,120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1, -4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,22,1,1,0,9,97,0,0,1,1, -0,11,117,0,0,0,1,3,2,0,11,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, -118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0, -59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0, -0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0, -0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105, -110,118,66,0,59,120,120,120,0,0,0,0,1,0,12,2,26,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99, -52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0, -0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101, -116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,27,1,1,0,9,97,0,0,1,1,0, -12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0, -0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4, -118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, -98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0, -0,18,117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0, -0,0,1,0,12,2,22,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,105,110,118,85,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,105,110,118,85,0,0, -0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,59,120,120,120,120, -0,0,0,0,1,0,6,2,26,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105, -118,101,99,50,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,6,2,27,1,1, -0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0, -0,18,117,0,47,20,0,0,1,0,6,2,27,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,47,20,0,0,1,0,6,2,21,1,1,0,5,97,0,0,1,1,0,6,117, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,48,20,0,0,1, -0,6,2,21,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118, -101,99,50,0,18,98,0,0,0,48,20,0,0,1,0,6,2,22,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,6,2,22,1,1,0,6,118,0,0,1, -1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,49, -20,0,0,1,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, -101,99,51,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,46,20,0,0,1,0,7,2,27,1,1,0, -5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0, -18,117,0,47,20,0,0,1,0,7,2,27,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,47,20,0,0,1,0,7,2,21,1,1,0,5,97,0,0,1,1,0,7,117,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,7, -2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, -99,51,0,18,98,0,0,0,48,20,0,0,1,0,7,2,22,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,7,2,22,1,1,0,7,118,0,0,1,1,0, -5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,49,20, -0,0,1,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, -101,99,52,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,46,20,0,0,1,0,8,2,27,1,1,0, -5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0, -18,117,0,47,20,0,0,1,0,8,2,27,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,47,20,0,0,1,0,8,2,21,1,1,0,5,97,0,0,1,1,0,8,117,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,8, -2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, -99,52,0,18,98,0,0,0,48,20,0,0,1,0,8,2,22,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,8,2,22,1,1,0,8,118,0,0,1,1,0, -5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,49,20, -0,0,1,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,6,2,27,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,110,101,103, -97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,7,2,27,1,1,0,7,118,0,0,0,1,4, -118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,8, -2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97, -108,0,0,18,118,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,0,1,4,118, -101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59, -120,121,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,0,0,1,0,12,2,27,1,1,0,12, -118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118, -0,0,0,0,1,0,13,2,27,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20, -0,0,1,0,14,2,27,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8, -48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,0,1,0,15,2,27,1,1,0, -15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,0,9,0,100,111,116,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,48,20,0,0,1,0,9,0,100,111,116,0,1,1,0,10,97,0,0,1, -1,0,10,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0, -59,121,0,18,98,0,59,121,0,48,46,20,0,0,1,0,9,0,100,111,116,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4, -118,101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0, -100,111,116,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,18,95,95,114, -101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,1,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,5,97,0,0,1,1,0,5,98,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2, -3,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0, -18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, +120,0,0,0,0,1,0,5,2,27,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95, +115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95, +105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5, +98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18, +97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108, +0,0,18,120,0,0,0,0,1,0,5,2,22,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,1,1, +120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116, +95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,26,1,1,0,6, +97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0, +0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0, +18,120,0,0,0,0,1,0,6,2,27,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52, +95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, +95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,21,1,1,0,6,97,0,0,1,1,0, +6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0, +18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97, +108,0,0,18,120,0,0,0,0,1,0,6,2,22,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,98,73,110,118,0,0,1, +1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102, +108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7, +2,26,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120, +0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86, +97,108,0,0,18,120,0,0,0,0,1,0,7,2,27,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118, +101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116, +95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,21,1,1,0,7, +97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114, +101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,22,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,98,73, +110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98, +0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102, +108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8, +2,26,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120, +0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86, +97,108,0,0,18,120,0,0,0,0,1,0,8,2,27,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118, +101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116, +95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,21,1,1,0,8, +97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114, +101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,22,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,98,73, +110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98, +0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,119,0,0,18,98,0,59,119,0,0,0,4,118,101,99,52, +95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97, +116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,2,26,1,1,0, +9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, +115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0, +1,0,9,2,21,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, +18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,22,1,1,0,9,97,0,0,1,1, +0,9,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118, +0,59,120,0,0,18,98,0,59,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95, +114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0, +1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, +0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95, +115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0, +0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,22,1, +1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18, +117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,118,0,0,18,119,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4, +118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18, +117,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116, +114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0, +11,2,21,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,22,1,1,0,11, +118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59, +120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59, +121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, +118,0,0,18,119,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100, +100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1, +1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97, +108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99, +52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0, +0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95, +114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119, +0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117, +0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4, +118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0, +18,119,0,0,0,0,1,0,10,2,26,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1, +0,10,2,26,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,27,1,1,0,9, +97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,27,1,1,0,10, +118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,21,1,1,0,9, +97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,21,1,1, +0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,22, +1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,105,110,118,85,0, +59,120,121,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, +4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0, +18,105,110,118,66,0,59,120,120,0,0,0,0,1,0,11,2,26,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101, +99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0, +0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52, +95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18, +98,0,59,120,120,120,0,0,0,0,1,0,11,2,27,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120, +120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101, +99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118, +0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,21,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0, +1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1, +1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,22,1, +1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, +118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,105,110,118,85,0,59, +120,121,122,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, 4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,105,110,118,66,0,0,0,4,102,108,111,97,116,95,116, -111,95,105,110,116,0,18,97,0,0,18,97,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,6, -117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0, -1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, -18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,3,2,0,6,1,105, -110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117, -0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0, -0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0, -0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,7,118, -0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0, -2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0, -0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0, -7,117,0,0,0,1,3,2,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105, +108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121, +122,0,0,18,105,110,118,66,0,59,120,120,120,0,0,0,0,1,0,12,2,26,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1, +4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0, +18,117,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, +95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,27,1,1,0,9, +97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, +86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9, +98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18, +118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120, +120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120, +120,120,0,0,0,0,1,0,12,2,22,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,105,110,118,85,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108, +111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95, +114,99,112,0,18,105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,105,110, +118,85,0,0,0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,59,120, +120,120,120,0,0,0,0,1,0,6,2,26,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,6, +2,27,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0, +18,97,0,0,0,18,117,0,47,20,0,0,1,0,6,2,27,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,47,20,0,0,1,0,6,2,21,1,1,0,5,97,0,0,1, +1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0, +48,20,0,0,1,0,6,2,21,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118, +0,58,105,118,101,99,50,0,18,98,0,0,0,48,20,0,0,1,0,6,2,22,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,6,2,22,1,1, +0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0, +18,98,0,0,0,49,20,0,0,1,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,46,20,0,0,1, +0,7,2,27,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99, +51,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,7,2,27,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,47,20,0,0,1,0,7,2,21,1,1,0,5,97,0, +0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117, +0,48,20,0,0,1,0,7,2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,58,105,118,101,99,51,0,18,98,0,0,0,48,20,0,0,1,0,7,2,22,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,7,2,22, +1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99, +51,0,18,98,0,0,0,49,20,0,0,1,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5, +98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,46,20,0, +0,1,0,8,2,27,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101, +99,52,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,8,2,27,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,47,20,0,0,1,0,8,2,21,1,1,0,5, +97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0, +18,117,0,48,20,0,0,1,0,8,2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,48,20,0,0,1,0,8,2,22,1,1,0,5,97,0,0,1,1,0,8,117,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,8, +2,22,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, +99,52,0,18,98,0,0,0,49,20,0,0,1,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,6,2,27,1,1,0,6,118,0,0,0,1,4, +118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,7, +2,27,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97, +108,0,0,18,118,0,0,0,0,1,0,8,2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0, +18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, +110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,2,27,1,1, +0,10,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,110, +101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0, +0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114, +101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,13,2,27,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, +18,109,0,16,10,49,0,57,54,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0, +57,54,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, +109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,0,9,0,100,111,116,0,1,1,0, +9,97,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,48,20,0,0,1,0,9,0, +100,111,116,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,59, +120,0,18,98,0,59,120,0,48,18,97,0,59,121,0,18,98,0,59,121,0,48,46,20,0,0,1,0,9,0,100,111,116,0,1,1, +0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0, +0,18,97,0,0,18,98,0,0,0,0,1,0,9,0,100,111,116,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99, +52,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,1,1,0,2,5, +97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0, +2,2,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,0, +18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0, +1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98, +0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,105,110,118,66,0, +0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,97,0,0,18,97,0,0,0,0,1,0,0,2,1,1,0,2,6,118, +0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0, +2,2,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0, +0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0, +6,117,0,0,0,1,3,2,0,6,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105, 110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0, 59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0, 18,118,0,0,18,105,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,118,0,0,18,122, -0,0,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0, -18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,115,117, -98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,8,117, +0,0,0,0,1,0,0,2,1,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0, +18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117, +98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,7,117, 0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0, -1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,3,2,0,8,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108, +1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,3,2,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108, 111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, 95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108, 116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,102,108,111,97,116,95,116,111, -95,105,110,116,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101, -99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,0,1,0,0,2,2,1,0,2, -9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,59,120,0,0, -18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,9,97,0,0,1,1,0,9, -98,0,0,0,1,3,2,0,9,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,98,0,0, -0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,119,0,0,0,0, -1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120, -121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,10, -117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59, -120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117, -0,59,120,121,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102, +95,105,110,116,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101, +99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0, +2,3,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,3,2,0,8,1,105,110, +118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59, +120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4, +118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4, +102,108,111,97,116,95,116,111,95,105,110,116,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,9,97,0,0, +1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0, +59,120,0,0,0,0,1,0,0,2,2,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114, +97,99,116,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1, +4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1, +0,0,2,4,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,119,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59, +120,0,0,18,97,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52, +95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0, +2,2,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118, +0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1, +1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18, +118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,3, +2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0, +4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,119,0,59,120, +121,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118, +0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4, +118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0, +0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,11, +117,0,0,0,1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0, +59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,119,0,59, +120,121,122,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100, +0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101, +99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,12, +118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18, +118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102, 108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,119,0,59,120,121,0,0,0,0,1,0,0,2,1,1, -0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18, -118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117, -98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,11, -118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120, -121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1, -119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,119,0,59,120,121,122,0,0,0,0,1,0, -0,2,1,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0, -18,117,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116, -114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0, -1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0, -2,4,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112, -0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0, -0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0, -0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0, -0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0, -0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0, -0,0,1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,9,18,118,0,59,120,0, -18,97,0,23,0,9,18,118,0,59,121,0,18,97,0,23,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,9,18, -118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,7,118,0,0,1,1,0,5, -97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0, -18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,59,120, -120,120,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0, -0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59, -121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1, -4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,2, -1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18, -118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0, -0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121, -0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,9,18,118,0,59,119,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2, -10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,0, -18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,3, -1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, -59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1, -3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0, -0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120, -121,0,0,18,105,110,118,65,0,59,120,120,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1, -0,0,2,2,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, -118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1, -0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18, -118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105, +114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119, +0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117, +0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0, +0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120, +121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,5,97,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120, +121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120, +120,0,0,0,9,18,118,0,59,120,0,18,97,0,23,0,9,18,118,0,59,121,0,18,97,0,23,0,0,1,0,0,2,4,1,0,2,6, +118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,0,1,0, +0,2,1,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0, +0,18,118,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,5,97, +0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,59, +120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122, +0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120, +0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2, +8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120, +120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116, +114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0, +1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18, +97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0, +18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,9,18,118,0,59,119,0, +18,97,0,24,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, +118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,9,97,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0, +59,120,120,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,4,1,0, +2,10,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112, +0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, +59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,65,0,59,120,120,0,0,0,0,1,0,0,2,1,1,0,2,11, +118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0, +18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, +115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0, +0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0, +1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, +118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121, +122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,65,0,59,120,120,120,0,0,0,0,1,0,0,2,1,1,0,2,12, +118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120, +120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,12, +118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118, +0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105, 110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0, -18,105,110,118,65,0,59,120,120,120,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101, -99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,12, -118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0, -18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0, -0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, -0,18,118,0,0,18,118,0,0,18,105,110,118,65,0,59,120,120,120,120,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0, -0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,0, -13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0, -57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57, -18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,0, -48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0, -0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50, -0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20, -0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0, -16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0, -48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10, -49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,0,48,46, -18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120,120,0,48,18,109, -0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16, -10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49, -20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15, -109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, +99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,105,110,118,65,0,59,120,120, +120,120,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13, +109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, 18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10, -50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, -16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, +57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48, +18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49, +0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49, +0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1, +0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, +16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9, 18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120, -120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,121,0,48,46,18,109,0, -16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16, -8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, -0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16, -10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122, -122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120, -120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,121,0,48,46,18,109,0,16, -10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10, -50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, -16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10, -51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,51,0,57,59,122,122,122, -122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59,119,119,119,119,0,48,46,20,0,0,1,0,15, -2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, -0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0, -13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13, -2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97, +120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50, +0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, +0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122, +122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110, +0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0, +48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1, +0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, +16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0, +57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10, +51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2, +21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48, +0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0, +48,46,18,109,0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120, +0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, +57,18,110,0,16,10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57, +59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48, +0,57,18,110,0,16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57, +59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48, +46,18,109,0,16,10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120, +0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, +57,18,110,0,16,10,51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, +59,119,119,119,119,0,48,46,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49, +20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, +0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0, +13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13, +2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, +57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97, 0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48, -0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20, -0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9, +0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20, +0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, +16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2,26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0, +1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, +47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2, +27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47, +20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, +0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0, +16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9, 98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2, -26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, -110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, -49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46, -20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, -0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50, -0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14, -110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0, -14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0, -1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110, -0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50, -0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109, -0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0, -57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2,27,1,1,0,9,97,0,0,1,1,0,15,110,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,9, -98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0,0,1,0,15,2,21,1,1,0,9,97,0,0, -1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,15,2,21,1,1,0, -15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,0,15,2, -22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, -110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, -49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,49,20,0,0,1,0, -15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, -0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, -0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,49,20,0, -0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16, -8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,0,48,46,20,0,0,1,0, -10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100, -111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, -100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21,1,1,0,14,109,0,0,1,1,0,11, -118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,0,48, -18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122, -122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,12, -2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0, -57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,121,0,48,46,18, -109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,118,0,59,119,119, -119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0, -0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21, -0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,13,110,0, -0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10, -49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20, -0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0,14, -110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0, -16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2,2,1,0,2,14,109,0, -0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,0,0,2,3,1,0,2, -14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,14,109,0,0, -1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,0,0,2,1,1,0,2, -15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16, -10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18, -109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1, -9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57, -22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10, -51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20, -0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, -50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1, -0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,0,0, -2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0, -57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23, -0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,14,109,0,0,1,1,0, -9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0, -16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57, -18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,0,0,2, -3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0, -57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,14,109,0,0,1,1,0,9,97,0,0, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9, +97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8, +48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1, +1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98, +0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2, +27,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, +110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, +49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0, +15,2,27,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, +0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, +0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0, +0,1,0,15,2,21,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18, +110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16, +10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57, +48,20,0,0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16, +10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57, +18,98,0,48,20,0,0,1,0,15,2,22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, +57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0, +18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0, +16,10,51,0,57,49,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +18,109,0,16,10,50,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, +16,10,51,0,57,18,98,0,49,20,0,0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0, +59,121,121,0,48,46,20,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21, +1,1,0,14,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18, +118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50, +0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, +57,0,0,20,0,0,1,0,12,2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121, +121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57, +18,118,0,59,119,119,119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, +57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10, +51,0,57,0,0,20,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110, +0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109, +0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0, +57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109, +0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14, +109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2, +2,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18, +109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22, +0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2, +4,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18, +109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24, +0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, +50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1, +0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51, +0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18, +109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57, +18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10, +50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2, +1,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, +57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22, +0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, +16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0, +9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2, +1,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, +57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0, +0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50, +0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2, +14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97, +0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18, +109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18, +97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9, +18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57, +18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1, +9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0, +57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0, 0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50, -0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, -21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10, -51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97, -0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16, -10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18, -97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,9,18,109,0, -16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57, -18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,9,18,109, -0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,118,0,18,118,0, -18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48, -20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0, -5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0, -18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49, -0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,25,1,0,2,7,118,0,0,0,1,9, -18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0, -18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49, -0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,25,1,0,2,9,97,0,0,0,1,9, -18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2, -25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, -12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, -15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, -49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, -48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, -0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2,24,1,0,2,5,97,0,0,0,1,9,18,97, -0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,24,1,0,2,6, -118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101, -116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101, -99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,24,1,0,2, -8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101, -116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24,1,0,2,10,118,0,0,0,1,9,18,118,0, -18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0, -20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, -46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,12,2,24,1,0,2,12,118,0,0,0,1,9,18, -118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118, -101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101, -99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,24, -1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0, -0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0, -0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,15,2,24,1,0,2,15,109,0,0,0,1,9,18,109, -0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10, -50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0, -57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,47,20,0,0,1,0,6,0,95,95,112, -111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, -18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,0,1,0,7,0,95,95,112,111,115,116,68, -101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, -118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95,95,112,111,115,116,68,101,99,114, -0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105, -118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,9,97, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,0, -1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,11,0, -95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,12,0,95,95,112, -111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0, -9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,13,0,95,95,112,111,115,116, -68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10, -49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,14,0,95,95,112, -111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0, -9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18, -109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109, -0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,15,0,95, -95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109, -0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0, -9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18, -109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109, -0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,9,0,95, -95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0, -20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2, -10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99, -50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,11,118,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49, -0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0, -46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0,95,95,112,111,115,116,73, -110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, -118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112,111,115,116,73,110,99,114, -0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105, -118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,8,118, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51, -0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,13,109,0,0,0,1,3,2, -0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49, -0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48, -0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,14,109,0,0,0,1, -3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0, -17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17, -49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0, -48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,15,109,0,0, -0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52, -0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, -49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, -48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, -103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0,0,0,0,1,0,1,2,15,1,1,0,5, -97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0, -0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0,0,4,102,108,111,97,116,95, -108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2,16,1,1,0,5,97,0,0,1,1,0,5, -98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,41,0,0,1,0,1, -2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108, -101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101, -0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1, -8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,43,0,0,1,0,1,2,17,1,1,0, -9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115, -115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18, -97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58, -102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0,0,1,0,0,0,112,114,105, -110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114,105,110,116,0,18,102,0,0, -0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105,110,116,95,112,114,105,110, -116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1,98,0,0,0,1,4,98,111,111,108, -95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,10,118,0,0, -0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69, -83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,11,118,0,0,0,1,9, -58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65, -0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0, -112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, -59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,6,118,0,0,0,1,9,58,112,114,105, -110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59, -121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7,118,0,0,0,1,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0, -0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116, -77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, -58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65, -0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0, -112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112, -114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, +0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109, +0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9, +18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18, +118,0,18,109,0,48,20,0,0,1,0,5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18, +95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58, +105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +7,2,25,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18, +95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0, +58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1, +0,9,2,25,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,18,97,0,20,0,0,1,0,10,2,25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49, +0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0, +0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,118,0,20,0,0,1,0,12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0, +17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13, +109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47, +20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0, +9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16, +8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49, +0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57, +18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,18,109,0,20,0,0,1,0,15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, +118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118, +101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2, +24,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, +97,0,20,0,0,1,0,6,2,24,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0, +0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18, +118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,0,1,0,8,2,24,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0, +0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18, +97,0,18,97,0,17,49,0,48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24, +1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95, +114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +12,2,24,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0, +57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,18,109,0,20,0,0,1,0,14,2,24,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, +15,2,24,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, +0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101, +99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16, +10,49,0,47,20,0,0,1,0,6,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20, +0,0,1,0,7,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95, +95,112,111,115,116,68,101,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0, +20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115, +116,68,101,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0, +18,97,0,17,49,0,48,0,0,47,20,0,0,1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0, +48,0,0,0,0,47,20,0,0,1,0,11,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, +47,20,0,0,1,0,12,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0, +1,0,13,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0, +0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0, +0,47,20,0,0,1,0,14,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17, +49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0, +48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0, +0,0,0,47,20,0,0,1,0,15,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0, +17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, +49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115, +116,73,110,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118, +0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110, +99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0, +2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101, +99,52,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0, +95,95,112,111,115,116,73,110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112, +111,115,116,73,110,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, +18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73, +110,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, +118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114, +0,1,0,2,13,109,0,0,0,1,3,2,0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110, +99,114,0,1,0,2,14,109,0,0,0,1,3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110, +99,114,0,1,0,2,15,109,0,0,0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, +58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58, +118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0, +0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0, +0,0,0,1,0,1,2,15,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102, +108,111,97,116,0,18,98,0,0,0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0, +0,4,102,108,111,97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2, +16,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0, +18,98,0,0,0,41,0,0,1,0,1,2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4, +102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95, +101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0, +5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0, +0,0,43,0,0,1,0,1,2,17,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108, +111,97,116,95,108,101,115,115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113, +117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0, +1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0, +0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114, +105,110,116,0,18,102,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105, +110,116,95,112,114,105,110,116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1, +98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77, +69,83,65,0,1,1,0,10,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, +58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, +83,65,0,1,1,0,11,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58, +112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, +18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112, +114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114, +105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1, +0,6,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105, +110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7, +118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110, +116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0, +0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77, +69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0, +9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83, +65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58, +112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, +18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112, +114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112, +114,105,110,116,77,69,83,65,0,1,1,0,4,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, 59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,4,118, -0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77, -69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0, -9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, -83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9, -58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77, -69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0, -9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69, -83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,15,109,0,0, -0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77, -69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0, -57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105, -110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1, -0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116, -0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95, -112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0, -1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0, -1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110, -116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 +116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0, +0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77, +69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0, +57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116, +77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49, +0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114, +105,110,116,77,69,83,65,0,1,1,0,15,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16, +8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114, +105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, +109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105, +110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17, +101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77, +69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112, +114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0, +0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105, +110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110, +116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 -- cgit v1.2.3 From 7c59a2b5a2699207e25638ff5be6b9cfd64560cc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 10:07:26 -0600 Subject: mesa: glsl: only allow one #version directive --- src/mesa/shader/slang/library/slang_pp_version.syn | 11 ++++++----- src/mesa/shader/slang/slang_preprocess.c | 4 ++++ 2 files changed, 10 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_pp_version.syn b/src/mesa/shader/slang/library/slang_pp_version.syn index d5e9317b5d..3fe1a57ba2 100644 --- a/src/mesa/shader/slang/library/slang_pp_version.syn +++ b/src/mesa/shader/slang/library/slang_pp_version.syn @@ -31,11 +31,9 @@ .syntax version_directive; version_directive - version_directive_1 .and .loop version_directive_2; + version_directive_1; version_directive_1 prior_optional_spaces .and optional_version_directive .and .true .emit $; -version_directive_2 - prior_optional_spaces .and version_directive_body .and .true .emit $; optional_version_directive version_directive_body .or .true .emit 10 .emit 1; @@ -45,13 +43,16 @@ version_directive_body new_line; version_number - version_number_110 .or version_number_120; + version_number_100 .or version_number_110 .or version_number_120; + +version_number_100 + leading_zeroes .and "100" .emit 0 .emit 1; version_number_110 leading_zeroes .and "110" .emit 10 .emit 1; version_number_120 - leading_zeroes .and "120" .emit 20 .emit 1; + leading_zeroes .and "120" .emit 20 .emit 1; leading_zeroes .loop zero; diff --git a/src/mesa/shader/slang/slang_preprocess.c b/src/mesa/shader/slang/slang_preprocess.c index bfe29b6bc9..786f6467c9 100644 --- a/src/mesa/shader/slang/slang_preprocess.c +++ b/src/mesa/shader/slang/slang_preprocess.c @@ -537,7 +537,11 @@ pp_state_init (pp_state *self, slang_info_log *elog) { self->line = 0; self->file = 1; +#if FEATURE_es2_glsl + self->version = 100; +#else self->version = 110; +#endif pp_symbols_init (&self->symbols); pp_ext_init (&self->ext); self->elog = elog; -- cgit v1.2.3 From e68a3ef1efc3873afb64acd15a0ea27a6ba54bda Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 10:21:55 -0600 Subject: mesa: glsl: added initializer size/type error checking --- src/mesa/shader/slang/slang_codegen.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index ad9cf06e27..92c96bef9e 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2687,11 +2687,12 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) if (!rhs) return NULL; - assert(rhs); + if (!rhs->Store || var->Store->Size != rhs->Store->Size) { + slang_info_log_error(A->log, "invalid assignment (wrong types)"); + return NULL; + } + init = new_node2(IR_MOVE, var, rhs); - /* - assert(rhs->Opcode != IR_SEQ); - */ n = new_seq(varDecl, init); } else { -- cgit v1.2.3 From 26ec3780b85765a681f3f365137ee2ecd63d37fc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 10:58:13 -0600 Subject: mesa: glsl: fix initialize size error check --- src/mesa/shader/slang/slang_codegen.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 92c96bef9e..70feed7326 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2687,7 +2687,7 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) if (!rhs) return NULL; - if (!rhs->Store || var->Store->Size != rhs->Store->Size) { + if (rhs->Store && var->Store->Size != rhs->Store->Size) { slang_info_log_error(A->log, "invalid assignment (wrong types)"); return NULL; } -- cgit v1.2.3 From 901c4db2ed7b98d2026fb9f0374a48eca9a8019f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 11:02:22 -0600 Subject: mesa: glsl: handle user-defined const vars in expression simplification --- src/mesa/shader/slang/slang_simplify.c | 46 ++++++++++++++++++++++++++++++---- 1 file changed, 41 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c index c4bf0175be..bc651962a1 100644 --- a/src/mesa/shader/slang/slang_simplify.c +++ b/src/mesa/shader/slang/slang_simplify.c @@ -94,6 +94,15 @@ _slang_lookup_constant(const char *name) } +static slang_operation_type +literal_type(slang_operation_type t1, slang_operation_type t2) +{ + if (t1 == SLANG_OPER_LITERAL_FLOAT || t2 == SLANG_OPER_LITERAL_FLOAT) + return SLANG_OPER_LITERAL_FLOAT; + else + return SLANG_OPER_LITERAL_INT; +} + /** * Recursively traverse an AST tree, applying simplifications wherever @@ -114,6 +123,7 @@ _slang_simplify(slang_operation *oper, if (oper->type == SLANG_OPER_IDENTIFIER) { /* see if it's a named constant */ GLint value = _slang_lookup_constant((char *) oper->a_id); + /*printf("value[%s] = %d\n", (char*) oper->a_id, value);*/ if (value >= 0) { oper->literal[0] = oper->literal[1] = @@ -122,6 +132,28 @@ _slang_simplify(slang_operation *oper, oper->type = SLANG_OPER_LITERAL_INT; return; } + /* look for user-defined constant */ + { + slang_variable *var; + var = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE); + if (var) { + if (var->type.qualifier == SLANG_QUAL_CONST && + var->initializer && + (var->initializer->type == SLANG_OPER_LITERAL_INT || + var->initializer->type == SLANG_OPER_LITERAL_FLOAT)) { + oper->literal[0] = var->initializer->literal[0]; + oper->literal[1] = var->initializer->literal[1]; + oper->literal[2] = var->initializer->literal[2]; + oper->literal[3] = var->initializer->literal[3]; + oper->type = var->initializer->type; + /* + printf("value[%s] = %f\n", + (char*) oper->a_id, oper->literal[0]); + */ + return; + } + } + } } /* first, simplify children */ @@ -146,8 +178,9 @@ _slang_simplify(slang_operation *oper, = oper->children[0].literal[i] + oper->children[1].literal[i]; } oper->literal_size = oper->children[0].literal_size; - slang_operation_destruct(oper); - oper->type = SLANG_OPER_LITERAL_FLOAT; + oper->type = literal_type(oper->children[0].type, + oper->children[1].type); + slang_operation_destruct(oper); /* frees unused children */ return; case SLANG_OPER_SUBTRACT: for (i = 0; i < 4; i++) { @@ -155,8 +188,9 @@ _slang_simplify(slang_operation *oper, = oper->children[0].literal[i] - oper->children[1].literal[i]; } oper->literal_size = oper->children[0].literal_size; + oper->type = literal_type(oper->children[0].type, + oper->children[1].type); slang_operation_destruct(oper); - oper->type = SLANG_OPER_LITERAL_FLOAT; return; case SLANG_OPER_MULTIPLY: for (i = 0; i < 4; i++) { @@ -164,8 +198,9 @@ _slang_simplify(slang_operation *oper, = oper->children[0].literal[i] * oper->children[1].literal[i]; } oper->literal_size = oper->children[0].literal_size; + oper->type = literal_type(oper->children[0].type, + oper->children[1].type); slang_operation_destruct(oper); - oper->type = SLANG_OPER_LITERAL_FLOAT; return; case SLANG_OPER_DIVIDE: for (i = 0; i < 4; i++) { @@ -173,8 +208,9 @@ _slang_simplify(slang_operation *oper, = oper->children[0].literal[i] / oper->children[1].literal[i]; } oper->literal_size = oper->children[0].literal_size; + oper->type = literal_type(oper->children[0].type, + oper->children[1].type); slang_operation_destruct(oper); - oper->type = SLANG_OPER_LITERAL_FLOAT; return; default: ; /* nothing */ -- cgit v1.2.3 From 1028458354773d748278d7719f695eb4c2ffc090 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 11:14:12 -0600 Subject: mesa: glsl: check that rhs of const var initializer is also const --- src/mesa/shader/slang/slang_codegen.c | 38 +++++++++++++++++++++++++++++++++++ 1 file changed, 38 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 70feed7326..691d10b8f1 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2621,6 +2621,32 @@ _slang_gen_return(slang_assemble_ctx * A, slang_operation *oper) } +/** + * Determine if the given operation/expression is const-valued. + */ +static GLboolean +_slang_is_constant_expr(const slang_operation *oper) +{ + slang_variable *var; + GLuint i; + + switch (oper->type) { + case SLANG_OPER_IDENTIFIER: + var = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE); + if (var && var->type.qualifier == SLANG_QUAL_CONST) + return GL_TRUE; + return GL_FALSE; + default: + for (i = 0; i < oper->num_children; i++) { + if (!_slang_is_constant_expr(&oper->children[i])) + return GL_FALSE; + } + return GL_TRUE; + } +} + + + /** * Generate IR tree for a variable declaration. */ @@ -2671,6 +2697,18 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) slang_info_log_error(A->log, "undefined variable '%s'", varName); return NULL; } + + if (v->type.qualifier == SLANG_QUAL_CONST) { + /* if the variable is const, the initializer must be a const + * expression as well. + */ + if (!_slang_is_constant_expr(v->initializer)) { + slang_info_log_error(A->log, + "initializer for %s not constant", varName); + return NULL; + } + } + #if 0 /* XXX make copy of this initializer? */ { -- cgit v1.2.3 From 02d223a7812c5cdf4bac5a83a8c9ce94a88fdba1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 13:05:49 -0600 Subject: mesa: glsl: stop code emit after error is recorded --- src/mesa/shader/slang/slang_emit.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 3d76c80b58..323fc511a6 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1633,12 +1633,18 @@ emit(slang_emit_info *emitInfo, slang_ir_node *n) if (!n) return NULL; + if (emitInfo->log->error_flag) { + return NULL; + } + switch (n->Opcode) { case IR_SEQ: /* sequence of two sub-trees */ assert(n->Children[0]); assert(n->Children[1]); emit(emitInfo, n->Children[0]); + if (emitInfo->log->error_flag) + return NULL; inst = emit(emitInfo, n->Children[1]); #if 0 assert(!n->Store); -- cgit v1.2.3 From 95ecc33330b3eea4ddc21e2345adf2a9c04a38c3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 13:06:19 -0600 Subject: mesa: glsl: add missing code emit for struct/field --- src/mesa/shader/slang/slang_emit.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index 323fc511a6..b1ebad7f17 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1621,6 +1621,10 @@ emit_struct_field(slang_emit_info *emitInfo, slang_ir_node *n) return NULL; } } + else { + /* do codegen for struct */ + emit(emitInfo, n->Children[0]); + } return NULL; /* no instruction */ } -- cgit v1.2.3 From c610c0d77d2e2d9fe0052f7d1316b3d348b536a0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 13:06:48 -0600 Subject: mesa: glsl: set literal_size for const vars --- src/mesa/shader/slang/slang_simplify.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_simplify.c b/src/mesa/shader/slang/slang_simplify.c index bc651962a1..6ef95c48f9 100644 --- a/src/mesa/shader/slang/slang_simplify.c +++ b/src/mesa/shader/slang/slang_simplify.c @@ -145,6 +145,7 @@ _slang_simplify(slang_operation *oper, oper->literal[1] = var->initializer->literal[1]; oper->literal[2] = var->initializer->literal[2]; oper->literal[3] = var->initializer->literal[3]; + oper->literal_size = var->initializer->literal_size; oper->type = var->initializer->type; /* printf("value[%s] = %f\n", -- cgit v1.2.3 From b57c53b2bc68e23796d59e66cdc5b45899c5585c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 13:07:05 -0600 Subject: mesa: glsl: added null ptr check --- src/mesa/shader/slang/slang_codegen.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 691d10b8f1..117e82a390 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -3385,6 +3385,8 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) for (i = 0; i < oper->num_children; i++) { slang_ir_node *n = _slang_gen_operation(A, &oper->children[i]); tree = new_seq(tree, n); + if (n) + tree->Store = n->Store; } if (oper->type == SLANG_OPER_NON_INLINED_CALL) { tree = new_function_call(tree, oper->label); -- cgit v1.2.3 From 250910a59d16cccb50ad8ebcb21da397fb8dcf91 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 13:11:27 -0600 Subject: mesa: glsl: error on const-qualified array declarations --- src/mesa/shader/slang/slang_codegen.c | 18 +++++++++++++----- 1 file changed, 13 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 117e82a390..d0453751f4 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2408,11 +2408,19 @@ _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) if (var->array_len > 0) { /* this is an array */ - /* round up element size to mult of 4 */ - GLint sz = (n->Store->Size + 3) & ~3; - /* mult by array size */ - sz *= var->array_len; - n->Store->Size = sz; + /* cannot be const-qualified */ + if (var->type.qualifier == SLANG_QUAL_CONST) { + slang_info_log_error(A->log, "array '%s' cannot be const", + (char*) var->a_name); + return NULL; + } + else { + /* round up element size to mult of 4 */ + GLint sz = (n->Store->Size + 3) & ~3; + /* mult by array size */ + sz *= var->array_len; + n->Store->Size = sz; + } } assert(n->Store->Size > 0); -- cgit v1.2.3 From e194c457fdabd7474da3b277245d7761311dd758 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 13:20:29 -0600 Subject: mesa: glsl: don't allow comparision of arrays --- src/mesa/shader/slang/slang_codegen.c | 48 ++++++++++++++++++++++------------- 1 file changed, 30 insertions(+), 18 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index d0453751f4..a9bd6369dc 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -3104,6 +3104,30 @@ _slang_gen_array_element(slang_assemble_ctx * A, slang_operation *oper) } +static slang_ir_node * +_slang_gen_compare(slang_assemble_ctx *A, slang_operation *oper, + slang_ir_opcode opcode) +{ + slang_typeinfo t0, t1; + + slang_typeinfo_construct(&t0); + _slang_typeof_operation(A, &oper->children[0], &t0); + + slang_typeinfo_construct(&t1); + _slang_typeof_operation(A, &oper->children[0], &t1); + + if (t0.spec.type == SLANG_SPEC_ARRAY || + t1.spec.type == SLANG_SPEC_ARRAY) { + slang_info_log_error(A->log, "Illegal array comparison"); + return NULL; + } + + return new_node2(opcode, + _slang_gen_operation(A, &oper->children[0]), + _slang_gen_operation(A, &oper->children[1])); +} + + #if 0 static void print_vars(slang_variable_scope *s) @@ -3212,29 +3236,17 @@ _slang_gen_operation(slang_assemble_ctx * A, slang_operation *oper) return new_node0(IR_KILL); case SLANG_OPER_EQUAL: - return new_node2(IR_EQUAL, - _slang_gen_operation(A, &oper->children[0]), - _slang_gen_operation(A, &oper->children[1])); + return _slang_gen_compare(A, oper, IR_EQUAL); case SLANG_OPER_NOTEQUAL: - return new_node2(IR_NOTEQUAL, - _slang_gen_operation(A, &oper->children[0]), - _slang_gen_operation(A, &oper->children[1])); + return _slang_gen_compare(A, oper, IR_NOTEQUAL); case SLANG_OPER_GREATER: - return new_node2(IR_SGT, - _slang_gen_operation(A, &oper->children[0]), - _slang_gen_operation(A, &oper->children[1])); + return _slang_gen_compare(A, oper, IR_SGT); case SLANG_OPER_LESS: - return new_node2(IR_SLT, - _slang_gen_operation(A, &oper->children[0]), - _slang_gen_operation(A, &oper->children[1])); + return _slang_gen_compare(A, oper, IR_SLT); case SLANG_OPER_GREATEREQUAL: - return new_node2(IR_SGE, - _slang_gen_operation(A, &oper->children[0]), - _slang_gen_operation(A, &oper->children[1])); + return _slang_gen_compare(A, oper, IR_SGE); case SLANG_OPER_LESSEQUAL: - return new_node2(IR_SLE, - _slang_gen_operation(A, &oper->children[0]), - _slang_gen_operation(A, &oper->children[1])); + return _slang_gen_compare(A, oper, IR_SLE); case SLANG_OPER_ADD: { slang_ir_node *n; -- cgit v1.2.3 From f70f6e076c9f1a2c06ec433ce93da5913eef457a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 15:21:25 -0600 Subject: mesa: glsl: additional type checking for ?: and = operators --- src/mesa/shader/slang/slang_codegen.c | 75 ++++++++++++++++++++++++++++++++--- 1 file changed, 69 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index a9bd6369dc..86db53c6ac 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2459,16 +2459,36 @@ _slang_gen_select(slang_assemble_ctx *A, slang_operation *oper) { slang_ir_node *cond, *ifNode, *trueExpr, *falseExpr, *trueNode, *falseNode; slang_ir_node *tmpDecl, *tmpVar, *tree; - slang_typeinfo type; - int size; + slang_typeinfo type0, type1, type2; + int size, isBool, isEqual; assert(oper->type == SLANG_OPER_SELECT); assert(oper->num_children == 3); + /* type of children[0] must be boolean */ + slang_typeinfo_construct(&type0); + _slang_typeof_operation(A, &oper->children[0], &type0); + isBool = (type0.spec.type == SLANG_SPEC_BOOL); + slang_typeinfo_destruct(&type0); + if (!isBool) { + slang_info_log_error(A->log, "selector type is not boolean"); + return NULL; + } + + slang_typeinfo_construct(&type1); + slang_typeinfo_construct(&type2); + _slang_typeof_operation(A, &oper->children[1], &type1); + _slang_typeof_operation(A, &oper->children[2], &type2); + isEqual = slang_type_specifier_equal(&type1.spec, &type2.spec); + slang_typeinfo_destruct(&type1); + slang_typeinfo_destruct(&type2); + if (!isEqual) { + slang_info_log_error(A->log, "incompatible types for ?: operator"); + return NULL; + } + /* size of x or y's type */ - slang_typeinfo_construct(&type); - _slang_typeof_operation(A, &oper->children[1], &type); - size = _slang_sizeof_type_specifier(&type.spec); + size = _slang_sizeof_type_specifier(&type1.spec); assert(size > 0); /* temporary var */ @@ -2800,6 +2820,32 @@ _slang_gen_swizzle(slang_ir_node *child, GLuint swizzle) } +/** + * Check if an assignment of type t1 to t0 is legal. + * XXX more cases needed. + */ +static GLboolean +_slang_assignment_compatible(const slang_typeinfo *t0, + const slang_typeinfo *t1) + +{ +#if 01 /* not used just yet - causes problems elsewhere */ + if (t0->spec.type == SLANG_SPEC_INT && + t1->spec.type == SLANG_SPEC_FLOAT) + return GL_FALSE; + + if (t0->spec.type == SLANG_SPEC_BOOL && + t1->spec.type == SLANG_SPEC_FLOAT) + return GL_FALSE; + + if (t0->spec.type == SLANG_SPEC_BOOL && + t1->spec.type == SLANG_SPEC_INT) + return GL_FALSE; +#endif + return GL_TRUE; +} + + /** * Generate IR tree for an assignment (=). */ @@ -2845,8 +2891,25 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper) } else { slang_ir_node *n, *lhs, *rhs; - lhs = _slang_gen_operation(A, &oper->children[0]); + /* lhs and rhs type checking */ + { + slang_typeinfo t0, t1; + + slang_typeinfo_construct(&t0); + _slang_typeof_operation(A, &oper->children[0], &t0); + + slang_typeinfo_construct(&t1); + _slang_typeof_operation(A, &oper->children[1], &t1); + + if (!_slang_assignment_compatible(&t0, &t1)) { + slang_info_log_error(A->log, + "illegal types in assignment"); + return NULL; + } + } + + lhs = _slang_gen_operation(A, &oper->children[0]); if (lhs) { if (!lhs->Store) { slang_info_log_error(A->log, -- cgit v1.2.3 From 50ecc38545ff59e9f99d092eebf555816ee3f7f0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 15:30:32 -0600 Subject: mesa: glsl: if/while/do condition must be boolean --- src/mesa/shader/slang/slang_codegen.c | 26 ++++++++++++++++++++++++-- 1 file changed, 24 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 86db53c6ac..8414f2ebde 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2049,6 +2049,23 @@ _slang_is_scalar_or_boolean(slang_assemble_ctx *A, slang_operation *oper) } +/** + * Test if an operation is boolean. + */ +static GLboolean +_slang_is_boolean(slang_assemble_ctx *A, slang_operation *oper) +{ + slang_typeinfo type; + GLboolean isBool; + + slang_typeinfo_construct(&type); + _slang_typeof_operation(A, oper, &type); + isBool = (type.spec.type == SLANG_SPEC_BOOL); + slang_typeinfo_destruct(&type); + return isBool; +} + + /** * Generate loop code using high-level IR_LOOP instruction */ @@ -2064,7 +2081,7 @@ _slang_gen_while(slang_assemble_ctx * A, const slang_operation *oper) GLboolean isConst, constTrue; /* type-check expression */ - if (!_slang_is_scalar_or_boolean(A, &oper->children[0])) { + if (!_slang_is_boolean(A, &oper->children[0])) { slang_info_log_error(A->log, "scalar/boolean expression expected for 'while'"); return NULL; } @@ -2127,7 +2144,7 @@ _slang_gen_do(slang_assemble_ctx * A, const slang_operation *oper) GLboolean isConst, constTrue; /* type-check expression */ - if (!_slang_is_scalar_or_boolean(A, &oper->children[1])) { + if (!_slang_is_boolean(A, &oper->children[1])) { slang_info_log_error(A->log, "scalar/boolean expression expected for 'do/while'"); return NULL; } @@ -2256,6 +2273,11 @@ _slang_gen_if(slang_assemble_ctx * A, const slang_operation *oper) GLboolean isConst, constTrue; /* type-check expression */ + if (!_slang_is_boolean(A, &oper->children[0])) { + slang_info_log_error(A->log, "boolean expression expected for 'while'"); + return NULL; + } + if (!_slang_is_scalar_or_boolean(A, &oper->children[0])) { slang_info_log_error(A->log, "scalar/boolean expression expected for 'if'"); return NULL; -- cgit v1.2.3 From 3b5f9588e79ea023ea2ecea755918ab35799a82a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 15:32:07 -0600 Subject: mesa: glsl: added vec4(ivec4) constructor --- src/mesa/shader/slang/library/slang_core.gc | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 0dfc03883f..218383e001 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -257,6 +257,11 @@ vec4 __constructor(const bvec4 b) __retVal = b; } +vec4 __constructor(const ivec4 i) +{ + __retVal = i; +} + vec4 __constructor(const vec3 v3, const float f) { // XXX this constructor shouldn't be needed anymore -- cgit v1.2.3 From 9564e58df052462ab9d3c62576bcf9e38be65c1e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 15:32:21 -0600 Subject: mesa: glsl: regenerated files --- src/mesa/shader/slang/library/slang_core_gc.h | 81 +++++++++++----------- .../shader/slang/library/slang_pp_version_syn.h | 10 +-- 2 files changed, 46 insertions(+), 45 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index 15a1d1776e..33a996f2d9 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -35,47 +35,48 @@ 0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,59,120,120,120,120,0,20,0,0,1,0,12,1,1,1,0,5,105,0, 0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,12,1,1,1,0,1,98, 0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,12,1,1,1,0,4,98, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,12,1,1,1,0,11,118,51,0,0,1,1,0,9,102, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,118,51,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,119,0,18,102,0,20,0,0,1,0,12,1,1,1,0,10,118,50,0,0,1,1,0,9,102,49,0,0,1,1,0,9, -102,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,118,50,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,122,0,18,102,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,50,0, -20,0,0,1,0,6,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18, -105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,105,0,59,120,120,0,20,0,0,1,0,6,1,1,1,0,9,102,0, -0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -0,0,18,102,0,59,120,120,0,0,0,0,1,0,6,1,1,1,0,1,98,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105, -110,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,7,1,1,1,0, -5,105,0,0,1,1,0,5,106,0,0,1,1,0,5,107,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,122,0,18,107,0,20,0,0,1,0,7,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,18,105,0,59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,18,102,0,59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,8,1,1,1,0,5,120,0,0,1,1,0, -5,121,0,0,1,1,0,5,122,0,0,1,1,0,5,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,8,1,1,1,0,5, -105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,8,1,1,1,0, -9,102,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0, -18,102,0,59,120,120,120,120,0,0,0,0,1,0,8,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -18,98,0,59,120,120,120,120,0,20,0,0,1,0,2,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50, -0,20,0,0,1,0,2,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,59,120, -120,0,20,0,0,1,0,2,1,1,1,0,9,102,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0, -48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,0,0,18,102,0,59,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,3,2, -1,6,1,122,101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,59,120,120,0,0,18,122,101, -114,111,0,0,0,0,1,0,2,1,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,3,2,1,6,1, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,12,1,1,1,0,8,105,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,18,105,0,20,0,0,1,0,12,1,1,1,0,11,118,51,0,0,1,1,0,9,102,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,18,118,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, +119,0,18,102,0,20,0,0,1,0,12,1,1,1,0,10,118,50,0,0,1,1,0,9,102,49,0,0,1,1,0,9,102,50,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,18,118,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, +122,0,18,102,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,50,0,20,0,0,1,0,6,1,1,1, +0,5,105,0,0,1,1,0,5,106,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,18,105,0,59,120,120,0,20,0,0,1,0,6,1,1,1,0,9,102,0,0,0,1,4,102,108, +111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,59, +120,120,0,0,0,0,1,0,6,1,1,1,0,1,98,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,7,1,1,1,0,5,105,0,0,1,1,0, +5,106,0,0,1,1,0,5,107,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,107, +0,20,0,0,1,0,7,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,105,0, +59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,18,102,0,59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,8,1,1,1,0,5,120,0,0,1,1,0,5,121,0,0,1,1,0,5, +122,0,0,1,1,0,5,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122, +0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,8,1,1,1,0,5,105,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,8,1,1,1,0,9,102,0,0,0,1, +4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,59, +120,120,120,120,0,0,0,0,1,0,8,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,59, +120,120,120,120,0,20,0,0,1,0,2,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,0,1, +0,2,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,59,120,120,0,20,0, +0,1,0,2,1,1,1,0,9,102,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0,48,0,0,0, +17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,102,0,59,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,3,2,1,6,1, 122,101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1, -0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,3,2,1, -11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0, -0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,59, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,59,120,120,0,0,18,122,101,114, +111,0,0,0,0,1,0,2,1,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,3,2,1,6,1,122, +101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3, +1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,3,2,1,11, +1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0, +4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,59, 120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,5,105,0,0,0,1,3,2,1,7,1,122,101,114,111,0, 2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101, 0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,59,120,120,120,0,0,18,122,101,114, diff --git a/src/mesa/shader/slang/library/slang_pp_version_syn.h b/src/mesa/shader/slang/library/slang_pp_version_syn.h index a75cf7509f..54aee3ff28 100644 --- a/src/mesa/shader/slang/library/slang_pp_version_syn.h +++ b/src/mesa/shader/slang/library/slang_pp_version_syn.h @@ -3,22 +3,22 @@ ".syntax version_directive;\n" "version_directive\n" -" version_directive_1 .and .loop version_directive_2;\n" +" version_directive_1;\n" "version_directive_1\n" " prior_optional_spaces .and optional_version_directive .and .true .emit $;\n" -"version_directive_2\n" -" prior_optional_spaces .and version_directive_body .and .true .emit $;\n" "optional_version_directive\n" " version_directive_body .or .true .emit 10 .emit 1;\n" "version_directive_body\n" " '#' .and optional_space .and \"version\" .and space .and version_number .and optional_space .and\n" " new_line;\n" "version_number\n" -" version_number_110 .or version_number_120;\n" +" version_number_100 .or version_number_110 .or version_number_120;\n" +"version_number_100\n" +" leading_zeroes .and \"100\" .emit 0 .emit 1;\n" "version_number_110\n" " leading_zeroes .and \"110\" .emit 10 .emit 1;\n" "version_number_120\n" -" leading_zeroes .and \"120\" .emit 20 .emit 1;\n" +" leading_zeroes .and \"120\" .emit 20 .emit 1;\n" "leading_zeroes\n" " .loop zero;\n" "zero\n" -- cgit v1.2.3 From 793bab2a368be6bbe93ab7f04b887f95c839c36b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 16:26:22 -0600 Subject: mesa: glsl: varying vars can't be user-defined structs --- src/mesa/shader/slang/slang_codegen.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 8414f2ebde..b69943022a 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -3594,7 +3594,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, else { slang_info_log_error(A->log, "invalid datatype for uniform variable %s", - (char *) var->a_name); + varName); } return GL_FALSE; } @@ -3616,6 +3616,13 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, if (dbg) printf("UNIFORM (sz %d) ", totalSize); } else if (var->type.qualifier == SLANG_QUAL_VARYING) { + if (var->type.specifier.type == SLANG_SPEC_STRUCT) { + slang_info_log_error(A->log, + "varying '%s' cannot be a structure type", + varName); + return GL_FALSE; + } + if (prog) { /* user-defined varying */ GLint varyingLoc = _mesa_add_varying(prog->Varying, varName, size); -- cgit v1.2.3 From 2d8551b30f38d111b60511700c690db1e5b2ab16 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 4 Aug 2008 16:27:14 -0600 Subject: mesa: glsl: check struct types in _slang_assignment_compatible() --- src/mesa/shader/slang/slang_codegen.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index b69943022a..1c87f56857 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2851,7 +2851,20 @@ _slang_assignment_compatible(const slang_typeinfo *t0, const slang_typeinfo *t1) { -#if 01 /* not used just yet - causes problems elsewhere */ +#if 0 + GLuint sz0 = _slang_sizeof_type_specifier(&t0->spec); + GLuint sz1 = _slang_sizeof_type_specifier(&t1->spec); + + if (sz0 != sz1) + return GL_FALSE; +#endif + + if (t0->spec.type == SLANG_SPEC_STRUCT && + t1->spec.type == SLANG_SPEC_STRUCT && + t0->spec._struct->a_name != t1->spec._struct->a_name) + return GL_FALSE; + +#if 0 /* not used just yet - causes problems elsewhere */ if (t0->spec.type == SLANG_SPEC_INT && t1->spec.type == SLANG_SPEC_FLOAT) return GL_FALSE; -- cgit v1.2.3 From 2eb4f86785ac7e4002904384b7ad0517e0bd6999 Mon Sep 17 00:00:00 2001 From: Jonathan White Date: Tue, 5 Aug 2008 08:13:50 -0600 Subject: egl: don't include stdint.h on Windows --- src/egl/main/eglconfigutil.h | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/egl/main/eglconfigutil.h b/src/egl/main/eglconfigutil.h index 7fe875a7d0..c477b94737 100644 --- a/src/egl/main/eglconfigutil.h +++ b/src/egl/main/eglconfigutil.h @@ -4,7 +4,9 @@ #include "eglconfig.h" #include "GL/internal/glcore.h" +#if (!defined(WIN32) && !defined(_WIN32_WCE)) #include "stdint.h" +#endif extern void -- cgit v1.2.3 From 749c1b9fbe4bd89e61dfc3657ad4f8adae20ff2b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 11:13:51 -0600 Subject: mesa: glsl: disable broken assignment type checking for now --- src/mesa/shader/slang/slang_codegen.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 1c87f56857..363c0a10c2 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2928,6 +2928,7 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper) slang_ir_node *n, *lhs, *rhs; /* lhs and rhs type checking */ + if (0) { slang_typeinfo t0, t1; -- cgit v1.2.3 From 1308ca6d2168c5c2f81a8e675687e9d9a4db1a28 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 16:18:39 -0600 Subject: mesa: glsl: re-org of intermediate/temp storage Simplify the code for allocating storage for intermediate results. Use fewer temps in some cases. Also, use new asm vec4_move intrinsic instead of regular assigments in various constructors. For example: float f; vec3 v; v.xyz = f; is not legal GLSL, so do this instead: __asm vec4_move v.xyz, f; // note: f will auto-expand into f.xxxx Plus, fix assorted bugs in structure comparison. --- .../shader/slang/library/slang_common_builtin.gc | 2 +- src/mesa/shader/slang/library/slang_core.gc | 275 +++++++++------------ src/mesa/shader/slang/slang_codegen.c | 61 +++-- src/mesa/shader/slang/slang_emit.c | 226 +++++++++-------- src/mesa/shader/slang/slang_ir.c | 9 +- src/mesa/shader/slang/slang_ir.h | 8 +- src/mesa/shader/slang/slang_vartable.c | 37 ++- 7 files changed, 304 insertions(+), 314 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index 3726335471..f41d1fb603 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -1418,7 +1418,7 @@ bvec4 greaterThan(const vec4 u, const vec4 v) bvec2 greaterThan(const ivec2 u, const ivec2 v) { - __asm vec4_sgt __retVal.xy, u, v; + __asm vec4_sgt __retVal.xy, u.xy, v.xy; } bvec3 greaterThan(const ivec3 u, const ivec3 v) diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 218383e001..840a0814c5 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -107,7 +107,7 @@ int __constructor(const float f) { - __asm float_to_int __retVal, f; + __asm vec4_to_ivec4 __retVal, f; } int __constructor(const bool b) @@ -122,14 +122,12 @@ int __constructor(const int i) bool __constructor(const int i) { - const float zero = 0.0; - __asm vec4_sne __retVal, i, zero; + __asm vec4_sne __retVal, i, 0.0; } bool __constructor(const float f) { - const float zero = 0.0; - __asm vec4_sne __retVal, f, zero; + __asm vec4_sne __retVal, f, 0.0; } bool __constructor(const bool b) @@ -139,12 +137,12 @@ bool __constructor(const bool b) float __constructor(const int i) { - __asm int_to_float __retVal, i; + __asm ivec4_to_vec4 __retVal, i; } float __constructor(const bool b) { - __retVal = b; + __asm ivec4_to_vec4 __retVal, b; } float __constructor(const float f) @@ -163,32 +161,33 @@ vec2 __constructor(const float x, const float y) vec2 __constructor(const float f) { - __retVal.xy = f.xx; + __asm vec4_move __retVal.xy, f; } vec2 __constructor(const int i) { - __retVal.xy = i.xx; + __asm ivec4_to_vec4 __retVal.xy, i; } vec2 __constructor(const bool b) { - __retVal.xy = b.xx; + __asm ivec4_to_vec4 __retVal.xy, b; } vec2 __constructor(const bvec2 b) { - __retVal = b; +// __retVal = b; + __asm ivec4_to_vec4 __retVal.xy, b; } vec2 __constructor(const vec3 v) { - __retVal.xy = v.xy; + __asm vec4_move __retVal.xy, v.xy; } vec2 __constructor(const vec4 v) { - __retVal.st = v.xy; + __asm vec4_move __retVal.xy, v.xy; } @@ -203,27 +202,28 @@ vec3 __constructor(const float x, const float y, const float z) vec3 __constructor(const float f) { - __retVal.xyz = f.xxx; + // Note: this could be "__retVal.xyz = f" but that's an illegal assignment + __asm vec4_move __retVal.xyz, f; } vec3 __constructor(const int i) { - __asm int_to_float __retVal.xyz, i.xxx; + __asm ivec4_to_vec4 __retVal.xyz, i; } vec3 __constructor(const bool b) { - __retVal.xyz = b.xxx; + __asm ivec4_to_vec4 __retVal.xyz, b; } vec3 __constructor(const bvec3 b) { - __retVal = b; + __asm ivec4_to_vec4 __retVal.xyz, b; } vec3 __constructor(const vec4 v) { - __retVal.xyz = v.xyz; + __asm vec4_move __retVal.xyz, v; } @@ -239,27 +239,28 @@ vec4 __constructor(const float x, const float y, const float z, const float w) vec4 __constructor(const float f) { - __retVal = f.xxxx; + // Note: this could be "__retVal = f" but that's an illegal assignment + __asm vec4_move __retVal, f; } vec4 __constructor(const int i) { - __retVal = i.xxxx; + __asm ivec4_to_vec4 __retVal, i; } vec4 __constructor(const bool b) { - __retVal = b.xxxx; + __asm ivec4_to_vec4 __retVal, b; } vec4 __constructor(const bvec4 b) { - __retVal = b; + __asm ivec4_to_vec4 __retVal, b; } vec4 __constructor(const ivec4 i) { - __retVal = i; + __asm ivec4_to_vec4 __retVal, i; } vec4 __constructor(const vec3 v3, const float f) @@ -288,17 +289,17 @@ ivec2 __constructor(const int i, const int j) ivec2 __constructor(const int i) { - __retVal.xy = i.xx; + __asm vec4_move __retVal.xy, i; } ivec2 __constructor(const float f) { - __asm float_to_int __retVal.xy, f.xx; + __asm vec4_to_ivec4 __retVal.xy, f; } ivec2 __constructor(const bool b) { - __asm float_to_int __retVal.xy, b.xx; + __asm vec4_to_ivec4 __retVal.xy, b; } @@ -313,17 +314,17 @@ ivec3 __constructor(const int i, const int j, const int k) ivec3 __constructor(const int i) { - __retVal.xyz = i.xxx; + __asm vec4_move __retVal.xyz, i; } ivec3 __constructor(const float f) { - __retVal.xyz = f.xxx; + __asm vec4_to_ivec4 __retVal.xyz, f; } ivec3 __constructor(const bool b) { - __retVal.xyz = b.xxx; + __asm vec4_move __retVal.xyz, b; } @@ -339,17 +340,17 @@ ivec4 __constructor(const int x, const int y, const int z, const int w) ivec4 __constructor(const int i) { - __retVal = i.xxxx; + __asm vec4_move __retVal, i; } ivec4 __constructor(const float f) { - __asm float_to_int __retVal, f.xxxx; + __asm vec4_to_ivec4 __retVal, f; } ivec4 __constructor(const bool b) { - __retVal = b.xxxx; + __asm vec4_to_ivec4 __retVal, b; } @@ -363,19 +364,17 @@ bvec2 __constructor(const bool b1, const bool b2) bvec2 __constructor(const bool b) { - __retVal.xy = b.xx; + __asm vec4_move __retVal.xy, b; } bvec2 __constructor(const float f) { - const vec2 zero = vec2(0.0, 0.0); - __asm vec4_sne __retVal.xy, f.xx, zero; + __asm vec4_sne __retVal.xy, f, 0.0; } bvec2 __constructor(const int i) { - const ivec2 zero = ivec2(0, 0); - __asm vec4_sne __retVal.xy, i.xx, zero; + __asm vec4_sne __retVal.xy, i, 0.0; } bvec2 __constructor(const vec2 v) @@ -385,8 +384,7 @@ bvec2 __constructor(const vec2 v) bvec2 __constructor(const ivec2 v) { - const ivec2 zero = ivec2(0, 0); - __asm vec4_sne __retVal.xy, v, zero; + __asm vec4_sne __retVal.xy, v, 0.0; } @@ -402,31 +400,27 @@ bvec3 __constructor(const bool b1, const bool b2, const bool b3) bvec3 __constructor(const bool b) { - __retVal.xyz = b.xxx; + __asm vec4_move __retVal.xyz, b; } bvec3 __constructor(const float f) { - const vec3 zero = vec3(0.0, 0.0, 0.0); - __asm vec4_sne __retVal.xyz, f.xxx, zero; + __asm vec4_sne __retVal.xyz, f, 0.0; } bvec3 __constructor(const int i) { - const ivec3 zero = ivec3(0, 0, 0); - __asm vec4_sne __retVal.xyz, i.xxx, zero; + __asm vec4_sne __retVal.xyz, i, 0.0; } bvec3 __constructor(const vec3 v) { - const vec3 zero = vec3(0.0, 0.0, 0.0); - __asm vec4_sne __retVal.xyz, v, zero; + __asm vec4_sne __retVal.xyz, v, 0.0; } bvec3 __constructor(const ivec3 v) { - const ivec3 zero = ivec3(0, 0, 0); - __asm vec4_sne __retVal.xyz, v, zero; + __asm vec4_sne __retVal.xyz, v, 0.0; } @@ -452,31 +446,27 @@ bvec4 __constructor(const float f1, const float f2, const float f3, const float bvec4 __constructor(const bool b) { - __retVal.xyzw = b.xxxx; + __asm vec4_move __retVal.xyzw, b; } bvec4 __constructor(const float f) { - const vec4 zero = vec4(0.0, 0.0, 0.0, 0.0); - __asm vec4_sne __retVal, f.xxxx, zero; + __asm vec4_sne __retVal.xyzw, f, 0.0; } bvec4 __constructor(const int i) { - const ivec4 zero = ivec4(0, 0, 0, 0); - __asm vec4_sne __retVal, i.xxxx, zero; + __asm vec4_sne __retVal.xyzw, i, 0.0; } bvec4 __constructor(const vec4 v) { - const vec4 zero = vec4(0.0, 0.0, 0.0, 0.0); - __asm vec4_sne __retVal, v, zero; + __asm vec4_sne __retVal.xyzw, v, 0.0; } bvec4 __constructor(const ivec4 v) { - const ivec4 zero = ivec4(0, 0, 0, 0); - __asm vec4_sne __retVal, v, zero; + __asm vec4_sne __retVal.xyzw, v, 0.0; } @@ -619,30 +609,17 @@ mat4 __constructor(const vec4 c0, const vec4 c1, const vec4 c2, const vec4 c3) int __operator + (const int a, const int b) { -// XXX If we ever have int registers, we'll do something like this: -// XXX For now, mostly treat ints as floats. -// float x, y; -// __asm int_to_float x, a; -// __asm int_to_float y, b; -// __asm vec4_add x.x, x.x, y.x; -// __asm float_to_int __retVal, x; - float x; - __asm vec4_add x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_add __retVal, a, b; } int __operator - (const int a, const int b) { - float x; - __asm vec4_subtract x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_subtract __retVal, a, b; } int __operator * (const int a, const int b) { - float x; - __asm vec4_multiply x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_multiply __retVal, a, b; } int __operator / (const int a, const int b) @@ -650,7 +627,7 @@ int __operator / (const int a, const int b) float bInv, x; __asm float_rcp bInv, b; __asm vec4_multiply x, a, bInv; - __asm float_to_int __retVal, x; + __asm vec4_to_ivec4 __retVal, x; } @@ -658,23 +635,17 @@ int __operator / (const int a, const int b) ivec2 __operator + (const ivec2 a, const ivec2 b) { - vec2 x; - __asm vec4_add x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_add __retVal, a, b; } ivec2 __operator - (const ivec2 a, const ivec2 b) { - vec2 x; - __asm vec4_subtract x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_subtract __retVal, a, b; } ivec2 __operator * (const ivec2 a, const ivec2 b) { - vec2 x; - __asm vec4_multiply x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_multiply __retVal, a, b; } ivec2 __operator / (const ivec2 a, const ivec2 b) @@ -683,7 +654,7 @@ ivec2 __operator / (const ivec2 a, const ivec2 b) __asm float_rcp bInv.x, b.x; __asm float_rcp bInv.y, b.y; __asm vec4_multiply x, a, bInv; - __asm float_to_int __retVal, x; + __asm vec4_to_ivec4 __retVal, x; } @@ -691,23 +662,17 @@ ivec2 __operator / (const ivec2 a, const ivec2 b) ivec3 __operator + (const ivec3 a, const ivec3 b) { - vec3 x; - __asm vec4_add x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_add __retVal, a, b; } ivec3 __operator - (const ivec3 a, const ivec3 b) { - vec3 x; - __asm vec4_subtract x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_subtract __retVal, a, b; } ivec3 __operator * (const ivec3 a, const ivec3 b) { - vec3 x; - __asm vec4_multiply x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_multiply __retVal, a, b; } ivec3 __operator / (const ivec3 a, const ivec3 b) @@ -717,7 +682,7 @@ ivec3 __operator / (const ivec3 a, const ivec3 b) __asm float_rcp bInv.y, b.y; __asm float_rcp bInv.z, b.z; __asm vec4_multiply x, a, bInv; - __asm float_to_int __retVal, x; + __asm vec4_to_ivec4 __retVal, x; } @@ -725,23 +690,17 @@ ivec3 __operator / (const ivec3 a, const ivec3 b) ivec4 __operator + (const ivec4 a, const ivec4 b) { - vec3 x; - __asm vec4_add x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_add __retVal, a, b; } ivec4 __operator - (const ivec4 a, const ivec4 b) { - vec4 x; - __asm vec4_subtract x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_subtract __retVal, a, b; } ivec4 __operator * (const ivec4 a, const ivec4 b) { - vec4 x; - __asm vec4_multiply x, a, b; - __asm float_to_int __retVal, x; + __asm vec4_multiply __retVal, a, b; } ivec4 __operator / (const ivec4 a, const ivec4 b) @@ -752,7 +711,7 @@ ivec4 __operator / (const ivec4 a, const ivec4 b) __asm float_rcp bInv.z, b.z; __asm float_rcp bInv.w, b.w; __asm vec4_multiply x, a, bInv; - __asm float_to_int __retVal, x; + __asm vec4_to_ivec4 __retVal, x; } @@ -760,24 +719,24 @@ ivec4 __operator / (const ivec4 a, const ivec4 b) float __operator + (const float a, const float b) { - __asm vec4_add __retVal.x, a, b; + __asm vec4_add __retVal, a, b; } float __operator - (const float a, const float b) { - __asm vec4_subtract __retVal.x, a, b; + __asm vec4_subtract __retVal, a, b; } float __operator * (const float a, const float b) { - __asm vec4_multiply __retVal.x, a, b; + __asm vec4_multiply __retVal, a, b; } float __operator / (const float a, const float b) { float bInv; - __asm float_rcp bInv.x, b.x; - __asm vec4_multiply __retVal.x, a, bInv; + __asm float_rcp bInv.x, b; + __asm vec4_multiply __retVal, a, bInv; } @@ -868,32 +827,32 @@ vec4 __operator / (const vec4 v, const vec4 u) vec2 __operator + (const float a, const vec2 u) { - __asm vec4_add __retVal.xy, a.xx, u.xy; + __asm vec4_add __retVal.xy, a, u.xy; } vec2 __operator + (const vec2 v, const float b) { - __asm vec4_add __retVal.xy, v.xy, b.xx; + __asm vec4_add __retVal.xy, v.xy, b; } vec2 __operator - (const float a, const vec2 u) { - __asm vec4_subtract __retVal.xy, a.xx, u.xy; + __asm vec4_subtract __retVal.xy, a, u.xy; } vec2 __operator - (const vec2 v, const float b) { - __asm vec4_subtract __retVal.xy, v.xy, b.xx; + __asm vec4_subtract __retVal.xy, v.xy, b; } vec2 __operator * (const float a, const vec2 u) { - __asm vec4_multiply __retVal.xy, a.xx, u.xy; + __asm vec4_multiply __retVal.xy, a, u.xy; } vec2 __operator * (const vec2 v, const float b) { - __asm vec4_multiply __retVal.xy, v.xy, b.xx; + __asm vec4_multiply __retVal.xy, v.xy, b; } vec2 __operator / (const float a, const vec2 u) @@ -901,14 +860,14 @@ vec2 __operator / (const float a, const vec2 u) vec2 invU; __asm float_rcp invU.x, u.x; __asm float_rcp invU.y, u.y; - __asm vec4_multiply __retVal.xy, a.xx, invU.xy; + __asm vec4_multiply __retVal.xy, a, invU.xy; } vec2 __operator / (const vec2 v, const float b) { float invB; __asm float_rcp invB, b; - __asm vec4_multiply __retVal.xy, v.xy, invB.xx; + __asm vec4_multiply __retVal.xy, v.xy, invB; } @@ -916,32 +875,32 @@ vec2 __operator / (const vec2 v, const float b) vec3 __operator + (const float a, const vec3 u) { - __asm vec4_add __retVal.xyz, a.xxx, u.xyz; + __asm vec4_add __retVal.xyz, a, u.xyz; } vec3 __operator + (const vec3 v, const float b) { - __asm vec4_add __retVal.xyz, v.xyz, b.xxx; + __asm vec4_add __retVal.xyz, v.xyz, b; } vec3 __operator - (const float a, const vec3 u) { - __asm vec4_subtract __retVal.xyz, a.xxx, u.xyz; + __asm vec4_subtract __retVal.xyz, a, u.xyz; } vec3 __operator - (const vec3 v, const float b) { - __asm vec4_subtract __retVal.xyz, v.xyz, b.xxx; + __asm vec4_subtract __retVal.xyz, v.xyz, b; } vec3 __operator * (const float a, const vec3 u) { - __asm vec4_multiply __retVal.xyz, a.xxx, u.xyz; + __asm vec4_multiply __retVal.xyz, a, u.xyz; } vec3 __operator * (const vec3 v, const float b) { - __asm vec4_multiply __retVal.xyz, v.xyz, b.xxx; + __asm vec4_multiply __retVal.xyz, v.xyz, b; } vec3 __operator / (const float a, const vec3 u) @@ -950,14 +909,14 @@ vec3 __operator / (const float a, const vec3 u) __asm float_rcp invU.x, u.x; __asm float_rcp invU.y, u.y; __asm float_rcp invU.z, u.z; - __asm vec4_multiply __retVal.xyz, a.xxx, invU.xyz; + __asm vec4_multiply __retVal.xyz, a, invU.xyz; } vec3 __operator / (const vec3 v, const float b) { float invB; __asm float_rcp invB, b; - __asm vec4_multiply __retVal.xyz, v.xyz, invB.xxx; + __asm vec4_multiply __retVal.xyz, v.xyz, invB; } @@ -965,32 +924,32 @@ vec3 __operator / (const vec3 v, const float b) vec4 __operator + (const float a, const vec4 u) { - __asm vec4_add __retVal, a.xxxx, u; + __asm vec4_add __retVal, a, u; } vec4 __operator + (const vec4 v, const float b) { - __asm vec4_add __retVal, v, b.xxxx; + __asm vec4_add __retVal, v, b; } vec4 __operator - (const float a, const vec4 u) { - __asm vec4_subtract __retVal, a.xxxx, u; + __asm vec4_subtract __retVal, a, u; } vec4 __operator - (const vec4 v, const float b) { - __asm vec4_subtract __retVal, v, b.xxxx; + __asm vec4_subtract __retVal, v, b; } vec4 __operator * (const float a, const vec4 u) { - __asm vec4_multiply __retVal, a.xxxx, u; + __asm vec4_multiply __retVal, a, u; } vec4 __operator * (const vec4 v, const float b) { - __asm vec4_multiply __retVal, v, b.xxxx; + __asm vec4_multiply __retVal, v, b; } vec4 __operator / (const float a, const vec4 u) @@ -1000,14 +959,14 @@ vec4 __operator / (const float a, const vec4 u) __asm float_rcp invU.y, u.y; __asm float_rcp invU.z, u.z; __asm float_rcp invU.w, u.w; - __asm vec4_multiply __retVal, a.xxxx, invU; + __asm vec4_multiply __retVal, a, invU; } vec4 __operator / (const vec4 v, const float b) { float invB; __asm float_rcp invB, b; - __asm vec4_multiply __retVal, v, invB.xxxx; + __asm vec4_multiply __retVal, v, invB; } @@ -1254,7 +1213,7 @@ void __operator /= (inout int a, const int b) float invB; __asm float_rcp invB, b; __asm vec4_multiply a, a, invB; - __asm float_to_int a, a; + __asm vec4_to_ivec4 a, a; } @@ -1281,7 +1240,7 @@ void __operator /= (inout ivec2 v, const ivec2 u) __asm float_rcp inv.x, u.x; __asm float_rcp inv.y, u.y; __asm vec4_multiply z, v, inv; - __asm float_to_int v, z; + __asm vec4_to_ivec4 v, z; } @@ -1308,7 +1267,7 @@ void __operator /= (inout ivec3 v, const ivec3 u) __asm float_rcp inv.x, u.x; __asm float_rcp inv.y, u.y; __asm vec4_multiply z, v, inv; - __asm float_to_int v, z; + __asm vec4_to_ivec4 v, z; } @@ -1335,7 +1294,7 @@ void __operator /= (inout ivec4 v, const ivec4 u) __asm float_rcp inv.x, u.x; __asm float_rcp inv.y, u.y; __asm vec4_multiply z, v, inv; - __asm float_to_int v, z; + __asm vec4_to_ivec4 v, z; } @@ -1450,17 +1409,17 @@ void __operator /= (inout vec4 v, const vec4 u) void __operator += (inout ivec2 v, const int a) { - __asm vec4_add v.xy, v.xy, a.xx; + __asm vec4_add v.xy, v.xy, a; } void __operator -= (inout ivec2 v, const int a) { - __asm vec4_subtract v.xy, v.xy, a.xx; + __asm vec4_subtract v.xy, v.xy, a; } void __operator *= (inout ivec2 v, const int a) { - __asm vec4_multiply v.xy, v.xy, a.xx; + __asm vec4_multiply v.xy, v.xy, a; v.x *= a; v.y *= a; } @@ -1477,17 +1436,17 @@ void __operator /= (inout ivec2 v, const int a) void __operator += (inout ivec3 v, const int a) { - __asm vec4_add v.xyz, v.xyz, a.xxx; + __asm vec4_add v.xyz, v.xyz, a; } void __operator -= (inout ivec3 v, const int a) { - __asm vec4_subtract v.xyz, v.xyz, a.xxx; + __asm vec4_subtract v.xyz, v.xyz, a; } void __operator *= (inout ivec3 v, const int a) { - __asm vec4_multiply v.xyz, v.xyz, a.xxx; + __asm vec4_multiply v.xyz, v.xyz, a; } void __operator /= (inout ivec3 v, const int a) @@ -1503,17 +1462,17 @@ void __operator /= (inout ivec3 v, const int a) void __operator += (inout ivec4 v, const int a) { - __asm vec4_add v, v, a.xxxx; + __asm vec4_add v, v, a; } void __operator -= (inout ivec4 v, const int a) { - __asm vec4_subtract v, v, a.xxxx; + __asm vec4_subtract v, v, a; } void __operator *= (inout ivec4 v, const int a) { - __asm vec4_multiply v, v, a.xxxx; + __asm vec4_multiply v, v, a; } void __operator /= (inout ivec4 v, const int a) @@ -1530,24 +1489,24 @@ void __operator /= (inout ivec4 v, const int a) void __operator += (inout vec2 v, const float a) { - __asm vec4_add v.xy, v, a.xx; + __asm vec4_add v.xy, v, a; } void __operator -= (inout vec2 v, const float a) { - __asm vec4_subtract v.xy, v, a.xx; + __asm vec4_subtract v.xy, v, a; } void __operator *= (inout vec2 v, const float a) { - __asm vec4_multiply v.xy, v, a.xx; + __asm vec4_multiply v.xy, v, a; } void __operator /= (inout vec2 v, const float a) { float invA; __asm float_rcp invA, a; - __asm vec4_multiply v.xy, v.xy, invA.xx; + __asm vec4_multiply v.xy, v.xy, invA; } @@ -1555,24 +1514,24 @@ void __operator /= (inout vec2 v, const float a) void __operator += (inout vec3 v, const float a) { - __asm vec4_add v.xyz, v, a.xxx; + __asm vec4_add v.xyz, v, a; } void __operator -= (inout vec3 v, const float a) { - __asm vec4_subtract v.xyz, v, a.xxx; + __asm vec4_subtract v.xyz, v, a; } void __operator *= (inout vec3 v, const float a) { - __asm vec4_multiply v.xyz, v, a.xxx; + __asm vec4_multiply v.xyz, v, a; } void __operator /= (inout vec3 v, const float a) { float invA; __asm float_rcp invA, a; - __asm vec4_multiply v.xyz, v.xyz, invA.xxx; + __asm vec4_multiply v.xyz, v.xyz, invA; } @@ -1580,24 +1539,24 @@ void __operator /= (inout vec3 v, const float a) void __operator += (inout vec4 v, const float a) { - __asm vec4_add v, v, a.xxxx; + __asm vec4_add v, v, a; } void __operator -= (inout vec4 v, const float a) { - __asm vec4_subtract v, v, a.xxxx; + __asm vec4_subtract v, v, a; } void __operator *= (inout vec4 v, const float a) { - __asm vec4_multiply v, v, a.xxxx; + __asm vec4_multiply v, v, a; } void __operator /= (inout vec4 v, const float a) { float invA; __asm float_rcp invA, a; - __asm vec4_multiply v, v, invA.xxxx; + __asm vec4_multiply v, v, invA; } diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 363c0a10c2..582b47c239 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -425,6 +425,7 @@ static slang_asm_info AsmInfo[] = { { "vec4_sle", IR_SLE, 1, 2 }, { "vec4_slt", IR_SLT, 1, 2 }, /* vec4 unary */ + { "vec4_move", IR_MOVE, 1, 1 }, { "vec4_floor", IR_FLOOR, 1, 1 }, { "vec4_frac", IR_FRAC, 1, 1 }, { "vec4_abs", IR_ABS, 1, 1 }, @@ -448,8 +449,8 @@ static slang_asm_info AsmInfo[] = { { "vec4_texp_rect", IR_TEX, 1, 2 },/* rectangle w/ projection */ /* unary op */ - { "int_to_float", IR_I_TO_F, 1, 1 }, - { "float_to_int", IR_F_TO_I, 1, 1 }, + { "ivec4_to_vec4", IR_I_TO_F, 1, 1 }, /* int[4] to float[4] */ + { "vec4_to_ivec4", IR_F_TO_I, 1, 1 }, /* float[4] to int[4] */ { "float_exp", IR_EXP, 1, 1 }, { "float_exp2", IR_EXP2, 1, 1 }, { "float_log2", IR_LOG2, 1, 1 }, @@ -1674,11 +1675,9 @@ _slang_gen_asm(slang_assemble_ctx *A, slang_operation *oper, return NULL; assert(!n->Store); - n->Store = get_store(n0); - n->Writemask = writemask; + n->Store = n0->Store; - assert(n->Store->File != PROGRAM_UNDEFINED || - n->Store->Parent); + assert(n->Store->File != PROGRAM_UNDEFINED || n->Store->Parent); _slang_free(n0); } @@ -1957,6 +1956,7 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name, slang_atom atom; slang_function *fun; GLboolean error; + slang_ir_node *n; atom = slang_atom_pool_atom(A->atoms, name); if (atom == SLANG_ATOM_NULL) @@ -2008,7 +2008,17 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name, assert(fun); } - return _slang_gen_function_call(A, fun, oper, dest); + n = _slang_gen_function_call(A, fun, oper, dest); + + if (n && !n->Store && !dest + && fun->header.type.specifier.type != SLANG_SPEC_VOID) { + /* setup n->Store for the result of the function call */ + GLint size = _slang_sizeof_type_specifier(&fun->header.type.specifier); + n->Store = _slang_new_ir_storage(PROGRAM_TEMPORARY, -1, size); + /*printf("Alloc storage for function result, size %d \n", size);*/ + } + + return n; } @@ -2524,13 +2534,13 @@ _slang_gen_select(slang_assemble_ctx *A, slang_operation *oper) tmpVar = new_node0(IR_VAR); tmpVar->Store = tmpDecl->Store; trueExpr = _slang_gen_operation(A, &oper->children[1]); - trueNode = new_node2(IR_MOVE, tmpVar, trueExpr); + trueNode = new_node2(IR_COPY, tmpVar, trueExpr); /* if-false body (child 2) */ tmpVar = new_node0(IR_VAR); tmpVar->Store = tmpDecl->Store; falseExpr = _slang_gen_operation(A, &oper->children[2]); - falseNode = new_node2(IR_MOVE, tmpVar, falseExpr); + falseNode = new_node2(IR_COPY, tmpVar, falseExpr); ifNode = new_if(cond, trueNode, falseNode); @@ -2711,7 +2721,7 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) assert(oper->num_children == 0 || oper->num_children == 1); v = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE); - /*printf("Declare %s at %p\n", varName, v);*/ + /*printf("Declare %s at %p\n", varName, (void *) v);*/ assert(v); #if 0 @@ -2736,7 +2746,8 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) rhs = _slang_gen_operation(A, &oper->children[0]); if (!rhs) return NULL; /* must have found an error */ - init = new_node2(IR_MOVE, var, rhs); + init = new_node2(IR_COPY, var, rhs); + /*assert(rhs->Opcode != IR_SEQ);*/ n = new_seq(varDecl, init); } @@ -2775,12 +2786,14 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) if (!rhs) return NULL; + /*assert(rhs->Store);*/ + if (rhs->Store && var->Store->Size != rhs->Store->Size) { slang_info_log_error(A->log, "invalid assignment (wrong types)"); return NULL; } - init = new_node2(IR_MOVE, var, rhs); + init = new_node2(IR_COPY, var, rhs); n = new_seq(varDecl, init); } else { @@ -2851,12 +2864,14 @@ _slang_assignment_compatible(const slang_typeinfo *t0, const slang_typeinfo *t1) { -#if 0 +#if 1 GLuint sz0 = _slang_sizeof_type_specifier(&t0->spec); GLuint sz1 = _slang_sizeof_type_specifier(&t1->spec); - if (sz0 != sz1) + if (sz0 != sz1) { + printf("size mismatch %u vs %u\n", sz0, sz1); return GL_FALSE; + } #endif if (t0->spec.type == SLANG_SPEC_STRUCT && @@ -2974,7 +2989,7 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper) */ rhs = _slang_gen_swizzle(rhs, newSwizzle); } - n = new_node2(IR_MOVE, lhs, rhs); + n = new_node2(IR_COPY, lhs, rhs); n->Writemask = writemask; return n; } @@ -3208,6 +3223,7 @@ _slang_gen_compare(slang_assemble_ctx *A, slang_operation *oper, slang_ir_opcode opcode) { slang_typeinfo t0, t1; + slang_ir_node *n; slang_typeinfo_construct(&t0); _slang_typeof_operation(A, &oper->children[0], &t0); @@ -3221,9 +3237,14 @@ _slang_gen_compare(slang_assemble_ctx *A, slang_operation *oper, return NULL; } - return new_node2(opcode, - _slang_gen_operation(A, &oper->children[0]), - _slang_gen_operation(A, &oper->children[1])); + n = new_node2(opcode, + _slang_gen_operation(A, &oper->children[0]), + _slang_gen_operation(A, &oper->children[1])); + + /* result is a bool (size 1) */ + n->Store = _slang_new_ir_storage(PROGRAM_TEMPORARY, -1, 1); + + return n; } @@ -3726,7 +3747,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, if (var->initializer) { slang_ir_node *lhs, *rhs, *init; - /* Generate IR_MOVE instruction to initialize the variable */ + /* Generate IR_COPY instruction to initialize the variable */ lhs = new_node0(IR_VAR); lhs->Var = var; lhs->Store = n->Store; @@ -3736,7 +3757,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, rhs = _slang_gen_operation(A, var->initializer); assert(rhs); - init = new_node2(IR_MOVE, lhs, rhs); + init = new_node2(IR_COPY, lhs, rhs); n = new_seq(n, init); } diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index b1ebad7f17..b902da8457 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -127,22 +127,33 @@ _slang_swizzle_swizzle(GLuint swz1, GLuint swz2) /** - * Allocate temporary storage for an intermediate result (such as for - * a multiply or add, etc. + * Allocate storage for the given node (if it hasn't already been allocated). + * + * Typically this is temporary storage for an intermediate result (such as + * for a multiply or add, etc). + * + * If n->Store does not exist it will be created and will be of the size + * specified by defaultSize. */ static GLboolean -alloc_temp_storage(slang_emit_info *emitInfo, slang_ir_node *n, GLint size) +alloc_node_storage(slang_emit_info *emitInfo, slang_ir_node *n, + GLint defaultSize) { assert(!n->Var); - assert(!n->Store); - assert(size > 0); - n->Store = _slang_new_ir_storage(PROGRAM_TEMPORARY, -1, size); - if (!_slang_alloc_temp(emitInfo->vt, n->Store)) { - slang_info_log_error(emitInfo->log, - "Ran out of registers, too many temporaries"); - _slang_free(n->Store); - n->Store = NULL; - return GL_FALSE; + if (!n->Store) { + assert(defaultSize > 0); + n->Store = _slang_new_ir_storage(PROGRAM_TEMPORARY, -1, defaultSize); + } + + /* now allocate actual register(s). I.e. set n->Store->Index >= 0 */ + if (n->Store->Index < 0) { + if (!_slang_alloc_temp(emitInfo->vt, n->Store)) { + slang_info_log_error(emitInfo->log, + "Ran out of registers, too many temporaries"); + _slang_free(n->Store); + n->Store = NULL; + return GL_FALSE; + } } return GL_TRUE; } @@ -153,7 +164,7 @@ alloc_temp_storage(slang_emit_info *emitInfo, slang_ir_node *n, GLint size) * Otherwise, no-op. */ static void -free_temp_storage(slang_var_table *vt, slang_ir_node *n) +free_node_storage(slang_var_table *vt, slang_ir_node *n) { if (n->Store->File == PROGRAM_TEMPORARY && n->Store->Index >= 0 && @@ -167,6 +178,22 @@ free_temp_storage(slang_var_table *vt, slang_ir_node *n) } +/** + * Helper function to allocate a short-term temporary. + * Free it with _slang_free_temp(). + */ +static GLboolean +alloc_local_temp(slang_emit_info *emitInfo, slang_ir_storage *temp, GLint size) +{ + assert(size >= 1); + assert(size <= 4); + _mesa_bzero(temp, sizeof(*temp)); + temp->Size = size; + temp->File = PROGRAM_TEMPORARY; + temp->Index = -1; + return _slang_alloc_temp(emitInfo->vt, temp); +} + /** * Remove any SWIZZLE_NIL terms from given swizzle mask. @@ -584,19 +611,15 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) } /* result storage */ - if (!n->Store) { - GLint size = info->ResultSize; - if (!alloc_temp_storage(emitInfo, n, size)) - return NULL; -#if 0000 /* this should work, but doesn't yet */ - if (size == 2) - n->Writemask = WRITEMASK_XY; - else if (size == 3) - n->Writemask = WRITEMASK_XYZ; - else if (size == 1) - n->Writemask = WRITEMASK_X << GET_SWZ(n->Store->Swizzle,0); -#endif - } + alloc_node_storage(emitInfo, n, -1); + assert(n->Store->Index >= 0); + if (n->Store->Size == 2) + n->Writemask = WRITEMASK_XY; + else if (n->Store->Size == 3) + n->Writemask = WRITEMASK_XYZ; + else if (n->Store->Size == 1) + n->Writemask = WRITEMASK_X << GET_SWZ(n->Store->Swizzle, 0); + storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); @@ -608,7 +631,7 @@ emit_arith(slang_emit_info *emitInfo, slang_ir_node *n) /* really free temps now */ for (i = 0; i < 3; i++) if (temps[i]) - free_temp_storage(emitInfo->vt, temps[i]); + free_node_storage(emitInfo->vt, temps[i]); /*_mesa_print_instruction(inst);*/ return inst; @@ -636,16 +659,15 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) return NULL; } + /* final result is 1 bool */ + if (!alloc_node_storage(emitInfo, n, 1)) + return NULL; + size = n->Children[0]->Store->Size; if (size == 1) { gl_inst_opcode opcode; - if (!n->Store) { - if (!alloc_temp_storage(emitInfo, n, 1)) /* 1 bool */ - return NULL; - } - opcode = n->Opcode == IR_EQUAL ? OPCODE_SEQ : OPCODE_SNE; inst = new_instruction(emitInfo, opcode); storage_to_src_reg(&inst->SrcReg[0], n->Children[0]->Store); @@ -655,11 +677,11 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) else if (size <= 4) { GLuint swizzle; gl_inst_opcode dotOp; - - assert(!n->Store); - if (!n->Store) { - if (!alloc_temp_storage(emitInfo, n, size)) /* 'size' bools */ - return NULL; + slang_ir_storage tempStore; + + if (!alloc_local_temp(emitInfo, &tempStore, 4)) { + return NULL; + /* out of temps */ } if (size == 4) { @@ -676,26 +698,25 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Y, SWIZZLE_Y); } - /* Compute equality, inequality (tmp1 = (A ?= B)) */ + /* Compute inequality (temp = (A != B)) */ inst = new_instruction(emitInfo, OPCODE_SNE); + storage_to_dst_reg(&inst->DstReg, &tempStore, n->Writemask); storage_to_src_reg(&inst->SrcReg[0], n->Children[0]->Store); storage_to_src_reg(&inst->SrcReg[1], n->Children[1]->Store); - storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); inst->Comment = _mesa_strdup("Compare values"); - /* Compute tmp2 = DOT(tmp1, tmp1) (reduction) */ + /* Compute val = DOT(temp, temp) (reduction) */ inst = new_instruction(emitInfo, dotOp); - storage_to_src_reg(&inst->SrcReg[0], n->Store); - storage_to_src_reg(&inst->SrcReg[1], n->Store); - inst->SrcReg[0].Swizzle = inst->SrcReg[1].Swizzle = swizzle; /*override*/ - free_temp_storage(emitInfo->vt, n); /* free tmp1 */ - if (!alloc_temp_storage(emitInfo, n, 1)) /* alloc tmp2 */ - return NULL; storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); + storage_to_src_reg(&inst->SrcReg[0], &tempStore); + storage_to_src_reg(&inst->SrcReg[1], &tempStore); + inst->SrcReg[0].Swizzle = inst->SrcReg[1].Swizzle = swizzle; /*override*/ inst->Comment = _mesa_strdup("Reduce vec to bool"); + _slang_free_temp(emitInfo->vt, &tempStore); /* free temp */ + if (n->Opcode == IR_EQUAL) { - /* compute tmp2.x = !tmp2.x via tmp2.x = (tmp2.x == 0) */ + /* compute val = !val.x with SEQ val, val, 0; */ inst = new_instruction(emitInfo, OPCODE_SEQ); storage_to_src_reg(&inst->SrcReg[0], n->Store); constant_to_src_reg(&inst->SrcReg[1], 0.0, emitInfo); @@ -708,73 +729,59 @@ emit_compare(slang_emit_info *emitInfo, slang_ir_node *n) * XXX this won't work reliably for structs with padding!! */ GLint i, num = (n->Children[0]->Store->Size + 3) / 4; - slang_ir_storage accTemp; + slang_ir_storage accTemp, sneTemp; - if (!n->Store) { - if (!alloc_temp_storage(emitInfo, n, 4)) - return NULL; - } + if (!alloc_local_temp(emitInfo, &accTemp, 4)) + return NULL; - accTemp.Size = 4; - accTemp.File = PROGRAM_TEMPORARY; - if (!_slang_alloc_temp(emitInfo->vt, &accTemp)) { + if (!alloc_local_temp(emitInfo, &sneTemp, 4)) return NULL; - /* out of temps */ - } for (i = 0; i < num; i++) { - /* SNE t0, left[i], right[i] */ + /* SNE sneTemp, left[i], right[i] */ inst = new_instruction(emitInfo, OPCODE_SNE); storage_to_src_reg(&inst->SrcReg[0], n->Children[0]->Store); storage_to_src_reg(&inst->SrcReg[1], n->Children[1]->Store); inst->SrcReg[0].Index += i; inst->SrcReg[1].Index += i; if (i == 0) { - inst->DstReg.File = accTemp.File; - inst->DstReg.Index = accTemp.Index; + storage_to_dst_reg(&inst->DstReg, &accTemp, WRITEMASK_XYZW); inst->Comment = _mesa_strdup("Begin struct/array comparison"); } else { - inst->DstReg.File = n->Store->File; - inst->DstReg.Index = n->Store->Index; - } - if (i > 0) { - /* ADD accTemp, accTemp, temp; # like logical-OR */ + storage_to_dst_reg(&inst->DstReg, &sneTemp, WRITEMASK_XYZW); + + /* ADD accTemp, accTemp, sneTemp; # like logical-OR */ inst = new_instruction(emitInfo, OPCODE_ADD); - inst->SrcReg[0].File = accTemp.File; - inst->SrcReg[0].Index = accTemp.Index; - inst->SrcReg[1].File = n->Store->File; - inst->SrcReg[1].Index = n->Store->Index; - inst->DstReg.File = accTemp.File; - inst->DstReg.Index = accTemp.Index; + storage_to_dst_reg(&inst->DstReg, &accTemp, WRITEMASK_XYZW); + storage_to_src_reg(&inst->SrcReg[0], &accTemp); + storage_to_src_reg(&inst->SrcReg[1], &sneTemp); } } /* compute accTemp.x || accTemp.y || accTemp.z || accTemp.w with DOT4 */ inst = new_instruction(emitInfo, OPCODE_DP4); - inst->SrcReg[0].File = accTemp.File; - inst->SrcReg[0].Index = accTemp.Index; - inst->SrcReg[1].File = accTemp.File; - inst->SrcReg[1].Index = accTemp.Index; - inst->DstReg.File = n->Store->File; - inst->DstReg.Index = n->Store->Index; + storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); + storage_to_src_reg(&inst->SrcReg[0], &accTemp); + storage_to_src_reg(&inst->SrcReg[1], &accTemp); inst->Comment = _mesa_strdup("End struct/array comparison"); if (n->Opcode == IR_EQUAL) { /* compute tmp.x = !tmp.x via tmp.x = (tmp.x == 0) */ inst = new_instruction(emitInfo, OPCODE_SEQ); + storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); storage_to_src_reg(&inst->SrcReg[0], n->Store); constant_to_src_reg(&inst->SrcReg[1], 0.0, emitInfo); - storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); inst->Comment = _mesa_strdup("Invert true/false"); } _slang_free_temp(emitInfo->vt, &accTemp); + _slang_free_temp(emitInfo->vt, &sneTemp); } /* free temps */ - free_temp_storage(emitInfo->vt, n->Children[0]); - free_temp_storage(emitInfo->vt, n->Children[1]); + free_node_storage(emitInfo->vt, n->Children[0]); + free_node_storage(emitInfo->vt, n->Children[1]); return inst; } @@ -827,9 +834,8 @@ emit_clamp(slang_emit_info *emitInfo, slang_ir_node *n) } #endif - if (!n->Store) - if (!alloc_temp_storage(emitInfo, n, n->Children[0]->Store->Size)) - return NULL; + if (!alloc_node_storage(emitInfo, n, n->Children[0]->Store->Size)) + return NULL; emit(emitInfo, n->Children[1]); emit(emitInfo, n->Children[2]); @@ -839,7 +845,7 @@ emit_clamp(slang_emit_info *emitInfo, slang_ir_node *n) * the intermediate result. Use a temp register instead. */ _mesa_bzero(&tmpNode, sizeof(tmpNode)); - alloc_temp_storage(emitInfo, &tmpNode, n->Store->Size); + alloc_node_storage(emitInfo, &tmpNode, n->Store->Size); /* tmp = max(ch[0], ch[1]) */ inst = new_instruction(emitInfo, OPCODE_MAX); @@ -853,7 +859,7 @@ emit_clamp(slang_emit_info *emitInfo, slang_ir_node *n) storage_to_src_reg(&inst->SrcReg[0], tmpNode.Store); storage_to_src_reg(&inst->SrcReg[1], n->Children[2]->Store); - free_temp_storage(emitInfo->vt, &tmpNode); + free_node_storage(emitInfo->vt, &tmpNode); return inst; } @@ -870,9 +876,8 @@ emit_negation(slang_emit_info *emitInfo, slang_ir_node *n) emit(emitInfo, n->Children[0]); - if (!n->Store) - if (!alloc_temp_storage(emitInfo, n, n->Children[0]->Store->Size)) - return NULL; + if (!alloc_node_storage(emitInfo, n, n->Children[0]->Store->Size)) + return NULL; inst = new_instruction(emitInfo, OPCODE_MOV); storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); @@ -1014,9 +1019,8 @@ emit_tex(slang_emit_info *emitInfo, slang_ir_node *n) inst = new_instruction(emitInfo, OPCODE_TXP); } - if (!n->Store) - if (!alloc_temp_storage(emitInfo, n, 4)) - return NULL; + if (!alloc_node_storage(emitInfo, n, 4)) + return NULL; storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); @@ -1045,12 +1049,15 @@ emit_tex(slang_emit_info *emitInfo, slang_ir_node *n) } +/** + * Assignment/copy + */ static struct prog_instruction * -emit_move(slang_emit_info *emitInfo, slang_ir_node *n) +emit_copy(slang_emit_info *emitInfo, slang_ir_node *n) { struct prog_instruction *inst; - assert(n->Opcode == IR_MOVE); + assert(n->Opcode == IR_COPY); /* lhs */ emit(emitInfo, n->Children[0]); @@ -1115,7 +1122,7 @@ emit_move(slang_emit_info *emitInfo, slang_ir_node *n) srcStore.Size = 4; while (size >= 4) { inst = new_instruction(emitInfo, OPCODE_MOV); - inst->Comment = _mesa_strdup("IR_MOVE block"); + inst->Comment = _mesa_strdup("IR_COPY block"); storage_to_dst_reg(&inst->DstReg, &dstStore, n->Writemask); storage_to_src_reg(&inst->SrcReg[0], &srcStore); srcStore.Index++; @@ -1135,7 +1142,7 @@ emit_move(slang_emit_info *emitInfo, slang_ir_node *n) inst->Comment = instruction_annotation(inst->Opcode, dstAnnot, srcAnnot, NULL, NULL); } - free_temp_storage(emitInfo->vt, n->Children[1]); + free_node_storage(emitInfo->vt, n->Children[1]); return inst; } } @@ -1184,7 +1191,7 @@ emit_cond(slang_emit_info *emitInfo, slang_ir_node *n) * is normally generated for the expression "i". * Generate a move instruction just to set condition codes. */ - if (!alloc_temp_storage(emitInfo, n, 1)) + if (!alloc_node_storage(emitInfo, n, 1)) return NULL; inst = new_instruction(emitInfo, OPCODE_MOV); inst->CondUpdate = GL_TRUE; @@ -1240,15 +1247,14 @@ emit_not(slang_emit_info *emitInfo, slang_ir_node *n) #endif /* else, invert using SEQ (v = v == 0) */ - if (!n->Store) - if (!alloc_temp_storage(emitInfo, n, n->Children[0]->Store->Size)) - return NULL; + if (!alloc_node_storage(emitInfo, n, n->Children[0]->Store->Size)) + return NULL; inst = new_instruction(emitInfo, OPCODE_SEQ); storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); storage_to_src_reg(&inst->SrcReg[0], n->Children[0]->Store); constant_to_src_reg(&inst->SrcReg[1], 0.0, emitInfo); - free_temp_storage(emitInfo->vt, n->Children[0]); + free_node_storage(emitInfo->vt, n->Children[0]); inst->Comment = _mesa_strdup("NOT"); return inst; @@ -1731,27 +1737,15 @@ emit(slang_emit_info *emitInfo, slang_ir_node *n) case IR_SWIZZLE: return emit_swizzle(emitInfo, n); - case IR_I_TO_F: - /* just move */ - emit(emitInfo, n->Children[0]); - inst = new_instruction(emitInfo, OPCODE_MOV); - if (!n->Store) { - if (!alloc_temp_storage(emitInfo, n, 1)) - return NULL; - } - storage_to_dst_reg(&inst->DstReg, n->Store, n->Writemask); - storage_to_src_reg(&inst->SrcReg[0], n->Children[0]->Store); - if (emitInfo->EmitComments) - inst->Comment = _mesa_strdup("int to float"); - return NULL; - /* Simple arithmetic */ /* unary */ + case IR_MOVE: case IR_RSQ: case IR_RCP: case IR_FLOOR: case IR_FRAC: case IR_F_TO_I: + case IR_I_TO_F: case IR_ABS: case IR_SIN: case IR_COS: @@ -1808,8 +1802,8 @@ emit(slang_emit_info *emitInfo, slang_ir_node *n) } return NULL; - case IR_MOVE: - return emit_move(emitInfo, n); + case IR_COPY: + return emit_copy(emitInfo, n); case IR_COND: return emit_cond(emitInfo, n); diff --git a/src/mesa/shader/slang/slang_ir.c b/src/mesa/shader/slang/slang_ir.c index 23d554234e..3a0b8bf3a0 100644 --- a/src/mesa/shader/slang/slang_ir.c +++ b/src/mesa/shader/slang/slang_ir.c @@ -54,7 +54,8 @@ static const slang_ir_info IrInfo[] = { { IR_NOTEQUAL, "IR_NOTEQUAL", OPCODE_NOP, 1, 2 }, /* unary ops */ - { IR_I_TO_F, "IR_I_TO_F", OPCODE_NOP, 1, 1 }, + { IR_MOVE, "IR_MOVE", OPCODE_MOV, 4, 1 }, + { IR_I_TO_F, "IR_I_TO_F", OPCODE_MOV, 4, 1 }, /* int[4] to float[4] */ { IR_F_TO_I, "IR_F_TO_I", OPCODE_INT, 4, 1 }, /* 4 floats to 4 ints */ { IR_EXP, "IR_EXP", OPCODE_EXP, 1, 1 }, { IR_EXP2, "IR_EXP2", OPCODE_EX2, 1, 1 }, @@ -82,7 +83,7 @@ static const slang_ir_info IrInfo[] = { { IR_KILL, "IR_KILL", OPCODE_NOP, 0, 0 }, { IR_COND, "IR_COND", OPCODE_NOP, 0, 0 }, { IR_CALL, "IR_CALL", OPCODE_NOP, 0, 0 }, - { IR_MOVE, "IR_MOVE", OPCODE_NOP, 0, 1 }, + { IR_COPY, "IR_COPY", OPCODE_NOP, 0, 1 }, { IR_NOT, "IR_NOT", OPCODE_NOP, 1, 1 }, { IR_VAR, "IR_VAR", OPCODE_NOP, 0, 0 }, { IR_VAR_DECL, "IR_VAR_DECL", OPCODE_NOP, 0, 0 }, @@ -326,8 +327,8 @@ _slang_print_ir_tree(const slang_ir_node *n, int indent) assert(!n->Children[1]); _slang_print_ir_tree(n->Children[0], indent + 3); break; - case IR_MOVE: - printf("MOVE (writemask = %s)\n", writemask_string(n->Writemask)); + case IR_COPY: + printf("COPY (writemask = %s)\n", writemask_string(n->Writemask)); _slang_print_ir_tree(n->Children[0], indent+3); _slang_print_ir_tree(n->Children[1], indent+3); break; diff --git a/src/mesa/shader/slang/slang_ir.h b/src/mesa/shader/slang/slang_ir.h index e4697ba3b4..f64f9a93b7 100644 --- a/src/mesa/shader/slang/slang_ir.h +++ b/src/mesa/shader/slang/slang_ir.h @@ -70,14 +70,14 @@ typedef enum /* n->Parent = ptr to parent IR_LOOP Node */ IR_BREAK, /* break loop */ - IR_BREAK_IF_TRUE, + IR_BREAK_IF_TRUE, /**< Children[0] = the condition expression */ IR_CONT_IF_TRUE, - /* Children[0] = the condition expression */ - IR_MOVE, + IR_COPY, /**< assignment/copy */ + IR_MOVE, /**< assembly MOV instruction */ /* vector ops: */ - IR_ADD, + IR_ADD, /**< assembly ADD instruction */ IR_SUB, IR_MUL, IR_DIV, diff --git a/src/mesa/shader/slang/slang_vartable.c b/src/mesa/shader/slang/slang_vartable.c index 68b4e00be0..9b607e6403 100644 --- a/src/mesa/shader/slang/slang_vartable.c +++ b/src/mesa/shader/slang/slang_vartable.c @@ -110,9 +110,10 @@ _slang_pop_var_table(slang_var_table *vt) slang_ir_storage *store = (slang_ir_storage *) t->Vars[i]->aux; GLint j; GLuint comp; - if (dbg) printf(" Free var %s, size %d at %d\n", + if (dbg) printf(" Free var %s, size %d at %d.%s\n", (char*) t->Vars[i]->a_name, store->Size, - store->Index); + store->Index, + _mesa_swizzle_string(store->Swizzle, 0, 0)); if (store->Size == 1) comp = GET_SWZ(store->Swizzle, 0); @@ -159,7 +160,7 @@ _slang_add_variable(slang_var_table *vt, slang_variable *v) assert(vt); t = vt->Top; assert(t); - if (dbg) printf("Adding var %s\n", (char *) v->a_name); + if (dbg) printf("Adding var %s, store %p\n", (char *) v->a_name, v->aux); t->Vars = (slang_variable **) _slang_realloc(t->Vars, t->NumVars * sizeof(slang_variable *), @@ -262,10 +263,11 @@ _slang_alloc_var(slang_var_table *vt, slang_ir_storage *store) } if (dbg) - printf("Alloc var sz %d at %d.%s (level %d)\n", + printf("Alloc var storage sz %d at %d.%s (level %d) store %p\n", store->Size, store->Index, _mesa_swizzle_string(store->Swizzle, 0, 0), - t->Level); + t->Level, + (void*) store); return GL_TRUE; } @@ -283,19 +285,29 @@ _slang_alloc_temp(slang_var_table *vt, slang_ir_storage *store) if (i < 0) return GL_FALSE; + assert(store->Index < 0); + store->Index = i / 4; if (store->Size == 1) { const GLuint comp = i % 4; store->Swizzle = MAKE_SWIZZLE4(comp, comp, comp, comp); - if (dbg) printf("Alloc temp sz %d at %d.%c (level %d)\n", - store->Size, store->Index, "xyzw"[comp], t->Level); } else { /* XXX improve swizzled for size=2/3, use for writemask... */ +#if 1 + if (store->Size == 2) { + store->Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, + SWIZZLE_NIL, SWIZZLE_NIL); + } +#endif store->Swizzle = SWIZZLE_NOOP; - if (dbg) printf("Alloc temp sz %d at %d.xyzw (level %d)\n", - store->Size, store->Index, t->Level); } + + if (dbg) printf("Alloc temp sz %d at %d.%s (level %d) store %p\n", + store->Size, store->Index, + _mesa_swizzle_string(store->Swizzle, 0, 0), t->Level, + (void *) store); + return GL_TRUE; } @@ -309,7 +321,10 @@ _slang_free_temp(slang_var_table *vt, slang_ir_storage *store) assert(store->Size > 0); assert(r >= 0); assert(r + store->Size <= vt->MaxRegisters * 4); - if (dbg) printf("Free temp sz %d at %d (level %d)\n", store->Size, r, t->Level); + if (dbg) printf("Free temp sz %d at %d.%s (level %d) store %p\n", + store->Size, r, + _mesa_swizzle_string(store->Swizzle, 0, 0), + t->Level, (void *) store); if (store->Size == 1) { const GLuint comp = GET_SWZ(store->Swizzle, 0); /* we can actually fail some of these assertions because of the @@ -319,8 +334,8 @@ _slang_free_temp(slang_var_table *vt, slang_ir_storage *store) assert(store->Swizzle == MAKE_SWIZZLE4(comp, comp, comp, comp)); assert(comp < 4); assert(t->ValSize[r * 4 + comp] == 1); - assert(t->Temps[r * 4 + comp] == TEMP); #endif + assert(t->Temps[r * 4 + comp] == TEMP); t->Temps[r * 4 + comp] = FREE; } else { -- cgit v1.2.3 From 8abeeb35d3c9810e191510ebe3194ae9b1b49f35 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 16:24:53 -0600 Subject: mesa: glsl: regenerated files --- .../shader/slang/library/slang_common_builtin_gc.h | 168 ++-- src/mesa/shader/slang/library/slang_core_gc.h | 1016 ++++++++++---------- 2 files changed, 579 insertions(+), 605 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index 0a43cad2a3..b7c7f4032a 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -621,90 +621,90 @@ 101,114,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18, 95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84, 104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104, -97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104, -97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116, -86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113, -117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84, -104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103, -114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4, -118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2, -0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1, -4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118, -0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0, -7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, -18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1, -0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0, -0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4, -118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0, -0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115, -101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0, -101,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95, -95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,6,117,0, -0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4, -118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18, -118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95, -115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97, -108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,3,117,0,0,1,1, -0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, -0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118, -101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0, -110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111, -116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69, -113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95, -114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0, -6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1, -0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, -0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0, -0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0, -110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111, -116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, -95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1,3, -2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59, -120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,3,118,0,0,0,1, -3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0, -59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18, -115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1, -1,0,4,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, -120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0, -59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,100,100,0,18, -115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0, -0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59, -121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,114, -111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,8,18,118,0,59,120,0,18,118,0,59,121,0,34,0,0,1,0,1,0,97, -108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114,111,100, -0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108, -0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,4,118, -0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,18,118,0, -59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0, -18,112,114,111,100,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, -114,101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,2,0, -110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4,118, -101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0, -48,0,0,0,0,0,1,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, -114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49, -68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95, -116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, +116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,3,0,103,114, +101,97,116,101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103, +114,101,97,116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115, +103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116, +101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95, +115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0, +103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1, +4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18, +118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0, +1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0, +0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117, +0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97, +108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97, +110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95, +95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0, +0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0, +1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0, +18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52, +95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117, +97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0, +1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1, +4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0, +2,0,101,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108, +0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1, +1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, +18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118, +101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4, +0,110,111,116,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117, +97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0, +7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0, +1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, +18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118, +101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4, +0,110,111,116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,2, +118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0, +0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0, +3,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120, +0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, +120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97, +110,121,0,1,1,0,4,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115, +117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18, +115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97, +100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101, +99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17, +48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0, +18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, +18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,8,18,118,0,59,120,0,18,118,0,59,121,0,34,0,0,1, +0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0, +0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114, +111,100,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1, +1,0,4,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0, +18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59, +120,0,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0, +2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4, +118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17, +48,0,48,0,0,0,0,0,1,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95, +95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101, +49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52, +95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, 111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115, 97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112, 49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index 33a996f2d9..ce7a5cc64a 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -2,535 +2,509 @@ /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ /* slang_core.gc */ -3,1,0,5,1,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101, +3,1,0,5,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101, 116,86,97,108,0,0,18,102,0,0,0,0,1,0,5,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, 98,0,20,0,0,1,0,5,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,20,0,0,1,0,1,1, -1,1,0,5,105,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,18,122,101,114,111,0,0,0,0,1,0,1,1,1,1,0,9,102, -0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, -114,101,116,86,97,108,0,0,18,102,0,0,18,122,101,114,111,0,0,0,0,1,0,1,1,1,1,0,1,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,5,105,0,0,0,1,4,105,110,116,95,116,111,95, -102,108,111,97,116,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,0,9,1,1,1,0,1,98,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,102,0,20,0,0,1,0,10,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,0, -1,0,10,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,102,0,59,120,120,0, -20,0,0,1,0,10,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,105,0,59, -120,120,0,20,0,0,1,0,10,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98, -0,59,120,120,0,20,0,0,1,0,10,1,1,1,0,2,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0, -0,1,0,10,1,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,118,0,59,120, -121,0,20,0,0,1,0,10,1,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,115,116,0,18,118, -0,59,120,121,0,20,0,0,1,0,11,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,0,1,0,11,1,1,1,0,9,102,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,18,102,0,59,120,120,120,0,20,0,0,1,0,11,1,1,1,0,5,105, -0,0,0,1,4,105,110,116,95,116,111,95,102,108,111,97,116,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,122,0,0,18,105,0,59,120,120,120,0,0,0,0,1,0,11,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,11,1,1,1,0,3,98,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,11,1,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,18,118,0,59,120,121,122,0,20,0,0,1,0,12,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1, -1,0,9,122,0,0,1,1,0,9,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0, -18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,12,1,1,1,0,9,102,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,59,120,120,120,120,0,20,0,0,1,0,12,1,1,1,0,5,105,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,12,1,1,1,0,1,98, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,59,120,120,120,120,0,20,0,0,1,0,12,1,1,1,0,4,98, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,12,1,1,1,0,8,105,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,18,105,0,20,0,0,1,0,12,1,1,1,0,11,118,51,0,0,1,1,0,9,102,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,18,118,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -119,0,18,102,0,20,0,0,1,0,12,1,1,1,0,10,118,50,0,0,1,1,0,9,102,49,0,0,1,1,0,9,102,50,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,18,118,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -122,0,18,102,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,50,0,20,0,0,1,0,6,1,1,1, -0,5,105,0,0,1,1,0,5,106,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,18,105,0,59,120,120,0,20,0,0,1,0,6,1,1,1,0,9,102,0,0,0,1,4,102,108, -111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,59, -120,120,0,0,0,0,1,0,6,1,1,1,0,1,98,0,0,0,1,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,7,1,1,1,0,5,105,0,0,1,1,0, -5,106,0,0,1,1,0,5,107,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,107, -0,20,0,0,1,0,7,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,105,0, -59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,18,102,0,59,120,120,120,0,20,0,0,1,0,7,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,8,1,1,1,0,5,120,0,0,1,1,0,5,121,0,0,1,1,0,5, -122,0,0,1,1,0,5,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,8,1,1,1,0,5,105,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,18,105,0,59,120,120,120,120,0,20,0,0,1,0,8,1,1,1,0,9,102,0,0,0,1, -4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,59, -120,120,120,120,0,0,0,0,1,0,8,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,98,0,59, -120,120,120,120,0,20,0,0,1,0,2,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,0,1, -0,2,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,59,120,120,0,20,0, -0,1,0,2,1,1,1,0,9,102,0,0,0,1,3,2,1,10,1,122,101,114,111,0,2,58,118,101,99,50,0,17,48,0,48,0,0,0, -17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,102,0,59,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,3,2,1,6,1, -122,101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,59,120,120,0,0,18,122,101,114, -111,0,0,0,0,1,0,2,1,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,3,2,1,6,1,122, -101,114,111,0,2,58,105,118,101,99,50,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3, -1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,18,98,0,59,120,120,120,0,20,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,3,2,1,11, -1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0, -4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,59, -120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,5,105,0,0,0,1,3,2,1,7,1,122,101,114,111,0, -2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,59,120,120,120,0,0,18,122,101,114, -111,0,0,0,0,1,0,3,1,1,1,0,11,118,0,0,0,1,3,2,1,11,1,122,101,114,111,0,2,58,118,101,99,51,0,17,48,0, -48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,3,1,1,1,0,7,118,0,0,0, -1,3,2,1,7,1,122,101,114,111,0,2,58,105,118,101,99,51,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18, -122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0,0,1,1,0,1,98, -52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,4,1,1,1,0,9,102,49,0,0,1,1,0,9,102,50, -0,0,1,1,0,9,102,51,0,0,1,1,0,9,102,52,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2,17,48,0,48,0,0,0,0,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,102,49,0,0,18,122, -101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18, -102,50,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, -108,0,59,122,0,0,18,102,51,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, -114,101,116,86,97,108,0,59,119,0,0,18,102,52,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,1,98,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,18,98,0,59,120,120,120,120,0,20,0,0,1, -0,4,1,1,1,0,9,102,0,0,0,1,3,2,1,12,1,122,101,114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48, -0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114, -101,116,86,97,108,0,0,18,102,0,59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,5, -105,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101,99,52,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0, -0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0, -59,120,120,120,120,0,0,18,122,101,114,111,0,0,0,0,1,0,4,1,1,1,0,12,118,0,0,0,1,3,2,1,12,1,122,101, -114,111,0,2,58,118,101,99,52,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0, -0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,122,101, -114,111,0,0,0,0,1,0,4,1,1,1,0,8,118,0,0,0,1,3,2,1,8,1,122,101,114,111,0,2,58,105,118,101,99,52,0, -16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,16,8,48,0,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, -114,101,116,86,97,108,0,0,18,118,0,0,18,122,101,114,111,0,0,0,0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1, -0,9,109,49,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121, -0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,0,13,1,1,1,0,9, -102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,49,0,57,59,120,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121, -0,18,102,0,20,0,0,1,0,13,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,105, -0,0,0,0,0,0,0,1,0,13,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,98,0,0,0, -0,0,0,0,1,0,13,1,1,1,0,10,99,48,0,0,1,1,0,10,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,14, -1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9, -109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59, -120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122, -0,18,109,50,50,0,20,0,0,1,0,14,1,1,1,0,9,102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102, -0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59,121,121,120,0,20,0,0,1,0,14,1,1,1,0,5,105,0,0,0, -1,8,58,109,97,116,51,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,14,1,1,1,0,1,98,0,0,0,1,8, -58,109,97,116,51,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,14,1,1,1,0,11,99,48,0,0,1,1,0, -11,99,49,0,0,1,1,0,11,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,0,15,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9, -109,50,48,0,0,1,1,0,9,109,51,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49, -0,0,1,1,0,9,109,51,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,1,1,0, -9,109,51,50,0,0,1,1,0,9,109,48,51,0,0,1,1,0,9,109,49,51,0,0,1,1,0,9,109,50,51,0,0,1,1,0,9,109,51, -51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119, -0,18,109,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20, +1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0, +17,48,0,48,0,0,0,0,0,1,0,1,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, +116,86,97,108,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,1,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,5,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118, +101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,0,9,1,1,1,0,1,98,0,0,0,1,4,105,118, +101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,9,1,1, +1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,20,0,0,1,0,10,1,1,1,0,9,120,0,0,1,1, +0,9,121,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,59,121,0,18,121,0,20,0,0,1,0,10,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,109,111,118, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,0,0,1,0,10,1,1,1,0,5,105,0,0,0,1, +4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,105,0,0,0,0,1,0,10,1,1,1,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,10,1,1,1,0,2,98,0,0,0,1,4,105,118,101, +99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0, +1,0,10,1,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108, +0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,10,1,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,109, +111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,11,1, +1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18, +120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,122,0,18,122,0,20,0,0,1,0,11,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,0,11,1,1,1,0,5,105,0,0,0,1,4, +105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,105,0,0,0,0,1,0,11,1,1,1,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,11,1,1,1,0,3,98,0,0,0,1,4,105, +118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, +98,0,0,0,0,1,0,11,1,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,0,18,118,0,0,0,0,1,0,12,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122, +0,0,1,1,0,9,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,12,1,1,1,0,9,102,0,0,0,1,4,118, +101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,0,0,1,0,12,1,1,1,0,5, +105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0, +18,105,0,0,0,0,1,0,12,1,1,1,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95, +95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,12,1,1,1,0,4,98,0,0,0,1,4,105,118,101,99,52,95,116, +111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,12,1,1,1,0,8,105,0,0,0, +1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0, +0,0,1,0,12,1,1,1,0,11,118,51,0,0,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,18,118,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,0,20,0,0,1,0,12,1,1,1,0, +10,118,50,0,0,1,1,0,9,102,49,0,0,1,1,0,9,102,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,18,118,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,102,49,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,119,0,18,102,50,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0, +18,106,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,105,0,0,0,0,1,0,6,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116, +111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,0,0,1,0,6,1, +1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,7,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,1,1,0,5,107,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121, +0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,107,0,20,0,0,1,0,7,1,1,1,0,5,105,0, +0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, +105,0,0,0,0,1,0,7,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,0,7,1,1,1,0,1,98,0,0,0,1,4,118,101,99, +52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,8,1, +1,1,0,5,120,0,0,1,1,0,5,121,0,0,1,1,0,5,122,0,0,1,1,0,5,119,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119, +0,20,0,0,1,0,8,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86, +97,108,0,0,18,105,0,0,0,0,1,0,8,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99, +52,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,0,0,1,0,8,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52, +95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,2,1,1,1,0,1, +98,49,0,0,1,1,0,1,98,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,0,2,1,1,1,0,1,98,0,0,0,1,4,118,101,99, +52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,2,1,1,1, +0,9,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,10,118,0,0, +0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17, +48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,1,1,1,0,1,98,49,0,0,1,1,0,1, +98,50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0, +18,98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0, +1,0,3,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,3,1,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0, +1,0,3,1,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,4,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0, +1,98,51,0,0,1,1,0,1,98,52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0, +18,98,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,4,1,1,1,0,9,102, +49,0,0,1,1,0,9,102,50,0,0,1,1,0,9,102,51,0,0,1,1,0,9,102,52,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2, +17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, +102,49,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, +108,0,59,121,0,0,18,102,50,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, +114,101,116,86,97,108,0,59,122,0,0,18,102,51,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,102,52,0,0,18,122,101,114,111,0,0,0,0,1,0, +4,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,119,0,0,18,98,0,0,0,0,1,0,4,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,4,1,1,1,0, +5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119, +0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,4,1,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,4,1,1, +1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, +119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0, +9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0, +18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,0,13,1,1,1,0,9,102,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0, +17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,0,1,0, +13,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,13, +1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,13,1,1,1, +0,10,99,48,0,0,1,1,0,10,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,14,1,1,1,0,9,109,48,48, +0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0, +9,109,50,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, +0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20, 0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101, 116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,59,119,0,18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0, -18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,51,0,57,59,120,0,18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0, -18,109,49,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0,0,1,0,15,1,1,1,0,9, -102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, -0,57,18,118,0,59,121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0, -59,121,121,121,120,0,20,0,0,1,0,15,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97, -116,0,18,105,0,0,0,0,0,0,0,1,0,15,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116, -0,18,98,0,0,0,0,0,0,0,1,0,15,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,1,1,0,12, -99,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101, +10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0, +18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0, +0,1,0,14,1,1,1,0,9,102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0, +0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,50,0,57,18,118,0,59,121,121,120,0,20,0,0,1,0,14,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,51,0, +58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,14,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,51,0,58, +102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,14,1,1,1,0,11,99,48,0,0,1,1,0,11,99,49,0,0,1,1,0,11, +99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101, 116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -99,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,5,2,26,1,1,0,5, -97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0,0, -18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,0,0,1,0,5,2,27,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95, -115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95, -105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5, -98,0,0,0,1,3,2,0,9,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18, -97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108, -0,0,18,120,0,0,0,0,1,0,5,2,22,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,1,1, -120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97,116, -95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,26,1,1,0,6, -97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120,0,0,18,97,0, -0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0, -18,120,0,0,0,0,1,0,6,2,27,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52, -95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111, -95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,21,1,1,0,6,97,0,0,1,1,0, -6,98,0,0,0,1,3,2,0,10,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0, -18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97, -108,0,0,18,120,0,0,0,0,1,0,6,2,22,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,98,73,110,118,0,0,1, -1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102, -108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7, -2,26,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120, -0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86, -97,108,0,0,18,120,0,0,0,0,1,0,7,2,27,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118, -101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116, -95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,21,1,1,0,7, -97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, -0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114, -101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,7,2,22,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,11,1,98,73, -110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98, -0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0, -0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102, -108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8, -2,26,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,11,1,120,0,0,0,4,118,101,99,52,95,97,100,100,0,18,120, -0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86, -97,108,0,0,18,120,0,0,0,0,1,0,8,2,27,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118, -101,99,52,95,115,117,98,116,114,97,99,116,0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116, -95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,21,1,1,0,8, -97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, -0,18,120,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,95,95,114, -101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,8,2,22,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,3,2,0,12,1,98,73, -110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98, -0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0, -0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,119,0,0,18,98,0,59,119,0,0,0,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,102,108,111,97, -116,95,116,111,95,105,110,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,2,26,1,1,0, -9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, -115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0, -1,0,9,2,21,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,22,1,1,0,9,97,0,0,1,1, -0,9,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118, -0,59,120,0,0,18,98,0,59,120,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95, -114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0, -1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, -0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95, -115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0, -0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,22,1, -1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18, -117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,118,0,0,18,119,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4, -118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18, -117,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116, -114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0, -11,2,21,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,22,1,1,0,11, -118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59, -120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59, -121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -118,0,0,18,119,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100, -100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1, -1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97, -108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99, -52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0, -0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119, -0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117, -0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4, -118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0, -18,119,0,0,0,0,1,0,10,2,26,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1, -0,10,2,26,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,27,1,1,0,9, -97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,27,1,1,0,10, -118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,21,1,1,0,9, -97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,21,1,1, -0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,10,2,22, -1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99, -112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,120,0,0,18,105,110,118,85,0, -59,120,121,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0, -18,105,110,118,66,0,59,120,120,0,0,0,0,1,0,11,2,26,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101, -99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0, -0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52, -95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18, -98,0,59,120,120,120,0,0,0,0,1,0,11,2,27,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120, -120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101, -99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118, -0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,21,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0, -1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,97,0,59,120,120,120,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1, -1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,11,2,22,1, -1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99, -112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, -118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,18,105,110,118,85,0,59, -120,121,122,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121, -122,0,0,18,105,110,118,66,0,59,120,120,120,0,0,0,0,1,0,12,2,26,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1, -4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0, -18,117,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, -95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,27,1,1,0,9, -97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, -86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9, -98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18, -118,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,12,2,21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120, -120,120,120,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,59,120,120, -120,120,0,0,0,0,1,0,12,2,22,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,105,110,118,85,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,120,120,120,0,0,18,105,110, -118,85,0,0,0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,59,120, -120,120,120,0,0,0,0,1,0,6,2,26,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,6, -2,27,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0, -18,97,0,0,0,18,117,0,47,20,0,0,1,0,6,2,27,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,47,20,0,0,1,0,6,2,21,1,1,0,5,97,0,0,1, -1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0, -48,20,0,0,1,0,6,2,21,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118, -0,58,105,118,101,99,50,0,18,98,0,0,0,48,20,0,0,1,0,6,2,22,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,6,2,22,1,1, -0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0, -18,98,0,0,0,49,20,0,0,1,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,46,20,0,0,1, -0,7,2,27,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99, -51,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,7,2,27,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,47,20,0,0,1,0,7,2,21,1,1,0,5,97,0, -0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117, -0,48,20,0,0,1,0,7,2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,58,105,118,101,99,51,0,18,98,0,0,0,48,20,0,0,1,0,7,2,22,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,7,2,22, -1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99, -51,0,18,98,0,0,0,49,20,0,0,1,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5, -98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,46,20,0, -0,1,0,8,2,27,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101, -99,52,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,8,2,27,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,47,20,0,0,1,0,8,2,21,1,1,0,5, -97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0, -18,117,0,48,20,0,0,1,0,8,2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,48,20,0,0,1,0,8,2,22,1,1,0,5,97,0,0,1,1,0,8,117,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,8, -2,22,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, -99,52,0,18,98,0,0,0,49,20,0,0,1,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,6,2,27,1,1,0,6,118,0,0,0,1,4, -118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,7, -2,27,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97, -108,0,0,18,118,0,0,0,0,1,0,8,2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0, -18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, -110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,2,27,1,1, -0,10,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,110, -101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0, -0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114, -101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,13,2,27,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -18,109,0,16,10,49,0,57,54,20,0,0,1,0,14,2,27,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0, -57,54,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,0,9,0,100,111,116,0,1,1,0, -9,97,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,48,20,0,0,1,0,9,0, -100,111,116,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,59, -120,0,18,98,0,59,120,0,48,18,97,0,59,121,0,18,98,0,59,121,0,48,46,20,0,0,1,0,9,0,100,111,116,0,1,1, -0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0, -0,18,97,0,0,18,98,0,0,0,0,1,0,9,0,100,111,116,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99, -52,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,1,1,0,2,5, -97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0, -2,2,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,0, -18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0, -1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98, -0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,105,110,118,66,0, -0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,97,0,0,18,97,0,0,0,0,1,0,0,2,1,1,0,2,6,118, -0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0, -2,2,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0, -0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0, -6,117,0,0,0,1,3,2,0,6,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105, -110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0, -59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0, -18,118,0,0,18,105,110,118,0,0,0,4,102,108,111,97,116,95,116,111,95,105,110,116,0,18,118,0,0,18,122, -0,0,0,0,1,0,0,2,1,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0, -18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117, -98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,7,117, -0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0, -1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,3,2,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,102,108,111,97,116,95,116,111, -95,105,110,116,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101, -99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0, -2,3,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,3,2,0,8,1,105,110, -118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59, -120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4, -118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4, -102,108,111,97,116,95,116,111,95,105,110,116,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,9,97,0,0, -1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0, -59,120,0,0,0,0,1,0,0,2,2,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114, -97,99,116,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1, -4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1, -0,0,2,4,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,119,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59, -120,0,0,18,97,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52, -95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0, -2,2,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118, -0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1, -1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18, -118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,3, -2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,119,0,59,120, -121,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118, -0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4, -118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0, -0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,11, +99,50,0,20,0,0,1,0,15,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9, +109,51,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,51,49, +0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,1,1,0,9,109,51,50,0,0,1,1,0, +9,109,48,51,0,0,1,1,0,9,109,49,51,0,0,1,1,0,9,109,50,51,0,0,1,1,0,9,109,51,51,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, +0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,109,51,48,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0, +18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0, +18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,109,49,51,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0,0,1,0,15,1,1,1,0,9,102,0,0,0,1,3,2,0,10, +1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0, +16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, +18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59, +121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0,59,121,121,121, +120,0,20,0,0,1,0,15,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,105,0,0, +0,0,0,0,0,1,0,15,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0, +0,0,1,0,15,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,1,1,0,12,99,51,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,5,2,26,1,1,0,5,97,0,0,1,1,0,5,98, +0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0, +1,0,5,2,27,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, +95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0, +1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0, +0,18,98,0,0,0,0,1,0,5,2,22,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,1,1,120,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116, +111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,26,1,1,0,6,97, +0,0,1,1,0,6,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0, +0,18,98,0,0,0,0,1,0,6,2,27,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114, +97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,6,2,21,1,1,0,6,97,0,0,1, +1,0,6,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, +108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,6,2,22,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,98,73,110, +118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0, +59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0, +0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0, +4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, +0,1,0,7,2,26,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101, +116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,7,2,27,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,4,118,101, +99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0, +1,0,7,2,21,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, +18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,7,2,22,1,1,0,7,97,0,0,1,1,0,7,98,0, +0,0,1,3,2,0,11,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110, +118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59, +121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18, +98,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98, +73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108, +0,0,18,120,0,0,0,0,1,0,8,2,26,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, +95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,8,2,27,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0, +1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, +18,98,0,0,0,0,1,0,8,2,21,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,8,2,22,1,1,0,8,97,0,0, +1,1,0,8,98,0,0,0,1,3,2,0,12,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73, +110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0, +59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,119,0,0, +18,98,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18, +98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97, +108,0,0,18,120,0,0,0,0,1,0,9,2,26,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100, +0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,1,1,0,9,98, +0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97, +0,0,18,98,0,0,0,0,1,0,9,2,21,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,22,1,1,0,9,97, +0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73, +110,118,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, +101,116,86,97,108,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0,1,1,0,10,117, +0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0, +18,117,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116, +114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2, +21,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0, +0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0, +0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0, +0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,118,0,0,18,119,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99, +52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1, +0,11,2,27,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,21,1,1,0,11, +118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,11, 117,0,0,0,1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0, 59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102, 108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,119,0,59, -120,121,122,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100, -0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101, -99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,12, -118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18, -118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119, -0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117, -0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0, -0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120, -121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,5,97,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120, -121,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,59,120, -120,0,0,0,9,18,118,0,59,120,0,18,97,0,23,0,9,18,118,0,59,121,0,18,97,0,23,0,0,1,0,0,2,4,1,0,2,6, -118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,0,1,0, -0,2,1,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0, -0,18,118,0,59,120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,7,118,0,0,1,1,0,5,97, -0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,59, -120,121,122,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122, -0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120, -0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2, -8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120, -120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116, -114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0, -1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18, -97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0, -18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,9,18,118,0,59,119,0, -18,97,0,24,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, -118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0,9,97,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0, -59,120,120,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,59,120,120,0,0,0,0,1,0,0,2,4,1,0, -2,10,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112, -0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, -59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,65,0,59,120,120,0,0,0,0,1,0,0,2,1,1,0,2,11, -118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0, -18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, -115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0, -0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121, -0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,59,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0, -1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, -118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121, -122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,65,0,59,120,120,120,0,0,0,0,1,0,0,2,1,1,0,2,12, -118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,59,120, -120,120,120,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98, -116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,3,1,0,2,12, -118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118, -0,0,18,97,0,59,120,120,120,120,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105, -110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,105,110,118,65,0,59,120,120, -120,120,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13, -109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48, -18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49, -0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49, -0,57,49,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1, -0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, -16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120, -120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50, -0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122, -122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110, -0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0, -48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1, -0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, -16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0, -57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10, -51,0,57,46,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2, -21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48, -0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0, -48,46,18,109,0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120, -0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, -57,18,110,0,16,10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57, -59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48, -0,57,18,110,0,16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57, -59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48, -46,18,109,0,16,10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120, -0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0, -57,18,110,0,16,10,51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, -59,119,119,119,119,0,48,46,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, +117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18, +119,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, +95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,12, +117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0, +18,118,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12, +2,22,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112, +0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0, +0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52, +95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0, +1,0,10,2,26,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0, +1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,0,10,2,27,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99, +52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18, +117,0,59,120,121,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0, +18,98,0,0,0,0,1,0,10,2,21,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0, +0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1, +0,10,2,22,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,105,110,118,85,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114, +99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,105,110,118,85,0,59, +120,121,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0, +18,105,110,118,66,0,0,0,0,1,0,11,2,26,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97, +100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0, +0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,11,2,27,1,1,0, +9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,27,1,1,0,11, +118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,11,2,21,1,1,0,9,97,0, +0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,21,1,1,0,11,118, +0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,11,2,22,1,1,0,9,97,0, +0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105, +110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, +85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59, +122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0,0,1,0,11, +2,22,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95, +114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118, +66,0,0,0,0,1,0,12,2,26,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95, +95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,9,98,0,0, +0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1, +0,12,2,27,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, +95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9,98,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118, +0,0,18,98,0,0,0,0,1,0,12,2,21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1, +0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, +101,116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1,0,12,2,22,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,3,2, +0,12,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18, +117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59, +121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,105, +110,118,85,0,0,0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, +4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,0,0,0, +1,0,6,2,26,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99, +50,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,6,2,27,1,1,0,5,97,0, +0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117, +0,47,20,0,0,1,0,6,2,27,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,58,105,118,101,99,50,0,18,98,0,0,0,47,20,0,0,1,0,6,2,21,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,6,2,21, +1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99, +50,0,18,98,0,0,0,48,20,0,0,1,0,6,2,22,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,6,2,22,1,1,0,6,118,0,0,1,1,0,5, +98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,49,20,0, +0,1,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101, +99,51,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,46,20,0,0,1,0,7,2,27,1,1,0,5, +97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0, +18,117,0,47,20,0,0,1,0,7,2,27,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,47,20,0,0,1,0,7,2,21,1,1,0,5,97,0,0,1,1,0,7,117,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,7, +2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, +99,51,0,18,98,0,0,0,48,20,0,0,1,0,7,2,22,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,7,2,22,1,1,0,7,118,0,0,1,1,0, +5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,49,20, +0,0,1,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, +101,99,52,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,46,20,0,0,1,0,8,2,27,1,1,0, +5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0, +18,117,0,47,20,0,0,1,0,8,2,27,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,47,20,0,0,1,0,8,2,21,1,1,0,5,97,0,0,1,1,0,8,117,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,8, +2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, +99,52,0,18,98,0,0,0,48,20,0,0,1,0,8,2,22,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,8,2,22,1,1,0,8,118,0,0,1,1,0, +5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,49,20, +0,0,1,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,6,2,27,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,110,101,103, +97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,7,2,27,1,1,0,7,118,0,0,0,1,4, +118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,8, +2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97, +108,0,0,18,118,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0, +18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,0,1,4,118, +101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59, +120,121,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,0,0,1,0,12,2,27,1,1,0,12, +118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118, +0,0,0,0,1,0,13,2,27,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20, +0,0,1,0,14,2,27,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8, +48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,0,1,0,15,2,27,1,1,0, +15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,0,9,0,100,111,116,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,48,20,0,0,1,0,9,0,100,111,116,0,1,1,0,10,97,0,0,1, +1,0,10,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0, +59,121,0,18,98,0,59,121,0,48,46,20,0,0,1,0,9,0,100,111,116,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4, +118,101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0, +100,111,116,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,18,95,95,114, +101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,1,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118, +101,99,52,95,97,100,100,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,5,97,0,0,1,1,0,5,98,0, +0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2, +3,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0, +18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, +4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,105,110,118,66,0,0,0,4,118,101,99,52,95,116,111, +95,105,118,101,99,52,0,18,97,0,0,18,97,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118, +101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,6, +117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0, +1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, +18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,3,2,0,6,1,105, +110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117, +0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0, +0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0, +0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,7, +118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1, +0,0,2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, +118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1, +1,0,7,117,0,0,0,1,3,2,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, +0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0, +18,118,0,0,18,105,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,118,0,0,18, +122,0,0,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0, +0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,8, +117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0, +0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,3,2,0,8,1,105,110,118,0,0,1,1,122,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,118,101,99,52,95,116,111,95, +105,118,101,99,52,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118, +101,99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,0,1,0,0,2,2,1, +0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,59,120,0, +0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,9,97,0,0,1,1, +0,9,98,0,0,0,1,3,2,0,9,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,98, +0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,119,0,0, +0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59, +120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0, +10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0, +59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18, +117,0,59,120,121,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,119,0,59,120,121,0,0,0,0,1,0,0,2, +1,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0, +0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2, +11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59, +120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11, +1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95, +114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,119,0,59,120,121,122,0,0,0,0, +1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118, +0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,12,117,0, +0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1, +0,0,2,4,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59, +121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59, +122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2,6, +118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120, +121,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2, +6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120, +121,0,0,18,118,0,59,120,121,0,0,18,97,0,0,0,9,18,118,0,59,120,0,18,97,0,23,0,9,18,118,0,59,121,0, +18,97,0,23,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18, +118,0,59,121,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97, +100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0,2,7, +118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121, +122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4, +118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121, +122,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24, +0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1, +1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0, +2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118, +0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1, +9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0, +9,18,118,0,59,119,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52, +95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0, +9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0, +18,97,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,9, +97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0, +0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118, +0,59,120,121,0,0,18,105,110,118,65,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101, +99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0,2,11,118, +0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0, +0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2, +11,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59, +120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,65,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1, +1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0, +2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18, +118,0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0, +0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18, +97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,105,110,118, +65,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16, +8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0, +1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110, +0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48,18,109,0, +16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49, +20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1,0,14,109,0, +0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50, +0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,0, +48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18, +110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16, +10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122,0, +48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10, +50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0,48,46, +18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1,0,14, +109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, +18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, +57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10, +50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0, +57,46,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,21,1,1, +0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59, +121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18, +109,0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120,0,48,18, +109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18, +110,0,16,10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59, +119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0, +57,18,110,0,16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59, +121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48,46, +18,109,0,16,10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48, +18,109,0,16,10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18, +110,0,16,10,51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59, +119,119,119,119,0,48,46,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, 101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95, 114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18, 95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0, -- cgit v1.2.3 From 013de307edd8590ae5a3444d3c18990ee4fd288e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 16:58:31 -0600 Subject: mesa: glsl: re-enable assignment type checking --- src/mesa/shader/slang/slang_codegen.c | 60 +++++++++++++++++------------------ 1 file changed, 30 insertions(+), 30 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 582b47c239..036196a8d2 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2860,36 +2860,46 @@ _slang_gen_swizzle(slang_ir_node *child, GLuint swizzle) * XXX more cases needed. */ static GLboolean -_slang_assignment_compatible(const slang_typeinfo *t0, - const slang_typeinfo *t1) - +_slang_assignment_compatible(slang_assemble_ctx *A, + slang_operation *op0, + slang_operation *op1) { -#if 1 - GLuint sz0 = _slang_sizeof_type_specifier(&t0->spec); - GLuint sz1 = _slang_sizeof_type_specifier(&t1->spec); + slang_typeinfo t0, t1; + GLuint sz0, sz1; + + + slang_typeinfo_construct(&t0); + _slang_typeof_operation(A, op0, &t0); + + slang_typeinfo_construct(&t1); + _slang_typeof_operation(A, op1, &t1); + + sz0 = _slang_sizeof_type_specifier(&t0.spec); + sz1 = _slang_sizeof_type_specifier(&t1.spec); +#if 1 if (sz0 != sz1) { printf("size mismatch %u vs %u\n", sz0, sz1); return GL_FALSE; } #endif - if (t0->spec.type == SLANG_SPEC_STRUCT && - t1->spec.type == SLANG_SPEC_STRUCT && - t0->spec._struct->a_name != t1->spec._struct->a_name) + if (t0.spec.type == SLANG_SPEC_STRUCT && + t1.spec.type == SLANG_SPEC_STRUCT && + t0.spec._struct->a_name != t1.spec._struct->a_name) return GL_FALSE; #if 0 /* not used just yet - causes problems elsewhere */ - if (t0->spec.type == SLANG_SPEC_INT && - t1->spec.type == SLANG_SPEC_FLOAT) + if (t0.spec.type == SLANG_SPEC_INT && + t1.spec.type == SLANG_SPEC_FLOAT) return GL_FALSE; - if (t0->spec.type == SLANG_SPEC_BOOL && - t1->spec.type == SLANG_SPEC_FLOAT) + if (t0.spec.type == SLANG_SPEC_BOOL && + t1.spec.type == SLANG_SPEC_FLOAT) return GL_FALSE; - if (t0->spec.type == SLANG_SPEC_BOOL && - t1->spec.type == SLANG_SPEC_INT) + if (t0.spec.type == SLANG_SPEC_BOOL && + t1.spec.type == SLANG_SPEC_INT) return GL_FALSE; #endif return GL_TRUE; @@ -2943,21 +2953,11 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper) slang_ir_node *n, *lhs, *rhs; /* lhs and rhs type checking */ - if (0) - { - slang_typeinfo t0, t1; - - slang_typeinfo_construct(&t0); - _slang_typeof_operation(A, &oper->children[0], &t0); - - slang_typeinfo_construct(&t1); - _slang_typeof_operation(A, &oper->children[1], &t1); - - if (!_slang_assignment_compatible(&t0, &t1)) { - slang_info_log_error(A->log, - "illegal types in assignment"); - return NULL; - } + if (!_slang_assignment_compatible(A, + &oper->children[0], + &oper->children[1])) { + slang_info_log_error(A->log, "illegal types in assignment"); + return NULL; } lhs = _slang_gen_operation(A, &oper->children[0]); -- cgit v1.2.3 From 7a63e317d1911166e02bdcda21614581ace5004b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 16:58:44 -0600 Subject: mesa: glsl: remove invalid use of f.x syntax where f is a float --- .../shader/slang/library/slang_common_builtin.gc | 87 +++++++++++----------- 1 file changed, 43 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index f41d1fb603..9f2a4bdd07 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -171,7 +171,7 @@ uniform gl_FogParameters gl_Fog; float radians(const float deg) { const float c = 3.1415926 / 180.0; - __asm vec4_multiply __retVal.x, deg, c; + __asm vec4_multiply __retVal, deg, c; } vec2 radians(const vec2 deg) @@ -198,7 +198,7 @@ vec4 radians(const vec4 deg) float degrees(const float rad) { const float c = 180.0 / 3.1415926; - __asm vec4_multiply __retVal.x, rad, c; + __asm vec4_multiply __retVal, rad, c; } vec2 degrees(const vec2 rad) @@ -224,7 +224,7 @@ vec4 degrees(const vec4 rad) float sin(const float radians) { - __asm float_sine __retVal.x, radians; + __asm float_sine __retVal, radians; } vec2 sin(const vec2 radians) @@ -253,7 +253,7 @@ vec4 sin(const vec4 radians) float cos(const float radians) { - __asm float_cosine __retVal.x, radians; + __asm float_cosine __retVal, radians; } vec2 cos(const vec2 radians) @@ -444,7 +444,7 @@ vec4 atan(const vec4 u, const vec4 v) float pow(const float a, const float b) { - __asm float_power __retVal.x, a, b; + __asm float_power __retVal, a, b; } vec2 pow(const vec2 a, const vec2 b) @@ -507,7 +507,7 @@ vec4 exp(const vec4 a) float log2(const float x) { - __asm float_log2 __retVal.x, x; + __asm float_log2 __retVal, x; } vec2 log2(const vec2 v) @@ -566,7 +566,7 @@ vec4 log(const vec4 v) float exp2(const float a) { - __asm float_exp2 __retVal.x, a; + __asm float_exp2 __retVal, a; } vec2 exp2(const vec2 a) @@ -597,7 +597,7 @@ float sqrt(const float x) { float r; __asm float_rsq r, x; - __asm float_rcp __retVal.x, r; + __asm float_rcp __retVal, r; } vec2 sqrt(const vec2 v) @@ -667,13 +667,13 @@ vec4 inversesqrt(const vec4 v) float normalize(const float x) { - __retVal.x = 1.0; + __retVal = 1.0; } vec2 normalize(const vec2 v) { const float s = inversesqrt(dot(v, v)); - __asm vec4_multiply __retVal.xy, v, s.xx; + __asm vec4_multiply __retVal.xy, v, s; } vec3 normalize(const vec3 v) @@ -686,7 +686,7 @@ vec3 normalize(const vec3 v) float tmp; __asm vec3_dot tmp, v, v; __asm float_rsq tmp, tmp; - __asm vec4_multiply __retVal.xyz, v, tmp.xxx; + __asm vec4_multiply __retVal.xyz, v, tmp; } vec4 normalize(const vec4 v) @@ -694,7 +694,7 @@ vec4 normalize(const vec4 v) float tmp; __asm vec4_dot tmp, v, v; __asm float_rsq tmp, tmp; - __asm vec4_multiply __retVal.xyz, v, tmp.xxx; + __asm vec4_multiply __retVal.xyz, v, tmp; } @@ -708,7 +708,7 @@ vec4 normalize(const vec4 v) float abs(const float a) { - __asm vec4_abs __retVal.x, a; + __asm vec4_abs __retVal, a; } vec2 abs(const vec2 a) @@ -732,9 +732,9 @@ vec4 abs(const vec4 a) float sign(const float x) { float p, n; - __asm vec4_sgt p.x, x, 0.0; // p = (x > 0) - __asm vec4_sgt n.x, 0.0, x; // n = (x < 0) - __asm vec4_subtract __retVal.x, p, n; // sign = p - n + __asm vec4_sgt p, x, 0.0; // p = (x > 0) + __asm vec4_sgt n, 0.0, x; // n = (x < 0) + __asm vec4_subtract __retVal, p, n; // sign = p - n } vec2 sign(const vec2 v) @@ -766,7 +766,7 @@ vec4 sign(const vec4 v) float floor(const float a) { - __asm vec4_floor __retVal.x, a; + __asm vec4_floor __retVal, a; } vec2 floor(const vec2 a) @@ -792,7 +792,7 @@ float ceil(const float a) // XXX this could be improved float b = -a; __asm vec4_floor b, b; - __retVal.x = -b; + __retVal = -b; } vec2 ceil(const vec2 a) @@ -821,7 +821,7 @@ vec4 ceil(const vec4 a) float fract(const float a) { - __asm vec4_frac __retVal.x, a; + __asm vec4_frac __retVal, a; } vec2 fract(const vec2 a) @@ -846,7 +846,7 @@ float mod(const float a, const float b) { float oneOverB; __asm float_rcp oneOverB, b; - __retVal.x = a - b * floor(a * oneOverB); + __retVal = a - b * floor(a * oneOverB); } vec2 mod(const vec2 a, const float b) @@ -908,7 +908,7 @@ vec4 mod(const vec4 a, const vec4 b) float min(const float a, const float b) { - __asm vec4_min __retVal.x, a.x, b.x; + __asm vec4_min __retVal, a, b; } vec2 min(const vec2 a, const vec2 b) @@ -928,17 +928,17 @@ vec4 min(const vec4 a, const vec4 b) vec2 min(const vec2 a, const float b) { - __asm vec4_min __retVal, a.xy, b.xx; + __asm vec4_min __retVal, a.xy, b; } vec3 min(const vec3 a, const float b) { - __asm vec4_min __retVal, a.xyz, b.xxx; + __asm vec4_min __retVal, a.xyz, b; } vec4 min(const vec4 a, const float b) { - __asm vec4_min __retVal, a, b.xxxx; + __asm vec4_min __retVal, a, b; } @@ -946,7 +946,7 @@ vec4 min(const vec4 a, const float b) float max(const float a, const float b) { - __asm vec4_max __retVal.x, a.x, b.x; + __asm vec4_max __retVal, a, b; } vec2 max(const vec2 a, const vec2 b) @@ -966,17 +966,17 @@ vec4 max(const vec4 a, const vec4 b) vec2 max(const vec2 a, const float b) { - __asm vec4_max __retVal, a.xy, b.xx; + __asm vec4_max __retVal, a.xy, b; } vec3 max(const vec3 a, const float b) { - __asm vec4_max __retVal, a.xyz, b.xxx; + __asm vec4_max __retVal, a.xyz, b; } vec4 max(const vec4 a, const float b) { - __asm vec4_max __retVal, a, b.xxxx; + __asm vec4_max __retVal, a, b; } @@ -1165,7 +1165,7 @@ float length(const vec3 v) float r; const float p = dot(v, v); // p = v.x * v.x + v.y * v.y + v.z * v.z __asm float_rsq r, p; // r = 1 / sqrt(p) - __asm float_rcp __retVal.x, r; // retVal = 1 / r + __asm float_rcp __retVal, r; // retVal = 1 / r } float length(const vec4 v) @@ -1173,7 +1173,7 @@ float length(const vec4 v) float r; const float p = dot(v, v); // p = v.x * v.x + v.y * v.y + ... __asm float_rsq r, p; // r = 1 / sqrt(p) - __asm float_rcp __retVal.x, r; // retVal = 1 / r + __asm float_rcp __retVal, r; // retVal = 1 / r } @@ -1219,7 +1219,7 @@ float faceforward(const float N, const float I, const float Nref) // this could probably be done better const float d = dot(Nref, I); float s; - __asm vec4_sgt s.x, 0.0, d; // s = (0.0 > d) ? 1 : 0 + __asm vec4_sgt s, 0.0, d; // s = (0.0 > d) ? 1 : 0 return mix(-N, N, s); } @@ -1228,7 +1228,7 @@ vec2 faceforward(const vec2 N, const vec2 I, const vec2 Nref) // this could probably be done better const float d = dot(Nref, I); float s; - __asm vec4_sgt s.x, 0.0, d; // s = (0.0 > d) ? 1 : 0 + __asm vec4_sgt s, 0.0, d; // s = (0.0 > d) ? 1 : 0 return mix(-N, N, s); } @@ -1237,7 +1237,7 @@ vec3 faceforward(const vec3 N, const vec3 I, const vec3 Nref) // this could probably be done better const float d = dot(Nref, I); float s; - __asm vec4_sgt s.x, 0.0, d; // s = (0.0 > d) ? 1 : 0 + __asm vec4_sgt s, 0.0, d; // s = (0.0 > d) ? 1 : 0 return mix(-N, N, s); } @@ -1246,7 +1246,7 @@ vec4 faceforward(const vec4 N, const vec4 I, const vec4 Nref) // this could probably be done better const float d = dot(Nref, I); float s; - __asm vec4_sgt s.x, 0.0, d; // s = (0.0 > d) ? 1 : 0 + __asm vec4_sgt s, 0.0, d; // s = (0.0 > d) ? 1 : 0 return mix(-N, N, s); } @@ -1596,26 +1596,25 @@ bool any(const bvec4 v) bool all (const bvec2 v) { float prod; - __asm vec4_multiply prod.x, v.x, v.y; - __asm vec4_sne __retVal.x, prod.x, 0.0; - return v.x && v.y; + __asm vec4_multiply prod, v.x, v.y; + __asm vec4_sne __retVal, prod, 0.0; } bool all (const bvec3 v) { float prod; - __asm vec4_multiply prod.x, v.x, v.y; - __asm vec4_multiply prod.x, prod.x, v.z; - __asm vec4_sne __retVal.x, prod.x, 0.0; + __asm vec4_multiply prod, v.x, v.y; + __asm vec4_multiply prod, prod, v.z; + __asm vec4_sne __retVal, prod, 0.0; } bool all (const bvec4 v) { float prod; - __asm vec4_multiply prod.x, v.x, v.y; - __asm vec4_multiply prod.x, prod.x, v.z; - __asm vec4_multiply prod.x, prod.x, v.w; - __asm vec4_sne __retVal.x, prod.x, 0.0; + __asm vec4_multiply prod, v.x, v.y; + __asm vec4_multiply prod, prod, v.z; + __asm vec4_multiply prod, prod, v.w; + __asm vec4_sne __retVal, prod, 0.0; } -- cgit v1.2.3 From 096b5fa52483eaf75fc56bee501aa757dccd8867 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 16:59:22 -0600 Subject: mesa: glsl: regenerated file --- .../shader/slang/library/slang_common_builtin_gc.h | 1110 ++++++++++---------- 1 file changed, 552 insertions(+), 558 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index b7c7f4032a..9301a57339 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -103,303 +103,301 @@ 108,95,70,111,103,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95,70,111,103,0,0,0,1,0,9,0,114, 97,100,105,97,110,115,0,1,1,0,9,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54, 0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,100,101,103,0,0,18,99,0,0,0,0,1,0,10,0,114,97,100,105,97,110,115, -0,1,1,0,10,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48, +101,116,86,97,108,0,0,18,100,101,103,0,0,18,99,0,0,0,0,1,0,10,0,114,97,100,105,97,110,115,0,1,1,0, +10,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0, +0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +0,0,18,100,101,103,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,0,11,0,114,97,100,105,97,110,115, +0,1,1,0,11,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48, 0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,0,0,18,100,101,103,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,0,11,0,114,97,100,105, -97,110,115,0,1,1,0,11,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49, -56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,100,101,103,0,59,120,121,122,0,0,18,99,0,59,120,120,120,0,0,0,0,1,0, -12,0,114,97,100,105,97,110,115,0,1,1,0,12,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53, -57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -95,95,114,101,116,86,97,108,0,0,18,100,101,103,0,0,18,99,0,59,120,120,120,120,0,0,0,0,1,0,9,0,100, -101,103,114,101,101,115,0,1,1,0,9,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0, -49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,114,97,100,0,0,18,99,0,0,0,0,1,0,10,0,100,101,103,114,101,101, -115,0,1,1,0,10,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50, -54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,114,97,100,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,0,11,0,100,101,103, -114,101,101,115,0,1,1,0,11,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52, +59,120,121,122,0,0,18,100,101,103,0,59,120,121,122,0,0,18,99,0,59,120,120,120,0,0,0,0,1,0,12,0,114, +97,100,105,97,110,115,0,1,1,0,12,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54, +0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, +101,116,86,97,108,0,0,18,100,101,103,0,0,18,99,0,59,120,120,120,120,0,0,0,0,1,0,9,0,100,101,103, +114,101,101,115,0,1,1,0,9,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52, 49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,0,18,114,97,100,0,59,120,121,122,0,0,18,99,0,59,120,120,120,0,0,0, -0,1,0,12,0,100,101,103,114,101,101,115,0,1,1,0,12,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48, -0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,59,120,120,120,120,0,0,0,0,1,0, -9,0,115,105,110,0,1,1,0,9,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,0,10,0, -115,105,110,0,1,1,0,10,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108, -111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97, -110,115,0,59,121,0,0,0,0,1,0,11,0,115,105,110,0,1,1,0,11,114,97,100,105,97,110,115,0,0,0,1,4,102, -108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105, -97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108, -0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,0,12,0, -115,105,110,0,1,1,0,12,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0, +116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,0,0,0,1,0,10,0,100,101,103,114,101,101,115,0,1,1,0,10, +114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4, +118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,114,97,100,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,0,11,0,100,101,103,114,101,101,115,0,1, +1,0,11,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0, +49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,114,97,100,0,59,120,121,122,0,0,18,99,0,59,120,120,120,0,0,0,0,1,0,12,0,100,101, +103,114,101,101,115,0,1,1,0,12,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49, +52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, +116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,59,120,120,120,120,0,0,0,0,1,0,9,0,115,105,110,0,1,1,0, +9,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116, +86,97,108,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,0,10,0,115,105,110,0,1,1,0,10,114,97,100,105, +97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18, +95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,0,1,0,11,0, +115,105,110,0,1,1,0,11,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0, 18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108, 111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97, 110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0, -59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,110,115,0,59,119,0,0,0,0,1,0,9,0, -99,111,115,0,1,1,0,9,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,0,10,0,99, -111,115,0,1,1,0,10,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110, +59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,0,12,0,115,105,110,0,1,1,0,12,114,97, +100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97, +108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110, +101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4, +102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100, +105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97, +108,0,59,119,0,0,18,114,97,100,105,97,110,115,0,59,119,0,0,0,0,1,0,9,0,99,111,115,0,1,1,0,9,114,97, +100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116, +86,97,108,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,0,10,0,99,111,115,0,1,1,0,10,114,97,100,105, +97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108, +0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111,115,105, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0, +0,1,0,11,0,99,111,115,0,1,1,0,11,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111, +115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59, +120,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121, +0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0, +18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,0,12,0, +99,111,115,0,1,1,0,12,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110, 101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4, 102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114, -97,100,105,97,110,115,0,59,121,0,0,0,0,1,0,11,0,99,111,115,0,1,1,0,11,114,97,100,105,97,110,115,0, -0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, -18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95, -95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97, -116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97, -110,115,0,59,122,0,0,0,0,1,0,12,0,99,111,115,0,1,1,0,12,114,97,100,105,97,110,115,0,0,0,1,4,102, -108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97, -100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114, -101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95, -99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115, -0,59,122,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0, -59,119,0,0,18,114,97,100,105,97,110,115,0,59,119,0,0,0,0,1,0,9,0,116,97,110,0,1,1,0,9,97,110,103, -108,101,0,0,0,1,3,2,1,9,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,9,1,99,0, -2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,10,0,116,97,110,0, -1,1,0,10,97,110,103,108,101,0,0,0,1,3,2,1,10,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0, -0,0,0,3,2,1,10,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0, -1,0,11,0,116,97,110,0,1,1,0,11,97,110,103,108,101,0,0,0,1,3,2,1,11,1,115,0,2,58,115,105,110,0,18, -97,110,103,108,101,0,0,0,0,0,3,2,1,11,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8, -18,115,0,18,99,0,49,0,0,1,0,12,0,116,97,110,0,1,1,0,12,97,110,103,108,101,0,0,0,1,3,2,1,12,1,115,0, -2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,12,1,99,0,2,58,99,111,115,0,18,97,110,103, -108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,9,0,97,115,105,110,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1, -97,48,0,2,17,49,0,53,55,48,55,50,56,56,0,0,0,0,3,2,1,9,1,97,49,0,2,17,48,0,50,49,50,49,49,52,52,0, -0,54,0,0,3,2,1,9,1,97,50,0,2,17,48,0,48,55,52,50,54,49,48,0,0,0,0,3,2,1,9,1,104,97,108,102,80,105, -0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,48,0,53,0,0,48,0,0,3,2,1,9,1,121,0,2,58,97,98,115,0,18,120, -0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,104,97,108,102,80,105,0,58,115,113,114,116,0,17,49, -0,48,0,0,18,121,0,47,0,0,18,97,48,0,18,121,0,18,97,49,0,18,97,50,0,18,121,0,48,46,48,46,48,47,58, -115,105,103,110,0,18,120,0,0,0,48,20,0,0,1,0,10,0,97,115,105,110,0,1,1,0,10,118,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,0,59,121,0,0,0,20,0,0,1,0,11,0,97,115,105,110, -0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,18,118,0,59, -120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,0,59,121,0,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,115,105,110,0,18,118,0,59,122,0,0,0,20,0, -0,1,0,12,0,97,115,105,110,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97, +97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114, +101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95, +99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,110,115, +0,59,119,0,0,0,0,1,0,9,0,116,97,110,0,1,1,0,9,97,110,103,108,101,0,0,0,1,3,2,1,9,1,115,0,2,58,115, +105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,9,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0, +0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,10,0,116,97,110,0,1,1,0,10,97,110,103,108,101,0,0,0,1,3,2,1, +10,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,10,1,99,0,2,58,99,111,115,0,18, +97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,11,0,116,97,110,0,1,1,0,11,97,110,103, +108,101,0,0,0,1,3,2,1,11,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,11,1,99, +0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,12,0,116,97,110, +0,1,1,0,12,97,110,103,108,101,0,0,0,1,3,2,1,12,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0, +0,0,0,0,3,2,1,12,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0, +0,1,0,9,0,97,115,105,110,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,97,48,0,2,17,49,0,53,55,48,55,50,56,56,0, +0,0,0,3,2,1,9,1,97,49,0,2,17,48,0,50,49,50,49,49,52,52,0,0,54,0,0,3,2,1,9,1,97,50,0,2,17,48,0,48, +55,52,50,54,49,48,0,0,0,0,3,2,1,9,1,104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17, +48,0,53,0,0,48,0,0,3,2,1,9,1,121,0,2,58,97,98,115,0,18,120,0,0,0,0,0,9,18,95,95,114,101,116,86,97, +108,0,18,104,97,108,102,80,105,0,58,115,113,114,116,0,17,49,0,48,0,0,18,121,0,47,0,0,18,97,48,0,18, +121,0,18,97,49,0,18,97,50,0,18,121,0,48,46,48,46,48,47,58,115,105,103,110,0,18,120,0,0,0,48,20,0,0, +1,0,10,0,97,115,105,110,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97, 115,105,110,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105, -110,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,115,105,110,0, -18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,115,105,110,0,18,118, -0,59,119,0,0,0,20,0,0,1,0,9,0,97,99,111,115,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,104,97,108,102,80,105, -0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,48,0,53,0,0,48,0,0,9,18,95,95,114,101,116,86,97,108,0,18, -104,97,108,102,80,105,0,58,97,115,105,110,0,18,120,0,0,0,47,20,0,0,1,0,10,0,97,99,111,115,0,1,1,0, -10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,0, -1,0,11,0,97,99,111,115,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99, -111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115, -0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115,0,18,118, -0,59,122,0,0,0,20,0,0,1,0,12,0,97,99,111,115,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58, -97,99,111,115,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,99, -111,115,0,18,118,0,59,119,0,0,0,20,0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,58,97,115,105,110,0,18,120,0,58,105,110,118,101,114,115,101,115,113,114,116,0, -18,120,0,18,120,0,48,17,49,0,48,0,0,46,0,0,48,0,0,20,0,0,1,0,10,0,97,116,97,110,0,1,1,0,10,121,95, -111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18, -121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, -97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97,110,0,1, -1,0,11,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116, +110,0,18,118,0,59,121,0,0,0,20,0,0,1,0,11,0,97,115,105,110,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,122,0,58,97,115,105,110,0,18,118,0,59,122,0,0,0,20,0,0,1,0,12,0,97,115,105,110,0,1,1,0,12, +118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,18,118,0,59,120,0,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,0,59,121,0,0,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,115,105,110,0,18,118,0,59,122,0,0,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,59,119,0,58,97,115,105,110,0,18,118,0,59,119,0,0,0,20,0,0,1,0,9,0,97,99, +111,115,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0, +17,48,0,53,0,0,48,0,0,9,18,95,95,114,101,116,86,97,108,0,18,104,97,108,102,80,105,0,58,97,115,105, +110,0,18,120,0,0,0,47,20,0,0,1,0,10,0,97,99,111,115,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,0,1,0,11,0,97,99,111,115,0,1,1,0,11, +118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115,0,18,118,0,59,122,0,0,0,20,0,0,1,0,12,0, +97,99,111,115,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115, +0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118, +0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115,0,18,118,0,59,122, +0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,99,111,115,0,18,118,0,59,119,0,0,0,20, +0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,97,115,105, +110,0,18,120,0,58,105,110,118,101,114,115,101,115,113,114,116,0,18,120,0,18,120,0,48,17,49,0,48,0, +0,46,0,0,48,0,0,20,0,0,1,0,10,0,97,116,97,110,0,1,1,0,10,121,95,111,118,101,114,95,120,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59, +120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,121,95,111,118, +101,114,95,120,0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97,110,0,1,1,0,11,121,95,111,118,101,114,95, +120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101, +114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18, +121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58, +97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,122,0,0,0,20,0,0,1,0,12,0,97,116,97,110,0,1, +1,0,12,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116, 97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, 59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114, 101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,122,0,0,0,20, -0,0,1,0,12,0,97,116,97,110,0,1,1,0,12,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59, -121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111,118, -101,114,95,120,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0, -18,121,95,111,118,101,114,95,120,0,59,119,0,0,0,20,0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,121,0,0,1,1, -0,9,120,0,0,0,1,3,2,0,9,1,114,0,0,0,10,58,97,98,115,0,18,120,0,0,0,17,49,0,48,0,45,52,0,41,0,2,9, -18,114,0,58,97,116,97,110,0,18,121,0,18,120,0,49,0,0,20,0,10,18,120,0,17,48,0,48,0,0,40,0,2,9,18, -114,0,18,114,0,58,115,105,103,110,0,18,121,0,0,0,17,51,0,49,52,49,53,57,51,0,0,48,46,20,0,0,9,14,0, -0,2,9,18,114,0,58,115,105,103,110,0,18,121,0,0,0,17,49,0,53,55,48,55,57,54,53,0,0,48,20,0,0,8,18, -114,0,0,0,1,0,10,0,97,116,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,0,1, -0,11,0,97,116,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,0,1, -0,12,0,97,116,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,18,117,0,59,119,0,0,18,118,0,59,119,0, -0,0,20,0,0,1,0,9,0,112,111,119,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,102,108,111,97,116,95,112,111, -119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,112,111, -119,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95, -114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95, -112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59, -121,0,0,0,0,1,0,11,0,112,111,119,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,102,108,111,97,116,95,112, -111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0, -0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18, -97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114, -101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,0,1,0,12,0,112,111,119,0,1, -1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101, -116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111, -119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0, -4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0, -59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116, -86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,18,98,0,59,119,0,0,0,0,1,0,9,0,101,120,112,0,1,1,0,9,97, -0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101, -114,0,18,95,95,114,101,116,86,97,108,0,0,18,101,0,0,18,97,0,0,0,0,1,0,10,0,101,120,112,0,1,1,0,10, -97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119, -101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4,102,108, -111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0,18,97,0, -59,121,0,0,0,0,1,0,11,0,101,120,112,0,1,1,0,11,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56, -0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, -18,101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116, -86,97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114, -0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18,97,0,59,122,0,0,0,0,1,0,12,0,101,120, -112,0,1,1,0,12,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95, -112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0, -4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0, -0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97, -108,0,59,122,0,0,18,101,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18, -95,95,114,101,116,86,97,108,0,59,119,0,0,18,101,0,0,18,97,0,59,119,0,0,0,0,1,0,9,0,108,111,103,50, -0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,120,0,0,0,0,1,0,10,0,108,111,103,50,0,1,1,0,10,118,0,0,0,1,4,102,108,111,97,116,95,108, -111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97, -116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,0,11, -0,108,111,103,50,0,1,1,0,11,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101, -116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95, -114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0, -18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,0,12,0,108,111,103,50,0,1,1, -0,12,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120, -0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108, -0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116, -86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114, -101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,9,0,108,111,103,0,1,1,0,9,120,0,0,0,1,3, -2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,120,0,0,0,18,99, -0,48,0,0,1,0,10,0,108,111,103,0,1,1,0,10,118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49, -56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,11,0,108,111,103,0,1,1,0,11, -118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18, -118,0,0,0,18,99,0,48,0,0,1,0,12,0,108,111,103,0,1,1,0,12,118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54, -57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,9,0,101,120, -112,50,0,1,1,0,9,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97, -108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,0,101,120,112,50,0,1,1,0,10,97,0,0,0,1,4,102,108,111,97,116, -95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111, -97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,0,1,0, -11,0,101,120,112,50,0,1,1,0,11,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95, -95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50, -0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,0,1,0,12,0,101,120,112,50,0,1,1, -0,12,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0, -0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0, -59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86, -97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101, -116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,0,0,1,0,9,0,115,113,114,116,0,1,1,0,9,120,0,0,0,1,3, -2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,120,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,0,1,0,10,0,115,113, -114,116,0,1,1,0,10,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0, -0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,0,1,0,11,0, -115,113,114,116,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0, -18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97, -108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,4, -102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,0,1,0,12,0,115,113,114,116,0,1,1, -0,12,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59, -120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114, -0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95, -114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114, -101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18, -118,0,59,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,119,0, -0,18,114,0,0,0,0,1,0,9,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,9,120,0,0,0,1,4,102, -108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,0,0,1,0,10, -0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,10,118,0,0,0,1,4,102,108,111,97,116,95,114, -115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116, -95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,0,11,0,105, -110,118,101,114,115,101,115,113,114,116,0,1,1,0,11,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113, -0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114, -115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116, -95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,0,12,0,105, -110,118,101,114,115,101,115,113,114,116,0,1,1,0,12,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113, -0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114, -115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116, -95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111, -97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,9, -0,110,111,114,109,97,108,105,122,101,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,17,49,0,48,0,0,20,0,0,1,0,10,0,110,111,114,109,97,108,105,122,101,0,1,1,0,10,118,0,0,0,1,3,2, -1,9,1,115,0,2,58,105,110,118,101,114,115,101,115,113,114,116,0,58,100,111,116,0,18,118,0,0,18,118, -0,0,0,0,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,118,0,0,18,115,0,59,120,120,0,0,0,0,1,0,11,0,110,111,114,109,97,108,105,122, -101,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118,101,99,51,95,100,111,116,0,18,116,109, -112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0,18,116, -109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,59,120,120,120,0,0,0,0,1,0,12,0,110,111,114,109, -97,108,105,122,101,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118,101,99,52,95,100,111, -116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109, -112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,59,120,120,120,0,0,0,0,1,0,9,0,97, -98,115,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120, -0,0,18,97,0,0,0,0,1,0,10,0,97,98,115,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,97,98,115,0,1,1,0,11,97,0,0,0,1,4, -118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0, -12,0,97,98,115,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108, -0,0,18,97,0,0,0,0,1,0,9,0,115,105,103,110,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,112,0,0,1,1,110,0,0,0,4, -118,101,99,52,95,115,103,116,0,18,112,0,59,120,0,0,18,120,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52, -95,115,103,116,0,18,110,0,59,120,0,0,17,48,0,48,0,0,0,18,120,0,0,0,4,118,101,99,52,95,115,117,98, -116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,0,0,18,110,0,0,0,0,1,0,10,0, -115,105,103,110,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103, -116,0,18,112,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18, -110,0,59,120,121,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,0,18,110,0,0,0,0,1,0,11,0,115,105,103, -110,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18, -112,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0, -59,120,121,122,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,0,18,110,0,0,0,0,1,0,12,0,115,105,103, -110,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18, -112,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0, -18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0, -18,112,0,0,18,110,0,0,0,0,1,0,9,0,102,108,111,111,114,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,102, -108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,0,102,108,111, -111,114,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,108,111,111,114,0,1,1,0,11,97,0,0,0,1,4,118,101,99, -52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0, -12,0,102,108,111,111,114,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95, -114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,99,101,105,108,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,98,0, -2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101, -116,86,97,108,0,59,120,0,18,98,0,54,20,0,0,1,0,10,0,99,101,105,108,0,1,1,0,10,97,0,0,0,1,3,2,0,10, -1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,18,98,0,54,20,0,0,1,0,11,0,99,101,105,108,0,1,1,0,11,97,0,0,0, -1,3,2,0,11,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0, -9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,98,0,54,20,0,0,1,0,12,0,99,101,105,108,0,1, -1,0,12,97,0,0,0,1,3,2,0,12,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98, -0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,9,0,102,114,97,99,116,0,1, -1,0,9,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -97,0,0,0,0,1,0,10,0,102,114,97,99,116,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,114,97,99,116,0,1,1,0,11, -97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, -18,97,0,0,0,0,1,0,12,0,102,114,97,99,116,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0, -18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,109,111,100,0,1,1,0,9,97,0,0,1,1,0,9,98,0, -0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, -110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,18, -98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0, -10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110, -101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0, -0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110, -101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,97,0, -18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1, -0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0, -0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95, -95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118, -101,114,66,0,48,0,0,48,47,20,0,0,1,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,3,2,0, -9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108, -111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,9,18,95, -95,114,101,116,86,97,108,0,59,120,0,18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18, -97,0,59,120,0,18,111,110,101,79,118,101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18, -111,110,101,79,118,101,114,66,121,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1, -0,11,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66, -121,0,0,1,1,111,110,101,79,118,101,114,66,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, +0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120, +0,59,119,0,0,0,20,0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,121,0,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,114,0, +0,0,10,58,97,98,115,0,18,120,0,0,0,17,49,0,48,0,45,52,0,41,0,2,9,18,114,0,58,97,116,97,110,0,18, +121,0,18,120,0,49,0,0,20,0,10,18,120,0,17,48,0,48,0,0,40,0,2,9,18,114,0,18,114,0,58,115,105,103, +110,0,18,121,0,0,0,17,51,0,49,52,49,53,57,51,0,0,48,46,20,0,0,9,14,0,0,2,9,18,114,0,58,115,105,103, +110,0,18,121,0,0,0,17,49,0,53,55,48,55,57,54,53,0,0,48,20,0,0,8,18,114,0,0,0,1,0,10,0,97,116,97, +110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116, +97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121, +0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97,110,0,1,1, +0,11,117,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0, +18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97, +116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, +122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,0,1,0,12,0,97,116,97,110,0, +1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110, +0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97, +116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, +122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,119,0,58,97,116,97,110,0,18,117,0,59,119,0,0,18,118,0,59,119,0,0,0,20,0,0,1,0,9,0,112, +111,119,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95, +114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,112,111,119,0,1,1,0,10,97,0,0,1,1,0,10, +98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0, +0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95, +114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,0,1,0,11,0,112,111,119, +0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114, +101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112, +111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0, +0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18, +97,0,59,122,0,0,18,98,0,59,122,0,0,0,0,1,0,12,0,112,111,119,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1, +4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0, +59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116, +86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119, +101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,4, +102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59, +119,0,0,18,98,0,59,119,0,0,0,0,1,0,9,0,101,120,112,0,1,1,0,9,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0, +55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97, +108,0,0,18,101,0,0,18,97,0,0,0,0,1,0,10,0,101,120,112,0,1,1,0,10,97,0,0,0,1,3,2,1,9,1,101,0,2,17, +50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86, +97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0, +18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,0,1,0,11,0,101,120,112, +0,1,1,0,11,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112, +111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4, +102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0, +18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108, +0,59,122,0,0,18,101,0,0,18,97,0,59,122,0,0,0,0,1,0,12,0,101,120,112,0,1,1,0,12,97,0,0,0,1,3,2,1,9, +1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95, +114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111, +119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,4,102, +108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18, +97,0,59,122,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0, +59,119,0,0,18,101,0,0,18,97,0,59,119,0,0,0,0,1,0,9,0,108,111,103,50,0,1,1,0,9,120,0,0,0,1,4,102, +108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,10,0,108, +111,103,50,0,1,1,0,10,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86, +97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114, +101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,0,11,0,108,111,103,50,0,1,1,0,11,118,0,0, +0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0, +59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0, +18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59, +122,0,0,18,118,0,59,122,0,0,0,0,1,0,12,0,108,111,103,50,0,1,1,0,12,118,0,0,0,1,4,102,108,111,97, +116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102, +108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0, +0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59, +122,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18, +118,0,59,119,0,0,0,0,1,0,9,0,108,111,103,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51, +49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,120,0,0,0,18,99,0,48,0,0,1,0,10,0,108,111,103,0, +1,1,0,10,118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103, +50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,11,0,108,111,103,0,1,1,0,11,118,0,0,0,1,3,2,1,9,1,99,0,2,17, +48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,12,0, +108,111,103,0,1,1,0,12,118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8, +58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,9,0,101,120,112,50,0,1,1,0,9,97,0,0,0,1,4,102, +108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,10,0,101, +120,112,50,0,1,1,0,10,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86, +97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101, +116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,0,1,0,11,0,101,120,112,50,0,1,1,0,11,97,0,0,0,1,4, +102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120, +0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0, +59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0, +18,97,0,59,122,0,0,0,0,1,0,12,0,101,120,112,50,0,1,1,0,12,97,0,0,0,1,4,102,108,111,97,116,95,101, +120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116, +95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111, +97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,4,102, +108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0, +0,0,1,0,9,0,115,113,114,116,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114, +115,113,0,18,114,0,0,18,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97, +108,0,0,18,114,0,0,0,0,1,0,10,0,115,113,114,116,0,1,1,0,10,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102, +108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113, +0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97, +108,0,59,121,0,0,18,114,0,0,0,0,1,0,11,0,115,113,114,116,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,114,0,0, +0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95, +114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114, +115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101, +116,86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0, +59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18, +114,0,0,0,0,1,0,12,0,115,113,114,116,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97, +116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95, +95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0, +0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, +121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,4,102,108, +111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,0,0,0,0,1,0,9,0,105,110,118,101,114,115,101,115, +113,114,116,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97, +108,0,59,120,0,0,18,120,0,0,0,0,1,0,10,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,10, +118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, +118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0, +0,18,118,0,59,121,0,0,0,0,1,0,11,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,11,118,0,0, +0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59, +120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118, +0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0, +18,118,0,59,122,0,0,0,0,1,0,12,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,12,118,0,0,0, +1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59, +120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118, +0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0, +18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59, +119,0,0,18,118,0,59,119,0,0,0,0,1,0,9,0,110,111,114,109,97,108,105,122,101,0,1,1,0,9,120,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,17,49,0,48,0,0,20,0,0,1,0,10,0,110,111,114,109,97,108,105,122,101, +0,1,1,0,10,118,0,0,0,1,3,2,1,9,1,115,0,2,58,105,110,118,101,114,115,101,115,113,114,116,0,58,100, +111,116,0,18,118,0,0,18,118,0,0,0,0,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,115,0,0,0,0,1,0,11,0,110,111,114,109,97, +108,105,122,101,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118,101,99,51,95,100,111,116, +0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112, +0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,0,12,0,110,111,114,109,97,108, +105,122,101,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118,101,99,52,95,100,111,116,0,18, +116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0, +18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,0,9,0,97,98,115,0,1,1,0,9,97,0,0,0, +1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,10,0,97,98, +115,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,97,0,0,0,0,1,0,11,0,97,98,115,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,12,0,97,98,115,0,1,1,0,12,97,0, +0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,115, +105,103,110,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0, +18,112,0,0,18,120,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0, +0,0,18,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108, +0,0,18,112,0,0,18,110,0,0,0,0,1,0,10,0,115,105,103,110,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,112,0,0,1, +1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0, +4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,121,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101, +99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,0, +18,110,0,0,0,0,1,0,11,0,115,105,103,110,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,112,0,0,1,1,110,0,0,0,4, +118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101, +99,52,95,115,103,116,0,18,110,0,59,120,121,122,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52, +95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,0, +18,110,0,0,0,0,1,0,12,0,115,105,103,110,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,112,0,0,1,1,110,0,0,0,4, +118,101,99,52,95,115,103,116,0,18,112,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103, +116,0,18,110,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, +18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,0,9,0,102,108,111,111,114,0,1,1,0,9, +97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0, +0,1,0,10,0,102,108,111,111,114,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,108,111,111,114,0,1,1,0,11, +97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,0,18,97,0,0,0,0,1,0,12,0,102,108,111,111,114,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102, +108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,99,101,105,108,0,1,1,0, +9,97,0,0,0,1,3,2,0,9,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18, +98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,10,0,99,101,105,108,0,1,1,0,10, +97,0,0,0,1,3,2,0,10,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18, +98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,54,20,0,0,1,0,11,0,99,101,105,108, +0,1,1,0,11,97,0,0,0,1,3,2,0,11,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18, +98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,98,0,54,20,0,0,1,0,12,0, +99,101,105,108,0,1,1,0,12,97,0,0,0,1,3,2,0,12,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108, +111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,9,0, +102,114,97,99,116,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86, +97,108,0,0,18,97,0,0,0,0,1,0,10,0,102,114,97,99,116,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102, +114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,114,97,99, +116,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,97,0,0,0,0,1,0,12,0,102,114,97,99,116,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95, +102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,109,111,100,0,1,1,0,9,97, +0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114, +99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97, +0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0, +1,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66, +0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111, +110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,9,98, +0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, +110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18, +97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0, +0,1,0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114, +66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9, +18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101, +79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1, +3,2,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0, +18,97,0,59,120,0,18,111,110,101,79,118,101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116, +86,97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0, +18,111,110,101,79,118,101,114,66,121,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0,1,1,0,11,97,0,0, +1,1,0,11,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114, +66,121,0,0,1,1,111,110,101,79,118,101,114,66,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, 110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, 110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, 110,101,79,118,101,114,66,122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0, @@ -422,285 +420,281 @@ 114,0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48,47,20,0,9,18,95,95,114,101, 116,86,97,108,0,59,119,0,18,97,0,59,119,0,18,98,0,59,119,0,58,102,108,111,111,114,0,18,97,0,59,119, 0,18,111,110,101,79,118,101,114,66,119,0,48,0,0,48,47,20,0,0,1,0,9,0,109,105,110,0,1,1,0,9,97,0,0, -1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -97,0,59,120,0,0,18,98,0,59,120,0,0,0,0,1,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1, -4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120, -121,0,0,18,98,0,59,120,121,0,0,0,0,1,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4, -118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120, -121,122,0,0,18,98,0,59,120,121,122,0,0,0,0,1,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0, -0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1, -0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95, -114,101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,11,0,109,105,110,0, -1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108, -0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,120,120,0,0,0,0,1,0,12,0,109,105,110,0,1,1,0,12,97,0, -0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, -18,98,0,59,120,120,120,120,0,0,0,0,1,0,9,0,109,97,120,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118, -101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0, -59,120,0,0,0,0,1,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,97, -120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0, -0,0,0,1,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122, -0,0,0,0,1,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0, -18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1, -1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120, -121,0,0,18,98,0,59,120,120,0,0,0,0,1,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118, -101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59, -120,120,120,0,0,0,0,1,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, -109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,59,120,120,120,120,0,0,0,0,1,0,9, -0,99,108,97,109,112,0,1,1,0,9,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120, -86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118, -97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,10,0,99,108,97,109,112, -0,1,1,0,10,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4, -118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109, -105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97, -108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95, -99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108, -0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,9, -109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109, -112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97, -120,86,97,108,0,0,0,0,1,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,10,109,105,110,86, -97,108,0,0,1,1,0,10,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95, -114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0, -0,0,0,1,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,11,109,105,110,86,97,108,0,0,1,1, -0,11,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86, -97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,12,0, -99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,12,109,105,110,86,97,108,0,0,1,1,0,12,109,97,120, -86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118, -97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,9,0,109,105,120,0,1,1, -0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101, -116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1, -0,10,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0, -0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1, -0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18, -121,0,0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,9,97,0,0,0,1, -4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0, -0,0,0,1,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52, -95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0, -109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,108,114,112, -0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1, -1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114, -101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,9,0,115,116,101,112,0,1,1,0,9,101, -100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, -108,0,59,120,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,10,0,115,116,101,112,0,1,1,0,10,101,100, -103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,11,0,115,116,101,112,0,1,1,0,11,101,100, -103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,12,0,115,116,101,112,0,1,1,0,12,101, -100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, -108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,10,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0, -0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,0,0,0,0,1,0,11,0,115,116,101,112,0,1,1,0,9,101, -100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,120,0,0,0,0,1,0,12,0,115,116, -101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95, -95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,120,120,0,0,0,0,1,0,9,0, -115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101, -49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,116,0,2,58,99,108,97,109,112,0,18,120,0,18,101,100,103,101,48, -0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0, -0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,10,0,115,109,111, -111,116,104,115,116,101,112,0,1,1,0,10,101,100,103,101,48,0,0,1,1,0,10,101,100,103,101,49,0,0,1,1, -0,10,118,0,0,0,1,3,2,0,10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18, -101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18, -116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,11,0,115,109,111,111,116, -104,115,116,101,112,0,1,1,0,11,101,100,103,101,48,0,0,1,1,0,11,101,100,103,101,49,0,0,1,1,0,11,118, -0,0,0,1,3,2,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100, -103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18, -116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,12,0,115,109,111,111,116,104,115, -116,101,112,0,1,1,0,12,101,100,103,101,48,0,0,1,1,0,12,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1, -3,2,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101, -49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0, -48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,10,0,115,109,111,111,116,104,115,116, -101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0, -10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18, -101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51, -0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1, -1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,116,0, -2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100, -103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0, -17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9, -101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,116,0,2,58, -99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101, -48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0, -48,0,0,18,116,0,48,47,48,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,9,120,0,0,0,1,8,58,97,98,115, -0,18,120,0,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,10,118,0,0,0,1,3,2,0,9,1,114,0,0,0,3,2, -1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0, -18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,114,0,0,0, -3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113, -0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,114,0,0,0, -3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113, -0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,114,0,0,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1, -3,2,1,9,1,100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103, -116,104,0,18,100,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118,0,0,1,1,0,10, -117,0,0,0,1,3,2,1,10,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, -108,101,110,103,116,104,0,18,100,50,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,11, -118,0,0,1,1,0,11,117,0,0,0,1,3,2,1,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116, -86,97,108,0,58,108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99, -101,0,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,12,1,100,52,0,2,18,118,0,18,117,0,47,0,0,9,18,95, -95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0,11,0,99,114,111, -115,115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115,115,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,9,0,102,97,99,101,102,111, -114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0, -2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95, -115,103,116,0,18,115,0,59,120,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18, -78,0,0,18,115,0,0,0,0,0,1,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10,78,0,0,1,1,0, -10,73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0, -18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,59,120,0,0,17,48,0, -48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,11,0,102,97, -99,101,102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0,1,1,0,11,78,114,101,102,0,0,0, -1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4, -118,101,99,52,95,115,103,116,0,18,115,0,59,120,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120, -0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,12,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1, -0,12,78,0,0,1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18, -78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0, -59,120,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0, -0,1,0,9,0,114,101,102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0, -58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,10,0,114,101,102,108,101,99,116,0, -1,1,0,10,73,0,0,1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0, -0,48,18,78,0,48,47,0,0,1,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,0,1,8, -18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,12,0,114, -101,102,108,101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111, -116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,9,0,114,101,102,114,97,99,116,0,1,1,0,9,73,0, -0,1,1,0,9,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18, -101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0, -18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116, -97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18, -107,0,0,0,46,18,78,0,48,47,0,0,1,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0, -0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48, -17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47, -48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48, -18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18, -78,0,48,47,0,0,1,0,11,0,114,101,102,114,97,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,1,1,0,9,101, +1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, +18,98,0,0,0,0,1,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109, +105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120, +121,0,0,0,0,1,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,105, +110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59, +120,121,122,0,0,0,0,1,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95, +109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,109,105,110,0,1, +1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0, +0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4, +118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98, +0,0,0,0,1,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0, +18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0,109,97,120,0,1,1,0,9,97,0,0,1,1, +0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98, +0,0,0,0,1,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0, +1,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0, +0,1,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95, +95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,9, +98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,0, +0,18,98,0,0,0,0,1,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97, +120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,12,0,109,97, +120,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86, +97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0,99,108,97,109,112,0,1,1,0,9,118,97,108,0,0,1,1,0,9,109, +105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0, +18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86, +97,108,0,0,0,0,1,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108, +0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101, +116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1, +0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109, +97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0, +18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,12,0,99,108,97, +109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0, +0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0, +18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,10,0,99,108,97,109,112,0,1,1,0,10, +118,97,108,0,0,1,1,0,10,109,105,110,86,97,108,0,0,1,1,0,10,109,97,120,86,97,108,0,0,0,1,4,118,101, +99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110, +86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0, +1,1,0,11,109,105,110,86,97,108,0,0,1,1,0,11,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108, +97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18, +109,97,120,86,97,108,0,0,0,0,1,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,12,109,105, +110,86,97,108,0,0,1,1,0,12,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18, +95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97, +108,0,0,0,0,1,0,9,0,109,105,120,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99, +52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,10, +0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114, +112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0,109,105,120, +0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95, +114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1,1,0,12,120, +0,0,1,1,0,12,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86, +97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10, +121,0,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18, +97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11, +97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0, +0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4, +118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0, +0,0,1,0,9,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95, +115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1, +0,10,0,115,116,101,112,0,1,1,0,10,101,100,103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115, +103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1, +0,11,0,115,116,101,112,0,1,1,0,11,101,100,103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115, +103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0, +0,1,0,12,0,115,116,101,112,0,1,1,0,12,101,100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95, +115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,10,0, +115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,0,0,0, +0,1,0,11,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, +115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0, +59,120,120,120,0,0,0,0,1,0,12,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1, +4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101, +0,59,120,120,120,120,0,0,0,0,1,0,9,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103, +101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,116,0,2,58,99,108,97,109, +112,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, +17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, +116,0,48,47,48,0,0,1,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,10,101,100,103,101,48, +0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,116,0,2,58,99,108,97,109,112,0, +18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48, +0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48, +47,48,0,0,1,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,11,101,100,103,101,48,0,0,1,1,0, +11,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0, +18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0, +0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0, +0,1,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,12,101,100,103,101,48,0,0,1,1,0,12,101, +100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101, +100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49, +0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,10, +0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101, +49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101, +48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0, +0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,11,0,115,109, +111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1, +1,0,11,118,0,0,0,1,3,2,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47, +18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8, +18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,12,0,115,109,111,111, +116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,12, +118,0,0,0,1,3,2,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101, +100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116, +0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,9,0,108,101,110,103,116,104, +0,1,1,0,9,120,0,0,0,1,8,58,97,98,115,0,18,120,0,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,10, +118,0,0,0,1,3,2,0,9,1,114,0,0,0,3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4, +102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0, +11,118,0,0,0,1,3,2,0,9,1,114,0,0,0,3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0, +4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0, +0,0,1,3,2,0,9,1,114,0,0,0,3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108, +111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95, +114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1, +1,0,9,121,0,0,0,1,3,2,1,9,1,100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, +108,101,110,103,116,104,0,18,100,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118, +0,0,1,1,0,10,117,0,0,0,1,3,2,1,10,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86, +97,108,0,58,108,101,110,103,116,104,0,18,100,50,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99,101, +0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,1,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95, +114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,9,0,100,105,115, +116,97,110,99,101,0,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,12,1,100,52,0,2,18,118,0,18,117,0, +47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0, +11,0,99,114,111,115,115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115, +115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,9,0,102,97, +99,101,102,111,114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0,1,3, +2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118, +101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54, +0,18,78,0,0,18,115,0,0,0,0,0,1,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10,78,0,0,1, +1,0,10,73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102, +0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0, +0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,11,0,102,97,99,101, +102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0,1,1,0,11,78,114,101,102,0,0,0,1,3,2,1, +9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101, +99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18, +78,0,0,18,115,0,0,0,0,0,1,0,12,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,12,78,0,0,1,1,0, +12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0, +18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0, +18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,9,0,114,101,102,108, +101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0, +18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,10,0,114,101,102,108,101,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78, +0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0, +11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58, +100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,12,0,114,101,102,108,101,99,116,0,1,1, +0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48, +18,78,0,48,47,0,0,1,0,9,0,114,101,102,114,97,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,1,1,0,9,101, 116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0, 58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10, 18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97, 0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0, -1,0,12,0,114,101,102,114,97,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3, +1,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3, 2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0, 18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0, 48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116, -0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,13,0,109,97, -116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97, -116,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49, -0,57,48,0,0,0,0,1,0,14,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1, -0,0,14,110,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0, -16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0, -1,0,15,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0,0, -0,1,8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0, -57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1,1, -0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, -18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0, -0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117, -0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4, -118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2, -0,108,101,115,115,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101, -115,115,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84, -104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101, -116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97, -108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,69,113, -117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97, -110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18, -95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,69, -113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110, -69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84, -104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,101, -0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114, -84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101, -114,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116, -101,114,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84, -104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,3,0,103,114, -101,97,116,101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103, -114,101,97,116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115, -103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116, -101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95, -115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0, -103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1, -4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18, -118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0, -1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0, -0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117, -0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97, -108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97, -110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95, -95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0, -0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0, -1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0, -18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52, -95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117, -97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0, -1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1, -4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0, -2,0,101,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108, -0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1, -1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, -18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4, -0,110,111,116,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117, -97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0, -7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0, -1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, -18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4, -0,110,111,116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,2, -118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0, -0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0, -3,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120, -0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, -120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95, -95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97, -110,121,0,1,1,0,4,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115, -117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18, -115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97, -100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101, -99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17, -48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0, -18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, -18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,8,18,118,0,59,120,0,18,118,0,59,121,0,34,0,0,1, -0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0, -0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114, -111,100,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1, -1,0,4,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, -121,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0, -18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,59, -120,0,0,18,112,114,111,100,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,114,111,100,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0, -2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4, -118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17, +0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,11,0,114,101, +102,114,97,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2, +17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73, +0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17, +48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73, +0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,12,0,114,101,102,114,97,99,116, +0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18, +101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111, +116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9, +14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115, +113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,13,0,109,97,116,114,105,120,67,111,109,112,77, +117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97,116,50,0,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,14,0,109,97, +116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1,0,0,14,110,0,0,0,1,8,58,109,97, +116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49, +0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,15,0,109,97,116,114,105,120, +67,111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0,0,0,1,8,58,109,97,116,52,0,18,109, +0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18, +109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0, +0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99, +52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0, +3,0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115, +108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0, +108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108, +116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104, +97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0, +7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,8,117,0,0, +1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, +18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10, +118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, +117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0, +1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0, +12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117, +0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0, +7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108, +0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97, +108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,10,117, +0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,11,117, +0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,12, +117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0, +6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, +117,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110, +0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97, +110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86, +97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117, +97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104, +97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101, +97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101, +99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103, +114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118, +101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118, +0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, +117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0, +8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118, +101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,3,0,101,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,101, +113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101, +113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, +114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,6,117,0,0,1, +1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118, +101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0, +0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,101, +113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1, +1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108, +0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3, +118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118, +101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0, +110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111, +116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69, +113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95, +114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0, +6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1, +0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0, +0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0, +110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111, +116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, +95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1,3, +2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59, +120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,3,118,0,0,0,1, +3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0, +59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18, +115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, +116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1, +1,0,4,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, +120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0, +59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,100,100,0,18, +115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0, +0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0, +4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0, +48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99, +52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121, +0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111, +100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0, +18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,4,118,0,0,0,1,3,2,0,9,1,112, +114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18, +118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112, +114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,119,0,0,0,4,118,101,99, +52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0, +1,0,2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1, +4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17, 48,0,48,0,0,0,0,0,1,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95, 95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101, 49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52, -- cgit v1.2.3 From ab0d13dd60c707f2c414fa9fe8e489a9e24a8aee Mon Sep 17 00:00:00 2001 From: Jonathan White Date: Tue, 5 Aug 2008 19:22:09 -0600 Subject: egl: changes for Windows --- src/egl/main/egldriver.c | 65 +++++++++++++++++++++++++++++------------------- 1 file changed, 39 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index edf85abe01..f5edb58f43 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -24,37 +24,45 @@ #include #include "eglx.h" #elif defined(_EGL_PLATFORM_WINDOWS) -/* XXX to do */ -#elif defined(_EGL_PLATFORM_WINCE) -/* XXX to do */ +/* Use static linking on Windows for now */ +#define WINDOWS_STATIC_LINK #endif - -static const char *DefaultDriverName = ":0"; -static const char *SysFS = "/sys/class"; - - - /** * Wrappers for dlopen/dlclose() */ #if defined(_EGL_PLATFORM_WINDOWS) - +#ifdef WINDOWS_STATIC_LINK + static const char *DefaultDriverName = "Windows EGL Static Library"; +#else + /* XXX Need to decide how to do dynamic name lookup on Windows */ + static const char *DefaultDriverName = "TBD"; +#endif + static const char *SysFS = NULL; typedef HMODULE lib_handle; static HMODULE open_library(const char *filename) { +#ifdef WINDOWS_STATIC_LINK + return 0; +#else return LoadLibrary(filename); +#endif } static void close_library(HMODULE lib) { +#ifdef WINDOWS_STATIC_LINK +#else FreeLibrary(lib); +#endif } - + #elif defined(_EGL_PLATFORM_X) + static const char *DefaultDriverName = ":0"; + static const char *SysFS = "/sys/class"; typedef void * lib_handle; @@ -72,8 +80,6 @@ static const char *SysFS = "/sys/class"; #endif - - /** * Given a card number, use sysfs to determine the DRI driver name. */ @@ -108,7 +114,6 @@ _eglChooseDRMDriver(int card) #endif } - /** * XXX this function is totally subject change!!! * @@ -130,7 +135,13 @@ _eglChooseDRMDriver(int card) const char * _eglChooseDriver(_EGLDisplay *dpy) { + /* Under Windows, the NativeDisplay is an HDC handle, therefore */ + /* it can't be interpreted as a string or a pointer. */ +#if defined(_EGL_PLATFORM_WINDOWS) + const char *displayString = NULL; +#else const char *displayString = (const char *) dpy->NativeDisplay; +#endif const char *driverName = NULL; (void) DefaultDriverName; @@ -173,15 +184,11 @@ _eglChooseDriver(_EGLDisplay *dpy) } } } - else { + else + { /* NativeDisplay is not a string! */ #if defined(_EGL_PLATFORM_X) driverName = _xeglChooseDriver(dpy); -#elif defined(_EGL_PLATFORM_WINDOWS) - /* XXX to do */ - driverName = _weglChooseDriver(dpy); -#elif defined(_EGL_PLATFORM_WINCE) - /* XXX to do */ #else driverName = DefaultDriverName; #endif @@ -209,25 +216,31 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) assert(driverName); #if defined(_EGL_PLATFORM_WINDOWS) +/* Use static linking on Windows for now */ +#ifdef WINDOWS_STATIC_LINK + mainFunc = (_EGLMain_t)_eglMain; +#else /* XXX untested */ sprintf(driverFilename, "%s.dll", driverName); _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename); + lib = open_library(driverFilename); + if (!lib) { + _eglLog(_EGL_WARNING, "Could not open %s", + driverFilename); + return NULL; + } + mainFunc = (_EGLMain_t) GetProcAddress(lib, "_eglMain"); +#endif #elif defined(_EGL_PLATFORM_X) /* XXX also prepend a directory path??? */ sprintf(driverFilename, "%s.so", driverName); _eglLog(_EGL_DEBUG, "dlopen(%s)", driverFilename); -#endif lib = open_library(driverFilename); - if (!lib) { _eglLog(_EGL_WARNING, "Could not open %s (%s)", driverFilename, dlerror()); return NULL; } - -#if defined(_EGL_PLATFORM_WINDOWS) - mainFunc = (_EGLMain_t) GetProcAddress(lib, "_eglMain"); -#elif defined(_EGL_PLATFORM_X) mainFunc = (_EGLMain_t) dlsym(lib, "_eglMain"); #endif -- cgit v1.2.3 From bf3040d4767702f09e042c58f0f5aa942ddb9a16 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 20:00:59 -0600 Subject: mesa: glsl: generate error for main(any args) --- src/mesa/shader/slang/slang_compile.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 82df9b95d3..5ce0e2e321 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -2025,6 +2025,12 @@ parse_code_unit(slang_parse_ctx * C, slang_code_unit * unit, A.vartable = o.vartable; A.log = C->L; + /* main() takes no parameters */ + if (mainFunc->param_count > 0) { + slang_info_log_error(A.log, "main() takes no arguments"); + return GL_FALSE; + } + _slang_codegen_function(&A, mainFunc); shader->Main = GL_TRUE; /* this shader defines main() */ -- cgit v1.2.3 From 6b888a10d37f43feae272d0927d087916e9e3c0d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 20:12:50 -0600 Subject: mesa: glsl: additional type checking for assignments, inequalities --- src/mesa/shader/slang/slang_codegen.c | 16 ++++++++++++++++ 1 file changed, 16 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 036196a8d2..94497cd833 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2889,6 +2889,10 @@ _slang_assignment_compatible(slang_assemble_ctx *A, t0.spec._struct->a_name != t1.spec._struct->a_name) return GL_FALSE; + if (t0.spec.type == SLANG_SPEC_FLOAT && + t1.spec.type == SLANG_SPEC_BOOL) + return GL_FALSE; + #if 0 /* not used just yet - causes problems elsewhere */ if (t0.spec.type == SLANG_SPEC_INT && t1.spec.type == SLANG_SPEC_FLOAT) @@ -3237,6 +3241,18 @@ _slang_gen_compare(slang_assemble_ctx *A, slang_operation *oper, return NULL; } + if (oper->type != SLANG_OPER_EQUAL && + oper->type != SLANG_OPER_NOTEQUAL) { + /* <, <=, >, >= can only be used with scalars */ + if ((t0.spec.type != SLANG_SPEC_INT && + t0.spec.type != SLANG_SPEC_FLOAT) || + (t1.spec.type != SLANG_SPEC_INT && + t1.spec.type != SLANG_SPEC_FLOAT)) { + slang_info_log_error(A->log, "Illegal type(s) for inequality operator"); + return NULL; + } + } + n = new_node2(opcode, _slang_gen_operation(A, &oper->children[0]), _slang_gen_operation(A, &oper->children[1])); -- cgit v1.2.3 From a4fad98dfb95a83de64e1685542a7e65c478cce7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 20:28:14 -0600 Subject: mesa: glsl: disallow user-defined functions/vars prefixed with gl_ --- src/mesa/shader/slang/slang_compile.c | 27 +++++++++++++++++++++++++++ 1 file changed, 27 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 5ce0e2e321..f90bbd8c56 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -60,6 +60,20 @@ #define TYPE_SPECIFIER_COUNT 32 +/** + * Check if the given identifier is legal. + */ +static GLboolean +legal_identifier(slang_atom name) +{ + /* "gl_" is a reserved prefix */ + if (_mesa_strncmp((char *) name, "gl_", 3) == 0) { + return GL_FALSE; + } + return GL_TRUE; +} + + /** * Allocate storage for a variable of 'size' bytes from given pool. * Return the allocated address for the variable. @@ -835,6 +849,12 @@ parse_statement(slang_parse_ctx * C, slang_output_ctx * O, o->type = SLANG_OPER_VARIABLE_DECL; o->locals->outer_scope = O->vars; o->a_id = O->vars->variables[i]->a_name; + + if (!legal_identifier(o->a_id)) { + slang_info_log_error(C->L, "illegal variable name '%s'", + (char *) o->a_id); + return 0; + } } } } @@ -1404,6 +1424,7 @@ parse_operator_name(slang_parse_ctx * C) return 0; } + static int parse_function_prototype(slang_parse_ctx * C, slang_output_ctx * O, slang_function * func) @@ -1441,6 +1462,12 @@ parse_function_prototype(slang_parse_ctx * C, slang_output_ctx * O, return 0; } + if (!legal_identifier(func->header.a_name)) { + slang_info_log_error(C->L, "illegal function name '%s'", + (char *) func->header.a_name); + return 0; + } + /* parse function parameters */ while (*C->I++ == PARAMETER_NEXT) { slang_variable *p = slang_variable_scope_grow(func->parameters); -- cgit v1.2.3 From efd730c5d3acaec035dfdbb0b7ed042e7c381a03 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 5 Aug 2008 20:56:18 -0600 Subject: mesa: glsl: code consolidation in _slang_gen_declaration() --- src/mesa/shader/slang/slang_codegen.c | 65 ++++++++++++++++------------------- 1 file changed, 29 insertions(+), 36 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 94497cd833..f891068322 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2717,6 +2717,7 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) slang_ir_node *varDecl; slang_variable *v; const char *varName = (char *) oper->a_id; + slang_operation *initializer; assert(oper->num_children == 0 || oper->num_children == 1); @@ -2733,26 +2734,24 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) varDecl = _slang_gen_var_decl(A, v); + /* check if the var has an initializer */ if (oper->num_children > 0) { - /* child is initializer */ - slang_ir_node *var, *init, *rhs; assert(oper->num_children == 1); - var = new_var(A, oper, oper->a_id); - if (!var) { - slang_info_log_error(A->log, "undefined variable '%s'", varName); - return NULL; - } - /* XXX make copy of this initializer? */ - rhs = _slang_gen_operation(A, &oper->children[0]); - if (!rhs) - return NULL; /* must have found an error */ - init = new_node2(IR_COPY, var, rhs); - - /*assert(rhs->Opcode != IR_SEQ);*/ - n = new_seq(varDecl, init); + initializer = &oper->children[0]; } else if (v->initializer) { - slang_ir_node *var, *init, *rhs; + initializer = v->initializer; + } + else { + initializer = NULL; + } + + if (initializer) { + slang_ir_node *var, *init; + + /* XXX todo: type check/compare var and initializer */ + + var = new_var(A, oper, oper->a_id); if (!var) { slang_info_log_error(A->log, "undefined variable '%s'", varName); @@ -2763,42 +2762,36 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) /* if the variable is const, the initializer must be a const * expression as well. */ - if (!_slang_is_constant_expr(v->initializer)) { +#if 0 + if (!_slang_is_constant_expr(initializer)) { slang_info_log_error(A->log, "initializer for %s not constant", varName); return NULL; } +#endif } -#if 0 - /* XXX make copy of this initializer? */ - { - slang_operation dup; - slang_operation_construct(&dup); - slang_operation_copy(&dup, v->initializer); - _slang_simplify(&dup, &A->space, A->atoms); - rhs = _slang_gen_operation(A, &dup); - } -#else - _slang_simplify(v->initializer, &A->space, A->atoms); - rhs = _slang_gen_operation(A, v->initializer); -#endif - if (!rhs) + _slang_simplify(initializer, &A->space, A->atoms); + + init = _slang_gen_operation(A, initializer); + if (!init) return NULL; - /*assert(rhs->Store);*/ + /*assert(init->Store);*/ - if (rhs->Store && var->Store->Size != rhs->Store->Size) { + /* XXX remove this when type checking is added above */ + if (init->Store && var->Store->Size != init->Store->Size) { slang_info_log_error(A->log, "invalid assignment (wrong types)"); return NULL; } - init = new_node2(IR_COPY, var, rhs); - n = new_seq(varDecl, init); + n = new_node2(IR_COPY, var, init); + n = new_seq(varDecl, n); } else { n = varDecl; } + return n; } @@ -2893,11 +2886,11 @@ _slang_assignment_compatible(slang_assemble_ctx *A, t1.spec.type == SLANG_SPEC_BOOL) return GL_FALSE; -#if 0 /* not used just yet - causes problems elsewhere */ if (t0.spec.type == SLANG_SPEC_INT && t1.spec.type == SLANG_SPEC_FLOAT) return GL_FALSE; +#if 0 /* not used just yet - causes problems elsewhere */ if (t0.spec.type == SLANG_SPEC_BOOL && t1.spec.type == SLANG_SPEC_FLOAT) return GL_FALSE; -- cgit v1.2.3 From d23b54a423b537fc08543299f9df086e831686fc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 08:39:54 -0600 Subject: fix some FBO/texture queries (bug 15296) --- src/mesa/main/fbobject.c | 14 ++++++++++++-- 1 file changed, 12 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 0ae69bdce7..960cc6da22 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -1525,7 +1525,12 @@ _mesa_GetFramebufferAttachmentParameterivEXT(GLenum target, GLenum attachment, return; case GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_CUBE_MAP_FACE_EXT: if (att->Type == GL_TEXTURE) { - *params = GL_TEXTURE_CUBE_MAP_POSITIVE_X + att->CubeMapFace; + if (att->Texture && att->Texture->Target == GL_TEXTURE_CUBE_MAP) { + *params = GL_TEXTURE_CUBE_MAP_POSITIVE_X + att->CubeMapFace; + } + else { + *params = 0; + } } else { _mesa_error(ctx, GL_INVALID_ENUM, @@ -1534,7 +1539,12 @@ _mesa_GetFramebufferAttachmentParameterivEXT(GLenum target, GLenum attachment, return; case GL_FRAMEBUFFER_ATTACHMENT_TEXTURE_3D_ZOFFSET_EXT: if (att->Type == GL_TEXTURE) { - *params = att->Zoffset; + if (att->Texture && att->Texture->Target == GL_TEXTURE_3D) { + *params = att->Zoffset; + } + else { + *params = 0; + } } else { _mesa_error(ctx, GL_INVALID_ENUM, -- cgit v1.2.3 From 8d1627c25a24f7bc14946bfe268d2d1c5dcde470 Mon Sep 17 00:00:00 2001 From: Haihao Xiang Date: Wed, 6 Aug 2008 09:05:09 -0600 Subject: mesa: glsl: count number of temp regs used --- src/mesa/shader/slang/slang_codegen.c | 2 -- src/mesa/shader/slang/slang_link.c | 35 ++++++++++++++++++++++++++++++++++- 2 files changed, 34 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index f891068322..142e240802 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2473,8 +2473,6 @@ _slang_gen_var_decl(slang_assemble_ctx *A, slang_variable *var) */ n->Store->Swizzle = SWIZZLE_NOOP; } - - A->program->NumTemporaries++; /* an approximation */ } return n; } diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 1de561930a..53de4232be 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -276,6 +276,36 @@ _slang_resolve_attributes(struct gl_shader_program *shProg, } +/** + * Scan program instructions to update the program's NumTemporaries field. + * Note: this implemenation relies on the code generator allocating + * temps in increasing order (0, 1, 2, ... ). + */ +static void +_slang_count_temporaries(struct gl_program *prog) +{ + GLuint i, j; + GLint maxIndex = -1; + + for (i = 0; i < prog->NumInstructions; i++) { + const struct prog_instruction *inst = prog->Instructions + i; + const GLuint numSrc = _mesa_num_inst_src_regs(inst->Opcode); + for (j = 0; j < numSrc; j++) { + if (inst->SrcReg[j].File == PROGRAM_TEMPORARY) { + if (maxIndex < inst->SrcReg[j].Index) + maxIndex = inst->SrcReg[j].Index; + } + if (inst->DstReg.File == PROGRAM_TEMPORARY) { + if (maxIndex < inst->DstReg.Index) + maxIndex = inst->DstReg.Index; + } + } + } + + prog->NumTemporaries = (GLuint) (maxIndex + 1); +} + + /** * Scan program instructions to update the program's InputsRead and * OutputsWritten fields. @@ -459,6 +489,7 @@ _slang_link(GLcontext *ctx, if (shProg->VertexProgram) { _slang_update_inputs_outputs(&shProg->VertexProgram->Base); + _slang_count_temporaries(&shProg->VertexProgram->Base); if (!(shProg->VertexProgram->Base.OutputsWritten & (1 << VERT_RESULT_HPOS))) { /* the vertex program did not compute a vertex position */ link_error(shProg, @@ -466,8 +497,10 @@ _slang_link(GLcontext *ctx, return; } } - if (shProg->FragmentProgram) + if (shProg->FragmentProgram) { + _slang_count_temporaries(&shProg->FragmentProgram->Base); _slang_update_inputs_outputs(&shProg->FragmentProgram->Base); + } /* Check that all the varying vars needed by the fragment shader are * actually produced by the vertex shader. -- cgit v1.2.3 From 93ed8f9fc9edddff62bbe4720c0b9029a1f8b6f5 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 11:15:42 -0600 Subject: mesa: glsl: fix SLANG_OPER_VARIABLE_DECL type checking case --- src/mesa/shader/slang/slang_typeinfo.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_typeinfo.c b/src/mesa/shader/slang/slang_typeinfo.c index 26f9f6ca85..8572118f3f 100644 --- a/src/mesa/shader/slang/slang_typeinfo.c +++ b/src/mesa/shader/slang/slang_typeinfo.c @@ -392,7 +392,6 @@ _slang_typeof_operation_(slang_operation * op, switch (op->type) { case SLANG_OPER_BLOCK_NO_NEW_SCOPE: case SLANG_OPER_BLOCK_NEW_SCOPE: - case SLANG_OPER_VARIABLE_DECL: case SLANG_OPER_ASM: case SLANG_OPER_BREAK: case SLANG_OPER_CONTINUE: @@ -477,6 +476,7 @@ _slang_typeof_operation_(slang_operation * op, } break; case SLANG_OPER_IDENTIFIER: + case SLANG_OPER_VARIABLE_DECL: { slang_variable *var; var = _slang_locate_variable(op->locals, op->a_id, GL_TRUE); -- cgit v1.2.3 From b9be69b85e8955fa9f773a65e7c3de5c4bb6e507 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 11:15:55 -0600 Subject: mesa: glsl: more assignment type checking Also that const declarations have initializers and that uniforms/samplers can't have initializers. --- src/mesa/shader/slang/slang_codegen.c | 147 ++++++++++++++++++++-------------- 1 file changed, 87 insertions(+), 60 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 142e240802..edb060a23f 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2704,6 +2704,62 @@ _slang_is_constant_expr(const slang_operation *oper) } +/** + * Check if an assignment of type t1 to t0 is legal. + * XXX more cases needed. + */ +static GLboolean +_slang_assignment_compatible(slang_assemble_ctx *A, + slang_operation *op0, + slang_operation *op1) +{ + slang_typeinfo t0, t1; + GLuint sz0, sz1; + + + slang_typeinfo_construct(&t0); + _slang_typeof_operation(A, op0, &t0); + + slang_typeinfo_construct(&t1); + _slang_typeof_operation(A, op1, &t1); + + sz0 = _slang_sizeof_type_specifier(&t0.spec); + sz1 = _slang_sizeof_type_specifier(&t1.spec); + +#if 1 + if (sz0 != sz1) { + /*printf("assignment size mismatch %u vs %u\n", sz0, sz1);*/ + return GL_FALSE; + } +#endif + + if (t0.spec.type == SLANG_SPEC_STRUCT && + t1.spec.type == SLANG_SPEC_STRUCT && + t0.spec._struct->a_name != t1.spec._struct->a_name) + return GL_FALSE; + + if (t0.spec.type == SLANG_SPEC_FLOAT && + t1.spec.type == SLANG_SPEC_BOOL) + return GL_FALSE; + +#if 0 /* not used just yet - causes problems elsewhere */ + if (t0.spec.type == SLANG_SPEC_INT && + t1.spec.type == SLANG_SPEC_FLOAT) + return GL_FALSE; +#endif + + if (t0.spec.type == SLANG_SPEC_BOOL && + t1.spec.type == SLANG_SPEC_FLOAT) + return GL_FALSE; + + if (t0.spec.type == SLANG_SPEC_BOOL && + t1.spec.type == SLANG_SPEC_INT) + return GL_FALSE; + + return GL_TRUE; +} + + /** * Generate IR tree for a variable declaration. @@ -2717,7 +2773,8 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) const char *varName = (char *) oper->a_id; slang_operation *initializer; - assert(oper->num_children == 0 || oper->num_children == 1); + assert(oper->type == SLANG_OPER_VARIABLE_DECL); + assert(oper->num_children <= 1); v = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE); /*printf("Declare %s at %p\n", varName, (void *) v);*/ @@ -2731,6 +2788,8 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) #endif varDecl = _slang_gen_var_decl(A, v); + if (!varDecl) + return NULL; /* check if the var has an initializer */ if (oper->num_children > 0) { @@ -2744,11 +2803,22 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) initializer = NULL; } + if (v->type.qualifier == SLANG_QUAL_CONST && !initializer) { + slang_info_log_error(A->log, + "const-qualified variable '%s' requires initializer", + varName); + return NULL; + } + + if (initializer) { slang_ir_node *var, *init; - /* XXX todo: type check/compare var and initializer */ - + /* type check/compare var and initializer */ + if (!_slang_assignment_compatible(A, oper, initializer)) { + slang_info_log_error(A->log, "illegal types in assignment"); + return NULL; + } var = new_var(A, oper, oper->a_id); if (!var) { @@ -2846,61 +2916,6 @@ _slang_gen_swizzle(slang_ir_node *child, GLuint swizzle) } -/** - * Check if an assignment of type t1 to t0 is legal. - * XXX more cases needed. - */ -static GLboolean -_slang_assignment_compatible(slang_assemble_ctx *A, - slang_operation *op0, - slang_operation *op1) -{ - slang_typeinfo t0, t1; - GLuint sz0, sz1; - - - slang_typeinfo_construct(&t0); - _slang_typeof_operation(A, op0, &t0); - - slang_typeinfo_construct(&t1); - _slang_typeof_operation(A, op1, &t1); - - sz0 = _slang_sizeof_type_specifier(&t0.spec); - sz1 = _slang_sizeof_type_specifier(&t1.spec); - -#if 1 - if (sz0 != sz1) { - printf("size mismatch %u vs %u\n", sz0, sz1); - return GL_FALSE; - } -#endif - - if (t0.spec.type == SLANG_SPEC_STRUCT && - t1.spec.type == SLANG_SPEC_STRUCT && - t0.spec._struct->a_name != t1.spec._struct->a_name) - return GL_FALSE; - - if (t0.spec.type == SLANG_SPEC_FLOAT && - t1.spec.type == SLANG_SPEC_BOOL) - return GL_FALSE; - - if (t0.spec.type == SLANG_SPEC_INT && - t1.spec.type == SLANG_SPEC_FLOAT) - return GL_FALSE; - -#if 0 /* not used just yet - causes problems elsewhere */ - if (t0.spec.type == SLANG_SPEC_BOOL && - t1.spec.type == SLANG_SPEC_FLOAT) - return GL_FALSE; - - if (t0.spec.type == SLANG_SPEC_BOOL && - t1.spec.type == SLANG_SPEC_INT) - return GL_FALSE; -#endif - return GL_TRUE; -} - - /** * Generate IR tree for an assignment (=). */ @@ -3606,14 +3621,26 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, * store->Index = sampler number (0..7, typically) * store->Size = texture type index (1D, 2D, 3D, cube, etc) */ - GLint sampNum = _mesa_add_sampler(prog->Parameters, varName, datatype); - store = _slang_new_ir_storage(PROGRAM_SAMPLER, sampNum, texIndex); + if (var->initializer) { + slang_info_log_error(A->log, "illegal assignment to '%s'", varName); + return GL_FALSE; + } + else { + GLint sampNum = _mesa_add_sampler(prog->Parameters, varName, datatype); + store = _slang_new_ir_storage(PROGRAM_SAMPLER, sampNum, texIndex); + } if (dbg) printf("SAMPLER "); } else if (var->type.qualifier == SLANG_QUAL_UNIFORM) { /* Uniform variable */ const GLint totalSize = array_size(size, var->array_len); const GLuint swizzle = _slang_var_swizzle(totalSize, 0); + + if (var->initializer) { + slang_info_log_error(A->log, "illegal assignment to '%s'", varName); + return GL_FALSE; + } + if (prog) { /* user-defined uniform */ if (datatype == GL_NONE) { -- cgit v1.2.3 From 64a795d8e60c42e7ec8c370507864faf3f85d99e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 11:22:36 -0600 Subject: mesa: glsl: check that attribute vars are of float/vec/mat type --- src/mesa/shader/slang/slang_codegen.c | 8 ++++++++ src/mesa/shader/slang/slang_typeinfo.c | 28 ++++++++++++++++++++++++++++ src/mesa/shader/slang/slang_typeinfo.h | 3 +++ 3 files changed, 39 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index edb060a23f..f717a22ac9 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -3726,6 +3726,14 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, if (dbg) printf("VARYING "); } else if (var->type.qualifier == SLANG_QUAL_ATTRIBUTE) { + /* attributes must be float, vec or mat */ + if (!_slang_type_is_float_vec_mat(var->type.specifier.type)) { + slang_info_log_error(A->log, + "attribute '%s' must be float/vector/matrix", + varName); + return GL_FALSE; + } + if (prog) { /* user-defined vertex attribute */ const GLint attr = -1; /* unknown */ diff --git a/src/mesa/shader/slang/slang_typeinfo.c b/src/mesa/shader/slang/slang_typeinfo.c index 8572118f3f..b3255512ba 100644 --- a/src/mesa/shader/slang/slang_typeinfo.c +++ b/src/mesa/shader/slang/slang_typeinfo.c @@ -874,6 +874,34 @@ _slang_type_is_vector(slang_type_specifier_type ty) } +/** + * Determine if a type is a float, float vector or float matrix. + * \return GL_TRUE if so, GL_FALSE otherwise + */ +GLboolean +_slang_type_is_float_vec_mat(slang_type_specifier_type ty) +{ + switch (ty) { + case SLANG_SPEC_FLOAT: + case SLANG_SPEC_VEC2: + case SLANG_SPEC_VEC3: + case SLANG_SPEC_VEC4: + case SLANG_SPEC_MAT2: + case SLANG_SPEC_MAT3: + case SLANG_SPEC_MAT4: + case SLANG_SPEC_MAT23: + case SLANG_SPEC_MAT32: + case SLANG_SPEC_MAT24: + case SLANG_SPEC_MAT42: + case SLANG_SPEC_MAT34: + case SLANG_SPEC_MAT43: + return GL_TRUE; + default: + return GL_FALSE; + } +} + + /** * Given a vector type, return the type of the vector's elements. * For a matrix, return the type of the columns. diff --git a/src/mesa/shader/slang/slang_typeinfo.h b/src/mesa/shader/slang/slang_typeinfo.h index 960e99bda2..b3ad06b65c 100644 --- a/src/mesa/shader/slang/slang_typeinfo.h +++ b/src/mesa/shader/slang/slang_typeinfo.h @@ -191,6 +191,9 @@ _slang_type_is_matrix(slang_type_specifier_type); extern GLboolean _slang_type_is_vector(slang_type_specifier_type); +extern GLboolean +_slang_type_is_float_vec_mat(slang_type_specifier_type); + extern slang_type_specifier_type _slang_type_base(slang_type_specifier_type); -- cgit v1.2.3 From acb81d0268bf96efd35196c619ca269ffbe6c0d3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 11:45:42 -0600 Subject: mesa: glsl: more type checking for attribute/varying/uniform vars/locals --- src/mesa/shader/slang/slang_codegen.c | 18 ++++++++++++++---- 1 file changed, 14 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index f717a22ac9..13cad6a412 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2777,8 +2777,18 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) assert(oper->num_children <= 1); v = _slang_locate_variable(oper->locals, oper->a_id, GL_TRUE); - /*printf("Declare %s at %p\n", varName, (void *) v);*/ - assert(v); + if (!v) + return NULL; /* "shouldn't happen" */ + + if (v->type.qualifier == SLANG_QUAL_ATTRIBUTE || + v->type.qualifier == SLANG_QUAL_VARYING || + v->type.qualifier == SLANG_QUAL_UNIFORM) { + /* can't declare attribute/uniform vars inside functions */ + slang_info_log_error(A->log, + "local variable '%s' cannot be an attribute/uniform/varying", + varName); + return NULL; + } #if 0 if (v->declared) { @@ -3685,9 +3695,9 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, if (dbg) printf("UNIFORM (sz %d) ", totalSize); } else if (var->type.qualifier == SLANG_QUAL_VARYING) { - if (var->type.specifier.type == SLANG_SPEC_STRUCT) { + if (!_slang_type_is_float_vec_mat(var->type.specifier.type)) { slang_info_log_error(A->log, - "varying '%s' cannot be a structure type", + "varying '%s' must be float/vector/matrix", varName); return GL_FALSE; } -- cgit v1.2.3 From ac553c6ce4718aa1e90f49c910c14e13d0d5de39 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 11:58:52 -0600 Subject: mesa: glsl: disallow initializers for varying vars --- src/mesa/shader/slang/slang_codegen.c | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 13cad6a412..e36b689b70 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -3647,7 +3647,7 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, const GLuint swizzle = _slang_var_swizzle(totalSize, 0); if (var->initializer) { - slang_info_log_error(A->log, "illegal assignment to '%s'", varName); + slang_info_log_error(A->log, "illegal initializer for uniform '%s'", varName); return GL_FALSE; } @@ -3695,13 +3695,21 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, if (dbg) printf("UNIFORM (sz %d) ", totalSize); } else if (var->type.qualifier == SLANG_QUAL_VARYING) { - if (!_slang_type_is_float_vec_mat(var->type.specifier.type)) { + /* varyings must be float, vec or mat */ + if (!_slang_type_is_float_vec_mat(var->type.specifier.type) && + var->type.specifier.type != SLANG_SPEC_ARRAY) { slang_info_log_error(A->log, "varying '%s' must be float/vector/matrix", varName); return GL_FALSE; } + if (var->initializer) { + slang_info_log_error(A->log, "illegal initializer for varying '%s'", + varName); + return GL_FALSE; + } + if (prog) { /* user-defined varying */ GLint varyingLoc = _mesa_add_varying(prog->Varying, varName, size); -- cgit v1.2.3 From 5fc6f6a33faed0dbfe67a0689d8f261633e21529 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 12:21:25 -0600 Subject: mesa: glsl: check for rect tex samplers --- src/mesa/shader/slang/slang_codegen.c | 10 +++++++++- 1 file changed, 9 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index e36b689b70..9c17dceb19 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -3635,7 +3635,15 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, slang_info_log_error(A->log, "illegal assignment to '%s'", varName); return GL_FALSE; } - else { +#if FEATURE_es2_glsl /* XXX should use FEATURE_texture_rect */ + /* disallow rect samplers */ + if (var->type.specifier.type == SLANG_SPEC_SAMPLER2DRECT || + var->type.specifier.type == SLANG_SPEC_SAMPLER2DRECTSHADOW) { + slang_info_log_error(A->log, "invalid sampler type for '%s'", varName); + return GL_FALSE; + } +#endif + { GLint sampNum = _mesa_add_sampler(prog->Parameters, varName, datatype); store = _slang_new_ir_storage(PROGRAM_SAMPLER, sampNum, texIndex); } -- cgit v1.2.3 From 18cd9c229a1fc8da8b7669b8d1d100f6bbeca183 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 12:45:14 -0600 Subject: mesa: glsl: fix error check in get_uniformfv() --- src/mesa/shader/shader_api.c | 13 +++++++++---- 1 file changed, 9 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 5c18e55dac..86b3d536e5 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -50,6 +50,11 @@ +#ifndef GL_PROGRAM_BINARY_LENGTH_OES +#define GL_PROGRAM_BINARY_LENGTH_OES 0x8741 +#endif + + /** * Allocate a new gl_shader_program object, initialize it. */ @@ -906,6 +911,9 @@ _mesa_get_programiv(GLcontext *ctx, GLuint program, if (*params > 0) (*params)++; /* add one for terminating zero */ break; + case GL_PROGRAM_BINARY_LENGTH_OES: + *params = 0; + break; default: _mesa_error(ctx, GL_INVALID_ENUM, "glGetProgramiv(pname)"); return; @@ -999,7 +1007,7 @@ get_uniformfv(GLcontext *ctx, GLuint program, GLint location, GLfloat *params) { struct gl_shader_program *shProg - = _mesa_lookup_shader_program(ctx, program); + = _mesa_lookup_shader_program_err(ctx, program, "glGetUniform[if]v"); if (shProg) { if (shProg->Uniforms && location >= 0 && location < (GLint) shProg->Uniforms->NumUniforms) { @@ -1033,9 +1041,6 @@ get_uniformfv(GLcontext *ctx, GLuint program, GLint location, _mesa_error(ctx, GL_INVALID_OPERATION, "glGetUniformfv(location)"); } } - else { - _mesa_error(ctx, GL_INVALID_OPERATION, "glGetUniformfv(program)"); - } return 0; } -- cgit v1.2.3 From 7a6eba54d064cadf15f93df2c1748cf5e474ef03 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 13:07:09 -0600 Subject: mesa: glsl: fix glGetUniform for matrix queries --- src/mesa/shader/shader_api.c | 129 ++++++++++++++++++++++++++----------------- 1 file changed, 79 insertions(+), 50 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 86b3d536e5..3a85f47674 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -996,6 +996,71 @@ _mesa_get_shader_source(GLcontext *ctx, GLuint shader, GLsizei maxLength, } +static void +get_matrix_dims(GLenum type, GLint *rows, GLint *cols) +{ + switch (type) { + case GL_FLOAT_MAT2: + *rows = *cols = 2; + break; + case GL_FLOAT_MAT2x3: + *rows = 3; + *cols = 2; + break; + case GL_FLOAT_MAT2x4: + *rows = 4; + *cols = 2; + break; + case GL_FLOAT_MAT3: + *rows = 3; + *cols = 3; + break; + case GL_FLOAT_MAT3x2: + *rows = 2; + *cols = 3; + break; + case GL_FLOAT_MAT3x4: + *rows = 4; + *cols = 3; + break; + case GL_FLOAT_MAT4: + *rows = 4; + *cols = 4; + break; + case GL_FLOAT_MAT4x2: + *rows = 2; + *cols = 4; + break; + case GL_FLOAT_MAT4x3: + *rows = 3; + *cols = 4; + break; + default: + *rows = *cols = 0; + } +} + + +/** + * Determine the number of rows and columns occupied by a uniform + * according to its datatype. + */ +static void +get_uniform_rows_cols(const struct gl_program_parameter *p, + GLint *rows, GLint *cols) +{ + get_matrix_dims(p->DataType, rows, cols); + if (*rows == 0 && *cols == 0) { + /* not a matrix type, probably a float or vector */ + *rows = p->Size / 4 + 1; + if (p->Size % 4 == 0) + *cols = 4; + else + *cols = p->Size % 4; + } +} + + #define MAX_UNIFORM_ELEMENTS 16 /** @@ -1012,7 +1077,6 @@ get_uniformfv(GLcontext *ctx, GLuint program, GLint location, if (shProg->Uniforms && location >= 0 && location < (GLint) shProg->Uniforms->NumUniforms) { GLint progPos; - GLuint i; const struct gl_program *prog = NULL; progPos = shProg->Uniforms->Uniforms[location].VertPos; @@ -1028,13 +1092,23 @@ get_uniformfv(GLcontext *ctx, GLuint program, GLint location, ASSERT(prog); if (prog) { + const struct gl_program_parameter *p = + &prog->Parameters->Parameters[progPos]; + GLint rows, cols, i, j, k; + /* See uniformiv() below */ - assert(prog->Parameters->Parameters[progPos].Size <= MAX_UNIFORM_ELEMENTS); + assert(p->Size <= MAX_UNIFORM_ELEMENTS); - for (i = 0; i < prog->Parameters->Parameters[progPos].Size; i++) { - params[i] = prog->Parameters->ParameterValues[progPos][i]; + get_uniform_rows_cols(p, &rows, &cols); + + k = 0; + for (i = 0; i < rows; i++) { + for (j = 0; j < cols; j++ ) { + params[k++] = prog->Parameters->ParameterValues[progPos+i][j]; + } } - return prog->Parameters->Parameters[progPos].Size; + + return p->Size; } } else { @@ -1415,51 +1489,6 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, } -static void -get_matrix_dims(GLenum type, GLint *rows, GLint *cols) -{ - switch (type) { - case GL_FLOAT_MAT2: - *rows = *cols = 2; - break; - case GL_FLOAT_MAT2x3: - *rows = 3; - *cols = 2; - break; - case GL_FLOAT_MAT2x4: - *rows = 4; - *cols = 2; - break; - case GL_FLOAT_MAT3: - *rows = 3; - *cols = 3; - break; - case GL_FLOAT_MAT3x2: - *rows = 2; - *cols = 3; - break; - case GL_FLOAT_MAT3x4: - *rows = 4; - *cols = 3; - break; - case GL_FLOAT_MAT4: - *rows = 4; - *cols = 4; - break; - case GL_FLOAT_MAT4x2: - *rows = 2; - *cols = 4; - break; - case GL_FLOAT_MAT4x3: - *rows = 3; - *cols = 4; - break; - default: - *rows = *cols = 0; - } -} - - static void set_program_uniform_matrix(GLcontext *ctx, struct gl_program *program, GLuint location, GLuint count, -- cgit v1.2.3 From 7e2458c7b55b2662067d8cce8a962d8c2bbd3a43 Mon Sep 17 00:00:00 2001 From: Jonathan White Date: Wed, 6 Aug 2008 13:40:03 -0600 Subject: egl: fix version handling --- src/egl/main/eglapi.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/egl/main/eglapi.c b/src/egl/main/eglapi.c index 49d1f3d0eb..9df938e188 100644 --- a/src/egl/main/eglapi.c +++ b/src/egl/main/eglapi.c @@ -63,6 +63,8 @@ eglGetDisplay(NativeDisplayType nativeDisplay) EGLBoolean EGLAPIENTRY eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor) { + EGLint major_int, minor_int; + if (dpy) { EGLBoolean retVal; _EGLDisplay *dpyPriv = _eglLookupDisplay(dpy); @@ -77,12 +79,19 @@ eglInitialize(EGLDisplay dpy, EGLint *major, EGLint *minor) } /* Initialize the particular driver now */ retVal = dpyPriv->Driver->API.Initialize(dpyPriv->Driver, dpy, - major, minor); + &major_int, &minor_int); - dpyPriv->Driver->APImajor = *major; - dpyPriv->Driver->APIminor = *minor; + dpyPriv->Driver->APImajor = major_int; + dpyPriv->Driver->APIminor = minor_int; snprintf(dpyPriv->Driver->Version, sizeof(dpyPriv->Driver->Version), - "%d.%d (%s)", *major, *minor, dpyPriv->Driver->Name); + "%d.%d (%s)", major_int, minor_int, dpyPriv->Driver->Name); + + /* Update applications version of major and minor if not NULL */ + if((major != NULL) && (minor != NULL)) + { + *major = major_int; + *minor = minor_int; + } return retVal; } -- cgit v1.2.3 From 8a40fe06d72aa8b0c3a03b92b60a0f5b4ea61987 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 16:26:47 -0600 Subject: mesa: glsl: fix a number of glUniform issues Additional error checking. Allow setting elements of uniform arrays. This involves encoding both a uniform location and a parameter offset in the value returned by glGetUniformLocation(). Limit glUniform[if]v()'s count to the size of the uniform array. When setting bool-valued uniforms, convert all float/int values to 0 or 1. --- src/mesa/shader/shader_api.c | 230 +++++++++++++++++++++++++++++++++++++------ 1 file changed, 199 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 3a85f47674..c2c58fffde 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -752,6 +752,36 @@ sizeof_glsl_type(GLenum type) } +static GLboolean +is_boolean_type(GLenum type) +{ + switch (type) { + case GL_BOOL: + case GL_BOOL_VEC2: + case GL_BOOL_VEC3: + case GL_BOOL_VEC4: + return GL_TRUE; + default: + return GL_FALSE; + } +} + + +static GLboolean +is_integer_type(GLenum type) +{ + switch (type) { + case GL_INT: + case GL_INT_VEC2: + case GL_INT_VEC3: + case GL_INT_VEC4: + return GL_TRUE; + default: + return GL_FALSE; + } +} + + static void _mesa_get_active_attrib(GLcontext *ctx, GLuint program, GLuint index, GLsizei maxLength, GLsizei *length, GLint *size, @@ -778,6 +808,30 @@ _mesa_get_active_attrib(GLcontext *ctx, GLuint program, GLuint index, } +static struct gl_program_parameter * +get_uniform_parameter(const struct gl_shader_program *shProg, GLuint index) +{ + const struct gl_program *prog; + GLint progPos; + + progPos = shProg->Uniforms->Uniforms[index].VertPos; + if (progPos >= 0) { + prog = &shProg->VertexProgram->Base; + } + else { + progPos = shProg->Uniforms->Uniforms[index].FragPos; + if (progPos >= 0) { + prog = &shProg->FragmentProgram->Base; + } + } + + if (!prog || progPos < 0) + return NULL; /* should never happen */ + + return &prog->Parameters->Parameters[progPos]; +} + + /** * Called via ctx->Driver.GetActiveUniform(). */ @@ -1147,12 +1201,42 @@ _mesa_get_uniformiv(GLcontext *ctx, GLuint program, GLint location, } +/** + * The value returned by GetUniformLocation actually encodes two things: + * 1. the index into the prog->Uniforms[] array for the uniform + * 2. an offset in the prog->ParameterValues[] array for specifying array + * elements or structure fields. + * This function merges those two values. + */ +static void +merge_location_offset(GLint *location, GLint offset) +{ + *location = *location | (offset << 16); +} + + +/** + * Seperate the uniform location and parameter offset. See above. + */ +static void +split_location_offset(GLint *location, GLint *offset) +{ + *offset = (*location >> 16); + *location = *location & 0xffff; +} + + /** * Called via ctx->Driver.GetUniformLocation(). + * + * The return value will encode two values, the uniform location and an + * offset (used for arrays, structs). */ static GLint _mesa_get_uniform_location(GLcontext *ctx, GLuint program, const GLchar *name) { + GLint offset = 0, location = -1; + struct gl_shader_program *shProg = _mesa_lookup_shader_program_err(ctx, program, "glGetUniformLocation"); @@ -1168,7 +1252,54 @@ _mesa_get_uniform_location(GLcontext *ctx, GLuint program, const GLchar *name) * actually used. */ - return _mesa_lookup_uniform(shProg->Uniforms, name); + /* XXX we need to be able to parse uniform names for structs and arrays + * such as: + * mymatrix[1] + * mystruct.field1 + */ + + { + /* handle 1-dimension arrays here... */ + char *c = strchr(name, '['); + if (c) { + /* truncate name at [ */ + const GLint len = c - name; + GLchar *newName = _mesa_malloc(len + 1); + if (!newName) + return -1; /* out of mem */ + _mesa_memcpy(newName, name, len); + newName[len] = 0; + + location = _mesa_lookup_uniform(shProg->Uniforms, newName); + if (location >= 0) { + const GLint element = _mesa_atoi(c + 1); + if (element > 0) { + /* get type of the uniform array element */ + struct gl_program_parameter *p; + p = get_uniform_parameter(shProg, location); + if (p) { + GLint rows, cols; + get_matrix_dims(p->DataType, &rows, &cols); + if (rows < 1) + rows = 1; + offset = element * rows; + } + } + } + + _mesa_free(newName); + } + } + + if (location < 0) { + location = _mesa_lookup_uniform(shProg->Uniforms, name); + } + + if (location >= 0) { + merge_location_offset(&location, offset); + } + + return location; } @@ -1341,23 +1472,33 @@ compatible_types(GLenum userType, GLenum targetType) /** * Set the value of a program's uniform variable. * \param program the program whose uniform to update - * \param location the location/index of the uniform + * \param index the index of the program parameter for the uniform + * \param offset additional parameter slot offset (for arrays) * \param type the datatype of the uniform * \param count the number of uniforms to set * \param elems number of elements per uniform * \param values the new values */ static void -set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, - GLenum type, GLsizei count, GLint elems, const void *values) +set_program_uniform(GLcontext *ctx, struct gl_program *program, + GLint index, GLint offset, + GLenum type, GLsizei count, GLint elems, + const void *values) { + assert(offset >= 0); + if (!compatible_types(type, - program->Parameters->Parameters[location].DataType)) { + program->Parameters->Parameters[index].DataType)) { _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(type mismatch)"); return; } - if (program->Parameters->Parameters[location].Type == PROGRAM_SAMPLER) { + if (index + offset > program->Parameters->Size) { + /* out of bounds! */ + return; + } + + if (program->Parameters->Parameters[index].Type == PROGRAM_SAMPLER) { /* This controls which texture unit which is used by a sampler */ GLuint texUnit, sampler; @@ -1369,7 +1510,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, return; } - sampler = (GLuint) program->Parameters->ParameterValues[location][0]; + sampler = (GLuint) program->Parameters->ParameterValues[index][0]; texUnit = ((GLuint *) values)[0]; /* check that the sampler (tex unit index) is legal */ @@ -1388,18 +1529,19 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, else { /* ordinary uniform variable */ GLsizei k, i; + GLint slots = (program->Parameters->Parameters[index].Size + 3) / 4; - if (count * elems > (GLint) program->Parameters->Parameters[location].Size) { + if (count * elems > (GLint) program->Parameters->Parameters[index].Size) { _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(count too large)"); return; } + if (count > slots) + count = slots; + for (k = 0; k < count; k++) { - GLfloat *uniformVal = program->Parameters->ParameterValues[location + k]; - if (type == GL_INT || - type == GL_INT_VEC2 || - type == GL_INT_VEC3 || - type == GL_INT_VEC4) { + GLfloat *uniformVal = program->Parameters->ParameterValues[index + offset + k]; + if (is_integer_type(type)) { const GLint *iValues = ((const GLint *) values) + k * elems; for (i = 0; i < elems; i++) { uniformVal[i] = (GLfloat) iValues[i]; @@ -1411,6 +1553,13 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program, GLint location, uniformVal[i] = fValues[i]; } } + + /* if the uniform is bool-valued, convert to 1.0 or 0.0 */ + if (is_boolean_type(program->Parameters->Parameters[index].DataType)) { + for (i = 0; i < elems; i++) { + uniformVal[i] = uniformVal[i] ? 1.0 : 0.0; + } + } } } } @@ -1424,7 +1573,7 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, const GLvoid *values, GLenum type) { struct gl_shader_program *shProg = ctx->Shader.CurrentProgram; - GLint elems; + GLint elems, offset; if (!shProg || !shProg->LinkStatus) { _mesa_error(ctx, GL_INVALID_OPERATION, "glUniform(program not linked)"); @@ -1434,6 +1583,8 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, if (location == -1) return; /* The standard specifies this as a no-op */ + split_location_offset(&location, &offset); + if (location < 0 || location >= (GLint) shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniform(location)"); return; @@ -1472,38 +1623,48 @@ _mesa_uniform(GLcontext *ctx, GLint location, GLsizei count, * shader. We may need to update one or both shader's uniform here: */ if (shProg->VertexProgram) { - GLint loc = shProg->Uniforms->Uniforms[location].VertPos; - if (loc >= 0) { + /* convert uniform location to program parameter index */ + GLint index = shProg->Uniforms->Uniforms[location].VertPos; + if (index >= 0) { set_program_uniform(ctx, &shProg->VertexProgram->Base, - loc, type, count, elems, values); + index, offset, type, count, elems, values); } } if (shProg->FragmentProgram) { - GLint loc = shProg->Uniforms->Uniforms[location].FragPos; - if (loc >= 0) { + /* convert uniform location to program parameter index */ + GLint index = shProg->Uniforms->Uniforms[location].FragPos; + if (index >= 0) { set_program_uniform(ctx, &shProg->FragmentProgram->Base, - loc, type, count, elems, values); + index, offset, type, count, elems, values); } } } +/** + * Set a matrix-valued program parameter. + */ static void set_program_uniform_matrix(GLcontext *ctx, struct gl_program *program, - GLuint location, GLuint count, - GLuint rows, GLuint cols, + GLuint index, GLuint offset, + GLuint count, GLuint rows, GLuint cols, GLboolean transpose, const GLfloat *values) { GLuint mat, row, col; - GLuint dst = location, src = 0; + GLuint dst = index + offset, src = 0; GLint nr, nc; /* check that the number of rows, columns is correct */ - get_matrix_dims(program->Parameters->Parameters[location].DataType, &nr, &nc); + get_matrix_dims(program->Parameters->Parameters[index].DataType, &nr, &nc); if (rows != nr || cols != nc) { _mesa_error(ctx, GL_INVALID_OPERATION, - "glUniformMatrix(matrix size mismatch"); + "glUniformMatrix(matrix size mismatch)"); + return; + } + + if (index + offset > program->Parameters->Size) { + /* out of bounds! */ return; } @@ -1544,6 +1705,7 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, GLenum matrixType, GLint location, GLsizei count, GLboolean transpose, const GLfloat *values) { + GLint offset; struct gl_shader_program *shProg = ctx->Shader.CurrentProgram; if (!shProg || !shProg->LinkStatus) { @@ -1555,6 +1717,8 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, if (location == -1) return; /* The standard specifies this as a no-op */ + split_location_offset(&location, &offset); + if (location < 0 || location >= (GLint) shProg->Uniforms->NumUniforms) { _mesa_error(ctx, GL_INVALID_VALUE, "glUniformMatrix(location)"); return; @@ -1567,18 +1731,22 @@ _mesa_uniform_matrix(GLcontext *ctx, GLint cols, GLint rows, FLUSH_VERTICES(ctx, _NEW_PROGRAM); if (shProg->VertexProgram) { - GLint loc = shProg->Uniforms->Uniforms[location].VertPos; - if (loc >= 0) { + /* convert uniform location to program parameter index */ + GLint index = shProg->Uniforms->Uniforms[location].VertPos; + if (index >= 0) { set_program_uniform_matrix(ctx, &shProg->VertexProgram->Base, - loc, count, rows, cols, transpose, values); + index, offset, + count, rows, cols, transpose, values); } } if (shProg->FragmentProgram) { - GLint loc = shProg->Uniforms->Uniforms[location].FragPos; - if (loc >= 0) { + /* convert uniform location to program parameter index */ + GLint index = shProg->Uniforms->Uniforms[location].FragPos; + if (index >= 0) { set_program_uniform_matrix(ctx, &shProg->FragmentProgram->Base, - loc, count, rows, cols, transpose, values); + index, offset, + count, rows, cols, transpose, values); } } } -- cgit v1.2.3 From 95d3642122f48439e356b781f62649553f97c7ee Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 16:29:10 -0600 Subject: mesa: glsl: additional error checking --- src/mesa/shader/slang/slang_link.c | 12 ++++++++++++ 1 file changed, 12 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 53de4232be..8f07406ee4 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -445,6 +445,18 @@ _slang_link(GLcontext *ctx, _mesa_problem(ctx, "unexpected shader target in slang_link()"); } +#if FEATURE_es2_glsl + /* must have both a vertex and fragment program for ES2 */ + if (!vertProg) { + link_error(shProg, "missing vertex shader\n"); + return; + } + if (!fragProg) { + link_error(shProg, "missing fragment shader\n"); + return; + } +#endif + /* * Make copies of the vertex/fragment programs now since we'll be * changing src/dst registers after merging the uniforms and varying vars. -- cgit v1.2.3 From 35b9aedf7fc29ccc1d8c969b11eeb5eca57c5af2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 16:29:40 -0600 Subject: mesa: glsl: report 'Syntax Error' instead of 'Invalid external declaration' --- src/mesa/shader/slang/library/slang_shader.syn | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_shader.syn b/src/mesa/shader/slang/library/slang_shader.syn index 52a792ad17..4505758520 100644 --- a/src/mesa/shader/slang/library/slang_shader.syn +++ b/src/mesa/shader/slang/library/slang_shader.syn @@ -242,7 +242,9 @@ .emtcode PARAMETER_ARRAY_NOT_PRESENT 0 .emtcode PARAMETER_ARRAY_PRESENT 1 -.errtext INVALID_EXTERNAL_DECLARATION "2001: Invalid external declaration." +/* INVALID_EXTERNAL_DECLARATION seems to be reported when there's */ +/* any syntax errors... */ +.errtext INVALID_EXTERNAL_DECLARATION "2001: Syntax error." .errtext INVALID_OPERATOR_OVERRIDE "2002: Invalid operator override." .errtext LBRACE_EXPECTED "2003: '{' expected but '$err_token$' found." .errtext LPAREN_EXPECTED "2004: '(' expected but '$err_token$' found." -- cgit v1.2.3 From 695fc325032f642dea6a10267a57c3d9758bea62 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 16:29:50 -0600 Subject: mesa: glsl: regenerated file --- src/mesa/shader/slang/library/slang_shader_syn.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_shader_syn.h b/src/mesa/shader/slang/library/slang_shader_syn.h index 82909d9423..0c4c14f9d1 100644 --- a/src/mesa/shader/slang/library/slang_shader_syn.h +++ b/src/mesa/shader/slang/library/slang_shader_syn.h @@ -137,7 +137,7 @@ ".emtcode PARAMETER_NEXT 1\n" ".emtcode PARAMETER_ARRAY_NOT_PRESENT 0\n" ".emtcode PARAMETER_ARRAY_PRESENT 1\n" -".errtext INVALID_EXTERNAL_DECLARATION \"2001: Invalid external declaration.\"\n" +".errtext INVALID_EXTERNAL_DECLARATION \"2001: Syntax error.\"\n" ".errtext INVALID_OPERATOR_OVERRIDE \"2002: Invalid operator override.\"\n" ".errtext LBRACE_EXPECTED \"2003: '{' expected but '$err_token$' found.\"\n" ".errtext LPAREN_EXPECTED \"2004: '(' expected but '$err_token$' found.\"\n" -- cgit v1.2.3 From bd953fdd549310638b6960f358b31a40c05ca680 Mon Sep 17 00:00:00 2001 From: Jonathan White Date: Wed, 6 Aug 2008 16:37:49 -0600 Subject: egl: fixes for Windows --- src/egl/main/egldriver.c | 5 ++++- src/egl/main/egllog.c | 4 ++++ 2 files changed, 8 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/egl/main/egldriver.c b/src/egl/main/egldriver.c index f5edb58f43..43b1f51903 100644 --- a/src/egl/main/egldriver.c +++ b/src/egl/main/egldriver.c @@ -146,10 +146,12 @@ _eglChooseDriver(_EGLDisplay *dpy) (void) DefaultDriverName; +#if defined(_EGL_PLATFORM_X) /* First, if the EGL_DRIVER env var is set, use that */ driverName = getenv("EGL_DRIVER"); if (driverName) return _eglstrdup(driverName); +#endif #if 0 if (!displayString) { @@ -217,7 +219,8 @@ _eglOpenDriver(_EGLDisplay *dpy, const char *driverName, const char *args) #if defined(_EGL_PLATFORM_WINDOWS) /* Use static linking on Windows for now */ -#ifdef WINDOWS_STATIC_LINK +#ifdef WINDOWS_STATIC_LINK + lib = 0; mainFunc = (_EGLMain_t)_eglMain; #else /* XXX untested */ diff --git a/src/egl/main/egllog.c b/src/egl/main/egllog.c index 23eb523eeb..1d7a0a388c 100644 --- a/src/egl/main/egllog.c +++ b/src/egl/main/egllog.c @@ -21,7 +21,11 @@ static EGLint ReportingLevel = -1; static void log_level_initialize(void) { +#if defined(_EGL_PLATFORM_X) char *log_env = getenv("EGL_LOG_LEVEL"); +#else + char *log_env = NULL; +#endif if (log_env == NULL) { ReportingLevel = FALLBACK_LOG_LEVEL; -- cgit v1.2.3 From ea0007cc4ca077c7e3951c4fda122bd242728d70 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 17:14:22 -0600 Subject: softpipe: add texture border color code --- src/gallium/drivers/softpipe/sp_tex_sample.c | 28 +++++++++++++++++++--------- 1 file changed, 19 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 63b3b91110..ed150527e2 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -601,15 +601,25 @@ get_texel(struct tgsi_sampler *sampler, unsigned face, unsigned level, int x, int y, int z, float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j) { - const int tx = x % TILE_SIZE; - const int ty = y % TILE_SIZE; - const struct softpipe_cached_tile *tile - = sp_get_cached_tile_tex(sampler->pipe, sampler->cache, - x, y, z, face, level); - rgba[0][j] = tile->data.color[ty][tx][0]; - rgba[1][j] = tile->data.color[ty][tx][1]; - rgba[2][j] = tile->data.color[ty][tx][2]; - rgba[3][j] = tile->data.color[ty][tx][3]; + if (x < 0 || x >= sampler->texture->width[level] || + y < 0 || y >= sampler->texture->height[level] || + z < 0 || z >= sampler->texture->depth[level]) { + rgba[0][j] = sampler->state->border_color[0]; + rgba[1][j] = sampler->state->border_color[1]; + rgba[2][j] = sampler->state->border_color[2]; + rgba[3][j] = sampler->state->border_color[3]; + } + else { + const int tx = x % TILE_SIZE; + const int ty = y % TILE_SIZE; + const struct softpipe_cached_tile *tile + = sp_get_cached_tile_tex(sampler->pipe, sampler->cache, + x, y, z, face, level); + rgba[0][j] = tile->data.color[ty][tx][0]; + rgba[1][j] = tile->data.color[ty][tx][1]; + rgba[2][j] = tile->data.color[ty][tx][2]; + rgba[3][j] = tile->data.color[ty][tx][3]; + } } -- cgit v1.2.3 From be66a8f43172327e3cdde27281e40377cacbb121 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 17:22:29 -0600 Subject: gallium: added PIPE_CAP_TEXTURE_MIRROR_CLAMP, PIPE_CAP_TEXTURE_MIRROR_REPEAT Check for these caps in state tracker and enable corresponding GL extensions if supported. --- src/gallium/drivers/softpipe/sp_screen.c | 4 ++++ src/gallium/include/pipe/p_defines.h | 3 +++ src/mesa/state_tracker/st_extensions.c | 10 ++++++++-- 3 files changed, 15 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index 3f9d4b0ed3..ceb5616b5d 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -72,6 +72,10 @@ softpipe_get_param(struct pipe_screen *screen, int param) return PIPE_MAX_COLOR_BUFS; case PIPE_CAP_OCCLUSION_QUERY: return 1; + case PIPE_CAP_TEXTURE_MIRROR_CLAMP: + return 1; + case PIPE_CAP_TEXTURE_MIRROR_REPEAT: + return 1; case PIPE_CAP_TEXTURE_SHADOW_MAP: return 1; case PIPE_CAP_MAX_TEXTURE_2D_LEVELS: diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index b1d100ef53..bdc6d4ef46 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -281,6 +281,9 @@ enum pipe_texture_target { #define PIPE_CAP_GUARD_BAND_TOP 21 /*< float */ #define PIPE_CAP_GUARD_BAND_RIGHT 22 /*< float */ #define PIPE_CAP_GUARD_BAND_BOTTOM 23 /*< float */ +#define PIPE_CAP_TEXTURE_MIRROR_CLAMP 24 +#define PIPE_CAP_TEXTURE_MIRROR_REPEAT 25 + #ifdef __cplusplus diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index cacf972a1b..b7444b298c 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -134,8 +134,6 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.ARB_texture_env_combine = GL_TRUE; ctx->Extensions.ARB_texture_env_crossbar = GL_TRUE; ctx->Extensions.ARB_texture_env_dot3 = GL_TRUE; - ctx->Extensions.ARB_texture_mirrored_repeat = GL_TRUE; /* XXX temp */ - ctx->Extensions.ARB_vertex_program = GL_TRUE; ctx->Extensions.ARB_vertex_buffer_object = GL_TRUE; @@ -179,6 +177,14 @@ void st_init_extensions(struct st_context *st) ctx->Extensions.ARB_shading_language_120 = GL_TRUE; } + if (screen->get_param(screen, PIPE_CAP_TEXTURE_MIRROR_REPEAT) > 0) { + ctx->Extensions.ARB_texture_mirrored_repeat = GL_TRUE; + } + + if (screen->get_param(screen, PIPE_CAP_TEXTURE_MIRROR_CLAMP) > 0) { + ctx->Extensions.EXT_texture_mirror_clamp = GL_TRUE; + } + if (screen->get_param(screen, PIPE_CAP_NPOT_TEXTURES)) { ctx->Extensions.ARB_texture_non_power_of_two = GL_TRUE; ctx->Extensions.NV_texture_rectangle = GL_TRUE; -- cgit v1.2.3 From a56ccb90c6a374c86158ac2323f844a4003560fa Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 6 Aug 2008 19:15:10 -0600 Subject: gallium: fix clipping/stride bugs in pipe_get_tile_raw(), pipe_put_tile_raw() We need to compute the default dst_stride and src_stride _before_ clipping. After clipping, the width value may have changed. This fixes visible tile glitches in some demos like progs/glsl/texdemo.c --- src/gallium/auxiliary/util/p_tile.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 2b6db43bee..d65e603785 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -51,12 +51,12 @@ pipe_get_tile_raw(struct pipe_surface *ps, { const void *src; - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - if (dst_stride == 0) dst_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + src = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ); assert(src); if(!src) @@ -79,12 +79,12 @@ pipe_put_tile_raw(struct pipe_surface *ps, { void *dst; - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - if (src_stride == 0) src_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + dst = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE); assert(dst); if(!dst) -- cgit v1.2.3 From d50d7a54de89e602a9951264878dfe06924e1adb Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 7 Aug 2008 16:22:34 +0200 Subject: softpipe: Silence compiler warnings on Windows. --- src/gallium/drivers/softpipe/sp_tex_sample.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index ed150527e2..8e39269533 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -601,9 +601,9 @@ get_texel(struct tgsi_sampler *sampler, unsigned face, unsigned level, int x, int y, int z, float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j) { - if (x < 0 || x >= sampler->texture->width[level] || - y < 0 || y >= sampler->texture->height[level] || - z < 0 || z >= sampler->texture->depth[level]) { + if (x < 0 || x >= (int) sampler->texture->width[level] || + y < 0 || y >= (int) sampler->texture->height[level] || + z < 0 || z >= (int) sampler->texture->depth[level]) { rgba[0][j] = sampler->state->border_color[0]; rgba[1][j] = sampler->state->border_color[1]; rgba[2][j] = sampler->state->border_color[2]; -- cgit v1.2.3 From c5c7130385b1b0d35b52a6df68d9558a92fabd86 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 7 Aug 2008 16:23:15 +0200 Subject: mesa: Silence compiler warnings on Windows. --- src/mesa/shader/shader_api.c | 4 ++-- src/mesa/shader/slang/slang_link.c | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index c2c58fffde..2b5d5ec585 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -1493,7 +1493,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program, return; } - if (index + offset > program->Parameters->Size) { + if (index + offset > (GLint) program->Parameters->Size) { /* out of bounds! */ return; } @@ -1557,7 +1557,7 @@ set_program_uniform(GLcontext *ctx, struct gl_program *program, /* if the uniform is bool-valued, convert to 1.0 or 0.0 */ if (is_boolean_type(program->Parameters->Parameters[index].DataType)) { for (i = 0; i < elems; i++) { - uniformVal[i] = uniformVal[i] ? 1.0 : 0.0; + uniformVal[i] = uniformVal[i] ? 1.0f : 0.0f; } } } diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index 8f07406ee4..d96380e03f 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -296,7 +296,7 @@ _slang_count_temporaries(struct gl_program *prog) maxIndex = inst->SrcReg[j].Index; } if (inst->DstReg.File == PROGRAM_TEMPORARY) { - if (maxIndex < inst->DstReg.Index) + if (maxIndex < (GLint) inst->DstReg.Index) maxIndex = inst->DstReg.Index; } } -- cgit v1.2.3 From 9a20cecc5cf41c5eda6cedd5f30416bc1b49ac91 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 7 Aug 2008 16:49:54 +0200 Subject: mesa: Extend GLSL syntax to accept type precision in full type specifiers. --- src/mesa/shader/slang/library/slang_shader.syn | 25 ++++++++++++-------- src/mesa/shader/slang/library/slang_shader_syn.h | 23 +++++++++++-------- src/mesa/shader/slang/slang_compile.c | 29 +++++++++++++++++++----- 3 files changed, 53 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_shader.syn b/src/mesa/shader/slang/library/slang_shader.syn index 4505758520..06bd8ac883 100644 --- a/src/mesa/shader/slang/library/slang_shader.syn +++ b/src/mesa/shader/slang/library/slang_shader.syn @@ -49,7 +49,7 @@ .syntax translation_unit; /* revision number - increment after each change affecting emitted output */ -.emtcode REVISION 3 +.emtcode REVISION 4 /* external declaration (or precision or invariant stmt) */ .emtcode EXTERNAL_NULL 0 @@ -59,9 +59,10 @@ .emtcode INVARIANT_STMT 4 /* precision */ -.emtcode PRECISION_LOW 0 -.emtcode PRECISION_MEDIUM 1 -.emtcode PRECISION_HIGH 2 +.emtcode PRECISION_DEFAULT 0 +.emtcode PRECISION_LOW 1 +.emtcode PRECISION_MEDIUM 2 +.emtcode PRECISION_HIGH 3 /* declaration */ .emtcode DECLARATION_FUNCTION_PROTOTYPE 1 @@ -855,15 +856,21 @@ single_declaration_6 /* ::= | + | + | */ fully_specified_type_space - fully_specified_type_1 .and type_specifier_space; + fully_specified_type_optqual .and fully_specified_type_optprec .and type_specifier_space; fully_specified_type_nospace - fully_specified_type_1 .and type_specifier_nospace; -fully_specified_type_1 - fully_specified_type_2 .or .true .emit TYPE_QUALIFIER_NONE; -fully_specified_type_2 + fully_specified_type_optqual .and fully_specified_type_optprec .and type_specifier_nospace; +fully_specified_type_optqual + fully_specified_type_qual .or .true .emit TYPE_QUALIFIER_NONE; +fully_specified_type_qual type_qualifier .and space; +fully_specified_type_optprec + fully_specified_type_prec .or .true .emit PRECISION_DEFAULT; +fully_specified_type_prec + precision .and space; /* ::= "const" diff --git a/src/mesa/shader/slang/library/slang_shader_syn.h b/src/mesa/shader/slang/library/slang_shader_syn.h index 0c4c14f9d1..1b258fcdc1 100644 --- a/src/mesa/shader/slang/library/slang_shader_syn.h +++ b/src/mesa/shader/slang/library/slang_shader_syn.h @@ -2,15 +2,16 @@ /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE .syn FILE */ ".syntax translation_unit;\n" -".emtcode REVISION 3\n" +".emtcode REVISION 4\n" ".emtcode EXTERNAL_NULL 0\n" ".emtcode EXTERNAL_FUNCTION_DEFINITION 1\n" ".emtcode EXTERNAL_DECLARATION 2\n" ".emtcode DEFAULT_PRECISION 3\n" ".emtcode INVARIANT_STMT 4\n" -".emtcode PRECISION_LOW 0\n" -".emtcode PRECISION_MEDIUM 1\n" -".emtcode PRECISION_HIGH 2\n" +".emtcode PRECISION_DEFAULT 0\n" +".emtcode PRECISION_LOW 1\n" +".emtcode PRECISION_MEDIUM 2\n" +".emtcode PRECISION_HIGH 3\n" ".emtcode DECLARATION_FUNCTION_PROTOTYPE 1\n" ".emtcode DECLARATION_INIT_DECLARATOR_LIST 2\n" ".emtcode FUNCTION_ORDINARY 0\n" @@ -413,13 +414,17 @@ "single_declaration_6\n" " constant_expression .emit VARIABLE_ARRAY_EXPLICIT .or .true .emit VARIABLE_ARRAY_UNKNOWN;\n" "fully_specified_type_space\n" -" fully_specified_type_1 .and type_specifier_space;\n" +" fully_specified_type_optqual .and fully_specified_type_optprec .and type_specifier_space;\n" "fully_specified_type_nospace\n" -" fully_specified_type_1 .and type_specifier_nospace;\n" -"fully_specified_type_1\n" -" fully_specified_type_2 .or .true .emit TYPE_QUALIFIER_NONE;\n" -"fully_specified_type_2\n" +" fully_specified_type_optqual .and fully_specified_type_optprec .and type_specifier_nospace;\n" +"fully_specified_type_optqual\n" +" fully_specified_type_qual .or .true .emit TYPE_QUALIFIER_NONE;\n" +"fully_specified_type_qual\n" " type_qualifier .and space;\n" +"fully_specified_type_optprec\n" +" fully_specified_type_prec .or .true .emit PRECISION_DEFAULT;\n" +"fully_specified_type_prec\n" +" precision .and space;\n" "type_qualifier\n" " \"const\" .emit TYPE_QUALIFIER_CONST .or\n" " .if (shader_type == 2) \"attribute\" .emit TYPE_QUALIFIER_ATTRIBUTE .or\n" diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index f90bbd8c56..2a1af39ee0 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -237,7 +237,7 @@ parse_float(slang_parse_ctx * C, float *number) } /* revision number - increment after each change affecting emitted output */ -#define REVISION 3 +#define REVISION 4 static int check_revision(slang_parse_ctx * C) @@ -691,14 +691,36 @@ parse_type_specifier(slang_parse_ctx * C, slang_output_ctx * O, return 1; } +#define PRECISION_DEFAULT 0 +#define PRECISION_LOW 1 +#define PRECISION_MEDIUM 2 +#define PRECISION_HIGH 3 + static int parse_fully_specified_type(slang_parse_ctx * C, slang_output_ctx * O, slang_fully_specified_type * type) { + GLuint precision; + if (!parse_type_qualifier(C, &type->qualifier)) return 0; + precision = *C->I++; if (!parse_type_specifier(C, O, &type->specifier)) return 0; + switch (precision) { + case PRECISION_DEFAULT: + /* TODO: Grab the default precision for the given type specifier. + */ + break; + case PRECISION_LOW: + case PRECISION_MEDIUM: + case PRECISION_HIGH: + /* TODO: Translate to mesa representation. + */ + break; + default: + return 0; + } return 1; } @@ -1869,11 +1891,6 @@ parse_declaration(slang_parse_ctx * C, slang_output_ctx * O) return 1; } - -#define PRECISION_LOW 0 -#define PRECISION_MEDIUM 1 -#define PRECISION_HIGH 2 - static int parse_default_precision(slang_parse_ctx * C, slang_output_ctx * O) { -- cgit v1.2.3 From 72a5e479789febb552ec783a1cba0ed628dfa427 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 6 Aug 2008 14:48:11 +0100 Subject: gallium: New function to dump surfaces. --- src/gallium/auxiliary/util/p_debug.c | 43 +++++++++++++++++++++++++++++++----- src/gallium/include/pipe/p_debug.h | 8 +++++-- 2 files changed, 44 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index cdc7e66361..a3db13f96d 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -50,12 +50,12 @@ #endif - - #include "pipe/p_compiler.h" #include "pipe/p_util.h" #include "pipe/p_debug.h" #include "pipe/p_format.h" +#include "pipe/p_state.h" +#include "pipe/p_inlines.h" #include "util/u_string.h" @@ -509,7 +509,7 @@ char *pf_sprint_name( char *str, enum pipe_format format ) void debug_dump_image(const char *prefix, unsigned format, unsigned cpp, unsigned width, unsigned height, - unsigned pitch, + unsigned stride, const void *data) { #ifdef PIPE_SUBSYSTEM_WINDOWS_DISPLAY @@ -530,7 +530,7 @@ void debug_dump_image(const char *prefix, for(i = 0; i < sizeof(filename); ++i) wfilename[i] = (WCHAR)filename[i]; - pMap = (unsigned char *)EngMapFile(wfilename, sizeof(header) + cpp*width*height, &iFile); + pMap = (unsigned char *)EngMapFile(wfilename, sizeof(header) + height*width*cpp, &iFile); if(!pMap) return; @@ -542,11 +542,44 @@ void debug_dump_image(const char *prefix, pMap += sizeof(header); for(i = 0; i < height; ++i) { - memcpy(pMap, (unsigned char *)data + cpp*pitch*i, cpp*width); + memcpy(pMap, (unsigned char *)data + stride*i, cpp*width); pMap += cpp*width; } EngUnmapFile(iFile); #endif } + +void debug_dump_surface(const char *prefix, + struct pipe_surface *surface) +{ + unsigned surface_usage; + void *data; + + if (!surface) + goto error1; + + /* XXX: force mappable surface */ + surface_usage = surface->usage; + surface->usage |= PIPE_BUFFER_USAGE_CPU_READ; + + data = pipe_surface_map(surface, + PIPE_BUFFER_USAGE_CPU_READ); + if(!data) + goto error2; + + debug_dump_image(prefix, + surface->format, + surface->block.size, + surface->nblocksx, + surface->nblocksy, + surface->stride, + data); + + pipe_surface_unmap(surface); +error2: + surface->usage = surface_usage; +error1: + ; +} #endif diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h index a5816c3cbb..6478ae2f08 100644 --- a/src/gallium/include/pipe/p_debug.h +++ b/src/gallium/include/pipe/p_debug.h @@ -332,13 +332,17 @@ debug_profile_stop(void); #ifdef DEBUG +struct pipe_surface; void debug_dump_image(const char *prefix, unsigned format, unsigned cpp, unsigned width, unsigned height, - unsigned pitch, + unsigned stride, const void *data); +void debug_dump_surface(const char *prefix, + struct pipe_surface *surface); #else -#define debug_dump_image(prefix, format, cpp, width, height, pitch, data) ((void)0) +#define debug_dump_image(prefix, format, cpp, width, height, stride, data) ((void)0) +#define debug_dump_surface(prefix, surface) ((void)0) #endif -- cgit v1.2.3 From 1ee500ac73fa657f02321c46cf5d9a4bfdea54de Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 6 Aug 2008 14:51:15 +0100 Subject: pipebuffer: Add an extra assertion to ensure buffers do not jump between lists. --- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index f599bee07e..ce41418a0f 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -142,12 +142,13 @@ _fenced_buffer_destroy(struct fenced_buffer *fenced_buf) static INLINE void -_fenced_buffer_remove(struct fenced_buffer *fenced_buf) +_fenced_buffer_remove(struct fenced_buffer_list *fenced_list, + struct fenced_buffer *fenced_buf) { - struct fenced_buffer_list *fenced_list = fenced_buf->list; struct pipe_winsys *winsys = fenced_list->winsys; assert(fenced_buf->fence); + assert(fenced_buf->list == fenced_list); winsys->fence_reference(winsys, &fenced_buf->fence, NULL); fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE; @@ -184,7 +185,8 @@ _fenced_buffer_finish(struct fenced_buffer *fenced_buf) return PIPE_ERROR; } /* Remove from the fenced list */ - _fenced_buffer_remove(fenced_buf); /* TODO: remove consequents */ + /* TODO: remove consequents */ + _fenced_buffer_remove(fenced_list, fenced_buf); } fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE; @@ -223,7 +225,7 @@ _fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list, assert(winsys->fence_signalled(winsys, fenced_buf->fence, 0) == 0); } - _fenced_buffer_remove(fenced_buf); + _fenced_buffer_remove(fenced_list, fenced_buf); curr = next; next = curr->next; @@ -248,7 +250,7 @@ fenced_buffer_destroy(struct pb_buffer *buf) do { fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head); assert(winsys->fence_signalled(winsys, fenced_buf->fence, 0) == 0); - _fenced_buffer_remove(fenced_buf); + _fenced_buffer_remove(fenced_list, fenced_buf); curr = prev; prev = curr->prev; } while (curr != &fenced_list->delayed); @@ -395,7 +397,7 @@ buffer_fence(struct pb_buffer *buf, _glthread_LOCK_MUTEX(fenced_list->mutex); if (fenced_buf->fence) - _fenced_buffer_remove(fenced_buf); + _fenced_buffer_remove(fenced_list, fenced_buf); if (fence) { winsys->fence_reference(winsys, &fenced_buf->fence, fence); fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE; -- cgit v1.2.3 From 35355f7610b69dcd2fdba451db4554478fe0acaa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 6 Aug 2008 21:36:25 +0100 Subject: trace: New pipe driver to trace incoming calls. Only pipe_screen calls traced, and only linux supported, for now. --- src/gallium/drivers/trace/README | 37 ++++ src/gallium/drivers/trace/SConscript | 15 ++ src/gallium/drivers/trace/tr_context.c | 41 ++++ src/gallium/drivers/trace/tr_context.h | 58 +++++ src/gallium/drivers/trace/tr_dump.c | 329 ++++++++++++++++++++++++++++ src/gallium/drivers/trace/tr_dump.h | 102 +++++++++ src/gallium/drivers/trace/tr_screen.c | 385 +++++++++++++++++++++++++++++++++ src/gallium/drivers/trace/tr_screen.h | 65 ++++++ src/gallium/drivers/trace/tr_state.c | 72 ++++++ src/gallium/drivers/trace/tr_state.h | 44 ++++ src/gallium/drivers/trace/tr_stream.c | 100 +++++++++ src/gallium/drivers/trace/tr_stream.h | 52 +++++ 12 files changed, 1300 insertions(+) create mode 100644 src/gallium/drivers/trace/README create mode 100644 src/gallium/drivers/trace/SConscript create mode 100644 src/gallium/drivers/trace/tr_context.c create mode 100644 src/gallium/drivers/trace/tr_context.h create mode 100644 src/gallium/drivers/trace/tr_dump.c create mode 100644 src/gallium/drivers/trace/tr_dump.h create mode 100644 src/gallium/drivers/trace/tr_screen.c create mode 100644 src/gallium/drivers/trace/tr_screen.h create mode 100644 src/gallium/drivers/trace/tr_state.c create mode 100644 src/gallium/drivers/trace/tr_state.h create mode 100644 src/gallium/drivers/trace/tr_stream.c create mode 100644 src/gallium/drivers/trace/tr_stream.h (limited to 'src') diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README new file mode 100644 index 0000000000..81da610bd5 --- /dev/null +++ b/src/gallium/drivers/trace/README @@ -0,0 +1,37 @@ +This directory contains a Gallium3D pipe driver which traces all incoming calls. + +To build, invoke scons on the top dir as + + scons statetrackers=mesa drivers=softpipe,i915simple,trace winsys=xlib + +To use do + + ln -s libGL.so build/linux-x86-debug/gallium/winsys/xlib/libGL.so.1 + export LD_LIBRARY_PATH=$PWD/build/linux-x86-debug/gallium/winsys/xlib + export GALLIUM_TRACE=y + +ensure the right libGL.so is being picked by doing + + ldd `which glxingo` + +and then try running + + glxinfo + +which should create a gallium.*.trace file, which is an XML file. You can view +copying trace.xsl and trace.css to the same directory, and opening with a +XSLT capable browser like Firefox or Internet Explorer. It often happens that +the trace file was not properly terminated, and a + + + +closing tag is missing from the file end. Add it before try to open or +further transform it by doing + + echo '' >> gallium.??.trace + + +This is still work in progress. + +-- +Jose Fonseca diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript new file mode 100644 index 0000000000..30225b5a54 --- /dev/null +++ b/src/gallium/drivers/trace/SConscript @@ -0,0 +1,15 @@ +Import('*') + +env = env.Clone() + +trace = env.ConvenienceLibrary( + target = 'trace', + source = [ + 'tr_context.c', + 'tr_dump.c', + 'tr_screen.c', + 'tr_state.c', + 'tr_stream.c', + ]) + +Export('trace') \ No newline at end of file diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c new file mode 100644 index 0000000000..67a67b3da6 --- /dev/null +++ b/src/gallium/drivers/trace/tr_context.c @@ -0,0 +1,41 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" + +#include "tr_stream.h" +#include "tr_dump.h" +#include "tr_state.h" +#include "tr_context.h" + + +struct pipe_context * +trace_context_create(struct pipe_context *pipe) +{ + /* TODO */ + return pipe; +} diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h new file mode 100644 index 0000000000..75b8d11a7a --- /dev/null +++ b/src/gallium/drivers/trace/tr_context.h @@ -0,0 +1,58 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TR_CONTEXT_H_ +#define TR_CONTEXT_H_ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_context.h" + + +struct trace_context +{ + struct pipe_context base; + + /* TODO */ +}; + + +static INLINE struct trace_context * +trace_context(struct pipe_context *context) +{ + assert(context); + return (struct trace_context *)context; +} + + + +struct pipe_context * +trace_context_create(struct pipe_context *context); + + +#endif /* TR_CONTEXT_H_ */ diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c new file mode 100644 index 0000000000..30c0986363 --- /dev/null +++ b/src/gallium/drivers/trace/tr_dump.c @@ -0,0 +1,329 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * @file + * Trace dumping functions. + * + * For now we just use standard XML for dumping the trace calls, as this is + * simple to write, parse, and visually inspect, but the actual representation + * is abstracted out of this file, so that we can switch to a binary + * representation if/when it becomes justified. + * + * @author Jose Fonseca + */ + + +#include "pipe/p_compiler.h" +#include "util/u_string.h" + +#include "tr_stream.h" +#include "tr_dump.h" + + +static INLINE void +trace_dump_write(struct trace_stream *stream, const char *s) +{ + trace_stream_write(stream, s, strlen(s)); +} + + +static INLINE void +trace_dump_writef(struct trace_stream *stream, const char *format, ...) +{ + char buf[1024]; + va_list ap; + va_start(ap, format); + util_vsnprintf(buf, sizeof(buf), format, ap); + va_end(ap); + trace_dump_write(stream, buf); +} + + +static INLINE void +trace_dump_escape(struct trace_stream *stream, const char *str) +{ + const unsigned char *p = (const unsigned char *)str; + unsigned char c; + while((c = *p++) != 0) { + if(c == '<') + trace_dump_write(stream, "<"); + else if(c == '>') + trace_dump_write(stream, ">"); + else if(c == '&') + trace_dump_write(stream, "&"); + else if(c == '\'') + trace_dump_write(stream, "'"); + else if(c == '\"') + trace_dump_write(stream, """); + else if(c >= 0x20 && c <= 0x7e) + trace_dump_writef(stream, "%c", c); + else + trace_dump_writef(stream, "&#%u;", c); + } +} + + +static INLINE void +trace_dump_indent(struct trace_stream *stream, unsigned level) +{ + unsigned i; + for(i = 0; i < level; ++i) + trace_dump_write(stream, "\t"); +} + + +static INLINE void +trace_dump_newline(struct trace_stream *stream) +{ + trace_dump_write(stream, "\n"); +} + + +static INLINE void +trace_dump_tag(struct trace_stream *stream, + const char *name) +{ + trace_dump_write(stream, "<"); + trace_dump_write(stream, name); + trace_dump_write(stream, "/>"); +} + + +static INLINE void +trace_dump_tag_begin(struct trace_stream *stream, + const char *name) +{ + trace_dump_write(stream, "<"); + trace_dump_write(stream, name); + trace_dump_write(stream, ">"); +} + +static INLINE void +trace_dump_tag_begin1(struct trace_stream *stream, + const char *name, + const char *attr1, const char *value1) +{ + trace_dump_write(stream, "<"); + trace_dump_write(stream, name); + trace_dump_write(stream, " "); + trace_dump_write(stream, attr1); + trace_dump_write(stream, "='"); + trace_dump_escape(stream, value1); + trace_dump_write(stream, "'>"); +} + + +static INLINE void +trace_dump_tag_begin2(struct trace_stream *stream, + const char *name, + const char *attr1, const char *value1, + const char *attr2, const char *value2) +{ + trace_dump_write(stream, "<"); + trace_dump_write(stream, name); + trace_dump_write(stream, " "); + trace_dump_write(stream, attr1); + trace_dump_write(stream, "=\'"); + trace_dump_escape(stream, value1); + trace_dump_write(stream, "\' "); + trace_dump_write(stream, attr2); + trace_dump_write(stream, "=\'"); + trace_dump_escape(stream, value2); + trace_dump_write(stream, "\'>"); +} + + +static INLINE void +trace_dump_tag_begin3(struct trace_stream *stream, + const char *name, + const char *attr1, const char *value1, + const char *attr2, const char *value2, + const char *attr3, const char *value3) +{ + trace_dump_write(stream, "<"); + trace_dump_write(stream, name); + trace_dump_write(stream, " "); + trace_dump_write(stream, attr1); + trace_dump_write(stream, "=\'"); + trace_dump_escape(stream, value1); + trace_dump_write(stream, "\' "); + trace_dump_write(stream, attr2); + trace_dump_write(stream, "=\'"); + trace_dump_escape(stream, value2); + trace_dump_write(stream, "\' "); + trace_dump_write(stream, attr3); + trace_dump_write(stream, "=\'"); + trace_dump_escape(stream, value3); + trace_dump_write(stream, "\'>"); +} + + +static INLINE void +trace_dump_tag_end(struct trace_stream *stream, + const char *name) +{ + trace_dump_write(stream, ""); +} + + +void trace_dump_trace_begin(struct trace_stream *stream, + unsigned version) +{ + trace_dump_write(stream, "\n"); + trace_dump_write(stream, "\n"); + trace_dump_writef(stream, "\n", version); +} + + +void trace_dump_trace_end(struct trace_stream *stream) +{ + trace_dump_write(stream, "\n"); +} + +void trace_dump_call_begin(struct trace_stream *stream, + const char *klass, const char *method) +{ + trace_dump_indent(stream, 1); + trace_dump_tag_begin2(stream, "call", "class", klass, "method", method); + trace_dump_newline(stream); +} + +void trace_dump_call_end(struct trace_stream *stream) +{ + trace_dump_indent(stream, 1); + trace_dump_tag_end(stream, "call"); + trace_dump_newline(stream); +} + +void trace_dump_arg_begin(struct trace_stream *stream, + const char *name) +{ + trace_dump_indent(stream, 2); + trace_dump_tag_begin1(stream, "arg", "name", name); +} + +void trace_dump_arg_end(struct trace_stream *stream) +{ + trace_dump_tag_end(stream, "arg"); + trace_dump_newline(stream); +} + +void trace_dump_ret_begin(struct trace_stream *stream) +{ + trace_dump_indent(stream, 2); + trace_dump_tag_begin(stream, "ret"); +} + +void trace_dump_ret_end(struct trace_stream *stream) +{ + trace_dump_tag_end(stream, "ret"); + trace_dump_newline(stream); +} + +void trace_dump_bool(struct trace_stream *stream, + int value) +{ + trace_dump_writef(stream, "%c", value ? '1' : '0'); +} + +void trace_dump_int(struct trace_stream *stream, + long int value) +{ + trace_dump_writef(stream, "%li", value); +} + +void trace_dump_uint(struct trace_stream *stream, + long unsigned value) +{ + trace_dump_writef(stream, "%lu", value); +} + +void trace_dump_float(struct trace_stream *stream, + double value) +{ + trace_dump_writef(stream, "%g", value); +} + +void trace_dump_string(struct trace_stream *stream, + const char *str) +{ + trace_dump_write(stream, ""); + trace_dump_escape(stream, str); + trace_dump_write(stream, ""); +} + +void trace_dump_array_begin(struct trace_stream *stream) +{ + trace_dump_write(stream, ""); +} + +void trace_dump_array_end(struct trace_stream *stream) +{ + trace_dump_write(stream, ""); +} + +void trace_dump_elem_begin(struct trace_stream *stream) +{ + trace_dump_write(stream, ""); +} + +void trace_dump_elem_end(struct trace_stream *stream) +{ + trace_dump_write(stream, ""); +} + +void trace_dump_struct_begin(struct trace_stream *stream, + const char *name) +{ + trace_dump_writef(stream, "", name); +} + +void trace_dump_struct_end(struct trace_stream *stream) +{ + trace_dump_write(stream, ""); +} + +void trace_dump_member_begin(struct trace_stream *stream, + const char *name) +{ + trace_dump_writef(stream, "", name); +} + +void trace_dump_member_end(struct trace_stream *stream) +{ + trace_dump_write(stream, ""); +} + +void trace_dump_ptr(struct trace_stream *stream, + const void *value) +{ + trace_dump_writef(stream, "%p", value); +} diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h new file mode 100644 index 0000000000..f086486591 --- /dev/null +++ b/src/gallium/drivers/trace/tr_dump.h @@ -0,0 +1,102 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation streams (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Trace data dumping primitives. + */ + +#ifndef TR_DUMP_H +#define TR_DUMP_H + + +struct trace_stream; + + +void trace_dump_trace_begin(struct trace_stream *stream, unsigned version); +void trace_dump_trace_end(struct trace_stream *stream); +void trace_dump_call_begin(struct trace_stream *stream, const char *klass, const char *method); +void trace_dump_call_end(struct trace_stream *stream); +void trace_dump_arg_begin(struct trace_stream *stream, const char *name); +void trace_dump_arg_end(struct trace_stream *stream); +void trace_dump_ret_begin(struct trace_stream *stream); +void trace_dump_ret_end(struct trace_stream *stream); +void trace_dump_bool(struct trace_stream *stream, int value); +void trace_dump_int(struct trace_stream *stream, long int value); +void trace_dump_uint(struct trace_stream *stream, long unsigned value); +void trace_dump_float(struct trace_stream *stream, double value); +void trace_dump_string(struct trace_stream *stream, const char *str); +void trace_dump_array_begin(struct trace_stream *stream); +void trace_dump_array_end(struct trace_stream *stream); +void trace_dump_elem_begin(struct trace_stream *stream); +void trace_dump_elem_end(struct trace_stream *stream); +void trace_dump_struct_begin(struct trace_stream *stream, const char *name); +void trace_dump_struct_end(struct trace_stream *stream); +void trace_dump_member_begin(struct trace_stream *stream, const char *name); +void trace_dump_member_end(struct trace_stream *stream); +void trace_dump_ptr(struct trace_stream *stream, const void *value); + + +/* + * Code saving macros. + */ + +#define trace_dump_arg(_stream, _type, _arg) \ + do { \ + trace_dump_arg_begin(_stream, #_arg); \ + trace_dump_##_type(_stream, _arg); \ + trace_dump_arg_end(_stream); \ + } while(0) + +#define trace_dump_ret(_stream, _type, _arg) \ + do { \ + trace_dump_ret_begin(_stream); \ + trace_dump_##_type(_stream, _arg); \ + trace_dump_ret_end(_stream); \ + } while(0) + +#define trace_dump_array(_stream, _type, _obj, _size) \ + do { \ + unsigned long idx; \ + trace_dump_array_begin(_stream); \ + for(idx = 0; idx < _size; ++idx) { \ + trace_dump_elem_begin(_stream); \ + trace_dump_##_type(_stream, _obj[idx]); \ + trace_dump_elem_end(_stream); \ + } \ + trace_dump_array_end(_stream); \ + } while(0) + +#define trace_dump_member(_stream, _type, _obj, _member) \ + do { \ + trace_dump_member_begin(_stream, #_member); \ + trace_dump_##_type(_stream, _obj->_member); \ + trace_dump_member_end(_stream); \ + } while(0) + + +#endif /* TR_DUMP_H */ diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c new file mode 100644 index 0000000000..20a2026e1d --- /dev/null +++ b/src/gallium/drivers/trace/tr_screen.c @@ -0,0 +1,385 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" + +#include "tr_stream.h" +#include "tr_dump.h" +#include "tr_state.h" +#include "tr_screen.h" + + +static const char * +trace_screen_get_name(struct pipe_screen *_screen) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + const char *result; + + trace_dump_call_begin(stream, "pipe_screen", "get_name"); + + trace_dump_arg(stream, ptr, screen); + + result = screen->get_name(screen); + + trace_dump_ret(stream, string, result); + + trace_dump_call_end(stream); + + return result; +} + + +static const char * +trace_screen_get_vendor(struct pipe_screen *_screen) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + const char *result; + + trace_dump_call_begin(stream, "pipe_screen", "get_vendor"); + + trace_dump_arg(stream, ptr, screen); + + result = screen->get_vendor(screen); + + trace_dump_ret(stream, string, result); + + trace_dump_call_end(stream); + + return result; +} + + +static int +trace_screen_get_param(struct pipe_screen *_screen, + int param) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + int result; + + trace_dump_call_begin(stream, "pipe_screen", "get_param"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, int, param); + + result = screen->get_param(screen, param); + + trace_dump_ret(stream, int, result); + + trace_dump_call_end(stream); + + return result; +} + + +static float +trace_screen_get_paramf(struct pipe_screen *_screen, + int param) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + float result; + + trace_dump_call_begin(stream, "pipe_screen", "get_paramf"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, int, param); + + result = screen->get_paramf(screen, param); + + trace_dump_ret(stream, float, result); + + trace_dump_call_end(stream); + + return result; +} + + +static boolean +trace_screen_is_format_supported(struct pipe_screen *_screen, + enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + boolean result; + + trace_dump_call_begin(stream, "pipe_screen", "is_format_supported"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, int, format); + trace_dump_arg(stream, int, target); + trace_dump_arg(stream, uint, tex_usage); + trace_dump_arg(stream, uint, geom_flags); + + result = screen->is_format_supported(screen, format, target, tex_usage, geom_flags); + + trace_dump_ret(stream, bool, result); + + trace_dump_call_end(stream); + + return result; +} + + +static struct pipe_texture * +trace_screen_texture_create(struct pipe_screen *_screen, + const struct pipe_texture *templat) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + struct pipe_texture *result; + + trace_dump_call_begin(stream, "pipe_screen", "texture_create"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, template, templat); + + result = screen->texture_create(screen, templat); + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static struct pipe_texture * +trace_screen_texture_blanket(struct pipe_screen *_screen, + const struct pipe_texture *templat, + const unsigned *ppitch, + struct pipe_buffer *buffer) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + unsigned pitch = *ppitch; + struct pipe_texture *result; + + trace_dump_call_begin(stream, "pipe_screen", "texture_blanket"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, template, templat); + trace_dump_arg(stream, uint, pitch); + trace_dump_arg(stream, ptr, buffer); + + result = screen->texture_blanket(screen, templat, ppitch, buffer); + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static void +trace_screen_texture_release(struct pipe_screen *_screen, + struct pipe_texture **ptexture) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + struct pipe_texture *texture = *ptexture; + + trace_dump_call_begin(stream, "pipe_screen", "texture_release"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, ptr, texture); + + screen->texture_release(screen, ptexture); + + trace_dump_call_end(stream); +} + +static struct pipe_surface * +trace_screen_get_tex_surface(struct pipe_screen *_screen, + struct pipe_texture *texture, + unsigned face, unsigned level, + unsigned zslice, + unsigned usage) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + struct pipe_surface *result; + + trace_dump_call_begin(stream, "pipe_screen", "get_tex_surface"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, ptr, texture); + trace_dump_arg(stream, uint, face); + trace_dump_arg(stream, uint, level); + trace_dump_arg(stream, uint, zslice); + trace_dump_arg(stream, uint, usage); + + result = screen->get_tex_surface(screen, texture, face, level, zslice, usage); + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static void +trace_screen_tex_surface_release(struct pipe_screen *_screen, + struct pipe_surface **psurface) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + struct pipe_surface *surface = *psurface; + + trace_dump_call_begin(stream, "pipe_screen", "tex_surface_release"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, ptr, surface); + + screen->tex_surface_release(screen, psurface); + + trace_dump_call_end(stream); +} + + +static void * +trace_screen_surface_map(struct pipe_screen *_screen, + struct pipe_surface *surface, + unsigned flags) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + struct pipe_surface *result; + + trace_dump_call_begin(stream, "pipe_screen", "surface_map"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, ptr, surface); + trace_dump_arg(stream, uint, flags); + + result = screen->surface_map(screen, surface, flags); + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static void +trace_screen_surface_unmap(struct pipe_screen *_screen, + struct pipe_surface *surface) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + + trace_dump_call_begin(stream, "pipe_screen", "surface_unmap"); + + trace_dump_arg(stream, ptr, screen); + trace_dump_arg(stream, ptr, surface); + + screen->surface_unmap(screen, surface); + + trace_dump_call_end(stream); +} + + +static void +trace_screen_destroy(struct pipe_screen *_screen) +{ + struct trace_screen *tr_scr = trace_screen(_screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_screen *screen = tr_scr->screen; + + trace_dump_call_begin(stream, "pipe_screen", "destroy"); + + trace_dump_arg(stream, ptr, screen); + + screen->destroy(screen); + + trace_dump_call_end(stream); + + trace_dump_trace_end(stream); + + trace_stream_close(tr_scr->stream); + + FREE(tr_scr); +} + + +struct pipe_screen * +trace_screen_create(struct pipe_screen *screen) +{ + struct trace_screen *tr_scr; + + if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) + return screen; + + tr_scr = CALLOC_STRUCT(trace_screen); + if(!tr_scr) + return NULL; + + tr_scr->base.winsys = screen->winsys; + tr_scr->base.destroy = trace_screen_destroy; + tr_scr->base.get_name = trace_screen_get_name; + tr_scr->base.get_vendor = trace_screen_get_vendor; + tr_scr->base.get_param = trace_screen_get_param; + tr_scr->base.get_paramf = trace_screen_get_paramf; + tr_scr->base.is_format_supported = trace_screen_is_format_supported; + tr_scr->base.texture_create = trace_screen_texture_create; + tr_scr->base.texture_blanket = trace_screen_texture_blanket; + tr_scr->base.texture_release = trace_screen_texture_release; + tr_scr->base.get_tex_surface = trace_screen_get_tex_surface; + tr_scr->base.tex_surface_release = trace_screen_tex_surface_release; + tr_scr->base.surface_map = trace_screen_surface_map; + tr_scr->base.surface_unmap = trace_screen_surface_unmap; + + tr_scr->screen = screen; + + tr_scr->stream = trace_stream_create("gallium", "trace"); + if(!tr_scr->stream) + return NULL; + + trace_dump_trace_begin(tr_scr->stream, 0); + + return &tr_scr->base; +} diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h new file mode 100644 index 0000000000..5d7d667f71 --- /dev/null +++ b/src/gallium/drivers/trace/tr_screen.h @@ -0,0 +1,65 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TR_SCREEN_H_ +#define TR_SCREEN_H_ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_screen.h" + + +struct trace_stream; + + +struct trace_screen +{ + struct pipe_screen base; + + struct pipe_screen *screen; + + struct trace_stream *stream; + + unsigned event_no; +}; + + +static INLINE struct trace_screen * +trace_screen(struct pipe_screen *screen) +{ + assert(screen); + return (struct trace_screen *)screen; +} + + + +struct pipe_screen * +trace_screen_create(struct pipe_screen *screen); + + +#endif /* TR_SCREEN_H_ */ diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c new file mode 100644 index 0000000000..1e0e87f7fd --- /dev/null +++ b/src/gallium/drivers/trace/tr_state.c @@ -0,0 +1,72 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" + +#include "tr_dump.h" +#include "tr_state.h" + + +void trace_dump_block(struct trace_stream *stream, + const struct pipe_format_block *block) +{ + trace_dump_struct_begin(stream, "pipe_format_block"); + trace_dump_member(stream, uint, block, size); + trace_dump_member(stream, uint, block, width); + trace_dump_member(stream, uint, block, height); + trace_dump_struct_end(stream); +} + + +void trace_dump_template(struct trace_stream *stream, + const struct pipe_texture *templat) +{ + trace_dump_struct_begin(stream, "pipe_texture"); + + trace_dump_member(stream, int, templat, target); + trace_dump_member(stream, int, templat, format); + + trace_dump_member_begin(stream, "width"); + trace_dump_array(stream, uint, templat->width, 1); + trace_dump_member_end(stream); + + trace_dump_member_begin(stream, "height"); + trace_dump_array(stream, uint, templat->height, 1); + trace_dump_member_end(stream); + + trace_dump_member_begin(stream, "block"); + trace_dump_block(stream, &templat->block); + trace_dump_member_end(stream); + + trace_dump_member(stream, uint, templat, last_level); + trace_dump_member(stream, uint, templat, tex_usage); + + trace_dump_struct_end(stream); +} + diff --git a/src/gallium/drivers/trace/tr_state.h b/src/gallium/drivers/trace/tr_state.h new file mode 100644 index 0000000000..e52524ee90 --- /dev/null +++ b/src/gallium/drivers/trace/tr_state.h @@ -0,0 +1,44 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation streams (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TR_STATE_H +#define TR_STATE_H + + +struct trace_stream; + +struct pipe_format_block; +struct pipe_texture; + +void trace_dump_block(struct trace_stream *stream, + const struct pipe_format_block *block); + +void trace_dump_template(struct trace_stream *stream, + const struct pipe_texture *templat); + + +#endif /* TR_STATE_H */ diff --git a/src/gallium/drivers/trace/tr_stream.c b/src/gallium/drivers/trace/tr_stream.c new file mode 100644 index 0000000000..14cc257e15 --- /dev/null +++ b/src/gallium/drivers/trace/tr_stream.c @@ -0,0 +1,100 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_config.h" + +#if defined(PIPE_OS_LINUX) +#include +#else +#error Unsupported platform +#endif + +#include "pipe/p_util.h" + +#include "tr_stream.h" + + +struct trace_stream +{ +#if defined(PIPE_OS_LINUX) + FILE *file; +#endif +}; + + +struct trace_stream * +trace_stream_create(const char *name, const char *ext) +{ + struct trace_stream *stream; + static unsigned file_no = 0; + char filename[1024]; + + stream = CALLOC_STRUCT(trace_stream); + if(!stream) + goto error1; + + snprintf(filename, sizeof(filename), "%s.%u.%s", name, file_no, ext); + +#if defined(PIPE_OS_LINUX) + stream->file = fopen(filename, "w"); + if(!stream->file) + goto error2; +#endif + + return stream; + +error2: + FREE(stream); +error1: + return NULL; +} + + +boolean +trace_stream_write(struct trace_stream *stream, const void *data, size_t size) +{ + if(!stream) + return FALSE; + +#if defined(PIPE_OS_LINUX) + return fwrite(data, size, 1, stream->file) == size ? TRUE : FALSE; +#endif +} + + +void +trace_stream_close(struct trace_stream *stream) +{ + if(!stream) + return; + +#if defined(PIPE_OS_LINUX) + fclose(stream->file); +#endif + + FREE(stream); +} diff --git a/src/gallium/drivers/trace/tr_stream.h b/src/gallium/drivers/trace/tr_stream.h new file mode 100644 index 0000000000..d50fed2691 --- /dev/null +++ b/src/gallium/drivers/trace/tr_stream.h @@ -0,0 +1,52 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation streams (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Cross-platform sequential access stream abstraction. + */ + +#ifndef TR_STREAM_H +#define TR_STREAM_H + + +#include "pipe/p_compiler.h" + +struct trace_stream; + + +struct trace_stream * +trace_stream_create(const char *name, const char *ext); + +boolean +trace_stream_write(struct trace_stream *stream, const void *data, size_t size); + +void +trace_stream_close(struct trace_stream *stream); + + +#endif /* TR_STREAM_H */ -- cgit v1.2.3 From 18fb8f148679ddddc5a781014d36a51afc304727 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 6 Aug 2008 21:37:00 +0100 Subject: xlib: Integrate with the trace pipe driver. --- src/gallium/winsys/xlib/SConscript | 54 ++++++++++++++++++++----------------- src/gallium/winsys/xlib/xm_winsys.c | 13 +++++++++ 2 files changed, 42 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/SConscript b/src/gallium/winsys/xlib/SConscript index 14a85ae0f2..8650f595a7 100644 --- a/src/gallium/winsys/xlib/SConscript +++ b/src/gallium/winsys/xlib/SConscript @@ -9,31 +9,35 @@ if env['platform'] == 'linux' \ and 'i965simple' in env['drivers'] \ and not env['dri']: - env = env.Clone() + env = env.Clone() - env.Append(CPPPATH = [ - '#/src/mesa', - '#/src/mesa/main', - ]) + env.Append(CPPPATH = [ + '#/src/mesa', + '#/src/mesa/main', + ]) - sources = [ - 'glxapi.c', - 'fakeglx.c', - 'xfonts.c', - 'xm_api.c', - 'xm_winsys.c', - 'xm_winsys_aub.c', - 'brw_aub.c', - ] - - drivers = [ - softpipe, - i965simple - ] + sources = [ + 'glxapi.c', + 'fakeglx.c', + 'xfonts.c', + 'xm_api.c', + 'xm_winsys.c', + 'xm_winsys_aub.c', + 'brw_aub.c', + ] + + drivers = [ + softpipe, + i965simple, + ] + + if 'trace' in env['drivers']: + env.Append(CPPDEFINES = 'GALLIUM_TRACE') + drivers += [trace] - # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions - env.SharedLibrary( - target ='GL', - source = sources, - LIBS = glapi + mesa + drivers + auxiliaries + env['LIBS'], - ) + # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions + env.SharedLibrary( + target ='GL', + source = sources, + LIBS = glapi + mesa + drivers + auxiliaries + env['LIBS'], + ) diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 9225ee510d..5a01b6167b 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -54,6 +54,11 @@ #define TILE_SIZE 32 /* avoid compilation errors */ #endif +#ifdef GALLIUM_TRACE +#include "trace/tr_screen.h" +#include "trace/tr_context.h" +#endif + #include "xm_winsys_aub.h" @@ -674,7 +679,15 @@ xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) { struct pipe_screen *screen = softpipe_create_screen(pws); +#ifdef GALLIUM_TRACE + screen = trace_screen_create(screen); +#endif + pipe = softpipe_create(screen, pws, NULL); + +#ifdef GALLIUM_TRACE + pipe = trace_context_create(pipe); +#endif } if (pipe) -- cgit v1.2.3 From f6e0514736bb763813a49ae5542b2c8648641595 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 6 Aug 2008 21:39:25 +0100 Subject: trace: Add missing XSL and CSS. --- src/gallium/drivers/trace/trace.css | 72 +++++++++++++++++++++++ src/gallium/drivers/trace/trace.xsl | 114 ++++++++++++++++++++++++++++++++++++ 2 files changed, 186 insertions(+) create mode 100644 src/gallium/drivers/trace/trace.css create mode 100644 src/gallium/drivers/trace/trace.xsl (limited to 'src') diff --git a/src/gallium/drivers/trace/trace.css b/src/gallium/drivers/trace/trace.css new file mode 100644 index 0000000000..d92c433791 --- /dev/null +++ b/src/gallium/drivers/trace/trace.css @@ -0,0 +1,72 @@ +/**************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc. + * + * This program is free software: you can redistribute it and/or modify it + * under the terms of the GNU Lesser General Public License as published + * by the Free Software Foundation, either version 3 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public License + * along with this program. If not, see . + * + ****************************************************************************/ + +body { + font-family: verdana, sans-serif; + font-size: 11px; + font-weight: normal; + text-align : left; +} + +ul.calls { + list-style: none; + margin-left: 0px; + padding-left: 0px; +} + +ul.args { + display:inline; + list-style: none; + margin-left: 0px; + padding-left: 0px; +} + +ul.args li { + display:inline; +} + +ul.elems { + list-style: none; + margin-left: 2em; + padding-left: 0px; +} + +ul.elems li { + display:block; +} + +.fun { + font-weight: bold; +} + +.var { + font-style: italic; +} + +.typ { + display: none; +} + +.lit { + color: #0000ff; +} + +.ptr { + color: #008000; +} diff --git a/src/gallium/drivers/trace/trace.xsl b/src/gallium/drivers/trace/trace.xsl new file mode 100644 index 0000000000..ace4020cf9 --- /dev/null +++ b/src/gallium/drivers/trace/trace.xsl @@ -0,0 +1,114 @@ + + + + + + + + + + + + + + Gallium Trace + + + +
    + +
+ + + + + +
  • + + + :: + + + ( +
      + +
    + ) + +
    • +     + + +
  • + + = + + + , + +
    • +     + + + = + + + + + + + + + + + + " + + " + + + + + { + + } + + + +
  • + + + , + +
    • +     + + + + + + + + + + + + +     -- cgit v1.2.3 From f2e19c34e06dfc33557a481f764fc75a5aef15ff Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 13:20:02 +0100 Subject: trace: Trace pipe context calls. --- src/gallium/drivers/trace/tr_context.c | 1017 +++++++++++++++++++++++++++++++- src/gallium/drivers/trace/tr_context.h | 10 +- src/gallium/drivers/trace/tr_dump.h | 7 + 3 files changed, 1027 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index 67a67b3da6..6161e35fa5 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -30,12 +30,1025 @@ #include "tr_stream.h" #include "tr_dump.h" #include "tr_state.h" +#include "tr_screen.h" #include "tr_context.h" +static INLINE void +trace_context_set_edgeflags(struct pipe_context *_pipe, + const unsigned *bitfield) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_edgeflags"); + + trace_dump_arg(stream, ptr, pipe); + /* FIXME: we don't know how big this array is */ + trace_dump_arg(stream, ptr, bitfield); + + pipe->set_edgeflags(pipe, bitfield);; + + trace_dump_call_end(stream); +} + + +static INLINE boolean +trace_context_draw_arrays(struct pipe_context *_pipe, + unsigned mode, unsigned start, unsigned count) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + boolean result; + + trace_dump_call_begin(stream, "pipe_context", "draw_arrays"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, uint, mode); + trace_dump_arg(stream, uint, start); + trace_dump_arg(stream, uint, count); + + result = pipe->draw_arrays(pipe, mode, start, count);; + + trace_dump_ret(stream, bool, result); + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE boolean +trace_context_draw_elements(struct pipe_context *_pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + boolean result; + + trace_dump_call_begin(stream, "pipe_context", "draw_elements"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, indexBuffer); + trace_dump_arg(stream, uint, indexSize); + trace_dump_arg(stream, uint, mode); + trace_dump_arg(stream, uint, start); + trace_dump_arg(stream, uint, count); + + result = pipe->draw_elements(pipe, indexBuffer, indexSize, mode, start, count);; + + trace_dump_ret(stream, bool, result); + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE boolean +trace_context_draw_range_elements(struct pipe_context *_pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned mode, + unsigned start, + unsigned count) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + boolean result; + + trace_dump_call_begin(stream, "pipe_context", "draw_range_elements"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, indexBuffer); + trace_dump_arg(stream, uint, indexSize); + trace_dump_arg(stream, uint, minIndex); + trace_dump_arg(stream, uint, maxIndex); + trace_dump_arg(stream, uint, mode); + trace_dump_arg(stream, uint, start); + trace_dump_arg(stream, uint, count); + + result = pipe->draw_range_elements(pipe, + indexBuffer, + indexSize, minIndex, maxIndex, + mode, start, count); + + trace_dump_ret(stream, bool, result); + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE struct pipe_query * +trace_context_create_query(struct pipe_context *_pipe, + unsigned query_type) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + struct pipe_query *result; + + trace_dump_call_begin(stream, "pipe_context", "create_query"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, uint, query_type); + + result = pipe->create_query(pipe, query_type);; + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE void +trace_context_destroy_query(struct pipe_context *_pipe, + struct pipe_query *query) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "destroy_query"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, query); + + pipe->destroy_query(pipe, query);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_begin_query(struct pipe_context *_pipe, + struct pipe_query *query) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "begin_query"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, query); + + pipe->begin_query(pipe, query);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_end_query(struct pipe_context *_pipe, + struct pipe_query *query) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "end_query"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, query); + + pipe->end_query(pipe, query); + + trace_dump_call_end(stream); +} + + +static INLINE boolean +trace_context_get_query_result(struct pipe_context *_pipe, + struct pipe_query *query, + boolean wait, + uint64 *presult) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + uint64 result; + boolean _result; + + trace_dump_call_begin(stream, "pipe_context", "get_query_result"); + + trace_dump_arg(stream, ptr, pipe); + + _result = pipe->get_query_result(pipe, query, wait, presult);; + result = *presult; + + trace_dump_arg(stream, uint, result); + trace_dump_ret(stream, bool, _result); + + trace_dump_call_end(stream); + + return _result; +} + + +static INLINE void * +trace_context_create_blend_state(struct pipe_context *_pipe, + const struct pipe_blend_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + void * result; + + trace_dump_call_begin(stream, "pipe_context", "create_blend_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + result = pipe->create_blend_state(pipe, state);; + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE void +trace_context_bind_blend_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "bind_blend_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->bind_blend_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_delete_blend_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "delete_blend_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->delete_blend_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void * +trace_context_create_sampler_state(struct pipe_context *_pipe, + const struct pipe_sampler_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + void * result; + + trace_dump_call_begin(stream, "pipe_context", "create_sampler_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + result = pipe->create_sampler_state(pipe, state);; + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE void +trace_context_bind_sampler_states(struct pipe_context *_pipe, + unsigned num_states, void **states) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "bind_sampler_states"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, uint, num_states); + trace_dump_arg_array(stream, ptr, states, num_states); + + pipe->bind_sampler_states(pipe, num_states, states);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_delete_sampler_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "delete_sampler_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->delete_sampler_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void * +trace_context_create_rasterizer_state(struct pipe_context *_pipe, + const struct pipe_rasterizer_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + void * result; + + trace_dump_call_begin(stream, "pipe_context", "create_rasterizer_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + result = pipe->create_rasterizer_state(pipe, state);; + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE void +trace_context_bind_rasterizer_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "bind_rasterizer_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->bind_rasterizer_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_delete_rasterizer_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "delete_rasterizer_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->delete_rasterizer_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void * +trace_context_create_depth_stencil_alpha_state(struct pipe_context *_pipe, + const struct pipe_depth_stencil_alpha_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + void * result; + + trace_dump_call_begin(stream, "pipe_context", "create_depth_stencil_alpha_state"); + + result = pipe->create_depth_stencil_alpha_state(pipe, state);; + + trace_dump_ret(stream, ptr, result); + trace_dump_arg(stream, ptr, state); + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE void +trace_context_bind_depth_stencil_alpha_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "bind_depth_stencil_alpha_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->bind_depth_stencil_alpha_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_delete_depth_stencil_alpha_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "delete_depth_stencil_alpha_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->delete_depth_stencil_alpha_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void * +trace_context_create_fs_state(struct pipe_context *_pipe, + const struct pipe_shader_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + void * result; + + trace_dump_call_begin(stream, "pipe_context", "create_fs_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + result = pipe->create_fs_state(pipe, state);; + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE void +trace_context_bind_fs_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "bind_fs_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->bind_fs_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_delete_fs_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "delete_fs_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->delete_fs_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void * +trace_context_create_vs_state(struct pipe_context *_pipe, + const struct pipe_shader_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + void * result; + + trace_dump_call_begin(stream, "pipe_context", "create_vs_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + result = pipe->create_vs_state(pipe, state);; + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static INLINE void +trace_context_bind_vs_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "bind_vs_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->bind_vs_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_delete_vs_state(struct pipe_context *_pipe, + void *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "delete_vs_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->delete_vs_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_blend_color(struct pipe_context *_pipe, + const struct pipe_blend_color *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_blend_color"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->set_blend_color(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_clip_state(struct pipe_context *_pipe, + const struct pipe_clip_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_clip_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->set_clip_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_constant_buffer(struct pipe_context *_pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buffer) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_constant_buffer"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, uint, shader); + trace_dump_arg(stream, uint, index); + trace_dump_arg(stream, ptr, buffer); + + pipe->set_constant_buffer(pipe, shader, index, buffer);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_framebuffer_state(struct pipe_context *_pipe, + const struct pipe_framebuffer_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_framebuffer_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->set_framebuffer_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_polygon_stipple(struct pipe_context *_pipe, + const struct pipe_poly_stipple *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_polygon_stipple"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->set_polygon_stipple(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_scissor_state(struct pipe_context *_pipe, + const struct pipe_scissor_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_scissor_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->set_scissor_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_viewport_state(struct pipe_context *_pipe, + const struct pipe_viewport_state *state) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_viewport_state"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, state); + + pipe->set_viewport_state(pipe, state);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_sampler_textures(struct pipe_context *_pipe, + unsigned num_textures, + struct pipe_texture **textures) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_sampler_textures"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, uint, num_textures); + trace_dump_arg_array(stream, ptr, textures, num_textures); + + pipe->set_sampler_textures(pipe, num_textures, textures);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_vertex_buffers(struct pipe_context *_pipe, + unsigned num_buffers, + const struct pipe_vertex_buffer *buffers) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_vertex_buffers"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, uint, num_buffers); + trace_dump_arg(stream, ptr, buffers); + //trace_dump_arg_array(stream, ptr, buffers, num_buffers); + + pipe->set_vertex_buffers(pipe, num_buffers, buffers);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_set_vertex_elements(struct pipe_context *_pipe, + unsigned num_elements, + const struct pipe_vertex_element *elements) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "set_vertex_elements"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, uint, num_elements); + trace_dump_arg(stream, ptr, elements); + //trace_dump_arg_array(stream, ptr, elements, num_elements); + + pipe->set_vertex_elements(pipe, num_elements, elements);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_surface_copy(struct pipe_context *_pipe, + boolean do_flip, + struct pipe_surface *dest, + unsigned destx, unsigned desty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "surface_copy"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, bool, do_flip); + trace_dump_arg(stream, ptr, dest); + trace_dump_arg(stream, uint, destx); + trace_dump_arg(stream, uint, desty); + trace_dump_arg(stream, ptr, src); + trace_dump_arg(stream, uint, srcx); + trace_dump_arg(stream, uint, srcy); + trace_dump_arg(stream, uint, width); + trace_dump_arg(stream, uint, height); + + pipe->surface_copy(pipe, do_flip, + dest, destx, desty, + src, srcx, srcy, width, height); + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_surface_fill(struct pipe_context *_pipe, + struct pipe_surface *dst, + unsigned dstx, unsigned dsty, + unsigned width, unsigned height, + unsigned value) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "surface_fill"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, dst); + trace_dump_arg(stream, uint, dstx); + trace_dump_arg(stream, uint, dsty); + trace_dump_arg(stream, uint, width); + trace_dump_arg(stream, uint, height); + + pipe->surface_fill(pipe, dst, dstx, dsty, width, height, value);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_clear(struct pipe_context *_pipe, + struct pipe_surface *surface, + unsigned clearValue) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "clear"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, ptr, surface); + trace_dump_arg(stream, uint, clearValue); + + pipe->clear(pipe, surface, clearValue);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_flush(struct pipe_context *_pipe, + unsigned flags, + struct pipe_fence_handle **fence) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "flush"); + + trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(stream, uint, flags); + trace_dump_arg(stream, ptr, fence); + + pipe->flush(pipe, flags, fence);; + + trace_dump_call_end(stream); +} + + +static INLINE void +trace_context_destroy(struct pipe_context *_pipe) +{ + struct trace_context *tr_ctx = trace_context(_pipe); + struct trace_screen *tr_scr = trace_screen(_pipe->screen); + struct trace_stream *stream = tr_scr->stream; + struct pipe_context *pipe = tr_ctx->pipe; + + trace_dump_call_begin(stream, "pipe_context", "destroy"); + + trace_dump_arg(stream, ptr, pipe); + + pipe->destroy(pipe); + + trace_dump_call_end(stream); + + FREE(tr_ctx); +} + + struct pipe_context * trace_context_create(struct pipe_context *pipe) { - /* TODO */ - return pipe; + struct trace_context *tr_ctx; + + if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) + return pipe; + + tr_ctx = CALLOC_STRUCT(trace_context); + if(!tr_ctx) + return NULL; + + tr_ctx->base.winsys = pipe->winsys; + tr_ctx->base.screen = pipe->screen; + tr_ctx->base.destroy = trace_context_destroy; + tr_ctx->base.set_edgeflags = trace_context_set_edgeflags; + tr_ctx->base.draw_arrays = trace_context_draw_arrays; + tr_ctx->base.draw_elements = trace_context_draw_elements; + tr_ctx->base.draw_range_elements = trace_context_draw_range_elements; + tr_ctx->base.create_query = trace_context_create_query; + tr_ctx->base.destroy_query = trace_context_destroy_query; + tr_ctx->base.begin_query = trace_context_begin_query; + tr_ctx->base.end_query = trace_context_end_query; + tr_ctx->base.get_query_result = trace_context_get_query_result; + tr_ctx->base.create_blend_state = trace_context_create_blend_state; + tr_ctx->base.bind_blend_state = trace_context_bind_blend_state; + tr_ctx->base.delete_blend_state = trace_context_delete_blend_state; + tr_ctx->base.create_sampler_state = trace_context_create_sampler_state; + tr_ctx->base.bind_sampler_states = trace_context_bind_sampler_states; + tr_ctx->base.delete_sampler_state = trace_context_delete_sampler_state; + tr_ctx->base.create_rasterizer_state = trace_context_create_rasterizer_state; + tr_ctx->base.bind_rasterizer_state = trace_context_bind_rasterizer_state; + tr_ctx->base.delete_rasterizer_state = trace_context_delete_rasterizer_state; + tr_ctx->base.create_depth_stencil_alpha_state = trace_context_create_depth_stencil_alpha_state; + tr_ctx->base.bind_depth_stencil_alpha_state = trace_context_bind_depth_stencil_alpha_state; + tr_ctx->base.delete_depth_stencil_alpha_state = trace_context_delete_depth_stencil_alpha_state; + tr_ctx->base.create_fs_state = trace_context_create_fs_state; + tr_ctx->base.bind_fs_state = trace_context_bind_fs_state; + tr_ctx->base.delete_fs_state = trace_context_delete_fs_state; + tr_ctx->base.create_vs_state = trace_context_create_vs_state; + tr_ctx->base.bind_vs_state = trace_context_bind_vs_state; + tr_ctx->base.delete_vs_state = trace_context_delete_vs_state; + tr_ctx->base.set_blend_color = trace_context_set_blend_color; + tr_ctx->base.set_clip_state = trace_context_set_clip_state; + tr_ctx->base.set_constant_buffer = trace_context_set_constant_buffer; + tr_ctx->base.set_framebuffer_state = trace_context_set_framebuffer_state; + tr_ctx->base.set_polygon_stipple = trace_context_set_polygon_stipple; + tr_ctx->base.set_scissor_state = trace_context_set_scissor_state; + tr_ctx->base.set_viewport_state = trace_context_set_viewport_state; + tr_ctx->base.set_sampler_textures = trace_context_set_sampler_textures; + tr_ctx->base.set_vertex_buffers = trace_context_set_vertex_buffers; + tr_ctx->base.set_vertex_elements = trace_context_set_vertex_elements; + tr_ctx->base.surface_copy = trace_context_surface_copy; + tr_ctx->base.surface_fill = trace_context_surface_fill; + tr_ctx->base.clear = trace_context_clear; + tr_ctx->base.flush = trace_context_flush; + + tr_ctx->pipe = pipe; + + return &tr_ctx->base; } diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h index 75b8d11a7a..80fb5980d6 100644 --- a/src/gallium/drivers/trace/tr_context.h +++ b/src/gallium/drivers/trace/tr_context.h @@ -38,21 +38,21 @@ struct trace_context { struct pipe_context base; - /* TODO */ + struct pipe_context *pipe; }; static INLINE struct trace_context * -trace_context(struct pipe_context *context) +trace_context(struct pipe_context *pipe) { - assert(context); - return (struct trace_context *)context; + assert(pipe); + return (struct trace_context *)pipe; } struct pipe_context * -trace_context_create(struct pipe_context *context); +trace_context_create(struct pipe_context *pipe); #endif /* TR_CONTEXT_H_ */ diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index f086486591..0be0812ce8 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -91,6 +91,13 @@ void trace_dump_ptr(struct trace_stream *stream, const void *value); trace_dump_array_end(_stream); \ } while(0) +#define trace_dump_arg_array(_stream, _type, _arg, _size) \ + do { \ + trace_dump_arg_begin(_stream, #_arg); \ + trace_dump_array(_stream, _type, _arg, _size); \ + trace_dump_arg_end(_stream); \ + } while(0) + #define trace_dump_member(_stream, _type, _obj, _member) \ do { \ trace_dump_member_begin(_stream, #_member); \ -- cgit v1.2.3 From 9dee60969df7ff263e05430e69ef26982fe2bd94 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 18:55:05 +0100 Subject: trace: Dump state. --- src/gallium/drivers/trace/tr_context.c | 37 +-- src/gallium/drivers/trace/tr_dump.c | 10 +- src/gallium/drivers/trace/tr_dump.h | 31 ++- src/gallium/drivers/trace/tr_state.c | 433 ++++++++++++++++++++++++++++++++- src/gallium/drivers/trace/tr_state.h | 57 ++++- src/gallium/drivers/trace/trace.xsl | 6 + 6 files changed, 549 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index 6161e35fa5..ee8ad5eb97 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -280,7 +280,7 @@ trace_context_create_blend_state(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "create_blend_state"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, blend_state, state); result = pipe->create_blend_state(pipe, state);; @@ -347,7 +347,7 @@ trace_context_create_sampler_state(struct pipe_context *_pipe, result = pipe->create_sampler_state(pipe, state);; - trace_dump_ret(stream, ptr, result); + trace_dump_ret(stream, sampler_state, result); trace_dump_call_end(stream); @@ -409,7 +409,7 @@ trace_context_create_rasterizer_state(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "create_rasterizer_state"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, rasterizer_state, state); result = pipe->create_rasterizer_state(pipe, state);; @@ -476,7 +476,7 @@ trace_context_create_depth_stencil_alpha_state(struct pipe_context *_pipe, result = pipe->create_depth_stencil_alpha_state(pipe, state);; trace_dump_ret(stream, ptr, result); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, depth_stencil_alpha_state, state); trace_dump_call_end(stream); @@ -537,7 +537,7 @@ trace_context_create_fs_state(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "create_fs_state"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, shader_state, state); result = pipe->create_fs_state(pipe, state);; @@ -602,7 +602,7 @@ trace_context_create_vs_state(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "create_vs_state"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, shader_state, state); result = pipe->create_vs_state(pipe, state);; @@ -666,7 +666,7 @@ trace_context_set_blend_color(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "set_blend_color"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, blend_color, state); pipe->set_blend_color(pipe, state);; @@ -686,7 +686,7 @@ trace_context_set_clip_state(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "set_clip_state"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, clip_state, state); pipe->set_clip_state(pipe, state);; @@ -709,7 +709,7 @@ trace_context_set_constant_buffer(struct pipe_context *_pipe, trace_dump_arg(stream, ptr, pipe); trace_dump_arg(stream, uint, shader); trace_dump_arg(stream, uint, index); - trace_dump_arg(stream, ptr, buffer); + trace_dump_arg(stream, constant_buffer, buffer); pipe->set_constant_buffer(pipe, shader, index, buffer);; @@ -729,7 +729,7 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "set_framebuffer_state"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, framebuffer_state, state); pipe->set_framebuffer_state(pipe, state);; @@ -749,7 +749,7 @@ trace_context_set_polygon_stipple(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "set_polygon_stipple"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, poly_stipple, state); pipe->set_polygon_stipple(pipe, state);; @@ -769,7 +769,7 @@ trace_context_set_scissor_state(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "set_scissor_state"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, scissor_state, state); pipe->set_scissor_state(pipe, state);; @@ -789,7 +789,7 @@ trace_context_set_viewport_state(struct pipe_context *_pipe, trace_dump_call_begin(stream, "pipe_context", "set_viewport_state"); trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(stream, viewport_state, state); pipe->set_viewport_state(pipe, state);; @@ -833,8 +833,10 @@ trace_context_set_vertex_buffers(struct pipe_context *_pipe, trace_dump_arg(stream, ptr, pipe); trace_dump_arg(stream, uint, num_buffers); - trace_dump_arg(stream, ptr, buffers); - //trace_dump_arg_array(stream, ptr, buffers, num_buffers); + + trace_dump_arg_begin(stream, "buffers"); + trace_dump_struct_array(stream, vertex_buffer, buffers, num_buffers); + trace_dump_arg_end(stream); pipe->set_vertex_buffers(pipe, num_buffers, buffers);; @@ -857,7 +859,10 @@ trace_context_set_vertex_elements(struct pipe_context *_pipe, trace_dump_arg(stream, ptr, pipe); trace_dump_arg(stream, uint, num_elements); trace_dump_arg(stream, ptr, elements); - //trace_dump_arg_array(stream, ptr, elements, num_elements); + + trace_dump_arg_begin(stream, "elements"); + trace_dump_struct_array(stream, vertex_element, elements, num_elements); + trace_dump_arg_end(stream); pipe->set_vertex_elements(pipe, num_elements, elements);; diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index 30c0986363..0591b9fc33 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -322,8 +322,16 @@ void trace_dump_member_end(struct trace_stream *stream) trace_dump_write(stream, ""); } +void trace_dump_null(struct trace_stream *stream) +{ + trace_dump_write(stream, ""); +} + void trace_dump_ptr(struct trace_stream *stream, const void *value) { - trace_dump_writef(stream, "%p", value); + if(value) + trace_dump_writef(stream, "%p", value); + else + trace_dump_null(stream); } diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index 0be0812ce8..5eeac8d3c5 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -34,6 +34,9 @@ #define TR_DUMP_H +#include "pipe/p_util.h" + + struct trace_stream; @@ -58,6 +61,7 @@ void trace_dump_struct_begin(struct trace_stream *stream, const char *name); void trace_dump_struct_end(struct trace_stream *stream); void trace_dump_member_begin(struct trace_stream *stream, const char *name); void trace_dump_member_end(struct trace_stream *stream); +void trace_dump_null(struct trace_stream *stream); void trace_dump_ptr(struct trace_stream *stream, const void *value); @@ -83,14 +87,33 @@ void trace_dump_ptr(struct trace_stream *stream, const void *value); do { \ unsigned long idx; \ trace_dump_array_begin(_stream); \ - for(idx = 0; idx < _size; ++idx) { \ + for(idx = 0; idx < (_size); ++idx) { \ + trace_dump_elem_begin(_stream); \ + trace_dump_##_type(_stream, (_obj)[idx]); \ + trace_dump_elem_end(_stream); \ + } \ + trace_dump_array_end(_stream); \ + } while(0) + +#define trace_dump_struct_array(_stream, _type, _obj, _size) \ + do { \ + unsigned long idx; \ + trace_dump_array_begin(_stream); \ + for(idx = 0; idx < (_size); ++idx) { \ trace_dump_elem_begin(_stream); \ - trace_dump_##_type(_stream, _obj[idx]); \ + trace_dump_##_type(_stream, &(_obj)[idx]); \ trace_dump_elem_end(_stream); \ } \ trace_dump_array_end(_stream); \ } while(0) +#define trace_dump_member(_stream, _type, _obj, _member) \ + do { \ + trace_dump_member_begin(_stream, #_member); \ + trace_dump_##_type(_stream, (_obj)->_member); \ + trace_dump_member_end(_stream); \ + } while(0) + #define trace_dump_arg_array(_stream, _type, _arg, _size) \ do { \ trace_dump_arg_begin(_stream, #_arg); \ @@ -98,10 +121,10 @@ void trace_dump_ptr(struct trace_stream *stream, const void *value); trace_dump_arg_end(_stream); \ } while(0) -#define trace_dump_member(_stream, _type, _obj, _member) \ +#define trace_dump_member_array(_stream, _type, _obj, _member) \ do { \ trace_dump_member_begin(_stream, #_member); \ - trace_dump_##_type(_stream, _obj->_member); \ + trace_dump_array(_stream, _type, (_obj)->_member, sizeof((_obj)->_member)/sizeof((_obj)->_member[0])); \ trace_dump_member_end(_stream); \ } while(0) diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index 1e0e87f7fd..f18f5c611e 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -27,12 +27,19 @@ #include "pipe/p_compiler.h" -#include "pipe/p_state.h" +#include "tgsi/tgsi_parse.h" #include "tr_dump.h" #include "tr_state.h" +void trace_dump_format(struct trace_stream *stream, + enum pipe_format format) +{ + trace_dump_int(stream, format); +} + + void trace_dump_block(struct trace_stream *stream, const struct pipe_format_block *block) { @@ -44,9 +51,32 @@ void trace_dump_block(struct trace_stream *stream, } +void trace_dump_buffer(struct trace_stream *stream, + const struct pipe_buffer *buffer) +{ + if(!buffer) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_buffer"); + trace_dump_member(stream, uint, buffer, alignment); + trace_dump_member(stream, uint, buffer, usage); + trace_dump_member(stream, uint, buffer, size); + /* TODO: buffer data */ + trace_dump_struct_end(stream); +} + + void trace_dump_template(struct trace_stream *stream, const struct pipe_texture *templat) { + assert(templat); + if(!templat) { + trace_dump_null(stream); + return; + } + trace_dump_struct_begin(stream, "pipe_texture"); trace_dump_member(stream, int, templat, target); @@ -70,3 +100,404 @@ void trace_dump_template(struct trace_stream *stream, trace_dump_struct_end(stream); } + +void trace_dump_rasterizer_state(struct trace_stream *stream, + const struct pipe_rasterizer_state *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_rasterizer_state"); + + trace_dump_member(stream, bool, state, flatshade); + trace_dump_member(stream, bool, state, light_twoside); + trace_dump_member(stream, uint, state, front_winding); + trace_dump_member(stream, uint, state, cull_mode); + trace_dump_member(stream, uint, state, fill_cw); + trace_dump_member(stream, uint, state, fill_ccw); + trace_dump_member(stream, bool, state, offset_cw); + trace_dump_member(stream, bool, state, offset_ccw); + trace_dump_member(stream, bool, state, scissor); + trace_dump_member(stream, bool, state, poly_smooth); + trace_dump_member(stream, bool, state, poly_stipple_enable); + trace_dump_member(stream, bool, state, point_smooth); + trace_dump_member(stream, bool, state, point_sprite); + trace_dump_member(stream, bool, state, point_size_per_vertex); + trace_dump_member(stream, bool, state, multisample); + trace_dump_member(stream, bool, state, line_smooth); + trace_dump_member(stream, bool, state, line_stipple_enable); + trace_dump_member(stream, uint, state, line_stipple_factor); + trace_dump_member(stream, uint, state, line_stipple_pattern); + trace_dump_member(stream, bool, state, line_last_pixel); + trace_dump_member(stream, bool, state, bypass_clipping); + trace_dump_member(stream, bool, state, bypass_vs); + trace_dump_member(stream, bool, state, origin_lower_left); + trace_dump_member(stream, bool, state, flatshade_first); + trace_dump_member(stream, bool, state, gl_rasterization_rules); + + trace_dump_member(stream, float, state, line_width); + trace_dump_member(stream, float, state, point_size); + trace_dump_member(stream, float, state, point_size_min); + trace_dump_member(stream, float, state, point_size_max); + trace_dump_member(stream, float, state, offset_units); + trace_dump_member(stream, float, state, offset_scale); + + trace_dump_member_array(stream, uint, state, sprite_coord_mode); + + trace_dump_struct_end(stream); +} + + +void trace_dump_poly_stipple(struct trace_stream *stream, + const struct pipe_poly_stipple *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_poly_stipple"); + + trace_dump_member_array(stream, uint, state, stipple); + + trace_dump_struct_end(stream); +} + + +void trace_dump_viewport_state(struct trace_stream *stream, + const struct pipe_viewport_state *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_viewport_state"); + + trace_dump_member_array(stream, float, state, scale); + trace_dump_member_array(stream, float, state, translate); + + trace_dump_struct_end(stream); +} + + +void trace_dump_scissor_state(struct trace_stream *stream, + const struct pipe_scissor_state *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_scissor_state"); + + trace_dump_member(stream, uint, state, minx); + trace_dump_member(stream, uint, state, miny); + trace_dump_member(stream, uint, state, maxx); + trace_dump_member(stream, uint, state, maxy); + + trace_dump_struct_end(stream); +} + + +void trace_dump_clip_state(struct trace_stream *stream, + const struct pipe_clip_state *state) +{ + unsigned i; + + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_scissor_state"); + + trace_dump_member_begin(stream, "ucp"); + trace_dump_array_begin(stream); + for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) + trace_dump_array(stream, float, state->ucp[i], 4); + trace_dump_array_end(stream); + trace_dump_member_end(stream); + + trace_dump_member(stream, uint, state, nr); + + trace_dump_struct_end(stream); +} + + +void trace_dump_constant_buffer(struct trace_stream *stream, + const struct pipe_constant_buffer *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_constant_buffer"); + + trace_dump_member(stream, buffer, state, buffer); + trace_dump_member(stream, uint, state, size); + + trace_dump_struct_end(stream); +} + + +void trace_dump_shader_state(struct trace_stream *stream, + const struct pipe_shader_state *state) +{ + uint32_t *p = (uint32_t *)state->tokens; + unsigned n = tgsi_num_tokens(state->tokens); + + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + assert(sizeof(struct tgsi_token) == 4); + + trace_dump_struct_begin(stream, "pipe_shader_state"); + + trace_dump_member_begin(stream, "tokens"); + trace_dump_array(stream, uint, p, n); + trace_dump_member_end(stream); + + trace_dump_struct_end(stream); +} + + +void trace_dump_depth_stencil_alpha_state(struct trace_stream *stream, + const struct pipe_depth_stencil_alpha_state *state) +{ + unsigned i; + + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_depth_stencil_alpha_state"); + + trace_dump_member_begin(stream, "depth"); + trace_dump_struct_begin(stream, ""); + trace_dump_member(stream, bool, &state->depth, enabled); + trace_dump_member(stream, bool, &state->depth, writemask); + trace_dump_member(stream, uint, &state->depth, func); + trace_dump_member(stream, bool, &state->depth, occlusion_count); + trace_dump_struct_end(stream); + trace_dump_member_end(stream); + + trace_dump_member_begin(stream, "stencil"); + trace_dump_array_begin(stream); + for(i = 0; i < Elements(state->stencil); ++i) { + trace_dump_elem_begin(stream); + trace_dump_struct_begin(stream, ""); + trace_dump_member(stream, bool, &state->stencil[i], enabled); + trace_dump_member(stream, uint, &state->stencil[i], func); + trace_dump_member(stream, uint, &state->stencil[i], fail_op); + trace_dump_member(stream, uint, &state->stencil[i], zpass_op); + trace_dump_member(stream, uint, &state->stencil[i], zfail_op); + trace_dump_member(stream, uint, &state->stencil[i], ref_value); + trace_dump_member(stream, uint, &state->stencil[i], value_mask); + trace_dump_member(stream, uint, &state->stencil[i], write_mask); + trace_dump_struct_end(stream); + trace_dump_elem_end(stream); + } + trace_dump_array_end(stream); + trace_dump_member_end(stream); + + trace_dump_member_begin(stream, "alpha"); + trace_dump_struct_begin(stream, ""); + trace_dump_member(stream, bool, &state->alpha, enabled); + trace_dump_member(stream, uint, &state->alpha, func); + trace_dump_member(stream, float, &state->alpha, ref); + trace_dump_struct_end(stream); + trace_dump_member_end(stream); + + trace_dump_struct_end(stream); +} + + +void trace_dump_blend_state(struct trace_stream *stream, + const struct pipe_blend_state *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_blend_state"); + + trace_dump_member(stream, bool, state, blend_enable); + + trace_dump_member(stream, uint, state, rgb_func); + trace_dump_member(stream, uint, state, rgb_src_factor); + trace_dump_member(stream, uint, state, rgb_dst_factor); + + trace_dump_member(stream, uint, state, alpha_func); + trace_dump_member(stream, uint, state, alpha_src_factor); + trace_dump_member(stream, uint, state, alpha_dst_factor); + + trace_dump_member(stream, bool, state, logicop_enable); + trace_dump_member(stream, uint, state, logicop_func); + + trace_dump_member(stream, uint, state, colormask); + trace_dump_member(stream, bool, state, dither); + + trace_dump_struct_end(stream); +} + + +void trace_dump_blend_color(struct trace_stream *stream, + const struct pipe_blend_color *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_blend_color"); + + trace_dump_member_array(stream, float, state, color); + + trace_dump_struct_end(stream); +} + + +void trace_dump_framebuffer_state(struct trace_stream *stream, + const struct pipe_framebuffer_state *state) +{ + trace_dump_struct_begin(stream, "pipe_framebuffer_state"); + + trace_dump_member(stream, uint, state, width); + trace_dump_member(stream, uint, state, height); + trace_dump_member(stream, uint, state, num_cbufs); + trace_dump_member_array(stream, ptr, state, cbufs); + trace_dump_member(stream, ptr, state, zsbuf); + + trace_dump_struct_end(stream); +} + + +void trace_dump_sampler_state(struct trace_stream *stream, + const struct pipe_sampler_state *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_sampler_state"); + + trace_dump_member(stream, uint, state, wrap_s); + trace_dump_member(stream, uint, state, wrap_t); + trace_dump_member(stream, uint, state, wrap_r); + trace_dump_member(stream, uint, state, min_img_filter); + trace_dump_member(stream, uint, state, min_mip_filter); + trace_dump_member(stream, uint, state, mag_img_filter); + trace_dump_member(stream, bool, state, compare_mode); + trace_dump_member(stream, uint, state, compare_func); + trace_dump_member(stream, bool, state, normalized_coords); + trace_dump_member(stream, uint, state, prefilter); + trace_dump_member(stream, float, state, shadow_ambient); + trace_dump_member(stream, float, state, lod_bias); + trace_dump_member(stream, float, state, min_lod); + trace_dump_member(stream, float, state, max_lod); + trace_dump_member_array(stream, float, state, border_color); + trace_dump_member(stream, float, state, max_anisotropy); + + trace_dump_struct_end(stream); +} + + +void trace_dump_surface(struct trace_stream *stream, + const struct pipe_surface *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_surface"); + + trace_dump_member(stream, buffer, state, buffer); + trace_dump_member(stream, format, state, format); + trace_dump_member(stream, uint, state, status); + trace_dump_member(stream, uint, state, clear_value); + trace_dump_member(stream, uint, state, width); + trace_dump_member(stream, uint, state, height); + + trace_dump_member_begin(stream, "block"); + trace_dump_block(stream, &state->block); + trace_dump_member_end(stream); + + trace_dump_member(stream, uint, state, nblocksx); + trace_dump_member(stream, uint, state, nblocksy); + trace_dump_member(stream, uint, state, stride); + trace_dump_member(stream, uint, state, layout); + trace_dump_member(stream, uint, state, offset); + trace_dump_member(stream, uint, state, refcount); + trace_dump_member(stream, uint, state, usage); + + trace_dump_member(stream, ptr, state, texture); + trace_dump_member(stream, uint, state, face); + trace_dump_member(stream, uint, state, level); + trace_dump_member(stream, uint, state, zslice); + + trace_dump_struct_end(stream); +} + + +void trace_dump_vertex_buffer(struct trace_stream *stream, + const struct pipe_vertex_buffer *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_vertex_buffer"); + + trace_dump_member(stream, uint, state, pitch); + trace_dump_member(stream, uint, state, max_index); + trace_dump_member(stream, uint, state, buffer_offset); + trace_dump_member(stream, buffer, state, buffer); + + trace_dump_struct_end(stream); +} + + +void trace_dump_vertex_element(struct trace_stream *stream, + const struct pipe_vertex_element *state) +{ + assert(state); + if(!state) { + trace_dump_null(stream); + return; + } + + trace_dump_struct_begin(stream, "pipe_vertex_element"); + + trace_dump_member(stream, uint, state, src_offset); + + trace_dump_member(stream, uint, state, vertex_buffer_index); + trace_dump_member(stream, uint, state, nr_components); + + trace_dump_member(stream, format, state, src_format); + + trace_dump_struct_end(stream); +} diff --git a/src/gallium/drivers/trace/tr_state.h b/src/gallium/drivers/trace/tr_state.h index e52524ee90..c1df63db6a 100644 --- a/src/gallium/drivers/trace/tr_state.h +++ b/src/gallium/drivers/trace/tr_state.h @@ -28,11 +28,13 @@ #ifndef TR_STATE_H #define TR_STATE_H +#include "pipe/p_format.h" +#include "pipe/p_state.h" +#include "pipe/p_shader_tokens.h" -struct trace_stream; -struct pipe_format_block; -struct pipe_texture; +void trace_dump_format(struct trace_stream *stream, + enum pipe_format format); void trace_dump_block(struct trace_stream *stream, const struct pipe_format_block *block); @@ -41,4 +43,53 @@ void trace_dump_template(struct trace_stream *stream, const struct pipe_texture *templat); +void trace_dump_rasterizer_state(struct trace_stream *stream, + const struct pipe_rasterizer_state *state); + +void trace_dump_poly_stipple(struct trace_stream *stream, + const struct pipe_poly_stipple *state); + +void trace_dump_viewport_state(struct trace_stream *stream, + const struct pipe_viewport_state *state); + +void trace_dump_scissor_state(struct trace_stream *stream, + const struct pipe_scissor_state *state); + +void trace_dump_clip_state(struct trace_stream *stream, + const struct pipe_clip_state *state); + +void trace_dump_constant_buffer(struct trace_stream *stream, + const struct pipe_constant_buffer *state); + +void trace_dump_token(struct trace_stream *stream, + const struct tgsi_token *token); + +void trace_dump_shader_state(struct trace_stream *stream, + const struct pipe_shader_state *state); + +void trace_dump_depth_stencil_alpha_state(struct trace_stream *stream, + const struct pipe_depth_stencil_alpha_state *state); + +void trace_dump_blend_state(struct trace_stream *stream, + const struct pipe_blend_state *state); + +void trace_dump_blend_color(struct trace_stream *stream, + const struct pipe_blend_color *state); + +void trace_dump_framebuffer_state(struct trace_stream *stream, + const struct pipe_framebuffer_state *state); + +void trace_dump_sampler_state(struct trace_stream *stream, + const struct pipe_sampler_state *state); + +void trace_dump_surface(struct trace_stream *stream, + const struct pipe_surface *state); + +void trace_dump_vertex_buffer(struct trace_stream *stream, + const struct pipe_vertex_buffer *state); + +void trace_dump_vertex_element(struct trace_stream *stream, + const struct pipe_vertex_element *state); + + #endif /* TR_STATE_H */ diff --git a/src/gallium/drivers/trace/trace.xsl b/src/gallium/drivers/trace/trace.xsl index ace4020cf9..194c8b2efd 100644 --- a/src/gallium/drivers/trace/trace.xsl +++ b/src/gallium/drivers/trace/trace.xsl @@ -100,6 +100,12 @@ along with this program. If not, see . + + + NULL + + + -- cgit v1.2.3 From fdafa1f8632bf4e92996291a094e98f4e58dc37c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 19:25:28 +0100 Subject: gallium: Simplify format->name conversion. --- src/gallium/auxiliary/util/p_debug.c | 7 +++---- src/gallium/include/pipe/p_format.h | 2 +- 2 files changed, 4 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index a3db13f96d..082b0e9fb5 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -335,7 +335,7 @@ const char * debug_dump_enum(const struct debug_named_value *names, unsigned long value) { - static char rest[256]; + static char rest[64]; while(names->name) { if(names->value == value) @@ -498,10 +498,9 @@ void debug_print_format(const char *msg, unsigned fmt ) } #endif -char *pf_sprint_name( char *str, enum pipe_format format ) +const char *pf_name( enum pipe_format format ) { - strcpy( str, debug_dump_enum(pipe_format_names, format) ); - return str; + return debug_dump_enum(pipe_format_names, format); } diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 947c445583..0a9cad8993 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -379,7 +379,7 @@ enum pipe_format { /** * Builds pipe format name from format token. */ -extern char *pf_sprint_name( char *str, enum pipe_format format ); +extern const char *pf_name( enum pipe_format format ); /** * Return bits for a particular component. -- cgit v1.2.3 From 1ffd6908d4153d647f8a3bf1ba9fe9d33c206185 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 19:25:54 +0100 Subject: mesa: pf_sprint_name->pf_name. --- src/mesa/state_tracker/st_draw.c | 3 +-- src/mesa/state_tracker/st_format.c | 9 +-------- 2 files changed, 2 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index 30c2bdb723..cc3d7450a9 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -339,7 +339,6 @@ st_draw_vbo(GLcontext *ctx, #if 0 { GLuint i; - char buf[100]; for (i = 0; i < vp->num_inputs; i++) { printf("buffers[%d].pitch = %u\n", i, vbuffer[i].pitch); printf("buffers[%d].max_index = %u\n", i, vbuffer[i].max_index); @@ -350,7 +349,7 @@ st_draw_vbo(GLcontext *ctx, printf("vlements[%d].src_offset = %u\n", i, velements[i].src_offset); printf("vlements[%d].vbuffer_index = %u\n", i, velements[i].vertex_buffer_index); printf("vlements[%d].nr_comps = %u\n", i, velements[i].nr_components); - printf("vlements[%d].format = %s\n", i, pf_sprint_name(buf, velements[i].src_format)); + printf("vlements[%d].format = %s\n", i, pf_name(velements[i].src_format)); } } #endif diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index b6d97ef659..2461f39622 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -91,14 +91,7 @@ st_get_format_info(enum pipe_format format, struct pipe_format_info *pinfo) info = format; #if 0 - { - char fmtname[256]; - - pf_sprint_name( fmtname, format ); - printf( - "%s\n", - fmtname ); - } + printf("%s\n", pf_name( format ) ); #endif /* Data type */ -- cgit v1.2.3 From a78f295f7ca0a93fdaeda559668fda013c2a03b0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 10:26:35 -0600 Subject: mesa: glsl: finish up support for precision qualifiers --- src/mesa/shader/slang/slang_compile.c | 21 +++++++++++++++++---- src/mesa/shader/slang/slang_compile_variable.h | 10 ++++++++++ 2 files changed, 27 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 2a1af39ee0..64c0dad611 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -707,20 +707,33 @@ parse_fully_specified_type(slang_parse_ctx * C, slang_output_ctx * O, precision = *C->I++; if (!parse_type_specifier(C, O, &type->specifier)) return 0; + switch (precision) { case PRECISION_DEFAULT: - /* TODO: Grab the default precision for the given type specifier. - */ + assert(type->specifier.type < TYPE_SPECIFIER_COUNT); + if (type->specifier.type < TYPE_SPECIFIER_COUNT) + type->precision = O->default_precision[type->specifier.type]; break; case PRECISION_LOW: + type->precision = SLANG_PREC_LOW; + break; case PRECISION_MEDIUM: + type->precision = SLANG_PREC_MEDIUM; + break; case PRECISION_HIGH: - /* TODO: Translate to mesa representation. - */ + type->precision = SLANG_PREC_HIGH; break; default: return 0; } + +#if !FEATURE_es2_glsl + if (precision != PRECISION_DEFAULT) { + slang_info_log_error(C->L, "precision qualifiers not allowed"); + return 0; + } +#endif + return 1; } diff --git a/src/mesa/shader/slang/slang_compile_variable.h b/src/mesa/shader/slang/slang_compile_variable.h index 9b0f85859a..d81a3d2869 100644 --- a/src/mesa/shader/slang/slang_compile_variable.h +++ b/src/mesa/shader/slang/slang_compile_variable.h @@ -51,10 +51,20 @@ slang_type_specifier_type_to_string(slang_type_specifier_type); +typedef enum slang_type_precision_ +{ + SLANG_PREC_DEFAULT, + SLANG_PREC_LOW, + SLANG_PREC_MEDIUM, + SLANG_PREC_HIGH +} slang_type_precision; + + typedef struct slang_fully_specified_type_ { slang_type_qualifier qualifier; slang_type_specifier specifier; + slang_type_precision precision; } slang_fully_specified_type; extern int -- cgit v1.2.3 From 14ac52e457ea71f2cfc761948f397cedab4d255b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 10:26:56 -0600 Subject: mesa: glsl: regenerated files --- src/mesa/shader/slang/library/slang_120_core_gc.h | 1332 ++++++++--------- .../slang/library/slang_builtin_120_common_gc.h | 188 +-- .../slang/library/slang_builtin_120_fragment_gc.h | 2 +- .../shader/slang/library/slang_common_builtin_gc.h | 1570 ++++++++++---------- src/mesa/shader/slang/library/slang_core_gc.h | 1562 +++++++++---------- .../slang/library/slang_fragment_builtin_gc.h | 190 +-- .../shader/slang/library/slang_vertex_builtin_gc.h | 175 +-- 7 files changed, 2522 insertions(+), 2497 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_120_core_gc.h b/src/mesa/shader/slang/library/slang_120_core_gc.h index 1bd781ffe5..3ee03e7ec3 100644 --- a/src/mesa/shader/slang/library/slang_120_core_gc.h +++ b/src/mesa/shader/slang/library/slang_120_core_gc.h @@ -2,709 +2,715 @@ /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ /* slang_120_core.gc */ -3,1,0,26,1,1,1,0,9,102,48,48,0,0,1,1,0,9,102,49,48,0,0,1,1,0,9,102,50,48,0,0,1,1,0,9,102,48,49,0,0, -1,1,0,9,102,49,49,0,0,1,1,0,9,102,50,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59, -120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114, +4,1,0,0,26,1,1,1,0,9,102,48,48,0,0,1,1,0,9,102,49,48,0,0,1,1,0,9,102,50,48,0,0,1,1,0,9,102,48,49,0, +0,1,1,0,9,102,49,49,0,0,1,1,0,9,102,50,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48, +0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114, 101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0, 16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122, -0,18,102,50,49,0,20,0,0,1,0,26,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97, -116,50,120,51,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0, -48,0,0,0,0,20,0,0,1,0,26,1,1,1,0,5,105,0,0,0,1,3,2,1,9,1,102,0,2,58,102,108,111,97,116,0,18,105,0, -0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,102,0,0,0,20,0,0,1,0,26,1, -1,1,0,1,98,0,0,0,1,3,2,1,9,1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,50,120,51,0,18,102,0,0,0,20,0,0,1,0,26,1,1,1,0,11,99,48,0,0,1,1,0,11, -99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,28,1,1,1,0,9,102,48,48,0,0,1,1,0,9,102,49,48,0, -0,1,1,0,9,102,50,48,0,0,1,1,0,9,102,51,48,0,0,1,1,0,9,102,48,49,0,0,1,1,0,9,102,49,49,0,0,1,1,0,9, -102,50,49,0,0,1,1,0,9,102,51,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0, -18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,8,48,0,57,59,119,0,18,102,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, +0,18,102,50,49,0,20,0,0,1,0,0,26,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, +97,116,50,120,51,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48, +0,48,0,0,0,0,20,0,0,1,0,0,26,1,1,1,0,5,105,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18, +105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,102,0,0,0,20,0,0,1,0, +0,26,1,1,1,0,1,98,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95, +114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,102,0,0,0,20,0,0,1,0,0,26,1,1,1,0,11,99,48,0, +0,1,1,0,11,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,0,28,1,1,1,0,9,102,48,48,0,0,1,1,0, +9,102,49,48,0,0,1,1,0,9,102,50,48,0,0,1,1,0,9,102,51,48,0,0,1,1,0,9,102,48,49,0,0,1,1,0,9,102,49, +49,0,0,1,1,0,9,102,50,49,0,0,1,1,0,9,102,51,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102, +49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,102,51,48,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, +59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50, +49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0,18,102,51,49,0,20,0,0,1,0,0,28, +1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,102,0,0,17, +48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48, +0,48,0,0,0,0,20,0,0,1,0,0,28,1,1,1,0,5,105,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18, +105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,102,0,0,0,20,0,0,1,0, +0,28,1,1,1,0,1,98,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95, +114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,102,0,0,0,20,0,0,1,0,0,28,1,1,1,0,12,99,48,0, +0,1,1,0,12,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,0,27,1,1,1,0,9,102,48,48,0,0,1,1,0, +9,102,49,48,0,0,1,1,0,9,102,48,49,0,0,1,1,0,9,102,49,49,0,0,1,1,0,9,102,48,50,0,0,1,1,0,9,102,49, +50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59, +121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,0,1,0,0,27,1,1, +1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,102,0,0,17,48,0, +48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,27,1,1,1,0,5, +105,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86, +97,108,0,58,109,97,116,51,120,50,0,18,102,0,0,0,20,0,0,1,0,0,27,1,1,1,0,1,98,0,0,0,1,3,2,1,0,9,1, +102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116, +51,120,50,0,18,102,0,0,0,20,0,0,1,0,0,27,1,1,1,0,10,99,48,0,0,1,1,0,10,99,49,0,0,1,1,0,10,99,50,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20, +0,0,1,0,0,30,1,1,1,0,9,102,48,48,0,0,1,1,0,9,102,49,48,0,0,1,1,0,9,102,50,48,0,0,1,1,0,9,102,51,48, +0,0,1,1,0,9,102,48,49,0,0,1,1,0,9,102,49,49,0,0,1,1,0,9,102,50,49,0,0,1,1,0,9,102,51,49,0,0,1,1,0, +9,102,48,50,0,0,1,1,0,9,102,49,50,0,0,1,1,0,9,102,50,50,0,0,1,1,0,9,102,51,50,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, +0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,102,51,48,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,59,122,0,18,102,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0, +18,102,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,59,122,0,18,102,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,50,0,57,59,119,0,18,102,51,50,0,20,0,0,1,0,0,30,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,58,109,97,116,51,120,52,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0, +17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18, +102,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,30,1,1,1,0,5,105,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108, +111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,102, +0,0,0,20,0,0,1,0,0,30,1,1,1,0,1,98,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,0, +0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,102,0,0,0,20,0,0,1,0,0,30,1,1, +1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,0,0,29,1,1,1,0,9,102,48,48,0,0,1,1,0, +9,102,49,48,0,0,1,1,0,9,102,48,49,0,0,1,1,0,9,102,49,49,0,0,1,1,0,9,102,48,50,0,0,1,1,0,9,102,49, +50,0,0,1,1,0,9,102,48,51,0,0,1,1,0,9,102,49,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102, +49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0,18,102,48, +51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,102,49,51,0,20,0,0,1,0,0,29, +1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,102,0,0,17, +48,0,48,0,0,0,17,48,0,48,0,0,0,16,10,52,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17, +48,0,48,0,0,0,0,20,0,0,1,0,0,29,1,1,1,0,5,105,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108,111,97,116,0, +18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,102,0,0,0,20,0,0, +1,0,0,29,1,1,1,0,1,98,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95, +95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,102,0,0,0,20,0,0,1,0,0,29,1,1,1,0,10,99,48, +0,0,1,1,0,10,99,49,0,0,1,1,0,10,99,50,0,0,1,1,0,10,99,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,51,0,57,18,99,51,0,20,0,0,1,0,0,31,1,1,1,0,9,102,48,48,0,0,1,1,0,9,102,49,48,0,0,1,1,0,9,102, +50,48,0,0,1,1,0,9,102,48,49,0,0,1,1,0,9,102,49,49,0,0,1,1,0,9,102,50,49,0,0,1,1,0,9,102,48,50,0,0, +1,1,0,9,102,49,50,0,0,1,1,0,9,102,50,50,0,0,1,1,0,9,102,48,51,0,0,1,1,0,9,102,49,51,0,0,1,1,0,9, +102,50,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, 0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102, 49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50,49,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0,18,102,51,49,0,20,0,0,1,0,28,1,1,1,0,9,102,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0, -48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1, -0,28,1,1,1,0,5,105,0,0,0,1,3,2,1,9,1,102,0,2,58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,102,0,0,0,20,0,0,1,0,28,1,1,1,0,1,98,0,0,0,1, -3,2,1,9,1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58, -109,97,116,50,120,52,0,18,102,0,0,0,20,0,0,1,0,28,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,99,49,0,20,0,0,1,0,27,1,1,1,0,9,102,48,48,0,0,1,1,0,9,102,49,48,0,0,1,1,0,9,102,48, -49,0,0,1,1,0,9,102,49,49,0,0,1,1,0,9,102,48,50,0,0,1,1,0,9,102,49,50,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18, -102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,49,49,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,0,1,0,27,1,1,1,0,9,102,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18, -102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,27,1,1,1,0,5,105,0,0,0,1,3,2,1,9,1,102,0,2, -58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120, -50,0,18,102,0,0,0,20,0,0,1,0,27,1,1,1,0,1,98,0,0,0,1,3,2,1,9,1,102,0,2,58,102,108,111,97,116,0,18, -98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,102,0,0,0,20,0,0,1,0, -27,1,1,1,0,10,99,48,0,0,1,1,0,10,99,49,0,0,1,1,0,10,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,0,30,1,1,1,0,9,102,48,48,0,0,1, -1,0,9,102,49,48,0,0,1,1,0,9,102,50,48,0,0,1,1,0,9,102,51,48,0,0,1,1,0,9,102,48,49,0,0,1,1,0,9,102, -49,49,0,0,1,1,0,9,102,50,49,0,0,1,1,0,9,102,51,49,0,0,1,1,0,9,102,48,50,0,0,1,1,0,9,102,49,50,0,0, -1,1,0,9,102,50,50,0,0,1,1,0,9,102,51,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59, -120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,102,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121, -0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,102,50,49,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0,18,102,51,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0, -18,102,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,119,0,18,102,51,50,0,20,0, -0,1,0,30,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,102, -0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0, -0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,30,1, -1,1,0,5,105,0,0,0,1,3,2,1,9,1,102,0,2,58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,51,120,52,0,18,102,0,0,0,20,0,0,1,0,30,1,1,1,0,1,98,0,0,0,1,3,2,1,9, -1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116, -51,120,52,0,18,102,0,0,0,20,0,0,1,0,30,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0,12,99,50,0,0, +95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +59,122,0,18,102,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0,18,102,48, +51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,102,49,51,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,102,50,51,0,20,0,0,1,0,0,31,1,1,1,0,9,102,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0, +48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0, +17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,31,1,1,1,0,5,105,0,0,0,1,3,2,1,0, +9,1,102,0,2,58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97, +116,52,120,51,0,18,102,0,0,0,20,0,0,1,0,0,31,1,1,1,0,1,98,0,0,0,1,3,2,1,0,9,1,102,0,2,58,102,108, +111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,102,0, +0,0,20,0,0,1,0,0,31,1,1,1,0,11,99,48,0,0,1,1,0,11,99,49,0,0,1,1,0,11,99,50,0,0,1,1,0,11,99,51,0,0, 0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97, 108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20, -0,0,1,0,29,1,1,1,0,9,102,48,48,0,0,1,1,0,9,102,49,48,0,0,1,1,0,9,102,48,49,0,0,1,1,0,9,102,49,49,0, -0,1,1,0,9,102,48,50,0,0,1,1,0,9,102,49,50,0,0,1,1,0,9,102,48,51,0,0,1,1,0,9,102,49,51,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102, -49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,59,120,0,18,102,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57, -59,121,0,18,102,49,51,0,20,0,0,1,0,29,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, -109,97,116,52,120,50,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,16,10,52,0,0,17,48,0,48,0,0,0, -17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,29,1,1,1,0,5,105,0,0,0,1,3,2,1,9,1, -102,0,2,58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116, -52,120,50,0,18,102,0,0,0,20,0,0,1,0,29,1,1,1,0,1,98,0,0,0,1,3,2,1,9,1,102,0,2,58,102,108,111,97, -116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,102,0,0,0,20, -0,0,1,0,29,1,1,1,0,10,99,48,0,0,1,1,0,10,99,49,0,0,1,1,0,10,99,50,0,0,1,1,0,10,99,51,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,31,1,1,1,0,9,102,48,48,0,0,1,1,0, -9,102,49,48,0,0,1,1,0,9,102,50,48,0,0,1,1,0,9,102,48,49,0,0,1,1,0,9,102,49,49,0,0,1,1,0,9,102,50, -49,0,0,1,1,0,9,102,48,50,0,0,1,1,0,9,102,49,50,0,0,1,1,0,9,102,50,50,0,0,1,1,0,9,102,48,51,0,0,1,1, -0,9,102,49,51,0,0,1,1,0,9,102,50,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120, -0,18,102,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,102,49,48,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,102,50,48,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,59,120,0,18,102,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,59,121,0,18,102,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0, -18,102,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,102,48,50,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,102,49,50,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,59,122,0,18,102,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,51,0,57,59,120,0,18,102,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0, -18,102,49,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,102,50,51,0,20,0, -0,1,0,31,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,102, -0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0, -0,17,48,0,48,0,0,0,18,102,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,31,1, -1,1,0,5,105,0,0,0,1,3,2,1,9,1,102,0,2,58,102,108,111,97,116,0,18,105,0,0,0,0,0,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,52,120,51,0,18,102,0,0,0,20,0,0,1,0,31,1,1,1,0,1,98,0,0,0,1,3,2,1,9, -1,102,0,2,58,102,108,111,97,116,0,18,98,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116, -52,120,51,0,18,102,0,0,0,20,0,0,1,0,31,1,1,1,0,11,99,48,0,0,1,1,0,11,99,49,0,0,1,1,0,11,99,50,0,0, -1,1,0,11,99,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, -0,57,18,99,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,13,1,1, -1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,13,1,1,1,0,27,109,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49, -0,57,0,0,20,0,0,1,0,13,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50, -0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,13,1,1,1,0,26,109,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10, -49,0,57,59,120,121,0,0,0,20,0,0,1,0,13,1,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20, -0,0,1,0,13,1,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0, -16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,0,13,1,1,1,0,30,109,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0, -18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,0,13,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59, -120,121,0,0,0,20,0,0,1,0,13,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97, -116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,0, -26,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,26,1,1,1,0,14,109, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,0,13,1,1,1,0,13,109,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,0,13,1,1,1,0,27,109,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0, +0,1,0,0,13,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0, +16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,0,13,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57, +59,120,121,0,0,0,20,0,0,1,0,0,13,1,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, +97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1, +0,0,13,1,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8, +48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,0,0,13,1,1,1,0,30,109,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18, +109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,0,0,13,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120, +121,0,0,0,20,0,0,1,0,0,13,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116, +50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,0,20,0,0,1,0,0,26, +1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,0,26,1,1,1,0,14,109, 0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,0,18, -109,0,16,10,49,0,57,0,0,20,0,0,1,0,26,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, -109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,26,1,1,1,0,28, -109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59, -120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,0,20,0,0,1,0,26,1,1,1,0,30,109,0,0,0,1,9, +109,0,16,10,49,0,57,0,0,20,0,0,1,0,0,26,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,0,26,1,1,1, +0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0, +57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,0,20,0,0,1,0,0,26,1,1,1,0,30,109,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59,120, +121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,0,20,0,0,1,0,0,26,1,1,1,0,15,109,0,0,0,1,9, 18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0, -0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,0,20,0,0,1,0,26,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16, -10,49,0,57,59,120,121,122,0,0,0,20,0,0,1,0,26,1,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0, -17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0, -0,0,0,20,0,0,1,0,26,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120, +0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,0,20,0,0,1,0,0,26,1,1,1,0,13,109,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8, +48,0,57,59,121,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59, +121,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,26,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17, +48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0, +0,20,0,0,1,0,0,26,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120, 51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,0,18,109,0, -16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,26,1,1,1, -0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0, -57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0, -18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,28,1,1,1,0,28,109,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,28,1,1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0, -28,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16, -8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,28,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59, -121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109, -0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,28,1,1, -1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0, -57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0, -18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122, -0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,28,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, -109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0, -16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57, -59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,28,1,1,1,0,13,109,0,0,0, +16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,28,1,1, +1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,0,28,1,1,1,0,30,109,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,0,18,109,0, +16,10,49,0,57,0,0,20,0,0,1,0,0,28,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, +97,116,50,120,52,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,0,28,1,1,1,0,26, +109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59, +120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109, +0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17, +48,0,48,0,0,0,0,20,0,0,1,0,0,28,1,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, +97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,8, +48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59, +121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,28,1,1,1,0,31,109,0,0,0, 1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18, -109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0, -18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,28,1,1,1,0,27,109, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120, -0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59, -120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,28,1,1,1, -0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0, -57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49, -0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0, -27,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,27,1,1,1,0,29,109, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,0,18, -109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,27,1,1,1,0,14,109,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0, -18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,27,1,1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109, -0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18, -109,0,16,10,50,0,57,59,121,0,0,0,20,0,0,1,0,27,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0, -18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,120, -0,0,18,109,0,16,10,50,0,57,59,121,0,0,0,20,0,0,1,0,27,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57, -59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17, -48,0,48,0,0,0,0,20,0,0,1,0,27,1,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97, -116,51,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,58,118,101,99,50,0,17,48,0,48,0,0, -0,0,0,0,20,0,0,1,0,27,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51, -120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57, -59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,27,1,1, -1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0, -57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49, -0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,14,1,1,1,0,14,109,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0, -57,0,0,20,0,0,1,0,14,1,1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0, -18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10, -50,0,57,59,120,121,122,0,0,0,20,0,0,1,0,14,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121, -122,0,0,18,109,0,16,10,50,0,57,59,120,121,122,0,0,0,20,0,0,1,0,14,1,1,1,0,26,109,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17, -48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,14,1,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57, -59,120,121,122,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,14,1,1,1,0,27,109,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18, -109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,0,20,0,0,1,0,14,1,1,1,0, -29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,17,48, -0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,0,20,0,0,1,0, -14,1,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0, -57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0, -0,0,20,0,0,1,0,30,1,1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,30, -1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8, -48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,30,1,1,1,0,14,109,0,0,0,1, -9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0, -18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,0,20,0,0,1,0,30,1,1, -1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0, -57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,0, -20,0,0,1,0,30,1,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0, -18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0, -0,0,0,0,20,0,0,1,0,30,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51, -120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0, -0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,30,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0, -18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48, -0,0,0,0,0,20,0,0,1,0,30,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51, -120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0, -0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,30,1,1,1,0,13, -109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0, -17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17, -48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,29,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,109,0,20,0,0,1,0,29,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0, -0,18,109,0,16,10,50,0,57,59,120,121,0,0,18,109,0,16,10,51,0,57,59,120,121,0,0,0,20,0,0,1,0,29,1,1, -1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0, -57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,18, -109,0,16,10,51,0,57,59,120,121,0,0,0,20,0,0,1,0,29,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0, -0,17,48,0,0,0,0,0,20,0,0,1,0,29,1,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, -97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18, -109,0,16,10,50,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,29,1,1,1,0,30,109,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121, -0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,17,48,0,0,0,0,17, -48,0,0,0,0,0,20,0,0,1,0,29,1,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116, -52,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0, -0,0,17,48,0,0,0,0,0,20,0,0,1,0,29,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, -97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,17, -48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,29,1,1,1,0,28,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18, -109,0,16,10,49,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0, -0,1,0,31,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,31,1,1,1,0, -15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57, +109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49, +0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,122,0,0,17,48,0,48,0,0, +0,0,20,0,0,1,0,0,28,1,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,50,120, +52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0, +48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17, +48,0,48,0,0,0,0,20,0,0,1,0,0,28,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, +97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,17,48,0,48,0, +0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0, +48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,28,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0, +17,48,0,48,0,0,0,17,48,0,48,0,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121, +0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,27,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,109,0,20,0,0,1,0,0,27,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57, +0,0,20,0,0,1,0,0,27,1,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120, +50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,0,27,1, +1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48, +0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10, +49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,121,0,0,0,20,0,0,1, +0,0,27,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109, +0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18, +109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,121,0,0, +0,20,0,0,1,0,0,27,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120, +50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59, +120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,27,1,1, +1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0, +57,0,18,109,0,16,10,49,0,57,0,58,118,101,99,50,0,17,48,0,48,0,0,0,0,0,0,20,0,0,1,0,0,27,1,1,1,0,26, +109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59, +120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57, +59,121,0,0,17,48,0,48,0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,27,1,1,1,0,28,109,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,8, +48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,121,0,0,17,48,0,48, +0,0,0,17,48,0,48,0,0,0,0,20,0,0,1,0,0,14,1,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +18,109,0,20,0,0,1,0,0,14,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116, +51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,0,14,1, +1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57, 59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50,0,57,59,120,121,122, -0,0,18,109,0,16,10,51,0,57,59,120,121,122,0,0,0,20,0,0,1,0,31,1,1,1,0,14,109,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0, -18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,31,1,1,1,0,30,109, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,59,120, -121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50,0,57,59,120,121,122,0,0,17, -48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,31,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49, -0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,18,109,0,16,10,51,0,57,0,17,48,0,0,0,0, -0,20,0,0,1,0,31,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51, -0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48, -0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,31,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57, -0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0, -20,0,0,1,0,31,1,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0, -18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,17,48,0,0,0,0, -17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,31,1,1,1,0,13, +0,0,0,20,0,0,1,0,0,14,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0, +18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10, +50,0,57,59,120,121,122,0,0,0,20,0,0,1,0,0,14,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0, +0,0,17,49,0,0,0,0,0,20,0,0,1,0,0,14,1,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, +109,97,116,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0, +17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,0,14,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0, +57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,0,20,0,0,1,0,0,14,1,1,1,0,29,109,0,0,0,1, +9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109, +0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,0,20,0,0,1,0,0,14,1,1,1,0, +13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,0,18,109,0,16,8,48,0,57,0,17,48, +0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0, +1,0,0,30,1,1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,0,30,1,1,1, +0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0, +57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,0,30,1,1,1,0,14,109,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0, +18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,0,20,0,0,1,0,0,30,1, +1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48, +0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0, +0,20,0,0,1,0,0,30,1,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120, +52,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17, +48,0,0,0,0,0,20,0,0,1,0,0,30,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97, +116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48, +0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,30,1,1,1,0,27,109,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0, +0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0, +17,48,0,0,0,0,0,20,0,0,1,0,0,30,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, +97,116,51,120,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17, +48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,30,1, +1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,51,120,52,0,18,109,0,16,8,48, +0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0, +0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,29,1,1,1,0,29,109,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,0,29,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59, +120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,18,109,0,16,10,51,0,57,59,120,121,0,0,0,20,0,0,1, +0,0,29,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109, +0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120, +121,0,0,18,109,0,16,10,51,0,57,59,120,121,0,0,0,20,0,0,1,0,0,29,1,1,1,0,27,109,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0, +17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,29,1,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120, +121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,29,1,1,1, +0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0, +57,59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,18,109,0,16,10,50,0,57,59,120,121,0,0,17, +48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,29,1,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0, +0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,29,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,121,0,0,18,109,0,16,10,49,0,57, +59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,29,1,1,1,0, +28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57, +59,120,121,0,0,18,109,0,16,10,49,0,57,59,120,121,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17, +48,0,0,0,0,0,20,0,0,1,0,0,31,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20, +0,0,1,0,0,31,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0, +18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10, +50,0,57,59,120,121,122,0,0,18,109,0,16,10,51,0,57,59,120,121,122,0,0,0,20,0,0,1,0,0,31,1,1,1,0,14, 109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0, -17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17, -48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,15,1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,109,0,20,0,0,1,0,15,1,1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,17,48, -0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,15,1,1,1,0,31,109,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10, -49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,18,109,0,16,10,51,0,57,0,17,49,0,0, -0,0,0,20,0,0,1,0,15,1,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0, -18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0, -0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,15,1,1,1,0,29,109,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17, -48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0, -0,0,17,48,0,0,0,0,18,109,0,16,10,51,0,57,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,15,1,1,1,0,14, -109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0, -0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0, -17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,15,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0, -57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0, -0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,15,1,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49, -0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0, -0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,15,1,1,1,0,13,109,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18, -109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48, -0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,0,2,1,1,0,2,26,109,0, -0,1,1,0,26,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,1,1,0,2,28,109,0,0,1,1,0,28,110,0,0,0,1,9,18,109,0,16,8,48,0, -57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,1,1,0, -2,27,109,0,0,1,1,0,27,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16, -10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0, -0,2,1,1,0,2,30,109,0,0,1,1,0,30,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9, -18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57, -21,0,0,1,0,0,2,1,1,0,2,29,109,0,0,1,1,0,29,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0, -57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16, -10,50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,1,1,0,2,31,109,0,0, -1,1,0,31,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10, -51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,26,109,0,0,1,1,0,26,110,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,0, -0,2,2,1,0,2,28,109,0,0,1,1,0,28,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9, -18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,2,2,1,0,2,27,109,0,0,1,1,0,27,110,0,0,0, -1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0, -57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,0,0,2,2,1,0,2,30,109,0,0,1,1,0,30, -110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0, -16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,0,0,2,2,1,0,2,29,109,0, -0,1,1,0,29,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10, -51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,2,2,1,0,2,31,109,0,0,1,1,0,31,110,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18, -109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22, -0,0,1,0,0,2,4,1,0,2,26,109,0,0,1,1,0,26,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,4,1,0,2,28,109,0,0,1,1,0,28, -110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0, -16,10,49,0,57,24,0,0,1,0,0,2,4,1,0,2,27,109,0,0,1,1,0,27,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0, -57,18,110,0,16,10,50,0,57,24,0,0,1,0,0,2,4,1,0,2,30,109,0,0,1,1,0,30,110,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0, -16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,0,1,0,0,2,4,1,0,2,29,109,0,0,1,1,0,29,110,0,0,0,1,9,18, -109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0, -9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0, -57,24,0,0,1,0,0,2,4,1,0,2,31,109,0,0,1,1,0,31,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48, +18,109,0,16,10,49,0,57,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0, +0,1,0,0,31,1,1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18, +109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10,49,0,57,59,120,121,122,0,0,18,109,0,16,10,50, +0,57,59,120,121,122,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,31,1,1,1,0,29,109, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0,17,48, +0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,18,109,0,16, +10,51,0,57,0,17,48,0,0,0,0,0,20,0,0,1,0,0,31,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17, +48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,31,1,1,1,0,27, +109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,0, +17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48, +0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,31,1,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,59,120,121,122,0,0,18,109,0,16,10, +49,0,57,59,120,121,122,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0, +17,48,0,0,0,0,0,20,0,0,1,0,0,31,1,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109, +97,116,52,120,51,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17, +48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,0,20,0,0,1,0,0,15, +1,1,1,0,15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,0,15,1,1,1,0,30,109, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,18,109,0,16, +10,49,0,57,0,18,109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20, +0,0,1,0,0,15,1,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0, +16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,48, +0,0,0,0,18,109,0,16,10,51,0,57,0,17,49,0,0,0,0,0,20,0,0,1,0,0,15,1,1,1,0,28,109,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57,0,18,109,0,16,10,49,0,57,0,17,48, +0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49, +0,0,0,0,0,20,0,0,1,0,0,15,1,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116, +52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17, +48,0,0,0,0,18,109,0,16,10,50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,51,0,57,0,17,48,0,0, +0,0,17,49,0,0,0,0,0,20,0,0,1,0,0,15,1,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58, +109,97,116,52,0,18,109,0,16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,18, +109,0,16,10,50,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0, +0,1,0,0,15,1,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0, +16,8,48,0,57,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17, +49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20,0,0,1,0,0,15, +1,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0,16,8,48,0,57, +0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10, +50,0,57,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,0,20, +0,0,1,0,0,15,1,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,109,97,116,52,0,18,109,0, +16,8,48,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,18,109,0,16,10,49,0,57,0,17,48,0,0,0,0,17,48,0,0,0,0,17, +48,0,0,0,0,17,48,0,0,0,0,17,49,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17,48,0,0,0,0,17, +49,0,0,0,0,0,20,0,0,1,0,0,0,2,1,1,0,2,26,109,0,0,1,1,0,26,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,0,2,1,1,0,2, +28,109,0,0,1,1,0,28,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16, +10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,0,2,1,1,0,2,27,109,0,0,1,1,0,27,110,0,0,0,1,9,18, +109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0, +9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,0,2,1,1,0,2,30,109,0,0,1,1,0,30,110,0, +0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10, +49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,0,2,1,1,0,2,29,109,0,0,1, +1,0,29,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,51, +0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,0,2,1,1,0,2,31,109,0,0,1,1,0,31,110,0,0,0,1,9,18,109,0,16, +8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109, +0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0, +1,0,0,0,2,2,1,0,2,26,109,0,0,1,1,0,26,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22, +0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,0,2,2,1,0,2,28,109,0,0,1,1,0,28,110, +0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10, +49,0,57,22,0,0,1,0,0,0,2,2,1,0,2,27,109,0,0,1,1,0,27,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0, +16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18, +110,0,16,10,50,0,57,22,0,0,1,0,0,0,2,2,1,0,2,30,109,0,0,1,1,0,30,110,0,0,0,1,9,18,109,0,16,8,48,0, +57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16, +10,50,0,57,18,110,0,16,10,50,0,57,22,0,0,1,0,0,0,2,2,1,0,2,29,109,0,0,1,1,0,29,110,0,0,0,1,9,18, +109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0, +9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0, +57,22,0,0,1,0,0,0,2,2,1,0,2,31,109,0,0,1,1,0,31,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8, +48,0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110, +0,16,10,50,0,57,22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,0,2,4,1,0,2,26, +109,0,0,1,1,0,26,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,0,2,4,1,0,2,28,109,0,0,1,1,0,28,110,0,0,0,1,9,18,109,0, +16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0, +0,0,2,4,1,0,2,27,109,0,0,1,1,0,27,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9, +18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57, +24,0,0,1,0,0,0,2,4,1,0,2,30,109,0,0,1,1,0,30,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48, 0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0, -16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,11,2,21,1,1,0,26,109, -0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16, -8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0, -18,109,0,16,10,49,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59, -120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,20, -0,0,1,0,12,2,21,1,1,0,28,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0, -18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59, -120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0, -57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109, -0,16,10,49,0,57,59,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,118,0,59,120,0, -18,109,0,16,8,48,0,57,59,119,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,119,0,48,46,20,0,0,1, -0,10,2,21,1,1,0,27,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118, -0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48, -46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,50,0,57,24,0,0,1,0,0,0,2,4,1,0,2,29,109,0,0,1,1,0,29,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0, +57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,0,2,4, +1,0,2,31,109,0,0,1,1,0,31,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109, +0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,9, +18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,11,2,21,1,1,0,26,109,0,0,1,1,0,10,118,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0, +48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0, 59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0, -57,59,121,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,20,0,0,1,0,12,2,21,1,1,0, -30,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18, +57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8, +48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,20,0,0,1,0,0,12,2,21,1, +1,0,28,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0, +18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48,18, +118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59, +122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57, +59,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,118,0,59,120,0,18,109,0,16,8,48, +0,57,59,119,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,119,0,48,46,20,0,0,1,0,0,10,2,21,1,1, +0,27,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18, 109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59, 122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18, 118,0,59,120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0, -48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16, -10,49,0,57,59,122,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,122,0,48,46,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,119,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,119,0,48,18,118,0,59, -121,0,18,109,0,16,10,49,0,57,59,119,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,119,0,48, -46,20,0,0,1,0,10,2,21,1,1,0,29,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57, -59,120,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,119,0,18,109,0, -16,10,51,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18, -109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59, -122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,48, -46,20,0,0,1,0,11,2,21,1,1,0,31,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57, -59,120,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,119,0,18,109,0, -16,10,51,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18, -109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59, -122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,48, -46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122, -0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0, -57,59,122,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,122,0,48,46,20,0,0,1,0,27,2,21,1,1,0, -13,109,0,0,1,1,0,27,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0, -16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0, -57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20, -0,0,1,0,29,2,21,1,1,0,13,109,0,0,1,1,0,29,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,20,0,0,1,0,0,12,2,21,1,1,0,30,109,0, +0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118,0,59,120,0,18,109,0,16,8, +48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48,46,18,118,0,59,122,0,18, +109,0,16,10,50,0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59, +120,0,18,109,0,16,8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18, +118,0,59,122,0,18,109,0,16,10,50,0,57,59,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59, +122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57, +59,122,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,122,0,48,46,20,0,9,18,95,95,114,101,116, +86,97,108,0,59,119,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,119,0,48,18,118,0,59,121,0,18,109, +0,16,10,49,0,57,59,119,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,119,0,48,46,20,0,0,1,0, +0,10,2,21,1,1,0,29,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,118, +0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0,48, +46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0, +57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16,8, +48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18, +109,0,16,10,50,0,57,59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,48,46,20,0,0, +1,0,0,11,2,21,1,1,0,31,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18, +118,0,59,120,0,18,109,0,16,8,48,0,57,59,120,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,120,0, +48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59,120,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51, +0,57,59,120,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,118,0,59,120,0,18,109,0,16, +8,48,0,57,59,121,0,48,18,118,0,59,121,0,18,109,0,16,10,49,0,57,59,121,0,48,46,18,118,0,59,122,0,18, +109,0,16,10,50,0,57,59,121,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,121,0,48,46,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,122,0,18,118,0,59,120,0,18,109,0,16,8,48,0,57,59,122,0,48,18, +118,0,59,121,0,18,109,0,16,10,49,0,57,59,122,0,48,46,18,118,0,59,122,0,18,109,0,16,10,50,0,57,59, +122,0,48,46,18,118,0,59,119,0,18,109,0,16,10,51,0,57,59,122,0,48,46,20,0,0,1,0,0,27,2,21,1,1,0,13, +109,0,0,1,1,0,27,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16, +8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57, +48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0, +1,0,0,29,2,21,1,1,0,13,109,0,0,1,1,0,29,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, 57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109, 0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110, 0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10, -51,0,57,48,20,0,0,1,0,26,2,21,1,1,0,26,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +51,0,57,48,20,0,0,1,0,0,26,2,21,1,1,0,26,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, 108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,14,2,21,1,1,0,26,109,0,0,1,1,0,27,110,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95, +49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,0,14,2,21,1,1,0,26,109,0,0,1,1,0,27,110,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95, 95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,31,2,21,1,1,0,26, +101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,0,31,2,21,1,1,0,26, 109,0,0,1,1,0,29,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16, 8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57, 48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,28,2, -21,1,1,0,28,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0, -16,10,49,0,57,48,20,0,0,1,0,30,2,21,1,1,0,28,109,0,0,1,1,0,27,110,0,0,0,1,9,18,95,95,114,101,116, +18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,0,28, +2,21,1,1,0,28,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, +0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0, +16,10,49,0,57,48,20,0,0,1,0,0,30,2,21,1,1,0,28,109,0,0,1,1,0,27,110,0,0,0,1,9,18,95,95,114,101,116, 86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0, 16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, -0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,15,2,21,1,1,0,28,109,0,0,1,1,0,29,110,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95, +0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,0,15,2,21,1,1,0,28,109,0,0,1,1,0,29,110,0,0,0,1, +9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95, 114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101, 116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,13,2,21,1,1,0,27,109,0,0,1,1,0, +108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,0,13,2,21,1,1,0,27,109,0,0,1,1,0, 26,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48, 20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1, -0,27,2,21,1,1,0,27,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18, -110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16, -10,50,0,57,48,20,0,0,1,0,29,2,21,1,1,0,27,109,0,0,1,1,0,31,110,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109, -0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,26,2,21,1,1,0,14,109,0,0,1,1,0,26,110,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,31,2,21,1,1,0,14,109,0,0,1, -1,0,31,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,28,2,21,1,1,0, -30,109,0,0,1,1,0,26,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0, -16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0, -57,48,20,0,0,1,0,30,2,21,1,1,0,30,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109, -0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,15,2,21,1,1,0,30,109,0,0,1,1,0,31,110,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,13,2,21,1,1,0,29,109,0,0,1,1,0,28,110,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,27,2,21,1,1, -0,29,109,0,0,1,1,0,30,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110, -0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49, -0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48, -20,0,0,1,0,29,2,21,1,1,0,29,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18, -110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16, -10,51,0,57,48,20,0,0,1,0,26,2,21,1,1,0,31,109,0,0,1,1,0,28,110,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,14,2,21,1,1,0,31,109,0,0,1,1,0,30,110,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95, -95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,31,2,21,1,1,0,31, -109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16, -8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57, -48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,28,2, -21,1,1,0,15,109,0,0,1,1,0,28,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +0,0,27,2,21,1,1,0,27,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, +18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0, +16,10,50,0,57,48,20,0,0,1,0,0,29,2,21,1,1,0,27,109,0,0,1,1,0,31,110,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, +0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18, +109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,0,26,2,21,1,1,0,14,109,0,0,1,1,0,26,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,0,31,2,21,1,1,0,14,109, +0,0,1,1,0,31,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48, +0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,0,28,2, +21,1,1,0,30,109,0,0,1,1,0,26,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, 18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0, -16,10,49,0,57,48,20,0,0,1,0,30,2,21,1,1,0,15,109,0,0,1,1,0,30,110,0,0,0,1,9,18,95,95,114,101,116, +16,10,49,0,57,48,20,0,0,1,0,0,30,2,21,1,1,0,30,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116, 86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0, 16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, -0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,0,2,3,1,0,2,26,109,0,0,1,1,0,13,110,0,0,0,1,9, -18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2,3,1,0,2,28,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18, -109,0,18,110,0,48,20,0,0,1,0,0,2,3,1,0,2,27,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18, -110,0,48,20,0,0,1,0,0,2,3,1,0,2,30,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20, -0,0,1,0,0,2,3,1,0,2,29,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2, -3,1,0,2,31,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,11,2,21,1,1,0, -10,118,0,0,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18, +0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,0,15,2,21,1,1,0,30,109,0,0,1,1,0,31,110,0,0,0,1, +9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,0,13,2,21,1,1,0,29,109,0,0,1,1,0, +28,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1, +0,0,27,2,21,1,1,0,29,109,0,0,1,1,0,30,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, +18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0, +16,10,50,0,57,48,20,0,0,1,0,0,29,2,21,1,1,0,29,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50, +0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18, +109,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,0,26,2,21,1,1,0,31,109,0,0,1,1,0,28,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,0,14,2,21,1,1,0,31,109, +0,0,1,1,0,30,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48, +0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1, +0,0,31,2,21,1,1,0,31,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, +18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0, +16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,18,110,0,16,10,51, +0,57,48,20,0,0,1,0,0,28,2,21,1,1,0,15,109,0,0,1,1,0,28,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,0,30,2,21,1,1,0,15,109,0,0,1,1,0,30,110,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,50,0,57,18,109,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,0,0,2,3,1,0,2,26, +109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,0,2,3,1,0,2,28,109,0,0,1, +1,0,13,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,0,2,3,1,0,2,27,109,0,0,1,1,0,14, +110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,0,2,3,1,0,2,30,109,0,0,1,1,0,14,110,0,0,0, +1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,0,2,3,1,0,2,29,109,0,0,1,1,0,15,110,0,0,0,1,9,18, +109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,0,2,3,1,0,2,31,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18, +109,0,18,110,0,48,20,0,0,1,0,0,11,2,21,1,1,0,10,118,0,0,1,1,0,27,109,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0, +1,0,0,12,2,21,1,1,0,10,118,0,0,1,1,0,29,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58, +100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121, +0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0,0,1,0,0,10,2,21,1,1,0, +11,118,0,0,1,1,0,26,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18, 118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116, -0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100, -111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,12,2,21,1,1,0,10,118,0,0,1,1,0,29,109,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0, -57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10, -49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0, -16,10,50,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18, -109,0,16,10,51,0,57,0,0,20,0,0,1,0,10,2,21,1,1,0,11,118,0,0,1,1,0,26,109,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0, -12,2,21,1,1,0,11,118,0,0,1,1,0,31,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100, +0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,0,12,2,21,1,1,0,11,118,0,0,1,1,0,31,109,0,0,0,1, +9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0, +57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10, +50,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0, +16,10,51,0,57,0,0,20,0,0,1,0,0,10,2,21,1,1,0,12,118,0,0,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,0, +11,2,21,1,1,0,12,118,0,0,1,1,0,30,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100, 111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, 100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122, -0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,51,0,57,0,0,20,0,0,1,0,10,2,21,1,1,0,12,118,0, -0,1,1,0,28,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18, -109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0, -0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21,1,1,0,12,118,0,0,1,1,0,30,109,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0, -20,0,0,1,0,0,2,1,1,0,2,26,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109, -0,16,10,49,0,57,18,97,0,21,0,0,1,0,0,2,1,1,0,2,28,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0, -57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,0,0,2,1,1,0,2,27,109,0,0,1,1,0,9,97,0,0, -0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50, -0,57,18,97,0,21,0,0,1,0,0,2,1,1,0,2,30,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, -21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,1,1,0,2, -29,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97, -0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,2,1,1,0, -2,31,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18, -97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,2,2, -1,0,2,26,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57, -18,97,0,22,0,0,1,0,0,2,2,1,0,2,28,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0, -9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,2,2,1,0,2,27,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0, -22,0,0,1,0,0,2,2,1,0,2,30,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109, -0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,0,0,2,2,1,0,2,29,109,0,0,1, -1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18, -109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,2,2,1,0,2,31,109,0, -0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9, -18,109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,26,109, -0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0, -0,1,0,0,2,3,1,0,2,28,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16, -10,49,0,57,18,97,0,23,0,0,1,0,0,2,3,1,0,2,27,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18, -97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,3, -1,0,2,30,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57, -18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,3,1,0,2,29,109,0,0,1,1,0,9,97,0,0,0,1, +0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0,57,0,0,20,0,0,1,0,0,0,2,1,1,0,2,26,109,0,0,1,1,0, +9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,0,0,0, +2,1,1,0,2,28,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, +57,18,97,0,21,0,0,1,0,0,0,2,1,1,0,2,27,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,0,2,1,1,0, +2,30,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18, +97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,0,2,1,1,0,2,29,109,0,0,1,1,0,9,97,0,0,0,1, +9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0, +57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,0,2,1,1,0,2,31,109,0,0,1,1,0,9,97,0, +0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10, +50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,0,2,2,1,0,2,26,109,0,0,1,1,0,9, +97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,0,2, +2,1,0,2,28,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0, +57,18,97,0,22,0,0,1,0,0,0,2,2,1,0,2,27,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,0,0,0,2,2,1,0, +2,30,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18, +97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,0,0,0,2,2,1,0,2,29,109,0,0,1,1,0,9,97,0,0,0,1, +9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0, +57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,0,2,2,1,0,2,31,109,0,0,1,1,0,9,97,0, +0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10, +50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,0,2,3,1,0,2,26,109,0,0,1,1,0,9, +97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,0,2, +3,1,0,2,28,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0, +57,18,97,0,23,0,0,1,0,0,0,2,3,1,0,2,27,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,0,2,3,1,0, +2,30,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18, +97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,0,2,3,1,0,2,29,109,0,0,1,1,0,9,97,0,0,0,1, 9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0, -57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,3,1,0,2,31,109,0,0,1,1,0,9,97,0,0, -0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50, -0,57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,26,109,0,0,1,1,0,9,97,0, -0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2,4,1,0,2, -28,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97, -0,24,0,0,1,0,0,2,4,1,0,2,27,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18, -109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,2,4,1,0,2,30,109,0, -0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9, -18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,2,4,1,0,2,29,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16, -8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0, -9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,4,1,0,2,31,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0, -24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,26,2,26,1,1,0,26,109,0,0,1,1,0,26,110,0,0,0,1,8, -58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,46,0,0,0,0,1,0,28,2,26,1,1,0,28,109,0,0,1,1,0,28,110,0,0,0,1,8,58,109,97, -116,50,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0, -16,10,49,0,57,46,0,0,0,0,1,0,27,2,26,1,1,0,27,109,0,0,1,1,0,27,110,0,0,0,1,8,58,109,97,116,51,120, -50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0, -57,46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,0,0,0,1,0,30,2,26,1,1,0,30,109,0,0,1,1, -0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0, -0,0,0,1,0,29,2,26,1,1,0,29,109,0,0,1,1,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8, -48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16, -10,50,0,57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,0,0,0,0,1, -0,31,2,26,1,1,0,31,109,0,0,1,1,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16,10,50,0, -57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,0,0,0,0,1,0,26,2, -27,1,1,0,26,109,0,0,1,1,0,26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,110, -0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,0,0,0,1,0,28,2,27,1,1,0,28, -109,0,0,1,1,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0, -57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,0,0,0,1,0,27,2,27,1,1,0,27,109,0,0,1,1, -0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0, -0,0,0,1,0,30,2,27,1,1,0,30,109,0,0,1,1,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8, -48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16, -10,50,0,57,18,110,0,16,10,50,0,57,47,0,0,0,0,1,0,29,2,27,1,1,0,29,109,0,0,1,1,0,29,110,0,0,0,1,8, -58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,18,109,0,16,10,51,0, -57,18,110,0,16,10,51,0,57,47,0,0,0,0,1,0,31,2,27,1,1,0,31,109,0,0,1,1,0,31,110,0,0,0,1,8,58,109,97, -116,52,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0, -16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,18,109,0,16,10,51,0,57,18, -110,0,16,10,51,0,57,47,0,0,0,0,1,0,26,2,22,1,1,0,26,109,0,0,1,1,0,26,110,0,0,0,1,8,58,109,97,116, -50,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10, -49,0,57,49,0,0,0,0,1,0,28,2,22,1,1,0,28,109,0,0,1,1,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0, +57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,0,2,3,1,0,2,31,109,0,0,1,1,0,9,97,0, +0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10, +50,0,57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,0,2,4,1,0,2,26,109,0,0,1,1,0,9, +97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,0,2, +4,1,0,2,28,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0, +57,18,97,0,24,0,0,1,0,0,0,2,4,1,0,2,27,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,0,2,4,1,0, +2,30,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18, +97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,0,2,4,1,0,2,29,109,0,0,1,1,0,9,97,0,0,0,1, +9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0, +57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,0,2,4,1,0,2,31,109,0,0,1,1,0,9,97,0, +0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10, +50,0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,26,2,26,1,1,0,26,109,0,0,1,1,0, +26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109, +0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,0,0,0,1,0,0,28,2,26,1,1,0,28,109,0,0,1,1,0,28,110,0,0, +0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49, +0,57,18,110,0,16,10,49,0,57,46,0,0,0,0,1,0,0,27,2,26,1,1,0,27,109,0,0,1,1,0,27,110,0,0,0,1,8,58, +109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,0,0,0,1,0,0,30,2,26,1, +1,0,30,109,0,0,1,1,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16, +8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,0,18,109,0,16,10,50,0,57,18,110,0, +16,10,50,0,57,46,0,0,0,0,1,0,0,29,2,26,1,1,0,29,109,0,0,1,1,0,29,110,0,0,0,1,8,58,109,97,116,52, +120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49, +0,57,46,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0,16,10, +51,0,57,46,0,0,0,0,1,0,0,31,2,26,1,1,0,31,109,0,0,1,1,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0, +18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46, +0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, +46,0,0,0,0,1,0,0,26,2,27,1,1,0,26,109,0,0,1,1,0,26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109, +0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,0,0,0, +1,0,0,28,2,27,1,1,0,28,109,0,0,1,1,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0, +57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,0,0,0,1,0,0,27,2, +27,1,1,0,27,109,0,0,1,1,0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,110, +0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18, +110,0,16,10,50,0,57,47,0,0,0,0,1,0,0,30,2,27,1,1,0,30,109,0,0,1,1,0,30,110,0,0,0,1,8,58,109,97,116, +51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109,0,16,10,49,0,57,18,110,0,16,10, +49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,0,0,0,1,0,0,29,2,27,1,1,0,29,109,0, +0,1,1,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47, +0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57, +47,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,0,0,0,0,1,0,0,31,2,27,1,1,0,31,109,0,0,1,1,0, +31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,0,18,109, +0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,0,18, +109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,0,0,0,0,1,0,0,26,2,22,1,1,0,26,109,0,0,1,1,0,26,110, +0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16, +10,49,0,57,18,110,0,16,10,49,0,57,49,0,0,0,0,1,0,0,28,2,22,1,1,0,28,109,0,0,1,1,0,28,110,0,0,0,1,8, +58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,49,0,0,0,0,1,0,0,27,2,22,1,1,0,27,109,0,0,1,1,0,27,110,0,0,0,1,8,58,109,97, +116,51,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0, +16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,0,0,0,1,0,0,30,2,22,1,1,0,30, +109,0,0,1,1,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0, +57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,10, +50,0,57,49,0,0,0,0,1,0,0,29,2,22,1,1,0,29,109,0,0,1,1,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0, 18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49, -0,0,0,0,1,0,27,2,22,1,1,0,27,109,0,0,1,1,0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16, -8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,18,109,0, -16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,0,0,0,1,0,30,2,22,1,1,0,30,109,0,0,1,1,0,30,110,0,0,0,1, -8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0, -57,18,110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,0,0,0,1,0,29,2, -22,1,1,0,29,109,0,0,1,1,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,18,110, -0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18, -110,0,16,10,50,0,57,49,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,0,0,0,0,1,0,31,2,22,1,1, -0,31,109,0,0,1,1,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8, -48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,18,109,0,16,10,50,0,57,18,110,0,16, -10,50,0,57,49,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,0,0,0,0,1,0,26,2,26,1,1,0,9,97,0, -0,1,1,0,26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18, -110,0,16,10,49,0,57,46,0,0,0,0,1,0,26,2,26,1,1,0,26,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,50, -120,51,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,0,0,0,1,0,28,2,26, -1,1,0,9,97,0,0,1,1,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,97,0,18,110,0,16,8,48,0,57,46,0, -18,97,0,18,110,0,16,10,49,0,57,46,0,0,0,0,1,0,28,2,26,1,1,0,28,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109, -97,116,50,120,52,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,0,0,0,1, -0,27,2,26,1,1,0,9,97,0,0,1,1,0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,97,0,18,110,0,16,8,48, -0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,0,0,0,1,0,27,2, -26,1,1,0,27,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,98,0, -46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,0,0,0,0,1,0,30,2,26,1,1, -0,9,97,0,0,1,1,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18, -97,0,18,110,0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,0,0,0,1,0,30,2,26,1,1,0,30, -109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0, -16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,0,0,0,0,1,0,29,2,26,1,1,0,29,109,0,0, -1,1,0,9,98,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10, -49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,0,18,109,0,16,10,51,0,57,18,98,0,46,0,0,0,0, -1,0,29,2,26,1,1,0,9,97,0,0,1,1,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0,18,97,0,18,110,0,16,8, -48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,18,97,0,18, -110,0,16,10,51,0,57,46,0,0,0,0,1,0,31,2,26,1,1,0,31,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,52, -120,51,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50, -0,57,18,98,0,46,0,18,109,0,16,10,51,0,57,18,98,0,46,0,0,0,0,1,0,31,2,26,1,1,0,9,97,0,0,1,1,0,31, -110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10, -49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,18,97,0,18,110,0,16,10,51,0,57,46,0,0,0,0,1,0,26, -2,27,1,1,0,9,97,0,0,1,1,0,26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,97,0,18,110,0,16,8,48,0,57, -47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,0,0,0,1,0,26,2,27,1,1,0,26,109,0,0,1,1,0,9,98,0,0,0,1,8, -58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,0, -0,0,1,0,28,2,27,1,1,0,9,97,0,0,1,1,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,97,0,18,110,0, -16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,0,0,0,1,0,28,2,27,1,1,0,28,109,0,0,1,1,0,9, -98,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57, -18,98,0,47,0,0,0,0,1,0,27,2,27,1,1,0,9,97,0,0,1,1,0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18, -97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57, -47,0,0,0,0,1,0,27,2,27,1,1,0,27,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16, -8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,0,0, -0,1,0,30,2,27,1,1,0,9,97,0,0,1,1,0,30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,97,0,18,110,0,16, -8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,47,0,0,0,0,1,0, -30,2,27,1,1,0,30,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18, -98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,0,0,0,1,0,29,2, -27,1,1,0,29,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,18,98,0, -47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,18,109,0,16,10,51,0, -57,18,98,0,47,0,0,0,0,1,0,29,2,27,1,1,0,9,97,0,0,1,1,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0, -18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0, -57,47,0,18,97,0,18,110,0,16,10,51,0,57,47,0,0,0,0,1,0,31,2,27,1,1,0,31,109,0,0,1,1,0,9,98,0,0,0,1, -8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0, -18,109,0,16,10,50,0,57,18,98,0,47,0,18,109,0,16,10,51,0,57,18,98,0,47,0,0,0,0,1,0,31,2,27,1,1,0,9, -97,0,0,1,1,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0, -18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,47,0,18,97,0,18,110,0,16,10,51,0,57,47, -0,0,0,0,1,0,26,2,21,1,1,0,9,97,0,0,1,1,0,26,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, -0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,26,2,21,1,1,0,26,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,28,2,21,1,1,0,9,97,0,0,1,1,0, -28,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,28, -2,21,1,1,0,28,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,98,0,48,20,0,0,1,0,27,2,21,1,1,0,9,97,0,0,1,1,0,27,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,27,2,21,1,1,0,27,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0,30,2,21,1,1,0,9,97,0,0,1,1,0,30,110,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,30,2,21,1,1,0,30, -109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, -18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0, -29,2,21,1,1,0,29,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, -0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, -0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0, -0,1,0,29,2,21,1,1,0,9,97,0,0,1,1,0,29,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, -18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18, -110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16, -10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57, -48,20,0,0,1,0,31,2,21,1,1,0,31,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16, -10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57, -18,98,0,48,20,0,0,1,0,31,2,21,1,1,0,9,97,0,0,1,1,0,31,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0, -18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0, -16,10,51,0,57,48,20,0,0,1,0,26,2,22,1,1,0,9,97,0,0,1,1,0,26,110,0,0,0,1,3,2,1,9,1,105,110,118,0,2, +0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, +49,0,0,0,0,1,0,0,31,2,22,1,1,0,31,109,0,0,1,1,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109, +0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,0,18, +109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,0, +0,0,0,1,0,0,26,2,26,1,1,0,9,97,0,0,1,1,0,26,110,0,0,0,1,8,58,109,97,116,50,120,51,0,18,97,0,18,110, +0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,0,0,0,1,0,0,26,2,26,1,1,0,26,109,0,0,1,1,0, +9,98,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57, +18,98,0,46,0,0,0,0,1,0,0,28,2,26,1,1,0,9,97,0,0,1,1,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0, +18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,0,0,0,1,0,0,28,2,26,1,1,0, +28,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18, +109,0,16,10,49,0,57,18,98,0,46,0,0,0,0,1,0,0,27,2,26,1,1,0,9,97,0,0,1,1,0,27,110,0,0,0,1,8,58,109, +97,116,51,120,50,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,18,97,0, +18,110,0,16,10,50,0,57,46,0,0,0,0,1,0,0,27,2,26,1,1,0,27,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97, +116,51,120,50,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16, +10,50,0,57,18,98,0,46,0,0,0,0,1,0,0,30,2,26,1,1,0,9,97,0,0,1,1,0,30,110,0,0,0,1,8,58,109,97,116,51, +120,52,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,18,97,0,18,110,0, +16,10,50,0,57,46,0,0,0,0,1,0,0,30,2,26,1,1,0,30,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,51,120, +52,0,18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57, +18,98,0,46,0,0,0,0,1,0,0,29,2,26,1,1,0,29,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,52,120,50,0, +18,109,0,16,8,48,0,57,18,98,0,46,0,18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18, +98,0,46,0,18,109,0,16,10,51,0,57,18,98,0,46,0,0,0,0,1,0,0,29,2,26,1,1,0,9,97,0,0,1,1,0,29,110,0,0, +0,1,8,58,109,97,116,52,120,50,0,18,97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57, +46,0,18,97,0,18,110,0,16,10,50,0,57,46,0,18,97,0,18,110,0,16,10,51,0,57,46,0,0,0,0,1,0,0,31,2,26,1, +1,0,31,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,18,98,0,46,0, +18,109,0,16,10,49,0,57,18,98,0,46,0,18,109,0,16,10,50,0,57,18,98,0,46,0,18,109,0,16,10,51,0,57,18, +98,0,46,0,0,0,0,1,0,0,31,2,26,1,1,0,9,97,0,0,1,1,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18, +97,0,18,110,0,16,8,48,0,57,46,0,18,97,0,18,110,0,16,10,49,0,57,46,0,18,97,0,18,110,0,16,10,50,0,57, +46,0,18,97,0,18,110,0,16,10,51,0,57,46,0,0,0,0,1,0,0,26,2,27,1,1,0,9,97,0,0,1,1,0,26,110,0,0,0,1,8, +58,109,97,116,50,120,51,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0,0, +0,0,1,0,0,26,2,27,1,1,0,26,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48, +0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,0,0,0,1,0,0,28,2,27,1,1,0,9,97,0,0,1,1,0,28, +110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10, +49,0,57,47,0,0,0,0,1,0,0,28,2,27,1,1,0,28,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,50,120,52,0, +18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,0,0,0,1,0,0,27,2,27,1,1,0,9, +97,0,0,1,1,0,27,110,0,0,0,1,8,58,109,97,116,51,120,50,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0, +18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,47,0,0,0,0,1,0,0,27,2,27,1,1,0,27,109,0, +0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10, +49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,0,0,0,1,0,0,30,2,27,1,1,0,9,97,0,0,1,1,0, +30,110,0,0,0,1,8,58,109,97,116,51,120,52,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16, +10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,47,0,0,0,0,1,0,0,30,2,27,1,1,0,30,109,0,0,1,1,0,9, +98,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57, +18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,0,0,0,1,0,0,29,2,27,1,1,0,29,109,0,0,1,1,0,9,98,0, +0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98, +0,47,0,18,109,0,16,10,50,0,57,18,98,0,47,0,18,109,0,16,10,51,0,57,18,98,0,47,0,0,0,0,1,0,0,29,2,27, +1,1,0,9,97,0,0,1,1,0,29,110,0,0,0,1,8,58,109,97,116,52,120,50,0,18,97,0,18,110,0,16,8,48,0,57,47,0, +18,97,0,18,110,0,16,10,49,0,57,47,0,18,97,0,18,110,0,16,10,50,0,57,47,0,18,97,0,18,110,0,16,10,51, +0,57,47,0,0,0,0,1,0,0,31,2,27,1,1,0,31,109,0,0,1,1,0,9,98,0,0,0,1,8,58,109,97,116,52,120,51,0,18, +109,0,16,8,48,0,57,18,98,0,47,0,18,109,0,16,10,49,0,57,18,98,0,47,0,18,109,0,16,10,50,0,57,18,98,0, +47,0,18,109,0,16,10,51,0,57,18,98,0,47,0,0,0,0,1,0,0,31,2,27,1,1,0,9,97,0,0,1,1,0,31,110,0,0,0,1,8, +58,109,97,116,52,120,51,0,18,97,0,18,110,0,16,8,48,0,57,47,0,18,97,0,18,110,0,16,10,49,0,57,47,0, +18,97,0,18,110,0,16,10,50,0,57,47,0,18,97,0,18,110,0,16,10,51,0,57,47,0,0,0,0,1,0,0,26,2,21,1,1,0, +9,97,0,0,1,1,0,26,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16, +8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57, +48,20,0,0,1,0,0,26,2,21,1,1,0,26,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16, +8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, +18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,0,28,2,21,1,1,0,9,97,0,0,1,1,0,28,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,0,28,2,21,1,1,0,28,109, +0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98, +0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0, +0,1,0,0,27,2,21,1,1,0,9,97,0,0,1,1,0,27,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0, +18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0, +16,10,50,0,57,48,20,0,0,1,0,0,27,2,21,1,1,0,27,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, +57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0,0,30,2,21,1,1,0,9,97,0,0,1,1,0,30,110,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,0,30,2,21,1,1,0,30,109,0,0,1,1, +0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0,0,29,2,21,1, +1,0,29,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98, +0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,48,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20,0,0,1,0,0,29, +2,21,1,1,0,9,97,0,0,1,1,0,29,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0, +18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16, +10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57, +48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0, +1,0,0,31,2,21,1,1,0,31,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16, +10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57, +18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48, +20,0,0,1,0,0,31,2,21,1,1,0,9,97,0,0,1,1,0,31,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +97,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18, +110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16, +10,51,0,57,48,20,0,0,1,0,0,26,2,22,1,1,0,9,97,0,0,1,1,0,26,110,0,0,0,1,3,2,1,0,9,1,105,110,118,0,2, 17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18, 110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18, -110,0,16,10,49,0,57,48,20,0,0,1,0,26,2,22,1,1,0,26,109,0,0,1,1,0,9,98,0,0,0,1,3,2,1,9,1,105,110, -118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, -49,0,57,18,105,110,118,0,48,20,0,0,1,0,28,2,22,1,1,0,9,97,0,0,1,1,0,28,110,0,0,0,1,3,2,1,9,1,105, -110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105, -110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105, -110,118,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,28,2,22,1,1,0,28,109,0,0,1,1,0,9,98,0,0,0,1,3,2,1,9, -1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, -109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,0,1,0,27,2,22,1,1,0,9,97,0,0,1,1,0,27,110,0,0,0,1,3,2, -1,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,27,2,22,1,1,0,27,109,0,0,1,1,0,9,98,0,0,0,1, -3,2,1,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +110,0,16,10,49,0,57,48,20,0,0,1,0,0,26,2,22,1,1,0,26,109,0,0,1,1,0,9,98,0,0,0,1,3,2,1,0,9,1,105, +110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, +16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16, +10,49,0,57,18,105,110,118,0,48,20,0,0,1,0,0,28,2,22,1,1,0,9,97,0,0,1,1,0,28,110,0,0,0,1,3,2,1,0,9, +1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, +105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,0,28,2,22,1,1,0,28,109,0,0,1,1,0,9,98,0,0,0,1,3, +2,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48, 0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,0,30,2,22,1,1,0,9,97,0,0,1,1,0,30,110,0,0,0, -1,3,2,1,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,30,2,22,1,1,0,30,109,0,0,1,1,0,9,98,0,0,0, -1,3,2,1,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,0,29,2,22,1,1,0,29,109,0,0,1,1,0,9,98,0,0,0, -1,3,2,1,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0, -57,18,109,0,16,10,51,0,57,18,105,110,118,0,48,20,0,0,1,0,29,2,22,1,1,0,9,97,0,0,1,1,0,29,110,0,0,0, -1,3,2,1,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0, -57,18,105,110,118,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,31,2,22,1,1,0,31,109,0,0,1,1,0,9,98,0,0,0, -1,3,2,1,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0, -57,18,109,0,16,10,51,0,57,18,105,110,118,0,48,20,0,0,1,0,31,2,22,1,1,0,9,97,0,0,1,1,0,31,110,0,0,0, -1,3,2,1,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, -0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0, -57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0, -57,18,105,110,118,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0,26,2,27,1,1,0,26,109,0,0,0,1,8,58,109,97, -116,50,120,51,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,0,0,0,1,0,28,2,27,1,1,0,28, -109,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,0,0, -0,1,0,27,2,27,1,1,0,27,109,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,54,0,18,109,0, -16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,0,0,0,1,0,30,2,27,1,1,0,30,109,0,0,0,1,8,58,109,97, -116,51,120,52,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0, -0,0,0,1,0,29,2,27,1,1,0,29,109,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,54,0,18, -109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1,0,31,2,27, -1,1,0,31,109,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57, -54,0,18,109,0,16,10,50,0,57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1,0,0,2,25,1,0,2,26,109,0,0,0,1, -9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,0,1,0,0,2,25,1,0,2,28,109,0,0,0,1,9,18, -109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,0,1,0,0,2,25,1,0,2,27,109,0,0,0,1,9,18,109,0, -16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,0,1,0,0,2,25,1,0,2, -30,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57, -52,0,0,1,0,0,2,25,1,0,2,29,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0, -9,18,109,0,16,10,50,0,57,52,0,9,18,109,0,16,10,51,0,57,52,0,0,1,0,0,2,25,1,0,2,31,109,0,0,0,1,9,18, -109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,9,18,109,0,16, -10,51,0,57,52,0,0,1,0,0,2,24,1,0,2,26,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49, -0,57,51,0,0,1,0,0,2,24,1,0,2,28,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57, -51,0,0,1,0,0,2,24,1,0,2,27,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0, -9,18,109,0,16,10,50,0,57,51,0,0,1,0,0,2,24,1,0,2,30,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18, -109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,0,1,0,0,2,24,1,0,2,29,109,0,0,0,1,9,18,109, -0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10, -51,0,57,51,0,0,1,0,0,2,24,1,0,2,31,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0, +57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,0,1,0,0,27,2,22,1,1,0,9,97,0,0,1,1,0,27,110,0,0, +0,1,3,2,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0, +16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,0,27,2,22,1,1,0,27,109,0,0,1,1,0, +9,98,0,0,0,1,3,2,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,0,0,30,2,22,1,1,0,9,97,0,0, +1,1,0,30,110,0,0,0,1,3,2,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49,0,0,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,0,1,0,0,30,2,22,1,1,0,30, +109,0,0,1,1,0,9,98,0,0,0,1,3,2,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,0,1,0,0,29,2,22, +1,1,0,29,109,0,0,1,1,0,9,98,0,0,0,1,3,2,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0,49,0,0,9,18, +95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,105,110,118,0,48,20,0,0,1,0,0, +29,2,22,1,1,0,9,97,0,0,1,1,0,29,110,0,0,0,1,3,2,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,97,0,49, +0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48,0,57,48,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0,57,48,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0,57,48,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,105,110,118,0,18,110,0,16,10,51,0,57,48,20,0,0,1, +0,0,31,2,22,1,1,0,31,109,0,0,1,1,0,9,98,0,0,0,1,3,2,1,0,9,1,105,110,118,0,2,17,49,0,48,0,0,18,98,0, +49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,105,110,118,0,48, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,105,110,118,0,48, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,105,110,118,0,48, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,105,110,118,0,48, +20,0,0,1,0,0,31,2,22,1,1,0,9,97,0,0,1,1,0,31,110,0,0,0,1,3,2,1,0,9,1,105,110,118,0,2,17,49,0,48,0, +0,18,97,0,49,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,105,110,118,0,18,110,0,16,8,48, +0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,105,110,118,0,18,110,0,16,10,49,0, +57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,105,110,118,0,18,110,0,16,10,50,0, +57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,105,110,118,0,18,110,0,16,10,51,0, +57,48,20,0,0,1,0,0,26,2,27,1,1,0,26,109,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57, +54,0,18,109,0,16,10,49,0,57,54,0,0,0,0,1,0,0,28,2,27,1,1,0,28,109,0,0,0,1,8,58,109,97,116,50,120, +52,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,0,0,0,1,0,0,27,2,27,1,1,0,27,109,0,0,0, +1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16, +10,50,0,57,54,0,0,0,0,1,0,0,30,2,27,1,1,0,30,109,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16,8, +48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0,57,54,0,0,0,0,1,0,0,29,2,27,1,1,0,29, +109,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18, +109,0,16,10,50,0,57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1,0,0,31,2,27,1,1,0,31,109,0,0,0,1,8,58, +109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,54,0,18,109,0,16,10,49,0,57,54,0,18,109,0,16,10,50,0, +57,54,0,18,109,0,16,10,51,0,57,54,0,0,0,0,1,0,0,0,2,25,1,0,2,26,109,0,0,0,1,9,18,109,0,16,8,48,0, +57,52,0,9,18,109,0,16,10,49,0,57,52,0,0,1,0,0,0,2,25,1,0,2,28,109,0,0,0,1,9,18,109,0,16,8,48,0,57, +52,0,9,18,109,0,16,10,49,0,57,52,0,0,1,0,0,0,2,25,1,0,2,27,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52, +0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,0,1,0,0,0,2,25,1,0,2,30,109,0,0,0,1, +9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,0,1,0,0,0, +2,25,1,0,2,29,109,0,0,0,1,9,18,109,0,16,8,48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16, +10,50,0,57,52,0,9,18,109,0,16,10,51,0,57,52,0,0,1,0,0,0,2,25,1,0,2,31,109,0,0,0,1,9,18,109,0,16,8, +48,0,57,52,0,9,18,109,0,16,10,49,0,57,52,0,9,18,109,0,16,10,50,0,57,52,0,9,18,109,0,16,10,51,0,57, +52,0,0,1,0,0,0,2,24,1,0,2,26,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51, +0,0,1,0,0,0,2,24,1,0,2,28,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,0, +1,0,0,0,2,24,1,0,2,27,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,18, +109,0,16,10,50,0,57,51,0,0,1,0,0,0,2,24,1,0,2,30,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109, +0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,0,1,0,0,0,2,24,1,0,2,29,109,0,0,0,1,9,18,109,0, +16,8,48,0,57,51,0,9,18,109,0,16,10,49,0,57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10,51, +0,57,51,0,0,1,0,0,0,2,24,1,0,2,31,109,0,0,0,1,9,18,109,0,16,8,48,0,57,51,0,9,18,109,0,16,10,49,0, 57,51,0,9,18,109,0,16,10,50,0,57,51,0,9,18,109,0,16,10,51,0,57,51,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h b/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h index fc1a944217..94ff0e74cf 100644 --- a/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h +++ b/src/mesa/shader/slang/library/slang_builtin_120_common_gc.h @@ -2,103 +2,103 @@ /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ /* slang_builtin_120_common.gc */ -3,1,0,26,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,26,109,0,0,1,0,0,26,110,0, -0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,28,0,109,97,116,114,105,120,67,111,109,112,77,117, -108,116,0,1,0,0,28,109,0,0,1,0,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,27,0,109, -97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,27,109,0,0,1,0,0,27,110,0,0,0,1,8,58,109, -97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110, -0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,30,0,109,97,116, -114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,30,109,0,0,1,0,0,30,110,0,0,0,1,8,58,109,97,116, -51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10, -49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,29,0,109,97,116,114,105, -120,67,111,109,112,77,117,108,116,0,1,0,0,29,109,0,0,1,0,0,29,110,0,0,0,1,8,58,109,97,116,52,120, -50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0, -57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10, -51,0,57,48,0,0,0,0,1,0,31,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,31,109,0, -0,1,0,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48, -0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57, -48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,0,13,0,111,117,116,101,114,80,114, -111,100,117,99,116,0,1,0,0,10,99,0,0,1,0,0,10,114,0,0,0,1,8,58,109,97,116,50,0,18,99,0,59,120,0,18, -114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48, -0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,0,0,0,1,0,14,0,111,117,116,101,114,80,114,111,100,117,99, -116,0,1,0,0,11,99,0,0,1,0,0,11,114,0,0,0,1,8,58,109,97,116,51,0,18,99,0,59,120,0,18,114,0,59,120,0, +4,1,0,0,26,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,26,109,0,0,1,0,0,26,110, +0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16, +10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,0,28,0,109,97,116,114,105,120,67,111,109,112,77, +117,108,116,0,1,0,0,28,109,0,0,1,0,0,28,110,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0, +57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,0,27,0, +109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,27,109,0,0,1,0,0,27,110,0,0,0,1,8,58, +109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,0,30,0,109, +97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,30,109,0,0,1,0,0,30,110,0,0,0,1,8,58,109, +97,116,51,120,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110, +0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,0,29,0,109,97, +116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,29,109,0,0,1,0,0,29,110,0,0,0,1,8,58,109,97, +116,52,120,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0, +16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18, +110,0,16,10,51,0,57,48,0,0,0,0,1,0,0,31,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1, +0,0,31,109,0,0,1,0,0,31,110,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,18,110,0,16, +8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0, +16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,0,0,13,0,111,117,116, +101,114,80,114,111,100,117,99,116,0,1,0,0,10,99,0,0,1,0,0,10,114,0,0,0,1,8,58,109,97,116,50,0,18, +99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18, +114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,0,0,0,1,0,0,14,0,111,117,116,101,114, +80,114,111,100,117,99,116,0,1,0,0,11,99,0,0,1,0,0,11,114,0,0,0,1,8,58,109,97,116,51,0,18,99,0,59, +120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59, +120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0, +59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0, +59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,0,0,0,1,0,0,15,0,111,117,116,101,114,80,114, +111,100,117,99,116,0,1,0,0,12,99,0,0,1,0,0,12,114,0,0,0,1,8,58,109,97,116,52,0,18,99,0,59,120,0,18, +114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48, +0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0, +18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0, +48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59, +122,0,18,114,0,59,122,0,48,0,18,99,0,59,119,0,18,114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59, +119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48,0,18,99,0,59,122,0,18,114,0,59,119,0,48,0,18,99,0, +59,119,0,18,114,0,59,119,0,48,0,0,0,0,1,0,0,26,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1, +0,0,11,99,0,0,1,0,0,10,114,0,0,0,1,8,58,109,97,116,50,120,51,0,18,99,0,59,120,0,18,114,0,59,120,0, 48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59, 120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59, -121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0, -59,122,0,18,114,0,59,122,0,48,0,0,0,0,1,0,15,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0, -0,12,99,0,0,1,0,0,12,114,0,0,0,1,8,58,109,97,116,52,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18, -99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18, -114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48, -0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0, -18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0, -48,0,18,99,0,59,119,0,18,114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59, -121,0,18,114,0,59,119,0,48,0,18,99,0,59,122,0,18,114,0,59,119,0,48,0,18,99,0,59,119,0,18,114,0,59, -119,0,48,0,0,0,0,1,0,26,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,11,99,0,0,1,0,0,10, -114,0,0,0,1,8,58,109,97,116,50,120,51,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0, -18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0, -48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,0,0,0,1,0,27, -0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,10,99,0,0,1,0,0,11,114,0,0,0,1,8,58,109,97, -116,51,120,50,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18, -99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18, -114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,0,0,0,1,0,28,0,111,117,116,101,114,80, -114,111,100,117,99,116,0,1,0,0,12,99,0,0,1,0,0,10,114,0,0,0,1,8,58,109,97,116,50,120,52,0,18,99,0, -59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0, -59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18, -99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,119,0,18, -114,0,59,121,0,48,0,0,0,0,1,0,29,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,10,99,0,0, -1,0,0,12,114,0,0,0,1,8,58,109,97,116,52,120,50,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0, -59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0, -59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18, -99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48,0,0,0,0,1,0,30,0,111, -117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,12,99,0,0,1,0,0,11,114,0,0,0,1,8,58,109,97,116, -51,120,52,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99, -0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114, -0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18, -99,0,59,119,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18, -114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,18,99,0,59,119,0,18,114,0,59,122,0,48, -0,0,0,0,1,0,31,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,11,99,0,0,1,0,0,12,114,0,0, -0,1,8,58,109,97,116,52,120,51,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0, -59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18, -99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18, -114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48, -0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0,48,0,18,99,0,59,122,0, -18,114,0,59,119,0,48,0,0,0,0,1,0,13,0,116,114,97,110,115,112,111,115,101,0,1,0,0,13,109,0,0,0,1,8, -58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16, -8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,0,0,0,1,0,14,0,116,114,97,110,115,112,111, -115,101,0,1,0,0,14,109,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16, -10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0, -16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109, -0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,0,0,0,1,0,15,0,116,114,97,110,115,112, -111,115,101,0,1,0,0,15,109,0,0,0,1,8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0, -16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,0,0,18, -109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0, -18,109,0,16,10,51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0, -0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59,122,0,0,18,109,0,16,8,48,0,57,59,119, -0,0,18,109,0,16,10,49,0,57,59,119,0,0,18,109,0,16,10,50,0,57,59,119,0,0,18,109,0,16,10,51,0,57,59, -119,0,0,0,0,0,1,0,26,0,116,114,97,110,115,112,111,115,101,0,1,0,0,27,109,0,0,0,1,8,58,109,97,116, -50,120,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0, -57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50, -0,57,59,121,0,0,0,0,0,1,0,27,0,116,114,97,110,115,112,111,115,101,0,1,0,0,26,109,0,0,0,1,8,58,109, -97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16, -8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0, -16,10,49,0,57,59,122,0,0,0,0,0,1,0,28,0,116,114,97,110,115,112,111,115,101,0,1,0,0,29,109,0,0,0,1, -8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18, -109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0, -18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121, -0,0,0,0,0,1,0,29,0,116,114,97,110,115,112,111,115,101,0,1,0,0,28,109,0,0,0,1,8,58,109,97,116,52, -120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,8,48,0,57, -59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0, -57,59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,0,0,0,1,0,30,0, -116,114,97,110,115,112,111,115,101,0,1,0,0,31,109,0,0,0,1,8,58,109,97,116,51,120,52,0,18,109,0,16, +121,0,48,0,0,0,0,1,0,0,27,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,10,99,0,0,1,0,0, +11,114,0,0,0,1,8,58,109,97,116,51,120,50,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121, +0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121, +0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,0,0,0,1,0,0, +28,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,12,99,0,0,1,0,0,10,114,0,0,0,1,8,58,109, +97,116,50,120,52,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0, +18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0, +18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0, +48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,0,0,0,1,0,0,29,0,111,117,116,101,114,80,114,111,100, +117,99,116,0,1,0,0,10,99,0,0,1,0,0,12,114,0,0,0,1,8,58,109,97,116,52,120,50,0,18,99,0,59,120,0,18, +114,0,59,120,0,48,0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0,18,114,0,59,121,0,48, +0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0, +18,114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59,119,0, +48,0,0,0,0,1,0,0,30,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0,0,12,99,0,0,1,0,0,11,114, +0,0,0,1,8,58,109,97,116,51,120,52,0,18,99,0,59,120,0,18,114,0,59,120,0,48,0,18,99,0,59,121,0,18, +114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,119,0,18,114,0,59,120,0,48, +0,18,99,0,59,120,0,18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0, +18,114,0,59,121,0,48,0,18,99,0,59,119,0,18,114,0,59,121,0,48,0,18,99,0,59,120,0,18,114,0,59,122,0, +48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59,122,0,18,114,0,59,122,0,48,0,18,99,0,59, +119,0,18,114,0,59,122,0,48,0,0,0,0,1,0,0,31,0,111,117,116,101,114,80,114,111,100,117,99,116,0,1,0, +0,11,99,0,0,1,0,0,12,114,0,0,0,1,8,58,109,97,116,52,120,51,0,18,99,0,59,120,0,18,114,0,59,120,0,48, +0,18,99,0,59,121,0,18,114,0,59,120,0,48,0,18,99,0,59,122,0,18,114,0,59,120,0,48,0,18,99,0,59,120,0, +18,114,0,59,121,0,48,0,18,99,0,59,121,0,18,114,0,59,121,0,48,0,18,99,0,59,122,0,18,114,0,59,121,0, +48,0,18,99,0,59,120,0,18,114,0,59,122,0,48,0,18,99,0,59,121,0,18,114,0,59,122,0,48,0,18,99,0,59, +122,0,18,114,0,59,122,0,48,0,18,99,0,59,120,0,18,114,0,59,119,0,48,0,18,99,0,59,121,0,18,114,0,59, +119,0,48,0,18,99,0,59,122,0,18,114,0,59,119,0,48,0,0,0,0,1,0,0,13,0,116,114,97,110,115,112,111,115, +101,0,1,0,0,13,109,0,0,0,1,8,58,109,97,116,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49, +0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,0,0,0,1,0,0,14, +0,116,114,97,110,115,112,111,115,101,0,1,0,0,14,109,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48, +0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8, +48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16, +8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,0,0,0,1,0, +0,15,0,116,114,97,110,115,112,111,115,101,0,1,0,0,15,109,0,0,0,1,8,58,109,97,116,52,0,18,109,0,16, 8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0, 16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109, 0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18, 109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,10,51,0,57,59,122,0,0, -0,0,0,1,0,31,0,116,114,97,110,115,112,111,115,101,0,1,0,0,30,109,0,0,0,1,8,58,109,97,116,52,120,51, -0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120, -0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59, -121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57, -59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,18,109,0,16,10,50,0, -57,59,119,0,0,0,0,0,0 +18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0,57,59,119,0,0,18,109,0,16,10,50,0,57,59,119,0, +0,18,109,0,16,10,51,0,57,59,119,0,0,0,0,0,1,0,0,26,0,116,114,97,110,115,112,111,115,101,0,1,0,0,27, +109,0,0,0,1,8,58,109,97,116,50,120,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59, +120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57, +59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,0,0,0,1,0,0,27,0,116,114,97,110,115,112,111,115,101,0, +1,0,0,26,109,0,0,0,1,8,58,109,97,116,51,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10, +49,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16, +8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,0,0,0,1,0,0,28,0,116,114,97,110,115,112,111, +115,101,0,1,0,0,29,109,0,0,0,1,8,58,109,97,116,50,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109, +0,16,10,49,0,57,59,120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,0,0,18, +109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0, +18,109,0,16,10,51,0,57,59,121,0,0,0,0,0,1,0,0,29,0,116,114,97,110,115,112,111,115,101,0,1,0,0,28, +109,0,0,0,1,8,58,109,97,116,52,120,50,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59, +120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57,59,121,0,0,18,109,0,16,8,48,0,57, +59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49,0, +57,59,119,0,0,0,0,0,1,0,0,30,0,116,114,97,110,115,112,111,115,101,0,1,0,0,31,109,0,0,0,1,8,58,109, +97,116,51,120,52,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59,120,0,0,18,109,0,16, +10,50,0,57,59,120,0,0,18,109,0,16,10,51,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0, +16,10,49,0,57,59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,10,51,0,57,59,121,0,0,18, +109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0,57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0, +18,109,0,16,10,51,0,57,59,122,0,0,0,0,0,1,0,0,31,0,116,114,97,110,115,112,111,115,101,0,1,0,0,30, +109,0,0,0,1,8,58,109,97,116,52,120,51,0,18,109,0,16,8,48,0,57,59,120,0,0,18,109,0,16,10,49,0,57,59, +120,0,0,18,109,0,16,10,50,0,57,59,120,0,0,18,109,0,16,8,48,0,57,59,121,0,0,18,109,0,16,10,49,0,57, +59,121,0,0,18,109,0,16,10,50,0,57,59,121,0,0,18,109,0,16,8,48,0,57,59,122,0,0,18,109,0,16,10,49,0, +57,59,122,0,0,18,109,0,16,10,50,0,57,59,122,0,0,18,109,0,16,8,48,0,57,59,119,0,0,18,109,0,16,10,49, +0,57,59,119,0,0,18,109,0,16,10,50,0,57,59,119,0,0,0,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h b/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h index 2400b273d8..72d86fb25d 100644 --- a/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h +++ b/src/mesa/shader/slang/library/slang_builtin_120_fragment_gc.h @@ -2,4 +2,4 @@ /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ /* slang_builtin_120_fragment.gc */ -3,2,2,3,10,1,103,108,95,80,111,105,110,116,67,111,111,114,100,0,0,0,0 +4,2,2,3,0,10,1,103,108,95,80,111,105,110,116,67,111,111,114,100,0,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index 9301a57339..c921f08a6b 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -2,814 +2,822 @@ /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ /* slang_common_builtin.gc */ -3,2,2,1,5,1,103,108,95,77,97,120,76,105,103,104,116,115,0,2,16,10,56,0,0,0,2,2,1,5,1,103,108,95,77, -97,120,67,108,105,112,80,108,97,110,101,115,0,2,16,10,54,0,0,0,2,2,1,5,1,103,108,95,77,97,120,84, -101,120,116,117,114,101,85,110,105,116,115,0,2,16,10,56,0,0,0,2,2,1,5,1,103,108,95,77,97,120,84, -101,120,116,117,114,101,67,111,111,114,100,115,0,2,16,10,56,0,0,0,2,2,1,5,1,103,108,95,77,97,120, -86,101,114,116,101,120,65,116,116,114,105,98,115,0,2,16,10,49,54,0,0,0,2,2,1,5,1,103,108,95,77,97, -120,86,101,114,116,101,120,85,110,105,102,111,114,109,67,111,109,112,111,110,101,110,116,115,0,2, -16,10,53,49,50,0,0,0,2,2,1,5,1,103,108,95,77,97,120,86,97,114,121,105,110,103,70,108,111,97,116, -115,0,2,16,10,51,50,0,0,0,2,2,1,5,1,103,108,95,77,97,120,86,101,114,116,101,120,84,101,120,116,117, -114,101,73,109,97,103,101,85,110,105,116,115,0,2,16,8,48,0,0,0,2,2,1,5,1,103,108,95,77,97,120,67, -111,109,98,105,110,101,100,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105,116,115,0,2,16, -10,50,0,0,0,2,2,1,5,1,103,108,95,77,97,120,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105, -116,115,0,2,16,10,50,0,0,0,2,2,1,5,1,103,108,95,77,97,120,70,114,97,103,109,101,110,116,85,110,105, -102,111,114,109,67,111,109,112,111,110,101,110,116,115,0,2,16,10,54,52,0,0,0,2,2,1,5,1,103,108,95, -77,97,120,68,114,97,119,66,117,102,102,101,114,115,0,2,16,10,49,0,0,0,2,2,4,15,1,103,108,95,77,111, -100,101,108,86,105,101,119,77,97,116,114,105,120,0,0,0,2,2,4,15,1,103,108,95,80,114,111,106,101,99, -116,105,111,110,77,97,116,114,105,120,0,0,0,2,2,4,15,1,103,108,95,77,111,100,101,108,86,105,101, -119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,0,0,2,2,4,15,1,103,108,95,84,101, -120,116,117,114,101,77,97,116,114,105,120,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101, -67,111,111,114,100,115,0,0,0,2,2,4,14,1,103,108,95,78,111,114,109,97,108,77,97,116,114,105,120,0,0, -0,2,2,4,14,1,95,95,78,111,114,109,97,108,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0, -0,0,2,2,4,15,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,73,110,118,101, -114,115,101,0,0,0,2,2,4,15,1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105, -120,73,110,118,101,114,115,101,0,0,0,2,2,4,15,1,103,108,95,77,111,100,101,108,86,105,101,119,80, -114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2,2,4,15, -1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,73,110,118,101,114,115,101,0,3,18, -103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,15,1,103,108,95, -77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0,0,0,2, -2,4,15,1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,84,114,97,110,115, -112,111,115,101,0,0,0,2,2,4,15,1,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101, -99,116,105,111,110,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0,0,0,2,2,4,15,1,103, -108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0,3,18, -103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,15,1,103,108,95, -77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,73,110,118,101,114,115,101,84,114,97,110, -115,112,111,115,101,0,0,0,2,2,4,15,1,103,108,95,80,114,111,106,101,99,116,105,111,110,77,97,116, -114,105,120,73,110,118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,0,0,2,2,4,15,1,103,108, -95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120, -73,110,118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,0,0,2,2,4,15,1,103,108,95,84,101, -120,116,117,114,101,77,97,116,114,105,120,73,110,118,101,114,115,101,84,114,97,110,115,112,111,115, -101,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,9,1, -103,108,95,78,111,114,109,97,108,83,99,97,108,101,0,0,0,2,2,0,24,103,108,95,68,101,112,116,104,82, -97,110,103,101,80,97,114,97,109,101,116,101,114,115,0,9,110,101,97,114,0,0,0,1,9,102,97,114,0,0,0, -1,9,100,105,102,102,0,0,0,0,0,0,2,2,4,25,103,108,95,68,101,112,116,104,82,97,110,103,101,80,97,114, -97,109,101,116,101,114,115,0,1,103,108,95,68,101,112,116,104,82,97,110,103,101,0,0,0,2,2,4,12,1, -103,108,95,67,108,105,112,80,108,97,110,101,0,3,18,103,108,95,77,97,120,67,108,105,112,80,108,97, -110,101,115,0,0,0,2,2,0,24,103,108,95,80,111,105,110,116,80,97,114,97,109,101,116,101,114,115,0,9, -115,105,122,101,0,0,0,1,9,115,105,122,101,77,105,110,0,0,0,1,9,115,105,122,101,77,97,120,0,0,0,1,9, -102,97,100,101,84,104,114,101,115,104,111,108,100,83,105,122,101,0,0,0,1,9,100,105,115,116,97,110, -99,101,67,111,110,115,116,97,110,116,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,100,105, -115,116,97,110,99,101,76,105,110,101,97,114,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9, -100,105,115,116,97,110,99,101,81,117,97,100,114,97,116,105,99,65,116,116,101,110,117,97,116,105, -111,110,0,0,0,0,0,0,2,2,4,25,103,108,95,80,111,105,110,116,80,97,114,97,109,101,116,101,114,115,0, -1,103,108,95,80,111,105,110,116,0,0,0,2,2,0,24,103,108,95,77,97,116,101,114,105,97,108,80,97,114, -97,109,101,116,101,114,115,0,12,101,109,105,115,115,105,111,110,0,0,0,1,12,97,109,98,105,101,110, -116,0,0,0,1,12,100,105,102,102,117,115,101,0,0,0,1,12,115,112,101,99,117,108,97,114,0,0,0,1,9,115, -104,105,110,105,110,101,115,115,0,0,0,0,0,0,2,2,4,25,103,108,95,77,97,116,101,114,105,97,108,80,97, -114,97,109,101,116,101,114,115,0,1,103,108,95,70,114,111,110,116,77,97,116,101,114,105,97,108,0,0, -0,2,2,4,25,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,101,116,101,114,115,0,1,103, -108,95,66,97,99,107,77,97,116,101,114,105,97,108,0,0,0,2,2,0,24,103,108,95,76,105,103,104,116,83, -111,117,114,99,101,80,97,114,97,109,101,116,101,114,115,0,12,97,109,98,105,101,110,116,0,0,0,1,12, -100,105,102,102,117,115,101,0,0,0,1,12,115,112,101,99,117,108,97,114,0,0,0,1,12,112,111,115,105, -116,105,111,110,0,0,0,1,12,104,97,108,102,86,101,99,116,111,114,0,0,0,1,11,115,112,111,116,68,105, -114,101,99,116,105,111,110,0,0,0,1,9,115,112,111,116,69,120,112,111,110,101,110,116,0,0,0,1,9,115, -112,111,116,67,117,116,111,102,102,0,0,0,1,9,115,112,111,116,67,111,115,67,117,116,111,102,102,0,0, -0,1,9,99,111,110,115,116,97,110,116,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,108,105, -110,101,97,114,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,113,117,97,100,114,97,116,105, -99,65,116,116,101,110,117,97,116,105,111,110,0,0,0,0,0,0,2,2,4,25,103,108,95,76,105,103,104,116,83, -111,117,114,99,101,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95,76,105,103,104,116,83,111, -117,114,99,101,0,3,18,103,108,95,77,97,120,76,105,103,104,116,115,0,0,0,2,2,0,24,103,108,95,76,105, -103,104,116,77,111,100,101,108,80,97,114,97,109,101,116,101,114,115,0,12,97,109,98,105,101,110,116, -0,0,0,0,0,0,2,2,4,25,103,108,95,76,105,103,104,116,77,111,100,101,108,80,97,114,97,109,101,116,101, -114,115,0,1,103,108,95,76,105,103,104,116,77,111,100,101,108,0,0,0,2,2,0,24,103,108,95,76,105,103, -104,116,77,111,100,101,108,80,114,111,100,117,99,116,115,0,12,115,99,101,110,101,67,111,108,111, -114,0,0,0,0,0,0,2,2,4,25,103,108,95,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99, -116,115,0,1,103,108,95,70,114,111,110,116,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117, -99,116,0,0,0,2,2,4,25,103,108,95,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116, -115,0,1,103,108,95,66,97,99,107,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,0, -0,0,2,2,0,24,103,108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,12,97,109,98,105,101, -110,116,0,0,0,1,12,100,105,102,102,117,115,101,0,0,0,1,12,115,112,101,99,117,108,97,114,0,0,0,0,0, -0,2,2,4,25,103,108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,1,103,108,95,70,114,111, -110,116,76,105,103,104,116,80,114,111,100,117,99,116,0,3,18,103,108,95,77,97,120,76,105,103,104, -116,115,0,0,0,2,2,4,25,103,108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,1,103,108,95, -66,97,99,107,76,105,103,104,116,80,114,111,100,117,99,116,0,3,18,103,108,95,77,97,120,76,105,103, -104,116,115,0,0,0,2,2,4,12,1,103,108,95,84,101,120,116,117,114,101,69,110,118,67,111,108,111,114,0, -3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105,116,115,0,0,0,2, -2,4,12,1,103,108,95,69,121,101,80,108,97,110,101,83,0,3,18,103,108,95,77,97,120,84,101,120,116,117, -114,101,67,111,111,114,100,115,0,0,0,2,2,4,12,1,103,108,95,69,121,101,80,108,97,110,101,84,0,3,18, -103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,12,1,103,108,95, -69,121,101,80,108,97,110,101,82,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111, -114,100,115,0,0,0,2,2,4,12,1,103,108,95,69,121,101,80,108,97,110,101,81,0,3,18,103,108,95,77,97, -120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,12,1,103,108,95,79,98,106,101,99, +4,2,2,1,0,5,1,103,108,95,77,97,120,76,105,103,104,116,115,0,2,16,10,56,0,0,0,2,2,1,0,5,1,103,108, +95,77,97,120,67,108,105,112,80,108,97,110,101,115,0,2,16,10,54,0,0,0,2,2,1,0,5,1,103,108,95,77,97, +120,84,101,120,116,117,114,101,85,110,105,116,115,0,2,16,10,56,0,0,0,2,2,1,0,5,1,103,108,95,77,97, +120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,2,16,10,56,0,0,0,2,2,1,0,5,1,103,108,95,77, +97,120,86,101,114,116,101,120,65,116,116,114,105,98,115,0,2,16,10,49,54,0,0,0,2,2,1,0,5,1,103,108, +95,77,97,120,86,101,114,116,101,120,85,110,105,102,111,114,109,67,111,109,112,111,110,101,110,116, +115,0,2,16,10,53,49,50,0,0,0,2,2,1,0,5,1,103,108,95,77,97,120,86,97,114,121,105,110,103,70,108,111, +97,116,115,0,2,16,10,51,50,0,0,0,2,2,1,0,5,1,103,108,95,77,97,120,86,101,114,116,101,120,84,101, +120,116,117,114,101,73,109,97,103,101,85,110,105,116,115,0,2,16,8,48,0,0,0,2,2,1,0,5,1,103,108,95, +77,97,120,67,111,109,98,105,110,101,100,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105, +116,115,0,2,16,10,50,0,0,0,2,2,1,0,5,1,103,108,95,77,97,120,84,101,120,116,117,114,101,73,109,97, +103,101,85,110,105,116,115,0,2,16,10,50,0,0,0,2,2,1,0,5,1,103,108,95,77,97,120,70,114,97,103,109, +101,110,116,85,110,105,102,111,114,109,67,111,109,112,111,110,101,110,116,115,0,2,16,10,54,52,0,0, +0,2,2,1,0,5,1,103,108,95,77,97,120,68,114,97,119,66,117,102,102,101,114,115,0,2,16,10,49,0,0,0,2,2, +4,0,15,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120,0,0,0,2,2,4,0,15,1,103, +108,95,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,0,0,2,2,4,0,15,1,103,108,95, +77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,0,0, +0,2,2,4,0,15,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,0,3,18,103,108,95,77,97, +120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,0,14,1,103,108,95,78,111,114,109, +97,108,77,97,116,114,105,120,0,0,0,2,2,4,0,14,1,95,95,78,111,114,109,97,108,77,97,116,114,105,120, +84,114,97,110,115,112,111,115,101,0,0,0,2,2,4,0,15,1,103,108,95,77,111,100,101,108,86,105,101,119, +77,97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2,2,4,0,15,1,103,108,95,80,114,111,106,101, +99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,101,0,0,0,2,2,4,0,15,1,103,108,95, +77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73, +110,118,101,114,115,101,0,0,0,2,2,4,0,15,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105, +120,73,110,118,101,114,115,101,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111, +114,100,115,0,0,0,2,2,4,0,15,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,114,105,120, +84,114,97,110,115,112,111,115,101,0,0,0,2,2,4,0,15,1,103,108,95,80,114,111,106,101,99,116,105,111, +110,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0,0,0,2,2,4,0,15,1,103,108,95,77,111, +100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,84,114,97, +110,115,112,111,115,101,0,0,0,2,2,4,0,15,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105, +120,84,114,97,110,115,112,111,115,101,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67, +111,111,114,100,115,0,0,0,2,2,4,0,15,1,103,108,95,77,111,100,101,108,86,105,101,119,77,97,116,114, +105,120,73,110,118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,0,0,2,2,4,0,15,1,103,108,95, +80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,101,84,114,97, +110,115,112,111,115,101,0,0,0,2,2,4,0,15,1,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111, +106,101,99,116,105,111,110,77,97,116,114,105,120,73,110,118,101,114,115,101,84,114,97,110,115,112, +111,115,101,0,0,0,2,2,4,0,15,1,103,108,95,84,101,120,116,117,114,101,77,97,116,114,105,120,73,110, +118,101,114,115,101,84,114,97,110,115,112,111,115,101,0,3,18,103,108,95,77,97,120,84,101,120,116, +117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,0,9,1,103,108,95,78,111,114,109,97,108,83,99,97,108, +101,0,0,0,2,2,0,0,24,103,108,95,68,101,112,116,104,82,97,110,103,101,80,97,114,97,109,101,116,101, +114,115,0,9,110,101,97,114,0,0,0,1,9,102,97,114,0,0,0,1,9,100,105,102,102,0,0,0,0,0,0,2,2,4,0,25, +103,108,95,68,101,112,116,104,82,97,110,103,101,80,97,114,97,109,101,116,101,114,115,0,1,103,108, +95,68,101,112,116,104,82,97,110,103,101,0,0,0,2,2,4,0,12,1,103,108,95,67,108,105,112,80,108,97,110, +101,0,3,18,103,108,95,77,97,120,67,108,105,112,80,108,97,110,101,115,0,0,0,2,2,0,0,24,103,108,95, +80,111,105,110,116,80,97,114,97,109,101,116,101,114,115,0,9,115,105,122,101,0,0,0,1,9,115,105,122, +101,77,105,110,0,0,0,1,9,115,105,122,101,77,97,120,0,0,0,1,9,102,97,100,101,84,104,114,101,115,104, +111,108,100,83,105,122,101,0,0,0,1,9,100,105,115,116,97,110,99,101,67,111,110,115,116,97,110,116, +65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,100,105,115,116,97,110,99,101,76,105,110,101, +97,114,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,100,105,115,116,97,110,99,101,81,117,97, +100,114,97,116,105,99,65,116,116,101,110,117,97,116,105,111,110,0,0,0,0,0,0,2,2,4,0,25,103,108,95, +80,111,105,110,116,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95,80,111,105,110,116,0,0,0,2, +2,0,0,24,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,101,116,101,114,115,0,12,101,109, +105,115,115,105,111,110,0,0,0,1,12,97,109,98,105,101,110,116,0,0,0,1,12,100,105,102,102,117,115, +101,0,0,0,1,12,115,112,101,99,117,108,97,114,0,0,0,1,9,115,104,105,110,105,110,101,115,115,0,0,0,0, +0,0,2,2,4,0,25,103,108,95,77,97,116,101,114,105,97,108,80,97,114,97,109,101,116,101,114,115,0,1, +103,108,95,70,114,111,110,116,77,97,116,101,114,105,97,108,0,0,0,2,2,4,0,25,103,108,95,77,97,116, +101,114,105,97,108,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95,66,97,99,107,77,97,116,101, +114,105,97,108,0,0,0,2,2,0,0,24,103,108,95,76,105,103,104,116,83,111,117,114,99,101,80,97,114,97, +109,101,116,101,114,115,0,12,97,109,98,105,101,110,116,0,0,0,1,12,100,105,102,102,117,115,101,0,0, +0,1,12,115,112,101,99,117,108,97,114,0,0,0,1,12,112,111,115,105,116,105,111,110,0,0,0,1,12,104,97, +108,102,86,101,99,116,111,114,0,0,0,1,11,115,112,111,116,68,105,114,101,99,116,105,111,110,0,0,0,1, +9,115,112,111,116,69,120,112,111,110,101,110,116,0,0,0,1,9,115,112,111,116,67,117,116,111,102,102, +0,0,0,1,9,115,112,111,116,67,111,115,67,117,116,111,102,102,0,0,0,1,9,99,111,110,115,116,97,110, +116,65,116,116,101,110,117,97,116,105,111,110,0,0,0,1,9,108,105,110,101,97,114,65,116,116,101,110, +117,97,116,105,111,110,0,0,0,1,9,113,117,97,100,114,97,116,105,99,65,116,116,101,110,117,97,116, +105,111,110,0,0,0,0,0,0,2,2,4,0,25,103,108,95,76,105,103,104,116,83,111,117,114,99,101,80,97,114, +97,109,101,116,101,114,115,0,1,103,108,95,76,105,103,104,116,83,111,117,114,99,101,0,3,18,103,108, +95,77,97,120,76,105,103,104,116,115,0,0,0,2,2,0,0,24,103,108,95,76,105,103,104,116,77,111,100,101, +108,80,97,114,97,109,101,116,101,114,115,0,12,97,109,98,105,101,110,116,0,0,0,0,0,0,2,2,4,0,25,103, +108,95,76,105,103,104,116,77,111,100,101,108,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95, +76,105,103,104,116,77,111,100,101,108,0,0,0,2,2,0,0,24,103,108,95,76,105,103,104,116,77,111,100, +101,108,80,114,111,100,117,99,116,115,0,12,115,99,101,110,101,67,111,108,111,114,0,0,0,0,0,0,2,2,4, +0,25,103,108,95,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,115,0,1,103,108,95, +70,114,111,110,116,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,0,0,0,2,2,4,0, +25,103,108,95,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,115,0,1,103,108,95, +66,97,99,107,76,105,103,104,116,77,111,100,101,108,80,114,111,100,117,99,116,0,0,0,2,2,0,0,24,103, +108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,12,97,109,98,105,101,110,116,0,0,0,1,12, +100,105,102,102,117,115,101,0,0,0,1,12,115,112,101,99,117,108,97,114,0,0,0,0,0,0,2,2,4,0,25,103, +108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,1,103,108,95,70,114,111,110,116,76,105, +103,104,116,80,114,111,100,117,99,116,0,3,18,103,108,95,77,97,120,76,105,103,104,116,115,0,0,0,2,2, +4,0,25,103,108,95,76,105,103,104,116,80,114,111,100,117,99,116,115,0,1,103,108,95,66,97,99,107,76, +105,103,104,116,80,114,111,100,117,99,116,0,3,18,103,108,95,77,97,120,76,105,103,104,116,115,0,0,0, +2,2,4,0,12,1,103,108,95,84,101,120,116,117,114,101,69,110,118,67,111,108,111,114,0,3,18,103,108,95, +77,97,120,84,101,120,116,117,114,101,73,109,97,103,101,85,110,105,116,115,0,0,0,2,2,4,0,12,1,103, +108,95,69,121,101,80,108,97,110,101,83,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67, +111,111,114,100,115,0,0,0,2,2,4,0,12,1,103,108,95,69,121,101,80,108,97,110,101,84,0,3,18,103,108, +95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,0,12,1,103,108,95,69, +121,101,80,108,97,110,101,82,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114, +100,115,0,0,0,2,2,4,0,12,1,103,108,95,69,121,101,80,108,97,110,101,81,0,3,18,103,108,95,77,97,120, +84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,0,12,1,103,108,95,79,98,106,101,99, 116,80,108,97,110,101,83,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100, -115,0,0,0,2,2,4,12,1,103,108,95,79,98,106,101,99,116,80,108,97,110,101,84,0,3,18,103,108,95,77,97, -120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,12,1,103,108,95,79,98,106,101,99, -116,80,108,97,110,101,82,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100, -115,0,0,0,2,2,4,12,1,103,108,95,79,98,106,101,99,116,80,108,97,110,101,81,0,3,18,103,108,95,77,97, -120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,0,24,103,108,95,70,111,103,80,97, -114,97,109,101,116,101,114,115,0,12,99,111,108,111,114,0,0,0,1,9,100,101,110,115,105,116,121,0,0,0, -1,9,115,116,97,114,116,0,0,0,1,9,101,110,100,0,0,0,1,9,115,99,97,108,101,0,0,0,0,0,0,2,2,4,25,103, -108,95,70,111,103,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95,70,111,103,0,0,0,1,0,9,0,114, -97,100,105,97,110,115,0,1,1,0,9,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54, -0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,0,18,100,101,103,0,0,18,99,0,0,0,0,1,0,10,0,114,97,100,105,97,110,115,0,1,1,0, -10,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0, -0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -0,0,18,100,101,103,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,0,11,0,114,97,100,105,97,110,115, -0,1,1,0,11,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48, -0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,122,0,0,18,100,101,103,0,59,120,121,122,0,0,18,99,0,59,120,120,120,0,0,0,0,1,0,12,0,114, -97,100,105,97,110,115,0,1,1,0,12,100,101,103,0,0,0,1,3,2,1,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54, -0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,0,18,100,101,103,0,0,18,99,0,59,120,120,120,120,0,0,0,0,1,0,9,0,100,101,103, -114,101,101,115,0,1,1,0,9,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52, -49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, -116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,0,0,0,1,0,10,0,100,101,103,114,101,101,115,0,1,1,0,10, -114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4, -118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, -18,114,97,100,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,0,11,0,100,101,103,114,101,101,115,0,1, -1,0,11,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0, -49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,114,97,100,0,59,120,121,122,0,0,18,99,0,59,120,120,120,0,0,0,0,1,0,12,0,100,101, -103,114,101,101,115,0,1,1,0,12,114,97,100,0,0,0,1,3,2,1,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49, -52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, -116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,59,120,120,120,120,0,0,0,0,1,0,9,0,115,105,110,0,1,1,0, -9,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116, -86,97,108,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,0,10,0,115,105,110,0,1,1,0,10,114,97,100,105, -97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,0,1,0,11,0, -115,105,110,0,1,1,0,11,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108, -111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97, -110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0, -59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,0,12,0,115,105,110,0,1,1,0,12,114,97, -100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110, -101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4, -102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100, -105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97, -108,0,59,119,0,0,18,114,97,100,105,97,110,115,0,59,119,0,0,0,0,1,0,9,0,99,111,115,0,1,1,0,9,114,97, +115,0,0,0,2,2,4,0,12,1,103,108,95,79,98,106,101,99,116,80,108,97,110,101,84,0,3,18,103,108,95,77, +97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,4,0,12,1,103,108,95,79,98,106, +101,99,116,80,108,97,110,101,82,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111, +114,100,115,0,0,0,2,2,4,0,12,1,103,108,95,79,98,106,101,99,116,80,108,97,110,101,81,0,3,18,103,108, +95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,0,0,24,103,108,95,70,111, +103,80,97,114,97,109,101,116,101,114,115,0,12,99,111,108,111,114,0,0,0,1,9,100,101,110,115,105,116, +121,0,0,0,1,9,115,116,97,114,116,0,0,0,1,9,101,110,100,0,0,0,1,9,115,99,97,108,101,0,0,0,0,0,0,2,2, +4,0,25,103,108,95,70,111,103,80,97,114,97,109,101,116,101,114,115,0,1,103,108,95,70,111,103,0,0,0, +1,0,0,9,0,114,97,100,105,97,110,115,0,1,1,0,9,100,101,103,0,0,0,1,3,2,1,0,9,1,99,0,2,17,51,0,49,52, +49,53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,95,95,114,101,116,86,97,108,0,0,18,100,101,103,0,0,18,99,0,0,0,0,1,0,0,10,0,114,97,100,105,97, +110,115,0,1,1,0,10,100,101,103,0,0,0,1,3,2,1,0,9,1,99,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17,49, +56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,100,101,103,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1,0,0,11,0,114, +97,100,105,97,110,115,0,1,1,0,11,100,101,103,0,0,0,1,3,2,1,0,9,1,99,0,2,17,51,0,49,52,49,53,57,50, +54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,100,101,103,0,59,120,121,122,0,0,18,99,0,59,120,120, +120,0,0,0,0,1,0,0,12,0,114,97,100,105,97,110,115,0,1,1,0,12,100,101,103,0,0,0,1,3,2,1,0,9,1,99,0,2, +17,51,0,49,52,49,53,57,50,54,0,0,17,49,56,48,0,48,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,100,101,103,0,0,18,99,0,59,120,120,120,120, +0,0,0,0,1,0,0,9,0,100,101,103,114,101,101,115,0,1,1,0,9,114,97,100,0,0,0,1,3,2,1,0,9,1,99,0,2,17, +49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,0,0,0,1,0,0,10,0,100, +101,103,114,101,101,115,0,1,1,0,10,114,97,100,0,0,0,1,3,2,1,0,9,1,99,0,2,17,49,56,48,0,48,0,0,17, +51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,0,18,114,97,100,0,59,120,121,0,0,18,99,0,59,120,120,0,0,0,0,1, +0,0,11,0,100,101,103,114,101,101,115,0,1,1,0,11,114,97,100,0,0,0,1,3,2,1,0,9,1,99,0,2,17,49,56,48, +0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,114,97,100,0,59,120,121,122,0,0,18,99, +0,59,120,120,120,0,0,0,0,1,0,0,12,0,100,101,103,114,101,101,115,0,1,1,0,12,114,97,100,0,0,0,1,3,2, +1,0,9,1,99,0,2,17,49,56,48,0,48,0,0,17,51,0,49,52,49,53,57,50,54,0,0,49,0,0,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,0,0,18,99,0,59,120, +120,120,120,0,0,0,0,1,0,0,9,0,115,105,110,0,1,1,0,9,114,97,100,105,97,110,115,0,0,0,1,4,102,108, +111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,105,97,110,115,0, +0,0,0,1,0,0,10,0,115,105,110,0,1,1,0,10,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95, +115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59, +120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18, +114,97,100,105,97,110,115,0,59,121,0,0,0,0,1,0,0,11,0,115,105,110,0,1,1,0,11,114,97,100,105,97,110, +115,0,0,0,1,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, +18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114, +101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95, +115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59, +122,0,0,0,0,1,0,0,12,0,115,105,110,0,1,1,0,12,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97, +116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115, +0,59,120,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0, +0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,115,105,110,101,0,18,95,95, +114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97, +116,95,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,110,115, +0,59,119,0,0,0,0,1,0,0,9,0,99,111,115,0,1,1,0,9,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97, +116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,114,97,100,105,97,110,115,0, +0,0,0,1,0,0,10,0,99,111,115,0,1,1,0,10,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95, +99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115, +0,59,120,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0, +59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,0,1,0,0,11,0,99,111,115,0,1,1,0,11,114,97, 100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116, -86,97,108,0,0,18,114,97,100,105,97,110,115,0,0,0,0,1,0,10,0,99,111,115,0,1,1,0,10,114,97,100,105, +86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111, +115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59, +121,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122, +0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,0,0,12,0,99,111,115,0,1,1,0,12,114,97,100,105, 97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108, 0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4,102,108,111,97,116,95,99,111,115,105, 110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0, -0,1,0,11,0,99,111,115,0,1,1,0,11,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111, -115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59, -120,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121, -0,0,18,114,97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,0,1,0,12,0, -99,111,115,0,1,1,0,12,114,97,100,105,97,110,115,0,0,0,1,4,102,108,111,97,116,95,99,111,115,105,110, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,97,100,105,97,110,115,0,59,120,0,0,0,4, -102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114, -97,100,105,97,110,115,0,59,121,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114, -101,116,86,97,108,0,59,122,0,0,18,114,97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95, -99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,110,115, -0,59,119,0,0,0,0,1,0,9,0,116,97,110,0,1,1,0,9,97,110,103,108,101,0,0,0,1,3,2,1,9,1,115,0,2,58,115, -105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,9,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0, -0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,10,0,116,97,110,0,1,1,0,10,97,110,103,108,101,0,0,0,1,3,2,1, -10,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,10,1,99,0,2,58,99,111,115,0,18, -97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,11,0,116,97,110,0,1,1,0,11,97,110,103, -108,101,0,0,0,1,3,2,1,11,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,11,1,99, -0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,12,0,116,97,110, -0,1,1,0,12,97,110,103,108,101,0,0,0,1,3,2,1,12,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0, -0,0,0,0,3,2,1,12,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0, -0,1,0,9,0,97,115,105,110,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,97,48,0,2,17,49,0,53,55,48,55,50,56,56,0, -0,0,0,3,2,1,9,1,97,49,0,2,17,48,0,50,49,50,49,49,52,52,0,0,54,0,0,3,2,1,9,1,97,50,0,2,17,48,0,48, -55,52,50,54,49,48,0,0,0,0,3,2,1,9,1,104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0,17, -48,0,53,0,0,48,0,0,3,2,1,9,1,121,0,2,58,97,98,115,0,18,120,0,0,0,0,0,9,18,95,95,114,101,116,86,97, -108,0,18,104,97,108,102,80,105,0,58,115,113,114,116,0,17,49,0,48,0,0,18,121,0,47,0,0,18,97,48,0,18, -121,0,18,97,49,0,18,97,50,0,18,121,0,48,46,48,46,48,47,58,115,105,103,110,0,18,120,0,0,0,48,20,0,0, -1,0,10,0,97,115,105,110,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97, -115,105,110,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105, -110,0,18,118,0,59,121,0,0,0,20,0,0,1,0,11,0,97,115,105,110,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,122,0,58,97,115,105,110,0,18,118,0,59,122,0,0,0,20,0,0,1,0,12,0,97,115,105,110,0,1,1,0,12, -118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,18,118,0,59,120,0,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,0,59,121,0,0,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,115,105,110,0,18,118,0,59,122,0,0,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,59,119,0,58,97,115,105,110,0,18,118,0,59,119,0,0,0,20,0,0,1,0,9,0,97,99, -111,115,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0,0, -17,48,0,53,0,0,48,0,0,9,18,95,95,114,101,116,86,97,108,0,18,104,97,108,102,80,105,0,58,97,115,105, -110,0,18,120,0,0,0,47,20,0,0,1,0,10,0,97,99,111,115,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,0,1,0,11,0,97,99,111,115,0,1,1,0,11, -118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115,0,18,118,0,59,122,0,0,0,20,0,0,1,0,12,0, -97,99,111,115,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115, -0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118, -0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115,0,18,118,0,59,122, -0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,99,111,115,0,18,118,0,59,119,0,0,0,20, -0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,97,115,105, -110,0,18,120,0,58,105,110,118,101,114,115,101,115,113,114,116,0,18,120,0,18,120,0,48,17,49,0,48,0, -0,46,0,0,48,0,0,20,0,0,1,0,10,0,97,116,97,110,0,1,1,0,10,121,95,111,118,101,114,95,120,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59, -120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,121,95,111,118, -101,114,95,120,0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97,110,0,1,1,0,11,121,95,111,118,101,114,95, -120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101, -114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18, -121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58, -97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,122,0,0,0,20,0,0,1,0,12,0,97,116,97,110,0,1, -1,0,12,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116, -97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114, +4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18, +114,97,100,105,97,110,115,0,59,122,0,0,0,4,102,108,111,97,116,95,99,111,115,105,110,101,0,18,95,95, +114,101,116,86,97,108,0,59,119,0,0,18,114,97,100,105,97,110,115,0,59,119,0,0,0,0,1,0,0,9,0,116,97, +110,0,1,1,0,9,97,110,103,108,101,0,0,0,1,3,2,1,0,9,1,115,0,2,58,115,105,110,0,18,97,110,103,108, +101,0,0,0,0,0,3,2,1,0,9,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99, +0,49,0,0,1,0,0,10,0,116,97,110,0,1,1,0,10,97,110,103,108,101,0,0,0,1,3,2,1,0,10,1,115,0,2,58,115, +105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,0,10,1,99,0,2,58,99,111,115,0,18,97,110,103,108, +101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,0,11,0,116,97,110,0,1,1,0,11,97,110,103,108,101,0,0,0, +1,3,2,1,0,11,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0,0,3,2,1,0,11,1,99,0,2,58,99, +111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1,0,0,12,0,116,97,110,0,1,1,0, +12,97,110,103,108,101,0,0,0,1,3,2,1,0,12,1,115,0,2,58,115,105,110,0,18,97,110,103,108,101,0,0,0,0, +0,3,2,1,0,12,1,99,0,2,58,99,111,115,0,18,97,110,103,108,101,0,0,0,0,0,8,18,115,0,18,99,0,49,0,0,1, +0,0,9,0,97,115,105,110,0,1,1,0,9,120,0,0,0,1,3,2,1,0,9,1,97,48,0,2,17,49,0,53,55,48,55,50,56,56,0, +0,0,0,3,2,1,0,9,1,97,49,0,2,17,48,0,50,49,50,49,49,52,52,0,0,54,0,0,3,2,1,0,9,1,97,50,0,2,17,48,0, +48,55,52,50,54,49,48,0,0,0,0,3,2,1,0,9,1,104,97,108,102,80,105,0,2,17,51,0,49,52,49,53,57,50,54,0, +0,17,48,0,53,0,0,48,0,0,3,2,1,0,9,1,121,0,2,58,97,98,115,0,18,120,0,0,0,0,0,9,18,95,95,114,101,116, +86,97,108,0,18,104,97,108,102,80,105,0,58,115,113,114,116,0,17,49,0,48,0,0,18,121,0,47,0,0,18,97, +48,0,18,121,0,18,97,49,0,18,97,50,0,18,121,0,48,46,48,46,48,47,58,115,105,103,110,0,18,120,0,0,0, +48,20,0,0,1,0,0,10,0,97,115,105,110,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,0,58,97,115,105,110,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0, +58,97,115,105,110,0,18,118,0,59,121,0,0,0,20,0,0,1,0,0,11,0,97,115,105,110,0,1,1,0,11,118,0,0,0,1, +9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,18,118,0,59,120,0,0,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,122,0,58,97,115,105,110,0,18,118,0,59,122,0,0,0,20,0,0,1,0,0,12,0,97,115, +105,110,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,115,105,110,0,18, +118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,115,105,110,0,18,118,0, +59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,115,105,110,0,18,118,0,59,122, +0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,115,105,110,0,18,118,0,59,119,0,0,0, +20,0,0,1,0,0,9,0,97,99,111,115,0,1,1,0,9,120,0,0,0,1,3,2,1,0,9,1,104,97,108,102,80,105,0,2,17,51,0, +49,52,49,53,57,50,54,0,0,17,48,0,53,0,0,48,0,0,9,18,95,95,114,101,116,86,97,108,0,18,104,97,108, +102,80,105,0,58,97,115,105,110,0,18,120,0,0,0,47,20,0,0,1,0,0,10,0,97,99,111,115,0,1,1,0,10,118,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,0,1,0,0,11, +0,97,99,111,115,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,99,111, +115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,99,111,115,0,18, +118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,99,111,115,0,18,118,0,59, +122,0,0,0,20,0,0,1,0,0,12,0,97,99,111,115,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,59,120,0,58,97,99,111,115,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121, +0,58,97,99,111,115,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97, +99,111,115,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,99,111, +115,0,18,118,0,59,119,0,0,0,20,0,0,1,0,0,9,0,97,116,97,110,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,58,97,115,105,110,0,18,120,0,58,105,110,118,101,114,115,101,115,113,114,116,0, +18,120,0,18,120,0,48,17,49,0,48,0,0,46,0,0,48,0,0,20,0,0,1,0,0,10,0,97,116,97,110,0,1,1,0,10,121, +95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0, +18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0, +58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,0,1,0,0,11,0,97,116,97,110, +0,1,1,0,11,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97, +116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,121,0,0,0,20,0,9,18,95,95,114, 101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,122,0,0,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120, -0,59,119,0,0,0,20,0,0,1,0,9,0,97,116,97,110,0,1,1,0,9,121,0,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,114,0, -0,0,10,58,97,98,115,0,18,120,0,0,0,17,49,0,48,0,45,52,0,41,0,2,9,18,114,0,58,97,116,97,110,0,18, -121,0,18,120,0,49,0,0,20,0,10,18,120,0,17,48,0,48,0,0,40,0,2,9,18,114,0,18,114,0,58,115,105,103, -110,0,18,121,0,0,0,17,51,0,49,52,49,53,57,51,0,0,48,46,20,0,0,9,14,0,0,2,9,18,114,0,58,115,105,103, -110,0,18,121,0,0,0,17,49,0,53,55,48,55,57,54,53,0,0,48,20,0,0,8,18,114,0,0,0,1,0,10,0,97,116,97, -110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116, -97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121, -0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,0,1,0,11,0,97,116,97,110,0,1,1, -0,11,117,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110,0, -18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97, -116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,0,1,0,12,0,97,116,97,110,0, -1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,97,116,97,110, -0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,97, -116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,119,0,58,97,116,97,110,0,18,117,0,59,119,0,0,18,118,0,59,119,0,0,0,20,0,0,1,0,9,0,112, -111,119,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95, -114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,112,111,119,0,1,1,0,10,97,0,0,1,1,0,10, -98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0, -0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95, -114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,0,1,0,11,0,112,111,119, -0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112, -111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0, -0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18, -97,0,59,122,0,0,18,98,0,59,122,0,0,0,0,1,0,12,0,112,111,119,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1, -4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0, -59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116, -86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119, -101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,4, -102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59, -119,0,0,18,98,0,59,119,0,0,0,0,1,0,9,0,101,120,112,0,1,1,0,9,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0, -55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97, -108,0,0,18,101,0,0,18,97,0,0,0,0,1,0,10,0,101,120,112,0,1,1,0,10,97,0,0,0,1,3,2,1,9,1,101,0,2,17, -50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86, -97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0, -18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,0,1,0,11,0,101,120,112, -0,1,1,0,11,97,0,0,0,1,3,2,1,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112, -111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4, -102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0, -18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108, -0,59,122,0,0,18,101,0,0,18,97,0,59,122,0,0,0,0,1,0,12,0,101,120,112,0,1,1,0,12,97,0,0,0,1,3,2,1,9, -1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95, -114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111, -119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,4,102, -108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18, -97,0,59,122,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0, -59,119,0,0,18,101,0,0,18,97,0,59,119,0,0,0,0,1,0,9,0,108,111,103,50,0,1,1,0,9,120,0,0,0,1,4,102, -108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,10,0,108, -111,103,50,0,1,1,0,10,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86, -97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114, -101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,0,11,0,108,111,103,50,0,1,1,0,11,118,0,0, -0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0, -59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0, -18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59, -122,0,0,18,118,0,59,122,0,0,0,0,1,0,12,0,108,111,103,50,0,1,1,0,12,118,0,0,0,1,4,102,108,111,97, -116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102, -108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0, -0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59, -122,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18, -118,0,59,119,0,0,0,0,1,0,9,0,108,111,103,0,1,1,0,9,120,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51, -49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,120,0,0,0,18,99,0,48,0,0,1,0,10,0,108,111,103,0, -1,1,0,10,118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103, -50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,11,0,108,111,103,0,1,1,0,11,118,0,0,0,1,3,2,1,9,1,99,0,2,17, -48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,12,0, -108,111,103,0,1,1,0,12,118,0,0,0,1,3,2,1,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8, -58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,9,0,101,120,112,50,0,1,1,0,9,97,0,0,0,1,4,102, -108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,10,0,101, -120,112,50,0,1,1,0,10,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86, -97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101, -116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,0,1,0,11,0,101,120,112,50,0,1,1,0,11,97,0,0,0,1,4, -102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120, -0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0, -59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0, -18,97,0,59,122,0,0,0,0,1,0,12,0,101,120,112,50,0,1,1,0,12,97,0,0,0,1,4,102,108,111,97,116,95,101, -120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116, -95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111, -97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,4,102, -108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0, -0,0,1,0,9,0,115,113,114,116,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114, -115,113,0,18,114,0,0,18,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97, -108,0,0,18,114,0,0,0,0,1,0,10,0,115,113,114,116,0,1,1,0,10,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102, -108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99, -112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113, -0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97, -108,0,59,121,0,0,18,114,0,0,0,0,1,0,11,0,115,113,114,116,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,114,0,0, -0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114, -115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101, -116,86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0, -59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18, -114,0,0,0,0,1,0,12,0,115,113,114,116,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,114,0,0,0,4,102,108,111,97, +0,0,1,0,0,12,0,97,116,97,110,0,1,1,0,12,121,95,111,118,101,114,95,120,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,59,120,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59,120,0,0,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,121,95,111,118,101,114,95,120,0,59, +121,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,121,95,111,118, +101,114,95,120,0,59,122,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0, +18,121,95,111,118,101,114,95,120,0,59,119,0,0,0,20,0,0,1,0,0,9,0,97,116,97,110,0,1,1,0,9,121,0,0,1, +1,0,9,120,0,0,0,1,3,2,0,0,9,1,114,0,0,0,10,58,97,98,115,0,18,120,0,0,0,17,49,0,48,0,45,52,0,41,0,2, +9,18,114,0,58,97,116,97,110,0,18,121,0,18,120,0,49,0,0,20,0,10,18,120,0,17,48,0,48,0,0,40,0,2,9,18, +114,0,18,114,0,58,115,105,103,110,0,18,121,0,0,0,17,51,0,49,52,49,53,57,51,0,0,48,46,20,0,0,9,14,0, +0,2,9,18,114,0,58,115,105,103,110,0,18,121,0,0,0,17,49,0,53,55,48,55,57,54,53,0,0,48,20,0,0,8,18, +114,0,0,0,1,0,0,10,0,97,116,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0, +0,1,0,0,11,0,97,116,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122,0, +0,0,20,0,0,1,0,0,12,0,97,116,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,59,120,0,58,97,116,97,110,0,18,117,0,59,120,0,0,18,118,0,59,120,0,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,58,97,116,97,110,0,18,117,0,59,121,0,0,18,118,0,59,121,0,0,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,97,116,97,110,0,18,117,0,59,122,0,0,18,118,0,59,122, +0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,97,116,97,110,0,18,117,0,59,119,0,0,18, +118,0,59,119,0,0,0,20,0,0,1,0,0,9,0,112,111,119,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,102,108,111, +97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0, +10,0,112,111,119,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114, +0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108, +111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0, +18,98,0,59,121,0,0,0,0,1,0,0,11,0,112,111,119,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,102,108,111, +97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18, +98,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0, +59,121,0,0,18,97,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0, +18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,0,1,0,0,12,0, +112,111,119,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,102,108,111,97,116,95,112,111,119,101,114,0,18, +95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97, +116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,18,98, +0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59, +122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18, +95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,18,98,0,59,119,0,0,0,0,1,0,0,9,0,101, +120,112,0,1,1,0,9,97,0,0,0,1,3,2,1,0,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97, +116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,0,18,101,0,0,18,97,0,0,0,0,1,0,0,10, +0,101,120,112,0,1,1,0,10,97,0,0,0,1,3,2,1,0,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108, +111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0, +59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121, +0,0,18,101,0,0,18,97,0,59,121,0,0,0,0,1,0,0,11,0,101,120,112,0,1,1,0,11,97,0,0,0,1,3,2,1,0,9,1,101, +0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101, +116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101, +114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97, +116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18,97,0,59,122, +0,0,0,0,1,0,0,12,0,101,120,112,0,1,1,0,12,97,0,0,0,1,3,2,1,0,9,1,101,0,2,17,50,0,55,49,56,50,56,0, +0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, +101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86, +97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0, +18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116, +95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,101,0,0,18,97,0,59,119,0,0, +0,0,1,0,0,9,0,108,111,103,50,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95, +95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,10,0,108,111,103,50,0,1,1,0,10,118,0,0,0,1,4,102, +108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0, +0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59, +121,0,0,0,0,1,0,0,11,0,108,111,103,50,0,1,1,0,11,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103, +50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108, +111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97, +116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,0,0, +12,0,108,111,103,50,0,1,1,0,12,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114, +101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18, +95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103, +50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,108, +111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,0,9,0,108,111, +103,0,1,1,0,9,120,0,0,0,1,3,2,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108, +111,103,50,0,18,120,0,0,0,18,99,0,48,0,0,1,0,0,10,0,108,111,103,0,1,1,0,10,118,0,0,0,1,3,2,1,0,9,1, +99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0, +0,1,0,0,11,0,108,111,103,0,1,1,0,11,118,0,0,0,1,3,2,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56, +49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,0,12,0,108,111,103,0,1,1,0,12,118, +0,0,0,1,3,2,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118, +0,0,0,18,99,0,48,0,0,1,0,0,9,0,101,120,112,50,0,1,1,0,9,97,0,0,0,1,4,102,108,111,97,116,95,101,120, +112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,10,0,101,120,112,50,0,1,1,0,10,97, +0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97, +0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0, +0,18,97,0,59,121,0,0,0,0,1,0,0,11,0,101,120,112,50,0,1,1,0,11,97,0,0,0,1,4,102,108,111,97,116,95, +101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97, +116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108, +111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,0, +1,0,0,12,0,101,120,112,50,0,1,1,0,12,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95, +114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0, +18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120, +112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95, +101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,0,0,1,0,0,9,0,115, +113,114,116,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18, +114,0,0,18,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18, +114,0,0,0,0,1,0,0,10,0,115,113,114,116,0,1,1,0,10,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4,102,108,111, +97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114, +0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, +121,0,0,18,114,0,0,0,0,1,0,0,11,0,115,113,114,116,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4, +102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114, +99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115, +113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116, +86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59, +122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114, +0,0,0,0,1,0,0,12,0,115,113,114,116,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4,102,108,111,97, 116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95, 95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0, 0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, 121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4,102, 108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,4,102,108, 111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,0,0,0,0,1,0,9,0,105,110,118,101,114,115,101,115, -113,114,116,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97, -108,0,59,120,0,0,18,120,0,0,0,0,1,0,10,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,10, +18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,0,0,0,0,1,0,0,9,0,105,110,118,101,114,115,101, +115,113,114,116,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116, +86,97,108,0,59,120,0,0,18,120,0,0,0,0,1,0,0,10,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1, +0,10,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, +18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59, +121,0,0,18,118,0,59,121,0,0,0,0,1,0,0,11,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,11, +118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, +118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0, +0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59, +122,0,0,18,118,0,59,122,0,0,0,0,1,0,0,12,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,12, 118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, 118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0, -0,18,118,0,59,121,0,0,0,0,1,0,11,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,11,118,0,0, -0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59, -120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118, -0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0, -18,118,0,59,122,0,0,0,0,1,0,12,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,12,118,0,0,0, -1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59, -120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118, -0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0, -18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59, -119,0,0,18,118,0,59,119,0,0,0,0,1,0,9,0,110,111,114,109,97,108,105,122,101,0,1,1,0,9,120,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,17,49,0,48,0,0,20,0,0,1,0,10,0,110,111,114,109,97,108,105,122,101, -0,1,1,0,10,118,0,0,0,1,3,2,1,9,1,115,0,2,58,105,110,118,101,114,115,101,115,113,114,116,0,58,100, -111,116,0,18,118,0,0,18,118,0,0,0,0,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,115,0,0,0,0,1,0,11,0,110,111,114,109,97, -108,105,122,101,0,1,1,0,11,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118,101,99,51,95,100,111,116, -0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112, -0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,0,12,0,110,111,114,109,97,108, -105,122,101,0,1,1,0,12,118,0,0,0,1,3,2,0,9,1,116,109,112,0,0,0,4,118,101,99,52,95,100,111,116,0,18, -116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0, -18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,0,9,0,97,98,115,0,1,1,0,9,97,0,0,0, -1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,10,0,97,98, -115,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,97,0,0,0,0,1,0,11,0,97,98,115,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,12,0,97,98,115,0,1,1,0,12,97,0, -0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,115, -105,103,110,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0, -18,112,0,0,18,120,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0, -0,0,18,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108, -0,0,18,112,0,0,18,110,0,0,0,0,1,0,10,0,115,105,103,110,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,112,0,0,1, -1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0, -4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,121,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101, -99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,0, -18,110,0,0,0,0,1,0,11,0,115,105,103,110,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,112,0,0,1,1,110,0,0,0,4, -118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101, -99,52,95,115,103,116,0,18,110,0,59,120,121,122,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52, -95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,0, -18,110,0,0,0,0,1,0,12,0,115,105,103,110,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,112,0,0,1,1,110,0,0,0,4, -118,101,99,52,95,115,103,116,0,18,112,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103, -116,0,18,110,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, -18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,0,9,0,102,108,111,111,114,0,1,1,0,9, -97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0, -0,1,0,10,0,102,108,111,111,114,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,108,111,111,114,0,1,1,0,11, -97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,97,0,0,0,0,1,0,12,0,102,108,111,111,114,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102, -108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,99,101,105,108,0,1,1,0, -9,97,0,0,0,1,3,2,0,9,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18, -98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,10,0,99,101,105,108,0,1,1,0,10, -97,0,0,0,1,3,2,0,10,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18, -98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,54,20,0,0,1,0,11,0,99,101,105,108, -0,1,1,0,11,97,0,0,0,1,3,2,0,11,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18, -98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,98,0,54,20,0,0,1,0,12,0, -99,101,105,108,0,1,1,0,12,97,0,0,0,1,3,2,0,12,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108, -111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,9,0, -102,114,97,99,116,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86, -97,108,0,0,18,97,0,0,0,0,1,0,10,0,102,114,97,99,116,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102, -114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,11,0,102,114,97,99, -116,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,97,0,0,0,0,1,0,12,0,102,114,97,99,116,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95, -102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,9,0,109,111,100,0,1,1,0,9,97, -0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97, -0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0, -1,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66, -0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111, -110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,9,98, -0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, -110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18, -97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0, -0,1,0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114, -66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9, -18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101, -79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1, -3,2,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0, -18,97,0,59,120,0,18,111,110,101,79,118,101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0, -18,111,110,101,79,118,101,114,66,121,0,48,0,0,48,47,20,0,0,1,0,11,0,109,111,100,0,1,1,0,11,97,0,0, -1,1,0,11,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114, -66,121,0,0,1,1,111,110,101,79,118,101,114,66,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, -110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, -110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111, -110,101,79,118,101,114,66,122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0, -18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0,18,111,110,101,79,118, -101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59,121,0,18, -98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79,118,101,114,66,121,0,48, -0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122,0,18,98,0,59,122,0,58, -102,108,111,111,114,0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48,47,20,0,0, -1,0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,3,2,0,9,1,111,110,101,79,118,101,114,66, -120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,1,1,111,110,101,79,118,101,114,66,122,0,0,1,1, -111,110,101,79,118,101,114,66,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, -101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, -101,114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, -101,114,66,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, -101,114,66,119,0,0,18,98,0,59,119,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,59,120, -0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0,18,111,110,101,79,118,101,114,66,120, -0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59,121,0, -58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79,118,101,114,66,121,0,48,0,0,48,47,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122,0,18,98,0,59,122,0,58,102,108,111,111, -114,0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48,47,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,119,0,18,97,0,59,119,0,18,98,0,59,119,0,58,102,108,111,111,114,0,18,97,0,59,119, -0,18,111,110,101,79,118,101,114,66,119,0,48,0,0,48,47,20,0,0,1,0,9,0,109,105,110,0,1,1,0,9,97,0,0, +0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59, +122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108, +0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,0,9,0,110,111,114,109,97,108,105,122,101,0,1,1,0,9,120,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,17,49,0,48,0,0,20,0,0,1,0,0,10,0,110,111,114,109,97,108, +105,122,101,0,1,1,0,10,118,0,0,0,1,3,2,1,0,9,1,115,0,2,58,105,110,118,101,114,115,101,115,113,114, +116,0,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,115,0,0,0,0,1,0,0,11,0,110, +111,114,109,97,108,105,122,101,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,116,109,112,0,0,0,4,118,101,99, +51,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113, +0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,0,0,12,0,110, +111,114,109,97,108,105,122,101,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,116,109,112,0,0,0,4,118,101,99, +52,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113, +0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,0,0,9,0,97,98, +115,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0, +0,0,0,1,0,0,10,0,97,98,115,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,11,0,97,98,115,0,1,1,0,11,97,0,0,0,1,4,118,101, +99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,12,0, +97,98,115,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0, +18,97,0,0,0,0,1,0,0,9,0,115,105,103,110,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9,1,112,0,0,1,1,110,0,0,0,4, +118,101,99,52,95,115,103,116,0,18,112,0,0,18,120,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103, +116,0,18,110,0,0,17,48,0,48,0,0,0,18,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, +18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,0,0,10,0,115,105,103,110,0,1,1,0,10, +118,0,0,0,1,3,2,0,0,10,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120, +121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,121,0,0,17, +48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,112,0,0,18,110,0,0,0,0,1,0,0,11,0,115,105,103,110,0,1,1,0,11,118,0,0, +0,1,3,2,0,0,11,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,122,0, +0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,121,122,0,0,17,48, +0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,112,0,0,18,110,0,0,0,0,1,0,0,12,0,115,105,103,110,0,1,1,0,12,118,0, +0,0,1,3,2,0,0,12,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,0,18,118,0,0,17, +48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118, +101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0, +0,0,0,1,0,0,9,0,102,108,111,111,114,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0, +18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,10,0,102,108,111,111,114,0,1,1,0,10,97,0,0, +0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97, +0,0,0,0,1,0,0,11,0,102,108,111,111,114,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111, +114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,12,0,102,108,111,111, +114,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108, +0,0,18,97,0,0,0,0,1,0,0,9,0,99,101,105,108,0,1,1,0,9,97,0,0,0,1,3,2,0,0,9,1,98,0,2,18,97,0,54,0,0, +4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0, +18,98,0,54,20,0,0,1,0,0,10,0,99,101,105,108,0,1,1,0,10,97,0,0,0,1,3,2,0,0,10,1,98,0,2,18,97,0,54,0, +0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108, +0,59,120,121,0,18,98,0,54,20,0,0,1,0,0,11,0,99,101,105,108,0,1,1,0,11,97,0,0,0,1,3,2,0,0,11,1,98,0, +2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,18,98,0,54,20,0,0,1,0,0,12,0,99,101,105,108,0,1,1,0,12,97,0,0,0,1, +3,2,0,0,12,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0, +9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,0,9,0,102,114,97,99,116,0,1,1,0,9,97,0,0, +0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,10,0, +102,114,97,99,116,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,11,0,102,114,97,99,116,0,1,1,0,11,97,0,0,0,1,4,118,101, +99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0, +12,0,102,114,97,99,116,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101, +116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,9,0,109,111,100,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0, +9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, +101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111, +114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,10,0,109,111,100,0,1,1, +0,10,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114, +66,0,48,0,0,48,47,20,0,0,1,0,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1, +111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, +114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,97,0,18,98,0,58,102, +108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,12,0,109, +111,100,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101, +116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66, +0,48,0,0,48,47,20,0,0,1,0,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,3,2,0,0,9,1,111, +110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,111,110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,111,110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,9,18,95,95,114,101, +116,86,97,108,0,59,120,0,18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120, +0,18,111,110,101,79,118,101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59, +121,0,18,97,0,59,121,0,18,98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101, +79,118,101,114,66,121,0,48,0,0,48,47,20,0,0,1,0,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,11,98,0, +0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0, +1,1,111,110,101,79,118,101,114,66,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79, +118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79, +118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79, +118,101,114,66,122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,59, +120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0,18,111,110,101,79,118,101,114,66, +120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59, +121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79,118,101,114,66,121,0,48,0,0,48, +47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122,0,18,98,0,59,122,0,58,102,108, +111,111,114,0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48,47,20,0,0,1,0,0,12, +0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,120, +0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,1,1,111,110,101,79,118,101,114,66,122,0,0,1,1,111, +110,101,79,118,101,114,66,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, +114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, +114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, +114,66,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, +114,66,119,0,0,18,98,0,59,119,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,59,120,0, +18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0,18,111,110,101,79,118,101,114,66,120,0, +48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59,121,0,58, +102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79,118,101,114,66,121,0,48,0,0,48,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122,0,18,98,0,59,122,0,58,102,108,111,111,114, +0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116, +86,97,108,0,59,119,0,18,97,0,59,119,0,18,98,0,59,119,0,58,102,108,111,111,114,0,18,97,0,59,119,0, +18,111,110,101,79,118,101,114,66,119,0,48,0,0,48,47,20,0,0,1,0,0,9,0,109,105,110,0,1,1,0,9,97,0,0, 1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, -18,98,0,0,0,0,1,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109, +18,98,0,0,0,0,1,0,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109, 105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120, -121,0,0,0,0,1,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,105, -110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59, -120,121,122,0,0,0,0,1,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95, -109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,109,105,110,0,1, -1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0, -0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4, -118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98, -0,0,0,0,1,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0, -18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0,109,97,120,0,1,1,0,9,97,0,0,1,1, -0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98, -0,0,0,0,1,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0, -1,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0, -0,1,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95, -95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,9, -98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,0, -0,18,98,0,0,0,0,1,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97, -120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,12,0,109,97, -120,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86, -97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0,99,108,97,109,112,0,1,1,0,9,118,97,108,0,0,1,1,0,9,109, +121,0,0,0,0,1,0,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109, +105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0, +59,120,121,122,0,0,0,0,1,0,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99, +52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,105, +110,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86, +97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,9, +98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121, +122,0,0,18,98,0,0,0,0,1,0,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52, +95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,0,109,97,120,0, +1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0, +0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99, +52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0, +59,120,121,0,0,0,0,1,0,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95, +109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98, +0,59,120,121,122,0,0,0,0,1,0,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99, +52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,97, +120,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86, +97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,9,98, +0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122, +0,0,18,98,0,0,0,0,1,0,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, +109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,0,99,108,97,109, +112,0,1,1,0,9,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0, +1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18, +109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,10,0,99,108,97,109,112,0,1,1,0,10, +118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99, +52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86, +97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0, +1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108, +97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18, +109,97,120,86,97,108,0,0,0,0,1,0,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,9,109, 105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0, 18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86, -97,108,0,0,0,0,1,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108, -0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101, -116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1, -0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109, -97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0, -18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,12,0,99,108,97, -109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0, -0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0, -18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,10,0,99,108,97,109,112,0,1,1,0,10, -118,97,108,0,0,1,1,0,10,109,105,110,86,97,108,0,0,1,1,0,10,109,97,120,86,97,108,0,0,0,1,4,118,101, -99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110, -86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0, -1,1,0,11,109,105,110,86,97,108,0,0,1,1,0,11,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108, -97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18, -109,97,120,86,97,108,0,0,0,0,1,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,12,109,105, -110,86,97,108,0,0,1,1,0,12,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18, -95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97, -108,0,0,0,0,1,0,9,0,109,105,120,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99, -52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,10, -0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114, -112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0,109,105,120, -0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95, -114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1,1,0,12,120, -0,0,1,1,0,12,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86, -97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10, -121,0,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18, -97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11, -97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0, -0,18,120,0,0,0,0,1,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4, -118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0, -0,0,1,0,9,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95, -115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1, -0,10,0,115,116,101,112,0,1,1,0,10,101,100,103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115, -103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1, -0,11,0,115,116,101,112,0,1,1,0,11,101,100,103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115, -103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0, -0,1,0,12,0,115,116,101,112,0,1,1,0,12,101,100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95, -115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,10,0, -115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,0,0,0, -0,1,0,11,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, -115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0, -59,120,120,120,0,0,0,0,1,0,12,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1, -4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101, -0,59,120,120,120,120,0,0,0,0,1,0,9,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103, -101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,9,1,116,0,2,58,99,108,97,109, +97,108,0,0,0,0,1,0,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,10,109,105,110,86,97, +108,0,0,1,1,0,10,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114, +101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0, +0,1,0,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,11,109,105,110,86,97,108,0,0,1,1,0, +11,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97, +108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,12,0, +99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,12,109,105,110,86,97,108,0,0,1,1,0,12,109,97,120, +86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118, +97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,9,0,109,105,120,0,1, +1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101, +116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1, +1,0,10,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108, +0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0, +1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, +18,121,0,0,18,120,0,0,0,0,1,0,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,9,97,0, +0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18, +120,0,0,0,0,1,0,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,10,97,0,0,0,1,4,118, +101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0, +1,0,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95, +108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,12,0, +109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,108,114,112, +0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,9,0,115,116,101,112, +0,1,1,0,9,101,100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114, +101,116,86,97,108,0,59,120,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0, +1,1,0,10,101,100,103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101, +112,0,1,1,0,11,101,100,103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115, +116,101,112,0,1,1,0,12,101,100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0, +18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101, +112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,0,0,0,0,1,0,0,11, +0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0,59,120, +120,120,0,0,0,0,1,0,0,12,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1,4, +118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101,0, +59,120,120,120,120,0,0,0,0,1,0,0,9,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103, +101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9,1,116,0,2,58,99,108,97,109, 112,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, 17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,10,101,100,103,101,48, -0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,116,0,2,58,99,108,97,109,112,0, -18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48, -0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48, -47,48,0,0,1,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,11,101,100,103,101,48,0,0,1,1,0, -11,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0, -18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0, -0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0, -0,1,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,12,101,100,103,101,48,0,0,1,1,0,12,101, -100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101, -100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49, -0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,10, -0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101, -49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101, -48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0, -0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,11,0,115,109, -111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1, -1,0,11,118,0,0,0,1,3,2,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47, -18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8, -18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,12,0,115,109,111,111, -116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,12, -118,0,0,0,1,3,2,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101, -100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116, -0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,9,0,108,101,110,103,116,104, -0,1,1,0,9,120,0,0,0,1,8,58,97,98,115,0,18,120,0,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,10, -118,0,0,0,1,3,2,0,9,1,114,0,0,0,3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4, -102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0, -11,118,0,0,0,1,3,2,0,9,1,114,0,0,0,3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0, -4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0, -0,0,1,3,2,0,9,1,114,0,0,0,3,2,1,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108, -111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95, -114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1, -1,0,9,121,0,0,0,1,3,2,1,9,1,100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, -108,101,110,103,116,104,0,18,100,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118, -0,0,1,1,0,10,117,0,0,0,1,3,2,1,10,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86, -97,108,0,58,108,101,110,103,116,104,0,18,100,50,0,0,0,20,0,0,1,0,9,0,100,105,115,116,97,110,99,101, -0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,1,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95, -114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,9,0,100,105,115, -116,97,110,99,101,0,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,12,1,100,52,0,2,18,118,0,18,117,0, -47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0, -11,0,99,114,111,115,115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115, -115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,9,0,102,97, -99,101,102,111,114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0,1,3, -2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118, -101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54, -0,18,78,0,0,18,115,0,0,0,0,0,1,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10,78,0,0,1, -1,0,10,73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102, -0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0, -0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,11,0,102,97,99,101, +116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,10,101,100,103,101, +48,0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99,108,97,109, +112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, +17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, +116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,11,101,100,103,101, +48,0,0,1,1,0,11,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97,109, +112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, +17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, +116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,12,101,100,103,101, +48,0,0,1,1,0,12,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109, +112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, +17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, +116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48, +0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99,108,97,109,112, +0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17, +48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0, +48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1, +1,0,9,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97,109,112,0,18, +118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0, +48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48, +47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1, +0,9,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118, +0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0, +0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48, +0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,9,120,0,0,0,1,8,58,97,98,115,0,18,120,0,0,0,0,0,1,0, +0,9,0,108,101,110,103,116,104,0,1,1,0,10,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0,2,58, +100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112, +0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0, +0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112, +0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0, +18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0, +0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0, +2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18, +112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1, +0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,3,2,1,0,9,1,100,0,2,18, +120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,0,0,0, +20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,1,0,10, +1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116, +104,0,18,100,50,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,11,118,0,0,1,1,0,11, +117,0,0,0,1,3,2,1,0,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, +108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0, +12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,0,12,1,100,52,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114, +101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0,0,11,0,99,114,111,115, +115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115,115,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,9,0,102,97,99,101,102,111, +114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100, +0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52, +95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0, +0,18,115,0,0,0,0,0,1,0,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10,78,0,0,1,1,0,10, +73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0, +18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0, +18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,11,0,102,97,99,101, 102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0,1,1,0,11,78,114,101,102,0,0,0,1,3,2,1, -9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101, -99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18, -78,0,0,18,115,0,0,0,0,0,1,0,12,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,12,78,0,0,1,1,0, -12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0, -18,73,0,0,0,0,0,3,2,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0, -18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,9,0,114,101,102,108, -101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0, -18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,10,0,114,101,102,108,101,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78, -0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0, -11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58, -100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,12,0,114,101,102,108,101,99,116,0,1,1, -0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48, -18,78,0,48,47,0,0,1,0,9,0,114,101,102,114,97,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,1,1,0,9,101, -116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0, -58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10, -18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97, -0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0, -1,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3, -2,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0, -18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0, -48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116, -0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,11,0,114,101, -102,114,97,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2, -17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73, -0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17, -48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73, -0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,12,0,114,101,102,114,97,99,116, -0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,9,1,107,0,2,17,49,0,48,0,0,18, -101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111, -116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9, -14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115, -113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,13,0,109,97,116,114,105,120,67,111,109,112,77, -117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97,116,50,0,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,14,0,109,97, -116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1,0,0,14,110,0,0,0,1,8,58,109,97, -116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49, -0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,15,0,109,97,116,114,105,120, -67,111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0,0,0,1,8,58,109,97,116,52,0,18,109, -0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18, -109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0, -0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99, -52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0, -3,0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115, -108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0, -108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108, -116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104, -97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0, -7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,8,117,0,0, -1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, -18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10, -118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0, -1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0, -12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0, -0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117, -0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0, -7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108, -0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97, -108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,10,117, -0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,11,117, -0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,12, -117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0, -18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0, -6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -117,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110, -0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97, -110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86, -97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117, -97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,103,114,101,97,116,101,114,84,104, -97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101, -97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101, -99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,103, -114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118, -101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118, -0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -117,0,0,18,118,0,0,0,0,1,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0, -8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0, -18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118, -101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,3,0,101,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,101, -113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,101, -113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, -114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1,1,0,6,117,0,0,1, -1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, -18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118, -101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0, -0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,101, -113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,101,113,117,97,108,0,1, -1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,101,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3, -118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, -18,117,0,0,18,118,0,0,0,0,1,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118, -101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0, -110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111, -116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69, -113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95, -114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0, -6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1, -0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, -0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0, -0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,3,0, -110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,4,0,110,111, -116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, -95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1,3, -2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59, -120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1,1,0,3,118,0,0,0,1, -3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0, -59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18, -115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,110,121,0,1, -1,0,4,118,0,0,0,1,3,2,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, +0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118, +101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54, +0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,12,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,12,78,0,0, +1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101, +102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0, +48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,9,0,114,101, +102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0, +18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,10,0,114,101,102,108,101,99,116,0,1,1,0,10,73,0,0, +1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48, +47,0,0,1,0,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,0,1,8,18,73,0,17,50, +0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,12,0,114,101,102,108, +101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0, +0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,9,0,114,101,102,114,97,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78, +0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0, +48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48, +47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0, +48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46, +18,78,0,48,47,0,0,1,0,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,1,1,0,9, +101,116,97,0,0,0,1,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0, +48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0, +0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101, +116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48, +47,0,0,1,0,0,11,0,114,101,102,114,97,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,1,1,0,9,101,116,97,0, +0,0,1,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100, +111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0, +17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100, +111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,0,12, +0,114,101,102,114,97,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9, +1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78, +0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0, +40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78, +0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,0,13,0,109,97,116, +114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97,116, +50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0, +57,48,0,0,0,0,1,0,0,14,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1, +0,0,14,110,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0, +16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0, +1,0,0,15,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0, +0,0,1,8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0, +57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1, +1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11, +118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118, +0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0, +0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52, +95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3, +0,108,101,115,115,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108, +101,115,115,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0, +18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97, +110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115, +84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108, +101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99, +52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101, +115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95, +115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0, +108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99, +52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0, +1,0,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4, +118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0, +2,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99, +52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0, +0,3,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101, +99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118, +101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0, +103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95, +115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59, +120,121,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0, +0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117, +0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8, +118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, +0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1, +1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0, +1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97, +110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18, +95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84, +104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101, +97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99, +52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0, +1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118, +0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0, +0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115, +101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101, +113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108, +0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, +97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0, +0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0, +18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52, +95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0, +0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18, +95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,2, +117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0, +0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117, +0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99, +52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111, +116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69, +113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113, +117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114, +101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,6, +117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1, +1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, +0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8, +118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, +0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99, +52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0, +0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0, +110,111,116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101, +0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,2,118, +0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0, +18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86, +97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0, +3,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, 120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0, -59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,100,100,0,18, -115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0, -0,0,1,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0, -4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0, -48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,9,1,112,114,111,100,0,0,0,4,118,101,99, -52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121, -0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111, -100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0, -18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,1,0,97,108,108,0,1,1,0,4,118,0,0,0,1,3,2,0,9,1,112, +59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1, +0,97,110,121,0,1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0, +18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100, +100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99, +52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4, +118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120, +0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0, +0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0, +0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112, +114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,112, 114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18, 118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112, -114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,119,0,0,0,4,118,101,99, -52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0, -1,0,2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1, -4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17, -48,0,48,0,0,0,0,0,1,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95, -95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101, -49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52, -95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, -111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115, -97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112, -49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, -100,0,59,120,121,121,121,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0, -16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, -120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, -111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,0,1,1,0,17,115,97,109,112,108,101, -114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114, -101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116, -101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,11,99, -111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108, -0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,12,0, -116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0, -12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86, -97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120, -116,117,114,101,51,68,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0, -1,4,118,101,99,52,95,116,101,120,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, -101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111, -106,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, -95,116,101,120,112,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, -18,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,67,117,98,101,0,1,1,0,19,115,97, -109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,99,117, -98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, -100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,49,68,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0, -11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100, -111,119,49,68,80,114,111,106,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100, -0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97, -109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,0,1,1, -0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, -120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, -114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,21,115,97,109,112,108, -101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95, -95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0, -12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1, -0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114, -101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,12,0,116, -101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101,114,0, -0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95, -95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121, -122,122,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22, +114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0, +1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0, +4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0, +18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112, +114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52, +95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1, +0,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,4,0,110,111,116,0,1,1,0,4,118,0, +0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0, +0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1, +0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97, +108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116, +117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114, +100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115, +97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,121,121,0,0,0,0,1,0,0,12,0,116,101, +120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111, +111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0, +18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114, +101,50,68,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101, +99,52,95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, +18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0, +17,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, +120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, +111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111, +106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, +95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, +18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,0,1,1,0,18,115,97,109, +112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,51,100,0,18, +95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1, +0,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111,106,0,1,1,0,18,115,97,109,112,108,101,114,0, +0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,51,100,0,18,95,95,114,101, +116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116, +101,120,116,117,114,101,67,117,98,101,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, +111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108, +0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111, +119,49,68,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101, +99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, +18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,20, 115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120, -112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, -99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,0,1,1,0,23,115,97, -109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114, -101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, -114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,80,114,111,106,0,1,1,0,23,115, -97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112, -95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, -111,114,100,0,0,0,0,1,0,9,0,110,111,105,115,101,49,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95, -110,111,105,115,101,49,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,0,110,111,105,115, -101,49,0,1,1,0,10,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,50,0,18,95,95,114,101, -116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,0,110,111,105,115,101,49,0,1,1,0,11,120,0,0,0,1,4,102,108, -111,97,116,95,110,111,105,115,101,51,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,9,0, -110,111,105,115,101,49,0,1,1,0,12,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,52,0,18, -95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,9,120,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0, -46,0,0,20,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54, -0,0,0,0,46,0,0,20,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0, -17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,0,10,0,110,111,105,115,101,50,0,1,1,0,12, -120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, -99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0, -0,20,0,0,1,0,11,0,110,111,105,115,101,51,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0, -58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,0,1,0,11,0,110, -111,105,115,101,51,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111, -105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115, -101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17, -53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,0,11,0,110,111,105,115,101,51,0,1,1,0,11, +112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, +114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,0,1,1,0,21,115,97,109,112,108,101,114,0,0, +1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116, +86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104, +97,100,111,119,50,68,80,114,111,106,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, +114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18, +115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114, +101,50,68,82,101,99,116,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0, +0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115, +97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50, +68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114, +100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108, +0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12, +0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101, +114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0, +18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,99,111,111,114,100,0,0, +0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,0,1,1,0,23,115,97,109,112,108,101,114,0, +0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95, +114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12, +0,115,104,97,100,111,119,50,68,82,101,99,116,80,114,111,106,0,1,1,0,23,115,97,109,112,108,101,114, +0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18, +95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1, +0,0,9,0,110,111,105,115,101,49,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101, +49,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,10, +120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,50,0,18,95,95,114,101,116,86,97,108,0,0,18, +120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,11,120,0,0,0,1,4,102,108,111,97,116,95,110, +111,105,115,101,51,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115, +101,49,0,1,1,0,12,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,52,0,18,95,95,114,101, +116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,9,120,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0, +0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, +110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0, +46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54, +54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,12,120,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17, +49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,0,1, +0,0,11,0,110,111,105,115,101,51,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0, +58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110, +111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111, +105,115,101,51,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105, +115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101, +49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53, +0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,11, 120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0, 20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, 99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95, 114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0, -52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,0,1,0,11,0,110,111,105,115, -101,51,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101, +52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105, +115,101,51,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115, +101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0, +18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0, +55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0, +18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17, +49,51,0,49,57,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,9,120,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52, +55,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120, +0,17,50,51,0,53,52,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,10,120,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57, +0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110, +111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0, +0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, +99,50,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115, +101,52,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101, 49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18, -120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55, -55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18, -120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49, -51,0,49,57,0,0,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1,1,0,9,120,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0, -46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,17,50, -51,0,53,52,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0, -0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115, -101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17, -50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1,1,0, -11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, -99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0, -52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,50,51,0,53,52,0,0,0, -17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0,0,1,0,12,0,110,111,105,115,101,52,0,1, -1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120, -0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118, -101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0, -46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58, -118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57, -0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120, -0,58,118,101,99,52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,17,51,55, -0,52,56,0,0,0,0,46,0,0,20,0,0,0 +120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, +99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,50,51, +0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105, +115,101,52,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115, +101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0, +18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0, +55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0, +18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17, +49,51,0,49,57,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115, +101,49,0,18,120,0,58,118,101,99,52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57, +49,0,0,0,17,51,55,0,52,56,0,0,0,0,46,0,0,20,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index ce7a5cc64a..416ee23d20 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -2,804 +2,812 @@ /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ /* slang_core.gc */ -3,1,0,5,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101, -116,86,97,108,0,0,18,102,0,0,0,0,1,0,5,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -98,0,20,0,0,1,0,5,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,20,0,0,1,0,1,1, -1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0, -17,48,0,48,0,0,0,0,0,1,0,1,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, -116,86,97,108,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,1,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,98,0,20,0,0,1,0,9,1,1,1,0,5,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118, -101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,0,9,1,1,1,0,1,98,0,0,0,1,4,105,118, -101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,9,1,1, -1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,20,0,0,1,0,10,1,1,1,0,9,120,0,0,1,1, -0,9,121,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,121,0,18,121,0,20,0,0,1,0,10,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,109,111,118, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,0,0,1,0,10,1,1,1,0,5,105,0,0,0,1, -4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, -18,105,0,0,0,0,1,0,10,1,1,1,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,10,1,1,1,0,2,98,0,0,0,1,4,105,118,101, -99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0, -1,0,10,1,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,10,1,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,109, -111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,11,1, -1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18, -120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,122,0,18,122,0,20,0,0,1,0,11,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,0,11,1,1,1,0,5,105,0,0,0,1,4, -105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, -0,18,105,0,0,0,0,1,0,11,1,1,1,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,11,1,1,1,0,3,98,0,0,0,1,4,105, -118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -98,0,0,0,0,1,0,11,1,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,118,0,0,0,0,1,0,12,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122, -0,0,1,1,0,9,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,12,1,1,1,0,9,102,0,0,0,1,4,118, -101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,0,0,1,0,12,1,1,1,0,5, -105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0, -18,105,0,0,0,0,1,0,12,1,1,1,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95, -95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,12,1,1,1,0,4,98,0,0,0,1,4,105,118,101,99,52,95,116, -111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,12,1,1,1,0,8,105,0,0,0, -1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0, -0,0,1,0,12,1,1,1,0,11,118,51,0,0,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,18,118,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102,0,20,0,0,1,0,12,1,1,1,0, -10,118,50,0,0,1,1,0,9,102,49,0,0,1,1,0,9,102,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,18,118,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,102,49,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,119,0,18,102,50,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0, -18,106,0,20,0,0,1,0,6,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,105,0,0,0,0,1,0,6,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116, -111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,0,0,1,0,6,1, -1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,7,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,1,1,0,5,107,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121, -0,18,106,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,107,0,20,0,0,1,0,7,1,1,1,0,5,105,0, -0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -105,0,0,0,0,1,0,7,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,0,0,1,0,7,1,1,1,0,1,98,0,0,0,1,4,118,101,99, -52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,8,1, -1,1,0,5,120,0,0,1,1,0,5,121,0,0,1,1,0,5,122,0,0,1,1,0,5,119,0,0,0,1,9,18,95,95,114,101,116,86,97, +4,1,0,0,5,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101, +116,86,97,108,0,0,18,102,0,0,0,0,1,0,0,5,1,1,1,0,1,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +18,98,0,20,0,0,1,0,0,5,1,1,1,0,5,105,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,105,0,20,0,0,1, +0,0,1,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18, +105,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, +95,114,101,116,86,97,108,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,1,1,1,0,1,98,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,98,0,20,0,0,1,0,0,9,1,1,1,0,5,105,0,0,0,1,4,105,118,101,99,52,95,116, +111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,0,0,9,1,1,1,0,1,98,0,0, +0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0, +0,0,1,0,0,9,1,1,1,0,9,102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,102,0,20,0,0,1,0,0,10,1,1, +1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,0,1,0,0,10,1,1,1,0,9,102,0,0,0,1,4,118,101,99, +52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,0,0,1,0,0,10,1, +1,1,0,5,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,105,0,0,0,0,1,0,0,10,1,1,1,0,1,98,0,0,0,1,4,105,118,101,99,52,95,116,111, +95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,10,1,1,1,0, +2,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,98,0,0,0,0,1,0,0,10,1,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,0,10,1,1,1,0,12,118,0, +0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118, +0,59,120,121,0,0,0,0,1,0,0,11,1,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,0,1,0,0,11,1,1,1,0,9,102,0,0,0,1,4,118, +101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,0,0, +1,0,0,11,1,1,1,0,5,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,105,0,0,0,0,1,0,0,11,1,1,1,0,1,98,0,0,0,1,4,105,118,101,99, +52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0, +1,0,0,11,1,1,1,0,3,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,0,11,1,1,1,0,12,118,0,0,0,1,4,118,101,99,52, +95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,0,0,1,0,0,12,1, +1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,122,0,0,1,1,0,9,119,0,0,0,1,9,18,95,95,114,101,116,86,97, 108,0,59,120,0,18,120,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95, 114,101,116,86,97,108,0,59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119, -0,20,0,0,1,0,8,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86, -97,108,0,0,18,105,0,0,0,0,1,0,8,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99, -52,0,18,95,95,114,101,116,86,97,108,0,0,18,102,0,0,0,0,1,0,8,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52, -95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,2,1,1,1,0,1, -98,49,0,0,1,1,0,1,98,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,0,2,1,1,1,0,1,98,0,0,0,1,4,118,101,99, -52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,2,1,1,1, -0,9,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, -18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,10,118,0,0, -0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17, -48,0,48,0,0,0,0,0,1,0,2,1,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,3,1,1,1,0,1,98,49,0,0,1,1,0,1, -98,50,0,0,1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0, -18,98,51,0,20,0,0,1,0,3,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,3,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0, -1,0,3,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,3,1,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0, -1,0,3,1,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,4,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0, -1,98,51,0,0,1,1,0,1,98,52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0, -18,98,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,4,1,1,1,0,9,102, -49,0,0,1,1,0,9,102,50,0,0,1,1,0,9,102,51,0,0,1,1,0,9,102,52,0,0,0,1,3,2,1,9,1,122,101,114,111,0,2, -17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -102,49,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, +0,20,0,0,1,0,0,12,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116, +86,97,108,0,0,18,102,0,0,0,0,1,0,0,12,1,1,1,0,5,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118, +101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,0,0,12,1,1,1,0,1,98,0,0,0,1,4,105, +118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,0, +12,1,1,1,0,4,98,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101,99,52,0,18,95,95,114,101,116,86, +97,108,0,0,18,98,0,0,0,0,1,0,0,12,1,1,1,0,8,105,0,0,0,1,4,105,118,101,99,52,95,116,111,95,118,101, +99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0,0,1,0,0,12,1,1,1,0,11,118,51,0,0,1,1,0,9, +102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,118,51,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,119,0,18,102,0,20,0,0,1,0,0,12,1,1,1,0,10,118,50,0,0,1,1,0,9,102,49,0,0,1,1, +0,9,102,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,118,50,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,122,0,18,102,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,102, +50,0,20,0,0,1,0,0,6,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,0,1,0,0,6,1,1,1,0,5, +105,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,105,0,0,0,0,1,0,0,6,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,0,0,1,0,0,6,1,1,1,0,1,98,0,0,0,1,4,118,101, +99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0, +0,0,1,0,0,7,1,1,1,0,5,105,0,0,1,1,0,5,106,0,0,1,1,0,5,107,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,59,120,0,18,105,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,106,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,122,0,18,107,0,20,0,0,1,0,0,7,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,109, +111,118,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,105,0,0,0,0,1,0,0,7,1,1,1,0,9, +102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,102,0,0,0,0,1,0,0,7,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,0,8,1,1,1,0,5,120,0,0,1,1,0, +5,121,0,0,1,1,0,5,122,0,0,1,1,0,5,119,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,120,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,122,0,18,122,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,119,0,20,0,0,1,0,0,8,1,1,1,0, +5,105,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,0,18,105,0,0,0, +0,1,0,0,8,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101, +116,86,97,108,0,0,18,102,0,0,0,0,1,0,0,8,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,116,111,95,105, +118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,0,2,1,1,1,0,1,98,49,0,0,1,1,0, +1,98,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,0,0,2,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,2,1,1,1,0,9,102,0,0,0,1, +4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,17,48, +0,48,0,0,0,0,0,1,0,0,2,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,0,2,1,1,1,0,10,118,0,0,0,1,4,118, +101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0, +0,0,0,0,1,0,0,2,1,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0, +1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51, +0,20,0,0,1,0,0,3,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,0,3,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3, +1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3,1,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3, +1,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,4,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98, +51,0,0,1,1,0,1,98,52,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18, +98,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,18,98,52,0,20,0,0,1,0,0,4,1,1,1,0,9,102, +49,0,0,1,1,0,9,102,50,0,0,1,1,0,9,102,51,0,0,1,1,0,9,102,52,0,0,0,1,3,2,1,0,9,1,122,101,114,111,0, +2,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, +18,102,49,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, 108,0,59,121,0,0,18,102,50,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, 114,101,116,86,97,108,0,59,122,0,0,18,102,51,0,0,18,122,101,114,111,0,0,0,4,118,101,99,52,95,115, 110,101,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,102,52,0,0,18,122,101,114,111,0,0,0,0,1,0, -4,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,119,0,0,18,98,0,0,0,0,1,0,4,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,4,1,1,1,0, -5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119, -0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,4,1,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,4,1,1, -1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, -119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0, -9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0, -18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,0,13,1,1,1,0,9,102,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0, -17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,102,0,20,0,0,1,0, -13,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,13, -1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,13,1,1,1, -0,10,99,48,0,0,1,1,0,10,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,0,1,0,14,1,1,1,0,9,109,48,48, -0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0, -9,109,50,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, -0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0, -18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0, -0,1,0,14,1,1,1,0,9,102,0,0,0,1,3,2,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0, -0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,118,0,59,121,121,120,0,20,0,0,1,0,14,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,51,0, -58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,14,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,51,0,58, -102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,14,1,1,1,0,11,99,48,0,0,1,1,0,11,99,49,0,0,1,1,0,11, -99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, -99,50,0,20,0,0,1,0,15,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9, -109,51,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,51,49, -0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9,109,50,50,0,0,1,1,0,9,109,51,50,0,0,1,1,0, -9,109,48,51,0,0,1,1,0,9,109,49,51,0,0,1,1,0,9,109,50,51,0,0,1,1,0,9,109,51,51,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,122, -0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,119,0,18,109,51,48,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,119,0, -18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,120,0, -18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,121,0,18,109,49,51,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0,18,109,50,51,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0,0,1,0,15,1,1,1,0,9,102,0,0,0,1,3,2,0,10, -1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57, -18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59, -121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,118,0,59,121,121,121, -120,0,20,0,0,1,0,15,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,105,0,0, -0,0,0,0,0,1,0,15,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,52,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0, -0,0,1,0,15,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0,12,99,50,0,0,1,1,0,12,99,51,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20,0,0,1,0,5,2,26,1,1,0,5,97,0,0,1,1,0,5,98, -0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0, -1,0,5,2,27,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0, -1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0, -0,18,98,0,0,0,0,1,0,5,2,22,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,1,1,120,0, -0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116, -111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,6,2,26,1,1,0,6,97, -0,0,1,1,0,6,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0, -0,18,98,0,0,0,0,1,0,6,2,27,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114, -97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,6,2,21,1,1,0,6,97,0,0,1, -1,0,6,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, -108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,6,2,22,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,10,1,98,73,110, -118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0, -59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0, -0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0, -4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, -0,1,0,7,2,26,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101, -116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,7,2,27,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,4,118,101, -99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0, -1,0,7,2,21,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, -18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,7,2,22,1,1,0,7,97,0,0,1,1,0,7,98,0, -0,0,1,3,2,0,11,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110, +0,4,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,119,0,0,18,98,0,0,0,0,1,0,0,4,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110,101,0, +18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,0,4,1, +1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,119,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,0,4,1,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0, +1,0,0,4,1,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,119,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,13,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109, +49,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +48,0,57,59,120,0,18,109,48,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18, +109,49,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,120,0,18,109,48,49,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0,20,0,0,1,0,0,13,1,1,1,0,9, +102,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,102,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,59,121,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,49,0,57,59,120,0,17,48,0,48,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121, +0,18,102,0,20,0,0,1,0,0,13,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18, +105,0,0,0,0,0,0,0,1,0,0,13,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,50,0,58,102,108,111,97,116,0,18,98, +0,0,0,0,0,0,0,1,0,0,13,1,1,1,0,10,99,48,0,0,1,1,0,10,99,49,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20, +0,0,1,0,0,14,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109,49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,48,49, +0,0,1,1,0,9,109,49,49,0,0,1,1,0,9,109,50,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0, +9,109,50,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48,48,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49, +0,57,59,120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109, +49,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,59,121,0,18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +59,122,0,18,109,50,50,0,20,0,0,1,0,0,14,1,1,1,0,9,102,0,0,0,1,3,2,0,0,10,1,118,0,2,58,118,101,99, +50,0,18,102,0,0,17,48,0,48,0,0,0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59, +120,121,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,118,0,59,121,121,120,0,20,0,0,1,0,0,14,1,1,1,0, +5,105,0,0,0,1,8,58,109,97,116,51,0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,0,14,1,1,1,0,1, +98,0,0,0,1,8,58,109,97,116,51,0,58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,0,14,1,1,1,0,11,99, +48,0,0,1,1,0,11,99,49,0,0,1,1,0,11,99,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +18,99,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,50,0,57,18,99,50,0,20,0,0,1,0,0,15,1,1,1,0,9,109,48,48,0,0,1,1,0,9,109, +49,48,0,0,1,1,0,9,109,50,48,0,0,1,1,0,9,109,51,48,0,0,1,1,0,9,109,48,49,0,0,1,1,0,9,109,49,49,0,0, +1,1,0,9,109,50,49,0,0,1,1,0,9,109,51,49,0,0,1,1,0,9,109,48,50,0,0,1,1,0,9,109,49,50,0,0,1,1,0,9, +109,50,50,0,0,1,1,0,9,109,51,50,0,0,1,1,0,9,109,48,51,0,0,1,1,0,9,109,49,51,0,0,1,1,0,9,109,50,51, +0,0,1,1,0,9,109,51,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,120,0,18,109,48, +48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,59,121,0,18,109,49,48,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,59,122,0,18,109,50,48,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,59,119,0,18,109,51,48,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59, +120,0,18,109,48,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,121,0,18,109,49,49,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,59,122,0,18,109,50,49,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,59,119,0,18,109,51,49,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,50,0,57,59,120,0,18,109,48,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,121, +0,18,109,49,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,122,0,18,109,50,50,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,59,119,0,18,109,51,50,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,51,0,57,59,120,0,18,109,48,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,51,0,57,59,121,0,18,109,49,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,122,0, +18,109,50,51,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,59,119,0,18,109,51,51,0,20,0, +0,1,0,0,15,1,1,1,0,9,102,0,0,0,1,3,2,0,0,10,1,118,0,2,58,118,101,99,50,0,18,102,0,0,17,48,0,48,0,0, +0,0,0,0,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,118,0,59,120,121,121,121,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,118,0,59,121,120,121,121,0,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,118,0,59,121,121,120,121,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,51,0,57,18,118,0,59,121,121,121,120,0,20,0,0,1,0,0,15,1,1,1,0,5,105,0,0,0,1,8,58,109,97,116,52, +0,58,102,108,111,97,116,0,18,105,0,0,0,0,0,0,0,1,0,0,15,1,1,1,0,1,98,0,0,0,1,8,58,109,97,116,52,0, +58,102,108,111,97,116,0,18,98,0,0,0,0,0,0,0,1,0,0,15,1,1,1,0,12,99,48,0,0,1,1,0,12,99,49,0,0,1,1,0, +12,99,50,0,0,1,1,0,12,99,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,99,48,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,99,49,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,99,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,99,51,0,20, +0,0,1,0,0,5,2,26,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114, +101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,5,2,27,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118, +101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0, +0,0,1,0,0,5,2,21,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,5,2,22,1,1,0,5,97,0,0,1,1,0, +5,98,0,0,0,1,3,2,0,0,9,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +98,73,110,118,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18, +97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101, +116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,6,2,26,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,4,118,101,99,52,95, +97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,6,2,27,1,1,0,6,97,0, +0,1,1,0,6,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97, +108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,6,2,21,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,4,118,101,99,52,95, +109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0, +6,2,22,1,1,0,6,97,0,0,1,1,0,6,98,0,0,0,1,3,2,0,0,10,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108, +111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95, +105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,7,2,26,1,1,0,7,97,0,0, +1,1,0,7,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18, +98,0,0,0,0,1,0,0,7,2,27,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97, +99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,7,2,21,1,1,0,7,97,0,0,1, +1,0,7,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97, +108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,7,2,22,1,1,0,7,97,0,0,1,1,0,7,98,0,0,0,1,3,2,0,0,11,1,98,73, +110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,120,0,0,18,98, +0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,121,0,0,18,98,0,59,121,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18,98,0,59,122,0,0,0,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110,118,0,0,0,4,118, +101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0, +0,8,2,26,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86, +97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,8,2,27,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,4,118,101,99,52, +95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0, +8,2,21,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,8,2,22,1,1,0,8,97,0,0,1,1,0,8,98,0,0, +0,1,3,2,0,0,12,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110, 118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59, 121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,122,0,0,18, -98,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98, -73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108, -0,0,18,120,0,0,0,0,1,0,8,2,26,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, -95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,8,2,27,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0, +98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,119,0,0,18,98,0,59, +119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18,98,73,110, +118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18, +120,0,0,0,0,1,0,0,9,2,26,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95, +95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,2,27,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0, 1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, -18,98,0,0,0,0,1,0,8,2,21,1,1,0,8,97,0,0,1,1,0,8,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,8,2,22,1,1,0,8,97,0,0, -1,1,0,8,98,0,0,0,1,3,2,0,12,1,98,73,110,118,0,0,1,1,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,98,73,110,118,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73, -110,118,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0, -59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73,110,118,0,59,119,0,0, -18,98,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,120,0,0,18,97,0,0,18, -98,73,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,95,95,114,101,116,86,97, -108,0,0,18,120,0,0,0,0,1,0,9,2,26,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100, -0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,1,1,0,9,98, -0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97, -0,0,18,98,0,0,0,0,1,0,9,2,21,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,2,22,1,1,0,9,97, -0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,98,73,110,118,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,98,73, -110,118,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0,1,1,0,10,117, -0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0, -18,117,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116, -114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2, -21,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0, -0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0, -0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0, +18,98,0,0,0,0,1,0,0,9,2,21,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,2,22,1,1,0,9, +97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,98,73,110,118,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +98,73,110,118,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95, +95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,73,110,118,0,0,0,0,1,0,0,10,2,26,1,1,0,10,118,0,0,1,1, +0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,118,0,0,18,117,0,0,0,0,1,0,0,10,2,27,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95, +115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0, +0,0,0,1,0,0,10,2,21,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,10,2, +22,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112, +0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0, +0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,118,0,0,18,119,0,0,0,0,1,0,0,11,2,26,1,1,0,11,118,0,0,1,1,0,11,117,0, +0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0, +0,18,117,0,0,0,0,1,0,0,11,2,27,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0, +1,0,0,11,2,21,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108, +121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,11,2,22, +1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0, +18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0, +0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +122,0,0,18,118,0,0,18,119,0,0,0,0,1,0,0,12,2,26,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99, +52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,12,2,27,1,1, +0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114, +101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,12,2,21,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0, +1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0, +0,18,117,0,0,0,0,1,0,0,12,2,22,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,0,12,1,119,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95, +114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119, +0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117, +0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, +0,0,18,118,0,0,18,119,0,0,0,0,1,0,0,10,2,26,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95, +97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0, +1,0,0,10,2,26,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,10,2,27,1,1,0,9,97,0,0, +1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,10,2,27,1,1,0,10,118,0,0,1,1,0,9, +98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,10,2,21,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0, +1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,10,2,21,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, +118,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,10,2,22,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,3,2,0,0,10,1, +105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0, +59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0, 0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,118,0,0,18,119,0,0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99, -52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1, -0,11,2,27,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, -18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,21,1,1,0,11, -118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,11,2,22,1,1,0,11,118,0,0,1,1,0,11, -117,0,0,0,1,3,2,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0, -59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18, -119,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18, -95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,12, +121,0,0,18,97,0,0,18,105,110,118,85,0,59,120,121,0,0,0,0,1,0,0,10,2,22,1,1,0,10,118,0,0,1,1,0,9,98, +0,0,0,1,3,2,0,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0, +0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108, +0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,105,110,118,66,0,0,0,0,1,0,0,11,2,26,1,1,0,9,97,0,0,1, +1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, +0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,0,11,2,26,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4, +118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120, +121,122,0,0,18,98,0,0,0,0,1,0,0,11,2,27,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0, +59,120,121,122,0,0,0,0,1,0,0,11,2,27,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121, +122,0,0,18,98,0,0,0,0,1,0,0,11,2,21,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117, +108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0, +59,120,121,122,0,0,0,0,1,0,0,11,2,21,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120, +121,122,0,0,18,98,0,0,0,0,1,0,0,11,2,22,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,3,2,0,0,11,1,105,110, +118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0, +0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, +97,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0,0,1,0,0,11,2,22,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0, +1,3,2,0,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18, +98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,66,0,0,0,0,1,0,0,12,2,26,1,1,0,9,97,0,0, +1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, +18,117,0,0,0,0,1,0,0,12,2,26,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0, +18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1,0,0,12,2,27,1,1,0,9,97,0,0,1,1,0,12, 117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0, -18,118,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,117,0,0,0,0,1,0,12, -2,22,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112, -0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0, -0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0, -0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52, -95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,119,0,0,0,0, -1,0,10,2,26,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0,0,0,0,1,0,10,2,26,1,1,0,10,118,0,0, -1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, -18,118,0,59,120,121,0,0,18,98,0,0,0,0,1,0,10,2,27,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99, -52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18, -117,0,59,120,121,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0, -18,98,0,0,0,0,1,0,10,2,21,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,117,0,59,120,121,0, -0,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,98,0,0,0,0,1, -0,10,2,22,1,1,0,9,97,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,105,110,118,85,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,18,105,110,118,85,0,59, -120,121,0,0,0,0,1,0,10,2,22,1,1,0,10,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0, -18,105,110,118,66,0,0,0,0,1,0,11,2,26,1,1,0,9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97, -100,100,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0, -0,0,0,1,0,11,2,26,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,11,2,27,1,1,0, -9,97,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,27,1,1,0,11, -118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,11,2,21,1,1,0,9,97,0, -0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,117,0,59,120,121,122,0,0,0,0,1,0,11,2,21,1,1,0,11,118, -0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,11,2,22,1,1,0,9,97,0, -0,1,1,0,11,117,0,0,0,1,3,2,0,11,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105, -110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, -85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59, -122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,18,105,110,118,85,0,59,120,121,122,0,0,0,0,1,0,11, -2,22,1,1,0,11,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118, -66,0,0,0,0,1,0,12,2,26,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,95, -95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,0,12,2,26,1,1,0,12,118,0,0,1,1,0,9,98,0,0, -0,1,4,118,101,99,52,95,97,100,100,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1, -0,12,2,27,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9,98,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118, -0,0,18,98,0,0,0,0,1,0,12,2,21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,0,12,2,21,1,1, -0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114, -101,116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1,0,12,2,22,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,3,2, -0,12,1,105,110,118,85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18, -117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59, -121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118, -101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,105, -110,118,85,0,0,0,0,1,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,0,0,0, -1,0,6,2,26,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99, -50,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,6,2,27,1,1,0,5,97,0, -0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117, -0,47,20,0,0,1,0,6,2,27,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,58,105,118,101,99,50,0,18,98,0,0,0,47,20,0,0,1,0,6,2,21,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,6,2,21, -1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99, -50,0,18,98,0,0,0,48,20,0,0,1,0,6,2,22,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,6,2,22,1,1,0,6,118,0,0,1,1,0,5, +18,97,0,0,18,117,0,0,0,0,1,0,0,12,2,27,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115, +117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,98,0,0,0,0,1,0,0,12,2, +21,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95, +95,114,101,116,86,97,108,0,0,18,97,0,0,18,117,0,0,0,0,1,0,0,12,2,21,1,1,0,12,118,0,0,1,1,0,9,98,0, +0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18, +118,0,0,18,98,0,0,0,0,1,0,0,12,2,22,1,1,0,9,97,0,0,1,1,0,12,117,0,0,0,1,3,2,0,0,12,1,105,110,118, +85,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,120,0,0,18,117,0,59,120,0,0,0, +4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,105,110,118,85,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111, +97,116,95,114,99,112,0,18,105,110,118,85,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,105,110,118,85,0,0,0, +0,1,0,0,12,2,22,1,1,0,12,118,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,105,110,118,66,0,0,0,4,102,108,111, +97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,105,110,118,66,0,0,0,0,1,0,0,6,2,26, +1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97, +0,0,0,18,117,0,46,20,0,0,1,0,0,6,2,26,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,46,20,0,0,1,0,0,6,2,27,1,1,0,5,97,0,0,1,1,0,6, +117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,47,20,0, +0,1,0,0,6,2,27,1,1,0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58, +105,118,101,99,50,0,18,98,0,0,0,47,20,0,0,1,0,0,6,2,21,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,0,6,2,21,1,1, +0,6,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0, +18,98,0,0,0,48,20,0,0,1,0,0,6,2,22,1,1,0,5,97,0,0,1,1,0,6,117,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,58,105,118,101,99,50,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,0,6,2,22,1,1,0,6,118,0,0,1,1,0,5, 98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,50,0,18,98,0,0,0,49,20,0, -0,1,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101, -99,51,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,46,20,0,0,1,0,7,2,27,1,1,0,5, -97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0, -18,117,0,47,20,0,0,1,0,7,2,27,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,47,20,0,0,1,0,7,2,21,1,1,0,5,97,0,0,1,1,0,7,117,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,7, -2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, -99,51,0,18,98,0,0,0,48,20,0,0,1,0,7,2,22,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,7,2,22,1,1,0,7,118,0,0,1,1,0, -5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,49,20, -0,0,1,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, -101,99,52,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,46,20,0,0,1,0,8,2,27,1,1,0, -5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0, -18,117,0,47,20,0,0,1,0,8,2,27,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,47,20,0,0,1,0,8,2,21,1,1,0,5,97,0,0,1,1,0,8,117,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,48,20,0,0,1,0,8, -2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101, -99,52,0,18,98,0,0,0,48,20,0,0,1,0,8,2,22,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,8,2,22,1,1,0,8,118,0,0,1,1,0, -5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,49,20, -0,0,1,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,6,2,27,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,110,101,103, -97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,7,2,27,1,1,0,7,118,0,0,0,1,4, -118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,8, -2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97, -108,0,0,18,118,0,0,0,0,1,0,9,2,27,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0, -18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,10,2,27,1,1,0,10,118,0,0,0,1,4,118, -101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59, -120,121,0,0,0,0,1,0,11,2,27,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,0,0,1,0,12,2,27,1,1,0,12, -118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118, -0,0,0,0,1,0,13,2,27,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20, -0,0,1,0,14,2,27,1,1,0,14,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8, -48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,0,1,0,15,2,27,1,1,0, -15,109,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -51,0,57,18,109,0,16,10,51,0,57,54,20,0,0,1,0,9,0,100,111,116,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,48,20,0,0,1,0,9,0,100,111,116,0,1,1,0,10,97,0,0,1, -1,0,10,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0, -59,121,0,18,98,0,59,121,0,48,46,20,0,0,1,0,9,0,100,111,116,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4, -118,101,99,51,95,100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,9,0, -100,111,116,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,18,95,95,114, -101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,1,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,2,1,0,2,5,97,0,0,1,1,0,5,98,0, -0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2, -3,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0, -18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,9,1,105,110,118,66,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,105,110,118,66,0,0,0,4,118,101,99,52,95,116,111, -95,105,118,101,99,52,0,18,97,0,0,18,97,0,0,0,0,1,0,0,2,1,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,6, +0,1,0,0,7,2,26,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, +101,99,51,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,0,7,2,26,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,46,20,0,0,1,0,0,7,2,27,1,1, +0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0, +0,18,117,0,47,20,0,0,1,0,0,7,2,27,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,118,0,58,105,118,101,99,51,0,18,98,0,0,0,47,20,0,0,1,0,0,7,2,21,1,1,0,5,97,0,0,1,1,0,7, +117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,48,20,0, +0,1,0,0,7,2,21,1,1,0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58, +105,118,101,99,51,0,18,98,0,0,0,48,20,0,0,1,0,0,7,2,22,1,1,0,5,97,0,0,1,1,0,7,117,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,58,105,118,101,99,51,0,18,97,0,0,0,18,117,0,49,20,0,0,1,0,0,7,2,22,1,1, +0,7,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,51,0, +18,98,0,0,0,49,20,0,0,1,0,0,8,2,26,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,46,20,0,0,1,0,0,8,2,26,1,1,0,8,118,0,0,1,1,0,5, +98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,46,20,0, +0,1,0,0,8,2,27,1,1,0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118, +101,99,52,0,18,97,0,0,0,18,117,0,47,20,0,0,1,0,0,8,2,27,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,47,20,0,0,1,0,0,8,2,21,1,1, +0,5,97,0,0,1,1,0,8,117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0, +0,18,117,0,48,20,0,0,1,0,0,8,2,21,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,118,0,58,105,118,101,99,52,0,18,98,0,0,0,48,20,0,0,1,0,0,8,2,22,1,1,0,5,97,0,0,1,1,0,8, +117,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,58,105,118,101,99,52,0,18,97,0,0,0,18,117,0,49,20,0, +0,1,0,0,8,2,22,1,1,0,8,118,0,0,1,1,0,5,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,58, +105,118,101,99,52,0,18,98,0,0,0,49,20,0,0,1,0,0,5,2,27,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,110, +101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,0,0,0,1,0,0,6,2,27,1,1,0, +6,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18, +118,0,0,0,0,1,0,0,7,2,27,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95, +114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,0,8,2,27,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,110, +101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,0,9,2,27,1,1,0,9,97,0,0, +0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97, +0,0,0,0,1,0,0,10,2,27,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,0,11,2,27,1,1,0,11,118,0,0, +0,1,4,118,101,99,52,95,110,101,103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,118,0,59,120,121,122,0,0,0,0,1,0,0,12,2,27,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,110,101, +103,97,116,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,0,0,1,0,0,13,2,27,1,1,0,13,109,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,0,1,0,0,14,2,27,1,1,0,14,109,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,50,0,57,18,109,0,16,10,50,0,57,54,20,0,0,1,0,0,15,2,27,1,1,0,15,109,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,49,0,57,18,109,0,16,10,49,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, +109,0,16,10,50,0,57,54,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0, +57,54,20,0,0,1,0,0,9,0,100,111,116,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,18,97,0,18,98,0,48,20,0,0,1,0,0,9,0,100,111,116,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,18,97,0,59,120,0,18,98,0,59,120,0,48,18,97,0,59,121,0,18,98,0,59, +121,0,48,46,20,0,0,1,0,0,9,0,100,111,116,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,51,95, +100,111,116,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,0,100,111,116,0,1, +1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,100,111,116,0,18,95,95,114,101,116,86,97,108, +0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,0,2,1,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,97, +100,100,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,0,2,2,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,4,118, +101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,0,2,3,1,0,2,5, +97,0,0,1,1,0,5,98,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,0,18,97,0,0, +18,98,0,0,0,0,1,0,0,0,2,4,1,0,2,5,97,0,0,1,1,0,5,98,0,0,0,1,3,2,0,0,9,1,105,110,118,66,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,105,110,118,66,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,97,0,0,18,97,0,0,18,105,110,118,66,0,0,0,4,118,101,99,52,95,116,111,95, +105,118,101,99,52,0,18,97,0,0,18,97,0,0,0,0,1,0,0,0,2,1,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118, +101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,2,1,0,2,6,118,0,0,1,1,0,6, 117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0, -1,0,0,2,3,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, -18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,3,2,0,6,1,105, -110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117, -0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0, -0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0, -0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,7, -118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1, -0,0,2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, -118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95, -109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1, -1,0,7,117,0,0,0,1,3,2,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118, -0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0, -18,118,0,0,18,105,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,118,0,0,18, -122,0,0,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0, -0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,8, -117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0, -0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,3,2,0,8,1,105,110,118,0,0,1,1,122,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,118,101,99,52,95,116,111,95, -105,118,101,99,52,0,18,118,0,0,18,122,0,0,0,0,1,0,0,2,1,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118, -101,99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0,59,120,0,0,0,0,1,0,0,2,2,1, -0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,97,0,59,120,0, -0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,117, -108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,2,4,1,0,2,9,97,0,0,1,1, -0,9,98,0,0,0,1,3,2,0,9,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,98, -0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,119,0,0, -0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59, -120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0, -10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0, -59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118, +1,0,0,0,2,3,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,4,1,0,2,6,118,0,0,1,1,0,6,117,0,0,0,1,3,2,0,0,6,1, +105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18, +117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59, +121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110, +118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,118,0,0,18,122,0,0,0,0,1,0,0,0,2,1, +1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0, +0,0,0,1,0,0,0,2,2,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99, +116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,3,1,0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,4,1, +0,2,7,118,0,0,1,1,0,7,117,0,0,0,1,3,2,0,0,7,1,105,110,118,0,0,1,1,122,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4,118,101,99,52,95,116,111,95,105,118,101,99, +52,0,18,118,0,0,18,122,0,0,0,0,1,0,0,0,2,1,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95, +97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,2,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1, +4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,3, +1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0, +0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,4,1,0,2,8,118,0,0,1,1,0,8,117,0,0,0,1,3,2,0,0,8,1,105,110, +118,0,0,1,1,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,120,0,0,18,117,0,59, +120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,0,59,121,0,0,18,117,0,59,121,0,0,0,4, +118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,122,0,0,18,118,0,0,18,105,110,118,0,0,0,4, +118,101,99,52,95,116,111,95,105,118,101,99,52,0,18,118,0,0,18,122,0,0,0,0,1,0,0,0,2,1,1,0,2,9,97,0, +0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,97,0,59,120,0,0,18,97,0,59,120,0,0,18,98,0, +59,120,0,0,0,0,1,0,0,0,2,2,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114, +97,99,116,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,0,2,3,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0, +1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0,59,120,0,0,18,97,0,0,18,98,0,0,0,0, +1,0,0,0,2,4,1,0,2,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,119,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,119,0,59,120,0,0,18,98,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,97,0, +59,120,0,0,18,97,0,0,18,119,0,0,0,0,1,0,0,0,2,1,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99, +52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0, +0,0,2,2,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18, +118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,0,2,3,1,0,2,10,118, +0,0,1,1,0,10,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121, +0,0,18,118,0,59,120,121,0,0,18,117,0,59,120,121,0,0,0,0,1,0,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117, +0,0,0,1,3,2,0,0,10,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0, +59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118, 101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18, -117,0,59,120,121,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,0,10,1,119,0,0,0,4, -102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,119,0,59,120,121,0,0,0,0,1,0,0,2, -1,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0, -0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,115, -117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2, -11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59, -120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,4,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,11, -1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95, -114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,119,0,59,120,121,122,0,0,0,0, -1,0,0,2,1,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118, -0,0,18,117,0,0,0,0,1,0,0,2,2,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98, -116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,12,117,0, -0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1, -0,0,2,4,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99, -112,0,18,119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59, -121,0,0,18,117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59, -122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101, -99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,119,0,0,0,0,1,0,0,2,1,1,0,2,6, -118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120, -121,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98, -116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2, -6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120, -121,0,0,18,118,0,59,120,121,0,0,18,97,0,0,0,9,18,118,0,59,120,0,18,97,0,23,0,9,18,118,0,59,121,0, -18,97,0,23,0,0,1,0,0,2,4,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18, -118,0,59,121,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97, -100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0,2,7, -118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121, -122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4, -118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121, -122,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24, -0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,8,118,0,0,1, -1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0, -2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118, -0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1, -9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0, -9,18,118,0,59,119,0,18,97,0,24,0,0,1,0,0,2,1,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52, -95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0,2,10,118,0,0,1,1,0, -9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0, -18,97,0,0,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, -112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2,10,118,0,0,1,1,0,9, -97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0, -0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118, -0,59,120,121,0,0,18,105,110,118,65,0,0,0,0,1,0,0,2,1,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101, -99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0,2,11,118, -0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0, -0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2, -11,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +119,0,59,120,121,0,0,0,0,1,0,0,0,2,1,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,97, +100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,2,1,0,2,11,118,0,0,1,1,0, +11,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18, +118,0,0,18,117,0,0,0,0,1,0,0,0,2,3,1,0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,4,1, +0,2,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,0,0,11,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +119,0,59,120,0,0,18,117,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18, +117,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0, +4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120, +121,122,0,0,18,119,0,59,120,121,122,0,0,0,0,1,0,0,0,2,1,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4, +118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,2,1,0,2,12,118,0,0,1, +1,0,12,117,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,117, +0,0,0,0,1,0,0,0,2,3,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,118,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,0,2,4,1,0,2,12,118,0,0,1,1,0,12,117,0,0,0, +1,3,2,0,0,12,1,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,120,0,0,18,117,0,59,120, +0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,119,0,59,121,0,0,18,117,0,59,121,0,0,0,4,102,108,111, +97,116,95,114,99,112,0,18,119,0,59,122,0,0,18,117,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,119,0,59,119,0,0,18,117,0,59,119,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121, +0,18,118,0,0,18,118,0,0,18,119,0,0,0,0,1,0,0,0,2,1,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99, +52,95,97,100,100,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,0,2,2,1,0,2, +6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121, +0,0,18,118,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,0,2,3,1,0,2,6,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101, +99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0,18,97,0, +0,0,9,18,118,0,59,120,0,18,97,0,23,0,9,18,118,0,59,121,0,18,97,0,23,0,0,1,0,0,0,2,4,1,0,2,6,118,0, +0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,0,1,0,0,0,2, +1,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,59,120,121,122,0,0, +18,118,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,0,2,2,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101, +99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18, +97,0,0,0,0,1,0,0,0,2,3,1,0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105, +112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,0,2,4,1, +0,2,7,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97,0,24,0,9,18,118,0,59,121,0,18,97,0,24,0, +9,18,118,0,59,122,0,18,97,0,24,0,0,1,0,0,0,2,1,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52, +95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,2,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1, +4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,3, +1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0, +18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,4,1,0,2,8,118,0,0,1,1,0,5,97,0,0,0,1,9,18,118,0,59,120,0,18,97, +0,24,0,9,18,118,0,59,121,0,18,97,0,24,0,9,18,118,0,59,122,0,18,97,0,24,0,9,18,118,0,59,119,0,18,97, +0,24,0,0,1,0,0,0,2,1,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0, +59,120,121,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,2,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101, +99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,0,2, +3,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118, +0,59,120,121,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,4,1,0,2,10,118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,0, +9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18,97,0,0,0,4, +118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59,120,121,0,0,18,118,0,59,120,121,0,0, +18,105,110,118,65,0,0,0,0,1,0,0,0,2,1,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97, +100,100,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,2,1,0,2,11,118,0,0,1,1,0, +9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,59,120,121,122,0,0,18,118, +0,0,18,97,0,0,0,0,1,0,0,0,2,3,1,0,2,11,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,118,0,59,120,121,122,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,4,1,0,2,11, +118,0,0,1,1,0,9,97,0,0,0,1,3,2,0,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0, 18,105,110,118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,59, -120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,65,0,0,0,0,1,0,0,2,1,1,0,2,12,118,0,0,1, -1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,2,1,0, -2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0,18, -118,0,0,18,97,0,0,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109,117,108, -116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,2,4,1,0,2,12,118,0,0,1,1,0,9,97,0, -0,0,1,3,2,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110,118,65,0,0,18, -97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,105,110,118, -65,0,0,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16, -8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0, -16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,27,1,1,0,13,109,0,0, -1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110, -0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,47,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,0,48,18,109,0, -16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,0,48,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49, -20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,27,1,1,0,14,109,0, +120,121,122,0,0,18,118,0,59,120,121,122,0,0,18,105,110,118,65,0,0,0,0,1,0,0,0,2,1,1,0,2,12,118,0,0, +1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,100,100,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,2, +1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,118,0,0, +18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,3,1,0,2,12,118,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0,18,97,0,0,0,0,1,0,0,0,2,4,1,0,2,12,118,0,0,1,1, +0,9,97,0,0,0,1,3,2,0,0,9,1,105,110,118,65,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,105,110, +118,65,0,0,18,97,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,118,0,0,18,118,0,0, +18,105,110,118,65,0,0,0,0,1,0,0,13,2,26,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,0,1,0,0,13, +2,27,1,1,0,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, +0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, +109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,0,1,0,0,13,2,21,1,1,0,13,109,0,0,1,1,0,13,110,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +59,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,0,48,46,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,0,48, +18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,0,48,46,20,0,0,1,0,0,13,2,22,1,1,0,13,109, +0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, +18,110,0,16,10,49,0,57,49,20,0,0,1,0,0,14,2,26,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20, +0,0,1,0,0,14,2,27,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, +0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0, +16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,0,14,2,21,1,1,0,14,109,0, 0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50, -0,57,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95, -114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,0, -48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18, -110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16, -10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122,0, -48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10, -50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,0,48,46, -18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0,48,46,20,0,0,1,0,14,2,22,1,1,0,14, -109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10, -50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,15,2,26,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,46,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0, -57,46,20,0,0,1,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108, -0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,47,20,0,0,1,0,15,2,21,1,1, +110,0,16,8,48,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121, +0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,0,48, +18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110, +0,16,10,49,0,57,59,122,122,122,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, +109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16, +10,50,0,57,59,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,0, +48,46,20,0,0,1,0,0,14,2,22,1,1,0,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, +0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,0,1,0,0,15,2,26,1, +1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8, +48,0,57,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16, +10,49,0,57,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109, +0,16,10,50,0,57,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18, +109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,46,20,0,0,1,0,0,15,2,27,1,1,0,15,109,0,0,1,1,0,15,110,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, +47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49, +0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16, +10,50,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110, +0,16,10,51,0,57,47,20,0,0,1,0,0,15,2,21,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,59,120,120,120,120,0,48, +18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18, +110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119, +119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18, +110,0,16,10,49,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121, +121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,49,0,57,59,122,122,122,122,0,48,46,18, +109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,50,0,57,59,120,120,120,120,0,48,18, +109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18, +110,0,16,10,50,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,50,0,57,59, +119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,8,48,0, +57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,51,0,57,59, +121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,51,0,57,59,122,122,122,122,0,48,46, +18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59,119,119,119,119,0,48,46,20,0,0,1,0,0,15,2,22,1,1, 0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,18,110,0,16,8,48,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,8,48,0,57,59, -121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,8,48,0,57,59,122,122,122,122,0,48,46,18, -109,0,16,10,51,0,57,18,110,0,16,8,48,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116, -86,97,108,0,16,10,49,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,49,0,57,59,120,120,120,120,0,48,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18, -110,0,16,10,49,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,49,0,57,59, -119,119,119,119,0,48,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,8,48,0, -57,18,110,0,16,10,50,0,57,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,110,0,16,10,50,0,57,59, -121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,59,122,122,122,122,0,48,46, -18,109,0,16,10,51,0,57,18,110,0,16,10,50,0,57,59,119,119,119,119,0,48,46,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,51,0,57,18,109,0,16,8,48,0,57,18,110,0,16,10,51,0,57,59,120,120,120,120,0,48, -18,109,0,16,10,49,0,57,18,110,0,16,10,51,0,57,59,121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18, -110,0,16,10,51,0,57,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,59, -119,119,119,119,0,48,46,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49, -20,0,0,1,0,13,2,26,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, -0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,13,2,26,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,0,1,0,13,2,27,1,1,0,9,97,0,0,1,1,0, -13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20, -0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,13, -2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0, -16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0, -57,18,98,0,47,20,0,0,1,0,13,2,21,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, -49,0,57,18,97,0,18,110,0,16,10,49,0,57,48,20,0,0,1,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114, -101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,13,2,22,1,1,0,9,97, -0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48, -0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20, -0,0,1,0,13,2,22,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, -57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0, -16,10,49,0,57,18,98,0,49,20,0,0,1,0,14,2,26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,0,1,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,0,14,2,27,1,1,0,9,97,0,0, -1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57, -47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,14,2, -27,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47, -20,0,0,1,0,14,2,21,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48, -0,57,18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97, -0,18,110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0, -16,10,50,0,57,48,20,0,0,1,0,14,2,21,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,18,98,0,48,20,0,0,1,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101, -116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97, -108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,49,20,0,0,1,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9, -98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9, -18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95, -114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,15,2,26,1,1,0,9, -97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8, -48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18, -95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,15,2,26,1, -1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, +0,57,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0, +16,10,50,0,57,18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18, +109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,49,20,0,0,1,0,0,13,2,26,1,1,0,9,97,0,0,1,1,0,13,110,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,0,1,0,0,13,2,26,1, +1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48, 0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0, -9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,15,2, -27,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, -110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, -49,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47, -20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0, -15,2,27,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109, -0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98, -0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0, -0,1,0,15,2,21,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, -18,97,0,18,110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18, -110,0,16,10,49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16, -10,50,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57, -48,20,0,0,1,0,15,2,21,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, +0,46,20,0,0,1,0,0,13,2,27,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, +57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,0,1,0,0,13,2,27,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,0,1,0,0,13,2,21,1,1,0,9, +97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8, +48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,48, +20,0,0,1,0,0,13,2,21,1,1,0,13,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8, 48,0,57,18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18, -109,0,16,10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16, -10,50,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57, -18,98,0,48,20,0,0,1,0,15,2,22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108, -0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, -57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0, -18,110,0,16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0, -16,10,51,0,57,49,20,0,0,1,0,15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16, -10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0, -16,10,51,0,57,18,98,0,49,20,0,0,1,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0, -59,121,121,0,48,46,20,0,0,1,0,10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, -116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,11,2,21, -1,1,0,14,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57,18, -118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10,50, -0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, -57,0,0,20,0,0,1,0,12,2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121, -121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10,51,0,57, -18,118,0,59,119,119,119,119,0,48,46,20,0,0,1,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,50,0, -57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10, -51,0,57,0,0,20,0,0,1,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110, -0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,2,2,1,0,2,13,109, -0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0, -57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,18,109, -0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0,2,1,1,0,2,14, -109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10, -49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,21,0,0,1,0,0,2, -2,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22, -0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,2, -4,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18, -109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24, -0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57, -21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10, -50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1, -0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0,16,10,49,0,57,18, -110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,22,0,9,18,109,0,16,10,51, -0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,18, -109,0,18,110,0,48,20,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57, -18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10, -50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,24,0,0,1,0,0,2, -1,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, -57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22, -0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0, -16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,13,109,0,0,1,1,0, -9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,0,1,0,0,2, -1,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, -57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0, -0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50, -0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, -23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2, -14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97, -0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,2,1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18, -109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18, -97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9, -18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57, -18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0,1,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1, -9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0, -57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23,0,0,1,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0, -0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50, -0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97,0,24,0,0,1,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109, -0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9, -18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,2,3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18, -118,0,18,109,0,48,20,0,0,1,0,5,2,25,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58, -105,118,101,99,50,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, -7,2,25,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18, -95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0, -58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1, -0,9,2,25,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,18,97,0,20,0,0,1,0,10,2,25,1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49, -0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,25,1,0,2,11,118,0, -0,0,1,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,118,0,20,0,0,1,0,12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0, -17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,25,1,0,2,13, -109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47, -20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0, -9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16, -8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49, -0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57, -18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97, -108,0,18,109,0,20,0,0,1,0,15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, -118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118, -101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,2, -24,1,0,2,5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -97,0,20,0,0,1,0,6,2,24,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18, -118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,0,1,0,8,2,24,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0, -0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18, -97,0,18,97,0,17,49,0,48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,10,2,24, -1,0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95, -114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, -12,2,24,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9, -18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8, -48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0, -57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86, -97,108,0,18,109,0,20,0,0,1,0,14,2,24,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48, -0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58, -118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0, -15,2,24,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17, -49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0, -48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, -0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,5,0,95,95,112,111,115,116,68,101, -99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16, -10,49,0,47,20,0,0,1,0,6,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20, -0,0,1,0,7,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,8,0,95, -95,112,111,115,116,68,101,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0, -20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115, -116,68,101,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0, -18,97,0,17,49,0,48,0,0,47,20,0,0,1,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10,118,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0, -48,0,0,0,0,47,20,0,0,1,0,11,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0, -47,20,0,0,1,0,12,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,0, -1,0,13,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101,116,86,97, -108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0, -0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0, -0,47,20,0,0,1,0,14,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17, -49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0, -48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0, -0,0,0,47,20,0,0,1,0,15,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0, +109,0,16,10,49,0,57,18,98,0,48,20,0,0,1,0,0,13,2,22,1,1,0,9,97,0,0,1,1,0,13,110,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,0,1,0,0,13,2,22,1,1,0,13,109,0,0, +1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0, +49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,0, +1,0,0,14,2,26,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18, +110,0,16,10,49,0,57,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16, +10,50,0,57,46,20,0,0,1,0,0,14,2,26,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16, +10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57, +18,109,0,16,10,50,0,57,18,98,0,46,20,0,0,1,0,0,14,2,27,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,47,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0,9,18,95,95,114,101,116,86,97, +108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,0,1,0,0,14,2,27,1,1,0,14,109,0,0,1,1,0, +9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,47,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,47,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,0,1,0,0,14,2,21,1,1,0, +9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16, +8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57, +48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48,20,0,0, +1,0,0,14,2,21,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57, +18,109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16, +10,49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57, +18,98,0,48,20,0,0,1,0,0,14,2,22,1,1,0,9,97,0,0,1,1,0,14,110,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10, +49,0,57,18,97,0,18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18, +97,0,18,110,0,16,10,50,0,57,49,20,0,0,1,0,0,14,2,22,1,1,0,14,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95, +114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116, +86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108, +0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,49,20,0,0,1,0,0,15,2,26,1,1,0,9,97,0,0,1,1,0,15,110, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0,57,46,20,0,9,18, +95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,46,20,0,9,18,95,95,114, +101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,46,20,0,9,18,95,95,114,101,116,86, +97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,46,20,0,0,1,0,0,15,2,26,1,1,0,15,109,0,0,1,1, +0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,46,20, +0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0,46,20,0,9,18,95, +95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,46,20,0,0,1,0,0,15,2,27,1,1,0,9,97,0, +0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18,110,0,16,8,48,0, +57,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10,49,0,57,47,20,0, +9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,47,20,0,9,18,95,95, +114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,47,20,0,0,1,0,0,15,2,27,1,1,0, +15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,109,0,16,8,48,0, +57,18,98,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10,49,0,57,18,98,0, +47,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18,98,0,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,47,20,0,0,1,0,0,15,2, +21,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18,97,0,18, +110,0,16,8,48,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0,18,110,0,16,10, +49,0,57,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0,16,10,50,0,57,48, +20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0,57,48,20,0,0,1,0, +0,15,2,21,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0,57,18, +109,0,16,8,48,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,109,0,16,10, +49,0,57,18,98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109,0,16,10,50,0,57,18, +98,0,48,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51,0,57,18,98,0,48,20, +0,0,1,0,0,15,2,22,1,1,0,9,97,0,0,1,1,0,15,110,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,16,8,48,0, +57,18,97,0,18,110,0,16,8,48,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0,57,18,97,0, +18,110,0,16,10,49,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,97,0,18,110,0, +16,10,50,0,57,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,97,0,18,110,0,16,10,51,0, +57,49,20,0,0,1,0,0,15,2,22,1,1,0,15,109,0,0,1,1,0,9,98,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +16,8,48,0,57,18,109,0,16,8,48,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,49,0, +57,18,109,0,16,10,49,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,50,0,57,18,109, +0,16,10,50,0,57,18,98,0,49,20,0,9,18,95,95,114,101,116,86,97,108,0,16,10,51,0,57,18,109,0,16,10,51, +0,57,18,98,0,49,20,0,0,1,0,0,10,2,21,1,1,0,13,109,0,0,1,1,0,10,118,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121, +121,0,48,46,20,0,0,1,0,0,10,2,21,1,1,0,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0,20,0,9,18,95,95,114,101, +116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0,57,0,0,20,0,0,1,0,0,11,2, +21,1,1,0,14,109,0,0,1,1,0,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,109,0,16,8,48,0,57, +18,118,0,59,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59,121,121,121,0,48,46,18,109,0,16,10, +50,0,57,18,118,0,59,122,122,122,0,48,46,20,0,0,1,0,0,11,2,21,1,1,0,11,118,0,0,1,1,0,14,109,0,0,0,1, +9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49,0, +57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10, +50,0,57,0,0,20,0,0,1,0,0,12,2,21,1,1,0,15,109,0,0,1,1,0,12,118,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,18,109,0,16,8,48,0,57,18,118,0,59,120,120,120,120,0,48,18,109,0,16,10,49,0,57,18,118,0,59, +121,121,121,121,0,48,46,18,109,0,16,10,50,0,57,18,118,0,59,122,122,122,122,0,48,46,18,109,0,16,10, +51,0,57,18,118,0,59,119,119,119,119,0,48,46,20,0,0,1,0,0,12,2,21,1,1,0,12,118,0,0,1,1,0,15,109,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,100,111,116,0,18,118,0,0,18,109,0,16,8,48,0,57, +0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,100,111,116,0,18,118,0,0,18,109,0,16,10,49, +0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,100,111,116,0,18,118,0,0,18,109,0,16, +10,50,0,57,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,100,111,116,0,18,118,0,0,18,109, +0,16,10,51,0,57,0,0,20,0,0,1,0,0,0,2,1,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0, +57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,0,1,0,0,0,2,2,1, +0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9,18,109,0, +16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,0,1,0,0,0,2,3,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18, +109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,0,2,4,1,0,2,13,109,0,0,1,1,0,13,110,0,0,0,1,9,18,109,0,16, +8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,0,1,0,0, +0,2,1,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,21,0,9, +18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57, +21,0,0,1,0,0,0,2,2,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48, +0,57,22,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0, +16,10,50,0,57,22,0,0,1,0,0,0,2,3,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,18,109,0,18,110, +0,48,20,0,0,1,0,0,0,2,4,1,0,2,14,109,0,0,1,1,0,14,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16, +8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,24,0,9,18,109,0,16,10,50,0,57,18, +110,0,16,10,50,0,57,24,0,0,1,0,0,0,2,1,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0, +57,18,110,0,16,8,48,0,57,21,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,21,0,9,18,109,0,16, +10,50,0,57,18,110,0,16,10,50,0,57,21,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,21,0,0,1,0, +0,0,2,2,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0,1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,22,0,9, +18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,22,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57, +22,0,9,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,22,0,0,1,0,0,0,2,3,1,0,2,15,109,0,0,1,1,0,15, +110,0,0,0,1,9,18,109,0,18,109,0,18,110,0,48,20,0,0,1,0,0,0,2,4,1,0,2,15,109,0,0,1,1,0,15,110,0,0,0, +1,9,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,24,0,9,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0, +57,24,0,9,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,24,0,9,18,109,0,16,10,51,0,57,18,110,0,16, +10,51,0,57,24,0,0,1,0,0,0,2,1,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0, +21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,0,1,0,0,0,2,2,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18, +109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0,57,18,97,0,22,0,0,1,0,0,0,2,3,1,0,2,13,109,0, +0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,0, +1,0,0,0,2,4,1,0,2,13,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109,0,16, +10,49,0,57,18,97,0,24,0,0,1,0,0,0,2,1,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57, +18,97,0,21,0,9,18,109,0,16,10,49,0,57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,0,1,0,0,0, +2,2,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10,49,0, +57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,0,1,0,0,0,2,3,1,0,2,14,109,0,0,1,1,0,9,97,0, +0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16,10,49,0,57,18,97,0,23,0,9,18,109,0,16,10, +50,0,57,18,97,0,23,0,0,1,0,0,0,2,4,1,0,2,14,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18, +97,0,24,0,9,18,109,0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,0,1,0,0,0,2, +1,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,21,0,9,18,109,0,16,10,49,0, +57,18,97,0,21,0,9,18,109,0,16,10,50,0,57,18,97,0,21,0,9,18,109,0,16,10,51,0,57,18,97,0,21,0,0,1,0, +0,0,2,2,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,22,0,9,18,109,0,16,10, +49,0,57,18,97,0,22,0,9,18,109,0,16,10,50,0,57,18,97,0,22,0,9,18,109,0,16,10,51,0,57,18,97,0,22,0,0, +1,0,0,0,2,3,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,23,0,9,18,109,0,16, +10,49,0,57,18,97,0,23,0,9,18,109,0,16,10,50,0,57,18,97,0,23,0,9,18,109,0,16,10,51,0,57,18,97,0,23, +0,0,1,0,0,0,2,4,1,0,2,15,109,0,0,1,1,0,9,97,0,0,0,1,9,18,109,0,16,8,48,0,57,18,97,0,24,0,9,18,109, +0,16,10,49,0,57,18,97,0,24,0,9,18,109,0,16,10,50,0,57,18,97,0,24,0,9,18,109,0,16,10,51,0,57,18,97, +0,24,0,0,1,0,0,0,2,3,1,0,2,10,118,0,0,1,1,0,13,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0, +1,0,0,0,2,3,1,0,2,11,118,0,0,1,1,0,14,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,0,2, +3,1,0,2,12,118,0,0,1,1,0,15,109,0,0,0,1,9,18,118,0,18,118,0,18,109,0,48,20,0,0,1,0,0,5,2,25,1,0,2, +5,97,0,0,0,1,9,18,97,0,18,97,0,16,10,49,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0, +0,1,0,0,6,2,25,1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20, +0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,0,7,2,25,1,0,2,7,118,0,0,0,1,9,18,118,0, +18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0, +20,0,0,1,0,0,8,2,25,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0, +47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,0,9,2,25,1,0,2,9,97,0,0,0,1,9,18,97, +0,18,97,0,17,49,0,48,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,0,10,2,25,1, +0,2,10,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95, +114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,0,11,2,25,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +0,12,2,25,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,0,13,2,25,1,0,2,13,109,0,0,0,1,9,18,109,0,16, +8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49, +0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116, +86,97,108,0,18,109,0,20,0,0,1,0,0,14,2,25,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16, +8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, +0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1, +0,0,15,2,25,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0, 17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17, 49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0, 48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0, -0,0,0,47,20,0,0,1,0,9,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114, -101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,10,0,95,95,112,111,115, -116,73,110,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118, -0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,11,0,95,95,112,111,115,116,73,110, -99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,12,0,95,95,112,111,115,116,73,110,99,114,0,1,0, -2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101, -99,52,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,5,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,5,97,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,6,0, -95,95,112,111,115,116,73,110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, -118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,0,1,0,7,0,95,95,112, -111,115,116,73,110,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9, -18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,8,0,95,95,112,111,115,116,73, -110,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18, -118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,13,0,95,95,112,111,115,116,73,110,99,114, -0,1,0,2,13,109,0,0,0,1,3,2,0,13,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0, -57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57, -58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,14,0,95,95,112,111,115,116,73,110, -99,114,0,1,0,2,14,109,0,0,0,1,3,2,0,14,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, -58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,15,0,95,95,112,111,115,116,73,110, -99,114,0,1,0,2,15,109,0,0,0,1,3,2,0,15,1,110,0,2,18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16, -8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49, -0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57, -58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58, -118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0,1,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0, -0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0, -0,0,0,1,0,1,2,15,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102, -108,111,97,116,0,18,98,0,0,0,40,0,0,1,0,1,2,16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,99,0,0, -0,4,102,108,111,97,116,95,108,101,115,115,0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,1,2, -16,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0, -18,98,0,0,0,41,0,0,1,0,1,2,18,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4, -102,108,111,97,116,95,108,101,115,115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95, -101,113,117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,18,1,1,0, -5,97,0,0,1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0, -0,0,43,0,0,1,0,1,2,17,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108, -111,97,116,95,108,101,115,115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113, -117,97,108,0,18,101,0,0,18,97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,1,2,17,1,1,0,5,97,0,0, -1,1,0,5,98,0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0, -0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114, -105,110,116,0,18,102,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105, -110,116,95,112,114,105,110,116,0,18,105,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1, -98,0,0,0,1,4,98,111,111,108,95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,112,114,105,110,116,77, -69,83,65,0,1,1,0,10,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9, -58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69, -83,65,0,1,1,0,11,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58, -112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, -18,118,0,59,122,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112, -114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1, -0,6,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105, -110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7, +0,0,0,47,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,0,5,2,24,1,0,2,5,97,0,0,0,1,9, +18,97,0,18,97,0,16,10,49,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,0,6,2,24, +1,0,2,6,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,46,20,0,9,18,95,95, +114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,0,7,2,24,1,0,2,7,118,0,0,0,1,9,18,118,0,18,118,0,58, +105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0, +0,8,2,24,1,0,2,8,118,0,0,0,1,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,46,20,0,9, +18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,0,9,2,24,1,0,2,9,97,0,0,0,1,9,18,97,0,18,97,0, +17,49,0,48,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,0,1,0,0,10,2,24,1,0,2,10, +118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,20,0,0,1,0,0,11,2,24,1,0,2,11,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101, +99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,0,12,2, +24,1,0,2,12,118,0,0,0,1,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,0,1,0,0,13,2,24,1,0,2,13,109,0,0,0,1,9,18,109,0,16,8,48,0, +57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18, +109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108, +0,18,109,0,20,0,0,1,0,0,14,2,24,1,0,2,14,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57, +58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58, +118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118, +101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,0,15, +2,24,1,0,2,15,109,0,0,0,1,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0, +48,0,0,0,0,46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0, +0,0,0,46,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0, +0,46,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46, +20,0,9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,0,1,0,0,5,0,95,95,112,111,115,116,68,101,99, +114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10, +49,0,47,20,0,0,1,0,0,6,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95,95,114, +101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0,47,20, +0,0,1,0,0,7,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,47,20,0,0,1,0,0,8, +0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,52,0,16,10,49,0,0,0,47,20,0,0,1,0,0,9,0,95,95,112, +111,115,116,68,101,99,114,0,1,0,2,9,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9, +18,97,0,18,97,0,17,49,0,48,0,0,47,20,0,0,1,0,0,10,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,10, +118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50, +0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,0,11,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,11,118,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0, +48,0,0,0,0,47,20,0,0,1,0,0,12,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0, +47,20,0,0,1,0,0,13,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,13,109,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17, +49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0, +48,0,0,0,0,47,20,0,0,1,0,0,14,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,14,109,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101, +99,51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99, +51,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0, +17,49,0,48,0,0,0,0,47,20,0,0,1,0,0,15,0,95,95,112,111,115,116,68,101,99,114,0,1,0,2,15,109,0,0,0,1, +9,18,95,95,114,101,116,86,97,108,0,18,109,0,20,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58, +118,101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118, +101,99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101, +99,52,0,17,49,0,48,0,0,0,0,47,20,0,9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99, +52,0,17,49,0,48,0,0,0,0,47,20,0,0,1,0,0,9,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,9,97,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0,16,10,49,0,46,20,0,0,1,0,0,10, +0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,10,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18, +118,0,20,0,9,18,118,0,18,118,0,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,0,11,0,95,95, +112,111,115,116,73,110,99,114,0,1,0,2,11,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0, +20,0,9,18,118,0,18,118,0,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,0,12,0,95,95,112,111, +115,116,73,110,99,114,0,1,0,2,12,118,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,118,0,20,0,9,18, +118,0,18,118,0,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,0,1,0,0,5,0,95,95,112,111,115,116,73, +110,99,114,0,1,0,2,5,97,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,97,0,20,0,9,18,97,0,18,97,0, +16,10,49,0,46,20,0,0,1,0,0,6,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,6,118,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,50,0,16,10,49,0,0,0, +46,20,0,0,1,0,0,7,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,7,118,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0, +1,0,0,8,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,8,118,0,0,0,1,9,18,95,95,114,101,116,86,97, +108,0,18,118,0,20,0,9,18,118,0,18,118,0,58,105,118,101,99,51,0,16,10,49,0,0,0,46,20,0,0,1,0,0,13,0, +95,95,112,111,115,116,73,110,99,114,0,1,0,2,13,109,0,0,0,1,3,2,0,0,13,1,110,0,2,18,109,0,0,0,9,18, +109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,9,18,109,0, +16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,50,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110,0,0,0, +1,0,0,14,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,14,109,0,0,0,1,3,2,0,0,14,1,110,0,2,18,109, +0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0, +9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,9,18, +109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,51,0,17,49,0,48,0,0,0,0,46,20,0,8,18,110, +0,0,0,1,0,0,15,0,95,95,112,111,115,116,73,110,99,114,0,1,0,2,15,109,0,0,0,1,3,2,0,0,15,1,110,0,2, +18,109,0,0,0,9,18,109,0,16,8,48,0,57,18,109,0,16,8,48,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0, +46,20,0,9,18,109,0,16,10,49,0,57,18,109,0,16,10,49,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46, +20,0,9,18,109,0,16,10,50,0,57,18,109,0,16,10,50,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0, +9,18,109,0,16,10,51,0,57,18,109,0,16,10,51,0,57,58,118,101,99,52,0,17,49,0,48,0,0,0,0,46,20,0,8,18, +110,0,0,0,1,0,0,1,2,15,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95, +114,101,116,86,97,108,0,59,120,0,0,18,98,0,0,18,97,0,0,0,0,1,0,0,1,2,15,1,1,0,5,97,0,0,1,1,0,5,98, +0,0,0,1,8,58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,40,0,0,1,0,0,1,2, +16,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,1,1,99,0,0,0,4,102,108,111,97,116,95,108,101,115,115, +0,18,99,0,0,18,98,0,0,18,97,0,0,0,8,18,99,0,0,0,1,0,0,1,2,16,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8, +58,102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,41,0,0,1,0,0,1,2,18,1,1,0, +9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115, +115,0,18,103,0,0,18,98,0,0,18,97,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18, +97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,0,1,2,18,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58, +102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,43,0,0,1,0,0,1,2,17,1,1,0,9, +97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,1,1,103,0,0,1,1,101,0,0,0,4,102,108,111,97,116,95,108,101,115, +115,0,18,103,0,0,18,97,0,0,18,98,0,0,0,4,102,108,111,97,116,95,101,113,117,97,108,0,18,101,0,0,18, +97,0,0,18,98,0,0,0,8,18,103,0,18,101,0,32,0,0,1,0,0,1,2,17,1,1,0,5,97,0,0,1,1,0,5,98,0,0,0,1,8,58, +102,108,111,97,116,0,18,97,0,0,0,58,102,108,111,97,116,0,18,98,0,0,0,42,0,0,1,0,0,0,0,112,114,105, +110,116,77,69,83,65,0,1,1,0,9,102,0,0,0,1,4,102,108,111,97,116,95,112,114,105,110,116,0,18,102,0,0, +0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,5,105,0,0,0,1,4,105,110,116,95,112,114,105, +110,116,0,18,105,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,1,98,0,0,0,1,4,98,111, +111,108,95,112,114,105,110,116,0,18,98,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0, +10,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110, +116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,11, 118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110, 116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0, -0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77, -69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0, -9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83, -65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58, +0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,12,118,0,0,0,1,9,58,112,114,105,110,116, +77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0, +0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83, +65,0,18,118,0,59,119,0,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,6,118,0,0,0,1,9, +58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65, +0,18,118,0,59,121,0,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,7,118,0,0,0,1,9,58, 112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0, -18,118,0,59,121,0,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112, -114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,112, -114,105,110,116,77,69,83,65,0,1,1,0,4,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, -59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110, -116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0, -0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77, -69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0, -57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116, -77,69,83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49, -0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,112,114, -105,110,116,77,69,83,65,0,1,1,0,15,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16, -8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114, -105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18, -109,0,16,10,51,0,57,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105, -110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17, -101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77, -69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,112, -114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0, -0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,20,101,0,0,0,1,4,105,110,116,95,112,114,105, -110,116,0,18,101,0,0,0,0,1,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,21,101,0,0,0,1,4,105,110, -116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 +18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,0, +112,114,105,110,116,77,69,83,65,0,1,1,0,8,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18, +118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114, +105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0, +59,119,0,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,2,118,0,0,0,1,9,58,112,114,105, +110,116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59, +121,0,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,3,118,0,0,0,1,9,58,112,114,105,110, +116,77,69,83,65,0,18,118,0,59,120,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0, +0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,122,0,0,0,0,0,1,0,0,0,0,112,114,105,110, +116,77,69,83,65,0,1,1,0,4,118,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,120,0,0,0, +0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,121,0,0,0,0,9,58,112,114,105,110,116,77,69,83, +65,0,18,118,0,59,122,0,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,118,0,59,119,0,0,0,0,0,1,0, +0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,13,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65, +0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0, +0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,14,109,0,0,0,1,9,58,112,114,105,110,116,77,69, +83,65,0,18,109,0,16,8,48,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57, +0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,0,1,0,0,0,0,112,114,105, +110,116,77,69,83,65,0,1,1,0,15,109,0,0,0,1,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,8,48, +0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0,16,10,49,0,57,0,0,0,9,58,112,114,105, +110,116,77,69,83,65,0,18,109,0,16,10,50,0,57,0,0,0,9,58,112,114,105,110,116,77,69,83,65,0,18,109,0, +16,10,51,0,57,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,16,101,0,0,0,1,4,105,110, +116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,17, +101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77, +69,83,65,0,1,1,0,18,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,0, +112,114,105,110,116,77,69,83,65,0,1,1,0,19,101,0,0,0,1,4,105,110,116,95,112,114,105,110,116,0,18, +101,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,20,101,0,0,0,1,4,105,110,116,95,112, +114,105,110,116,0,18,101,0,0,0,0,1,0,0,0,0,112,114,105,110,116,77,69,83,65,0,1,1,0,21,101,0,0,0,1, +4,105,110,116,95,112,114,105,110,116,0,18,101,0,0,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h b/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h index 8b8202ea2f..ae69c37ebe 100644 --- a/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_fragment_builtin_gc.h @@ -2,98 +2,100 @@ /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ /* slang_fragment_builtin.gc */ -3,2,2,6,12,1,103,108,95,70,114,97,103,67,111,111,114,100,0,0,0,2,2,6,1,1,103,108,95,70,114,111,110, -116,70,97,99,105,110,103,0,0,0,2,2,5,12,1,103,108,95,70,114,97,103,67,111,108,111,114,0,0,0,2,2,5, -12,1,103,108,95,70,114,97,103,68,97,116,97,0,3,18,103,108,95,77,97,120,68,114,97,119,66,117,102, -102,101,114,115,0,0,0,2,2,5,9,1,103,108,95,70,114,97,103,68,101,112,116,104,0,0,0,2,2,3,12,1,103, -108,95,67,111,108,111,114,0,0,0,2,2,3,12,1,103,108,95,83,101,99,111,110,100,97,114,121,67,111,108, -111,114,0,0,0,2,2,3,12,1,103,108,95,84,101,120,67,111,111,114,100,0,3,18,103,108,95,77,97,120,84, -101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,3,9,1,103,108,95,70,111,103,70,114,97,103, -67,111,111,114,100,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49,68,0,1,1,0,16,115,97,109,112,108, -101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,12,1,99,111,111,114, -100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111, -114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95, -114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0, -12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1, -1,0,10,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,12,1,112,99,111,111,114,100,0,0, -0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0, -59,121,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95, +4,2,2,6,0,12,1,103,108,95,70,114,97,103,67,111,111,114,100,0,0,0,2,2,6,0,1,1,103,108,95,70,114,111, +110,116,70,97,99,105,110,103,0,0,0,2,2,5,0,12,1,103,108,95,70,114,97,103,67,111,108,111,114,0,0,0, +2,2,5,0,12,1,103,108,95,70,114,97,103,68,97,116,97,0,3,18,103,108,95,77,97,120,68,114,97,119,66, +117,102,102,101,114,115,0,0,0,2,2,5,0,9,1,103,108,95,70,114,97,103,68,101,112,116,104,0,0,0,2,2,3, +0,12,1,103,108,95,67,111,108,111,114,0,0,0,2,2,3,0,12,1,103,108,95,83,101,99,111,110,100,97,114, +121,67,111,108,111,114,0,0,0,2,2,3,0,12,1,103,108,95,84,101,120,67,111,111,114,100,0,3,18,103,108, +95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0,0,2,2,3,0,9,1,103,108,95,70,111, +103,70,114,97,103,67,111,111,114,100,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,0,1,1,0,16, +115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0, +0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0, +20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120, +98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, +100,52,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109, +112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,0,12,1,112, +99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0, +18,99,111,111,114,100,0,59,121,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0, +20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109, +112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68, +80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9, +98,105,97,115,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59, +120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111, +111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95, +114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0, +0,12,0,116,101,120,116,117,114,101,50,68,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111, +111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111, +111,114,100,52,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,20,0,9,18,99,111,111,114,100,52, +0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116, +86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,116, +101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,11,99, +111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18, +112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0, +59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95, +116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18, +112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0, +17,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3, +2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111, +114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59, +119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86, +97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101, +120,116,117,114,101,51,68,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0, +0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100, +52,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,20,0,9,18,99,111,111,114,100,52,0, +59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116, +86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,116, +101,120,116,117,114,101,51,68,80,114,111,106,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,12,99, +111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18, +112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111, +114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118, +101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, +114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,67,117,98,101,0,1, +1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0, +1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111, +111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52, +95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0, +0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,0,1,1,0,20,115,97,109, +112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,0,12,1,99, +111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20, +0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98, +49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, +100,52,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,20,115,97,109,112, +108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,0,12,1,112,99, +111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18, +99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100, +0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95, 116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18, -112,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0, -16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3, -2,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100, -0,59,120,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98, -105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18, -115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114, -101,50,68,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,1,1,0,9,98,105, -97,115,0,0,0,1,3,2,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,0,18, -99,111,111,114,100,0,59,120,121,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20, -0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, -108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,80,114, -111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,98,105, -97,115,0,0,0,1,3,2,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0, -18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111, -114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114, -101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,12,0, -116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0, -12,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,12,1,112,99,111,111,114,100,0,0,0,9, -18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114, -100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99, -52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, -18,112,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,51,68,0,1,1,0,18,115,97,109, -112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,12,1,99, -111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,59, -120,121,122,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52, -95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18, -99,111,111,114,100,52,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111,106,0,1,1,0,18, -115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0, -12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111,114, -100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59, -119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86, -97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101, -120,116,117,114,101,67,117,98,101,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111, -114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111, -114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0, -18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,12,0,115,104,97,100, -111,119,49,68,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,98, -105,97,115,0,0,0,1,3,2,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121, -122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4, -118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, -101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106, -0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0, -0,0,1,3,2,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111, -114,100,0,59,120,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0, -18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20, -0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, -108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,0,1,1,0,21, -115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0, -12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114, -100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116, -101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, -111,114,100,52,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,21,115,97,109, -112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,12,1,112, -99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120, -121,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111, -111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118, -101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,9,0,100,70,100,120,0,1,1,0,9,112,0,0,0,1,4,118,101, -99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,0,59,120,120,120,120,0,0, -0,0,1,0,10,0,100,70,100,120,0,1,1,0,10,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,112,0,59,120,121,121,121,0,0,0,0,1,0,11,0,100,70,100,120,0,1, -1,0,11,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -122,0,0,18,112,0,59,120,121,122,122,0,0,0,0,1,0,12,0,100,70,100,120,0,1,1,0,12,112,0,0,0,1,4,118, -101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,0,0,1,0,9,0,100,70,100, -121,0,1,1,0,9,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120, -0,0,18,112,0,59,120,120,120,120,0,0,0,0,1,0,10,0,100,70,100,121,0,1,1,0,10,112,0,0,0,1,4,118,101, -99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,59,120,121,121,121, -0,0,0,0,1,0,11,0,100,70,100,121,0,1,1,0,11,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,59,120,121,122,122,0,0,0,0,1,0,12,0,100,70,100, -121,0,1,1,0,12,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,0,18, -112,0,0,0,0,1,0,9,0,102,119,105,100,116,104,0,1,1,0,9,112,0,0,0,1,8,58,97,98,115,0,58,100,70,100, -120,0,18,112,0,0,0,0,0,58,97,98,115,0,58,100,70,100,121,0,18,112,0,0,0,0,0,46,0,0,1,0,10,0,102,119, -105,100,116,104,0,1,1,0,10,112,0,0,0,1,8,58,97,98,115,0,58,100,70,100,120,0,18,112,0,0,0,0,0,58,97, -98,115,0,58,100,70,100,121,0,18,112,0,0,0,0,0,46,0,0,1,0,11,0,102,119,105,100,116,104,0,1,1,0,11, -112,0,0,0,1,8,58,97,98,115,0,58,100,70,100,120,0,18,112,0,0,0,0,0,58,97,98,115,0,58,100,70,100,121, -0,18,112,0,0,0,0,0,46,0,0,1,0,12,0,102,119,105,100,116,104,0,1,1,0,12,112,0,0,0,1,8,58,97,98,115,0, -58,100,70,100,120,0,18,112,0,0,0,0,0,58,97,98,115,0,58,100,70,100,121,0,18,112,0,0,0,0,0,46,0,0,0 +112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,0,1,1,0,21,115,97,109,112, +108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,0,12,1,99,111, +111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9, +18,99,111,111,114,100,52,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52,95,116,101,120,98,50, +100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100, +52,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,21,115,97,109,112,108, +101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,98,105,97,115,0,0,0,1,3,2,0,0,12,1,112,99,111, +111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0, +18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114, +100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,98,105,97,115,0,20,0,4,118,101,99,52, +95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18, +112,99,111,111,114,100,0,0,0,0,1,0,0,9,0,100,70,100,120,0,1,1,0,9,112,0,0,0,1,4,118,101,99,52,95, +100,100,120,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,112,0,59,120,120,120,120,0,0,0,0,1,0, +0,10,0,100,70,100,120,0,1,1,0,10,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,112,0,59,120,121,121,121,0,0,0,0,1,0,0,11,0,100,70,100,120,0,1,1,0, +11,112,0,0,0,1,4,118,101,99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,112,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0,100,70,100,120,0,1,1,0,12,112,0,0,0,1,4,118,101, +99,52,95,100,100,120,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,0,0,1,0,0,9,0,100,70,100,121, +0,1,1,0,9,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, +18,112,0,59,120,120,120,120,0,0,0,0,1,0,0,10,0,100,70,100,121,0,1,1,0,10,112,0,0,0,1,4,118,101,99, +52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,59,120,121,121,121,0, +0,0,0,1,0,0,11,0,100,70,100,121,0,1,1,0,11,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0,100,70, +100,121,0,1,1,0,12,112,0,0,0,1,4,118,101,99,52,95,100,100,121,0,18,95,95,114,101,116,86,97,108,0,0, +18,112,0,0,0,0,1,0,0,9,0,102,119,105,100,116,104,0,1,1,0,9,112,0,0,0,1,8,58,97,98,115,0,58,100,70, +100,120,0,18,112,0,0,0,0,0,58,97,98,115,0,58,100,70,100,121,0,18,112,0,0,0,0,0,46,0,0,1,0,0,10,0, +102,119,105,100,116,104,0,1,1,0,10,112,0,0,0,1,8,58,97,98,115,0,58,100,70,100,120,0,18,112,0,0,0,0, +0,58,97,98,115,0,58,100,70,100,121,0,18,112,0,0,0,0,0,46,0,0,1,0,0,11,0,102,119,105,100,116,104,0, +1,1,0,11,112,0,0,0,1,8,58,97,98,115,0,58,100,70,100,120,0,18,112,0,0,0,0,0,58,97,98,115,0,58,100, +70,100,121,0,18,112,0,0,0,0,0,46,0,0,1,0,0,12,0,102,119,105,100,116,104,0,1,1,0,12,112,0,0,0,1,8, +58,97,98,115,0,58,100,70,100,120,0,18,112,0,0,0,0,0,58,97,98,115,0,58,100,70,100,121,0,18,112,0,0, +0,0,0,46,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h b/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h index d47d62f909..1a08a54d7e 100644 --- a/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h @@ -2,92 +2,93 @@ /* DO NOT EDIT - THIS FILE IS AUTOMATICALLY GENERATED FROM THE FOLLOWING FILE: */ /* slang_vertex_builtin.gc */ -3,2,2,5,12,1,103,108,95,80,111,115,105,116,105,111,110,0,0,0,2,2,5,9,1,103,108,95,80,111,105,110, -116,83,105,122,101,0,0,0,2,2,5,12,1,103,108,95,67,108,105,112,86,101,114,116,101,120,0,0,0,2,2,2, -12,1,103,108,95,67,111,108,111,114,0,0,0,2,2,2,12,1,103,108,95,83,101,99,111,110,100,97,114,121,67, -111,108,111,114,0,0,0,2,2,2,11,1,103,108,95,78,111,114,109,97,108,0,0,0,2,2,2,12,1,103,108,95,86, -101,114,116,101,120,0,0,0,2,2,2,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100, -48,0,0,0,2,2,2,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,49,0,0,0,2,2,2,12, -1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,50,0,0,0,2,2,2,12,1,103,108,95,77, -117,108,116,105,84,101,120,67,111,111,114,100,51,0,0,0,2,2,2,12,1,103,108,95,77,117,108,116,105,84, -101,120,67,111,111,114,100,52,0,0,0,2,2,2,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111, -114,100,53,0,0,0,2,2,2,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,54,0,0,0,2, -2,2,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,55,0,0,0,2,2,2,9,1,103,108,95, -70,111,103,67,111,111,114,100,0,0,0,2,2,3,12,1,103,108,95,70,114,111,110,116,67,111,108,111,114,0, -0,0,2,2,3,12,1,103,108,95,66,97,99,107,67,111,108,111,114,0,0,0,2,2,3,12,1,103,108,95,70,114,111, -110,116,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,3,12,1,103,108,95,66,97,99, -107,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0,0,2,2,3,12,1,103,108,95,84,101,120,67, -111,111,114,100,0,3,18,103,108,95,77,97,120,84,101,120,116,117,114,101,67,111,111,114,100,115,0,0, -0,2,2,3,9,1,103,108,95,70,111,103,70,114,97,103,67,111,111,114,100,0,0,0,1,0,12,0,102,116,114,97, -110,115,102,111,114,109,0,0,1,9,18,95,95,114,101,116,86,97,108,0,18,103,108,95,86,101,114,116,101, -120,0,18,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97, -116,114,105,120,84,114,97,110,115,112,111,115,101,0,48,20,0,0,1,0,12,0,116,101,120,116,117,114,101, -49,68,76,111,100,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,1,1,0,9, -108,111,100,0,0,0,1,3,2,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0, -18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101, -99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114, -0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,76, -111,100,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,1,1,0,9,108,111, -100,0,0,0,1,3,2,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99, -111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,121,0,49,20,0,9,18,112,99,111,111,114,100,0, -59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86, -97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101, -120,116,117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1, -0,12,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,12,1,112,99,111,111,114,100,0,0,0,9, -18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59, -122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116, -101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99, -111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,76,111,100,0,1,1,0,17,115,97, -109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,12,1,99, -111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,0,18,99,111,111,114,100,0,59,120, -121,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101, -120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, -114,100,52,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,76,111,100,0,1,1,0,17, -115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0, -12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100, -0,59,120,121,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18, -108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0, -18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117, -114,101,50,68,80,114,111,106,76,111,100,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111, -111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99, -111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,122, -0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101, -120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111, -111,114,100,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,51,68,76,111,100,0,1,1,0,18,115,97,109, -112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,12,1,99,111, -111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120, -121,122,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116, -101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, -111,114,100,52,0,0,0,0,1,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111,106,76,111,100,0,1,1, -0,18,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3, -2,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111, -114,100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0, -59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86, -97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,12,0,116,101, -120,116,117,114,101,67,117,98,101,76,111,100,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99, -111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111, +4,2,2,5,0,12,1,103,108,95,80,111,115,105,116,105,111,110,0,0,0,2,2,5,0,9,1,103,108,95,80,111,105, +110,116,83,105,122,101,0,0,0,2,2,5,0,12,1,103,108,95,67,108,105,112,86,101,114,116,101,120,0,0,0,2, +2,2,0,12,1,103,108,95,67,111,108,111,114,0,0,0,2,2,2,0,12,1,103,108,95,83,101,99,111,110,100,97, +114,121,67,111,108,111,114,0,0,0,2,2,2,0,11,1,103,108,95,78,111,114,109,97,108,0,0,0,2,2,2,0,12,1, +103,108,95,86,101,114,116,101,120,0,0,0,2,2,2,0,12,1,103,108,95,77,117,108,116,105,84,101,120,67, +111,111,114,100,48,0,0,0,2,2,2,0,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100, +49,0,0,0,2,2,2,0,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,50,0,0,0,2,2,2,0, +12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114,100,51,0,0,0,2,2,2,0,12,1,103,108,95, +77,117,108,116,105,84,101,120,67,111,111,114,100,52,0,0,0,2,2,2,0,12,1,103,108,95,77,117,108,116, +105,84,101,120,67,111,111,114,100,53,0,0,0,2,2,2,0,12,1,103,108,95,77,117,108,116,105,84,101,120, +67,111,111,114,100,54,0,0,0,2,2,2,0,12,1,103,108,95,77,117,108,116,105,84,101,120,67,111,111,114, +100,55,0,0,0,2,2,2,0,9,1,103,108,95,70,111,103,67,111,111,114,100,0,0,0,2,2,3,0,12,1,103,108,95,70, +114,111,110,116,67,111,108,111,114,0,0,0,2,2,3,0,12,1,103,108,95,66,97,99,107,67,111,108,111,114,0, +0,0,2,2,3,0,12,1,103,108,95,70,114,111,110,116,83,101,99,111,110,100,97,114,121,67,111,108,111,114, +0,0,0,2,2,3,0,12,1,103,108,95,66,97,99,107,83,101,99,111,110,100,97,114,121,67,111,108,111,114,0,0, +0,2,2,3,0,12,1,103,108,95,84,101,120,67,111,111,114,100,0,3,18,103,108,95,77,97,120,84,101,120,116, +117,114,101,67,111,111,114,100,115,0,0,0,2,2,3,0,9,1,103,108,95,70,111,103,70,114,97,103,67,111, +111,114,100,0,0,0,1,0,0,12,0,102,116,114,97,110,115,102,111,114,109,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,18,103,108,95,86,101,114,116,101,120,0,18,103,108,95,77,111,100,101,108,86,105,101,119, +80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0,48, +20,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,76,111,100,0,1,1,0,16,115,97,109,112,108,101, +114,0,0,1,1,0,9,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100, +52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114, +100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114, +101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12, +0,116,101,120,116,117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0,16,115,97,109,112,108,101, +114,0,0,1,1,0,10,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111, +114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111, +111,114,100,0,59,121,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118, +101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, +114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111, +106,76,111,100,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9, +108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120, +0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111, +114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114, +101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12, +0,116,101,120,116,117,114,101,50,68,76,111,100,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,10, +99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18, +99,111,111,114,100,52,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,20,0,9,18,99,111,111,114, +100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114, +101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12, +0,116,101,120,116,117,114,101,50,68,80,114,111,106,76,111,100,0,1,1,0,17,115,97,109,112,108,101, +114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111, +114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18, +99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0, +4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, +108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80, +114,111,106,76,111,100,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1, +1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0, +59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18, +112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0, +18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0, +0,0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,76,111,100,0,1,1,0,18,115,97,109,112,108,101,114, +0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52, +0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,20,0, +9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,51, +100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100, +52,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111,106,76,111,100,0,1,1,0,18,115, +97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12, +1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111,114, +100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59, +119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97, +108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120, +116,117,114,101,67,117,98,101,76,111,100,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, +111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111, 111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119, 0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,12,0,115,104,97,100, -111,119,49,68,76,111,100,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0, -1,1,0,9,108,111,100,0,0,0,1,3,2,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59, -120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0, -20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109, -112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,12,0,115,104,97,100,111,119,49,68,80,114, -111,106,76,111,100,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0, -9,108,111,100,0,0,0,1,3,2,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120, -0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111, -114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18, -108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0, -18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,12,0,115,104,97,100,111, -119,50,68,76,111,100,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1, -0,9,108,111,100,0,0,0,1,3,2,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120, -121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0, -4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, -108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,12,0,115,104,97,100,111,119,50,68,80,114,111, -106,76,111,100,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9, -108,111,100,0,0,0,1,3,2,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120, -121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99, -111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59, -119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,0 +108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97, +100,111,119,49,68,76,111,100,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100, +0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100, +52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108, +111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18, +115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119, +49,68,80,114,111,106,76,111,100,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114, +100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111, +114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18, +112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114, +100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101, +116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0, +115,104,97,100,111,119,50,68,76,111,100,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, +111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111, +111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119, +0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108, +0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97,100, +111,119,50,68,80,114,111,106,76,111,100,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111, +111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99, +111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,119, +0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99, +111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95, +95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,0 -- cgit v1.2.3 From 2c2d6c90fe910e9ba9f84650ce0e80c34f165eac Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 10:29:11 -0600 Subject: mesa: fix glBindTexture comment/error string --- src/mesa/main/texobj.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c index b77a00dd15..c163a8158f 100644 --- a/src/mesa/main/texobj.c +++ b/src/mesa/main/texobj.c @@ -875,9 +875,9 @@ _mesa_BindTexture( GLenum target, GLuint texName ) if (newTexObj) { /* error checking */ if (newTexObj->Target != 0 && newTexObj->Target != target) { - /* the named texture object's dimensions don't match the target */ + /* the named texture object's target doesn't match the given target */ _mesa_error( ctx, GL_INVALID_OPERATION, - "glBindTexture(wrong dimensionality)" ); + "glBindTexture(target mismatch)" ); return; } if (newTexObj->Target == 0 && target == GL_TEXTURE_RECTANGLE_NV) { -- cgit v1.2.3 From c8f319138355a5365303e6649a953422ecc88c5f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 11:04:05 -0600 Subject: gallium: enable the call to r16_put_tile_rgba(), silences warning --- src/gallium/auxiliary/util/p_tile.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index d65e603785..5521a6d08f 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -891,7 +891,7 @@ pipe_put_tile_rgba(struct pipe_surface *ps, /*a8_l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ break; case PIPE_FORMAT_R16_SNORM: - /*r16_put_tile_rgba((short *) packed, w, h, p, src_stride);*/ + r16_put_tile_rgba((short *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_R16G16B16A16_SNORM: r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); -- cgit v1.2.3 From 600aa7501f95c520ad97ac894cfd1a639562f94c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 13:30:24 -0600 Subject: gallium: implement a bunch of missing put_tile functions --- src/gallium/auxiliary/util/p_tile.c | 163 ++++++++++++++++++++++++++++++------ 1 file changed, 137 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index 5521a6d08f..e366039a27 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -109,7 +109,7 @@ pipe_put_tile_raw(struct pipe_surface *ps, /*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ static void -a8r8g8b8_get_tile_rgba(unsigned *src, +a8r8g8b8_get_tile_rgba(const unsigned *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -156,7 +156,7 @@ a8r8g8b8_put_tile_rgba(unsigned *dst, /*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ static void -x8r8g8b8_get_tile_rgba(unsigned *src, +x8r8g8b8_get_tile_rgba(const unsigned *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -202,7 +202,7 @@ x8r8g8b8_put_tile_rgba(unsigned *dst, /*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ static void -b8g8r8a8_get_tile_rgba(unsigned *src, +b8g8r8a8_get_tile_rgba(const unsigned *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -249,7 +249,7 @@ b8g8r8a8_put_tile_rgba(unsigned *dst, /*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ static void -a1r5g5b5_get_tile_rgba(ushort *src, +a1r5g5b5_get_tile_rgba(const ushort *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -270,10 +270,37 @@ a1r5g5b5_get_tile_rgba(ushort *src, } +static void +a1r5g5b5_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); + UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + r = r >> 3; /* 5 bits */ + g = g >> 3; /* 5 bits */ + b = b >> 3; /* 5 bits */ + a = a >> 7; /* 1 bit */ + *dst++ = (a << 15) | (r << 10) | (g << 5) | b; + } + p += src_stride; + } +} + + /*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ static void -a4r4g4b4_get_tile_rgba(ushort *src, +a4r4g4b4_get_tile_rgba(const ushort *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -324,7 +351,7 @@ a4r4g4b4_put_tile_rgba(ushort *dst, /*** PIPE_FORMAT_R5G6B5_UNORM ***/ static void -r5g6b5_get_tile_rgba(ushort *src, +r5g6b5_get_tile_rgba(const ushort *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -373,7 +400,7 @@ r5g6b5_put_tile_rgba(ushort *dst, * Return each Z value as four floats in [0,1]. */ static void -z16_get_tile_rgba(ushort *src, +z16_get_tile_rgba(const ushort *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -399,7 +426,7 @@ z16_get_tile_rgba(ushort *src, /*** PIPE_FORMAT_L8_UNORM ***/ static void -l8_get_tile_rgba(ubyte *src, +l8_get_tile_rgba(const ubyte *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -419,10 +446,31 @@ l8_get_tile_rgba(ubyte *src, } +static void +l8_put_tile_rgba(ubyte *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + *dst++ = r; + } + p += src_stride; + } +} + + + /*** PIPE_FORMAT_A8_UNORM ***/ static void -a8_get_tile_rgba(ubyte *src, +a8_get_tile_rgba(const ubyte *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -441,10 +489,32 @@ a8_get_tile_rgba(ubyte *src, } } + +static void +a8_put_tile_rgba(ubyte *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned a; + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + *dst++ = a; + } + p += src_stride; + } +} + + + /*** PIPE_FORMAT_R16_SNORM ***/ static void -r16_get_tile_rgba(short *src, +r16_get_tile_rgba(const short *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -485,7 +555,7 @@ r16_put_tile_rgba(short *dst, /*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ static void -r16g16b16a16_get_tile_rgba(short *src, +r16g16b16a16_get_tile_rgba(const short *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -530,7 +600,7 @@ r16g16b16a16_put_tile_rgba(short *dst, /*** PIPE_FORMAT_I8_UNORM ***/ static void -i8_get_tile_rgba(ubyte *src, +i8_get_tile_rgba(const ubyte *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -550,13 +620,33 @@ i8_get_tile_rgba(ubyte *src, } +static void +i8_put_tile_rgba(ubyte *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + *dst++ = r; + } + p += src_stride; + } +} + + /*** PIPE_FORMAT_A8L8_UNORM ***/ static void -a8_l8_get_tile_rgba(ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) +a8l8_get_tile_rgba(const ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) { unsigned i, j; @@ -574,6 +664,27 @@ a8_l8_get_tile_rgba(ushort *src, } +static void +a8l8_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, a; + UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); + UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); + *dst++ = (a << 8) | r; + } + p += src_stride; + } +} + + /*** PIPE_FORMAT_Z32_UNORM ***/ @@ -582,7 +693,7 @@ a8_l8_get_tile_rgba(ushort *src, * Return each Z value as four floats in [0,1]. */ static void -z32_get_tile_rgba(unsigned *src, +z32_get_tile_rgba(const unsigned *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -609,7 +720,7 @@ z32_get_tile_rgba(unsigned *src, * Return Z component as four float in [0,1]. Stencil part ignored. */ static void -s8z24_get_tile_rgba(unsigned *src, +s8z24_get_tile_rgba(const unsigned *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -636,7 +747,7 @@ s8z24_get_tile_rgba(unsigned *src, * Return Z component as four float in [0,1]. Stencil part ignored. */ static void -z24s8_get_tile_rgba(unsigned *src, +z24s8_get_tile_rgba(const unsigned *src, unsigned w, unsigned h, float *p, unsigned dst_stride) @@ -663,7 +774,7 @@ z24s8_get_tile_rgba(unsigned *src, * Convert YCbCr (or YCrCb) to RGBA. */ static void -ycbcr_get_tile_rgba(ushort *src, +ycbcr_get_tile_rgba(const ushort *src, unsigned w, unsigned h, float *p, unsigned dst_stride, @@ -780,7 +891,7 @@ pipe_tile_raw_to_rgba(enum pipe_format format, i8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_A8L8_UNORM: - a8_l8_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); + a8l8_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); break; case PIPE_FORMAT_R16_SNORM: r16_get_tile_rgba((short *) src, w, h, dst, dst_stride); @@ -867,7 +978,7 @@ pipe_put_tile_rgba(struct pipe_surface *ps, b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_A1R5G5B5_UNORM: - /*a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ + a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_R5G6B5_UNORM: r5g6b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); @@ -879,16 +990,16 @@ pipe_put_tile_rgba(struct pipe_surface *ps, a4r4g4b4_put_tile_rgba((ushort *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_L8_UNORM: - /*l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ + l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_A8_UNORM: - /*a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ + a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_I8_UNORM: - /*i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);*/ + i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_A8L8_UNORM: - /*a8_l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ + a8l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride); break; case PIPE_FORMAT_R16_SNORM: r16_put_tile_rgba((short *) packed, w, h, p, src_stride); -- cgit v1.2.3 From b1ff7dac537947d412bf423a73e7eacd76f90d84 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 13:30:49 -0600 Subject: gallium: new/better debug code (disabled) --- src/gallium/drivers/softpipe/sp_setup.c | 4 ++-- src/gallium/drivers/softpipe/sp_tex_sample.c | 7 +++++++ 2 files changed, 9 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 4321ca46f4..b48eec730b 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -284,9 +284,9 @@ static void print_vertex(const struct setup_context *setup, const float (*v)[4]) { int i; - debug_printf("Vertex: (%p)\n", v); + debug_printf(" Vertex: (%p)\n", v); for (i = 0; i < setup->quad.nr_attrs; i++) { - debug_printf(" %d: %f %f %f %f\n", i, + debug_printf(" %d: %f %f %f %f\n", i, v[i][0], v[i][1], v[i][2], v[i][3]); } } diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 8e39269533..5d7a8cc61a 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -619,6 +619,13 @@ get_texel(struct tgsi_sampler *sampler, rgba[1][j] = tile->data.color[ty][tx][1]; rgba[2][j] = tile->data.color[ty][tx][2]; rgba[3][j] = tile->data.color[ty][tx][3]; + if (0) + { + char fmt[100]; + pf_sprint_name( fmt, sampler->texture->format); + printf("Get texel %f %f %f %f from %s\n", + rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j], fmt); + } } } -- cgit v1.2.3 From ae2195caf56d2eb782475254c68858a25ee7c857 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 13:31:57 -0600 Subject: gallium: s/printf/debug_printf/ --- src/gallium/drivers/softpipe/sp_tex_sample.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 5d7a8cc61a..508db5aa61 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -623,8 +623,8 @@ get_texel(struct tgsi_sampler *sampler, { char fmt[100]; pf_sprint_name( fmt, sampler->texture->format); - printf("Get texel %f %f %f %f from %s\n", - rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j], fmt); + debug_printf("Get texel %f %f %f %f from %s\n", + rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j], fmt); } } } -- cgit v1.2.3 From 275b09f2569c8bea5dcca1643c93aa6a43d7e2a2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 13:32:11 -0600 Subject: gallium: fix a few bugs on the fallback CopyTex[Sub]Image() path. --- src/mesa/state_tracker/st_cb_texture.c | 12 ++++-------- 1 file changed, 4 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 1f94a0b9ef..d8d2d2aaea 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1073,9 +1073,9 @@ fallback_copy_texsubimage(GLcontext *ctx, st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); - /* determine bottom-to-top vs. top-to-bottom order */ + /* determine bottom-to-top vs. top-to-bottom order for src renderbuffer */ if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { - destY = height - 1 - destY; + srcY = strb->Base.Height - 1 - srcY; yStep = -1; } else { @@ -1083,9 +1083,6 @@ fallback_copy_texsubimage(GLcontext *ctx, } src_surf = strb->surface; - src_surf = screen->get_tex_surface(screen, strb->texture, face, level, destZ, - PIPE_BUFFER_USAGE_CPU_READ); - dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -1098,7 +1095,7 @@ fallback_copy_texsubimage(GLcontext *ctx, const GLboolean scaleOrBias = (ctx->Pixel.DepthScale != 1.0F || ctx->Pixel.DepthBias != 0.0F); - for (row = 0; row < height; row++, srcY++, destY += yStep) { + for (row = 0; row < height; row++, srcY += yStep, destY++) { uint data[MAX_WIDTH]; pipe_get_tile_z(src_surf, srcX, srcY, width, 1, data); if (scaleOrBias) { @@ -1109,7 +1106,7 @@ fallback_copy_texsubimage(GLcontext *ctx, } else { /* RGBA format */ - for (row = 0; row < height; row++, srcY++, destY += yStep) { + for (row = 0; row < height; row++, srcY += yStep, destY++) { float data[4 * MAX_WIDTH]; pipe_get_tile_rgba(src_surf, srcX, srcY, width, 1, data); /* XXX we're ignoring convolution for now */ @@ -1123,7 +1120,6 @@ fallback_copy_texsubimage(GLcontext *ctx, } screen->tex_surface_release(screen, &dest_surf); - screen->tex_surface_release(screen, &src_surf); } -- cgit v1.2.3 From 8fb55dab783f2de5111e7440093e1458fce5fb3d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 13:35:21 -0600 Subject: gallium: s/pf_sprint_name/pf_name/ --- src/gallium/drivers/softpipe/sp_tex_sample.c | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 508db5aa61..01f4d0ca81 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -621,10 +621,9 @@ get_texel(struct tgsi_sampler *sampler, rgba[3][j] = tile->data.color[ty][tx][3]; if (0) { - char fmt[100]; - pf_sprint_name( fmt, sampler->texture->format); debug_printf("Get texel %f %f %f %f from %s\n", - rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j], fmt); + rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j], + pf_name(sampler->texture->format)); } } } -- cgit v1.2.3 From ce2137a6a4c2648a77b9697a3aa0479c9c61bacc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 19:46:39 +0100 Subject: trace: Dump format names. --- src/gallium/drivers/trace/tr_dump.c | 8 ++++++++ src/gallium/drivers/trace/tr_dump.h | 1 + src/gallium/drivers/trace/tr_screen.c | 2 +- src/gallium/drivers/trace/tr_state.c | 4 ++-- src/gallium/drivers/trace/trace.xsl | 2 +- 5 files changed, 13 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index 0591b9fc33..359a903ab9 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -280,6 +280,14 @@ void trace_dump_string(struct trace_stream *stream, trace_dump_write(stream, ""); } +void trace_dump_enum(struct trace_stream *stream, + const char *value) +{ + trace_dump_write(stream, ""); + trace_dump_escape(stream, value); + trace_dump_write(stream, ""); +} + void trace_dump_array_begin(struct trace_stream *stream) { trace_dump_write(stream, ""); diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index 5eeac8d3c5..7b15da3033 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -53,6 +53,7 @@ void trace_dump_int(struct trace_stream *stream, long int value); void trace_dump_uint(struct trace_stream *stream, long unsigned value); void trace_dump_float(struct trace_stream *stream, double value); void trace_dump_string(struct trace_stream *stream, const char *str); +void trace_dump_enum(struct trace_stream *stream, const char *value); void trace_dump_array_begin(struct trace_stream *stream); void trace_dump_array_end(struct trace_stream *stream); void trace_dump_elem_begin(struct trace_stream *stream); diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index 20a2026e1d..3a48654609 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -140,7 +140,7 @@ trace_screen_is_format_supported(struct pipe_screen *_screen, trace_dump_call_begin(stream, "pipe_screen", "is_format_supported"); trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, int, format); + trace_dump_arg(stream, format, format); trace_dump_arg(stream, int, target); trace_dump_arg(stream, uint, tex_usage); trace_dump_arg(stream, uint, geom_flags); diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index f18f5c611e..f17006dd81 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -36,7 +36,7 @@ void trace_dump_format(struct trace_stream *stream, enum pipe_format format) { - trace_dump_int(stream, format); + trace_dump_enum(stream, pf_name(format) ); } @@ -80,7 +80,7 @@ void trace_dump_template(struct trace_stream *stream, trace_dump_struct_begin(stream, "pipe_texture"); trace_dump_member(stream, int, templat, target); - trace_dump_member(stream, int, templat, format); + trace_dump_member(stream, format, templat, format); trace_dump_member_begin(stream, "width"); trace_dump_array(stream, uint, templat->width, 1); diff --git a/src/gallium/drivers/trace/trace.xsl b/src/gallium/drivers/trace/trace.xsl index 194c8b2efd..3c84c4c35a 100644 --- a/src/gallium/drivers/trace/trace.xsl +++ b/src/gallium/drivers/trace/trace.xsl @@ -71,7 +71,7 @@ along with this program. If not, see . - + -- cgit v1.2.3 From 05e90964cfd99bb8cad2d9146ab042730052b8a4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 19:58:20 +0100 Subject: trace: Merge the CSS into the XSL. --- src/gallium/drivers/trace/trace.css | 72 ------------------------------------- src/gallium/drivers/trace/trace.xsl | 58 ++++++++++++++++++++++++++++-- 2 files changed, 56 insertions(+), 74 deletions(-) delete mode 100644 src/gallium/drivers/trace/trace.css (limited to 'src') diff --git a/src/gallium/drivers/trace/trace.css b/src/gallium/drivers/trace/trace.css deleted file mode 100644 index d92c433791..0000000000 --- a/src/gallium/drivers/trace/trace.css +++ /dev/null @@ -1,72 +0,0 @@ -/**************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc. - * - * This program is free software: you can redistribute it and/or modify it - * under the terms of the GNU Lesser General Public License as published - * by the Free Software Foundation, either version 3 of the License, or - * (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU Lesser General Public License for more details. - * - * You should have received a copy of the GNU Lesser General Public License - * along with this program. If not, see . - * - ****************************************************************************/ - -body { - font-family: verdana, sans-serif; - font-size: 11px; - font-weight: normal; - text-align : left; -} - -ul.calls { - list-style: none; - margin-left: 0px; - padding-left: 0px; -} - -ul.args { - display:inline; - list-style: none; - margin-left: 0px; - padding-left: 0px; -} - -ul.args li { - display:inline; -} - -ul.elems { - list-style: none; - margin-left: 2em; - padding-left: 0px; -} - -ul.elems li { - display:block; -} - -.fun { - font-weight: bold; -} - -.var { - font-style: italic; -} - -.typ { - display: none; -} - -.lit { - color: #0000ff; -} - -.ptr { - color: #008000; -} diff --git a/src/gallium/drivers/trace/trace.xsl b/src/gallium/drivers/trace/trace.xsl index 3c84c4c35a..5a876e0f21 100644 --- a/src/gallium/drivers/trace/trace.xsl +++ b/src/gallium/drivers/trace/trace.xsl @@ -29,8 +29,62 @@ along with this program. If not, see . Gallium Trace - +
      @@ -71,7 +125,7 @@ along with this program. If not, see . - + -- cgit v1.2.3 From 6a82ea2ed21c47749676e765311baed97f5dea9f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 20:57:03 +0100 Subject: trace: Simplify HTML output. --- src/gallium/drivers/trace/trace.xsl | 47 ++++++------------------------------- 1 file changed, 7 insertions(+), 40 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/trace.xsl b/src/gallium/drivers/trace/trace.xsl index 5a876e0f21..13039d44c4 100644 --- a/src/gallium/drivers/trace/trace.xsl +++ b/src/gallium/drivers/trace/trace.xsl @@ -38,33 +38,6 @@ along with this program. If not, see . text-align : left; } - ul.calls { - list-style: none; - margin-left: 0px; - padding-left: 0px; - } - - ul.args { - display:inline; - list-style: none; - margin-left: 0px; - padding-left: 0px; - } - - ul.args li { - display:inline; - } - - ul.elems { - list-style: none; - margin-left: 2em; - padding-left: 0px; - } - - ul.elems li { - display:block; - } - .fun { font-weight: bold; } @@ -86,9 +59,9 @@ along with this program. If not, see . } -
        +
          -
      +       @@ -101,23 +74,19 @@ along with this program. If not, see . ( -
        - -
      + )       -
    • = , -
      • @@ -146,12 +115,10 @@ along with this program. If not, see . -
    • - - - , - -
      • + + + , +       -- cgit v1.2.3 From fdb7dc889f4251183915c811566ced083fdac40d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 15:05:13 -0600 Subject: softpipe: add support for PIPE_FORMAT_X8Z24_UNORM, PIPE_FORMAT_Z24X8_UNORM --- src/gallium/drivers/softpipe/sp_quad_depth_test.c | 14 ++++++++++++++ src/gallium/drivers/softpipe/sp_tile_cache.c | 2 ++ 2 files changed, 16 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c index 33888abcc5..0c82692c6e 100644 --- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c @@ -104,6 +104,8 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } } break; + case PIPE_FORMAT_X8Z24_UNORM: + /* fall-through */ case PIPE_FORMAT_S8Z24_UNORM: { float scale = (float) ((1 << 24) - 1); @@ -119,6 +121,8 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) } } break; + case PIPE_FORMAT_Z24X8_UNORM: + /* fall-through */ case PIPE_FORMAT_Z24S8_UNORM: { float scale = (float) ((1 << 24) - 1); @@ -209,6 +213,9 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) tile->data.depth16[y][x] = (ushort) bzzzz[j]; } break; + case PIPE_FORMAT_X8Z24_UNORM: + /* fall-through */ + /* (yes, this falls through to a different case than above) */ case PIPE_FORMAT_Z32_UNORM: for (j = 0; j < QUAD_SIZE; j++) { int x = quad->x0 % TILE_SIZE + (j & 1); @@ -234,6 +241,13 @@ sp_depth_test_quad(struct quad_stage *qs, struct quad_header *quad) tile->data.depth32[y][x] = z24s8; } break; + case PIPE_FORMAT_Z24X8_UNORM: + for (j = 0; j < QUAD_SIZE; j++) { + int x = quad->x0 % TILE_SIZE + (j & 1); + int y = quad->y0 % TILE_SIZE + (j >> 1); + tile->data.depth32[y][x] = bzzzz[j] << 8; + } + break; default: assert(0); } diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 5d10234945..57c12ffe33 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -170,7 +170,9 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc, PIPE_BUFFER_USAGE_CPU_WRITE); tc->depth_stencil = (ps->format == PIPE_FORMAT_S8Z24_UNORM || + ps->format == PIPE_FORMAT_X8Z24_UNORM || ps->format == PIPE_FORMAT_Z24S8_UNORM || + ps->format == PIPE_FORMAT_Z24X8_UNORM || ps->format == PIPE_FORMAT_Z16_UNORM || ps->format == PIPE_FORMAT_Z32_UNORM || ps->format == PIPE_FORMAT_S8_UNORM); -- cgit v1.2.3 From c1c4ff28f0af0b14ae64eb01ef1d087b987af2c1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 15:11:54 -0600 Subject: gallium: use PIPE_FORMAT_X8Z24_UNORM for 24-bit Z but no stencil --- src/gallium/winsys/xlib/xm_api.c | 5 +++++ 1 file changed, 5 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c index 8a32c54349..4e5441a13d 100644 --- a/src/gallium/winsys/xlib/xm_api.c +++ b/src/gallium/winsys/xlib/xm_api.c @@ -363,7 +363,12 @@ create_xmesa_buffer(XMesaDrawable d, BufferType type, stencilFormat = PIPE_FORMAT_S8_UNORM; } else { + /* no stencil */ stencilFormat = PIPE_FORMAT_NONE; + if (depthFormat == PIPE_FORMAT_S8Z24_UNORM) { + /* use 24-bit Z, undefined stencil channel */ + depthFormat = PIPE_FORMAT_X8Z24_UNORM; + } } -- cgit v1.2.3 From 4e070f10c0a2af8a563eadf7fdb09b11d9067a99 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 7 Aug 2008 17:59:41 -0600 Subject: gallium: clean-up, remove dead code, update comments --- src/mesa/state_tracker/st_cb_texture.c | 330 +++++++++------------------------ 1 file changed, 89 insertions(+), 241 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index d8d2d2aaea..202b69f59d 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -115,22 +115,7 @@ compressed_num_bytes(GLuint mesaFormat) } -static GLboolean -st_IsTextureResident(GLcontext * ctx, struct gl_texture_object *texObj) -{ -#if 0 - struct intel_context *intel = intel_context(ctx); - struct st_texture_object *stObj = st_texture_object(texObj); - - return - stObj->pt && - stObj->pt->region && - intel_is_region_resident(intel, stObj->pt->region); -#endif - return 1; -} - - +/** called via ctx->Driver.NewTextureImage() */ static struct gl_texture_image * st_NewTextureImage(GLcontext * ctx) { @@ -140,6 +125,7 @@ st_NewTextureImage(GLcontext * ctx) } +/** called via ctx->Driver.NewTextureObject() */ static struct gl_texture_object * st_NewTextureObject(GLcontext * ctx, GLuint name, GLenum target) { @@ -151,6 +137,7 @@ st_NewTextureObject(GLcontext * ctx, GLuint name, GLenum target) return &obj->base; } +/** called via ctx->Driver.DeleteTextureImage() */ static void st_DeleteTextureObject(GLcontext *ctx, struct gl_texture_object *texObj) @@ -163,6 +150,7 @@ st_DeleteTextureObject(GLcontext *ctx, } +/** called via ctx->Driver.FreeTexImageData() */ static void st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) { @@ -181,7 +169,7 @@ st_FreeTextureImageData(GLcontext * ctx, struct gl_texture_image *texImage) } -/* ================================================================ +/** * From linux kernel i386 header files, copes with odd sizes better * than COPY_DWORDS would: * XXX Put this in src/mesa/main/imports.h ??? @@ -207,7 +195,8 @@ __memcpy(void *to, const void *from, size_t n) #endif -/* The system memcpy (at least on ubuntu 5.10) has problems copying +/** + * The system memcpy (at least on ubuntu 5.10) has problems copying * to agp (writecombined) memory from a source which isn't 64-byte * aligned - there is a 4x performance falloff. * @@ -230,25 +219,14 @@ do_memcpy(void *dest, const void *src, size_t n) } -/* Functions to store texture images. Where possible, textures - * will be created or further instantiated with image data, otherwise - * images will be stored in malloc'd memory. A validation step is - * required to pull those images into a texture, or otherwise - * decide a fallback is required. - */ - - static int logbase2(int n) { - GLint i = 1; - GLint log2 = 0; - + GLint i = 1, log2 = 0; while (n > i) { i *= 2; log2++; } - return log2; } @@ -356,94 +334,6 @@ guess_and_alloc_texture(struct st_context *st, } -/* There are actually quite a few combinations this will work for, - * more than what I've listed here. - */ -static GLboolean -check_pbo_format(GLint internalFormat, - GLenum format, GLenum type, - const struct gl_texture_format *mesa_format) -{ - switch (internalFormat) { - case 4: - case GL_RGBA: - return (format == GL_BGRA && - (type == GL_UNSIGNED_BYTE || - type == GL_UNSIGNED_INT_8_8_8_8_REV) && - mesa_format == &_mesa_texformat_argb8888); - case 3: - case GL_RGB: - return (format == GL_RGB && - type == GL_UNSIGNED_SHORT_5_6_5 && - mesa_format == &_mesa_texformat_rgb565); - case GL_YCBCR_MESA: - return (type == GL_UNSIGNED_SHORT_8_8_MESA || type == GL_UNSIGNED_BYTE); - default: - return GL_FALSE; - } -} - - -/* XXX: Do this for TexSubImage also: - */ -static GLboolean -try_pbo_upload(GLcontext *ctx, - struct st_texture_image *stImage, - const struct gl_pixelstore_attrib *unpack, - GLint internalFormat, - GLint width, GLint height, - GLenum format, GLenum type, const void *pixels) -{ - return GL_FALSE; /* XXX fix flushing/locking/blitting below */ -#if 000 - struct intel_context *intel = intel_context(ctx); - struct intel_buffer_object *pbo = intel_buffer_object(unpack->BufferObj); - GLuint src_offset, src_stride; - GLuint dst_offset, dst_stride; - - if (!pbo || - ctx._ImageTransferState || - unpack->SkipPixels || unpack->SkipRows) { - _mesa_printf("%s: failure 1\n", __FUNCTION__); - return GL_FALSE; - } - - src_offset = (GLuint) pixels; - - if (unpack->RowLength > 0) - src_stride = unpack->RowLength; - else - src_stride = width; - - dst_offset = st_texture_image_offset(stImage->pt, - stImage->face, - stImage->level); - - dst_stride = stImage->pt->pitch; - - { - struct _DriBufferObject *src_buffer = - intel_bufferobj_buffer(intel, pbo, INTEL_READ); - - /* Temporary hack: cast to _DriBufferObject: - */ - struct _DriBufferObject *dst_buffer = - (struct _DriBufferObject *)stImage->pt->region->buffer; - - - intelEmitCopyBlit(intel, - stImage->pt->cpp, - src_stride, src_buffer, src_offset, - dst_stride, dst_buffer, dst_offset, - 0, 0, 0, 0, width, height, - GL_COPY); - } - - return GL_TRUE; -#endif -} - - /** * Adjust pixel unpack params and image dimensions to strip off the * texture border. @@ -482,6 +372,9 @@ strip_texture_border(GLint border, } +/** + * Do glTexImage1/2/3D(). + */ static void st_TexImage(GLcontext * ctx, GLint dims, @@ -608,36 +501,6 @@ st_TexImage(GLcontext * ctx, if (!stImage->pt) DBG("XXX: Image did not fit into texture - storing in local memory!\n"); -#if 0 /* XXX FIX when st_buffer_objects are in place */ - /* PBO fastpaths: - */ - if (dims <= 2 && - stImage->pt && - intel_buffer_object(unpack->BufferObj) && - check_pbo_format(internalFormat, format, - type, texImage->TexFormat)) { - - DBG("trying pbo upload\n"); - - - - /* Otherwise, attempt to use the blitter for PBO image uploads. - */ - if (try_pbo_upload(intel, stImage, unpack, - internalFormat, - width, height, format, type, pixels)) { - DBG("pbo upload succeeded\n"); - return; - } - - DBG("pbo upload failed\n"); - } -#else - (void) try_pbo_upload; - (void) check_pbo_format; -#endif - - /* st_CopyTexImage calls this function with pixels == NULL, with * the expectation that the texture will be set up but nothing * more will be done. This is where those calls return: @@ -734,14 +597,14 @@ st_TexImage(GLcontext * ctx, static void st_TexImage3D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint depth, - GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint depth, + GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) { st_TexImage(ctx, 3, target, level, internalFormat, width, height, depth, border, @@ -751,13 +614,13 @@ st_TexImage3D(GLcontext * ctx, static void st_TexImage2D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) { st_TexImage(ctx, 2, target, level, internalFormat, width, height, 1, border, @@ -767,13 +630,13 @@ st_TexImage2D(GLcontext * ctx, static void st_TexImage1D(GLcontext * ctx, - GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint border, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *unpack, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) + GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint border, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *unpack, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) { st_TexImage(ctx, 1, target, level, internalFormat, width, 1, 1, border, @@ -782,12 +645,12 @@ st_TexImage1D(GLcontext * ctx, static void -st_CompressedTexImage2D( GLcontext *ctx, GLenum target, GLint level, - GLint internalFormat, - GLint width, GLint height, GLint border, - GLsizei imageSize, const GLvoid *data, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage ) +st_CompressedTexImage2D(GLcontext *ctx, GLenum target, GLint level, + GLint internalFormat, + GLint width, GLint height, GLint border, + GLsizei imageSize, const GLvoid *data, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) { st_TexImage(ctx, 2, target, level, internalFormat, width, height, 1, border, @@ -806,9 +669,10 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, struct gl_texture_image *texImage, int compressed) { struct st_texture_image *stImage = st_texture_image(texImage); - GLuint dstImageStride = _mesa_image_image_stride(&ctx->Pack, texImage->Width, - texImage->Height, format, - type); + GLuint dstImageStride = _mesa_image_image_stride(&ctx->Pack, + texImage->Width, + texImage->Height, + format, type); GLuint depth; GLuint i; GLubyte *dest; @@ -868,9 +732,9 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, static void st_GetTexImage(GLcontext * ctx, GLenum target, GLint level, - GLenum format, GLenum type, GLvoid * pixels, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) + GLenum format, GLenum type, GLvoid * pixels, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) { st_get_tex_image(ctx, target, level, format, type, pixels, texObj, texImage, 0); @@ -879,9 +743,9 @@ st_GetTexImage(GLcontext * ctx, GLenum target, GLint level, static void st_GetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, - GLvoid *pixels, - const struct gl_texture_object *texObj, - const struct gl_texture_image *texImage) + GLvoid *pixels, + const struct gl_texture_object *texObj, + const struct gl_texture_image *texImage) { st_get_tex_image(ctx, target, level, 0, 0, pixels, (struct gl_texture_object *) texObj, @@ -892,14 +756,14 @@ st_GetCompressedTexImage(GLcontext *ctx, GLenum target, GLint level, static void st_TexSubimage(GLcontext * ctx, - GLint dims, - GLenum target, GLint level, - GLint xoffset, GLint yoffset, GLint zoffset, - GLint width, GLint height, GLint depth, - GLenum format, GLenum type, const void *pixels, - const struct gl_pixelstore_attrib *packing, - struct gl_texture_object *texObj, - struct gl_texture_image *texImage) + GLint dims, + GLenum target, GLint level, + GLint xoffset, GLint yoffset, GLint zoffset, + GLint width, GLint height, GLint depth, + GLenum format, GLenum type, const void *pixels, + const struct gl_pixelstore_attrib *packing, + struct gl_texture_object *texObj, + struct gl_texture_image *texImage) { struct st_texture_image *stImage = st_texture_image(texImage); GLuint dstRowStride; @@ -989,7 +853,6 @@ st_TexSubImage3D(GLcontext * ctx, } - static void st_TexSubImage2D(GLcontext * ctx, GLenum target, @@ -1071,8 +934,6 @@ fallback_copy_texsubimage(GLcontext *ctx, struct pipe_surface *src_surf, *dest_surf; GLint row, yStep; - st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); - /* determine bottom-to-top vs. top-to-bottom order for src renderbuffer */ if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { srcY = strb->Base.Height - 1 - srcY; @@ -1123,18 +984,15 @@ fallback_copy_texsubimage(GLcontext *ctx, } - - /** - * Do a CopyTex[Sub]Image using an optimized hardware (blit) path. - * Note that the region to copy has already been clip tested. - * - * Note: srcY=0=Bottom of renderbuffer + * Do a CopyTex[Sub]Image1/2/3D() using a hardware (blit) path if possible. + * Note that the region to copy has already been clipped so we know we + * won't read from outside the source renderbuffer's bounds. * - * \return GL_TRUE if success, GL_FALSE if failure (use a fallback) + * Note: srcY=0=Bottom of renderbuffer (GL convention) */ static void -do_copy_texsubimage(GLcontext *ctx, +st_copy_texsubimage(GLcontext *ctx, GLenum target, GLint level, GLint destX, GLint destY, GLint destZ, GLint srcX, GLint srcY, @@ -1147,7 +1005,7 @@ do_copy_texsubimage(GLcontext *ctx, struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); struct st_texture_image *stImage = st_texture_image(texImage); - GLenum baseFormat = texImage->InternalFormat; + const GLenum texBaseFormat = texImage->InternalFormat; struct gl_framebuffer *fb = ctx->ReadBuffer; struct st_renderbuffer *strb; struct pipe_context *pipe = ctx->st->pipe; @@ -1155,20 +1013,17 @@ do_copy_texsubimage(GLcontext *ctx, uint dest_format, src_format; GLboolean use_fallback = GL_TRUE; - (void) texImage; - - /* XX need this?*/ st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); /* determine if copying depth or color data */ - if (baseFormat == GL_DEPTH_COMPONENT) { + if (texBaseFormat == GL_DEPTH_COMPONENT) { strb = st_renderbuffer(fb->_DepthBuffer); } - else if (baseFormat == GL_DEPTH_STENCIL_EXT) { + else if (texBaseFormat == GL_DEPTH_STENCIL_EXT) { strb = st_renderbuffer(fb->_StencilBuffer); } else { - /* baseFormat == GL_RGB, GL_RGBA, GL_ALPHA, etc */ + /* texBaseFormat == GL_RGB, GL_RGBA, GL_ALPHA, etc */ strb = st_renderbuffer(fb->_ColorReadBuffer); } @@ -1192,6 +1047,7 @@ do_copy_texsubimage(GLcontext *ctx, assert(dest_surface->buffer); if (src_format == dest_format) { + /* use surface_copy() / blit */ boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP); pipe->surface_copy(pipe, do_flip, @@ -1205,14 +1061,15 @@ do_copy_texsubimage(GLcontext *ctx, width, height); use_fallback = GL_FALSE; } - else if (screen->is_format_supported(screen, strb->surface->format, + else if (screen->is_format_supported(screen, src_format, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_SAMPLER, 0) && - screen->is_format_supported(screen, dest_surface->format, + screen->is_format_supported(screen, dest_format, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_RENDER_TARGET, 0)) { + /* draw textured quad to do the copy */ boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP); int srcY0, srcY1; if (do_flip) { @@ -1238,8 +1095,9 @@ do_copy_texsubimage(GLcontext *ctx, } if (use_fallback) { + /* software fallback */ fallback_copy_texsubimage(ctx, target, level, - strb, stImage, baseFormat, + strb, stImage, texBaseFormat, destX, destY, destZ, srcX, srcY, width, height); } @@ -1276,9 +1134,9 @@ st_CopyTexImage1D(GLcontext * ctx, GLenum target, GLint level, GL_RGBA, CHAN_TYPE, NULL, &ctx->DefaultPacking, texObj, texImage); - do_copy_texsubimage(ctx, target, level, - 0, 0, 0, - x, y, width, 1); + st_copy_texsubimage(ctx, target, level, + 0, 0, 0, /* destX,Y,Z */ + x, y, width, 1); /* src X, Y, size */ } @@ -1295,11 +1153,6 @@ st_CopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, struct gl_texture_image *texImage = _mesa_select_tex_image(ctx, texObj, target, level); -#if 0 - if (border) - goto fail; -#endif - /* Setup or redefine the texture object, texture and texture * image. Don't populate yet. */ @@ -1308,10 +1161,9 @@ st_CopyTexImage2D(GLcontext * ctx, GLenum target, GLint level, GL_RGBA, CHAN_TYPE, NULL, &ctx->DefaultPacking, texObj, texImage); - - do_copy_texsubimage(ctx, target, level, - 0, 0, 0, - x, y, width, height); + st_copy_texsubimage(ctx, target, level, + 0, 0, 0, /* destX,Y,Z */ + x, y, width, height); /* src X, Y, size */ } @@ -1321,9 +1173,9 @@ st_CopyTexSubImage1D(GLcontext * ctx, GLenum target, GLint level, { const GLint yoffset = 0, zoffset = 0; const GLsizei height = 1; - do_copy_texsubimage(ctx, target, level, - xoffset, yoffset, zoffset, - x, y, width, height); + st_copy_texsubimage(ctx, target, level, + xoffset, yoffset, zoffset, /* destX,Y,Z */ + x, y, width, height); /* src X, Y, size */ } @@ -1333,9 +1185,9 @@ st_CopyTexSubImage2D(GLcontext * ctx, GLenum target, GLint level, GLint x, GLint y, GLsizei width, GLsizei height) { const GLint zoffset = 0; - do_copy_texsubimage(ctx, target, level, - xoffset, yoffset, zoffset, - x, y, width, height); + st_copy_texsubimage(ctx, target, level, + xoffset, yoffset, zoffset, /* destX,Y,Z */ + x, y, width, height); /* src X, Y, size */ } @@ -1344,14 +1196,12 @@ st_CopyTexSubImage3D(GLcontext * ctx, GLenum target, GLint level, GLint xoffset, GLint yoffset, GLint zoffset, GLint x, GLint y, GLsizei width, GLsizei height) { - do_copy_texsubimage(ctx, target, level, - xoffset, yoffset, zoffset, - x, y, width, height); + st_copy_texsubimage(ctx, target, level, + xoffset, yoffset, zoffset, /* destX,Y,Z */ + x, y, width, height); /* src X, Y, size */ } - - /** * Compute which mipmap levels that really need to be sent to the hardware. * This depends on the base image size, GL_TEXTURE_MIN_LOD, @@ -1383,7 +1233,8 @@ calculate_first_last_level(struct st_texture_object *stObj) } else { firstLevel = 0; - lastLevel = MIN2(tObj->MaxLevel, (int) tObj->Image[0][tObj->BaseLevel]->WidthLog2); + lastLevel = MIN2(tObj->MaxLevel, + (int) tObj->Image[0][tObj->BaseLevel]->WidthLog2); } break; case GL_TEXTURE_RECTANGLE_NV: @@ -1554,8 +1405,6 @@ st_finalize_texture(GLcontext *ctx, } - - void st_init_texture_functions(struct dd_function_table *functions) { @@ -1585,7 +1434,6 @@ st_init_texture_functions(struct dd_function_table *functions) functions->DeleteTexture = st_DeleteTextureObject; functions->FreeTexImageData = st_FreeTextureImageData; functions->UpdateTexturePalette = 0; - functions->IsTextureResident = st_IsTextureResident; functions->TextureMemCpy = do_memcpy; -- cgit v1.2.3 From be36f7869e8ecc4b00e414557a9699ba373e6bdd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 23:32:45 +0100 Subject: i965: Remove extraneous arg to debug_printf. --- src/gallium/drivers/i965simple/brw_context.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h index f00eb34f92..3079485180 100644 --- a/src/gallium/drivers/i965simple/brw_context.h +++ b/src/gallium/drivers/i965simple/brw_context.h @@ -188,7 +188,7 @@ extern int BRW_DEBUG; } while(0) #define PRINT(...) do { \ - debug_printf(brw->pipe.winsys, __VA_ARGS__); \ + debug_printf(__VA_ARGS__); \ } while(0) struct brw_state_flags { -- cgit v1.2.3 From 9dcb956a0618931c97693f7c74493cf296cfe86c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 8 Aug 2008 12:23:21 +0100 Subject: gallium: Add destroy callback to all *_winsys interfaces. For consistency and to simplify these objects' destruction. --- src/gallium/drivers/cell/ppu/cell_screen.c | 5 +++++ src/gallium/drivers/i915simple/i915_context.c | 3 +++ src/gallium/drivers/i915simple/i915_screen.c | 5 +++++ src/gallium/drivers/i915simple/i915_winsys.h | 2 ++ src/gallium/drivers/i965simple/brw_context.c | 3 +++ src/gallium/drivers/i965simple/brw_screen.c | 5 +++++ src/gallium/drivers/i965simple/brw_winsys.h | 2 ++ src/gallium/drivers/softpipe/sp_screen.c | 5 +++++ src/gallium/include/pipe/p_winsys.h | 2 ++ 9 files changed, 32 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c index cf9b68b695..2bf441a0c5 100644 --- a/src/gallium/drivers/cell/ppu/cell_screen.c +++ b/src/gallium/drivers/cell/ppu/cell_screen.c @@ -132,6 +132,11 @@ cell_is_format_supported( struct pipe_screen *screen, static void cell_destroy_screen( struct pipe_screen *screen ) { + struct pipe_winsys *winsys = screen->winsys; + + if(winsys->destroy) + winsys->destroy(winsys); + FREE(screen); } diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index 4c01b8d5b1..e3d19017b5 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -44,6 +44,9 @@ static void i915_destroy( struct pipe_context *pipe ) struct i915_context *i915 = i915_context( pipe ); draw_destroy( i915->draw ); + + if(i915->winsys) + i915->winsys->destroy(i915->winsys); FREE( i915 ); } diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 4b1b8af7da..0afa17bed8 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -193,6 +193,11 @@ i915_is_format_supported( struct pipe_screen *screen, static void i915_destroy_screen( struct pipe_screen *screen ) { + struct pipe_winsys *winsys = screen->winsys; + + if(winsys->destroy) + winsys->destroy(winsys); + FREE(screen); } diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h index 9afaa16a62..81904c2a74 100644 --- a/src/gallium/drivers/i915simple/i915_winsys.h +++ b/src/gallium/drivers/i915simple/i915_winsys.h @@ -75,6 +75,8 @@ struct pipe_screen; */ struct i915_winsys { + void (*destroy)( struct i915_winsys *sws ); + /** * Get the current batch buffer from the winsys. */ diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c index a276cc0535..8326f7b9c4 100644 --- a/src/gallium/drivers/i965simple/brw_context.c +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -52,6 +52,9 @@ static void brw_destroy(struct pipe_context *pipe) { struct brw_context *brw = brw_context(pipe); + if(brw->winsys->destroy) + brw->winsys->destroy(brw->winsys); + FREE(brw); } diff --git a/src/gallium/drivers/i965simple/brw_screen.c b/src/gallium/drivers/i965simple/brw_screen.c index 6d8f24d1c4..fadfbf94ab 100644 --- a/src/gallium/drivers/i965simple/brw_screen.c +++ b/src/gallium/drivers/i965simple/brw_screen.c @@ -206,6 +206,11 @@ brw_is_format_supported( struct pipe_screen *screen, static void brw_destroy_screen( struct pipe_screen *screen ) { + struct pipe_winsys *winsys = screen->winsys; + + if(winsys->destroy) + winsys->destroy(winsys); + FREE(screen); } diff --git a/src/gallium/drivers/i965simple/brw_winsys.h b/src/gallium/drivers/i965simple/brw_winsys.h index b67bd73750..ec1e400418 100644 --- a/src/gallium/drivers/i965simple/brw_winsys.h +++ b/src/gallium/drivers/i965simple/brw_winsys.h @@ -112,6 +112,8 @@ enum brw_cache_id { */ struct brw_winsys { + void (*destroy)(struct brw_winsys *); + /** * Reserve space on batch buffer. * diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index ceb5616b5d..f6b3d7ac24 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -139,6 +139,11 @@ softpipe_is_format_supported( struct pipe_screen *screen, static void softpipe_destroy_screen( struct pipe_screen *screen ) { + struct pipe_winsys *winsys = screen->winsys; + + if(winsys->destroy) + winsys->destroy(winsys); + FREE(screen); } diff --git a/src/gallium/include/pipe/p_winsys.h b/src/gallium/include/pipe/p_winsys.h index 7ebc285192..5d18291dc6 100644 --- a/src/gallium/include/pipe/p_winsys.h +++ b/src/gallium/include/pipe/p_winsys.h @@ -62,6 +62,8 @@ struct pipe_surface; */ struct pipe_winsys { + void (*destroy)( struct pipe_winsys *ws ); + /** Returns name of this winsys interface */ const char *(*get_name)( struct pipe_winsys *ws ); -- cgit v1.2.3 From 74d649d9a9e85be6dfec170b018b31626d40914b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 8 Aug 2008 12:31:23 +0100 Subject: trace: Prevent from internal calls from pipe_context to pipe_screen from being traced. --- src/gallium/drivers/trace/tr_context.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index ee8ad5eb97..242a03ccb0 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -999,11 +999,14 @@ trace_context_destroy(struct pipe_context *_pipe) struct pipe_context * trace_context_create(struct pipe_context *pipe) { + struct trace_screen *tr_scr; struct trace_context *tr_ctx; if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) return pipe; + tr_scr = trace_screen(pipe->screen); + tr_ctx = CALLOC_STRUCT(trace_context); if(!tr_ctx) return NULL; @@ -1055,5 +1058,8 @@ trace_context_create(struct pipe_context *pipe) tr_ctx->pipe = pipe; + /* We don't want to trace the pipe calls */ + pipe->screen = tr_scr->screen; + return &tr_ctx->base; } -- cgit v1.2.3 From 48cf46a29d8bccb0d83ae7e53e4ded44492cdb5f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 8 Aug 2008 09:08:44 -0600 Subject: mesa: fix out-of-bounds memory reads in swizzle_copy() --- src/mesa/main/texstore.c | 118 +++++++++++++++++++++++++++++++++++------------ 1 file changed, 89 insertions(+), 29 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index 97fdfd39b2..9fe989aaeb 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -2,7 +2,7 @@ * Mesa 3-D graphics library * Version: 7.1 * - * Copyright (C) 1999-2007 Brian Paul All Rights Reserved. + * Copyright (C) 1999-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -678,56 +678,116 @@ static void swizzle_copy(GLubyte *dst, GLuint dstComponents, const GLubyte *src, GLuint srcComponents, const GLubyte *map, GLuint count) { +#define SWZ_CPY(dst, src, count, dstComps, srcComps) \ + do { \ + GLuint i; \ + for (i = 0; i < count; i++) { \ + GLuint j; \ + if (srcComps == 4) { \ + COPY_4UBV(tmp, src); \ + } \ + else { \ + for (j = 0; j < srcComps; j++) { \ + tmp[j] = src[j]; \ + } \ + } \ + src += srcComps; \ + for (j = 0; j < dstComps; j++) { \ + dst[j] = tmp[map[j]]; \ + } \ + dst += dstComps; \ + } \ + } while (0) + GLubyte tmp[6]; - GLuint i; tmp[ZERO] = 0x0; tmp[ONE] = 0xff; + ASSERT(srcComponents <= 4); + ASSERT(dstComponents <= 4); + switch (dstComponents) { case 4: - for (i = 0; i < count; i++) { - COPY_4UBV(tmp, src); - src += srcComponents; - dst[0] = tmp[map[0]]; - dst[1] = tmp[map[1]]; - dst[2] = tmp[map[2]]; - dst[3] = tmp[map[3]]; - dst += 4; + switch (srcComponents) { + case 4: + SWZ_CPY(dst, src, count, 4, 4); + break; + case 3: + SWZ_CPY(dst, src, count, 4, 3); + break; + case 2: + SWZ_CPY(dst, src, count, 4, 2); + break; + case 1: + SWZ_CPY(dst, src, count, 4, 1); + break; + default: + ; } break; case 3: - for (i = 0; i < count; i++) { - COPY_4UBV(tmp, src); - src += srcComponents; - dst[0] = tmp[map[0]]; - dst[1] = tmp[map[1]]; - dst[2] = tmp[map[2]]; - dst += 3; + switch (srcComponents) { + case 4: + SWZ_CPY(dst, src, count, 3, 4); + break; + case 3: + SWZ_CPY(dst, src, count, 3, 3); + break; + case 2: + SWZ_CPY(dst, src, count, 3, 2); + break; + case 1: + SWZ_CPY(dst, src, count, 3, 1); + break; + default: + ; } break; case 2: - for (i = 0; i < count; i++) { - COPY_4UBV(tmp, src); - src += srcComponents; - dst[0] = tmp[map[0]]; - dst[1] = tmp[map[1]]; - dst += 2; + switch (srcComponents) { + case 4: + SWZ_CPY(dst, src, count, 2, 4); + break; + case 3: + SWZ_CPY(dst, src, count, 2, 3); + break; + case 2: + SWZ_CPY(dst, src, count, 2, 2); + break; + case 1: + SWZ_CPY(dst, src, count, 2, 1); + break; + default: + ; } break; case 1: - /* XXX investigate valgrind invalid read when running demos/texenv.c */ - for (i = 0; i < count; i++) { - COPY_4UBV(tmp, src); - src += srcComponents; - dst[0] = tmp[map[0]]; - dst += 1; + switch (srcComponents) { + case 4: + SWZ_CPY(dst, src, count, 1, 4); + break; + case 3: + SWZ_CPY(dst, src, count, 1, 3); + break; + case 2: + SWZ_CPY(dst, src, count, 1, 2); + break; + case 1: + SWZ_CPY(dst, src, count, 1, 1); + break; + default: + ; } break; + default: + ; } +#undef SWZ_CPY } + static const GLubyte map_identity[6] = { 0, 1, 2, 3, ZERO, ONE }; static const GLubyte map_3210[6] = { 3, 2, 1, 0, ZERO, ONE }; -- cgit v1.2.3 From 27858226dc4eb3e64d980ba8530c2fd2e39218cf Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 8 Aug 2008 10:39:56 -0600 Subject: gallium: fix additional glCopyTex[Sub]Image issues Use the TexFormat->StoreImage() routine for fallbacks. This handles the case of copying RGBA framebuffer data into an RGBA texture when the base format is GL_RGB (or GL_LUMINANCE, etc). In that case, we need to set the texture's alpha=1 and override the framebuffer's alpha. --- src/mesa/state_tracker/st_cb_texture.c | 118 +++++++++++++++++++++++---------- 1 file changed, 84 insertions(+), 34 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 202b69f59d..147ca5b1a2 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -932,16 +932,6 @@ fallback_copy_texsubimage(GLcontext *ctx, const uint face = texture_face(target); struct pipe_texture *pt = stImage->pt; struct pipe_surface *src_surf, *dest_surf; - GLint row, yStep; - - /* determine bottom-to-top vs. top-to-bottom order for src renderbuffer */ - if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { - srcY = strb->Base.Height - 1 - srcY; - yStep = -1; - } - else { - yStep = 1; - } src_surf = strb->surface; dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ, @@ -949,13 +939,21 @@ fallback_copy_texsubimage(GLcontext *ctx, assert(width <= MAX_WIDTH); - /* - * To avoid a large temp memory allocation, do copy row by row. - */ if (baseFormat == GL_DEPTH_COMPONENT) { const GLboolean scaleOrBias = (ctx->Pixel.DepthScale != 1.0F || ctx->Pixel.DepthBias != 0.0F); + GLint row, yStep; + /* determine bottom-to-top vs. top-to-bottom order for src buffer */ + if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { + srcY = strb->Base.Height - 1 - srcY; + yStep = -1; + } + else { + yStep = 1; + } + + /* To avoid a large temp memory allocation, do copy row by row */ for (row = 0; row < height; row++, srcY += yStep, destY++) { uint data[MAX_WIDTH]; pipe_get_tile_z(src_surf, srcX, srcY, width, 1, data); @@ -967,17 +965,54 @@ fallback_copy_texsubimage(GLcontext *ctx, } else { /* RGBA format */ - for (row = 0; row < height; row++, srcY += yStep, destY++) { - float data[4 * MAX_WIDTH]; - pipe_get_tile_rgba(src_surf, srcX, srcY, width, 1, data); - /* XXX we're ignoring convolution for now */ - if (ctx->_ImageTransferState) { - _mesa_apply_rgba_transfer_ops(ctx, - ctx->_ImageTransferState & ~IMAGE_CONVOLUTION_BIT, - width, (GLfloat (*)[4]) data); + GLfloat *tempSrc = + (GLfloat *) _mesa_malloc(width * height * 4 * sizeof(GLfloat)); + GLvoid *texDest = + st_texture_image_map(ctx->st, stImage, 0,PIPE_BUFFER_USAGE_CPU_WRITE); + + if (tempSrc && texDest) { + const GLint dims = 2; + struct gl_texture_image *texImage = &stImage->base; + GLint dstRowStride = stImage->surface->stride; + struct gl_pixelstore_attrib unpack = ctx->DefaultPacking; + + if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { + /* need to invert src */ + srcY = strb->Base.Height - srcY - height; + unpack.Invert = GL_TRUE; } - pipe_put_tile_rgba(dest_surf, destX, destY, width, 1, data); + + /* get float/RGBA image from framebuffer */ + /* XXX this usually involves a lot of int/float conversion. + * try to avoid that someday. + */ + pipe_get_tile_rgba(src_surf, srcX, srcY, width, height, tempSrc); + + /* Store into texture memory. + * Note that this does some special things such as pixel transfer + * ops and format conversion. In particular, if the dest tex format + * is actually RGBA but the user created the texture as GL_RGB we + * need to fill-in/override the alpha channel with 1.0. + */ + texImage->TexFormat->StoreImage(ctx, dims, + texImage->_BaseFormat, + texImage->TexFormat, + texDest, + destX, destY, destZ, + dstRowStride, + texImage->ImageOffsets, + width, height, 1, + GL_RGBA, GL_FLOAT, tempSrc, /* src */ + &unpack); } + else { + _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexSubImage"); + } + + if (tempSrc) + _mesa_free(tempSrc); + if (texDest) + st_texture_image_unmap(ctx->st, stImage); } screen->tex_surface_release(screen, &dest_surf); @@ -1010,8 +1045,9 @@ st_copy_texsubimage(GLcontext *ctx, struct st_renderbuffer *strb; struct pipe_context *pipe = ctx->st->pipe; struct pipe_screen *screen = pipe->screen; - uint dest_format, src_format; + enum pipe_format dest_format, src_format; GLboolean use_fallback = GL_TRUE; + GLboolean matching_base_formats; st_flush(ctx->st, PIPE_FLUSH_RENDER_CACHE, NULL); @@ -1034,21 +1070,28 @@ st_copy_texsubimage(GLcontext *ctx, src_format = strb->surface->format; dest_format = stImage->pt->format; - if (ctx->_ImageTransferState == 0x0) { - /* do blit-style copy */ - struct pipe_surface *dest_surface; - - dest_surface = screen->get_tex_surface(screen, stImage->pt, - stImage->face, - stImage->level, destZ, - PIPE_BUFFER_USAGE_GPU_WRITE); + /* + * Determine if the src framebuffer and dest texture have the same + * base format. We need this to detect a case such as the framebuffer + * being GL_RGBA but the texture being GL_RGB. If the actual hardware + * texture format stores RGBA we need to set A=1 (overriding the + * framebuffer's alpha values). We can't do that with the blit or + * textured-quad paths. + */ + matching_base_formats = (strb->Base._BaseFormat == texImage->_BaseFormat); - assert(strb->surface->buffer); - assert(dest_surface->buffer); + if (matching_base_formats && ctx->_ImageTransferState == 0x0) { + /* try potential hardware path */ + struct pipe_surface *dest_surface = NULL; if (src_format == dest_format) { /* use surface_copy() / blit */ boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP); + + dest_surface = screen->get_tex_surface(screen, stImage->pt, + stImage->face, stImage->level, + destZ, + PIPE_BUFFER_USAGE_GPU_WRITE); pipe->surface_copy(pipe, do_flip, /* dest */ @@ -1072,6 +1115,12 @@ st_copy_texsubimage(GLcontext *ctx, /* draw textured quad to do the copy */ boolean do_flip = (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP); int srcY0, srcY1; + + dest_surface = screen->get_tex_surface(screen, stImage->pt, + stImage->face, stImage->level, + destZ, + PIPE_BUFFER_USAGE_GPU_WRITE); + if (do_flip) { srcY1 = strb->Base.Height - srcY - height; srcY0 = srcY1 + height; @@ -1091,7 +1140,8 @@ st_copy_texsubimage(GLcontext *ctx, use_fallback = GL_FALSE; } - pipe_surface_reference(&dest_surface, NULL); + if (dest_surface) + pipe_surface_reference(&dest_surface, NULL); } if (use_fallback) { -- cgit v1.2.3 From 966e199e409a1b52eef88e48997442250997f45e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 8 Aug 2008 12:29:48 -0600 Subject: mesa: fix some pixel transfer state tests for depth formats --- src/mesa/main/texstore.c | 14 +++++++++----- 1 file changed, 9 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c index 9fe989aaeb..e821d9f367 100644 --- a/src/mesa/main/texstore.c +++ b/src/mesa/main/texstore.c @@ -1160,7 +1160,8 @@ _mesa_texstore_z32(TEXSTORE_PARAMS) ASSERT(dstFormat == &_mesa_texformat_z32); ASSERT(dstFormat->TexelBytes == sizeof(GLuint)); - if (!ctx->_ImageTransferState && + if (ctx->Pixel.DepthScale == 1.0f && + ctx->Pixel.DepthBias == 0.0f && !srcPacking->SwapBytes && baseInternalFormat == GL_DEPTH_COMPONENT && srcFormat == GL_DEPTH_COMPONENT && @@ -1207,7 +1208,8 @@ _mesa_texstore_z16(TEXSTORE_PARAMS) ASSERT(dstFormat == &_mesa_texformat_z16); ASSERT(dstFormat->TexelBytes == sizeof(GLushort)); - if (!ctx->_ImageTransferState && + if (ctx->Pixel.DepthScale == 1.0f && + ctx->Pixel.DepthBias == 0.0f && !srcPacking->SwapBytes && baseInternalFormat == GL_DEPTH_COMPONENT && srcFormat == GL_DEPTH_COMPONENT && @@ -2405,7 +2407,8 @@ _mesa_texstore_z24_s8(TEXSTORE_PARAMS) ASSERT(srcFormat == GL_DEPTH_STENCIL_EXT); ASSERT(srcType == GL_UNSIGNED_INT_24_8_EXT); - if (!ctx->_ImageTransferState && + if (ctx->Pixel.DepthScale == 1.0f && + ctx->Pixel.DepthBias == 0.0f && !srcPacking->SwapBytes) { /* simple path */ memcpy_texture(ctx, dims, @@ -2476,7 +2479,7 @@ _mesa_texstore_s8_z24(TEXSTORE_PARAMS) ASSERT(srcFormat == GL_DEPTH_STENCIL_EXT || srcFormat == GL_DEPTH_COMPONENT); ASSERT(srcFormat != GL_DEPTH_STENCIL_EXT || srcType == GL_UNSIGNED_INT_24_8_EXT); - /* Incase we only upload depth we need to preserve the stencil */ + /* In case we only upload depth we need to preserve the stencil */ if (srcFormat == GL_DEPTH_COMPONENT) { for (img = 0; img < srcDepth; img++) { GLuint *dstRow = (GLuint *) dstAddr @@ -2504,7 +2507,8 @@ _mesa_texstore_s8_z24(TEXSTORE_PARAMS) dstRow += dstRowStride / sizeof(GLuint); } } - } else { + } + else { for (img = 0; img < srcDepth; img++) { GLuint *dstRow = (GLuint *) dstAddr + dstImageOffsets[dstZoffset + img] -- cgit v1.2.3 From 74b14fe6ddbece8bc662aac4d3b2b18d8d853486 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 8 Aug 2008 13:06:54 -0600 Subject: mesa: fix some feature tests --- src/mesa/main/context.c | 6 ------ src/mesa/main/fbobject.c | 2 ++ 2 files changed, 2 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 27e5e2fcce..32460e92c3 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -120,9 +120,7 @@ #include "macros.h" #include "matrix.h" #include "multisample.h" -#if FEATURE_pixel_transfer #include "pixel.h" -#endif #include "pixelstore.h" #include "points.h" #include "polygon.h" @@ -1035,11 +1033,7 @@ init_attrib_groups(GLcontext *ctx) _mesa_init_lighting( ctx ); _mesa_init_matrix( ctx ); _mesa_init_multisample( ctx ); -#if FEATURE_pixel_transfer _mesa_init_pixel( ctx ); -#else - ctx->Pixel.ReadBuffer = ctx->Visual.doubleBufferMode ? GL_BACK : GL_FRONT; -#endif _mesa_init_pixelstore( ctx ); _mesa_init_point( ctx ); _mesa_init_polygon( ctx ); diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c index 960cc6da22..b5605a199c 100644 --- a/src/mesa/main/fbobject.c +++ b/src/mesa/main/fbobject.c @@ -508,6 +508,7 @@ _mesa_test_framebuffer_completeness(GLcontext *ctx, struct gl_framebuffer *fb) } } +#ifndef FEATURE_OES_framebuffer_object /* Check that all DrawBuffers are present */ for (j = 0; j < ctx->Const.MaxDrawBuffers; j++) { if (fb->ColorDrawBuffer[j] != GL_NONE) { @@ -533,6 +534,7 @@ _mesa_test_framebuffer_completeness(GLcontext *ctx, struct gl_framebuffer *fb) return; } } +#endif if (numImages == 0) { fb->_Status = GL_FRAMEBUFFER_INCOMPLETE_MISSING_ATTACHMENT_EXT; -- cgit v1.2.3 From fcfe63805d1a4b1815dec7d85a21772d02ac12b8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 8 Aug 2008 22:13:13 +0100 Subject: trace: Remove unused code. --- src/gallium/drivers/trace/tr_screen.h | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h index 5d7d667f71..40b844778f 100644 --- a/src/gallium/drivers/trace/tr_screen.h +++ b/src/gallium/drivers/trace/tr_screen.h @@ -44,8 +44,6 @@ struct trace_screen struct pipe_screen *screen; struct trace_stream *stream; - - unsigned event_no; }; -- cgit v1.2.3 From 6c7aff209ca3b310008dd345836ebc020d2db004 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 8 Aug 2008 23:11:56 +0100 Subject: trace: Trace pipe_winsys calls. --- src/gallium/drivers/trace/SConscript | 1 + src/gallium/drivers/trace/tr_screen.c | 9 +- src/gallium/drivers/trace/tr_winsys.c | 410 ++++++++++++++++++++++++++++++++++ src/gallium/drivers/trace/tr_winsys.h | 63 ++++++ src/gallium/winsys/xlib/xm_winsys.c | 7 +- 5 files changed, 482 insertions(+), 8 deletions(-) create mode 100644 src/gallium/drivers/trace/tr_winsys.c create mode 100644 src/gallium/drivers/trace/tr_winsys.h (limited to 'src') diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript index 30225b5a54..35507e21e4 100644 --- a/src/gallium/drivers/trace/SConscript +++ b/src/gallium/drivers/trace/SConscript @@ -10,6 +10,7 @@ trace = env.ConvenienceLibrary( 'tr_screen.c', 'tr_state.c', 'tr_stream.c', + 'tr_winsys.c', ]) Export('trace') \ No newline at end of file diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index 3a48654609..b40d56bcff 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -30,6 +30,7 @@ #include "tr_stream.h" #include "tr_dump.h" #include "tr_state.h" +#include "tr_winsys.h" #include "tr_screen.h" @@ -338,10 +339,6 @@ trace_screen_destroy(struct pipe_screen *_screen) trace_dump_call_end(stream); - trace_dump_trace_end(stream); - - trace_stream_close(tr_scr->stream); - FREE(tr_scr); } @@ -375,11 +372,9 @@ trace_screen_create(struct pipe_screen *screen) tr_scr->screen = screen; - tr_scr->stream = trace_stream_create("gallium", "trace"); + tr_scr->stream = trace_winsys(screen->winsys)->stream; if(!tr_scr->stream) return NULL; - trace_dump_trace_begin(tr_scr->stream, 0); - return &tr_scr->base; } diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c new file mode 100644 index 0000000000..f1ef7cbebb --- /dev/null +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -0,0 +1,410 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" + +#include "tr_stream.h" +#include "tr_dump.h" +#include "tr_state.h" +#include "tr_winsys.h" + + +static const char * +trace_winsys_get_name(struct pipe_winsys *_winsys) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + const char *result; + + trace_dump_call_begin(stream, "pipe_winsys", "get_name"); + + trace_dump_arg(stream, ptr, winsys); + + result = winsys->get_name(winsys); + + trace_dump_ret(stream, string, result); + + trace_dump_call_end(stream); + + return result; +} + + +static void +trace_winsys_flush_frontbuffer(struct pipe_winsys *_winsys, + struct pipe_surface *surface, + void *context_private) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + + trace_dump_call_begin(stream, "pipe_winsys", "flush_frontbuffer"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, surface); + trace_dump_arg(stream, ptr, context_private); + + winsys->flush_frontbuffer(winsys, surface, context_private); + + trace_dump_call_end(stream); +} + + +static struct pipe_surface * +trace_winsys_surface_alloc(struct pipe_winsys *_winsys) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + struct pipe_surface *result; + + trace_dump_call_begin(stream, "pipe_winsys", "surface_alloc"); + + trace_dump_arg(stream, ptr, winsys); + + result = winsys->surface_alloc(winsys); + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static int +trace_winsys_surface_alloc_storage(struct pipe_winsys *_winsys, + struct pipe_surface *surface, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags, + unsigned tex_usage) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + int result; + + trace_dump_call_begin(stream, "pipe_winsys", "surface_alloc_storage"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, surface); + trace_dump_arg(stream, uint, width); + trace_dump_arg(stream, uint, height); + trace_dump_arg(stream, format, format); + trace_dump_arg(stream, uint, flags); + trace_dump_arg(stream, uint, tex_usage); + + result = winsys->surface_alloc_storage(winsys, + surface, + width, height, + format, + flags, + tex_usage); + + trace_dump_ret(stream, int, result); + + trace_dump_call_end(stream); + + return result; +} + + +static void +trace_winsys_surface_release(struct pipe_winsys *_winsys, + struct pipe_surface **psurface) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + struct pipe_surface *surface = *psurface; + + trace_dump_call_begin(stream, "pipe_winsys", "surface_release"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, surface); + + winsys->surface_release(winsys, psurface); + + trace_dump_call_end(stream); +} + + +static struct pipe_buffer * +trace_winsys_buffer_create(struct pipe_winsys *_winsys, + unsigned alignment, + unsigned usage, + unsigned size) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + struct pipe_buffer *result; + + trace_dump_call_begin(stream, "pipe_winsys", "buffer_create"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, uint, alignment); + trace_dump_arg(stream, uint, usage); + trace_dump_arg(stream, uint, size); + + result = winsys->buffer_create(winsys, alignment, usage, size); + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static struct pipe_buffer * +trace_winsys_user_buffer_create(struct pipe_winsys *_winsys, + void *ptr, + unsigned bytes) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + struct pipe_buffer *result; + + trace_dump_call_begin(stream, "pipe_winsys", "user_buffer_create"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, ptr); + trace_dump_arg(stream, uint, bytes); + + result = winsys->user_buffer_create(winsys, ptr, bytes); + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static void * +trace_winsys_buffer_map(struct pipe_winsys *_winsys, + struct pipe_buffer *buffer, + unsigned usage) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + void *result; + + trace_dump_call_begin(stream, "pipe_winsys", "buffer_map"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, buffer); + trace_dump_arg(stream, uint, usage); + + result = winsys->buffer_map(winsys, buffer, usage); + + trace_dump_ret(stream, ptr, result); + + trace_dump_call_end(stream); + + return result; +} + + +static void +trace_winsys_buffer_unmap(struct pipe_winsys *_winsys, + struct pipe_buffer *buffer) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + + trace_dump_call_begin(stream, "pipe_winsys", "buffer_unmap"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, buffer); + + winsys->buffer_unmap(winsys, buffer); + + trace_dump_call_end(stream); +} + + +static void +trace_winsys_buffer_destroy(struct pipe_winsys *_winsys, + struct pipe_buffer *buffer) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + + trace_dump_call_begin(stream, "pipe_winsys", "buffer_destroy"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, buffer); + + winsys->buffer_destroy(winsys, buffer); + + trace_dump_call_end(stream); +} + + +static void +trace_winsys_fence_reference(struct pipe_winsys *_winsys, + struct pipe_fence_handle **pdst, + struct pipe_fence_handle *src) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + struct pipe_fence_handle *dst = *pdst; + + trace_dump_call_begin(stream, "pipe_winsys", "fence_reference"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, dst); + trace_dump_arg(stream, ptr, src); + + winsys->fence_reference(winsys, pdst, src); + + trace_dump_call_end(stream); +} + + +static int +trace_winsys_fence_signalled(struct pipe_winsys *_winsys, + struct pipe_fence_handle *fence, + unsigned flag) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + int result; + + trace_dump_call_begin(stream, "pipe_winsys", "fence_signalled"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, fence); + trace_dump_arg(stream, uint, flag); + + result = winsys->fence_signalled(winsys, fence, flag); + + trace_dump_ret(stream, int, result); + + trace_dump_call_end(stream); + + return result; +} + + +static int +trace_winsys_fence_finish(struct pipe_winsys *_winsys, + struct pipe_fence_handle *fence, + unsigned flag) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + int result; + + trace_dump_call_begin(stream, "pipe_winsys", "fence_finish"); + + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, fence); + trace_dump_arg(stream, uint, flag); + + result = winsys->fence_finish(winsys, fence, flag); + + trace_dump_ret(stream, int, result); + + trace_dump_call_end(stream); + + return result; +} + + +static void +trace_winsys_destroy(struct pipe_winsys *_winsys) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct trace_stream *stream = tr_ws->stream; + struct pipe_winsys *winsys = tr_ws->winsys; + + trace_dump_call_begin(stream, "pipe_winsys", "destroy"); + + trace_dump_arg(stream, ptr, winsys); + + winsys->destroy(winsys); + + trace_dump_call_end(stream); + + trace_dump_trace_end(stream); + + trace_stream_close(tr_ws->stream); + + FREE(tr_ws); +} + + +struct pipe_winsys * +trace_winsys_create(struct pipe_winsys *winsys) +{ + struct trace_winsys *tr_ws; + + if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) + return winsys; + + tr_ws = CALLOC_STRUCT(trace_winsys); + if(!tr_ws) + return NULL; + + tr_ws->base.destroy = trace_winsys_destroy; + tr_ws->base.get_name = trace_winsys_get_name; + tr_ws->base.flush_frontbuffer = trace_winsys_flush_frontbuffer; + tr_ws->base.surface_alloc = trace_winsys_surface_alloc; + tr_ws->base.surface_alloc_storage = trace_winsys_surface_alloc_storage; + tr_ws->base.surface_release = trace_winsys_surface_release; + tr_ws->base.buffer_create = trace_winsys_buffer_create; + tr_ws->base.user_buffer_create = trace_winsys_user_buffer_create; + tr_ws->base.buffer_map = trace_winsys_buffer_map; + tr_ws->base.buffer_unmap = trace_winsys_buffer_unmap; + tr_ws->base.buffer_destroy = trace_winsys_buffer_destroy; + tr_ws->base.fence_reference = trace_winsys_fence_reference; + tr_ws->base.fence_signalled = trace_winsys_fence_signalled; + tr_ws->base.fence_finish = trace_winsys_fence_finish; + + tr_ws->winsys = winsys; + + tr_ws->stream = trace_stream_create("gallium", "trace"); + if(!tr_ws->stream) + return NULL; + + trace_dump_trace_begin(tr_ws->stream, 0); + + return &tr_ws->base; +} diff --git a/src/gallium/drivers/trace/tr_winsys.h b/src/gallium/drivers/trace/tr_winsys.h new file mode 100644 index 0000000000..353b0ea7b6 --- /dev/null +++ b/src/gallium/drivers/trace/tr_winsys.h @@ -0,0 +1,63 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TR_WINSYS_H_ +#define TR_WINSYS_H_ + + +#include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "pipe/p_winsys.h" + + +struct trace_stream; + + +struct trace_winsys +{ + struct pipe_winsys base; + + struct pipe_winsys *winsys; + + struct trace_stream *stream; +}; + + +static INLINE struct trace_winsys * +trace_winsys(struct pipe_winsys *winsys) +{ + assert(winsys); + return (struct trace_winsys *)winsys; +} + + + +struct pipe_winsys * +trace_winsys_create(struct pipe_winsys *winsys); + + +#endif /* TR_WINSYS_H_ */ diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 5a01b6167b..6071a5ad5e 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -55,6 +55,7 @@ #endif #ifdef GALLIUM_TRACE +#include "trace/tr_winsys.h" #include "trace/tr_screen.h" #include "trace/tr_context.h" #endif @@ -650,7 +651,11 @@ xmesa_get_pipe_winsys(struct xmesa_visual *xm_vis) ws->base.get_name = xm_get_name; } - return &ws->base; +#ifdef GALLIUM_TRACE + return trace_winsys_create(&ws->base); +#else + return &ws->base; +#endif } -- cgit v1.2.3 From b65259de6c0a2e77550bbef6b291c6d09dfb5867 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 8 Aug 2008 23:53:53 +0100 Subject: trace: Allow to dump binary data. --- src/gallium/drivers/trace/tr_dump.c | 19 ++++++++++++++++++ src/gallium/drivers/trace/tr_dump.h | 1 + src/gallium/drivers/trace/tr_state.c | 38 +++++++++++------------------------- src/gallium/drivers/trace/trace.xsl | 6 ++++++ 4 files changed, 37 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index 359a903ab9..f545de30f4 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -272,6 +272,25 @@ void trace_dump_float(struct trace_stream *stream, trace_dump_writef(stream, "%g", value); } +void trace_dump_bytes(struct trace_stream *stream, + const void *data, + long unsigned size) +{ + static char hex_table[] = "0123456789ABCDE"; + const uint8_t *p = data; + long unsigned i; + trace_dump_write(stream, ""); + for(i = 0; i < size; ++i) { + uint8_t byte = *p++; + char str[3]; + str[0] = hex_table[byte >> 4]; + str[1] = hex_table[byte & 0xf]; + str[2] = 0; + trace_dump_write(stream, str); + } + trace_dump_write(stream, ""); +} + void trace_dump_string(struct trace_stream *stream, const char *str) { diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index 7b15da3033..b2367c3288 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -52,6 +52,7 @@ void trace_dump_bool(struct trace_stream *stream, int value); void trace_dump_int(struct trace_stream *stream, long int value); void trace_dump_uint(struct trace_stream *stream, long unsigned value); void trace_dump_float(struct trace_stream *stream, double value); +void trace_dump_bytes(struct trace_stream *stream, const void *data, long unsigned size); void trace_dump_string(struct trace_stream *stream, const char *str); void trace_dump_enum(struct trace_stream *stream, const char *value); void trace_dump_array_begin(struct trace_stream *stream); diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index f17006dd81..e0e48185fa 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -51,23 +51,6 @@ void trace_dump_block(struct trace_stream *stream, } -void trace_dump_buffer(struct trace_stream *stream, - const struct pipe_buffer *buffer) -{ - if(!buffer) { - trace_dump_null(stream); - return; - } - - trace_dump_struct_begin(stream, "pipe_buffer"); - trace_dump_member(stream, uint, buffer, alignment); - trace_dump_member(stream, uint, buffer, usage); - trace_dump_member(stream, uint, buffer, size); - /* TODO: buffer data */ - trace_dump_struct_end(stream); -} - - void trace_dump_template(struct trace_stream *stream, const struct pipe_texture *templat) { @@ -162,7 +145,11 @@ void trace_dump_poly_stipple(struct trace_stream *stream, trace_dump_struct_begin(stream, "pipe_poly_stipple"); - trace_dump_member_array(stream, uint, state, stipple); + trace_dump_member_begin(stream, "stipple"); + trace_dump_bytes(stream, + state->stipple, + sizeof(state->stipple)); + trace_dump_member_end(stream); trace_dump_struct_end(stream); } @@ -243,7 +230,7 @@ void trace_dump_constant_buffer(struct trace_stream *stream, trace_dump_struct_begin(stream, "pipe_constant_buffer"); - trace_dump_member(stream, buffer, state, buffer); + trace_dump_member(stream, ptr, state, buffer); trace_dump_member(stream, uint, state, size); trace_dump_struct_end(stream); @@ -253,21 +240,18 @@ void trace_dump_constant_buffer(struct trace_stream *stream, void trace_dump_shader_state(struct trace_stream *stream, const struct pipe_shader_state *state) { - uint32_t *p = (uint32_t *)state->tokens; - unsigned n = tgsi_num_tokens(state->tokens); - assert(state); if(!state) { trace_dump_null(stream); return; } - assert(sizeof(struct tgsi_token) == 4); - trace_dump_struct_begin(stream, "pipe_shader_state"); trace_dump_member_begin(stream, "tokens"); - trace_dump_array(stream, uint, p, n); + trace_dump_bytes(stream, + state->tokens, + sizeof(struct tgsi_token) * tgsi_num_tokens(state->tokens)); trace_dump_member_end(stream); trace_dump_struct_end(stream); @@ -433,7 +417,7 @@ void trace_dump_surface(struct trace_stream *stream, trace_dump_struct_begin(stream, "pipe_surface"); - trace_dump_member(stream, buffer, state, buffer); + trace_dump_member(stream, ptr, state, buffer); trace_dump_member(stream, format, state, format); trace_dump_member(stream, uint, state, status); trace_dump_member(stream, uint, state, clear_value); @@ -475,7 +459,7 @@ void trace_dump_vertex_buffer(struct trace_stream *stream, trace_dump_member(stream, uint, state, pitch); trace_dump_member(stream, uint, state, max_index); trace_dump_member(stream, uint, state, buffer_offset); - trace_dump_member(stream, buffer, state, buffer); + trace_dump_member(stream, ptr, state, buffer); trace_dump_struct_end(stream); } diff --git a/src/gallium/drivers/trace/trace.xsl b/src/gallium/drivers/trace/trace.xsl index 13039d44c4..991e5bf935 100644 --- a/src/gallium/drivers/trace/trace.xsl +++ b/src/gallium/drivers/trace/trace.xsl @@ -100,6 +100,12 @@ along with this program. If not, see . + + + ... + + + " -- cgit v1.2.3 From 3e1974f94ef8796a5ac9e750d47ccb63c677a85b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 9 Aug 2008 11:51:29 +0100 Subject: util: Utility functions to print to a string buffer without overflowing. --- src/gallium/auxiliary/util/u_string.h | 37 +++++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_string.h b/src/gallium/auxiliary/util/u_string.h index 73c88d87b4..abc3232b49 100644 --- a/src/gallium/auxiliary/util/u_string.h +++ b/src/gallium/auxiliary/util/u_string.h @@ -176,6 +176,43 @@ util_memmove(void *dest, const void *src, size_t n) #endif +/** + * Printable string buffer + */ +struct util_strbuf +{ + char *str; + char *ptr; + size_t left; +}; + + +static INLINE void +util_strbuf_init(struct util_strbuf *sbuf, char *str, size_t size) +{ + sbuf->str = str; + sbuf->str[0] = 0; + sbuf->ptr = sbuf->str; + sbuf->left = size; +} + + +static INLINE void +util_strbuf_printf(struct util_strbuf *sbuf, const char *format, ...) +{ + if(sbuf->left > 1) { + size_t written; + va_list ap; + va_start(ap, format); + written = util_vsnprintf(sbuf->ptr, sbuf->left, format, ap); + va_end(ap); + sbuf->ptr += written; + sbuf->left -= written; + } +} + + + #ifdef __cplusplus } #endif -- cgit v1.2.3 From 673489fa5cde4ce8d49918f20f007201a17bc45e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 9 Aug 2008 11:52:27 +0100 Subject: tgsi: Dump shaders to a string too. Again. --- src/gallium/auxiliary/tgsi/tgsi_dump.c | 166 ++++++++++++++++++++++++--------- src/gallium/auxiliary/tgsi/tgsi_dump.h | 7 ++ 2 files changed, 127 insertions(+), 46 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index 0eedb1c91e..29bb530b4d 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "pipe/p_debug.h" +#include "util/u_string.h" #include "tgsi_dump.h" #include "tgsi_iterate.h" @@ -34,28 +35,31 @@ struct dump_ctx struct tgsi_iterate_context iter; uint instno; + + struct util_strbuf *sbuf; }; static void dump_enum( + struct util_strbuf *sbuf, uint e, const char **enums, uint enum_count ) { if (e >= enum_count) - debug_printf( "%u", e ); + util_strbuf_printf( sbuf, "%u", e ); else - debug_printf( "%s", enums[e] ); + util_strbuf_printf( sbuf, "%s", enums[e] ); } -#define EOL() debug_printf( "\n" ) -#define TXT(S) debug_printf( "%s", S ) -#define CHR(C) debug_printf( "%c", C ) -#define UIX(I) debug_printf( "0x%x", I ) -#define UID(I) debug_printf( "%u", I ) -#define SID(I) debug_printf( "%d", I ) -#define FLT(F) debug_printf( "%10.4f", F ) -#define ENM(E,ENUMS) dump_enum( E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) +#define EOL() util_strbuf_printf( sbuf, "\n" ) +#define TXT(S) util_strbuf_printf( sbuf, "%s", S ) +#define CHR(C) util_strbuf_printf( sbuf, "%c", C ) +#define UIX(I) util_strbuf_printf( sbuf, "0x%x", I ) +#define UID(I) util_strbuf_printf( sbuf, "%u", I ) +#define SID(I) util_strbuf_printf( sbuf, "%d", I ) +#define FLT(F) util_strbuf_printf( sbuf, "%10.4f", F ) +#define ENM(E,ENUMS) dump_enum( sbuf, E, ENUMS, sizeof( ENUMS ) / sizeof( *ENUMS ) ) static const char *processor_type_names[] = { @@ -266,6 +270,7 @@ static const char *modulate_names[TGSI_MODULATE_COUNT] = static void _dump_register( + struct util_strbuf *sbuf, uint file, int first, int last ) @@ -282,6 +287,7 @@ _dump_register( static void _dump_register_ind( + struct util_strbuf *sbuf, uint file, int index, uint ind_file, @@ -303,6 +309,7 @@ _dump_register_ind( static void _dump_writemask( + struct util_strbuf *sbuf, uint writemask ) { if (writemask != TGSI_WRITEMASK_XYZW) { @@ -318,17 +325,23 @@ _dump_writemask( } } -void -tgsi_dump_declaration( - const struct tgsi_full_declaration *decl ) +static boolean +iter_declaration( + struct tgsi_iterate_context *iter, + struct tgsi_full_declaration *decl ) { + struct dump_ctx *ctx = (struct dump_ctx *) iter; + struct util_strbuf *sbuf = ctx->sbuf; + TXT( "DCL " ); _dump_register( + sbuf, decl->Declaration.File, decl->DeclarationRange.First, decl->DeclarationRange.Last ); _dump_writemask( + sbuf, decl->Declaration.UsageMask ); if (decl->Declaration.Semantic) { @@ -346,21 +359,35 @@ tgsi_dump_declaration( ENM( decl->Declaration.Interpolate, interpolate_names ); EOL(); -} -static boolean -iter_declaration( - struct tgsi_iterate_context *iter, - struct tgsi_full_declaration *decl ) -{ - tgsi_dump_declaration( decl ); return TRUE; } void -tgsi_dump_immediate( - const struct tgsi_full_immediate *imm ) +tgsi_dump_declaration( + const struct tgsi_full_declaration *decl ) { + static char str[1024]; + struct util_strbuf sbuf; + struct dump_ctx ctx; + + util_strbuf_init(&sbuf, str, sizeof(str)); + + ctx.sbuf = &sbuf; + + iter_declaration( &ctx.iter, (struct tgsi_full_declaration *)decl ); + + debug_printf("%s", str); +} + +static boolean +iter_immediate( + struct tgsi_iterate_context *iter, + struct tgsi_full_immediate *imm ) +{ + struct dump_ctx *ctx = (struct dump_ctx *) iter; + struct util_strbuf *sbuf = ctx->sbuf; + uint i; TXT( "IMM " ); @@ -382,22 +409,36 @@ tgsi_dump_immediate( TXT( " }" ); EOL(); -} -static boolean -iter_immediate( - struct tgsi_iterate_context *iter, - struct tgsi_full_immediate *imm ) -{ - tgsi_dump_immediate( imm ); return TRUE; } void -tgsi_dump_instruction( - const struct tgsi_full_instruction *inst, - uint instno ) +tgsi_dump_immediate( + const struct tgsi_full_immediate *imm ) { + static char str[1024]; + struct util_strbuf sbuf; + struct dump_ctx ctx; + + util_strbuf_init(&sbuf, str, sizeof(str)); + + ctx.sbuf = &sbuf; + + iter_immediate( &ctx.iter, (struct tgsi_full_immediate *)imm ); + + debug_printf("%s", str); +} + +static boolean +iter_instruction( + struct tgsi_iterate_context *iter, + struct tgsi_full_instruction *inst ) +{ + struct dump_ctx *ctx = (struct dump_ctx *) iter; + struct util_strbuf *sbuf = ctx->sbuf; + uint instno = ctx->instno++; + uint i; boolean first_reg = TRUE; @@ -426,11 +467,12 @@ tgsi_dump_instruction( CHR( ' ' ); _dump_register( + sbuf, dst->DstRegister.File, dst->DstRegister.Index, dst->DstRegister.Index ); ENM( dst->DstRegisterExtModulate.Modulate, modulate_names ); - _dump_writemask( dst->DstRegister.WriteMask ); + _dump_writemask( sbuf, dst->DstRegister.WriteMask ); first_reg = FALSE; } @@ -457,6 +499,7 @@ tgsi_dump_instruction( if (src->SrcRegister.Indirect) { _dump_register_ind( + sbuf, src->SrcRegister.File, src->SrcRegister.Index, src->SrcRegisterInd.File, @@ -464,6 +507,7 @@ tgsi_dump_instruction( } else { _dump_register( + sbuf, src->SrcRegister.File, src->SrcRegister.Index, src->SrcRegister.Index ); @@ -529,38 +573,55 @@ tgsi_dump_instruction( } EOL(); + + return TRUE; } -static boolean -iter_instruction( - struct tgsi_iterate_context *iter, - struct tgsi_full_instruction *inst ) +void +tgsi_dump_instruction( + const struct tgsi_full_instruction *inst, + uint instno ) { - struct dump_ctx *ctx = (struct dump_ctx *) iter; + static char str[1024]; + struct util_strbuf sbuf; + struct dump_ctx ctx; - tgsi_dump_instruction( inst, ctx->instno++ ); - return TRUE; + util_strbuf_init(&sbuf, str, sizeof(str)); + + ctx.instno = instno; + ctx.sbuf = &sbuf; + + iter_instruction( &ctx.iter, (struct tgsi_full_instruction *)inst ); + + debug_printf("%s", str); } static boolean prolog( - struct tgsi_iterate_context *ctx ) + struct tgsi_iterate_context *iter ) { - ENM( ctx->processor.Processor, processor_type_names ); - UID( ctx->version.MajorVersion ); + struct dump_ctx *ctx = (struct dump_ctx *) iter; + struct util_strbuf *sbuf = ctx->sbuf; + ENM( iter->processor.Processor, processor_type_names ); + UID( iter->version.MajorVersion ); CHR( '.' ); - UID( ctx->version.MinorVersion ); + UID( iter->version.MinorVersion ); EOL(); return TRUE; } void -tgsi_dump( +tgsi_dump_str( const struct tgsi_token *tokens, - uint flags ) + uint flags, + char *str, + size_t size) { + struct util_strbuf sbuf; struct dump_ctx ctx; + util_strbuf_init(&sbuf, str, size); + /* sanity checks */ assert( strcmp( opcode_names[TGSI_OPCODE_CONT], "CONT" ) == 0 ); assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); @@ -572,6 +633,19 @@ tgsi_dump( ctx.iter.epilog = NULL; ctx.instno = 0; + ctx.sbuf = &sbuf; tgsi_iterate_shader( tokens, &ctx.iter ); } + +void +tgsi_dump( + const struct tgsi_token *tokens, + uint flags ) +{ + static char str[4096]; + + tgsi_dump_str(tokens, flags, str, sizeof(str)); + + debug_printf("%s", str); +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.h b/src/gallium/auxiliary/tgsi/tgsi_dump.h index 51c230b5db..ad1e647ec9 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.h +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.h @@ -34,6 +34,13 @@ extern "C" { #endif +void +tgsi_dump_str( + const struct tgsi_token *tokens, + uint flags, + char *str, + size_t size); + void tgsi_dump( const struct tgsi_token *tokens, -- cgit v1.2.3 From 696067e781977ad54bb31b3843355701124f1b22 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 9 Aug 2008 11:52:53 +0100 Subject: trace: Dump shaders as text. --- src/gallium/drivers/trace/tr_state.c | 9 ++++---- src/gallium/drivers/trace/trace.xsl | 40 +++++++++++++++++++++++++++++++++++- 2 files changed, 44 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index e0e48185fa..9ffe77146d 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -27,7 +27,7 @@ #include "pipe/p_compiler.h" -#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_dump.h" #include "tr_dump.h" #include "tr_state.h" @@ -240,18 +240,19 @@ void trace_dump_constant_buffer(struct trace_stream *stream, void trace_dump_shader_state(struct trace_stream *stream, const struct pipe_shader_state *state) { + static char str[8192]; assert(state); if(!state) { trace_dump_null(stream); return; } + tgsi_dump_str(state->tokens, 0, str, sizeof(str)); + trace_dump_struct_begin(stream, "pipe_shader_state"); trace_dump_member_begin(stream, "tokens"); - trace_dump_bytes(stream, - state->tokens, - sizeof(struct tgsi_token) * tgsi_num_tokens(state->tokens)); + trace_dump_string(stream, str); trace_dump_member_end(stream); trace_dump_struct_end(stream); diff --git a/src/gallium/drivers/trace/trace.xsl b/src/gallium/drivers/trace/trace.xsl index 991e5bf935..9cd621e7ab 100644 --- a/src/gallium/drivers/trace/trace.xsl +++ b/src/gallium/drivers/trace/trace.xsl @@ -109,7 +109,9 @@ along with this program. If not, see . " - + + + " @@ -144,4 +146,40 @@ along with this program. If not, see . + + + + + + +
      + + + +       + + + +       +       + + + + + + + + + + + + + + + + + + + + -- cgit v1.2.3 From dc31bb5076df914ad16e063fdcc46fd2ecba9dbb Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 9 Aug 2008 11:53:16 +0100 Subject: gallium: Invert include order. --- src/gallium/include/pipe/p_format.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 0a9cad8993..d9aa5792a4 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -28,11 +28,11 @@ #ifndef PIPE_FORMAT_H #define PIPE_FORMAT_H -#include "util/u_string.h" - #include "p_compiler.h" #include "p_debug.h" +#include "util/u_string.h" + #ifdef __cplusplus extern "C" { #endif -- cgit v1.2.3 From 376f2cbb190389807c8ba6df401e06743ead9eb8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 9 Aug 2008 11:53:56 +0100 Subject: trace: Prevent tracing internal pipe driver calls. --- src/gallium/drivers/trace/tr_context.c | 9 ++++----- src/gallium/drivers/trace/tr_screen.c | 3 +++ 2 files changed, 7 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index 242a03ccb0..47a217ec7c 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -30,6 +30,7 @@ #include "tr_stream.h" #include "tr_dump.h" #include "tr_state.h" +#include "tr_winsys.h" #include "tr_screen.h" #include "tr_context.h" @@ -999,14 +1000,11 @@ trace_context_destroy(struct pipe_context *_pipe) struct pipe_context * trace_context_create(struct pipe_context *pipe) { - struct trace_screen *tr_scr; struct trace_context *tr_ctx; if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) return pipe; - tr_scr = trace_screen(pipe->screen); - tr_ctx = CALLOC_STRUCT(trace_context); if(!tr_ctx) return NULL; @@ -1058,8 +1056,9 @@ trace_context_create(struct pipe_context *pipe) tr_ctx->pipe = pipe; - /* We don't want to trace the pipe calls */ - pipe->screen = tr_scr->screen; + /* We don't want to trace the internal pipe calls */ + pipe->winsys = trace_winsys(pipe->winsys)->winsys; + pipe->screen = trace_screen(pipe->screen)->screen; return &tr_ctx->base; } diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index b40d56bcff..de885abae2 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -376,5 +376,8 @@ trace_screen_create(struct pipe_screen *screen) if(!tr_scr->stream) return NULL; + /* We don't want to trace the internal pipe calls */ + screen->winsys = trace_winsys(screen->winsys)->winsys; + return &tr_scr->base; } -- cgit v1.2.3 From 5549d35db5323829702099af6e53a8dd7c451524 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 9 Aug 2008 11:54:35 +0100 Subject: trace: Dump writes to pipe_buffers. --- src/gallium/drivers/trace/tr_winsys.c | 54 ++++++++++++++++++++++++++++++++++- src/gallium/drivers/trace/tr_winsys.h | 3 ++ 2 files changed, 56 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index f1ef7cbebb..964da5677b 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "pipe/p_util.h" +#include "util/u_hash_table.h" #include "tr_stream.h" #include "tr_dump.h" @@ -33,6 +34,18 @@ #include "tr_winsys.h" +static unsigned trace_buffer_hash(void *buffer) +{ + return (unsigned)(uintptr_t)buffer; +} + + +static int trace_buffer_compare(void *buffer1, void *buffer2) +{ + return (char *)buffer2 - (char *)buffer1; +} + + static const char * trace_winsys_get_name(struct pipe_winsys *_winsys) { @@ -232,6 +245,13 @@ trace_winsys_buffer_map(struct pipe_winsys *_winsys, trace_dump_call_end(stream); + if(result) { + if(usage & PIPE_BUFFER_USAGE_CPU_WRITE) { + assert(!hash_table_get(tr_ws->buffer_maps, buffer)); + hash_table_set(tr_ws->buffer_maps, buffer, result); + } + } + return result; } @@ -243,6 +263,30 @@ trace_winsys_buffer_unmap(struct pipe_winsys *_winsys, struct trace_winsys *tr_ws = trace_winsys(_winsys); struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; + const void *map; + + map = hash_table_get(tr_ws->buffer_maps, buffer); + if(map) { + trace_dump_call_begin(stream, "", "memcpy"); + + trace_dump_arg_begin(stream, "dst"); + trace_dump_ptr(stream, map); + trace_dump_arg_end(stream); + + trace_dump_arg_begin(stream, "src"); + trace_dump_bytes(stream, map, buffer->size); + trace_dump_arg_end(stream); + + trace_dump_arg_begin(stream, "size"); + trace_dump_uint(stream, buffer->size); + trace_dump_arg_end(stream); + + trace_dump_call_end(stream); + + winsys->buffer_unmap(winsys, buffer); + + hash_table_remove(tr_ws->buffer_maps, buffer); + } trace_dump_call_begin(stream, "pipe_winsys", "buffer_unmap"); @@ -365,6 +409,8 @@ trace_winsys_destroy(struct pipe_winsys *_winsys) trace_dump_trace_end(stream); + hash_table_destroy(tr_ws->buffer_maps); + trace_stream_close(tr_ws->stream); FREE(tr_ws); @@ -399,11 +445,17 @@ trace_winsys_create(struct pipe_winsys *winsys) tr_ws->base.fence_finish = trace_winsys_fence_finish; tr_ws->winsys = winsys; - + tr_ws->stream = trace_stream_create("gallium", "trace"); if(!tr_ws->stream) return NULL; + tr_ws->buffer_maps = hash_table_create(trace_buffer_hash, + trace_buffer_compare); + if(!tr_ws->buffer_maps) + return NULL; + + trace_dump_trace_begin(tr_ws->stream, 0); return &tr_ws->base; diff --git a/src/gallium/drivers/trace/tr_winsys.h b/src/gallium/drivers/trace/tr_winsys.h index 353b0ea7b6..a3576da867 100644 --- a/src/gallium/drivers/trace/tr_winsys.h +++ b/src/gallium/drivers/trace/tr_winsys.h @@ -34,6 +34,7 @@ #include "pipe/p_winsys.h" +struct hash_table; struct trace_stream; @@ -44,6 +45,8 @@ struct trace_winsys struct pipe_winsys *winsys; struct trace_stream *stream; + + struct hash_table *buffer_maps; }; -- cgit v1.2.3 From 22604727e469f4a9881f722be319670d5cde4519 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 14:59:26 -0600 Subject: gallium: emit sprite coords (gl_PointCoord) --- src/gallium/auxiliary/draw/draw_pipe_wide_point.c | 21 +++++++++++++++++++++ 1 file changed, 21 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index d40a07f4ae..54590984c6 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -50,6 +50,8 @@ struct widepoint_stage { uint num_texcoords; int psize_slot; + + int point_coord_fs_input; /**< input for pointcoord (and fog) */ }; @@ -84,6 +86,13 @@ static void set_texcoords(const struct widepoint_stage *wide, v->data[j][3] = tc[3]; } } + + if (wide->point_coord_fs_input >= 0) { + /* put gl_PointCoord into extra vertex output's zw components */ + uint k = wide->stage.draw->extra_vp_outputs.slot; + v->data[k][2] = tc[0]; + v->data[k][3] = tc[1]; + } } @@ -208,6 +217,18 @@ static void widepoint_first_point( struct draw_stage *stage, } } wide->num_texcoords = j; + + /* find fragment shader PointCoord/Fog input */ + wide->point_coord_fs_input = 0; /* XXX fix this! */ + + /* setup extra vp output */ + draw->extra_vp_outputs.semantic_name = TGSI_SEMANTIC_FOG; + draw->extra_vp_outputs.semantic_index = 0; + draw->extra_vp_outputs.slot = draw->vs.num_vs_outputs; + } + else { + wide->point_coord_fs_input = -1; + draw->extra_vp_outputs.slot = 0; } wide->psize_slot = -1; -- cgit v1.2.3 From d506fc0acf68c861c4e160a3d931a3f81757242b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 15:01:35 -0600 Subject: gallium: debug/print vertex tweak --- src/gallium/drivers/softpipe/sp_setup.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index b48eec730b..14aacd73c2 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -1272,10 +1272,7 @@ void setup_prepare( struct setup_context *setup ) } /* Note: nr_attrs is only used for debugging (vertex printing) */ - { - const struct sp_fragment_shader *fs = setup->softpipe->fs; - setup->quad.nr_attrs = fs->info.num_inputs + 1; /* +1 for vert pos */ - } + setup->quad.nr_attrs = draw_num_vs_outputs(sp->draw); sp->quad.first->begin(sp->quad.first); -- cgit v1.2.3 From e552140a9a55fcadfe7f28c74a7e0aa1c68cdc93 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 15:04:55 -0600 Subject: gallium: added _NEW_PROGRAM to dependencies --- src/mesa/state_tracker/st_atom_rasterizer.c | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_rasterizer.c b/src/mesa/state_tracker/st_atom_rasterizer.c index 9f4e1c1c69..e286dc5116 100644 --- a/src/mesa/state_tracker/st_atom_rasterizer.c +++ b/src/mesa/state_tracker/st_atom_rasterizer.c @@ -267,11 +267,17 @@ static void update_raster_state( struct st_context *st ) } const struct st_tracked_state st_update_rasterizer = { - "st_update_rasterizer", /* name */ - { /* dirty */ - (_NEW_LIGHT | _NEW_POLYGON | _NEW_LINE | _NEW_SCISSOR | /* mesa */ - _NEW_POINT | _NEW_BUFFERS | _NEW_MULTISAMPLE), - 0, /* st */ + "st_update_rasterizer", /* name */ + { + (_NEW_BUFFERS | + _NEW_LIGHT | + _NEW_LINE | + _NEW_MULTISAMPLE | + _NEW_POINT | + _NEW_POLYGON | + _NEW_PROGRAM | + _NEW_SCISSOR), /* mesa state dependencies*/ + 0, /* state tracker dependencies */ }, - update_raster_state /* update */ + update_raster_state /* update function */ }; -- cgit v1.2.3 From a0a614480931b281d13e7f2be2c50d20681d9de4 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 15:05:27 -0600 Subject: gallium: added comment --- src/mesa/state_tracker/st_program.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index c25c668329..59314a3232 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -187,6 +187,12 @@ st_translate_vertex_program(struct st_context *st, if (stvp->Base.Base.OutputsWritten & (1 << attr)) { GLuint slot; + /* XXX + * Pass in the fragment program's input's semantic info. + * Use the generic semantic indexes from there, instead of + * guessing below. + */ + if (outputMapping) { slot = outputMapping[attr]; assert(slot != ~0); -- cgit v1.2.3 From 6fccd8d0649436ad1a4b8b812ead563573cb9cc3 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 15:10:18 -0600 Subject: mesa: glsl: add missing sampler types in sizeof_glsl_type(), bug 17079 --- src/mesa/shader/shader_api.c | 15 +++++++++++++-- 1 file changed, 13 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c index 2b5d5ec585..d8b210be53 100644 --- a/src/mesa/shader/shader_api.c +++ b/src/mesa/shader/shader_api.c @@ -1,8 +1,8 @@ /* * Mesa 3-D graphics library - * Version: 7.0 + * Version: 7.1 * - * Copyright (C) 2004-2007 Brian Paul All Rights Reserved. + * Copyright (C) 2004-2008 Brian Paul All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -720,6 +720,17 @@ sizeof_glsl_type(GLenum type) case GL_FLOAT: case GL_INT: case GL_BOOL: + case GL_SAMPLER_1D: + case GL_SAMPLER_2D: + case GL_SAMPLER_3D: + case GL_SAMPLER_CUBE: + case GL_SAMPLER_1D_SHADOW: + case GL_SAMPLER_2D_SHADOW: + case GL_SAMPLER_2D_RECT_ARB: + case GL_SAMPLER_2D_RECT_SHADOW_ARB: + case GL_SAMPLER_1D_ARRAY_SHADOW_EXT: + case GL_SAMPLER_2D_ARRAY_SHADOW_EXT: + case GL_SAMPLER_CUBE_SHADOW_EXT: return 1; case GL_FLOAT_VEC2: case GL_INT_VEC2: -- cgit v1.2.3 From 90ffd2736c2edc967a5a72d7afeebccf8b791692 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 16:07:40 -0600 Subject: mesa: glsl: fix bugs in refract() functions, plus minor optimization --- .../shader/slang/library/slang_common_builtin.gc | 26 +++++++++++++--------- 1 file changed, 15 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index 9f2a4bdd07..3182841b1f 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -1277,34 +1277,38 @@ vec4 reflect(const vec4 I, const vec4 N) float refract(const float I, const float N, const float eta) { - float k = 1.0 - eta * eta * (1.0 - dot(N, I) * dot(N, I)); + float n_dot_i = dot(N, I); + float k = 1.0 - eta * eta * (1.0 - n_dot_i * n_dot_i); if (k < 0.0) return 0.0; - return eta * I - (eta * dot(N, I) + sqrt(k)) * N; + return eta * I - (eta * n_dot_i + sqrt(k)) * N; } vec2 refract(const vec2 I, const vec2 N, const float eta) { - float k = 1.0 - eta * eta * (1.0 - dot(N, I) * dot(N, I)); + float n_dot_i = dot(N, I); + float k = 1.0 - eta * eta * (1.0 - n_dot_i * n_dot_i); if (k < 0.0) - return 0.0; - return eta * I - (eta * dot(N, I) + sqrt(k)) * N; + return vec2(0.0); + return eta * I - (eta * n_dot_i + sqrt(k)) * N; } vec3 refract(const vec3 I, const vec3 N, const float eta) { - float k = 1.0 - eta * eta * (1.0 - dot(N, I) * dot(N, I)); + float n_dot_i = dot(N, I); + float k = 1.0 - eta * eta * (1.0 - n_dot_i * n_dot_i); if (k < 0.0) - return 0.0; - return eta * I - (eta * dot(N, I) + sqrt(k)) * N; + return vec3(0.0); + return eta * I - (eta * n_dot_i + sqrt(k)) * N; } vec4 refract(const vec4 I, const vec4 N, const float eta) { - float k = 1.0 - eta * eta * (1.0 - dot(N, I) * dot(N, I)); + float n_dot_i = dot(N, I); + float k = 1.0 - eta * eta * (1.0 - n_dot_i * n_dot_i); if (k < 0.0) - return 0.0; - return eta * I - (eta * dot(N, I) + sqrt(k)) * N; + return vec4(0.0); + return eta * I - (eta * n_dot_i + sqrt(k)) * N; } -- cgit v1.2.3 From 2023f7f91d8c40bca54db972f836c969c82ffbca Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 16:13:39 -0600 Subject: mesa: glsl: use SGE instead of SGT in step() functions --- src/mesa/shader/slang/library/slang_common_builtin.gc | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index 3182841b1f..45504f4089 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -1056,45 +1056,45 @@ vec4 mix(const vec4 x, const vec4 y, const vec4 a) } -//// step (untested) +//// step float step(const float edge, const float x) { - __asm vec4_sgt __retVal.x, x, edge; + __asm vec4_sge __retVal, x, edge; } vec2 step(const vec2 edge, const vec2 x) { - __asm vec4_sgt __retVal.xy, x, edge; + __asm vec4_sge __retVal.xy, x, edge; } vec3 step(const vec3 edge, const vec3 x) { - __asm vec4_sgt __retVal.xyz, x, edge; + __asm vec4_sge __retVal.xyz, x, edge; } vec4 step(const vec4 edge, const vec4 x) { - __asm vec4_sgt __retVal, x, edge; + __asm vec4_sge __retVal, x, edge; } vec2 step(const float edge, const vec2 v) { - __asm vec4_sgt __retVal.xy, v, edge.xx; + __asm vec4_sge __retVal.xy, v, edge; } vec3 step(const float edge, const vec3 v) { - __asm vec4_sgt __retVal.xyz, v, edge.xxx; + __asm vec4_sge __retVal.xyz, v, edge; } vec4 step(const float edge, const vec4 v) { - __asm vec4_sgt __retVal, v, edge.xxxx; + __asm vec4_sge __retVal, v, edge; } -//// smoothstep (untested) +//// smoothstep float smoothstep(const float edge0, const float edge1, const float x) { -- cgit v1.2.3 From 1c30e56cd228726225af5ff7916da8dede43e34f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 17:10:24 -0600 Subject: mesa: glsl: change refract() functions to allow inlining --- .../shader/slang/library/slang_common_builtin.gc | 28 +++++++++++++++------- 1 file changed, 20 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index 45504f4089..18b38af758 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -1279,36 +1279,48 @@ float refract(const float I, const float N, const float eta) { float n_dot_i = dot(N, I); float k = 1.0 - eta * eta * (1.0 - n_dot_i * n_dot_i); + float retval; if (k < 0.0) - return 0.0; - return eta * I - (eta * n_dot_i + sqrt(k)) * N; + retval = 0.0; + else + retval = eta * I - (eta * n_dot_i + sqrt(k)) * N; + return retval; } vec2 refract(const vec2 I, const vec2 N, const float eta) { float n_dot_i = dot(N, I); float k = 1.0 - eta * eta * (1.0 - n_dot_i * n_dot_i); + vec2 retval; if (k < 0.0) - return vec2(0.0); - return eta * I - (eta * n_dot_i + sqrt(k)) * N; + retval = vec2(0.0); + else + retval = eta * I - (eta * n_dot_i + sqrt(k)) * N; + return retval; } vec3 refract(const vec3 I, const vec3 N, const float eta) { float n_dot_i = dot(N, I); float k = 1.0 - eta * eta * (1.0 - n_dot_i * n_dot_i); + vec3 retval; if (k < 0.0) - return vec3(0.0); - return eta * I - (eta * n_dot_i + sqrt(k)) * N; + retval = vec3(0.0); + else + retval = eta * I - (eta * n_dot_i + sqrt(k)) * N; + return retval; } vec4 refract(const vec4 I, const vec4 N, const float eta) { float n_dot_i = dot(N, I); float k = 1.0 - eta * eta * (1.0 - n_dot_i * n_dot_i); + vec4 retval; if (k < 0.0) - return vec4(0.0); - return eta * I - (eta * n_dot_i + sqrt(k)) * N; + retval = vec4(0.0); + else + retval = eta * I - (eta * n_dot_i + sqrt(k)) * N; + return retval; } -- cgit v1.2.3 From 40436b31a68735bb7ffb97b411492cada22a54a2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 17:10:39 -0600 Subject: mesa: glsl: regenerated file --- .../shader/slang/library/slang_common_builtin_gc.h | 666 +++++++++++---------- 1 file changed, 337 insertions(+), 329 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index c921f08a6b..7b837a9173 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -481,343 +481,351 @@ 108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,12,0, 109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,108,114,112, 0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,9,0,115,116,101,112, -0,1,1,0,9,101,100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0, -1,1,0,10,101,100,103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101, -112,0,1,1,0,11,101,100,103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115, -116,101,112,0,1,1,0,12,101,100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95,115,103,116,0, -18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101, -112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,59,120,120,0,0,0,0,1,0,0,11, -0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0,59,120, -120,120,0,0,0,0,1,0,0,12,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1,4, -118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101,0, -59,120,120,120,120,0,0,0,0,1,0,0,9,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103, -101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9,1,116,0,2,58,99,108,97,109, -112,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, -17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,10,101,100,103,101, -48,0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99,108,97,109, -112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, -17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,11,101,100,103,101, -48,0,0,1,1,0,11,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97,109, +0,1,1,0,9,101,100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114, +101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0,1,1,0,10, +101,100,103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101,112,0,1,1,0, +11,101,100,103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115,116,101,112,0, +1,1,0,12,101,100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114, +101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0,1,1,0,9, +101,100,103,101,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101,112,0,1,1,0,9, +101,100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115,116,101,112,0,1,1, +0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116, +86,97,108,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,9,0,115,109,111,111,116,104,115,116,101, +112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9, +1,116,0,2,58,99,108,97,109,112,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18, +101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51, +0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0, +1,1,0,10,101,100,103,101,48,0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1, +116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101, +100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48, +0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1, +0,11,101,100,103,101,48,0,0,1,1,0,11,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116, +0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100, +103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0, +17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,12, +101,100,103,101,48,0,0,1,1,0,12,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2, +58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103, +101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17, +50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101, +100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99, +108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48, +0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48, +0,0,18,116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103, +101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97, +109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47, +49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0, +18,116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101, +48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109, 112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, 17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,12,101,100,103,101, -48,0,0,1,1,0,12,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109, -112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, -17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48, -0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99,108,97,109,112, -0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17, -48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0, -48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1, -1,0,9,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97,109,112,0,18, -118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0, -48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48, -47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1, -0,9,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118, -0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0, -0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48, -0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,9,120,0,0,0,1,8,58,97,98,115,0,18,120,0,0,0,0,0,1,0, -0,9,0,108,101,110,103,116,104,0,1,1,0,10,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0,2,58, -100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112, -0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0, -0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112, -0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0, -18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0, -0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0, -2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18, -112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1, -0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,3,2,1,0,9,1,100,0,2,18, -120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,0,0,0, -20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,1,0,10, -1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116, -104,0,18,100,50,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,11,118,0,0,1,1,0,11, -117,0,0,0,1,3,2,1,0,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, -108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0, -12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,0,12,1,100,52,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114, -101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0,0,11,0,99,114,111,115, -115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115,115,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,9,0,102,97,99,101,102,111, -114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100, -0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52, -95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0, -0,18,115,0,0,0,0,0,1,0,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10,78,0,0,1,1,0,10, -73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0, -18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0, -18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,11,0,102,97,99,101, -102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0,1,1,0,11,78,114,101,102,0,0,0,1,3,2,1, -0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118, -101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54, -0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,12,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,12,78,0,0, -1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101, -102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0, -48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,9,0,114,101, -102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0, -18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,10,0,114,101,102,108,101,99,116,0,1,1,0,10,73,0,0, -1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48, -47,0,0,1,0,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,0,1,8,18,73,0,17,50, -0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,12,0,114,101,102,108, -101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0, -0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,9,0,114,101,102,114,97,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78, -0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0, -48,17,49,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48, -47,48,47,0,0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0, -48,18,101,116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46, -18,78,0,48,47,0,0,1,0,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,1,1,0,9, -101,116,97,0,0,0,1,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0, -48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0, -0,10,18,107,0,17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101, -116,97,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48, -47,0,0,1,0,0,11,0,114,101,102,114,97,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,1,1,0,9,101,116,97,0, -0,0,1,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100, -111,116,0,18,78,0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0, -17,48,0,48,0,0,40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100, -111,116,0,18,78,0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,0,12, -0,114,101,102,114,97,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9, -1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,58,100,111,116,0,18,78, -0,0,18,73,0,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,47,48,47,0,0,10,18,107,0,17,48,0,48,0,0, -40,0,8,17,48,0,48,0,0,0,9,14,0,8,18,101,116,97,0,18,73,0,48,18,101,116,97,0,58,100,111,116,0,18,78, -0,0,18,73,0,0,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,0,0,1,0,0,13,0,109,97,116, -114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97,116, -50,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0, -57,48,0,0,0,0,1,0,0,14,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1, -0,0,14,110,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0, -16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0, -1,0,0,15,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0, -0,0,1,8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57, -18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0, -57,18,110,0,16,10,51,0,57,48,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1, -1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, -0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11, -118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, -18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118, -0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0, -0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52, -95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3, -0,108,101,115,115,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108, -101,115,115,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0, -18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97, -110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115, -84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108, -101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99, -52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101, -115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95, -115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0, -108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99, -52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0, -1,0,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4, -118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0, -2,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99, -52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0, -0,3,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101, -99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118, -101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0, -103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95, -115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59, -120,121,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0, -0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117, -0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8, -118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, -0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1, -1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, -0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0, -1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97, -110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18, +116,0,48,47,48,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,9,120,0,0,0,1,8,58,97,98,115,0,18,120, +0,0,0,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,10,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9, +1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18, +114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120, +0,0,18,114,0,0,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0, +3,2,1,0,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115, +113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108, +0,0,18,114,0,0,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0, +3,2,1,0,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115, +113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108, +0,0,18,114,0,0,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,3, +2,1,0,9,1,100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103, +116,104,0,18,100,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118,0,0,1,1,0,10, +117,0,0,0,1,3,2,1,0,10,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, +108,101,110,103,116,104,0,18,100,50,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0, +11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,1,0,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114, +101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116, +97,110,99,101,0,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,0,12,1,100,52,0,2,18,118,0,18,117,0,47, +0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0,0, +11,0,99,114,111,115,115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115, +115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,9,0,102, +97,99,101,102,111,114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0, +1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0, +4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78, +0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10, +78,0,0,1,1,0,10,73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78, +114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0, +17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,11, +0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0,1,1,0,11,78,114,101, +102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1, +115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105, +120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,12,0,102,97,99,101,102,111,114,119,97,114,100, +0,1,1,0,12,78,0,0,1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111, +116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0, +18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0, +0,1,0,0,9,0,114,101,102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0, +0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,10,0,114,101,102,108,101,99, +116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18, +73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78, +0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0, +12,0,114,101,102,108,101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58, +100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,9,0,114,101,102,114,97,99,116,0,1,1, +0,9,73,0,0,1,1,0,9,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58, +100,111,116,0,18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101, +116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48, +47,0,0,3,2,0,0,9,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116, +118,97,108,0,17,48,0,48,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101, +116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0, +8,18,114,101,116,118,97,108,0,0,0,1,0,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10, +78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78, +0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0, +48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,10,1, +114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118, +101,99,50,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18, +101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47, +20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,11,0,114,101,102,114,97,99,116,0,1,1,0,11,73,0,0,1,1, +0,11,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0, +18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17, +49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0, +11,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58, +118,101,99,51,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48, +18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48, +47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,12,0,114,101,102,114,97,99,116,0,1,1,0,12,73,0,0,1, +1,0,12,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116, +0,18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48, +17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0, +0,12,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0, +58,118,101,99,52,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0, +48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0, +48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,13,0,109,97,116,114,105,120,67,111,109,112,77, +117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97,116,50,0,18,109,0,16,8,48,0,57,18, +110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,0,14,0,109,97, +116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1,0,0,14,110,0,0,0,1,8,58,109,97, +116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49, +0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,0,15,0,109,97,116,114,105, +120,67,111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0,0,0,1,8,58,109,97,116,52,0,18, +109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0, +18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, +48,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118, +101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52, +95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0, +0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115, +108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115, +84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97, +110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,0,1, +1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108, +0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,10, +117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0, +1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,69,113, +117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114, +101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117, +97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,69, +113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104, +97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18, +95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84, +104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114, +84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116, +101,114,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18, 95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84, -104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101, -0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101, -97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99, -52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0, -1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118, -0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0, -0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115, +104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,0,3,0,103, +114,101,97,116,101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115, +103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0, +103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95, +115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101, +97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101, +99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1, +0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118, +0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, +117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1, +0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108, +0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108, +0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110, +69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116, +101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95, +115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117, +97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,11, +117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12, +118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, +0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115, 101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101, -113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108, -0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, -97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0, -0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0, -18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52, -95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0, -0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18, -95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,2, -117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0, -0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117, -0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99, -52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111, -116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69, -113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113, -117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114, -101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,6, -117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1, -1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, -0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8, -118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, -0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99, -52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0, -0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115, -110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0, -110,111,116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101, -0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,2,118, -0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0, -18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86, -97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0, -3,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59, -120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0, -59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1, -0,97,110,121,0,1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0, -18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100, -100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99, -52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120, -0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0, -0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0, -0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112, -114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,112, +113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1, +1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108, +0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4, +118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0, +0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115, +101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0, +101,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, +114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1, +0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108, +0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,11,117, +0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,12,117,0,0,1, +1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, +18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118, +101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0, +4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113, +117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0, +1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0, +4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0, +4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0, +0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0, +59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,115,117,109, +0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59, +121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0, +18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, +18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,4,118,0,0,0,1,3,2,0, +0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59, +120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115, +117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120, +0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, +114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97, +108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52, +95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1, +0,0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4, +118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0, +18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112, +114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,112, 114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18, 118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112, -114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0, -1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0, -4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0, -18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112, -114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52, -95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1, -0,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,4,0,110,111,116,0,1,1,0,4,118,0, -0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0, -0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1, -0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116, -117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114, -100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115, -97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,121,121,0,0,0,0,1,0,0,12,0,116,101, -120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111, -111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0, -18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114, -101,50,68,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101, -99,52,95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, -18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0, -17,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, -120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, -111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111, -106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, +114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116, +105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,119,0,0,0,4,118,101,99, +52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0, +1,0,0,2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116, +86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3,0,110,111,116,0,1,1,0,3,118,0,0, +0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0, +0,17,48,0,48,0,0,0,0,0,1,0,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113, +0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,12,0,116,101,120,116, +117,114,101,49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4, +118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, +114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106, +0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95, +116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18, +99,111,111,114,100,0,59,120,121,121,121,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80, +114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118, +101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, +114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,0,1,1,0,17,115, +97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50, +100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0, +0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108, +101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95, +95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121, +122,122,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109, +112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100, +0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0, +0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,11, +99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,51,100,0,18,95,95,114,101,116,86,97,108, +0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117, +114,101,51,68,80,114,111,106,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100, +0,0,0,1,4,118,101,99,52,95,116,101,120,112,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97, +109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,67, +117,98,101,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101, +99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, +114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,0,1,1,0,20,115,97, +109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100, +0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0, +0,1,0,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,20,115,97,109,112,108,101,114,0,0, +1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101, +116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115, +104,97,100,111,119,50,68,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0, +0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, +108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,80,114,111, +106,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, 95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, -18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,0,1,1,0,18,115,97,109, -112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,51,100,0,18, -95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1, -0,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111,106,0,1,1,0,18,115,97,109,112,108,101,114,0, -0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,51,100,0,18,95,95,114,101, -116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116, -101,120,116,117,114,101,67,117,98,101,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, -111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108, -0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111, -119,49,68,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101, -99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, -18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,20, -115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120, -112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, -114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,0,1,1,0,21,115,97,109,112,108,101,114,0,0, -1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116, -86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104, -97,100,111,119,50,68,80,114,111,106,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, -114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18, -115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114, -101,50,68,82,101,99,116,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0, -0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115, -97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50, -68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114, -100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108, -0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12, -0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101, -114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0, -18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,99,111,111,114,100,0,0, -0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,0,1,1,0,23,115,97,109,112,108,101,114,0, -0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95, -114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12, -0,115,104,97,100,111,119,50,68,82,101,99,116,80,114,111,106,0,1,1,0,23,115,97,109,112,108,101,114, -0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1, -0,0,9,0,110,111,105,115,101,49,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101, -49,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,10, -120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,50,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,11,120,0,0,0,1,4,102,108,111,97,116,95,110, -111,105,115,101,51,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115, -101,49,0,1,1,0,12,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,52,0,18,95,95,114,101, -116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,9,120,0,0,0,1,9,18,95, +18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,0,1,1,0, +22,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, +120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, +111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106, +0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95, +116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, +114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101, +50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, +114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97, +108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97, +100,111,119,50,68,82,101,99,116,0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114, +100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0, +18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119, +50,68,82,101,99,116,80,114,111,106,0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, +114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97, +108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,9,0,110,111,105,115, +101,49,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,49,0,18,95,95,114,101,116, +86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,10,120,0,0,0,1,4,102,108,111, +97,116,95,110,111,105,115,101,50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110, +111,105,115,101,49,0,1,1,0,11,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,51,0,18,95, +95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,12,120,0,0,0, +1,4,102,108,111,97,116,95,110,111,105,115,101,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, +0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, +110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115, +101,50,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101, +49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18, +120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,0,1,0,0,10,0,110, +111,105,115,101,50,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111, +105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115, +101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0, +0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116, +86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17, +55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105, +115,101,51,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115, +101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0, +18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111, +105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1, +0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0, +0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118, +101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49, +55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,11,120,0,0,0,1,9,18,95, 95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0, -0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51, +52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56, +53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,12,120,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0, +17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9, +18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0, +17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0, +20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, +59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0, +58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,17,50,51,0,53,52,0,0,46,0,0,20,0, +0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, 120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, 110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0, -46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54, -54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,12,120,0,0, -0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18, +46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58, +118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,50,51,0,53,52,0,0,0, +17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,11,120,0,0,0,1,9, +18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57, +0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97, +108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49, +55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0, +58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0, +0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,12,120,0,0,0, +1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18, 95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17, -49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,0,1, -0,0,11,0,110,111,105,115,101,51,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0, -58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110, -111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111, -105,115,101,51,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105, -115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101, -49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95, -95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53, -0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,11, -120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, -99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0, -52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105, -115,101,51,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115, -101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0, -18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0, -55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0, -18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17, -49,51,0,49,57,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,9,120,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52, -55,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120, -0,17,50,51,0,53,52,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,10,120,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57, -0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110, -111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0, -0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, -99,50,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115, -101,52,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101, -49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18, -120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0, -20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, -99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,50,51, -0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105, -115,101,52,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115, -101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0, -18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0, -55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0, -18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17, -49,51,0,49,57,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115, -101,49,0,18,120,0,58,118,101,99,52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57, -49,0,0,0,17,51,55,0,52,56,0,0,0,0,46,0,0,20,0,0,0 +49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17, +53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20, +0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99, +52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,17,51,55,0,52,56,0,0,0,0, +46,0,0,20,0,0,0 -- cgit v1.2.3 From 66ef96f6dc4cd898edb862e45965795fac46caed Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 12 Aug 2008 11:30:50 +0200 Subject: softpipe: Include missing header. --- src/gallium/drivers/softpipe/sp_setup.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index 14aacd73c2..b7f2f16307 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -39,6 +39,7 @@ #include "sp_quad.h" #include "sp_state.h" #include "sp_prim_setup.h" +#include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vertex.h" #include "pipe/p_util.h" -- cgit v1.2.3 From 23e8c92543d95c216f3459a5618611c079e86174 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 9 Aug 2008 15:07:40 +0100 Subject: python: Update the documentation. --- src/gallium/state_trackers/python/README | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/README b/src/gallium/state_trackers/python/README index 75341aa28c..8ab88ab2f8 100644 --- a/src/gallium/state_trackers/python/README +++ b/src/gallium/state_trackers/python/README @@ -7,7 +7,7 @@ To build you'll need: * Python (with development packages) * SCons * SWIG -* Python Imaging Library (for the samples) +* Python Imaging Library with TK support (for the samples) Invoke scons on the top dir as @@ -19,9 +19,9 @@ To use do and then try running - python src/gallium/state_trackers/python/samples/simple.py + python src/gallium/state_trackers/python/samples/tri.py -which should create a simple.png +which should show a triangle. This is still in experimental phase, and there many limitations to what you can -- cgit v1.2.3 From 91f6032919f8e5718004bb7ac0ee2b015fb403d7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 10 Aug 2008 16:24:15 +0100 Subject: trace: Trace texture depth. --- src/gallium/drivers/trace/tr_state.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index 9ffe77146d..3c7b007bfe 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -73,6 +73,10 @@ void trace_dump_template(struct trace_stream *stream, trace_dump_array(stream, uint, templat->height, 1); trace_dump_member_end(stream); + trace_dump_member_begin(stream, "depth"); + trace_dump_array(stream, uint, templat->depth, 1); + trace_dump_member_end(stream); + trace_dump_member_begin(stream, "block"); trace_dump_block(stream, &templat->block); trace_dump_member_end(stream); -- cgit v1.2.3 From 94cf4f15c3a94311ffeb670459e285d2c70a5d7e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 10 Aug 2008 16:24:43 +0100 Subject: trace: Trace winsys/screen/context creation. --- src/gallium/drivers/trace/tr_context.c | 135 +++++++++++++-------------------- src/gallium/drivers/trace/tr_context.h | 5 ++ src/gallium/drivers/trace/tr_screen.c | 13 +++- src/gallium/drivers/trace/tr_winsys.c | 8 +- 4 files changed, 71 insertions(+), 90 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index 47a217ec7c..c9c15b2bb9 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -40,8 +40,7 @@ trace_context_set_edgeflags(struct pipe_context *_pipe, const unsigned *bitfield) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_edgeflags"); @@ -61,8 +60,7 @@ trace_context_draw_arrays(struct pipe_context *_pipe, unsigned mode, unsigned start, unsigned count) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; boolean result; @@ -90,8 +88,7 @@ trace_context_draw_elements(struct pipe_context *_pipe, unsigned mode, unsigned start, unsigned count) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; boolean result; @@ -125,8 +122,7 @@ trace_context_draw_range_elements(struct pipe_context *_pipe, unsigned count) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; boolean result; @@ -159,8 +155,7 @@ trace_context_create_query(struct pipe_context *_pipe, unsigned query_type) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; struct pipe_query *result; @@ -184,8 +179,7 @@ trace_context_destroy_query(struct pipe_context *_pipe, struct pipe_query *query) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "destroy_query"); @@ -204,8 +198,7 @@ trace_context_begin_query(struct pipe_context *_pipe, struct pipe_query *query) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "begin_query"); @@ -224,8 +217,7 @@ trace_context_end_query(struct pipe_context *_pipe, struct pipe_query *query) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "end_query"); @@ -246,8 +238,7 @@ trace_context_get_query_result(struct pipe_context *_pipe, uint64 *presult) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; uint64 result; boolean _result; @@ -273,8 +264,7 @@ trace_context_create_blend_state(struct pipe_context *_pipe, const struct pipe_blend_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; @@ -296,8 +286,7 @@ trace_context_bind_blend_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "bind_blend_state"); @@ -316,8 +305,7 @@ trace_context_delete_blend_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "delete_blend_state"); @@ -336,8 +324,7 @@ trace_context_create_sampler_state(struct pipe_context *_pipe, const struct pipe_sampler_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; @@ -361,8 +348,7 @@ trace_context_bind_sampler_states(struct pipe_context *_pipe, unsigned num_states, void **states) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "bind_sampler_states"); @@ -382,8 +368,7 @@ trace_context_delete_sampler_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "delete_sampler_state"); @@ -402,8 +387,7 @@ trace_context_create_rasterizer_state(struct pipe_context *_pipe, const struct pipe_rasterizer_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; @@ -427,8 +411,7 @@ trace_context_bind_rasterizer_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "bind_rasterizer_state"); @@ -447,8 +430,7 @@ trace_context_delete_rasterizer_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "delete_rasterizer_state"); @@ -467,8 +449,7 @@ trace_context_create_depth_stencil_alpha_state(struct pipe_context *_pipe, const struct pipe_depth_stencil_alpha_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; @@ -490,8 +471,7 @@ trace_context_bind_depth_stencil_alpha_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "bind_depth_stencil_alpha_state"); @@ -510,8 +490,7 @@ trace_context_delete_depth_stencil_alpha_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "delete_depth_stencil_alpha_state"); @@ -530,8 +509,7 @@ trace_context_create_fs_state(struct pipe_context *_pipe, const struct pipe_shader_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; @@ -555,8 +533,7 @@ trace_context_bind_fs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "bind_fs_state"); @@ -575,8 +552,7 @@ trace_context_delete_fs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "delete_fs_state"); @@ -595,8 +571,7 @@ trace_context_create_vs_state(struct pipe_context *_pipe, const struct pipe_shader_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; @@ -620,8 +595,7 @@ trace_context_bind_vs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "bind_vs_state"); @@ -640,8 +614,7 @@ trace_context_delete_vs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "delete_vs_state"); @@ -660,8 +633,7 @@ trace_context_set_blend_color(struct pipe_context *_pipe, const struct pipe_blend_color *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_blend_color"); @@ -680,8 +652,7 @@ trace_context_set_clip_state(struct pipe_context *_pipe, const struct pipe_clip_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_clip_state"); @@ -701,8 +672,7 @@ trace_context_set_constant_buffer(struct pipe_context *_pipe, const struct pipe_constant_buffer *buffer) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_constant_buffer"); @@ -723,8 +693,7 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe, const struct pipe_framebuffer_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_framebuffer_state"); @@ -743,8 +712,7 @@ trace_context_set_polygon_stipple(struct pipe_context *_pipe, const struct pipe_poly_stipple *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_polygon_stipple"); @@ -763,8 +731,7 @@ trace_context_set_scissor_state(struct pipe_context *_pipe, const struct pipe_scissor_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_scissor_state"); @@ -783,8 +750,7 @@ trace_context_set_viewport_state(struct pipe_context *_pipe, const struct pipe_viewport_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_viewport_state"); @@ -804,8 +770,7 @@ trace_context_set_sampler_textures(struct pipe_context *_pipe, struct pipe_texture **textures) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_sampler_textures"); @@ -826,8 +791,7 @@ trace_context_set_vertex_buffers(struct pipe_context *_pipe, const struct pipe_vertex_buffer *buffers) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_vertex_buffers"); @@ -851,8 +815,7 @@ trace_context_set_vertex_elements(struct pipe_context *_pipe, const struct pipe_vertex_element *elements) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "set_vertex_elements"); @@ -881,8 +844,7 @@ trace_context_surface_copy(struct pipe_context *_pipe, unsigned width, unsigned height) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "surface_copy"); @@ -914,8 +876,7 @@ trace_context_surface_fill(struct pipe_context *_pipe, unsigned value) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "surface_fill"); @@ -939,8 +900,7 @@ trace_context_clear(struct pipe_context *_pipe, unsigned clearValue) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "clear"); @@ -961,8 +921,7 @@ trace_context_flush(struct pipe_context *_pipe, struct pipe_fence_handle **fence) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "flush"); @@ -981,8 +940,7 @@ static INLINE void trace_context_destroy(struct pipe_context *_pipe) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_screen *tr_scr = trace_screen(_pipe->screen); - struct trace_stream *stream = tr_scr->stream; + struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; trace_dump_call_begin(stream, "pipe_context", "destroy"); @@ -1000,6 +958,7 @@ trace_context_destroy(struct pipe_context *_pipe) struct pipe_context * trace_context_create(struct pipe_context *pipe) { + struct trace_stream *stream; struct trace_context *tr_ctx; if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) @@ -1055,10 +1014,18 @@ trace_context_create(struct pipe_context *pipe) tr_ctx->base.flush = trace_context_flush; tr_ctx->pipe = pipe; + tr_ctx->stream = stream = trace_winsys(pipe->winsys)->stream; /* We don't want to trace the internal pipe calls */ pipe->winsys = trace_winsys(pipe->winsys)->winsys; pipe->screen = trace_screen(pipe->screen)->screen; + trace_dump_call_begin(stream, "", "pipe_context_create"); + trace_dump_arg_begin(stream, "screen"); + trace_dump_ptr(stream, pipe->screen); + trace_dump_arg_end(stream); + trace_dump_ret(stream, ptr, pipe); + trace_dump_call_end(stream); + return &tr_ctx->base; } diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h index 80fb5980d6..2c0b0c72e4 100644 --- a/src/gallium/drivers/trace/tr_context.h +++ b/src/gallium/drivers/trace/tr_context.h @@ -34,11 +34,16 @@ #include "pipe/p_context.h" +struct trace_stream; + + struct trace_context { struct pipe_context base; struct pipe_context *pipe; + + struct trace_stream *stream; }; diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index de885abae2..27c218039e 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -346,6 +346,7 @@ trace_screen_destroy(struct pipe_screen *_screen) struct pipe_screen * trace_screen_create(struct pipe_screen *screen) { + struct trace_stream *stream; struct trace_screen *tr_scr; if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) @@ -371,13 +372,17 @@ trace_screen_create(struct pipe_screen *screen) tr_scr->base.surface_unmap = trace_screen_surface_unmap; tr_scr->screen = screen; - - tr_scr->stream = trace_winsys(screen->winsys)->stream; - if(!tr_scr->stream) - return NULL; + tr_scr->stream = stream = trace_winsys(screen->winsys)->stream; /* We don't want to trace the internal pipe calls */ screen->winsys = trace_winsys(screen->winsys)->winsys; + trace_dump_call_begin(stream, "", "pipe_screen_create"); + trace_dump_arg_begin(stream, "winsys"); + trace_dump_ptr(stream, screen->winsys); + trace_dump_arg_end(stream); + trace_dump_ret(stream, ptr, screen); + trace_dump_call_end(stream); + return &tr_scr->base; } diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index 964da5677b..128e502ffc 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -420,6 +420,7 @@ trace_winsys_destroy(struct pipe_winsys *_winsys) struct pipe_winsys * trace_winsys_create(struct pipe_winsys *winsys) { + struct trace_stream *stream; struct trace_winsys *tr_ws; if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) @@ -446,7 +447,7 @@ trace_winsys_create(struct pipe_winsys *winsys) tr_ws->winsys = winsys; - tr_ws->stream = trace_stream_create("gallium", "trace"); + tr_ws->stream = stream = trace_stream_create("gallium", "trace"); if(!tr_ws->stream) return NULL; @@ -455,8 +456,11 @@ trace_winsys_create(struct pipe_winsys *winsys) if(!tr_ws->buffer_maps) return NULL; - trace_dump_trace_begin(tr_ws->stream, 0); + trace_dump_call_begin(stream, "", "pipe_winsys_create"); + trace_dump_ret(stream, ptr, winsys); + trace_dump_call_end(stream); + return &tr_ws->base; } -- cgit v1.2.3 From e5a606883f24980dd8e2378c25e6fb3b8f1937ce Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 10 Aug 2008 18:51:23 +0100 Subject: trace: Fix hexadecimal dumping. --- src/gallium/drivers/trace/tr_dump.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index f545de30f4..5269c4ddc8 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -276,17 +276,16 @@ void trace_dump_bytes(struct trace_stream *stream, const void *data, long unsigned size) { - static char hex_table[] = "0123456789ABCDE"; + static const char hex_table[16] = "0123456789ABCDEF"; const uint8_t *p = data; long unsigned i; trace_dump_write(stream, ""); for(i = 0; i < size; ++i) { uint8_t byte = *p++; - char str[3]; - str[0] = hex_table[byte >> 4]; - str[1] = hex_table[byte & 0xf]; - str[2] = 0; - trace_dump_write(stream, str); + char hex[2]; + hex[0] = hex_table[byte >> 4]; + hex[1] = hex_table[byte & 0xf]; + trace_stream_write(stream, hex, 2); } trace_dump_write(stream, ""); } -- cgit v1.2.3 From a304d271185af935974850055da74ec8263b00e2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 10 Aug 2008 18:52:00 +0100 Subject: python: Allow writing to buffers. --- src/gallium/state_trackers/python/gallium.i | 37 +++++++++++++++++++++------ src/gallium/state_trackers/python/st_device.c | 33 ++++++++++++++++++++++++ src/gallium/state_trackers/python/st_device.h | 14 ++++++++++ 3 files changed, 76 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 284ecb827d..8f731ca905 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -65,8 +65,7 @@ %rename(Context) st_context; %rename(Texture) pipe_texture; %rename(Surface) pipe_surface; - -%rename(Buffer) pipe_buffer; +%rename(Buffer) st_buffer; %rename(BlendColor) pipe_blend_color; %rename(Blend) pipe_blend_state; @@ -88,13 +87,13 @@ %nodefaultctor st_context; %nodefaultctor pipe_texture; %nodefaultctor pipe_surface; -%nodefaultctor pipe_buffer; +%nodefaultctor st_buffer; %nodefaultdtor st_device; %nodefaultdtor st_context; %nodefaultdtor pipe_texture; %nodefaultdtor pipe_surface; -%nodefaultdtor pipe_buffer; +%nodefaultdtor st_buffer; %ignore pipe_texture::screen; @@ -114,6 +113,9 @@ struct st_device { struct st_context { }; +struct st_buffer { +}; + %newobject st_device::texture_create; %newobject st_device::context_create; @@ -197,9 +199,9 @@ struct st_context { return $self->screen->texture_create($self->screen, &templat); } - struct pipe_buffer * + struct st_buffer * buffer_create(unsigned size, unsigned alignment = 0, unsigned usage = 0) { - return $self->screen->winsys->buffer_create($self->screen->winsys, alignment, usage, size); + return st_buffer_create($self, alignment, usage, size); } }; @@ -323,10 +325,10 @@ struct st_context { $self->pipe->draw_arrays($self->pipe, mode, start, count); } - void draw_elements( struct pipe_buffer *indexBuffer, + void draw_elements( struct st_buffer *indexBuffer, unsigned indexSize, unsigned mode, unsigned start, unsigned count) { - $self->pipe->draw_elements($self->pipe, indexBuffer, indexSize, mode, start, count); + $self->pipe->draw_elements($self->pipe, indexBuffer->buffer, indexSize, mode, start, count); } void draw_vertices(unsigned prim, @@ -522,6 +524,25 @@ error1: }; +%extend st_buffer { + + ~st_buffer() { + st_buffer_destroy($self); + } + + void write( const char *STRING, unsigned LENGTH, unsigned offset = 0) { + struct pipe_winsys *winsys = $self->st_dev->screen->winsys; + char *map; + + map = winsys->buffer_map(winsys, $self->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); + if(!map) { + memcpy(map + offset, STRING, LENGTH); + winsys->buffer_unmap(winsys, $self->buffer); + } + } +}; + + %extend pipe_framebuffer_state { pipe_framebuffer_state(void) { diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index 2e53a83eea..fbd56712e0 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -266,3 +266,36 @@ st_context_create(struct st_device *st_dev) return st_ctx; } + + +void +st_buffer_destroy(struct st_buffer *st_buf) +{ + if(st_buf) { + struct pipe_winsys *winsys = st_buf->st_dev->screen->winsys; + pipe_buffer_reference(winsys, &st_buf->buffer, NULL); + FREE(st_buf); + } +} + + +struct st_buffer * +st_buffer_create(struct st_device *st_dev, + unsigned alignment, unsigned usage, unsigned size) +{ + struct pipe_winsys *winsys = st_dev->screen->winsys; + struct st_buffer *st_buf; + + st_buf = CALLOC_STRUCT(st_buffer); + if(!st_buf) + return NULL; + + st_buf->st_dev = st_dev; + + st_buf->buffer = winsys->buffer_create(winsys, alignment, usage, size); + if(!st_buf->buffer) + st_buffer_destroy(st_buf); + + return st_buf; +} + diff --git a/src/gallium/state_trackers/python/st_device.h b/src/gallium/state_trackers/python/st_device.h index 2d95c2da73..5b7adbe1a0 100644 --- a/src/gallium/state_trackers/python/st_device.h +++ b/src/gallium/state_trackers/python/st_device.h @@ -38,6 +38,13 @@ struct pipe_context; struct st_winsys; +struct st_buffer { + struct st_device *st_dev; + + struct pipe_buffer *buffer; +}; + + struct st_context { struct st_device *st_dev; @@ -65,6 +72,13 @@ struct st_device { }; +struct st_buffer * +st_buffer_create(struct st_device *st_dev, + unsigned alignment, unsigned usage, unsigned size); + +void +st_buffer_destroy(struct st_buffer *st_buf); + struct st_context * st_context_create(struct st_device *st_dev); -- cgit v1.2.3 From a318325b516b53fc321669453e4abe53c51f917a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sun, 10 Aug 2008 18:54:10 +0100 Subject: trace: Zero the buffers to avoid dumping uninitialized memory. --- src/gallium/drivers/trace/tr_winsys.c | 19 +++++++++++++++---- 1 file changed, 15 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index 128e502ffc..60a69626c0 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "pipe/p_util.h" +#include "pipe/p_state.h" #include "util/u_hash_table.h" #include "tr_stream.h" @@ -178,7 +179,7 @@ trace_winsys_buffer_create(struct pipe_winsys *_winsys, struct trace_winsys *tr_ws = trace_winsys(_winsys); struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; - struct pipe_buffer *result; + struct pipe_buffer *buffer; trace_dump_call_begin(stream, "pipe_winsys", "buffer_create"); @@ -187,13 +188,23 @@ trace_winsys_buffer_create(struct pipe_winsys *_winsys, trace_dump_arg(stream, uint, usage); trace_dump_arg(stream, uint, size); - result = winsys->buffer_create(winsys, alignment, usage, size); + buffer = winsys->buffer_create(winsys, alignment, usage, size); - trace_dump_ret(stream, ptr, result); + trace_dump_ret(stream, ptr, buffer); trace_dump_call_end(stream); + + /* Zero the buffer to avoid dumping uninitialized memory */ + if(buffer->usage & PIPE_BUFFER_USAGE_CPU_WRITE) { + void *map; + map = winsys->buffer_map(winsys, buffer, PIPE_BUFFER_USAGE_CPU_WRITE); + if(map) { + memset(map, 0, buffer->size); + winsys->buffer_unmap(winsys, buffer); + } + } - return result; + return buffer; } -- cgit v1.2.3 From 8dbb3011a11ff3b6e8ebcf19e64ed4fbef17356b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 00:15:07 +0100 Subject: trace: Replace buffer_map+memcpy+buffer_unmap by buffer_write --- src/gallium/drivers/trace/tr_winsys.c | 39 +++++++---------------------------- 1 file changed, 8 insertions(+), 31 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index 60a69626c0..0a43f1f7f4 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -240,30 +240,18 @@ trace_winsys_buffer_map(struct pipe_winsys *_winsys, unsigned usage) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; - void *result; - - trace_dump_call_begin(stream, "pipe_winsys", "buffer_map"); - - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, buffer); - trace_dump_arg(stream, uint, usage); - - result = winsys->buffer_map(winsys, buffer, usage); - - trace_dump_ret(stream, ptr, result); - - trace_dump_call_end(stream); + void *map; - if(result) { + map = winsys->buffer_map(winsys, buffer, usage); + if(map) { if(usage & PIPE_BUFFER_USAGE_CPU_WRITE) { assert(!hash_table_get(tr_ws->buffer_maps, buffer)); - hash_table_set(tr_ws->buffer_maps, buffer, result); + hash_table_set(tr_ws->buffer_maps, buffer, map); } } - return result; + return map; } @@ -278,13 +266,11 @@ trace_winsys_buffer_unmap(struct pipe_winsys *_winsys, map = hash_table_get(tr_ws->buffer_maps, buffer); if(map) { - trace_dump_call_begin(stream, "", "memcpy"); + trace_dump_call_begin(stream, "pipe_winsys", "buffer_write"); - trace_dump_arg_begin(stream, "dst"); - trace_dump_ptr(stream, map); - trace_dump_arg_end(stream); + trace_dump_arg(stream, ptr, buffer); - trace_dump_arg_begin(stream, "src"); + trace_dump_arg_begin(stream, "data"); trace_dump_bytes(stream, map, buffer->size); trace_dump_arg_end(stream); @@ -294,19 +280,10 @@ trace_winsys_buffer_unmap(struct pipe_winsys *_winsys, trace_dump_call_end(stream); - winsys->buffer_unmap(winsys, buffer); - hash_table_remove(tr_ws->buffer_maps, buffer); } - trace_dump_call_begin(stream, "pipe_winsys", "buffer_unmap"); - - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, buffer); - winsys->buffer_unmap(winsys, buffer); - - trace_dump_call_end(stream); } -- cgit v1.2.3 From b5f7dd0c08dbe3f140630345c331aeccaf2a7a94 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 00:16:20 +0100 Subject: python: Add unsigned arrays. --- src/gallium/state_trackers/python/gallium.i | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 8f731ca905..9d9a50b520 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -58,6 +58,7 @@ %include "carrays.i" %array_class(unsigned char, ByteArray); %array_class(int, IntArray); +%array_class(unsigned, UnsignedArray); %array_class(float, FloatArray); @@ -606,4 +607,4 @@ error1: void dump(unsigned flags = 0) { tgsi_dump($self->tokens, flags); } -} \ No newline at end of file +} -- cgit v1.2.3 From e7f1ac39be5ab8c627f66a4a8488697c274079e6 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 14:06:24 +0100 Subject: python: Split the interface definition file in smaller ones. --- src/gallium/state_trackers/python/gallium.i | 524 +------------------------- src/gallium/state_trackers/python/p_context.i | 232 ++++++++++++ src/gallium/state_trackers/python/p_device.i | 130 +++++++ src/gallium/state_trackers/python/p_state.i | 103 +++++ src/gallium/state_trackers/python/p_texture.i | 192 ++++++++++ 5 files changed, 662 insertions(+), 519 deletions(-) create mode 100644 src/gallium/state_trackers/python/p_context.i create mode 100644 src/gallium/state_trackers/python/p_device.i create mode 100644 src/gallium/state_trackers/python/p_state.i create mode 100644 src/gallium/state_trackers/python/p_texture.i (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 9d9a50b520..4e21eef7be 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -84,527 +84,13 @@ %rename(VertexElement) pipe_vertex_element; %rename(Viewport) pipe_viewport_state; -%nodefaultctor st_device; -%nodefaultctor st_context; -%nodefaultctor pipe_texture; -%nodefaultctor pipe_surface; -%nodefaultctor st_buffer; -%nodefaultdtor st_device; -%nodefaultdtor st_context; -%nodefaultdtor pipe_texture; -%nodefaultdtor pipe_surface; -%nodefaultdtor st_buffer; - -%ignore pipe_texture::screen; - -%ignore pipe_surface::winsys; -%immutable pipe_surface::texture; -%immutable pipe_surface::buffer; - -%include "p_format.i"; %include "pipe/p_defines.h"; -%include "pipe/p_state.h"; %include "pipe/p_shader_tokens.h"; +%include "p_format.i" +%include "p_device.i" +%include "p_context.i" +%include "p_texture.i" +%include "p_state.i" -struct st_device { -}; - -struct st_context { -}; - -struct st_buffer { -}; - - -%newobject st_device::texture_create; -%newobject st_device::context_create; -%newobject st_device::buffer_create; - -%extend st_device { - - st_device(int hardware = 1) { - return st_device_create(hardware ? TRUE : FALSE); - } - - ~st_device() { - st_device_destroy($self); - } - - const char * get_name( void ) { - return $self->screen->get_name($self->screen); - } - - const char * get_vendor( void ) { - return $self->screen->get_vendor($self->screen); - } - - /** - * Query an integer-valued capability/parameter/limit - * \param param one of PIPE_CAP_x - */ - int get_param( int param ) { - return $self->screen->get_param($self->screen, param); - } - - /** - * Query a float-valued capability/parameter/limit - * \param param one of PIPE_CAP_x - */ - float get_paramf( int param ) { - return $self->screen->get_paramf($self->screen, param); - } - - /** - * Check if the given pipe_format is supported as a texture or - * drawing surface. - * \param type one of PIPE_TEXTURE, PIPE_SURFACE - */ - int is_format_supported( enum pipe_format format, - enum pipe_texture_target target, - unsigned tex_usage, - unsigned geom_flags ) { - return $self->screen->is_format_supported( $self->screen, - format, - target, - tex_usage, - geom_flags ); - } - - struct st_context * - context_create(void) { - return st_context_create($self); - } - - struct pipe_texture * - texture_create( - enum pipe_format format, - unsigned width, - unsigned height, - unsigned depth = 1, - unsigned last_level = 0, - enum pipe_texture_target target = PIPE_TEXTURE_2D, - unsigned tex_usage = 0 - ) { - struct pipe_texture templat; - memset(&templat, 0, sizeof(templat)); - templat.format = format; - pf_get_block(templat.format, &templat.block); - templat.width[0] = width; - templat.height[0] = height; - templat.depth[0] = depth; - templat.last_level = last_level; - templat.target = target; - templat.tex_usage = tex_usage; - return $self->screen->texture_create($self->screen, &templat); - } - - struct st_buffer * - buffer_create(unsigned size, unsigned alignment = 0, unsigned usage = 0) { - return st_buffer_create($self, alignment, usage, size); - } - -}; - - -%extend st_context { - - ~st_context() { - st_context_destroy($self); - } - - /* - * State functions (create/bind/destroy state objects) - */ - - void set_blend( const struct pipe_blend_state *state ) { - cso_set_blend($self->cso, state); - } - - void set_sampler( unsigned index, const struct pipe_sampler_state *state ) { - cso_single_sampler($self->cso, index, state); - cso_single_sampler_done($self->cso); - } - - void set_rasterizer( const struct pipe_rasterizer_state *state ) { - cso_set_rasterizer($self->cso, state); - } - - void set_depth_stencil_alpha(const struct pipe_depth_stencil_alpha_state *state) { - cso_set_depth_stencil_alpha($self->cso, state); - } - - void set_fragment_shader( const struct pipe_shader_state *state ) { - void *fs; - - fs = $self->pipe->create_fs_state($self->pipe, state); - if(!fs) - return; - - if(cso_set_fragment_shader_handle($self->cso, fs) != PIPE_OK) - return; - - cso_delete_fragment_shader($self->cso, $self->fs); - $self->fs = fs; - } - - void set_vertex_shader( const struct pipe_shader_state *state ) { - void *vs; - - vs = $self->pipe->create_vs_state($self->pipe, state); - if(!vs) - return; - - if(cso_set_vertex_shader_handle($self->cso, vs) != PIPE_OK) - return; - - cso_delete_vertex_shader($self->cso, $self->vs); - $self->vs = vs; - } - - /* - * Parameter-like state (or properties) - */ - - void set_blend_color(const struct pipe_blend_color *state ) { - cso_set_blend_color($self->cso, state); - } - - void set_clip(const struct pipe_clip_state *state ) { - $self->pipe->set_clip_state($self->pipe, state); - } - - void set_constant_buffer(unsigned shader, unsigned index, - const struct pipe_constant_buffer *buf ) { - $self->pipe->set_constant_buffer($self->pipe, shader, index, buf); - } - - void set_framebuffer(const struct pipe_framebuffer_state *state ) { - cso_set_framebuffer($self->cso, state); - } - - void set_polygon_stipple(const struct pipe_poly_stipple *state ) { - $self->pipe->set_polygon_stipple($self->pipe, state); - } - - void set_scissor(const struct pipe_scissor_state *state ) { - $self->pipe->set_scissor_state($self->pipe, state); - } - - void set_viewport(const struct pipe_viewport_state *state) { - cso_set_viewport($self->cso, state); - } - - void set_sampler_texture(unsigned index, - struct pipe_texture *texture) { - if(!texture) - texture = $self->default_texture; - pipe_texture_reference(&$self->sampler_textures[index], texture); - $self->pipe->set_sampler_textures($self->pipe, - PIPE_MAX_SAMPLERS, - $self->sampler_textures); - } - - void set_vertex_buffer(unsigned index, - const struct pipe_vertex_buffer *buffer) { - memcpy(&$self->vertex_buffers[index], buffer, sizeof(*buffer)); - $self->pipe->set_vertex_buffers($self->pipe, PIPE_MAX_ATTRIBS, $self->vertex_buffers); - } - - void set_vertex_element(unsigned index, - const struct pipe_vertex_element *element) { - memcpy(&$self->vertex_elements[index], element, sizeof(*element)); - $self->pipe->set_vertex_elements($self->pipe, PIPE_MAX_ATTRIBS, $self->vertex_elements); - } - - /* - * Draw functions - */ - - void draw_arrays(unsigned mode, unsigned start, unsigned count) { - $self->pipe->draw_arrays($self->pipe, mode, start, count); - } - - void draw_elements( struct st_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count) { - $self->pipe->draw_elements($self->pipe, indexBuffer->buffer, indexSize, mode, start, count); - } - - void draw_vertices(unsigned prim, - unsigned num_verts, - unsigned num_attribs, - const float *vertices) - { - struct pipe_context *pipe = $self->pipe; - struct pipe_winsys *winsys = pipe->winsys; - struct pipe_buffer *vbuf; - float *map; - unsigned size; - - size = num_verts * num_attribs * 4 * sizeof(float); - - vbuf = winsys->buffer_create(winsys, - 32, - PIPE_BUFFER_USAGE_VERTEX, - size); - if(!vbuf) - goto error1; - - map = winsys->buffer_map(winsys, vbuf, PIPE_BUFFER_USAGE_CPU_WRITE); - if (!map) - goto error2; - memcpy(map, vertices, size); - pipe->winsys->buffer_unmap(pipe->winsys, vbuf); - - util_draw_vertex_buffer(pipe, vbuf, prim, num_verts, num_attribs); - -error2: - pipe_buffer_reference(pipe->winsys, &vbuf, NULL); -error1: - ; - } - - void - flush(void) { - struct pipe_fence_handle *fence = NULL; - $self->pipe->flush($self->pipe, PIPE_FLUSH_RENDER_CACHE, &fence); - /* TODO: allow asynchronous operation */ - $self->pipe->winsys->fence_finish( $self->pipe->winsys, fence, 0 ); - $self->pipe->winsys->fence_reference( $self->pipe->winsys, &fence, NULL ); - } - - /* - * Surface functions - */ - - void surface_copy(int do_flip, - struct pipe_surface *dest, - unsigned destx, unsigned desty, - struct pipe_surface *src, - unsigned srcx, unsigned srcy, - unsigned width, unsigned height) { - $self->pipe->surface_copy($self->pipe, do_flip, dest, destx, desty, src, srcx, srcy, width, height); - } - - void surface_fill(struct pipe_surface *dst, - unsigned x, unsigned y, - unsigned width, unsigned height, - unsigned value) { - $self->pipe->surface_fill($self->pipe, dst, x, y, width, height, value); - } - - void surface_clear(struct pipe_surface *surface, unsigned value = 0) { - $self->pipe->clear($self->pipe, surface, value); - } - -}; - - -%newobject pipe_texture::get_surface; - -%extend pipe_texture { - - ~pipe_texture() { - struct pipe_texture *ptr = $self; - pipe_texture_reference(&ptr, NULL); - } - - unsigned get_width(unsigned level=0) { - return $self->width[level]; - } - - unsigned get_height(unsigned level=0) { - return $self->height[level]; - } - - unsigned get_depth(unsigned level=0) { - return $self->depth[level]; - } - - unsigned get_nblocksx(unsigned level=0) { - return $self->nblocksx[level]; - } - - unsigned get_nblocksy(unsigned level=0) { - return $self->nblocksy[level]; - } - - /** Get a surface which is a "view" into a texture */ - struct pipe_surface * - get_surface(unsigned face=0, unsigned level=0, unsigned zslice=0, unsigned usage=0 ) - { - struct pipe_screen *screen = $self->screen; - return screen->get_tex_surface(screen, $self, face, level, zslice, usage); - } - -}; - - -%extend pipe_surface { - - ~pipe_surface() { - struct pipe_surface *ptr = $self; - pipe_surface_reference(&ptr, NULL); - } - - // gets mapped to pipe_surface_map automatically - void * map( unsigned flags ); - - // gets mapped to pipe_surface_unmap automatically - void unmap( void ); - - void - get_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, unsigned char *raw, unsigned stride) { - pipe_get_tile_raw($self, x, y, w, h, raw, stride); - } - - void - put_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned char *raw, unsigned stride) { - pipe_put_tile_raw($self, x, y, w, h, raw, stride); - } - - void - get_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, float *rgba) { - pipe_get_tile_rgba($self, x, y, w, h, rgba); - } - - void - put_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba) { - pipe_put_tile_rgba($self, x, y, w, h, rgba); - } - - void - get_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, unsigned *z) { - pipe_get_tile_z($self, x, y, w, h, z); - } - - void - put_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned *z) { - pipe_put_tile_z($self, x, y, w, h, z); - } - - void - sample_rgba(float *rgba) { - st_sample_surface($self, rgba); - } - - unsigned - compare_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba, float tol = 0.0) - { - float *rgba2; - const float *p1; - const float *p2; - unsigned i, j, n; - - rgba2 = MALLOC(h*w*4*sizeof(float)); - if(!rgba2) - return ~0; - - pipe_get_tile_rgba($self, x, y, w, h, rgba2); - - p1 = rgba; - p2 = rgba2; - n = 0; - for(i = h*w; i; --i) { - unsigned differs = 0; - for(j = 4; j; --j) { - float delta = *p2++ - *p1++; - if (delta < -tol || delta > tol) - differs = 1; - } - n += differs; - } - - FREE(rgba2); - - return n; - } - -}; - - -%extend st_buffer { - - ~st_buffer() { - st_buffer_destroy($self); - } - - void write( const char *STRING, unsigned LENGTH, unsigned offset = 0) { - struct pipe_winsys *winsys = $self->st_dev->screen->winsys; - char *map; - - map = winsys->buffer_map(winsys, $self->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); - if(!map) { - memcpy(map + offset, STRING, LENGTH); - winsys->buffer_unmap(winsys, $self->buffer); - } - } -}; - - -%extend pipe_framebuffer_state { - - pipe_framebuffer_state(void) { - return CALLOC_STRUCT(pipe_framebuffer_state); - } - - ~pipe_framebuffer_state() { - unsigned index; - for(index = 0; index < PIPE_MAX_COLOR_BUFS; ++index) - pipe_surface_reference(&$self->cbufs[index], NULL); - pipe_surface_reference(&$self->zsbuf, NULL); - FREE($self); - } - - void - set_cbuf(unsigned index, struct pipe_surface *surface) { - pipe_surface_reference(&$self->cbufs[index], surface); - } - - void - set_zsbuf(struct pipe_surface *surface) { - pipe_surface_reference(&$self->zsbuf, surface); - } - -}; - - -%extend pipe_shader_state { - - pipe_shader_state(const char *text, unsigned num_tokens = 1024) { - struct tgsi_token *tokens; - struct pipe_shader_state *shader; - - tokens = MALLOC(num_tokens * sizeof(struct tgsi_token)); - if(!tokens) - goto error1; - - if(tgsi_text_translate(text, tokens, num_tokens ) != TRUE) - goto error2; - - shader = CALLOC_STRUCT(pipe_shader_state); - if(!shader) - goto error3; - - shader->tokens = tokens; - - return shader; - -error3: -error2: - FREE(tokens); -error1: - return NULL; - } - - ~pipe_shader_state() { - FREE((void*)$self->tokens); - FREE($self); - } - - void dump(unsigned flags = 0) { - tgsi_dump($self->tokens, flags); - } -} diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i new file mode 100644 index 0000000000..172dc7256c --- /dev/null +++ b/src/gallium/state_trackers/python/p_context.i @@ -0,0 +1,232 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca + */ + +%nodefaultctor st_context; +%nodefaultdtor st_context; + +struct st_context { +}; + +%extend st_context { + + ~st_context() { + st_context_destroy($self); + } + + /* + * State functions (create/bind/destroy state objects) + */ + + void set_blend( const struct pipe_blend_state *state ) { + cso_set_blend($self->cso, state); + } + + void set_sampler( unsigned index, const struct pipe_sampler_state *state ) { + cso_single_sampler($self->cso, index, state); + cso_single_sampler_done($self->cso); + } + + void set_rasterizer( const struct pipe_rasterizer_state *state ) { + cso_set_rasterizer($self->cso, state); + } + + void set_depth_stencil_alpha(const struct pipe_depth_stencil_alpha_state *state) { + cso_set_depth_stencil_alpha($self->cso, state); + } + + void set_fragment_shader( const struct pipe_shader_state *state ) { + void *fs; + + fs = $self->pipe->create_fs_state($self->pipe, state); + if(!fs) + return; + + if(cso_set_fragment_shader_handle($self->cso, fs) != PIPE_OK) + return; + + cso_delete_fragment_shader($self->cso, $self->fs); + $self->fs = fs; + } + + void set_vertex_shader( const struct pipe_shader_state *state ) { + void *vs; + + vs = $self->pipe->create_vs_state($self->pipe, state); + if(!vs) + return; + + if(cso_set_vertex_shader_handle($self->cso, vs) != PIPE_OK) + return; + + cso_delete_vertex_shader($self->cso, $self->vs); + $self->vs = vs; + } + + /* + * Parameter-like state (or properties) + */ + + void set_blend_color(const struct pipe_blend_color *state ) { + cso_set_blend_color($self->cso, state); + } + + void set_clip(const struct pipe_clip_state *state ) { + $self->pipe->set_clip_state($self->pipe, state); + } + + void set_constant_buffer(unsigned shader, unsigned index, + const struct pipe_constant_buffer *buf ) { + $self->pipe->set_constant_buffer($self->pipe, shader, index, buf); + } + + void set_framebuffer(const struct pipe_framebuffer_state *state ) { + cso_set_framebuffer($self->cso, state); + } + + void set_polygon_stipple(const struct pipe_poly_stipple *state ) { + $self->pipe->set_polygon_stipple($self->pipe, state); + } + + void set_scissor(const struct pipe_scissor_state *state ) { + $self->pipe->set_scissor_state($self->pipe, state); + } + + void set_viewport(const struct pipe_viewport_state *state) { + cso_set_viewport($self->cso, state); + } + + void set_sampler_texture(unsigned index, + struct pipe_texture *texture) { + if(!texture) + texture = $self->default_texture; + pipe_texture_reference(&$self->sampler_textures[index], texture); + $self->pipe->set_sampler_textures($self->pipe, + PIPE_MAX_SAMPLERS, + $self->sampler_textures); + } + + void set_vertex_buffer(unsigned index, + const struct pipe_vertex_buffer *buffer) { + memcpy(&$self->vertex_buffers[index], buffer, sizeof(*buffer)); + $self->pipe->set_vertex_buffers($self->pipe, PIPE_MAX_ATTRIBS, $self->vertex_buffers); + } + + void set_vertex_element(unsigned index, + const struct pipe_vertex_element *element) { + memcpy(&$self->vertex_elements[index], element, sizeof(*element)); + $self->pipe->set_vertex_elements($self->pipe, PIPE_MAX_ATTRIBS, $self->vertex_elements); + } + + /* + * Draw functions + */ + + void draw_arrays(unsigned mode, unsigned start, unsigned count) { + $self->pipe->draw_arrays($self->pipe, mode, start, count); + } + + void draw_elements( struct st_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) { + $self->pipe->draw_elements($self->pipe, indexBuffer->buffer, indexSize, mode, start, count); + } + + void draw_vertices(unsigned prim, + unsigned num_verts, + unsigned num_attribs, + const float *vertices) + { + struct pipe_context *pipe = $self->pipe; + struct pipe_winsys *winsys = pipe->winsys; + struct pipe_buffer *vbuf; + float *map; + unsigned size; + + size = num_verts * num_attribs * 4 * sizeof(float); + + vbuf = winsys->buffer_create(winsys, + 32, + PIPE_BUFFER_USAGE_VERTEX, + size); + if(!vbuf) + goto error1; + + map = winsys->buffer_map(winsys, vbuf, PIPE_BUFFER_USAGE_CPU_WRITE); + if (!map) + goto error2; + memcpy(map, vertices, size); + pipe->winsys->buffer_unmap(pipe->winsys, vbuf); + + util_draw_vertex_buffer(pipe, vbuf, prim, num_verts, num_attribs); + +error2: + pipe_buffer_reference(pipe->winsys, &vbuf, NULL); +error1: + ; + } + + void + flush(void) { + struct pipe_fence_handle *fence = NULL; + $self->pipe->flush($self->pipe, PIPE_FLUSH_RENDER_CACHE, &fence); + /* TODO: allow asynchronous operation */ + $self->pipe->winsys->fence_finish( $self->pipe->winsys, fence, 0 ); + $self->pipe->winsys->fence_reference( $self->pipe->winsys, &fence, NULL ); + } + + /* + * Surface functions + */ + + void surface_copy(int do_flip, + struct pipe_surface *dest, + unsigned destx, unsigned desty, + struct pipe_surface *src, + unsigned srcx, unsigned srcy, + unsigned width, unsigned height) { + $self->pipe->surface_copy($self->pipe, do_flip, dest, destx, desty, src, srcx, srcy, width, height); + } + + void surface_fill(struct pipe_surface *dst, + unsigned x, unsigned y, + unsigned width, unsigned height, + unsigned value) { + $self->pipe->surface_fill($self->pipe, dst, x, y, width, height, value); + } + + void surface_clear(struct pipe_surface *surface, unsigned value = 0) { + $self->pipe->clear($self->pipe, surface, value); + } + +}; diff --git a/src/gallium/state_trackers/python/p_device.i b/src/gallium/state_trackers/python/p_device.i new file mode 100644 index 0000000000..84fd2e4349 --- /dev/null +++ b/src/gallium/state_trackers/python/p_device.i @@ -0,0 +1,130 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca + */ + + +%nodefaultctor st_device; +%nodefaultdtor st_device; + + +struct st_device { +}; + +%newobject st_device::texture_create; +%newobject st_device::context_create; +%newobject st_device::buffer_create; + +%extend st_device { + + st_device(int hardware = 1) { + return st_device_create(hardware ? TRUE : FALSE); + } + + ~st_device() { + st_device_destroy($self); + } + + const char * get_name( void ) { + return $self->screen->get_name($self->screen); + } + + const char * get_vendor( void ) { + return $self->screen->get_vendor($self->screen); + } + + /** + * Query an integer-valued capability/parameter/limit + * \param param one of PIPE_CAP_x + */ + int get_param( int param ) { + return $self->screen->get_param($self->screen, param); + } + + /** + * Query a float-valued capability/parameter/limit + * \param param one of PIPE_CAP_x + */ + float get_paramf( int param ) { + return $self->screen->get_paramf($self->screen, param); + } + + /** + * Check if the given pipe_format is supported as a texture or + * drawing surface. + * \param type one of PIPE_TEXTURE, PIPE_SURFACE + */ + int is_format_supported( enum pipe_format format, + enum pipe_texture_target target, + unsigned tex_usage, + unsigned geom_flags ) { + return $self->screen->is_format_supported( $self->screen, + format, + target, + tex_usage, + geom_flags ); + } + + struct st_context * + context_create(void) { + return st_context_create($self); + } + + struct pipe_texture * + texture_create( + enum pipe_format format, + unsigned width, + unsigned height, + unsigned depth = 1, + unsigned last_level = 0, + enum pipe_texture_target target = PIPE_TEXTURE_2D, + unsigned tex_usage = 0 + ) { + struct pipe_texture templat; + memset(&templat, 0, sizeof(templat)); + templat.format = format; + pf_get_block(templat.format, &templat.block); + templat.width[0] = width; + templat.height[0] = height; + templat.depth[0] = depth; + templat.last_level = last_level; + templat.target = target; + templat.tex_usage = tex_usage; + return $self->screen->texture_create($self->screen, &templat); + } + + struct st_buffer * + buffer_create(unsigned size, unsigned alignment = 0, unsigned usage = 0) { + return st_buffer_create($self, alignment, usage, size); + } + +}; diff --git a/src/gallium/state_trackers/python/p_state.i b/src/gallium/state_trackers/python/p_state.i new file mode 100644 index 0000000000..9834229b4b --- /dev/null +++ b/src/gallium/state_trackers/python/p_state.i @@ -0,0 +1,103 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca + */ + +%module gallium; + +%include "pipe/p_state.h"; + + +%extend pipe_framebuffer_state { + + pipe_framebuffer_state(void) { + return CALLOC_STRUCT(pipe_framebuffer_state); + } + + ~pipe_framebuffer_state() { + unsigned index; + for(index = 0; index < PIPE_MAX_COLOR_BUFS; ++index) + pipe_surface_reference(&$self->cbufs[index], NULL); + pipe_surface_reference(&$self->zsbuf, NULL); + FREE($self); + } + + void + set_cbuf(unsigned index, struct pipe_surface *surface) { + pipe_surface_reference(&$self->cbufs[index], surface); + } + + void + set_zsbuf(struct pipe_surface *surface) { + pipe_surface_reference(&$self->zsbuf, surface); + } + +}; + + +%extend pipe_shader_state { + + pipe_shader_state(const char *text, unsigned num_tokens = 1024) { + struct tgsi_token *tokens; + struct pipe_shader_state *shader; + + tokens = MALLOC(num_tokens * sizeof(struct tgsi_token)); + if(!tokens) + goto error1; + + if(tgsi_text_translate(text, tokens, num_tokens ) != TRUE) + goto error2; + + shader = CALLOC_STRUCT(pipe_shader_state); + if(!shader) + goto error3; + + shader->tokens = tokens; + + return shader; + +error3: +error2: + FREE(tokens); +error1: + return NULL; + } + + ~pipe_shader_state() { + FREE((void*)$self->tokens); + FREE($self); + } + + void dump(unsigned flags = 0) { + tgsi_dump($self->tokens, flags); + } +} diff --git a/src/gallium/state_trackers/python/p_texture.i b/src/gallium/state_trackers/python/p_texture.i new file mode 100644 index 0000000000..fd38e0733f --- /dev/null +++ b/src/gallium/state_trackers/python/p_texture.i @@ -0,0 +1,192 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * SWIG interface definion for Gallium types. + * + * @author Jose Fonseca + */ + + +%nodefaultctor pipe_texture; +%nodefaultctor pipe_surface; +%nodefaultctor st_buffer; + +%nodefaultdtor pipe_texture; +%nodefaultdtor pipe_surface; +%nodefaultdtor st_buffer; + +%ignore pipe_texture::screen; + +%ignore pipe_surface::winsys; +%immutable pipe_surface::texture; +%immutable pipe_surface::buffer; + +%newobject pipe_texture::get_surface; + + +%extend pipe_texture { + + ~pipe_texture() { + struct pipe_texture *ptr = $self; + pipe_texture_reference(&ptr, NULL); + } + + unsigned get_width(unsigned level=0) { + return $self->width[level]; + } + + unsigned get_height(unsigned level=0) { + return $self->height[level]; + } + + unsigned get_depth(unsigned level=0) { + return $self->depth[level]; + } + + unsigned get_nblocksx(unsigned level=0) { + return $self->nblocksx[level]; + } + + unsigned get_nblocksy(unsigned level=0) { + return $self->nblocksy[level]; + } + + /** Get a surface which is a "view" into a texture */ + struct pipe_surface * + get_surface(unsigned face=0, unsigned level=0, unsigned zslice=0, unsigned usage=0 ) + { + struct pipe_screen *screen = $self->screen; + return screen->get_tex_surface(screen, $self, face, level, zslice, usage); + } + +}; + + +%extend pipe_surface { + + ~pipe_surface() { + struct pipe_surface *ptr = $self; + pipe_surface_reference(&ptr, NULL); + } + + // gets mapped to pipe_surface_map automatically + void * map( unsigned flags ); + + // gets mapped to pipe_surface_unmap automatically + void unmap( void ); + + void + get_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, unsigned char *raw, unsigned stride) { + pipe_get_tile_raw($self, x, y, w, h, raw, stride); + } + + void + put_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned char *raw, unsigned stride) { + pipe_put_tile_raw($self, x, y, w, h, raw, stride); + } + + void + get_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, float *rgba) { + pipe_get_tile_rgba($self, x, y, w, h, rgba); + } + + void + put_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba) { + pipe_put_tile_rgba($self, x, y, w, h, rgba); + } + + void + get_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, unsigned *z) { + pipe_get_tile_z($self, x, y, w, h, z); + } + + void + put_tile_z(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned *z) { + pipe_put_tile_z($self, x, y, w, h, z); + } + + void + sample_rgba(float *rgba) { + st_sample_surface($self, rgba); + } + + unsigned + compare_tile_rgba(unsigned x, unsigned y, unsigned w, unsigned h, const float *rgba, float tol = 0.0) + { + float *rgba2; + const float *p1; + const float *p2; + unsigned i, j, n; + + rgba2 = MALLOC(h*w*4*sizeof(float)); + if(!rgba2) + return ~0; + + pipe_get_tile_rgba($self, x, y, w, h, rgba2); + + p1 = rgba; + p2 = rgba2; + n = 0; + for(i = h*w; i; --i) { + unsigned differs = 0; + for(j = 4; j; --j) { + float delta = *p2++ - *p1++; + if (delta < -tol || delta > tol) + differs = 1; + } + n += differs; + } + + FREE(rgba2); + + return n; + } + +}; + +struct st_buffer { +}; + +%extend st_buffer { + + ~st_buffer() { + st_buffer_destroy($self); + } + + void write( const char *STRING, unsigned LENGTH, unsigned offset = 0) { + struct pipe_winsys *winsys = $self->st_dev->screen->winsys; + char *map; + + map = winsys->buffer_map(winsys, $self->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); + if(!map) { + memcpy(map + offset, STRING, LENGTH); + winsys->buffer_unmap(winsys, $self->buffer); + } + } +}; -- cgit v1.2.3 From ae0c9b56d4bec52f9accabbcaf8d42ef41a0153c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 14:31:04 +0100 Subject: python: Simplify setting the constant buffers. --- src/gallium/state_trackers/python/p_context.i | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i index 172dc7256c..df07f3bc47 100644 --- a/src/gallium/state_trackers/python/p_context.i +++ b/src/gallium/state_trackers/python/p_context.i @@ -106,8 +106,13 @@ struct st_context { } void set_constant_buffer(unsigned shader, unsigned index, - const struct pipe_constant_buffer *buf ) { - $self->pipe->set_constant_buffer($self->pipe, shader, index, buf); + const struct pipe_constant_buffer *buf ) + { + struct pipe_constant_buffer state; + memset(&state, 0, sizeof(state)); + state.buffer = buf->buffer; + state.size = buf->buffer->size; + $self->pipe->set_constant_buffer($self->pipe, shader, index, &state); } void set_framebuffer(const struct pipe_framebuffer_state *state ) { -- cgit v1.2.3 From ca826d79a64d46e7f3bffff22640f148c044cdad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 16:55:30 +0100 Subject: gallium: Name pipe_depth_stencil_alpha_state member structures. So that the previously anonymous depth/stencil/alpha structures can be identified in the traces. This is just syntactic sugar: it does not break source or binary compatibility. --- src/gallium/include/pipe/p_state.h | 51 ++++++++++++++++++++++---------------- 1 file changed, 30 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h index 2401305eb7..4d0f3597e6 100644 --- a/src/gallium/include/pipe/p_state.h +++ b/src/gallium/include/pipe/p_state.h @@ -170,29 +170,38 @@ struct pipe_shader_state }; +struct pipe_depth_state { + unsigned enabled:1; /**< depth test enabled? */ + unsigned writemask:1; /**< allow depth buffer writes? */ + unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ + unsigned occlusion_count:1; /**< do occlusion counting? */ +}; + + +struct pipe_stencil_state { + unsigned enabled:1; /**< stencil[0]: stencil enabled, stencil[1]: two-side enabled */ + unsigned func:3; /**< PIPE_FUNC_x */ + unsigned fail_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned zpass_op:3; /**< PIPE_STENCIL_OP_x */ + unsigned zfail_op:3; /**< PIPE_STENCIL_OP_x */ + ubyte ref_value; + ubyte value_mask; + ubyte write_mask; +}; + + +struct pipe_alpha_state { + unsigned enabled:1; + unsigned func:3; /**< PIPE_FUNC_x */ + float ref; /**< reference value */ +}; + + struct pipe_depth_stencil_alpha_state { - struct { - unsigned enabled:1; /**< depth test enabled? */ - unsigned writemask:1; /**< allow depth buffer writes? */ - unsigned func:3; /**< depth test func (PIPE_FUNC_x) */ - unsigned occlusion_count:1; /**< do occlusion counting? */ - } depth; - struct { - unsigned enabled:1; /**< stencil[0]: stencil enabled, stencil[1]: two-side enabled */ - unsigned func:3; /**< PIPE_FUNC_x */ - unsigned fail_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned zpass_op:3; /**< PIPE_STENCIL_OP_x */ - unsigned zfail_op:3; /**< PIPE_STENCIL_OP_x */ - ubyte ref_value; - ubyte value_mask; - ubyte write_mask; - } stencil[2]; /**< [0] = front, [1] = back */ - struct { - unsigned enabled:1; - unsigned func:3; /**< PIPE_FUNC_x */ - float ref; /**< reference value */ - } alpha; + struct pipe_depth_state depth; + struct pipe_stencil_state stencil[2]; /**< [0] = front, [1] = back */ + struct pipe_alpha_state alpha; }; -- cgit v1.2.3 From 166b939d523a8683f1aee819f73e002e627a49ba Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 16:56:08 +0100 Subject: trace: Dump pipe_{depth,stencil,alpha}_state names. --- src/gallium/drivers/trace/tr_state.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index 3c7b007bfe..961705b9ee 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -277,7 +277,7 @@ void trace_dump_depth_stencil_alpha_state(struct trace_stream *stream, trace_dump_struct_begin(stream, "pipe_depth_stencil_alpha_state"); trace_dump_member_begin(stream, "depth"); - trace_dump_struct_begin(stream, ""); + trace_dump_struct_begin(stream, "pipe_depth_state"); trace_dump_member(stream, bool, &state->depth, enabled); trace_dump_member(stream, bool, &state->depth, writemask); trace_dump_member(stream, uint, &state->depth, func); @@ -289,7 +289,7 @@ void trace_dump_depth_stencil_alpha_state(struct trace_stream *stream, trace_dump_array_begin(stream); for(i = 0; i < Elements(state->stencil); ++i) { trace_dump_elem_begin(stream); - trace_dump_struct_begin(stream, ""); + trace_dump_struct_begin(stream, "pipe_stencil_state"); trace_dump_member(stream, bool, &state->stencil[i], enabled); trace_dump_member(stream, uint, &state->stencil[i], func); trace_dump_member(stream, uint, &state->stencil[i], fail_op); @@ -305,7 +305,7 @@ void trace_dump_depth_stencil_alpha_state(struct trace_stream *stream, trace_dump_member_end(stream); trace_dump_member_begin(stream, "alpha"); - trace_dump_struct_begin(stream, ""); + trace_dump_struct_begin(stream, "pipe_alpha_state"); trace_dump_member(stream, bool, &state->alpha, enabled); trace_dump_member(stream, uint, &state->alpha, func); trace_dump_member(stream, float, &state->alpha, ref); -- cgit v1.2.3 From e69e94d3010b33cf690a2e4e54b61437a1836617 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 16:57:50 +0100 Subject: python: Bindings for pipe_{depth,stencil,alpha}_state. --- src/gallium/state_trackers/python/gallium.i | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 4e21eef7be..1792f63c3c 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -72,6 +72,9 @@ %rename(Blend) pipe_blend_state; %rename(Clip) pipe_clip_state; %rename(ConstantBuffer) pipe_constant_buffer; +%rename(Depth) pipe_depth_state; +%rename(Stencil) pipe_stencil_state; +%rename(Alpha) pipe_alpha_state; %rename(DepthStencilAlpha) pipe_depth_stencil_alpha_state; %rename(FormatBlock) pipe_format_block; %rename(Framebuffer) pipe_framebuffer_state; @@ -86,7 +89,6 @@ %include "pipe/p_defines.h"; -%include "pipe/p_shader_tokens.h"; %include "p_format.i" %include "p_device.i" -- cgit v1.2.3 From d4d8f803884584525a36713b76ce04802658bd0f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 17:08:27 +0100 Subject: trace: Fix create_depth_stencil_alpha_state trace. --- src/gallium/drivers/trace/tr_context.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index c9c15b2bb9..1df6842b71 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -457,8 +457,10 @@ trace_context_create_depth_stencil_alpha_state(struct pipe_context *_pipe, result = pipe->create_depth_stencil_alpha_state(pipe, state);; - trace_dump_ret(stream, ptr, result); + trace_dump_arg(stream, ptr, pipe); trace_dump_arg(stream, depth_stencil_alpha_state, state); + + trace_dump_ret(stream, ptr, result); trace_dump_call_end(stream); -- cgit v1.2.3 From e08072c5db137e3fc19964ea8705c62cb00b6343 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 17:08:53 +0100 Subject: python: Bindings for p_compiler.h types. --- src/gallium/state_trackers/python/gallium.i | 1 + src/gallium/state_trackers/python/p_compiler.i | 29 ++++++++++++++++++++++++++ 2 files changed, 30 insertions(+) create mode 100644 src/gallium/state_trackers/python/p_compiler.i (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 1792f63c3c..8d7a49cee7 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -88,6 +88,7 @@ %rename(Viewport) pipe_viewport_state; +%include "p_compiler.i" %include "pipe/p_defines.h"; %include "p_format.i" diff --git a/src/gallium/state_trackers/python/p_compiler.i b/src/gallium/state_trackers/python/p_compiler.i new file mode 100644 index 0000000000..15f6ba5b9d --- /dev/null +++ b/src/gallium/state_trackers/python/p_compiler.i @@ -0,0 +1,29 @@ + /************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +typedef unsigned char ubyte; -- cgit v1.2.3 From e54fa77d130582ee48b699917324040ef254ed74 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 18:09:36 +0100 Subject: trace: Dump polygon stipple state as an array. --- src/gallium/drivers/trace/tr_state.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index 961705b9ee..e074ae7abc 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -150,9 +150,10 @@ void trace_dump_poly_stipple(struct trace_stream *stream, trace_dump_struct_begin(stream, "pipe_poly_stipple"); trace_dump_member_begin(stream, "stipple"); - trace_dump_bytes(stream, + trace_dump_array(stream, + uint, state->stipple, - sizeof(state->stipple)); + Elements(state->stipple)); trace_dump_member_end(stream); trace_dump_struct_end(stream); -- cgit v1.2.3 From 7b8fa937eb9bbc63a59f2db399781c32b20f339f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 11 Aug 2008 18:29:04 +0100 Subject: trace: Fix create_blend_state dump. --- src/gallium/drivers/trace/tr_context.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index 1df6842b71..3e098128c9 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -275,6 +275,8 @@ trace_context_create_blend_state(struct pipe_context *_pipe, result = pipe->create_blend_state(pipe, state);; + trace_dump_ret(stream, ptr, result); + trace_dump_call_end(stream); return result; -- cgit v1.2.3 From 4a77fadd102264f1bf2caec3263e193eb831dade Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 12 Aug 2008 11:32:42 +0100 Subject: trace: More dump fixes. --- src/gallium/drivers/trace/tr_context.c | 1 - src/gallium/drivers/trace/tr_winsys.c | 2 ++ 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index 3e098128c9..f82126370c 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -826,7 +826,6 @@ trace_context_set_vertex_elements(struct pipe_context *_pipe, trace_dump_arg(stream, ptr, pipe); trace_dump_arg(stream, uint, num_elements); - trace_dump_arg(stream, ptr, elements); trace_dump_arg_begin(stream, "elements"); trace_dump_struct_array(stream, vertex_element, elements, num_elements); diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index 0a43f1f7f4..eec84a1981 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -268,6 +268,8 @@ trace_winsys_buffer_unmap(struct pipe_winsys *_winsys, if(map) { trace_dump_call_begin(stream, "pipe_winsys", "buffer_write"); + trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(stream, ptr, buffer); trace_dump_arg_begin(stream, "data"); -- cgit v1.2.3 From 506b8ebdba73f85c169599ce861b339e97370eb2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 12 Aug 2008 11:33:20 +0100 Subject: python: Use st_buffers instead of pipe_buffers. --- src/gallium/state_trackers/python/p_context.i | 32 ++++++++++++++++++++++----- 1 file changed, 26 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i index df07f3bc47..521a2dee07 100644 --- a/src/gallium/state_trackers/python/p_context.i +++ b/src/gallium/state_trackers/python/p_context.i @@ -106,12 +106,12 @@ struct st_context { } void set_constant_buffer(unsigned shader, unsigned index, - const struct pipe_constant_buffer *buf ) + struct st_buffer *buffer ) { struct pipe_constant_buffer state; memset(&state, 0, sizeof(state)); - state.buffer = buf->buffer; - state.size = buf->buffer->size; + state.buffer = buffer->buffer; + state.size = buffer->buffer->size; $self->pipe->set_constant_buffer($self->pipe, shader, index, &state); } @@ -142,9 +142,29 @@ struct st_context { } void set_vertex_buffer(unsigned index, - const struct pipe_vertex_buffer *buffer) { - memcpy(&$self->vertex_buffers[index], buffer, sizeof(*buffer)); - $self->pipe->set_vertex_buffers($self->pipe, PIPE_MAX_ATTRIBS, $self->vertex_buffers); + unsigned pitch, + unsigned max_index, + unsigned buffer_offset, + struct st_buffer *buffer) + { + unsigned i, num_vertex_buffers; + struct pipe_vertex_buffer state; + + memset(&state, 0, sizeof(state)); + state.pitch = pitch; + state.max_index = max_index; + state.buffer_offset = buffer_offset; + state.buffer = buffer->buffer; + + memcpy(&$self->vertex_buffers[index], &state, sizeof(state)); + + for(i = 0; i < PIPE_MAX_ATTRIBS; ++i) + if(self->vertex_buffers[i].buffer) + num_vertex_buffers = i + 1; + + $self->pipe->set_vertex_buffers($self->pipe, + num_vertex_buffers, + $self->vertex_buffers); } void set_vertex_element(unsigned index, -- cgit v1.2.3 From ebb5855821f64fd8c0129d218a9b46ee7c0e012d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 09:03:30 +0100 Subject: glut: Remove EOF characters. --- src/glut/os2/WarpWin.cpp | 1 - src/glut/os2/glut_cindex.cpp | 1 - src/glut/os2/glut_gamemode.cpp | 1 - src/glut/os2/glut_win.cpp | 1 - src/glut/os2/glut_winmisc.cpp | 1 - src/glut/os2/os2_glx.cpp | 1 - src/glut/os2/os2_menu.cpp | 1 - src/glut/os2/os2_winproc.cpp | 1 - 8 files changed, 8 deletions(-) (limited to 'src') diff --git a/src/glut/os2/WarpWin.cpp b/src/glut/os2/WarpWin.cpp index bd5a6e80c7..71cc8e1bc9 100644 --- a/src/glut/os2/WarpWin.cpp +++ b/src/glut/os2/WarpWin.cpp @@ -417,4 +417,3 @@ void ScreenToClient( HWND hwnd, POINTL *point) point->y -= swp_src.y; } - \ No newline at end of file diff --git a/src/glut/os2/glut_cindex.cpp b/src/glut/os2/glut_cindex.cpp index 3484185051..fa6d2f12d6 100644 --- a/src/glut/os2/glut_cindex.cpp +++ b/src/glut/os2/glut_cindex.cpp @@ -256,4 +256,3 @@ bla bla bla #endif } /* ENDCENTRY */ - \ No newline at end of file diff --git a/src/glut/os2/glut_gamemode.cpp b/src/glut/os2/glut_gamemode.cpp index 39918fdf39..1756e8a4b5 100644 --- a/src/glut/os2/glut_gamemode.cpp +++ b/src/glut/os2/glut_gamemode.cpp @@ -677,4 +677,3 @@ glutGameModeGet(GLenum mode) return -1; } } - \ No newline at end of file diff --git a/src/glut/os2/glut_win.cpp b/src/glut/os2/glut_win.cpp index 53ff5d5d95..63757d7152 100644 --- a/src/glut/os2/glut_win.cpp +++ b/src/glut/os2/glut_win.cpp @@ -1218,4 +1218,3 @@ glutReshapeFunc(GLUTreshapeCB reshapeFunc) __glutCurrentWindow->reshape = __glutDefaultReshape; } } - \ No newline at end of file diff --git a/src/glut/os2/glut_winmisc.cpp b/src/glut/os2/glut_winmisc.cpp index 456d19a8c1..7cd6699924 100644 --- a/src/glut/os2/glut_winmisc.cpp +++ b/src/glut/os2/glut_winmisc.cpp @@ -124,4 +124,3 @@ glutHideWindow(void) } /* ENDCENTRY */ - \ No newline at end of file diff --git a/src/glut/os2/os2_glx.cpp b/src/glut/os2/os2_glx.cpp index 5e135bc17e..bf116e49c9 100644 --- a/src/glut/os2/os2_glx.cpp +++ b/src/glut/os2/os2_glx.cpp @@ -143,4 +143,3 @@ XVisualInfo * glXChooseVisual(int mode) #if POKA #endif /* POKA */ - \ No newline at end of file diff --git a/src/glut/os2/os2_menu.cpp b/src/glut/os2/os2_menu.cpp index 56f2cdef8b..f140e9c79b 100644 --- a/src/glut/os2/os2_menu.cpp +++ b/src/glut/os2/os2_menu.cpp @@ -530,4 +530,3 @@ glutDetachMenu(int button) } } - \ No newline at end of file diff --git a/src/glut/os2/os2_winproc.cpp b/src/glut/os2/os2_winproc.cpp index 6ffe0d4624..9cbdec1cd2 100644 --- a/src/glut/os2/os2_winproc.cpp +++ b/src/glut/os2/os2_winproc.cpp @@ -1294,4 +1294,3 @@ Bool __glutSetWindowText(Window window, char *text) } #endif - \ No newline at end of file -- cgit v1.2.3 From cf16285d1bcb9f172a930f9d4f3402e379b53c33 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 09:03:52 +0100 Subject: gdi: Remove CVS keyword. --- src/gallium/winsys/gdi/wgl.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/gdi/wgl.c b/src/gallium/winsys/gdi/wgl.c index 50126014a8..3ce470480d 100644 --- a/src/gallium/winsys/gdi/wgl.c +++ b/src/gallium/winsys/gdi/wgl.c @@ -1,5 +1,3 @@ -/* $Id: wgl.c,v 1.12 2006/03/30 07:58:24 kschultz Exp $ */ - /* * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Library General Public -- cgit v1.2.3 From 3b5ee3d6de2c08faf69c701bf05d8f33ccd01502 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 09:12:55 +0100 Subject: gallium: Allow compilation inside X. --- src/gallium/auxiliary/util/u_string.h | 12 ++++++------ src/gallium/include/pipe/p_compiler.h | 5 +++++ 2 files changed, 11 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_string.h b/src/gallium/auxiliary/util/u_string.h index abc3232b49..08c89bbf77 100644 --- a/src/gallium/auxiliary/util/u_string.h +++ b/src/gallium/auxiliary/util/u_string.h @@ -28,14 +28,14 @@ /** * @file * Platform independent functions for string manipulation. - * + * * @author Jose Fonseca */ #ifndef U_STRING_H_ #define U_STRING_H_ -#ifndef WIN32 +#if !defined(WIN32) && !defined(XF86_LIBC_H) #include #endif #include @@ -48,19 +48,19 @@ extern "C" { #endif - + #ifdef WIN32 - + int util_vsnprintf(char *, size_t, const char *, va_list); int util_snprintf(char *str, size_t size, const char *format, ...); -static INLINE void +static INLINE void util_vsprintf(char *str, const char *format, va_list ap) { util_vsnprintf(str, (size_t)-1, format, ap); } -static INLINE void +static INLINE void util_sprintf(char *str, const char *format, ...) { va_list ap; diff --git a/src/gallium/include/pipe/p_compiler.h b/src/gallium/include/pipe/p_compiler.h index 2afb8464c7..4d64c74a4a 100644 --- a/src/gallium/include/pipe/p_compiler.h +++ b/src/gallium/include/pipe/p_compiler.h @@ -31,8 +31,13 @@ #include "p_config.h" +#ifndef XFree86Server #include #include +#else +#include "xf86_ansic.h" +#include "xf86_libc.h" +#endif #if defined(_WIN32) && !defined(__WIN32__) -- cgit v1.2.3 From 80d3a653f0172f01be694a29456c70f1f4da1812 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 09:13:11 +0100 Subject: tgsi: Prevent division by zero. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index 8b430548bc..b93f3d5222 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -393,10 +393,18 @@ micro_div( const union tgsi_exec_channel *src0, const union tgsi_exec_channel *src1 ) { - dst->f[0] = src0->f[0] / src1->f[0]; - dst->f[1] = src0->f[1] / src1->f[1]; - dst->f[2] = src0->f[2] / src1->f[2]; - dst->f[3] = src0->f[3] / src1->f[3]; + if (src1->f[0] != 0) { + dst->f[0] = src0->f[0] / src1->f[0]; + } + if (src1->f[1] != 0) { + dst->f[1] = src0->f[1] / src1->f[1]; + } + if (src1->f[2] != 0) { + dst->f[2] = src0->f[2] / src1->f[2]; + } + if (src1->f[3] != 0) { + dst->f[3] = src0->f[3] / src1->f[3]; + } } static void -- cgit v1.2.3 From 0f5a2ebec4c16e71bde7132edb8c2209b4000c61 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 09:15:32 +0100 Subject: gallium: Disable debug break by default on windows. --- src/gallium/auxiliary/util/p_debug.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 082b0e9fb5..2c2f2f8931 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -324,7 +324,11 @@ void _debug_assert_fail(const char *expr, const char *function) { _debug_printf("%s:%u:%s: Assertion `%s' failed.\n", file, line, function, expr); +#if defined(PIPE_OS_WINDOWS) + if (debug_get_bool_option("GALLIUM_ABORT_ON_ASSERT", FALSE)) +#else if (debug_get_bool_option("GALLIUM_ABORT_ON_ASSERT", TRUE)) +#endif debug_break(); else _debug_printf("continuing...\n"); -- cgit v1.2.3 From 3eea56dc93eeeb7f0155e2113ffea6a3fae48dd9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 7 Aug 2008 09:41:05 +0100 Subject: mesa: Rename sources to sources.mak. --- src/mesa/Makefile | 2 +- src/mesa/sources | 347 --------------------------------------------------- src/mesa/sources.mak | 347 +++++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 348 insertions(+), 348 deletions(-) delete mode 100644 src/mesa/sources create mode 100644 src/mesa/sources.mak (limited to 'src') diff --git a/src/mesa/Makefile b/src/mesa/Makefile index cf75baa706..cd42cd654b 100644 --- a/src/mesa/Makefile +++ b/src/mesa/Makefile @@ -3,7 +3,7 @@ TOP = ../.. include $(TOP)/configs/current -include sources +include sources.mak .SUFFIXES : .cpp diff --git a/src/mesa/sources b/src/mesa/sources deleted file mode 100644 index 0e0e10979b..0000000000 --- a/src/mesa/sources +++ /dev/null @@ -1,347 +0,0 @@ -### Lists of source files, included by Makefiles - -MAIN_SOURCES = \ - main/api_arrayelt.c \ - main/api_exec.c \ - main/api_loopback.c \ - main/api_noop.c \ - main/api_validate.c \ - main/accum.c \ - main/attrib.c \ - main/arrayobj.c \ - main/blend.c \ - main/bufferobj.c \ - main/buffers.c \ - main/clear.c \ - main/clip.c \ - main/colortab.c \ - main/context.c \ - main/convolve.c \ - main/debug.c \ - main/depth.c \ - main/depthstencil.c \ - main/dlist.c \ - main/drawpix.c \ - main/enable.c \ - main/enums.c \ - main/eval.c \ - main/execmem.c \ - main/extensions.c \ - main/fbobject.c \ - main/feedback.c \ - main/ffvertex_prog.c \ - main/fog.c \ - main/framebuffer.c \ - main/get.c \ - main/getstring.c \ - main/hash.c \ - main/hint.c \ - main/histogram.c \ - main/image.c \ - main/imports.c \ - main/light.c \ - main/lines.c \ - main/matrix.c \ - main/mipmap.c \ - main/mm.c \ - main/multisample.c \ - main/pixel.c \ - main/pixelstore.c \ - main/points.c \ - main/polygon.c \ - main/queryobj.c \ - main/rastpos.c \ - main/rbadaptors.c \ - main/readpix.c \ - main/renderbuffer.c \ - main/scissor.c \ - main/shaders.c \ - main/state.c \ - main/stencil.c \ - main/texcompress.c \ - main/texcompress_s3tc.c \ - main/texcompress_fxt1.c \ - main/texenv.c \ - main/texenvprogram.c \ - main/texformat.c \ - main/texgen.c \ - main/teximage.c \ - main/texobj.c \ - main/texparam.c \ - main/texrender.c \ - main/texstate.c \ - main/texstore.c \ - main/varray.c \ - main/vtxfmt.c - -GLAPI_SOURCES = \ - main/dispatch.c \ - glapi/glapi.c \ - glapi/glapi_getproc.c \ - glapi/glthread.c - -MATH_SOURCES = \ - math/m_debug_clip.c \ - math/m_debug_norm.c \ - math/m_debug_xform.c \ - math/m_eval.c \ - math/m_matrix.c \ - math/m_translate.c \ - math/m_vector.c \ - math/m_xform.c - -__SWRAST_SOURCES = \ - swrast/s_aaline.c \ - swrast/s_aatriangle.c \ - swrast/s_accum.c \ - swrast/s_alpha.c \ - swrast/s_atifragshader.c \ - swrast/s_bitmap.c \ - swrast/s_blend.c \ - swrast/s_blit.c \ - swrast/s_buffers.c \ - swrast/s_copypix.c \ - swrast/s_context.c \ - swrast/s_depth.c \ - swrast/s_drawpix.c \ - swrast/s_feedback.c \ - swrast/s_fog.c \ - swrast/s_fragprog.c \ - swrast/s_imaging.c \ - swrast/s_lines.c \ - swrast/s_logic.c \ - swrast/s_masking.c \ - swrast/s_points.c \ - swrast/s_readpix.c \ - swrast/s_span.c \ - swrast/s_stencil.c \ - swrast/s_texcombine.c \ - swrast/s_texfilter.c \ - swrast/s_texstore.c \ - swrast/s_triangle.c \ - swrast/s_zoom.c - -__SWRAST_SETUP_SOURCES = \ - swrast_setup/ss_context.c \ - swrast_setup/ss_triangle.c - -__TNL_SOURCES = \ - tnl/t_context.c \ - tnl/t_pipeline.c \ - tnl/t_draw.c \ - tnl/t_rasterpos.c \ - tnl/t_vb_program.c \ - tnl/t_vb_render.c \ - tnl/t_vb_texgen.c \ - tnl/t_vb_texmat.c \ - tnl/t_vb_vertex.c \ - tnl/t_vb_cull.c \ - tnl/t_vb_fog.c \ - tnl/t_vb_light.c \ - tnl/t_vb_normals.c \ - tnl/t_vb_points.c \ - tnl/t_vp_build.c \ - tnl/t_vertex.c \ - tnl/t_vertex_sse.c \ - tnl/t_vertex_generic.c - -VBO_SOURCES = \ - vbo/vbo_context.c \ - vbo/vbo_exec.c \ - vbo/vbo_exec_api.c \ - vbo/vbo_exec_array.c \ - vbo/vbo_exec_draw.c \ - vbo/vbo_exec_eval.c \ - vbo/vbo_rebase.c \ - vbo/vbo_split.c \ - vbo/vbo_split_copy.c \ - vbo/vbo_split_inplace.c \ - vbo/vbo_save.c \ - vbo/vbo_save_api.c \ - vbo/vbo_save_draw.c \ - vbo/vbo_save_loopback.c - -__VF_SOURCES = \ - vf/vf.c \ - vf/vf_generic.c \ - vf/vf_sse.c - -STATETRACKER_SOURCES = \ - state_tracker/st_atom.c \ - state_tracker/st_atom_blend.c \ - state_tracker/st_atom_clip.c \ - state_tracker/st_atom_constbuf.c \ - state_tracker/st_atom_depth.c \ - state_tracker/st_atom_framebuffer.c \ - state_tracker/st_atom_pixeltransfer.c \ - state_tracker/st_atom_sampler.c \ - state_tracker/st_atom_scissor.c \ - state_tracker/st_atom_shader.c \ - state_tracker/st_atom_rasterizer.c \ - state_tracker/st_atom_stipple.c \ - state_tracker/st_atom_texture.c \ - state_tracker/st_atom_viewport.c \ - state_tracker/st_cb_accum.c \ - state_tracker/st_cb_bitmap.c \ - state_tracker/st_cb_blit.c \ - state_tracker/st_cb_bufferobjects.c \ - state_tracker/st_cb_clear.c \ - state_tracker/st_cb_flush.c \ - state_tracker/st_cb_drawpixels.c \ - state_tracker/st_cb_fbo.c \ - state_tracker/st_cb_feedback.c \ - state_tracker/st_cb_program.c \ - state_tracker/st_cb_queryobj.c \ - state_tracker/st_cb_rasterpos.c \ - state_tracker/st_cb_readpixels.c \ - state_tracker/st_cb_strings.c \ - state_tracker/st_cb_texture.c \ - state_tracker/st_api.c \ - state_tracker/st_context.c \ - state_tracker/st_debug.c \ - state_tracker/st_draw.c \ - state_tracker/st_extensions.c \ - state_tracker/st_format.c \ - state_tracker/st_framebuffer.c \ - state_tracker/st_gen_mipmap.c \ - state_tracker/st_mesa_to_tgsi.c \ - state_tracker/st_program.c \ - state_tracker/st_texture.c - -SHADER_SOURCES = \ - shader/arbprogparse.c \ - shader/arbprogram.c \ - shader/atifragshader.c \ - shader/grammar/grammar_mesa.c \ - shader/nvfragparse.c \ - shader/nvprogram.c \ - shader/nvvertparse.c \ - shader/program.c \ - shader/prog_cache.c \ - shader/prog_debug.c \ - shader/prog_execute.c \ - shader/prog_instruction.c \ - shader/prog_parameter.c \ - shader/prog_print.c \ - shader/prog_statevars.c \ - shader/prog_uniform.c \ - shader/programopt.c \ - shader/shader_api.c \ - -SLANG_SOURCES = \ - shader/slang/slang_builtin.c \ - shader/slang/slang_codegen.c \ - shader/slang/slang_compile.c \ - shader/slang/slang_compile_function.c \ - shader/slang/slang_compile_operation.c \ - shader/slang/slang_compile_struct.c \ - shader/slang/slang_compile_variable.c \ - shader/slang/slang_emit.c \ - shader/slang/slang_ir.c \ - shader/slang/slang_label.c \ - shader/slang/slang_library_noise.c \ - shader/slang/slang_link.c \ - shader/slang/slang_log.c \ - shader/slang/slang_mem.c \ - shader/slang/slang_preprocess.c \ - shader/slang/slang_print.c \ - shader/slang/slang_simplify.c \ - shader/slang/slang_storage.c \ - shader/slang/slang_typeinfo.c \ - shader/slang/slang_vartable.c \ - shader/slang/slang_utility.c - -ASM_C_SOURCES = \ - x86/common_x86.c \ - x86/x86.c \ - x86/3dnow.c \ - x86/sse.c \ - sparc/sparc.c \ - ppc/common_ppc.c \ - x86-64/x86-64.c - -X86_SOURCES = \ - x86/common_x86_asm.S \ - x86/x86_xform2.S \ - x86/x86_xform3.S \ - x86/x86_xform4.S \ - x86/x86_cliptest.S \ - x86/mmx_blend.S \ - x86/3dnow_xform1.S \ - x86/3dnow_xform2.S \ - x86/3dnow_xform3.S \ - x86/3dnow_xform4.S \ - x86/3dnow_normal.S \ - x86/sse_xform1.S \ - x86/sse_xform2.S \ - x86/sse_xform3.S \ - x86/sse_xform4.S \ - x86/sse_normal.S \ - x86/read_rgba_span_x86.S - -X86_API = \ - x86/glapi_x86.S - -X86-64_SOURCES = \ - x86-64/xform4.S - -X86-64_API = \ - x86-64/glapi_x86-64.S - -SPARC_SOURCES = \ - sparc/clip.S \ - sparc/norm.S \ - sparc/xform.S - -SPARC_API = \ - sparc/glapi_sparc.S - -__COMMON_DRIVER_SOURCES = \ - drivers/common/driverfuncs.c - - - -### All the core C sources - -MESA_SOURCES = \ - $(MAIN_SOURCES) \ - $(MATH_SOURCES) \ - $(VBO_SOURCES) \ - $(STATETRACKER_SOURCES) \ - $(TNL_SOURCES) \ - $(SHADER_SOURCES) \ - $(SWRAST_SOURCES) \ - $(SWRAST_SETUP_SOURCES) \ - $(ASM_C_SOURCES) \ - $(SLANG_SOURCES) - -ALL_SOURCES = \ - $(MESA_SOURCES) \ - $(GLAPI_SOURCES) \ - $(MESA_ASM_SOURCES) \ - $(COMMON_DRIVER_SOURCES) - - -### Object files - -MESA_OBJECTS = \ - $(MESA_SOURCES:.c=.o) \ - $(MESA_ASM_SOURCES:.S=.o) - -GLAPI_OBJECTS = \ - $(GLAPI_SOURCES:.c=.o) \ - $(GLAPI_ASM_SOURCES:.S=.o) - - -COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o) - - - -### Include directories - -INCLUDE_DIRS = \ - -I$(TOP)/include \ - -I$(TOP)/src/mesa \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/src/gallium/auxiliary diff --git a/src/mesa/sources.mak b/src/mesa/sources.mak new file mode 100644 index 0000000000..0e0e10979b --- /dev/null +++ b/src/mesa/sources.mak @@ -0,0 +1,347 @@ +### Lists of source files, included by Makefiles + +MAIN_SOURCES = \ + main/api_arrayelt.c \ + main/api_exec.c \ + main/api_loopback.c \ + main/api_noop.c \ + main/api_validate.c \ + main/accum.c \ + main/attrib.c \ + main/arrayobj.c \ + main/blend.c \ + main/bufferobj.c \ + main/buffers.c \ + main/clear.c \ + main/clip.c \ + main/colortab.c \ + main/context.c \ + main/convolve.c \ + main/debug.c \ + main/depth.c \ + main/depthstencil.c \ + main/dlist.c \ + main/drawpix.c \ + main/enable.c \ + main/enums.c \ + main/eval.c \ + main/execmem.c \ + main/extensions.c \ + main/fbobject.c \ + main/feedback.c \ + main/ffvertex_prog.c \ + main/fog.c \ + main/framebuffer.c \ + main/get.c \ + main/getstring.c \ + main/hash.c \ + main/hint.c \ + main/histogram.c \ + main/image.c \ + main/imports.c \ + main/light.c \ + main/lines.c \ + main/matrix.c \ + main/mipmap.c \ + main/mm.c \ + main/multisample.c \ + main/pixel.c \ + main/pixelstore.c \ + main/points.c \ + main/polygon.c \ + main/queryobj.c \ + main/rastpos.c \ + main/rbadaptors.c \ + main/readpix.c \ + main/renderbuffer.c \ + main/scissor.c \ + main/shaders.c \ + main/state.c \ + main/stencil.c \ + main/texcompress.c \ + main/texcompress_s3tc.c \ + main/texcompress_fxt1.c \ + main/texenv.c \ + main/texenvprogram.c \ + main/texformat.c \ + main/texgen.c \ + main/teximage.c \ + main/texobj.c \ + main/texparam.c \ + main/texrender.c \ + main/texstate.c \ + main/texstore.c \ + main/varray.c \ + main/vtxfmt.c + +GLAPI_SOURCES = \ + main/dispatch.c \ + glapi/glapi.c \ + glapi/glapi_getproc.c \ + glapi/glthread.c + +MATH_SOURCES = \ + math/m_debug_clip.c \ + math/m_debug_norm.c \ + math/m_debug_xform.c \ + math/m_eval.c \ + math/m_matrix.c \ + math/m_translate.c \ + math/m_vector.c \ + math/m_xform.c + +__SWRAST_SOURCES = \ + swrast/s_aaline.c \ + swrast/s_aatriangle.c \ + swrast/s_accum.c \ + swrast/s_alpha.c \ + swrast/s_atifragshader.c \ + swrast/s_bitmap.c \ + swrast/s_blend.c \ + swrast/s_blit.c \ + swrast/s_buffers.c \ + swrast/s_copypix.c \ + swrast/s_context.c \ + swrast/s_depth.c \ + swrast/s_drawpix.c \ + swrast/s_feedback.c \ + swrast/s_fog.c \ + swrast/s_fragprog.c \ + swrast/s_imaging.c \ + swrast/s_lines.c \ + swrast/s_logic.c \ + swrast/s_masking.c \ + swrast/s_points.c \ + swrast/s_readpix.c \ + swrast/s_span.c \ + swrast/s_stencil.c \ + swrast/s_texcombine.c \ + swrast/s_texfilter.c \ + swrast/s_texstore.c \ + swrast/s_triangle.c \ + swrast/s_zoom.c + +__SWRAST_SETUP_SOURCES = \ + swrast_setup/ss_context.c \ + swrast_setup/ss_triangle.c + +__TNL_SOURCES = \ + tnl/t_context.c \ + tnl/t_pipeline.c \ + tnl/t_draw.c \ + tnl/t_rasterpos.c \ + tnl/t_vb_program.c \ + tnl/t_vb_render.c \ + tnl/t_vb_texgen.c \ + tnl/t_vb_texmat.c \ + tnl/t_vb_vertex.c \ + tnl/t_vb_cull.c \ + tnl/t_vb_fog.c \ + tnl/t_vb_light.c \ + tnl/t_vb_normals.c \ + tnl/t_vb_points.c \ + tnl/t_vp_build.c \ + tnl/t_vertex.c \ + tnl/t_vertex_sse.c \ + tnl/t_vertex_generic.c + +VBO_SOURCES = \ + vbo/vbo_context.c \ + vbo/vbo_exec.c \ + vbo/vbo_exec_api.c \ + vbo/vbo_exec_array.c \ + vbo/vbo_exec_draw.c \ + vbo/vbo_exec_eval.c \ + vbo/vbo_rebase.c \ + vbo/vbo_split.c \ + vbo/vbo_split_copy.c \ + vbo/vbo_split_inplace.c \ + vbo/vbo_save.c \ + vbo/vbo_save_api.c \ + vbo/vbo_save_draw.c \ + vbo/vbo_save_loopback.c + +__VF_SOURCES = \ + vf/vf.c \ + vf/vf_generic.c \ + vf/vf_sse.c + +STATETRACKER_SOURCES = \ + state_tracker/st_atom.c \ + state_tracker/st_atom_blend.c \ + state_tracker/st_atom_clip.c \ + state_tracker/st_atom_constbuf.c \ + state_tracker/st_atom_depth.c \ + state_tracker/st_atom_framebuffer.c \ + state_tracker/st_atom_pixeltransfer.c \ + state_tracker/st_atom_sampler.c \ + state_tracker/st_atom_scissor.c \ + state_tracker/st_atom_shader.c \ + state_tracker/st_atom_rasterizer.c \ + state_tracker/st_atom_stipple.c \ + state_tracker/st_atom_texture.c \ + state_tracker/st_atom_viewport.c \ + state_tracker/st_cb_accum.c \ + state_tracker/st_cb_bitmap.c \ + state_tracker/st_cb_blit.c \ + state_tracker/st_cb_bufferobjects.c \ + state_tracker/st_cb_clear.c \ + state_tracker/st_cb_flush.c \ + state_tracker/st_cb_drawpixels.c \ + state_tracker/st_cb_fbo.c \ + state_tracker/st_cb_feedback.c \ + state_tracker/st_cb_program.c \ + state_tracker/st_cb_queryobj.c \ + state_tracker/st_cb_rasterpos.c \ + state_tracker/st_cb_readpixels.c \ + state_tracker/st_cb_strings.c \ + state_tracker/st_cb_texture.c \ + state_tracker/st_api.c \ + state_tracker/st_context.c \ + state_tracker/st_debug.c \ + state_tracker/st_draw.c \ + state_tracker/st_extensions.c \ + state_tracker/st_format.c \ + state_tracker/st_framebuffer.c \ + state_tracker/st_gen_mipmap.c \ + state_tracker/st_mesa_to_tgsi.c \ + state_tracker/st_program.c \ + state_tracker/st_texture.c + +SHADER_SOURCES = \ + shader/arbprogparse.c \ + shader/arbprogram.c \ + shader/atifragshader.c \ + shader/grammar/grammar_mesa.c \ + shader/nvfragparse.c \ + shader/nvprogram.c \ + shader/nvvertparse.c \ + shader/program.c \ + shader/prog_cache.c \ + shader/prog_debug.c \ + shader/prog_execute.c \ + shader/prog_instruction.c \ + shader/prog_parameter.c \ + shader/prog_print.c \ + shader/prog_statevars.c \ + shader/prog_uniform.c \ + shader/programopt.c \ + shader/shader_api.c \ + +SLANG_SOURCES = \ + shader/slang/slang_builtin.c \ + shader/slang/slang_codegen.c \ + shader/slang/slang_compile.c \ + shader/slang/slang_compile_function.c \ + shader/slang/slang_compile_operation.c \ + shader/slang/slang_compile_struct.c \ + shader/slang/slang_compile_variable.c \ + shader/slang/slang_emit.c \ + shader/slang/slang_ir.c \ + shader/slang/slang_label.c \ + shader/slang/slang_library_noise.c \ + shader/slang/slang_link.c \ + shader/slang/slang_log.c \ + shader/slang/slang_mem.c \ + shader/slang/slang_preprocess.c \ + shader/slang/slang_print.c \ + shader/slang/slang_simplify.c \ + shader/slang/slang_storage.c \ + shader/slang/slang_typeinfo.c \ + shader/slang/slang_vartable.c \ + shader/slang/slang_utility.c + +ASM_C_SOURCES = \ + x86/common_x86.c \ + x86/x86.c \ + x86/3dnow.c \ + x86/sse.c \ + sparc/sparc.c \ + ppc/common_ppc.c \ + x86-64/x86-64.c + +X86_SOURCES = \ + x86/common_x86_asm.S \ + x86/x86_xform2.S \ + x86/x86_xform3.S \ + x86/x86_xform4.S \ + x86/x86_cliptest.S \ + x86/mmx_blend.S \ + x86/3dnow_xform1.S \ + x86/3dnow_xform2.S \ + x86/3dnow_xform3.S \ + x86/3dnow_xform4.S \ + x86/3dnow_normal.S \ + x86/sse_xform1.S \ + x86/sse_xform2.S \ + x86/sse_xform3.S \ + x86/sse_xform4.S \ + x86/sse_normal.S \ + x86/read_rgba_span_x86.S + +X86_API = \ + x86/glapi_x86.S + +X86-64_SOURCES = \ + x86-64/xform4.S + +X86-64_API = \ + x86-64/glapi_x86-64.S + +SPARC_SOURCES = \ + sparc/clip.S \ + sparc/norm.S \ + sparc/xform.S + +SPARC_API = \ + sparc/glapi_sparc.S + +__COMMON_DRIVER_SOURCES = \ + drivers/common/driverfuncs.c + + + +### All the core C sources + +MESA_SOURCES = \ + $(MAIN_SOURCES) \ + $(MATH_SOURCES) \ + $(VBO_SOURCES) \ + $(STATETRACKER_SOURCES) \ + $(TNL_SOURCES) \ + $(SHADER_SOURCES) \ + $(SWRAST_SOURCES) \ + $(SWRAST_SETUP_SOURCES) \ + $(ASM_C_SOURCES) \ + $(SLANG_SOURCES) + +ALL_SOURCES = \ + $(MESA_SOURCES) \ + $(GLAPI_SOURCES) \ + $(MESA_ASM_SOURCES) \ + $(COMMON_DRIVER_SOURCES) + + +### Object files + +MESA_OBJECTS = \ + $(MESA_SOURCES:.c=.o) \ + $(MESA_ASM_SOURCES:.S=.o) + +GLAPI_OBJECTS = \ + $(GLAPI_SOURCES:.c=.o) \ + $(GLAPI_ASM_SOURCES:.S=.o) + + +COMMON_DRIVER_OBJECTS = $(COMMON_DRIVER_SOURCES:.c=.o) + + + +### Include directories + +INCLUDE_DIRS = \ + -I$(TOP)/include \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/auxiliary -- cgit v1.2.3 From 42b262d01a2b0f676ead098c09e33e387b8a5e3a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 11 Aug 2008 17:35:21 -0600 Subject: mesa: glsl: better mod() functions --- .../shader/slang/library/slang_common_builtin.gc | 36 +++++++++------------- 1 file changed, 15 insertions(+), 21 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index 18b38af758..e908e6c940 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -872,35 +872,29 @@ vec4 mod(const vec4 a, const float b) vec2 mod(const vec2 a, const vec2 b) { - float oneOverBx, oneOverBy; - __asm float_rcp oneOverBx, b.x; - __asm float_rcp oneOverBy, b.y; - __retVal.x = a.x - b.x * floor(a.x * oneOverBx); - __retVal.y = a.y - b.y * floor(a.y * oneOverBy); + vec2 oneOverB; + __asm float_rcp oneOverB.x, b.x; + __asm float_rcp oneOverB.y, b.y; + __retVal = a - b * floor(a * oneOverB); } vec3 mod(const vec3 a, const vec3 b) { - float oneOverBx, oneOverBy, oneOverBz; - __asm float_rcp oneOverBx, b.x; - __asm float_rcp oneOverBy, b.y; - __asm float_rcp oneOverBz, b.z; - __retVal.x = a.x - b.x * floor(a.x * oneOverBx); - __retVal.y = a.y - b.y * floor(a.y * oneOverBy); - __retVal.z = a.z - b.z * floor(a.z * oneOverBz); + vec3 oneOverB; + __asm float_rcp oneOverB.x, b.x; + __asm float_rcp oneOverB.y, b.y; + __asm float_rcp oneOverB.z, b.z; + __retVal = a - b * floor(a * oneOverB); } vec4 mod(const vec4 a, const vec4 b) { - float oneOverBx, oneOverBy, oneOverBz, oneOverBw; - __asm float_rcp oneOverBx, b.x; - __asm float_rcp oneOverBy, b.y; - __asm float_rcp oneOverBz, b.z; - __asm float_rcp oneOverBw, b.w; - __retVal.x = a.x - b.x * floor(a.x * oneOverBx); - __retVal.y = a.y - b.y * floor(a.y * oneOverBy); - __retVal.z = a.z - b.z * floor(a.z * oneOverBz); - __retVal.w = a.w - b.w * floor(a.w * oneOverBw); + vec4 oneOverB; + __asm float_rcp oneOverB.x, b.x; + __asm float_rcp oneOverB.y, b.y; + __asm float_rcp oneOverB.z, b.z; + __asm float_rcp oneOverB.w, b.w; + __retVal = a - b * floor(a * oneOverB); } -- cgit v1.2.3 From 3e0bbd404a3ec87c11653f63494d7e0aeee86aa1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 08:49:57 -0600 Subject: mesa: glsl: additional constructors --- src/mesa/shader/slang/library/slang_core.gc | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_core.gc b/src/mesa/shader/slang/library/slang_core.gc index 840a0814c5..10a6bb5cb1 100644 --- a/src/mesa/shader/slang/library/slang_core.gc +++ b/src/mesa/shader/slang/library/slang_core.gc @@ -362,6 +362,13 @@ bvec2 __constructor(const bool b1, const bool b2) __retVal.y = b2; } +bvec2 __constructor(const int i1, const int i2) +{ + __asm vec4_sne __retVal.x, i1, 0.0; + __asm vec4_sne __retVal.y, i2, 0.0; +} + + bvec2 __constructor(const bool b) { __asm vec4_move __retVal.xy, b; @@ -398,6 +405,13 @@ bvec3 __constructor(const bool b1, const bool b2, const bool b3) __retVal.z = b3; } +bvec3 __constructor(const float f1, const float f2, const float f3) +{ + __asm vec4_sne __retVal.x, f1, 0.0; + __asm vec4_sne __retVal.y, f2, 0.0; + __asm vec4_sne __retVal.z, f3, 0.0; +} + bvec3 __constructor(const bool b) { __asm vec4_move __retVal.xyz, b; -- cgit v1.2.3 From ef82311b3db49fa62c916a100d05b1b05d21f466 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 08:50:14 -0600 Subject: mesa: glsl: regenerated files --- .../shader/slang/library/slang_common_builtin_gc.h | 821 ++++++++++----------- src/mesa/shader/slang/library/slang_core_gc.h | 31 +- 2 files changed, 422 insertions(+), 430 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index 7b837a9173..3b07c15785 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -392,440 +392,425 @@ 111,100,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102, 108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101, 116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66, -0,48,0,0,48,47,20,0,0,1,0,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,3,2,0,0,9,1,111, -110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,111,110,101,79,118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,111,110,101,79,118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,9,18,95,95,114,101, -116,86,97,108,0,59,120,0,18,97,0,59,120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120, -0,18,111,110,101,79,118,101,114,66,120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59, -121,0,18,97,0,59,121,0,18,98,0,59,121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101, -79,118,101,114,66,121,0,48,0,0,48,47,20,0,0,1,0,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,11,98,0, -0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,120,0,0,1,1,111,110,101,79,118,101,114,66,121,0,0, -1,1,111,110,101,79,118,101,114,66,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79, -118,101,114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79, -118,101,114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79, -118,101,114,66,122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,59, -120,0,18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0,18,111,110,101,79,118,101,114,66, -120,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59, -121,0,58,102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79,118,101,114,66,121,0,48,0,0,48, -47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122,0,18,98,0,59,122,0,58,102,108, -111,111,114,0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48,47,20,0,0,1,0,0,12, -0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,120, -0,0,1,1,111,110,101,79,118,101,114,66,121,0,0,1,1,111,110,101,79,118,101,114,66,122,0,0,1,1,111, -110,101,79,118,101,114,66,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, -114,66,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, -114,66,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, -114,66,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, -114,66,119,0,0,18,98,0,59,119,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,97,0,59,120,0, -18,98,0,59,120,0,58,102,108,111,111,114,0,18,97,0,59,120,0,18,111,110,101,79,118,101,114,66,120,0, -48,0,0,48,47,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,18,97,0,59,121,0,18,98,0,59,121,0,58, -102,108,111,111,114,0,18,97,0,59,121,0,18,111,110,101,79,118,101,114,66,121,0,48,0,0,48,47,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,122,0,18,97,0,59,122,0,18,98,0,59,122,0,58,102,108,111,111,114, -0,18,97,0,59,122,0,18,111,110,101,79,118,101,114,66,122,0,48,0,0,48,47,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,119,0,18,97,0,59,119,0,18,98,0,59,119,0,58,102,108,111,111,114,0,18,97,0,59,119,0, -18,111,110,101,79,118,101,114,66,119,0,48,0,0,48,47,20,0,0,1,0,0,9,0,109,105,110,0,1,1,0,9,97,0,0, -1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, -18,98,0,0,0,0,1,0,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109, -105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120, -121,0,0,0,0,1,0,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109, -105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0, -59,120,121,122,0,0,0,0,1,0,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99, -52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,105, -110,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86, -97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,9, -98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121, -122,0,0,18,98,0,0,0,0,1,0,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52, -95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,0,109,97,120,0, -1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0, -0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99, -52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0, -59,120,121,0,0,0,0,1,0,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95, -109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98, -0,59,120,121,122,0,0,0,0,1,0,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99, -52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,97, -120,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86, -97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,9,98, -0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122, -0,0,18,98,0,0,0,0,1,0,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95, -109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,0,99,108,97,109, -112,0,1,1,0,9,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0, -1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18, -109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,10,0,99,108,97,109,112,0,1,1,0,10, -118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99, -52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86, -97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0, +0,48,0,0,48,47,20,0,0,1,0,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,3,2,0,0,10,1, +111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, +114,66,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79, +118,101,114,66,0,59,121,0,0,18,98,0,59,121,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98, +0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,11, +0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,3,2,0,0,11,1,111,110,101,79,118,101,114,66,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,120,0,0,18,98,0,59, +120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,121,0,0,18,98, +0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,122,0,0, +18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0, +18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,12,0,109,111,100,0,1,1,0,12, +97,0,0,1,1,0,12,98,0,0,0,1,3,2,0,0,12,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116, +95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111, +97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,122,0,0,18,98,0,59,122,0,0,0, +4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,119,0,0,18,98,0,59,119, +0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111, +110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,9,0,109,105,110,0,1,1,0,9,97,0,0,1,1,0,9,98, +0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0, +0,1,0,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,1,0, +0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0, +0,1,0,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18, +95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1, +1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59, +120,121,0,0,18,98,0,0,0,0,1,0,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99, +52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1, +0,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95, +95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,0,109,97,120,0,1,1,0,9,97,0,0,1,1,0,9, +98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0, +0,0,1,0,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,1,0, +0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0, +0,1,0,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18, +95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1, +1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120, +121,0,0,18,98,0,0,0,0,1,0,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52, +95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,0, +12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114, +101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,0,99,108,97,109,112,0,1,1,0,9,118,97,108,0,0, 1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108, 97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18, -109,97,120,86,97,108,0,0,0,0,1,0,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,9,109, +109,97,120,86,97,108,0,0,0,0,1,0,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,9,109, 105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0, 18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86, -97,108,0,0,0,0,1,0,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,10,109,105,110,86,97, -108,0,0,1,1,0,10,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114, +97,108,0,0,0,0,1,0,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,9,109,105,110,86,97, +108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114, 101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0, -0,1,0,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,11,109,105,110,86,97,108,0,0,1,1,0, -11,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97, -108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,12,0, -99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,12,109,105,110,86,97,108,0,0,1,1,0,12,109,97,120, -86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118, -97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,9,0,109,105,120,0,1, -1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101, -116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1, -1,0,10,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108, -0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0, +0,1,0,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9, +109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108, +0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,10,0,99, +108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,10,109,105,110,86,97,108,0,0,1,1,0,10,109,97,120,86, +97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97, +108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,11,0,99,108,97,109,112, +0,1,1,0,11,118,97,108,0,0,1,1,0,11,109,105,110,86,97,108,0,0,1,1,0,11,109,97,120,86,97,108,0,0,0,1, +4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109, +105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,12,0,99,108,97,109,112,0,1,1,0,12,118, +97,108,0,0,1,1,0,12,109,105,110,86,97,108,0,0,1,1,0,12,109,97,120,86,97,108,0,0,0,1,4,118,101,99, +52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86, +97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,9,0,109,105,120,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0, 1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, -18,121,0,0,18,120,0,0,0,0,1,0,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,9,97,0, +18,121,0,0,18,120,0,0,0,0,1,0,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,9,97,0, 0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18, -120,0,0,0,0,1,0,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,10,97,0,0,0,1,4,118, +120,0,0,0,0,1,0,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,9,97,0,0,0,1,4,118, 101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0, -1,0,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95, -108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,12,0, -109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,108,114,112, -0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,9,0,115,116,101,112, -0,1,1,0,9,101,100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114, -101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0,1,1,0,10, -101,100,103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101,112,0,1,1,0, -11,101,100,103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115,116,101,112,0, -1,1,0,12,101,100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114, -101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0,1,1,0,9, -101,100,103,101,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101,112,0,1,1,0,9, -101,100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115,116,101,112,0,1,1, -0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116, -86,97,108,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,9,0,115,109,111,111,116,104,115,116,101, -112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9, -1,116,0,2,58,99,108,97,109,112,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18, -101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51, -0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0, -1,1,0,10,101,100,103,101,48,0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1, -116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101, -100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48, -0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1, -0,11,101,100,103,101,48,0,0,1,1,0,11,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116, -0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100, -103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0, -17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,12, -101,100,103,101,48,0,0,1,1,0,12,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2, -58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103, -101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17, -50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101, -100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99, -108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48, -0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48, -0,0,18,116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103, -101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97, -109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47, -49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0, -18,116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101, -48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109, +1,0,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, +108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,10,0, +109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,108,114,112, +0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,11,0,109,105,120,0, +1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95, +114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,12,0,109,105,120,0,1,1,0,12, +120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116, +86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,9,0,115,116,101,112,0,1,1,0,9,101,100,103, +101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18, +120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0,1,1,0,10,101,100,103,101,0,0,1,1,0, +10,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, +120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101,112,0,1,1,0,11,101,100,103,101,0,0,1,1,0, +11,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115,116,101,112,0,1,1,0,12,101,100,103,101,0,0, +1,1,0,12,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0, +0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,10,118, +0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0, +0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,11,118, +0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, +118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0, +12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, +101,100,103,101,0,0,0,0,1,0,0,9,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103, +101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9,1,116,0,2,58,99,108,97,109, +112,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, +17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, +116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,10,101,100,103,101, +48,0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99,108,97,109, 112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, 17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,9,120,0,0,0,1,8,58,97,98,115,0,18,120, -0,0,0,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,10,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9, -1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18, -114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120, -0,0,18,114,0,0,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0, -3,2,1,0,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115, -113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108, -0,0,18,114,0,0,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0, -3,2,1,0,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115, -113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108, -0,0,18,114,0,0,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,3, -2,1,0,9,1,100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103, -116,104,0,18,100,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118,0,0,1,1,0,10, -117,0,0,0,1,3,2,1,0,10,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, -108,101,110,103,116,104,0,18,100,50,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0, -11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,1,0,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114, -101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116, -97,110,99,101,0,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,0,12,1,100,52,0,2,18,118,0,18,117,0,47, -0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0,0, -11,0,99,114,111,115,115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115, -115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,9,0,102, -97,99,101,102,111,114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0, -1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0, -4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78, -0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10, -78,0,0,1,1,0,10,73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78, -114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0, -17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,11, -0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0,1,1,0,11,78,114,101, -102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1, -115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105, -120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,12,0,102,97,99,101,102,111,114,119,97,114,100, -0,1,1,0,12,78,0,0,1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111, -116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0, -18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0, -0,1,0,0,9,0,114,101,102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0, -0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,10,0,114,101,102,108,101,99, -116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18, -73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78, -0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0, -12,0,114,101,102,108,101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58, -100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,9,0,114,101,102,114,97,99,116,0,1,1, -0,9,73,0,0,1,1,0,9,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58, -100,111,116,0,18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101, -116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48, -47,0,0,3,2,0,0,9,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116, -118,97,108,0,17,48,0,48,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101, +116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,11,101,100,103,101, +48,0,0,1,1,0,11,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97,109, +112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, +17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, +116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,12,101,100,103,101, +48,0,0,1,1,0,12,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109, +112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, +17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, +116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48, +0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99,108,97,109,112, +0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17, +48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0, +48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1, +1,0,9,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97,109,112,0,18, +118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0, +48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48, +47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1, +0,9,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118, +0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0, +0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48, +0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,9,120,0,0,0,1,8,58,97,98,115,0,18,120,0,0,0,0,0,1,0, +0,9,0,108,101,110,103,116,104,0,1,1,0,10,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0,2,58, +100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112, +0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0, +0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112, +0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0, +18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0, +0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0, +2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18, +112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1, +0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,3,2,1,0,9,1,100,0,2,18, +120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,0,0,0, +20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,1,0,10, +1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116, +104,0,18,100,50,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,11,118,0,0,1,1,0,11, +117,0,0,0,1,3,2,1,0,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, +108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0, +12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,0,12,1,100,52,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114, +101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0,0,11,0,99,114,111,115, +115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115,115,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,9,0,102,97,99,101,102,111, +114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100, +0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52, +95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0, +0,18,115,0,0,0,0,0,1,0,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10,78,0,0,1,1,0,10, +73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0, +18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0, +18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,11,0,102,97,99,101, +102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0,1,1,0,11,78,114,101,102,0,0,0,1,3,2,1, +0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118, +101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54, +0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,12,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,12,78,0,0, +1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101, +102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0, +48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,9,0,114,101, +102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0, +18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,10,0,114,101,102,108,101,99,116,0,1,1,0,10,73,0,0, +1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48, +47,0,0,1,0,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,0,1,8,18,73,0,17,50, +0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,12,0,114,101,102,108, +101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0, +0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,9,0,114,101,102,114,97,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78, +0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78,0, +0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48, +0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,9,1,114, +101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,17,48,0,48, +0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100, +111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118, +97,108,0,0,0,1,0,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,1,1,0,9,101, +116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78,0,0,18,73,0,0,0,0, +0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95, +100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,10,1,114,101,116,118, +97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101,99,50,0,17, +48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18, +110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114, +101,116,118,97,108,0,0,0,1,0,0,11,0,114,101,102,114,97,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,1, +1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78,0,0,18, +73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0, +18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,11,1,114,101, +116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101,99, +51,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116, +97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8, +18,114,101,116,118,97,108,0,0,0,1,0,0,12,0,114,101,102,114,97,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78, +0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78,0, +0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48, +0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,12,1,114, +101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101, +99,52,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101, 116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0, -8,18,114,101,116,118,97,108,0,0,0,1,0,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10, -78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78, -0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0, -48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,10,1, -114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118, -101,99,50,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18, -101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47, -20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,11,0,114,101,102,114,97,99,116,0,1,1,0,11,73,0,0,1,1, -0,11,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0, -18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17, -49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0, -11,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58, -118,101,99,51,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48, -18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48, -47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,12,0,114,101,102,114,97,99,116,0,1,1,0,12,73,0,0,1, -1,0,12,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116, -0,18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48, -17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0, -0,12,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0, -58,118,101,99,52,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0, -48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0, -48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,13,0,109,97,116,114,105,120,67,111,109,112,77, -117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97,116,50,0,18,109,0,16,8,48,0,57,18, -110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,0,14,0,109,97, -116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1,0,0,14,110,0,0,0,1,8,58,109,97, -116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49, -0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,0,15,0,109,97,116,114,105, -120,67,111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0,0,0,1,8,58,109,97,116,52,0,18, -109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0, -18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57, -48,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118, -101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52, -95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0, -0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115, -108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115, -84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97, -110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,0,1, -1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108, -0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,10, -117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0, -1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97, +8,18,114,101,116,118,97,108,0,0,0,1,0,0,13,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116, +0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97,116,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8, +48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,0,14,0,109,97,116,114, +105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1,0,0,14,110,0,0,0,1,8,58,109,97,116,51,0, +18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48, +0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,0,15,0,109,97,116,114,105,120,67, +111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0,0,0,1,8,58,109,97,116,52,0,18,109,0, +16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109, +0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0, +0,1,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52, +95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3, +0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108, +101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0, +18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97, +110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0, +7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,8,117,0, +0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0, +0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1, +0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117, +0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0, +1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97, +108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1, +0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0, +59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108, +0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97, 108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,69,113, -117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114, -101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117, -97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,69, -113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104, -97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18, -95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84, -104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114, -84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116, -101,114,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18, -95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84, -104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,0,3,0,103, -114,101,97,116,101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115, -103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0, -103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95, -115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101, -97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101, -99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1, -0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118, -0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1, -0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108, -0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108, -0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97, +117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101, +116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,0,1, +1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, 108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110, -69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116, -101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95, -115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117, -97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,11, -117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12, +0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104, +97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101, +116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,0,1, +1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108, +0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,0,3,0,103,114,101,97,116, +101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97, +116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0, +18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114, +84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114, +101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118, +101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0, +0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1, +0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18, +118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0, +1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0, +1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97, +110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95, +95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,10,117, +0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0, +0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0, +0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99, +52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113, +117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,7, +117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8, 118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, -0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115, +0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115, 101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101, -113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114, +113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114, 101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1, -1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108, -0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4, -118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0, -0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115, -101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0, -101,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, -114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1, -0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,11,117, -0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,12,117,0,0,1, -1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0, -18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0, -0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95, -115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0, -4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113, -117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0, -1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0, -4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0, -18,117,0,0,18,118,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0, -4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0, -0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0, -59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,115,117,109, -0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59, -121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0, -18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, -18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,4,118,0,0,0,1,3,2,0, -0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59, -120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115, -117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120, -0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95, -114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97, -108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52, -95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1, -0,0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4, -118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0, -18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112, -114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,112, -114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18, -118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112, -114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116, -105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,119,0,0,0,4,118,101,99, -52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0, -1,0,0,2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3,0,110,111,116,0,1,1,0,3,118,0,0, -0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0, -0,17,48,0,48,0,0,0,0,0,1,0,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113, -0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,12,0,116,101,120,116, -117,114,101,49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4, -118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106, -0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95, -116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18, -99,111,111,114,100,0,59,120,121,121,121,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80, -114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118, -101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,0,1,1,0,17,115, -97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50, +1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108, +0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10, +118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, +117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0, +0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0, +0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4, +118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0, +2,0,110,111,116,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110, +101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111, +116,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69, +113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114, +101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2, +117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1, +1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, +0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4, +118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, +0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52, +95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99, +52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48, +0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101, +99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118, +101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122, +0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109, +0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,115,117, +109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0, +59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0, +0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109, +0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108, +0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,2,118,0, +0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, +112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0, +1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, +108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109, +117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0, +4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0, +48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0, +59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112, +114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112, +114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,2,0,110,111,116, +0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95, +115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0, +1,0,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116, +86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,0,1,1,0, +16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, +120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, +114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109, +112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100, +0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59, +120,121,121,121,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115, +97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112, +49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, +100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,0,1,1,0,17,115,97,109,112,108,101,114,0,0, +1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116, +86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101, +120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, +111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0, +18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0, +116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0, +12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86, +97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120, +116,117,114,101,51,68,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0, +1,4,118,101,99,52,95,116,101,120,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, +101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111, +106,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, +95,116,101,120,112,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, +18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,67,117,98,101,0,1,1,0,19,115, +97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,99, +117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, +114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,0,1,1,0,20,115,97,109,112,108,101,114,0,0, +1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116, +86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104, +97,100,111,119,49,68,80,114,111,106,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, +114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18, +115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50, +68,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52, +95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, +111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,21,115,97, +109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50, 100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0, -0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108, -101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95, -95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121, -122,122,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109, -112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100, -0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0, -0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,11, -99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,51,100,0,18,95,95,114,101,116,86,97,108, -0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117, -114,101,51,68,80,114,111,106,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100, -0,0,0,1,4,118,101,99,52,95,116,101,120,112,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97, -109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,67, -117,98,101,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101, -99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,0,1,1,0,20,115,97, -109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100, -0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0, -0,1,0,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,20,115,97,109,112,108,101,114,0,0, -1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101, -116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115, -104,97,100,111,119,50,68,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0, -0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, -108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,80,114,111, -106,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, -95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, -18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,0,1,1,0, -22,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, -120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, -111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106, -0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95, -116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101, -50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, -114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97, -100,111,119,50,68,82,101,99,116,0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114, -100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0, -18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119, -50,68,82,101,99,116,80,114,111,106,0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, -114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,9,0,110,111,105,115, -101,49,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,49,0,18,95,95,114,101,116, -86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,10,120,0,0,0,1,4,102,108,111, -97,116,95,110,111,105,115,101,50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110, -111,105,115,101,49,0,1,1,0,11,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,51,0,18,95, -95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,12,120,0,0,0, -1,4,102,108,111,97,116,95,110,111,105,115,101,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0, -0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, -110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115, -101,50,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101, -49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18, -120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,0,1,0,0,10,0,110, -111,105,115,101,50,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111, +0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,0,1,1,0,22,115,97,109,112,108,101, +114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18, +95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1, +0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112, +108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101, +99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, +100,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114, +111,106,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99, +52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, +101,114,0,0,18,99,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116, +0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95, +116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0, +0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,80,114,111, +106,0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, +95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, +101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,9,120,0,0,0,1,4, +102,108,111,97,116,95,110,111,105,115,101,49,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1, +0,0,9,0,110,111,105,115,101,49,0,1,1,0,10,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101, +50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,11, +120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,51,0,18,95,95,114,101,116,86,97,108,0,0,18, +120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,12,120,0,0,0,1,4,102,108,111,97,116,95,110, +111,105,115,101,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,10,0,110,111,105,115, +101,50,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49, +0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120, +0,17,49,57,0,51,52,0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,10,120,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57, +0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,11,120, +0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0, +9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51, +0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,0,0,10,0,110, +111,105,115,101,50,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111, 105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115, -101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0, -0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17, -55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105, -115,101,51,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115, -101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0, -18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111, -105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1, -0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0, -0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118, -101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49, -55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,11,120,0,0,0,1,9,18,95, -95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51, -52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56, -53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,12,120,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0, -17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9, -18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0, -17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0, -20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0, -59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0, -58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,17,50,51,0,53,52,0,0,46,0,0,20,0, -0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0, +0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,9,120,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52, +55,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101, +116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0, +17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115, +101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1, +0,0,11,0,110,111,105,115,101,51,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0, +58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110, +111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51, +0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49, +0,18,120,0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0, +46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54, +54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56, +53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115, +101,52,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49, +0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120, +0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115, +101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110, +111,105,115,101,49,0,18,120,0,17,50,51,0,53,52,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52, +0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18, +120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58, +118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86, +97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17, +49,55,0,56,53,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115, +101,49,0,18,120,0,58,118,101,99,50,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0, +1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120, +0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110, +111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51, +0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49, +0,18,120,0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0, +46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58, +118,101,99,51,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0, +0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, 120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, -110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0, -46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58, -118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,50,51,0,53,52,0,0,0, -17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,11,120,0,0,0,1,9, -18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57, -0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97, -108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49, -55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0, -58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0, -0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,12,120,0,0,0, -1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17, -49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18, -95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17, -53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20, -0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99, -52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,17,51,55,0,52,56,0,0,0,0, -46,0,0,20,0,0,0 +110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17, +51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58, +110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17, +49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119, +0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49, +0,0,0,17,51,49,0,57,49,0,0,0,17,51,55,0,52,56,0,0,0,0,46,0,0,20,0,0,0 diff --git a/src/mesa/shader/slang/library/slang_core_gc.h b/src/mesa/shader/slang/library/slang_core_gc.h index 416ee23d20..e8b8041914 100644 --- a/src/mesa/shader/slang/library/slang_core_gc.h +++ b/src/mesa/shader/slang/library/slang_core_gc.h @@ -67,18 +67,25 @@ 116,86,97,108,0,0,18,102,0,0,0,0,1,0,0,8,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,116,111,95,105, 118,101,99,52,0,18,95,95,114,101,116,86,97,108,0,0,18,98,0,0,0,0,1,0,0,2,1,1,1,0,1,98,49,0,0,1,1,0, 1,98,50,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116, -86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,0,0,2,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,2,1,1,1,0,9,102,0,0,0,1, -4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,17,48, -0,48,0,0,0,0,0,1,0,0,2,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,0,2,1,1,1,0,10,118,0,0,0,1,4,118, -101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0, -0,0,0,0,1,0,0,2,1,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0, -1,1,0,1,98,51,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114, -101,116,86,97,108,0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51, -0,20,0,0,1,0,0,3,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,0,3,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110, +86,97,108,0,59,121,0,18,98,50,0,20,0,0,1,0,0,2,1,1,1,0,5,105,49,0,0,1,1,0,5,105,50,0,0,0,1,4,118, +101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,105,49,0,0,17,48,0,48,0, +0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,105,50,0,0, +17,48,0,48,0,0,0,0,0,1,0,0,2,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,2,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52, +95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1, +0,0,2,1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,0,2,1,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,2, +1,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3,1,1,1,0,1,98,49,0,0,1,1,0,1,98,50,0,0,1,1,0,1,98,51,0, +0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,18,98,49,0,20,0,9,18,95,95,114,101,116,86,97,108, +0,59,121,0,18,98,50,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,18,98,51,0,20,0,0,1,0,0,3,1, +1,1,0,9,102,49,0,0,1,1,0,9,102,50,0,0,1,1,0,9,102,51,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,0,0,18,102,49,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115, +110,101,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,102,50,0,0,17,48,0,48,0,0,0,0,4,118,101, +99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,102,51,0,0,17,48,0,48,0,0,0, +0,0,1,0,0,3,1,1,1,0,1,98,0,0,0,1,4,118,101,99,52,95,109,111,118,101,0,18,95,95,114,101,116,86,97, +108,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,0,3,1,1,1,0,9,102,0,0,0,1,4,118,101,99,52,95,115,110, 101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,102,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3, 1,1,1,0,5,105,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121, 122,0,0,18,105,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3,1,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110, -- cgit v1.2.3 From 0fe0dc32e44b62e51b20645288360c7ce97ea436 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 08:50:29 -0600 Subject: mesa: glsl: better error messages --- src/mesa/shader/slang/slang_codegen.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 9c17dceb19..a1d862f4ef 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2826,7 +2826,7 @@ _slang_gen_declaration(slang_assemble_ctx *A, slang_operation *oper) /* type check/compare var and initializer */ if (!_slang_assignment_compatible(A, oper, initializer)) { - slang_info_log_error(A->log, "illegal types in assignment"); + slang_info_log_error(A->log, "incompatible types in assignment"); return NULL; } @@ -2976,7 +2976,7 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper) if (!_slang_assignment_compatible(A, &oper->children[0], &oper->children[1])) { - slang_info_log_error(A->log, "illegal types in assignment"); + slang_info_log_error(A->log, "incompatible types in assignment"); return NULL; } @@ -3264,7 +3264,7 @@ _slang_gen_compare(slang_assemble_ctx *A, slang_operation *oper, t0.spec.type != SLANG_SPEC_FLOAT) || (t1.spec.type != SLANG_SPEC_INT && t1.spec.type != SLANG_SPEC_FLOAT)) { - slang_info_log_error(A->log, "Illegal type(s) for inequality operator"); + slang_info_log_error(A->log, "Incompatible type(s) for inequality operator"); return NULL; } } -- cgit v1.2.3 From 49c2ee9d99e25c8977c67c490af2416f85d3092b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 10:00:02 -0600 Subject: mesa: allow _mesa_remove_output_reads() to work on fragment programs --- src/mesa/shader/programopt.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c index d6a3231055..56f1eb832e 100644 --- a/src/mesa/shader/programopt.c +++ b/src/mesa/shader/programopt.c @@ -368,9 +368,9 @@ _mesa_count_texture_instructions(struct gl_program *prog) /** * Scan/rewrite program to remove reads of custom (output) registers. - * The passed type has to be either PROGRAM_VARYING or PROGRAM_OUTPUT. - * In GLSL vertex shaders, varying vars can be read and written. - * Normally, vertex varying vars are implemented as output registers. + * The passed type has to be either PROGRAM_OUTPUT or PROGRAM_VARYING + * (for vertex shaders). + * In GLSL shaders, varying vars can be read and written. * On some hardware, trying to read an output register causes trouble. * So, rewrite the program to use a temporary register in this case. */ @@ -381,8 +381,8 @@ _mesa_remove_output_reads(struct gl_program *prog, enum register_file type) GLint outputMap[VERT_RESULT_MAX]; GLuint numVaryingReads = 0; - assert(prog->Target == GL_VERTEX_PROGRAM_ARB); assert(type == PROGRAM_VARYING || type == PROGRAM_OUTPUT); + assert(prog->Target == GL_VERTEX_PROGRAM_ARB || type != PROGRAM_VARYING); for (i = 0; i < VERT_RESULT_MAX; i++) outputMap[i] = -1; -- cgit v1.2.3 From af3d9dba562813ffed71691bffd7faf6665c4487 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 10:00:36 -0600 Subject: mesa: set point state --- src/mesa/main/points.c | 4 ++++ src/mesa/shader/program.c | 4 ++++ 2 files changed, 8 insertions(+) (limited to 'src') diff --git a/src/mesa/main/points.c b/src/mesa/main/points.c index fbedbcb22c..d16344a42c 100644 --- a/src/mesa/main/points.c +++ b/src/mesa/main/points.c @@ -244,7 +244,11 @@ _mesa_init_point(GLcontext *ctx) ctx->Point.MaxSize = MAX2(ctx->Const.MaxPointSize, ctx->Const.MaxPointSizeAA); ctx->Point.Threshold = 1.0; +#if FEATURE_es2_glsl + ctx->Point.PointSprite = GL_TRUE; /* GL_ARB/NV_point_sprite */ +#else ctx->Point.PointSprite = GL_FALSE; /* GL_ARB/NV_point_sprite */ +#endif ctx->Point.SpriteRMode = GL_ZERO; /* GL_NV_point_sprite (only!) */ ctx->Point.SpriteOrigin = GL_UPPER_LEFT; /* GL_ARB_point_sprite */ for (i = 0; i < MAX_TEXTURE_UNITS; i++) { diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c index 02e23aaa3a..b03dd24d11 100644 --- a/src/mesa/shader/program.c +++ b/src/mesa/shader/program.c @@ -58,7 +58,11 @@ _mesa_init_program(GLcontext *ctx) #if FEATURE_NV_vertex_program || FEATURE_ARB_vertex_program ctx->VertexProgram.Enabled = GL_FALSE; +#if FEATURE_es2_glsl + ctx->VertexProgram.PointSizeEnabled = GL_TRUE; +#else ctx->VertexProgram.PointSizeEnabled = GL_FALSE; +#endif ctx->VertexProgram.TwoSideEnabled = GL_FALSE; _mesa_reference_vertprog(ctx, &ctx->VertexProgram.Current, ctx->Shared->DefaultVertexProgram); -- cgit v1.2.3 From ee7296d3199c0b0c0b6695f553a4d0919760fd41 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 10:00:59 -0600 Subject: mesa: glsl: remove reads of output regs from fragment shaders too --- src/mesa/shader/slang/slang_compile.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c index 64c0dad611..9cf0f8b81a 100644 --- a/src/mesa/shader/slang/slang_compile.c +++ b/src/mesa/shader/slang/slang_compile.c @@ -2431,19 +2431,20 @@ _slang_compile(GLcontext *ctx, struct gl_shader *shader) _slang_delete_mempool((slang_mempool *) ctx->Shader.MemPool); ctx->Shader.MemPool = NULL; - if (shader->Type == GL_VERTEX_SHADER) { - /* remove any reads of varying (output) registers */ + /* remove any reads of output registers */ #if 0 - printf("Pre-remove output reads:\n"); - _mesa_print_program(shader->Programs); + printf("Pre-remove output reads:\n"); + _mesa_print_program(shader->Program); #endif + _mesa_remove_output_reads(shader->Program, PROGRAM_OUTPUT); + if (shader->Type == GL_VERTEX_SHADER) { + /* and remove writes to varying vars in vertex programs */ _mesa_remove_output_reads(shader->Program, PROGRAM_VARYING); - _mesa_remove_output_reads(shader->Program, PROGRAM_OUTPUT); + } #if 0 - printf("Post-remove output reads:\n"); - _mesa_print_program(shader->Programs); + printf("Post-remove output reads:\n"); + _mesa_print_program(shader->Program); #endif - } return success; } -- cgit v1.2.3 From 8f243d659397bf625379daad220be199d933b0ec Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 11:47:46 -0600 Subject: mesa: glsl: enable bzero() in slang_typeinfo_construct() --- src/mesa/shader/slang/slang_typeinfo.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_typeinfo.c b/src/mesa/shader/slang/slang_typeinfo.c index b3255512ba..b2f0f14995 100644 --- a/src/mesa/shader/slang/slang_typeinfo.c +++ b/src/mesa/shader/slang/slang_typeinfo.c @@ -273,7 +273,7 @@ slang_type_specifier_compatible(const slang_type_specifier * x, GLboolean slang_typeinfo_construct(slang_typeinfo * ti) { - /*_mesa_bzero(ti, sizeof(*ti));*/ + _mesa_bzero(ti, sizeof(*ti)); slang_type_specifier_ctr(&ti->spec); ti->array_len = 0; return GL_TRUE; -- cgit v1.2.3 From faad6655946968dd16ab30cc8d5fbd5a09321976 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 12:00:57 -0600 Subject: gallium: distinguish between KIL and KILP Note: KIL (unconditional) not done yet. --- src/gallium/auxiliary/tgsi/tgsi_sse2.c | 20 ++++++++++++++++++-- 1 file changed, 18 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 0cb1f11ef2..f3a202ae89 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -1002,7 +1002,7 @@ emit_store( */ static void -emit_kil( +emit_kilp( struct x86_function *func, const struct tgsi_full_src_register *reg ) { @@ -1096,6 +1096,15 @@ emit_kil( x86_make_reg( file_REG32, reg_AX ) ); } + +static void +emit_kil( + struct x86_function *func ) +{ + /* XXX todo / fix me */ +} + + static void emit_setcc( struct x86_function *func, @@ -1609,8 +1618,15 @@ emit_instruction( return 0; break; + case TGSI_OPCODE_KILP: + /* predicated kill */ + emit_kilp( func, &inst->FullSrcRegisters[0] ); + break; + case TGSI_OPCODE_KIL: - emit_kil( func, &inst->FullSrcRegisters[0] ); + /* unconditional kill */ + emit_kil( func ); + return 0; /* XXX fix me */ break; case TGSI_OPCODE_PK2H: -- cgit v1.2.3 From 5af60556edc97bc1d4fd57933e99e9661dbc5fa0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 17:41:21 -0600 Subject: mesa: glsl: fix comment --- src/mesa/shader/slang/slang_emit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index b902da8457..f0c3791f2a 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -991,7 +991,7 @@ emit_kill(slang_emit_info *emitInfo) * Note that ARB-KILL depends on sign of vector operand. */ inst = new_instruction(emitInfo, OPCODE_KIL_NV); - inst->DstReg.CondMask = COND_TR; /* always branch */ + inst->DstReg.CondMask = COND_TR; /* always kill */ assert(emitInfo->prog->Target == GL_FRAGMENT_PROGRAM_ARB); fp = (struct gl_fragment_program *) emitInfo->prog; -- cgit v1.2.3 From c01fbc7866d7cd5cf4263dffec6d9591470b4c23 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 17:41:57 -0600 Subject: mesa: texture crop rect state --- src/mesa/main/glheader.h | 4 ++++ src/mesa/main/mtypes.h | 2 +- src/mesa/main/texparam.c | 30 ++++++++++++++++++++++++++++++ 3 files changed, 35 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/main/glheader.h b/src/mesa/main/glheader.h index 3131a356b8..f0f97c218c 100644 --- a/src/mesa/main/glheader.h +++ b/src/mesa/main/glheader.h @@ -176,6 +176,10 @@ #endif +#ifndef GL_OES_draw_texture +#define GL_TEXTURE_CROP_RECT_OES 0x8B9D +#endif + #if !defined(CAPI) && defined(WIN32) && !defined(BUILD_FOR_SNAP) #define CAPI _cdecl diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h index a95f02b889..71a4ca55bc 100644 --- a/src/mesa/main/mtypes.h +++ b/src/mesa/main/mtypes.h @@ -1434,6 +1434,7 @@ struct gl_texture_object GLenum DepthMode; /**< GL_ARB_depth_texture */ GLint _MaxLevel; /**< actual max mipmap level (q in the spec) */ GLfloat _MaxLambda; /**< = _MaxLevel - BaseLevel (q - b in spec) */ + GLint CropRect[4]; /**< GL_OES_draw_texture */ GLboolean GenerateMipmap; /**< GL_SGIS_generate_mipmap */ GLboolean _Complete; /**< Is texture object complete? */ @@ -1443,7 +1444,6 @@ struct gl_texture_object /** GL_EXT_paletted_texture */ struct gl_color_table Palette; - /** * \name For device driver. * Note: instead of attaching driver data to this pointer, it's preferable diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c index af288c4e18..3f3b448dbc 100644 --- a/src/mesa/main/texparam.c +++ b/src/mesa/main/texparam.c @@ -409,6 +409,14 @@ _mesa_TexParameterfv( GLenum target, GLenum pname, const GLfloat *params ) } } break; +#ifdef FEATURE_OES_draw_texture + case GL_TEXTURE_CROP_RECT_OES: + texObj->CropRect[0] = (GLint) params[0]; + texObj->CropRect[1] = (GLint) params[1]; + texObj->CropRect[2] = (GLint) params[2]; + texObj->CropRect[3] = (GLint) params[3]; + break; +#endif default: _mesa_error(ctx, GL_INVALID_ENUM, @@ -447,6 +455,12 @@ _mesa_TexParameteriv( GLenum target, GLenum pname, const GLint *params ) fparam[2] = INT_TO_FLOAT(params[2]); fparam[3] = INT_TO_FLOAT(params[3]); } + else if (pname == GL_TEXTURE_CROP_RECT_OES) { + fparam[0] = (GLfloat) params[0]; + fparam[1] = (GLfloat) params[1]; + fparam[2] = (GLfloat) params[2]; + fparam[3] = (GLfloat) params[3]; + } else { if (pname == GL_TEXTURE_PRIORITY) fparam[0] = INT_TO_FLOAT(params[0]); @@ -885,6 +899,14 @@ _mesa_GetTexParameterfv( GLenum target, GLenum pname, GLfloat *params ) else error = 1; break; +#ifdef FEATURE_OES_draw_texture + case GL_TEXTURE_CROP_RECT_OES: + params[0] = obj->CropRect[0]; + params[0] = obj->CropRect[1]; + params[0] = obj->CropRect[2]; + params[0] = obj->CropRect[3]; + break; +#endif default: error = 1; break; @@ -1027,6 +1049,14 @@ _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params ) return; } break; +#ifdef FEATURE_OES_draw_texture + case GL_TEXTURE_CROP_RECT_OES: + params[0] = obj->CropRect[0]; + params[0] = obj->CropRect[1]; + params[0] = obj->CropRect[2]; + params[0] = obj->CropRect[3]; + break; +#endif default: ; /* silence warnings */ } -- cgit v1.2.3 From a6880ace4ca4a3cfde1a7a7578a8134e8f32d31d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 17:42:13 -0600 Subject: mesa: glsl: fix error detection of writing to read-only variables --- src/mesa/shader/slang/slang_codegen.c | 58 +++++++++++++++++++++++++---------- 1 file changed, 41 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index a1d862f4ef..db25106d03 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2926,6 +2926,24 @@ _slang_gen_swizzle(slang_ir_node *child, GLuint swizzle) } +static GLboolean +is_store_writable(const slang_assemble_ctx *A, const slang_ir_storage *store) +{ + while (store->Parent) + store = store->Parent; + + if (!(store->File == PROGRAM_OUTPUT || + store->File == PROGRAM_TEMPORARY || + (store->File == PROGRAM_VARYING && + A->program->Target == GL_VERTEX_PROGRAM_ARB))) { + return GL_FALSE; + } + else { + return GL_TRUE; + } +} + + /** * Generate IR tree for an assignment (=). */ @@ -2981,21 +2999,21 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper) } lhs = _slang_gen_operation(A, &oper->children[0]); - if (lhs) { - if (!lhs->Store) { - slang_info_log_error(A->log, - "invalid left hand side for assignment"); - return NULL; - } - if (!(lhs->Store->File == PROGRAM_OUTPUT || - lhs->Store->File == PROGRAM_TEMPORARY || - (lhs->Store->File == PROGRAM_VARYING && - A->program->Target == GL_VERTEX_PROGRAM_ARB) || - lhs->Store->File == PROGRAM_UNDEFINED)) { - slang_info_log_error(A->log, - "illegal assignment to read-only l-value"); - return NULL; - } + if (!lhs) { + return NULL; + } + + if (!lhs->Store) { + slang_info_log_error(A->log, + "invalid left hand side for assignment"); + return NULL; + } + + /* check that lhs is writable */ + if (!is_store_writable(A, lhs->Store)) { + slang_info_log_error(A->log, + "illegal assignment to read-only l-value"); + return NULL; } rhs = _slang_gen_operation(A, &oper->children[1]); @@ -3663,7 +3681,13 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, /* user-defined uniform */ if (datatype == GL_NONE) { if (var->type.specifier.type == SLANG_SPEC_STRUCT) { - _mesa_problem(NULL, "user-declared uniform structs not supported yet"); + /* temporary work-around */ + GLenum datatype = GL_FLOAT; + GLint uniformLoc = _mesa_add_uniform(prog->Parameters, varName, + totalSize, datatype); + store = _slang_new_ir_storage_swz(PROGRAM_UNIFORM, uniformLoc, + totalSize, swizzle); + /* XXX what we need to do is unroll the struct into its * basic types, creating a uniform variable for each. * For example: @@ -3682,8 +3706,8 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, slang_info_log_error(A->log, "invalid datatype for uniform variable %s", varName); + return GL_FALSE; } - return GL_FALSE; } else { GLint uniformLoc = _mesa_add_uniform(prog->Parameters, varName, -- cgit v1.2.3 From 4424c668756a6639d3c042c297b63ea47b629735 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 12 Aug 2008 18:02:44 -0600 Subject: gallium: more context clean-up --- src/mesa/state_tracker/st_context.c | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 00447fff08..6e4a376d44 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -45,6 +45,9 @@ #include "st_cb_drawpixels.h" #include "st_cb_rasterpos.h" #endif +#ifdef FEATURE_OES_draw_texture +#include "st_cb_drawtex.h" +#endif #include "st_cb_fbo.h" #if FEATURE_feedback #include "st_cb_feedback.h" @@ -180,6 +183,9 @@ static void st_destroy_context_priv( struct st_context *st ) st_destroy_bitmap(st); st_destroy_drawpix(st); #endif +#ifdef FEATURE_OES_draw_texture + st_destroy_drawtex(st); +#endif _vbo_DestroyContext(st->ctx); -- cgit v1.2.3 From 9804ab9accb3c6607c64a65bebe70124578240cd Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 09:15:27 +0200 Subject: gallium: Make a note that OPCODE_KIL is a conditional kill. Map OPCODE_TEXKILL to OPCODE_KIL. --- src/gallium/include/pipe/p_shader_tokens.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index ed931d24b9..3ab0c7deaf 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -240,6 +240,7 @@ struct tgsi_immediate_float32 * GL_ARB_fragment_program */ #define TGSI_OPCODE_CMP 66 +#define TGSI_OPCODE_KIL 116 /* conditional kill */ #define TGSI_OPCODE_SCS 67 #define TGSI_OPCODE_TXB 68 @@ -327,7 +328,7 @@ struct tgsi_immediate_float32 /* * ps_1_1 */ -#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KILP +#define TGSI_OPCODE_TEXKILL TGSI_OPCODE_KIL /* * ps_1_2 @@ -386,7 +387,6 @@ struct tgsi_immediate_float32 * vs_2_x */ -#define TGSI_OPCODE_KIL 116 /* unpredicated kill */ #define TGSI_OPCODE_END 117 /* aka HALT */ #define TGSI_OPCODE_LAST 119 -- cgit v1.2.3 From 0c8f4c25ff1cdf8d4cad21789e0c73b41aa29c98 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 10:52:55 +0200 Subject: draw: Use KIL instead of KILP. --- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 4 ++-- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index 13b4401521..c7f4349cb3 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -301,9 +301,9 @@ aa_transform_inst(struct tgsi_transform_context *ctx, newInst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_W; ctx->emit_instruction(ctx, &newInst); - /* KILP -t0.yyyy; # if b, KILL */ + /* KIL -tmp0.yyyy; # if -tmp0.y < 0, KILL */ newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_KILP; + newInst.Instruction.Opcode = TGSI_OPCODE_KIL; newInst.Instruction.NumDstRegs = 0; newInst.Instruction.NumSrcRegs = 1; newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index d3bd9baddd..e97136fa1f 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -299,9 +299,9 @@ pstip_transform_inst(struct tgsi_transform_context *ctx, newInst.FullSrcRegisters[1].SrcRegister.Index = pctx->freeSampler; ctx->emit_instruction(ctx, &newInst); - /* KILP texTemp; # if texTemp < 0, KILL fragment */ + /* KIL -texTemp; # if -texTemp < 0, KILL fragment */ newInst = tgsi_default_full_instruction(); - newInst.Instruction.Opcode = TGSI_OPCODE_KILP; + newInst.Instruction.Opcode = TGSI_OPCODE_KIL; newInst.Instruction.NumDstRegs = 0; newInst.Instruction.NumSrcRegs = 1; newInst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY; -- cgit v1.2.3 From f633b14b92352d8abc2a178a1f6ff80a58662bb1 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 10:58:54 +0200 Subject: gallivm: Translate KIL instead of KILP. --- src/gallium/auxiliary/gallivm/instructions.cpp | 4 ++-- src/gallium/auxiliary/gallivm/instructions.h | 2 +- src/gallium/auxiliary/gallivm/llvm_builtins.c | 2 +- src/gallium/auxiliary/gallivm/tgsitollvm.cpp | 18 +++++++++--------- 4 files changed, 13 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/gallivm/instructions.cpp b/src/gallium/auxiliary/gallivm/instructions.cpp index 1a98491b82..035224e8f3 100644 --- a/src/gallium/auxiliary/gallivm/instructions.cpp +++ b/src/gallium/auxiliary/gallivm/instructions.cpp @@ -878,9 +878,9 @@ llvm::Value * Instructions::scs(llvm::Value *in) return call; } -llvm::Value * Instructions::kilp(llvm::Value *in) +llvm::Value * Instructions::kil(llvm::Value *in) { - llvm::Function *func = m_mod->getFunction("kilp"); + llvm::Function *func = m_mod->getFunction("kil"); assert(func); CallInst *call = m_builder.CreateCall(func, in, name("kilpres")); diff --git a/src/gallium/auxiliary/gallivm/instructions.h b/src/gallium/auxiliary/gallivm/instructions.h index 3a476928b6..d286ce80c7 100644 --- a/src/gallium/auxiliary/gallivm/instructions.h +++ b/src/gallium/auxiliary/gallivm/instructions.h @@ -79,7 +79,7 @@ public: llvm::Value *floor(llvm::Value *in); llvm::Value *frc(llvm::Value *in); void ifop(llvm::Value *in); - llvm::Value *kilp(llvm::Value *in); + llvm::Value *kil(llvm::Value *in); llvm::Value *lerp(llvm::Value *in1, llvm::Value *in2, llvm::Value *in3); llvm::Value *lit(llvm::Value *in); diff --git a/src/gallium/auxiliary/gallivm/llvm_builtins.c b/src/gallium/auxiliary/gallivm/llvm_builtins.c index 6b9d626ed4..d5a003a48b 100644 --- a/src/gallium/auxiliary/gallivm/llvm_builtins.c +++ b/src/gallium/auxiliary/gallivm/llvm_builtins.c @@ -105,7 +105,7 @@ inline float4 vsin(float4 val) return result; } -inline int kilp(float4 val) +inline int kil(float4 val) { if (val.x < 0 || val.y < 0 || val.z < 0 || val.w < 0) return 1; diff --git a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp index b14e2affd6..cc1516a45e 100644 --- a/src/gallium/auxiliary/gallivm/tgsitollvm.cpp +++ b/src/gallium/auxiliary/gallivm/tgsitollvm.cpp @@ -396,11 +396,7 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_DDY: break; - case TGSI_OPCODE_KILP: { - out = instr->kilp(inputs[0]); - storage->setKilElement(out); - return; - } + case TGSI_OPCODE_KILP: break; case TGSI_OPCODE_PK2H: break; @@ -602,7 +598,11 @@ translate_instruction(llvm::Module *module, break; case TGSI_OPCODE_BREAKC: break; - case TGSI_OPCODE_KIL: + case TGSI_OPCODE_KIL: { + out = instr->kil(inputs[0]); + storage->setKilElement(out); + return; + } break; case TGSI_OPCODE_END: instr->end(); @@ -799,8 +799,7 @@ translate_instructionir(llvm::Module *module, break; case TGSI_OPCODE_DDY: break; - case TGSI_OPCODE_KILP: { - } + case TGSI_OPCODE_KILP: break; case TGSI_OPCODE_PK2H: break; @@ -967,7 +966,8 @@ translate_instructionir(llvm::Module *module, break; case TGSI_OPCODE_BREAKC: break; - case TGSI_OPCODE_KIL: + case TGSI_OPCODE_KIL: { + } break; case TGSI_OPCODE_END: instr->end(); -- cgit v1.2.3 From f0874d1a6b832e1bb29256661cb8beab3ddeb528 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 11:09:20 +0200 Subject: tgsi: Swap meanings of KIL and KILP opcodes. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 22 ++++++++++++++++++---- src/gallium/auxiliary/tgsi/tgsi_sse2.c | 12 ++++++------ 2 files changed, 24 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index b93f3d5222..c4ba667d32 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1189,8 +1189,8 @@ store_dest( * Kill fragment if any of the four values is less than zero. */ static void -exec_kilp(struct tgsi_exec_machine *mach, - const struct tgsi_full_instruction *inst) +exec_kil(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) { uint uniquemask; uint chan_index; @@ -1226,6 +1226,21 @@ exec_kilp(struct tgsi_exec_machine *mach, mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; } +/** + * Execute NVIDIA-style KIL which is predicated by a condition code. + * Kill fragment if the condition code is TRUE. + */ +static void +exec_kilp(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + + /* TODO: build kilmask from CC mask */ + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; +} + /* * Fetch a texel using STR texture coordinates. @@ -1971,8 +1986,7 @@ exec_instruction( break; case TGSI_OPCODE_KIL: - /* for enabled ExecMask bits, set the killed bit */ - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; + exec_kil (mach, inst); break; case TGSI_OPCODE_PK2H: diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index f3a202ae89..47dc06faf6 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -1002,7 +1002,7 @@ emit_store( */ static void -emit_kilp( +emit_kil( struct x86_function *func, const struct tgsi_full_src_register *reg ) { @@ -1098,7 +1098,7 @@ emit_kilp( static void -emit_kil( +emit_kilp( struct x86_function *func ) { /* XXX todo / fix me */ @@ -1620,13 +1620,13 @@ emit_instruction( case TGSI_OPCODE_KILP: /* predicated kill */ - emit_kilp( func, &inst->FullSrcRegisters[0] ); + emit_kilp( func ); + return 0; /* XXX fix me */ break; case TGSI_OPCODE_KIL: - /* unconditional kill */ - emit_kil( func ); - return 0; /* XXX fix me */ + /* conditional kill */ + emit_kil( func, &inst->FullSrcRegisters[0] ); break; case TGSI_OPCODE_PK2H: -- cgit v1.2.3 From e8e75c8c865bb5bbff9db2682b130c8d147f3a38 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 11:10:58 +0200 Subject: cell: KILP is a predicated discard, KIL is a conditional discard. --- src/gallium/drivers/cell/spu/spu_exec.c | 21 +++++++++++++++++---- 1 file changed, 17 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 96393732ed..42e5022f30 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -603,8 +603,8 @@ store_dest( * Kill fragment if any of the four values is less than zero. */ static void -exec_kilp(struct spu_exec_machine *mach, - const struct tgsi_full_instruction *inst) +exec_kil(struct spu_exec_machine *mach, + const struct tgsi_full_instruction *inst) { uint uniquemask; uint chan_index; @@ -640,6 +640,20 @@ exec_kilp(struct spu_exec_machine *mach, mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; } +/** + * Execute NVIDIA-style KIL which is predicated by a condition code. + * Kill fragment if the condition code is TRUE. + */ +static void +exec_kilp(struct tgsi_exec_machine *mach, + const struct tgsi_full_instruction *inst) +{ + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + + /* TODO: build kilmask from CC mask */ + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; +} /* * Fetch a texel using STR texture coordinates. @@ -1336,8 +1350,7 @@ exec_instruction( break; case TGSI_OPCODE_KIL: - /* for enabled ExecMask bits, set the killed bit */ - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= mach->ExecMask; + exec_kil (mach, inst); break; case TGSI_OPCODE_PK2H: -- cgit v1.2.3 From db38708c43d7e9bbc744893ad2e9c2a77e9ec15c Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 11:13:46 +0200 Subject: i915: Swap meanings of KIL and KILP. --- src/gallium/drivers/i915simple/i915_fpc_translate.c | 16 ++++------------ 1 file changed, 4 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 04507ab8ad..64432982c4 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -560,18 +560,6 @@ i915_translate_instruction(struct i915_fp_compile *p, break; case TGSI_OPCODE_KIL: - /* unconditional kill */ - assert(0); /* not tested yet */ -#if 0 - src0 = src_vector(p, &inst->FullSrcRegisters[0]); - tmp = i915_get_utemp(p); - - i915_emit_texld(p, tmp, A0_DEST_CHANNEL_ALL, /* use a dummy dest reg */ - 0, src0, T0_TEXKILL); -#endif - break; - - case TGSI_OPCODE_KILP: /* kill if src[0].x < 0 || src[0].y < 0 ... */ src0 = src_vector(p, &inst->FullSrcRegisters[0]); tmp = i915_get_utemp(p); @@ -584,6 +572,10 @@ i915_translate_instruction(struct i915_fp_compile *p, T0_TEXKILL); /* opcode */ break; + case TGSI_OPCODE_KILP: + assert(0); /* not tested yet */ + break; + case TGSI_OPCODE_LG2: src0 = src_vector(p, &inst->FullSrcRegisters[0]); -- cgit v1.2.3 From e7e992b12ab3717eadaf055fe2fab88c3be8a092 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 11:19:24 +0200 Subject: mesa: KIL -> KIL, KIL_NV -> KILP. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 6565107b10..198d406b59 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -376,13 +376,13 @@ compile_instruction( fullinst->Instruction.Opcode = TGSI_OPCODE_INT; break; case OPCODE_KIL: - /* predicated w/ a register */ - fullinst->Instruction.Opcode = TGSI_OPCODE_KILP; + /* conditional */ + fullinst->Instruction.Opcode = TGSI_OPCODE_KIL; break; case OPCODE_KIL_NV: - /* unpredicated */ + /* predicated */ assert(inst->DstReg.CondMask == COND_TR); - fullinst->Instruction.Opcode = TGSI_OPCODE_KIL; + fullinst->Instruction.Opcode = TGSI_OPCODE_KILP; break; case OPCODE_LG2: fullinst->Instruction.Opcode = TGSI_OPCODE_LG2; -- cgit v1.2.3 From d8be393cb60d908f98f0edb74c1c7964e8f690fe Mon Sep 17 00:00:00 2001 From: Zack Rusin Date: Wed, 13 Aug 2008 07:14:47 -0400 Subject: fix compilation --- src/glx/x11/Makefile | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/glx/x11/Makefile b/src/glx/x11/Makefile index b404727f08..c99bcf26a9 100644 --- a/src/glx/x11/Makefile +++ b/src/glx/x11/Makefile @@ -33,7 +33,7 @@ SOURCES = \ dri_glx.c \ XF86dri.c -include $(TOP)/src/mesa/sources +include $(TOP)/src/mesa/sources.mak MESA_GLAPI_ASM_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_ASM_SOURCES)) MESA_GLAPI_SOURCES = $(addprefix $(TOP)/src/mesa/, $(GLAPI_SOURCES)) -- cgit v1.2.3 From 94f9faab31f7fbf5f14d23d993f9830fa51ce076 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 17:16:59 +0200 Subject: tgsi: Opcode information. --- src/gallium/auxiliary/tgsi/tgsi_info.c | 161 +++++++++++++++++++++++++++++++++ src/gallium/auxiliary/tgsi/tgsi_info.h | 53 +++++++++++ 2 files changed, 214 insertions(+) create mode 100644 src/gallium/auxiliary/tgsi/tgsi_info.c create mode 100644 src/gallium/auxiliary/tgsi/tgsi_info.h (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.c b/src/gallium/auxiliary/tgsi/tgsi_info.c new file mode 100644 index 0000000000..a4899cd4c2 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_info.c @@ -0,0 +1,161 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_debug.h" +#include "tgsi_info.h" + +static const struct tgsi_opcode_info opcode_info[TGSI_OPCODE_LAST] = +{ + { 1, 1, 0, 0, "ARL" }, + { 1, 1, 0, 0, "MOV" }, + { 1, 1, 0, 0, "LIT" }, + { 1, 1, 0, 0, "RCP" }, + { 1, 1, 0, 0, "RSQ" }, + { 1, 1, 0, 0, "EXP" }, + { 1, 1, 0, 0, "LOG" }, + { 1, 2, 0, 0, "MUL" }, + { 1, 2, 0, 0, "ADD" }, + { 1, 2, 0, 0, "DP3" }, + { 1, 2, 0, 0, "DP4" }, + { 1, 2, 0, 0, "DST" }, + { 1, 2, 0, 0, "MIN" }, + { 1, 2, 0, 0, "MAX" }, + { 1, 2, 0, 0, "SLT" }, + { 1, 2, 0, 0, "SGE" }, + { 1, 3, 0, 0, "MAD" }, + { 1, 2, 0, 0, "SUB" }, + { 1, 3, 0, 0, "LERP" }, + { 1, 3, 0, 0, "CND" }, + { 1, 3, 0, 0, "CND0" }, + { 1, 3, 0, 0, "DOT2ADD" }, + { 1, 2, 0, 0, "INDEX" }, + { 1, 1, 0, 0, "NEGATE" }, + { 1, 1, 0, 0, "FRAC" }, + { 1, 3, 0, 0, "CLAMP" }, + { 1, 1, 0, 0, "FLOOR" }, + { 1, 1, 0, 0, "ROUND" }, + { 1, 1, 0, 0, "EXPBASE2" }, + { 1, 1, 0, 0, "LOGBASE2" }, + { 1, 2, 0, 0, "POWER" }, + { 1, 2, 0, 0, "CROSSPRODUCT" }, + { 1, 2, 0, 0, "MULTIPLYMATRIX" }, + { 1, 1, 0, 0, "ABS" }, + { 1, 1, 0, 0, "RCC" }, + { 1, 2, 0, 0, "DPH" }, + { 1, 1, 0, 0, "COS" }, + { 1, 1, 0, 0, "DDX" }, + { 1, 1, 0, 0, "DDY" }, + { 0, 1, 0, 0, "KILP" }, + { 1, 1, 0, 0, "PK2H" }, + { 1, 1, 0, 0, "PK2US" }, + { 1, 1, 0, 0, "PK4B" }, + { 1, 1, 0, 0, "PK4UB" }, + { 1, 2, 0, 0, "RFL" }, + { 1, 2, 0, 0, "SEQ" }, + { 1, 2, 0, 0, "SFL" }, + { 1, 2, 0, 0, "SGT" }, + { 1, 1, 0, 0, "SIN" }, + { 1, 2, 0, 0, "SLE" }, + { 1, 2, 0, 0, "SNE" }, + { 1, 2, 0, 0, "STR" }, + { 1, 2, 1, 0, "TEX" }, + { 1, 4, 1, 0, "TXD" }, + { 1, 2, 1, 0, "TXP" }, + { 1, 1, 0, 0, "UP2H" }, + { 1, 1, 0, 0, "UP2US" }, + { 1, 1, 0, 0, "UP4B" }, + { 1, 1, 0, 0, "UP4UB" }, + { 1, 3, 0, 0, "X2D" }, + { 1, 1, 0, 0, "ARA" }, + { 1, 1, 0, 0, "ARR" }, + { 0, 1, 0, 0, "BRA" }, + { 0, 0, 0, 1, "CAL" }, + { 0, 0, 0, 0, "RET" }, + { 1, 1, 0, 0, "SSG" }, + { 1, 3, 0, 0, "CMP" }, + { 1, 1, 0, 0, "SCS" }, + { 1, 2, 1, 0, "TXB" }, + { 1, 1, 0, 0, "NRM" }, + { 1, 2, 0, 0, "DIV" }, + { 1, 2, 0, 0, "DP2" }, + { 1, 2, 1, 0, "TXL" }, + { 0, 0, 0, 0, "BRK" }, + { 0, 1, 0, 1, "IF" }, + { 0, 0, 0, 0, "LOOP" }, + { 0, 1, 0, 0, "REP" }, + { 0, 0, 0, 1, "ELSE" }, + { 0, 0, 0, 0, "ENDIF" }, + { 0, 0, 0, 0, "ENDLOOP" }, + { 0, 0, 0, 0, "ENDREP" }, + { 0, 1, 0, 0, "PUSHA" }, + { 1, 0, 0, 0, "POPA" }, + { 1, 1, 0, 0, "CEIL" }, + { 1, 1, 0, 0, "I2F" }, + { 1, 1, 0, 0, "NOT" }, + { 1, 1, 0, 0, "TRUNC" }, + { 1, 2, 0, 0, "SHL" }, + { 1, 2, 0, 0, "SHR" }, + { 1, 2, 0, 0, "AND" }, + { 1, 2, 0, 0, "OR" }, + { 1, 2, 0, 0, "MOD" }, + { 1, 2, 0, 0, "XOR" }, + { 1, 3, 0, 0, "SAD" }, + { 1, 2, 1, 0, "TXF" }, + { 1, 2, 1, 0, "TXQ" }, + { 0, 0, 0, 0, "CONT" }, + { 0, 0, 0, 0, "EMIT" }, + { 0, 0, 0, 0, "ENDPRIM" }, + { 0, 0, 0, 1, "BGNLOOP2" }, + { 0, 0, 0, 0, "BGNSUB" }, + { 0, 0, 0, 1, "ENDLOOP2" }, + { 0, 0, 0, 0, "ENDSUB" }, + { 1, 1, 0, 0, "NOISE1" }, + { 1, 1, 0, 0, "NOISE2" }, + { 1, 1, 0, 0, "NOISE3" }, + { 1, 1, 0, 0, "NOISE4" }, + { 0, 0, 0, 0, "NOP" }, + { 1, 2, 0, 0, "M4X3" }, + { 1, 2, 0, 0, "M3X4" }, + { 1, 2, 0, 0, "M3X3" }, + { 1, 2, 0, 0, "M3X2" }, + { 1, 1, 0, 0, "NRM4" }, + { 0, 1, 0, 0, "CALLNZ" }, + { 0, 1, 0, 0, "IFC" }, + { 0, 1, 0, 0, "BREAKC" }, + { 0, 0, 0, 0, "KIL" }, + { 0, 0, 0, 0, "END" }, + { 1, 1, 0, 0, "SWZ" } +}; + +const struct tgsi_opcode_info * +tgsi_get_opcode_info( uint opcode ) +{ + if (opcode < TGSI_OPCODE_LAST) + return &opcode_info[opcode]; + assert( 0 ); + return NULL; +} diff --git a/src/gallium/auxiliary/tgsi/tgsi_info.h b/src/gallium/auxiliary/tgsi/tgsi_info.h new file mode 100644 index 0000000000..7230bdaae3 --- /dev/null +++ b/src/gallium/auxiliary/tgsi/tgsi_info.h @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TGSI_INFO_H +#define TGSI_INFO_H + +#include "pipe/p_shader_tokens.h" + +#if defined __cplusplus +extern "C" { +#endif + +struct tgsi_opcode_info +{ + uint num_dst; + uint num_src; + boolean is_tex; + boolean is_branch; + const char *mnemonic; +}; + +const struct tgsi_opcode_info * +tgsi_get_opcode_info( uint opcode ); + +#if defined __cplusplus +} +#endif + +#endif /* TGSI_INFO_H */ -- cgit v1.2.3 From 668ac2572548b7818b89a424f44e9c9c59786df0 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 17:19:20 +0200 Subject: scons: List tgsi_info.c. --- src/gallium/auxiliary/tgsi/SConscript | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/SConscript b/src/gallium/auxiliary/tgsi/SConscript index 03982e2194..45bf3f6d57 100644 --- a/src/gallium/auxiliary/tgsi/SConscript +++ b/src/gallium/auxiliary/tgsi/SConscript @@ -7,6 +7,7 @@ tgsi = env.ConvenienceLibrary( 'tgsi_dump.c', 'tgsi_dump_c.c', 'tgsi_exec.c', + 'tgsi_info.c', 'tgsi_iterate.c', 'tgsi_parse.c', 'tgsi_sanity.c', -- cgit v1.2.3 From 6d83a0cc6846adb546794c0483694fdb1d1b4664 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 17:20:14 +0200 Subject: make: List tgsi_info.c. --- src/gallium/auxiliary/tgsi/Makefile | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/Makefile b/src/gallium/auxiliary/tgsi/Makefile index bbeff1304d..806a2bd4c5 100644 --- a/src/gallium/auxiliary/tgsi/Makefile +++ b/src/gallium/auxiliary/tgsi/Makefile @@ -7,6 +7,7 @@ C_SOURCES = \ tgsi_build.c \ tgsi_dump.c \ tgsi_exec.c \ + tgsi_info.c \ tgsi_iterate.c \ tgsi_parse.c \ tgsi_scan.c \ -- cgit v1.2.3 From 79e52779bdcd1fa0ff8e1cbdc7555746205ee519 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 17:22:34 +0200 Subject: tgsi: Use tgsi_opcode_info. --- src/gallium/auxiliary/tgsi/tgsi_text.c | 137 +-------------------------------- 1 file changed, 3 insertions(+), 134 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c index 35cb3055bb..7cdf1b67c5 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/tgsi_text.c @@ -28,6 +28,7 @@ #include "pipe/p_debug.h" #include "tgsi_text.h" #include "tgsi_build.h" +#include "tgsi_info.h" #include "tgsi_parse.h" #include "tgsi_sanity.h" #include "tgsi_util.h" @@ -719,138 +720,6 @@ parse_src_operand( return TRUE; } -struct opcode_info -{ - uint num_dst; - uint num_src; - uint is_tex; - uint is_branch; - const char *mnemonic; -}; - -static const struct opcode_info opcode_info[TGSI_OPCODE_LAST] = -{ - { 1, 1, 0, 0, "ARL" }, - { 1, 1, 0, 0, "MOV" }, - { 1, 1, 0, 0, "LIT" }, - { 1, 1, 0, 0, "RCP" }, - { 1, 1, 0, 0, "RSQ" }, - { 1, 1, 0, 0, "EXP" }, - { 1, 1, 0, 0, "LOG" }, - { 1, 2, 0, 0, "MUL" }, - { 1, 2, 0, 0, "ADD" }, - { 1, 2, 0, 0, "DP3" }, - { 1, 2, 0, 0, "DP4" }, - { 1, 2, 0, 0, "DST" }, - { 1, 2, 0, 0, "MIN" }, - { 1, 2, 0, 0, "MAX" }, - { 1, 2, 0, 0, "SLT" }, - { 1, 2, 0, 0, "SGE" }, - { 1, 3, 0, 0, "MAD" }, - { 1, 2, 0, 0, "SUB" }, - { 1, 3, 0, 0, "LERP" }, - { 1, 3, 0, 0, "CND" }, - { 1, 3, 0, 0, "CND0" }, - { 1, 3, 0, 0, "DOT2ADD" }, - { 1, 2, 0, 0, "INDEX" }, - { 1, 1, 0, 0, "NEGATE" }, - { 1, 1, 0, 0, "FRAC" }, - { 1, 3, 0, 0, "CLAMP" }, - { 1, 1, 0, 0, "FLOOR" }, - { 1, 1, 0, 0, "ROUND" }, - { 1, 1, 0, 0, "EXPBASE2" }, - { 1, 1, 0, 0, "LOGBASE2" }, - { 1, 2, 0, 0, "POWER" }, - { 1, 2, 0, 0, "CROSSPRODUCT" }, - { 1, 2, 0, 0, "MULTIPLYMATRIX" }, - { 1, 1, 0, 0, "ABS" }, - { 1, 1, 0, 0, "RCC" }, - { 1, 2, 0, 0, "DPH" }, - { 1, 1, 0, 0, "COS" }, - { 1, 1, 0, 0, "DDX" }, - { 1, 1, 0, 0, "DDY" }, - { 0, 1, 0, 0, "KILP" }, - { 1, 1, 0, 0, "PK2H" }, - { 1, 1, 0, 0, "PK2US" }, - { 1, 1, 0, 0, "PK4B" }, - { 1, 1, 0, 0, "PK4UB" }, - { 1, 2, 0, 0, "RFL" }, - { 1, 2, 0, 0, "SEQ" }, - { 1, 2, 0, 0, "SFL" }, - { 1, 2, 0, 0, "SGT" }, - { 1, 1, 0, 0, "SIN" }, - { 1, 2, 0, 0, "SLE" }, - { 1, 2, 0, 0, "SNE" }, - { 1, 2, 0, 0, "STR" }, - { 1, 2, 1, 0, "TEX" }, - { 1, 4, 1, 0, "TXD" }, - { 1, 2, 1, 0, "TXP" }, - { 1, 1, 0, 0, "UP2H" }, - { 1, 1, 0, 0, "UP2US" }, - { 1, 1, 0, 0, "UP4B" }, - { 1, 1, 0, 0, "UP4UB" }, - { 1, 3, 0, 0, "X2D" }, - { 1, 1, 0, 0, "ARA" }, - { 1, 1, 0, 0, "ARR" }, - { 0, 1, 0, 0, "BRA" }, - { 0, 0, 0, 1, "CAL" }, - { 0, 0, 0, 0, "RET" }, - { 1, 1, 0, 0, "SSG" }, - { 1, 3, 0, 0, "CMP" }, - { 1, 1, 0, 0, "SCS" }, - { 1, 2, 1, 0, "TXB" }, - { 1, 1, 0, 0, "NRM" }, - { 1, 2, 0, 0, "DIV" }, - { 1, 2, 0, 0, "DP2" }, - { 1, 2, 1, 0, "TXL" }, - { 0, 0, 0, 0, "BRK" }, - { 0, 1, 0, 1, "IF" }, - { 0, 0, 0, 0, "LOOP" }, - { 0, 1, 0, 0, "REP" }, - { 0, 0, 0, 1, "ELSE" }, - { 0, 0, 0, 0, "ENDIF" }, - { 0, 0, 0, 0, "ENDLOOP" }, - { 0, 0, 0, 0, "ENDREP" }, - { 0, 1, 0, 0, "PUSHA" }, - { 1, 0, 0, 0, "POPA" }, - { 1, 1, 0, 0, "CEIL" }, - { 1, 1, 0, 0, "I2F" }, - { 1, 1, 0, 0, "NOT" }, - { 1, 1, 0, 0, "TRUNC" }, - { 1, 2, 0, 0, "SHL" }, - { 1, 2, 0, 0, "SHR" }, - { 1, 2, 0, 0, "AND" }, - { 1, 2, 0, 0, "OR" }, - { 1, 2, 0, 0, "MOD" }, - { 1, 2, 0, 0, "XOR" }, - { 1, 3, 0, 0, "SAD" }, - { 1, 2, 1, 0, "TXF" }, - { 1, 2, 1, 0, "TXQ" }, - { 0, 0, 0, 0, "CONT" }, - { 0, 0, 0, 0, "EMIT" }, - { 0, 0, 0, 0, "ENDPRIM" }, - { 0, 0, 0, 1, "BGNLOOP2" }, - { 0, 0, 0, 0, "BGNSUB" }, - { 0, 0, 0, 1, "ENDLOOP2" }, - { 0, 0, 0, 0, "ENDSUB" }, - { 1, 1, 0, 0, "NOISE1" }, - { 1, 1, 0, 0, "NOISE2" }, - { 1, 1, 0, 0, "NOISE3" }, - { 1, 1, 0, 0, "NOISE4" }, - { 0, 0, 0, 0, "NOP" }, - { 1, 2, 0, 0, "M4X3" }, - { 1, 2, 0, 0, "M3X4" }, - { 1, 2, 0, 0, "M3X3" }, - { 1, 2, 0, 0, "M3X2" }, - { 1, 1, 0, 0, "NRM4" }, - { 0, 1, 0, 0, "CALLNZ" }, - { 0, 1, 0, 0, "IFC" }, - { 0, 1, 0, 0, "BREAKC" }, - { 0, 0, 0, 0, "KIL" }, - { 0, 0, 0, 0, "END" }, - { 1, 1, 0, 0, "SWZ" } -}; - static const char *texture_names[TGSI_TEXTURE_COUNT] = { "UNKNOWN", @@ -871,7 +740,7 @@ parse_instruction( { uint i; uint saturate = TGSI_SAT_NONE; - const struct opcode_info *info; + const struct tgsi_opcode_info *info; struct tgsi_full_instruction inst; uint advance; @@ -881,7 +750,7 @@ parse_instruction( for (i = 0; i < TGSI_OPCODE_LAST; i++) { const char *cur = ctx->cur; - info = &opcode_info[i]; + info = tgsi_get_opcode_info( i ); if (str_match_no_case( &cur, info->mnemonic )) { if (str_match_no_case( &cur, "_SATNV" )) saturate = TGSI_SAT_MINUS_PLUS_ONE; -- cgit v1.2.3 From 2caaba8195d1019702246bd7f0c02aa95364a8bd Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 17:27:15 +0200 Subject: tgsi: Use tgsi_info to dump opcode names. --- src/gallium/auxiliary/tgsi/tgsi_dump.c | 130 +----------------------------- src/gallium/auxiliary/tgsi/tgsi_dump_c.c | 133 +------------------------------ 2 files changed, 6 insertions(+), 257 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index 29bb530b4d..4309d1bc76 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -28,6 +28,7 @@ #include "pipe/p_debug.h" #include "util/u_string.h" #include "tgsi_dump.h" +#include "tgsi_info.h" #include "tgsi_iterate.h" struct dump_ctx @@ -103,129 +104,6 @@ static const char *immediate_type_names[] = "FLT32" }; -static const char *opcode_names[TGSI_OPCODE_LAST] = -{ - "ARL", - "MOV", - "LIT", - "RCP", - "RSQ", - "EXP", - "LOG", - "MUL", - "ADD", - "DP3", - "DP4", - "DST", - "MIN", - "MAX", - "SLT", - "SGE", - "MAD", - "SUB", - "LERP", - "CND", - "CND0", - "DOT2ADD", - "INDEX", - "NEGATE", - "FRAC", - "CLAMP", - "FLOOR", - "ROUND", - "EXPBASE2", - "LOGBASE2", - "POWER", - "CROSSPRODUCT", - "MULTIPLYMATRIX", - "ABS", - "RCC", - "DPH", - "COS", - "DDX", - "DDY", - "KILP", - "PK2H", - "PK2US", - "PK4B", - "PK4UB", - "RFL", - "SEQ", - "SFL", - "SGT", - "SIN", - "SLE", - "SNE", - "STR", - "TEX", - "TXD", - "TXP", - "UP2H", - "UP2US", - "UP4B", - "UP4UB", - "X2D", - "ARA", - "ARR", - "BRA", - "CAL", - "RET", - "SSG", - "CMP", - "SCS", - "TXB", - "NRM", - "DIV", - "DP2", - "TXL", - "BRK", - "IF", - "LOOP", - "REP", - "ELSE", - "ENDIF", - "ENDLOOP", - "ENDREP", - "PUSHA", - "POPA", - "CEIL", - "I2F", - "NOT", - "TRUNC", - "SHL", - "SHR", - "AND", - "OR", - "MOD", - "XOR", - "SAD", - "TXF", - "TXQ", - "CONT", - "EMIT", - "ENDPRIM", - "BGNLOOP2", - "BGNSUB", - "ENDLOOP2", - "ENDSUB", - "NOISE1", - "NOISE2", - "NOISE3", - "NOISE4", - "NOP", - "M4X3", - "M3X4", - "M3X3", - "M3X2", - "NRM4", - "CALLNZ", - "IFC", - "BREAKC", - "KIL", - "END", - "SWZ" -}; - static const char *swizzle_names[] = { "x", @@ -444,7 +322,7 @@ iter_instruction( UID( instno ); CHR( ':' ); - ENM( inst->Instruction.Opcode, opcode_names ); + TXT( tgsi_get_opcode_info( inst->Instruction.Opcode )->mnemonic ); switch (inst->Instruction.Saturate) { case TGSI_SAT_NONE: @@ -622,10 +500,6 @@ tgsi_dump_str( util_strbuf_init(&sbuf, str, size); - /* sanity checks */ - assert( strcmp( opcode_names[TGSI_OPCODE_CONT], "CONT" ) == 0 ); - assert( strcmp( opcode_names[TGSI_OPCODE_END], "END" ) == 0 ); - ctx.iter.prolog = prolog; ctx.iter.iterate_instruction = iter_instruction; ctx.iter.iterate_declaration = iter_declaration; diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c index eabd74bd6d..1025866a25 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c @@ -29,8 +29,9 @@ #include "pipe/p_util.h" #include "util/u_string.h" #include "tgsi_dump_c.h" -#include "tgsi_parse.h" #include "tgsi_build.h" +#include "tgsi_info.h" +#include "tgsi_parse.h" static void dump_enum( @@ -104,128 +105,6 @@ static const char *TGSI_IMMS[] = "IMM_FLOAT32" }; -static const char *TGSI_OPCODES[TGSI_OPCODE_LAST] = -{ - "OPCODE_ARL", - "OPCODE_MOV", - "OPCODE_LIT", - "OPCODE_RCP", - "OPCODE_RSQ", - "OPCODE_EXP", - "OPCODE_LOG", - "OPCODE_MUL", - "OPCODE_ADD", - "OPCODE_DP3", - "OPCODE_DP4", - "OPCODE_DST", - "OPCODE_MIN", - "OPCODE_MAX", - "OPCODE_SLT", - "OPCODE_SGE", - "OPCODE_MAD", - "OPCODE_SUB", - "OPCODE_LERP", - "OPCODE_CND", - "OPCODE_CND0", - "OPCODE_DOT2ADD", - "OPCODE_INDEX", - "OPCODE_NEGATE", - "OPCODE_FRAC", - "OPCODE_CLAMP", - "OPCODE_FLOOR", - "OPCODE_ROUND", - "OPCODE_EXPBASE2", - "OPCODE_LOGBASE2", - "OPCODE_POWER", - "OPCODE_CROSSPRODUCT", - "OPCODE_MULTIPLYMATRIX", - "OPCODE_ABS", - "OPCODE_RCC", - "OPCODE_DPH", - "OPCODE_COS", - "OPCODE_DDX", - "OPCODE_DDY", - "OPCODE_KILP", - "OPCODE_PK2H", - "OPCODE_PK2US", - "OPCODE_PK4B", - "OPCODE_PK4UB", - "OPCODE_RFL", - "OPCODE_SEQ", - "OPCODE_SFL", - "OPCODE_SGT", - "OPCODE_SIN", - "OPCODE_SLE", - "OPCODE_SNE", - "OPCODE_STR", - "OPCODE_TEX", - "OPCODE_TXD", - "OPCODE_TXP", - "OPCODE_UP2H", - "OPCODE_UP2US", - "OPCODE_UP4B", - "OPCODE_UP4UB", - "OPCODE_X2D", - "OPCODE_ARA", - "OPCODE_ARR", - "OPCODE_BRA", - "OPCODE_CAL", - "OPCODE_RET", - "OPCODE_SSG", - "OPCODE_CMP", - "OPCODE_SCS", - "OPCODE_TXB", - "OPCODE_NRM", - "OPCODE_DIV", - "OPCODE_DP2", - "OPCODE_TXL", - "OPCODE_BRK", - "OPCODE_IF", - "OPCODE_LOOP", - "OPCODE_REP", - "OPCODE_ELSE", - "OPCODE_ENDIF", - "OPCODE_ENDLOOP", - "OPCODE_ENDREP", - "OPCODE_PUSHA", - "OPCODE_POPA", - "OPCODE_CEIL", - "OPCODE_I2F", - "OPCODE_NOT", - "OPCODE_TRUNC", - "OPCODE_SHL", - "OPCODE_SHR", - "OPCODE_AND", - "OPCODE_OR", - "OPCODE_MOD", - "OPCODE_XOR", - "OPCODE_SAD", - "OPCODE_TXF", - "OPCODE_TXQ", - "OPCODE_CONT", - "OPCODE_EMIT", - "OPCODE_ENDPRIM", - "OPCODE_BGNLOOP2", - "OPCODE_BGNSUB", - "OPCODE_ENDLOOP2", - "OPCODE_ENDSUB", - "OPCODE_NOISE1", - "OPCODE_NOISE2", - "OPCODE_NOISE3", - "OPCODE_NOISE4", - "OPCODE_NOP", - "OPCODE_M4X3", - "OPCODE_M3X4", - "OPCODE_M3X3", - "OPCODE_M3X2", - "OPCODE_NRM4", - "OPCODE_CALLNZ", - "OPCODE_IFC", - "OPCODE_BREAKC", - "OPCODE_KIL", - "OPCODE_END" -}; - static const char *TGSI_SATS[] = { "SAT_NONE", @@ -428,8 +307,8 @@ dump_instruction_verbose( { unsigned i; - TXT( "\nOpcode : " ); - ENM( inst->Instruction.Opcode, TGSI_OPCODES ); + TXT( "\nOpcode : OPCODE_" ); + TXT( tgsi_get_opcode_info( inst->Instruction.Opcode )->mnemonic ); if( deflt || fi->Instruction.Saturate != inst->Instruction.Saturate ) { TXT( "\nSaturate : " ); ENM( inst->Instruction.Saturate, TGSI_SATS ); @@ -770,10 +649,6 @@ tgsi_dump_c( uint deflt = flags & TGSI_DUMP_C_DEFAULT; uint instno = 0; - /* sanity checks */ - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_CONT], "OPCODE_CONT") == 0); - assert(strcmp(TGSI_OPCODES[TGSI_OPCODE_END], "OPCODE_END") == 0); - tgsi_parse_init( &parse, tokens ); TXT( "tgsi-dump begin -----------------" ); -- cgit v1.2.3 From 4b929b32d03a58d80cacbd63c172dbd7221c8a8f Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 17:31:26 +0200 Subject: tgsi: Validate instruction opcode and operand counts. --- src/gallium/auxiliary/tgsi/tgsi_sanity.c | 15 +++++++++++++++ 1 file changed, 15 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c index 2e3ec96b5b..0b5bdd6ba1 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c @@ -27,6 +27,7 @@ #include "pipe/p_debug.h" #include "tgsi_sanity.h" +#include "tgsi_info.h" #include "tgsi_iterate.h" #define MAX_REGISTERS 256 @@ -170,6 +171,7 @@ iter_instruction( struct tgsi_full_instruction *inst ) { struct sanity_check_ctx *ctx = (struct sanity_check_ctx *) iter; + const struct tgsi_opcode_info *info; uint i; /* There must be no other instructions after END. @@ -181,6 +183,19 @@ iter_instruction( ctx->index_of_END = ctx->num_instructions; } + info = tgsi_get_opcode_info( inst->Instruction.Opcode ); + if (info == NULL) { + report_error( ctx, "Invalid instruction opcode" ); + return TRUE; + } + + if (info->num_dst != inst->Instruction.NumDstRegs) { + report_error( ctx, "Invalid number of destination operands" ); + } + if (info->num_src != inst->Instruction.NumSrcRegs) { + report_error( ctx, "Invalid number of source operands" ); + } + /* Check destination and source registers' validity. * Mark the registers as used. */ -- cgit v1.2.3 From bfdb1d55d58d70044af9fcd6f8465179145581dd Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 19:28:10 +0200 Subject: tgsi: Fix typo. --- src/gallium/auxiliary/tgsi/tgsi_build.c | 4 ++-- src/gallium/auxiliary/tgsi/tgsi_build.h | 2 +- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c index 742ef14c35..050b448fe7 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/tgsi_build.c @@ -820,7 +820,7 @@ tgsi_build_instruction_ext_nv( unsigned cond_swizzle_z, unsigned cond_swizzle_w, unsigned cond_dst_update, - unsigned cond_flow_update, + unsigned cond_flow_enable, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ) @@ -837,7 +837,7 @@ tgsi_build_instruction_ext_nv( instruction_ext_nv.CondSwizzleZ = cond_swizzle_z; instruction_ext_nv.CondSwizzleW = cond_swizzle_w; instruction_ext_nv.CondDstUpdate = cond_dst_update; - instruction_ext_nv.CondFlowEnable = cond_flow_update; + instruction_ext_nv.CondFlowEnable = cond_flow_enable; prev_token->Extended = 1; instruction_grow( instruction, header ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h index ed25830248..6ae7f324f8 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.h +++ b/src/gallium/auxiliary/tgsi/tgsi_build.h @@ -170,7 +170,7 @@ tgsi_build_instruction_ext_nv( unsigned cond_swizzle_z, unsigned cond_swizzle_w, unsigned cond_dst_update, - unsigned cond_flow_update, + unsigned cond_flow_enable, struct tgsi_token *prev_token, struct tgsi_instruction *instruction, struct tgsi_header *header ); -- cgit v1.2.3 From 83a5a225d773c119857f58672fd51c1d425f21d3 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 19:31:13 +0200 Subject: tgsi: Use a homebrew version of toupper(). --- src/gallium/auxiliary/tgsi/tgsi_text.c | 27 +++++++++++++++++---------- 1 file changed, 17 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c index 7cdf1b67c5..9454563361 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_text.c +++ b/src/gallium/auxiliary/tgsi/tgsi_text.c @@ -51,11 +51,18 @@ static boolean is_digit_alpha_underscore( const char *cur ) return is_digit( cur ) || is_alpha_underscore( cur ); } +static boolean uprcase( char c ) +{ + if (c >= 'a' && c <= 'z') + return c += 'A' - 'a'; + return c; +} + static boolean str_match_no_case( const char **pcur, const char *str ) { const char *cur = *pcur; - while (*str != '\0' && *str == toupper( *cur )) { + while (*str != '\0' && *str == uprcase( *cur )) { str++; cur++; } @@ -131,7 +138,7 @@ static boolean parse_float( const char **pcur, float *val ) } if (!integral_part && !fractional_part) return FALSE; - if (toupper( *cur ) == 'E') { + if (uprcase( *cur ) == 'E') { cur++; if (*cur == '-' || *cur == '+') cur++; @@ -258,19 +265,19 @@ parse_opt_writemask( cur++; *writemask = TGSI_WRITEMASK_NONE; eat_opt_white( &cur ); - if (toupper( *cur ) == 'X') { + if (uprcase( *cur ) == 'X') { cur++; *writemask |= TGSI_WRITEMASK_X; } - if (toupper( *cur ) == 'Y') { + if (uprcase( *cur ) == 'Y') { cur++; *writemask |= TGSI_WRITEMASK_Y; } - if (toupper( *cur ) == 'Z') { + if (uprcase( *cur ) == 'Z') { cur++; *writemask |= TGSI_WRITEMASK_Z; } - if (toupper( *cur ) == 'W') { + if (uprcase( *cur ) == 'W') { cur++; *writemask |= TGSI_WRITEMASK_W; } @@ -516,13 +523,13 @@ parse_optional_swizzle( cur++; eat_opt_white( &cur ); for (i = 0; i < 4; i++) { - if (toupper( *cur ) == 'X') + if (uprcase( *cur ) == 'X') swizzle[i] = TGSI_SWIZZLE_X; - else if (toupper( *cur ) == 'Y') + else if (uprcase( *cur ) == 'Y') swizzle[i] = TGSI_SWIZZLE_Y; - else if (toupper( *cur ) == 'Z') + else if (uprcase( *cur ) == 'Z') swizzle[i] = TGSI_SWIZZLE_Z; - else if (toupper( *cur ) == 'W') + else if (uprcase( *cur ) == 'W') swizzle[i] = TGSI_SWIZZLE_W; else { if (*cur == '0') -- cgit v1.2.3 From 3a1af846fe13c2a9b8c24eb7e37e11a9b42668d5 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Wed, 13 Aug 2008 20:09:56 +0200 Subject: tgsi: Initial code for KILP, needs CCs working to be complete. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 24 +++++++++++++++++++++--- 1 file changed, 21 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index c4ba667d32..df2a7cc4b3 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1234,11 +1234,29 @@ static void exec_kilp(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst) { - uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + if (inst->InstructionExtNv.CondFlowEnable) { + uint swizzle[4]; + uint chan_index; + uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ - /* TODO: build kilmask from CC mask */ + swizzle[0] = inst->InstructionExtNv.CondSwizzleX; + swizzle[1] = inst->InstructionExtNv.CondSwizzleY; + swizzle[2] = inst->InstructionExtNv.CondSwizzleZ; + swizzle[3] = inst->InstructionExtNv.CondSwizzleW; - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; + for (chan_index = 0; chan_index < 4; chan_index++) + { + uint i; + + for (i = 0; i < 4; i++) { + /* TODO: evaluate the condition code */ + if (0) + kilmask |= 1 << i; + } + } + + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; + } } -- cgit v1.2.3 From f3048ad90ed2e4583f0f7aaf35a0f4aa581942dd Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 13 Aug 2008 14:59:48 -0600 Subject: gallium: in fallback_copy_texsubimage() create new tex surface for CPU read access Was trying to use the strb->surface but it's made for GPU read/write only. --- src/mesa/state_tracker/st_cb_texture.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 147ca5b1a2..fceb260d70 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -933,7 +933,12 @@ fallback_copy_texsubimage(GLcontext *ctx, struct pipe_texture *pt = stImage->pt; struct pipe_surface *src_surf, *dest_surf; - src_surf = strb->surface; + /* We'd use strb->surface, here but it's created for GPU read/write only */ + src_surf = pipe->screen->get_tex_surface( pipe->screen, + strb->texture, + 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_READ); + dest_surf = screen->get_tex_surface(screen, pt, face, level, destZ, PIPE_BUFFER_USAGE_CPU_WRITE); @@ -1016,6 +1021,7 @@ fallback_copy_texsubimage(GLcontext *ctx, } screen->tex_surface_release(screen, &dest_surf); + screen->tex_surface_release(screen, &src_surf); } -- cgit v1.2.3 From 19ff2326e94a7a773bcb1dd9bfd22197b999daf9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 13 Aug 2008 17:20:40 -0600 Subject: gallium: fix exec_kilp(), fix Exec/FuncMask test for TGSI_OPCODE_RET Fixes a few glean glsl regressions. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 14 ++++++++++---- 1 file changed, 10 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index df2a7cc4b3..a96209db1a 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1234,10 +1234,13 @@ static void exec_kilp(struct tgsi_exec_machine *mach, const struct tgsi_full_instruction *inst) { + uint kilmask; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + if (inst->InstructionExtNv.CondFlowEnable) { uint swizzle[4]; uint chan_index; - uint kilmask = 0; /* bit 0 = pixel 0, bit 1 = pixel 1, etc */ + + kilmask = 0x0; swizzle[0] = inst->InstructionExtNv.CondSwizzleX; swizzle[1] = inst->InstructionExtNv.CondSwizzleY; @@ -1254,9 +1257,12 @@ exec_kilp(struct tgsi_exec_machine *mach, kilmask |= 1 << i; } } - - mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; } + else { + /* "unconditional" kil */ + kilmask = mach->ExecMask; + } + mach->Temps[TEMP_KILMASK_I].xyzw[TEMP_KILMASK_C].u[0] |= kilmask; } @@ -2176,7 +2182,7 @@ exec_instruction( mach->FuncMask &= ~mach->ExecMask; UPDATE_EXEC_MASK(mach); - if (mach->ExecMask == 0x0) { + if (mach->FuncMask == 0x0) { /* really return now (otherwise, keep executing */ if (mach->CallStackTop == 0) { -- cgit v1.2.3 From 73467e1080d94e6d5cfb52f71c845c09a78adcd9 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Thu, 14 Aug 2008 09:52:15 +0100 Subject: check for winsys->destroy before calling --- src/gallium/drivers/i915simple/i915_context.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index e3d19017b5..e2bf5ab678 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -45,7 +45,7 @@ static void i915_destroy( struct pipe_context *pipe ) draw_destroy( i915->draw ); - if(i915->winsys) + if(i915->winsys->destroy) i915->winsys->destroy(i915->winsys); FREE( i915 ); -- cgit v1.2.3 From d756f9512d295531ff6a600c736a68ac2dcff58b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 13 Aug 2008 12:13:54 +0100 Subject: translate: Draw can request up to PIPE_MAX_ATTRIBS + 1 vertex elements. --- src/gallium/auxiliary/translate/translate.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h index c3b754a902..650cd81fa6 100644 --- a/src/gallium/auxiliary/translate/translate.h +++ b/src/gallium/auxiliary/translate/translate.h @@ -57,7 +57,7 @@ struct translate_element struct translate_key { unsigned output_stride; unsigned nr_elements; - struct translate_element element[PIPE_MAX_ATTRIBS]; + struct translate_element element[PIPE_MAX_ATTRIBS + 1]; }; -- cgit v1.2.3 From 0ecaa37e49ff3a745b821ba7027eebeed938547a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 13 Aug 2008 12:31:53 +0100 Subject: python: Bindings fixes. --- src/gallium/state_trackers/python/gallium.i | 2 ++ src/gallium/state_trackers/python/p_context.i | 32 ++++++++++++++++++--------- src/gallium/state_trackers/python/p_state.i | 6 +++++ src/gallium/state_trackers/python/st_device.h | 4 ++++ 4 files changed, 34 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 8d7a49cee7..641b19e940 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -55,6 +55,8 @@ %} +%include "typemaps.i" + %include "carrays.i" %array_class(unsigned char, ByteArray); %array_class(int, IntArray); diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i index 521a2dee07..a23c805852 100644 --- a/src/gallium/state_trackers/python/p_context.i +++ b/src/gallium/state_trackers/python/p_context.i @@ -110,7 +110,7 @@ struct st_context { { struct pipe_constant_buffer state; memset(&state, 0, sizeof(state)); - state.buffer = buffer->buffer; + state.buffer = buffer ? buffer->buffer : NULL; state.size = buffer->buffer->size; $self->pipe->set_constant_buffer($self->pipe, shader, index, &state); } @@ -154,23 +154,31 @@ struct st_context { state.pitch = pitch; state.max_index = max_index; state.buffer_offset = buffer_offset; - state.buffer = buffer->buffer; + state.buffer = buffer ? buffer->buffer : NULL; memcpy(&$self->vertex_buffers[index], &state, sizeof(state)); for(i = 0; i < PIPE_MAX_ATTRIBS; ++i) if(self->vertex_buffers[i].buffer) - num_vertex_buffers = i + 1; + $self->num_vertex_buffers = i + 1; $self->pipe->set_vertex_buffers($self->pipe, - num_vertex_buffers, + $self->num_vertex_buffers, $self->vertex_buffers); } void set_vertex_element(unsigned index, - const struct pipe_vertex_element *element) { + const struct pipe_vertex_element *element) + { memcpy(&$self->vertex_elements[index], element, sizeof(*element)); - $self->pipe->set_vertex_elements($self->pipe, PIPE_MAX_ATTRIBS, $self->vertex_elements); + } + + void set_vertex_elements(unsigned num) + { + $self->num_vertex_elements = num; + $self->pipe->set_vertex_elements($self->pipe, + $self->num_vertex_elements, + $self->vertex_elements); } /* @@ -183,8 +191,12 @@ struct st_context { void draw_elements( struct st_buffer *indexBuffer, unsigned indexSize, - unsigned mode, unsigned start, unsigned count) { - $self->pipe->draw_elements($self->pipe, indexBuffer->buffer, indexSize, mode, start, count); + unsigned mode, unsigned start, unsigned count) + { + $self->pipe->draw_elements($self->pipe, + indexBuffer->buffer, + indexSize, + mode, start, count); } void draw_vertices(unsigned prim, @@ -222,9 +234,9 @@ error1: } void - flush(void) { + flush(unsigned flags = 0) { struct pipe_fence_handle *fence = NULL; - $self->pipe->flush($self->pipe, PIPE_FLUSH_RENDER_CACHE, &fence); + $self->pipe->flush($self->pipe, flags | PIPE_FLUSH_RENDER_CACHE, &fence); /* TODO: allow asynchronous operation */ $self->pipe->winsys->fence_finish( $self->pipe->winsys, fence, 0 ); $self->pipe->winsys->fence_reference( $self->pipe->winsys, &fence, NULL ); diff --git a/src/gallium/state_trackers/python/p_state.i b/src/gallium/state_trackers/python/p_state.i index 9834229b4b..7f5760b3b6 100644 --- a/src/gallium/state_trackers/python/p_state.i +++ b/src/gallium/state_trackers/python/p_state.i @@ -34,9 +34,15 @@ %module gallium; +%ignore winsys; +%ignore pipe_vertex_buffer::buffer; + %include "pipe/p_state.h"; +%array_class(struct pipe_stencil_state, StencilArray); + + %extend pipe_framebuffer_state { pipe_framebuffer_state(void) { diff --git a/src/gallium/state_trackers/python/st_device.h b/src/gallium/state_trackers/python/st_device.h index 5b7adbe1a0..644c263b53 100644 --- a/src/gallium/state_trackers/python/st_device.h +++ b/src/gallium/state_trackers/python/st_device.h @@ -57,7 +57,11 @@ struct st_context { struct pipe_texture *default_texture; struct pipe_texture *sampler_textures[PIPE_MAX_SAMPLERS]; + + unsigned num_vertex_buffers; struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; + + unsigned num_vertex_elements; struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS]; }; -- cgit v1.2.3 From 50c6092355edf9e978a3c34cb816d4833bcf92fd Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 13 Aug 2008 12:41:19 +0100 Subject: python: Simplify st_winsys. --- src/gallium/state_trackers/python/st_device.c | 8 +++----- .../state_trackers/python/st_hardpipe_winsys.c | 18 +----------------- .../state_trackers/python/st_softpipe_winsys.c | 21 +++++---------------- src/gallium/state_trackers/python/st_winsys.h | 6 ------ 4 files changed, 9 insertions(+), 44 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index fbd56712e0..c82261e4e7 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -41,7 +41,7 @@ static void st_device_really_destroy(struct st_device *st_dev) { if(st_dev->screen) - st_dev->st_ws->screen_destroy(st_dev->screen); + st_dev->screen->destroy(st_dev->screen); FREE(st_dev); } @@ -61,9 +61,7 @@ st_device_create_from_st_winsys(const struct st_winsys *st_ws) struct st_device *st_dev; if(!st_ws->screen_create || - !st_ws->screen_destroy || - !st_ws->context_create || - !st_ws->context_destroy) + !st_ws->context_create) return NULL; st_dev = CALLOC_STRUCT(st_device); @@ -106,7 +104,7 @@ st_context_destroy(struct st_context *st_ctx) } if(st_ctx->pipe) - st_ctx->st_dev->st_ws->context_destroy(st_ctx->pipe); + st_ctx->pipe->destroy(st_ctx->pipe); for(i = 0; i < PIPE_MAX_SAMPLERS; ++i) pipe_texture_reference(&st_ctx->sampler_textures[i], NULL); diff --git a/src/gallium/state_trackers/python/st_hardpipe_winsys.c b/src/gallium/state_trackers/python/st_hardpipe_winsys.c index 1e04998232..8b33c70fd7 100644 --- a/src/gallium/state_trackers/python/st_hardpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_hardpipe_winsys.c @@ -42,13 +42,6 @@ extern void init_gallium(void); void (*force_init_gallium_linkage)(void) = &init_gallium; -static void -st_hardpipe_screen_destroy(struct pipe_screen *screen) -{ - st_softpipe_winsys.screen_destroy(screen); -} - - static struct pipe_screen * st_hardpipe_screen_create(void) { @@ -56,13 +49,6 @@ st_hardpipe_screen_create(void) } -static void -st_hardpipe_context_destroy(struct pipe_context *pipe) -{ - st_softpipe_winsys.context_destroy(pipe); -} - - static struct pipe_context * st_hardpipe_context_create(struct pipe_screen *screen) { @@ -72,7 +58,5 @@ st_hardpipe_context_create(struct pipe_screen *screen) const struct st_winsys st_hardpipe_winsys = { &st_hardpipe_screen_create, - &st_hardpipe_screen_destroy, - &st_hardpipe_context_create, - &st_hardpipe_context_destroy + &st_hardpipe_context_create }; diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c index 1fda70ca00..6ea3c9a5cf 100644 --- a/src/gallium/state_trackers/python/st_softpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c @@ -253,13 +253,9 @@ st_softpipe_fence_finish(struct pipe_winsys *winsys, } -static void -st_softpipe_screen_destroy(struct pipe_screen *screen) +static void +st_softpipe_destroy(struct pipe_winsys *winsys) { - struct pipe_winsys *winsys = screen->winsys; - - screen->destroy(screen); - FREE(winsys); } @@ -274,6 +270,8 @@ st_softpipe_screen_create(void) if(!winsys) return NULL; + winsys->destroy = st_softpipe_destroy; + winsys->buffer_create = st_softpipe_buffer_create; winsys->user_buffer_create = st_softpipe_user_buffer_create; winsys->buffer_map = st_softpipe_buffer_map; @@ -293,19 +291,12 @@ st_softpipe_screen_create(void) screen = softpipe_create_screen(winsys); if(!screen) - FREE(winsys); + st_softpipe_destroy(winsys); return screen; } -static void -st_softpipe_context_destroy(struct pipe_context *pipe) -{ - pipe->destroy(pipe); -} - - static struct pipe_context * st_softpipe_context_create(struct pipe_screen *screen) { @@ -315,7 +306,5 @@ st_softpipe_context_create(struct pipe_screen *screen) const struct st_winsys st_softpipe_winsys = { &st_softpipe_screen_create, - &st_softpipe_screen_destroy, &st_softpipe_context_create, - &st_softpipe_context_destroy }; diff --git a/src/gallium/state_trackers/python/st_winsys.h b/src/gallium/state_trackers/python/st_winsys.h index 43db8b6bff..b8cb612d86 100644 --- a/src/gallium/state_trackers/python/st_winsys.h +++ b/src/gallium/state_trackers/python/st_winsys.h @@ -39,14 +39,8 @@ struct st_winsys struct pipe_screen * (*screen_create)(void); - void - (*screen_destroy)(struct pipe_screen *screen); - struct pipe_context * (*context_create)(struct pipe_screen *screen); - - void - (*context_destroy)(struct pipe_context *pipe); }; -- cgit v1.2.3 From a8540ef9fb22bbedc788c374fda3354047df110f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 09:44:49 +0100 Subject: gallium: Catch errors from posix_memalign. --- src/gallium/include/pipe/p_util.h | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index eec413842f..4e1a996122 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -213,7 +213,8 @@ align_malloc(size_t bytes, uint alignment) { #if defined(HAVE_POSIX_MEMALIGN) void *mem; - (void) posix_memalign(& mem, alignment, bytes); + if(posix_memalign(& mem, alignment, bytes) != 0) + return NULL; return mem; #else char *ptr, *buf; -- cgit v1.2.3 From 449cb6b37b54bd6e4a7058e97739d9634ccefaa7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 10:10:53 +0100 Subject: mesa: posix_memalign requires alignment to be multiple of sizeof(void*) --- src/gallium/include/pipe/p_util.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 4e1a996122..b4ab70a3fc 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -213,6 +213,7 @@ align_malloc(size_t bytes, uint alignment) { #if defined(HAVE_POSIX_MEMALIGN) void *mem; + alignment = (alignment + (uint)sizeof(void*) - 1) & ~((uint)sizeof(void*) - 1); if(posix_memalign(& mem, alignment, bytes) != 0) return NULL; return mem; -- cgit v1.2.3 From c67f353c0f0d3872c330e5a8cdb13334a5e2e239 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 10:11:51 +0100 Subject: mesa: Don't use alignments < 4. --- src/mesa/state_tracker/st_atom_constbuf.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c index 2856e0f0e0..d3aadf5074 100644 --- a/src/mesa/state_tracker/st_atom_constbuf.c +++ b/src/mesa/state_tracker/st_atom_constbuf.c @@ -74,7 +74,7 @@ void st_upload_constants( struct st_context *st, * avoid gratuitous rendering synchronization. */ pipe_reference_buffer(pipe, &cbuf->buffer, NULL ); - cbuf->buffer = pipe_buffer_create(pipe, 1, PIPE_BUFFER_USAGE_CONSTANT, + cbuf->buffer = pipe_buffer_create(pipe, 16, PIPE_BUFFER_USAGE_CONSTANT, paramBytes ); if (0) -- cgit v1.2.3 From 3c90678ea69ee8be832e16d42a1b8049a49535e3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 10:46:38 +0100 Subject: trace: Separate the trace screen/context vs the original screen/context. --- src/gallium/drivers/trace/tr_context.c | 13 +++++-------- src/gallium/drivers/trace/tr_context.h | 3 ++- src/gallium/drivers/trace/tr_screen.c | 24 ++++++++++++++++++------ src/gallium/drivers/trace/tr_winsys.c | 23 +++++++---------------- src/gallium/drivers/trace/tr_winsys.h | 3 ++- 5 files changed, 34 insertions(+), 32 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index f82126370c..e43fc01c55 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -959,7 +959,8 @@ trace_context_destroy(struct pipe_context *_pipe) struct pipe_context * -trace_context_create(struct pipe_context *pipe) +trace_context_create(struct pipe_screen *screen, + struct pipe_context *pipe) { struct trace_stream *stream; struct trace_context *tr_ctx; @@ -971,8 +972,8 @@ trace_context_create(struct pipe_context *pipe) if(!tr_ctx) return NULL; - tr_ctx->base.winsys = pipe->winsys; - tr_ctx->base.screen = pipe->screen; + tr_ctx->base.winsys = screen->winsys; + tr_ctx->base.screen = screen; tr_ctx->base.destroy = trace_context_destroy; tr_ctx->base.set_edgeflags = trace_context_set_edgeflags; tr_ctx->base.draw_arrays = trace_context_draw_arrays; @@ -1017,11 +1018,7 @@ trace_context_create(struct pipe_context *pipe) tr_ctx->base.flush = trace_context_flush; tr_ctx->pipe = pipe; - tr_ctx->stream = stream = trace_winsys(pipe->winsys)->stream; - - /* We don't want to trace the internal pipe calls */ - pipe->winsys = trace_winsys(pipe->winsys)->winsys; - pipe->screen = trace_screen(pipe->screen)->screen; + tr_ctx->stream = stream = trace_screen(screen)->stream; trace_dump_call_begin(stream, "", "pipe_context_create"); trace_dump_arg_begin(stream, "screen"); diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h index 2c0b0c72e4..1aa822ba02 100644 --- a/src/gallium/drivers/trace/tr_context.h +++ b/src/gallium/drivers/trace/tr_context.h @@ -57,7 +57,8 @@ trace_context(struct pipe_context *pipe) struct pipe_context * -trace_context_create(struct pipe_context *pipe); +trace_context_create(struct pipe_screen *screen, + struct pipe_context *pipe); #endif /* TR_CONTEXT_H_ */ diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index 27c218039e..0e253123ae 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -338,7 +338,11 @@ trace_screen_destroy(struct pipe_screen *_screen) screen->destroy(screen); trace_dump_call_end(stream); - + + trace_dump_trace_end(stream); + + trace_stream_close(stream); + FREE(tr_scr); } @@ -348,6 +352,7 @@ trace_screen_create(struct pipe_screen *screen) { struct trace_stream *stream; struct trace_screen *tr_scr; + struct pipe_winsys *winsys; if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) return screen; @@ -356,7 +361,17 @@ trace_screen_create(struct pipe_screen *screen) if(!tr_scr) return NULL; - tr_scr->base.winsys = screen->winsys; + tr_scr->stream = stream = trace_stream_create("gallium", "trace"); + if(!tr_scr->stream) + return NULL; + + trace_dump_trace_begin(stream, 0); + + winsys = trace_winsys_create(stream, screen->winsys); + if(!winsys) + return NULL; + + tr_scr->base.winsys = winsys; tr_scr->base.destroy = trace_screen_destroy; tr_scr->base.get_name = trace_screen_get_name; tr_scr->base.get_vendor = trace_screen_get_vendor; @@ -372,10 +387,7 @@ trace_screen_create(struct pipe_screen *screen) tr_scr->base.surface_unmap = trace_screen_surface_unmap; tr_scr->screen = screen; - tr_scr->stream = stream = trace_winsys(screen->winsys)->stream; - - /* We don't want to trace the internal pipe calls */ - screen->winsys = trace_winsys(screen->winsys)->winsys; + tr_scr->stream = stream = trace_winsys(winsys)->stream; trace_dump_call_begin(stream, "", "pipe_screen_create"); trace_dump_arg_begin(stream, "winsys"); diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index eec84a1981..524049148d 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -393,28 +393,24 @@ trace_winsys_destroy(struct pipe_winsys *_winsys) trace_dump_arg(stream, ptr, winsys); - winsys->destroy(winsys); + /* + winsys->destroy(winsys); + */ trace_dump_call_end(stream); - trace_dump_trace_end(stream); - hash_table_destroy(tr_ws->buffer_maps); - trace_stream_close(tr_ws->stream); - FREE(tr_ws); } struct pipe_winsys * -trace_winsys_create(struct pipe_winsys *winsys) +trace_winsys_create(struct trace_stream *stream, + struct pipe_winsys *winsys) { - struct trace_stream *stream; - struct trace_winsys *tr_ws; - if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) - return winsys; + struct trace_winsys *tr_ws; tr_ws = CALLOC_STRUCT(trace_winsys); if(!tr_ws) @@ -436,18 +432,13 @@ trace_winsys_create(struct pipe_winsys *winsys) tr_ws->base.fence_finish = trace_winsys_fence_finish; tr_ws->winsys = winsys; - - tr_ws->stream = stream = trace_stream_create("gallium", "trace"); - if(!tr_ws->stream) - return NULL; + tr_ws->stream = stream; tr_ws->buffer_maps = hash_table_create(trace_buffer_hash, trace_buffer_compare); if(!tr_ws->buffer_maps) return NULL; - trace_dump_trace_begin(tr_ws->stream, 0); - trace_dump_call_begin(stream, "", "pipe_winsys_create"); trace_dump_ret(stream, ptr, winsys); trace_dump_call_end(stream); diff --git a/src/gallium/drivers/trace/tr_winsys.h b/src/gallium/drivers/trace/tr_winsys.h index a3576da867..704d2c57c8 100644 --- a/src/gallium/drivers/trace/tr_winsys.h +++ b/src/gallium/drivers/trace/tr_winsys.h @@ -60,7 +60,8 @@ trace_winsys(struct pipe_winsys *winsys) struct pipe_winsys * -trace_winsys_create(struct pipe_winsys *winsys); +trace_winsys_create(struct trace_stream *stream, + struct pipe_winsys *winsys); #endif /* TR_WINSYS_H_ */ -- cgit v1.2.3 From 969ff9e9bf3dfae08d087bbc47ce43f17d107bfc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 10:47:39 +0100 Subject: xlib: Use trace usage. --- src/gallium/winsys/xlib/xm_api.c | 5 ++++- src/gallium/winsys/xlib/xm_winsys.c | 15 ++++----------- 2 files changed, 8 insertions(+), 12 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c index 4e5441a13d..d2f92a0875 100644 --- a/src/gallium/winsys/xlib/xm_api.c +++ b/src/gallium/winsys/xlib/xm_api.c @@ -67,6 +67,7 @@ #include "state_tracker/st_public.h" #include "state_tracker/st_context.h" #include "pipe/p_defines.h" +#include "pipe/p_screen.h" #include "pipe/p_context.h" #include "xm_winsys_aub.h" @@ -833,7 +834,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) fail: if (c->st) st_destroy_context(c->st); - if (pipe) + else if (pipe) pipe->destroy(pipe); FREE(c); return NULL; @@ -844,7 +845,9 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list ) PUBLIC void XMesaDestroyContext( XMesaContext c ) { + struct pipe_screen *screen = c->st->pipe->screen; st_destroy_context(c->st); + screen->destroy(screen); _mesa_free(c); } diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 6071a5ad5e..3ab4c67cae 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -55,7 +55,6 @@ #endif #ifdef GALLIUM_TRACE -#include "trace/tr_winsys.h" #include "trace/tr_screen.h" #include "trace/tr_context.h" #endif @@ -651,11 +650,7 @@ xmesa_get_pipe_winsys(struct xmesa_visual *xm_vis) ws->base.get_name = xm_get_name; } -#ifdef GALLIUM_TRACE - return trace_winsys_create(&ws->base); -#else - return &ws->base; -#endif + return &ws->base; } @@ -684,14 +679,12 @@ xmesa_create_pipe_context(XMesaContext xmesa, uint pixelformat) { struct pipe_screen *screen = softpipe_create_screen(pws); -#ifdef GALLIUM_TRACE - screen = trace_screen_create(screen); -#endif - pipe = softpipe_create(screen, pws, NULL); #ifdef GALLIUM_TRACE - pipe = trace_context_create(pipe); + screen = trace_screen_create(screen); + + pipe = trace_context_create(screen, pipe); #endif } -- cgit v1.2.3 From 90ed1742d55a246ffa63e57f291b18a1a43e41ba Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 10:47:55 +0100 Subject: python: Remove unused var. --- src/gallium/state_trackers/python/p_context.i | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i index a23c805852..535783b641 100644 --- a/src/gallium/state_trackers/python/p_context.i +++ b/src/gallium/state_trackers/python/p_context.i @@ -147,7 +147,7 @@ struct st_context { unsigned buffer_offset, struct st_buffer *buffer) { - unsigned i, num_vertex_buffers; + unsigned i; struct pipe_vertex_buffer state; memset(&state, 0, sizeof(state)); -- cgit v1.2.3 From 92675f6e22a1caa11146c2e9f3b2fb2285fdccea Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 10:48:29 +0100 Subject: python: Fix typo in buffer::write. --- src/gallium/state_trackers/python/p_texture.i | 14 +++++++++++++- 1 file changed, 13 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/p_texture.i b/src/gallium/state_trackers/python/p_texture.i index fd38e0733f..06bc105465 100644 --- a/src/gallium/state_trackers/python/p_texture.i +++ b/src/gallium/state_trackers/python/p_texture.i @@ -183,8 +183,20 @@ struct st_buffer { struct pipe_winsys *winsys = $self->st_dev->screen->winsys; char *map; + assert($self->buffer->refcount); + + if(offset > $self->buffer->size) { + PyErr_SetString(PyExc_ValueError, "offset must be smaller than buffer size"); + return; + } + + if(offset + LENGTH > $self->buffer->size) { + PyErr_SetString(PyExc_ValueError, "data length must fit inside the buffer"); + return; + } + map = winsys->buffer_map(winsys, $self->buffer, PIPE_BUFFER_USAGE_CPU_WRITE); - if(!map) { + if(map) { memcpy(map + offset, STRING, LENGTH); winsys->buffer_unmap(winsys, $self->buffer); } -- cgit v1.2.3 From ffaa4e816fb1d7360e6c14cbbfd4df8e027b9573 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 10:58:23 +0100 Subject: python: Allow to use trace pipe driver with python. --- src/gallium/state_trackers/python/SConscript | 2 +- src/gallium/state_trackers/python/st_device.c | 35 ++++++++++++++++++++++----- src/gallium/state_trackers/python/st_device.h | 4 ++- 3 files changed, 33 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/SConscript b/src/gallium/state_trackers/python/SConscript index 973d96d55a..1581182aec 100644 --- a/src/gallium/state_trackers/python/SConscript +++ b/src/gallium/state_trackers/python/SConscript @@ -30,5 +30,5 @@ if 'python' in env['statetrackers']: source = [ 'st_hardpipe_winsys.c', ], - LIBS = [pyst, softpipe] + auxiliaries + env['LIBS'], + LIBS = [pyst, softpipe, trace] + auxiliaries + env['LIBS'], ) diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index c82261e4e7..a1889539dc 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -33,6 +33,9 @@ #include "pipe/p_inlines.h" #include "cso_cache/cso_context.h" #include "util/u_simple_shaders.h" +#include "trace/tr_screen.h" +#include "trace/tr_context.h" + #include "st_device.h" #include "st_winsys.h" @@ -71,9 +74,17 @@ st_device_create_from_st_winsys(const struct st_winsys *st_ws) st_dev->refcount = 1; st_dev->st_ws = st_ws; - st_dev->screen = st_ws->screen_create(); - if(!st_dev->screen) + st_dev->real_screen = st_ws->screen_create(); + if(!st_dev->real_screen) { + st_device_destroy(st_dev); + return NULL; + } + + st_dev->screen = trace_screen_create(st_dev->real_screen); + if(!st_dev->screen) { st_device_destroy(st_dev); + return NULL; + } return st_dev; } @@ -130,13 +141,23 @@ st_context_create(struct st_device *st_dev) st_ctx->st_dev = st_dev; ++st_dev->refcount; - st_ctx->pipe = st_dev->st_ws->context_create(st_dev->screen); - if(!st_ctx->pipe) + st_ctx->real_pipe = st_dev->st_ws->context_create(st_dev->real_screen); + if(!st_ctx->real_pipe) { st_context_destroy(st_ctx); + return NULL; + } + st_ctx->pipe = trace_context_create(st_dev->screen, st_ctx->real_pipe); + if(!st_ctx->pipe) { + st_context_destroy(st_ctx); + return NULL; + } + st_ctx->cso = cso_create_context(st_ctx->pipe); - if(!st_ctx->cso) + if(!st_ctx->cso) { st_context_destroy(st_ctx); + return NULL; + } /* disabled blending/masking */ { @@ -291,8 +312,10 @@ st_buffer_create(struct st_device *st_dev, st_buf->st_dev = st_dev; st_buf->buffer = winsys->buffer_create(winsys, alignment, usage, size); - if(!st_buf->buffer) + if(!st_buf->buffer) { st_buffer_destroy(st_buf); + return NULL; + } return st_buf; } diff --git a/src/gallium/state_trackers/python/st_device.h b/src/gallium/state_trackers/python/st_device.h index 644c263b53..7cfe6de9f6 100644 --- a/src/gallium/state_trackers/python/st_device.h +++ b/src/gallium/state_trackers/python/st_device.h @@ -48,6 +48,7 @@ struct st_buffer { struct st_context { struct st_device *st_dev; + struct pipe_context *real_pipe; struct pipe_context *pipe; struct cso_context *cso; @@ -68,7 +69,8 @@ struct st_context { struct st_device { const struct st_winsys *st_ws; - + + struct pipe_screen *real_screen; struct pipe_screen *screen; /* FIXME: we also need to refcount for textures and surfaces... */ -- cgit v1.2.3 From 468c9775cbc863b024f41fccad3d6bb967e10ea8 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 11:00:22 +0100 Subject: python/retrace: Application capable of replaying gallium traces. At the moment it is capable of replaying trivial/tri kind of apps. See README for status. --- src/gallium/state_trackers/python/retrace/README | 18 + .../state_trackers/python/retrace/interpreter.py | 418 +++++++++++++++++++++ src/gallium/state_trackers/python/retrace/model.py | 151 ++++++++ .../state_trackers/python/retrace/parser.py | 332 ++++++++++++++++ 4 files changed, 919 insertions(+) create mode 100644 src/gallium/state_trackers/python/retrace/README create mode 100755 src/gallium/state_trackers/python/retrace/interpreter.py create mode 100755 src/gallium/state_trackers/python/retrace/model.py create mode 100755 src/gallium/state_trackers/python/retrace/parser.py (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/README b/src/gallium/state_trackers/python/retrace/README new file mode 100644 index 0000000000..c0ad8ae6c1 --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/README @@ -0,0 +1,18 @@ +This is + + +To use it follow the instructions in src/gallium/drivers/trace/README and +src/gallium/state_trackers/python/README, and then do + + python src/gallium/state_trackers/python/samples/retrace/interpreter.py filename.trace + + + + +This is still work in progress: +- not everything is captured/replayed + - surface/textures contents +- any tiny error will result in a crash + +-- +Jose Fonseca diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py new file mode 100755 index 0000000000..ae62bc04a8 --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -0,0 +1,418 @@ +#!/usr/bin/env python +############################################################################# +# +# Copyright 2008 Tungsten Graphics, Inc. +# +# This program is free software: you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published +# by the Free Software Foundation, either version 3 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public License +# along with this program. If not, see . +# +############################################################################# + + +import sys +import gallium +import model +from parser import TraceParser + + +def make_image(surface): + pixels = gallium.FloatArray(surface.height*surface.width*4) + surface.get_tile_rgba(0, 0, surface.width, surface.height, pixels) + + import Image + outimage = Image.new( + mode='RGB', + size=(surface.width, surface.height), + color=(0,0,0)) + outpixels = outimage.load() + for y in range(0, surface.height): + for x in range(0, surface.width): + offset = (y*surface.width + x)*4 + r, g, b, a = [int(pixels[offset + ch]*255) for ch in range(4)] + outpixels[x, y] = r, g, b + return outimage + +def save_image(filename, surface): + outimage = make_image(surface) + outimage.save(filename, "PNG") + +def show_image(surface): + outimage = make_image(surface) + + import Tkinter as tk + from PIL import Image, ImageTk + root = tk.Tk() + + root.title('background image') + + image1 = ImageTk.PhotoImage(outimage) + w = image1.width() + h = image1.height() + x = 100 + y = 100 + root.geometry("%dx%d+%d+%d" % (w, h, x, y)) + panel1 = tk.Label(root, image=image1) + panel1.pack(side='top', fill='both', expand='yes') + panel1.image = image1 + root.mainloop() + + + + +class Struct: + """C-like struct""" + + # A basic Python class can pass as a C-like structure + pass + + +struct_factories = { + "pipe_blend_color": gallium.BlendColor, + "pipe_blend_state": gallium.Blend, + "pipe_clip_state": gallium.Clip, + #"pipe_constant_buffer": gallium.ConstantBuffer, + "pipe_depth_state": gallium.Depth, + "pipe_stencil_state": gallium.Stencil, + "pipe_alpha_state": gallium.Alpha, + "pipe_depth_stencil_alpha_state": gallium.DepthStencilAlpha, + "pipe_format_block": gallium.FormatBlock, + #"pipe_framebuffer_state": gallium.Framebuffer, + "pipe_poly_stipple": gallium.PolyStipple, + "pipe_rasterizer_state": gallium.Rasterizer, + "pipe_sampler_state": gallium.Sampler, + "pipe_scissor_state": gallium.Scissor, + #"pipe_shader_state": gallium.Shader, + #"pipe_vertex_buffer": gallium.VertexBuffer, + "pipe_vertex_element": gallium.VertexElement, + "pipe_viewport_state": gallium.Viewport, + #"pipe_texture": gallium.Texture, +} + + +member_array_factories = { + "pipe_rasterizer_state": {"sprite_coord_mode": gallium.ByteArray}, + "pipe_poly_stipple": {"stipple": gallium.UnsignedArray}, + "pipe_viewport_state": {"scale": gallium.FloatArray, "translate": gallium.FloatArray}, + "pipe_clip_state": {"ucp": gallium.FloatArray}, + "pipe_depth_stencil_alpha_state": {"stencil": gallium.StencilArray}, + "pipe_blend_color": {"color": gallium.FloatArray}, +} + + +class Translator(model.Visitor): + """Translate model arguments into regular Python objects""" + + def __init__(self, interpreter): + self.interpreter = interpreter + self.result = None + + def visit(self, node): + self.result = None + node.visit(self) + return self.result + + def visit_literal(self, node): + self.result = node.value + + def visit_named_constant(self, node): + # lookup the named constant in the gallium module + self.result = getattr(gallium, node.name) + + def visit_array(self, node): + array = [] + for element in node.elements: + array.append(self.visit(element)) + self.result = array + + def visit_struct(self, node): + struct_factory = struct_factories.get(node.name, Struct) + struct = struct_factory() + for member_name, member_node in node.members: + member_value = self.visit(member_node) + try: + array_factory = member_array_factories[node.name][member_name] + except KeyError: + pass + else: + assert isinstance(member_value, list) + array = array_factory(len(member_value)) + for i in range(len(member_value)): + array[i] = member_value[i] + member_value = array + #print node.name, member_name, member_value + assert isinstance(struct, Struct) or hasattr(struct, member_name) + setattr(struct, member_name, member_value) + self.result = struct + + def visit_pointer(self, node): + self.result = self.interpreter.lookup_object(node.address) + + +class Object: + + def __init__(self, interpreter, real): + self.interpreter = interpreter + self.real = real + + +class Global(Object): + + def __init__(self, interpreter, real): + self.interpreter = interpreter + self.real = real + + def pipe_winsys_create(self): + return Winsys(self.interpreter, gallium.Device()) + + def pipe_screen_create(self, winsys): + return Screen(self.interpreter, winsys.real) + + def pipe_context_create(self, screen): + context = screen.real.context_create() + return Context(self.interpreter, context) + + +class Winsys(Object): + + def __init__(self, interpreter, real): + self.interpreter = interpreter + self.real = real + + def get_name(self): + pass + + def buffer_create(self, alignment, usage, size): + return self.real.buffer_create(size, alignment, usage) + + def buffer_destroy(self, buffer): + pass + + def buffer_write(self, buffer, data, size): + buffer.write(data, size) + buffer.write(data, size) + + +class Screen(Object): + + def get_name(self): + pass + + def get_vendor(self): + pass + + def get_param(self, param): + pass + + def get_paramf(self, param): + pass + + def is_format_supported(self, format, target, tex_usage, geom_flags): + return self.real.is_format_supported(format, target, tex_usage, geom_flags) + + def texture_create(self, template): + return self.real.texture_create( + format = template.format, + width = template.width[0], + height = template.height[0], + depth = template.depth[0], + last_level = template.last_level, + target = template.target, + tex_usage = template.tex_usage, + ) + + def texture_release(self, texture): + self.interpreter.unregister_object(texture) + + def get_tex_surface(self, texture, face, level, zslice, usage): + return texture.get_surface(face, level, zslice, usage) + + def tex_surface_release(self, surface): + self.interpreter.unregister_object(surface) + + def surface_map(self, surface, flags): + return None + + def surface_unmap(self, surface): + pass + + +class Context(Object): + + def __init__(self, interpreter, real): + Object.__init__(self, interpreter, real) + self.cbufs = [] + self.zsbuf = None + + def create_blend_state(self, state): + return state + + def bind_blend_state(self, state): + self.real.set_blend(state) + + def create_sampler_state(self, state): + return state + + def bind_sampler_states(self, n, states): + for i in range(n): + self.real.set_sampler(i, states[i]) + + def create_rasterizer_state(self, state): + return state + + def bind_rasterizer_state(self, state): + self.real.set_rasterizer(state) + + def create_depth_stencil_alpha_state(self, state): + return state + + def bind_depth_stencil_alpha_state(self, state): + self.real.set_depth_stencil_alpha(state) + + def create_fs_state(self, state): + tokens = str(state.tokens) + shader = gallium.Shader(tokens) + return shader + + create_vs_state = create_fs_state + + def bind_fs_state(self, state): + self.real.set_fragment_shader(state) + + def bind_vs_state(self, state): + self.real.set_vertex_shader(state) + + def set_blend_color(self, state): + self.real.set_blend_color(state) + + def set_clip_state(self, state): + self.real.set_clip(state) + + def set_constant_buffer(self, shader, index, state): + self.real.set_constant_buffer(shader, index, state.buffer) + + def set_framebuffer_state(self, state): + _state = gallium.Framebuffer() + _state.width = state.width + _state.height = state.height + _state.num_cbufs = state.num_cbufs + for i in range(len(state.cbufs)): + _state.set_cbuf(i, state.cbufs[i]) + _state.set_zsbuf(state.zsbuf) + self.real.set_framebuffer(_state) + + self.cbufs = state.cbufs + self.zsbuf = state.zsbuf + + def set_polygon_stipple(self, state): + self.real.set_polygon_stipple(state) + + def set_scissor_state(self, state): + self.real.set_scissor(state) + + def set_viewport_state(self, state): + self.real.set_viewport(state) + + def set_sampler_textures(self, n, textures): + for i in range(n): + self.real.set_sampler_textures(textures[i]) + + def set_vertex_buffers(self, n, vbufs): + for i in range(n): + vbuf = vbufs[i] + self.real.set_vertex_buffer( + i, + pitch = vbuf.pitch, + max_index = vbuf.max_index, + buffer_offset = vbuf.buffer_offset, + buffer = vbuf.buffer, + ) + + def set_vertex_elements(self, n, elements): + for i in range(n): + self.real.set_vertex_element(i, elements[i]) + self.real.set_vertex_elements(n) + + def set_edgeflags(self, bitfield): + # FIXME + pass + + def draw_arrays(self, mode, start, count): + self.real.draw_arrays(mode, start, count) + self._update() + + def flush(self, flags, fence): + self.real.flush(flags) + + def clear(self, surface, value): + self.real.surface_clear(surface, value) + + def _update(self): + self.real.flush() + + if self.cbufs and self.cbufs[0]: + show_image(self.cbufs[0]) + + +class Interpreter: + + def __init__(self): + self.objects = {} + self.result = None + self.globl = Global(self, None) + + def register_object(self, address, object): + self.objects[address] = object + + def unregister_object(self, object): + # FIXME: + pass + + def lookup_object(self, address): + return self.objects[address] + + def interpret(self, trace): + for call in trace.calls: + self.interpret_call(call) + + def interpret_call(self, call): + sys.stderr.write("%s\n" % call) + + args = [self.interpret_arg(arg) for name, arg in call.args] + + if call.klass: + obj = args[0] + args = args[1:] + else: + obj = self.globl + + method = getattr(obj, call.method) + ret = method(*args) + + if call.ret and isinstance(call.ret, model.Pointer): + self.register_object(call.ret.address, ret) + + def interpret_arg(self, node): + translator = Translator(self) + return translator.visit(node) + + +def main(): + for arg in sys.argv[1:]: + parser = TraceParser(open(arg, 'rt')) + trace = parser.parse() + interpreter = Interpreter() + interpreter.interpret(trace) + + +if __name__ == '__main__': + main() diff --git a/src/gallium/state_trackers/python/retrace/model.py b/src/gallium/state_trackers/python/retrace/model.py new file mode 100755 index 0000000000..c17b5db9f8 --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/model.py @@ -0,0 +1,151 @@ +#!/usr/bin/env python +############################################################################# +# +# Copyright 2008 Tungsten Graphics, Inc. +# +# This program is free software: you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published +# by the Free Software Foundation, either version 3 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public License +# along with this program. If not, see . +# +############################################################################# + + +'''Trace data model.''' + + +class Node: + + def visit(self, visitor): + raise NotImplementedError + + +class Literal(Node): + + def __init__(self, value): + self.value = value + + def visit(self, visitor): + visitor.visit_literal(self) + + def __str__(self): + if isinstance(self.value, str) and len(self.value) > 32: + return '...' + else: + return repr(self.value) + + +class NamedConstant(Node): + + def __init__(self, name): + self.name = name + + def visit(self, visitor): + visitor.visit_named_constant(self) + + def __str__(self): + return self.name + + +class Array(Node): + + def __init__(self, elements): + self.elements = elements + + def visit(self, visitor): + visitor.visit_array(self) + + def __str__(self): + return '{' + ', '.join([str(value) for value in self.elements]) + '}' + + +class Struct(Node): + + def __init__(self, name, members): + self.name = name + self.members = members + + def visit(self, visitor): + visitor.visit_struct(self) + + def __str__(self): + return '{' + ', '.join([name + ' = ' + str(value) for name, value in self.members]) + '}' + + +class Pointer(Node): + + def __init__(self, address): + self.address = address + + def visit(self, visitor): + visitor.visit_pointer(self) + + def __str__(self): + return hex(self.address) + + +class Call: + + def __init__(self, klass, method, args, ret): + self.klass = klass + self.method = method + self.args = args + self.ret = ret + + def visit(self, visitor): + visitor.visit_call(self) + + def __str__(self): + s = self.method + if self.klass: + s = self.klass + '::' + s + s += '(' + ', '.join([name + ' = ' + str(value) for name, value in self.args]) + ')' + if self.ret is not None: + s += ' = ' + str(self.ret) + return s + + +class Trace: + + def __init__(self, calls): + self.calls = calls + + def visit(self, visitor): + visitor.visit_trace(self) + + def __str__(self): + return '\n'.join([str(call) for call in self.calls]) + + +class Visitor: + + def visit_literal(self, node): + raise NotImplementedError + + def visit_named_constant(self, node): + raise NotImplementedError + + def visit_array(self, node): + raise NotImplementedError + + def visit_struct(self, node): + raise NotImplementedError + + def visit_pointer(self, node): + raise NotImplementedError + + def visit_call(self, node): + raise NotImplementedError + + def visit_trace(self, node): + raise NotImplementedError + + diff --git a/src/gallium/state_trackers/python/retrace/parser.py b/src/gallium/state_trackers/python/retrace/parser.py new file mode 100755 index 0000000000..571d8f6eae --- /dev/null +++ b/src/gallium/state_trackers/python/retrace/parser.py @@ -0,0 +1,332 @@ +#!/usr/bin/env python +############################################################################# +# +# Copyright 2008 Tungsten Graphics, Inc. +# +# This program is free software: you can redistribute it and/or modify it +# under the terms of the GNU Lesser General Public License as published +# by the Free Software Foundation, either version 3 of the License, or +# (at your option) any later version. +# +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU Lesser General Public License for more details. +# +# You should have received a copy of the GNU Lesser General Public License +# along with this program. If not, see . +# +############################################################################# + + +import sys +import xml.parsers.expat +import binascii + +from model import * + + +ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF = range(4) + + +class XmlToken: + + def __init__(self, type, name_or_data, attrs = None, line = None, column = None): + assert type in (ELEMENT_START, ELEMENT_END, CHARACTER_DATA, EOF) + self.type = type + self.name_or_data = name_or_data + self.attrs = attrs + self.line = line + self.column = column + + def __str__(self): + if self.type == ELEMENT_START: + return '<' + self.name_or_data + ' ...>' + if self.type == ELEMENT_END: + return '' + if self.type == CHARACTER_DATA: + return self.name_or_data + if self.type == EOF: + return 'end of file' + assert 0 + + +class XmlTokenizer: + """Expat based XML tokenizer.""" + + def __init__(self, fp, skip_ws = True): + self.fp = fp + self.tokens = [] + self.index = 0 + self.final = False + self.skip_ws = skip_ws + + self.character_pos = 0, 0 + self.character_data = '' + + self.parser = xml.parsers.expat.ParserCreate() + self.parser.StartElementHandler = self.handle_element_start + self.parser.EndElementHandler = self.handle_element_end + self.parser.CharacterDataHandler = self.handle_character_data + + def handle_element_start(self, name, attributes): + self.finish_character_data() + line, column = self.pos() + token = XmlToken(ELEMENT_START, name, attributes, line, column) + self.tokens.append(token) + + def handle_element_end(self, name): + self.finish_character_data() + line, column = self.pos() + token = XmlToken(ELEMENT_END, name, None, line, column) + self.tokens.append(token) + + def handle_character_data(self, data): + if not self.character_data: + self.character_pos = self.pos() + self.character_data += data + + def finish_character_data(self): + if self.character_data: + if not self.skip_ws or not self.character_data.isspace(): + line, column = self.character_pos + token = XmlToken(CHARACTER_DATA, self.character_data, None, line, column) + self.tokens.append(token) + self.character_data = '' + + def next(self): + size = 16*1024 + while self.index >= len(self.tokens) and not self.final: + self.tokens = [] + self.index = 0 + data = self.fp.read(size) + self.final = len(data) < size + try: + self.parser.Parse(data, self.final) + except xml.parsers.expat.ExpatError, e: + #if e.code == xml.parsers.expat.errors.XML_ERROR_NO_ELEMENTS: + if e.code == 3: + pass + else: + raise e + if self.index >= len(self.tokens): + line, column = self.pos() + token = XmlToken(EOF, None, None, line, column) + else: + token = self.tokens[self.index] + self.index += 1 + return token + + def pos(self): + return self.parser.CurrentLineNumber, self.parser.CurrentColumnNumber + + +class TokenMismatch(Exception): + + def __init__(self, expected, found): + self.expected = expected + self.found = found + + def __str__(self): + return '%u:%u: %s expected, %s found' % (self.found.line, self.found.column, str(self.expected), str(self.found)) + + + +class XmlParser: + """Base XML document parser.""" + + def __init__(self, fp): + self.tokenizer = XmlTokenizer(fp) + self.consume() + + def consume(self): + self.token = self.tokenizer.next() + + def match_element_start(self, name): + return self.token.type == ELEMENT_START and self.token.name_or_data == name + + def match_element_end(self, name): + return self.token.type == ELEMENT_END and self.token.name_or_data == name + + def element_start(self, name): + while self.token.type == CHARACTER_DATA: + self.consume() + if self.token.type != ELEMENT_START: + raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token) + if self.token.name_or_data != name: + raise TokenMismatch(XmlToken(ELEMENT_START, name), self.token) + attrs = self.token.attrs + self.consume() + return attrs + + def element_end(self, name): + while self.token.type == CHARACTER_DATA: + self.consume() + if self.token.type != ELEMENT_END: + raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token) + if self.token.name_or_data != name: + raise TokenMismatch(XmlToken(ELEMENT_END, name), self.token) + self.consume() + + def character_data(self, strip = True): + data = '' + while self.token.type == CHARACTER_DATA: + data += self.token.name_or_data + self.consume() + if strip: + data = data.strip() + return data + + +class TraceParser(XmlParser): + + def parse(self): + self.element_start('trace') + calls = [] + while self.token.type not in (ELEMENT_END, EOF): + calls.append(self.parse_call()) + if self.token.type != EOF: + self.element_end('trace') + return Trace(calls) + + def parse_call(self): + attrs = self.element_start('call') + klass = attrs['class'] + method = attrs['method'] + args = [] + ret = None + while self.token.type == ELEMENT_START: + if self.token.name_or_data == 'arg': + arg = self.parse_arg() + args.append(arg) + elif self.token.name_or_data == 'ret': + ret = self.parse_ret() + elif self.token.name_or_data == 'call': + # ignore nested function calls + self.parse_call() + else: + raise TokenMismatch(" or ", self.token) + self.element_end('call') + + return Call(klass, method, args, ret) + + def parse_arg(self): + attrs = self.element_start('arg') + name = attrs['name'] + value = self.parse_value() + self.element_end('arg') + + return name, value + + def parse_ret(self): + attrs = self.element_start('ret') + value = self.parse_value() + self.element_end('ret') + + return value + + def parse_value(self): + expected_tokens = ('null', 'bool', 'int', 'uint', 'float', 'string', 'enum', 'array', 'struct', 'ptr', 'bytes') + if self.token.type == ELEMENT_START: + if self.token.name_or_data in expected_tokens: + method = getattr(self, 'parse_' + self.token.name_or_data) + return method() + raise TokenMismatch(" or " .join(expected_tokens), self.token) + + def parse_null(self): + self.element_start('null') + self.element_end('null') + return Literal(None) + + def parse_bool(self): + self.element_start('bool') + value = int(self.character_data()) + self.element_end('bool') + return Literal(value) + + def parse_int(self): + self.element_start('int') + value = int(self.character_data()) + self.element_end('int') + return Literal(value) + + def parse_uint(self): + self.element_start('uint') + value = int(self.character_data()) + self.element_end('uint') + return Literal(value) + + def parse_float(self): + self.element_start('float') + value = float(self.character_data()) + self.element_end('float') + return Literal(value) + + def parse_enum(self): + self.element_start('enum') + name = self.character_data() + self.element_end('enum') + return NamedConstant(name) + + def parse_string(self): + self.element_start('string') + value = self.character_data() + self.element_end('string') + return Literal(value) + + def parse_bytes(self): + self.element_start('bytes') + value = binascii.a2b_hex(self.character_data()) + self.element_end('bytes') + return Literal(value) + + def parse_array(self): + self.element_start('array') + elems = [] + while self.token.type != ELEMENT_END: + elems.append(self.parse_elem()) + self.element_end('array') + return Array(elems) + + def parse_elem(self): + self.element_start('elem') + value = self.parse_value() + self.element_end('elem') + return value + + def parse_struct(self): + attrs = self.element_start('struct') + name = attrs['name'] + members = [] + while self.token.type != ELEMENT_END: + members.append(self.parse_member()) + self.element_end('struct') + return Struct(name, members) + + def parse_member(self): + attrs = self.element_start('member') + name = attrs['name'] + value = self.parse_value() + self.element_end('member') + + return name, value + + def parse_ptr(self): + self.element_start('ptr') + address = self.character_data() + self.element_end('ptr') + + address = int(address, 16) + + return Pointer(address) + + +def main(): + for arg in sys.argv[1:]: + parser = TraceParser(open(arg, 'rt')) + trace = parser.parse() + print trace + + +if __name__ == '__main__': + main() -- cgit v1.2.3 From 31bb6c0d17eb3265c9596bd25d5e6fabfefe4d61 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 11:00:40 +0100 Subject: python: Update status. --- src/gallium/state_trackers/python/README | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/README b/src/gallium/state_trackers/python/README index 8ab88ab2f8..8f45fb6d1b 100644 --- a/src/gallium/state_trackers/python/README +++ b/src/gallium/state_trackers/python/README @@ -24,9 +24,10 @@ and then try running which should show a triangle. -This is still in experimental phase, and there many limitations to what you can -do with from Python. - +This is still work in progress: +- errors are not handled properly and almost always result in crash +- state atoms with array members are awkward to set +- there no efficient way to view images -- Jose Fonseca -- cgit v1.2.3 From 19aee90179135387c3236c38b207cc47176226ad Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 11:00:49 +0100 Subject: trace: Update status. --- src/gallium/drivers/trace/README | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README index 81da610bd5..4b763f2488 100644 --- a/src/gallium/drivers/trace/README +++ b/src/gallium/drivers/trace/README @@ -12,7 +12,7 @@ To use do ensure the right libGL.so is being picked by doing - ldd `which glxingo` + ldd `which glxinfo` and then try running @@ -31,7 +31,9 @@ further transform it by doing echo '' >> gallium.??.trace -This is still work in progress. +This is still work in progress, namely: +- surface writes are not traced +- no way to know the start/end of a frame -- Jose Fonseca -- cgit v1.2.3 From 6c46b49d12b33b04546bee137f75ca7649e9a4ef Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 14 Aug 2008 10:53:59 -0600 Subject: gallium: added queries to determin GL_EXT_packed_depth_stencil support --- src/mesa/state_tracker/st_extensions.c | 20 ++++++++++++++++++++ 1 file changed, 20 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c index b7444b298c..60fd440ef7 100644 --- a/src/mesa/state_tracker/st_extensions.c +++ b/src/mesa/state_tracker/st_extensions.c @@ -218,6 +218,26 @@ void st_init_extensions(struct st_context *st) /*ctx->Extensions.ARB_shadow_ambient = GL_TRUE;*/ } + /* GL_EXT_packed_depth_stencil requires both the ability to render to + * a depth/stencil buffer and texture from depth/stencil source. + */ + if (screen->is_format_supported(screen, PIPE_FORMAT_Z24S8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0) && + screen->is_format_supported(screen, PIPE_FORMAT_Z24S8_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { + ctx->Extensions.EXT_packed_depth_stencil = GL_TRUE; + } + else if (screen->is_format_supported(screen, PIPE_FORMAT_S8Z24_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_DEPTH_STENCIL, 0) && + screen->is_format_supported(screen, PIPE_FORMAT_S8Z24_UNORM, + PIPE_TEXTURE_2D, + PIPE_TEXTURE_USAGE_SAMPLER, 0)) { + ctx->Extensions.EXT_packed_depth_stencil = GL_TRUE; + } + if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_SRGB, PIPE_TEXTURE_2D, PIPE_TEXTURE_USAGE_SAMPLER, 0)) { -- cgit v1.2.3 From 8f6d9e12be0be086ca2aab0b56dff8d2181addd9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 14 Aug 2008 15:38:09 -0600 Subject: gallium: use a default texture in update_textures(), update_samplers() when needed The default texture is used when the current fragment shader has texture sample instructions but the user has not provided/bound a texture. --- src/mesa/state_tracker/st_atom_sampler.c | 12 ++++++---- src/mesa/state_tracker/st_atom_texture.c | 34 ++++++++++++++++++--------- src/mesa/state_tracker/st_cb_texture.c | 40 ++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_texture.h | 4 ++++ src/mesa/state_tracker/st_context.c | 5 ++++ src/mesa/state_tracker/st_context.h | 2 ++ 6 files changed, 82 insertions(+), 15 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c index 3ba6a971f6..cef61fb55c 100644 --- a/src/mesa/state_tracker/st_atom_sampler.c +++ b/src/mesa/state_tracker/st_atom_sampler.c @@ -35,6 +35,7 @@ #include "main/macros.h" #include "st_context.h" +#include "st_cb_texture.h" #include "st_atom.h" #include "st_program.h" #include "pipe/p_context.h" @@ -125,6 +126,8 @@ update_samplers(struct st_context *st) st->state.num_samplers = 0; + /*printf("%s samplers used = 0x%x\n", __FUNCTION__, fs->Base.Base.SamplersUsed);*/ + /* loop over sampler units (aka tex image units) */ for (su = 0; su < st->ctx->Const.MaxTextureImageUnits; su++) { struct pipe_sampler_state *sampler = st->state.samplers + su; @@ -136,8 +139,9 @@ update_samplers(struct st_context *st) const struct gl_texture_object *texobj = st->ctx->Texture.Unit[texUnit]._Current; - if (!texobj) - continue; + if (!texobj) { + texobj = st_get_default_texture(st); + } sampler->wrap_s = gl_wrap_to_sp(texobj->WrapS); sampler->wrap_t = gl_wrap_to_sp(texobj->WrapT); @@ -184,11 +188,11 @@ update_samplers(struct st_context *st) st->state.num_samplers = su + 1; - /* XXX more sampler state here */ - + /*printf("%s su=%u non-null\n", __FUNCTION__, su);*/ cso_single_sampler(st->cso_context, su, sampler); } else { + /*printf("%s su=%u null\n", __FUNCTION__, su);*/ cso_single_sampler(st->cso_context, su, NULL); } } diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c index 1ec671ed48..fb03766ff5 100644 --- a/src/mesa/state_tracker/st_atom_texture.c +++ b/src/mesa/state_tracker/st_atom_texture.c @@ -49,6 +49,8 @@ update_textures(struct st_context *st) st->state.num_textures = 0; + /*printf("%s samplers used = 0x%x\n", __FUNCTION__, fprog->Base.SamplersUsed);*/ + for (su = 0; su < st->ctx->Const.MaxTextureCoordUnits; su++) { struct pipe_texture *pt = NULL; @@ -56,24 +58,34 @@ update_textures(struct st_context *st) const GLuint texUnit = fprog->Base.SamplerUnits[su]; struct gl_texture_object *texObj = st->ctx->Texture.Unit[texUnit]._Current; - struct st_texture_object *stObj = st_texture_object(texObj); - - if (texObj) { - GLboolean flush, retval; + struct st_texture_object *stObj; + GLboolean flush, retval; - retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); - if (!retval) { - /* out of mem */ - /* missing texture */ - continue; - } + if (!texObj) { + texObj = st_get_default_texture(st); + } + stObj = st_texture_object(texObj); - st->state.num_textures = su + 1; + retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush); + if (!retval) { + /* out of mem */ + continue; } + st->state.num_textures = su + 1; + pt = st_get_stobj_texture(stObj); } + /* + if (pt) { + printf("%s su=%u non-null\n", __FUNCTION__, su); + } + else { + printf("%s su=%u null\n", __FUNCTION__, su); + } + */ + pipe_texture_reference(&st->state.sampler_texture[su], pt); } diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index fceb260d70..f291531f81 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1461,6 +1461,46 @@ st_finalize_texture(GLcontext *ctx, } +/** + * Returns pointer to a default/dummy texture. + * This is typically used when the current shader has tex/sample instructions + * but the user has not provided a (any) texture(s). + */ +struct gl_texture_object * +st_get_default_texture(struct st_context *st) +{ + if (!st->default_texture) { + static const GLenum target = GL_TEXTURE_2D; + GLubyte pixels[16][16][4]; + struct gl_texture_object *texObj; + struct gl_texture_image *texImg; + + texObj = st->ctx->Driver.NewTextureObject(st->ctx, 0, target); + + texImg = _mesa_get_tex_image(st->ctx, texObj, target, 0); + + _mesa_init_teximage_fields(st->ctx, target, texImg, + 16, 16, 1, 0, /* w, h, d, border */ + GL_RGBA); + + st_TexImage(st->ctx, 2, target, + 0, GL_RGBA, /* level, intformat */ + 16, 16, 1, 0, /* w, h, d, border */ + GL_RGBA, GL_UNSIGNED_BYTE, pixels, + &st->ctx->DefaultPacking, + texObj, texImg, + 0, 0); + + texObj->MinFilter = GL_NEAREST; + texObj->MagFilter = GL_NEAREST; + texObj->_Complete = GL_TRUE; + + st->default_texture = texObj; + } + return st->default_texture; +} + + void st_init_texture_functions(struct dd_function_table *functions) { diff --git a/src/mesa/state_tracker/st_cb_texture.h b/src/mesa/state_tracker/st_cb_texture.h index 843745fcd6..f1fe0339cd 100644 --- a/src/mesa/state_tracker/st_cb_texture.h +++ b/src/mesa/state_tracker/st_cb_texture.h @@ -37,6 +37,10 @@ st_finalize_texture(GLcontext *ctx, GLboolean *needFlush); +extern struct gl_texture_object * +st_get_default_texture(struct st_context *st); + + extern void st_init_texture_functions(struct dd_function_table *functions); diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 6e4a376d44..d014ace59f 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -199,6 +199,11 @@ static void st_destroy_context_priv( struct st_context *st ) } } + if (st->default_texture) { + st->ctx->Driver.DeleteTexture(st->ctx, st->default_texture); + st->default_texture = NULL; + } + free( st ); } diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h index 96333902a9..4314d9af5c 100644 --- a/src/mesa/state_tracker/st_context.h +++ b/src/mesa/state_tracker/st_context.h @@ -132,6 +132,8 @@ struct st_context struct st_vertex_program *vp; /**< Currently bound vertex program */ struct st_fragment_program *fp; /**< Currently bound fragment program */ + struct gl_texture_object *default_texture; + struct { struct gl_program_cache *cache; struct st_fragment_program *program; /**< cur pixel transfer prog */ -- cgit v1.2.3 From a7b818d53a95b549bbff942a9cb91272e0799dd5 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Thu, 14 Aug 2008 15:44:09 -0600 Subject: gallium: init default texture image to gray --- src/mesa/state_tracker/st_cb_texture.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index f291531f81..cf194a937f 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1475,6 +1475,9 @@ st_get_default_texture(struct st_context *st) struct gl_texture_object *texObj; struct gl_texture_image *texImg; + /* init image to gray */ + memset(pixels, 127, sizeof(pixels)); + texObj = st->ctx->Driver.NewTextureObject(st->ctx, 0, target); texImg = _mesa_get_tex_image(st->ctx, texObj, target, 0); -- cgit v1.2.3 From df3d694851fd99b6ea88c339b5153944824d3d3a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 11:15:13 +0100 Subject: python/retrace: Commit unsaved changes to the doc. --- src/gallium/state_trackers/python/retrace/README | 5 ++--- 1 file changed, 2 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/README b/src/gallium/state_trackers/python/retrace/README index c0ad8ae6c1..822cd11404 100644 --- a/src/gallium/state_trackers/python/retrace/README +++ b/src/gallium/state_trackers/python/retrace/README @@ -1,4 +1,5 @@ -This is +This is an application written in python to replay the traces captured by the + trace pipe driver. To use it follow the instructions in src/gallium/drivers/trace/README and @@ -7,8 +8,6 @@ src/gallium/state_trackers/python/README, and then do python src/gallium/state_trackers/python/samples/retrace/interpreter.py filename.trace - - This is still work in progress: - not everything is captured/replayed - surface/textures contents -- cgit v1.2.3 From daa481a9c0825018d2320dcd8d9e2d7ddcfd46a0 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 11:39:41 +0100 Subject: gallium: New PIPE_FLUSH_FRAME flag to signal the end of a frame. --- src/gallium/include/pipe/p_defines.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h index bdc6d4ef46..cda10a2f06 100644 --- a/src/gallium/include/pipe/p_defines.h +++ b/src/gallium/include/pipe/p_defines.h @@ -211,6 +211,7 @@ enum pipe_texture_target { #define PIPE_FLUSH_RENDER_CACHE 0x1 #define PIPE_FLUSH_TEXTURE_CACHE 0x2 #define PIPE_FLUSH_SWAPBUFFERS 0x4 +#define PIPE_FLUSH_FRAME 0x8 /**< Mark the end of a frame */ /** -- cgit v1.2.3 From 2f34c851c01321b177bc3074755e97af270ec994 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 11:40:33 +0100 Subject: mesa: Issue PIPE_FLUSH_FRAME in glSwapBuffers/glFlush/glFinish. --- src/mesa/state_tracker/st_cb_flush.c | 2 +- src/mesa/state_tracker/st_framebuffer.c | 4 +++- 2 files changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_flush.c b/src/mesa/state_tracker/st_cb_flush.c index 5e866b0d41..d8f9537d2d 100644 --- a/src/mesa/state_tracker/st_cb_flush.c +++ b/src/mesa/state_tracker/st_cb_flush.c @@ -91,7 +91,7 @@ void st_finish( struct st_context *st ) { struct pipe_fence_handle *fence = NULL; - st_flush(st, PIPE_FLUSH_RENDER_CACHE, &fence); + st_flush(st, PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_FRAME, &fence); st->pipe->winsys->fence_finish(st->pipe->winsys, fence, 0); st->pipe->winsys->fence_reference(st->pipe->winsys, &fence, NULL); diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c index 039a9cfe89..4012cf73e5 100644 --- a/src/mesa/state_tracker/st_framebuffer.c +++ b/src/mesa/state_tracker/st_framebuffer.c @@ -265,7 +265,9 @@ st_notify_swapbuffers(struct st_framebuffer *stfb) if (ctx && ctx->DrawBuffer == &stfb->Base) { st_flush( ctx->st, - PIPE_FLUSH_RENDER_CACHE | PIPE_FLUSH_SWAPBUFFERS, + PIPE_FLUSH_RENDER_CACHE | + PIPE_FLUSH_SWAPBUFFERS | + PIPE_FLUSH_FRAME, NULL ); ctx->st->frontbuffer_status = FRONT_STATUS_COPY_OF_BACK; } -- cgit v1.2.3 From 196167e9d5c84f9f6dfe6f15b3e2f2c3ec6825dc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 12:50:52 +0100 Subject: trace: Make stream a global variable. This not only simplifies the code, but allows to use atexit() to ensure the log is closed when applications don't exit cleanly. --- src/gallium/drivers/trace/README | 11 +- src/gallium/drivers/trace/tr_context.c | 497 +++++++++++++++------------------ src/gallium/drivers/trace/tr_context.h | 5 - src/gallium/drivers/trace/tr_dump.c | 337 +++++++++++----------- src/gallium/drivers/trace/tr_dump.h | 120 ++++---- src/gallium/drivers/trace/tr_screen.c | 196 ++++++------- src/gallium/drivers/trace/tr_screen.h | 5 - src/gallium/drivers/trace/tr_state.c | 497 ++++++++++++++++----------------- src/gallium/drivers/trace/tr_state.h | 59 ++-- src/gallium/drivers/trace/tr_stream.h | 5 +- src/gallium/drivers/trace/tr_winsys.c | 185 ++++++------ src/gallium/drivers/trace/tr_winsys.h | 6 +- 12 files changed, 906 insertions(+), 1017 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README index 4b763f2488..5752448b93 100644 --- a/src/gallium/drivers/trace/README +++ b/src/gallium/drivers/trace/README @@ -20,16 +20,7 @@ and then try running which should create a gallium.*.trace file, which is an XML file. You can view copying trace.xsl and trace.css to the same directory, and opening with a -XSLT capable browser like Firefox or Internet Explorer. It often happens that -the trace file was not properly terminated, and a - - - -closing tag is missing from the file end. Add it before try to open or -further transform it by doing - - echo '' >> gallium.??.trace - +XSLT capable browser like Firefox or Internet Explorer. This is still work in progress, namely: - surface writes are not traced diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index e43fc01c55..868b4f010d 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -26,12 +26,10 @@ **************************************************************************/ #include "pipe/p_util.h" +#include "pipe/p_screen.h" -#include "tr_stream.h" #include "tr_dump.h" #include "tr_state.h" -#include "tr_winsys.h" -#include "tr_screen.h" #include "tr_context.h" @@ -40,18 +38,17 @@ trace_context_set_edgeflags(struct pipe_context *_pipe, const unsigned *bitfield) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_edgeflags"); + trace_dump_call_begin("pipe_context", "set_edgeflags"); - trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(ptr, pipe); /* FIXME: we don't know how big this array is */ - trace_dump_arg(stream, ptr, bitfield); + trace_dump_arg(ptr, bitfield); pipe->set_edgeflags(pipe, bitfield);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -60,22 +57,21 @@ trace_context_draw_arrays(struct pipe_context *_pipe, unsigned mode, unsigned start, unsigned count) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; boolean result; - trace_dump_call_begin(stream, "pipe_context", "draw_arrays"); + trace_dump_call_begin("pipe_context", "draw_arrays"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, uint, mode); - trace_dump_arg(stream, uint, start); - trace_dump_arg(stream, uint, count); + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, mode); + trace_dump_arg(uint, start); + trace_dump_arg(uint, count); result = pipe->draw_arrays(pipe, mode, start, count);; - trace_dump_ret(stream, bool, result); + trace_dump_ret(bool, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -88,24 +84,23 @@ trace_context_draw_elements(struct pipe_context *_pipe, unsigned mode, unsigned start, unsigned count) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; boolean result; - trace_dump_call_begin(stream, "pipe_context", "draw_elements"); + trace_dump_call_begin("pipe_context", "draw_elements"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, indexBuffer); - trace_dump_arg(stream, uint, indexSize); - trace_dump_arg(stream, uint, mode); - trace_dump_arg(stream, uint, start); - trace_dump_arg(stream, uint, count); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, indexBuffer); + trace_dump_arg(uint, indexSize); + trace_dump_arg(uint, mode); + trace_dump_arg(uint, start); + trace_dump_arg(uint, count); result = pipe->draw_elements(pipe, indexBuffer, indexSize, mode, start, count);; - trace_dump_ret(stream, bool, result); + trace_dump_ret(bool, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -122,29 +117,28 @@ trace_context_draw_range_elements(struct pipe_context *_pipe, unsigned count) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; boolean result; - trace_dump_call_begin(stream, "pipe_context", "draw_range_elements"); + trace_dump_call_begin("pipe_context", "draw_range_elements"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, indexBuffer); - trace_dump_arg(stream, uint, indexSize); - trace_dump_arg(stream, uint, minIndex); - trace_dump_arg(stream, uint, maxIndex); - trace_dump_arg(stream, uint, mode); - trace_dump_arg(stream, uint, start); - trace_dump_arg(stream, uint, count); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, indexBuffer); + trace_dump_arg(uint, indexSize); + trace_dump_arg(uint, minIndex); + trace_dump_arg(uint, maxIndex); + trace_dump_arg(uint, mode); + trace_dump_arg(uint, start); + trace_dump_arg(uint, count); result = pipe->draw_range_elements(pipe, indexBuffer, indexSize, minIndex, maxIndex, mode, start, count); - trace_dump_ret(stream, bool, result); + trace_dump_ret(bool, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -155,20 +149,19 @@ trace_context_create_query(struct pipe_context *_pipe, unsigned query_type) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; struct pipe_query *result; - trace_dump_call_begin(stream, "pipe_context", "create_query"); + trace_dump_call_begin("pipe_context", "create_query"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, uint, query_type); + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, query_type); result = pipe->create_query(pipe, query_type);; - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -179,17 +172,16 @@ trace_context_destroy_query(struct pipe_context *_pipe, struct pipe_query *query) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "destroy_query"); + trace_dump_call_begin("pipe_context", "destroy_query"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, query); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, query); pipe->destroy_query(pipe, query);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -198,17 +190,16 @@ trace_context_begin_query(struct pipe_context *_pipe, struct pipe_query *query) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "begin_query"); + trace_dump_call_begin("pipe_context", "begin_query"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, query); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, query); pipe->begin_query(pipe, query);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -217,17 +208,16 @@ trace_context_end_query(struct pipe_context *_pipe, struct pipe_query *query) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "end_query"); + trace_dump_call_begin("pipe_context", "end_query"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, query); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, query); pipe->end_query(pipe, query); - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -238,22 +228,21 @@ trace_context_get_query_result(struct pipe_context *_pipe, uint64 *presult) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; uint64 result; boolean _result; - trace_dump_call_begin(stream, "pipe_context", "get_query_result"); + trace_dump_call_begin("pipe_context", "get_query_result"); - trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(ptr, pipe); _result = pipe->get_query_result(pipe, query, wait, presult);; result = *presult; - trace_dump_arg(stream, uint, result); - trace_dump_ret(stream, bool, _result); + trace_dump_arg(uint, result); + trace_dump_ret(bool, _result); - trace_dump_call_end(stream); + trace_dump_call_end(); return _result; } @@ -264,20 +253,19 @@ trace_context_create_blend_state(struct pipe_context *_pipe, const struct pipe_blend_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; - trace_dump_call_begin(stream, "pipe_context", "create_blend_state"); + trace_dump_call_begin("pipe_context", "create_blend_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, blend_state, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(blend_state, state); result = pipe->create_blend_state(pipe, state);; - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -288,17 +276,16 @@ trace_context_bind_blend_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "bind_blend_state"); + trace_dump_call_begin("pipe_context", "bind_blend_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->bind_blend_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -307,17 +294,16 @@ trace_context_delete_blend_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "delete_blend_state"); + trace_dump_call_begin("pipe_context", "delete_blend_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->delete_blend_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -326,20 +312,19 @@ trace_context_create_sampler_state(struct pipe_context *_pipe, const struct pipe_sampler_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; - trace_dump_call_begin(stream, "pipe_context", "create_sampler_state"); + trace_dump_call_begin("pipe_context", "create_sampler_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); result = pipe->create_sampler_state(pipe, state);; - trace_dump_ret(stream, sampler_state, result); + trace_dump_ret(sampler_state, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -350,18 +335,17 @@ trace_context_bind_sampler_states(struct pipe_context *_pipe, unsigned num_states, void **states) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "bind_sampler_states"); + trace_dump_call_begin("pipe_context", "bind_sampler_states"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, uint, num_states); - trace_dump_arg_array(stream, ptr, states, num_states); + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, num_states); + trace_dump_arg_array(ptr, states, num_states); pipe->bind_sampler_states(pipe, num_states, states);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -370,17 +354,16 @@ trace_context_delete_sampler_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "delete_sampler_state"); + trace_dump_call_begin("pipe_context", "delete_sampler_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->delete_sampler_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -389,20 +372,19 @@ trace_context_create_rasterizer_state(struct pipe_context *_pipe, const struct pipe_rasterizer_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; - trace_dump_call_begin(stream, "pipe_context", "create_rasterizer_state"); + trace_dump_call_begin("pipe_context", "create_rasterizer_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, rasterizer_state, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(rasterizer_state, state); result = pipe->create_rasterizer_state(pipe, state);; - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -413,17 +395,16 @@ trace_context_bind_rasterizer_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "bind_rasterizer_state"); + trace_dump_call_begin("pipe_context", "bind_rasterizer_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->bind_rasterizer_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -432,17 +413,16 @@ trace_context_delete_rasterizer_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "delete_rasterizer_state"); + trace_dump_call_begin("pipe_context", "delete_rasterizer_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->delete_rasterizer_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -451,20 +431,19 @@ trace_context_create_depth_stencil_alpha_state(struct pipe_context *_pipe, const struct pipe_depth_stencil_alpha_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; - trace_dump_call_begin(stream, "pipe_context", "create_depth_stencil_alpha_state"); + trace_dump_call_begin("pipe_context", "create_depth_stencil_alpha_state"); result = pipe->create_depth_stencil_alpha_state(pipe, state);; - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, depth_stencil_alpha_state, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(depth_stencil_alpha_state, state); - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -475,17 +454,16 @@ trace_context_bind_depth_stencil_alpha_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "bind_depth_stencil_alpha_state"); + trace_dump_call_begin("pipe_context", "bind_depth_stencil_alpha_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->bind_depth_stencil_alpha_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -494,17 +472,16 @@ trace_context_delete_depth_stencil_alpha_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "delete_depth_stencil_alpha_state"); + trace_dump_call_begin("pipe_context", "delete_depth_stencil_alpha_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->delete_depth_stencil_alpha_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -513,20 +490,19 @@ trace_context_create_fs_state(struct pipe_context *_pipe, const struct pipe_shader_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; - trace_dump_call_begin(stream, "pipe_context", "create_fs_state"); + trace_dump_call_begin("pipe_context", "create_fs_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, shader_state, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(shader_state, state); result = pipe->create_fs_state(pipe, state);; - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -537,17 +513,16 @@ trace_context_bind_fs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "bind_fs_state"); + trace_dump_call_begin("pipe_context", "bind_fs_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->bind_fs_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -556,17 +531,16 @@ trace_context_delete_fs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "delete_fs_state"); + trace_dump_call_begin("pipe_context", "delete_fs_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->delete_fs_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -575,20 +549,19 @@ trace_context_create_vs_state(struct pipe_context *_pipe, const struct pipe_shader_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; void * result; - trace_dump_call_begin(stream, "pipe_context", "create_vs_state"); + trace_dump_call_begin("pipe_context", "create_vs_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, shader_state, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(shader_state, state); result = pipe->create_vs_state(pipe, state);; - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -599,17 +572,16 @@ trace_context_bind_vs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "bind_vs_state"); + trace_dump_call_begin("pipe_context", "bind_vs_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->bind_vs_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -618,17 +590,16 @@ trace_context_delete_vs_state(struct pipe_context *_pipe, void *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "delete_vs_state"); + trace_dump_call_begin("pipe_context", "delete_vs_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, state); pipe->delete_vs_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -637,17 +608,16 @@ trace_context_set_blend_color(struct pipe_context *_pipe, const struct pipe_blend_color *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_blend_color"); + trace_dump_call_begin("pipe_context", "set_blend_color"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, blend_color, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(blend_color, state); pipe->set_blend_color(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -656,17 +626,16 @@ trace_context_set_clip_state(struct pipe_context *_pipe, const struct pipe_clip_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_clip_state"); + trace_dump_call_begin("pipe_context", "set_clip_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, clip_state, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(clip_state, state); pipe->set_clip_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -676,19 +645,18 @@ trace_context_set_constant_buffer(struct pipe_context *_pipe, const struct pipe_constant_buffer *buffer) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_constant_buffer"); + trace_dump_call_begin("pipe_context", "set_constant_buffer"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, uint, shader); - trace_dump_arg(stream, uint, index); - trace_dump_arg(stream, constant_buffer, buffer); + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, shader); + trace_dump_arg(uint, index); + trace_dump_arg(constant_buffer, buffer); pipe->set_constant_buffer(pipe, shader, index, buffer);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -697,17 +665,16 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe, const struct pipe_framebuffer_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_framebuffer_state"); + trace_dump_call_begin("pipe_context", "set_framebuffer_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, framebuffer_state, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(framebuffer_state, state); pipe->set_framebuffer_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -716,17 +683,16 @@ trace_context_set_polygon_stipple(struct pipe_context *_pipe, const struct pipe_poly_stipple *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_polygon_stipple"); + trace_dump_call_begin("pipe_context", "set_polygon_stipple"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, poly_stipple, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(poly_stipple, state); pipe->set_polygon_stipple(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -735,17 +701,16 @@ trace_context_set_scissor_state(struct pipe_context *_pipe, const struct pipe_scissor_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_scissor_state"); + trace_dump_call_begin("pipe_context", "set_scissor_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, scissor_state, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(scissor_state, state); pipe->set_scissor_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -754,17 +719,16 @@ trace_context_set_viewport_state(struct pipe_context *_pipe, const struct pipe_viewport_state *state) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_viewport_state"); + trace_dump_call_begin("pipe_context", "set_viewport_state"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, viewport_state, state); + trace_dump_arg(ptr, pipe); + trace_dump_arg(viewport_state, state); pipe->set_viewport_state(pipe, state);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -774,18 +738,17 @@ trace_context_set_sampler_textures(struct pipe_context *_pipe, struct pipe_texture **textures) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_sampler_textures"); + trace_dump_call_begin("pipe_context", "set_sampler_textures"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, uint, num_textures); - trace_dump_arg_array(stream, ptr, textures, num_textures); + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, num_textures); + trace_dump_arg_array(ptr, textures, num_textures); pipe->set_sampler_textures(pipe, num_textures, textures);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -795,21 +758,20 @@ trace_context_set_vertex_buffers(struct pipe_context *_pipe, const struct pipe_vertex_buffer *buffers) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_vertex_buffers"); + trace_dump_call_begin("pipe_context", "set_vertex_buffers"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, uint, num_buffers); + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, num_buffers); - trace_dump_arg_begin(stream, "buffers"); - trace_dump_struct_array(stream, vertex_buffer, buffers, num_buffers); - trace_dump_arg_end(stream); + trace_dump_arg_begin("buffers"); + trace_dump_struct_array(vertex_buffer, buffers, num_buffers); + trace_dump_arg_end(); pipe->set_vertex_buffers(pipe, num_buffers, buffers);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -819,21 +781,20 @@ trace_context_set_vertex_elements(struct pipe_context *_pipe, const struct pipe_vertex_element *elements) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "set_vertex_elements"); + trace_dump_call_begin("pipe_context", "set_vertex_elements"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, uint, num_elements); + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, num_elements); - trace_dump_arg_begin(stream, "elements"); - trace_dump_struct_array(stream, vertex_element, elements, num_elements); - trace_dump_arg_end(stream); + trace_dump_arg_begin("elements"); + trace_dump_struct_array(vertex_element, elements, num_elements); + trace_dump_arg_end(); pipe->set_vertex_elements(pipe, num_elements, elements);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -847,27 +808,26 @@ trace_context_surface_copy(struct pipe_context *_pipe, unsigned width, unsigned height) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "surface_copy"); + trace_dump_call_begin("pipe_context", "surface_copy"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, bool, do_flip); - trace_dump_arg(stream, ptr, dest); - trace_dump_arg(stream, uint, destx); - trace_dump_arg(stream, uint, desty); - trace_dump_arg(stream, ptr, src); - trace_dump_arg(stream, uint, srcx); - trace_dump_arg(stream, uint, srcy); - trace_dump_arg(stream, uint, width); - trace_dump_arg(stream, uint, height); + trace_dump_arg(ptr, pipe); + trace_dump_arg(bool, do_flip); + trace_dump_arg(ptr, dest); + trace_dump_arg(uint, destx); + trace_dump_arg(uint, desty); + trace_dump_arg(ptr, src); + trace_dump_arg(uint, srcx); + trace_dump_arg(uint, srcy); + trace_dump_arg(uint, width); + trace_dump_arg(uint, height); pipe->surface_copy(pipe, do_flip, dest, destx, desty, src, srcx, srcy, width, height); - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -879,21 +839,20 @@ trace_context_surface_fill(struct pipe_context *_pipe, unsigned value) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "surface_fill"); + trace_dump_call_begin("pipe_context", "surface_fill"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, dst); - trace_dump_arg(stream, uint, dstx); - trace_dump_arg(stream, uint, dsty); - trace_dump_arg(stream, uint, width); - trace_dump_arg(stream, uint, height); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, dst); + trace_dump_arg(uint, dstx); + trace_dump_arg(uint, dsty); + trace_dump_arg(uint, width); + trace_dump_arg(uint, height); pipe->surface_fill(pipe, dst, dstx, dsty, width, height, value);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -903,18 +862,17 @@ trace_context_clear(struct pipe_context *_pipe, unsigned clearValue) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "clear"); + trace_dump_call_begin("pipe_context", "clear"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, ptr, surface); - trace_dump_arg(stream, uint, clearValue); + trace_dump_arg(ptr, pipe); + trace_dump_arg(ptr, surface); + trace_dump_arg(uint, clearValue); pipe->clear(pipe, surface, clearValue);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -924,18 +882,17 @@ trace_context_flush(struct pipe_context *_pipe, struct pipe_fence_handle **fence) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "flush"); + trace_dump_call_begin("pipe_context", "flush"); - trace_dump_arg(stream, ptr, pipe); - trace_dump_arg(stream, uint, flags); - trace_dump_arg(stream, ptr, fence); + trace_dump_arg(ptr, pipe); + trace_dump_arg(uint, flags); + trace_dump_arg(ptr, fence); pipe->flush(pipe, flags, fence);; - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -943,16 +900,15 @@ static INLINE void trace_context_destroy(struct pipe_context *_pipe) { struct trace_context *tr_ctx = trace_context(_pipe); - struct trace_stream *stream = tr_ctx->stream; struct pipe_context *pipe = tr_ctx->pipe; - trace_dump_call_begin(stream, "pipe_context", "destroy"); + trace_dump_call_begin("pipe_context", "destroy"); - trace_dump_arg(stream, ptr, pipe); + trace_dump_arg(ptr, pipe); pipe->destroy(pipe); - trace_dump_call_end(stream); + trace_dump_call_end(); FREE(tr_ctx); } @@ -962,15 +918,14 @@ struct pipe_context * trace_context_create(struct pipe_screen *screen, struct pipe_context *pipe) { - struct trace_stream *stream; struct trace_context *tr_ctx; - if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) - return pipe; + if(!pipe) + goto error1; tr_ctx = CALLOC_STRUCT(trace_context); if(!tr_ctx) - return NULL; + goto error1; tr_ctx->base.winsys = screen->winsys; tr_ctx->base.screen = screen; @@ -1018,14 +973,16 @@ trace_context_create(struct pipe_screen *screen, tr_ctx->base.flush = trace_context_flush; tr_ctx->pipe = pipe; - tr_ctx->stream = stream = trace_screen(screen)->stream; - trace_dump_call_begin(stream, "", "pipe_context_create"); - trace_dump_arg_begin(stream, "screen"); - trace_dump_ptr(stream, pipe->screen); - trace_dump_arg_end(stream); - trace_dump_ret(stream, ptr, pipe); - trace_dump_call_end(stream); + trace_dump_call_begin("", "pipe_context_create"); + trace_dump_arg_begin("screen"); + trace_dump_ptr(pipe->screen); + trace_dump_arg_end(); + trace_dump_ret(ptr, pipe); + trace_dump_call_end(); return &tr_ctx->base; + +error1: + return pipe; } diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h index 1aa822ba02..679371e310 100644 --- a/src/gallium/drivers/trace/tr_context.h +++ b/src/gallium/drivers/trace/tr_context.h @@ -34,16 +34,11 @@ #include "pipe/p_context.h" -struct trace_stream; - - struct trace_context { struct pipe_context base; struct pipe_context *pipe; - - struct trace_stream *stream; }; diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index 5269c4ddc8..6ebb639b7c 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -38,6 +38,11 @@ * @author Jose Fonseca */ +#include "pipe/p_config.h" + +#if defined(PIPE_OS_LINUX) +#include +#endif #include "pipe/p_compiler.h" #include "util/u_string.h" @@ -46,318 +51,328 @@ #include "tr_dump.h" +static struct trace_stream *stream = NULL; + + static INLINE void -trace_dump_write(struct trace_stream *stream, const char *s) +trace_dump_write(const char *buf, size_t size) { - trace_stream_write(stream, s, strlen(s)); + if(stream) + trace_stream_write(stream, buf, size); } static INLINE void -trace_dump_writef(struct trace_stream *stream, const char *format, ...) +trace_dump_writes(const char *s) { - char buf[1024]; + trace_dump_write(s, strlen(s)); +} + + +static INLINE void +trace_dump_writef(const char *format, ...) +{ + static char buf[1024]; + unsigned len; va_list ap; va_start(ap, format); - util_vsnprintf(buf, sizeof(buf), format, ap); + len = util_vsnprintf(buf, sizeof(buf), format, ap); va_end(ap); - trace_dump_write(stream, buf); + trace_dump_write(buf, len); } static INLINE void -trace_dump_escape(struct trace_stream *stream, const char *str) +trace_dump_escape(const char *str) { const unsigned char *p = (const unsigned char *)str; unsigned char c; while((c = *p++) != 0) { if(c == '<') - trace_dump_write(stream, "<"); + trace_dump_writes("<"); else if(c == '>') - trace_dump_write(stream, ">"); + trace_dump_writes(">"); else if(c == '&') - trace_dump_write(stream, "&"); + trace_dump_writes("&"); else if(c == '\'') - trace_dump_write(stream, "'"); + trace_dump_writes("'"); else if(c == '\"') - trace_dump_write(stream, """); + trace_dump_writes("""); else if(c >= 0x20 && c <= 0x7e) - trace_dump_writef(stream, "%c", c); + trace_dump_writef("%c", c); else - trace_dump_writef(stream, "&#%u;", c); + trace_dump_writef("&#%u;", c); } } static INLINE void -trace_dump_indent(struct trace_stream *stream, unsigned level) +trace_dump_indent(unsigned level) { unsigned i; for(i = 0; i < level; ++i) - trace_dump_write(stream, "\t"); + trace_dump_writes("\t"); } static INLINE void -trace_dump_newline(struct trace_stream *stream) +trace_dump_newline(void) { - trace_dump_write(stream, "\n"); + trace_dump_writes("\n"); } static INLINE void -trace_dump_tag(struct trace_stream *stream, - const char *name) +trace_dump_tag(const char *name) { - trace_dump_write(stream, "<"); - trace_dump_write(stream, name); - trace_dump_write(stream, "/>"); + trace_dump_writes("<"); + trace_dump_writes(name); + trace_dump_writes("/>"); } static INLINE void -trace_dump_tag_begin(struct trace_stream *stream, - const char *name) +trace_dump_tag_begin(const char *name) { - trace_dump_write(stream, "<"); - trace_dump_write(stream, name); - trace_dump_write(stream, ">"); + trace_dump_writes("<"); + trace_dump_writes(name); + trace_dump_writes(">"); } static INLINE void -trace_dump_tag_begin1(struct trace_stream *stream, - const char *name, +trace_dump_tag_begin1(const char *name, const char *attr1, const char *value1) { - trace_dump_write(stream, "<"); - trace_dump_write(stream, name); - trace_dump_write(stream, " "); - trace_dump_write(stream, attr1); - trace_dump_write(stream, "='"); - trace_dump_escape(stream, value1); - trace_dump_write(stream, "'>"); + trace_dump_writes("<"); + trace_dump_writes(name); + trace_dump_writes(" "); + trace_dump_writes(attr1); + trace_dump_writes("='"); + trace_dump_escape(value1); + trace_dump_writes("'>"); } static INLINE void -trace_dump_tag_begin2(struct trace_stream *stream, - const char *name, +trace_dump_tag_begin2(const char *name, const char *attr1, const char *value1, const char *attr2, const char *value2) { - trace_dump_write(stream, "<"); - trace_dump_write(stream, name); - trace_dump_write(stream, " "); - trace_dump_write(stream, attr1); - trace_dump_write(stream, "=\'"); - trace_dump_escape(stream, value1); - trace_dump_write(stream, "\' "); - trace_dump_write(stream, attr2); - trace_dump_write(stream, "=\'"); - trace_dump_escape(stream, value2); - trace_dump_write(stream, "\'>"); + trace_dump_writes("<"); + trace_dump_writes(name); + trace_dump_writes(" "); + trace_dump_writes(attr1); + trace_dump_writes("=\'"); + trace_dump_escape(value1); + trace_dump_writes("\' "); + trace_dump_writes(attr2); + trace_dump_writes("=\'"); + trace_dump_escape(value2); + trace_dump_writes("\'>"); } static INLINE void -trace_dump_tag_begin3(struct trace_stream *stream, - const char *name, +trace_dump_tag_begin3(const char *name, const char *attr1, const char *value1, const char *attr2, const char *value2, const char *attr3, const char *value3) { - trace_dump_write(stream, "<"); - trace_dump_write(stream, name); - trace_dump_write(stream, " "); - trace_dump_write(stream, attr1); - trace_dump_write(stream, "=\'"); - trace_dump_escape(stream, value1); - trace_dump_write(stream, "\' "); - trace_dump_write(stream, attr2); - trace_dump_write(stream, "=\'"); - trace_dump_escape(stream, value2); - trace_dump_write(stream, "\' "); - trace_dump_write(stream, attr3); - trace_dump_write(stream, "=\'"); - trace_dump_escape(stream, value3); - trace_dump_write(stream, "\'>"); + trace_dump_writes("<"); + trace_dump_writes(name); + trace_dump_writes(" "); + trace_dump_writes(attr1); + trace_dump_writes("=\'"); + trace_dump_escape(value1); + trace_dump_writes("\' "); + trace_dump_writes(attr2); + trace_dump_writes("=\'"); + trace_dump_escape(value2); + trace_dump_writes("\' "); + trace_dump_writes(attr3); + trace_dump_writes("=\'"); + trace_dump_escape(value3); + trace_dump_writes("\'>"); } static INLINE void -trace_dump_tag_end(struct trace_stream *stream, - const char *name) -{ - trace_dump_write(stream, ""); -} - - -void trace_dump_trace_begin(struct trace_stream *stream, - unsigned version) -{ - trace_dump_write(stream, "\n"); - trace_dump_write(stream, "\n"); - trace_dump_writef(stream, "\n", version); -} - - -void trace_dump_trace_end(struct trace_stream *stream) -{ - trace_dump_write(stream, "\n"); +trace_dump_tag_end(const char *name) +{ + trace_dump_writes(""); +} + +boolean trace_dump_trace_begin() +{ + if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) + return FALSE; + + stream = trace_stream_create("gallium", "trace"); + if(!stream) + return FALSE; + + trace_dump_writes("\n"); + trace_dump_writes("\n"); + trace_dump_writes("\n"); + +#if defined(PIPE_OS_LINUX) + /* Linux applications rarely cleanup GL / Gallium resources so catch + * application exit here */ + atexit(trace_dump_trace_end); +#endif + + return TRUE; +} + +void trace_dump_trace_end(void) +{ + if(stream) { + trace_dump_writes("\n"); + trace_stream_close(stream); + stream = NULL; + } } -void trace_dump_call_begin(struct trace_stream *stream, - const char *klass, const char *method) +void trace_dump_call_begin(const char *klass, const char *method) { - trace_dump_indent(stream, 1); - trace_dump_tag_begin2(stream, "call", "class", klass, "method", method); - trace_dump_newline(stream); + trace_dump_indent(1); + trace_dump_tag_begin2("call", "class", klass, "method", method); + trace_dump_newline(); } -void trace_dump_call_end(struct trace_stream *stream) +void trace_dump_call_end(void) { - trace_dump_indent(stream, 1); - trace_dump_tag_end(stream, "call"); - trace_dump_newline(stream); + trace_dump_indent(1); + trace_dump_tag_end("call"); + trace_dump_newline(); } -void trace_dump_arg_begin(struct trace_stream *stream, - const char *name) +void trace_dump_arg_begin(const char *name) { - trace_dump_indent(stream, 2); - trace_dump_tag_begin1(stream, "arg", "name", name); + trace_dump_indent(2); + trace_dump_tag_begin1("arg", "name", name); } -void trace_dump_arg_end(struct trace_stream *stream) +void trace_dump_arg_end(void) { - trace_dump_tag_end(stream, "arg"); - trace_dump_newline(stream); + trace_dump_tag_end("arg"); + trace_dump_newline(); } -void trace_dump_ret_begin(struct trace_stream *stream) +void trace_dump_ret_begin(void) { - trace_dump_indent(stream, 2); - trace_dump_tag_begin(stream, "ret"); + trace_dump_indent(2); + trace_dump_tag_begin("ret"); } -void trace_dump_ret_end(struct trace_stream *stream) +void trace_dump_ret_end(void) { - trace_dump_tag_end(stream, "ret"); - trace_dump_newline(stream); + trace_dump_tag_end("ret"); + trace_dump_newline(); } -void trace_dump_bool(struct trace_stream *stream, - int value) +void trace_dump_bool(int value) { - trace_dump_writef(stream, "%c", value ? '1' : '0'); + trace_dump_writef("%c", value ? '1' : '0'); } -void trace_dump_int(struct trace_stream *stream, - long int value) +void trace_dump_int(long int value) { - trace_dump_writef(stream, "%li", value); + trace_dump_writef("%li", value); } -void trace_dump_uint(struct trace_stream *stream, - long unsigned value) +void trace_dump_uint(long unsigned value) { - trace_dump_writef(stream, "%lu", value); + trace_dump_writef("%lu", value); } -void trace_dump_float(struct trace_stream *stream, - double value) +void trace_dump_float(double value) { - trace_dump_writef(stream, "%g", value); + trace_dump_writef("%g", value); } -void trace_dump_bytes(struct trace_stream *stream, - const void *data, +void trace_dump_bytes(const void *data, long unsigned size) { static const char hex_table[16] = "0123456789ABCDEF"; const uint8_t *p = data; long unsigned i; - trace_dump_write(stream, ""); + trace_dump_writes(""); for(i = 0; i < size; ++i) { uint8_t byte = *p++; char hex[2]; hex[0] = hex_table[byte >> 4]; hex[1] = hex_table[byte & 0xf]; - trace_stream_write(stream, hex, 2); + trace_dump_write(hex, 2); } - trace_dump_write(stream, ""); + trace_dump_writes(""); } -void trace_dump_string(struct trace_stream *stream, - const char *str) +void trace_dump_string(const char *str) { - trace_dump_write(stream, ""); - trace_dump_escape(stream, str); - trace_dump_write(stream, ""); + trace_dump_writes(""); + trace_dump_escape(str); + trace_dump_writes(""); } -void trace_dump_enum(struct trace_stream *stream, - const char *value) +void trace_dump_enum(const char *value) { - trace_dump_write(stream, ""); - trace_dump_escape(stream, value); - trace_dump_write(stream, ""); + trace_dump_writes(""); + trace_dump_escape(value); + trace_dump_writes(""); } -void trace_dump_array_begin(struct trace_stream *stream) +void trace_dump_array_begin(void) { - trace_dump_write(stream, ""); + trace_dump_writes(""); } -void trace_dump_array_end(struct trace_stream *stream) +void trace_dump_array_end(void) { - trace_dump_write(stream, ""); + trace_dump_writes(""); } -void trace_dump_elem_begin(struct trace_stream *stream) +void trace_dump_elem_begin(void) { - trace_dump_write(stream, ""); + trace_dump_writes(""); } -void trace_dump_elem_end(struct trace_stream *stream) +void trace_dump_elem_end(void) { - trace_dump_write(stream, ""); + trace_dump_writes(""); } -void trace_dump_struct_begin(struct trace_stream *stream, - const char *name) +void trace_dump_struct_begin(const char *name) { - trace_dump_writef(stream, "", name); + trace_dump_writef("", name); } -void trace_dump_struct_end(struct trace_stream *stream) +void trace_dump_struct_end(void) { - trace_dump_write(stream, ""); + trace_dump_writes(""); } -void trace_dump_member_begin(struct trace_stream *stream, - const char *name) +void trace_dump_member_begin(const char *name) { - trace_dump_writef(stream, "", name); + trace_dump_writef("", name); } -void trace_dump_member_end(struct trace_stream *stream) +void trace_dump_member_end(void) { - trace_dump_write(stream, ""); + trace_dump_writes(""); } -void trace_dump_null(struct trace_stream *stream) +void trace_dump_null(void) { - trace_dump_write(stream, ""); + trace_dump_writes(""); } -void trace_dump_ptr(struct trace_stream *stream, - const void *value) +void trace_dump_ptr(const void *value) { if(value) - trace_dump_writef(stream, "%p", value); + trace_dump_writef("%p", value); else - trace_dump_null(stream); + trace_dump_null(); } diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index b2367c3288..0beb1023b1 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -4,7 +4,7 @@ * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation streams (the + * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sub license, and/or sell copies of the Software, and to @@ -34,100 +34,98 @@ #define TR_DUMP_H +#include "pipe/p_compiler.h" #include "pipe/p_util.h" -struct trace_stream; - - -void trace_dump_trace_begin(struct trace_stream *stream, unsigned version); -void trace_dump_trace_end(struct trace_stream *stream); -void trace_dump_call_begin(struct trace_stream *stream, const char *klass, const char *method); -void trace_dump_call_end(struct trace_stream *stream); -void trace_dump_arg_begin(struct trace_stream *stream, const char *name); -void trace_dump_arg_end(struct trace_stream *stream); -void trace_dump_ret_begin(struct trace_stream *stream); -void trace_dump_ret_end(struct trace_stream *stream); -void trace_dump_bool(struct trace_stream *stream, int value); -void trace_dump_int(struct trace_stream *stream, long int value); -void trace_dump_uint(struct trace_stream *stream, long unsigned value); -void trace_dump_float(struct trace_stream *stream, double value); -void trace_dump_bytes(struct trace_stream *stream, const void *data, long unsigned size); -void trace_dump_string(struct trace_stream *stream, const char *str); -void trace_dump_enum(struct trace_stream *stream, const char *value); -void trace_dump_array_begin(struct trace_stream *stream); -void trace_dump_array_end(struct trace_stream *stream); -void trace_dump_elem_begin(struct trace_stream *stream); -void trace_dump_elem_end(struct trace_stream *stream); -void trace_dump_struct_begin(struct trace_stream *stream, const char *name); -void trace_dump_struct_end(struct trace_stream *stream); -void trace_dump_member_begin(struct trace_stream *stream, const char *name); -void trace_dump_member_end(struct trace_stream *stream); -void trace_dump_null(struct trace_stream *stream); -void trace_dump_ptr(struct trace_stream *stream, const void *value); +boolean trace_dump_trace_begin(void); +void trace_dump_trace_end(void); +void trace_dump_call_begin(const char *klass, const char *method); +void trace_dump_call_end(void); +void trace_dump_arg_begin(const char *name); +void trace_dump_arg_end(void); +void trace_dump_ret_begin(void); +void trace_dump_ret_end(void); +void trace_dump_bool(int value); +void trace_dump_int(long int value); +void trace_dump_uint(long unsigned value); +void trace_dump_float(double value); +void trace_dump_bytes(const void *data, long unsigned size); +void trace_dump_string(const char *str); +void trace_dump_enum(const char *value); +void trace_dump_array_begin(void); +void trace_dump_array_end(void); +void trace_dump_elem_begin(void); +void trace_dump_elem_end(void); +void trace_dump_struct_begin(const char *name); +void trace_dump_struct_end(void); +void trace_dump_member_begin(const char *name); +void trace_dump_member_end(void); +void trace_dump_null(void); +void trace_dump_ptr(const void *value); /* * Code saving macros. */ -#define trace_dump_arg(_stream, _type, _arg) \ +#define trace_dump_arg(_type, _arg) \ do { \ - trace_dump_arg_begin(_stream, #_arg); \ - trace_dump_##_type(_stream, _arg); \ - trace_dump_arg_end(_stream); \ + trace_dump_arg_begin(#_arg); \ + trace_dump_##_type(_arg); \ + trace_dump_arg_end(); \ } while(0) -#define trace_dump_ret(_stream, _type, _arg) \ +#define trace_dump_ret(_type, _arg) \ do { \ - trace_dump_ret_begin(_stream); \ - trace_dump_##_type(_stream, _arg); \ - trace_dump_ret_end(_stream); \ + trace_dump_ret_begin(); \ + trace_dump_##_type(_arg); \ + trace_dump_ret_end(); \ } while(0) -#define trace_dump_array(_stream, _type, _obj, _size) \ +#define trace_dump_array(_type, _obj, _size) \ do { \ unsigned long idx; \ - trace_dump_array_begin(_stream); \ + trace_dump_array_begin(); \ for(idx = 0; idx < (_size); ++idx) { \ - trace_dump_elem_begin(_stream); \ - trace_dump_##_type(_stream, (_obj)[idx]); \ - trace_dump_elem_end(_stream); \ + trace_dump_elem_begin(); \ + trace_dump_##_type((_obj)[idx]); \ + trace_dump_elem_end(); \ } \ - trace_dump_array_end(_stream); \ + trace_dump_array_end(); \ } while(0) -#define trace_dump_struct_array(_stream, _type, _obj, _size) \ +#define trace_dump_struct_array(_type, _obj, _size) \ do { \ unsigned long idx; \ - trace_dump_array_begin(_stream); \ + trace_dump_array_begin(); \ for(idx = 0; idx < (_size); ++idx) { \ - trace_dump_elem_begin(_stream); \ - trace_dump_##_type(_stream, &(_obj)[idx]); \ - trace_dump_elem_end(_stream); \ + trace_dump_elem_begin(); \ + trace_dump_##_type(&(_obj)[idx]); \ + trace_dump_elem_end(); \ } \ - trace_dump_array_end(_stream); \ + trace_dump_array_end(); \ } while(0) -#define trace_dump_member(_stream, _type, _obj, _member) \ +#define trace_dump_member(_type, _obj, _member) \ do { \ - trace_dump_member_begin(_stream, #_member); \ - trace_dump_##_type(_stream, (_obj)->_member); \ - trace_dump_member_end(_stream); \ + trace_dump_member_begin(#_member); \ + trace_dump_##_type((_obj)->_member); \ + trace_dump_member_end(); \ } while(0) -#define trace_dump_arg_array(_stream, _type, _arg, _size) \ +#define trace_dump_arg_array(_type, _arg, _size) \ do { \ - trace_dump_arg_begin(_stream, #_arg); \ - trace_dump_array(_stream, _type, _arg, _size); \ - trace_dump_arg_end(_stream); \ + trace_dump_arg_begin(#_arg); \ + trace_dump_array(_type, _arg, _size); \ + trace_dump_arg_end(); \ } while(0) -#define trace_dump_member_array(_stream, _type, _obj, _member) \ +#define trace_dump_member_array(_type, _obj, _member) \ do { \ - trace_dump_member_begin(_stream, #_member); \ - trace_dump_array(_stream, _type, (_obj)->_member, sizeof((_obj)->_member)/sizeof((_obj)->_member[0])); \ - trace_dump_member_end(_stream); \ + trace_dump_member_begin(#_member); \ + trace_dump_array(_type, (_obj)->_member, sizeof((_obj)->_member)/sizeof((_obj)->_member[0])); \ + trace_dump_member_end(); \ } while(0) diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index 0e253123ae..68165a4553 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -27,7 +27,6 @@ #include "pipe/p_util.h" -#include "tr_stream.h" #include "tr_dump.h" #include "tr_state.h" #include "tr_winsys.h" @@ -38,19 +37,18 @@ static const char * trace_screen_get_name(struct pipe_screen *_screen) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; const char *result; - trace_dump_call_begin(stream, "pipe_screen", "get_name"); + trace_dump_call_begin("pipe_screen", "get_name"); - trace_dump_arg(stream, ptr, screen); + trace_dump_arg(ptr, screen); result = screen->get_name(screen); - trace_dump_ret(stream, string, result); + trace_dump_ret(string, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -60,19 +58,18 @@ static const char * trace_screen_get_vendor(struct pipe_screen *_screen) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; const char *result; - trace_dump_call_begin(stream, "pipe_screen", "get_vendor"); + trace_dump_call_begin("pipe_screen", "get_vendor"); - trace_dump_arg(stream, ptr, screen); + trace_dump_arg(ptr, screen); result = screen->get_vendor(screen); - trace_dump_ret(stream, string, result); + trace_dump_ret(string, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -83,20 +80,19 @@ trace_screen_get_param(struct pipe_screen *_screen, int param) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; int result; - trace_dump_call_begin(stream, "pipe_screen", "get_param"); + trace_dump_call_begin("pipe_screen", "get_param"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, int, param); + trace_dump_arg(ptr, screen); + trace_dump_arg(int, param); result = screen->get_param(screen, param); - trace_dump_ret(stream, int, result); + trace_dump_ret(int, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -107,20 +103,19 @@ trace_screen_get_paramf(struct pipe_screen *_screen, int param) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; float result; - trace_dump_call_begin(stream, "pipe_screen", "get_paramf"); + trace_dump_call_begin("pipe_screen", "get_paramf"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, int, param); + trace_dump_arg(ptr, screen); + trace_dump_arg(int, param); result = screen->get_paramf(screen, param); - trace_dump_ret(stream, float, result); + trace_dump_ret(float, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -134,23 +129,22 @@ trace_screen_is_format_supported(struct pipe_screen *_screen, unsigned geom_flags) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; boolean result; - trace_dump_call_begin(stream, "pipe_screen", "is_format_supported"); + trace_dump_call_begin("pipe_screen", "is_format_supported"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, format, format); - trace_dump_arg(stream, int, target); - trace_dump_arg(stream, uint, tex_usage); - trace_dump_arg(stream, uint, geom_flags); + trace_dump_arg(ptr, screen); + trace_dump_arg(format, format); + trace_dump_arg(int, target); + trace_dump_arg(uint, tex_usage); + trace_dump_arg(uint, geom_flags); result = screen->is_format_supported(screen, format, target, tex_usage, geom_flags); - trace_dump_ret(stream, bool, result); + trace_dump_ret(bool, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -161,20 +155,19 @@ trace_screen_texture_create(struct pipe_screen *_screen, const struct pipe_texture *templat) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; struct pipe_texture *result; - trace_dump_call_begin(stream, "pipe_screen", "texture_create"); + trace_dump_call_begin("pipe_screen", "texture_create"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, template, templat); + trace_dump_arg(ptr, screen); + trace_dump_arg(template, templat); result = screen->texture_create(screen, templat); - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -187,23 +180,22 @@ trace_screen_texture_blanket(struct pipe_screen *_screen, struct pipe_buffer *buffer) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; unsigned pitch = *ppitch; struct pipe_texture *result; - trace_dump_call_begin(stream, "pipe_screen", "texture_blanket"); + trace_dump_call_begin("pipe_screen", "texture_blanket"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, template, templat); - trace_dump_arg(stream, uint, pitch); - trace_dump_arg(stream, ptr, buffer); + trace_dump_arg(ptr, screen); + trace_dump_arg(template, templat); + trace_dump_arg(uint, pitch); + trace_dump_arg(ptr, buffer); result = screen->texture_blanket(screen, templat, ppitch, buffer); - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -214,18 +206,17 @@ trace_screen_texture_release(struct pipe_screen *_screen, struct pipe_texture **ptexture) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; struct pipe_texture *texture = *ptexture; - trace_dump_call_begin(stream, "pipe_screen", "texture_release"); + trace_dump_call_begin("pipe_screen", "texture_release"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, ptr, texture); + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, texture); screen->texture_release(screen, ptexture); - trace_dump_call_end(stream); + trace_dump_call_end(); } static struct pipe_surface * @@ -236,24 +227,23 @@ trace_screen_get_tex_surface(struct pipe_screen *_screen, unsigned usage) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; struct pipe_surface *result; - trace_dump_call_begin(stream, "pipe_screen", "get_tex_surface"); + trace_dump_call_begin("pipe_screen", "get_tex_surface"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, ptr, texture); - trace_dump_arg(stream, uint, face); - trace_dump_arg(stream, uint, level); - trace_dump_arg(stream, uint, zslice); - trace_dump_arg(stream, uint, usage); + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, texture); + trace_dump_arg(uint, face); + trace_dump_arg(uint, level); + trace_dump_arg(uint, zslice); + trace_dump_arg(uint, usage); result = screen->get_tex_surface(screen, texture, face, level, zslice, usage); - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -264,18 +254,17 @@ trace_screen_tex_surface_release(struct pipe_screen *_screen, struct pipe_surface **psurface) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; struct pipe_surface *surface = *psurface; - trace_dump_call_begin(stream, "pipe_screen", "tex_surface_release"); + trace_dump_call_begin("pipe_screen", "tex_surface_release"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, ptr, surface); + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, surface); screen->tex_surface_release(screen, psurface); - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -285,21 +274,20 @@ trace_screen_surface_map(struct pipe_screen *_screen, unsigned flags) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; struct pipe_surface *result; - trace_dump_call_begin(stream, "pipe_screen", "surface_map"); + trace_dump_call_begin("pipe_screen", "surface_map"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, ptr, surface); - trace_dump_arg(stream, uint, flags); + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, surface); + trace_dump_arg(uint, flags); result = screen->surface_map(screen, surface, flags); - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -310,17 +298,16 @@ trace_screen_surface_unmap(struct pipe_screen *_screen, struct pipe_surface *surface) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; - trace_dump_call_begin(stream, "pipe_screen", "surface_unmap"); + trace_dump_call_begin("pipe_screen", "surface_unmap"); - trace_dump_arg(stream, ptr, screen); - trace_dump_arg(stream, ptr, surface); + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, surface); screen->surface_unmap(screen, surface); - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -328,20 +315,17 @@ static void trace_screen_destroy(struct pipe_screen *_screen) { struct trace_screen *tr_scr = trace_screen(_screen); - struct trace_stream *stream = tr_scr->stream; struct pipe_screen *screen = tr_scr->screen; - trace_dump_call_begin(stream, "pipe_screen", "destroy"); + trace_dump_call_begin("pipe_screen", "destroy"); - trace_dump_arg(stream, ptr, screen); + trace_dump_arg(ptr, screen); screen->destroy(screen); - trace_dump_call_end(stream); + trace_dump_call_end(); - trace_dump_trace_end(stream); - - trace_stream_close(stream); + trace_dump_trace_end(); FREE(tr_scr); } @@ -350,26 +334,22 @@ trace_screen_destroy(struct pipe_screen *_screen) struct pipe_screen * trace_screen_create(struct pipe_screen *screen) { - struct trace_stream *stream; struct trace_screen *tr_scr; struct pipe_winsys *winsys; - if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) - return screen; - - tr_scr = CALLOC_STRUCT(trace_screen); - if(!tr_scr) - return NULL; + if(!screen) + goto error1; - tr_scr->stream = stream = trace_stream_create("gallium", "trace"); - if(!tr_scr->stream) - return NULL; + if(!trace_dump_trace_begin()) + goto error1; - trace_dump_trace_begin(stream, 0); + tr_scr = CALLOC_STRUCT(trace_screen); + if(!tr_scr) + goto error2; - winsys = trace_winsys_create(stream, screen->winsys); + winsys = trace_winsys_create(screen->winsys); if(!winsys) - return NULL; + goto error3; tr_scr->base.winsys = winsys; tr_scr->base.destroy = trace_screen_destroy; @@ -387,14 +367,20 @@ trace_screen_create(struct pipe_screen *screen) tr_scr->base.surface_unmap = trace_screen_surface_unmap; tr_scr->screen = screen; - tr_scr->stream = stream = trace_winsys(winsys)->stream; - trace_dump_call_begin(stream, "", "pipe_screen_create"); - trace_dump_arg_begin(stream, "winsys"); - trace_dump_ptr(stream, screen->winsys); - trace_dump_arg_end(stream); - trace_dump_ret(stream, ptr, screen); - trace_dump_call_end(stream); + trace_dump_call_begin("", "pipe_screen_create"); + trace_dump_arg_begin("winsys"); + trace_dump_ptr(screen->winsys); + trace_dump_arg_end(); + trace_dump_ret(ptr, screen); + trace_dump_call_end(); return &tr_scr->base; + +error3: + FREE(tr_scr); +error2: + trace_dump_trace_end(); +error1: + return screen; } diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h index 40b844778f..446c4af6a6 100644 --- a/src/gallium/drivers/trace/tr_screen.h +++ b/src/gallium/drivers/trace/tr_screen.h @@ -34,16 +34,11 @@ #include "pipe/p_screen.h" -struct trace_stream; - - struct trace_screen { struct pipe_screen base; struct pipe_screen *screen; - - struct trace_stream *stream; }; diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index e074ae7abc..7b35e89e75 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -33,461 +33,442 @@ #include "tr_state.h" -void trace_dump_format(struct trace_stream *stream, - enum pipe_format format) +void trace_dump_format(enum pipe_format format) { - trace_dump_enum(stream, pf_name(format) ); + trace_dump_enum(pf_name(format) ); } -void trace_dump_block(struct trace_stream *stream, - const struct pipe_format_block *block) +void trace_dump_block(const struct pipe_format_block *block) { - trace_dump_struct_begin(stream, "pipe_format_block"); - trace_dump_member(stream, uint, block, size); - trace_dump_member(stream, uint, block, width); - trace_dump_member(stream, uint, block, height); - trace_dump_struct_end(stream); + trace_dump_struct_begin("pipe_format_block"); + trace_dump_member(uint, block, size); + trace_dump_member(uint, block, width); + trace_dump_member(uint, block, height); + trace_dump_struct_end(); } -void trace_dump_template(struct trace_stream *stream, - const struct pipe_texture *templat) +void trace_dump_template(const struct pipe_texture *templat) { assert(templat); if(!templat) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_texture"); + trace_dump_struct_begin("pipe_texture"); - trace_dump_member(stream, int, templat, target); - trace_dump_member(stream, format, templat, format); + trace_dump_member(int, templat, target); + trace_dump_member(format, templat, format); - trace_dump_member_begin(stream, "width"); - trace_dump_array(stream, uint, templat->width, 1); - trace_dump_member_end(stream); + trace_dump_member_begin("width"); + trace_dump_array(uint, templat->width, 1); + trace_dump_member_end(); - trace_dump_member_begin(stream, "height"); - trace_dump_array(stream, uint, templat->height, 1); - trace_dump_member_end(stream); + trace_dump_member_begin("height"); + trace_dump_array(uint, templat->height, 1); + trace_dump_member_end(); - trace_dump_member_begin(stream, "depth"); - trace_dump_array(stream, uint, templat->depth, 1); - trace_dump_member_end(stream); + trace_dump_member_begin("depth"); + trace_dump_array(uint, templat->depth, 1); + trace_dump_member_end(); - trace_dump_member_begin(stream, "block"); - trace_dump_block(stream, &templat->block); - trace_dump_member_end(stream); + trace_dump_member_begin("block"); + trace_dump_block(&templat->block); + trace_dump_member_end(); - trace_dump_member(stream, uint, templat, last_level); - trace_dump_member(stream, uint, templat, tex_usage); + trace_dump_member(uint, templat, last_level); + trace_dump_member(uint, templat, tex_usage); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_rasterizer_state(struct trace_stream *stream, - const struct pipe_rasterizer_state *state) +void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_rasterizer_state"); - - trace_dump_member(stream, bool, state, flatshade); - trace_dump_member(stream, bool, state, light_twoside); - trace_dump_member(stream, uint, state, front_winding); - trace_dump_member(stream, uint, state, cull_mode); - trace_dump_member(stream, uint, state, fill_cw); - trace_dump_member(stream, uint, state, fill_ccw); - trace_dump_member(stream, bool, state, offset_cw); - trace_dump_member(stream, bool, state, offset_ccw); - trace_dump_member(stream, bool, state, scissor); - trace_dump_member(stream, bool, state, poly_smooth); - trace_dump_member(stream, bool, state, poly_stipple_enable); - trace_dump_member(stream, bool, state, point_smooth); - trace_dump_member(stream, bool, state, point_sprite); - trace_dump_member(stream, bool, state, point_size_per_vertex); - trace_dump_member(stream, bool, state, multisample); - trace_dump_member(stream, bool, state, line_smooth); - trace_dump_member(stream, bool, state, line_stipple_enable); - trace_dump_member(stream, uint, state, line_stipple_factor); - trace_dump_member(stream, uint, state, line_stipple_pattern); - trace_dump_member(stream, bool, state, line_last_pixel); - trace_dump_member(stream, bool, state, bypass_clipping); - trace_dump_member(stream, bool, state, bypass_vs); - trace_dump_member(stream, bool, state, origin_lower_left); - trace_dump_member(stream, bool, state, flatshade_first); - trace_dump_member(stream, bool, state, gl_rasterization_rules); - - trace_dump_member(stream, float, state, line_width); - trace_dump_member(stream, float, state, point_size); - trace_dump_member(stream, float, state, point_size_min); - trace_dump_member(stream, float, state, point_size_max); - trace_dump_member(stream, float, state, offset_units); - trace_dump_member(stream, float, state, offset_scale); + trace_dump_struct_begin("pipe_rasterizer_state"); + + trace_dump_member(bool, state, flatshade); + trace_dump_member(bool, state, light_twoside); + trace_dump_member(uint, state, front_winding); + trace_dump_member(uint, state, cull_mode); + trace_dump_member(uint, state, fill_cw); + trace_dump_member(uint, state, fill_ccw); + trace_dump_member(bool, state, offset_cw); + trace_dump_member(bool, state, offset_ccw); + trace_dump_member(bool, state, scissor); + trace_dump_member(bool, state, poly_smooth); + trace_dump_member(bool, state, poly_stipple_enable); + trace_dump_member(bool, state, point_smooth); + trace_dump_member(bool, state, point_sprite); + trace_dump_member(bool, state, point_size_per_vertex); + trace_dump_member(bool, state, multisample); + trace_dump_member(bool, state, line_smooth); + trace_dump_member(bool, state, line_stipple_enable); + trace_dump_member(uint, state, line_stipple_factor); + trace_dump_member(uint, state, line_stipple_pattern); + trace_dump_member(bool, state, line_last_pixel); + trace_dump_member(bool, state, bypass_clipping); + trace_dump_member(bool, state, bypass_vs); + trace_dump_member(bool, state, origin_lower_left); + trace_dump_member(bool, state, flatshade_first); + trace_dump_member(bool, state, gl_rasterization_rules); + + trace_dump_member(float, state, line_width); + trace_dump_member(float, state, point_size); + trace_dump_member(float, state, point_size_min); + trace_dump_member(float, state, point_size_max); + trace_dump_member(float, state, offset_units); + trace_dump_member(float, state, offset_scale); - trace_dump_member_array(stream, uint, state, sprite_coord_mode); + trace_dump_member_array(uint, state, sprite_coord_mode); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_poly_stipple(struct trace_stream *stream, - const struct pipe_poly_stipple *state) +void trace_dump_poly_stipple(const struct pipe_poly_stipple *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_poly_stipple"); + trace_dump_struct_begin("pipe_poly_stipple"); - trace_dump_member_begin(stream, "stipple"); - trace_dump_array(stream, - uint, + trace_dump_member_begin("stipple"); + trace_dump_array(uint, state->stipple, Elements(state->stipple)); - trace_dump_member_end(stream); + trace_dump_member_end(); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_viewport_state(struct trace_stream *stream, - const struct pipe_viewport_state *state) +void trace_dump_viewport_state(const struct pipe_viewport_state *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_viewport_state"); + trace_dump_struct_begin("pipe_viewport_state"); - trace_dump_member_array(stream, float, state, scale); - trace_dump_member_array(stream, float, state, translate); + trace_dump_member_array(float, state, scale); + trace_dump_member_array(float, state, translate); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_scissor_state(struct trace_stream *stream, - const struct pipe_scissor_state *state) +void trace_dump_scissor_state(const struct pipe_scissor_state *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_scissor_state"); + trace_dump_struct_begin("pipe_scissor_state"); - trace_dump_member(stream, uint, state, minx); - trace_dump_member(stream, uint, state, miny); - trace_dump_member(stream, uint, state, maxx); - trace_dump_member(stream, uint, state, maxy); + trace_dump_member(uint, state, minx); + trace_dump_member(uint, state, miny); + trace_dump_member(uint, state, maxx); + trace_dump_member(uint, state, maxy); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_clip_state(struct trace_stream *stream, - const struct pipe_clip_state *state) +void trace_dump_clip_state(const struct pipe_clip_state *state) { unsigned i; assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_scissor_state"); + trace_dump_struct_begin("pipe_scissor_state"); - trace_dump_member_begin(stream, "ucp"); - trace_dump_array_begin(stream); + trace_dump_member_begin("ucp"); + trace_dump_array_begin(); for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) - trace_dump_array(stream, float, state->ucp[i], 4); - trace_dump_array_end(stream); - trace_dump_member_end(stream); + trace_dump_array(float, state->ucp[i], 4); + trace_dump_array_end(); + trace_dump_member_end(); - trace_dump_member(stream, uint, state, nr); + trace_dump_member(uint, state, nr); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_constant_buffer(struct trace_stream *stream, - const struct pipe_constant_buffer *state) +void trace_dump_constant_buffer(const struct pipe_constant_buffer *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_constant_buffer"); + trace_dump_struct_begin("pipe_constant_buffer"); - trace_dump_member(stream, ptr, state, buffer); - trace_dump_member(stream, uint, state, size); + trace_dump_member(ptr, state, buffer); + trace_dump_member(uint, state, size); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_shader_state(struct trace_stream *stream, - const struct pipe_shader_state *state) +void trace_dump_shader_state(const struct pipe_shader_state *state) { static char str[8192]; assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } tgsi_dump_str(state->tokens, 0, str, sizeof(str)); - trace_dump_struct_begin(stream, "pipe_shader_state"); + trace_dump_struct_begin("pipe_shader_state"); - trace_dump_member_begin(stream, "tokens"); - trace_dump_string(stream, str); - trace_dump_member_end(stream); + trace_dump_member_begin("tokens"); + trace_dump_string(str); + trace_dump_member_end(); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_depth_stencil_alpha_state(struct trace_stream *stream, - const struct pipe_depth_stencil_alpha_state *state) +void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state) { unsigned i; assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_depth_stencil_alpha_state"); + trace_dump_struct_begin("pipe_depth_stencil_alpha_state"); - trace_dump_member_begin(stream, "depth"); - trace_dump_struct_begin(stream, "pipe_depth_state"); - trace_dump_member(stream, bool, &state->depth, enabled); - trace_dump_member(stream, bool, &state->depth, writemask); - trace_dump_member(stream, uint, &state->depth, func); - trace_dump_member(stream, bool, &state->depth, occlusion_count); - trace_dump_struct_end(stream); - trace_dump_member_end(stream); + trace_dump_member_begin("depth"); + trace_dump_struct_begin("pipe_depth_state"); + trace_dump_member(bool, &state->depth, enabled); + trace_dump_member(bool, &state->depth, writemask); + trace_dump_member(uint, &state->depth, func); + trace_dump_member(bool, &state->depth, occlusion_count); + trace_dump_struct_end(); + trace_dump_member_end(); - trace_dump_member_begin(stream, "stencil"); - trace_dump_array_begin(stream); + trace_dump_member_begin("stencil"); + trace_dump_array_begin(); for(i = 0; i < Elements(state->stencil); ++i) { - trace_dump_elem_begin(stream); - trace_dump_struct_begin(stream, "pipe_stencil_state"); - trace_dump_member(stream, bool, &state->stencil[i], enabled); - trace_dump_member(stream, uint, &state->stencil[i], func); - trace_dump_member(stream, uint, &state->stencil[i], fail_op); - trace_dump_member(stream, uint, &state->stencil[i], zpass_op); - trace_dump_member(stream, uint, &state->stencil[i], zfail_op); - trace_dump_member(stream, uint, &state->stencil[i], ref_value); - trace_dump_member(stream, uint, &state->stencil[i], value_mask); - trace_dump_member(stream, uint, &state->stencil[i], write_mask); - trace_dump_struct_end(stream); - trace_dump_elem_end(stream); + trace_dump_elem_begin(); + trace_dump_struct_begin("pipe_stencil_state"); + trace_dump_member(bool, &state->stencil[i], enabled); + trace_dump_member(uint, &state->stencil[i], func); + trace_dump_member(uint, &state->stencil[i], fail_op); + trace_dump_member(uint, &state->stencil[i], zpass_op); + trace_dump_member(uint, &state->stencil[i], zfail_op); + trace_dump_member(uint, &state->stencil[i], ref_value); + trace_dump_member(uint, &state->stencil[i], value_mask); + trace_dump_member(uint, &state->stencil[i], write_mask); + trace_dump_struct_end(); + trace_dump_elem_end(); } - trace_dump_array_end(stream); - trace_dump_member_end(stream); - - trace_dump_member_begin(stream, "alpha"); - trace_dump_struct_begin(stream, "pipe_alpha_state"); - trace_dump_member(stream, bool, &state->alpha, enabled); - trace_dump_member(stream, uint, &state->alpha, func); - trace_dump_member(stream, float, &state->alpha, ref); - trace_dump_struct_end(stream); - trace_dump_member_end(stream); - - trace_dump_struct_end(stream); + trace_dump_array_end(); + trace_dump_member_end(); + + trace_dump_member_begin("alpha"); + trace_dump_struct_begin("pipe_alpha_state"); + trace_dump_member(bool, &state->alpha, enabled); + trace_dump_member(uint, &state->alpha, func); + trace_dump_member(float, &state->alpha, ref); + trace_dump_struct_end(); + trace_dump_member_end(); + + trace_dump_struct_end(); } -void trace_dump_blend_state(struct trace_stream *stream, - const struct pipe_blend_state *state) +void trace_dump_blend_state(const struct pipe_blend_state *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_blend_state"); + trace_dump_struct_begin("pipe_blend_state"); - trace_dump_member(stream, bool, state, blend_enable); + trace_dump_member(bool, state, blend_enable); - trace_dump_member(stream, uint, state, rgb_func); - trace_dump_member(stream, uint, state, rgb_src_factor); - trace_dump_member(stream, uint, state, rgb_dst_factor); + trace_dump_member(uint, state, rgb_func); + trace_dump_member(uint, state, rgb_src_factor); + trace_dump_member(uint, state, rgb_dst_factor); - trace_dump_member(stream, uint, state, alpha_func); - trace_dump_member(stream, uint, state, alpha_src_factor); - trace_dump_member(stream, uint, state, alpha_dst_factor); + trace_dump_member(uint, state, alpha_func); + trace_dump_member(uint, state, alpha_src_factor); + trace_dump_member(uint, state, alpha_dst_factor); - trace_dump_member(stream, bool, state, logicop_enable); - trace_dump_member(stream, uint, state, logicop_func); + trace_dump_member(bool, state, logicop_enable); + trace_dump_member(uint, state, logicop_func); - trace_dump_member(stream, uint, state, colormask); - trace_dump_member(stream, bool, state, dither); + trace_dump_member(uint, state, colormask); + trace_dump_member(bool, state, dither); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_blend_color(struct trace_stream *stream, - const struct pipe_blend_color *state) +void trace_dump_blend_color(const struct pipe_blend_color *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_blend_color"); + trace_dump_struct_begin("pipe_blend_color"); - trace_dump_member_array(stream, float, state, color); + trace_dump_member_array(float, state, color); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_framebuffer_state(struct trace_stream *stream, - const struct pipe_framebuffer_state *state) +void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state) { - trace_dump_struct_begin(stream, "pipe_framebuffer_state"); + trace_dump_struct_begin("pipe_framebuffer_state"); - trace_dump_member(stream, uint, state, width); - trace_dump_member(stream, uint, state, height); - trace_dump_member(stream, uint, state, num_cbufs); - trace_dump_member_array(stream, ptr, state, cbufs); - trace_dump_member(stream, ptr, state, zsbuf); + trace_dump_member(uint, state, width); + trace_dump_member(uint, state, height); + trace_dump_member(uint, state, num_cbufs); + trace_dump_member_array(ptr, state, cbufs); + trace_dump_member(ptr, state, zsbuf); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_sampler_state(struct trace_stream *stream, - const struct pipe_sampler_state *state) +void trace_dump_sampler_state(const struct pipe_sampler_state *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_sampler_state"); - - trace_dump_member(stream, uint, state, wrap_s); - trace_dump_member(stream, uint, state, wrap_t); - trace_dump_member(stream, uint, state, wrap_r); - trace_dump_member(stream, uint, state, min_img_filter); - trace_dump_member(stream, uint, state, min_mip_filter); - trace_dump_member(stream, uint, state, mag_img_filter); - trace_dump_member(stream, bool, state, compare_mode); - trace_dump_member(stream, uint, state, compare_func); - trace_dump_member(stream, bool, state, normalized_coords); - trace_dump_member(stream, uint, state, prefilter); - trace_dump_member(stream, float, state, shadow_ambient); - trace_dump_member(stream, float, state, lod_bias); - trace_dump_member(stream, float, state, min_lod); - trace_dump_member(stream, float, state, max_lod); - trace_dump_member_array(stream, float, state, border_color); - trace_dump_member(stream, float, state, max_anisotropy); - - trace_dump_struct_end(stream); + trace_dump_struct_begin("pipe_sampler_state"); + + trace_dump_member(uint, state, wrap_s); + trace_dump_member(uint, state, wrap_t); + trace_dump_member(uint, state, wrap_r); + trace_dump_member(uint, state, min_img_filter); + trace_dump_member(uint, state, min_mip_filter); + trace_dump_member(uint, state, mag_img_filter); + trace_dump_member(bool, state, compare_mode); + trace_dump_member(uint, state, compare_func); + trace_dump_member(bool, state, normalized_coords); + trace_dump_member(uint, state, prefilter); + trace_dump_member(float, state, shadow_ambient); + trace_dump_member(float, state, lod_bias); + trace_dump_member(float, state, min_lod); + trace_dump_member(float, state, max_lod); + trace_dump_member_array(float, state, border_color); + trace_dump_member(float, state, max_anisotropy); + + trace_dump_struct_end(); } -void trace_dump_surface(struct trace_stream *stream, - const struct pipe_surface *state) +void trace_dump_surface(const struct pipe_surface *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_surface"); + trace_dump_struct_begin("pipe_surface"); - trace_dump_member(stream, ptr, state, buffer); - trace_dump_member(stream, format, state, format); - trace_dump_member(stream, uint, state, status); - trace_dump_member(stream, uint, state, clear_value); - trace_dump_member(stream, uint, state, width); - trace_dump_member(stream, uint, state, height); + trace_dump_member(ptr, state, buffer); + trace_dump_member(format, state, format); + trace_dump_member(uint, state, status); + trace_dump_member(uint, state, clear_value); + trace_dump_member(uint, state, width); + trace_dump_member(uint, state, height); - trace_dump_member_begin(stream, "block"); - trace_dump_block(stream, &state->block); - trace_dump_member_end(stream); + trace_dump_member_begin("block"); + trace_dump_block(&state->block); + trace_dump_member_end(); - trace_dump_member(stream, uint, state, nblocksx); - trace_dump_member(stream, uint, state, nblocksy); - trace_dump_member(stream, uint, state, stride); - trace_dump_member(stream, uint, state, layout); - trace_dump_member(stream, uint, state, offset); - trace_dump_member(stream, uint, state, refcount); - trace_dump_member(stream, uint, state, usage); - - trace_dump_member(stream, ptr, state, texture); - trace_dump_member(stream, uint, state, face); - trace_dump_member(stream, uint, state, level); - trace_dump_member(stream, uint, state, zslice); - - trace_dump_struct_end(stream); + trace_dump_member(uint, state, nblocksx); + trace_dump_member(uint, state, nblocksy); + trace_dump_member(uint, state, stride); + trace_dump_member(uint, state, layout); + trace_dump_member(uint, state, offset); + trace_dump_member(uint, state, refcount); + trace_dump_member(uint, state, usage); + + trace_dump_member(ptr, state, texture); + trace_dump_member(uint, state, face); + trace_dump_member(uint, state, level); + trace_dump_member(uint, state, zslice); + + trace_dump_struct_end(); } -void trace_dump_vertex_buffer(struct trace_stream *stream, - const struct pipe_vertex_buffer *state) +void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_vertex_buffer"); + trace_dump_struct_begin("pipe_vertex_buffer"); - trace_dump_member(stream, uint, state, pitch); - trace_dump_member(stream, uint, state, max_index); - trace_dump_member(stream, uint, state, buffer_offset); - trace_dump_member(stream, ptr, state, buffer); + trace_dump_member(uint, state, pitch); + trace_dump_member(uint, state, max_index); + trace_dump_member(uint, state, buffer_offset); + trace_dump_member(ptr, state, buffer); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } -void trace_dump_vertex_element(struct trace_stream *stream, - const struct pipe_vertex_element *state) +void trace_dump_vertex_element(const struct pipe_vertex_element *state) { assert(state); if(!state) { - trace_dump_null(stream); + trace_dump_null(); return; } - trace_dump_struct_begin(stream, "pipe_vertex_element"); + trace_dump_struct_begin("pipe_vertex_element"); - trace_dump_member(stream, uint, state, src_offset); + trace_dump_member(uint, state, src_offset); - trace_dump_member(stream, uint, state, vertex_buffer_index); - trace_dump_member(stream, uint, state, nr_components); + trace_dump_member(uint, state, vertex_buffer_index); + trace_dump_member(uint, state, nr_components); - trace_dump_member(stream, format, state, src_format); + trace_dump_member(format, state, src_format); - trace_dump_struct_end(stream); + trace_dump_struct_end(); } diff --git a/src/gallium/drivers/trace/tr_state.h b/src/gallium/drivers/trace/tr_state.h index c1df63db6a..5ae533dc66 100644 --- a/src/gallium/drivers/trace/tr_state.h +++ b/src/gallium/drivers/trace/tr_state.h @@ -4,7 +4,7 @@ * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation streams (the + * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sub license, and/or sell copies of the Software, and to @@ -33,63 +33,44 @@ #include "pipe/p_shader_tokens.h" -void trace_dump_format(struct trace_stream *stream, - enum pipe_format format); +void trace_dump_format(enum pipe_format format); -void trace_dump_block(struct trace_stream *stream, - const struct pipe_format_block *block); +void trace_dump_block(const struct pipe_format_block *block); -void trace_dump_template(struct trace_stream *stream, - const struct pipe_texture *templat); +void trace_dump_template(const struct pipe_texture *templat); -void trace_dump_rasterizer_state(struct trace_stream *stream, - const struct pipe_rasterizer_state *state); +void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state); -void trace_dump_poly_stipple(struct trace_stream *stream, - const struct pipe_poly_stipple *state); +void trace_dump_poly_stipple(const struct pipe_poly_stipple *state); -void trace_dump_viewport_state(struct trace_stream *stream, - const struct pipe_viewport_state *state); +void trace_dump_viewport_state(const struct pipe_viewport_state *state); -void trace_dump_scissor_state(struct trace_stream *stream, - const struct pipe_scissor_state *state); +void trace_dump_scissor_state(const struct pipe_scissor_state *state); -void trace_dump_clip_state(struct trace_stream *stream, - const struct pipe_clip_state *state); +void trace_dump_clip_state(const struct pipe_clip_state *state); -void trace_dump_constant_buffer(struct trace_stream *stream, - const struct pipe_constant_buffer *state); +void trace_dump_constant_buffer(const struct pipe_constant_buffer *state); -void trace_dump_token(struct trace_stream *stream, - const struct tgsi_token *token); +void trace_dump_token(const struct tgsi_token *token); -void trace_dump_shader_state(struct trace_stream *stream, - const struct pipe_shader_state *state); +void trace_dump_shader_state(const struct pipe_shader_state *state); -void trace_dump_depth_stencil_alpha_state(struct trace_stream *stream, - const struct pipe_depth_stencil_alpha_state *state); +void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_state *state); -void trace_dump_blend_state(struct trace_stream *stream, - const struct pipe_blend_state *state); +void trace_dump_blend_state(const struct pipe_blend_state *state); -void trace_dump_blend_color(struct trace_stream *stream, - const struct pipe_blend_color *state); +void trace_dump_blend_color(const struct pipe_blend_color *state); -void trace_dump_framebuffer_state(struct trace_stream *stream, - const struct pipe_framebuffer_state *state); +void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state); -void trace_dump_sampler_state(struct trace_stream *stream, - const struct pipe_sampler_state *state); +void trace_dump_sampler_state(const struct pipe_sampler_state *state); -void trace_dump_surface(struct trace_stream *stream, - const struct pipe_surface *state); +void trace_dump_surface(const struct pipe_surface *state); -void trace_dump_vertex_buffer(struct trace_stream *stream, - const struct pipe_vertex_buffer *state); +void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state); -void trace_dump_vertex_element(struct trace_stream *stream, - const struct pipe_vertex_element *state); +void trace_dump_vertex_element(const struct pipe_vertex_element *state); #endif /* TR_STATE_H */ diff --git a/src/gallium/drivers/trace/tr_stream.h b/src/gallium/drivers/trace/tr_stream.h index d50fed2691..53e854aa91 100644 --- a/src/gallium/drivers/trace/tr_stream.h +++ b/src/gallium/drivers/trace/tr_stream.h @@ -4,7 +4,7 @@ * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation streams (the + * copy of this software and associated documentation files (the * "Software"), to deal in the Software without restriction, including * without limitation the rights to use, copy, modify, merge, publish, * distribute, sub license, and/or sell copies of the Software, and to @@ -28,6 +28,9 @@ /** * @file * Cross-platform sequential access stream abstraction. + * + * These are really general purpose file access functions, and might one day + * be moved into the util module. */ #ifndef TR_STREAM_H diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index 524049148d..be76c0716f 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -29,7 +29,6 @@ #include "pipe/p_state.h" #include "util/u_hash_table.h" -#include "tr_stream.h" #include "tr_dump.h" #include "tr_state.h" #include "tr_winsys.h" @@ -51,19 +50,18 @@ static const char * trace_winsys_get_name(struct pipe_winsys *_winsys) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; const char *result; - trace_dump_call_begin(stream, "pipe_winsys", "get_name"); + trace_dump_call_begin("pipe_winsys", "get_name"); - trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(ptr, winsys); result = winsys->get_name(winsys); - trace_dump_ret(stream, string, result); + trace_dump_ret(string, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -75,18 +73,17 @@ trace_winsys_flush_frontbuffer(struct pipe_winsys *_winsys, void *context_private) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; - trace_dump_call_begin(stream, "pipe_winsys", "flush_frontbuffer"); + trace_dump_call_begin("pipe_winsys", "flush_frontbuffer"); - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, surface); - trace_dump_arg(stream, ptr, context_private); + trace_dump_arg(ptr, winsys); + trace_dump_arg(ptr, surface); + trace_dump_arg(ptr, context_private); winsys->flush_frontbuffer(winsys, surface, context_private); - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -94,19 +91,18 @@ static struct pipe_surface * trace_winsys_surface_alloc(struct pipe_winsys *_winsys) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; struct pipe_surface *result; - trace_dump_call_begin(stream, "pipe_winsys", "surface_alloc"); + trace_dump_call_begin("pipe_winsys", "surface_alloc"); - trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(ptr, winsys); result = winsys->surface_alloc(winsys); - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -121,19 +117,18 @@ trace_winsys_surface_alloc_storage(struct pipe_winsys *_winsys, unsigned tex_usage) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; int result; - trace_dump_call_begin(stream, "pipe_winsys", "surface_alloc_storage"); + trace_dump_call_begin("pipe_winsys", "surface_alloc_storage"); - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, surface); - trace_dump_arg(stream, uint, width); - trace_dump_arg(stream, uint, height); - trace_dump_arg(stream, format, format); - trace_dump_arg(stream, uint, flags); - trace_dump_arg(stream, uint, tex_usage); + trace_dump_arg(ptr, winsys); + trace_dump_arg(ptr, surface); + trace_dump_arg(uint, width); + trace_dump_arg(uint, height); + trace_dump_arg(format, format); + trace_dump_arg(uint, flags); + trace_dump_arg(uint, tex_usage); result = winsys->surface_alloc_storage(winsys, surface, @@ -142,9 +137,9 @@ trace_winsys_surface_alloc_storage(struct pipe_winsys *_winsys, flags, tex_usage); - trace_dump_ret(stream, int, result); + trace_dump_ret(int, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -155,18 +150,17 @@ trace_winsys_surface_release(struct pipe_winsys *_winsys, struct pipe_surface **psurface) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; struct pipe_surface *surface = *psurface; - trace_dump_call_begin(stream, "pipe_winsys", "surface_release"); + trace_dump_call_begin("pipe_winsys", "surface_release"); - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, surface); + trace_dump_arg(ptr, winsys); + trace_dump_arg(ptr, surface); winsys->surface_release(winsys, psurface); - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -177,22 +171,21 @@ trace_winsys_buffer_create(struct pipe_winsys *_winsys, unsigned size) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; struct pipe_buffer *buffer; - trace_dump_call_begin(stream, "pipe_winsys", "buffer_create"); + trace_dump_call_begin("pipe_winsys", "buffer_create"); - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, uint, alignment); - trace_dump_arg(stream, uint, usage); - trace_dump_arg(stream, uint, size); + trace_dump_arg(ptr, winsys); + trace_dump_arg(uint, alignment); + trace_dump_arg(uint, usage); + trace_dump_arg(uint, size); buffer = winsys->buffer_create(winsys, alignment, usage, size); - trace_dump_ret(stream, ptr, buffer); + trace_dump_ret(ptr, buffer); - trace_dump_call_end(stream); + trace_dump_call_end(); /* Zero the buffer to avoid dumping uninitialized memory */ if(buffer->usage & PIPE_BUFFER_USAGE_CPU_WRITE) { @@ -214,21 +207,20 @@ trace_winsys_user_buffer_create(struct pipe_winsys *_winsys, unsigned bytes) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; struct pipe_buffer *result; - trace_dump_call_begin(stream, "pipe_winsys", "user_buffer_create"); + trace_dump_call_begin("pipe_winsys", "user_buffer_create"); - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, ptr); - trace_dump_arg(stream, uint, bytes); + trace_dump_arg(ptr, winsys); + trace_dump_arg(ptr, ptr); + trace_dump_arg(uint, bytes); result = winsys->user_buffer_create(winsys, ptr, bytes); - trace_dump_ret(stream, ptr, result); + trace_dump_ret(ptr, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -260,27 +252,26 @@ trace_winsys_buffer_unmap(struct pipe_winsys *_winsys, struct pipe_buffer *buffer) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; const void *map; map = hash_table_get(tr_ws->buffer_maps, buffer); if(map) { - trace_dump_call_begin(stream, "pipe_winsys", "buffer_write"); + trace_dump_call_begin("pipe_winsys", "buffer_write"); - trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(ptr, winsys); - trace_dump_arg(stream, ptr, buffer); + trace_dump_arg(ptr, buffer); - trace_dump_arg_begin(stream, "data"); - trace_dump_bytes(stream, map, buffer->size); - trace_dump_arg_end(stream); + trace_dump_arg_begin("data"); + trace_dump_bytes(map, buffer->size); + trace_dump_arg_end(); - trace_dump_arg_begin(stream, "size"); - trace_dump_uint(stream, buffer->size); - trace_dump_arg_end(stream); + trace_dump_arg_begin("size"); + trace_dump_uint(buffer->size); + trace_dump_arg_end(); - trace_dump_call_end(stream); + trace_dump_call_end(); hash_table_remove(tr_ws->buffer_maps, buffer); } @@ -294,17 +285,16 @@ trace_winsys_buffer_destroy(struct pipe_winsys *_winsys, struct pipe_buffer *buffer) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; - trace_dump_call_begin(stream, "pipe_winsys", "buffer_destroy"); + trace_dump_call_begin("pipe_winsys", "buffer_destroy"); - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, buffer); + trace_dump_arg(ptr, winsys); + trace_dump_arg(ptr, buffer); winsys->buffer_destroy(winsys, buffer); - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -314,19 +304,18 @@ trace_winsys_fence_reference(struct pipe_winsys *_winsys, struct pipe_fence_handle *src) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; struct pipe_fence_handle *dst = *pdst; - trace_dump_call_begin(stream, "pipe_winsys", "fence_reference"); + trace_dump_call_begin("pipe_winsys", "fence_reference"); - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, dst); - trace_dump_arg(stream, ptr, src); + trace_dump_arg(ptr, winsys); + trace_dump_arg(ptr, dst); + trace_dump_arg(ptr, src); winsys->fence_reference(winsys, pdst, src); - trace_dump_call_end(stream); + trace_dump_call_end(); } @@ -336,21 +325,20 @@ trace_winsys_fence_signalled(struct pipe_winsys *_winsys, unsigned flag) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; int result; - trace_dump_call_begin(stream, "pipe_winsys", "fence_signalled"); + trace_dump_call_begin("pipe_winsys", "fence_signalled"); - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, fence); - trace_dump_arg(stream, uint, flag); + trace_dump_arg(ptr, winsys); + trace_dump_arg(ptr, fence); + trace_dump_arg(uint, flag); result = winsys->fence_signalled(winsys, fence, flag); - trace_dump_ret(stream, int, result); + trace_dump_ret(int, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -362,21 +350,20 @@ trace_winsys_fence_finish(struct pipe_winsys *_winsys, unsigned flag) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; int result; - trace_dump_call_begin(stream, "pipe_winsys", "fence_finish"); + trace_dump_call_begin("pipe_winsys", "fence_finish"); - trace_dump_arg(stream, ptr, winsys); - trace_dump_arg(stream, ptr, fence); - trace_dump_arg(stream, uint, flag); + trace_dump_arg(ptr, winsys); + trace_dump_arg(ptr, fence); + trace_dump_arg(uint, flag); result = winsys->fence_finish(winsys, fence, flag); - trace_dump_ret(stream, int, result); + trace_dump_ret(int, result); - trace_dump_call_end(stream); + trace_dump_call_end(); return result; } @@ -386,18 +373,17 @@ static void trace_winsys_destroy(struct pipe_winsys *_winsys) { struct trace_winsys *tr_ws = trace_winsys(_winsys); - struct trace_stream *stream = tr_ws->stream; struct pipe_winsys *winsys = tr_ws->winsys; - trace_dump_call_begin(stream, "pipe_winsys", "destroy"); + trace_dump_call_begin("pipe_winsys", "destroy"); - trace_dump_arg(stream, ptr, winsys); + trace_dump_arg(ptr, winsys); /* winsys->destroy(winsys); */ - trace_dump_call_end(stream); + trace_dump_call_end(); hash_table_destroy(tr_ws->buffer_maps); @@ -406,15 +392,16 @@ trace_winsys_destroy(struct pipe_winsys *_winsys) struct pipe_winsys * -trace_winsys_create(struct trace_stream *stream, - struct pipe_winsys *winsys) +trace_winsys_create(struct pipe_winsys *winsys) { - struct trace_winsys *tr_ws; + if(!winsys) + goto error1; + tr_ws = CALLOC_STRUCT(trace_winsys); if(!tr_ws) - return NULL; + goto error1; tr_ws->base.destroy = trace_winsys_destroy; tr_ws->base.get_name = trace_winsys_get_name; @@ -432,16 +419,20 @@ trace_winsys_create(struct trace_stream *stream, tr_ws->base.fence_finish = trace_winsys_fence_finish; tr_ws->winsys = winsys; - tr_ws->stream = stream; tr_ws->buffer_maps = hash_table_create(trace_buffer_hash, trace_buffer_compare); if(!tr_ws->buffer_maps) - return NULL; + goto error2; - trace_dump_call_begin(stream, "", "pipe_winsys_create"); - trace_dump_ret(stream, ptr, winsys); - trace_dump_call_end(stream); + trace_dump_call_begin("", "pipe_winsys_create"); + trace_dump_ret(ptr, winsys); + trace_dump_call_end(); return &tr_ws->base; + +error2: + FREE(tr_ws); +error1: + return winsys; } diff --git a/src/gallium/drivers/trace/tr_winsys.h b/src/gallium/drivers/trace/tr_winsys.h index 704d2c57c8..2218117347 100644 --- a/src/gallium/drivers/trace/tr_winsys.h +++ b/src/gallium/drivers/trace/tr_winsys.h @@ -35,7 +35,6 @@ struct hash_table; -struct trace_stream; struct trace_winsys @@ -44,8 +43,6 @@ struct trace_winsys struct pipe_winsys *winsys; - struct trace_stream *stream; - struct hash_table *buffer_maps; }; @@ -60,8 +57,7 @@ trace_winsys(struct pipe_winsys *winsys) struct pipe_winsys * -trace_winsys_create(struct trace_stream *stream, - struct pipe_winsys *winsys); +trace_winsys_create(struct pipe_winsys *winsys); #endif /* TR_WINSYS_H_ */ -- cgit v1.2.3 From 424dea98d47e77f61efc79134b230f2046061ebe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 14:34:33 +0100 Subject: trace: Trace surface contents. --- src/gallium/drivers/trace/README | 4 -- src/gallium/drivers/trace/tr_screen.c | 75 +++++++++++++++++++++++++---------- src/gallium/drivers/trace/tr_screen.h | 5 +++ 3 files changed, 60 insertions(+), 24 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README index 5752448b93..5f0ae2c641 100644 --- a/src/gallium/drivers/trace/README +++ b/src/gallium/drivers/trace/README @@ -22,9 +22,5 @@ which should create a gallium.*.trace file, which is an XML file. You can view copying trace.xsl and trace.css to the same directory, and opening with a XSLT capable browser like Firefox or Internet Explorer. -This is still work in progress, namely: -- surface writes are not traced -- no way to know the start/end of a frame - -- Jose Fonseca diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index 68165a4553..2eecfdc187 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -26,6 +26,7 @@ **************************************************************************/ #include "pipe/p_util.h" +#include "util/u_hash_table.h" #include "tr_dump.h" #include "tr_state.h" @@ -33,6 +34,18 @@ #include "tr_screen.h" +static unsigned trace_surface_hash(void *surface) +{ + return (unsigned)(uintptr_t)surface; +} + + +static int trace_surface_compare(void *surface1, void *surface2) +{ + return (char *)surface2 - (char *)surface1; +} + + static const char * trace_screen_get_name(struct pipe_screen *_screen) { @@ -275,21 +288,17 @@ trace_screen_surface_map(struct pipe_screen *_screen, { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; - struct pipe_surface *result; - - trace_dump_call_begin("pipe_screen", "surface_map"); + void *map; - trace_dump_arg(ptr, screen); - trace_dump_arg(ptr, surface); - trace_dump_arg(uint, flags); - - result = screen->surface_map(screen, surface, flags); + map = screen->surface_map(screen, surface, flags); + if(map) { + if(flags & PIPE_BUFFER_USAGE_CPU_WRITE) { + assert(!hash_table_get(tr_scr->surface_maps, surface)); + hash_table_set(tr_scr->surface_maps, surface, map); + } + } - trace_dump_ret(ptr, result); - - trace_dump_call_end(); - - return result; + return map; } @@ -299,15 +308,36 @@ trace_screen_surface_unmap(struct pipe_screen *_screen, { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; - - trace_dump_call_begin("pipe_screen", "surface_unmap"); - - trace_dump_arg(ptr, screen); - trace_dump_arg(ptr, surface); + const void *map; + + map = hash_table_get(tr_scr->surface_maps, surface); + if(map) { + size_t size = surface->nblocksy * surface->stride; + + trace_dump_call_begin("pipe_winsys", "surface_write"); + + trace_dump_arg(ptr, screen); + + trace_dump_arg(ptr, surface); + + trace_dump_arg_begin("data"); + trace_dump_bytes(map, size); + trace_dump_arg_end(); + + trace_dump_arg_begin("stride"); + trace_dump_uint(surface->stride); + trace_dump_arg_end(); + + trace_dump_arg_begin("size"); + trace_dump_uint(size); + trace_dump_arg_end(); + + trace_dump_call_end(); + + hash_table_remove(tr_scr->surface_maps, surface); + } screen->surface_unmap(screen, surface); - - trace_dump_call_end(); } @@ -368,6 +398,11 @@ trace_screen_create(struct pipe_screen *screen) tr_scr->screen = screen; + tr_scr->surface_maps = hash_table_create(trace_surface_hash, + trace_surface_compare); + if(!tr_scr->surface_maps) + goto error3; + trace_dump_call_begin("", "pipe_screen_create"); trace_dump_arg_begin("winsys"); trace_dump_ptr(screen->winsys); diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h index 446c4af6a6..90103aa922 100644 --- a/src/gallium/drivers/trace/tr_screen.h +++ b/src/gallium/drivers/trace/tr_screen.h @@ -34,11 +34,16 @@ #include "pipe/p_screen.h" +struct hash_table; + + struct trace_screen { struct pipe_screen base; struct pipe_screen *screen; + + struct hash_table *surface_maps; }; -- cgit v1.2.3 From 37336b7d9e7f8bed8765e9d8e4f61f804e4d8107 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 14:35:25 +0100 Subject: python: Fix put/get_tile_raw bindings. --- src/gallium/state_trackers/python/p_texture.i | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/p_texture.i b/src/gallium/state_trackers/python/p_texture.i index 06bc105465..f23afea167 100644 --- a/src/gallium/state_trackers/python/p_texture.i +++ b/src/gallium/state_trackers/python/p_texture.i @@ -102,12 +102,12 @@ void unmap( void ); void - get_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, unsigned char *raw, unsigned stride) { + get_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, char *raw, unsigned stride) { pipe_get_tile_raw($self, x, y, w, h, raw, stride); } void - put_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, const unsigned char *raw, unsigned stride) { + put_tile_raw(unsigned x, unsigned y, unsigned w, unsigned h, const char *raw, unsigned stride) { pipe_put_tile_raw($self, x, y, w, h, raw, stride); } -- cgit v1.2.3 From 4fffc9d63f117cfc5d5b2400536e8757c786cc1e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 14:38:58 +0100 Subject: trace: Several fixes. --- src/gallium/drivers/trace/tr_context.c | 8 +++++--- src/gallium/drivers/trace/tr_winsys.c | 16 ++++++++++------ 2 files changed, 15 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index 868b4f010d..e81cab252c 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -318,11 +318,11 @@ trace_context_create_sampler_state(struct pipe_context *_pipe, trace_dump_call_begin("pipe_context", "create_sampler_state"); trace_dump_arg(ptr, pipe); - trace_dump_arg(ptr, state); + trace_dump_arg(sampler_state, state); result = pipe->create_sampler_state(pipe, state);; - trace_dump_ret(sampler_state, result); + trace_dump_ret(ptr, result); trace_dump_call_end(); @@ -888,10 +888,12 @@ trace_context_flush(struct pipe_context *_pipe, trace_dump_arg(ptr, pipe); trace_dump_arg(uint, flags); - trace_dump_arg(ptr, fence); pipe->flush(pipe, flags, fence);; + if(fence) + trace_dump_ret(ptr, *fence); + trace_dump_call_end(); } diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index be76c0716f..356579809a 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -74,12 +74,14 @@ trace_winsys_flush_frontbuffer(struct pipe_winsys *_winsys, { struct trace_winsys *tr_ws = trace_winsys(_winsys); struct pipe_winsys *winsys = tr_ws->winsys; - + trace_dump_call_begin("pipe_winsys", "flush_frontbuffer"); trace_dump_arg(ptr, winsys); trace_dump_arg(ptr, surface); + /* XXX: hide, as there is nothing we can do with this trace_dump_arg(ptr, context_private); + */ winsys->flush_frontbuffer(winsys, surface, context_private); @@ -203,8 +205,8 @@ trace_winsys_buffer_create(struct pipe_winsys *_winsys, static struct pipe_buffer * trace_winsys_user_buffer_create(struct pipe_winsys *_winsys, - void *ptr, - unsigned bytes) + void *data, + unsigned size) { struct trace_winsys *tr_ws = trace_winsys(_winsys); struct pipe_winsys *winsys = tr_ws->winsys; @@ -213,10 +215,12 @@ trace_winsys_user_buffer_create(struct pipe_winsys *_winsys, trace_dump_call_begin("pipe_winsys", "user_buffer_create"); trace_dump_arg(ptr, winsys); - trace_dump_arg(ptr, ptr); - trace_dump_arg(uint, bytes); + trace_dump_arg_begin("data"); + trace_dump_bytes(data, size); + trace_dump_arg_end(); + trace_dump_arg(uint, size); - result = winsys->user_buffer_create(winsys, ptr, bytes); + result = winsys->user_buffer_create(winsys, data, size); trace_dump_ret(ptr, result); -- cgit v1.2.3 From 14fe0d62ee8cf1dd48ec8d373ee3dcb3fe06fd77 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 14:40:27 +0100 Subject: python/retrace: Keep addresses as strings. To simplify looking up these in the trace. --- src/gallium/state_trackers/python/retrace/model.py | 2 +- src/gallium/state_trackers/python/retrace/parser.py | 2 -- 2 files changed, 1 insertion(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/model.py b/src/gallium/state_trackers/python/retrace/model.py index c17b5db9f8..a17a765914 100755 --- a/src/gallium/state_trackers/python/retrace/model.py +++ b/src/gallium/state_trackers/python/retrace/model.py @@ -89,7 +89,7 @@ class Pointer(Node): visitor.visit_pointer(self) def __str__(self): - return hex(self.address) + return self.address class Call: diff --git a/src/gallium/state_trackers/python/retrace/parser.py b/src/gallium/state_trackers/python/retrace/parser.py index 571d8f6eae..9ee47f6e19 100755 --- a/src/gallium/state_trackers/python/retrace/parser.py +++ b/src/gallium/state_trackers/python/retrace/parser.py @@ -316,8 +316,6 @@ class TraceParser(XmlParser): address = self.character_data() self.element_end('ptr') - address = int(address, 16) - return Pointer(address) -- cgit v1.2.3 From c9751522b0ee1908c79f3f9d37b508ac0680bd16 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 14 Aug 2008 22:03:35 +0100 Subject: pyhon/retrace: Retrace surface contents. Now capable of replaying trivial/quad-tex-2d --- .../state_trackers/python/retrace/interpreter.py | 35 ++++++++++++++++------ 1 file changed, 26 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index ae62bc04a8..00b6376623 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -105,7 +105,8 @@ member_array_factories = { "pipe_viewport_state": {"scale": gallium.FloatArray, "translate": gallium.FloatArray}, "pipe_clip_state": {"ucp": gallium.FloatArray}, "pipe_depth_stencil_alpha_state": {"stencil": gallium.StencilArray}, - "pipe_blend_color": {"color": gallium.FloatArray}, + "pipe_blend_color": {"color": gallium.FloatArray}, + "pipe_sampler_state": {"border_color": gallium.FloatArray}, } @@ -191,6 +192,15 @@ class Winsys(Object): def get_name(self): pass + def user_buffer_create(self, data, size): + # We don't really care to distinguish between user and regular buffers + buffer = self.real.buffer_create(size, + 4, + gallium.PIPE_BUFFER_USAGE_CPU_READ | + gallium.PIPE_BUFFER_USAGE_CPU_WRITE ) + buffer.write(data, size) + return buffer + def buffer_create(self, alignment, usage, size): return self.real.buffer_create(size, alignment, usage) @@ -199,7 +209,15 @@ class Winsys(Object): def buffer_write(self, buffer, data, size): buffer.write(data, size) - buffer.write(data, size) + + def fence_finish(self, fence, flags): + pass + + def fence_reference(self, dst, src): + pass + + def flush_frontbuffer(self, surface): + pass class Screen(Object): @@ -239,11 +257,9 @@ class Screen(Object): def tex_surface_release(self, surface): self.interpreter.unregister_object(surface) - def surface_map(self, surface, flags): - return None - - def surface_unmap(self, surface): - pass + def surface_write(self, surface, data, stride, size): + assert surface.nblocksy * stride == size + surface.put_tile_raw(0, 0, surface.width, surface.height, data, stride) class Context(Object): @@ -324,7 +340,7 @@ class Context(Object): def set_sampler_textures(self, n, textures): for i in range(n): - self.real.set_sampler_textures(textures[i]) + self.real.set_sampler_texture(i, textures[i]) def set_vertex_buffers(self, n, vbufs): for i in range(n): @@ -350,8 +366,9 @@ class Context(Object): self.real.draw_arrays(mode, start, count) self._update() - def flush(self, flags, fence): + def flush(self, flags): self.real.flush(flags) + return None def clear(self, surface, value): self.real.surface_clear(surface, value) -- cgit v1.2.3 From f121d0e54f39d8f6361dcf0bf4d938ccb5ae4b5e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Aug 2008 10:24:09 +0100 Subject: trace: Allow multiple screens. Flush after call. --- src/gallium/drivers/trace/tr_dump.c | 53 ++++++++++++++++++++++++----------- src/gallium/drivers/trace/tr_dump.h | 1 + src/gallium/drivers/trace/tr_stream.c | 12 ++++++++ src/gallium/drivers/trace/tr_stream.h | 3 ++ 4 files changed, 53 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index 6ebb639b7c..c711a56b94 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -52,6 +52,7 @@ static struct trace_stream *stream = NULL; +static unsigned refcount = 0; static INLINE void @@ -204,35 +205,54 @@ trace_dump_tag_end(const char *name) trace_dump_writes(">"); } +static void +trace_dump_trace_close(void) +{ + if(stream) { + trace_dump_writes("\n"); + trace_stream_close(stream); + stream = NULL; + refcount = 0; + } +} + boolean trace_dump_trace_begin() { if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) return FALSE; - stream = trace_stream_create("gallium", "trace"); - if(!stream) - return FALSE; - - trace_dump_writes("\n"); - trace_dump_writes("\n"); - trace_dump_writes("\n"); + if(!stream) { + stream = trace_stream_create("gallium", "trace"); + if(!stream) + return FALSE; + + trace_dump_writes("\n"); + trace_dump_writes("\n"); + trace_dump_writes("\n"); + #if defined(PIPE_OS_LINUX) - /* Linux applications rarely cleanup GL / Gallium resources so catch - * application exit here */ - atexit(trace_dump_trace_end); + /* Linux applications rarely cleanup GL / Gallium resources so catch + * application exit here */ + atexit(trace_dump_trace_close); #endif + } + + ++refcount; return TRUE; } +boolean trace_dump_enabled(void) +{ + return stream ? TRUE : FALSE; +} + void trace_dump_trace_end(void) { - if(stream) { - trace_dump_writes("\n"); - trace_stream_close(stream); - stream = NULL; - } + if(stream) + if(!--refcount) + trace_dump_trace_close(); } void trace_dump_call_begin(const char *klass, const char *method) @@ -247,6 +267,7 @@ void trace_dump_call_end(void) trace_dump_indent(1); trace_dump_tag_end("call"); trace_dump_newline(); + trace_stream_flush(stream); } void trace_dump_arg_begin(const char *name) @@ -372,7 +393,7 @@ void trace_dump_null(void) void trace_dump_ptr(const void *value) { if(value) - trace_dump_writef("%p", value); + trace_dump_writef("0x%08lx", (unsigned long)(uintptr_t)value); else trace_dump_null(); } diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index 0beb1023b1..14176a78e9 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -39,6 +39,7 @@ boolean trace_dump_trace_begin(void); +boolean trace_dump_enabled(void); void trace_dump_trace_end(void); void trace_dump_call_begin(const char *klass, const char *method); void trace_dump_call_end(void); diff --git a/src/gallium/drivers/trace/tr_stream.c b/src/gallium/drivers/trace/tr_stream.c index 14cc257e15..aecc1286b8 100644 --- a/src/gallium/drivers/trace/tr_stream.c +++ b/src/gallium/drivers/trace/tr_stream.c @@ -86,6 +86,18 @@ trace_stream_write(struct trace_stream *stream, const void *data, size_t size) } +void +trace_stream_flush(struct trace_stream *stream) +{ + if(!stream) + return; + +#if defined(PIPE_OS_LINUX) + fflush(stream->file); +#endif +} + + void trace_stream_close(struct trace_stream *stream) { diff --git a/src/gallium/drivers/trace/tr_stream.h b/src/gallium/drivers/trace/tr_stream.h index 53e854aa91..679c4a725d 100644 --- a/src/gallium/drivers/trace/tr_stream.h +++ b/src/gallium/drivers/trace/tr_stream.h @@ -48,6 +48,9 @@ trace_stream_create(const char *name, const char *ext); boolean trace_stream_write(struct trace_stream *stream, const void *data, size_t size); +void +trace_stream_flush(struct trace_stream *stream); + void trace_stream_close(struct trace_stream *stream); -- cgit v1.2.3 From f40de50def1b7ee75dd320b151c025b7ddff45be Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Aug 2008 10:31:23 +0100 Subject: trace: Wrap all textures and surface created by the pipe driver. That is, Unfortunately, this causes a regression in softpipe, where the output gets tyled. --- src/gallium/drivers/trace/SConscript | 11 ++-- src/gallium/drivers/trace/tr_context.c | 73 ++++++++++++++++++++- src/gallium/drivers/trace/tr_screen.c | 104 +++++++++++++++++++++--------- src/gallium/drivers/trace/tr_screen.h | 16 +---- src/gallium/drivers/trace/tr_texture.c | 112 +++++++++++++++++++++++++++++++++ src/gallium/drivers/trace/tr_texture.h | 95 ++++++++++++++++++++++++++++ src/gallium/drivers/trace/tr_winsys.c | 16 +++++ 7 files changed, 378 insertions(+), 49 deletions(-) create mode 100644 src/gallium/drivers/trace/tr_texture.c create mode 100644 src/gallium/drivers/trace/tr_texture.h (limited to 'src') diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript index 35507e21e4..a21ced9176 100644 --- a/src/gallium/drivers/trace/SConscript +++ b/src/gallium/drivers/trace/SConscript @@ -3,14 +3,15 @@ Import('*') env = env.Clone() trace = env.ConvenienceLibrary( - target = 'trace', - source = [ + target = 'trace', + source = [ 'tr_context.c', 'tr_dump.c', - 'tr_screen.c', + 'tr_screen.c', 'tr_state.c', - 'tr_stream.c', + 'tr_stream.c', + 'tr_texture.c', 'tr_winsys.c', - ]) + ]) Export('trace') \ No newline at end of file diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index e81cab252c..529bed3c6b 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -30,9 +30,53 @@ #include "tr_dump.h" #include "tr_state.h" +#include "tr_screen.h" +#include "tr_texture.h" #include "tr_context.h" +static INLINE struct pipe_texture * +trace_texture_unwrap(struct trace_context *tr_ctx, + struct pipe_texture *texture) +{ + struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); + struct trace_texture *tr_tex; + + if(!texture) + return NULL; + + tr_tex = trace_texture(tr_scr, texture); + + assert(tr_tex->texture); + assert(tr_tex->texture->screen == tr_scr->screen); + return tr_tex->texture; +} + + +static INLINE struct pipe_surface * +trace_surface_unwrap(struct trace_context *tr_ctx, + struct pipe_surface *surface) +{ + struct trace_screen *tr_scr = trace_screen(tr_ctx->base.screen); + struct trace_texture *tr_tex; + struct trace_surface *tr_surf; + + if(!surface) + return NULL; + + assert(surface->texture); + if(!surface->texture) + return surface; + + tr_tex = trace_texture(tr_scr, surface->texture); + tr_surf = trace_surface(tr_tex, surface); + + assert(tr_surf->surface); + assert(tr_surf->surface->texture->screen == tr_scr->screen); + return tr_surf->surface; +} + + static INLINE void trace_context_set_edgeflags(struct pipe_context *_pipe, const unsigned *bitfield) @@ -666,7 +710,18 @@ trace_context_set_framebuffer_state(struct pipe_context *_pipe, { struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; - + struct pipe_framebuffer_state unwrapped_state; + unsigned i; + + /* Unwrap the input state */ + memcpy(&unwrapped_state, state, sizeof(unwrapped_state)); + for(i = 0; i < state->num_cbufs; ++i) + unwrapped_state.cbufs[i] = trace_surface_unwrap(tr_ctx, state->cbufs[i]); + for(i = state->num_cbufs; i < PIPE_MAX_COLOR_BUFS; ++i) + unwrapped_state.cbufs[i] = NULL; + unwrapped_state.zsbuf = trace_surface_unwrap(tr_ctx, state->zsbuf); + state = &unwrapped_state; + trace_dump_call_begin("pipe_context", "set_framebuffer_state"); trace_dump_arg(ptr, pipe); @@ -739,6 +794,12 @@ trace_context_set_sampler_textures(struct pipe_context *_pipe, { struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; + struct pipe_texture *unwrapped_textures[PIPE_MAX_SAMPLERS]; + unsigned i; + + for(i = 0; i < num_textures; ++i) + unwrapped_textures[i] = trace_texture_unwrap(tr_ctx, textures[i]); + textures = unwrapped_textures; trace_dump_call_begin("pipe_context", "set_sampler_textures"); @@ -810,6 +871,9 @@ trace_context_surface_copy(struct pipe_context *_pipe, struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; + dest = trace_surface_unwrap(tr_ctx, dest); + src = trace_surface_unwrap(tr_ctx, src); + trace_dump_call_begin("pipe_context", "surface_copy"); trace_dump_arg(ptr, pipe); @@ -841,6 +905,8 @@ trace_context_surface_fill(struct pipe_context *_pipe, struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; + dst = trace_surface_unwrap(tr_ctx, dst); + trace_dump_call_begin("pipe_context", "surface_fill"); trace_dump_arg(ptr, pipe); @@ -864,6 +930,8 @@ trace_context_clear(struct pipe_context *_pipe, struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; + surface = trace_surface_unwrap(tr_ctx, surface); + trace_dump_call_begin("pipe_context", "clear"); trace_dump_arg(ptr, pipe); @@ -925,6 +993,9 @@ trace_context_create(struct pipe_screen *screen, if(!pipe) goto error1; + if(!trace_dump_enabled()) + goto error1; + tr_ctx = CALLOC_STRUCT(trace_context); if(!tr_ctx) goto error1; diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index 2eecfdc187..cca8597a87 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -26,26 +26,14 @@ **************************************************************************/ #include "pipe/p_util.h" -#include "util/u_hash_table.h" #include "tr_dump.h" #include "tr_state.h" #include "tr_winsys.h" +#include "tr_texture.h" #include "tr_screen.h" -static unsigned trace_surface_hash(void *surface) -{ - return (unsigned)(uintptr_t)surface; -} - - -static int trace_surface_compare(void *surface1, void *surface2) -{ - return (char *)surface2 - (char *)surface1; -} - - static const char * trace_screen_get_name(struct pipe_screen *_screen) { @@ -182,6 +170,8 @@ trace_screen_texture_create(struct pipe_screen *_screen, trace_dump_call_end(); + result = trace_texture_create(tr_scr, result); + return result; } @@ -210,6 +200,8 @@ trace_screen_texture_blanket(struct pipe_screen *_screen, trace_dump_call_end(); + result = trace_texture_create(tr_scr, result); + return result; } @@ -220,18 +212,34 @@ trace_screen_texture_release(struct pipe_screen *_screen, { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; - struct pipe_texture *texture = *ptexture; + struct trace_texture *tr_tex; + struct pipe_texture *texture; + + assert(ptexture); + if(*ptexture) { + tr_tex = trace_texture(tr_scr, *ptexture); + texture = tr_tex->texture; + assert(texture->screen == screen); + } + else + texture = NULL; trace_dump_call_begin("pipe_screen", "texture_release"); trace_dump_arg(ptr, screen); trace_dump_arg(ptr, texture); - screen->texture_release(screen, ptexture); + if (*ptexture) { + if (!--(*ptexture)->refcount) + trace_texture_destroy(tr_scr, *ptexture); + + *ptexture = NULL; + } trace_dump_call_end(); } + static struct pipe_surface * trace_screen_get_tex_surface(struct pipe_screen *_screen, struct pipe_texture *texture, @@ -241,8 +249,14 @@ trace_screen_get_tex_surface(struct pipe_screen *_screen, { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; + struct trace_texture *tr_tex; struct pipe_surface *result; + assert(texture); + tr_tex = trace_texture(tr_scr, texture); + texture = tr_tex->texture; + assert(texture->screen == screen); + trace_dump_call_begin("pipe_screen", "get_tex_surface"); trace_dump_arg(ptr, screen); @@ -258,6 +272,8 @@ trace_screen_get_tex_surface(struct pipe_screen *_screen, trace_dump_call_end(); + result = trace_surface_create(tr_tex, result); + return result; } @@ -268,14 +284,30 @@ trace_screen_tex_surface_release(struct pipe_screen *_screen, { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; - struct pipe_surface *surface = *psurface; + struct trace_texture *tr_tex; + struct trace_surface *tr_surf; + struct pipe_surface *surface; + + assert(psurface); + if(*psurface) { + tr_tex = trace_texture(tr_scr, (*psurface)->texture); + tr_surf = trace_surface(tr_tex, *psurface); + surface = tr_surf->surface; + } + else + surface = NULL; trace_dump_call_begin("pipe_screen", "tex_surface_release"); trace_dump_arg(ptr, screen); trace_dump_arg(ptr, surface); - screen->tex_surface_release(screen, psurface); + if (*psurface) { + if (!--(*psurface)->refcount) + trace_surface_destroy(tr_tex, *psurface); + + *psurface = NULL; + } trace_dump_call_end(); } @@ -288,13 +320,19 @@ trace_screen_surface_map(struct pipe_screen *_screen, { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; + struct trace_texture *tr_tex; + struct trace_surface *tr_surf; void *map; + tr_tex = trace_texture(tr_scr, surface->texture); + tr_surf = trace_surface(tr_tex, surface); + surface = tr_surf->surface; + map = screen->surface_map(screen, surface, flags); if(map) { if(flags & PIPE_BUFFER_USAGE_CPU_WRITE) { - assert(!hash_table_get(tr_scr->surface_maps, surface)); - hash_table_set(tr_scr->surface_maps, surface, map); + assert(!tr_surf->map); + tr_surf->map = map; } } @@ -308,10 +346,14 @@ trace_screen_surface_unmap(struct pipe_screen *_screen, { struct trace_screen *tr_scr = trace_screen(_screen); struct pipe_screen *screen = tr_scr->screen; - const void *map; + struct trace_texture *tr_tex; + struct trace_surface *tr_surf; - map = hash_table_get(tr_scr->surface_maps, surface); - if(map) { + tr_tex = trace_texture(tr_scr, surface->texture); + tr_surf = trace_surface(tr_tex, surface); + surface = tr_surf->surface; + + if(tr_surf->map) { size_t size = surface->nblocksy * surface->stride; trace_dump_call_begin("pipe_winsys", "surface_write"); @@ -321,7 +363,7 @@ trace_screen_surface_unmap(struct pipe_screen *_screen, trace_dump_arg(ptr, surface); trace_dump_arg_begin("data"); - trace_dump_bytes(map, size); + trace_dump_bytes(tr_surf->map, size); trace_dump_arg_end(); trace_dump_arg_begin("stride"); @@ -334,7 +376,7 @@ trace_screen_surface_unmap(struct pipe_screen *_screen, trace_dump_call_end(); - hash_table_remove(tr_scr->surface_maps, surface); + tr_surf->map = NULL; } screen->surface_unmap(screen, surface); @@ -398,11 +440,6 @@ trace_screen_create(struct pipe_screen *screen) tr_scr->screen = screen; - tr_scr->surface_maps = hash_table_create(trace_surface_hash, - trace_surface_compare); - if(!tr_scr->surface_maps) - goto error3; - trace_dump_call_begin("", "pipe_screen_create"); trace_dump_arg_begin("winsys"); trace_dump_ptr(screen->winsys); @@ -419,3 +456,12 @@ error2: error1: return screen; } + + +struct trace_screen * +trace_screen(struct pipe_screen *screen) +{ + assert(screen); + assert(screen->destroy == trace_screen_destroy); + return (struct trace_screen *)screen; +} diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h index 90103aa922..698b84811d 100644 --- a/src/gallium/drivers/trace/tr_screen.h +++ b/src/gallium/drivers/trace/tr_screen.h @@ -29,31 +29,19 @@ #define TR_SCREEN_H_ -#include "pipe/p_compiler.h" -#include "pipe/p_debug.h" #include "pipe/p_screen.h" -struct hash_table; - - struct trace_screen { struct pipe_screen base; struct pipe_screen *screen; - - struct hash_table *surface_maps; }; -static INLINE struct trace_screen * -trace_screen(struct pipe_screen *screen) -{ - assert(screen); - return (struct trace_screen *)screen; -} - +struct trace_screen * +trace_screen(struct pipe_screen *screen); struct pipe_screen * diff --git a/src/gallium/drivers/trace/tr_texture.c b/src/gallium/drivers/trace/tr_texture.c new file mode 100644 index 0000000000..99ba74d366 --- /dev/null +++ b/src/gallium/drivers/trace/tr_texture.c @@ -0,0 +1,112 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" +#include "util/u_hash_table.h" + +#include "tr_screen.h" +#include "tr_texture.h" + + +struct pipe_texture * +trace_texture_create(struct trace_screen *tr_scr, + struct pipe_texture *texture) +{ + struct trace_texture *tr_tex; + + if(!texture) + goto error; + + assert(texture->screen == tr_scr->screen); + + tr_tex = CALLOC_STRUCT(trace_texture); + if(!tr_tex) + goto error; + + memcpy(&tr_tex->base, texture, sizeof(struct pipe_texture)); + tr_tex->base.screen = &tr_scr->base; + tr_tex->texture = texture; + + return &tr_tex->base; + +error: + pipe_texture_reference(&texture, NULL); + return NULL; +} + + +void +trace_texture_destroy(struct trace_screen *tr_scr, + struct pipe_texture *texture) +{ + struct trace_texture *tr_tex = trace_texture(tr_scr, texture); + pipe_texture_reference(&tr_tex->texture, NULL); + FREE(tr_tex); +} + + +struct pipe_surface * +trace_surface_create(struct trace_texture *tr_tex, + struct pipe_surface *surface) +{ + struct trace_surface *tr_surf; + + if(!surface) + goto error; + + assert(surface->texture == tr_tex->texture); + + tr_surf = CALLOC_STRUCT(trace_surface); + if(!tr_surf) + goto error; + + memcpy(&tr_surf->base, surface, sizeof(struct pipe_surface)); + + tr_surf->base.winsys = tr_tex->base.screen->winsys; + tr_surf->base.texture = NULL; + pipe_texture_reference(&tr_surf->base.texture, &tr_tex->base); + tr_surf->surface = surface; + + return &tr_surf->base; + +error: + pipe_surface_reference(&surface, NULL); + return NULL; +} + + +void +trace_surface_destroy(struct trace_texture *tr_tex, + struct pipe_surface *surface) +{ + struct trace_surface *tr_surf = trace_surface(tr_tex, surface); + pipe_texture_reference(&tr_surf->base.texture, NULL); + pipe_surface_reference(&tr_surf->surface, NULL); + FREE(tr_surf); +} + diff --git a/src/gallium/drivers/trace/tr_texture.h b/src/gallium/drivers/trace/tr_texture.h new file mode 100644 index 0000000000..9e72edb8a3 --- /dev/null +++ b/src/gallium/drivers/trace/tr_texture.h @@ -0,0 +1,95 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef TR_TEXTURE_H_ +#define TR_TEXTURE_H_ + + +#include "pipe/p_compiler.h" +#include "pipe/p_state.h" + +#include "tr_screen.h" + + +struct trace_texture +{ + struct pipe_texture base; + + struct pipe_texture *texture; +}; + + +struct trace_surface +{ + struct pipe_surface base; + + struct pipe_surface *surface; + + void *map; +}; + + +static INLINE struct trace_texture * +trace_texture(struct trace_screen *tr_scr, + struct pipe_texture *texture) +{ + if(!texture) + return NULL; + assert(texture->screen == &tr_scr->base); + return (struct trace_texture *)texture; +} + + +static INLINE struct trace_surface * +trace_surface(struct trace_texture *tr_tex, + struct pipe_surface *surface) +{ + if(!surface) + return NULL; + assert(surface->texture == &tr_tex->base); + return (struct trace_surface *)surface; +} + + +struct pipe_texture * +trace_texture_create(struct trace_screen *tr_scr, + struct pipe_texture *texture); + +void +trace_texture_destroy(struct trace_screen *tr_scr, + struct pipe_texture *texture); + +struct pipe_surface * +trace_surface_create(struct trace_texture *tr_tex, + struct pipe_surface *surface); + +void +trace_surface_destroy(struct trace_texture *tr_tex, + struct pipe_surface *surface); + + +#endif /* TR_TEXTURE_H_ */ diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index 356579809a..c426211408 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -31,6 +31,8 @@ #include "tr_dump.h" #include "tr_state.h" +#include "tr_screen.h" +#include "tr_texture.h" #include "tr_winsys.h" @@ -75,6 +77,14 @@ trace_winsys_flush_frontbuffer(struct pipe_winsys *_winsys, struct trace_winsys *tr_ws = trace_winsys(_winsys); struct pipe_winsys *winsys = tr_ws->winsys; + assert(surface); + if(surface->texture) { + struct trace_screen *tr_scr = trace_screen(surface->texture->screen); + struct trace_screen *tr_tex = trace_texture(tr_scr, surface->texture); + struct trace_surface *tr_surf = trace_surface(tr_tex, surface); + surface = tr_surf->surface; + } + trace_dump_call_begin("pipe_winsys", "flush_frontbuffer"); trace_dump_arg(ptr, winsys); @@ -106,6 +116,8 @@ trace_winsys_surface_alloc(struct pipe_winsys *_winsys) trace_dump_call_end(); + assert(!result || !result->texture); + return result; } @@ -122,6 +134,8 @@ trace_winsys_surface_alloc_storage(struct pipe_winsys *_winsys, struct pipe_winsys *winsys = tr_ws->winsys; int result; + assert(surface && !surface->texture); + trace_dump_call_begin("pipe_winsys", "surface_alloc_storage"); trace_dump_arg(ptr, winsys); @@ -155,6 +169,8 @@ trace_winsys_surface_release(struct pipe_winsys *_winsys, struct pipe_winsys *winsys = tr_ws->winsys; struct pipe_surface *surface = *psurface; + assert(psurface && *psurface && !(*psurface)->texture); + trace_dump_call_begin("pipe_winsys", "surface_release"); trace_dump_arg(ptr, winsys); -- cgit v1.2.3 From 52ecb8f56368180b5dec303d1d77a8d0596aaef7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Aug 2008 10:33:35 +0100 Subject: python/retrace: Update frame when PIPE_FLUSH_FRAME spotted. --- src/gallium/state_trackers/python/retrace/interpreter.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index 00b6376623..fa9a65365c 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -364,10 +364,11 @@ class Context(Object): def draw_arrays(self, mode, start, count): self.real.draw_arrays(mode, start, count) - self._update() def flush(self, flags): self.real.flush(flags) + if flags & gallium.PIPE_FLUSH_FRAME: + self._update() return None def clear(self, surface, value): -- cgit v1.2.3 From 9d58b2a432d44fc2861f2df1a611188f92dc288f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Aug 2008 11:20:57 +0100 Subject: trace: Fix typo. --- src/gallium/drivers/trace/tr_winsys.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index c426211408..120006ea9f 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -80,7 +80,7 @@ trace_winsys_flush_frontbuffer(struct pipe_winsys *_winsys, assert(surface); if(surface->texture) { struct trace_screen *tr_scr = trace_screen(surface->texture->screen); - struct trace_screen *tr_tex = trace_texture(tr_scr, surface->texture); + struct trace_texture *tr_tex = trace_texture(tr_scr, surface->texture); struct trace_surface *tr_surf = trace_surface(tr_tex, surface); surface = tr_surf->surface; } -- cgit v1.2.3 From 4b1b5ca9bd4d861ac8654c510f12b52a1a646ec4 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 15 Aug 2008 11:24:06 +0100 Subject: xlib: Kill xmesa_surface. A winsys cannot expect that the surfaces passed display_surface are the surfaces it created, as surface are now in generate texture views created by pipe_screen. Indeed, the fact it was working so far was mere luck. This fixes a weird tiled output when using the trace pipe driver. Winsys' surfaces should die. --- src/gallium/winsys/xlib/xm_winsys.c | 62 ++++++++++++------------------------- 1 file changed, 19 insertions(+), 43 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 3ab4c67cae..4b4dc56e84 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -81,18 +81,6 @@ struct xm_buffer }; -/** - * Subclass of pipe_surface for Xlib winsys - */ -struct xmesa_surface -{ - struct pipe_surface surface; - - int tileSize; - boolean no_swap; -}; - - /** * Subclass of pipe_winsys for Xlib winsys */ @@ -105,14 +93,6 @@ struct xmesa_pipe_winsys -/** Cast wrapper */ -static INLINE struct xmesa_surface * -xmesa_surface(struct pipe_surface *ps) -{ - return (struct xmesa_surface *) ps; -} - - /** Cast wrapper */ static INLINE struct xm_buffer * xm_buffer( struct pipe_buffer *buf ) @@ -358,13 +338,24 @@ xmesa_display_surface(XMesaBuffer b, const struct pipe_surface *surf) { XImage *ximage; struct xm_buffer *xm_buf = xm_buffer(surf->buffer); - const struct xmesa_surface *xm_surf - = xmesa_surface((struct pipe_surface *) surf); + static boolean no_swap = 0; + static boolean firsttime = 1; + static int tileSize = 0; - if (xm_surf->no_swap) + if (firsttime) { + no_swap = getenv("SP_NO_RAST") != NULL; +#ifdef GALLIUM_CELL + if (!getenv("GALLIUM_NOCELL")) { + tileSize = 32; /** probably temporary */ + } +#endif + firsttime = 0; + } + + if (no_swap) return; - if (xm_surf->tileSize) { + if (tileSize) { xmesa_display_surface_tiled(b, surf); return; } @@ -531,29 +522,14 @@ xm_surface_alloc_storage(struct pipe_winsys *winsys, static struct pipe_surface * xm_surface_alloc(struct pipe_winsys *ws) { - struct xmesa_surface *xms = CALLOC_STRUCT(xmesa_surface); - static boolean no_swap = 0; - static boolean firsttime = 1; - - if (firsttime) { - no_swap = getenv("SP_NO_RAST") != NULL; - firsttime = 0; - } + struct pipe_surface *surface = CALLOC_STRUCT(pipe_surface); assert(ws); - xms->surface.refcount = 1; - xms->surface.winsys = ws; + surface->refcount = 1; + surface->winsys = ws; -#ifdef GALLIUM_CELL - if (!getenv("GALLIUM_NOCELL")) { - xms->tileSize = 32; /** probably temporary */ - } -#endif - - xms->no_swap = no_swap; - - return &xms->surface; + return surface; } -- cgit v1.2.3 From 52a6b7e6da8cabf7104c17becb81dea26c44db18 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 15 Aug 2008 11:50:29 -0600 Subject: gallium: in st_draw_vbo() use ctx->Current.Attrib[] values when arrays are missing/null fixes potential segfaults when vertex arrays are missing --- src/mesa/state_tracker/st_draw.c | 31 +++++++++++++++++++++---------- 1 file changed, 21 insertions(+), 10 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c index cc3d7450a9..e1bc108eae 100644 --- a/src/mesa/state_tracker/st_draw.c +++ b/src/mesa/state_tracker/st_draw.c @@ -289,6 +289,7 @@ st_draw_vbo(GLcontext *ctx, for (attr = 0; attr < vp->num_inputs; attr++) { const GLuint mesaAttr = vp->index_to_input[attr]; struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj; + GLsizei stride = arrays[mesaAttr]->StrideB; if (bufobj && bufobj->Name) { /* Attribute data is in a VBO. @@ -307,17 +308,27 @@ st_draw_vbo(GLcontext *ctx, /* attribute data is in user-space memory, not a VBO */ uint bytes; - if (!arrays[mesaAttr]->StrideB) { - bytes = arrays[mesaAttr]->Size - * _mesa_sizeof_type(arrays[mesaAttr]->Type); - } else { - bytes = arrays[mesaAttr]->StrideB * (max_index + 1); + /* wrap user data */ + if (arrays[mesaAttr]->Ptr) { + /* user's vertex array */ + if (arrays[mesaAttr]->StrideB) { + bytes = arrays[mesaAttr]->StrideB * (max_index + 1); + } + else { + bytes = arrays[mesaAttr]->Size + * _mesa_sizeof_type(arrays[mesaAttr]->Type); + } + vbuffer[attr].buffer = pipe_user_buffer_create(pipe, + (void *) arrays[mesaAttr]->Ptr, bytes); + } + else { + /* no array, use ctx->Current.Attrib[] value */ + bytes = sizeof(ctx->Current.Attrib[0]); + vbuffer[attr].buffer = pipe_user_buffer_create(pipe, + (void *) ctx->Current.Attrib[mesaAttr], bytes); + stride = 0; } - /* wrap user data */ - vbuffer[attr].buffer - = pipe_user_buffer_create(pipe, (void *) arrays[mesaAttr]->Ptr, - bytes); vbuffer[attr].buffer_offset = 0; velements[attr].src_offset = 0; } @@ -325,7 +336,7 @@ st_draw_vbo(GLcontext *ctx, assert(velements[attr].src_offset <= 2048); /* 11-bit field */ /* common-case setup */ - vbuffer[attr].pitch = arrays[mesaAttr]->StrideB; /* in bytes */ + vbuffer[attr].pitch = stride; /* in bytes */ vbuffer[attr].max_index = max_index; velements[attr].vertex_buffer_index = attr; velements[attr].nr_components = arrays[mesaAttr]->Size; -- cgit v1.2.3 From 2e40e44bf43d968e64d313cbb3b325a9772d5389 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 15 Aug 2008 15:00:18 -0600 Subject: mesa: glsl: fix linking of varying vars which are arrays --- src/mesa/shader/slang/slang_codegen.c | 6 ++- src/mesa/shader/slang/slang_link.c | 80 +++++++++++++++++------------------ 2 files changed, 44 insertions(+), 42 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index db25106d03..f96c655759 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -3727,6 +3727,8 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, if (dbg) printf("UNIFORM (sz %d) ", totalSize); } else if (var->type.qualifier == SLANG_QUAL_VARYING) { + const GLint totalSize = array_size(size, var->array_len); + /* varyings must be float, vec or mat */ if (!_slang_type_is_float_vec_mat(var->type.specifier.type) && var->type.specifier.type != SLANG_SPEC_ARRAY) { @@ -3744,10 +3746,10 @@ _slang_codegen_global_variable(slang_assemble_ctx *A, slang_variable *var, if (prog) { /* user-defined varying */ - GLint varyingLoc = _mesa_add_varying(prog->Varying, varName, size); + GLint varyingLoc = _mesa_add_varying(prog->Varying, varName, totalSize); GLuint swizzle = _slang_var_swizzle(size, 0); store = _slang_new_ir_storage_swz(PROGRAM_VARYING, varyingLoc, - size, swizzle); + totalSize, swizzle); } else { /* pre-defined varying, like gl_Color or gl_TexCoord */ diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index d96380e03f..e33272eff2 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -42,6 +42,20 @@ #include "slang_link.h" +/** + * Record a linking error. + */ +static void +link_error(struct gl_shader_program *shProg, const char *msg) +{ + if (shProg->InfoLog) { + _mesa_free(shProg->InfoLog); + } + shProg->InfoLog = _mesa_strdup(msg); + shProg->LinkStatus = GL_FALSE; +} + + /** * Linking varying vars involves rearranging varying vars so that the @@ -52,7 +66,6 @@ static GLboolean link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog) { GLuint *map, i, firstVarying, newFile; - GLbitfield varsWritten, varsRead; map = (GLuint *) malloc(prog->Varying->NumParameters * sizeof(GLuint)); if (!map) @@ -60,14 +73,13 @@ link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog) for (i = 0; i < prog->Varying->NumParameters; i++) { /* see if this varying is in the linked varying list */ - const struct gl_program_parameter *var - = prog->Varying->Parameters + i; - + const struct gl_program_parameter *var = prog->Varying->Parameters + i; GLint j = _mesa_lookup_parameter_index(shProg->Varying, -1, var->Name); if (j >= 0) { /* already in list, check size */ if (var->Size != shProg->Varying->Parameters[j].Size) { /* error */ + link_error(shProg, "mismatched varying variable types"); return GL_FALSE; } } @@ -75,9 +87,19 @@ link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog) /* not already in linked list */ j = _mesa_add_varying(shProg->Varying, var->Name, var->Size); } - ASSERT(j >= 0); - map[i] = j; + /* map varying[i] to varying[j]. + * Note: the loop here takes care of arrays or large (sz>4) vars. + */ + { + GLint sz = var->Size; + while (sz > 0) { + /*printf("Link varying from %d to %d\n", i, j);*/ + map[i++] = j++; + sz -= 4; + } + i--; /* go back one */ + } } @@ -96,9 +118,6 @@ link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog) newFile = PROGRAM_INPUT; } - /* keep track of which varying vars we read and write */ - varsWritten = varsRead = 0x0; - /* OK, now scan the program/shader instructions looking for varying vars, * replacing the old index with the new index. */ @@ -109,30 +128,22 @@ link_varying_vars(struct gl_shader_program *shProg, struct gl_program *prog) if (inst->DstReg.File == PROGRAM_VARYING) { inst->DstReg.File = newFile; inst->DstReg.Index = map[ inst->DstReg.Index ] + firstVarying; - varsWritten |= (1 << inst->DstReg.Index); } for (j = 0; j < 3; j++) { if (inst->SrcReg[j].File == PROGRAM_VARYING) { inst->SrcReg[j].File = newFile; inst->SrcReg[j].Index = map[ inst->SrcReg[j].Index ] + firstVarying; - varsRead |= (1 << inst->SrcReg[j].Index); } } } - if (prog->Target == GL_VERTEX_PROGRAM_ARB) { - prog->OutputsWritten |= varsWritten; - /*printf("VERT OUTPUTS: 0x%x \n", varsWritten);*/ - } - else { - assert(prog->Target == GL_FRAGMENT_PROGRAM_ARB); - prog->InputsRead |= varsRead; - /*printf("FRAG INPUTS: 0x%x\n", varsRead);*/ - } - free(map); + /* these will get recomputed before linking is completed */ + prog->InputsRead = 0x0; + prog->OutputsWritten = 0x0; + return GL_TRUE; } @@ -381,21 +392,6 @@ fragment_program(struct gl_program *prog) } -/** - * Record a linking error. - */ -static void -link_error(struct gl_shader_program *shProg, const char *msg) -{ - if (shProg->InfoLog) { - _mesa_free(shProg->InfoLog); - } - shProg->InfoLog = _mesa_strdup(msg); - shProg->LinkStatus = GL_FALSE; -} - - - /** * Shader linker. Currently: * @@ -478,10 +474,14 @@ _slang_link(GLcontext *ctx, } /* link varying vars */ - if (shProg->VertexProgram) - link_varying_vars(shProg, &shProg->VertexProgram->Base); - if (shProg->FragmentProgram) - link_varying_vars(shProg, &shProg->FragmentProgram->Base); + if (shProg->VertexProgram) { + if (!link_varying_vars(shProg, &shProg->VertexProgram->Base)) + return; + } + if (shProg->FragmentProgram) { + if (!link_varying_vars(shProg, &shProg->FragmentProgram->Base)) + return; + } /* link uniform vars */ if (shProg->VertexProgram) -- cgit v1.2.3 From 404166217ac4ae7f5ef656be7352c98b9c8197da Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Hanno=20B=C3=B6ck?= Date: Fri, 15 Aug 2008 17:14:54 -0600 Subject: fix DESTDIR omissions --- src/glu/Makefile | 4 ++-- src/glw/Makefile | 8 ++++---- 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/glu/Makefile b/src/glu/Makefile index 5ddc50946a..b8c55db6d0 100644 --- a/src/glu/Makefile +++ b/src/glu/Makefile @@ -14,8 +14,8 @@ default: $(TOP)/configs/current done install: - $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) - $(INSTALL) $(TOP)/$(LIB_DIR)/libGLU.* $(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) -d $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) $(TOP)/$(LIB_DIR)/libGLU.* $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) clean: @for dir in $(SUBDIRS) ; do \ diff --git a/src/glw/Makefile b/src/glw/Makefile index 12ab2facd4..86a0bbaeac 100644 --- a/src/glw/Makefile +++ b/src/glw/Makefile @@ -26,10 +26,10 @@ OBJECTS = $(GLW_SOURCES:.c=.o) default: $(TOP)/$(LIB_DIR)/$(GLW_LIB_NAME) install: - $(INSTALL) -d $(INSTALL_DIR)/include/GL - $(INSTALL) -d $(INSTALL_DIR)/$(LIB_DIR) - $(INSTALL) -m 644 *.h $(INSTALL_DIR)/include/GL - $(INSTALL) $(TOP)/$(LIB_DIR)/libGLw.* $(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) -d $(DESTDIR)$(INSTALL_DIR)/include/GL + $(INSTALL) -d $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) + $(INSTALL) -m 644 *.h $(DESTDIR)$(INSTALL_DIR)/include/GL + $(INSTALL) $(TOP)/$(LIB_DIR)/libGLw.* $(DESTDIR)$(INSTALL_DIR)/$(LIB_DIR) clean: -rm depend depend.bak -- cgit v1.2.3 From 2037b1381c129c74ba87a092484258608583d34e Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Sun, 17 Aug 2008 19:14:00 +0100 Subject: rather than use CRTC 1 (aka LVDS), try and setup all connectors/encoders and get the same output on all. --- src/gallium/winsys/egl_drm/intel/intel_egl.c | 111 ++++++++++++++++++++++----- 1 file changed, 92 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c index 55f4d92afb..1851babaf6 100644 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ b/src/gallium/winsys/egl_drm/intel/intel_egl.c @@ -66,7 +66,67 @@ egl_drm_create_device(int drmFD) return device; } -__GLcontextModes* _gl_context_modes_create( unsigned count, size_t minimum_size ); +static void +_egl_context_modes_destroy(__GLcontextModes *modes) +{ + _eglLog(_EGL_DEBUG, "%s", __FUNCTION__); + + while (modes) { + __GLcontextModes * const next = modes->next; + free(modes); + modes = next; + } +} +/** + * Create a linked list of 'count' GLcontextModes. + * These are used during the client/server visual negotiation phase, + * then discarded. + */ +static __GLcontextModes * +_egl_context_modes_create(unsigned count, size_t minimum_size) +{ + /* This code copied from libGLX, and modified */ + const size_t size = (minimum_size > sizeof(__GLcontextModes)) + ? minimum_size : sizeof(__GLcontextModes); + __GLcontextModes * head = NULL; + __GLcontextModes ** next; + unsigned i; + + _eglLog(_EGL_DEBUG, "%s %d %d", __FUNCTION__, count, minimum_size); + + next = & head; + for (i = 0 ; i < count ; i++) { + *next = (__GLcontextModes *) calloc(1, size); + if (*next == NULL) { + _egl_context_modes_destroy(head); + head = NULL; + break; + } + + (*next)->doubleBufferMode = 1; + (*next)->visualID = GLX_DONT_CARE; + (*next)->visualType = GLX_DONT_CARE; + (*next)->visualRating = GLX_NONE; + (*next)->transparentPixel = GLX_NONE; + (*next)->transparentRed = GLX_DONT_CARE; + (*next)->transparentGreen = GLX_DONT_CARE; + (*next)->transparentBlue = GLX_DONT_CARE; + (*next)->transparentAlpha = GLX_DONT_CARE; + (*next)->transparentIndex = GLX_DONT_CARE; + (*next)->xRenderable = GLX_DONT_CARE; + (*next)->fbconfigID = GLX_DONT_CARE; + (*next)->swapMethod = GLX_SWAP_UNDEFINED_OML; + (*next)->bindToTextureRgb = GLX_DONT_CARE; + (*next)->bindToTextureRgba = GLX_DONT_CARE; + (*next)->bindToMipmapTexture = GLX_DONT_CARE; + (*next)->bindToTextureTargets = 0; + (*next)->yInverted = GLX_DONT_CARE; + + next = & ((*next)->next); + } + + return head; +} struct drm_screen; @@ -102,7 +162,6 @@ struct drm_screen /* currently only support one connector */ drmModeConnectorPtr connector; - uint32_t connectorID; /* Has this screen been shown */ int shown; @@ -190,7 +249,6 @@ drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) screen = malloc(sizeof(struct drm_screen)); memset(screen, 0, sizeof(*screen)); - screen->connectorID = res->connectors[i]; screen->connector = connector; _eglInitScreen(&screen->base); _eglAddScreen(disp, &screen->base); @@ -225,17 +283,20 @@ static void drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen) { struct drm_driver *drm_drv = (struct drm_driver *)drv; + unsigned int i; intel_bind_frontbuffer(screen->surf->drawable, NULL); screen->surf = NULL; - drmModeSetCrtc( - drm_drv->device->drmFD, - drm_drv->res->crtcs[1], - 0, // FD - 0, 0, - NULL, 0, // List of output ids - NULL); + for (i = 0; i < drm_drv->res->count_crtcs; i++) { + drmModeSetCrtc( + drm_drv->device->drmFD, + drm_drv->res->crtcs[i], + 0, // FD + 0, 0, + NULL, 0, // List of output ids + NULL); + } drmModeRmFB(drm_drv->device->drmFD, screen->fbID); drmModeFreeFB(screen->fb); @@ -560,9 +621,10 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, struct drm_surface *surf = lookup_drm_surface(surface); struct drm_screen *scrn = lookup_drm_screen(dpy, screen); _EGLMode *mode = _eglLookupMode(dpy, m); - size_t pitch = 2048 * 4; + size_t pitch = mode->Width * 4; size_t size = mode->Height * pitch; int ret; + unsigned int i,j,k; if (scrn->shown) drm_takedown_shown_screen(drv, scrn); @@ -596,14 +658,25 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, if (!scrn->mode) goto err_fb; - ret = drmModeSetCrtc( - drm_drv->device->drmFD, - drm_drv->res->crtcs[1], - scrn->fbID, - 0, 0, - &scrn->connectorID, 1, - scrn->mode); - + for (j = 0; j < drm_drv->res->count_connectors; j++) { + drmModeConnector *con = drmModeGetConnector(drm_drv->device->drmFD, drm_drv->res->connectors[j]); + for (k = 0; k < con->count_encoders; k++) { + drmModeEncoder *enc = drmModeGetEncoder(drm_drv->device->drmFD, con->encoders[k]); + for (i = 0; i < drm_drv->res->count_crtcs; i++) { + if (enc->possible_crtcs & (1<device->drmFD, + drm_drv->res->crtcs[i], + scrn->fbID, + 0, 0, + &drm_drv->res->connectors[j], 1, + scrn->mode); + /* skip the other crtcs now */ + i = drm_drv->res->count_crtcs; + } + } + } + } scrn->front.handle = scrn->buffer.handle; scrn->front.pitch = pitch; -- cgit v1.2.3 From ccf1910dd4b2a8ccd04ddbdf725b6dd3f8026eee Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Sun, 17 Aug 2008 20:17:18 +0100 Subject: consolidate intel directories. we now have src/gallium/winsys/drm/intel/{common,dri,egl} --- src/gallium/winsys/SConscript | 4 +- src/gallium/winsys/common/Makefile | 20 - src/gallium/winsys/common/Makefile.template | 64 -- src/gallium/winsys/common/intel_drm/Makefile | 23 - src/gallium/winsys/common/intel_drm/glthread.h | 359 -------- .../winsys/common/intel_drm/intel_be_batchbuffer.c | 429 --------- .../winsys/common/intel_drm/intel_be_batchbuffer.h | 69 -- .../winsys/common/intel_drm/intel_be_context.c | 107 --- .../winsys/common/intel_drm/intel_be_context.h | 40 - .../winsys/common/intel_drm/intel_be_device.c | 308 ------- .../winsys/common/intel_drm/intel_be_device.h | 72 -- .../winsys/common/intel_drm/ws_dri_bufmgr.c | 949 -------------------- .../winsys/common/intel_drm/ws_dri_bufmgr.h | 138 --- .../winsys/common/intel_drm/ws_dri_bufpool.h | 102 --- .../winsys/common/intel_drm/ws_dri_drmpool.c | 268 ------ .../winsys/common/intel_drm/ws_dri_fencemgr.c | 377 -------- .../winsys/common/intel_drm/ws_dri_fencemgr.h | 115 --- .../winsys/common/intel_drm/ws_dri_mallocpool.c | 161 ---- .../winsys/common/intel_drm/ws_dri_slabpool.c | 968 --------------------- src/gallium/winsys/dri/Makefile | 38 - src/gallium/winsys/dri/Makefile.template | 125 --- src/gallium/winsys/dri/SConscript | 54 -- src/gallium/winsys/dri/intel/Makefile | 34 - src/gallium/winsys/dri/intel/SConscript | 41 - src/gallium/winsys/dri/intel/intel_batchbuffer.h | 24 - src/gallium/winsys/dri/intel/intel_context.c | 337 ------- src/gallium/winsys/dri/intel/intel_context.h | 164 ---- src/gallium/winsys/dri/intel/intel_lock.c | 102 --- src/gallium/winsys/dri/intel/intel_reg.h | 53 -- src/gallium/winsys/dri/intel/intel_screen.c | 607 ------------- src/gallium/winsys/dri/intel/intel_screen.h | 122 --- src/gallium/winsys/dri/intel/intel_swapbuffers.c | 261 ------ src/gallium/winsys/dri/intel/intel_swapbuffers.h | 47 - .../winsys/dri/intel/intel_winsys_softpipe.c | 82 -- .../winsys/dri/intel/intel_winsys_softpipe.h | 39 - src/gallium/winsys/dri/intel/server/i830_common.h | 255 ------ src/gallium/winsys/dri/intel/server/i830_dri.h | 62 -- src/gallium/winsys/drm/Makefile | 38 + src/gallium/winsys/drm/Makefile.template | 125 +++ src/gallium/winsys/drm/SConscript | 54 ++ src/gallium/winsys/drm/intel/Makefile | 25 + src/gallium/winsys/drm/intel/common/Makefile | 23 + .../winsys/drm/intel/common/Makefile.template | 64 ++ src/gallium/winsys/drm/intel/common/glthread.h | 359 ++++++++ .../winsys/drm/intel/common/intel_be_batchbuffer.c | 429 +++++++++ .../winsys/drm/intel/common/intel_be_batchbuffer.h | 69 ++ .../winsys/drm/intel/common/intel_be_context.c | 107 +++ .../winsys/drm/intel/common/intel_be_context.h | 40 + .../winsys/drm/intel/common/intel_be_device.c | 308 +++++++ .../winsys/drm/intel/common/intel_be_device.h | 72 ++ .../winsys/drm/intel/common/ws_dri_bufmgr.c | 949 ++++++++++++++++++++ .../winsys/drm/intel/common/ws_dri_bufmgr.h | 138 +++ .../winsys/drm/intel/common/ws_dri_bufpool.h | 102 +++ .../winsys/drm/intel/common/ws_dri_drmpool.c | 268 ++++++ .../winsys/drm/intel/common/ws_dri_fencemgr.c | 377 ++++++++ .../winsys/drm/intel/common/ws_dri_fencemgr.h | 115 +++ .../winsys/drm/intel/common/ws_dri_mallocpool.c | 161 ++++ .../winsys/drm/intel/common/ws_dri_slabpool.c | 968 +++++++++++++++++++++ src/gallium/winsys/drm/intel/dri/Makefile | 33 + src/gallium/winsys/drm/intel/dri/SConscript | 41 + .../winsys/drm/intel/dri/intel_batchbuffer.h | 24 + src/gallium/winsys/drm/intel/dri/intel_context.c | 337 +++++++ src/gallium/winsys/drm/intel/dri/intel_context.h | 164 ++++ src/gallium/winsys/drm/intel/dri/intel_lock.c | 102 +++ src/gallium/winsys/drm/intel/dri/intel_reg.h | 53 ++ src/gallium/winsys/drm/intel/dri/intel_screen.c | 607 +++++++++++++ src/gallium/winsys/drm/intel/dri/intel_screen.h | 122 +++ .../winsys/drm/intel/dri/intel_swapbuffers.c | 261 ++++++ .../winsys/drm/intel/dri/intel_swapbuffers.h | 47 + .../winsys/drm/intel/dri/intel_winsys_softpipe.c | 82 ++ .../winsys/drm/intel/dri/intel_winsys_softpipe.h | 39 + .../winsys/drm/intel/dri/server/i830_common.h | 255 ++++++ src/gallium/winsys/drm/intel/dri/server/i830_dri.h | 62 ++ src/gallium/winsys/drm/intel/egl/Makefile | 28 + src/gallium/winsys/drm/intel/egl/SConscript | 39 + .../winsys/drm/intel/egl/intel_batchbuffer.h | 24 + src/gallium/winsys/drm/intel/egl/intel_context.c | 242 ++++++ src/gallium/winsys/drm/intel/egl/intel_context.h | 118 +++ src/gallium/winsys/drm/intel/egl/intel_device.c | 137 +++ src/gallium/winsys/drm/intel/egl/intel_device.h | 50 ++ src/gallium/winsys/drm/intel/egl/intel_egl.c | 796 +++++++++++++++++ src/gallium/winsys/drm/intel/egl/intel_egl.h | 53 ++ src/gallium/winsys/drm/intel/egl/intel_reg.h | 53 ++ .../winsys/drm/intel/egl/intel_swapbuffers.c | 111 +++ src/gallium/winsys/egl_drm/Makefile | 38 - src/gallium/winsys/egl_drm/Makefile.template | 117 --- src/gallium/winsys/egl_drm/intel/Makefile | 29 - src/gallium/winsys/egl_drm/intel/SConscript | 39 - .../winsys/egl_drm/intel/intel_batchbuffer.h | 24 - src/gallium/winsys/egl_drm/intel/intel_context.c | 242 ------ src/gallium/winsys/egl_drm/intel/intel_context.h | 118 --- src/gallium/winsys/egl_drm/intel/intel_device.c | 137 --- src/gallium/winsys/egl_drm/intel/intel_device.h | 50 -- src/gallium/winsys/egl_drm/intel/intel_egl.c | 796 ----------------- src/gallium/winsys/egl_drm/intel/intel_egl.h | 53 -- src/gallium/winsys/egl_drm/intel/intel_reg.h | 53 -- .../winsys/egl_drm/intel/intel_swapbuffers.c | 111 --- src/gallium/winsys/gdi/wmesa.c | 70 -- 98 files changed, 8673 insertions(+), 8895 deletions(-) delete mode 100644 src/gallium/winsys/common/Makefile delete mode 100644 src/gallium/winsys/common/Makefile.template delete mode 100644 src/gallium/winsys/common/intel_drm/Makefile delete mode 100644 src/gallium/winsys/common/intel_drm/glthread.h delete mode 100644 src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.c delete mode 100644 src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.h delete mode 100644 src/gallium/winsys/common/intel_drm/intel_be_context.c delete mode 100644 src/gallium/winsys/common/intel_drm/intel_be_context.h delete mode 100644 src/gallium/winsys/common/intel_drm/intel_be_device.c delete mode 100644 src/gallium/winsys/common/intel_drm/intel_be_device.h delete mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.c delete mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.h delete mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_bufpool.h delete mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_drmpool.c delete mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.c delete mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.h delete mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_mallocpool.c delete mode 100644 src/gallium/winsys/common/intel_drm/ws_dri_slabpool.c delete mode 100644 src/gallium/winsys/dri/Makefile delete mode 100644 src/gallium/winsys/dri/Makefile.template delete mode 100644 src/gallium/winsys/dri/SConscript delete mode 100644 src/gallium/winsys/dri/intel/Makefile delete mode 100644 src/gallium/winsys/dri/intel/SConscript delete mode 100644 src/gallium/winsys/dri/intel/intel_batchbuffer.h delete mode 100644 src/gallium/winsys/dri/intel/intel_context.c delete mode 100644 src/gallium/winsys/dri/intel/intel_context.h delete mode 100644 src/gallium/winsys/dri/intel/intel_lock.c delete mode 100644 src/gallium/winsys/dri/intel/intel_reg.h delete mode 100644 src/gallium/winsys/dri/intel/intel_screen.c delete mode 100644 src/gallium/winsys/dri/intel/intel_screen.h delete mode 100644 src/gallium/winsys/dri/intel/intel_swapbuffers.c delete mode 100644 src/gallium/winsys/dri/intel/intel_swapbuffers.h delete mode 100644 src/gallium/winsys/dri/intel/intel_winsys_softpipe.c delete mode 100644 src/gallium/winsys/dri/intel/intel_winsys_softpipe.h delete mode 100644 src/gallium/winsys/dri/intel/server/i830_common.h delete mode 100644 src/gallium/winsys/dri/intel/server/i830_dri.h create mode 100644 src/gallium/winsys/drm/Makefile create mode 100644 src/gallium/winsys/drm/Makefile.template create mode 100644 src/gallium/winsys/drm/SConscript create mode 100644 src/gallium/winsys/drm/intel/Makefile create mode 100644 src/gallium/winsys/drm/intel/common/Makefile create mode 100644 src/gallium/winsys/drm/intel/common/Makefile.template create mode 100644 src/gallium/winsys/drm/intel/common/glthread.h create mode 100644 src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c create mode 100644 src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h create mode 100644 src/gallium/winsys/drm/intel/common/intel_be_context.c create mode 100644 src/gallium/winsys/drm/intel/common/intel_be_context.h create mode 100644 src/gallium/winsys/drm/intel/common/intel_be_device.c create mode 100644 src/gallium/winsys/drm/intel/common/intel_be_device.h create mode 100644 src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c create mode 100644 src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h create mode 100644 src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h create mode 100644 src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c create mode 100644 src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c create mode 100644 src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h create mode 100644 src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c create mode 100644 src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c create mode 100644 src/gallium/winsys/drm/intel/dri/Makefile create mode 100644 src/gallium/winsys/drm/intel/dri/SConscript create mode 100644 src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h create mode 100644 src/gallium/winsys/drm/intel/dri/intel_context.c create mode 100644 src/gallium/winsys/drm/intel/dri/intel_context.h create mode 100644 src/gallium/winsys/drm/intel/dri/intel_lock.c create mode 100644 src/gallium/winsys/drm/intel/dri/intel_reg.h create mode 100644 src/gallium/winsys/drm/intel/dri/intel_screen.c create mode 100644 src/gallium/winsys/drm/intel/dri/intel_screen.h create mode 100644 src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c create mode 100644 src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h create mode 100644 src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c create mode 100644 src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h create mode 100644 src/gallium/winsys/drm/intel/dri/server/i830_common.h create mode 100644 src/gallium/winsys/drm/intel/dri/server/i830_dri.h create mode 100644 src/gallium/winsys/drm/intel/egl/Makefile create mode 100644 src/gallium/winsys/drm/intel/egl/SConscript create mode 100644 src/gallium/winsys/drm/intel/egl/intel_batchbuffer.h create mode 100644 src/gallium/winsys/drm/intel/egl/intel_context.c create mode 100644 src/gallium/winsys/drm/intel/egl/intel_context.h create mode 100644 src/gallium/winsys/drm/intel/egl/intel_device.c create mode 100644 src/gallium/winsys/drm/intel/egl/intel_device.h create mode 100644 src/gallium/winsys/drm/intel/egl/intel_egl.c create mode 100644 src/gallium/winsys/drm/intel/egl/intel_egl.h create mode 100644 src/gallium/winsys/drm/intel/egl/intel_reg.h create mode 100644 src/gallium/winsys/drm/intel/egl/intel_swapbuffers.c delete mode 100644 src/gallium/winsys/egl_drm/Makefile delete mode 100644 src/gallium/winsys/egl_drm/Makefile.template delete mode 100644 src/gallium/winsys/egl_drm/intel/Makefile delete mode 100644 src/gallium/winsys/egl_drm/intel/SConscript delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_context.c delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_context.h delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_device.c delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_device.h delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_egl.c delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_egl.h delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_reg.h delete mode 100644 src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c (limited to 'src') diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript index bf1718e7a5..ddab9efc86 100644 --- a/src/gallium/winsys/SConscript +++ b/src/gallium/winsys/SConscript @@ -1,8 +1,8 @@ Import('*') -if env['dri']: +if env['drm']: SConscript([ - 'dri/SConscript', + 'drm/SConscript', ]) if 'xlib' in env['winsys']: diff --git a/src/gallium/winsys/common/Makefile b/src/gallium/winsys/common/Makefile deleted file mode 100644 index 4c0f3545a5..0000000000 --- a/src/gallium/winsys/common/Makefile +++ /dev/null @@ -1,20 +0,0 @@ -TOP = ../../../.. -include $(TOP)/configs/current - - -SUBDIRS = $(GALLIUM_WINSYS_COMMON_DIRS) - - -default: subdirs - - -subdirs: - @for dir in $(SUBDIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ - fi \ - done - - -clean: - rm -f `find . -name \*.[oa]` diff --git a/src/gallium/winsys/common/Makefile.template b/src/gallium/winsys/common/Makefile.template deleted file mode 100644 index 67af778157..0000000000 --- a/src/gallium/winsys/common/Makefile.template +++ /dev/null @@ -1,64 +0,0 @@ -# -*-makefile-*- - - -# We still have a dependency on the "dri" buffer manager. Most likely -# the interface can be reused in non-dri environments, and also as a -# frontend to simpler memory managers. -# -COMMON_SOURCES = - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(CPP_SOURCES:.cpp=.o) \ - $(ASM_SOURCES:.S=.o) - - -### Include directories -INCLUDES = \ - -I. \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/auxiliary \ - -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/include \ - $(DRIVER_INCLUDES) - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.cpp.o: - $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - - -##### TARGETS ##### - -default: depend symlinks $(LIBNAME) - - -$(LIBNAME): $(OBJECTS) Makefile $(TOP)/src/gallium/winsys/common/Makefile.template - $(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS) - - -depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ - $(ASM_SOURCES) 2> /dev/null - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean:: - -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) - -rm -f depend depend.bak - - -include depend diff --git a/src/gallium/winsys/common/intel_drm/Makefile b/src/gallium/winsys/common/intel_drm/Makefile deleted file mode 100644 index 913dbeff20..0000000000 --- a/src/gallium/winsys/common/intel_drm/Makefile +++ /dev/null @@ -1,23 +0,0 @@ -TOP = ../../../../.. -include $(TOP)/configs/current - -LIBNAME = inteldrm - -C_SOURCES = \ - intel_be_batchbuffer.c \ - intel_be_context.c \ - intel_be_device.c \ - ws_dri_bufmgr.c \ - ws_dri_drmpool.c \ - ws_dri_fencemgr.c \ - ws_dri_mallocpool.c \ - ws_dri_slabpool.c - - -include ../Makefile.template - -DRIVER_DEFINES = $(shell pkg-config libdrm --cflags \ - && pkg-config libdrm --atleast-version=2.3.1 \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") -symlinks: - diff --git a/src/gallium/winsys/common/intel_drm/glthread.h b/src/gallium/winsys/common/intel_drm/glthread.h deleted file mode 100644 index b8e9d5f59b..0000000000 --- a/src/gallium/winsys/common/intel_drm/glthread.h +++ /dev/null @@ -1,359 +0,0 @@ -/* - * Mesa 3-D graphics library - * Version: 6.5.2 - * - * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included - * in all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN - * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN - * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - */ - - -/* - * Thread support for gl dispatch. - * - * Initial version by John Stone (j.stone@acm.org) (johns@cs.umr.edu) - * and Christoph Poliwoda (poliwoda@volumegraphics.com) - * Revised by Keith Whitwell - * Adapted for new gl dispatcher by Brian Paul - * - * - * - * DOCUMENTATION - * - * This thread module exports the following types: - * _glthread_TSD Thread-specific data area - * _glthread_Thread Thread datatype - * _glthread_Mutex Mutual exclusion lock - * - * Macros: - * _glthread_DECLARE_STATIC_MUTEX(name) Declare a non-local mutex - * _glthread_INIT_MUTEX(name) Initialize a mutex - * _glthread_LOCK_MUTEX(name) Lock a mutex - * _glthread_UNLOCK_MUTEX(name) Unlock a mutex - * - * Functions: - * _glthread_GetID(v) Get integer thread ID - * _glthread_InitTSD() Initialize thread-specific data - * _glthread_GetTSD() Get thread-specific data - * _glthread_SetTSD() Set thread-specific data - * - */ - -/* - * If this file is accidentally included by a non-threaded build, - * it should not cause the build to fail, or otherwise cause problems. - * In general, it should only be included when needed however. - */ - -#ifndef GLTHREAD_H -#define GLTHREAD_H - - -#if defined(USE_MGL_NAMESPACE) -#define _glapi_Dispatch _mglapi_Dispatch -#endif - - - -#if (defined(PTHREADS) || defined(SOLARIS_THREADS) ||\ - defined(WIN32_THREADS) || defined(USE_XTHREADS) || defined(BEOS_THREADS)) \ - && !defined(THREADS) -# define THREADS -#endif - -#ifdef VMS -#include -#endif - -/* - * POSIX threads. This should be your choice in the Unix world - * whenever possible. When building with POSIX threads, be sure - * to enable any compiler flags which will cause the MT-safe - * libc (if one exists) to be used when linking, as well as any - * header macros for MT-safe errno, etc. For Solaris, this is the -mt - * compiler flag. On Solaris with gcc, use -D_REENTRANT to enable - * proper compiling for MT-safe libc etc. - */ -#if defined(PTHREADS) -#include /* POSIX threads headers */ - -typedef struct { - pthread_key_t key; - int initMagic; -} _glthread_TSD; - -typedef pthread_t _glthread_Thread; - -typedef pthread_mutex_t _glthread_Mutex; - -#define _glthread_DECLARE_STATIC_MUTEX(name) \ - static _glthread_Mutex name = PTHREAD_MUTEX_INITIALIZER - -#define _glthread_INIT_MUTEX(name) \ - pthread_mutex_init(&(name), NULL) - -#define _glthread_DESTROY_MUTEX(name) \ - pthread_mutex_destroy(&(name)) - -#define _glthread_LOCK_MUTEX(name) \ - (void) pthread_mutex_lock(&(name)) - -#define _glthread_UNLOCK_MUTEX(name) \ - (void) pthread_mutex_unlock(&(name)) - -typedef pthread_cond_t _glthread_Cond; - -#define _glthread_DECLARE_STATIC_COND(name) \ - static _glthread_Cond name = PTHREAD_COND_INITIALIZER - -#define _glthread_INIT_COND(cond) \ - pthread_cond_init(&(cond), NULL) - -#define _glthread_DESTROY_COND(name) \ - pthread_cond_destroy(&(name)) - -#define _glthread_COND_WAIT(cond, mutex) \ - pthread_cond_wait(&(cond), &(mutex)) - -#define _glthread_COND_SIGNAL(cond) \ - pthread_cond_signal(&(cond)) - -#define _glthread_COND_BROADCAST(cond) \ - pthread_cond_broadcast(&(cond)) - - -#else /* PTHREADS */ - -typedef unsigned int _glthread_Cond; -#define _glthread_DECLARE_STATIC_COND(name) \ -// #warning Condition variables not implemented. - -#define _glthread_INIT_COND(cond) \ - abort(); - -#define _glthread_DESTROY_COND(name) \ - abort(); - -#define _glthread_COND_WAIT(cond, mutex) \ - abort(); - -#define _glthread_COND_SIGNAL(cond) \ - abort(); - -#define _glthread_COND_BROADCAST(cond) \ - abort(); - -#endif - - -/* - * Solaris threads. Use only up to Solaris 2.4. - * Solaris 2.5 and higher provide POSIX threads. - * Be sure to compile with -mt on the Solaris compilers, or - * use -D_REENTRANT if using gcc. - */ -#ifdef SOLARIS_THREADS -#include - -typedef struct { - thread_key_t key; - mutex_t keylock; - int initMagic; -} _glthread_TSD; - -typedef thread_t _glthread_Thread; - -typedef mutex_t _glthread_Mutex; - -/* XXX need to really implement mutex-related macros */ -#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 -#define _glthread_INIT_MUTEX(name) (void) name -#define _glthread_DESTROY_MUTEX(name) (void) name -#define _glthread_LOCK_MUTEX(name) (void) name -#define _glthread_UNLOCK_MUTEX(name) (void) name - -#endif /* SOLARIS_THREADS */ - - - - -/* - * Windows threads. Should work with Windows NT and 95. - * IMPORTANT: Link with multithreaded runtime library when THREADS are - * used! - */ -#ifdef WIN32_THREADS -#include - -typedef struct { - DWORD key; - int initMagic; -} _glthread_TSD; - -typedef HANDLE _glthread_Thread; - -typedef CRITICAL_SECTION _glthread_Mutex; - -#define _glthread_DECLARE_STATIC_MUTEX(name) /*static*/ _glthread_Mutex name = {0,0,0,0,0,0} -#define _glthread_INIT_MUTEX(name) InitializeCriticalSection(&name) -#define _glthread_DESTROY_MUTEX(name) DeleteCriticalSection(&name) -#define _glthread_LOCK_MUTEX(name) EnterCriticalSection(&name) -#define _glthread_UNLOCK_MUTEX(name) LeaveCriticalSection(&name) - -#endif /* WIN32_THREADS */ - - - - -/* - * XFree86 has its own thread wrapper, Xthreads.h - * We wrap it again for GL. - */ -#ifdef USE_XTHREADS -#include - -typedef struct { - xthread_key_t key; - int initMagic; -} _glthread_TSD; - -typedef xthread_t _glthread_Thread; - -typedef xmutex_rec _glthread_Mutex; - -#ifdef XMUTEX_INITIALIZER -#define _glthread_DECLARE_STATIC_MUTEX(name) \ - static _glthread_Mutex name = XMUTEX_INITIALIZER -#else -#define _glthread_DECLARE_STATIC_MUTEX(name) \ - static _glthread_Mutex name -#endif - -#define _glthread_INIT_MUTEX(name) \ - xmutex_init(&(name)) - -#define _glthread_DESTROY_MUTEX(name) \ - xmutex_clear(&(name)) - -#define _glthread_LOCK_MUTEX(name) \ - (void) xmutex_lock(&(name)) - -#define _glthread_UNLOCK_MUTEX(name) \ - (void) xmutex_unlock(&(name)) - -#endif /* USE_XTHREADS */ - - - -/* - * BeOS threads. R5.x required. - */ -#ifdef BEOS_THREADS - -#include -#include - -typedef struct { - int32 key; - int initMagic; -} _glthread_TSD; - -typedef thread_id _glthread_Thread; - -/* Use Benaphore, aka speeder semaphore */ -typedef struct { - int32 lock; - sem_id sem; -} benaphore; -typedef benaphore _glthread_Mutex; - -#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = { 0, 0 } -#define _glthread_INIT_MUTEX(name) name.sem = create_sem(0, #name"_benaphore"), name.lock = 0 -#define _glthread_DESTROY_MUTEX(name) delete_sem(name.sem), name.lock = 0 -#define _glthread_LOCK_MUTEX(name) if (name.sem == 0) _glthread_INIT_MUTEX(name); \ - if (atomic_add(&(name.lock), 1) >= 1) acquire_sem(name.sem) -#define _glthread_UNLOCK_MUTEX(name) if (atomic_add(&(name.lock), -1) > 1) release_sem(name.sem) - -#endif /* BEOS_THREADS */ - - - -#ifndef THREADS - -/* - * THREADS not defined - */ - -typedef GLuint _glthread_TSD; - -typedef GLuint _glthread_Thread; - -typedef GLuint _glthread_Mutex; - -#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 - -#define _glthread_INIT_MUTEX(name) (void) name - -#define _glthread_DESTROY_MUTEX(name) (void) name - -#define _glthread_LOCK_MUTEX(name) (void) name - -#define _glthread_UNLOCK_MUTEX(name) (void) name - -#endif /* THREADS */ - - - -/* - * Platform independent thread specific data API. - */ - -extern unsigned long -_glthread_GetID(void); - - -extern void -_glthread_InitTSD(_glthread_TSD *); - - -extern void * -_glthread_GetTSD(_glthread_TSD *); - - -extern void -_glthread_SetTSD(_glthread_TSD *, void *); - -#if defined(GLX_USE_TLS) - -extern __thread struct _glapi_table * _glapi_tls_Dispatch - __attribute__((tls_model("initial-exec"))); - -#define GET_DISPATCH() _glapi_tls_Dispatch - -#elif !defined(GL_CALL) -# if defined(THREADS) -# define GET_DISPATCH() \ - ((__builtin_expect( _glapi_Dispatch != NULL, 1 )) \ - ? _glapi_Dispatch : _glapi_get_dispatch()) -# else -# define GET_DISPATCH() _glapi_Dispatch -# endif /* defined(THREADS) */ -#endif /* ndef GL_CALL */ - - -#endif /* THREADS_H */ diff --git a/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.c b/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.c deleted file mode 100644 index bc13a5761e..0000000000 --- a/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.c +++ /dev/null @@ -1,429 +0,0 @@ - -#include "intel_be_batchbuffer.h" -#include "intel_be_context.h" -#include "intel_be_device.h" -#include - -#include "xf86drm.h" - -static void -intel_realloc_relocs(struct intel_be_batchbuffer *batch, int num_relocs) -{ - unsigned long size = num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER; - - size *= sizeof(uint32_t); - batch->reloc = realloc(batch->reloc, size); - batch->reloc_size = num_relocs; -} - - -void -intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch) -{ - /* - * Get a new, free batchbuffer. - */ - drmBO *bo; - struct drm_bo_info_req *req; - - driBOUnrefUserList(batch->list); - driBOResetList(batch->list); - - /* base.size is the size available to the i915simple driver */ - batch->base.size = batch->device->max_batch_size - BATCH_RESERVED; - batch->base.actual_size = batch->device->max_batch_size; - driBOData(batch->buffer, batch->base.actual_size, NULL, NULL, 0); - - /* - * Add the batchbuffer to the validate list. - */ - - driBOAddListItem(batch->list, batch->buffer, - DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_EXE | DRM_BO_MASK_MEM, - &batch->dest_location, &batch->node); - - req = &batch->node->bo_arg.d.req.bo_req; - - /* - * Set up information needed for us to make relocations - * relative to the underlying drm buffer objects. - */ - - driReadLockKernelBO(); - bo = driBOKernel(batch->buffer); - req->presumed_offset = (uint64_t) bo->offset; - req->hint = DRM_BO_HINT_PRESUMED_OFFSET; - batch->drmBOVirtual = (uint8_t *) bo->virtual; - driReadUnlockKernelBO(); - - /* - * Adjust the relocation buffer size. - */ - - if (batch->reloc_size > INTEL_MAX_RELOCS || - batch->reloc == NULL) - intel_realloc_relocs(batch, INTEL_DEFAULT_RELOCS); - - assert(batch->reloc != NULL); - batch->reloc[0] = 0; /* No relocs yet. */ - batch->reloc[1] = 1; /* Reloc type 1 */ - batch->reloc[2] = 0; /* Only a single relocation list. */ - batch->reloc[3] = 0; /* Only a single relocation list. */ - - batch->base.map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); - batch->poolOffset = driBOPoolOffset(batch->buffer); - batch->base.ptr = batch->base.map; - batch->dirty_state = ~0; - batch->nr_relocs = 0; - batch->flags = 0; - batch->id = 0;//batch->intel->intelScreen->batch_id++; -} - -/*====================================================================== - * Public functions - */ -struct intel_be_batchbuffer * -intel_be_batchbuffer_alloc(struct intel_be_context *intel) -{ - struct intel_be_batchbuffer *batch = calloc(sizeof(*batch), 1); - - batch->intel = intel; - batch->device = intel->device; - - driGenBuffers(intel->device->batchPool, "batchbuffer", 1, - &batch->buffer, 4096, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); - batch->last_fence = NULL; - batch->list = driBOCreateList(20); - batch->reloc = NULL; - intel_be_batchbuffer_reset(batch); - return batch; -} - -void -intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch) -{ - if (batch->last_fence) { - driFenceFinish(batch->last_fence, - DRM_FENCE_TYPE_EXE, FALSE); - driFenceUnReference(&batch->last_fence); - } - if (batch->base.map) { - driBOUnmap(batch->buffer); - batch->base.map = NULL; - } - driBOUnReference(batch->buffer); - driBOFreeList(batch->list); - if (batch->reloc) - free(batch->reloc); - batch->buffer = NULL; - free(batch); -} - -void -intel_be_offset_relocation(struct intel_be_batchbuffer *batch, - unsigned pre_add, - struct _DriBufferObject *driBO, - uint64_t val_flags, - uint64_t val_mask) -{ - int itemLoc; - struct _drmBONode *node; - uint32_t *reloc; - struct drm_bo_info_req *req; - - driBOAddListItem(batch->list, driBO, val_flags, val_mask, - &itemLoc, &node); - req = &node->bo_arg.d.req.bo_req; - - if (!(req->hint & DRM_BO_HINT_PRESUMED_OFFSET)) { - - /* - * Stop other threads from tampering with the underlying - * drmBO while we're reading its offset. - */ - - driReadLockKernelBO(); - req->presumed_offset = (uint64_t) driBOKernel(driBO)->offset; - driReadUnlockKernelBO(); - req->hint = DRM_BO_HINT_PRESUMED_OFFSET; - } - - pre_add += driBOPoolOffset(driBO); - - if (batch->nr_relocs == batch->reloc_size) - intel_realloc_relocs(batch, batch->reloc_size * 2); - - reloc = batch->reloc + - (I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE); - - reloc[0] = ((uint8_t *)batch->base.ptr - batch->drmBOVirtual); - i915_batchbuffer_dword(&batch->base, req->presumed_offset + pre_add); - reloc[1] = pre_add; - reloc[2] = itemLoc; - reloc[3] = batch->dest_location; - batch->nr_relocs++; -} - -static void -i915_drm_copy_reply(const struct drm_bo_info_rep * rep, drmBO * buf) -{ - buf->handle = rep->handle; - buf->flags = rep->flags; - buf->size = rep->size; - buf->offset = rep->offset; - buf->mapHandle = rep->arg_handle; - buf->proposedFlags = rep->proposed_flags; - buf->start = rep->buffer_start; - buf->fenceFlags = rep->fence_flags; - buf->replyFlags = rep->rep_flags; - buf->pageAlignment = rep->page_alignment; -} - -static int -i915_execbuf(struct intel_be_batchbuffer *batch, - unsigned int used, - boolean ignore_cliprects, - drmBOList *list, - struct drm_i915_execbuffer *ea) -{ -// struct intel_be_context *intel = batch->intel; - drmBONode *node; - drmMMListHead *l; - struct drm_i915_op_arg *arg, *first; - struct drm_bo_op_req *req; - struct drm_bo_info_rep *rep; - uint64_t *prevNext = NULL; - drmBO *buf; - int ret = 0; - uint32_t count = 0; - - first = NULL; - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - - arg = &node->bo_arg; - req = &arg->d.req; - - if (!first) - first = arg; - - if (prevNext) - *prevNext = (unsigned long)arg; - - prevNext = &arg->next; - req->bo_req.handle = node->buf->handle; - req->op = drm_bo_validate; - req->bo_req.flags = node->arg0; - req->bo_req.mask = node->arg1; - req->bo_req.hint |= 0; - count++; - } - - memset(ea, 0, sizeof(*ea)); - ea->num_buffers = count; - ea->batch.start = batch->poolOffset; - ea->batch.used = used; -#if 0 /* ZZZ JB: no cliprects used */ - ea->batch.cliprects = intel->pClipRects; - ea->batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; - ea->batch.DR1 = 0; - ea->batch.DR4 = 0;((((GLuint) intel->drawX) & 0xffff) | - (((GLuint) intel->drawY) << 16)); -#else - ea->batch.cliprects = NULL; - ea->batch.num_cliprects = 0; - ea->batch.DR1 = 0; - ea->batch.DR4 = 0; -#endif - ea->fence_arg.flags = DRM_I915_FENCE_FLAG_FLUSHED; - ea->ops_list = (unsigned long) first; - first->reloc_ptr = (unsigned long) batch->reloc; - batch->reloc[0] = batch->nr_relocs; - - //return -EFAULT; - do { - ret = drmCommandWriteRead(batch->device->fd, DRM_I915_EXECBUFFER, ea, - sizeof(*ea)); - } while (ret == -EAGAIN); - - if (ret != 0) - return ret; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - arg = &node->bo_arg; - rep = &arg->d.rep.bo_info; - - if (!arg->handled) { - return -EFAULT; - } - if (arg->d.rep.ret) - return arg->d.rep.ret; - - buf = node->buf; - i915_drm_copy_reply(rep, buf); - } - return 0; -} - -/* TODO: Push this whole function into bufmgr. - */ -static struct _DriFenceObject * -do_flush_locked(struct intel_be_batchbuffer *batch, - unsigned int used, - boolean ignore_cliprects, boolean allow_unlock) -{ - struct intel_be_context *intel = batch->intel; - struct _DriFenceObject *fo; - drmFence fence; - drmBOList *boList; - struct drm_i915_execbuffer ea; - int ret = 0; - - driBOValidateUserList(batch->list); - boList = driGetdrmBOList(batch->list); - -#if 0 /* ZZZ JB Allways run */ - if (!(intel->numClipRects == 0 && !ignore_cliprects)) { -#else - if (1) { -#endif - ret = i915_execbuf(batch, used, ignore_cliprects, boList, &ea); - } else { - driPutdrmBOList(batch->list); - fo = NULL; - goto out; - } - driPutdrmBOList(batch->list); - if (ret) - abort(); - - if (ea.fence_arg.error != 0) { - - /* - * The hardware has been idled by the kernel. - * Don't fence the driBOs. - */ - - if (batch->last_fence) - driFenceUnReference(&batch->last_fence); -#if 0 /* ZZZ JB: no _mesa_* funcs in gallium */ - _mesa_printf("fence error\n"); -#endif - batch->last_fence = NULL; - fo = NULL; - goto out; - } - - fence.handle = ea.fence_arg.handle; - fence.fence_class = ea.fence_arg.fence_class; - fence.type = ea.fence_arg.type; - fence.flags = ea.fence_arg.flags; - fence.signaled = ea.fence_arg.signaled; - - fo = driBOFenceUserList(batch->device->fenceMgr, batch->list, - "SuperFence", &fence); - - if (driFenceType(fo) & DRM_I915_FENCE_TYPE_RW) { - if (batch->last_fence) - driFenceUnReference(&batch->last_fence); - /* - * FIXME: Context last fence?? - */ - batch->last_fence = fo; - driFenceReference(fo); - } - out: -#if 0 /* ZZZ JB: fix this */ - intel->vtbl.lost_hardware(intel); -#else - (void)intel; -#endif - return fo; -} - - -struct _DriFenceObject * -intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch) -{ - struct intel_be_context *intel = batch->intel; - unsigned int used = batch->base.ptr - batch->base.map; - boolean was_locked = batch->intel->hardware_locked(intel); - struct _DriFenceObject *fence; - - if (used == 0) { - driFenceReference(batch->last_fence); - return batch->last_fence; - } - - /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a - * performance drain that we would like to avoid. - */ -#if 0 /* ZZZ JB: what should we do here? */ - if (used & 4) { - ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->base.ptr)[1] = 0; - ((int *) batch->base.ptr)[2] = MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); - ((int *) batch->base.ptr)[1] = MI_BATCH_BUFFER_END; - used += 8; - } -#else - if (used & 4) { - ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->base.ptr)[1] = 0; - ((int *) batch->base.ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; - used += 12; - } - else { - ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; - ((int *) batch->base.ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; - used += 8; - } -#endif - driBOUnmap(batch->buffer); - batch->base.ptr = NULL; - batch->base.map = NULL; - - /* TODO: Just pass the relocation list and dma buffer up to the - * kernel. - */ - if (!was_locked) - intel->hardware_lock(intel); - - fence = do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), - FALSE); - - if (!was_locked) - intel->hardware_unlock(intel); - - /* Reset the buffer: - */ - intel_be_batchbuffer_reset(batch); - return fence; -} - -void -intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch) -{ - struct _DriFenceObject *fence = intel_be_batchbuffer_flush(batch); - driFenceFinish(fence, driFenceType(fence), FALSE); - driFenceUnReference(&fence); -} - -#if 0 -void -intel_be_batchbuffer_data(struct intel_be_batchbuffer *batch, - const void *data, unsigned int bytes, unsigned int flags) -{ - assert((bytes & 3) == 0); - intel_batchbuffer_require_space(batch, bytes, flags); - memcpy(batch->base.ptr, data, bytes); - batch->base.ptr += bytes; -} -#endif diff --git a/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.h b/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.h deleted file mode 100644 index f150e3a674..0000000000 --- a/src/gallium/winsys/common/intel_drm/intel_be_batchbuffer.h +++ /dev/null @@ -1,69 +0,0 @@ - -#ifndef INTEL_BE_BATCHBUFFER_H -#define INTEL_BE_BATCHBUFFER_H - -#include "i915simple/i915_batch.h" - -#include "ws_dri_bufmgr.h" - -#define BATCH_RESERVED 16 - -#define INTEL_DEFAULT_RELOCS 100 -#define INTEL_MAX_RELOCS 400 - -#define INTEL_BATCH_NO_CLIPRECTS 0x1 -#define INTEL_BATCH_CLIPRECTS 0x2 - -struct intel_be_context; -struct intel_be_device; - -struct intel_be_batchbuffer -{ - struct i915_batchbuffer base; - - struct intel_be_context *intel; - struct intel_be_device *device; - - struct _DriBufferObject *buffer; - struct _DriFenceObject *last_fence; - uint32_t flags; - - struct _DriBufferList *list; - size_t list_count; - - uint32_t *reloc; - size_t reloc_size; - size_t nr_relocs; - - uint32_t dirty_state; - uint32_t id; - - uint32_t poolOffset; - uint8_t *drmBOVirtual; - struct _drmBONode *node; /* Validation list node for this buffer */ - int dest_location; /* Validation list sequence for this buffer */ -}; - -struct intel_be_batchbuffer * -intel_be_batchbuffer_alloc(struct intel_be_context *intel); - -void -intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch); - -void -intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch); - -struct _DriFenceObject * -intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch); - -void -intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch); - -void -intel_be_offset_relocation(struct intel_be_batchbuffer *batch, - unsigned pre_add, - struct _DriBufferObject *driBO, - uint64_t val_flags, - uint64_t val_mask); - -#endif diff --git a/src/gallium/winsys/common/intel_drm/intel_be_context.c b/src/gallium/winsys/common/intel_drm/intel_be_context.c deleted file mode 100644 index 1af39674f4..0000000000 --- a/src/gallium/winsys/common/intel_drm/intel_be_context.c +++ /dev/null @@ -1,107 +0,0 @@ - -/* - * Authors: Jakob Bornecrantz - */ - -#include "ws_dri_fencemgr.h" -#include "intel_be_device.h" -#include "intel_be_context.h" -#include "intel_be_batchbuffer.h" - -static INLINE struct intel_be_context * -intel_be_context(struct i915_winsys *sws) -{ - return (struct intel_be_context *)sws; -} - -/* Simple batchbuffer interface: - */ - -static struct i915_batchbuffer* -intel_i915_batch_get(struct i915_winsys *sws) -{ - struct intel_be_context *intel = intel_be_context(sws); - return &intel->batch->base; -} - -static void intel_i915_batch_reloc(struct i915_winsys *sws, - struct pipe_buffer *buf, - unsigned access_flags, - unsigned delta) -{ - struct intel_be_context *intel = intel_be_context(sws); - - unsigned flags = DRM_BO_FLAG_MEM_TT; - unsigned mask = DRM_BO_MASK_MEM; - - if (access_flags & I915_BUFFER_ACCESS_WRITE) { - flags |= DRM_BO_FLAG_WRITE; - mask |= DRM_BO_FLAG_WRITE; - } - - if (access_flags & I915_BUFFER_ACCESS_READ) { - flags |= DRM_BO_FLAG_READ; - mask |= DRM_BO_FLAG_READ; - } - - intel_be_offset_relocation(intel->batch, - delta, - dri_bo(buf), - flags, - mask); -} - -static void intel_i915_batch_flush(struct i915_winsys *sws, - struct pipe_fence_handle **fence) -{ - struct intel_be_context *intel = intel_be_context(sws); - - union { - struct _DriFenceObject *dri; - struct pipe_fence_handle *pipe; - } fu; - - if (fence) - assert(!*fence); - - fu.dri = intel_be_batchbuffer_flush(intel->batch); - - if (!fu.dri) { - assert(0); - *fence = NULL; - return; - } - - if (fu.dri) { - if (fence) - *fence = fu.pipe; - else - driFenceUnReference(&fu.dri); - } - -} - -boolean -intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device) -{ - assert(intel); - assert(device); - - intel->device = device; - - /* TODO move framebuffer createion to the driver */ - - intel->base.batch_get = intel_i915_batch_get; - intel->base.batch_reloc = intel_i915_batch_reloc; - intel->base.batch_flush = intel_i915_batch_flush; - - intel->batch = intel_be_batchbuffer_alloc(intel); - - return true; -} - -void -intel_be_destroy_context(struct intel_be_context *intel) -{ - intel_be_batchbuffer_free(intel->batch); -} diff --git a/src/gallium/winsys/common/intel_drm/intel_be_context.h b/src/gallium/winsys/common/intel_drm/intel_be_context.h deleted file mode 100644 index d5cbc93594..0000000000 --- a/src/gallium/winsys/common/intel_drm/intel_be_context.h +++ /dev/null @@ -1,40 +0,0 @@ -/* These need to be diffrent from the intel winsys */ -#ifndef INTEL_BE_CONTEXT_H -#define INTEL_BE_CONTEXT_H - -#include "i915simple/i915_winsys.h" - -struct intel_be_context -{ - /** Interface to i915simple driver */ - struct i915_winsys base; - - struct intel_be_device *device; - struct intel_be_batchbuffer *batch; - - /* - * Hardware lock functions. - * - * Needs to be filled in by the winsys. - */ - void (*hardware_lock)(struct intel_be_context *context); - void (*hardware_unlock)(struct intel_be_context *context); - boolean (*hardware_locked)(struct intel_be_context *context); -}; - -/** - * Intialize a allocated intel_be_context struct. - * - * Remember to set the hardware_* functions. - */ -boolean -intel_be_init_context(struct intel_be_context *intel, - struct intel_be_device *device); - -/** - * Destroy a intel_be_context. - * Does not free the struct that is up to the winsys. - */ -void -intel_be_destroy_context(struct intel_be_context *intel); -#endif diff --git a/src/gallium/winsys/common/intel_drm/intel_be_device.c b/src/gallium/winsys/common/intel_drm/intel_be_device.c deleted file mode 100644 index 8db0329615..0000000000 --- a/src/gallium/winsys/common/intel_drm/intel_be_device.c +++ /dev/null @@ -1,308 +0,0 @@ - - -/* - * Authors: Keith Whitwell - * Jakob Bornecrantz - */ - -#include "intel_be_device.h" -#include "ws_dri_bufmgr.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" - -#include "pipe/p_winsys.h" -#include "pipe/p_defines.h" -#include "pipe/p_state.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" - -#include "i915simple/i915_screen.h" - -/* Turn a pipe winsys into an intel/pipe winsys: - */ -static INLINE struct intel_be_device * -intel_be_device( struct pipe_winsys *winsys ) -{ - return (struct intel_be_device *)winsys; -} - - -/* - * Buffer functions. - * - * Most callbacks map direcly onto dri_bufmgr operations: - */ - -static void *intel_be_buffer_map(struct pipe_winsys *winsys, - struct pipe_buffer *buf, - unsigned flags ) -{ - unsigned drm_flags = 0; - - if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) - drm_flags |= DRM_BO_FLAG_WRITE; - - if (flags & PIPE_BUFFER_USAGE_CPU_READ) - drm_flags |= DRM_BO_FLAG_READ; - - return driBOMap( dri_bo(buf), drm_flags, 0 ); -} - -static void intel_be_buffer_unmap(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnmap( dri_bo(buf) ); -} - -static void -intel_be_buffer_destroy(struct pipe_winsys *winsys, - struct pipe_buffer *buf) -{ - driBOUnReference( dri_bo(buf) ); - FREE(buf); -} - -static struct pipe_buffer * -intel_be_buffer_create(struct pipe_winsys *winsys, - unsigned alignment, - unsigned usage, - unsigned size ) -{ - struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer ); - struct intel_be_device *iws = intel_be_device(winsys); - unsigned flags = 0; - struct _DriBufferPool *pool; - - buffer->base.refcount = 1; - buffer->base.alignment = alignment; - buffer->base.usage = usage; - buffer->base.size = size; - - if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { - flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; - pool = iws->mallocPool; - } else if (usage & PIPE_BUFFER_USAGE_CUSTOM) { - /* For vertex buffers */ - flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; - pool = iws->vertexPool; - } else { - flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; - pool = iws->regionPool; - } - - if (usage & PIPE_BUFFER_USAGE_GPU_READ) - flags |= DRM_BO_FLAG_READ; - - if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) - flags |= DRM_BO_FLAG_WRITE; - - /* drm complains if we don't set any read/write flags. - */ - if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) - flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; - - buffer->pool = pool; - driGenBuffers( buffer->pool, - "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); - - driBOData( buffer->driBO, size, NULL, buffer->pool, 0 ); - - return &buffer->base; -} - - -static struct pipe_buffer * -intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) -{ - struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer ); - struct intel_be_device *iws = intel_be_device(winsys); - - driGenUserBuffer( iws->regionPool, - "pipe user buffer", &buffer->driBO, ptr, bytes ); - - buffer->base.refcount = 1; - - return &buffer->base; -} - -struct pipe_buffer * -intel_be_buffer_from_handle(struct intel_be_device *device, - const char* name, unsigned handle) -{ - struct intel_be_buffer *be_buf = malloc(sizeof(*be_buf)); - struct pipe_buffer *buffer; - - if (!be_buf) - goto err; - - memset(be_buf, 0, sizeof(*be_buf)); - - driGenBuffers(device->staticPool, name, 1, &be_buf->driBO, 0, 0, 0); - driBOSetReferenced(be_buf->driBO, handle); - - if (0) /** XXX TODO check error */ - goto err_bo; - - buffer = &be_buf->base; - buffer->refcount = 1; - buffer->alignment = 0; - buffer->usage = 0; - buffer->size = driBOSize(be_buf->driBO); - - return buffer; -err_bo: - free(be_buf); -err: - return NULL; -} - - -/* - * Surface functions. - * - * Deprecated! - */ - -static struct pipe_surface * -intel_i915_surface_alloc(struct pipe_winsys *winsys) -{ - assert((size_t)"intel_i915_surface_alloc is deprecated" & 0); - return NULL; -} - -static int -intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, - struct pipe_surface *surf, - unsigned width, unsigned height, - enum pipe_format format, - unsigned flags, - unsigned tex_usage) -{ - assert((size_t)"intel_i915_surface_alloc_storage is deprecated" & 0); - return -1; -} - -static void -intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) -{ - assert((size_t)"intel_i915_surface_release is deprecated" & 0); -} - - -/* - * Fence functions - */ - -static void -intel_be_fence_reference( struct pipe_winsys *sws, - struct pipe_fence_handle **ptr, - struct pipe_fence_handle *fence ) -{ - if (*ptr) - driFenceUnReference((struct _DriFenceObject **)ptr); - - if (fence) - *ptr = (struct pipe_fence_handle *)driFenceReference((struct _DriFenceObject *)fence); -} - -static int -intel_be_fence_signalled( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ) -{ - return driFenceSignaled((struct _DriFenceObject *)fence, flag); -} - -static int -intel_be_fence_finish( struct pipe_winsys *sws, - struct pipe_fence_handle *fence, - unsigned flag ) -{ - return driFenceFinish((struct _DriFenceObject *)fence, flag, 0); -} - - -/* - * Misc functions - */ - -boolean -intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) -{ - dev->fd = fd; - dev->max_batch_size = 16 * 4096; - dev->max_vertex_size = 128 * 4096; - - dev->base.buffer_create = intel_be_buffer_create; - dev->base.user_buffer_create = intel_be_user_buffer_create; - dev->base.buffer_map = intel_be_buffer_map; - dev->base.buffer_unmap = intel_be_buffer_unmap; - dev->base.buffer_destroy = intel_be_buffer_destroy; - dev->base.surface_alloc = intel_i915_surface_alloc; - dev->base.surface_alloc_storage = intel_i915_surface_alloc_storage; - dev->base.surface_release = intel_i915_surface_release; - dev->base.fence_reference = intel_be_fence_reference; - dev->base.fence_signalled = intel_be_fence_signalled; - dev->base.fence_finish = intel_be_fence_finish; - -#if 0 /* Set by the winsys */ - dev->base.flush_frontbuffer = intel_flush_frontbuffer; - dev->base.get_name = intel_get_name; -#endif - - dev->fMan = driInitFreeSlabManager(10, 10); - dev->fenceMgr = driFenceMgrTTMInit(dev->fd); - - dev->mallocPool = driMallocPoolInit(); - dev->staticPool = driDRMPoolInit(dev->fd); - /* Sizes: 64 128 256 512 1024 2048 4096 8192 16384 32768 */ - dev->regionPool = driSlabPoolInit(dev->fd, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - 64, - 10, 120, 4096 * 64, 0, - dev->fMan); - - dev->vertexPool = driSlabPoolInit(dev->fd, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT, - dev->max_vertex_size, - 1, 120, dev->max_vertex_size * 4, 0, - dev->fMan); - - dev->batchPool = driSlabPoolInit(dev->fd, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - DRM_BO_FLAG_EXE | - DRM_BO_FLAG_MEM_TT, - dev->max_batch_size, - 1, 40, dev->max_batch_size * 16, 0, - dev->fMan); - - /* Fill in this struct with callbacks that i915simple will need to - * communicate with the window system, buffer manager, etc. - */ - dev->screen = i915_create_screen(&dev->base, id); - - return true; -} - -void -intel_be_destroy_device(struct intel_be_device *dev) -{ - driPoolTakeDown(dev->mallocPool); - driPoolTakeDown(dev->staticPool); - driPoolTakeDown(dev->regionPool); - driPoolTakeDown(dev->vertexPool); - driPoolTakeDown(dev->batchPool); - - /** TODO takedown fenceMgr and fMan */ -} diff --git a/src/gallium/winsys/common/intel_drm/intel_be_device.h b/src/gallium/winsys/common/intel_drm/intel_be_device.h deleted file mode 100644 index 3f8b3f585c..0000000000 --- a/src/gallium/winsys/common/intel_drm/intel_be_device.h +++ /dev/null @@ -1,72 +0,0 @@ -#ifndef INTEL_DRM_DEVICE_H -#define INTEL_DRM_DEVICE_H - -#include "pipe/p_winsys.h" -#include "pipe/p_context.h" - -/* - * Device - */ - -struct intel_be_device -{ - struct pipe_winsys base; - - /** - * Hw level screen - */ - struct pipe_screen *screen; - - int fd; /**< Drm file discriptor */ - - size_t max_batch_size; - size_t max_vertex_size; - - struct _DriFenceMgr *fenceMgr; - - struct _DriBufferPool *batchPool; - struct _DriBufferPool *regionPool; - struct _DriBufferPool *mallocPool; - struct _DriBufferPool *vertexPool; - struct _DriBufferPool *staticPool; - struct _DriFreeSlabManager *fMan; -}; - -boolean -intel_be_init_device(struct intel_be_device *device, int fd, unsigned id); - -void -intel_be_destroy_device(struct intel_be_device *dev); - -/* - * Buffer - */ - -struct intel_be_buffer { - struct pipe_buffer base; - struct _DriBufferPool *pool; - struct _DriBufferObject *driBO; -}; - -/** - * Create a be buffer from a drm bo handle - * - * Takes a reference - */ -struct pipe_buffer * -intel_be_buffer_from_handle(struct intel_be_device *device, - const char* name, unsigned handle); - -static INLINE struct intel_be_buffer * -intel_be_buffer(struct pipe_buffer *buf) -{ - return (struct intel_be_buffer *)buf; -} - -static INLINE struct _DriBufferObject * -dri_bo(struct pipe_buffer *buf) -{ - return intel_be_buffer(buf)->driBO; -} - -#endif diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.c b/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.c deleted file mode 100644 index b6d901f85e..0000000000 --- a/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.c +++ /dev/null @@ -1,949 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - * Keith Whitwell - */ - -#include -#include -#include -#include "glthread.h" -#include "errno.h" -#include "ws_dri_bufmgr.h" -#include "string.h" -#include "pipe/p_debug.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" - - -/* - * This lock is here to protect drmBO structs changing underneath us during a - * validate list call, since validatelist cannot take individiual locks for - * each drmBO. Validatelist takes this lock in write mode. Any access to an - * individual drmBO should take this lock in read mode, since in that case, the - * driBufferObject mutex will protect the access. Locking order is - * driBufferObject mutex - > this rw lock. - */ - -_glthread_DECLARE_STATIC_MUTEX(bmMutex); -_glthread_DECLARE_STATIC_COND(bmCond); - -static int kernelReaders = 0; -static int num_buffers = 0; -static int num_user_buffers = 0; - -static drmBO *drmBOListBuf(void *iterator) -{ - drmBONode *node; - drmMMListHead *l = (drmMMListHead *) iterator; - node = DRMLISTENTRY(drmBONode, l, head); - return node->buf; -} - -static void *drmBOListIterator(drmBOList *list) -{ - void *ret = list->list.next; - - if (ret == &list->list) - return NULL; - return ret; -} - -static void *drmBOListNext(drmBOList *list, void *iterator) -{ - void *ret; - - drmMMListHead *l = (drmMMListHead *) iterator; - ret = l->next; - if (ret == &list->list) - return NULL; - return ret; -} - -static drmBONode *drmAddListItem(drmBOList *list, drmBO *item, - uint64_t arg0, - uint64_t arg1) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->free.next; - if (l == &list->free) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - return NULL; - } - list->numCurrent++; - } - else { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - } - node->buf = item; - node->arg0 = arg0; - node->arg1 = arg1; - DRMLISTADD(&node->head, &list->list); - list->numOnList++; - return node; -} - -static int drmAddValidateItem(drmBOList *list, drmBO *buf, uint64_t flags, - uint64_t mask, int *newItem) -{ - drmBONode *node, *cur; - drmMMListHead *l; - - *newItem = 0; - cur = NULL; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - if (node->buf == buf) { - cur = node; - break; - } - } - if (!cur) { - cur = drmAddListItem(list, buf, flags, mask); - if (!cur) { - return -ENOMEM; - } - *newItem = 1; - cur->arg0 = flags; - cur->arg1 = mask; - } - else { - uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; - uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; - - if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { - return -EINVAL; - } - - cur->arg1 |= mask; - cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); - - if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && - (cur->arg0 & DRM_BO_MASK_MEM) == 0) { - return -EINVAL; - } - } - return 0; -} - -static void drmBOFreeList(drmBOList *list) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->list.next; - while(l != &list->list) { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - l = list->list.next; - list->numCurrent--; - list->numOnList--; - } - - l = list->free.next; - while(l != &list->free) { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - l = list->free.next; - list->numCurrent--; - } -} - -static int drmAdjustListNodes(drmBOList *list) -{ - drmBONode *node; - drmMMListHead *l; - int ret = 0; - - while(list->numCurrent < list->numTarget) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - ret = -ENOMEM; - break; - } - list->numCurrent++; - DRMLISTADD(&node->head, &list->free); - } - - while(list->numCurrent > list->numTarget) { - l = list->free.next; - if (l == &list->free) - break; - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - free(node); - list->numCurrent--; - } - return ret; -} - -static int drmBOCreateList(int numTarget, drmBOList *list) -{ - DRMINITLISTHEAD(&list->list); - DRMINITLISTHEAD(&list->free); - list->numTarget = numTarget; - list->numCurrent = 0; - list->numOnList = 0; - return drmAdjustListNodes(list); -} - -static int drmBOResetList(drmBOList *list) -{ - drmMMListHead *l; - int ret; - - ret = drmAdjustListNodes(list); - if (ret) - return ret; - - l = list->list.next; - while (l != &list->list) { - DRMLISTDEL(l); - DRMLISTADD(l, &list->free); - list->numOnList--; - l = list->list.next; - } - return drmAdjustListNodes(list); -} - -void driWriteLockKernelBO(void) -{ - _glthread_LOCK_MUTEX(bmMutex); - while(kernelReaders != 0) - _glthread_COND_WAIT(bmCond, bmMutex); -} - -void driWriteUnlockKernelBO(void) -{ - _glthread_UNLOCK_MUTEX(bmMutex); -} - -void driReadLockKernelBO(void) -{ - _glthread_LOCK_MUTEX(bmMutex); - kernelReaders++; - _glthread_UNLOCK_MUTEX(bmMutex); -} - -void driReadUnlockKernelBO(void) -{ - _glthread_LOCK_MUTEX(bmMutex); - if (--kernelReaders == 0) - _glthread_COND_BROADCAST(bmCond); - _glthread_UNLOCK_MUTEX(bmMutex); -} - - - - -/* - * TODO: Introduce fence pools in the same way as - * buffer object pools. - */ - -typedef struct _DriBufferObject -{ - DriBufferPool *pool; - _glthread_Mutex mutex; - int refCount; - const char *name; - uint64_t flags; - unsigned hint; - unsigned alignment; - unsigned createdByReference; - void *private; - /* user-space buffer: */ - unsigned userBuffer; - void *userData; - unsigned userSize; -} DriBufferObject; - -typedef struct _DriBufferList { - drmBOList drmBuffers; /* List of kernel buffers needing validation */ - drmBOList driBuffers; /* List of user-space buffers needing validation */ -} DriBufferList; - - -void -bmError(int val, const char *file, const char *function, int line) -{ - printf("Fatal video memory manager error \"%s\".\n" - "Check kernel logs or set the LIBGL_DEBUG\n" - "environment variable to \"verbose\" for more info.\n" - "Detected in file %s, line %d, function %s.\n", - strerror(-val), file, line, function); -#ifndef NDEBUG - abort(); -#else - abort(); -#endif -} - -extern drmBO * -driBOKernel(struct _DriBufferObject *buf) -{ - drmBO *ret; - - driReadLockKernelBO(); - _glthread_LOCK_MUTEX(buf->mutex); - assert(buf->private != NULL); - ret = buf->pool->kernel(buf->pool, buf->private); - if (!ret) - BM_CKFATAL(-EINVAL); - _glthread_UNLOCK_MUTEX(buf->mutex); - driReadUnlockKernelBO(); - - return ret; -} - -void -driBOWaitIdle(struct _DriBufferObject *buf, int lazy) -{ - - /* - * This function may block. Is it sane to keep the mutex held during - * that time?? - */ - - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy)); - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -void * -driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint) -{ - void *virtual; - int retval; - - if (buf->userBuffer) { - return buf->userData; - } - - _glthread_LOCK_MUTEX(buf->mutex); - assert(buf->private != NULL); - retval = buf->pool->map(buf->pool, buf->private, flags, hint, - &buf->mutex, &virtual); - _glthread_UNLOCK_MUTEX(buf->mutex); - - return retval == 0 ? virtual : NULL; -} - -void -driBOUnmap(struct _DriBufferObject *buf) -{ - if (buf->userBuffer) - return; - - assert(buf->private != NULL); - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -unsigned long -driBOOffset(struct _DriBufferObject *buf) -{ - unsigned long ret; - - assert(buf->private != NULL); - - _glthread_LOCK_MUTEX(buf->mutex); - ret = buf->pool->offset(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - return ret; -} - -unsigned long -driBOPoolOffset(struct _DriBufferObject *buf) -{ - unsigned long ret; - - assert(buf->private != NULL); - - _glthread_LOCK_MUTEX(buf->mutex); - ret = buf->pool->poolOffset(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - return ret; -} - -uint64_t -driBOFlags(struct _DriBufferObject *buf) -{ - uint64_t ret; - - assert(buf->private != NULL); - - driReadLockKernelBO(); - _glthread_LOCK_MUTEX(buf->mutex); - ret = buf->pool->flags(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - driReadUnlockKernelBO(); - return ret; -} - -struct _DriBufferObject * -driBOReference(struct _DriBufferObject *buf) -{ - _glthread_LOCK_MUTEX(buf->mutex); - if (++buf->refCount == 1) { - _glthread_UNLOCK_MUTEX(buf->mutex); - BM_CKFATAL(-EINVAL); - } - _glthread_UNLOCK_MUTEX(buf->mutex); - return buf; -} - -void -driBOUnReference(struct _DriBufferObject *buf) -{ - int tmp; - - if (!buf) - return; - - _glthread_LOCK_MUTEX(buf->mutex); - tmp = --buf->refCount; - if (!tmp) { - _glthread_UNLOCK_MUTEX(buf->mutex); - if (buf->private) { - if (buf->createdByReference) - buf->pool->unreference(buf->pool, buf->private); - else - buf->pool->destroy(buf->pool, buf->private); - } - if (buf->userBuffer) - num_user_buffers--; - else - num_buffers--; - free(buf); - } else - _glthread_UNLOCK_MUTEX(buf->mutex); - -} - - -int -driBOData(struct _DriBufferObject *buf, - unsigned size, const void *data, - DriBufferPool *newPool, - uint64_t flags) -{ - void *virtual = NULL; - int newBuffer; - int retval = 0; - struct _DriBufferPool *pool; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - _glthread_LOCK_MUTEX(buf->mutex); - pool = buf->pool; - - if (pool == NULL && newPool != NULL) { - buf->pool = newPool; - pool = newPool; - } - if (newPool == NULL) - newPool = pool; - - if (!pool->create) { - assert((size_t)"driBOData called on invalid buffer\n" & 0); - BM_CKFATAL(-EINVAL); - } - - newBuffer = (!buf->private || pool != newPool || - pool->size(pool, buf->private) < size); - - if (!flags) - flags = buf->flags; - - if (newBuffer) { - - if (buf->createdByReference) { - assert((size_t)"driBOData requiring resizing called on shared buffer.\n" & 0); - BM_CKFATAL(-EINVAL); - } - - if (buf->private) - buf->pool->destroy(buf->pool, buf->private); - - pool = newPool; - buf->pool = newPool; - buf->private = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, - buf->alignment); - if (!buf->private) - retval = -ENOMEM; - - if (retval == 0) - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, - DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual); - } else if (pool->map(pool, buf->private, DRM_BO_FLAG_WRITE, - DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual)) { - /* - * Buffer is busy. need to create a new one. - */ - - void *newBuf; - - newBuf = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, - buf->alignment); - if (newBuf) { - buf->pool->destroy(buf->pool, buf->private); - buf->private = newBuf; - } - - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); - } else { - uint64_t flag_diff = flags ^ buf->flags; - - /* - * We might need to change buffer flags. - */ - - if (flag_diff){ - assert(pool->setStatus != NULL); - BM_CKFATAL(pool->unmap(pool, buf->private)); - BM_CKFATAL(pool->setStatus(pool, buf->private, flag_diff, - buf->flags)); - if (!data) - goto out; - - retval = pool->map(pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); - } - } - - if (retval == 0) { - if (data) - memcpy(virtual, data, size); - - BM_CKFATAL(pool->unmap(pool, buf->private)); - } - - out: - _glthread_UNLOCK_MUTEX(buf->mutex); - - return retval; -} - -void -driBOSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, const void *data) -{ - void *virtual; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - _glthread_LOCK_MUTEX(buf->mutex); - if (size && data) { - BM_CKFATAL(buf->pool->map(buf->pool, buf->private, - DRM_BO_FLAG_WRITE, 0, &buf->mutex, - &virtual)); - memcpy((unsigned char *) virtual + offset, data, size); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - } - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -void -driBOGetSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, void *data) -{ - void *virtual; - - assert(!buf->userBuffer); /* XXX just do a memcpy? */ - - _glthread_LOCK_MUTEX(buf->mutex); - if (size && data) { - BM_CKFATAL(buf->pool->map(buf->pool, buf->private, - DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual)); - memcpy(data, (unsigned char *) virtual + offset, size); - BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); - } - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -void -driBOSetReferenced(struct _DriBufferObject *buf, - unsigned long handle) -{ - _glthread_LOCK_MUTEX(buf->mutex); - if (buf->private != NULL) { - assert((size_t)"Invalid buffer for setReferenced\n" & 0); - BM_CKFATAL(-EINVAL); - - } - if (buf->pool->reference == NULL) { - assert((size_t)"Invalid buffer pool for setReferenced\n" & 0); - BM_CKFATAL(-EINVAL); - } - buf->private = buf->pool->reference(buf->pool, handle); - if (!buf->private) { - assert((size_t)"Invalid buffer pool for setStatic\n" & 0); - BM_CKFATAL(-ENOMEM); - } - buf->createdByReference = TRUE; - buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags; - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -int -driGenBuffers(struct _DriBufferPool *pool, - const char *name, - unsigned n, - struct _DriBufferObject *buffers[], - unsigned alignment, uint64_t flags, unsigned hint) -{ - struct _DriBufferObject *buf; - int i; - - flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM | - DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; - - ++num_buffers; - - assert(pool); - - for (i = 0; i < n; ++i) { - buf = (struct _DriBufferObject *) calloc(1, sizeof(*buf)); - if (!buf) - return -ENOMEM; - - _glthread_INIT_MUTEX(buf->mutex); - _glthread_LOCK_MUTEX(buf->mutex); - buf->refCount = 1; - buf->flags = flags; - buf->hint = hint; - buf->name = name; - buf->alignment = alignment; - buf->pool = pool; - buf->createdByReference = 0; - _glthread_UNLOCK_MUTEX(buf->mutex); - buffers[i] = buf; - } - return 0; -} - -void -driGenUserBuffer(struct _DriBufferPool *pool, - const char *name, - struct _DriBufferObject **buffers, - void *ptr, unsigned bytes) -{ - const unsigned alignment = 1, flags = 0, hint = 0; - - --num_buffers; /* JB: is inced in GenBuffes */ - driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); - ++num_user_buffers; - - (*buffers)->userBuffer = 1; - (*buffers)->userData = ptr; - (*buffers)->userSize = bytes; -} - -void -driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]) -{ - int i; - - for (i = 0; i < n; ++i) { - driBOUnReference(buffers[i]); - } -} - - -void -driInitBufMgr(int fd) -{ - ; -} - -/* - * Note that lists are per-context and don't need mutex protection. - */ - -struct _DriBufferList * -driBOCreateList(int target) -{ - struct _DriBufferList *list = calloc(sizeof(*list), 1); - - BM_CKFATAL(drmBOCreateList(target, &list->drmBuffers)); - BM_CKFATAL(drmBOCreateList(target, &list->driBuffers)); - return list; -} - -int -driBOResetList(struct _DriBufferList * list) -{ - int ret; - ret = drmBOResetList(&list->drmBuffers); - if (ret) - return ret; - ret = drmBOResetList(&list->driBuffers); - return ret; -} - -void -driBOFreeList(struct _DriBufferList * list) -{ - drmBOFreeList(&list->drmBuffers); - drmBOFreeList(&list->driBuffers); - free(list); -} - - -/* - * Copied from libdrm, because it is needed by driAddValidateItem. - */ - -static drmBONode * -driAddListItem(drmBOList * list, drmBO * item, - uint64_t arg0, uint64_t arg1) -{ - drmBONode *node; - drmMMListHead *l; - - l = list->free.next; - if (l == &list->free) { - node = (drmBONode *) malloc(sizeof(*node)); - if (!node) { - return NULL; - } - list->numCurrent++; - } else { - DRMLISTDEL(l); - node = DRMLISTENTRY(drmBONode, l, head); - } - memset(&node->bo_arg, 0, sizeof(node->bo_arg)); - node->buf = item; - node->arg0 = arg0; - node->arg1 = arg1; - DRMLISTADDTAIL(&node->head, &list->list); - list->numOnList++; - return node; -} - -/* - * Slightly modified version compared to the libdrm version. - * This one returns the list index of the buffer put on the list. - */ - -static int -driAddValidateItem(drmBOList * list, drmBO * buf, uint64_t flags, - uint64_t mask, int *itemLoc, - struct _drmBONode **pnode) -{ - drmBONode *node, *cur; - drmMMListHead *l; - int count = 0; - - cur = NULL; - - for (l = list->list.next; l != &list->list; l = l->next) { - node = DRMLISTENTRY(drmBONode, l, head); - if (node->buf == buf) { - cur = node; - break; - } - count++; - } - if (!cur) { - cur = driAddListItem(list, buf, flags, mask); - if (!cur) - return -ENOMEM; - - cur->arg0 = flags; - cur->arg1 = mask; - } else { - uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; - uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; - - if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { - return -EINVAL; - } - - cur->arg1 |= mask; - cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); - - if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && - (cur->arg0 & DRM_BO_MASK_MEM) == 0) { - return -EINVAL; - } - } - *itemLoc = count; - *pnode = cur; - return 0; -} - - -void -driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf, - uint64_t flags, uint64_t mask, int *itemLoc, - struct _drmBONode **node) -{ - int newItem; - - _glthread_LOCK_MUTEX(buf->mutex); - BM_CKFATAL(driAddValidateItem(&list->drmBuffers, - buf->pool->kernel(buf->pool, buf->private), - flags, mask, itemLoc, node)); - BM_CKFATAL(drmAddValidateItem(&list->driBuffers, (drmBO *) buf, - flags, mask, &newItem)); - if (newItem) - buf->refCount++; - - _glthread_UNLOCK_MUTEX(buf->mutex); -} - -drmBOList *driGetdrmBOList(struct _DriBufferList *list) -{ - driWriteLockKernelBO(); - return &list->drmBuffers; -} - -void driPutdrmBOList(struct _DriBufferList *list) -{ - driWriteUnlockKernelBO(); -} - - -void -driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence) -{ - _glthread_LOCK_MUTEX(buf->mutex); - if (buf->pool->fence) - BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence)); - _glthread_UNLOCK_MUTEX(buf->mutex); - -} - -void -driBOUnrefUserList(struct _DriBufferList *list) -{ - struct _DriBufferObject *buf; - void *curBuf; - - curBuf = drmBOListIterator(&list->driBuffers); - while (curBuf) { - buf = (struct _DriBufferObject *)drmBOListBuf(curBuf); - driBOUnReference(buf); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } -} - -struct _DriFenceObject * -driBOFenceUserList(struct _DriFenceMgr *mgr, - struct _DriBufferList *list, const char *name, - drmFence *kFence) -{ - struct _DriFenceObject *fence; - struct _DriBufferObject *buf; - void *curBuf; - - fence = driFenceCreate(mgr, kFence->fence_class, kFence->type, - kFence, sizeof(*kFence)); - curBuf = drmBOListIterator(&list->driBuffers); - - /* - * User-space fencing callbacks. - */ - - while (curBuf) { - buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); - driBOFence(buf, fence); - driBOUnReference(buf); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } - - driBOResetList(list); - return fence; -} - -void -driBOValidateUserList(struct _DriBufferList * list) -{ - void *curBuf; - struct _DriBufferObject *buf; - - curBuf = drmBOListIterator(&list->driBuffers); - - /* - * User-space validation callbacks. - */ - - while (curBuf) { - buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); - _glthread_LOCK_MUTEX(buf->mutex); - if (buf->pool->validate) - BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex)); - _glthread_UNLOCK_MUTEX(buf->mutex); - curBuf = drmBOListNext(&list->driBuffers, curBuf); - } -} - - -void -driPoolTakeDown(struct _DriBufferPool *pool) -{ - pool->takeDown(pool); - -} - -unsigned long -driBOSize(struct _DriBufferObject *buf) -{ - unsigned long size; - - _glthread_LOCK_MUTEX(buf->mutex); - size = buf->pool->size(buf->pool, buf->private); - _glthread_UNLOCK_MUTEX(buf->mutex); - - return size; - -} - -drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list) -{ - return &list->drmBuffers; -} - -drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list) -{ - return &list->driBuffers; -} - diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.h b/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.h deleted file mode 100644 index e6c0cff0a0..0000000000 --- a/src/gallium/winsys/common/intel_drm/ws_dri_bufmgr.h +++ /dev/null @@ -1,138 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - * Keith Whitwell - */ - -#ifndef _PSB_BUFMGR_H_ -#define _PSB_BUFMGR_H_ -#include -#include "i915_drm.h" -#include "ws_dri_fencemgr.h" - -typedef struct _drmBONode -{ - drmMMListHead head; - drmBO *buf; - struct drm_i915_op_arg bo_arg; - uint64_t arg0; - uint64_t arg1; -} drmBONode; - -typedef struct _drmBOList { - unsigned numTarget; - unsigned numCurrent; - unsigned numOnList; - drmMMListHead list; - drmMMListHead free; -} drmBOList; - - -struct _DriFenceObject; -struct _DriBufferObject; -struct _DriBufferPool; -struct _DriBufferList; - -/* - * Return a pointer to the libdrm buffer object this DriBufferObject - * uses. - */ - -extern drmBO *driBOKernel(struct _DriBufferObject *buf); -extern void *driBOMap(struct _DriBufferObject *buf, unsigned flags, - unsigned hint); -extern void driBOUnmap(struct _DriBufferObject *buf); -extern unsigned long driBOOffset(struct _DriBufferObject *buf); -extern unsigned long driBOPoolOffset(struct _DriBufferObject *buf); - -extern uint64_t driBOFlags(struct _DriBufferObject *buf); -extern struct _DriBufferObject *driBOReference(struct _DriBufferObject *buf); -extern void driBOUnReference(struct _DriBufferObject *buf); - -extern int driBOData(struct _DriBufferObject *r_buf, - unsigned size, const void *data, - struct _DriBufferPool *pool, uint64_t flags); - -extern void driBOSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, - const void *data); -extern void driBOGetSubData(struct _DriBufferObject *buf, - unsigned long offset, unsigned long size, - void *data); -extern int driGenBuffers(struct _DriBufferPool *pool, - const char *name, - unsigned n, - struct _DriBufferObject *buffers[], - unsigned alignment, uint64_t flags, unsigned hint); -extern void driGenUserBuffer(struct _DriBufferPool *pool, - const char *name, - struct _DriBufferObject *buffers[], - void *ptr, unsigned bytes); -extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]); -extern void driInitBufMgr(int fd); -extern struct _DriBufferList *driBOCreateList(int target); -extern int driBOResetList(struct _DriBufferList * list); -extern void driBOAddListItem(struct _DriBufferList * list, - struct _DriBufferObject *buf, - uint64_t flags, uint64_t mask, int *itemLoc, - struct _drmBONode **node); - -extern void driBOValidateList(int fd, struct _DriBufferList * list); -extern void driBOFreeList(struct _DriBufferList * list); -extern struct _DriFenceObject *driBOFenceUserList(struct _DriFenceMgr *mgr, - struct _DriBufferList *list, - const char *name, - drmFence *kFence); -extern void driBOUnrefUserList(struct _DriBufferList *list); -extern void driBOValidateUserList(struct _DriBufferList * list); -extern drmBOList *driGetdrmBOList(struct _DriBufferList *list); -extern void driPutdrmBOList(struct _DriBufferList *list); - -extern void driBOFence(struct _DriBufferObject *buf, - struct _DriFenceObject *fence); - -extern void driPoolTakeDown(struct _DriBufferPool *pool); -extern void driBOSetReferenced(struct _DriBufferObject *buf, - unsigned long handle); -unsigned long driBOSize(struct _DriBufferObject *buf); -extern void driBOWaitIdle(struct _DriBufferObject *buf, int lazy); -extern void driPoolTakeDown(struct _DriBufferPool *pool); - -extern void driReadLockKernelBO(void); -extern void driReadUnlockKernelBO(void); -extern void driWriteLockKernelBO(void); -extern void driWriteUnlockKernelBO(void); - -/* - * For debugging purposes. - */ - -extern drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list); -extern drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list); -#endif diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_bufpool.h b/src/gallium/winsys/common/intel_drm/ws_dri_bufpool.h deleted file mode 100644 index bf60798924..0000000000 --- a/src/gallium/winsys/common/intel_drm/ws_dri_bufpool.h +++ /dev/null @@ -1,102 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#ifndef _PSB_BUFPOOL_H_ -#define _PSB_BUFPOOL_H_ - -#include -#include -struct _DriFenceObject; - -typedef struct _DriBufferPool -{ - int fd; - int (*map) (struct _DriBufferPool * pool, void *private, - unsigned flags, int hint, _glthread_Mutex *mutex, - void **virtual); - int (*unmap) (struct _DriBufferPool * pool, void *private); - int (*destroy) (struct _DriBufferPool * pool, void *private); - unsigned long (*offset) (struct _DriBufferPool * pool, void *private); - unsigned long (*poolOffset) (struct _DriBufferPool * pool, void *private); - uint64_t (*flags) (struct _DriBufferPool * pool, void *private); - unsigned long (*size) (struct _DriBufferPool * pool, void *private); - void *(*create) (struct _DriBufferPool * pool, unsigned long size, - uint64_t flags, unsigned hint, unsigned alignment); - void *(*reference) (struct _DriBufferPool * pool, unsigned handle); - int (*unreference) (struct _DriBufferPool * pool, void *private); - int (*fence) (struct _DriBufferPool * pool, void *private, - struct _DriFenceObject * fence); - drmBO *(*kernel) (struct _DriBufferPool * pool, void *private); - int (*validate) (struct _DriBufferPool * pool, void *private, _glthread_Mutex *mutex); - int (*waitIdle) (struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, - int lazy); - int (*setStatus) (struct _DriBufferPool *pool, void *private, - uint64_t flag_diff, uint64_t old_flags); - void (*takeDown) (struct _DriBufferPool * pool); - void *data; -} DriBufferPool; - -extern void bmError(int val, const char *file, const char *function, - int line); -#define BM_CKFATAL(val) \ - do{ \ - int tstVal = (val); \ - if (tstVal) \ - bmError(tstVal, __FILE__, __FUNCTION__, __LINE__); \ - } while(0); - - -/* - * Builtin pools. - */ - -/* - * Kernel buffer objects. Size in multiples of page size. Page size aligned. - */ - -extern struct _DriBufferPool *driDRMPoolInit(int fd); -extern struct _DriBufferPool *driMallocPoolInit(void); - -struct _DriFreeSlabManager; -extern struct _DriBufferPool * driSlabPoolInit(int fd, uint64_t flags, - uint64_t validMask, - uint32_t smallestSize, - uint32_t numSizes, - uint32_t desiredNumBuffers, - uint32_t maxSlabSize, - uint32_t pageAlignment, - struct _DriFreeSlabManager *fMan); -extern void driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan); -extern struct _DriFreeSlabManager * -driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec); - - -#endif diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_drmpool.c b/src/gallium/winsys/common/intel_drm/ws_dri_drmpool.c deleted file mode 100644 index 40929efa2f..0000000000 --- a/src/gallium/winsys/common/intel_drm/ws_dri_drmpool.c +++ /dev/null @@ -1,268 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#include -#include -#include -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" -#include "assert.h" - -/* - * Buffer pool implementation using DRM buffer objects as DRI buffer objects. - */ - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, uint64_t flags, unsigned hint, - unsigned alignment) -{ - drmBO *buf = (drmBO *) malloc(sizeof(*buf)); - int ret; - unsigned pageSize = getpagesize(); - - if (!buf) - return NULL; - - if ((alignment > pageSize) && (alignment % pageSize)) { - free(buf); - return NULL; - } - - ret = drmBOCreate(pool->fd, size, alignment / pageSize, - NULL, - flags, hint, buf); - if (ret) { - free(buf); - return NULL; - } - - return (void *) buf; -} - -static void * -pool_reference(struct _DriBufferPool *pool, unsigned handle) -{ - drmBO *buf = (drmBO *) malloc(sizeof(*buf)); - int ret; - - if (!buf) - return NULL; - - ret = drmBOReference(pool->fd, handle, buf); - - if (ret) { - free(buf); - return NULL; - } - - return (void *) buf; -} - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - int ret; - drmBO *buf = (drmBO *) private; - driReadLockKernelBO(); - ret = drmBOUnreference(pool->fd, buf); - free(buf); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_unreference(struct _DriBufferPool *pool, void *private) -{ - int ret; - drmBO *buf = (drmBO *) private; - driReadLockKernelBO(); - ret = drmBOUnreference(pool->fd, buf); - free(buf); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, _glthread_Mutex *mutex, void **virtual) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOMap(pool->fd, buf, flags, hint, virtual); - driReadUnlockKernelBO(); - return ret; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOUnmap(pool->fd, buf); - driReadUnlockKernelBO(); - - return ret; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - unsigned long offset; - - driReadLockKernelBO(); - assert(buf->flags & DRM_BO_FLAG_NO_MOVE); - offset = buf->offset; - driReadUnlockKernelBO(); - - return buf->offset; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - return 0; -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - uint64_t flags; - - driReadLockKernelBO(); - flags = buf->flags; - driReadUnlockKernelBO(); - - return flags; -} - - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - drmBO *buf = (drmBO *) private; - unsigned long size; - - driReadLockKernelBO(); - size = buf->size; - driReadUnlockKernelBO(); - - return buf->size; -} - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - /* - * Noop. The kernel handles all fencing. - */ - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - return (drmBO *) private; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, - int lazy) -{ - drmBO *buf = (drmBO *) private; - int ret; - - driReadLockKernelBO(); - ret = drmBOWaitIdle(pool->fd, buf, (lazy) ? DRM_BO_HINT_WAIT_LAZY:0); - driReadUnlockKernelBO(); - - return ret; -} - - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - free(pool); -} - -/*static int -pool_setStatus(struct _DriBufferPool *pool, void *private, - uint64_t flag_diff, uint64_t old_flags) -{ - drmBO *buf = (drmBO *) private; - uint64_t new_flags = old_flags ^ flag_diff; - int ret; - - driReadLockKernelBO(); - ret = drmBOSetStatus(pool->fd, buf, new_flags, flag_diff, - 0, 0, 0); - driReadUnlockKernelBO(); - return ret; -}*/ - -struct _DriBufferPool * -driDRMPoolInit(int fd) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - - if (!pool) - return NULL; - - pool->fd = fd; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->poolOffset = &pool_poolOffset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = NULL; - pool->waitIdle = &pool_waitIdle; - pool->takeDown = &pool_takedown; - pool->reference = &pool_reference; - pool->unreference = &pool_unreference; - pool->data = NULL; - return pool; -} diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.c b/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.c deleted file mode 100644 index b56bc269da..0000000000 --- a/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.c +++ /dev/null @@ -1,377 +0,0 @@ -#include "ws_dri_fencemgr.h" -#include "glthread.h" -#include -#include -#include - -/* - * Note: Locking order is - * _DriFenceObject::mutex - * _DriFenceMgr::mutex - */ - -struct _DriFenceMgr { - /* - * Constant members. Need no mutex protection. - */ - struct _DriFenceMgrCreateInfo info; - void *private; - - /* - * These members are protected by this->mutex - */ - _glthread_Mutex mutex; - int refCount; - drmMMListHead *heads; - int num_fences; -}; - -struct _DriFenceObject { - - /* - * These members are constant and need no mutex protection. - */ - struct _DriFenceMgr *mgr; - uint32_t fence_class; - uint32_t fence_type; - - /* - * These members are protected by mgr->mutex. - */ - drmMMListHead head; - int refCount; - - /* - * These members are protected by this->mutex. - */ - _glthread_Mutex mutex; - uint32_t signaled_type; - void *private; -}; - -uint32_t -driFenceType(struct _DriFenceObject *fence) -{ - return fence->fence_type; -} - -struct _DriFenceMgr * -driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info) -{ - struct _DriFenceMgr *tmp; - uint32_t i; - - tmp = calloc(1, sizeof(*tmp)); - if (!tmp) - return NULL; - - _glthread_INIT_MUTEX(tmp->mutex); - _glthread_LOCK_MUTEX(tmp->mutex); - tmp->refCount = 1; - tmp->info = *info; - tmp->num_fences = 0; - tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads)); - if (!tmp->heads) - goto out_err; - - for (i=0; iinfo.num_classes; ++i) { - DRMINITLISTHEAD(&tmp->heads[i]); - } - _glthread_UNLOCK_MUTEX(tmp->mutex); - return tmp; - - out_err: - if (tmp) - free(tmp); - return NULL; -} - -static void -driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr) -{ - struct _DriFenceMgr *mgr = *pMgr; - - *pMgr = NULL; - if (--mgr->refCount == 0) - free(mgr); - else - _glthread_UNLOCK_MUTEX(mgr->mutex); -} - -void -driFenceMgrUnReference(struct _DriFenceMgr **pMgr) -{ - _glthread_LOCK_MUTEX((*pMgr)->mutex); - driFenceMgrUnrefUnlock(pMgr); -} - -static void -driFenceUnReferenceLocked(struct _DriFenceObject **pFence) -{ - struct _DriFenceObject *fence = *pFence; - struct _DriFenceMgr *mgr = fence->mgr; - - *pFence = NULL; - if (--fence->refCount == 0) { - DRMLISTDELINIT(&fence->head); - if (fence->private) - mgr->info.unreference(mgr, &fence->private); - --mgr->num_fences; - fence->mgr = NULL; - --mgr->refCount; - free(fence); - - } -} - - -static void -driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr, - drmMMListHead *list, - uint32_t fence_class, - uint32_t fence_type) -{ - struct _DriFenceObject *entry; - drmMMListHead *prev; - - while(list != &mgr->heads[fence_class]) { - entry = DRMLISTENTRY(struct _DriFenceObject, list, head); - - /* - * Up refcount so that entry doesn't disappear from under us - * when we unlock-relock mgr to get the correct locking order. - */ - - ++entry->refCount; - _glthread_UNLOCK_MUTEX(mgr->mutex); - _glthread_LOCK_MUTEX(entry->mutex); - _glthread_LOCK_MUTEX(mgr->mutex); - - prev = list->prev; - - - - if (list->prev == list) { - - /* - * Somebody else removed the entry from the list. - */ - - _glthread_UNLOCK_MUTEX(entry->mutex); - driFenceUnReferenceLocked(&entry); - return; - } - - entry->signaled_type |= (fence_type & entry->fence_type); - if (entry->signaled_type == entry->fence_type) { - DRMLISTDELINIT(list); - mgr->info.unreference(mgr, &entry->private); - } - _glthread_UNLOCK_MUTEX(entry->mutex); - driFenceUnReferenceLocked(&entry); - list = prev; - } -} - - -int -driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, - int lazy_hint) -{ - struct _DriFenceMgr *mgr = fence->mgr; - int ret = 0; - - _glthread_LOCK_MUTEX(fence->mutex); - - if ((fence->signaled_type & fence_type) == fence_type) - goto out0; - - ret = mgr->info.finish(mgr, fence->private, fence_type, lazy_hint); - if (ret) - goto out0; - - _glthread_LOCK_MUTEX(mgr->mutex); - _glthread_UNLOCK_MUTEX(fence->mutex); - - driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, - fence_type); - _glthread_UNLOCK_MUTEX(mgr->mutex); - return 0; - - out0: - _glthread_UNLOCK_MUTEX(fence->mutex); - return ret; -} - -uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence) -{ - uint32_t ret; - - _glthread_LOCK_MUTEX(fence->mutex); - ret = fence->signaled_type; - _glthread_UNLOCK_MUTEX(fence->mutex); - - return ret; -} - -int -driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type, - uint32_t *signaled) -{ - int ret = 0; - struct _DriFenceMgr *mgr; - - _glthread_LOCK_MUTEX(fence->mutex); - mgr = fence->mgr; - *signaled = fence->signaled_type; - if ((fence->signaled_type & flush_type) == flush_type) - goto out0; - - ret = mgr->info.signaled(mgr, fence->private, flush_type, signaled); - if (ret) { - *signaled = fence->signaled_type; - goto out0; - } - - if ((fence->signaled_type | *signaled) == fence->signaled_type) - goto out0; - - _glthread_LOCK_MUTEX(mgr->mutex); - _glthread_UNLOCK_MUTEX(fence->mutex); - - driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, - *signaled); - - _glthread_UNLOCK_MUTEX(mgr->mutex); - return 0; - out0: - _glthread_UNLOCK_MUTEX(fence->mutex); - return ret; -} - -struct _DriFenceObject * -driFenceReference(struct _DriFenceObject *fence) -{ - _glthread_LOCK_MUTEX(fence->mgr->mutex); - ++fence->refCount; - _glthread_UNLOCK_MUTEX(fence->mgr->mutex); - return fence; -} - -void -driFenceUnReference(struct _DriFenceObject **pFence) -{ - struct _DriFenceMgr *mgr; - - if (*pFence == NULL) - return; - - mgr = (*pFence)->mgr; - _glthread_LOCK_MUTEX(mgr->mutex); - ++mgr->refCount; - driFenceUnReferenceLocked(pFence); - driFenceMgrUnrefUnlock(&mgr); -} - -struct _DriFenceObject -*driFenceCreate(struct _DriFenceMgr *mgr, uint32_t fence_class, - uint32_t fence_type, void *private, size_t private_size) -{ - struct _DriFenceObject *fence; - size_t fence_size = sizeof(*fence); - - if (private_size) - fence_size = ((fence_size + 15) & ~15); - - fence = calloc(1, fence_size + private_size); - - if (!fence) { - int ret = mgr->info.finish(mgr, private, fence_type, 0); - - if (ret) - usleep(10000000); - - return NULL; - } - - _glthread_INIT_MUTEX(fence->mutex); - _glthread_LOCK_MUTEX(fence->mutex); - _glthread_LOCK_MUTEX(mgr->mutex); - fence->refCount = 1; - DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]); - fence->mgr = mgr; - ++mgr->refCount; - ++mgr->num_fences; - _glthread_UNLOCK_MUTEX(mgr->mutex); - fence->fence_class = fence_class; - fence->fence_type = fence_type; - fence->signaled_type = 0; - fence->private = private; - if (private_size) { - fence->private = (void *)(((uint8_t *) fence) + fence_size); - memcpy(fence->private, private, private_size); - } - - _glthread_UNLOCK_MUTEX(fence->mutex); - return fence; -} - - -static int -tSignaled(struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, - uint32_t *signaled_type) -{ - long fd = (long) mgr->private; - int dummy; - drmFence *fence = (drmFence *) private; - int ret; - - *signaled_type = 0; - ret = drmFenceSignaled((int) fd, fence, flush_type, &dummy); - if (ret) - return ret; - - *signaled_type = fence->signaled; - - return 0; -} - -static int -tFinish(struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, - int lazy_hint) -{ - long fd = (long) mgr->private; - unsigned flags = lazy_hint ? DRM_FENCE_FLAG_WAIT_LAZY : 0; - - return drmFenceWait((int)fd, flags, (drmFence *) private, fence_type); -} - -static int -tUnref(struct _DriFenceMgr *mgr, void **private) -{ - long fd = (long) mgr->private; - drmFence *fence = (drmFence *) *private; - *private = NULL; - - return drmFenceUnreference(fd, fence); -} - -struct _DriFenceMgr *driFenceMgrTTMInit(int fd) -{ - struct _DriFenceMgrCreateInfo info; - struct _DriFenceMgr *mgr; - - info.flags = DRI_FENCE_CLASS_ORDERED; - info.num_classes = 4; - info.signaled = tSignaled; - info.finish = tFinish; - info.unreference = tUnref; - - mgr = driFenceMgrCreate(&info); - if (mgr == NULL) - return NULL; - - mgr->private = (void *) (long) fd; - return mgr; -} - diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.h b/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.h deleted file mode 100644 index 4ea58dfe18..0000000000 --- a/src/gallium/winsys/common/intel_drm/ws_dri_fencemgr.h +++ /dev/null @@ -1,115 +0,0 @@ -#ifndef DRI_FENCEMGR_H -#define DRI_FENCEMGR_H - -#include -#include - -struct _DriFenceObject; -struct _DriFenceMgr; - -/* - * Do a quick check to see if the fence manager has registered the fence - * object as signaled. Note that this function may return a false negative - * answer. - */ -extern uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence); - -/* - * Check if the fence object is signaled. This function can be substantially - * more expensive to call than the above function, but will not return a false - * negative answer. The argument "flush_type" sets the types that the - * underlying mechanism must make sure will eventually signal. - */ -extern int driFenceSignaledType(struct _DriFenceObject *fence, - uint32_t flush_type, uint32_t *signaled); - -/* - * Convenience functions. - */ - -static inline int driFenceSignaled(struct _DriFenceObject *fence, - uint32_t flush_type) -{ - uint32_t signaled_types; - int ret = driFenceSignaledType(fence, flush_type, &signaled_types); - if (ret) - return 0; - return ((signaled_types & flush_type) == flush_type); -} - -static inline int driFenceSignaledCached(struct _DriFenceObject *fence, - uint32_t flush_type) -{ - uint32_t signaled_types = - driFenceSignaledTypeCached(fence); - - return ((signaled_types & flush_type) == flush_type); -} - -/* - * Reference a fence object. - */ -extern struct _DriFenceObject *driFenceReference(struct _DriFenceObject *fence); - -/* - * Unreference a fence object. The fence object pointer will be reset to NULL. - */ - -extern void driFenceUnReference(struct _DriFenceObject **pFence); - - -/* - * Wait for a fence to signal the indicated fence_type. - * If "lazy_hint" is true, it indicates that the wait may sleep to avoid - * busy-wait polling. - */ -extern int driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, - int lazy_hint); - -/* - * Create a DriFenceObject for manager "mgr". - * - * "private" is a pointer that should be used for the callbacks in - * struct _DriFenceMgrCreateInfo. - * - * if private_size is nonzero, then the info stored at *private, with size - * private size will be copied and the fence manager will instead use a - * pointer to the copied data for the callbacks in - * struct _DriFenceMgrCreateInfo. In that case, the object pointed to by - * "private" may be destroyed after the call to driFenceCreate. - */ -extern struct _DriFenceObject *driFenceCreate(struct _DriFenceMgr *mgr, - uint32_t fence_class, - uint32_t fence_type, - void *private, - size_t private_size); - -extern uint32_t driFenceType(struct _DriFenceObject *fence); - -/* - * Fence creations are ordered. If a fence signals a fence_type, - * it is safe to assume that all fences of the same class that was - * created before that fence has signaled the same type. - */ - -#define DRI_FENCE_CLASS_ORDERED (1 << 0) - -struct _DriFenceMgrCreateInfo { - uint32_t flags; - uint32_t num_classes; - int (*signaled) (struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, - uint32_t *signaled_type); - int (*finish) (struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, int lazy_hint); - int (*unreference) (struct _DriFenceMgr *mgr, void **private); -}; - -extern struct _DriFenceMgr * -driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info); - -void -driFenceMgrUnReference(struct _DriFenceMgr **pMgr); - -extern struct _DriFenceMgr * -driFenceMgrTTMInit(int fd); - -#endif diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_mallocpool.c b/src/gallium/winsys/common/intel_drm/ws_dri_mallocpool.c deleted file mode 100644 index a80555c9c7..0000000000 --- a/src/gallium/winsys/common/intel_drm/ws_dri_mallocpool.c +++ /dev/null @@ -1,161 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, TX., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellström - */ - -#include -#include -#include -#include "pipe/p_debug.h" -#include "glthread.h" -#include "ws_dri_bufpool.h" -#include "ws_dri_bufmgr.h" - -static void * -pool_create(struct _DriBufferPool *pool, - unsigned long size, uint64_t flags, unsigned hint, - unsigned alignment) -{ - unsigned long *private = malloc(size + 2*sizeof(unsigned long)); - if ((flags & DRM_BO_MASK_MEM) != DRM_BO_FLAG_MEM_LOCAL) - abort(); - - *private = size; - return (void *)private; -} - - -static int -pool_destroy(struct _DriBufferPool *pool, void *private) -{ - free(private); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *pool, void *private, - _glthread_Mutex *mutex, int lazy) -{ - return 0; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, _glthread_Mutex *mutex, void **virtual) -{ - *virtual = (void *)((unsigned long *)private + 2); - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - /* - * BUG - */ - abort(); - return 0UL; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - /* - * BUG - */ - abort(); -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - return DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - return *(unsigned long *) private; -} - - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - abort(); - return 0UL; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - abort(); - return NULL; -} - -static void -pool_takedown(struct _DriBufferPool *pool) -{ - free(pool); -} - - -struct _DriBufferPool * -driMallocPoolInit(void) -{ - struct _DriBufferPool *pool; - - pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); - if (!pool) - return NULL; - - pool->data = NULL; - pool->fd = -1; - pool->map = &pool_map; - pool->unmap = &pool_unmap; - pool->destroy = &pool_destroy; - pool->offset = &pool_offset; - pool->poolOffset = &pool_poolOffset; - pool->flags = &pool_flags; - pool->size = &pool_size; - pool->create = &pool_create; - pool->fence = &pool_fence; - pool->kernel = &pool_kernel; - pool->validate = NULL; - pool->waitIdle = &pool_waitIdle; - pool->takeDown = &pool_takedown; - return pool; -} diff --git a/src/gallium/winsys/common/intel_drm/ws_dri_slabpool.c b/src/gallium/winsys/common/intel_drm/ws_dri_slabpool.c deleted file mode 100644 index dfcf6d6b19..0000000000 --- a/src/gallium/winsys/common/intel_drm/ws_dri_slabpool.c +++ /dev/null @@ -1,968 +0,0 @@ -/************************************************************************** - * - * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Thomas Hellstrom - */ - -#include -#include -#include -#include -#include -#include "ws_dri_bufpool.h" -#include "ws_dri_fencemgr.h" -#include "ws_dri_bufmgr.h" -#include "glthread.h" - -#define DRI_SLABPOOL_ALLOC_RETRIES 100 - -struct _DriSlab; - -struct _DriSlabBuffer { - int isSlabBuffer; - drmBO *bo; - struct _DriFenceObject *fence; - struct _DriSlab *parent; - drmMMListHead head; - uint32_t mapCount; - uint32_t start; - uint32_t fenceType; - int unFenced; - _glthread_Cond event; -}; - -struct _DriKernelBO { - int fd; - drmBO bo; - drmMMListHead timeoutHead; - drmMMListHead head; - struct timeval timeFreed; - uint32_t pageAlignment; - void *virtual; -}; - -struct _DriSlab{ - drmMMListHead head; - drmMMListHead freeBuffers; - uint32_t numBuffers; - uint32_t numFree; - struct _DriSlabBuffer *buffers; - struct _DriSlabSizeHeader *header; - struct _DriKernelBO *kbo; -}; - - -struct _DriSlabSizeHeader { - drmMMListHead slabs; - drmMMListHead freeSlabs; - drmMMListHead delayedBuffers; - uint32_t numDelayed; - struct _DriSlabPool *slabPool; - uint32_t bufSize; - _glthread_Mutex mutex; -}; - -struct _DriFreeSlabManager { - struct timeval slabTimeout; - struct timeval checkInterval; - struct timeval nextCheck; - drmMMListHead timeoutList; - drmMMListHead unCached; - drmMMListHead cached; - _glthread_Mutex mutex; -}; - - -struct _DriSlabPool { - - /* - * The data of this structure remains constant after - * initialization and thus needs no mutex protection. - */ - - struct _DriFreeSlabManager *fMan; - uint64_t proposedFlags; - uint64_t validMask; - uint32_t *bucketSizes; - uint32_t numBuckets; - uint32_t pageSize; - int fd; - int pageAlignment; - int maxSlabSize; - int desiredNumBuffers; - struct _DriSlabSizeHeader *headers; -}; - -/* - * FIXME: Perhaps arrange timeout slabs in size buckets for fast - * retreival?? - */ - - -static inline int -driTimeAfterEq(struct timeval *arg1, struct timeval *arg2) -{ - return ((arg1->tv_sec > arg2->tv_sec) || - ((arg1->tv_sec == arg2->tv_sec) && - (arg1->tv_usec > arg2->tv_usec))); -} - -static inline void -driTimeAdd(struct timeval *arg, struct timeval *add) -{ - unsigned int sec; - - arg->tv_sec += add->tv_sec; - arg->tv_usec += add->tv_usec; - sec = arg->tv_usec / 1000000; - arg->tv_sec += sec; - arg->tv_usec -= sec*1000000; -} - -static void -driFreeKernelBO(struct _DriKernelBO *kbo) -{ - if (!kbo) - return; - - (void) drmBOUnreference(kbo->fd, &kbo->bo); - free(kbo); -} - - -static void -driFreeTimeoutKBOsLocked(struct _DriFreeSlabManager *fMan, - struct timeval *time) -{ - drmMMListHead *list, *next; - struct _DriKernelBO *kbo; - - if (!driTimeAfterEq(time, &fMan->nextCheck)) - return; - - for (list = fMan->timeoutList.next, next = list->next; - list != &fMan->timeoutList; - list = next, next = list->next) { - - kbo = DRMLISTENTRY(struct _DriKernelBO, list, timeoutHead); - - if (!driTimeAfterEq(time, &kbo->timeFreed)) - break; - - DRMLISTDELINIT(&kbo->timeoutHead); - DRMLISTDELINIT(&kbo->head); - driFreeKernelBO(kbo); - } - - fMan->nextCheck = *time; - driTimeAdd(&fMan->nextCheck, &fMan->checkInterval); -} - - -/* - * Add a _DriKernelBO to the free slab manager. - * This means that it is available for reuse, but if it's not - * reused in a while, it will be freed. - */ - -static void -driSetKernelBOFree(struct _DriFreeSlabManager *fMan, - struct _DriKernelBO *kbo) -{ - struct timeval time; - - _glthread_LOCK_MUTEX(fMan->mutex); - gettimeofday(&time, NULL); - driTimeAdd(&time, &fMan->slabTimeout); - - kbo->timeFreed = time; - - if (kbo->bo.flags & DRM_BO_FLAG_CACHED) - DRMLISTADD(&kbo->head, &fMan->cached); - else - DRMLISTADD(&kbo->head, &fMan->unCached); - - DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList); - driFreeTimeoutKBOsLocked(fMan, &time); - - _glthread_UNLOCK_MUTEX(fMan->mutex); -} - -/* - * Get a _DriKernelBO for us to use as storage for a slab. - * - */ - -static struct _DriKernelBO * -driAllocKernelBO(struct _DriSlabSizeHeader *header) - -{ - struct _DriSlabPool *slabPool = header->slabPool; - struct _DriFreeSlabManager *fMan = slabPool->fMan; - drmMMListHead *list, *next, *head; - uint32_t size = header->bufSize * slabPool->desiredNumBuffers; - struct _DriKernelBO *kbo; - struct _DriKernelBO *kboTmp; - int ret; - - /* - * FIXME: We should perhaps allow some variation in slabsize in order - * to efficiently reuse slabs. - */ - - size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize; - size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1); - _glthread_LOCK_MUTEX(fMan->mutex); - - kbo = NULL; - - retry: - head = (slabPool->proposedFlags & DRM_BO_FLAG_CACHED) ? - &fMan->cached : &fMan->unCached; - - for (list = head->next, next = list->next; - list != head; - list = next, next = list->next) { - - kboTmp = DRMLISTENTRY(struct _DriKernelBO, list, head); - - if ((kboTmp->bo.size == size) && - (slabPool->pageAlignment == 0 || - (kboTmp->pageAlignment % slabPool->pageAlignment) == 0)) { - - if (!kbo) - kbo = kboTmp; - - if ((kbo->bo.proposedFlags ^ slabPool->proposedFlags) == 0) - break; - - } - } - - if (kbo) { - DRMLISTDELINIT(&kbo->head); - DRMLISTDELINIT(&kbo->timeoutHead); - } - - _glthread_UNLOCK_MUTEX(fMan->mutex); - - if (kbo) { - uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags; - - ret = 0; - if (new_mask) { - ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, - new_mask, DRM_BO_HINT_DONT_FENCE, 0, 0); - } - if (ret == 0) - return kbo; - - driFreeKernelBO(kbo); - kbo = NULL; - goto retry; - } - - kbo = calloc(1, sizeof(struct _DriKernelBO)); - if (!kbo) - return NULL; - - kbo->fd = slabPool->fd; - DRMINITLISTHEAD(&kbo->head); - DRMINITLISTHEAD(&kbo->timeoutHead); - ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL, - slabPool->proposedFlags, - DRM_BO_HINT_DONT_FENCE, &kbo->bo); - if (ret) - goto out_err0; - - ret = drmBOMap(kbo->fd, &kbo->bo, - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0, &kbo->virtual); - - if (ret) - goto out_err1; - - ret = drmBOUnmap(kbo->fd, &kbo->bo); - if (ret) - goto out_err1; - - return kbo; - - out_err1: - drmBOUnreference(kbo->fd, &kbo->bo); - out_err0: - free(kbo); - return NULL; -} - - -static int -driAllocSlab(struct _DriSlabSizeHeader *header) -{ - struct _DriSlab *slab; - struct _DriSlabBuffer *buf; - uint32_t numBuffers; - int ret; - int i; - - slab = calloc(1, sizeof(*slab)); - if (!slab) - return -ENOMEM; - - slab->kbo = driAllocKernelBO(header); - if (!slab->kbo) { - ret = -ENOMEM; - goto out_err0; - } - - numBuffers = slab->kbo->bo.size / header->bufSize; - - slab->buffers = calloc(numBuffers, sizeof(*slab->buffers)); - if (!slab->buffers) { - ret = -ENOMEM; - goto out_err1; - } - - DRMINITLISTHEAD(&slab->head); - DRMINITLISTHEAD(&slab->freeBuffers); - slab->numBuffers = numBuffers; - slab->numFree = 0; - slab->header = header; - - buf = slab->buffers; - for (i=0; i < numBuffers; ++i) { - buf->parent = slab; - buf->start = i* header->bufSize; - buf->mapCount = 0; - buf->isSlabBuffer = 1; - _glthread_INIT_COND(buf->event); - DRMLISTADDTAIL(&buf->head, &slab->freeBuffers); - slab->numFree++; - buf++; - } - - DRMLISTADDTAIL(&slab->head, &header->slabs); - - return 0; - - out_err1: - driSetKernelBOFree(header->slabPool->fMan, slab->kbo); - free(slab->buffers); - out_err0: - free(slab); - return ret; -} - -/* - * Delete a buffer from the slab header delayed list and put - * it on the slab free list. - */ - -static void -driSlabFreeBufferLocked(struct _DriSlabBuffer *buf) -{ - struct _DriSlab *slab = buf->parent; - struct _DriSlabSizeHeader *header = slab->header; - drmMMListHead *list = &buf->head; - - DRMLISTDEL(list); - DRMLISTADDTAIL(list, &slab->freeBuffers); - slab->numFree++; - - if (slab->head.next == &slab->head) - DRMLISTADDTAIL(&slab->head, &header->slabs); - - if (slab->numFree == slab->numBuffers) { - list = &slab->head; - DRMLISTDEL(list); - DRMLISTADDTAIL(list, &header->freeSlabs); - } - - if (header->slabs.next == &header->slabs || - slab->numFree != slab->numBuffers) { - - drmMMListHead *next; - struct _DriFreeSlabManager *fMan = header->slabPool->fMan; - - for (list = header->freeSlabs.next, next = list->next; - list != &header->freeSlabs; - list = next, next = list->next) { - - slab = DRMLISTENTRY(struct _DriSlab, list, head); - - DRMLISTDELINIT(list); - driSetKernelBOFree(fMan, slab->kbo); - free(slab->buffers); - free(slab); - } - } -} - -static void -driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait) -{ - drmMMListHead *list, *prev, *first; - struct _DriSlabBuffer *buf; - struct _DriSlab *slab; - int firstWasSignaled = 1; - int signaled; - int i; - int ret; - - /* - * Rerun the freeing test if the youngest tested buffer - * was signaled, since there might be more idle buffers - * in the delay list. - */ - - while (firstWasSignaled) { - firstWasSignaled = 0; - signaled = 0; - first = header->delayedBuffers.next; - - /* Only examine the oldest 1/3 of delayed buffers: - */ - if (header->numDelayed > 3) { - for (i = 0; i < header->numDelayed; i += 3) { - first = first->next; - } - } - - for (list = first, prev = list->prev; - list != &header->delayedBuffers; - list = prev, prev = list->prev) { - buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); - slab = buf->parent; - - if (!signaled) { - if (wait) { - ret = driFenceFinish(buf->fence, buf->fenceType, 0); - if (ret) - break; - signaled = 1; - wait = 0; - } else { - signaled = driFenceSignaled(buf->fence, buf->fenceType); - } - if (signaled) { - if (list == first) - firstWasSignaled = 1; - driFenceUnReference(&buf->fence); - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { - driFenceUnReference(&buf->fence); - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - } - } -} - - -static struct _DriSlabBuffer * -driSlabAllocBuffer(struct _DriSlabSizeHeader *header) -{ - static struct _DriSlabBuffer *buf; - struct _DriSlab *slab; - drmMMListHead *list; - int count = DRI_SLABPOOL_ALLOC_RETRIES; - - _glthread_LOCK_MUTEX(header->mutex); - while(header->slabs.next == &header->slabs && count > 0) { - driSlabCheckFreeLocked(header, 0); - if (header->slabs.next != &header->slabs) - break; - - _glthread_UNLOCK_MUTEX(header->mutex); - if (count != DRI_SLABPOOL_ALLOC_RETRIES) - usleep(1); - _glthread_LOCK_MUTEX(header->mutex); - (void) driAllocSlab(header); - count--; - } - - list = header->slabs.next; - if (list == &header->slabs) { - _glthread_UNLOCK_MUTEX(header->mutex); - return NULL; - } - slab = DRMLISTENTRY(struct _DriSlab, list, head); - if (--slab->numFree == 0) - DRMLISTDELINIT(list); - - list = slab->freeBuffers.next; - DRMLISTDELINIT(list); - - _glthread_UNLOCK_MUTEX(header->mutex); - buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); - return buf; -} - -static void * -pool_create(struct _DriBufferPool *driPool, unsigned long size, - uint64_t flags, unsigned hint, unsigned alignment) -{ - struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; - struct _DriSlabSizeHeader *header; - struct _DriSlabBuffer *buf; - void *dummy; - int i; - int ret; - - /* - * FIXME: Check for compatibility. - */ - - header = pool->headers; - for (i=0; inumBuckets; ++i) { - if (header->bufSize >= size) - break; - header++; - } - - if (i < pool->numBuckets) - return driSlabAllocBuffer(header); - - - /* - * Fall back to allocate a buffer object directly from DRM. - * and wrap it in a driBO structure. - */ - - - buf = calloc(1, sizeof(*buf)); - - if (!buf) - return NULL; - - buf->bo = calloc(1, sizeof(*buf->bo)); - if (!buf->bo) - goto out_err0; - - if (alignment) { - if ((alignment < pool->pageSize) && (pool->pageSize % alignment)) - goto out_err1; - if ((alignment > pool->pageSize) && (alignment % pool->pageSize)) - goto out_err1; - } - - ret = drmBOCreate(pool->fd, size, alignment / pool->pageSize, NULL, - flags, hint, buf->bo); - if (ret) - goto out_err1; - - ret = drmBOMap(pool->fd, buf->bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, - 0, &dummy); - if (ret) - goto out_err2; - - ret = drmBOUnmap(pool->fd, buf->bo); - if (ret) - goto out_err2; - - return buf; - out_err2: - drmBOUnreference(pool->fd, buf->bo); - out_err1: - free(buf->bo); - out_err0: - free(buf); - return NULL; -} - -static int -pool_destroy(struct _DriBufferPool *driPool, void *private) -{ - struct _DriSlabBuffer *buf = - (struct _DriSlabBuffer *) private; - struct _DriSlab *slab; - struct _DriSlabSizeHeader *header; - - if (!buf->isSlabBuffer) { - struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; - int ret; - - ret = drmBOUnreference(pool->fd, buf->bo); - free(buf->bo); - free(buf); - return ret; - } - - slab = buf->parent; - header = slab->header; - - _glthread_LOCK_MUTEX(header->mutex); - buf->unFenced = 0; - buf->mapCount = 0; - - if (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)) { - DRMLISTADDTAIL(&buf->head, &header->delayedBuffers); - header->numDelayed++; - } else { - if (buf->fence) - driFenceUnReference(&buf->fence); - driSlabFreeBufferLocked(buf); - } - - _glthread_UNLOCK_MUTEX(header->mutex); - return 0; -} - -static int -pool_waitIdle(struct _DriBufferPool *driPool, void *private, - _glthread_Mutex *mutex, int lazy) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - while(buf->unFenced) - _glthread_COND_WAIT(buf->event, *mutex); - - if (!buf->fence) - return 0; - - driFenceFinish(buf->fence, buf->fenceType, lazy); - driFenceUnReference(&buf->fence); - - return 0; -} - -static int -pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, - int hint, _glthread_Mutex *mutex, void **virtual) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - int busy; - - if (buf->isSlabBuffer) - busy = buf->unFenced || (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)); - else - busy = buf->fence && !driFenceSignaled(buf->fence, buf->fenceType); - - - if (busy) { - if (hint & DRM_BO_HINT_DONT_BLOCK) - return -EBUSY; - else { - (void) pool_waitIdle(pool, private, mutex, 0); - } - } - - ++buf->mapCount; - *virtual = (buf->isSlabBuffer) ? - (void *) ((uint8_t *) buf->parent->kbo->virtual + buf->start) : - (void *) buf->bo->virtual; - - return 0; -} - -static int -pool_unmap(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - --buf->mapCount; - if (buf->mapCount == 0 && buf->isSlabBuffer) - _glthread_COND_BROADCAST(buf->event); - - return 0; -} - -static unsigned long -pool_offset(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - struct _DriSlab *slab; - struct _DriSlabSizeHeader *header; - - if (!buf->isSlabBuffer) { - assert(buf->bo->proposedFlags & DRM_BO_FLAG_NO_MOVE); - return buf->bo->offset; - } - - slab = buf->parent; - header = slab->header; - - (void) header; - assert(header->slabPool->proposedFlags & DRM_BO_FLAG_NO_MOVE); - return slab->kbo->bo.offset + buf->start; -} - -static unsigned long -pool_poolOffset(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - return buf->start; -} - -static uint64_t -pool_flags(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - if (!buf->isSlabBuffer) - return buf->bo->flags; - - return buf->parent->kbo->bo.flags; -} - -static unsigned long -pool_size(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - if (!buf->isSlabBuffer) - return buf->bo->size; - - return buf->parent->header->bufSize; -} - -static int -pool_fence(struct _DriBufferPool *pool, void *private, - struct _DriFenceObject *fence) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - drmBO *bo; - - if (buf->fence) - driFenceUnReference(&buf->fence); - - buf->fence = driFenceReference(fence); - bo = (buf->isSlabBuffer) ? - &buf->parent->kbo->bo: - buf->bo; - buf->fenceType = bo->fenceFlags; - - buf->unFenced = 0; - _glthread_COND_BROADCAST(buf->event); - - return 0; -} - -static drmBO * -pool_kernel(struct _DriBufferPool *pool, void *private) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - return (buf->isSlabBuffer) ? &buf->parent->kbo->bo : buf->bo; -} - -static int -pool_validate(struct _DriBufferPool *pool, void *private, - _glthread_Mutex *mutex) -{ - struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; - - if (!buf->isSlabBuffer) - return 0; - - while(buf->mapCount != 0) - _glthread_COND_WAIT(buf->event, *mutex); - - buf->unFenced = 1; - return 0; -} - - -struct _DriFreeSlabManager * -driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec) -{ - struct _DriFreeSlabManager *tmp; - - tmp = calloc(1, sizeof(*tmp)); - if (!tmp) - return NULL; - - _glthread_INIT_MUTEX(tmp->mutex); - _glthread_LOCK_MUTEX(tmp->mutex); - tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000; - tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000; - tmp->slabTimeout.tv_usec -= tmp->slabTimeout.tv_sec*1000000; - - tmp->checkInterval.tv_usec = checkIntervalMsec*1000; - tmp->checkInterval.tv_sec = tmp->checkInterval.tv_usec / 1000000; - tmp->checkInterval.tv_usec -= tmp->checkInterval.tv_sec*1000000; - - gettimeofday(&tmp->nextCheck, NULL); - driTimeAdd(&tmp->nextCheck, &tmp->checkInterval); - DRMINITLISTHEAD(&tmp->timeoutList); - DRMINITLISTHEAD(&tmp->unCached); - DRMINITLISTHEAD(&tmp->cached); - _glthread_UNLOCK_MUTEX(tmp->mutex); - - return tmp; -} - -void -driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan) -{ - struct timeval time; - - time = fMan->nextCheck; - driTimeAdd(&time, &fMan->checkInterval); - - _glthread_LOCK_MUTEX(fMan->mutex); - driFreeTimeoutKBOsLocked(fMan, &time); - _glthread_UNLOCK_MUTEX(fMan->mutex); - - assert(fMan->timeoutList.next == &fMan->timeoutList); - assert(fMan->unCached.next == &fMan->unCached); - assert(fMan->cached.next == &fMan->cached); - - free(fMan); -} - -static void -driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size, - struct _DriSlabSizeHeader *header) -{ - _glthread_INIT_MUTEX(header->mutex); - _glthread_LOCK_MUTEX(header->mutex); - - DRMINITLISTHEAD(&header->slabs); - DRMINITLISTHEAD(&header->freeSlabs); - DRMINITLISTHEAD(&header->delayedBuffers); - - header->numDelayed = 0; - header->slabPool = pool; - header->bufSize = size; - - _glthread_UNLOCK_MUTEX(header->mutex); -} - -static void -driFinishSizeHeader(struct _DriSlabSizeHeader *header) -{ - drmMMListHead *list, *next; - struct _DriSlabBuffer *buf; - - _glthread_LOCK_MUTEX(header->mutex); - for (list = header->delayedBuffers.next, next = list->next; - list != &header->delayedBuffers; - list = next, next = list->next) { - - buf = DRMLISTENTRY(struct _DriSlabBuffer, list , head); - if (buf->fence) { - (void) driFenceFinish(buf->fence, buf->fenceType, 0); - driFenceUnReference(&buf->fence); - } - header->numDelayed--; - driSlabFreeBufferLocked(buf); - } - _glthread_UNLOCK_MUTEX(header->mutex); -} - -static void -pool_takedown(struct _DriBufferPool *driPool) -{ - struct _DriSlabPool *pool = driPool->data; - int i; - - for (i=0; inumBuckets; ++i) { - driFinishSizeHeader(&pool->headers[i]); - } - - free(pool->headers); - free(pool->bucketSizes); - free(pool); - free(driPool); -} - -struct _DriBufferPool * -driSlabPoolInit(int fd, uint64_t flags, - uint64_t validMask, - uint32_t smallestSize, - uint32_t numSizes, - uint32_t desiredNumBuffers, - uint32_t maxSlabSize, - uint32_t pageAlignment, - struct _DriFreeSlabManager *fMan) -{ - struct _DriBufferPool *driPool; - struct _DriSlabPool *pool; - uint32_t i; - - driPool = calloc(1, sizeof(*driPool)); - if (!driPool) - return NULL; - - pool = calloc(1, sizeof(*pool)); - if (!pool) - goto out_err0; - - pool->bucketSizes = calloc(numSizes, sizeof(*pool->bucketSizes)); - if (!pool->bucketSizes) - goto out_err1; - - pool->headers = calloc(numSizes, sizeof(*pool->headers)); - if (!pool->headers) - goto out_err2; - - pool->fMan = fMan; - pool->proposedFlags = flags; - pool->validMask = validMask; - pool->numBuckets = numSizes; - pool->pageSize = getpagesize(); - pool->fd = fd; - pool->pageAlignment = pageAlignment; - pool->maxSlabSize = maxSlabSize; - pool->desiredNumBuffers = desiredNumBuffers; - - for (i=0; inumBuckets; ++i) { - pool->bucketSizes[i] = (smallestSize << i); - driInitSizeHeader(pool, pool->bucketSizes[i], - &pool->headers[i]); - } - - driPool->data = (void *) pool; - driPool->map = &pool_map; - driPool->unmap = &pool_unmap; - driPool->destroy = &pool_destroy; - driPool->offset = &pool_offset; - driPool->poolOffset = &pool_poolOffset; - driPool->flags = &pool_flags; - driPool->size = &pool_size; - driPool->create = &pool_create; - driPool->fence = &pool_fence; - driPool->kernel = &pool_kernel; - driPool->validate = &pool_validate; - driPool->waitIdle = &pool_waitIdle; - driPool->takeDown = &pool_takedown; - - return driPool; - - out_err2: - free(pool->bucketSizes); - out_err1: - free(pool); - out_err0: - free(driPool); - - return NULL; -} diff --git a/src/gallium/winsys/dri/Makefile b/src/gallium/winsys/dri/Makefile deleted file mode 100644 index f466ce6c3c..0000000000 --- a/src/gallium/winsys/dri/Makefile +++ /dev/null @@ -1,38 +0,0 @@ -# src/mesa/drivers/dri/Makefile - -TOP = ../../../.. - -include $(TOP)/configs/current - - - -default: $(TOP)/$(LIB_DIR) subdirs - - -$(TOP)/$(LIB_DIR): - -mkdir $(TOP)/$(LIB_DIR) - - -subdirs: - @for dir in $(DRI_DIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ - fi \ - done - - -install: - @for dir in $(DRI_DIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE) install) || exit 1 ; \ - fi \ - done - - -clean: - @for dir in $(DRI_DIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE) clean) ; \ - fi \ - done - -rm -f common/*.o diff --git a/src/gallium/winsys/dri/Makefile.template b/src/gallium/winsys/dri/Makefile.template deleted file mode 100644 index 80e817b808..0000000000 --- a/src/gallium/winsys/dri/Makefile.template +++ /dev/null @@ -1,125 +0,0 @@ -# -*-makefile-*- - -MESA_MODULES = \ - $(TOP)/src/mesa/libmesa.a \ - $(GALLIUM_AUXILIARIES) - -COMMON_GALLIUM_SOURCES = \ - $(TOP)/src/mesa/drivers/dri/common/utils.c \ - $(TOP)/src/mesa/drivers/dri/common/vblank.c \ - $(TOP)/src/mesa/drivers/dri/common/dri_util.c \ - $(TOP)/src/mesa/drivers/dri/common/xmlconfig.c - -COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \ - $(TOP)/src/mesa/drivers/common/driverfuncs.c \ - $(TOP)/src/mesa/drivers/dri/common/texmem.c \ - $(TOP)/src/mesa/drivers/dri/common/drirenderbuffer.c - -COMMON_BM_SOURCES = \ - $(TOP)/src/mesa/drivers/dri/common/dri_bufmgr.c \ - $(TOP)/src/mesa/drivers/dri/common/dri_drmpool.c - - -ifeq ($(WINDOW_SYSTEM),dri) -WINOBJ= -WINLIB= -INCLUDES = $(SHARED_INCLUDES) $(EXPAT_INCLUDES) - -OBJECTS = \ - $(C_SOURCES:.c=.o) \ - $(ASM_SOURCES:.S=.o) - -else -# miniglx -WINOBJ= -WINLIB=-L$(MESA)/src/glx/mini -MINIGLX_INCLUDES = -I$(TOP)/src/glx/mini -INCLUDES = $(MINIGLX_INCLUDES) \ - $(SHARED_INCLUDES) \ - $(PCIACCESS_CFLAGS) - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(MINIGLX_SOURCES:.c=.o) \ - $(ASM_SOURCES:.S=.o) -endif - - -### Include directories -SHARED_INCLUDES = \ - -I. \ - -I$(TOP)/src/mesa/drivers/dri/common \ - -Iserver \ - -I$(TOP)/include \ - -I$(TOP)/include/GL/internal \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/auxiliary \ - -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/src/gallium/winsys/common \ - -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/main \ - -I$(TOP)/src/mesa/glapi \ - -I$(TOP)/src/mesa/math \ - -I$(TOP)/src/mesa/transform \ - -I$(TOP)/src/mesa/shader \ - -I$(TOP)/src/mesa/swrast \ - -I$(TOP)/src/mesa/swrast_setup \ - -I$(TOP)/src/egl/main \ - -I$(TOP)/src/egl/drivers/dri \ - $(LIBDRM_CFLAGS) - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - - -##### TARGETS ##### - -default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR)/$(LIBNAME_EGL) - - -$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template - $(TOP)/bin/mklib -noprefix -o $@ \ - $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) - -$(LIBNAME_EGL): $(WINSYS_OBJECTS) $(LIBS) - $(TOP)/bin/mklib -o $(LIBNAME_EGL) \ - -linker "$(CC)" \ - -noprefix \ - $(OBJECTS) $(MKLIB_OPTIONS) $(WINSYS_OBJECTS) $(PIPE_DRIVERS) $(WINOBJ) $(DRI_LIB_DEPS) \ - --whole-archive $(LIBS) $(GALLIUM_AUXILIARIES) --no-whole-archive - -$(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) - $(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR) - -$(TOP)/$(LIB_DIR)/$(LIBNAME_EGL): $(LIBNAME_EGL) - $(INSTALL) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR) - -depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ - $(ASM_SOURCES) 2> /dev/null - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean: - -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) - -rm -f depend depend.bak - - -install: $(LIBNAME) - $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) - $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) - - -include depend diff --git a/src/gallium/winsys/dri/SConscript b/src/gallium/winsys/dri/SConscript deleted file mode 100644 index aef5210a32..0000000000 --- a/src/gallium/winsys/dri/SConscript +++ /dev/null @@ -1,54 +0,0 @@ -Import('*') - -if env['dri']: - - drienv = env.Clone() - - drienv.Replace(CPPPATH = [ - '#src/mesa/drivers/dri/common', - '#include', - '#include/GL/internal', - '#src/gallium/include', - '#src/gallium/auxiliary', - '#src/gallium/drivers', - '#src/mesa', - '#src/mesa/main', - '#src/mesa/glapi', - '#src/mesa/math', - '#src/mesa/transform', - '#src/mesa/shader', - '#src/mesa/swrast', - '#src/mesa/swrast_setup', - '#src/egl/main', - '#src/egl/drivers/dri', - ]) - - drienv.ParseConfig('pkg-config --cflags --libs libdrm') - - COMMON_GALLIUM_SOURCES = [ - '#src/mesa/drivers/dri/common/utils.c', - '#src/mesa/drivers/dri/common/vblank.c', - '#src/mesa/drivers/dri/common/dri_util.c', - '#src/mesa/drivers/dri/common/xmlconfig.c', - ] - - COMMON_BM_SOURCES = [ - '#src/mesa/drivers/dri/common/dri_bufmgr.c', - '#src/mesa/drivers/dri/common/dri_drmpool.c', - ] - - Export([ - 'drienv', - 'COMMON_GALLIUM_SOURCES', - 'COMMON_BM_SOURCES', - ]) - - # TODO: Installation - #install: $(LIBNAME) - # $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) - # $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) - - if 'intel' in env['winsys']: - SConscript([ - 'intel/SConscript', - ]) diff --git a/src/gallium/winsys/dri/intel/Makefile b/src/gallium/winsys/dri/intel/Makefile deleted file mode 100644 index e0716ea28e..0000000000 --- a/src/gallium/winsys/dri/intel/Makefile +++ /dev/null @@ -1,34 +0,0 @@ - -TOP = ../../../../.. -include $(TOP)/configs/current - -LIBNAME = i915_dri.so -LIBNAME_EGL = egl_i915_dri.so - -PIPE_DRIVERS = \ - $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ - $(TOP)/src/gallium/winsys/common/intel_drm/libinteldrm.a \ - $(TOP)/src/gallium/drivers/i915simple/libi915simple.a - - -DRIVER_SOURCES = \ - intel_winsys_softpipe.c \ - intel_swapbuffers.c \ - intel_context.c \ - intel_lock.c \ - intel_screen.c - -C_SOURCES = \ - $(COMMON_GALLIUM_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -DRIVER_DEFINES = -I$(TOP)/src/mesa/drivers/dri/intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") - -include ../Makefile.template - -#intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c - -symlinks: diff --git a/src/gallium/winsys/dri/intel/SConscript b/src/gallium/winsys/dri/intel/SConscript deleted file mode 100644 index 6a4f50afcc..0000000000 --- a/src/gallium/winsys/dri/intel/SConscript +++ /dev/null @@ -1,41 +0,0 @@ -Import('*') - -if 'mesa' in env['statetrackers']: - - env = drienv.Clone() - - env.Append(CPPPATH = [ - '../intel', - 'server' - ]) - - #MINIGLX_SOURCES = server/intel_dri.c - - DRIVER_SOURCES = [ - 'intel_winsys_pipe.c', - 'intel_winsys_softpipe.c', - 'intel_winsys_i915.c', - 'intel_batchbuffer.c', - 'intel_swapbuffers.c', - 'intel_context.c', - 'intel_lock.c', - 'intel_screen.c', - 'intel_batchpool.c', - ] - - sources = \ - COMMON_GALLIUM_SOURCES + \ - COMMON_BM_SOURCES + \ - DRIVER_SOURCES - - drivers = [ - softpipe, - i915simple - ] - - # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions - env.SharedLibrary( - target ='i915tex_dri.so', - source = sources, - LIBS = drivers + mesa + auxiliaries + env['LIBS'], - ) diff --git a/src/gallium/winsys/dri/intel/intel_batchbuffer.h b/src/gallium/winsys/dri/intel/intel_batchbuffer.h deleted file mode 100644 index 1fa2719845..0000000000 --- a/src/gallium/winsys/dri/intel/intel_batchbuffer.h +++ /dev/null @@ -1,24 +0,0 @@ -#ifndef INTEL_BATCHBUFFER_H -#define INTEL_BATCHBUFFER_H - -#include "intel_drm/intel_be_batchbuffer.h" - -/* - * Need to redefine the BATCH defines - */ - -#undef BEGIN_BATCH -#define BEGIN_BATCH(dwords, relocs) \ - (i915_batchbuffer_check(&intel->base.batch->base, dwords, relocs)) - -#undef OUT_BATCH -#define OUT_BATCH(d) \ - i915_batchbuffer_dword(&intel->base.batch->base, d) - -#undef OUT_RELOC -#define OUT_RELOC(buf,flags,mask,delta) do { \ - assert((delta) >= 0); \ - intel_be_offset_relocation(intel->base.batch, delta, buf, flags, mask); \ -} while (0) - -#endif diff --git a/src/gallium/winsys/dri/intel/intel_context.c b/src/gallium/winsys/dri/intel/intel_context.c deleted file mode 100644 index 97ef731aaa..0000000000 --- a/src/gallium/winsys/dri/intel/intel_context.c +++ /dev/null @@ -1,337 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "i830_dri.h" - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_swapbuffers.h" -#include "intel_batchbuffer.h" -#include "intel_winsys_softpipe.h" - -#include "i915simple/i915_screen.h" - -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "pipe/p_defines.h" -#include "pipe/p_context.h" - -#include "utils.h" - - -#ifdef DEBUG -int __intel_debug = 0; -#endif - - -#define need_GL_ARB_multisample -#define need_GL_ARB_point_parameters -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_buffer_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_window_pos -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_equation_separate -#define need_GL_EXT_blend_func_separate -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_cull_vertex -#define need_GL_EXT_fog_coord -#define need_GL_EXT_framebuffer_object -#define need_GL_EXT_multi_draw_arrays -#define need_GL_EXT_secondary_color -#define need_GL_NV_vertex_program -#include "extension_helper.h" - - -/** - * Extension strings exported by the intel driver. - * - * \note - * It appears that ARB_texture_env_crossbar has "disappeared" compared to the - * old i830-specific driver. - */ -const struct dri_extension card_extensions[] = { - {"GL_ARB_multisample", GL_ARB_multisample_functions}, - {"GL_ARB_multitexture", NULL}, - {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, - {"GL_ARB_texture_border_clamp", NULL}, - {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, - {"GL_ARB_texture_cube_map", NULL}, - {"GL_ARB_texture_env_add", NULL}, - {"GL_ARB_texture_env_combine", NULL}, - {"GL_ARB_texture_env_dot3", NULL}, - {"GL_ARB_texture_mirrored_repeat", NULL}, - {"GL_ARB_texture_rectangle", NULL}, - {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, - {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, - {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, - {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, - {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, - {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_subtract", NULL}, - {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, - {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, - {"GL_EXT_packed_depth_stencil", NULL}, - {"GL_EXT_pixel_buffer_object", NULL}, - {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, - {"GL_EXT_stencil_wrap", NULL}, - {"GL_EXT_texture_edge_clamp", NULL}, - {"GL_EXT_texture_env_combine", NULL}, - {"GL_EXT_texture_env_dot3", NULL}, - {"GL_EXT_texture_filter_anisotropic", NULL}, - {"GL_EXT_texture_lod_bias", NULL}, - {"GL_3DFX_texture_compression_FXT1", NULL}, - {"GL_APPLE_client_storage", NULL}, - {"GL_MESA_pack_invert", NULL}, - {"GL_MESA_ycbcr_texture", NULL}, - {"GL_NV_blend_square", NULL}, - {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, - {"GL_NV_vertex_program1_1", NULL}, - {"GL_SGIS_generate_mipmap", NULL }, - {NULL, NULL} -}; - - - -#ifdef DEBUG -static const struct dri_debug_control debug_control[] = { - {"ioctl", DEBUG_IOCTL}, - {"bat", DEBUG_BATCH}, - {"lock", DEBUG_LOCK}, - {"swap", DEBUG_SWAP}, - {NULL, 0} -}; -#endif - - - -static void -intel_lock_hardware(struct intel_be_context *context) -{ - struct intel_context *intel = (struct intel_context *)context; - LOCK_HARDWARE(intel); -} - -static void -intel_unlock_hardware(struct intel_be_context *context) -{ - struct intel_context *intel = (struct intel_context *)context; - UNLOCK_HARDWARE(intel); -} - -static boolean -intel_locked_hardware(struct intel_be_context *context) -{ - struct intel_context *intel = (struct intel_context *)context; - return intel->locked ? TRUE : FALSE; -} - -GLboolean -intelCreateContext(const __GLcontextModes * visual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate) -{ - struct intel_context *intel = CALLOC_STRUCT(intel_context); - __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; - struct intel_screen *intelScreen = intel_screen(sPriv); - drmI830Sarea *saPriv = intelScreen->sarea; - int fthrottle_mode; - GLboolean havePools; - struct pipe_context *pipe; - struct st_context *st_share = NULL; - - if (sharedContextPrivate) { - st_share = ((struct intel_context *) sharedContextPrivate)->st; - } - - driContextPriv->driverPrivate = intel; - intel->intelScreen = intelScreen; - intel->driScreen = sPriv; - intel->sarea = saPriv; - - driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, - intel->driScreen->myNum, "i915"); - - - /* - * memory pools - */ - DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - // ZZZ JB should be per screen and not be done per context - havePools = intelCreatePools(sPriv); - DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); - if (!havePools) - return GL_FALSE; - - - /* Dri stuff */ - intel->hHWContext = driContextPriv->hHWContext; - intel->driFd = sPriv->fd; - intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; - - fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); - intel->iw.irq_seq = -1; - intel->irqsEmitted = 0; - - intel->last_swap_fence = NULL; - intel->first_swap_fence = NULL; - -#ifdef DEBUG - __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); -#endif - intel->base.hardware_lock = intel_lock_hardware; - intel->base.hardware_unlock = intel_unlock_hardware; - intel->base.hardware_locked = intel_locked_hardware; - - intel_be_init_context(&intel->base, &intelScreen->base); - - /* - * Pipe-related setup - */ - if (getenv("INTEL_SP")) { - /* use softpipe driver instead of hw */ - pipe = intel_create_softpipe( intel, &intelScreen->base.base ); - } - else { - switch (intel->intelScreen->deviceID) { - case PCI_CHIP_I945_G: - case PCI_CHIP_I945_GM: - case PCI_CHIP_I945_GME: - case PCI_CHIP_G33_G: - case PCI_CHIP_Q33_G: - case PCI_CHIP_Q35_G: - case PCI_CHIP_I915_G: - case PCI_CHIP_I915_GM: - pipe = i915_create_context(intelScreen->base.screen, - &intelScreen->base.base, - &intel->base.base); - break; - default: - fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", - intel->intelScreen->deviceID, __FUNCTION__); - - pipe = intel_create_softpipe( intel, &intelScreen->base.base ); - break; - } - } - - pipe->priv = intel; - - intel->st = st_create_context(pipe, visual, st_share); - - driInitExtensions( intel->st->ctx, card_extensions, GL_TRUE ); - - return GL_TRUE; -} - - -void -intelDestroyContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = intel_context(driContextPriv); - - assert(intel); /* should never be null */ - if (intel) { - st_finish(intel->st); - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(&intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - if (intel->first_swap_fence) { - driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); - driFenceUnReference(&intel->first_swap_fence); - intel->first_swap_fence = NULL; - } - - if (intel->intelScreen->dummyContext == intel) - intel->intelScreen->dummyContext = NULL; - - st_destroy_context(intel->st); - intel_be_destroy_context(&intel->base); - free(intel); - } -} - - -GLboolean -intelUnbindContext(__DRIcontextPrivate * driContextPriv) -{ - struct intel_context *intel = intel_context(driContextPriv); - st_flush(intel->st, PIPE_FLUSH_RENDER_CACHE, NULL); - /* XXX make_current(NULL)? */ - return GL_TRUE; -} - - -GLboolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv) -{ - if (driContextPriv) { - struct intel_context *intel = intel_context(driContextPriv); - struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv); - struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv); - - assert(draw_fb->stfb); - assert(read_fb->stfb); - - /* This is for situations in which we need a rendering context but - * there may not be any currently bound. - */ - intel->intelScreen->dummyContext = intel; - - st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); - - if ((intel->driDrawable != driDrawPriv) || - (intel->lastStamp != driDrawPriv->lastStamp)) { - intel->driDrawable = driDrawPriv; - intelUpdateWindowSize(driDrawPriv); - intel->lastStamp = driDrawPriv->lastStamp; - } - - /* The size of the draw buffer will have been updated above. - * If the readbuffer is a different window, check/update its size now. - */ - if (driReadPriv != driDrawPriv) { - intelUpdateWindowSize(driReadPriv); - } - - } - else { - st_make_current(NULL, NULL, NULL); - } - - return GL_TRUE; -} diff --git a/src/gallium/winsys/dri/intel/intel_context.h b/src/gallium/winsys/dri/intel/intel_context.h deleted file mode 100644 index ced18da143..0000000000 --- a/src/gallium/winsys/dri/intel/intel_context.h +++ /dev/null @@ -1,164 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_CONTEXT_H -#define INTEL_CONTEXT_H - -#include -#include "drm.h" - -#include "pipe/p_debug.h" - -#include "intel_screen.h" -#include "i915_drm.h" - -#include "intel_drm/intel_be_context.h" - - -struct pipe_context; -struct intel_context; -struct _DriBufferObject; -struct st_context; - - -#define INTEL_MAX_FIXUP 64 - -/** - * Intel rendering context, contains a state tracker and intel-specific info. - */ -struct intel_context -{ - struct intel_be_context base; - struct st_context *st; - - struct _DriFenceObject *last_swap_fence; - struct _DriFenceObject *first_swap_fence; - -// struct intel_batchbuffer *batch; - - boolean locked; - char *prevLockFile; - int prevLockLine; - - uint irqsEmitted; - drm_i915_irq_wait_t iw; - - drm_context_t hHWContext; - drmLock *driHwLock; - int driFd; - - __DRIdrawablePrivate *driDrawable; - __DRIscreenPrivate *driScreen; - struct intel_screen *intelScreen; - drmI830Sarea *sarea; - - uint lastStamp; - - /** - * Configuration cache - */ - driOptionCache optionCache; -}; - - - -/** - * Intel framebuffer. - */ -struct intel_framebuffer -{ - struct st_framebuffer *stfb; - - /* other fields TBD */ - int other; -}; - - - - -/* These are functions now: - */ -void LOCK_HARDWARE( struct intel_context *intel ); -void UNLOCK_HARDWARE( struct intel_context *intel ); - -extern char *__progname; - - - -/* ================================================================ - * Debugging: - */ -#ifdef DEBUG -extern int __intel_debug; - -#define DEBUG_SWAP 0x1 -#define DEBUG_LOCK 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_BATCH 0x8 - -#define DBG(flag, ...) do { \ - if (__intel_debug & (DEBUG_##flag)) \ - printf(__VA_ARGS__); \ -} while(0) - -#else -#define DBG(flag, ...) -#endif - - - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GME 0x27AE -#define PCI_CHIP_G33_G 0x29C2 -#define PCI_CHIP_Q35_G 0x29B2 -#define PCI_CHIP_Q33_G 0x29D2 - - -/** Cast wrapper */ -static INLINE struct intel_context * -intel_context(__DRIcontextPrivate *driContextPriv) -{ - return (struct intel_context *) driContextPriv->driverPrivate; -} - - -/** Cast wrapper */ -static INLINE struct intel_framebuffer * -intel_framebuffer(__DRIdrawablePrivate * driDrawPriv) -{ - return (struct intel_framebuffer *) driDrawPriv->driverPrivate; -} - - -#endif diff --git a/src/gallium/winsys/dri/intel/intel_lock.c b/src/gallium/winsys/dri/intel/intel_lock.c deleted file mode 100644 index 406284c98f..0000000000 --- a/src/gallium/winsys/dri/intel/intel_lock.c +++ /dev/null @@ -1,102 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "main/glheader.h" -#include "glapi/glthread.h" -#include -#include "state_tracker/st_public.h" -#include "intel_context.h" -#include "i830_dri.h" - - - -_glthread_DECLARE_STATIC_MUTEX( lockMutex ); - - -static void -intelContendedLock(struct intel_context *intel, uint flags) -{ - __DRIdrawablePrivate *dPriv = intel->driDrawable; - __DRIscreenPrivate *sPriv = intel->driScreen; - struct intel_screen *intelScreen = intel_screen(sPriv); - drmI830Sarea *sarea = intel->sarea; - - drmGetLock(intel->driFd, intel->hHWContext, flags); - - DBG(LOCK, "%s - got contended lock\n", __progname); - - /* If the window moved, may need to set a new cliprect now. - * - * NOTE: This releases and regains the hw lock, so all state - * checking must be done *after* this call: - */ - if (dPriv) - DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); - - if (sarea->width != intelScreen->front.width || - sarea->height != intelScreen->front.height) { - - intelUpdateScreenRotation(sPriv, sarea); - } -} - - -/* Lock the hardware and validate our state. - */ -void LOCK_HARDWARE( struct intel_context *intel ) -{ - char __ret = 0; - - _glthread_LOCK_MUTEX(lockMutex); - assert(!intel->locked); - - DRM_CAS(intel->driHwLock, intel->hHWContext, - (DRM_LOCK_HELD|intel->hHWContext), __ret); - - if (__ret) - intelContendedLock( intel, 0 ); - - DBG(LOCK, "%s - locked\n", __progname); - - intel->locked = 1; -} - - -/* Unlock the hardware using the global current context - */ -void UNLOCK_HARDWARE( struct intel_context *intel ) -{ - assert(intel->locked); - intel->locked = 0; - - DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); - - _glthread_UNLOCK_MUTEX(lockMutex); - - DBG(LOCK, "%s - unlocked\n", __progname); -} diff --git a/src/gallium/winsys/dri/intel/intel_reg.h b/src/gallium/winsys/dri/intel/intel_reg.h deleted file mode 100644 index 4f33bee438..0000000000 --- a/src/gallium/winsys/dri/intel/intel_reg.h +++ /dev/null @@ -1,53 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef _INTEL_REG_H_ -#define _INTEL_REG_H_ - - -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 - -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) - -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) - -#define MI_WAIT_FOR_EVENT ((0x3<<23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) - -#define MI_BATCH_BUFFER_END (0xA<<23) - - -#endif diff --git a/src/gallium/winsys/dri/intel/intel_screen.c b/src/gallium/winsys/dri/intel/intel_screen.c deleted file mode 100644 index b3022fd17a..0000000000 --- a/src/gallium/winsys/dri/intel/intel_screen.c +++ /dev/null @@ -1,607 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "utils.h" -#include "vblank.h" -#include "xmlpool.h" - -#include "intel_context.h" -#include "intel_screen.h" -#include "intel_batchbuffer.h" -#include "intel_swapbuffers.h" - -#include "i830_dri.h" -#include "intel_drm/ws_dri_bufpool.h" - -#include "pipe/p_context.h" -#include "pipe/p_screen.h" -#include "pipe/p_inlines.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_cb_fbo.h" - -static void -intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle); - -static void -intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle) -{ - struct pipe_screen *screen = intelScreen->base.screen; - struct pipe_texture *texture; - struct pipe_texture templat; - struct pipe_surface *surface; - struct pipe_buffer *buffer; - unsigned pitch; - - assert(intelScreen->front.cpp == 4); - - buffer = intel_be_buffer_from_handle(&intelScreen->base, - "front", handle); - - if (!buffer) - return; - - intelScreen->front.buffer = dri_bo(buffer); - - memset(&templat, 0, sizeof(templat)); - templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET; - templat.target = PIPE_TEXTURE_2D; - templat.last_level = 0; - templat.depth[0] = 1; - templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; - templat.width[0] = intelScreen->front.width; - templat.height[0] = intelScreen->front.height; - pf_get_block(templat.format, &templat.block); - pitch = intelScreen->front.pitch; - - texture = screen->texture_blanket(screen, - &templat, - &pitch, - buffer); - - /* Unref the buffer we don't need it anyways */ - pipe_buffer_reference(screen->winsys, &buffer, NULL); - - surface = screen->get_tex_surface(screen, - texture, - 0, - 0, - 0, - PIPE_BUFFER_USAGE_GPU_WRITE); - - intelScreen->front.texture = texture; - intelScreen->front.surface = surface; -} - -PUBLIC const char __driConfigOptions[] = - DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE - DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) - DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) - DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY -// DRI_CONF_FORCE_S3TC_ENABLE(false) - DRI_CONF_ALLOW_LARGE_TEXTURES(1) - DRI_CONF_SECTION_END DRI_CONF_END; - -const uint __driNConfigOptions = 3; - -#ifdef USE_NEW_INTERFACE -static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; -#endif /*USE_NEW_INTERFACE */ - -extern const struct dri_extension card_extensions[]; - - - - -static void -intelPrintDRIInfo(struct intel_screen * intelScreen, - __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) -{ - fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", - intelScreen->front.size, intelScreen->front.offset, - intelScreen->front.pitch); - fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); -} - - -#if 0 -static void -intelPrintSAREA(const drmI830Sarea * sarea) -{ - fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, - sarea->height); - fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); - fprintf(stderr, - "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->front_offset, sarea->front_size, - (unsigned) sarea->front_handle); - fprintf(stderr, - "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->back_offset, sarea->back_size, - (unsigned) sarea->back_handle); - fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->depth_offset, sarea->depth_size, - (unsigned) sarea->depth_handle); - fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", - sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); - fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); - fprintf(stderr, - "SAREA: rotated offset: 0x%08x size: 0x%x\n", - sarea->rotated_offset, sarea->rotated_size); - fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); -} -#endif - - -/** - * Use the information in the sarea to update the screen parameters - * related to screen rotation. Needs to be called locked. - */ -void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) -{ - struct intel_screen *intelScreen = intel_screen(sPriv); - - if (intelScreen->front.map) { - drmUnmap(intelScreen->front.map, intelScreen->front.size); - intelScreen->front.map = NULL; - } - - if (intelScreen->front.buffer) - driDeleteBuffers(1, &intelScreen->front.buffer); - - intelScreen->front.width = sarea->width; - intelScreen->front.height = sarea->height; - intelScreen->front.offset = sarea->front_offset; - intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; - intelScreen->front.size = sarea->front_size; - intelScreen->front.handle = sarea->front_handle; - - assert( sarea->front_size >= - intelScreen->front.pitch * intelScreen->front.height ); - -#if 0 /* JB not important */ - if (!sarea->front_handle) - return; - - if (drmMap(sPriv->fd, - sarea->front_handle, - intelScreen->front.size, - (drmAddress *) & intelScreen->front.map) != 0) { - fprintf(stderr, "drmMap(frontbuffer) failed!\n"); - return; - } -#endif - -#if 0 /* JB */ - if (intelScreen->staticPool) { - driGenBuffers(intelScreen->staticPool, "static region", 1, - &intelScreen->front.buffer, 64, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | - DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); - - driBOSetStatic(intelScreen->front.buffer, - intelScreen->front.offset, - intelScreen->front.pitch * intelScreen->front.height, - intelScreen->front.map, 0); - } -#else - if (intelScreen->base.staticPool) { - if (intelScreen->front.buffer) { - driBOUnReference(intelScreen->front.buffer); - pipe_surface_reference(&intelScreen->front.surface, NULL); - pipe_texture_reference(&intelScreen->front.texture, NULL); - } - intelCreateSurface(intelScreen, &intelScreen->base.base, sarea->front_bo_handle); - } -#endif -} - - -boolean -intelCreatePools(__DRIscreenPrivate * sPriv) -{ - //unsigned batchPoolSize = 1024*1024; - struct intel_screen *intelScreen = intel_screen(sPriv); - - if (intelScreen->havePools) - return GL_TRUE; - - intelScreen->havePools = GL_TRUE; - - intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - return GL_TRUE; -} - -static const char * -intel_get_name( struct pipe_winsys *winsys ) -{ - return "Intel/DRI/ttm"; -} - -/* - * The state tracker (should!) keep track of whether the fake - * frontbuffer has been touched by any rendering since the last time - * we copied its contents to the real frontbuffer. Our task is easy: - */ -static void -intel_flush_frontbuffer( struct pipe_winsys *winsys, - struct pipe_surface *surf, - void *context_private) -{ - struct intel_context *intel = (struct intel_context *) context_private; - __DRIdrawablePrivate *dPriv = intel->driDrawable; - - intelDisplaySurface(dPriv, surf, NULL); -} - -static boolean -intelInitDriver(__DRIscreenPrivate * sPriv) -{ - struct intel_screen *intelScreen; - I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; - - PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = - (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> - getProcAddress("glxEnableExtension")); - void *const psc = sPriv->psc->screenConfigs; - - if (sPriv->devPrivSize != sizeof(I830DRIRec)) { - fprintf(stderr, - "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); - return GL_FALSE; - } - - /* Allocate the private area */ - intelScreen = CALLOC_STRUCT(intel_screen); - if (!intelScreen) - return GL_FALSE; - - /* parse information in __driConfigOptions */ - driParseOptionInfo(&intelScreen->optionCache, - __driConfigOptions, __driNConfigOptions); - - sPriv->private = (void *) intelScreen; - - intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + - gDRIPriv->sarea_priv_offset); - intelScreen->deviceID = gDRIPriv->deviceID; - intelScreen->front.cpp = gDRIPriv->cpp; - intelScreen->drmMinor = sPriv->drmMinor; - - assert(gDRIPriv->bitsPerPixel == 16 || - gDRIPriv->bitsPerPixel == 32); - - intelUpdateScreenRotation(sPriv, intelScreen->sarea); - - if (0) - intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); - - if (glx_enable_extension != NULL) { - (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); - (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); - (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); - (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); - } - - intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; - intelScreen->base.base.get_name = intel_get_name; - intel_be_init_device(&intelScreen->base, sPriv->fd, intelScreen->deviceID); - - return GL_TRUE; -} - - -static void -intelDestroyScreen(__DRIscreenPrivate * sPriv) -{ - struct intel_screen *intelScreen = intel_screen(sPriv); - - intel_be_destroy_device(&intelScreen->base); - /* intelUnmapScreenRegions(intelScreen); */ - - FREE(intelScreen); - sPriv->private = NULL; -} - - -/** - * This is called when we need to set up GL rendering to a new X window. - */ -static boolean -intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, - __DRIdrawablePrivate * driDrawPriv, - const __GLcontextModes * visual, boolean isPixmap) -{ - if (isPixmap) { - return GL_FALSE; /* not implemented */ - } - else { - enum pipe_format colorFormat, depthFormat, stencilFormat; - struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); - - if (!intelfb) - return GL_FALSE; - - if (visual->redBits == 5) - colorFormat = PIPE_FORMAT_R5G6B5_UNORM; - else - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (visual->depthBits == 16) - depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (visual->depthBits == 24) - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - else - depthFormat = PIPE_FORMAT_NONE; - - if (visual->stencilBits == 8) - stencilFormat = PIPE_FORMAT_S8Z24_UNORM; - else - stencilFormat = PIPE_FORMAT_NONE; - - intelfb->stfb = st_create_framebuffer(visual, - colorFormat, - depthFormat, - stencilFormat, - driDrawPriv->w, - driDrawPriv->h, - (void*) intelfb); - if (!intelfb->stfb) { - free(intelfb); - return GL_FALSE; - } - - driDrawPriv->driverPrivate = (void *) intelfb; - return GL_TRUE; - } -} - -static void -intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) -{ - struct intel_framebuffer *intelfb = intel_framebuffer(driDrawPriv); - assert(intelfb->stfb); - st_unreference_framebuffer(&intelfb->stfb); - free(intelfb); -} - - -/** - * Get information about previous buffer swaps. - */ -static int -intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) -{ - if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) - || (sInfo == NULL)) { - return -1; - } - - return 0; -} - - -static void -intelSetTexOffset(__DRIcontext *pDRICtx, int texname, - unsigned long long offset, int depth, uint pitch) -{ - abort(); -#if 0 - struct intel_context *intel = (struct intel_context*) - ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; - struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); - struct st_texture_object *stObj = st_texture_object(tObj); - - if (!stObj) - return; - - if (stObj->pt) - st->pipe->texture_release(intel->st->pipe, &stObj->pt); - - stObj->imageOverride = GL_TRUE; - stObj->depthOverride = depth; - stObj->pitchOverride = pitch; - - if (offset) - stObj->textureOffset = offset; -#endif -} - - -static const struct __DriverAPIRec intelAPI = { - .InitDriver = intelInitDriver, - .DestroyScreen = intelDestroyScreen, - .CreateContext = intelCreateContext, - .DestroyContext = intelDestroyContext, - .CreateBuffer = intelCreateBuffer, - .DestroyBuffer = intelDestroyBuffer, - .SwapBuffers = intelSwapBuffers, - .MakeCurrent = intelMakeCurrent, - .UnbindContext = intelUnbindContext, - .GetSwapInfo = intelGetSwapInfo, - .GetMSC = driGetMSC32, - .WaitForMSC = driWaitForMSC32, - .WaitForSBC = NULL, - .SwapBuffersMSC = NULL, - .CopySubBuffer = intelCopySubBuffer, - .setTexOffset = intelSetTexOffset, -}; - - -static __GLcontextModes * -intelFillInModes(unsigned pixel_bits, unsigned depth_bits, - unsigned stencil_bits, boolean have_back_buffer) -{ - __GLcontextModes *modes; - __GLcontextModes *m; - unsigned num_modes; - unsigned depth_buffer_factor; - unsigned back_buffer_factor; - GLenum fb_format; - GLenum fb_type; - - /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't - * support pageflipping at all. - */ - static const GLenum back_buffer_modes[] = { - GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML - }; - - uint8_t depth_bits_array[3]; - uint8_t stencil_bits_array[3]; - uint8_t msaa_samples_array[1]; - - - depth_bits_array[0] = 0; - depth_bits_array[1] = depth_bits; - depth_bits_array[2] = depth_bits; - msaa_samples_array[0] = 0; - - /* Just like with the accumulation buffer, always provide some modes - * with a stencil buffer. It will be a sw fallback, but some apps won't - * care about that. - */ - stencil_bits_array[0] = 0; - stencil_bits_array[1] = 0; - if (depth_bits == 24) - stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; - - stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; - - depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; - back_buffer_factor = (have_back_buffer) ? 3 : 1; - - num_modes = depth_buffer_factor * back_buffer_factor * 4; - - if (pixel_bits == 16) { - fb_format = GL_RGB; - fb_type = GL_UNSIGNED_SHORT_5_6_5; - } - else { - fb_format = GL_BGRA; - fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; - } - - modes = - (*dri_interface->createContextModes) (num_modes, - sizeof(__GLcontextModes)); - m = modes; - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, msaa_samples_array, 1, GLX_TRUE_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - if (!driFillInModes(&m, fb_format, fb_type, - depth_bits_array, stencil_bits_array, - depth_buffer_factor, back_buffer_modes, - back_buffer_factor, msaa_samples_array, 1, GLX_DIRECT_COLOR)) { - fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, - __LINE__); - return NULL; - } - - /* Mark the visual as slow if there are "fake" stencil bits. - */ - for (m = modes; m != NULL; m = m->next) { - if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { - m->visualRating = GLX_SLOW_CONFIG; - } - } - - return modes; -} - - -/** - * This is the bootstrap function for the driver. libGL supplies all of the - * requisite information about the system, and the driver initializes itself. - * This routine also fills in the linked list pointed to by \c driver_modes - * with the \c __GLcontextModes that the driver can support for windows or - * pbuffers. - * - * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on - * failure. - */ -PUBLIC void * -__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, - __DRIscreen * psc, - const __GLcontextModes * modes, - const __DRIversion * ddx_version, - const __DRIversion * dri_version, - const __DRIversion * drm_version, - const __DRIframebuffer * frame_buffer, - drmAddress pSAREA, int fd, - int internal_api_version, - const __DRIinterfaceMethods * interface, - __GLcontextModes ** driver_modes) -{ - __DRIscreenPrivate *psp; - static const __DRIversion ddx_expected = { 1, 7, 0 }; - static const __DRIversion dri_expected = { 4, 0, 0 }; - static const __DRIversion drm_expected = { 1, 7, 0 }; - - dri_interface = interface; - - if (!driCheckDriDdxDrmVersions2("i915", - dri_version, &dri_expected, - ddx_version, &ddx_expected, - drm_version, &drm_expected)) { - return NULL; - } - - psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, - ddx_version, dri_version, drm_version, - frame_buffer, pSAREA, fd, - internal_api_version, &intelAPI); - - if (psp != NULL) { - I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; - *driver_modes = intelFillInModes(dri_priv->cpp * 8, - (dri_priv->cpp == 2) ? 16 : 24, - (dri_priv->cpp == 2) ? 0 : 8, 1); - - /* Calling driInitExtensions here, with a NULL context pointer, - * does not actually enable the extensions. It just makes sure - * that all the dispatch offsets for all the extensions that - * *might* be enables are known. This is needed because the - * dispatch offsets need to be known when _mesa_context_create - * is called, but we can't enable the extensions until we have a - * context pointer. - * - * Hello chicken. Hello egg. How are you two today? - */ - driInitExtensions(NULL, card_extensions, GL_FALSE); - } - - return (void *) psp; -} - diff --git a/src/gallium/winsys/dri/intel/intel_screen.h b/src/gallium/winsys/dri/intel/intel_screen.h deleted file mode 100644 index e62f9e71ec..0000000000 --- a/src/gallium/winsys/dri/intel/intel_screen.h +++ /dev/null @@ -1,122 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef _INTEL_SCREEN_H_ -#define _INTEL_SCREEN_H_ - -#include "dri_util.h" -#include "i830_common.h" -#include "xmlconfig.h" -#include "intel_drm/ws_dri_bufpool.h" - -#include "pipe/p_compiler.h" - -#include "intel_drm/intel_be_device.h" - -struct intel_screen -{ - struct intel_be_device base; - - struct { - drm_handle_t handle; - - /* We create a static dri buffer for the frontbuffer. - */ - struct _DriBufferObject *buffer; - struct pipe_surface *surface; - struct pipe_texture *texture; - - char *map; /* memory map */ - int offset; /* from start of video mem, in bytes */ - int pitch; /* row stride, in bytes */ - int width; - int height; - int size; - int cpp; /* for front and back buffers */ - } front; - - int deviceID; - int drmMinor; - - drmI830Sarea *sarea; - - /** - * Configuration cache with default values for all contexts - */ - driOptionCache optionCache; - - boolean havePools; - - /** - * Temporary(?) context to use for SwapBuffers or other situations in - * which we need a rendering context, but none is currently bound. - */ - struct intel_context *dummyContext; - - /* - * New stuff form the i915tex integration - */ - unsigned batch_id; - - - struct pipe_winsys *winsys; -}; - - - -/** cast wrapper */ -static INLINE struct intel_screen * -intel_screen(__DRIscreenPrivate *sPriv) -{ - return (struct intel_screen *) sPriv->private; -} - - -extern void -intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); - - -extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); - -extern boolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); - -extern boolean -intelMakeCurrent(__DRIcontextPrivate * driContextPriv, - __DRIdrawablePrivate * driDrawPriv, - __DRIdrawablePrivate * driReadPriv); - - -extern boolean -intelCreatePools(__DRIscreenPrivate *sPriv); - -extern boolean -intelCreateContext(const __GLcontextModes * visual, - __DRIcontextPrivate * driContextPriv, - void *sharedContextPrivate); - - -#endif diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.c b/src/gallium/winsys/dri/intel/intel_swapbuffers.c deleted file mode 100644 index f751f97524..0000000000 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.c +++ /dev/null @@ -1,261 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_screen.h" -#include "intel_context.h" -#include "intel_swapbuffers.h" - -#include "intel_reg.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_cb_fbo.h" - -#include "intel_drm/ws_dri_bufmgr.h" -#include "intel_batchbuffer.h" - -/** - * Display a colorbuffer surface in an X window. - * Used for SwapBuffers and flushing front buffer rendering. - * - * \param dPriv the window/drawable to display into - * \param surf the surface to display - * \param rect optional subrect of surface to display (may be NULL). - */ -void -intelDisplaySurface(__DRIdrawablePrivate *dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t *rect) -{ - struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); - struct intel_context *intel = intelScreen->dummyContext; - - DBG(SWAP, "%s\n", __FUNCTION__); - - if (!intel) { - /* XXX this is where some kind of extra/meta context could be useful */ - return; - } - - if (intel->last_swap_fence) { - driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); - driFenceUnReference(&intel->last_swap_fence); - intel->last_swap_fence = NULL; - } - intel->last_swap_fence = intel->first_swap_fence; - intel->first_swap_fence = NULL; - - /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets - * should work regardless. - */ - LOCK_HARDWARE(intel); - /* if this drawable isn't currently bound the LOCK_HARDWARE done on the - * current context (which is what intelScreenContext should return) might - * not get a contended lock and thus cliprects not updated (tests/manywin) - */ - if (intel_context(dPriv->driContextPriv) != intel) - DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); - - - if (dPriv && dPriv->numClipRects) { - const int srcWidth = surf->width; - const int srcHeight = surf->height; - const int nbox = dPriv->numClipRects; - const drm_clip_rect_t *pbox = dPriv->pClipRects; - const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; - const int cpp = intelScreen->front.cpp; - const int srcpitch = surf->stride / cpp; - int BR13, CMD; - int i; - - ASSERT(surf->buffer); - ASSERT(surf->cpp == cpp); - - DBG(SWAP, "screen pitch %d src surface pitch %d\n", - pitch, surf->stride); - - if (cpp == 2) { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); - CMD = XY_SRC_COPY_BLT_CMD; - } - else { - BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - } - - for (i = 0; i < nbox; i++, pbox++) { - drm_clip_rect_t box; - drm_clip_rect_t sbox; - - if (pbox->x1 > pbox->x2 || - pbox->y1 > pbox->y2 || - pbox->x2 > intelScreen->front.width || - pbox->y2 > intelScreen->front.height) { - /* invalid cliprect, skip it */ - continue; - } - - box = *pbox; - - if (rect) { - /* intersect cliprect with user-provided src rect */ - drm_clip_rect_t rrect; - - rrect.x1 = dPriv->x + rect->x1; - rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; - rrect.x2 = rect->x2 + rrect.x1; - rrect.y2 = rect->y2 + rrect.y1; - if (rrect.x1 > box.x1) - box.x1 = rrect.x1; - if (rrect.y1 > box.y1) - box.y1 = rrect.y1; - if (rrect.x2 < box.x2) - box.x2 = rrect.x2; - if (rrect.y2 < box.y2) - box.y2 = rrect.y2; - - if (box.x1 > box.x2 || box.y1 > box.y2) - continue; - } - - /* restrict blit to size of actually rendered area */ - if (box.x2 - box.x1 > srcWidth) - box.x2 = srcWidth + box.x1; - if (box.y2 - box.y1 > srcHeight) - box.y2 = srcHeight + box.y1; - - DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", - box.x1, box.x2, box.y1, box.y2); - - sbox.x1 = box.x1 - dPriv->x; - sbox.y1 = box.y1 - dPriv->y; - - assert(box.x1 < box.x2); - assert(box.y1 < box.y2); - - /* XXX this could be done with pipe->surface_copy() */ - /* XXX should have its own batch buffer */ - if (!BEGIN_BATCH(8, 2)) { - /* - * Since we share this batch buffer with a context - * we can't flush it since that risks a GPU lockup - */ - assert(0); - continue; - } - - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((box.y1 << 16) | box.x1); - OUT_BATCH((box.y2 << 16) | box.x2); - - OUT_RELOC(intelScreen->front.buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - OUT_BATCH((sbox.y1 << 16) | sbox.x1); - OUT_BATCH((srcpitch * cpp) & 0xffff); - OUT_RELOC(dri_bo(surf->buffer), - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - } - - if (intel->first_swap_fence) - driFenceUnReference(&intel->first_swap_fence); - intel->first_swap_fence = intel_be_batchbuffer_flush(intel->base.batch); - } - - UNLOCK_HARDWARE(intel); - - if (intel->lastStamp != dPriv->lastStamp) { - intelUpdateWindowSize(dPriv); - intel->lastStamp = dPriv->lastStamp; - } -} - - - -/** - * This will be called whenever the currently bound window is moved/resized. - */ -void -intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) -{ - struct intel_framebuffer *intelfb = intel_framebuffer(dPriv); - assert(intelfb->stfb); - st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); -} - - - -void -intelSwapBuffers(__DRIdrawablePrivate * dPriv) -{ - struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; - - assert(intel_fb); - assert(intel_fb->stfb); - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, NULL); - st_notify_swapbuffers_complete(intel_fb->stfb); - } -} - - -/** - * Called via glXCopySubBufferMESA() to copy a subrect of the back - * buffer to the front buffer/screen. - */ -void -intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) -{ - struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); - struct pipe_surface *back_surf; - - assert(intel_fb); - assert(intel_fb->stfb); - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, - ST_SURFACE_BACK_LEFT); - if (back_surf) { - drm_clip_rect_t rect; - rect.x1 = x; - rect.y1 = y; - rect.x2 = w; - rect.y2 = h; - - st_notify_swapbuffers(intel_fb->stfb); - intelDisplaySurface(dPriv, back_surf, &rect); - } -} diff --git a/src/gallium/winsys/dri/intel/intel_swapbuffers.h b/src/gallium/winsys/dri/intel/intel_swapbuffers.h deleted file mode 100644 index 46c9bab3af..0000000000 --- a/src/gallium/winsys/dri/intel/intel_swapbuffers.h +++ /dev/null @@ -1,47 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_SWAPBUFFERS_H -#define INTEL_SWAPBUFFERS_H - - -struct pipe_surface; - - -extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, - struct pipe_surface *surf, - const drm_clip_rect_t * rect); - -extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); - -extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, - int x, int y, int w, int h); - -extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); - - -#endif /* INTEL_SWAPBUFFERS_H */ diff --git a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c deleted file mode 100644 index 0d98d16cf1..0000000000 --- a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.c +++ /dev/null @@ -1,82 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, - * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR - * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE - * USE OR OTHER DEALINGS IN THE SOFTWARE. - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * - **************************************************************************/ -/* - * Authors: Keith Whitwell - */ - -#include "intel_context.h" -#include "intel_winsys_softpipe.h" -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_format.h" -#include "softpipe/sp_winsys.h" - - -struct intel_softpipe_winsys { - struct softpipe_winsys sws; - struct intel_context *intel; -}; - -/** - * Return list of surface formats supported by this driver. - */ -static boolean -intel_is_format_supported(struct softpipe_winsys *sws, - enum pipe_format format) -{ - switch(format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - case PIPE_FORMAT_R5G6B5_UNORM: - case PIPE_FORMAT_S8Z24_UNORM: - return TRUE; - default: - return FALSE; - } -} - - -/** - * Create rendering context which uses software rendering. - */ -struct pipe_context * -intel_create_softpipe( struct intel_context *intel, - struct pipe_winsys *winsys ) -{ - struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); - struct pipe_screen *screen = softpipe_create_screen(winsys); - - /* Fill in this struct with callbacks that softpipe will need to - * communicate with the window system, buffer manager, etc. - */ - isws->sws.is_format_supported = intel_is_format_supported; - isws->intel = intel; - - /* Create the softpipe context: - */ - return softpipe_create( screen, winsys, &isws->sws ); -} diff --git a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.h b/src/gallium/winsys/dri/intel/intel_winsys_softpipe.h deleted file mode 100644 index 5fa14cb749..0000000000 --- a/src/gallium/winsys/dri/intel/intel_winsys_softpipe.h +++ /dev/null @@ -1,39 +0,0 @@ -/************************************************************************** - * - * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_SOFTPIPE_H -#define INTEL_SOFTPIPE_H - -struct pipe_winsys; -struct pipe_context; -struct intel_context; - -struct pipe_context * -intel_create_softpipe( struct intel_context *intel, - struct pipe_winsys *winsys ); - -#endif diff --git a/src/gallium/winsys/dri/intel/server/i830_common.h b/src/gallium/winsys/dri/intel/server/i830_common.h deleted file mode 100644 index 3452ddb3c9..0000000000 --- a/src/gallium/winsys/dri/intel/server/i830_common.h +++ /dev/null @@ -1,255 +0,0 @@ -/************************************************************************** - -Copyright 2001 VA Linux Systems Inc., Fremont, California. -Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas. - -All Rights Reserved. - -Permission is hereby granted, free of charge, to any person obtaining a -copy of this software and associated documentation files (the "Software"), -to deal in the Software without restriction, including without limitation -on the rights to use, copy, modify, merge, publish, distribute, sub -license, and/or sell copies of the Software, and to permit persons to whom -the Software is furnished to do so, subject to the following conditions: - -The above copyright notice and this permission notice (including the next -paragraph) shall be included in all copies or substantial portions of the -Software. - -THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR -IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, -FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL -ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, -DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR -OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE -USE OR OTHER DEALINGS IN THE SOFTWARE. - -**************************************************************************/ - - -#ifndef _I830_COMMON_H_ -#define _I830_COMMON_H_ - - -#define I830_NR_TEX_REGIONS 255 /* maximum due to use of chars for next/prev */ -#define I830_LOG_MIN_TEX_REGION_SIZE 14 - - -/* Driver specific DRM command indices - * NOTE: these are not OS specific, but they are driver specific - */ -#define DRM_I830_INIT 0x00 -#define DRM_I830_FLUSH 0x01 -#define DRM_I830_FLIP 0x02 -#define DRM_I830_BATCHBUFFER 0x03 -#define DRM_I830_IRQ_EMIT 0x04 -#define DRM_I830_IRQ_WAIT 0x05 -#define DRM_I830_GETPARAM 0x06 -#define DRM_I830_SETPARAM 0x07 -#define DRM_I830_ALLOC 0x08 -#define DRM_I830_FREE 0x09 -#define DRM_I830_INIT_HEAP 0x0a -#define DRM_I830_CMDBUFFER 0x0b -#define DRM_I830_DESTROY_HEAP 0x0c -#define DRM_I830_SET_VBLANK_PIPE 0x0d -#define DRM_I830_GET_VBLANK_PIPE 0x0e -#define DRM_I830_MMIO 0x10 - -typedef struct { - enum { - I830_INIT_DMA = 0x01, - I830_CLEANUP_DMA = 0x02, - I830_RESUME_DMA = 0x03 - } func; - unsigned int mmio_offset; - int sarea_priv_offset; - unsigned int ring_start; - unsigned int ring_end; - unsigned int ring_size; - unsigned int front_offset; - unsigned int back_offset; - unsigned int depth_offset; - unsigned int w; - unsigned int h; - unsigned int pitch; - unsigned int pitch_bits; - unsigned int back_pitch; - unsigned int depth_pitch; - unsigned int cpp; - unsigned int chipset; -} drmI830Init; - -typedef struct { - drmTextureRegion texList[I830_NR_TEX_REGIONS+1]; - int last_upload; /* last time texture was uploaded */ - int last_enqueue; /* last time a buffer was enqueued */ - int last_dispatch; /* age of the most recently dispatched buffer */ - int ctxOwner; /* last context to upload state */ - /** Last context that used the buffer manager. */ - int texAge; - int pf_enabled; /* is pageflipping allowed? */ - int pf_active; - int pf_current_page; /* which buffer is being displayed? */ - int perf_boxes; /* performance boxes to be displayed */ - int width, height; /* screen size in pixels */ - - drm_handle_t front_handle; - int front_offset; - int front_size; - - drm_handle_t back_handle; - int back_offset; - int back_size; - - drm_handle_t depth_handle; - int depth_offset; - int depth_size; - - drm_handle_t tex_handle; - int tex_offset; - int tex_size; - int log_tex_granularity; - int pitch; - int rotation; /* 0, 90, 180 or 270 */ - int rotated_offset; - int rotated_size; - int rotated_pitch; - int virtualX, virtualY; - - unsigned int front_tiled; - unsigned int back_tiled; - unsigned int depth_tiled; - unsigned int rotated_tiled; - unsigned int rotated2_tiled; - - int planeA_x; - int planeA_y; - int planeA_w; - int planeA_h; - int planeB_x; - int planeB_y; - int planeB_w; - int planeB_h; - - /* Triple buffering */ - drm_handle_t third_handle; - int third_offset; - int third_size; - unsigned int third_tiled; - - /* buffer object handles for the static buffers. May change - * over the lifetime of the client, though it doesn't in our current - * implementation. - */ - unsigned int front_bo_handle; - unsigned int back_bo_handle; - unsigned int third_bo_handle; - unsigned int depth_bo_handle; -} drmI830Sarea; - -/* Flags for perf_boxes - */ -#define I830_BOX_RING_EMPTY 0x1 /* populated by kernel */ -#define I830_BOX_FLIP 0x2 /* populated by kernel */ -#define I830_BOX_WAIT 0x4 /* populated by kernel & client */ -#define I830_BOX_TEXTURE_LOAD 0x8 /* populated by kernel */ -#define I830_BOX_LOST_CONTEXT 0x10 /* populated by client */ - - -typedef struct { - int start; /* agp offset */ - int used; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830BatchBuffer; - -typedef struct { - char *buf; /* agp offset */ - int sz; /* nr bytes in use */ - int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ - int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ - int num_cliprects; /* mulitpass with multiple cliprects? */ - drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ -} drmI830CmdBuffer; - -typedef struct { - int *irq_seq; -} drmI830IrqEmit; - -typedef struct { - int irq_seq; -} drmI830IrqWait; - -typedef struct { - int param; - int *value; -} drmI830GetParam; - -#define I830_PARAM_IRQ_ACTIVE 1 -#define I830_PARAM_ALLOW_BATCHBUFFER 2 - -typedef struct { - int param; - int value; -} drmI830SetParam; - -#define I830_SETPARAM_USE_MI_BATCHBUFFER_START 1 -#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY 2 -#define I830_SETPARAM_ALLOW_BATCHBUFFER 3 - - -/* A memory manager for regions of shared memory: - */ -#define I830_MEM_REGION_AGP 1 - -typedef struct { - int region; - int alignment; - int size; - int *region_offset; /* offset from start of fb or agp */ -} drmI830MemAlloc; - -typedef struct { - int region; - int region_offset; -} drmI830MemFree; - -typedef struct { - int region; - int size; - int start; -} drmI830MemInitHeap; - -typedef struct { - int region; -} drmI830MemDestroyHeap; - -#define DRM_I830_VBLANK_PIPE_A 1 -#define DRM_I830_VBLANK_PIPE_B 2 - -typedef struct { - int pipe; -} drmI830VBlankPipe; - -#define MMIO_READ 0 -#define MMIO_WRITE 1 - -#define MMIO_REGS_IA_PRIMATIVES_COUNT 0 -#define MMIO_REGS_IA_VERTICES_COUNT 1 -#define MMIO_REGS_VS_INVOCATION_COUNT 2 -#define MMIO_REGS_GS_PRIMITIVES_COUNT 3 -#define MMIO_REGS_GS_INVOCATION_COUNT 4 -#define MMIO_REGS_CL_PRIMITIVES_COUNT 5 -#define MMIO_REGS_CL_INVOCATION_COUNT 6 -#define MMIO_REGS_PS_INVOCATION_COUNT 7 -#define MMIO_REGS_PS_DEPTH_COUNT 8 - -typedef struct { - unsigned int read_write:1; - unsigned int reg:31; - void __user *data; -} drmI830MMIO; - -#endif /* _I830_DRM_H_ */ diff --git a/src/gallium/winsys/dri/intel/server/i830_dri.h b/src/gallium/winsys/dri/intel/server/i830_dri.h deleted file mode 100644 index 0d514b6c38..0000000000 --- a/src/gallium/winsys/dri/intel/server/i830_dri.h +++ /dev/null @@ -1,62 +0,0 @@ - -#ifndef _I830_DRI_H -#define _I830_DRI_H - -#include "xf86drm.h" -#include "i830_common.h" - -#define I830_MAX_DRAWABLES 256 - -#define I830_MAJOR_VERSION 1 -#define I830_MINOR_VERSION 7 -#define I830_PATCHLEVEL 2 - -#define I830_REG_SIZE 0x80000 - -typedef struct _I830DRIRec { - drm_handle_t regs; - drmSize regsSize; - - drmSize unused1; /* backbufferSize */ - drm_handle_t unused2; /* backbuffer */ - - drmSize unused3; /* depthbufferSize */ - drm_handle_t unused4; /* depthbuffer */ - - drmSize unused5; /* rotatedSize */ - drm_handle_t unused6; /* rotatedbuffer */ - - drm_handle_t unused7; /* textures */ - int unused8; /* textureSize */ - - drm_handle_t unused9; /* agp_buffers */ - drmSize unused10; /* agp_buf_size */ - - int deviceID; - int width; - int height; - int mem; - int cpp; - int bitsPerPixel; - - int unused11[8]; /* was front/back/depth/rotated offset/pitch */ - - int unused12; /* logTextureGranularity */ - int unused13; /* textureOffset */ - - int irq; - int sarea_priv_offset; -} I830DRIRec, *I830DRIPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830ConfigPrivRec, *I830ConfigPrivPtr; - -typedef struct { - /* Nothing here yet */ - int dummy; -} I830DRIContextRec, *I830DRIContextPtr; - - -#endif diff --git a/src/gallium/winsys/drm/Makefile b/src/gallium/winsys/drm/Makefile new file mode 100644 index 0000000000..f466ce6c3c --- /dev/null +++ b/src/gallium/winsys/drm/Makefile @@ -0,0 +1,38 @@ +# src/mesa/drivers/dri/Makefile + +TOP = ../../../.. + +include $(TOP)/configs/current + + + +default: $(TOP)/$(LIB_DIR) subdirs + + +$(TOP)/$(LIB_DIR): + -mkdir $(TOP)/$(LIB_DIR) + + +subdirs: + @for dir in $(DRI_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +install: + @for dir in $(DRI_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) install) || exit 1 ; \ + fi \ + done + + +clean: + @for dir in $(DRI_DIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE) clean) ; \ + fi \ + done + -rm -f common/*.o diff --git a/src/gallium/winsys/drm/Makefile.template b/src/gallium/winsys/drm/Makefile.template new file mode 100644 index 0000000000..80e817b808 --- /dev/null +++ b/src/gallium/winsys/drm/Makefile.template @@ -0,0 +1,125 @@ +# -*-makefile-*- + +MESA_MODULES = \ + $(TOP)/src/mesa/libmesa.a \ + $(GALLIUM_AUXILIARIES) + +COMMON_GALLIUM_SOURCES = \ + $(TOP)/src/mesa/drivers/dri/common/utils.c \ + $(TOP)/src/mesa/drivers/dri/common/vblank.c \ + $(TOP)/src/mesa/drivers/dri/common/dri_util.c \ + $(TOP)/src/mesa/drivers/dri/common/xmlconfig.c + +COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \ + $(TOP)/src/mesa/drivers/common/driverfuncs.c \ + $(TOP)/src/mesa/drivers/dri/common/texmem.c \ + $(TOP)/src/mesa/drivers/dri/common/drirenderbuffer.c + +COMMON_BM_SOURCES = \ + $(TOP)/src/mesa/drivers/dri/common/dri_bufmgr.c \ + $(TOP)/src/mesa/drivers/dri/common/dri_drmpool.c + + +ifeq ($(WINDOW_SYSTEM),dri) +WINOBJ= +WINLIB= +INCLUDES = $(SHARED_INCLUDES) $(EXPAT_INCLUDES) + +OBJECTS = \ + $(C_SOURCES:.c=.o) \ + $(ASM_SOURCES:.S=.o) + +else +# miniglx +WINOBJ= +WINLIB=-L$(MESA)/src/glx/mini +MINIGLX_INCLUDES = -I$(TOP)/src/glx/mini +INCLUDES = $(MINIGLX_INCLUDES) \ + $(SHARED_INCLUDES) \ + $(PCIACCESS_CFLAGS) + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(MINIGLX_SOURCES:.c=.o) \ + $(ASM_SOURCES:.S=.o) +endif + + +### Include directories +SHARED_INCLUDES = \ + -I. \ + -I$(TOP)/src/mesa/drivers/dri/common \ + -Iserver \ + -I$(TOP)/include \ + -I$(TOP)/include/GL/internal \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/auxiliary \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/src/gallium/winsys/common \ + -I$(TOP)/src/mesa \ + -I$(TOP)/src/mesa/main \ + -I$(TOP)/src/mesa/glapi \ + -I$(TOP)/src/mesa/math \ + -I$(TOP)/src/mesa/transform \ + -I$(TOP)/src/mesa/shader \ + -I$(TOP)/src/mesa/swrast \ + -I$(TOP)/src/mesa/swrast_setup \ + -I$(TOP)/src/egl/main \ + -I$(TOP)/src/egl/drivers/dri \ + $(LIBDRM_CFLAGS) + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + + +##### TARGETS ##### + +default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR)/$(LIBNAME_EGL) + + +$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template + $(TOP)/bin/mklib -noprefix -o $@ \ + $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) + +$(LIBNAME_EGL): $(WINSYS_OBJECTS) $(LIBS) + $(TOP)/bin/mklib -o $(LIBNAME_EGL) \ + -linker "$(CC)" \ + -noprefix \ + $(OBJECTS) $(MKLIB_OPTIONS) $(WINSYS_OBJECTS) $(PIPE_DRIVERS) $(WINOBJ) $(DRI_LIB_DEPS) \ + --whole-archive $(LIBS) $(GALLIUM_AUXILIARIES) --no-whole-archive + +$(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) + $(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR) + +$(TOP)/$(LIB_DIR)/$(LIBNAME_EGL): $(LIBNAME_EGL) + $(INSTALL) $(LIBNAME_EGL) $(TOP)/$(LIB_DIR) + +depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ + $(ASM_SOURCES) 2> /dev/null + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean: + -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) + -rm -f depend depend.bak + + +install: $(LIBNAME) + $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) + $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) + + +include depend diff --git a/src/gallium/winsys/drm/SConscript b/src/gallium/winsys/drm/SConscript new file mode 100644 index 0000000000..aef5210a32 --- /dev/null +++ b/src/gallium/winsys/drm/SConscript @@ -0,0 +1,54 @@ +Import('*') + +if env['dri']: + + drienv = env.Clone() + + drienv.Replace(CPPPATH = [ + '#src/mesa/drivers/dri/common', + '#include', + '#include/GL/internal', + '#src/gallium/include', + '#src/gallium/auxiliary', + '#src/gallium/drivers', + '#src/mesa', + '#src/mesa/main', + '#src/mesa/glapi', + '#src/mesa/math', + '#src/mesa/transform', + '#src/mesa/shader', + '#src/mesa/swrast', + '#src/mesa/swrast_setup', + '#src/egl/main', + '#src/egl/drivers/dri', + ]) + + drienv.ParseConfig('pkg-config --cflags --libs libdrm') + + COMMON_GALLIUM_SOURCES = [ + '#src/mesa/drivers/dri/common/utils.c', + '#src/mesa/drivers/dri/common/vblank.c', + '#src/mesa/drivers/dri/common/dri_util.c', + '#src/mesa/drivers/dri/common/xmlconfig.c', + ] + + COMMON_BM_SOURCES = [ + '#src/mesa/drivers/dri/common/dri_bufmgr.c', + '#src/mesa/drivers/dri/common/dri_drmpool.c', + ] + + Export([ + 'drienv', + 'COMMON_GALLIUM_SOURCES', + 'COMMON_BM_SOURCES', + ]) + + # TODO: Installation + #install: $(LIBNAME) + # $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) + # $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) + + if 'intel' in env['winsys']: + SConscript([ + 'intel/SConscript', + ]) diff --git a/src/gallium/winsys/drm/intel/Makefile b/src/gallium/winsys/drm/intel/Makefile new file mode 100644 index 0000000000..a670ac044d --- /dev/null +++ b/src/gallium/winsys/drm/intel/Makefile @@ -0,0 +1,25 @@ +TOP = ../../../../.. +include $(TOP)/configs/current + + +SUBDIRS = common dri egl + + +default: subdirs + + +subdirs: + @for dir in $(SUBDIRS) ; do \ + if [ -d $$dir ] ; then \ + (cd $$dir && $(MAKE)) || exit 1 ; \ + fi \ + done + + +clean: + rm -f `find . -name \*.[oa]` + rm -f `find . -name depend` + + +# Dummy install target +install: diff --git a/src/gallium/winsys/drm/intel/common/Makefile b/src/gallium/winsys/drm/intel/common/Makefile new file mode 100644 index 0000000000..bf1a7d691f --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/Makefile @@ -0,0 +1,23 @@ +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = inteldrm + +C_SOURCES = \ + intel_be_batchbuffer.c \ + intel_be_context.c \ + intel_be_device.c \ + ws_dri_bufmgr.c \ + ws_dri_drmpool.c \ + ws_dri_fencemgr.c \ + ws_dri_mallocpool.c \ + ws_dri_slabpool.c + + +include ./Makefile.template + +DRIVER_DEFINES = $(shell pkg-config libdrm --cflags \ + && pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") +symlinks: + diff --git a/src/gallium/winsys/drm/intel/common/Makefile.template b/src/gallium/winsys/drm/intel/common/Makefile.template new file mode 100644 index 0000000000..02ed363a43 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/Makefile.template @@ -0,0 +1,64 @@ +# -*-makefile-*- + + +# We still have a dependency on the "dri" buffer manager. Most likely +# the interface can be reused in non-dri environments, and also as a +# frontend to simpler memory managers. +# +COMMON_SOURCES = + +OBJECTS = $(C_SOURCES:.c=.o) \ + $(CPP_SOURCES:.cpp=.o) \ + $(ASM_SOURCES:.S=.o) + + +### Include directories +INCLUDES = \ + -I. \ + -I$(TOP)/src/gallium/include \ + -I$(TOP)/src/gallium/auxiliary \ + -I$(TOP)/src/gallium/drivers \ + -I$(TOP)/include \ + $(DRIVER_INCLUDES) + + +##### RULES ##### + +.c.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.cpp.o: + $(CXX) -c $(INCLUDES) $(CXXFLAGS) $(DRIVER_DEFINES) $< -o $@ + +.S.o: + $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ + + +##### TARGETS ##### + +default: depend symlinks $(LIBNAME) + + +$(LIBNAME): $(OBJECTS) Makefile Makefile.template + $(TOP)/bin/mklib -o $@ -static $(OBJECTS) $(DRIVER_LIBS) + + +depend: $(C_SOURCES) $(CPP_SOURCES) $(ASM_SOURCES) $(SYMLINKS) + rm -f depend + touch depend + $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) $(CPP_SOURCES) \ + $(ASM_SOURCES) 2> /dev/null + + +# Emacs tags +tags: + etags `find . -name \*.[ch]` `find ../include` + + +# Remove .o and backup files +clean:: + -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) + -rm -f depend depend.bak + + +include depend diff --git a/src/gallium/winsys/drm/intel/common/glthread.h b/src/gallium/winsys/drm/intel/common/glthread.h new file mode 100644 index 0000000000..b8e9d5f59b --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/glthread.h @@ -0,0 +1,359 @@ +/* + * Mesa 3-D graphics library + * Version: 6.5.2 + * + * Copyright (C) 1999-2006 Brian Paul All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + + +/* + * Thread support for gl dispatch. + * + * Initial version by John Stone (j.stone@acm.org) (johns@cs.umr.edu) + * and Christoph Poliwoda (poliwoda@volumegraphics.com) + * Revised by Keith Whitwell + * Adapted for new gl dispatcher by Brian Paul + * + * + * + * DOCUMENTATION + * + * This thread module exports the following types: + * _glthread_TSD Thread-specific data area + * _glthread_Thread Thread datatype + * _glthread_Mutex Mutual exclusion lock + * + * Macros: + * _glthread_DECLARE_STATIC_MUTEX(name) Declare a non-local mutex + * _glthread_INIT_MUTEX(name) Initialize a mutex + * _glthread_LOCK_MUTEX(name) Lock a mutex + * _glthread_UNLOCK_MUTEX(name) Unlock a mutex + * + * Functions: + * _glthread_GetID(v) Get integer thread ID + * _glthread_InitTSD() Initialize thread-specific data + * _glthread_GetTSD() Get thread-specific data + * _glthread_SetTSD() Set thread-specific data + * + */ + +/* + * If this file is accidentally included by a non-threaded build, + * it should not cause the build to fail, or otherwise cause problems. + * In general, it should only be included when needed however. + */ + +#ifndef GLTHREAD_H +#define GLTHREAD_H + + +#if defined(USE_MGL_NAMESPACE) +#define _glapi_Dispatch _mglapi_Dispatch +#endif + + + +#if (defined(PTHREADS) || defined(SOLARIS_THREADS) ||\ + defined(WIN32_THREADS) || defined(USE_XTHREADS) || defined(BEOS_THREADS)) \ + && !defined(THREADS) +# define THREADS +#endif + +#ifdef VMS +#include +#endif + +/* + * POSIX threads. This should be your choice in the Unix world + * whenever possible. When building with POSIX threads, be sure + * to enable any compiler flags which will cause the MT-safe + * libc (if one exists) to be used when linking, as well as any + * header macros for MT-safe errno, etc. For Solaris, this is the -mt + * compiler flag. On Solaris with gcc, use -D_REENTRANT to enable + * proper compiling for MT-safe libc etc. + */ +#if defined(PTHREADS) +#include /* POSIX threads headers */ + +typedef struct { + pthread_key_t key; + int initMagic; +} _glthread_TSD; + +typedef pthread_t _glthread_Thread; + +typedef pthread_mutex_t _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) \ + static _glthread_Mutex name = PTHREAD_MUTEX_INITIALIZER + +#define _glthread_INIT_MUTEX(name) \ + pthread_mutex_init(&(name), NULL) + +#define _glthread_DESTROY_MUTEX(name) \ + pthread_mutex_destroy(&(name)) + +#define _glthread_LOCK_MUTEX(name) \ + (void) pthread_mutex_lock(&(name)) + +#define _glthread_UNLOCK_MUTEX(name) \ + (void) pthread_mutex_unlock(&(name)) + +typedef pthread_cond_t _glthread_Cond; + +#define _glthread_DECLARE_STATIC_COND(name) \ + static _glthread_Cond name = PTHREAD_COND_INITIALIZER + +#define _glthread_INIT_COND(cond) \ + pthread_cond_init(&(cond), NULL) + +#define _glthread_DESTROY_COND(name) \ + pthread_cond_destroy(&(name)) + +#define _glthread_COND_WAIT(cond, mutex) \ + pthread_cond_wait(&(cond), &(mutex)) + +#define _glthread_COND_SIGNAL(cond) \ + pthread_cond_signal(&(cond)) + +#define _glthread_COND_BROADCAST(cond) \ + pthread_cond_broadcast(&(cond)) + + +#else /* PTHREADS */ + +typedef unsigned int _glthread_Cond; +#define _glthread_DECLARE_STATIC_COND(name) \ +// #warning Condition variables not implemented. + +#define _glthread_INIT_COND(cond) \ + abort(); + +#define _glthread_DESTROY_COND(name) \ + abort(); + +#define _glthread_COND_WAIT(cond, mutex) \ + abort(); + +#define _glthread_COND_SIGNAL(cond) \ + abort(); + +#define _glthread_COND_BROADCAST(cond) \ + abort(); + +#endif + + +/* + * Solaris threads. Use only up to Solaris 2.4. + * Solaris 2.5 and higher provide POSIX threads. + * Be sure to compile with -mt on the Solaris compilers, or + * use -D_REENTRANT if using gcc. + */ +#ifdef SOLARIS_THREADS +#include + +typedef struct { + thread_key_t key; + mutex_t keylock; + int initMagic; +} _glthread_TSD; + +typedef thread_t _glthread_Thread; + +typedef mutex_t _glthread_Mutex; + +/* XXX need to really implement mutex-related macros */ +#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 +#define _glthread_INIT_MUTEX(name) (void) name +#define _glthread_DESTROY_MUTEX(name) (void) name +#define _glthread_LOCK_MUTEX(name) (void) name +#define _glthread_UNLOCK_MUTEX(name) (void) name + +#endif /* SOLARIS_THREADS */ + + + + +/* + * Windows threads. Should work with Windows NT and 95. + * IMPORTANT: Link with multithreaded runtime library when THREADS are + * used! + */ +#ifdef WIN32_THREADS +#include + +typedef struct { + DWORD key; + int initMagic; +} _glthread_TSD; + +typedef HANDLE _glthread_Thread; + +typedef CRITICAL_SECTION _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) /*static*/ _glthread_Mutex name = {0,0,0,0,0,0} +#define _glthread_INIT_MUTEX(name) InitializeCriticalSection(&name) +#define _glthread_DESTROY_MUTEX(name) DeleteCriticalSection(&name) +#define _glthread_LOCK_MUTEX(name) EnterCriticalSection(&name) +#define _glthread_UNLOCK_MUTEX(name) LeaveCriticalSection(&name) + +#endif /* WIN32_THREADS */ + + + + +/* + * XFree86 has its own thread wrapper, Xthreads.h + * We wrap it again for GL. + */ +#ifdef USE_XTHREADS +#include + +typedef struct { + xthread_key_t key; + int initMagic; +} _glthread_TSD; + +typedef xthread_t _glthread_Thread; + +typedef xmutex_rec _glthread_Mutex; + +#ifdef XMUTEX_INITIALIZER +#define _glthread_DECLARE_STATIC_MUTEX(name) \ + static _glthread_Mutex name = XMUTEX_INITIALIZER +#else +#define _glthread_DECLARE_STATIC_MUTEX(name) \ + static _glthread_Mutex name +#endif + +#define _glthread_INIT_MUTEX(name) \ + xmutex_init(&(name)) + +#define _glthread_DESTROY_MUTEX(name) \ + xmutex_clear(&(name)) + +#define _glthread_LOCK_MUTEX(name) \ + (void) xmutex_lock(&(name)) + +#define _glthread_UNLOCK_MUTEX(name) \ + (void) xmutex_unlock(&(name)) + +#endif /* USE_XTHREADS */ + + + +/* + * BeOS threads. R5.x required. + */ +#ifdef BEOS_THREADS + +#include +#include + +typedef struct { + int32 key; + int initMagic; +} _glthread_TSD; + +typedef thread_id _glthread_Thread; + +/* Use Benaphore, aka speeder semaphore */ +typedef struct { + int32 lock; + sem_id sem; +} benaphore; +typedef benaphore _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = { 0, 0 } +#define _glthread_INIT_MUTEX(name) name.sem = create_sem(0, #name"_benaphore"), name.lock = 0 +#define _glthread_DESTROY_MUTEX(name) delete_sem(name.sem), name.lock = 0 +#define _glthread_LOCK_MUTEX(name) if (name.sem == 0) _glthread_INIT_MUTEX(name); \ + if (atomic_add(&(name.lock), 1) >= 1) acquire_sem(name.sem) +#define _glthread_UNLOCK_MUTEX(name) if (atomic_add(&(name.lock), -1) > 1) release_sem(name.sem) + +#endif /* BEOS_THREADS */ + + + +#ifndef THREADS + +/* + * THREADS not defined + */ + +typedef GLuint _glthread_TSD; + +typedef GLuint _glthread_Thread; + +typedef GLuint _glthread_Mutex; + +#define _glthread_DECLARE_STATIC_MUTEX(name) static _glthread_Mutex name = 0 + +#define _glthread_INIT_MUTEX(name) (void) name + +#define _glthread_DESTROY_MUTEX(name) (void) name + +#define _glthread_LOCK_MUTEX(name) (void) name + +#define _glthread_UNLOCK_MUTEX(name) (void) name + +#endif /* THREADS */ + + + +/* + * Platform independent thread specific data API. + */ + +extern unsigned long +_glthread_GetID(void); + + +extern void +_glthread_InitTSD(_glthread_TSD *); + + +extern void * +_glthread_GetTSD(_glthread_TSD *); + + +extern void +_glthread_SetTSD(_glthread_TSD *, void *); + +#if defined(GLX_USE_TLS) + +extern __thread struct _glapi_table * _glapi_tls_Dispatch + __attribute__((tls_model("initial-exec"))); + +#define GET_DISPATCH() _glapi_tls_Dispatch + +#elif !defined(GL_CALL) +# if defined(THREADS) +# define GET_DISPATCH() \ + ((__builtin_expect( _glapi_Dispatch != NULL, 1 )) \ + ? _glapi_Dispatch : _glapi_get_dispatch()) +# else +# define GET_DISPATCH() _glapi_Dispatch +# endif /* defined(THREADS) */ +#endif /* ndef GL_CALL */ + + +#endif /* THREADS_H */ diff --git a/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c new file mode 100644 index 0000000000..bc13a5761e --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.c @@ -0,0 +1,429 @@ + +#include "intel_be_batchbuffer.h" +#include "intel_be_context.h" +#include "intel_be_device.h" +#include + +#include "xf86drm.h" + +static void +intel_realloc_relocs(struct intel_be_batchbuffer *batch, int num_relocs) +{ + unsigned long size = num_relocs * I915_RELOC0_STRIDE + I915_RELOC_HEADER; + + size *= sizeof(uint32_t); + batch->reloc = realloc(batch->reloc, size); + batch->reloc_size = num_relocs; +} + + +void +intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch) +{ + /* + * Get a new, free batchbuffer. + */ + drmBO *bo; + struct drm_bo_info_req *req; + + driBOUnrefUserList(batch->list); + driBOResetList(batch->list); + + /* base.size is the size available to the i915simple driver */ + batch->base.size = batch->device->max_batch_size - BATCH_RESERVED; + batch->base.actual_size = batch->device->max_batch_size; + driBOData(batch->buffer, batch->base.actual_size, NULL, NULL, 0); + + /* + * Add the batchbuffer to the validate list. + */ + + driBOAddListItem(batch->list, batch->buffer, + DRM_BO_FLAG_EXE | DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_EXE | DRM_BO_MASK_MEM, + &batch->dest_location, &batch->node); + + req = &batch->node->bo_arg.d.req.bo_req; + + /* + * Set up information needed for us to make relocations + * relative to the underlying drm buffer objects. + */ + + driReadLockKernelBO(); + bo = driBOKernel(batch->buffer); + req->presumed_offset = (uint64_t) bo->offset; + req->hint = DRM_BO_HINT_PRESUMED_OFFSET; + batch->drmBOVirtual = (uint8_t *) bo->virtual; + driReadUnlockKernelBO(); + + /* + * Adjust the relocation buffer size. + */ + + if (batch->reloc_size > INTEL_MAX_RELOCS || + batch->reloc == NULL) + intel_realloc_relocs(batch, INTEL_DEFAULT_RELOCS); + + assert(batch->reloc != NULL); + batch->reloc[0] = 0; /* No relocs yet. */ + batch->reloc[1] = 1; /* Reloc type 1 */ + batch->reloc[2] = 0; /* Only a single relocation list. */ + batch->reloc[3] = 0; /* Only a single relocation list. */ + + batch->base.map = driBOMap(batch->buffer, DRM_BO_FLAG_WRITE, 0); + batch->poolOffset = driBOPoolOffset(batch->buffer); + batch->base.ptr = batch->base.map; + batch->dirty_state = ~0; + batch->nr_relocs = 0; + batch->flags = 0; + batch->id = 0;//batch->intel->intelScreen->batch_id++; +} + +/*====================================================================== + * Public functions + */ +struct intel_be_batchbuffer * +intel_be_batchbuffer_alloc(struct intel_be_context *intel) +{ + struct intel_be_batchbuffer *batch = calloc(sizeof(*batch), 1); + + batch->intel = intel; + batch->device = intel->device; + + driGenBuffers(intel->device->batchPool, "batchbuffer", 1, + &batch->buffer, 4096, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_EXE, 0); + batch->last_fence = NULL; + batch->list = driBOCreateList(20); + batch->reloc = NULL; + intel_be_batchbuffer_reset(batch); + return batch; +} + +void +intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch) +{ + if (batch->last_fence) { + driFenceFinish(batch->last_fence, + DRM_FENCE_TYPE_EXE, FALSE); + driFenceUnReference(&batch->last_fence); + } + if (batch->base.map) { + driBOUnmap(batch->buffer); + batch->base.map = NULL; + } + driBOUnReference(batch->buffer); + driBOFreeList(batch->list); + if (batch->reloc) + free(batch->reloc); + batch->buffer = NULL; + free(batch); +} + +void +intel_be_offset_relocation(struct intel_be_batchbuffer *batch, + unsigned pre_add, + struct _DriBufferObject *driBO, + uint64_t val_flags, + uint64_t val_mask) +{ + int itemLoc; + struct _drmBONode *node; + uint32_t *reloc; + struct drm_bo_info_req *req; + + driBOAddListItem(batch->list, driBO, val_flags, val_mask, + &itemLoc, &node); + req = &node->bo_arg.d.req.bo_req; + + if (!(req->hint & DRM_BO_HINT_PRESUMED_OFFSET)) { + + /* + * Stop other threads from tampering with the underlying + * drmBO while we're reading its offset. + */ + + driReadLockKernelBO(); + req->presumed_offset = (uint64_t) driBOKernel(driBO)->offset; + driReadUnlockKernelBO(); + req->hint = DRM_BO_HINT_PRESUMED_OFFSET; + } + + pre_add += driBOPoolOffset(driBO); + + if (batch->nr_relocs == batch->reloc_size) + intel_realloc_relocs(batch, batch->reloc_size * 2); + + reloc = batch->reloc + + (I915_RELOC_HEADER + batch->nr_relocs * I915_RELOC0_STRIDE); + + reloc[0] = ((uint8_t *)batch->base.ptr - batch->drmBOVirtual); + i915_batchbuffer_dword(&batch->base, req->presumed_offset + pre_add); + reloc[1] = pre_add; + reloc[2] = itemLoc; + reloc[3] = batch->dest_location; + batch->nr_relocs++; +} + +static void +i915_drm_copy_reply(const struct drm_bo_info_rep * rep, drmBO * buf) +{ + buf->handle = rep->handle; + buf->flags = rep->flags; + buf->size = rep->size; + buf->offset = rep->offset; + buf->mapHandle = rep->arg_handle; + buf->proposedFlags = rep->proposed_flags; + buf->start = rep->buffer_start; + buf->fenceFlags = rep->fence_flags; + buf->replyFlags = rep->rep_flags; + buf->pageAlignment = rep->page_alignment; +} + +static int +i915_execbuf(struct intel_be_batchbuffer *batch, + unsigned int used, + boolean ignore_cliprects, + drmBOList *list, + struct drm_i915_execbuffer *ea) +{ +// struct intel_be_context *intel = batch->intel; + drmBONode *node; + drmMMListHead *l; + struct drm_i915_op_arg *arg, *first; + struct drm_bo_op_req *req; + struct drm_bo_info_rep *rep; + uint64_t *prevNext = NULL; + drmBO *buf; + int ret = 0; + uint32_t count = 0; + + first = NULL; + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + + arg = &node->bo_arg; + req = &arg->d.req; + + if (!first) + first = arg; + + if (prevNext) + *prevNext = (unsigned long)arg; + + prevNext = &arg->next; + req->bo_req.handle = node->buf->handle; + req->op = drm_bo_validate; + req->bo_req.flags = node->arg0; + req->bo_req.mask = node->arg1; + req->bo_req.hint |= 0; + count++; + } + + memset(ea, 0, sizeof(*ea)); + ea->num_buffers = count; + ea->batch.start = batch->poolOffset; + ea->batch.used = used; +#if 0 /* ZZZ JB: no cliprects used */ + ea->batch.cliprects = intel->pClipRects; + ea->batch.num_cliprects = ignore_cliprects ? 0 : intel->numClipRects; + ea->batch.DR1 = 0; + ea->batch.DR4 = 0;((((GLuint) intel->drawX) & 0xffff) | + (((GLuint) intel->drawY) << 16)); +#else + ea->batch.cliprects = NULL; + ea->batch.num_cliprects = 0; + ea->batch.DR1 = 0; + ea->batch.DR4 = 0; +#endif + ea->fence_arg.flags = DRM_I915_FENCE_FLAG_FLUSHED; + ea->ops_list = (unsigned long) first; + first->reloc_ptr = (unsigned long) batch->reloc; + batch->reloc[0] = batch->nr_relocs; + + //return -EFAULT; + do { + ret = drmCommandWriteRead(batch->device->fd, DRM_I915_EXECBUFFER, ea, + sizeof(*ea)); + } while (ret == -EAGAIN); + + if (ret != 0) + return ret; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + arg = &node->bo_arg; + rep = &arg->d.rep.bo_info; + + if (!arg->handled) { + return -EFAULT; + } + if (arg->d.rep.ret) + return arg->d.rep.ret; + + buf = node->buf; + i915_drm_copy_reply(rep, buf); + } + return 0; +} + +/* TODO: Push this whole function into bufmgr. + */ +static struct _DriFenceObject * +do_flush_locked(struct intel_be_batchbuffer *batch, + unsigned int used, + boolean ignore_cliprects, boolean allow_unlock) +{ + struct intel_be_context *intel = batch->intel; + struct _DriFenceObject *fo; + drmFence fence; + drmBOList *boList; + struct drm_i915_execbuffer ea; + int ret = 0; + + driBOValidateUserList(batch->list); + boList = driGetdrmBOList(batch->list); + +#if 0 /* ZZZ JB Allways run */ + if (!(intel->numClipRects == 0 && !ignore_cliprects)) { +#else + if (1) { +#endif + ret = i915_execbuf(batch, used, ignore_cliprects, boList, &ea); + } else { + driPutdrmBOList(batch->list); + fo = NULL; + goto out; + } + driPutdrmBOList(batch->list); + if (ret) + abort(); + + if (ea.fence_arg.error != 0) { + + /* + * The hardware has been idled by the kernel. + * Don't fence the driBOs. + */ + + if (batch->last_fence) + driFenceUnReference(&batch->last_fence); +#if 0 /* ZZZ JB: no _mesa_* funcs in gallium */ + _mesa_printf("fence error\n"); +#endif + batch->last_fence = NULL; + fo = NULL; + goto out; + } + + fence.handle = ea.fence_arg.handle; + fence.fence_class = ea.fence_arg.fence_class; + fence.type = ea.fence_arg.type; + fence.flags = ea.fence_arg.flags; + fence.signaled = ea.fence_arg.signaled; + + fo = driBOFenceUserList(batch->device->fenceMgr, batch->list, + "SuperFence", &fence); + + if (driFenceType(fo) & DRM_I915_FENCE_TYPE_RW) { + if (batch->last_fence) + driFenceUnReference(&batch->last_fence); + /* + * FIXME: Context last fence?? + */ + batch->last_fence = fo; + driFenceReference(fo); + } + out: +#if 0 /* ZZZ JB: fix this */ + intel->vtbl.lost_hardware(intel); +#else + (void)intel; +#endif + return fo; +} + + +struct _DriFenceObject * +intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch) +{ + struct intel_be_context *intel = batch->intel; + unsigned int used = batch->base.ptr - batch->base.map; + boolean was_locked = batch->intel->hardware_locked(intel); + struct _DriFenceObject *fence; + + if (used == 0) { + driFenceReference(batch->last_fence); + return batch->last_fence; + } + + /* Add the MI_BATCH_BUFFER_END. Always add an MI_FLUSH - this is a + * performance drain that we would like to avoid. + */ +#if 0 /* ZZZ JB: what should we do here? */ + if (used & 4) { + ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->base.ptr)[1] = 0; + ((int *) batch->base.ptr)[2] = MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->base.ptr)[0] = intel->vtbl.flush_cmd(); + ((int *) batch->base.ptr)[1] = MI_BATCH_BUFFER_END; + used += 8; + } +#else + if (used & 4) { + ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->base.ptr)[1] = 0; + ((int *) batch->base.ptr)[2] = (0xA<<23); // MI_BATCH_BUFFER_END; + used += 12; + } + else { + ((int *) batch->base.ptr)[0] = ((0<<29)|(4<<23)); // MI_FLUSH; + ((int *) batch->base.ptr)[1] = (0xA<<23); // MI_BATCH_BUFFER_END; + used += 8; + } +#endif + driBOUnmap(batch->buffer); + batch->base.ptr = NULL; + batch->base.map = NULL; + + /* TODO: Just pass the relocation list and dma buffer up to the + * kernel. + */ + if (!was_locked) + intel->hardware_lock(intel); + + fence = do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS), + FALSE); + + if (!was_locked) + intel->hardware_unlock(intel); + + /* Reset the buffer: + */ + intel_be_batchbuffer_reset(batch); + return fence; +} + +void +intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch) +{ + struct _DriFenceObject *fence = intel_be_batchbuffer_flush(batch); + driFenceFinish(fence, driFenceType(fence), FALSE); + driFenceUnReference(&fence); +} + +#if 0 +void +intel_be_batchbuffer_data(struct intel_be_batchbuffer *batch, + const void *data, unsigned int bytes, unsigned int flags) +{ + assert((bytes & 3) == 0); + intel_batchbuffer_require_space(batch, bytes, flags); + memcpy(batch->base.ptr, data, bytes); + batch->base.ptr += bytes; +} +#endif diff --git a/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h new file mode 100644 index 0000000000..f150e3a674 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/intel_be_batchbuffer.h @@ -0,0 +1,69 @@ + +#ifndef INTEL_BE_BATCHBUFFER_H +#define INTEL_BE_BATCHBUFFER_H + +#include "i915simple/i915_batch.h" + +#include "ws_dri_bufmgr.h" + +#define BATCH_RESERVED 16 + +#define INTEL_DEFAULT_RELOCS 100 +#define INTEL_MAX_RELOCS 400 + +#define INTEL_BATCH_NO_CLIPRECTS 0x1 +#define INTEL_BATCH_CLIPRECTS 0x2 + +struct intel_be_context; +struct intel_be_device; + +struct intel_be_batchbuffer +{ + struct i915_batchbuffer base; + + struct intel_be_context *intel; + struct intel_be_device *device; + + struct _DriBufferObject *buffer; + struct _DriFenceObject *last_fence; + uint32_t flags; + + struct _DriBufferList *list; + size_t list_count; + + uint32_t *reloc; + size_t reloc_size; + size_t nr_relocs; + + uint32_t dirty_state; + uint32_t id; + + uint32_t poolOffset; + uint8_t *drmBOVirtual; + struct _drmBONode *node; /* Validation list node for this buffer */ + int dest_location; /* Validation list sequence for this buffer */ +}; + +struct intel_be_batchbuffer * +intel_be_batchbuffer_alloc(struct intel_be_context *intel); + +void +intel_be_batchbuffer_free(struct intel_be_batchbuffer *batch); + +void +intel_be_batchbuffer_finish(struct intel_be_batchbuffer *batch); + +struct _DriFenceObject * +intel_be_batchbuffer_flush(struct intel_be_batchbuffer *batch); + +void +intel_be_batchbuffer_reset(struct intel_be_batchbuffer *batch); + +void +intel_be_offset_relocation(struct intel_be_batchbuffer *batch, + unsigned pre_add, + struct _DriBufferObject *driBO, + uint64_t val_flags, + uint64_t val_mask); + +#endif diff --git a/src/gallium/winsys/drm/intel/common/intel_be_context.c b/src/gallium/winsys/drm/intel/common/intel_be_context.c new file mode 100644 index 0000000000..1af39674f4 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/intel_be_context.c @@ -0,0 +1,107 @@ + +/* + * Authors: Jakob Bornecrantz + */ + +#include "ws_dri_fencemgr.h" +#include "intel_be_device.h" +#include "intel_be_context.h" +#include "intel_be_batchbuffer.h" + +static INLINE struct intel_be_context * +intel_be_context(struct i915_winsys *sws) +{ + return (struct intel_be_context *)sws; +} + +/* Simple batchbuffer interface: + */ + +static struct i915_batchbuffer* +intel_i915_batch_get(struct i915_winsys *sws) +{ + struct intel_be_context *intel = intel_be_context(sws); + return &intel->batch->base; +} + +static void intel_i915_batch_reloc(struct i915_winsys *sws, + struct pipe_buffer *buf, + unsigned access_flags, + unsigned delta) +{ + struct intel_be_context *intel = intel_be_context(sws); + + unsigned flags = DRM_BO_FLAG_MEM_TT; + unsigned mask = DRM_BO_MASK_MEM; + + if (access_flags & I915_BUFFER_ACCESS_WRITE) { + flags |= DRM_BO_FLAG_WRITE; + mask |= DRM_BO_FLAG_WRITE; + } + + if (access_flags & I915_BUFFER_ACCESS_READ) { + flags |= DRM_BO_FLAG_READ; + mask |= DRM_BO_FLAG_READ; + } + + intel_be_offset_relocation(intel->batch, + delta, + dri_bo(buf), + flags, + mask); +} + +static void intel_i915_batch_flush(struct i915_winsys *sws, + struct pipe_fence_handle **fence) +{ + struct intel_be_context *intel = intel_be_context(sws); + + union { + struct _DriFenceObject *dri; + struct pipe_fence_handle *pipe; + } fu; + + if (fence) + assert(!*fence); + + fu.dri = intel_be_batchbuffer_flush(intel->batch); + + if (!fu.dri) { + assert(0); + *fence = NULL; + return; + } + + if (fu.dri) { + if (fence) + *fence = fu.pipe; + else + driFenceUnReference(&fu.dri); + } + +} + +boolean +intel_be_init_context(struct intel_be_context *intel, struct intel_be_device *device) +{ + assert(intel); + assert(device); + + intel->device = device; + + /* TODO move framebuffer createion to the driver */ + + intel->base.batch_get = intel_i915_batch_get; + intel->base.batch_reloc = intel_i915_batch_reloc; + intel->base.batch_flush = intel_i915_batch_flush; + + intel->batch = intel_be_batchbuffer_alloc(intel); + + return true; +} + +void +intel_be_destroy_context(struct intel_be_context *intel) +{ + intel_be_batchbuffer_free(intel->batch); +} diff --git a/src/gallium/winsys/drm/intel/common/intel_be_context.h b/src/gallium/winsys/drm/intel/common/intel_be_context.h new file mode 100644 index 0000000000..d5cbc93594 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/intel_be_context.h @@ -0,0 +1,40 @@ +/* These need to be diffrent from the intel winsys */ +#ifndef INTEL_BE_CONTEXT_H +#define INTEL_BE_CONTEXT_H + +#include "i915simple/i915_winsys.h" + +struct intel_be_context +{ + /** Interface to i915simple driver */ + struct i915_winsys base; + + struct intel_be_device *device; + struct intel_be_batchbuffer *batch; + + /* + * Hardware lock functions. + * + * Needs to be filled in by the winsys. + */ + void (*hardware_lock)(struct intel_be_context *context); + void (*hardware_unlock)(struct intel_be_context *context); + boolean (*hardware_locked)(struct intel_be_context *context); +}; + +/** + * Intialize a allocated intel_be_context struct. + * + * Remember to set the hardware_* functions. + */ +boolean +intel_be_init_context(struct intel_be_context *intel, + struct intel_be_device *device); + +/** + * Destroy a intel_be_context. + * Does not free the struct that is up to the winsys. + */ +void +intel_be_destroy_context(struct intel_be_context *intel); +#endif diff --git a/src/gallium/winsys/drm/intel/common/intel_be_device.c b/src/gallium/winsys/drm/intel/common/intel_be_device.c new file mode 100644 index 0000000000..8db0329615 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/intel_be_device.c @@ -0,0 +1,308 @@ + + +/* + * Authors: Keith Whitwell + * Jakob Bornecrantz + */ + +#include "intel_be_device.h" +#include "ws_dri_bufmgr.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" + +#include "pipe/p_winsys.h" +#include "pipe/p_defines.h" +#include "pipe/p_state.h" +#include "pipe/p_util.h" +#include "pipe/p_inlines.h" + +#include "i915simple/i915_screen.h" + +/* Turn a pipe winsys into an intel/pipe winsys: + */ +static INLINE struct intel_be_device * +intel_be_device( struct pipe_winsys *winsys ) +{ + return (struct intel_be_device *)winsys; +} + + +/* + * Buffer functions. + * + * Most callbacks map direcly onto dri_bufmgr operations: + */ + +static void *intel_be_buffer_map(struct pipe_winsys *winsys, + struct pipe_buffer *buf, + unsigned flags ) +{ + unsigned drm_flags = 0; + + if (flags & PIPE_BUFFER_USAGE_CPU_WRITE) + drm_flags |= DRM_BO_FLAG_WRITE; + + if (flags & PIPE_BUFFER_USAGE_CPU_READ) + drm_flags |= DRM_BO_FLAG_READ; + + return driBOMap( dri_bo(buf), drm_flags, 0 ); +} + +static void intel_be_buffer_unmap(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + driBOUnmap( dri_bo(buf) ); +} + +static void +intel_be_buffer_destroy(struct pipe_winsys *winsys, + struct pipe_buffer *buf) +{ + driBOUnReference( dri_bo(buf) ); + FREE(buf); +} + +static struct pipe_buffer * +intel_be_buffer_create(struct pipe_winsys *winsys, + unsigned alignment, + unsigned usage, + unsigned size ) +{ + struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer ); + struct intel_be_device *iws = intel_be_device(winsys); + unsigned flags = 0; + struct _DriBufferPool *pool; + + buffer->base.refcount = 1; + buffer->base.alignment = alignment; + buffer->base.usage = usage; + buffer->base.size = size; + + if (usage & (PIPE_BUFFER_USAGE_VERTEX | PIPE_BUFFER_USAGE_CONSTANT)) { + flags |= DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; + pool = iws->mallocPool; + } else if (usage & PIPE_BUFFER_USAGE_CUSTOM) { + /* For vertex buffers */ + flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; + pool = iws->vertexPool; + } else { + flags |= DRM_BO_FLAG_MEM_VRAM | DRM_BO_FLAG_MEM_TT; + pool = iws->regionPool; + } + + if (usage & PIPE_BUFFER_USAGE_GPU_READ) + flags |= DRM_BO_FLAG_READ; + + if (usage & PIPE_BUFFER_USAGE_GPU_WRITE) + flags |= DRM_BO_FLAG_WRITE; + + /* drm complains if we don't set any read/write flags. + */ + if ((flags & (DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE)) == 0) + flags |= DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + + buffer->pool = pool; + driGenBuffers( buffer->pool, + "pipe buffer", 1, &buffer->driBO, alignment, flags, 0 ); + + driBOData( buffer->driBO, size, NULL, buffer->pool, 0 ); + + return &buffer->base; +} + + +static struct pipe_buffer * +intel_be_user_buffer_create(struct pipe_winsys *winsys, void *ptr, unsigned bytes) +{ + struct intel_be_buffer *buffer = CALLOC_STRUCT( intel_be_buffer ); + struct intel_be_device *iws = intel_be_device(winsys); + + driGenUserBuffer( iws->regionPool, + "pipe user buffer", &buffer->driBO, ptr, bytes ); + + buffer->base.refcount = 1; + + return &buffer->base; +} + +struct pipe_buffer * +intel_be_buffer_from_handle(struct intel_be_device *device, + const char* name, unsigned handle) +{ + struct intel_be_buffer *be_buf = malloc(sizeof(*be_buf)); + struct pipe_buffer *buffer; + + if (!be_buf) + goto err; + + memset(be_buf, 0, sizeof(*be_buf)); + + driGenBuffers(device->staticPool, name, 1, &be_buf->driBO, 0, 0, 0); + driBOSetReferenced(be_buf->driBO, handle); + + if (0) /** XXX TODO check error */ + goto err_bo; + + buffer = &be_buf->base; + buffer->refcount = 1; + buffer->alignment = 0; + buffer->usage = 0; + buffer->size = driBOSize(be_buf->driBO); + + return buffer; +err_bo: + free(be_buf); +err: + return NULL; +} + + +/* + * Surface functions. + * + * Deprecated! + */ + +static struct pipe_surface * +intel_i915_surface_alloc(struct pipe_winsys *winsys) +{ + assert((size_t)"intel_i915_surface_alloc is deprecated" & 0); + return NULL; +} + +static int +intel_i915_surface_alloc_storage(struct pipe_winsys *winsys, + struct pipe_surface *surf, + unsigned width, unsigned height, + enum pipe_format format, + unsigned flags, + unsigned tex_usage) +{ + assert((size_t)"intel_i915_surface_alloc_storage is deprecated" & 0); + return -1; +} + +static void +intel_i915_surface_release(struct pipe_winsys *winsys, struct pipe_surface **s) +{ + assert((size_t)"intel_i915_surface_release is deprecated" & 0); +} + + +/* + * Fence functions + */ + +static void +intel_be_fence_reference( struct pipe_winsys *sws, + struct pipe_fence_handle **ptr, + struct pipe_fence_handle *fence ) +{ + if (*ptr) + driFenceUnReference((struct _DriFenceObject **)ptr); + + if (fence) + *ptr = (struct pipe_fence_handle *)driFenceReference((struct _DriFenceObject *)fence); +} + +static int +intel_be_fence_signalled( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ) +{ + return driFenceSignaled((struct _DriFenceObject *)fence, flag); +} + +static int +intel_be_fence_finish( struct pipe_winsys *sws, + struct pipe_fence_handle *fence, + unsigned flag ) +{ + return driFenceFinish((struct _DriFenceObject *)fence, flag, 0); +} + + +/* + * Misc functions + */ + +boolean +intel_be_init_device(struct intel_be_device *dev, int fd, unsigned id) +{ + dev->fd = fd; + dev->max_batch_size = 16 * 4096; + dev->max_vertex_size = 128 * 4096; + + dev->base.buffer_create = intel_be_buffer_create; + dev->base.user_buffer_create = intel_be_user_buffer_create; + dev->base.buffer_map = intel_be_buffer_map; + dev->base.buffer_unmap = intel_be_buffer_unmap; + dev->base.buffer_destroy = intel_be_buffer_destroy; + dev->base.surface_alloc = intel_i915_surface_alloc; + dev->base.surface_alloc_storage = intel_i915_surface_alloc_storage; + dev->base.surface_release = intel_i915_surface_release; + dev->base.fence_reference = intel_be_fence_reference; + dev->base.fence_signalled = intel_be_fence_signalled; + dev->base.fence_finish = intel_be_fence_finish; + +#if 0 /* Set by the winsys */ + dev->base.flush_frontbuffer = intel_flush_frontbuffer; + dev->base.get_name = intel_get_name; +#endif + + dev->fMan = driInitFreeSlabManager(10, 10); + dev->fenceMgr = driFenceMgrTTMInit(dev->fd); + + dev->mallocPool = driMallocPoolInit(); + dev->staticPool = driDRMPoolInit(dev->fd); + /* Sizes: 64 128 256 512 1024 2048 4096 8192 16384 32768 */ + dev->regionPool = driSlabPoolInit(dev->fd, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + 64, + 10, 120, 4096 * 64, 0, + dev->fMan); + + dev->vertexPool = driSlabPoolInit(dev->fd, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT, + dev->max_vertex_size, + 1, 120, dev->max_vertex_size * 4, 0, + dev->fMan); + + dev->batchPool = driSlabPoolInit(dev->fd, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + DRM_BO_FLAG_EXE | + DRM_BO_FLAG_MEM_TT, + dev->max_batch_size, + 1, 40, dev->max_batch_size * 16, 0, + dev->fMan); + + /* Fill in this struct with callbacks that i915simple will need to + * communicate with the window system, buffer manager, etc. + */ + dev->screen = i915_create_screen(&dev->base, id); + + return true; +} + +void +intel_be_destroy_device(struct intel_be_device *dev) +{ + driPoolTakeDown(dev->mallocPool); + driPoolTakeDown(dev->staticPool); + driPoolTakeDown(dev->regionPool); + driPoolTakeDown(dev->vertexPool); + driPoolTakeDown(dev->batchPool); + + /** TODO takedown fenceMgr and fMan */ +} diff --git a/src/gallium/winsys/drm/intel/common/intel_be_device.h b/src/gallium/winsys/drm/intel/common/intel_be_device.h new file mode 100644 index 0000000000..3f8b3f585c --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/intel_be_device.h @@ -0,0 +1,72 @@ +#ifndef INTEL_DRM_DEVICE_H +#define INTEL_DRM_DEVICE_H + +#include "pipe/p_winsys.h" +#include "pipe/p_context.h" + +/* + * Device + */ + +struct intel_be_device +{ + struct pipe_winsys base; + + /** + * Hw level screen + */ + struct pipe_screen *screen; + + int fd; /**< Drm file discriptor */ + + size_t max_batch_size; + size_t max_vertex_size; + + struct _DriFenceMgr *fenceMgr; + + struct _DriBufferPool *batchPool; + struct _DriBufferPool *regionPool; + struct _DriBufferPool *mallocPool; + struct _DriBufferPool *vertexPool; + struct _DriBufferPool *staticPool; + struct _DriFreeSlabManager *fMan; +}; + +boolean +intel_be_init_device(struct intel_be_device *device, int fd, unsigned id); + +void +intel_be_destroy_device(struct intel_be_device *dev); + +/* + * Buffer + */ + +struct intel_be_buffer { + struct pipe_buffer base; + struct _DriBufferPool *pool; + struct _DriBufferObject *driBO; +}; + +/** + * Create a be buffer from a drm bo handle + * + * Takes a reference + */ +struct pipe_buffer * +intel_be_buffer_from_handle(struct intel_be_device *device, + const char* name, unsigned handle); + +static INLINE struct intel_be_buffer * +intel_be_buffer(struct pipe_buffer *buf) +{ + return (struct intel_be_buffer *)buf; +} + +static INLINE struct _DriBufferObject * +dri_bo(struct pipe_buffer *buf) +{ + return intel_be_buffer(buf)->driBO; +} + +#endif diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c new file mode 100644 index 0000000000..b6d901f85e --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.c @@ -0,0 +1,949 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + * Keith Whitwell + */ + +#include +#include +#include +#include "glthread.h" +#include "errno.h" +#include "ws_dri_bufmgr.h" +#include "string.h" +#include "pipe/p_debug.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" + + +/* + * This lock is here to protect drmBO structs changing underneath us during a + * validate list call, since validatelist cannot take individiual locks for + * each drmBO. Validatelist takes this lock in write mode. Any access to an + * individual drmBO should take this lock in read mode, since in that case, the + * driBufferObject mutex will protect the access. Locking order is + * driBufferObject mutex - > this rw lock. + */ + +_glthread_DECLARE_STATIC_MUTEX(bmMutex); +_glthread_DECLARE_STATIC_COND(bmCond); + +static int kernelReaders = 0; +static int num_buffers = 0; +static int num_user_buffers = 0; + +static drmBO *drmBOListBuf(void *iterator) +{ + drmBONode *node; + drmMMListHead *l = (drmMMListHead *) iterator; + node = DRMLISTENTRY(drmBONode, l, head); + return node->buf; +} + +static void *drmBOListIterator(drmBOList *list) +{ + void *ret = list->list.next; + + if (ret == &list->list) + return NULL; + return ret; +} + +static void *drmBOListNext(drmBOList *list, void *iterator) +{ + void *ret; + + drmMMListHead *l = (drmMMListHead *) iterator; + ret = l->next; + if (ret == &list->list) + return NULL; + return ret; +} + +static drmBONode *drmAddListItem(drmBOList *list, drmBO *item, + uint64_t arg0, + uint64_t arg1) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->free.next; + if (l == &list->free) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + return NULL; + } + list->numCurrent++; + } + else { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + } + node->buf = item; + node->arg0 = arg0; + node->arg1 = arg1; + DRMLISTADD(&node->head, &list->list); + list->numOnList++; + return node; +} + +static int drmAddValidateItem(drmBOList *list, drmBO *buf, uint64_t flags, + uint64_t mask, int *newItem) +{ + drmBONode *node, *cur; + drmMMListHead *l; + + *newItem = 0; + cur = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + if (node->buf == buf) { + cur = node; + break; + } + } + if (!cur) { + cur = drmAddListItem(list, buf, flags, mask); + if (!cur) { + return -ENOMEM; + } + *newItem = 1; + cur->arg0 = flags; + cur->arg1 = mask; + } + else { + uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; + uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; + + if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { + return -EINVAL; + } + + cur->arg1 |= mask; + cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); + + if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && + (cur->arg0 & DRM_BO_MASK_MEM) == 0) { + return -EINVAL; + } + } + return 0; +} + +static void drmBOFreeList(drmBOList *list) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->list.next; + while(l != &list->list) { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + l = list->list.next; + list->numCurrent--; + list->numOnList--; + } + + l = list->free.next; + while(l != &list->free) { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + l = list->free.next; + list->numCurrent--; + } +} + +static int drmAdjustListNodes(drmBOList *list) +{ + drmBONode *node; + drmMMListHead *l; + int ret = 0; + + while(list->numCurrent < list->numTarget) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + ret = -ENOMEM; + break; + } + list->numCurrent++; + DRMLISTADD(&node->head, &list->free); + } + + while(list->numCurrent > list->numTarget) { + l = list->free.next; + if (l == &list->free) + break; + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + free(node); + list->numCurrent--; + } + return ret; +} + +static int drmBOCreateList(int numTarget, drmBOList *list) +{ + DRMINITLISTHEAD(&list->list); + DRMINITLISTHEAD(&list->free); + list->numTarget = numTarget; + list->numCurrent = 0; + list->numOnList = 0; + return drmAdjustListNodes(list); +} + +static int drmBOResetList(drmBOList *list) +{ + drmMMListHead *l; + int ret; + + ret = drmAdjustListNodes(list); + if (ret) + return ret; + + l = list->list.next; + while (l != &list->list) { + DRMLISTDEL(l); + DRMLISTADD(l, &list->free); + list->numOnList--; + l = list->list.next; + } + return drmAdjustListNodes(list); +} + +void driWriteLockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + while(kernelReaders != 0) + _glthread_COND_WAIT(bmCond, bmMutex); +} + +void driWriteUnlockKernelBO(void) +{ + _glthread_UNLOCK_MUTEX(bmMutex); +} + +void driReadLockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + kernelReaders++; + _glthread_UNLOCK_MUTEX(bmMutex); +} + +void driReadUnlockKernelBO(void) +{ + _glthread_LOCK_MUTEX(bmMutex); + if (--kernelReaders == 0) + _glthread_COND_BROADCAST(bmCond); + _glthread_UNLOCK_MUTEX(bmMutex); +} + + + + +/* + * TODO: Introduce fence pools in the same way as + * buffer object pools. + */ + +typedef struct _DriBufferObject +{ + DriBufferPool *pool; + _glthread_Mutex mutex; + int refCount; + const char *name; + uint64_t flags; + unsigned hint; + unsigned alignment; + unsigned createdByReference; + void *private; + /* user-space buffer: */ + unsigned userBuffer; + void *userData; + unsigned userSize; +} DriBufferObject; + +typedef struct _DriBufferList { + drmBOList drmBuffers; /* List of kernel buffers needing validation */ + drmBOList driBuffers; /* List of user-space buffers needing validation */ +} DriBufferList; + + +void +bmError(int val, const char *file, const char *function, int line) +{ + printf("Fatal video memory manager error \"%s\".\n" + "Check kernel logs or set the LIBGL_DEBUG\n" + "environment variable to \"verbose\" for more info.\n" + "Detected in file %s, line %d, function %s.\n", + strerror(-val), file, line, function); +#ifndef NDEBUG + abort(); +#else + abort(); +#endif +} + +extern drmBO * +driBOKernel(struct _DriBufferObject *buf) +{ + drmBO *ret; + + driReadLockKernelBO(); + _glthread_LOCK_MUTEX(buf->mutex); + assert(buf->private != NULL); + ret = buf->pool->kernel(buf->pool, buf->private); + if (!ret) + BM_CKFATAL(-EINVAL); + _glthread_UNLOCK_MUTEX(buf->mutex); + driReadUnlockKernelBO(); + + return ret; +} + +void +driBOWaitIdle(struct _DriBufferObject *buf, int lazy) +{ + + /* + * This function may block. Is it sane to keep the mutex held during + * that time?? + */ + + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(buf->pool->waitIdle(buf->pool, buf->private, &buf->mutex, lazy)); + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void * +driBOMap(struct _DriBufferObject *buf, unsigned flags, unsigned hint) +{ + void *virtual; + int retval; + + if (buf->userBuffer) { + return buf->userData; + } + + _glthread_LOCK_MUTEX(buf->mutex); + assert(buf->private != NULL); + retval = buf->pool->map(buf->pool, buf->private, flags, hint, + &buf->mutex, &virtual); + _glthread_UNLOCK_MUTEX(buf->mutex); + + return retval == 0 ? virtual : NULL; +} + +void +driBOUnmap(struct _DriBufferObject *buf) +{ + if (buf->userBuffer) + return; + + assert(buf->private != NULL); + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +unsigned long +driBOOffset(struct _DriBufferObject *buf) +{ + unsigned long ret; + + assert(buf->private != NULL); + + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->offset(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + return ret; +} + +unsigned long +driBOPoolOffset(struct _DriBufferObject *buf) +{ + unsigned long ret; + + assert(buf->private != NULL); + + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->poolOffset(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + return ret; +} + +uint64_t +driBOFlags(struct _DriBufferObject *buf) +{ + uint64_t ret; + + assert(buf->private != NULL); + + driReadLockKernelBO(); + _glthread_LOCK_MUTEX(buf->mutex); + ret = buf->pool->flags(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + driReadUnlockKernelBO(); + return ret; +} + +struct _DriBufferObject * +driBOReference(struct _DriBufferObject *buf) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (++buf->refCount == 1) { + _glthread_UNLOCK_MUTEX(buf->mutex); + BM_CKFATAL(-EINVAL); + } + _glthread_UNLOCK_MUTEX(buf->mutex); + return buf; +} + +void +driBOUnReference(struct _DriBufferObject *buf) +{ + int tmp; + + if (!buf) + return; + + _glthread_LOCK_MUTEX(buf->mutex); + tmp = --buf->refCount; + if (!tmp) { + _glthread_UNLOCK_MUTEX(buf->mutex); + if (buf->private) { + if (buf->createdByReference) + buf->pool->unreference(buf->pool, buf->private); + else + buf->pool->destroy(buf->pool, buf->private); + } + if (buf->userBuffer) + num_user_buffers--; + else + num_buffers--; + free(buf); + } else + _glthread_UNLOCK_MUTEX(buf->mutex); + +} + + +int +driBOData(struct _DriBufferObject *buf, + unsigned size, const void *data, + DriBufferPool *newPool, + uint64_t flags) +{ + void *virtual = NULL; + int newBuffer; + int retval = 0; + struct _DriBufferPool *pool; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + pool = buf->pool; + + if (pool == NULL && newPool != NULL) { + buf->pool = newPool; + pool = newPool; + } + if (newPool == NULL) + newPool = pool; + + if (!pool->create) { + assert((size_t)"driBOData called on invalid buffer\n" & 0); + BM_CKFATAL(-EINVAL); + } + + newBuffer = (!buf->private || pool != newPool || + pool->size(pool, buf->private) < size); + + if (!flags) + flags = buf->flags; + + if (newBuffer) { + + if (buf->createdByReference) { + assert((size_t)"driBOData requiring resizing called on shared buffer.\n" & 0); + BM_CKFATAL(-EINVAL); + } + + if (buf->private) + buf->pool->destroy(buf->pool, buf->private); + + pool = newPool; + buf->pool = newPool; + buf->private = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, + buf->alignment); + if (!buf->private) + retval = -ENOMEM; + + if (retval == 0) + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, + DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual); + } else if (pool->map(pool, buf->private, DRM_BO_FLAG_WRITE, + DRM_BO_HINT_DONT_BLOCK, &buf->mutex, &virtual)) { + /* + * Buffer is busy. need to create a new one. + */ + + void *newBuf; + + newBuf = pool->create(pool, size, flags, DRM_BO_HINT_DONT_FENCE, + buf->alignment); + if (newBuf) { + buf->pool->destroy(buf->pool, buf->private); + buf->private = newBuf; + } + + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); + } else { + uint64_t flag_diff = flags ^ buf->flags; + + /* + * We might need to change buffer flags. + */ + + if (flag_diff){ + assert(pool->setStatus != NULL); + BM_CKFATAL(pool->unmap(pool, buf->private)); + BM_CKFATAL(pool->setStatus(pool, buf->private, flag_diff, + buf->flags)); + if (!data) + goto out; + + retval = pool->map(pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, &virtual); + } + } + + if (retval == 0) { + if (data) + memcpy(virtual, data, size); + + BM_CKFATAL(pool->unmap(pool, buf->private)); + } + + out: + _glthread_UNLOCK_MUTEX(buf->mutex); + + return retval; +} + +void +driBOSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, const void *data) +{ + void *virtual; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + if (size && data) { + BM_CKFATAL(buf->pool->map(buf->pool, buf->private, + DRM_BO_FLAG_WRITE, 0, &buf->mutex, + &virtual)); + memcpy((unsigned char *) virtual + offset, data, size); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + } + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void +driBOGetSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, void *data) +{ + void *virtual; + + assert(!buf->userBuffer); /* XXX just do a memcpy? */ + + _glthread_LOCK_MUTEX(buf->mutex); + if (size && data) { + BM_CKFATAL(buf->pool->map(buf->pool, buf->private, + DRM_BO_FLAG_READ, 0, &buf->mutex, &virtual)); + memcpy(data, (unsigned char *) virtual + offset, size); + BM_CKFATAL(buf->pool->unmap(buf->pool, buf->private)); + } + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +void +driBOSetReferenced(struct _DriBufferObject *buf, + unsigned long handle) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->private != NULL) { + assert((size_t)"Invalid buffer for setReferenced\n" & 0); + BM_CKFATAL(-EINVAL); + + } + if (buf->pool->reference == NULL) { + assert((size_t)"Invalid buffer pool for setReferenced\n" & 0); + BM_CKFATAL(-EINVAL); + } + buf->private = buf->pool->reference(buf->pool, handle); + if (!buf->private) { + assert((size_t)"Invalid buffer pool for setStatic\n" & 0); + BM_CKFATAL(-ENOMEM); + } + buf->createdByReference = TRUE; + buf->flags = buf->pool->kernel(buf->pool, buf->private)->flags; + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +int +driGenBuffers(struct _DriBufferPool *pool, + const char *name, + unsigned n, + struct _DriBufferObject *buffers[], + unsigned alignment, uint64_t flags, unsigned hint) +{ + struct _DriBufferObject *buf; + int i; + + flags = (flags) ? flags : DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_MEM_VRAM | + DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE; + + ++num_buffers; + + assert(pool); + + for (i = 0; i < n; ++i) { + buf = (struct _DriBufferObject *) calloc(1, sizeof(*buf)); + if (!buf) + return -ENOMEM; + + _glthread_INIT_MUTEX(buf->mutex); + _glthread_LOCK_MUTEX(buf->mutex); + buf->refCount = 1; + buf->flags = flags; + buf->hint = hint; + buf->name = name; + buf->alignment = alignment; + buf->pool = pool; + buf->createdByReference = 0; + _glthread_UNLOCK_MUTEX(buf->mutex); + buffers[i] = buf; + } + return 0; +} + +void +driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject **buffers, + void *ptr, unsigned bytes) +{ + const unsigned alignment = 1, flags = 0, hint = 0; + + --num_buffers; /* JB: is inced in GenBuffes */ + driGenBuffers(pool, name, 1, buffers, alignment, flags, hint); + ++num_user_buffers; + + (*buffers)->userBuffer = 1; + (*buffers)->userData = ptr; + (*buffers)->userSize = bytes; +} + +void +driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]) +{ + int i; + + for (i = 0; i < n; ++i) { + driBOUnReference(buffers[i]); + } +} + + +void +driInitBufMgr(int fd) +{ + ; +} + +/* + * Note that lists are per-context and don't need mutex protection. + */ + +struct _DriBufferList * +driBOCreateList(int target) +{ + struct _DriBufferList *list = calloc(sizeof(*list), 1); + + BM_CKFATAL(drmBOCreateList(target, &list->drmBuffers)); + BM_CKFATAL(drmBOCreateList(target, &list->driBuffers)); + return list; +} + +int +driBOResetList(struct _DriBufferList * list) +{ + int ret; + ret = drmBOResetList(&list->drmBuffers); + if (ret) + return ret; + ret = drmBOResetList(&list->driBuffers); + return ret; +} + +void +driBOFreeList(struct _DriBufferList * list) +{ + drmBOFreeList(&list->drmBuffers); + drmBOFreeList(&list->driBuffers); + free(list); +} + + +/* + * Copied from libdrm, because it is needed by driAddValidateItem. + */ + +static drmBONode * +driAddListItem(drmBOList * list, drmBO * item, + uint64_t arg0, uint64_t arg1) +{ + drmBONode *node; + drmMMListHead *l; + + l = list->free.next; + if (l == &list->free) { + node = (drmBONode *) malloc(sizeof(*node)); + if (!node) { + return NULL; + } + list->numCurrent++; + } else { + DRMLISTDEL(l); + node = DRMLISTENTRY(drmBONode, l, head); + } + memset(&node->bo_arg, 0, sizeof(node->bo_arg)); + node->buf = item; + node->arg0 = arg0; + node->arg1 = arg1; + DRMLISTADDTAIL(&node->head, &list->list); + list->numOnList++; + return node; +} + +/* + * Slightly modified version compared to the libdrm version. + * This one returns the list index of the buffer put on the list. + */ + +static int +driAddValidateItem(drmBOList * list, drmBO * buf, uint64_t flags, + uint64_t mask, int *itemLoc, + struct _drmBONode **pnode) +{ + drmBONode *node, *cur; + drmMMListHead *l; + int count = 0; + + cur = NULL; + + for (l = list->list.next; l != &list->list; l = l->next) { + node = DRMLISTENTRY(drmBONode, l, head); + if (node->buf == buf) { + cur = node; + break; + } + count++; + } + if (!cur) { + cur = driAddListItem(list, buf, flags, mask); + if (!cur) + return -ENOMEM; + + cur->arg0 = flags; + cur->arg1 = mask; + } else { + uint64_t memFlags = cur->arg0 & flags & DRM_BO_MASK_MEM; + uint64_t accFlags = (cur->arg0 | flags) & ~DRM_BO_MASK_MEM; + + if (mask & cur->arg1 & ~DRM_BO_MASK_MEM & (cur->arg0 ^ flags)) { + return -EINVAL; + } + + cur->arg1 |= mask; + cur->arg0 = (cur->arg0 & ~mask) | ((memFlags | accFlags) & mask); + + if (((cur->arg1 & DRM_BO_MASK_MEM) != 0) && + (cur->arg0 & DRM_BO_MASK_MEM) == 0) { + return -EINVAL; + } + } + *itemLoc = count; + *pnode = cur; + return 0; +} + + +void +driBOAddListItem(struct _DriBufferList * list, struct _DriBufferObject *buf, + uint64_t flags, uint64_t mask, int *itemLoc, + struct _drmBONode **node) +{ + int newItem; + + _glthread_LOCK_MUTEX(buf->mutex); + BM_CKFATAL(driAddValidateItem(&list->drmBuffers, + buf->pool->kernel(buf->pool, buf->private), + flags, mask, itemLoc, node)); + BM_CKFATAL(drmAddValidateItem(&list->driBuffers, (drmBO *) buf, + flags, mask, &newItem)); + if (newItem) + buf->refCount++; + + _glthread_UNLOCK_MUTEX(buf->mutex); +} + +drmBOList *driGetdrmBOList(struct _DriBufferList *list) +{ + driWriteLockKernelBO(); + return &list->drmBuffers; +} + +void driPutdrmBOList(struct _DriBufferList *list) +{ + driWriteUnlockKernelBO(); +} + + +void +driBOFence(struct _DriBufferObject *buf, struct _DriFenceObject *fence) +{ + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->pool->fence) + BM_CKFATAL(buf->pool->fence(buf->pool, buf->private, fence)); + _glthread_UNLOCK_MUTEX(buf->mutex); + +} + +void +driBOUnrefUserList(struct _DriBufferList *list) +{ + struct _DriBufferObject *buf; + void *curBuf; + + curBuf = drmBOListIterator(&list->driBuffers); + while (curBuf) { + buf = (struct _DriBufferObject *)drmBOListBuf(curBuf); + driBOUnReference(buf); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } +} + +struct _DriFenceObject * +driBOFenceUserList(struct _DriFenceMgr *mgr, + struct _DriBufferList *list, const char *name, + drmFence *kFence) +{ + struct _DriFenceObject *fence; + struct _DriBufferObject *buf; + void *curBuf; + + fence = driFenceCreate(mgr, kFence->fence_class, kFence->type, + kFence, sizeof(*kFence)); + curBuf = drmBOListIterator(&list->driBuffers); + + /* + * User-space fencing callbacks. + */ + + while (curBuf) { + buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); + driBOFence(buf, fence); + driBOUnReference(buf); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } + + driBOResetList(list); + return fence; +} + +void +driBOValidateUserList(struct _DriBufferList * list) +{ + void *curBuf; + struct _DriBufferObject *buf; + + curBuf = drmBOListIterator(&list->driBuffers); + + /* + * User-space validation callbacks. + */ + + while (curBuf) { + buf = (struct _DriBufferObject *) drmBOListBuf(curBuf); + _glthread_LOCK_MUTEX(buf->mutex); + if (buf->pool->validate) + BM_CKFATAL(buf->pool->validate(buf->pool, buf->private, &buf->mutex)); + _glthread_UNLOCK_MUTEX(buf->mutex); + curBuf = drmBOListNext(&list->driBuffers, curBuf); + } +} + + +void +driPoolTakeDown(struct _DriBufferPool *pool) +{ + pool->takeDown(pool); + +} + +unsigned long +driBOSize(struct _DriBufferObject *buf) +{ + unsigned long size; + + _glthread_LOCK_MUTEX(buf->mutex); + size = buf->pool->size(buf->pool, buf->private); + _glthread_UNLOCK_MUTEX(buf->mutex); + + return size; + +} + +drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list) +{ + return &list->drmBuffers; +} + +drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list) +{ + return &list->driBuffers; +} + diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h new file mode 100644 index 0000000000..e6c0cff0a0 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/ws_dri_bufmgr.h @@ -0,0 +1,138 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + * Keith Whitwell + */ + +#ifndef _PSB_BUFMGR_H_ +#define _PSB_BUFMGR_H_ +#include +#include "i915_drm.h" +#include "ws_dri_fencemgr.h" + +typedef struct _drmBONode +{ + drmMMListHead head; + drmBO *buf; + struct drm_i915_op_arg bo_arg; + uint64_t arg0; + uint64_t arg1; +} drmBONode; + +typedef struct _drmBOList { + unsigned numTarget; + unsigned numCurrent; + unsigned numOnList; + drmMMListHead list; + drmMMListHead free; +} drmBOList; + + +struct _DriFenceObject; +struct _DriBufferObject; +struct _DriBufferPool; +struct _DriBufferList; + +/* + * Return a pointer to the libdrm buffer object this DriBufferObject + * uses. + */ + +extern drmBO *driBOKernel(struct _DriBufferObject *buf); +extern void *driBOMap(struct _DriBufferObject *buf, unsigned flags, + unsigned hint); +extern void driBOUnmap(struct _DriBufferObject *buf); +extern unsigned long driBOOffset(struct _DriBufferObject *buf); +extern unsigned long driBOPoolOffset(struct _DriBufferObject *buf); + +extern uint64_t driBOFlags(struct _DriBufferObject *buf); +extern struct _DriBufferObject *driBOReference(struct _DriBufferObject *buf); +extern void driBOUnReference(struct _DriBufferObject *buf); + +extern int driBOData(struct _DriBufferObject *r_buf, + unsigned size, const void *data, + struct _DriBufferPool *pool, uint64_t flags); + +extern void driBOSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, + const void *data); +extern void driBOGetSubData(struct _DriBufferObject *buf, + unsigned long offset, unsigned long size, + void *data); +extern int driGenBuffers(struct _DriBufferPool *pool, + const char *name, + unsigned n, + struct _DriBufferObject *buffers[], + unsigned alignment, uint64_t flags, unsigned hint); +extern void driGenUserBuffer(struct _DriBufferPool *pool, + const char *name, + struct _DriBufferObject *buffers[], + void *ptr, unsigned bytes); +extern void driDeleteBuffers(unsigned n, struct _DriBufferObject *buffers[]); +extern void driInitBufMgr(int fd); +extern struct _DriBufferList *driBOCreateList(int target); +extern int driBOResetList(struct _DriBufferList * list); +extern void driBOAddListItem(struct _DriBufferList * list, + struct _DriBufferObject *buf, + uint64_t flags, uint64_t mask, int *itemLoc, + struct _drmBONode **node); + +extern void driBOValidateList(int fd, struct _DriBufferList * list); +extern void driBOFreeList(struct _DriBufferList * list); +extern struct _DriFenceObject *driBOFenceUserList(struct _DriFenceMgr *mgr, + struct _DriBufferList *list, + const char *name, + drmFence *kFence); +extern void driBOUnrefUserList(struct _DriBufferList *list); +extern void driBOValidateUserList(struct _DriBufferList * list); +extern drmBOList *driGetdrmBOList(struct _DriBufferList *list); +extern void driPutdrmBOList(struct _DriBufferList *list); + +extern void driBOFence(struct _DriBufferObject *buf, + struct _DriFenceObject *fence); + +extern void driPoolTakeDown(struct _DriBufferPool *pool); +extern void driBOSetReferenced(struct _DriBufferObject *buf, + unsigned long handle); +unsigned long driBOSize(struct _DriBufferObject *buf); +extern void driBOWaitIdle(struct _DriBufferObject *buf, int lazy); +extern void driPoolTakeDown(struct _DriBufferPool *pool); + +extern void driReadLockKernelBO(void); +extern void driReadUnlockKernelBO(void); +extern void driWriteLockKernelBO(void); +extern void driWriteUnlockKernelBO(void); + +/* + * For debugging purposes. + */ + +extern drmBOList *driBOGetDRMBuffers(struct _DriBufferList *list); +extern drmBOList *driBOGetDRIBuffers(struct _DriBufferList *list); +#endif diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h b/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h new file mode 100644 index 0000000000..bf60798924 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/ws_dri_bufpool.h @@ -0,0 +1,102 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#ifndef _PSB_BUFPOOL_H_ +#define _PSB_BUFPOOL_H_ + +#include +#include +struct _DriFenceObject; + +typedef struct _DriBufferPool +{ + int fd; + int (*map) (struct _DriBufferPool * pool, void *private, + unsigned flags, int hint, _glthread_Mutex *mutex, + void **virtual); + int (*unmap) (struct _DriBufferPool * pool, void *private); + int (*destroy) (struct _DriBufferPool * pool, void *private); + unsigned long (*offset) (struct _DriBufferPool * pool, void *private); + unsigned long (*poolOffset) (struct _DriBufferPool * pool, void *private); + uint64_t (*flags) (struct _DriBufferPool * pool, void *private); + unsigned long (*size) (struct _DriBufferPool * pool, void *private); + void *(*create) (struct _DriBufferPool * pool, unsigned long size, + uint64_t flags, unsigned hint, unsigned alignment); + void *(*reference) (struct _DriBufferPool * pool, unsigned handle); + int (*unreference) (struct _DriBufferPool * pool, void *private); + int (*fence) (struct _DriBufferPool * pool, void *private, + struct _DriFenceObject * fence); + drmBO *(*kernel) (struct _DriBufferPool * pool, void *private); + int (*validate) (struct _DriBufferPool * pool, void *private, _glthread_Mutex *mutex); + int (*waitIdle) (struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, + int lazy); + int (*setStatus) (struct _DriBufferPool *pool, void *private, + uint64_t flag_diff, uint64_t old_flags); + void (*takeDown) (struct _DriBufferPool * pool); + void *data; +} DriBufferPool; + +extern void bmError(int val, const char *file, const char *function, + int line); +#define BM_CKFATAL(val) \ + do{ \ + int tstVal = (val); \ + if (tstVal) \ + bmError(tstVal, __FILE__, __FUNCTION__, __LINE__); \ + } while(0); + + +/* + * Builtin pools. + */ + +/* + * Kernel buffer objects. Size in multiples of page size. Page size aligned. + */ + +extern struct _DriBufferPool *driDRMPoolInit(int fd); +extern struct _DriBufferPool *driMallocPoolInit(void); + +struct _DriFreeSlabManager; +extern struct _DriBufferPool * driSlabPoolInit(int fd, uint64_t flags, + uint64_t validMask, + uint32_t smallestSize, + uint32_t numSizes, + uint32_t desiredNumBuffers, + uint32_t maxSlabSize, + uint32_t pageAlignment, + struct _DriFreeSlabManager *fMan); +extern void driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan); +extern struct _DriFreeSlabManager * +driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec); + + +#endif diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c new file mode 100644 index 0000000000..40929efa2f --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/ws_dri_drmpool.c @@ -0,0 +1,268 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" +#include "assert.h" + +/* + * Buffer pool implementation using DRM buffer objects as DRI buffer objects. + */ + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, uint64_t flags, unsigned hint, + unsigned alignment) +{ + drmBO *buf = (drmBO *) malloc(sizeof(*buf)); + int ret; + unsigned pageSize = getpagesize(); + + if (!buf) + return NULL; + + if ((alignment > pageSize) && (alignment % pageSize)) { + free(buf); + return NULL; + } + + ret = drmBOCreate(pool->fd, size, alignment / pageSize, + NULL, + flags, hint, buf); + if (ret) { + free(buf); + return NULL; + } + + return (void *) buf; +} + +static void * +pool_reference(struct _DriBufferPool *pool, unsigned handle) +{ + drmBO *buf = (drmBO *) malloc(sizeof(*buf)); + int ret; + + if (!buf) + return NULL; + + ret = drmBOReference(pool->fd, handle, buf); + + if (ret) { + free(buf); + return NULL; + } + + return (void *) buf; +} + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + int ret; + drmBO *buf = (drmBO *) private; + driReadLockKernelBO(); + ret = drmBOUnreference(pool->fd, buf); + free(buf); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_unreference(struct _DriBufferPool *pool, void *private) +{ + int ret; + drmBO *buf = (drmBO *) private; + driReadLockKernelBO(); + ret = drmBOUnreference(pool->fd, buf); + free(buf); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOMap(pool->fd, buf, flags, hint, virtual); + driReadUnlockKernelBO(); + return ret; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOUnmap(pool->fd, buf); + driReadUnlockKernelBO(); + + return ret; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + unsigned long offset; + + driReadLockKernelBO(); + assert(buf->flags & DRM_BO_FLAG_NO_MOVE); + offset = buf->offset; + driReadUnlockKernelBO(); + + return buf->offset; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + return 0; +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + uint64_t flags; + + driReadLockKernelBO(); + flags = buf->flags; + driReadUnlockKernelBO(); + + return flags; +} + + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + drmBO *buf = (drmBO *) private; + unsigned long size; + + driReadLockKernelBO(); + size = buf->size; + driReadUnlockKernelBO(); + + return buf->size; +} + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + /* + * Noop. The kernel handles all fencing. + */ + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + return (drmBO *) private; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, _glthread_Mutex *mutex, + int lazy) +{ + drmBO *buf = (drmBO *) private; + int ret; + + driReadLockKernelBO(); + ret = drmBOWaitIdle(pool->fd, buf, (lazy) ? DRM_BO_HINT_WAIT_LAZY:0); + driReadUnlockKernelBO(); + + return ret; +} + + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + free(pool); +} + +/*static int +pool_setStatus(struct _DriBufferPool *pool, void *private, + uint64_t flag_diff, uint64_t old_flags) +{ + drmBO *buf = (drmBO *) private; + uint64_t new_flags = old_flags ^ flag_diff; + int ret; + + driReadLockKernelBO(); + ret = drmBOSetStatus(pool->fd, buf, new_flags, flag_diff, + 0, 0, 0); + driReadUnlockKernelBO(); + return ret; +}*/ + +struct _DriBufferPool * +driDRMPoolInit(int fd) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + + if (!pool) + return NULL; + + pool->fd = fd; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->poolOffset = &pool_poolOffset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = NULL; + pool->waitIdle = &pool_waitIdle; + pool->takeDown = &pool_takedown; + pool->reference = &pool_reference; + pool->unreference = &pool_unreference; + pool->data = NULL; + return pool; +} diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c new file mode 100644 index 0000000000..b56bc269da --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.c @@ -0,0 +1,377 @@ +#include "ws_dri_fencemgr.h" +#include "glthread.h" +#include +#include +#include + +/* + * Note: Locking order is + * _DriFenceObject::mutex + * _DriFenceMgr::mutex + */ + +struct _DriFenceMgr { + /* + * Constant members. Need no mutex protection. + */ + struct _DriFenceMgrCreateInfo info; + void *private; + + /* + * These members are protected by this->mutex + */ + _glthread_Mutex mutex; + int refCount; + drmMMListHead *heads; + int num_fences; +}; + +struct _DriFenceObject { + + /* + * These members are constant and need no mutex protection. + */ + struct _DriFenceMgr *mgr; + uint32_t fence_class; + uint32_t fence_type; + + /* + * These members are protected by mgr->mutex. + */ + drmMMListHead head; + int refCount; + + /* + * These members are protected by this->mutex. + */ + _glthread_Mutex mutex; + uint32_t signaled_type; + void *private; +}; + +uint32_t +driFenceType(struct _DriFenceObject *fence) +{ + return fence->fence_type; +} + +struct _DriFenceMgr * +driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info) +{ + struct _DriFenceMgr *tmp; + uint32_t i; + + tmp = calloc(1, sizeof(*tmp)); + if (!tmp) + return NULL; + + _glthread_INIT_MUTEX(tmp->mutex); + _glthread_LOCK_MUTEX(tmp->mutex); + tmp->refCount = 1; + tmp->info = *info; + tmp->num_fences = 0; + tmp->heads = calloc(tmp->info.num_classes, sizeof(*tmp->heads)); + if (!tmp->heads) + goto out_err; + + for (i=0; iinfo.num_classes; ++i) { + DRMINITLISTHEAD(&tmp->heads[i]); + } + _glthread_UNLOCK_MUTEX(tmp->mutex); + return tmp; + + out_err: + if (tmp) + free(tmp); + return NULL; +} + +static void +driFenceMgrUnrefUnlock(struct _DriFenceMgr **pMgr) +{ + struct _DriFenceMgr *mgr = *pMgr; + + *pMgr = NULL; + if (--mgr->refCount == 0) + free(mgr); + else + _glthread_UNLOCK_MUTEX(mgr->mutex); +} + +void +driFenceMgrUnReference(struct _DriFenceMgr **pMgr) +{ + _glthread_LOCK_MUTEX((*pMgr)->mutex); + driFenceMgrUnrefUnlock(pMgr); +} + +static void +driFenceUnReferenceLocked(struct _DriFenceObject **pFence) +{ + struct _DriFenceObject *fence = *pFence; + struct _DriFenceMgr *mgr = fence->mgr; + + *pFence = NULL; + if (--fence->refCount == 0) { + DRMLISTDELINIT(&fence->head); + if (fence->private) + mgr->info.unreference(mgr, &fence->private); + --mgr->num_fences; + fence->mgr = NULL; + --mgr->refCount; + free(fence); + + } +} + + +static void +driSignalPreviousFencesLocked(struct _DriFenceMgr *mgr, + drmMMListHead *list, + uint32_t fence_class, + uint32_t fence_type) +{ + struct _DriFenceObject *entry; + drmMMListHead *prev; + + while(list != &mgr->heads[fence_class]) { + entry = DRMLISTENTRY(struct _DriFenceObject, list, head); + + /* + * Up refcount so that entry doesn't disappear from under us + * when we unlock-relock mgr to get the correct locking order. + */ + + ++entry->refCount; + _glthread_UNLOCK_MUTEX(mgr->mutex); + _glthread_LOCK_MUTEX(entry->mutex); + _glthread_LOCK_MUTEX(mgr->mutex); + + prev = list->prev; + + + + if (list->prev == list) { + + /* + * Somebody else removed the entry from the list. + */ + + _glthread_UNLOCK_MUTEX(entry->mutex); + driFenceUnReferenceLocked(&entry); + return; + } + + entry->signaled_type |= (fence_type & entry->fence_type); + if (entry->signaled_type == entry->fence_type) { + DRMLISTDELINIT(list); + mgr->info.unreference(mgr, &entry->private); + } + _glthread_UNLOCK_MUTEX(entry->mutex); + driFenceUnReferenceLocked(&entry); + list = prev; + } +} + + +int +driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, + int lazy_hint) +{ + struct _DriFenceMgr *mgr = fence->mgr; + int ret = 0; + + _glthread_LOCK_MUTEX(fence->mutex); + + if ((fence->signaled_type & fence_type) == fence_type) + goto out0; + + ret = mgr->info.finish(mgr, fence->private, fence_type, lazy_hint); + if (ret) + goto out0; + + _glthread_LOCK_MUTEX(mgr->mutex); + _glthread_UNLOCK_MUTEX(fence->mutex); + + driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, + fence_type); + _glthread_UNLOCK_MUTEX(mgr->mutex); + return 0; + + out0: + _glthread_UNLOCK_MUTEX(fence->mutex); + return ret; +} + +uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence) +{ + uint32_t ret; + + _glthread_LOCK_MUTEX(fence->mutex); + ret = fence->signaled_type; + _glthread_UNLOCK_MUTEX(fence->mutex); + + return ret; +} + +int +driFenceSignaledType(struct _DriFenceObject *fence, uint32_t flush_type, + uint32_t *signaled) +{ + int ret = 0; + struct _DriFenceMgr *mgr; + + _glthread_LOCK_MUTEX(fence->mutex); + mgr = fence->mgr; + *signaled = fence->signaled_type; + if ((fence->signaled_type & flush_type) == flush_type) + goto out0; + + ret = mgr->info.signaled(mgr, fence->private, flush_type, signaled); + if (ret) { + *signaled = fence->signaled_type; + goto out0; + } + + if ((fence->signaled_type | *signaled) == fence->signaled_type) + goto out0; + + _glthread_LOCK_MUTEX(mgr->mutex); + _glthread_UNLOCK_MUTEX(fence->mutex); + + driSignalPreviousFencesLocked(mgr, &fence->head, fence->fence_class, + *signaled); + + _glthread_UNLOCK_MUTEX(mgr->mutex); + return 0; + out0: + _glthread_UNLOCK_MUTEX(fence->mutex); + return ret; +} + +struct _DriFenceObject * +driFenceReference(struct _DriFenceObject *fence) +{ + _glthread_LOCK_MUTEX(fence->mgr->mutex); + ++fence->refCount; + _glthread_UNLOCK_MUTEX(fence->mgr->mutex); + return fence; +} + +void +driFenceUnReference(struct _DriFenceObject **pFence) +{ + struct _DriFenceMgr *mgr; + + if (*pFence == NULL) + return; + + mgr = (*pFence)->mgr; + _glthread_LOCK_MUTEX(mgr->mutex); + ++mgr->refCount; + driFenceUnReferenceLocked(pFence); + driFenceMgrUnrefUnlock(&mgr); +} + +struct _DriFenceObject +*driFenceCreate(struct _DriFenceMgr *mgr, uint32_t fence_class, + uint32_t fence_type, void *private, size_t private_size) +{ + struct _DriFenceObject *fence; + size_t fence_size = sizeof(*fence); + + if (private_size) + fence_size = ((fence_size + 15) & ~15); + + fence = calloc(1, fence_size + private_size); + + if (!fence) { + int ret = mgr->info.finish(mgr, private, fence_type, 0); + + if (ret) + usleep(10000000); + + return NULL; + } + + _glthread_INIT_MUTEX(fence->mutex); + _glthread_LOCK_MUTEX(fence->mutex); + _glthread_LOCK_MUTEX(mgr->mutex); + fence->refCount = 1; + DRMLISTADDTAIL(&fence->head, &mgr->heads[fence_class]); + fence->mgr = mgr; + ++mgr->refCount; + ++mgr->num_fences; + _glthread_UNLOCK_MUTEX(mgr->mutex); + fence->fence_class = fence_class; + fence->fence_type = fence_type; + fence->signaled_type = 0; + fence->private = private; + if (private_size) { + fence->private = (void *)(((uint8_t *) fence) + fence_size); + memcpy(fence->private, private, private_size); + } + + _glthread_UNLOCK_MUTEX(fence->mutex); + return fence; +} + + +static int +tSignaled(struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, + uint32_t *signaled_type) +{ + long fd = (long) mgr->private; + int dummy; + drmFence *fence = (drmFence *) private; + int ret; + + *signaled_type = 0; + ret = drmFenceSignaled((int) fd, fence, flush_type, &dummy); + if (ret) + return ret; + + *signaled_type = fence->signaled; + + return 0; +} + +static int +tFinish(struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, + int lazy_hint) +{ + long fd = (long) mgr->private; + unsigned flags = lazy_hint ? DRM_FENCE_FLAG_WAIT_LAZY : 0; + + return drmFenceWait((int)fd, flags, (drmFence *) private, fence_type); +} + +static int +tUnref(struct _DriFenceMgr *mgr, void **private) +{ + long fd = (long) mgr->private; + drmFence *fence = (drmFence *) *private; + *private = NULL; + + return drmFenceUnreference(fd, fence); +} + +struct _DriFenceMgr *driFenceMgrTTMInit(int fd) +{ + struct _DriFenceMgrCreateInfo info; + struct _DriFenceMgr *mgr; + + info.flags = DRI_FENCE_CLASS_ORDERED; + info.num_classes = 4; + info.signaled = tSignaled; + info.finish = tFinish; + info.unreference = tUnref; + + mgr = driFenceMgrCreate(&info); + if (mgr == NULL) + return NULL; + + mgr->private = (void *) (long) fd; + return mgr; +} + diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h new file mode 100644 index 0000000000..4ea58dfe18 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/ws_dri_fencemgr.h @@ -0,0 +1,115 @@ +#ifndef DRI_FENCEMGR_H +#define DRI_FENCEMGR_H + +#include +#include + +struct _DriFenceObject; +struct _DriFenceMgr; + +/* + * Do a quick check to see if the fence manager has registered the fence + * object as signaled. Note that this function may return a false negative + * answer. + */ +extern uint32_t driFenceSignaledTypeCached(struct _DriFenceObject *fence); + +/* + * Check if the fence object is signaled. This function can be substantially + * more expensive to call than the above function, but will not return a false + * negative answer. The argument "flush_type" sets the types that the + * underlying mechanism must make sure will eventually signal. + */ +extern int driFenceSignaledType(struct _DriFenceObject *fence, + uint32_t flush_type, uint32_t *signaled); + +/* + * Convenience functions. + */ + +static inline int driFenceSignaled(struct _DriFenceObject *fence, + uint32_t flush_type) +{ + uint32_t signaled_types; + int ret = driFenceSignaledType(fence, flush_type, &signaled_types); + if (ret) + return 0; + return ((signaled_types & flush_type) == flush_type); +} + +static inline int driFenceSignaledCached(struct _DriFenceObject *fence, + uint32_t flush_type) +{ + uint32_t signaled_types = + driFenceSignaledTypeCached(fence); + + return ((signaled_types & flush_type) == flush_type); +} + +/* + * Reference a fence object. + */ +extern struct _DriFenceObject *driFenceReference(struct _DriFenceObject *fence); + +/* + * Unreference a fence object. The fence object pointer will be reset to NULL. + */ + +extern void driFenceUnReference(struct _DriFenceObject **pFence); + + +/* + * Wait for a fence to signal the indicated fence_type. + * If "lazy_hint" is true, it indicates that the wait may sleep to avoid + * busy-wait polling. + */ +extern int driFenceFinish(struct _DriFenceObject *fence, uint32_t fence_type, + int lazy_hint); + +/* + * Create a DriFenceObject for manager "mgr". + * + * "private" is a pointer that should be used for the callbacks in + * struct _DriFenceMgrCreateInfo. + * + * if private_size is nonzero, then the info stored at *private, with size + * private size will be copied and the fence manager will instead use a + * pointer to the copied data for the callbacks in + * struct _DriFenceMgrCreateInfo. In that case, the object pointed to by + * "private" may be destroyed after the call to driFenceCreate. + */ +extern struct _DriFenceObject *driFenceCreate(struct _DriFenceMgr *mgr, + uint32_t fence_class, + uint32_t fence_type, + void *private, + size_t private_size); + +extern uint32_t driFenceType(struct _DriFenceObject *fence); + +/* + * Fence creations are ordered. If a fence signals a fence_type, + * it is safe to assume that all fences of the same class that was + * created before that fence has signaled the same type. + */ + +#define DRI_FENCE_CLASS_ORDERED (1 << 0) + +struct _DriFenceMgrCreateInfo { + uint32_t flags; + uint32_t num_classes; + int (*signaled) (struct _DriFenceMgr *mgr, void *private, uint32_t flush_type, + uint32_t *signaled_type); + int (*finish) (struct _DriFenceMgr *mgr, void *private, uint32_t fence_type, int lazy_hint); + int (*unreference) (struct _DriFenceMgr *mgr, void **private); +}; + +extern struct _DriFenceMgr * +driFenceMgrCreate(const struct _DriFenceMgrCreateInfo *info); + +void +driFenceMgrUnReference(struct _DriFenceMgr **pMgr); + +extern struct _DriFenceMgr * +driFenceMgrTTMInit(int fd); + +#endif diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c new file mode 100644 index 0000000000..a80555c9c7 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/ws_dri_mallocpool.c @@ -0,0 +1,161 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellström + */ + +#include +#include +#include +#include "pipe/p_debug.h" +#include "glthread.h" +#include "ws_dri_bufpool.h" +#include "ws_dri_bufmgr.h" + +static void * +pool_create(struct _DriBufferPool *pool, + unsigned long size, uint64_t flags, unsigned hint, + unsigned alignment) +{ + unsigned long *private = malloc(size + 2*sizeof(unsigned long)); + if ((flags & DRM_BO_MASK_MEM) != DRM_BO_FLAG_MEM_LOCAL) + abort(); + + *private = size; + return (void *)private; +} + + +static int +pool_destroy(struct _DriBufferPool *pool, void *private) +{ + free(private); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *pool, void *private, + _glthread_Mutex *mutex, int lazy) +{ + return 0; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + *virtual = (void *)((unsigned long *)private + 2); + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + /* + * BUG + */ + abort(); + return 0UL; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + /* + * BUG + */ + abort(); +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + return DRM_BO_FLAG_MEM_LOCAL | DRM_BO_FLAG_CACHED; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + return *(unsigned long *) private; +} + + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + abort(); + return 0UL; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + abort(); + return NULL; +} + +static void +pool_takedown(struct _DriBufferPool *pool) +{ + free(pool); +} + + +struct _DriBufferPool * +driMallocPoolInit(void) +{ + struct _DriBufferPool *pool; + + pool = (struct _DriBufferPool *) malloc(sizeof(*pool)); + if (!pool) + return NULL; + + pool->data = NULL; + pool->fd = -1; + pool->map = &pool_map; + pool->unmap = &pool_unmap; + pool->destroy = &pool_destroy; + pool->offset = &pool_offset; + pool->poolOffset = &pool_poolOffset; + pool->flags = &pool_flags; + pool->size = &pool_size; + pool->create = &pool_create; + pool->fence = &pool_fence; + pool->kernel = &pool_kernel; + pool->validate = NULL; + pool->waitIdle = &pool_waitIdle; + pool->takeDown = &pool_takedown; + return pool; +} diff --git a/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c b/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c new file mode 100644 index 0000000000..dfcf6d6b19 --- /dev/null +++ b/src/gallium/winsys/drm/intel/common/ws_dri_slabpool.c @@ -0,0 +1,968 @@ +/************************************************************************** + * + * Copyright 2006-2008 Tungsten Graphics, Inc., Cedar Park, TX., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Thomas Hellstrom + */ + +#include +#include +#include +#include +#include +#include "ws_dri_bufpool.h" +#include "ws_dri_fencemgr.h" +#include "ws_dri_bufmgr.h" +#include "glthread.h" + +#define DRI_SLABPOOL_ALLOC_RETRIES 100 + +struct _DriSlab; + +struct _DriSlabBuffer { + int isSlabBuffer; + drmBO *bo; + struct _DriFenceObject *fence; + struct _DriSlab *parent; + drmMMListHead head; + uint32_t mapCount; + uint32_t start; + uint32_t fenceType; + int unFenced; + _glthread_Cond event; +}; + +struct _DriKernelBO { + int fd; + drmBO bo; + drmMMListHead timeoutHead; + drmMMListHead head; + struct timeval timeFreed; + uint32_t pageAlignment; + void *virtual; +}; + +struct _DriSlab{ + drmMMListHead head; + drmMMListHead freeBuffers; + uint32_t numBuffers; + uint32_t numFree; + struct _DriSlabBuffer *buffers; + struct _DriSlabSizeHeader *header; + struct _DriKernelBO *kbo; +}; + + +struct _DriSlabSizeHeader { + drmMMListHead slabs; + drmMMListHead freeSlabs; + drmMMListHead delayedBuffers; + uint32_t numDelayed; + struct _DriSlabPool *slabPool; + uint32_t bufSize; + _glthread_Mutex mutex; +}; + +struct _DriFreeSlabManager { + struct timeval slabTimeout; + struct timeval checkInterval; + struct timeval nextCheck; + drmMMListHead timeoutList; + drmMMListHead unCached; + drmMMListHead cached; + _glthread_Mutex mutex; +}; + + +struct _DriSlabPool { + + /* + * The data of this structure remains constant after + * initialization and thus needs no mutex protection. + */ + + struct _DriFreeSlabManager *fMan; + uint64_t proposedFlags; + uint64_t validMask; + uint32_t *bucketSizes; + uint32_t numBuckets; + uint32_t pageSize; + int fd; + int pageAlignment; + int maxSlabSize; + int desiredNumBuffers; + struct _DriSlabSizeHeader *headers; +}; + +/* + * FIXME: Perhaps arrange timeout slabs in size buckets for fast + * retreival?? + */ + + +static inline int +driTimeAfterEq(struct timeval *arg1, struct timeval *arg2) +{ + return ((arg1->tv_sec > arg2->tv_sec) || + ((arg1->tv_sec == arg2->tv_sec) && + (arg1->tv_usec > arg2->tv_usec))); +} + +static inline void +driTimeAdd(struct timeval *arg, struct timeval *add) +{ + unsigned int sec; + + arg->tv_sec += add->tv_sec; + arg->tv_usec += add->tv_usec; + sec = arg->tv_usec / 1000000; + arg->tv_sec += sec; + arg->tv_usec -= sec*1000000; +} + +static void +driFreeKernelBO(struct _DriKernelBO *kbo) +{ + if (!kbo) + return; + + (void) drmBOUnreference(kbo->fd, &kbo->bo); + free(kbo); +} + + +static void +driFreeTimeoutKBOsLocked(struct _DriFreeSlabManager *fMan, + struct timeval *time) +{ + drmMMListHead *list, *next; + struct _DriKernelBO *kbo; + + if (!driTimeAfterEq(time, &fMan->nextCheck)) + return; + + for (list = fMan->timeoutList.next, next = list->next; + list != &fMan->timeoutList; + list = next, next = list->next) { + + kbo = DRMLISTENTRY(struct _DriKernelBO, list, timeoutHead); + + if (!driTimeAfterEq(time, &kbo->timeFreed)) + break; + + DRMLISTDELINIT(&kbo->timeoutHead); + DRMLISTDELINIT(&kbo->head); + driFreeKernelBO(kbo); + } + + fMan->nextCheck = *time; + driTimeAdd(&fMan->nextCheck, &fMan->checkInterval); +} + + +/* + * Add a _DriKernelBO to the free slab manager. + * This means that it is available for reuse, but if it's not + * reused in a while, it will be freed. + */ + +static void +driSetKernelBOFree(struct _DriFreeSlabManager *fMan, + struct _DriKernelBO *kbo) +{ + struct timeval time; + + _glthread_LOCK_MUTEX(fMan->mutex); + gettimeofday(&time, NULL); + driTimeAdd(&time, &fMan->slabTimeout); + + kbo->timeFreed = time; + + if (kbo->bo.flags & DRM_BO_FLAG_CACHED) + DRMLISTADD(&kbo->head, &fMan->cached); + else + DRMLISTADD(&kbo->head, &fMan->unCached); + + DRMLISTADDTAIL(&kbo->timeoutHead, &fMan->timeoutList); + driFreeTimeoutKBOsLocked(fMan, &time); + + _glthread_UNLOCK_MUTEX(fMan->mutex); +} + +/* + * Get a _DriKernelBO for us to use as storage for a slab. + * + */ + +static struct _DriKernelBO * +driAllocKernelBO(struct _DriSlabSizeHeader *header) + +{ + struct _DriSlabPool *slabPool = header->slabPool; + struct _DriFreeSlabManager *fMan = slabPool->fMan; + drmMMListHead *list, *next, *head; + uint32_t size = header->bufSize * slabPool->desiredNumBuffers; + struct _DriKernelBO *kbo; + struct _DriKernelBO *kboTmp; + int ret; + + /* + * FIXME: We should perhaps allow some variation in slabsize in order + * to efficiently reuse slabs. + */ + + size = (size <= slabPool->maxSlabSize) ? size : slabPool->maxSlabSize; + size = (size + slabPool->pageSize - 1) & ~(slabPool->pageSize - 1); + _glthread_LOCK_MUTEX(fMan->mutex); + + kbo = NULL; + + retry: + head = (slabPool->proposedFlags & DRM_BO_FLAG_CACHED) ? + &fMan->cached : &fMan->unCached; + + for (list = head->next, next = list->next; + list != head; + list = next, next = list->next) { + + kboTmp = DRMLISTENTRY(struct _DriKernelBO, list, head); + + if ((kboTmp->bo.size == size) && + (slabPool->pageAlignment == 0 || + (kboTmp->pageAlignment % slabPool->pageAlignment) == 0)) { + + if (!kbo) + kbo = kboTmp; + + if ((kbo->bo.proposedFlags ^ slabPool->proposedFlags) == 0) + break; + + } + } + + if (kbo) { + DRMLISTDELINIT(&kbo->head); + DRMLISTDELINIT(&kbo->timeoutHead); + } + + _glthread_UNLOCK_MUTEX(fMan->mutex); + + if (kbo) { + uint64_t new_mask = kbo->bo.proposedFlags ^ slabPool->proposedFlags; + + ret = 0; + if (new_mask) { + ret = drmBOSetStatus(kbo->fd, &kbo->bo, slabPool->proposedFlags, + new_mask, DRM_BO_HINT_DONT_FENCE, 0, 0); + } + if (ret == 0) + return kbo; + + driFreeKernelBO(kbo); + kbo = NULL; + goto retry; + } + + kbo = calloc(1, sizeof(struct _DriKernelBO)); + if (!kbo) + return NULL; + + kbo->fd = slabPool->fd; + DRMINITLISTHEAD(&kbo->head); + DRMINITLISTHEAD(&kbo->timeoutHead); + ret = drmBOCreate(kbo->fd, size, slabPool->pageAlignment, NULL, + slabPool->proposedFlags, + DRM_BO_HINT_DONT_FENCE, &kbo->bo); + if (ret) + goto out_err0; + + ret = drmBOMap(kbo->fd, &kbo->bo, + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, &kbo->virtual); + + if (ret) + goto out_err1; + + ret = drmBOUnmap(kbo->fd, &kbo->bo); + if (ret) + goto out_err1; + + return kbo; + + out_err1: + drmBOUnreference(kbo->fd, &kbo->bo); + out_err0: + free(kbo); + return NULL; +} + + +static int +driAllocSlab(struct _DriSlabSizeHeader *header) +{ + struct _DriSlab *slab; + struct _DriSlabBuffer *buf; + uint32_t numBuffers; + int ret; + int i; + + slab = calloc(1, sizeof(*slab)); + if (!slab) + return -ENOMEM; + + slab->kbo = driAllocKernelBO(header); + if (!slab->kbo) { + ret = -ENOMEM; + goto out_err0; + } + + numBuffers = slab->kbo->bo.size / header->bufSize; + + slab->buffers = calloc(numBuffers, sizeof(*slab->buffers)); + if (!slab->buffers) { + ret = -ENOMEM; + goto out_err1; + } + + DRMINITLISTHEAD(&slab->head); + DRMINITLISTHEAD(&slab->freeBuffers); + slab->numBuffers = numBuffers; + slab->numFree = 0; + slab->header = header; + + buf = slab->buffers; + for (i=0; i < numBuffers; ++i) { + buf->parent = slab; + buf->start = i* header->bufSize; + buf->mapCount = 0; + buf->isSlabBuffer = 1; + _glthread_INIT_COND(buf->event); + DRMLISTADDTAIL(&buf->head, &slab->freeBuffers); + slab->numFree++; + buf++; + } + + DRMLISTADDTAIL(&slab->head, &header->slabs); + + return 0; + + out_err1: + driSetKernelBOFree(header->slabPool->fMan, slab->kbo); + free(slab->buffers); + out_err0: + free(slab); + return ret; +} + +/* + * Delete a buffer from the slab header delayed list and put + * it on the slab free list. + */ + +static void +driSlabFreeBufferLocked(struct _DriSlabBuffer *buf) +{ + struct _DriSlab *slab = buf->parent; + struct _DriSlabSizeHeader *header = slab->header; + drmMMListHead *list = &buf->head; + + DRMLISTDEL(list); + DRMLISTADDTAIL(list, &slab->freeBuffers); + slab->numFree++; + + if (slab->head.next == &slab->head) + DRMLISTADDTAIL(&slab->head, &header->slabs); + + if (slab->numFree == slab->numBuffers) { + list = &slab->head; + DRMLISTDEL(list); + DRMLISTADDTAIL(list, &header->freeSlabs); + } + + if (header->slabs.next == &header->slabs || + slab->numFree != slab->numBuffers) { + + drmMMListHead *next; + struct _DriFreeSlabManager *fMan = header->slabPool->fMan; + + for (list = header->freeSlabs.next, next = list->next; + list != &header->freeSlabs; + list = next, next = list->next) { + + slab = DRMLISTENTRY(struct _DriSlab, list, head); + + DRMLISTDELINIT(list); + driSetKernelBOFree(fMan, slab->kbo); + free(slab->buffers); + free(slab); + } + } +} + +static void +driSlabCheckFreeLocked(struct _DriSlabSizeHeader *header, int wait) +{ + drmMMListHead *list, *prev, *first; + struct _DriSlabBuffer *buf; + struct _DriSlab *slab; + int firstWasSignaled = 1; + int signaled; + int i; + int ret; + + /* + * Rerun the freeing test if the youngest tested buffer + * was signaled, since there might be more idle buffers + * in the delay list. + */ + + while (firstWasSignaled) { + firstWasSignaled = 0; + signaled = 0; + first = header->delayedBuffers.next; + + /* Only examine the oldest 1/3 of delayed buffers: + */ + if (header->numDelayed > 3) { + for (i = 0; i < header->numDelayed; i += 3) { + first = first->next; + } + } + + for (list = first, prev = list->prev; + list != &header->delayedBuffers; + list = prev, prev = list->prev) { + buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); + slab = buf->parent; + + if (!signaled) { + if (wait) { + ret = driFenceFinish(buf->fence, buf->fenceType, 0); + if (ret) + break; + signaled = 1; + wait = 0; + } else { + signaled = driFenceSignaled(buf->fence, buf->fenceType); + } + if (signaled) { + if (list == first) + firstWasSignaled = 1; + driFenceUnReference(&buf->fence); + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + } else if (driFenceSignaledCached(buf->fence, buf->fenceType)) { + driFenceUnReference(&buf->fence); + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + } + } +} + + +static struct _DriSlabBuffer * +driSlabAllocBuffer(struct _DriSlabSizeHeader *header) +{ + static struct _DriSlabBuffer *buf; + struct _DriSlab *slab; + drmMMListHead *list; + int count = DRI_SLABPOOL_ALLOC_RETRIES; + + _glthread_LOCK_MUTEX(header->mutex); + while(header->slabs.next == &header->slabs && count > 0) { + driSlabCheckFreeLocked(header, 0); + if (header->slabs.next != &header->slabs) + break; + + _glthread_UNLOCK_MUTEX(header->mutex); + if (count != DRI_SLABPOOL_ALLOC_RETRIES) + usleep(1); + _glthread_LOCK_MUTEX(header->mutex); + (void) driAllocSlab(header); + count--; + } + + list = header->slabs.next; + if (list == &header->slabs) { + _glthread_UNLOCK_MUTEX(header->mutex); + return NULL; + } + slab = DRMLISTENTRY(struct _DriSlab, list, head); + if (--slab->numFree == 0) + DRMLISTDELINIT(list); + + list = slab->freeBuffers.next; + DRMLISTDELINIT(list); + + _glthread_UNLOCK_MUTEX(header->mutex); + buf = DRMLISTENTRY(struct _DriSlabBuffer, list, head); + return buf; +} + +static void * +pool_create(struct _DriBufferPool *driPool, unsigned long size, + uint64_t flags, unsigned hint, unsigned alignment) +{ + struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; + struct _DriSlabSizeHeader *header; + struct _DriSlabBuffer *buf; + void *dummy; + int i; + int ret; + + /* + * FIXME: Check for compatibility. + */ + + header = pool->headers; + for (i=0; inumBuckets; ++i) { + if (header->bufSize >= size) + break; + header++; + } + + if (i < pool->numBuckets) + return driSlabAllocBuffer(header); + + + /* + * Fall back to allocate a buffer object directly from DRM. + * and wrap it in a driBO structure. + */ + + + buf = calloc(1, sizeof(*buf)); + + if (!buf) + return NULL; + + buf->bo = calloc(1, sizeof(*buf->bo)); + if (!buf->bo) + goto out_err0; + + if (alignment) { + if ((alignment < pool->pageSize) && (pool->pageSize % alignment)) + goto out_err1; + if ((alignment > pool->pageSize) && (alignment % pool->pageSize)) + goto out_err1; + } + + ret = drmBOCreate(pool->fd, size, alignment / pool->pageSize, NULL, + flags, hint, buf->bo); + if (ret) + goto out_err1; + + ret = drmBOMap(pool->fd, buf->bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, + 0, &dummy); + if (ret) + goto out_err2; + + ret = drmBOUnmap(pool->fd, buf->bo); + if (ret) + goto out_err2; + + return buf; + out_err2: + drmBOUnreference(pool->fd, buf->bo); + out_err1: + free(buf->bo); + out_err0: + free(buf); + return NULL; +} + +static int +pool_destroy(struct _DriBufferPool *driPool, void *private) +{ + struct _DriSlabBuffer *buf = + (struct _DriSlabBuffer *) private; + struct _DriSlab *slab; + struct _DriSlabSizeHeader *header; + + if (!buf->isSlabBuffer) { + struct _DriSlabPool *pool = (struct _DriSlabPool *) driPool->data; + int ret; + + ret = drmBOUnreference(pool->fd, buf->bo); + free(buf->bo); + free(buf); + return ret; + } + + slab = buf->parent; + header = slab->header; + + _glthread_LOCK_MUTEX(header->mutex); + buf->unFenced = 0; + buf->mapCount = 0; + + if (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)) { + DRMLISTADDTAIL(&buf->head, &header->delayedBuffers); + header->numDelayed++; + } else { + if (buf->fence) + driFenceUnReference(&buf->fence); + driSlabFreeBufferLocked(buf); + } + + _glthread_UNLOCK_MUTEX(header->mutex); + return 0; +} + +static int +pool_waitIdle(struct _DriBufferPool *driPool, void *private, + _glthread_Mutex *mutex, int lazy) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + while(buf->unFenced) + _glthread_COND_WAIT(buf->event, *mutex); + + if (!buf->fence) + return 0; + + driFenceFinish(buf->fence, buf->fenceType, lazy); + driFenceUnReference(&buf->fence); + + return 0; +} + +static int +pool_map(struct _DriBufferPool *pool, void *private, unsigned flags, + int hint, _glthread_Mutex *mutex, void **virtual) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + int busy; + + if (buf->isSlabBuffer) + busy = buf->unFenced || (buf->fence && !driFenceSignaledCached(buf->fence, buf->fenceType)); + else + busy = buf->fence && !driFenceSignaled(buf->fence, buf->fenceType); + + + if (busy) { + if (hint & DRM_BO_HINT_DONT_BLOCK) + return -EBUSY; + else { + (void) pool_waitIdle(pool, private, mutex, 0); + } + } + + ++buf->mapCount; + *virtual = (buf->isSlabBuffer) ? + (void *) ((uint8_t *) buf->parent->kbo->virtual + buf->start) : + (void *) buf->bo->virtual; + + return 0; +} + +static int +pool_unmap(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + --buf->mapCount; + if (buf->mapCount == 0 && buf->isSlabBuffer) + _glthread_COND_BROADCAST(buf->event); + + return 0; +} + +static unsigned long +pool_offset(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + struct _DriSlab *slab; + struct _DriSlabSizeHeader *header; + + if (!buf->isSlabBuffer) { + assert(buf->bo->proposedFlags & DRM_BO_FLAG_NO_MOVE); + return buf->bo->offset; + } + + slab = buf->parent; + header = slab->header; + + (void) header; + assert(header->slabPool->proposedFlags & DRM_BO_FLAG_NO_MOVE); + return slab->kbo->bo.offset + buf->start; +} + +static unsigned long +pool_poolOffset(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + return buf->start; +} + +static uint64_t +pool_flags(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + if (!buf->isSlabBuffer) + return buf->bo->flags; + + return buf->parent->kbo->bo.flags; +} + +static unsigned long +pool_size(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + if (!buf->isSlabBuffer) + return buf->bo->size; + + return buf->parent->header->bufSize; +} + +static int +pool_fence(struct _DriBufferPool *pool, void *private, + struct _DriFenceObject *fence) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + drmBO *bo; + + if (buf->fence) + driFenceUnReference(&buf->fence); + + buf->fence = driFenceReference(fence); + bo = (buf->isSlabBuffer) ? + &buf->parent->kbo->bo: + buf->bo; + buf->fenceType = bo->fenceFlags; + + buf->unFenced = 0; + _glthread_COND_BROADCAST(buf->event); + + return 0; +} + +static drmBO * +pool_kernel(struct _DriBufferPool *pool, void *private) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + return (buf->isSlabBuffer) ? &buf->parent->kbo->bo : buf->bo; +} + +static int +pool_validate(struct _DriBufferPool *pool, void *private, + _glthread_Mutex *mutex) +{ + struct _DriSlabBuffer *buf = (struct _DriSlabBuffer *) private; + + if (!buf->isSlabBuffer) + return 0; + + while(buf->mapCount != 0) + _glthread_COND_WAIT(buf->event, *mutex); + + buf->unFenced = 1; + return 0; +} + + +struct _DriFreeSlabManager * +driInitFreeSlabManager(uint32_t checkIntervalMsec, uint32_t slabTimeoutMsec) +{ + struct _DriFreeSlabManager *tmp; + + tmp = calloc(1, sizeof(*tmp)); + if (!tmp) + return NULL; + + _glthread_INIT_MUTEX(tmp->mutex); + _glthread_LOCK_MUTEX(tmp->mutex); + tmp->slabTimeout.tv_usec = slabTimeoutMsec*1000; + tmp->slabTimeout.tv_sec = tmp->slabTimeout.tv_usec / 1000000; + tmp->slabTimeout.tv_usec -= tmp->slabTimeout.tv_sec*1000000; + + tmp->checkInterval.tv_usec = checkIntervalMsec*1000; + tmp->checkInterval.tv_sec = tmp->checkInterval.tv_usec / 1000000; + tmp->checkInterval.tv_usec -= tmp->checkInterval.tv_sec*1000000; + + gettimeofday(&tmp->nextCheck, NULL); + driTimeAdd(&tmp->nextCheck, &tmp->checkInterval); + DRMINITLISTHEAD(&tmp->timeoutList); + DRMINITLISTHEAD(&tmp->unCached); + DRMINITLISTHEAD(&tmp->cached); + _glthread_UNLOCK_MUTEX(tmp->mutex); + + return tmp; +} + +void +driFinishFreeSlabManager(struct _DriFreeSlabManager *fMan) +{ + struct timeval time; + + time = fMan->nextCheck; + driTimeAdd(&time, &fMan->checkInterval); + + _glthread_LOCK_MUTEX(fMan->mutex); + driFreeTimeoutKBOsLocked(fMan, &time); + _glthread_UNLOCK_MUTEX(fMan->mutex); + + assert(fMan->timeoutList.next == &fMan->timeoutList); + assert(fMan->unCached.next == &fMan->unCached); + assert(fMan->cached.next == &fMan->cached); + + free(fMan); +} + +static void +driInitSizeHeader(struct _DriSlabPool *pool, uint32_t size, + struct _DriSlabSizeHeader *header) +{ + _glthread_INIT_MUTEX(header->mutex); + _glthread_LOCK_MUTEX(header->mutex); + + DRMINITLISTHEAD(&header->slabs); + DRMINITLISTHEAD(&header->freeSlabs); + DRMINITLISTHEAD(&header->delayedBuffers); + + header->numDelayed = 0; + header->slabPool = pool; + header->bufSize = size; + + _glthread_UNLOCK_MUTEX(header->mutex); +} + +static void +driFinishSizeHeader(struct _DriSlabSizeHeader *header) +{ + drmMMListHead *list, *next; + struct _DriSlabBuffer *buf; + + _glthread_LOCK_MUTEX(header->mutex); + for (list = header->delayedBuffers.next, next = list->next; + list != &header->delayedBuffers; + list = next, next = list->next) { + + buf = DRMLISTENTRY(struct _DriSlabBuffer, list , head); + if (buf->fence) { + (void) driFenceFinish(buf->fence, buf->fenceType, 0); + driFenceUnReference(&buf->fence); + } + header->numDelayed--; + driSlabFreeBufferLocked(buf); + } + _glthread_UNLOCK_MUTEX(header->mutex); +} + +static void +pool_takedown(struct _DriBufferPool *driPool) +{ + struct _DriSlabPool *pool = driPool->data; + int i; + + for (i=0; inumBuckets; ++i) { + driFinishSizeHeader(&pool->headers[i]); + } + + free(pool->headers); + free(pool->bucketSizes); + free(pool); + free(driPool); +} + +struct _DriBufferPool * +driSlabPoolInit(int fd, uint64_t flags, + uint64_t validMask, + uint32_t smallestSize, + uint32_t numSizes, + uint32_t desiredNumBuffers, + uint32_t maxSlabSize, + uint32_t pageAlignment, + struct _DriFreeSlabManager *fMan) +{ + struct _DriBufferPool *driPool; + struct _DriSlabPool *pool; + uint32_t i; + + driPool = calloc(1, sizeof(*driPool)); + if (!driPool) + return NULL; + + pool = calloc(1, sizeof(*pool)); + if (!pool) + goto out_err0; + + pool->bucketSizes = calloc(numSizes, sizeof(*pool->bucketSizes)); + if (!pool->bucketSizes) + goto out_err1; + + pool->headers = calloc(numSizes, sizeof(*pool->headers)); + if (!pool->headers) + goto out_err2; + + pool->fMan = fMan; + pool->proposedFlags = flags; + pool->validMask = validMask; + pool->numBuckets = numSizes; + pool->pageSize = getpagesize(); + pool->fd = fd; + pool->pageAlignment = pageAlignment; + pool->maxSlabSize = maxSlabSize; + pool->desiredNumBuffers = desiredNumBuffers; + + for (i=0; inumBuckets; ++i) { + pool->bucketSizes[i] = (smallestSize << i); + driInitSizeHeader(pool, pool->bucketSizes[i], + &pool->headers[i]); + } + + driPool->data = (void *) pool; + driPool->map = &pool_map; + driPool->unmap = &pool_unmap; + driPool->destroy = &pool_destroy; + driPool->offset = &pool_offset; + driPool->poolOffset = &pool_poolOffset; + driPool->flags = &pool_flags; + driPool->size = &pool_size; + driPool->create = &pool_create; + driPool->fence = &pool_fence; + driPool->kernel = &pool_kernel; + driPool->validate = &pool_validate; + driPool->waitIdle = &pool_waitIdle; + driPool->takeDown = &pool_takedown; + + return driPool; + + out_err2: + free(pool->bucketSizes); + out_err1: + free(pool); + out_err0: + free(driPool); + + return NULL; +} diff --git a/src/gallium/winsys/drm/intel/dri/Makefile b/src/gallium/winsys/drm/intel/dri/Makefile new file mode 100644 index 0000000000..2046441a22 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/Makefile @@ -0,0 +1,33 @@ +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = i915_dri.so +LIBNAME_EGL = egl_i915_dri.so + +PIPE_DRIVERS = \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + ../common/libinteldrm.a \ + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a + + +DRIVER_SOURCES = \ + intel_winsys_softpipe.c \ + intel_swapbuffers.c \ + intel_context.c \ + intel_lock.c \ + intel_screen.c + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +DRIVER_DEFINES = -I../common $(shell pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") + +include ../../Makefile.template + +#intel_tex_layout.o: $(TOP)/src/mesa/drivers/dri/intel/intel_tex_layout.c + +symlinks: diff --git a/src/gallium/winsys/drm/intel/dri/SConscript b/src/gallium/winsys/drm/intel/dri/SConscript new file mode 100644 index 0000000000..6a4f50afcc --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/SConscript @@ -0,0 +1,41 @@ +Import('*') + +if 'mesa' in env['statetrackers']: + + env = drienv.Clone() + + env.Append(CPPPATH = [ + '../intel', + 'server' + ]) + + #MINIGLX_SOURCES = server/intel_dri.c + + DRIVER_SOURCES = [ + 'intel_winsys_pipe.c', + 'intel_winsys_softpipe.c', + 'intel_winsys_i915.c', + 'intel_batchbuffer.c', + 'intel_swapbuffers.c', + 'intel_context.c', + 'intel_lock.c', + 'intel_screen.c', + 'intel_batchpool.c', + ] + + sources = \ + COMMON_GALLIUM_SOURCES + \ + COMMON_BM_SOURCES + \ + DRIVER_SOURCES + + drivers = [ + softpipe, + i915simple + ] + + # TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions + env.SharedLibrary( + target ='i915tex_dri.so', + source = sources, + LIBS = drivers + mesa + auxiliaries + env['LIBS'], + ) diff --git a/src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h b/src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h new file mode 100644 index 0000000000..3e95326168 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_batchbuffer.h @@ -0,0 +1,24 @@ +#ifndef INTEL_BATCHBUFFER_H +#define INTEL_BATCHBUFFER_H + +#include "intel_be_batchbuffer.h" + +/* + * Need to redefine the BATCH defines + */ + +#undef BEGIN_BATCH +#define BEGIN_BATCH(dwords, relocs) \ + (i915_batchbuffer_check(&intel->base.batch->base, dwords, relocs)) + +#undef OUT_BATCH +#define OUT_BATCH(d) \ + i915_batchbuffer_dword(&intel->base.batch->base, d) + +#undef OUT_RELOC +#define OUT_RELOC(buf,flags,mask,delta) do { \ + assert((delta) >= 0); \ + intel_be_offset_relocation(intel->base.batch, delta, buf, flags, mask); \ +} while (0) + +#endif diff --git a/src/gallium/winsys/drm/intel/dri/intel_context.c b/src/gallium/winsys/drm/intel/dri/intel_context.c new file mode 100644 index 0000000000..97ef731aaa --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_context.c @@ -0,0 +1,337 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "i830_dri.h" + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_swapbuffers.h" +#include "intel_batchbuffer.h" +#include "intel_winsys_softpipe.h" + +#include "i915simple/i915_screen.h" + +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_context.h" + +#include "utils.h" + + +#ifdef DEBUG +int __intel_debug = 0; +#endif + + +#define need_GL_ARB_multisample +#define need_GL_ARB_point_parameters +#define need_GL_ARB_texture_compression +#define need_GL_ARB_vertex_buffer_object +#define need_GL_ARB_vertex_program +#define need_GL_ARB_window_pos +#define need_GL_EXT_blend_color +#define need_GL_EXT_blend_equation_separate +#define need_GL_EXT_blend_func_separate +#define need_GL_EXT_blend_minmax +#define need_GL_EXT_cull_vertex +#define need_GL_EXT_fog_coord +#define need_GL_EXT_framebuffer_object +#define need_GL_EXT_multi_draw_arrays +#define need_GL_EXT_secondary_color +#define need_GL_NV_vertex_program +#include "extension_helper.h" + + +/** + * Extension strings exported by the intel driver. + * + * \note + * It appears that ARB_texture_env_crossbar has "disappeared" compared to the + * old i830-specific driver. + */ +const struct dri_extension card_extensions[] = { + {"GL_ARB_multisample", GL_ARB_multisample_functions}, + {"GL_ARB_multitexture", NULL}, + {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_texture_border_clamp", NULL}, + {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, + {"GL_ARB_texture_cube_map", NULL}, + {"GL_ARB_texture_env_add", NULL}, + {"GL_ARB_texture_env_combine", NULL}, + {"GL_ARB_texture_env_dot3", NULL}, + {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_rectangle", NULL}, + {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, + {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, + {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, + {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, + {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_subtract", NULL}, + {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, + {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, + {"GL_EXT_packed_depth_stencil", NULL}, + {"GL_EXT_pixel_buffer_object", NULL}, + {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, + {"GL_EXT_stencil_wrap", NULL}, + {"GL_EXT_texture_edge_clamp", NULL}, + {"GL_EXT_texture_env_combine", NULL}, + {"GL_EXT_texture_env_dot3", NULL}, + {"GL_EXT_texture_filter_anisotropic", NULL}, + {"GL_EXT_texture_lod_bias", NULL}, + {"GL_3DFX_texture_compression_FXT1", NULL}, + {"GL_APPLE_client_storage", NULL}, + {"GL_MESA_pack_invert", NULL}, + {"GL_MESA_ycbcr_texture", NULL}, + {"GL_NV_blend_square", NULL}, + {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, + {"GL_NV_vertex_program1_1", NULL}, + {"GL_SGIS_generate_mipmap", NULL }, + {NULL, NULL} +}; + + + +#ifdef DEBUG +static const struct dri_debug_control debug_control[] = { + {"ioctl", DEBUG_IOCTL}, + {"bat", DEBUG_BATCH}, + {"lock", DEBUG_LOCK}, + {"swap", DEBUG_SWAP}, + {NULL, 0} +}; +#endif + + + +static void +intel_lock_hardware(struct intel_be_context *context) +{ + struct intel_context *intel = (struct intel_context *)context; + LOCK_HARDWARE(intel); +} + +static void +intel_unlock_hardware(struct intel_be_context *context) +{ + struct intel_context *intel = (struct intel_context *)context; + UNLOCK_HARDWARE(intel); +} + +static boolean +intel_locked_hardware(struct intel_be_context *context) +{ + struct intel_context *intel = (struct intel_context *)context; + return intel->locked ? TRUE : FALSE; +} + +GLboolean +intelCreateContext(const __GLcontextModes * visual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate) +{ + struct intel_context *intel = CALLOC_STRUCT(intel_context); + __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv; + struct intel_screen *intelScreen = intel_screen(sPriv); + drmI830Sarea *saPriv = intelScreen->sarea; + int fthrottle_mode; + GLboolean havePools; + struct pipe_context *pipe; + struct st_context *st_share = NULL; + + if (sharedContextPrivate) { + st_share = ((struct intel_context *) sharedContextPrivate)->st; + } + + driContextPriv->driverPrivate = intel; + intel->intelScreen = intelScreen; + intel->driScreen = sPriv; + intel->sarea = saPriv; + + driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache, + intel->driScreen->myNum, "i915"); + + + /* + * memory pools + */ + DRM_LIGHT_LOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + // ZZZ JB should be per screen and not be done per context + havePools = intelCreatePools(sPriv); + DRM_UNLOCK(sPriv->fd, &sPriv->pSAREA->lock, driContextPriv->hHWContext); + if (!havePools) + return GL_FALSE; + + + /* Dri stuff */ + intel->hHWContext = driContextPriv->hHWContext; + intel->driFd = sPriv->fd; + intel->driHwLock = (drmLock *) & sPriv->pSAREA->lock; + + fthrottle_mode = driQueryOptioni(&intel->optionCache, "fthrottle_mode"); + intel->iw.irq_seq = -1; + intel->irqsEmitted = 0; + + intel->last_swap_fence = NULL; + intel->first_swap_fence = NULL; + +#ifdef DEBUG + __intel_debug = driParseDebugString(getenv("INTEL_DEBUG"), debug_control); +#endif + intel->base.hardware_lock = intel_lock_hardware; + intel->base.hardware_unlock = intel_unlock_hardware; + intel->base.hardware_locked = intel_locked_hardware; + + intel_be_init_context(&intel->base, &intelScreen->base); + + /* + * Pipe-related setup + */ + if (getenv("INTEL_SP")) { + /* use softpipe driver instead of hw */ + pipe = intel_create_softpipe( intel, &intelScreen->base.base ); + } + else { + switch (intel->intelScreen->deviceID) { + case PCI_CHIP_I945_G: + case PCI_CHIP_I945_GM: + case PCI_CHIP_I945_GME: + case PCI_CHIP_G33_G: + case PCI_CHIP_Q33_G: + case PCI_CHIP_Q35_G: + case PCI_CHIP_I915_G: + case PCI_CHIP_I915_GM: + pipe = i915_create_context(intelScreen->base.screen, + &intelScreen->base.base, + &intel->base.base); + break; + default: + fprintf(stderr, "Unknown PCIID %x in %s, using software driver\n", + intel->intelScreen->deviceID, __FUNCTION__); + + pipe = intel_create_softpipe( intel, &intelScreen->base.base ); + break; + } + } + + pipe->priv = intel; + + intel->st = st_create_context(pipe, visual, st_share); + + driInitExtensions( intel->st->ctx, card_extensions, GL_TRUE ); + + return GL_TRUE; +} + + +void +intelDestroyContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = intel_context(driContextPriv); + + assert(intel); /* should never be null */ + if (intel) { + st_finish(intel->st); + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(&intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + if (intel->first_swap_fence) { + driFenceFinish(intel->first_swap_fence, DRM_FENCE_TYPE_EXE, GL_TRUE); + driFenceUnReference(&intel->first_swap_fence); + intel->first_swap_fence = NULL; + } + + if (intel->intelScreen->dummyContext == intel) + intel->intelScreen->dummyContext = NULL; + + st_destroy_context(intel->st); + intel_be_destroy_context(&intel->base); + free(intel); + } +} + + +GLboolean +intelUnbindContext(__DRIcontextPrivate * driContextPriv) +{ + struct intel_context *intel = intel_context(driContextPriv); + st_flush(intel->st, PIPE_FLUSH_RENDER_CACHE, NULL); + /* XXX make_current(NULL)? */ + return GL_TRUE; +} + + +GLboolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv) +{ + if (driContextPriv) { + struct intel_context *intel = intel_context(driContextPriv); + struct intel_framebuffer *draw_fb = intel_framebuffer(driDrawPriv); + struct intel_framebuffer *read_fb = intel_framebuffer(driReadPriv); + + assert(draw_fb->stfb); + assert(read_fb->stfb); + + /* This is for situations in which we need a rendering context but + * there may not be any currently bound. + */ + intel->intelScreen->dummyContext = intel; + + st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); + + if ((intel->driDrawable != driDrawPriv) || + (intel->lastStamp != driDrawPriv->lastStamp)) { + intel->driDrawable = driDrawPriv; + intelUpdateWindowSize(driDrawPriv); + intel->lastStamp = driDrawPriv->lastStamp; + } + + /* The size of the draw buffer will have been updated above. + * If the readbuffer is a different window, check/update its size now. + */ + if (driReadPriv != driDrawPriv) { + intelUpdateWindowSize(driReadPriv); + } + + } + else { + st_make_current(NULL, NULL, NULL); + } + + return GL_TRUE; +} diff --git a/src/gallium/winsys/drm/intel/dri/intel_context.h b/src/gallium/winsys/drm/intel/dri/intel_context.h new file mode 100644 index 0000000000..5d22a422af --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_context.h @@ -0,0 +1,164 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_CONTEXT_H +#define INTEL_CONTEXT_H + +#include +#include "drm.h" + +#include "pipe/p_debug.h" + +#include "intel_screen.h" +#include "i915_drm.h" + +#include "intel_be_context.h" + + +struct pipe_context; +struct intel_context; +struct _DriBufferObject; +struct st_context; + + +#define INTEL_MAX_FIXUP 64 + +/** + * Intel rendering context, contains a state tracker and intel-specific info. + */ +struct intel_context +{ + struct intel_be_context base; + struct st_context *st; + + struct _DriFenceObject *last_swap_fence; + struct _DriFenceObject *first_swap_fence; + +// struct intel_batchbuffer *batch; + + boolean locked; + char *prevLockFile; + int prevLockLine; + + uint irqsEmitted; + drm_i915_irq_wait_t iw; + + drm_context_t hHWContext; + drmLock *driHwLock; + int driFd; + + __DRIdrawablePrivate *driDrawable; + __DRIscreenPrivate *driScreen; + struct intel_screen *intelScreen; + drmI830Sarea *sarea; + + uint lastStamp; + + /** + * Configuration cache + */ + driOptionCache optionCache; +}; + + + +/** + * Intel framebuffer. + */ +struct intel_framebuffer +{ + struct st_framebuffer *stfb; + + /* other fields TBD */ + int other; +}; + + + + +/* These are functions now: + */ +void LOCK_HARDWARE( struct intel_context *intel ); +void UNLOCK_HARDWARE( struct intel_context *intel ); + +extern char *__progname; + + + +/* ================================================================ + * Debugging: + */ +#ifdef DEBUG +extern int __intel_debug; + +#define DEBUG_SWAP 0x1 +#define DEBUG_LOCK 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_BATCH 0x8 + +#define DBG(flag, ...) do { \ + if (__intel_debug & (DEBUG_##flag)) \ + printf(__VA_ARGS__); \ +} while(0) + +#else +#define DBG(flag, ...) +#endif + + + +#define PCI_CHIP_845_G 0x2562 +#define PCI_CHIP_I830_M 0x3577 +#define PCI_CHIP_I855_GM 0x3582 +#define PCI_CHIP_I865_G 0x2572 +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GME 0x27AE +#define PCI_CHIP_G33_G 0x29C2 +#define PCI_CHIP_Q35_G 0x29B2 +#define PCI_CHIP_Q33_G 0x29D2 + + +/** Cast wrapper */ +static INLINE struct intel_context * +intel_context(__DRIcontextPrivate *driContextPriv) +{ + return (struct intel_context *) driContextPriv->driverPrivate; +} + + +/** Cast wrapper */ +static INLINE struct intel_framebuffer * +intel_framebuffer(__DRIdrawablePrivate * driDrawPriv) +{ + return (struct intel_framebuffer *) driDrawPriv->driverPrivate; +} + + +#endif diff --git a/src/gallium/winsys/drm/intel/dri/intel_lock.c b/src/gallium/winsys/drm/intel/dri/intel_lock.c new file mode 100644 index 0000000000..406284c98f --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_lock.c @@ -0,0 +1,102 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "main/glheader.h" +#include "glapi/glthread.h" +#include +#include "state_tracker/st_public.h" +#include "intel_context.h" +#include "i830_dri.h" + + + +_glthread_DECLARE_STATIC_MUTEX( lockMutex ); + + +static void +intelContendedLock(struct intel_context *intel, uint flags) +{ + __DRIdrawablePrivate *dPriv = intel->driDrawable; + __DRIscreenPrivate *sPriv = intel->driScreen; + struct intel_screen *intelScreen = intel_screen(sPriv); + drmI830Sarea *sarea = intel->sarea; + + drmGetLock(intel->driFd, intel->hHWContext, flags); + + DBG(LOCK, "%s - got contended lock\n", __progname); + + /* If the window moved, may need to set a new cliprect now. + * + * NOTE: This releases and regains the hw lock, so all state + * checking must be done *after* this call: + */ + if (dPriv) + DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv); + + if (sarea->width != intelScreen->front.width || + sarea->height != intelScreen->front.height) { + + intelUpdateScreenRotation(sPriv, sarea); + } +} + + +/* Lock the hardware and validate our state. + */ +void LOCK_HARDWARE( struct intel_context *intel ) +{ + char __ret = 0; + + _glthread_LOCK_MUTEX(lockMutex); + assert(!intel->locked); + + DRM_CAS(intel->driHwLock, intel->hHWContext, + (DRM_LOCK_HELD|intel->hHWContext), __ret); + + if (__ret) + intelContendedLock( intel, 0 ); + + DBG(LOCK, "%s - locked\n", __progname); + + intel->locked = 1; +} + + +/* Unlock the hardware using the global current context + */ +void UNLOCK_HARDWARE( struct intel_context *intel ) +{ + assert(intel->locked); + intel->locked = 0; + + DRM_UNLOCK(intel->driFd, intel->driHwLock, intel->hHWContext); + + _glthread_UNLOCK_MUTEX(lockMutex); + + DBG(LOCK, "%s - unlocked\n", __progname); +} diff --git a/src/gallium/winsys/drm/intel/dri/intel_reg.h b/src/gallium/winsys/drm/intel/dri/intel_reg.h new file mode 100644 index 0000000000..4f33bee438 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_reg.h @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef _INTEL_REG_H_ +#define _INTEL_REG_H_ + + +#define BR00_BITBLT_CLIENT 0x40000000 +#define BR00_OP_COLOR_BLT 0x10000000 +#define BR00_OP_SRC_COPY_BLT 0x10C00000 +#define BR13_SOLID_PATTERN 0x80000000 + +#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) +#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) +#define XY_COLOR_BLT_WRITE_RGB (1<<20) + +#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) +#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) +#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) + +#define MI_WAIT_FOR_EVENT ((0x3<<23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +#define MI_BATCH_BUFFER_END (0xA<<23) + + +#endif diff --git a/src/gallium/winsys/drm/intel/dri/intel_screen.c b/src/gallium/winsys/drm/intel/dri/intel_screen.c new file mode 100644 index 0000000000..46d4861e77 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_screen.c @@ -0,0 +1,607 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "utils.h" +#include "vblank.h" +#include "xmlpool.h" + +#include "intel_context.h" +#include "intel_screen.h" +#include "intel_batchbuffer.h" +#include "intel_swapbuffers.h" + +#include "i830_dri.h" +#include "ws_dri_bufpool.h" + +#include "pipe/p_context.h" +#include "pipe/p_screen.h" +#include "pipe/p_inlines.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_cb_fbo.h" + +static void +intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle); + +static void +intelCreateSurface(struct intel_screen *intelScreen, struct pipe_winsys *winsys, unsigned handle) +{ + struct pipe_screen *screen = intelScreen->base.screen; + struct pipe_texture *texture; + struct pipe_texture templat; + struct pipe_surface *surface; + struct pipe_buffer *buffer; + unsigned pitch; + + assert(intelScreen->front.cpp == 4); + + buffer = intel_be_buffer_from_handle(&intelScreen->base, + "front", handle); + + if (!buffer) + return; + + intelScreen->front.buffer = dri_bo(buffer); + + memset(&templat, 0, sizeof(templat)); + templat.tex_usage |= PIPE_TEXTURE_USAGE_DISPLAY_TARGET; + templat.target = PIPE_TEXTURE_2D; + templat.last_level = 0; + templat.depth[0] = 1; + templat.format = PIPE_FORMAT_A8R8G8B8_UNORM; + templat.width[0] = intelScreen->front.width; + templat.height[0] = intelScreen->front.height; + pf_get_block(templat.format, &templat.block); + pitch = intelScreen->front.pitch; + + texture = screen->texture_blanket(screen, + &templat, + &pitch, + buffer); + + /* Unref the buffer we don't need it anyways */ + pipe_buffer_reference(screen->winsys, &buffer, NULL); + + surface = screen->get_tex_surface(screen, + texture, + 0, + 0, + 0, + PIPE_BUFFER_USAGE_GPU_WRITE); + + intelScreen->front.texture = texture; + intelScreen->front.surface = surface; +} + +PUBLIC const char __driConfigOptions[] = + DRI_CONF_BEGIN DRI_CONF_SECTION_PERFORMANCE + DRI_CONF_FTHROTTLE_MODE(DRI_CONF_FTHROTTLE_IRQS) + DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0) + DRI_CONF_SECTION_END DRI_CONF_SECTION_QUALITY +// DRI_CONF_FORCE_S3TC_ENABLE(false) + DRI_CONF_ALLOW_LARGE_TEXTURES(1) + DRI_CONF_SECTION_END DRI_CONF_END; + +const uint __driNConfigOptions = 3; + +#ifdef USE_NEW_INTERFACE +static PFNGLXCREATECONTEXTMODES create_context_modes = NULL; +#endif /*USE_NEW_INTERFACE */ + +extern const struct dri_extension card_extensions[]; + + + + +static void +intelPrintDRIInfo(struct intel_screen * intelScreen, + __DRIscreenPrivate * sPriv, I830DRIPtr gDRIPriv) +{ + fprintf(stderr, "*** Front size: 0x%x offset: 0x%x pitch: %d\n", + intelScreen->front.size, intelScreen->front.offset, + intelScreen->front.pitch); + fprintf(stderr, "*** Memory : 0x%x\n", gDRIPriv->mem); +} + + +#if 0 +static void +intelPrintSAREA(const drmI830Sarea * sarea) +{ + fprintf(stderr, "SAREA: sarea width %d height %d\n", sarea->width, + sarea->height); + fprintf(stderr, "SAREA: pitch: %d\n", sarea->pitch); + fprintf(stderr, + "SAREA: front offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->front_offset, sarea->front_size, + (unsigned) sarea->front_handle); + fprintf(stderr, + "SAREA: back offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->back_offset, sarea->back_size, + (unsigned) sarea->back_handle); + fprintf(stderr, "SAREA: depth offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->depth_offset, sarea->depth_size, + (unsigned) sarea->depth_handle); + fprintf(stderr, "SAREA: tex offset: 0x%08x size: 0x%x handle: 0x%x\n", + sarea->tex_offset, sarea->tex_size, (unsigned) sarea->tex_handle); + fprintf(stderr, "SAREA: rotation: %d\n", sarea->rotation); + fprintf(stderr, + "SAREA: rotated offset: 0x%08x size: 0x%x\n", + sarea->rotated_offset, sarea->rotated_size); + fprintf(stderr, "SAREA: rotated pitch: %d\n", sarea->rotated_pitch); +} +#endif + + +/** + * Use the information in the sarea to update the screen parameters + * related to screen rotation. Needs to be called locked. + */ +void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea) +{ + struct intel_screen *intelScreen = intel_screen(sPriv); + + if (intelScreen->front.map) { + drmUnmap(intelScreen->front.map, intelScreen->front.size); + intelScreen->front.map = NULL; + } + + if (intelScreen->front.buffer) + driDeleteBuffers(1, &intelScreen->front.buffer); + + intelScreen->front.width = sarea->width; + intelScreen->front.height = sarea->height; + intelScreen->front.offset = sarea->front_offset; + intelScreen->front.pitch = sarea->pitch * intelScreen->front.cpp; + intelScreen->front.size = sarea->front_size; + intelScreen->front.handle = sarea->front_handle; + + assert( sarea->front_size >= + intelScreen->front.pitch * intelScreen->front.height ); + +#if 0 /* JB not important */ + if (!sarea->front_handle) + return; + + if (drmMap(sPriv->fd, + sarea->front_handle, + intelScreen->front.size, + (drmAddress *) & intelScreen->front.map) != 0) { + fprintf(stderr, "drmMap(frontbuffer) failed!\n"); + return; + } +#endif + +#if 0 /* JB */ + if (intelScreen->staticPool) { + driGenBuffers(intelScreen->staticPool, "static region", 1, + &intelScreen->front.buffer, 64, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_NO_MOVE | + DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0); + + driBOSetStatic(intelScreen->front.buffer, + intelScreen->front.offset, + intelScreen->front.pitch * intelScreen->front.height, + intelScreen->front.map, 0); + } +#else + if (intelScreen->base.staticPool) { + if (intelScreen->front.buffer) { + driBOUnReference(intelScreen->front.buffer); + pipe_surface_reference(&intelScreen->front.surface, NULL); + pipe_texture_reference(&intelScreen->front.texture, NULL); + } + intelCreateSurface(intelScreen, &intelScreen->base.base, sarea->front_bo_handle); + } +#endif +} + + +boolean +intelCreatePools(__DRIscreenPrivate * sPriv) +{ + //unsigned batchPoolSize = 1024*1024; + struct intel_screen *intelScreen = intel_screen(sPriv); + + if (intelScreen->havePools) + return GL_TRUE; + + intelScreen->havePools = GL_TRUE; + + intelUpdateScreenRotation(sPriv, intelScreen->sarea); + + return GL_TRUE; +} + +static const char * +intel_get_name( struct pipe_winsys *winsys ) +{ + return "Intel/DRI/ttm"; +} + +/* + * The state tracker (should!) keep track of whether the fake + * frontbuffer has been touched by any rendering since the last time + * we copied its contents to the real frontbuffer. Our task is easy: + */ +static void +intel_flush_frontbuffer( struct pipe_winsys *winsys, + struct pipe_surface *surf, + void *context_private) +{ + struct intel_context *intel = (struct intel_context *) context_private; + __DRIdrawablePrivate *dPriv = intel->driDrawable; + + intelDisplaySurface(dPriv, surf, NULL); +} + +static boolean +intelInitDriver(__DRIscreenPrivate * sPriv) +{ + struct intel_screen *intelScreen; + I830DRIPtr gDRIPriv = (I830DRIPtr) sPriv->pDevPriv; + + PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension = + (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface-> + getProcAddress("glxEnableExtension")); + void *const psc = sPriv->psc->screenConfigs; + + if (sPriv->devPrivSize != sizeof(I830DRIRec)) { + fprintf(stderr, + "\nERROR! sizeof(I830DRIRec) does not match passed size from device driver\n"); + return GL_FALSE; + } + + /* Allocate the private area */ + intelScreen = CALLOC_STRUCT(intel_screen); + if (!intelScreen) + return GL_FALSE; + + /* parse information in __driConfigOptions */ + driParseOptionInfo(&intelScreen->optionCache, + __driConfigOptions, __driNConfigOptions); + + sPriv->private = (void *) intelScreen; + + intelScreen->sarea = (drmI830Sarea *) (((GLubyte *) sPriv->pSAREA) + + gDRIPriv->sarea_priv_offset); + intelScreen->deviceID = gDRIPriv->deviceID; + intelScreen->front.cpp = gDRIPriv->cpp; + intelScreen->drmMinor = sPriv->drmMinor; + + assert(gDRIPriv->bitsPerPixel == 16 || + gDRIPriv->bitsPerPixel == 32); + + intelUpdateScreenRotation(sPriv, intelScreen->sarea); + + if (0) + intelPrintDRIInfo(intelScreen, sPriv, gDRIPriv); + + if (glx_enable_extension != NULL) { + (*glx_enable_extension) (psc, "GLX_SGI_swap_control"); + (*glx_enable_extension) (psc, "GLX_SGI_video_sync"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_control"); + (*glx_enable_extension) (psc, "GLX_MESA_swap_frame_usage"); + (*glx_enable_extension) (psc, "GLX_SGI_make_current_read"); + } + + intelScreen->base.base.flush_frontbuffer = intel_flush_frontbuffer; + intelScreen->base.base.get_name = intel_get_name; + intel_be_init_device(&intelScreen->base, sPriv->fd, intelScreen->deviceID); + + return GL_TRUE; +} + + +static void +intelDestroyScreen(__DRIscreenPrivate * sPriv) +{ + struct intel_screen *intelScreen = intel_screen(sPriv); + + intel_be_destroy_device(&intelScreen->base); + /* intelUnmapScreenRegions(intelScreen); */ + + FREE(intelScreen); + sPriv->private = NULL; +} + + +/** + * This is called when we need to set up GL rendering to a new X window. + */ +static boolean +intelCreateBuffer(__DRIscreenPrivate * driScrnPriv, + __DRIdrawablePrivate * driDrawPriv, + const __GLcontextModes * visual, boolean isPixmap) +{ + if (isPixmap) { + return GL_FALSE; /* not implemented */ + } + else { + enum pipe_format colorFormat, depthFormat, stencilFormat; + struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); + + if (!intelfb) + return GL_FALSE; + + if (visual->redBits == 5) + colorFormat = PIPE_FORMAT_R5G6B5_UNORM; + else + colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; + + if (visual->depthBits == 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (visual->depthBits == 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_NONE; + + if (visual->stencilBits == 8) + stencilFormat = PIPE_FORMAT_S8Z24_UNORM; + else + stencilFormat = PIPE_FORMAT_NONE; + + intelfb->stfb = st_create_framebuffer(visual, + colorFormat, + depthFormat, + stencilFormat, + driDrawPriv->w, + driDrawPriv->h, + (void*) intelfb); + if (!intelfb->stfb) { + free(intelfb); + return GL_FALSE; + } + + driDrawPriv->driverPrivate = (void *) intelfb; + return GL_TRUE; + } +} + +static void +intelDestroyBuffer(__DRIdrawablePrivate * driDrawPriv) +{ + struct intel_framebuffer *intelfb = intel_framebuffer(driDrawPriv); + assert(intelfb->stfb); + st_unreference_framebuffer(&intelfb->stfb); + free(intelfb); +} + + +/** + * Get information about previous buffer swaps. + */ +static int +intelGetSwapInfo(__DRIdrawablePrivate * dPriv, __DRIswapInfo * sInfo) +{ + if ((dPriv == NULL) || (dPriv->driverPrivate == NULL) + || (sInfo == NULL)) { + return -1; + } + + return 0; +} + + +static void +intelSetTexOffset(__DRIcontext *pDRICtx, int texname, + unsigned long long offset, int depth, uint pitch) +{ + abort(); +#if 0 + struct intel_context *intel = (struct intel_context*) + ((__DRIcontextPrivate*)pDRICtx->private)->driverPrivate; + struct gl_texture_object *tObj = _mesa_lookup_texture(&intel->ctx, texname); + struct st_texture_object *stObj = st_texture_object(tObj); + + if (!stObj) + return; + + if (stObj->pt) + st->pipe->texture_release(intel->st->pipe, &stObj->pt); + + stObj->imageOverride = GL_TRUE; + stObj->depthOverride = depth; + stObj->pitchOverride = pitch; + + if (offset) + stObj->textureOffset = offset; +#endif +} + + +static const struct __DriverAPIRec intelAPI = { + .InitDriver = intelInitDriver, + .DestroyScreen = intelDestroyScreen, + .CreateContext = intelCreateContext, + .DestroyContext = intelDestroyContext, + .CreateBuffer = intelCreateBuffer, + .DestroyBuffer = intelDestroyBuffer, + .SwapBuffers = intelSwapBuffers, + .MakeCurrent = intelMakeCurrent, + .UnbindContext = intelUnbindContext, + .GetSwapInfo = intelGetSwapInfo, + .GetMSC = driGetMSC32, + .WaitForMSC = driWaitForMSC32, + .WaitForSBC = NULL, + .SwapBuffersMSC = NULL, + .CopySubBuffer = intelCopySubBuffer, + .setTexOffset = intelSetTexOffset, +}; + + +static __GLcontextModes * +intelFillInModes(unsigned pixel_bits, unsigned depth_bits, + unsigned stencil_bits, boolean have_back_buffer) +{ + __GLcontextModes *modes; + __GLcontextModes *m; + unsigned num_modes; + unsigned depth_buffer_factor; + unsigned back_buffer_factor; + GLenum fb_format; + GLenum fb_type; + + /* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't + * support pageflipping at all. + */ + static const GLenum back_buffer_modes[] = { + GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML + }; + + uint8_t depth_bits_array[3]; + uint8_t stencil_bits_array[3]; + uint8_t msaa_samples_array[1]; + + + depth_bits_array[0] = 0; + depth_bits_array[1] = depth_bits; + depth_bits_array[2] = depth_bits; + msaa_samples_array[0] = 0; + + /* Just like with the accumulation buffer, always provide some modes + * with a stencil buffer. It will be a sw fallback, but some apps won't + * care about that. + */ + stencil_bits_array[0] = 0; + stencil_bits_array[1] = 0; + if (depth_bits == 24) + stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits; + + stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits; + + depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1; + back_buffer_factor = (have_back_buffer) ? 3 : 1; + + num_modes = depth_buffer_factor * back_buffer_factor * 4; + + if (pixel_bits == 16) { + fb_format = GL_RGB; + fb_type = GL_UNSIGNED_SHORT_5_6_5; + } + else { + fb_format = GL_BGRA; + fb_type = GL_UNSIGNED_INT_8_8_8_8_REV; + } + + modes = + (*dri_interface->createContextModes) (num_modes, + sizeof(__GLcontextModes)); + m = modes; + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, msaa_samples_array, 1, GLX_TRUE_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + if (!driFillInModes(&m, fb_format, fb_type, + depth_bits_array, stencil_bits_array, + depth_buffer_factor, back_buffer_modes, + back_buffer_factor, msaa_samples_array, 1, GLX_DIRECT_COLOR)) { + fprintf(stderr, "[%s:%u] Error creating FBConfig!\n", __func__, + __LINE__); + return NULL; + } + + /* Mark the visual as slow if there are "fake" stencil bits. + */ + for (m = modes; m != NULL; m = m->next) { + if ((m->stencilBits != 0) && (m->stencilBits != stencil_bits)) { + m->visualRating = GLX_SLOW_CONFIG; + } + } + + return modes; +} + + +/** + * This is the bootstrap function for the driver. libGL supplies all of the + * requisite information about the system, and the driver initializes itself. + * This routine also fills in the linked list pointed to by \c driver_modes + * with the \c __GLcontextModes that the driver can support for windows or + * pbuffers. + * + * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on + * failure. + */ +PUBLIC void * +__driCreateNewScreen_20050727(__DRInativeDisplay * dpy, int scrn, + __DRIscreen * psc, + const __GLcontextModes * modes, + const __DRIversion * ddx_version, + const __DRIversion * dri_version, + const __DRIversion * drm_version, + const __DRIframebuffer * frame_buffer, + drmAddress pSAREA, int fd, + int internal_api_version, + const __DRIinterfaceMethods * interface, + __GLcontextModes ** driver_modes) +{ + __DRIscreenPrivate *psp; + static const __DRIversion ddx_expected = { 1, 7, 0 }; + static const __DRIversion dri_expected = { 4, 0, 0 }; + static const __DRIversion drm_expected = { 1, 7, 0 }; + + dri_interface = interface; + + if (!driCheckDriDdxDrmVersions2("i915", + dri_version, &dri_expected, + ddx_version, &ddx_expected, + drm_version, &drm_expected)) { + return NULL; + } + + psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL, + ddx_version, dri_version, drm_version, + frame_buffer, pSAREA, fd, + internal_api_version, &intelAPI); + + if (psp != NULL) { + I830DRIPtr dri_priv = (I830DRIPtr) psp->pDevPriv; + *driver_modes = intelFillInModes(dri_priv->cpp * 8, + (dri_priv->cpp == 2) ? 16 : 24, + (dri_priv->cpp == 2) ? 0 : 8, 1); + + /* Calling driInitExtensions here, with a NULL context pointer, + * does not actually enable the extensions. It just makes sure + * that all the dispatch offsets for all the extensions that + * *might* be enables are known. This is needed because the + * dispatch offsets need to be known when _mesa_context_create + * is called, but we can't enable the extensions until we have a + * context pointer. + * + * Hello chicken. Hello egg. How are you two today? + */ + driInitExtensions(NULL, card_extensions, GL_FALSE); + } + + return (void *) psp; +} + diff --git a/src/gallium/winsys/drm/intel/dri/intel_screen.h b/src/gallium/winsys/drm/intel/dri/intel_screen.h new file mode 100644 index 0000000000..0bb43a915c --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_screen.h @@ -0,0 +1,122 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _INTEL_SCREEN_H_ +#define _INTEL_SCREEN_H_ + +#include "dri_util.h" +#include "i830_common.h" +#include "xmlconfig.h" +#include "ws_dri_bufpool.h" + +#include "pipe/p_compiler.h" + +#include "intel_be_device.h" + +struct intel_screen +{ + struct intel_be_device base; + + struct { + drm_handle_t handle; + + /* We create a static dri buffer for the frontbuffer. + */ + struct _DriBufferObject *buffer; + struct pipe_surface *surface; + struct pipe_texture *texture; + + char *map; /* memory map */ + int offset; /* from start of video mem, in bytes */ + int pitch; /* row stride, in bytes */ + int width; + int height; + int size; + int cpp; /* for front and back buffers */ + } front; + + int deviceID; + int drmMinor; + + drmI830Sarea *sarea; + + /** + * Configuration cache with default values for all contexts + */ + driOptionCache optionCache; + + boolean havePools; + + /** + * Temporary(?) context to use for SwapBuffers or other situations in + * which we need a rendering context, but none is currently bound. + */ + struct intel_context *dummyContext; + + /* + * New stuff form the i915tex integration + */ + unsigned batch_id; + + + struct pipe_winsys *winsys; +}; + + + +/** cast wrapper */ +static INLINE struct intel_screen * +intel_screen(__DRIscreenPrivate *sPriv) +{ + return (struct intel_screen *) sPriv->private; +} + + +extern void +intelUpdateScreenRotation(__DRIscreenPrivate * sPriv, drmI830Sarea * sarea); + + +extern void intelDestroyContext(__DRIcontextPrivate * driContextPriv); + +extern boolean intelUnbindContext(__DRIcontextPrivate * driContextPriv); + +extern boolean +intelMakeCurrent(__DRIcontextPrivate * driContextPriv, + __DRIdrawablePrivate * driDrawPriv, + __DRIdrawablePrivate * driReadPriv); + + +extern boolean +intelCreatePools(__DRIscreenPrivate *sPriv); + +extern boolean +intelCreateContext(const __GLcontextModes * visual, + __DRIcontextPrivate * driContextPriv, + void *sharedContextPrivate); + + +#endif diff --git a/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c new file mode 100644 index 0000000000..8a18bfd9a4 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.c @@ -0,0 +1,261 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_screen.h" +#include "intel_context.h" +#include "intel_swapbuffers.h" + +#include "intel_reg.h" + +#include "pipe/p_context.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_cb_fbo.h" + +#include "ws_dri_bufmgr.h" +#include "intel_batchbuffer.h" + +/** + * Display a colorbuffer surface in an X window. + * Used for SwapBuffers and flushing front buffer rendering. + * + * \param dPriv the window/drawable to display into + * \param surf the surface to display + * \param rect optional subrect of surface to display (may be NULL). + */ +void +intelDisplaySurface(__DRIdrawablePrivate *dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t *rect) +{ + struct intel_screen *intelScreen = intel_screen(dPriv->driScreenPriv); + struct intel_context *intel = intelScreen->dummyContext; + + DBG(SWAP, "%s\n", __FUNCTION__); + + if (!intel) { + /* XXX this is where some kind of extra/meta context could be useful */ + return; + } + + if (intel->last_swap_fence) { + driFenceFinish(intel->last_swap_fence, DRM_FENCE_TYPE_EXE, TRUE); + driFenceUnReference(&intel->last_swap_fence); + intel->last_swap_fence = NULL; + } + intel->last_swap_fence = intel->first_swap_fence; + intel->first_swap_fence = NULL; + + /* The LOCK_HARDWARE is required for the cliprects. Buffer offsets + * should work regardless. + */ + LOCK_HARDWARE(intel); + /* if this drawable isn't currently bound the LOCK_HARDWARE done on the + * current context (which is what intelScreenContext should return) might + * not get a contended lock and thus cliprects not updated (tests/manywin) + */ + if (intel_context(dPriv->driContextPriv) != intel) + DRI_VALIDATE_DRAWABLE_INFO(intel->driScreen, dPriv); + + + if (dPriv && dPriv->numClipRects) { + const int srcWidth = surf->width; + const int srcHeight = surf->height; + const int nbox = dPriv->numClipRects; + const drm_clip_rect_t *pbox = dPriv->pClipRects; + const int pitch = intelScreen->front.pitch / intelScreen->front.cpp; + const int cpp = intelScreen->front.cpp; + const int srcpitch = surf->stride / cpp; + int BR13, CMD; + int i; + + ASSERT(surf->buffer); + ASSERT(surf->cpp == cpp); + + DBG(SWAP, "screen pitch %d src surface pitch %d\n", + pitch, surf->stride); + + if (cpp == 2) { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24); + CMD = XY_SRC_COPY_BLT_CMD; + } + else { + BR13 = (pitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + } + + for (i = 0; i < nbox; i++, pbox++) { + drm_clip_rect_t box; + drm_clip_rect_t sbox; + + if (pbox->x1 > pbox->x2 || + pbox->y1 > pbox->y2 || + pbox->x2 > intelScreen->front.width || + pbox->y2 > intelScreen->front.height) { + /* invalid cliprect, skip it */ + continue; + } + + box = *pbox; + + if (rect) { + /* intersect cliprect with user-provided src rect */ + drm_clip_rect_t rrect; + + rrect.x1 = dPriv->x + rect->x1; + rrect.y1 = (dPriv->h - rect->y1 - rect->y2) + dPriv->y; + rrect.x2 = rect->x2 + rrect.x1; + rrect.y2 = rect->y2 + rrect.y1; + if (rrect.x1 > box.x1) + box.x1 = rrect.x1; + if (rrect.y1 > box.y1) + box.y1 = rrect.y1; + if (rrect.x2 < box.x2) + box.x2 = rrect.x2; + if (rrect.y2 < box.y2) + box.y2 = rrect.y2; + + if (box.x1 > box.x2 || box.y1 > box.y2) + continue; + } + + /* restrict blit to size of actually rendered area */ + if (box.x2 - box.x1 > srcWidth) + box.x2 = srcWidth + box.x1; + if (box.y2 - box.y1 > srcHeight) + box.y2 = srcHeight + box.y1; + + DBG(SWAP, "box x1 x2 y1 y2 %d %d %d %d\n", + box.x1, box.x2, box.y1, box.y2); + + sbox.x1 = box.x1 - dPriv->x; + sbox.y1 = box.y1 - dPriv->y; + + assert(box.x1 < box.x2); + assert(box.y1 < box.y2); + + /* XXX this could be done with pipe->surface_copy() */ + /* XXX should have its own batch buffer */ + if (!BEGIN_BATCH(8, 2)) { + /* + * Since we share this batch buffer with a context + * we can't flush it since that risks a GPU lockup + */ + assert(0); + continue; + } + + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((box.y1 << 16) | box.x1); + OUT_BATCH((box.y2 << 16) | box.x2); + + OUT_RELOC(intelScreen->front.buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + OUT_BATCH((sbox.y1 << 16) | sbox.x1); + OUT_BATCH((srcpitch * cpp) & 0xffff); + OUT_RELOC(dri_bo(surf->buffer), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + } + + if (intel->first_swap_fence) + driFenceUnReference(&intel->first_swap_fence); + intel->first_swap_fence = intel_be_batchbuffer_flush(intel->base.batch); + } + + UNLOCK_HARDWARE(intel); + + if (intel->lastStamp != dPriv->lastStamp) { + intelUpdateWindowSize(dPriv); + intel->lastStamp = dPriv->lastStamp; + } +} + + + +/** + * This will be called whenever the currently bound window is moved/resized. + */ +void +intelUpdateWindowSize(__DRIdrawablePrivate *dPriv) +{ + struct intel_framebuffer *intelfb = intel_framebuffer(dPriv); + assert(intelfb->stfb); + st_resize_framebuffer(intelfb->stfb, dPriv->w, dPriv->h); +} + + + +void +intelSwapBuffers(__DRIdrawablePrivate * dPriv) +{ + struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); + struct pipe_surface *back_surf; + + assert(intel_fb); + assert(intel_fb->stfb); + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { + st_notify_swapbuffers(intel_fb->stfb); + intelDisplaySurface(dPriv, back_surf, NULL); + st_notify_swapbuffers_complete(intel_fb->stfb); + } +} + + +/** + * Called via glXCopySubBufferMESA() to copy a subrect of the back + * buffer to the front buffer/screen. + */ +void +intelCopySubBuffer(__DRIdrawablePrivate * dPriv, int x, int y, int w, int h) +{ + struct intel_framebuffer *intel_fb = intel_framebuffer(dPriv); + struct pipe_surface *back_surf; + + assert(intel_fb); + assert(intel_fb->stfb); + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, + ST_SURFACE_BACK_LEFT); + if (back_surf) { + drm_clip_rect_t rect; + rect.x1 = x; + rect.y1 = y; + rect.x2 = w; + rect.y2 = h; + + st_notify_swapbuffers(intel_fb->stfb); + intelDisplaySurface(dPriv, back_surf, &rect); + } +} diff --git a/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h new file mode 100644 index 0000000000..46c9bab3af --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_swapbuffers.h @@ -0,0 +1,47 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_SWAPBUFFERS_H +#define INTEL_SWAPBUFFERS_H + + +struct pipe_surface; + + +extern void intelDisplaySurface(__DRIdrawablePrivate * dPriv, + struct pipe_surface *surf, + const drm_clip_rect_t * rect); + +extern void intelSwapBuffers(__DRIdrawablePrivate * dPriv); + +extern void intelCopySubBuffer(__DRIdrawablePrivate * dPriv, + int x, int y, int w, int h); + +extern void intelUpdateWindowSize(__DRIdrawablePrivate *dPriv); + + +#endif /* INTEL_SWAPBUFFERS_H */ diff --git a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c new file mode 100644 index 0000000000..0d98d16cf1 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c @@ -0,0 +1,82 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * + **************************************************************************/ +/* + * Authors: Keith Whitwell + */ + +#include "intel_context.h" +#include "intel_winsys_softpipe.h" +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_format.h" +#include "softpipe/sp_winsys.h" + + +struct intel_softpipe_winsys { + struct softpipe_winsys sws; + struct intel_context *intel; +}; + +/** + * Return list of surface formats supported by this driver. + */ +static boolean +intel_is_format_supported(struct softpipe_winsys *sws, + enum pipe_format format) +{ + switch(format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + case PIPE_FORMAT_R5G6B5_UNORM: + case PIPE_FORMAT_S8Z24_UNORM: + return TRUE; + default: + return FALSE; + } +} + + +/** + * Create rendering context which uses software rendering. + */ +struct pipe_context * +intel_create_softpipe( struct intel_context *intel, + struct pipe_winsys *winsys ) +{ + struct intel_softpipe_winsys *isws = CALLOC_STRUCT( intel_softpipe_winsys ); + struct pipe_screen *screen = softpipe_create_screen(winsys); + + /* Fill in this struct with callbacks that softpipe will need to + * communicate with the window system, buffer manager, etc. + */ + isws->sws.is_format_supported = intel_is_format_supported; + isws->intel = intel; + + /* Create the softpipe context: + */ + return softpipe_create( screen, winsys, &isws->sws ); +} diff --git a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h new file mode 100644 index 0000000000..5fa14cb749 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.h @@ -0,0 +1,39 @@ +/************************************************************************** + * + * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_SOFTPIPE_H +#define INTEL_SOFTPIPE_H + +struct pipe_winsys; +struct pipe_context; +struct intel_context; + +struct pipe_context * +intel_create_softpipe( struct intel_context *intel, + struct pipe_winsys *winsys ); + +#endif diff --git a/src/gallium/winsys/drm/intel/dri/server/i830_common.h b/src/gallium/winsys/drm/intel/dri/server/i830_common.h new file mode 100644 index 0000000000..3452ddb3c9 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/server/i830_common.h @@ -0,0 +1,255 @@ +/************************************************************************** + +Copyright 2001 VA Linux Systems Inc., Fremont, California. +Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas. + +All Rights Reserved. + +Permission is hereby granted, free of charge, to any person obtaining a +copy of this software and associated documentation files (the "Software"), +to deal in the Software without restriction, including without limitation +on the rights to use, copy, modify, merge, publish, distribute, sub +license, and/or sell copies of the Software, and to permit persons to whom +the Software is furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice (including the next +paragraph) shall be included in all copies or substantial portions of the +Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL +ATI, VA LINUX SYSTEMS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, +DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR +OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE +USE OR OTHER DEALINGS IN THE SOFTWARE. + +**************************************************************************/ + + +#ifndef _I830_COMMON_H_ +#define _I830_COMMON_H_ + + +#define I830_NR_TEX_REGIONS 255 /* maximum due to use of chars for next/prev */ +#define I830_LOG_MIN_TEX_REGION_SIZE 14 + + +/* Driver specific DRM command indices + * NOTE: these are not OS specific, but they are driver specific + */ +#define DRM_I830_INIT 0x00 +#define DRM_I830_FLUSH 0x01 +#define DRM_I830_FLIP 0x02 +#define DRM_I830_BATCHBUFFER 0x03 +#define DRM_I830_IRQ_EMIT 0x04 +#define DRM_I830_IRQ_WAIT 0x05 +#define DRM_I830_GETPARAM 0x06 +#define DRM_I830_SETPARAM 0x07 +#define DRM_I830_ALLOC 0x08 +#define DRM_I830_FREE 0x09 +#define DRM_I830_INIT_HEAP 0x0a +#define DRM_I830_CMDBUFFER 0x0b +#define DRM_I830_DESTROY_HEAP 0x0c +#define DRM_I830_SET_VBLANK_PIPE 0x0d +#define DRM_I830_GET_VBLANK_PIPE 0x0e +#define DRM_I830_MMIO 0x10 + +typedef struct { + enum { + I830_INIT_DMA = 0x01, + I830_CLEANUP_DMA = 0x02, + I830_RESUME_DMA = 0x03 + } func; + unsigned int mmio_offset; + int sarea_priv_offset; + unsigned int ring_start; + unsigned int ring_end; + unsigned int ring_size; + unsigned int front_offset; + unsigned int back_offset; + unsigned int depth_offset; + unsigned int w; + unsigned int h; + unsigned int pitch; + unsigned int pitch_bits; + unsigned int back_pitch; + unsigned int depth_pitch; + unsigned int cpp; + unsigned int chipset; +} drmI830Init; + +typedef struct { + drmTextureRegion texList[I830_NR_TEX_REGIONS+1]; + int last_upload; /* last time texture was uploaded */ + int last_enqueue; /* last time a buffer was enqueued */ + int last_dispatch; /* age of the most recently dispatched buffer */ + int ctxOwner; /* last context to upload state */ + /** Last context that used the buffer manager. */ + int texAge; + int pf_enabled; /* is pageflipping allowed? */ + int pf_active; + int pf_current_page; /* which buffer is being displayed? */ + int perf_boxes; /* performance boxes to be displayed */ + int width, height; /* screen size in pixels */ + + drm_handle_t front_handle; + int front_offset; + int front_size; + + drm_handle_t back_handle; + int back_offset; + int back_size; + + drm_handle_t depth_handle; + int depth_offset; + int depth_size; + + drm_handle_t tex_handle; + int tex_offset; + int tex_size; + int log_tex_granularity; + int pitch; + int rotation; /* 0, 90, 180 or 270 */ + int rotated_offset; + int rotated_size; + int rotated_pitch; + int virtualX, virtualY; + + unsigned int front_tiled; + unsigned int back_tiled; + unsigned int depth_tiled; + unsigned int rotated_tiled; + unsigned int rotated2_tiled; + + int planeA_x; + int planeA_y; + int planeA_w; + int planeA_h; + int planeB_x; + int planeB_y; + int planeB_w; + int planeB_h; + + /* Triple buffering */ + drm_handle_t third_handle; + int third_offset; + int third_size; + unsigned int third_tiled; + + /* buffer object handles for the static buffers. May change + * over the lifetime of the client, though it doesn't in our current + * implementation. + */ + unsigned int front_bo_handle; + unsigned int back_bo_handle; + unsigned int third_bo_handle; + unsigned int depth_bo_handle; +} drmI830Sarea; + +/* Flags for perf_boxes + */ +#define I830_BOX_RING_EMPTY 0x1 /* populated by kernel */ +#define I830_BOX_FLIP 0x2 /* populated by kernel */ +#define I830_BOX_WAIT 0x4 /* populated by kernel & client */ +#define I830_BOX_TEXTURE_LOAD 0x8 /* populated by kernel */ +#define I830_BOX_LOST_CONTEXT 0x10 /* populated by client */ + + +typedef struct { + int start; /* agp offset */ + int used; /* nr bytes in use */ + int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ + int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ + int num_cliprects; /* mulitpass with multiple cliprects? */ + drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ +} drmI830BatchBuffer; + +typedef struct { + char *buf; /* agp offset */ + int sz; /* nr bytes in use */ + int DR1; /* hw flags for GFX_OP_DRAWRECT_INFO */ + int DR4; /* window origin for GFX_OP_DRAWRECT_INFO*/ + int num_cliprects; /* mulitpass with multiple cliprects? */ + drm_clip_rect_t *cliprects; /* pointer to userspace cliprects */ +} drmI830CmdBuffer; + +typedef struct { + int *irq_seq; +} drmI830IrqEmit; + +typedef struct { + int irq_seq; +} drmI830IrqWait; + +typedef struct { + int param; + int *value; +} drmI830GetParam; + +#define I830_PARAM_IRQ_ACTIVE 1 +#define I830_PARAM_ALLOW_BATCHBUFFER 2 + +typedef struct { + int param; + int value; +} drmI830SetParam; + +#define I830_SETPARAM_USE_MI_BATCHBUFFER_START 1 +#define I830_SETPARAM_TEX_LRU_LOG_GRANULARITY 2 +#define I830_SETPARAM_ALLOW_BATCHBUFFER 3 + + +/* A memory manager for regions of shared memory: + */ +#define I830_MEM_REGION_AGP 1 + +typedef struct { + int region; + int alignment; + int size; + int *region_offset; /* offset from start of fb or agp */ +} drmI830MemAlloc; + +typedef struct { + int region; + int region_offset; +} drmI830MemFree; + +typedef struct { + int region; + int size; + int start; +} drmI830MemInitHeap; + +typedef struct { + int region; +} drmI830MemDestroyHeap; + +#define DRM_I830_VBLANK_PIPE_A 1 +#define DRM_I830_VBLANK_PIPE_B 2 + +typedef struct { + int pipe; +} drmI830VBlankPipe; + +#define MMIO_READ 0 +#define MMIO_WRITE 1 + +#define MMIO_REGS_IA_PRIMATIVES_COUNT 0 +#define MMIO_REGS_IA_VERTICES_COUNT 1 +#define MMIO_REGS_VS_INVOCATION_COUNT 2 +#define MMIO_REGS_GS_PRIMITIVES_COUNT 3 +#define MMIO_REGS_GS_INVOCATION_COUNT 4 +#define MMIO_REGS_CL_PRIMITIVES_COUNT 5 +#define MMIO_REGS_CL_INVOCATION_COUNT 6 +#define MMIO_REGS_PS_INVOCATION_COUNT 7 +#define MMIO_REGS_PS_DEPTH_COUNT 8 + +typedef struct { + unsigned int read_write:1; + unsigned int reg:31; + void __user *data; +} drmI830MMIO; + +#endif /* _I830_DRM_H_ */ diff --git a/src/gallium/winsys/drm/intel/dri/server/i830_dri.h b/src/gallium/winsys/drm/intel/dri/server/i830_dri.h new file mode 100644 index 0000000000..0d514b6c38 --- /dev/null +++ b/src/gallium/winsys/drm/intel/dri/server/i830_dri.h @@ -0,0 +1,62 @@ + +#ifndef _I830_DRI_H +#define _I830_DRI_H + +#include "xf86drm.h" +#include "i830_common.h" + +#define I830_MAX_DRAWABLES 256 + +#define I830_MAJOR_VERSION 1 +#define I830_MINOR_VERSION 7 +#define I830_PATCHLEVEL 2 + +#define I830_REG_SIZE 0x80000 + +typedef struct _I830DRIRec { + drm_handle_t regs; + drmSize regsSize; + + drmSize unused1; /* backbufferSize */ + drm_handle_t unused2; /* backbuffer */ + + drmSize unused3; /* depthbufferSize */ + drm_handle_t unused4; /* depthbuffer */ + + drmSize unused5; /* rotatedSize */ + drm_handle_t unused6; /* rotatedbuffer */ + + drm_handle_t unused7; /* textures */ + int unused8; /* textureSize */ + + drm_handle_t unused9; /* agp_buffers */ + drmSize unused10; /* agp_buf_size */ + + int deviceID; + int width; + int height; + int mem; + int cpp; + int bitsPerPixel; + + int unused11[8]; /* was front/back/depth/rotated offset/pitch */ + + int unused12; /* logTextureGranularity */ + int unused13; /* textureOffset */ + + int irq; + int sarea_priv_offset; +} I830DRIRec, *I830DRIPtr; + +typedef struct { + /* Nothing here yet */ + int dummy; +} I830ConfigPrivRec, *I830ConfigPrivPtr; + +typedef struct { + /* Nothing here yet */ + int dummy; +} I830DRIContextRec, *I830DRIContextPtr; + + +#endif diff --git a/src/gallium/winsys/drm/intel/egl/Makefile b/src/gallium/winsys/drm/intel/egl/Makefile new file mode 100644 index 0000000000..f0b5a44389 --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/Makefile @@ -0,0 +1,28 @@ +TOP = ../../../../../.. +include $(TOP)/configs/current + +LIBNAME = EGL_i915.so + +PIPE_DRIVERS = \ + $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ + $(TOP)/src/gallium/drivers/i915simple/libi915simple.a \ + ../common/libinteldrm.a + +DRIVER_SOURCES = \ + intel_swapbuffers.c \ + intel_context.c \ + intel_device.c \ + intel_egl.c + +C_SOURCES = \ + $(COMMON_GALLIUM_SOURCES) \ + $(DRIVER_SOURCES) + +ASM_SOURCES = + +DRIVER_DEFINES = -I../common $(shell pkg-config libdrm --atleast-version=2.3.1 \ + && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") + +include ../../Makefile.template + +symlinks: diff --git a/src/gallium/winsys/drm/intel/egl/SConscript b/src/gallium/winsys/drm/intel/egl/SConscript new file mode 100644 index 0000000000..0ad19d42a8 --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/SConscript @@ -0,0 +1,39 @@ +Import('*') + +env = drienv.Clone() + +env.Append(CPPPATH = [ + '../intel', + 'server' +]) + +#MINIGLX_SOURCES = server/intel_dri.c + +DRIVER_SOURCES = [ + 'intel_winsys_pipe.c', + 'intel_winsys_softpipe.c', + 'intel_winsys_i915.c', + 'intel_batchbuffer.c', + 'intel_swapbuffers.c', + 'intel_context.c', + 'intel_lock.c', + 'intel_screen.c', + 'intel_batchpool.c', +] + +sources = \ + COMMON_GALLIUM_SOURCES + \ + COMMON_BM_SOURCES + \ + DRIVER_SOURCES + +drivers = [ + softpipe, + i915simple +] + +# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions +env.SharedLibrary( + target ='i915tex_dri.so', + source = sources, + LIBS = drivers + mesa + auxiliaries + env['LIBS'], +) \ No newline at end of file diff --git a/src/gallium/winsys/drm/intel/egl/intel_batchbuffer.h b/src/gallium/winsys/drm/intel/egl/intel_batchbuffer.h new file mode 100644 index 0000000000..3e95326168 --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/intel_batchbuffer.h @@ -0,0 +1,24 @@ +#ifndef INTEL_BATCHBUFFER_H +#define INTEL_BATCHBUFFER_H + +#include "intel_be_batchbuffer.h" + +/* + * Need to redefine the BATCH defines + */ + +#undef BEGIN_BATCH +#define BEGIN_BATCH(dwords, relocs) \ + (i915_batchbuffer_check(&intel->base.batch->base, dwords, relocs)) + +#undef OUT_BATCH +#define OUT_BATCH(d) \ + i915_batchbuffer_dword(&intel->base.batch->base, d) + +#undef OUT_RELOC +#define OUT_RELOC(buf,flags,mask,delta) do { \ + assert((delta) >= 0); \ + intel_be_offset_relocation(intel->base.batch, delta, buf, flags, mask); \ +} while (0) + +#endif diff --git a/src/gallium/winsys/drm/intel/egl/intel_context.c b/src/gallium/winsys/drm/intel/egl/intel_context.c new file mode 100644 index 0000000000..927addb834 --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/intel_context.c @@ -0,0 +1,242 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "i915simple/i915_screen.h" + +#include "intel_device.h" +#include "intel_context.h" +#include "intel_batchbuffer.h" + +#include "state_tracker/st_public.h" +#include "pipe/p_defines.h" +#include "pipe/p_context.h" +#include "intel_egl.h" +#include "utils.h" + +#ifdef DEBUG +int __intel_debug = 0; +#endif + + +#define need_GL_ARB_multisample +#define need_GL_ARB_point_parameters +#define need_GL_ARB_texture_compression +#define need_GL_ARB_vertex_buffer_object +#define need_GL_ARB_vertex_program +#define need_GL_ARB_window_pos +#define need_GL_EXT_blend_color +#define need_GL_EXT_blend_equation_separate +#define need_GL_EXT_blend_func_separate +#define need_GL_EXT_blend_minmax +#define need_GL_EXT_cull_vertex +#define need_GL_EXT_fog_coord +#define need_GL_EXT_framebuffer_object +#define need_GL_EXT_multi_draw_arrays +#define need_GL_EXT_secondary_color +#define need_GL_NV_vertex_program +#include "extension_helper.h" + + +/** + * Extension strings exported by the intel driver. + * + * \note + * It appears that ARB_texture_env_crossbar has "disappeared" compared to the + * old i830-specific driver. + */ +const struct dri_extension card_extensions[] = { + {"GL_ARB_multisample", GL_ARB_multisample_functions}, + {"GL_ARB_multitexture", NULL}, + {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, + {"GL_ARB_texture_border_clamp", NULL}, + {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, + {"GL_ARB_texture_cube_map", NULL}, + {"GL_ARB_texture_env_add", NULL}, + {"GL_ARB_texture_env_combine", NULL}, + {"GL_ARB_texture_env_dot3", NULL}, + {"GL_ARB_texture_mirrored_repeat", NULL}, + {"GL_ARB_texture_rectangle", NULL}, + {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, + {"GL_ARB_pixel_buffer_object", NULL}, + {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, + {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, + {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, + {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, + {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, + {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, + {"GL_EXT_blend_subtract", NULL}, + {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, + {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, + {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, + {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, + {"GL_EXT_packed_depth_stencil", NULL}, + {"GL_EXT_pixel_buffer_object", NULL}, + {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, + {"GL_EXT_stencil_wrap", NULL}, + {"GL_EXT_texture_edge_clamp", NULL}, + {"GL_EXT_texture_env_combine", NULL}, + {"GL_EXT_texture_env_dot3", NULL}, + {"GL_EXT_texture_filter_anisotropic", NULL}, + {"GL_EXT_texture_lod_bias", NULL}, + {"GL_3DFX_texture_compression_FXT1", NULL}, + {"GL_APPLE_client_storage", NULL}, + {"GL_MESA_pack_invert", NULL}, + {"GL_MESA_ycbcr_texture", NULL}, + {"GL_NV_blend_square", NULL}, + {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, + {"GL_NV_vertex_program1_1", NULL}, + {"GL_SGIS_generate_mipmap", NULL }, + {NULL, NULL} +}; + + +/* + * Hardware lock functions. + * Doesn't do anything in EGL + */ + +static void +intel_lock_hardware(struct intel_be_context *context) +{ + (void)context; +} + +static void +intel_unlock_hardware(struct intel_be_context *context) +{ + (void)context; +} + +static boolean +intel_locked_hardware(struct intel_be_context *context) +{ + (void)context; + return FALSE; +} + + +/* + * Misc functions. + */ + +int +intel_create_context(struct egl_drm_context *egl_context, const __GLcontextModes *visual, void *sharedContextPrivate) +{ + struct intel_context *intel = CALLOC_STRUCT(intel_context); + struct intel_device *device = (struct intel_device *)egl_context->device->priv; + struct pipe_context *pipe; + struct st_context *st_share = NULL; + + egl_context->priv = intel; + + intel->intel_device = device; + intel->egl_context = egl_context; + intel->egl_device = egl_context->device; + + intel->base.hardware_lock = intel_lock_hardware; + intel->base.hardware_unlock = intel_unlock_hardware; + intel->base.hardware_locked = intel_locked_hardware; + + intel_be_init_context(&intel->base, &device->base); + +#if 0 + pipe = intel_create_softpipe(intel, screen->winsys); +#else + pipe = i915_create_context(device->pipe, &device->base.base, &intel->base.base); +#endif + + pipe->priv = intel; + + intel->st = st_create_context(pipe, visual, st_share); + + device->dummy = intel; + + return TRUE; +} + +int +intel_destroy_context(struct egl_drm_context *egl_context) +{ + struct intel_context *intel = egl_context->priv; + + if (intel->intel_device->dummy == intel) + intel->intel_device->dummy = NULL; + + st_destroy_context(intel->st); + intel_be_destroy_context(&intel->base); + free(intel); + return TRUE; +} + +void +intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *draw, struct egl_drm_drawable *read) +{ + if (context) { + struct intel_context *intel = (struct intel_context *)context->priv; + struct intel_framebuffer *draw_fb = (struct intel_framebuffer *)draw->priv; + struct intel_framebuffer *read_fb = (struct intel_framebuffer *)read->priv; + + assert(draw_fb->stfb); + assert(read_fb->stfb); + + st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); + + intel->egl_drawable = draw; + + st_resize_framebuffer(draw_fb->stfb, draw->w, draw->h); + + if (draw != read) + st_resize_framebuffer(read_fb->stfb, read->w, read->h); + + } else { + st_make_current(NULL, NULL, NULL); + } +} + +void +intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer *front) +{ + struct intel_device *device = (struct intel_device *)draw->device->priv; + struct intel_framebuffer *draw_fb = (struct intel_framebuffer *)draw->priv; + + if (draw_fb->front_buffer) + driBOUnReference(draw_fb->front_buffer); + + draw_fb->front_buffer = NULL; + draw_fb->front = NULL; + + /* to unbind just call this function with front == NULL */ + if (!front) + return; + + draw_fb->front = front; + + driGenBuffers(device->base.staticPool, "front", 1, &draw_fb->front_buffer, 0, 0, 0); + driBOSetReferenced(draw_fb->front_buffer, front->handle); + + st_resize_framebuffer(draw_fb->stfb, draw->w, draw->h); +} diff --git a/src/gallium/winsys/drm/intel/egl/intel_context.h b/src/gallium/winsys/drm/intel/egl/intel_context.h new file mode 100644 index 0000000000..477fdec7f7 --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/intel_context.h @@ -0,0 +1,118 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef INTEL_CONTEXT_H +#define INTEL_CONTEXT_H + +#include "pipe/p_debug.h" +#include "intel_be_context.h" + + +struct st_context; +struct egl_drm_device; +struct egl_drm_context; +struct egl_drm_frontbuffer; + + +/** + * Intel rendering context, contains a state tracker and intel-specific info. + */ +struct intel_context +{ + struct intel_be_context base; + + struct st_context *st; + + struct intel_device *intel_device; + + /* new egl stuff */ + struct egl_drm_device *egl_device; + struct egl_drm_context *egl_context; + struct egl_drm_drawable *egl_drawable; +}; + + + +/** + * Intel framebuffer. + */ +struct intel_framebuffer +{ + struct st_framebuffer *stfb; + + struct intel_device *device; + struct _DriBufferObject *front_buffer; + struct egl_drm_frontbuffer *front; +}; + + + + +/* These are functions now: + */ +void LOCK_HARDWARE( struct intel_context *intel ); +void UNLOCK_HARDWARE( struct intel_context *intel ); + +extern char *__progname; + + + +/* ================================================================ + * Debugging: + */ +#ifdef DEBUG +extern int __intel_debug; + +#define DEBUG_SWAP 0x1 +#define DEBUG_LOCK 0x2 +#define DEBUG_IOCTL 0x4 +#define DEBUG_BATCH 0x8 + +#define DBG(flag, ...) do { \ + if (__intel_debug & (DEBUG_##flag)) \ + printf(__VA_ARGS__); \ +} while(0) + +#else +#define DBG(flag, ...) +#endif + + +#define PCI_CHIP_845_G 0x2562 +#define PCI_CHIP_I830_M 0x3577 +#define PCI_CHIP_I855_GM 0x3582 +#define PCI_CHIP_I865_G 0x2572 +#define PCI_CHIP_I915_G 0x2582 +#define PCI_CHIP_I915_GM 0x2592 +#define PCI_CHIP_I945_G 0x2772 +#define PCI_CHIP_I945_GM 0x27A2 +#define PCI_CHIP_I945_GME 0x27AE +#define PCI_CHIP_G33_G 0x29C2 +#define PCI_CHIP_Q35_G 0x29B2 +#define PCI_CHIP_Q33_G 0x29D2 + +#endif diff --git a/src/gallium/winsys/drm/intel/egl/intel_device.c b/src/gallium/winsys/drm/intel/egl/intel_device.c new file mode 100644 index 0000000000..b9649cbec7 --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/intel_device.c @@ -0,0 +1,137 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#include "utils.h" + +#include "state_tracker/st_public.h" +#include "i915simple/i915_screen.h" + +#include "intel_context.h" +#include "intel_device.h" +#include "intel_batchbuffer.h" +#include "intel_egl.h" + + +extern const struct dri_extension card_extensions[]; + + +int +intel_create_device(struct egl_drm_device *device) +{ + struct intel_device *intel_device; + + /* Allocate the private area */ + intel_device = CALLOC_STRUCT(intel_device); + if (!intel_device) + return FALSE; + + device->priv = (void *)intel_device; + intel_device->device = device; + + intel_device->deviceID = device->deviceID; + + intel_be_init_device(&intel_device->base, device->drmFD, intel_device->deviceID); + + intel_device->pipe = i915_create_screen(&intel_device->base.base, intel_device->deviceID); + + /* hack */ + driInitExtensions(NULL, card_extensions, GL_FALSE); + + return TRUE; +} + +int +intel_destroy_device(struct egl_drm_device *device) +{ + struct intel_device *intel_device = (struct intel_device *)device->priv; + + intel_be_destroy_device(&intel_device->base); + + free(intel_device); + device->priv = NULL; + + return TRUE; +} + +int +intel_create_drawable(struct egl_drm_drawable *drawable, + const __GLcontextModes * visual) +{ + enum pipe_format colorFormat, depthFormat, stencilFormat; + struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); + + if (!intelfb) + return GL_FALSE; + + intelfb->device = drawable->device->priv; + + if (visual->redBits == 5) + colorFormat = PIPE_FORMAT_R5G6B5_UNORM; + else + colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; + + if (visual->depthBits == 16) + depthFormat = PIPE_FORMAT_Z16_UNORM; + else if (visual->depthBits == 24) + depthFormat = PIPE_FORMAT_S8Z24_UNORM; + else + depthFormat = PIPE_FORMAT_NONE; + + if (visual->stencilBits == 8) + stencilFormat = PIPE_FORMAT_S8Z24_UNORM; + else + stencilFormat = PIPE_FORMAT_NONE; + + intelfb->stfb = st_create_framebuffer(visual, + colorFormat, + depthFormat, + stencilFormat, + drawable->w, + drawable->h, + (void*) intelfb); + + if (!intelfb->stfb) { + free(intelfb); + return GL_FALSE; + } + + drawable->priv = (void *) intelfb; + return GL_TRUE; +} + +int +intel_destroy_drawable(struct egl_drm_drawable *drawable) +{ + struct intel_framebuffer *intelfb = (struct intel_framebuffer *)drawable->priv; + drawable->priv = NULL; + + assert(intelfb->stfb); + st_unreference_framebuffer(&intelfb->stfb); + free(intelfb); + return TRUE; +} diff --git a/src/gallium/winsys/drm/intel/egl/intel_device.h b/src/gallium/winsys/drm/intel/egl/intel_device.h new file mode 100644 index 0000000000..323a7c2aef --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/intel_device.h @@ -0,0 +1,50 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef _INTEL_SCREEN_H_ +#define _INTEL_SCREEN_H_ + +#include "intel_be_device.h" + +#include "pipe/p_compiler.h" + +struct pipe_screen; +struct egl_drm_device; +struct intel_context; + +struct intel_device +{ + struct intel_be_device base; + struct pipe_screen *pipe; + + int deviceID; + struct egl_drm_device *device; + + struct intel_context *dummy; +}; + +#endif diff --git a/src/gallium/winsys/drm/intel/egl/intel_egl.c b/src/gallium/winsys/drm/intel/egl/intel_egl.c new file mode 100644 index 0000000000..b89c5c508a --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/intel_egl.c @@ -0,0 +1,796 @@ + +#include +#include +#include +#include +#include + +#include "eglconfig.h" +#include "eglcontext.h" +#include "egldisplay.h" +#include "egldriver.h" +#include "eglglobals.h" +#include "eglmode.h" +#include "eglscreen.h" +#include "eglsurface.h" +#include "egllog.h" + +#include "intel_egl.h" + +#include "xf86drm.h" +#include "xf86drmMode.h" + +#include "intel_context.h" + +#include "state_tracker/st_public.h" + +#define MAX_SCREENS 16 + +static void +drm_get_device_id(struct egl_drm_device *device) +{ + char path[512]; + FILE *file; + + /* TODO get the real minor */ + int minor = 0; + + snprintf(path, sizeof(path), "/sys/class/drm/card%d/device/device", minor); + file = fopen(path, "r"); + if (!file) { + _eglLog(_EGL_WARNING, "Could not retrive device ID\n"); + return; + } + + fgets(path, sizeof( path ), file); + sscanf(path, "%x", &device->deviceID); + fclose(file); +} + +static struct egl_drm_device* +egl_drm_create_device(int drmFD) +{ + struct egl_drm_device *device = malloc(sizeof(*device)); + memset(device, 0, sizeof(*device)); + device->drmFD = drmFD; + + device->version = drmGetVersion(device->drmFD); + + drm_get_device_id(device); + + if (!intel_create_device(device)) { + free(device); + return NULL; + } + + return device; +} + +static void +_egl_context_modes_destroy(__GLcontextModes *modes) +{ + _eglLog(_EGL_DEBUG, "%s", __FUNCTION__); + + while (modes) { + __GLcontextModes * const next = modes->next; + free(modes); + modes = next; + } +} +/** + * Create a linked list of 'count' GLcontextModes. + * These are used during the client/server visual negotiation phase, + * then discarded. + */ +static __GLcontextModes * +_egl_context_modes_create(unsigned count, size_t minimum_size) +{ + /* This code copied from libGLX, and modified */ + const size_t size = (minimum_size > sizeof(__GLcontextModes)) + ? minimum_size : sizeof(__GLcontextModes); + __GLcontextModes * head = NULL; + __GLcontextModes ** next; + unsigned i; + + _eglLog(_EGL_DEBUG, "%s %d %d", __FUNCTION__, count, minimum_size); + + next = & head; + for (i = 0 ; i < count ; i++) { + *next = (__GLcontextModes *) calloc(1, size); + if (*next == NULL) { + _egl_context_modes_destroy(head); + head = NULL; + break; + } + + (*next)->doubleBufferMode = 1; + (*next)->visualID = GLX_DONT_CARE; + (*next)->visualType = GLX_DONT_CARE; + (*next)->visualRating = GLX_NONE; + (*next)->transparentPixel = GLX_NONE; + (*next)->transparentRed = GLX_DONT_CARE; + (*next)->transparentGreen = GLX_DONT_CARE; + (*next)->transparentBlue = GLX_DONT_CARE; + (*next)->transparentAlpha = GLX_DONT_CARE; + (*next)->transparentIndex = GLX_DONT_CARE; + (*next)->xRenderable = GLX_DONT_CARE; + (*next)->fbconfigID = GLX_DONT_CARE; + (*next)->swapMethod = GLX_SWAP_UNDEFINED_OML; + (*next)->bindToTextureRgb = GLX_DONT_CARE; + (*next)->bindToTextureRgba = GLX_DONT_CARE; + (*next)->bindToMipmapTexture = GLX_DONT_CARE; + (*next)->bindToTextureTargets = 0; + (*next)->yInverted = GLX_DONT_CARE; + + next = & ((*next)->next); + } + + return head; +} + +struct drm_screen; + +struct drm_driver +{ + _EGLDriver base; /* base class/object */ + + drmModeResPtr res; + + struct drm_screen *screens[MAX_SCREENS]; + size_t count_screens; + + struct egl_drm_device *device; +}; + +struct drm_surface +{ + _EGLSurface base; /* base class/object */ + + struct egl_drm_drawable *drawable; +}; + +struct drm_context +{ + _EGLContext base; /* base class/object */ + + struct egl_drm_context *context; +}; + +struct drm_screen +{ + _EGLScreen base; + + /* currently only support one connector */ + drmModeConnectorPtr connector; + + /* Has this screen been shown */ + int shown; + + /* Surface that is currently attached to this screen */ + struct drm_surface *surf; + + /* backing buffer */ + drmBO buffer; + + /* framebuffer */ + drmModeFBPtr fb; + uint32_t fbID; + + /* crtc and mode used */ + drmModeCrtcPtr crtc; + uint32_t crtcID; + + struct drm_mode_modeinfo *mode; + + /* geometry of the screen */ + struct egl_drm_frontbuffer front; +}; + +static void +drm_update_res(struct drm_driver *drm_drv) +{ + drmModeFreeResources(drm_drv->res); + drm_drv->res = drmModeGetResources(drm_drv->device->drmFD); +} + +static void +drm_add_modes_from_connector(_EGLScreen *screen, drmModeConnectorPtr connector) +{ + struct drm_mode_modeinfo *m; + int i; + + for (i = 0; i < connector->count_modes; i++) { + m = &connector->modes[i]; + _eglAddNewMode(screen, m->hdisplay, m->vdisplay, m->vrefresh, m->name); + } +} + + +static EGLBoolean +drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) +{ + _EGLDisplay *disp = _eglLookupDisplay(dpy); + struct drm_driver *drm_drv = (struct drm_driver *)drv; + struct drm_screen *screen = NULL; + drmModeConnectorPtr connector = NULL; + drmModeResPtr res = NULL; + unsigned count_connectors = 0; + int num_screens = 0; + + EGLint i; + int fd; + + fd = drmOpen("i915", NULL); + if (fd < 0) { + return EGL_FALSE; + } + + drm_drv->device = egl_drm_create_device(fd); + if (!drm_drv->device) { + drmClose(fd); + return EGL_FALSE; + } + + drm_update_res(drm_drv); + res = drm_drv->res; + if (res) + count_connectors = res->count_connectors; + + for(i = 0; i < count_connectors && i < MAX_SCREENS; i++) { + connector = drmModeGetConnector(fd, res->connectors[i]); + + if (!connector) + continue; + + if (connector->connection != DRM_MODE_CONNECTED) { + drmModeFreeConnector(connector); + continue; + } + + screen = malloc(sizeof(struct drm_screen)); + memset(screen, 0, sizeof(*screen)); + screen->connector = connector; + _eglInitScreen(&screen->base); + _eglAddScreen(disp, &screen->base); + drm_add_modes_from_connector(&screen->base, connector); + drm_drv->screens[num_screens++] = screen; + } + drm_drv->count_screens = num_screens; + + /* for now we only have one config */ + _EGLConfig *config = calloc(1, sizeof(*config)); + memset(config, 1, sizeof(*config)); + _eglInitConfig(config, 1); + _eglSetConfigAttrib(config, EGL_RED_SIZE, 8); + _eglSetConfigAttrib(config, EGL_GREEN_SIZE, 8); + _eglSetConfigAttrib(config, EGL_BLUE_SIZE, 8); + _eglSetConfigAttrib(config, EGL_ALPHA_SIZE, 8); + _eglSetConfigAttrib(config, EGL_BUFFER_SIZE, 32); + _eglSetConfigAttrib(config, EGL_DEPTH_SIZE, 24); + _eglSetConfigAttrib(config, EGL_STENCIL_SIZE, 8); + _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_PBUFFER_BIT); + _eglAddConfig(disp, config); + + drv->Initialized = EGL_TRUE; + + *major = 1; + *minor = 4; + + return EGL_TRUE; +} + +static void +drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen) +{ + struct drm_driver *drm_drv = (struct drm_driver *)drv; + unsigned int i; + + intel_bind_frontbuffer(screen->surf->drawable, NULL); + screen->surf = NULL; + + for (i = 0; i < drm_drv->res->count_crtcs; i++) { + drmModeSetCrtc( + drm_drv->device->drmFD, + drm_drv->res->crtcs[i], + 0, // FD + 0, 0, + NULL, 0, // List of output ids + NULL); + } + + drmModeRmFB(drm_drv->device->drmFD, screen->fbID); + drmModeFreeFB(screen->fb); + screen->fb = NULL; + + drmBOUnreference(drm_drv->device->drmFD, &screen->buffer); + + screen->shown = 0; +} + +static EGLBoolean +drm_terminate(_EGLDriver *drv, EGLDisplay dpy) +{ + struct drm_driver *drm_drv = (struct drm_driver *)drv; + struct drm_screen *screen; + int i = 0; + + intel_destroy_device(drm_drv->device); + drmFreeVersion(drm_drv->device->version); + + for (i = 0; i < drm_drv->count_screens; i++) { + screen = drm_drv->screens[i]; + + if (screen->shown) + drm_takedown_shown_screen(drv, screen); + + drmModeFreeConnector(screen->connector); + _eglDestroyScreen(&screen->base); + drm_drv->screens[i] = NULL; + } + + drmClose(drm_drv->device->drmFD); + + free(drm_drv->device); + + _eglCleanupDisplay(_eglLookupDisplay(dpy)); + free(drm_drv); + + return EGL_TRUE; +} + + +static struct drm_context * +lookup_drm_context(EGLContext context) +{ + _EGLContext *c = _eglLookupContext(context); + return (struct drm_context *) c; +} + + +static struct drm_surface * +lookup_drm_surface(EGLSurface surface) +{ + _EGLSurface *s = _eglLookupSurface(surface); + return (struct drm_surface *) s; +} + +static struct drm_screen * +lookup_drm_screen(EGLDisplay dpy, EGLScreenMESA screen) +{ + _EGLScreen *s = _eglLookupScreen(dpy, screen); + return (struct drm_screen *) s; +} + +static __GLcontextModes* +visual_from_config(_EGLConfig *conf) +{ + __GLcontextModes *visual; + (void)conf; + + visual = _egl_context_modes_create(1, sizeof(*visual)); + visual->redBits = 8; + visual->greenBits = 8; + visual->blueBits = 8; + visual->alphaBits = 8; + + visual->rgbBits = 32; + visual->doubleBufferMode = 1; + + visual->depthBits = 24; + visual->haveDepthBuffer = visual->depthBits > 0; + visual->stencilBits = 8; + visual->haveStencilBuffer = visual->stencilBits > 0; + + return visual; +} + + + +static EGLContext +drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list) +{ + struct drm_driver *drm_drv = (struct drm_driver *)drv; + struct drm_context *c; + struct drm_egl_context *share = NULL; + _EGLConfig *conf; + int i; + int ret; + __GLcontextModes *visual; + struct egl_drm_context *context; + + conf = _eglLookupConfig(drv, dpy, config); + if (!conf) { + _eglError(EGL_BAD_CONFIG, "eglCreateContext"); + return EGL_NO_CONTEXT; + } + + for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { + switch (attrib_list[i]) { + /* no attribs defined for now */ + default: + _eglError(EGL_BAD_ATTRIBUTE, "eglCreateContext"); + return EGL_NO_CONTEXT; + } + } + + c = (struct drm_context *) calloc(1, sizeof(struct drm_context)); + if (!c) + return EGL_NO_CONTEXT; + + _eglInitContext(drv, dpy, &c->base, config, attrib_list); + + context = malloc(sizeof(*context)); + memset(context, 0, sizeof(*context)); + + if (!context) + goto err_c; + + context->device = drm_drv->device; + visual = visual_from_config(conf); + + ret = intel_create_context(context, visual, share); + free(visual); + + if (!ret) + goto err_gl; + + c->context = context; + + /* generate handle and insert into hash table */ + _eglSaveContext(&c->base); + assert(_eglGetContextHandle(&c->base)); + + return _eglGetContextHandle(&c->base); +err_gl: + free(context); +err_c: + free(c); + return EGL_NO_CONTEXT; +} + +static EGLBoolean +drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context) +{ + struct drm_context *fc = lookup_drm_context(context); + _eglRemoveContext(&fc->base); + if (fc->base.IsBound) { + fc->base.DeletePending = EGL_TRUE; + } else { + intel_destroy_context(fc->context); + free(fc->context); + free(fc); + } + return EGL_TRUE; +} + + +static EGLSurface +drm_create_window_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list) +{ + return EGL_NO_SURFACE; +} + + +static EGLSurface +drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list) +{ + return EGL_NO_SURFACE; +} + + +static EGLSurface +drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, + const EGLint *attrib_list) +{ + struct drm_driver *drm_drv = (struct drm_driver *)drv; + int i; + int ret; + int width = -1; + int height = -1; + struct drm_surface *surf = NULL; + struct egl_drm_drawable *drawable = NULL; + __GLcontextModes *visual; + _EGLConfig *conf; + + conf = _eglLookupConfig(drv, dpy, config); + if (!conf) { + _eglError(EGL_BAD_CONFIG, "eglCreatePbufferSurface"); + return EGL_NO_CONTEXT; + } + + for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { + switch (attrib_list[i]) { + case EGL_WIDTH: + width = attrib_list[++i]; + break; + case EGL_HEIGHT: + height = attrib_list[++i]; + break; + default: + _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface"); + return EGL_NO_SURFACE; + } + } + + if (width < 1 || height < 1) { + _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface"); + return EGL_NO_SURFACE; + } + + surf = (struct drm_surface *) calloc(1, sizeof(struct drm_surface)); + if (!surf) + goto err; + + if (!_eglInitSurface(drv, dpy, &surf->base, EGL_PBUFFER_BIT, config, attrib_list)) + goto err_surf; + + drawable = malloc(sizeof(*drawable)); + memset(drawable, 0, sizeof(*drawable)); + + drawable->w = width; + drawable->h = height; + + visual = visual_from_config(conf); + + drawable->device = drm_drv->device; + ret = intel_create_drawable(drawable, visual); + free(visual); + + if (!ret) + goto err_draw; + + surf->drawable = drawable; + + _eglSaveSurface(&surf->base); + return surf->base.Handle; + +err_draw: + free(drawable); +err_surf: + free(surf); +err: + return EGL_NO_SURFACE; +} + +static EGLSurface +drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, + const EGLint *attrib_list) +{ + EGLSurface surf = drm_create_pbuffer_surface(drv, dpy, cfg, attrib_list); + + return surf; +} + +static struct drm_mode_modeinfo * +drm_find_mode(drmModeConnectorPtr connector, _EGLMode *mode) +{ + int i; + struct drm_mode_modeinfo *m; + + for (i = 0; i < connector->count_modes; i++) { + m = &connector->modes[i]; + if (m->hdisplay == mode->Width && m->vdisplay == mode->Height && m->vrefresh == mode->RefreshRate) + break; + m = NULL; + } + + return m; +} +static void +draw(size_t x, size_t y, size_t w, size_t h, size_t pitch, size_t v, unsigned int *ptr) +{ + int i, j; + + for (i = x; i < x + w; i++) + for(j = y; j < y + h; j++) + ptr[(i * pitch / 4) + j] = v; + +} + +static void +prettyColors(int fd, unsigned int handle, size_t pitch) +{ + drmBO bo; + unsigned int *ptr; + void *p; + int i; + + drmBOReference(fd, handle, &bo); + drmBOMap(fd, &bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, &p); + ptr = (unsigned int*)p; + + for (i = 0; i < (bo.size / 4); i++) + ptr[i] = 0xFFFFFFFF; + + for (i = 0; i < 4; i++) + draw(i * 40, i * 40, 40, 40, pitch, 0, ptr); + + + draw(200, 100, 40, 40, pitch, 0xff00ff, ptr); + draw(100, 200, 40, 40, pitch, 0xff00ff, ptr); + + drmBOUnmap(fd, &bo); +} + +static EGLBoolean +drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, + EGLScreenMESA screen, + EGLSurface surface, EGLModeMESA m) +{ + struct drm_driver *drm_drv = (struct drm_driver *)drv; + struct drm_surface *surf = lookup_drm_surface(surface); + struct drm_screen *scrn = lookup_drm_screen(dpy, screen); + _EGLMode *mode = _eglLookupMode(dpy, m); + size_t pitch = mode->Width * 4; + size_t size = mode->Height * pitch; + int ret; + unsigned int i,j,k; + + if (scrn->shown) + drm_takedown_shown_screen(drv, scrn); + + ret = drmBOCreate(drm_drv->device->drmFD, size, 0, 0, + DRM_BO_FLAG_READ | + DRM_BO_FLAG_WRITE | + DRM_BO_FLAG_MEM_TT | + DRM_BO_FLAG_MEM_VRAM | + DRM_BO_FLAG_NO_EVICT, + DRM_BO_HINT_DONT_FENCE, &scrn->buffer); + + if (ret) + return EGL_FALSE; + + prettyColors(drm_drv->device->drmFD, scrn->buffer.handle, pitch); + + ret = drmModeAddFB(drm_drv->device->drmFD, mode->Width, mode->Height, + 32, 32, pitch, + scrn->buffer.handle, + &scrn->fbID); + + if (ret) + goto err_bo; + + scrn->fb = drmModeGetFB(drm_drv->device->drmFD, scrn->fbID); + if (!scrn->fb) + goto err_bo; + + scrn->mode = drm_find_mode(scrn->connector, mode); + if (!scrn->mode) + goto err_fb; + + for (j = 0; j < drm_drv->res->count_connectors; j++) { + drmModeConnector *con = drmModeGetConnector(drm_drv->device->drmFD, drm_drv->res->connectors[j]); + for (k = 0; k < con->count_encoders; k++) { + drmModeEncoder *enc = drmModeGetEncoder(drm_drv->device->drmFD, con->encoders[k]); + for (i = 0; i < drm_drv->res->count_crtcs; i++) { + if (enc->possible_crtcs & (1<device->drmFD, + drm_drv->res->crtcs[i], + scrn->fbID, + 0, 0, + &drm_drv->res->connectors[j], 1, + scrn->mode); + /* skip the other crtcs now */ + i = drm_drv->res->count_crtcs; + } + } + } + } + + scrn->front.handle = scrn->buffer.handle; + scrn->front.pitch = pitch; + scrn->front.width = mode->Width; + scrn->front.height = mode->Height; + + scrn->surf = surf; + intel_bind_frontbuffer(surf->drawable, &scrn->front); + + scrn->shown = 1; + + return EGL_TRUE; + +err_fb: + /* TODO remove fb */ + +err_bo: + drmBOUnreference(drm_drv->device->drmFD, &scrn->buffer); + return EGL_FALSE; +} + +static EGLBoolean +drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) +{ + struct drm_surface *fs = lookup_drm_surface(surface); + _eglRemoveSurface(&fs->base); + if (fs->base.IsBound) { + fs->base.DeletePending = EGL_TRUE; + } else { + intel_bind_frontbuffer(fs->drawable, NULL); + intel_destroy_drawable(fs->drawable); + free(fs->drawable); + free(fs); + } + return EGL_TRUE; +} + + +static EGLBoolean +drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context) +{ + struct drm_surface *readSurf = lookup_drm_surface(read); + struct drm_surface *drawSurf = lookup_drm_surface(draw); + struct drm_context *ctx = lookup_drm_context(context); + EGLBoolean b; + + b = _eglMakeCurrent(drv, dpy, draw, read, context); + if (!b) + return EGL_FALSE; + + if (ctx) { + if (!drawSurf || !readSurf) + return EGL_FALSE; + + intel_make_current(ctx->context, drawSurf->drawable, readSurf->drawable); + } else { + intel_make_current(NULL, NULL, NULL); + } + + return EGL_TRUE; +} + +static EGLBoolean +drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) +{ + struct drm_surface *surf = lookup_drm_surface(draw); + if (!surf) + return EGL_FALSE; + + /* error checking */ + if (!_eglSwapBuffers(drv, dpy, draw)) + return EGL_FALSE; + + intel_swap_buffers(surf->drawable); + return EGL_TRUE; +} + + +/** + * The bootstrap function. Return a new drm_driver object and + * plug in API functions. + */ +_EGLDriver * +_eglMain(_EGLDisplay *dpy, const char *args) +{ + struct drm_driver *drm; + + drm = (struct drm_driver *) calloc(1, sizeof(struct drm_driver)); + if (!drm) { + return NULL; + } + + /* First fill in the dispatch table with defaults */ + _eglInitDriverFallbacks(&drm->base); + /* then plug in our Drm-specific functions */ + drm->base.API.Initialize = drm_initialize; + drm->base.API.Terminate = drm_terminate; + drm->base.API.CreateContext = drm_create_context; + drm->base.API.MakeCurrent = drm_make_current; + drm->base.API.CreateWindowSurface = drm_create_window_surface; + drm->base.API.CreatePixmapSurface = drm_create_pixmap_surface; + drm->base.API.CreatePbufferSurface = drm_create_pbuffer_surface; + drm->base.API.DestroySurface = drm_destroy_surface; + drm->base.API.DestroyContext = drm_destroy_context; + drm->base.API.CreateScreenSurfaceMESA = drm_create_screen_surface_mesa; + drm->base.API.ShowScreenSurfaceMESA = drm_show_screen_surface_mesa; + drm->base.API.SwapBuffers = drm_swap_buffers; + + drm->base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; + drm->base.Name = "DRM/Gallium"; + + /* enable supported extensions */ + drm->base.Extensions.MESA_screen_surface = EGL_TRUE; + drm->base.Extensions.MESA_copy_context = EGL_TRUE; + + return &drm->base; +} diff --git a/src/gallium/winsys/drm/intel/egl/intel_egl.h b/src/gallium/winsys/drm/intel/egl/intel_egl.h new file mode 100644 index 0000000000..1ee27e0847 --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/intel_egl.h @@ -0,0 +1,53 @@ + +#ifndef _INTEL_EGL_H_ +#define _INTEL_EGL_H_ + +#include + +struct egl_drm_device +{ + void *priv; + int drmFD; + + drmVersionPtr version; + int deviceID; +}; + +struct egl_drm_context +{ + void *priv; + struct egl_drm_device *device; +}; + +struct egl_drm_drawable +{ + void *priv; + struct egl_drm_device *device; + size_t h; + size_t w; +}; + +struct egl_drm_frontbuffer +{ + uint32_t handle; + uint32_t pitch; + uint32_t width; + uint32_t height; +}; + +#include "GL/internal/glcore.h" + +int intel_create_device(struct egl_drm_device *device); +int intel_destroy_device(struct egl_drm_device *device); + +int intel_create_context(struct egl_drm_context *context, const __GLcontextModes *visual, void *sharedContextPrivate); +int intel_destroy_context(struct egl_drm_context *context); + +int intel_create_drawable(struct egl_drm_drawable *drawable, const __GLcontextModes * visual); +int intel_destroy_drawable(struct egl_drm_drawable *drawable); + +void intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *draw, struct egl_drm_drawable *read); +void intel_swap_buffers(struct egl_drm_drawable *draw); +void intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer *front); + +#endif diff --git a/src/gallium/winsys/drm/intel/egl/intel_reg.h b/src/gallium/winsys/drm/intel/egl/intel_reg.h new file mode 100644 index 0000000000..4f33bee438 --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/intel_reg.h @@ -0,0 +1,53 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef _INTEL_REG_H_ +#define _INTEL_REG_H_ + + +#define BR00_BITBLT_CLIENT 0x40000000 +#define BR00_OP_COLOR_BLT 0x10000000 +#define BR00_OP_SRC_COPY_BLT 0x10C00000 +#define BR13_SOLID_PATTERN 0x80000000 + +#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) +#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) +#define XY_COLOR_BLT_WRITE_RGB (1<<20) + +#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) +#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) +#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) + +#define MI_WAIT_FOR_EVENT ((0x3<<23)) +#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) +#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) + +#define MI_BATCH_BUFFER_END (0xA<<23) + + +#endif diff --git a/src/gallium/winsys/drm/intel/egl/intel_swapbuffers.c b/src/gallium/winsys/drm/intel/egl/intel_swapbuffers.c new file mode 100644 index 0000000000..2edcbc79ff --- /dev/null +++ b/src/gallium/winsys/drm/intel/egl/intel_swapbuffers.c @@ -0,0 +1,111 @@ +/************************************************************************** + * + * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#include "intel_device.h" +#include "intel_context.h" +#include "intel_batchbuffer.h" +#include "intel_reg.h" + +#include "pipe/p_context.h" +#include "state_tracker/st_public.h" +#include "state_tracker/st_context.h" +#include "state_tracker/st_cb_fbo.h" +#include "intel_egl.h" + + +static void +intel_display_surface(struct egl_drm_drawable *draw, + struct pipe_surface *surf); + +void intel_swap_buffers(struct egl_drm_drawable *draw) +{ + struct intel_framebuffer *intel_fb = (struct intel_framebuffer *)draw->priv; + struct pipe_surface *back_surf; + + assert(intel_fb); + assert(intel_fb->stfb); + + back_surf = st_get_framebuffer_surface(intel_fb->stfb, ST_SURFACE_BACK_LEFT); + if (back_surf) { + st_notify_swapbuffers(intel_fb->stfb); + if (intel_fb->front) + intel_display_surface(draw, back_surf); + st_notify_swapbuffers_complete(intel_fb->stfb); + } +} + +static void +intel_display_surface(struct egl_drm_drawable *draw, + struct pipe_surface *surf) +{ + struct intel_context *intel = NULL; + struct intel_framebuffer *intel_fb = (struct intel_framebuffer *)draw->priv; + struct _DriFenceObject *fence; + + //const int srcWidth = surf->width; + //const int srcHeight = surf->height; + + intel = intel_fb->device->dummy; + if (!intel) { + printf("No dummy context\n"); + return; + } + + const int dstWidth = intel_fb->front->width; + const int dstHeight = intel_fb->front->height; + const int dstPitch = intel_fb->front->pitch / 4;//draw->front.cpp; + + const int cpp = 4;//intel_fb->front->cpp; + const int srcPitch = surf->stride / cpp; + + int BR13, CMD; + //int i; + + BR13 = (dstPitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); + CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | + XY_SRC_COPY_BLT_WRITE_RGB); + + BEGIN_BATCH(8, 2); + OUT_BATCH(CMD); + OUT_BATCH(BR13); + OUT_BATCH((0 << 16) | 0); + OUT_BATCH((dstHeight << 16) | dstWidth); + + OUT_RELOC(intel_fb->front_buffer, + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, + DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); + + OUT_BATCH((0 << 16) | 0); + OUT_BATCH((srcPitch * cpp) & 0xffff); + OUT_RELOC(dri_bo(surf->buffer), + DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, + DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); + + fence = intel_be_batchbuffer_flush(intel->base.batch); + driFenceUnReference(&fence); + intel_be_batchbuffer_finish(intel->base.batch); +} diff --git a/src/gallium/winsys/egl_drm/Makefile b/src/gallium/winsys/egl_drm/Makefile deleted file mode 100644 index 4139d9e71f..0000000000 --- a/src/gallium/winsys/egl_drm/Makefile +++ /dev/null @@ -1,38 +0,0 @@ -# src/mesa/drivers/egl_drm/Makefile - -TOP = ../../../.. - -include $(TOP)/configs/current - - - -default: $(TOP)/$(LIB_DIR) subdirs - - -$(TOP)/$(LIB_DIR): - -mkdir $(TOP)/$(LIB_DIR) - - -subdirs: - @for dir in $(DRI_DIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE)) || exit 1 ; \ - fi \ - done - - -install: - @for dir in $(DRI_DIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE) install) || exit 1 ; \ - fi \ - done - - -clean: - @for dir in $(DRI_DIRS) ; do \ - if [ -d $$dir ] ; then \ - (cd $$dir && $(MAKE) clean) ; \ - fi \ - done - -rm -f common/*.o diff --git a/src/gallium/winsys/egl_drm/Makefile.template b/src/gallium/winsys/egl_drm/Makefile.template deleted file mode 100644 index 07abfa53f3..0000000000 --- a/src/gallium/winsys/egl_drm/Makefile.template +++ /dev/null @@ -1,117 +0,0 @@ -# -*-makefile-*- - -MESA_MODULES = \ - $(TOP)/src/mesa/libmesa.a \ - $(GALLIUM_AUXILIARIES) - -COMMON_GALLIUM_SOURCES = \ - $(TOP)/src/mesa/drivers/dri/common/utils.c \ - $(TOP)/src/mesa/drivers/dri/common/vblank.c \ - $(TOP)/src/mesa/drivers/dri/common/dri_util.c \ - $(TOP)/src/mesa/drivers/dri/common/xmlconfig.c - -COMMON_SOURCES = $(COMMON_GALLIUM_SOURCES) \ - $(TOP)/src/mesa/drivers/common/driverfuncs.c \ - $(TOP)/src/mesa/drivers/dri/common/texmem.c \ - $(TOP)/src/mesa/drivers/dri/common/drirenderbuffer.c - -COMMON_BM_SOURCES = \ - $(TOP)/src/mesa/drivers/dri/common/dri_bufmgr.c \ - $(TOP)/src/mesa/drivers/dri/common/dri_drmpool.c - - -ifeq ($(WINDOW_SYSTEM),dri) -WINOBJ= -WINLIB= -INCLUDES = $(SHARED_INCLUDES) $(EXPAT_INCLUDES) - -OBJECTS = \ - $(C_SOURCES:.c=.o) \ - $(ASM_SOURCES:.S=.o) - -else -# miniglx -WINOBJ= -WINLIB=-L$(MESA)/src/glx/mini -MINIGLX_INCLUDES = -I$(TOP)/src/glx/mini -INCLUDES = $(MINIGLX_INCLUDES) \ - $(SHARED_INCLUDES) \ - $(PCIACCESS_CFLAGS) - -OBJECTS = $(C_SOURCES:.c=.o) \ - $(MINIGLX_SOURCES:.c=.o) \ - $(ASM_SOURCES:.S=.o) -endif - - -### Include directories -SHARED_INCLUDES = \ - -I. \ - -I$(TOP)/src/mesa/drivers/dri/common \ - -Iserver \ - -I$(TOP)/include \ - -I$(TOP)/include/GL/internal \ - -I$(TOP)/src/gallium/include \ - -I$(TOP)/src/gallium/auxiliary \ - -I$(TOP)/src/gallium/drivers \ - -I$(TOP)/src/gallium/winsys/common \ - -I$(TOP)/src/mesa \ - -I$(TOP)/src/mesa/main \ - -I$(TOP)/src/mesa/glapi \ - -I$(TOP)/src/mesa/math \ - -I$(TOP)/src/mesa/transform \ - -I$(TOP)/src/mesa/shader \ - -I$(TOP)/src/mesa/swrast \ - -I$(TOP)/src/mesa/swrast_setup \ - -I$(TOP)/src/egl/main \ - -I$(TOP)/src/egl/drivers/dri \ - $(LIBDRM_CFLAGS) - - -##### RULES ##### - -.c.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - -.S.o: - $(CC) -c $(INCLUDES) $(CFLAGS) $(DRIVER_DEFINES) $< -o $@ - - -##### TARGETS ##### - -default: depend symlinks $(LIBNAME) $(TOP)/$(LIB_DIR)/$(LIBNAME) - - -$(LIBNAME): $(OBJECTS) $(MESA_MODULES) $(PIPE_DRIVERS) $(WINOBJ) Makefile $(TOP)/src/mesa/drivers/dri/Makefile.template - $(TOP)/bin/mklib -noprefix -o $@ \ - $(OBJECTS) $(PIPE_DRIVERS) $(MESA_MODULES) $(WINOBJ) $(DRI_LIB_DEPS) - - -$(TOP)/$(LIB_DIR)/$(LIBNAME): $(LIBNAME) - $(INSTALL) $(LIBNAME) $(TOP)/$(LIB_DIR) - - -depend: $(C_SOURCES) $(ASM_SOURCES) $(SYMLINKS) - rm -f depend - touch depend - $(MKDEP) $(MKDEP_OPTIONS) $(DRIVER_DEFINES) $(INCLUDES) $(C_SOURCES) \ - $(ASM_SOURCES) 2> /dev/null - - -# Emacs tags -tags: - etags `find . -name \*.[ch]` `find ../include` - - -# Remove .o and backup files -clean: - -rm -f *.o */*.o *~ *.so *~ server/*.o $(SYMLINKS) - -rm -f depend depend.bak - - -install: $(LIBNAME) - $(INSTALL) -d $(DRI_DRIVER_INSTALL_DIR) - $(INSTALL) -m 755 $(LIBNAME) $(DRI_DRIVER_INSTALL_DIR) - - -include depend diff --git a/src/gallium/winsys/egl_drm/intel/Makefile b/src/gallium/winsys/egl_drm/intel/Makefile deleted file mode 100644 index e67b49f3ad..0000000000 --- a/src/gallium/winsys/egl_drm/intel/Makefile +++ /dev/null @@ -1,29 +0,0 @@ - -TOP = ../../../../.. -include $(TOP)/configs/current - -LIBNAME = EGL_i915.so - -PIPE_DRIVERS = \ - $(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \ - $(TOP)/src/gallium/drivers/i915simple/libi915simple.a \ - $(TOP)/src/gallium/winsys/common/intel_drm/libinteldrm.a - -DRIVER_SOURCES = \ - intel_swapbuffers.c \ - intel_context.c \ - intel_device.c \ - intel_egl.c - -C_SOURCES = \ - $(COMMON_GALLIUM_SOURCES) \ - $(DRIVER_SOURCES) - -ASM_SOURCES = - -DRIVER_DEFINES = -I$(TOP)/src/mesa/drivers/dri/intel $(shell pkg-config libdrm --atleast-version=2.3.1 \ - && echo "-DDRM_VBLANK_FLIP=DRM_VBLANK_FLIP") - -include ../Makefile.template - -symlinks: diff --git a/src/gallium/winsys/egl_drm/intel/SConscript b/src/gallium/winsys/egl_drm/intel/SConscript deleted file mode 100644 index 0ad19d42a8..0000000000 --- a/src/gallium/winsys/egl_drm/intel/SConscript +++ /dev/null @@ -1,39 +0,0 @@ -Import('*') - -env = drienv.Clone() - -env.Append(CPPPATH = [ - '../intel', - 'server' -]) - -#MINIGLX_SOURCES = server/intel_dri.c - -DRIVER_SOURCES = [ - 'intel_winsys_pipe.c', - 'intel_winsys_softpipe.c', - 'intel_winsys_i915.c', - 'intel_batchbuffer.c', - 'intel_swapbuffers.c', - 'intel_context.c', - 'intel_lock.c', - 'intel_screen.c', - 'intel_batchpool.c', -] - -sources = \ - COMMON_GALLIUM_SOURCES + \ - COMMON_BM_SOURCES + \ - DRIVER_SOURCES - -drivers = [ - softpipe, - i915simple -] - -# TODO: write a wrapper function http://www.scons.org/wiki/WrapperFunctions -env.SharedLibrary( - target ='i915tex_dri.so', - source = sources, - LIBS = drivers + mesa + auxiliaries + env['LIBS'], -) \ No newline at end of file diff --git a/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h b/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h deleted file mode 100644 index 1fa2719845..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_batchbuffer.h +++ /dev/null @@ -1,24 +0,0 @@ -#ifndef INTEL_BATCHBUFFER_H -#define INTEL_BATCHBUFFER_H - -#include "intel_drm/intel_be_batchbuffer.h" - -/* - * Need to redefine the BATCH defines - */ - -#undef BEGIN_BATCH -#define BEGIN_BATCH(dwords, relocs) \ - (i915_batchbuffer_check(&intel->base.batch->base, dwords, relocs)) - -#undef OUT_BATCH -#define OUT_BATCH(d) \ - i915_batchbuffer_dword(&intel->base.batch->base, d) - -#undef OUT_RELOC -#define OUT_RELOC(buf,flags,mask,delta) do { \ - assert((delta) >= 0); \ - intel_be_offset_relocation(intel->base.batch, delta, buf, flags, mask); \ -} while (0) - -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.c b/src/gallium/winsys/egl_drm/intel/intel_context.c deleted file mode 100644 index 927addb834..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_context.c +++ /dev/null @@ -1,242 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "i915simple/i915_screen.h" - -#include "intel_device.h" -#include "intel_context.h" -#include "intel_batchbuffer.h" - -#include "state_tracker/st_public.h" -#include "pipe/p_defines.h" -#include "pipe/p_context.h" -#include "intel_egl.h" -#include "utils.h" - -#ifdef DEBUG -int __intel_debug = 0; -#endif - - -#define need_GL_ARB_multisample -#define need_GL_ARB_point_parameters -#define need_GL_ARB_texture_compression -#define need_GL_ARB_vertex_buffer_object -#define need_GL_ARB_vertex_program -#define need_GL_ARB_window_pos -#define need_GL_EXT_blend_color -#define need_GL_EXT_blend_equation_separate -#define need_GL_EXT_blend_func_separate -#define need_GL_EXT_blend_minmax -#define need_GL_EXT_cull_vertex -#define need_GL_EXT_fog_coord -#define need_GL_EXT_framebuffer_object -#define need_GL_EXT_multi_draw_arrays -#define need_GL_EXT_secondary_color -#define need_GL_NV_vertex_program -#include "extension_helper.h" - - -/** - * Extension strings exported by the intel driver. - * - * \note - * It appears that ARB_texture_env_crossbar has "disappeared" compared to the - * old i830-specific driver. - */ -const struct dri_extension card_extensions[] = { - {"GL_ARB_multisample", GL_ARB_multisample_functions}, - {"GL_ARB_multitexture", NULL}, - {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions}, - {"GL_ARB_texture_border_clamp", NULL}, - {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions}, - {"GL_ARB_texture_cube_map", NULL}, - {"GL_ARB_texture_env_add", NULL}, - {"GL_ARB_texture_env_combine", NULL}, - {"GL_ARB_texture_env_dot3", NULL}, - {"GL_ARB_texture_mirrored_repeat", NULL}, - {"GL_ARB_texture_rectangle", NULL}, - {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions}, - {"GL_ARB_pixel_buffer_object", NULL}, - {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions}, - {"GL_ARB_window_pos", GL_ARB_window_pos_functions}, - {"GL_EXT_blend_color", GL_EXT_blend_color_functions}, - {"GL_EXT_blend_equation_separate", GL_EXT_blend_equation_separate_functions}, - {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions}, - {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions}, - {"GL_EXT_blend_subtract", NULL}, - {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions}, - {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions}, - {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions}, - {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions}, - {"GL_EXT_packed_depth_stencil", NULL}, - {"GL_EXT_pixel_buffer_object", NULL}, - {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions}, - {"GL_EXT_stencil_wrap", NULL}, - {"GL_EXT_texture_edge_clamp", NULL}, - {"GL_EXT_texture_env_combine", NULL}, - {"GL_EXT_texture_env_dot3", NULL}, - {"GL_EXT_texture_filter_anisotropic", NULL}, - {"GL_EXT_texture_lod_bias", NULL}, - {"GL_3DFX_texture_compression_FXT1", NULL}, - {"GL_APPLE_client_storage", NULL}, - {"GL_MESA_pack_invert", NULL}, - {"GL_MESA_ycbcr_texture", NULL}, - {"GL_NV_blend_square", NULL}, - {"GL_NV_vertex_program", GL_NV_vertex_program_functions}, - {"GL_NV_vertex_program1_1", NULL}, - {"GL_SGIS_generate_mipmap", NULL }, - {NULL, NULL} -}; - - -/* - * Hardware lock functions. - * Doesn't do anything in EGL - */ - -static void -intel_lock_hardware(struct intel_be_context *context) -{ - (void)context; -} - -static void -intel_unlock_hardware(struct intel_be_context *context) -{ - (void)context; -} - -static boolean -intel_locked_hardware(struct intel_be_context *context) -{ - (void)context; - return FALSE; -} - - -/* - * Misc functions. - */ - -int -intel_create_context(struct egl_drm_context *egl_context, const __GLcontextModes *visual, void *sharedContextPrivate) -{ - struct intel_context *intel = CALLOC_STRUCT(intel_context); - struct intel_device *device = (struct intel_device *)egl_context->device->priv; - struct pipe_context *pipe; - struct st_context *st_share = NULL; - - egl_context->priv = intel; - - intel->intel_device = device; - intel->egl_context = egl_context; - intel->egl_device = egl_context->device; - - intel->base.hardware_lock = intel_lock_hardware; - intel->base.hardware_unlock = intel_unlock_hardware; - intel->base.hardware_locked = intel_locked_hardware; - - intel_be_init_context(&intel->base, &device->base); - -#if 0 - pipe = intel_create_softpipe(intel, screen->winsys); -#else - pipe = i915_create_context(device->pipe, &device->base.base, &intel->base.base); -#endif - - pipe->priv = intel; - - intel->st = st_create_context(pipe, visual, st_share); - - device->dummy = intel; - - return TRUE; -} - -int -intel_destroy_context(struct egl_drm_context *egl_context) -{ - struct intel_context *intel = egl_context->priv; - - if (intel->intel_device->dummy == intel) - intel->intel_device->dummy = NULL; - - st_destroy_context(intel->st); - intel_be_destroy_context(&intel->base); - free(intel); - return TRUE; -} - -void -intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *draw, struct egl_drm_drawable *read) -{ - if (context) { - struct intel_context *intel = (struct intel_context *)context->priv; - struct intel_framebuffer *draw_fb = (struct intel_framebuffer *)draw->priv; - struct intel_framebuffer *read_fb = (struct intel_framebuffer *)read->priv; - - assert(draw_fb->stfb); - assert(read_fb->stfb); - - st_make_current(intel->st, draw_fb->stfb, read_fb->stfb); - - intel->egl_drawable = draw; - - st_resize_framebuffer(draw_fb->stfb, draw->w, draw->h); - - if (draw != read) - st_resize_framebuffer(read_fb->stfb, read->w, read->h); - - } else { - st_make_current(NULL, NULL, NULL); - } -} - -void -intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer *front) -{ - struct intel_device *device = (struct intel_device *)draw->device->priv; - struct intel_framebuffer *draw_fb = (struct intel_framebuffer *)draw->priv; - - if (draw_fb->front_buffer) - driBOUnReference(draw_fb->front_buffer); - - draw_fb->front_buffer = NULL; - draw_fb->front = NULL; - - /* to unbind just call this function with front == NULL */ - if (!front) - return; - - draw_fb->front = front; - - driGenBuffers(device->base.staticPool, "front", 1, &draw_fb->front_buffer, 0, 0, 0); - driBOSetReferenced(draw_fb->front_buffer, front->handle); - - st_resize_framebuffer(draw_fb->stfb, draw->w, draw->h); -} diff --git a/src/gallium/winsys/egl_drm/intel/intel_context.h b/src/gallium/winsys/egl_drm/intel/intel_context.h deleted file mode 100644 index dfa4720b08..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_context.h +++ /dev/null @@ -1,118 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef INTEL_CONTEXT_H -#define INTEL_CONTEXT_H - -#include "pipe/p_debug.h" -#include "intel_drm/intel_be_context.h" - - -struct st_context; -struct egl_drm_device; -struct egl_drm_context; -struct egl_drm_frontbuffer; - - -/** - * Intel rendering context, contains a state tracker and intel-specific info. - */ -struct intel_context -{ - struct intel_be_context base; - - struct st_context *st; - - struct intel_device *intel_device; - - /* new egl stuff */ - struct egl_drm_device *egl_device; - struct egl_drm_context *egl_context; - struct egl_drm_drawable *egl_drawable; -}; - - - -/** - * Intel framebuffer. - */ -struct intel_framebuffer -{ - struct st_framebuffer *stfb; - - struct intel_device *device; - struct _DriBufferObject *front_buffer; - struct egl_drm_frontbuffer *front; -}; - - - - -/* These are functions now: - */ -void LOCK_HARDWARE( struct intel_context *intel ); -void UNLOCK_HARDWARE( struct intel_context *intel ); - -extern char *__progname; - - - -/* ================================================================ - * Debugging: - */ -#ifdef DEBUG -extern int __intel_debug; - -#define DEBUG_SWAP 0x1 -#define DEBUG_LOCK 0x2 -#define DEBUG_IOCTL 0x4 -#define DEBUG_BATCH 0x8 - -#define DBG(flag, ...) do { \ - if (__intel_debug & (DEBUG_##flag)) \ - printf(__VA_ARGS__); \ -} while(0) - -#else -#define DBG(flag, ...) -#endif - - -#define PCI_CHIP_845_G 0x2562 -#define PCI_CHIP_I830_M 0x3577 -#define PCI_CHIP_I855_GM 0x3582 -#define PCI_CHIP_I865_G 0x2572 -#define PCI_CHIP_I915_G 0x2582 -#define PCI_CHIP_I915_GM 0x2592 -#define PCI_CHIP_I945_G 0x2772 -#define PCI_CHIP_I945_GM 0x27A2 -#define PCI_CHIP_I945_GME 0x27AE -#define PCI_CHIP_G33_G 0x29C2 -#define PCI_CHIP_Q35_G 0x29B2 -#define PCI_CHIP_Q33_G 0x29D2 - -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_device.c b/src/gallium/winsys/egl_drm/intel/intel_device.c deleted file mode 100644 index b9649cbec7..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_device.c +++ /dev/null @@ -1,137 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#include "utils.h" - -#include "state_tracker/st_public.h" -#include "i915simple/i915_screen.h" - -#include "intel_context.h" -#include "intel_device.h" -#include "intel_batchbuffer.h" -#include "intel_egl.h" - - -extern const struct dri_extension card_extensions[]; - - -int -intel_create_device(struct egl_drm_device *device) -{ - struct intel_device *intel_device; - - /* Allocate the private area */ - intel_device = CALLOC_STRUCT(intel_device); - if (!intel_device) - return FALSE; - - device->priv = (void *)intel_device; - intel_device->device = device; - - intel_device->deviceID = device->deviceID; - - intel_be_init_device(&intel_device->base, device->drmFD, intel_device->deviceID); - - intel_device->pipe = i915_create_screen(&intel_device->base.base, intel_device->deviceID); - - /* hack */ - driInitExtensions(NULL, card_extensions, GL_FALSE); - - return TRUE; -} - -int -intel_destroy_device(struct egl_drm_device *device) -{ - struct intel_device *intel_device = (struct intel_device *)device->priv; - - intel_be_destroy_device(&intel_device->base); - - free(intel_device); - device->priv = NULL; - - return TRUE; -} - -int -intel_create_drawable(struct egl_drm_drawable *drawable, - const __GLcontextModes * visual) -{ - enum pipe_format colorFormat, depthFormat, stencilFormat; - struct intel_framebuffer *intelfb = CALLOC_STRUCT(intel_framebuffer); - - if (!intelfb) - return GL_FALSE; - - intelfb->device = drawable->device->priv; - - if (visual->redBits == 5) - colorFormat = PIPE_FORMAT_R5G6B5_UNORM; - else - colorFormat = PIPE_FORMAT_A8R8G8B8_UNORM; - - if (visual->depthBits == 16) - depthFormat = PIPE_FORMAT_Z16_UNORM; - else if (visual->depthBits == 24) - depthFormat = PIPE_FORMAT_S8Z24_UNORM; - else - depthFormat = PIPE_FORMAT_NONE; - - if (visual->stencilBits == 8) - stencilFormat = PIPE_FORMAT_S8Z24_UNORM; - else - stencilFormat = PIPE_FORMAT_NONE; - - intelfb->stfb = st_create_framebuffer(visual, - colorFormat, - depthFormat, - stencilFormat, - drawable->w, - drawable->h, - (void*) intelfb); - - if (!intelfb->stfb) { - free(intelfb); - return GL_FALSE; - } - - drawable->priv = (void *) intelfb; - return GL_TRUE; -} - -int -intel_destroy_drawable(struct egl_drm_drawable *drawable) -{ - struct intel_framebuffer *intelfb = (struct intel_framebuffer *)drawable->priv; - drawable->priv = NULL; - - assert(intelfb->stfb); - st_unreference_framebuffer(&intelfb->stfb); - free(intelfb); - return TRUE; -} diff --git a/src/gallium/winsys/egl_drm/intel/intel_device.h b/src/gallium/winsys/egl_drm/intel/intel_device.h deleted file mode 100644 index 2f9d4f887e..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_device.h +++ /dev/null @@ -1,50 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef _INTEL_SCREEN_H_ -#define _INTEL_SCREEN_H_ - -#include "intel_drm/intel_be_device.h" - -#include "pipe/p_compiler.h" - -struct pipe_screen; -struct egl_drm_device; -struct intel_context; - -struct intel_device -{ - struct intel_be_device base; - struct pipe_screen *pipe; - - int deviceID; - struct egl_drm_device *device; - - struct intel_context *dummy; -}; - -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.c b/src/gallium/winsys/egl_drm/intel/intel_egl.c deleted file mode 100644 index 1851babaf6..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.c +++ /dev/null @@ -1,796 +0,0 @@ - -#include -#include -#include -#include -#include - -#include "eglconfig.h" -#include "eglcontext.h" -#include "egldisplay.h" -#include "egldriver.h" -#include "eglglobals.h" -#include "eglmode.h" -#include "eglscreen.h" -#include "eglsurface.h" -#include "egllog.h" - -#include "intel_egl.h" - -#include "xf86drm.h" -#include "xf86drmMode.h" - -#include "intel_context.h" - -#include "state_tracker/st_public.h" - -#define MAX_SCREENS 16 - -static void -drm_get_device_id(struct egl_drm_device *device) -{ - char path[512]; - FILE *file; - - /* TODO get the real minor */ - int minor = 0; - - snprintf(path, sizeof(path), "/sys/class/drm/card%d/device/device", minor); - file = fopen(path, "r"); - if (!file) { - _eglLog(_EGL_WARNING, "Could not retrive device ID\n"); - return; - } - - fgets(path, sizeof( path ), file); - sscanf(path, "%x", &device->deviceID); - fclose(file); -} - -static struct egl_drm_device* -egl_drm_create_device(int drmFD) -{ - struct egl_drm_device *device = malloc(sizeof(*device)); - memset(device, 0, sizeof(*device)); - device->drmFD = drmFD; - - device->version = drmGetVersion(device->drmFD); - - drm_get_device_id(device); - - if (!intel_create_device(device)) { - free(device); - return NULL; - } - - return device; -} - -static void -_egl_context_modes_destroy(__GLcontextModes *modes) -{ - _eglLog(_EGL_DEBUG, "%s", __FUNCTION__); - - while (modes) { - __GLcontextModes * const next = modes->next; - free(modes); - modes = next; - } -} -/** - * Create a linked list of 'count' GLcontextModes. - * These are used during the client/server visual negotiation phase, - * then discarded. - */ -static __GLcontextModes * -_egl_context_modes_create(unsigned count, size_t minimum_size) -{ - /* This code copied from libGLX, and modified */ - const size_t size = (minimum_size > sizeof(__GLcontextModes)) - ? minimum_size : sizeof(__GLcontextModes); - __GLcontextModes * head = NULL; - __GLcontextModes ** next; - unsigned i; - - _eglLog(_EGL_DEBUG, "%s %d %d", __FUNCTION__, count, minimum_size); - - next = & head; - for (i = 0 ; i < count ; i++) { - *next = (__GLcontextModes *) calloc(1, size); - if (*next == NULL) { - _egl_context_modes_destroy(head); - head = NULL; - break; - } - - (*next)->doubleBufferMode = 1; - (*next)->visualID = GLX_DONT_CARE; - (*next)->visualType = GLX_DONT_CARE; - (*next)->visualRating = GLX_NONE; - (*next)->transparentPixel = GLX_NONE; - (*next)->transparentRed = GLX_DONT_CARE; - (*next)->transparentGreen = GLX_DONT_CARE; - (*next)->transparentBlue = GLX_DONT_CARE; - (*next)->transparentAlpha = GLX_DONT_CARE; - (*next)->transparentIndex = GLX_DONT_CARE; - (*next)->xRenderable = GLX_DONT_CARE; - (*next)->fbconfigID = GLX_DONT_CARE; - (*next)->swapMethod = GLX_SWAP_UNDEFINED_OML; - (*next)->bindToTextureRgb = GLX_DONT_CARE; - (*next)->bindToTextureRgba = GLX_DONT_CARE; - (*next)->bindToMipmapTexture = GLX_DONT_CARE; - (*next)->bindToTextureTargets = 0; - (*next)->yInverted = GLX_DONT_CARE; - - next = & ((*next)->next); - } - - return head; -} - -struct drm_screen; - -struct drm_driver -{ - _EGLDriver base; /* base class/object */ - - drmModeResPtr res; - - struct drm_screen *screens[MAX_SCREENS]; - size_t count_screens; - - struct egl_drm_device *device; -}; - -struct drm_surface -{ - _EGLSurface base; /* base class/object */ - - struct egl_drm_drawable *drawable; -}; - -struct drm_context -{ - _EGLContext base; /* base class/object */ - - struct egl_drm_context *context; -}; - -struct drm_screen -{ - _EGLScreen base; - - /* currently only support one connector */ - drmModeConnectorPtr connector; - - /* Has this screen been shown */ - int shown; - - /* Surface that is currently attached to this screen */ - struct drm_surface *surf; - - /* backing buffer */ - drmBO buffer; - - /* framebuffer */ - drmModeFBPtr fb; - uint32_t fbID; - - /* crtc and mode used */ - drmModeCrtcPtr crtc; - uint32_t crtcID; - - struct drm_mode_modeinfo *mode; - - /* geometry of the screen */ - struct egl_drm_frontbuffer front; -}; - -static void -drm_update_res(struct drm_driver *drm_drv) -{ - drmModeFreeResources(drm_drv->res); - drm_drv->res = drmModeGetResources(drm_drv->device->drmFD); -} - -static void -drm_add_modes_from_connector(_EGLScreen *screen, drmModeConnectorPtr connector) -{ - struct drm_mode_modeinfo *m; - int i; - - for (i = 0; i < connector->count_modes; i++) { - m = &connector->modes[i]; - _eglAddNewMode(screen, m->hdisplay, m->vdisplay, m->vrefresh, m->name); - } -} - - -static EGLBoolean -drm_initialize(_EGLDriver *drv, EGLDisplay dpy, EGLint *major, EGLint *minor) -{ - _EGLDisplay *disp = _eglLookupDisplay(dpy); - struct drm_driver *drm_drv = (struct drm_driver *)drv; - struct drm_screen *screen = NULL; - drmModeConnectorPtr connector = NULL; - drmModeResPtr res = NULL; - unsigned count_connectors = 0; - int num_screens = 0; - - EGLint i; - int fd; - - fd = drmOpen("i915", NULL); - if (fd < 0) { - return EGL_FALSE; - } - - drm_drv->device = egl_drm_create_device(fd); - if (!drm_drv->device) { - drmClose(fd); - return EGL_FALSE; - } - - drm_update_res(drm_drv); - res = drm_drv->res; - if (res) - count_connectors = res->count_connectors; - - for(i = 0; i < count_connectors && i < MAX_SCREENS; i++) { - connector = drmModeGetConnector(fd, res->connectors[i]); - - if (!connector) - continue; - - if (connector->connection != DRM_MODE_CONNECTED) { - drmModeFreeConnector(connector); - continue; - } - - screen = malloc(sizeof(struct drm_screen)); - memset(screen, 0, sizeof(*screen)); - screen->connector = connector; - _eglInitScreen(&screen->base); - _eglAddScreen(disp, &screen->base); - drm_add_modes_from_connector(&screen->base, connector); - drm_drv->screens[num_screens++] = screen; - } - drm_drv->count_screens = num_screens; - - /* for now we only have one config */ - _EGLConfig *config = calloc(1, sizeof(*config)); - memset(config, 1, sizeof(*config)); - _eglInitConfig(config, 1); - _eglSetConfigAttrib(config, EGL_RED_SIZE, 8); - _eglSetConfigAttrib(config, EGL_GREEN_SIZE, 8); - _eglSetConfigAttrib(config, EGL_BLUE_SIZE, 8); - _eglSetConfigAttrib(config, EGL_ALPHA_SIZE, 8); - _eglSetConfigAttrib(config, EGL_BUFFER_SIZE, 32); - _eglSetConfigAttrib(config, EGL_DEPTH_SIZE, 24); - _eglSetConfigAttrib(config, EGL_STENCIL_SIZE, 8); - _eglSetConfigAttrib(config, EGL_SURFACE_TYPE, EGL_PBUFFER_BIT); - _eglAddConfig(disp, config); - - drv->Initialized = EGL_TRUE; - - *major = 1; - *minor = 4; - - return EGL_TRUE; -} - -static void -drm_takedown_shown_screen(_EGLDriver *drv, struct drm_screen *screen) -{ - struct drm_driver *drm_drv = (struct drm_driver *)drv; - unsigned int i; - - intel_bind_frontbuffer(screen->surf->drawable, NULL); - screen->surf = NULL; - - for (i = 0; i < drm_drv->res->count_crtcs; i++) { - drmModeSetCrtc( - drm_drv->device->drmFD, - drm_drv->res->crtcs[i], - 0, // FD - 0, 0, - NULL, 0, // List of output ids - NULL); - } - - drmModeRmFB(drm_drv->device->drmFD, screen->fbID); - drmModeFreeFB(screen->fb); - screen->fb = NULL; - - drmBOUnreference(drm_drv->device->drmFD, &screen->buffer); - - screen->shown = 0; -} - -static EGLBoolean -drm_terminate(_EGLDriver *drv, EGLDisplay dpy) -{ - struct drm_driver *drm_drv = (struct drm_driver *)drv; - struct drm_screen *screen; - int i = 0; - - intel_destroy_device(drm_drv->device); - drmFreeVersion(drm_drv->device->version); - - for (i = 0; i < drm_drv->count_screens; i++) { - screen = drm_drv->screens[i]; - - if (screen->shown) - drm_takedown_shown_screen(drv, screen); - - drmModeFreeConnector(screen->connector); - _eglDestroyScreen(&screen->base); - drm_drv->screens[i] = NULL; - } - - drmClose(drm_drv->device->drmFD); - - free(drm_drv->device); - - _eglCleanupDisplay(_eglLookupDisplay(dpy)); - free(drm_drv); - - return EGL_TRUE; -} - - -static struct drm_context * -lookup_drm_context(EGLContext context) -{ - _EGLContext *c = _eglLookupContext(context); - return (struct drm_context *) c; -} - - -static struct drm_surface * -lookup_drm_surface(EGLSurface surface) -{ - _EGLSurface *s = _eglLookupSurface(surface); - return (struct drm_surface *) s; -} - -static struct drm_screen * -lookup_drm_screen(EGLDisplay dpy, EGLScreenMESA screen) -{ - _EGLScreen *s = _eglLookupScreen(dpy, screen); - return (struct drm_screen *) s; -} - -static __GLcontextModes* -visual_from_config(_EGLConfig *conf) -{ - __GLcontextModes *visual; - (void)conf; - - visual = _gl_context_modes_create(1, sizeof(*visual)); - visual->redBits = 8; - visual->greenBits = 8; - visual->blueBits = 8; - visual->alphaBits = 8; - - visual->rgbBits = 32; - visual->doubleBufferMode = 1; - - visual->depthBits = 24; - visual->haveDepthBuffer = visual->depthBits > 0; - visual->stencilBits = 8; - visual->haveStencilBuffer = visual->stencilBits > 0; - - return visual; -} - - - -static EGLContext -drm_create_context(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext share_list, const EGLint *attrib_list) -{ - struct drm_driver *drm_drv = (struct drm_driver *)drv; - struct drm_context *c; - struct drm_egl_context *share = NULL; - _EGLConfig *conf; - int i; - int ret; - __GLcontextModes *visual; - struct egl_drm_context *context; - - conf = _eglLookupConfig(drv, dpy, config); - if (!conf) { - _eglError(EGL_BAD_CONFIG, "eglCreateContext"); - return EGL_NO_CONTEXT; - } - - for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { - switch (attrib_list[i]) { - /* no attribs defined for now */ - default: - _eglError(EGL_BAD_ATTRIBUTE, "eglCreateContext"); - return EGL_NO_CONTEXT; - } - } - - c = (struct drm_context *) calloc(1, sizeof(struct drm_context)); - if (!c) - return EGL_NO_CONTEXT; - - _eglInitContext(drv, dpy, &c->base, config, attrib_list); - - context = malloc(sizeof(*context)); - memset(context, 0, sizeof(*context)); - - if (!context) - goto err_c; - - context->device = drm_drv->device; - visual = visual_from_config(conf); - - ret = intel_create_context(context, visual, share); - free(visual); - - if (!ret) - goto err_gl; - - c->context = context; - - /* generate handle and insert into hash table */ - _eglSaveContext(&c->base); - assert(_eglGetContextHandle(&c->base)); - - return _eglGetContextHandle(&c->base); -err_gl: - free(context); -err_c: - free(c); - return EGL_NO_CONTEXT; -} - -static EGLBoolean -drm_destroy_context(_EGLDriver *drv, EGLDisplay dpy, EGLContext context) -{ - struct drm_context *fc = lookup_drm_context(context); - _eglRemoveContext(&fc->base); - if (fc->base.IsBound) { - fc->base.DeletePending = EGL_TRUE; - } else { - intel_destroy_context(fc->context); - free(fc->context); - free(fc); - } - return EGL_TRUE; -} - - -static EGLSurface -drm_create_window_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativeWindowType window, const EGLint *attrib_list) -{ - return EGL_NO_SURFACE; -} - - -static EGLSurface -drm_create_pixmap_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, NativePixmapType pixmap, const EGLint *attrib_list) -{ - return EGL_NO_SURFACE; -} - - -static EGLSurface -drm_create_pbuffer_surface(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, - const EGLint *attrib_list) -{ - struct drm_driver *drm_drv = (struct drm_driver *)drv; - int i; - int ret; - int width = -1; - int height = -1; - struct drm_surface *surf = NULL; - struct egl_drm_drawable *drawable = NULL; - __GLcontextModes *visual; - _EGLConfig *conf; - - conf = _eglLookupConfig(drv, dpy, config); - if (!conf) { - _eglError(EGL_BAD_CONFIG, "eglCreatePbufferSurface"); - return EGL_NO_CONTEXT; - } - - for (i = 0; attrib_list && attrib_list[i] != EGL_NONE; i++) { - switch (attrib_list[i]) { - case EGL_WIDTH: - width = attrib_list[++i]; - break; - case EGL_HEIGHT: - height = attrib_list[++i]; - break; - default: - _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface"); - return EGL_NO_SURFACE; - } - } - - if (width < 1 || height < 1) { - _eglError(EGL_BAD_ATTRIBUTE, "eglCreatePbufferSurface"); - return EGL_NO_SURFACE; - } - - surf = (struct drm_surface *) calloc(1, sizeof(struct drm_surface)); - if (!surf) - goto err; - - if (!_eglInitSurface(drv, dpy, &surf->base, EGL_PBUFFER_BIT, config, attrib_list)) - goto err_surf; - - drawable = malloc(sizeof(*drawable)); - memset(drawable, 0, sizeof(*drawable)); - - drawable->w = width; - drawable->h = height; - - visual = visual_from_config(conf); - - drawable->device = drm_drv->device; - ret = intel_create_drawable(drawable, visual); - free(visual); - - if (!ret) - goto err_draw; - - surf->drawable = drawable; - - _eglSaveSurface(&surf->base); - return surf->base.Handle; - -err_draw: - free(drawable); -err_surf: - free(surf); -err: - return EGL_NO_SURFACE; -} - -static EGLSurface -drm_create_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, EGLConfig cfg, - const EGLint *attrib_list) -{ - EGLSurface surf = drm_create_pbuffer_surface(drv, dpy, cfg, attrib_list); - - return surf; -} - -static struct drm_mode_modeinfo * -drm_find_mode(drmModeConnectorPtr connector, _EGLMode *mode) -{ - int i; - struct drm_mode_modeinfo *m; - - for (i = 0; i < connector->count_modes; i++) { - m = &connector->modes[i]; - if (m->hdisplay == mode->Width && m->vdisplay == mode->Height && m->vrefresh == mode->RefreshRate) - break; - m = NULL; - } - - return m; -} -static void -draw(size_t x, size_t y, size_t w, size_t h, size_t pitch, size_t v, unsigned int *ptr) -{ - int i, j; - - for (i = x; i < x + w; i++) - for(j = y; j < y + h; j++) - ptr[(i * pitch / 4) + j] = v; - -} - -static void -prettyColors(int fd, unsigned int handle, size_t pitch) -{ - drmBO bo; - unsigned int *ptr; - void *p; - int i; - - drmBOReference(fd, handle, &bo); - drmBOMap(fd, &bo, DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE, 0, &p); - ptr = (unsigned int*)p; - - for (i = 0; i < (bo.size / 4); i++) - ptr[i] = 0xFFFFFFFF; - - for (i = 0; i < 4; i++) - draw(i * 40, i * 40, 40, 40, pitch, 0, ptr); - - - draw(200, 100, 40, 40, pitch, 0xff00ff, ptr); - draw(100, 200, 40, 40, pitch, 0xff00ff, ptr); - - drmBOUnmap(fd, &bo); -} - -static EGLBoolean -drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, - EGLScreenMESA screen, - EGLSurface surface, EGLModeMESA m) -{ - struct drm_driver *drm_drv = (struct drm_driver *)drv; - struct drm_surface *surf = lookup_drm_surface(surface); - struct drm_screen *scrn = lookup_drm_screen(dpy, screen); - _EGLMode *mode = _eglLookupMode(dpy, m); - size_t pitch = mode->Width * 4; - size_t size = mode->Height * pitch; - int ret; - unsigned int i,j,k; - - if (scrn->shown) - drm_takedown_shown_screen(drv, scrn); - - ret = drmBOCreate(drm_drv->device->drmFD, size, 0, 0, - DRM_BO_FLAG_READ | - DRM_BO_FLAG_WRITE | - DRM_BO_FLAG_MEM_TT | - DRM_BO_FLAG_MEM_VRAM | - DRM_BO_FLAG_NO_EVICT, - DRM_BO_HINT_DONT_FENCE, &scrn->buffer); - - if (ret) - return EGL_FALSE; - - prettyColors(drm_drv->device->drmFD, scrn->buffer.handle, pitch); - - ret = drmModeAddFB(drm_drv->device->drmFD, mode->Width, mode->Height, - 32, 32, pitch, - scrn->buffer.handle, - &scrn->fbID); - - if (ret) - goto err_bo; - - scrn->fb = drmModeGetFB(drm_drv->device->drmFD, scrn->fbID); - if (!scrn->fb) - goto err_bo; - - scrn->mode = drm_find_mode(scrn->connector, mode); - if (!scrn->mode) - goto err_fb; - - for (j = 0; j < drm_drv->res->count_connectors; j++) { - drmModeConnector *con = drmModeGetConnector(drm_drv->device->drmFD, drm_drv->res->connectors[j]); - for (k = 0; k < con->count_encoders; k++) { - drmModeEncoder *enc = drmModeGetEncoder(drm_drv->device->drmFD, con->encoders[k]); - for (i = 0; i < drm_drv->res->count_crtcs; i++) { - if (enc->possible_crtcs & (1<device->drmFD, - drm_drv->res->crtcs[i], - scrn->fbID, - 0, 0, - &drm_drv->res->connectors[j], 1, - scrn->mode); - /* skip the other crtcs now */ - i = drm_drv->res->count_crtcs; - } - } - } - } - - scrn->front.handle = scrn->buffer.handle; - scrn->front.pitch = pitch; - scrn->front.width = mode->Width; - scrn->front.height = mode->Height; - - scrn->surf = surf; - intel_bind_frontbuffer(surf->drawable, &scrn->front); - - scrn->shown = 1; - - return EGL_TRUE; - -err_fb: - /* TODO remove fb */ - -err_bo: - drmBOUnreference(drm_drv->device->drmFD, &scrn->buffer); - return EGL_FALSE; -} - -static EGLBoolean -drm_destroy_surface(_EGLDriver *drv, EGLDisplay dpy, EGLSurface surface) -{ - struct drm_surface *fs = lookup_drm_surface(surface); - _eglRemoveSurface(&fs->base); - if (fs->base.IsBound) { - fs->base.DeletePending = EGL_TRUE; - } else { - intel_bind_frontbuffer(fs->drawable, NULL); - intel_destroy_drawable(fs->drawable); - free(fs->drawable); - free(fs); - } - return EGL_TRUE; -} - - -static EGLBoolean -drm_make_current(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw, EGLSurface read, EGLContext context) -{ - struct drm_surface *readSurf = lookup_drm_surface(read); - struct drm_surface *drawSurf = lookup_drm_surface(draw); - struct drm_context *ctx = lookup_drm_context(context); - EGLBoolean b; - - b = _eglMakeCurrent(drv, dpy, draw, read, context); - if (!b) - return EGL_FALSE; - - if (ctx) { - if (!drawSurf || !readSurf) - return EGL_FALSE; - - intel_make_current(ctx->context, drawSurf->drawable, readSurf->drawable); - } else { - intel_make_current(NULL, NULL, NULL); - } - - return EGL_TRUE; -} - -static EGLBoolean -drm_swap_buffers(_EGLDriver *drv, EGLDisplay dpy, EGLSurface draw) -{ - struct drm_surface *surf = lookup_drm_surface(draw); - if (!surf) - return EGL_FALSE; - - /* error checking */ - if (!_eglSwapBuffers(drv, dpy, draw)) - return EGL_FALSE; - - intel_swap_buffers(surf->drawable); - return EGL_TRUE; -} - - -/** - * The bootstrap function. Return a new drm_driver object and - * plug in API functions. - */ -_EGLDriver * -_eglMain(_EGLDisplay *dpy, const char *args) -{ - struct drm_driver *drm; - - drm = (struct drm_driver *) calloc(1, sizeof(struct drm_driver)); - if (!drm) { - return NULL; - } - - /* First fill in the dispatch table with defaults */ - _eglInitDriverFallbacks(&drm->base); - /* then plug in our Drm-specific functions */ - drm->base.API.Initialize = drm_initialize; - drm->base.API.Terminate = drm_terminate; - drm->base.API.CreateContext = drm_create_context; - drm->base.API.MakeCurrent = drm_make_current; - drm->base.API.CreateWindowSurface = drm_create_window_surface; - drm->base.API.CreatePixmapSurface = drm_create_pixmap_surface; - drm->base.API.CreatePbufferSurface = drm_create_pbuffer_surface; - drm->base.API.DestroySurface = drm_destroy_surface; - drm->base.API.DestroyContext = drm_destroy_context; - drm->base.API.CreateScreenSurfaceMESA = drm_create_screen_surface_mesa; - drm->base.API.ShowScreenSurfaceMESA = drm_show_screen_surface_mesa; - drm->base.API.SwapBuffers = drm_swap_buffers; - - drm->base.ClientAPIsMask = EGL_OPENGL_BIT /*| EGL_OPENGL_ES_BIT*/; - drm->base.Name = "DRM/Gallium"; - - /* enable supported extensions */ - drm->base.Extensions.MESA_screen_surface = EGL_TRUE; - drm->base.Extensions.MESA_copy_context = EGL_TRUE; - - return &drm->base; -} diff --git a/src/gallium/winsys/egl_drm/intel/intel_egl.h b/src/gallium/winsys/egl_drm/intel/intel_egl.h deleted file mode 100644 index 1ee27e0847..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_egl.h +++ /dev/null @@ -1,53 +0,0 @@ - -#ifndef _INTEL_EGL_H_ -#define _INTEL_EGL_H_ - -#include - -struct egl_drm_device -{ - void *priv; - int drmFD; - - drmVersionPtr version; - int deviceID; -}; - -struct egl_drm_context -{ - void *priv; - struct egl_drm_device *device; -}; - -struct egl_drm_drawable -{ - void *priv; - struct egl_drm_device *device; - size_t h; - size_t w; -}; - -struct egl_drm_frontbuffer -{ - uint32_t handle; - uint32_t pitch; - uint32_t width; - uint32_t height; -}; - -#include "GL/internal/glcore.h" - -int intel_create_device(struct egl_drm_device *device); -int intel_destroy_device(struct egl_drm_device *device); - -int intel_create_context(struct egl_drm_context *context, const __GLcontextModes *visual, void *sharedContextPrivate); -int intel_destroy_context(struct egl_drm_context *context); - -int intel_create_drawable(struct egl_drm_drawable *drawable, const __GLcontextModes * visual); -int intel_destroy_drawable(struct egl_drm_drawable *drawable); - -void intel_make_current(struct egl_drm_context *context, struct egl_drm_drawable *draw, struct egl_drm_drawable *read); -void intel_swap_buffers(struct egl_drm_drawable *draw); -void intel_bind_frontbuffer(struct egl_drm_drawable *draw, struct egl_drm_frontbuffer *front); - -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_reg.h b/src/gallium/winsys/egl_drm/intel/intel_reg.h deleted file mode 100644 index 4f33bee438..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_reg.h +++ /dev/null @@ -1,53 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - - -#ifndef _INTEL_REG_H_ -#define _INTEL_REG_H_ - - -#define BR00_BITBLT_CLIENT 0x40000000 -#define BR00_OP_COLOR_BLT 0x10000000 -#define BR00_OP_SRC_COPY_BLT 0x10C00000 -#define BR13_SOLID_PATTERN 0x80000000 - -#define XY_COLOR_BLT_CMD ((2<<29)|(0x50<<22)|0x4) -#define XY_COLOR_BLT_WRITE_ALPHA (1<<21) -#define XY_COLOR_BLT_WRITE_RGB (1<<20) - -#define XY_SRC_COPY_BLT_CMD ((2<<29)|(0x53<<22)|6) -#define XY_SRC_COPY_BLT_WRITE_ALPHA (1<<21) -#define XY_SRC_COPY_BLT_WRITE_RGB (1<<20) - -#define MI_WAIT_FOR_EVENT ((0x3<<23)) -#define MI_WAIT_FOR_PLANE_B_FLIP (1<<6) -#define MI_WAIT_FOR_PLANE_A_FLIP (1<<2) - -#define MI_BATCH_BUFFER_END (0xA<<23) - - -#endif diff --git a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c b/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c deleted file mode 100644 index 2edcbc79ff..0000000000 --- a/src/gallium/winsys/egl_drm/intel/intel_swapbuffers.c +++ /dev/null @@ -1,111 +0,0 @@ -/************************************************************************** - * - * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "intel_device.h" -#include "intel_context.h" -#include "intel_batchbuffer.h" -#include "intel_reg.h" - -#include "pipe/p_context.h" -#include "state_tracker/st_public.h" -#include "state_tracker/st_context.h" -#include "state_tracker/st_cb_fbo.h" -#include "intel_egl.h" - - -static void -intel_display_surface(struct egl_drm_drawable *draw, - struct pipe_surface *surf); - -void intel_swap_buffers(struct egl_drm_drawable *draw) -{ - struct intel_framebuffer *intel_fb = (struct intel_framebuffer *)draw->priv; - struct pipe_surface *back_surf; - - assert(intel_fb); - assert(intel_fb->stfb); - - back_surf = st_get_framebuffer_surface(intel_fb->stfb, ST_SURFACE_BACK_LEFT); - if (back_surf) { - st_notify_swapbuffers(intel_fb->stfb); - if (intel_fb->front) - intel_display_surface(draw, back_surf); - st_notify_swapbuffers_complete(intel_fb->stfb); - } -} - -static void -intel_display_surface(struct egl_drm_drawable *draw, - struct pipe_surface *surf) -{ - struct intel_context *intel = NULL; - struct intel_framebuffer *intel_fb = (struct intel_framebuffer *)draw->priv; - struct _DriFenceObject *fence; - - //const int srcWidth = surf->width; - //const int srcHeight = surf->height; - - intel = intel_fb->device->dummy; - if (!intel) { - printf("No dummy context\n"); - return; - } - - const int dstWidth = intel_fb->front->width; - const int dstHeight = intel_fb->front->height; - const int dstPitch = intel_fb->front->pitch / 4;//draw->front.cpp; - - const int cpp = 4;//intel_fb->front->cpp; - const int srcPitch = surf->stride / cpp; - - int BR13, CMD; - //int i; - - BR13 = (dstPitch * cpp) | (0xCC << 16) | (1 << 24) | (1 << 25); - CMD = (XY_SRC_COPY_BLT_CMD | XY_SRC_COPY_BLT_WRITE_ALPHA | - XY_SRC_COPY_BLT_WRITE_RGB); - - BEGIN_BATCH(8, 2); - OUT_BATCH(CMD); - OUT_BATCH(BR13); - OUT_BATCH((0 << 16) | 0); - OUT_BATCH((dstHeight << 16) | dstWidth); - - OUT_RELOC(intel_fb->front_buffer, - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_WRITE, - DRM_BO_MASK_MEM | DRM_BO_FLAG_WRITE, 0); - - OUT_BATCH((0 << 16) | 0); - OUT_BATCH((srcPitch * cpp) & 0xffff); - OUT_RELOC(dri_bo(surf->buffer), - DRM_BO_FLAG_MEM_TT | DRM_BO_FLAG_READ, - DRM_BO_MASK_MEM | DRM_BO_FLAG_READ, 0); - - fence = intel_be_batchbuffer_flush(intel->base.batch); - driFenceUnReference(&fence); - intel_be_batchbuffer_finish(intel->base.batch); -} diff --git a/src/gallium/winsys/gdi/wmesa.c b/src/gallium/winsys/gdi/wmesa.c index 86b085ab84..ff52ceb8c4 100644 --- a/src/gallium/winsys/gdi/wmesa.c +++ b/src/gallium/winsys/gdi/wmesa.c @@ -118,48 +118,6 @@ static void wmSetPixelFormat(WMesaFramebuffer pwfb, HDC hDC) } } - -/** - * Create DIB for back buffer. - * We write into this memory with the span routines and then blit it - * to the window on a buffer swap. - */ -BOOL wmCreateBackingStore(WMesaFramebuffer pwfb, long lxSize, long lySize) -{ - HDC hdc = pwfb->hDC; - BITMAPINFO bmi; - LPBITMAPINFO pbmi = &bmi; - HDC hic; - - pbmi->bmiHeader.biSize = sizeof(BITMAPINFOHEADER); - pbmi->bmiHeader.biWidth = lxSize; - pbmi->bmiHeader.biHeight= -lySize; - pbmi->bmiHeader.biPlanes = 1; - pbmi->bmiHeader.biBitCount = pwfb->cColorBits; - pbmi->bmiHeader.biCompression = BI_RGB; - pbmi->bmiHeader.biSizeImage = 0; - pbmi->bmiHeader.biXPelsPerMeter = 0; - pbmi->bmiHeader.biYPelsPerMeter = 0; - pbmi->bmiHeader.biClrUsed = 0; - pbmi->bmiHeader.biClrImportant = 0; - - hic = CreateIC("display", NULL, NULL, NULL); - pwfb->dib_hDC = CreateCompatibleDC(hic); - - pwfb->hbmDIB = CreateDIBSection(hic, - pbmi, - DIB_RGB_COLORS, - (void **)&(pwfb->pbPixels), - 0, - 0); - pwfb->hOldBitmap = SelectObject(pwfb->dib_hDC, pwfb->hbmDIB); - - DeleteDC(hic); - - wmSetPixelFormat(pwfb, pwfb->hDC); - return TRUE; -} - /** * Create a new WMesaFramebuffer object which will correspond to the * given HDC (Window handle). @@ -201,10 +159,6 @@ wmesa_new_framebuffer(HDC hdc, GLvisual *visual, GLuint width, GLuint height) pwfb->cColorBits = GetDeviceCaps(hdc, BITSPIXEL); -#if 0 - wmCreateBackingStore(pwfb, width, height); -#endif - pwfb->hDC = hdc; /* insert at head of list */ pwfb->next = FirstFramebuffer; @@ -266,16 +220,6 @@ static WMesaContext wmesa_context(const GLcontext *ctx) return (WMesaContext) ctx; } -static wmDeleteBackingStore(WMesaFramebuffer pwfb) -{ - if (pwfb->hbmDIB) { - SelectObject(pwfb->dib_hDC, pwfb->hOldBitmap); - DeleteDC(pwfb->dib_hDC); - DeleteObject(pwfb->hbmDIB); - } -} - - /** * Find the width and height of the window named by hdc. */ @@ -383,11 +327,6 @@ wm_flush_frontbuffer(struct pipe_winsys *pws, struct wm_buffer *wm_buf; BITMAPINFO bmi, *pbmi; -#if 0 - if (pwfb) - BitBlt(pwfb->hDC, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, - pwfb->dib_hDC, 0, 0, SRCCOPY); -#else wm_buf = wm_buffer(surf->buffer); pbmi = &bmi; @@ -405,7 +344,6 @@ wm_flush_frontbuffer(struct pipe_winsys *pws, pbmi->bmiHeader.biClrImportant = 0; StretchDIBits(pwfb->hDC, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, wm_buf->data, pbmi, 0, SRCCOPY); -#endif } @@ -706,9 +644,6 @@ void WMesaDestroyContext( WMesaContext pwc ) /* clean up frame buffer resources */ pwfb = wmesa_lookup_framebuffer(pwc->hDC); if (pwfb) { -#if 0 - wmDeleteBackingStore(pwfb); -#endif wmesa_free_framebuffer(pwc->hDC); } @@ -779,10 +714,6 @@ void WMesaSwapBuffers( HDC hdc ) */ st_notify_swapbuffers(pwfb->stfb); -#if 0 - BitBlt(pwfb->hDC, 0, 0, pwfb->stfb->Base.Width, pwfb->stfb->Base.Height, - pwfb->dib_hDC, 0, 0, SRCCOPY); -#else surf = st_get_framebuffer_surface(pwfb->stfb, ST_SURFACE_BACK_LEFT); wm_buf = wm_buffer(surf->buffer); @@ -809,7 +740,6 @@ void WMesaSwapBuffers( HDC hdc ) st_resize_framebuffer(pwfb->stfb, width, height); } -#endif } /* This is hopefully a temporary hack to define some needed dispatch -- cgit v1.2.3 From 9bbc55116f36a357ee75d2766b0adb039eaca806 Mon Sep 17 00:00:00 2001 From: Alan Hourihane Date: Sun, 17 Aug 2008 21:55:37 +0100 Subject: if we can't find a mode, return first. At least we should see the top left portion if we've got larger screens on other CRTC's --- src/gallium/winsys/drm/intel/egl/intel_egl.c | 10 +++++----- 1 file changed, 5 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/winsys/drm/intel/egl/intel_egl.c b/src/gallium/winsys/drm/intel/egl/intel_egl.c index b89c5c508a..3204ed3131 100644 --- a/src/gallium/winsys/drm/intel/egl/intel_egl.c +++ b/src/gallium/winsys/drm/intel/egl/intel_egl.c @@ -571,7 +571,7 @@ drm_find_mode(drmModeConnectorPtr connector, _EGLMode *mode) m = &connector->modes[i]; if (m->hdisplay == mode->Width && m->vdisplay == mode->Height && m->vrefresh == mode->RefreshRate) break; - m = NULL; + m = &connector->modes[0]; /* if we can't find one, return first */ } return m; @@ -654,12 +654,12 @@ drm_show_screen_surface_mesa(_EGLDriver *drv, EGLDisplay dpy, if (!scrn->fb) goto err_bo; - scrn->mode = drm_find_mode(scrn->connector, mode); - if (!scrn->mode) - goto err_fb; - for (j = 0; j < drm_drv->res->count_connectors; j++) { drmModeConnector *con = drmModeGetConnector(drm_drv->device->drmFD, drm_drv->res->connectors[j]); + scrn->mode = drm_find_mode(con, mode); + if (!scrn->mode) + goto err_fb; + for (k = 0; k < con->count_encoders; k++) { drmModeEncoder *enc = drmModeGetEncoder(drm_drv->device->drmFD, con->encoders[k]); for (i = 0; i < drm_drv->res->count_crtcs; i++) { -- cgit v1.2.3 From 6fdc057887bd5e1e5f5ae2781529df050b3a31c6 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 14:40:02 +0200 Subject: scons: Fix build. --- src/gallium/winsys/SConscript | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/SConscript b/src/gallium/winsys/SConscript index ddab9efc86..30c3378dff 100644 --- a/src/gallium/winsys/SConscript +++ b/src/gallium/winsys/SConscript @@ -1,6 +1,6 @@ Import('*') -if env['drm']: +if env['dri']: SConscript([ 'drm/SConscript', ]) -- cgit v1.2.3 From 087ee474ed1dcf9b09c13227d442e6580f828f57 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 18 Aug 2008 08:49:41 -0600 Subject: mesa: rearrange some code in _mesa_BindTexture() to fix error detection bug 17173 Also, move GL_TEXTURE_RECTANGLE init code into separate function. --- src/mesa/main/texobj.c | 101 +++++++++++++++++++++++++++++-------------------- 1 file changed, 59 insertions(+), 42 deletions(-) (limited to 'src') diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c index c163a8158f..2a54ff7ff9 100644 --- a/src/mesa/main/texobj.c +++ b/src/mesa/main/texobj.c @@ -141,6 +141,34 @@ _mesa_initialize_texture_object( struct gl_texture_object *obj, } +/** + * Some texture initialization can't be finished until we know which + * target it's getting bound to (GL_TEXTURE_1D/2D/etc). + */ +static void +finish_texture_init(GLcontext *ctx, GLenum target, + struct gl_texture_object *obj) +{ + assert(obj->Target == 0); + + if (target == GL_TEXTURE_RECTANGLE_NV) { + /* have to init wrap and filter state here - kind of klunky */ + obj->WrapS = GL_CLAMP_TO_EDGE; + obj->WrapT = GL_CLAMP_TO_EDGE; + obj->WrapR = GL_CLAMP_TO_EDGE; + obj->MinFilter = GL_LINEAR; + if (ctx->Driver.TexParameter) { + static const GLfloat fparam_wrap[1] = {(GLfloat) GL_CLAMP_TO_EDGE}; + static const GLfloat fparam_filter[1] = {(GLfloat) GL_LINEAR}; + ctx->Driver.TexParameter(ctx, target, obj, GL_TEXTURE_WRAP_S, fparam_wrap); + ctx->Driver.TexParameter(ctx, target, obj, GL_TEXTURE_WRAP_T, fparam_wrap); + ctx->Driver.TexParameter(ctx, target, obj, GL_TEXTURE_WRAP_R, fparam_wrap); + ctx->Driver.TexParameter(ctx, target, obj, GL_TEXTURE_MIN_FILTER, fparam_filter); + } + } +} + + /** * Deallocate a texture object struct. It should have already been * removed from the texture object pool. @@ -830,44 +858,45 @@ _mesa_BindTexture( GLenum target, GLuint texName ) GET_CURRENT_CONTEXT(ctx); const GLuint unit = ctx->Texture.CurrentUnit; struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit]; - struct gl_texture_object *newTexObj = NULL; + struct gl_texture_object *newTexObj = NULL, *defaultTexObj = NULL; ASSERT_OUTSIDE_BEGIN_END(ctx); if (MESA_VERBOSE & (VERBOSE_API|VERBOSE_TEXTURE)) _mesa_debug(ctx, "glBindTexture %s %d\n", _mesa_lookup_enum_by_nr(target), (GLint) texName); + switch (target) { + case GL_TEXTURE_1D: + defaultTexObj = ctx->Shared->Default1D; + break; + case GL_TEXTURE_2D: + defaultTexObj = ctx->Shared->Default2D; + break; + case GL_TEXTURE_3D: + defaultTexObj = ctx->Shared->Default3D; + break; + case GL_TEXTURE_CUBE_MAP_ARB: + defaultTexObj = ctx->Shared->DefaultCubeMap; + break; + case GL_TEXTURE_RECTANGLE_NV: + defaultTexObj = ctx->Shared->DefaultRect; + break; + case GL_TEXTURE_1D_ARRAY_EXT: + defaultTexObj = ctx->Shared->Default1DArray; + break; + case GL_TEXTURE_2D_ARRAY_EXT: + defaultTexObj = ctx->Shared->Default2DArray; + break; + default: + _mesa_error(ctx, GL_INVALID_ENUM, "glBindTexture(target)"); + return; + } + /* * Get pointer to new texture object (newTexObj) */ if (texName == 0) { - /* newTexObj = a default texture object */ - switch (target) { - case GL_TEXTURE_1D: - newTexObj = ctx->Shared->Default1D; - break; - case GL_TEXTURE_2D: - newTexObj = ctx->Shared->Default2D; - break; - case GL_TEXTURE_3D: - newTexObj = ctx->Shared->Default3D; - break; - case GL_TEXTURE_CUBE_MAP_ARB: - newTexObj = ctx->Shared->DefaultCubeMap; - break; - case GL_TEXTURE_RECTANGLE_NV: - newTexObj = ctx->Shared->DefaultRect; - break; - case GL_TEXTURE_1D_ARRAY_EXT: - newTexObj = ctx->Shared->Default1DArray; - break; - case GL_TEXTURE_2D_ARRAY_EXT: - newTexObj = ctx->Shared->Default2DArray; - break; - default: - _mesa_error(ctx, GL_INVALID_ENUM, "glBindTexture(target)"); - return; - } + newTexObj = defaultTexObj; } else { /* non-default texture object */ @@ -880,20 +909,8 @@ _mesa_BindTexture( GLenum target, GLuint texName ) "glBindTexture(target mismatch)" ); return; } - if (newTexObj->Target == 0 && target == GL_TEXTURE_RECTANGLE_NV) { - /* have to init wrap and filter state here - kind of klunky */ - newTexObj->WrapS = GL_CLAMP_TO_EDGE; - newTexObj->WrapT = GL_CLAMP_TO_EDGE; - newTexObj->WrapR = GL_CLAMP_TO_EDGE; - newTexObj->MinFilter = GL_LINEAR; - if (ctx->Driver.TexParameter) { - static const GLfloat fparam_wrap[1] = {(GLfloat) GL_CLAMP_TO_EDGE}; - static const GLfloat fparam_filter[1] = {(GLfloat) GL_LINEAR}; - (*ctx->Driver.TexParameter)( ctx, target, newTexObj, GL_TEXTURE_WRAP_S, fparam_wrap ); - (*ctx->Driver.TexParameter)( ctx, target, newTexObj, GL_TEXTURE_WRAP_T, fparam_wrap ); - (*ctx->Driver.TexParameter)( ctx, target, newTexObj, GL_TEXTURE_WRAP_R, fparam_wrap ); - (*ctx->Driver.TexParameter)( ctx, target, newTexObj, GL_TEXTURE_MIN_FILTER, fparam_filter ); - } + if (newTexObj->Target == 0) { + finish_texture_init(ctx, target, newTexObj); } } else { -- cgit v1.2.3 From e2da7edd642198d7c515dbc0b9ba77d4286c3262 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 15:42:26 +0200 Subject: tgsi: Add condition code (CC) register. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 13 +++++++++++-- src/gallium/auxiliary/tgsi/tgsi_exec.h | 29 ++++++++++++++++++++++++----- 2 files changed, 35 insertions(+), 7 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index a96209db1a..7974915211 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -26,7 +26,7 @@ **************************************************************************/ /** - * TGSI interpretor/executor. + * TGSI interpreter/executor. * * Flow control information: * @@ -88,6 +88,8 @@ #define TEMP_OUTPUT_C TGSI_EXEC_TEMP_OUTPUT_C #define TEMP_PRIMITIVE_I TGSI_EXEC_TEMP_PRIMITIVE_I #define TEMP_PRIMITIVE_C TGSI_EXEC_TEMP_PRIMITIVE_C +#define TEMP_CC_I TGSI_EXEC_TEMP_CC_I +#define TEMP_CC_C TGSI_EXEC_TEMP_CC_C #define TEMP_3_I TGSI_EXEC_TEMP_THREE_I #define TEMP_3_C TGSI_EXEC_TEMP_THREE_C #define TEMP_HALF_I TGSI_EXEC_TEMP_HALF_I @@ -2539,6 +2541,13 @@ tgsi_exec_machine_run( struct tgsi_exec_machine *mach ) mach->Primitives[0] = 0; } + for (i = 0; i < QUAD_SIZE; i++) { + mach->Temps[TEMP_CC_I].xyzw[TEMP_CC_C].u[i] = + (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_X_SHIFT) | + (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_Y_SHIFT) | + (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_Z_SHIFT) | + (TGSI_EXEC_CC_EQ << TGSI_EXEC_CC_W_SHIFT); + } /* execute declarations (interpolants) */ for (i = 0; i < mach->NumDeclarations; i++) { diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h index 4f30650b07..c4e649e69c 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.h +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -25,7 +25,7 @@ * **************************************************************************/ -#if !defined TGSI_EXEC_H +#ifndef TGSI_EXEC_H #define TGSI_EXEC_H #include "pipe/p_compiler.h" @@ -140,11 +140,30 @@ struct tgsi_exec_labels #define TGSI_EXEC_TEMP_PRIMITIVE_I (TGSI_EXEC_NUM_TEMPS + 2) #define TGSI_EXEC_TEMP_PRIMITIVE_C 2 -#define TGSI_EXEC_TEMP_THREE_I (TGSI_EXEC_NUM_TEMPS + 2) -#define TGSI_EXEC_TEMP_THREE_C 3 +/* NVIDIA condition code (CC) vector + */ +#define TGSI_EXEC_CC_GT 0x01 +#define TGSI_EXEC_CC_EQ 0x02 +#define TGSI_EXEC_CC_LT 0x04 +#define TGSI_EXEC_CC_UN 0x08 + +#define TGSI_EXEC_CC_X_MASK 0x000000ff +#define TGSI_EXEC_CC_X_SHIFT 0 +#define TGSI_EXEC_CC_Y_MASK 0x0000ff00 +#define TGSI_EXEC_CC_Y_SHIFT 8 +#define TGSI_EXEC_CC_Z_MASK 0x00ff0000 +#define TGSI_EXEC_CC_Z_SHIFT 16 +#define TGSI_EXEC_CC_W_MASK 0xff000000 +#define TGSI_EXEC_CC_W_SHIFT 24 + +#define TGSI_EXEC_TEMP_CC_I (TGSI_EXEC_NUM_TEMPS + 2) +#define TGSI_EXEC_TEMP_CC_C 3 + +#define TGSI_EXEC_TEMP_THREE_I (TGSI_EXEC_NUM_TEMPS + 3) +#define TGSI_EXEC_TEMP_THREE_C 0 #define TGSI_EXEC_TEMP_HALF_I (TGSI_EXEC_NUM_TEMPS + 3) -#define TGSI_EXEC_TEMP_HALF_C 0 +#define TGSI_EXEC_TEMP_HALF_C 1 #define TGSI_EXEC_TEMP_R0 (TGSI_EXEC_NUM_TEMPS + 4) -- cgit v1.2.3 From 880b751e8e21cab21a0d522346300f463fd9c634 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 15:57:39 +0200 Subject: tgsi: Cosmetic changes. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 31 ++++++++++++------------------- 1 file changed, 12 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index 7974915211..6b19ab5c8e 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -63,6 +63,11 @@ #define TILE_BOTTOM_LEFT 2 #define TILE_BOTTOM_RIGHT 3 +#define CHAN_X 0 +#define CHAN_Y 1 +#define CHAN_Z 2 +#define CHAN_W 3 + /* * Shorthand locations of various utility registers (_I = Index, _C = Channel) */ @@ -96,9 +101,6 @@ #define TEMP_HALF_C TGSI_EXEC_TEMP_HALF_C #define TEMP_R0 TGSI_EXEC_TEMP_R0 -#define FOR_EACH_CHANNEL(CHAN)\ - for (CHAN = 0; CHAN < 4; CHAN++) - #define IS_CHANNEL_ENABLED(INST, CHAN)\ ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) @@ -106,11 +108,11 @@ ((INST).FullDstRegisters[1].DstRegister.WriteMask & (1 << (CHAN))) #define FOR_EACH_ENABLED_CHANNEL(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ + for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)\ if (IS_CHANNEL_ENABLED( INST, CHAN )) #define FOR_EACH_ENABLED_CHANNEL2(INST, CHAN)\ - FOR_EACH_CHANNEL( CHAN )\ + for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)\ if (IS_CHANNEL_ENABLED2( INST, CHAN )) @@ -118,14 +120,6 @@ #define UPDATE_EXEC_MASK(MACH) \ MACH->ExecMask = MACH->CondMask & MACH->LoopMask & MACH->ContMask & MACH->FuncMask - -#define CHAN_X 0 -#define CHAN_Y 1 -#define CHAN_Z 2 -#define CHAN_W 3 - - - /** * Initialize machine state by expanding tokens to full instructions, * allocating temporary storage, setting up constants, etc. @@ -1128,7 +1122,7 @@ store_dest( { union tgsi_exec_channel *dst; - switch( reg->DstRegister.File ) { + switch (reg->DstRegister.File) { case TGSI_FILE_NULL: return; @@ -1138,7 +1132,7 @@ store_dest( break; case TGSI_FILE_TEMPORARY: - assert(reg->DstRegister.Index < TGSI_EXEC_NUM_TEMPS); + assert( reg->DstRegister.Index < TGSI_EXEC_NUM_TEMPS ); dst = &mach->Temps[reg->DstRegister.Index].xyzw[chan_index]; break; @@ -1151,8 +1145,7 @@ store_dest( return; } - switch (inst->Instruction.Saturate) - { + switch (inst->Instruction.Saturate) { case TGSI_SAT_NONE: if (mach->ExecMask & 0x1) dst->i[0] = chan->i[0]; @@ -1166,8 +1159,8 @@ store_dest( case TGSI_SAT_ZERO_ONE: /* XXX need to obey ExecMask here */ - micro_max(dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C]); - micro_min(dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C]); + micro_max( dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); + micro_min( dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); break; case TGSI_SAT_MINUS_PLUS_ONE: -- cgit v1.2.3 From 6aacca106b0619d87015aece1a0b1d6332910926 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 17:03:56 +0200 Subject: tgsi: Respect condition codes when storing destination register. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 121 ++++++++++++++++++++++++++++++--- 1 file changed, 113 insertions(+), 8 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index 6b19ab5c8e..1217ee71a5 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1120,7 +1120,9 @@ store_dest( const struct tgsi_full_instruction *inst, uint chan_index ) { + uint i; union tgsi_exec_channel *dst; + uint execmask = mach->ExecMask; switch (reg->DstRegister.File) { case TGSI_FILE_NULL: @@ -1145,16 +1147,119 @@ store_dest( return; } + if (inst->InstructionExtNv.CondFlowEnable) { + union tgsi_exec_channel *cc = &mach->Temps[TEMP_CC_I].xyzw[TEMP_CC_C]; + uint swizzle; + uint shift; + uint mask; + uint test; + + /* Only CC0 supported. + */ + assert( inst->InstructionExtNv.CondFlowIndex < 1 ); + + switch (chan_index) { + case CHAN_X: + swizzle = inst->InstructionExtNv.CondSwizzleX; + break; + case CHAN_Y: + swizzle = inst->InstructionExtNv.CondSwizzleY; + break; + case CHAN_Z: + swizzle = inst->InstructionExtNv.CondSwizzleZ; + break; + case CHAN_W: + swizzle = inst->InstructionExtNv.CondSwizzleW; + break; + default: + assert( 0 ); + return; + } + + switch (swizzle) { + case TGSI_SWIZZLE_X: + shift = TGSI_EXEC_CC_X_SHIFT; + mask = TGSI_EXEC_CC_X_MASK; + break; + case TGSI_SWIZZLE_Y: + shift = TGSI_EXEC_CC_Y_SHIFT; + mask = TGSI_EXEC_CC_Y_MASK; + break; + case TGSI_SWIZZLE_Z: + shift = TGSI_EXEC_CC_Z_SHIFT; + mask = TGSI_EXEC_CC_Z_MASK; + break; + case TGSI_SWIZZLE_W: + shift = TGSI_EXEC_CC_W_SHIFT; + mask = TGSI_EXEC_CC_W_MASK; + break; + default: + assert( 0 ); + return; + } + + switch (inst->InstructionExtNv.CondMask) { + case TGSI_CC_GT: + test = ~(TGSI_EXEC_CC_GT << shift) & mask; + for (i = 0; i < QUAD_SIZE; i++) + if (cc->u[i] & test) + execmask &= ~(1 << i); + break; + + case TGSI_CC_EQ: + test = ~(TGSI_EXEC_CC_EQ << shift) & mask; + for (i = 0; i < QUAD_SIZE; i++) + if (cc->u[i] & test) + execmask &= ~(1 << i); + break; + + case TGSI_CC_LT: + test = ~(TGSI_EXEC_CC_LT << shift) & mask; + for (i = 0; i < QUAD_SIZE; i++) + if (cc->u[i] & test) + execmask &= ~(1 << i); + break; + + case TGSI_CC_GE: + test = ~((TGSI_EXEC_CC_GT | TGSI_EXEC_CC_EQ) << shift) & mask; + for (i = 0; i < QUAD_SIZE; i++) + if (cc->u[i] & test) + execmask &= ~(1 << i); + break; + + case TGSI_CC_LE: + test = ~((TGSI_EXEC_CC_LT | TGSI_EXEC_CC_EQ) << shift) & mask; + for (i = 0; i < QUAD_SIZE; i++) + if (cc->u[i] & test) + execmask &= ~(1 << i); + break; + + case TGSI_CC_NE: + test = ~((TGSI_EXEC_CC_GT | TGSI_EXEC_CC_LT | TGSI_EXEC_CC_UN) << shift) & mask; + for (i = 0; i < QUAD_SIZE; i++) + if (cc->u[i] & test) + execmask &= ~(1 << i); + break; + + case TGSI_CC_TR: + break; + + case TGSI_CC_FL: + for (i = 0; i < QUAD_SIZE; i++) + execmask &= ~(1 << i); + break; + + default: + assert( 0 ); + return; + } + } + switch (inst->Instruction.Saturate) { case TGSI_SAT_NONE: - if (mach->ExecMask & 0x1) - dst->i[0] = chan->i[0]; - if (mach->ExecMask & 0x2) - dst->i[1] = chan->i[1]; - if (mach->ExecMask & 0x4) - dst->i[2] = chan->i[2]; - if (mach->ExecMask & 0x8) - dst->i[3] = chan->i[3]; + for (i = 0; i < QUAD_SIZE; i++) + if (execmask & (1 << i)) + dst->i[i] = chan->i[i]; break; case TGSI_SAT_ZERO_ONE: -- cgit v1.2.3 From 56c30bf17b9f57efdb93ae5d1b801677535a9651 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 17:09:20 +0200 Subject: tgsi: Saturate modifier obeys ExecMask. Implement NVIDIA [-1;+1] saturate mode. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 22 ++++++++++++++++++---- 1 file changed, 18 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index 1217ee71a5..dfd22270c8 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1263,13 +1263,27 @@ store_dest( break; case TGSI_SAT_ZERO_ONE: - /* XXX need to obey ExecMask here */ - micro_max( dst, chan, &mach->Temps[TEMP_0_I].xyzw[TEMP_0_C] ); - micro_min( dst, dst, &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C] ); + for (i = 0; i < QUAD_SIZE; i++) + if (execmask & (1 << i)) { + if (chan->f[i] < 0.0f) + dst->f[i] = 0.0f; + else if (chan->f[i] > 1.0f) + dst->f[i] = 1.0f; + else + dst->i[i] = chan->i[i]; + } break; case TGSI_SAT_MINUS_PLUS_ONE: - assert( 0 ); + for (i = 0; i < QUAD_SIZE; i++) + if (execmask & (1 << i)) { + if (chan->f[i] < -1.0f) + dst->f[i] = -1.0f; + else if (chan->f[i] > 1.0f) + dst->f[i] = 1.0f; + else + dst->i[i] = chan->i[i]; + } break; default: -- cgit v1.2.3 From e9ec60097cfed372f202aa64aa04674448881f0e Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 17:18:58 +0200 Subject: tgsi: Update condition code vector when storing dest register. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 49 +++++++++++++++++++++++++++++++++- 1 file changed, 48 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index dfd22270c8..a2235de059 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1121,12 +1121,14 @@ store_dest( uint chan_index ) { uint i; + union tgsi_exec_channel null; union tgsi_exec_channel *dst; uint execmask = mach->ExecMask; switch (reg->DstRegister.File) { case TGSI_FILE_NULL: - return; + dst = &null; + break; case TGSI_FILE_OUTPUT: dst = &mach->Outputs[mach->Temps[TEMP_OUTPUT_I].xyzw[TEMP_OUTPUT_C].u[0] @@ -1289,6 +1291,51 @@ store_dest( default: assert( 0 ); } + + if (inst->InstructionExtNv.CondDstUpdate) { + union tgsi_exec_channel *cc = &mach->Temps[TEMP_CC_I].xyzw[TEMP_CC_C]; + uint shift; + uint mask; + + /* Only CC0 supported. + */ + assert( inst->InstructionExtNv.CondDstIndex < 1 ); + + switch (chan_index) { + case CHAN_X: + shift = TGSI_EXEC_CC_X_SHIFT; + mask = ~TGSI_EXEC_CC_X_MASK; + break; + case CHAN_Y: + shift = TGSI_EXEC_CC_Y_SHIFT; + mask = ~TGSI_EXEC_CC_Y_MASK; + break; + case CHAN_Z: + shift = TGSI_EXEC_CC_Z_SHIFT; + mask = ~TGSI_EXEC_CC_Z_MASK; + break; + case CHAN_W: + shift = TGSI_EXEC_CC_W_SHIFT; + mask = ~TGSI_EXEC_CC_W_MASK; + break; + default: + assert( 0 ); + return; + } + + for (i = 0; i < QUAD_SIZE; i++) + if (execmask & (1 << i)) { + cc->u[i] &= mask; + if (dst->f[i] < 0.0f) + cc->u[i] |= TGSI_EXEC_CC_LT << shift; + else if (dst->f[i] > 0.0f) + cc->u[i] |= TGSI_EXEC_CC_GT << shift; + else if (dst->f[i] == 0.0f) + cc->u[i] |= TGSI_EXEC_CC_EQ << shift; + else + cc->u[i] |= TGSI_EXEC_CC_UN << shift; + } + } } #define FETCH(VAL,INDEX,CHAN)\ -- cgit v1.2.3 From 235981d8719aa7f67bfe547bf96e5343020e0373 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 17:49:01 +0200 Subject: gallium: Remove TGSI_CC_UN. --- src/gallium/include/pipe/p_shader_tokens.h | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index 3ab0c7deaf..f3e339619d 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -469,12 +469,11 @@ struct tgsi_instruction_ext #define TGSI_CC_GT 0 #define TGSI_CC_EQ 1 #define TGSI_CC_LT 2 -#define TGSI_CC_UN 3 -#define TGSI_CC_GE 4 -#define TGSI_CC_LE 5 -#define TGSI_CC_NE 6 -#define TGSI_CC_TR 7 -#define TGSI_CC_FL 8 +#define TGSI_CC_GE 3 +#define TGSI_CC_LE 4 +#define TGSI_CC_NE 5 +#define TGSI_CC_TR 6 +#define TGSI_CC_FL 7 #define TGSI_SWIZZLE_X 0 #define TGSI_SWIZZLE_Y 1 -- cgit v1.2.3 From 145a45e9d6807add8229f1fc9c63fbb6951b1b88 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 18 Aug 2008 13:15:09 -0600 Subject: gallium: fix do_flip bug on glCopyTexImage / surface_copy() path --- src/mesa/state_tracker/st_cb_texture.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index cf194a937f..6177ac63f0 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -1098,6 +1098,10 @@ st_copy_texsubimage(GLcontext *ctx, stImage->face, stImage->level, destZ, PIPE_BUFFER_USAGE_GPU_WRITE); + if (do_flip) + srcY = strb->surface->height - srcY - height; + + /* for surface_copy(), y=0=top, always */ pipe->surface_copy(pipe, do_flip, /* dest */ -- cgit v1.2.3 From 6a31bb6ad8251ae977327e64562f373a89f55c70 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 16 Aug 2008 18:45:00 +0100 Subject: trace: Get the trace file from the GALLIUM_TRACE option itself. --- src/gallium/drivers/trace/README | 9 ++++----- src/gallium/drivers/trace/tr_dump.c | 7 +++++-- src/gallium/drivers/trace/tr_stream.c | 6 +----- src/gallium/drivers/trace/tr_stream.h | 3 ++- 4 files changed, 12 insertions(+), 13 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README index 5f0ae2c641..7e98ec2454 100644 --- a/src/gallium/drivers/trace/README +++ b/src/gallium/drivers/trace/README @@ -8,7 +8,6 @@ To use do ln -s libGL.so build/linux-x86-debug/gallium/winsys/xlib/libGL.so.1 export LD_LIBRARY_PATH=$PWD/build/linux-x86-debug/gallium/winsys/xlib - export GALLIUM_TRACE=y ensure the right libGL.so is being picked by doing @@ -16,11 +15,11 @@ ensure the right libGL.so is being picked by doing and then try running - glxinfo + GALLIUM_TRACE=tri.trace ./progs/trivial/tri -which should create a gallium.*.trace file, which is an XML file. You can view -copying trace.xsl and trace.css to the same directory, and opening with a -XSLT capable browser like Firefox or Internet Explorer. +which should create a tri.trace file, which is an XML file. You can view copying +trace.xsl to the same directory, and opening with a XSLT capable browser like +Firefox or Internet Explorer. -- Jose Fonseca diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index c711a56b94..ecd0d6830d 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -218,12 +218,15 @@ trace_dump_trace_close(void) boolean trace_dump_trace_begin() { - if(!debug_get_bool_option("GALLIUM_TRACE", FALSE)) + const char *filename; + + filename = debug_get_option("GALLIUM_TRACE", NULL); + if(!filename) return FALSE; if(!stream) { - stream = trace_stream_create("gallium", "trace"); + stream = trace_stream_create(filename); if(!stream) return FALSE; diff --git a/src/gallium/drivers/trace/tr_stream.c b/src/gallium/drivers/trace/tr_stream.c index aecc1286b8..d008dbac27 100644 --- a/src/gallium/drivers/trace/tr_stream.c +++ b/src/gallium/drivers/trace/tr_stream.c @@ -47,18 +47,14 @@ struct trace_stream struct trace_stream * -trace_stream_create(const char *name, const char *ext) +trace_stream_create(const char *filename) { struct trace_stream *stream; - static unsigned file_no = 0; - char filename[1024]; stream = CALLOC_STRUCT(trace_stream); if(!stream) goto error1; - snprintf(filename, sizeof(filename), "%s.%u.%s", name, file_no, ext); - #if defined(PIPE_OS_LINUX) stream->file = fopen(filename, "w"); if(!stream->file) diff --git a/src/gallium/drivers/trace/tr_stream.h b/src/gallium/drivers/trace/tr_stream.h index 679c4a725d..6111174d6a 100644 --- a/src/gallium/drivers/trace/tr_stream.h +++ b/src/gallium/drivers/trace/tr_stream.h @@ -39,11 +39,12 @@ #include "pipe/p_compiler.h" + struct trace_stream; struct trace_stream * -trace_stream_create(const char *name, const char *ext); +trace_stream_create(const char *filename); boolean trace_stream_write(struct trace_stream *stream, const void *data, size_t size); -- cgit v1.2.3 From d042f415fc9edcc174573fc2cc06afa373a7ef9b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 16 Aug 2008 19:44:37 +0100 Subject: trace: Use long longs to ensure covering 64bits integers. --- src/gallium/drivers/trace/tr_dump.c | 8 ++++---- src/gallium/drivers/trace/tr_dump.h | 4 ++-- 2 files changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index ecd0d6830d..1613a626df 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -302,14 +302,14 @@ void trace_dump_bool(int value) trace_dump_writef("%c", value ? '1' : '0'); } -void trace_dump_int(long int value) +void trace_dump_int(long long int value) { - trace_dump_writef("%li", value); + trace_dump_writef("%lli", value); } -void trace_dump_uint(long unsigned value) +void trace_dump_uint(long long unsigned value) { - trace_dump_writef("%lu", value); + trace_dump_writef("%llu", value); } void trace_dump_float(double value) diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index 14176a78e9..6ddc8fc15c 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -48,8 +48,8 @@ void trace_dump_arg_end(void); void trace_dump_ret_begin(void); void trace_dump_ret_end(void); void trace_dump_bool(int value); -void trace_dump_int(long int value); -void trace_dump_uint(long unsigned value); +void trace_dump_int(long long int value); +void trace_dump_uint(long long unsigned value); void trace_dump_float(double value); void trace_dump_bytes(const void *data, long unsigned size); void trace_dump_string(const char *str); -- cgit v1.2.3 From 747762f379f05d127865de77615d6e4dd46be191 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Sat, 16 Aug 2008 19:45:05 +0100 Subject: trace: Preliminary stream implementation for GDI. --- src/gallium/drivers/trace/SConscript | 3 +- src/gallium/drivers/trace/tr_stream.c | 108 ----------------------- src/gallium/drivers/trace/tr_stream_stdc.c | 104 ++++++++++++++++++++++ src/gallium/drivers/trace/tr_stream_wd.c | 134 +++++++++++++++++++++++++++++ 4 files changed, 240 insertions(+), 109 deletions(-) delete mode 100644 src/gallium/drivers/trace/tr_stream.c create mode 100644 src/gallium/drivers/trace/tr_stream_stdc.c create mode 100644 src/gallium/drivers/trace/tr_stream_wd.c (limited to 'src') diff --git a/src/gallium/drivers/trace/SConscript b/src/gallium/drivers/trace/SConscript index a21ced9176..5c49468c4e 100644 --- a/src/gallium/drivers/trace/SConscript +++ b/src/gallium/drivers/trace/SConscript @@ -9,7 +9,8 @@ trace = env.ConvenienceLibrary( 'tr_dump.c', 'tr_screen.c', 'tr_state.c', - 'tr_stream.c', + 'tr_stream_stdc.c', + 'tr_stream_wd.c', 'tr_texture.c', 'tr_winsys.c', ]) diff --git a/src/gallium/drivers/trace/tr_stream.c b/src/gallium/drivers/trace/tr_stream.c deleted file mode 100644 index d008dbac27..0000000000 --- a/src/gallium/drivers/trace/tr_stream.c +++ /dev/null @@ -1,108 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#include "pipe/p_config.h" - -#if defined(PIPE_OS_LINUX) -#include -#else -#error Unsupported platform -#endif - -#include "pipe/p_util.h" - -#include "tr_stream.h" - - -struct trace_stream -{ -#if defined(PIPE_OS_LINUX) - FILE *file; -#endif -}; - - -struct trace_stream * -trace_stream_create(const char *filename) -{ - struct trace_stream *stream; - - stream = CALLOC_STRUCT(trace_stream); - if(!stream) - goto error1; - -#if defined(PIPE_OS_LINUX) - stream->file = fopen(filename, "w"); - if(!stream->file) - goto error2; -#endif - - return stream; - -error2: - FREE(stream); -error1: - return NULL; -} - - -boolean -trace_stream_write(struct trace_stream *stream, const void *data, size_t size) -{ - if(!stream) - return FALSE; - -#if defined(PIPE_OS_LINUX) - return fwrite(data, size, 1, stream->file) == size ? TRUE : FALSE; -#endif -} - - -void -trace_stream_flush(struct trace_stream *stream) -{ - if(!stream) - return; - -#if defined(PIPE_OS_LINUX) - fflush(stream->file); -#endif -} - - -void -trace_stream_close(struct trace_stream *stream) -{ - if(!stream) - return; - -#if defined(PIPE_OS_LINUX) - fclose(stream->file); -#endif - - FREE(stream); -} diff --git a/src/gallium/drivers/trace/tr_stream_stdc.c b/src/gallium/drivers/trace/tr_stream_stdc.c new file mode 100644 index 0000000000..4c77e1c995 --- /dev/null +++ b/src/gallium/drivers/trace/tr_stream_stdc.c @@ -0,0 +1,104 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Stream implementation based on the Standard C Library. + */ + +#include "pipe/p_config.h" + +#if defined(PIPE_OS_LINUX) + +#include + +#include "pipe/p_util.h" + +#include "tr_stream.h" + + +struct trace_stream +{ + FILE *file; +}; + + +struct trace_stream * +trace_stream_create(const char *filename) +{ + struct trace_stream *stream; + + stream = CALLOC_STRUCT(trace_stream); + if(!stream) + goto error1; + + stream->file = fopen(filename, "w"); + if(!stream->file) + goto error2; + + return stream; + +error2: + FREE(stream); +error1: + return NULL; +} + + +boolean +trace_stream_write(struct trace_stream *stream, const void *data, size_t size) +{ + if(!stream) + return FALSE; + + return fwrite(data, size, 1, stream->file) == size ? TRUE : FALSE; +} + + +void +trace_stream_flush(struct trace_stream *stream) +{ + if(!stream) + return; + + fflush(stream->file); +} + + +void +trace_stream_close(struct trace_stream *stream) +{ + if(!stream) + return; + + fclose(stream->file); + + FREE(stream); +} + + +#endif diff --git a/src/gallium/drivers/trace/tr_stream_wd.c b/src/gallium/drivers/trace/tr_stream_wd.c new file mode 100644 index 0000000000..9a05c0cec8 --- /dev/null +++ b/src/gallium/drivers/trace/tr_stream_wd.c @@ -0,0 +1,134 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * @file + * Stream implementation for the Windows Display driver. + */ + +#include "pipe/p_config.h" + +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) + +#include +#include + +#include "pipe/p_util.h" + +#include "tr_stream.h" + + +#define MAP_FILE_SIZE (4*1024*1024) + + +struct trace_stream +{ + WCHAR wFileName[MAX_PATH + 1]; + ULONG_PTR iFile; + char *pMap; + size_t written; +}; + + +struct trace_stream * +trace_stream_create(const char *filename) +{ + struct trace_stream *stream; + ULONG BytesInUnicodeString; + + stream = CALLOC_STRUCT(trace_stream); + if(!stream) + goto error1; + + EngMultiByteToUnicodeN( + stream->wFileName, + sizeof(stream->wFileName), + &BytesInUnicodeString, + (char *)filename, + strlen(filename)); + + stream->pMap = EngMapFile(stream->wFileName, MAP_FILE_SIZE, &stream->iFile); + if(!stream->pMap) + goto error2; + + return stream; + +error2: + FREE(stream); +error1: + return NULL; +} + + +boolean +trace_stream_write(struct trace_stream *stream, const void *data, size_t size) +{ + boolean ret; + + if(!stream) + return FALSE; + + if(stream->written + size > MAP_FILE_SIZE) { + ret = FALSE; + size = MAP_FILE_SIZE - stream->written; + } + else + ret = TRUE; + + memcpy(stream->pMap + stream->written, data, size); + stream->written += size; + + return ret; +} + + +void +trace_stream_flush(struct trace_stream *stream) +{ + (void)stream; +} + + +void +trace_stream_close(struct trace_stream *stream) +{ + if(!stream) + return; + + EngUnmapFile(stream->iFile); + if(stream->written < MAP_FILE_SIZE) { + /* Truncate file size */ + stream->pMap = EngMapFile(stream->wFileName, stream->written, &stream->iFile); + if(stream->pMap) + EngUnmapFile(stream->iFile); + } + + FREE(stream); +} + + +#endif -- cgit v1.2.3 From 90a1c6e4032571a1c3e43daeb357068ba14136fe Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Mon, 18 Aug 2008 12:42:08 +0100 Subject: trace: Explain how to integrate with a state tracker or winsys. --- src/gallium/drivers/trace/README | 45 +++++++++++++++++++++++++++++++++++++--- 1 file changed, 42 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/README b/src/gallium/drivers/trace/README index 7e98ec2454..e7a2f12b02 100644 --- a/src/gallium/drivers/trace/README +++ b/src/gallium/drivers/trace/README @@ -1,9 +1,20 @@ + TRACE PIPE DRIVER + + += About = + This directory contains a Gallium3D pipe driver which traces all incoming calls. + += Build Instructions = + To build, invoke scons on the top dir as scons statetrackers=mesa drivers=softpipe,i915simple,trace winsys=xlib + += Usage = + To use do ln -s libGL.so build/linux-x86-debug/gallium/winsys/xlib/libGL.so.1 @@ -11,15 +22,43 @@ To use do ensure the right libGL.so is being picked by doing - ldd `which glxinfo` + ldd progs/trivial/tri and then try running - GALLIUM_TRACE=tri.trace ./progs/trivial/tri + GALLIUM_TRACE=tri.trace progs/trivial/tri which should create a tri.trace file, which is an XML file. You can view copying -trace.xsl to the same directory, and opening with a XSLT capable browser like +trace.xsl to the same directory, and opening with a XSLT capable browser such as Firefox or Internet Explorer. + += Integrating = + +You can integrate the trace pipe driver either inside the state tracker or the +winsys. The procedure on both cases is the same. Let's assume you have a +pipe_screen and a pipe_context pair obtained by the usual means (variable and +function names are just for illustration purposes): + + real_screen = real_screen_create(...); + + real_context = real_context_create(...); + +The trace screen and pipe_context is then created by doing + + trace_screen = trace_screen_create(real_screen); + + trace_context = trace_context_create(trace_screen, real_context); + +You can then simply use trace_screen and trace_context instead of real_screen +and real_context. + +Do not call trace_winsys_create. Simply pass trace_screen->winsys or +trace_context->winsys in places you would pass winsys. + +You can create as many contexts you wish. Just ensure that you don't mistake +trace_screen with real_screen when creating them. + + -- Jose Fonseca -- cgit v1.2.3 From e3f3e22cf7e9af4c5416d37329b7b8ee50f0cfcb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 18 Aug 2008 16:10:01 -0600 Subject: gallium: fix vertex program output translation/mapping bug In some cases, the vertex program output's semantic info didn't match up with the fragment program's input semantic info. This info is now explicitly passed into the st_translate_fragment_program() function. --- src/mesa/state_tracker/st_atom_shader.c | 60 ++++++++++++++++++++----- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_program.c | 74 +++++++++++++++++++------------ src/mesa/state_tracker/st_program.h | 8 +++- 4 files changed, 103 insertions(+), 41 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c index 4fa304ede0..6594e59e94 100644 --- a/src/mesa/state_tracker/st_atom_shader.c +++ b/src/mesa/state_tracker/st_atom_shader.c @@ -39,6 +39,7 @@ #include "main/imports.h" #include "main/mtypes.h" +#include "main/macros.h" #include "shader/program.h" #include "pipe/p_context.h" @@ -71,6 +72,8 @@ struct translated_vertex_program /** Maps VERT_RESULT_x to slot */ GLuint output_to_slot[VERT_RESULT_MAX]; + ubyte output_to_semantic_name[VERT_RESULT_MAX]; + ubyte output_to_semantic_index[VERT_RESULT_MAX]; /** Pointer to the translated vertex program */ struct st_vertex_program *vp; @@ -176,17 +179,22 @@ find_translated_vp(struct st_context *st, const GLbitfield outputsWritten = stvp->Base.Base.OutputsWritten; GLuint numVpOuts = 0; GLboolean emitPntSize = GL_FALSE, emitBFC0 = GL_FALSE, emitBFC1 = GL_FALSE; + GLint maxGeneric; /* Compute mapping of vertex program outputs to slots, which depends * on the fragment program's input->slot mapping. */ for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { - /* set default: */ + /* set defaults: */ xvp->output_to_slot[outAttr] = UNUSED; + xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_COUNT; + xvp->output_to_semantic_index[outAttr] = 99; if (outAttr == VERT_RESULT_HPOS) { /* always put xformed position into slot zero */ xvp->output_to_slot[VERT_RESULT_HPOS] = 0; + xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_POSITION; + xvp->output_to_semantic_index[outAttr] = 0; numVpOuts++; } else if (outputsWritten & (1 << outAttr)) { @@ -195,8 +203,11 @@ find_translated_vp(struct st_context *st, if (fpInAttrib >= 0) { GLuint fpInSlot = stfp->input_to_slot[fpInAttrib]; if (fpInSlot != ~0) { + /* match this vp output to the fp input */ GLuint vpOutSlot = stfp->input_map[fpInSlot]; xvp->output_to_slot[outAttr] = vpOutSlot; + xvp->output_to_semantic_name[outAttr] = stfp->input_semantic_name[fpInSlot]; + xvp->output_to_semantic_index[outAttr] = stfp->input_semantic_index[fpInSlot]; numVpOuts++; } } @@ -208,19 +219,27 @@ find_translated_vp(struct st_context *st, emitBFC1 = GL_TRUE; } #if 0 /*debug*/ - printf("assign output_to_slot[%d] = %d\n", outAttr, + printf("assign vp output_to_slot[%d] = %d\n", outAttr, xvp->output_to_slot[outAttr]); #endif } /* must do these last */ - if (emitPntSize) + if (emitPntSize) { xvp->output_to_slot[VERT_RESULT_PSIZ] = numVpOuts++; - if (emitBFC0) + xvp->output_to_semantic_name[VERT_RESULT_PSIZ] = TGSI_SEMANTIC_PSIZE; + xvp->output_to_semantic_index[VERT_RESULT_PSIZ] = 0; + } + if (emitBFC0) { xvp->output_to_slot[VERT_RESULT_BFC0] = numVpOuts++; - if (emitBFC1) + xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR; + xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 0; + } + if (emitBFC1) { xvp->output_to_slot[VERT_RESULT_BFC1] = numVpOuts++; - + xvp->output_to_semantic_name[VERT_RESULT_BFC0] = TGSI_SEMANTIC_COLOR; + xvp->output_to_semantic_index[VERT_RESULT_BFC0] = 1; + } /* Unneeded vertex program outputs will go to this slot. * We could use this info to do dead code elimination in the @@ -228,14 +247,33 @@ find_translated_vp(struct st_context *st, */ dummySlot = numVpOuts; - /* Map vert program outputs that aren't used to the dummy slot */ + /* find max GENERIC slot index */ + maxGeneric = -1; + for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { + if (xvp->output_to_semantic_name[outAttr] == TGSI_SEMANTIC_GENERIC) { + maxGeneric = MAX2(maxGeneric, + xvp->output_to_semantic_index[outAttr]); + } + } + + /* Map vert program outputs that aren't used to the dummy slot + * (and an unused generic attribute slot). + */ for (outAttr = 0; outAttr < VERT_RESULT_MAX; outAttr++) { if (outputsWritten & (1 << outAttr)) { - if (xvp->output_to_slot[outAttr] == UNUSED) + if (xvp->output_to_slot[outAttr] == UNUSED) { xvp->output_to_slot[outAttr] = dummySlot; + xvp->output_to_semantic_name[outAttr] = TGSI_SEMANTIC_GENERIC; + xvp->output_to_semantic_index[outAttr] = maxGeneric + 1; + } + + assert(xvp->output_to_semantic_name[outAttr] != TGSI_SEMANTIC_COUNT); + assert(xvp->output_to_semantic_index[outAttr] != 99); } + + #if 0 /*debug*/ - printf("output_to_slot[%d] = %d\n", outAttr, + printf("vp output_to_slot[%d] = %d\n", outAttr, xvp->output_to_slot[outAttr]); #endif @@ -243,7 +281,9 @@ find_translated_vp(struct st_context *st, assert(stvp->Base.Base.NumInstructions > 1); - st_translate_vertex_program(st, stvp, xvp->output_to_slot); + st_translate_vertex_program(st, stvp, xvp->output_to_slot, + xvp->output_to_semantic_name, + xvp->output_to_semantic_index); xvp->vp = stvp; diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index db0c9fbd09..0c5e21d4ff 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -294,7 +294,7 @@ st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor) } stvp = (struct st_vertex_program *) p; - st_translate_vertex_program(st, stvp, NULL); + st_translate_vertex_program(st, stvp, NULL, NULL, NULL); st->drawpix.vert_shaders[passColor] = stvp; diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 59314a3232..325d95e865 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -71,15 +71,17 @@ mem_dup(const void *src, uint size) /** * Translate a Mesa vertex shader into a TGSI shader. - * \param outputMapping to map vertex program output registers to TGSI - * output slots + * \param outputMapping to map vertex program output registers (VERT_RESULT_x) + * to TGSI output slots * \param tokensOut destination for TGSI tokens * \return pointer to cached pipe_shader object. */ void st_translate_vertex_program(struct st_context *st, struct st_vertex_program *stvp, - const GLuint outputMapping[]) + const GLuint outputMapping[], + const ubyte *outputSemanticName, + const ubyte *outputSemanticIndex) { struct pipe_context *pipe = st->pipe; struct tgsi_token tokens[ST_MAX_SHADER_TOKENS]; @@ -172,6 +174,20 @@ st_translate_vertex_program(struct st_context *st, } } +#if 0 + if (outputMapping && outputSemanticName) { + printf("VERT_RESULT written out_slot semantic_name semantic_index\n"); + for (attr = 0; attr < VERT_RESULT_MAX; attr++) { + printf(" %-2d %c %3d %2d %2d\n", + attr, + ((stvp->Base.Base.OutputsWritten & (1 << attr)) ? 'Y' : ' '), + outputMapping[attr], + outputSemanticName[attr], + outputSemanticIndex[attr]); + } + } +#endif + /* initialize output semantics to defaults */ for (i = 0; i < PIPE_MAX_SHADER_OUTPUTS; i++) { vs_output_semantic_name[i] = TGSI_SEMANTIC_GENERIC; @@ -203,10 +219,6 @@ st_translate_vertex_program(struct st_context *st, defaultOutputMapping[attr] = slot; } - /* - printf("Output %u -> slot %u\n", attr, slot); - */ - switch (attr) { case VERT_RESULT_HPOS: assert(slot == 0); @@ -248,15 +260,21 @@ st_translate_vertex_program(struct st_context *st, case VERT_RESULT_TEX5: case VERT_RESULT_TEX6: case VERT_RESULT_TEX7: - vs_output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_output_semantic_index[slot] = num_generic++; - break; + /* fall-through */ case VERT_RESULT_VAR0: /* fall-through */ default: - assert(attr - VERT_RESULT_VAR0 < MAX_VARYING); - vs_output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - vs_output_semantic_index[slot] = num_generic++; + if (outputSemanticName) { + /* use provided semantic into */ + assert(outputSemanticName[attr] != TGSI_SEMANTIC_COUNT); + vs_output_semantic_name[slot] = outputSemanticName[attr]; + vs_output_semantic_index[slot] = outputSemanticIndex[attr]; + } + else { + /* use default semantic info */ + vs_output_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + vs_output_semantic_index[slot] = num_generic++; + } } } } @@ -347,8 +365,6 @@ st_translate_fragment_program(struct st_context *st, GLuint num_generic = 0; GLuint num_tokens; - ubyte fs_input_semantic_name[PIPE_MAX_SHADER_INPUTS]; - ubyte fs_input_semantic_index[PIPE_MAX_SHADER_INPUTS]; uint fs_num_inputs = 0; ubyte fs_output_semantic_name[PIPE_MAX_SHADER_OUTPUTS]; @@ -378,23 +394,23 @@ st_translate_fragment_program(struct st_context *st, switch (attr) { case FRAG_ATTRIB_WPOS: - fs_input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; - fs_input_semantic_index[slot] = 0; + stfp->input_semantic_name[slot] = TGSI_SEMANTIC_POSITION; + stfp->input_semantic_index[slot] = 0; interpMode[slot] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_COL0: - fs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - fs_input_semantic_index[slot] = 0; + stfp->input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + stfp->input_semantic_index[slot] = 0; interpMode[slot] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_COL1: - fs_input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; - fs_input_semantic_index[slot] = 1; + stfp->input_semantic_name[slot] = TGSI_SEMANTIC_COLOR; + stfp->input_semantic_index[slot] = 1; interpMode[slot] = TGSI_INTERPOLATE_LINEAR; break; case FRAG_ATTRIB_FOGC: - fs_input_semantic_name[slot] = TGSI_SEMANTIC_FOG; - fs_input_semantic_index[slot] = 0; + stfp->input_semantic_name[slot] = TGSI_SEMANTIC_FOG; + stfp->input_semantic_index[slot] = 0; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; case FRAG_ATTRIB_TEX0: @@ -405,15 +421,15 @@ st_translate_fragment_program(struct st_context *st, case FRAG_ATTRIB_TEX5: case FRAG_ATTRIB_TEX6: case FRAG_ATTRIB_TEX7: - fs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - fs_input_semantic_index[slot] = num_generic++; + stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + stfp->input_semantic_index[slot] = num_generic++; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; break; case FRAG_ATTRIB_VAR0: /* fall-through */ default: - fs_input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; - fs_input_semantic_index[slot] = num_generic++; + stfp->input_semantic_name[slot] = TGSI_SEMANTIC_GENERIC; + stfp->input_semantic_index[slot] = num_generic++; interpMode[slot] = TGSI_INTERPOLATE_PERSPECTIVE; } } @@ -467,8 +483,8 @@ st_translate_fragment_program(struct st_context *st, /* inputs */ fs_num_inputs, inputMapping, - fs_input_semantic_name, - fs_input_semantic_index, + stfp->input_semantic_name, + stfp->input_semantic_index, interpMode, /* outputs */ fs_num_outputs, diff --git a/src/mesa/state_tracker/st_program.h b/src/mesa/state_tracker/st_program.h index 086e9391aa..078e2c42a6 100644 --- a/src/mesa/state_tracker/st_program.h +++ b/src/mesa/state_tracker/st_program.h @@ -58,6 +58,9 @@ struct st_fragment_program /** map FP input back to VP output */ GLuint input_map[PIPE_MAX_SHADER_INPUTS]; + ubyte input_semantic_name[PIPE_MAX_SHADER_INPUTS]; + ubyte input_semantic_index[PIPE_MAX_SHADER_INPUTS]; + struct pipe_shader_state state; void *driver_shader; @@ -143,6 +146,9 @@ st_translate_fragment_program(struct st_context *st, extern void st_translate_vertex_program(struct st_context *st, struct st_vertex_program *vp, - const GLuint vert_output_to_slot[]); + const GLuint vert_output_to_slot[], + const ubyte *fs_input_semantic_name, + const ubyte *fs_input_semantic_index); + #endif -- cgit v1.2.3 From a88de345cd2e8d79de5f5da36223b1db6adf1b3e Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 18 Aug 2008 16:12:39 -0600 Subject: egl: don't put the example/demo driver in the lib/ dir --- src/egl/drivers/demo/Makefile | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'src') diff --git a/src/egl/drivers/demo/Makefile b/src/egl/drivers/demo/Makefile index 5c05633df5..e5a7be2438 100644 --- a/src/egl/drivers/demo/Makefile +++ b/src/egl/drivers/demo/Makefile @@ -21,8 +21,7 @@ default: $(TOP)/$(LIB_DIR)/demodriver.so $(TOP)/$(LIB_DIR)/demodriver.so: $(OBJECTS) - $(TOP)/bin/mklib -o demodriver.so -noprefix \ - -install $(TOP)/$(LIB_DIR) $(OBJECTS) + $(TOP)/bin/mklib -o demodriver.so -noprefix $(OBJECTS) -- cgit v1.2.3 From c71161f1b1996d72fdc5398539c8de7cb0c34b6d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 18 Aug 2008 16:30:10 -0600 Subject: gallium: use PIPE_TEXTURE_USAGE_RENDER_TARGET for stencil renderbuffers --- src/mesa/state_tracker/st_format.c | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c index 2461f39622..a8ae30a454 100644 --- a/src/mesa/state_tracker/st_format.c +++ b/src/mesa/state_tracker/st_format.c @@ -342,6 +342,9 @@ default_depth_format(struct pipe_screen *screen, /** * Given an OpenGL internalFormat value for a texture or surface, return * the best matching PIPE_FORMAT_x, or PIPE_FORMAT_NONE if there's no match. + * \param target one of PIPE_TEXTURE_x + * \param tex_usage either PIPE_TEXTURE_USAGE_RENDER_TARGET + * or PIPE_TEXTURE_USAGE_SAMPLER */ enum pipe_format st_choose_format(struct pipe_context *pipe, GLint internalFormat, @@ -515,14 +518,35 @@ st_choose_format(struct pipe_context *pipe, GLint internalFormat, } +static GLboolean +is_stencil_format(GLenum format) +{ + switch (format) { + case GL_STENCIL_INDEX: + case GL_STENCIL_INDEX1_EXT: + case GL_STENCIL_INDEX4_EXT: + case GL_STENCIL_INDEX8_EXT: + case GL_STENCIL_INDEX16_EXT: + case GL_DEPTH_STENCIL_EXT: + case GL_DEPTH24_STENCIL8_EXT: + return GL_TRUE; + default: + return GL_FALSE; + } +} + /** * Called by FBO code to choose a PIPE_FORMAT_ for drawing surfaces. */ enum pipe_format st_choose_renderbuffer_format(struct pipe_context *pipe, GLint internalFormat) { - return st_choose_format(pipe, internalFormat, PIPE_TEXTURE_2D, - PIPE_TEXTURE_USAGE_RENDER_TARGET); + uint usage; + if (is_stencil_format(internalFormat)) + usage = PIPE_TEXTURE_USAGE_DEPTH_STENCIL; + else + usage = PIPE_TEXTURE_USAGE_RENDER_TARGET; + return st_choose_format(pipe, internalFormat, PIPE_TEXTURE_2D, usage); } -- cgit v1.2.3 From 93305bd6808787964c0088a88c428ecd1208aa44 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 18:09:23 +0200 Subject: tgsi: Use NUM_CHANNELS. --- src/gallium/auxiliary/tgsi/tgsi_sse2.c | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 47dc06faf6..0bf260c4dd 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -44,13 +44,13 @@ #define FOR_EACH_CHANNEL( CHAN )\ - for( CHAN = 0; CHAN < 4; CHAN++ ) + for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++) #define IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ ((INST).FullDstRegisters[0].DstRegister.WriteMask & (1 << (CHAN))) #define IF_IS_DST0_CHANNEL_ENABLED( INST, CHAN )\ - if( IS_DST0_CHANNEL_ENABLED( INST, CHAN )) + if (IS_DST0_CHANNEL_ENABLED( INST, CHAN )) #define FOR_EACH_DST0_ENABLED_CHANNEL( INST, CHAN )\ FOR_EACH_CHANNEL( CHAN )\ @@ -1172,7 +1172,7 @@ emit_instruction( { unsigned chan_index; - switch( inst->Instruction.Opcode ) { + switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: #if 0 /* XXX this isn't working properly (see glean vertProg1 test) */ -- cgit v1.2.3 From 2b512c0135bc8512cc80009ea7430f7cc0c869d6 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 22:21:32 +0200 Subject: tgsi: Workaround debug output buffer size limitations. --- src/gallium/auxiliary/tgsi/tgsi_dump.c | 21 ++++++++++++++++++--- 1 file changed, 18 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c index 4309d1bc76..a168c94928 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c @@ -518,8 +518,23 @@ tgsi_dump( uint flags ) { static char str[4096]; - + uint len; + char *p = str; + tgsi_dump_str(tokens, flags, str, sizeof(str)); - - debug_printf("%s", str); + + /* Workaround output buffer size limitations. + */ + len = strlen( str ); + while (len > 256) { + char piggy_bank; + + piggy_bank = p[256]; + p[256] = '\0'; + debug_printf( "%s", p ); + p[256] = piggy_bank; + p += 256; + len -= 256; + } + debug_printf( "%s", p ); } -- cgit v1.2.3 From 81dbc31f8b21aecdb91802dd4481057bd670f227 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 22:38:48 +0200 Subject: st: Fix immediate construction. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 198d406b59..372e4672f6 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -1,6 +1,6 @@ /************************************************************************** * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. * All Rights Reserved. * * Permission is hereby granted, free of charge, to any person obtaining a @@ -175,10 +175,11 @@ static struct tgsi_full_immediate make_immediate(const float *value, uint size) { struct tgsi_full_immediate imm; - imm.Immediate.Type = TGSI_TOKEN_TYPE_IMMEDIATE; - imm.Immediate.Size = 1 + size; /* one for the token itself */ + + imm = tgsi_default_full_immediate(); + imm.Immediate.Size += size; imm.Immediate.DataType = TGSI_IMM_FLOAT32; - imm.u.ImmediateFloat32 = (struct tgsi_immediate_float32 *) value; + imm.u.Pointer = value; return imm; } -- cgit v1.2.3 From 3eeaa943e264d9c61abfe5f4cf6351fdba0be4fd Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 22:51:15 +0200 Subject: st: Translate address register declaration. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 28 ++++++++++++++++++++++++++++ 1 file changed, 28 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index 372e4672f6..bcb23eefc3 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -610,6 +610,19 @@ make_temp_decl( return decl; } +static struct tgsi_full_declaration +make_addr_decl( + GLuint start_index, + GLuint end_index ) +{ + struct tgsi_full_declaration decl; + + decl = tgsi_default_full_declaration(); + decl.Declaration.File = TGSI_FILE_ADDRESS; + decl.DeclarationRange.First = start_index; + decl.DeclarationRange.Last = end_index; + return decl; +} static struct tgsi_full_declaration make_sampler_decl(GLuint index) @@ -828,6 +841,21 @@ tgsi_translate_mesa_program( } } + /* Address register. + */ + if (program->NumAddressRegs > 0) { + struct tgsi_full_declaration fulldecl; + + assert( program->NumAddressRegs == 1 ); + + fulldecl = make_addr_decl( 0, 0 ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); + } + /* immediates/literals */ memset(immediates, ~0, sizeof(immediates)); -- cgit v1.2.3 From 5d0a079fd00e5c9f1e9c96dd7bd17ecee57c52ce Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 18 Aug 2008 23:20:00 +0200 Subject: st: Do not translate program constants to immediates when indirect addressing used. --- src/mesa/state_tracker/st_mesa_to_tgsi.c | 136 +++++++++++++++++++------------ 1 file changed, 84 insertions(+), 52 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_mesa_to_tgsi.c b/src/mesa/state_tracker/st_mesa_to_tgsi.c index bcb23eefc3..5ec9fddd7f 100644 --- a/src/mesa/state_tracker/st_mesa_to_tgsi.c +++ b/src/mesa/state_tracker/st_mesa_to_tgsi.c @@ -47,7 +47,8 @@ static GLuint map_register_file( enum register_file file, GLuint index, - const GLuint immediateMapping[] ) + const GLuint immediateMapping[], + GLboolean indirectAccess ) { switch( file ) { case PROGRAM_UNDEFINED: @@ -66,11 +67,13 @@ map_register_file( case PROGRAM_STATE_VAR: case PROGRAM_NAMED_PARAM: case PROGRAM_UNIFORM: - if (immediateMapping && immediateMapping[index] != ~0) + if (!indirectAccess && immediateMapping && immediateMapping[index] != ~0) return TGSI_FILE_IMMEDIATE; else return TGSI_FILE_CONSTANT; case PROGRAM_CONSTANT: + if (indirectAccess) + return TGSI_FILE_CONSTANT; return TGSI_FILE_IMMEDIATE; case PROGRAM_INPUT: return TGSI_FILE_INPUT; @@ -98,7 +101,8 @@ map_register_file_index( GLuint index, const GLuint inputMapping[], const GLuint outputMapping[], - const GLuint immediateMapping[]) + const GLuint immediateMapping[], + GLboolean indirectAccess ) { switch( file ) { case TGSI_FILE_INPUT: @@ -109,6 +113,8 @@ map_register_file_index( return outputMapping[index]; case TGSI_FILE_IMMEDIATE: + if (indirectAccess) + return index; return immediateMapping[index]; default: @@ -190,6 +196,7 @@ compile_instruction( const GLuint inputMapping[], const GLuint outputMapping[], const GLuint immediateMapping[], + GLboolean indirectAccess, GLuint preamble_size, GLuint processor, GLboolean *insideSubroutine) @@ -205,29 +212,32 @@ compile_instruction( fullinst->Instruction.NumSrcRegs = _mesa_num_inst_src_regs( inst->Opcode ); fulldst = &fullinst->FullDstRegisters[0]; - fulldst->DstRegister.File = map_register_file( inst->DstReg.File, 0, NULL ); + fulldst->DstRegister.File = map_register_file( inst->DstReg.File, 0, NULL, GL_FALSE ); fulldst->DstRegister.Index = map_register_file_index( fulldst->DstRegister.File, inst->DstReg.Index, inputMapping, outputMapping, - NULL - ); + NULL, + GL_FALSE ); fulldst->DstRegister.WriteMask = convert_writemask( inst->DstReg.WriteMask ); - for( i = 0; i < fullinst->Instruction.NumSrcRegs; i++ ) { + for (i = 0; i < fullinst->Instruction.NumSrcRegs; i++) { GLuint j; fullsrc = &fullinst->FullSrcRegisters[i]; - fullsrc->SrcRegister.File = map_register_file( inst->SrcReg[i].File, - inst->SrcReg[i].Index, - immediateMapping ); + fullsrc->SrcRegister.File = map_register_file( + inst->SrcReg[i].File, + inst->SrcReg[i].Index, + immediateMapping, + indirectAccess ); fullsrc->SrcRegister.Index = map_register_file_index( fullsrc->SrcRegister.File, inst->SrcReg[i].Index, inputMapping, outputMapping, - immediateMapping); + immediateMapping, + indirectAccess ); /* swizzle (ext swizzle also depends on negation) */ { @@ -721,6 +731,7 @@ tgsi_translate_mesa_program( GLuint immediates[1000]; GLuint numImmediates = 0; GLboolean insideSubroutine = GL_FALSE; + GLboolean indirectAccess = GL_FALSE; assert(procType == TGSI_PROCESSOR_FRAGMENT || procType == TGSI_PROCESSOR_VERTEX); @@ -833,15 +844,15 @@ tgsi_translate_mesa_program( inside_range = GL_FALSE; fulldecl = make_temp_decl( start_range, i - 1 ); ti += tgsi_build_full_declaration( - &fulldecl, - &tokens[ti], - header, - maxTokens - ti ); + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); } } } - /* Address register. + /* Declare address register. */ if (program->NumAddressRegs > 0) { struct tgsi_full_declaration fulldecl; @@ -854,39 +865,57 @@ tgsi_translate_mesa_program( &tokens[ti], header, maxTokens - ti ); + + indirectAccess = GL_TRUE; } /* immediates/literals */ memset(immediates, ~0, sizeof(immediates)); - for (i = 0; program->Parameters && i < program->Parameters->NumParameters; - i++) { - if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { - struct tgsi_full_immediate fullimm - = make_immediate(program->Parameters->ParameterValues[i], 4); - ti += tgsi_build_full_immediate(&fullimm, - &tokens[ti], - header, - maxTokens - ti); - immediates[i] = numImmediates; - numImmediates++; + /* Emit immediates only when there is no address register in use. + * FIXME: Be smarter and recognize param arrays -- indirect addressing is + * only valid within the referenced array. + */ + if (program->Parameters && !indirectAccess) { + for (i = 0; i < program->Parameters->NumParameters; i++) { + if (program->Parameters->Parameters[i].Type == PROGRAM_CONSTANT) { + struct tgsi_full_immediate fullimm; + + fullimm = make_immediate( program->Parameters->ParameterValues[i], 4 ); + ti += tgsi_build_full_immediate( + &fullimm, + &tokens[ti], + header, + maxTokens - ti ); + immediates[i] = numImmediates; + numImmediates++; + } } } /* constant buffer refs */ - { + if (program->Parameters) { GLint start = -1, end = -1; - for (i = 0; - program->Parameters && i < program->Parameters->NumParameters; - i++) { + for (i = 0; i < program->Parameters->NumParameters; i++) { GLboolean emit = (i == program->Parameters->NumParameters - 1); + GLboolean matches; switch (program->Parameters->Parameters[i].Type) { case PROGRAM_ENV_PARAM: case PROGRAM_STATE_VAR: case PROGRAM_NAMED_PARAM: case PROGRAM_UNIFORM: + matches = GL_TRUE; + break; + case PROGRAM_CONSTANT: + matches = indirectAccess; + break; + default: + matches = GL_FALSE; + } + + if (matches) { if (start == -1) { /* begin a sequence */ start = i; @@ -896,8 +925,8 @@ tgsi_translate_mesa_program( /* continue sequence */ end = i; } - break; - default: + } + else { if (start != -1) { /* end of sequence */ emit = GL_TRUE; @@ -906,11 +935,13 @@ tgsi_translate_mesa_program( if (emit && start >= 0) { struct tgsi_full_declaration fulldecl; + fulldecl = make_constant_decl( start, end ); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); start = end = -1; } } @@ -920,25 +951,27 @@ tgsi_translate_mesa_program( for (i = 0; i < 8; i++) { if (program->SamplersUsed & (1 << i)) { struct tgsi_full_declaration fulldecl; + fulldecl = make_sampler_decl( i ); - ti += tgsi_build_full_declaration(&fulldecl, - &tokens[ti], - header, - maxTokens - ti ); + ti += tgsi_build_full_declaration( + &fulldecl, + &tokens[ti], + header, + maxTokens - ti ); } } - - for( i = 0; i < program->NumInstructions; i++ ) { + for (i = 0; i < program->NumInstructions; i++) { compile_instruction( - &program->Instructions[i], - &fullinst, - inputMapping, - outputMapping, - immediates, - preamble_size, - procType, - &insideSubroutine); + &program->Instructions[i], + &fullinst, + inputMapping, + outputMapping, + immediates, + indirectAccess, + preamble_size, + procType, + &insideSubroutine ); ti += tgsi_build_full_instruction( &fullinst, @@ -949,4 +982,3 @@ tgsi_translate_mesa_program( return ti; } - -- cgit v1.2.3 From de3083be7197cccb8dbf223d644ebdb78a8c809d Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 19 Aug 2008 00:42:30 +0200 Subject: tgsi: Fix ARL opcode in SSE2 codegen. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 6 +-- src/gallium/auxiliary/tgsi/tgsi_sse2.c | 87 +++++++++++++++++++++++++--------- 2 files changed, 68 insertions(+), 25 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index a2235de059..b793e1ebae 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1688,9 +1688,9 @@ exec_instruction( switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { - FETCH( &r[0], 0, chan_index ); - micro_f2it( &r[0], &r[0] ); - STORE( &r[0], 0, chan_index ); + FETCH( &r[0], 0, chan_index ); + micro_f2it( &r[0], &r[0] ); + STORE( &r[0], 0, chan_index ); } break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 0bf260c4dd..ca48d422e0 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -62,6 +62,7 @@ #define CHAN_W 3 #define TEMP_R0 TGSI_EXEC_TEMP_R0 +#define TEMP_ADDR TGSI_EXEC_TEMP_ADDR /** * X86 utility functions. @@ -215,19 +216,61 @@ emit_ret( static void emit_const( struct x86_function *func, - unsigned xmm, - unsigned vec, - unsigned chan ) -{ - sse_movss( - func, - make_xmm( xmm ), - get_const( vec, chan ) ); - sse_shufps( - func, - make_xmm( xmm ), - make_xmm( xmm ), - SHUF( 0, 0, 0, 0 ) ); + uint xmm, + int vec, + uint chan, + uint indirect, + uint indirectFile, + int indirectIndex ) +{ + if (indirect) { + struct x86_reg r0 = get_input_base(); + struct x86_reg r1 = get_output_base(); + uint i; + + assert( indirectFile == TGSI_FILE_ADDRESS ); + assert( indirectIndex == 0 ); + + x86_push( func, r0 ); + x86_push( func, r1 ); + + for (i = 0; i < QUAD_SIZE; i++) { + x86_lea( func, r0, get_const( vec, chan ) ); + x86_mov( func, r1, x86_make_disp( get_temp( TEMP_ADDR, CHAN_X ), i * 4 ) ); + + /* Quick hack to multiply by 16 -- need to add SHL to rtasm. + */ + x86_add( func, r1, r1 ); + x86_add( func, r1, r1 ); + x86_add( func, r1, r1 ); + x86_add( func, r1, r1 ); + + x86_add( func, r0, r1 ); + x86_mov( func, r1, x86_deref( r0 ) ); + x86_mov( func, x86_make_disp( get_temp( TEMP_R0, CHAN_X ), i * 4 ), r1 ); + } + + x86_pop( func, r1 ); + x86_pop( func, r0 ); + + sse_movaps( + func, + make_xmm( xmm ), + get_temp( TEMP_R0, CHAN_X ) ); + } + else { + assert( vec >= 0 ); + + sse_movss( + func, + make_xmm( xmm ), + get_const( vec, chan ) ); + sse_shufps( + func, + make_xmm( xmm ), + make_xmm( xmm ), + SHUF( 0, 0, 0, 0 ) ); + } } static void @@ -369,10 +412,12 @@ emit_addrs( unsigned vec, unsigned chan ) { + assert( vec == 0 ); + emit_temps( func, xmm, - vec + TGSI_EXEC_NUM_TEMPS, + vec + TGSI_EXEC_TEMP_ADDR, chan ); } @@ -855,18 +900,21 @@ emit_fetch( { unsigned swizzle = tgsi_util_get_full_src_register_extswizzle( reg, chan_index ); - switch( swizzle ) { + switch (swizzle) { case TGSI_EXTSWIZZLE_X: case TGSI_EXTSWIZZLE_Y: case TGSI_EXTSWIZZLE_Z: case TGSI_EXTSWIZZLE_W: - switch( reg->SrcRegister.File ) { + switch (reg->SrcRegister.File) { case TGSI_FILE_CONSTANT: emit_const( func, xmm, reg->SrcRegister.Index, - swizzle ); + swizzle, + reg->SrcRegister.Indirect, + reg->SrcRegisterInd.File, + reg->SrcRegisterInd.Index ); break; case TGSI_FILE_IMMEDIATE: @@ -1174,16 +1222,11 @@ emit_instruction( switch (inst->Instruction.Opcode) { case TGSI_OPCODE_ARL: -#if 0 - /* XXX this isn't working properly (see glean vertProg1 test) */ FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { FETCH( func, *inst, 0, 0, chan_index ); emit_f2it( func, 0 ); STORE( func, *inst, 0, 0, chan_index ); } -#else - return 0; -#endif break; case TGSI_OPCODE_MOV: -- cgit v1.2.3 From f9e14e0daca3b2f3662a68413c2fa71c05375b8e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Aug 2008 00:18:19 +0100 Subject: xlib: Revert destroying pipe_screen (temporary). --- src/gallium/winsys/xlib/xm_api.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_api.c b/src/gallium/winsys/xlib/xm_api.c index d2f92a0875..7256340420 100644 --- a/src/gallium/winsys/xlib/xm_api.c +++ b/src/gallium/winsys/xlib/xm_api.c @@ -847,7 +847,10 @@ void XMesaDestroyContext( XMesaContext c ) { struct pipe_screen *screen = c->st->pipe->screen; st_destroy_context(c->st); + /* FIXME: We should destroy the screen here, but if we do so, surfaces may + * outlive it, causing segfaults screen->destroy(screen); + */ _mesa_free(c); } -- cgit v1.2.3 From 7f2d349b2d27cb1f422677fdac1a9d92f61d8fce Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Aug 2008 02:24:13 +0100 Subject: mesa: Call pipe->destroy on context destruction for all platforms. --- src/mesa/state_tracker/st_context.c | 3 --- 1 file changed, 3 deletions(-) (limited to 'src') diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index d014ace59f..3b126e6c8f 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -228,10 +228,7 @@ void st_destroy_context( struct st_context *st ) cso_destroy_context(cso); - /* Temporary workaround for mismatched pipe create and pipe destroy */ -#if !defined(PIPE_SUBSYSTEM_WINDOWS_USER) && !defined(PIPE_SUBSYSTEM_WINDOWS_CE) pipe->destroy( pipe ); -#endif free(ctx); } -- cgit v1.2.3 From 0d9d2045e86a249a36d8eeebf59979a162c164af Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 19 Aug 2008 11:47:30 +0200 Subject: tgsi: Implement EXP opcode for SSE2. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 8 ++--- src/gallium/auxiliary/tgsi/tgsi_sse2.c | 64 ++++++++++++++++++++++++++-------- 2 files changed, 53 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index b793e1ebae..d584a4e4fa 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1756,18 +1756,18 @@ exec_instruction( micro_flr( &r[1], &r[0] ); /* r1 = floor(r0) */ if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { micro_exp2( &r[2], &r[1] ); /* r2 = 2 ^ r1 */ - STORE( &r[2], 0, CHAN_X ); /* store r2 */ + STORE( &r[2], 0, CHAN_X ); /* store r2 */ } if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { micro_sub( &r[2], &r[0], &r[1] ); /* r2 = r0 - r1 */ - STORE( &r[2], 0, CHAN_Y ); /* store r2 */ + STORE( &r[2], 0, CHAN_Y ); /* store r2 */ } if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { micro_exp2( &r[2], &r[0] ); /* r2 = 2 ^ r0 */ - STORE( &r[2], 0, CHAN_Z ); /* store r2 */ + STORE( &r[2], 0, CHAN_Z ); /* store r2 */ } if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); } break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index ca48d422e0..684bc081e5 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -61,6 +61,9 @@ #define CHAN_Z 2 #define CHAN_W 3 +#define TEMP_ONE_I TGSI_EXEC_TEMP_ONE_I +#define TEMP_ONE_C TGSI_EXEC_TEMP_ONE_C + #define TEMP_R0 TGSI_EXEC_TEMP_R0 #define TEMP_ADDR TGSI_EXEC_TEMP_ADDR @@ -958,8 +961,8 @@ emit_fetch( emit_tempf( func, xmm, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); + TEMP_ONE_I, + TEMP_ONE_C ); break; default: @@ -1173,8 +1176,8 @@ emit_setcc( func, make_xmm( 0 ), get_temp( - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ) ); + TEMP_ONE_I, + TEMP_ONE_C ) ); STORE( func, *inst, 0, 0, chan_index ); } } @@ -1243,8 +1246,8 @@ emit_instruction( emit_tempf( func, 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C); + TEMP_ONE_I, + TEMP_ONE_C); if( IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) ) { STORE( func, *inst, 0, 0, CHAN_X ); } @@ -1329,7 +1332,38 @@ emit_instruction( break; case TGSI_OPCODE_EXP: - return 0; + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( func, *inst, 0, 0, CHAN_X ); + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y )) { + emit_MOV( func, 1, 0 ); + emit_flr( func, 1 ); + /* dst.x = ex2(floor(src.x)) */ + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X )) { + emit_MOV( func, 2, 1 ); + emit_ex2( func, 2 ); + STORE( func, *inst, 2, 0, CHAN_X ); + } + /* dst.y = src.x - floor(src.x) */ + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y )) { + emit_MOV( func, 2, 0 ); + emit_sub( func, 2, 1 ); + STORE( func, *inst, 2, 0, CHAN_Y ); + } + } + /* dst.z = ex2(src.x) */ + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z )) { + emit_ex2( func, 0 ); + STORE( func, *inst, 0, 0, CHAN_Z ); + } + } + /* dst.w = 1.0 */ + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W )) { + emit_tempf( func, 0, TEMP_ONE_I, TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } break; case TGSI_OPCODE_LOG: @@ -1399,8 +1433,8 @@ emit_instruction( emit_tempf( func, 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); + TEMP_ONE_I, + TEMP_ONE_C ); STORE( func, *inst, 0, 0, CHAN_X ); } IF_IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) { @@ -1603,8 +1637,8 @@ emit_instruction( emit_tempf( func, 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); + TEMP_ONE_I, + TEMP_ONE_C ); STORE( func, *inst, 0, 0, CHAN_W ); } break; @@ -1731,8 +1765,8 @@ emit_instruction( emit_tempf( func, 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); + TEMP_ONE_I, + TEMP_ONE_C ); FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) { STORE( func, *inst, 0, 0, chan_index ); } @@ -1820,8 +1854,8 @@ emit_instruction( emit_tempf( func, 0, - TGSI_EXEC_TEMP_ONE_I, - TGSI_EXEC_TEMP_ONE_C ); + TEMP_ONE_I, + TEMP_ONE_C ); STORE( func, *inst, 0, 0, CHAN_W ); } break; -- cgit v1.2.3 From 4405e428e4c2a80172c803cc3c4933d546bf2b4d Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 19 Aug 2008 12:07:25 +0200 Subject: tgsi: Implement LOG opcode for SSE2 codegen. --- src/gallium/auxiliary/tgsi/tgsi_exec.c | 8 ++++---- src/gallium/auxiliary/tgsi/tgsi_sse2.c | 33 ++++++++++++++++++++++++++++++++- 2 files changed, 36 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index d584a4e4fa..88a34a6961 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -1777,18 +1777,18 @@ exec_instruction( micro_lg2( &r[1], &r[2] ); /* r1 = lg2(r2) */ micro_flr( &r[0], &r[1] ); /* r0 = floor(r1) */ if (IS_CHANNEL_ENABLED( *inst, CHAN_X )) { - STORE( &r[0], 0, CHAN_X ); + STORE( &r[0], 0, CHAN_X ); } if (IS_CHANNEL_ENABLED( *inst, CHAN_Y )) { micro_exp2( &r[0], &r[0] ); /* r0 = 2 ^ r0 */ micro_div( &r[0], &r[2], &r[0] ); /* r0 = r2 / r0 */ - STORE( &r[0], 0, CHAN_Y ); + STORE( &r[0], 0, CHAN_Y ); } if (IS_CHANNEL_ENABLED( *inst, CHAN_Z )) { - STORE( &r[1], 0, CHAN_Z ); + STORE( &r[1], 0, CHAN_Z ); } if (IS_CHANNEL_ENABLED( *inst, CHAN_W )) { - STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); + STORE( &mach->Temps[TEMP_1_I].xyzw[TEMP_1_C], 0, CHAN_W ); } break; diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 684bc081e5..485e5a0e6f 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -1367,7 +1367,38 @@ emit_instruction( break; case TGSI_OPCODE_LOG: - return 0; + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z )) { + FETCH( func, *inst, 0, 0, CHAN_X ); + emit_abs( func, 0 ); + emit_MOV( func, 1, 0 ); + emit_lg2( func, 1 ); + /* dst.z = lg2(abs(src.x)) */ + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Z )) { + STORE( func, *inst, 1, 0, CHAN_Z ); + } + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X ) || + IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y )) { + emit_flr( func, 1 ); + /* dst.x = floor(lg2(abs(src.x))) */ + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_X )) { + STORE( func, *inst, 1, 0, CHAN_X ); + } + /* dst.x = abs(src)/ex2(floor(lg2(abs(src.x)))) */ + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_Y )) { + emit_ex2( func, 1 ); + emit_rcp( func, 1, 1 ); + emit_mul( func, 0, 1 ); + STORE( func, *inst, 0, 0, CHAN_Y ); + } + } + } + /* dst.w = 1.0 */ + if (IS_DST0_CHANNEL_ENABLED( *inst, CHAN_W )) { + emit_tempf( func, 0, TEMP_ONE_I, TEMP_ONE_C ); + STORE( func, *inst, 0, 0, CHAN_W ); + } break; case TGSI_OPCODE_MUL: -- cgit v1.2.3 From 146a0fba0011e21a4831a4b86969d7cd3f093bf6 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Tue, 19 Aug 2008 15:32:09 +0200 Subject: gallium: Add note about vs_2_0 EXPP mapping. --- src/gallium/include/pipe/p_shader_tokens.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h index f3e339619d..a562e3a6b1 100644 --- a/src/gallium/include/pipe/p_shader_tokens.h +++ b/src/gallium/include/pipe/p_shader_tokens.h @@ -382,6 +382,7 @@ struct tgsi_immediate_float32 */ #define TGSI_OPCODE_SGN TGSI_OPCODE_SSG #define TGSI_OPCODE_MOVA TGSI_OPCODE_ARR +/* EXPP - use TGSI_OPCODE_EX2 */ /* * vs_2_x -- cgit v1.2.3 From df9bd01ea0c8a4d5b5ffecad4482ca6b4f2db33d Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 10:05:11 -0600 Subject: mesa: glsl: put var emit/ref code into separate functions --- src/mesa/shader/slang/slang_emit.c | 134 ++++++++++++++++++++++--------------- 1 file changed, 81 insertions(+), 53 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index f0c3791f2a..cad983443a 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1636,6 +1636,84 @@ emit_struct_field(slang_emit_info *emitInfo, slang_ir_node *n) } +/** + * Emit code for a variable declaration. + * This usually doesn't result in any code generation, but just + * memory allocation. + */ +static struct prog_instruction * +emit_var_decl(slang_emit_info *emitInfo, slang_ir_node *n) +{ + struct prog_instruction *inst; + + assert(n->Store); + assert(n->Store->File != PROGRAM_UNDEFINED); + assert(n->Store->Size > 0); + /*assert(n->Store->Index < 0);*/ + + if (!n->Var || n->Var->isTemp) { + /* a nameless/temporary variable, will be freed after first use */ + /*NEW*/ + if (n->Store->Index < 0 && !_slang_alloc_temp(emitInfo->vt, n->Store)) { + slang_info_log_error(emitInfo->log, + "Ran out of registers, too many temporaries"); + return NULL; + } + } + else { + /* a regular variable */ + _slang_add_variable(emitInfo->vt, n->Var); + if (!_slang_alloc_var(emitInfo->vt, n->Store)) { + slang_info_log_error(emitInfo->log, + "Ran out of registers, too many variables"); + return NULL; + } + /* + printf("IR_VAR_DECL %s %d store %p\n", + (char*) n->Var->a_name, n->Store->Index, (void*) n->Store); + */ + assert(n->Var->aux == n->Store); + } + if (emitInfo->EmitComments) { + /* emit NOP with comment describing the variable's storage location */ + char s[1000]; + sprintf(s, "TEMP[%d]%s = variable %s (size %d)", + n->Store->Index, + _mesa_swizzle_string(n->Store->Swizzle, 0, GL_FALSE), + (n->Var ? (char *) n->Var->a_name : "anonymous"), + n->Store->Size); + inst = emit_comment(emitInfo, s); + return inst; + } + return NULL; +} + + +/** + * Emit code for a reference to a variable. + * Actually, no code is generated but we may do some memory alloation. + * In particular, state vars (uniforms) are allocated on an as-needed basis. + */ +static struct prog_instruction * +emit_var_ref(slang_emit_info *emitInfo, slang_ir_node *n) +{ + assert(n->Store); + assert(n->Store->File != PROGRAM_UNDEFINED); + + if (n->Store->File == PROGRAM_STATE_VAR && n->Store->Index < 0) { + n->Store->Index = _slang_alloc_statevar(n, emitInfo->prog->Parameters); + } + + if (n->Store->Index < 0) { + /* probably ran out of registers */ + return NULL; + } + assert(n->Store->Size > 0); + + return NULL; +} + + static struct prog_instruction * emit(slang_emit_info *emitInfo, slang_ir_node *n) { @@ -1671,64 +1749,14 @@ emit(slang_emit_info *emitInfo, slang_ir_node *n) case IR_VAR_DECL: /* Variable declaration - allocate a register for it */ - assert(n->Store); - assert(n->Store->File != PROGRAM_UNDEFINED); - assert(n->Store->Size > 0); - /*assert(n->Store->Index < 0);*/ - if (!n->Var || n->Var->isTemp) { - /* a nameless/temporary variable, will be freed after first use */ - /*NEW*/ - if (n->Store->Index < 0 && !_slang_alloc_temp(emitInfo->vt, n->Store)) { - slang_info_log_error(emitInfo->log, - "Ran out of registers, too many temporaries"); - return NULL; - } - } - else { - /* a regular variable */ - _slang_add_variable(emitInfo->vt, n->Var); - if (!_slang_alloc_var(emitInfo->vt, n->Store)) { - slang_info_log_error(emitInfo->log, - "Ran out of registers, too many variables"); - return NULL; - } - /* - printf("IR_VAR_DECL %s %d store %p\n", - (char*) n->Var->a_name, n->Store->Index, (void*) n->Store); - */ - assert(n->Var->aux == n->Store); - } - if (emitInfo->EmitComments) { - /* emit NOP with comment describing the variable's storage location */ - char s[1000]; - sprintf(s, "TEMP[%d]%s = variable %s (size %d)", - n->Store->Index, - _mesa_swizzle_string(n->Store->Swizzle, 0, GL_FALSE), - (n->Var ? (char *) n->Var->a_name : "anonymous"), - n->Store->Size); - inst = emit_comment(emitInfo, s); - return inst; - } - return NULL; + inst = emit_var_decl(emitInfo, n); + return inst; case IR_VAR: /* Reference to a variable * Storage should have already been resolved/allocated. */ - assert(n->Store); - assert(n->Store->File != PROGRAM_UNDEFINED); - - if (n->Store->File == PROGRAM_STATE_VAR && - n->Store->Index < 0) { - n->Store->Index = _slang_alloc_statevar(n, emitInfo->prog->Parameters); - } - - if (n->Store->Index < 0) { - /* probably ran out of registers */ - return NULL; - } - assert(n->Store->Size > 0); - break; + return emit_var_ref(emitInfo, n); case IR_ELEMENT: return emit_array_element(emitInfo, n); -- cgit v1.2.3 From 3e1706f720b211f27a4aa88af2141db0c793adbb Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 10:27:06 -0600 Subject: mesa: glsl: add usage tracking for uniform vars Unused uniforms are no longer included in the active uniforms list. --- src/mesa/shader/prog_parameter.c | 27 +++++++++++++++++++++++++-- src/mesa/shader/prog_parameter.h | 5 +++++ src/mesa/shader/slang/slang_emit.c | 4 ++++ src/mesa/shader/slang/slang_link.c | 2 +- 4 files changed, 35 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/prog_parameter.c b/src/mesa/shader/prog_parameter.c index e0d2096b30..2dfd923a0f 100644 --- a/src/mesa/shader/prog_parameter.c +++ b/src/mesa/shader/prog_parameter.c @@ -277,6 +277,26 @@ _mesa_add_uniform(struct gl_program_parameter_list *paramList, } +/** + * Mark the named uniform as 'used'. + */ +void +_mesa_use_uniform(struct gl_program_parameter_list *paramList, + const char *name) +{ + GLuint i; + for (i = 0; i < paramList->NumParameters; i++) { + struct gl_program_parameter *p = paramList->Parameters + i; + if (p->Type == PROGRAM_UNIFORM && _mesa_strcmp(p->Name, name) == 0) { + p->Used = GL_TRUE; + /* Note that large uniforms may occupy several slots so we're + * not done searching yet. + */ + } + } +} + + /** * Add a sampler to the parameter list. * \param name uniform's name @@ -591,21 +611,24 @@ _mesa_clone_parameter_list(const struct gl_program_parameter_list *list) /** Not too efficient, but correct */ for (i = 0; i < list->NumParameters; i++) { struct gl_program_parameter *p = list->Parameters + i; + struct gl_program_parameter *pCopy; GLuint size = MIN2(p->Size, 4); GLint j = _mesa_add_parameter(clone, p->Type, p->Name, size, p->DataType, list->ParameterValues[i], NULL); ASSERT(j >= 0); + pCopy = clone->Parameters + j; + pCopy->Used = p->Used; /* copy state indexes */ if (p->Type == PROGRAM_STATE_VAR) { GLint k; - struct gl_program_parameter *q = clone->Parameters + j; for (k = 0; k < STATE_LENGTH; k++) { - q->StateIndexes[k] = p->StateIndexes[k]; + pCopy->StateIndexes[k] = p->StateIndexes[k]; } } else { clone->Parameters[j].Size = p->Size; } + } clone->StateFlags = list->StateFlags; diff --git a/src/mesa/shader/prog_parameter.h b/src/mesa/shader/prog_parameter.h index ac5c629fab..e8d3e9ec33 100644 --- a/src/mesa/shader/prog_parameter.h +++ b/src/mesa/shader/prog_parameter.h @@ -48,6 +48,7 @@ struct gl_program_parameter enum register_file Type; /**< PROGRAM_NAMED_PARAM, CONSTANT or STATE_VAR */ GLenum DataType; /**< GL_FLOAT, GL_FLOAT_VEC2, etc */ GLuint Size; /**< Number of components (1..4) */ + GLboolean Used; /**< Helper flag for GLSL uniform tracking */ /** * A sequence of STATE_* tokens and integers to identify GL state. */ @@ -112,6 +113,10 @@ extern GLint _mesa_add_uniform(struct gl_program_parameter_list *paramList, const char *name, GLuint size, GLenum datatype); +extern void +_mesa_use_uniform(struct gl_program_parameter_list *paramList, + const char *name); + extern GLint _mesa_add_sampler(struct gl_program_parameter_list *paramList, const char *name, GLenum datatype); diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index cad983443a..c5a0fcb6c7 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1703,6 +1703,10 @@ emit_var_ref(slang_emit_info *emitInfo, slang_ir_node *n) if (n->Store->File == PROGRAM_STATE_VAR && n->Store->Index < 0) { n->Store->Index = _slang_alloc_statevar(n, emitInfo->prog->Parameters); } + else if (n->Store->File == PROGRAM_UNIFORM) { + /* mark var as used */ + _mesa_use_uniform(emitInfo->prog->Parameters, (char *) n->Var->a_name); + } if (n->Store->Index < 0) { /* probably ran out of registers */ diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c index e33272eff2..57dbfc2388 100644 --- a/src/mesa/shader/slang/slang_link.c +++ b/src/mesa/shader/slang/slang_link.c @@ -173,7 +173,7 @@ link_uniform_vars(struct gl_shader_program *shProg, * Furthermore, we'll need to fix the state-var's size/datatype info. */ - if (p->Type == PROGRAM_UNIFORM || + if ((p->Type == PROGRAM_UNIFORM && p->Used) || p->Type == PROGRAM_SAMPLER) { _mesa_append_uniform(shProg->Uniforms, p->Name, prog->Target, i); } -- cgit v1.2.3 From c4cf2513e60a90a5b3a6238684a6b51e1f501864 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 11:23:25 -0600 Subject: mesa: glsl: limit function matching through casting --- src/mesa/shader/slang/slang_codegen.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index f96c655759..e5a2c61d4f 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1980,6 +1980,12 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name, fun = _slang_locate_struct_constructor(A, name); } + /* + * At this point, some heuristics are used to try to find a function + * that matches the calling signature by means of casting or "unrolling" + * of constructors. + */ + if (!fun && _slang_is_vec_mat_type(name)) { /* Next, if this call looks like a vec() or mat() constructor call, * try "unwinding" the args to satisfy a constructor. @@ -1995,7 +2001,7 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name, } } - if (!fun) { + if (!fun && _slang_is_vec_mat_type(name)) { /* Next, try casting args to the types of the formal parameters */ int numArgs = oper->num_children; fun = _slang_find_function_by_argc(A->space.funcs, name, numArgs); @@ -2008,6 +2014,13 @@ _slang_gen_function_call_name(slang_assemble_ctx *A, const char *name, assert(fun); } + if (!fun) { + slang_info_log_error(A->log, + "Function '%s' not found (check argument types)", + name); + return NULL; + } + n = _slang_gen_function_call(A, fun, oper, dest); if (n && !n->Store && !dest -- cgit v1.2.3 From bec1519a3202f94c5642dd94ce89cc2ff767aa3b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 11:31:10 -0600 Subject: mesa: glsl: added some post incr/decr error checks --- src/mesa/shader/slang/slang_codegen.c | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index e5a2c61d4f..6cadd2c250 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -2729,7 +2729,11 @@ _slang_assignment_compatible(slang_assemble_ctx *A, slang_typeinfo t0, t1; GLuint sz0, sz1; - + if (op0->type == SLANG_OPER_POSTINCREMENT || + op0->type == SLANG_OPER_POSTDECREMENT) { + return GL_FALSE; + } + slang_typeinfo_construct(&t0); _slang_typeof_operation(A, op0, &t0); -- cgit v1.2.3 From ff01a12b23cad44e4c93e4a2c8cdaa5d18166252 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 12:04:35 -0600 Subject: mesa: glsl: more writemask error checking --- src/mesa/shader/slang/slang_codegen.c | 22 +++++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_codegen.c b/src/mesa/shader/slang/slang_codegen.c index 6cadd2c250..7006e86958 100644 --- a/src/mesa/shader/slang/slang_codegen.c +++ b/src/mesa/shader/slang/slang_codegen.c @@ -1481,7 +1481,7 @@ _slang_simple_writemask(GLuint writemask, GLuint swizzle) * \return GL_FALSE for simple writemasks, GL_TRUE for non-simple */ static GLboolean -swizzle_to_writemask(GLuint swizzle, +swizzle_to_writemask(slang_assemble_ctx *A, GLuint swizzle, GLuint *writemaskOut, GLuint *swizzleOut) { GLuint mask = 0x0, newSwizzle[4]; @@ -1495,6 +1495,18 @@ swizzle_to_writemask(GLuint swizzle, break; } assert(swz >= 0 && swz <= 3); + + if (swizzle != SWIZZLE_XXXX && + swizzle != SWIZZLE_YYYY && + swizzle != SWIZZLE_ZZZZ && + swizzle != SWIZZLE_WWWW && + (mask & (1 << swz))) { + /* a channel can't be specified twice (ex: ".xyyz") */ + slang_info_log_error(A->log, "Invalid writemask '%s'", + _mesa_swizzle_string(swizzle, 0, 0)); + return GL_FALSE; + } + mask |= (1 << swz); } assert(mask <= 0xf); @@ -1590,11 +1602,11 @@ resolve_swizzle(const slang_operation *oper) * As above, but produce a writemask. */ static GLuint -resolve_writemask(const slang_operation *oper) +resolve_writemask(slang_assemble_ctx *A, const slang_operation *oper) { GLuint swizzle = resolve_swizzle(oper); GLuint writemask, swizzleOut; - swizzle_to_writemask(swizzle, &writemask, &swizzleOut); + swizzle_to_writemask(A, swizzle, &writemask, &swizzleOut); return writemask; } @@ -1668,7 +1680,7 @@ _slang_gen_asm(slang_assemble_ctx *A, slang_operation *oper, dest_oper = &oper->children[0]; - writemask = resolve_writemask(dest_oper); + writemask = resolve_writemask(A, dest_oper); n0 = _slang_gen_operation(A, dest_oper); if (!n0) @@ -3037,7 +3049,7 @@ _slang_gen_assignment(slang_assemble_ctx * A, slang_operation *oper) if (lhs && rhs) { /* convert lhs swizzle into writemask */ GLuint writemask, newSwizzle; - if (!swizzle_to_writemask(lhs->Store->Swizzle, + if (!swizzle_to_writemask(A, lhs->Store->Swizzle, &writemask, &newSwizzle)) { /* Non-simple writemask, need to swizzle right hand side in * order to put components into the right place. -- cgit v1.2.3 From 200d6dcc83ea9ac0bf6f1506214e0bd9b65714f2 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Aug 2008 17:08:41 +0100 Subject: trace: Support C++. --- src/gallium/drivers/trace/tr_context.h | 9 +++++++++ src/gallium/drivers/trace/tr_screen.h | 9 +++++++++ 2 files changed, 18 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.h b/src/gallium/drivers/trace/tr_context.h index 679371e310..7831900ec2 100644 --- a/src/gallium/drivers/trace/tr_context.h +++ b/src/gallium/drivers/trace/tr_context.h @@ -34,6 +34,11 @@ #include "pipe/p_context.h" +#ifdef __cplusplus +extern "C" { +#endif + + struct trace_context { struct pipe_context base; @@ -56,4 +61,8 @@ trace_context_create(struct pipe_screen *screen, struct pipe_context *pipe); +#ifdef __cplusplus +} +#endif + #endif /* TR_CONTEXT_H_ */ diff --git a/src/gallium/drivers/trace/tr_screen.h b/src/gallium/drivers/trace/tr_screen.h index 698b84811d..93fefdb9a5 100644 --- a/src/gallium/drivers/trace/tr_screen.h +++ b/src/gallium/drivers/trace/tr_screen.h @@ -32,6 +32,11 @@ #include "pipe/p_screen.h" +#ifdef __cplusplus +extern "C" { +#endif + + struct trace_screen { struct pipe_screen base; @@ -48,4 +53,8 @@ struct pipe_screen * trace_screen_create(struct pipe_screen *screen); +#ifdef __cplusplus +} +#endif + #endif /* TR_SCREEN_H_ */ -- cgit v1.2.3 From 5f1ef11ad2e3016ac4029489d9429d7f93737fe5 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Aug 2008 18:25:38 +0100 Subject: python/retrace: Trim null chars. They are often left in memory mapped files, and are not part of the XML accept chars anyway. --- src/gallium/state_trackers/python/retrace/parser.py | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/parser.py b/src/gallium/state_trackers/python/retrace/parser.py index 9ee47f6e19..d02fd7f741 100755 --- a/src/gallium/state_trackers/python/retrace/parser.py +++ b/src/gallium/state_trackers/python/retrace/parser.py @@ -101,6 +101,7 @@ class XmlTokenizer: self.index = 0 data = self.fp.read(size) self.final = len(data) < size + data = data.rstrip('\0') try: self.parser.Parse(data, self.final) except xml.parsers.expat.ExpatError, e: -- cgit v1.2.3 From 9adfc57a4c4deeb86d8f62491b94d6a44586eb1e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Aug 2008 19:11:36 +0100 Subject: python: Handle null state. --- src/gallium/state_trackers/python/p_context.i | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i index 535783b641..496f738ad5 100644 --- a/src/gallium/state_trackers/python/p_context.i +++ b/src/gallium/state_trackers/python/p_context.i @@ -68,6 +68,11 @@ struct st_context { void set_fragment_shader( const struct pipe_shader_state *state ) { void *fs; + if(!state) { + cso_set_fragment_shader_handle($self->cso, NULL); + return; + } + fs = $self->pipe->create_fs_state($self->pipe, state); if(!fs) return; @@ -82,6 +87,11 @@ struct st_context { void set_vertex_shader( const struct pipe_shader_state *state ) { void *vs; + if(!state) { + cso_set_vertex_shader_handle($self->cso, NULL); + return; + } + vs = $self->pipe->create_vs_state($self->pipe, state); if(!vs) return; -- cgit v1.2.3 From d27ffb8c6d4551371995608a6d752dfeb26c2351 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Aug 2008 19:11:47 +0100 Subject: trace: Fix pipe_clip_state dump. --- src/gallium/drivers/trace/tr_state.c | 23 ++++++----------------- 1 file changed, 6 insertions(+), 17 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index 7b35e89e75..30ab5a8fdc 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -51,7 +51,6 @@ void trace_dump_block(const struct pipe_format_block *block) void trace_dump_template(const struct pipe_texture *templat) { - assert(templat); if(!templat) { trace_dump_null(); return; @@ -87,7 +86,6 @@ void trace_dump_template(const struct pipe_texture *templat) void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -136,7 +134,6 @@ void trace_dump_rasterizer_state(const struct pipe_rasterizer_state *state) void trace_dump_poly_stipple(const struct pipe_poly_stipple *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -156,7 +153,6 @@ void trace_dump_poly_stipple(const struct pipe_poly_stipple *state) void trace_dump_viewport_state(const struct pipe_viewport_state *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -173,7 +169,6 @@ void trace_dump_viewport_state(const struct pipe_viewport_state *state) void trace_dump_scissor_state(const struct pipe_scissor_state *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -194,18 +189,20 @@ void trace_dump_clip_state(const struct pipe_clip_state *state) { unsigned i; - assert(state); if(!state) { trace_dump_null(); return; } - trace_dump_struct_begin("pipe_scissor_state"); + trace_dump_struct_begin("pipe_clip_state"); trace_dump_member_begin("ucp"); trace_dump_array_begin(); - for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) + for(i = 0; i < PIPE_MAX_CLIP_PLANES; ++i) { + trace_dump_elem_begin(); trace_dump_array(float, state->ucp[i], 4); + trace_dump_elem_end(); + } trace_dump_array_end(); trace_dump_member_end(); @@ -217,7 +214,6 @@ void trace_dump_clip_state(const struct pipe_clip_state *state) void trace_dump_constant_buffer(const struct pipe_constant_buffer *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -235,7 +231,7 @@ void trace_dump_constant_buffer(const struct pipe_constant_buffer *state) void trace_dump_shader_state(const struct pipe_shader_state *state) { static char str[8192]; - assert(state); + if(!state) { trace_dump_null(); return; @@ -257,7 +253,6 @@ void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_ { unsigned i; - assert(state); if(!state) { trace_dump_null(); return; @@ -307,7 +302,6 @@ void trace_dump_depth_stencil_alpha_state(const struct pipe_depth_stencil_alpha_ void trace_dump_blend_state(const struct pipe_blend_state *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -337,7 +331,6 @@ void trace_dump_blend_state(const struct pipe_blend_state *state) void trace_dump_blend_color(const struct pipe_blend_color *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -367,7 +360,6 @@ void trace_dump_framebuffer_state(const struct pipe_framebuffer_state *state) void trace_dump_sampler_state(const struct pipe_sampler_state *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -398,7 +390,6 @@ void trace_dump_sampler_state(const struct pipe_sampler_state *state) void trace_dump_surface(const struct pipe_surface *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -436,7 +427,6 @@ void trace_dump_surface(const struct pipe_surface *state) void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state) { - assert(state); if(!state) { trace_dump_null(); return; @@ -455,7 +445,6 @@ void trace_dump_vertex_buffer(const struct pipe_vertex_buffer *state) void trace_dump_vertex_element(const struct pipe_vertex_element *state) { - assert(state); if(!state) { trace_dump_null(); return; -- cgit v1.2.3 From d7f8b95e097121a8c33332f920115795853d6ad7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Aug 2008 19:12:22 +0100 Subject: python/retrace: Add several missing functions. --- .../state_trackers/python/retrace/interpreter.py | 57 ++++++++++++++++++---- 1 file changed, 48 insertions(+), 9 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index fa9a65365c..a7a7341e06 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -79,7 +79,7 @@ class Struct: struct_factories = { "pipe_blend_color": gallium.BlendColor, "pipe_blend_state": gallium.Blend, - "pipe_clip_state": gallium.Clip, + #"pipe_clip_state": gallium.Clip, #"pipe_constant_buffer": gallium.ConstantBuffer, "pipe_depth_state": gallium.Depth, "pipe_stencil_state": gallium.Stencil, @@ -103,7 +103,7 @@ member_array_factories = { "pipe_rasterizer_state": {"sprite_coord_mode": gallium.ByteArray}, "pipe_poly_stipple": {"stipple": gallium.UnsignedArray}, "pipe_viewport_state": {"scale": gallium.FloatArray, "translate": gallium.FloatArray}, - "pipe_clip_state": {"ucp": gallium.FloatArray}, + #"pipe_clip_state": {"ucp": gallium.FloatArray}, "pipe_depth_stencil_alpha_state": {"stencil": gallium.StencilArray}, "pipe_blend_color": {"color": gallium.FloatArray}, "pipe_sampler_state": {"border_color": gallium.FloatArray}, @@ -219,6 +219,12 @@ class Winsys(Object): def flush_frontbuffer(self, surface): pass + def surface_alloc(self): + return None + + def surface_release(self, surface): + pass + class Screen(Object): @@ -269,15 +275,25 @@ class Context(Object): self.cbufs = [] self.zsbuf = None + def destroy(self): + pass + def create_blend_state(self, state): return state def bind_blend_state(self, state): - self.real.set_blend(state) - + if state is not None: + self.real.set_blend(state) + + def delete_blend_state(self, state): + pass + def create_sampler_state(self, state): return state + def delete_sampler_state(self, state): + pass + def bind_sampler_states(self, n, states): for i in range(n): self.real.set_sampler(i, states[i]) @@ -286,14 +302,22 @@ class Context(Object): return state def bind_rasterizer_state(self, state): - self.real.set_rasterizer(state) + if state is not None: + self.real.set_rasterizer(state) + def delete_rasterizer_state(self, state): + pass + def create_depth_stencil_alpha_state(self, state): return state def bind_depth_stencil_alpha_state(self, state): - self.real.set_depth_stencil_alpha(state) - + if state is not None: + self.real.set_depth_stencil_alpha(state) + + def delete_depth_stencil_alpha_state(self, state): + pass + def create_fs_state(self, state): tokens = str(state.tokens) shader = gallium.Shader(tokens) @@ -307,14 +331,29 @@ class Context(Object): def bind_vs_state(self, state): self.real.set_vertex_shader(state) + def delete_fs_state(self, state): + pass + + delete_vs_state = delete_fs_state + def set_blend_color(self, state): self.real.set_blend_color(state) def set_clip_state(self, state): - self.real.set_clip(state) + _state = gallium.Clip() + _state.nr = state.nr + if state.nr: + # FIXME + ucp = gallium.FloatArray(gallium.PIPE_MAX_CLIP_PLANES*4) + for i in range(len(state.ucp)): + for j in range(len(state.ucp[i])): + ucp[i*4 + j] = state.ucp[i][j] + _state.ucp = ucp + self.real.set_clip(_state) def set_constant_buffer(self, shader, index, state): - self.real.set_constant_buffer(shader, index, state.buffer) + if state is not None: + self.real.set_constant_buffer(shader, index, state.buffer) def set_framebuffer_state(self, state): _state = gallium.Framebuffer() -- cgit v1.2.3 From 9e3aaa50483b61d644306d483feed564c69018bf Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Tue, 19 Aug 2008 21:33:59 +0100 Subject: gallium: WinCE build fixes. --- src/gallium/auxiliary/util/u_snprintf.c | 9 +++++++++ src/gallium/include/pipe/p_util.h | 6 ++++++ 2 files changed, 15 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_snprintf.c b/src/gallium/auxiliary/util/u_snprintf.c index 7fa84d8bec..0d54299b28 100644 --- a/src/gallium/auxiliary/util/u_snprintf.c +++ b/src/gallium/auxiliary/util/u_snprintf.c @@ -334,6 +334,15 @@ static void *mymemcpy(void *, void *, size_t); #endif /* HAVE_UINTPTR_T || defined(uintptr_t) */ #endif /* !defined(UINTPTR_T) */ +/* WinCE5.0 does not have uintptr_t defined */ +#if (_WIN32_WCE < 600) +#ifdef UINTPTR_T +#undef UINTPTR_T +#endif +#define UINTPTR_T unsigned long int +#endif + + /* Support for ptrdiff_t. */ #ifndef PTRDIFF_T #if HAVE_PTRDIFF_T || defined(ptrdiff_t) diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index b4ab70a3fc..473a8d94ab 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -99,6 +99,12 @@ __inline double __cdecl atan2(double val) #include #endif + /* Define ENOMEM for WINCE */ +#if (_WIN32_WCE < 600) +#ifndef ENOMEM +#define ENOMEM 12 +#endif +#endif #ifdef __cplusplus extern "C" { -- cgit v1.2.3 From e90c0c976d89d108a8220b538e02be173cf308fe Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 14:24:27 -0600 Subject: gallium: do a proper implementation of GL_OES_read_format Examine the currently bound color buffer's format to see if there's a good format/type match. --- src/mesa/SConscript | 1 + src/mesa/sources.mak | 1 + src/mesa/state_tracker/st_cb_get.c | 97 +++++++++++++++++++++++++++++++++++++ src/mesa/state_tracker/st_cb_get.h | 37 ++++++++++++++ src/mesa/state_tracker/st_context.c | 2 + 5 files changed, 138 insertions(+) create mode 100644 src/mesa/state_tracker/st_cb_get.c create mode 100644 src/mesa/state_tracker/st_cb_get.h (limited to 'src') diff --git a/src/mesa/SConscript b/src/mesa/SConscript index 2c74dc7dd0..af8dfcb493 100644 --- a/src/mesa/SConscript +++ b/src/mesa/SConscript @@ -153,6 +153,7 @@ if env['platform'] != 'winddk': 'state_tracker/st_cb_flush.c', 'state_tracker/st_cb_drawpixels.c', 'state_tracker/st_cb_fbo.c', + 'state_tracker/st_cb_get.c', 'state_tracker/st_cb_feedback.c', 'state_tracker/st_cb_program.c', 'state_tracker/st_cb_queryobj.c', diff --git a/src/mesa/sources.mak b/src/mesa/sources.mak index 0e0e10979b..14fd8597a8 100644 --- a/src/mesa/sources.mak +++ b/src/mesa/sources.mak @@ -187,6 +187,7 @@ STATETRACKER_SOURCES = \ state_tracker/st_cb_bufferobjects.c \ state_tracker/st_cb_clear.c \ state_tracker/st_cb_flush.c \ + state_tracker/st_cb_get.c \ state_tracker/st_cb_drawpixels.c \ state_tracker/st_cb_fbo.c \ state_tracker/st_cb_feedback.c \ diff --git a/src/mesa/state_tracker/st_cb_get.c b/src/mesa/state_tracker/st_cb_get.c new file mode 100644 index 0000000000..e7d7f03bc9 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_get.c @@ -0,0 +1,97 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * glGet functions + * + * \author Brian Paul + */ + +#include "main/imports.h" +#include "main/context.h" + +#include "pipe/p_defines.h" + +#include "st_cb_fbo.h" +#include "st_cb_get.h" + + + +/** + * Examine the current color read buffer format to determine + * which GL pixel format/type combo is the best match. + */ +static void +get_preferred_read_format_type(GLcontext *ctx, GLint *format, GLint *type) +{ + struct gl_framebuffer *fb = ctx->ReadBuffer; + struct st_renderbuffer *strb = st_renderbuffer(fb->_ColorReadBuffer); + + /* defaults */ + *format = ctx->Const.ColorReadFormat; + *type = ctx->Const.ColorReadType; + + if (strb) { + /* XXX could add more cases here... */ + if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM) { + *format = GL_BGRA; + if (_mesa_little_endian()) + *type = GL_UNSIGNED_INT_8_8_8_8_REV; + else + *type = GL_UNSIGNED_INT_8_8_8_8; + } + } +} + + +/** + * We only intercept the OES preferred ReadPixels format/type. + * Everything else goes to the default _mesa_GetIntegerv. + */ +static GLboolean +st_GetIntegerv(GLcontext *ctx, GLenum pname, GLint *params) +{ + GLint dummy; + + switch (pname) { + case GL_IMPLEMENTATION_COLOR_READ_TYPE_OES: + get_preferred_read_format_type(ctx, &dummy, params); + return GL_TRUE; + case GL_IMPLEMENTATION_COLOR_READ_FORMAT_OES: + get_preferred_read_format_type(ctx, params, &dummy); + return GL_TRUE; + default: + return GL_FALSE; + } +} + + +void st_init_get_functions(struct dd_function_table *functions) +{ + functions->GetIntegerv = st_GetIntegerv; +} diff --git a/src/mesa/state_tracker/st_cb_get.h b/src/mesa/state_tracker/st_cb_get.h new file mode 100644 index 0000000000..8e9f3e9306 --- /dev/null +++ b/src/mesa/state_tracker/st_cb_get.h @@ -0,0 +1,37 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +#ifndef ST_CB_GET_H +#define ST_CB_GET_H + + +extern void +st_init_get_functions(struct dd_function_table *functions); + + +#endif diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c index 3b126e6c8f..83b0be06da 100644 --- a/src/mesa/state_tracker/st_context.c +++ b/src/mesa/state_tracker/st_context.c @@ -49,6 +49,7 @@ #include "st_cb_drawtex.h" #endif #include "st_cb_fbo.h" +#include "st_cb_get.h" #if FEATURE_feedback #include "st_cb_feedback.h" #endif @@ -293,6 +294,7 @@ void st_init_driver_functions(struct dd_function_table *functions) st_init_rasterpos_functions(functions); #endif st_init_fbo_functions(functions); + st_init_get_functions(functions); #if FEATURE_feedback st_init_feedback_functions(functions); #endif -- cgit v1.2.3 From 84a496746e1b01f4b11c4a44dd698b904ec83572 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 14:34:59 -0600 Subject: gallium: implement a fast-path for glReadPixels for common cases --- src/mesa/state_tracker/st_cb_readpixels.c | 126 ++++++++++++++++++++++++++++++ 1 file changed, 126 insertions(+) (limited to 'src') diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index eb71779cc9..39f5856f94 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -162,6 +162,124 @@ st_get_color_read_renderbuffer(GLcontext *ctx) } +/** + * Try to do glReadPixels in a fast manner for common cases. + * \return GL_TRUE for success, GL_FALSE for failure + */ +static GLboolean +st_fast_readpixels(GLcontext *ctx, struct st_renderbuffer *strb, + GLint x, GLint y, GLsizei width, GLsizei height, + GLenum format, GLenum type, + const struct gl_pixelstore_attrib *pack, + GLvoid *dest) +{ + enum combination { + A8R8G8B8_UNORM_TO_RGBA_UBYTE, + A8R8G8B8_UNORM_TO_RGB_UBYTE, + A8R8G8B8_UNORM_TO_BGRA_UINT + } combo; + + if (ctx->_ImageTransferState) + return GL_FALSE; + + if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM && + format == GL_RGBA && type == GL_UNSIGNED_BYTE) { + combo = A8R8G8B8_UNORM_TO_RGBA_UBYTE; + } + else if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM && + format == GL_RGB && type == GL_UNSIGNED_BYTE) { + combo = A8R8G8B8_UNORM_TO_RGB_UBYTE; + } + else if (strb->format == PIPE_FORMAT_A8R8G8B8_UNORM && + format == GL_BGRA && type == GL_UNSIGNED_INT_8_8_8_8_REV) { + combo = A8R8G8B8_UNORM_TO_BGRA_UINT; + } + else { + return GL_FALSE; + } + + /*printf("st_fast_readpixels combo %d\n", (GLint) combo);*/ + + { + struct pipe_context *pipe = ctx->st->pipe; + struct pipe_screen *screen = pipe->screen; + struct pipe_surface *surf; + const GLubyte *map; + GLubyte *dst; + GLint row, col, dy, dstStride; + + surf = screen->get_tex_surface(screen, strb->texture, 0, 0, 0, + PIPE_BUFFER_USAGE_CPU_READ); + if (!surf) { + return GL_FALSE; + } + + map = screen->surface_map(screen, surf, PIPE_BUFFER_USAGE_CPU_READ); + if (!map) { + pipe_surface_reference(&surf, NULL); + return GL_FALSE; + } + + if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) { + y = surf->height - y - 1; + dy = -1; + } + else { + dy = 1; + } + + dst = _mesa_image_address2d(pack, dest, width, height, + format, type, 0, 0); + dstStride = _mesa_image_row_stride(pack, width, format, type); + + switch (combo) { + case A8R8G8B8_UNORM_TO_RGBA_UBYTE: + for (row = 0; row < height; row++) { + const GLubyte *src = map + y * surf->stride + x * 4; + for (col = 0; col < width; col++) { + GLuint pixel = ((GLuint *) src)[col]; + dst[col*4+0] = (pixel >> 16) & 0xff; + dst[col*4+1] = (pixel >> 8) & 0xff; + dst[col*4+2] = (pixel >> 0) & 0xff; + dst[col*4+3] = (pixel >> 24) & 0xff; + } + dst += dstStride; + y += dy; + } + break; + case A8R8G8B8_UNORM_TO_RGB_UBYTE: + for (row = 0; row < height; row++) { + const GLubyte *src = map + y * surf->stride + x * 4; + for (col = 0; col < width; col++) { + GLuint pixel = ((GLuint *) src)[col]; + dst[col*3+0] = (pixel >> 16) & 0xff; + dst[col*3+1] = (pixel >> 8) & 0xff; + dst[col*3+2] = (pixel >> 0) & 0xff; + } + dst += dstStride; + y += dy; + } + break; + case A8R8G8B8_UNORM_TO_BGRA_UINT: + for (row = 0; row < height; row++) { + const GLubyte *src = map + y * surf->stride + x * 4; + memcpy(dst, src, 4 * width); + dst += dstStride; + y += dy; + } + break; + default: + ; /* nothing */ + } + + screen->surface_unmap(screen, surf); + pipe_surface_reference(&surf, NULL); + } + + return GL_TRUE; +} + + /** * Do glReadPixels by getting rows from the framebuffer surface with * get_tile(). Convert to requested format/type with Mesa image routines. @@ -219,6 +337,14 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height, if (!strb) return; + /* try a fast-path readpixels before anything else */ + if (st_fast_readpixels(ctx, strb, x, y, width, height, + format, type, pack, dest)) { + /* success! */ + _mesa_unmap_readpix_pbo(ctx, &clippedPacking); + return; + } + if (format == GL_RGBA && type == GL_FLOAT) { /* write tile(row) directly into user's buffer */ df = (GLfloat *) _mesa_image_address2d(&clippedPacking, dest, width, -- cgit v1.2.3 From 63c0970dca9c13ab83ea24f108b41f75f2a290a7 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 17:04:48 -0600 Subject: gallium: fix an assertion --- src/gallium/auxiliary/draw/draw_vs_varient.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index ad0b829afa..994ce3e889 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -298,7 +298,7 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float); emit.element[i].output_format = draw_translate_vinfo_format(key->element[i].out.format); emit.element[i].output_offset = key->element[i].out.offset; - assert(emit.element[i].input_offset < fetch.output_stride); + assert(emit.element[i].input_offset <= fetch.output_stride); } else { emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT; -- cgit v1.2.3 From 7f9959ae8394f8e52a180f5e261b0f9470f6c5bc Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 17:41:34 -0600 Subject: gallium: fix do_flip bug in sp_surface_copy() Surfaces are always in y=0=top raster order so the caller should invert the Y coordinate if needed; don't do it in sp_surface_copy(). Fixes a glCopyTexture regression. --- src/gallium/drivers/softpipe/sp_surface.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index 7dc15c38d1..4f1bb881cb 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -39,6 +39,9 @@ * Copy a rectangular region from one surface to another. * Surfaces must have same bpp. * + * Note that it's always the case that Y=0=top of the raster. + * If do_flip is non-zero, the region being copied will be flipped vertically. + * * Assumes all values are within bounds -- no checking at this level - * do it higher up if required. */ @@ -72,7 +75,7 @@ sp_surface_copy(struct pipe_context *pipe, width, height, src_map, do_flip ? -(int) src->stride : src->stride, - srcx, do_flip ? src->height - 1 - srcy : srcy); + srcx, srcy); pipe->screen->surface_unmap(pipe->screen, src); pipe->screen->surface_unmap(pipe->screen, dst); -- cgit v1.2.3 From 815cdcfbc0740c66b901361620c88d99541bdad2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Tue, 19 Aug 2008 18:14:15 -0600 Subject: mesa: allow for extra per-context init --- src/mesa/main/context.c | 4 ++++ src/mesa/main/context.h | 3 +++ src/mesa/main/points.c | 4 ---- 3 files changed, 7 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c index 32460e92c3..ed3faecf0d 100644 --- a/src/mesa/main/context.c +++ b/src/mesa/main/context.c @@ -1214,6 +1214,10 @@ _mesa_initialize_context(GLcontext *ctx, ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE; } +#ifdef FEATURE_extra_context_init + _mesa_initialize_context_extra(ctx); +#endif + ctx->FirstTimeCurrent = GL_TRUE; return GL_TRUE; diff --git a/src/mesa/main/context.h b/src/mesa/main/context.h index 099912aa15..9423b66a7d 100644 --- a/src/mesa/main/context.h +++ b/src/mesa/main/context.h @@ -114,6 +114,9 @@ _mesa_initialize_context( GLcontext *ctx, const struct dd_function_table *driverFunctions, void *driverContext ); +extern void +_mesa_initialize_context_extra(GLcontext *ctx); + extern void _mesa_free_context_data( GLcontext *ctx ); diff --git a/src/mesa/main/points.c b/src/mesa/main/points.c index d16344a42c..fbedbcb22c 100644 --- a/src/mesa/main/points.c +++ b/src/mesa/main/points.c @@ -244,11 +244,7 @@ _mesa_init_point(GLcontext *ctx) ctx->Point.MaxSize = MAX2(ctx->Const.MaxPointSize, ctx->Const.MaxPointSizeAA); ctx->Point.Threshold = 1.0; -#if FEATURE_es2_glsl - ctx->Point.PointSprite = GL_TRUE; /* GL_ARB/NV_point_sprite */ -#else ctx->Point.PointSprite = GL_FALSE; /* GL_ARB/NV_point_sprite */ -#endif ctx->Point.SpriteRMode = GL_ZERO; /* GL_NV_point_sprite (only!) */ ctx->Point.SpriteOrigin = GL_UPPER_LEFT; /* GL_ARB_point_sprite */ for (i = 0; i < MAX_TEXTURE_UNITS; i++) { -- cgit v1.2.3 From 1d881b8e8472ff49482b2b0404ac29160f096771 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 20 Aug 2008 14:02:43 +0100 Subject: python: bindings for draw_range_elements. --- src/gallium/state_trackers/python/p_context.i | 10 ++++++++++ 1 file changed, 10 insertions(+) (limited to 'src') diff --git a/src/gallium/state_trackers/python/p_context.i b/src/gallium/state_trackers/python/p_context.i index 496f738ad5..0b2621f7c3 100644 --- a/src/gallium/state_trackers/python/p_context.i +++ b/src/gallium/state_trackers/python/p_context.i @@ -209,6 +209,16 @@ struct st_context { mode, start, count); } + void draw_range_elements( struct st_buffer *indexBuffer, + unsigned indexSize, unsigned minIndex, unsigned maxIndex, + unsigned mode, unsigned start, unsigned count) + { + $self->pipe->draw_range_elements($self->pipe, + indexBuffer->buffer, + indexSize, minIndex, maxIndex, + mode, start, count); + } + void draw_vertices(unsigned prim, unsigned num_verts, unsigned num_attribs, -- cgit v1.2.3 From 34cffce33413fe7701975d0d4bd54207bc44cacc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Wed, 20 Aug 2008 14:03:12 +0100 Subject: python/retrace: Retrace draw_elements and draw_range_elements. --- src/gallium/state_trackers/python/retrace/interpreter.py | 6 ++++++ 1 file changed, 6 insertions(+) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index a7a7341e06..7502c096b6 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -403,6 +403,12 @@ class Context(Object): def draw_arrays(self, mode, start, count): self.real.draw_arrays(mode, start, count) + + def draw_elements(self, indexBuffer, indexSize, mode, start, count): + self.real.draw_elements(indexBuffer, indexSize, mode, start, count) + + def draw_range_elements(self, indexBuffer, indexSize, minIndex, maxIndex, mode, start, count): + self.real.draw_range_elements(indexBuffer, indexSize, minIndex, maxIndex, mode, start, count) def flush(self, flags): self.real.flush(flags) -- cgit v1.2.3 From f6abdb20437b1b8d27d8c45c0787017dfcad3497 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 20 Aug 2008 10:27:30 -0600 Subject: gallium: fix typo in LINE() macro (replace i+1 with i1 var) We were sometimes referencing an invalid vertex. Fixes progs/trivial/line-clip.c test among others. --- src/gallium/auxiliary/draw/draw_pipe.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 3355c871ee..1db43876ef 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -238,7 +238,7 @@ void draw_pipeline_run( struct draw_context *draw, do_line( draw, \ flags, \ verts + stride * ((i0) & ~DRAW_PIPE_FLAG_MASK), \ - verts + stride * (i+1)) + verts + stride * (i1)) #define POINT(i0) \ do_point( draw, \ -- cgit v1.2.3 From 0fae7648987d8264f85a9b6b6d7f903bff82a0f0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 20 Aug 2008 10:31:38 -0600 Subject: gallium: test for and cull prims with inf/nan vertices in sp_setup.c code. --- src/gallium/drivers/softpipe/sp_setup.c | 45 +++++++++++++++++++++++++-------- 1 file changed, 34 insertions(+), 11 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index b7f2f16307..dd0562e7ef 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -104,6 +104,16 @@ struct setup_context { +/** + * Test if x is NaN or +/- infinity. + */ +static INLINE boolean +is_inf_or_nan(float x) +{ + union fi tmp; + tmp.f = x; + return !(int)((unsigned int)((tmp.i & 0x7fffffff)-0x7f800000) >> 31); +} static boolean cull_tri( struct setup_context *setup, @@ -293,6 +303,9 @@ static void print_vertex(const struct setup_context *setup, } #endif +/** + * \return FALSE if coords are inf/nan (cull the tri), TRUE otherwise + */ static boolean setup_sort_vertices( struct setup_context *setup, float det, const float (*v0)[4], @@ -370,10 +383,13 @@ static boolean setup_sort_vertices( struct setup_context *setup, setup->ebot.dx * setup->emaj.dy); setup->oneoverarea = 1.0f / area; + /* debug_printf("%s one-over-area %f area %f det %f\n", __FUNCTION__, setup->oneoverarea, area, det ); */ + if (is_inf_or_nan(setup->oneoverarea)) + return FALSE; } /* We need to know if this is a front or back-facing triangle for: @@ -742,7 +758,8 @@ void setup_tri( struct setup_context *setup, if (cull_tri( setup, det )) return; - setup_sort_vertices( setup, det, v0, v1, v2 ); + if (!setup_sort_vertices( setup, det, v0, v1, v2 )) + return; setup_tri_coefficients( setup ); setup_tri_edges( setup ); @@ -827,7 +844,7 @@ line_persp_coeff(struct setup_context *setup, * Compute the setup->coef[] array dadx, dady, a0 values. * Must be called after setup->vmin,vmax are initialized. */ -static INLINE void +static INLINE boolean setup_line_coefficients(struct setup_context *setup, const float (*v0)[4], const float (*v1)[4]) @@ -836,6 +853,7 @@ setup_line_coefficients(struct setup_context *setup, const struct sp_fragment_shader *spfs = softpipe->fs; const struct vertex_info *vinfo = softpipe_get_vertex_info(softpipe); uint fragSlot; + float area; /* use setup->vmin, vmax to point to vertices */ setup->vprovoke = v1; @@ -844,9 +862,12 @@ setup_line_coefficients(struct setup_context *setup, setup->emaj.dx = setup->vmax[0][0] - setup->vmin[0][0]; setup->emaj.dy = setup->vmax[0][1] - setup->vmin[0][1]; - /* NOTE: this is not really 1/area */ - setup->oneoverarea = 1.0f / (setup->emaj.dx * setup->emaj.dx + - setup->emaj.dy * setup->emaj.dy); + + /* NOTE: this is not really area but something proportional to it */ + area = setup->emaj.dx * setup->emaj.dx + setup->emaj.dy * setup->emaj.dy; + if (area == 0.0f || is_inf_or_nan(area)) + return FALSE; + setup->oneoverarea = 1.0f / area; /* z and w are done by linear interpolation: */ @@ -886,6 +907,7 @@ setup_line_coefficients(struct setup_context *setup, setup->coef[fragSlot].dady[1] = 0.0; } } + return TRUE; } @@ -942,18 +964,19 @@ setup_line(struct setup_context *setup, print_vertex(setup, v1); #endif - assert(v0[0][0] < 1.0e9); - assert(v0[0][1] < 1.0e9); - assert(v1[0][0] < 1.0e9); - assert(v1[0][1] < 1.0e9); - if (setup->softpipe->no_rast) return; if (dx == 0 && dy == 0) return; - setup_line_coefficients(setup, v0, v1); + if (!setup_line_coefficients(setup, v0, v1)) + return; + + assert(v0[0][0] < 1.0e9); + assert(v0[0][1] < 1.0e9); + assert(v1[0][0] < 1.0e9); + assert(v1[0][1] < 1.0e9); if (dx < 0) { dx = -dx; /* make positive */ -- cgit v1.2.3 From ef33697c7e11d93d8b94eadf286f4096734b613f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 20 Aug 2008 11:06:53 -0600 Subject: mesa: glsl: better implementation of ftransform() --- src/mesa/shader/slang/library/slang_vertex_builtin.gc | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_vertex_builtin.gc b/src/mesa/shader/slang/library/slang_vertex_builtin.gc index 20c924a30d..17e86d9a0e 100644 --- a/src/mesa/shader/slang/library/slang_vertex_builtin.gc +++ b/src/mesa/shader/slang/library/slang_vertex_builtin.gc @@ -57,7 +57,10 @@ varying float gl_FogFragCoord; vec4 ftransform() { - __retVal = gl_Vertex * gl_ModelViewProjectionMatrixTranspose; + __retVal = gl_ModelViewProjectionMatrix[0] * gl_Vertex.xxxx + + gl_ModelViewProjectionMatrix[1] * gl_Vertex.yyyy + + gl_ModelViewProjectionMatrix[2] * gl_Vertex.zzzz + + gl_ModelViewProjectionMatrix[3] * gl_Vertex.wwww; } -- cgit v1.2.3 From 29bee15b1e8bea9162d9b9343c98b22b766694b1 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 20 Aug 2008 11:07:25 -0600 Subject: mesa: glsl: regenerated file --- .../shader/slang/library/slang_vertex_builtin_gc.h | 146 +++++++++++---------- 1 file changed, 76 insertions(+), 70 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h b/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h index 1a08a54d7e..63d33af78e 100644 --- a/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_vertex_builtin_gc.h @@ -20,75 +20,81 @@ 0,2,2,3,0,12,1,103,108,95,84,101,120,67,111,111,114,100,0,3,18,103,108,95,77,97,120,84,101,120,116, 117,114,101,67,111,111,114,100,115,0,0,0,2,2,3,0,9,1,103,108,95,70,111,103,70,114,97,103,67,111, 111,114,100,0,0,0,1,0,0,12,0,102,116,114,97,110,115,102,111,114,109,0,0,1,9,18,95,95,114,101,116, -86,97,108,0,18,103,108,95,86,101,114,116,101,120,0,18,103,108,95,77,111,100,101,108,86,105,101,119, -80,114,111,106,101,99,116,105,111,110,77,97,116,114,105,120,84,114,97,110,115,112,111,115,101,0,48, -20,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,76,111,100,0,1,1,0,16,115,97,109,112,108,101, -114,0,0,1,1,0,9,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100, -52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114, -100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114, -101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12, -0,116,101,120,116,117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0,16,115,97,109,112,108,101, -114,0,0,1,1,0,10,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111, -114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111, -111,114,100,0,59,121,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118, -101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, -114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111, -106,76,111,100,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9, -108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120, -0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111, -114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114, -101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12, -0,116,101,120,116,117,114,101,50,68,76,111,100,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,10, -99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18, -99,111,111,114,100,52,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,20,0,9,18,99,111,111,114, -100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114, -101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12, -0,116,101,120,116,117,114,101,50,68,80,114,111,106,76,111,100,0,1,1,0,17,115,97,109,112,108,101, -114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111, -114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18, -99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0, -4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, -108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80, -114,111,106,76,111,100,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1, -1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0, -59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18, -112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0, -18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0, -0,0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,76,111,100,0,1,1,0,18,115,97,109,112,108,101,114, -0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52, -0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,20,0, -9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,51, -100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100, -52,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111,106,76,111,100,0,1,1,0,18,115, -97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12, -1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111,114, -100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59, -119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120, -116,117,114,101,67,117,98,101,76,111,100,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, -111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111, -111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119, -0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97, -108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97, -100,111,119,49,68,76,111,100,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100, -0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100, -52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108, +86,97,108,0,18,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110, +77,97,116,114,105,120,0,16,8,48,0,57,18,103,108,95,86,101,114,116,101,120,0,59,120,120,120,120,0, +48,18,103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116, +114,105,120,0,16,10,49,0,57,18,103,108,95,86,101,114,116,101,120,0,59,121,121,121,121,0,48,46,18, +103,108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114, +105,120,0,16,10,50,0,57,18,103,108,95,86,101,114,116,101,120,0,59,122,122,122,122,0,48,46,18,103, +108,95,77,111,100,101,108,86,105,101,119,80,114,111,106,101,99,116,105,111,110,77,97,116,114,105, +120,0,16,10,51,0,57,18,103,108,95,86,101,114,116,101,120,0,59,119,119,119,119,0,48,46,20,0,0,1,0,0, +12,0,116,101,120,116,117,114,101,49,68,76,111,100,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0, +9,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18, +99,111,111,114,100,52,0,59,120,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119, +0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108, +0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,116,101,120,116, +117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,10,99, +111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112, +99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,121,0,49, +20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120, +98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111, +114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,76,111,100,0,1,1,0,16, +115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0, +12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0, +59,120,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108, 111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18, -115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119, -49,68,80,114,111,106,76,111,100,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114, -100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111, -114,100,0,59,120,0,18,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18, -112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114, -100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101, -116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0, -115,104,97,100,111,119,50,68,76,111,100,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, -111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111, -111,114,100,52,0,59,120,121,122,0,18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119, -0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108, -0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97,100, -111,119,50,68,80,114,111,106,76,111,100,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111, +115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117, +114,101,50,68,76,111,100,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0, +1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0, +59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18, +108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0, +18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,116,101,120,116,117, +114,101,50,68,80,114,111,106,76,111,100,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, 111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99, -111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,119, -0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99, -111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95, -95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,0 +111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114,100,0,59,122, +0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101, +120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,112,99,111, +111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,76,111,100,0,1,1,0, +17,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2, +0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114, +100,0,59,120,121,0,18,99,111,111,114,100,0,59,122,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0, +18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0, +0,18,115,97,109,112,108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116, +117,114,101,51,68,76,111,100,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100, +0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100, +52,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,20,0,9,18,99,111,111,114,100,52,0, +59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86, +97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,116,101, +120,116,117,114,101,51,68,80,114,111,106,76,111,100,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1, +0,12,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2,0,0,12,1,112,99,111,111,114,100,0,0,0, +9,18,112,99,111,111,114,100,0,59,120,121,122,0,18,99,111,111,114,100,0,59,120,121,122,0,18,99,111, +111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118, +101,99,52,95,116,101,120,98,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, +114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,67,117,98,101,76, +111,100,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,108,111, +100,0,0,0,1,3,2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0, +18,99,111,111,114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101, +99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, +114,0,0,18,99,111,111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,76,111,100,0,1,1, +0,20,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3, +2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111, +114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116, +101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, +111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106,76,111,100,0,1,1,0, +20,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2, +0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,0,18,99,111,111,114,100, +0,59,120,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0,18,99, +111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0,4,118, +101,99,52,95,116,101,120,98,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101, +114,0,0,18,112,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,76,111,100,0,1,1, +0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3, +2,0,0,12,1,99,111,111,114,100,52,0,0,0,9,18,99,111,111,114,100,52,0,59,120,121,122,0,18,99,111,111, +114,100,0,20,0,9,18,99,111,111,114,100,52,0,59,119,0,18,108,111,100,0,20,0,4,118,101,99,52,95,116, +101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, +111,114,100,52,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,80,114,111,106,76,111,100,0,1,1,0, +21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,1,1,0,9,108,111,100,0,0,0,1,3,2, +0,0,12,1,112,99,111,111,114,100,0,0,0,9,18,112,99,111,111,114,100,0,59,120,121,0,18,99,111,111,114, +100,0,59,120,121,0,18,99,111,111,114,100,0,59,119,0,49,20,0,9,18,112,99,111,111,114,100,0,59,122,0, +18,99,111,111,114,100,0,59,122,0,20,0,9,18,112,99,111,111,114,100,0,59,119,0,18,108,111,100,0,20,0, +4,118,101,99,52,95,116,101,120,98,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112, +108,101,114,0,0,18,112,99,111,111,114,100,0,0,0,0,0 -- cgit v1.2.3 From e55093bf2f4ff32a691f63409b9fea321cf8eb0f Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 20 Aug 2008 14:39:10 -0600 Subject: mesa: glsl: add check for exceeding max number of constants/uniforms --- src/mesa/shader/slang/slang_emit.c | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index c5a0fcb6c7..f5f41eb116 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -1971,6 +1971,7 @@ _slang_emit_code(slang_ir_node *n, slang_var_table *vt, GET_CURRENT_CONTEXT(ctx); GLboolean success; slang_emit_info emitInfo; + GLuint maxUniforms; emitInfo.log = log; emitInfo.vt = vt; @@ -1987,6 +1988,19 @@ _slang_emit_code(slang_ir_node *n, slang_var_table *vt, emitInfo.EmitHighLevelInstructions = GL_TRUE; } + /* Check uniform/constant limits */ + if (prog->Target == GL_FRAGMENT_PROGRAM_ARB) { + maxUniforms = ctx->Const.FragmentProgram.MaxUniformComponents / 4; + } + else { + assert(prog->Target == GL_VERTEX_PROGRAM_ARB); + maxUniforms = ctx->Const.VertexProgram.MaxUniformComponents / 4; + } + if (prog->Parameters->NumParameters > maxUniforms) { + slang_info_log_error(log, "Constant/uniform register limit exceeded"); + return GL_FALSE; + } + (void) emit(&emitInfo, n); /* finish up by adding the END opcode to program */ -- cgit v1.2.3 From ad9c862ffbb42d733595052d5769405ef7a98a75 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Wed, 20 Aug 2008 14:40:02 -0600 Subject: mesa: glsl: fix a swizzle bug in storage_to_src_reg() Need to remove the 'nil' components before swizzling a swizzle --- src/mesa/shader/slang/slang_emit.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c index f5f41eb116..9e8daa1051 100644 --- a/src/mesa/shader/slang/slang_emit.c +++ b/src/mesa/shader/slang/slang_emit.c @@ -108,7 +108,9 @@ writemask_to_swizzle(GLuint writemask) /** - * Swizzle a swizzle. That is, return swz2(swz1) + * Swizzle a swizzle (function composition). + * That is, return swz2(swz1), or said another way: swz1.szw2 + * Example: swizzle_swizzle(".zwxx", ".xxyw") yields ".zzwx" */ GLuint _slang_swizzle_swizzle(GLuint swz1, GLuint swz2) @@ -279,7 +281,7 @@ storage_to_src_reg(struct prog_src_register *src, const slang_ir_storage *st) while (st->Parent) { st = st->Parent; index += st->Index; - swizzle = _slang_swizzle_swizzle(st->Swizzle, swizzle); + swizzle = _slang_swizzle_swizzle(fix_swizzle(st->Swizzle), swizzle); } assert(st->File >= 0); -- cgit v1.2.3 From 0fff3e4ea991ce2f841739ee8c8e8937452e56fa Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 21 Aug 2008 13:39:52 +0100 Subject: trace: Split the output stream on windows. Because windows limits the ammount of memory that can be mapped. --- src/gallium/drivers/trace/tr_stream_wd.c | 102 +++++++++++++++++++++++-------- 1 file changed, 75 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_stream_wd.c b/src/gallium/drivers/trace/tr_stream_wd.c index 9a05c0cec8..c375f50b95 100644 --- a/src/gallium/drivers/trace/tr_stream_wd.c +++ b/src/gallium/drivers/trace/tr_stream_wd.c @@ -47,32 +47,73 @@ struct trace_stream { + char filename[MAX_PATH + 1]; WCHAR wFileName[MAX_PATH + 1]; ULONG_PTR iFile; char *pMap; size_t written; + unsigned suffix; }; +static INLINE boolean +trace_stream_map(struct trace_stream *stream) +{ + ULONG BytesInUnicodeString; + static char filename[MAX_PATH + 1]; + unsigned filename_len; + + filename_len = util_snprintf(filename, + sizeof(filename), + "\\??\\%s.%03u", + stream->filename, + stream->suffix++); + + EngMultiByteToUnicodeN( + stream->wFileName, + sizeof(stream->wFileName), + &BytesInUnicodeString, + filename, + filename_len); + + stream->pMap = EngMapFile(stream->wFileName, MAP_FILE_SIZE, &stream->iFile); + if(!stream->pMap) + return FALSE; + + memset(stream->pMap, 0, MAP_FILE_SIZE); + stream->written = 0; + + return TRUE; +} + + +static INLINE void +trace_stream_unmap(struct trace_stream *stream) +{ + EngUnmapFile(stream->iFile); + if(stream->written < MAP_FILE_SIZE) { + /* Truncate file size */ + stream->pMap = EngMapFile(stream->wFileName, stream->written, &stream->iFile); + if(stream->pMap) + EngUnmapFile(stream->iFile); + } + + stream->pMap = NULL; +} + + struct trace_stream * trace_stream_create(const char *filename) { struct trace_stream *stream; - ULONG BytesInUnicodeString; stream = CALLOC_STRUCT(trace_stream); if(!stream) goto error1; - EngMultiByteToUnicodeN( - stream->wFileName, - sizeof(stream->wFileName), - &BytesInUnicodeString, - (char *)filename, - strlen(filename)); + strncpy(stream->filename, filename, sizeof(stream->filename)); - stream->pMap = EngMapFile(stream->wFileName, MAP_FILE_SIZE, &stream->iFile); - if(!stream->pMap) + if(!trace_stream_map(stream)) goto error2; return stream; @@ -84,25 +125,38 @@ error1: } +static INLINE void +trace_stream_copy(struct trace_stream *stream, const char *data, size_t size) +{ + assert(stream->written + size <= MAP_FILE_SIZE); + memcpy(stream->pMap + stream->written, data, size); + stream->written += size; +} + + boolean trace_stream_write(struct trace_stream *stream, const void *data, size_t size) { - boolean ret; - if(!stream) return FALSE; - if(stream->written + size > MAP_FILE_SIZE) { - ret = FALSE; - size = MAP_FILE_SIZE - stream->written; - } - else - ret = TRUE; + if(!stream->pMap) + return FALSE; - memcpy(stream->pMap + stream->written, data, size); - stream->written += size; + while(stream->written + size > MAP_FILE_SIZE) { + size_t step = MAP_FILE_SIZE - stream->written; + trace_stream_copy(stream, data, step); + data = (const char *)data + step; + size -= step; + + trace_stream_unmap(stream); + if(!trace_stream_map(stream)) + return FALSE; + } + + trace_stream_copy(stream, data, size); - return ret; + return TRUE; } @@ -119,13 +173,7 @@ trace_stream_close(struct trace_stream *stream) if(!stream) return; - EngUnmapFile(stream->iFile); - if(stream->written < MAP_FILE_SIZE) { - /* Truncate file size */ - stream->pMap = EngMapFile(stream->wFileName, stream->written, &stream->iFile); - if(stream->pMap) - EngUnmapFile(stream->iFile); - } + trace_stream_unmap(stream); FREE(stream); } -- cgit v1.2.3 From 34d12c1787116c254e528dd981810b7b78b7a2ee Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 21 Aug 2008 13:57:59 +0100 Subject: trace: Hack to detect writes to user buffers. It often happens that new data is written directly to the user buffers without mapping/unmapping. This hack marks user buffers and dumps them before passing them to pipe context. --- src/gallium/drivers/trace/tr_context.c | 11 ++++++++++ src/gallium/drivers/trace/tr_winsys.c | 40 ++++++++++++++++++++++++++++++++++ src/gallium/drivers/trace/tr_winsys.h | 13 +++++++++++ 3 files changed, 64 insertions(+) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index 529bed3c6b..f16359e8ad 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -32,6 +32,7 @@ #include "tr_state.h" #include "tr_screen.h" #include "tr_texture.h" +#include "tr_winsys.h" #include "tr_context.h" @@ -131,6 +132,8 @@ trace_context_draw_elements(struct pipe_context *_pipe, struct pipe_context *pipe = tr_ctx->pipe; boolean result; + trace_winsys_user_buffer_update(_pipe->winsys, indexBuffer); + trace_dump_call_begin("pipe_context", "draw_elements"); trace_dump_arg(ptr, pipe); @@ -164,6 +167,8 @@ trace_context_draw_range_elements(struct pipe_context *_pipe, struct pipe_context *pipe = tr_ctx->pipe; boolean result; + trace_winsys_user_buffer_update(_pipe->winsys, indexBuffer); + trace_dump_call_begin("pipe_context", "draw_range_elements"); trace_dump_arg(ptr, pipe); @@ -691,6 +696,8 @@ trace_context_set_constant_buffer(struct pipe_context *_pipe, struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; + trace_winsys_user_buffer_update(_pipe->winsys, (struct pipe_buffer *)buffer); + trace_dump_call_begin("pipe_context", "set_constant_buffer"); trace_dump_arg(ptr, pipe); @@ -820,6 +827,10 @@ trace_context_set_vertex_buffers(struct pipe_context *_pipe, { struct trace_context *tr_ctx = trace_context(_pipe); struct pipe_context *pipe = tr_ctx->pipe; + unsigned i; + + for(i = 0; i < num_buffers; ++i) + trace_winsys_user_buffer_update(_pipe->winsys, buffers[i].buffer); trace_dump_call_begin("pipe_context", "set_vertex_buffers"); diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index 120006ea9f..2c7a6f893b 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -242,10 +242,50 @@ trace_winsys_user_buffer_create(struct pipe_winsys *_winsys, trace_dump_call_end(); + /* XXX: Mark the user buffers. (we should wrap pipe_buffers, but is is + * impossible to do so while texture-less surfaces are still around */ + if(result) { + assert(!(result->usage & TRACE_BUFFER_USAGE_USER)); + result->usage |= TRACE_BUFFER_USAGE_USER; + } + return result; } +void +trace_winsys_user_buffer_update(struct pipe_winsys *_winsys, + struct pipe_buffer *buffer) +{ + struct trace_winsys *tr_ws = trace_winsys(_winsys); + struct pipe_winsys *winsys = tr_ws->winsys; + const void *map; + + if(buffer && buffer->usage & TRACE_BUFFER_USAGE_USER) { + map = winsys->buffer_map(winsys, buffer, PIPE_BUFFER_USAGE_CPU_READ); + if(map) { + trace_dump_call_begin("pipe_winsys", "buffer_write"); + + trace_dump_arg(ptr, winsys); + + trace_dump_arg(ptr, buffer); + + trace_dump_arg_begin("data"); + trace_dump_bytes(map, buffer->size); + trace_dump_arg_end(); + + trace_dump_arg_begin("size"); + trace_dump_uint(buffer->size); + trace_dump_arg_end(); + + trace_dump_call_end(); + + winsys->buffer_unmap(winsys, buffer); + } + } +} + + static void * trace_winsys_buffer_map(struct pipe_winsys *_winsys, struct pipe_buffer *buffer, diff --git a/src/gallium/drivers/trace/tr_winsys.h b/src/gallium/drivers/trace/tr_winsys.h index 2218117347..062ddf66a0 100644 --- a/src/gallium/drivers/trace/tr_winsys.h +++ b/src/gallium/drivers/trace/tr_winsys.h @@ -34,6 +34,14 @@ #include "pipe/p_winsys.h" +/** + * It often happens that new data is written directly to the user buffers + * without mapping/unmapping. This flag marks user buffers, so that their + * contents can be dumpped before being used by the pipe context. + */ +#define TRACE_BUFFER_USAGE_USER (1 << 31) + + struct hash_table; @@ -60,4 +68,9 @@ struct pipe_winsys * trace_winsys_create(struct pipe_winsys *winsys); +void +trace_winsys_user_buffer_update(struct pipe_winsys *winsys, + struct pipe_buffer *buffer); + + #endif /* TR_WINSYS_H_ */ -- cgit v1.2.3 From 10624065b0dc631164d786b2c39f35655e55eadf Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Thu, 21 Aug 2008 18:45:43 +0100 Subject: python/retrace: Support gziped traces. --- src/gallium/state_trackers/python/retrace/interpreter.py | 7 ++++++- src/gallium/state_trackers/python/retrace/parser.py | 10 ++++++++-- 2 files changed, 14 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index 7502c096b6..f596a54d10 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -471,7 +471,12 @@ class Interpreter: def main(): for arg in sys.argv[1:]: - parser = TraceParser(open(arg, 'rt')) + if arg.endswith('.gz'): + import gzip + stream = gzip.GzipFile(arg, 'rt') + else: + stream = open(arg, 'rt') + parser = TraceParser(stream) trace = parser.parse() interpreter = Interpreter() interpreter.interpret(trace) diff --git a/src/gallium/state_trackers/python/retrace/parser.py b/src/gallium/state_trackers/python/retrace/parser.py index d02fd7f741..2ee4d3068d 100755 --- a/src/gallium/state_trackers/python/retrace/parser.py +++ b/src/gallium/state_trackers/python/retrace/parser.py @@ -322,9 +322,15 @@ class TraceParser(XmlParser): def main(): for arg in sys.argv[1:]: - parser = TraceParser(open(arg, 'rt')) + if arg.endswith('.gz'): + import gzip + stream = gzip.GzipFile(arg, 'rt') + else: + stream = open(arg, 'rt') + parser = TraceParser(stream) trace = parser.parse() - print trace + for call in trace.calls: + print call if __name__ == '__main__': -- cgit v1.2.3 From 7f3c3683ce50bad42411547d4b78e03d7f20e498 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 22 Aug 2008 01:23:48 +0100 Subject: python/retrace: Process the trace call-by-call (instead of reading everything into memory). --- .../state_trackers/python/retrace/interpreter.py | 24 +++++--------------- .../state_trackers/python/retrace/parser.py | 26 ++++++++++++++-------- 2 files changed, 23 insertions(+), 27 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index f596a54d10..a2f924f528 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -22,7 +22,7 @@ import sys import gallium import model -from parser import TraceParser +import parser def make_image(surface): @@ -426,9 +426,10 @@ class Context(Object): show_image(self.cbufs[0]) -class Interpreter: +class Interpreter(parser.TraceParser): - def __init__(self): + def __init__(self, stream): + parser.TraceParser.__init__(self, stream) self.objects = {} self.result = None self.globl = Global(self, None) @@ -447,7 +448,7 @@ class Interpreter: for call in trace.calls: self.interpret_call(call) - def interpret_call(self, call): + def handle_call(self, call): sys.stderr.write("%s\n" % call) args = [self.interpret_arg(arg) for name, arg in call.args] @@ -469,18 +470,5 @@ class Interpreter: return translator.visit(node) -def main(): - for arg in sys.argv[1:]: - if arg.endswith('.gz'): - import gzip - stream = gzip.GzipFile(arg, 'rt') - else: - stream = open(arg, 'rt') - parser = TraceParser(stream) - trace = parser.parse() - interpreter = Interpreter() - interpreter.interpret(trace) - - if __name__ == '__main__': - main() + parser.main(Interpreter) diff --git a/src/gallium/state_trackers/python/retrace/parser.py b/src/gallium/state_trackers/python/retrace/parser.py index 2ee4d3068d..6bc75ad685 100755 --- a/src/gallium/state_trackers/python/retrace/parser.py +++ b/src/gallium/state_trackers/python/retrace/parser.py @@ -183,12 +183,11 @@ class TraceParser(XmlParser): def parse(self): self.element_start('trace') - calls = [] while self.token.type not in (ELEMENT_END, EOF): - calls.append(self.parse_call()) + call = self.parse_call() + self.handle_call(call) if self.token.type != EOF: self.element_end('trace') - return Trace(calls) def parse_call(self): attrs = self.element_start('call') @@ -319,19 +318,28 @@ class TraceParser(XmlParser): return Pointer(address) + def handle_call(self, call): + + pass + + +class TraceDumper(TraceParser): + + + def handle_call(self, call): + print call + -def main(): +def main(ParserFactory): for arg in sys.argv[1:]: if arg.endswith('.gz'): import gzip stream = gzip.GzipFile(arg, 'rt') else: stream = open(arg, 'rt') - parser = TraceParser(stream) - trace = parser.parse() - for call in trace.calls: - print call + parser = ParserFactory(stream) + parser.parse() if __name__ == '__main__': - main() + main(TraceDumper) -- cgit v1.2.3 From 152d00d199f2c8c46a96be450cfd017b4798b4d1 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 22 Aug 2008 03:13:27 +0100 Subject: trace: Use a 4 hexadecimal digit suffix. --- src/gallium/drivers/trace/tr_stream_wd.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_stream_wd.c b/src/gallium/drivers/trace/tr_stream_wd.c index c375f50b95..45309dcb4e 100644 --- a/src/gallium/drivers/trace/tr_stream_wd.c +++ b/src/gallium/drivers/trace/tr_stream_wd.c @@ -65,7 +65,7 @@ trace_stream_map(struct trace_stream *stream) filename_len = util_snprintf(filename, sizeof(filename), - "\\??\\%s.%03u", + "\\??\\%s.%04x", stream->filename, stream->suffix++); -- cgit v1.2.3 From 807a7487ff6f0af60240ce467bd1ac160b0b054e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 22 Aug 2008 03:14:24 +0100 Subject: trace: Don't trace texture/surfaces releases, only destructions. --- src/gallium/drivers/trace/tr_screen.c | 34 ++++++++++++++++++---------------- 1 file changed, 18 insertions(+), 16 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index cca8597a87..a6467ec35f 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -224,19 +224,20 @@ trace_screen_texture_release(struct pipe_screen *_screen, else texture = NULL; - trace_dump_call_begin("pipe_screen", "texture_release"); - - trace_dump_arg(ptr, screen); - trace_dump_arg(ptr, texture); - if (*ptexture) { - if (!--(*ptexture)->refcount) + if (!--(*ptexture)->refcount) { + trace_dump_call_begin("pipe_screen", "texture_destroy"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, texture); + trace_texture_destroy(tr_scr, *ptexture); + + trace_dump_call_end(); + } *ptexture = NULL; } - - trace_dump_call_end(); } @@ -297,19 +298,20 @@ trace_screen_tex_surface_release(struct pipe_screen *_screen, else surface = NULL; - trace_dump_call_begin("pipe_screen", "tex_surface_release"); - - trace_dump_arg(ptr, screen); - trace_dump_arg(ptr, surface); - if (*psurface) { - if (!--(*psurface)->refcount) + if (!--(*psurface)->refcount) { + trace_dump_call_begin("pipe_screen", "tex_surface_destroy"); + + trace_dump_arg(ptr, screen); + trace_dump_arg(ptr, surface); + trace_surface_destroy(tr_tex, *psurface); + + trace_dump_call_end(); + } *psurface = NULL; } - - trace_dump_call_end(); } -- cgit v1.2.3 From d32f51d4717686d626b0f7ec9a92c3c9af50009f Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Jos=C3=A9=20Fonseca?= Date: Fri, 22 Aug 2008 03:14:51 +0100 Subject: python/retrace: Interpret texture/surfaces destructions. --- src/gallium/state_trackers/python/retrace/interpreter.py | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/state_trackers/python/retrace/interpreter.py b/src/gallium/state_trackers/python/retrace/interpreter.py index a2f924f528..351a6e739b 100755 --- a/src/gallium/state_trackers/python/retrace/interpreter.py +++ b/src/gallium/state_trackers/python/retrace/interpreter.py @@ -254,15 +254,21 @@ class Screen(Object): tex_usage = template.tex_usage, ) - def texture_release(self, texture): + def texture_destroy(self, texture): self.interpreter.unregister_object(texture) + def texture_release(self, surface): + pass + def get_tex_surface(self, texture, face, level, zslice, usage): return texture.get_surface(face, level, zslice, usage) - def tex_surface_release(self, surface): + def tex_surface_destroy(self, surface): self.interpreter.unregister_object(surface) + def tex_surface_release(self, surface): + pass + def surface_write(self, surface, data, stride, size): assert surface.nblocksy * stride == size surface.put_tile_raw(0, 0, surface.width, surface.height, data, stride) -- cgit v1.2.3 From 9e2b867b3f2e9afc9e9f9178788ae07f6be1f3c0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 15:13:47 -0600 Subject: gallium: new u_math.[ch] files for math functions So far, optimized/low-precision versions of exp(), exp2(), log2(), pow(). --- src/gallium/auxiliary/util/Makefile | 1 + src/gallium/auxiliary/util/SConscript | 1 + src/gallium/auxiliary/util/u_math.c | 60 ++++++++++++++ src/gallium/auxiliary/util/u_math.h | 144 ++++++++++++++++++++++++++++++++++ 4 files changed, 206 insertions(+) create mode 100644 src/gallium/auxiliary/util/u_math.c create mode 100644 src/gallium/auxiliary/util/u_math.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index db3d810a2e..be1b97b535 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -12,6 +12,7 @@ C_SOURCES = \ u_gen_mipmap.c \ u_handle_table.c \ u_hash_table.c \ + u_math.c \ u_mm.c \ u_simple_shaders.c \ u_snprintf.c \ diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 0309de1ac2..7865307a7a 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -13,6 +13,7 @@ util = env.ConvenienceLibrary( 'u_gen_mipmap.c', 'u_handle_table.c', 'u_hash_table.c', + 'u_math.c', 'u_mm.c', 'u_simple_shaders.c', 'u_snprintf.c', diff --git a/src/gallium/auxiliary/util/u_math.c b/src/gallium/auxiliary/util/u_math.c new file mode 100644 index 0000000000..8bf6551b6e --- /dev/null +++ b/src/gallium/auxiliary/util/u_math.c @@ -0,0 +1,60 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + + +#include "util/u_math.h" + + + +float pow2_table[POW2_TABLE_SIZE]; + + +static void +init_pow2_table(void) +{ + int i; + for (i = 0; i < POW2_TABLE_SIZE; i++) { + pow2_table[i] = pow(2.0, i / POW2_TABLE_SCALE); + } +} + + +/** + * One time init for math utilities. + */ +void +util_init_math(void) +{ + static boolean initialized = FALSE; + if (!initialized) { + init_pow2_table(); + initialized = TRUE; + } +} + + diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h new file mode 100644 index 0000000000..e179ce700f --- /dev/null +++ b/src/gallium/auxiliary/util/u_math.h @@ -0,0 +1,144 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Math utilities and approximations for common math functions. + * Reduced precision is usually acceptable in shaders... + */ + + +#ifndef U_MATH_H +#define U_MATH_H + + +#include "pipe/p_util.h" +#include "util/u_math.h" + + + +#define POW2_TABLE_SIZE 256 +#define POW2_TABLE_SCALE ((float) (POW2_TABLE_SIZE-1)) +extern float pow2_table[POW2_TABLE_SIZE]; + + +extern void +util_init_math(void); + + + + +/** + * Fast approximation to exp(x). + * Compute with base 2 exponents: exp(x) = exp2(log2(e) * x) + * Note: log2(e) is a constant, k = 1.44269 + * So, exp(x) = exp2(k * x); + * Identity: exp2(a + b) = exp2(a) * exp2(b) + * Let ipart = int(k*x) + * Let fpart = k*x - ipart; + * So, exp2(k*x) = exp2(ipart) * exp2(fpart) + * Compute exp2(ipart) with i << ipart + * Compute exp2(fpart) with lookup table. + */ +static INLINE float +util_fast_exp(float x) +{ + if (x >= 0.0f) { + float k = 1.44269f; /* = log2(e) */ + float kx = k * x; + int ipart = (int) kx; + float fpart = kx - (float) ipart; + float y = (float) (1 << ipart) + * pow2_table[(int) (fpart * POW2_TABLE_SCALE)]; + return y; + } + else { + /* exp(-x) = 1.0 / exp(x) */ + float k = -1.44269f; + float kx = k * x; + int ipart = (int) kx; + float fpart = kx - (float) ipart; + float y = (float) (1 << ipart) + * pow2_table[(int) (fpart * POW2_TABLE_SCALE)]; + return 1.0f / y; + } +} + + +/** + * Fast version of 2^x + * XXX the above function could be implemented in terms of this one. + */ +static INLINE float +util_fast_exp2(float x) +{ + if (x >= 0.0f) { + int ipart = (int) x; + float fpart = x - (float) ipart; + float y = (float) (1 << ipart) + * pow2_table[(int) (fpart * POW2_TABLE_SCALE)]; + return y; + } + else { + /* exp(-x) = 1.0 / exp(x) */ + int ipart = (int) -x; + float fpart = -x - (float) ipart; + float y = (float) (1 << ipart) + * pow2_table[(int) (fpart * POW2_TABLE_SCALE)]; + return 1.0f / y; + } +} + + +/** + * Based on code from http://www.flipcode.com/totd/ + */ +static INLINE float +util_fast_log2(float val) +{ + union fi num; + int log_2; + num.f = val; + log_2 = ((num.i >> 23) & 255) - 128; + num.i &= ~(255 << 23); + num.i += 127 << 23; + num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3; + return num.f + log_2; +} + + +static INLINE float +util_fast_pow(float x, float y) +{ + /* XXX this test may need adjustment */ + if (y >= 3.0 && -0.02f <= x && x <= 0.02f) + return 0.0f; + return util_fast_exp2(util_fast_log2(x) * y); +} + + +#endif /* U_MATH_H */ -- cgit v1.2.3 From 47d4b958cfaac080a97bf0ba21d78b3ce37b62a9 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 15:14:36 -0600 Subject: mesa: glsl: implement exp() functions in terms of EXP asm instruction, not pow --- .../shader/slang/library/slang_common_builtin.gc | 29 +++++++++++----------- 1 file changed, 15 insertions(+), 14 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin.gc b/src/mesa/shader/slang/library/slang_common_builtin.gc index e908e6c940..561e94f6ba 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin.gc +++ b/src/mesa/shader/slang/library/slang_common_builtin.gc @@ -473,32 +473,33 @@ vec4 pow(const vec4 a, const vec4 b) float exp(const float a) { - const float e = 2.71828; - __asm float_power __retVal, e, a; + // NOTE: log2(e) = 1.44269502 + float t = a * 1.44269502; + __asm float_exp2 __retVal, t; } vec2 exp(const vec2 a) { - const float e = 2.71828; - __asm float_power __retVal.x, e, a.x; - __asm float_power __retVal.y, e, a.y; + vec2 t = a * 1.44269502; + __asm float_exp2 __retVal.x, t.x; + __asm float_exp2 __retVal.y, t.y; } vec3 exp(const vec3 a) { - const float e = 2.71828; - __asm float_power __retVal.x, e, a.x; - __asm float_power __retVal.y, e, a.y; - __asm float_power __retVal.z, e, a.z; + vec3 t = a * 1.44269502; + __asm float_exp2 __retVal.x, t.x; + __asm float_exp2 __retVal.y, t.y; + __asm float_exp2 __retVal.z, t.z; } vec4 exp(const vec4 a) { - const float e = 2.71828; - __asm float_power __retVal.x, e, a.x; - __asm float_power __retVal.y, e, a.y; - __asm float_power __retVal.z, e, a.z; - __asm float_power __retVal.w, e, a.w; + vec4 t = a * 1.44269502; + __asm float_exp2 __retVal.x, t.x; + __asm float_exp2 __retVal.y, t.y; + __asm float_exp2 __retVal.z, t.z; + __asm float_exp2 __retVal.w, t.w; } -- cgit v1.2.3 From 07bebb839d8f5e35209c446133743892084a672c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 15:14:47 -0600 Subject: mesa: glsl: regenerated file --- .../shader/slang/library/slang_common_builtin_gc.h | 1104 ++++++++++---------- 1 file changed, 552 insertions(+), 552 deletions(-) (limited to 'src') diff --git a/src/mesa/shader/slang/library/slang_common_builtin_gc.h b/src/mesa/shader/slang/library/slang_common_builtin_gc.h index 3b07c15785..d20240543a 100644 --- a/src/mesa/shader/slang/library/slang_common_builtin_gc.h +++ b/src/mesa/shader/slang/library/slang_common_builtin_gc.h @@ -244,573 +244,573 @@ 0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59, 122,0,0,18,97,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18, 95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,18,98,0,59,119,0,0,0,0,1,0,0,9,0,101, -120,112,0,1,1,0,9,97,0,0,0,1,3,2,1,0,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97, -116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,0,18,101,0,0,18,97,0,0,0,0,1,0,0,10, -0,101,120,112,0,1,1,0,10,97,0,0,0,1,3,2,1,0,9,1,101,0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108, -111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0, -59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,121, -0,0,18,101,0,0,18,97,0,59,121,0,0,0,0,1,0,0,11,0,101,120,112,0,1,1,0,11,97,0,0,0,1,3,2,1,0,9,1,101, -0,2,17,50,0,55,49,56,50,56,0,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101, -116,86,97,108,0,59,120,0,0,18,101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101, -114,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97, -116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18,97,0,59,122, -0,0,0,0,1,0,0,12,0,101,120,112,0,1,1,0,12,97,0,0,0,1,3,2,1,0,9,1,101,0,2,17,50,0,55,49,56,50,56,0, -0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -101,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0,18,95,95,114,101,116,86, -97,108,0,59,121,0,0,18,101,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,112,111,119,101,114,0, -18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,101,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116, -95,112,111,119,101,114,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,101,0,0,18,97,0,59,119,0,0, -0,0,1,0,0,9,0,108,111,103,50,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95, -95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,10,0,108,111,103,50,0,1,1,0,10,118,0,0,0,1,4,102, -108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0, -0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59, -121,0,0,0,0,1,0,0,11,0,108,111,103,50,0,1,1,0,11,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103, -50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108, -111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97, -116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,0,0, -12,0,108,111,103,50,0,1,1,0,12,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114, -101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18, -95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103, -50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,108, -111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,0,9,0,108,111, -103,0,1,1,0,9,120,0,0,0,1,3,2,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108, -111,103,50,0,18,120,0,0,0,18,99,0,48,0,0,1,0,0,10,0,108,111,103,0,1,1,0,10,118,0,0,0,1,3,2,1,0,9,1, +120,112,0,1,1,0,9,97,0,0,0,1,3,2,0,0,9,1,116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0, +0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,116,0,0,0,0,1,0,0, +10,0,101,120,112,0,1,1,0,10,97,0,0,0,1,3,2,0,0,10,1,116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48, +50,0,0,48,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, +18,116,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59, +121,0,0,18,116,0,59,121,0,0,0,0,1,0,0,11,0,101,120,112,0,1,1,0,11,97,0,0,0,1,3,2,0,0,11,1,116,0,2, +18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95, +95,114,101,116,86,97,108,0,59,120,0,0,18,116,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50, +0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,116,0,59,121,0,0,0,4,102,108,111,97,116,95,101, +120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,116,0,59,122,0,0,0,0,1,0,0,12,0,101, +120,112,0,1,1,0,12,97,0,0,0,1,3,2,0,0,12,1,116,0,2,18,97,0,17,49,0,52,52,50,54,57,53,48,50,0,0,48, +0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,116,0, +59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0, +18,116,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59, +122,0,0,18,116,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97, +108,0,59,119,0,0,18,116,0,59,119,0,0,0,0,1,0,0,9,0,108,111,103,50,0,1,1,0,9,120,0,0,0,1,4,102,108, +111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,10,0,108, +111,103,50,0,1,1,0,10,118,0,0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86, +97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114, +101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,0,0,11,0,108,111,103,50,0,1,1,0,11,118,0, +0,0,1,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118, +0,59,120,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0, +0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0, +59,122,0,0,18,118,0,59,122,0,0,0,0,1,0,0,12,0,108,111,103,50,0,1,1,0,12,118,0,0,0,1,4,102,108,111, +97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102, +108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0, +0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59, +122,0,0,0,4,102,108,111,97,116,95,108,111,103,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18, +118,0,59,119,0,0,0,0,1,0,0,9,0,108,111,103,0,1,1,0,9,120,0,0,0,1,3,2,1,0,9,1,99,0,2,17,48,0,54,57, +51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,120,0,0,0,18,99,0,48,0,0,1,0,0,10,0,108,111, +103,0,1,1,0,10,118,0,0,0,1,3,2,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108, +111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,0,11,0,108,111,103,0,1,1,0,11,118,0,0,0,1,3,2,1,0,9,1, 99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0, -0,1,0,0,11,0,108,111,103,0,1,1,0,11,118,0,0,0,1,3,2,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56, -49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,0,12,0,108,111,103,0,1,1,0,12,118, -0,0,0,1,3,2,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56,49,0,0,0,0,8,58,108,111,103,50,0,18,118, -0,0,0,18,99,0,48,0,0,1,0,0,9,0,101,120,112,50,0,1,1,0,9,97,0,0,0,1,4,102,108,111,97,116,95,101,120, -112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,10,0,101,120,112,50,0,1,1,0,10,97, -0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97, -0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0, -0,18,97,0,59,121,0,0,0,0,1,0,0,11,0,101,120,112,50,0,1,1,0,11,97,0,0,0,1,4,102,108,111,97,116,95, -101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97, -116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108, -111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,0, -1,0,0,12,0,101,120,112,50,0,1,1,0,12,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95, +0,1,0,0,12,0,108,111,103,0,1,1,0,12,118,0,0,0,1,3,2,1,0,9,1,99,0,2,17,48,0,54,57,51,49,52,55,49,56, +49,0,0,0,0,8,58,108,111,103,50,0,18,118,0,0,0,18,99,0,48,0,0,1,0,0,9,0,101,120,112,50,0,1,1,0,9,97, +0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0, +1,0,0,10,0,101,120,112,50,0,1,1,0,10,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95, 114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0, -18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120, -112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,4,102,108,111,97,116,95, -101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,97,0,59,119,0,0,0,0,1,0,0,9,0,115, -113,114,116,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18, -114,0,0,18,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18, -114,0,0,0,0,1,0,0,10,0,115,113,114,116,0,1,1,0,10,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4,102,108,111, -97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, -95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114, -0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, -121,0,0,18,114,0,0,0,0,1,0,0,11,0,115,113,114,116,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4, -102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114, -99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115, -113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116, -86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59, -122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114, -0,0,0,0,1,0,0,12,0,115,113,114,116,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4,102,108,111,97, -116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95, -95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0, -0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59, -121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,4,102,108, -111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,119,0,0,0,4,102,108,111,97,116,95,114,99,112,0, -18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,0,0,0,0,1,0,0,9,0,105,110,118,101,114,115,101, -115,113,114,116,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116, -86,97,108,0,59,120,0,0,18,120,0,0,0,0,1,0,0,10,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1, -0,10,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0, -18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59, -121,0,0,18,118,0,59,121,0,0,0,0,1,0,0,11,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,11, -118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0, -0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59, -122,0,0,18,118,0,59,122,0,0,0,0,1,0,0,12,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,12, -118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, -118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,121,0, -0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59, -122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114,101,116,86,97,108, -0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,0,9,0,110,111,114,109,97,108,105,122,101,0,1,1,0,9,120,0, -0,0,1,9,18,95,95,114,101,116,86,97,108,0,17,49,0,48,0,0,20,0,0,1,0,0,10,0,110,111,114,109,97,108, -105,122,101,0,1,1,0,10,118,0,0,0,1,3,2,1,0,9,1,115,0,2,58,105,110,118,101,114,115,101,115,113,114, -116,0,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,115,0,0,0,0,1,0,0,11,0,110, -111,114,109,97,108,105,122,101,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,116,109,112,0,0,0,4,118,101,99, -51,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113, -0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,0,0,12,0,110, -111,114,109,97,108,105,122,101,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,116,109,112,0,0,0,4,118,101,99, -52,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0,0,0,4,102,108,111,97,116,95,114,115,113, -0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,116,109,112,0,0,0,0,1,0,0,9,0,97,98, -115,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0, -0,0,0,1,0,0,10,0,97,98,115,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,11,0,97,98,115,0,1,1,0,11,97,0,0,0,1,4,118,101, -99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,12,0, -97,98,115,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,0, -18,97,0,0,0,0,1,0,0,9,0,115,105,103,110,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9,1,112,0,0,1,1,110,0,0,0,4, -118,101,99,52,95,115,103,116,0,18,112,0,0,18,120,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103, -116,0,18,110,0,0,17,48,0,48,0,0,0,18,120,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0, -18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,0,0,10,0,115,105,103,110,0,1,1,0,10, -118,0,0,0,1,3,2,0,0,10,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120, -121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,121,0,0,17, -48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116, -86,97,108,0,59,120,121,0,0,18,112,0,0,18,110,0,0,0,0,1,0,0,11,0,115,105,103,110,0,1,1,0,11,118,0,0, -0,1,3,2,0,0,11,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,59,120,121,122,0, -0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,59,120,121,122,0,0,17,48, -0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,112,0,0,18,110,0,0,0,0,1,0,0,12,0,115,105,103,110,0,1,1,0,12,118,0, -0,0,1,3,2,0,0,12,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,0,18,118,0,0,17, -48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118, -101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0, -0,0,0,1,0,0,9,0,102,108,111,111,114,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0, -18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,10,0,102,108,111,111,114,0,1,1,0,10,97,0,0, -0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97, -0,0,0,0,1,0,0,11,0,102,108,111,111,114,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111, -114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,12,0,102,108,111,111, -114,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108, -0,0,18,97,0,0,0,0,1,0,0,9,0,99,101,105,108,0,1,1,0,9,97,0,0,0,1,3,2,0,0,9,1,98,0,2,18,97,0,54,0,0, -4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0, -18,98,0,54,20,0,0,1,0,0,10,0,99,101,105,108,0,1,1,0,10,97,0,0,0,1,3,2,0,0,10,1,98,0,2,18,97,0,54,0, -0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,18,98,0,54,20,0,0,1,0,0,11,0,99,101,105,108,0,1,1,0,11,97,0,0,0,1,3,2,0,0,11,1,98,0, -2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101, -116,86,97,108,0,59,120,121,122,0,18,98,0,54,20,0,0,1,0,0,12,0,99,101,105,108,0,1,1,0,12,97,0,0,0,1, -3,2,0,0,12,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0, -9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,0,9,0,102,114,97,99,116,0,1,1,0,9,97,0,0, -0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,10,0, -102,114,97,99,116,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,11,0,102,114,97,99,116,0,1,1,0,11,97,0,0,0,1,4,118,101, -99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0, -12,0,102,114,97,99,116,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101, -116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,9,0,109,111,100,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0, -9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118, -101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111, -114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,10,0,109,111,100,0,1,1, -0,10,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97, -116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114, -66,0,48,0,0,48,47,20,0,0,1,0,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1, +18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59,121,0,0,0,0,1,0,0,11,0,101,120,112,50,0,1,1, +0,11,97,0,0,0,1,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0, +0,18,97,0,59,120,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0, +59,121,0,0,18,97,0,59,121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86, +97,108,0,59,122,0,0,18,97,0,59,122,0,0,0,0,1,0,0,12,0,101,120,112,50,0,1,1,0,12,97,0,0,0,1,4,102, +108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,97,0,59,120,0,0, +0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,97,0,59, +121,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,122,0,0,18, +97,0,59,122,0,0,0,4,102,108,111,97,116,95,101,120,112,50,0,18,95,95,114,101,116,86,97,108,0,59,119, +0,0,18,97,0,59,119,0,0,0,0,1,0,0,9,0,115,113,114,116,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4, +102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,0,0,10,0,115,113,114,116,0,1,1,0,10,118,0,0,0, +1,3,2,0,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102, +108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,0,1,0,0,11,0,115,113,114,116,0,1,1, +0,11,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0, +59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18, +114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97, +116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116, +95,114,115,113,0,18,114,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95, +114,101,116,86,97,108,0,59,122,0,0,18,114,0,0,0,0,1,0,0,12,0,115,113,114,116,0,1,1,0,12,118,0,0,0, +1,3,2,0,0,9,1,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,120,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,4,102, +108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113, +0,18,114,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97, +108,0,59,122,0,0,18,114,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,118,0,59,119,0,0, +0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,114,0,0,0,0, +1,0,0,9,0,105,110,118,101,114,115,101,115,113,114,116,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116, +95,114,115,113,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,120,0,0,0,0,1,0,0,10,0,105,110,118, +101,114,115,101,115,113,114,116,0,1,1,0,10,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95, +95,114,101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0, +18,95,95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,0,1,0,0,11,0,105,110,118,101,114, +115,101,115,113,114,116,0,1,1,0,11,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114, +101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95, +95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0, +18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,0,1,0,0,12,0,105,110,118,101,114, +115,101,115,113,114,116,0,1,1,0,12,118,0,0,0,1,4,102,108,111,97,116,95,114,115,113,0,18,95,95,114, +101,116,86,97,108,0,59,120,0,0,18,118,0,59,120,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,95, +95,114,101,116,86,97,108,0,59,121,0,0,18,118,0,59,121,0,0,0,4,102,108,111,97,116,95,114,115,113,0, +18,95,95,114,101,116,86,97,108,0,59,122,0,0,18,118,0,59,122,0,0,0,4,102,108,111,97,116,95,114,115, +113,0,18,95,95,114,101,116,86,97,108,0,59,119,0,0,18,118,0,59,119,0,0,0,0,1,0,0,9,0,110,111,114, +109,97,108,105,122,101,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,17,49,0,48,0,0,20, +0,0,1,0,0,10,0,110,111,114,109,97,108,105,122,101,0,1,1,0,10,118,0,0,0,1,3,2,1,0,9,1,115,0,2,58, +105,110,118,101,114,115,101,115,113,114,116,0,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,0,0,4, +118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, +18,118,0,0,18,115,0,0,0,0,1,0,0,11,0,110,111,114,109,97,108,105,122,101,0,1,1,0,11,118,0,0,0,1,3,2, +0,0,9,1,116,109,112,0,0,0,4,118,101,99,51,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0, +0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52, +95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0, +0,18,116,109,112,0,0,0,0,1,0,0,12,0,110,111,114,109,97,108,105,122,101,0,1,1,0,12,118,0,0,0,1,3,2, +0,0,9,1,116,109,112,0,0,0,4,118,101,99,52,95,100,111,116,0,18,116,109,112,0,0,18,118,0,0,18,118,0, +0,0,4,102,108,111,97,116,95,114,115,113,0,18,116,109,112,0,0,18,116,109,112,0,0,0,4,118,101,99,52, +95,109,117,108,116,105,112,108,121,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0, +0,18,116,109,112,0,0,0,0,1,0,0,9,0,97,98,115,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0, +18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,10,0,97,98,115,0,1,1,0,10,97,0,0,0,1,4,118, +101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,11,0, +97,98,115,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,97,98,115,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,97,0,0,0,0,1,0,0,12,0,97,98,115,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,97,98, +115,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,9,0,115,105,103,110,0,1,1,0,9,120,0,0, +0,1,3,2,0,0,9,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52,95,115,103,116,0,18,112,0,0,18,120,0,0,17,48, +0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18,110,0,0,17,48,0,48,0,0,0,18,120,0,0,0,4,118,101, +99,52,95,115,117,98,116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0, +0,1,0,0,10,0,115,105,103,110,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,112,0,0,1,1,110,0,0,0,4,118,101, +99,52,95,115,103,116,0,18,112,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95, +115,103,116,0,18,110,0,59,120,121,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,112,0,0,18,110,0,0,0,0,1,0, +0,11,0,115,105,103,110,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,112,0,0,1,1,110,0,0,0,4,118,101,99,52, +95,115,103,116,0,18,112,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115, +103,116,0,18,110,0,59,120,121,122,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98, +116,114,97,99,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,112,0,0,18,110,0,0,0,0, +1,0,0,12,0,115,105,103,110,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,112,0,0,1,1,110,0,0,0,4,118,101,99, +52,95,115,103,116,0,18,112,0,0,18,118,0,0,17,48,0,48,0,0,0,0,4,118,101,99,52,95,115,103,116,0,18, +110,0,0,17,48,0,48,0,0,0,18,118,0,0,0,4,118,101,99,52,95,115,117,98,116,114,97,99,116,0,18,95,95, +114,101,116,86,97,108,0,0,18,112,0,0,18,110,0,0,0,0,1,0,0,9,0,102,108,111,111,114,0,1,1,0,9,97,0,0, +0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0, +0,10,0,102,108,111,111,114,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95, +114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0,0,11,0,102,108,111,111,114,0,1,1,0,11,97, +0,0,0,1,4,118,101,99,52,95,102,108,111,111,114,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,97,0,0,0,0,1,0,0,12,0,102,108,111,111,114,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,102,108, +111,111,114,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,9,0,99,101,105,108,0,1,1,0,9, +97,0,0,0,1,3,2,0,0,9,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0,18, +98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54,20,0,0,1,0,0,10,0,99,101,105,108,0,1,1,0,10, +97,0,0,0,1,3,2,0,0,10,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111,111,114,0,18,98,0,0, +18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,98,0,54,20,0,0,1,0,0,11,0,99,101, +105,108,0,1,1,0,11,97,0,0,0,1,3,2,0,0,11,1,98,0,2,18,97,0,54,0,0,4,118,101,99,52,95,102,108,111, +111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,98,0,54,20, +0,0,1,0,0,12,0,99,101,105,108,0,1,1,0,12,97,0,0,0,1,3,2,0,0,12,1,98,0,2,18,97,0,54,0,0,4,118,101, +99,52,95,102,108,111,111,114,0,18,98,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,98,0,54, +20,0,0,1,0,0,9,0,102,114,97,99,116,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95, +95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,10,0,102,114,97,99,116,0,1,1,0,10,97,0,0,0,1,4, +118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,0,0,0,1,0, +0,11,0,102,114,97,99,116,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,97,0,0,0,0,1,0,0,12,0,102,114,97,99,116,0,1,1,0,12,97,0,0,0, +1,4,118,101,99,52,95,102,114,97,99,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,0,0,1,0,0,9,0, +109,111,100,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4, +102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114, +101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114, +66,0,48,0,0,48,47,20,0,0,1,0,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1, 111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, -114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,18,97,0,18,98,0,58,102, -108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,12,0,109, -111,100,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101, +114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,59,120,121,0,18,97,0,18,98,0,58,102,108, +111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,11,0,109,111, +100,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,3,2,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108, +111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116, +86,97,108,0,59,120,121,122,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79, +118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,3, +2,0,0,9,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101, +79,118,101,114,66,0,0,18,98,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108, +111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,10,0,109,111, +100,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,3,2,0,0,10,1,111,110,101,79,118,101,114,66,0,0,0,4,102, +108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,120,0,0,18,98,0,59,120,0,0,0, +4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,121,0,0,18,98,0,59,121, +0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111, +110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,11,0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,11, +98,0,0,0,1,3,2,0,0,11,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0, +18,111,110,101,79,118,101,114,66,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99, +112,0,18,111,110,101,79,118,101,114,66,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95, +114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,122,0,0,18,98,0,59,122,0,0,0,9,18,95,95,114,101, 116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66, -0,48,0,0,48,47,20,0,0,1,0,0,10,0,109,111,100,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,3,2,0,0,10,1, +0,48,0,0,48,47,20,0,0,1,0,0,12,0,109,111,100,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,3,2,0,0,12,1, 111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101, 114,66,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79, -118,101,114,66,0,59,121,0,0,18,98,0,59,121,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98, -0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,11, -0,109,111,100,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,3,2,0,0,11,1,111,110,101,79,118,101,114,66,0,0, -0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,120,0,0,18,98,0,59, -120,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,121,0,0,18,98, -0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,122,0,0, -18,98,0,59,122,0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0, -18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,12,0,109,111,100,0,1,1,0,12, -97,0,0,1,1,0,12,98,0,0,0,1,3,2,0,0,12,1,111,110,101,79,118,101,114,66,0,0,0,4,102,108,111,97,116, -95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,120,0,0,18,98,0,59,120,0,0,0,4,102,108,111, -97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102, -108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,122,0,0,18,98,0,59,122,0,0,0, -4,102,108,111,97,116,95,114,99,112,0,18,111,110,101,79,118,101,114,66,0,59,119,0,0,18,98,0,59,119, -0,0,0,9,18,95,95,114,101,116,86,97,108,0,18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111, -110,101,79,118,101,114,66,0,48,0,0,48,47,20,0,0,1,0,0,9,0,109,105,110,0,1,1,0,9,97,0,0,1,1,0,9,98, -0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0, -0,1,0,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,1,0, -0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0, -0,1,0,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18, -95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1, -1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59, -120,121,0,0,18,98,0,0,0,0,1,0,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99, -52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1, -0,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95, -95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,0,109,97,120,0,1,1,0,9,97,0,0,1,1,0,9, -98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0, -0,0,1,0,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18, -95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,1,0, -0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,11,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95, -114,101,116,86,97,108,0,59,120,121,122,0,0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0, -0,1,0,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18, -95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1, -1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120, -121,0,0,18,98,0,0,0,0,1,0,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52, -95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,0, -12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114, -101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,9,0,99,108,97,109,112,0,1,1,0,9,118,97,108,0,0, -1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108, +118,101,114,66,0,59,121,0,0,18,98,0,59,121,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,111,110, +101,79,118,101,114,66,0,59,122,0,0,18,98,0,59,122,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18, +111,110,101,79,118,101,114,66,0,59,119,0,0,18,98,0,59,119,0,0,0,9,18,95,95,114,101,116,86,97,108,0, +18,97,0,18,98,0,58,102,108,111,111,114,0,18,97,0,18,111,110,101,79,118,101,114,66,0,48,0,0,48,47, +20,0,0,1,0,0,9,0,109,105,110,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0, +18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,105,110,0,1,1,0,10,97,0, +0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,1,0,0,11,0,109,105,110,0,1,1,0,11,97,0,0,1,1, +0,11,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0,0,1,0,0,12,0,109,105,110,0,1,1,0,12,97,0, +0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0, +0,18,98,0,0,0,0,1,0,0,10,0,109,105,110,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109, +105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,11,0,109, +105,110,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116, +86,97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,0,12,0,109,105,110,0,1,1,0,12,97,0,0,1, +1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,105,110,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18, +98,0,0,0,0,1,0,0,9,0,109,97,120,0,1,1,0,9,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120, +0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0,0,0,0,1,0,0,10,0,109,97,120,0,1,1,0,10,97,0, +0,1,1,0,10,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121, +0,0,18,97,0,59,120,121,0,0,18,98,0,59,120,121,0,0,0,0,1,0,0,11,0,109,97,120,0,1,1,0,11,97,0,0,1,1, +0,11,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, +0,18,97,0,59,120,121,122,0,0,18,98,0,59,120,121,122,0,0,0,0,1,0,0,12,0,109,97,120,0,1,1,0,12,97,0, +0,1,1,0,12,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, +18,98,0,0,0,0,1,0,0,10,0,109,97,120,0,1,1,0,10,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97, +120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,59,120,121,0,0,18,98,0,0,0,0,1,0,0,11,0,109,97, +120,0,1,1,0,11,97,0,0,1,1,0,9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86, +97,108,0,0,18,97,0,59,120,121,122,0,0,18,98,0,0,0,0,1,0,0,12,0,109,97,120,0,1,1,0,12,97,0,0,1,1,0, +9,98,0,0,0,1,4,118,101,99,52,95,109,97,120,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,98,0, +0,0,0,1,0,0,9,0,99,108,97,109,112,0,1,1,0,9,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0, +9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97, +108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,10,0, +99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120, +86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118, +97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,11,0,99,108,97,109, +112,0,1,1,0,11,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0, +1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18, +109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,12,0,99,108,97,109,112,0,1,1,0,12, +118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99, +52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86, +97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0, +1,1,0,10,109,105,110,86,97,108,0,0,1,1,0,10,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108, 97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18, -109,97,120,86,97,108,0,0,0,0,1,0,0,10,0,99,108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,9,109, -105,110,86,97,108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0, +109,97,120,86,97,108,0,0,0,0,1,0,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,11,109, +105,110,86,97,108,0,0,1,1,0,11,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0, 18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86, -97,108,0,0,0,0,1,0,0,11,0,99,108,97,109,112,0,1,1,0,11,118,97,108,0,0,1,1,0,9,109,105,110,86,97, -108,0,0,1,1,0,9,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114, +97,108,0,0,0,0,1,0,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,12,109,105,110,86,97, +108,0,0,1,1,0,12,109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114, 101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0, -0,1,0,0,12,0,99,108,97,109,112,0,1,1,0,12,118,97,108,0,0,1,1,0,9,109,105,110,86,97,108,0,0,1,1,0,9, -109,97,120,86,97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108, -0,0,18,118,97,108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,10,0,99, -108,97,109,112,0,1,1,0,10,118,97,108,0,0,1,1,0,10,109,105,110,86,97,108,0,0,1,1,0,10,109,97,120,86, -97,108,0,0,0,1,4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97, -108,0,0,18,109,105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,11,0,99,108,97,109,112, -0,1,1,0,11,118,97,108,0,0,1,1,0,11,109,105,110,86,97,108,0,0,1,1,0,11,109,97,120,86,97,108,0,0,0,1, -4,118,101,99,52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109, -105,110,86,97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,12,0,99,108,97,109,112,0,1,1,0,12,118, -97,108,0,0,1,1,0,12,109,105,110,86,97,108,0,0,1,1,0,12,109,97,120,86,97,108,0,0,0,1,4,118,101,99, -52,95,99,108,97,109,112,0,18,95,95,114,101,116,86,97,108,0,0,18,118,97,108,0,0,18,109,105,110,86, -97,108,0,0,18,109,97,120,86,97,108,0,0,0,0,1,0,0,9,0,109,105,120,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0, -1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0, -18,121,0,0,18,120,0,0,0,0,1,0,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,9,97,0, -0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18, -120,0,0,0,0,1,0,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,9,97,0,0,0,1,4,118, -101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0, -1,0,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, +0,1,0,0,9,0,109,105,120,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95, 108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,10,0, -109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,108,114,112, +109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112, 0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,11,0,109,105,120,0, -1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95, -114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,12,0,109,105,120,0,1,1,0,12, -120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116, -86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,9,0,115,116,101,112,0,1,1,0,9,101,100,103, -101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0,1,1,0,10,101,100,103,101,0,0,1,1,0, -10,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101,112,0,1,1,0,11,101,100,103,101,0,0,1,1,0, -11,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0, -0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115,116,101,112,0,1,1,0,12,101,100,103,101,0,0, -1,1,0,12,120,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0, -0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,10,118, -0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0, -0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,11,118, -0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18, -118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0, -12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18, -101,100,103,101,0,0,0,0,1,0,0,9,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103, -101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,9,120,0,0,0,1,3,2,0,0,9,1,116,0,2,58,99,108,97,109, -112,0,18,120,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, -17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,10,101,100,103,101, -48,0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99,108,97,109, -112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, -17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,11,101,100,103,101, -48,0,0,1,1,0,11,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97,109, -112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, -17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,12,101,100,103,101, -48,0,0,1,1,0,12,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109, -112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0, -17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18, -116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48, -0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99,108,97,109,112, -0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17, -48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0, -48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1, -1,0,9,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97,109,112,0,18, -118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0, -48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48, -47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1, -0,9,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118, -0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0, -0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48, -0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,9,120,0,0,0,1,8,58,97,98,115,0,18,120,0,0,0,0,0,1,0, -0,9,0,108,101,110,103,116,104,0,1,1,0,10,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0,2,58, -100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18,112, -0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,114,0,0, -0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,11,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112, -0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0, -18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0, -0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0,0,0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0, -2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111,97,116,95,114,115,113,0,18,114,0,0,18, -112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1, -0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0,1,1,0,9,121,0,0,0,1,3,2,1,0,9,1,100,0,2,18, -120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,0,0,0, -20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,1,0,10, -1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116, -104,0,18,100,50,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,11,118,0,0,1,1,0,11, -117,0,0,0,1,3,2,1,0,11,1,100,51,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58, -108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0, -12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,0,12,1,100,52,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114, -101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0,0,0,20,0,0,1,0,0,11,0,99,114,111,115, -115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51,95,99,114,111,115,115,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0,0,0,1,0,0,9,0,102,97,99,101,102,111, -114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100, -0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52, -95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0, -0,18,115,0,0,0,0,0,1,0,0,10,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,10,78,0,0,1,1,0,10, -73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0, -18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0, -18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,11,0,102,97,99,101, -102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0,0,1,1,0,11,78,114,101,102,0,0,0,1,3,2,1, -0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118, -101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54, -0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,12,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,12,78,0,0, -1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101, -102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0, -48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,9,0,114,101, -102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0, -18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,10,0,114,101,102,108,101,99,116,0,1,1,0,10,73,0,0, -1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48, -47,0,0,1,0,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,0,1,8,18,73,0,17,50, -0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,12,0,114,101,102,108, -101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0, -0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,9,0,114,101,102,114,97,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78, -0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78,0, -0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48, -0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,9,1,114, -101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,17,48,0,48, -0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100, -111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118, -97,108,0,0,0,1,0,0,10,0,114,101,102,114,97,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,1,1,0,9,101, -116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78,0,0,18,73,0,0,0,0, -0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95, -100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,10,1,114,101,116,118, -97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101,99,50,0,17, -48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116,97,0,18, -110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114, -101,116,118,97,108,0,0,0,1,0,0,11,0,114,101,102,114,97,99,116,0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,1, -1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78,0,0,18, -73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0, -18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,11,1,114,101, -116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101,99, -51,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101,116, -97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0,8, -18,114,101,116,118,97,108,0,0,0,1,0,0,12,0,114,101,102,114,97,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78, -0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2,58,100,111,116,0,18,78,0, -0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18,101,116,97,0,48,17,49,0,48, -0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47,48,47,0,0,3,2,0,0,12,1,114, -101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101,116,118,97,108,0,58,118,101, -99,52,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97,0,18,73,0,48,18,101, -116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0,46,18,78,0,48,47,20,0, -8,18,114,101,116,118,97,108,0,0,0,1,0,0,13,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116, -0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97,116,50,0,18,109,0,16,8,48,0,57,18,110,0,16,8, -48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0,0,0,1,0,0,14,0,109,97,116,114, -105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1,0,0,14,110,0,0,0,1,8,58,109,97,116,51,0, -18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48, -0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,0,15,0,109,97,116,114,105,120,67, -111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0,0,0,1,8,58,109,97,116,52,0,18,109,0, -16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,18,109, -0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57,18,110,0,16,10,51,0,57,48,0,0,0, -0,1,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52, -95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3, -0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108, -116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108, -101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0, -18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97, -110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0, -7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,8,117,0, -0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0, -0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1, -0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0, -18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117, -0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0, -1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97, -108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1, -0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0, -59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,69,113,117,97,108, -0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,69,113, -117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101, -116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,0,1, -1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110, -0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86, -97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104, -97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101, -116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,0,1, -1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108, -0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59,120,121,0,0,0,0,1,0,0,3,0,103,114,101,97,116, -101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97, -116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,116,0, -18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114, -84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114, -101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118, -101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0, -0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1, -0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18, -118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0, -1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, -0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0, -1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97, -108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97, -110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95, -95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,10,117, -0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120, -121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0, -0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0, -0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99, -52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113, -117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101, -116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,7, -117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8, -118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, -0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115, -101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101, -113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114, -101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1, -1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108, -0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10, -118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, -117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0, -0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0, -0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4, -118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0, -2,0,110,111,116,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110, -101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111, -116,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, -95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69, -113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114, -101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2, -117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59, -120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1, -1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122, -0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4, -118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118, -0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52, -95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99, +1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114, +101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,12,0,109,105,120,0,1,1,0,12,120,0, +0,1,1,0,12,121,0,0,1,1,0,9,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97, +108,0,0,18,97,0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,10,0,109,105,120,0,1,1,0,10,120,0,0,1,1,0,10,121, +0,0,1,1,0,10,97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97, +0,0,18,121,0,0,18,120,0,0,0,0,1,0,0,11,0,109,105,120,0,1,1,0,11,120,0,0,1,1,0,11,121,0,0,1,1,0,11, +97,0,0,0,1,4,118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0, +0,18,120,0,0,0,0,1,0,0,12,0,109,105,120,0,1,1,0,12,120,0,0,1,1,0,12,121,0,0,1,1,0,12,97,0,0,0,1,4, +118,101,99,52,95,108,114,112,0,18,95,95,114,101,116,86,97,108,0,0,18,97,0,0,18,121,0,0,18,120,0,0, +0,0,1,0,0,9,0,115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,9,120,0,0,0,1,4,118,101,99,52,95, +115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0, +115,116,101,112,0,1,1,0,10,101,100,103,101,0,0,1,1,0,10,120,0,0,0,1,4,118,101,99,52,95,115,103,101, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0, +115,116,101,112,0,1,1,0,11,101,100,103,101,0,0,1,1,0,11,120,0,0,0,1,4,118,101,99,52,95,115,103,101, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0, +12,0,115,116,101,112,0,1,1,0,12,101,100,103,101,0,0,1,1,0,12,120,0,0,0,1,4,118,101,99,52,95,115, +103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,18,101,100,103,101,0,0,0,0,1,0,0,10,0,115, +116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,11,0,115, +116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18, +95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,12,0, +115,116,101,112,0,1,1,0,9,101,100,103,101,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101, +0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,18,101,100,103,101,0,0,0,0,1,0,0,9,0,115,109,111, +111,116,104,115,116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0, +9,120,0,0,0,1,3,2,0,0,9,1,116,0,2,58,99,108,97,109,112,0,18,120,0,18,101,100,103,101,48,0,47,18, +101,100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18, +116,0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111, +116,104,115,116,101,112,0,1,1,0,10,101,100,103,101,48,0,0,1,1,0,10,101,100,103,101,49,0,0,1,1,0,10, +118,0,0,0,1,3,2,0,0,10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101, +100,103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116, +0,18,116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116, +104,115,116,101,112,0,1,1,0,11,101,100,103,101,48,0,0,1,1,0,11,101,100,103,101,49,0,0,1,1,0,11,118, +0,0,0,1,3,2,0,0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100, +103,101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18, +116,0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104, +115,116,101,112,0,1,1,0,12,101,100,103,101,48,0,0,1,1,0,12,101,100,103,101,49,0,0,1,1,0,12,118,0,0, +0,1,3,2,0,0,12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103, +101,49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116, +0,48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,10,0,115,109,111,111,116,104,115, +116,101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,10,118,0,0,0,1,3, +2,0,0,10,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101, +49,0,18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0, +48,17,51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,11,0,115,109,111,111,116,104,115,116, +101,112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,11,118,0,0,0,1,3,2,0, +0,11,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0, +18,101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17, +51,0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,12,0,115,109,111,111,116,104,115,116,101, +112,0,1,1,0,9,101,100,103,101,48,0,0,1,1,0,9,101,100,103,101,49,0,0,1,1,0,12,118,0,0,0,1,3,2,0,0, +12,1,116,0,2,58,99,108,97,109,112,0,18,118,0,18,101,100,103,101,48,0,47,18,101,100,103,101,49,0,18, +101,100,103,101,48,0,47,49,0,17,48,0,48,0,0,0,17,49,0,48,0,0,0,0,0,0,8,18,116,0,18,116,0,48,17,51, +0,48,0,0,17,50,0,48,0,0,18,116,0,48,47,48,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,9,120,0,0, +0,1,8,58,97,98,115,0,18,120,0,0,0,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,10,118,0,0,0,1,3,2, +0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108,111, +97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95,114, +101,116,86,97,108,0,59,120,0,0,18,114,0,0,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,11,118,0,0, +0,1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102, +108,111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95, +95,114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,0,0,9,0,108,101,110,103,116,104,0,1,1,0,12,118,0,0,0, +1,3,2,0,0,9,1,114,0,0,0,3,2,1,0,9,1,112,0,2,58,100,111,116,0,18,118,0,0,18,118,0,0,0,0,0,4,102,108, +111,97,116,95,114,115,113,0,18,114,0,0,18,112,0,0,0,4,102,108,111,97,116,95,114,99,112,0,18,95,95, +114,101,116,86,97,108,0,0,18,114,0,0,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0,9,120,0,0, +1,1,0,9,121,0,0,0,1,3,2,1,0,9,1,100,0,2,18,120,0,18,121,0,47,0,0,9,18,95,95,114,101,116,86,97,108, +0,58,108,101,110,103,116,104,0,18,100,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116,97,110,99,101,0,1,1,0, +10,118,0,0,1,1,0,10,117,0,0,0,1,3,2,1,0,10,1,100,50,0,2,18,118,0,18,117,0,47,0,0,9,18,95,95,114, +101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,50,0,0,0,20,0,0,1,0,0,9,0,100,105,115,116, +97,110,99,101,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,3,2,1,0,11,1,100,51,0,2,18,118,0,18,117,0,47, +0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,51,0,0,0,20,0,0,1,0,0,9, +0,100,105,115,116,97,110,99,101,0,1,1,0,12,118,0,0,1,1,0,12,117,0,0,0,1,3,2,1,0,12,1,100,52,0,2,18, +118,0,18,117,0,47,0,0,9,18,95,95,114,101,116,86,97,108,0,58,108,101,110,103,116,104,0,18,100,52,0, +0,0,20,0,0,1,0,0,11,0,99,114,111,115,115,0,1,1,0,11,118,0,0,1,1,0,11,117,0,0,0,1,4,118,101,99,51, +95,99,114,111,115,115,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,18,117,0,0, +0,0,1,0,0,9,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,9,78,0,0,1,1,0,9,73,0,0,1,1,0,9,78, +114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2, +0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58, +109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,10,0,102,97,99,101,102,111,114,119,97, +114,100,0,1,1,0,10,78,0,0,1,1,0,10,73,0,0,1,1,0,10,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58, +100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115, +103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18, +115,0,0,0,0,0,1,0,0,11,0,102,97,99,101,102,111,114,119,97,114,100,0,1,1,0,11,78,0,0,1,1,0,11,73,0, +0,1,1,0,11,78,114,101,102,0,0,0,1,3,2,1,0,9,1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73, +0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101,99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18, +100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18,78,0,0,18,115,0,0,0,0,0,1,0,0,12,0,102,97,99,101,102, +111,114,119,97,114,100,0,1,1,0,12,78,0,0,1,1,0,12,73,0,0,1,1,0,12,78,114,101,102,0,0,0,1,3,2,1,0,9, +1,100,0,2,58,100,111,116,0,18,78,114,101,102,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,115,0,0,0,4,118,101, +99,52,95,115,103,116,0,18,115,0,0,17,48,0,48,0,0,0,18,100,0,0,0,8,58,109,105,120,0,18,78,0,54,0,18, +78,0,0,18,115,0,0,0,0,0,1,0,0,9,0,114,101,102,108,101,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,0,1,8, +18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,10,0,114, +101,102,108,101,99,116,0,1,1,0,10,73,0,0,1,1,0,10,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111, +116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,11,0,114,101,102,108,101,99,116,0,1,1,0,11, +73,0,0,1,1,0,11,78,0,0,0,1,8,18,73,0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18, +78,0,48,47,0,0,1,0,0,12,0,114,101,102,108,101,99,116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,0,1,8,18,73, +0,17,50,0,48,0,0,58,100,111,116,0,18,78,0,0,18,73,0,0,0,48,18,78,0,48,47,0,0,1,0,0,9,0,114,101,102, +114,97,99,116,0,1,1,0,9,73,0,0,1,1,0,9,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100, +111,116,95,105,0,2,58,100,111,116,0,18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0, +18,101,116,97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111, +116,95,105,0,48,47,48,47,0,0,3,2,0,0,9,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0, +40,0,9,18,114,101,116,118,97,108,0,17,48,0,48,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101,116,97, +0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107,0,0,0, +46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,10,0,114,101,102,114,97,99,116,0,1,1, +0,10,73,0,0,1,1,0,10,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105,0,2, +58,100,111,116,0,18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0,18, +101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48,47, +48,47,0,0,3,2,0,0,10,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114,101, +116,118,97,108,0,58,118,101,99,50,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18,101, +116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18,107, +0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,11,0,114,101,102,114,97,99,116, +0,1,1,0,11,73,0,0,1,1,0,11,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95,105, +0,2,58,100,111,116,0,18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116,97,0, +18,101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105,0,48, +47,48,47,0,0,3,2,0,0,11,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18,114, +101,116,118,97,108,0,58,118,101,99,51,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0,18, +101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0,18, +107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,12,0,114,101,102,114,97,99, +116,0,1,1,0,12,73,0,0,1,1,0,12,78,0,0,1,1,0,9,101,116,97,0,0,0,1,3,2,0,0,9,1,110,95,100,111,116,95, +105,0,2,58,100,111,116,0,18,78,0,0,18,73,0,0,0,0,0,3,2,0,0,9,1,107,0,2,17,49,0,48,0,0,18,101,116, +97,0,18,101,116,97,0,48,17,49,0,48,0,0,18,110,95,100,111,116,95,105,0,18,110,95,100,111,116,95,105, +0,48,47,48,47,0,0,3,2,0,0,12,1,114,101,116,118,97,108,0,0,0,10,18,107,0,17,48,0,48,0,0,40,0,9,18, +114,101,116,118,97,108,0,58,118,101,99,52,0,17,48,0,48,0,0,0,0,20,0,9,18,114,101,116,118,97,108,0, +18,101,116,97,0,18,73,0,48,18,101,116,97,0,18,110,95,100,111,116,95,105,0,48,58,115,113,114,116,0, +18,107,0,0,0,46,18,78,0,48,47,20,0,8,18,114,101,116,118,97,108,0,0,0,1,0,0,13,0,109,97,116,114,105, +120,67,111,109,112,77,117,108,116,0,1,0,0,13,109,0,0,1,0,0,13,110,0,0,0,1,8,58,109,97,116,50,0,18, +109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18,110,0,16,10,49,0,57,48,0,0, +0,0,1,0,0,14,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,14,109,0,0,1,0,0,14, +110,0,0,0,1,8,58,109,97,116,51,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10, +49,0,57,18,110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,0,0,0,1,0,0, +15,0,109,97,116,114,105,120,67,111,109,112,77,117,108,116,0,1,0,0,15,109,0,0,1,0,0,15,110,0,0,0,1, +8,58,109,97,116,52,0,18,109,0,16,8,48,0,57,18,110,0,16,8,48,0,57,48,0,18,109,0,16,10,49,0,57,18, +110,0,16,10,49,0,57,48,0,18,109,0,16,10,50,0,57,18,110,0,16,10,50,0,57,48,0,18,109,0,16,10,51,0,57, +18,110,0,16,10,51,0,57,48,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0, +10,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18, +117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0, +0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0, +0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4, +118,101,99,52,95,115,108,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0, +2,0,108,101,115,115,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101, +115,115,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115, +84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,108,116,0,18,95,95,114, +101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84,104,97,110,69,113,117, +97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115,115,84,104,97,110, +69,113,117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,108,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,108,101,115,115, +84,104,97,110,69,113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,108, +101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,108,101,115,115,84, +104,97,110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,108,101, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,108,101,115, +115,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115, +108,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0, +108,101,115,115,84,104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99, +52,95,115,108,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114, +101,97,116,101,114,84,104,97,110,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115, +103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103, +114,101,97,116,101,114,84,104,97,110,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95, +115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0, +4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99, +52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114, +101,97,116,101,114,84,104,97,110,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103, +116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,59,120,121,0,0,18,118,0,59,120,121, +0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4, +118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18, +118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0, +1,4,118,101,99,52,95,115,103,116,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1, +0,0,2,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118, +0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0, +0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,11, +117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,59, +120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,103,114,101,97,116,101,114,84,104,97,110,69, +113,117,97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95,95, +114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,103,114,101,97,116,101,114,84,104,97, +110,69,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,103,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,103,114,101,97,116, +101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95, +115,103,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0, +4,0,103,114,101,97,116,101,114,84,104,97,110,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0, +1,4,118,101,99,52,95,115,103,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1, +0,0,2,0,101,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,101,113, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117, +97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101, +116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0, +12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,6,117,0,0,1,1,0,6,118,0,0,0,1,4, +118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0, +0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0,7,118,0,0,0,1,4,118,101,99,52,95,115, +101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0, +101,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95, +114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,101,113,117,97,108,0,1,1,0,2,117,0,0, +1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0, +0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,101,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4, +118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18, +118,0,0,0,0,1,0,0,4,0,101,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95, +115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116, +69,113,117,97,108,0,1,1,0,10,117,0,0,1,1,0,10,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, +95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113, +117,97,108,0,1,1,0,11,117,0,0,1,1,0,11,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114, +101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117, +97,108,0,1,1,0,12,117,0,0,1,1,0,12,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101, +116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,6,117, +0,0,1,1,0,6,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120, +121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0,110,111,116,69,113,117,97,108,0,1,1,0,7,117,0,0,1,1,0, +7,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0, +18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111,116,69,113,117,97,108,0,1,1,0,8,117,0,0,1,1,0,8,118,0, +0,0,1,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0, +0,1,0,0,2,0,110,111,116,69,113,117,97,108,0,1,1,0,2,117,0,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95, +115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,3,0, +110,111,116,69,113,117,97,108,0,1,1,0,3,117,0,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,110,101, +0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,4,0,110,111, +116,69,113,117,97,108,0,1,1,0,4,117,0,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,110,101,0,18,95, +95,114,101,116,86,97,108,0,0,18,117,0,0,18,118,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,2,118,0,0,0,1, +3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0, +59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0, +59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,3,118,0,0, +0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18, +118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0, +18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114, +101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110, +121,0,1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117, +109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115, +117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,100, +100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99, 52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48, -0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,115,117,109,0,0,0,4,118,101, -99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118, -101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0,0,18,118,0,59,122, -0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,59,120,0,0,18,115,117,109, -0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,110,121,0,1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,115,117, -109,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,118,0,59,120,0,0,18,118,0, -59,121,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109,0,59,120,0, -0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,97,100,100,0,18,115,117,109,0,59,120,0,0,18,115,117,109, -0,59,120,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108, -0,59,120,0,0,18,115,117,109,0,59,120,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,2,118,0, -0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18, -112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0, -1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112, -108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109, -117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0, -4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0, -48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,4,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118, +0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,2,118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118, 101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0, -59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112, -114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112, -114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18, -95,95,114,101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,2,0,110,111,116, -0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121, -0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95, -115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0, -1,0,0,4,0,110,111,116,0,1,1,0,4,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116, -86,97,108,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,0,1,1,0, -16,115,97,109,112,108,101,114,0,0,1,1,0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, -120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, -114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109, -112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100, -0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59, -120,121,121,121,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115, -97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112, -49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, -100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,0,1,1,0,17,115,97,109,112,108,101,114,0,0, -1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116, -86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101, -120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, -111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0, -18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0, -116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0, -12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86, -97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120, -116,117,114,101,51,68,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0, -1,4,118,101,99,52,95,116,101,120,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, -101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111, -106,0,1,1,0,18,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, -95,116,101,120,112,51,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, -18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,67,117,98,101,0,1,1,0,19,115, -97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,99, -117,98,101,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, -114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,0,1,1,0,20,115,97,109,112,108,101,114,0,0, -1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116, +59,121,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114,111, +100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,3,118,0,0,0,1,3,2,0,0,9,1,112,114,111, +100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,118,0,59, +120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111, +100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114, +101,116,86,97,108,0,0,18,112,114,111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,1,0,97,108,108,0,1,1,0,4, +118,0,0,0,1,3,2,0,0,9,1,112,114,111,100,0,0,0,4,118,101,99,52,95,109,117,108,116,105,112,108,121,0, +18,112,114,111,100,0,0,18,118,0,59,120,0,0,18,118,0,59,121,0,0,0,4,118,101,99,52,95,109,117,108, +116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18,118,0,59,122,0,0,0,4,118, +101,99,52,95,109,117,108,116,105,112,108,121,0,18,112,114,111,100,0,0,18,112,114,111,100,0,0,18, +118,0,59,119,0,0,0,4,118,101,99,52,95,115,110,101,0,18,95,95,114,101,116,86,97,108,0,0,18,112,114, +111,100,0,0,17,48,0,48,0,0,0,0,0,1,0,0,2,0,110,111,116,0,1,1,0,2,118,0,0,0,1,4,118,101,99,52,95, +115,101,113,0,18,95,95,114,101,116,86,97,108,0,59,120,121,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0, +0,3,0,110,111,116,0,1,1,0,3,118,0,0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86, +97,108,0,59,120,121,122,0,0,18,118,0,0,17,48,0,48,0,0,0,0,0,1,0,0,4,0,110,111,116,0,1,1,0,4,118,0, +0,0,1,4,118,101,99,52,95,115,101,113,0,18,95,95,114,101,116,86,97,108,0,0,18,118,0,0,17,48,0,48,0, +0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,49,68,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1, +0,9,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97, +108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116, +117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114, +100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115, +97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,121,121,0,0,0,0,1,0,0,12,0,116,101, +120,116,117,114,101,49,68,80,114,111,106,0,1,1,0,16,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111, +111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0, +18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114, +101,50,68,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101, +99,52,95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, +18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111,106,0,1,1,0, +17,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101, +120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111, +111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,80,114,111, +106,0,1,1,0,17,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, +95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, +18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,51,68,0,1,1,0,18,115,97,109, +112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,51,100,0,18, +95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1, +0,0,12,0,116,101,120,116,117,114,101,51,68,80,114,111,106,0,1,1,0,18,115,97,109,112,108,101,114,0, +0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,51,100,0,18,95,95,114,101, +116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116, +101,120,116,117,114,101,67,117,98,101,0,1,1,0,19,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111, +111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,99,117,98,101,0,18,95,95,114,101,116,86,97,108, +0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111, +119,49,68,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101, +99,52,95,116,101,120,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0, +18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,49,68,80,114,111,106,0,1,1,0,20, +115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120, +112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111, +114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,0,1,1,0,21,115,97,109,112,108,101,114,0,0, +1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,50,100,0,18,95,95,114,101,116, 86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104, -97,100,111,119,49,68,80,114,111,106,0,1,1,0,20,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, -114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,49,100,0,18,95,95,114,101,116,86,97,108,0,0,18, -115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50, -68,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52, -95,116,101,120,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99, -111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,80,114,111,106,0,1,1,0,21,115,97, -109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50, -100,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0, -0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,0,1,1,0,22,115,97,109,112,108,101, -114,0,0,1,1,0,10,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18, +97,100,111,119,50,68,80,114,111,106,0,1,1,0,21,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111, +114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,50,100,0,18,95,95,114,101,116,86,97,108,0,0,18, +115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114, +101,50,68,82,101,99,116,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,10,99,111,111,114,100,0,0, +0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115, +97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50, +68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114, +100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108, +0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,59,120,121,122,122,0,0,0,0,1,0,0,12, +0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112,108,101, +114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0, +18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,99,111,111,114,100,0,0, +0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,0,1,1,0,23,115,97,109,112,108,101,114,0, +0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,95,114,101,99,116,0,18,95,95, +114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,12, +0,115,104,97,100,111,119,50,68,82,101,99,116,80,114,111,106,0,1,1,0,23,115,97,109,112,108,101,114, +0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101,99,116,0,18, 95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114,100,0,0,0,0,1, -0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114,111,106,0,1,1,0,22,115,97,109,112, -108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95,116,101,120,112,95,114,101, -99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0,0,18,99,111,111,114, -100,0,59,120,121,122,122,0,0,0,0,1,0,0,12,0,116,101,120,116,117,114,101,50,68,82,101,99,116,80,114, -111,106,0,1,1,0,22,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99, -52,95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, -101,114,0,0,18,99,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116, -0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,11,99,111,111,114,100,0,0,0,1,4,118,101,99,52,95, -116,101,120,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108,101,114,0, -0,18,99,111,111,114,100,0,0,0,0,1,0,0,12,0,115,104,97,100,111,119,50,68,82,101,99,116,80,114,111, -106,0,1,1,0,23,115,97,109,112,108,101,114,0,0,1,1,0,12,99,111,111,114,100,0,0,0,1,4,118,101,99,52, -95,116,101,120,112,95,114,101,99,116,0,18,95,95,114,101,116,86,97,108,0,0,18,115,97,109,112,108, -101,114,0,0,18,99,111,111,114,100,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,9,120,0,0,0,1,4, -102,108,111,97,116,95,110,111,105,115,101,49,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1, -0,0,9,0,110,111,105,115,101,49,0,1,1,0,10,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101, -50,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,11, -120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,51,0,18,95,95,114,101,116,86,97,108,0,0,18, -120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,12,120,0,0,0,1,4,102,108,111,97,116,95,110, -111,105,115,101,52,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,10,0,110,111,105,115, -101,50,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49, -0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120, -0,17,49,57,0,51,52,0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,10,120,0,0,0,1,9,18, -95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, -114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57, -0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,11,120, -0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0, -9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51, -0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,0,0,10,0,110, -111,105,115,101,50,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111, -105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115, -101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0, -0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,9,120,0,0,0,1,9,18, +0,0,9,0,110,111,105,115,101,49,0,1,1,0,9,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101, +49,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,10, +120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,50,0,18,95,95,114,101,116,86,97,108,0,0,18, +120,0,0,0,0,1,0,0,9,0,110,111,105,115,101,49,0,1,1,0,11,120,0,0,0,1,4,102,108,111,97,116,95,110, +111,105,115,101,51,0,18,95,95,114,101,116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,9,0,110,111,105,115, +101,49,0,1,1,0,12,120,0,0,0,1,4,102,108,111,97,116,95,110,111,105,115,101,52,0,18,95,95,114,101, +116,86,97,108,0,0,18,120,0,0,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,9,120,0,0,0,1,9,18,95, +95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114, +101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0, +0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, +120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, +110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0, +46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86, +97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54, +54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,0,1,0,0,10,0,110,111,105,115,101,50,0,1,1,0,12,120,0,0, +0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18, +95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17, +49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,0,1, +0,0,11,0,110,111,105,115,101,51,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0, +58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110, +111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, +59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111, +105,115,101,51,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105, +115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101, +49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95, +95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53, +0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,11, +120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, +99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,53,0, +52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105, +115,101,51,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115, +101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0, +18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0, +55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0, +18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17, +49,51,0,49,57,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,9,120,0,0,0,1,9,18, 95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, 114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,17,49,57,0,51,52,0,0,46,0,0, 20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,17,53,0,52, -55,0,0,46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101, -116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57,0,51,52,0,0,0, -17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115, -101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0,0,20,0,0,1, -0,0,11,0,110,111,105,115,101,51,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0, -58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110, -111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51, -0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49, -0,18,120,0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0, -46,0,0,20,0,0,1,0,0,11,0,110,111,105,115,101,51,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86, -97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54, -54,0,0,0,17,51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0, -59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56, -53,0,0,0,17,49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115, -101,52,0,1,1,0,9,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49, -0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120, -0,17,49,57,0,51,52,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115, -101,49,0,18,120,0,17,53,0,52,55,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110, -111,105,115,101,49,0,18,120,0,17,50,51,0,53,52,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52, -0,1,1,0,10,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18, -120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58, -118,101,99,50,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86, -97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17, -49,55,0,56,53,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115, -101,49,0,18,120,0,58,118,101,99,50,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0, -1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120, -0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110, -111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51, -0,50,51,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49, -0,18,120,0,58,118,101,99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0, -46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58, -118,101,99,51,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0, -0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59, -120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58, -110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17, -51,0,50,51,0,0,0,17,50,0,55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58, -110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17, -49,49,0,48,52,0,0,0,17,49,51,0,49,57,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119, -0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49, -0,0,0,17,51,49,0,57,49,0,0,0,17,51,55,0,52,56,0,0,0,0,46,0,0,20,0,0,0 +55,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120, +0,17,50,51,0,53,52,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115,101,52,0,1,1,0,10,120,0,0,0,1,9,18, +95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101,49,0,18,120,0,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,49,57, +0,51,52,0,0,0,17,55,0,54,54,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110, +111,105,115,101,49,0,18,120,0,58,118,101,99,50,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,0,46,0, +0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, +99,50,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105,115, +101,52,0,1,1,0,11,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115,101, +49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0,18, +120,0,58,118,101,99,51,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,0,46,0,0, +20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101, +99,51,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,0,46,0,0,20,0,9,18,95,95, +114,101,116,86,97,108,0,59,119,0,58,110,111,105,115,101,49,0,18,120,0,58,118,101,99,51,0,17,50,51, +0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57,49,0,0,0,0,46,0,0,20,0,0,1,0,0,12,0,110,111,105, +115,101,52,0,1,1,0,12,120,0,0,0,1,9,18,95,95,114,101,116,86,97,108,0,59,120,0,58,110,111,105,115, +101,49,0,18,120,0,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,121,0,58,110,111,105,115,101,49,0, +18,120,0,58,118,101,99,52,0,17,49,57,0,51,52,0,0,0,17,55,0,54,54,0,0,0,17,51,0,50,51,0,0,0,17,50,0, +55,55,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,122,0,58,110,111,105,115,101,49,0, +18,120,0,58,118,101,99,52,0,17,53,0,52,55,0,0,0,17,49,55,0,56,53,0,0,0,17,49,49,0,48,52,0,0,0,17, +49,51,0,49,57,0,0,0,0,46,0,0,20,0,9,18,95,95,114,101,116,86,97,108,0,59,119,0,58,110,111,105,115, +101,49,0,18,120,0,58,118,101,99,52,0,17,50,51,0,53,52,0,0,0,17,50,57,0,49,49,0,0,0,17,51,49,0,57, +49,0,0,0,17,51,55,0,52,56,0,0,0,0,46,0,0,20,0,0,0 -- cgit v1.2.3 From 1c2ff4d9e65563c071747a9c3bd907bd24706da0 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 15:16:43 -0600 Subject: gallium: use new util_fast_exp2(), _log2(), pow() functions New code surrounded with #if FAST_MATH to allow comparing against original code if we need to debug. --- src/gallium/auxiliary/draw/draw_vs_aos.c | 66 +++++++++++++++++++++++++++++++- src/gallium/auxiliary/tgsi/tgsi_exec.c | 30 +++++++++++++++ src/gallium/auxiliary/tgsi/tgsi_sse2.c | 19 +++++++++ 3 files changed, 113 insertions(+), 2 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 441877d46f..41bdd012d5 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -31,6 +31,7 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" +#include "util/u_math.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi/tgsi_exec.h" @@ -43,6 +44,7 @@ #ifdef PIPE_ARCH_X86 #define DISASSEM 0 +#define FAST_MATH 1 static const char *files[] = { @@ -1380,14 +1382,28 @@ static boolean emit_MAD( struct aos_compilation *cp, const struct tgsi_full_inst return TRUE; } + + /* A wrapper for powf(). * Makes sure it is cdecl and operates on floats. */ static float PIPE_CDECL _powerf( float x, float y ) { +#if FAST_MATH + return util_fast_pow(x, y); +#else return powf( x, y ); +#endif } +#if FAST_MATH +static float PIPE_CDECL _exp2(float x) +{ + return util_fast_exp2(x); +} +#endif + + /* Really not sufficient -- need to check for conditions that could * generate inf/nan values, which will slow things down hugely. */ @@ -1442,6 +1458,48 @@ static boolean emit_POW( struct aos_compilation *cp, const struct tgsi_full_inst } +#if FAST_MATH +static boolean emit_EXPBASE2( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) +{ + uint i; + + /* For absolute correctness, need to spill/invalidate all XMM regs + * too. + */ + for (i = 0; i < 8; i++) { + if (cp->xmm[i].dirty) + spill(cp, i); + aos_release_xmm_reg(cp, i); + } + + /* Push caller-save (ie scratch) regs. + */ + x86_cdecl_caller_push_regs( cp->func ); + + x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, -4) ); + + x87_fld_src( cp, &op->FullSrcRegisters[0], 0 ); + x87_fstp( cp->func, x86_make_disp( cp->stack_ESP, 0 ) ); + + /* tmp_EAX has been pushed & will be restored below */ + x86_mov_reg_imm( cp->func, cp->tmp_EAX, (unsigned long) _exp2 ); + x86_call( cp->func, cp->tmp_EAX ); + + x86_lea( cp->func, cp->stack_ESP, x86_make_disp(cp->stack_ESP, 4) ); + + x86_cdecl_caller_pop_regs( cp->func ); + + /* Note retval on x87 stack: + */ + cp->func->x87_stack++; + + x87_fstp_dest4( cp, &op->FullDstRegisters[0] ); + + return TRUE; +} +#endif + + static boolean emit_RCP( struct aos_compilation *cp, const struct tgsi_full_instruction *op ) { struct x86_reg arg0 = fetch_src(cp, &op->FullSrcRegisters[0]); @@ -1662,7 +1720,9 @@ emit_instruction( struct aos_compilation *cp, return emit_RND(cp, inst); case TGSI_OPCODE_EXPBASE2: -#if 0 +#if FAST_MATH + return emit_EXPBASE2(cp, inst); +#elif 0 /* this seems to fail for "larger" exponents. * See glean tvertProg1's EX2 test. */ @@ -1827,6 +1887,8 @@ static boolean build_vertex_program( struct draw_vs_varient_aos_sse *varient, struct aos_compilation cp; unsigned fixup, label; + util_init_math(); + tgsi_parse_init( &parse, varient->base.vs->state.tokens ); memset(&cp, 0, sizeof(cp)); @@ -2135,4 +2197,4 @@ struct draw_vs_varient *draw_vs_varient_aos_sse( struct draw_vertex_shader *vs, -#endif +#endif /* PIPE_ARCH_X86 */ diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index 88a34a6961..e28b56c842 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -57,6 +57,9 @@ #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi_exec.h" +#include "util/u_math.h" + +#define FAST_MATH 1 #define TILE_TOP_LEFT 0 #define TILE_TOP_RIGHT 1 @@ -145,6 +148,8 @@ tgsi_exec_machine_bind_shader( tgsi_dump(tokens, 0); #endif + util_init_math(); + mach->Tokens = tokens; mach->Samplers = samplers; @@ -448,10 +453,17 @@ micro_exp2( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src) { +#if FAST_MATH + dst->f[0] = util_fast_exp2( src->f[0] ); + dst->f[1] = util_fast_exp2( src->f[1] ); + dst->f[2] = util_fast_exp2( src->f[2] ); + dst->f[3] = util_fast_exp2( src->f[3] ); +#else dst->f[0] = powf( 2.0f, src->f[0] ); dst->f[1] = powf( 2.0f, src->f[1] ); dst->f[2] = powf( 2.0f, src->f[2] ); dst->f[3] = powf( 2.0f, src->f[3] ); +#endif } static void @@ -528,10 +540,17 @@ micro_lg2( union tgsi_exec_channel *dst, const union tgsi_exec_channel *src ) { +#if FAST_MATH + dst->f[0] = util_fast_log2( src->f[0] ); + dst->f[1] = util_fast_log2( src->f[1] ); + dst->f[2] = util_fast_log2( src->f[2] ); + dst->f[3] = util_fast_log2( src->f[3] ); +#else dst->f[0] = logf( src->f[0] ) * 1.442695f; dst->f[1] = logf( src->f[1] ) * 1.442695f; dst->f[2] = logf( src->f[2] ) * 1.442695f; dst->f[3] = logf( src->f[3] ) * 1.442695f; +#endif } static void @@ -796,10 +815,17 @@ micro_pow( const union tgsi_exec_channel *src0, const union tgsi_exec_channel *src1 ) { +#if FAST_MATH + dst->f[0] = util_fast_pow( src0->f[0], src1->f[0] ); + dst->f[1] = util_fast_pow( src0->f[1], src1->f[1] ); + dst->f[2] = util_fast_pow( src0->f[2], src1->f[2] ); + dst->f[3] = util_fast_pow( src0->f[3], src1->f[3] ); +#else dst->f[0] = powf( src0->f[0], src1->f[0] ); dst->f[1] = powf( src0->f[1], src1->f[1] ); dst->f[2] = powf( src0->f[2], src1->f[2] ); dst->f[3] = powf( src0->f[3], src1->f[3] ); +#endif } static void @@ -2024,7 +2050,11 @@ exec_instruction( /* TGSI_OPCODE_EX2 */ FETCH(&r[0], 0, CHAN_X); +#if FAST_MATH + micro_exp2( &r[0], &r[0] ); +#else micro_pow( &r[0], &mach->Temps[TEMP_2_I].xyzw[TEMP_2_C], &r[0] ); +#endif FOR_EACH_ENABLED_CHANNEL( *inst, chan_index ) { STORE( &r[0], 0, chan_index ); diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 485e5a0e6f..e390607023 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -27,6 +27,7 @@ #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" +#include "util/u_math.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi_exec.h" @@ -42,6 +43,8 @@ */ #define HIGH_PRECISION 1 +#define FAST_MATH 1 + #define FOR_EACH_CHANNEL( CHAN )\ for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++) @@ -623,10 +626,17 @@ ex24f( { const unsigned X = 0; +#if FAST_MATH + store[X + 0] = util_fast_exp2( store[X + 0] ); + store[X + 1] = util_fast_exp2( store[X + 1] ); + store[X + 2] = util_fast_exp2( store[X + 2] ); + store[X + 3] = util_fast_exp2( store[X + 3] ); +#else store[X + 0] = powf( 2.0f, store[X + 0] ); store[X + 1] = powf( 2.0f, store[X + 1] ); store[X + 2] = powf( 2.0f, store[X + 2] ); store[X + 3] = powf( 2.0f, store[X + 3] ); +#endif } static void @@ -762,10 +772,17 @@ pow4f( { const unsigned X = 0; +#if FAST_MATH + store[X + 0] = util_fast_pow( store[X + 0], store[X + 4] ); + store[X + 1] = util_fast_pow( store[X + 1], store[X + 5] ); + store[X + 2] = util_fast_pow( store[X + 2], store[X + 6] ); + store[X + 3] = util_fast_pow( store[X + 3], store[X + 7] ); +#else store[X + 0] = powf( store[X + 0], store[X + 4] ); store[X + 1] = powf( store[X + 1], store[X + 5] ); store[X + 2] = powf( store[X + 2], store[X + 6] ); store[X + 3] = powf( store[X + 3], store[X + 7] ); +#endif } static void @@ -2235,6 +2252,8 @@ tgsi_emit_sse2( unsigned ok = 1; uint num_immediates = 0; + util_init_math(); + func->csr = func->store; tgsi_parse_init( &parse, tokens ); -- cgit v1.2.3 From 1a46dcc8a927dfb38ca1381e7b3dafb789f8257c Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 15:25:21 -0600 Subject: gallium: replace LOG2() macro with util_fast_log2() inline func --- src/gallium/auxiliary/tgsi/tgsi_sse2.c | 8 ++++---- src/gallium/drivers/softpipe/sp_context.c | 3 +++ src/gallium/drivers/softpipe/sp_tex_sample.c | 3 ++- src/gallium/include/pipe/p_util.h | 15 --------------- 4 files changed, 9 insertions(+), 20 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index e390607023..00ed4da450 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -713,10 +713,10 @@ lg24f( { const unsigned X = 0; - store[X + 0] = LOG2( store[X + 0] ); - store[X + 1] = LOG2( store[X + 1] ); - store[X + 2] = LOG2( store[X + 2] ); - store[X + 3] = LOG2( store[X + 3] ); + store[X + 0] = util_fast_log2( store[X + 0] ); + store[X + 1] = util_fast_log2( store[X + 1] ); + store[X + 2] = util_fast_log2( store[X + 2] ); + store[X + 3] = util_fast_log2( store[X + 3] ); } static void diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 626c3a9d4e..9b1313bc83 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -33,6 +33,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/p_util.h" +#include "util/u_math.h" #include "sp_clear.h" #include "sp_context.h" #include "sp_flush.h" @@ -128,6 +129,8 @@ softpipe_create( struct pipe_screen *screen, struct softpipe_context *softpipe = CALLOC_STRUCT(softpipe_context); uint i; + util_init_math(); + #ifdef PIPE_ARCH_X86 softpipe->use_sse = !debug_get_bool_option( "GALLIUM_NOSSE", FALSE ); #else diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 01f4d0ca81..d7680ffa81 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -41,6 +41,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "tgsi/tgsi_exec.h" +#include "util/u_math.h" /* @@ -499,7 +500,7 @@ compute_lambda(struct tgsi_sampler *sampler, rho = MAX2(rho, max); } - lambda = LOG2(rho); + lambda = util_fast_log2(rho); lambda += lodbias + sampler->state->lod_bias; lambda = CLAMP(lambda, sampler->state->min_lod, sampler->state->max_lod); diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 473a8d94ab..660192b28e 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -443,21 +443,6 @@ static INLINE int iround(float f) #define FABSF(x) ((float) fabs(x)) #endif -/* Pretty fast, and accurate. - * Based on code from http://www.flipcode.com/totd/ - */ -static INLINE float LOG2(float val) -{ - union fi num; - int log_2; - - num.f = val; - log_2 = ((num.i >> 23) & 255) - 128; - num.i &= ~(255 << 23); - num.i += 127 << 23; - num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3; - return num.f + log_2; -} #if defined(__GNUC__) #define CEILF(x) ceilf(x) -- cgit v1.2.3 From 120270def74149b240926f827b80ca073536d462 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 15:49:36 -0600 Subject: gallium: stop using FABSF() macro --- src/gallium/auxiliary/draw/draw_pipe_offset.c | 4 ++-- src/gallium/auxiliary/draw/draw_pipe_wide_line.c | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c index 8f1650e55c..2f5865741c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_offset.c +++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c @@ -84,8 +84,8 @@ static void do_offset_tri( struct draw_stage *stage, float a = ey*fz - ez*fy; float b = ez*fx - ex*fz; - float dzdx = FABSF(a * inv_det); - float dzdy = FABSF(b * inv_det); + float dzdx = fabsf(a * inv_det); + float dzdy = fabsf(b * inv_det); float zoffset = offset->units + MAX2(dzdx, dzdy) * offset->scale; diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c index 29649f5787..48ec2f1239 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c @@ -73,8 +73,8 @@ static void wideline_line( struct draw_stage *stage, float *pos2 = v2->data[pos]; float *pos3 = v3->data[pos]; - const float dx = FABSF(pos0[0] - pos2[0]); - const float dy = FABSF(pos0[1] - pos2[1]); + const float dx = fabsf(pos0[0] - pos2[0]); + const float dy = fabsf(pos0[1] - pos2[1]); /* small tweak to meet GL specification */ const float bias = 0.125f; -- cgit v1.2.3 From 9935e3b7303da656e258d4bd5bc799ffbfbc737b Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 15:51:02 -0600 Subject: gallium: stop using ifloor(), FABSF(), etc --- src/gallium/drivers/softpipe/sp_setup.c | 14 +++--- src/gallium/drivers/softpipe/sp_tex_sample.c | 72 ++++++++++++++-------------- 2 files changed, 44 insertions(+), 42 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index dd0562e7ef..c8c55fa6e8 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -44,6 +44,8 @@ #include "draw/draw_vertex.h" #include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" +#include "util/u_math.h" + #define DEBUG_VERTS 0 #define DEBUG_FRAGS 0 @@ -611,18 +613,18 @@ static void setup_tri_edges( struct setup_context *setup ) float vmid_y = setup->vmid[0][1] - 0.5f; float vmax_y = setup->vmax[0][1] - 0.5f; - setup->emaj.sy = CEILF(vmin_y); - setup->emaj.lines = (int) CEILF(vmax_y - setup->emaj.sy); + setup->emaj.sy = ceilf(vmin_y); + setup->emaj.lines = (int) ceilf(vmax_y - setup->emaj.sy); setup->emaj.dxdy = setup->emaj.dx / setup->emaj.dy; setup->emaj.sx = vmin_x + (setup->emaj.sy - vmin_y) * setup->emaj.dxdy; - setup->etop.sy = CEILF(vmid_y); - setup->etop.lines = (int) CEILF(vmax_y - setup->etop.sy); + setup->etop.sy = ceilf(vmid_y); + setup->etop.lines = (int) ceilf(vmax_y - setup->etop.sy); setup->etop.dxdy = setup->etop.dx / setup->etop.dy; setup->etop.sx = vmid_x + (setup->etop.sy - vmid_y) * setup->etop.dxdy; - setup->ebot.sy = CEILF(vmin_y); - setup->ebot.lines = (int) CEILF(vmid_y - setup->ebot.sy); + setup->ebot.sy = ceilf(vmin_y); + setup->ebot.lines = (int) ceilf(vmid_y - setup->ebot.sy); setup->ebot.dxdy = setup->ebot.dx / setup->ebot.dy; setup->ebot.sx = vmin_x + (setup->ebot.sy - vmin_y) * setup->ebot.dxdy; } diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index d7680ffa81..58a95d13e1 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -51,7 +51,7 @@ * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0. * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x). */ -#define FRAC(f) ((f) - ifloor(f)) +#define FRAC(f) ((f) - util_ifloor(f)) /** @@ -100,7 +100,7 @@ nearest_texcoord(unsigned wrapMode, float s, unsigned size) case PIPE_TEX_WRAP_REPEAT: /* s limited to [0,1) */ /* i limited to [0,size-1] */ - i = ifloor(s * size); + i = util_ifloor(s * size); i = REMAINDER(i, size); return i; case PIPE_TEX_WRAP_CLAMP: @@ -111,7 +111,7 @@ nearest_texcoord(unsigned wrapMode, float s, unsigned size) else if (s >= 1.0F) i = size - 1; else - i = ifloor(s * size); + i = util_ifloor(s * size); return i; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: { @@ -124,7 +124,7 @@ nearest_texcoord(unsigned wrapMode, float s, unsigned size) else if (s > max) i = size - 1; else - i = ifloor(s * size); + i = util_ifloor(s * size); } return i; case PIPE_TEX_WRAP_CLAMP_TO_BORDER: @@ -138,14 +138,14 @@ nearest_texcoord(unsigned wrapMode, float s, unsigned size) else if (s >= max) i = size; else - i = ifloor(s * size); + i = util_ifloor(s * size); } return i; case PIPE_TEX_WRAP_MIRROR_REPEAT: { const float min = 1.0F / (2.0F * size); const float max = 1.0F - min; - const int flr = ifloor(s); + const int flr = util_ifloor(s); float u; if (flr & 1) u = 1.0F - (s - (float) flr); @@ -156,20 +156,20 @@ nearest_texcoord(unsigned wrapMode, float s, unsigned size) else if (u > max) i = size - 1; else - i = ifloor(u * size); + i = util_ifloor(u * size); } return i; case PIPE_TEX_WRAP_MIRROR_CLAMP: { /* s limited to [0,1] */ /* i limited to [0,size-1] */ - const float u = FABSF(s); + const float u = fabsf(s); if (u <= 0.0F) i = 0; else if (u >= 1.0F) i = size - 1; else - i = ifloor(u * size); + i = util_ifloor(u * size); } return i; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: @@ -178,13 +178,13 @@ nearest_texcoord(unsigned wrapMode, float s, unsigned size) /* i limited to [0, size-1] */ const float min = 1.0F / (2.0F * size); const float max = 1.0F - min; - const float u = FABSF(s); + const float u = fabsf(s); if (u < min) i = 0; else if (u > max) i = size - 1; else - i = ifloor(u * size); + i = util_ifloor(u * size); } return i; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER: @@ -193,13 +193,13 @@ nearest_texcoord(unsigned wrapMode, float s, unsigned size) /* i limited to [0, size-1] */ const float min = -1.0F / (2.0F * size); const float max = 1.0F - min; - const float u = FABSF(s); + const float u = fabsf(s); if (u < min) i = -1; else if (u > max) i = size; else - i = ifloor(u * size); + i = util_ifloor(u * size); } return i; default: @@ -226,7 +226,7 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, switch (wrapMode) { case PIPE_TEX_WRAP_REPEAT: u = s * size - 0.5F; - *i0 = REMAINDER(ifloor(u), size); + *i0 = REMAINDER(util_ifloor(u), size); *i1 = REMAINDER(*i0 + 1, size); break; case PIPE_TEX_WRAP_CLAMP: @@ -237,7 +237,7 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, else u = s * size; u -= 0.5F; - *i0 = ifloor(u); + *i0 = util_ifloor(u); *i1 = *i0 + 1; break; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: @@ -248,7 +248,7 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, else u = s * size; u -= 0.5F; - *i0 = ifloor(u); + *i0 = util_ifloor(u); *i1 = *i0 + 1; if (*i0 < 0) *i0 = 0; @@ -266,19 +266,19 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, else u = s * size; u -= 0.5F; - *i0 = ifloor(u); + *i0 = util_ifloor(u); *i1 = *i0 + 1; } break; case PIPE_TEX_WRAP_MIRROR_REPEAT: { - const int flr = ifloor(s); + const int flr = util_ifloor(s); if (flr & 1) u = 1.0F - (s - (float) flr); else u = s - (float) flr; u = (u * size) - 0.5F; - *i0 = ifloor(u); + *i0 = util_ifloor(u); *i1 = *i0 + 1; if (*i0 < 0) *i0 = 0; @@ -287,23 +287,23 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, } break; case PIPE_TEX_WRAP_MIRROR_CLAMP: - u = FABSF(s); + u = fabsf(s); if (u >= 1.0F) u = (float) size; else u *= size; u -= 0.5F; - *i0 = ifloor(u); + *i0 = util_ifloor(u); *i1 = *i0 + 1; break; case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE: - u = FABSF(s); + u = fabsf(s); if (u >= 1.0F) u = (float) size; else u *= size; u -= 0.5F; - *i0 = ifloor(u); + *i0 = util_ifloor(u); *i1 = *i0 + 1; if (*i0 < 0) *i0 = 0; @@ -314,7 +314,7 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, { const float min = -1.0F / (2.0F * size); const float max = 1.0F - min; - u = FABSF(s); + u = fabsf(s); if (u <= min) u = min * size; else if (u >= max) @@ -322,7 +322,7 @@ linear_texcoord(unsigned wrapMode, float s, unsigned size, else u *= size; u -= 0.5F; - *i0 = ifloor(u); + *i0 = util_ifloor(u); *i1 = *i0 + 1; } break; @@ -343,12 +343,12 @@ nearest_texcoord_unnorm(unsigned wrapMode, float s, unsigned size) int i; switch (wrapMode) { case PIPE_TEX_WRAP_CLAMP: - i = ifloor(s); + i = util_ifloor(s); return CLAMP(i, 0, (int) size-1); case PIPE_TEX_WRAP_CLAMP_TO_EDGE: /* fall-through */ case PIPE_TEX_WRAP_CLAMP_TO_BORDER: - return ifloor( CLAMP(s, 0.5F, (float) size - 0.5F) ); + return util_ifloor( CLAMP(s, 0.5F, (float) size - 0.5F) ); default: assert(0); return 0; @@ -368,7 +368,7 @@ linear_texcoord_unnorm(unsigned wrapMode, float s, unsigned size, case PIPE_TEX_WRAP_CLAMP: /* Not exactly what the spec says, but it matches NVIDIA output */ s = CLAMP(s - 0.5F, 0.0f, (float) size - 1.0f); - *i0 = ifloor(s); + *i0 = util_ifloor(s); *i1 = *i0 + 1; break; case PIPE_TEX_WRAP_CLAMP_TO_EDGE: @@ -376,7 +376,7 @@ linear_texcoord_unnorm(unsigned wrapMode, float s, unsigned size, case PIPE_TEX_WRAP_CLAMP_TO_BORDER: s = CLAMP(s, 0.5F, (float) size - 0.5F); s -= 0.5F; - *i0 = ifloor(s); + *i0 = util_ifloor(s); *i1 = *i0 + 1; if (*i1 > (int) size - 1) *i1 = size - 1; @@ -402,7 +402,7 @@ choose_cube_face(float rx, float ry, float rz, float *newS, float *newT) +rz TEXTURE_CUBE_MAP_POSITIVE_Z_EXT +rx -ry rz -rz TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT -rx -ry rz */ - const float arx = FABSF(rx), ary = FABSF(ry), arz = FABSF(rz); + const float arx = fabsf(rx), ary = fabsf(ry), arz = fabsf(rz); unsigned face; float sc, tc, ma; @@ -477,16 +477,16 @@ compute_lambda(struct tgsi_sampler *sampler, { float dsdx = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT]; float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT]; - dsdx = FABSF(dsdx); - dsdy = FABSF(dsdy); + dsdx = fabsf(dsdx); + dsdy = fabsf(dsdy); rho = MAX2(dsdx, dsdy) * sampler->texture->width[0]; } if (t) { float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT]; float dtdy = t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT]; float max; - dtdx = FABSF(dtdx); - dtdy = FABSF(dtdy); + dtdx = fabsf(dtdx); + dtdy = fabsf(dtdy); max = MAX2(dtdx, dtdy) * sampler->texture->height[0]; rho = MAX2(rho, max); } @@ -494,8 +494,8 @@ compute_lambda(struct tgsi_sampler *sampler, float dpdx = p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT]; float dpdy = p[QUAD_TOP_LEFT] - p[QUAD_BOTTOM_LEFT]; float max; - dpdx = FABSF(dpdx); - dpdy = FABSF(dpdy); + dpdx = fabsf(dpdx); + dpdy = fabsf(dpdy); max = MAX2(dpdx, dpdy) * sampler->texture->depth[0]; rho = MAX2(rho, max); } -- cgit v1.2.3 From a22bdd42d77bc858814f37d657fa940a520dbe56 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 15:51:38 -0600 Subject: gallium: move math macros from p_util.h to u_math.h More can be done... --- src/gallium/auxiliary/util/u_math.h | 49 +++++++++++++++++++++++++++++++++++ src/gallium/include/pipe/p_util.h | 51 ------------------------------------- 2 files changed, 49 insertions(+), 51 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h index e179ce700f..09c55e437f 100644 --- a/src/gallium/auxiliary/util/u_math.h +++ b/src/gallium/auxiliary/util/u_math.h @@ -29,6 +29,9 @@ /** * Math utilities and approximations for common math functions. * Reduced precision is usually acceptable in shaders... + * + * "fast" is used in the names of functions which are low-precision, + * or at least lower-precision than the normal C lib functions. */ @@ -36,6 +39,7 @@ #define U_MATH_H +#include "pipe/p_compiler.h" #include "pipe/p_util.h" #include "util/u_math.h" @@ -141,4 +145,49 @@ util_fast_pow(float x, float y) } + +/** + * Floor(x), returned as int. + */ +static INLINE int +util_ifloor(float f) +{ + int ai, bi; + double af, bf; + union fi u; + af = (3 << 22) + 0.5 + (double)f; + bf = (3 << 22) + 0.5 - (double)f; + u.f = (float) af; ai = u.i; + u.f = (float) bf; bi = u.i; + return (ai - bi) >> 1; +} + + +/** + * Round float to nearest int. + */ +static INLINE int +util_iround(float f) +{ +#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86) + int r; + __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st"); + return r; +#elif defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86) + int r; + _asm { + fld f + fistp r + } + return r; +#else + if (f >= 0.0f) + return (int) (f + 0.5f); + else + return (int) (f - 0.5f); +#endif +} + + + #endif /* U_MATH_H */ diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 660192b28e..8f5cb4ddc5 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -399,57 +399,6 @@ do { \ } while (0) -static INLINE int ifloor(float f) -{ - int ai, bi; - double af, bf; - union fi u; - - af = (3 << 22) + 0.5 + (double)f; - bf = (3 << 22) + 0.5 - (double)f; - u.f = (float) af; ai = u.i; - u.f = (float) bf; bi = u.i; - return (ai - bi) >> 1; -} - - -#if defined(PIPE_CC_GCC) && defined(PIPE_ARCH_X86) -static INLINE int iround(float f) -{ - int r; - __asm__ ("fistpl %0" : "=m" (r) : "t" (f) : "st"); - return r; -} -#elif defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86) -static INLINE int iround(float f) -{ - int r; - _asm { - fld f - fistp r - } - return r; -} -#else -#define IROUND(f) ((int) (((f) >= 0.0F) ? ((f) + 0.5F) : ((f) - 0.5F))) -#endif - - -/* Could maybe have an inline version of this? - */ -#if defined(__GNUC__) -#define FABSF(x) fabsf(x) -#else -#define FABSF(x) ((float) fabs(x)) -#endif - - -#if defined(__GNUC__) -#define CEILF(x) ceilf(x) -#else -#define CEILF(x) ((float) ceil(x)) -#endif - static INLINE int align(int value, int alignment) { return (value + alignment - 1) & ~(alignment - 1); -- cgit v1.2.3 From f9c04d55d045b3cf3ffef24f805166e3995096e2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 15:53:28 -0600 Subject: gallium: insert __cplusplus/extern wrappings --- src/gallium/auxiliary/util/u_math.h | 9 +++++++++ 1 file changed, 9 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h index 09c55e437f..a541d30a5d 100644 --- a/src/gallium/auxiliary/util/u_math.h +++ b/src/gallium/auxiliary/util/u_math.h @@ -44,12 +44,17 @@ #include "util/u_math.h" +#ifdef __cplusplus +extern "C" { +#endif + #define POW2_TABLE_SIZE 256 #define POW2_TABLE_SCALE ((float) (POW2_TABLE_SIZE-1)) extern float pow2_table[POW2_TABLE_SIZE]; + extern void util_init_math(void); @@ -190,4 +195,8 @@ util_iround(float f) +#ifdef __cplusplus +} +#endif + #endif /* U_MATH_H */ -- cgit v1.2.3 From a13475ff0057f1de8e3bc08d6ca42b9e135a3f5a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 16:09:37 -0600 Subject: gallium: replace align_int() with align() The two functions are identical. Removed align_int() from p_util.h --- src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 6 +++--- src/gallium/drivers/i915simple/i915_texture.c | 8 ++++---- src/gallium/drivers/i965simple/brw_tex_layout.c | 8 ++++---- src/gallium/include/pipe/p_util.h | 11 +++-------- 4 files changed, 14 insertions(+), 19 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index 0aec4b71ba..be3535ed9e 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -119,7 +119,7 @@ static void fetch_pipeline_run( struct draw_pt_middle_end *middle, struct draw_context *draw = fpme->draw; struct draw_vertex_shader *shader = draw->vs.vertex_shader; unsigned opt = fpme->opt; - unsigned alloc_count = align_int( fetch_count, 4 ); + unsigned alloc_count = align( fetch_count, 4 ); struct vertex_header *pipeline_verts = (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count); @@ -195,7 +195,7 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, struct draw_context *draw = fpme->draw; struct draw_vertex_shader *shader = draw->vs.vertex_shader; unsigned opt = fpme->opt; - unsigned alloc_count = align_int( count, 4 ); + unsigned alloc_count = align( count, 4 ); struct vertex_header *pipeline_verts = (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count); @@ -271,7 +271,7 @@ static void fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle, struct draw_context *draw = fpme->draw; struct draw_vertex_shader *shader = draw->vs.vertex_shader; unsigned opt = fpme->opt; - unsigned alloc_count = align_int( count, 4 ); + unsigned alloc_count = align( count, 4 ); struct vertex_header *pipeline_verts = (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count); diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index cf4964b26b..ca0fb8761b 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -220,7 +220,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) */ if (pt->last_level > 0) { unsigned mip1_nblocksx - = align_int(pf_get_nblocksx(&pt->block, minify(width)), align_x) + = align(pf_get_nblocksx(&pt->block, minify(width)), align_x) + pf_get_nblocksx(&pt->block, minify(minify(width))); if (mip1_nblocksx > nblocksx) @@ -229,14 +229,14 @@ i945_miptree_layout_2d( struct i915_texture *tex ) /* Pitch must be a whole number of dwords */ - tex->stride = align_int(tex->stride, 64); + tex->stride = align(tex->stride, 64); tex->total_nblocksy = 0; for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_level_info(tex, level, 1, width, height, 1); i915_miptree_set_image_offset(tex, level, 0, x, y); - nblocksy = align_int(nblocksy, align_y); + nblocksy = align(nblocksy, align_y); /* Because the images are packed better, the final offset * might not be the maximal one: @@ -246,7 +246,7 @@ i945_miptree_layout_2d( struct i915_texture *tex ) /* Layout_below: step right after second mipmap level. */ if (level == 1) { - x += align_int(nblocksx, align_x); + x += align(nblocksx, align_x); } else { y += nblocksy; diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 8c7725605b..9b6cf81723 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -144,7 +144,7 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) */ if (pt->last_level > 0) { unsigned mip1_nblocksx - = align_int(pf_get_nblocksx(&pt->block, minify(width)), align_x) + = align(pf_get_nblocksx(&pt->block, minify(width)), align_x) + pf_get_nblocksx(&pt->block, minify(minify(width))); if (mip1_nblocksx > nblocksx) @@ -153,14 +153,14 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) /* Pitch must be a whole number of dwords */ - tex->stride = align_int(tex->stride, 64); + tex->stride = align(tex->stride, 64); tex->total_nblocksy = 0; for (level = 0; level <= pt->last_level; level++) { intel_miptree_set_level_info(tex, level, 1, x, y, width, height, 1); - nblocksy = align_int(nblocksy, align_y); + nblocksy = align(nblocksy, align_y); /* Because the images are packed better, the final offset * might not be the maximal one: @@ -170,7 +170,7 @@ static void i945_miptree_layout_2d(struct brw_texture *tex) /* Layout_below: step right after second mipmap level. */ if (level == 1) { - x += align_int(nblocksx, align_x); + x += align(nblocksx, align_x); } else { y += nblocksy; diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index 8f5cb4ddc5..cac0039e12 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -289,13 +289,14 @@ align16( void *unaligned ) } -static INLINE int align_int(int x, int align) +static INLINE int align(int value, int alignment) { - return (x + align - 1) & ~(align - 1); + return (value + alignment - 1) & ~(alignment - 1); } + #if defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86) static INLINE unsigned ffs( unsigned u ) { @@ -399,12 +400,6 @@ do { \ } while (0) -static INLINE int align(int value, int alignment) -{ - return (value + alignment - 1) & ~(alignment - 1); -} - - /* util/p_util.c */ extern void pipe_copy_rect(ubyte * dst, const struct pipe_format_block *block, -- cgit v1.2.3 From fe1e39afbb147deab60ecc932c24f921b46f1364 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 16:19:22 -0600 Subject: gallium: move pipe_copy_rect(), pipe_fill_rect() protos into new u_rect.h header --- src/gallium/auxiliary/util/p_util.c | 1 + src/gallium/auxiliary/util/u_rect.h | 54 +++++++++++++++++++++++++++ src/gallium/drivers/cell/ppu/cell_surface.c | 1 + src/gallium/drivers/i915simple/i915_surface.c | 1 + src/gallium/drivers/i965simple/brw_surface.c | 1 + src/gallium/drivers/softpipe/sp_surface.c | 1 + src/gallium/include/pipe/p_util.h | 13 ------- src/mesa/state_tracker/st_texture.c | 1 + 8 files changed, 60 insertions(+), 13 deletions(-) create mode 100644 src/gallium/auxiliary/util/u_rect.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_util.c b/src/gallium/auxiliary/util/p_util.c index 271be4edf1..787881b192 100644 --- a/src/gallium/auxiliary/util/p_util.c +++ b/src/gallium/auxiliary/util/p_util.c @@ -33,6 +33,7 @@ #include "pipe/p_defines.h" #include "pipe/p_util.h" #include "pipe/p_format.h" +#include "util/u_rect.h" /** diff --git a/src/gallium/auxiliary/util/u_rect.h b/src/gallium/auxiliary/util/u_rect.h new file mode 100644 index 0000000000..fba4808864 --- /dev/null +++ b/src/gallium/auxiliary/util/u_rect.h @@ -0,0 +1,54 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Pipe copy/fill rect helpers. + */ + + +#ifndef U_RECT_H +#define U_RECT_H + + +#include "pipe/p_format.h" + + +extern void +pipe_copy_rect(ubyte * dst, const struct pipe_format_block *block, + unsigned dst_stride, unsigned dst_x, unsigned dst_y, + unsigned width, unsigned height, const ubyte * src, + int src_stride, unsigned src_x, int src_y); + +extern void +pipe_fill_rect(ubyte * dst, const struct pipe_format_block *block, + unsigned dst_stride, unsigned dst_x, unsigned dst_y, + unsigned width, unsigned height, uint32_t value); + + + +#endif /* U_RECT_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c index 5549eb496d..01ffa31c2c 100644 --- a/src/gallium/drivers/cell/ppu/cell_surface.c +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -30,6 +30,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "util/p_tile.h" +#include "util/u_rect.h" #include "cell_context.h" #include "cell_surface.h" diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index 4430e81626..17b5125e56 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -34,6 +34,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "util/p_tile.h" +#include "util/u_rect.h" /* Assumes all values are within bounds -- no checking at this level - diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index 0be3dfc743..69da252285 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -33,6 +33,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "util/p_tile.h" +#include "util/u_rect.h" diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index 4f1bb881cb..bfbae234f1 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -30,6 +30,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "util/p_tile.h" +#include "util/u_rect.h" #include "sp_context.h" #include "sp_surface.h" diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h index cac0039e12..4a3fca5962 100644 --- a/src/gallium/include/pipe/p_util.h +++ b/src/gallium/include/pipe/p_util.h @@ -31,7 +31,6 @@ #include "p_config.h" #include "p_compiler.h" #include "p_debug.h" -#include "p_format.h" #include "p_pointer.h" #if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) @@ -400,18 +399,6 @@ do { \ } while (0) -/* util/p_util.c - */ -extern void pipe_copy_rect(ubyte * dst, const struct pipe_format_block *block, - unsigned dst_stride, unsigned dst_x, unsigned dst_y, - unsigned width, unsigned height, const ubyte * src, - int src_stride, unsigned src_x, int src_y); - -extern void -pipe_fill_rect(ubyte * dst, const struct pipe_format_block *block, - unsigned dst_stride, unsigned dst_x, unsigned dst_y, - unsigned width, unsigned height, uint32_t value); - #if defined(_MSC_VER) #if _MSC_VER < 1400 && !defined(__cplusplus) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 289b78b38b..63046a0ecc 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -38,6 +38,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "util/u_rect.h" #define DBG if(0) printf -- cgit v1.2.3 From 2b87174ec306673c59386102890f35907be497ef Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Fri, 22 Aug 2008 16:22:35 -0600 Subject: gallium: rename p_util.c to u_rect.c (it only contains rect copy/fill helpers) --- src/gallium/auxiliary/util/Makefile | 2 +- src/gallium/auxiliary/util/SConscript | 2 +- src/gallium/auxiliary/util/p_util.c | 151 ---------------------------------- src/gallium/auxiliary/util/u_rect.c | 151 ++++++++++++++++++++++++++++++++++ 4 files changed, 153 insertions(+), 153 deletions(-) delete mode 100644 src/gallium/auxiliary/util/p_util.c create mode 100644 src/gallium/auxiliary/util/u_rect.c (limited to 'src') diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index be1b97b535..6eebf6d29b 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -6,7 +6,6 @@ LIBNAME = util C_SOURCES = \ p_debug.c \ p_tile.c \ - p_util.c \ u_blit.c \ u_draw_quad.c \ u_gen_mipmap.c \ @@ -14,6 +13,7 @@ C_SOURCES = \ u_hash_table.c \ u_math.c \ u_mm.c \ + u_rect.c \ u_simple_shaders.c \ u_snprintf.c \ u_time.c diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 7865307a7a..94382fe1f9 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -7,7 +7,6 @@ util = env.ConvenienceLibrary( 'p_debug_mem.c', 'p_debug_prof.c', 'p_tile.c', - 'p_util.c', 'u_blit.c', 'u_draw_quad.c', 'u_gen_mipmap.c', @@ -15,6 +14,7 @@ util = env.ConvenienceLibrary( 'u_hash_table.c', 'u_math.c', 'u_mm.c', + 'u_rect.c', 'u_simple_shaders.c', 'u_snprintf.c', 'u_time.c', diff --git a/src/gallium/auxiliary/util/p_util.c b/src/gallium/auxiliary/util/p_util.c deleted file mode 100644 index 787881b192..0000000000 --- a/src/gallium/auxiliary/util/p_util.c +++ /dev/null @@ -1,151 +0,0 @@ -/************************************************************************** - * - * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * Miscellaneous utility functions. - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_format.h" -#include "util/u_rect.h" - - -/** - * Copy 2D rect from one place to another. - * Position and sizes are in pixels. - * src_pitch may be negative to do vertical flip of pixels from source. - */ -void -pipe_copy_rect(ubyte * dst, - const struct pipe_format_block *block, - unsigned dst_stride, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - const ubyte * src, - int src_stride, - unsigned src_x, - int src_y) -{ - unsigned i; - int src_stride_pos = src_stride < 0 ? -src_stride : src_stride; - - assert(block->size > 0); - assert(block->width > 0); - assert(block->height > 0); - assert(src_x >= 0); - assert(src_y >= 0); - assert(dst_x >= 0); - assert(dst_y >= 0); - - dst_x /= block->width; - dst_y /= block->height; - width = (width + block->width - 1)/block->width; - height = (height + block->height - 1)/block->height; - src_x /= block->width; - src_y /= block->height; - - dst += dst_x * block->size; - src += src_x * block->size; - dst += dst_y * dst_stride; - src += src_y * src_stride_pos; - width *= block->size; - - if (width == dst_stride && width == src_stride) - memcpy(dst, src, height * width); - else { - for (i = 0; i < height; i++) { - memcpy(dst, src, width); - dst += dst_stride; - src += src_stride; - } - } -} - -void -pipe_fill_rect(ubyte * dst, - const struct pipe_format_block *block, - unsigned dst_stride, - unsigned dst_x, - unsigned dst_y, - unsigned width, - unsigned height, - uint32_t value) -{ - unsigned i, j; - unsigned width_size; - - assert(block->size > 0); - assert(block->width > 0); - assert(block->height > 0); - assert(dst_x >= 0); - assert(dst_y >= 0); - - dst_x /= block->width; - dst_y /= block->height; - width = (width + block->width - 1)/block->width; - height = (height + block->height - 1)/block->height; - - dst += dst_x * block->size; - dst += dst_y * dst_stride; - width_size = width * block->size; - - switch (block->size) { - case 1: - if(dst_stride == width_size) - memset(dst, (ubyte) value, height * width_size); - else { - for (i = 0; i < height; i++) { - memset(dst, (ubyte) value, width_size); - dst += dst_stride; - } - } - break; - case 2: - for (i = 0; i < height; i++) { - uint16_t *row = (uint16_t *)dst; - for (j = 0; j < width; j++) - *row++ = (uint16_t) value; - dst += dst_stride; - } - break; - case 4: - for (i = 0; i < height; i++) { - uint32_t *row = (uint32_t *)dst; - for (j = 0; j < width; j++) - *row++ = value; - dst += dst_stride; - } - break; - default: - assert(0); - break; - } -} diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c new file mode 100644 index 0000000000..94e447b9d5 --- /dev/null +++ b/src/gallium/auxiliary/util/u_rect.c @@ -0,0 +1,151 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * Rectangle-related helper functions. + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_util.h" +#include "pipe/p_format.h" +#include "util/u_rect.h" + + +/** + * Copy 2D rect from one place to another. + * Position and sizes are in pixels. + * src_pitch may be negative to do vertical flip of pixels from source. + */ +void +pipe_copy_rect(ubyte * dst, + const struct pipe_format_block *block, + unsigned dst_stride, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + const ubyte * src, + int src_stride, + unsigned src_x, + int src_y) +{ + unsigned i; + int src_stride_pos = src_stride < 0 ? -src_stride : src_stride; + + assert(block->size > 0); + assert(block->width > 0); + assert(block->height > 0); + assert(src_x >= 0); + assert(src_y >= 0); + assert(dst_x >= 0); + assert(dst_y >= 0); + + dst_x /= block->width; + dst_y /= block->height; + width = (width + block->width - 1)/block->width; + height = (height + block->height - 1)/block->height; + src_x /= block->width; + src_y /= block->height; + + dst += dst_x * block->size; + src += src_x * block->size; + dst += dst_y * dst_stride; + src += src_y * src_stride_pos; + width *= block->size; + + if (width == dst_stride && width == src_stride) + memcpy(dst, src, height * width); + else { + for (i = 0; i < height; i++) { + memcpy(dst, src, width); + dst += dst_stride; + src += src_stride; + } + } +} + +void +pipe_fill_rect(ubyte * dst, + const struct pipe_format_block *block, + unsigned dst_stride, + unsigned dst_x, + unsigned dst_y, + unsigned width, + unsigned height, + uint32_t value) +{ + unsigned i, j; + unsigned width_size; + + assert(block->size > 0); + assert(block->width > 0); + assert(block->height > 0); + assert(dst_x >= 0); + assert(dst_y >= 0); + + dst_x /= block->width; + dst_y /= block->height; + width = (width + block->width - 1)/block->width; + height = (height + block->height - 1)/block->height; + + dst += dst_x * block->size; + dst += dst_y * dst_stride; + width_size = width * block->size; + + switch (block->size) { + case 1: + if(dst_stride == width_size) + memset(dst, (ubyte) value, height * width_size); + else { + for (i = 0; i < height; i++) { + memset(dst, (ubyte) value, width_size); + dst += dst_stride; + } + } + break; + case 2: + for (i = 0; i < height; i++) { + uint16_t *row = (uint16_t *)dst; + for (j = 0; j < width; j++) + *row++ = (uint16_t) value; + dst += dst_stride; + } + break; + case 4: + for (i = 0; i < height; i++) { + uint32_t *row = (uint32_t *)dst; + for (j = 0; j < width; j++) + *row++ = value; + dst += dst_stride; + } + break; + default: + assert(0); + break; + } +} -- cgit v1.2.3 From f3dfd5969d447515635c077b700a2d4957208167 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 23 Aug 2008 12:30:39 +0200 Subject: util: Include missing u_rect.h. --- src/gallium/auxiliary/util/p_tile.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c index e366039a27..8219041c80 100644 --- a/src/gallium/auxiliary/util/p_tile.c +++ b/src/gallium/auxiliary/util/p_tile.c @@ -37,7 +37,7 @@ #include "pipe/p_inlines.h" #include "p_tile.h" - +#include "u_rect.h" /** -- cgit v1.2.3 From e7ff7f78be6c14b6c48e451d6d1f597af379f8f8 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 23 Aug 2008 12:31:16 +0200 Subject: util: Silence compiler warnings on Windows. --- src/gallium/auxiliary/util/u_math.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_math.c b/src/gallium/auxiliary/util/u_math.c index 8bf6551b6e..0729114d6a 100644 --- a/src/gallium/auxiliary/util/u_math.c +++ b/src/gallium/auxiliary/util/u_math.c @@ -39,7 +39,7 @@ init_pow2_table(void) { int i; for (i = 0; i < POW2_TABLE_SIZE; i++) { - pow2_table[i] = pow(2.0, i / POW2_TABLE_SCALE); + pow2_table[i] = (float) pow(2.0, i / POW2_TABLE_SCALE); } } -- cgit v1.2.3 From ec7415642d7eb192164e7a513198b86756de484c Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Sat, 23 Aug 2008 12:31:50 +0200 Subject: trace: Include u_string.h. --- src/gallium/drivers/trace/tr_stream_wd.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/drivers/trace/tr_stream_wd.c b/src/gallium/drivers/trace/tr_stream_wd.c index 45309dcb4e..b3b65f0971 100644 --- a/src/gallium/drivers/trace/tr_stream_wd.c +++ b/src/gallium/drivers/trace/tr_stream_wd.c @@ -38,6 +38,7 @@ #include #include "pipe/p_util.h" +#include "util/u_string.h" #include "tr_stream.h" -- cgit v1.2.3 From 4f25420bdd834e81a3e22733304efc5261c2998a Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sun, 24 Aug 2008 17:48:55 -0600 Subject: gallium: refactor/replace p_util.h with util/u_memory.h and util/u_math.h Also, rename p_tile.[ch] to u_tile.[ch] --- src/gallium/README.portability | 4 +- src/gallium/auxiliary/cso_cache/cso_cache.c | 3 +- src/gallium/auxiliary/cso_cache/cso_context.c | 2 +- src/gallium/auxiliary/cso_cache/cso_hash.c | 2 +- src/gallium/auxiliary/draw/draw_context.c | 3 +- src/gallium/auxiliary/draw/draw_pipe.c | 1 - src/gallium/auxiliary/draw/draw_pipe_aaline.c | 3 +- src/gallium/auxiliary/draw/draw_pipe_aapoint.c | 4 +- src/gallium/auxiliary/draw/draw_pipe_clip.c | 4 +- src/gallium/auxiliary/draw/draw_pipe_cull.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_flatshade.c | 4 +- src/gallium/auxiliary/draw/draw_pipe_offset.c | 3 +- src/gallium/auxiliary/draw/draw_pipe_pstipple.c | 4 +- src/gallium/auxiliary/draw/draw_pipe_stipple.c | 6 +- src/gallium/auxiliary/draw/draw_pipe_twoside.c | 3 +- src/gallium/auxiliary/draw/draw_pipe_unfilled.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_util.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_validate.c | 2 +- src/gallium/auxiliary/draw/draw_pipe_vbuf.c | 3 +- src/gallium/auxiliary/draw/draw_pipe_wide_line.c | 3 +- src/gallium/auxiliary/draw/draw_pipe_wide_point.c | 3 +- src/gallium/auxiliary/draw/draw_pt.c | 1 - src/gallium/auxiliary/draw/draw_pt_emit.c | 2 +- src/gallium/auxiliary/draw/draw_pt_fetch.c | 2 +- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 2 +- .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 3 +- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 3 +- src/gallium/auxiliary/draw/draw_pt_post_vs.c | 2 +- src/gallium/auxiliary/draw/draw_pt_util.c | 1 - src/gallium/auxiliary/draw/draw_pt_varray.c | 4 +- src/gallium/auxiliary/draw/draw_pt_vcache.c | 2 +- src/gallium/auxiliary/draw/draw_vbuf.h | 2 - src/gallium/auxiliary/draw/draw_vs.c | 6 +- src/gallium/auxiliary/draw/draw_vs_aos.c | 4 +- src/gallium/auxiliary/draw/draw_vs_aos_io.c | 2 +- src/gallium/auxiliary/draw/draw_vs_aos_machine.c | 3 +- src/gallium/auxiliary/draw/draw_vs_exec.c | 3 +- src/gallium/auxiliary/draw/draw_vs_llvm.c | 1 - src/gallium/auxiliary/draw/draw_vs_sse.c | 3 +- src/gallium/auxiliary/draw/draw_vs_varient.c | 3 +- src/gallium/auxiliary/gallivm/gallivm_cpu.cpp | 3 +- src/gallium/auxiliary/gallivm/instructions.cpp | 2 +- src/gallium/auxiliary/gallivm/instructionssoa.cpp | 2 +- .../auxiliary/pipebuffer/pb_buffer_fenced.c | 2 +- .../auxiliary/pipebuffer/pb_buffer_malloc.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c | 2 +- .../auxiliary/pipebuffer/pb_bufmgr_fenced.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_validate.c | 2 +- src/gallium/auxiliary/pipebuffer/pb_winsys.c | 2 +- src/gallium/auxiliary/rtasm/rtasm_execmem.c | 2 +- src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c | 2 +- src/gallium/auxiliary/sct/sct.c | 2 +- src/gallium/auxiliary/tgsi/tgsi_build.c | 1 - src/gallium/auxiliary/tgsi/tgsi_build.h | 4 + src/gallium/auxiliary/tgsi/tgsi_dump_c.c | 1 - src/gallium/auxiliary/tgsi/tgsi_exec.c | 2 +- src/gallium/auxiliary/tgsi/tgsi_parse.c | 2 +- src/gallium/auxiliary/tgsi/tgsi_scan.c | 6 +- src/gallium/auxiliary/tgsi/tgsi_sse2.c | 2 +- src/gallium/auxiliary/tgsi/tgsi_transform.c | 1 + src/gallium/auxiliary/tgsi/tgsi_transform.h | 1 - src/gallium/auxiliary/tgsi/tgsi_util.c | 1 - src/gallium/auxiliary/translate/translate.c | 1 - src/gallium/auxiliary/translate/translate_cache.c | 2 +- .../auxiliary/translate/translate_generic.c | 2 +- src/gallium/auxiliary/translate/translate_sse.c | 2 +- src/gallium/auxiliary/util/Makefile | 2 +- src/gallium/auxiliary/util/SConscript | 2 +- src/gallium/auxiliary/util/p_debug.c | 1 - src/gallium/auxiliary/util/u_blit.c | 5 +- src/gallium/auxiliary/util/u_gen_mipmap.c | 2 +- src/gallium/auxiliary/util/u_handle_table.c | 4 +- src/gallium/auxiliary/util/u_hash_table.c | 5 +- src/gallium/auxiliary/util/u_math.h | 240 +++- src/gallium/auxiliary/util/u_memory.h | 222 ++++ src/gallium/auxiliary/util/u_mm.c | 2 +- src/gallium/auxiliary/util/u_pack_color.h | 36 +- src/gallium/auxiliary/util/u_pointer.h | 107 ++ src/gallium/auxiliary/util/u_rect.c | 1 - src/gallium/auxiliary/util/u_simple_shaders.c | 2 +- src/gallium/auxiliary/util/u_tile.c | 1169 ++++++++++++++++++++ src/gallium/auxiliary/util/u_tile.h | 101 ++ src/gallium/drivers/cell/common.h | 1 - src/gallium/drivers/cell/ppu/cell_clear.c | 2 +- src/gallium/drivers/cell/ppu/cell_context.c | 2 +- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 2 +- src/gallium/drivers/cell/ppu/cell_render.c | 2 +- src/gallium/drivers/cell/ppu/cell_screen.c | 2 +- src/gallium/drivers/cell/ppu/cell_state_derived.c | 2 +- src/gallium/drivers/cell/ppu/cell_state_emit.c | 2 +- src/gallium/drivers/cell/ppu/cell_state_shader.c | 2 +- src/gallium/drivers/cell/ppu/cell_surface.c | 2 +- src/gallium/drivers/cell/ppu/cell_texture.c | 2 +- src/gallium/drivers/cell/ppu/cell_winsys.c | 2 +- src/gallium/drivers/cell/spu/spu_exec.c | 1 - src/gallium/drivers/cell/spu/spu_tri.c | 1 - src/gallium/drivers/cell/spu/spu_util.c | 1 - src/gallium/drivers/cell/spu/spu_vertex_fetch.c | 1 - src/gallium/drivers/cell/spu/spu_vertex_shader.c | 1 - src/gallium/drivers/failover/fo_context.c | 2 +- src/gallium/drivers/i915simple/i915_context.c | 2 +- src/gallium/drivers/i915simple/i915_debug_fp.c | 2 +- src/gallium/drivers/i915simple/i915_fpc.h | 1 - .../drivers/i915simple/i915_fpc_translate.c | 2 + src/gallium/drivers/i915simple/i915_prim_emit.c | 4 +- src/gallium/drivers/i915simple/i915_prim_vbuf.c | 3 +- src/gallium/drivers/i915simple/i915_screen.c | 2 +- src/gallium/drivers/i915simple/i915_state.c | 3 +- .../drivers/i915simple/i915_state_derived.c | 2 +- .../drivers/i915simple/i915_state_dynamic.c | 4 +- .../drivers/i915simple/i915_state_immediate.c | 2 +- .../drivers/i915simple/i915_state_sampler.c | 2 +- src/gallium/drivers/i915simple/i915_surface.c | 3 +- src/gallium/drivers/i915simple/i915_texture.c | 3 +- src/gallium/drivers/i965simple/brw_cc.c | 6 +- src/gallium/drivers/i965simple/brw_clip_state.c | 3 +- src/gallium/drivers/i965simple/brw_context.c | 2 +- src/gallium/drivers/i965simple/brw_curbe.c | 3 +- src/gallium/drivers/i965simple/brw_draw_upload.c | 1 + src/gallium/drivers/i965simple/brw_gs_state.c | 3 +- src/gallium/drivers/i965simple/brw_screen.c | 2 +- src/gallium/drivers/i965simple/brw_sf_state.c | 5 +- src/gallium/drivers/i965simple/brw_shader_info.c | 2 +- src/gallium/drivers/i965simple/brw_state.c | 2 +- src/gallium/drivers/i965simple/brw_state_batch.c | 2 +- src/gallium/drivers/i965simple/brw_state_cache.c | 2 +- src/gallium/drivers/i965simple/brw_state_pool.c | 3 +- src/gallium/drivers/i965simple/brw_state_upload.c | 2 +- src/gallium/drivers/i965simple/brw_surface.c | 3 +- src/gallium/drivers/i965simple/brw_tex_layout.c | 8 +- src/gallium/drivers/i965simple/brw_vs_state.c | 3 +- src/gallium/drivers/i965simple/brw_wm.c | 2 +- src/gallium/drivers/i965simple/brw_wm_decl.c | 3 +- src/gallium/drivers/i965simple/brw_wm_glsl.c | 3 +- .../drivers/i965simple/brw_wm_sampler_state.c | 3 +- src/gallium/drivers/i965simple/brw_wm_state.c | 3 +- src/gallium/drivers/softpipe/sp_context.c | 2 +- src/gallium/drivers/softpipe/sp_fs_exec.c | 2 +- src/gallium/drivers/softpipe/sp_fs_llvm.c | 2 +- src/gallium/drivers/softpipe/sp_fs_sse.c | 2 +- src/gallium/drivers/softpipe/sp_prim_setup.c | 2 +- src/gallium/drivers/softpipe/sp_prim_vbuf.c | 1 + src/gallium/drivers/softpipe/sp_quad_alpha_test.c | 2 +- src/gallium/drivers/softpipe/sp_quad_blend.c | 29 +- src/gallium/drivers/softpipe/sp_quad_bufloop.c | 2 +- src/gallium/drivers/softpipe/sp_quad_colormask.c | 3 +- src/gallium/drivers/softpipe/sp_quad_coverage.c | 2 +- src/gallium/drivers/softpipe/sp_quad_depth_test.c | 2 +- src/gallium/drivers/softpipe/sp_quad_earlyz.c | 2 +- src/gallium/drivers/softpipe/sp_quad_fs.c | 3 +- src/gallium/drivers/softpipe/sp_quad_occlusion.c | 2 +- src/gallium/drivers/softpipe/sp_quad_output.c | 2 +- src/gallium/drivers/softpipe/sp_quad_stencil.c | 2 +- src/gallium/drivers/softpipe/sp_quad_stipple.c | 2 +- src/gallium/drivers/softpipe/sp_query.c | 2 +- src/gallium/drivers/softpipe/sp_screen.c | 2 +- src/gallium/drivers/softpipe/sp_setup.c | 2 +- src/gallium/drivers/softpipe/sp_state_blend.c | 2 +- src/gallium/drivers/softpipe/sp_state_derived.c | 3 +- src/gallium/drivers/softpipe/sp_state_fs.c | 2 +- src/gallium/drivers/softpipe/sp_state_rasterizer.c | 2 +- src/gallium/drivers/softpipe/sp_state_sampler.c | 2 +- src/gallium/drivers/softpipe/sp_surface.c | 3 +- src/gallium/drivers/softpipe/sp_tex_sample.c | 2 +- src/gallium/drivers/softpipe/sp_texture.c | 3 +- src/gallium/drivers/softpipe/sp_tile_cache.c | 4 +- src/gallium/drivers/trace/tr_context.c | 2 +- src/gallium/drivers/trace/tr_dump.c | 2 + src/gallium/drivers/trace/tr_dump.h | 1 - src/gallium/drivers/trace/tr_screen.c | 2 +- src/gallium/drivers/trace/tr_state.c | 1 + src/gallium/drivers/trace/tr_stream_stdc.c | 2 +- src/gallium/drivers/trace/tr_stream_wd.c | 2 +- src/gallium/drivers/trace/tr_texture.c | 2 +- src/gallium/drivers/trace/tr_winsys.c | 3 +- src/gallium/include/pipe/p_util.h | 460 -------- src/gallium/state_trackers/python/gallium.i | 2 +- src/gallium/state_trackers/python/st_device.c | 3 +- src/gallium/state_trackers/python/st_sample.c | 5 +- .../state_trackers/python/st_softpipe_winsys.c | 3 +- .../winsys/drm/intel/common/intel_be_device.c | 2 +- .../winsys/drm/intel/dri/intel_winsys_softpipe.c | 2 +- src/gallium/winsys/egl_xlib/egl_xlib.c | 2 +- src/gallium/winsys/egl_xlib/sw_winsys.c | 3 +- src/gallium/winsys/gdi/wmesa.c | 2 +- src/gallium/winsys/xlib/brw_aub.c | 1 - src/gallium/winsys/xlib/xm_winsys.c | 3 +- src/gallium/winsys/xlib/xm_winsys_aub.c | 2 +- src/mesa/state_tracker/acc2.c | 319 ++++++ src/mesa/state_tracker/st_cb_accum.c | 2 +- src/mesa/state_tracker/st_cb_bitmap.c | 2 +- src/mesa/state_tracker/st_cb_drawpixels.c | 2 +- src/mesa/state_tracker/st_cb_readpixels.c | 2 +- src/mesa/state_tracker/st_cb_texture.c | 2 +- src/mesa/state_tracker/st_program.c | 2 +- src/mesa/state_tracker/st_texture.c | 1 - 201 files changed, 2453 insertions(+), 686 deletions(-) create mode 100644 src/gallium/auxiliary/util/u_memory.h create mode 100644 src/gallium/auxiliary/util/u_pointer.h create mode 100644 src/gallium/auxiliary/util/u_tile.c create mode 100644 src/gallium/auxiliary/util/u_tile.h delete mode 100644 src/gallium/include/pipe/p_util.h create mode 100644 src/mesa/state_tracker/acc2.c (limited to 'src') diff --git a/src/gallium/README.portability b/src/gallium/README.portability index d5d5987a7f..adecf4bb79 100644 --- a/src/gallium/README.portability +++ b/src/gallium/README.portability @@ -35,8 +35,8 @@ not available in Windows Kernel Mode. Use the appropriate p_*.h include. * Use MALLOC, CALLOC, FREE instead of the malloc, calloc, free functions. -* Use align_pointer() function defined in p_util.h for aligning pointers in a -portable way. +* Use align_pointer() function defined in u_memory.h for aligning pointers + in a portable way. == Debugging == diff --git a/src/gallium/auxiliary/cso_cache/cso_cache.c b/src/gallium/auxiliary/cso_cache/cso_cache.c index 36dc46ff80..6b1754ea00 100644 --- a/src/gallium/auxiliary/cso_cache/cso_cache.c +++ b/src/gallium/auxiliary/cso_cache/cso_cache.c @@ -28,9 +28,10 @@ /* Authors: Zack Rusin */ -#include "pipe/p_util.h" #include "pipe/p_debug.h" +#include "util/u_memory.h" + #include "cso_cache.h" #include "cso_hash.h" diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c index 86e4d46a20..f22ba40824 100644 --- a/src/gallium/auxiliary/cso_cache/cso_context.c +++ b/src/gallium/auxiliary/cso_cache/cso_context.c @@ -36,7 +36,7 @@ */ #include "pipe/p_state.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "tgsi/tgsi_parse.h" diff --git a/src/gallium/auxiliary/cso_cache/cso_hash.c b/src/gallium/auxiliary/cso_cache/cso_hash.c index 0646efd952..7f0044c5a7 100644 --- a/src/gallium/auxiliary/cso_cache/cso_hash.c +++ b/src/gallium/auxiliary/cso_cache/cso_hash.c @@ -31,7 +31,7 @@ */ #include "pipe/p_debug.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "cso_hash.h" diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 2f263cf06a..1c26cb31a3 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -31,7 +31,8 @@ */ -#include "pipe/p_util.h" +#include "util/u_memory.h" +#include "util/u_math.h" #include "draw_context.h" #include "draw_vbuf.h" #include "draw_vs.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c index 1db43876ef..3cde9d36d3 100644 --- a/src/gallium/auxiliary/draw/draw_pipe.c +++ b/src/gallium/auxiliary/draw/draw_pipe.c @@ -30,7 +30,6 @@ * Keith Whitwell */ -#include "pipe/p_util.h" #include "draw/draw_private.h" #include "draw/draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_aaline.c b/src/gallium/auxiliary/draw/draw_pipe_aaline.c index 991304b2c8..20841bb5d6 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aaline.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aaline.c @@ -32,11 +32,12 @@ */ -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "tgsi/tgsi_transform.h" #include "tgsi/tgsi_dump.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c index c7f4349cb3..2c1cacbdb4 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_aapoint.c +++ b/src/gallium/auxiliary/draw/draw_pipe_aapoint.c @@ -38,7 +38,6 @@ */ -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" @@ -47,6 +46,9 @@ #include "tgsi/tgsi_transform.h" #include "tgsi/tgsi_dump.h" +#include "util/u_math.h" +#include "util/u_memory.h" + #include "draw_context.h" #include "draw_vs.h" #include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_clip.c b/src/gallium/auxiliary/draw/draw_pipe_clip.c index fa10f8efca..3265dcd154 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_clip.c +++ b/src/gallium/auxiliary/draw/draw_pipe_clip.c @@ -32,7 +32,9 @@ */ -#include "pipe/p_util.h" +#include "util/u_memory.h" +#include "util/u_math.h" + #include "pipe/p_shader_tokens.h" #include "draw_vs.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_cull.c b/src/gallium/auxiliary/draw/draw_pipe_cull.c index d0d22a38e0..053be5f050 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_cull.c +++ b/src/gallium/auxiliary/draw/draw_pipe_cull.c @@ -33,7 +33,7 @@ */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_defines.h" #include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c index 4741b22d02..43d1fecc4d 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_flatshade.c +++ b/src/gallium/auxiliary/draw/draw_pipe_flatshade.c @@ -28,7 +28,9 @@ /* Authors: Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" + #include "pipe/p_shader_tokens.h" #include "draw_vs.h" #include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_offset.c b/src/gallium/auxiliary/draw/draw_pipe_offset.c index 2f5865741c..1fea5e6dcb 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_offset.c +++ b/src/gallium/auxiliary/draw/draw_pipe_offset.c @@ -32,7 +32,8 @@ * \author Brian Paul */ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c index e97136fa1f..b764d9c518 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_pstipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_pstipple.c @@ -34,12 +34,14 @@ */ -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" +#include "util/u_math.h" +#include "util/u_memory.h" + #include "tgsi/tgsi_transform.h" #include "tgsi/tgsi_dump.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_stipple.c b/src/gallium/auxiliary/draw/draw_pipe_stipple.c index bf0db18a68..b65e2aa102 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_stipple.c +++ b/src/gallium/auxiliary/draw/draw_pipe_stipple.c @@ -36,10 +36,12 @@ */ -#include "pipe/p_util.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" -#include "draw_pipe.h" +#include "util/u_math.h" +#include "util/u_memory.h" + +#include "draw/draw_pipe.h" /** Subclass of draw_stage */ diff --git a/src/gallium/auxiliary/draw/draw_pipe_twoside.c b/src/gallium/auxiliary/draw/draw_pipe_twoside.c index 3ac825f565..c329d92339 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_twoside.c +++ b/src/gallium/auxiliary/draw/draw_pipe_twoside.c @@ -28,7 +28,8 @@ /* Authors: Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" #include "draw_vs.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c index 8f97fdedaa..68835fd1a5 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_unfilled.c +++ b/src/gallium/auxiliary/draw/draw_pipe_unfilled.c @@ -33,7 +33,7 @@ /* Authors: Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_defines.h" #include "draw_private.h" #include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_util.c b/src/gallium/auxiliary/draw/draw_pipe_util.c index 04438f4dd0..e22e5fed0c 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_util.c +++ b/src/gallium/auxiliary/draw/draw_pipe_util.c @@ -30,7 +30,7 @@ * Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "draw/draw_private.h" #include "draw/draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_validate.c b/src/gallium/auxiliary/draw/draw_pipe_validate.c index 6be1d369c3..f34c68728e 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_validate.c +++ b/src/gallium/auxiliary/draw/draw_pipe_validate.c @@ -28,7 +28,7 @@ /* Authors: Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_defines.h" #include "draw_private.h" #include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c index a6fde77a0e..c0cf4269db 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_vbuf.c +++ b/src/gallium/auxiliary/draw/draw_pipe_vbuf.c @@ -35,7 +35,8 @@ #include "pipe/p_debug.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "draw_vbuf.h" #include "draw_private.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c index 48ec2f1239..184e363594 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_line.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_line.c @@ -28,9 +28,10 @@ /* Authors: Keith Whitwell */ -#include "pipe/p_util.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "draw_private.h" #include "draw_pipe.h" diff --git a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c index 54590984c6..4f1326053d 100644 --- a/src/gallium/auxiliary/draw/draw_pipe_wide_point.c +++ b/src/gallium/auxiliary/draw/draw_pipe_wide_point.c @@ -28,7 +28,8 @@ /* Authors: Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" #include "draw_vs.h" diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c index 85a75525c8..669c11c993 100644 --- a/src/gallium/auxiliary/draw/draw_pt.c +++ b/src/gallium/auxiliary/draw/draw_pt.c @@ -30,7 +30,6 @@ * Keith Whitwell */ -#include "pipe/p_util.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_pt.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_emit.c b/src/gallium/auxiliary/draw/draw_pt_emit.c index 40f05cb9e0..d4eca80588 100644 --- a/src/gallium/auxiliary/draw/draw_pt_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_emit.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vbuf.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c index 07f4c99164..6377f896fb 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vbuf.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 4a1f3b0953..0684c93d10 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -30,7 +30,7 @@ * Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vbuf.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index fdf9b6fe6a..87094f3092 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -31,7 +31,8 @@ */ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vbuf.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index be3535ed9e..f617aac9f7 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -25,7 +25,8 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "draw/draw_context.h" #include "draw/draw_vbuf.h" #include "draw/draw_vertex.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c index af6306b1c6..96dc706b99 100644 --- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c +++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_context.h" #include "draw/draw_context.h" #include "draw/draw_private.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_util.c b/src/gallium/auxiliary/draw/draw_pt_util.c index 32c8a9632c..3bc7939c55 100644 --- a/src/gallium/auxiliary/draw/draw_pt_util.c +++ b/src/gallium/auxiliary/draw/draw_pt_util.c @@ -30,7 +30,6 @@ * Keith Whitwell */ -#include "pipe/p_util.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_pt.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_varray.c b/src/gallium/auxiliary/draw/draw_pt_varray.c index 46e722a154..c15afe65f1 100644 --- a/src/gallium/auxiliary/draw/draw_pt_varray.c +++ b/src/gallium/auxiliary/draw/draw_pt_varray.c @@ -25,7 +25,9 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" + #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_pt.h" diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index cda2987c9e..b8b5de729d 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -30,7 +30,7 @@ * Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_pt.h" diff --git a/src/gallium/auxiliary/draw/draw_vbuf.h b/src/gallium/auxiliary/draw/draw_vbuf.h index e90f37872a..62247ccd9f 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.h +++ b/src/gallium/auxiliary/draw/draw_vbuf.h @@ -37,8 +37,6 @@ #define DRAW_VBUF_H_ -#include "pipe/p_util.h" - struct draw_context; struct vertex_info; diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c index f798b20492..34adbd49b0 100644 --- a/src/gallium/auxiliary/draw/draw_vs.c +++ b/src/gallium/auxiliary/draw/draw_vs.c @@ -31,11 +31,15 @@ * Brian Paul */ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" + #include "pipe/p_shader_tokens.h" + #include "draw_private.h" #include "draw_context.h" #include "draw_vs.h" + #include "translate/translate.h" #include "translate/translate_cache.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_aos.c b/src/gallium/auxiliary/draw/draw_vs_aos.c index 41bdd012d5..760fcb389f 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos.c @@ -29,9 +29,9 @@ */ -#include "pipe/p_util.h" -#include "pipe/p_shader_tokens.h" +#include "util/u_memory.h" #include "util/u_math.h" +#include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi/tgsi_exec.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_io.c b/src/gallium/auxiliary/draw/draw_vs_aos_io.c index eda677cc62..ab3c5b94a5 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_io.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_io.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c index e029b7b4bb..b358bd2df4 100644 --- a/src/gallium/auxiliary/draw/draw_vs_aos_machine.c +++ b/src/gallium/auxiliary/draw/draw_vs_aos_machine.c @@ -29,8 +29,9 @@ #include "pipe/p_config.h" -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi/tgsi_exec.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_exec.c b/src/gallium/auxiliary/draw/draw_vs_exec.c index e26903d8cc..44563803f9 100644 --- a/src/gallium/auxiliary/draw/draw_vs_exec.c +++ b/src/gallium/auxiliary/draw/draw_vs_exec.c @@ -31,7 +31,8 @@ * Brian Paul */ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "pipe/p_shader_tokens.h" #include "draw_private.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_llvm.c b/src/gallium/auxiliary/draw/draw_vs_llvm.c index fc03473b91..2ce30b9a02 100644 --- a/src/gallium/auxiliary/draw/draw_vs_llvm.c +++ b/src/gallium/auxiliary/draw/draw_vs_llvm.c @@ -32,7 +32,6 @@ * Brian Paul */ -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "draw_private.h" #include "draw_context.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_sse.c b/src/gallium/auxiliary/draw/draw_vs_sse.c index 61f0c084c3..0efabd9de8 100644 --- a/src/gallium/auxiliary/draw/draw_vs_sse.c +++ b/src/gallium/auxiliary/draw/draw_vs_sse.c @@ -31,13 +31,14 @@ * Brian Paul */ +#include "util/u_math.h" +#include "util/u_memory.h" #include "pipe/p_config.h" #include "draw_vs.h" #if defined(PIPE_ARCH_X86) -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "draw_private.h" diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index 994ce3e889..4daf05dae7 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -30,7 +30,8 @@ * Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_memory.h" +#include "util/u_math.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vbuf.h" diff --git a/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp index cf5b978837..e64bfb1c6c 100644 --- a/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp +++ b/src/gallium/auxiliary/gallivm/gallivm_cpu.cpp @@ -41,11 +41,12 @@ #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" -#include "pipe/p_util.h" #include "tgsi/tgsi_exec.h" #include "tgsi/tgsi_dump.h" +#include "util/u_memory.h" + #include #include #include diff --git a/src/gallium/auxiliary/gallivm/instructions.cpp b/src/gallium/auxiliary/gallivm/instructions.cpp index 035224e8f3..a82dc30306 100644 --- a/src/gallium/auxiliary/gallivm/instructions.cpp +++ b/src/gallium/auxiliary/gallivm/instructions.cpp @@ -35,7 +35,7 @@ #include "storage.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include #include diff --git a/src/gallium/auxiliary/gallivm/instructionssoa.cpp b/src/gallium/auxiliary/gallivm/instructionssoa.cpp index 76049ade7c..efddc04e81 100644 --- a/src/gallium/auxiliary/gallivm/instructionssoa.cpp +++ b/src/gallium/auxiliary/gallivm/instructionssoa.cpp @@ -29,7 +29,7 @@ #include "storagesoa.h" #include "pipe/p_shader_tokens.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include #include diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index ce41418a0f..8ae052e875 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -45,7 +45,7 @@ #include "pipe/p_debug.h" #include "pipe/p_winsys.h" #include "pipe/p_thread.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "util/u_double_list.h" #include "pb_buffer.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c index e90d2e5623..20fc87b39d 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_malloc.c @@ -35,7 +35,7 @@ #include "pipe/p_debug.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pb_buffer.h" #include "pb_bufmgr.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c index 0d2d6c0c1b..2afaeafa1a 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_alt.c @@ -35,7 +35,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_debug.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pb_buffer.h" #include "pb_bufmgr.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c index bed4bec4fe..b914c2d0fe 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_cache.c @@ -38,7 +38,7 @@ #include "pipe/p_debug.h" #include "pipe/p_winsys.h" #include "pipe/p_thread.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "util/u_double_list.h" #include "util/u_time.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c index d02e3500ff..5e518370d0 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c @@ -37,7 +37,7 @@ #include "pipe/p_debug.h" #include "pipe/p_winsys.h" #include "pipe/p_thread.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "util/u_double_list.h" #include "util/u_time.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c index 05efd8ce41..8fc63ce648 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c @@ -35,7 +35,7 @@ #include "pipe/p_debug.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pb_buffer.h" #include "pb_buffer_fenced.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c index c51e582611..b40eb6cc90 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_mm.c @@ -36,7 +36,7 @@ #include "pipe/p_defines.h" #include "pipe/p_debug.h" #include "pipe/p_thread.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "util/u_double_list.h" #include "util/u_mm.h" #include "pb_buffer.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c index 95af08929a..93d2cc9635 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_pool.c @@ -39,7 +39,7 @@ #include "pipe/p_debug.h" #include "pipe/p_thread.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "util/u_double_list.h" #include "pb_buffer.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c index 598d9ce310..af307e265a 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_slab.c @@ -39,7 +39,7 @@ #include "pipe/p_debug.h" #include "pipe/p_thread.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "util/u_double_list.h" #include "util/u_time.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.c b/src/gallium/auxiliary/pipebuffer/pb_validate.c index 362fd896f3..1e54fc39d4 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_validate.c +++ b/src/gallium/auxiliary/pipebuffer/pb_validate.c @@ -35,7 +35,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_error.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_debug.h" #include "pb_buffer.h" diff --git a/src/gallium/auxiliary/pipebuffer/pb_winsys.c b/src/gallium/auxiliary/pipebuffer/pb_winsys.c index 978944091f..28d137dbc4 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_winsys.c +++ b/src/gallium/auxiliary/pipebuffer/pb_winsys.c @@ -35,7 +35,7 @@ #include "pipe/p_winsys.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pb_buffer.h" diff --git a/src/gallium/auxiliary/rtasm/rtasm_execmem.c b/src/gallium/auxiliary/rtasm/rtasm_execmem.c index 300c1c2d9d..dfa5c35ab6 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_execmem.c +++ b/src/gallium/auxiliary/rtasm/rtasm_execmem.c @@ -33,7 +33,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_debug.h" #include "pipe/p_thread.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "rtasm_execmem.h" diff --git a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c index 7f6bf577b2..285ddc0e3f 100644 --- a/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c +++ b/src/gallium/auxiliary/rtasm/rtasm_ppc_spe.c @@ -30,7 +30,7 @@ */ #include "pipe/p_compiler.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "rtasm_ppc_spe.h" #ifdef GALLIUM_CELL diff --git a/src/gallium/auxiliary/sct/sct.c b/src/gallium/auxiliary/sct/sct.c index 5e4126e014..49bb7ea92e 100644 --- a/src/gallium/auxiliary/sct/sct.c +++ b/src/gallium/auxiliary/sct/sct.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_state.h" #include "pipe/p_inlines.h" #include "sct.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.c b/src/gallium/auxiliary/tgsi/tgsi_build.c index 050b448fe7..74614d3688 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.c +++ b/src/gallium/auxiliary/tgsi/tgsi_build.c @@ -26,7 +26,6 @@ **************************************************************************/ #include "pipe/p_debug.h" -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi_build.h" #include "tgsi_parse.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_build.h b/src/gallium/auxiliary/tgsi/tgsi_build.h index 6ae7f324f8..7d6234746a 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_build.h +++ b/src/gallium/auxiliary/tgsi/tgsi_build.h @@ -28,6 +28,10 @@ #ifndef TGSI_BUILD_H #define TGSI_BUILD_H + +struct tgsi_token; + + #if defined __cplusplus extern "C" { #endif diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c index 1025866a25..be25cb45a0 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_dump_c.c +++ b/src/gallium/auxiliary/tgsi/tgsi_dump_c.c @@ -26,7 +26,6 @@ **************************************************************************/ #include "pipe/p_debug.h" -#include "pipe/p_util.h" #include "util/u_string.h" #include "tgsi_dump_c.h" #include "tgsi_build.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c index e28b56c842..fb573fe1f0 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_exec.c +++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c @@ -52,11 +52,11 @@ #include "pipe/p_compiler.h" #include "pipe/p_state.h" -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" #include "tgsi_exec.h" +#include "util/u_memory.h" #include "util/u_math.h" #define FAST_MATH 1 diff --git a/src/gallium/auxiliary/tgsi/tgsi_parse.c b/src/gallium/auxiliary/tgsi/tgsi_parse.c index d16f0cdcad..3757486ba9 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_parse.c +++ b/src/gallium/auxiliary/tgsi/tgsi_parse.c @@ -26,10 +26,10 @@ **************************************************************************/ #include "pipe/p_debug.h" -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi_parse.h" #include "tgsi_build.h" +#include "util/u_memory.h" void tgsi_full_token_init( diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c index 59bcf10b53..be4870a498 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_scan.c +++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c @@ -33,11 +33,11 @@ */ -#include "tgsi_scan.h" -#include "tgsi/tgsi_parse.h" +#include "util/u_math.h" #include "tgsi/tgsi_build.h" +#include "tgsi/tgsi_parse.h" +#include "tgsi/tgsi_scan.h" -#include "pipe/p_util.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/tgsi_sse2.c index 00ed4da450..626724ad4e 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_sse2.c +++ b/src/gallium/auxiliary/tgsi/tgsi_sse2.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "pipe/p_debug.h" #include "pipe/p_shader_tokens.h" #include "util/u_math.h" #include "tgsi/tgsi_parse.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.c b/src/gallium/auxiliary/tgsi/tgsi_transform.c index 357f77b05a..ea87da31e5 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_transform.c +++ b/src/gallium/auxiliary/tgsi/tgsi_transform.c @@ -31,6 +31,7 @@ * Authors: Brian Paul */ +#include "pipe/p_debug.h" #include "tgsi_transform.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_transform.h b/src/gallium/auxiliary/tgsi/tgsi_transform.h index 3da0b38271..a121adbaef 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_transform.h +++ b/src/gallium/auxiliary/tgsi/tgsi_transform.h @@ -29,7 +29,6 @@ #define TGSI_TRANSFORM_H -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_build.h" diff --git a/src/gallium/auxiliary/tgsi/tgsi_util.c b/src/gallium/auxiliary/tgsi/tgsi_util.c index 09486e649e..50101a9bb0 100644 --- a/src/gallium/auxiliary/tgsi/tgsi_util.c +++ b/src/gallium/auxiliary/tgsi/tgsi_util.c @@ -26,7 +26,6 @@ **************************************************************************/ #include "pipe/p_debug.h" -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi_parse.h" #include "tgsi_build.h" diff --git a/src/gallium/auxiliary/translate/translate.c b/src/gallium/auxiliary/translate/translate.c index b93fbf9033..7678903f75 100644 --- a/src/gallium/auxiliary/translate/translate.c +++ b/src/gallium/auxiliary/translate/translate.c @@ -31,7 +31,6 @@ */ #include "pipe/p_config.h" -#include "pipe/p_util.h" #include "pipe/p_state.h" #include "translate.h" diff --git a/src/gallium/auxiliary/translate/translate_cache.c b/src/gallium/auxiliary/translate/translate_cache.c index 115dc9287e..d8069a149c 100644 --- a/src/gallium/auxiliary/translate/translate_cache.c +++ b/src/gallium/auxiliary/translate/translate_cache.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_state.h" #include "translate.h" #include "translate_cache.h" diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c index 4c8179ffa8..4d336f47ea 100644 --- a/src/gallium/auxiliary/translate/translate_generic.c +++ b/src/gallium/auxiliary/translate/translate_generic.c @@ -30,7 +30,7 @@ * Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_state.h" #include "translate.h" diff --git a/src/gallium/auxiliary/translate/translate_sse.c b/src/gallium/auxiliary/translate/translate_sse.c index 18a212ac1c..7955186e16 100644 --- a/src/gallium/auxiliary/translate/translate_sse.c +++ b/src/gallium/auxiliary/translate/translate_sse.c @@ -28,7 +28,7 @@ #include "pipe/p_config.h" #include "pipe/p_compiler.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "util/u_simple_list.h" #include "translate.h" diff --git a/src/gallium/auxiliary/util/Makefile b/src/gallium/auxiliary/util/Makefile index 6eebf6d29b..6e5fd26c05 100644 --- a/src/gallium/auxiliary/util/Makefile +++ b/src/gallium/auxiliary/util/Makefile @@ -5,7 +5,6 @@ LIBNAME = util C_SOURCES = \ p_debug.c \ - p_tile.c \ u_blit.c \ u_draw_quad.c \ u_gen_mipmap.c \ @@ -16,6 +15,7 @@ C_SOURCES = \ u_rect.c \ u_simple_shaders.c \ u_snprintf.c \ + u_tile.c \ u_time.c include ../../Makefile.template diff --git a/src/gallium/auxiliary/util/SConscript b/src/gallium/auxiliary/util/SConscript index 94382fe1f9..ce3fad7068 100644 --- a/src/gallium/auxiliary/util/SConscript +++ b/src/gallium/auxiliary/util/SConscript @@ -6,7 +6,6 @@ util = env.ConvenienceLibrary( 'p_debug.c', 'p_debug_mem.c', 'p_debug_prof.c', - 'p_tile.c', 'u_blit.c', 'u_draw_quad.c', 'u_gen_mipmap.c', @@ -17,6 +16,7 @@ util = env.ConvenienceLibrary( 'u_rect.c', 'u_simple_shaders.c', 'u_snprintf.c', + 'u_tile.c', 'u_time.c', ]) diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c index 2c2f2f8931..7d1dba5a24 100644 --- a/src/gallium/auxiliary/util/p_debug.c +++ b/src/gallium/auxiliary/util/p_debug.c @@ -51,7 +51,6 @@ #endif #include "pipe/p_compiler.h" -#include "pipe/p_util.h" #include "pipe/p_debug.h" #include "pipe/p_format.h" #include "pipe/p_state.h" diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c index ae087df4cf..05399f9885 100644 --- a/src/gallium/auxiliary/util/u_blit.c +++ b/src/gallium/auxiliary/util/u_blit.c @@ -37,12 +37,13 @@ #include "pipe/p_debug.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" #include "pipe/p_winsys.h" #include "pipe/p_shader_tokens.h" -#include "util/u_draw_quad.h" #include "util/u_blit.h" +#include "util/u_draw_quad.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "util/u_simple_shaders.h" #include "cso_cache/cso_context.h" diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 8713ff5d58..c1e2c19f87 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -37,10 +37,10 @@ #include "pipe/p_debug.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" #include "pipe/p_winsys.h" #include "pipe/p_shader_tokens.h" +#include "util/u_memory.h" #include "util/u_draw_quad.h" #include "util/u_gen_mipmap.h" #include "util/u_simple_shaders.h" diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c index 2176a00959..2c40011923 100644 --- a/src/gallium/auxiliary/util/u_handle_table.c +++ b/src/gallium/auxiliary/util/u_handle_table.c @@ -35,9 +35,9 @@ #include "pipe/p_compiler.h" #include "pipe/p_debug.h" -#include "pipe/p_util.h" -#include "u_handle_table.h" +#include "util/u_memory.h" +#include "util/u_handle_table.h" #define HANDLE_TABLE_INITIAL_SIZE 16 diff --git a/src/gallium/auxiliary/util/u_hash_table.c b/src/gallium/auxiliary/util/u_hash_table.c index dd5eca7fca..0bc8de9632 100644 --- a/src/gallium/auxiliary/util/u_hash_table.c +++ b/src/gallium/auxiliary/util/u_hash_table.c @@ -40,10 +40,11 @@ #include "pipe/p_compiler.h" #include "pipe/p_debug.h" -#include "pipe/p_util.h" #include "cso_cache/cso_hash.h" -#include "u_hash_table.h" + +#include "util/u_memory.h" +#include "util/u_hash_table.h" struct hash_table diff --git a/src/gallium/auxiliary/util/u_math.h b/src/gallium/auxiliary/util/u_math.h index a541d30a5d..9b4ca39371 100644 --- a/src/gallium/auxiliary/util/u_math.h +++ b/src/gallium/auxiliary/util/u_math.h @@ -40,8 +40,6 @@ #include "pipe/p_compiler.h" -#include "pipe/p_util.h" -#include "util/u_math.h" #ifdef __cplusplus @@ -49,6 +47,132 @@ extern "C" { #endif +#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) +__inline double ceil(double val) +{ + double ceil_val; + + if((val - (long) val) == 0) { + ceil_val = val; + } + else { + if(val > 0) { + ceil_val = (long) val + 1; + } + else { + ceil_val = (long) val; + } + } + + return ceil_val; +} + +#ifndef PIPE_SUBSYSTEM_WINDOWS_CE +__inline double floor(double val) +{ + double floor_val; + + if((val - (long) val) == 0) { + floor_val = val; + } + else { + if(val > 0) { + floor_val = (long) val; + } + else { + floor_val = (long) val - 1; + } + } + + return floor_val; +} +#endif + +#pragma function(pow) +__inline double __cdecl pow(double val, double exponent) +{ + /* XXX */ + assert(0); + return 0; +} + +#pragma function(log) +__inline double __cdecl log(double val) +{ + /* XXX */ + assert(0); + return 0; +} + +#pragma function(atan2) +__inline double __cdecl atan2(double val) +{ + /* XXX */ + assert(0); + return 0; +} +#else +#include +#include +#endif + + +#if defined(_MSC_VER) +#if _MSC_VER < 1400 && !defined(__cplusplus) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) + +static INLINE float cosf( float f ) +{ + return (float) cos( (double) f ); +} + +static INLINE float sinf( float f ) +{ + return (float) sin( (double) f ); +} + +static INLINE float ceilf( float f ) +{ + return (float) ceil( (double) f ); +} + +static INLINE float floorf( float f ) +{ + return (float) floor( (double) f ); +} + +static INLINE float powf( float f, float g ) +{ + return (float) pow( (double) f, (double) g ); +} + +static INLINE float sqrtf( float f ) +{ + return (float) sqrt( (double) f ); +} + +static INLINE float fabsf( float f ) +{ + return (float) fabs( (double) f ); +} + +static INLINE float logf( float f ) +{ + return (float) log( (double) f ); +} + +#else +/* Work-around an extra semi-colon in VS 2005 logf definition */ +#ifdef logf +#undef logf +#define logf(x) ((float)log((double)(x))) +#endif /* logf */ +#endif +#endif /* _MSC_VER */ + + + + + #define POW2_TABLE_SIZE 256 #define POW2_TABLE_SCALE ((float) (POW2_TABLE_SIZE-1)) extern float pow2_table[POW2_TABLE_SIZE]; @@ -59,6 +183,11 @@ extern void util_init_math(void); +union fi { + float f; + int i; + unsigned ui; +}; /** @@ -195,6 +324,113 @@ util_iround(float f) +#if defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86) +/** + * Find first bit set in word. Least significant bit is 1. + * Return 0 if no bits set. + */ +static INLINE +unsigned ffs( unsigned u ) +{ + unsigned i; + + if( u == 0 ) { + return 0; + } + + __asm bsf eax, [u] + __asm inc eax + __asm mov [i], eax + + return i; +} +#endif + + +/** + * Return float bits. + */ +static INLINE unsigned +fui( float f ) +{ + union fi fi; + fi.f = f; + return fi.ui; +} + + + +static INLINE float +ubyte_to_float(ubyte ub) +{ + return (float) ub * (1.0f / 255.0f); +} + + +/** + * Convert float in [0,1] to ubyte in [0,255] with clamping. + */ +static INLINE ubyte +float_to_ubyte(float f) +{ + const int ieee_0996 = 0x3f7f0000; /* 0.996 or so */ + union fi tmp; + + tmp.f = f; + if (tmp.i < 0) { + return (ubyte) 0; + } + else if (tmp.i >= ieee_0996) { + return (ubyte) 255; + } + else { + tmp.f = tmp.f * (255.0f/256.0f) + 32768.0f; + return (ubyte) tmp.i; + } +} + + + +#define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) + +#define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) +#define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) + + +static INLINE int +align(int value, int alignment) +{ + return (value + alignment - 1) & ~(alignment - 1); +} + + +#ifndef COPY_4V +#define COPY_4V( DST, SRC ) \ +do { \ + (DST)[0] = (SRC)[0]; \ + (DST)[1] = (SRC)[1]; \ + (DST)[2] = (SRC)[2]; \ + (DST)[3] = (SRC)[3]; \ +} while (0) +#endif + + +#ifndef COPY_4FV +#define COPY_4FV( DST, SRC ) COPY_4V(DST, SRC) +#endif + + +#ifndef ASSIGN_4V +#define ASSIGN_4V( DST, V0, V1, V2, V3 ) \ +do { \ + (DST)[0] = (V0); \ + (DST)[1] = (V1); \ + (DST)[2] = (V2); \ + (DST)[3] = (V3); \ +} while (0) +#endif + + #ifdef __cplusplus } #endif diff --git a/src/gallium/auxiliary/util/u_memory.h b/src/gallium/auxiliary/util/u_memory.h new file mode 100644 index 0000000000..148a5cb997 --- /dev/null +++ b/src/gallium/auxiliary/util/u_memory.h @@ -0,0 +1,222 @@ +/************************************************************************** + * + * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + +/** + * Memory functions + */ + + +#ifndef U_MEMORY_H +#define U_MEMORY_H + + +#include "util/u_pointer.h" + + + /* Define ENOMEM for WINCE */ +#if (_WIN32_WCE < 600) +#ifndef ENOMEM +#define ENOMEM 12 +#endif +#endif + + + +#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && defined(DEBUG) + +/* memory debugging */ + +#include "p_debug.h" + +#define MALLOC( _size ) \ + debug_malloc( __FILE__, __LINE__, __FUNCTION__, _size ) +#define CALLOC( _count, _size ) \ + debug_calloc(__FILE__, __LINE__, __FUNCTION__, _count, _size ) +#define FREE( _ptr ) \ + debug_free( __FILE__, __LINE__, __FUNCTION__, _ptr ) +#define REALLOC( _ptr, _old_size, _size ) \ + debug_realloc( __FILE__, __LINE__, __FUNCTION__, _ptr, _old_size, _size ) + +#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) + +void * __stdcall +EngAllocMem( + unsigned long Flags, + unsigned long MemSize, + unsigned long Tag ); + +void __stdcall +EngFreeMem( + void *Mem ); + +#define MALLOC( _size ) EngAllocMem( 0, _size, 'D3AG' ) +#define _FREE( _ptr ) EngFreeMem( _ptr ) + +#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) + +void * +ExAllocatePool( + unsigned long PoolType, + size_t NumberOfBytes); + +void +ExFreePool(void *P); + +#define MALLOC(_size) ExAllocatePool(0, _size) +#define _FREE(_ptr) ExFreePool(_ptr) + +#else + +#define MALLOC( SIZE ) malloc( SIZE ) +#define CALLOC( COUNT, SIZE ) calloc( COUNT, SIZE ) +#define FREE( PTR ) free( PTR ) +#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) + +#endif + + +#ifndef CALLOC +static INLINE void * +CALLOC( unsigned count, unsigned size ) +{ + void *ptr = MALLOC( count * size ); + if( ptr ) { + memset( ptr, 0, count * size ); + } + return ptr; +} +#endif /* !CALLOC */ + +#ifndef FREE +static INLINE void +FREE( void *ptr ) +{ + if( ptr ) { + _FREE( ptr ); + } +} +#endif /* !FREE */ + +#ifndef REALLOC +static INLINE void * +REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) +{ + void *new_ptr = NULL; + + if (new_size != 0) { + unsigned copy_size = old_size < new_size ? old_size : new_size; + new_ptr = MALLOC( new_size ); + if (new_ptr && old_ptr && copy_size) { + memcpy( new_ptr, old_ptr, copy_size ); + } + } + + FREE( old_ptr ); + return new_ptr; +} +#endif /* !REALLOC */ + + +#define MALLOC_STRUCT(T) (struct T *) MALLOC(sizeof(struct T)) + +#define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) + + +/** + * Return memory on given byte alignment + */ +static INLINE void * +align_malloc(size_t bytes, uint alignment) +{ +#if defined(HAVE_POSIX_MEMALIGN) + void *mem; + alignment = (alignment + (uint)sizeof(void*) - 1) & ~((uint)sizeof(void*) - 1); + if(posix_memalign(& mem, alignment, bytes) != 0) + return NULL; + return mem; +#else + char *ptr, *buf; + + assert( alignment > 0 ); + + ptr = (char *) MALLOC(bytes + alignment + sizeof(void *)); + if (!ptr) + return NULL; + + buf = (char *) align_pointer( ptr + sizeof(void *), alignment ); + *(char **)(buf - sizeof(void *)) = ptr; + + return buf; +#endif /* defined(HAVE_POSIX_MEMALIGN) */ +} + +/** + * Free memory returned by align_malloc(). + */ +static INLINE void +align_free(void *ptr) +{ +#if defined(HAVE_POSIX_MEMALIGN) + FREE(ptr); +#else + void **cubbyHole = (void **) ((char *) ptr - sizeof(void *)); + void *realAddr = *cubbyHole; + FREE(realAddr); +#endif /* defined(HAVE_POSIX_MEMALIGN) */ +} + + +/** + * Duplicate a block of memory. + */ +static INLINE void * +mem_dup(const void *src, uint size) +{ + void *dup = MALLOC(size); + if (dup) + memcpy(dup, src, size); + return dup; +} + + +/** + * Number of elements in an array. + */ +#ifndef Elements +#define Elements(x) (sizeof(x)/sizeof((x)[0])) +#endif + + +/** + * Offset of a field in a struct, in bytes. + */ +#define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER)) + + + +#endif /* U_MEMORY_H */ diff --git a/src/gallium/auxiliary/util/u_mm.c b/src/gallium/auxiliary/util/u_mm.c index b49ae074e0..0f51dd5977 100644 --- a/src/gallium/auxiliary/util/u_mm.c +++ b/src/gallium/auxiliary/util/u_mm.c @@ -24,9 +24,9 @@ #include "pipe/p_compiler.h" -#include "pipe/p_util.h" #include "pipe/p_debug.h" +#include "util/u_memory.h" #include "util/u_mm.h" diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h index 06abb34d5a..39e4ae9d07 100644 --- a/src/gallium/auxiliary/util/u_pack_color.h +++ b/src/gallium/auxiliary/util/u_pack_color.h @@ -37,6 +37,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_format.h" +#include "util/u_math.h" /** @@ -150,10 +151,10 @@ util_pack_color(const float rgba[4], enum pipe_format format, void *dest) if (pf_size_x(format) <= 8) { /* format uses 8-bit components or less */ - UNCLAMPED_FLOAT_TO_UBYTE(r, rgba[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, rgba[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, rgba[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, rgba[3]); + r = float_to_ubyte(rgba[0]); + g = float_to_ubyte(rgba[1]); + b = float_to_ubyte(rgba[2]); + a = float_to_ubyte(rgba[3]); } switch (format) { @@ -286,4 +287,31 @@ util_pack_z(enum pipe_format format, double z) } +/** + * Pack 4 ubytes into a 4-byte word + */ +static INLINE unsigned +pack_ub4(ubyte b0, ubyte b1, ubyte b2, ubyte b3) +{ + return ((((unsigned int)b0) << 0) | + (((unsigned int)b1) << 8) | + (((unsigned int)b2) << 16) | + (((unsigned int)b3) << 24)); +} + + +/** + * Pack/convert 4 floats into one 4-byte word. + */ +static INLINE unsigned +pack_ui32_float4(float a, float b, float c, float d) +{ + return pack_ub4( float_to_ubyte(a), + float_to_ubyte(b), + float_to_ubyte(c), + float_to_ubyte(d) ); +} + + + #endif /* U_PACK_COLOR_H */ diff --git a/src/gallium/auxiliary/util/u_pointer.h b/src/gallium/auxiliary/util/u_pointer.h new file mode 100644 index 0000000000..e1af9f11cb --- /dev/null +++ b/src/gallium/auxiliary/util/u_pointer.h @@ -0,0 +1,107 @@ +/************************************************************************** + * + * Copyright 2007-2008 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef U_POINTER_H +#define U_POINTER_H + +#include "pipe/p_compiler.h" + +#ifdef __cplusplus +extern "C" { +#endif + +static INLINE intptr_t +pointer_to_intptr( const void *p ) +{ + union { + const void *p; + intptr_t i; + } pi; + pi.p = p; + return pi.i; +} + +static INLINE void * +intptr_to_pointer( intptr_t i ) +{ + union { + void *p; + intptr_t i; + } pi; + pi.i = i; + return pi.p; +} + +static INLINE uintptr_t +pointer_to_uintptr( const void *ptr ) +{ + union { + const void *p; + uintptr_t u; + } pu; + pu.p = ptr; + return pu.u; +} + +static INLINE void * +uintptr_to_pointer( uintptr_t u ) +{ + union { + void *p; + uintptr_t u; + } pu; + pu.u = u; + return pu.p; +} + +/** + * Return a pointer aligned to next multiple of N bytes. + */ +static INLINE void * +align_pointer( const void *unaligned, uintptr_t alignment ) +{ + uintptr_t aligned = (pointer_to_uintptr( unaligned ) + alignment - 1) & ~(alignment - 1); + return uintptr_to_pointer( aligned ); +} + + +/** + * Return a pointer aligned to next multiple of 16 bytes. + */ +static INLINE void * +align16( void *unaligned ) +{ + return align_pointer( unaligned, 16 ); +} + + + +#ifdef __cplusplus +} +#endif + +#endif /* U_POINTER_H */ diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c index 94e447b9d5..b31ab5415f 100644 --- a/src/gallium/auxiliary/util/u_rect.c +++ b/src/gallium/auxiliary/util/u_rect.c @@ -31,7 +31,6 @@ #include "pipe/p_defines.h" -#include "pipe/p_util.h" #include "pipe/p_format.h" #include "util/u_rect.h" diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c index c34fb6ee33..f06d13c2c4 100644 --- a/src/gallium/auxiliary/util/u_simple_shaders.c +++ b/src/gallium/auxiliary/util/u_simple_shaders.c @@ -37,10 +37,10 @@ #include "pipe/p_debug.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" #include "pipe/p_winsys.h" #include "pipe/p_shader_tokens.h" +#include "util/u_memory.h" #include "util/u_simple_shaders.h" #include "tgsi/tgsi_build.h" diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c new file mode 100644 index 0000000000..853c503f4f --- /dev/null +++ b/src/gallium/auxiliary/util/u_tile.c @@ -0,0 +1,1169 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +/** + * RGBA/float tile get/put functions. + * Usable both by drivers and state trackers. + * Surfaces should already be in a mapped state. + */ + + +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" + +#include "util/u_math.h" +#include "util/u_memory.h" +#include "util/u_rect.h" +#include "util/u_tile.h" + + +/** + * Move raw block of pixels from surface to user memory. + * This should be usable by any hw driver that has mappable surfaces. + */ +void +pipe_get_tile_raw(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *dst, int dst_stride) +{ + const void *src; + + if (dst_stride == 0) + dst_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + src = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ); + assert(src); + if(!src) + return; + + pipe_copy_rect(dst, &ps->block, dst_stride, 0, 0, w, h, src, ps->stride, x, y); + + pipe_surface_unmap(ps); +} + + +/** + * Move raw block of pixels from user memory to surface. + * This should be usable by any hw driver that has mappable surfaces. + */ +void +pipe_put_tile_raw(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *src, int src_stride) +{ + void *dst; + + if (src_stride == 0) + src_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + dst = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE); + assert(dst); + if(!dst) + return; + + pipe_copy_rect(dst, &ps->block, ps->stride, x, y, w, h, src, src_stride, 0, 0); + + pipe_surface_unmap(ps); +} + + + + +/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */ +#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) + +#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ + us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) + + + +/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ + +static void +a8r8g8b8_get_tile_rgba(const unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; + pRow[0] = ubyte_to_float((pixel >> 16) & 0xff); + pRow[1] = ubyte_to_float((pixel >> 8) & 0xff); + pRow[2] = ubyte_to_float((pixel >> 0) & 0xff); + pRow[3] = ubyte_to_float((pixel >> 24) & 0xff); + } + p += dst_stride; + } +} + + +static void +a8r8g8b8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + r = float_to_ubyte(pRow[0]); + g = float_to_ubyte(pRow[1]); + b = float_to_ubyte(pRow[2]); + a = float_to_ubyte(pRow[3]); + *dst++ = (a << 24) | (r << 16) | (g << 8) | b; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ + +static void +x8r8g8b8_get_tile_rgba(const unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; + pRow[0] = ubyte_to_float((pixel >> 16) & 0xff); + pRow[1] = ubyte_to_float((pixel >> 8) & 0xff); + pRow[2] = ubyte_to_float((pixel >> 0) & 0xff); + pRow[3] = ubyte_to_float(0xff); + } + p += dst_stride; + } +} + + +static void +x8r8g8b8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b; + r = float_to_ubyte(pRow[0]); + g = float_to_ubyte(pRow[1]); + b = float_to_ubyte(pRow[2]); + *dst++ = (0xff << 24) | (r << 16) | (g << 8) | b; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ + +static void +b8g8r8a8_get_tile_rgba(const unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const unsigned pixel = *src++; + pRow[0] = ubyte_to_float((pixel >> 8) & 0xff); + pRow[1] = ubyte_to_float((pixel >> 16) & 0xff); + pRow[2] = ubyte_to_float((pixel >> 24) & 0xff); + pRow[3] = ubyte_to_float((pixel >> 0) & 0xff); + } + p += dst_stride; + } +} + + +static void +b8g8r8a8_put_tile_rgba(unsigned *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + r = float_to_ubyte(pRow[0]); + g = float_to_ubyte(pRow[1]); + b = float_to_ubyte(pRow[2]); + a = float_to_ubyte(pRow[3]); + *dst++ = (b << 24) | (g << 16) | (r << 8) | a; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ + +static void +a1r5g5b5_get_tile_rgba(const ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); + pRow[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); + pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + pRow[3] = ((pixel >> 15) ) * 1.0f; + } + p += dst_stride; + } +} + + +static void +a1r5g5b5_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + r = float_to_ubyte(pRow[0]); + g = float_to_ubyte(pRow[1]); + b = float_to_ubyte(pRow[2]); + a = float_to_ubyte(pRow[3]); + r = r >> 3; /* 5 bits */ + g = g >> 3; /* 5 bits */ + b = b >> 3; /* 5 bits */ + a = a >> 7; /* 1 bit */ + *dst++ = (a << 15) | (r << 10) | (g << 5) | b; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ + +static void +a4r4g4b4_get_tile_rgba(const ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); + pRow[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); + pRow[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); + pRow[3] = ((pixel >> 12) ) * (1.0f / 15.0f); + } + p += dst_stride; + } +} + + +static void +a4r4g4b4_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, g, b, a; + r = float_to_ubyte(pRow[0]); + g = float_to_ubyte(pRow[1]); + b = float_to_ubyte(pRow[2]); + a = float_to_ubyte(pRow[3]); + r >>= 4; + g >>= 4; + b >>= 4; + a >>= 4; + *dst++ = (a << 12) | (r << 16) | (g << 4) | b; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_R5G6B5_UNORM ***/ + +static void +r5g6b5_get_tile_rgba(const ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + const ushort pixel = *src++; + pRow[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); + pRow[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); + pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); + pRow[3] = 1.0f; + } + p += dst_stride; + } +} + + +static void +r5g6b5_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); + uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); + uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); + *dst++ = (r << 11) | (g << 5) | (b); + } + p += src_stride; + } +} + + + +/*** PIPE_FORMAT_Z16_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z16_get_tile_rgba(const ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const float scale = 1.0f / 65535.0f; + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = *src++ * scale; + } + p += dst_stride; + } +} + + + + +/*** PIPE_FORMAT_L8_UNORM ***/ + +static void +l8_get_tile_rgba(const ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = ubyte_to_float(*src); + pRow[3] = 1.0; + } + p += dst_stride; + } +} + + +static void +l8_put_tile_rgba(ubyte *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r; + r = float_to_ubyte(pRow[0]); + *dst++ = r; + } + p += src_stride; + } +} + + + +/*** PIPE_FORMAT_A8_UNORM ***/ + +static void +a8_get_tile_rgba(const ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = 0.0; + pRow[3] = ubyte_to_float(*src); + } + p += dst_stride; + } +} + + +static void +a8_put_tile_rgba(ubyte *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned a; + a = float_to_ubyte(pRow[3]); + *dst++ = a; + } + p += src_stride; + } +} + + + +/*** PIPE_FORMAT_R16_SNORM ***/ + +static void +r16_get_tile_rgba(const short *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = SHORT_TO_FLOAT(src[0]); + pRow[1] = + pRow[2] = 0.0; + pRow[3] = 1.0; + } + p += dst_stride; + } +} + + +static void +r16_put_tile_rgba(short *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, dst++, pRow += 4) { + UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ + +static void +r16g16b16a16_get_tile_rgba(const short *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src += 4, pRow += 4) { + pRow[0] = SHORT_TO_FLOAT(src[0]); + pRow[1] = SHORT_TO_FLOAT(src[1]); + pRow[2] = SHORT_TO_FLOAT(src[2]); + pRow[3] = SHORT_TO_FLOAT(src[3]); + } + p += dst_stride; + } +} + + +static void +r16g16b16a16_put_tile_rgba(short *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, dst += 4, pRow += 4) { + UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); + UNCLAMPED_FLOAT_TO_SHORT(dst[1], pRow[1]); + UNCLAMPED_FLOAT_TO_SHORT(dst[2], pRow[2]); + UNCLAMPED_FLOAT_TO_SHORT(dst[3], pRow[3]); + } + p += src_stride; + } +} + + + +/*** PIPE_FORMAT_I8_UNORM ***/ + +static void +i8_get_tile_rgba(const ubyte *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, src++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = ubyte_to_float(*src); + } + p += dst_stride; + } +} + + +static void +i8_put_tile_rgba(ubyte *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r; + r = float_to_ubyte(pRow[0]); + *dst++ = r; + } + p += src_stride; + } +} + + +/*** PIPE_FORMAT_A8L8_UNORM ***/ + +static void +a8l8_get_tile_rgba(const ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + ushort p = *src++; + pRow[0] = + pRow[1] = + pRow[2] = ubyte_to_float(p & 0xff); + pRow[3] = ubyte_to_float(p >> 8); + } + p += dst_stride; + } +} + + +static void +a8l8_put_tile_rgba(ushort *dst, + unsigned w, unsigned h, + const float *p, + unsigned src_stride) +{ + unsigned i, j; + + for (i = 0; i < h; i++) { + const float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + unsigned r, a; + r = float_to_ubyte(pRow[0]); + a = float_to_ubyte(pRow[3]); + *dst++ = (a << 8) | r; + } + p += src_stride; + } +} + + + + +/*** PIPE_FORMAT_Z32_UNORM ***/ + +/** + * Return each Z value as four floats in [0,1]. + */ +static void +z32_get_tile_rgba(const unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const double scale = 1.0 / (double) 0xffffffff; + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (*src++ * scale); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_S8Z24_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +s8z24_get_tile_rgba(const unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (scale * (*src++ & 0xffffff)); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_Z24S8_UNORM ***/ + +/** + * Return Z component as four float in [0,1]. Stencil part ignored. + */ +static void +z24s8_get_tile_rgba(const unsigned *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride) +{ + const double scale = 1.0 / ((1 << 24) - 1); + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + for (j = 0; j < w; j++, pRow += 4) { + pRow[0] = + pRow[1] = + pRow[2] = + pRow[3] = (float) (scale * (*src++ >> 8)); + } + p += dst_stride; + } +} + + +/*** PIPE_FORMAT_YCBCR / PIPE_FORMAT_YCBCR_REV ***/ + +/** + * Convert YCbCr (or YCrCb) to RGBA. + */ +static void +ycbcr_get_tile_rgba(const ushort *src, + unsigned w, unsigned h, + float *p, + unsigned dst_stride, + boolean rev) +{ + const float scale = 1.0f / 255.0f; + unsigned i, j; + + for (i = 0; i < h; i++) { + float *pRow = p; + /* do two texels at a time */ + for (j = 0; j < (w & ~1); j += 2, src += 2) { + const ushort t0 = src[0]; + const ushort t1 = src[1]; + const ubyte y0 = (t0 >> 8) & 0xff; /* luminance */ + const ubyte y1 = (t1 >> 8) & 0xff; /* luminance */ + ubyte cb, cr; + float r, g, b; + + if (rev) { + cb = t1 & 0xff; /* chroma U */ + cr = t0 & 0xff; /* chroma V */ + } + else { + cb = t0 & 0xff; /* chroma U */ + cr = t1 & 0xff; /* chroma V */ + } + + /* even pixel: y0,cr,cb */ + r = 1.164f * (y0-16) + 1.596f * (cr-128); + g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128); + b = 1.164f * (y0-16) + 2.018f * (cb-128); + pRow[0] = r * scale; + pRow[1] = g * scale; + pRow[2] = b * scale; + pRow[3] = 1.0f; + pRow += 4; + + /* odd pixel: use y1,cr,cb */ + r = 1.164f * (y1-16) + 1.596f * (cr-128); + g = 1.164f * (y1-16) - 0.813f * (cr-128) - 0.391f * (cb-128); + b = 1.164f * (y1-16) + 2.018f * (cb-128); + pRow[0] = r * scale; + pRow[1] = g * scale; + pRow[2] = b * scale; + pRow[3] = 1.0f; + pRow += 4; + + } + /* do the last texel */ + if (w & 1) { + const ushort t0 = src[0]; + const ushort t1 = src[1]; + const ubyte y0 = (t0 >> 8) & 0xff; /* luminance */ + ubyte cb, cr; + float r, g, b; + + if (rev) { + cb = t1 & 0xff; /* chroma U */ + cr = t0 & 0xff; /* chroma V */ + } + else { + cb = t0 & 0xff; /* chroma U */ + cr = t1 & 0xff; /* chroma V */ + } + + /* even pixel: y0,cr,cb */ + r = 1.164f * (y0-16) + 1.596f * (cr-128); + g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128); + b = 1.164f * (y0-16) + 2.018f * (cb-128); + pRow[0] = r * scale; + pRow[1] = g * scale; + pRow[2] = b * scale; + pRow[3] = 1.0f; + pRow += 4; + } + p += dst_stride; + } +} + + +void +pipe_tile_raw_to_rgba(enum pipe_format format, + void *src, + uint w, uint h, + float *dst, unsigned dst_stride) +{ + switch (format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_X8R8G8B8_UNORM: + x8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + a1r5g5b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_A4R4G4B4_UNORM: + a4r4g4b4_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g6b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_L8_UNORM: + l8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_A8_UNORM: + a8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_I8_UNORM: + i8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_A8L8_UNORM: + a8l8_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_R16_SNORM: + r16_get_tile_rgba((short *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_get_tile_rgba((short *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_Z16_UNORM: + z16_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_Z32_UNORM: + z32_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: + s8z24_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_Z24S8_UNORM: + z24s8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); + break; + case PIPE_FORMAT_YCBCR: + ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, FALSE); + break; + case PIPE_FORMAT_YCBCR_REV: + ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, TRUE); + break; + default: + assert(0); + } +} + + +void +pipe_get_tile_rgba(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p) +{ + unsigned dst_stride = w * 4; + void *packed; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); + + if (!packed) + return; + + if(ps->format == PIPE_FORMAT_YCBCR || ps->format == PIPE_FORMAT_YCBCR_REV) + assert((x & 1) == 0); + + pipe_get_tile_raw(ps, x, y, w, h, packed, 0); + + pipe_tile_raw_to_rgba(ps->format, packed, w, h, p, dst_stride); + + FREE(packed); +} + + +void +pipe_put_tile_rgba(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p) +{ + unsigned src_stride = w * 4; + void *packed; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); + + if (!packed) + return; + + switch (ps->format) { + case PIPE_FORMAT_A8R8G8B8_UNORM: + a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_X8R8G8B8_UNORM: + x8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_B8G8R8A8_UNORM: + b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_A1R5G5B5_UNORM: + a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_R5G6B5_UNORM: + r5g6b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_R8G8B8A8_UNORM: + assert(0); + break; + case PIPE_FORMAT_A4R4G4B4_UNORM: + a4r4g4b4_put_tile_rgba((ushort *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_L8_UNORM: + l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_A8_UNORM: + a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_I8_UNORM: + i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_A8L8_UNORM: + a8l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_R16_SNORM: + r16_put_tile_rgba((short *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_R16G16B16A16_SNORM: + r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); + break; + case PIPE_FORMAT_Z16_UNORM: + /*z16_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_Z32_UNORM: + /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: + /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ + break; + case PIPE_FORMAT_Z24S8_UNORM: + /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ + break; + default: + assert(0); + } + + pipe_put_tile_raw(ps, x, y, w, h, packed, 0); + + FREE(packed); +} + + +/** + * Get a block of Z values, converted to 32-bit range. + */ +void +pipe_get_tile_z(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + uint *z) +{ + const uint dstStride = w; + ubyte *map; + uint *pDest = z; + uint i, j; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ); + if (!map) { + assert(0); + return; + } + + switch (ps->format) { + case PIPE_FORMAT_Z32_UNORM: + { + const uint *pSrc + = (const uint *)(map + y * ps->stride + x*4); + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, 4 * w); + pDest += dstStride; + pSrc += ps->stride/4; + } + } + break; + case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: + { + const uint *pSrc + = (const uint *)(map + y * ps->stride + x*4); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 24-bit Z to 32-bit Z */ + pDest[j] = (pSrc[j] << 8) | (pSrc[j] & 0xff); + } + pDest += dstStride; + pSrc += ps->stride/4; + } + } + break; + case PIPE_FORMAT_Z16_UNORM: + { + const ushort *pSrc + = (const ushort *)(map + y * ps->stride + x*2); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 16-bit Z to 32-bit Z */ + pDest[j] = (pSrc[j] << 16) | pSrc[j]; + } + pDest += dstStride; + pSrc += ps->stride/2; + } + } + break; + default: + assert(0); + } + + pipe_surface_unmap(ps); +} + + +void +pipe_put_tile_z(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const uint *zSrc) +{ + const uint srcStride = w; + const uint *pSrc = zSrc; + ubyte *map; + uint i, j; + + if (pipe_clip_tile(x, y, &w, &h, ps)) + return; + + map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE); + if (!map) { + assert(0); + return; + } + + switch (ps->format) { + case PIPE_FORMAT_Z32_UNORM: + { + uint *pDest = (uint *) (map + y * ps->stride + x*4); + for (i = 0; i < h; i++) { + memcpy(pDest, pSrc, 4 * w); + pDest += ps->stride/4; + pSrc += srcStride; + } + } + break; + case PIPE_FORMAT_S8Z24_UNORM: + case PIPE_FORMAT_X8Z24_UNORM: + { + uint *pDest = (uint *) (map + y * ps->stride + x*4); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 32-bit Z to 24-bit Z (0 stencil) */ + pDest[j] = pSrc[j] >> 8; + } + pDest += ps->stride/4; + pSrc += srcStride; + } + } + break; + case PIPE_FORMAT_Z16_UNORM: + { + ushort *pDest = (ushort *) (map + y * ps->stride + x*2); + for (i = 0; i < h; i++) { + for (j = 0; j < w; j++) { + /* convert 32-bit Z to 16-bit Z */ + pDest[j] = pSrc[j] >> 16; + } + pDest += ps->stride/2; + pSrc += srcStride; + } + } + break; + default: + assert(0); + } + + pipe_surface_unmap(ps); +} + + diff --git a/src/gallium/auxiliary/util/u_tile.h b/src/gallium/auxiliary/util/u_tile.h new file mode 100644 index 0000000000..a8ac805308 --- /dev/null +++ b/src/gallium/auxiliary/util/u_tile.h @@ -0,0 +1,101 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + +#ifndef P_TILE_H +#define P_TILE_H + +#include "pipe/p_compiler.h" + +struct pipe_surface; + + +/** + * Clip tile against surface dims. + * \return TRUE if tile is totally clipped, FALSE otherwise + */ +static INLINE boolean +pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps) +{ + if (x >= ps->width) + return TRUE; + if (y >= ps->height) + return TRUE; + if (x + *w > ps->width) + *w = ps->width - x; + if (y + *h > ps->height) + *h = ps->height - y; + return FALSE; +} + +#ifdef __cplusplus +extern "C" { +#endif + +void +pipe_get_tile_raw(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + void *p, int dst_stride); + +void +pipe_put_tile_raw(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const void *p, int src_stride); + + +void +pipe_get_tile_rgba(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + float *p); + +void +pipe_put_tile_rgba(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const float *p); + + +void +pipe_get_tile_z(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + uint *z); + +void +pipe_put_tile_z(struct pipe_surface *ps, + uint x, uint y, uint w, uint h, + const uint *z); + +void +pipe_tile_raw_to_rgba(enum pipe_format format, + void *src, + uint w, uint h, + float *dst, unsigned dst_stride); + + +#ifdef __cplusplus +} +#endif + +#endif diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h index f430e88b9c..6bace0bb11 100644 --- a/src/gallium/drivers/cell/common.h +++ b/src/gallium/drivers/cell/common.h @@ -34,7 +34,6 @@ #define CELL_COMMON_H #include "pipe/p_compiler.h" -#include "pipe/p_util.h" #include "pipe/p_format.h" #include "pipe/p_state.h" diff --git a/src/gallium/drivers/cell/ppu/cell_clear.c b/src/gallium/drivers/cell/ppu/cell_clear.c index 3ffe09add6..cee0917b63 100644 --- a/src/gallium/drivers/cell/ppu/cell_clear.c +++ b/src/gallium/drivers/cell/ppu/cell_clear.c @@ -34,7 +34,7 @@ #include #include #include "pipe/p_inlines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "cell/common.h" #include "cell_clear.h" #include "cell_context.h" diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index 12eb5aa254..5af95a3c10 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -35,7 +35,7 @@ #include "pipe/p_defines.h" #include "pipe/p_format.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_winsys.h" #include "pipe/p_screen.h" diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 67b87f16d7..971d65d09e 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -30,7 +30,7 @@ * Brian Paul */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "draw/draw_context.h" #include "cell_context.h" diff --git a/src/gallium/drivers/cell/ppu/cell_render.c b/src/gallium/drivers/cell/ppu/cell_render.c index b663b37622..dd25ae880e 100644 --- a/src/gallium/drivers/cell/ppu/cell_render.c +++ b/src/gallium/drivers/cell/ppu/cell_render.c @@ -33,7 +33,7 @@ #include "cell_context.h" #include "cell_render.h" #include "cell_spu.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "draw/draw_private.h" diff --git a/src/gallium/drivers/cell/ppu/cell_screen.c b/src/gallium/drivers/cell/ppu/cell_screen.c index 2bf441a0c5..139b3719b6 100644 --- a/src/gallium/drivers/cell/ppu/cell_screen.c +++ b/src/gallium/drivers/cell/ppu/cell_screen.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_winsys.h" #include "pipe/p_defines.h" #include "pipe/p_screen.h" diff --git a/src/gallium/drivers/cell/ppu/cell_state_derived.c b/src/gallium/drivers/cell/ppu/cell_state_derived.c index 5480534ad9..8ab938a02a 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_derived.c +++ b/src/gallium/drivers/cell/ppu/cell_state_derived.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" #include "draw/draw_vertex.h" diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 9cae67f091..3646a0ee4f 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "cell_context.h" #include "cell_state.h" #include "cell_state_emit.h" diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c index f5707f2bb8..cd96b317fa 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -26,7 +26,7 @@ **************************************************************************/ #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "draw/draw_context.h" diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c index 01ffa31c2c..2d31ad89a6 100644 --- a/src/gallium/drivers/cell/ppu/cell_surface.c +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -26,7 +26,7 @@ **************************************************************************/ #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "util/p_tile.h" diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index 533b64227d..1add81373d 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -33,7 +33,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_winsys.h" #include "cell_context.h" diff --git a/src/gallium/drivers/cell/ppu/cell_winsys.c b/src/gallium/drivers/cell/ppu/cell_winsys.c index ebabce3c8f..d570bbd2f9 100644 --- a/src/gallium/drivers/cell/ppu/cell_winsys.c +++ b/src/gallium/drivers/cell/ppu/cell_winsys.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "cell_winsys.h" diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c index 42e5022f30..89c61136a4 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.c +++ b/src/gallium/drivers/cell/spu/spu_exec.c @@ -63,7 +63,6 @@ #include "pipe/p_compiler.h" #include "pipe/p_state.h" -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_util.h" diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index ab4ff8160a..8944ef171e 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -32,7 +32,6 @@ #include #include "pipe/p_compiler.h" #include "pipe/p_format.h" -#include "pipe/p_util.h" #include "spu_colorpack.h" #include "spu_main.h" #include "spu_texture.h" diff --git a/src/gallium/drivers/cell/spu/spu_util.c b/src/gallium/drivers/cell/spu/spu_util.c index 74ab2bbd1f..dbcf4b0eb9 100644 --- a/src/gallium/drivers/cell/spu/spu_util.c +++ b/src/gallium/drivers/cell/spu/spu_util.c @@ -1,4 +1,3 @@ -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" //#include "tgsi_build.h" diff --git a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c index 219fd90cc0..26f2363749 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_fetch.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_fetch.c @@ -32,7 +32,6 @@ * Ian Romanick */ -#include "pipe/p_util.h" #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" #include "spu_exec.h" diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.c b/src/gallium/drivers/cell/spu/spu_vertex_shader.c index 3119a78c06..a1e81975e6 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_shader.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.c @@ -34,7 +34,6 @@ #include -#include "pipe/p_util.h" #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" #include "spu_vertex_shader.h" diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c index 014a3e31d5..10c4ffc209 100644 --- a/src/gallium/drivers/failover/fo_context.c +++ b/src/gallium/drivers/failover/fo_context.c @@ -28,7 +28,7 @@ #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_context.h" #include "fo_context.h" diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c index e2bf5ab678..c6776716a2 100644 --- a/src/gallium/drivers/i915simple/i915_context.c +++ b/src/gallium/drivers/i915simple/i915_context.c @@ -35,7 +35,7 @@ #include "draw/draw_context.h" #include "pipe/p_defines.h" #include "pipe/p_winsys.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_screen.h" diff --git a/src/gallium/drivers/i915simple/i915_debug_fp.c b/src/gallium/drivers/i915simple/i915_debug_fp.c index c024a051a5..48be3e1472 100644 --- a/src/gallium/drivers/i915simple/i915_debug_fp.c +++ b/src/gallium/drivers/i915simple/i915_debug_fp.c @@ -29,7 +29,7 @@ #include "i915_reg.h" #include "i915_debug.h" #include "pipe/p_winsys.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" static void diff --git a/src/gallium/drivers/i915simple/i915_fpc.h b/src/gallium/drivers/i915simple/i915_fpc.h index 80a9576304..2f0f99d046 100644 --- a/src/gallium/drivers/i915simple/i915_fpc.h +++ b/src/gallium/drivers/i915simple/i915_fpc.h @@ -29,7 +29,6 @@ #ifndef I915_FPC_H #define I915_FPC_H -#include "pipe/p_util.h" #include "i915_context.h" #include "i915_reg.h" diff --git a/src/gallium/drivers/i915simple/i915_fpc_translate.c b/src/gallium/drivers/i915simple/i915_fpc_translate.c index 64432982c4..34b4a846c1 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_translate.c +++ b/src/gallium/drivers/i915simple/i915_fpc_translate.c @@ -33,6 +33,8 @@ #include "i915_fpc.h" #include "pipe/p_shader_tokens.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "util/u_string.h" #include "tgsi/tgsi_parse.h" #include "tgsi/tgsi_dump.h" diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c index 9ffa460138..d194c2fb15 100644 --- a/src/gallium/drivers/i915simple/i915_prim_emit.c +++ b/src/gallium/drivers/i915simple/i915_prim_emit.c @@ -27,7 +27,9 @@ #include "draw/draw_pipe.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" +#include "util/u_pack_color.h" #include "i915_context.h" #include "i915_winsys.h" diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c index aef3682bbf..e4ece55098 100644 --- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c +++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c @@ -41,9 +41,10 @@ #include "draw/draw_context.h" #include "draw/draw_vbuf.h" #include "pipe/p_debug.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "i915_context.h" #include "i915_reg.h" diff --git a/src/gallium/drivers/i915simple/i915_screen.c b/src/gallium/drivers/i915simple/i915_screen.c index 0afa17bed8..e9e40c3f0b 100644 --- a/src/gallium/drivers/i915simple/i915_screen.c +++ b/src/gallium/drivers/i915simple/i915_screen.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_winsys.h" #include "util/u_string.h" diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c index e8521b385e..d2487d8277 100644 --- a/src/gallium/drivers/i915simple/i915_state.c +++ b/src/gallium/drivers/i915simple/i915_state.c @@ -31,8 +31,9 @@ #include "draw/draw_context.h" #include "pipe/p_winsys.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "tgsi/tgsi_parse.h" #include "i915_context.h" diff --git a/src/gallium/drivers/i915simple/i915_state_derived.c b/src/gallium/drivers/i915simple/i915_state_derived.c index 4daccec6e0..488615067c 100644 --- a/src/gallium/drivers/i915simple/i915_state_derived.c +++ b/src/gallium/drivers/i915simple/i915_state_derived.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" #include "draw/draw_vertex.h" diff --git a/src/gallium/drivers/i915simple/i915_state_dynamic.c b/src/gallium/drivers/i915simple/i915_state_dynamic.c index 8cfbdddd19..86126a5a15 100644 --- a/src/gallium/drivers/i915simple/i915_state_dynamic.c +++ b/src/gallium/drivers/i915simple/i915_state_dynamic.c @@ -30,7 +30,9 @@ #include "i915_context.h" #include "i915_reg.h" #include "i915_state.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" +#include "util/u_pack_color.h" #define FILE_DEBUG_FLAG DEBUG_STATE diff --git a/src/gallium/drivers/i915simple/i915_state_immediate.c b/src/gallium/drivers/i915simple/i915_state_immediate.c index 2501f2d7cb..8c16bb4e27 100644 --- a/src/gallium/drivers/i915simple/i915_state_immediate.c +++ b/src/gallium/drivers/i915simple/i915_state_immediate.c @@ -33,7 +33,7 @@ #include "i915_context.h" #include "i915_state.h" #include "i915_reg.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" /* All state expressable with the LOAD_STATE_IMMEDIATE_1 packet. diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c index 7868f21ca6..c09c10601b 100644 --- a/src/gallium/drivers/i915simple/i915_state_sampler.c +++ b/src/gallium/drivers/i915simple/i915_state_sampler.c @@ -27,7 +27,7 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "i915_state_inlines.h" #include "i915_context.h" diff --git a/src/gallium/drivers/i915simple/i915_surface.c b/src/gallium/drivers/i915simple/i915_surface.c index 17b5125e56..62f1926644 100644 --- a/src/gallium/drivers/i915simple/i915_surface.c +++ b/src/gallium/drivers/i915simple/i915_surface.c @@ -30,10 +30,9 @@ #include "i915_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "util/p_tile.h" +#include "util/u_tile.h" #include "util/u_rect.h" diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c index ca0fb8761b..32344da4d5 100644 --- a/src/gallium/drivers/i915simple/i915_texture.c +++ b/src/gallium/drivers/i915simple/i915_texture.c @@ -34,8 +34,9 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" #include "pipe/p_winsys.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "i915_context.h" #include "i915_texture.h" diff --git a/src/gallium/drivers/i965simple/brw_cc.c b/src/gallium/drivers/i965simple/brw_cc.c index 337e4f95f6..79d4150383 100644 --- a/src/gallium/drivers/i965simple/brw_cc.c +++ b/src/gallium/drivers/i965simple/brw_cc.c @@ -29,7 +29,8 @@ * Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "brw_context.h" #include "brw_state.h" @@ -232,8 +233,7 @@ static void upload_cc_unit( struct brw_context *brw ) cc.cc3.alpha_test_func = brw_translate_compare_func(brw->attribs.DepthStencil->alpha.func); - UNCLAMPED_FLOAT_TO_UBYTE(cc.cc7.alpha_ref.ub[0], - brw->attribs.DepthStencil->alpha.ref); + cc.cc7.alpha_ref.ub[0] = float_to_ubyte(brw->attribs.DepthStencil->alpha.ref); cc.cc3.alpha_test_format = BRW_ALPHATEST_FORMAT_UNORM8; } diff --git a/src/gallium/drivers/i965simple/brw_clip_state.c b/src/gallium/drivers/i965simple/brw_clip_state.c index ea5c05a279..8e78dd51be 100644 --- a/src/gallium/drivers/i965simple/brw_clip_state.c +++ b/src/gallium/drivers/i965simple/brw_clip_state.c @@ -32,7 +32,8 @@ #include "brw_context.h" #include "brw_state.h" #include "brw_defines.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" static void upload_clip_unit( struct brw_context *brw ) diff --git a/src/gallium/drivers/i965simple/brw_context.c b/src/gallium/drivers/i965simple/brw_context.c index 8326f7b9c4..96920df008 100644 --- a/src/gallium/drivers/i965simple/brw_context.c +++ b/src/gallium/drivers/i965simple/brw_context.c @@ -39,7 +39,7 @@ #include "pipe/p_winsys.h" #include "pipe/p_context.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_screen.h" diff --git a/src/gallium/drivers/i965simple/brw_curbe.c b/src/gallium/drivers/i965simple/brw_curbe.c index 52bbd525c1..824ee7fd6d 100644 --- a/src/gallium/drivers/i965simple/brw_curbe.c +++ b/src/gallium/drivers/i965simple/brw_curbe.c @@ -39,7 +39,8 @@ #include "brw_wm.h" #include "pipe/p_state.h" #include "pipe/p_winsys.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #define FILE_DEBUG_FLAG DEBUG_FALLBACKS diff --git a/src/gallium/drivers/i965simple/brw_draw_upload.c b/src/gallium/drivers/i965simple/brw_draw_upload.c index 9c0c78c236..7c20ea52af 100644 --- a/src/gallium/drivers/i965simple/brw_draw_upload.c +++ b/src/gallium/drivers/i965simple/brw_draw_upload.c @@ -33,6 +33,7 @@ #include "brw_context.h" #include "brw_state.h" + struct brw_array_state { union header_union header; diff --git a/src/gallium/drivers/i965simple/brw_gs_state.c b/src/gallium/drivers/i965simple/brw_gs_state.c index 3932e9e939..5b8016b2e9 100644 --- a/src/gallium/drivers/i965simple/brw_gs_state.c +++ b/src/gallium/drivers/i965simple/brw_gs_state.c @@ -34,7 +34,8 @@ #include "brw_context.h" #include "brw_state.h" #include "brw_defines.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" diff --git a/src/gallium/drivers/i965simple/brw_screen.c b/src/gallium/drivers/i965simple/brw_screen.c index fadfbf94ab..ab7cd624b2 100644 --- a/src/gallium/drivers/i965simple/brw_screen.c +++ b/src/gallium/drivers/i965simple/brw_screen.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_winsys.h" #include "util/u_string.h" diff --git a/src/gallium/drivers/i965simple/brw_sf_state.c b/src/gallium/drivers/i965simple/brw_sf_state.c index 9acd3ea61b..2a5de61c21 100644 --- a/src/gallium/drivers/i965simple/brw_sf_state.c +++ b/src/gallium/drivers/i965simple/brw_sf_state.c @@ -30,11 +30,12 @@ */ - #include "brw_context.h" #include "brw_state.h" #include "brw_defines.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" + static void upload_sf_vp(struct brw_context *brw) { diff --git a/src/gallium/drivers/i965simple/brw_shader_info.c b/src/gallium/drivers/i965simple/brw_shader_info.c index 30f37a99d4..86d877d7ef 100644 --- a/src/gallium/drivers/i965simple/brw_shader_info.c +++ b/src/gallium/drivers/i965simple/brw_shader_info.c @@ -1,7 +1,7 @@ #include "brw_context.h" #include "brw_state.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c index 27ca32843d..af46cb546f 100644 --- a/src/gallium/drivers/i965simple/brw_state.c +++ b/src/gallium/drivers/i965simple/brw_state.c @@ -31,7 +31,7 @@ #include "pipe/p_winsys.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_dump.h" diff --git a/src/gallium/drivers/i965simple/brw_state_batch.c b/src/gallium/drivers/i965simple/brw_state_batch.c index 35db76b594..43a1c89fc4 100644 --- a/src/gallium/drivers/i965simple/brw_state_batch.c +++ b/src/gallium/drivers/i965simple/brw_state_batch.c @@ -32,7 +32,7 @@ #include "brw_state.h" #include "brw_winsys.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" /* A facility similar to the data caching code above, which aims to * prevent identical commands being issued repeatedly. diff --git a/src/gallium/drivers/i965simple/brw_state_cache.c b/src/gallium/drivers/i965simple/brw_state_cache.c index b3a5124461..094248fa69 100644 --- a/src/gallium/drivers/i965simple/brw_state_cache.c +++ b/src/gallium/drivers/i965simple/brw_state_cache.c @@ -38,7 +38,7 @@ #include "brw_sf.h" #include "brw_gs.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" diff --git a/src/gallium/drivers/i965simple/brw_state_pool.c b/src/gallium/drivers/i965simple/brw_state_pool.c index f3174bfe0a..78d4c0e411 100644 --- a/src/gallium/drivers/i965simple/brw_state_pool.c +++ b/src/gallium/drivers/i965simple/brw_state_pool.c @@ -43,7 +43,8 @@ */ #include "pipe/p_winsys.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "brw_context.h" #include "brw_state.h" diff --git a/src/gallium/drivers/i965simple/brw_state_upload.c b/src/gallium/drivers/i965simple/brw_state_upload.c index e727601e1e..bac9161b5f 100644 --- a/src/gallium/drivers/i965simple/brw_state_upload.c +++ b/src/gallium/drivers/i965simple/brw_state_upload.c @@ -33,7 +33,7 @@ #include "brw_context.h" #include "brw_state.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" /* This is used to initialize brw->state.atoms[]. We could use this * list directly except for a single atom, brw_constant_buffer, which diff --git a/src/gallium/drivers/i965simple/brw_surface.c b/src/gallium/drivers/i965simple/brw_surface.c index 69da252285..b89756c47b 100644 --- a/src/gallium/drivers/i965simple/brw_surface.c +++ b/src/gallium/drivers/i965simple/brw_surface.c @@ -29,10 +29,9 @@ #include "brw_context.h" #include "brw_state.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "util/p_tile.h" +#include "util/u_tile.h" #include "util/u_rect.h" diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c index 9b6cf81723..05eda9d1f2 100644 --- a/src/gallium/drivers/i965simple/brw_tex_layout.c +++ b/src/gallium/drivers/i965simple/brw_tex_layout.c @@ -33,16 +33,16 @@ /* Code to layout images in a mipmap tree for i965. */ -#include "brw_tex_layout.h" - #include "pipe/p_state.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" - +#include "util/u_math.h" +#include "util/u_memory.h" #include "brw_context.h" +#include "brw_tex_layout.h" + #define FILE_DEBUG_FLAG DEBUG_TEXTURE diff --git a/src/gallium/drivers/i965simple/brw_vs_state.c b/src/gallium/drivers/i965simple/brw_vs_state.c index c73469929c..1eaff87892 100644 --- a/src/gallium/drivers/i965simple/brw_vs_state.c +++ b/src/gallium/drivers/i965simple/brw_vs_state.c @@ -34,7 +34,8 @@ #include "brw_state.h" #include "brw_defines.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" static void upload_vs_unit( struct brw_context *brw ) { diff --git a/src/gallium/drivers/i965simple/brw_wm.c b/src/gallium/drivers/i965simple/brw_wm.c index 7fc5f59a98..8de565b96c 100644 --- a/src/gallium/drivers/i965simple/brw_wm.c +++ b/src/gallium/drivers/i965simple/brw_wm.c @@ -35,7 +35,7 @@ #include "brw_wm.h" #include "brw_eu.h" #include "brw_state.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" diff --git a/src/gallium/drivers/i965simple/brw_wm_decl.c b/src/gallium/drivers/i965simple/brw_wm_decl.c index e6f1a44817..d50e66f613 100644 --- a/src/gallium/drivers/i965simple/brw_wm_decl.c +++ b/src/gallium/drivers/i965simple/brw_wm_decl.c @@ -2,7 +2,8 @@ #include "brw_context.h" #include "brw_eu.h" #include "brw_wm.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" diff --git a/src/gallium/drivers/i965simple/brw_wm_glsl.c b/src/gallium/drivers/i965simple/brw_wm_glsl.c index 6a4a5aef09..ab6410aa60 100644 --- a/src/gallium/drivers/i965simple/brw_wm_glsl.c +++ b/src/gallium/drivers/i965simple/brw_wm_glsl.c @@ -2,7 +2,8 @@ #include "brw_context.h" #include "brw_eu.h" #include "brw_wm.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "pipe/p_shader_tokens.h" #include "tgsi/tgsi_parse.h" diff --git a/src/gallium/drivers/i965simple/brw_wm_sampler_state.c b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c index b9eaee56ee..52b2909a65 100644 --- a/src/gallium/drivers/i965simple/brw_wm_sampler_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_sampler_state.c @@ -34,7 +34,8 @@ #include "brw_state.h" #include "brw_defines.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #define COMPAREFUNC_ALWAYS 0 diff --git a/src/gallium/drivers/i965simple/brw_wm_state.c b/src/gallium/drivers/i965simple/brw_wm_state.c index f3aa36b07f..37a9bf919c 100644 --- a/src/gallium/drivers/i965simple/brw_wm_state.c +++ b/src/gallium/drivers/i965simple/brw_wm_state.c @@ -34,7 +34,8 @@ #include "brw_state.h" #include "brw_defines.h" #include "brw_wm.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" /*********************************************************************** * WM unit - fragment programs and rasterization diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c index 9b1313bc83..dda90f760a 100644 --- a/src/gallium/drivers/softpipe/sp_context.c +++ b/src/gallium/drivers/softpipe/sp_context.c @@ -32,8 +32,8 @@ #include "draw/draw_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" #include "util/u_math.h" +#include "util/u_memory.h" #include "sp_clear.h" #include "sp_context.h" #include "sp_flush.h" diff --git a/src/gallium/drivers/softpipe/sp_fs_exec.c b/src/gallium/drivers/softpipe/sp_fs_exec.c index cc171bbc39..d0456731be 100644 --- a/src/gallium/drivers/softpipe/sp_fs_exec.c +++ b/src/gallium/drivers/softpipe/sp_fs_exec.c @@ -34,7 +34,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "tgsi/tgsi_exec.h" #include "tgsi/tgsi_parse.h" diff --git a/src/gallium/drivers/softpipe/sp_fs_llvm.c b/src/gallium/drivers/softpipe/sp_fs_llvm.c index 20226da78c..34adac5226 100644 --- a/src/gallium/drivers/softpipe/sp_fs_llvm.c +++ b/src/gallium/drivers/softpipe/sp_fs_llvm.c @@ -36,7 +36,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "tgsi/tgsi_sse2.h" diff --git a/src/gallium/drivers/softpipe/sp_fs_sse.c b/src/gallium/drivers/softpipe/sp_fs_sse.c index 8b7da7c747..35653a8e48 100644 --- a/src/gallium/drivers/softpipe/sp_fs_sse.c +++ b/src/gallium/drivers/softpipe/sp_fs_sse.c @@ -34,7 +34,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "tgsi/tgsi_exec.h" #include "tgsi/tgsi_sse2.h" diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c index 941ab62e00..038ff04d4f 100644 --- a/src/gallium/drivers/softpipe/sp_prim_setup.c +++ b/src/gallium/drivers/softpipe/sp_prim_setup.c @@ -41,7 +41,7 @@ #include "sp_prim_setup.h" #include "draw/draw_pipe.h" #include "draw/draw_vertex.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" /** * Triangle setup info (derived from draw_stage). diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c index e9fae951e0..425e13cd28 100644 --- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c +++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c @@ -43,6 +43,7 @@ #include "sp_setup.h" #include "draw/draw_context.h" #include "draw/draw_vbuf.h" +#include "util/u_memory.h" #define SP_MAX_VBUF_INDEXES 1024 diff --git a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c index 7a42b08ef5..7d3580fb4f 100644 --- a/src/gallium/drivers/softpipe/sp_quad_alpha_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_alpha_test.c @@ -7,7 +7,7 @@ #include "sp_headers.h" #include "sp_quad.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" static void diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c index 74c6bff84a..a834accb86 100644 --- a/src/gallium/drivers/softpipe/sp_quad_blend.c +++ b/src/gallium/drivers/softpipe/sp_quad_blend.c @@ -31,7 +31,8 @@ */ #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" @@ -128,15 +129,15 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) /* convert to ubyte */ for (j = 0; j < 4; j++) { /* loop over R,G,B,A channels */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][0], dest[j][0]); /* P0 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][1], dest[j][1]); /* P1 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][2], dest[j][2]); /* P2 */ - UNCLAMPED_FLOAT_TO_UBYTE(dst[j][3], dest[j][3]); /* P3 */ - - UNCLAMPED_FLOAT_TO_UBYTE(src[j][0], quadColor[j][0]); /* P0 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][1], quadColor[j][1]); /* P1 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][2], quadColor[j][2]); /* P2 */ - UNCLAMPED_FLOAT_TO_UBYTE(src[j][3], quadColor[j][3]); /* P3 */ + dst[j][0] = float_to_ubyte(dest[j][0]); /* P0 */ + dst[j][1] = float_to_ubyte(dest[j][1]); /* P1 */ + dst[j][2] = float_to_ubyte(dest[j][2]); /* P2 */ + dst[j][3] = float_to_ubyte(dest[j][3]); /* P3 */ + + src[j][0] = float_to_ubyte(quadColor[j][0]); /* P0 */ + src[j][1] = float_to_ubyte(quadColor[j][1]); /* P1 */ + src[j][2] = float_to_ubyte(quadColor[j][2]); /* P2 */ + src[j][3] = float_to_ubyte(quadColor[j][3]); /* P3 */ } switch (softpipe->blend->logicop_func) { @@ -209,10 +210,10 @@ logicop_quad(struct quad_stage *qs, struct quad_header *quad) } for (j = 0; j < 4; j++) { - quadColor[j][0] = UBYTE_TO_FLOAT(res[j][0]); - quadColor[j][1] = UBYTE_TO_FLOAT(res[j][1]); - quadColor[j][2] = UBYTE_TO_FLOAT(res[j][2]); - quadColor[j][3] = UBYTE_TO_FLOAT(res[j][3]); + quadColor[j][0] = ubyte_to_float(res[j][0]); + quadColor[j][1] = ubyte_to_float(res[j][1]); + quadColor[j][2] = ubyte_to_float(res[j][2]); + quadColor[j][3] = ubyte_to_float(res[j][3]); } } diff --git a/src/gallium/drivers/softpipe/sp_quad_bufloop.c b/src/gallium/drivers/softpipe/sp_quad_bufloop.c index b3db428ef1..92e9af09c1 100644 --- a/src/gallium/drivers/softpipe/sp_quad_bufloop.c +++ b/src/gallium/drivers/softpipe/sp_quad_bufloop.c @@ -1,5 +1,5 @@ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_colormask.c b/src/gallium/drivers/softpipe/sp_quad_colormask.c index 7fe080990b..f72f31db97 100644 --- a/src/gallium/drivers/softpipe/sp_quad_colormask.c +++ b/src/gallium/drivers/softpipe/sp_quad_colormask.c @@ -31,7 +31,8 @@ */ #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_coverage.c b/src/gallium/drivers/softpipe/sp_quad_coverage.c index dd5ebb2296..ad907ec25f 100644 --- a/src/gallium/drivers/softpipe/sp_quad_coverage.c +++ b/src/gallium/drivers/softpipe/sp_quad_coverage.c @@ -33,7 +33,7 @@ #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_quad.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c index 0c82692c6e..227cb2014e 100644 --- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c +++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c @@ -30,7 +30,7 @@ */ #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_earlyz.c b/src/gallium/drivers/softpipe/sp_quad_earlyz.c index 22ea99049f..5a66a86699 100644 --- a/src/gallium/drivers/softpipe/sp_quad_earlyz.c +++ b/src/gallium/drivers/softpipe/sp_quad_earlyz.c @@ -30,7 +30,7 @@ */ #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "sp_headers.h" #include "sp_quad.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c index 8c88c192f8..5499ba5361 100644 --- a/src/gallium/drivers/softpipe/sp_quad_fs.c +++ b/src/gallium/drivers/softpipe/sp_quad_fs.c @@ -35,7 +35,8 @@ * all the enabled attributes run contiguously. */ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "pipe/p_defines.h" #include "pipe/p_shader_tokens.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_occlusion.c b/src/gallium/drivers/softpipe/sp_quad_occlusion.c index 54254df1f1..db13e73ae3 100644 --- a/src/gallium/drivers/softpipe/sp_quad_occlusion.c +++ b/src/gallium/drivers/softpipe/sp_quad_occlusion.c @@ -33,7 +33,7 @@ #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_output.c b/src/gallium/drivers/softpipe/sp_quad_output.c index 40083138a4..b64646a449 100644 --- a/src/gallium/drivers/softpipe/sp_quad_output.c +++ b/src/gallium/drivers/softpipe/sp_quad_output.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_headers.h" #include "sp_surface.h" diff --git a/src/gallium/drivers/softpipe/sp_quad_stencil.c b/src/gallium/drivers/softpipe/sp_quad_stencil.c index b4c7e942fa..ce9562e07c 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stencil.c +++ b/src/gallium/drivers/softpipe/sp_quad_stencil.c @@ -10,7 +10,7 @@ #include "sp_tile_cache.h" #include "sp_quad.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" /** Only 8-bit stencil supported */ diff --git a/src/gallium/drivers/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c index f1e9b80e09..a39ecc2e9d 100644 --- a/src/gallium/drivers/softpipe/sp_quad_stipple.c +++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c @@ -7,7 +7,7 @@ #include "sp_headers.h" #include "sp_quad.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" /** diff --git a/src/gallium/drivers/softpipe/sp_query.c b/src/gallium/drivers/softpipe/sp_query.c index adf9ccf64c..2106ee1d23 100644 --- a/src/gallium/drivers/softpipe/sp_query.c +++ b/src/gallium/drivers/softpipe/sp_query.c @@ -32,7 +32,7 @@ #include "draw/draw_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_query.h" diff --git a/src/gallium/drivers/softpipe/sp_screen.c b/src/gallium/drivers/softpipe/sp_screen.c index f6b3d7ac24..9644dbd168 100644 --- a/src/gallium/drivers/softpipe/sp_screen.c +++ b/src/gallium/drivers/softpipe/sp_screen.c @@ -26,7 +26,7 @@ **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_winsys.h" #include "pipe/p_defines.h" #include "pipe/p_screen.h" diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c index c8c55fa6e8..87336ab6e3 100644 --- a/src/gallium/drivers/softpipe/sp_setup.c +++ b/src/gallium/drivers/softpipe/sp_setup.c @@ -42,9 +42,9 @@ #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vertex.h" -#include "pipe/p_util.h" #include "pipe/p_shader_tokens.h" #include "util/u_math.h" +#include "util/u_memory.h" #define DEBUG_VERTS 0 diff --git a/src/gallium/drivers/softpipe/sp_state_blend.c b/src/gallium/drivers/softpipe/sp_state_blend.c index 2d40d6bd8f..384fe559af 100644 --- a/src/gallium/drivers/softpipe/sp_state_blend.c +++ b/src/gallium/drivers/softpipe/sp_state_blend.c @@ -28,7 +28,7 @@ /* Authors: Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_state.h" diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c index f10a1fa471..6b6a4c3ff3 100644 --- a/src/gallium/drivers/softpipe/sp_state_derived.c +++ b/src/gallium/drivers/softpipe/sp_state_derived.c @@ -25,7 +25,8 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "pipe/p_shader_tokens.h" #include "draw/draw_context.h" #include "draw/draw_vertex.h" diff --git a/src/gallium/drivers/softpipe/sp_state_fs.c b/src/gallium/drivers/softpipe/sp_state_fs.c index 76fe6bfef9..1be461b3a4 100644 --- a/src/gallium/drivers/softpipe/sp_state_fs.c +++ b/src/gallium/drivers/softpipe/sp_state_fs.c @@ -30,7 +30,7 @@ #include "sp_fs.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" #include "pipe/p_shader_tokens.h" diff --git a/src/gallium/drivers/softpipe/sp_state_rasterizer.c b/src/gallium/drivers/softpipe/sp_state_rasterizer.c index 98e04352db..87b7219683 100644 --- a/src/gallium/drivers/softpipe/sp_state_rasterizer.c +++ b/src/gallium/drivers/softpipe/sp_state_rasterizer.c @@ -26,7 +26,7 @@ **************************************************************************/ #include "pipe/p_defines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_state.h" #include "draw/draw_context.h" diff --git a/src/gallium/drivers/softpipe/sp_state_sampler.c b/src/gallium/drivers/softpipe/sp_state_sampler.c index 033288a0aa..99a28c0d7e 100644 --- a/src/gallium/drivers/softpipe/sp_state_sampler.c +++ b/src/gallium/drivers/softpipe/sp_state_sampler.c @@ -29,7 +29,7 @@ * Brian Paul */ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "draw/draw_context.h" diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c index bfbae234f1..389aceb27c 100644 --- a/src/gallium/drivers/softpipe/sp_surface.c +++ b/src/gallium/drivers/softpipe/sp_surface.c @@ -26,10 +26,9 @@ **************************************************************************/ #include "pipe/p_defines.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "util/p_tile.h" +#include "util/u_tile.h" #include "util/u_rect.h" #include "sp_context.h" #include "sp_surface.h" diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c index 58a95d13e1..49250ec084 100644 --- a/src/gallium/drivers/softpipe/sp_tex_sample.c +++ b/src/gallium/drivers/softpipe/sp_tex_sample.c @@ -39,9 +39,9 @@ #include "sp_tile_cache.h" #include "pipe/p_context.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" #include "tgsi/tgsi_exec.h" #include "util/u_math.h" +#include "util/u_memory.h" /* diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index f775591352..3a737d6f72 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -33,8 +33,9 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" #include "pipe/p_winsys.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "sp_context.h" #include "sp_state.h" diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 57c12ffe33..b50c984513 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -32,9 +32,9 @@ * Brian Paul */ -#include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "util/p_tile.h" +#include "util/u_memory.h" +#include "util/u_tile.h" #include "sp_context.h" #include "sp_surface.h" #include "sp_texture.h" diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c index f16359e8ad..1dd7719379 100644 --- a/src/gallium/drivers/trace/tr_context.c +++ b/src/gallium/drivers/trace/tr_context.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_screen.h" #include "tr_dump.h" diff --git a/src/gallium/drivers/trace/tr_dump.c b/src/gallium/drivers/trace/tr_dump.c index 1613a626df..48032c1617 100644 --- a/src/gallium/drivers/trace/tr_dump.c +++ b/src/gallium/drivers/trace/tr_dump.c @@ -45,6 +45,8 @@ #endif #include "pipe/p_compiler.h" +#include "pipe/p_debug.h" +#include "util/u_memory.h" #include "util/u_string.h" #include "tr_stream.h" diff --git a/src/gallium/drivers/trace/tr_dump.h b/src/gallium/drivers/trace/tr_dump.h index 6ddc8fc15c..76a53731b3 100644 --- a/src/gallium/drivers/trace/tr_dump.h +++ b/src/gallium/drivers/trace/tr_dump.h @@ -35,7 +35,6 @@ #include "pipe/p_compiler.h" -#include "pipe/p_util.h" boolean trace_dump_trace_begin(void); diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index a6467ec35f..8789f86b1a 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -25,7 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "tr_dump.h" #include "tr_state.h" diff --git a/src/gallium/drivers/trace/tr_state.c b/src/gallium/drivers/trace/tr_state.c index 30ab5a8fdc..986d939e0c 100644 --- a/src/gallium/drivers/trace/tr_state.c +++ b/src/gallium/drivers/trace/tr_state.c @@ -27,6 +27,7 @@ #include "pipe/p_compiler.h" +#include "util/u_memory.h" #include "tgsi/tgsi_dump.h" #include "tr_dump.h" diff --git a/src/gallium/drivers/trace/tr_stream_stdc.c b/src/gallium/drivers/trace/tr_stream_stdc.c index 4c77e1c995..4c19ec0b24 100644 --- a/src/gallium/drivers/trace/tr_stream_stdc.c +++ b/src/gallium/drivers/trace/tr_stream_stdc.c @@ -36,7 +36,7 @@ #include -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "tr_stream.h" diff --git a/src/gallium/drivers/trace/tr_stream_wd.c b/src/gallium/drivers/trace/tr_stream_wd.c index b3b65f0971..704eb15bd7 100644 --- a/src/gallium/drivers/trace/tr_stream_wd.c +++ b/src/gallium/drivers/trace/tr_stream_wd.c @@ -37,7 +37,7 @@ #include #include -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "util/u_string.h" #include "tr_stream.h" diff --git a/src/gallium/drivers/trace/tr_texture.c b/src/gallium/drivers/trace/tr_texture.c index 99ba74d366..440a78704a 100644 --- a/src/gallium/drivers/trace/tr_texture.c +++ b/src/gallium/drivers/trace/tr_texture.c @@ -25,9 +25,9 @@ * **************************************************************************/ -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "util/u_hash_table.h" +#include "util/u_memory.h" #include "tr_screen.h" #include "tr_texture.h" diff --git a/src/gallium/drivers/trace/tr_winsys.c b/src/gallium/drivers/trace/tr_winsys.c index 2c7a6f893b..177835854e 100644 --- a/src/gallium/drivers/trace/tr_winsys.c +++ b/src/gallium/drivers/trace/tr_winsys.c @@ -25,8 +25,7 @@ * **************************************************************************/ -#include "pipe/p_util.h" -#include "pipe/p_state.h" +#include "util/u_memory.h" #include "util/u_hash_table.h" #include "tr_dump.h" diff --git a/src/gallium/include/pipe/p_util.h b/src/gallium/include/pipe/p_util.h deleted file mode 100644 index 4a3fca5962..0000000000 --- a/src/gallium/include/pipe/p_util.h +++ /dev/null @@ -1,460 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef P_UTIL_H -#define P_UTIL_H - -#include "p_config.h" -#include "p_compiler.h" -#include "p_debug.h" -#include "p_pointer.h" - -#if defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) -__inline double ceil(double val) -{ - double ceil_val; - - if((val - (long) val) == 0) { - ceil_val = val; - } else { - if(val > 0) { - ceil_val = (long) val + 1; - } else { - ceil_val = (long) val; - } - } - - return ceil_val; -} - -#ifndef PIPE_SUBSYSTEM_WINDOWS_CE -__inline double floor(double val) -{ - double floor_val; - - if((val - (long) val) == 0) { - floor_val = val; - } else { - if(val > 0) { - floor_val = (long) val; - } else { - floor_val = (long) val - 1; - } - } - - return floor_val; -} -#endif - -#pragma function(pow) -__inline double __cdecl pow(double val, double exponent) -{ - /* XXX */ - assert(0); - return 0; -} - -#pragma function(log) -__inline double __cdecl log(double val) -{ - /* XXX */ - assert(0); - return 0; -} - -#pragma function(atan2) -__inline double __cdecl atan2(double val) -{ - /* XXX */ - assert(0); - return 0; -} -#else -#include -#include -#endif - - /* Define ENOMEM for WINCE */ -#if (_WIN32_WCE < 600) -#ifndef ENOMEM -#define ENOMEM 12 -#endif -#endif - -#ifdef __cplusplus -extern "C" { -#endif - - -#if defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) && defined(DEBUG) - -/* memory debugging */ - -#include "p_debug.h" - -#define MALLOC( _size ) \ - debug_malloc( __FILE__, __LINE__, __FUNCTION__, _size ) -#define CALLOC( _count, _size ) \ - debug_calloc(__FILE__, __LINE__, __FUNCTION__, _count, _size ) -#define FREE( _ptr ) \ - debug_free( __FILE__, __LINE__, __FUNCTION__, _ptr ) -#define REALLOC( _ptr, _old_size, _size ) \ - debug_realloc( __FILE__, __LINE__, __FUNCTION__, _ptr, _old_size, _size ) - -#elif defined(PIPE_SUBSYSTEM_WINDOWS_DISPLAY) - -void * __stdcall -EngAllocMem( - unsigned long Flags, - unsigned long MemSize, - unsigned long Tag ); - -void __stdcall -EngFreeMem( - void *Mem ); - -#define MALLOC( _size ) EngAllocMem( 0, _size, 'D3AG' ) -#define _FREE( _ptr ) EngFreeMem( _ptr ) - -#elif defined(PIPE_SUBSYSTEM_WINDOWS_MINIPORT) - -void * -ExAllocatePool( - unsigned long PoolType, - size_t NumberOfBytes); - -void -ExFreePool(void *P); - -#define MALLOC(_size) ExAllocatePool(0, _size) -#define _FREE(_ptr) ExFreePool(_ptr) - -#else - -#define MALLOC( SIZE ) malloc( SIZE ) -#define CALLOC( COUNT, SIZE ) calloc( COUNT, SIZE ) -#define FREE( PTR ) free( PTR ) -#define REALLOC( OLDPTR, OLDSIZE, NEWSIZE ) realloc( OLDPTR, NEWSIZE ) - -#endif - - -#ifndef CALLOC -static INLINE void * -CALLOC( unsigned count, unsigned size ) -{ - void *ptr = MALLOC( count * size ); - if( ptr ) { - memset( ptr, 0, count * size ); - } - return ptr; -} -#endif /* !CALLOC */ - -#ifndef FREE -static INLINE void -FREE( void *ptr ) -{ - if( ptr ) { - _FREE( ptr ); - } -} -#endif /* !FREE */ - -#ifndef REALLOC -static INLINE void * -REALLOC( void *old_ptr, unsigned old_size, unsigned new_size ) -{ - void *new_ptr = NULL; - - if (new_size != 0) { - unsigned copy_size = old_size < new_size ? old_size : new_size; - new_ptr = MALLOC( new_size ); - if (new_ptr && old_ptr && copy_size) { - memcpy( new_ptr, old_ptr, copy_size ); - } - } - - FREE( old_ptr ); - return new_ptr; -} -#endif /* !REALLOC */ - - -#define MALLOC_STRUCT(T) (struct T *) MALLOC(sizeof(struct T)) - -#define CALLOC_STRUCT(T) (struct T *) CALLOC(1, sizeof(struct T)) - - -/** - * Return memory on given byte alignment - */ -static INLINE void * -align_malloc(size_t bytes, uint alignment) -{ -#if defined(HAVE_POSIX_MEMALIGN) - void *mem; - alignment = (alignment + (uint)sizeof(void*) - 1) & ~((uint)sizeof(void*) - 1); - if(posix_memalign(& mem, alignment, bytes) != 0) - return NULL; - return mem; -#else - char *ptr, *buf; - - assert( alignment > 0 ); - - ptr = (char *) MALLOC(bytes + alignment + sizeof(void *)); - if (!ptr) - return NULL; - - buf = (char *) align_pointer( ptr + sizeof(void *), alignment ); - *(char **)(buf - sizeof(void *)) = ptr; - - return buf; -#endif /* defined(HAVE_POSIX_MEMALIGN) */ -} - -/** - * Free memory returned by align_malloc(). - */ -static INLINE void -align_free(void *ptr) -{ -#if defined(HAVE_POSIX_MEMALIGN) - FREE(ptr); -#else - void **cubbyHole = (void **) ((char *) ptr - sizeof(void *)); - void *realAddr = *cubbyHole; - FREE(realAddr); -#endif /* defined(HAVE_POSIX_MEMALIGN) */ -} - - - -/** - * Duplicate a block of memory. - */ -static INLINE void * -mem_dup(const void *src, uint size) -{ - void *dup = MALLOC(size); - if (dup) - memcpy(dup, src, size); - return dup; -} - - - -#define CLAMP( X, MIN, MAX ) ( (X)<(MIN) ? (MIN) : ((X)>(MAX) ? (MAX) : (X)) ) -#define MIN2( A, B ) ( (A)<(B) ? (A) : (B) ) -#define MAX2( A, B ) ( (A)>(B) ? (A) : (B) ) - -#ifndef Elements -#define Elements(x) (sizeof(x)/sizeof((x)[0])) -#endif -#define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER)) - -/** - * Return a pointer aligned to next multiple of 16 bytes. - */ -static INLINE void * -align16( void *unaligned ) -{ - return align_pointer( unaligned, 16 ); -} - - -static INLINE int align(int value, int alignment) -{ - return (value + alignment - 1) & ~(alignment - 1); -} - - - - -#if defined(PIPE_CC_MSVC) && defined(PIPE_ARCH_X86) -static INLINE unsigned ffs( unsigned u ) -{ - unsigned i; - - if( u == 0 ) { - return 0; - } - - __asm bsf eax, [u] - __asm inc eax - __asm mov [i], eax - - return i; -} -#endif - -union fi { - float f; - int i; - unsigned ui; -}; - -#define UBYTE_TO_FLOAT( ub ) ((float)(ub) / 255.0F) - -#define IEEE_0996 0x3f7f0000 /* 0.996 or so */ - -/* This function/macro is sensitive to precision. Test very carefully - * if you change it! - */ -#define UNCLAMPED_FLOAT_TO_UBYTE(UB, F) \ - do { \ - union fi __tmp; \ - __tmp.f = (F); \ - if (__tmp.i < 0) \ - UB = (ubyte) 0; \ - else if (__tmp.i >= IEEE_0996) \ - UB = (ubyte) 255; \ - else { \ - __tmp.f = __tmp.f * (255.0f/256.0f) + 32768.0f; \ - UB = (ubyte) __tmp.i; \ - } \ - } while (0) - - - -static INLINE unsigned pack_ub4( unsigned char b0, - unsigned char b1, - unsigned char b2, - unsigned char b3 ) -{ - return ((((unsigned int)b0) << 0) | - (((unsigned int)b1) << 8) | - (((unsigned int)b2) << 16) | - (((unsigned int)b3) << 24)); -} - -static INLINE unsigned fui( float f ) -{ - union fi fi; - fi.f = f; - return fi.ui; -} - -static INLINE unsigned char float_to_ubyte( float f ) -{ - unsigned char ub; - UNCLAMPED_FLOAT_TO_UBYTE(ub, f); - return ub; -} - -static INLINE unsigned pack_ui32_float4( float a, - float b, - float c, - float d ) -{ - return pack_ub4( float_to_ubyte(a), - float_to_ubyte(b), - float_to_ubyte(c), - float_to_ubyte(d) ); -} - -#define COPY_4V( DST, SRC ) \ -do { \ - (DST)[0] = (SRC)[0]; \ - (DST)[1] = (SRC)[1]; \ - (DST)[2] = (SRC)[2]; \ - (DST)[3] = (SRC)[3]; \ -} while (0) - - -#define COPY_4FV( DST, SRC ) COPY_4V(DST, SRC) - - -#define ASSIGN_4V( DST, V0, V1, V2, V3 ) \ -do { \ - (DST)[0] = (V0); \ - (DST)[1] = (V1); \ - (DST)[2] = (V2); \ - (DST)[3] = (V3); \ -} while (0) - - - -#if defined(_MSC_VER) -#if _MSC_VER < 1400 && !defined(__cplusplus) || defined(PIPE_SUBSYSTEM_WINDOWS_CE) - -static INLINE float cosf( float f ) -{ - return (float) cos( (double) f ); -} - -static INLINE float sinf( float f ) -{ - return (float) sin( (double) f ); -} - -static INLINE float ceilf( float f ) -{ - return (float) ceil( (double) f ); -} - -static INLINE float floorf( float f ) -{ - return (float) floor( (double) f ); -} - -static INLINE float powf( float f, float g ) -{ - return (float) pow( (double) f, (double) g ); -} - -static INLINE float sqrtf( float f ) -{ - return (float) sqrt( (double) f ); -} - -static INLINE float fabsf( float f ) -{ - return (float) fabs( (double) f ); -} - -static INLINE float logf( float f ) -{ - return (float) log( (double) f ); -} - -#else -/* Work-around an extra semi-colon in VS 2005 logf definition */ -#ifdef logf -#undef logf -#define logf(x) ((float)log((double)(x))) -#endif /* logf */ -#endif -#endif /* _MSC_VER */ - - -#ifdef __cplusplus -} -#endif - -#endif diff --git a/src/gallium/state_trackers/python/gallium.i b/src/gallium/state_trackers/python/gallium.i index 641b19e940..a67372c623 100644 --- a/src/gallium/state_trackers/python/gallium.i +++ b/src/gallium/state_trackers/python/gallium.i @@ -42,7 +42,7 @@ #include "pipe/p_screen.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_shader_tokens.h" #include "cso_cache/cso_context.h" #include "util/u_draw_quad.h" diff --git a/src/gallium/state_trackers/python/st_device.c b/src/gallium/state_trackers/python/st_device.c index a1889539dc..f71d85dd9b 100644 --- a/src/gallium/state_trackers/python/st_device.c +++ b/src/gallium/state_trackers/python/st_device.c @@ -26,12 +26,13 @@ **************************************************************************/ -#include "pipe/p_util.h" #include "pipe/p_winsys.h" #include "pipe/p_context.h" #include "pipe/p_shader_tokens.h" #include "pipe/p_inlines.h" #include "cso_cache/cso_context.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "util/u_simple_shaders.h" #include "trace/tr_screen.h" #include "trace/tr_context.h" diff --git a/src/gallium/state_trackers/python/st_sample.c b/src/gallium/state_trackers/python/st_sample.c index b47c7be293..7765df3c4a 100644 --- a/src/gallium/state_trackers/python/st_sample.c +++ b/src/gallium/state_trackers/python/st_sample.c @@ -29,9 +29,10 @@ #include "pipe/p_compiler.h" #include "pipe/p_format.h" #include "pipe/p_state.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" -#include "util/p_tile.h" +#include "util/u_tile.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "st_sample.h" diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c index 6ea3c9a5cf..2d4f5434b3 100644 --- a/src/gallium/state_trackers/python/st_softpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c @@ -39,8 +39,9 @@ #include "pipe/p_winsys.h" #include "pipe/p_format.h" #include "pipe/p_context.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "softpipe/sp_winsys.h" #include "st_winsys.h" diff --git a/src/gallium/winsys/drm/intel/common/intel_be_device.c b/src/gallium/winsys/drm/intel/common/intel_be_device.c index 8db0329615..019ee5cbd2 100644 --- a/src/gallium/winsys/drm/intel/common/intel_be_device.c +++ b/src/gallium/winsys/drm/intel/common/intel_be_device.c @@ -13,8 +13,8 @@ #include "pipe/p_winsys.h" #include "pipe/p_defines.h" #include "pipe/p_state.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "util/u_memory.h" #include "i915simple/i915_screen.h" diff --git a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c index 0d98d16cf1..20920a2052 100644 --- a/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c +++ b/src/gallium/winsys/drm/intel/dri/intel_winsys_softpipe.c @@ -32,8 +32,8 @@ #include "intel_context.h" #include "intel_winsys_softpipe.h" #include "pipe/p_defines.h" -#include "pipe/p_util.h" #include "pipe/p_format.h" +#include "util/u_memory.h" #include "softpipe/sp_winsys.h" diff --git a/src/gallium/winsys/egl_xlib/egl_xlib.c b/src/gallium/winsys/egl_xlib/egl_xlib.c index 829732eea8..e9f821d276 100644 --- a/src/gallium/winsys/egl_xlib/egl_xlib.c +++ b/src/gallium/winsys/egl_xlib/egl_xlib.c @@ -38,8 +38,8 @@ #include "pipe/p_compiler.h" #include "pipe/p_format.h" #include "pipe/p_state.h" -#include "pipe/p_util.h" #include "pipe/p_winsys.h" +#include "util/u_memory.h" #include "softpipe/sp_winsys.h" #include "eglconfig.h" diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c index f4199e6f89..ae81d7f801 100644 --- a/src/gallium/winsys/egl_xlib/sw_winsys.c +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -37,8 +37,9 @@ #include "pipe/p_winsys.h" #include "pipe/p_state.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "sw_winsys.h" diff --git a/src/gallium/winsys/gdi/wmesa.c b/src/gallium/winsys/gdi/wmesa.c index ff52ceb8c4..730fb1b541 100644 --- a/src/gallium/winsys/gdi/wmesa.c +++ b/src/gallium/winsys/gdi/wmesa.c @@ -12,8 +12,8 @@ #include "pipe/p_winsys.h" #include "pipe/p_format.h" #include "pipe/p_context.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "util/u_memory.h" #include "softpipe/sp_winsys.h" #include "glapi/glapi.h" #include "colors.h" diff --git a/src/gallium/winsys/xlib/brw_aub.c b/src/gallium/winsys/xlib/brw_aub.c index 6e814ce5d1..f319802962 100644 --- a/src/gallium/winsys/xlib/brw_aub.c +++ b/src/gallium/winsys/xlib/brw_aub.c @@ -34,7 +34,6 @@ #include "brw_aub.h" #include "pipe/p_context.h" #include "pipe/p_state.h" -#include "pipe/p_util.h" #include "pipe/p_debug.h" diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c index 4b4dc56e84..68ead7f528 100644 --- a/src/gallium/winsys/xlib/xm_winsys.c +++ b/src/gallium/winsys/xlib/xm_winsys.c @@ -42,8 +42,9 @@ #include "pipe/p_winsys.h" #include "pipe/p_format.h" #include "pipe/p_context.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "softpipe/sp_winsys.h" #ifdef GALLIUM_CELL diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c index 7fc9debdd5..3439367636 100644 --- a/src/gallium/winsys/xlib/xm_winsys_aub.c +++ b/src/gallium/winsys/xlib/xm_winsys_aub.c @@ -37,7 +37,7 @@ #include "xmesaP.h" #include "pipe/p_winsys.h" -#include "pipe/p_util.h" +#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "i965simple/brw_winsys.h" #include "i965simple/brw_screen.h" diff --git a/src/mesa/state_tracker/acc2.c b/src/mesa/state_tracker/acc2.c new file mode 100644 index 0000000000..fa5de2b764 --- /dev/null +++ b/src/mesa/state_tracker/acc2.c @@ -0,0 +1,319 @@ +/************************************************************************** + * + * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. + * All Rights Reserved. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the + * "Software"), to deal in the Software without restriction, including + * without limitation the rights to use, copy, modify, merge, publish, + * distribute, sub license, and/or sell copies of the Software, and to + * permit persons to whom the Software is furnished to do so, subject to + * the following conditions: + * + * The above copyright notice and this permission notice (including the + * next paragraph) shall be included in all copies or substantial portions + * of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF + * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. + * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR + * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, + * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE + * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + * + **************************************************************************/ + + /* + * Authors: + * Brian Paul + */ + +#include "main/imports.h" +#include "main/image.h" +#include "main/macros.h" + +#include "st_context.h" +#include "st_cb_accum.h" +#include "st_cb_fbo.h" +#include "st_draw.h" +#include "st_format.h" +#include "pipe/p_context.h" +#include "pipe/p_defines.h" +#include "pipe/p_inlines.h" +#include "util/p_tile.h" + + +#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ + us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) + + +/** + * For hardware that supports deep color buffers, we could accelerate + * most/all the accum operations with blending/texturing. + * For now, just use the get/put_tile() functions and do things in software. + */ + + +static void +acc_get_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps, + uint x, uint y, uint w, uint h, float *p) +{ + const enum pipe_format f = acc_ps->format; + const int cpp = acc_ps->cpp; + + acc_ps->format = PIPE_FORMAT_R16G16B16A16_SNORM; + acc_ps->cpp = 8; + + pipe_get_tile_rgba(pipe, acc_ps, x, y, w, h, p); + + acc_ps->format = f; + acc_ps->cpp = cpp; +} + + +static void +acc_put_tile_rgba(struct pipe_context *pipe, struct pipe_surface *acc_ps, + uint x, uint y, uint w, uint h, const float *p) +{ + enum pipe_format f = acc_ps->format; + const int cpp = acc_ps->cpp; + + acc_ps->format = PIPE_FORMAT_R16G16B16A16_SNORM; + acc_ps->cpp = 8; + + pipe_put_tile_rgba(pipe, acc_ps, x, y, w, h, p); + + acc_ps->format = f; + acc_ps->cpp = cpp; +} + + + +void +st_clear_accum_buffer(GLcontext *ctx, struct gl_renderbuffer *rb) +{ + struct pipe_context *pipe = ctx->st->pipe; + struct st_renderbuffer *acc_strb = st_renderbuffer(rb); + struct pipe_surface *acc_ps = acc_strb->surface; + const GLint xpos = ctx->DrawBuffer->_Xmin; + const GLint ypos = ctx->DrawBuffer->_Ymin; + const GLint width = ctx->DrawBuffer->_Xmax - xpos; + const GLint height = ctx->DrawBuffer->_Ymax - ypos; + const GLfloat r = ctx->Accum.ClearColor[0]; + const GLfloat g = ctx->Accum.ClearColor[1]; + const GLfloat b = ctx->Accum.ClearColor[2]; + const GLfloat a = ctx->Accum.ClearColor[3]; + GLfloat *accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + int i; + +#if 1 + GLvoid *map; + + map = pipe_surface_map(acc_ps); + switch (acc_strb->format) { + case PIPE_FORMAT_R16G16B16A16_SNORM: + { + GLshort r = FLOAT_TO_SHORT(ctx->Accum.ClearColor[0]); + GLshort g = FLOAT_TO_SHORT(ctx->Accum.ClearColor[1]); + GLshort b = FLOAT_TO_SHORT(ctx->Accum.ClearColor[2]); + GLshort a = FLOAT_TO_SHORT(ctx->Accum.ClearColor[3]); + int i, j; + for (i = 0; i < height; i++) { + GLshort *dst = ((GLshort *) map + + ((ypos + i) * acc_ps->pitch + xpos) * 4); + for (j = 0; j < width; j++) { + dst[0] = r; + dst[1] = g; + dst[2] = b; + dst[3] = a; + dst += 4; + } + } + } + break; + default: + _mesa_problem(ctx, "unexpected format in st_clear_accum_buffer()"); + } + + pipe_surface_unmap(acc_ps); + +#else + for (i = 0; i < width * height; i++) { + accBuf[i*4+0] = r; + accBuf[i*4+1] = g; + accBuf[i*4+2] = b; + accBuf[i*4+3] = a; + } + + acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); +#endif +} + + +/** For ADD/MULT */ +static void +accum_mad(struct pipe_context *pipe, GLfloat scale, GLfloat bias, + GLint xpos, GLint ypos, GLint width, GLint height, + struct pipe_surface *acc_ps) +{ + GLfloat *accBuf; + GLint i; + + accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + pipe_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + + for (i = 0; i < 4 * width * height; i++) { + accBuf[i] = accBuf[i] * scale + bias; + } + + pipe_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + + free(accBuf); +} + + +static void +accum_accum(struct pipe_context *pipe, GLfloat value, + GLint xpos, GLint ypos, GLint width, GLint height, + struct pipe_surface *acc_ps, + struct pipe_surface *color_ps) +{ + GLfloat *colorBuf, *accBuf; + GLint i; + + colorBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + accBuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, colorBuf); + acc_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + + for (i = 0; i < 4 * width * height; i++) { + accBuf[i] = accBuf[i] + colorBuf[i] * value; + } + + acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, accBuf); + + free(colorBuf); + free(accBuf); +} + + +static void +accum_load(struct pipe_context *pipe, GLfloat value, + GLint xpos, GLint ypos, GLint width, GLint height, + struct pipe_surface *acc_ps, + struct pipe_surface *color_ps) +{ + GLfloat *buf; + GLint i; + + buf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, buf); + + for (i = 0; i < 4 * width * height; i++) { + buf[i] = buf[i] * value; + } + + acc_put_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, buf); + + free(buf); +} + + +static void +accum_return(GLcontext *ctx, GLfloat value, + GLint xpos, GLint ypos, GLint width, GLint height, + struct pipe_surface *acc_ps, + struct pipe_surface *color_ps) +{ + struct pipe_context *pipe = ctx->st->pipe; + const GLubyte *colormask = ctx->Color.ColorMask; + GLfloat *abuf, *cbuf = NULL; + GLint i, ch; + + abuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + + acc_get_tile_rgba(pipe, acc_ps, xpos, ypos, width, height, abuf); + + if (!colormask[0] || !colormask[1] || !colormask[2] || !colormask[3]) { + cbuf = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat)); + pipe_get_tile_rgba(pipe, color_ps, xpos, ypos, width, height, cbuf); + } + + for (i = 0; i < width * height; i++) { + for (ch = 0; ch < 4; ch++) { + if (colormask[ch]) { + GLfloat val = abuf[i * 4 + ch] * value; + abuf[i * 4 + ch] = CLAMP(val, 0.0, 1.0); + } + else { + abuf[i * 4 + ch] = cbuf[i * 4 + ch]; + } + } + } + + pipe_put_tile_rgba(pipe, color_ps, xpos, ypos, width, height, abuf); + + free(abuf); + if (cbuf) + free(cbuf); +} + + +static void +st_Accum(GLcontext *ctx, GLenum op, GLfloat value) +{ + struct st_context *st = ctx->st; + struct pipe_context *pipe = st->pipe; + struct st_renderbuffer *acc_strb + = st_renderbuffer(ctx->DrawBuffer->Attachment[BUFFER_ACCUM].Renderbuffer); + struct st_renderbuffer *color_strb + = st_renderbuffer(ctx->ReadBuffer->_ColorReadBuffer); + struct pipe_surface *acc_ps = acc_strb->surface; + struct pipe_surface *color_ps = color_strb->surface; + + const GLint xpos = ctx->DrawBuffer->_Xmin; + const GLint ypos = ctx->DrawBuffer->_Ymin; + const GLint width = ctx->DrawBuffer->_Xmax - xpos; + const GLint height = ctx->DrawBuffer->_Ymax - ypos; + + /* make sure color bufs aren't cached */ + pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL); + + switch (op) { + case GL_ADD: + if (value != 0.0F) { + accum_mad(pipe, 1.0, value, xpos, ypos, width, height, acc_ps); + } + break; + case GL_MULT: + if (value != 1.0F) { + accum_mad(pipe, value, 0.0, xpos, ypos, width, height, acc_ps); + } + break; + case GL_ACCUM: + if (value != 0.0F) { + accum_accum(pipe, value, xpos, ypos, width, height, acc_ps, color_ps); + } + break; + case GL_LOAD: + accum_load(pipe, value, xpos, ypos, width, height, acc_ps, color_ps); + break; + case GL_RETURN: + accum_return(ctx, value, xpos, ypos, width, height, acc_ps, color_ps); + break; + default: + assert(0); + } +} + + + +void st_init_accum_functions(struct dd_function_table *functions) +{ + functions->Accum = st_Accum; +} diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c index a992e08ff6..cf3a99e7e9 100644 --- a/src/mesa/state_tracker/st_cb_accum.c +++ b/src/mesa/state_tracker/st_cb_accum.c @@ -42,7 +42,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "util/p_tile.h" +#include "util/u_tile.h" #define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c index d5696a909f..a0c305d66f 100644 --- a/src/mesa/state_tracker/st_cb_bitmap.c +++ b/src/mesa/state_tracker/st_cb_bitmap.c @@ -50,7 +50,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "util/p_tile.h" +#include "util/u_tile.h" #include "util/u_draw_quad.h" #include "util/u_simple_shaders.h" #include "shader/prog_instruction.h" diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index 0c5e21d4ff..4ec7c752df 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -55,7 +55,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "util/p_tile.h" +#include "util/u_tile.h" #include "util/u_draw_quad.h" #include "shader/prog_instruction.h" #include "cso_cache/cso_context.h" diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c index 39f5856f94..c801532788 100644 --- a/src/mesa/state_tracker/st_cb_readpixels.c +++ b/src/mesa/state_tracker/st_cb_readpixels.c @@ -41,7 +41,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "util/p_tile.h" +#include "util/u_tile.h" #include "st_context.h" #include "st_cb_bitmap.h" #include "st_cb_readpixels.h" diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 6177ac63f0..16bbf3d80f 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -51,7 +51,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "util/p_tile.h" +#include "util/u_tile.h" #include "util/u_blit.h" diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c index 325d95e865..936a6e32ea 100644 --- a/src/mesa/state_tracker/st_program.c +++ b/src/mesa/state_tracker/st_program.c @@ -55,7 +55,7 @@ #define TGSI_DEBUG 0 -/** XXX we should use the version of this from p_util.h but including +/** XXX we should use the version of this from u_memory.h but including * that header causes symbol collisions. */ static INLINE void * diff --git a/src/mesa/state_tracker/st_texture.c b/src/mesa/state_tracker/st_texture.c index 63046a0ecc..73cebff33f 100644 --- a/src/mesa/state_tracker/st_texture.c +++ b/src/mesa/state_tracker/st_texture.c @@ -36,7 +36,6 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "pipe/p_util.h" #include "pipe/p_inlines.h" #include "util/u_rect.h" -- cgit v1.2.3 From 4c84e940d009fe56c1f249507028daadea749ef2 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Sun, 24 Aug 2008 17:49:11 -0600 Subject: gallium: remove old tile util files --- src/gallium/auxiliary/util/p_tile.c | 1168 ----------------------------------- src/gallium/auxiliary/util/p_tile.h | 101 --- 2 files changed, 1269 deletions(-) delete mode 100644 src/gallium/auxiliary/util/p_tile.c delete mode 100644 src/gallium/auxiliary/util/p_tile.h (limited to 'src') diff --git a/src/gallium/auxiliary/util/p_tile.c b/src/gallium/auxiliary/util/p_tile.c deleted file mode 100644 index 8219041c80..0000000000 --- a/src/gallium/auxiliary/util/p_tile.c +++ /dev/null @@ -1,1168 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -/** - * RGBA/float tile get/put functions. - * Usable both by drivers and state trackers. - * Surfaces should already be in a mapped state. - */ - - -#include "pipe/p_defines.h" -#include "pipe/p_util.h" -#include "pipe/p_inlines.h" - -#include "p_tile.h" -#include "u_rect.h" - - -/** - * Move raw block of pixels from surface to user memory. - * This should be usable by any hw driver that has mappable surfaces. - */ -void -pipe_get_tile_raw(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *dst, int dst_stride) -{ - const void *src; - - if (dst_stride == 0) - dst_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - src = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ); - assert(src); - if(!src) - return; - - pipe_copy_rect(dst, &ps->block, dst_stride, 0, 0, w, h, src, ps->stride, x, y); - - pipe_surface_unmap(ps); -} - - -/** - * Move raw block of pixels from user memory to surface. - * This should be usable by any hw driver that has mappable surfaces. - */ -void -pipe_put_tile_raw(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *src, int src_stride) -{ - void *dst; - - if (src_stride == 0) - src_stride = pf_get_nblocksx(&ps->block, w) * ps->block.size; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - dst = pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE); - assert(dst); - if(!dst) - return; - - pipe_copy_rect(dst, &ps->block, ps->stride, x, y, w, h, src, src_stride, 0, 0); - - pipe_surface_unmap(ps); -} - - - - -/** Convert short in [-32768,32767] to GLfloat in [-1.0,1.0] */ -#define SHORT_TO_FLOAT(S) ((2.0F * (S) + 1.0F) * (1.0F/65535.0F)) - -#define UNCLAMPED_FLOAT_TO_SHORT(us, f) \ - us = ( (short) ( CLAMP((f), -1.0, 1.0) * 32767.0F) ) - - - -/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ - -static void -a8r8g8b8_get_tile_rgba(const unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const unsigned pixel = *src++; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - } - p += dst_stride; - } -} - - -static void -a8r8g8b8_put_tile_rgba(unsigned *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - *dst++ = (a << 24) | (r << 16) | (g << 8) | b; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/ - -static void -x8r8g8b8_get_tile_rgba(const unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const unsigned pixel = *src++; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - pRow[3] = UBYTE_TO_FLOAT(0xff); - } - p += dst_stride; - } -} - - -static void -x8r8g8b8_put_tile_rgba(unsigned *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, g, b; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - *dst++ = (0xff << 24) | (r << 16) | (g << 8) | b; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/ - -static void -b8g8r8a8_get_tile_rgba(const unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const unsigned pixel = *src++; - pRow[0] = UBYTE_TO_FLOAT((pixel >> 8) & 0xff); - pRow[1] = UBYTE_TO_FLOAT((pixel >> 16) & 0xff); - pRow[2] = UBYTE_TO_FLOAT((pixel >> 24) & 0xff); - pRow[3] = UBYTE_TO_FLOAT((pixel >> 0) & 0xff); - } - p += dst_stride; - } -} - - -static void -b8g8r8a8_put_tile_rgba(unsigned *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - *dst++ = (b << 24) | (g << 16) | (r << 8) | a; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_A1R5G5B5_UNORM ***/ - -static void -a1r5g5b5_get_tile_rgba(const ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const ushort pixel = *src++; - pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f); - pRow[1] = ((pixel >> 5) & 0x1f) * (1.0f / 31.0f); - pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - pRow[3] = ((pixel >> 15) ) * 1.0f; - } - p += dst_stride; - } -} - - -static void -a1r5g5b5_put_tile_rgba(ushort *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - r = r >> 3; /* 5 bits */ - g = g >> 3; /* 5 bits */ - b = b >> 3; /* 5 bits */ - a = a >> 7; /* 1 bit */ - *dst++ = (a << 15) | (r << 10) | (g << 5) | b; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_A4R4G4B4_UNORM ***/ - -static void -a4r4g4b4_get_tile_rgba(const ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const ushort pixel = *src++; - pRow[0] = ((pixel >> 8) & 0xf) * (1.0f / 15.0f); - pRow[1] = ((pixel >> 4) & 0xf) * (1.0f / 15.0f); - pRow[2] = ((pixel ) & 0xf) * (1.0f / 15.0f); - pRow[3] = ((pixel >> 12) ) * (1.0f / 15.0f); - } - p += dst_stride; - } -} - - -static void -a4r4g4b4_put_tile_rgba(ushort *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, g, b, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(g, pRow[1]); - UNCLAMPED_FLOAT_TO_UBYTE(b, pRow[2]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - r >>= 4; - g >>= 4; - b >>= 4; - a >>= 4; - *dst++ = (a << 12) | (r << 16) | (g << 4) | b; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_R5G6B5_UNORM ***/ - -static void -r5g6b5_get_tile_rgba(const ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - const ushort pixel = *src++; - pRow[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f); - pRow[1] = ((pixel >> 5) & 0x3f) * (1.0f / 63.0f); - pRow[2] = ((pixel ) & 0x1f) * (1.0f / 31.0f); - pRow[3] = 1.0f; - } - p += dst_stride; - } -} - - -static void -r5g6b5_put_tile_rgba(ushort *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0); - uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0); - uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0); - *dst++ = (r << 11) | (g << 5) | (b); - } - p += src_stride; - } -} - - - -/*** PIPE_FORMAT_Z16_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z16_get_tile_rgba(const ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const float scale = 1.0f / 65535.0f; - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = *src++ * scale; - } - p += dst_stride; - } -} - - - - -/*** PIPE_FORMAT_L8_UNORM ***/ - -static void -l8_get_tile_rgba(const ubyte *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(*src); - pRow[3] = 1.0; - } - p += dst_stride; - } -} - - -static void -l8_put_tile_rgba(ubyte *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - *dst++ = r; - } - p += src_stride; - } -} - - - -/*** PIPE_FORMAT_A8_UNORM ***/ - -static void -a8_get_tile_rgba(const ubyte *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = 0.0; - pRow[3] = UBYTE_TO_FLOAT(*src); - } - p += dst_stride; - } -} - - -static void -a8_put_tile_rgba(ubyte *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned a; - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - *dst++ = a; - } - p += src_stride; - } -} - - - -/*** PIPE_FORMAT_R16_SNORM ***/ - -static void -r16_get_tile_rgba(const short *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = SHORT_TO_FLOAT(src[0]); - pRow[1] = - pRow[2] = 0.0; - pRow[3] = 1.0; - } - p += dst_stride; - } -} - - -static void -r16_put_tile_rgba(short *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, dst++, pRow += 4) { - UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/ - -static void -r16g16b16a16_get_tile_rgba(const short *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src += 4, pRow += 4) { - pRow[0] = SHORT_TO_FLOAT(src[0]); - pRow[1] = SHORT_TO_FLOAT(src[1]); - pRow[2] = SHORT_TO_FLOAT(src[2]); - pRow[3] = SHORT_TO_FLOAT(src[3]); - } - p += dst_stride; - } -} - - -static void -r16g16b16a16_put_tile_rgba(short *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, dst += 4, pRow += 4) { - UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]); - UNCLAMPED_FLOAT_TO_SHORT(dst[1], pRow[1]); - UNCLAMPED_FLOAT_TO_SHORT(dst[2], pRow[2]); - UNCLAMPED_FLOAT_TO_SHORT(dst[3], pRow[3]); - } - p += src_stride; - } -} - - - -/*** PIPE_FORMAT_I8_UNORM ***/ - -static void -i8_get_tile_rgba(const ubyte *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, src++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = UBYTE_TO_FLOAT(*src); - } - p += dst_stride; - } -} - - -static void -i8_put_tile_rgba(ubyte *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - *dst++ = r; - } - p += src_stride; - } -} - - -/*** PIPE_FORMAT_A8L8_UNORM ***/ - -static void -a8l8_get_tile_rgba(const ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - ushort p = *src++; - pRow[0] = - pRow[1] = - pRow[2] = UBYTE_TO_FLOAT(p & 0xff); - pRow[3] = UBYTE_TO_FLOAT(p >> 8); - } - p += dst_stride; - } -} - - -static void -a8l8_put_tile_rgba(ushort *dst, - unsigned w, unsigned h, - const float *p, - unsigned src_stride) -{ - unsigned i, j; - - for (i = 0; i < h; i++) { - const float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - unsigned r, a; - UNCLAMPED_FLOAT_TO_UBYTE(r, pRow[0]); - UNCLAMPED_FLOAT_TO_UBYTE(a, pRow[3]); - *dst++ = (a << 8) | r; - } - p += src_stride; - } -} - - - - -/*** PIPE_FORMAT_Z32_UNORM ***/ - -/** - * Return each Z value as four floats in [0,1]. - */ -static void -z32_get_tile_rgba(const unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const double scale = 1.0 / (double) 0xffffffff; - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (float) (*src++ * scale); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_S8Z24_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -s8z24_get_tile_rgba(const unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (float) (scale * (*src++ & 0xffffff)); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_Z24S8_UNORM ***/ - -/** - * Return Z component as four float in [0,1]. Stencil part ignored. - */ -static void -z24s8_get_tile_rgba(const unsigned *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride) -{ - const double scale = 1.0 / ((1 << 24) - 1); - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - for (j = 0; j < w; j++, pRow += 4) { - pRow[0] = - pRow[1] = - pRow[2] = - pRow[3] = (float) (scale * (*src++ >> 8)); - } - p += dst_stride; - } -} - - -/*** PIPE_FORMAT_YCBCR / PIPE_FORMAT_YCBCR_REV ***/ - -/** - * Convert YCbCr (or YCrCb) to RGBA. - */ -static void -ycbcr_get_tile_rgba(const ushort *src, - unsigned w, unsigned h, - float *p, - unsigned dst_stride, - boolean rev) -{ - const float scale = 1.0f / 255.0f; - unsigned i, j; - - for (i = 0; i < h; i++) { - float *pRow = p; - /* do two texels at a time */ - for (j = 0; j < (w & ~1); j += 2, src += 2) { - const ushort t0 = src[0]; - const ushort t1 = src[1]; - const ubyte y0 = (t0 >> 8) & 0xff; /* luminance */ - const ubyte y1 = (t1 >> 8) & 0xff; /* luminance */ - ubyte cb, cr; - float r, g, b; - - if (rev) { - cb = t1 & 0xff; /* chroma U */ - cr = t0 & 0xff; /* chroma V */ - } - else { - cb = t0 & 0xff; /* chroma U */ - cr = t1 & 0xff; /* chroma V */ - } - - /* even pixel: y0,cr,cb */ - r = 1.164f * (y0-16) + 1.596f * (cr-128); - g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128); - b = 1.164f * (y0-16) + 2.018f * (cb-128); - pRow[0] = r * scale; - pRow[1] = g * scale; - pRow[2] = b * scale; - pRow[3] = 1.0f; - pRow += 4; - - /* odd pixel: use y1,cr,cb */ - r = 1.164f * (y1-16) + 1.596f * (cr-128); - g = 1.164f * (y1-16) - 0.813f * (cr-128) - 0.391f * (cb-128); - b = 1.164f * (y1-16) + 2.018f * (cb-128); - pRow[0] = r * scale; - pRow[1] = g * scale; - pRow[2] = b * scale; - pRow[3] = 1.0f; - pRow += 4; - - } - /* do the last texel */ - if (w & 1) { - const ushort t0 = src[0]; - const ushort t1 = src[1]; - const ubyte y0 = (t0 >> 8) & 0xff; /* luminance */ - ubyte cb, cr; - float r, g, b; - - if (rev) { - cb = t1 & 0xff; /* chroma U */ - cr = t0 & 0xff; /* chroma V */ - } - else { - cb = t0 & 0xff; /* chroma U */ - cr = t1 & 0xff; /* chroma V */ - } - - /* even pixel: y0,cr,cb */ - r = 1.164f * (y0-16) + 1.596f * (cr-128); - g = 1.164f * (y0-16) - 0.813f * (cr-128) - 0.391f * (cb-128); - b = 1.164f * (y0-16) + 2.018f * (cb-128); - pRow[0] = r * scale; - pRow[1] = g * scale; - pRow[2] = b * scale; - pRow[3] = 1.0f; - pRow += 4; - } - p += dst_stride; - } -} - - -void -pipe_tile_raw_to_rgba(enum pipe_format format, - void *src, - uint w, uint h, - float *dst, unsigned dst_stride) -{ - switch (format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_X8R8G8B8_UNORM: - x8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - a1r5g5b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_A4R4G4B4_UNORM: - a4r4g4b4_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_R5G6B5_UNORM: - r5g6b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_L8_UNORM: - l8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_A8_UNORM: - a8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_I8_UNORM: - i8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_A8L8_UNORM: - a8l8_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_R16_SNORM: - r16_get_tile_rgba((short *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_get_tile_rgba((short *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_Z16_UNORM: - z16_get_tile_rgba((ushort *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_Z32_UNORM: - z32_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_S8Z24_UNORM: - case PIPE_FORMAT_X8Z24_UNORM: - s8z24_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_Z24S8_UNORM: - z24s8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride); - break; - case PIPE_FORMAT_YCBCR: - ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, FALSE); - break; - case PIPE_FORMAT_YCBCR_REV: - ycbcr_get_tile_rgba((ushort *) src, w, h, dst, dst_stride, TRUE); - break; - default: - assert(0); - } -} - - -void -pipe_get_tile_rgba(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p) -{ - unsigned dst_stride = w * 4; - void *packed; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); - - if (!packed) - return; - - if(ps->format == PIPE_FORMAT_YCBCR || ps->format == PIPE_FORMAT_YCBCR_REV) - assert((x & 1) == 0); - - pipe_get_tile_raw(ps, x, y, w, h, packed, 0); - - pipe_tile_raw_to_rgba(ps->format, packed, w, h, p, dst_stride); - - FREE(packed); -} - - -void -pipe_put_tile_rgba(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p) -{ - unsigned src_stride = w * 4; - void *packed; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - packed = MALLOC(pf_get_nblocks(&ps->block, w, h) * ps->block.size); - - if (!packed) - return; - - switch (ps->format) { - case PIPE_FORMAT_A8R8G8B8_UNORM: - a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_X8R8G8B8_UNORM: - x8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_B8G8R8A8_UNORM: - b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_A1R5G5B5_UNORM: - a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_R5G6B5_UNORM: - r5g6b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_R8G8B8A8_UNORM: - assert(0); - break; - case PIPE_FORMAT_A4R4G4B4_UNORM: - a4r4g4b4_put_tile_rgba((ushort *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_L8_UNORM: - l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_A8_UNORM: - a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_I8_UNORM: - i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_A8L8_UNORM: - a8l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_R16_SNORM: - r16_put_tile_rgba((short *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_R16G16B16A16_SNORM: - r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride); - break; - case PIPE_FORMAT_Z16_UNORM: - /*z16_put_tile_rgba((ushort *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_Z32_UNORM: - /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_S8Z24_UNORM: - case PIPE_FORMAT_X8Z24_UNORM: - /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ - break; - case PIPE_FORMAT_Z24S8_UNORM: - /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/ - break; - default: - assert(0); - } - - pipe_put_tile_raw(ps, x, y, w, h, packed, 0); - - FREE(packed); -} - - -/** - * Get a block of Z values, converted to 32-bit range. - */ -void -pipe_get_tile_z(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - uint *z) -{ - const uint dstStride = w; - ubyte *map; - uint *pDest = z; - uint i, j; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_READ); - if (!map) { - assert(0); - return; - } - - switch (ps->format) { - case PIPE_FORMAT_Z32_UNORM: - { - const uint *pSrc - = (const uint *)(map + y * ps->stride + x*4); - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, 4 * w); - pDest += dstStride; - pSrc += ps->stride/4; - } - } - break; - case PIPE_FORMAT_S8Z24_UNORM: - case PIPE_FORMAT_X8Z24_UNORM: - { - const uint *pSrc - = (const uint *)(map + y * ps->stride + x*4); - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - /* convert 24-bit Z to 32-bit Z */ - pDest[j] = (pSrc[j] << 8) | (pSrc[j] & 0xff); - } - pDest += dstStride; - pSrc += ps->stride/4; - } - } - break; - case PIPE_FORMAT_Z16_UNORM: - { - const ushort *pSrc - = (const ushort *)(map + y * ps->stride + x*2); - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - /* convert 16-bit Z to 32-bit Z */ - pDest[j] = (pSrc[j] << 16) | pSrc[j]; - } - pDest += dstStride; - pSrc += ps->stride/2; - } - } - break; - default: - assert(0); - } - - pipe_surface_unmap(ps); -} - - -void -pipe_put_tile_z(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const uint *zSrc) -{ - const uint srcStride = w; - const uint *pSrc = zSrc; - ubyte *map; - uint i, j; - - if (pipe_clip_tile(x, y, &w, &h, ps)) - return; - - map = (ubyte *)pipe_surface_map(ps, PIPE_BUFFER_USAGE_CPU_WRITE); - if (!map) { - assert(0); - return; - } - - switch (ps->format) { - case PIPE_FORMAT_Z32_UNORM: - { - uint *pDest = (uint *) (map + y * ps->stride + x*4); - for (i = 0; i < h; i++) { - memcpy(pDest, pSrc, 4 * w); - pDest += ps->stride/4; - pSrc += srcStride; - } - } - break; - case PIPE_FORMAT_S8Z24_UNORM: - case PIPE_FORMAT_X8Z24_UNORM: - { - uint *pDest = (uint *) (map + y * ps->stride + x*4); - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - /* convert 32-bit Z to 24-bit Z (0 stencil) */ - pDest[j] = pSrc[j] >> 8; - } - pDest += ps->stride/4; - pSrc += srcStride; - } - } - break; - case PIPE_FORMAT_Z16_UNORM: - { - ushort *pDest = (ushort *) (map + y * ps->stride + x*2); - for (i = 0; i < h; i++) { - for (j = 0; j < w; j++) { - /* convert 32-bit Z to 16-bit Z */ - pDest[j] = pSrc[j] >> 16; - } - pDest += ps->stride/2; - pSrc += srcStride; - } - } - break; - default: - assert(0); - } - - pipe_surface_unmap(ps); -} - - diff --git a/src/gallium/auxiliary/util/p_tile.h b/src/gallium/auxiliary/util/p_tile.h deleted file mode 100644 index a8ac805308..0000000000 --- a/src/gallium/auxiliary/util/p_tile.h +++ /dev/null @@ -1,101 +0,0 @@ -/************************************************************************** - * - * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas. - * All Rights Reserved. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the - * "Software"), to deal in the Software without restriction, including - * without limitation the rights to use, copy, modify, merge, publish, - * distribute, sub license, and/or sell copies of the Software, and to - * permit persons to whom the Software is furnished to do so, subject to - * the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS - * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF - * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. - * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR - * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, - * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE - * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. - * - **************************************************************************/ - -#ifndef P_TILE_H -#define P_TILE_H - -#include "pipe/p_compiler.h" - -struct pipe_surface; - - -/** - * Clip tile against surface dims. - * \return TRUE if tile is totally clipped, FALSE otherwise - */ -static INLINE boolean -pipe_clip_tile(uint x, uint y, uint *w, uint *h, const struct pipe_surface *ps) -{ - if (x >= ps->width) - return TRUE; - if (y >= ps->height) - return TRUE; - if (x + *w > ps->width) - *w = ps->width - x; - if (y + *h > ps->height) - *h = ps->height - y; - return FALSE; -} - -#ifdef __cplusplus -extern "C" { -#endif - -void -pipe_get_tile_raw(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - void *p, int dst_stride); - -void -pipe_put_tile_raw(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const void *p, int src_stride); - - -void -pipe_get_tile_rgba(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - float *p); - -void -pipe_put_tile_rgba(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const float *p); - - -void -pipe_get_tile_z(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - uint *z); - -void -pipe_put_tile_z(struct pipe_surface *ps, - uint x, uint y, uint w, uint h, - const uint *z); - -void -pipe_tile_raw_to_rgba(enum pipe_format format, - void *src, - uint w, uint h, - float *dst, unsigned dst_stride); - - -#ifdef __cplusplus -} -#endif - -#endif -- cgit v1.2.3 From 6fd2feaad7029a2f6d02bcf7039cbe72f53c615c Mon Sep 17 00:00:00 2001 From: Brian Date: Sun, 24 Aug 2008 18:10:50 -0600 Subject: gallium: include u_memory.h, u_math.h --- src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c | 1 + src/gallium/winsys/xlib/brw_aub.c | 1 + 2 files changed, 2 insertions(+) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c index 5e518370d0..4cb8c3bb55 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c +++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c @@ -37,6 +37,7 @@ #include "pipe/p_debug.h" #include "pipe/p_winsys.h" #include "pipe/p_thread.h" +#include "util/u_math.h" #include "util/u_memory.h" #include "util/u_double_list.h" #include "util/u_time.h" diff --git a/src/gallium/winsys/xlib/brw_aub.c b/src/gallium/winsys/xlib/brw_aub.c index f319802962..9e96efaa53 100644 --- a/src/gallium/winsys/xlib/brw_aub.c +++ b/src/gallium/winsys/xlib/brw_aub.c @@ -35,6 +35,7 @@ #include "pipe/p_context.h" #include "pipe/p_state.h" #include "pipe/p_debug.h" +#include "util/u_memory.h" struct brw_aubfile { -- cgit v1.2.3 From 2c4661f8fce8a27f2082c6ac498f9fb188878476 Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Mon, 25 Aug 2008 10:42:00 +0200 Subject: gallium: Add missing includes. --- src/gallium/auxiliary/util/u_memory.h | 2 +- src/gallium/drivers/i915simple/i915_fpc_emit.c | 1 + 2 files changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_memory.h b/src/gallium/auxiliary/util/u_memory.h index 148a5cb997..ec32d60fbe 100644 --- a/src/gallium/auxiliary/util/u_memory.h +++ b/src/gallium/auxiliary/util/u_memory.h @@ -51,7 +51,7 @@ /* memory debugging */ -#include "p_debug.h" +#include "pipe/p_debug.h" #define MALLOC( _size ) \ debug_malloc( __FILE__, __LINE__, __FUNCTION__, _size ) diff --git a/src/gallium/drivers/i915simple/i915_fpc_emit.c b/src/gallium/drivers/i915simple/i915_fpc_emit.c index 4bdeefb449..b054ce41d3 100644 --- a/src/gallium/drivers/i915simple/i915_fpc_emit.c +++ b/src/gallium/drivers/i915simple/i915_fpc_emit.c @@ -28,6 +28,7 @@ #include "i915_reg.h" #include "i915_context.h" #include "i915_fpc.h" +#include "util/u_math.h" #define A0_DEST( reg ) (((reg)&UREG_TYPE_NR_MASK)>>UREG_A0_DEST_SHIFT_LEFT) -- cgit v1.2.3 From 85813e36f2646e00ac163fb6e6e78e1d2f73eb34 Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 25 Aug 2008 08:32:50 -0600 Subject: gallium: include u_math.h --- src/gallium/winsys/xlib/xm_winsys_aub.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src') diff --git a/src/gallium/winsys/xlib/xm_winsys_aub.c b/src/gallium/winsys/xlib/xm_winsys_aub.c index 3439367636..35c4ebc4ba 100644 --- a/src/gallium/winsys/xlib/xm_winsys_aub.c +++ b/src/gallium/winsys/xlib/xm_winsys_aub.c @@ -37,8 +37,9 @@ #include "xmesaP.h" #include "pipe/p_winsys.h" -#include "util/u_memory.h" #include "pipe/p_inlines.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "i965simple/brw_winsys.h" #include "i965simple/brw_screen.h" #include "brw_aub.h" -- cgit v1.2.3 From fd06d01b80fd9b2924682686e943d819bfcf027b Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Aug 2008 11:29:39 -0600 Subject: gallium: include on linux to get sched_yield() proto --- src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c index 8ae052e875..410d336fef 100644 --- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c +++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c @@ -38,6 +38,7 @@ #if defined(PIPE_OS_LINUX) #include +#include #endif #include "pipe/p_compiler.h" -- cgit v1.2.3 From a644c5a850242376e6ab03f7f4bdbfb8a232490f Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Aug 2008 11:30:20 -0600 Subject: gallium: include p_debug.h for non-HAVE_POSIX_MEMALIGN --- src/gallium/auxiliary/util/u_memory.h | 1 + 1 file changed, 1 insertion(+) (limited to 'src') diff --git a/src/gallium/auxiliary/util/u_memory.h b/src/gallium/auxiliary/util/u_memory.h index ec32d60fbe..8fbbb4e55d 100644 --- a/src/gallium/auxiliary/util/u_memory.h +++ b/src/gallium/auxiliary/util/u_memory.h @@ -36,6 +36,7 @@ #include "util/u_pointer.h" +#include "pipe/p_debug.h" /* Define ENOMEM for WINCE */ -- cgit v1.2.3 From 60ac76175b6457ecc1cd8bd7a25cb79b2d529434 Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Aug 2008 11:30:57 -0600 Subject: gallium: added const qualifiers on some draw funcs --- src/gallium/auxiliary/draw/draw_context.c | 4 ++-- src/gallium/auxiliary/draw/draw_context.h | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c index 1c26cb31a3..36751c2621 100644 --- a/src/gallium/auxiliary/draw/draw_context.c +++ b/src/gallium/auxiliary/draw/draw_context.c @@ -288,7 +288,7 @@ draw_enable_point_sprites(struct draw_context *draw, boolean enable) * work for the drivers. */ int -draw_find_vs_output(struct draw_context *draw, +draw_find_vs_output(const struct draw_context *draw, uint semantic_name, uint semantic_index) { const struct draw_vertex_shader *vs = draw->vs.vertex_shader; @@ -314,7 +314,7 @@ draw_find_vs_output(struct draw_context *draw, * Return number of vertex shader outputs. */ uint -draw_num_vs_outputs(struct draw_context *draw) +draw_num_vs_outputs(const struct draw_context *draw) { uint count = draw->vs.vertex_shader->info.num_outputs; if (draw->extra_vp_outputs.slot > 0) diff --git a/src/gallium/auxiliary/draw/draw_context.h b/src/gallium/auxiliary/draw/draw_context.h index b8f2bfa332..0ab3681b64 100644 --- a/src/gallium/auxiliary/draw/draw_context.h +++ b/src/gallium/auxiliary/draw/draw_context.h @@ -84,11 +84,11 @@ draw_install_pstipple_stage(struct draw_context *draw, struct pipe_context *pipe int -draw_find_vs_output(struct draw_context *draw, +draw_find_vs_output(const struct draw_context *draw, uint semantic_name, uint semantic_index); uint -draw_num_vs_outputs(struct draw_context *draw); +draw_num_vs_outputs(const struct draw_context *draw); -- cgit v1.2.3 From 6ba9fb9b6693904054ad4e1506ba42e324334b0a Mon Sep 17 00:00:00 2001 From: Brian Date: Mon, 25 Aug 2008 11:31:59 -0600 Subject: cell: asst fixes to get driver building/running again. Note that SPU vertex transformation is disabled at this time. --- src/gallium/drivers/cell/ppu/cell_clear.c | 4 +- src/gallium/drivers/cell/ppu/cell_context.c | 4 ++ src/gallium/drivers/cell/ppu/cell_draw_arrays.c | 35 ++++++++++-- src/gallium/drivers/cell/ppu/cell_draw_arrays.h | 24 ++++++-- src/gallium/drivers/cell/ppu/cell_pipe_state.c | 6 +- src/gallium/drivers/cell/ppu/cell_state_emit.c | 14 ++--- src/gallium/drivers/cell/ppu/cell_state_shader.c | 4 +- src/gallium/drivers/cell/ppu/cell_surface.c | 11 ++-- src/gallium/drivers/cell/ppu/cell_texture.c | 69 ++++++++++++++++++++--- src/gallium/drivers/cell/ppu/cell_texture.h | 1 + src/gallium/drivers/cell/ppu/cell_vbuf.c | 1 + src/gallium/drivers/cell/ppu/cell_vertex_fetch.c | 4 ++ src/gallium/drivers/cell/ppu/cell_vertex_shader.c | 5 ++ src/gallium/drivers/cell/spu/spu_exec.h | 2 +- src/gallium/drivers/cell/spu/spu_render.c | 2 +- src/gallium/drivers/cell/spu/spu_tri.c | 1 + src/gallium/drivers/cell/spu/spu_util.c | 2 + src/gallium/drivers/cell/spu/spu_vertex_shader.c | 26 ++++++++- 18 files changed, 176 insertions(+), 39 deletions(-) (limited to 'src') diff --git a/src/gallium/drivers/cell/ppu/cell_clear.c b/src/gallium/drivers/cell/ppu/cell_clear.c index cee0917b63..a421c95c8e 100644 --- a/src/gallium/drivers/cell/ppu/cell_clear.c +++ b/src/gallium/drivers/cell/ppu/cell_clear.c @@ -48,6 +48,7 @@ void cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, unsigned clearValue) { + struct pipe_screen *screen = pipe->screen; struct cell_context *cell = cell_context(pipe); uint surfIndex; @@ -56,7 +57,8 @@ cell_clear_surface(struct pipe_context *pipe, struct pipe_surface *ps, if (!cell->cbuf_map[0]) - cell->cbuf_map[0] = pipe_surface_map(ps); + cell->cbuf_map[0] = screen->surface_map(screen, ps, + PIPE_BUFFER_USAGE_GPU_WRITE); if (ps == cell->framebuffer.zsbuf) { surfIndex = 1; diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c index 5af95a3c10..9ff4e86943 100644 --- a/src/gallium/drivers/cell/ppu/cell_context.c +++ b/src/gallium/drivers/cell/ppu/cell_context.c @@ -73,11 +73,13 @@ cell_draw_create(struct cell_context *cell) { struct draw_context *draw = draw_create(); +#if 0 /* broken */ if (getenv("GALLIUM_CELL_VS")) { /* plug in SPU-based vertex transformation code */ draw->shader_queue_flush = cell_vertex_shader_queue_flush; draw->driver_private = cell; } +#endif return draw; } @@ -108,6 +110,8 @@ cell_create_context(struct pipe_screen *screen, cell->pipe.draw_arrays = cell_draw_arrays; cell->pipe.draw_elements = cell_draw_elements; + cell->pipe.draw_range_elements = cell_draw_range_elements; + cell->pipe.set_edgeflags = cell_set_edgeflags; cell->pipe.clear = cell_clear_surface; cell->pipe.flush = cell_flush; diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c index 6e08cf6fe8..f02dffe124 100644 --- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c @@ -59,7 +59,8 @@ cell_map_constant_buffers(struct cell_context *sp) } draw_set_mapped_constant_buffer(sp->draw, - sp->mapped_constants[PIPE_SHADER_VERTEX]); + sp->mapped_constants[PIPE_SHADER_VERTEX], + sp->constants[PIPE_SHADER_VERTEX].size); } static void @@ -92,10 +93,12 @@ cell_draw_arrays(struct pipe_context *pipe, unsigned mode, * XXX should the element buffer be specified/bound with a separate function? */ boolean -cell_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count) +cell_draw_range_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned min_index, + unsigned max_index, + unsigned mode, unsigned start, unsigned count) { struct cell_context *sp = cell_context(pipe); struct draw_context *draw = sp->draw; @@ -152,3 +155,25 @@ cell_draw_elements(struct pipe_context *pipe, return TRUE; } + + +boolean +cell_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count) +{ + return cell_draw_range_elements( pipe, indexBuffer, + indexSize, + 0, 0xffffffff, + mode, start, count ); +} + + + +void +cell_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags) +{ + struct cell_context *cell = cell_context(pipe); + draw_set_edgeflags(cell->draw, edgeflags); +} diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.h b/src/gallium/drivers/cell/ppu/cell_draw_arrays.h index d5df4aa05f..cd35ec17b4 100644 --- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.h +++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.h @@ -29,14 +29,26 @@ #define CELL_DRAW_ARRAYS_H -boolean cell_draw_arrays(struct pipe_context *pipe, unsigned mode, - unsigned start, unsigned count); +extern boolean +cell_draw_arrays(struct pipe_context *pipe, unsigned mode, + unsigned start, unsigned count); -boolean cell_draw_elements(struct pipe_context *pipe, - struct pipe_buffer *indexBuffer, - unsigned indexSize, - unsigned mode, unsigned start, unsigned count); +extern boolean +cell_draw_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned mode, unsigned start, unsigned count); +extern boolean +cell_draw_range_elements(struct pipe_context *pipe, + struct pipe_buffer *indexBuffer, + unsigned indexSize, + unsigned min_index, + unsigned max_index, + unsigned mode, unsigned start, unsigned count); + +extern void +cell_set_edgeflags(struct pipe_context *pipe, const unsigned *edgeflags); #endif /* CELL_DRAW_ARRAYS_H */ diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c index 971d65d09e..fe5437023b 100644 --- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c +++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c @@ -294,6 +294,8 @@ cell_set_framebuffer_state(struct pipe_context *pipe, struct pipe_surface *csurf = fb->cbufs[0]; struct pipe_surface *zsurf = fb->zsbuf; uint i; + uint flags = (PIPE_BUFFER_USAGE_GPU_WRITE | + PIPE_BUFFER_USAGE_GPU_READ); /* unmap old surfaces */ for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) { @@ -313,10 +315,10 @@ cell_set_framebuffer_state(struct pipe_context *pipe, /* map new surfaces */ if (csurf) - cell->cbuf_map[0] = pipe_surface_map(csurf); + cell->cbuf_map[0] = pipe_surface_map(csurf, flags); if (zsurf) - cell->zsbuf_map = pipe_surface_map(zsurf); + cell->zsbuf_map = pipe_surface_map(zsurf, flags); cell->dirty |= CELL_NEW_FRAMEBUFFER; } diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c index 3646a0ee4f..9d88c1cf3d 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_emit.c +++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c @@ -162,13 +162,13 @@ cell_emit_state(struct cell_context *cell) const struct draw_context *const draw = cell->draw; struct cell_shader_info info; - info.num_outputs = draw->num_vs_outputs; - info.declarations = (uintptr_t) draw->machine.Declarations; - info.num_declarations = draw->machine.NumDeclarations; - info.instructions = (uintptr_t) draw->machine.Instructions; - info.num_instructions = draw->machine.NumInstructions; - info.immediates = (uintptr_t) draw->machine.Imms; - info.num_immediates = draw->machine.ImmLimit / 4; + info.num_outputs = draw_num_vs_outputs(draw); + info.declarations = (uintptr_t) draw->vs.machine.Declarations; + info.num_declarations = draw->vs.machine.NumDeclarations; + info.instructions = (uintptr_t) draw->vs.machine.Instructions; + info.num_instructions = draw->vs.machine.NumInstructions; + info.immediates = (uintptr_t) draw->vs.machine.Imms; + info.num_immediates = draw->vs.machine.ImmLimit / 4; emit_state_cmd(cell, CELL_CMD_STATE_BIND_VS, & info, sizeof(info)); diff --git a/src/gallium/drivers/cell/ppu/cell_state_shader.c b/src/gallium/drivers/cell/ppu/cell_state_shader.c index cd96b317fa..86bcad05e9 100644 --- a/src/gallium/drivers/cell/ppu/cell_state_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_state_shader.c @@ -41,7 +41,7 @@ static INLINE struct cell_fragment_shader_state * cell_fragment_shader_state(void *shader) { - return (struct pipe_shader_state *) shader; + return (struct cell_fragment_shader_state *) shader; } @@ -49,7 +49,7 @@ cell_fragment_shader_state(void *shader) static INLINE struct cell_vertex_shader_state * cell_vertex_shader_state(void *shader) { - return (struct pipe_shader_state *) shader; + return (struct cell_vertex_shader_state *) shader; } diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c index 2d31ad89a6..d9e3b510dc 100644 --- a/src/gallium/drivers/cell/ppu/cell_surface.c +++ b/src/gallium/drivers/cell/ppu/cell_surface.c @@ -26,11 +26,12 @@ **************************************************************************/ #include "pipe/p_defines.h" -#include "util/u_memory.h" #include "pipe/p_inlines.h" #include "pipe/p_winsys.h" -#include "util/p_tile.h" +#include "util/u_memory.h" #include "util/u_rect.h" +#include "util/u_tile.h" + #include "cell_context.h" #include "cell_surface.h" @@ -46,12 +47,12 @@ cell_surface_copy(struct pipe_context *pipe, { assert( dst->cpp == src->cpp ); - pipe_copy_rect(pipe_surface_map(dst), + pipe_copy_rect(pipe_surface_map(dst, PIPE_BUFFER_USAGE_CPU_WRITE), &dst->block, dst->stride, dstx, dsty, width, height, - pipe_surface_map(src), + pipe_surface_map(src, PIPE_BUFFER_USAGE_CPU_READ), do_flip ? -src->stride : src->stride, srcx, do_flip ? height - 1 - srcy : srcy); @@ -81,7 +82,7 @@ cell_surface_fill(struct pipe_context *pipe, unsigned width, unsigned height, unsigned value) { unsigned i, j; - void *dst_map = pipe_surface_map(dst); + void *dst_map = pipe_surface_map(dst, PIPE_BUFFER_USAGE_CPU_WRITE); assert(dst->stride > 0); diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index 1add81373d..5a0942bbd6 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -33,8 +33,9 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" -#include "util/u_memory.h" #include "pipe/p_winsys.h" +#include "util/u_math.h" +#include "util/u_memory.h" #include "cell_context.h" #include "cell_state.h" @@ -68,11 +69,13 @@ cell_texture_layout(struct cell_texture * spt) pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width); pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height); + spt->stride[level] = pt->nblocksx[level] * pt->block.size; + spt->level_offset[level] = spt->buffer_size; spt->buffer_size += (pt->nblocksy[level] * ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * - pt->nblocksx[level] * pt->block.size; + pt->nblocksx[level] * pt->block.size); width = minify(width); height = minify(height); @@ -147,7 +150,8 @@ cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture, static struct pipe_surface * cell_get_tex_surface_screen(struct pipe_screen *screen, struct pipe_texture *pt, - unsigned face, unsigned level, unsigned zslice) + unsigned face, unsigned level, unsigned zslice, + unsigned usage) { struct pipe_winsys *ws = screen->winsys; struct cell_texture *spt = cell_texture(pt); @@ -166,6 +170,10 @@ cell_get_tex_surface_screen(struct pipe_screen *screen, ps->nblocksy = pt->nblocksy[level]; ps->stride = spt->stride[level]; ps->offset = spt->level_offset[level]; + ps->usage = usage; + + /* XXX may need to override usage flags (see sp_texture.c) */ + if (pt->target == PIPE_TEXTURE_CUBE || pt->target == PIPE_TEXTURE_3D) { ps->offset += ((pt->target == PIPE_TEXTURE_CUBE) ? face : zslice) * @@ -228,10 +236,11 @@ cell_tile_texture(struct cell_context *cell, assert(w % TILE_SIZE == 0); assert(h % TILE_SIZE == 0); - surf = screen->get_tex_surface(screen, &texture->base, face, level, zslice); + surf = screen->get_tex_surface(screen, &texture->base, face, level, zslice, + PIPE_BUFFER_USAGE_CPU_WRITE); ASSERT(surf); - src = (const uint *) pipe_surface_map(surf); + src = (const uint *) pipe_surface_map(surf, PIPE_BUFFER_USAGE_CPU_WRITE); if (texture->tiled_data) { align_free(texture->tiled_data); @@ -246,11 +255,12 @@ cell_tile_texture(struct cell_context *cell, } - void cell_update_texture_mapping(struct cell_context *cell) { +#if 0 uint face = 0, level = 0, zslice = 0; +#endif uint i; for (i = 0; i < CELL_MAX_SAMPLERS; i++) { @@ -275,10 +285,52 @@ cell_update_texture_mapping(struct cell_context *cell) } +static void * +cell_surface_map( struct pipe_screen *screen, + struct pipe_surface *surface, + unsigned flags ) +{ + ubyte *map; + + if (flags & ~surface->usage) { + assert(0); + return NULL; + } + + map = screen->winsys->buffer_map( screen->winsys, surface->buffer, flags ); + if (map == NULL) + return NULL; + + /* May want to different things here depending on read/write nature + * of the map: + */ + if (surface->texture && + (flags & PIPE_BUFFER_USAGE_CPU_WRITE)) + { + /* Do something to notify sharing contexts of a texture change. + * In softpipe, that would mean flushing the texture cache. + */ +#if 0 + cell_screen(screen)->timestamp++; +#endif + } + + return map + surface->offset; +} + + +static void +cell_surface_unmap(struct pipe_screen *screen, + struct pipe_surface *surface) +{ + screen->winsys->buffer_unmap( screen->winsys, surface->buffer ); +} + + void cell_init_texture_functions(struct cell_context *cell) { - cell->pipe.texture_update = cell_texture_update; + /*cell->pipe.texture_update = cell_texture_update;*/ } void @@ -287,4 +339,7 @@ cell_init_screen_texture_funcs(struct pipe_screen *screen) screen->texture_create = cell_texture_create_screen; screen->texture_release = cell_texture_release_screen; screen->get_tex_surface = cell_get_tex_surface_screen; + + screen->surface_map = cell_surface_map; + screen->surface_unmap = cell_surface_unmap; } diff --git a/src/gallium/drivers/cell/ppu/cell_texture.h b/src/gallium/drivers/cell/ppu/cell_texture.h index fcee069d05..6d37e95ebc 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.h +++ b/src/gallium/drivers/cell/ppu/cell_texture.h @@ -41,6 +41,7 @@ struct cell_texture struct pipe_texture base; unsigned long level_offset[PIPE_MAX_TEXTURE_LEVELS]; + unsigned long stride[PIPE_MAX_TEXTURE_LEVELS]; /* The data is held here: */ diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.c b/src/gallium/drivers/cell/ppu/cell_vbuf.c index 3a181b585c..e4230c7a5f 100644 --- a/src/gallium/drivers/cell/ppu/cell_vbuf.c +++ b/src/gallium/drivers/cell/ppu/cell_vbuf.c @@ -37,6 +37,7 @@ #include "cell_spu.h" #include "cell_vbuf.h" #include "draw/draw_vbuf.h" +#include "util/u_memory.h" /** Allow vertex data to be inlined after RENDER command */ diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c index 49d5443cde..2ece0250f6 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c @@ -260,6 +260,7 @@ emit_fetch(struct spe_function *p, void cell_update_vertex_fetch(struct draw_context *draw) { +#if 0 struct cell_context *const cell = (struct cell_context *) draw->driver_private; struct spe_function *p = &cell->attrib_fetch; @@ -337,4 +338,7 @@ void cell_update_vertex_fetch(struct draw_context *draw) cell->attrib_fetch_offsets[function_index[i]]; } } +#else + assert(0); +#endif } diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c index f753960a0f..3658947715 100644 --- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c +++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c @@ -32,6 +32,7 @@ #include "pipe/p_defines.h" #include "pipe/p_context.h" #include "pipe/p_winsys.h" +#include "util/u_math.h" #include "cell_context.h" #include "cell_draw_arrays.h" @@ -50,6 +51,7 @@ void cell_vertex_shader_queue_flush(struct draw_context *draw) { +#if 0 struct cell_context *const cell = (struct cell_context *) draw->driver_private; struct cell_command_vs *const vs = &cell_global.command[0].vs; @@ -138,4 +140,7 @@ cell_vertex_shader_queue_flush(struct draw_context *draw) draw->vs.post_nr = draw->vs.queue_nr; draw->vs.queue_nr = 0; +#else + assert(0); +#endif } diff --git a/src/gallium/drivers/cell/spu/spu_exec.h b/src/gallium/drivers/cell/spu/spu_exec.h index c68f78f59b..8605679940 100644 --- a/src/gallium/drivers/cell/spu/spu_exec.h +++ b/src/gallium/drivers/cell/spu/spu_exec.h @@ -99,7 +99,7 @@ struct spu_exec_machine * 1 address */ struct spu_exec_vector Temps[TGSI_EXEC_NUM_TEMPS - + TGSI_EXEC_NUM_ADDRS + 1] + + TGSI_EXEC_NUM_TEMP_EXTRAS + 1] ALIGN16_ATTRIB; struct spu_exec_vector *Addrs; diff --git a/src/gallium/drivers/cell/spu/spu_render.c b/src/gallium/drivers/cell/spu/spu_render.c index 6df59abd36..305dc98881 100644 --- a/src/gallium/drivers/cell/spu/spu_render.c +++ b/src/gallium/drivers/cell/spu/spu_render.c @@ -35,7 +35,7 @@ #include "spu_tri.h" #include "spu_tile.h" #include "cell/common.h" - +#include "util/u_memory.h" /** diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index 8944ef171e..2a4e0b423c 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -32,6 +32,7 @@ #include #include "pipe/p_compiler.h" #include "pipe/p_format.h" +#include "util/u_math.h" #include "spu_colorpack.h" #include "spu_main.h" #include "spu_texture.h" diff --git a/src/gallium/drivers/cell/spu/spu_util.c b/src/gallium/drivers/cell/spu/spu_util.c index dbcf4b0eb9..b25ca4eafc 100644 --- a/src/gallium/drivers/cell/spu/spu_util.c +++ b/src/gallium/drivers/cell/spu/spu_util.c @@ -1,4 +1,6 @@ + #include "pipe/p_shader_tokens.h" +#include "pipe/p_debug.h" #include "tgsi/tgsi_parse.h" //#include "tgsi_build.h" #include "tgsi/tgsi_util.h" diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.c b/src/gallium/drivers/cell/spu/spu_vertex_shader.c index a1e81975e6..f81d19fea1 100644 --- a/src/gallium/drivers/cell/spu/spu_vertex_shader.c +++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.c @@ -36,13 +36,35 @@ #include "pipe/p_state.h" #include "pipe/p_shader_tokens.h" -#include "spu_vertex_shader.h" -#include "spu_exec.h" +#include "util/u_math.h" #include "draw/draw_private.h" #include "draw/draw_context.h" #include "cell/common.h" +#include "spu_vertex_shader.h" +#include "spu_exec.h" #include "spu_main.h" + +#define MAX_VERTEX_SIZE ((2 + PIPE_MAX_SHADER_OUTPUTS) * 4 * sizeof(float)) + + +#define CLIP_RIGHT_BIT 0x01 +#define CLIP_LEFT_BIT 0x02 +#define CLIP_TOP_BIT 0x04 +#define CLIP_BOTTOM_BIT 0x08 +#define CLIP_FAR_BIT 0x10 +#define CLIP_NEAR_BIT 0x20 + + +static INLINE float +dot4(const float *a, const float *b) +{ + return (a[0]*b[0] + + a[1]*b[1] + + a[2]*b[2] + + a[3]*b[3]); +} + static INLINE unsigned compute_clipmask(const float *clip, /*const*/ float plane[][4], unsigned nr) { -- cgit v1.2.3 From 026e7731e549e0777c010348460fd48b3d75a843 Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 25 Aug 2008 10:59:45 +0100 Subject: vbo: seed initial max_element value with a more likely candidate --- src/mesa/vbo/vbo_exec_array.c | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-) (limited to 'src') diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c index bf97956286..0f9d8da356 100644 --- a/src/mesa/vbo/vbo_exec_array.c +++ b/src/mesa/vbo/vbo_exec_array.c @@ -47,9 +47,9 @@ static void get_minmax_index( GLuint count, GLuint type, switch(type) { case GL_UNSIGNED_INT: { const GLuint *ui_indices = (const GLuint *)indices; - GLuint max_ui = ui_indices[0]; + GLuint max_ui = ui_indices[count-1]; GLuint min_ui = ui_indices[0]; - for (i = 1; i < count; i++) { + for (i = 0; i < count; i++) { if (ui_indices[i] > max_ui) max_ui = ui_indices[i]; if (ui_indices[i] < min_ui) min_ui = ui_indices[i]; } @@ -59,9 +59,9 @@ static void get_minmax_index( GLuint count, GLuint type, } case GL_UNSIGNED_SHORT: { const GLushort *us_indices = (const GLushort *)indices; - GLuint max_us = us_indices[0]; + GLuint max_us = us_indices[count-1]; GLuint min_us = us_indices[0]; - for (i = 1; i < count; i++) { + for (i = 0; i < count; i++) { if (us_indices[i] > max_us) max_us = us_indices[i]; if (us_indices[i] < min_us) min_us = us_indices[i]; } @@ -71,9 +71,9 @@ static void get_minmax_index( GLuint count, GLuint type, } case GL_UNSIGNED_BYTE: { const GLubyte *ub_indices = (const GLubyte *)indices; - GLuint max_ub = ub_indices[0]; + GLuint max_ub = ub_indices[count-1]; GLuint min_ub = ub_indices[0]; - for (i = 1; i < count; i++) { + for (i = 0; i < count; i++) { if (ub_indices[i] > max_ub) max_ub = ub_indices[i]; if (ub_indices[i] < min_ub) min_ub = ub_indices[i]; } -- cgit v1.2.3 From e6887a5752774c18cf527477fdd3e57e4893ff3b Mon Sep 17 00:00:00 2001 From: Keith Whitwell Date: Mon, 25 Aug 2008 11:19:24 +0100 Subject: draw: attempt atomic submit of large drawelements calls --- src/gallium/auxiliary/draw/draw_pt.h | 4 ++-- src/gallium/auxiliary/draw/draw_pt_fetch_emit.c | 9 ++++----- .../auxiliary/draw/draw_pt_fetch_shade_emit.c | 7 ++++--- .../auxiliary/draw/draw_pt_fetch_shade_pipeline.c | 11 ++++------- src/gallium/auxiliary/draw/draw_pt_vcache.c | 21 +++++++++++++-------- src/gallium/auxiliary/draw/draw_vbuf.h | 3 ++- 6 files changed, 29 insertions(+), 26 deletions(-) (limited to 'src') diff --git a/src/gallium/auxiliary/draw/draw_pt.h b/src/gallium/auxiliary/draw/draw_pt.h index 3d2a9c78b7..c02f229110 100644 --- a/src/gallium/auxiliary/draw/draw_pt.h +++ b/src/gallium/auxiliary/draw/draw_pt.h @@ -98,9 +98,9 @@ struct draw_pt_middle_end { unsigned count); /* Transform all vertices in a linear range and then draw them with - * the supplied element list. + * the supplied element list. May fail and return FALSE. */ - void (*run_linear_elts)( struct draw_pt_middle_end *, + boolean (*run_linear_elts)( struct draw_pt_middle_end *, unsigned fetch_start, unsigned fetch_count, const ushort *draw_elts, diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c index 0684c93d10..5a4db6cfe5 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_emit.c @@ -324,7 +324,7 @@ static void fetch_emit_run_linear( struct draw_pt_middle_end *middle, } -static void fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle, +static boolean fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle, unsigned start, unsigned count, const ushort *draw_elts, @@ -341,10 +341,8 @@ static void fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle, hw_verts = draw->render->allocate_vertices( draw->render, (ushort)feme->translate->key.output_stride, (ushort)count ); - if (!hw_verts) { - assert(0); - return; - } + if (!hw_verts) + return FALSE; /* Single routine to fetch vertices and emit HW verts. */ @@ -367,6 +365,7 @@ static void fetch_emit_run_linear_elts( struct draw_pt_middle_end *middle, feme->translate->key.output_stride, count ); + return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c index 87094f3092..73fc70c1bc 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c @@ -332,7 +332,7 @@ fse_run(struct draw_pt_middle_end *middle, -static void fse_run_linear_elts( struct draw_pt_middle_end *middle, +static boolean fse_run_linear_elts( struct draw_pt_middle_end *middle, unsigned start, unsigned count, const ushort *draw_elts, @@ -351,8 +351,7 @@ static void fse_run_linear_elts( struct draw_pt_middle_end *middle, (ushort)count ); if (!hw_verts) { - assert(0); - return; + return FALSE; } /* Single routine to fetch vertices, run shader and emit HW verts. @@ -374,6 +373,8 @@ static void fse_run_linear_elts( struct draw_pt_middle_end *middle, hw_verts, fse->key.output_stride, count ); + + return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c index f617aac9f7..ec3b41c320 100644 --- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c +++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c @@ -262,7 +262,7 @@ static void fetch_pipeline_linear_run( struct draw_pt_middle_end *middle, -static void fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle, +static boolean fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle, unsigned start, unsigned count, const ushort *draw_elts, @@ -277,12 +277,8 @@ static void fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle, struct vertex_header *pipeline_verts = (struct vertex_header *)MALLOC(fpme->vertex_size * alloc_count); - if (!pipeline_verts) { - /* Not much we can do here - just skip the rendering. - */ - assert(0); - return; - } + if (!pipeline_verts) + return FALSE; /* Fetch into our vertex buffer */ @@ -336,6 +332,7 @@ static void fetch_pipeline_linear_run_elts( struct draw_pt_middle_end *middle, } FREE(pipeline_verts); + return TRUE; } diff --git a/src/gallium/auxiliary/draw/draw_pt_vcache.c b/src/gallium/auxiliary/draw/draw_pt_vcache.c index b8b5de729d..e8467b2ae3 100644 --- a/src/gallium/auxiliary/draw/draw_pt_vcache.c +++ b/src/gallium/auxiliary/draw/draw_pt_vcache.c @@ -306,6 +306,7 @@ static void vcache_check_run( struct draw_pt_front_end *frontend, unsigned fetch_count = max_index + 1 - min_index; const ushort *transformed_elts; ushort *storage = NULL; + boolean ok; if (0) debug_printf("fetch_count %d fetch_max %d draw_count %d\n", fetch_count, @@ -313,7 +314,6 @@ static void vcache_check_run( struct draw_pt_front_end *frontend, draw_count); if (max_index == 0xffffffff || - fetch_count >= vcache->fetch_max || fetch_count > draw_count) { if (0) debug_printf("fail\n"); goto fail; @@ -395,14 +395,19 @@ static void vcache_check_run( struct draw_pt_front_end *frontend, transformed_elts = storage; } - vcache->middle->run_linear_elts( vcache->middle, - min_index, /* start */ - fetch_count, - transformed_elts, - draw_count ); - + ok = vcache->middle->run_linear_elts( vcache->middle, + min_index, /* start */ + fetch_count, + transformed_elts, + draw_count ); + FREE(storage); - return; + + if (ok) + return; + + debug_printf("failed to execute atomic draw elts for %d/%d, splitting up\n", + fetch_count, draw_count); fail: vcache_run( frontend, get_elt, elts, draw_count ); diff --git a/src/gallium/auxiliary/draw/draw_vbuf.h b/src/gallium/auxiliary/draw/draw_vbuf.h index 62247ccd9f..b0aa2df309 100644 --- a/src/gallium/auxiliary/draw/draw_vbuf.h +++ b/src/gallium/auxiliary/draw/draw_vbuf.h @@ -79,7 +79,8 @@ struct vbuf_render { boolean (*set_primitive)( struct vbuf_render *, unsigned prim ); /** - * DrawElements, note indices are ushort: + * DrawElements, note indices are ushort. The driver must complete + * this call, if necessary splitting the index list itself. */ void (*draw)( struct vbuf_render *, const ushort *indices, -- cgit v1.2.3